1// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause
2/*
3 * Copyright (C) 2003-2014, 2018-2021, 2023 Intel Corporation
4 * Copyright (C) 2013-2015 Intel Mobile Communications GmbH
5 * Copyright (C) 2016-2017 Intel Deutschland GmbH
6 */
7#include <linux/etherdevice.h>
8#include <linux/ieee80211.h>
9#include <linux/slab.h>
10#include <linux/sched.h>
11#include <net/ip6_checksum.h>
12#include <net/tso.h>
13
14#include "iwl-debug.h"
15#include "iwl-csr.h"
16#include "iwl-prph.h"
17#include "iwl-io.h"
18#include "iwl-scd.h"
19#include "iwl-op-mode.h"
20#include "internal.h"
21#include "fw/api/tx.h"
22
23/*************** DMA-QUEUE-GENERAL-FUNCTIONS  *****
24 * DMA services
25 *
26 * Theory of operation
27 *
28 * A Tx or Rx queue resides in host DRAM, and is comprised of a circular buffer
29 * of buffer descriptors, each of which points to one or more data buffers for
30 * the device to read from or fill.  Driver and device exchange status of each
31 * queue via "read" and "write" pointers.  Driver keeps minimum of 2 empty
32 * entries in each circular buffer, to protect against confusing empty and full
33 * queue states.
34 *
35 * The device reads or writes the data in the queues via the device's several
36 * DMA/FIFO channels.  Each queue is mapped to a single DMA channel.
37 *
38 * For Tx queue, there are low mark and high mark limits. If, after queuing
39 * the packet for Tx, free space become < low mark, Tx queue stopped. When
40 * reclaiming packets (on 'tx done IRQ), if free space become > high mark,
41 * Tx queue resumed.
42 *
43 ***************************************************/
44
45
46int iwl_pcie_alloc_dma_ptr(struct iwl_trans *trans,
47			   struct iwl_dma_ptr *ptr, size_t size)
48{
49	if (WARN_ON(ptr->addr))
50		return -EINVAL;
51
52	ptr->addr = dma_alloc_coherent(trans->dev, size,
53				       &ptr->dma, GFP_KERNEL);
54	if (!ptr->addr)
55		return -ENOMEM;
56	ptr->size = size;
57	return 0;
58}
59
60void iwl_pcie_free_dma_ptr(struct iwl_trans *trans, struct iwl_dma_ptr *ptr)
61{
62	if (unlikely(!ptr->addr))
63		return;
64
65	dma_free_coherent(trans->dev, ptr->size, ptr->addr, ptr->dma);
66	memset(ptr, 0, sizeof(*ptr));
67}
68
69/*
70 * iwl_pcie_txq_inc_wr_ptr - Send new write index to hardware
71 */
72static void iwl_pcie_txq_inc_wr_ptr(struct iwl_trans *trans,
73				    struct iwl_txq *txq)
74{
75	u32 reg = 0;
76	int txq_id = txq->id;
77
78	lockdep_assert_held(&txq->lock);
79
80	/*
81	 * explicitly wake up the NIC if:
82	 * 1. shadow registers aren't enabled
83	 * 2. NIC is woken up for CMD regardless of shadow outside this function
84	 * 3. there is a chance that the NIC is asleep
85	 */
86	if (!trans->trans_cfg->base_params->shadow_reg_enable &&
87	    txq_id != trans->txqs.cmd.q_id &&
88	    test_bit(STATUS_TPOWER_PMI, &trans->status)) {
89		/*
90		 * wake up nic if it's powered down ...
91		 * uCode will wake up, and interrupt us again, so next
92		 * time we'll skip this part.
93		 */
94		reg = iwl_read32(trans, CSR_UCODE_DRV_GP1);
95
96		if (reg & CSR_UCODE_DRV_GP1_BIT_MAC_SLEEP) {
97			IWL_DEBUG_INFO(trans, "Tx queue %d requesting wakeup, GP1 = 0x%x\n",
98				       txq_id, reg);
99			iwl_set_bit(trans, CSR_GP_CNTRL,
100				    CSR_GP_CNTRL_REG_FLAG_MAC_ACCESS_REQ);
101			txq->need_update = true;
102			return;
103		}
104	}
105
106	/*
107	 * if not in power-save mode, uCode will never sleep when we're
108	 * trying to tx (during RFKILL, we're not trying to tx).
109	 */
110	IWL_DEBUG_TX(trans, "Q:%d WR: 0x%x\n", txq_id, txq->write_ptr);
111	if (!txq->block)
112		iwl_write32(trans, HBUS_TARG_WRPTR,
113			    txq->write_ptr | (txq_id << 8));
114}
115
116void iwl_pcie_txq_check_wrptrs(struct iwl_trans *trans)
117{
118	int i;
119
120	for (i = 0; i < trans->trans_cfg->base_params->num_of_queues; i++) {
121		struct iwl_txq *txq = trans->txqs.txq[i];
122
123		if (!test_bit(i, trans->txqs.queue_used))
124			continue;
125
126		spin_lock_bh(&txq->lock);
127		if (txq->need_update) {
128			iwl_pcie_txq_inc_wr_ptr(trans, txq);
129			txq->need_update = false;
130		}
131		spin_unlock_bh(&txq->lock);
132	}
133}
134
135static int iwl_pcie_txq_build_tfd(struct iwl_trans *trans, struct iwl_txq *txq,
136				  dma_addr_t addr, u16 len, bool reset)
137{
138	void *tfd;
139	u32 num_tbs;
140
141	tfd = (u8 *)txq->tfds + trans->txqs.tfd.size * txq->write_ptr;
142
143	if (reset)
144		memset(tfd, 0, trans->txqs.tfd.size);
145
146	num_tbs = iwl_txq_gen1_tfd_get_num_tbs(trans, tfd);
147
148	/* Each TFD can point to a maximum max_tbs Tx buffers */
149	if (num_tbs >= trans->txqs.tfd.max_tbs) {
150		IWL_ERR(trans, "Error can not send more than %d chunks\n",
151			trans->txqs.tfd.max_tbs);
152		return -EINVAL;
153	}
154
155	if (WARN(addr & ~IWL_TX_DMA_MASK,
156		 "Unaligned address = %llx\n", (unsigned long long)addr))
157		return -EINVAL;
158
159	iwl_pcie_gen1_tfd_set_tb(trans, tfd, num_tbs, addr, len);
160
161	return num_tbs;
162}
163
164static void iwl_pcie_clear_cmd_in_flight(struct iwl_trans *trans)
165{
166	struct iwl_trans_pcie *trans_pcie = IWL_TRANS_GET_PCIE_TRANS(trans);
167
168	if (!trans->trans_cfg->base_params->apmg_wake_up_wa)
169		return;
170
171	spin_lock(&trans_pcie->reg_lock);
172
173	if (WARN_ON(!trans_pcie->cmd_hold_nic_awake)) {
174		spin_unlock(&trans_pcie->reg_lock);
175		return;
176	}
177
178	trans_pcie->cmd_hold_nic_awake = false;
179	__iwl_trans_pcie_clear_bit(trans, CSR_GP_CNTRL,
180				   CSR_GP_CNTRL_REG_FLAG_MAC_ACCESS_REQ);
181	spin_unlock(&trans_pcie->reg_lock);
182}
183
184/*
185 * iwl_pcie_txq_unmap -  Unmap any remaining DMA mappings and free skb's
186 */
187static void iwl_pcie_txq_unmap(struct iwl_trans *trans, int txq_id)
188{
189	struct iwl_txq *txq = trans->txqs.txq[txq_id];
190
191	if (!txq) {
192		IWL_ERR(trans, "Trying to free a queue that wasn't allocated?\n");
193		return;
194	}
195
196	spin_lock_bh(&txq->lock);
197	while (txq->write_ptr != txq->read_ptr) {
198		IWL_DEBUG_TX_REPLY(trans, "Q %d Free %d\n",
199				   txq_id, txq->read_ptr);
200
201		if (txq_id != trans->txqs.cmd.q_id) {
202			struct sk_buff *skb = txq->entries[txq->read_ptr].skb;
203
204			if (WARN_ON_ONCE(!skb))
205				continue;
206
207			iwl_txq_free_tso_page(trans, skb);
208		}
209		iwl_txq_free_tfd(trans, txq);
210		txq->read_ptr = iwl_txq_inc_wrap(trans, txq->read_ptr);
211
212		if (txq->read_ptr == txq->write_ptr &&
213		    txq_id == trans->txqs.cmd.q_id)
214			iwl_pcie_clear_cmd_in_flight(trans);
215	}
216
217	while (!skb_queue_empty(&txq->overflow_q)) {
218		struct sk_buff *skb = __skb_dequeue(&txq->overflow_q);
219
220		iwl_op_mode_free_skb(trans->op_mode, skb);
221	}
222
223	spin_unlock_bh(&txq->lock);
224
225	/* just in case - this queue may have been stopped */
226	iwl_wake_queue(trans, txq);
227}
228
229/*
230 * iwl_pcie_txq_free - Deallocate DMA queue.
231 * @txq: Transmit queue to deallocate.
232 *
233 * Empty queue by removing and destroying all BD's.
234 * Free all buffers.
235 * 0-fill, but do not free "txq" descriptor structure.
236 */
237static void iwl_pcie_txq_free(struct iwl_trans *trans, int txq_id)
238{
239	struct iwl_txq *txq = trans->txqs.txq[txq_id];
240	struct device *dev = trans->dev;
241	int i;
242
243	if (WARN_ON(!txq))
244		return;
245
246	iwl_pcie_txq_unmap(trans, txq_id);
247
248	/* De-alloc array of command/tx buffers */
249	if (txq_id == trans->txqs.cmd.q_id)
250		for (i = 0; i < txq->n_window; i++) {
251			kfree_sensitive(txq->entries[i].cmd);
252			kfree_sensitive(txq->entries[i].free_buf);
253		}
254
255	/* De-alloc circular buffer of TFDs */
256	if (txq->tfds) {
257		dma_free_coherent(dev,
258				  trans->txqs.tfd.size *
259				  trans->trans_cfg->base_params->max_tfd_queue_size,
260				  txq->tfds, txq->dma_addr);
261		txq->dma_addr = 0;
262		txq->tfds = NULL;
263
264		dma_free_coherent(dev,
265				  sizeof(*txq->first_tb_bufs) * txq->n_window,
266				  txq->first_tb_bufs, txq->first_tb_dma);
267	}
268
269	kfree(txq->entries);
270	txq->entries = NULL;
271
272	del_timer_sync(&txq->stuck_timer);
273
274	/* 0-fill queue descriptor structure */
275	memset(txq, 0, sizeof(*txq));
276}
277
278void iwl_pcie_tx_start(struct iwl_trans *trans, u32 scd_base_addr)
279{
280	struct iwl_trans_pcie *trans_pcie = IWL_TRANS_GET_PCIE_TRANS(trans);
281	int nq = trans->trans_cfg->base_params->num_of_queues;
282	int chan;
283	u32 reg_val;
284	int clear_dwords = (SCD_TRANS_TBL_OFFSET_QUEUE(nq) -
285				SCD_CONTEXT_MEM_LOWER_BOUND) / sizeof(u32);
286
287	/* make sure all queue are not stopped/used */
288	memset(trans->txqs.queue_stopped, 0,
289	       sizeof(trans->txqs.queue_stopped));
290	memset(trans->txqs.queue_used, 0, sizeof(trans->txqs.queue_used));
291
292	trans_pcie->scd_base_addr =
293		iwl_read_prph(trans, SCD_SRAM_BASE_ADDR);
294
295	WARN_ON(scd_base_addr != 0 &&
296		scd_base_addr != trans_pcie->scd_base_addr);
297
298	/* reset context data, TX status and translation data */
299	iwl_trans_write_mem(trans, trans_pcie->scd_base_addr +
300				   SCD_CONTEXT_MEM_LOWER_BOUND,
301			    NULL, clear_dwords);
302
303	iwl_write_prph(trans, SCD_DRAM_BASE_ADDR,
304		       trans->txqs.scd_bc_tbls.dma >> 10);
305
306	/* The chain extension of the SCD doesn't work well. This feature is
307	 * enabled by default by the HW, so we need to disable it manually.
308	 */
309	if (trans->trans_cfg->base_params->scd_chain_ext_wa)
310		iwl_write_prph(trans, SCD_CHAINEXT_EN, 0);
311
312	iwl_trans_ac_txq_enable(trans, trans->txqs.cmd.q_id,
313				trans->txqs.cmd.fifo,
314				trans->txqs.cmd.wdg_timeout);
315
316	/* Activate all Tx DMA/FIFO channels */
317	iwl_scd_activate_fifos(trans);
318
319	/* Enable DMA channel */
320	for (chan = 0; chan < FH_TCSR_CHNL_NUM; chan++)
321		iwl_write_direct32(trans, FH_TCSR_CHNL_TX_CONFIG_REG(chan),
322				   FH_TCSR_TX_CONFIG_REG_VAL_DMA_CHNL_ENABLE |
323				   FH_TCSR_TX_CONFIG_REG_VAL_DMA_CREDIT_ENABLE);
324
325	/* Update FH chicken bits */
326	reg_val = iwl_read_direct32(trans, FH_TX_CHICKEN_BITS_REG);
327	iwl_write_direct32(trans, FH_TX_CHICKEN_BITS_REG,
328			   reg_val | FH_TX_CHICKEN_BITS_SCD_AUTO_RETRY_EN);
329
330	/* Enable L1-Active */
331	if (trans->trans_cfg->device_family < IWL_DEVICE_FAMILY_8000)
332		iwl_clear_bits_prph(trans, APMG_PCIDEV_STT_REG,
333				    APMG_PCIDEV_STT_VAL_L1_ACT_DIS);
334}
335
336void iwl_trans_pcie_tx_reset(struct iwl_trans *trans)
337{
338	struct iwl_trans_pcie *trans_pcie = IWL_TRANS_GET_PCIE_TRANS(trans);
339	int txq_id;
340
341	/*
342	 * we should never get here in gen2 trans mode return early to avoid
343	 * having invalid accesses
344	 */
345	if (WARN_ON_ONCE(trans->trans_cfg->gen2))
346		return;
347
348	for (txq_id = 0; txq_id < trans->trans_cfg->base_params->num_of_queues;
349	     txq_id++) {
350		struct iwl_txq *txq = trans->txqs.txq[txq_id];
351		if (trans->trans_cfg->gen2)
352			iwl_write_direct64(trans,
353					   FH_MEM_CBBC_QUEUE(trans, txq_id),
354					   txq->dma_addr);
355		else
356			iwl_write_direct32(trans,
357					   FH_MEM_CBBC_QUEUE(trans, txq_id),
358					   txq->dma_addr >> 8);
359		iwl_pcie_txq_unmap(trans, txq_id);
360		txq->read_ptr = 0;
361		txq->write_ptr = 0;
362	}
363
364	/* Tell NIC where to find the "keep warm" buffer */
365	iwl_write_direct32(trans, FH_KW_MEM_ADDR_REG,
366			   trans_pcie->kw.dma >> 4);
367
368	/*
369	 * Send 0 as the scd_base_addr since the device may have be reset
370	 * while we were in WoWLAN in which case SCD_SRAM_BASE_ADDR will
371	 * contain garbage.
372	 */
373	iwl_pcie_tx_start(trans, 0);
374}
375
376static void iwl_pcie_tx_stop_fh(struct iwl_trans *trans)
377{
378	struct iwl_trans_pcie *trans_pcie = IWL_TRANS_GET_PCIE_TRANS(trans);
379	int ch, ret;
380	u32 mask = 0;
381
382	spin_lock_bh(&trans_pcie->irq_lock);
383
384	if (!iwl_trans_grab_nic_access(trans))
385		goto out;
386
387	/* Stop each Tx DMA channel */
388	for (ch = 0; ch < FH_TCSR_CHNL_NUM; ch++) {
389		iwl_write32(trans, FH_TCSR_CHNL_TX_CONFIG_REG(ch), 0x0);
390		mask |= FH_TSSR_TX_STATUS_REG_MSK_CHNL_IDLE(ch);
391	}
392
393	/* Wait for DMA channels to be idle */
394	ret = iwl_poll_bit(trans, FH_TSSR_TX_STATUS_REG, mask, mask, 5000);
395	if (ret < 0)
396		IWL_ERR(trans,
397			"Failing on timeout while stopping DMA channel %d [0x%08x]\n",
398			ch, iwl_read32(trans, FH_TSSR_TX_STATUS_REG));
399
400	iwl_trans_release_nic_access(trans);
401
402out:
403	spin_unlock_bh(&trans_pcie->irq_lock);
404}
405
406/*
407 * iwl_pcie_tx_stop - Stop all Tx DMA channels
408 */
409int iwl_pcie_tx_stop(struct iwl_trans *trans)
410{
411	struct iwl_trans_pcie *trans_pcie = IWL_TRANS_GET_PCIE_TRANS(trans);
412	int txq_id;
413
414	/* Turn off all Tx DMA fifos */
415	iwl_scd_deactivate_fifos(trans);
416
417	/* Turn off all Tx DMA channels */
418	iwl_pcie_tx_stop_fh(trans);
419
420	/*
421	 * This function can be called before the op_mode disabled the
422	 * queues. This happens when we have an rfkill interrupt.
423	 * Since we stop Tx altogether - mark the queues as stopped.
424	 */
425	memset(trans->txqs.queue_stopped, 0,
426	       sizeof(trans->txqs.queue_stopped));
427	memset(trans->txqs.queue_used, 0, sizeof(trans->txqs.queue_used));
428
429	/* This can happen: start_hw, stop_device */
430	if (!trans_pcie->txq_memory)
431		return 0;
432
433	/* Unmap DMA from host system and free skb's */
434	for (txq_id = 0; txq_id < trans->trans_cfg->base_params->num_of_queues;
435	     txq_id++)
436		iwl_pcie_txq_unmap(trans, txq_id);
437
438	return 0;
439}
440
441/*
442 * iwl_trans_tx_free - Free TXQ Context
443 *
444 * Destroy all TX DMA queues and structures
445 */
446void iwl_pcie_tx_free(struct iwl_trans *trans)
447{
448	int txq_id;
449	struct iwl_trans_pcie *trans_pcie = IWL_TRANS_GET_PCIE_TRANS(trans);
450
451	memset(trans->txqs.queue_used, 0, sizeof(trans->txqs.queue_used));
452
453	/* Tx queues */
454	if (trans_pcie->txq_memory) {
455		for (txq_id = 0;
456		     txq_id < trans->trans_cfg->base_params->num_of_queues;
457		     txq_id++) {
458			iwl_pcie_txq_free(trans, txq_id);
459			trans->txqs.txq[txq_id] = NULL;
460		}
461	}
462
463	kfree(trans_pcie->txq_memory);
464	trans_pcie->txq_memory = NULL;
465
466	iwl_pcie_free_dma_ptr(trans, &trans_pcie->kw);
467
468	iwl_pcie_free_dma_ptr(trans, &trans->txqs.scd_bc_tbls);
469}
470
471/*
472 * iwl_pcie_tx_alloc - allocate TX context
473 * Allocate all Tx DMA structures and initialize them
474 */
475static int iwl_pcie_tx_alloc(struct iwl_trans *trans)
476{
477	int ret;
478	int txq_id, slots_num;
479	struct iwl_trans_pcie *trans_pcie = IWL_TRANS_GET_PCIE_TRANS(trans);
480	u16 bc_tbls_size = trans->trans_cfg->base_params->num_of_queues;
481
482	if (WARN_ON(trans->trans_cfg->device_family >= IWL_DEVICE_FAMILY_AX210))
483		return -EINVAL;
484
485	bc_tbls_size *= sizeof(struct iwlagn_scd_bc_tbl);
486
487	/*It is not allowed to alloc twice, so warn when this happens.
488	 * We cannot rely on the previous allocation, so free and fail */
489	if (WARN_ON(trans_pcie->txq_memory)) {
490		ret = -EINVAL;
491		goto error;
492	}
493
494	ret = iwl_pcie_alloc_dma_ptr(trans, &trans->txqs.scd_bc_tbls,
495				     bc_tbls_size);
496	if (ret) {
497		IWL_ERR(trans, "Scheduler BC Table allocation failed\n");
498		goto error;
499	}
500
501	/* Alloc keep-warm buffer */
502	ret = iwl_pcie_alloc_dma_ptr(trans, &trans_pcie->kw, IWL_KW_SIZE);
503	if (ret) {
504		IWL_ERR(trans, "Keep Warm allocation failed\n");
505		goto error;
506	}
507
508	trans_pcie->txq_memory =
509		kcalloc(trans->trans_cfg->base_params->num_of_queues,
510			sizeof(struct iwl_txq), GFP_KERNEL);
511	if (!trans_pcie->txq_memory) {
512		IWL_ERR(trans, "Not enough memory for txq\n");
513		ret = -ENOMEM;
514		goto error;
515	}
516
517	/* Alloc and init all Tx queues, including the command queue (#4/#9) */
518	for (txq_id = 0; txq_id < trans->trans_cfg->base_params->num_of_queues;
519	     txq_id++) {
520		bool cmd_queue = (txq_id == trans->txqs.cmd.q_id);
521
522		if (cmd_queue)
523			slots_num = max_t(u32, IWL_CMD_QUEUE_SIZE,
524					  trans->cfg->min_txq_size);
525		else
526			slots_num = max_t(u32, IWL_DEFAULT_QUEUE_SIZE,
527					  trans->cfg->min_ba_txq_size);
528		trans->txqs.txq[txq_id] = &trans_pcie->txq_memory[txq_id];
529		ret = iwl_txq_alloc(trans, trans->txqs.txq[txq_id], slots_num,
530				    cmd_queue);
531		if (ret) {
532			IWL_ERR(trans, "Tx %d queue alloc failed\n", txq_id);
533			goto error;
534		}
535		trans->txqs.txq[txq_id]->id = txq_id;
536	}
537
538	return 0;
539
540error:
541	iwl_pcie_tx_free(trans);
542
543	return ret;
544}
545
546int iwl_pcie_tx_init(struct iwl_trans *trans)
547{
548	struct iwl_trans_pcie *trans_pcie = IWL_TRANS_GET_PCIE_TRANS(trans);
549	int ret;
550	int txq_id, slots_num;
551	bool alloc = false;
552
553	if (!trans_pcie->txq_memory) {
554		ret = iwl_pcie_tx_alloc(trans);
555		if (ret)
556			goto error;
557		alloc = true;
558	}
559
560	spin_lock_bh(&trans_pcie->irq_lock);
561
562	/* Turn off all Tx DMA fifos */
563	iwl_scd_deactivate_fifos(trans);
564
565	/* Tell NIC where to find the "keep warm" buffer */
566	iwl_write_direct32(trans, FH_KW_MEM_ADDR_REG,
567			   trans_pcie->kw.dma >> 4);
568
569	spin_unlock_bh(&trans_pcie->irq_lock);
570
571	/* Alloc and init all Tx queues, including the command queue (#4/#9) */
572	for (txq_id = 0; txq_id < trans->trans_cfg->base_params->num_of_queues;
573	     txq_id++) {
574		bool cmd_queue = (txq_id == trans->txqs.cmd.q_id);
575
576		if (cmd_queue)
577			slots_num = max_t(u32, IWL_CMD_QUEUE_SIZE,
578					  trans->cfg->min_txq_size);
579		else
580			slots_num = max_t(u32, IWL_DEFAULT_QUEUE_SIZE,
581					  trans->cfg->min_ba_txq_size);
582		ret = iwl_txq_init(trans, trans->txqs.txq[txq_id], slots_num,
583				   cmd_queue);
584		if (ret) {
585			IWL_ERR(trans, "Tx %d queue init failed\n", txq_id);
586			goto error;
587		}
588
589		/*
590		 * Tell nic where to find circular buffer of TFDs for a
591		 * given Tx queue, and enable the DMA channel used for that
592		 * queue.
593		 * Circular buffer (TFD queue in DRAM) physical base address
594		 */
595		iwl_write_direct32(trans, FH_MEM_CBBC_QUEUE(trans, txq_id),
596				   trans->txqs.txq[txq_id]->dma_addr >> 8);
597	}
598
599	iwl_set_bits_prph(trans, SCD_GP_CTRL, SCD_GP_CTRL_AUTO_ACTIVE_MODE);
600	if (trans->trans_cfg->base_params->num_of_queues > 20)
601		iwl_set_bits_prph(trans, SCD_GP_CTRL,
602				  SCD_GP_CTRL_ENABLE_31_QUEUES);
603
604	return 0;
605error:
606	/*Upon error, free only if we allocated something */
607	if (alloc)
608		iwl_pcie_tx_free(trans);
609	return ret;
610}
611
612static int iwl_pcie_set_cmd_in_flight(struct iwl_trans *trans,
613				      const struct iwl_host_cmd *cmd)
614{
615	struct iwl_trans_pcie *trans_pcie = IWL_TRANS_GET_PCIE_TRANS(trans);
616
617	/* Make sure the NIC is still alive in the bus */
618	if (test_bit(STATUS_TRANS_DEAD, &trans->status))
619		return -ENODEV;
620
621	if (!trans->trans_cfg->base_params->apmg_wake_up_wa)
622		return 0;
623
624	/*
625	 * wake up the NIC to make sure that the firmware will see the host
626	 * command - we will let the NIC sleep once all the host commands
627	 * returned. This needs to be done only on NICs that have
628	 * apmg_wake_up_wa set (see above.)
629	 */
630	if (!_iwl_trans_pcie_grab_nic_access(trans))
631		return -EIO;
632
633	/*
634	 * In iwl_trans_grab_nic_access(), we've acquired the reg_lock.
635	 * There, we also returned immediately if cmd_hold_nic_awake is
636	 * already true, so it's OK to unconditionally set it to true.
637	 */
638	trans_pcie->cmd_hold_nic_awake = true;
639	spin_unlock(&trans_pcie->reg_lock);
640
641	return 0;
642}
643
644/*
645 * iwl_pcie_cmdq_reclaim - Reclaim TX command queue entries already Tx'd
646 *
647 * When FW advances 'R' index, all entries between old and new 'R' index
648 * need to be reclaimed. As result, some free space forms.  If there is
649 * enough free space (> low mark), wake the stack that feeds us.
650 */
651static void iwl_pcie_cmdq_reclaim(struct iwl_trans *trans, int txq_id, int idx)
652{
653	struct iwl_txq *txq = trans->txqs.txq[txq_id];
654	int nfreed = 0;
655	u16 r;
656
657	lockdep_assert_held(&txq->lock);
658
659	idx = iwl_txq_get_cmd_index(txq, idx);
660	r = iwl_txq_get_cmd_index(txq, txq->read_ptr);
661
662	if (idx >= trans->trans_cfg->base_params->max_tfd_queue_size ||
663	    (!iwl_txq_used(txq, idx))) {
664		WARN_ONCE(test_bit(txq_id, trans->txqs.queue_used),
665			  "%s: Read index for DMA queue txq id (%d), index %d is out of range [0-%d] %d %d.\n",
666			  __func__, txq_id, idx,
667			  trans->trans_cfg->base_params->max_tfd_queue_size,
668			  txq->write_ptr, txq->read_ptr);
669		return;
670	}
671
672	for (idx = iwl_txq_inc_wrap(trans, idx); r != idx;
673	     r = iwl_txq_inc_wrap(trans, r)) {
674		txq->read_ptr = iwl_txq_inc_wrap(trans, txq->read_ptr);
675
676		if (nfreed++ > 0) {
677			IWL_ERR(trans, "HCMD skipped: index (%d) %d %d\n",
678				idx, txq->write_ptr, r);
679			iwl_force_nmi(trans);
680		}
681	}
682
683	if (txq->read_ptr == txq->write_ptr)
684		iwl_pcie_clear_cmd_in_flight(trans);
685
686	iwl_txq_progress(txq);
687}
688
689static int iwl_pcie_txq_set_ratid_map(struct iwl_trans *trans, u16 ra_tid,
690				 u16 txq_id)
691{
692	struct iwl_trans_pcie *trans_pcie = IWL_TRANS_GET_PCIE_TRANS(trans);
693	u32 tbl_dw_addr;
694	u32 tbl_dw;
695	u16 scd_q2ratid;
696
697	scd_q2ratid = ra_tid & SCD_QUEUE_RA_TID_MAP_RATID_MSK;
698
699	tbl_dw_addr = trans_pcie->scd_base_addr +
700			SCD_TRANS_TBL_OFFSET_QUEUE(txq_id);
701
702	tbl_dw = iwl_trans_read_mem32(trans, tbl_dw_addr);
703
704	if (txq_id & 0x1)
705		tbl_dw = (scd_q2ratid << 16) | (tbl_dw & 0x0000FFFF);
706	else
707		tbl_dw = scd_q2ratid | (tbl_dw & 0xFFFF0000);
708
709	iwl_trans_write_mem32(trans, tbl_dw_addr, tbl_dw);
710
711	return 0;
712}
713
714/* Receiver address (actually, Rx station's index into station table),
715 * combined with Traffic ID (QOS priority), in format used by Tx Scheduler */
716#define BUILD_RAxTID(sta_id, tid)	(((sta_id) << 4) + (tid))
717
718bool iwl_trans_pcie_txq_enable(struct iwl_trans *trans, int txq_id, u16 ssn,
719			       const struct iwl_trans_txq_scd_cfg *cfg,
720			       unsigned int wdg_timeout)
721{
722	struct iwl_trans_pcie *trans_pcie = IWL_TRANS_GET_PCIE_TRANS(trans);
723	struct iwl_txq *txq = trans->txqs.txq[txq_id];
724	int fifo = -1;
725	bool scd_bug = false;
726
727	if (test_and_set_bit(txq_id, trans->txqs.queue_used))
728		WARN_ONCE(1, "queue %d already used - expect issues", txq_id);
729
730	txq->wd_timeout = msecs_to_jiffies(wdg_timeout);
731
732	if (cfg) {
733		fifo = cfg->fifo;
734
735		/* Disable the scheduler prior configuring the cmd queue */
736		if (txq_id == trans->txqs.cmd.q_id &&
737		    trans_pcie->scd_set_active)
738			iwl_scd_enable_set_active(trans, 0);
739
740		/* Stop this Tx queue before configuring it */
741		iwl_scd_txq_set_inactive(trans, txq_id);
742
743		/* Set this queue as a chain-building queue unless it is CMD */
744		if (txq_id != trans->txqs.cmd.q_id)
745			iwl_scd_txq_set_chain(trans, txq_id);
746
747		if (cfg->aggregate) {
748			u16 ra_tid = BUILD_RAxTID(cfg->sta_id, cfg->tid);
749
750			/* Map receiver-address / traffic-ID to this queue */
751			iwl_pcie_txq_set_ratid_map(trans, ra_tid, txq_id);
752
753			/* enable aggregations for the queue */
754			iwl_scd_txq_enable_agg(trans, txq_id);
755			txq->ampdu = true;
756		} else {
757			/*
758			 * disable aggregations for the queue, this will also
759			 * make the ra_tid mapping configuration irrelevant
760			 * since it is now a non-AGG queue.
761			 */
762			iwl_scd_txq_disable_agg(trans, txq_id);
763
764			ssn = txq->read_ptr;
765		}
766	} else {
767		/*
768		 * If we need to move the SCD write pointer by steps of
769		 * 0x40, 0x80 or 0xc0, it gets stuck. Avoids this and let
770		 * the op_mode know by returning true later.
771		 * Do this only in case cfg is NULL since this trick can
772		 * be done only if we have DQA enabled which is true for mvm
773		 * only. And mvm never sets a cfg pointer.
774		 * This is really ugly, but this is the easiest way out for
775		 * this sad hardware issue.
776		 * This bug has been fixed on devices 9000 and up.
777		 */
778		scd_bug = !trans->trans_cfg->mq_rx_supported &&
779			!((ssn - txq->write_ptr) & 0x3f) &&
780			(ssn != txq->write_ptr);
781		if (scd_bug)
782			ssn++;
783	}
784
785	/* Place first TFD at index corresponding to start sequence number.
786	 * Assumes that ssn_idx is valid (!= 0xFFF) */
787	txq->read_ptr = (ssn & 0xff);
788	txq->write_ptr = (ssn & 0xff);
789	iwl_write_direct32(trans, HBUS_TARG_WRPTR,
790			   (ssn & 0xff) | (txq_id << 8));
791
792	if (cfg) {
793		u8 frame_limit = cfg->frame_limit;
794
795		iwl_write_prph(trans, SCD_QUEUE_RDPTR(txq_id), ssn);
796
797		/* Set up Tx window size and frame limit for this queue */
798		iwl_trans_write_mem32(trans, trans_pcie->scd_base_addr +
799				SCD_CONTEXT_QUEUE_OFFSET(txq_id), 0);
800		iwl_trans_write_mem32(trans,
801			trans_pcie->scd_base_addr +
802			SCD_CONTEXT_QUEUE_OFFSET(txq_id) + sizeof(u32),
803			SCD_QUEUE_CTX_REG2_VAL(WIN_SIZE, frame_limit) |
804			SCD_QUEUE_CTX_REG2_VAL(FRAME_LIMIT, frame_limit));
805
806		/* Set up status area in SRAM, map to Tx DMA/FIFO, activate */
807		iwl_write_prph(trans, SCD_QUEUE_STATUS_BITS(txq_id),
808			       (1 << SCD_QUEUE_STTS_REG_POS_ACTIVE) |
809			       (cfg->fifo << SCD_QUEUE_STTS_REG_POS_TXF) |
810			       (1 << SCD_QUEUE_STTS_REG_POS_WSL) |
811			       SCD_QUEUE_STTS_REG_MSK);
812
813		/* enable the scheduler for this queue (only) */
814		if (txq_id == trans->txqs.cmd.q_id &&
815		    trans_pcie->scd_set_active)
816			iwl_scd_enable_set_active(trans, BIT(txq_id));
817
818		IWL_DEBUG_TX_QUEUES(trans,
819				    "Activate queue %d on FIFO %d WrPtr: %d\n",
820				    txq_id, fifo, ssn & 0xff);
821	} else {
822		IWL_DEBUG_TX_QUEUES(trans,
823				    "Activate queue %d WrPtr: %d\n",
824				    txq_id, ssn & 0xff);
825	}
826
827	return scd_bug;
828}
829
830void iwl_trans_pcie_txq_set_shared_mode(struct iwl_trans *trans, u32 txq_id,
831					bool shared_mode)
832{
833	struct iwl_txq *txq = trans->txqs.txq[txq_id];
834
835	txq->ampdu = !shared_mode;
836}
837
838void iwl_trans_pcie_txq_disable(struct iwl_trans *trans, int txq_id,
839				bool configure_scd)
840{
841	struct iwl_trans_pcie *trans_pcie = IWL_TRANS_GET_PCIE_TRANS(trans);
842	u32 stts_addr = trans_pcie->scd_base_addr +
843			SCD_TX_STTS_QUEUE_OFFSET(txq_id);
844	static const u32 zero_val[4] = {};
845
846	trans->txqs.txq[txq_id]->frozen_expiry_remainder = 0;
847	trans->txqs.txq[txq_id]->frozen = false;
848
849	/*
850	 * Upon HW Rfkill - we stop the device, and then stop the queues
851	 * in the op_mode. Just for the sake of the simplicity of the op_mode,
852	 * allow the op_mode to call txq_disable after it already called
853	 * stop_device.
854	 */
855	if (!test_and_clear_bit(txq_id, trans->txqs.queue_used)) {
856		WARN_ONCE(test_bit(STATUS_DEVICE_ENABLED, &trans->status),
857			  "queue %d not used", txq_id);
858		return;
859	}
860
861	if (configure_scd) {
862		iwl_scd_txq_set_inactive(trans, txq_id);
863
864		iwl_trans_write_mem(trans, stts_addr, (const void *)zero_val,
865				    ARRAY_SIZE(zero_val));
866	}
867
868	iwl_pcie_txq_unmap(trans, txq_id);
869	trans->txqs.txq[txq_id]->ampdu = false;
870
871	IWL_DEBUG_TX_QUEUES(trans, "Deactivate queue %d\n", txq_id);
872}
873
874/*************** HOST COMMAND QUEUE FUNCTIONS   *****/
875
876/*
877 * iwl_pcie_enqueue_hcmd - enqueue a uCode command
878 * @priv: device private data point
879 * @cmd: a pointer to the ucode command structure
880 *
881 * The function returns < 0 values to indicate the operation
882 * failed. On success, it returns the index (>= 0) of command in the
883 * command queue.
884 */
885int iwl_pcie_enqueue_hcmd(struct iwl_trans *trans,
886			  struct iwl_host_cmd *cmd)
887{
888	struct iwl_txq *txq = trans->txqs.txq[trans->txqs.cmd.q_id];
889	struct iwl_device_cmd *out_cmd;
890	struct iwl_cmd_meta *out_meta;
891	void *dup_buf = NULL;
892	dma_addr_t phys_addr;
893	int idx;
894	u16 copy_size, cmd_size, tb0_size;
895	bool had_nocopy = false;
896	u8 group_id = iwl_cmd_groupid(cmd->id);
897	int i, ret;
898	u32 cmd_pos;
899	const u8 *cmddata[IWL_MAX_CMD_TBS_PER_TFD];
900	u16 cmdlen[IWL_MAX_CMD_TBS_PER_TFD];
901	unsigned long flags;
902
903	if (WARN(!trans->wide_cmd_header &&
904		 group_id > IWL_ALWAYS_LONG_GROUP,
905		 "unsupported wide command %#x\n", cmd->id))
906		return -EINVAL;
907
908	if (group_id != 0) {
909		copy_size = sizeof(struct iwl_cmd_header_wide);
910		cmd_size = sizeof(struct iwl_cmd_header_wide);
911	} else {
912		copy_size = sizeof(struct iwl_cmd_header);
913		cmd_size = sizeof(struct iwl_cmd_header);
914	}
915
916	/* need one for the header if the first is NOCOPY */
917	BUILD_BUG_ON(IWL_MAX_CMD_TBS_PER_TFD > IWL_NUM_OF_TBS - 1);
918
919	for (i = 0; i < IWL_MAX_CMD_TBS_PER_TFD; i++) {
920		cmddata[i] = cmd->data[i];
921		cmdlen[i] = cmd->len[i];
922
923		if (!cmd->len[i])
924			continue;
925
926		/* need at least IWL_FIRST_TB_SIZE copied */
927		if (copy_size < IWL_FIRST_TB_SIZE) {
928			int copy = IWL_FIRST_TB_SIZE - copy_size;
929
930			if (copy > cmdlen[i])
931				copy = cmdlen[i];
932			cmdlen[i] -= copy;
933			cmddata[i] += copy;
934			copy_size += copy;
935		}
936
937		if (cmd->dataflags[i] & IWL_HCMD_DFL_NOCOPY) {
938			had_nocopy = true;
939			if (WARN_ON(cmd->dataflags[i] & IWL_HCMD_DFL_DUP)) {
940				idx = -EINVAL;
941				goto free_dup_buf;
942			}
943		} else if (cmd->dataflags[i] & IWL_HCMD_DFL_DUP) {
944			/*
945			 * This is also a chunk that isn't copied
946			 * to the static buffer so set had_nocopy.
947			 */
948			had_nocopy = true;
949
950			/* only allowed once */
951			if (WARN_ON(dup_buf)) {
952				idx = -EINVAL;
953				goto free_dup_buf;
954			}
955
956			dup_buf = kmemdup(cmddata[i], cmdlen[i],
957					  GFP_ATOMIC);
958			if (!dup_buf)
959				return -ENOMEM;
960		} else {
961			/* NOCOPY must not be followed by normal! */
962			if (WARN_ON(had_nocopy)) {
963				idx = -EINVAL;
964				goto free_dup_buf;
965			}
966			copy_size += cmdlen[i];
967		}
968		cmd_size += cmd->len[i];
969	}
970
971	/*
972	 * If any of the command structures end up being larger than
973	 * the TFD_MAX_PAYLOAD_SIZE and they aren't dynamically
974	 * allocated into separate TFDs, then we will need to
975	 * increase the size of the buffers.
976	 */
977	if (WARN(copy_size > TFD_MAX_PAYLOAD_SIZE,
978		 "Command %s (%#x) is too large (%d bytes)\n",
979		 iwl_get_cmd_string(trans, cmd->id),
980		 cmd->id, copy_size)) {
981		idx = -EINVAL;
982		goto free_dup_buf;
983	}
984
985	spin_lock_irqsave(&txq->lock, flags);
986
987	if (iwl_txq_space(trans, txq) < ((cmd->flags & CMD_ASYNC) ? 2 : 1)) {
988		spin_unlock_irqrestore(&txq->lock, flags);
989
990		IWL_ERR(trans, "No space in command queue\n");
991		iwl_op_mode_cmd_queue_full(trans->op_mode);
992		idx = -ENOSPC;
993		goto free_dup_buf;
994	}
995
996	idx = iwl_txq_get_cmd_index(txq, txq->write_ptr);
997	out_cmd = txq->entries[idx].cmd;
998	out_meta = &txq->entries[idx].meta;
999
1000	memset(out_meta, 0, sizeof(*out_meta));	/* re-initialize to NULL */
1001	if (cmd->flags & CMD_WANT_SKB)
1002		out_meta->source = cmd;
1003
1004	/* set up the header */
1005	if (group_id != 0) {
1006		out_cmd->hdr_wide.cmd = iwl_cmd_opcode(cmd->id);
1007		out_cmd->hdr_wide.group_id = group_id;
1008		out_cmd->hdr_wide.version = iwl_cmd_version(cmd->id);
1009		out_cmd->hdr_wide.length =
1010			cpu_to_le16(cmd_size -
1011				    sizeof(struct iwl_cmd_header_wide));
1012		out_cmd->hdr_wide.reserved = 0;
1013		out_cmd->hdr_wide.sequence =
1014			cpu_to_le16(QUEUE_TO_SEQ(trans->txqs.cmd.q_id) |
1015						 INDEX_TO_SEQ(txq->write_ptr));
1016
1017		cmd_pos = sizeof(struct iwl_cmd_header_wide);
1018		copy_size = sizeof(struct iwl_cmd_header_wide);
1019	} else {
1020		out_cmd->hdr.cmd = iwl_cmd_opcode(cmd->id);
1021		out_cmd->hdr.sequence =
1022			cpu_to_le16(QUEUE_TO_SEQ(trans->txqs.cmd.q_id) |
1023						 INDEX_TO_SEQ(txq->write_ptr));
1024		out_cmd->hdr.group_id = 0;
1025
1026		cmd_pos = sizeof(struct iwl_cmd_header);
1027		copy_size = sizeof(struct iwl_cmd_header);
1028	}
1029
1030	/* and copy the data that needs to be copied */
1031	for (i = 0; i < IWL_MAX_CMD_TBS_PER_TFD; i++) {
1032		int copy;
1033
1034		if (!cmd->len[i])
1035			continue;
1036
1037		/* copy everything if not nocopy/dup */
1038		if (!(cmd->dataflags[i] & (IWL_HCMD_DFL_NOCOPY |
1039					   IWL_HCMD_DFL_DUP))) {
1040			copy = cmd->len[i];
1041
1042			memcpy((u8 *)out_cmd + cmd_pos, cmd->data[i], copy);
1043			cmd_pos += copy;
1044			copy_size += copy;
1045			continue;
1046		}
1047
1048		/*
1049		 * Otherwise we need at least IWL_FIRST_TB_SIZE copied
1050		 * in total (for bi-directional DMA), but copy up to what
1051		 * we can fit into the payload for debug dump purposes.
1052		 */
1053		copy = min_t(int, TFD_MAX_PAYLOAD_SIZE - cmd_pos, cmd->len[i]);
1054
1055		memcpy((u8 *)out_cmd + cmd_pos, cmd->data[i], copy);
1056		cmd_pos += copy;
1057
1058		/* However, treat copy_size the proper way, we need it below */
1059		if (copy_size < IWL_FIRST_TB_SIZE) {
1060			copy = IWL_FIRST_TB_SIZE - copy_size;
1061
1062			if (copy > cmd->len[i])
1063				copy = cmd->len[i];
1064			copy_size += copy;
1065		}
1066	}
1067
1068	IWL_DEBUG_HC(trans,
1069		     "Sending command %s (%.2x.%.2x), seq: 0x%04X, %d bytes at %d[%d]:%d\n",
1070		     iwl_get_cmd_string(trans, cmd->id),
1071		     group_id, out_cmd->hdr.cmd,
1072		     le16_to_cpu(out_cmd->hdr.sequence),
1073		     cmd_size, txq->write_ptr, idx, trans->txqs.cmd.q_id);
1074
1075	/* start the TFD with the minimum copy bytes */
1076	tb0_size = min_t(int, copy_size, IWL_FIRST_TB_SIZE);
1077	memcpy(&txq->first_tb_bufs[idx], &out_cmd->hdr, tb0_size);
1078	iwl_pcie_txq_build_tfd(trans, txq,
1079			       iwl_txq_get_first_tb_dma(txq, idx),
1080			       tb0_size, true);
1081
1082	/* map first command fragment, if any remains */
1083	if (copy_size > tb0_size) {
1084		phys_addr = dma_map_single(trans->dev,
1085					   ((u8 *)&out_cmd->hdr) + tb0_size,
1086					   copy_size - tb0_size,
1087					   DMA_TO_DEVICE);
1088		if (dma_mapping_error(trans->dev, phys_addr)) {
1089			iwl_txq_gen1_tfd_unmap(trans, out_meta, txq,
1090					       txq->write_ptr);
1091			idx = -ENOMEM;
1092			goto out;
1093		}
1094
1095		iwl_pcie_txq_build_tfd(trans, txq, phys_addr,
1096				       copy_size - tb0_size, false);
1097	}
1098
1099	/* map the remaining (adjusted) nocopy/dup fragments */
1100	for (i = 0; i < IWL_MAX_CMD_TBS_PER_TFD; i++) {
1101		void *data = (void *)(uintptr_t)cmddata[i];
1102
1103		if (!cmdlen[i])
1104			continue;
1105		if (!(cmd->dataflags[i] & (IWL_HCMD_DFL_NOCOPY |
1106					   IWL_HCMD_DFL_DUP)))
1107			continue;
1108		if (cmd->dataflags[i] & IWL_HCMD_DFL_DUP)
1109			data = dup_buf;
1110		phys_addr = dma_map_single(trans->dev, data,
1111					   cmdlen[i], DMA_TO_DEVICE);
1112		if (dma_mapping_error(trans->dev, phys_addr)) {
1113			iwl_txq_gen1_tfd_unmap(trans, out_meta, txq,
1114					       txq->write_ptr);
1115			idx = -ENOMEM;
1116			goto out;
1117		}
1118
1119		iwl_pcie_txq_build_tfd(trans, txq, phys_addr, cmdlen[i], false);
1120	}
1121
1122	BUILD_BUG_ON(IWL_TFH_NUM_TBS > sizeof(out_meta->tbs) * BITS_PER_BYTE);
1123	out_meta->flags = cmd->flags;
1124	if (WARN_ON_ONCE(txq->entries[idx].free_buf))
1125		kfree_sensitive(txq->entries[idx].free_buf);
1126	txq->entries[idx].free_buf = dup_buf;
1127
1128	trace_iwlwifi_dev_hcmd(trans->dev, cmd, cmd_size, &out_cmd->hdr_wide);
1129
1130	/* start timer if queue currently empty */
1131	if (txq->read_ptr == txq->write_ptr && txq->wd_timeout)
1132		mod_timer(&txq->stuck_timer, jiffies + txq->wd_timeout);
1133
1134	ret = iwl_pcie_set_cmd_in_flight(trans, cmd);
1135	if (ret < 0) {
1136		idx = ret;
1137		goto out;
1138	}
1139
1140	/* Increment and update queue's write index */
1141	txq->write_ptr = iwl_txq_inc_wrap(trans, txq->write_ptr);
1142	iwl_pcie_txq_inc_wr_ptr(trans, txq);
1143
1144 out:
1145	spin_unlock_irqrestore(&txq->lock, flags);
1146 free_dup_buf:
1147	if (idx < 0)
1148		kfree(dup_buf);
1149	return idx;
1150}
1151
1152/*
1153 * iwl_pcie_hcmd_complete - Pull unused buffers off the queue and reclaim them
1154 * @rxb: Rx buffer to reclaim
1155 */
1156void iwl_pcie_hcmd_complete(struct iwl_trans *trans,
1157			    struct iwl_rx_cmd_buffer *rxb)
1158{
1159	struct iwl_rx_packet *pkt = rxb_addr(rxb);
1160	u16 sequence = le16_to_cpu(pkt->hdr.sequence);
1161	u8 group_id;
1162	u32 cmd_id;
1163	int txq_id = SEQ_TO_QUEUE(sequence);
1164	int index = SEQ_TO_INDEX(sequence);
1165	int cmd_index;
1166	struct iwl_device_cmd *cmd;
1167	struct iwl_cmd_meta *meta;
1168	struct iwl_trans_pcie *trans_pcie = IWL_TRANS_GET_PCIE_TRANS(trans);
1169	struct iwl_txq *txq = trans->txqs.txq[trans->txqs.cmd.q_id];
1170
1171	/* If a Tx command is being handled and it isn't in the actual
1172	 * command queue then there a command routing bug has been introduced
1173	 * in the queue management code. */
1174	if (WARN(txq_id != trans->txqs.cmd.q_id,
1175		 "wrong command queue %d (should be %d), sequence 0x%X readp=%d writep=%d\n",
1176		 txq_id, trans->txqs.cmd.q_id, sequence, txq->read_ptr,
1177		 txq->write_ptr)) {
1178		iwl_print_hex_error(trans, pkt, 32);
1179		return;
1180	}
1181
1182	spin_lock_bh(&txq->lock);
1183
1184	cmd_index = iwl_txq_get_cmd_index(txq, index);
1185	cmd = txq->entries[cmd_index].cmd;
1186	meta = &txq->entries[cmd_index].meta;
1187	group_id = cmd->hdr.group_id;
1188	cmd_id = WIDE_ID(group_id, cmd->hdr.cmd);
1189
1190	if (trans->trans_cfg->gen2)
1191		iwl_txq_gen2_tfd_unmap(trans, meta,
1192				       iwl_txq_get_tfd(trans, txq, index));
1193	else
1194		iwl_txq_gen1_tfd_unmap(trans, meta, txq, index);
1195
1196	/* Input error checking is done when commands are added to queue. */
1197	if (meta->flags & CMD_WANT_SKB) {
1198		struct page *p = rxb_steal_page(rxb);
1199
1200		meta->source->resp_pkt = pkt;
1201		meta->source->_rx_page_addr = (unsigned long)page_address(p);
1202		meta->source->_rx_page_order = trans_pcie->rx_page_order;
1203	}
1204
1205	if (meta->flags & CMD_WANT_ASYNC_CALLBACK)
1206		iwl_op_mode_async_cb(trans->op_mode, cmd);
1207
1208	iwl_pcie_cmdq_reclaim(trans, txq_id, index);
1209
1210	if (!(meta->flags & CMD_ASYNC)) {
1211		if (!test_bit(STATUS_SYNC_HCMD_ACTIVE, &trans->status)) {
1212			IWL_WARN(trans,
1213				 "HCMD_ACTIVE already clear for command %s\n",
1214				 iwl_get_cmd_string(trans, cmd_id));
1215		}
1216		clear_bit(STATUS_SYNC_HCMD_ACTIVE, &trans->status);
1217		IWL_DEBUG_INFO(trans, "Clearing HCMD_ACTIVE for command %s\n",
1218			       iwl_get_cmd_string(trans, cmd_id));
1219		wake_up(&trans->wait_command_queue);
1220	}
1221
1222	meta->flags = 0;
1223
1224	spin_unlock_bh(&txq->lock);
1225}
1226
1227static int iwl_fill_data_tbs(struct iwl_trans *trans, struct sk_buff *skb,
1228			     struct iwl_txq *txq, u8 hdr_len,
1229			     struct iwl_cmd_meta *out_meta)
1230{
1231	u16 head_tb_len;
1232	int i;
1233
1234	/*
1235	 * Set up TFD's third entry to point directly to remainder
1236	 * of skb's head, if any
1237	 */
1238	head_tb_len = skb_headlen(skb) - hdr_len;
1239
1240	if (head_tb_len > 0) {
1241		dma_addr_t tb_phys = dma_map_single(trans->dev,
1242						    skb->data + hdr_len,
1243						    head_tb_len, DMA_TO_DEVICE);
1244		if (unlikely(dma_mapping_error(trans->dev, tb_phys)))
1245			return -EINVAL;
1246		trace_iwlwifi_dev_tx_tb(trans->dev, skb, skb->data + hdr_len,
1247					tb_phys, head_tb_len);
1248		iwl_pcie_txq_build_tfd(trans, txq, tb_phys, head_tb_len, false);
1249	}
1250
1251	/* set up the remaining entries to point to the data */
1252	for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1253		const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1254		dma_addr_t tb_phys;
1255		int tb_idx;
1256
1257		if (!skb_frag_size(frag))
1258			continue;
1259
1260		tb_phys = skb_frag_dma_map(trans->dev, frag, 0,
1261					   skb_frag_size(frag), DMA_TO_DEVICE);
1262
1263		if (unlikely(dma_mapping_error(trans->dev, tb_phys)))
1264			return -EINVAL;
1265		trace_iwlwifi_dev_tx_tb(trans->dev, skb, skb_frag_address(frag),
1266					tb_phys, skb_frag_size(frag));
1267		tb_idx = iwl_pcie_txq_build_tfd(trans, txq, tb_phys,
1268						skb_frag_size(frag), false);
1269		if (tb_idx < 0)
1270			return tb_idx;
1271
1272		out_meta->tbs |= BIT(tb_idx);
1273	}
1274
1275	return 0;
1276}
1277
1278#ifdef CONFIG_INET
1279static int iwl_fill_data_tbs_amsdu(struct iwl_trans *trans, struct sk_buff *skb,
1280				   struct iwl_txq *txq, u8 hdr_len,
1281				   struct iwl_cmd_meta *out_meta,
1282				   struct iwl_device_tx_cmd *dev_cmd,
1283				   u16 tb1_len)
1284{
1285	struct iwl_tx_cmd *tx_cmd = (void *)dev_cmd->payload;
1286	struct ieee80211_hdr *hdr = (void *)skb->data;
1287	unsigned int snap_ip_tcp_hdrlen, ip_hdrlen, total_len, hdr_room;
1288	unsigned int mss = skb_shinfo(skb)->gso_size;
1289	u16 length, iv_len, amsdu_pad;
1290	u8 *start_hdr;
1291	struct iwl_tso_hdr_page *hdr_page;
1292	struct tso_t tso;
1293
1294	/* if the packet is protected, then it must be CCMP or GCMP */
1295	BUILD_BUG_ON(IEEE80211_CCMP_HDR_LEN != IEEE80211_GCMP_HDR_LEN);
1296	iv_len = ieee80211_has_protected(hdr->frame_control) ?
1297		IEEE80211_CCMP_HDR_LEN : 0;
1298
1299	trace_iwlwifi_dev_tx(trans->dev, skb,
1300			     iwl_txq_get_tfd(trans, txq, txq->write_ptr),
1301			     trans->txqs.tfd.size,
1302			     &dev_cmd->hdr, IWL_FIRST_TB_SIZE + tb1_len, 0);
1303
1304	ip_hdrlen = skb_transport_header(skb) - skb_network_header(skb);
1305	snap_ip_tcp_hdrlen = 8 + ip_hdrlen + tcp_hdrlen(skb);
1306	total_len = skb->len - snap_ip_tcp_hdrlen - hdr_len - iv_len;
1307	amsdu_pad = 0;
1308
1309	/* total amount of header we may need for this A-MSDU */
1310	hdr_room = DIV_ROUND_UP(total_len, mss) *
1311		(3 + snap_ip_tcp_hdrlen + sizeof(struct ethhdr)) + iv_len;
1312
1313	/* Our device supports 9 segments at most, it will fit in 1 page */
1314	hdr_page = get_page_hdr(trans, hdr_room, skb);
1315	if (!hdr_page)
1316		return -ENOMEM;
1317
1318	start_hdr = hdr_page->pos;
1319	memcpy(hdr_page->pos, skb->data + hdr_len, iv_len);
1320	hdr_page->pos += iv_len;
1321
1322	/*
1323	 * Pull the ieee80211 header + IV to be able to use TSO core,
1324	 * we will restore it for the tx_status flow.
1325	 */
1326	skb_pull(skb, hdr_len + iv_len);
1327
1328	/*
1329	 * Remove the length of all the headers that we don't actually
1330	 * have in the MPDU by themselves, but that we duplicate into
1331	 * all the different MSDUs inside the A-MSDU.
1332	 */
1333	le16_add_cpu(&tx_cmd->len, -snap_ip_tcp_hdrlen);
1334
1335	tso_start(skb, &tso);
1336
1337	while (total_len) {
1338		/* this is the data left for this subframe */
1339		unsigned int data_left =
1340			min_t(unsigned int, mss, total_len);
1341		unsigned int hdr_tb_len;
1342		dma_addr_t hdr_tb_phys;
1343		u8 *subf_hdrs_start = hdr_page->pos;
1344
1345		total_len -= data_left;
1346
1347		memset(hdr_page->pos, 0, amsdu_pad);
1348		hdr_page->pos += amsdu_pad;
1349		amsdu_pad = (4 - (sizeof(struct ethhdr) + snap_ip_tcp_hdrlen +
1350				  data_left)) & 0x3;
1351		ether_addr_copy(hdr_page->pos, ieee80211_get_DA(hdr));
1352		hdr_page->pos += ETH_ALEN;
1353		ether_addr_copy(hdr_page->pos, ieee80211_get_SA(hdr));
1354		hdr_page->pos += ETH_ALEN;
1355
1356		length = snap_ip_tcp_hdrlen + data_left;
1357		*((__be16 *)hdr_page->pos) = cpu_to_be16(length);
1358		hdr_page->pos += sizeof(length);
1359
1360		/*
1361		 * This will copy the SNAP as well which will be considered
1362		 * as MAC header.
1363		 */
1364		tso_build_hdr(skb, hdr_page->pos, &tso, data_left, !total_len);
1365
1366		hdr_page->pos += snap_ip_tcp_hdrlen;
1367
1368		hdr_tb_len = hdr_page->pos - start_hdr;
1369		hdr_tb_phys = dma_map_single(trans->dev, start_hdr,
1370					     hdr_tb_len, DMA_TO_DEVICE);
1371		if (unlikely(dma_mapping_error(trans->dev, hdr_tb_phys)))
1372			return -EINVAL;
1373		iwl_pcie_txq_build_tfd(trans, txq, hdr_tb_phys,
1374				       hdr_tb_len, false);
1375		trace_iwlwifi_dev_tx_tb(trans->dev, skb, start_hdr,
1376					hdr_tb_phys, hdr_tb_len);
1377		/* add this subframe's headers' length to the tx_cmd */
1378		le16_add_cpu(&tx_cmd->len, hdr_page->pos - subf_hdrs_start);
1379
1380		/* prepare the start_hdr for the next subframe */
1381		start_hdr = hdr_page->pos;
1382
1383		/* put the payload */
1384		while (data_left) {
1385			unsigned int size = min_t(unsigned int, tso.size,
1386						  data_left);
1387			dma_addr_t tb_phys;
1388
1389			tb_phys = dma_map_single(trans->dev, tso.data,
1390						 size, DMA_TO_DEVICE);
1391			if (unlikely(dma_mapping_error(trans->dev, tb_phys)))
1392				return -EINVAL;
1393
1394			iwl_pcie_txq_build_tfd(trans, txq, tb_phys,
1395					       size, false);
1396			trace_iwlwifi_dev_tx_tb(trans->dev, skb, tso.data,
1397						tb_phys, size);
1398
1399			data_left -= size;
1400			tso_build_data(skb, &tso, size);
1401		}
1402	}
1403
1404	/* re -add the WiFi header and IV */
1405	skb_push(skb, hdr_len + iv_len);
1406
1407	return 0;
1408}
1409#else /* CONFIG_INET */
1410static int iwl_fill_data_tbs_amsdu(struct iwl_trans *trans, struct sk_buff *skb,
1411				   struct iwl_txq *txq, u8 hdr_len,
1412				   struct iwl_cmd_meta *out_meta,
1413				   struct iwl_device_tx_cmd *dev_cmd,
1414				   u16 tb1_len)
1415{
1416	/* No A-MSDU without CONFIG_INET */
1417	WARN_ON(1);
1418
1419	return -1;
1420}
1421#endif /* CONFIG_INET */
1422
1423int iwl_trans_pcie_tx(struct iwl_trans *trans, struct sk_buff *skb,
1424		      struct iwl_device_tx_cmd *dev_cmd, int txq_id)
1425{
1426	struct ieee80211_hdr *hdr;
1427	struct iwl_tx_cmd *tx_cmd = (struct iwl_tx_cmd *)dev_cmd->payload;
1428	struct iwl_cmd_meta *out_meta;
1429	struct iwl_txq *txq;
1430	dma_addr_t tb0_phys, tb1_phys, scratch_phys;
1431	void *tb1_addr;
1432	void *tfd;
1433	u16 len, tb1_len;
1434	bool wait_write_ptr;
1435	__le16 fc;
1436	u8 hdr_len;
1437	u16 wifi_seq;
1438	bool amsdu;
1439
1440	txq = trans->txqs.txq[txq_id];
1441
1442	if (WARN_ONCE(!test_bit(txq_id, trans->txqs.queue_used),
1443		      "TX on unused queue %d\n", txq_id))
1444		return -EINVAL;
1445
1446	if (skb_is_nonlinear(skb) &&
1447	    skb_shinfo(skb)->nr_frags > IWL_TRANS_MAX_FRAGS(trans) &&
1448	    __skb_linearize(skb))
1449		return -ENOMEM;
1450
1451	/* mac80211 always puts the full header into the SKB's head,
1452	 * so there's no need to check if it's readable there
1453	 */
1454	hdr = (struct ieee80211_hdr *)skb->data;
1455	fc = hdr->frame_control;
1456	hdr_len = ieee80211_hdrlen(fc);
1457
1458	spin_lock(&txq->lock);
1459
1460	if (iwl_txq_space(trans, txq) < txq->high_mark) {
1461		iwl_txq_stop(trans, txq);
1462
1463		/* don't put the packet on the ring, if there is no room */
1464		if (unlikely(iwl_txq_space(trans, txq) < 3)) {
1465			struct iwl_device_tx_cmd **dev_cmd_ptr;
1466
1467			dev_cmd_ptr = (void *)((u8 *)skb->cb +
1468					       trans->txqs.dev_cmd_offs);
1469
1470			*dev_cmd_ptr = dev_cmd;
1471			__skb_queue_tail(&txq->overflow_q, skb);
1472
1473			spin_unlock(&txq->lock);
1474			return 0;
1475		}
1476	}
1477
1478	/* In AGG mode, the index in the ring must correspond to the WiFi
1479	 * sequence number. This is a HW requirements to help the SCD to parse
1480	 * the BA.
1481	 * Check here that the packets are in the right place on the ring.
1482	 */
1483	wifi_seq = IEEE80211_SEQ_TO_SN(le16_to_cpu(hdr->seq_ctrl));
1484	WARN_ONCE(txq->ampdu &&
1485		  (wifi_seq & 0xff) != txq->write_ptr,
1486		  "Q: %d WiFi Seq %d tfdNum %d",
1487		  txq_id, wifi_seq, txq->write_ptr);
1488
1489	/* Set up driver data for this TFD */
1490	txq->entries[txq->write_ptr].skb = skb;
1491	txq->entries[txq->write_ptr].cmd = dev_cmd;
1492
1493	dev_cmd->hdr.sequence =
1494		cpu_to_le16((u16)(QUEUE_TO_SEQ(txq_id) |
1495			    INDEX_TO_SEQ(txq->write_ptr)));
1496
1497	tb0_phys = iwl_txq_get_first_tb_dma(txq, txq->write_ptr);
1498	scratch_phys = tb0_phys + sizeof(struct iwl_cmd_header) +
1499		       offsetof(struct iwl_tx_cmd, scratch);
1500
1501	tx_cmd->dram_lsb_ptr = cpu_to_le32(scratch_phys);
1502	tx_cmd->dram_msb_ptr = iwl_get_dma_hi_addr(scratch_phys);
1503
1504	/* Set up first empty entry in queue's array of Tx/cmd buffers */
1505	out_meta = &txq->entries[txq->write_ptr].meta;
1506	out_meta->flags = 0;
1507
1508	/*
1509	 * The second TB (tb1) points to the remainder of the TX command
1510	 * and the 802.11 header - dword aligned size
1511	 * (This calculation modifies the TX command, so do it before the
1512	 * setup of the first TB)
1513	 */
1514	len = sizeof(struct iwl_tx_cmd) + sizeof(struct iwl_cmd_header) +
1515	      hdr_len - IWL_FIRST_TB_SIZE;
1516	/* do not align A-MSDU to dword as the subframe header aligns it */
1517	amsdu = ieee80211_is_data_qos(fc) &&
1518		(*ieee80211_get_qos_ctl(hdr) &
1519		 IEEE80211_QOS_CTL_A_MSDU_PRESENT);
1520	if (!amsdu) {
1521		tb1_len = ALIGN(len, 4);
1522		/* Tell NIC about any 2-byte padding after MAC header */
1523		if (tb1_len != len)
1524			tx_cmd->tx_flags |= cpu_to_le32(TX_CMD_FLG_MH_PAD);
1525	} else {
1526		tb1_len = len;
1527	}
1528
1529	/*
1530	 * The first TB points to bi-directional DMA data, we'll
1531	 * memcpy the data into it later.
1532	 */
1533	iwl_pcie_txq_build_tfd(trans, txq, tb0_phys,
1534			       IWL_FIRST_TB_SIZE, true);
1535
1536	/* there must be data left over for TB1 or this code must be changed */
1537	BUILD_BUG_ON(sizeof(struct iwl_tx_cmd) < IWL_FIRST_TB_SIZE);
1538	BUILD_BUG_ON(sizeof(struct iwl_cmd_header) +
1539		     offsetofend(struct iwl_tx_cmd, scratch) >
1540		     IWL_FIRST_TB_SIZE);
1541
1542	/* map the data for TB1 */
1543	tb1_addr = ((u8 *)&dev_cmd->hdr) + IWL_FIRST_TB_SIZE;
1544	tb1_phys = dma_map_single(trans->dev, tb1_addr, tb1_len, DMA_TO_DEVICE);
1545	if (unlikely(dma_mapping_error(trans->dev, tb1_phys)))
1546		goto out_err;
1547	iwl_pcie_txq_build_tfd(trans, txq, tb1_phys, tb1_len, false);
1548
1549	trace_iwlwifi_dev_tx(trans->dev, skb,
1550			     iwl_txq_get_tfd(trans, txq, txq->write_ptr),
1551			     trans->txqs.tfd.size,
1552			     &dev_cmd->hdr, IWL_FIRST_TB_SIZE + tb1_len,
1553			     hdr_len);
1554
1555	/*
1556	 * If gso_size wasn't set, don't give the frame "amsdu treatment"
1557	 * (adding subframes, etc.).
1558	 * This can happen in some testing flows when the amsdu was already
1559	 * pre-built, and we just need to send the resulting skb.
1560	 */
1561	if (amsdu && skb_shinfo(skb)->gso_size) {
1562		if (unlikely(iwl_fill_data_tbs_amsdu(trans, skb, txq, hdr_len,
1563						     out_meta, dev_cmd,
1564						     tb1_len)))
1565			goto out_err;
1566	} else {
1567		struct sk_buff *frag;
1568
1569		if (unlikely(iwl_fill_data_tbs(trans, skb, txq, hdr_len,
1570					       out_meta)))
1571			goto out_err;
1572
1573		skb_walk_frags(skb, frag) {
1574			if (unlikely(iwl_fill_data_tbs(trans, frag, txq, 0,
1575						       out_meta)))
1576				goto out_err;
1577		}
1578	}
1579
1580	/* building the A-MSDU might have changed this data, so memcpy it now */
1581	memcpy(&txq->first_tb_bufs[txq->write_ptr], dev_cmd, IWL_FIRST_TB_SIZE);
1582
1583	tfd = iwl_txq_get_tfd(trans, txq, txq->write_ptr);
1584	/* Set up entry for this TFD in Tx byte-count array */
1585	iwl_txq_gen1_update_byte_cnt_tbl(trans, txq, le16_to_cpu(tx_cmd->len),
1586					 iwl_txq_gen1_tfd_get_num_tbs(trans,
1587								      tfd));
1588
1589	wait_write_ptr = ieee80211_has_morefrags(fc);
1590
1591	/* start timer if queue currently empty */
1592	if (txq->read_ptr == txq->write_ptr && txq->wd_timeout) {
1593		/*
1594		 * If the TXQ is active, then set the timer, if not,
1595		 * set the timer in remainder so that the timer will
1596		 * be armed with the right value when the station will
1597		 * wake up.
1598		 */
1599		if (!txq->frozen)
1600			mod_timer(&txq->stuck_timer,
1601				  jiffies + txq->wd_timeout);
1602		else
1603			txq->frozen_expiry_remainder = txq->wd_timeout;
1604	}
1605
1606	/* Tell device the write index *just past* this latest filled TFD */
1607	txq->write_ptr = iwl_txq_inc_wrap(trans, txq->write_ptr);
1608	if (!wait_write_ptr)
1609		iwl_pcie_txq_inc_wr_ptr(trans, txq);
1610
1611	/*
1612	 * At this point the frame is "transmitted" successfully
1613	 * and we will get a TX status notification eventually.
1614	 */
1615	spin_unlock(&txq->lock);
1616	return 0;
1617out_err:
1618	iwl_txq_gen1_tfd_unmap(trans, out_meta, txq, txq->write_ptr);
1619	spin_unlock(&txq->lock);
1620	return -1;
1621}
1622