Free Electrons

Embedded Linux Experts

  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 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
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
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
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
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
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
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
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
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
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
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
479
480
481
482
483
484
485
486
487
488
489
490
491
492
/*
 * Intel Wireless Multicomm 3200 WiFi driver
 *
 * Copyright (C) 2009 Intel Corporation. All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 *
 *   * Redistributions of source code must retain the above copyright
 *     notice, this list of conditions and the following disclaimer.
 *   * Redistributions in binary form must reproduce the above copyright
 *     notice, this list of conditions and the following disclaimer in
 *     the documentation and/or other materials provided with the
 *     distribution.
 *   * Neither the name of Intel Corporation nor the names of its
 *     contributors may be used to endorse or promote products derived
 *     from this software without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *
 *
 * Intel Corporation <ilw@linux.intel.com>
 * Samuel Ortiz <samuel.ortiz@intel.com>
 * Zhu Yi <yi.zhu@intel.com>
 *
 */

/*
 * iwm Tx theory of operation:
 *
 * 1) We receive a 802.3 frame from the stack
 * 2) We convert it to a 802.11 frame [iwm_xmit_frame]
 * 3) We queue it to its corresponding tx queue [iwm_xmit_frame]
 * 4) We schedule the tx worker. There is one worker per tx
 *    queue. [iwm_xmit_frame]
 * 5) The tx worker is scheduled
 * 6) We go through every queued skb on the tx queue, and for each
 *    and every one of them: [iwm_tx_worker]
 *    a) We check if we have enough Tx credits (see below for a Tx
 *       credits description) for the frame length. [iwm_tx_worker]
 *    b) If we do, we aggregate the Tx frame into a UDMA one, by
 *       concatenating one REPLY_TX command per Tx frame. [iwm_tx_worker]
 *    c) When we run out of credits, or when we reach the maximum
 *       concatenation size, we actually send the concatenated UDMA
 *       frame. [iwm_tx_worker]
 *
 * When we run out of Tx credits, the skbs are filling the tx queue,
 * and eventually we will stop the netdev queue. [iwm_tx_worker]
 * The tx queue is emptied as we're getting new tx credits, by
 * scheduling the tx_worker. [iwm_tx_credit_inc]
 * The netdev queue is started again when we have enough tx credits,
 * and when our tx queue has some reasonable amout of space available
 * (i.e. half of the max size). [iwm_tx_worker]
 */

#include <linux/skbuff.h>
#include <linux/netdevice.h>
#include <linux/ieee80211.h>

#include "iwm.h"
#include "debug.h"
#include "commands.h"
#include "hal.h"
#include "umac.h"
#include "bus.h"

#define IWM_UMAC_PAGE_ALLOC_WRAP 0xffff

#define BYTES_TO_PAGES(n)	 (1 + ((n) >> ilog2(IWM_UMAC_PAGE_SIZE)) - \
				 (((n) & (IWM_UMAC_PAGE_SIZE - 1)) == 0))

#define pool_id_to_queue(id)	 ((id < IWM_TX_CMD_QUEUE) ? id : id - 1)
#define queue_to_pool_id(q)	 ((q < IWM_TX_CMD_QUEUE) ? q : q + 1)

/* require to hold tx_credit lock */
static int iwm_tx_credit_get(struct iwm_tx_credit *tx_credit, int id)
{
	struct pool_entry *pool = &tx_credit->pools[id];
	struct spool_entry *spool = &tx_credit->spools[pool->sid];
	int spool_pages;

	/* number of pages can be taken from spool by this pool */
	spool_pages = spool->max_pages - spool->alloc_pages +
		      max(pool->min_pages - pool->alloc_pages, 0);

	return min(pool->max_pages - pool->alloc_pages, spool_pages);
}

static bool iwm_tx_credit_ok(struct iwm_priv *iwm, int id, int nb)
{
	u32 npages = BYTES_TO_PAGES(nb);

	if (npages <= iwm_tx_credit_get(&iwm->tx_credit, id))
		return 1;

	set_bit(id, &iwm->tx_credit.full_pools_map);

	IWM_DBG_TX(iwm, DBG, "LINK: stop txq[%d], available credit: %d\n",
		   pool_id_to_queue(id),
		   iwm_tx_credit_get(&iwm->tx_credit, id));

	return 0;
}

void iwm_tx_credit_inc(struct iwm_priv *iwm, int id, int total_freed_pages)
{
	struct pool_entry *pool;
	struct spool_entry *spool;
	int freed_pages;
	int queue;

	BUG_ON(id >= IWM_MACS_OUT_GROUPS);

	pool = &iwm->tx_credit.pools[id];
	spool = &iwm->tx_credit.spools[pool->sid];

	freed_pages = total_freed_pages - pool->total_freed_pages;
	IWM_DBG_TX(iwm, DBG, "Free %d pages for pool[%d]\n", freed_pages, id);

	if (!freed_pages) {
		IWM_DBG_TX(iwm, DBG, "No pages are freed by UMAC\n");
		return;
	} else if (freed_pages < 0)
		freed_pages += IWM_UMAC_PAGE_ALLOC_WRAP + 1;

	if (pool->alloc_pages > pool->min_pages) {
		int spool_pages = pool->alloc_pages - pool->min_pages;
		spool_pages = min(spool_pages, freed_pages);
		spool->alloc_pages -= spool_pages;
	}

	pool->alloc_pages -= freed_pages;
	pool->total_freed_pages = total_freed_pages;

	IWM_DBG_TX(iwm, DBG, "Pool[%d] pages alloc: %d, total_freed: %d, "
		   "Spool[%d] pages alloc: %d\n", id, pool->alloc_pages,
		   pool->total_freed_pages, pool->sid, spool->alloc_pages);

	if (test_bit(id, &iwm->tx_credit.full_pools_map) &&
	    (pool->alloc_pages < pool->max_pages / 2)) {
		clear_bit(id, &iwm->tx_credit.full_pools_map);

		queue = pool_id_to_queue(id);

		IWM_DBG_TX(iwm, DBG, "LINK: start txq[%d], available "
			   "credit: %d\n", queue,
			   iwm_tx_credit_get(&iwm->tx_credit, id));
		queue_work(iwm->txq[queue].wq, &iwm->txq[queue].worker);
	}
}

static void iwm_tx_credit_dec(struct iwm_priv *iwm, int id, int alloc_pages)
{
	struct pool_entry *pool;
	struct spool_entry *spool;
	int spool_pages;

	IWM_DBG_TX(iwm, DBG, "Allocate %d pages for pool[%d]\n",
		   alloc_pages, id);

	BUG_ON(id >= IWM_MACS_OUT_GROUPS);

	pool = &iwm->tx_credit.pools[id];
	spool = &iwm->tx_credit.spools[pool->sid];

	spool_pages = pool->alloc_pages + alloc_pages - pool->min_pages;

	if (pool->alloc_pages >= pool->min_pages)
		spool->alloc_pages += alloc_pages;
	else if (spool_pages > 0)
		spool->alloc_pages += spool_pages;

	pool->alloc_pages += alloc_pages;

	IWM_DBG_TX(iwm, DBG, "Pool[%d] pages alloc: %d, total_freed: %d, "
		   "Spool[%d] pages alloc: %d\n", id, pool->alloc_pages,
		   pool->total_freed_pages, pool->sid, spool->alloc_pages);
}

int iwm_tx_credit_alloc(struct iwm_priv *iwm, int id, int nb)
{
	u32 npages = BYTES_TO_PAGES(nb);
	int ret = 0;

	spin_lock(&iwm->tx_credit.lock);

	if (!iwm_tx_credit_ok(iwm, id, nb)) {
		IWM_DBG_TX(iwm, DBG, "No credit avaliable for pool[%d]\n", id);
		ret = -ENOSPC;
		goto out;
	}

	iwm_tx_credit_dec(iwm, id, npages);

 out:
	spin_unlock(&iwm->tx_credit.lock);
	return ret;
}

/*
 * Since we're on an SDIO or USB bus, we are not sharing memory
 * for storing to be transmitted frames. The host needs to push
 * them upstream. As a consequence there needs to be a way for
 * the target to let us know if it can actually take more TX frames
 * or not. This is what Tx credits are for.
 *
 * For each Tx HW queue, we have a Tx pool, and then we have one
 * unique super pool (spool), which is actually a global pool of
 * all the UMAC pages.
 * For each Tx pool we have a min_pages, a max_pages fields, and a
 * alloc_pages fields. The alloc_pages tracks the number of pages
 * currently allocated from the tx pool.
 * Here are the rules to check if given a tx frame we have enough
 * tx credits for it:
 * 1) We translate the frame length into a number of UMAC pages.
 *    Let's call them n_pages.
 * 2) For the corresponding tx pool, we check if n_pages +
 *    pool->alloc_pages is higher than pool->min_pages. min_pages
 *    represent a set of pre-allocated pages on the tx pool. If
 *    that's the case, then we need to allocate those pages from
 *    the spool. We can do so until we reach spool->max_pages.
 * 3) Each tx pool is not allowed to allocate more than pool->max_pages
 *    from the spool, so once we're over min_pages, we can allocate
 *    pages from the spool, but not more than max_pages.
 *
 * When the tx code path needs to send a tx frame, it checks first
 * if it has enough tx credits, following those rules. [iwm_tx_credit_get]
 * If it does, it then updates the pool and spool counters and
 * then send the frame. [iwm_tx_credit_alloc and iwm_tx_credit_dec]
 * On the other side, when the UMAC is done transmitting frames, it
 * will send a credit update notification to the host. This is when
 * the pool and spool counters gets to be decreased. [iwm_tx_credit_inc,
 * called from rx.c:iwm_ntf_tx_credit_update]
 *
 */
void iwm_tx_credit_init_pools(struct iwm_priv *iwm,
			      struct iwm_umac_notif_alive *alive)
{
	int i, sid, pool_pages;

	spin_lock(&iwm->tx_credit.lock);

	iwm->tx_credit.pool_nr = le16_to_cpu(alive->page_grp_count);
	iwm->tx_credit.full_pools_map = 0;
	memset(&iwm->tx_credit.spools[0], 0, sizeof(struct spool_entry));

	IWM_DBG_TX(iwm, DBG, "Pools number is %d\n", iwm->tx_credit.pool_nr);

	for (i = 0; i < iwm->tx_credit.pool_nr; i++) {
		__le32 page_grp_state = alive->page_grp_state[i];

		iwm->tx_credit.pools[i].id = GET_VAL32(page_grp_state,
				UMAC_ALIVE_PAGE_STS_GRP_NUM);
		iwm->tx_credit.pools[i].sid = GET_VAL32(page_grp_state,
				UMAC_ALIVE_PAGE_STS_SGRP_NUM);
		iwm->tx_credit.pools[i].min_pages = GET_VAL32(page_grp_state,
				UMAC_ALIVE_PAGE_STS_GRP_MIN_SIZE);
		iwm->tx_credit.pools[i].max_pages = GET_VAL32(page_grp_state,
				UMAC_ALIVE_PAGE_STS_GRP_MAX_SIZE);
		iwm->tx_credit.pools[i].alloc_pages = 0;
		iwm->tx_credit.pools[i].total_freed_pages = 0;

		sid = iwm->tx_credit.pools[i].sid;
		pool_pages = iwm->tx_credit.pools[i].min_pages;

		if (iwm->tx_credit.spools[sid].max_pages == 0) {
			iwm->tx_credit.spools[sid].id = sid;
			iwm->tx_credit.spools[sid].max_pages =
				GET_VAL32(page_grp_state,
					  UMAC_ALIVE_PAGE_STS_SGRP_MAX_SIZE);
			iwm->tx_credit.spools[sid].alloc_pages = 0;
		}

		iwm->tx_credit.spools[sid].alloc_pages += pool_pages;

		IWM_DBG_TX(iwm, DBG, "Pool idx: %d, id: %d, sid: %d, capacity "
			   "min: %d, max: %d, pool alloc: %d, total_free: %d, "
			   "super poll alloc: %d\n",
			   i, iwm->tx_credit.pools[i].id,
			   iwm->tx_credit.pools[i].sid,
			   iwm->tx_credit.pools[i].min_pages,
			   iwm->tx_credit.pools[i].max_pages,
			   iwm->tx_credit.pools[i].alloc_pages,
			   iwm->tx_credit.pools[i].total_freed_pages,
			   iwm->tx_credit.spools[sid].alloc_pages);
	}

	spin_unlock(&iwm->tx_credit.lock);
}

#define IWM_UDMA_HDR_LEN	sizeof(struct iwm_umac_wifi_out_hdr)

static int iwm_tx_build_packet(struct iwm_priv *iwm, struct sk_buff *skb,
			       int pool_id, u8 *buf)
{
	struct iwm_umac_wifi_out_hdr *hdr = (struct iwm_umac_wifi_out_hdr *)buf;
	struct iwm_udma_wifi_cmd udma_cmd;
	struct iwm_umac_cmd umac_cmd;
	struct iwm_tx_info *tx_info = skb_to_tx_info(skb);

	udma_cmd.count = cpu_to_le16(skb->len +
				     sizeof(struct iwm_umac_fw_cmd_hdr));
	/* set EOP to 0 here. iwm_udma_wifi_hdr_set_eop() will be
	 * called later to set EOP for the last packet. */
	udma_cmd.eop = 0;
	udma_cmd.credit_group = pool_id;
	udma_cmd.ra_tid = tx_info->sta << 4 | tx_info->tid;
	udma_cmd.lmac_offset = 0;

	umac_cmd.id = REPLY_TX;
	umac_cmd.count = cpu_to_le16(skb->len);
	umac_cmd.color = tx_info->color;
	umac_cmd.resp = 0;
	umac_cmd.seq_num = cpu_to_le16(iwm_alloc_wifi_cmd_seq(iwm));

	iwm_build_udma_wifi_hdr(iwm, &hdr->hw_hdr, &udma_cmd);
	iwm_build_umac_hdr(iwm, &hdr->sw_hdr, &umac_cmd);

	memcpy(buf + sizeof(*hdr), skb->data, skb->len);

	return 0;
}

static int iwm_tx_send_concat_packets(struct iwm_priv *iwm,
				      struct iwm_tx_queue *txq)
{
	int ret;

	if (!txq->concat_count)
		return 0;

	IWM_DBG_TX(iwm, DBG, "Send concatenated Tx: queue %d, %d bytes\n",
		   txq->id, txq->concat_count);

	/* mark EOP for the last packet */
	iwm_udma_wifi_hdr_set_eop(iwm, txq->concat_ptr, 1);

	ret = iwm_bus_send_chunk(iwm, txq->concat_buf, txq->concat_count);

	txq->concat_count = 0;
	txq->concat_ptr = txq->concat_buf;

	return ret;
}

#define CONFIG_IWM_TX_CONCATENATED 1

void iwm_tx_worker(struct work_struct *work)
{
	struct iwm_priv *iwm;
	struct iwm_tx_info *tx_info = NULL;
	struct sk_buff *skb;
	int cmdlen, ret;
	struct iwm_tx_queue *txq;
	int pool_id;

	txq = container_of(work, struct iwm_tx_queue, worker);
	iwm = container_of(txq, struct iwm_priv, txq[txq->id]);

	pool_id = queue_to_pool_id(txq->id);

	while (!test_bit(pool_id, &iwm->tx_credit.full_pools_map) &&
	       !skb_queue_empty(&txq->queue)) {

		skb = skb_dequeue(&txq->queue);
		tx_info = skb_to_tx_info(skb);
		cmdlen = IWM_UDMA_HDR_LEN + skb->len;

		IWM_DBG_TX(iwm, DBG, "Tx frame on queue %d: skb: 0x%p, sta: "
			   "%d, color: %d\n", txq->id, skb, tx_info->sta,
			   tx_info->color);

#if !CONFIG_IWM_TX_CONCATENATED
		/* temporarily keep this to comparing the performance */
		ret = iwm_send_packet(iwm, skb, pool_id);
#else

		if (txq->concat_count + cmdlen > IWM_HAL_CONCATENATE_BUF_SIZE)
			iwm_tx_send_concat_packets(iwm, txq);

		ret = iwm_tx_credit_alloc(iwm, pool_id, cmdlen);
		if (ret) {
			IWM_DBG_TX(iwm, DBG, "not enough tx_credit for queue "
				   "%d, Tx worker stopped\n", txq->id);
			skb_queue_head(&txq->queue, skb);
			break;
		}

		txq->concat_ptr = txq->concat_buf + txq->concat_count;
		iwm_tx_build_packet(iwm, skb, pool_id, txq->concat_ptr);
		txq->concat_count += ALIGN(cmdlen, 16);
#endif
		kfree_skb(skb);
	}

	iwm_tx_send_concat_packets(iwm, txq);

	if (__netif_subqueue_stopped(iwm_to_ndev(iwm), txq->id) &&
	    !test_bit(pool_id, &iwm->tx_credit.full_pools_map) &&
	    (skb_queue_len(&txq->queue) < IWM_TX_LIST_SIZE / 2)) {
		IWM_DBG_TX(iwm, DBG, "LINK: start netif_subqueue[%d]", txq->id);
		netif_wake_subqueue(iwm_to_ndev(iwm), txq->id);
	}
}

int iwm_xmit_frame(struct sk_buff *skb, struct net_device *netdev)
{
	struct iwm_priv *iwm = ndev_to_iwm(netdev);
	struct net_device *ndev = iwm_to_ndev(iwm);
	struct wireless_dev *wdev = iwm_to_wdev(iwm);
	u8 *dst_addr;
	struct iwm_tx_info *tx_info;
	struct iwm_tx_queue *txq;
	struct iwm_sta_info *sta_info;
	u8 sta_id;
	u16 queue;
	int ret;

	if (!test_bit(IWM_STATUS_ASSOCIATED, &iwm->status)) {
		IWM_DBG_TX(iwm, DBG, "LINK: stop netif_all_queues: "
			   "not associated\n");
		netif_tx_stop_all_queues(netdev);
		goto drop;
	}

	queue = skb_get_queue_mapping(skb);
	BUG_ON(queue >= IWM_TX_DATA_QUEUES); /* no iPAN yet */

	txq = &iwm->txq[queue];

	/* No free space for Tx, tx_worker is too slow */
	if (skb_queue_len(&txq->queue) > IWM_TX_LIST_SIZE) {
		IWM_DBG_TX(iwm, DBG, "LINK: stop netif_subqueue[%d]\n", queue);
		netif_stop_subqueue(netdev, queue);
		return NETDEV_TX_BUSY;
	}

	ret = ieee80211_data_from_8023(skb, netdev->dev_addr, wdev->iftype,
				       iwm->bssid, 0);
	if (ret) {
		IWM_ERR(iwm, "build wifi header failed\n");
		goto drop;
	}

	dst_addr = ((struct ieee80211_hdr *)(skb->data))->addr1;

	for (sta_id = 0; sta_id < IWM_STA_TABLE_NUM; sta_id++) {
		sta_info = &iwm->sta_table[sta_id];
		if (sta_info->valid &&
		    !memcmp(dst_addr, sta_info->addr, ETH_ALEN))
			break;
	}

	if (sta_id == IWM_STA_TABLE_NUM) {
		IWM_ERR(iwm, "STA %pM not found in sta_table, Tx ignored\n",
			dst_addr);
		goto drop;
	}

	tx_info = skb_to_tx_info(skb);
	tx_info->sta = sta_id;
	tx_info->color = sta_info->color;
	/* UMAC uses TID 8 (vs. 0) for non QoS packets */
	if (sta_info->qos)
		tx_info->tid = skb->priority;
	else
		tx_info->tid = IWM_UMAC_MGMT_TID;

	skb_queue_tail(&iwm->txq[queue].queue, skb);

	queue_work(iwm->txq[queue].wq, &iwm->txq[queue].worker);

	ndev->stats.tx_packets++;
	ndev->stats.tx_bytes += skb->len;
	return NETDEV_TX_OK;

 drop:
	ndev->stats.tx_dropped++;
	dev_kfree_skb_any(skb);
	return NETDEV_TX_OK;
}