octeon_droq.c 25.6 KB
Newer Older
1
/**********************************************************************
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
 * Author: Cavium, Inc.
 *
 * Contact: support@cavium.com
 *          Please include "LiquidIO" in the subject.
 *
 * Copyright (c) 2003-2016 Cavium, Inc.
 *
 * This file is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License, Version 2, as
 * published by the Free Software Foundation.
 *
 * This file is distributed in the hope that it will be useful, but
 * AS-IS and WITHOUT ANY WARRANTY; without even the implied warranty
 * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE, TITLE, or
 * NONINFRINGEMENT.  See the GNU General Public License for more details.
 ***********************************************************************/
18 19
#include <linux/pci.h>
#include <linux/netdevice.h>
20
#include <linux/vmalloc.h>
21 22 23 24 25 26 27 28 29
#include "liquidio_common.h"
#include "octeon_droq.h"
#include "octeon_iq.h"
#include "response_manager.h"
#include "octeon_device.h"
#include "octeon_main.h"
#include "octeon_network.h"
#include "cn66xx_regs.h"
#include "cn66xx_device.h"
30
#include "cn23xx_pf_device.h"
R
Raghu Vatsavayi 已提交
31
#include "cn23xx_vf_device.h"
32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89

struct niclist {
	struct list_head list;
	void *ptr;
};

struct __dispatch {
	struct list_head list;
	struct octeon_recv_info *rinfo;
	octeon_dispatch_fn_t disp_fn;
};

/** Get the argument that the user set when registering dispatch
 *  function for a given opcode/subcode.
 *  @param  octeon_dev - the octeon device pointer.
 *  @param  opcode     - the opcode for which the dispatch argument
 *                       is to be checked.
 *  @param  subcode    - the subcode for which the dispatch argument
 *                       is to be checked.
 *  @return  Success: void * (argument to the dispatch function)
 *  @return  Failure: NULL
 *
 */
static inline void *octeon_get_dispatch_arg(struct octeon_device *octeon_dev,
					    u16 opcode, u16 subcode)
{
	int idx;
	struct list_head *dispatch;
	void *fn_arg = NULL;
	u16 combined_opcode = OPCODE_SUBCODE(opcode, subcode);

	idx = combined_opcode & OCTEON_OPCODE_MASK;

	spin_lock_bh(&octeon_dev->dispatch.lock);

	if (octeon_dev->dispatch.count == 0) {
		spin_unlock_bh(&octeon_dev->dispatch.lock);
		return NULL;
	}

	if (octeon_dev->dispatch.dlist[idx].opcode == combined_opcode) {
		fn_arg = octeon_dev->dispatch.dlist[idx].arg;
	} else {
		list_for_each(dispatch,
			      &octeon_dev->dispatch.dlist[idx].list) {
			if (((struct octeon_dispatch *)dispatch)->opcode ==
			    combined_opcode) {
				fn_arg = ((struct octeon_dispatch *)
					  dispatch)->arg;
				break;
			}
		}
	}

	spin_unlock_bh(&octeon_dev->dispatch.lock);
	return fn_arg;
}

90
/** Check for packets on Droq. This function should be called with lock held.
91 92 93
 *  @param  droq - Droq on which count is checked.
 *  @return Returns packet count.
 */
R
Raghu Vatsavayi 已提交
94
u32 octeon_droq_check_hw_for_pkts(struct octeon_droq *droq)
95 96
{
	u32 pkt_count = 0;
97
	u32 last_count;
98 99 100

	pkt_count = readl(droq->pkts_sent_reg);

101 102 103 104 105 106 107 108
	last_count = pkt_count - droq->pkt_count;
	droq->pkt_count = pkt_count;

	/* we shall write to cnts  at napi irq enable or end of droq tasklet */
	if (last_count)
		atomic_add(last_count, &droq->pkts_pending);

	return last_count;
109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143
}

static void octeon_droq_compute_max_packet_bufs(struct octeon_droq *droq)
{
	u32 count = 0;

	/* max_empty_descs is the max. no. of descs that can have no buffers.
	 * If the empty desc count goes beyond this value, we cannot safely
	 * read in a 64K packet sent by Octeon
	 * (64K is max pkt size from Octeon)
	 */
	droq->max_empty_descs = 0;

	do {
		droq->max_empty_descs++;
		count += droq->buffer_size;
	} while (count < (64 * 1024));

	droq->max_empty_descs = droq->max_count - droq->max_empty_descs;
}

static void octeon_droq_reset_indices(struct octeon_droq *droq)
{
	droq->read_idx = 0;
	droq->write_idx = 0;
	droq->refill_idx = 0;
	droq->refill_count = 0;
	atomic_set(&droq->pkts_pending, 0);
}

static void
octeon_droq_destroy_ring_buffers(struct octeon_device *oct,
				 struct octeon_droq *droq)
{
	u32 i;
144
	struct octeon_skb_page_info *pg_info;
145 146

	for (i = 0; i < droq->max_count; i++) {
147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163
		pg_info = &droq->recv_buf_list[i].pg_info;

		if (pg_info->dma)
			lio_unmap_ring(oct->pci_dev,
				       (u64)pg_info->dma);
		pg_info->dma = 0;

		if (pg_info->page)
			recv_buffer_destroy(droq->recv_buf_list[i].buffer,
					    pg_info);

		if (droq->desc_ring && droq->desc_ring[i].info_ptr)
			lio_unmap_ring_info(oct->pci_dev,
					    (u64)droq->
					    desc_ring[i].info_ptr,
					    OCT_DROQ_INFO_SIZE);
		droq->recv_buf_list[i].buffer = NULL;
164 165 166 167 168 169 170 171 172 173 174 175 176 177
	}

	octeon_droq_reset_indices(droq);
}

static int
octeon_droq_setup_ring_buffers(struct octeon_device *oct,
			       struct octeon_droq *droq)
{
	u32 i;
	void *buf;
	struct octeon_droq_desc *desc_ring = droq->desc_ring;

	for (i = 0; i < droq->max_count; i++) {
178
		buf = recv_buffer_alloc(oct, &droq->recv_buf_list[i].pg_info);
179 180 181 182

		if (!buf) {
			dev_err(&oct->pci_dev->dev, "%s buffer alloc failed\n",
				__func__);
183
			droq->stats.rx_alloc_failure++;
184 185 186 187 188 189 190 191 192 193
			return -ENOMEM;
		}

		droq->recv_buf_list[i].buffer = buf;
		droq->recv_buf_list[i].data = get_rbd(buf);
		droq->info_list[i].length = 0;

		/* map ring buffers into memory */
		desc_ring[i].info_ptr = lio_map_ring_info(droq, i);
		desc_ring[i].buffer_ptr =
194
			lio_map_ring(droq->recv_buf_list[i].buffer);
195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210
	}

	octeon_droq_reset_indices(droq);

	octeon_droq_compute_max_packet_bufs(droq);

	return 0;
}

int octeon_delete_droq(struct octeon_device *oct, u32 q_no)
{
	struct octeon_droq *droq = oct->droq[q_no];

	dev_dbg(&oct->pci_dev->dev, "%s[%d]\n", __func__, q_no);

	octeon_droq_destroy_ring_buffers(oct, droq);
211
	vfree(droq->recv_buf_list);
212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236

	if (droq->info_base_addr)
		cnnic_free_aligned_dma(oct->pci_dev, droq->info_list,
				       droq->info_alloc_size,
				       droq->info_base_addr,
				       droq->info_list_dma);

	if (droq->desc_ring)
		lio_dma_free(oct, (droq->max_count * OCT_DROQ_DESC_SIZE),
			     droq->desc_ring, droq->desc_ring_dma);

	memset(droq, 0, OCT_DROQ_SIZE);

	return 0;
}

int octeon_init_droq(struct octeon_device *oct,
		     u32 q_no,
		     u32 num_descs,
		     u32 desc_size,
		     void *app_ctx)
{
	struct octeon_droq *droq;
	u32 desc_ring_size = 0, c_num_descs = 0, c_buf_size = 0;
	u32 c_pkts_per_intr = 0, c_refill_threshold = 0;
237
	int numa_node = dev_to_node(&oct->pci_dev->dev);
238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253

	dev_dbg(&oct->pci_dev->dev, "%s[%d]\n", __func__, q_no);

	droq = oct->droq[q_no];
	memset(droq, 0, OCT_DROQ_SIZE);

	droq->oct_dev = oct;
	droq->q_no = q_no;
	if (app_ctx)
		droq->app_ctx = app_ctx;
	else
		droq->app_ctx = (void *)(size_t)q_no;

	c_num_descs = num_descs;
	c_buf_size = desc_size;
	if (OCTEON_CN6XXX(oct)) {
254
		struct octeon_config *conf6x = CHIP_CONF(oct, cn6xxx);
255 256

		c_pkts_per_intr = (u32)CFG_GET_OQ_PKTS_PER_INTR(conf6x);
257 258
		c_refill_threshold =
			(u32)CFG_GET_OQ_REFILL_THRESHOLD(conf6x);
259
	} else if (OCTEON_CN23XX_PF(oct)) {
260
		struct octeon_config *conf23 = CHIP_CONF(oct, cn23xx_pf);
261

R
Raghu Vatsavayi 已提交
262 263 264 265 266
		c_pkts_per_intr = (u32)CFG_GET_OQ_PKTS_PER_INTR(conf23);
		c_refill_threshold = (u32)CFG_GET_OQ_REFILL_THRESHOLD(conf23);
	} else if (OCTEON_CN23XX_VF(oct)) {
		struct octeon_config *conf23 = CHIP_CONF(oct, cn23xx_vf);

267 268
		c_pkts_per_intr = (u32)CFG_GET_OQ_PKTS_PER_INTR(conf23);
		c_refill_threshold = (u32)CFG_GET_OQ_REFILL_THRESHOLD(conf23);
269 270
	} else {
		return 1;
271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291
	}

	droq->max_count = c_num_descs;
	droq->buffer_size = c_buf_size;

	desc_ring_size = droq->max_count * OCT_DROQ_DESC_SIZE;
	droq->desc_ring = lio_dma_alloc(oct, desc_ring_size,
					(dma_addr_t *)&droq->desc_ring_dma);

	if (!droq->desc_ring) {
		dev_err(&oct->pci_dev->dev,
			"Output queue %d ring alloc failed\n", q_no);
		return 1;
	}

	dev_dbg(&oct->pci_dev->dev, "droq[%d]: desc_ring: virt: 0x%p, dma: %lx\n",
		q_no, droq->desc_ring, droq->desc_ring_dma);
	dev_dbg(&oct->pci_dev->dev, "droq[%d]: num_desc: %d\n", q_no,
		droq->max_count);

	droq->info_list =
292 293 294 295 296
		cnnic_numa_alloc_aligned_dma((droq->max_count *
					      OCT_DROQ_INFO_SIZE),
					     &droq->info_alloc_size,
					     &droq->info_base_addr,
					     numa_node);
297 298 299 300 301 302 303 304
	if (!droq->info_list) {
		dev_err(&oct->pci_dev->dev, "Cannot allocate memory for info list.\n");
		lio_dma_free(oct, (droq->max_count * OCT_DROQ_DESC_SIZE),
			     droq->desc_ring, droq->desc_ring_dma);
		return 1;
	}

	droq->recv_buf_list = (struct octeon_recv_buffer *)
305 306 307 308 309 310
			      vmalloc_node(droq->max_count *
						OCT_DROQ_RECVBUF_SIZE,
						numa_node);
	if (!droq->recv_buf_list)
		droq->recv_buf_list = (struct octeon_recv_buffer *)
				      vmalloc(droq->max_count *
311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332
						OCT_DROQ_RECVBUF_SIZE);
	if (!droq->recv_buf_list) {
		dev_err(&oct->pci_dev->dev, "Output queue recv buf list alloc failed\n");
		goto init_droq_fail;
	}

	if (octeon_droq_setup_ring_buffers(oct, droq))
		goto init_droq_fail;

	droq->pkts_per_intr = c_pkts_per_intr;
	droq->refill_threshold = c_refill_threshold;

	dev_dbg(&oct->pci_dev->dev, "DROQ INIT: max_empty_descs: %d\n",
		droq->max_empty_descs);

	spin_lock_init(&droq->lock);

	INIT_LIST_HEAD(&droq->dispatch_list);

	/* For 56xx Pass1, this function won't be called, so no checks. */
	oct->fn_list.setup_oq_regs(oct, q_no);

R
Raghu Vatsavayi 已提交
333
	oct->io_qmask.oq |= BIT_ULL(q_no);
334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370

	return 0;

init_droq_fail:
	octeon_delete_droq(oct, q_no);
	return 1;
}

/* octeon_create_recv_info
 * Parameters:
 *  octeon_dev - pointer to the octeon device structure
 *  droq       - droq in which the packet arrived.
 *  buf_cnt    - no. of buffers used by the packet.
 *  idx        - index in the descriptor for the first buffer in the packet.
 * Description:
 *  Allocates a recv_info_t and copies the buffer addresses for packet data
 *  into the recv_pkt space which starts at an 8B offset from recv_info_t.
 *  Flags the descriptors for refill later. If available descriptors go
 *  below the threshold to receive a 64K pkt, new buffers are first allocated
 *  before the recv_pkt_t is created.
 *  This routine will be called in interrupt context.
 * Returns:
 *  Success: Pointer to recv_info_t
 *  Failure: NULL.
 * Locks:
 *  The droq->lock is held when this routine is called.
 */
static inline struct octeon_recv_info *octeon_create_recv_info(
		struct octeon_device *octeon_dev,
		struct octeon_droq *droq,
		u32 buf_cnt,
		u32 idx)
{
	struct octeon_droq_info *info;
	struct octeon_recv_pkt *recv_pkt;
	struct octeon_recv_info *recv_info;
	u32 i, bytes_left;
371
	struct octeon_skb_page_info *pg_info;
372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388

	info = &droq->info_list[idx];

	recv_info = octeon_alloc_recv_info(sizeof(struct __dispatch));
	if (!recv_info)
		return NULL;

	recv_pkt = recv_info->recv_pkt;
	recv_pkt->rh = info->rh;
	recv_pkt->length = (u32)info->length;
	recv_pkt->buffer_count = (u16)buf_cnt;
	recv_pkt->octeon_id = (u16)octeon_dev->octeon_id;

	i = 0;
	bytes_left = (u32)info->length;

	while (buf_cnt) {
389 390 391 392 393 394 395 396
		{
			pg_info = &droq->recv_buf_list[idx].pg_info;

			lio_unmap_ring(octeon_dev->pci_dev,
				       (u64)pg_info->dma);
			pg_info->page = NULL;
			pg_info->dma = 0;
		}
397 398 399 400 401 402 403 404

		recv_pkt->buffer_size[i] =
			(bytes_left >=
			 droq->buffer_size) ? droq->buffer_size : bytes_left;

		recv_pkt->buffer_ptr[i] = droq->recv_buf_list[idx].buffer;
		droq->recv_buf_list[idx].buffer = NULL;

405
		idx = incr_index(idx, 1, droq->max_count);
406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435
		bytes_left -= droq->buffer_size;
		i++;
		buf_cnt--;
	}

	return recv_info;
}

/* If we were not able to refill all buffers, try to move around
 * the buffers that were not dispatched.
 */
static inline u32
octeon_droq_refill_pullup_descs(struct octeon_droq *droq,
				struct octeon_droq_desc *desc_ring)
{
	u32 desc_refilled = 0;

	u32 refill_index = droq->refill_idx;

	while (refill_index != droq->read_idx) {
		if (droq->recv_buf_list[refill_index].buffer) {
			droq->recv_buf_list[droq->refill_idx].buffer =
				droq->recv_buf_list[refill_index].buffer;
			droq->recv_buf_list[droq->refill_idx].data =
				droq->recv_buf_list[refill_index].data;
			desc_ring[droq->refill_idx].buffer_ptr =
				desc_ring[refill_index].buffer_ptr;
			droq->recv_buf_list[refill_index].buffer = NULL;
			desc_ring[refill_index].buffer_ptr = 0;
			do {
436 437 438
				droq->refill_idx = incr_index(droq->refill_idx,
							      1,
							      droq->max_count);
439 440 441 442 443
				desc_refilled++;
				droq->refill_count--;
			} while (droq->recv_buf_list[droq->refill_idx].
				 buffer);
		}
444
		refill_index = incr_index(refill_index, 1, droq->max_count);
445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468
	}                       /* while */
	return desc_refilled;
}

/* octeon_droq_refill
 * Parameters:
 *  droq       - droq in which descriptors require new buffers.
 * Description:
 *  Called during normal DROQ processing in interrupt mode or by the poll
 *  thread to refill the descriptors from which buffers were dispatched
 *  to upper layers. Attempts to allocate new buffers. If that fails, moves
 *  up buffers (that were not dispatched) to form a contiguous ring.
 * Returns:
 *  No of descriptors refilled.
 * Locks:
 *  This routine is called with droq->lock held.
 */
static u32
octeon_droq_refill(struct octeon_device *octeon_dev, struct octeon_droq *droq)
{
	struct octeon_droq_desc *desc_ring;
	void *buf = NULL;
	u8 *data;
	u32 desc_refilled = 0;
469
	struct octeon_skb_page_info *pg_info;
470 471 472 473 474 475 476 477 478

	desc_ring = droq->desc_ring;

	while (droq->refill_count && (desc_refilled < droq->max_count)) {
		/* If a valid buffer exists (happens if there is no dispatch),
		 * reuse
		 * the buffer, else allocate.
		 */
		if (!droq->recv_buf_list[droq->refill_idx].buffer) {
479 480 481 482 483 484 485 486 487
			pg_info =
				&droq->recv_buf_list[droq->refill_idx].pg_info;
			/* Either recycle the existing pages or go for
			 * new page alloc
			 */
			if (pg_info->page)
				buf = recv_buffer_reuse(octeon_dev, pg_info);
			else
				buf = recv_buffer_alloc(octeon_dev, pg_info);
488 489 490
			/* If a buffer could not be allocated, no point in
			 * continuing
			 */
491 492
			if (!buf) {
				droq->stats.rx_alloc_failure++;
493
				break;
494
			}
495 496 497 498 499 500 501 502 503 504 505
			droq->recv_buf_list[droq->refill_idx].buffer =
				buf;
			data = get_rbd(buf);
		} else {
			data = get_rbd(droq->recv_buf_list
				       [droq->refill_idx].buffer);
		}

		droq->recv_buf_list[droq->refill_idx].data = data;

		desc_ring[droq->refill_idx].buffer_ptr =
506 507
			lio_map_ring(droq->recv_buf_list[droq->
				     refill_idx].buffer);
508 509 510
		/* Reset any previous values in the length field. */
		droq->info_list[droq->refill_idx].length = 0;

511 512
		droq->refill_idx = incr_index(droq->refill_idx, 1,
					      droq->max_count);
513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566
		desc_refilled++;
		droq->refill_count--;
	}

	if (droq->refill_count)
		desc_refilled +=
			octeon_droq_refill_pullup_descs(droq, desc_ring);

	/* if droq->refill_count
	 * The refill count would not change in pass two. We only moved buffers
	 * to close the gap in the ring, but we would still have the same no. of
	 * buffers to refill.
	 */
	return desc_refilled;
}

static inline u32
octeon_droq_get_bufcount(u32 buf_size, u32 total_len)
{
	u32 buf_cnt = 0;

	while (total_len > (buf_size * buf_cnt))
		buf_cnt++;
	return buf_cnt;
}

static int
octeon_droq_dispatch_pkt(struct octeon_device *oct,
			 struct octeon_droq *droq,
			 union octeon_rh *rh,
			 struct octeon_droq_info *info)
{
	u32 cnt;
	octeon_dispatch_fn_t disp_fn;
	struct octeon_recv_info *rinfo;

	cnt = octeon_droq_get_bufcount(droq->buffer_size, (u32)info->length);

	disp_fn = octeon_get_dispatch(oct, (u16)rh->r.opcode,
				      (u16)rh->r.subcode);
	if (disp_fn) {
		rinfo = octeon_create_recv_info(oct, droq, cnt, droq->read_idx);
		if (rinfo) {
			struct __dispatch *rdisp = rinfo->rsvd;

			rdisp->rinfo = rinfo;
			rdisp->disp_fn = disp_fn;
			rinfo->recv_pkt->rh = *rh;
			list_add_tail(&rdisp->list,
				      &droq->dispatch_list);
		} else {
			droq->stats.dropped_nomem++;
		}
	} else {
567 568 569
		dev_err(&oct->pci_dev->dev, "DROQ: No dispatch function (opcode %u/%u)\n",
			(unsigned int)rh->r.opcode,
			(unsigned int)rh->r.subcode);
570
		droq->stats.dropped_nodispatch++;
R
Raghu Vatsavayi 已提交
571
	}
572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596

	return cnt;
}

static inline void octeon_droq_drop_packets(struct octeon_device *oct,
					    struct octeon_droq *droq,
					    u32 cnt)
{
	u32 i = 0, buf_cnt;
	struct octeon_droq_info *info;

	for (i = 0; i < cnt; i++) {
		info = &droq->info_list[droq->read_idx];
		octeon_swap_8B_data((u64 *)info, 2);

		if (info->length) {
			info->length -= OCT_RH_SIZE;
			droq->stats.bytes_received += info->length;
			buf_cnt = octeon_droq_get_bufcount(droq->buffer_size,
							   (u32)info->length);
		} else {
			dev_err(&oct->pci_dev->dev, "DROQ: In drop: pkt with len 0\n");
			buf_cnt = 1;
		}

597 598
		droq->read_idx = incr_index(droq->read_idx, buf_cnt,
					    droq->max_count);
599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616
		droq->refill_count += buf_cnt;
	}
}

static u32
octeon_droq_fast_process_packets(struct octeon_device *oct,
				 struct octeon_droq *droq,
				 u32 pkts_to_process)
{
	struct octeon_droq_info *info;
	union octeon_rh *rh;
	u32 pkt, total_len = 0, pkt_count;

	pkt_count = pkts_to_process;

	for (pkt = 0; pkt < pkt_count; pkt++) {
		u32 pkt_len = 0;
		struct sk_buff *nicbuf = NULL;
617 618
		struct octeon_skb_page_info *pg_info;
		void *buf;
619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637

		info = &droq->info_list[droq->read_idx];
		octeon_swap_8B_data((u64 *)info, 2);

		if (!info->length) {
			dev_err(&oct->pci_dev->dev,
				"DROQ[%d] idx: %d len:0, pkt_cnt: %d\n",
				droq->q_no, droq->read_idx, pkt_count);
			print_hex_dump_bytes("", DUMP_PREFIX_ADDRESS,
					     (u8 *)info,
					     OCT_DROQ_INFO_SIZE);
			break;
		}

		/* Len of resp hdr in included in the received data len. */
		info->length -= OCT_RH_SIZE;
		rh = &info->rh;

		total_len += (u32)info->length;
638
		if (opcode_slow_path(rh)) {
639 640 641
			u32 buf_cnt;

			buf_cnt = octeon_droq_dispatch_pkt(oct, droq, rh, info);
642 643
			droq->read_idx = incr_index(droq->read_idx,
						    buf_cnt, droq->max_count);
644 645 646 647 648 649
			droq->refill_count += buf_cnt;
		} else {
			if (info->length <= droq->buffer_size) {
				pkt_len = (u32)info->length;
				nicbuf = droq->recv_buf_list[
					droq->read_idx].buffer;
650 651 652 653
				pg_info = &droq->recv_buf_list[
					droq->read_idx].pg_info;
				if (recv_buffer_recycle(oct, pg_info))
					pg_info->page = NULL;
654 655
				droq->recv_buf_list[droq->read_idx].buffer =
					NULL;
656

657 658
				droq->read_idx = incr_index(droq->read_idx, 1,
							    droq->max_count);
659 660
				droq->refill_count++;
			} else {
661
				nicbuf = octeon_fast_packet_alloc((u32)
662 663 664 665 666 667
								  info->length);
				pkt_len = 0;
				/* nicbuf allocation can fail. We'll handle it
				 * inside the loop.
				 */
				while (pkt_len < info->length) {
668
					int cpy_len, idx = droq->read_idx;
669

670 671
					cpy_len = ((pkt_len + droq->buffer_size)
						   > info->length) ?
672 673 674 675 676 677 678
						((u32)info->length - pkt_len) :
						droq->buffer_size;

					if (nicbuf) {
						octeon_fast_packet_next(droq,
									nicbuf,
									cpy_len,
679 680 681 682 683 684 685 686
									idx);
						buf = droq->recv_buf_list[idx].
							buffer;
						recv_buffer_fast_free(buf);
						droq->recv_buf_list[idx].buffer
							= NULL;
					} else {
						droq->stats.rx_alloc_failure++;
687 688 689
					}

					pkt_len += cpy_len;
690 691 692
					droq->read_idx =
						incr_index(droq->read_idx, 1,
							   droq->max_count);
693 694 695 696 697
					droq->refill_count++;
				}
			}

			if (nicbuf) {
698
				if (droq->ops.fptr) {
699
					droq->ops.fptr(oct->octeon_id,
700
						       nicbuf, pkt_len,
701 702
						       rh, &droq->napi,
						       droq->ops.farg);
703
				} else {
704
					recv_buffer_free(nicbuf);
705
				}
706 707 708 709 710 711 712
			}
		}

		if (droq->refill_count >= droq->refill_threshold) {
			int desc_refilled = octeon_droq_refill(oct, droq);

			/* Flush the droq descriptor data to memory to be sure
713 714 715
			 * that when we update the credits the data in memory
			 * is accurate.
			 */
716 717 718 719 720 721
			wmb();
			writel((desc_refilled), droq->pkts_credit_reg);
			/* make sure mmio write completes */
			mmiowb();
		}

722
	}                       /* for (each packet)... */
723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745

	/* Increment refill_count by the number of buffers processed. */
	droq->stats.pkts_received += pkt;
	droq->stats.bytes_received += total_len;

	if ((droq->ops.drop_on_max) && (pkts_to_process - pkt)) {
		octeon_droq_drop_packets(oct, droq, (pkts_to_process - pkt));

		droq->stats.dropped_toomany += (pkts_to_process - pkt);
		return pkts_to_process;
	}

	return pkt;
}

int
octeon_droq_process_packets(struct octeon_device *oct,
			    struct octeon_droq *droq,
			    u32 budget)
{
	u32 pkt_count = 0, pkts_processed = 0;
	struct list_head *tmp, *tmp2;

746 747 748 749
	/* Grab the droq lock */
	spin_lock(&droq->lock);

	octeon_droq_check_hw_for_pkts(droq);
750
	pkt_count = atomic_read(&droq->pkts_pending);
751 752 753

	if (!pkt_count) {
		spin_unlock(&droq->lock);
754
		return 0;
755
	}
756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803

	if (pkt_count > budget)
		pkt_count = budget;

	pkts_processed = octeon_droq_fast_process_packets(oct, droq, pkt_count);

	atomic_sub(pkts_processed, &droq->pkts_pending);

	/* Release the spin lock */
	spin_unlock(&droq->lock);

	list_for_each_safe(tmp, tmp2, &droq->dispatch_list) {
		struct __dispatch *rdisp = (struct __dispatch *)tmp;

		list_del(tmp);
		rdisp->disp_fn(rdisp->rinfo,
			       octeon_get_dispatch_arg
			       (oct,
				(u16)rdisp->rinfo->recv_pkt->rh.r.opcode,
				(u16)rdisp->rinfo->recv_pkt->rh.r.subcode));
	}

	/* If there are packets pending. schedule tasklet again */
	if (atomic_read(&droq->pkts_pending))
		return 1;

	return 0;
}

/**
 * Utility function to poll for packets. check_hw_for_packets must be
 * called before calling this routine.
 */

static int
octeon_droq_process_poll_pkts(struct octeon_device *oct,
			      struct octeon_droq *droq, u32 budget)
{
	struct list_head *tmp, *tmp2;
	u32 pkts_available = 0, pkts_processed = 0;
	u32 total_pkts_processed = 0;

	if (budget > droq->max_count)
		budget = droq->max_count;

	spin_lock(&droq->lock);

	while (total_pkts_processed < budget) {
804 805
		octeon_droq_check_hw_for_pkts(droq);

806 807
		pkts_available = min((budget - total_pkts_processed),
				     (u32)(atomic_read(&droq->pkts_pending)));
808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887

		if (pkts_available == 0)
			break;

		pkts_processed =
			octeon_droq_fast_process_packets(oct, droq,
							 pkts_available);

		atomic_sub(pkts_processed, &droq->pkts_pending);

		total_pkts_processed += pkts_processed;
	}

	spin_unlock(&droq->lock);

	list_for_each_safe(tmp, tmp2, &droq->dispatch_list) {
		struct __dispatch *rdisp = (struct __dispatch *)tmp;

		list_del(tmp);
		rdisp->disp_fn(rdisp->rinfo,
			       octeon_get_dispatch_arg
			       (oct,
				(u16)rdisp->rinfo->recv_pkt->rh.r.opcode,
				(u16)rdisp->rinfo->recv_pkt->rh.r.subcode));
	}

	return total_pkts_processed;
}

int
octeon_process_droq_poll_cmd(struct octeon_device *oct, u32 q_no, int cmd,
			     u32 arg)
{
	struct octeon_droq *droq;

	droq = oct->droq[q_no];

	if (cmd == POLL_EVENT_PROCESS_PKTS)
		return octeon_droq_process_poll_pkts(oct, droq, arg);

	if (cmd == POLL_EVENT_PENDING_PKTS) {
		u32 pkt_cnt = atomic_read(&droq->pkts_pending);

		return  octeon_droq_process_packets(oct, droq, pkt_cnt);
	}

	if (cmd == POLL_EVENT_ENABLE_INTR) {
		u32 value;
		unsigned long flags;

		/* Enable Pkt Interrupt */
		switch (oct->chip_id) {
		case OCTEON_CN66XX:
		case OCTEON_CN68XX: {
			struct octeon_cn6xxx *cn6xxx =
				(struct octeon_cn6xxx *)oct->chip;
			spin_lock_irqsave
				(&cn6xxx->lock_for_droq_int_enb_reg, flags);
			value =
				octeon_read_csr(oct,
						CN6XXX_SLI_PKT_TIME_INT_ENB);
			value |= (1 << q_no);
			octeon_write_csr(oct,
					 CN6XXX_SLI_PKT_TIME_INT_ENB,
					 value);
			value =
				octeon_read_csr(oct,
						CN6XXX_SLI_PKT_CNT_INT_ENB);
			value |= (1 << q_no);
			octeon_write_csr(oct,
					 CN6XXX_SLI_PKT_CNT_INT_ENB,
					 value);

			/* don't bother flushing the enables */

			spin_unlock_irqrestore
				(&cn6xxx->lock_for_droq_int_enb_reg, flags);
			return 0;
		}
		break;
R
Raghu Vatsavayi 已提交
888 889 890 891
		case OCTEON_CN23XX_PF_VID: {
			lio_enable_irq(oct->droq[q_no], oct->instr_queue[q_no]);
		}
		break;
R
Raghu Vatsavayi 已提交
892 893 894 895

		case OCTEON_CN23XX_VF_VID:
			lio_enable_irq(oct->droq[q_no], oct->instr_queue[q_no]);
		break;
896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966
		}
		return 0;
	}

	dev_err(&oct->pci_dev->dev, "%s Unknown command: %d\n", __func__, cmd);
	return -EINVAL;
}

int octeon_register_droq_ops(struct octeon_device *oct, u32 q_no,
			     struct octeon_droq_ops *ops)
{
	struct octeon_droq *droq;
	unsigned long flags;
	struct octeon_config *oct_cfg = NULL;

	oct_cfg = octeon_get_conf(oct);

	if (!oct_cfg)
		return -EINVAL;

	if (!(ops)) {
		dev_err(&oct->pci_dev->dev, "%s: droq_ops pointer is NULL\n",
			__func__);
		return -EINVAL;
	}

	if (q_no >= CFG_GET_OQ_MAX_Q(oct_cfg)) {
		dev_err(&oct->pci_dev->dev, "%s: droq id (%d) exceeds MAX (%d)\n",
			__func__, q_no, (oct->num_oqs - 1));
		return -EINVAL;
	}

	droq = oct->droq[q_no];

	spin_lock_irqsave(&droq->lock, flags);

	memcpy(&droq->ops, ops, sizeof(struct octeon_droq_ops));

	spin_unlock_irqrestore(&droq->lock, flags);

	return 0;
}

int octeon_unregister_droq_ops(struct octeon_device *oct, u32 q_no)
{
	unsigned long flags;
	struct octeon_droq *droq;
	struct octeon_config *oct_cfg = NULL;

	oct_cfg = octeon_get_conf(oct);

	if (!oct_cfg)
		return -EINVAL;

	if (q_no >= CFG_GET_OQ_MAX_Q(oct_cfg)) {
		dev_err(&oct->pci_dev->dev, "%s: droq id (%d) exceeds MAX (%d)\n",
			__func__, q_no, oct->num_oqs - 1);
		return -EINVAL;
	}

	droq = oct->droq[q_no];

	if (!droq) {
		dev_info(&oct->pci_dev->dev,
			 "Droq id (%d) not available.\n", q_no);
		return 0;
	}

	spin_lock_irqsave(&droq->lock, flags);

	droq->ops.fptr = NULL;
967
	droq->ops.farg = NULL;
968 969 970 971 972 973 974 975 976 977 978 979
	droq->ops.drop_on_max = 0;

	spin_unlock_irqrestore(&droq->lock, flags);

	return 0;
}

int octeon_create_droq(struct octeon_device *oct,
		       u32 q_no, u32 num_descs,
		       u32 desc_size, void *app_ctx)
{
	struct octeon_droq *droq;
980
	int numa_node = dev_to_node(&oct->pci_dev->dev);
981 982 983 984 985 986 987 988

	if (oct->droq[q_no]) {
		dev_dbg(&oct->pci_dev->dev, "Droq already in use. Cannot create droq %d again\n",
			q_no);
		return 1;
	}

	/* Allocate the DS for the new droq. */
989 990 991
	droq = vmalloc_node(sizeof(*droq), numa_node);
	if (!droq)
		droq = vmalloc(sizeof(*droq));
992
	if (!droq)
R
Raghu Vatsavayi 已提交
993 994
		return -1;

995 996 997 998 999 1000 1001
	memset(droq, 0, sizeof(struct octeon_droq));

	/*Disable the pkt o/p for this Q  */
	octeon_set_droq_pkt_op(oct, q_no, 0);
	oct->droq[q_no] = droq;

	/* Initialize the Droq */
R
Raghu Vatsavayi 已提交
1002 1003 1004 1005 1006
	if (octeon_init_droq(oct, q_no, num_descs, desc_size, app_ctx)) {
		vfree(oct->droq[q_no]);
		oct->droq[q_no] = NULL;
		return -1;
	}
1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019

	oct->num_oqs++;

	dev_dbg(&oct->pci_dev->dev, "%s: Total number of OQ: %d\n", __func__,
		oct->num_oqs);

	/* Global Droq register settings */

	/* As of now not required, as setting are done for all 32 Droqs at
	 * the same time.
	 */
	return 0;
}