hpsa.c 250.9 KB
Newer Older
1 2
/*
 *    Disk Array driver for HP Smart Array SAS controllers
3
 *    Copyright 2000, 2014 Hewlett-Packard Development Company, L.P.
4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
 *
 *    This program is free software; you can redistribute it and/or modify
 *    it under the terms of the GNU General Public License as published by
 *    the Free Software Foundation; version 2 of the License.
 *
 *    This program is distributed in the hope that it will be useful,
 *    but WITHOUT ANY WARRANTY; without even the implied warranty of
 *    MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
 *    NON INFRINGEMENT.  See the GNU General Public License for more details.
 *
 *    You should have received a copy of the GNU General Public License
 *    along with this program; if not, write to the Free Software
 *    Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 *
 *    Questions/Comments/Bugfixes to iss_storagedev@hp.com
 *
 */

#include <linux/module.h>
#include <linux/interrupt.h>
#include <linux/types.h>
#include <linux/pci.h>
M
Matthew Garrett 已提交
26
#include <linux/pci-aspm.h>
27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44
#include <linux/kernel.h>
#include <linux/slab.h>
#include <linux/delay.h>
#include <linux/fs.h>
#include <linux/timer.h>
#include <linux/init.h>
#include <linux/spinlock.h>
#include <linux/compat.h>
#include <linux/blktrace_api.h>
#include <linux/uaccess.h>
#include <linux/io.h>
#include <linux/dma-mapping.h>
#include <linux/completion.h>
#include <linux/moduleparam.h>
#include <scsi/scsi.h>
#include <scsi/scsi_cmnd.h>
#include <scsi/scsi_device.h>
#include <scsi/scsi_host.h>
45
#include <scsi/scsi_tcq.h>
46
#include <scsi/scsi_eh.h>
47
#include <scsi/scsi_dbg.h>
48 49 50
#include <linux/cciss_ioctl.h>
#include <linux/string.h>
#include <linux/bitmap.h>
A
Arun Sharma 已提交
51
#include <linux/atomic.h>
52
#include <linux/jiffies.h>
D
Don Brace 已提交
53
#include <linux/percpu-defs.h>
54
#include <linux/percpu.h>
D
Don Brace 已提交
55
#include <asm/unaligned.h>
56
#include <asm/div64.h>
57 58 59 60
#include "hpsa_cmd.h"
#include "hpsa.h"

/* HPSA_DRIVER_VERSION must be 3 byte values (0-255) separated by '.' */
D
Don Brace 已提交
61
#define HPSA_DRIVER_VERSION "3.4.10-0"
62
#define DRIVER_NAME "HP HPSA Driver (v " HPSA_DRIVER_VERSION ")"
63
#define HPSA "hpsa"
64

65 66 67 68 69
/* How long to wait for CISS doorbell communication */
#define CLEAR_EVENT_WAIT_INTERVAL 20	/* ms for each msleep() call */
#define MODE_CHANGE_WAIT_INTERVAL 10	/* ms for each msleep() call */
#define MAX_CLEAR_EVENT_WAIT 30000	/* times 20 ms = 600 s */
#define MAX_MODE_CHANGE_WAIT 2000	/* times 10 ms = 20 s */
70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86
#define MAX_IOCTL_CONFIG_WAIT 1000

/*define how many times we will try a command because of bus resets */
#define MAX_CMD_RETRIES 3

/* Embedded module documentation macros - see modules.h */
MODULE_AUTHOR("Hewlett-Packard Company");
MODULE_DESCRIPTION("Driver for HP Smart Array Controller version " \
	HPSA_DRIVER_VERSION);
MODULE_SUPPORTED_DEVICE("HP Smart Array Controllers");
MODULE_VERSION(HPSA_DRIVER_VERSION);
MODULE_LICENSE("GPL");

static int hpsa_allow_any;
module_param(hpsa_allow_any, int, S_IRUGO|S_IWUSR);
MODULE_PARM_DESC(hpsa_allow_any,
		"Allow hpsa driver to access unknown HP Smart Array hardware");
87 88 89 90
static int hpsa_simple_mode;
module_param(hpsa_simple_mode, int, S_IRUGO|S_IWUSR);
MODULE_PARM_DESC(hpsa_simple_mode,
	"Use 'simple mode' rather than 'performant mode'");
91 92 93 94 95 96 97 98

/* define the PCI info for the cards we can control */
static const struct pci_device_id hpsa_pci_device_id[] = {
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSE,     0x103C, 0x3241},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSE,     0x103C, 0x3243},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSE,     0x103C, 0x3245},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSE,     0x103C, 0x3247},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSE,     0x103C, 0x3249},
99 100
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSE,     0x103C, 0x324A},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSE,     0x103C, 0x324B},
101
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSE,     0x103C, 0x3233},
102 103 104 105 106 107 108
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSF,     0x103C, 0x3350},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSF,     0x103C, 0x3351},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSF,     0x103C, 0x3352},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSF,     0x103C, 0x3353},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSF,     0x103C, 0x3354},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSF,     0x103C, 0x3355},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSF,     0x103C, 0x3356},
109 110 111 112 113 114
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSH,     0x103C, 0x1921},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSH,     0x103C, 0x1922},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSH,     0x103C, 0x1923},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSH,     0x103C, 0x1924},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSH,     0x103C, 0x1926},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSH,     0x103C, 0x1928},
115 116 117 118 119 120 121 122 123 124
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSH,     0x103C, 0x1929},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSI,     0x103C, 0x21BD},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSI,     0x103C, 0x21BE},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSI,     0x103C, 0x21BF},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSI,     0x103C, 0x21C0},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSI,     0x103C, 0x21C1},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSI,     0x103C, 0x21C2},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSI,     0x103C, 0x21C3},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSI,     0x103C, 0x21C4},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSI,     0x103C, 0x21C5},
125
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSI,     0x103C, 0x21C6},
126 127 128
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSI,     0x103C, 0x21C7},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSI,     0x103C, 0x21C8},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSI,     0x103C, 0x21C9},
129 130 131 132 133
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSI,     0x103C, 0x21CA},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSI,     0x103C, 0x21CB},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSI,     0x103C, 0x21CC},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSI,     0x103C, 0x21CD},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSI,     0x103C, 0x21CE},
D
Don Brace 已提交
134
	{PCI_VENDOR_ID_ADAPTEC2, 0x0290, 0x9005, 0x0580},
135 136 137 138 139
	{PCI_VENDOR_ID_HP_3PAR, 0x0075, 0x1590, 0x0076},
	{PCI_VENDOR_ID_HP_3PAR, 0x0075, 0x1590, 0x0087},
	{PCI_VENDOR_ID_HP_3PAR, 0x0075, 0x1590, 0x007D},
	{PCI_VENDOR_ID_HP_3PAR, 0x0075, 0x1590, 0x0088},
	{PCI_VENDOR_ID_HP, 0x333f, 0x103c, 0x333f},
140
	{PCI_VENDOR_ID_HP,     PCI_ANY_ID,	PCI_ANY_ID, PCI_ANY_ID,
141
		PCI_CLASS_STORAGE_RAID << 8, 0xffff << 8, 0},
142 143 144 145 146 147 148 149 150 151 152 153 154 155 156
	{0,}
};

MODULE_DEVICE_TABLE(pci, hpsa_pci_device_id);

/*  board_id = Subsystem Device ID & Vendor ID
 *  product = Marketing Name for the board
 *  access = Address of the struct of function pointers
 */
static struct board_type products[] = {
	{0x3241103C, "Smart Array P212", &SA5_access},
	{0x3243103C, "Smart Array P410", &SA5_access},
	{0x3245103C, "Smart Array P410i", &SA5_access},
	{0x3247103C, "Smart Array P411", &SA5_access},
	{0x3249103C, "Smart Array P812", &SA5_access},
157 158
	{0x324A103C, "Smart Array P712m", &SA5_access},
	{0x324B103C, "Smart Array P711m", &SA5_access},
159
	{0x3233103C, "HP StorageWorks 1210m", &SA5_access}, /* alias of 333f */
160 161 162 163 164 165 166
	{0x3350103C, "Smart Array P222", &SA5_access},
	{0x3351103C, "Smart Array P420", &SA5_access},
	{0x3352103C, "Smart Array P421", &SA5_access},
	{0x3353103C, "Smart Array P822", &SA5_access},
	{0x3354103C, "Smart Array P420i", &SA5_access},
	{0x3355103C, "Smart Array P220i", &SA5_access},
	{0x3356103C, "Smart Array P721m", &SA5_access},
167 168 169 170 171 172 173
	{0x1921103C, "Smart Array P830i", &SA5_access},
	{0x1922103C, "Smart Array P430", &SA5_access},
	{0x1923103C, "Smart Array P431", &SA5_access},
	{0x1924103C, "Smart Array P830", &SA5_access},
	{0x1926103C, "Smart Array P731m", &SA5_access},
	{0x1928103C, "Smart Array P230i", &SA5_access},
	{0x1929103C, "Smart Array P530", &SA5_access},
174 175 176 177
	{0x21BD103C, "Smart Array P244br", &SA5_access},
	{0x21BE103C, "Smart Array P741m", &SA5_access},
	{0x21BF103C, "Smart HBA H240ar", &SA5_access},
	{0x21C0103C, "Smart Array P440ar", &SA5_access},
178
	{0x21C1103C, "Smart Array P840ar", &SA5_access},
179 180
	{0x21C2103C, "Smart Array P440", &SA5_access},
	{0x21C3103C, "Smart Array P441", &SA5_access},
181
	{0x21C4103C, "Smart Array", &SA5_access},
182 183 184 185
	{0x21C5103C, "Smart Array P841", &SA5_access},
	{0x21C6103C, "Smart HBA H244br", &SA5_access},
	{0x21C7103C, "Smart HBA H240", &SA5_access},
	{0x21C8103C, "Smart HBA H241", &SA5_access},
186
	{0x21C9103C, "Smart Array", &SA5_access},
187 188
	{0x21CA103C, "Smart Array P246br", &SA5_access},
	{0x21CB103C, "Smart Array P840", &SA5_access},
189 190
	{0x21CC103C, "Smart Array", &SA5_access},
	{0x21CD103C, "Smart Array", &SA5_access},
191
	{0x21CE103C, "Smart HBA", &SA5_access},
D
Don Brace 已提交
192
	{0x05809005, "SmartHBA-SA", &SA5_access},
193 194 195 196 197
	{0x00761590, "HP Storage P1224 Array Controller", &SA5_access},
	{0x00871590, "HP Storage P1224e Array Controller", &SA5_access},
	{0x007D1590, "HP Storage P1228 Array Controller", &SA5_access},
	{0x00881590, "HP Storage P1228e Array Controller", &SA5_access},
	{0x333f103c, "HP StorageWorks 1210m Array Controller", &SA5_access},
198 199 200
	{0xFFFF103C, "Unknown Smart Array", &SA5_access},
};

201 202 203 204
#define SCSI_CMD_BUSY ((struct scsi_cmnd *)&hpsa_cmd_busy)
static const struct scsi_cmnd hpsa_cmd_busy;
#define SCSI_CMD_IDLE ((struct scsi_cmnd *)&hpsa_cmd_idle)
static const struct scsi_cmnd hpsa_cmd_idle;
205 206
static int number_of_controllers;

207 208
static irqreturn_t do_hpsa_intr_intx(int irq, void *dev_id);
static irqreturn_t do_hpsa_intr_msi(int irq, void *dev_id);
D
Don Brace 已提交
209
static int hpsa_ioctl(struct scsi_device *dev, int cmd, void __user *arg);
210 211

#ifdef CONFIG_COMPAT
D
Don Brace 已提交
212 213
static int hpsa_compat_ioctl(struct scsi_device *dev, int cmd,
	void __user *arg);
214 215 216 217
#endif

static void cmd_free(struct ctlr_info *h, struct CommandList *c);
static struct CommandList *cmd_alloc(struct ctlr_info *h);
218 219 220
static void cmd_tagged_free(struct ctlr_info *h, struct CommandList *c);
static struct CommandList *cmd_tagged_alloc(struct ctlr_info *h,
					    struct scsi_cmnd *scmd);
221
static int fill_cmd(struct CommandList *c, u8 cmd, struct ctlr_info *h,
222
	void *buff, size_t size, u16 page_code, unsigned char *scsi3addr,
223
	int cmd_type);
224
static void hpsa_free_cmd_pool(struct ctlr_info *h);
225
#define VPD_PAGE (1 << 8)
226

J
Jeff Garzik 已提交
227
static int hpsa_scsi_queue_command(struct Scsi_Host *h, struct scsi_cmnd *cmd);
228 229 230
static void hpsa_scan_start(struct Scsi_Host *);
static int hpsa_scan_finished(struct Scsi_Host *sh,
	unsigned long elapsed_time);
D
Don Brace 已提交
231
static int hpsa_change_queue_depth(struct scsi_device *sdev, int qdepth);
232 233

static int hpsa_eh_device_reset_handler(struct scsi_cmnd *scsicmd);
234
static int hpsa_eh_abort_handler(struct scsi_cmnd *scsicmd);
235
static int hpsa_slave_alloc(struct scsi_device *sdev);
236
static int hpsa_slave_configure(struct scsi_device *sdev);
237 238 239 240 241 242 243
static void hpsa_slave_destroy(struct scsi_device *sdev);

static void hpsa_update_scsi_devices(struct ctlr_info *h, int hostno);
static int check_for_unit_attention(struct ctlr_info *h,
	struct CommandList *c);
static void check_ioctl_unit_attention(struct ctlr_info *h,
	struct CommandList *c);
244 245
/* performant mode helper functions */
static void calc_bucket_map(int *bucket, int num_buckets,
D
Don Brace 已提交
246
	int nsgs, int min_blocks, u32 *bucket_map);
R
Robert Elliott 已提交
247 248
static void hpsa_free_performant_mode(struct ctlr_info *h);
static int hpsa_put_ctlr_into_performant_mode(struct ctlr_info *h);
249
static inline u32 next_command(struct ctlr_info *h, u8 q);
250 251 252 253 254 255 256 257
static int hpsa_find_cfg_addrs(struct pci_dev *pdev, void __iomem *vaddr,
			       u32 *cfg_base_addr, u64 *cfg_base_addr_index,
			       u64 *cfg_offset);
static int hpsa_pci_find_memory_BAR(struct pci_dev *pdev,
				    unsigned long *memory_bar);
static int hpsa_lookup_board_id(struct pci_dev *pdev, u32 *board_id);
static int hpsa_wait_for_board_state(struct pci_dev *pdev, void __iomem *vaddr,
				     int wait_for_ready);
258
static inline void finish_cmd(struct CommandList *c);
259
static int hpsa_wait_for_mode_change_ack(struct ctlr_info *h);
260 261
#define BOARD_NOT_READY 0
#define BOARD_READY 1
262
static void hpsa_drain_accel_commands(struct ctlr_info *h);
263
static void hpsa_flush_cache(struct ctlr_info *h);
264 265
static int hpsa_scsi_ioaccel_queue_command(struct ctlr_info *h,
	struct CommandList *c, u32 ioaccel_handle, u8 *cdb, int cdb_len,
266
	u8 *scsi3addr, struct hpsa_scsi_dev_t *phys_disk);
267
static void hpsa_command_resubmit_worker(struct work_struct *work);
268 269
static u32 lockup_detected(struct ctlr_info *h);
static int detect_controller_lockup(struct ctlr_info *h);
270 271 272 273 274 275 276

static inline struct ctlr_info *sdev_to_hba(struct scsi_device *sdev)
{
	unsigned long *priv = shost_priv(sdev->host);
	return (struct ctlr_info *) *priv;
}

277 278 279 280 281 282
static inline struct ctlr_info *shost_to_hba(struct Scsi_Host *sh)
{
	unsigned long *priv = shost_priv(sh);
	return (struct ctlr_info *) *priv;
}

283 284 285 286 287
static inline bool hpsa_is_cmd_idle(struct CommandList *c)
{
	return c->scsi_cmd == SCSI_CMD_IDLE;
}

W
Webb Scales 已提交
288 289 290 291 292
static inline bool hpsa_is_pending_event(struct CommandList *c)
{
	return c->abort_pending || c->reset_pending;
}

293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314
/* extract sense key, asc, and ascq from sense data.  -1 means invalid. */
static void decode_sense_data(const u8 *sense_data, int sense_data_len,
			u8 *sense_key, u8 *asc, u8 *ascq)
{
	struct scsi_sense_hdr sshdr;
	bool rc;

	*sense_key = -1;
	*asc = -1;
	*ascq = -1;

	if (sense_data_len < 1)
		return;

	rc = scsi_normalize_sense(sense_data, sense_data_len, &sshdr);
	if (rc) {
		*sense_key = sshdr.sense_key;
		*asc = sshdr.asc;
		*ascq = sshdr.ascq;
	}
}

315 316 317
static int check_for_unit_attention(struct ctlr_info *h,
	struct CommandList *c)
{
318 319 320 321 322 323 324 325 326 327 328
	u8 sense_key, asc, ascq;
	int sense_len;

	if (c->err_info->SenseLen > sizeof(c->err_info->SenseInfo))
		sense_len = sizeof(c->err_info->SenseInfo);
	else
		sense_len = c->err_info->SenseLen;

	decode_sense_data(c->err_info->SenseInfo, sense_len,
				&sense_key, &asc, &ascq);
	if (sense_key != UNIT_ATTENTION || asc == -1)
329 330
		return 0;

331
	switch (asc) {
332
	case STATE_CHANGED:
333
		dev_warn(&h->pdev->dev,
334 335
			"%s: a state change detected, command retried\n",
			h->devname);
336 337
		break;
	case LUN_FAILED:
338
		dev_warn(&h->pdev->dev,
339
			"%s: LUN failure detected\n", h->devname);
340 341
		break;
	case REPORT_LUNS_CHANGED:
342
		dev_warn(&h->pdev->dev,
343
			"%s: report LUN data changed\n", h->devname);
344
	/*
345 346
	 * Note: this REPORT_LUNS_CHANGED condition only occurs on the external
	 * target (array) devices.
347 348 349
	 */
		break;
	case POWER_OR_RESET:
350 351 352
		dev_warn(&h->pdev->dev,
			"%s: a power on or device reset detected\n",
			h->devname);
353 354
		break;
	case UNIT_ATTENTION_CLEARED:
355 356 357
		dev_warn(&h->pdev->dev,
			"%s: unit attention cleared by another initiator\n",
			h->devname);
358 359
		break;
	default:
360 361 362
		dev_warn(&h->pdev->dev,
			"%s: unknown unit attention detected\n",
			h->devname);
363 364 365 366 367
		break;
	}
	return 1;
}

368 369 370 371 372 373 374 375 376 377
static int check_for_busy(struct ctlr_info *h, struct CommandList *c)
{
	if (c->err_info->CommandStatus != CMD_TARGET_STATUS ||
		(c->err_info->ScsiStatus != SAM_STAT_BUSY &&
		 c->err_info->ScsiStatus != SAM_STAT_TASK_SET_FULL))
		return 0;
	dev_warn(&h->pdev->dev, HPSA "device busy");
	return 1;
}

378 379 380 381 382 383 384 385 386 387 388 389 390 391
static u32 lockup_detected(struct ctlr_info *h);
static ssize_t host_show_lockup_detected(struct device *dev,
		struct device_attribute *attr, char *buf)
{
	int ld;
	struct ctlr_info *h;
	struct Scsi_Host *shost = class_to_shost(dev);

	h = shost_to_hba(shost);
	ld = lockup_detected(h);

	return sprintf(buf, "ld=%d\n", ld);
}

392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415
static ssize_t host_store_hp_ssd_smart_path_status(struct device *dev,
					 struct device_attribute *attr,
					 const char *buf, size_t count)
{
	int status, len;
	struct ctlr_info *h;
	struct Scsi_Host *shost = class_to_shost(dev);
	char tmpbuf[10];

	if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SYS_RAWIO))
		return -EACCES;
	len = count > sizeof(tmpbuf) - 1 ? sizeof(tmpbuf) - 1 : count;
	strncpy(tmpbuf, buf, len);
	tmpbuf[len] = '\0';
	if (sscanf(tmpbuf, "%d", &status) != 1)
		return -EINVAL;
	h = shost_to_hba(shost);
	h->acciopath_status = !!status;
	dev_warn(&h->pdev->dev,
		"hpsa: HP SSD Smart Path %s via sysfs update.\n",
		h->acciopath_status ? "enabled" : "disabled");
	return count;
}

416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440
static ssize_t host_store_raid_offload_debug(struct device *dev,
					 struct device_attribute *attr,
					 const char *buf, size_t count)
{
	int debug_level, len;
	struct ctlr_info *h;
	struct Scsi_Host *shost = class_to_shost(dev);
	char tmpbuf[10];

	if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SYS_RAWIO))
		return -EACCES;
	len = count > sizeof(tmpbuf) - 1 ? sizeof(tmpbuf) - 1 : count;
	strncpy(tmpbuf, buf, len);
	tmpbuf[len] = '\0';
	if (sscanf(tmpbuf, "%d", &debug_level) != 1)
		return -EINVAL;
	if (debug_level < 0)
		debug_level = 0;
	h = shost_to_hba(shost);
	h->raid_offload_debug = debug_level;
	dev_warn(&h->pdev->dev, "hpsa: Set raid_offload_debug level = %d\n",
		h->raid_offload_debug);
	return count;
}

441 442 443 444 445 446
static ssize_t host_store_rescan(struct device *dev,
				 struct device_attribute *attr,
				 const char *buf, size_t count)
{
	struct ctlr_info *h;
	struct Scsi_Host *shost = class_to_shost(dev);
447
	h = shost_to_hba(shost);
M
Mike Miller 已提交
448
	hpsa_scan_start(h->scsi_host);
449 450 451
	return count;
}

452 453 454 455 456 457 458 459 460 461 462 463 464 465 466
static ssize_t host_show_firmware_revision(struct device *dev,
	     struct device_attribute *attr, char *buf)
{
	struct ctlr_info *h;
	struct Scsi_Host *shost = class_to_shost(dev);
	unsigned char *fwrev;

	h = shost_to_hba(shost);
	if (!h->hba_inquiry_data)
		return 0;
	fwrev = &h->hba_inquiry_data[32];
	return snprintf(buf, 20, "%c%c%c%c\n",
		fwrev[0], fwrev[1], fwrev[2], fwrev[3]);
}

467 468 469 470 471 472
static ssize_t host_show_commands_outstanding(struct device *dev,
	     struct device_attribute *attr, char *buf)
{
	struct Scsi_Host *shost = class_to_shost(dev);
	struct ctlr_info *h = shost_to_hba(shost);

473 474
	return snprintf(buf, 20, "%d\n",
			atomic_read(&h->commands_outstanding));
475 476
}

477 478 479 480 481 482 483 484
static ssize_t host_show_transport_mode(struct device *dev,
	struct device_attribute *attr, char *buf)
{
	struct ctlr_info *h;
	struct Scsi_Host *shost = class_to_shost(dev);

	h = shost_to_hba(shost);
	return snprintf(buf, 20, "%s\n",
485
		h->transMethod & CFGTBL_Trans_Performant ?
486 487 488
			"performant" : "simple");
}

489 490 491 492 493 494 495 496 497 498 499
static ssize_t host_show_hp_ssd_smart_path_status(struct device *dev,
	struct device_attribute *attr, char *buf)
{
	struct ctlr_info *h;
	struct Scsi_Host *shost = class_to_shost(dev);

	h = shost_to_hba(shost);
	return snprintf(buf, 30, "HP SSD Smart Path %s\n",
		(h->acciopath_status == 1) ?  "enabled" : "disabled");
}

500
/* List of controllers which cannot be hard reset on kexec with reset_devices */
501 502
static u32 unresettable_controller[] = {
	0x324a103C, /* Smart Array P712m */
S
Stephen Cameron 已提交
503
	0x324b103C, /* Smart Array P711m */
504 505 506 507 508 509 510 511 512 513
	0x3223103C, /* Smart Array P800 */
	0x3234103C, /* Smart Array P400 */
	0x3235103C, /* Smart Array P400i */
	0x3211103C, /* Smart Array E200i */
	0x3212103C, /* Smart Array E200 */
	0x3213103C, /* Smart Array E200i */
	0x3214103C, /* Smart Array E200i */
	0x3215103C, /* Smart Array E200i */
	0x3237103C, /* Smart Array E500 */
	0x323D103C, /* Smart Array P700m */
514
	0x40800E11, /* Smart Array 5i */
515 516
	0x409C0E11, /* Smart Array 6400 */
	0x409D0E11, /* Smart Array 6400 EM */
517 518 519 520 521 522
	0x40700E11, /* Smart Array 5300 */
	0x40820E11, /* Smart Array 532 */
	0x40830E11, /* Smart Array 5312 */
	0x409A0E11, /* Smart Array 641 */
	0x409B0E11, /* Smart Array 642 */
	0x40910E11, /* Smart Array 6i */
523 524
};

525 526
/* List of controllers which cannot even be soft reset */
static u32 soft_unresettable_controller[] = {
527
	0x40800E11, /* Smart Array 5i */
528 529 530 531 532 533
	0x40700E11, /* Smart Array 5300 */
	0x40820E11, /* Smart Array 532 */
	0x40830E11, /* Smart Array 5312 */
	0x409A0E11, /* Smart Array 641 */
	0x409B0E11, /* Smart Array 642 */
	0x40910E11, /* Smart Array 6i */
534 535 536 537 538 539 540 541 542 543 544
	/* Exclude 640x boards.  These are two pci devices in one slot
	 * which share a battery backed cache module.  One controls the
	 * cache, the other accesses the cache through the one that controls
	 * it.  If we reset the one controlling the cache, the other will
	 * likely not be happy.  Just forbid resetting this conjoined mess.
	 * The 640x isn't really supported by hpsa anyway.
	 */
	0x409C0E11, /* Smart Array 6400 */
	0x409D0E11, /* Smart Array 6400 EM */
};

S
Stephen Cameron 已提交
545 546 547 548 549 550 551
static u32 needs_abort_tags_swizzled[] = {
	0x323D103C, /* Smart Array P700m */
	0x324a103C, /* Smart Array P712m */
	0x324b103C, /* SmartArray P711m */
};

static int board_id_in_array(u32 a[], int nelems, u32 board_id)
552 553 554
{
	int i;

S
Stephen Cameron 已提交
555 556 557 558
	for (i = 0; i < nelems; i++)
		if (a[i] == board_id)
			return 1;
	return 0;
559 560
}

S
Stephen Cameron 已提交
561
static int ctlr_is_hard_resettable(u32 board_id)
562
{
S
Stephen Cameron 已提交
563 564 565
	return !board_id_in_array(unresettable_controller,
			ARRAY_SIZE(unresettable_controller), board_id);
}
566

S
Stephen Cameron 已提交
567 568 569 570
static int ctlr_is_soft_resettable(u32 board_id)
{
	return !board_id_in_array(soft_unresettable_controller,
			ARRAY_SIZE(soft_unresettable_controller), board_id);
571 572
}

573 574 575 576 577 578
static int ctlr_is_resettable(u32 board_id)
{
	return ctlr_is_hard_resettable(board_id) ||
		ctlr_is_soft_resettable(board_id);
}

S
Stephen Cameron 已提交
579 580 581 582 583 584
static int ctlr_needs_abort_tags_swizzled(u32 board_id)
{
	return board_id_in_array(needs_abort_tags_swizzled,
			ARRAY_SIZE(needs_abort_tags_swizzled), board_id);
}

585 586 587 588 589 590 591
static ssize_t host_show_resettable(struct device *dev,
	struct device_attribute *attr, char *buf)
{
	struct ctlr_info *h;
	struct Scsi_Host *shost = class_to_shost(dev);

	h = shost_to_hba(shost);
592
	return snprintf(buf, 20, "%d\n", ctlr_is_resettable(h->board_id));
593 594
}

595 596 597 598 599
static inline int is_logical_dev_addr_mode(unsigned char scsi3addr[])
{
	return (scsi3addr[3] & 0xC0) == 0x40;
}

600 601
static const char * const raid_label[] = { "0", "4", "1(+0)", "5", "5+1", "6",
	"1(+0)ADM", "UNKNOWN"
602
};
603 604 605 606 607 608 609
#define HPSA_RAID_0	0
#define HPSA_RAID_4	1
#define HPSA_RAID_1	2	/* also used for RAID 10 */
#define HPSA_RAID_5	3	/* also used for RAID 50 */
#define HPSA_RAID_51	4
#define HPSA_RAID_6	5	/* also used for RAID 60 */
#define HPSA_RAID_ADM	6	/* also used for RAID 1+0 ADM */
610 611 612 613 614 615
#define RAID_UNKNOWN (ARRAY_SIZE(raid_label) - 1)

static ssize_t raid_level_show(struct device *dev,
	     struct device_attribute *attr, char *buf)
{
	ssize_t l = 0;
616
	unsigned char rlevel;
617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639
	struct ctlr_info *h;
	struct scsi_device *sdev;
	struct hpsa_scsi_dev_t *hdev;
	unsigned long flags;

	sdev = to_scsi_device(dev);
	h = sdev_to_hba(sdev);
	spin_lock_irqsave(&h->lock, flags);
	hdev = sdev->hostdata;
	if (!hdev) {
		spin_unlock_irqrestore(&h->lock, flags);
		return -ENODEV;
	}

	/* Is this even a logical drive? */
	if (!is_logical_dev_addr_mode(hdev->scsi3addr)) {
		spin_unlock_irqrestore(&h->lock, flags);
		l = snprintf(buf, PAGE_SIZE, "N/A\n");
		return l;
	}

	rlevel = hdev->raid_level;
	spin_unlock_irqrestore(&h->lock, flags);
640
	if (rlevel > RAID_UNKNOWN)
641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697
		rlevel = RAID_UNKNOWN;
	l = snprintf(buf, PAGE_SIZE, "RAID %s\n", raid_label[rlevel]);
	return l;
}

static ssize_t lunid_show(struct device *dev,
	     struct device_attribute *attr, char *buf)
{
	struct ctlr_info *h;
	struct scsi_device *sdev;
	struct hpsa_scsi_dev_t *hdev;
	unsigned long flags;
	unsigned char lunid[8];

	sdev = to_scsi_device(dev);
	h = sdev_to_hba(sdev);
	spin_lock_irqsave(&h->lock, flags);
	hdev = sdev->hostdata;
	if (!hdev) {
		spin_unlock_irqrestore(&h->lock, flags);
		return -ENODEV;
	}
	memcpy(lunid, hdev->scsi3addr, sizeof(lunid));
	spin_unlock_irqrestore(&h->lock, flags);
	return snprintf(buf, 20, "0x%02x%02x%02x%02x%02x%02x%02x%02x\n",
		lunid[0], lunid[1], lunid[2], lunid[3],
		lunid[4], lunid[5], lunid[6], lunid[7]);
}

static ssize_t unique_id_show(struct device *dev,
	     struct device_attribute *attr, char *buf)
{
	struct ctlr_info *h;
	struct scsi_device *sdev;
	struct hpsa_scsi_dev_t *hdev;
	unsigned long flags;
	unsigned char sn[16];

	sdev = to_scsi_device(dev);
	h = sdev_to_hba(sdev);
	spin_lock_irqsave(&h->lock, flags);
	hdev = sdev->hostdata;
	if (!hdev) {
		spin_unlock_irqrestore(&h->lock, flags);
		return -ENODEV;
	}
	memcpy(sn, hdev->device_id, sizeof(sn));
	spin_unlock_irqrestore(&h->lock, flags);
	return snprintf(buf, 16 * 2 + 2,
			"%02X%02X%02X%02X%02X%02X%02X%02X"
			"%02X%02X%02X%02X%02X%02X%02X%02X\n",
			sn[0], sn[1], sn[2], sn[3],
			sn[4], sn[5], sn[6], sn[7],
			sn[8], sn[9], sn[10], sn[11],
			sn[12], sn[13], sn[14], sn[15]);
}

698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719
static ssize_t host_show_hp_ssd_smart_path_enabled(struct device *dev,
	     struct device_attribute *attr, char *buf)
{
	struct ctlr_info *h;
	struct scsi_device *sdev;
	struct hpsa_scsi_dev_t *hdev;
	unsigned long flags;
	int offload_enabled;

	sdev = to_scsi_device(dev);
	h = sdev_to_hba(sdev);
	spin_lock_irqsave(&h->lock, flags);
	hdev = sdev->hostdata;
	if (!hdev) {
		spin_unlock_irqrestore(&h->lock, flags);
		return -ENODEV;
	}
	offload_enabled = hdev->offload_enabled;
	spin_unlock_irqrestore(&h->lock, flags);
	return snprintf(buf, 20, "%d\n", offload_enabled);
}

720 721 722 723
static DEVICE_ATTR(raid_level, S_IRUGO, raid_level_show, NULL);
static DEVICE_ATTR(lunid, S_IRUGO, lunid_show, NULL);
static DEVICE_ATTR(unique_id, S_IRUGO, unique_id_show, NULL);
static DEVICE_ATTR(rescan, S_IWUSR, NULL, host_store_rescan);
724 725
static DEVICE_ATTR(hp_ssd_smart_path_enabled, S_IRUGO,
			host_show_hp_ssd_smart_path_enabled, NULL);
726 727 728
static DEVICE_ATTR(hp_ssd_smart_path_status, S_IWUSR|S_IRUGO|S_IROTH,
		host_show_hp_ssd_smart_path_status,
		host_store_hp_ssd_smart_path_status);
729 730
static DEVICE_ATTR(raid_offload_debug, S_IWUSR, NULL,
			host_store_raid_offload_debug);
731 732 733 734 735 736
static DEVICE_ATTR(firmware_revision, S_IRUGO,
	host_show_firmware_revision, NULL);
static DEVICE_ATTR(commands_outstanding, S_IRUGO,
	host_show_commands_outstanding, NULL);
static DEVICE_ATTR(transport_mode, S_IRUGO,
	host_show_transport_mode, NULL);
737 738
static DEVICE_ATTR(resettable, S_IRUGO,
	host_show_resettable, NULL);
739 740
static DEVICE_ATTR(lockup_detected, S_IRUGO,
	host_show_lockup_detected, NULL);
741 742 743 744 745

static struct device_attribute *hpsa_sdev_attrs[] = {
	&dev_attr_raid_level,
	&dev_attr_lunid,
	&dev_attr_unique_id,
746
	&dev_attr_hp_ssd_smart_path_enabled,
747
	&dev_attr_lockup_detected,
748 749 750 751 752 753 754 755
	NULL,
};

static struct device_attribute *hpsa_shost_attrs[] = {
	&dev_attr_rescan,
	&dev_attr_firmware_revision,
	&dev_attr_commands_outstanding,
	&dev_attr_transport_mode,
756
	&dev_attr_resettable,
757
	&dev_attr_hp_ssd_smart_path_status,
758
	&dev_attr_raid_offload_debug,
759 760 761
	NULL,
};

762 763 764
#define HPSA_NRESERVED_CMDS	(HPSA_CMDS_RESERVED_FOR_ABORTS + \
		HPSA_CMDS_RESERVED_FOR_DRIVER + HPSA_MAX_CONCURRENT_PASSTHRUS)

765 766
static struct scsi_host_template hpsa_driver_template = {
	.module			= THIS_MODULE,
767 768
	.name			= HPSA,
	.proc_name		= HPSA,
769 770 771
	.queuecommand		= hpsa_scsi_queue_command,
	.scan_start		= hpsa_scan_start,
	.scan_finished		= hpsa_scan_finished,
D
Don Brace 已提交
772
	.change_queue_depth	= hpsa_change_queue_depth,
773 774
	.this_id		= -1,
	.use_clustering		= ENABLE_CLUSTERING,
775
	.eh_abort_handler	= hpsa_eh_abort_handler,
776 777 778
	.eh_device_reset_handler = hpsa_eh_device_reset_handler,
	.ioctl			= hpsa_ioctl,
	.slave_alloc		= hpsa_slave_alloc,
779
	.slave_configure	= hpsa_slave_configure,
780 781 782 783 784 785
	.slave_destroy		= hpsa_slave_destroy,
#ifdef CONFIG_COMPAT
	.compat_ioctl		= hpsa_compat_ioctl,
#endif
	.sdev_attrs = hpsa_sdev_attrs,
	.shost_attrs = hpsa_shost_attrs,
786
	.max_sectors = 8192,
787
	.no_write_same = 1,
788 789
};

790
static inline u32 next_command(struct ctlr_info *h, u8 q)
791 792
{
	u32 a;
793
	struct reply_queue_buffer *rq = &h->reply_queue[q];
794

795 796 797
	if (h->transMethod & CFGTBL_Trans_io_accel1)
		return h->access.command_completed(h, q);

798
	if (unlikely(!(h->transMethod & CFGTBL_Trans_Performant)))
799
		return h->access.command_completed(h, q);
800

801 802 803
	if ((rq->head[rq->current_entry] & 1) == rq->wraparound) {
		a = rq->head[rq->current_entry];
		rq->current_entry++;
804
		atomic_dec(&h->commands_outstanding);
805 806 807 808
	} else {
		a = FIFO_EMPTY;
	}
	/* Check for wraparound */
809 810 811
	if (rq->current_entry == h->max_commands) {
		rq->current_entry = 0;
		rq->wraparound ^= 1;
812 813 814 815
	}
	return a;
}

816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841
/*
 * There are some special bits in the bus address of the
 * command that we have to set for the controller to know
 * how to process the command:
 *
 * Normal performant mode:
 * bit 0: 1 means performant mode, 0 means simple mode.
 * bits 1-3 = block fetch table entry
 * bits 4-6 = command type (== 0)
 *
 * ioaccel1 mode:
 * bit 0 = "performant mode" bit.
 * bits 1-3 = block fetch table entry
 * bits 4-6 = command type (== 110)
 * (command type is needed because ioaccel1 mode
 * commands are submitted through the same register as normal
 * mode commands, so this is how the controller knows whether
 * the command is normal mode or ioaccel1 mode.)
 *
 * ioaccel2 mode:
 * bit 0 = "performant mode" bit.
 * bits 1-4 = block fetch table entry (note extra bit)
 * bits 4-6 = not needed, because ioaccel2 mode has
 * a separate special register for submitting commands.
 */

842 843
/*
 * set_performant_mode: Modify the tag for cciss performant
844 845 846
 * set bit 0 for pull model, bits 3-1 for block fetch
 * register number
 */
847 848 849
#define DEFAULT_REPLY_QUEUE (-1)
static void set_performant_mode(struct ctlr_info *h, struct CommandList *c,
					int reply_queue)
850
{
851
	if (likely(h->transMethod & CFGTBL_Trans_Performant)) {
852
		c->busaddr |= 1 | (h->blockFetchTable[c->Header.SGList] << 1);
853 854 855
		if (unlikely(!h->msix_vector))
			return;
		if (likely(reply_queue == DEFAULT_REPLY_QUEUE))
856
			c->Header.ReplyQueue =
857
				raw_smp_processor_id() % h->nreply_queues;
858 859
		else
			c->Header.ReplyQueue = reply_queue % h->nreply_queues;
860
	}
861 862
}

863
static void set_ioaccel1_performant_mode(struct ctlr_info *h,
864 865
						struct CommandList *c,
						int reply_queue)
866 867 868
{
	struct io_accel1_cmd *cp = &h->ioaccel_cmd_pool[c->cmdindex];

869 870
	/*
	 * Tell the controller to post the reply to the queue for this
871 872
	 * processor.  This seems to give the best I/O throughput.
	 */
873 874 875 876 877 878
	if (likely(reply_queue == DEFAULT_REPLY_QUEUE))
		cp->ReplyQueue = smp_processor_id() % h->nreply_queues;
	else
		cp->ReplyQueue = reply_queue % h->nreply_queues;
	/*
	 * Set the bits in the address sent down to include:
879 880 881 882 883 884 885 886
	 *  - performant mode bit (bit 0)
	 *  - pull count (bits 1-3)
	 *  - command type (bits 4-6)
	 */
	c->busaddr |= 1 | (h->ioaccel1_blockFetchTable[c->Header.SGList] << 1) |
					IOACCEL1_BUSADDR_CMDTYPE;
}

887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908
static void set_ioaccel2_tmf_performant_mode(struct ctlr_info *h,
						struct CommandList *c,
						int reply_queue)
{
	struct hpsa_tmf_struct *cp = (struct hpsa_tmf_struct *)
		&h->ioaccel2_cmd_pool[c->cmdindex];

	/* Tell the controller to post the reply to the queue for this
	 * processor.  This seems to give the best I/O throughput.
	 */
	if (likely(reply_queue == DEFAULT_REPLY_QUEUE))
		cp->reply_queue = smp_processor_id() % h->nreply_queues;
	else
		cp->reply_queue = reply_queue % h->nreply_queues;
	/* Set the bits in the address sent down to include:
	 *  - performant mode bit not used in ioaccel mode 2
	 *  - pull count (bits 0-3)
	 *  - command type isn't needed for ioaccel2
	 */
	c->busaddr |= h->ioaccel2_blockFetchTable[0];
}

909
static void set_ioaccel2_performant_mode(struct ctlr_info *h,
910 911
						struct CommandList *c,
						int reply_queue)
912 913 914
{
	struct io_accel2_cmd *cp = &h->ioaccel2_cmd_pool[c->cmdindex];

915 916
	/*
	 * Tell the controller to post the reply to the queue for this
917 918
	 * processor.  This seems to give the best I/O throughput.
	 */
919 920 921 922 923 924
	if (likely(reply_queue == DEFAULT_REPLY_QUEUE))
		cp->reply_queue = smp_processor_id() % h->nreply_queues;
	else
		cp->reply_queue = reply_queue % h->nreply_queues;
	/*
	 * Set the bits in the address sent down to include:
925 926 927 928 929 930 931
	 *  - performant mode bit not used in ioaccel mode 2
	 *  - pull count (bits 0-3)
	 *  - command type isn't needed for ioaccel2
	 */
	c->busaddr |= (h->ioaccel2_blockFetchTable[cp->sg_count]);
}

932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960
static int is_firmware_flash_cmd(u8 *cdb)
{
	return cdb[0] == BMIC_WRITE && cdb[6] == BMIC_FLASH_FIRMWARE;
}

/*
 * During firmware flash, the heartbeat register may not update as frequently
 * as it should.  So we dial down lockup detection during firmware flash. and
 * dial it back up when firmware flash completes.
 */
#define HEARTBEAT_SAMPLE_INTERVAL_DURING_FLASH (240 * HZ)
#define HEARTBEAT_SAMPLE_INTERVAL (30 * HZ)
static void dial_down_lockup_detection_during_fw_flash(struct ctlr_info *h,
		struct CommandList *c)
{
	if (!is_firmware_flash_cmd(c->Request.CDB))
		return;
	atomic_inc(&h->firmware_flash_in_progress);
	h->heartbeat_sample_interval = HEARTBEAT_SAMPLE_INTERVAL_DURING_FLASH;
}

static void dial_up_lockup_detection_on_fw_flash_complete(struct ctlr_info *h,
		struct CommandList *c)
{
	if (is_firmware_flash_cmd(c->Request.CDB) &&
		atomic_dec_and_test(&h->firmware_flash_in_progress))
		h->heartbeat_sample_interval = HEARTBEAT_SAMPLE_INTERVAL;
}

961 962
static void __enqueue_cmd_and_start_io(struct ctlr_info *h,
	struct CommandList *c, int reply_queue)
963
{
964 965
	dial_down_lockup_detection_during_fw_flash(h, c);
	atomic_inc(&h->commands_outstanding);
966 967
	switch (c->cmd_type) {
	case CMD_IOACCEL1:
968
		set_ioaccel1_performant_mode(h, c, reply_queue);
969
		writel(c->busaddr, h->vaddr + SA5_REQUEST_PORT_OFFSET);
970 971
		break;
	case CMD_IOACCEL2:
972
		set_ioaccel2_performant_mode(h, c, reply_queue);
973
		writel(c->busaddr, h->vaddr + IOACCEL2_INBOUND_POSTQ_32);
974
		break;
975 976 977 978
	case IOACCEL2_TMF:
		set_ioaccel2_tmf_performant_mode(h, c, reply_queue);
		writel(c->busaddr, h->vaddr + IOACCEL2_INBOUND_POSTQ_32);
		break;
979
	default:
980
		set_performant_mode(h, c, reply_queue);
981
		h->access.submit_command(h, c);
982
	}
983 984
}

985
static void enqueue_cmd_and_start_io(struct ctlr_info *h, struct CommandList *c)
986
{
W
Webb Scales 已提交
987
	if (unlikely(hpsa_is_pending_event(c)))
988 989
		return finish_cmd(c);

990 991 992
	__enqueue_cmd_and_start_io(h, c, DEFAULT_REPLY_QUEUE);
}

993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006
static inline int is_hba_lunid(unsigned char scsi3addr[])
{
	return memcmp(scsi3addr, RAID_CTLR_LUNID, 8) == 0;
}

static inline int is_scsi_rev_5(struct ctlr_info *h)
{
	if (!h->hba_inquiry_data)
		return 0;
	if ((h->hba_inquiry_data[2] & 0x07) == 5)
		return 1;
	return 0;
}

1007 1008 1009 1010 1011 1012 1013
static int hpsa_find_target_lun(struct ctlr_info *h,
	unsigned char scsi3addr[], int bus, int *target, int *lun)
{
	/* finds an unused bus, target, lun for a new physical device
	 * assumes h->devlock is held
	 */
	int i, found = 0;
1014
	DECLARE_BITMAP(lun_taken, HPSA_MAX_DEVICES);
1015

1016
	bitmap_zero(lun_taken, HPSA_MAX_DEVICES);
1017 1018 1019

	for (i = 0; i < h->ndevices; i++) {
		if (h->dev[i]->bus == bus && h->dev[i]->target != -1)
1020
			__set_bit(h->dev[i]->target, lun_taken);
1021 1022
	}

1023 1024 1025 1026 1027 1028
	i = find_first_zero_bit(lun_taken, HPSA_MAX_DEVICES);
	if (i < HPSA_MAX_DEVICES) {
		/* *bus = 1; */
		*target = i;
		*lun = 0;
		found = 1;
1029 1030 1031 1032
	}
	return !found;
}

1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049
static inline void hpsa_show_dev_msg(const char *level, struct ctlr_info *h,
	struct hpsa_scsi_dev_t *dev, char *description)
{
	dev_printk(level, &h->pdev->dev,
			"scsi %d:%d:%d:%d: %s %s %.8s %.16s RAID-%s SSDSmartPathCap%c En%c Exp=%d\n",
			h->scsi_host->host_no, dev->bus, dev->target, dev->lun,
			description,
			scsi_device_type(dev->devtype),
			dev->vendor,
			dev->model,
			dev->raid_level > RAID_UNKNOWN ?
				"RAID-?" : raid_label[dev->raid_level],
			dev->offload_config ? '+' : '-',
			dev->offload_enabled ? '+' : '-',
			dev->expose_state);
}

1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060
/* Add an entry into h->dev[] array. */
static int hpsa_scsi_add_entry(struct ctlr_info *h, int hostno,
		struct hpsa_scsi_dev_t *device,
		struct hpsa_scsi_dev_t *added[], int *nadded)
{
	/* assumes h->devlock is held */
	int n = h->ndevices;
	int i;
	unsigned char addr1[8], addr2[8];
	struct hpsa_scsi_dev_t *sd;

1061
	if (n >= HPSA_MAX_DEVICES) {
1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073
		dev_err(&h->pdev->dev, "too many devices, some will be "
			"inaccessible.\n");
		return -1;
	}

	/* physical devices do not have lun or target assigned until now. */
	if (device->lun != -1)
		/* Logical device, lun is already assigned. */
		goto lun_assigned;

	/* If this device a non-zero lun of a multi-lun device
	 * byte 4 of the 8-byte LUN addr will contain the logical
D
Don Brace 已提交
1074
	 * unit no, zero otherwise.
1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116
	 */
	if (device->scsi3addr[4] == 0) {
		/* This is not a non-zero lun of a multi-lun device */
		if (hpsa_find_target_lun(h, device->scsi3addr,
			device->bus, &device->target, &device->lun) != 0)
			return -1;
		goto lun_assigned;
	}

	/* This is a non-zero lun of a multi-lun device.
	 * Search through our list and find the device which
	 * has the same 8 byte LUN address, excepting byte 4.
	 * Assign the same bus and target for this new LUN.
	 * Use the logical unit number from the firmware.
	 */
	memcpy(addr1, device->scsi3addr, 8);
	addr1[4] = 0;
	for (i = 0; i < n; i++) {
		sd = h->dev[i];
		memcpy(addr2, sd->scsi3addr, 8);
		addr2[4] = 0;
		/* differ only in byte 4? */
		if (memcmp(addr1, addr2, 8) == 0) {
			device->bus = sd->bus;
			device->target = sd->target;
			device->lun = device->scsi3addr[4];
			break;
		}
	}
	if (device->lun == -1) {
		dev_warn(&h->pdev->dev, "physical device with no LUN=0,"
			" suspect firmware bug or unsupported hardware "
			"configuration.\n");
			return -1;
	}

lun_assigned:

	h->dev[n] = device;
	h->ndevices++;
	added[*nadded] = device;
	(*nadded)++;
1117 1118
	hpsa_show_dev_msg(KERN_INFO, h, device,
		device->expose_state & HPSA_SCSI_ADD ? "added" : "masked");
1119 1120
	device->offload_to_be_enabled = device->offload_enabled;
	device->offload_enabled = 0;
1121 1122 1123
	return 0;
}

1124 1125 1126 1127
/* Update an entry in h->dev[] array. */
static void hpsa_scsi_update_entry(struct ctlr_info *h, int hostno,
	int entry, struct hpsa_scsi_dev_t *new_entry)
{
1128
	int offload_enabled;
1129 1130 1131 1132 1133
	/* assumes h->devlock is held */
	BUG_ON(entry < 0 || entry >= HPSA_MAX_DEVICES);

	/* Raid level changed. */
	h->dev[entry]->raid_level = new_entry->raid_level;
1134

1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147
	/* Raid offload parameters changed.  Careful about the ordering. */
	if (new_entry->offload_config && new_entry->offload_enabled) {
		/*
		 * if drive is newly offload_enabled, we want to copy the
		 * raid map data first.  If previously offload_enabled and
		 * offload_config were set, raid map data had better be
		 * the same as it was before.  if raid map data is changed
		 * then it had better be the case that
		 * h->dev[entry]->offload_enabled is currently 0.
		 */
		h->dev[entry]->raid_map = new_entry->raid_map;
		h->dev[entry]->ioaccel_handle = new_entry->ioaccel_handle;
	}
1148 1149 1150 1151 1152
	if (new_entry->hba_ioaccel_enabled) {
		h->dev[entry]->ioaccel_handle = new_entry->ioaccel_handle;
		wmb(); /* set ioaccel_handle *before* hba_ioaccel_enabled */
	}
	h->dev[entry]->hba_ioaccel_enabled = new_entry->hba_ioaccel_enabled;
1153
	h->dev[entry]->offload_config = new_entry->offload_config;
1154
	h->dev[entry]->offload_to_mirror = new_entry->offload_to_mirror;
1155
	h->dev[entry]->queue_depth = new_entry->queue_depth;
1156

1157 1158 1159 1160 1161 1162 1163 1164 1165
	/*
	 * We can turn off ioaccel offload now, but need to delay turning
	 * it on until we can update h->dev[entry]->phys_disk[], but we
	 * can't do that until all the devices are updated.
	 */
	h->dev[entry]->offload_to_be_enabled = new_entry->offload_enabled;
	if (!new_entry->offload_enabled)
		h->dev[entry]->offload_enabled = 0;

1166 1167
	offload_enabled = h->dev[entry]->offload_enabled;
	h->dev[entry]->offload_enabled = h->dev[entry]->offload_to_be_enabled;
1168
	hpsa_show_dev_msg(KERN_INFO, h, h->dev[entry], "updated");
1169
	h->dev[entry]->offload_enabled = offload_enabled;
1170 1171
}

1172 1173 1174 1175 1176 1177 1178
/* Replace an entry from h->dev[] array. */
static void hpsa_scsi_replace_entry(struct ctlr_info *h, int hostno,
	int entry, struct hpsa_scsi_dev_t *new_entry,
	struct hpsa_scsi_dev_t *added[], int *nadded,
	struct hpsa_scsi_dev_t *removed[], int *nremoved)
{
	/* assumes h->devlock is held */
1179
	BUG_ON(entry < 0 || entry >= HPSA_MAX_DEVICES);
1180 1181
	removed[*nremoved] = h->dev[entry];
	(*nremoved)++;
1182 1183 1184 1185 1186 1187 1188 1189 1190 1191

	/*
	 * New physical devices won't have target/lun assigned yet
	 * so we need to preserve the values in the slot we are replacing.
	 */
	if (new_entry->target == -1) {
		new_entry->target = h->dev[entry]->target;
		new_entry->lun = h->dev[entry]->lun;
	}

1192 1193 1194
	h->dev[entry] = new_entry;
	added[*nadded] = new_entry;
	(*nadded)++;
1195
	hpsa_show_dev_msg(KERN_INFO, h, new_entry, "replaced");
1196 1197
	new_entry->offload_to_be_enabled = new_entry->offload_enabled;
	new_entry->offload_enabled = 0;
1198 1199
}

1200 1201 1202 1203 1204 1205 1206 1207
/* Remove an entry from h->dev[] array. */
static void hpsa_scsi_remove_entry(struct ctlr_info *h, int hostno, int entry,
	struct hpsa_scsi_dev_t *removed[], int *nremoved)
{
	/* assumes h->devlock is held */
	int i;
	struct hpsa_scsi_dev_t *sd;

1208
	BUG_ON(entry < 0 || entry >= HPSA_MAX_DEVICES);
1209 1210 1211 1212 1213 1214 1215 1216

	sd = h->dev[entry];
	removed[*nremoved] = h->dev[entry];
	(*nremoved)++;

	for (i = entry; i < h->ndevices-1; i++)
		h->dev[i] = h->dev[i+1];
	h->ndevices--;
1217
	hpsa_show_dev_msg(KERN_INFO, h, sd, "removed");
1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275
}

#define SCSI3ADDR_EQ(a, b) ( \
	(a)[7] == (b)[7] && \
	(a)[6] == (b)[6] && \
	(a)[5] == (b)[5] && \
	(a)[4] == (b)[4] && \
	(a)[3] == (b)[3] && \
	(a)[2] == (b)[2] && \
	(a)[1] == (b)[1] && \
	(a)[0] == (b)[0])

static void fixup_botched_add(struct ctlr_info *h,
	struct hpsa_scsi_dev_t *added)
{
	/* called when scsi_add_device fails in order to re-adjust
	 * h->dev[] to match the mid layer's view.
	 */
	unsigned long flags;
	int i, j;

	spin_lock_irqsave(&h->lock, flags);
	for (i = 0; i < h->ndevices; i++) {
		if (h->dev[i] == added) {
			for (j = i; j < h->ndevices-1; j++)
				h->dev[j] = h->dev[j+1];
			h->ndevices--;
			break;
		}
	}
	spin_unlock_irqrestore(&h->lock, flags);
	kfree(added);
}

static inline int device_is_the_same(struct hpsa_scsi_dev_t *dev1,
	struct hpsa_scsi_dev_t *dev2)
{
	/* we compare everything except lun and target as these
	 * are not yet assigned.  Compare parts likely
	 * to differ first
	 */
	if (memcmp(dev1->scsi3addr, dev2->scsi3addr,
		sizeof(dev1->scsi3addr)) != 0)
		return 0;
	if (memcmp(dev1->device_id, dev2->device_id,
		sizeof(dev1->device_id)) != 0)
		return 0;
	if (memcmp(dev1->model, dev2->model, sizeof(dev1->model)) != 0)
		return 0;
	if (memcmp(dev1->vendor, dev2->vendor, sizeof(dev1->vendor)) != 0)
		return 0;
	if (dev1->devtype != dev2->devtype)
		return 0;
	if (dev1->bus != dev2->bus)
		return 0;
	return 1;
}

1276 1277 1278 1279 1280 1281 1282 1283 1284
static inline int device_updated(struct hpsa_scsi_dev_t *dev1,
	struct hpsa_scsi_dev_t *dev2)
{
	/* Device attributes that can change, but don't mean
	 * that the device is a different device, nor that the OS
	 * needs to be told anything about the change.
	 */
	if (dev1->raid_level != dev2->raid_level)
		return 1;
1285 1286 1287 1288
	if (dev1->offload_config != dev2->offload_config)
		return 1;
	if (dev1->offload_enabled != dev2->offload_enabled)
		return 1;
1289 1290
	if (dev1->queue_depth != dev2->queue_depth)
		return 1;
1291 1292 1293
	return 0;
}

1294 1295 1296
/* Find needle in haystack.  If exact match found, return DEVICE_SAME,
 * and return needle location in *index.  If scsi3addr matches, but not
 * vendor, model, serial num, etc. return DEVICE_CHANGED, and return needle
1297 1298 1299 1300
 * location in *index.
 * In the case of a minor device attribute change, such as RAID level, just
 * return DEVICE_UPDATED, along with the updated device's location in index.
 * If needle not found, return DEVICE_NOT_FOUND.
1301 1302 1303 1304 1305 1306 1307 1308 1309
 */
static int hpsa_scsi_find_entry(struct hpsa_scsi_dev_t *needle,
	struct hpsa_scsi_dev_t *haystack[], int haystack_size,
	int *index)
{
	int i;
#define DEVICE_NOT_FOUND 0
#define DEVICE_CHANGED 1
#define DEVICE_SAME 2
1310
#define DEVICE_UPDATED 3
1311
	for (i = 0; i < haystack_size; i++) {
1312 1313
		if (haystack[i] == NULL) /* previously removed. */
			continue;
1314 1315
		if (SCSI3ADDR_EQ(needle->scsi3addr, haystack[i]->scsi3addr)) {
			*index = i;
1316 1317 1318
			if (device_is_the_same(needle, haystack[i])) {
				if (device_updated(needle, haystack[i]))
					return DEVICE_UPDATED;
1319
				return DEVICE_SAME;
1320
			} else {
1321 1322 1323
				/* Keep offline devices offline */
				if (needle->volume_offline)
					return DEVICE_NOT_FOUND;
1324
				return DEVICE_CHANGED;
1325
			}
1326 1327 1328 1329 1330 1331
		}
	}
	*index = -1;
	return DEVICE_NOT_FOUND;
}

1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435
static void hpsa_monitor_offline_device(struct ctlr_info *h,
					unsigned char scsi3addr[])
{
	struct offline_device_entry *device;
	unsigned long flags;

	/* Check to see if device is already on the list */
	spin_lock_irqsave(&h->offline_device_lock, flags);
	list_for_each_entry(device, &h->offline_device_list, offline_list) {
		if (memcmp(device->scsi3addr, scsi3addr,
			sizeof(device->scsi3addr)) == 0) {
			spin_unlock_irqrestore(&h->offline_device_lock, flags);
			return;
		}
	}
	spin_unlock_irqrestore(&h->offline_device_lock, flags);

	/* Device is not on the list, add it. */
	device = kmalloc(sizeof(*device), GFP_KERNEL);
	if (!device) {
		dev_warn(&h->pdev->dev, "out of memory in %s\n", __func__);
		return;
	}
	memcpy(device->scsi3addr, scsi3addr, sizeof(device->scsi3addr));
	spin_lock_irqsave(&h->offline_device_lock, flags);
	list_add_tail(&device->offline_list, &h->offline_device_list);
	spin_unlock_irqrestore(&h->offline_device_lock, flags);
}

/* Print a message explaining various offline volume states */
static void hpsa_show_volume_status(struct ctlr_info *h,
	struct hpsa_scsi_dev_t *sd)
{
	if (sd->volume_offline == HPSA_VPD_LV_STATUS_UNSUPPORTED)
		dev_info(&h->pdev->dev,
			"C%d:B%d:T%d:L%d Volume status is not available through vital product data pages.\n",
			h->scsi_host->host_no,
			sd->bus, sd->target, sd->lun);
	switch (sd->volume_offline) {
	case HPSA_LV_OK:
		break;
	case HPSA_LV_UNDERGOING_ERASE:
		dev_info(&h->pdev->dev,
			"C%d:B%d:T%d:L%d Volume is undergoing background erase process.\n",
			h->scsi_host->host_no,
			sd->bus, sd->target, sd->lun);
		break;
	case HPSA_LV_UNDERGOING_RPI:
		dev_info(&h->pdev->dev,
			"C%d:B%d:T%d:L%d Volume is undergoing rapid parity initialization process.\n",
			h->scsi_host->host_no,
			sd->bus, sd->target, sd->lun);
		break;
	case HPSA_LV_PENDING_RPI:
		dev_info(&h->pdev->dev,
				"C%d:B%d:T%d:L%d Volume is queued for rapid parity initialization process.\n",
				h->scsi_host->host_no,
				sd->bus, sd->target, sd->lun);
		break;
	case HPSA_LV_ENCRYPTED_NO_KEY:
		dev_info(&h->pdev->dev,
			"C%d:B%d:T%d:L%d Volume is encrypted and cannot be accessed because key is not present.\n",
			h->scsi_host->host_no,
			sd->bus, sd->target, sd->lun);
		break;
	case HPSA_LV_PLAINTEXT_IN_ENCRYPT_ONLY_CONTROLLER:
		dev_info(&h->pdev->dev,
			"C%d:B%d:T%d:L%d Volume is not encrypted and cannot be accessed because controller is in encryption-only mode.\n",
			h->scsi_host->host_no,
			sd->bus, sd->target, sd->lun);
		break;
	case HPSA_LV_UNDERGOING_ENCRYPTION:
		dev_info(&h->pdev->dev,
			"C%d:B%d:T%d:L%d Volume is undergoing encryption process.\n",
			h->scsi_host->host_no,
			sd->bus, sd->target, sd->lun);
		break;
	case HPSA_LV_UNDERGOING_ENCRYPTION_REKEYING:
		dev_info(&h->pdev->dev,
			"C%d:B%d:T%d:L%d Volume is undergoing encryption re-keying process.\n",
			h->scsi_host->host_no,
			sd->bus, sd->target, sd->lun);
		break;
	case HPSA_LV_ENCRYPTED_IN_NON_ENCRYPTED_CONTROLLER:
		dev_info(&h->pdev->dev,
			"C%d:B%d:T%d:L%d Volume is encrypted and cannot be accessed because controller does not have encryption enabled.\n",
			h->scsi_host->host_no,
			sd->bus, sd->target, sd->lun);
		break;
	case HPSA_LV_PENDING_ENCRYPTION:
		dev_info(&h->pdev->dev,
			"C%d:B%d:T%d:L%d Volume is pending migration to encrypted state, but process has not started.\n",
			h->scsi_host->host_no,
			sd->bus, sd->target, sd->lun);
		break;
	case HPSA_LV_PENDING_ENCRYPTION_REKEYING:
		dev_info(&h->pdev->dev,
			"C%d:B%d:T%d:L%d Volume is encrypted and is pending encryption rekeying.\n",
			h->scsi_host->host_no,
			sd->bus, sd->target, sd->lun);
		break;
	}
}

1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458
/*
 * Figure the list of physical drive pointers for a logical drive with
 * raid offload configured.
 */
static void hpsa_figure_phys_disk_ptrs(struct ctlr_info *h,
				struct hpsa_scsi_dev_t *dev[], int ndevices,
				struct hpsa_scsi_dev_t *logical_drive)
{
	struct raid_map_data *map = &logical_drive->raid_map;
	struct raid_map_disk_data *dd = &map->data[0];
	int i, j;
	int total_disks_per_row = le16_to_cpu(map->data_disks_per_row) +
				le16_to_cpu(map->metadata_disks_per_row);
	int nraid_map_entries = le16_to_cpu(map->row_cnt) *
				le16_to_cpu(map->layout_map_count) *
				total_disks_per_row;
	int nphys_disk = le16_to_cpu(map->layout_map_count) *
				total_disks_per_row;
	int qdepth;

	if (nraid_map_entries > RAID_MAP_MAX_ENTRIES)
		nraid_map_entries = RAID_MAP_MAX_ENTRIES;

W
Webb Scales 已提交
1459 1460
	logical_drive->nphysical_disks = nraid_map_entries;

1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489
	qdepth = 0;
	for (i = 0; i < nraid_map_entries; i++) {
		logical_drive->phys_disk[i] = NULL;
		if (!logical_drive->offload_config)
			continue;
		for (j = 0; j < ndevices; j++) {
			if (dev[j]->devtype != TYPE_DISK)
				continue;
			if (is_logical_dev_addr_mode(dev[j]->scsi3addr))
				continue;
			if (dev[j]->ioaccel_handle != dd[i].ioaccel_handle)
				continue;

			logical_drive->phys_disk[i] = dev[j];
			if (i < nphys_disk)
				qdepth = min(h->nr_cmds, qdepth +
				    logical_drive->phys_disk[i]->queue_depth);
			break;
		}

		/*
		 * This can happen if a physical drive is removed and
		 * the logical drive is degraded.  In that case, the RAID
		 * map data will refer to a physical disk which isn't actually
		 * present.  And in that case offload_enabled should already
		 * be 0, but we'll turn it off here just in case
		 */
		if (!logical_drive->phys_disk[i]) {
			logical_drive->offload_enabled = 0;
1490 1491
			logical_drive->offload_to_be_enabled = 0;
			logical_drive->queue_depth = 8;
1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513
		}
	}
	if (nraid_map_entries)
		/*
		 * This is correct for reads, too high for full stripe writes,
		 * way too high for partial stripe writes
		 */
		logical_drive->queue_depth = qdepth;
	else
		logical_drive->queue_depth = h->nr_cmds;
}

static void hpsa_update_log_drive_phys_drive_ptrs(struct ctlr_info *h,
				struct hpsa_scsi_dev_t *dev[], int ndevices)
{
	int i;

	for (i = 0; i < ndevices; i++) {
		if (dev[i]->devtype != TYPE_DISK)
			continue;
		if (!is_logical_dev_addr_mode(dev[i]->scsi3addr))
			continue;
1514 1515 1516 1517 1518 1519 1520 1521 1522 1523

		/*
		 * If offload is currently enabled, the RAID map and
		 * phys_disk[] assignment *better* not be changing
		 * and since it isn't changing, we do not need to
		 * update it.
		 */
		if (dev[i]->offload_enabled)
			continue;

1524 1525 1526 1527
		hpsa_figure_phys_disk_ptrs(h, dev, ndevices, dev[i]);
	}
}

1528
static void adjust_hpsa_scsi_table(struct ctlr_info *h, int hostno,
1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541
	struct hpsa_scsi_dev_t *sd[], int nsds)
{
	/* sd contains scsi3 addresses and devtypes, and inquiry
	 * data.  This function takes what's in sd to be the current
	 * reality and updates h->dev[] to reflect that reality.
	 */
	int i, entry, device_change, changes = 0;
	struct hpsa_scsi_dev_t *csd;
	unsigned long flags;
	struct hpsa_scsi_dev_t **added, **removed;
	int nadded, nremoved;
	struct Scsi_Host *sh = NULL;

1542 1543
	added = kzalloc(sizeof(*added) * HPSA_MAX_DEVICES, GFP_KERNEL);
	removed = kzalloc(sizeof(*removed) * HPSA_MAX_DEVICES, GFP_KERNEL);
1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556

	if (!added || !removed) {
		dev_warn(&h->pdev->dev, "out of memory in "
			"adjust_hpsa_scsi_table\n");
		goto free_and_out;
	}

	spin_lock_irqsave(&h->devlock, flags);

	/* find any devices in h->dev[] that are not in
	 * sd[] and remove them from h->dev[], and for any
	 * devices which have changed, remove the old device
	 * info and add the new device info.
1557 1558
	 * If minor device attributes change, just update
	 * the existing device structure.
1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572
	 */
	i = 0;
	nremoved = 0;
	nadded = 0;
	while (i < h->ndevices) {
		csd = h->dev[i];
		device_change = hpsa_scsi_find_entry(csd, sd, nsds, &entry);
		if (device_change == DEVICE_NOT_FOUND) {
			changes++;
			hpsa_scsi_remove_entry(h, hostno, i,
				removed, &nremoved);
			continue; /* remove ^^^, hence i not incremented */
		} else if (device_change == DEVICE_CHANGED) {
			changes++;
1573 1574
			hpsa_scsi_replace_entry(h, hostno, i, sd[entry],
				added, &nadded, removed, &nremoved);
1575 1576 1577 1578
			/* Set it to NULL to prevent it from being freed
			 * at the bottom of hpsa_update_scsi_devices()
			 */
			sd[entry] = NULL;
1579 1580
		} else if (device_change == DEVICE_UPDATED) {
			hpsa_scsi_update_entry(h, hostno, i, sd[entry]);
1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591
		}
		i++;
	}

	/* Now, make sure every device listed in sd[] is also
	 * listed in h->dev[], adding them if they aren't found
	 */

	for (i = 0; i < nsds; i++) {
		if (!sd[i]) /* if already added above. */
			continue;
1592 1593 1594 1595 1596 1597 1598 1599

		/* Don't add devices which are NOT READY, FORMAT IN PROGRESS
		 * as the SCSI mid-layer does not handle such devices well.
		 * It relentlessly loops sending TUR at 3Hz, then READ(10)
		 * at 160Hz, and prevents the system from coming up.
		 */
		if (sd[i]->volume_offline) {
			hpsa_show_volume_status(h, sd[i]);
1600
			hpsa_show_dev_msg(KERN_INFO, h, sd[i], "offline");
1601 1602 1603
			continue;
		}

1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619
		device_change = hpsa_scsi_find_entry(sd[i], h->dev,
					h->ndevices, &entry);
		if (device_change == DEVICE_NOT_FOUND) {
			changes++;
			if (hpsa_scsi_add_entry(h, hostno, sd[i],
				added, &nadded) != 0)
				break;
			sd[i] = NULL; /* prevent from being freed later. */
		} else if (device_change == DEVICE_CHANGED) {
			/* should never happen... */
			changes++;
			dev_warn(&h->pdev->dev,
				"device unexpectedly changed.\n");
			/* but if it does happen, we just ignore that device */
		}
	}
1620 1621 1622 1623 1624 1625 1626 1627
	hpsa_update_log_drive_phys_drive_ptrs(h, h->dev, h->ndevices);

	/* Now that h->dev[]->phys_disk[] is coherent, we can enable
	 * any logical drives that need it enabled.
	 */
	for (i = 0; i < h->ndevices; i++)
		h->dev[i]->offload_enabled = h->dev[i]->offload_to_be_enabled;

1628 1629
	spin_unlock_irqrestore(&h->devlock, flags);

1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640
	/* Monitor devices which are in one of several NOT READY states to be
	 * brought online later. This must be done without holding h->devlock,
	 * so don't touch h->dev[]
	 */
	for (i = 0; i < nsds; i++) {
		if (!sd[i]) /* if already added above. */
			continue;
		if (sd[i]->volume_offline)
			hpsa_monitor_offline_device(h, sd[i]->scsi3addr);
	}

1641 1642 1643 1644 1645 1646 1647 1648 1649 1650
	/* Don't notify scsi mid layer of any changes the first time through
	 * (or if there are no changes) scsi_scan_host will do it later the
	 * first time through.
	 */
	if (hostno == -1 || !changes)
		goto free_and_out;

	sh = h->scsi_host;
	/* Notify scsi mid layer of any removed devices */
	for (i = 0; i < nremoved; i++) {
1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663
		if (removed[i]->expose_state & HPSA_SCSI_ADD) {
			struct scsi_device *sdev =
				scsi_device_lookup(sh, removed[i]->bus,
					removed[i]->target, removed[i]->lun);
			if (sdev != NULL) {
				scsi_remove_device(sdev);
				scsi_device_put(sdev);
			} else {
				/*
				 * We don't expect to get here.
				 * future cmds to this device will get selection
				 * timeout as if the device was gone.
				 */
1664 1665
				hpsa_show_dev_msg(KERN_WARNING, h, removed[i],
					"didn't find device for removal.");
1666
			}
1667 1668 1669 1670 1671 1672 1673
		}
		kfree(removed[i]);
		removed[i] = NULL;
	}

	/* Notify scsi mid layer of any added devices */
	for (i = 0; i < nadded; i++) {
1674 1675
		if (!(added[i]->expose_state & HPSA_SCSI_ADD))
			continue;
1676 1677 1678
		if (scsi_add_device(sh, added[i]->bus,
			added[i]->target, added[i]->lun) == 0)
			continue;
1679 1680
		hpsa_show_dev_msg(KERN_WARNING, h, added[i],
					"addition failed, device not added.");
1681 1682 1683 1684
		/* now we have to remove it from h->dev,
		 * since it didn't get added to scsi mid layer
		 */
		fixup_botched_add(h, added[i]);
R
Robert Elliott 已提交
1685
		added[i] = NULL;
1686 1687 1688 1689 1690 1691 1692 1693
	}

free_and_out:
	kfree(added);
	kfree(removed);
}

/*
1694
 * Lookup bus/target/lun and return corresponding struct hpsa_scsi_dev_t *
1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720
 * Assume's h->devlock is held.
 */
static struct hpsa_scsi_dev_t *lookup_hpsa_scsi_dev(struct ctlr_info *h,
	int bus, int target, int lun)
{
	int i;
	struct hpsa_scsi_dev_t *sd;

	for (i = 0; i < h->ndevices; i++) {
		sd = h->dev[i];
		if (sd->bus == bus && sd->target == target && sd->lun == lun)
			return sd;
	}
	return NULL;
}

static int hpsa_slave_alloc(struct scsi_device *sdev)
{
	struct hpsa_scsi_dev_t *sd;
	unsigned long flags;
	struct ctlr_info *h;

	h = sdev_to_hba(sdev);
	spin_lock_irqsave(&h->devlock, flags);
	sd = lookup_hpsa_scsi_dev(h, sdev_channel(sdev),
		sdev_id(sdev), sdev->lun);
1721
	if (likely(sd)) {
1722
		atomic_set(&sd->ioaccel_cmds_out, 0);
1723 1724 1725
		sdev->hostdata = (sd->expose_state & HPSA_SCSI_ADD) ? sd : NULL;
	} else
		sdev->hostdata = NULL;
1726 1727 1728 1729
	spin_unlock_irqrestore(&h->devlock, flags);
	return 0;
}

1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749
/* configure scsi device based on internal per-device structure */
static int hpsa_slave_configure(struct scsi_device *sdev)
{
	struct hpsa_scsi_dev_t *sd;
	int queue_depth;

	sd = sdev->hostdata;
	sdev->no_uld_attach = !sd || !(sd->expose_state & HPSA_ULD_ATTACH);

	if (sd)
		queue_depth = sd->queue_depth != 0 ?
			sd->queue_depth : sdev->host->can_queue;
	else
		queue_depth = sdev->host->can_queue;

	scsi_change_queue_depth(sdev, queue_depth);

	return 0;
}

1750 1751
static void hpsa_slave_destroy(struct scsi_device *sdev)
{
1752
	/* nothing to do. */
1753 1754
}

1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794
static void hpsa_free_ioaccel2_sg_chain_blocks(struct ctlr_info *h)
{
	int i;

	if (!h->ioaccel2_cmd_sg_list)
		return;
	for (i = 0; i < h->nr_cmds; i++) {
		kfree(h->ioaccel2_cmd_sg_list[i]);
		h->ioaccel2_cmd_sg_list[i] = NULL;
	}
	kfree(h->ioaccel2_cmd_sg_list);
	h->ioaccel2_cmd_sg_list = NULL;
}

static int hpsa_allocate_ioaccel2_sg_chain_blocks(struct ctlr_info *h)
{
	int i;

	if (h->chainsize <= 0)
		return 0;

	h->ioaccel2_cmd_sg_list =
		kzalloc(sizeof(*h->ioaccel2_cmd_sg_list) * h->nr_cmds,
					GFP_KERNEL);
	if (!h->ioaccel2_cmd_sg_list)
		return -ENOMEM;
	for (i = 0; i < h->nr_cmds; i++) {
		h->ioaccel2_cmd_sg_list[i] =
			kmalloc(sizeof(*h->ioaccel2_cmd_sg_list[i]) *
					h->maxsgentries, GFP_KERNEL);
		if (!h->ioaccel2_cmd_sg_list[i])
			goto clean;
	}
	return 0;

clean:
	hpsa_free_ioaccel2_sg_chain_blocks(h);
	return -ENOMEM;
}

1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808
static void hpsa_free_sg_chain_blocks(struct ctlr_info *h)
{
	int i;

	if (!h->cmd_sg_list)
		return;
	for (i = 0; i < h->nr_cmds; i++) {
		kfree(h->cmd_sg_list[i]);
		h->cmd_sg_list[i] = NULL;
	}
	kfree(h->cmd_sg_list);
	h->cmd_sg_list = NULL;
}

R
Robert Elliott 已提交
1809
static int hpsa_alloc_sg_chain_blocks(struct ctlr_info *h)
1810 1811 1812 1813 1814 1815 1816 1817
{
	int i;

	if (h->chainsize <= 0)
		return 0;

	h->cmd_sg_list = kzalloc(sizeof(*h->cmd_sg_list) * h->nr_cmds,
				GFP_KERNEL);
1818 1819
	if (!h->cmd_sg_list) {
		dev_err(&h->pdev->dev, "Failed to allocate SG list\n");
1820
		return -ENOMEM;
1821
	}
1822 1823 1824
	for (i = 0; i < h->nr_cmds; i++) {
		h->cmd_sg_list[i] = kmalloc(sizeof(*h->cmd_sg_list[i]) *
						h->chainsize, GFP_KERNEL);
1825 1826
		if (!h->cmd_sg_list[i]) {
			dev_err(&h->pdev->dev, "Failed to allocate cmd SG\n");
1827
			goto clean;
1828
		}
1829 1830 1831 1832 1833 1834 1835 1836
	}
	return 0;

clean:
	hpsa_free_sg_chain_blocks(h);
	return -ENOMEM;
}

1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869
static int hpsa_map_ioaccel2_sg_chain_block(struct ctlr_info *h,
	struct io_accel2_cmd *cp, struct CommandList *c)
{
	struct ioaccel2_sg_element *chain_block;
	u64 temp64;
	u32 chain_size;

	chain_block = h->ioaccel2_cmd_sg_list[c->cmdindex];
	chain_size = le32_to_cpu(cp->data_len);
	temp64 = pci_map_single(h->pdev, chain_block, chain_size,
				PCI_DMA_TODEVICE);
	if (dma_mapping_error(&h->pdev->dev, temp64)) {
		/* prevent subsequent unmapping */
		cp->sg->address = 0;
		return -1;
	}
	cp->sg->address = cpu_to_le64(temp64);
	return 0;
}

static void hpsa_unmap_ioaccel2_sg_chain_block(struct ctlr_info *h,
	struct io_accel2_cmd *cp)
{
	struct ioaccel2_sg_element *chain_sg;
	u64 temp64;
	u32 chain_size;

	chain_sg = cp->sg;
	temp64 = le64_to_cpu(chain_sg->address);
	chain_size = le32_to_cpu(cp->data_len);
	pci_unmap_single(h->pdev, temp64, chain_size, PCI_DMA_TODEVICE);
}

1870
static int hpsa_map_sg_chain_block(struct ctlr_info *h,
1871 1872 1873 1874
	struct CommandList *c)
{
	struct SGDescriptor *chain_sg, *chain_block;
	u64 temp64;
1875
	u32 chain_len;
1876 1877 1878

	chain_sg = &c->SG[h->max_cmd_sg_entries - 1];
	chain_block = h->cmd_sg_list[c->cmdindex];
1879 1880
	chain_sg->Ext = cpu_to_le32(HPSA_SG_CHAIN);
	chain_len = sizeof(*chain_sg) *
D
Don Brace 已提交
1881
		(le16_to_cpu(c->Header.SGTotal) - h->max_cmd_sg_entries);
1882 1883
	chain_sg->Len = cpu_to_le32(chain_len);
	temp64 = pci_map_single(h->pdev, chain_block, chain_len,
1884
				PCI_DMA_TODEVICE);
1885 1886
	if (dma_mapping_error(&h->pdev->dev, temp64)) {
		/* prevent subsequent unmapping */
1887
		chain_sg->Addr = cpu_to_le64(0);
1888 1889
		return -1;
	}
1890
	chain_sg->Addr = cpu_to_le64(temp64);
1891
	return 0;
1892 1893 1894 1895 1896 1897 1898
}

static void hpsa_unmap_sg_chain_block(struct ctlr_info *h,
	struct CommandList *c)
{
	struct SGDescriptor *chain_sg;

1899
	if (le16_to_cpu(c->Header.SGTotal) <= h->max_cmd_sg_entries)
1900 1901 1902
		return;

	chain_sg = &c->SG[h->max_cmd_sg_entries - 1];
1903 1904
	pci_unmap_single(h->pdev, le64_to_cpu(chain_sg->Addr),
			le32_to_cpu(chain_sg->Len), PCI_DMA_TODEVICE);
1905 1906
}

1907 1908 1909 1910 1911 1912

/* Decode the various types of errors on ioaccel2 path.
 * Return 1 for any error that should generate a RAID path retry.
 * Return 0 for errors that don't require a RAID path retry.
 */
static int handle_ioaccel_mode2_error(struct ctlr_info *h,
1913 1914 1915 1916 1917
					struct CommandList *c,
					struct scsi_cmnd *cmd,
					struct io_accel2_cmd *c2)
{
	int data_len;
1918
	int retry = 0;
1919
	u32 ioaccel2_resid = 0;
1920 1921 1922 1923 1924 1925 1926

	switch (c2->error_data.serv_response) {
	case IOACCEL2_SERV_RESPONSE_COMPLETE:
		switch (c2->error_data.status) {
		case IOACCEL2_STATUS_SR_TASK_COMP_GOOD:
			break;
		case IOACCEL2_STATUS_SR_TASK_COMP_CHK_COND:
1927
			cmd->result |= SAM_STAT_CHECK_CONDITION;
1928
			if (c2->error_data.data_present !=
1929 1930 1931
					IOACCEL2_SENSE_DATA_PRESENT) {
				memset(cmd->sense_buffer, 0,
					SCSI_SENSE_BUFFERSIZE);
1932
				break;
1933
			}
1934 1935 1936 1937 1938 1939 1940 1941 1942
			/* copy the sense data */
			data_len = c2->error_data.sense_data_len;
			if (data_len > SCSI_SENSE_BUFFERSIZE)
				data_len = SCSI_SENSE_BUFFERSIZE;
			if (data_len > sizeof(c2->error_data.sense_data_buff))
				data_len =
					sizeof(c2->error_data.sense_data_buff);
			memcpy(cmd->sense_buffer,
				c2->error_data.sense_data_buff, data_len);
1943
			retry = 1;
1944 1945
			break;
		case IOACCEL2_STATUS_SR_TASK_COMP_BUSY:
1946
			retry = 1;
1947 1948
			break;
		case IOACCEL2_STATUS_SR_TASK_COMP_RES_CON:
1949
			retry = 1;
1950 1951
			break;
		case IOACCEL2_STATUS_SR_TASK_COMP_SET_FULL:
1952
			retry = 1;
1953 1954
			break;
		case IOACCEL2_STATUS_SR_TASK_COMP_ABORTED:
1955
			retry = 1;
1956 1957
			break;
		default:
1958
			retry = 1;
1959 1960 1961 1962
			break;
		}
		break;
	case IOACCEL2_SERV_RESPONSE_FAILURE:
1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984
		switch (c2->error_data.status) {
		case IOACCEL2_STATUS_SR_IO_ERROR:
		case IOACCEL2_STATUS_SR_IO_ABORTED:
		case IOACCEL2_STATUS_SR_OVERRUN:
			retry = 1;
			break;
		case IOACCEL2_STATUS_SR_UNDERRUN:
			cmd->result = (DID_OK << 16);		/* host byte */
			cmd->result |= (COMMAND_COMPLETE << 8);	/* msg byte */
			ioaccel2_resid = get_unaligned_le32(
						&c2->error_data.resid_cnt[0]);
			scsi_set_resid(cmd, ioaccel2_resid);
			break;
		case IOACCEL2_STATUS_SR_NO_PATH_TO_DEVICE:
		case IOACCEL2_STATUS_SR_INVALID_DEVICE:
		case IOACCEL2_STATUS_SR_IOACCEL_DISABLED:
			/* We will get an event from ctlr to trigger rescan */
			retry = 1;
			break;
		default:
			retry = 1;
		}
1985 1986 1987 1988 1989 1990
		break;
	case IOACCEL2_SERV_RESPONSE_TMF_COMPLETE:
		break;
	case IOACCEL2_SERV_RESPONSE_TMF_SUCCESS:
		break;
	case IOACCEL2_SERV_RESPONSE_TMF_REJECTED:
1991
		retry = 1;
1992 1993 1994 1995
		break;
	case IOACCEL2_SERV_RESPONSE_TMF_WRONG_LUN:
		break;
	default:
1996
		retry = 1;
1997 1998
		break;
	}
1999 2000

	return retry;	/* retry on raid path? */
2001 2002
}

2003 2004 2005
static void hpsa_cmd_resolve_events(struct ctlr_info *h,
		struct CommandList *c)
{
W
Webb Scales 已提交
2006 2007
	bool do_wake = false;

2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
	/*
	 * Prevent the following race in the abort handler:
	 *
	 * 1. LLD is requested to abort a SCSI command
	 * 2. The SCSI command completes
	 * 3. The struct CommandList associated with step 2 is made available
	 * 4. New I/O request to LLD to another LUN re-uses struct CommandList
	 * 5. Abort handler follows scsi_cmnd->host_scribble and
	 *    finds struct CommandList and tries to aborts it
	 * Now we have aborted the wrong command.
	 *
W
Webb Scales 已提交
2019 2020
	 * Reset c->scsi_cmd here so that the abort or reset handler will know
	 * this command has completed.  Then, check to see if the handler is
2021 2022 2023
	 * waiting for this command, and, if so, wake it.
	 */
	c->scsi_cmd = SCSI_CMD_IDLE;
W
Webb Scales 已提交
2024
	mb();	/* Declare command idle before checking for pending events. */
2025
	if (c->abort_pending) {
W
Webb Scales 已提交
2026
		do_wake = true;
2027 2028
		c->abort_pending = false;
	}
W
Webb Scales 已提交
2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047
	if (c->reset_pending) {
		unsigned long flags;
		struct hpsa_scsi_dev_t *dev;

		/*
		 * There appears to be a reset pending; lock the lock and
		 * reconfirm.  If so, then decrement the count of outstanding
		 * commands and wake the reset command if this is the last one.
		 */
		spin_lock_irqsave(&h->lock, flags);
		dev = c->reset_pending;		/* Re-fetch under the lock. */
		if (dev && atomic_dec_and_test(&dev->reset_cmds_out))
			do_wake = true;
		c->reset_pending = NULL;
		spin_unlock_irqrestore(&h->lock, flags);
	}

	if (do_wake)
		wake_up_all(&h->event_sync_wait_queue);
2048 2049
}

2050 2051 2052 2053 2054 2055 2056
static void hpsa_cmd_resolve_and_free(struct ctlr_info *h,
				      struct CommandList *c)
{
	hpsa_cmd_resolve_events(h, c);
	cmd_tagged_free(h, c);
}

2057 2058 2059
static void hpsa_cmd_free_and_done(struct ctlr_info *h,
		struct CommandList *c, struct scsi_cmnd *cmd)
{
2060
	hpsa_cmd_resolve_and_free(h, c);
2061 2062 2063 2064 2065 2066 2067 2068 2069
	cmd->scsi_done(cmd);
}

static void hpsa_retry_cmd(struct ctlr_info *h, struct CommandList *c)
{
	INIT_WORK(&c->work, hpsa_command_resubmit_worker);
	queue_work_on(raw_smp_processor_id(), h->resubmit_wq, &c->work);
}

2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080
static void hpsa_set_scsi_cmd_aborted(struct scsi_cmnd *cmd)
{
	cmd->result = DID_ABORT << 16;
}

static void hpsa_cmd_abort_and_free(struct ctlr_info *h, struct CommandList *c,
				    struct scsi_cmnd *cmd)
{
	hpsa_set_scsi_cmd_aborted(cmd);
	dev_warn(&h->pdev->dev, "CDB %16phN was aborted with status 0x%x\n",
			 c->Request.CDB, c->err_info->ScsiStatus);
2081
	hpsa_cmd_resolve_and_free(h, c);
2082 2083
}

2084 2085 2086 2087 2088 2089 2090 2091
static void process_ioaccel2_completion(struct ctlr_info *h,
		struct CommandList *c, struct scsi_cmnd *cmd,
		struct hpsa_scsi_dev_t *dev)
{
	struct io_accel2_cmd *c2 = &h->ioaccel2_cmd_pool[c->cmdindex];

	/* check for good status */
	if (likely(c2->error_data.serv_response == 0 &&
2092 2093
			c2->error_data.status == 0))
		return hpsa_cmd_free_and_done(h, c, cmd);
2094

2095 2096
	/*
	 * Any RAID offload error results in retry which will use
2097 2098 2099 2100 2101 2102
	 * the normal I/O path so the controller can handle whatever's
	 * wrong.
	 */
	if (is_logical_dev_addr_mode(dev->scsi3addr) &&
		c2->error_data.serv_response ==
			IOACCEL2_SERV_RESPONSE_FAILURE) {
2103 2104 2105
		if (c2->error_data.status ==
			IOACCEL2_STATUS_SR_IOACCEL_DISABLED)
			dev->offload_enabled = 0;
2106 2107

		return hpsa_retry_cmd(h, c);
2108
	}
2109 2110

	if (handle_ioaccel_mode2_error(h, c, cmd, c2))
2111
		return hpsa_retry_cmd(h, c);
2112

2113
	return hpsa_cmd_free_and_done(h, c, cmd);
2114 2115
}

2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143
/* Returns 0 on success, < 0 otherwise. */
static int hpsa_evaluate_tmf_status(struct ctlr_info *h,
					struct CommandList *cp)
{
	u8 tmf_status = cp->err_info->ScsiStatus;

	switch (tmf_status) {
	case CISS_TMF_COMPLETE:
		/*
		 * CISS_TMF_COMPLETE never happens, instead,
		 * ei->CommandStatus == 0 for this case.
		 */
	case CISS_TMF_SUCCESS:
		return 0;
	case CISS_TMF_INVALID_FRAME:
	case CISS_TMF_NOT_SUPPORTED:
	case CISS_TMF_FAILED:
	case CISS_TMF_WRONG_LUN:
	case CISS_TMF_OVERLAPPED_TAG:
		break;
	default:
		dev_warn(&h->pdev->dev, "Unknown TMF status: 0x%02x\n",
				tmf_status);
		break;
	}
	return -tmf_status;
}

2144
static void complete_scsi_command(struct CommandList *cp)
2145 2146 2147 2148
{
	struct scsi_cmnd *cmd;
	struct ctlr_info *h;
	struct ErrorInfo *ei;
2149
	struct hpsa_scsi_dev_t *dev;
2150
	struct io_accel2_cmd *c2;
2151

2152 2153 2154
	u8 sense_key;
	u8 asc;      /* additional sense code */
	u8 ascq;     /* additional sense code qualifier */
2155
	unsigned long sense_data_size;
2156 2157

	ei = cp->err_info;
2158
	cmd = cp->scsi_cmd;
2159
	h = cp->h;
2160
	dev = cmd->device->hostdata;
2161
	c2 = &h->ioaccel2_cmd_pool[cp->cmdindex];
2162 2163

	scsi_dma_unmap(cmd); /* undo the DMA mappings */
2164
	if ((cp->cmd_type == CMD_SCSI) &&
D
Don Brace 已提交
2165
		(le16_to_cpu(cp->Header.SGTotal) > h->max_cmd_sg_entries))
2166
		hpsa_unmap_sg_chain_block(h, cp);
2167

2168 2169 2170 2171
	if ((cp->cmd_type == CMD_IOACCEL2) &&
		(c2->sg[0].chain_indicator == IOACCEL2_CHAIN))
		hpsa_unmap_ioaccel2_sg_chain_block(h, c2);

2172 2173
	cmd->result = (DID_OK << 16); 		/* host byte */
	cmd->result |= (COMMAND_COMPLETE << 8);	/* msg byte */
2174

2175 2176 2177
	if (cp->cmd_type == CMD_IOACCEL2 || cp->cmd_type == CMD_IOACCEL1)
		atomic_dec(&cp->phys_disk->ioaccel_cmds_out);

2178 2179 2180 2181 2182 2183 2184 2185
	/*
	 * We check for lockup status here as it may be set for
	 * CMD_SCSI, CMD_IOACCEL1 and CMD_IOACCEL2 commands by
	 * fail_all_oustanding_cmds()
	 */
	if (unlikely(ei->CommandStatus == CMD_CTLR_LOCKUP)) {
		/* DID_NO_CONNECT will prevent a retry */
		cmd->result = DID_NO_CONNECT << 16;
2186
		return hpsa_cmd_free_and_done(h, cp, cmd);
2187 2188
	}

W
Webb Scales 已提交
2189 2190 2191 2192 2193 2194 2195
	if ((unlikely(hpsa_is_pending_event(cp)))) {
		if (cp->reset_pending)
			return hpsa_cmd_resolve_and_free(h, cp);
		if (cp->abort_pending)
			return hpsa_cmd_abort_and_free(h, cp, cmd);
	}

2196 2197 2198
	if (cp->cmd_type == CMD_IOACCEL2)
		return process_ioaccel2_completion(h, cp, cmd, dev);

2199
	scsi_set_resid(cmd, ei->ResidualCnt);
2200 2201
	if (ei->CommandStatus == 0)
		return hpsa_cmd_free_and_done(h, cp, cmd);
2202

2203 2204 2205 2206 2207
	/* For I/O accelerator commands, copy over some fields to the normal
	 * CISS header used below for error handling.
	 */
	if (cp->cmd_type == CMD_IOACCEL1) {
		struct io_accel1_cmd *c = &h->ioaccel_cmd_pool[cp->cmdindex];
D
Don Brace 已提交
2208 2209 2210 2211
		cp->Header.SGList = scsi_sg_count(cmd);
		cp->Header.SGTotal = cpu_to_le16(cp->Header.SGList);
		cp->Request.CDBLen = le16_to_cpu(c->io_flags) &
			IOACCEL1_IOFLAGS_CDBLEN_MASK;
2212
		cp->Header.tag = c->tag;
2213 2214
		memcpy(cp->Header.LUN.LunAddrBytes, c->CISS_LUN, 8);
		memcpy(cp->Request.CDB, c->CDB, cp->Request.CDBLen);
2215 2216 2217 2218 2219 2220 2221 2222

		/* Any RAID offload error results in retry which will use
		 * the normal I/O path so the controller can handle whatever's
		 * wrong.
		 */
		if (is_logical_dev_addr_mode(dev->scsi3addr)) {
			if (ei->CommandStatus == CMD_IOACCEL_DISABLED)
				dev->offload_enabled = 0;
W
Webb Scales 已提交
2223
			return hpsa_retry_cmd(h, cp);
2224
		}
2225 2226
	}

2227 2228 2229 2230
	/* an error has occurred */
	switch (ei->CommandStatus) {

	case CMD_TARGET_STATUS:
2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242
		cmd->result |= ei->ScsiStatus;
		/* copy the sense data */
		if (SCSI_SENSE_BUFFERSIZE < sizeof(ei->SenseInfo))
			sense_data_size = SCSI_SENSE_BUFFERSIZE;
		else
			sense_data_size = sizeof(ei->SenseInfo);
		if (ei->SenseLen < sense_data_size)
			sense_data_size = ei->SenseLen;
		memcpy(cmd->sense_buffer, ei->SenseInfo, sense_data_size);
		if (ei->ScsiStatus)
			decode_sense_data(ei->SenseInfo, sense_data_size,
				&sense_key, &asc, &ascq);
2243
		if (ei->ScsiStatus == SAM_STAT_CHECK_CONDITION) {
2244
			if (sense_key == ABORTED_COMMAND) {
2245
				cmd->result |= DID_SOFT_ERROR << 16;
2246 2247
				break;
			}
2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282
			break;
		}
		/* Problem was not a check condition
		 * Pass it up to the upper layers...
		 */
		if (ei->ScsiStatus) {
			dev_warn(&h->pdev->dev, "cp %p has status 0x%x "
				"Sense: 0x%x, ASC: 0x%x, ASCQ: 0x%x, "
				"Returning result: 0x%x\n",
				cp, ei->ScsiStatus,
				sense_key, asc, ascq,
				cmd->result);
		} else {  /* scsi status is zero??? How??? */
			dev_warn(&h->pdev->dev, "cp %p SCSI status was 0. "
				"Returning no connection.\n", cp),

			/* Ordinarily, this case should never happen,
			 * but there is a bug in some released firmware
			 * revisions that allows it to happen if, for
			 * example, a 4100 backplane loses power and
			 * the tape drive is in it.  We assume that
			 * it's a fatal error of some kind because we
			 * can't show that it wasn't. We will make it
			 * look like selection timeout since that is
			 * the most common reason for this to occur,
			 * and it's severe enough.
			 */

			cmd->result = DID_NO_CONNECT << 16;
		}
		break;

	case CMD_DATA_UNDERRUN: /* let mid layer handle it. */
		break;
	case CMD_DATA_OVERRUN:
2283 2284
		dev_warn(&h->pdev->dev,
			"CDB %16phN data overrun\n", cp->Request.CDB);
2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298
		break;
	case CMD_INVALID: {
		/* print_bytes(cp, sizeof(*cp), 1, 0);
		print_cmd(cp); */
		/* We get CMD_INVALID if you address a non-existent device
		 * instead of a selection timeout (no response).  You will
		 * see this if you yank out a drive, then try to access it.
		 * This is kind of a shame because it means that any other
		 * CMD_INVALID (e.g. driver bug) will get interpreted as a
		 * missing target. */
		cmd->result = DID_NO_CONNECT << 16;
	}
		break;
	case CMD_PROTOCOL_ERR:
2299
		cmd->result = DID_ERROR << 16;
2300 2301
		dev_warn(&h->pdev->dev, "CDB %16phN : protocol error\n",
				cp->Request.CDB);
2302 2303 2304
		break;
	case CMD_HARDWARE_ERR:
		cmd->result = DID_ERROR << 16;
2305 2306
		dev_warn(&h->pdev->dev, "CDB %16phN : hardware error\n",
			cp->Request.CDB);
2307 2308 2309
		break;
	case CMD_CONNECTION_LOST:
		cmd->result = DID_ERROR << 16;
2310 2311
		dev_warn(&h->pdev->dev, "CDB %16phN : connection lost\n",
			cp->Request.CDB);
2312 2313
		break;
	case CMD_ABORTED:
2314 2315
		/* Return now to avoid calling scsi_done(). */
		return hpsa_cmd_abort_and_free(h, cp, cmd);
2316 2317
	case CMD_ABORT_FAILED:
		cmd->result = DID_ERROR << 16;
2318 2319
		dev_warn(&h->pdev->dev, "CDB %16phN : abort failed\n",
			cp->Request.CDB);
2320 2321
		break;
	case CMD_UNSOLICITED_ABORT:
2322
		cmd->result = DID_SOFT_ERROR << 16; /* retry the command */
2323 2324
		dev_warn(&h->pdev->dev, "CDB %16phN : unsolicited abort\n",
			cp->Request.CDB);
2325 2326 2327
		break;
	case CMD_TIMEOUT:
		cmd->result = DID_TIME_OUT << 16;
2328 2329
		dev_warn(&h->pdev->dev, "CDB %16phN timed out\n",
			cp->Request.CDB);
2330
		break;
2331 2332 2333 2334
	case CMD_UNABORTABLE:
		cmd->result = DID_ERROR << 16;
		dev_warn(&h->pdev->dev, "Command unabortable\n");
		break;
2335 2336 2337 2338
	case CMD_TMF_STATUS:
		if (hpsa_evaluate_tmf_status(h, cp)) /* TMF failed? */
			cmd->result = DID_ERROR << 16;
		break;
2339 2340 2341 2342 2343 2344 2345 2346
	case CMD_IOACCEL_DISABLED:
		/* This only handles the direct pass-through case since RAID
		 * offload is handled above.  Just attempt a retry.
		 */
		cmd->result = DID_SOFT_ERROR << 16;
		dev_warn(&h->pdev->dev,
				"cp %p had HP SSD Smart Path error\n", cp);
		break;
2347 2348 2349 2350 2351
	default:
		cmd->result = DID_ERROR << 16;
		dev_warn(&h->pdev->dev, "cp %p returned unknown status %x\n",
				cp, ei->CommandStatus);
	}
2352 2353

	return hpsa_cmd_free_and_done(h, cp, cmd);
2354 2355 2356 2357 2358 2359 2360
}

static void hpsa_pci_unmap(struct pci_dev *pdev,
	struct CommandList *c, int sg_used, int data_direction)
{
	int i;

2361 2362 2363 2364
	for (i = 0; i < sg_used; i++)
		pci_unmap_single(pdev, (dma_addr_t) le64_to_cpu(c->SG[i].Addr),
				le32_to_cpu(c->SG[i].Len),
				data_direction);
2365 2366
}

2367
static int hpsa_map_one(struct pci_dev *pdev,
2368 2369 2370 2371 2372
		struct CommandList *cp,
		unsigned char *buf,
		size_t buflen,
		int data_direction)
{
2373
	u64 addr64;
2374 2375 2376

	if (buflen == 0 || data_direction == PCI_DMA_NONE) {
		cp->Header.SGList = 0;
2377
		cp->Header.SGTotal = cpu_to_le16(0);
2378
		return 0;
2379 2380
	}

2381
	addr64 = pci_map_single(pdev, buf, buflen, data_direction);
2382
	if (dma_mapping_error(&pdev->dev, addr64)) {
2383
		/* Prevent subsequent unmap of something never mapped */
2384
		cp->Header.SGList = 0;
2385
		cp->Header.SGTotal = cpu_to_le16(0);
2386
		return -1;
2387
	}
2388 2389 2390 2391 2392
	cp->SG[0].Addr = cpu_to_le64(addr64);
	cp->SG[0].Len = cpu_to_le32(buflen);
	cp->SG[0].Ext = cpu_to_le32(HPSA_SG_LAST); /* we are not chaining */
	cp->Header.SGList = 1;   /* no. SGs contig in this cmd */
	cp->Header.SGTotal = cpu_to_le16(1); /* total sgs in cmd list */
2393
	return 0;
2394 2395
}

2396 2397 2398 2399
#define NO_TIMEOUT ((unsigned long) -1)
#define DEFAULT_TIMEOUT 30000 /* milliseconds */
static int hpsa_scsi_do_simple_cmd_core(struct ctlr_info *h,
	struct CommandList *c, int reply_queue, unsigned long timeout_msecs)
2400 2401 2402 2403
{
	DECLARE_COMPLETION_ONSTACK(wait);

	c->waiting = &wait;
2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425
	__enqueue_cmd_and_start_io(h, c, reply_queue);
	if (timeout_msecs == NO_TIMEOUT) {
		/* TODO: get rid of this no-timeout thing */
		wait_for_completion_io(&wait);
		return IO_OK;
	}
	if (!wait_for_completion_io_timeout(&wait,
					msecs_to_jiffies(timeout_msecs))) {
		dev_warn(&h->pdev->dev, "Command timed out.\n");
		return -ETIMEDOUT;
	}
	return IO_OK;
}

static int hpsa_scsi_do_simple_cmd(struct ctlr_info *h, struct CommandList *c,
				   int reply_queue, unsigned long timeout_msecs)
{
	if (unlikely(lockup_detected(h))) {
		c->err_info->CommandStatus = CMD_CTLR_LOCKUP;
		return IO_OK;
	}
	return hpsa_scsi_do_simple_cmd_core(h, c, reply_queue, timeout_msecs);
2426 2427
}

2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439
static u32 lockup_detected(struct ctlr_info *h)
{
	int cpu;
	u32 rc, *lockup_detected;

	cpu = get_cpu();
	lockup_detected = per_cpu_ptr(h->lockup_detected, cpu);
	rc = *lockup_detected;
	put_cpu();
	return rc;
}

2440
#define MAX_DRIVER_CMD_RETRIES 25
2441 2442
static int hpsa_scsi_do_simple_cmd_with_retry(struct ctlr_info *h,
	struct CommandList *c, int data_direction, unsigned long timeout_msecs)
2443
{
2444
	int backoff_time = 10, retry_count = 0;
2445
	int rc;
2446 2447

	do {
2448
		memset(c->err_info, 0, sizeof(*c->err_info));
2449 2450 2451 2452
		rc = hpsa_scsi_do_simple_cmd(h, c, DEFAULT_REPLY_QUEUE,
						  timeout_msecs);
		if (rc)
			break;
2453
		retry_count++;
2454 2455 2456 2457 2458
		if (retry_count > 3) {
			msleep(backoff_time);
			if (backoff_time < 1000)
				backoff_time *= 2;
		}
2459
	} while ((check_for_unit_attention(h, c) ||
2460 2461
			check_for_busy(h, c)) &&
			retry_count <= MAX_DRIVER_CMD_RETRIES);
2462
	hpsa_pci_unmap(h->pdev, c, 1, data_direction);
2463 2464 2465
	if (retry_count > MAX_DRIVER_CMD_RETRIES)
		rc = -EIO;
	return rc;
2466 2467
}

2468 2469
static void hpsa_print_cmd(struct ctlr_info *h, char *txt,
				struct CommandList *c)
2470
{
2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487
	const u8 *cdb = c->Request.CDB;
	const u8 *lun = c->Header.LUN.LunAddrBytes;

	dev_warn(&h->pdev->dev, "%s: LUN:%02x%02x%02x%02x%02x%02x%02x%02x"
	" CDB:%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x\n",
		txt, lun[0], lun[1], lun[2], lun[3],
		lun[4], lun[5], lun[6], lun[7],
		cdb[0], cdb[1], cdb[2], cdb[3],
		cdb[4], cdb[5], cdb[6], cdb[7],
		cdb[8], cdb[9], cdb[10], cdb[11],
		cdb[12], cdb[13], cdb[14], cdb[15]);
}

static void hpsa_scsi_interpret_error(struct ctlr_info *h,
			struct CommandList *cp)
{
	const struct ErrorInfo *ei = cp->err_info;
2488
	struct device *d = &cp->h->pdev->dev;
2489 2490
	u8 sense_key, asc, ascq;
	int sense_len;
2491 2492 2493

	switch (ei->CommandStatus) {
	case CMD_TARGET_STATUS:
2494 2495 2496 2497 2498 2499
		if (ei->SenseLen > sizeof(ei->SenseInfo))
			sense_len = sizeof(ei->SenseInfo);
		else
			sense_len = ei->SenseLen;
		decode_sense_data(ei->SenseInfo, sense_len,
					&sense_key, &asc, &ascq);
2500 2501
		hpsa_print_cmd(h, "SCSI status", cp);
		if (ei->ScsiStatus == SAM_STAT_CHECK_CONDITION)
2502 2503
			dev_warn(d, "SCSI Status = 02, Sense key = 0x%02x, ASC = 0x%02x, ASCQ = 0x%02x\n",
				sense_key, asc, ascq);
2504
		else
2505
			dev_warn(d, "SCSI Status = 0x%02x\n", ei->ScsiStatus);
2506 2507 2508 2509 2510 2511 2512 2513 2514
		if (ei->ScsiStatus == 0)
			dev_warn(d, "SCSI status is abnormally zero.  "
			"(probably indicates selection timeout "
			"reported incorrectly due to a known "
			"firmware bug, circa July, 2001.)\n");
		break;
	case CMD_DATA_UNDERRUN: /* let mid layer handle it. */
		break;
	case CMD_DATA_OVERRUN:
2515
		hpsa_print_cmd(h, "overrun condition", cp);
2516 2517 2518 2519 2520
		break;
	case CMD_INVALID: {
		/* controller unfortunately reports SCSI passthru's
		 * to non-existent targets as invalid commands.
		 */
2521 2522
		hpsa_print_cmd(h, "invalid command", cp);
		dev_warn(d, "probably means device no longer present\n");
2523 2524 2525
		}
		break;
	case CMD_PROTOCOL_ERR:
2526
		hpsa_print_cmd(h, "protocol error", cp);
2527 2528
		break;
	case CMD_HARDWARE_ERR:
2529
		hpsa_print_cmd(h, "hardware error", cp);
2530 2531
		break;
	case CMD_CONNECTION_LOST:
2532
		hpsa_print_cmd(h, "connection lost", cp);
2533 2534
		break;
	case CMD_ABORTED:
2535
		hpsa_print_cmd(h, "aborted", cp);
2536 2537
		break;
	case CMD_ABORT_FAILED:
2538
		hpsa_print_cmd(h, "abort failed", cp);
2539 2540
		break;
	case CMD_UNSOLICITED_ABORT:
2541
		hpsa_print_cmd(h, "unsolicited abort", cp);
2542 2543
		break;
	case CMD_TIMEOUT:
2544
		hpsa_print_cmd(h, "timed out", cp);
2545
		break;
2546
	case CMD_UNABORTABLE:
2547
		hpsa_print_cmd(h, "unabortable", cp);
2548
		break;
2549 2550 2551
	case CMD_CTLR_LOCKUP:
		hpsa_print_cmd(h, "controller lockup detected", cp);
		break;
2552
	default:
2553 2554
		hpsa_print_cmd(h, "unknown status", cp);
		dev_warn(d, "Unknown command status %x\n",
2555 2556 2557 2558 2559
				ei->CommandStatus);
	}
}

static int hpsa_scsi_do_inquiry(struct ctlr_info *h, unsigned char *scsi3addr,
2560
			u16 page, unsigned char *buf,
2561 2562 2563 2564 2565 2566
			unsigned char bufsize)
{
	int rc = IO_OK;
	struct CommandList *c;
	struct ErrorInfo *ei;

2567
	c = cmd_alloc(h);
2568

2569 2570 2571 2572 2573
	if (fill_cmd(c, HPSA_INQUIRY, h, buf, bufsize,
			page, scsi3addr, TYPE_CMD)) {
		rc = -1;
		goto out;
	}
2574 2575 2576 2577
	rc = hpsa_scsi_do_simple_cmd_with_retry(h, c,
					PCI_DMA_FROMDEVICE, NO_TIMEOUT);
	if (rc)
		goto out;
2578 2579
	ei = c->err_info;
	if (ei->CommandStatus != 0 && ei->CommandStatus != CMD_DATA_UNDERRUN) {
2580
		hpsa_scsi_interpret_error(h, c);
2581 2582
		rc = -1;
	}
2583
out:
2584
	cmd_free(h, c);
2585 2586 2587
	return rc;
}

2588 2589 2590 2591 2592 2593 2594 2595
static int hpsa_bmic_ctrl_mode_sense(struct ctlr_info *h,
		unsigned char *scsi3addr, unsigned char page,
		struct bmic_controller_parameters *buf, size_t bufsize)
{
	int rc = IO_OK;
	struct CommandList *c;
	struct ErrorInfo *ei;

2596
	c = cmd_alloc(h);
2597 2598 2599 2600 2601
	if (fill_cmd(c, BMIC_SENSE_CONTROLLER_PARAMETERS, h, buf, bufsize,
			page, scsi3addr, TYPE_CMD)) {
		rc = -1;
		goto out;
	}
2602 2603 2604 2605
	rc = hpsa_scsi_do_simple_cmd_with_retry(h, c,
			PCI_DMA_FROMDEVICE, NO_TIMEOUT);
	if (rc)
		goto out;
2606 2607 2608 2609 2610 2611
	ei = c->err_info;
	if (ei->CommandStatus != 0 && ei->CommandStatus != CMD_DATA_UNDERRUN) {
		hpsa_scsi_interpret_error(h, c);
		rc = -1;
	}
out:
2612
	cmd_free(h, c);
2613
	return rc;
2614
}
2615

2616
static int hpsa_send_reset(struct ctlr_info *h, unsigned char *scsi3addr,
2617
	u8 reset_type, int reply_queue)
2618 2619 2620 2621 2622
{
	int rc = IO_OK;
	struct CommandList *c;
	struct ErrorInfo *ei;

2623
	c = cmd_alloc(h);
2624 2625


2626
	/* fill_cmd can't fail here, no data buffer to map. */
2627 2628 2629
	(void) fill_cmd(c, HPSA_DEVICE_RESET_MSG, h, NULL, 0, 0,
			scsi3addr, TYPE_MSG);
	c->Request.CDB[1] = reset_type; /* fill_cmd defaults to LUN reset */
2630 2631 2632 2633 2634
	rc = hpsa_scsi_do_simple_cmd(h, c, reply_queue, NO_TIMEOUT);
	if (rc) {
		dev_warn(&h->pdev->dev, "Failed to send reset command\n");
		goto out;
	}
2635 2636 2637 2638
	/* no unmap needed here because no data xfer. */

	ei = c->err_info;
	if (ei->CommandStatus != 0) {
2639
		hpsa_scsi_interpret_error(h, c);
2640 2641
		rc = -1;
	}
2642
out:
2643
	cmd_free(h, c);
2644 2645 2646
	return rc;
}

W
Webb Scales 已提交
2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751
static bool hpsa_cmd_dev_match(struct ctlr_info *h, struct CommandList *c,
			       struct hpsa_scsi_dev_t *dev,
			       unsigned char *scsi3addr)
{
	int i;
	bool match = false;
	struct io_accel2_cmd *c2 = &h->ioaccel2_cmd_pool[c->cmdindex];
	struct hpsa_tmf_struct *ac = (struct hpsa_tmf_struct *) c2;

	if (hpsa_is_cmd_idle(c))
		return false;

	switch (c->cmd_type) {
	case CMD_SCSI:
	case CMD_IOCTL_PEND:
		match = !memcmp(scsi3addr, &c->Header.LUN.LunAddrBytes,
				sizeof(c->Header.LUN.LunAddrBytes));
		break;

	case CMD_IOACCEL1:
	case CMD_IOACCEL2:
		if (c->phys_disk == dev) {
			/* HBA mode match */
			match = true;
		} else {
			/* Possible RAID mode -- check each phys dev. */
			/* FIXME:  Do we need to take out a lock here?  If
			 * so, we could just call hpsa_get_pdisk_of_ioaccel2()
			 * instead. */
			for (i = 0; i < dev->nphysical_disks && !match; i++) {
				/* FIXME: an alternate test might be
				 *
				 * match = dev->phys_disk[i]->ioaccel_handle
				 *              == c2->scsi_nexus;      */
				match = dev->phys_disk[i] == c->phys_disk;
			}
		}
		break;

	case IOACCEL2_TMF:
		for (i = 0; i < dev->nphysical_disks && !match; i++) {
			match = dev->phys_disk[i]->ioaccel_handle ==
					le32_to_cpu(ac->it_nexus);
		}
		break;

	case 0:		/* The command is in the middle of being initialized. */
		match = false;
		break;

	default:
		dev_err(&h->pdev->dev, "unexpected cmd_type: %d\n",
			c->cmd_type);
		BUG();
	}

	return match;
}

static int hpsa_do_reset(struct ctlr_info *h, struct hpsa_scsi_dev_t *dev,
	unsigned char *scsi3addr, u8 reset_type, int reply_queue)
{
	int i;
	int rc = 0;

	/* We can really only handle one reset at a time */
	if (mutex_lock_interruptible(&h->reset_mutex) == -EINTR) {
		dev_warn(&h->pdev->dev, "concurrent reset wait interrupted.\n");
		return -EINTR;
	}

	BUG_ON(atomic_read(&dev->reset_cmds_out) != 0);

	for (i = 0; i < h->nr_cmds; i++) {
		struct CommandList *c = h->cmd_pool + i;
		int refcount = atomic_inc_return(&c->refcount);

		if (refcount > 1 && hpsa_cmd_dev_match(h, c, dev, scsi3addr)) {
			unsigned long flags;

			/*
			 * Mark the target command as having a reset pending,
			 * then lock a lock so that the command cannot complete
			 * while we're considering it.  If the command is not
			 * idle then count it; otherwise revoke the event.
			 */
			c->reset_pending = dev;
			spin_lock_irqsave(&h->lock, flags);	/* Implied MB */
			if (!hpsa_is_cmd_idle(c))
				atomic_inc(&dev->reset_cmds_out);
			else
				c->reset_pending = NULL;
			spin_unlock_irqrestore(&h->lock, flags);
		}

		cmd_free(h, c);
	}

	rc = hpsa_send_reset(h, scsi3addr, reset_type, reply_queue);
	if (!rc)
		wait_event(h->event_sync_wait_queue,
			atomic_read(&dev->reset_cmds_out) == 0 ||
			lockup_detected(h));

	if (unlikely(lockup_detected(h))) {
D
Don Brace 已提交
2752 2753 2754 2755
		dev_warn(&h->pdev->dev,
			 "Controller lockup detected during reset wait\n");
		rc = -ENODEV;
	}
W
Webb Scales 已提交
2756 2757 2758 2759 2760 2761 2762 2763

	if (unlikely(rc))
		atomic_set(&dev->reset_cmds_out, 0);

	mutex_unlock(&h->reset_mutex);
	return rc;
}

2764 2765 2766 2767 2768 2769 2770 2771 2772 2773
static void hpsa_get_raid_level(struct ctlr_info *h,
	unsigned char *scsi3addr, unsigned char *raid_level)
{
	int rc;
	unsigned char *buf;

	*raid_level = RAID_UNKNOWN;
	buf = kzalloc(64, GFP_KERNEL);
	if (!buf)
		return;
2774
	rc = hpsa_scsi_do_inquiry(h, scsi3addr, VPD_PAGE | 0xC1, buf, 64);
2775 2776 2777 2778 2779 2780 2781 2782
	if (rc == 0)
		*raid_level = buf[8];
	if (*raid_level > RAID_UNKNOWN)
		*raid_level = RAID_UNKNOWN;
	kfree(buf);
	return;
}

2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794
#define HPSA_MAP_DEBUG
#ifdef HPSA_MAP_DEBUG
static void hpsa_debug_map_buff(struct ctlr_info *h, int rc,
				struct raid_map_data *map_buff)
{
	struct raid_map_disk_data *dd = &map_buff->data[0];
	int map, row, col;
	u16 map_cnt, row_cnt, disks_per_row;

	if (rc != 0)
		return;

2795 2796 2797 2798
	/* Show details only if debugging has been activated. */
	if (h->raid_offload_debug < 2)
		return;

2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822
	dev_info(&h->pdev->dev, "structure_size = %u\n",
				le32_to_cpu(map_buff->structure_size));
	dev_info(&h->pdev->dev, "volume_blk_size = %u\n",
			le32_to_cpu(map_buff->volume_blk_size));
	dev_info(&h->pdev->dev, "volume_blk_cnt = 0x%llx\n",
			le64_to_cpu(map_buff->volume_blk_cnt));
	dev_info(&h->pdev->dev, "physicalBlockShift = %u\n",
			map_buff->phys_blk_shift);
	dev_info(&h->pdev->dev, "parity_rotation_shift = %u\n",
			map_buff->parity_rotation_shift);
	dev_info(&h->pdev->dev, "strip_size = %u\n",
			le16_to_cpu(map_buff->strip_size));
	dev_info(&h->pdev->dev, "disk_starting_blk = 0x%llx\n",
			le64_to_cpu(map_buff->disk_starting_blk));
	dev_info(&h->pdev->dev, "disk_blk_cnt = 0x%llx\n",
			le64_to_cpu(map_buff->disk_blk_cnt));
	dev_info(&h->pdev->dev, "data_disks_per_row = %u\n",
			le16_to_cpu(map_buff->data_disks_per_row));
	dev_info(&h->pdev->dev, "metadata_disks_per_row = %u\n",
			le16_to_cpu(map_buff->metadata_disks_per_row));
	dev_info(&h->pdev->dev, "row_cnt = %u\n",
			le16_to_cpu(map_buff->row_cnt));
	dev_info(&h->pdev->dev, "layout_map_count = %u\n",
			le16_to_cpu(map_buff->layout_map_count));
D
Don Brace 已提交
2823
	dev_info(&h->pdev->dev, "flags = 0x%x\n",
2824
			le16_to_cpu(map_buff->flags));
D
Don Brace 已提交
2825 2826 2827
	dev_info(&h->pdev->dev, "encrypytion = %s\n",
			le16_to_cpu(map_buff->flags) &
			RAID_MAP_FLAG_ENCRYPT_ON ?  "ON" : "OFF");
2828 2829
	dev_info(&h->pdev->dev, "dekindex = %u\n",
			le16_to_cpu(map_buff->dekindex));
2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867
	map_cnt = le16_to_cpu(map_buff->layout_map_count);
	for (map = 0; map < map_cnt; map++) {
		dev_info(&h->pdev->dev, "Map%u:\n", map);
		row_cnt = le16_to_cpu(map_buff->row_cnt);
		for (row = 0; row < row_cnt; row++) {
			dev_info(&h->pdev->dev, "  Row%u:\n", row);
			disks_per_row =
				le16_to_cpu(map_buff->data_disks_per_row);
			for (col = 0; col < disks_per_row; col++, dd++)
				dev_info(&h->pdev->dev,
					"    D%02u: h=0x%04x xor=%u,%u\n",
					col, dd->ioaccel_handle,
					dd->xor_mult[0], dd->xor_mult[1]);
			disks_per_row =
				le16_to_cpu(map_buff->metadata_disks_per_row);
			for (col = 0; col < disks_per_row; col++, dd++)
				dev_info(&h->pdev->dev,
					"    M%02u: h=0x%04x xor=%u,%u\n",
					col, dd->ioaccel_handle,
					dd->xor_mult[0], dd->xor_mult[1]);
		}
	}
}
#else
static void hpsa_debug_map_buff(__attribute__((unused)) struct ctlr_info *h,
			__attribute__((unused)) int rc,
			__attribute__((unused)) struct raid_map_data *map_buff)
{
}
#endif

static int hpsa_get_raid_map(struct ctlr_info *h,
	unsigned char *scsi3addr, struct hpsa_scsi_dev_t *this_device)
{
	int rc = 0;
	struct CommandList *c;
	struct ErrorInfo *ei;

2868
	c = cmd_alloc(h);
2869

2870 2871 2872
	if (fill_cmd(c, HPSA_GET_RAID_MAP, h, &this_device->raid_map,
			sizeof(this_device->raid_map), 0,
			scsi3addr, TYPE_CMD)) {
2873 2874 2875
		dev_warn(&h->pdev->dev, "hpsa_get_raid_map fill_cmd failed\n");
		cmd_free(h, c);
		return -1;
2876
	}
2877 2878 2879 2880
	rc = hpsa_scsi_do_simple_cmd_with_retry(h, c,
					PCI_DMA_FROMDEVICE, NO_TIMEOUT);
	if (rc)
		goto out;
2881 2882
	ei = c->err_info;
	if (ei->CommandStatus != 0 && ei->CommandStatus != CMD_DATA_UNDERRUN) {
2883
		hpsa_scsi_interpret_error(h, c);
2884 2885
		rc = -1;
		goto out;
2886
	}
2887
	cmd_free(h, c);
2888 2889 2890 2891 2892 2893 2894 2895 2896

	/* @todo in the future, dynamically allocate RAID map memory */
	if (le32_to_cpu(this_device->raid_map.structure_size) >
				sizeof(this_device->raid_map)) {
		dev_warn(&h->pdev->dev, "RAID map size is too large!\n");
		rc = -1;
	}
	hpsa_debug_map_buff(h, rc, &this_device->raid_map);
	return rc;
2897 2898 2899
out:
	cmd_free(h, c);
	return rc;
2900 2901
}

2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918
static int hpsa_bmic_id_physical_device(struct ctlr_info *h,
		unsigned char scsi3addr[], u16 bmic_device_index,
		struct bmic_identify_physical_device *buf, size_t bufsize)
{
	int rc = IO_OK;
	struct CommandList *c;
	struct ErrorInfo *ei;

	c = cmd_alloc(h);
	rc = fill_cmd(c, BMIC_IDENTIFY_PHYSICAL_DEVICE, h, buf, bufsize,
		0, RAID_CTLR_LUNID, TYPE_CMD);
	if (rc)
		goto out;

	c->Request.CDB[2] = bmic_device_index & 0xff;
	c->Request.CDB[9] = (bmic_device_index >> 8) & 0xff;

2919 2920
	hpsa_scsi_do_simple_cmd_with_retry(h, c, PCI_DMA_FROMDEVICE,
						NO_TIMEOUT);
2921 2922 2923 2924 2925 2926 2927 2928 2929 2930
	ei = c->err_info;
	if (ei->CommandStatus != 0 && ei->CommandStatus != CMD_DATA_UNDERRUN) {
		hpsa_scsi_interpret_error(h, c);
		rc = -1;
	}
out:
	cmd_free(h, c);
	return rc;
}

2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973
static int hpsa_vpd_page_supported(struct ctlr_info *h,
	unsigned char scsi3addr[], u8 page)
{
	int rc;
	int i;
	int pages;
	unsigned char *buf, bufsize;

	buf = kzalloc(256, GFP_KERNEL);
	if (!buf)
		return 0;

	/* Get the size of the page list first */
	rc = hpsa_scsi_do_inquiry(h, scsi3addr,
				VPD_PAGE | HPSA_VPD_SUPPORTED_PAGES,
				buf, HPSA_VPD_HEADER_SZ);
	if (rc != 0)
		goto exit_unsupported;
	pages = buf[3];
	if ((pages + HPSA_VPD_HEADER_SZ) <= 255)
		bufsize = pages + HPSA_VPD_HEADER_SZ;
	else
		bufsize = 255;

	/* Get the whole VPD page list */
	rc = hpsa_scsi_do_inquiry(h, scsi3addr,
				VPD_PAGE | HPSA_VPD_SUPPORTED_PAGES,
				buf, bufsize);
	if (rc != 0)
		goto exit_unsupported;

	pages = buf[3];
	for (i = 1; i <= pages; i++)
		if (buf[3 + i] == page)
			goto exit_supported;
exit_unsupported:
	kfree(buf);
	return 0;
exit_supported:
	kfree(buf);
	return 1;
}

2974 2975 2976 2977 2978 2979 2980 2981 2982
static void hpsa_get_ioaccel_status(struct ctlr_info *h,
	unsigned char *scsi3addr, struct hpsa_scsi_dev_t *this_device)
{
	int rc;
	unsigned char *buf;
	u8 ioaccel_status;

	this_device->offload_config = 0;
	this_device->offload_enabled = 0;
2983
	this_device->offload_to_be_enabled = 0;
2984 2985 2986 2987

	buf = kzalloc(64, GFP_KERNEL);
	if (!buf)
		return;
2988 2989
	if (!hpsa_vpd_page_supported(h, scsi3addr, HPSA_VPD_LV_IOACCEL_STATUS))
		goto out;
2990
	rc = hpsa_scsi_do_inquiry(h, scsi3addr,
2991
			VPD_PAGE | HPSA_VPD_LV_IOACCEL_STATUS, buf, 64);
2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006
	if (rc != 0)
		goto out;

#define IOACCEL_STATUS_BYTE 4
#define OFFLOAD_CONFIGURED_BIT 0x01
#define OFFLOAD_ENABLED_BIT 0x02
	ioaccel_status = buf[IOACCEL_STATUS_BYTE];
	this_device->offload_config =
		!!(ioaccel_status & OFFLOAD_CONFIGURED_BIT);
	if (this_device->offload_config) {
		this_device->offload_enabled =
			!!(ioaccel_status & OFFLOAD_ENABLED_BIT);
		if (hpsa_get_raid_map(h, scsi3addr, this_device))
			this_device->offload_enabled = 0;
	}
3007
	this_device->offload_to_be_enabled = this_device->offload_enabled;
3008 3009 3010 3011 3012
out:
	kfree(buf);
	return;
}

3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023
/* Get the device id from inquiry page 0x83 */
static int hpsa_get_device_id(struct ctlr_info *h, unsigned char *scsi3addr,
	unsigned char *device_id, int buflen)
{
	int rc;
	unsigned char *buf;

	if (buflen > 16)
		buflen = 16;
	buf = kzalloc(64, GFP_KERNEL);
	if (!buf)
3024
		return -ENOMEM;
3025
	rc = hpsa_scsi_do_inquiry(h, scsi3addr, VPD_PAGE | 0x83, buf, 64);
3026 3027 3028 3029 3030 3031 3032
	if (rc == 0)
		memcpy(device_id, &buf[8], buflen);
	kfree(buf);
	return rc != 0;
}

static int hpsa_scsi_do_report_luns(struct ctlr_info *h, int logical,
3033
		void *buf, int bufsize,
3034 3035 3036 3037 3038 3039 3040
		int extended_response)
{
	int rc = IO_OK;
	struct CommandList *c;
	unsigned char scsi3addr[8];
	struct ErrorInfo *ei;

3041
	c = cmd_alloc(h);
3042

3043 3044
	/* address the controller */
	memset(scsi3addr, 0, sizeof(scsi3addr));
3045 3046 3047 3048 3049
	if (fill_cmd(c, logical ? HPSA_REPORT_LOG : HPSA_REPORT_PHYS, h,
		buf, bufsize, 0, scsi3addr, TYPE_CMD)) {
		rc = -1;
		goto out;
	}
3050 3051
	if (extended_response)
		c->Request.CDB[1] = extended_response;
3052 3053 3054 3055
	rc = hpsa_scsi_do_simple_cmd_with_retry(h, c,
					PCI_DMA_FROMDEVICE, NO_TIMEOUT);
	if (rc)
		goto out;
3056 3057 3058
	ei = c->err_info;
	if (ei->CommandStatus != 0 &&
	    ei->CommandStatus != CMD_DATA_UNDERRUN) {
3059
		hpsa_scsi_interpret_error(h, c);
3060
		rc = -1;
3061
	} else {
3062 3063 3064
		struct ReportLUNdata *rld = buf;

		if (rld->extended_response_flag != extended_response) {
3065 3066 3067
			dev_err(&h->pdev->dev,
				"report luns requested format %u, got %u\n",
				extended_response,
3068
				rld->extended_response_flag);
3069 3070
			rc = -1;
		}
3071
	}
3072
out:
3073
	cmd_free(h, c);
3074 3075 3076 3077
	return rc;
}

static inline int hpsa_scsi_do_report_phys_luns(struct ctlr_info *h,
3078
		struct ReportExtendedLUNdata *buf, int bufsize)
3079
{
3080 3081
	return hpsa_scsi_do_report_luns(h, 0, buf, bufsize,
						HPSA_REPORT_PHYS_EXTENDED);
3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097
}

static inline int hpsa_scsi_do_report_log_luns(struct ctlr_info *h,
		struct ReportLUNdata *buf, int bufsize)
{
	return hpsa_scsi_do_report_luns(h, 1, buf, bufsize, 0);
}

static inline void hpsa_set_bus_target_lun(struct hpsa_scsi_dev_t *device,
	int bus, int target, int lun)
{
	device->bus = bus;
	device->target = target;
	device->lun = lun;
}

3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111
/* Use VPD inquiry to get details of volume status */
static int hpsa_get_volume_status(struct ctlr_info *h,
					unsigned char scsi3addr[])
{
	int rc;
	int status;
	int size;
	unsigned char *buf;

	buf = kzalloc(64, GFP_KERNEL);
	if (!buf)
		return HPSA_VPD_LV_STATUS_UNSUPPORTED;

	/* Does controller have VPD for logical volume status? */
3112
	if (!hpsa_vpd_page_supported(h, scsi3addr, HPSA_VPD_LV_STATUS))
3113 3114 3115 3116 3117
		goto exit_failed;

	/* Get the size of the VPD return buffer */
	rc = hpsa_scsi_do_inquiry(h, scsi3addr, VPD_PAGE | HPSA_VPD_LV_STATUS,
					buf, HPSA_VPD_HEADER_SZ);
3118
	if (rc != 0)
3119 3120 3121 3122 3123 3124
		goto exit_failed;
	size = buf[3];

	/* Now get the whole VPD buffer */
	rc = hpsa_scsi_do_inquiry(h, scsi3addr, VPD_PAGE | HPSA_VPD_LV_STATUS,
					buf, size + HPSA_VPD_HEADER_SZ);
3125
	if (rc != 0)
3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138
		goto exit_failed;
	status = buf[4]; /* status byte */

	kfree(buf);
	return status;
exit_failed:
	kfree(buf);
	return HPSA_VPD_LV_STATUS_UNSUPPORTED;
}

/* Determine offline status of a volume.
 * Return either:
 *  0 (not offline)
3139
 *  0xff (offline for unknown reasons)
3140 3141 3142
 *  # (integer code indicating one of several NOT READY states
 *     describing why a volume is to be kept offline)
 */
3143
static int hpsa_volume_offline(struct ctlr_info *h,
3144 3145 3146
					unsigned char scsi3addr[])
{
	struct CommandList *c;
3147 3148 3149
	unsigned char *sense;
	u8 sense_key, asc, ascq;
	int sense_len;
3150
	int rc, ldstat = 0;
3151 3152 3153 3154 3155 3156 3157
	u16 cmd_status;
	u8 scsi_status;
#define ASC_LUN_NOT_READY 0x04
#define ASCQ_LUN_NOT_READY_FORMAT_IN_PROGRESS 0x04
#define ASCQ_LUN_NOT_READY_INITIALIZING_CMD_REQ 0x02

	c = cmd_alloc(h);
3158

3159
	(void) fill_cmd(c, TEST_UNIT_READY, h, NULL, 0, 0, scsi3addr, TYPE_CMD);
3160 3161 3162 3163 3164
	rc = hpsa_scsi_do_simple_cmd(h, c, DEFAULT_REPLY_QUEUE, NO_TIMEOUT);
	if (rc) {
		cmd_free(h, c);
		return 0;
	}
3165
	sense = c->err_info->SenseInfo;
3166 3167 3168 3169 3170
	if (c->err_info->SenseLen > sizeof(c->err_info->SenseInfo))
		sense_len = sizeof(c->err_info->SenseInfo);
	else
		sense_len = c->err_info->SenseLen;
	decode_sense_data(sense, sense_len, &sense_key, &asc, &ascq);
3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209
	cmd_status = c->err_info->CommandStatus;
	scsi_status = c->err_info->ScsiStatus;
	cmd_free(h, c);
	/* Is the volume 'not ready'? */
	if (cmd_status != CMD_TARGET_STATUS ||
		scsi_status != SAM_STAT_CHECK_CONDITION ||
		sense_key != NOT_READY ||
		asc != ASC_LUN_NOT_READY)  {
		return 0;
	}

	/* Determine the reason for not ready state */
	ldstat = hpsa_get_volume_status(h, scsi3addr);

	/* Keep volume offline in certain cases: */
	switch (ldstat) {
	case HPSA_LV_UNDERGOING_ERASE:
	case HPSA_LV_UNDERGOING_RPI:
	case HPSA_LV_PENDING_RPI:
	case HPSA_LV_ENCRYPTED_NO_KEY:
	case HPSA_LV_PLAINTEXT_IN_ENCRYPT_ONLY_CONTROLLER:
	case HPSA_LV_UNDERGOING_ENCRYPTION:
	case HPSA_LV_UNDERGOING_ENCRYPTION_REKEYING:
	case HPSA_LV_ENCRYPTED_IN_NON_ENCRYPTED_CONTROLLER:
		return ldstat;
	case HPSA_VPD_LV_STATUS_UNSUPPORTED:
		/* If VPD status page isn't available,
		 * use ASC/ASCQ to determine state
		 */
		if ((ascq == ASCQ_LUN_NOT_READY_FORMAT_IN_PROGRESS) ||
			(ascq == ASCQ_LUN_NOT_READY_INITIALIZING_CMD_REQ))
			return ldstat;
		break;
	default:
		break;
	}
	return 0;
}

S
Stephen Cameron 已提交
3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231
/*
 * Find out if a logical device supports aborts by simply trying one.
 * Smart Array may claim not to support aborts on logical drives, but
 * if a MSA2000 * is connected, the drives on that will be presented
 * by the Smart Array as logical drives, and aborts may be sent to
 * those devices successfully.  So the simplest way to find out is
 * to simply try an abort and see how the device responds.
 */
static int hpsa_device_supports_aborts(struct ctlr_info *h,
					unsigned char *scsi3addr)
{
	struct CommandList *c;
	struct ErrorInfo *ei;
	int rc = 0;

	u64 tag = (u64) -1; /* bogus tag */

	/* Assume that physical devices support aborts */
	if (!is_logical_dev_addr_mode(scsi3addr))
		return 1;

	c = cmd_alloc(h);
3232

S
Stephen Cameron 已提交
3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244
	(void) fill_cmd(c, HPSA_ABORT_MSG, h, &tag, 0, 0, scsi3addr, TYPE_MSG);
	(void) hpsa_scsi_do_simple_cmd(h, c, DEFAULT_REPLY_QUEUE, NO_TIMEOUT);
	/* no unmap needed here because no data xfer. */
	ei = c->err_info;
	switch (ei->CommandStatus) {
	case CMD_INVALID:
		rc = 0;
		break;
	case CMD_UNABORTABLE:
	case CMD_ABORT_FAILED:
		rc = 1;
		break;
3245 3246 3247
	case CMD_TMF_STATUS:
		rc = hpsa_evaluate_tmf_status(h, c);
		break;
S
Stephen Cameron 已提交
3248 3249 3250 3251 3252 3253 3254 3255
	default:
		rc = 0;
		break;
	}
	cmd_free(h, c);
	return rc;
}

3256
static int hpsa_update_device_info(struct ctlr_info *h,
3257 3258
	unsigned char scsi3addr[], struct hpsa_scsi_dev_t *this_device,
	unsigned char *is_OBDR_device)
3259
{
3260 3261 3262 3263 3264 3265

#define OBDR_SIG_OFFSET 43
#define OBDR_TAPE_SIG "$DR-10"
#define OBDR_SIG_LEN (sizeof(OBDR_TAPE_SIG) - 1)
#define OBDR_TAPE_INQ_SIZE (OBDR_SIG_OFFSET + OBDR_SIG_LEN)

3266
	unsigned char *inq_buff;
3267
	unsigned char *obdr_sig;
3268

3269
	inq_buff = kzalloc(OBDR_TAPE_INQ_SIZE, GFP_KERNEL);
3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293
	if (!inq_buff)
		goto bail_out;

	/* Do an inquiry to the device to see what it is. */
	if (hpsa_scsi_do_inquiry(h, scsi3addr, 0, inq_buff,
		(unsigned char) OBDR_TAPE_INQ_SIZE) != 0) {
		/* Inquiry failed (msg printed already) */
		dev_err(&h->pdev->dev,
			"hpsa_update_device_info: inquiry failed\n");
		goto bail_out;
	}

	this_device->devtype = (inq_buff[0] & 0x1f);
	memcpy(this_device->scsi3addr, scsi3addr, 8);
	memcpy(this_device->vendor, &inq_buff[8],
		sizeof(this_device->vendor));
	memcpy(this_device->model, &inq_buff[16],
		sizeof(this_device->model));
	memset(this_device->device_id, 0,
		sizeof(this_device->device_id));
	hpsa_get_device_id(h, scsi3addr, this_device->device_id,
		sizeof(this_device->device_id));

	if (this_device->devtype == TYPE_DISK &&
3294
		is_logical_dev_addr_mode(scsi3addr)) {
3295 3296
		int volume_offline;

3297
		hpsa_get_raid_level(h, scsi3addr, &this_device->raid_level);
3298 3299
		if (h->fw_support & MISC_FW_RAID_OFFLOAD_BASIC)
			hpsa_get_ioaccel_status(h, scsi3addr, this_device);
3300 3301 3302 3303
		volume_offline = hpsa_volume_offline(h, scsi3addr);
		if (volume_offline < 0 || volume_offline > 0xff)
			volume_offline = HPSA_VPD_LV_STATUS_UNSUPPORTED;
		this_device->volume_offline = volume_offline & 0xff;
3304
	} else {
3305
		this_device->raid_level = RAID_UNKNOWN;
3306 3307
		this_device->offload_config = 0;
		this_device->offload_enabled = 0;
3308
		this_device->offload_to_be_enabled = 0;
3309
		this_device->hba_ioaccel_enabled = 0;
3310
		this_device->volume_offline = 0;
3311
		this_device->queue_depth = h->nr_cmds;
3312
	}
3313

3314 3315 3316 3317 3318 3319 3320 3321 3322
	if (is_OBDR_device) {
		/* See if this is a One-Button-Disaster-Recovery device
		 * by looking for "$DR-10" at offset 43 in inquiry data.
		 */
		obdr_sig = &inq_buff[OBDR_SIG_OFFSET];
		*is_OBDR_device = (this_device->devtype == TYPE_ROM &&
					strncmp(obdr_sig, OBDR_TAPE_SIG,
						OBDR_SIG_LEN) == 0);
	}
3323 3324 3325 3326 3327 3328 3329 3330
	kfree(inq_buff);
	return 0;

bail_out:
	kfree(inq_buff);
	return 1;
}

S
Stephen Cameron 已提交
3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355
static void hpsa_update_device_supports_aborts(struct ctlr_info *h,
			struct hpsa_scsi_dev_t *dev, u8 *scsi3addr)
{
	unsigned long flags;
	int rc, entry;
	/*
	 * See if this device supports aborts.  If we already know
	 * the device, we already know if it supports aborts, otherwise
	 * we have to find out if it supports aborts by trying one.
	 */
	spin_lock_irqsave(&h->devlock, flags);
	rc = hpsa_scsi_find_entry(dev, h->dev, h->ndevices, &entry);
	if ((rc == DEVICE_SAME || rc == DEVICE_UPDATED) &&
		entry >= 0 && entry < h->ndevices) {
		dev->supports_aborts = h->dev[entry]->supports_aborts;
		spin_unlock_irqrestore(&h->devlock, flags);
	} else {
		spin_unlock_irqrestore(&h->devlock, flags);
		dev->supports_aborts =
				hpsa_device_supports_aborts(h, scsi3addr);
		if (dev->supports_aborts < 0)
			dev->supports_aborts = 0;
	}
}

3356
static unsigned char *ext_target_model[] = {
3357 3358 3359 3360
	"MSA2012",
	"MSA2024",
	"MSA2312",
	"MSA2324",
3361
	"P2000 G3 SAS",
3362
	"MSA 2040 SAS",
3363 3364 3365
	NULL,
};

3366
static int is_ext_target(struct ctlr_info *h, struct hpsa_scsi_dev_t *device)
3367 3368 3369
{
	int i;

3370 3371 3372
	for (i = 0; ext_target_model[i]; i++)
		if (strncmp(device->model, ext_target_model[i],
			strlen(ext_target_model[i])) == 0)
3373 3374 3375 3376 3377
			return 1;
	return 0;
}

/* Helper function to assign bus, target, lun mapping of devices.
3378
 * Puts non-external target logical volumes on bus 0, external target logical
3379 3380 3381 3382 3383 3384
 * volumes on bus 1, physical devices on bus 2. and the hba on bus 3.
 * Logical drive target and lun are assigned at this time, but
 * physical device lun and target assignment are deferred (assigned
 * in hpsa_find_target_lun, called by hpsa_scsi_add_entry.)
 */
static void figure_bus_target_lun(struct ctlr_info *h,
3385
	u8 *lunaddrbytes, struct hpsa_scsi_dev_t *device)
3386
{
3387 3388 3389 3390
	u32 lunid = le32_to_cpu(*((__le32 *) lunaddrbytes));

	if (!is_logical_dev_addr_mode(lunaddrbytes)) {
		/* physical device, target and lun filled in later */
3391
		if (is_hba_lunid(lunaddrbytes))
3392
			hpsa_set_bus_target_lun(device, 3, 0, lunid & 0x3fff);
3393
		else
3394 3395 3396 3397 3398
			/* defer target, lun assignment for physical devices */
			hpsa_set_bus_target_lun(device, 2, -1, -1);
		return;
	}
	/* It's a logical device */
3399 3400
	if (is_ext_target(h, device)) {
		/* external target way, put logicals on bus 1
3401 3402 3403 3404 3405 3406
		 * and match target/lun numbers box
		 * reports, other smart array, bus 0, target 0, match lunid
		 */
		hpsa_set_bus_target_lun(device,
			1, (lunid >> 16) & 0x3fff, lunid & 0x00ff);
		return;
3407
	}
3408
	hpsa_set_bus_target_lun(device, 0, 0, lunid & 0x3fff);
3409 3410 3411 3412
}

/*
 * If there is no lun 0 on a target, linux won't find any devices.
3413
 * For the external targets (arrays), we have to manually detect the enclosure
3414 3415 3416 3417 3418 3419 3420 3421
 * which is at lun zero, as CCISS_REPORT_PHYSICAL_LUNS doesn't report
 * it for some reason.  *tmpdevice is the target we're adding,
 * this_device is a pointer into the current element of currentsd[]
 * that we're building up in update_scsi_devices(), below.
 * lunzerobits is a bitmap that tracks which targets already have a
 * lun 0 assigned.
 * Returns 1 if an enclosure was added, 0 if not.
 */
3422
static int add_ext_target_dev(struct ctlr_info *h,
3423
	struct hpsa_scsi_dev_t *tmpdevice,
3424
	struct hpsa_scsi_dev_t *this_device, u8 *lunaddrbytes,
3425
	unsigned long lunzerobits[], int *n_ext_target_devs)
3426 3427 3428
{
	unsigned char scsi3addr[8];

3429
	if (test_bit(tmpdevice->target, lunzerobits))
3430 3431 3432 3433 3434
		return 0; /* There is already a lun 0 on this target. */

	if (!is_logical_dev_addr_mode(lunaddrbytes))
		return 0; /* It's the logical targets that may lack lun 0. */

3435 3436
	if (!is_ext_target(h, tmpdevice))
		return 0; /* Only external target devices have this problem. */
3437

3438
	if (tmpdevice->lun == 0) /* if lun is 0, then we have a lun 0. */
3439 3440
		return 0;

3441
	memset(scsi3addr, 0, 8);
3442
	scsi3addr[3] = tmpdevice->target;
3443 3444 3445
	if (is_hba_lunid(scsi3addr))
		return 0; /* Don't add the RAID controller here. */

3446 3447 3448
	if (is_scsi_rev_5(h))
		return 0; /* p1210m doesn't need to do this. */

3449
	if (*n_ext_target_devs >= MAX_EXT_TARGETS) {
3450 3451
		dev_warn(&h->pdev->dev, "Maximum number of external "
			"target devices exceeded.  Check your hardware "
3452 3453 3454 3455
			"configuration.");
		return 0;
	}

3456
	if (hpsa_update_device_info(h, scsi3addr, this_device, NULL))
3457
		return 0;
3458
	(*n_ext_target_devs)++;
3459 3460
	hpsa_set_bus_target_lun(this_device,
				tmpdevice->bus, tmpdevice->target, 0);
S
Stephen Cameron 已提交
3461
	hpsa_update_device_supports_aborts(h, this_device, scsi3addr);
3462
	set_bit(tmpdevice->target, lunzerobits);
3463 3464 3465
	return 1;
}

3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476
/*
 * Get address of physical disk used for an ioaccel2 mode command:
 *	1. Extract ioaccel2 handle from the command.
 *	2. Find a matching ioaccel2 handle from list of physical disks.
 *	3. Return:
 *		1 and set scsi3addr to address of matching physical
 *		0 if no matching physical disk was found.
 */
static int hpsa_get_pdisk_of_ioaccel2(struct ctlr_info *h,
	struct CommandList *ioaccel2_cmd_to_abort, unsigned char *scsi3addr)
{
3477 3478 3479
	struct io_accel2_cmd *c2 =
			&h->ioaccel2_cmd_pool[ioaccel2_cmd_to_abort->cmdindex];
	unsigned long flags;
3480 3481
	int i;

3482 3483 3484 3485 3486 3487 3488 3489 3490 3491
	spin_lock_irqsave(&h->devlock, flags);
	for (i = 0; i < h->ndevices; i++)
		if (h->dev[i]->ioaccel_handle == le32_to_cpu(c2->scsi_nexus)) {
			memcpy(scsi3addr, h->dev[i]->scsi3addr,
				sizeof(h->dev[i]->scsi3addr));
			spin_unlock_irqrestore(&h->devlock, flags);
			return 1;
		}
	spin_unlock_irqrestore(&h->devlock, flags);
	return 0;
3492
}
3493

3494 3495 3496 3497 3498 3499 3500
/*
 * Do CISS_REPORT_PHYS and CISS_REPORT_LOG.  Data is returned in physdev,
 * logdev.  The number of luns in physdev and logdev are returned in
 * *nphysicals and *nlogicals, respectively.
 * Returns 0 on success, -1 otherwise.
 */
static int hpsa_gather_lun_info(struct ctlr_info *h,
3501
	struct ReportExtendedLUNdata *physdev, u32 *nphysicals,
3502
	struct ReportLUNdata *logdev, u32 *nlogicals)
3503
{
3504
	if (hpsa_scsi_do_report_phys_luns(h, physdev, sizeof(*physdev))) {
3505 3506 3507
		dev_err(&h->pdev->dev, "report physical LUNs failed.\n");
		return -1;
	}
3508
	*nphysicals = be32_to_cpu(*((__be32 *)physdev->LUNListLength)) / 24;
3509
	if (*nphysicals > HPSA_MAX_PHYS_LUN) {
3510 3511
		dev_warn(&h->pdev->dev, "maximum physical LUNs (%d) exceeded. %d LUNs ignored.\n",
			HPSA_MAX_PHYS_LUN, *nphysicals - HPSA_MAX_PHYS_LUN);
3512 3513
		*nphysicals = HPSA_MAX_PHYS_LUN;
	}
3514
	if (hpsa_scsi_do_report_log_luns(h, logdev, sizeof(*logdev))) {
3515 3516 3517
		dev_err(&h->pdev->dev, "report logical LUNs failed.\n");
		return -1;
	}
3518
	*nlogicals = be32_to_cpu(*((__be32 *) logdev->LUNListLength)) / 8;
3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536
	/* Reject Logicals in excess of our max capability. */
	if (*nlogicals > HPSA_MAX_LUN) {
		dev_warn(&h->pdev->dev,
			"maximum logical LUNs (%d) exceeded.  "
			"%d LUNs ignored.\n", HPSA_MAX_LUN,
			*nlogicals - HPSA_MAX_LUN);
			*nlogicals = HPSA_MAX_LUN;
	}
	if (*nlogicals + *nphysicals > HPSA_MAX_PHYS_LUN) {
		dev_warn(&h->pdev->dev,
			"maximum logical + physical LUNs (%d) exceeded. "
			"%d LUNs ignored.\n", HPSA_MAX_PHYS_LUN,
			*nphysicals + *nlogicals - HPSA_MAX_PHYS_LUN);
		*nlogicals = HPSA_MAX_PHYS_LUN - *nphysicals;
	}
	return 0;
}

D
Don Brace 已提交
3537 3538
static u8 *figure_lunaddrbytes(struct ctlr_info *h, int raid_ctlr_position,
	int i, int nphysicals, int nlogicals,
3539
	struct ReportExtendedLUNdata *physdev_list,
3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553
	struct ReportLUNdata *logdev_list)
{
	/* Helper function, figure out where the LUN ID info is coming from
	 * given index i, lists of physical and logical devices, where in
	 * the list the raid controller is supposed to appear (first or last)
	 */

	int logicals_start = nphysicals + (raid_ctlr_position == 0);
	int last_device = nphysicals + nlogicals + (raid_ctlr_position == 0);

	if (i == raid_ctlr_position)
		return RAID_CTLR_LUNID;

	if (i < logicals_start)
3554 3555
		return &physdev_list->LUN[i -
				(raid_ctlr_position == 0)].lunid[0];
3556 3557 3558 3559 3560 3561 3562 3563

	if (i < last_device)
		return &logdev_list->LUN[i - nphysicals -
			(raid_ctlr_position == 0)][0];
	BUG();
	return NULL;
}

3564 3565 3566
static int hpsa_hba_mode_enabled(struct ctlr_info *h)
{
	int rc;
3567
	int hba_mode_enabled;
3568 3569 3570 3571 3572
	struct bmic_controller_parameters *ctlr_params;
	ctlr_params = kzalloc(sizeof(struct bmic_controller_parameters),
		GFP_KERNEL);

	if (!ctlr_params)
3573
		return -ENOMEM;
3574 3575
	rc = hpsa_bmic_ctrl_mode_sense(h, RAID_CTLR_LUNID, 0, ctlr_params,
		sizeof(struct bmic_controller_parameters));
3576
	if (rc) {
3577
		kfree(ctlr_params);
3578
		return rc;
3579
	}
3580 3581 3582 3583 3584

	hba_mode_enabled =
		((ctlr_params->nvram_flags & HBA_MODE_ENABLED_FLAG) != 0);
	kfree(ctlr_params);
	return hba_mode_enabled;
3585 3586
}

3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597
/* get physical drive ioaccel handle and queue depth */
static void hpsa_get_ioaccel_drive_info(struct ctlr_info *h,
		struct hpsa_scsi_dev_t *dev,
		u8 *lunaddrbytes,
		struct bmic_identify_physical_device *id_phys)
{
	int rc;
	struct ext_report_lun_entry *rle =
		(struct ext_report_lun_entry *) lunaddrbytes;

	dev->ioaccel_handle = rle->ioaccel_handle;
3598 3599
	if (PHYS_IOACCEL(lunaddrbytes) && dev->ioaccel_handle)
		dev->hba_ioaccel_enabled = 1;
3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613
	memset(id_phys, 0, sizeof(*id_phys));
	rc = hpsa_bmic_id_physical_device(h, lunaddrbytes,
			GET_BMIC_DRIVE_NUMBER(lunaddrbytes), id_phys,
			sizeof(*id_phys));
	if (!rc)
		/* Reserve space for FW operations */
#define DRIVE_CMDS_RESERVED_FOR_FW 2
#define DRIVE_QUEUE_DEPTH 7
		dev->queue_depth =
			le16_to_cpu(id_phys->current_queue_depth_limit) -
				DRIVE_CMDS_RESERVED_FOR_FW;
	else
		dev->queue_depth = DRIVE_QUEUE_DEPTH; /* conservative */
	atomic_set(&dev->ioaccel_cmds_out, 0);
W
Webb Scales 已提交
3614
	atomic_set(&dev->reset_cmds_out, 0);
3615 3616
}

3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628
static void hpsa_update_scsi_devices(struct ctlr_info *h, int hostno)
{
	/* the idea here is we could get notified
	 * that some devices have changed, so we do a report
	 * physical luns and report logical luns cmd, and adjust
	 * our list of devices accordingly.
	 *
	 * The scsi3addr's of devices won't change so long as the
	 * adapter is not reset.  That means we can rescan and
	 * tell which devices we already know about, vs. new
	 * devices, vs.  disappearing devices.
	 */
3629
	struct ReportExtendedLUNdata *physdev_list = NULL;
3630
	struct ReportLUNdata *logdev_list = NULL;
3631
	struct bmic_identify_physical_device *id_phys = NULL;
3632 3633 3634
	u32 nphysicals = 0;
	u32 nlogicals = 0;
	u32 ndev_allocated = 0;
3635 3636
	struct hpsa_scsi_dev_t **currentsd, *this_device, *tmpdevice;
	int ncurrent = 0;
3637
	int i, n_ext_target_devs, ndevs_to_allocate;
3638
	int raid_ctlr_position;
3639
	int rescan_hba_mode;
3640
	DECLARE_BITMAP(lunzerobits, MAX_EXT_TARGETS);
3641

3642
	currentsd = kzalloc(sizeof(*currentsd) * HPSA_MAX_DEVICES, GFP_KERNEL);
3643 3644
	physdev_list = kzalloc(sizeof(*physdev_list), GFP_KERNEL);
	logdev_list = kzalloc(sizeof(*logdev_list), GFP_KERNEL);
3645
	tmpdevice = kzalloc(sizeof(*tmpdevice), GFP_KERNEL);
3646
	id_phys = kzalloc(sizeof(*id_phys), GFP_KERNEL);
3647

3648 3649
	if (!currentsd || !physdev_list || !logdev_list ||
		!tmpdevice || !id_phys) {
3650 3651 3652 3653 3654
		dev_err(&h->pdev->dev, "out of memory\n");
		goto out;
	}
	memset(lunzerobits, 0, sizeof(lunzerobits));

3655
	rescan_hba_mode = hpsa_hba_mode_enabled(h);
3656 3657
	if (rescan_hba_mode < 0)
		goto out;
3658 3659 3660 3661 3662 3663 3664 3665

	if (!h->hba_mode_enabled && rescan_hba_mode)
		dev_warn(&h->pdev->dev, "HBA mode enabled\n");
	else if (h->hba_mode_enabled && !rescan_hba_mode)
		dev_warn(&h->pdev->dev, "HBA mode disabled\n");

	h->hba_mode_enabled = rescan_hba_mode;

3666 3667
	if (hpsa_gather_lun_info(h, physdev_list, &nphysicals,
			logdev_list, &nlogicals))
3668 3669
		goto out;

3670 3671 3672
	/* We might see up to the maximum number of logical and physical disks
	 * plus external target devices, and a device for the local RAID
	 * controller.
3673
	 */
3674
	ndevs_to_allocate = nphysicals + nlogicals + MAX_EXT_TARGETS + 1;
3675 3676 3677

	/* Allocate the per device structures */
	for (i = 0; i < ndevs_to_allocate; i++) {
3678 3679 3680 3681 3682 3683 3684
		if (i >= HPSA_MAX_DEVICES) {
			dev_warn(&h->pdev->dev, "maximum devices (%d) exceeded."
				"  %d devices ignored.\n", HPSA_MAX_DEVICES,
				ndevs_to_allocate - HPSA_MAX_DEVICES);
			break;
		}

3685 3686 3687 3688 3689 3690 3691 3692 3693
		currentsd[i] = kzalloc(sizeof(*currentsd[i]), GFP_KERNEL);
		if (!currentsd[i]) {
			dev_warn(&h->pdev->dev, "out of memory at %s:%d\n",
				__FILE__, __LINE__);
			goto out;
		}
		ndev_allocated++;
	}

3694
	if (is_scsi_rev_5(h))
3695 3696 3697 3698
		raid_ctlr_position = 0;
	else
		raid_ctlr_position = nphysicals + nlogicals;

3699
	/* adjust our table of devices */
3700
	n_ext_target_devs = 0;
3701
	for (i = 0; i < nphysicals + nlogicals + 1; i++) {
3702
		u8 *lunaddrbytes, is_OBDR = 0;
3703 3704

		/* Figure out where the LUN ID info is coming from */
3705 3706
		lunaddrbytes = figure_lunaddrbytes(h, raid_ctlr_position,
			i, nphysicals, nlogicals, physdev_list, logdev_list);
3707 3708 3709 3710 3711 3712

		/* skip masked non-disk devices */
		if (MASKED_DEVICE(lunaddrbytes))
			if (i < nphysicals + (raid_ctlr_position == 0) &&
				NON_DISK_PHYS_DEV(lunaddrbytes))
				continue;
3713 3714

		/* Get device type, vendor, model, device id */
3715 3716
		if (hpsa_update_device_info(h, lunaddrbytes, tmpdevice,
							&is_OBDR))
3717
			continue; /* skip it if we can't talk to it. */
3718
		figure_bus_target_lun(h, lunaddrbytes, tmpdevice);
S
Stephen Cameron 已提交
3719
		hpsa_update_device_supports_aborts(h, tmpdevice, lunaddrbytes);
3720 3721 3722
		this_device = currentsd[ncurrent];

		/*
3723
		 * For external target devices, we have to insert a LUN 0 which
3724 3725 3726 3727 3728
		 * doesn't show up in CCISS_REPORT_PHYSICAL data, but there
		 * is nonetheless an enclosure device there.  We have to
		 * present that otherwise linux won't find anything if
		 * there is no lun 0.
		 */
3729
		if (add_ext_target_dev(h, tmpdevice, this_device,
3730
				lunaddrbytes, lunzerobits,
3731
				&n_ext_target_devs)) {
3732 3733 3734 3735 3736 3737
			ncurrent++;
			this_device = currentsd[ncurrent];
		}

		*this_device = *tmpdevice;

3738 3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749
		/* do not expose masked devices */
		if (MASKED_DEVICE(lunaddrbytes) &&
			i < nphysicals + (raid_ctlr_position == 0)) {
			if (h->hba_mode_enabled)
				dev_warn(&h->pdev->dev,
					"Masked physical device detected\n");
			this_device->expose_state = HPSA_DO_NOT_EXPOSE;
		} else {
			this_device->expose_state =
					HPSA_SG_ATTACH | HPSA_ULD_ATTACH;
		}

3750
		switch (this_device->devtype) {
3751
		case TYPE_ROM:
3752 3753 3754 3755 3756 3757 3758
			/* We don't *really* support actual CD-ROM devices,
			 * just "One Button Disaster Recovery" tape drive
			 * which temporarily pretends to be a CD-ROM drive.
			 * So we check that the device is really an OBDR tape
			 * device by checking for "$DR-10" in bytes 43-48 of
			 * the inquiry data.
			 */
3759 3760
			if (is_OBDR)
				ncurrent++;
3761 3762
			break;
		case TYPE_DISK:
3763
			if (i >= nphysicals) {
3764 3765
				ncurrent++;
				break;
3766
			}
3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778

			if (h->hba_mode_enabled)
				/* never use raid mapper in HBA mode */
				this_device->offload_enabled = 0;
			else if (!(h->transMethod & CFGTBL_Trans_io_accel1 ||
				h->transMethod & CFGTBL_Trans_io_accel2))
				break;

			hpsa_get_ioaccel_drive_info(h, this_device,
						lunaddrbytes, id_phys);
			atomic_set(&this_device->ioaccel_cmds_out, 0);
			ncurrent++;
3779 3780 3781 3782 3783
			break;
		case TYPE_TAPE:
		case TYPE_MEDIUM_CHANGER:
			ncurrent++;
			break;
3784 3785 3786 3787
		case TYPE_ENCLOSURE:
			if (h->hba_mode_enabled)
				ncurrent++;
			break;
3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799 3800
		case TYPE_RAID:
			/* Only present the Smartarray HBA as a RAID controller.
			 * If it's a RAID controller other than the HBA itself
			 * (an external RAID controller, MSA500 or similar)
			 * don't present it.
			 */
			if (!is_hba_lunid(lunaddrbytes))
				break;
			ncurrent++;
			break;
		default:
			break;
		}
3801
		if (ncurrent >= HPSA_MAX_DEVICES)
3802 3803 3804 3805 3806 3807 3808 3809 3810 3811
			break;
	}
	adjust_hpsa_scsi_table(h, hostno, currentsd, ncurrent);
out:
	kfree(tmpdevice);
	for (i = 0; i < ndev_allocated; i++)
		kfree(currentsd[i]);
	kfree(currentsd);
	kfree(physdev_list);
	kfree(logdev_list);
3812
	kfree(id_phys);
3813 3814
}

3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825
static void hpsa_set_sg_descriptor(struct SGDescriptor *desc,
				   struct scatterlist *sg)
{
	u64 addr64 = (u64) sg_dma_address(sg);
	unsigned int len = sg_dma_len(sg);

	desc->Addr = cpu_to_le64(addr64);
	desc->Len = cpu_to_le32(len);
	desc->Ext = 0;
}

3826 3827
/*
 * hpsa_scatter_gather takes a struct scsi_cmnd, (cmd), and does the pci
3828 3829 3830
 * dma mapping  and fills in the scatter gather entries of the
 * hpsa command, cp.
 */
3831
static int hpsa_scatter_gather(struct ctlr_info *h,
3832 3833 3834 3835
		struct CommandList *cp,
		struct scsi_cmnd *cmd)
{
	struct scatterlist *sg;
3836
	int use_sg, i, sg_limit, chained, last_sg;
3837
	struct SGDescriptor *curr_sg;
3838

3839
	BUG_ON(scsi_sg_count(cmd) > h->maxsgentries);
3840 3841 3842 3843 3844 3845 3846 3847

	use_sg = scsi_dma_map(cmd);
	if (use_sg < 0)
		return use_sg;

	if (!use_sg)
		goto sglist_finished;

3848 3849 3850 3851 3852 3853 3854
	/*
	 * If the number of entries is greater than the max for a single list,
	 * then we have a chained list; we will set up all but one entry in the
	 * first list (the last entry is saved for link information);
	 * otherwise, we don't have a chained list and we'll set up at each of
	 * the entries in the one list.
	 */
3855
	curr_sg = cp->SG;
3856 3857 3858 3859
	chained = use_sg > h->max_cmd_sg_entries;
	sg_limit = chained ? h->max_cmd_sg_entries - 1 : use_sg;
	last_sg = scsi_sg_count(cmd) - 1;
	scsi_for_each_sg(cmd, sg, sg_limit, i) {
3860
		hpsa_set_sg_descriptor(curr_sg, sg);
3861 3862
		curr_sg++;
	}
3863

3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878
	if (chained) {
		/*
		 * Continue with the chained list.  Set curr_sg to the chained
		 * list.  Modify the limit to the total count less the entries
		 * we've already set up.  Resume the scan at the list entry
		 * where the previous loop left off.
		 */
		curr_sg = h->cmd_sg_list[cp->cmdindex];
		sg_limit = use_sg - sg_limit;
		for_each_sg(sg, sg, sg_limit, i) {
			hpsa_set_sg_descriptor(curr_sg, sg);
			curr_sg++;
		}
	}

3879
	/* Back the pointer up to the last entry and mark it as "last". */
3880
	(curr_sg - 1)->Ext = cpu_to_le32(HPSA_SG_LAST);
3881 3882 3883 3884 3885 3886

	if (use_sg + chained > h->maxSG)
		h->maxSG = use_sg + chained;

	if (chained) {
		cp->Header.SGList = h->max_cmd_sg_entries;
3887
		cp->Header.SGTotal = cpu_to_le16(use_sg + 1);
3888 3889 3890 3891
		if (hpsa_map_sg_chain_block(h, cp)) {
			scsi_dma_unmap(cmd);
			return -1;
		}
3892
		return 0;
3893 3894 3895 3896
	}

sglist_finished:

3897
	cp->Header.SGList = (u8) use_sg;   /* no. SGs contig in this cmd */
3898
	cp->Header.SGTotal = cpu_to_le16(use_sg); /* total sgs in cmd list */
3899 3900 3901
	return 0;
}

3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949
#define IO_ACCEL_INELIGIBLE (1)
static int fixup_ioaccel_cdb(u8 *cdb, int *cdb_len)
{
	int is_write = 0;
	u32 block;
	u32 block_cnt;

	/* Perform some CDB fixups if needed using 10 byte reads/writes only */
	switch (cdb[0]) {
	case WRITE_6:
	case WRITE_12:
		is_write = 1;
	case READ_6:
	case READ_12:
		if (*cdb_len == 6) {
			block = (((u32) cdb[2]) << 8) | cdb[3];
			block_cnt = cdb[4];
		} else {
			BUG_ON(*cdb_len != 12);
			block = (((u32) cdb[2]) << 24) |
				(((u32) cdb[3]) << 16) |
				(((u32) cdb[4]) << 8) |
				cdb[5];
			block_cnt =
				(((u32) cdb[6]) << 24) |
				(((u32) cdb[7]) << 16) |
				(((u32) cdb[8]) << 8) |
				cdb[9];
		}
		if (block_cnt > 0xffff)
			return IO_ACCEL_INELIGIBLE;

		cdb[0] = is_write ? WRITE_10 : READ_10;
		cdb[1] = 0;
		cdb[2] = (u8) (block >> 24);
		cdb[3] = (u8) (block >> 16);
		cdb[4] = (u8) (block >> 8);
		cdb[5] = (u8) (block);
		cdb[6] = 0;
		cdb[7] = (u8) (block_cnt >> 8);
		cdb[8] = (u8) (block_cnt);
		cdb[9] = 0;
		*cdb_len = 10;
		break;
	}
	return 0;
}

3950
static int hpsa_scsi_ioaccel1_queue_command(struct ctlr_info *h,
3951
	struct CommandList *c, u32 ioaccel_handle, u8 *cdb, int cdb_len,
3952
	u8 *scsi3addr, struct hpsa_scsi_dev_t *phys_disk)
3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963
{
	struct scsi_cmnd *cmd = c->scsi_cmd;
	struct io_accel1_cmd *cp = &h->ioaccel_cmd_pool[c->cmdindex];
	unsigned int len;
	unsigned int total_len = 0;
	struct scatterlist *sg;
	u64 addr64;
	int use_sg, i;
	struct SGDescriptor *curr_sg;
	u32 control = IOACCEL1_CONTROL_SIMPLEQUEUE;

3964
	/* TODO: implement chaining support */
3965 3966
	if (scsi_sg_count(cmd) > h->ioaccel_maxsg) {
		atomic_dec(&phys_disk->ioaccel_cmds_out);
3967
		return IO_ACCEL_INELIGIBLE;
3968
	}
3969

3970 3971
	BUG_ON(cmd->cmd_len > IOACCEL1_IOFLAGS_CDBLEN_MAX);

3972 3973
	if (fixup_ioaccel_cdb(cdb, &cdb_len)) {
		atomic_dec(&phys_disk->ioaccel_cmds_out);
3974
		return IO_ACCEL_INELIGIBLE;
3975
	}
3976

3977 3978 3979 3980 3981 3982 3983 3984
	c->cmd_type = CMD_IOACCEL1;

	/* Adjust the DMA address to point to the accelerated command buffer */
	c->busaddr = (u32) h->ioaccel_cmd_pool_dhandle +
				(c->cmdindex * sizeof(*cp));
	BUG_ON(c->busaddr & 0x0000007F);

	use_sg = scsi_dma_map(cmd);
3985 3986
	if (use_sg < 0) {
		atomic_dec(&phys_disk->ioaccel_cmds_out);
3987
		return use_sg;
3988
	}
3989 3990 3991 3992 3993 3994 3995

	if (use_sg) {
		curr_sg = cp->SG;
		scsi_for_each_sg(cmd, sg, use_sg, i) {
			addr64 = (u64) sg_dma_address(sg);
			len  = sg_dma_len(sg);
			total_len += len;
3996 3997 3998
			curr_sg->Addr = cpu_to_le64(addr64);
			curr_sg->Len = cpu_to_le32(len);
			curr_sg->Ext = cpu_to_le32(0);
3999 4000
			curr_sg++;
		}
4001
		(--curr_sg)->Ext = cpu_to_le32(HPSA_SG_LAST);
4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022

		switch (cmd->sc_data_direction) {
		case DMA_TO_DEVICE:
			control |= IOACCEL1_CONTROL_DATA_OUT;
			break;
		case DMA_FROM_DEVICE:
			control |= IOACCEL1_CONTROL_DATA_IN;
			break;
		case DMA_NONE:
			control |= IOACCEL1_CONTROL_NODATAXFER;
			break;
		default:
			dev_err(&h->pdev->dev, "unknown data direction: %d\n",
			cmd->sc_data_direction);
			BUG();
			break;
		}
	} else {
		control |= IOACCEL1_CONTROL_NODATAXFER;
	}

4023
	c->Header.SGList = use_sg;
4024
	/* Fill out the command structure to submit */
D
Don Brace 已提交
4025 4026 4027 4028 4029
	cp->dev_handle = cpu_to_le16(ioaccel_handle & 0xFFFF);
	cp->transfer_len = cpu_to_le32(total_len);
	cp->io_flags = cpu_to_le16(IOACCEL1_IOFLAGS_IO_REQ |
			(cdb_len & IOACCEL1_IOFLAGS_CDBLEN_MASK));
	cp->control = cpu_to_le32(control);
4030 4031
	memcpy(cp->CDB, cdb, cdb_len);
	memcpy(cp->CISS_LUN, scsi3addr, 8);
4032
	/* Tag was already set at init time. */
4033
	enqueue_cmd_and_start_io(h, c);
4034 4035
	return 0;
}
4036

4037 4038 4039 4040 4041 4042 4043 4044 4045 4046
/*
 * Queue a command directly to a device behind the controller using the
 * I/O accelerator path.
 */
static int hpsa_scsi_ioaccel_direct_map(struct ctlr_info *h,
	struct CommandList *c)
{
	struct scsi_cmnd *cmd = c->scsi_cmd;
	struct hpsa_scsi_dev_t *dev = cmd->device->hostdata;

4047 4048
	c->phys_disk = dev;

4049
	return hpsa_scsi_ioaccel_queue_command(h, c, dev->ioaccel_handle,
4050
		cmd->cmnd, cmd->cmd_len, dev->scsi3addr, dev);
4051 4052
}

4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064
/*
 * Set encryption parameters for the ioaccel2 request
 */
static void set_encrypt_ioaccel2(struct ctlr_info *h,
	struct CommandList *c, struct io_accel2_cmd *cp)
{
	struct scsi_cmnd *cmd = c->scsi_cmd;
	struct hpsa_scsi_dev_t *dev = cmd->device->hostdata;
	struct raid_map_data *map = &dev->raid_map;
	u64 first_block;

	/* Are we doing encryption on this device */
D
Don Brace 已提交
4065
	if (!(le16_to_cpu(map->flags) & RAID_MAP_FLAG_ENCRYPT_ON))
4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080
		return;
	/* Set the data encryption key index. */
	cp->dekindex = map->dekindex;

	/* Set the encryption enable flag, encoded into direction field. */
	cp->direction |= IOACCEL2_DIRECTION_ENCRYPT_MASK;

	/* Set encryption tweak values based on logical block address
	 * If block size is 512, tweak value is LBA.
	 * For other block sizes, tweak is (LBA * block size)/ 512)
	 */
	switch (cmd->cmnd[0]) {
	/* Required? 6-byte cdbs eliminated by fixup_ioaccel_cdb */
	case WRITE_6:
	case READ_6:
D
Don Brace 已提交
4081
		first_block = get_unaligned_be16(&cmd->cmnd[2]);
4082 4083 4084 4085 4086 4087
		break;
	case WRITE_10:
	case READ_10:
	/* Required? 12-byte cdbs eliminated by fixup_ioaccel_cdb */
	case WRITE_12:
	case READ_12:
D
Don Brace 已提交
4088
		first_block = get_unaligned_be32(&cmd->cmnd[2]);
4089 4090 4091
		break;
	case WRITE_16:
	case READ_16:
D
Don Brace 已提交
4092
		first_block = get_unaligned_be64(&cmd->cmnd[2]);
4093 4094 4095
		break;
	default:
		dev_err(&h->pdev->dev,
D
Don Brace 已提交
4096 4097
			"ERROR: %s: size (0x%x) not supported for encryption\n",
			__func__, cmd->cmnd[0]);
4098 4099 4100
		BUG();
		break;
	}
D
Don Brace 已提交
4101 4102 4103 4104 4105 4106 4107

	if (le32_to_cpu(map->volume_blk_size) != 512)
		first_block = first_block *
				le32_to_cpu(map->volume_blk_size)/512;

	cp->tweak_lower = cpu_to_le32(first_block);
	cp->tweak_upper = cpu_to_le32(first_block >> 32);
4108 4109
}

4110 4111
static int hpsa_scsi_ioaccel2_queue_command(struct ctlr_info *h,
	struct CommandList *c, u32 ioaccel_handle, u8 *cdb, int cdb_len,
4112
	u8 *scsi3addr, struct hpsa_scsi_dev_t *phys_disk)
4113 4114 4115 4116 4117 4118 4119 4120 4121 4122
{
	struct scsi_cmnd *cmd = c->scsi_cmd;
	struct io_accel2_cmd *cp = &h->ioaccel2_cmd_pool[c->cmdindex];
	struct ioaccel2_sg_element *curr_sg;
	int use_sg, i;
	struct scatterlist *sg;
	u64 addr64;
	u32 len;
	u32 total_len = 0;

4123
	BUG_ON(scsi_sg_count(cmd) > h->maxsgentries);
4124

4125 4126
	if (fixup_ioaccel_cdb(cdb, &cdb_len)) {
		atomic_dec(&phys_disk->ioaccel_cmds_out);
4127
		return IO_ACCEL_INELIGIBLE;
4128 4129
	}

4130 4131 4132 4133 4134 4135 4136 4137 4138 4139
	c->cmd_type = CMD_IOACCEL2;
	/* Adjust the DMA address to point to the accelerated command buffer */
	c->busaddr = (u32) h->ioaccel2_cmd_pool_dhandle +
				(c->cmdindex * sizeof(*cp));
	BUG_ON(c->busaddr & 0x0000007F);

	memset(cp, 0, sizeof(*cp));
	cp->IU_type = IOACCEL2_IU_TYPE;

	use_sg = scsi_dma_map(cmd);
4140 4141
	if (use_sg < 0) {
		atomic_dec(&phys_disk->ioaccel_cmds_out);
4142
		return use_sg;
4143
	}
4144 4145 4146

	if (use_sg) {
		curr_sg = cp->sg;
4147 4148 4149 4150 4151 4152 4153 4154 4155 4156 4157 4158
		if (use_sg > h->ioaccel_maxsg) {
			addr64 = le64_to_cpu(
				h->ioaccel2_cmd_sg_list[c->cmdindex]->address);
			curr_sg->address = cpu_to_le64(addr64);
			curr_sg->length = 0;
			curr_sg->reserved[0] = 0;
			curr_sg->reserved[1] = 0;
			curr_sg->reserved[2] = 0;
			curr_sg->chain_indicator = 0x80;

			curr_sg = h->ioaccel2_cmd_sg_list[c->cmdindex];
		}
4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173
		scsi_for_each_sg(cmd, sg, use_sg, i) {
			addr64 = (u64) sg_dma_address(sg);
			len  = sg_dma_len(sg);
			total_len += len;
			curr_sg->address = cpu_to_le64(addr64);
			curr_sg->length = cpu_to_le32(len);
			curr_sg->reserved[0] = 0;
			curr_sg->reserved[1] = 0;
			curr_sg->reserved[2] = 0;
			curr_sg->chain_indicator = 0;
			curr_sg++;
		}

		switch (cmd->sc_data_direction) {
		case DMA_TO_DEVICE:
4174 4175
			cp->direction &= ~IOACCEL2_DIRECTION_MASK;
			cp->direction |= IOACCEL2_DIR_DATA_OUT;
4176 4177
			break;
		case DMA_FROM_DEVICE:
4178 4179
			cp->direction &= ~IOACCEL2_DIRECTION_MASK;
			cp->direction |= IOACCEL2_DIR_DATA_IN;
4180 4181
			break;
		case DMA_NONE:
4182 4183
			cp->direction &= ~IOACCEL2_DIRECTION_MASK;
			cp->direction |= IOACCEL2_DIR_NO_DATA;
4184 4185 4186 4187 4188 4189 4190 4191
			break;
		default:
			dev_err(&h->pdev->dev, "unknown data direction: %d\n",
				cmd->sc_data_direction);
			BUG();
			break;
		}
	} else {
4192 4193
		cp->direction &= ~IOACCEL2_DIRECTION_MASK;
		cp->direction |= IOACCEL2_DIR_NO_DATA;
4194
	}
4195 4196 4197 4198

	/* Set encryption parameters, if necessary */
	set_encrypt_ioaccel2(h, c, cp);

D
Don Brace 已提交
4199
	cp->scsi_nexus = cpu_to_le32(ioaccel_handle);
4200
	cp->Tag = cpu_to_le32(c->cmdindex << DIRECT_LOOKUP_SHIFT);
4201 4202 4203 4204 4205
	memcpy(cp->cdb, cdb, sizeof(cp->cdb));

	cp->data_len = cpu_to_le32(total_len);
	cp->err_ptr = cpu_to_le64(c->busaddr +
			offsetof(struct io_accel2_cmd, error_data));
4206
	cp->err_len = cpu_to_le32(sizeof(cp->error_data));
4207

4208 4209 4210 4211 4212 4213 4214 4215 4216 4217 4218
	/* fill in sg elements */
	if (use_sg > h->ioaccel_maxsg) {
		cp->sg_count = 1;
		if (hpsa_map_ioaccel2_sg_chain_block(h, cp, c)) {
			atomic_dec(&phys_disk->ioaccel_cmds_out);
			scsi_dma_unmap(cmd);
			return -1;
		}
	} else
		cp->sg_count = (u8) use_sg;

4219 4220 4221 4222 4223 4224 4225 4226 4227
	enqueue_cmd_and_start_io(h, c);
	return 0;
}

/*
 * Queue a command to the correct I/O accelerator path.
 */
static int hpsa_scsi_ioaccel_queue_command(struct ctlr_info *h,
	struct CommandList *c, u32 ioaccel_handle, u8 *cdb, int cdb_len,
4228
	u8 *scsi3addr, struct hpsa_scsi_dev_t *phys_disk)
4229
{
4230 4231 4232 4233 4234 4235
	/* Try to honor the device's queue depth */
	if (atomic_inc_return(&phys_disk->ioaccel_cmds_out) >
					phys_disk->queue_depth) {
		atomic_dec(&phys_disk->ioaccel_cmds_out);
		return IO_ACCEL_INELIGIBLE;
	}
4236 4237
	if (h->transMethod & CFGTBL_Trans_io_accel1)
		return hpsa_scsi_ioaccel1_queue_command(h, c, ioaccel_handle,
4238 4239
						cdb, cdb_len, scsi3addr,
						phys_disk);
4240 4241
	else
		return hpsa_scsi_ioaccel2_queue_command(h, c, ioaccel_handle,
4242 4243
						cdb, cdb_len, scsi3addr,
						phys_disk);
4244 4245
}

4246 4247 4248 4249 4250
static void raid_map_helper(struct raid_map_data *map,
		int offload_to_mirror, u32 *map_index, u32 *current_group)
{
	if (offload_to_mirror == 0)  {
		/* use physical disk in the first mirrored group. */
D
Don Brace 已提交
4251
		*map_index %= le16_to_cpu(map->data_disks_per_row);
4252 4253 4254 4255
		return;
	}
	do {
		/* determine mirror group that *map_index indicates */
D
Don Brace 已提交
4256 4257
		*current_group = *map_index /
			le16_to_cpu(map->data_disks_per_row);
4258 4259
		if (offload_to_mirror == *current_group)
			continue;
D
Don Brace 已提交
4260
		if (*current_group < le16_to_cpu(map->layout_map_count) - 1) {
4261
			/* select map index from next group */
D
Don Brace 已提交
4262
			*map_index += le16_to_cpu(map->data_disks_per_row);
4263 4264 4265
			(*current_group)++;
		} else {
			/* select map index from first group */
D
Don Brace 已提交
4266
			*map_index %= le16_to_cpu(map->data_disks_per_row);
4267 4268 4269 4270 4271
			*current_group = 0;
		}
	} while (offload_to_mirror != *current_group);
}

4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289
/*
 * Attempt to perform offload RAID mapping for a logical volume I/O.
 */
static int hpsa_scsi_ioaccel_raid_map(struct ctlr_info *h,
	struct CommandList *c)
{
	struct scsi_cmnd *cmd = c->scsi_cmd;
	struct hpsa_scsi_dev_t *dev = cmd->device->hostdata;
	struct raid_map_data *map = &dev->raid_map;
	struct raid_map_disk_data *dd = &map->data[0];
	int is_write = 0;
	u32 map_index;
	u64 first_block, last_block;
	u32 block_cnt;
	u32 blocks_per_row;
	u64 first_row, last_row;
	u32 first_row_offset, last_row_offset;
	u32 first_column, last_column;
4290 4291 4292 4293 4294 4295 4296 4297
	u64 r0_first_row, r0_last_row;
	u32 r5or6_blocks_per_row;
	u64 r5or6_first_row, r5or6_last_row;
	u32 r5or6_first_row_offset, r5or6_last_row_offset;
	u32 r5or6_first_column, r5or6_last_column;
	u32 total_disks_per_row;
	u32 stripesize;
	u32 first_group, last_group, current_group;
4298 4299 4300 4301 4302 4303
	u32 map_row;
	u32 disk_handle;
	u64 disk_block;
	u32 disk_block_cnt;
	u8 cdb[16];
	u8 cdb_len;
D
Don Brace 已提交
4304
	u16 strip_size;
4305 4306 4307
#if BITS_PER_LONG == 32
	u64 tmpdiv;
#endif
4308
	int offload_to_mirror;
4309 4310 4311 4312 4313 4314 4315 4316 4317 4318

	/* check for valid opcode, get LBA and block count */
	switch (cmd->cmnd[0]) {
	case WRITE_6:
		is_write = 1;
	case READ_6:
		first_block =
			(((u64) cmd->cmnd[2]) << 8) |
			cmd->cmnd[3];
		block_cnt = cmd->cmnd[4];
4319 4320
		if (block_cnt == 0)
			block_cnt = 256;
4321 4322 4323 4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352 4353 4354 4355 4356 4357 4358 4359 4360 4361 4362 4363 4364 4365 4366 4367 4368 4369 4370 4371 4372 4373 4374 4375
		break;
	case WRITE_10:
		is_write = 1;
	case READ_10:
		first_block =
			(((u64) cmd->cmnd[2]) << 24) |
			(((u64) cmd->cmnd[3]) << 16) |
			(((u64) cmd->cmnd[4]) << 8) |
			cmd->cmnd[5];
		block_cnt =
			(((u32) cmd->cmnd[7]) << 8) |
			cmd->cmnd[8];
		break;
	case WRITE_12:
		is_write = 1;
	case READ_12:
		first_block =
			(((u64) cmd->cmnd[2]) << 24) |
			(((u64) cmd->cmnd[3]) << 16) |
			(((u64) cmd->cmnd[4]) << 8) |
			cmd->cmnd[5];
		block_cnt =
			(((u32) cmd->cmnd[6]) << 24) |
			(((u32) cmd->cmnd[7]) << 16) |
			(((u32) cmd->cmnd[8]) << 8) |
		cmd->cmnd[9];
		break;
	case WRITE_16:
		is_write = 1;
	case READ_16:
		first_block =
			(((u64) cmd->cmnd[2]) << 56) |
			(((u64) cmd->cmnd[3]) << 48) |
			(((u64) cmd->cmnd[4]) << 40) |
			(((u64) cmd->cmnd[5]) << 32) |
			(((u64) cmd->cmnd[6]) << 24) |
			(((u64) cmd->cmnd[7]) << 16) |
			(((u64) cmd->cmnd[8]) << 8) |
			cmd->cmnd[9];
		block_cnt =
			(((u32) cmd->cmnd[10]) << 24) |
			(((u32) cmd->cmnd[11]) << 16) |
			(((u32) cmd->cmnd[12]) << 8) |
			cmd->cmnd[13];
		break;
	default:
		return IO_ACCEL_INELIGIBLE; /* process via normal I/O path */
	}
	last_block = first_block + block_cnt - 1;

	/* check for write to non-RAID-0 */
	if (is_write && dev->raid_level != 0)
		return IO_ACCEL_INELIGIBLE;

	/* check for invalid block or wraparound */
D
Don Brace 已提交
4376 4377
	if (last_block >= le64_to_cpu(map->volume_blk_cnt) ||
		last_block < first_block)
4378 4379 4380
		return IO_ACCEL_INELIGIBLE;

	/* calculate stripe information for the request */
D
Don Brace 已提交
4381 4382 4383
	blocks_per_row = le16_to_cpu(map->data_disks_per_row) *
				le16_to_cpu(map->strip_size);
	strip_size = le16_to_cpu(map->strip_size);
4384 4385 4386 4387 4388 4389 4390 4391 4392 4393
#if BITS_PER_LONG == 32
	tmpdiv = first_block;
	(void) do_div(tmpdiv, blocks_per_row);
	first_row = tmpdiv;
	tmpdiv = last_block;
	(void) do_div(tmpdiv, blocks_per_row);
	last_row = tmpdiv;
	first_row_offset = (u32) (first_block - (first_row * blocks_per_row));
	last_row_offset = (u32) (last_block - (last_row * blocks_per_row));
	tmpdiv = first_row_offset;
D
Don Brace 已提交
4394
	(void) do_div(tmpdiv, strip_size);
4395 4396
	first_column = tmpdiv;
	tmpdiv = last_row_offset;
D
Don Brace 已提交
4397
	(void) do_div(tmpdiv, strip_size);
4398 4399 4400 4401 4402 4403
	last_column = tmpdiv;
#else
	first_row = first_block / blocks_per_row;
	last_row = last_block / blocks_per_row;
	first_row_offset = (u32) (first_block - (first_row * blocks_per_row));
	last_row_offset = (u32) (last_block - (last_row * blocks_per_row));
D
Don Brace 已提交
4404 4405
	first_column = first_row_offset / strip_size;
	last_column = last_row_offset / strip_size;
4406 4407 4408 4409 4410 4411 4412
#endif

	/* if this isn't a single row/column then give to the controller */
	if ((first_row != last_row) || (first_column != last_column))
		return IO_ACCEL_INELIGIBLE;

	/* proceeding with driver mapping */
D
Don Brace 已提交
4413 4414
	total_disks_per_row = le16_to_cpu(map->data_disks_per_row) +
				le16_to_cpu(map->metadata_disks_per_row);
4415
	map_row = ((u32)(first_row >> map->parity_rotation_shift)) %
D
Don Brace 已提交
4416
				le16_to_cpu(map->row_cnt);
4417 4418 4419 4420 4421 4422 4423 4424 4425
	map_index = (map_row * total_disks_per_row) + first_column;

	switch (dev->raid_level) {
	case HPSA_RAID_0:
		break; /* nothing special to do */
	case HPSA_RAID_1:
		/* Handles load balance across RAID 1 members.
		 * (2-drive R1 and R10 with even # of drives.)
		 * Appropriate for SSDs, not optimal for HDDs
4426
		 */
D
Don Brace 已提交
4427
		BUG_ON(le16_to_cpu(map->layout_map_count) != 2);
4428
		if (dev->offload_to_mirror)
D
Don Brace 已提交
4429
			map_index += le16_to_cpu(map->data_disks_per_row);
4430
		dev->offload_to_mirror = !dev->offload_to_mirror;
4431 4432 4433 4434 4435
		break;
	case HPSA_RAID_ADM:
		/* Handles N-way mirrors  (R1-ADM)
		 * and R10 with # of drives divisible by 3.)
		 */
D
Don Brace 已提交
4436
		BUG_ON(le16_to_cpu(map->layout_map_count) != 3);
4437 4438 4439 4440 4441 4442

		offload_to_mirror = dev->offload_to_mirror;
		raid_map_helper(map, offload_to_mirror,
				&map_index, &current_group);
		/* set mirror group to use next time */
		offload_to_mirror =
D
Don Brace 已提交
4443 4444
			(offload_to_mirror >=
			le16_to_cpu(map->layout_map_count) - 1)
4445 4446 4447 4448 4449 4450 4451 4452 4453
			? 0 : offload_to_mirror + 1;
		dev->offload_to_mirror = offload_to_mirror;
		/* Avoid direct use of dev->offload_to_mirror within this
		 * function since multiple threads might simultaneously
		 * increment it beyond the range of dev->layout_map_count -1.
		 */
		break;
	case HPSA_RAID_5:
	case HPSA_RAID_6:
D
Don Brace 已提交
4454
		if (le16_to_cpu(map->layout_map_count) <= 1)
4455 4456 4457 4458
			break;

		/* Verify first and last block are in same RAID group */
		r5or6_blocks_per_row =
D
Don Brace 已提交
4459 4460
			le16_to_cpu(map->strip_size) *
			le16_to_cpu(map->data_disks_per_row);
4461
		BUG_ON(r5or6_blocks_per_row == 0);
D
Don Brace 已提交
4462 4463
		stripesize = r5or6_blocks_per_row *
			le16_to_cpu(map->layout_map_count);
4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478
#if BITS_PER_LONG == 32
		tmpdiv = first_block;
		first_group = do_div(tmpdiv, stripesize);
		tmpdiv = first_group;
		(void) do_div(tmpdiv, r5or6_blocks_per_row);
		first_group = tmpdiv;
		tmpdiv = last_block;
		last_group = do_div(tmpdiv, stripesize);
		tmpdiv = last_group;
		(void) do_div(tmpdiv, r5or6_blocks_per_row);
		last_group = tmpdiv;
#else
		first_group = (first_block % stripesize) / r5or6_blocks_per_row;
		last_group = (last_block % stripesize) / r5or6_blocks_per_row;
#endif
4479
		if (first_group != last_group)
4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501 4502 4503 4504 4505 4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525
			return IO_ACCEL_INELIGIBLE;

		/* Verify request is in a single row of RAID 5/6 */
#if BITS_PER_LONG == 32
		tmpdiv = first_block;
		(void) do_div(tmpdiv, stripesize);
		first_row = r5or6_first_row = r0_first_row = tmpdiv;
		tmpdiv = last_block;
		(void) do_div(tmpdiv, stripesize);
		r5or6_last_row = r0_last_row = tmpdiv;
#else
		first_row = r5or6_first_row = r0_first_row =
						first_block / stripesize;
		r5or6_last_row = r0_last_row = last_block / stripesize;
#endif
		if (r5or6_first_row != r5or6_last_row)
			return IO_ACCEL_INELIGIBLE;


		/* Verify request is in a single column */
#if BITS_PER_LONG == 32
		tmpdiv = first_block;
		first_row_offset = do_div(tmpdiv, stripesize);
		tmpdiv = first_row_offset;
		first_row_offset = (u32) do_div(tmpdiv, r5or6_blocks_per_row);
		r5or6_first_row_offset = first_row_offset;
		tmpdiv = last_block;
		r5or6_last_row_offset = do_div(tmpdiv, stripesize);
		tmpdiv = r5or6_last_row_offset;
		r5or6_last_row_offset = do_div(tmpdiv, r5or6_blocks_per_row);
		tmpdiv = r5or6_first_row_offset;
		(void) do_div(tmpdiv, map->strip_size);
		first_column = r5or6_first_column = tmpdiv;
		tmpdiv = r5or6_last_row_offset;
		(void) do_div(tmpdiv, map->strip_size);
		r5or6_last_column = tmpdiv;
#else
		first_row_offset = r5or6_first_row_offset =
			(u32)((first_block % stripesize) %
						r5or6_blocks_per_row);

		r5or6_last_row_offset =
			(u32)((last_block % stripesize) %
						r5or6_blocks_per_row);

		first_column = r5or6_first_column =
D
Don Brace 已提交
4526
			r5or6_first_row_offset / le16_to_cpu(map->strip_size);
4527
		r5or6_last_column =
D
Don Brace 已提交
4528
			r5or6_last_row_offset / le16_to_cpu(map->strip_size);
4529 4530 4531 4532 4533 4534
#endif
		if (r5or6_first_column != r5or6_last_column)
			return IO_ACCEL_INELIGIBLE;

		/* Request is eligible */
		map_row = ((u32)(first_row >> map->parity_rotation_shift)) %
D
Don Brace 已提交
4535
			le16_to_cpu(map->row_cnt);
4536 4537

		map_index = (first_group *
D
Don Brace 已提交
4538
			(le16_to_cpu(map->row_cnt) * total_disks_per_row)) +
4539 4540 4541 4542
			(map_row * total_disks_per_row) + first_column;
		break;
	default:
		return IO_ACCEL_INELIGIBLE;
4543
	}
4544

4545 4546 4547
	if (unlikely(map_index >= RAID_MAP_MAX_ENTRIES))
		return IO_ACCEL_INELIGIBLE;

4548 4549
	c->phys_disk = dev->phys_disk[map_index];

4550
	disk_handle = dd[map_index].ioaccel_handle;
D
Don Brace 已提交
4551 4552 4553 4554
	disk_block = le64_to_cpu(map->disk_starting_blk) +
			first_row * le16_to_cpu(map->strip_size) +
			(first_row_offset - first_column *
			le16_to_cpu(map->strip_size));
4555 4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596
	disk_block_cnt = block_cnt;

	/* handle differing logical/physical block sizes */
	if (map->phys_blk_shift) {
		disk_block <<= map->phys_blk_shift;
		disk_block_cnt <<= map->phys_blk_shift;
	}
	BUG_ON(disk_block_cnt > 0xffff);

	/* build the new CDB for the physical disk I/O */
	if (disk_block > 0xffffffff) {
		cdb[0] = is_write ? WRITE_16 : READ_16;
		cdb[1] = 0;
		cdb[2] = (u8) (disk_block >> 56);
		cdb[3] = (u8) (disk_block >> 48);
		cdb[4] = (u8) (disk_block >> 40);
		cdb[5] = (u8) (disk_block >> 32);
		cdb[6] = (u8) (disk_block >> 24);
		cdb[7] = (u8) (disk_block >> 16);
		cdb[8] = (u8) (disk_block >> 8);
		cdb[9] = (u8) (disk_block);
		cdb[10] = (u8) (disk_block_cnt >> 24);
		cdb[11] = (u8) (disk_block_cnt >> 16);
		cdb[12] = (u8) (disk_block_cnt >> 8);
		cdb[13] = (u8) (disk_block_cnt);
		cdb[14] = 0;
		cdb[15] = 0;
		cdb_len = 16;
	} else {
		cdb[0] = is_write ? WRITE_10 : READ_10;
		cdb[1] = 0;
		cdb[2] = (u8) (disk_block >> 24);
		cdb[3] = (u8) (disk_block >> 16);
		cdb[4] = (u8) (disk_block >> 8);
		cdb[5] = (u8) (disk_block);
		cdb[6] = 0;
		cdb[7] = (u8) (disk_block_cnt >> 8);
		cdb[8] = (u8) (disk_block_cnt);
		cdb[9] = 0;
		cdb_len = 10;
	}
	return hpsa_scsi_ioaccel_queue_command(h, c, disk_handle, cdb, cdb_len,
4597 4598
						dev->scsi3addr,
						dev->phys_disk[map_index]);
4599 4600
}

4601 4602 4603 4604 4605
/*
 * Submit commands down the "normal" RAID stack path
 * All callers to hpsa_ciss_submit must check lockup_detected
 * beforehand, before (opt.) and after calling cmd_alloc
 */
4606 4607 4608
static int hpsa_ciss_submit(struct ctlr_info *h,
	struct CommandList *c, struct scsi_cmnd *cmd,
	unsigned char scsi3addr[])
4609 4610 4611 4612 4613 4614
{
	cmd->host_scribble = (unsigned char *) c;
	c->cmd_type = CMD_SCSI;
	c->scsi_cmd = cmd;
	c->Header.ReplyQueue = 0;  /* unused in simple mode */
	memcpy(&c->Header.LUN.LunAddrBytes[0], &scsi3addr[0], 8);
4615
	c->Header.tag = cpu_to_le64((c->cmdindex << DIRECT_LOOKUP_SHIFT));
4616 4617 4618 4619 4620 4621 4622 4623 4624

	/* Fill in the request block... */

	c->Request.Timeout = 0;
	BUG_ON(cmd->cmd_len > sizeof(c->Request.CDB));
	c->Request.CDBLen = cmd->cmd_len;
	memcpy(c->Request.CDB, cmd->cmnd, cmd->cmd_len);
	switch (cmd->sc_data_direction) {
	case DMA_TO_DEVICE:
4625 4626
		c->Request.type_attr_dir =
			TYPE_ATTR_DIR(TYPE_CMD, ATTR_SIMPLE, XFER_WRITE);
4627 4628
		break;
	case DMA_FROM_DEVICE:
4629 4630
		c->Request.type_attr_dir =
			TYPE_ATTR_DIR(TYPE_CMD, ATTR_SIMPLE, XFER_READ);
4631 4632
		break;
	case DMA_NONE:
4633 4634
		c->Request.type_attr_dir =
			TYPE_ATTR_DIR(TYPE_CMD, ATTR_SIMPLE, XFER_NONE);
4635 4636 4637 4638 4639 4640 4641
		break;
	case DMA_BIDIRECTIONAL:
		/* This can happen if a buggy application does a scsi passthru
		 * and sets both inlen and outlen to non-zero. ( see
		 * ../scsi/scsi_ioctl.c:scsi_ioctl_send_command() )
		 */

4642 4643
		c->Request.type_attr_dir =
			TYPE_ATTR_DIR(TYPE_CMD, ATTR_SIMPLE, XFER_RSVD);
4644 4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660
		/* This is technically wrong, and hpsa controllers should
		 * reject it with CMD_INVALID, which is the most correct
		 * response, but non-fibre backends appear to let it
		 * slide by, and give the same results as if this field
		 * were set correctly.  Either way is acceptable for
		 * our purposes here.
		 */

		break;

	default:
		dev_err(&h->pdev->dev, "unknown data direction: %d\n",
			cmd->sc_data_direction);
		BUG();
		break;
	}

4661
	if (hpsa_scatter_gather(h, c, cmd) < 0) { /* Fill SG list */
4662
		hpsa_cmd_resolve_and_free(h, c);
4663 4664 4665 4666 4667 4668 4669
		return SCSI_MLQUEUE_HOST_BUSY;
	}
	enqueue_cmd_and_start_io(h, c);
	/* the cmd'll come back via intr handler in complete_scsi_command()  */
	return 0;
}

4670 4671 4672 4673 4674 4675 4676 4677 4678 4679 4680 4681 4682 4683 4684 4685 4686 4687
static void hpsa_cmd_init(struct ctlr_info *h, int index,
				struct CommandList *c)
{
	dma_addr_t cmd_dma_handle, err_dma_handle;

	/* Zero out all of commandlist except the last field, refcount */
	memset(c, 0, offsetof(struct CommandList, refcount));
	c->Header.tag = cpu_to_le64((u64) (index << DIRECT_LOOKUP_SHIFT));
	cmd_dma_handle = h->cmd_pool_dhandle + index * sizeof(*c);
	c->err_info = h->errinfo_pool + index;
	memset(c->err_info, 0, sizeof(*c->err_info));
	err_dma_handle = h->errinfo_pool_dhandle
	    + index * sizeof(*c->err_info);
	c->cmdindex = index;
	c->busaddr = (u32) cmd_dma_handle;
	c->ErrDesc.Addr = cpu_to_le64((u64) err_dma_handle);
	c->ErrDesc.Len = cpu_to_le32((u32) sizeof(*c->err_info));
	c->h = h;
4688
	c->scsi_cmd = SCSI_CMD_IDLE;
4689 4690 4691 4692 4693 4694 4695 4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707
}

static void hpsa_preinitialize_commands(struct ctlr_info *h)
{
	int i;

	for (i = 0; i < h->nr_cmds; i++) {
		struct CommandList *c = h->cmd_pool + i;

		hpsa_cmd_init(h, i, c);
		atomic_set(&c->refcount, 0);
	}
}

static inline void hpsa_cmd_partial_init(struct ctlr_info *h, int index,
				struct CommandList *c)
{
	dma_addr_t cmd_dma_handle = h->cmd_pool_dhandle + index * sizeof(*c);

4708 4709
	BUG_ON(c->cmdindex != index);

4710 4711 4712 4713 4714
	memset(c->Request.CDB, 0, sizeof(c->Request.CDB));
	memset(c->err_info, 0, sizeof(*c->err_info));
	c->busaddr = (u32) cmd_dma_handle;
}

4715 4716 4717 4718 4719 4720 4721 4722 4723 4724 4725 4726 4727 4728 4729 4730
static int hpsa_ioaccel_submit(struct ctlr_info *h,
		struct CommandList *c, struct scsi_cmnd *cmd,
		unsigned char *scsi3addr)
{
	struct hpsa_scsi_dev_t *dev = cmd->device->hostdata;
	int rc = IO_ACCEL_INELIGIBLE;

	cmd->host_scribble = (unsigned char *) c;

	if (dev->offload_enabled) {
		hpsa_cmd_init(h, c->cmdindex, c);
		c->cmd_type = CMD_SCSI;
		c->scsi_cmd = cmd;
		rc = hpsa_scsi_ioaccel_raid_map(h, c);
		if (rc < 0)     /* scsi_dma_map failed. */
			rc = SCSI_MLQUEUE_HOST_BUSY;
4731
	} else if (dev->hba_ioaccel_enabled) {
4732 4733 4734 4735 4736 4737 4738 4739 4740 4741
		hpsa_cmd_init(h, c->cmdindex, c);
		c->cmd_type = CMD_SCSI;
		c->scsi_cmd = cmd;
		rc = hpsa_scsi_ioaccel_direct_map(h, c);
		if (rc < 0)     /* scsi_dma_map failed. */
			rc = SCSI_MLQUEUE_HOST_BUSY;
	}
	return rc;
}

4742 4743 4744 4745
static void hpsa_command_resubmit_worker(struct work_struct *work)
{
	struct scsi_cmnd *cmd;
	struct hpsa_scsi_dev_t *dev;
4746
	struct CommandList *c = container_of(work, struct CommandList, work);
4747 4748 4749 4750 4751

	cmd = c->scsi_cmd;
	dev = cmd->device->hostdata;
	if (!dev) {
		cmd->result = DID_NO_CONNECT << 16;
4752
		return hpsa_cmd_free_and_done(c->h, c, cmd);
4753
	}
W
Webb Scales 已提交
4754 4755
	if (c->reset_pending)
		return hpsa_cmd_resolve_and_free(c->h, c);
4756 4757
	if (c->abort_pending)
		return hpsa_cmd_abort_and_free(c->h, c, cmd);
4758 4759 4760 4761 4762 4763 4764 4765 4766 4767 4768 4769 4770 4771 4772 4773 4774
	if (c->cmd_type == CMD_IOACCEL2) {
		struct ctlr_info *h = c->h;
		struct io_accel2_cmd *c2 = &h->ioaccel2_cmd_pool[c->cmdindex];
		int rc;

		if (c2->error_data.serv_response ==
				IOACCEL2_STATUS_SR_TASK_COMP_SET_FULL) {
			rc = hpsa_ioaccel_submit(h, c, cmd, dev->scsi3addr);
			if (rc == 0)
				return;
			if (rc == SCSI_MLQUEUE_HOST_BUSY) {
				/*
				 * If we get here, it means dma mapping failed.
				 * Try again via scsi mid layer, which will
				 * then get SCSI_MLQUEUE_HOST_BUSY.
				 */
				cmd->result = DID_IMM_RETRY << 16;
4775
				return hpsa_cmd_free_and_done(h, c, cmd);
4776 4777 4778 4779
			}
			/* else, fall thru and resubmit down CISS path */
		}
	}
4780
	hpsa_cmd_partial_init(c->h, c->cmdindex, c);
4781 4782 4783 4784 4785
	if (hpsa_ciss_submit(c->h, c, cmd, dev->scsi3addr)) {
		/*
		 * If we get here, it means dma mapping failed. Try
		 * again via scsi mid layer, which will then get
		 * SCSI_MLQUEUE_HOST_BUSY.
4786 4787 4788
		 *
		 * hpsa_ciss_submit will have already freed c
		 * if it encountered a dma mapping failure.
4789 4790 4791 4792 4793 4794
		 */
		cmd->result = DID_IMM_RETRY << 16;
		cmd->scsi_done(cmd);
	}
}

4795 4796 4797 4798 4799 4800 4801 4802 4803 4804 4805
/* Running in struct Scsi_Host->host_lock less mode */
static int hpsa_scsi_queue_command(struct Scsi_Host *sh, struct scsi_cmnd *cmd)
{
	struct ctlr_info *h;
	struct hpsa_scsi_dev_t *dev;
	unsigned char scsi3addr[8];
	struct CommandList *c;
	int rc = 0;

	/* Get the ptr to our adapter structure out of cmd->host. */
	h = sdev_to_hba(cmd->device);
4806 4807 4808

	BUG_ON(cmd->request->tag < 0);

4809 4810 4811 4812 4813 4814 4815
	dev = cmd->device->hostdata;
	if (!dev) {
		cmd->result = DID_NO_CONNECT << 16;
		cmd->scsi_done(cmd);
		return 0;
	}

4816
	memcpy(scsi3addr, dev->scsi3addr, sizeof(scsi3addr));
4817

4818
	if (unlikely(lockup_detected(h))) {
4819
		cmd->result = DID_NO_CONNECT << 16;
4820 4821 4822
		cmd->scsi_done(cmd);
		return 0;
	}
4823
	c = cmd_tagged_alloc(h, cmd);
4824

4825 4826
	/*
	 * Call alternate submit routine for I/O accelerated commands.
4827 4828 4829 4830 4831
	 * Retries always go down the normal I/O path.
	 */
	if (likely(cmd->retries == 0 &&
		cmd->request->cmd_type == REQ_TYPE_FS &&
		h->acciopath_status)) {
4832 4833 4834 4835
		rc = hpsa_ioaccel_submit(h, c, cmd, scsi3addr);
		if (rc == 0)
			return 0;
		if (rc == SCSI_MLQUEUE_HOST_BUSY) {
4836
			hpsa_cmd_resolve_and_free(h, c);
4837
			return SCSI_MLQUEUE_HOST_BUSY;
4838 4839 4840 4841 4842
		}
	}
	return hpsa_ciss_submit(h, c, cmd, scsi3addr);
}

4843
static void hpsa_scan_complete(struct ctlr_info *h)
4844 4845 4846
{
	unsigned long flags;

4847 4848 4849 4850
	spin_lock_irqsave(&h->scan_lock, flags);
	h->scan_finished = 1;
	wake_up_all(&h->scan_wait_queue);
	spin_unlock_irqrestore(&h->scan_lock, flags);
4851 4852
}

4853 4854 4855 4856 4857
static void hpsa_scan_start(struct Scsi_Host *sh)
{
	struct ctlr_info *h = shost_to_hba(sh);
	unsigned long flags;

4858 4859 4860 4861 4862 4863 4864 4865
	/*
	 * Don't let rescans be initiated on a controller known to be locked
	 * up.  If the controller locks up *during* a rescan, that thread is
	 * probably hosed, but at least we can prevent new rescan threads from
	 * piling up on a locked up controller.
	 */
	if (unlikely(lockup_detected(h)))
		return hpsa_scan_complete(h);
4866

4867 4868 4869 4870 4871 4872 4873 4874 4875 4876 4877 4878 4879 4880 4881 4882
	/* wait until any scan already in progress is finished. */
	while (1) {
		spin_lock_irqsave(&h->scan_lock, flags);
		if (h->scan_finished)
			break;
		spin_unlock_irqrestore(&h->scan_lock, flags);
		wait_event(h->scan_wait_queue, h->scan_finished);
		/* Note: We don't need to worry about a race between this
		 * thread and driver unload because the midlayer will
		 * have incremented the reference count, so unload won't
		 * happen if we're in here.
		 */
	}
	h->scan_finished = 0; /* mark scan as in progress */
	spin_unlock_irqrestore(&h->scan_lock, flags);

4883 4884
	if (unlikely(lockup_detected(h)))
		return hpsa_scan_complete(h);
4885

4886 4887
	hpsa_update_scsi_devices(h, h->scsi_host->host_no);

4888
	hpsa_scan_complete(h);
4889 4890
}

D
Don Brace 已提交
4891 4892
static int hpsa_change_queue_depth(struct scsi_device *sdev, int qdepth)
{
4893 4894 4895 4896
	struct hpsa_scsi_dev_t *logical_drive = sdev->hostdata;

	if (!logical_drive)
		return -ENODEV;
D
Don Brace 已提交
4897 4898 4899

	if (qdepth < 1)
		qdepth = 1;
4900 4901 4902 4903
	else if (qdepth > logical_drive->queue_depth)
		qdepth = logical_drive->queue_depth;

	return scsi_change_queue_depth(sdev, qdepth);
D
Don Brace 已提交
4904 4905
}

4906 4907 4908 4909 4910 4911 4912 4913 4914 4915 4916 4917 4918
static int hpsa_scan_finished(struct Scsi_Host *sh,
	unsigned long elapsed_time)
{
	struct ctlr_info *h = shost_to_hba(sh);
	unsigned long flags;
	int finished;

	spin_lock_irqsave(&h->scan_lock, flags);
	finished = h->scan_finished;
	spin_unlock_irqrestore(&h->scan_lock, flags);
	return finished;
}

4919
static int hpsa_scsi_host_alloc(struct ctlr_info *h)
4920
{
4921 4922
	struct Scsi_Host *sh;
	int error;
4923

4924
	sh = scsi_host_alloc(&hpsa_driver_template, sizeof(h));
4925 4926 4927 4928
	if (sh == NULL) {
		dev_err(&h->pdev->dev, "scsi_host_alloc failed\n");
		return -ENOMEM;
	}
4929 4930 4931 4932 4933 4934 4935 4936

	sh->io_port = 0;
	sh->n_io_port = 0;
	sh->this_id = -1;
	sh->max_channel = 3;
	sh->max_cmd_len = MAX_COMMAND_SIZE;
	sh->max_lun = HPSA_MAX_LUN;
	sh->max_id = HPSA_MAX_LUN;
4937
	sh->can_queue = h->nr_cmds - HPSA_NRESERVED_CMDS;
4938
	sh->cmd_per_lun = sh->can_queue;
4939 4940 4941 4942
	sh->sg_tablesize = h->maxsgentries;
	sh->hostdata[0] = (unsigned long) h;
	sh->irq = h->intr[h->intr_mode];
	sh->unique_id = sh->irq;
4943 4944 4945 4946 4947
	error = scsi_init_shared_tag_map(sh, sh->can_queue);
	if (error) {
		dev_err(&h->pdev->dev,
			"%s: scsi_init_shared_tag_map failed for controller %d\n",
			__func__, h->ctlr);
4948 4949
			scsi_host_put(sh);
			return error;
4950
	}
4951
	h->scsi_host = sh;
4952
	return 0;
4953
}
4954

4955 4956 4957 4958 4959 4960 4961 4962 4963 4964 4965
static int hpsa_scsi_add_host(struct ctlr_info *h)
{
	int rv;

	rv = scsi_add_host(h->scsi_host, &h->pdev->dev);
	if (rv) {
		dev_err(&h->pdev->dev, "scsi_add_host failed\n");
		return rv;
	}
	scsi_scan_host(h->scsi_host);
	return 0;
4966 4967
}

4968 4969 4970 4971 4972 4973 4974 4975 4976 4977 4978 4979 4980 4981 4982 4983 4984
/*
 * The block layer has already gone to the trouble of picking out a unique,
 * small-integer tag for this request.  We use an offset from that value as
 * an index to select our command block.  (The offset allows us to reserve the
 * low-numbered entries for our own uses.)
 */
static int hpsa_get_cmd_index(struct scsi_cmnd *scmd)
{
	int idx = scmd->request->tag;

	if (idx < 0)
		return idx;

	/* Offset to leave space for internal cmds. */
	return idx += HPSA_NRESERVED_CMDS;
}

4985 4986 4987 4988 4989 4990 4991 4992 4993 4994 4995 4996 4997 4998 4999 5000 5001 5002 5003 5004 5005 5006 5007 5008 5009 5010 5011 5012 5013 5014 5015 5016 5017 5018 5019 5020 5021 5022 5023 5024 5025 5026 5027
/*
 * Send a TEST_UNIT_READY command to the specified LUN using the specified
 * reply queue; returns zero if the unit is ready, and non-zero otherwise.
 */
static int hpsa_send_test_unit_ready(struct ctlr_info *h,
				struct CommandList *c, unsigned char lunaddr[],
				int reply_queue)
{
	int rc;

	/* Send the Test Unit Ready, fill_cmd can't fail, no mapping */
	(void) fill_cmd(c, TEST_UNIT_READY, h,
			NULL, 0, 0, lunaddr, TYPE_CMD);
	rc = hpsa_scsi_do_simple_cmd(h, c, reply_queue, NO_TIMEOUT);
	if (rc)
		return rc;
	/* no unmap needed here because no data xfer. */

	/* Check if the unit is already ready. */
	if (c->err_info->CommandStatus == CMD_SUCCESS)
		return 0;

	/*
	 * The first command sent after reset will receive "unit attention" to
	 * indicate that the LUN has been reset...this is actually what we're
	 * looking for (but, success is good too).
	 */
	if (c->err_info->CommandStatus == CMD_TARGET_STATUS &&
		c->err_info->ScsiStatus == SAM_STAT_CHECK_CONDITION &&
			(c->err_info->SenseInfo[2] == NO_SENSE ||
			 c->err_info->SenseInfo[2] == UNIT_ATTENTION))
		return 0;

	return 1;
}

/*
 * Wait for a TEST_UNIT_READY command to complete, retrying as necessary;
 * returns zero when the unit is ready, and non-zero when giving up.
 */
static int hpsa_wait_for_test_unit_ready(struct ctlr_info *h,
				struct CommandList *c,
				unsigned char lunaddr[], int reply_queue)
5028
{
5029
	int rc;
5030 5031 5032 5033
	int count = 0;
	int waittime = 1; /* seconds */

	/* Send test unit ready until device ready, or give up. */
5034
	for (count = 0; count < HPSA_TUR_RETRY_LIMIT; count++) {
5035

5036 5037
		/*
		 * Wait for a bit.  do this first, because if we send
5038 5039 5040
		 * the TUR right away, the reset will just abort it.
		 */
		msleep(1000 * waittime);
5041 5042 5043 5044

		rc = hpsa_send_test_unit_ready(h, c, lunaddr, reply_queue);
		if (!rc)
			break;
5045 5046 5047

		/* Increase wait time with each try, up to a point. */
		if (waittime < HPSA_MAX_WAIT_INTERVAL_SECS)
5048
			waittime *= 2;
5049

5050 5051 5052 5053
		dev_warn(&h->pdev->dev,
			 "waiting %d secs for device to become ready.\n",
			 waittime);
	}
5054

5055 5056
	return rc;
}
5057

5058 5059 5060 5061 5062 5063 5064 5065 5066 5067 5068 5069 5070 5071 5072 5073 5074 5075 5076 5077 5078 5079 5080 5081 5082 5083 5084 5085
static int wait_for_device_to_become_ready(struct ctlr_info *h,
					   unsigned char lunaddr[],
					   int reply_queue)
{
	int first_queue;
	int last_queue;
	int rq;
	int rc = 0;
	struct CommandList *c;

	c = cmd_alloc(h);

	/*
	 * If no specific reply queue was requested, then send the TUR
	 * repeatedly, requesting a reply on each reply queue; otherwise execute
	 * the loop exactly once using only the specified queue.
	 */
	if (reply_queue == DEFAULT_REPLY_QUEUE) {
		first_queue = 0;
		last_queue = h->nreply_queues - 1;
	} else {
		first_queue = reply_queue;
		last_queue = reply_queue;
	}

	for (rq = first_queue; rq <= last_queue; rq++) {
		rc = hpsa_wait_for_test_unit_ready(h, c, lunaddr, rq);
		if (rc)
5086 5087 5088 5089 5090 5091 5092 5093
			break;
	}

	if (rc)
		dev_warn(&h->pdev->dev, "giving up on device.\n");
	else
		dev_warn(&h->pdev->dev, "device is ready.\n");

5094
	cmd_free(h, c);
5095 5096 5097 5098 5099 5100 5101 5102 5103 5104 5105
	return rc;
}

/* Need at least one of these error handlers to keep ../scsi/hosts.c from
 * complaining.  Doing a host- or bus-reset can't do anything good here.
 */
static int hpsa_eh_device_reset_handler(struct scsi_cmnd *scsicmd)
{
	int rc;
	struct ctlr_info *h;
	struct hpsa_scsi_dev_t *dev;
5106
	char msg[48];
5107 5108 5109 5110 5111

	/* find the controller to which the command to be aborted was sent */
	h = sdev_to_hba(scsicmd->device);
	if (h == NULL) /* paranoia */
		return FAILED;
5112 5113 5114 5115

	if (lockup_detected(h))
		return FAILED;

5116 5117
	dev = scsicmd->device->hostdata;
	if (!dev) {
W
Webb Scales 已提交
5118
		dev_err(&h->pdev->dev, "%s: device lookup failed\n", __func__);
5119 5120
		return FAILED;
	}
5121 5122 5123

	/* if controller locked up, we can guarantee command won't complete */
	if (lockup_detected(h)) {
5124 5125 5126
		snprintf(msg, sizeof(msg),
			 "cmd %d RESET FAILED, lockup detected",
			 hpsa_get_cmd_index(scsicmd));
5127
		hpsa_show_dev_msg(KERN_WARNING, h, dev, msg);
5128 5129 5130 5131 5132
		return FAILED;
	}

	/* this reset request might be the result of a lockup; check */
	if (detect_controller_lockup(h)) {
5133 5134 5135
		snprintf(msg, sizeof(msg),
			 "cmd %d RESET FAILED, new lockup detected",
			 hpsa_get_cmd_index(scsicmd));
5136
		hpsa_show_dev_msg(KERN_WARNING, h, dev, msg);
5137 5138 5139
		return FAILED;
	}

W
Webb Scales 已提交
5140 5141 5142 5143
	/* Do not attempt on controller */
	if (is_hba_lunid(dev->scsi3addr))
		return SUCCESS;

5144 5145
	hpsa_show_dev_msg(KERN_WARNING, h, dev, "resetting");

5146
	/* send a reset to the SCSI LUN which the command was sent to */
W
Webb Scales 已提交
5147 5148
	rc = hpsa_do_reset(h, dev, dev->scsi3addr, HPSA_RESET_TYPE_LUN,
			   DEFAULT_REPLY_QUEUE);
5149 5150
	snprintf(msg, sizeof(msg), "reset %s",
		 rc == 0 ? "completed successfully" : "failed");
W
Webb Scales 已提交
5151 5152
	hpsa_show_dev_msg(KERN_WARNING, h, dev, msg);
	return rc == 0 ? SUCCESS : FAILED;
5153 5154
}

5155 5156 5157 5158 5159 5160 5161 5162 5163 5164 5165 5166 5167 5168 5169
static void swizzle_abort_tag(u8 *tag)
{
	u8 original_tag[8];

	memcpy(original_tag, tag, 8);
	tag[0] = original_tag[3];
	tag[1] = original_tag[2];
	tag[2] = original_tag[1];
	tag[3] = original_tag[0];
	tag[4] = original_tag[7];
	tag[5] = original_tag[6];
	tag[6] = original_tag[5];
	tag[7] = original_tag[4];
}

5170
static void hpsa_get_tag(struct ctlr_info *h,
D
Don Brace 已提交
5171
	struct CommandList *c, __le32 *taglower, __le32 *tagupper)
5172
{
D
Don Brace 已提交
5173
	u64 tag;
5174 5175 5176
	if (c->cmd_type == CMD_IOACCEL1) {
		struct io_accel1_cmd *cm1 = (struct io_accel1_cmd *)
			&h->ioaccel_cmd_pool[c->cmdindex];
D
Don Brace 已提交
5177 5178 5179
		tag = le64_to_cpu(cm1->tag);
		*tagupper = cpu_to_le32(tag >> 32);
		*taglower = cpu_to_le32(tag);
5180 5181 5182 5183 5184
		return;
	}
	if (c->cmd_type == CMD_IOACCEL2) {
		struct io_accel2_cmd *cm2 = (struct io_accel2_cmd *)
			&h->ioaccel2_cmd_pool[c->cmdindex];
5185 5186 5187
		/* upper tag not used in ioaccel2 mode */
		memset(tagupper, 0, sizeof(*tagupper));
		*taglower = cm2->Tag;
5188
		return;
5189
	}
D
Don Brace 已提交
5190 5191 5192
	tag = le64_to_cpu(c->Header.tag);
	*tagupper = cpu_to_le32(tag >> 32);
	*taglower = cpu_to_le32(tag);
5193 5194
}

5195
static int hpsa_send_abort(struct ctlr_info *h, unsigned char *scsi3addr,
S
Stephen Cameron 已提交
5196
	struct CommandList *abort, int reply_queue)
5197 5198 5199 5200
{
	int rc = IO_OK;
	struct CommandList *c;
	struct ErrorInfo *ei;
D
Don Brace 已提交
5201
	__le32 tagupper, taglower;
5202

5203
	c = cmd_alloc(h);
5204

5205
	/* fill_cmd can't fail here, no buffer to map */
S
Stephen Cameron 已提交
5206
	(void) fill_cmd(c, HPSA_ABORT_MSG, h, &abort->Header.tag,
5207
		0, 0, scsi3addr, TYPE_MSG);
S
Stephen Cameron 已提交
5208
	if (h->needs_abort_tags_swizzled)
5209
		swizzle_abort_tag(&c->Request.CDB[4]);
5210
	(void) hpsa_scsi_do_simple_cmd(h, c, reply_queue, NO_TIMEOUT);
5211
	hpsa_get_tag(h, abort, &taglower, &tagupper);
5212
	dev_dbg(&h->pdev->dev, "%s: Tag:0x%08x:%08x: do_simple_cmd(abort) completed.\n",
5213
		__func__, tagupper, taglower);
5214 5215 5216 5217 5218 5219
	/* no unmap needed here because no data xfer. */

	ei = c->err_info;
	switch (ei->CommandStatus) {
	case CMD_SUCCESS:
		break;
5220 5221 5222
	case CMD_TMF_STATUS:
		rc = hpsa_evaluate_tmf_status(h, c);
		break;
5223 5224 5225 5226 5227
	case CMD_UNABORTABLE: /* Very common, don't make noise. */
		rc = -1;
		break;
	default:
		dev_dbg(&h->pdev->dev, "%s: Tag:0x%08x:%08x: interpreting error.\n",
5228
			__func__, tagupper, taglower);
5229
		hpsa_scsi_interpret_error(h, c);
5230 5231 5232
		rc = -1;
		break;
	}
5233
	cmd_free(h, c);
5234 5235
	dev_dbg(&h->pdev->dev, "%s: Tag:0x%08x:%08x: Finished.\n",
		__func__, tagupper, taglower);
5236 5237 5238
	return rc;
}

5239 5240 5241 5242 5243 5244 5245
static void setup_ioaccel2_abort_cmd(struct CommandList *c, struct ctlr_info *h,
	struct CommandList *command_to_abort, int reply_queue)
{
	struct io_accel2_cmd *c2 = &h->ioaccel2_cmd_pool[c->cmdindex];
	struct hpsa_tmf_struct *ac = (struct hpsa_tmf_struct *) c2;
	struct io_accel2_cmd *c2a =
		&h->ioaccel2_cmd_pool[command_to_abort->cmdindex];
5246
	struct scsi_cmnd *scmd = command_to_abort->scsi_cmd;
5247 5248 5249 5250 5251 5252 5253 5254 5255 5256 5257 5258 5259 5260
	struct hpsa_scsi_dev_t *dev = scmd->device->hostdata;

	/*
	 * We're overlaying struct hpsa_tmf_struct on top of something which
	 * was allocated as a struct io_accel2_cmd, so we better be sure it
	 * actually fits, and doesn't overrun the error info space.
	 */
	BUILD_BUG_ON(sizeof(struct hpsa_tmf_struct) >
			sizeof(struct io_accel2_cmd));
	BUG_ON(offsetof(struct io_accel2_cmd, error_data) <
			offsetof(struct hpsa_tmf_struct, error_len) +
				sizeof(ac->error_len));

	c->cmd_type = IOACCEL2_TMF;
5261 5262
	c->scsi_cmd = SCSI_CMD_BUSY;

5263 5264 5265 5266 5267 5268 5269 5270 5271 5272 5273 5274 5275 5276 5277 5278 5279 5280
	/* Adjust the DMA address to point to the accelerated command buffer */
	c->busaddr = (u32) h->ioaccel2_cmd_pool_dhandle +
				(c->cmdindex * sizeof(struct io_accel2_cmd));
	BUG_ON(c->busaddr & 0x0000007F);

	memset(ac, 0, sizeof(*c2)); /* yes this is correct */
	ac->iu_type = IOACCEL2_IU_TMF_TYPE;
	ac->reply_queue = reply_queue;
	ac->tmf = IOACCEL2_TMF_ABORT;
	ac->it_nexus = cpu_to_le32(dev->ioaccel_handle);
	memset(ac->lun_id, 0, sizeof(ac->lun_id));
	ac->tag = cpu_to_le64(c->cmdindex << DIRECT_LOOKUP_SHIFT);
	ac->abort_tag = cpu_to_le64(le32_to_cpu(c2a->Tag));
	ac->error_ptr = cpu_to_le64(c->busaddr +
			offsetof(struct io_accel2_cmd, error_data));
	ac->error_len = cpu_to_le32(sizeof(c2->error_data));
}

5281 5282 5283 5284 5285 5286 5287 5288
/* ioaccel2 path firmware cannot handle abort task requests.
 * Change abort requests to physical target reset, and send to the
 * address of the physical disk used for the ioaccel 2 command.
 * Return 0 on success (IO_OK)
 *	 -1 on failure
 */

static int hpsa_send_reset_as_abort_ioaccel2(struct ctlr_info *h,
5289
	unsigned char *scsi3addr, struct CommandList *abort, int reply_queue)
5290 5291 5292 5293 5294 5295 5296 5297
{
	int rc = IO_OK;
	struct scsi_cmnd *scmd; /* scsi command within request being aborted */
	struct hpsa_scsi_dev_t *dev; /* device to which scsi cmd was sent */
	unsigned char phys_scsi3addr[8]; /* addr of phys disk with volume */
	unsigned char *psa = &phys_scsi3addr[0];

	/* Get a pointer to the hpsa logical device. */
5298
	scmd = abort->scsi_cmd;
5299 5300 5301 5302 5303 5304 5305
	dev = (struct hpsa_scsi_dev_t *)(scmd->device->hostdata);
	if (dev == NULL) {
		dev_warn(&h->pdev->dev,
			"Cannot abort: no device pointer for command.\n");
			return -1; /* not abortable */
	}

5306 5307
	if (h->raid_offload_debug > 0)
		dev_info(&h->pdev->dev,
5308
			"scsi %d:%d:%d:%d %s scsi3addr 0x%02x%02x%02x%02x%02x%02x%02x%02x\n",
5309
			h->scsi_host->host_no, dev->bus, dev->target, dev->lun,
5310
			"Reset as abort",
5311 5312 5313
			scsi3addr[0], scsi3addr[1], scsi3addr[2], scsi3addr[3],
			scsi3addr[4], scsi3addr[5], scsi3addr[6], scsi3addr[7]);

5314 5315 5316 5317 5318 5319 5320 5321 5322 5323 5324 5325 5326
	if (!dev->offload_enabled) {
		dev_warn(&h->pdev->dev,
			"Can't abort: device is not operating in HP SSD Smart Path mode.\n");
		return -1; /* not abortable */
	}

	/* Incoming scsi3addr is logical addr. We need physical disk addr. */
	if (!hpsa_get_pdisk_of_ioaccel2(h, abort, psa)) {
		dev_warn(&h->pdev->dev, "Can't abort: Failed lookup of physical address.\n");
		return -1; /* not abortable */
	}

	/* send the reset */
5327 5328 5329 5330 5331
	if (h->raid_offload_debug > 0)
		dev_info(&h->pdev->dev,
			"Reset as abort: Resetting physical device at scsi3addr 0x%02x%02x%02x%02x%02x%02x%02x%02x\n",
			psa[0], psa[1], psa[2], psa[3],
			psa[4], psa[5], psa[6], psa[7]);
W
Webb Scales 已提交
5332
	rc = hpsa_do_reset(h, dev, psa, HPSA_RESET_TYPE_TARGET, reply_queue);
5333 5334 5335 5336 5337 5338 5339 5340 5341
	if (rc != 0) {
		dev_warn(&h->pdev->dev,
			"Reset as abort: Failed on physical device at scsi3addr 0x%02x%02x%02x%02x%02x%02x%02x%02x\n",
			psa[0], psa[1], psa[2], psa[3],
			psa[4], psa[5], psa[6], psa[7]);
		return rc; /* failed to reset */
	}

	/* wait for device to recover */
5342
	if (wait_for_device_to_become_ready(h, psa, reply_queue) != 0) {
5343 5344 5345 5346 5347 5348 5349 5350 5351 5352 5353 5354 5355 5356 5357 5358
		dev_warn(&h->pdev->dev,
			"Reset as abort: Failed: Device never recovered from reset: 0x%02x%02x%02x%02x%02x%02x%02x%02x\n",
			psa[0], psa[1], psa[2], psa[3],
			psa[4], psa[5], psa[6], psa[7]);
		return -1;  /* failed to recover */
	}

	/* device recovered */
	dev_info(&h->pdev->dev,
		"Reset as abort: Device recovered from reset: scsi3addr 0x%02x%02x%02x%02x%02x%02x%02x%02x\n",
		psa[0], psa[1], psa[2], psa[3],
		psa[4], psa[5], psa[6], psa[7]);

	return rc; /* success */
}

5359 5360 5361 5362 5363 5364 5365 5366 5367 5368 5369 5370 5371 5372 5373 5374 5375 5376 5377 5378 5379 5380 5381 5382 5383 5384 5385 5386 5387 5388 5389 5390 5391 5392 5393 5394 5395 5396 5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407
static int hpsa_send_abort_ioaccel2(struct ctlr_info *h,
	struct CommandList *abort, int reply_queue)
{
	int rc = IO_OK;
	struct CommandList *c;
	__le32 taglower, tagupper;
	struct hpsa_scsi_dev_t *dev;
	struct io_accel2_cmd *c2;

	dev = abort->scsi_cmd->device->hostdata;
	if (!dev->offload_enabled && !dev->hba_ioaccel_enabled)
		return -1;

	c = cmd_alloc(h);
	setup_ioaccel2_abort_cmd(c, h, abort, reply_queue);
	c2 = &h->ioaccel2_cmd_pool[c->cmdindex];
	(void) hpsa_scsi_do_simple_cmd(h, c, reply_queue, NO_TIMEOUT);
	hpsa_get_tag(h, abort, &taglower, &tagupper);
	dev_dbg(&h->pdev->dev,
		"%s: Tag:0x%08x:%08x: do_simple_cmd(ioaccel2 abort) completed.\n",
		__func__, tagupper, taglower);
	/* no unmap needed here because no data xfer. */

	dev_dbg(&h->pdev->dev,
		"%s: Tag:0x%08x:%08x: abort service response = 0x%02x.\n",
		__func__, tagupper, taglower, c2->error_data.serv_response);
	switch (c2->error_data.serv_response) {
	case IOACCEL2_SERV_RESPONSE_TMF_COMPLETE:
	case IOACCEL2_SERV_RESPONSE_TMF_SUCCESS:
		rc = 0;
		break;
	case IOACCEL2_SERV_RESPONSE_TMF_REJECTED:
	case IOACCEL2_SERV_RESPONSE_FAILURE:
	case IOACCEL2_SERV_RESPONSE_TMF_WRONG_LUN:
		rc = -1;
		break;
	default:
		dev_warn(&h->pdev->dev,
			"%s: Tag:0x%08x:%08x: unknown abort service response 0x%02x\n",
			__func__, tagupper, taglower,
			c2->error_data.serv_response);
		rc = -1;
	}
	cmd_free(h, c);
	dev_dbg(&h->pdev->dev, "%s: Tag:0x%08x:%08x: Finished.\n", __func__,
		tagupper, taglower);
	return rc;
}

5408
static int hpsa_send_abort_both_ways(struct ctlr_info *h,
5409
	unsigned char *scsi3addr, struct CommandList *abort, int reply_queue)
5410
{
5411 5412
	/*
	 * ioccelerator mode 2 commands should be aborted via the
5413
	 * accelerated path, since RAID path is unaware of these commands,
5414 5415
	 * but not all underlying firmware can handle abort TMF.
	 * Change abort to physical device reset when abort TMF is unsupported.
5416
	 */
5417 5418 5419 5420 5421 5422
	if (abort->cmd_type == CMD_IOACCEL2) {
		if (HPSATMF_IOACCEL_ENABLED & h->TMFSupportFlags)
			return hpsa_send_abort_ioaccel2(h, abort,
						reply_queue);
		else
			return hpsa_send_reset_as_abort_ioaccel2(h, scsi3addr,
5423
							abort, reply_queue);
5424
	}
S
Stephen Cameron 已提交
5425
	return hpsa_send_abort(h, scsi3addr, abort, reply_queue);
5426
}
5427

5428 5429 5430 5431 5432 5433 5434
/* Find out which reply queue a command was meant to return on */
static int hpsa_extract_reply_queue(struct ctlr_info *h,
					struct CommandList *c)
{
	if (c->cmd_type == CMD_IOACCEL2)
		return h->ioaccel2_cmd_pool[c->cmdindex].reply_queue;
	return c->Header.ReplyQueue;
5435 5436
}

S
Stephen Cameron 已提交
5437 5438 5439 5440 5441 5442 5443 5444 5445 5446 5447 5448
/*
 * Limit concurrency of abort commands to prevent
 * over-subscription of commands
 */
static inline int wait_for_available_abort_cmd(struct ctlr_info *h)
{
#define ABORT_CMD_WAIT_MSECS 5000
	return !wait_event_timeout(h->abort_cmd_wait_queue,
			atomic_dec_if_positive(&h->abort_cmds_available) >= 0,
			msecs_to_jiffies(ABORT_CMD_WAIT_MSECS));
}

5449 5450 5451 5452 5453 5454 5455
/* Send an abort for the specified command.
 *	If the device and controller support it,
 *		send a task abort request.
 */
static int hpsa_eh_abort_handler(struct scsi_cmnd *sc)
{

5456
	int rc;
5457 5458 5459 5460 5461 5462
	struct ctlr_info *h;
	struct hpsa_scsi_dev_t *dev;
	struct CommandList *abort; /* pointer to command to be aborted */
	struct scsi_cmnd *as;	/* ptr to scsi cmd inside aborted command. */
	char msg[256];		/* For debug messaging. */
	int ml = 0;
D
Don Brace 已提交
5463
	__le32 tagupper, taglower;
5464 5465 5466 5467
	int refcount, reply_queue;

	if (sc == NULL)
		return FAILED;
5468

S
Stephen Cameron 已提交
5469 5470 5471
	if (sc->device == NULL)
		return FAILED;

5472 5473
	/* Find the controller of the command to be aborted */
	h = sdev_to_hba(sc->device);
S
Stephen Cameron 已提交
5474
	if (h == NULL)
5475 5476
		return FAILED;

5477 5478 5479 5480 5481
	/* Find the device of the command to be aborted */
	dev = sc->device->hostdata;
	if (!dev) {
		dev_err(&h->pdev->dev, "%s FAILED, Device lookup failed.\n",
				msg);
5482
		return FAILED;
5483 5484 5485 5486 5487 5488 5489 5490 5491 5492 5493 5494 5495 5496 5497
	}

	/* If controller locked up, we can guarantee command won't complete */
	if (lockup_detected(h)) {
		hpsa_show_dev_msg(KERN_WARNING, h, dev,
					"ABORT FAILED, lockup detected");
		return FAILED;
	}

	/* This is a good time to check if controller lockup has occurred */
	if (detect_controller_lockup(h)) {
		hpsa_show_dev_msg(KERN_WARNING, h, dev,
					"ABORT FAILED, new lockup detected");
		return FAILED;
	}
5498

5499 5500 5501 5502 5503 5504
	/* Check that controller supports some kind of task abort */
	if (!(HPSATMF_PHYS_TASK_ABORT & h->TMFSupportFlags) &&
		!(HPSATMF_LOG_TASK_ABORT & h->TMFSupportFlags))
		return FAILED;

	memset(msg, 0, sizeof(msg));
5505
	ml += sprintf(msg+ml, "scsi %d:%d:%d:%llu %s %p",
5506
		h->scsi_host->host_no, sc->device->channel,
5507
		sc->device->id, sc->device->lun,
5508
		"Aborting command", sc);
5509 5510 5511 5512

	/* Get SCSI command to be aborted */
	abort = (struct CommandList *) sc->host_scribble;
	if (abort == NULL) {
5513 5514 5515 5516 5517 5518 5519
		/* This can happen if the command already completed. */
		return SUCCESS;
	}
	refcount = atomic_inc_return(&abort->refcount);
	if (refcount == 1) { /* Command is done already. */
		cmd_free(h, abort);
		return SUCCESS;
5520
	}
S
Stephen Cameron 已提交
5521 5522 5523 5524 5525 5526 5527 5528

	/* Don't bother trying the abort if we know it won't work. */
	if (abort->cmd_type != CMD_IOACCEL2 &&
		abort->cmd_type != CMD_IOACCEL1 && !dev->supports_aborts) {
		cmd_free(h, abort);
		return FAILED;
	}

5529 5530 5531 5532 5533 5534 5535 5536 5537 5538
	/*
	 * Check that we're aborting the right command.
	 * It's possible the CommandList already completed and got re-used.
	 */
	if (abort->scsi_cmd != sc) {
		cmd_free(h, abort);
		return SUCCESS;
	}

	abort->abort_pending = true;
5539
	hpsa_get_tag(h, abort, &taglower, &tagupper);
5540
	reply_queue = hpsa_extract_reply_queue(h, abort);
5541
	ml += sprintf(msg+ml, "Tag:0x%08x:%08x ", tagupper, taglower);
5542
	as  = abort->scsi_cmd;
5543
	if (as != NULL)
5544 5545 5546 5547 5548
		ml += sprintf(msg+ml,
			"CDBLen: %d CDB: 0x%02x%02x... SN: 0x%lx ",
			as->cmd_len, as->cmnd[0], as->cmnd[1],
			as->serial_number);
	dev_warn(&h->pdev->dev, "%s BEING SENT\n", msg);
5549
	hpsa_show_dev_msg(KERN_WARNING, h, dev, "Aborting command");
5550

5551 5552 5553 5554 5555
	/*
	 * Command is in flight, or possibly already completed
	 * by the firmware (but not to the scsi mid layer) but we can't
	 * distinguish which.  Send the abort down.
	 */
S
Stephen Cameron 已提交
5556 5557
	if (wait_for_available_abort_cmd(h)) {
		dev_warn(&h->pdev->dev,
5558 5559
			"%s FAILED, timeout waiting for an abort command to become available.\n",
			msg);
S
Stephen Cameron 已提交
5560 5561 5562
		cmd_free(h, abort);
		return FAILED;
	}
5563
	rc = hpsa_send_abort_both_ways(h, dev->scsi3addr, abort, reply_queue);
S
Stephen Cameron 已提交
5564 5565
	atomic_inc(&h->abort_cmds_available);
	wake_up_all(&h->abort_cmd_wait_queue);
5566
	if (rc != 0) {
5567
		dev_warn(&h->pdev->dev, "%s SENT, FAILED\n", msg);
5568
		hpsa_show_dev_msg(KERN_WARNING, h, dev,
5569
				"FAILED to abort command");
5570
		cmd_free(h, abort);
5571 5572
		return FAILED;
	}
5573
	dev_info(&h->pdev->dev, "%s SENT, SUCCESS\n", msg);
W
Webb Scales 已提交
5574
	wait_event(h->event_sync_wait_queue,
5575
		   abort->scsi_cmd != sc || lockup_detected(h));
5576
	cmd_free(h, abort);
5577
	return !lockup_detected(h) ? SUCCESS : FAILED;
5578 5579
}

5580 5581 5582 5583 5584 5585 5586 5587 5588 5589 5590 5591 5592 5593 5594 5595 5596 5597 5598 5599 5600 5601 5602 5603 5604 5605 5606 5607 5608 5609 5610 5611 5612 5613 5614 5615 5616 5617 5618 5619 5620 5621 5622 5623 5624 5625 5626 5627 5628 5629 5630 5631
/*
 * For operations with an associated SCSI command, a command block is allocated
 * at init, and managed by cmd_tagged_alloc() and cmd_tagged_free() using the
 * block request tag as an index into a table of entries.  cmd_tagged_free() is
 * the complement, although cmd_free() may be called instead.
 */
static struct CommandList *cmd_tagged_alloc(struct ctlr_info *h,
					    struct scsi_cmnd *scmd)
{
	int idx = hpsa_get_cmd_index(scmd);
	struct CommandList *c = h->cmd_pool + idx;

	if (idx < HPSA_NRESERVED_CMDS || idx >= h->nr_cmds) {
		dev_err(&h->pdev->dev, "Bad block tag: %d not in [%d..%d]\n",
			idx, HPSA_NRESERVED_CMDS, h->nr_cmds - 1);
		/* The index value comes from the block layer, so if it's out of
		 * bounds, it's probably not our bug.
		 */
		BUG();
	}

	atomic_inc(&c->refcount);
	if (unlikely(!hpsa_is_cmd_idle(c))) {
		/*
		 * We expect that the SCSI layer will hand us a unique tag
		 * value.  Thus, there should never be a collision here between
		 * two requests...because if the selected command isn't idle
		 * then someone is going to be very disappointed.
		 */
		dev_err(&h->pdev->dev,
			"tag collision (tag=%d) in cmd_tagged_alloc().\n",
			idx);
		if (c->scsi_cmd != NULL)
			scsi_print_command(c->scsi_cmd);
		scsi_print_command(scmd);
	}

	hpsa_cmd_partial_init(h, idx, c);
	return c;
}

static void cmd_tagged_free(struct ctlr_info *h, struct CommandList *c)
{
	/*
	 * Release our reference to the block.  We don't need to do anything
	 * else to free it, because it is accessed by index.  (There's no point
	 * in checking the result of the decrement, since we cannot guarantee
	 * that there isn't a concurrent abort which is also accessing it.)
	 */
	(void)atomic_dec(&c->refcount);
}

5632 5633 5634 5635 5636
/*
 * For operations that cannot sleep, a command block is allocated at init,
 * and managed by cmd_alloc() and cmd_free() using a simple bitmap to track
 * which ones are free or in use.  Lock must be held when calling this.
 * cmd_free() is the complement.
5637 5638
 * This function never gives up and returns NULL.  If it hangs,
 * another thread must call cmd_free() to free some tags.
5639
 */
5640

5641 5642 5643
static struct CommandList *cmd_alloc(struct ctlr_info *h)
{
	struct CommandList *c;
5644
	int refcount, i;
5645
	int offset = 0;
5646

5647 5648
	/*
	 * There is some *extremely* small but non-zero chance that that
5649 5650 5651 5652 5653 5654 5655 5656
	 * multiple threads could get in here, and one thread could
	 * be scanning through the list of bits looking for a free
	 * one, but the free ones are always behind him, and other
	 * threads sneak in behind him and eat them before he can
	 * get to them, so that while there is always a free one, a
	 * very unlucky thread might be starved anyway, never able to
	 * beat the other threads.  In reality, this happens so
	 * infrequently as to be indistinguishable from never.
5657 5658 5659 5660 5661 5662 5663
	 *
	 * Note that we start allocating commands before the SCSI host structure
	 * is initialized.  Since the search starts at bit zero, this
	 * all works, since we have at least one command structure available;
	 * however, it means that the structures with the low indexes have to be
	 * reserved for driver-initiated requests, while requests from the block
	 * layer will use the higher indexes.
5664
	 */
5665

5666
	for (;;) {
5667 5668 5669 5670
		i = find_next_zero_bit(h->cmd_pool_bits,
					HPSA_NRESERVED_CMDS,
					offset);
		if (unlikely(i >= HPSA_NRESERVED_CMDS)) {
5671 5672 5673 5674 5675 5676 5677
			offset = 0;
			continue;
		}
		c = h->cmd_pool + i;
		refcount = atomic_inc_return(&c->refcount);
		if (unlikely(refcount > 1)) {
			cmd_free(h, c); /* already in use */
5678
			offset = (i + 1) % HPSA_NRESERVED_CMDS;
5679 5680 5681 5682 5683 5684
			continue;
		}
		set_bit(i & (BITS_PER_LONG - 1),
			h->cmd_pool_bits + (i / BITS_PER_LONG));
		break; /* it's ours now. */
	}
5685
	hpsa_cmd_partial_init(h, i, c);
5686 5687 5688
	return c;
}

5689 5690 5691 5692 5693 5694
/*
 * This is the complementary operation to cmd_alloc().  Note, however, in some
 * corner cases it may also be used to free blocks allocated by
 * cmd_tagged_alloc() in which case the ref-count decrement does the trick and
 * the clear-bit is harmless.
 */
5695 5696
static void cmd_free(struct ctlr_info *h, struct CommandList *c)
{
5697 5698
	if (atomic_dec_and_test(&c->refcount)) {
		int i;
5699

5700 5701 5702 5703
		i = c - h->cmd_pool;
		clear_bit(i & (BITS_PER_LONG - 1),
			  h->cmd_pool_bits + (i / BITS_PER_LONG));
	}
5704 5705 5706 5707
}

#ifdef CONFIG_COMPAT

D
Don Brace 已提交
5708 5709
static int hpsa_ioctl32_passthru(struct scsi_device *dev, int cmd,
	void __user *arg)
5710 5711 5712 5713 5714 5715 5716 5717
{
	IOCTL32_Command_struct __user *arg32 =
	    (IOCTL32_Command_struct __user *) arg;
	IOCTL_Command_struct arg64;
	IOCTL_Command_struct __user *p = compat_alloc_user_space(sizeof(arg64));
	int err;
	u32 cp;

5718
	memset(&arg64, 0, sizeof(arg64));
5719 5720 5721 5722 5723 5724 5725 5726 5727 5728 5729 5730 5731 5732 5733
	err = 0;
	err |= copy_from_user(&arg64.LUN_info, &arg32->LUN_info,
			   sizeof(arg64.LUN_info));
	err |= copy_from_user(&arg64.Request, &arg32->Request,
			   sizeof(arg64.Request));
	err |= copy_from_user(&arg64.error_info, &arg32->error_info,
			   sizeof(arg64.error_info));
	err |= get_user(arg64.buf_size, &arg32->buf_size);
	err |= get_user(cp, &arg32->buf);
	arg64.buf = compat_ptr(cp);
	err |= copy_to_user(p, &arg64, sizeof(arg64));

	if (err)
		return -EFAULT;

D
Don Brace 已提交
5734
	err = hpsa_ioctl(dev, CCISS_PASSTHRU, p);
5735 5736 5737 5738 5739 5740 5741 5742 5743 5744
	if (err)
		return err;
	err |= copy_in_user(&arg32->error_info, &p->error_info,
			 sizeof(arg32->error_info));
	if (err)
		return -EFAULT;
	return err;
}

static int hpsa_ioctl32_big_passthru(struct scsi_device *dev,
D
Don Brace 已提交
5745
	int cmd, void __user *arg)
5746 5747 5748 5749 5750 5751 5752 5753 5754
{
	BIG_IOCTL32_Command_struct __user *arg32 =
	    (BIG_IOCTL32_Command_struct __user *) arg;
	BIG_IOCTL_Command_struct arg64;
	BIG_IOCTL_Command_struct __user *p =
	    compat_alloc_user_space(sizeof(arg64));
	int err;
	u32 cp;

5755
	memset(&arg64, 0, sizeof(arg64));
5756 5757 5758 5759 5760 5761 5762 5763 5764 5765 5766 5767 5768 5769 5770 5771
	err = 0;
	err |= copy_from_user(&arg64.LUN_info, &arg32->LUN_info,
			   sizeof(arg64.LUN_info));
	err |= copy_from_user(&arg64.Request, &arg32->Request,
			   sizeof(arg64.Request));
	err |= copy_from_user(&arg64.error_info, &arg32->error_info,
			   sizeof(arg64.error_info));
	err |= get_user(arg64.buf_size, &arg32->buf_size);
	err |= get_user(arg64.malloc_size, &arg32->malloc_size);
	err |= get_user(cp, &arg32->buf);
	arg64.buf = compat_ptr(cp);
	err |= copy_to_user(p, &arg64, sizeof(arg64));

	if (err)
		return -EFAULT;

D
Don Brace 已提交
5772
	err = hpsa_ioctl(dev, CCISS_BIG_PASSTHRU, p);
5773 5774 5775 5776 5777 5778 5779 5780
	if (err)
		return err;
	err |= copy_in_user(&arg32->error_info, &p->error_info,
			 sizeof(arg32->error_info));
	if (err)
		return -EFAULT;
	return err;
}
5781

D
Don Brace 已提交
5782
static int hpsa_compat_ioctl(struct scsi_device *dev, int cmd, void __user *arg)
5783 5784 5785 5786 5787 5788 5789 5790 5791 5792 5793 5794 5795 5796 5797 5798 5799 5800 5801 5802 5803 5804 5805 5806 5807 5808 5809 5810
{
	switch (cmd) {
	case CCISS_GETPCIINFO:
	case CCISS_GETINTINFO:
	case CCISS_SETINTINFO:
	case CCISS_GETNODENAME:
	case CCISS_SETNODENAME:
	case CCISS_GETHEARTBEAT:
	case CCISS_GETBUSTYPES:
	case CCISS_GETFIRMVER:
	case CCISS_GETDRIVVER:
	case CCISS_REVALIDVOLS:
	case CCISS_DEREGDISK:
	case CCISS_REGNEWDISK:
	case CCISS_REGNEWD:
	case CCISS_RESCANDISK:
	case CCISS_GETLUNINFO:
		return hpsa_ioctl(dev, cmd, arg);

	case CCISS_PASSTHRU32:
		return hpsa_ioctl32_passthru(dev, cmd, arg);
	case CCISS_BIG_PASSTHRU32:
		return hpsa_ioctl32_big_passthru(dev, cmd, arg);

	default:
		return -ENOIOCTLCMD;
	}
}
5811 5812 5813 5814 5815 5816 5817 5818 5819 5820 5821 5822 5823 5824 5825 5826 5827 5828 5829 5830 5831 5832 5833 5834 5835 5836 5837 5838 5839 5840 5841 5842 5843 5844 5845 5846 5847 5848 5849 5850 5851 5852 5853 5854 5855
#endif

static int hpsa_getpciinfo_ioctl(struct ctlr_info *h, void __user *argp)
{
	struct hpsa_pci_info pciinfo;

	if (!argp)
		return -EINVAL;
	pciinfo.domain = pci_domain_nr(h->pdev->bus);
	pciinfo.bus = h->pdev->bus->number;
	pciinfo.dev_fn = h->pdev->devfn;
	pciinfo.board_id = h->board_id;
	if (copy_to_user(argp, &pciinfo, sizeof(pciinfo)))
		return -EFAULT;
	return 0;
}

static int hpsa_getdrivver_ioctl(struct ctlr_info *h, void __user *argp)
{
	DriverVer_type DriverVer;
	unsigned char vmaj, vmin, vsubmin;
	int rc;

	rc = sscanf(HPSA_DRIVER_VERSION, "%hhu.%hhu.%hhu",
		&vmaj, &vmin, &vsubmin);
	if (rc != 3) {
		dev_info(&h->pdev->dev, "driver version string '%s' "
			"unrecognized.", HPSA_DRIVER_VERSION);
		vmaj = 0;
		vmin = 0;
		vsubmin = 0;
	}
	DriverVer = (vmaj << 16) | (vmin << 8) | vsubmin;
	if (!argp)
		return -EINVAL;
	if (copy_to_user(argp, &DriverVer, sizeof(DriverVer_type)))
		return -EFAULT;
	return 0;
}

static int hpsa_passthru_ioctl(struct ctlr_info *h, void __user *argp)
{
	IOCTL_Command_struct iocommand;
	struct CommandList *c;
	char *buff = NULL;
5856
	u64 temp64;
5857
	int rc = 0;
5858 5859 5860 5861 5862 5863 5864 5865 5866 5867 5868 5869 5870 5871

	if (!argp)
		return -EINVAL;
	if (!capable(CAP_SYS_RAWIO))
		return -EPERM;
	if (copy_from_user(&iocommand, argp, sizeof(iocommand)))
		return -EFAULT;
	if ((iocommand.buf_size < 1) &&
	    (iocommand.Request.Type.Direction != XFER_NONE)) {
		return -EINVAL;
	}
	if (iocommand.buf_size > 0) {
		buff = kmalloc(iocommand.buf_size, GFP_KERNEL);
		if (buff == NULL)
5872
			return -ENOMEM;
5873
		if (iocommand.Request.Type.Direction & XFER_WRITE) {
5874 5875 5876
			/* Copy the data into the buffer we created */
			if (copy_from_user(buff, iocommand.buf,
				iocommand.buf_size)) {
5877 5878
				rc = -EFAULT;
				goto out_kfree;
5879 5880 5881
			}
		} else {
			memset(buff, 0, iocommand.buf_size);
5882
		}
5883
	}
5884
	c = cmd_alloc(h);
5885

5886 5887
	/* Fill in the command type */
	c->cmd_type = CMD_IOCTL_PEND;
5888
	c->scsi_cmd = SCSI_CMD_BUSY;
5889 5890 5891 5892
	/* Fill in Command Header */
	c->Header.ReplyQueue = 0; /* unused in simple mode */
	if (iocommand.buf_size > 0) {	/* buffer to fill */
		c->Header.SGList = 1;
5893
		c->Header.SGTotal = cpu_to_le16(1);
5894 5895
	} else	{ /* no buffers to fill */
		c->Header.SGList = 0;
5896
		c->Header.SGTotal = cpu_to_le16(0);
5897 5898 5899 5900 5901 5902 5903 5904 5905
	}
	memcpy(&c->Header.LUN, &iocommand.LUN_info, sizeof(c->Header.LUN));

	/* Fill in Request block */
	memcpy(&c->Request, &iocommand.Request,
		sizeof(c->Request));

	/* Fill in the scatter gather information */
	if (iocommand.buf_size > 0) {
5906
		temp64 = pci_map_single(h->pdev, buff,
5907
			iocommand.buf_size, PCI_DMA_BIDIRECTIONAL);
5908 5909 5910
		if (dma_mapping_error(&h->pdev->dev, (dma_addr_t) temp64)) {
			c->SG[0].Addr = cpu_to_le64(0);
			c->SG[0].Len = cpu_to_le32(0);
5911 5912 5913
			rc = -ENOMEM;
			goto out;
		}
5914 5915 5916
		c->SG[0].Addr = cpu_to_le64(temp64);
		c->SG[0].Len = cpu_to_le32(iocommand.buf_size);
		c->SG[0].Ext = cpu_to_le32(HPSA_SG_LAST); /* not chaining */
5917
	}
5918
	rc = hpsa_scsi_do_simple_cmd(h, c, DEFAULT_REPLY_QUEUE, NO_TIMEOUT);
5919 5920
	if (iocommand.buf_size > 0)
		hpsa_pci_unmap(h->pdev, c, 1, PCI_DMA_BIDIRECTIONAL);
5921
	check_ioctl_unit_attention(h, c);
5922 5923 5924 5925
	if (rc) {
		rc = -EIO;
		goto out;
	}
5926 5927 5928 5929 5930

	/* Copy the error information out */
	memcpy(&iocommand.error_info, c->err_info,
		sizeof(iocommand.error_info));
	if (copy_to_user(argp, &iocommand, sizeof(iocommand))) {
5931 5932
		rc = -EFAULT;
		goto out;
5933
	}
5934
	if ((iocommand.Request.Type.Direction & XFER_READ) &&
5935
		iocommand.buf_size > 0) {
5936 5937
		/* Copy the data out of the buffer we created */
		if (copy_to_user(iocommand.buf, buff, iocommand.buf_size)) {
5938 5939
			rc = -EFAULT;
			goto out;
5940 5941
		}
	}
5942
out:
5943
	cmd_free(h, c);
5944 5945 5946
out_kfree:
	kfree(buff);
	return rc;
5947 5948 5949 5950 5951 5952 5953 5954
}

static int hpsa_big_passthru_ioctl(struct ctlr_info *h, void __user *argp)
{
	BIG_IOCTL_Command_struct *ioc;
	struct CommandList *c;
	unsigned char **buff = NULL;
	int *buff_size = NULL;
5955
	u64 temp64;
5956 5957
	BYTE sg_used = 0;
	int status = 0;
5958 5959
	u32 left;
	u32 sz;
5960 5961 5962 5963 5964 5965 5966 5967 5968 5969 5970 5971 5972 5973 5974 5975 5976 5977 5978 5979 5980 5981 5982 5983 5984 5985
	BYTE __user *data_ptr;

	if (!argp)
		return -EINVAL;
	if (!capable(CAP_SYS_RAWIO))
		return -EPERM;
	ioc = (BIG_IOCTL_Command_struct *)
	    kmalloc(sizeof(*ioc), GFP_KERNEL);
	if (!ioc) {
		status = -ENOMEM;
		goto cleanup1;
	}
	if (copy_from_user(ioc, argp, sizeof(*ioc))) {
		status = -EFAULT;
		goto cleanup1;
	}
	if ((ioc->buf_size < 1) &&
	    (ioc->Request.Type.Direction != XFER_NONE)) {
		status = -EINVAL;
		goto cleanup1;
	}
	/* Check kmalloc limits  using all SGs */
	if (ioc->malloc_size > MAX_KMALLOC_SIZE) {
		status = -EINVAL;
		goto cleanup1;
	}
5986
	if (ioc->buf_size > ioc->malloc_size * SG_ENTRIES_IN_CMD) {
5987 5988 5989
		status = -EINVAL;
		goto cleanup1;
	}
5990
	buff = kzalloc(SG_ENTRIES_IN_CMD * sizeof(char *), GFP_KERNEL);
5991 5992 5993 5994
	if (!buff) {
		status = -ENOMEM;
		goto cleanup1;
	}
5995
	buff_size = kmalloc(SG_ENTRIES_IN_CMD * sizeof(int), GFP_KERNEL);
5996 5997 5998 5999 6000 6001 6002 6003 6004 6005 6006 6007 6008 6009
	if (!buff_size) {
		status = -ENOMEM;
		goto cleanup1;
	}
	left = ioc->buf_size;
	data_ptr = ioc->buf;
	while (left) {
		sz = (left > ioc->malloc_size) ? ioc->malloc_size : left;
		buff_size[sg_used] = sz;
		buff[sg_used] = kmalloc(sz, GFP_KERNEL);
		if (buff[sg_used] == NULL) {
			status = -ENOMEM;
			goto cleanup1;
		}
6010
		if (ioc->Request.Type.Direction & XFER_WRITE) {
6011
			if (copy_from_user(buff[sg_used], data_ptr, sz)) {
6012
				status = -EFAULT;
6013 6014 6015 6016 6017 6018 6019 6020
				goto cleanup1;
			}
		} else
			memset(buff[sg_used], 0, sz);
		left -= sz;
		data_ptr += sz;
		sg_used++;
	}
6021
	c = cmd_alloc(h);
6022

6023
	c->cmd_type = CMD_IOCTL_PEND;
6024
	c->scsi_cmd = SCSI_CMD_BUSY;
6025
	c->Header.ReplyQueue = 0;
6026 6027
	c->Header.SGList = (u8) sg_used;
	c->Header.SGTotal = cpu_to_le16(sg_used);
6028 6029 6030 6031 6032
	memcpy(&c->Header.LUN, &ioc->LUN_info, sizeof(c->Header.LUN));
	memcpy(&c->Request, &ioc->Request, sizeof(c->Request));
	if (ioc->buf_size > 0) {
		int i;
		for (i = 0; i < sg_used; i++) {
6033
			temp64 = pci_map_single(h->pdev, buff[i],
6034
				    buff_size[i], PCI_DMA_BIDIRECTIONAL);
6035 6036 6037 6038
			if (dma_mapping_error(&h->pdev->dev,
							(dma_addr_t) temp64)) {
				c->SG[i].Addr = cpu_to_le64(0);
				c->SG[i].Len = cpu_to_le32(0);
6039 6040 6041
				hpsa_pci_unmap(h->pdev, c, i,
					PCI_DMA_BIDIRECTIONAL);
				status = -ENOMEM;
6042
				goto cleanup0;
6043
			}
6044 6045 6046
			c->SG[i].Addr = cpu_to_le64(temp64);
			c->SG[i].Len = cpu_to_le32(buff_size[i]);
			c->SG[i].Ext = cpu_to_le32(0);
6047
		}
6048
		c->SG[--i].Ext = cpu_to_le32(HPSA_SG_LAST);
6049
	}
6050
	status = hpsa_scsi_do_simple_cmd(h, c, DEFAULT_REPLY_QUEUE, NO_TIMEOUT);
6051 6052
	if (sg_used)
		hpsa_pci_unmap(h->pdev, c, sg_used, PCI_DMA_BIDIRECTIONAL);
6053
	check_ioctl_unit_attention(h, c);
6054 6055 6056 6057 6058
	if (status) {
		status = -EIO;
		goto cleanup0;
	}

6059 6060 6061 6062
	/* Copy the error information out */
	memcpy(&ioc->error_info, c->err_info, sizeof(ioc->error_info));
	if (copy_to_user(argp, ioc, sizeof(*ioc))) {
		status = -EFAULT;
6063
		goto cleanup0;
6064
	}
6065
	if ((ioc->Request.Type.Direction & XFER_READ) && ioc->buf_size > 0) {
D
Don Brace 已提交
6066 6067
		int i;

6068 6069 6070 6071 6072
		/* Copy the data out of the buffer we created */
		BYTE __user *ptr = ioc->buf;
		for (i = 0; i < sg_used; i++) {
			if (copy_to_user(ptr, buff[i], buff_size[i])) {
				status = -EFAULT;
6073
				goto cleanup0;
6074 6075 6076 6077 6078
			}
			ptr += buff_size[i];
		}
	}
	status = 0;
6079
cleanup0:
6080
	cmd_free(h, c);
6081 6082
cleanup1:
	if (buff) {
D
Don Brace 已提交
6083 6084
		int i;

6085 6086 6087 6088 6089 6090 6091 6092 6093 6094 6095 6096 6097 6098 6099 6100
		for (i = 0; i < sg_used; i++)
			kfree(buff[i]);
		kfree(buff);
	}
	kfree(buff_size);
	kfree(ioc);
	return status;
}

static void check_ioctl_unit_attention(struct ctlr_info *h,
	struct CommandList *c)
{
	if (c->err_info->CommandStatus == CMD_TARGET_STATUS &&
			c->err_info->ScsiStatus != SAM_STAT_CHECK_CONDITION)
		(void) check_for_unit_attention(h, c);
}
6101

6102 6103 6104
/*
 * ioctl
 */
D
Don Brace 已提交
6105
static int hpsa_ioctl(struct scsi_device *dev, int cmd, void __user *arg)
6106 6107 6108
{
	struct ctlr_info *h;
	void __user *argp = (void __user *)arg;
6109
	int rc;
6110 6111 6112 6113 6114 6115 6116

	h = sdev_to_hba(dev);

	switch (cmd) {
	case CCISS_DEREGDISK:
	case CCISS_REGNEWDISK:
	case CCISS_REGNEWD:
6117
		hpsa_scan_start(h->scsi_host);
6118 6119 6120 6121 6122 6123
		return 0;
	case CCISS_GETPCIINFO:
		return hpsa_getpciinfo_ioctl(h, argp);
	case CCISS_GETDRIVVER:
		return hpsa_getdrivver_ioctl(h, argp);
	case CCISS_PASSTHRU:
6124
		if (atomic_dec_if_positive(&h->passthru_cmds_avail) < 0)
6125 6126
			return -EAGAIN;
		rc = hpsa_passthru_ioctl(h, argp);
6127
		atomic_inc(&h->passthru_cmds_avail);
6128
		return rc;
6129
	case CCISS_BIG_PASSTHRU:
6130
		if (atomic_dec_if_positive(&h->passthru_cmds_avail) < 0)
6131 6132
			return -EAGAIN;
		rc = hpsa_big_passthru_ioctl(h, argp);
6133
		atomic_inc(&h->passthru_cmds_avail);
6134
		return rc;
6135 6136 6137 6138 6139
	default:
		return -ENOTTY;
	}
}

6140
static void hpsa_send_host_reset(struct ctlr_info *h, unsigned char *scsi3addr,
6141
				u8 reset_type)
6142 6143 6144 6145
{
	struct CommandList *c;

	c = cmd_alloc(h);
6146

6147 6148
	/* fill_cmd can't fail here, no data buffer to map */
	(void) fill_cmd(c, HPSA_DEVICE_RESET_MSG, h, NULL, 0, 0,
6149 6150 6151 6152 6153 6154 6155 6156
		RAID_CTLR_LUNID, TYPE_MSG);
	c->Request.CDB[1] = reset_type; /* fill_cmd defaults to target reset */
	c->waiting = NULL;
	enqueue_cmd_and_start_io(h, c);
	/* Don't wait for completion, the reset won't complete.  Don't free
	 * the command either.  This is the last command we will send before
	 * re-initializing everything, so it doesn't matter and won't leak.
	 */
6157
	return;
6158 6159
}

6160
static int fill_cmd(struct CommandList *c, u8 cmd, struct ctlr_info *h,
6161
	void *buff, size_t size, u16 page_code, unsigned char *scsi3addr,
6162 6163 6164
	int cmd_type)
{
	int pci_dir = XFER_NONE;
S
Stephen Cameron 已提交
6165
	u64 tag; /* for commands to be aborted */
6166 6167

	c->cmd_type = CMD_IOCTL_PEND;
6168
	c->scsi_cmd = SCSI_CMD_BUSY;
6169 6170 6171
	c->Header.ReplyQueue = 0;
	if (buff != NULL && size > 0) {
		c->Header.SGList = 1;
6172
		c->Header.SGTotal = cpu_to_le16(1);
6173 6174
	} else {
		c->Header.SGList = 0;
6175
		c->Header.SGTotal = cpu_to_le16(0);
6176 6177 6178 6179 6180 6181 6182
	}
	memcpy(c->Header.LUN.LunAddrBytes, scsi3addr, 8);

	if (cmd_type == TYPE_CMD) {
		switch (cmd) {
		case HPSA_INQUIRY:
			/* are we trying to read a vital product page */
6183
			if (page_code & VPD_PAGE) {
6184
				c->Request.CDB[1] = 0x01;
6185
				c->Request.CDB[2] = (page_code & 0xff);
6186 6187
			}
			c->Request.CDBLen = 6;
6188 6189
			c->Request.type_attr_dir =
				TYPE_ATTR_DIR(cmd_type, ATTR_SIMPLE, XFER_READ);
6190 6191 6192 6193 6194 6195 6196 6197 6198 6199
			c->Request.Timeout = 0;
			c->Request.CDB[0] = HPSA_INQUIRY;
			c->Request.CDB[4] = size & 0xFF;
			break;
		case HPSA_REPORT_LOG:
		case HPSA_REPORT_PHYS:
			/* Talking to controller so It's a physical command
			   mode = 00 target = 0.  Nothing to write.
			 */
			c->Request.CDBLen = 12;
6200 6201
			c->Request.type_attr_dir =
				TYPE_ATTR_DIR(cmd_type, ATTR_SIMPLE, XFER_READ);
6202 6203 6204 6205 6206 6207 6208 6209 6210
			c->Request.Timeout = 0;
			c->Request.CDB[0] = cmd;
			c->Request.CDB[6] = (size >> 24) & 0xFF; /* MSB */
			c->Request.CDB[7] = (size >> 16) & 0xFF;
			c->Request.CDB[8] = (size >> 8) & 0xFF;
			c->Request.CDB[9] = size & 0xFF;
			break;
		case HPSA_CACHE_FLUSH:
			c->Request.CDBLen = 12;
6211 6212 6213
			c->Request.type_attr_dir =
					TYPE_ATTR_DIR(cmd_type,
						ATTR_SIMPLE, XFER_WRITE);
6214 6215 6216
			c->Request.Timeout = 0;
			c->Request.CDB[0] = BMIC_WRITE;
			c->Request.CDB[6] = BMIC_CACHE_FLUSH;
6217 6218
			c->Request.CDB[7] = (size >> 8) & 0xFF;
			c->Request.CDB[8] = size & 0xFF;
6219 6220 6221
			break;
		case TEST_UNIT_READY:
			c->Request.CDBLen = 6;
6222 6223
			c->Request.type_attr_dir =
				TYPE_ATTR_DIR(cmd_type, ATTR_SIMPLE, XFER_NONE);
6224 6225
			c->Request.Timeout = 0;
			break;
6226 6227
		case HPSA_GET_RAID_MAP:
			c->Request.CDBLen = 12;
6228 6229
			c->Request.type_attr_dir =
				TYPE_ATTR_DIR(cmd_type, ATTR_SIMPLE, XFER_READ);
6230 6231 6232 6233 6234 6235 6236 6237
			c->Request.Timeout = 0;
			c->Request.CDB[0] = HPSA_CISS_READ;
			c->Request.CDB[1] = cmd;
			c->Request.CDB[6] = (size >> 24) & 0xFF; /* MSB */
			c->Request.CDB[7] = (size >> 16) & 0xFF;
			c->Request.CDB[8] = (size >> 8) & 0xFF;
			c->Request.CDB[9] = size & 0xFF;
			break;
6238 6239
		case BMIC_SENSE_CONTROLLER_PARAMETERS:
			c->Request.CDBLen = 10;
6240 6241
			c->Request.type_attr_dir =
				TYPE_ATTR_DIR(cmd_type, ATTR_SIMPLE, XFER_READ);
6242 6243 6244 6245 6246 6247
			c->Request.Timeout = 0;
			c->Request.CDB[0] = BMIC_READ;
			c->Request.CDB[6] = BMIC_SENSE_CONTROLLER_PARAMETERS;
			c->Request.CDB[7] = (size >> 16) & 0xFF;
			c->Request.CDB[8] = (size >> 8) & 0xFF;
			break;
6248 6249 6250 6251 6252 6253 6254 6255 6256 6257
		case BMIC_IDENTIFY_PHYSICAL_DEVICE:
			c->Request.CDBLen = 10;
			c->Request.type_attr_dir =
				TYPE_ATTR_DIR(cmd_type, ATTR_SIMPLE, XFER_READ);
			c->Request.Timeout = 0;
			c->Request.CDB[0] = BMIC_READ;
			c->Request.CDB[6] = BMIC_IDENTIFY_PHYSICAL_DEVICE;
			c->Request.CDB[7] = (size >> 16) & 0xFF;
			c->Request.CDB[8] = (size >> 8) & 0XFF;
			break;
6258 6259 6260
		default:
			dev_warn(&h->pdev->dev, "unknown command 0x%c\n", cmd);
			BUG();
6261
			return -1;
6262 6263 6264 6265 6266 6267
		}
	} else if (cmd_type == TYPE_MSG) {
		switch (cmd) {

		case  HPSA_DEVICE_RESET_MSG:
			c->Request.CDBLen = 16;
6268 6269
			c->Request.type_attr_dir =
				TYPE_ATTR_DIR(cmd_type, ATTR_SIMPLE, XFER_NONE);
6270
			c->Request.Timeout = 0; /* Don't time out */
6271 6272
			memset(&c->Request.CDB[0], 0, sizeof(c->Request.CDB));
			c->Request.CDB[0] =  cmd;
6273
			c->Request.CDB[1] = HPSA_RESET_TYPE_LUN;
6274 6275 6276 6277 6278 6279
			/* If bytes 4-7 are zero, it means reset the */
			/* LunID device */
			c->Request.CDB[4] = 0x00;
			c->Request.CDB[5] = 0x00;
			c->Request.CDB[6] = 0x00;
			c->Request.CDB[7] = 0x00;
6280 6281
			break;
		case  HPSA_ABORT_MSG:
S
Stephen Cameron 已提交
6282
			memcpy(&tag, buff, sizeof(tag));
D
Don Brace 已提交
6283
			dev_dbg(&h->pdev->dev,
S
Stephen Cameron 已提交
6284 6285
				"Abort Tag:0x%016llx using rqst Tag:0x%016llx",
				tag, c->Header.tag);
6286
			c->Request.CDBLen = 16;
6287 6288 6289
			c->Request.type_attr_dir =
					TYPE_ATTR_DIR(cmd_type,
						ATTR_SIMPLE, XFER_WRITE);
6290 6291 6292 6293 6294 6295
			c->Request.Timeout = 0; /* Don't time out */
			c->Request.CDB[0] = HPSA_TASK_MANAGEMENT;
			c->Request.CDB[1] = HPSA_TMF_ABORT_TASK;
			c->Request.CDB[2] = 0x00; /* reserved */
			c->Request.CDB[3] = 0x00; /* reserved */
			/* Tag to abort goes in CDB[4]-CDB[11] */
S
Stephen Cameron 已提交
6296
			memcpy(&c->Request.CDB[4], &tag, sizeof(tag));
6297 6298 6299 6300
			c->Request.CDB[12] = 0x00; /* reserved */
			c->Request.CDB[13] = 0x00; /* reserved */
			c->Request.CDB[14] = 0x00; /* reserved */
			c->Request.CDB[15] = 0x00; /* reserved */
6301 6302 6303 6304 6305 6306 6307 6308 6309 6310 6311
		break;
		default:
			dev_warn(&h->pdev->dev, "unknown message type %d\n",
				cmd);
			BUG();
		}
	} else {
		dev_warn(&h->pdev->dev, "unknown command type %d\n", cmd_type);
		BUG();
	}

6312
	switch (GET_DIR(c->Request.type_attr_dir)) {
6313 6314 6315 6316 6317 6318 6319 6320 6321 6322 6323 6324
	case XFER_READ:
		pci_dir = PCI_DMA_FROMDEVICE;
		break;
	case XFER_WRITE:
		pci_dir = PCI_DMA_TODEVICE;
		break;
	case XFER_NONE:
		pci_dir = PCI_DMA_NONE;
		break;
	default:
		pci_dir = PCI_DMA_BIDIRECTIONAL;
	}
6325 6326 6327
	if (hpsa_map_one(h->pdev, c, buff, size, pci_dir))
		return -1;
	return 0;
6328 6329 6330 6331 6332 6333 6334 6335 6336
}

/*
 * Map (physical) PCI mem into (virtual) kernel space
 */
static void __iomem *remap_pci_mem(ulong base, ulong size)
{
	ulong page_base = ((ulong) base) & PAGE_MASK;
	ulong page_offs = ((ulong) base) - page_base;
6337 6338
	void __iomem *page_remapped = ioremap_nocache(page_base,
		page_offs + size);
6339 6340 6341 6342

	return page_remapped ? (page_remapped + page_offs) : NULL;
}

6343
static inline unsigned long get_next_completion(struct ctlr_info *h, u8 q)
6344
{
6345
	return h->access.command_completed(h, q);
6346 6347
}

6348
static inline bool interrupt_pending(struct ctlr_info *h)
6349 6350 6351 6352 6353 6354
{
	return h->access.intr_pending(h);
}

static inline long interrupt_not_for_us(struct ctlr_info *h)
{
6355 6356
	return (h->access.intr_pending(h) == 0) ||
		(h->interrupts_enabled == 0);
6357 6358
}

6359 6360
static inline int bad_tag(struct ctlr_info *h, u32 tag_index,
	u32 raw_tag)
6361 6362 6363 6364 6365 6366 6367 6368
{
	if (unlikely(tag_index >= h->nr_cmds)) {
		dev_warn(&h->pdev->dev, "bad tag 0x%08x ignored.\n", raw_tag);
		return 1;
	}
	return 0;
}

6369
static inline void finish_cmd(struct CommandList *c)
6370
{
6371
	dial_up_lockup_detection_on_fw_flash_complete(c->h, c);
6372 6373
	if (likely(c->cmd_type == CMD_IOACCEL1 || c->cmd_type == CMD_SCSI
			|| c->cmd_type == CMD_IOACCEL2))
6374
		complete_scsi_command(c);
6375
	else if (c->cmd_type == CMD_IOCTL_PEND || c->cmd_type == IOACCEL2_TMF)
6376
		complete(c->waiting);
6377 6378
}

6379 6380

static inline u32 hpsa_tag_discard_error_bits(struct ctlr_info *h, u32 tag)
6381
{
6382 6383
#define HPSA_PERF_ERROR_BITS ((1 << DIRECT_LOOKUP_SHIFT) - 1)
#define HPSA_SIMPLE_ERROR_BITS 0x03
6384
	if (unlikely(!(h->transMethod & CFGTBL_Trans_Performant)))
6385 6386
		return tag & ~HPSA_SIMPLE_ERROR_BITS;
	return tag & ~HPSA_PERF_ERROR_BITS;
6387 6388
}

6389
/* process completion of an indexed ("direct lookup") command */
6390
static inline void process_indexed_cmd(struct ctlr_info *h,
6391 6392 6393 6394 6395
	u32 raw_tag)
{
	u32 tag_index;
	struct CommandList *c;

6396
	tag_index = raw_tag >> DIRECT_LOOKUP_SHIFT;
6397 6398 6399 6400
	if (!bad_tag(h, tag_index, raw_tag)) {
		c = h->cmd_pool + tag_index;
		finish_cmd(c);
	}
6401 6402
}

6403 6404 6405 6406 6407 6408 6409 6410 6411 6412 6413 6414 6415 6416 6417 6418 6419 6420 6421
/* Some controllers, like p400, will give us one interrupt
 * after a soft reset, even if we turned interrupts off.
 * Only need to check for this in the hpsa_xxx_discard_completions
 * functions.
 */
static int ignore_bogus_interrupt(struct ctlr_info *h)
{
	if (likely(!reset_devices))
		return 0;

	if (likely(h->interrupts_enabled))
		return 0;

	dev_info(&h->pdev->dev, "Received interrupt while interrupts disabled "
		"(known firmware bug.)  Ignoring.\n");

	return 1;
}

6422 6423 6424 6425 6426 6427
/*
 * Convert &h->q[x] (passed to interrupt handlers) back to h.
 * Relies on (h-q[x] == x) being true for x such that
 * 0 <= x < MAX_REPLY_QUEUES.
 */
static struct ctlr_info *queue_to_hba(u8 *queue)
6428
{
6429 6430 6431 6432 6433 6434 6435
	return container_of((queue - *queue), struct ctlr_info, q[0]);
}

static irqreturn_t hpsa_intx_discard_completions(int irq, void *queue)
{
	struct ctlr_info *h = queue_to_hba(queue);
	u8 q = *(u8 *) queue;
6436 6437 6438 6439 6440 6441 6442
	u32 raw_tag;

	if (ignore_bogus_interrupt(h))
		return IRQ_NONE;

	if (interrupt_not_for_us(h))
		return IRQ_NONE;
6443
	h->last_intr_timestamp = get_jiffies_64();
6444
	while (interrupt_pending(h)) {
6445
		raw_tag = get_next_completion(h, q);
6446
		while (raw_tag != FIFO_EMPTY)
6447
			raw_tag = next_command(h, q);
6448 6449 6450 6451
	}
	return IRQ_HANDLED;
}

6452
static irqreturn_t hpsa_msix_discard_completions(int irq, void *queue)
6453
{
6454
	struct ctlr_info *h = queue_to_hba(queue);
6455
	u32 raw_tag;
6456
	u8 q = *(u8 *) queue;
6457 6458 6459 6460

	if (ignore_bogus_interrupt(h))
		return IRQ_NONE;

6461
	h->last_intr_timestamp = get_jiffies_64();
6462
	raw_tag = get_next_completion(h, q);
6463
	while (raw_tag != FIFO_EMPTY)
6464
		raw_tag = next_command(h, q);
6465 6466 6467
	return IRQ_HANDLED;
}

6468
static irqreturn_t do_hpsa_intr_intx(int irq, void *queue)
6469
{
6470
	struct ctlr_info *h = queue_to_hba((u8 *) queue);
6471
	u32 raw_tag;
6472
	u8 q = *(u8 *) queue;
6473 6474 6475

	if (interrupt_not_for_us(h))
		return IRQ_NONE;
6476
	h->last_intr_timestamp = get_jiffies_64();
6477
	while (interrupt_pending(h)) {
6478
		raw_tag = get_next_completion(h, q);
6479
		while (raw_tag != FIFO_EMPTY) {
6480
			process_indexed_cmd(h, raw_tag);
6481
			raw_tag = next_command(h, q);
6482 6483 6484 6485 6486
		}
	}
	return IRQ_HANDLED;
}

6487
static irqreturn_t do_hpsa_intr_msi(int irq, void *queue)
6488
{
6489
	struct ctlr_info *h = queue_to_hba(queue);
6490
	u32 raw_tag;
6491
	u8 q = *(u8 *) queue;
6492

6493
	h->last_intr_timestamp = get_jiffies_64();
6494
	raw_tag = get_next_completion(h, q);
6495
	while (raw_tag != FIFO_EMPTY) {
6496
		process_indexed_cmd(h, raw_tag);
6497
		raw_tag = next_command(h, q);
6498 6499 6500 6501
	}
	return IRQ_HANDLED;
}

6502 6503 6504 6505
/* Send a message CDB to the firmware. Careful, this only works
 * in simple mode, not performant mode due to the tag lookup.
 * We only ever use this immediately after a controller reset.
 */
6506 6507
static int hpsa_message(struct pci_dev *pdev, unsigned char opcode,
			unsigned char type)
6508 6509 6510 6511 6512 6513 6514 6515 6516 6517
{
	struct Command {
		struct CommandListHeader CommandHeader;
		struct RequestBlock Request;
		struct ErrDescriptor ErrorDescriptor;
	};
	struct Command *cmd;
	static const size_t cmd_sz = sizeof(*cmd) +
					sizeof(cmd->ErrorDescriptor);
	dma_addr_t paddr64;
D
Don Brace 已提交
6518 6519
	__le32 paddr32;
	u32 tag;
6520 6521 6522 6523 6524 6525 6526 6527 6528 6529 6530 6531 6532 6533
	void __iomem *vaddr;
	int i, err;

	vaddr = pci_ioremap_bar(pdev, 0);
	if (vaddr == NULL)
		return -ENOMEM;

	/* The Inbound Post Queue only accepts 32-bit physical addresses for the
	 * CCISS commands, so they must be allocated from the lower 4GiB of
	 * memory.
	 */
	err = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(32));
	if (err) {
		iounmap(vaddr);
6534
		return err;
6535 6536 6537 6538 6539 6540 6541 6542 6543 6544 6545 6546
	}

	cmd = pci_alloc_consistent(pdev, cmd_sz, &paddr64);
	if (cmd == NULL) {
		iounmap(vaddr);
		return -ENOMEM;
	}

	/* This must fit, because of the 32-bit consistent DMA mask.  Also,
	 * although there's no guarantee, we assume that the address is at
	 * least 4-byte aligned (most likely, it's page-aligned).
	 */
D
Don Brace 已提交
6547
	paddr32 = cpu_to_le32(paddr64);
6548 6549 6550

	cmd->CommandHeader.ReplyQueue = 0;
	cmd->CommandHeader.SGList = 0;
6551
	cmd->CommandHeader.SGTotal = cpu_to_le16(0);
D
Don Brace 已提交
6552
	cmd->CommandHeader.tag = cpu_to_le64(paddr64);
6553 6554 6555
	memset(&cmd->CommandHeader.LUN.LunAddrBytes, 0, 8);

	cmd->Request.CDBLen = 16;
6556 6557
	cmd->Request.type_attr_dir =
			TYPE_ATTR_DIR(TYPE_MSG, ATTR_HEADOFQUEUE, XFER_NONE);
6558 6559 6560 6561
	cmd->Request.Timeout = 0; /* Don't time out */
	cmd->Request.CDB[0] = opcode;
	cmd->Request.CDB[1] = type;
	memset(&cmd->Request.CDB[2], 0, 14); /* rest of the CDB is reserved */
6562
	cmd->ErrorDescriptor.Addr =
D
Don Brace 已提交
6563
			cpu_to_le64((le32_to_cpu(paddr32) + sizeof(*cmd)));
6564
	cmd->ErrorDescriptor.Len = cpu_to_le32(sizeof(struct ErrorInfo));
6565

D
Don Brace 已提交
6566
	writel(le32_to_cpu(paddr32), vaddr + SA5_REQUEST_PORT_OFFSET);
6567 6568 6569

	for (i = 0; i < HPSA_MSG_SEND_RETRY_LIMIT; i++) {
		tag = readl(vaddr + SA5_REPLY_PORT_OFFSET);
D
Don Brace 已提交
6570
		if ((tag & ~HPSA_SIMPLE_ERROR_BITS) == paddr64)
6571 6572 6573 6574 6575 6576 6577 6578 6579 6580 6581 6582 6583 6584 6585 6586 6587 6588 6589 6590 6591 6592 6593 6594 6595 6596 6597 6598 6599 6600
			break;
		msleep(HPSA_MSG_SEND_RETRY_INTERVAL_MSECS);
	}

	iounmap(vaddr);

	/* we leak the DMA buffer here ... no choice since the controller could
	 *  still complete the command.
	 */
	if (i == HPSA_MSG_SEND_RETRY_LIMIT) {
		dev_err(&pdev->dev, "controller message %02x:%02x timed out\n",
			opcode, type);
		return -ETIMEDOUT;
	}

	pci_free_consistent(pdev, cmd_sz, cmd, paddr64);

	if (tag & HPSA_ERROR_BIT) {
		dev_err(&pdev->dev, "controller message %02x:%02x failed\n",
			opcode, type);
		return -EIO;
	}

	dev_info(&pdev->dev, "controller message %02x:%02x succeeded\n",
		opcode, type);
	return 0;
}

#define hpsa_noop(p) hpsa_message(p, 3, 0)

6601
static int hpsa_controller_hard_reset(struct pci_dev *pdev,
D
Don Brace 已提交
6602
	void __iomem *vaddr, u32 use_doorbell)
6603 6604 6605 6606 6607 6608 6609 6610
{

	if (use_doorbell) {
		/* For everything after the P600, the PCI power state method
		 * of resetting the controller doesn't work, so we have this
		 * other way using the doorbell register.
		 */
		dev_info(&pdev->dev, "using doorbell to reset controller\n");
6611
		writel(use_doorbell, vaddr + SA5_DOORBELL);
6612

6613
		/* PMC hardware guys tell us we need a 10 second delay after
6614 6615 6616 6617
		 * doorbell reset and before any attempt to talk to the board
		 * at all to ensure that this actually works and doesn't fall
		 * over in some weird corner cases.
		 */
6618
		msleep(10000);
6619 6620 6621 6622 6623 6624 6625 6626 6627
	} else { /* Try to do it the PCI power state way */

		/* Quoting from the Open CISS Specification: "The Power
		 * Management Control/Status Register (CSR) controls the power
		 * state of the device.  The normal operating state is D0,
		 * CSR=00h.  The software off state is D3, CSR=03h.  To reset
		 * the controller, place the interface device in D3 then to D0,
		 * this causes a secondary PCI reset which will reset the
		 * controller." */
6628 6629 6630

		int rc = 0;

6631
		dev_info(&pdev->dev, "using PCI PM to reset controller\n");
6632

6633
		/* enter the D3hot power management state */
6634 6635 6636
		rc = pci_set_power_state(pdev, PCI_D3hot);
		if (rc)
			return rc;
6637 6638 6639 6640

		msleep(500);

		/* enter the D0 power management state */
6641 6642 6643
		rc = pci_set_power_state(pdev, PCI_D0);
		if (rc)
			return rc;
6644 6645 6646 6647 6648 6649 6650

		/*
		 * The P600 requires a small delay when changing states.
		 * Otherwise we may think the board did not reset and we bail.
		 * This for kdump only and is particular to the P600.
		 */
		msleep(500);
6651 6652 6653 6654
	}
	return 0;
}

6655
static void init_driver_version(char *driver_version, int len)
6656 6657
{
	memset(driver_version, 0, len);
6658
	strncpy(driver_version, HPSA " " HPSA_DRIVER_VERSION, len - 1);
6659 6660
}

6661
static int write_driver_ver_to_cfgtable(struct CfgTable __iomem *cfgtable)
6662 6663 6664 6665 6666 6667 6668 6669 6670 6671 6672 6673 6674 6675 6676
{
	char *driver_version;
	int i, size = sizeof(cfgtable->driver_version);

	driver_version = kmalloc(size, GFP_KERNEL);
	if (!driver_version)
		return -ENOMEM;

	init_driver_version(driver_version, size);
	for (i = 0; i < size; i++)
		writeb(driver_version[i], &cfgtable->driver_version[i]);
	kfree(driver_version);
	return 0;
}

6677 6678
static void read_driver_ver_from_cfgtable(struct CfgTable __iomem *cfgtable,
					  unsigned char *driver_ver)
6679 6680 6681 6682 6683 6684 6685
{
	int i;

	for (i = 0; i < sizeof(cfgtable->driver_version); i++)
		driver_ver[i] = readb(&cfgtable->driver_version[i]);
}

6686
static int controller_reset_failed(struct CfgTable __iomem *cfgtable)
6687 6688 6689 6690 6691 6692 6693 6694 6695 6696 6697 6698 6699 6700 6701 6702 6703 6704 6705
{

	char *driver_ver, *old_driver_ver;
	int rc, size = sizeof(cfgtable->driver_version);

	old_driver_ver = kmalloc(2 * size, GFP_KERNEL);
	if (!old_driver_ver)
		return -ENOMEM;
	driver_ver = old_driver_ver + size;

	/* After a reset, the 32 bytes of "driver version" in the cfgtable
	 * should have been changed, otherwise we know the reset failed.
	 */
	init_driver_version(old_driver_ver, size);
	read_driver_ver_from_cfgtable(cfgtable, driver_ver);
	rc = !memcmp(driver_ver, old_driver_ver, size);
	kfree(old_driver_ver);
	return rc;
}
6706
/* This does a hard reset of the controller using PCI power management
6707
 * states or the using the doorbell register.
6708
 */
6709
static int hpsa_kdump_hard_reset_controller(struct pci_dev *pdev, u32 board_id)
6710
{
6711 6712 6713 6714 6715
	u64 cfg_offset;
	u32 cfg_base_addr;
	u64 cfg_base_addr_index;
	void __iomem *vaddr;
	unsigned long paddr;
6716
	u32 misc_fw_support;
6717
	int rc;
6718
	struct CfgTable __iomem *cfgtable;
6719
	u32 use_doorbell;
6720
	u16 command_register;
6721

6722 6723
	/* For controllers as old as the P600, this is very nearly
	 * the same thing as
6724 6725 6726 6727 6728 6729
	 *
	 * pci_save_state(pci_dev);
	 * pci_set_power_state(pci_dev, PCI_D3hot);
	 * pci_set_power_state(pci_dev, PCI_D0);
	 * pci_restore_state(pci_dev);
	 *
6730 6731 6732
	 * For controllers newer than the P600, the pci power state
	 * method of resetting doesn't work so we have another way
	 * using the doorbell register.
6733
	 */
6734

6735 6736
	if (!ctlr_is_resettable(board_id)) {
		dev_warn(&pdev->dev, "Controller not resettable\n");
6737 6738
		return -ENODEV;
	}
6739 6740 6741 6742

	/* if controller is soft- but not hard resettable... */
	if (!ctlr_is_hard_resettable(board_id))
		return -ENOTSUPP; /* try soft reset later. */
6743

6744 6745 6746
	/* Save the PCI command register */
	pci_read_config_word(pdev, 4, &command_register);
	pci_save_state(pdev);
6747

6748 6749 6750 6751 6752 6753 6754
	/* find the first memory BAR, so we can find the cfg table */
	rc = hpsa_pci_find_memory_BAR(pdev, &paddr);
	if (rc)
		return rc;
	vaddr = remap_pci_mem(paddr, 0x250);
	if (!vaddr)
		return -ENOMEM;
6755

6756 6757 6758 6759 6760 6761 6762 6763 6764 6765 6766
	/* find cfgtable in order to check if reset via doorbell is supported */
	rc = hpsa_find_cfg_addrs(pdev, vaddr, &cfg_base_addr,
					&cfg_base_addr_index, &cfg_offset);
	if (rc)
		goto unmap_vaddr;
	cfgtable = remap_pci_mem(pci_resource_start(pdev,
		       cfg_base_addr_index) + cfg_offset, sizeof(*cfgtable));
	if (!cfgtable) {
		rc = -ENOMEM;
		goto unmap_vaddr;
	}
6767 6768
	rc = write_driver_ver_to_cfgtable(cfgtable);
	if (rc)
6769
		goto unmap_cfgtable;
6770

6771 6772 6773
	/* If reset via doorbell register is supported, use that.
	 * There are two such methods.  Favor the newest method.
	 */
6774
	misc_fw_support = readl(&cfgtable->misc_fw_support);
6775 6776 6777 6778 6779 6780
	use_doorbell = misc_fw_support & MISC_FW_DOORBELL_RESET2;
	if (use_doorbell) {
		use_doorbell = DOORBELL_CTLR_RESET2;
	} else {
		use_doorbell = misc_fw_support & MISC_FW_DOORBELL_RESET;
		if (use_doorbell) {
6781 6782
			dev_warn(&pdev->dev,
				"Soft reset not supported. Firmware update is required.\n");
6783
			rc = -ENOTSUPP; /* try soft reset */
6784 6785 6786
			goto unmap_cfgtable;
		}
	}
6787

6788 6789 6790
	rc = hpsa_controller_hard_reset(pdev, vaddr, use_doorbell);
	if (rc)
		goto unmap_cfgtable;
6791

6792 6793
	pci_restore_state(pdev);
	pci_write_config_word(pdev, 4, command_register);
6794

6795 6796 6797 6798
	/* Some devices (notably the HP Smart Array 5i Controller)
	   need a little pause here */
	msleep(HPSA_POST_RESET_PAUSE_MSECS);

6799 6800 6801
	rc = hpsa_wait_for_board_state(pdev, vaddr, BOARD_READY);
	if (rc) {
		dev_warn(&pdev->dev,
6802
			"Failed waiting for board to become ready after hard reset\n");
6803 6804 6805
		goto unmap_cfgtable;
	}

6806 6807 6808 6809
	rc = controller_reset_failed(vaddr);
	if (rc < 0)
		goto unmap_cfgtable;
	if (rc) {
6810 6811 6812
		dev_warn(&pdev->dev, "Unable to successfully reset "
			"controller. Will try soft reset.\n");
		rc = -ENOTSUPP;
6813
	} else {
6814
		dev_info(&pdev->dev, "board ready after hard reset.\n");
6815 6816 6817 6818 6819 6820 6821 6822
	}

unmap_cfgtable:
	iounmap(cfgtable);

unmap_vaddr:
	iounmap(vaddr);
	return rc;
6823 6824 6825 6826 6827 6828 6829
}

/*
 *  We cannot read the structure directly, for portability we must use
 *   the io functions.
 *   This is for debug only.
 */
D
Don Brace 已提交
6830
static void print_cfg_table(struct device *dev, struct CfgTable __iomem *tb)
6831
{
6832
#ifdef HPSA_DEBUG
6833 6834 6835 6836 6837 6838 6839 6840 6841 6842 6843 6844 6845 6846 6847 6848 6849 6850 6851 6852
	int i;
	char temp_name[17];

	dev_info(dev, "Controller Configuration information\n");
	dev_info(dev, "------------------------------------\n");
	for (i = 0; i < 4; i++)
		temp_name[i] = readb(&(tb->Signature[i]));
	temp_name[4] = '\0';
	dev_info(dev, "   Signature = %s\n", temp_name);
	dev_info(dev, "   Spec Number = %d\n", readl(&(tb->SpecValence)));
	dev_info(dev, "   Transport methods supported = 0x%x\n",
	       readl(&(tb->TransportSupport)));
	dev_info(dev, "   Transport methods active = 0x%x\n",
	       readl(&(tb->TransportActive)));
	dev_info(dev, "   Requested transport Method = 0x%x\n",
	       readl(&(tb->HostWrite.TransportRequest)));
	dev_info(dev, "   Coalesce Interrupt Delay = 0x%x\n",
	       readl(&(tb->HostWrite.CoalIntDelay)));
	dev_info(dev, "   Coalesce Interrupt Count = 0x%x\n",
	       readl(&(tb->HostWrite.CoalIntCount)));
6853
	dev_info(dev, "   Max outstanding commands = %d\n",
6854 6855 6856 6857 6858 6859 6860 6861 6862
	       readl(&(tb->CmdsOutMax)));
	dev_info(dev, "   Bus Types = 0x%x\n", readl(&(tb->BusTypes)));
	for (i = 0; i < 16; i++)
		temp_name[i] = readb(&(tb->ServerName[i]));
	temp_name[16] = '\0';
	dev_info(dev, "   Server Name = %s\n", temp_name);
	dev_info(dev, "   Heartbeat Counter = 0x%x\n\n\n",
		readl(&(tb->HeartBeat)));
#endif				/* HPSA_DEBUG */
6863
}
6864 6865 6866 6867 6868 6869 6870 6871 6872 6873 6874 6875 6876 6877 6878 6879 6880 6881 6882 6883 6884 6885 6886 6887 6888 6889 6890 6891 6892 6893 6894 6895 6896 6897 6898 6899

static int find_PCI_BAR_index(struct pci_dev *pdev, unsigned long pci_bar_addr)
{
	int i, offset, mem_type, bar_type;

	if (pci_bar_addr == PCI_BASE_ADDRESS_0)	/* looking for BAR zero? */
		return 0;
	offset = 0;
	for (i = 0; i < DEVICE_COUNT_RESOURCE; i++) {
		bar_type = pci_resource_flags(pdev, i) & PCI_BASE_ADDRESS_SPACE;
		if (bar_type == PCI_BASE_ADDRESS_SPACE_IO)
			offset += 4;
		else {
			mem_type = pci_resource_flags(pdev, i) &
			    PCI_BASE_ADDRESS_MEM_TYPE_MASK;
			switch (mem_type) {
			case PCI_BASE_ADDRESS_MEM_TYPE_32:
			case PCI_BASE_ADDRESS_MEM_TYPE_1M:
				offset += 4;	/* 32 bit */
				break;
			case PCI_BASE_ADDRESS_MEM_TYPE_64:
				offset += 8;
				break;
			default:	/* reserved in PCI 2.2 */
				dev_warn(&pdev->dev,
				       "base address is invalid\n");
				return -1;
				break;
			}
		}
		if (offset == pci_bar_addr - PCI_BASE_ADDRESS_0)
			return i + 1;
	}
	return -1;
}

6900 6901 6902 6903 6904
static void hpsa_disable_interrupt_mode(struct ctlr_info *h)
{
	if (h->msix_vector) {
		if (h->pdev->msix_enabled)
			pci_disable_msix(h->pdev);
R
Robert Elliott 已提交
6905
		h->msix_vector = 0;
6906 6907 6908
	} else if (h->msi_vector) {
		if (h->pdev->msi_enabled)
			pci_disable_msi(h->pdev);
R
Robert Elliott 已提交
6909
		h->msi_vector = 0;
6910 6911 6912
	}
}

6913
/* If MSI/MSI-X is supported by the kernel we will try to enable it on
6914
 * controllers that are capable. If not, we use legacy INTx mode.
6915
 */
6916
static void hpsa_interrupt_mode(struct ctlr_info *h)
6917 6918
{
#ifdef CONFIG_PCI_MSI
6919 6920 6921 6922 6923 6924 6925
	int err, i;
	struct msix_entry hpsa_msix_entries[MAX_REPLY_QUEUES];

	for (i = 0; i < MAX_REPLY_QUEUES; i++) {
		hpsa_msix_entries[i].vector = 0;
		hpsa_msix_entries[i].entry = i;
	}
6926 6927

	/* Some boards advertise MSI but don't really support it */
6928 6929
	if ((h->board_id == 0x40700E11) || (h->board_id == 0x40800E11) ||
	    (h->board_id == 0x40820E11) || (h->board_id == 0x40830E11))
6930
		goto default_int_mode;
6931
	if (pci_find_capability(h->pdev, PCI_CAP_ID_MSIX)) {
6932
		dev_info(&h->pdev->dev, "MSI-X capable controller\n");
6933
		h->msix_vector = MAX_REPLY_QUEUES;
6934 6935
		if (h->msix_vector > num_online_cpus())
			h->msix_vector = num_online_cpus();
6936 6937 6938 6939 6940 6941 6942
		err = pci_enable_msix_range(h->pdev, hpsa_msix_entries,
					    1, h->msix_vector);
		if (err < 0) {
			dev_warn(&h->pdev->dev, "MSI-X init failed %d\n", err);
			h->msix_vector = 0;
			goto single_msi_mode;
		} else if (err < h->msix_vector) {
6943
			dev_warn(&h->pdev->dev, "only %d MSI-X vectors "
6944 6945
			       "available\n", err);
		}
6946 6947 6948 6949
		h->msix_vector = err;
		for (i = 0; i < h->msix_vector; i++)
			h->intr[i] = hpsa_msix_entries[i].vector;
		return;
6950
	}
6951
single_msi_mode:
6952
	if (pci_find_capability(h->pdev, PCI_CAP_ID_MSI)) {
6953
		dev_info(&h->pdev->dev, "MSI capable controller\n");
6954
		if (!pci_enable_msi(h->pdev))
6955 6956
			h->msi_vector = 1;
		else
6957
			dev_warn(&h->pdev->dev, "MSI init failed\n");
6958 6959 6960 6961
	}
default_int_mode:
#endif				/* CONFIG_PCI_MSI */
	/* if we get here we're going to use the default interrupt mode */
6962
	h->intr[h->intr_mode] = h->pdev->irq;
6963 6964
}

6965
static int hpsa_lookup_board_id(struct pci_dev *pdev, u32 *board_id)
6966 6967 6968 6969 6970 6971 6972 6973 6974 6975 6976 6977 6978
{
	int i;
	u32 subsystem_vendor_id, subsystem_device_id;

	subsystem_vendor_id = pdev->subsystem_vendor;
	subsystem_device_id = pdev->subsystem_device;
	*board_id = ((subsystem_device_id << 16) & 0xffff0000) |
		    subsystem_vendor_id;

	for (i = 0; i < ARRAY_SIZE(products); i++)
		if (*board_id == products[i].board_id)
			return i;

6979 6980 6981
	if ((subsystem_vendor_id != PCI_VENDOR_ID_HP &&
		subsystem_vendor_id != PCI_VENDOR_ID_COMPAQ) ||
		!hpsa_allow_any) {
6982 6983 6984 6985 6986 6987 6988
		dev_warn(&pdev->dev, "unrecognized board ID: "
			"0x%08x, ignoring.\n", *board_id);
			return -ENODEV;
	}
	return ARRAY_SIZE(products) - 1; /* generic unknown smart array */
}

6989 6990
static int hpsa_pci_find_memory_BAR(struct pci_dev *pdev,
				    unsigned long *memory_bar)
6991 6992 6993 6994
{
	int i;

	for (i = 0; i < DEVICE_COUNT_RESOURCE; i++)
6995
		if (pci_resource_flags(pdev, i) & IORESOURCE_MEM) {
6996
			/* addressing mode bits already removed */
6997 6998
			*memory_bar = pci_resource_start(pdev, i);
			dev_dbg(&pdev->dev, "memory BAR = %lx\n",
6999 7000 7001
				*memory_bar);
			return 0;
		}
7002
	dev_warn(&pdev->dev, "no memory BAR found\n");
7003 7004 7005
	return -ENODEV;
}

7006 7007
static int hpsa_wait_for_board_state(struct pci_dev *pdev, void __iomem *vaddr,
				     int wait_for_ready)
7008
{
7009
	int i, iterations;
7010
	u32 scratchpad;
7011 7012 7013 7014
	if (wait_for_ready)
		iterations = HPSA_BOARD_READY_ITERATIONS;
	else
		iterations = HPSA_BOARD_NOT_READY_ITERATIONS;
7015

7016 7017 7018 7019 7020 7021 7022 7023 7024
	for (i = 0; i < iterations; i++) {
		scratchpad = readl(vaddr + SA5_SCRATCHPAD_OFFSET);
		if (wait_for_ready) {
			if (scratchpad == HPSA_FIRMWARE_READY)
				return 0;
		} else {
			if (scratchpad != HPSA_FIRMWARE_READY)
				return 0;
		}
7025 7026
		msleep(HPSA_BOARD_READY_POLL_INTERVAL_MSECS);
	}
7027
	dev_warn(&pdev->dev, "board not ready, timed out.\n");
7028 7029 7030
	return -ENODEV;
}

7031 7032 7033
static int hpsa_find_cfg_addrs(struct pci_dev *pdev, void __iomem *vaddr,
			       u32 *cfg_base_addr, u64 *cfg_base_addr_index,
			       u64 *cfg_offset)
7034 7035 7036 7037 7038 7039 7040 7041 7042 7043 7044 7045
{
	*cfg_base_addr = readl(vaddr + SA5_CTCFG_OFFSET);
	*cfg_offset = readl(vaddr + SA5_CTMEM_OFFSET);
	*cfg_base_addr &= (u32) 0x0000ffff;
	*cfg_base_addr_index = find_PCI_BAR_index(pdev, *cfg_base_addr);
	if (*cfg_base_addr_index == -1) {
		dev_warn(&pdev->dev, "cannot find cfg_base_addr_index\n");
		return -ENODEV;
	}
	return 0;
}

R
Robert Elliott 已提交
7046 7047
static void hpsa_free_cfgtables(struct ctlr_info *h)
{
R
Robert Elliott 已提交
7048
	if (h->transtable) {
R
Robert Elliott 已提交
7049
		iounmap(h->transtable);
R
Robert Elliott 已提交
7050 7051 7052
		h->transtable = NULL;
	}
	if (h->cfgtable) {
R
Robert Elliott 已提交
7053
		iounmap(h->cfgtable);
R
Robert Elliott 已提交
7054 7055
		h->cfgtable = NULL;
	}
R
Robert Elliott 已提交
7056 7057 7058 7059 7060
}

/* Find and map CISS config table and transfer table
+ * several items must be unmapped (freed) later
+ * */
7061
static int hpsa_find_cfgtables(struct ctlr_info *h)
7062
{
7063 7064 7065
	u64 cfg_offset;
	u32 cfg_base_addr;
	u64 cfg_base_addr_index;
7066
	u32 trans_offset;
7067
	int rc;
7068

7069 7070 7071 7072
	rc = hpsa_find_cfg_addrs(h->pdev, h->vaddr, &cfg_base_addr,
		&cfg_base_addr_index, &cfg_offset);
	if (rc)
		return rc;
7073
	h->cfgtable = remap_pci_mem(pci_resource_start(h->pdev,
7074
		       cfg_base_addr_index) + cfg_offset, sizeof(*h->cfgtable));
7075 7076
	if (!h->cfgtable) {
		dev_err(&h->pdev->dev, "Failed mapping cfgtable\n");
7077
		return -ENOMEM;
7078
	}
7079 7080 7081
	rc = write_driver_ver_to_cfgtable(h->cfgtable);
	if (rc)
		return rc;
7082
	/* Find performant mode table. */
7083
	trans_offset = readl(&h->cfgtable->TransMethodOffset);
7084 7085 7086
	h->transtable = remap_pci_mem(pci_resource_start(h->pdev,
				cfg_base_addr_index)+cfg_offset+trans_offset,
				sizeof(*h->transtable));
R
Robert Elliott 已提交
7087 7088 7089
	if (!h->transtable) {
		dev_err(&h->pdev->dev, "Failed mapping transfer table\n");
		hpsa_free_cfgtables(h);
7090
		return -ENOMEM;
R
Robert Elliott 已提交
7091
	}
7092 7093 7094
	return 0;
}

7095
static void hpsa_get_max_perf_mode_cmds(struct ctlr_info *h)
7096
{
7097 7098 7099 7100
#define MIN_MAX_COMMANDS 16
	BUILD_BUG_ON(MIN_MAX_COMMANDS <= HPSA_NRESERVED_CMDS);

	h->max_commands = readl(&h->cfgtable->MaxPerformantModeCommands);
7101 7102 7103 7104 7105

	/* Limit commands in memory limited kdump scenario. */
	if (reset_devices && h->max_commands > 32)
		h->max_commands = 32;

7106 7107 7108 7109 7110 7111
	if (h->max_commands < MIN_MAX_COMMANDS) {
		dev_warn(&h->pdev->dev,
			"Controller reports max supported commands of %d Using %d instead. Ensure that firmware is up to date.\n",
			h->max_commands,
			MIN_MAX_COMMANDS);
		h->max_commands = MIN_MAX_COMMANDS;
7112 7113 7114
	}
}

7115 7116 7117 7118 7119 7120 7121 7122 7123
/* If the controller reports that the total max sg entries is greater than 512,
 * then we know that chained SG blocks work.  (Original smart arrays did not
 * support chained SG blocks and would return zero for max sg entries.)
 */
static int hpsa_supports_chained_sg_blocks(struct ctlr_info *h)
{
	return h->maxsgentries > 512;
}

7124 7125 7126 7127
/* Interrogate the hardware for some limits:
 * max commands, max SG elements without chaining, and with chaining,
 * SG chain block size, etc.
 */
7128
static void hpsa_find_board_params(struct ctlr_info *h)
7129
{
7130
	hpsa_get_max_perf_mode_cmds(h);
7131
	h->nr_cmds = h->max_commands;
7132
	h->maxsgentries = readl(&(h->cfgtable->MaxScatterGatherElements));
7133
	h->fw_support = readl(&(h->cfgtable->misc_fw_support));
7134 7135
	if (hpsa_supports_chained_sg_blocks(h)) {
		/* Limit in-command s/g elements to 32 save dma'able memory. */
7136
		h->max_cmd_sg_entries = 32;
7137
		h->chainsize = h->maxsgentries - h->max_cmd_sg_entries;
7138 7139
		h->maxsgentries--; /* save one for chain pointer */
	} else {
7140 7141 7142 7143 7144 7145
		/*
		 * Original smart arrays supported at most 31 s/g entries
		 * embedded inline in the command (trying to use more
		 * would lock up the controller)
		 */
		h->max_cmd_sg_entries = 31;
7146
		h->maxsgentries = 31; /* default to traditional values */
7147
		h->chainsize = 0;
7148
	}
7149 7150 7151

	/* Find out what task management functions are supported and cache */
	h->TMFSupportFlags = readl(&(h->cfgtable->TMFSupportFlags));
7152 7153 7154 7155
	if (!(HPSATMF_PHYS_TASK_ABORT & h->TMFSupportFlags))
		dev_warn(&h->pdev->dev, "Physical aborts not supported\n");
	if (!(HPSATMF_LOG_TASK_ABORT & h->TMFSupportFlags))
		dev_warn(&h->pdev->dev, "Logical aborts not supported\n");
7156 7157
	if (!(HPSATMF_IOACCEL_ENABLED & h->TMFSupportFlags))
		dev_warn(&h->pdev->dev, "HP SSD Smart Path aborts not supported\n");
7158 7159
}

7160 7161
static inline bool hpsa_CISS_signature_present(struct ctlr_info *h)
{
A
Akinobu Mita 已提交
7162
	if (!check_signature(h->cfgtable->Signature, "CISS", 4)) {
7163
		dev_err(&h->pdev->dev, "not a valid CISS config table\n");
7164 7165 7166 7167 7168
		return false;
	}
	return true;
}

7169
static inline void hpsa_set_driver_support_bits(struct ctlr_info *h)
7170
{
7171
	u32 driver_support;
7172

7173
	driver_support = readl(&(h->cfgtable->driver_support));
A
Arnd Bergmann 已提交
7174 7175
	/* Need to enable prefetch in the SCSI core for 6400 in x86 */
#ifdef CONFIG_X86
7176
	driver_support |= ENABLE_SCSI_PREFETCH;
7177
#endif
7178 7179
	driver_support |= ENABLE_UNIT_ATTN;
	writel(driver_support, &(h->cfgtable->driver_support));
7180 7181
}

7182 7183 7184 7185 7186 7187 7188 7189 7190 7191 7192 7193 7194 7195
/* Disable DMA prefetch for the P600.  Otherwise an ASIC bug may result
 * in a prefetch beyond physical memory.
 */
static inline void hpsa_p600_dma_prefetch_quirk(struct ctlr_info *h)
{
	u32 dma_prefetch;

	if (h->board_id != 0x3225103C)
		return;
	dma_prefetch = readl(h->vaddr + I2O_DMA1_CFG);
	dma_prefetch |= 0x8000;
	writel(dma_prefetch, h->vaddr + I2O_DMA1_CFG);
}

7196
static int hpsa_wait_for_clear_event_notify_ack(struct ctlr_info *h)
7197 7198 7199 7200 7201
{
	int i;
	u32 doorbell_value;
	unsigned long flags;
	/* wait until the clear_event_notify bit 6 is cleared by controller. */
7202
	for (i = 0; i < MAX_CLEAR_EVENT_WAIT; i++) {
7203 7204 7205 7206
		spin_lock_irqsave(&h->lock, flags);
		doorbell_value = readl(h->vaddr + SA5_DOORBELL);
		spin_unlock_irqrestore(&h->lock, flags);
		if (!(doorbell_value & DOORBELL_CLEAR_EVENTS))
7207
			goto done;
7208
		/* delay and try again */
7209
		msleep(CLEAR_EVENT_WAIT_INTERVAL);
7210
	}
7211 7212 7213
	return -ENODEV;
done:
	return 0;
7214 7215
}

7216
static int hpsa_wait_for_mode_change_ack(struct ctlr_info *h)
7217 7218
{
	int i;
7219 7220
	u32 doorbell_value;
	unsigned long flags;
7221 7222 7223 7224 7225

	/* under certain very rare conditions, this can take awhile.
	 * (e.g.: hot replace a failed 144GB drive in a RAID 5 set right
	 * as we enter this code.)
	 */
7226
	for (i = 0; i < MAX_MODE_CHANGE_WAIT; i++) {
7227 7228
		if (h->remove_in_progress)
			goto done;
7229 7230 7231
		spin_lock_irqsave(&h->lock, flags);
		doorbell_value = readl(h->vaddr + SA5_DOORBELL);
		spin_unlock_irqrestore(&h->lock, flags);
D
Dan Carpenter 已提交
7232
		if (!(doorbell_value & CFGTBL_ChangeReq))
7233
			goto done;
7234
		/* delay and try again */
7235
		msleep(MODE_CHANGE_WAIT_INTERVAL);
7236
	}
7237 7238 7239
	return -ENODEV;
done:
	return 0;
7240 7241
}

7242
/* return -ENODEV or other reason on error, 0 on success */
7243
static int hpsa_enter_simple_mode(struct ctlr_info *h)
7244 7245 7246 7247 7248 7249 7250 7251
{
	u32 trans_support;

	trans_support = readl(&(h->cfgtable->TransportSupport));
	if (!(trans_support & SIMPLE_MODE))
		return -ENOTSUPP;

	h->max_commands = readl(&(h->cfgtable->CmdsOutMax));
7252

7253 7254
	/* Update the field, and then ring the doorbell */
	writel(CFGTBL_Trans_Simple, &(h->cfgtable->HostWrite.TransportRequest));
7255
	writel(0, &h->cfgtable->HostWrite.command_pool_addr_hi);
7256
	writel(CFGTBL_ChangeReq, h->vaddr + SA5_DOORBELL);
7257 7258
	if (hpsa_wait_for_mode_change_ack(h))
		goto error;
7259
	print_cfg_table(&h->pdev->dev, h->cfgtable);
7260 7261
	if (!(readl(&(h->cfgtable->TransportActive)) & CFGTBL_Trans_Simple))
		goto error;
7262
	h->transMethod = CFGTBL_Trans_Simple;
7263
	return 0;
7264
error:
7265
	dev_err(&h->pdev->dev, "failed to enter simple mode\n");
7266
	return -ENODEV;
7267 7268
}

R
Robert Elliott 已提交
7269 7270 7271 7272 7273
/* free items allocated or mapped by hpsa_pci_init */
static void hpsa_free_pci_init(struct ctlr_info *h)
{
	hpsa_free_cfgtables(h);			/* pci_init 4 */
	iounmap(h->vaddr);			/* pci_init 3 */
R
Robert Elliott 已提交
7274
	h->vaddr = NULL;
R
Robert Elliott 已提交
7275
	hpsa_disable_interrupt_mode(h);		/* pci_init 2 */
7276 7277 7278 7279
	/*
	 * call pci_disable_device before pci_release_regions per
	 * Documentation/PCI/pci.txt
	 */
R
Robert Elliott 已提交
7280
	pci_disable_device(h->pdev);		/* pci_init 1 */
7281
	pci_release_regions(h->pdev);		/* pci_init 2 */
R
Robert Elliott 已提交
7282 7283 7284
}

/* several items must be freed later */
7285
static int hpsa_pci_init(struct ctlr_info *h)
7286
{
7287
	int prod_index, err;
7288

7289 7290
	prod_index = hpsa_lookup_board_id(h->pdev, &h->board_id);
	if (prod_index < 0)
7291
		return prod_index;
7292 7293
	h->product_name = products[prod_index].product_name;
	h->access = *(products[prod_index].access);
7294

S
Stephen Cameron 已提交
7295 7296 7297
	h->needs_abort_tags_swizzled =
		ctlr_needs_abort_tags_swizzled(h->board_id);

M
Matthew Garrett 已提交
7298 7299 7300
	pci_disable_link_state(h->pdev, PCIE_LINK_STATE_L0S |
			       PCIE_LINK_STATE_L1 | PCIE_LINK_STATE_CLKPM);

7301
	err = pci_enable_device(h->pdev);
7302
	if (err) {
R
Robert Elliott 已提交
7303
		dev_err(&h->pdev->dev, "failed to enable PCI device\n");
7304
		pci_disable_device(h->pdev);
7305 7306 7307
		return err;
	}

7308
	err = pci_request_regions(h->pdev, HPSA);
7309
	if (err) {
7310
		dev_err(&h->pdev->dev,
R
Robert Elliott 已提交
7311
			"failed to obtain PCI resources\n");
7312 7313
		pci_disable_device(h->pdev);
		return err;
7314
	}
7315 7316 7317

	pci_set_master(h->pdev);

7318
	hpsa_interrupt_mode(h);
7319
	err = hpsa_pci_find_memory_BAR(h->pdev, &h->paddr);
7320
	if (err)
R
Robert Elliott 已提交
7321
		goto clean2;	/* intmode+region, pci */
7322
	h->vaddr = remap_pci_mem(h->paddr, 0x250);
7323
	if (!h->vaddr) {
R
Robert Elliott 已提交
7324
		dev_err(&h->pdev->dev, "failed to remap PCI mem\n");
7325
		err = -ENOMEM;
R
Robert Elliott 已提交
7326
		goto clean2;	/* intmode+region, pci */
7327
	}
7328
	err = hpsa_wait_for_board_state(h->pdev, h->vaddr, BOARD_READY);
7329
	if (err)
R
Robert Elliott 已提交
7330
		goto clean3;	/* vaddr, intmode+region, pci */
7331 7332
	err = hpsa_find_cfgtables(h);
	if (err)
R
Robert Elliott 已提交
7333
		goto clean3;	/* vaddr, intmode+region, pci */
7334
	hpsa_find_board_params(h);
7335

7336
	if (!hpsa_CISS_signature_present(h)) {
7337
		err = -ENODEV;
R
Robert Elliott 已提交
7338
		goto clean4;	/* cfgtables, vaddr, intmode+region, pci */
7339
	}
7340
	hpsa_set_driver_support_bits(h);
7341
	hpsa_p600_dma_prefetch_quirk(h);
7342 7343
	err = hpsa_enter_simple_mode(h);
	if (err)
R
Robert Elliott 已提交
7344
		goto clean4;	/* cfgtables, vaddr, intmode+region, pci */
7345 7346
	return 0;

R
Robert Elliott 已提交
7347 7348 7349 7350
clean4:	/* cfgtables, vaddr, intmode+region, pci */
	hpsa_free_cfgtables(h);
clean3:	/* vaddr, intmode+region, pci */
	iounmap(h->vaddr);
R
Robert Elliott 已提交
7351
	h->vaddr = NULL;
R
Robert Elliott 已提交
7352 7353
clean2:	/* intmode+region, pci */
	hpsa_disable_interrupt_mode(h);
7354 7355 7356 7357
	/*
	 * call pci_disable_device before pci_release_regions per
	 * Documentation/PCI/pci.txt
	 */
R
Robert Elliott 已提交
7358
	pci_disable_device(h->pdev);
7359
	pci_release_regions(h->pdev);
7360 7361 7362
	return err;
}

7363
static void hpsa_hba_inquiry(struct ctlr_info *h)
7364 7365 7366 7367 7368 7369 7370 7371 7372 7373 7374 7375 7376 7377 7378
{
	int rc;

#define HBA_INQUIRY_BYTE_COUNT 64
	h->hba_inquiry_data = kmalloc(HBA_INQUIRY_BYTE_COUNT, GFP_KERNEL);
	if (!h->hba_inquiry_data)
		return;
	rc = hpsa_scsi_do_inquiry(h, RAID_CTLR_LUNID, 0,
		h->hba_inquiry_data, HBA_INQUIRY_BYTE_COUNT);
	if (rc != 0) {
		kfree(h->hba_inquiry_data);
		h->hba_inquiry_data = NULL;
	}
}

7379
static int hpsa_init_reset_devices(struct pci_dev *pdev, u32 board_id)
7380
{
7381
	int rc, i;
7382
	void __iomem *vaddr;
7383 7384 7385 7386

	if (!reset_devices)
		return 0;

7387 7388 7389 7390 7391 7392 7393 7394 7395 7396 7397 7398 7399 7400 7401 7402
	/* kdump kernel is loading, we don't know in which state is
	 * the pci interface. The dev->enable_cnt is equal zero
	 * so we call enable+disable, wait a while and switch it on.
	 */
	rc = pci_enable_device(pdev);
	if (rc) {
		dev_warn(&pdev->dev, "Failed to enable PCI device\n");
		return -ENODEV;
	}
	pci_disable_device(pdev);
	msleep(260);			/* a randomly chosen number */
	rc = pci_enable_device(pdev);
	if (rc) {
		dev_warn(&pdev->dev, "failed to enable device.\n");
		return -ENODEV;
	}
7403

7404
	pci_set_master(pdev);
7405

7406 7407 7408 7409 7410 7411 7412 7413
	vaddr = pci_ioremap_bar(pdev, 0);
	if (vaddr == NULL) {
		rc = -ENOMEM;
		goto out_disable;
	}
	writel(SA5_INTR_OFF, vaddr + SA5_REPLY_INTR_MASK_OFFSET);
	iounmap(vaddr);

7414
	/* Reset the controller with a PCI power-cycle or via doorbell */
7415
	rc = hpsa_kdump_hard_reset_controller(pdev, board_id);
7416

7417 7418
	/* -ENOTSUPP here means we cannot reset the controller
	 * but it's already (and still) up and running in
7419 7420
	 * "performant mode".  Or, it might be 640x, which can't reset
	 * due to concerns about shared bbwc between 6402/6404 pair.
7421
	 */
7422
	if (rc)
7423
		goto out_disable;
7424 7425

	/* Now try to get the controller to respond to a no-op */
7426
	dev_info(&pdev->dev, "Waiting for controller to respond to no-op\n");
7427 7428 7429 7430 7431 7432 7433
	for (i = 0; i < HPSA_POST_RESET_NOOP_RETRIES; i++) {
		if (hpsa_noop(pdev) == 0)
			break;
		else
			dev_warn(&pdev->dev, "no-op failed%s\n",
					(i < 11 ? "; re-trying" : ""));
	}
7434 7435 7436 7437 7438

out_disable:

	pci_disable_device(pdev);
	return rc;
7439 7440
}

7441 7442 7443
static void hpsa_free_cmd_pool(struct ctlr_info *h)
{
	kfree(h->cmd_pool_bits);
R
Robert Elliott 已提交
7444 7445
	h->cmd_pool_bits = NULL;
	if (h->cmd_pool) {
7446 7447 7448 7449
		pci_free_consistent(h->pdev,
				h->nr_cmds * sizeof(struct CommandList),
				h->cmd_pool,
				h->cmd_pool_dhandle);
R
Robert Elliott 已提交
7450 7451 7452 7453
		h->cmd_pool = NULL;
		h->cmd_pool_dhandle = 0;
	}
	if (h->errinfo_pool) {
7454 7455 7456 7457
		pci_free_consistent(h->pdev,
				h->nr_cmds * sizeof(struct ErrorInfo),
				h->errinfo_pool,
				h->errinfo_pool_dhandle);
R
Robert Elliott 已提交
7458 7459 7460
		h->errinfo_pool = NULL;
		h->errinfo_pool_dhandle = 0;
	}
7461 7462
}

7463
static int hpsa_alloc_cmd_pool(struct ctlr_info *h)
7464 7465 7466 7467 7468 7469 7470 7471 7472 7473 7474 7475 7476 7477
{
	h->cmd_pool_bits = kzalloc(
		DIV_ROUND_UP(h->nr_cmds, BITS_PER_LONG) *
		sizeof(unsigned long), GFP_KERNEL);
	h->cmd_pool = pci_alloc_consistent(h->pdev,
		    h->nr_cmds * sizeof(*h->cmd_pool),
		    &(h->cmd_pool_dhandle));
	h->errinfo_pool = pci_alloc_consistent(h->pdev,
		    h->nr_cmds * sizeof(*h->errinfo_pool),
		    &(h->errinfo_pool_dhandle));
	if ((h->cmd_pool_bits == NULL)
	    || (h->cmd_pool == NULL)
	    || (h->errinfo_pool == NULL)) {
		dev_err(&h->pdev->dev, "out of memory in %s", __func__);
7478
		goto clean_up;
7479
	}
7480
	hpsa_preinitialize_commands(h);
7481
	return 0;
7482 7483 7484
clean_up:
	hpsa_free_cmd_pool(h);
	return -ENOMEM;
7485 7486
}

7487 7488
static void hpsa_irq_affinity_hints(struct ctlr_info *h)
{
7489
	int i, cpu;
7490 7491 7492

	cpu = cpumask_first(cpu_online_mask);
	for (i = 0; i < h->msix_vector; i++) {
7493
		irq_set_affinity_hint(h->intr[i], get_cpu_mask(cpu));
7494 7495 7496 7497
		cpu = cpumask_next(cpu, cpu_online_mask);
	}
}

7498 7499 7500 7501 7502 7503 7504 7505 7506 7507
/* clear affinity hints and free MSI-X, MSI, or legacy INTx vectors */
static void hpsa_free_irqs(struct ctlr_info *h)
{
	int i;

	if (!h->msix_vector || h->intr_mode != PERF_MODE_INT) {
		/* Single reply queue, only one irq to free */
		i = h->intr_mode;
		irq_set_affinity_hint(h->intr[i], NULL);
		free_irq(h->intr[i], &h->q[i]);
R
Robert Elliott 已提交
7508
		h->q[i] = 0;
7509 7510 7511 7512 7513 7514
		return;
	}

	for (i = 0; i < h->msix_vector; i++) {
		irq_set_affinity_hint(h->intr[i], NULL);
		free_irq(h->intr[i], &h->q[i]);
R
Robert Elliott 已提交
7515
		h->q[i] = 0;
7516
	}
7517 7518
	for (; i < MAX_REPLY_QUEUES; i++)
		h->q[i] = 0;
7519 7520
}

7521 7522
/* returns 0 on success; cleans up and returns -Enn on error */
static int hpsa_request_irqs(struct ctlr_info *h,
7523 7524 7525
	irqreturn_t (*msixhandler)(int, void *),
	irqreturn_t (*intxhandler)(int, void *))
{
7526
	int rc, i;
7527

7528 7529 7530 7531 7532 7533 7534
	/*
	 * initialize h->q[x] = x so that interrupt handlers know which
	 * queue to process.
	 */
	for (i = 0; i < MAX_REPLY_QUEUES; i++)
		h->q[i] = (u8) i;

7535
	if (h->intr_mode == PERF_MODE_INT && h->msix_vector > 0) {
7536
		/* If performant mode and MSI-X, use multiple reply queues */
7537
		for (i = 0; i < h->msix_vector; i++) {
7538
			sprintf(h->intrname[i], "%s-msix%d", h->devname, i);
7539
			rc = request_irq(h->intr[i], msixhandler,
7540
					0, h->intrname[i],
7541
					&h->q[i]);
7542 7543 7544 7545 7546 7547 7548 7549 7550 7551 7552 7553 7554 7555 7556
			if (rc) {
				int j;

				dev_err(&h->pdev->dev,
					"failed to get irq %d for %s\n",
				       h->intr[i], h->devname);
				for (j = 0; j < i; j++) {
					free_irq(h->intr[j], &h->q[j]);
					h->q[j] = 0;
				}
				for (; j < MAX_REPLY_QUEUES; j++)
					h->q[j] = 0;
				return rc;
			}
		}
7557
		hpsa_irq_affinity_hints(h);
7558 7559
	} else {
		/* Use single reply pool */
7560
		if (h->msix_vector > 0 || h->msi_vector) {
7561 7562 7563 7564 7565 7566
			if (h->msix_vector)
				sprintf(h->intrname[h->intr_mode],
					"%s-msix", h->devname);
			else
				sprintf(h->intrname[h->intr_mode],
					"%s-msi", h->devname);
7567
			rc = request_irq(h->intr[h->intr_mode],
7568 7569
				msixhandler, 0,
				h->intrname[h->intr_mode],
7570 7571
				&h->q[h->intr_mode]);
		} else {
7572 7573
			sprintf(h->intrname[h->intr_mode],
				"%s-intx", h->devname);
7574
			rc = request_irq(h->intr[h->intr_mode],
7575 7576
				intxhandler, IRQF_SHARED,
				h->intrname[h->intr_mode],
7577 7578
				&h->q[h->intr_mode]);
		}
R
Robert Elliott 已提交
7579
		irq_set_affinity_hint(h->intr[h->intr_mode], NULL);
7580
	}
7581
	if (rc) {
R
Robert Elliott 已提交
7582
		dev_err(&h->pdev->dev, "failed to get irq %d for %s\n",
7583
		       h->intr[h->intr_mode], h->devname);
R
Robert Elliott 已提交
7584
		hpsa_free_irqs(h);
7585 7586 7587 7588 7589
		return -ENODEV;
	}
	return 0;
}

7590
static int hpsa_kdump_soft_reset(struct ctlr_info *h)
7591
{
7592
	int rc;
7593
	hpsa_send_host_reset(h, RAID_CTLR_LUNID, HPSA_RESET_TYPE_CONTROLLER);
7594 7595

	dev_info(&h->pdev->dev, "Waiting for board to soft reset.\n");
7596 7597
	rc = hpsa_wait_for_board_state(h->pdev, h->vaddr, BOARD_NOT_READY);
	if (rc) {
7598
		dev_warn(&h->pdev->dev, "Soft reset had no effect.\n");
7599
		return rc;
7600 7601 7602
	}

	dev_info(&h->pdev->dev, "Board reset, awaiting READY status.\n");
7603 7604
	rc = hpsa_wait_for_board_state(h->pdev, h->vaddr, BOARD_READY);
	if (rc) {
7605 7606
		dev_warn(&h->pdev->dev, "Board failed to become ready "
			"after soft reset.\n");
7607
		return rc;
7608 7609 7610 7611 7612
	}

	return 0;
}

7613 7614 7615 7616 7617 7618 7619
static void hpsa_free_reply_queues(struct ctlr_info *h)
{
	int i;

	for (i = 0; i < h->nreply_queues; i++) {
		if (!h->reply_queue[i].head)
			continue;
7620 7621 7622 7623
		pci_free_consistent(h->pdev,
					h->reply_queue_size,
					h->reply_queue[i].head,
					h->reply_queue[i].busaddr);
7624 7625 7626
		h->reply_queue[i].head = NULL;
		h->reply_queue[i].busaddr = 0;
	}
R
Robert Elliott 已提交
7627
	h->reply_queue_size = 0;
7628 7629
}

7630 7631
static void hpsa_undo_allocations_after_kdump_soft_reset(struct ctlr_info *h)
{
R
Robert Elliott 已提交
7632 7633 7634 7635
	hpsa_free_performant_mode(h);		/* init_one 7 */
	hpsa_free_sg_chain_blocks(h);		/* init_one 6 */
	hpsa_free_cmd_pool(h);			/* init_one 5 */
	hpsa_free_irqs(h);			/* init_one 4 */
7636 7637 7638
	scsi_host_put(h->scsi_host);		/* init_one 3 */
	h->scsi_host = NULL;			/* init_one 3 */
	hpsa_free_pci_init(h);			/* init_one 2_5 */
7639 7640 7641 7642 7643 7644 7645 7646 7647 7648
	free_percpu(h->lockup_detected);	/* init_one 2 */
	h->lockup_detected = NULL;		/* init_one 2 */
	if (h->resubmit_wq) {
		destroy_workqueue(h->resubmit_wq);	/* init_one 1 */
		h->resubmit_wq = NULL;
	}
	if (h->rescan_ctlr_wq) {
		destroy_workqueue(h->rescan_ctlr_wq);
		h->rescan_ctlr_wq = NULL;
	}
R
Robert Elliott 已提交
7649
	kfree(h);				/* init_one 1 */
7650 7651
}

7652
/* Called when controller lockup detected. */
7653
static void fail_all_outstanding_cmds(struct ctlr_info *h)
7654
{
7655 7656
	int i, refcount;
	struct CommandList *c;
7657
	int failcount = 0;
7658

7659
	flush_workqueue(h->resubmit_wq); /* ensure all cmds are fully built */
7660 7661
	for (i = 0; i < h->nr_cmds; i++) {
		c = h->cmd_pool + i;
7662 7663
		refcount = atomic_inc_return(&c->refcount);
		if (refcount > 1) {
7664
			c->err_info->CommandStatus = CMD_CTLR_LOCKUP;
7665
			finish_cmd(c);
7666
			atomic_dec(&h->commands_outstanding);
7667
			failcount++;
7668 7669
		}
		cmd_free(h, c);
7670
	}
7671 7672
	dev_warn(&h->pdev->dev,
		"failed %d commands in fail_all\n", failcount);
7673 7674
}

7675 7676
static void set_lockup_detected_for_all_cpus(struct ctlr_info *h, u32 value)
{
7677
	int cpu;
7678

7679
	for_each_online_cpu(cpu) {
7680 7681 7682 7683 7684 7685 7686
		u32 *lockup_detected;
		lockup_detected = per_cpu_ptr(h->lockup_detected, cpu);
		*lockup_detected = value;
	}
	wmb(); /* be sure the per-cpu variables are out to memory */
}

7687 7688 7689
static void controller_lockup_detected(struct ctlr_info *h)
{
	unsigned long flags;
7690
	u32 lockup_detected;
7691 7692 7693

	h->access.set_intr_mask(h, HPSA_INTR_OFF);
	spin_lock_irqsave(&h->lock, flags);
7694 7695 7696 7697
	lockup_detected = readl(h->vaddr + SA5_SCRATCHPAD_OFFSET);
	if (!lockup_detected) {
		/* no heartbeat, but controller gave us a zero. */
		dev_warn(&h->pdev->dev,
7698 7699
			"lockup detected after %d but scratchpad register is zero\n",
			h->heartbeat_sample_interval / HZ);
7700 7701 7702
		lockup_detected = 0xffffffff;
	}
	set_lockup_detected_for_all_cpus(h, lockup_detected);
7703
	spin_unlock_irqrestore(&h->lock, flags);
7704 7705
	dev_warn(&h->pdev->dev, "Controller lockup detected: 0x%08x after %d\n",
			lockup_detected, h->heartbeat_sample_interval / HZ);
7706
	pci_disable_device(h->pdev);
7707
	fail_all_outstanding_cmds(h);
7708 7709
}

7710
static int detect_controller_lockup(struct ctlr_info *h)
7711 7712 7713 7714 7715 7716 7717 7718
{
	u64 now;
	u32 heartbeat;
	unsigned long flags;

	now = get_jiffies_64();
	/* If we've received an interrupt recently, we're ok. */
	if (time_after64(h->last_intr_timestamp +
7719
				(h->heartbeat_sample_interval), now))
7720
		return false;
7721 7722 7723 7724 7725 7726 7727

	/*
	 * If we've already checked the heartbeat recently, we're ok.
	 * This could happen if someone sends us a signal. We
	 * otherwise don't care about signals in this thread.
	 */
	if (time_after64(h->last_heartbeat_timestamp +
7728
				(h->heartbeat_sample_interval), now))
7729
		return false;
7730 7731 7732 7733 7734 7735 7736

	/* If heartbeat has not changed since we last looked, we're not ok. */
	spin_lock_irqsave(&h->lock, flags);
	heartbeat = readl(&h->cfgtable->HeartBeat);
	spin_unlock_irqrestore(&h->lock, flags);
	if (h->last_heartbeat == heartbeat) {
		controller_lockup_detected(h);
7737
		return true;
7738 7739 7740 7741 7742
	}

	/* We're ok. */
	h->last_heartbeat = heartbeat;
	h->last_heartbeat_timestamp = now;
7743
	return false;
7744 7745
}

7746
static void hpsa_ack_ctlr_events(struct ctlr_info *h)
7747 7748 7749 7750
{
	int i;
	char *event_type;

7751 7752 7753
	if (!(h->fw_support & MISC_FW_EVENT_NOTIFY))
		return;

7754
	/* Ask the controller to clear the events we're handling. */
7755 7756
	if ((h->transMethod & (CFGTBL_Trans_io_accel1
			| CFGTBL_Trans_io_accel2)) &&
7757 7758 7759 7760 7761 7762 7763 7764 7765 7766 7767
		(h->events & HPSA_EVENT_NOTIFY_ACCEL_IO_PATH_STATE_CHANGE ||
		 h->events & HPSA_EVENT_NOTIFY_ACCEL_IO_PATH_CONFIG_CHANGE)) {

		if (h->events & HPSA_EVENT_NOTIFY_ACCEL_IO_PATH_STATE_CHANGE)
			event_type = "state change";
		if (h->events & HPSA_EVENT_NOTIFY_ACCEL_IO_PATH_CONFIG_CHANGE)
			event_type = "configuration change";
		/* Stop sending new RAID offload reqs via the IO accelerator */
		scsi_block_requests(h->scsi_host);
		for (i = 0; i < h->ndevices; i++)
			h->dev[i]->offload_enabled = 0;
7768
		hpsa_drain_accel_commands(h);
7769 7770 7771 7772 7773 7774 7775 7776 7777 7778 7779 7780 7781 7782 7783 7784 7785 7786 7787 7788
		/* Set 'accelerator path config change' bit */
		dev_warn(&h->pdev->dev,
			"Acknowledging event: 0x%08x (HP SSD Smart Path %s)\n",
			h->events, event_type);
		writel(h->events, &(h->cfgtable->clear_event_notify));
		/* Set the "clear event notify field update" bit 6 */
		writel(DOORBELL_CLEAR_EVENTS, h->vaddr + SA5_DOORBELL);
		/* Wait until ctlr clears 'clear event notify field', bit 6 */
		hpsa_wait_for_clear_event_notify_ack(h);
		scsi_unblock_requests(h->scsi_host);
	} else {
		/* Acknowledge controller notification events. */
		writel(h->events, &(h->cfgtable->clear_event_notify));
		writel(DOORBELL_CLEAR_EVENTS, h->vaddr + SA5_DOORBELL);
		hpsa_wait_for_clear_event_notify_ack(h);
#if 0
		writel(CFGTBL_ChangeReq, h->vaddr + SA5_DOORBELL);
		hpsa_wait_for_mode_change_ack(h);
#endif
	}
7789
	return;
7790 7791 7792 7793
}

/* Check a register on the controller to see if there are configuration
 * changes (added/changed/removed logical drives, etc.) which mean that
7794 7795
 * we should rescan the controller for devices.
 * Also check flag for driver-initiated rescan.
7796
 */
7797
static int hpsa_ctlr_needs_rescan(struct ctlr_info *h)
7798 7799
{
	if (!(h->fw_support & MISC_FW_EVENT_NOTIFY))
7800
		return 0;
7801 7802

	h->events = readl(&(h->cfgtable->event_notify));
7803 7804
	return h->events & RESCAN_REQUIRED_EVENT_BITS;
}
7805

7806 7807 7808 7809 7810 7811 7812 7813 7814 7815 7816 7817 7818 7819
/*
 * Check if any of the offline devices have become ready
 */
static int hpsa_offline_devices_ready(struct ctlr_info *h)
{
	unsigned long flags;
	struct offline_device_entry *d;
	struct list_head *this, *tmp;

	spin_lock_irqsave(&h->offline_device_lock, flags);
	list_for_each_safe(this, tmp, &h->offline_device_list) {
		d = list_entry(this, struct offline_device_entry,
				offline_list);
		spin_unlock_irqrestore(&h->offline_device_lock, flags);
7820 7821 7822 7823
		if (!hpsa_volume_offline(h, d->scsi3addr)) {
			spin_lock_irqsave(&h->offline_device_lock, flags);
			list_del(&d->offline_list);
			spin_unlock_irqrestore(&h->offline_device_lock, flags);
7824
			return 1;
7825
		}
7826 7827 7828 7829
		spin_lock_irqsave(&h->offline_device_lock, flags);
	}
	spin_unlock_irqrestore(&h->offline_device_lock, flags);
	return 0;
7830 7831
}

7832
static void hpsa_rescan_ctlr_worker(struct work_struct *work)
7833 7834
{
	unsigned long flags;
7835
	struct ctlr_info *h = container_of(to_delayed_work(work),
7836 7837 7838 7839
					struct ctlr_info, rescan_ctlr_work);


	if (h->remove_in_progress)
7840
		return;
7841 7842 7843 7844 7845 7846 7847

	if (hpsa_ctlr_needs_rescan(h) || hpsa_offline_devices_ready(h)) {
		scsi_host_get(h->scsi_host);
		hpsa_ack_ctlr_events(h);
		hpsa_scan_start(h->scsi_host);
		scsi_host_put(h->scsi_host);
	}
7848
	spin_lock_irqsave(&h->lock, flags);
7849 7850 7851 7852 7853 7854 7855 7856 7857 7858 7859 7860 7861 7862
	if (!h->remove_in_progress)
		queue_delayed_work(h->rescan_ctlr_wq, &h->rescan_ctlr_work,
				h->heartbeat_sample_interval);
	spin_unlock_irqrestore(&h->lock, flags);
}

static void hpsa_monitor_ctlr_worker(struct work_struct *work)
{
	unsigned long flags;
	struct ctlr_info *h = container_of(to_delayed_work(work),
					struct ctlr_info, monitor_ctlr_work);

	detect_controller_lockup(h);
	if (lockup_detected(h))
7863
		return;
7864 7865 7866 7867

	spin_lock_irqsave(&h->lock, flags);
	if (!h->remove_in_progress)
		schedule_delayed_work(&h->monitor_ctlr_work,
7868 7869
				h->heartbeat_sample_interval);
	spin_unlock_irqrestore(&h->lock, flags);
7870 7871
}

7872 7873 7874 7875 7876
static struct workqueue_struct *hpsa_create_controller_wq(struct ctlr_info *h,
						char *name)
{
	struct workqueue_struct *wq = NULL;

7877
	wq = alloc_ordered_workqueue("%s_%d_hpsa", 0, name, h->ctlr);
7878 7879 7880 7881 7882 7883
	if (!wq)
		dev_err(&h->pdev->dev, "failed to create %s workqueue\n", name);

	return wq;
}

7884
static int hpsa_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
7885
{
7886
	int dac, rc;
7887
	struct ctlr_info *h;
7888 7889
	int try_soft_reset = 0;
	unsigned long flags;
7890
	u32 board_id;
7891 7892 7893 7894

	if (number_of_controllers == 0)
		printk(KERN_INFO DRIVER_NAME "\n");

7895 7896 7897 7898 7899 7900 7901
	rc = hpsa_lookup_board_id(pdev, &board_id);
	if (rc < 0) {
		dev_warn(&pdev->dev, "Board ID not found\n");
		return rc;
	}

	rc = hpsa_init_reset_devices(pdev, board_id);
7902 7903 7904 7905 7906 7907 7908 7909 7910 7911 7912 7913 7914
	if (rc) {
		if (rc != -ENOTSUPP)
			return rc;
		/* If the reset fails in a particular way (it has no way to do
		 * a proper hard reset, so returns -ENOTSUPP) we can try to do
		 * a soft reset once we get the controller configured up to the
		 * point that it can accept a command.
		 */
		try_soft_reset = 1;
		rc = 0;
	}

reinit_after_soft_reset:
7915

7916 7917 7918 7919 7920
	/* Command structures must be aligned on a 32-byte boundary because
	 * the 5 lower bits of the address are used by the hardware. and by
	 * the driver.  See comments in hpsa.h for more info.
	 */
	BUILD_BUG_ON(sizeof(struct CommandList) % COMMANDLIST_ALIGNMENT);
7921
	h = kzalloc(sizeof(*h), GFP_KERNEL);
R
Robert Elliott 已提交
7922 7923
	if (!h) {
		dev_err(&pdev->dev, "Failed to allocate controller head\n");
7924
		return -ENOMEM;
R
Robert Elliott 已提交
7925
	}
7926

7927
	h->pdev = pdev;
R
Robert Elliott 已提交
7928

7929
	h->intr_mode = hpsa_simple_mode ? SIMPLE_MODE_INT : PERF_MODE_INT;
7930
	INIT_LIST_HEAD(&h->offline_device_list);
7931
	spin_lock_init(&h->lock);
7932
	spin_lock_init(&h->offline_device_lock);
7933
	spin_lock_init(&h->scan_lock);
7934
	atomic_set(&h->passthru_cmds_avail, HPSA_MAX_CONCURRENT_PASSTHRUS);
S
Stephen Cameron 已提交
7935
	atomic_set(&h->abort_cmds_available, HPSA_CMDS_RESERVED_FOR_ABORTS);
7936 7937 7938

	/* Allocate and clear per-cpu variable lockup_detected */
	h->lockup_detected = alloc_percpu(u32);
7939
	if (!h->lockup_detected) {
R
Robert Elliott 已提交
7940
		dev_err(&h->pdev->dev, "Failed to allocate lockup detector\n");
7941
		rc = -ENOMEM;
7942
		goto clean1;	/* aer/h */
7943
	}
7944 7945
	set_lockup_detected_for_all_cpus(h, 0);

7946
	rc = hpsa_pci_init(h);
R
Robert Elliott 已提交
7947
	if (rc)
7948 7949 7950 7951 7952 7953 7954
		goto clean2;	/* lu, aer/h */

	/* relies on h-> settings made by hpsa_pci_init, including
	 * interrupt_mode h->intr */
	rc = hpsa_scsi_host_alloc(h);
	if (rc)
		goto clean2_5;	/* pci, lu, aer/h */
7955

7956
	sprintf(h->devname, HPSA "%d", h->scsi_host->host_no);
7957 7958 7959 7960
	h->ctlr = number_of_controllers;
	number_of_controllers++;

	/* configure PCI DMA stuff */
7961 7962
	rc = pci_set_dma_mask(pdev, DMA_BIT_MASK(64));
	if (rc == 0) {
7963
		dac = 1;
7964 7965 7966 7967 7968 7969
	} else {
		rc = pci_set_dma_mask(pdev, DMA_BIT_MASK(32));
		if (rc == 0) {
			dac = 0;
		} else {
			dev_err(&pdev->dev, "no suitable DMA available\n");
7970
			goto clean3;	/* shost, pci, lu, aer/h */
7971
		}
7972 7973 7974 7975
	}

	/* make sure the board interrupts are off */
	h->access.set_intr_mask(h, HPSA_INTR_OFF);
7976

R
Robert Elliott 已提交
7977 7978
	rc = hpsa_request_irqs(h, do_hpsa_intr_msi, do_hpsa_intr_intx);
	if (rc)
7979
		goto clean3;	/* shost, pci, lu, aer/h */
7980
	rc = hpsa_alloc_cmd_pool(h);
7981
	if (rc)
7982
		goto clean4;	/* irq, shost, pci, lu, aer/h */
R
Robert Elliott 已提交
7983 7984
	rc = hpsa_alloc_sg_chain_blocks(h);
	if (rc)
7985
		goto clean5;	/* cmd, irq, shost, pci, lu, aer/h */
7986
	init_waitqueue_head(&h->scan_wait_queue);
S
Stephen Cameron 已提交
7987
	init_waitqueue_head(&h->abort_cmd_wait_queue);
W
Webb Scales 已提交
7988 7989
	init_waitqueue_head(&h->event_sync_wait_queue);
	mutex_init(&h->reset_mutex);
7990
	h->scan_finished = 1; /* no scan currently in progress */
7991 7992

	pci_set_drvdata(pdev, h);
7993
	h->ndevices = 0;
7994
	h->hba_mode_enabled = 0;
7995

7996
	spin_lock_init(&h->devlock);
R
Robert Elliott 已提交
7997 7998
	rc = hpsa_put_ctlr_into_performant_mode(h);
	if (rc)
7999 8000 8001 8002 8003 8004
		goto clean6; /* sg, cmd, irq, shost, pci, lu, aer/h */

	/* hook into SCSI subsystem */
	rc = hpsa_scsi_add_host(h);
	if (rc)
		goto clean7; /* perf, sg, cmd, irq, shost, pci, lu, aer/h */
8005 8006 8007 8008 8009 8010 8011 8012 8013 8014 8015 8016 8017

	/* create the resubmit workqueue */
	h->rescan_ctlr_wq = hpsa_create_controller_wq(h, "rescan");
	if (!h->rescan_ctlr_wq) {
		rc = -ENOMEM;
		goto clean7;
	}

	h->resubmit_wq = hpsa_create_controller_wq(h, "resubmit");
	if (!h->resubmit_wq) {
		rc = -ENOMEM;
		goto clean7;	/* aer/h */
	}
8018

R
Robert Elliott 已提交
8019 8020
	/*
	 * At this point, the controller is ready to take commands.
8021 8022 8023 8024 8025 8026 8027 8028 8029 8030 8031 8032 8033 8034 8035
	 * Now, if reset_devices and the hard reset didn't work, try
	 * the soft reset and see if that works.
	 */
	if (try_soft_reset) {

		/* This is kind of gross.  We may or may not get a completion
		 * from the soft reset command, and if we do, then the value
		 * from the fifo may or may not be valid.  So, we wait 10 secs
		 * after the reset throwing away any completions we get during
		 * that time.  Unregister the interrupt handler and register
		 * fake ones to scoop up any residual completions.
		 */
		spin_lock_irqsave(&h->lock, flags);
		h->access.set_intr_mask(h, HPSA_INTR_OFF);
		spin_unlock_irqrestore(&h->lock, flags);
8036
		hpsa_free_irqs(h);
8037
		rc = hpsa_request_irqs(h, hpsa_msix_discard_completions,
8038 8039
					hpsa_intx_discard_completions);
		if (rc) {
8040 8041
			dev_warn(&h->pdev->dev,
				"Failed to request_irq after soft reset.\n");
8042
			/*
8043 8044 8045 8046 8047 8048 8049 8050 8051
			 * cannot goto clean7 or free_irqs will be called
			 * again. Instead, do its work
			 */
			hpsa_free_performant_mode(h);	/* clean7 */
			hpsa_free_sg_chain_blocks(h);	/* clean6 */
			hpsa_free_cmd_pool(h);		/* clean5 */
			/*
			 * skip hpsa_free_irqs(h) clean4 since that
			 * was just called before request_irqs failed
8052 8053
			 */
			goto clean3;
8054 8055 8056 8057 8058
		}

		rc = hpsa_kdump_soft_reset(h);
		if (rc)
			/* Neither hard nor soft reset worked, we're hosed. */
8059
			goto clean9;
8060 8061 8062 8063 8064 8065 8066 8067 8068 8069 8070 8071 8072 8073 8074 8075 8076 8077 8078 8079

		dev_info(&h->pdev->dev, "Board READY.\n");
		dev_info(&h->pdev->dev,
			"Waiting for stale completions to drain.\n");
		h->access.set_intr_mask(h, HPSA_INTR_ON);
		msleep(10000);
		h->access.set_intr_mask(h, HPSA_INTR_OFF);

		rc = controller_reset_failed(h->cfgtable);
		if (rc)
			dev_info(&h->pdev->dev,
				"Soft reset appears to have failed.\n");

		/* since the controller's reset, we have to go back and re-init
		 * everything.  Easiest to just forget what we've done and do it
		 * all over again.
		 */
		hpsa_undo_allocations_after_kdump_soft_reset(h);
		try_soft_reset = 0;
		if (rc)
8080
			/* don't goto clean, we already unallocated */
8081 8082 8083 8084
			return -ENODEV;

		goto reinit_after_soft_reset;
	}
8085

R
Robert Elliott 已提交
8086 8087
	/* Enable Accelerated IO path at driver layer */
	h->acciopath_status = 1;
8088

8089

8090 8091 8092
	/* Turn the interrupts on so we can service requests */
	h->access.set_intr_mask(h, HPSA_INTR_ON);

8093
	hpsa_hba_inquiry(h);
8094 8095 8096 8097 8098 8099

	/* Monitor the controller for firmware lockups */
	h->heartbeat_sample_interval = HEARTBEAT_SAMPLE_INTERVAL;
	INIT_DELAYED_WORK(&h->monitor_ctlr_work, hpsa_monitor_ctlr_worker);
	schedule_delayed_work(&h->monitor_ctlr_work,
				h->heartbeat_sample_interval);
8100 8101 8102
	INIT_DELAYED_WORK(&h->rescan_ctlr_work, hpsa_rescan_ctlr_worker);
	queue_delayed_work(h->rescan_ctlr_wq, &h->rescan_ctlr_work,
				h->heartbeat_sample_interval);
8103
	return 0;
8104

8105
clean9: /* wq, sh, perf, sg, cmd, irq, shost, pci, lu, aer/h */
R
Robert Elliott 已提交
8106
	kfree(h->hba_inquiry_data);
8107
clean7: /* perf, sg, cmd, irq, shost, pci, lu, aer/h */
R
Robert Elliott 已提交
8108 8109 8110
	hpsa_free_performant_mode(h);
	h->access.set_intr_mask(h, HPSA_INTR_OFF);
clean6: /* sg, cmd, irq, pci, lockup, wq/aer/h */
8111
	hpsa_free_sg_chain_blocks(h);
8112
clean5: /* cmd, irq, shost, pci, lu, aer/h */
8113
	hpsa_free_cmd_pool(h);
8114
clean4: /* irq, shost, pci, lu, aer/h */
8115
	hpsa_free_irqs(h);
8116 8117 8118 8119
clean3: /* shost, pci, lu, aer/h */
	scsi_host_put(h->scsi_host);
	h->scsi_host = NULL;
clean2_5: /* pci, lu, aer/h */
R
Robert Elliott 已提交
8120
	hpsa_free_pci_init(h);
8121
clean2: /* lu, aer/h */
R
Robert Elliott 已提交
8122 8123 8124 8125 8126 8127
	if (h->lockup_detected) {
		free_percpu(h->lockup_detected);
		h->lockup_detected = NULL;
	}
clean1:	/* wq/aer/h */
	if (h->resubmit_wq) {
8128
		destroy_workqueue(h->resubmit_wq);
R
Robert Elliott 已提交
8129 8130 8131
		h->resubmit_wq = NULL;
	}
	if (h->rescan_ctlr_wq) {
8132
		destroy_workqueue(h->rescan_ctlr_wq);
R
Robert Elliott 已提交
8133 8134
		h->rescan_ctlr_wq = NULL;
	}
8135
	kfree(h);
8136
	return rc;
8137 8138 8139 8140 8141 8142
}

static void hpsa_flush_cache(struct ctlr_info *h)
{
	char *flush_buf;
	struct CommandList *c;
8143
	int rc;
8144

8145
	if (unlikely(lockup_detected(h)))
8146
		return;
8147 8148 8149 8150
	flush_buf = kzalloc(4, GFP_KERNEL);
	if (!flush_buf)
		return;

8151
	c = cmd_alloc(h);
8152

8153 8154 8155 8156
	if (fill_cmd(c, HPSA_CACHE_FLUSH, h, flush_buf, 4, 0,
		RAID_CTLR_LUNID, TYPE_CMD)) {
		goto out;
	}
8157 8158 8159 8160
	rc = hpsa_scsi_do_simple_cmd_with_retry(h, c,
					PCI_DMA_TODEVICE, NO_TIMEOUT);
	if (rc)
		goto out;
8161
	if (c->err_info->CommandStatus != 0)
8162
out:
8163 8164
		dev_warn(&h->pdev->dev,
			"error flushing cache on controller\n");
8165
	cmd_free(h, c);
8166 8167 8168 8169 8170 8171 8172 8173 8174 8175 8176 8177 8178 8179
	kfree(flush_buf);
}

static void hpsa_shutdown(struct pci_dev *pdev)
{
	struct ctlr_info *h;

	h = pci_get_drvdata(pdev);
	/* Turn board interrupts off  and send the flush cache command
	 * sendcmd will turn off interrupt, and send the flush...
	 * To write all data in the battery backed cache to disks
	 */
	hpsa_flush_cache(h);
	h->access.set_intr_mask(h, HPSA_INTR_OFF);
R
Robert Elliott 已提交
8180
	hpsa_free_irqs(h);			/* init_one 4 */
8181
	hpsa_disable_interrupt_mode(h);		/* pci_init 2 */
8182 8183
}

8184
static void hpsa_free_device_info(struct ctlr_info *h)
8185 8186 8187
{
	int i;

R
Robert Elliott 已提交
8188
	for (i = 0; i < h->ndevices; i++) {
8189
		kfree(h->dev[i]);
R
Robert Elliott 已提交
8190 8191
		h->dev[i] = NULL;
	}
8192 8193
}

8194
static void hpsa_remove_one(struct pci_dev *pdev)
8195 8196
{
	struct ctlr_info *h;
8197
	unsigned long flags;
8198 8199

	if (pci_get_drvdata(pdev) == NULL) {
8200
		dev_err(&pdev->dev, "unable to remove device\n");
8201 8202 8203
		return;
	}
	h = pci_get_drvdata(pdev);
8204 8205 8206 8207 8208

	/* Get rid of any controller monitoring work items */
	spin_lock_irqsave(&h->lock, flags);
	h->remove_in_progress = 1;
	spin_unlock_irqrestore(&h->lock, flags);
8209 8210 8211 8212
	cancel_delayed_work_sync(&h->monitor_ctlr_work);
	cancel_delayed_work_sync(&h->rescan_ctlr_work);
	destroy_workqueue(h->rescan_ctlr_wq);
	destroy_workqueue(h->resubmit_wq);
8213

R
Robert Elliott 已提交
8214
	/* includes hpsa_free_irqs - init_one 4 */
R
Robert Elliott 已提交
8215
	/* includes hpsa_disable_interrupt_mode - pci_init 2 */
8216
	hpsa_shutdown(pdev);
8217

R
Robert Elliott 已提交
8218 8219
	hpsa_free_device_info(h);		/* scan */

8220 8221 8222 8223 8224
	kfree(h->hba_inquiry_data);			/* init_one 10 */
	h->hba_inquiry_data = NULL;			/* init_one 10 */
	if (h->scsi_host)
		scsi_remove_host(h->scsi_host);		/* init_one 8 */
	hpsa_free_ioaccel2_sg_chain_blocks(h);
R
Robert Elliott 已提交
8225 8226 8227 8228 8229
	hpsa_free_performant_mode(h);			/* init_one 7 */
	hpsa_free_sg_chain_blocks(h);			/* init_one 6 */
	hpsa_free_cmd_pool(h);				/* init_one 5 */

	/* hpsa_free_irqs already called via hpsa_shutdown init_one 4 */
R
Robert Elliott 已提交
8230

8231 8232 8233
	scsi_host_put(h->scsi_host);			/* init_one 3 */
	h->scsi_host = NULL;				/* init_one 3 */

R
Robert Elliott 已提交
8234
	/* includes hpsa_disable_interrupt_mode - pci_init 2 */
8235
	hpsa_free_pci_init(h);				/* init_one 2.5 */
R
Robert Elliott 已提交
8236

R
Robert Elliott 已提交
8237 8238 8239 8240
	free_percpu(h->lockup_detected);		/* init_one 2 */
	h->lockup_detected = NULL;			/* init_one 2 */
	/* (void) pci_disable_pcie_error_reporting(pdev); */	/* init_one 1 */
	kfree(h);					/* init_one 1 */
8241 8242 8243 8244 8245 8246 8247 8248 8249 8250 8251 8252 8253 8254
}

static int hpsa_suspend(__attribute__((unused)) struct pci_dev *pdev,
	__attribute__((unused)) pm_message_t state)
{
	return -ENOSYS;
}

static int hpsa_resume(__attribute__((unused)) struct pci_dev *pdev)
{
	return -ENOSYS;
}

static struct pci_driver hpsa_pci_driver = {
8255
	.name = HPSA,
8256
	.probe = hpsa_init_one,
8257
	.remove = hpsa_remove_one,
8258 8259 8260 8261 8262 8263
	.id_table = hpsa_pci_device_id,	/* id_table */
	.shutdown = hpsa_shutdown,
	.suspend = hpsa_suspend,
	.resume = hpsa_resume,
};

8264 8265 8266 8267 8268 8269 8270 8271 8272 8273 8274 8275 8276
/* Fill in bucket_map[], given nsgs (the max number of
 * scatter gather elements supported) and bucket[],
 * which is an array of 8 integers.  The bucket[] array
 * contains 8 different DMA transfer sizes (in 16
 * byte increments) which the controller uses to fetch
 * commands.  This function fills in bucket_map[], which
 * maps a given number of scatter gather elements to one of
 * the 8 DMA transfer sizes.  The point of it is to allow the
 * controller to only do as much DMA as needed to fetch the
 * command, with the DMA transfer size encoded in the lower
 * bits of the command address.
 */
static void  calc_bucket_map(int bucket[], int num_buckets,
D
Don Brace 已提交
8277
	int nsgs, int min_blocks, u32 *bucket_map)
8278 8279 8280 8281 8282 8283
{
	int i, j, b, size;

	/* Note, bucket_map must have nsgs+1 entries. */
	for (i = 0; i <= nsgs; i++) {
		/* Compute size of a command with i SG entries */
8284
		size = i + min_blocks;
8285 8286
		b = num_buckets; /* Assume the biggest bucket */
		/* Find the bucket that is just big enough */
8287
		for (j = 0; j < num_buckets; j++) {
8288 8289 8290 8291 8292 8293 8294 8295 8296 8297
			if (bucket[j] >= size) {
				b = j;
				break;
			}
		}
		/* for a command with i SG entries, use bucket b. */
		bucket_map[i] = b;
	}
}

R
Robert Elliott 已提交
8298 8299 8300 8301
/*
 * return -ENODEV on err, 0 on success (or no action)
 * allocates numerous items that must be freed later
 */
8302
static int hpsa_enter_performant_mode(struct ctlr_info *h, u32 trans_support)
8303
{
8304 8305
	int i;
	unsigned long register_value;
8306 8307
	unsigned long transMethod = CFGTBL_Trans_Performant |
			(trans_support & CFGTBL_Trans_use_short_tags) |
8308 8309 8310
				CFGTBL_Trans_enable_directed_msix |
			(trans_support & (CFGTBL_Trans_io_accel1 |
				CFGTBL_Trans_io_accel2));
8311
	struct access_method access = SA5_performant_access;
8312 8313 8314 8315 8316 8317 8318 8319 8320 8321 8322

	/* This is a bit complicated.  There are 8 registers on
	 * the controller which we write to to tell it 8 different
	 * sizes of commands which there may be.  It's a way of
	 * reducing the DMA done to fetch each command.  Encoded into
	 * each command's tag are 3 bits which communicate to the controller
	 * which of the eight sizes that command fits within.  The size of
	 * each command depends on how many scatter gather entries there are.
	 * Each SG entry requires 16 bytes.  The eight registers are programmed
	 * with the number of 16-byte blocks a command of that size requires.
	 * The smallest command possible requires 5 such 16 byte blocks.
8323
	 * the largest command possible requires SG_ENTRIES_IN_CMD + 4 16-byte
8324 8325 8326 8327 8328 8329
	 * blocks.  Note, this only extends to the SG entries contained
	 * within the command block, and does not extend to chained blocks
	 * of SG elements.   bft[] contains the eight values we write to
	 * the registers.  They are not evenly distributed, but have more
	 * sizes for small commands, and fewer sizes for larger commands.
	 */
8330
	int bft[8] = {5, 6, 8, 10, 12, 20, 28, SG_ENTRIES_IN_CMD + 4};
8331 8332 8333 8334 8335 8336 8337 8338 8339 8340
#define MIN_IOACCEL2_BFT_ENTRY 5
#define HPSA_IOACCEL2_HEADER_SZ 4
	int bft2[16] = {MIN_IOACCEL2_BFT_ENTRY, 6, 7, 8, 9, 10, 11, 12,
			13, 14, 15, 16, 17, 18, 19,
			HPSA_IOACCEL2_HEADER_SZ + IOACCEL2_MAXSGENTRIES};
	BUILD_BUG_ON(ARRAY_SIZE(bft2) != 16);
	BUILD_BUG_ON(ARRAY_SIZE(bft) != 8);
	BUILD_BUG_ON(offsetof(struct io_accel2_cmd, sg) >
				 16 * MIN_IOACCEL2_BFT_ENTRY);
	BUILD_BUG_ON(sizeof(struct ioaccel2_sg_element) != 16);
8341
	BUILD_BUG_ON(28 > SG_ENTRIES_IN_CMD + 4);
8342 8343 8344 8345 8346 8347
	/*  5 = 1 s/g entry or 4k
	 *  6 = 2 s/g entry or 8k
	 *  8 = 4 s/g entry or 16k
	 * 10 = 6 s/g entry or 24k
	 */

8348 8349 8350 8351 8352 8353 8354
	/* If the controller supports either ioaccel method then
	 * we can also use the RAID stack submit path that does not
	 * perform the superfluous readl() after each command submission.
	 */
	if (trans_support & (CFGTBL_Trans_io_accel1 | CFGTBL_Trans_io_accel2))
		access = SA5_performant_access_no_read;

8355
	/* Controller spec: zero out this buffer. */
8356 8357
	for (i = 0; i < h->nreply_queues; i++)
		memset(h->reply_queue[i].head, 0, h->reply_queue_size);
8358

8359 8360
	bft[7] = SG_ENTRIES_IN_CMD + 4;
	calc_bucket_map(bft, ARRAY_SIZE(bft),
8361
				SG_ENTRIES_IN_CMD, 4, h->blockFetchTable);
8362 8363 8364 8365 8366
	for (i = 0; i < 8; i++)
		writel(bft[i], &h->transtable->BlockFetch[i]);

	/* size of controller ring buffer */
	writel(h->max_commands, &h->transtable->RepQSize);
8367
	writel(h->nreply_queues, &h->transtable->RepQCount);
8368 8369
	writel(0, &h->transtable->RepQCtrAddrLow32);
	writel(0, &h->transtable->RepQCtrAddrHigh32);
8370 8371 8372

	for (i = 0; i < h->nreply_queues; i++) {
		writel(0, &h->transtable->RepQAddr[i].upper);
8373
		writel(h->reply_queue[i].busaddr,
8374 8375 8376
			&h->transtable->RepQAddr[i].lower);
	}

8377
	writel(0, &h->cfgtable->HostWrite.command_pool_addr_hi);
8378 8379 8380 8381 8382 8383 8384 8385
	writel(transMethod, &(h->cfgtable->HostWrite.TransportRequest));
	/*
	 * enable outbound interrupt coalescing in accelerator mode;
	 */
	if (trans_support & CFGTBL_Trans_io_accel1) {
		access = SA5_ioaccel_mode1_access;
		writel(10, &h->cfgtable->HostWrite.CoalIntDelay);
		writel(4, &h->cfgtable->HostWrite.CoalIntCount);
8386 8387 8388 8389 8390 8391
	} else {
		if (trans_support & CFGTBL_Trans_io_accel2) {
			access = SA5_ioaccel_mode2_access;
			writel(10, &h->cfgtable->HostWrite.CoalIntDelay);
			writel(4, &h->cfgtable->HostWrite.CoalIntCount);
		}
8392
	}
8393
	writel(CFGTBL_ChangeReq, h->vaddr + SA5_DOORBELL);
8394 8395 8396 8397 8398
	if (hpsa_wait_for_mode_change_ack(h)) {
		dev_err(&h->pdev->dev,
			"performant mode problem - doorbell timeout\n");
		return -ENODEV;
	}
8399 8400
	register_value = readl(&(h->cfgtable->TransportActive));
	if (!(register_value & CFGTBL_Trans_Performant)) {
8401 8402
		dev_err(&h->pdev->dev,
			"performant mode problem - transport not active\n");
8403
		return -ENODEV;
8404
	}
8405
	/* Change the access methods to the performant access methods */
8406 8407 8408
	h->access = access;
	h->transMethod = transMethod;

8409 8410
	if (!((trans_support & CFGTBL_Trans_io_accel1) ||
		(trans_support & CFGTBL_Trans_io_accel2)))
8411
		return 0;
8412

8413 8414 8415 8416 8417 8418 8419 8420 8421 8422
	if (trans_support & CFGTBL_Trans_io_accel1) {
		/* Set up I/O accelerator mode */
		for (i = 0; i < h->nreply_queues; i++) {
			writel(i, h->vaddr + IOACCEL_MODE1_REPLY_QUEUE_INDEX);
			h->reply_queue[i].current_entry =
				readl(h->vaddr + IOACCEL_MODE1_PRODUCER_INDEX);
		}
		bft[7] = h->ioaccel_maxsg + 8;
		calc_bucket_map(bft, ARRAY_SIZE(bft), h->ioaccel_maxsg, 8,
				h->ioaccel1_blockFetchTable);
8423

8424
		/* initialize all reply queue entries to unused */
8425 8426 8427 8428
		for (i = 0; i < h->nreply_queues; i++)
			memset(h->reply_queue[i].head,
				(u8) IOACCEL_MODE1_REPLY_UNUSED,
				h->reply_queue_size);
8429

8430 8431 8432 8433 8434 8435 8436 8437 8438 8439 8440
		/* set all the constant fields in the accelerator command
		 * frames once at init time to save CPU cycles later.
		 */
		for (i = 0; i < h->nr_cmds; i++) {
			struct io_accel1_cmd *cp = &h->ioaccel_cmd_pool[i];

			cp->function = IOACCEL1_FUNCTION_SCSIIO;
			cp->err_info = (u32) (h->errinfo_pool_dhandle +
					(i * sizeof(struct ErrorInfo)));
			cp->err_info_len = sizeof(struct ErrorInfo);
			cp->sgl_offset = IOACCEL1_SGLOFFSET;
D
Don Brace 已提交
8441 8442
			cp->host_context_flags =
				cpu_to_le16(IOACCEL1_HCFLAGS_CISS_FORMAT);
8443 8444
			cp->timeout_sec = 0;
			cp->ReplyQueue = 0;
8445
			cp->tag =
8446
				cpu_to_le64((i << DIRECT_LOOKUP_SHIFT));
8447 8448
			cp->host_addr =
				cpu_to_le64(h->ioaccel_cmd_pool_dhandle +
8449 8450 8451 8452 8453 8454 8455 8456 8457 8458 8459 8460 8461 8462 8463 8464 8465 8466 8467 8468 8469 8470 8471 8472
					(i * sizeof(struct io_accel1_cmd)));
		}
	} else if (trans_support & CFGTBL_Trans_io_accel2) {
		u64 cfg_offset, cfg_base_addr_index;
		u32 bft2_offset, cfg_base_addr;
		int rc;

		rc = hpsa_find_cfg_addrs(h->pdev, h->vaddr, &cfg_base_addr,
			&cfg_base_addr_index, &cfg_offset);
		BUILD_BUG_ON(offsetof(struct io_accel2_cmd, sg) != 64);
		bft2[15] = h->ioaccel_maxsg + HPSA_IOACCEL2_HEADER_SZ;
		calc_bucket_map(bft2, ARRAY_SIZE(bft2), h->ioaccel_maxsg,
				4, h->ioaccel2_blockFetchTable);
		bft2_offset = readl(&h->cfgtable->io_accel_request_size_offset);
		BUILD_BUG_ON(offsetof(struct CfgTable,
				io_accel_request_size_offset) != 0xb8);
		h->ioaccel2_bft2_regs =
			remap_pci_mem(pci_resource_start(h->pdev,
					cfg_base_addr_index) +
					cfg_offset + bft2_offset,
					ARRAY_SIZE(bft2) *
					sizeof(*h->ioaccel2_bft2_regs));
		for (i = 0; i < ARRAY_SIZE(bft2); i++)
			writel(bft2[i], &h->ioaccel2_bft2_regs[i]);
8473
	}
8474
	writel(CFGTBL_ChangeReq, h->vaddr + SA5_DOORBELL);
8475 8476 8477 8478 8479 8480
	if (hpsa_wait_for_mode_change_ack(h)) {
		dev_err(&h->pdev->dev,
			"performant mode problem - enabling ioaccel mode\n");
		return -ENODEV;
	}
	return 0;
8481 8482
}

8483 8484 8485
/* Free ioaccel1 mode command blocks and block fetch table */
static void hpsa_free_ioaccel1_cmd_and_bft(struct ctlr_info *h)
{
R
Robert Elliott 已提交
8486
	if (h->ioaccel_cmd_pool) {
8487 8488 8489 8490
		pci_free_consistent(h->pdev,
			h->nr_cmds * sizeof(*h->ioaccel_cmd_pool),
			h->ioaccel_cmd_pool,
			h->ioaccel_cmd_pool_dhandle);
R
Robert Elliott 已提交
8491 8492 8493
		h->ioaccel_cmd_pool = NULL;
		h->ioaccel_cmd_pool_dhandle = 0;
	}
8494
	kfree(h->ioaccel1_blockFetchTable);
R
Robert Elliott 已提交
8495
	h->ioaccel1_blockFetchTable = NULL;
8496 8497
}

8498 8499
/* Allocate ioaccel1 mode command blocks and block fetch table */
static int hpsa_alloc_ioaccel1_cmd_and_bft(struct ctlr_info *h)
8500
{
8501 8502 8503 8504 8505
	h->ioaccel_maxsg =
		readl(&(h->cfgtable->io_accel_max_embedded_sg_count));
	if (h->ioaccel_maxsg > IOACCEL1_MAXSGENTRIES)
		h->ioaccel_maxsg = IOACCEL1_MAXSGENTRIES;

8506 8507 8508 8509 8510 8511 8512 8513 8514 8515 8516 8517
	/* Command structures must be aligned on a 128-byte boundary
	 * because the 7 lower bits of the address are used by the
	 * hardware.
	 */
	BUILD_BUG_ON(sizeof(struct io_accel1_cmd) %
			IOACCEL1_COMMANDLIST_ALIGNMENT);
	h->ioaccel_cmd_pool =
		pci_alloc_consistent(h->pdev,
			h->nr_cmds * sizeof(*h->ioaccel_cmd_pool),
			&(h->ioaccel_cmd_pool_dhandle));

	h->ioaccel1_blockFetchTable =
8518
		kmalloc(((h->ioaccel_maxsg + 1) *
8519 8520 8521 8522 8523 8524 8525 8526 8527 8528 8529
				sizeof(u32)), GFP_KERNEL);

	if ((h->ioaccel_cmd_pool == NULL) ||
		(h->ioaccel1_blockFetchTable == NULL))
		goto clean_up;

	memset(h->ioaccel_cmd_pool, 0,
		h->nr_cmds * sizeof(*h->ioaccel_cmd_pool));
	return 0;

clean_up:
8530
	hpsa_free_ioaccel1_cmd_and_bft(h);
8531
	return -ENOMEM;
8532 8533
}

8534 8535 8536
/* Free ioaccel2 mode command blocks and block fetch table */
static void hpsa_free_ioaccel2_cmd_and_bft(struct ctlr_info *h)
{
8537 8538
	hpsa_free_ioaccel2_sg_chain_blocks(h);

R
Robert Elliott 已提交
8539
	if (h->ioaccel2_cmd_pool) {
8540 8541 8542 8543
		pci_free_consistent(h->pdev,
			h->nr_cmds * sizeof(*h->ioaccel2_cmd_pool),
			h->ioaccel2_cmd_pool,
			h->ioaccel2_cmd_pool_dhandle);
R
Robert Elliott 已提交
8544 8545 8546
		h->ioaccel2_cmd_pool = NULL;
		h->ioaccel2_cmd_pool_dhandle = 0;
	}
8547
	kfree(h->ioaccel2_blockFetchTable);
R
Robert Elliott 已提交
8548
	h->ioaccel2_blockFetchTable = NULL;
8549 8550
}

8551 8552
/* Allocate ioaccel2 mode command blocks and block fetch table */
static int hpsa_alloc_ioaccel2_cmd_and_bft(struct ctlr_info *h)
8553
{
8554 8555
	int rc;

8556 8557 8558 8559 8560 8561 8562 8563 8564 8565 8566 8567 8568 8569 8570 8571 8572 8573 8574
	/* Allocate ioaccel2 mode command blocks and block fetch table */

	h->ioaccel_maxsg =
		readl(&(h->cfgtable->io_accel_max_embedded_sg_count));
	if (h->ioaccel_maxsg > IOACCEL2_MAXSGENTRIES)
		h->ioaccel_maxsg = IOACCEL2_MAXSGENTRIES;

	BUILD_BUG_ON(sizeof(struct io_accel2_cmd) %
			IOACCEL2_COMMANDLIST_ALIGNMENT);
	h->ioaccel2_cmd_pool =
		pci_alloc_consistent(h->pdev,
			h->nr_cmds * sizeof(*h->ioaccel2_cmd_pool),
			&(h->ioaccel2_cmd_pool_dhandle));

	h->ioaccel2_blockFetchTable =
		kmalloc(((h->ioaccel_maxsg + 1) *
				sizeof(u32)), GFP_KERNEL);

	if ((h->ioaccel2_cmd_pool == NULL) ||
8575 8576 8577 8578 8579 8580 8581
		(h->ioaccel2_blockFetchTable == NULL)) {
		rc = -ENOMEM;
		goto clean_up;
	}

	rc = hpsa_allocate_ioaccel2_sg_chain_blocks(h);
	if (rc)
8582 8583 8584 8585 8586 8587 8588
		goto clean_up;

	memset(h->ioaccel2_cmd_pool, 0,
		h->nr_cmds * sizeof(*h->ioaccel2_cmd_pool));
	return 0;

clean_up:
8589
	hpsa_free_ioaccel2_cmd_and_bft(h);
8590
	return rc;
8591 8592
}

R
Robert Elliott 已提交
8593 8594 8595 8596 8597 8598 8599 8600 8601 8602 8603 8604 8605 8606
/* Free items allocated by hpsa_put_ctlr_into_performant_mode */
static void hpsa_free_performant_mode(struct ctlr_info *h)
{
	kfree(h->blockFetchTable);
	h->blockFetchTable = NULL;
	hpsa_free_reply_queues(h);
	hpsa_free_ioaccel1_cmd_and_bft(h);
	hpsa_free_ioaccel2_cmd_and_bft(h);
}

/* return -ENODEV on error, 0 on success (or no action)
 * allocates numerous items that must be freed later
 */
static int hpsa_put_ctlr_into_performant_mode(struct ctlr_info *h)
8607 8608
{
	u32 trans_support;
8609 8610
	unsigned long transMethod = CFGTBL_Trans_Performant |
					CFGTBL_Trans_use_short_tags;
R
Robert Elliott 已提交
8611
	int i, rc;
8612

8613
	if (hpsa_simple_mode)
R
Robert Elliott 已提交
8614
		return 0;
8615

8616 8617
	trans_support = readl(&(h->cfgtable->TransportSupport));
	if (!(trans_support & PERFORMANT_MODE))
R
Robert Elliott 已提交
8618
		return 0;
8619

8620 8621 8622 8623
	/* Check for I/O accelerator mode support */
	if (trans_support & CFGTBL_Trans_io_accel1) {
		transMethod |= CFGTBL_Trans_io_accel1 |
				CFGTBL_Trans_enable_directed_msix;
R
Robert Elliott 已提交
8624 8625 8626 8627 8628
		rc = hpsa_alloc_ioaccel1_cmd_and_bft(h);
		if (rc)
			return rc;
	} else if (trans_support & CFGTBL_Trans_io_accel2) {
		transMethod |= CFGTBL_Trans_io_accel2 |
8629
				CFGTBL_Trans_enable_directed_msix;
R
Robert Elliott 已提交
8630 8631 8632
		rc = hpsa_alloc_ioaccel2_cmd_and_bft(h);
		if (rc)
			return rc;
8633 8634
	}

8635
	h->nreply_queues = h->msix_vector > 0 ? h->msix_vector : 1;
8636
	hpsa_get_max_perf_mode_cmds(h);
8637
	/* Performant mode ring buffer and supporting data structures */
8638
	h->reply_queue_size = h->max_commands * sizeof(u64);
8639

8640
	for (i = 0; i < h->nreply_queues; i++) {
8641 8642 8643
		h->reply_queue[i].head = pci_alloc_consistent(h->pdev,
						h->reply_queue_size,
						&(h->reply_queue[i].busaddr));
R
Robert Elliott 已提交
8644 8645 8646 8647
		if (!h->reply_queue[i].head) {
			rc = -ENOMEM;
			goto clean1;	/* rq, ioaccel */
		}
8648 8649 8650 8651 8652
		h->reply_queue[i].size = h->max_commands;
		h->reply_queue[i].wraparound = 1;  /* spec: init to 1 */
		h->reply_queue[i].current_entry = 0;
	}

8653
	/* Need a block fetch table for performant mode */
8654
	h->blockFetchTable = kmalloc(((SG_ENTRIES_IN_CMD + 1) *
8655
				sizeof(u32)), GFP_KERNEL);
R
Robert Elliott 已提交
8656 8657 8658 8659
	if (!h->blockFetchTable) {
		rc = -ENOMEM;
		goto clean1;	/* rq, ioaccel */
	}
8660

R
Robert Elliott 已提交
8661 8662 8663 8664
	rc = hpsa_enter_performant_mode(h, trans_support);
	if (rc)
		goto clean2;	/* bft, rq, ioaccel */
	return 0;
8665

R
Robert Elliott 已提交
8666
clean2:	/* bft, rq, ioaccel */
8667
	kfree(h->blockFetchTable);
R
Robert Elliott 已提交
8668 8669 8670 8671 8672 8673
	h->blockFetchTable = NULL;
clean1:	/* rq, ioaccel */
	hpsa_free_reply_queues(h);
	hpsa_free_ioaccel1_cmd_and_bft(h);
	hpsa_free_ioaccel2_cmd_and_bft(h);
	return rc;
8674 8675
}

8676
static int is_accelerated_cmd(struct CommandList *c)
8677
{
8678 8679 8680 8681 8682 8683
	return c->cmd_type == CMD_IOACCEL1 || c->cmd_type == CMD_IOACCEL2;
}

static void hpsa_drain_accel_commands(struct ctlr_info *h)
{
	struct CommandList *c = NULL;
8684
	int i, accel_cmds_out;
8685
	int refcount;
8686

8687
	do { /* wait for all outstanding ioaccel commands to drain out */
8688
		accel_cmds_out = 0;
8689 8690
		for (i = 0; i < h->nr_cmds; i++) {
			c = h->cmd_pool + i;
8691 8692 8693 8694
			refcount = atomic_inc_return(&c->refcount);
			if (refcount > 1) /* Command is allocated */
				accel_cmds_out += is_accelerated_cmd(c);
			cmd_free(h, c);
8695
		}
8696
		if (accel_cmds_out <= 0)
8697
			break;
8698 8699 8700 8701
		msleep(100);
	} while (1);
}

8702 8703 8704 8705 8706 8707
/*
 *  This is it.  Register the PCI driver information for the cards we control
 *  the OS will call our registered routines when it finds one of our cards.
 */
static int __init hpsa_init(void)
{
M
Mike Miller 已提交
8708
	return pci_register_driver(&hpsa_pci_driver);
8709 8710 8711 8712 8713 8714 8715
}

static void __exit hpsa_cleanup(void)
{
	pci_unregister_driver(&hpsa_pci_driver);
}

8716 8717
static void __attribute__((unused)) verify_offsets(void)
{
8718 8719 8720 8721 8722 8723 8724 8725 8726 8727 8728 8729 8730 8731 8732 8733 8734 8735 8736 8737 8738 8739
#define VERIFY_OFFSET(member, offset) \
	BUILD_BUG_ON(offsetof(struct raid_map_data, member) != offset)

	VERIFY_OFFSET(structure_size, 0);
	VERIFY_OFFSET(volume_blk_size, 4);
	VERIFY_OFFSET(volume_blk_cnt, 8);
	VERIFY_OFFSET(phys_blk_shift, 16);
	VERIFY_OFFSET(parity_rotation_shift, 17);
	VERIFY_OFFSET(strip_size, 18);
	VERIFY_OFFSET(disk_starting_blk, 20);
	VERIFY_OFFSET(disk_blk_cnt, 28);
	VERIFY_OFFSET(data_disks_per_row, 36);
	VERIFY_OFFSET(metadata_disks_per_row, 38);
	VERIFY_OFFSET(row_cnt, 40);
	VERIFY_OFFSET(layout_map_count, 42);
	VERIFY_OFFSET(flags, 44);
	VERIFY_OFFSET(dekindex, 46);
	/* VERIFY_OFFSET(reserved, 48 */
	VERIFY_OFFSET(data, 64);

#undef VERIFY_OFFSET

8740 8741 8742 8743 8744 8745 8746 8747 8748 8749 8750 8751 8752 8753 8754 8755 8756 8757 8758 8759 8760 8761
#define VERIFY_OFFSET(member, offset) \
	BUILD_BUG_ON(offsetof(struct io_accel2_cmd, member) != offset)

	VERIFY_OFFSET(IU_type, 0);
	VERIFY_OFFSET(direction, 1);
	VERIFY_OFFSET(reply_queue, 2);
	/* VERIFY_OFFSET(reserved1, 3);  */
	VERIFY_OFFSET(scsi_nexus, 4);
	VERIFY_OFFSET(Tag, 8);
	VERIFY_OFFSET(cdb, 16);
	VERIFY_OFFSET(cciss_lun, 32);
	VERIFY_OFFSET(data_len, 40);
	VERIFY_OFFSET(cmd_priority_task_attr, 44);
	VERIFY_OFFSET(sg_count, 45);
	/* VERIFY_OFFSET(reserved3 */
	VERIFY_OFFSET(err_ptr, 48);
	VERIFY_OFFSET(err_len, 56);
	/* VERIFY_OFFSET(reserved4  */
	VERIFY_OFFSET(sg, 64);

#undef VERIFY_OFFSET

8762 8763 8764 8765 8766 8767 8768 8769 8770 8771 8772 8773 8774 8775 8776 8777 8778 8779 8780 8781 8782 8783 8784 8785 8786
#define VERIFY_OFFSET(member, offset) \
	BUILD_BUG_ON(offsetof(struct io_accel1_cmd, member) != offset)

	VERIFY_OFFSET(dev_handle, 0x00);
	VERIFY_OFFSET(reserved1, 0x02);
	VERIFY_OFFSET(function, 0x03);
	VERIFY_OFFSET(reserved2, 0x04);
	VERIFY_OFFSET(err_info, 0x0C);
	VERIFY_OFFSET(reserved3, 0x10);
	VERIFY_OFFSET(err_info_len, 0x12);
	VERIFY_OFFSET(reserved4, 0x13);
	VERIFY_OFFSET(sgl_offset, 0x14);
	VERIFY_OFFSET(reserved5, 0x15);
	VERIFY_OFFSET(transfer_len, 0x1C);
	VERIFY_OFFSET(reserved6, 0x20);
	VERIFY_OFFSET(io_flags, 0x24);
	VERIFY_OFFSET(reserved7, 0x26);
	VERIFY_OFFSET(LUN, 0x34);
	VERIFY_OFFSET(control, 0x3C);
	VERIFY_OFFSET(CDB, 0x40);
	VERIFY_OFFSET(reserved8, 0x50);
	VERIFY_OFFSET(host_context_flags, 0x60);
	VERIFY_OFFSET(timeout_sec, 0x62);
	VERIFY_OFFSET(ReplyQueue, 0x64);
	VERIFY_OFFSET(reserved9, 0x65);
8787
	VERIFY_OFFSET(tag, 0x68);
8788 8789 8790 8791 8792 8793
	VERIFY_OFFSET(host_addr, 0x70);
	VERIFY_OFFSET(CISS_LUN, 0x78);
	VERIFY_OFFSET(SG, 0x78 + 8);
#undef VERIFY_OFFSET
}

8794 8795
module_init(hpsa_init);
module_exit(hpsa_cleanup);