cciss.c 148.4 KB
Newer Older
L
Linus Torvalds 已提交
1
/*
M
Mike Miller 已提交
2 3
 *    Disk Array driver for HP Smart Array controllers.
 *    (C) Copyright 2000, 2007 Hewlett-Packard Development Company, L.P.
L
Linus Torvalds 已提交
4 5 6
 *
 *    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
M
Mike Miller 已提交
7
 *    the Free Software Foundation; version 2 of the License.
L
Linus Torvalds 已提交
8 9 10
 *
 *    This program is distributed in the hope that it will be useful,
 *    but WITHOUT ANY WARRANTY; without even the implied warranty of
M
Mike Miller 已提交
11 12
 *    MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. See the GNU
 *    General Public License for more details.
L
Linus Torvalds 已提交
13 14 15
 *
 *    You should have received a copy of the GNU General Public License
 *    along with this program; if not, write to the Free Software
M
Mike Miller 已提交
16 17
 *    Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
 *    02111-1307, USA.
L
Linus Torvalds 已提交
18 19 20 21 22 23 24 25 26
 *
 *    Questions/Comments/Bugfixes to iss_storagedev@hp.com
 *
 */

#include <linux/module.h>
#include <linux/interrupt.h>
#include <linux/types.h>
#include <linux/pci.h>
27
#include <linux/pci-aspm.h>
L
Linus Torvalds 已提交
28 29 30 31 32 33 34 35 36
#include <linux/kernel.h>
#include <linux/slab.h>
#include <linux/delay.h>
#include <linux/major.h>
#include <linux/fs.h>
#include <linux/bio.h>
#include <linux/blkpg.h>
#include <linux/timer.h>
#include <linux/proc_fs.h>
37
#include <linux/seq_file.h>
38
#include <linux/init.h>
39
#include <linux/jiffies.h>
L
Linus Torvalds 已提交
40 41 42
#include <linux/hdreg.h>
#include <linux/spinlock.h>
#include <linux/compat.h>
43
#include <linux/mutex.h>
L
Linus Torvalds 已提交
44 45 46
#include <asm/uaccess.h>
#include <asm/io.h>

47
#include <linux/dma-mapping.h>
L
Linus Torvalds 已提交
48 49 50
#include <linux/blkdev.h>
#include <linux/genhd.h>
#include <linux/completion.h>
51
#include <scsi/scsi.h>
52 53 54
#include <scsi/sg.h>
#include <scsi/scsi_ioctl.h>
#include <linux/cdrom.h>
55
#include <linux/scatterlist.h>
56
#include <linux/kthread.h>
L
Linus Torvalds 已提交
57 58

#define CCISS_DRIVER_VERSION(maj,min,submin) ((maj<<16)|(min<<8)|(submin))
59 60
#define DRIVER_NAME "HP CISS Driver (v 3.6.26)"
#define DRIVER_VERSION CCISS_DRIVER_VERSION(3, 6, 26)
L
Linus Torvalds 已提交
61 62 63

/* Embedded module documentation macros - see modules.h */
MODULE_AUTHOR("Hewlett-Packard Company");
M
Mike Miller 已提交
64
MODULE_DESCRIPTION("Driver for HP Smart Array Controllers");
65 66
MODULE_SUPPORTED_DEVICE("HP Smart Array Controllers");
MODULE_VERSION("3.6.26");
L
Linus Torvalds 已提交
67
MODULE_LICENSE("GPL");
68 69 70 71
static int cciss_tape_cmds = 6;
module_param(cciss_tape_cmds, int, 0644);
MODULE_PARM_DESC(cciss_tape_cmds,
	"number of commands to allocate for tape devices (default: 6)");
72 73 74 75
static int cciss_simple_mode;
module_param(cciss_simple_mode, int, S_IRUGO|S_IWUSR);
MODULE_PARM_DESC(cciss_simple_mode,
	"Use 'simple mode' rather than 'performant mode'");
L
Linus Torvalds 已提交
76

77
static DEFINE_MUTEX(cciss_mutex);
J
Jens Axboe 已提交
78
static struct proc_dir_entry *proc_cciss;
79

L
Linus Torvalds 已提交
80 81 82 83 84 85
#include "cciss_cmd.h"
#include "cciss.h"
#include <linux/cciss_ioctl.h>

/* define the PCI info for the cards we can control */
static const struct pci_device_id cciss_pci_device_id[] = {
86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103
	{PCI_VENDOR_ID_COMPAQ, PCI_DEVICE_ID_COMPAQ_CISS,  0x0E11, 0x4070},
	{PCI_VENDOR_ID_COMPAQ, PCI_DEVICE_ID_COMPAQ_CISSB, 0x0E11, 0x4080},
	{PCI_VENDOR_ID_COMPAQ, PCI_DEVICE_ID_COMPAQ_CISSB, 0x0E11, 0x4082},
	{PCI_VENDOR_ID_COMPAQ, PCI_DEVICE_ID_COMPAQ_CISSB, 0x0E11, 0x4083},
	{PCI_VENDOR_ID_COMPAQ, PCI_DEVICE_ID_COMPAQ_CISSC, 0x0E11, 0x4091},
	{PCI_VENDOR_ID_COMPAQ, PCI_DEVICE_ID_COMPAQ_CISSC, 0x0E11, 0x409A},
	{PCI_VENDOR_ID_COMPAQ, PCI_DEVICE_ID_COMPAQ_CISSC, 0x0E11, 0x409B},
	{PCI_VENDOR_ID_COMPAQ, PCI_DEVICE_ID_COMPAQ_CISSC, 0x0E11, 0x409C},
	{PCI_VENDOR_ID_COMPAQ, PCI_DEVICE_ID_COMPAQ_CISSC, 0x0E11, 0x409D},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSA,     0x103C, 0x3225},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSC,     0x103C, 0x3223},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSC,     0x103C, 0x3234},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSC,     0x103C, 0x3235},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSD,     0x103C, 0x3211},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSD,     0x103C, 0x3212},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSD,     0x103C, 0x3213},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSD,     0x103C, 0x3214},
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSD,     0x103C, 0x3215},
104
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSC,     0x103C, 0x3237},
105
	{PCI_VENDOR_ID_HP,     PCI_DEVICE_ID_HP_CISSC,     0x103C, 0x323D},
L
Linus Torvalds 已提交
106 107
	{0,}
};
108

L
Linus Torvalds 已提交
109 110 111 112
MODULE_DEVICE_TABLE(pci, cciss_pci_device_id);

/*  board_id = Subsystem Device ID & Vendor ID
 *  product = Marketing Name for the board
113
 *  access = Address of the struct of function pointers
L
Linus Torvalds 已提交
114 115
 */
static struct board_type products[] = {
116 117 118 119 120 121 122 123 124 125
	{0x40700E11, "Smart Array 5300", &SA5_access},
	{0x40800E11, "Smart Array 5i", &SA5B_access},
	{0x40820E11, "Smart Array 532", &SA5B_access},
	{0x40830E11, "Smart Array 5312", &SA5B_access},
	{0x409A0E11, "Smart Array 641", &SA5_access},
	{0x409B0E11, "Smart Array 642", &SA5_access},
	{0x409C0E11, "Smart Array 6400", &SA5_access},
	{0x409D0E11, "Smart Array 6400 EM", &SA5_access},
	{0x40910E11, "Smart Array 6i", &SA5_access},
	{0x3225103C, "Smart Array P600", &SA5_access},
126 127
	{0x3223103C, "Smart Array P800", &SA5_access},
	{0x3234103C, "Smart Array P400", &SA5_access},
128 129 130 131 132 133 134
	{0x3235103C, "Smart Array P400i", &SA5_access},
	{0x3211103C, "Smart Array E200i", &SA5_access},
	{0x3212103C, "Smart Array E200", &SA5_access},
	{0x3213103C, "Smart Array E200i", &SA5_access},
	{0x3214103C, "Smart Array E200i", &SA5_access},
	{0x3215103C, "Smart Array E200i", &SA5_access},
	{0x3237103C, "Smart Array E500", &SA5_access},
135 136
	{0x3223103C, "Smart Array P800", &SA5_access},
	{0x3234103C, "Smart Array P400", &SA5_access},
137
	{0x323D103C, "Smart Array P700m", &SA5_access},
L
Linus Torvalds 已提交
138 139
};

140
/* How long to wait (in milliseconds) for board to go into simple mode */
141
#define MAX_CONFIG_WAIT 30000
L
Linus Torvalds 已提交
142 143 144 145 146 147 148 149 150 151 152 153
#define MAX_IOCTL_CONFIG_WAIT 1000

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

#define MAX_CTLR	32

/* Originally cciss driver only supports 8 major numbers */
#define MAX_CTLR_ORIG 	8

static ctlr_info_t *hba[MAX_CTLR];

154 155 156 157
static struct task_struct *cciss_scan_thread;
static DEFINE_MUTEX(scan_mutex);
static LIST_HEAD(scan_q);

158
static void do_cciss_request(struct request_queue *q);
M
Mike Miller 已提交
159 160
static irqreturn_t do_cciss_intx(int irq, void *dev_id);
static irqreturn_t do_cciss_msix_intr(int irq, void *dev_id);
A
Al Viro 已提交
161
static int cciss_open(struct block_device *bdev, fmode_t mode);
162
static int cciss_unlocked_open(struct block_device *bdev, fmode_t mode);
A
Al Viro 已提交
163
static int cciss_release(struct gendisk *disk, fmode_t mode);
164 165
static int do_ioctl(struct block_device *bdev, fmode_t mode,
		    unsigned int cmd, unsigned long arg);
A
Al Viro 已提交
166
static int cciss_ioctl(struct block_device *bdev, fmode_t mode,
167
		       unsigned int cmd, unsigned long arg);
168
static int cciss_getgeo(struct block_device *bdev, struct hd_geometry *geo);
L
Linus Torvalds 已提交
169 170

static int cciss_revalidate(struct gendisk *disk);
171
static int rebuild_lun_table(ctlr_info_t *h, int first_time, int via_ioctl);
172
static int deregister_disk(ctlr_info_t *h, int drv_index,
173
			   int clear_all, int via_ioctl);
L
Linus Torvalds 已提交
174

175
static void cciss_read_capacity(ctlr_info_t *h, int logvol,
176
			sector_t *total_size, unsigned int *block_size);
177
static void cciss_read_capacity_16(ctlr_info_t *h, int logvol,
178
			sector_t *total_size, unsigned int *block_size);
179
static void cciss_geometry_inquiry(ctlr_info_t *h, int logvol,
180
			sector_t total_size,
181
			unsigned int block_size, InquiryData_struct *inq_buff,
182
				   drive_info_struct *drv);
183
static void __devinit cciss_interrupt_mode(ctlr_info_t *);
184
static int __devinit cciss_enter_simple_mode(struct ctlr_info *h);
185
static void start_io(ctlr_info_t *h);
186
static int sendcmd_withirq(ctlr_info_t *h, __u8 cmd, void *buff, size_t size,
187 188
			__u8 page_code, unsigned char scsi3addr[],
			int cmd_type);
189 190 191
static int sendcmd_withirq_core(ctlr_info_t *h, CommandList_struct *c,
	int attempt_retry);
static int process_sendcmd_error(ctlr_info_t *h, CommandList_struct *c);
L
Linus Torvalds 已提交
192

193
static int add_to_scan_list(struct ctlr_info *h);
194 195
static int scan_thread(void *data);
static int check_for_unit_attention(ctlr_info_t *h, CommandList_struct *c);
196 197
static void cciss_hba_release(struct device *dev);
static void cciss_device_release(struct device *dev);
198
static void cciss_free_gendisk(ctlr_info_t *h, int drv_index);
199
static void cciss_free_drive_info(ctlr_info_t *h, int drv_index);
200
static inline u32 next_command(ctlr_info_t *h);
S
Stephen M. Cameron 已提交
201 202 203 204 205
static int __devinit cciss_find_cfg_addrs(struct pci_dev *pdev,
	void __iomem *vaddr, u32 *cfg_base_addr, u64 *cfg_base_addr_index,
	u64 *cfg_offset);
static int __devinit cciss_pci_find_memory_BAR(struct pci_dev *pdev,
	unsigned long *memory_bar);
206
static inline u32 cciss_tag_discard_error_bits(ctlr_info_t *h, u32 tag);
207 208
static __devinit int write_driver_ver_to_cfgtable(
	CfgTable_struct __iomem *cfgtable);
209

210 211 212 213
/* performant mode helper functions */
static void  calc_bucket_map(int *bucket, int num_buckets, int nsgs,
				int *bucket_map);
static void cciss_put_controller_into_performant_mode(ctlr_info_t *h);
214

L
Linus Torvalds 已提交
215
#ifdef CONFIG_PROC_FS
216
static void cciss_procinit(ctlr_info_t *h);
L
Linus Torvalds 已提交
217
#else
218
static void cciss_procinit(ctlr_info_t *h)
219 220 221
{
}
#endif				/* CONFIG_PROC_FS */
L
Linus Torvalds 已提交
222 223

#ifdef CONFIG_COMPAT
A
Al Viro 已提交
224 225
static int cciss_compat_ioctl(struct block_device *, fmode_t,
			      unsigned, unsigned long);
L
Linus Torvalds 已提交
226 227
#endif

228
static const struct block_device_operations cciss_fops = {
229
	.owner = THIS_MODULE,
230
	.open = cciss_unlocked_open,
A
Al Viro 已提交
231
	.release = cciss_release,
232
	.ioctl = do_ioctl,
233
	.getgeo = cciss_getgeo,
L
Linus Torvalds 已提交
234
#ifdef CONFIG_COMPAT
A
Al Viro 已提交
235
	.compat_ioctl = cciss_compat_ioctl,
L
Linus Torvalds 已提交
236
#endif
237
	.revalidate_disk = cciss_revalidate,
L
Linus Torvalds 已提交
238 239
};

240 241 242 243 244 245
/* set_performant_mode: Modify the tag for cciss performant
 * set bit 0 for pull model, bits 3-1 for block fetch
 * register number
 */
static void set_performant_mode(ctlr_info_t *h, CommandList_struct *c)
{
246
	if (likely(h->transMethod & CFGTBL_Trans_Performant))
247 248 249
		c->busaddr |= 1 | (h->blockFetchTable[c->Header.SGList] << 1);
}

L
Linus Torvalds 已提交
250 251 252
/*
 * Enqueuing and dequeuing functions for cmdlists.
 */
253
static inline void addQ(struct list_head *list, CommandList_struct *c)
L
Linus Torvalds 已提交
254
{
255
	list_add_tail(&c->list, list);
L
Linus Torvalds 已提交
256 257
}

258
static inline void removeQ(CommandList_struct *c)
L
Linus Torvalds 已提交
259
{
260 261 262 263 264 265 266 267
	/*
	 * After kexec/dump some commands might still
	 * be in flight, which the firmware will try
	 * to complete. Resetting the firmware doesn't work
	 * with old fw revisions, so we have to mark
	 * them off as 'stale' to prevent the driver from
	 * falling over.
	 */
268
	if (WARN_ON(list_empty(&c->list))) {
269
		c->cmd_type = CMD_MSG_STALE;
270
		return;
271
	}
272

273
	list_del_init(&c->list);
L
Linus Torvalds 已提交
274 275
}

M
Mike Miller 已提交
276 277 278 279
static void enqueue_cmd_and_start_io(ctlr_info_t *h,
	CommandList_struct *c)
{
	unsigned long flags;
280
	set_performant_mode(h, c);
M
Mike Miller 已提交
281 282 283
	spin_lock_irqsave(&h->lock, flags);
	addQ(&h->reqQ, c);
	h->Qdepth++;
284 285
	if (h->Qdepth > h->maxQsinceinit)
		h->maxQsinceinit = h->Qdepth;
M
Mike Miller 已提交
286 287 288 289
	start_io(h);
	spin_unlock_irqrestore(&h->lock, flags);
}

290
static void cciss_free_sg_chain_blocks(SGDescriptor_struct **cmd_sg_list,
291 292 293 294 295 296 297
	int nr_cmds)
{
	int i;

	if (!cmd_sg_list)
		return;
	for (i = 0; i < nr_cmds; i++) {
298 299
		kfree(cmd_sg_list[i]);
		cmd_sg_list[i] = NULL;
300 301 302 303
	}
	kfree(cmd_sg_list);
}

304 305
static SGDescriptor_struct **cciss_allocate_sg_chain_blocks(
	ctlr_info_t *h, int chainsize, int nr_cmds)
306 307
{
	int j;
308
	SGDescriptor_struct **cmd_sg_list;
309 310 311 312 313 314 315 316 317 318 319

	if (chainsize <= 0)
		return NULL;

	cmd_sg_list = kmalloc(sizeof(*cmd_sg_list) * nr_cmds, GFP_KERNEL);
	if (!cmd_sg_list)
		return NULL;

	/* Build up chain blocks for each command */
	for (j = 0; j < nr_cmds; j++) {
		/* Need a block of chainsized s/g elements. */
320 321 322
		cmd_sg_list[j] = kmalloc((chainsize *
			sizeof(*cmd_sg_list[j])), GFP_KERNEL);
		if (!cmd_sg_list[j]) {
323 324 325 326 327 328 329 330 331 332 333
			dev_err(&h->pdev->dev, "Cannot get memory "
				"for s/g chains.\n");
			goto clean;
		}
	}
	return cmd_sg_list;
clean:
	cciss_free_sg_chain_blocks(cmd_sg_list, nr_cmds);
	return NULL;
}

334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362
static void cciss_unmap_sg_chain_block(ctlr_info_t *h, CommandList_struct *c)
{
	SGDescriptor_struct *chain_sg;
	u64bit temp64;

	if (c->Header.SGTotal <= h->max_cmd_sgentries)
		return;

	chain_sg = &c->SG[h->max_cmd_sgentries - 1];
	temp64.val32.lower = chain_sg->Addr.lower;
	temp64.val32.upper = chain_sg->Addr.upper;
	pci_unmap_single(h->pdev, temp64.val, chain_sg->Len, PCI_DMA_TODEVICE);
}

static void cciss_map_sg_chain_block(ctlr_info_t *h, CommandList_struct *c,
	SGDescriptor_struct *chain_block, int len)
{
	SGDescriptor_struct *chain_sg;
	u64bit temp64;

	chain_sg = &c->SG[h->max_cmd_sgentries - 1];
	chain_sg->Ext = CCISS_SG_CHAIN;
	chain_sg->Len = len;
	temp64.val = pci_map_single(h->pdev, chain_block, len,
				PCI_DMA_TODEVICE);
	chain_sg->Addr.lower = temp64.val32.lower;
	chain_sg->Addr.upper = temp64.val32.upper;
}

L
Linus Torvalds 已提交
363 364
#include "cciss_scsi.c"		/* For SCSI tape support */

365 366 367
static const char *raid_label[] = { "0", "4", "1(1+0)", "5", "5+1", "ADG",
	"UNKNOWN"
};
K
Kulikov Vasiliy 已提交
368
#define RAID_UNKNOWN (ARRAY_SIZE(raid_label)-1)
R
Randy Dunlap 已提交
369

L
Linus Torvalds 已提交
370 371 372 373 374 375 376
#ifdef CONFIG_PROC_FS

/*
 * Report information about this controller.
 */
#define ENG_GIG 1000000000
#define ENG_GIG_FACTOR (ENG_GIG/512)
377
#define ENGAGE_SCSI	"engage scsi"
L
Linus Torvalds 已提交
378

379
static void cciss_seq_show_header(struct seq_file *seq)
L
Linus Torvalds 已提交
380
{
381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396
	ctlr_info_t *h = seq->private;

	seq_printf(seq, "%s: HP %s Controller\n"
		"Board ID: 0x%08lx\n"
		"Firmware Version: %c%c%c%c\n"
		"IRQ: %d\n"
		"Logical drives: %d\n"
		"Current Q depth: %d\n"
		"Current # commands on controller: %d\n"
		"Max Q depth since init: %d\n"
		"Max # commands on controller since init: %d\n"
		"Max SG entries since init: %d\n",
		h->devname,
		h->product_name,
		(unsigned long)h->board_id,
		h->firm_ver[0], h->firm_ver[1], h->firm_ver[2],
397
		h->firm_ver[3], (unsigned int)h->intr[h->intr_mode],
398 399 400 401 402
		h->num_luns,
		h->Qdepth, h->commands_outstanding,
		h->maxQsinceinit, h->max_outstanding, h->maxSG);

#ifdef CONFIG_CISS_SCSI_TAPE
403
	cciss_seq_tape_report(seq, h);
404 405
#endif /* CONFIG_CISS_SCSI_TAPE */
}
L
Linus Torvalds 已提交
406

407 408 409 410
static void *cciss_seq_start(struct seq_file *seq, loff_t *pos)
{
	ctlr_info_t *h = seq->private;
	unsigned long flags;
L
Linus Torvalds 已提交
411 412 413 414

	/* prevent displaying bogus info during configuration
	 * or deconfiguration of a logical volume
	 */
415
	spin_lock_irqsave(&h->lock, flags);
L
Linus Torvalds 已提交
416
	if (h->busy_configuring) {
417
		spin_unlock_irqrestore(&h->lock, flags);
418
		return ERR_PTR(-EBUSY);
L
Linus Torvalds 已提交
419 420
	}
	h->busy_configuring = 1;
421
	spin_unlock_irqrestore(&h->lock, flags);
L
Linus Torvalds 已提交
422

423 424 425 426 427 428 429 430 431 432 433 434
	if (*pos == 0)
		cciss_seq_show_header(seq);

	return pos;
}

static int cciss_seq_show(struct seq_file *seq, void *v)
{
	sector_t vol_sz, vol_sz_frac;
	ctlr_info_t *h = seq->private;
	unsigned ctlr = h->ctlr;
	loff_t *pos = v;
435
	drive_info_struct *drv = h->drv[*pos];
436 437 438 439

	if (*pos > h->highest_lun)
		return 0;

440 441 442
	if (drv == NULL) /* it's possible for h->drv[] to have holes. */
		return 0;

443 444 445 446 447 448 449 450
	if (drv->heads == 0)
		return 0;

	vol_sz = drv->nr_blocks;
	vol_sz_frac = sector_div(vol_sz, ENG_GIG_FACTOR);
	vol_sz_frac *= 100;
	sector_div(vol_sz_frac, ENG_GIG_FACTOR);

451
	if (drv->raid_level < 0 || drv->raid_level > RAID_UNKNOWN)
452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478
		drv->raid_level = RAID_UNKNOWN;
	seq_printf(seq, "cciss/c%dd%d:"
			"\t%4u.%02uGB\tRAID %s\n",
			ctlr, (int) *pos, (int)vol_sz, (int)vol_sz_frac,
			raid_label[drv->raid_level]);
	return 0;
}

static void *cciss_seq_next(struct seq_file *seq, void *v, loff_t *pos)
{
	ctlr_info_t *h = seq->private;

	if (*pos > h->highest_lun)
		return NULL;
	*pos += 1;

	return pos;
}

static void cciss_seq_stop(struct seq_file *seq, void *v)
{
	ctlr_info_t *h = seq->private;

	/* Only reset h->busy_configuring if we succeeded in setting
	 * it during cciss_seq_start. */
	if (v == ERR_PTR(-EBUSY))
		return;
479

L
Linus Torvalds 已提交
480 481 482
	h->busy_configuring = 0;
}

J
James Morris 已提交
483
static const struct seq_operations cciss_seq_ops = {
484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503
	.start = cciss_seq_start,
	.show  = cciss_seq_show,
	.next  = cciss_seq_next,
	.stop  = cciss_seq_stop,
};

static int cciss_seq_open(struct inode *inode, struct file *file)
{
	int ret = seq_open(file, &cciss_seq_ops);
	struct seq_file *seq = file->private_data;

	if (!ret)
		seq->private = PDE(inode)->data;

	return ret;
}

static ssize_t
cciss_proc_write(struct file *file, const char __user *buf,
		 size_t length, loff_t *ppos)
L
Linus Torvalds 已提交
504
{
505 506 507 508 509
	int err;
	char *buffer;

#ifndef CONFIG_CISS_SCSI_TAPE
	return -EINVAL;
L
Linus Torvalds 已提交
510 511
#endif

512
	if (!buf || length > PAGE_SIZE - 1)
513
		return -EINVAL;
514 515 516 517 518 519 520 521 522 523 524 525 526 527 528

	buffer = (char *)__get_free_page(GFP_KERNEL);
	if (!buffer)
		return -ENOMEM;

	err = -EFAULT;
	if (copy_from_user(buffer, buf, length))
		goto out;
	buffer[length] = '\0';

#ifdef CONFIG_CISS_SCSI_TAPE
	if (strncmp(ENGAGE_SCSI, buffer, sizeof ENGAGE_SCSI - 1) == 0) {
		struct seq_file *seq = file->private_data;
		ctlr_info_t *h = seq->private;

529
		err = cciss_engage_scsi(h);
530
		if (err == 0)
531 532 533 534
			err = length;
	} else
#endif /* CONFIG_CISS_SCSI_TAPE */
		err = -EINVAL;
535 536
	/* might be nice to have "disengage" too, but it's not
	   safely possible. (only 1 module use count, lock issues.) */
537 538 539 540

out:
	free_page((unsigned long)buffer);
	return err;
L
Linus Torvalds 已提交
541 542
}

543
static const struct file_operations cciss_proc_fops = {
544 545 546 547 548 549 550 551
	.owner	 = THIS_MODULE,
	.open    = cciss_seq_open,
	.read    = seq_read,
	.llseek  = seq_lseek,
	.release = seq_release,
	.write	 = cciss_proc_write,
};

552
static void __devinit cciss_procinit(ctlr_info_t *h)
L
Linus Torvalds 已提交
553 554 555
{
	struct proc_dir_entry *pde;

556
	if (proc_cciss == NULL)
A
Alexey Dobriyan 已提交
557
		proc_cciss = proc_mkdir("driver/cciss", NULL);
558 559
	if (!proc_cciss)
		return;
560
	pde = proc_create_data(h->devname, S_IWUSR | S_IRUSR | S_IRGRP |
561
					S_IROTH, proc_cciss,
562
					&cciss_proc_fops, h);
L
Linus Torvalds 已提交
563
}
564
#endif				/* CONFIG_PROC_FS */
L
Linus Torvalds 已提交
565

566 567 568 569 570
#define MAX_PRODUCT_NAME_LEN 19

#define to_hba(n) container_of(n, struct ctlr_info, dev)
#define to_drv(n) container_of(n, drive_info_struct, dev)

571
/* List of controllers which cannot be hard reset on kexec with reset_devices */
572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588
static u32 unresettable_controller[] = {
	0x324a103C, /* Smart Array P712m */
	0x324b103C, /* SmartArray P711m */
	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 */
	0x409C0E11, /* Smart Array 6400 */
	0x409D0E11, /* Smart Array 6400 EM */
};

589 590 591 592 593 594 595
/* List of controllers which cannot even be soft reset */
static u32 soft_unresettable_controller[] = {
	0x409C0E11, /* Smart Array 6400 */
	0x409D0E11, /* Smart Array 6400 EM */
};

static int ctlr_is_hard_resettable(u32 board_id)
596 597 598 599
{
	int i;

	for (i = 0; i < ARRAY_SIZE(unresettable_controller); i++)
600
		if (unresettable_controller[i] == board_id)
601 602 603 604
			return 0;
	return 1;
}

605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620
static int ctlr_is_soft_resettable(u32 board_id)
{
	int i;

	for (i = 0; i < ARRAY_SIZE(soft_unresettable_controller); i++)
		if (soft_unresettable_controller[i] == board_id)
			return 0;
	return 1;
}

static int ctlr_is_resettable(u32 board_id)
{
	return ctlr_is_hard_resettable(board_id) ||
		ctlr_is_soft_resettable(board_id);
}

621 622 623 624 625 626
static ssize_t host_show_resettable(struct device *dev,
				    struct device_attribute *attr,
				    char *buf)
{
	struct ctlr_info *h = to_hba(dev);

627
	return snprintf(buf, 20, "%d\n", ctlr_is_resettable(h->board_id));
628 629 630
}
static DEVICE_ATTR(resettable, S_IRUGO, host_show_resettable, NULL);

631 632 633 634 635 636 637 638 639 640 641 642
static ssize_t host_store_rescan(struct device *dev,
				 struct device_attribute *attr,
				 const char *buf, size_t count)
{
	struct ctlr_info *h = to_hba(dev);

	add_to_scan_list(h);
	wake_up_process(cciss_scan_thread);
	wait_for_completion_interruptible(&h->scan_wait);

	return count;
}
A
Alex Chiang 已提交
643
static DEVICE_ATTR(rescan, S_IWUSR, NULL, host_store_rescan);
644

645 646 647 648 649 650 651 652 653 654 655 656
static ssize_t host_show_transport_mode(struct device *dev,
				 struct device_attribute *attr,
				 char *buf)
{
	struct ctlr_info *h = to_hba(dev);

	return snprintf(buf, 20, "%s\n",
		h->transMethod & CFGTBL_Trans_Performant ?
			"performant" : "simple");
}
static DEVICE_ATTR(transport_mode, S_IRUGO, host_show_transport_mode, NULL);

657 658 659 660 661 662 663 664 665 666
static ssize_t dev_show_unique_id(struct device *dev,
				 struct device_attribute *attr,
				 char *buf)
{
	drive_info_struct *drv = to_drv(dev);
	struct ctlr_info *h = to_hba(drv->dev.parent);
	__u8 sn[16];
	unsigned long flags;
	int ret = 0;

667
	spin_lock_irqsave(&h->lock, flags);
668 669 670 671
	if (h->busy_configuring)
		ret = -EBUSY;
	else
		memcpy(sn, drv->serial_no, sizeof(sn));
672
	spin_unlock_irqrestore(&h->lock, flags);
673 674 675 676 677 678 679 680 681 682 683 684

	if (ret)
		return ret;
	else
		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]);
}
A
Alex Chiang 已提交
685
static DEVICE_ATTR(unique_id, S_IRUGO, dev_show_unique_id, NULL);
686 687 688 689 690 691 692 693 694 695 696

static ssize_t dev_show_vendor(struct device *dev,
			       struct device_attribute *attr,
			       char *buf)
{
	drive_info_struct *drv = to_drv(dev);
	struct ctlr_info *h = to_hba(drv->dev.parent);
	char vendor[VENDOR_LEN + 1];
	unsigned long flags;
	int ret = 0;

697
	spin_lock_irqsave(&h->lock, flags);
698 699 700 701
	if (h->busy_configuring)
		ret = -EBUSY;
	else
		memcpy(vendor, drv->vendor, VENDOR_LEN + 1);
702
	spin_unlock_irqrestore(&h->lock, flags);
703 704 705 706 707 708

	if (ret)
		return ret;
	else
		return snprintf(buf, sizeof(vendor) + 1, "%s\n", drv->vendor);
}
A
Alex Chiang 已提交
709
static DEVICE_ATTR(vendor, S_IRUGO, dev_show_vendor, NULL);
710 711 712 713 714 715 716 717 718 719 720

static ssize_t dev_show_model(struct device *dev,
			      struct device_attribute *attr,
			      char *buf)
{
	drive_info_struct *drv = to_drv(dev);
	struct ctlr_info *h = to_hba(drv->dev.parent);
	char model[MODEL_LEN + 1];
	unsigned long flags;
	int ret = 0;

721
	spin_lock_irqsave(&h->lock, flags);
722 723 724 725
	if (h->busy_configuring)
		ret = -EBUSY;
	else
		memcpy(model, drv->model, MODEL_LEN + 1);
726
	spin_unlock_irqrestore(&h->lock, flags);
727 728 729 730 731 732

	if (ret)
		return ret;
	else
		return snprintf(buf, sizeof(model) + 1, "%s\n", drv->model);
}
A
Alex Chiang 已提交
733
static DEVICE_ATTR(model, S_IRUGO, dev_show_model, NULL);
734 735 736 737 738 739 740 741 742 743 744

static ssize_t dev_show_rev(struct device *dev,
			    struct device_attribute *attr,
			    char *buf)
{
	drive_info_struct *drv = to_drv(dev);
	struct ctlr_info *h = to_hba(drv->dev.parent);
	char rev[REV_LEN + 1];
	unsigned long flags;
	int ret = 0;

745
	spin_lock_irqsave(&h->lock, flags);
746 747 748 749
	if (h->busy_configuring)
		ret = -EBUSY;
	else
		memcpy(rev, drv->rev, REV_LEN + 1);
750
	spin_unlock_irqrestore(&h->lock, flags);
751 752 753 754 755 756

	if (ret)
		return ret;
	else
		return snprintf(buf, sizeof(rev) + 1, "%s\n", drv->rev);
}
A
Alex Chiang 已提交
757
static DEVICE_ATTR(rev, S_IRUGO, dev_show_rev, NULL);
758

759 760 761
static ssize_t cciss_show_lunid(struct device *dev,
				struct device_attribute *attr, char *buf)
{
762 763
	drive_info_struct *drv = to_drv(dev);
	struct ctlr_info *h = to_hba(drv->dev.parent);
764 765 766
	unsigned long flags;
	unsigned char lunid[8];

767
	spin_lock_irqsave(&h->lock, flags);
768
	if (h->busy_configuring) {
769
		spin_unlock_irqrestore(&h->lock, flags);
770 771 772
		return -EBUSY;
	}
	if (!drv->heads) {
773
		spin_unlock_irqrestore(&h->lock, flags);
774 775 776
		return -ENOTTY;
	}
	memcpy(lunid, drv->LunID, sizeof(lunid));
777
	spin_unlock_irqrestore(&h->lock, flags);
778 779 780 781
	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]);
}
A
Alex Chiang 已提交
782
static DEVICE_ATTR(lunid, S_IRUGO, cciss_show_lunid, NULL);
783

784 785 786
static ssize_t cciss_show_raid_level(struct device *dev,
				     struct device_attribute *attr, char *buf)
{
787 788
	drive_info_struct *drv = to_drv(dev);
	struct ctlr_info *h = to_hba(drv->dev.parent);
789 790 791
	int raid;
	unsigned long flags;

792
	spin_lock_irqsave(&h->lock, flags);
793
	if (h->busy_configuring) {
794
		spin_unlock_irqrestore(&h->lock, flags);
795 796 797
		return -EBUSY;
	}
	raid = drv->raid_level;
798
	spin_unlock_irqrestore(&h->lock, flags);
799 800 801 802 803 804
	if (raid < 0 || raid > RAID_UNKNOWN)
		raid = RAID_UNKNOWN;

	return snprintf(buf, strlen(raid_label[raid]) + 7, "RAID %s\n",
			raid_label[raid]);
}
A
Alex Chiang 已提交
805
static DEVICE_ATTR(raid_level, S_IRUGO, cciss_show_raid_level, NULL);
806

807 808 809
static ssize_t cciss_show_usage_count(struct device *dev,
				      struct device_attribute *attr, char *buf)
{
810 811
	drive_info_struct *drv = to_drv(dev);
	struct ctlr_info *h = to_hba(drv->dev.parent);
812 813 814
	unsigned long flags;
	int count;

815
	spin_lock_irqsave(&h->lock, flags);
816
	if (h->busy_configuring) {
817
		spin_unlock_irqrestore(&h->lock, flags);
818 819 820
		return -EBUSY;
	}
	count = drv->usage_count;
821
	spin_unlock_irqrestore(&h->lock, flags);
822 823
	return snprintf(buf, 20, "%d\n", count);
}
A
Alex Chiang 已提交
824
static DEVICE_ATTR(usage_count, S_IRUGO, cciss_show_usage_count, NULL);
825

826 827
static struct attribute *cciss_host_attrs[] = {
	&dev_attr_rescan.attr,
828
	&dev_attr_resettable.attr,
829
	&dev_attr_transport_mode.attr,
830 831 832 833 834 835 836
	NULL
};

static struct attribute_group cciss_host_attr_group = {
	.attrs = cciss_host_attrs,
};

837
static const struct attribute_group *cciss_host_attr_groups[] = {
838 839 840 841 842 843 844
	&cciss_host_attr_group,
	NULL
};

static struct device_type cciss_host_type = {
	.name		= "cciss_host",
	.groups		= cciss_host_attr_groups,
845
	.release	= cciss_hba_release,
846 847
};

848 849 850 851 852
static struct attribute *cciss_dev_attrs[] = {
	&dev_attr_unique_id.attr,
	&dev_attr_model.attr,
	&dev_attr_vendor.attr,
	&dev_attr_rev.attr,
853
	&dev_attr_lunid.attr,
854
	&dev_attr_raid_level.attr,
855
	&dev_attr_usage_count.attr,
856 857 858 859 860 861 862
	NULL
};

static struct attribute_group cciss_dev_attr_group = {
	.attrs = cciss_dev_attrs,
};

863
static const struct attribute_group *cciss_dev_attr_groups[] = {
864 865 866 867 868 869 870
	&cciss_dev_attr_group,
	NULL
};

static struct device_type cciss_dev_type = {
	.name		= "cciss_device",
	.groups		= cciss_dev_attr_groups,
871
	.release	= cciss_device_release,
872 873 874 875 876 877
};

static struct bus_type cciss_bus_type = {
	.name		= "cciss",
};

878 879 880 881 882 883 884 885 886 887 888
/*
 * cciss_hba_release is called when the reference count
 * of h->dev goes to zero.
 */
static void cciss_hba_release(struct device *dev)
{
	/*
	 * nothing to do, but need this to avoid a warning
	 * about not having a release handler from lib/kref.c.
	 */
}
889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911

/*
 * Initialize sysfs entry for each controller.  This sets up and registers
 * the 'cciss#' directory for each individual controller under
 * /sys/bus/pci/devices/<dev>/.
 */
static int cciss_create_hba_sysfs_entry(struct ctlr_info *h)
{
	device_initialize(&h->dev);
	h->dev.type = &cciss_host_type;
	h->dev.bus = &cciss_bus_type;
	dev_set_name(&h->dev, "%s", h->devname);
	h->dev.parent = &h->pdev->dev;

	return device_add(&h->dev);
}

/*
 * Remove sysfs entries for an hba.
 */
static void cciss_destroy_hba_sysfs_entry(struct ctlr_info *h)
{
	device_del(&h->dev);
912 913 914 915
	put_device(&h->dev); /* final put. */
}

/* cciss_device_release is called when the reference count
916
 * of h->drv[x]dev goes to zero.
917 918 919
 */
static void cciss_device_release(struct device *dev)
{
920 921
	drive_info_struct *drv = to_drv(dev);
	kfree(drv);
922 923 924 925 926 927 928 929
}

/*
 * Initialize sysfs for each logical drive.  This sets up and registers
 * the 'c#d#' directory for each individual logical drive under
 * /sys/bus/pci/devices/<dev/ccis#/. We also create a link from
 * /sys/block/cciss!c#d# to this entry.
 */
930
static long cciss_create_ld_sysfs_entry(struct ctlr_info *h,
931 932
				       int drv_index)
{
933 934
	struct device *dev;

935
	if (h->drv[drv_index]->device_initialized)
936 937
		return 0;

938
	dev = &h->drv[drv_index]->dev;
939 940 941 942 943
	device_initialize(dev);
	dev->type = &cciss_dev_type;
	dev->bus = &cciss_bus_type;
	dev_set_name(dev, "c%dd%d", h->ctlr, drv_index);
	dev->parent = &h->dev;
944
	h->drv[drv_index]->device_initialized = 1;
945
	return device_add(dev);
946 947 948 949 950
}

/*
 * Remove sysfs entries for a logical drive.
 */
951 952
static void cciss_destroy_ld_sysfs_entry(struct ctlr_info *h, int drv_index,
	int ctlr_exiting)
953
{
954
	struct device *dev = &h->drv[drv_index]->dev;
955 956 957 958 959

	/* special case for c*d0, we only destroy it on controller exit */
	if (drv_index == 0 && !ctlr_exiting)
		return;

960 961
	device_del(dev);
	put_device(dev); /* the "final" put. */
962
	h->drv[drv_index] = NULL;
963 964
}

965 966
/*
 * For operations that cannot sleep, a command block is allocated at init,
L
Linus Torvalds 已提交
967
 * and managed by cmd_alloc() and cmd_free() using a simple bitmap to track
968
 * which ones are free or in use.
969
 */
970
static CommandList_struct *cmd_alloc(ctlr_info_t *h)
L
Linus Torvalds 已提交
971 972
{
	CommandList_struct *c;
973
	int i;
L
Linus Torvalds 已提交
974 975 976
	u64bit temp64;
	dma_addr_t cmd_dma_handle, err_dma_handle;

977 978 979
	do {
		i = find_first_zero_bit(h->cmd_pool_bits, h->nr_cmds);
		if (i == h->nr_cmds)
980
			return NULL;
981 982 983 984 985 986 987 988 989 990
	} while (test_and_set_bit(i & (BITS_PER_LONG - 1),
		  h->cmd_pool_bits + (i / BITS_PER_LONG)) != 0);
	c = h->cmd_pool + i;
	memset(c, 0, sizeof(CommandList_struct));
	cmd_dma_handle = h->cmd_pool_dhandle + i * sizeof(CommandList_struct);
	c->err_info = h->errinfo_pool + i;
	memset(c->err_info, 0, sizeof(ErrorInfo_struct));
	err_dma_handle = h->errinfo_pool_dhandle
	    + i * sizeof(ErrorInfo_struct);
	h->nr_allocs++;
L
Linus Torvalds 已提交
991

992
	c->cmdindex = i;
993

994
	INIT_LIST_HEAD(&c->list);
995 996 997 998 999
	c->busaddr = (__u32) cmd_dma_handle;
	temp64.val = (__u64) err_dma_handle;
	c->ErrDesc.Addr.lower = temp64.val32.lower;
	c->ErrDesc.Addr.upper = temp64.val32.upper;
	c->ErrDesc.Len = sizeof(ErrorInfo_struct);
1000

1001 1002 1003
	c->ctlr = h->ctlr;
	return c;
}
1004

1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029
/* allocate a command using pci_alloc_consistent, used for ioctls,
 * etc., not for the main i/o path.
 */
static CommandList_struct *cmd_special_alloc(ctlr_info_t *h)
{
	CommandList_struct *c;
	u64bit temp64;
	dma_addr_t cmd_dma_handle, err_dma_handle;

	c = (CommandList_struct *) pci_alloc_consistent(h->pdev,
		sizeof(CommandList_struct), &cmd_dma_handle);
	if (c == NULL)
		return NULL;
	memset(c, 0, sizeof(CommandList_struct));

	c->cmdindex = -1;

	c->err_info = (ErrorInfo_struct *)
	    pci_alloc_consistent(h->pdev, sizeof(ErrorInfo_struct),
		    &err_dma_handle);

	if (c->err_info == NULL) {
		pci_free_consistent(h->pdev,
			sizeof(CommandList_struct), c, cmd_dma_handle);
		return NULL;
1030
	}
1031
	memset(c->err_info, 0, sizeof(ErrorInfo_struct));
L
Linus Torvalds 已提交
1032

1033
	INIT_LIST_HEAD(&c->list);
L
Linus Torvalds 已提交
1034
	c->busaddr = (__u32) cmd_dma_handle;
1035
	temp64.val = (__u64) err_dma_handle;
L
Linus Torvalds 已提交
1036 1037 1038 1039
	c->ErrDesc.Addr.lower = temp64.val32.lower;
	c->ErrDesc.Addr.upper = temp64.val32.upper;
	c->ErrDesc.Len = sizeof(ErrorInfo_struct);

1040 1041
	c->ctlr = h->ctlr;
	return c;
L
Linus Torvalds 已提交
1042 1043
}

1044
static void cmd_free(ctlr_info_t *h, CommandList_struct *c)
L
Linus Torvalds 已提交
1045 1046
{
	int i;
1047 1048 1049 1050 1051 1052 1053 1054 1055

	i = c - h->cmd_pool;
	clear_bit(i & (BITS_PER_LONG - 1),
		  h->cmd_pool_bits + (i / BITS_PER_LONG));
	h->nr_frees++;
}

static void cmd_special_free(ctlr_info_t *h, CommandList_struct *c)
{
L
Linus Torvalds 已提交
1056 1057
	u64bit temp64;

1058 1059 1060 1061
	temp64.val32.lower = c->ErrDesc.Addr.lower;
	temp64.val32.upper = c->ErrDesc.Addr.upper;
	pci_free_consistent(h->pdev, sizeof(ErrorInfo_struct),
			    c->err_info, (dma_addr_t) temp64.val);
1062 1063
	pci_free_consistent(h->pdev, sizeof(CommandList_struct), c,
		(dma_addr_t) cciss_tag_discard_error_bits(h, (u32) c->busaddr));
L
Linus Torvalds 已提交
1064 1065 1066 1067
}

static inline ctlr_info_t *get_host(struct gendisk *disk)
{
1068
	return disk->queue->queuedata;
L
Linus Torvalds 已提交
1069 1070 1071 1072 1073 1074 1075 1076 1077 1078
}

static inline drive_info_struct *get_drv(struct gendisk *disk)
{
	return disk->private_data;
}

/*
 * Open.  Make sure the device is really there.
 */
A
Al Viro 已提交
1079
static int cciss_open(struct block_device *bdev, fmode_t mode)
L
Linus Torvalds 已提交
1080
{
1081
	ctlr_info_t *h = get_host(bdev->bd_disk);
A
Al Viro 已提交
1082
	drive_info_struct *drv = get_drv(bdev->bd_disk);
L
Linus Torvalds 已提交
1083

1084
	dev_dbg(&h->pdev->dev, "cciss_open %s\n", bdev->bd_disk->disk_name);
1085
	if (drv->busy_configuring)
1086
		return -EBUSY;
L
Linus Torvalds 已提交
1087 1088 1089 1090 1091 1092 1093 1094
	/*
	 * Root is allowed to open raw volume zero even if it's not configured
	 * so array config can still work. Root is also allowed to open any
	 * volume that has a LUN ID, so it can issue IOCTL to reread the
	 * disk information.  I don't think I really like this
	 * but I'm already using way to many device nodes to claim another one
	 * for "raw controller".
	 */
1095
	if (drv->heads == 0) {
A
Al Viro 已提交
1096
		if (MINOR(bdev->bd_dev) != 0) {	/* not node 0? */
L
Linus Torvalds 已提交
1097
			/* if not node 0 make sure it is a partition = 0 */
A
Al Viro 已提交
1098
			if (MINOR(bdev->bd_dev) & 0x0f) {
1099
				return -ENXIO;
L
Linus Torvalds 已提交
1100
				/* if it is, make sure we have a LUN ID */
1101 1102
			} else if (memcmp(drv->LunID, CTLR_LUNID,
				sizeof(drv->LunID))) {
L
Linus Torvalds 已提交
1103 1104 1105 1106 1107 1108 1109
				return -ENXIO;
			}
		}
		if (!capable(CAP_SYS_ADMIN))
			return -EPERM;
	}
	drv->usage_count++;
1110
	h->usage_count++;
L
Linus Torvalds 已提交
1111 1112
	return 0;
}
1113

1114 1115 1116 1117
static int cciss_unlocked_open(struct block_device *bdev, fmode_t mode)
{
	int ret;

1118
	mutex_lock(&cciss_mutex);
1119
	ret = cciss_open(bdev, mode);
1120
	mutex_unlock(&cciss_mutex);
1121 1122 1123 1124

	return ret;
}

L
Linus Torvalds 已提交
1125 1126 1127
/*
 * Close.  Sync first.
 */
A
Al Viro 已提交
1128
static int cciss_release(struct gendisk *disk, fmode_t mode)
L
Linus Torvalds 已提交
1129
{
1130
	ctlr_info_t *h;
1131
	drive_info_struct *drv;
L
Linus Torvalds 已提交
1132

1133
	mutex_lock(&cciss_mutex);
1134
	h = get_host(disk);
1135
	drv = get_drv(disk);
1136
	dev_dbg(&h->pdev->dev, "cciss_release %s\n", disk->disk_name);
L
Linus Torvalds 已提交
1137
	drv->usage_count--;
1138
	h->usage_count--;
1139
	mutex_unlock(&cciss_mutex);
L
Linus Torvalds 已提交
1140 1141 1142
	return 0;
}

A
Al Viro 已提交
1143 1144
static int do_ioctl(struct block_device *bdev, fmode_t mode,
		    unsigned cmd, unsigned long arg)
L
Linus Torvalds 已提交
1145 1146
{
	int ret;
1147
	mutex_lock(&cciss_mutex);
A
Al Viro 已提交
1148
	ret = cciss_ioctl(bdev, mode, cmd, arg);
1149
	mutex_unlock(&cciss_mutex);
L
Linus Torvalds 已提交
1150 1151 1152
	return ret;
}

1153 1154
#ifdef CONFIG_COMPAT

A
Al Viro 已提交
1155 1156 1157 1158
static int cciss_ioctl32_passthru(struct block_device *bdev, fmode_t mode,
				  unsigned cmd, unsigned long arg);
static int cciss_ioctl32_big_passthru(struct block_device *bdev, fmode_t mode,
				      unsigned cmd, unsigned long arg);
L
Linus Torvalds 已提交
1159

A
Al Viro 已提交
1160 1161
static int cciss_compat_ioctl(struct block_device *bdev, fmode_t mode,
			      unsigned cmd, unsigned long arg)
L
Linus Torvalds 已提交
1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178
{
	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:
A
Al Viro 已提交
1179
		return do_ioctl(bdev, mode, cmd, arg);
L
Linus Torvalds 已提交
1180 1181

	case CCISS_PASSTHRU32:
A
Al Viro 已提交
1182
		return cciss_ioctl32_passthru(bdev, mode, cmd, arg);
L
Linus Torvalds 已提交
1183
	case CCISS_BIG_PASSTHRU32:
A
Al Viro 已提交
1184
		return cciss_ioctl32_big_passthru(bdev, mode, cmd, arg);
L
Linus Torvalds 已提交
1185 1186 1187 1188 1189 1190

	default:
		return -ENOIOCTLCMD;
	}
}

A
Al Viro 已提交
1191 1192
static int cciss_ioctl32_passthru(struct block_device *bdev, fmode_t mode,
				  unsigned cmd, unsigned long arg)
L
Linus Torvalds 已提交
1193 1194
{
	IOCTL32_Command_struct __user *arg32 =
1195
	    (IOCTL32_Command_struct __user *) arg;
L
Linus Torvalds 已提交
1196 1197 1198 1199 1200 1201
	IOCTL_Command_struct arg64;
	IOCTL_Command_struct __user *p = compat_alloc_user_space(sizeof(arg64));
	int err;
	u32 cp;

	err = 0;
1202 1203 1204 1205 1206 1207 1208 1209 1210
	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));
L
Linus Torvalds 已提交
1211 1212 1213 1214 1215 1216 1217 1218
	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;

A
Al Viro 已提交
1219
	err = do_ioctl(bdev, mode, CCISS_PASSTHRU, (unsigned long)p);
L
Linus Torvalds 已提交
1220 1221
	if (err)
		return err;
1222 1223 1224
	err |=
	    copy_in_user(&arg32->error_info, &p->error_info,
			 sizeof(arg32->error_info));
L
Linus Torvalds 已提交
1225 1226 1227 1228 1229
	if (err)
		return -EFAULT;
	return err;
}

A
Al Viro 已提交
1230 1231
static int cciss_ioctl32_big_passthru(struct block_device *bdev, fmode_t mode,
				      unsigned cmd, unsigned long arg)
L
Linus Torvalds 已提交
1232 1233
{
	BIG_IOCTL32_Command_struct __user *arg32 =
1234
	    (BIG_IOCTL32_Command_struct __user *) arg;
L
Linus Torvalds 已提交
1235
	BIG_IOCTL_Command_struct arg64;
1236 1237
	BIG_IOCTL_Command_struct __user *p =
	    compat_alloc_user_space(sizeof(arg64));
L
Linus Torvalds 已提交
1238 1239 1240
	int err;
	u32 cp;

1241
	memset(&arg64, 0, sizeof(arg64));
L
Linus Torvalds 已提交
1242
	err = 0;
1243 1244 1245 1246 1247 1248 1249 1250 1251
	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));
L
Linus Torvalds 已提交
1252 1253 1254 1255 1256 1257 1258
	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)
1259
		return -EFAULT;
L
Linus Torvalds 已提交
1260

A
Al Viro 已提交
1261
	err = do_ioctl(bdev, mode, CCISS_BIG_PASSTHRU, (unsigned long)p);
L
Linus Torvalds 已提交
1262 1263
	if (err)
		return err;
1264 1265 1266
	err |=
	    copy_in_user(&arg32->error_info, &p->error_info,
			 sizeof(arg32->error_info));
L
Linus Torvalds 已提交
1267 1268 1269 1270 1271
	if (err)
		return -EFAULT;
	return err;
}
#endif
1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285

static int cciss_getgeo(struct block_device *bdev, struct hd_geometry *geo)
{
	drive_info_struct *drv = get_drv(bdev->bd_disk);

	if (!drv->cylinders)
		return -ENXIO;

	geo->heads = drv->heads;
	geo->sectors = drv->sectors;
	geo->cylinders = drv->cylinders;
	return 0;
}

1286
static void check_ioctl_unit_attention(ctlr_info_t *h, CommandList_struct *c)
1287 1288 1289
{
	if (c->err_info->CommandStatus == CMD_TARGET_STATUS &&
			c->err_info->ScsiStatus != SAM_STAT_CHECK_CONDITION)
1290
		(void)check_for_unit_attention(h, c);
1291
}
1292 1293

static int cciss_getpciinfo(ctlr_info_t *h, void __user *argp)
L
Linus Torvalds 已提交
1294
{
1295
	cciss_pci_info_struct pciinfo;
L
Linus Torvalds 已提交
1296

1297 1298 1299 1300 1301 1302 1303 1304 1305 1306
	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(cciss_pci_info_struct)))
		return -EFAULT;
	return 0;
}
L
Linus Torvalds 已提交
1307

1308 1309 1310
static int cciss_getintinfo(ctlr_info_t *h, void __user *argp)
{
	cciss_coalint_struct intinfo;
L
Linus Torvalds 已提交
1311

1312 1313 1314 1315 1316 1317 1318 1319 1320
	if (!argp)
		return -EINVAL;
	intinfo.delay = readl(&h->cfgtable->HostWrite.CoalIntDelay);
	intinfo.count = readl(&h->cfgtable->HostWrite.CoalIntCount);
	if (copy_to_user
	    (argp, &intinfo, sizeof(cciss_coalint_struct)))
		return -EFAULT;
	return 0;
}
L
Linus Torvalds 已提交
1321

1322 1323 1324 1325 1326
static int cciss_setintinfo(ctlr_info_t *h, void __user *argp)
{
	cciss_coalint_struct intinfo;
	unsigned long flags;
	int i;
L
Linus Torvalds 已提交
1327

1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351
	if (!argp)
		return -EINVAL;
	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
	if (copy_from_user(&intinfo, argp, sizeof(intinfo)))
		return -EFAULT;
	if ((intinfo.delay == 0) && (intinfo.count == 0))
		return -EINVAL;
	spin_lock_irqsave(&h->lock, flags);
	/* Update the field, and then ring the doorbell */
	writel(intinfo.delay, &(h->cfgtable->HostWrite.CoalIntDelay));
	writel(intinfo.count, &(h->cfgtable->HostWrite.CoalIntCount));
	writel(CFGTBL_ChangeReq, h->vaddr + SA5_DOORBELL);

	for (i = 0; i < MAX_IOCTL_CONFIG_WAIT; i++) {
		if (!(readl(h->vaddr + SA5_DOORBELL) & CFGTBL_ChangeReq))
			break;
		udelay(1000); /* delay and try again */
	}
	spin_unlock_irqrestore(&h->lock, flags);
	if (i >= MAX_IOCTL_CONFIG_WAIT)
		return -EAGAIN;
	return 0;
}
L
Linus Torvalds 已提交
1352

1353 1354 1355 1356
static int cciss_getnodename(ctlr_info_t *h, void __user *argp)
{
	NodeName_type NodeName;
	int i;
L
Linus Torvalds 已提交
1357

1358 1359 1360 1361 1362 1363 1364 1365
	if (!argp)
		return -EINVAL;
	for (i = 0; i < 16; i++)
		NodeName[i] = readb(&h->cfgtable->ServerName[i]);
	if (copy_to_user(argp, NodeName, sizeof(NodeName_type)))
		return -EFAULT;
	return 0;
}
1366

1367 1368 1369 1370 1371
static int cciss_setnodename(ctlr_info_t *h, void __user *argp)
{
	NodeName_type NodeName;
	unsigned long flags;
	int i;
1372

1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393
	if (!argp)
		return -EINVAL;
	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
	if (copy_from_user(NodeName, argp, sizeof(NodeName_type)))
		return -EFAULT;
	spin_lock_irqsave(&h->lock, flags);
	/* Update the field, and then ring the doorbell */
	for (i = 0; i < 16; i++)
		writeb(NodeName[i], &h->cfgtable->ServerName[i]);
	writel(CFGTBL_ChangeReq, h->vaddr + SA5_DOORBELL);
	for (i = 0; i < MAX_IOCTL_CONFIG_WAIT; i++) {
		if (!(readl(h->vaddr + SA5_DOORBELL) & CFGTBL_ChangeReq))
			break;
		udelay(1000); /* delay and try again */
	}
	spin_unlock_irqrestore(&h->lock, flags);
	if (i >= MAX_IOCTL_CONFIG_WAIT)
		return -EAGAIN;
	return 0;
}
1394

1395 1396 1397
static int cciss_getheartbeat(ctlr_info_t *h, void __user *argp)
{
	Heartbeat_type heartbeat;
1398

1399 1400 1401 1402 1403 1404 1405
	if (!argp)
		return -EINVAL;
	heartbeat = readl(&h->cfgtable->HeartBeat);
	if (copy_to_user(argp, &heartbeat, sizeof(Heartbeat_type)))
		return -EFAULT;
	return 0;
}
1406

1407 1408 1409
static int cciss_getbustypes(ctlr_info_t *h, void __user *argp)
{
	BusTypes_type BusTypes;
1410

1411 1412 1413 1414 1415 1416 1417 1418
	if (!argp)
		return -EINVAL;
	BusTypes = readl(&h->cfgtable->BusTypes);
	if (copy_to_user(argp, &BusTypes, sizeof(BusTypes_type)))
		return -EFAULT;
	return 0;
}

1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432
static int cciss_getfirmver(ctlr_info_t *h, void __user *argp)
{
	FirmwareVer_type firmware;

	if (!argp)
		return -EINVAL;
	memcpy(firmware, h->firm_ver, 4);

	if (copy_to_user
	    (argp, firmware, sizeof(FirmwareVer_type)))
		return -EFAULT;
	return 0;
}

1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443
static int cciss_getdrivver(ctlr_info_t *h, void __user *argp)
{
	DriverVer_type DriverVer = DRIVER_VERSION;

	if (!argp)
		return -EINVAL;
	if (copy_to_user(argp, &DriverVer, sizeof(DriverVer_type)))
		return -EFAULT;
	return 0;
}

1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459
static int cciss_getluninfo(ctlr_info_t *h,
	struct gendisk *disk, void __user *argp)
{
	LogvolInfo_struct luninfo;
	drive_info_struct *drv = get_drv(disk);

	if (!argp)
		return -EINVAL;
	memcpy(&luninfo.LunID, drv->LunID, sizeof(luninfo.LunID));
	luninfo.num_opens = drv->usage_count;
	luninfo.num_parts = 0;
	if (copy_to_user(argp, &luninfo, sizeof(LogvolInfo_struct)))
		return -EFAULT;
	return 0;
}

1460 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 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549
static int cciss_passthru(ctlr_info_t *h, void __user *argp)
{
	IOCTL_Command_struct iocommand;
	CommandList_struct *c;
	char *buff = NULL;
	u64bit temp64;
	DECLARE_COMPLETION_ONSTACK(wait);

	if (!argp)
		return -EINVAL;

	if (!capable(CAP_SYS_RAWIO))
		return -EPERM;

	if (copy_from_user
	    (&iocommand, argp, sizeof(IOCTL_Command_struct)))
		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)
			return -EFAULT;
	}
	if (iocommand.Request.Type.Direction == XFER_WRITE) {
		/* Copy the data into the buffer we created */
		if (copy_from_user(buff, iocommand.buf, iocommand.buf_size)) {
			kfree(buff);
			return -EFAULT;
		}
	} else {
		memset(buff, 0, iocommand.buf_size);
	}
	c = cmd_special_alloc(h);
	if (!c) {
		kfree(buff);
		return -ENOMEM;
	}
	/* Fill in the command type */
	c->cmd_type = CMD_IOCTL_PEND;
	/* Fill in Command Header */
	c->Header.ReplyQueue = 0;   /* unused in simple mode */
	if (iocommand.buf_size > 0) { /* buffer to fill */
		c->Header.SGList = 1;
		c->Header.SGTotal = 1;
	} else { /* no buffers to fill */
		c->Header.SGList = 0;
		c->Header.SGTotal = 0;
	}
	c->Header.LUN = iocommand.LUN_info;
	/* use the kernel address the cmd block for tag */
	c->Header.Tag.lower = c->busaddr;

	/* Fill in Request block */
	c->Request = iocommand.Request;

	/* Fill in the scatter gather information */
	if (iocommand.buf_size > 0) {
		temp64.val = pci_map_single(h->pdev, buff,
			iocommand.buf_size, PCI_DMA_BIDIRECTIONAL);
		c->SG[0].Addr.lower = temp64.val32.lower;
		c->SG[0].Addr.upper = temp64.val32.upper;
		c->SG[0].Len = iocommand.buf_size;
		c->SG[0].Ext = 0;  /* we are not chaining */
	}
	c->waiting = &wait;

	enqueue_cmd_and_start_io(h, c);
	wait_for_completion(&wait);

	/* unlock the buffers from DMA */
	temp64.val32.lower = c->SG[0].Addr.lower;
	temp64.val32.upper = c->SG[0].Addr.upper;
	pci_unmap_single(h->pdev, (dma_addr_t) temp64.val, iocommand.buf_size,
			 PCI_DMA_BIDIRECTIONAL);
	check_ioctl_unit_attention(h, c);

	/* Copy the error information out */
	iocommand.error_info = *(c->err_info);
	if (copy_to_user(argp, &iocommand, sizeof(IOCTL_Command_struct))) {
		kfree(buff);
		cmd_special_free(h, c);
		return -EFAULT;
	}

	if (iocommand.Request.Type.Direction == XFER_READ) {
		/* Copy the data out of the buffer we created */
		if (copy_to_user(iocommand.buf, buff, iocommand.buf_size)) {
1550
			kfree(buff);
1551
			cmd_special_free(h, c);
1552
			return -EFAULT;
L
Linus Torvalds 已提交
1553
		}
1554 1555 1556 1557 1558 1559
	}
	kfree(buff);
	cmd_special_free(h, c);
	return 0;
}

1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578
static int cciss_bigpassthru(ctlr_info_t *h, void __user *argp)
{
	BIG_IOCTL_Command_struct *ioc;
	CommandList_struct *c;
	unsigned char **buff = NULL;
	int *buff_size = NULL;
	u64bit temp64;
	BYTE sg_used = 0;
	int status = 0;
	int i;
	DECLARE_COMPLETION_ONSTACK(wait);
	__u32 left;
	__u32 sz;
	BYTE __user *data_ptr;

	if (!argp)
		return -EINVAL;
	if (!capable(CAP_SYS_RAWIO))
		return -EPERM;
1579
	ioc = kmalloc(sizeof(*ioc), GFP_KERNEL);
1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623
	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;
	}
	if (ioc->buf_size > ioc->malloc_size * MAXSGENTRIES) {
		status = -EINVAL;
		goto cleanup1;
	}
	buff = kzalloc(MAXSGENTRIES * sizeof(char *), GFP_KERNEL);
	if (!buff) {
		status = -ENOMEM;
		goto cleanup1;
	}
	buff_size = kmalloc(MAXSGENTRIES * sizeof(int), GFP_KERNEL);
	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;
		}
		if (ioc->Request.Type.Direction == XFER_WRITE) {
			if (copy_from_user(buff[sg_used], data_ptr, sz)) {
1624 1625 1626
				status = -EFAULT;
				goto cleanup1;
			}
1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640
		} else {
			memset(buff[sg_used], 0, sz);
		}
		left -= sz;
		data_ptr += sz;
		sg_used++;
	}
	c = cmd_special_alloc(h);
	if (!c) {
		status = -ENOMEM;
		goto cleanup1;
	}
	c->cmd_type = CMD_IOCTL_PEND;
	c->Header.ReplyQueue = 0;
1641 1642
	c->Header.SGList = sg_used;
	c->Header.SGTotal = sg_used;
1643 1644 1645 1646
	c->Header.LUN = ioc->LUN_info;
	c->Header.Tag.lower = c->busaddr;

	c->Request = ioc->Request;
1647 1648
	for (i = 0; i < sg_used; i++) {
		temp64.val = pci_map_single(h->pdev, buff[i], buff_size[i],
1649
				    PCI_DMA_BIDIRECTIONAL);
1650 1651 1652 1653
		c->SG[i].Addr.lower = temp64.val32.lower;
		c->SG[i].Addr.upper = temp64.val32.upper;
		c->SG[i].Len = buff_size[i];
		c->SG[i].Ext = 0;	/* we are not chaining */
1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678
	}
	c->waiting = &wait;
	enqueue_cmd_and_start_io(h, c);
	wait_for_completion(&wait);
	/* unlock the buffers from DMA */
	for (i = 0; i < sg_used; i++) {
		temp64.val32.lower = c->SG[i].Addr.lower;
		temp64.val32.upper = c->SG[i].Addr.upper;
		pci_unmap_single(h->pdev,
			(dma_addr_t) temp64.val, buff_size[i],
			PCI_DMA_BIDIRECTIONAL);
	}
	check_ioctl_unit_attention(h, c);
	/* Copy the error information out */
	ioc->error_info = *(c->err_info);
	if (copy_to_user(argp, ioc, sizeof(*ioc))) {
		cmd_special_free(h, c);
		status = -EFAULT;
		goto cleanup1;
	}
	if (ioc->Request.Type.Direction == XFER_READ) {
		/* 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])) {
1679
				cmd_special_free(h, c);
1680 1681 1682
				status = -EFAULT;
				goto cleanup1;
			}
1683
			ptr += buff_size[i];
L
Linus Torvalds 已提交
1684
		}
1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698
	}
	cmd_special_free(h, c);
	status = 0;
cleanup1:
	if (buff) {
		for (i = 0; i < sg_used; i++)
			kfree(buff[i]);
		kfree(buff);
	}
	kfree(buff_size);
	kfree(ioc);
	return status;
}

A
Al Viro 已提交
1699
static int cciss_ioctl(struct block_device *bdev, fmode_t mode,
1700
	unsigned int cmd, unsigned long arg)
L
Linus Torvalds 已提交
1701 1702
{
	struct gendisk *disk = bdev->bd_disk;
1703
	ctlr_info_t *h = get_host(disk);
L
Linus Torvalds 已提交
1704 1705
	void __user *argp = (void __user *)arg;

1706 1707
	dev_dbg(&h->pdev->dev, "cciss_ioctl: Called with cmd=%x %lx\n",
		cmd, arg);
1708
	switch (cmd) {
L
Linus Torvalds 已提交
1709
	case CCISS_GETPCIINFO:
1710
		return cciss_getpciinfo(h, argp);
L
Linus Torvalds 已提交
1711
	case CCISS_GETINTINFO:
1712
		return cciss_getintinfo(h, argp);
L
Linus Torvalds 已提交
1713
	case CCISS_SETINTINFO:
1714
		return cciss_setintinfo(h, argp);
L
Linus Torvalds 已提交
1715
	case CCISS_GETNODENAME:
1716
		return cciss_getnodename(h, argp);
L
Linus Torvalds 已提交
1717
	case CCISS_SETNODENAME:
1718
		return cciss_setnodename(h, argp);
L
Linus Torvalds 已提交
1719
	case CCISS_GETHEARTBEAT:
1720
		return cciss_getheartbeat(h, argp);
L
Linus Torvalds 已提交
1721
	case CCISS_GETBUSTYPES:
1722
		return cciss_getbustypes(h, argp);
L
Linus Torvalds 已提交
1723
	case CCISS_GETFIRMVER:
1724
		return cciss_getfirmver(h, argp);
1725
	case CCISS_GETDRIVVER:
1726
		return cciss_getdrivver(h, argp);
M
Mike Miller 已提交
1727 1728
	case CCISS_DEREGDISK:
	case CCISS_REGNEWD:
L
Linus Torvalds 已提交
1729
	case CCISS_REVALIDVOLS:
1730
		return rebuild_lun_table(h, 0, 1);
1731 1732
	case CCISS_GETLUNINFO:
		return cciss_getluninfo(h, disk, argp);
L
Linus Torvalds 已提交
1733
	case CCISS_PASSTHRU:
1734
		return cciss_passthru(h, argp);
1735 1736
	case CCISS_BIG_PASSTHRU:
		return cciss_bigpassthru(h, argp);
1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748

	/* scsi_cmd_ioctl handles these, below, though some are not */
	/* very meaningful for cciss.  SG_IO is the main one people want. */

	case SG_GET_VERSION_NUM:
	case SG_SET_TIMEOUT:
	case SG_GET_TIMEOUT:
	case SG_GET_RESERVED_SIZE:
	case SG_SET_RESERVED_SIZE:
	case SG_EMULATED_HOST:
	case SG_IO:
	case SCSI_IOCTL_SEND_COMMAND:
A
Al Viro 已提交
1749
		return scsi_cmd_ioctl(disk->queue, disk, mode, cmd, argp);
1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760

	/* scsi_cmd_ioctl would normally handle these, below, but */
	/* they aren't a good fit for cciss, as CD-ROMs are */
	/* not supported, and we don't have any bus/target/lun */
	/* which we present to the kernel. */

	case CDROM_SEND_PACKET:
	case CDROMCLOSETRAY:
	case CDROMEJECT:
	case SCSI_IOCTL_GET_IDLUN:
	case SCSI_IOCTL_GET_BUS_NUMBER:
L
Linus Torvalds 已提交
1761 1762 1763 1764 1765
	default:
		return -ENOTTY;
	}
}

1766 1767 1768 1769 1770 1771 1772 1773 1774 1775
static void cciss_check_queues(ctlr_info_t *h)
{
	int start_queue = h->next_to_run;
	int i;

	/* check to see if we have maxed out the number of commands that can
	 * be placed on the queue.  If so then exit.  We do this check here
	 * in case the interrupt we serviced was from an ioctl and did not
	 * free any new commands.
	 */
1776
	if ((find_first_zero_bit(h->cmd_pool_bits, h->nr_cmds)) == h->nr_cmds)
1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787
		return;

	/* We have room on the queue for more commands.  Now we need to queue
	 * them up.  We will also keep track of the next queue to run so
	 * that every queue gets a chance to be started first.
	 */
	for (i = 0; i < h->highest_lun + 1; i++) {
		int curr_queue = (start_queue + i) % (h->highest_lun + 1);
		/* make sure the disk has been added and the drive is real
		 * because this can be called from the middle of init_one.
		 */
1788 1789 1790 1791
		if (!h->drv[curr_queue])
			continue;
		if (!(h->drv[curr_queue]->queue) ||
			!(h->drv[curr_queue]->heads))
1792 1793 1794 1795 1796 1797
			continue;
		blk_start_queue(h->gendisk[curr_queue]->queue);

		/* check to see if we have maxed out the number of commands
		 * that can be placed on the queue.
		 */
1798
		if ((find_first_zero_bit(h->cmd_pool_bits, h->nr_cmds)) == h->nr_cmds) {
1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810
			if (curr_queue == start_queue) {
				h->next_to_run =
				    (start_queue + 1) % (h->highest_lun + 1);
				break;
			} else {
				h->next_to_run = curr_queue;
				break;
			}
		}
	}
}

1811 1812
static void cciss_softirq_done(struct request *rq)
{
1813 1814 1815
	CommandList_struct *c = rq->completion_data;
	ctlr_info_t *h = hba[c->ctlr];
	SGDescriptor_struct *curr_sg = c->SG;
1816
	u64bit temp64;
M
Mike Miller 已提交
1817
	unsigned long flags;
1818
	int i, ddir;
1819
	int sg_index = 0;
1820

1821
	if (c->Request.Type.Direction == XFER_READ)
1822 1823 1824 1825 1826 1827
		ddir = PCI_DMA_FROMDEVICE;
	else
		ddir = PCI_DMA_TODEVICE;

	/* command did not need to be retried */
	/* unmap the DMA mapping for all the scatter gather elements */
1828
	for (i = 0; i < c->Header.SGList; i++) {
1829
		if (curr_sg[sg_index].Ext == CCISS_SG_CHAIN) {
1830
			cciss_unmap_sg_chain_block(h, c);
1831
			/* Point to the next block */
1832
			curr_sg = h->cmd_sg_list[c->cmdindex];
1833 1834 1835 1836 1837 1838 1839
			sg_index = 0;
		}
		temp64.val32.lower = curr_sg[sg_index].Addr.lower;
		temp64.val32.upper = curr_sg[sg_index].Addr.upper;
		pci_unmap_page(h->pdev, temp64.val, curr_sg[sg_index].Len,
				ddir);
		++sg_index;
1840 1841
	}

1842
	dev_dbg(&h->pdev->dev, "Done with %p\n", rq);
1843

T
Tejun Heo 已提交
1844
	/* set the residual count for pc requests */
1845
	if (rq->cmd_type == REQ_TYPE_BLOCK_PC)
1846
		rq->resid_len = c->err_info->ResidualCnt;
1847

T
Tejun Heo 已提交
1848
	blk_end_request_all(rq, (rq->errors == 0) ? 0 : -EIO);
1849

1850
	spin_lock_irqsave(&h->lock, flags);
1851
	cmd_free(h, c);
1852
	cciss_check_queues(h);
1853 1854 1855
	spin_unlock_irqrestore(&h->lock, flags);
}

1856 1857
static inline void log_unit_to_scsi3addr(ctlr_info_t *h,
	unsigned char scsi3addr[], uint32_t log_unit)
1858
{
1859 1860
	memcpy(scsi3addr, h->drv[log_unit]->LunID,
		sizeof(h->drv[log_unit]->LunID));
1861 1862
}

1863 1864 1865 1866
/* This function gets the SCSI vendor, model, and revision of a logical drive
 * via the inquiry page 0.  Model, vendor, and rev are set to empty strings if
 * they cannot be read.
 */
1867
static void cciss_get_device_descr(ctlr_info_t *h, int logvol,
1868 1869 1870 1871
				   char *vendor, char *model, char *rev)
{
	int rc;
	InquiryData_struct *inq_buf;
1872
	unsigned char scsi3addr[8];
1873 1874 1875 1876 1877 1878 1879 1880 1881

	*vendor = '\0';
	*model = '\0';
	*rev = '\0';

	inq_buf = kzalloc(sizeof(InquiryData_struct), GFP_KERNEL);
	if (!inq_buf)
		return;

1882 1883
	log_unit_to_scsi3addr(h, scsi3addr, logvol);
	rc = sendcmd_withirq(h, CISS_INQUIRY, inq_buf, sizeof(*inq_buf), 0,
1884
			scsi3addr, TYPE_CMD);
1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897
	if (rc == IO_OK) {
		memcpy(vendor, &inq_buf->data_byte[8], VENDOR_LEN);
		vendor[VENDOR_LEN] = '\0';
		memcpy(model, &inq_buf->data_byte[16], MODEL_LEN);
		model[MODEL_LEN] = '\0';
		memcpy(rev, &inq_buf->data_byte[32], REV_LEN);
		rev[REV_LEN] = '\0';
	}

	kfree(inq_buf);
	return;
}

1898 1899 1900 1901 1902
/* This function gets the serial number of a logical drive via
 * inquiry page 0x83.  Serial no. is 16 bytes.  If the serial
 * number cannot be had, for whatever reason, 16 bytes of 0xff
 * are returned instead.
 */
1903
static void cciss_get_serial_no(ctlr_info_t *h, int logvol,
1904 1905 1906 1907 1908
				unsigned char *serial_no, int buflen)
{
#define PAGE_83_INQ_BYTES 64
	int rc;
	unsigned char *buf;
1909
	unsigned char scsi3addr[8];
1910 1911 1912 1913 1914 1915 1916 1917

	if (buflen > 16)
		buflen = 16;
	memset(serial_no, 0xff, buflen);
	buf = kzalloc(PAGE_83_INQ_BYTES, GFP_KERNEL);
	if (!buf)
		return;
	memset(serial_no, 0, buflen);
1918 1919
	log_unit_to_scsi3addr(h, scsi3addr, logvol);
	rc = sendcmd_withirq(h, CISS_INQUIRY, buf,
1920
		PAGE_83_INQ_BYTES, 0x83, scsi3addr, TYPE_CMD);
1921 1922 1923 1924 1925 1926
	if (rc == IO_OK)
		memcpy(serial_no, &buf[8], buflen);
	kfree(buf);
	return;
}

1927 1928 1929 1930
/*
 * cciss_add_disk sets up the block device queue for a logical drive
 */
static int cciss_add_disk(ctlr_info_t *h, struct gendisk *disk,
M
Mike Miller 已提交
1931 1932 1933
				int drv_index)
{
	disk->queue = blk_init_queue(do_cciss_request, &h->lock);
1934 1935
	if (!disk->queue)
		goto init_queue_failure;
M
Mike Miller 已提交
1936 1937 1938 1939
	sprintf(disk->disk_name, "cciss/c%dd%d", h->ctlr, drv_index);
	disk->major = h->major;
	disk->first_minor = drv_index << NWD_SHIFT;
	disk->fops = &cciss_fops;
1940 1941 1942 1943
	if (cciss_create_ld_sysfs_entry(h, drv_index))
		goto cleanup_queue;
	disk->private_data = h->drv[drv_index];
	disk->driverfs_dev = &h->drv[drv_index]->dev;
M
Mike Miller 已提交
1944 1945 1946 1947 1948

	/* Set up queue information */
	blk_queue_bounce_limit(disk->queue, h->pdev->dma_mask);

	/* This is a hardware imposed limit. */
1949
	blk_queue_max_segments(disk->queue, h->maxsgentries);
M
Mike Miller 已提交
1950

1951
	blk_queue_max_hw_sectors(disk->queue, h->cciss_max_sectors);
M
Mike Miller 已提交
1952 1953 1954 1955 1956

	blk_queue_softirq_done(disk->queue, cciss_softirq_done);

	disk->queue->queuedata = h;

1957
	blk_queue_logical_block_size(disk->queue,
1958
				     h->drv[drv_index]->block_size);
M
Mike Miller 已提交
1959 1960

	/* Make sure all queue data is written out before */
1961
	/* setting h->drv[drv_index]->queue, as setting this */
M
Mike Miller 已提交
1962 1963
	/* allows the interrupt handler to start the queue */
	wmb();
1964
	h->drv[drv_index]->queue = disk->queue;
M
Mike Miller 已提交
1965
	add_disk(disk);
1966 1967 1968 1969 1970
	return 0;

cleanup_queue:
	blk_cleanup_queue(disk->queue);
	disk->queue = NULL;
1971
init_queue_failure:
1972
	return -1;
M
Mike Miller 已提交
1973 1974
}

1975
/* This function will check the usage_count of the drive to be updated/added.
1976 1977 1978 1979 1980 1981 1982 1983
 * If the usage_count is zero and it is a heretofore unknown drive, or,
 * the drive's capacity, geometry, or serial number has changed,
 * then the drive information will be updated and the disk will be
 * re-registered with the kernel.  If these conditions don't hold,
 * then it will be left alone for the next reboot.  The exception to this
 * is disk 0 which will always be left registered with the kernel since it
 * is also the controller node.  Any changes to disk 0 will show up on
 * the next reboot.
1984
 */
1985 1986
static void cciss_update_drive_info(ctlr_info_t *h, int drv_index,
	int first_time, int via_ioctl)
1987
{
1988 1989 1990
	struct gendisk *disk;
	InquiryData_struct *inq_buff = NULL;
	unsigned int block_size;
1991
	sector_t total_size;
1992 1993
	unsigned long flags = 0;
	int ret = 0;
1994 1995 1996 1997
	drive_info_struct *drvinfo;

	/* Get information about the disk and modify the driver structure */
	inq_buff = kmalloc(sizeof(InquiryData_struct), GFP_KERNEL);
1998
	drvinfo = kzalloc(sizeof(*drvinfo), GFP_KERNEL);
1999 2000 2001 2002 2003
	if (inq_buff == NULL || drvinfo == NULL)
		goto mem_msg;

	/* testing to see if 16-byte CDBs are already being used */
	if (h->cciss_read == CCISS_READ_16) {
2004
		cciss_read_capacity_16(h, drv_index,
2005 2006 2007
			&total_size, &block_size);

	} else {
2008
		cciss_read_capacity(h, drv_index, &total_size, &block_size);
2009 2010 2011 2012
		/* if read_capacity returns all F's this volume is >2TB */
		/* in size so we switch to 16-byte CDB's for all */
		/* read/write ops */
		if (total_size == 0xFFFFFFFFULL) {
2013
			cciss_read_capacity_16(h, drv_index,
2014 2015 2016 2017 2018 2019 2020 2021 2022
			&total_size, &block_size);
			h->cciss_read = CCISS_READ_16;
			h->cciss_write = CCISS_WRITE_16;
		} else {
			h->cciss_read = CCISS_READ_10;
			h->cciss_write = CCISS_WRITE_10;
		}
	}

2023
	cciss_geometry_inquiry(h, drv_index, total_size, block_size,
2024 2025 2026 2027
			       inq_buff, drvinfo);
	drvinfo->block_size = block_size;
	drvinfo->nr_blocks = total_size + 1;

2028
	cciss_get_device_descr(h, drv_index, drvinfo->vendor,
2029
				drvinfo->model, drvinfo->rev);
2030
	cciss_get_serial_no(h, drv_index, drvinfo->serial_no,
2031
			sizeof(drvinfo->serial_no));
2032 2033 2034
	/* Save the lunid in case we deregister the disk, below. */
	memcpy(drvinfo->LunID, h->drv[drv_index]->LunID,
		sizeof(drvinfo->LunID));
2035 2036

	/* Is it the same disk we already know, and nothing's changed? */
2037
	if (h->drv[drv_index]->raid_level != -1 &&
2038
		((memcmp(drvinfo->serial_no,
2039 2040 2041 2042 2043 2044
				h->drv[drv_index]->serial_no, 16) == 0) &&
		drvinfo->block_size == h->drv[drv_index]->block_size &&
		drvinfo->nr_blocks == h->drv[drv_index]->nr_blocks &&
		drvinfo->heads == h->drv[drv_index]->heads &&
		drvinfo->sectors == h->drv[drv_index]->sectors &&
		drvinfo->cylinders == h->drv[drv_index]->cylinders))
2045 2046 2047
			/* The disk is unchanged, nothing to update */
			goto freeret;

M
Mike Miller 已提交
2048 2049 2050 2051 2052 2053
	/* If we get here it's not the same disk, or something's changed,
	 * so we need to * deregister it, and re-register it, if it's not
	 * in use.
	 * If the disk already exists then deregister it before proceeding
	 * (unless it's the first disk (for the controller node).
	 */
2054
	if (h->drv[drv_index]->raid_level != -1 && drv_index != 0) {
2055
		dev_warn(&h->pdev->dev, "disk %d has changed.\n", drv_index);
2056
		spin_lock_irqsave(&h->lock, flags);
2057
		h->drv[drv_index]->busy_configuring = 1;
2058
		spin_unlock_irqrestore(&h->lock, flags);
2059

2060
		/* deregister_disk sets h->drv[drv_index]->queue = NULL
M
Mike Miller 已提交
2061 2062 2063
		 * which keeps the interrupt handler from starting
		 * the queue.
		 */
2064
		ret = deregister_disk(h, drv_index, 0, via_ioctl);
2065 2066 2067 2068
	}

	/* If the disk is in use return */
	if (ret)
2069 2070
		goto freeret;

M
Mike Miller 已提交
2071
	/* Save the new information from cciss_geometry_inquiry
2072 2073
	 * and serial number inquiry.  If the disk was deregistered
	 * above, then h->drv[drv_index] will be NULL.
M
Mike Miller 已提交
2074
	 */
2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092
	if (h->drv[drv_index] == NULL) {
		drvinfo->device_initialized = 0;
		h->drv[drv_index] = drvinfo;
		drvinfo = NULL; /* so it won't be freed below. */
	} else {
		/* special case for cxd0 */
		h->drv[drv_index]->block_size = drvinfo->block_size;
		h->drv[drv_index]->nr_blocks = drvinfo->nr_blocks;
		h->drv[drv_index]->heads = drvinfo->heads;
		h->drv[drv_index]->sectors = drvinfo->sectors;
		h->drv[drv_index]->cylinders = drvinfo->cylinders;
		h->drv[drv_index]->raid_level = drvinfo->raid_level;
		memcpy(h->drv[drv_index]->serial_no, drvinfo->serial_no, 16);
		memcpy(h->drv[drv_index]->vendor, drvinfo->vendor,
			VENDOR_LEN + 1);
		memcpy(h->drv[drv_index]->model, drvinfo->model, MODEL_LEN + 1);
		memcpy(h->drv[drv_index]->rev, drvinfo->rev, REV_LEN + 1);
	}
2093 2094 2095

	++h->num_luns;
	disk = h->gendisk[drv_index];
2096
	set_capacity(disk, h->drv[drv_index]->nr_blocks);
2097

M
Mike Miller 已提交
2098 2099 2100 2101 2102 2103
	/* If it's not disk 0 (drv_index != 0)
	 * or if it was disk 0, but there was previously
	 * no actual corresponding configured logical drive
	 * (raid_leve == -1) then we want to update the
	 * logical drive's information.
	 */
2104 2105 2106
	if (drv_index || first_time) {
		if (cciss_add_disk(h, disk, drv_index) != 0) {
			cciss_free_gendisk(h, drv_index);
2107
			cciss_free_drive_info(h, drv_index);
2108 2109
			dev_warn(&h->pdev->dev, "could not update disk %d\n",
				drv_index);
2110 2111 2112
			--h->num_luns;
		}
	}
2113

M
Mike Miller 已提交
2114
freeret:
2115
	kfree(inq_buff);
2116
	kfree(drvinfo);
2117
	return;
M
Mike Miller 已提交
2118
mem_msg:
2119
	dev_err(&h->pdev->dev, "out of memory\n");
2120 2121 2122 2123
	goto freeret;
}

/* This function will find the first index of the controllers drive array
2124 2125 2126 2127 2128 2129 2130 2131
 * that has a null drv pointer and allocate the drive info struct and
 * will return that index   This is where new drives will be added.
 * If the index to be returned is greater than the highest_lun index for
 * the controller then highest_lun is set * to this new index.
 * If there are no available indexes or if tha allocation fails, then -1
 * is returned.  * "controller_node" is used to know if this is a real
 * logical drive, or just the controller node, which determines if this
 * counts towards highest_lun.
2132
 */
2133
static int cciss_alloc_drive_info(ctlr_info_t *h, int controller_node)
2134 2135
{
	int i;
2136
	drive_info_struct *drv;
2137

2138
	/* Search for an empty slot for our drive info */
2139
	for (i = 0; i < CISS_MAX_LUN; i++) {
2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159

		/* if not cxd0 case, and it's occupied, skip it. */
		if (h->drv[i] && i != 0)
			continue;
		/*
		 * If it's cxd0 case, and drv is alloc'ed already, and a
		 * disk is configured there, skip it.
		 */
		if (i == 0 && h->drv[i] && h->drv[i]->raid_level != -1)
			continue;

		/*
		 * We've found an empty slot.  Update highest_lun
		 * provided this isn't just the fake cxd0 controller node.
		 */
		if (i > h->highest_lun && !controller_node)
			h->highest_lun = i;

		/* If adding a real disk at cxd0, and it's already alloc'ed */
		if (i == 0 && h->drv[i] != NULL)
2160
			return i;
2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171

		/*
		 * Found an empty slot, not already alloc'ed.  Allocate it.
		 * Mark it with raid_level == -1, so we know it's new later on.
		 */
		drv = kzalloc(sizeof(*drv), GFP_KERNEL);
		if (!drv)
			return -1;
		drv->raid_level = -1; /* so we know it's new */
		h->drv[i] = drv;
		return i;
2172 2173 2174 2175
	}
	return -1;
}

2176 2177 2178 2179 2180 2181
static void cciss_free_drive_info(ctlr_info_t *h, int drv_index)
{
	kfree(h->drv[drv_index]);
	h->drv[drv_index] = NULL;
}

2182 2183 2184 2185 2186 2187
static void cciss_free_gendisk(ctlr_info_t *h, int drv_index)
{
	put_disk(h->gendisk[drv_index]);
	h->gendisk[drv_index] = NULL;
}

M
Mike Miller 已提交
2188 2189 2190 2191 2192 2193 2194 2195 2196
/* cciss_add_gendisk finds a free hba[]->drv structure
 * and allocates a gendisk if needed, and sets the lunid
 * in the drvinfo structure.   It returns the index into
 * the ->drv[] array, or -1 if none are free.
 * is_controller_node indicates whether highest_lun should
 * count this disk, or if it's only being added to provide
 * a means to talk to the controller in case no logical
 * drives have yet been configured.
 */
2197 2198
static int cciss_add_gendisk(ctlr_info_t *h, unsigned char lunid[],
	int controller_node)
M
Mike Miller 已提交
2199 2200 2201
{
	int drv_index;

2202
	drv_index = cciss_alloc_drive_info(h, controller_node);
M
Mike Miller 已提交
2203 2204
	if (drv_index == -1)
		return -1;
2205

M
Mike Miller 已提交
2206 2207 2208 2209 2210
	/*Check if the gendisk needs to be allocated */
	if (!h->gendisk[drv_index]) {
		h->gendisk[drv_index] =
			alloc_disk(1 << NWD_SHIFT);
		if (!h->gendisk[drv_index]) {
2211 2212 2213
			dev_err(&h->pdev->dev,
				"could not allocate a new disk %d\n",
				drv_index);
2214
			goto err_free_drive_info;
M
Mike Miller 已提交
2215 2216
		}
	}
2217 2218 2219
	memcpy(h->drv[drv_index]->LunID, lunid,
		sizeof(h->drv[drv_index]->LunID));
	if (cciss_create_ld_sysfs_entry(h, drv_index))
2220
		goto err_free_disk;
M
Mike Miller 已提交
2221 2222 2223
	/* Don't need to mark this busy because nobody */
	/* else knows about this disk yet to contend */
	/* for access to it. */
2224
	h->drv[drv_index]->busy_configuring = 0;
M
Mike Miller 已提交
2225 2226
	wmb();
	return drv_index;
2227 2228

err_free_disk:
2229
	cciss_free_gendisk(h, drv_index);
2230 2231
err_free_drive_info:
	cciss_free_drive_info(h, drv_index);
2232
	return -1;
M
Mike Miller 已提交
2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247
}

/* This is for the special case of a controller which
 * has no logical drives.  In this case, we still need
 * to register a disk so the controller can be accessed
 * by the Array Config Utility.
 */
static void cciss_add_controller_node(ctlr_info_t *h)
{
	struct gendisk *disk;
	int drv_index;

	if (h->gendisk[0] != NULL) /* already did this? Then bail. */
		return;

2248
	drv_index = cciss_add_gendisk(h, CTLR_LUNID, 1);
2249 2250
	if (drv_index == -1)
		goto error;
2251 2252 2253 2254 2255 2256 2257
	h->drv[drv_index]->block_size = 512;
	h->drv[drv_index]->nr_blocks = 0;
	h->drv[drv_index]->heads = 0;
	h->drv[drv_index]->sectors = 0;
	h->drv[drv_index]->cylinders = 0;
	h->drv[drv_index]->raid_level = -1;
	memset(h->drv[drv_index]->serial_no, 0, 16);
M
Mike Miller 已提交
2258
	disk = h->gendisk[drv_index];
2259 2260 2261
	if (cciss_add_disk(h, disk, drv_index) == 0)
		return;
	cciss_free_gendisk(h, drv_index);
2262
	cciss_free_drive_info(h, drv_index);
2263
error:
2264
	dev_warn(&h->pdev->dev, "could not add disk 0.\n");
2265
	return;
M
Mike Miller 已提交
2266 2267
}

2268
/* This function will add and remove logical drives from the Logical
2269
 * drive array of the controller and maintain persistency of ordering
2270 2271 2272 2273 2274
 * so that mount points are preserved until the next reboot.  This allows
 * for the removal of logical drives in the middle of the drive array
 * without a re-ordering of those drives.
 * INPUT
 * h		= The controller to perform the operations on
2275
 */
2276 2277
static int rebuild_lun_table(ctlr_info_t *h, int first_time,
	int via_ioctl)
L
Linus Torvalds 已提交
2278
{
2279 2280 2281 2282 2283 2284 2285
	int num_luns;
	ReportLunData_struct *ld_buff = NULL;
	int return_code;
	int listlength = 0;
	int i;
	int drv_found;
	int drv_index = 0;
2286
	unsigned char lunid[8] = CTLR_LUNID;
L
Linus Torvalds 已提交
2287
	unsigned long flags;
2288

M
Mike Miller 已提交
2289 2290 2291
	if (!capable(CAP_SYS_RAWIO))
		return -EPERM;

2292
	/* Set busy_configuring flag for this operation */
2293
	spin_lock_irqsave(&h->lock, flags);
2294
	if (h->busy_configuring) {
2295
		spin_unlock_irqrestore(&h->lock, flags);
2296 2297 2298
		return -EBUSY;
	}
	h->busy_configuring = 1;
2299
	spin_unlock_irqrestore(&h->lock, flags);
2300

2301 2302 2303 2304
	ld_buff = kzalloc(sizeof(ReportLunData_struct), GFP_KERNEL);
	if (ld_buff == NULL)
		goto mem_msg;

2305
	return_code = sendcmd_withirq(h, CISS_REPORT_LOG, ld_buff,
2306 2307
				      sizeof(ReportLunData_struct),
				      0, CTLR_LUNID, TYPE_CMD);
2308

2309 2310 2311
	if (return_code == IO_OK)
		listlength = be32_to_cpu(*(__be32 *) ld_buff->LUNListLength);
	else {	/* reading number of logical volumes failed */
2312 2313
		dev_warn(&h->pdev->dev,
			"report logical volume command failed\n");
2314 2315 2316 2317 2318 2319 2320
		listlength = 0;
		goto freeret;
	}

	num_luns = listlength / 8;	/* 8 bytes per entry */
	if (num_luns > CISS_MAX_LUN) {
		num_luns = CISS_MAX_LUN;
2321
		dev_warn(&h->pdev->dev, "more luns configured"
2322 2323 2324 2325
		       " on controller than can be handled by"
		       " this driver.\n");
	}

M
Mike Miller 已提交
2326 2327 2328 2329 2330 2331 2332 2333
	if (num_luns == 0)
		cciss_add_controller_node(h);

	/* Compare controller drive array to driver's drive array
	 * to see if any drives are missing on the controller due
	 * to action of Array Config Utility (user deletes drive)
	 * and deregister logical drives which have disappeared.
	 */
2334 2335 2336
	for (i = 0; i <= h->highest_lun; i++) {
		int j;
		drv_found = 0;
2337 2338

		/* skip holes in the array from already deleted drives */
2339
		if (h->drv[i] == NULL)
2340 2341
			continue;

2342
		for (j = 0; j < num_luns; j++) {
2343
			memcpy(lunid, &ld_buff->LUN[j][0], sizeof(lunid));
2344
			if (memcmp(h->drv[i]->LunID, lunid,
2345
				sizeof(lunid)) == 0) {
2346 2347 2348 2349 2350 2351
				drv_found = 1;
				break;
			}
		}
		if (!drv_found) {
			/* Deregister it from the OS, it's gone. */
2352
			spin_lock_irqsave(&h->lock, flags);
2353
			h->drv[i]->busy_configuring = 1;
2354
			spin_unlock_irqrestore(&h->lock, flags);
2355
			return_code = deregister_disk(h, i, 1, via_ioctl);
2356 2357
			if (h->drv[i] != NULL)
				h->drv[i]->busy_configuring = 0;
2358
		}
2359
	}
2360

2361 2362 2363 2364 2365 2366 2367
	/* Compare controller drive array to driver's drive array.
	 * Check for updates in the drive information and any new drives
	 * on the controller due to ACU adding logical drives, or changing
	 * a logical drive's size, etc.  Reregister any new/changed drives
	 */
	for (i = 0; i < num_luns; i++) {
		int j;
2368

2369
		drv_found = 0;
2370

2371
		memcpy(lunid, &ld_buff->LUN[i][0], sizeof(lunid));
2372 2373 2374 2375 2376 2377
		/* Find if the LUN is already in the drive array
		 * of the driver.  If so then update its info
		 * if not in use.  If it does not exist then find
		 * the first free index and add it.
		 */
		for (j = 0; j <= h->highest_lun; j++) {
2378 2379 2380
			if (h->drv[j] != NULL &&
				memcmp(h->drv[j]->LunID, lunid,
					sizeof(h->drv[j]->LunID)) == 0) {
2381 2382 2383
				drv_index = j;
				drv_found = 1;
				break;
2384
			}
2385
		}
2386

2387 2388
		/* check if the drive was found already in the array */
		if (!drv_found) {
2389
			drv_index = cciss_add_gendisk(h, lunid, 0);
2390 2391 2392
			if (drv_index == -1)
				goto freeret;
		}
2393
		cciss_update_drive_info(h, drv_index, first_time, via_ioctl);
2394
	}		/* end for */
2395

M
Mike Miller 已提交
2396
freeret:
2397 2398 2399 2400 2401
	kfree(ld_buff);
	h->busy_configuring = 0;
	/* We return -1 here to tell the ACU that we have registered/updated
	 * all of the drives that we can and to keep it from calling us
	 * additional times.
2402
	 */
2403
	return -1;
M
Mike Miller 已提交
2404
mem_msg:
2405
	dev_err(&h->pdev->dev, "out of memory\n");
2406
	h->busy_configuring = 0;
2407 2408 2409
	goto freeret;
}

2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428
static void cciss_clear_drive_info(drive_info_struct *drive_info)
{
	/* zero out the disk size info */
	drive_info->nr_blocks = 0;
	drive_info->block_size = 0;
	drive_info->heads = 0;
	drive_info->sectors = 0;
	drive_info->cylinders = 0;
	drive_info->raid_level = -1;
	memset(drive_info->serial_no, 0, sizeof(drive_info->serial_no));
	memset(drive_info->model, 0, sizeof(drive_info->model));
	memset(drive_info->rev, 0, sizeof(drive_info->rev));
	memset(drive_info->vendor, 0, sizeof(drive_info->vendor));
	/*
	 * don't clear the LUNID though, we need to remember which
	 * one this one is.
	 */
}

2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439
/* This function will deregister the disk and it's queue from the
 * kernel.  It must be called with the controller lock held and the
 * drv structures busy_configuring flag set.  It's parameters are:
 *
 * disk = This is the disk to be deregistered
 * drv  = This is the drive_info_struct associated with the disk to be
 *        deregistered.  It contains information about the disk used
 *        by the driver.
 * clear_all = This flag determines whether or not the disk information
 *             is going to be completely cleared out and the highest_lun
 *             reset.  Sometimes we want to clear out information about
2440
 *             the disk in preparation for re-adding it.  In this case
2441 2442
 *             the highest_lun should be left unchanged and the LunID
 *             should not be cleared.
2443 2444 2445 2446 2447 2448
 * via_ioctl
 *    This indicates whether we've reached this path via ioctl.
 *    This affects the maximum usage count allowed for c0d0 to be messed with.
 *    If this path is reached via ioctl(), then the max_usage_count will
 *    be 1, as the process calling ioctl() has got to have the device open.
 *    If we get here via sysfs, then the max usage count will be zero.
2449
*/
2450
static int deregister_disk(ctlr_info_t *h, int drv_index,
2451
			   int clear_all, int via_ioctl)
2452
{
2453
	int i;
2454 2455
	struct gendisk *disk;
	drive_info_struct *drv;
2456
	int recalculate_highest_lun;
L
Linus Torvalds 已提交
2457 2458 2459 2460

	if (!capable(CAP_SYS_RAWIO))
		return -EPERM;

2461
	drv = h->drv[drv_index];
2462 2463
	disk = h->gendisk[drv_index];

L
Linus Torvalds 已提交
2464
	/* make sure logical volume is NOT is use */
2465
	if (clear_all || (h->gendisk[0] == disk)) {
2466
		if (drv->usage_count > via_ioctl)
2467 2468 2469
			return -EBUSY;
	} else if (drv->usage_count > 0)
		return -EBUSY;
L
Linus Torvalds 已提交
2470

2471 2472
	recalculate_highest_lun = (drv == h->drv[h->highest_lun]);

2473 2474 2475
	/* invalidate the devices and deregister the disk.  If it is disk
	 * zero do not deregister it but just zero out it's values.  This
	 * allows us to delete disk zero but keep the controller registered.
2476 2477
	 */
	if (h->gendisk[0] != disk) {
2478
		struct request_queue *q = disk->queue;
2479
		if (disk->flags & GENHD_FL_UP) {
2480
			cciss_destroy_ld_sysfs_entry(h, drv_index, 0);
2481 2482
			del_gendisk(disk);
		}
2483
		if (q)
2484 2485 2486 2487 2488 2489
			blk_cleanup_queue(q);
		/* If clear_all is set then we are deleting the logical
		 * drive, not just refreshing its info.  For drives
		 * other than disk 0 we will call put_disk.  We do not
		 * do this for disk 0 as we need it to be able to
		 * configure the controller.
2490
		 */
2491 2492 2493 2494 2495
		if (clear_all){
			/* This isn't pretty, but we need to find the
			 * disk in our array and NULL our the pointer.
			 * This is so that we will call alloc_disk if
			 * this index is used again later.
2496
			 */
2497
			for (i=0; i < CISS_MAX_LUN; i++){
2498
				if (h->gendisk[i] == disk) {
2499 2500
					h->gendisk[i] = NULL;
					break;
2501 2502
				}
			}
2503
			put_disk(disk);
2504
		}
2505 2506
	} else {
		set_capacity(disk, 0);
2507
		cciss_clear_drive_info(drv);
2508 2509 2510 2511
	}

	--h->num_luns;

2512 2513
	/* if it was the last disk, find the new hightest lun */
	if (clear_all && recalculate_highest_lun) {
2514
		int newhighest = -1;
2515 2516 2517 2518
		for (i = 0; i <= h->highest_lun; i++) {
			/* if the disk has size > 0, it is available */
			if (h->drv[i] && h->drv[i]->heads)
				newhighest = i;
L
Linus Torvalds 已提交
2519
		}
2520
		h->highest_lun = newhighest;
2521
	}
2522
	return 0;
L
Linus Torvalds 已提交
2523
}
2524

2525
static int fill_cmd(ctlr_info_t *h, CommandList_struct *c, __u8 cmd, void *buff,
2526 2527
		size_t size, __u8 page_code, unsigned char *scsi3addr,
		int cmd_type)
L
Linus Torvalds 已提交
2528 2529 2530 2531 2532 2533
{
	u64bit buff_dma_handle;
	int status = IO_OK;

	c->cmd_type = CMD_IOCTL_PEND;
	c->Header.ReplyQueue = 0;
2534
	if (buff != NULL) {
L
Linus Torvalds 已提交
2535
		c->Header.SGList = 1;
2536
		c->Header.SGTotal = 1;
L
Linus Torvalds 已提交
2537 2538
	} else {
		c->Header.SGList = 0;
2539
		c->Header.SGTotal = 0;
L
Linus Torvalds 已提交
2540 2541
	}
	c->Header.Tag.lower = c->busaddr;
2542
	memcpy(c->Header.LUN.LunAddrBytes, scsi3addr, 8);
L
Linus Torvalds 已提交
2543 2544 2545

	c->Request.Type.Type = cmd_type;
	if (cmd_type == TYPE_CMD) {
2546 2547
		switch (cmd) {
		case CISS_INQUIRY:
L
Linus Torvalds 已提交
2548
			/* are we trying to read a vital product page */
2549
			if (page_code != 0) {
L
Linus Torvalds 已提交
2550 2551 2552 2553
				c->Request.CDB[1] = 0x01;
				c->Request.CDB[2] = page_code;
			}
			c->Request.CDBLen = 6;
2554
			c->Request.Type.Attribute = ATTR_SIMPLE;
L
Linus Torvalds 已提交
2555 2556
			c->Request.Type.Direction = XFER_READ;
			c->Request.Timeout = 0;
2557 2558 2559
			c->Request.CDB[0] = CISS_INQUIRY;
			c->Request.CDB[4] = size & 0xFF;
			break;
L
Linus Torvalds 已提交
2560 2561
		case CISS_REPORT_LOG:
		case CISS_REPORT_PHYS:
2562
			/* Talking to controller so It's a physical command
L
Linus Torvalds 已提交
2563
			   mode = 00 target = 0.  Nothing to write.
2564
			 */
L
Linus Torvalds 已提交
2565 2566 2567 2568 2569
			c->Request.CDBLen = 12;
			c->Request.Type.Attribute = ATTR_SIMPLE;
			c->Request.Type.Direction = XFER_READ;
			c->Request.Timeout = 0;
			c->Request.CDB[0] = cmd;
D
dann frazier 已提交
2570
			c->Request.CDB[6] = (size >> 24) & 0xFF; /* MSB */
L
Linus Torvalds 已提交
2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581
			c->Request.CDB[7] = (size >> 16) & 0xFF;
			c->Request.CDB[8] = (size >> 8) & 0xFF;
			c->Request.CDB[9] = size & 0xFF;
			break;

		case CCISS_READ_CAPACITY:
			c->Request.CDBLen = 10;
			c->Request.Type.Attribute = ATTR_SIMPLE;
			c->Request.Type.Direction = XFER_READ;
			c->Request.Timeout = 0;
			c->Request.CDB[0] = cmd;
2582
			break;
2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596
		case CCISS_READ_CAPACITY_16:
			c->Request.CDBLen = 16;
			c->Request.Type.Attribute = ATTR_SIMPLE;
			c->Request.Type.Direction = XFER_READ;
			c->Request.Timeout = 0;
			c->Request.CDB[0] = cmd;
			c->Request.CDB[1] = 0x10;
			c->Request.CDB[10] = (size >> 24) & 0xFF;
			c->Request.CDB[11] = (size >> 16) & 0xFF;
			c->Request.CDB[12] = (size >> 8) & 0xFF;
			c->Request.CDB[13] = size & 0xFF;
			c->Request.Timeout = 0;
			c->Request.CDB[0] = cmd;
			break;
L
Linus Torvalds 已提交
2597 2598 2599 2600 2601 2602 2603
		case CCISS_CACHE_FLUSH:
			c->Request.CDBLen = 12;
			c->Request.Type.Attribute = ATTR_SIMPLE;
			c->Request.Type.Direction = XFER_WRITE;
			c->Request.Timeout = 0;
			c->Request.CDB[0] = BMIC_WRITE;
			c->Request.CDB[6] = BMIC_CACHE_FLUSH;
2604
			break;
2605 2606 2607 2608 2609 2610
		case TEST_UNIT_READY:
			c->Request.CDBLen = 6;
			c->Request.Type.Attribute = ATTR_SIMPLE;
			c->Request.Type.Direction = XFER_NONE;
			c->Request.Timeout = 0;
			break;
L
Linus Torvalds 已提交
2611
		default:
2612
			dev_warn(&h->pdev->dev, "Unknown Command 0x%c\n", cmd);
2613
			return IO_ERROR;
L
Linus Torvalds 已提交
2614 2615 2616
		}
	} else if (cmd_type == TYPE_MSG) {
		switch (cmd) {
2617
		case CCISS_ABORT_MSG:
2618 2619 2620 2621
			c->Request.CDBLen = 12;
			c->Request.Type.Attribute = ATTR_SIMPLE;
			c->Request.Type.Direction = XFER_WRITE;
			c->Request.Timeout = 0;
2622 2623
			c->Request.CDB[0] = cmd;	/* abort */
			c->Request.CDB[1] = 0;	/* abort a command */
2624 2625 2626
			/* buff contains the tag of the command to abort */
			memcpy(&c->Request.CDB[4], buff, 8);
			break;
2627
		case CCISS_RESET_MSG:
2628
			c->Request.CDBLen = 16;
2629
			c->Request.Type.Attribute = ATTR_SIMPLE;
2630
			c->Request.Type.Direction = XFER_NONE;
2631 2632
			c->Request.Timeout = 0;
			memset(&c->Request.CDB[0], 0, sizeof(c->Request.CDB));
2633
			c->Request.CDB[0] = cmd;	/* reset */
2634
			c->Request.CDB[1] = CCISS_RESET_TYPE_TARGET;
2635
			break;
2636
		case CCISS_NOOP_MSG:
L
Linus Torvalds 已提交
2637 2638 2639 2640 2641 2642 2643
			c->Request.CDBLen = 1;
			c->Request.Type.Attribute = ATTR_SIMPLE;
			c->Request.Type.Direction = XFER_WRITE;
			c->Request.Timeout = 0;
			c->Request.CDB[0] = cmd;
			break;
		default:
2644 2645
			dev_warn(&h->pdev->dev,
				"unknown message type %d\n", cmd);
L
Linus Torvalds 已提交
2646 2647 2648
			return IO_ERROR;
		}
	} else {
2649
		dev_warn(&h->pdev->dev, "unknown command type %d\n", cmd_type);
L
Linus Torvalds 已提交
2650 2651 2652 2653 2654
		return IO_ERROR;
	}
	/* Fill in the scatter gather information */
	if (size > 0) {
		buff_dma_handle.val = (__u64) pci_map_single(h->pdev,
2655 2656
							     buff, size,
							     PCI_DMA_BIDIRECTIONAL);
L
Linus Torvalds 已提交
2657 2658 2659
		c->SG[0].Addr.lower = buff_dma_handle.val32.lower;
		c->SG[0].Addr.upper = buff_dma_handle.val32.upper;
		c->SG[0].Len = size;
2660
		c->SG[0].Ext = 0;	/* we are not chaining */
L
Linus Torvalds 已提交
2661 2662 2663
	}
	return status;
}
2664

J
Jens Axboe 已提交
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
static int __devinit cciss_send_reset(ctlr_info_t *h, unsigned char *scsi3addr,
	u8 reset_type)
{
	CommandList_struct *c;
	int return_status;

	c = cmd_alloc(h);
	if (!c)
		return -ENOMEM;
	return_status = fill_cmd(h, c, CCISS_RESET_MSG, NULL, 0, 0,
		CTLR_LUNID, TYPE_MSG);
	c->Request.CDB[1] = reset_type; /* fill_cmd defaults to target reset */
	if (return_status != IO_OK) {
		cmd_special_free(h, c);
		return return_status;
	}
	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.
	 */
	return 0;
}

2690 2691 2692 2693 2694 2695 2696 2697 2698 2699
static int check_target_status(ctlr_info_t *h, CommandList_struct *c)
{
	switch (c->err_info->ScsiStatus) {
	case SAM_STAT_GOOD:
		return IO_OK;
	case SAM_STAT_CHECK_CONDITION:
		switch (0xf & c->err_info->SenseInfo[2]) {
		case 0: return IO_OK; /* no sense */
		case 1: return IO_OK; /* recovered error */
		default:
2700 2701
			if (check_for_unit_attention(h, c))
				return IO_NEEDS_RETRY;
2702
			dev_warn(&h->pdev->dev, "cmd 0x%02x "
2703
				"check condition, sense key = 0x%02x\n",
2704
				c->Request.CDB[0], c->err_info->SenseInfo[2]);
2705 2706 2707
		}
		break;
	default:
2708 2709
		dev_warn(&h->pdev->dev, "cmd 0x%02x"
			"scsi status = 0x%02x\n",
2710 2711 2712 2713 2714 2715
			c->Request.CDB[0], c->err_info->ScsiStatus);
		break;
	}
	return IO_ERROR;
}

2716
static int process_sendcmd_error(ctlr_info_t *h, CommandList_struct *c)
L
Linus Torvalds 已提交
2717
{
2718
	int return_status = IO_OK;
2719

2720 2721
	if (c->err_info->CommandStatus == CMD_SUCCESS)
		return IO_OK;
2722 2723 2724

	switch (c->err_info->CommandStatus) {
	case CMD_TARGET_STATUS:
2725
		return_status = check_target_status(h, c);
2726 2727 2728 2729 2730 2731
		break;
	case CMD_DATA_UNDERRUN:
	case CMD_DATA_OVERRUN:
		/* expected for inquiry and report lun commands */
		break;
	case CMD_INVALID:
2732
		dev_warn(&h->pdev->dev, "cmd 0x%02x is "
2733 2734 2735 2736
		       "reported invalid\n", c->Request.CDB[0]);
		return_status = IO_ERROR;
		break;
	case CMD_PROTOCOL_ERR:
2737 2738
		dev_warn(&h->pdev->dev, "cmd 0x%02x has "
		       "protocol error\n", c->Request.CDB[0]);
2739 2740 2741
		return_status = IO_ERROR;
		break;
	case CMD_HARDWARE_ERR:
2742
		dev_warn(&h->pdev->dev, "cmd 0x%02x had "
2743 2744 2745 2746
		       " hardware error\n", c->Request.CDB[0]);
		return_status = IO_ERROR;
		break;
	case CMD_CONNECTION_LOST:
2747
		dev_warn(&h->pdev->dev, "cmd 0x%02x had "
2748 2749 2750 2751
		       "connection lost\n", c->Request.CDB[0]);
		return_status = IO_ERROR;
		break;
	case CMD_ABORTED:
2752
		dev_warn(&h->pdev->dev, "cmd 0x%02x was "
2753 2754 2755 2756
		       "aborted\n", c->Request.CDB[0]);
		return_status = IO_ERROR;
		break;
	case CMD_ABORT_FAILED:
2757
		dev_warn(&h->pdev->dev, "cmd 0x%02x reports "
2758 2759 2760 2761
		       "abort failed\n", c->Request.CDB[0]);
		return_status = IO_ERROR;
		break;
	case CMD_UNSOLICITED_ABORT:
2762
		dev_warn(&h->pdev->dev, "unsolicited abort 0x%02x\n",
2763
			c->Request.CDB[0]);
2764
		return_status = IO_NEEDS_RETRY;
2765
		break;
2766 2767 2768 2769
	case CMD_UNABORTABLE:
		dev_warn(&h->pdev->dev, "cmd unabortable\n");
		return_status = IO_ERROR;
		break;
2770
	default:
2771
		dev_warn(&h->pdev->dev, "cmd 0x%02x returned "
2772 2773 2774
		       "unknown status %x\n", c->Request.CDB[0],
		       c->err_info->CommandStatus);
		return_status = IO_ERROR;
2775
	}
2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787
	return return_status;
}

static int sendcmd_withirq_core(ctlr_info_t *h, CommandList_struct *c,
	int attempt_retry)
{
	DECLARE_COMPLETION_ONSTACK(wait);
	u64bit buff_dma_handle;
	int return_status = IO_OK;

resend_cmd2:
	c->waiting = &wait;
M
Mike Miller 已提交
2788
	enqueue_cmd_and_start_io(h, c);
2789 2790 2791 2792 2793 2794 2795 2796 2797 2798

	wait_for_completion(&wait);

	if (c->err_info->CommandStatus == 0 || !attempt_retry)
		goto command_done;

	return_status = process_sendcmd_error(h, c);

	if (return_status == IO_NEEDS_RETRY &&
		c->retry_count < MAX_CMD_RETRIES) {
2799
		dev_warn(&h->pdev->dev, "retrying 0x%02x\n",
2800 2801 2802 2803 2804 2805 2806 2807
			c->Request.CDB[0]);
		c->retry_count++;
		/* erase the old error information */
		memset(c->err_info, 0, sizeof(ErrorInfo_struct));
		return_status = IO_OK;
		INIT_COMPLETION(wait);
		goto resend_cmd2;
	}
2808 2809

command_done:
L
Linus Torvalds 已提交
2810
	/* unlock the buffers from DMA */
2811 2812
	buff_dma_handle.val32.lower = c->SG[0].Addr.lower;
	buff_dma_handle.val32.upper = c->SG[0].Addr.upper;
2813 2814
	pci_unmap_single(h->pdev, (dma_addr_t) buff_dma_handle.val,
			 c->SG[0].Len, PCI_DMA_BIDIRECTIONAL);
2815 2816 2817
	return return_status;
}

2818
static int sendcmd_withirq(ctlr_info_t *h, __u8 cmd, void *buff, size_t size,
2819 2820
			   __u8 page_code, unsigned char scsi3addr[],
			int cmd_type)
2821 2822 2823 2824
{
	CommandList_struct *c;
	int return_status;

2825
	c = cmd_special_alloc(h);
2826 2827
	if (!c)
		return -ENOMEM;
2828
	return_status = fill_cmd(h, c, cmd, buff, size, page_code,
2829
		scsi3addr, cmd_type);
2830
	if (return_status == IO_OK)
2831 2832
		return_status = sendcmd_withirq_core(h, c, 1);

2833
	cmd_special_free(h, c);
2834
	return return_status;
L
Linus Torvalds 已提交
2835
}
2836

2837
static void cciss_geometry_inquiry(ctlr_info_t *h, int logvol,
2838
				   sector_t total_size,
2839 2840 2841
				   unsigned int block_size,
				   InquiryData_struct *inq_buff,
				   drive_info_struct *drv)
L
Linus Torvalds 已提交
2842 2843
{
	int return_code;
2844
	unsigned long t;
2845
	unsigned char scsi3addr[8];
2846

L
Linus Torvalds 已提交
2847
	memset(inq_buff, 0, sizeof(InquiryData_struct));
2848 2849
	log_unit_to_scsi3addr(h, scsi3addr, logvol);
	return_code = sendcmd_withirq(h, CISS_INQUIRY, inq_buff,
2850
			sizeof(*inq_buff), 0xC1, scsi3addr, TYPE_CMD);
L
Linus Torvalds 已提交
2851
	if (return_code == IO_OK) {
2852
		if (inq_buff->data_byte[8] == 0xFF) {
2853 2854
			dev_warn(&h->pdev->dev,
			       "reading geometry failed, volume "
2855
			       "does not support reading geometry\n");
L
Linus Torvalds 已提交
2856
			drv->heads = 255;
D
dann frazier 已提交
2857
			drv->sectors = 32;	/* Sectors per track */
2858
			drv->cylinders = total_size + 1;
2859
			drv->raid_level = RAID_UNKNOWN;
L
Linus Torvalds 已提交
2860 2861 2862 2863 2864 2865
		} else {
			drv->heads = inq_buff->data_byte[6];
			drv->sectors = inq_buff->data_byte[7];
			drv->cylinders = (inq_buff->data_byte[4] & 0xff) << 8;
			drv->cylinders += inq_buff->data_byte[5];
			drv->raid_level = inq_buff->data_byte[8];
2866 2867
		}
		drv->block_size = block_size;
2868
		drv->nr_blocks = total_size + 1;
2869 2870
		t = drv->heads * drv->sectors;
		if (t > 1) {
2871 2872
			sector_t real_size = total_size + 1;
			unsigned long rem = sector_div(real_size, t);
2873
			if (rem)
2874 2875
				real_size++;
			drv->cylinders = real_size;
L
Linus Torvalds 已提交
2876
		}
2877
	} else {		/* Get geometry failed */
2878
		dev_warn(&h->pdev->dev, "reading geometry failed\n");
L
Linus Torvalds 已提交
2879 2880
	}
}
2881

L
Linus Torvalds 已提交
2882
static void
2883
cciss_read_capacity(ctlr_info_t *h, int logvol, sector_t *total_size,
2884
		    unsigned int *block_size)
L
Linus Torvalds 已提交
2885
{
2886
	ReadCapdata_struct *buf;
L
Linus Torvalds 已提交
2887
	int return_code;
2888
	unsigned char scsi3addr[8];
2889 2890 2891

	buf = kzalloc(sizeof(ReadCapdata_struct), GFP_KERNEL);
	if (!buf) {
2892
		dev_warn(&h->pdev->dev, "out of memory\n");
2893 2894
		return;
	}
2895

2896 2897
	log_unit_to_scsi3addr(h, scsi3addr, logvol);
	return_code = sendcmd_withirq(h, CCISS_READ_CAPACITY, buf,
2898
		sizeof(ReadCapdata_struct), 0, scsi3addr, TYPE_CMD);
L
Linus Torvalds 已提交
2899
	if (return_code == IO_OK) {
A
Al Viro 已提交
2900 2901
		*total_size = be32_to_cpu(*(__be32 *) buf->total_size);
		*block_size = be32_to_cpu(*(__be32 *) buf->block_size);
2902
	} else {		/* read capacity command failed */
2903
		dev_warn(&h->pdev->dev, "read capacity failed\n");
L
Linus Torvalds 已提交
2904 2905 2906
		*total_size = 0;
		*block_size = BLOCK_SIZE;
	}
2907 2908 2909
	kfree(buf);
}

2910
static void cciss_read_capacity_16(ctlr_info_t *h, int logvol,
2911
	sector_t *total_size, unsigned int *block_size)
2912 2913 2914
{
	ReadCapdata_struct_16 *buf;
	int return_code;
2915
	unsigned char scsi3addr[8];
2916 2917 2918

	buf = kzalloc(sizeof(ReadCapdata_struct_16), GFP_KERNEL);
	if (!buf) {
2919
		dev_warn(&h->pdev->dev, "out of memory\n");
2920 2921
		return;
	}
2922

2923 2924 2925
	log_unit_to_scsi3addr(h, scsi3addr, logvol);
	return_code = sendcmd_withirq(h, CCISS_READ_CAPACITY_16,
		buf, sizeof(ReadCapdata_struct_16),
2926
			0, scsi3addr, TYPE_CMD);
2927
	if (return_code == IO_OK) {
A
Al Viro 已提交
2928 2929
		*total_size = be64_to_cpu(*(__be64 *) buf->total_size);
		*block_size = be32_to_cpu(*(__be32 *) buf->block_size);
2930
	} else {		/* read capacity command failed */
2931
		dev_warn(&h->pdev->dev, "read capacity failed\n");
2932 2933 2934
		*total_size = 0;
		*block_size = BLOCK_SIZE;
	}
2935
	dev_info(&h->pdev->dev, "      blocks= %llu block_size= %d\n",
2936
	       (unsigned long long)*total_size+1, *block_size);
2937
	kfree(buf);
L
Linus Torvalds 已提交
2938 2939 2940 2941 2942 2943 2944
}

static int cciss_revalidate(struct gendisk *disk)
{
	ctlr_info_t *h = get_host(disk);
	drive_info_struct *drv = get_drv(disk);
	int logvol;
2945
	int FOUND = 0;
L
Linus Torvalds 已提交
2946
	unsigned int block_size;
2947
	sector_t total_size;
L
Linus Torvalds 已提交
2948 2949
	InquiryData_struct *inq_buff = NULL;

2950
	for (logvol = 0; logvol <= h->highest_lun; logvol++) {
2951
		if (!h->drv[logvol])
2952
			continue;
2953
		if (memcmp(h->drv[logvol]->LunID, drv->LunID,
2954
			sizeof(drv->LunID)) == 0) {
2955
			FOUND = 1;
L
Linus Torvalds 已提交
2956 2957 2958 2959
			break;
		}
	}

2960 2961
	if (!FOUND)
		return 1;
L
Linus Torvalds 已提交
2962

2963 2964
	inq_buff = kmalloc(sizeof(InquiryData_struct), GFP_KERNEL);
	if (inq_buff == NULL) {
2965
		dev_warn(&h->pdev->dev, "out of memory\n");
2966 2967
		return 1;
	}
2968
	if (h->cciss_read == CCISS_READ_10) {
2969
		cciss_read_capacity(h, logvol,
2970 2971
					&total_size, &block_size);
	} else {
2972
		cciss_read_capacity_16(h, logvol,
2973 2974
					&total_size, &block_size);
	}
2975
	cciss_geometry_inquiry(h, logvol, total_size, block_size,
2976
			       inq_buff, drv);
L
Linus Torvalds 已提交
2977

2978
	blk_queue_logical_block_size(drv->queue, drv->block_size);
L
Linus Torvalds 已提交
2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989
	set_capacity(disk, drv->nr_blocks);

	kfree(inq_buff);
	return 0;
}

/*
 * Map (physical) PCI mem into (virtual) kernel space
 */
static void __iomem *remap_pci_mem(ulong base, ulong size)
{
2990 2991 2992
	ulong page_base = ((ulong) base) & PAGE_MASK;
	ulong page_offs = ((ulong) base) - page_base;
	void __iomem *page_remapped = ioremap(page_base, page_offs + size);
L
Linus Torvalds 已提交
2993

2994
	return page_remapped ? (page_remapped + page_offs) : NULL;
L
Linus Torvalds 已提交
2995 2996
}

2997 2998 2999 3000 3001
/*
 * Takes jobs of the Q and sends them to the hardware, then puts it on
 * the Q to wait for completion.
 */
static void start_io(ctlr_info_t *h)
L
Linus Torvalds 已提交
3002 3003
{
	CommandList_struct *c;
3004

3005 3006
	while (!list_empty(&h->reqQ)) {
		c = list_entry(h->reqQ.next, CommandList_struct, list);
L
Linus Torvalds 已提交
3007 3008
		/* can't do anything if fifo is full */
		if ((h->access.fifo_full(h))) {
3009
			dev_warn(&h->pdev->dev, "fifo full\n");
L
Linus Torvalds 已提交
3010 3011 3012
			break;
		}

3013
		/* Get the first entry from the Request Q */
3014
		removeQ(c);
L
Linus Torvalds 已提交
3015
		h->Qdepth--;
3016 3017

		/* Tell the controller execute command */
L
Linus Torvalds 已提交
3018
		h->access.submit_command(h, c);
3019 3020

		/* Put job onto the completed Q */
3021
		addQ(&h->cmpQ, c);
L
Linus Torvalds 已提交
3022 3023
	}
}
3024

3025
/* Assumes that h->lock is held. */
L
Linus Torvalds 已提交
3026 3027
/* Zeros out the error record and then resends the command back */
/* to the controller */
3028
static inline void resend_cciss_cmd(ctlr_info_t *h, CommandList_struct *c)
L
Linus Torvalds 已提交
3029 3030 3031 3032 3033
{
	/* erase the old error information */
	memset(c->err_info, 0, sizeof(ErrorInfo_struct));

	/* add it to software queue and then send it to the controller */
3034
	addQ(&h->reqQ, c);
L
Linus Torvalds 已提交
3035
	h->Qdepth++;
3036
	if (h->Qdepth > h->maxQsinceinit)
L
Linus Torvalds 已提交
3037 3038 3039 3040
		h->maxQsinceinit = h->Qdepth;

	start_io(h);
}
3041

3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052
static inline unsigned int make_status_bytes(unsigned int scsi_status_byte,
	unsigned int msg_byte, unsigned int host_byte,
	unsigned int driver_byte)
{
	/* inverse of macros in scsi.h */
	return (scsi_status_byte & 0xff) |
		((msg_byte & 0xff) << 8) |
		((host_byte & 0xff) << 16) |
		((driver_byte & 0xff) << 24);
}

3053 3054
static inline int evaluate_target_status(ctlr_info_t *h,
			CommandList_struct *cmd, int *retry_cmd)
3055 3056
{
	unsigned char sense_key;
3057 3058 3059
	unsigned char status_byte, msg_byte, host_byte, driver_byte;
	int error_value;

3060
	*retry_cmd = 0;
3061 3062 3063 3064 3065
	/* If we get in here, it means we got "target status", that is, scsi status */
	status_byte = cmd->err_info->ScsiStatus;
	driver_byte = DRIVER_OK;
	msg_byte = cmd->err_info->CommandStatus; /* correct?  seems too device specific */

3066
	if (cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC)
3067 3068 3069 3070 3071 3072
		host_byte = DID_PASSTHROUGH;
	else
		host_byte = DID_OK;

	error_value = make_status_bytes(status_byte, msg_byte,
		host_byte, driver_byte);
3073

3074
	if (cmd->err_info->ScsiStatus != SAM_STAT_CHECK_CONDITION) {
3075
		if (cmd->rq->cmd_type != REQ_TYPE_BLOCK_PC)
3076
			dev_warn(&h->pdev->dev, "cmd %p "
3077 3078
			       "has SCSI Status 0x%x\n",
			       cmd, cmd->err_info->ScsiStatus);
3079
		return error_value;
3080 3081 3082 3083 3084
	}

	/* check the sense key */
	sense_key = 0xf & cmd->err_info->SenseInfo[2];
	/* no status or recovered error */
3085 3086
	if (((sense_key == 0x0) || (sense_key == 0x1)) &&
	    (cmd->rq->cmd_type != REQ_TYPE_BLOCK_PC))
3087
		error_value = 0;
3088

3089
	if (check_for_unit_attention(h, cmd)) {
3090
		*retry_cmd = !(cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC);
3091 3092 3093
		return 0;
	}

3094 3095
	/* Not SG_IO or similar? */
	if (cmd->rq->cmd_type != REQ_TYPE_BLOCK_PC) {
3096
		if (error_value != 0)
3097
			dev_warn(&h->pdev->dev, "cmd %p has CHECK CONDITION"
3098
			       " sense key = 0x%x\n", cmd, sense_key);
3099
		return error_value;
3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110
	}

	/* SG_IO or similar, copy sense data back */
	if (cmd->rq->sense) {
		if (cmd->rq->sense_len > cmd->err_info->SenseLen)
			cmd->rq->sense_len = cmd->err_info->SenseLen;
		memcpy(cmd->rq->sense, cmd->err_info->SenseInfo,
			cmd->rq->sense_len);
	} else
		cmd->rq->sense_len = 0;

3111
	return error_value;
3112 3113
}

3114
/* checks the status of the job and calls complete buffers to mark all
3115 3116
 * buffers for the completed job. Note that this function does not need
 * to hold the hba/queue lock.
3117 3118 3119
 */
static inline void complete_command(ctlr_info_t *h, CommandList_struct *cmd,
				    int timeout)
L
Linus Torvalds 已提交
3120 3121
{
	int retry_cmd = 0;
3122 3123 3124
	struct request *rq = cmd->rq;

	rq->errors = 0;
3125

L
Linus Torvalds 已提交
3126
	if (timeout)
3127
		rq->errors = make_status_bytes(0, 0, 0, DRIVER_TIMEOUT);
L
Linus Torvalds 已提交
3128

3129 3130
	if (cmd->err_info->CommandStatus == 0)	/* no error has occurred */
		goto after_error_processing;
3131

3132 3133
	switch (cmd->err_info->CommandStatus) {
	case CMD_TARGET_STATUS:
3134
		rq->errors = evaluate_target_status(h, cmd, &retry_cmd);
3135 3136
		break;
	case CMD_DATA_UNDERRUN:
3137
		if (cmd->rq->cmd_type == REQ_TYPE_FS) {
3138
			dev_warn(&h->pdev->dev, "cmd %p has"
3139 3140
			       " completed with data underrun "
			       "reported\n", cmd);
T
Tejun Heo 已提交
3141
			cmd->rq->resid_len = cmd->err_info->ResidualCnt;
3142
		}
3143 3144
		break;
	case CMD_DATA_OVERRUN:
3145
		if (cmd->rq->cmd_type == REQ_TYPE_FS)
3146
			dev_warn(&h->pdev->dev, "cciss: cmd %p has"
3147 3148
			       " completed with data overrun "
			       "reported\n", cmd);
3149 3150
		break;
	case CMD_INVALID:
3151
		dev_warn(&h->pdev->dev, "cciss: cmd %p is "
3152
		       "reported invalid\n", cmd);
3153 3154
		rq->errors = make_status_bytes(SAM_STAT_GOOD,
			cmd->err_info->CommandStatus, DRIVER_OK,
3155 3156
			(cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ?
				DID_PASSTHROUGH : DID_ERROR);
3157 3158
		break;
	case CMD_PROTOCOL_ERR:
3159 3160
		dev_warn(&h->pdev->dev, "cciss: cmd %p has "
		       "protocol error\n", cmd);
3161 3162
		rq->errors = make_status_bytes(SAM_STAT_GOOD,
			cmd->err_info->CommandStatus, DRIVER_OK,
3163 3164
			(cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ?
				DID_PASSTHROUGH : DID_ERROR);
3165 3166
		break;
	case CMD_HARDWARE_ERR:
3167
		dev_warn(&h->pdev->dev, "cciss: cmd %p had "
3168
		       " hardware error\n", cmd);
3169 3170
		rq->errors = make_status_bytes(SAM_STAT_GOOD,
			cmd->err_info->CommandStatus, DRIVER_OK,
3171 3172
			(cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ?
				DID_PASSTHROUGH : DID_ERROR);
3173 3174
		break;
	case CMD_CONNECTION_LOST:
3175
		dev_warn(&h->pdev->dev, "cciss: cmd %p had "
3176
		       "connection lost\n", cmd);
3177 3178
		rq->errors = make_status_bytes(SAM_STAT_GOOD,
			cmd->err_info->CommandStatus, DRIVER_OK,
3179 3180
			(cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ?
				DID_PASSTHROUGH : DID_ERROR);
3181 3182
		break;
	case CMD_ABORTED:
3183
		dev_warn(&h->pdev->dev, "cciss: cmd %p was "
3184
		       "aborted\n", cmd);
3185 3186
		rq->errors = make_status_bytes(SAM_STAT_GOOD,
			cmd->err_info->CommandStatus, DRIVER_OK,
3187 3188
			(cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ?
				DID_PASSTHROUGH : DID_ABORT);
3189 3190
		break;
	case CMD_ABORT_FAILED:
3191
		dev_warn(&h->pdev->dev, "cciss: cmd %p reports "
3192
		       "abort failed\n", cmd);
3193 3194
		rq->errors = make_status_bytes(SAM_STAT_GOOD,
			cmd->err_info->CommandStatus, DRIVER_OK,
3195 3196
			(cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ?
				DID_PASSTHROUGH : DID_ERROR);
3197 3198
		break;
	case CMD_UNSOLICITED_ABORT:
3199
		dev_warn(&h->pdev->dev, "cciss%d: unsolicited "
3200 3201 3202
		       "abort %p\n", h->ctlr, cmd);
		if (cmd->retry_count < MAX_CMD_RETRIES) {
			retry_cmd = 1;
3203
			dev_warn(&h->pdev->dev, "retrying %p\n", cmd);
3204 3205
			cmd->retry_count++;
		} else
3206 3207
			dev_warn(&h->pdev->dev,
				"%p retried too many times\n", cmd);
3208 3209
		rq->errors = make_status_bytes(SAM_STAT_GOOD,
			cmd->err_info->CommandStatus, DRIVER_OK,
3210 3211
			(cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ?
				DID_PASSTHROUGH : DID_ABORT);
3212 3213
		break;
	case CMD_TIMEOUT:
3214
		dev_warn(&h->pdev->dev, "cmd %p timedout\n", cmd);
3215 3216
		rq->errors = make_status_bytes(SAM_STAT_GOOD,
			cmd->err_info->CommandStatus, DRIVER_OK,
3217 3218
			(cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ?
				DID_PASSTHROUGH : DID_ERROR);
3219
		break;
3220 3221 3222 3223 3224 3225 3226
	case CMD_UNABORTABLE:
		dev_warn(&h->pdev->dev, "cmd %p unabortable\n", cmd);
		rq->errors = make_status_bytes(SAM_STAT_GOOD,
			cmd->err_info->CommandStatus, DRIVER_OK,
			cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC ?
				DID_PASSTHROUGH : DID_ERROR);
		break;
3227
	default:
3228
		dev_warn(&h->pdev->dev, "cmd %p returned "
3229 3230
		       "unknown status %x\n", cmd,
		       cmd->err_info->CommandStatus);
3231 3232
		rq->errors = make_status_bytes(SAM_STAT_GOOD,
			cmd->err_info->CommandStatus, DRIVER_OK,
3233 3234
			(cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ?
				DID_PASSTHROUGH : DID_ERROR);
L
Linus Torvalds 已提交
3235
	}
3236 3237 3238

after_error_processing:

L
Linus Torvalds 已提交
3239
	/* We need to return this command */
3240 3241
	if (retry_cmd) {
		resend_cciss_cmd(h, cmd);
L
Linus Torvalds 已提交
3242
		return;
3243
	}
3244
	cmd->rq->completion_data = cmd;
3245
	blk_complete_request(cmd->rq);
L
Linus Torvalds 已提交
3246 3247
}

M
Mike Miller 已提交
3248 3249
static inline u32 cciss_tag_contains_index(u32 tag)
{
3250
#define DIRECT_LOOKUP_BIT 0x10
M
Mike Miller 已提交
3251 3252 3253 3254 3255
	return tag & DIRECT_LOOKUP_BIT;
}

static inline u32 cciss_tag_to_index(u32 tag)
{
3256
#define DIRECT_LOOKUP_SHIFT 5
M
Mike Miller 已提交
3257 3258 3259
	return tag >> DIRECT_LOOKUP_SHIFT;
}

3260
static inline u32 cciss_tag_discard_error_bits(ctlr_info_t *h, u32 tag)
M
Mike Miller 已提交
3261
{
3262 3263 3264 3265 3266
#define CCISS_PERF_ERROR_BITS ((1 << DIRECT_LOOKUP_SHIFT) - 1)
#define CCISS_SIMPLE_ERROR_BITS 0x03
	if (likely(h->transMethod & CFGTBL_Trans_Performant))
		return tag & ~CCISS_PERF_ERROR_BITS;
	return tag & ~CCISS_SIMPLE_ERROR_BITS;
M
Mike Miller 已提交
3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278
}

static inline void cciss_mark_tag_indexed(u32 *tag)
{
	*tag |= DIRECT_LOOKUP_BIT;
}

static inline void cciss_set_tag_index(u32 *tag, u32 index)
{
	*tag |= (index << DIRECT_LOOKUP_SHIFT);
}

3279 3280
/*
 * Get a request and submit it to the controller.
L
Linus Torvalds 已提交
3281
 */
3282
static void do_cciss_request(struct request_queue *q)
L
Linus Torvalds 已提交
3283
{
3284
	ctlr_info_t *h = q->queuedata;
L
Linus Torvalds 已提交
3285
	CommandList_struct *c;
3286 3287
	sector_t start_blk;
	int seg;
L
Linus Torvalds 已提交
3288 3289
	struct request *creq;
	u64bit temp64;
3290 3291
	struct scatterlist *tmp_sg;
	SGDescriptor_struct *curr_sg;
L
Linus Torvalds 已提交
3292 3293
	drive_info_struct *drv;
	int i, dir;
3294 3295
	int sg_index = 0;
	int chained = 0;
L
Linus Torvalds 已提交
3296

3297
      queue:
3298
	creq = blk_peek_request(q);
L
Linus Torvalds 已提交
3299 3300 3301
	if (!creq)
		goto startio;

3302
	BUG_ON(creq->nr_phys_segments > h->maxsgentries);
L
Linus Torvalds 已提交
3303

3304 3305
	c = cmd_alloc(h);
	if (!c)
L
Linus Torvalds 已提交
3306 3307
		goto full;

3308
	blk_start_request(creq);
L
Linus Torvalds 已提交
3309

3310
	tmp_sg = h->scatter_list[c->cmdindex];
L
Linus Torvalds 已提交
3311 3312 3313 3314
	spin_unlock_irq(q->queue_lock);

	c->cmd_type = CMD_RWREQ;
	c->rq = creq;
3315 3316

	/* fill in the request */
L
Linus Torvalds 已提交
3317
	drv = creq->rq_disk->private_data;
D
dann frazier 已提交
3318
	c->Header.ReplyQueue = 0;	/* unused in simple mode */
3319 3320 3321
	/* got command from pool, so use the command block index instead */
	/* for direct lookups. */
	/* The first 2 bits are reserved for controller error reporting. */
M
Mike Miller 已提交
3322 3323
	cciss_set_tag_index(&c->Header.Tag.lower, c->cmdindex);
	cciss_mark_tag_indexed(&c->Header.Tag.lower);
3324
	memcpy(&c->Header.LUN, drv->LunID, sizeof(drv->LunID));
D
dann frazier 已提交
3325 3326
	c->Request.CDBLen = 10;	/* 12 byte commands not in FW yet; */
	c->Request.Type.Type = TYPE_CMD;	/* It is a command. */
3327 3328
	c->Request.Type.Attribute = ATTR_SIMPLE;
	c->Request.Type.Direction =
3329
	    (rq_data_dir(creq) == READ) ? XFER_READ : XFER_WRITE;
D
dann frazier 已提交
3330
	c->Request.Timeout = 0;	/* Don't time out */
3331
	c->Request.CDB[0] =
3332
	    (rq_data_dir(creq) == READ) ? h->cciss_read : h->cciss_write;
3333
	start_blk = blk_rq_pos(creq);
3334
	dev_dbg(&h->pdev->dev, "sector =%d nr_sectors=%d\n",
3335
	       (int)blk_rq_pos(creq), (int)blk_rq_sectors(creq));
3336
	sg_init_table(tmp_sg, h->maxsgentries);
L
Linus Torvalds 已提交
3337 3338
	seg = blk_rq_map_sg(q, creq, tmp_sg);

3339
	/* get the DMA records for the setup */
L
Linus Torvalds 已提交
3340 3341 3342 3343 3344
	if (c->Request.Type.Direction == XFER_READ)
		dir = PCI_DMA_FROMDEVICE;
	else
		dir = PCI_DMA_TODEVICE;

3345 3346 3347 3348
	curr_sg = c->SG;
	sg_index = 0;
	chained = 0;

3349
	for (i = 0; i < seg; i++) {
3350 3351 3352
		if (((sg_index+1) == (h->max_cmd_sgentries)) &&
			!chained && ((seg - i) > 1)) {
			/* Point to next chain block. */
3353
			curr_sg = h->cmd_sg_list[c->cmdindex];
3354 3355 3356 3357
			sg_index = 0;
			chained = 1;
		}
		curr_sg[sg_index].Len = tmp_sg[i].length;
J
Jens Axboe 已提交
3358
		temp64.val = (__u64) pci_map_page(h->pdev, sg_page(&tmp_sg[i]),
3359 3360 3361 3362 3363 3364
						tmp_sg[i].offset,
						tmp_sg[i].length, dir);
		curr_sg[sg_index].Addr.lower = temp64.val32.lower;
		curr_sg[sg_index].Addr.upper = temp64.val32.upper;
		curr_sg[sg_index].Ext = 0;  /* we are not chaining */
		++sg_index;
L
Linus Torvalds 已提交
3365
	}
3366 3367 3368 3369
	if (chained)
		cciss_map_sg_chain_block(h, c, h->cmd_sg_list[c->cmdindex],
			(seg - (h->max_cmd_sgentries - 1)) *
				sizeof(SGDescriptor_struct));
3370

3371 3372 3373
	/* track how many SG entries we are using */
	if (seg > h->maxSG)
		h->maxSG = seg;
L
Linus Torvalds 已提交
3374

3375
	dev_dbg(&h->pdev->dev, "Submitting %u sectors in %d segments "
3376 3377
			"chained[%d]\n",
			blk_rq_sectors(creq), seg, chained);
L
Linus Torvalds 已提交
3378

3379 3380 3381 3382
	c->Header.SGTotal = seg + chained;
	if (seg <= h->max_cmd_sgentries)
		c->Header.SGList = c->Header.SGTotal;
	else
3383
		c->Header.SGList = h->max_cmd_sgentries;
3384
	set_performant_mode(h, c);
3385

3386
	if (likely(creq->cmd_type == REQ_TYPE_FS)) {
3387 3388
		if(h->cciss_read == CCISS_READ_10) {
			c->Request.CDB[1] = 0;
D
dann frazier 已提交
3389
			c->Request.CDB[2] = (start_blk >> 24) & 0xff; /* MSB */
3390 3391 3392
			c->Request.CDB[3] = (start_blk >> 16) & 0xff;
			c->Request.CDB[4] = (start_blk >> 8) & 0xff;
			c->Request.CDB[5] = start_blk & 0xff;
D
dann frazier 已提交
3393
			c->Request.CDB[6] = 0; /* (sect >> 24) & 0xff; MSB */
3394 3395
			c->Request.CDB[7] = (blk_rq_sectors(creq) >> 8) & 0xff;
			c->Request.CDB[8] = blk_rq_sectors(creq) & 0xff;
3396 3397
			c->Request.CDB[9] = c->Request.CDB[11] = c->Request.CDB[12] = 0;
		} else {
3398 3399
			u32 upper32 = upper_32_bits(start_blk);

3400 3401
			c->Request.CDBLen = 16;
			c->Request.CDB[1]= 0;
D
dann frazier 已提交
3402
			c->Request.CDB[2]= (upper32 >> 24) & 0xff; /* MSB */
3403 3404 3405
			c->Request.CDB[3]= (upper32 >> 16) & 0xff;
			c->Request.CDB[4]= (upper32 >>  8) & 0xff;
			c->Request.CDB[5]= upper32 & 0xff;
3406 3407 3408 3409
			c->Request.CDB[6]= (start_blk >> 24) & 0xff;
			c->Request.CDB[7]= (start_blk >> 16) & 0xff;
			c->Request.CDB[8]= (start_blk >>  8) & 0xff;
			c->Request.CDB[9]= start_blk & 0xff;
3410 3411 3412 3413
			c->Request.CDB[10]= (blk_rq_sectors(creq) >> 24) & 0xff;
			c->Request.CDB[11]= (blk_rq_sectors(creq) >> 16) & 0xff;
			c->Request.CDB[12]= (blk_rq_sectors(creq) >>  8) & 0xff;
			c->Request.CDB[13]= blk_rq_sectors(creq) & 0xff;
3414 3415
			c->Request.CDB[14] = c->Request.CDB[15] = 0;
		}
3416
	} else if (creq->cmd_type == REQ_TYPE_BLOCK_PC) {
3417 3418
		c->Request.CDBLen = creq->cmd_len;
		memcpy(c->Request.CDB, creq->cmd, BLK_MAX_CDB);
3419
	} else {
3420 3421
		dev_warn(&h->pdev->dev, "bad request type %d\n",
			creq->cmd_type);
3422
		BUG();
3423
	}
L
Linus Torvalds 已提交
3424 3425 3426

	spin_lock_irq(q->queue_lock);

3427
	addQ(&h->reqQ, c);
L
Linus Torvalds 已提交
3428
	h->Qdepth++;
3429 3430
	if (h->Qdepth > h->maxQsinceinit)
		h->maxQsinceinit = h->Qdepth;
L
Linus Torvalds 已提交
3431 3432

	goto queue;
3433
full:
L
Linus Torvalds 已提交
3434
	blk_stop_queue(q);
3435
startio:
L
Linus Torvalds 已提交
3436 3437
	/* We will already have the driver lock here so not need
	 * to lock it.
3438
	 */
L
Linus Torvalds 已提交
3439 3440 3441
	start_io(h);
}

3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453
static inline unsigned long get_next_completion(ctlr_info_t *h)
{
	return h->access.command_completed(h);
}

static inline int interrupt_pending(ctlr_info_t *h)
{
	return h->access.intr_pending(h);
}

static inline long interrupt_not_for_us(ctlr_info_t *h)
{
3454
	return ((h->access.intr_pending(h) == 0) ||
3455
		(h->interrupts_enabled == 0));
3456 3457
}

M
Mike Miller 已提交
3458 3459
static inline int bad_tag(ctlr_info_t *h, u32 tag_index,
			u32 raw_tag)
L
Linus Torvalds 已提交
3460
{
M
Mike Miller 已提交
3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481
	if (unlikely(tag_index >= h->nr_cmds)) {
		dev_warn(&h->pdev->dev, "bad tag 0x%08x ignored.\n", raw_tag);
		return 1;
	}
	return 0;
}

static inline void finish_cmd(ctlr_info_t *h, CommandList_struct *c,
				u32 raw_tag)
{
	removeQ(c);
	if (likely(c->cmd_type == CMD_RWREQ))
		complete_command(h, c, 0);
	else if (c->cmd_type == CMD_IOCTL_PEND)
		complete(c->waiting);
#ifdef CONFIG_CISS_SCSI_TAPE
	else if (c->cmd_type == CMD_SCSI)
		complete_scsi_command(c, 0, raw_tag);
#endif
}

3482 3483 3484 3485
static inline u32 next_command(ctlr_info_t *h)
{
	u32 a;

3486
	if (unlikely(!(h->transMethod & CFGTBL_Trans_Performant)))
3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503
		return h->access.command_completed(h);

	if ((*(h->reply_pool_head) & 1) == (h->reply_pool_wraparound)) {
		a = *(h->reply_pool_head); /* Next cmd in ring buffer */
		(h->reply_pool_head)++;
		h->commands_outstanding--;
	} else {
		a = FIFO_EMPTY;
	}
	/* Check for wraparound */
	if (h->reply_pool_head == (h->reply_pool + h->max_commands)) {
		h->reply_pool_head = h->reply_pool;
		h->reply_pool_wraparound ^= 1;
	}
	return a;
}

M
Mike Miller 已提交
3504 3505 3506 3507
/* process completion of an indexed ("direct lookup") command */
static inline u32 process_indexed_cmd(ctlr_info_t *h, u32 raw_tag)
{
	u32 tag_index;
L
Linus Torvalds 已提交
3508
	CommandList_struct *c;
M
Mike Miller 已提交
3509 3510 3511

	tag_index = cciss_tag_to_index(raw_tag);
	if (bad_tag(h, tag_index, raw_tag))
3512
		return next_command(h);
M
Mike Miller 已提交
3513 3514
	c = h->cmd_pool + tag_index;
	finish_cmd(h, c, raw_tag);
3515
	return next_command(h);
M
Mike Miller 已提交
3516 3517 3518 3519 3520 3521 3522 3523
}

/* process completion of a non-indexed command */
static inline u32 process_nonindexed_cmd(ctlr_info_t *h, u32 raw_tag)
{
	CommandList_struct *c = NULL;
	__u32 busaddr_masked, tag_masked;

3524
	tag_masked = cciss_tag_discard_error_bits(h, raw_tag);
3525
	list_for_each_entry(c, &h->cmpQ, list) {
3526
		busaddr_masked = cciss_tag_discard_error_bits(h, c->busaddr);
M
Mike Miller 已提交
3527 3528
		if (busaddr_masked == tag_masked) {
			finish_cmd(h, c, raw_tag);
3529
			return next_command(h);
M
Mike Miller 已提交
3530 3531 3532
		}
	}
	bad_tag(h, h->nr_cmds + 1, raw_tag);
3533
	return next_command(h);
M
Mike Miller 已提交
3534 3535
}

3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592
/* 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 cciss_xxx_discard_completions
 * functions.
 */
static int ignore_bogus_interrupt(ctlr_info_t *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;
}

static irqreturn_t cciss_intx_discard_completions(int irq, void *dev_id)
{
	ctlr_info_t *h = dev_id;
	unsigned long flags;
	u32 raw_tag;

	if (ignore_bogus_interrupt(h))
		return IRQ_NONE;

	if (interrupt_not_for_us(h))
		return IRQ_NONE;
	spin_lock_irqsave(&h->lock, flags);
	while (interrupt_pending(h)) {
		raw_tag = get_next_completion(h);
		while (raw_tag != FIFO_EMPTY)
			raw_tag = next_command(h);
	}
	spin_unlock_irqrestore(&h->lock, flags);
	return IRQ_HANDLED;
}

static irqreturn_t cciss_msix_discard_completions(int irq, void *dev_id)
{
	ctlr_info_t *h = dev_id;
	unsigned long flags;
	u32 raw_tag;

	if (ignore_bogus_interrupt(h))
		return IRQ_NONE;

	spin_lock_irqsave(&h->lock, flags);
	raw_tag = get_next_completion(h);
	while (raw_tag != FIFO_EMPTY)
		raw_tag = next_command(h);
	spin_unlock_irqrestore(&h->lock, flags);
	return IRQ_HANDLED;
}

M
Mike Miller 已提交
3593 3594 3595
static irqreturn_t do_cciss_intx(int irq, void *dev_id)
{
	ctlr_info_t *h = dev_id;
L
Linus Torvalds 已提交
3596
	unsigned long flags;
M
Mike Miller 已提交
3597
	u32 raw_tag;
L
Linus Torvalds 已提交
3598

3599
	if (interrupt_not_for_us(h))
L
Linus Torvalds 已提交
3600
		return IRQ_NONE;
3601
	spin_lock_irqsave(&h->lock, flags);
3602
	while (interrupt_pending(h)) {
M
Mike Miller 已提交
3603 3604 3605 3606 3607 3608
		raw_tag = get_next_completion(h);
		while (raw_tag != FIFO_EMPTY) {
			if (cciss_tag_contains_index(raw_tag))
				raw_tag = process_indexed_cmd(h, raw_tag);
			else
				raw_tag = process_nonindexed_cmd(h, raw_tag);
L
Linus Torvalds 已提交
3609 3610
		}
	}
3611
	spin_unlock_irqrestore(&h->lock, flags);
M
Mike Miller 已提交
3612 3613
	return IRQ_HANDLED;
}
L
Linus Torvalds 已提交
3614

M
Mike Miller 已提交
3615 3616 3617 3618 3619 3620 3621 3622
/* Add a second interrupt handler for MSI/MSI-X mode. In this mode we never
 * check the interrupt pending register because it is not set.
 */
static irqreturn_t do_cciss_msix_intr(int irq, void *dev_id)
{
	ctlr_info_t *h = dev_id;
	unsigned long flags;
	u32 raw_tag;
3623

3624
	spin_lock_irqsave(&h->lock, flags);
M
Mike Miller 已提交
3625 3626 3627 3628 3629 3630
	raw_tag = get_next_completion(h);
	while (raw_tag != FIFO_EMPTY) {
		if (cciss_tag_contains_index(raw_tag))
			raw_tag = process_indexed_cmd(h, raw_tag);
		else
			raw_tag = process_nonindexed_cmd(h, raw_tag);
L
Linus Torvalds 已提交
3631
	}
3632
	spin_unlock_irqrestore(&h->lock, flags);
L
Linus Torvalds 已提交
3633 3634
	return IRQ_HANDLED;
}
3635

3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680
/**
 * add_to_scan_list() - add controller to rescan queue
 * @h:		      Pointer to the controller.
 *
 * Adds the controller to the rescan queue if not already on the queue.
 *
 * returns 1 if added to the queue, 0 if skipped (could be on the
 * queue already, or the controller could be initializing or shutting
 * down).
 **/
static int add_to_scan_list(struct ctlr_info *h)
{
	struct ctlr_info *test_h;
	int found = 0;
	int ret = 0;

	if (h->busy_initializing)
		return 0;

	if (!mutex_trylock(&h->busy_shutting_down))
		return 0;

	mutex_lock(&scan_mutex);
	list_for_each_entry(test_h, &scan_q, scan_list) {
		if (test_h == h) {
			found = 1;
			break;
		}
	}
	if (!found && !h->busy_scanning) {
		INIT_COMPLETION(h->scan_wait);
		list_add_tail(&h->scan_list, &scan_q);
		ret = 1;
	}
	mutex_unlock(&scan_mutex);
	mutex_unlock(&h->busy_shutting_down);

	return ret;
}

/**
 * remove_from_scan_list() - remove controller from rescan queue
 * @h:			   Pointer to the controller.
 *
 * Removes the controller from the rescan queue if present. Blocks if
3681 3682 3683 3684 3685 3686 3687
 * the controller is currently conducting a rescan.  The controller
 * can be in one of three states:
 * 1. Doesn't need a scan
 * 2. On the scan list, but not scanning yet (we remove it)
 * 3. Busy scanning (and not on the list). In this case we want to wait for
 *    the scan to complete to make sure the scanning thread for this
 *    controller is completely idle.
3688 3689 3690 3691 3692 3693 3694
 **/
static void remove_from_scan_list(struct ctlr_info *h)
{
	struct ctlr_info *test_h, *tmp_h;

	mutex_lock(&scan_mutex);
	list_for_each_entry_safe(test_h, tmp_h, &scan_q, scan_list) {
3695
		if (test_h == h) { /* state 2. */
3696 3697 3698 3699 3700 3701
			list_del(&h->scan_list);
			complete_all(&h->scan_wait);
			mutex_unlock(&scan_mutex);
			return;
		}
	}
3702 3703
	if (h->busy_scanning) { /* state 3. */
		mutex_unlock(&scan_mutex);
3704
		wait_for_completion(&h->scan_wait);
3705 3706 3707
	} else { /* state 1, nothing to do. */
		mutex_unlock(&scan_mutex);
	}
3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721
}

/**
 * scan_thread() - kernel thread used to rescan controllers
 * @data:	 Ignored.
 *
 * A kernel thread used scan for drive topology changes on
 * controllers. The thread processes only one controller at a time
 * using a queue.  Controllers are added to the queue using
 * add_to_scan_list() and removed from the queue either after done
 * processing or using remove_from_scan_list().
 *
 * returns 0.
 **/
3722 3723
static int scan_thread(void *data)
{
3724
	struct ctlr_info *h;
3725

3726 3727 3728
	while (1) {
		set_current_state(TASK_INTERRUPTIBLE);
		schedule();
3729 3730
		if (kthread_should_stop())
			break;
3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745

		while (1) {
			mutex_lock(&scan_mutex);
			if (list_empty(&scan_q)) {
				mutex_unlock(&scan_mutex);
				break;
			}

			h = list_entry(scan_q.next,
				       struct ctlr_info,
				       scan_list);
			list_del(&h->scan_list);
			h->busy_scanning = 1;
			mutex_unlock(&scan_mutex);

3746 3747 3748 3749 3750
			rebuild_lun_table(h, 0, 0);
			complete_all(&h->scan_wait);
			mutex_lock(&scan_mutex);
			h->busy_scanning = 0;
			mutex_unlock(&scan_mutex);
3751
		}
3752
	}
3753

3754 3755 3756 3757 3758 3759 3760 3761 3762 3763
	return 0;
}

static int check_for_unit_attention(ctlr_info_t *h, CommandList_struct *c)
{
	if (c->err_info->SenseInfo[2] != UNIT_ATTENTION)
		return 0;

	switch (c->err_info->SenseInfo[12]) {
	case STATE_CHANGED:
3764 3765
		dev_warn(&h->pdev->dev, "a state change "
			"detected, command retried\n");
3766 3767 3768
		return 1;
	break;
	case LUN_FAILED:
3769 3770
		dev_warn(&h->pdev->dev, "LUN failure "
			"detected, action required\n");
3771 3772 3773
		return 1;
	break;
	case REPORT_LUNS_CHANGED:
3774
		dev_warn(&h->pdev->dev, "report LUN data changed\n");
3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790
	/*
	 * Here, we could call add_to_scan_list and wake up the scan thread,
	 * except that it's quite likely that we will get more than one
	 * REPORT_LUNS_CHANGED condition in quick succession, which means
	 * that those which occur after the first one will likely happen
	 * *during* the scan_thread's rescan.  And the rescan code is not
	 * robust enough to restart in the middle, undoing what it has already
	 * done, and it's not clear that it's even possible to do this, since
	 * part of what it does is notify the block layer, which starts
	 * doing it's own i/o to read partition tables and so on, and the
	 * driver doesn't have visibility to know what might need undoing.
	 * In any event, if possible, it is horribly complicated to get right
	 * so we just don't do it for now.
	 *
	 * Note: this REPORT_LUNS_CHANGED condition only occurs on the MSA2012.
	 */
3791 3792 3793
		return 1;
	break;
	case POWER_OR_RESET:
3794 3795
		dev_warn(&h->pdev->dev,
			"a power on or device reset detected\n");
3796 3797 3798
		return 1;
	break;
	case UNIT_ATTENTION_CLEARED:
3799 3800
		dev_warn(&h->pdev->dev,
			"unit attention cleared by another initiator\n");
3801 3802 3803
		return 1;
	break;
	default:
3804 3805
		dev_warn(&h->pdev->dev, "unknown unit attention detected\n");
		return 1;
3806 3807 3808
	}
}

3809
/*
3810
 *  We cannot read the structure directly, for portability we must use
L
Linus Torvalds 已提交
3811
 *   the io functions.
3812
 *   This is for debug only.
L
Linus Torvalds 已提交
3813
 */
3814
static void print_cfg_table(ctlr_info_t *h)
L
Linus Torvalds 已提交
3815 3816 3817
{
	int i;
	char temp_name[17];
3818
	CfgTable_struct *tb = h->cfgtable;
L
Linus Torvalds 已提交
3819

3820 3821
	dev_dbg(&h->pdev->dev, "Controller Configuration information\n");
	dev_dbg(&h->pdev->dev, "------------------------------------\n");
3822
	for (i = 0; i < 4; i++)
L
Linus Torvalds 已提交
3823
		temp_name[i] = readb(&(tb->Signature[i]));
3824
	temp_name[4] = '\0';
3825 3826 3827 3828
	dev_dbg(&h->pdev->dev, "   Signature = %s\n", temp_name);
	dev_dbg(&h->pdev->dev, "   Spec Number = %d\n",
		readl(&(tb->SpecValence)));
	dev_dbg(&h->pdev->dev, "   Transport methods supported = 0x%x\n",
3829
	       readl(&(tb->TransportSupport)));
3830
	dev_dbg(&h->pdev->dev, "   Transport methods active = 0x%x\n",
3831
	       readl(&(tb->TransportActive)));
3832
	dev_dbg(&h->pdev->dev, "   Requested transport Method = 0x%x\n",
3833
	       readl(&(tb->HostWrite.TransportRequest)));
3834
	dev_dbg(&h->pdev->dev, "   Coalesce Interrupt Delay = 0x%x\n",
3835
	       readl(&(tb->HostWrite.CoalIntDelay)));
3836
	dev_dbg(&h->pdev->dev, "   Coalesce Interrupt Count = 0x%x\n",
3837
	       readl(&(tb->HostWrite.CoalIntCount)));
3838
	dev_dbg(&h->pdev->dev, "   Max outstanding commands = 0x%d\n",
3839
	       readl(&(tb->CmdsOutMax)));
3840 3841
	dev_dbg(&h->pdev->dev, "   Bus Types = 0x%x\n",
		readl(&(tb->BusTypes)));
3842
	for (i = 0; i < 16; i++)
L
Linus Torvalds 已提交
3843 3844
		temp_name[i] = readb(&(tb->ServerName[i]));
	temp_name[16] = '\0';
3845 3846 3847
	dev_dbg(&h->pdev->dev, "   Server Name = %s\n", temp_name);
	dev_dbg(&h->pdev->dev, "   Heartbeat Counter = 0x%x\n\n\n",
		readl(&(tb->HeartBeat)));
L
Linus Torvalds 已提交
3848 3849
}

3850
static int find_PCI_BAR_index(struct pci_dev *pdev, unsigned long pci_bar_addr)
L
Linus Torvalds 已提交
3851 3852
{
	int i, offset, mem_type, bar_type;
3853
	if (pci_bar_addr == PCI_BASE_ADDRESS_0)	/* looking for BAR zero? */
L
Linus Torvalds 已提交
3854 3855
		return 0;
	offset = 0;
3856 3857
	for (i = 0; i < DEVICE_COUNT_RESOURCE; i++) {
		bar_type = pci_resource_flags(pdev, i) & PCI_BASE_ADDRESS_SPACE;
L
Linus Torvalds 已提交
3858 3859 3860 3861
		if (bar_type == PCI_BASE_ADDRESS_SPACE_IO)
			offset += 4;
		else {
			mem_type = pci_resource_flags(pdev, i) &
3862
			    PCI_BASE_ADDRESS_MEM_TYPE_MASK;
L
Linus Torvalds 已提交
3863
			switch (mem_type) {
3864 3865 3866 3867 3868 3869 3870 3871
			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 */
3872
				dev_warn(&pdev->dev,
3873 3874
				       "Base address is invalid\n");
				return -1;
L
Linus Torvalds 已提交
3875 3876 3877
				break;
			}
		}
3878 3879
		if (offset == pci_bar_addr - PCI_BASE_ADDRESS_0)
			return i + 1;
L
Linus Torvalds 已提交
3880 3881 3882 3883
	}
	return -1;
}

3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920
/* 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,
	int nsgs, int *bucket_map)
{
	int i, j, b, size;

	/* even a command with 0 SGs requires 4 blocks */
#define MINIMUM_TRANSFER_BLOCKS 4
#define NUM_BUCKETS 8
	/* Note, bucket_map must have nsgs+1 entries. */
	for (i = 0; i <= nsgs; i++) {
		/* Compute size of a command with i SG entries */
		size = i + MINIMUM_TRANSFER_BLOCKS;
		b = num_buckets; /* Assume the biggest bucket */
		/* Find the bucket that is just big enough */
		for (j = 0; j < 8; j++) {
			if (bucket[j] >= size) {
				b = j;
				break;
			}
		}
		/* for a command with i SG entries, use bucket b. */
		bucket_map[i] = b;
	}
}

3921 3922 3923 3924 3925 3926 3927 3928 3929 3930
static void __devinit cciss_wait_for_mode_change_ack(ctlr_info_t *h)
{
	int i;

	/* 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.) */
	for (i = 0; i < MAX_CONFIG_WAIT; i++) {
		if (!(readl(h->vaddr + SA5_DOORBELL) & CFGTBL_ChangeReq))
			break;
3931
		usleep_range(10000, 20000);
3932 3933 3934
	}
}

3935 3936
static __devinit void cciss_enter_performant_mode(ctlr_info_t *h,
	u32 use_short_tags)
3937 3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954
{
	/* 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.
	 * the largest command possible requires MAXSGENTRIES + 4 16-byte
	 * 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.
	 */
3955
	__u32 trans_offset;
3956
	int bft[8] = { 5, 6, 8, 10, 12, 20, 28, MAXSGENTRIES + 4};
3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990
			/*
			 *  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
			 */
	unsigned long register_value;
	BUILD_BUG_ON(28 > MAXSGENTRIES + 4);

	h->reply_pool_wraparound = 1; /* spec: init to 1 */

	/* Controller spec: zero out this buffer. */
	memset(h->reply_pool, 0, h->max_commands * sizeof(__u64));
	h->reply_pool_head = h->reply_pool;

	trans_offset = readl(&(h->cfgtable->TransMethodOffset));
	calc_bucket_map(bft, ARRAY_SIZE(bft), h->maxsgentries,
				h->blockFetchTable);
	writel(bft[0], &h->transtable->BlockFetch0);
	writel(bft[1], &h->transtable->BlockFetch1);
	writel(bft[2], &h->transtable->BlockFetch2);
	writel(bft[3], &h->transtable->BlockFetch3);
	writel(bft[4], &h->transtable->BlockFetch4);
	writel(bft[5], &h->transtable->BlockFetch5);
	writel(bft[6], &h->transtable->BlockFetch6);
	writel(bft[7], &h->transtable->BlockFetch7);

	/* size of controller ring buffer */
	writel(h->max_commands, &h->transtable->RepQSize);
	writel(1, &h->transtable->RepQCount);
	writel(0, &h->transtable->RepQCtrAddrLow32);
	writel(0, &h->transtable->RepQCtrAddrHigh32);
	writel(h->reply_pool_dhandle, &h->transtable->RepQAddr0Low32);
	writel(0, &h->transtable->RepQAddr0High32);
3991
	writel(CFGTBL_Trans_Performant | use_short_tags,
3992 3993 3994
			&(h->cfgtable->HostWrite.TransportRequest));

	writel(CFGTBL_ChangeReq, h->vaddr + SA5_DOORBELL);
3995
	cciss_wait_for_mode_change_ack(h);
3996
	register_value = readl(&(h->cfgtable->TransportActive));
3997
	if (!(register_value & CFGTBL_Trans_Performant))
3998
		dev_warn(&h->pdev->dev, "cciss: unable to get board into"
3999
					" performant mode\n");
4000 4001 4002 4003 4004 4005
}

static void __devinit cciss_put_controller_into_performant_mode(ctlr_info_t *h)
{
	__u32 trans_support;

4006 4007 4008
	if (cciss_simple_mode)
		return;

4009 4010 4011 4012 4013 4014 4015
	dev_dbg(&h->pdev->dev, "Trying to put board into Performant mode\n");
	/* Attempt to put controller into performant mode if supported */
	/* Does board support performant mode? */
	trans_support = readl(&(h->cfgtable->TransportSupport));
	if (!(trans_support & PERFORMANT_MODE))
		return;

4016
	dev_dbg(&h->pdev->dev, "Placing controller into performant mode\n");
4017 4018 4019 4020 4021
	/* Performant mode demands commands on a 32 byte boundary
	 * pci_alloc_consistent aligns on page boundarys already.
	 * Just need to check if divisible by 32
	 */
	if ((sizeof(CommandList_struct) % 32) != 0) {
4022
		dev_warn(&h->pdev->dev, "%s %d %s\n",
4023 4024 4025
			"cciss info: command size[",
			(int)sizeof(CommandList_struct),
			"] not divisible by 32, no performant mode..\n");
4026 4027 4028
		return;
	}

4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040
	/* Performant mode ring buffer and supporting data structures */
	h->reply_pool = (__u64 *)pci_alloc_consistent(
		h->pdev, h->max_commands * sizeof(__u64),
		&(h->reply_pool_dhandle));

	/* Need a block fetch table for performant mode */
	h->blockFetchTable = kmalloc(((h->maxsgentries+1) *
		sizeof(__u32)), GFP_KERNEL);

	if ((h->reply_pool == NULL) || (h->blockFetchTable == NULL))
		goto clean_up;

4041 4042
	cciss_enter_performant_mode(h,
		trans_support & CFGTBL_Trans_use_short_tags);
4043

4044 4045
	/* Change the access methods to the performant access methods */
	h->access = SA5_performant_access;
4046
	h->transMethod = CFGTBL_Trans_Performant;
4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059

	return;
clean_up:
	kfree(h->blockFetchTable);
	if (h->reply_pool)
		pci_free_consistent(h->pdev,
				h->max_commands * sizeof(__u64),
				h->reply_pool,
				h->reply_pool_dhandle);
	return;

} /* cciss_put_controller_into_performant_mode */

4060 4061 4062 4063
/* If MSI/MSI-X is supported by the kernel we will try to enable it on
 * controllers that are capable. If not, we use IO-APIC mode.
 */

4064
static void __devinit cciss_interrupt_mode(ctlr_info_t *h)
4065 4066
{
#ifdef CONFIG_PCI_MSI
4067 4068 4069 4070
	int err;
	struct msix_entry cciss_msix_entries[4] = { {0, 0}, {0, 1},
	{0, 2}, {0, 3}
	};
4071 4072

	/* Some boards advertise MSI but don't really support it */
4073 4074
	if ((h->board_id == 0x40700E11) || (h->board_id == 0x40800E11) ||
	    (h->board_id == 0x40820E11) || (h->board_id == 0x40830E11))
4075 4076
		goto default_int_mode;

4077 4078
	if (pci_find_capability(h->pdev, PCI_CAP_ID_MSIX)) {
		err = pci_enable_msix(h->pdev, cciss_msix_entries, 4);
4079
		if (!err) {
4080 4081 4082 4083 4084
			h->intr[0] = cciss_msix_entries[0].vector;
			h->intr[1] = cciss_msix_entries[1].vector;
			h->intr[2] = cciss_msix_entries[2].vector;
			h->intr[3] = cciss_msix_entries[3].vector;
			h->msix_vector = 1;
4085 4086 4087
			return;
		}
		if (err > 0) {
4088 4089
			dev_warn(&h->pdev->dev,
				"only %d MSI-X vectors available\n", err);
4090
			goto default_int_mode;
4091
		} else {
4092 4093
			dev_warn(&h->pdev->dev,
				"MSI-X init failed %d\n", err);
4094
			goto default_int_mode;
4095 4096
		}
	}
4097 4098 4099 4100
	if (pci_find_capability(h->pdev, PCI_CAP_ID_MSI)) {
		if (!pci_enable_msi(h->pdev))
			h->msi_vector = 1;
		else
4101
			dev_warn(&h->pdev->dev, "MSI init failed\n");
4102
	}
4103
default_int_mode:
4104
#endif				/* CONFIG_PCI_MSI */
4105
	/* if we get here we're going to use the default interrupt mode */
4106
	h->intr[h->intr_mode] = h->pdev->irq;
4107 4108 4109
	return;
}

4110
static int __devinit cciss_lookup_board_id(struct pci_dev *pdev, u32 *board_id)
L
Linus Torvalds 已提交
4111
{
4112 4113
	int i;
	u32 subsystem_vendor_id, subsystem_device_id;
4114 4115 4116

	subsystem_vendor_id = pdev->subsystem_vendor;
	subsystem_device_id = pdev->subsystem_device;
4117 4118
	*board_id = ((subsystem_device_id << 16) & 0xffff0000) |
			subsystem_vendor_id;
4119

4120
	for (i = 0; i < ARRAY_SIZE(products); i++)
4121 4122 4123 4124 4125 4126
		if (*board_id == products[i].board_id)
			return i;
	dev_warn(&pdev->dev, "unrecognized board ID: 0x%08x, ignoring.\n",
		*board_id);
	return -ENODEV;
}
L
Linus Torvalds 已提交
4127

4128 4129 4130
static inline bool cciss_board_disabled(ctlr_info_t *h)
{
	u16 command;
L
Linus Torvalds 已提交
4131

4132 4133 4134
	(void) pci_read_config_word(h->pdev, PCI_COMMAND, &command);
	return ((command & PCI_COMMAND_MEMORY) == 0);
}
L
Linus Torvalds 已提交
4135

4136 4137 4138 4139
static int __devinit cciss_pci_find_memory_BAR(struct pci_dev *pdev,
	unsigned long *memory_bar)
{
	int i;
4140

4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151
	for (i = 0; i < DEVICE_COUNT_RESOURCE; i++)
		if (pci_resource_flags(pdev, i) & IORESOURCE_MEM) {
			/* addressing mode bits already removed */
			*memory_bar = pci_resource_start(pdev, i);
			dev_dbg(&pdev->dev, "memory BAR = %lx\n",
				*memory_bar);
			return 0;
		}
	dev_warn(&pdev->dev, "no memory BAR found\n");
	return -ENODEV;
}
L
Linus Torvalds 已提交
4152

4153 4154 4155 4156
static int __devinit cciss_wait_for_board_state(struct pci_dev *pdev,
	void __iomem *vaddr, int wait_for_ready)
#define BOARD_READY 1
#define BOARD_NOT_READY 0
4157
{
4158
	int i, iterations;
4159
	u32 scratchpad;
L
Linus Torvalds 已提交
4160

4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174
	if (wait_for_ready)
		iterations = CCISS_BOARD_READY_ITERATIONS;
	else
		iterations = CCISS_BOARD_NOT_READY_ITERATIONS;

	for (i = 0; i < iterations; i++) {
		scratchpad = readl(vaddr + SA5_SCRATCHPAD_OFFSET);
		if (wait_for_ready) {
			if (scratchpad == CCISS_FIRMWARE_READY)
				return 0;
		} else {
			if (scratchpad != CCISS_FIRMWARE_READY)
				return 0;
		}
4175
		msleep(CCISS_BOARD_READY_POLL_INTERVAL_MSECS);
4176
	}
4177
	dev_warn(&pdev->dev, "board not ready, timed out.\n");
4178 4179
	return -ENODEV;
}
4180

4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195
static int __devinit cciss_find_cfg_addrs(struct pci_dev *pdev,
	void __iomem *vaddr, u32 *cfg_base_addr, u64 *cfg_base_addr_index,
	u64 *cfg_offset)
{
	*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, "
			"*cfg_base_addr = 0x%08x\n", *cfg_base_addr);
		return -ENODEV;
	}
	return 0;
}
L
Linus Torvalds 已提交
4196

4197 4198 4199 4200 4201 4202
static int __devinit cciss_find_cfgtables(ctlr_info_t *h)
{
	u64 cfg_offset;
	u32 cfg_base_addr;
	u64 cfg_base_addr_index;
	u32 trans_offset;
4203
	int rc;
L
Linus Torvalds 已提交
4204

4205 4206 4207 4208
	rc = cciss_find_cfg_addrs(h->pdev, h->vaddr, &cfg_base_addr,
		&cfg_base_addr_index, &cfg_offset);
	if (rc)
		return rc;
4209
	h->cfgtable = remap_pci_mem(pci_resource_start(h->pdev,
4210
		cfg_base_addr_index) + cfg_offset, sizeof(h->cfgtable));
4211 4212
	if (!h->cfgtable)
		return -ENOMEM;
4213 4214 4215
	rc = write_driver_ver_to_cfgtable(h->cfgtable);
	if (rc)
		return rc;
4216
	/* Find performant mode table. */
4217
	trans_offset = readl(&h->cfgtable->TransMethodOffset);
4218 4219 4220 4221 4222 4223 4224
	h->transtable = remap_pci_mem(pci_resource_start(h->pdev,
				cfg_base_addr_index)+cfg_offset+trans_offset,
				sizeof(*h->transtable));
	if (!h->transtable)
		return -ENOMEM;
	return 0;
}
L
Linus Torvalds 已提交
4225

4226 4227 4228
static void __devinit cciss_get_max_perf_mode_cmds(struct ctlr_info *h)
{
	h->max_commands = readl(&(h->cfgtable->MaxPerformantModeCommands));
4229 4230 4231 4232 4233

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

4234 4235 4236 4237 4238 4239
	if (h->max_commands < 16) {
		dev_warn(&h->pdev->dev, "Controller reports "
			"max supported commands of %d, an obvious lie. "
			"Using 16.  Ensure that firmware is up to date.\n",
			h->max_commands);
		h->max_commands = 16;
L
Linus Torvalds 已提交
4240
	}
4241
}
L
Linus Torvalds 已提交
4242

4243 4244 4245 4246 4247 4248
/* Interrogate the hardware for some limits:
 * max commands, max SG elements without chaining, and with chaining,
 * SG chain block size, etc.
 */
static void __devinit cciss_find_board_params(ctlr_info_t *h)
{
4249
	cciss_get_max_perf_mode_cmds(h);
4250
	h->nr_cmds = h->max_commands - 4 - cciss_tape_cmds;
4251
	h->maxsgentries = readl(&(h->cfgtable->MaxSGElements));
4252
	/*
4253
	 * Limit in-command s/g elements to 32 save dma'able memory.
4254 4255
	 * Howvever spec says if 0, use 31
	 */
4256 4257 4258 4259 4260
	h->max_cmd_sgentries = 31;
	if (h->maxsgentries > 512) {
		h->max_cmd_sgentries = 32;
		h->chainsize = h->maxsgentries - h->max_cmd_sgentries + 1;
		h->maxsgentries--; /* save one for chain pointer */
4261
	} else {
4262 4263
		h->maxsgentries = 31; /* default to traditional values */
		h->chainsize = 0;
4264
	}
4265
}
4266

4267 4268 4269 4270 4271 4272 4273 4274
static inline bool CISS_signature_present(ctlr_info_t *h)
{
	if ((readb(&h->cfgtable->Signature[0]) != 'C') ||
	    (readb(&h->cfgtable->Signature[1]) != 'I') ||
	    (readb(&h->cfgtable->Signature[2]) != 'S') ||
	    (readb(&h->cfgtable->Signature[3]) != 'S')) {
		dev_warn(&h->pdev->dev, "not a valid CISS config table\n");
		return false;
L
Linus Torvalds 已提交
4275
	}
4276 4277 4278
	return true;
}

4279 4280 4281
/* Need to enable prefetch in the SCSI core for 6400 in x86 */
static inline void cciss_enable_scsi_prefetch(ctlr_info_t *h)
{
L
Linus Torvalds 已提交
4282
#ifdef CONFIG_X86
4283 4284 4285 4286 4287
	u32 prefetch;

	prefetch = readl(&(h->cfgtable->SCSI_Prefetch));
	prefetch |= 0x100;
	writel(prefetch, &(h->cfgtable->SCSI_Prefetch));
L
Linus Torvalds 已提交
4288
#endif
4289
}
L
Linus Torvalds 已提交
4290

4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308
/* Disable DMA prefetch for the P600.  Otherwise an ASIC bug may result
 * in a prefetch beyond physical memory.
 */
static inline void cciss_p600_dma_prefetch_quirk(ctlr_info_t *h)
{
	u32 dma_prefetch;
	__u32 dma_refetch;

	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);
	pci_read_config_dword(h->pdev, PCI_COMMAND_PARITY, &dma_refetch);
	dma_refetch |= 0x1;
	pci_write_config_dword(h->pdev, PCI_COMMAND_PARITY, dma_refetch);
}

4309
static int __devinit cciss_pci_init(ctlr_info_t *h)
4310
{
4311
	int prod_index, err;
4312

4313
	prod_index = cciss_lookup_board_id(h->pdev, &h->board_id);
4314
	if (prod_index < 0)
4315
		return -ENODEV;
4316 4317
	h->product_name = products[prod_index].product_name;
	h->access = *(products[prod_index].access);
L
Linus Torvalds 已提交
4318

4319
	if (cciss_board_disabled(h)) {
4320
		dev_warn(&h->pdev->dev, "controller appears to be disabled\n");
4321
		return -ENODEV;
L
Linus Torvalds 已提交
4322
	}
4323 4324 4325 4326

	pci_disable_link_state(h->pdev, PCIE_LINK_STATE_L0S |
				PCIE_LINK_STATE_L1 | PCIE_LINK_STATE_CLKPM);

4327
	err = pci_enable_device(h->pdev);
4328
	if (err) {
4329
		dev_warn(&h->pdev->dev, "Unable to Enable PCI device\n");
4330
		return err;
4331 4332
	}

4333
	err = pci_request_regions(h->pdev, "cciss");
4334
	if (err) {
4335 4336
		dev_warn(&h->pdev->dev,
			"Cannot obtain PCI resources, aborting\n");
4337
		return err;
4338
	}
L
Linus Torvalds 已提交
4339

4340 4341
	dev_dbg(&h->pdev->dev, "irq = %x\n", h->pdev->irq);
	dev_dbg(&h->pdev->dev, "board_id = %x\n", h->board_id);
L
Linus Torvalds 已提交
4342

4343 4344 4345
/* If the kernel supports MSI/MSI-X we will try to enable that functionality,
 * else we use the IO-APIC interrupt assigned to us by system ROM.
 */
4346 4347
	cciss_interrupt_mode(h);
	err = cciss_pci_find_memory_BAR(h->pdev, &h->paddr);
4348
	if (err)
4349
		goto err_out_free_res;
4350 4351
	h->vaddr = remap_pci_mem(h->paddr, 0x250);
	if (!h->vaddr) {
4352 4353
		err = -ENOMEM;
		goto err_out_free_res;
4354
	}
4355
	err = cciss_wait_for_board_state(h->pdev, h->vaddr, BOARD_READY);
4356
	if (err)
4357
		goto err_out_free_res;
4358
	err = cciss_find_cfgtables(h);
4359
	if (err)
4360
		goto err_out_free_res;
4361
	print_cfg_table(h);
4362
	cciss_find_board_params(h);
L
Linus Torvalds 已提交
4363

4364
	if (!CISS_signature_present(h)) {
4365
		err = -ENODEV;
4366
		goto err_out_free_res;
L
Linus Torvalds 已提交
4367
	}
4368 4369
	cciss_enable_scsi_prefetch(h);
	cciss_p600_dma_prefetch_quirk(h);
4370 4371 4372
	err = cciss_enter_simple_mode(h);
	if (err)
		goto err_out_free_res;
4373
	cciss_put_controller_into_performant_mode(h);
L
Linus Torvalds 已提交
4374 4375
	return 0;

4376
err_out_free_res:
4377 4378 4379 4380
	/*
	 * Deliberately omit pci_disable_device(): it does something nasty to
	 * Smart Array controllers that pci_enable_device does not undo
	 */
4381 4382 4383 4384 4385 4386 4387
	if (h->transtable)
		iounmap(h->transtable);
	if (h->cfgtable)
		iounmap(h->cfgtable);
	if (h->vaddr)
		iounmap(h->vaddr);
	pci_release_regions(h->pdev);
4388
	return err;
L
Linus Torvalds 已提交
4389 4390
}

M
Mike Miller 已提交
4391 4392
/* Function to find the first free pointer into our hba[] array
 * Returns -1 if no free entries are left.
4393
 */
4394
static int alloc_cciss_hba(struct pci_dev *pdev)
L
Linus Torvalds 已提交
4395
{
4396
	int i;
L
Linus Torvalds 已提交
4397

4398
	for (i = 0; i < MAX_CTLR; i++) {
L
Linus Torvalds 已提交
4399
		if (!hba[i]) {
4400
			ctlr_info_t *h;
J
Jesper Juhl 已提交
4401

4402 4403
			h = kzalloc(sizeof(ctlr_info_t), GFP_KERNEL);
			if (!h)
L
Linus Torvalds 已提交
4404
				goto Enomem;
4405
			hba[i] = h;
L
Linus Torvalds 已提交
4406 4407 4408
			return i;
		}
	}
4409
	dev_warn(&pdev->dev, "This driver supports a maximum"
4410
	       " of %d controllers.\n", MAX_CTLR);
4411 4412
	return -1;
Enomem:
4413
	dev_warn(&pdev->dev, "out of memory.\n");
L
Linus Torvalds 已提交
4414 4415 4416
	return -1;
}

4417
static void free_hba(ctlr_info_t *h)
L
Linus Torvalds 已提交
4418
{
4419
	int i;
L
Linus Torvalds 已提交
4420

4421
	hba[h->ctlr] = NULL;
4422 4423 4424 4425
	for (i = 0; i < h->highest_lun + 1; i++)
		if (h->gendisk[i] != NULL)
			put_disk(h->gendisk[i]);
	kfree(h);
L
Linus Torvalds 已提交
4426 4427
}

4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449
/* Send a message CDB to the firmware. */
static __devinit int cciss_message(struct pci_dev *pdev, unsigned char opcode, unsigned char type)
{
	typedef struct {
		CommandListHeader_struct CommandHeader;
		RequestBlock_struct Request;
		ErrDescriptor_struct ErrorDescriptor;
	} Command;
	static const size_t cmd_sz = sizeof(Command) + sizeof(ErrorInfo_struct);
	Command *cmd;
	dma_addr_t paddr64;
	uint32_t paddr32, tag;
	void __iomem *vaddr;
	int i, err;

	vaddr = ioremap_nocache(pci_resource_start(pdev, 0), pci_resource_len(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. */
4450
	err = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(32));
4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492
	if (err) {
		iounmap(vaddr);
		return -ENOMEM;
	}

	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). */
	paddr32 = paddr64;

	cmd->CommandHeader.ReplyQueue = 0;
	cmd->CommandHeader.SGList = 0;
	cmd->CommandHeader.SGTotal = 0;
	cmd->CommandHeader.Tag.lower = paddr32;
	cmd->CommandHeader.Tag.upper = 0;
	memset(&cmd->CommandHeader.LUN.LunAddrBytes, 0, 8);

	cmd->Request.CDBLen = 16;
	cmd->Request.Type.Type = TYPE_MSG;
	cmd->Request.Type.Attribute = ATTR_HEADOFQUEUE;
	cmd->Request.Type.Direction = XFER_NONE;
	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); /* the rest of the CDB is reserved */

	cmd->ErrorDescriptor.Addr.lower = paddr32 + sizeof(Command);
	cmd->ErrorDescriptor.Addr.upper = 0;
	cmd->ErrorDescriptor.Len = sizeof(ErrorInfo_struct);

	writel(paddr32, vaddr + SA5_REQUEST_PORT_OFFSET);

	for (i = 0; i < 10; i++) {
		tag = readl(vaddr + SA5_REPLY_PORT_OFFSET);
		if ((tag & ~3) == paddr32)
			break;
4493
		msleep(CCISS_POST_RESET_NOOP_TIMEOUT_MSECS);
4494 4495 4496 4497 4498 4499 4500
	}

	iounmap(vaddr);

	/* we leak the DMA buffer here ... no choice since the controller could
	   still complete the command. */
	if (i == 10) {
4501 4502
		dev_err(&pdev->dev,
			"controller message %02x:%02x timed out\n",
4503 4504 4505 4506 4507 4508 4509
			opcode, type);
		return -ETIMEDOUT;
	}

	pci_free_consistent(pdev, cmd_sz, cmd, paddr64);

	if (tag & 2) {
4510
		dev_err(&pdev->dev, "controller message %02x:%02x failed\n",
4511 4512 4513 4514
			opcode, type);
		return -EIO;
	}

4515
	dev_info(&pdev->dev, "controller message %02x:%02x succeeded\n",
4516 4517 4518 4519 4520 4521
		opcode, type);
	return 0;
}

#define cciss_noop(p) cciss_message(p, 3, 0)

S
Stephen M. Cameron 已提交
4522
static int cciss_controller_hard_reset(struct pci_dev *pdev,
4523
	void * __iomem vaddr, u32 use_doorbell)
4524
{
S
Stephen M. Cameron 已提交
4525 4526
	u16 pmcsr;
	int pos;
4527

S
Stephen M. Cameron 已提交
4528 4529 4530 4531 4532 4533
	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");
4534
		writel(use_doorbell, vaddr + SA5_DOORBELL);
S
Stephen M. Cameron 已提交
4535 4536 4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557
	} 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." */

		pos = pci_find_capability(pdev, PCI_CAP_ID_PM);
		if (pos == 0) {
			dev_err(&pdev->dev,
				"cciss_controller_hard_reset: "
				"PCI PM not supported\n");
			return -ENODEV;
		}
		dev_info(&pdev->dev, "using PCI PM to reset controller\n");
		/* enter the D3hot power management state */
		pci_read_config_word(pdev, pos + PCI_PM_CTRL, &pmcsr);
		pmcsr &= ~PCI_PM_CTRL_STATE_MASK;
		pmcsr |= PCI_D3hot;
		pci_write_config_word(pdev, pos + PCI_PM_CTRL, pmcsr);
4558

S
Stephen M. Cameron 已提交
4559
		msleep(500);
4560

S
Stephen M. Cameron 已提交
4561 4562 4563 4564
		/* enter the D0 power management state */
		pmcsr &= ~PCI_PM_CTRL_STATE_MASK;
		pmcsr |= PCI_D0;
		pci_write_config_word(pdev, pos + PCI_PM_CTRL, pmcsr);
4565 4566 4567 4568 4569 4570 4571

		/*
		 * 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);
S
Stephen M. Cameron 已提交
4572 4573 4574
	}
	return 0;
}
4575

4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629
static __devinit void init_driver_version(char *driver_version, int len)
{
	memset(driver_version, 0, len);
	strncpy(driver_version, "cciss " DRIVER_NAME, len - 1);
}

static __devinit int write_driver_ver_to_cfgtable(
	CfgTable_struct __iomem *cfgtable)
{
	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;
}

static __devinit void read_driver_ver_from_cfgtable(
	CfgTable_struct __iomem *cfgtable, unsigned char *driver_ver)
{
	int i;

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

static __devinit int controller_reset_failed(
	CfgTable_struct __iomem *cfgtable)
{

	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;
}

S
Stephen M. Cameron 已提交
4630 4631 4632 4633 4634 4635 4636 4637 4638
/* This does a hard reset of the controller using PCI power management
 * states or using the doorbell register. */
static __devinit int cciss_kdump_hard_reset_controller(struct pci_dev *pdev)
{
	u64 cfg_offset;
	u32 cfg_base_addr;
	u64 cfg_base_addr_index;
	void __iomem *vaddr;
	unsigned long paddr;
4639
	u32 misc_fw_support;
4640
	int rc;
S
Stephen M. Cameron 已提交
4641
	CfgTable_struct __iomem *cfgtable;
4642
	u32 use_doorbell;
4643
	u32 board_id;
4644
	u16 command_register;
S
Stephen M. Cameron 已提交
4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657

	/* For controllers as old a the p600, this is very nearly
	 * the same thing as
	 *
	 * 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);
	 *
	 * 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.
	 */
4658

4659 4660 4661 4662 4663 4664 4665
	/* 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.
	 */
	cciss_lookup_board_id(pdev, &board_id);
4666
	if (!ctlr_is_resettable(board_id)) {
4667 4668
		dev_warn(&pdev->dev, "Cannot reset Smart Array 640x "
				"due to shared cache module.");
4669 4670 4671
		return -ENODEV;
	}

4672 4673 4674 4675
	/* if controller is soft- but not hard resettable... */
	if (!ctlr_is_hard_resettable(board_id))
		return -ENOTSUPP; /* try soft reset later. */

4676 4677 4678 4679 4680 4681 4682
	/* Save the PCI command register */
	pci_read_config_word(pdev, 4, &command_register);
	/* Turn the board off.  This is so that later pci_restore_state()
	 * won't turn the board on before the rest of config space is ready.
	 */
	pci_disable_device(pdev);
	pci_save_state(pdev);
4683

S
Stephen M. Cameron 已提交
4684 4685 4686 4687 4688 4689 4690
	/* find the first memory BAR, so we can find the cfg table */
	rc = cciss_pci_find_memory_BAR(pdev, &paddr);
	if (rc)
		return rc;
	vaddr = remap_pci_mem(paddr, 0x250);
	if (!vaddr)
		return -ENOMEM;
4691

S
Stephen M. Cameron 已提交
4692 4693 4694 4695 4696 4697 4698 4699 4700 4701 4702
	/* find cfgtable in order to check if reset via doorbell is supported */
	rc = cciss_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;
	}
4703 4704 4705
	rc = write_driver_ver_to_cfgtable(cfgtable);
	if (rc)
		goto unmap_vaddr;
4706

4707 4708
	/* If reset via doorbell register is supported, use that.
	 * There are two such methods.  Favor the newest method.
4709
	 */
4710 4711 4712 4713 4714 4715
	misc_fw_support = readl(&cfgtable->misc_fw_support);
	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;
4716 4717 4718 4719 4720 4721 4722 4723
		if (use_doorbell) {
			dev_warn(&pdev->dev, "Controller claims that "
				"'Bit 2 doorbell reset' is "
				"supported, but not 'bit 5 doorbell reset'.  "
				"Firmware update is recommended.\n");
			rc = -ENOTSUPP; /* use the soft reset */
			goto unmap_cfgtable;
		}
4724
	}
4725

S
Stephen M. Cameron 已提交
4726 4727 4728
	rc = cciss_controller_hard_reset(pdev, vaddr, use_doorbell);
	if (rc)
		goto unmap_cfgtable;
4729 4730 4731 4732 4733
	pci_restore_state(pdev);
	rc = pci_enable_device(pdev);
	if (rc) {
		dev_warn(&pdev->dev, "failed to enable device.\n");
		goto unmap_cfgtable;
4734
	}
4735
	pci_write_config_word(pdev, 4, command_register);
4736

S
Stephen M. Cameron 已提交
4737 4738 4739 4740
	/* Some devices (notably the HP Smart Array 5i Controller)
	   need a little pause here */
	msleep(CCISS_POST_RESET_PAUSE_MSECS);

4741
	/* Wait for board to become not ready, then ready. */
4742
	dev_info(&pdev->dev, "Waiting for board to reset.\n");
4743
	rc = cciss_wait_for_board_state(pdev, vaddr, BOARD_NOT_READY);
4744 4745 4746 4747 4748 4749
	if (rc) {
		dev_warn(&pdev->dev, "Failed waiting for board to hard reset."
				"  Will try soft reset.\n");
		rc = -ENOTSUPP; /* Not expected, but try soft reset later */
		goto unmap_cfgtable;
	}
4750 4751 4752
	rc = cciss_wait_for_board_state(pdev, vaddr, BOARD_READY);
	if (rc) {
		dev_warn(&pdev->dev,
4753 4754
			"failed waiting for board to become ready "
			"after hard reset\n");
4755 4756 4757
		goto unmap_cfgtable;
	}

4758 4759 4760 4761
	rc = controller_reset_failed(vaddr);
	if (rc < 0)
		goto unmap_cfgtable;
	if (rc) {
4762 4763 4764
		dev_warn(&pdev->dev, "Unable to successfully hard reset "
			"controller. Will try soft reset.\n");
		rc = -ENOTSUPP; /* Not expected, but try soft reset later */
4765
	} else {
4766
		dev_info(&pdev->dev, "Board ready after hard reset.\n");
S
Stephen M. Cameron 已提交
4767 4768 4769 4770 4771 4772 4773 4774
	}

unmap_cfgtable:
	iounmap(cfgtable);

unmap_vaddr:
	iounmap(vaddr);
	return rc;
4775 4776
}

4777 4778
static __devinit int cciss_init_reset_devices(struct pci_dev *pdev)
{
S
Stephen M. Cameron 已提交
4779
	int rc, i;
4780 4781 4782 4783

	if (!reset_devices)
		return 0;

S
Stephen M. Cameron 已提交
4784 4785
	/* Reset the controller with a PCI power-cycle or via doorbell */
	rc = cciss_kdump_hard_reset_controller(pdev);
4786

S
Stephen M. Cameron 已提交
4787 4788
	/* -ENOTSUPP here means we cannot reset the controller
	 * but it's already (and still) up and running in
4789 4790
	 * "performant mode".  Or, it might be 640x, which can't reset
	 * due to concerns about shared bbwc between 6402/6404 pair.
S
Stephen M. Cameron 已提交
4791 4792
	 */
	if (rc == -ENOTSUPP)
4793
		return rc; /* just try to do the kdump anyhow. */
S
Stephen M. Cameron 已提交
4794 4795
	if (rc)
		return -ENODEV;
4796 4797

	/* Now try to get the controller to respond to a no-op */
4798
	dev_warn(&pdev->dev, "Waiting for controller to respond to no-op\n");
4799 4800 4801 4802 4803 4804 4805 4806 4807
	for (i = 0; i < CCISS_POST_RESET_NOOP_RETRIES; i++) {
		if (cciss_noop(pdev) == 0)
			break;
		else
			dev_warn(&pdev->dev, "no-op failed%s\n",
				(i < CCISS_POST_RESET_NOOP_RETRIES - 1 ?
					"; re-trying" : ""));
		msleep(CCISS_POST_RESET_NOOP_INTERVAL_MSECS);
	}
4808 4809 4810
	return 0;
}

4811 4812 4813 4814 4815 4816 4817 4818 4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830
static __devinit int cciss_allocate_cmd_pool(ctlr_info_t *h)
{
	h->cmd_pool_bits = kmalloc(
		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(CommandList_struct),
		&(h->cmd_pool_dhandle));
	h->errinfo_pool = pci_alloc_consistent(h->pdev,
		h->nr_cmds * sizeof(ErrorInfo_struct),
		&(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");
		return -ENOMEM;
	}
	return 0;
}

4831 4832 4833 4834 4835 4836 4837 4838 4839 4840 4841 4842 4843 4844 4845 4846 4847 4848 4849 4850 4851 4852 4853 4854 4855 4856 4857 4858 4859 4860 4861 4862 4863
static __devinit int cciss_allocate_scatterlists(ctlr_info_t *h)
{
	int i;

	/* zero it, so that on free we need not know how many were alloc'ed */
	h->scatter_list = kzalloc(h->max_commands *
				sizeof(struct scatterlist *), GFP_KERNEL);
	if (!h->scatter_list)
		return -ENOMEM;

	for (i = 0; i < h->nr_cmds; i++) {
		h->scatter_list[i] = kmalloc(sizeof(struct scatterlist) *
						h->maxsgentries, GFP_KERNEL);
		if (h->scatter_list[i] == NULL) {
			dev_err(&h->pdev->dev, "could not allocate "
				"s/g lists\n");
			return -ENOMEM;
		}
	}
	return 0;
}

static void cciss_free_scatterlists(ctlr_info_t *h)
{
	int i;

	if (h->scatter_list) {
		for (i = 0; i < h->nr_cmds; i++)
			kfree(h->scatter_list[i]);
		kfree(h->scatter_list);
	}
}

4864 4865 4866 4867 4868 4869 4870 4871 4872 4873 4874 4875 4876
static void cciss_free_cmd_pool(ctlr_info_t *h)
{
	kfree(h->cmd_pool_bits);
	if (h->cmd_pool)
		pci_free_consistent(h->pdev,
			h->nr_cmds * sizeof(CommandList_struct),
			h->cmd_pool, h->cmd_pool_dhandle);
	if (h->errinfo_pool)
		pci_free_consistent(h->pdev,
			h->nr_cmds * sizeof(ErrorInfo_struct),
			h->errinfo_pool, h->errinfo_pool_dhandle);
}

4877 4878 4879 4880 4881
static int cciss_request_irq(ctlr_info_t *h,
	irqreturn_t (*msixhandler)(int, void *),
	irqreturn_t (*intxhandler)(int, void *))
{
	if (h->msix_vector || h->msi_vector) {
4882
		if (!request_irq(h->intr[h->intr_mode], msixhandler,
4883 4884 4885
				IRQF_DISABLED, h->devname, h))
			return 0;
		dev_err(&h->pdev->dev, "Unable to get msi irq %d"
4886
			" for %s\n", h->intr[h->intr_mode],
4887 4888 4889 4890
			h->devname);
		return -1;
	}

4891
	if (!request_irq(h->intr[h->intr_mode], intxhandler,
4892 4893 4894
			IRQF_DISABLED, h->devname, h))
		return 0;
	dev_err(&h->pdev->dev, "Unable to get irq %d for %s\n",
4895
		h->intr[h->intr_mode], h->devname);
4896 4897 4898
	return -1;
}

4899 4900 4901 4902 4903 4904 4905 4906 4907 4908 4909 4910 4911 4912 4913 4914 4915 4916 4917 4918 4919 4920 4921 4922 4923 4924 4925
static int __devinit cciss_kdump_soft_reset(ctlr_info_t *h)
{
	if (cciss_send_reset(h, CTLR_LUNID, CCISS_RESET_TYPE_CONTROLLER)) {
		dev_warn(&h->pdev->dev, "Resetting array controller failed.\n");
		return -EIO;
	}

	dev_info(&h->pdev->dev, "Waiting for board to soft reset.\n");
	if (cciss_wait_for_board_state(h->pdev, h->vaddr, BOARD_NOT_READY)) {
		dev_warn(&h->pdev->dev, "Soft reset had no effect.\n");
		return -1;
	}

	dev_info(&h->pdev->dev, "Board reset, awaiting READY status.\n");
	if (cciss_wait_for_board_state(h->pdev, h->vaddr, BOARD_READY)) {
		dev_warn(&h->pdev->dev, "Board failed to become ready "
			"after soft reset.\n");
		return -1;
	}

	return 0;
}

static void cciss_undo_allocations_after_kdump_soft_reset(ctlr_info_t *h)
{
	int ctlr = h->ctlr;

4926
	free_irq(h->intr[h->intr_mode], h);
4927 4928 4929 4930 4931 4932 4933 4934 4935 4936 4937 4938 4939 4940 4941 4942 4943 4944 4945 4946 4947 4948 4949 4950 4951 4952
#ifdef CONFIG_PCI_MSI
	if (h->msix_vector)
		pci_disable_msix(h->pdev);
	else if (h->msi_vector)
		pci_disable_msi(h->pdev);
#endif /* CONFIG_PCI_MSI */
	cciss_free_sg_chain_blocks(h->cmd_sg_list, h->nr_cmds);
	cciss_free_scatterlists(h);
	cciss_free_cmd_pool(h);
	kfree(h->blockFetchTable);
	if (h->reply_pool)
		pci_free_consistent(h->pdev, h->max_commands * sizeof(__u64),
				h->reply_pool, h->reply_pool_dhandle);
	if (h->transtable)
		iounmap(h->transtable);
	if (h->cfgtable)
		iounmap(h->cfgtable);
	if (h->vaddr)
		iounmap(h->vaddr);
	unregister_blkdev(h->major, h->devname);
	cciss_destroy_hba_sysfs_entry(h);
	pci_release_regions(h->pdev);
	kfree(h);
	hba[ctlr] = NULL;
}

L
Linus Torvalds 已提交
4953 4954 4955 4956 4957 4958
/*
 *  This is it.  Find all the controllers and register them.  I really hate
 *  stealing all these major device numbers.
 *  returns the number of block devices registered.
 */
static int __devinit cciss_init_one(struct pci_dev *pdev,
4959
				    const struct pci_device_id *ent)
L
Linus Torvalds 已提交
4960 4961
{
	int i;
4962
	int j = 0;
L
Linus Torvalds 已提交
4963
	int rc;
4964
	int try_soft_reset = 0;
4965
	int dac, return_code;
4966
	InquiryData_struct *inq_buff;
4967
	ctlr_info_t *h;
4968
	unsigned long flags;
L
Linus Torvalds 已提交
4969

4970
	rc = cciss_init_reset_devices(pdev);
4971 4972 4973 4974 4975 4976 4977 4978 4979 4980 4981 4982 4983 4984
	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:

4985
	i = alloc_cciss_hba(pdev);
4986
	if (i < 0)
4987
		return -1;
4988

4989 4990 4991
	h = hba[i];
	h->pdev = pdev;
	h->busy_initializing = 1;
4992
	h->intr_mode = cciss_simple_mode ? SIMPLE_MODE_INT : PERF_MODE_INT;
4993 4994
	INIT_LIST_HEAD(&h->cmpQ);
	INIT_LIST_HEAD(&h->reqQ);
4995
	mutex_init(&h->busy_shutting_down);
4996

4997
	if (cciss_pci_init(h) != 0)
4998
		goto clean_no_release_regions;
L
Linus Torvalds 已提交
4999

5000 5001
	sprintf(h->devname, "cciss%d", i);
	h->ctlr = i;
L
Linus Torvalds 已提交
5002

5003 5004 5005 5006 5007
	if (cciss_tape_cmds < 2)
		cciss_tape_cmds = 2;
	if (cciss_tape_cmds > 16)
		cciss_tape_cmds = 16;

5008
	init_completion(&h->scan_wait);
5009

5010
	if (cciss_create_hba_sysfs_entry(h))
5011 5012
		goto clean0;

L
Linus Torvalds 已提交
5013
	/* configure PCI DMA stuff */
5014
	if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(64)))
5015
		dac = 1;
5016
	else if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(32)))
5017
		dac = 0;
L
Linus Torvalds 已提交
5018
	else {
5019
		dev_err(&h->pdev->dev, "no suitable DMA available\n");
L
Linus Torvalds 已提交
5020 5021 5022 5023 5024 5025 5026 5027 5028
		goto clean1;
	}

	/*
	 * register with the major number, or get a dynamic major number
	 * by passing 0 as argument.  This is done for greater than
	 * 8 controller support.
	 */
	if (i < MAX_CTLR_ORIG)
5029 5030
		h->major = COMPAQ_CISS_MAJOR + i;
	rc = register_blkdev(h->major, h->devname);
5031
	if (rc == -EBUSY || rc == -EINVAL) {
5032 5033
		dev_err(&h->pdev->dev,
		       "Unable to get major number %d for %s "
5034
		       "on hba %d\n", h->major, h->devname, i);
L
Linus Torvalds 已提交
5035
		goto clean1;
5036
	} else {
L
Linus Torvalds 已提交
5037
		if (i >= MAX_CTLR_ORIG)
5038
			h->major = rc;
L
Linus Torvalds 已提交
5039 5040 5041
	}

	/* make sure the board interrupts are off */
5042
	h->access.set_intr_mask(h, CCISS_INTR_OFF);
5043 5044 5045
	rc = cciss_request_irq(h, do_cciss_msix_intr, do_cciss_intx);
	if (rc)
		goto clean2;
5046

5047
	dev_info(&h->pdev->dev, "%s: <0x%x> at PCI %s IRQ %d%s using DAC\n",
5048
	       h->devname, pdev->device, pci_name(pdev),
5049
	       h->intr[h->intr_mode], dac ? "" : " not");
5050

5051
	if (cciss_allocate_cmd_pool(h))
L
Linus Torvalds 已提交
5052
		goto clean4;
5053

5054
	if (cciss_allocate_scatterlists(h))
D
Dan Carpenter 已提交
5055 5056
		goto clean4;

5057 5058 5059
	h->cmd_sg_list = cciss_allocate_sg_chain_blocks(h,
		h->chainsize, h->nr_cmds);
	if (!h->cmd_sg_list && h->chainsize > 0)
5060 5061
		goto clean4;

5062
	spin_lock_init(&h->lock);
L
Linus Torvalds 已提交
5063

5064
	/* Initialize the pdev driver private data.
5065 5066
	   have it point to h.  */
	pci_set_drvdata(pdev, h);
5067 5068
	/* command and error info recs zeroed out before
	   they are used */
5069 5070
	memset(h->cmd_pool_bits, 0,
	       DIV_ROUND_UP(h->nr_cmds, BITS_PER_LONG)
J
Julia Lawall 已提交
5071
			* sizeof(unsigned long));
L
Linus Torvalds 已提交
5072

5073 5074
	h->num_luns = 0;
	h->highest_lun = -1;
M
Mike Miller 已提交
5075
	for (j = 0; j < CISS_MAX_LUN; j++) {
5076 5077
		h->drv[j] = NULL;
		h->gendisk[j] = NULL;
M
Mike Miller 已提交
5078
	}
L
Linus Torvalds 已提交
5079

5080 5081 5082 5083 5084 5085 5086 5087 5088 5089 5090 5091 5092 5093 5094 5095
	/* At this point, the controller is ready to take commands.
	 * 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, CCISS_INTR_OFF);
		spin_unlock_irqrestore(&h->lock, flags);
5096
		free_irq(h->intr[h->intr_mode], h);
5097 5098 5099 5100 5101 5102 5103 5104 5105 5106 5107 5108 5109 5110 5111 5112 5113 5114 5115 5116 5117 5118 5119 5120 5121 5122 5123 5124 5125 5126 5127 5128 5129 5130 5131 5132 5133 5134 5135
		rc = cciss_request_irq(h, cciss_msix_discard_completions,
					cciss_intx_discard_completions);
		if (rc) {
			dev_warn(&h->pdev->dev, "Failed to request_irq after "
				"soft reset.\n");
			goto clean4;
		}

		rc = cciss_kdump_soft_reset(h);
		if (rc) {
			dev_warn(&h->pdev->dev, "Soft reset failed.\n");
			goto clean4;
		}

		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, CCISS_INTR_ON);
		msleep(10000);
		h->access.set_intr_mask(h, CCISS_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.
		 */
		cciss_undo_allocations_after_kdump_soft_reset(h);
		try_soft_reset = 0;
		if (rc)
			/* don't go to clean4, we already unallocated */
			return -ENODEV;

		goto reinit_after_soft_reset;
	}

5136
	cciss_scsi_setup(h);
L
Linus Torvalds 已提交
5137 5138

	/* Turn the interrupts on so we can service requests */
5139
	h->access.set_intr_mask(h, CCISS_INTR_ON);
L
Linus Torvalds 已提交
5140

5141 5142 5143
	/* Get the firmware version */
	inq_buff = kzalloc(sizeof(InquiryData_struct), GFP_KERNEL);
	if (inq_buff == NULL) {
5144
		dev_err(&h->pdev->dev, "out of memory\n");
5145 5146 5147
		goto clean4;
	}

5148
	return_code = sendcmd_withirq(h, CISS_INQUIRY, inq_buff,
5149
		sizeof(InquiryData_struct), 0, CTLR_LUNID, TYPE_CMD);
5150
	if (return_code == IO_OK) {
5151 5152 5153 5154
		h->firm_ver[0] = inq_buff->data_byte[32];
		h->firm_ver[1] = inq_buff->data_byte[33];
		h->firm_ver[2] = inq_buff->data_byte[34];
		h->firm_ver[3] = inq_buff->data_byte[35];
5155
	} else {	 /* send command failed */
5156
		dev_warn(&h->pdev->dev, "unable to determine firmware"
5157 5158
			" version of controller\n");
	}
5159
	kfree(inq_buff);
5160

5161
	cciss_procinit(h);
5162

5163
	h->cciss_max_sectors = 8192;
5164

5165 5166
	rebuild_lun_table(h, 1, 0);
	h->busy_initializing = 0;
5167
	return 1;
L
Linus Torvalds 已提交
5168

M
Mike Miller 已提交
5169
clean4:
5170
	cciss_free_cmd_pool(h);
5171
	cciss_free_scatterlists(h);
5172
	cciss_free_sg_chain_blocks(h->cmd_sg_list, h->nr_cmds);
5173
	free_irq(h->intr[h->intr_mode], h);
M
Mike Miller 已提交
5174
clean2:
5175
	unregister_blkdev(h->major, h->devname);
M
Mike Miller 已提交
5176
clean1:
5177
	cciss_destroy_hba_sysfs_entry(h);
5178
clean0:
5179 5180
	pci_release_regions(pdev);
clean_no_release_regions:
5181
	h->busy_initializing = 0;
5182

5183 5184 5185 5186
	/*
	 * Deliberately omit pci_disable_device(): it does something nasty to
	 * Smart Array controllers that pci_enable_device does not undo
	 */
5187
	pci_set_drvdata(pdev, NULL);
5188
	free_hba(h);
5189
	return -1;
L
Linus Torvalds 已提交
5190 5191
}

5192
static void cciss_shutdown(struct pci_dev *pdev)
L
Linus Torvalds 已提交
5193
{
5194 5195
	ctlr_info_t *h;
	char *flush_buf;
5196
	int return_code;
L
Linus Torvalds 已提交
5197

5198 5199 5200
	h = pci_get_drvdata(pdev);
	flush_buf = kzalloc(4, GFP_KERNEL);
	if (!flush_buf) {
5201
		dev_warn(&h->pdev->dev, "cache not flushed, out of memory.\n");
5202 5203
		return;
	}
5204 5205
	/* write all data in the battery backed cache to disk */
	memset(flush_buf, 0, 4);
5206
	return_code = sendcmd_withirq(h, CCISS_CACHE_FLUSH, flush_buf,
5207 5208 5209
		4, 0, CTLR_LUNID, TYPE_CMD);
	kfree(flush_buf);
	if (return_code != IO_OK)
5210
		dev_warn(&h->pdev->dev, "Error flushing cache\n");
5211
	h->access.set_intr_mask(h, CCISS_INTR_OFF);
5212
	free_irq(h->intr[h->intr_mode], h);
5213 5214
}

5215 5216 5217 5218 5219 5220 5221 5222 5223 5224 5225 5226 5227 5228 5229 5230 5231 5232 5233 5234 5235 5236
static int __devinit cciss_enter_simple_mode(struct ctlr_info *h)
{
	u32 trans_support;

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

	h->max_commands = readl(&(h->cfgtable->CmdsOutMax));
	writel(CFGTBL_Trans_Simple, &(h->cfgtable->HostWrite.TransportRequest));
	writel(CFGTBL_ChangeReq, h->vaddr + SA5_DOORBELL);
	cciss_wait_for_mode_change_ack(h);
	print_cfg_table(h);
	if (!(readl(&(h->cfgtable->TransportActive)) & CFGTBL_Trans_Simple)) {
		dev_warn(&h->pdev->dev, "unable to get board into simple mode\n");
		return -ENODEV;
	}
	h->transMethod = CFGTBL_Trans_Simple;
	return 0;
}


5237 5238
static void __devexit cciss_remove_one(struct pci_dev *pdev)
{
5239
	ctlr_info_t *h;
5240 5241
	int i, j;

5242
	if (pci_get_drvdata(pdev) == NULL) {
5243
		dev_err(&pdev->dev, "Unable to remove device\n");
L
Linus Torvalds 已提交
5244 5245
		return;
	}
5246

5247 5248
	h = pci_get_drvdata(pdev);
	i = h->ctlr;
5249
	if (hba[i] == NULL) {
5250
		dev_err(&pdev->dev, "device appears to already be removed\n");
L
Linus Torvalds 已提交
5251 5252
		return;
	}
5253

5254
	mutex_lock(&h->busy_shutting_down);
5255

5256 5257 5258
	remove_from_scan_list(h);
	remove_proc_entry(h->devname, proc_cciss);
	unregister_blkdev(h->major, h->devname);
5259 5260 5261

	/* remove it from the disk list */
	for (j = 0; j < CISS_MAX_LUN; j++) {
5262
		struct gendisk *disk = h->gendisk[j];
5263
		if (disk) {
5264
			struct request_queue *q = disk->queue;
5265

5266
			if (disk->flags & GENHD_FL_UP) {
5267
				cciss_destroy_ld_sysfs_entry(h, j, 1);
5268
				del_gendisk(disk);
5269
			}
5270 5271 5272 5273 5274
			if (q)
				blk_cleanup_queue(q);
		}
	}

5275
#ifdef CONFIG_CISS_SCSI_TAPE
5276
	cciss_unregister_scsi(h);	/* unhook from SCSI subsystem */
5277
#endif
5278

5279
	cciss_shutdown(pdev);
5280 5281

#ifdef CONFIG_PCI_MSI
5282 5283 5284 5285
	if (h->msix_vector)
		pci_disable_msix(h->pdev);
	else if (h->msi_vector)
		pci_disable_msi(h->pdev);
5286
#endif				/* CONFIG_PCI_MSI */
5287

5288 5289 5290
	iounmap(h->transtable);
	iounmap(h->cfgtable);
	iounmap(h->vaddr);
L
Linus Torvalds 已提交
5291

5292
	cciss_free_cmd_pool(h);
5293
	/* Free up sg elements */
5294 5295 5296 5297
	for (j = 0; j < h->nr_cmds; j++)
		kfree(h->scatter_list[j]);
	kfree(h->scatter_list);
	cciss_free_sg_chain_blocks(h->cmd_sg_list, h->nr_cmds);
5298 5299 5300 5301
	kfree(h->blockFetchTable);
	if (h->reply_pool)
		pci_free_consistent(h->pdev, h->max_commands * sizeof(__u64),
				h->reply_pool, h->reply_pool_dhandle);
5302 5303 5304 5305
	/*
	 * Deliberately omit pci_disable_device(): it does something nasty to
	 * Smart Array controllers that pci_enable_device does not undo
	 */
5306
	pci_release_regions(pdev);
5307
	pci_set_drvdata(pdev, NULL);
5308 5309 5310
	cciss_destroy_hba_sysfs_entry(h);
	mutex_unlock(&h->busy_shutting_down);
	free_hba(h);
5311
}
L
Linus Torvalds 已提交
5312 5313

static struct pci_driver cciss_pci_driver = {
5314 5315 5316 5317
	.name = "cciss",
	.probe = cciss_init_one,
	.remove = __devexit_p(cciss_remove_one),
	.id_table = cciss_pci_device_id,	/* id_table */
5318
	.shutdown = cciss_shutdown,
L
Linus Torvalds 已提交
5319 5320 5321 5322
};

/*
 *  This is it.  Register the PCI driver information for the cards we control
5323
 *  the OS will call our registered routines when it finds one of our cards.
L
Linus Torvalds 已提交
5324 5325 5326
 */
static int __init cciss_init(void)
{
5327 5328
	int err;

5329 5330 5331 5332 5333
	/*
	 * The hardware requires that commands are aligned on a 64-bit
	 * boundary. Given that we use pci_alloc_consistent() to allocate an
	 * array of them, the size must be a multiple of 8 bytes.
	 */
5334
	BUILD_BUG_ON(sizeof(CommandList_struct) % COMMANDLIST_ALIGNMENT);
L
Linus Torvalds 已提交
5335 5336
	printk(KERN_INFO DRIVER_NAME "\n");

5337 5338 5339 5340
	err = bus_register(&cciss_bus_type);
	if (err)
		return err;

5341 5342 5343 5344 5345 5346 5347
	/* Start the scan thread */
	cciss_scan_thread = kthread_run(scan_thread, NULL, "cciss_scan");
	if (IS_ERR(cciss_scan_thread)) {
		err = PTR_ERR(cciss_scan_thread);
		goto err_bus_unregister;
	}

L
Linus Torvalds 已提交
5348
	/* Register for our PCI devices */
5349 5350
	err = pci_register_driver(&cciss_pci_driver);
	if (err)
5351
		goto err_thread_stop;
5352

5353
	return err;
5354

5355 5356 5357
err_thread_stop:
	kthread_stop(cciss_scan_thread);
err_bus_unregister:
5358
	bus_unregister(&cciss_bus_type);
5359

5360
	return err;
L
Linus Torvalds 已提交
5361 5362 5363 5364 5365 5366 5367 5368
}

static void __exit cciss_cleanup(void)
{
	int i;

	pci_unregister_driver(&cciss_pci_driver);
	/* double check that all controller entrys have been removed */
5369 5370
	for (i = 0; i < MAX_CTLR; i++) {
		if (hba[i] != NULL) {
5371 5372
			dev_warn(&hba[i]->pdev->dev,
				"had to remove controller\n");
L
Linus Torvalds 已提交
5373 5374 5375
			cciss_remove_one(hba[i]->pdev);
		}
	}
5376
	kthread_stop(cciss_scan_thread);
5377 5378
	if (proc_cciss)
		remove_proc_entry("driver/cciss", NULL);
5379
	bus_unregister(&cciss_bus_type);
L
Linus Torvalds 已提交
5380 5381 5382 5383
}

module_init(cciss_init);
module_exit(cciss_cleanup);