1 // SPDX-License-Identifier: GPL-2.0
2 /* Marvell OcteonTx2 RVU Physical Function ethernet driver
3  *
4  * Copyright (C) 2020 Marvell International Ltd.
5  *
6  * This program is free software; you can redistribute it and/or modify
7  * it under the terms of the GNU General Public License version 2 as
8  * published by the Free Software Foundation.
9  */
10 
11 #include <linux/module.h>
12 #include <linux/interrupt.h>
13 #include <linux/pci.h>
14 #include <linux/etherdevice.h>
15 #include <linux/of.h>
16 #include <linux/if_vlan.h>
17 #include <linux/iommu.h>
18 #include <net/ip.h>
19 
20 #include "otx2_reg.h"
21 #include "otx2_common.h"
22 #include "otx2_txrx.h"
23 #include "otx2_struct.h"
24 #include "otx2_ptp.h"
25 #include "cn10k.h"
26 #include <rvu_trace.h>
27 
28 #define DRV_NAME	"rvu_nicpf"
29 #define DRV_STRING	"Marvell RVU NIC Physical Function Driver"
30 
31 /* Supported devices */
32 static const struct pci_device_id otx2_pf_id_table[] = {
33 	{ PCI_DEVICE(PCI_VENDOR_ID_CAVIUM, PCI_DEVID_OCTEONTX2_RVU_PF) },
34 	{ 0, }  /* end of table */
35 };
36 
37 MODULE_AUTHOR("Sunil Goutham <sgoutham@marvell.com>");
38 MODULE_DESCRIPTION(DRV_STRING);
39 MODULE_LICENSE("GPL v2");
40 MODULE_DEVICE_TABLE(pci, otx2_pf_id_table);
41 
42 static void otx2_vf_link_event_task(struct work_struct *work);
43 
44 enum {
45 	TYPE_PFAF,
46 	TYPE_PFVF,
47 };
48 
49 static int otx2_config_hw_tx_tstamp(struct otx2_nic *pfvf, bool enable);
50 static int otx2_config_hw_rx_tstamp(struct otx2_nic *pfvf, bool enable);
51 
52 static int otx2_change_mtu(struct net_device *netdev, int new_mtu)
53 {
54 	bool if_up = netif_running(netdev);
55 	int err = 0;
56 
57 	if (if_up)
58 		otx2_stop(netdev);
59 
60 	netdev_info(netdev, "Changing MTU from %d to %d\n",
61 		    netdev->mtu, new_mtu);
62 	netdev->mtu = new_mtu;
63 
64 	if (if_up)
65 		err = otx2_open(netdev);
66 
67 	return err;
68 }
69 
70 static void otx2_disable_flr_me_intr(struct otx2_nic *pf)
71 {
72 	int irq, vfs = pf->total_vfs;
73 
74 	/* Disable VFs ME interrupts */
75 	otx2_write64(pf, RVU_PF_VFME_INT_ENA_W1CX(0), INTR_MASK(vfs));
76 	irq = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFME0);
77 	free_irq(irq, pf);
78 
79 	/* Disable VFs FLR interrupts */
80 	otx2_write64(pf, RVU_PF_VFFLR_INT_ENA_W1CX(0), INTR_MASK(vfs));
81 	irq = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFFLR0);
82 	free_irq(irq, pf);
83 
84 	if (vfs <= 64)
85 		return;
86 
87 	otx2_write64(pf, RVU_PF_VFME_INT_ENA_W1CX(1), INTR_MASK(vfs - 64));
88 	irq = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFME1);
89 	free_irq(irq, pf);
90 
91 	otx2_write64(pf, RVU_PF_VFFLR_INT_ENA_W1CX(1), INTR_MASK(vfs - 64));
92 	irq = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFFLR1);
93 	free_irq(irq, pf);
94 }
95 
96 static void otx2_flr_wq_destroy(struct otx2_nic *pf)
97 {
98 	if (!pf->flr_wq)
99 		return;
100 	destroy_workqueue(pf->flr_wq);
101 	pf->flr_wq = NULL;
102 	devm_kfree(pf->dev, pf->flr_wrk);
103 }
104 
105 static void otx2_flr_handler(struct work_struct *work)
106 {
107 	struct flr_work *flrwork = container_of(work, struct flr_work, work);
108 	struct otx2_nic *pf = flrwork->pf;
109 	struct mbox *mbox = &pf->mbox;
110 	struct msg_req *req;
111 	int vf, reg = 0;
112 
113 	vf = flrwork - pf->flr_wrk;
114 
115 	mutex_lock(&mbox->lock);
116 	req = otx2_mbox_alloc_msg_vf_flr(mbox);
117 	if (!req) {
118 		mutex_unlock(&mbox->lock);
119 		return;
120 	}
121 	req->hdr.pcifunc &= RVU_PFVF_FUNC_MASK;
122 	req->hdr.pcifunc |= (vf + 1) & RVU_PFVF_FUNC_MASK;
123 
124 	if (!otx2_sync_mbox_msg(&pf->mbox)) {
125 		if (vf >= 64) {
126 			reg = 1;
127 			vf = vf - 64;
128 		}
129 		/* clear transcation pending bit */
130 		otx2_write64(pf, RVU_PF_VFTRPENDX(reg), BIT_ULL(vf));
131 		otx2_write64(pf, RVU_PF_VFFLR_INT_ENA_W1SX(reg), BIT_ULL(vf));
132 	}
133 
134 	mutex_unlock(&mbox->lock);
135 }
136 
137 static irqreturn_t otx2_pf_flr_intr_handler(int irq, void *pf_irq)
138 {
139 	struct otx2_nic *pf = (struct otx2_nic *)pf_irq;
140 	int reg, dev, vf, start_vf, num_reg = 1;
141 	u64 intr;
142 
143 	if (pf->total_vfs > 64)
144 		num_reg = 2;
145 
146 	for (reg = 0; reg < num_reg; reg++) {
147 		intr = otx2_read64(pf, RVU_PF_VFFLR_INTX(reg));
148 		if (!intr)
149 			continue;
150 		start_vf = 64 * reg;
151 		for (vf = 0; vf < 64; vf++) {
152 			if (!(intr & BIT_ULL(vf)))
153 				continue;
154 			dev = vf + start_vf;
155 			queue_work(pf->flr_wq, &pf->flr_wrk[dev].work);
156 			/* Clear interrupt */
157 			otx2_write64(pf, RVU_PF_VFFLR_INTX(reg), BIT_ULL(vf));
158 			/* Disable the interrupt */
159 			otx2_write64(pf, RVU_PF_VFFLR_INT_ENA_W1CX(reg),
160 				     BIT_ULL(vf));
161 		}
162 	}
163 	return IRQ_HANDLED;
164 }
165 
166 static irqreturn_t otx2_pf_me_intr_handler(int irq, void *pf_irq)
167 {
168 	struct otx2_nic *pf = (struct otx2_nic *)pf_irq;
169 	int vf, reg, num_reg = 1;
170 	u64 intr;
171 
172 	if (pf->total_vfs > 64)
173 		num_reg = 2;
174 
175 	for (reg = 0; reg < num_reg; reg++) {
176 		intr = otx2_read64(pf, RVU_PF_VFME_INTX(reg));
177 		if (!intr)
178 			continue;
179 		for (vf = 0; vf < 64; vf++) {
180 			if (!(intr & BIT_ULL(vf)))
181 				continue;
182 			/* clear trpend bit */
183 			otx2_write64(pf, RVU_PF_VFTRPENDX(reg), BIT_ULL(vf));
184 			/* clear interrupt */
185 			otx2_write64(pf, RVU_PF_VFME_INTX(reg), BIT_ULL(vf));
186 		}
187 	}
188 	return IRQ_HANDLED;
189 }
190 
191 static int otx2_register_flr_me_intr(struct otx2_nic *pf, int numvfs)
192 {
193 	struct otx2_hw *hw = &pf->hw;
194 	char *irq_name;
195 	int ret;
196 
197 	/* Register ME interrupt handler*/
198 	irq_name = &hw->irq_name[RVU_PF_INT_VEC_VFME0 * NAME_SIZE];
199 	snprintf(irq_name, NAME_SIZE, "RVUPF%d_ME0", rvu_get_pf(pf->pcifunc));
200 	ret = request_irq(pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFME0),
201 			  otx2_pf_me_intr_handler, 0, irq_name, pf);
202 	if (ret) {
203 		dev_err(pf->dev,
204 			"RVUPF: IRQ registration failed for ME0\n");
205 	}
206 
207 	/* Register FLR interrupt handler */
208 	irq_name = &hw->irq_name[RVU_PF_INT_VEC_VFFLR0 * NAME_SIZE];
209 	snprintf(irq_name, NAME_SIZE, "RVUPF%d_FLR0", rvu_get_pf(pf->pcifunc));
210 	ret = request_irq(pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFFLR0),
211 			  otx2_pf_flr_intr_handler, 0, irq_name, pf);
212 	if (ret) {
213 		dev_err(pf->dev,
214 			"RVUPF: IRQ registration failed for FLR0\n");
215 		return ret;
216 	}
217 
218 	if (numvfs > 64) {
219 		irq_name = &hw->irq_name[RVU_PF_INT_VEC_VFME1 * NAME_SIZE];
220 		snprintf(irq_name, NAME_SIZE, "RVUPF%d_ME1",
221 			 rvu_get_pf(pf->pcifunc));
222 		ret = request_irq(pci_irq_vector
223 				  (pf->pdev, RVU_PF_INT_VEC_VFME1),
224 				  otx2_pf_me_intr_handler, 0, irq_name, pf);
225 		if (ret) {
226 			dev_err(pf->dev,
227 				"RVUPF: IRQ registration failed for ME1\n");
228 		}
229 		irq_name = &hw->irq_name[RVU_PF_INT_VEC_VFFLR1 * NAME_SIZE];
230 		snprintf(irq_name, NAME_SIZE, "RVUPF%d_FLR1",
231 			 rvu_get_pf(pf->pcifunc));
232 		ret = request_irq(pci_irq_vector
233 				  (pf->pdev, RVU_PF_INT_VEC_VFFLR1),
234 				  otx2_pf_flr_intr_handler, 0, irq_name, pf);
235 		if (ret) {
236 			dev_err(pf->dev,
237 				"RVUPF: IRQ registration failed for FLR1\n");
238 			return ret;
239 		}
240 	}
241 
242 	/* Enable ME interrupt for all VFs*/
243 	otx2_write64(pf, RVU_PF_VFME_INTX(0), INTR_MASK(numvfs));
244 	otx2_write64(pf, RVU_PF_VFME_INT_ENA_W1SX(0), INTR_MASK(numvfs));
245 
246 	/* Enable FLR interrupt for all VFs*/
247 	otx2_write64(pf, RVU_PF_VFFLR_INTX(0), INTR_MASK(numvfs));
248 	otx2_write64(pf, RVU_PF_VFFLR_INT_ENA_W1SX(0), INTR_MASK(numvfs));
249 
250 	if (numvfs > 64) {
251 		numvfs -= 64;
252 
253 		otx2_write64(pf, RVU_PF_VFME_INTX(1), INTR_MASK(numvfs));
254 		otx2_write64(pf, RVU_PF_VFME_INT_ENA_W1SX(1),
255 			     INTR_MASK(numvfs));
256 
257 		otx2_write64(pf, RVU_PF_VFFLR_INTX(1), INTR_MASK(numvfs));
258 		otx2_write64(pf, RVU_PF_VFFLR_INT_ENA_W1SX(1),
259 			     INTR_MASK(numvfs));
260 	}
261 	return 0;
262 }
263 
264 static int otx2_pf_flr_init(struct otx2_nic *pf, int num_vfs)
265 {
266 	int vf;
267 
268 	pf->flr_wq = alloc_workqueue("otx2_pf_flr_wq",
269 				     WQ_UNBOUND | WQ_HIGHPRI, 1);
270 	if (!pf->flr_wq)
271 		return -ENOMEM;
272 
273 	pf->flr_wrk = devm_kcalloc(pf->dev, num_vfs,
274 				   sizeof(struct flr_work), GFP_KERNEL);
275 	if (!pf->flr_wrk) {
276 		destroy_workqueue(pf->flr_wq);
277 		return -ENOMEM;
278 	}
279 
280 	for (vf = 0; vf < num_vfs; vf++) {
281 		pf->flr_wrk[vf].pf = pf;
282 		INIT_WORK(&pf->flr_wrk[vf].work, otx2_flr_handler);
283 	}
284 
285 	return 0;
286 }
287 
288 static void otx2_queue_work(struct mbox *mw, struct workqueue_struct *mbox_wq,
289 			    int first, int mdevs, u64 intr, int type)
290 {
291 	struct otx2_mbox_dev *mdev;
292 	struct otx2_mbox *mbox;
293 	struct mbox_hdr *hdr;
294 	int i;
295 
296 	for (i = first; i < mdevs; i++) {
297 		/* start from 0 */
298 		if (!(intr & BIT_ULL(i - first)))
299 			continue;
300 
301 		mbox = &mw->mbox;
302 		mdev = &mbox->dev[i];
303 		if (type == TYPE_PFAF)
304 			otx2_sync_mbox_bbuf(mbox, i);
305 		hdr = mdev->mbase + mbox->rx_start;
306 		/* The hdr->num_msgs is set to zero immediately in the interrupt
307 		 * handler to  ensure that it holds a correct value next time
308 		 * when the interrupt handler is called.
309 		 * pf->mbox.num_msgs holds the data for use in pfaf_mbox_handler
310 		 * pf>mbox.up_num_msgs holds the data for use in
311 		 * pfaf_mbox_up_handler.
312 		 */
313 		if (hdr->num_msgs) {
314 			mw[i].num_msgs = hdr->num_msgs;
315 			hdr->num_msgs = 0;
316 			if (type == TYPE_PFAF)
317 				memset(mbox->hwbase + mbox->rx_start, 0,
318 				       ALIGN(sizeof(struct mbox_hdr),
319 					     sizeof(u64)));
320 
321 			queue_work(mbox_wq, &mw[i].mbox_wrk);
322 		}
323 
324 		mbox = &mw->mbox_up;
325 		mdev = &mbox->dev[i];
326 		if (type == TYPE_PFAF)
327 			otx2_sync_mbox_bbuf(mbox, i);
328 		hdr = mdev->mbase + mbox->rx_start;
329 		if (hdr->num_msgs) {
330 			mw[i].up_num_msgs = hdr->num_msgs;
331 			hdr->num_msgs = 0;
332 			if (type == TYPE_PFAF)
333 				memset(mbox->hwbase + mbox->rx_start, 0,
334 				       ALIGN(sizeof(struct mbox_hdr),
335 					     sizeof(u64)));
336 
337 			queue_work(mbox_wq, &mw[i].mbox_up_wrk);
338 		}
339 	}
340 }
341 
342 static void otx2_forward_msg_pfvf(struct otx2_mbox_dev *mdev,
343 				  struct otx2_mbox *pfvf_mbox, void *bbuf_base,
344 				  int devid)
345 {
346 	struct otx2_mbox_dev *src_mdev = mdev;
347 	int offset;
348 
349 	/* Msgs are already copied, trigger VF's mbox irq */
350 	smp_wmb();
351 
352 	offset = pfvf_mbox->trigger | (devid << pfvf_mbox->tr_shift);
353 	writeq(1, (void __iomem *)pfvf_mbox->reg_base + offset);
354 
355 	/* Restore VF's mbox bounce buffer region address */
356 	src_mdev->mbase = bbuf_base;
357 }
358 
359 static int otx2_forward_vf_mbox_msgs(struct otx2_nic *pf,
360 				     struct otx2_mbox *src_mbox,
361 				     int dir, int vf, int num_msgs)
362 {
363 	struct otx2_mbox_dev *src_mdev, *dst_mdev;
364 	struct mbox_hdr *mbox_hdr;
365 	struct mbox_hdr *req_hdr;
366 	struct mbox *dst_mbox;
367 	int dst_size, err;
368 
369 	if (dir == MBOX_DIR_PFAF) {
370 		/* Set VF's mailbox memory as PF's bounce buffer memory, so
371 		 * that explicit copying of VF's msgs to PF=>AF mbox region
372 		 * and AF=>PF responses to VF's mbox region can be avoided.
373 		 */
374 		src_mdev = &src_mbox->dev[vf];
375 		mbox_hdr = src_mbox->hwbase +
376 				src_mbox->rx_start + (vf * MBOX_SIZE);
377 
378 		dst_mbox = &pf->mbox;
379 		dst_size = dst_mbox->mbox.tx_size -
380 				ALIGN(sizeof(*mbox_hdr), MBOX_MSG_ALIGN);
381 		/* Check if msgs fit into destination area and has valid size */
382 		if (mbox_hdr->msg_size > dst_size || !mbox_hdr->msg_size)
383 			return -EINVAL;
384 
385 		dst_mdev = &dst_mbox->mbox.dev[0];
386 
387 		mutex_lock(&pf->mbox.lock);
388 		dst_mdev->mbase = src_mdev->mbase;
389 		dst_mdev->msg_size = mbox_hdr->msg_size;
390 		dst_mdev->num_msgs = num_msgs;
391 		err = otx2_sync_mbox_msg(dst_mbox);
392 		if (err) {
393 			dev_warn(pf->dev,
394 				 "AF not responding to VF%d messages\n", vf);
395 			/* restore PF mbase and exit */
396 			dst_mdev->mbase = pf->mbox.bbuf_base;
397 			mutex_unlock(&pf->mbox.lock);
398 			return err;
399 		}
400 		/* At this point, all the VF messages sent to AF are acked
401 		 * with proper responses and responses are copied to VF
402 		 * mailbox hence raise interrupt to VF.
403 		 */
404 		req_hdr = (struct mbox_hdr *)(dst_mdev->mbase +
405 					      dst_mbox->mbox.rx_start);
406 		req_hdr->num_msgs = num_msgs;
407 
408 		otx2_forward_msg_pfvf(dst_mdev, &pf->mbox_pfvf[0].mbox,
409 				      pf->mbox.bbuf_base, vf);
410 		mutex_unlock(&pf->mbox.lock);
411 	} else if (dir == MBOX_DIR_PFVF_UP) {
412 		src_mdev = &src_mbox->dev[0];
413 		mbox_hdr = src_mbox->hwbase + src_mbox->rx_start;
414 		req_hdr = (struct mbox_hdr *)(src_mdev->mbase +
415 					      src_mbox->rx_start);
416 		req_hdr->num_msgs = num_msgs;
417 
418 		dst_mbox = &pf->mbox_pfvf[0];
419 		dst_size = dst_mbox->mbox_up.tx_size -
420 				ALIGN(sizeof(*mbox_hdr), MBOX_MSG_ALIGN);
421 		/* Check if msgs fit into destination area */
422 		if (mbox_hdr->msg_size > dst_size)
423 			return -EINVAL;
424 
425 		dst_mdev = &dst_mbox->mbox_up.dev[vf];
426 		dst_mdev->mbase = src_mdev->mbase;
427 		dst_mdev->msg_size = mbox_hdr->msg_size;
428 		dst_mdev->num_msgs = mbox_hdr->num_msgs;
429 		err = otx2_sync_mbox_up_msg(dst_mbox, vf);
430 		if (err) {
431 			dev_warn(pf->dev,
432 				 "VF%d is not responding to mailbox\n", vf);
433 			return err;
434 		}
435 	} else if (dir == MBOX_DIR_VFPF_UP) {
436 		req_hdr = (struct mbox_hdr *)(src_mbox->dev[0].mbase +
437 					      src_mbox->rx_start);
438 		req_hdr->num_msgs = num_msgs;
439 		otx2_forward_msg_pfvf(&pf->mbox_pfvf->mbox_up.dev[vf],
440 				      &pf->mbox.mbox_up,
441 				      pf->mbox_pfvf[vf].bbuf_base,
442 				      0);
443 	}
444 
445 	return 0;
446 }
447 
448 static void otx2_pfvf_mbox_handler(struct work_struct *work)
449 {
450 	struct mbox_msghdr *msg = NULL;
451 	int offset, vf_idx, id, err;
452 	struct otx2_mbox_dev *mdev;
453 	struct mbox_hdr *req_hdr;
454 	struct otx2_mbox *mbox;
455 	struct mbox *vf_mbox;
456 	struct otx2_nic *pf;
457 
458 	vf_mbox = container_of(work, struct mbox, mbox_wrk);
459 	pf = vf_mbox->pfvf;
460 	vf_idx = vf_mbox - pf->mbox_pfvf;
461 
462 	mbox = &pf->mbox_pfvf[0].mbox;
463 	mdev = &mbox->dev[vf_idx];
464 	req_hdr = (struct mbox_hdr *)(mdev->mbase + mbox->rx_start);
465 
466 	offset = ALIGN(sizeof(*req_hdr), MBOX_MSG_ALIGN);
467 
468 	for (id = 0; id < vf_mbox->num_msgs; id++) {
469 		msg = (struct mbox_msghdr *)(mdev->mbase + mbox->rx_start +
470 					     offset);
471 
472 		if (msg->sig != OTX2_MBOX_REQ_SIG)
473 			goto inval_msg;
474 
475 		/* Set VF's number in each of the msg */
476 		msg->pcifunc &= RVU_PFVF_FUNC_MASK;
477 		msg->pcifunc |= (vf_idx + 1) & RVU_PFVF_FUNC_MASK;
478 		offset = msg->next_msgoff;
479 	}
480 	err = otx2_forward_vf_mbox_msgs(pf, mbox, MBOX_DIR_PFAF, vf_idx,
481 					vf_mbox->num_msgs);
482 	if (err)
483 		goto inval_msg;
484 	return;
485 
486 inval_msg:
487 	otx2_reply_invalid_msg(mbox, vf_idx, 0, msg->id);
488 	otx2_mbox_msg_send(mbox, vf_idx);
489 }
490 
491 static void otx2_pfvf_mbox_up_handler(struct work_struct *work)
492 {
493 	struct mbox *vf_mbox = container_of(work, struct mbox, mbox_up_wrk);
494 	struct otx2_nic *pf = vf_mbox->pfvf;
495 	struct otx2_mbox_dev *mdev;
496 	int offset, id, vf_idx = 0;
497 	struct mbox_hdr *rsp_hdr;
498 	struct mbox_msghdr *msg;
499 	struct otx2_mbox *mbox;
500 
501 	vf_idx = vf_mbox - pf->mbox_pfvf;
502 	mbox = &pf->mbox_pfvf[0].mbox_up;
503 	mdev = &mbox->dev[vf_idx];
504 
505 	rsp_hdr = (struct mbox_hdr *)(mdev->mbase + mbox->rx_start);
506 	offset = mbox->rx_start + ALIGN(sizeof(*rsp_hdr), MBOX_MSG_ALIGN);
507 
508 	for (id = 0; id < vf_mbox->up_num_msgs; id++) {
509 		msg = mdev->mbase + offset;
510 
511 		if (msg->id >= MBOX_MSG_MAX) {
512 			dev_err(pf->dev,
513 				"Mbox msg with unknown ID 0x%x\n", msg->id);
514 			goto end;
515 		}
516 
517 		if (msg->sig != OTX2_MBOX_RSP_SIG) {
518 			dev_err(pf->dev,
519 				"Mbox msg with wrong signature %x, ID 0x%x\n",
520 				msg->sig, msg->id);
521 			goto end;
522 		}
523 
524 		switch (msg->id) {
525 		case MBOX_MSG_CGX_LINK_EVENT:
526 			break;
527 		default:
528 			if (msg->rc)
529 				dev_err(pf->dev,
530 					"Mbox msg response has err %d, ID 0x%x\n",
531 					msg->rc, msg->id);
532 			break;
533 		}
534 
535 end:
536 		offset = mbox->rx_start + msg->next_msgoff;
537 		if (mdev->msgs_acked == (vf_mbox->up_num_msgs - 1))
538 			__otx2_mbox_reset(mbox, 0);
539 		mdev->msgs_acked++;
540 	}
541 }
542 
543 static irqreturn_t otx2_pfvf_mbox_intr_handler(int irq, void *pf_irq)
544 {
545 	struct otx2_nic *pf = (struct otx2_nic *)(pf_irq);
546 	int vfs = pf->total_vfs;
547 	struct mbox *mbox;
548 	u64 intr;
549 
550 	mbox = pf->mbox_pfvf;
551 	/* Handle VF interrupts */
552 	if (vfs > 64) {
553 		intr = otx2_read64(pf, RVU_PF_VFPF_MBOX_INTX(1));
554 		otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(1), intr);
555 		otx2_queue_work(mbox, pf->mbox_pfvf_wq, 64, vfs, intr,
556 				TYPE_PFVF);
557 		vfs -= 64;
558 	}
559 
560 	intr = otx2_read64(pf, RVU_PF_VFPF_MBOX_INTX(0));
561 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(0), intr);
562 
563 	otx2_queue_work(mbox, pf->mbox_pfvf_wq, 0, vfs, intr, TYPE_PFVF);
564 
565 	trace_otx2_msg_interrupt(mbox->mbox.pdev, "VF(s) to PF", intr);
566 
567 	return IRQ_HANDLED;
568 }
569 
570 static int otx2_pfvf_mbox_init(struct otx2_nic *pf, int numvfs)
571 {
572 	void __iomem *hwbase;
573 	struct mbox *mbox;
574 	int err, vf;
575 	u64 base;
576 
577 	if (!numvfs)
578 		return -EINVAL;
579 
580 	pf->mbox_pfvf = devm_kcalloc(&pf->pdev->dev, numvfs,
581 				     sizeof(struct mbox), GFP_KERNEL);
582 	if (!pf->mbox_pfvf)
583 		return -ENOMEM;
584 
585 	pf->mbox_pfvf_wq = alloc_workqueue("otx2_pfvf_mailbox",
586 					   WQ_UNBOUND | WQ_HIGHPRI |
587 					   WQ_MEM_RECLAIM, 1);
588 	if (!pf->mbox_pfvf_wq)
589 		return -ENOMEM;
590 
591 	/* On CN10K platform, PF <-> VF mailbox region follows after
592 	 * PF <-> AF mailbox region.
593 	 */
594 	if (test_bit(CN10K_MBOX, &pf->hw.cap_flag))
595 		base = pci_resource_start(pf->pdev, PCI_MBOX_BAR_NUM) +
596 		       MBOX_SIZE;
597 	else
598 		base = readq((void __iomem *)((u64)pf->reg_base +
599 					      RVU_PF_VF_BAR4_ADDR));
600 
601 	hwbase = ioremap_wc(base, MBOX_SIZE * pf->total_vfs);
602 	if (!hwbase) {
603 		err = -ENOMEM;
604 		goto free_wq;
605 	}
606 
607 	mbox = &pf->mbox_pfvf[0];
608 	err = otx2_mbox_init(&mbox->mbox, hwbase, pf->pdev, pf->reg_base,
609 			     MBOX_DIR_PFVF, numvfs);
610 	if (err)
611 		goto free_iomem;
612 
613 	err = otx2_mbox_init(&mbox->mbox_up, hwbase, pf->pdev, pf->reg_base,
614 			     MBOX_DIR_PFVF_UP, numvfs);
615 	if (err)
616 		goto free_iomem;
617 
618 	for (vf = 0; vf < numvfs; vf++) {
619 		mbox->pfvf = pf;
620 		INIT_WORK(&mbox->mbox_wrk, otx2_pfvf_mbox_handler);
621 		INIT_WORK(&mbox->mbox_up_wrk, otx2_pfvf_mbox_up_handler);
622 		mbox++;
623 	}
624 
625 	return 0;
626 
627 free_iomem:
628 	if (hwbase)
629 		iounmap(hwbase);
630 free_wq:
631 	destroy_workqueue(pf->mbox_pfvf_wq);
632 	return err;
633 }
634 
635 static void otx2_pfvf_mbox_destroy(struct otx2_nic *pf)
636 {
637 	struct mbox *mbox = &pf->mbox_pfvf[0];
638 
639 	if (!mbox)
640 		return;
641 
642 	if (pf->mbox_pfvf_wq) {
643 		destroy_workqueue(pf->mbox_pfvf_wq);
644 		pf->mbox_pfvf_wq = NULL;
645 	}
646 
647 	if (mbox->mbox.hwbase)
648 		iounmap(mbox->mbox.hwbase);
649 
650 	otx2_mbox_destroy(&mbox->mbox);
651 }
652 
653 static void otx2_enable_pfvf_mbox_intr(struct otx2_nic *pf, int numvfs)
654 {
655 	/* Clear PF <=> VF mailbox IRQ */
656 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(0), ~0ull);
657 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(1), ~0ull);
658 
659 	/* Enable PF <=> VF mailbox IRQ */
660 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INT_ENA_W1SX(0), INTR_MASK(numvfs));
661 	if (numvfs > 64) {
662 		numvfs -= 64;
663 		otx2_write64(pf, RVU_PF_VFPF_MBOX_INT_ENA_W1SX(1),
664 			     INTR_MASK(numvfs));
665 	}
666 }
667 
668 static void otx2_disable_pfvf_mbox_intr(struct otx2_nic *pf, int numvfs)
669 {
670 	int vector;
671 
672 	/* Disable PF <=> VF mailbox IRQ */
673 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INT_ENA_W1CX(0), ~0ull);
674 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INT_ENA_W1CX(1), ~0ull);
675 
676 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(0), ~0ull);
677 	vector = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFPF_MBOX0);
678 	free_irq(vector, pf);
679 
680 	if (numvfs > 64) {
681 		otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(1), ~0ull);
682 		vector = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFPF_MBOX1);
683 		free_irq(vector, pf);
684 	}
685 }
686 
687 static int otx2_register_pfvf_mbox_intr(struct otx2_nic *pf, int numvfs)
688 {
689 	struct otx2_hw *hw = &pf->hw;
690 	char *irq_name;
691 	int err;
692 
693 	/* Register MBOX0 interrupt handler */
694 	irq_name = &hw->irq_name[RVU_PF_INT_VEC_VFPF_MBOX0 * NAME_SIZE];
695 	if (pf->pcifunc)
696 		snprintf(irq_name, NAME_SIZE,
697 			 "RVUPF%d_VF Mbox0", rvu_get_pf(pf->pcifunc));
698 	else
699 		snprintf(irq_name, NAME_SIZE, "RVUPF_VF Mbox0");
700 	err = request_irq(pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFPF_MBOX0),
701 			  otx2_pfvf_mbox_intr_handler, 0, irq_name, pf);
702 	if (err) {
703 		dev_err(pf->dev,
704 			"RVUPF: IRQ registration failed for PFVF mbox0 irq\n");
705 		return err;
706 	}
707 
708 	if (numvfs > 64) {
709 		/* Register MBOX1 interrupt handler */
710 		irq_name = &hw->irq_name[RVU_PF_INT_VEC_VFPF_MBOX1 * NAME_SIZE];
711 		if (pf->pcifunc)
712 			snprintf(irq_name, NAME_SIZE,
713 				 "RVUPF%d_VF Mbox1", rvu_get_pf(pf->pcifunc));
714 		else
715 			snprintf(irq_name, NAME_SIZE, "RVUPF_VF Mbox1");
716 		err = request_irq(pci_irq_vector(pf->pdev,
717 						 RVU_PF_INT_VEC_VFPF_MBOX1),
718 						 otx2_pfvf_mbox_intr_handler,
719 						 0, irq_name, pf);
720 		if (err) {
721 			dev_err(pf->dev,
722 				"RVUPF: IRQ registration failed for PFVF mbox1 irq\n");
723 			return err;
724 		}
725 	}
726 
727 	otx2_enable_pfvf_mbox_intr(pf, numvfs);
728 
729 	return 0;
730 }
731 
732 static void otx2_process_pfaf_mbox_msg(struct otx2_nic *pf,
733 				       struct mbox_msghdr *msg)
734 {
735 	int devid;
736 
737 	if (msg->id >= MBOX_MSG_MAX) {
738 		dev_err(pf->dev,
739 			"Mbox msg with unknown ID 0x%x\n", msg->id);
740 		return;
741 	}
742 
743 	if (msg->sig != OTX2_MBOX_RSP_SIG) {
744 		dev_err(pf->dev,
745 			"Mbox msg with wrong signature %x, ID 0x%x\n",
746 			 msg->sig, msg->id);
747 		return;
748 	}
749 
750 	/* message response heading VF */
751 	devid = msg->pcifunc & RVU_PFVF_FUNC_MASK;
752 	if (devid) {
753 		struct otx2_vf_config *config = &pf->vf_configs[devid - 1];
754 		struct delayed_work *dwork;
755 
756 		switch (msg->id) {
757 		case MBOX_MSG_NIX_LF_START_RX:
758 			config->intf_down = false;
759 			dwork = &config->link_event_work;
760 			schedule_delayed_work(dwork, msecs_to_jiffies(100));
761 			break;
762 		case MBOX_MSG_NIX_LF_STOP_RX:
763 			config->intf_down = true;
764 			break;
765 		}
766 
767 		return;
768 	}
769 
770 	switch (msg->id) {
771 	case MBOX_MSG_READY:
772 		pf->pcifunc = msg->pcifunc;
773 		break;
774 	case MBOX_MSG_MSIX_OFFSET:
775 		mbox_handler_msix_offset(pf, (struct msix_offset_rsp *)msg);
776 		break;
777 	case MBOX_MSG_NPA_LF_ALLOC:
778 		mbox_handler_npa_lf_alloc(pf, (struct npa_lf_alloc_rsp *)msg);
779 		break;
780 	case MBOX_MSG_NIX_LF_ALLOC:
781 		mbox_handler_nix_lf_alloc(pf, (struct nix_lf_alloc_rsp *)msg);
782 		break;
783 	case MBOX_MSG_NIX_TXSCH_ALLOC:
784 		mbox_handler_nix_txsch_alloc(pf,
785 					     (struct nix_txsch_alloc_rsp *)msg);
786 		break;
787 	case MBOX_MSG_NIX_BP_ENABLE:
788 		mbox_handler_nix_bp_enable(pf, (struct nix_bp_cfg_rsp *)msg);
789 		break;
790 	case MBOX_MSG_CGX_STATS:
791 		mbox_handler_cgx_stats(pf, (struct cgx_stats_rsp *)msg);
792 		break;
793 	case MBOX_MSG_CGX_FEC_STATS:
794 		mbox_handler_cgx_fec_stats(pf, (struct cgx_fec_stats_rsp *)msg);
795 		break;
796 	default:
797 		if (msg->rc)
798 			dev_err(pf->dev,
799 				"Mbox msg response has err %d, ID 0x%x\n",
800 				msg->rc, msg->id);
801 		break;
802 	}
803 }
804 
805 static void otx2_pfaf_mbox_handler(struct work_struct *work)
806 {
807 	struct otx2_mbox_dev *mdev;
808 	struct mbox_hdr *rsp_hdr;
809 	struct mbox_msghdr *msg;
810 	struct otx2_mbox *mbox;
811 	struct mbox *af_mbox;
812 	struct otx2_nic *pf;
813 	int offset, id;
814 
815 	af_mbox = container_of(work, struct mbox, mbox_wrk);
816 	mbox = &af_mbox->mbox;
817 	mdev = &mbox->dev[0];
818 	rsp_hdr = (struct mbox_hdr *)(mdev->mbase + mbox->rx_start);
819 
820 	offset = mbox->rx_start + ALIGN(sizeof(*rsp_hdr), MBOX_MSG_ALIGN);
821 	pf = af_mbox->pfvf;
822 
823 	for (id = 0; id < af_mbox->num_msgs; id++) {
824 		msg = (struct mbox_msghdr *)(mdev->mbase + offset);
825 		otx2_process_pfaf_mbox_msg(pf, msg);
826 		offset = mbox->rx_start + msg->next_msgoff;
827 		if (mdev->msgs_acked == (af_mbox->num_msgs - 1))
828 			__otx2_mbox_reset(mbox, 0);
829 		mdev->msgs_acked++;
830 	}
831 
832 }
833 
834 static void otx2_handle_link_event(struct otx2_nic *pf)
835 {
836 	struct cgx_link_user_info *linfo = &pf->linfo;
837 	struct net_device *netdev = pf->netdev;
838 
839 	pr_info("%s NIC Link is %s %d Mbps %s duplex\n", netdev->name,
840 		linfo->link_up ? "UP" : "DOWN", linfo->speed,
841 		linfo->full_duplex ? "Full" : "Half");
842 	if (linfo->link_up) {
843 		netif_carrier_on(netdev);
844 		netif_tx_start_all_queues(netdev);
845 	} else {
846 		netif_tx_stop_all_queues(netdev);
847 		netif_carrier_off(netdev);
848 	}
849 }
850 
851 int otx2_mbox_up_handler_cgx_link_event(struct otx2_nic *pf,
852 					struct cgx_link_info_msg *msg,
853 					struct msg_rsp *rsp)
854 {
855 	int i;
856 
857 	/* Copy the link info sent by AF */
858 	pf->linfo = msg->link_info;
859 
860 	/* notify VFs about link event */
861 	for (i = 0; i < pci_num_vf(pf->pdev); i++) {
862 		struct otx2_vf_config *config = &pf->vf_configs[i];
863 		struct delayed_work *dwork = &config->link_event_work;
864 
865 		if (config->intf_down)
866 			continue;
867 
868 		schedule_delayed_work(dwork, msecs_to_jiffies(100));
869 	}
870 
871 	/* interface has not been fully configured yet */
872 	if (pf->flags & OTX2_FLAG_INTF_DOWN)
873 		return 0;
874 
875 	otx2_handle_link_event(pf);
876 	return 0;
877 }
878 
879 static int otx2_process_mbox_msg_up(struct otx2_nic *pf,
880 				    struct mbox_msghdr *req)
881 {
882 	/* Check if valid, if not reply with a invalid msg */
883 	if (req->sig != OTX2_MBOX_REQ_SIG) {
884 		otx2_reply_invalid_msg(&pf->mbox.mbox_up, 0, 0, req->id);
885 		return -ENODEV;
886 	}
887 
888 	switch (req->id) {
889 #define M(_name, _id, _fn_name, _req_type, _rsp_type)			\
890 	case _id: {							\
891 		struct _rsp_type *rsp;					\
892 		int err;						\
893 									\
894 		rsp = (struct _rsp_type *)otx2_mbox_alloc_msg(		\
895 			&pf->mbox.mbox_up, 0,				\
896 			sizeof(struct _rsp_type));			\
897 		if (!rsp)						\
898 			return -ENOMEM;					\
899 									\
900 		rsp->hdr.id = _id;					\
901 		rsp->hdr.sig = OTX2_MBOX_RSP_SIG;			\
902 		rsp->hdr.pcifunc = 0;					\
903 		rsp->hdr.rc = 0;					\
904 									\
905 		err = otx2_mbox_up_handler_ ## _fn_name(		\
906 			pf, (struct _req_type *)req, rsp);		\
907 		return err;						\
908 	}
909 MBOX_UP_CGX_MESSAGES
910 #undef M
911 		break;
912 	default:
913 		otx2_reply_invalid_msg(&pf->mbox.mbox_up, 0, 0, req->id);
914 		return -ENODEV;
915 	}
916 	return 0;
917 }
918 
919 static void otx2_pfaf_mbox_up_handler(struct work_struct *work)
920 {
921 	struct mbox *af_mbox = container_of(work, struct mbox, mbox_up_wrk);
922 	struct otx2_mbox *mbox = &af_mbox->mbox_up;
923 	struct otx2_mbox_dev *mdev = &mbox->dev[0];
924 	struct otx2_nic *pf = af_mbox->pfvf;
925 	int offset, id, devid = 0;
926 	struct mbox_hdr *rsp_hdr;
927 	struct mbox_msghdr *msg;
928 
929 	rsp_hdr = (struct mbox_hdr *)(mdev->mbase + mbox->rx_start);
930 
931 	offset = mbox->rx_start + ALIGN(sizeof(*rsp_hdr), MBOX_MSG_ALIGN);
932 
933 	for (id = 0; id < af_mbox->up_num_msgs; id++) {
934 		msg = (struct mbox_msghdr *)(mdev->mbase + offset);
935 
936 		devid = msg->pcifunc & RVU_PFVF_FUNC_MASK;
937 		/* Skip processing VF's messages */
938 		if (!devid)
939 			otx2_process_mbox_msg_up(pf, msg);
940 		offset = mbox->rx_start + msg->next_msgoff;
941 	}
942 	if (devid) {
943 		otx2_forward_vf_mbox_msgs(pf, &pf->mbox.mbox_up,
944 					  MBOX_DIR_PFVF_UP, devid - 1,
945 					  af_mbox->up_num_msgs);
946 		return;
947 	}
948 
949 	otx2_mbox_msg_send(mbox, 0);
950 }
951 
952 static irqreturn_t otx2_pfaf_mbox_intr_handler(int irq, void *pf_irq)
953 {
954 	struct otx2_nic *pf = (struct otx2_nic *)pf_irq;
955 	struct mbox *mbox;
956 
957 	/* Clear the IRQ */
958 	otx2_write64(pf, RVU_PF_INT, BIT_ULL(0));
959 
960 	mbox = &pf->mbox;
961 
962 	trace_otx2_msg_interrupt(mbox->mbox.pdev, "AF to PF", BIT_ULL(0));
963 
964 	otx2_queue_work(mbox, pf->mbox_wq, 0, 1, 1, TYPE_PFAF);
965 
966 	return IRQ_HANDLED;
967 }
968 
969 static void otx2_disable_mbox_intr(struct otx2_nic *pf)
970 {
971 	int vector = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_AFPF_MBOX);
972 
973 	/* Disable AF => PF mailbox IRQ */
974 	otx2_write64(pf, RVU_PF_INT_ENA_W1C, BIT_ULL(0));
975 	free_irq(vector, pf);
976 }
977 
978 static int otx2_register_mbox_intr(struct otx2_nic *pf, bool probe_af)
979 {
980 	struct otx2_hw *hw = &pf->hw;
981 	struct msg_req *req;
982 	char *irq_name;
983 	int err;
984 
985 	/* Register mailbox interrupt handler */
986 	irq_name = &hw->irq_name[RVU_PF_INT_VEC_AFPF_MBOX * NAME_SIZE];
987 	snprintf(irq_name, NAME_SIZE, "RVUPFAF Mbox");
988 	err = request_irq(pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_AFPF_MBOX),
989 			  otx2_pfaf_mbox_intr_handler, 0, irq_name, pf);
990 	if (err) {
991 		dev_err(pf->dev,
992 			"RVUPF: IRQ registration failed for PFAF mbox irq\n");
993 		return err;
994 	}
995 
996 	/* Enable mailbox interrupt for msgs coming from AF.
997 	 * First clear to avoid spurious interrupts, if any.
998 	 */
999 	otx2_write64(pf, RVU_PF_INT, BIT_ULL(0));
1000 	otx2_write64(pf, RVU_PF_INT_ENA_W1S, BIT_ULL(0));
1001 
1002 	if (!probe_af)
1003 		return 0;
1004 
1005 	/* Check mailbox communication with AF */
1006 	req = otx2_mbox_alloc_msg_ready(&pf->mbox);
1007 	if (!req) {
1008 		otx2_disable_mbox_intr(pf);
1009 		return -ENOMEM;
1010 	}
1011 	err = otx2_sync_mbox_msg(&pf->mbox);
1012 	if (err) {
1013 		dev_warn(pf->dev,
1014 			 "AF not responding to mailbox, deferring probe\n");
1015 		otx2_disable_mbox_intr(pf);
1016 		return -EPROBE_DEFER;
1017 	}
1018 
1019 	return 0;
1020 }
1021 
1022 static void otx2_pfaf_mbox_destroy(struct otx2_nic *pf)
1023 {
1024 	struct mbox *mbox = &pf->mbox;
1025 
1026 	if (pf->mbox_wq) {
1027 		destroy_workqueue(pf->mbox_wq);
1028 		pf->mbox_wq = NULL;
1029 	}
1030 
1031 	if (mbox->mbox.hwbase)
1032 		iounmap((void __iomem *)mbox->mbox.hwbase);
1033 
1034 	otx2_mbox_destroy(&mbox->mbox);
1035 	otx2_mbox_destroy(&mbox->mbox_up);
1036 }
1037 
1038 static int otx2_pfaf_mbox_init(struct otx2_nic *pf)
1039 {
1040 	struct mbox *mbox = &pf->mbox;
1041 	void __iomem *hwbase;
1042 	int err;
1043 
1044 	mbox->pfvf = pf;
1045 	pf->mbox_wq = alloc_workqueue("otx2_pfaf_mailbox",
1046 				      WQ_UNBOUND | WQ_HIGHPRI |
1047 				      WQ_MEM_RECLAIM, 1);
1048 	if (!pf->mbox_wq)
1049 		return -ENOMEM;
1050 
1051 	/* Mailbox is a reserved memory (in RAM) region shared between
1052 	 * admin function (i.e AF) and this PF, shouldn't be mapped as
1053 	 * device memory to allow unaligned accesses.
1054 	 */
1055 	hwbase = ioremap_wc(pci_resource_start(pf->pdev, PCI_MBOX_BAR_NUM),
1056 			    MBOX_SIZE);
1057 	if (!hwbase) {
1058 		dev_err(pf->dev, "Unable to map PFAF mailbox region\n");
1059 		err = -ENOMEM;
1060 		goto exit;
1061 	}
1062 
1063 	err = otx2_mbox_init(&mbox->mbox, hwbase, pf->pdev, pf->reg_base,
1064 			     MBOX_DIR_PFAF, 1);
1065 	if (err)
1066 		goto exit;
1067 
1068 	err = otx2_mbox_init(&mbox->mbox_up, hwbase, pf->pdev, pf->reg_base,
1069 			     MBOX_DIR_PFAF_UP, 1);
1070 	if (err)
1071 		goto exit;
1072 
1073 	err = otx2_mbox_bbuf_init(mbox, pf->pdev);
1074 	if (err)
1075 		goto exit;
1076 
1077 	INIT_WORK(&mbox->mbox_wrk, otx2_pfaf_mbox_handler);
1078 	INIT_WORK(&mbox->mbox_up_wrk, otx2_pfaf_mbox_up_handler);
1079 	mutex_init(&mbox->lock);
1080 
1081 	return 0;
1082 exit:
1083 	otx2_pfaf_mbox_destroy(pf);
1084 	return err;
1085 }
1086 
1087 static int otx2_cgx_config_linkevents(struct otx2_nic *pf, bool enable)
1088 {
1089 	struct msg_req *msg;
1090 	int err;
1091 
1092 	mutex_lock(&pf->mbox.lock);
1093 	if (enable)
1094 		msg = otx2_mbox_alloc_msg_cgx_start_linkevents(&pf->mbox);
1095 	else
1096 		msg = otx2_mbox_alloc_msg_cgx_stop_linkevents(&pf->mbox);
1097 
1098 	if (!msg) {
1099 		mutex_unlock(&pf->mbox.lock);
1100 		return -ENOMEM;
1101 	}
1102 
1103 	err = otx2_sync_mbox_msg(&pf->mbox);
1104 	mutex_unlock(&pf->mbox.lock);
1105 	return err;
1106 }
1107 
1108 static int otx2_cgx_config_loopback(struct otx2_nic *pf, bool enable)
1109 {
1110 	struct msg_req *msg;
1111 	int err;
1112 
1113 	mutex_lock(&pf->mbox.lock);
1114 	if (enable)
1115 		msg = otx2_mbox_alloc_msg_cgx_intlbk_enable(&pf->mbox);
1116 	else
1117 		msg = otx2_mbox_alloc_msg_cgx_intlbk_disable(&pf->mbox);
1118 
1119 	if (!msg) {
1120 		mutex_unlock(&pf->mbox.lock);
1121 		return -ENOMEM;
1122 	}
1123 
1124 	err = otx2_sync_mbox_msg(&pf->mbox);
1125 	mutex_unlock(&pf->mbox.lock);
1126 	return err;
1127 }
1128 
1129 int otx2_set_real_num_queues(struct net_device *netdev,
1130 			     int tx_queues, int rx_queues)
1131 {
1132 	int err;
1133 
1134 	err = netif_set_real_num_tx_queues(netdev, tx_queues);
1135 	if (err) {
1136 		netdev_err(netdev,
1137 			   "Failed to set no of Tx queues: %d\n", tx_queues);
1138 		return err;
1139 	}
1140 
1141 	err = netif_set_real_num_rx_queues(netdev, rx_queues);
1142 	if (err)
1143 		netdev_err(netdev,
1144 			   "Failed to set no of Rx queues: %d\n", rx_queues);
1145 	return err;
1146 }
1147 EXPORT_SYMBOL(otx2_set_real_num_queues);
1148 
1149 static irqreturn_t otx2_q_intr_handler(int irq, void *data)
1150 {
1151 	struct otx2_nic *pf = data;
1152 	u64 val, *ptr;
1153 	u64 qidx = 0;
1154 
1155 	/* CQ */
1156 	for (qidx = 0; qidx < pf->qset.cq_cnt; qidx++) {
1157 		ptr = otx2_get_regaddr(pf, NIX_LF_CQ_OP_INT);
1158 		val = otx2_atomic64_add((qidx << 44), ptr);
1159 
1160 		otx2_write64(pf, NIX_LF_CQ_OP_INT, (qidx << 44) |
1161 			     (val & NIX_CQERRINT_BITS));
1162 		if (!(val & (NIX_CQERRINT_BITS | BIT_ULL(42))))
1163 			continue;
1164 
1165 		if (val & BIT_ULL(42)) {
1166 			netdev_err(pf->netdev, "CQ%lld: error reading NIX_LF_CQ_OP_INT, NIX_LF_ERR_INT 0x%llx\n",
1167 				   qidx, otx2_read64(pf, NIX_LF_ERR_INT));
1168 		} else {
1169 			if (val & BIT_ULL(NIX_CQERRINT_DOOR_ERR))
1170 				netdev_err(pf->netdev, "CQ%lld: Doorbell error",
1171 					   qidx);
1172 			if (val & BIT_ULL(NIX_CQERRINT_CQE_FAULT))
1173 				netdev_err(pf->netdev, "CQ%lld: Memory fault on CQE write to LLC/DRAM",
1174 					   qidx);
1175 		}
1176 
1177 		schedule_work(&pf->reset_task);
1178 	}
1179 
1180 	/* SQ */
1181 	for (qidx = 0; qidx < pf->hw.tx_queues; qidx++) {
1182 		ptr = otx2_get_regaddr(pf, NIX_LF_SQ_OP_INT);
1183 		val = otx2_atomic64_add((qidx << 44), ptr);
1184 		otx2_write64(pf, NIX_LF_SQ_OP_INT, (qidx << 44) |
1185 			     (val & NIX_SQINT_BITS));
1186 
1187 		if (!(val & (NIX_SQINT_BITS | BIT_ULL(42))))
1188 			continue;
1189 
1190 		if (val & BIT_ULL(42)) {
1191 			netdev_err(pf->netdev, "SQ%lld: error reading NIX_LF_SQ_OP_INT, NIX_LF_ERR_INT 0x%llx\n",
1192 				   qidx, otx2_read64(pf, NIX_LF_ERR_INT));
1193 		} else {
1194 			if (val & BIT_ULL(NIX_SQINT_LMT_ERR)) {
1195 				netdev_err(pf->netdev, "SQ%lld: LMT store error NIX_LF_SQ_OP_ERR_DBG:0x%llx",
1196 					   qidx,
1197 					   otx2_read64(pf,
1198 						       NIX_LF_SQ_OP_ERR_DBG));
1199 				otx2_write64(pf, NIX_LF_SQ_OP_ERR_DBG,
1200 					     BIT_ULL(44));
1201 			}
1202 			if (val & BIT_ULL(NIX_SQINT_MNQ_ERR)) {
1203 				netdev_err(pf->netdev, "SQ%lld: Meta-descriptor enqueue error NIX_LF_MNQ_ERR_DGB:0x%llx\n",
1204 					   qidx,
1205 					   otx2_read64(pf, NIX_LF_MNQ_ERR_DBG));
1206 				otx2_write64(pf, NIX_LF_MNQ_ERR_DBG,
1207 					     BIT_ULL(44));
1208 			}
1209 			if (val & BIT_ULL(NIX_SQINT_SEND_ERR)) {
1210 				netdev_err(pf->netdev, "SQ%lld: Send error, NIX_LF_SEND_ERR_DBG 0x%llx",
1211 					   qidx,
1212 					   otx2_read64(pf,
1213 						       NIX_LF_SEND_ERR_DBG));
1214 				otx2_write64(pf, NIX_LF_SEND_ERR_DBG,
1215 					     BIT_ULL(44));
1216 			}
1217 			if (val & BIT_ULL(NIX_SQINT_SQB_ALLOC_FAIL))
1218 				netdev_err(pf->netdev, "SQ%lld: SQB allocation failed",
1219 					   qidx);
1220 		}
1221 
1222 		schedule_work(&pf->reset_task);
1223 	}
1224 
1225 	return IRQ_HANDLED;
1226 }
1227 
1228 static irqreturn_t otx2_cq_intr_handler(int irq, void *cq_irq)
1229 {
1230 	struct otx2_cq_poll *cq_poll = (struct otx2_cq_poll *)cq_irq;
1231 	struct otx2_nic *pf = (struct otx2_nic *)cq_poll->dev;
1232 	int qidx = cq_poll->cint_idx;
1233 
1234 	/* Disable interrupts.
1235 	 *
1236 	 * Completion interrupts behave in a level-triggered interrupt
1237 	 * fashion, and hence have to be cleared only after it is serviced.
1238 	 */
1239 	otx2_write64(pf, NIX_LF_CINTX_ENA_W1C(qidx), BIT_ULL(0));
1240 
1241 	/* Schedule NAPI */
1242 	napi_schedule_irqoff(&cq_poll->napi);
1243 
1244 	return IRQ_HANDLED;
1245 }
1246 
1247 static void otx2_disable_napi(struct otx2_nic *pf)
1248 {
1249 	struct otx2_qset *qset = &pf->qset;
1250 	struct otx2_cq_poll *cq_poll;
1251 	int qidx;
1252 
1253 	for (qidx = 0; qidx < pf->hw.cint_cnt; qidx++) {
1254 		cq_poll = &qset->napi[qidx];
1255 		napi_disable(&cq_poll->napi);
1256 		netif_napi_del(&cq_poll->napi);
1257 	}
1258 }
1259 
1260 static void otx2_free_cq_res(struct otx2_nic *pf)
1261 {
1262 	struct otx2_qset *qset = &pf->qset;
1263 	struct otx2_cq_queue *cq;
1264 	int qidx;
1265 
1266 	/* Disable CQs */
1267 	otx2_ctx_disable(&pf->mbox, NIX_AQ_CTYPE_CQ, false);
1268 	for (qidx = 0; qidx < qset->cq_cnt; qidx++) {
1269 		cq = &qset->cq[qidx];
1270 		qmem_free(pf->dev, cq->cqe);
1271 	}
1272 }
1273 
1274 static void otx2_free_sq_res(struct otx2_nic *pf)
1275 {
1276 	struct otx2_qset *qset = &pf->qset;
1277 	struct otx2_snd_queue *sq;
1278 	int qidx;
1279 
1280 	/* Disable SQs */
1281 	otx2_ctx_disable(&pf->mbox, NIX_AQ_CTYPE_SQ, false);
1282 	/* Free SQB pointers */
1283 	otx2_sq_free_sqbs(pf);
1284 	for (qidx = 0; qidx < pf->hw.tx_queues; qidx++) {
1285 		sq = &qset->sq[qidx];
1286 		qmem_free(pf->dev, sq->sqe);
1287 		qmem_free(pf->dev, sq->tso_hdrs);
1288 		kfree(sq->sg);
1289 		kfree(sq->sqb_ptrs);
1290 	}
1291 }
1292 
1293 static int otx2_get_rbuf_size(struct otx2_nic *pf, int mtu)
1294 {
1295 	int frame_size;
1296 	int total_size;
1297 	int rbuf_size;
1298 
1299 	/* The data transferred by NIX to memory consists of actual packet
1300 	 * plus additional data which has timestamp and/or EDSA/HIGIG2
1301 	 * headers if interface is configured in corresponding modes.
1302 	 * NIX transfers entire data using 6 segments/buffers and writes
1303 	 * a CQE_RX descriptor with those segment addresses. First segment
1304 	 * has additional data prepended to packet. Also software omits a
1305 	 * headroom of 128 bytes and sizeof(struct skb_shared_info) in
1306 	 * each segment. Hence the total size of memory needed
1307 	 * to receive a packet with 'mtu' is:
1308 	 * frame size =  mtu + additional data;
1309 	 * memory = frame_size + (headroom + struct skb_shared_info size) * 6;
1310 	 * each receive buffer size = memory / 6;
1311 	 */
1312 	frame_size = mtu + OTX2_ETH_HLEN + OTX2_HW_TIMESTAMP_LEN;
1313 	total_size = frame_size + (OTX2_HEAD_ROOM +
1314 		     OTX2_DATA_ALIGN(sizeof(struct skb_shared_info))) * 6;
1315 	rbuf_size = total_size / 6;
1316 
1317 	return ALIGN(rbuf_size, 2048);
1318 }
1319 
1320 static int otx2_init_hw_resources(struct otx2_nic *pf)
1321 {
1322 	struct nix_lf_free_req *free_req;
1323 	struct mbox *mbox = &pf->mbox;
1324 	struct otx2_hw *hw = &pf->hw;
1325 	struct msg_req *req;
1326 	int err = 0, lvl;
1327 
1328 	/* Set required NPA LF's pool counts
1329 	 * Auras and Pools are used in a 1:1 mapping,
1330 	 * so, aura count = pool count.
1331 	 */
1332 	hw->rqpool_cnt = hw->rx_queues;
1333 	hw->sqpool_cnt = hw->tx_queues;
1334 	hw->pool_cnt = hw->rqpool_cnt + hw->sqpool_cnt;
1335 
1336 	pf->max_frs = pf->netdev->mtu + OTX2_ETH_HLEN + OTX2_HW_TIMESTAMP_LEN;
1337 
1338 	pf->rbsize = otx2_get_rbuf_size(pf, pf->netdev->mtu);
1339 
1340 	mutex_lock(&mbox->lock);
1341 	/* NPA init */
1342 	err = otx2_config_npa(pf);
1343 	if (err)
1344 		goto exit;
1345 
1346 	/* NIX init */
1347 	err = otx2_config_nix(pf);
1348 	if (err)
1349 		goto err_free_npa_lf;
1350 
1351 	/* Enable backpressure */
1352 	otx2_nix_config_bp(pf, true);
1353 
1354 	/* Init Auras and pools used by NIX RQ, for free buffer ptrs */
1355 	err = otx2_rq_aura_pool_init(pf);
1356 	if (err) {
1357 		mutex_unlock(&mbox->lock);
1358 		goto err_free_nix_lf;
1359 	}
1360 	/* Init Auras and pools used by NIX SQ, for queueing SQEs */
1361 	err = otx2_sq_aura_pool_init(pf);
1362 	if (err) {
1363 		mutex_unlock(&mbox->lock);
1364 		goto err_free_rq_ptrs;
1365 	}
1366 
1367 	err = otx2_txsch_alloc(pf);
1368 	if (err) {
1369 		mutex_unlock(&mbox->lock);
1370 		goto err_free_sq_ptrs;
1371 	}
1372 
1373 	err = otx2_config_nix_queues(pf);
1374 	if (err) {
1375 		mutex_unlock(&mbox->lock);
1376 		goto err_free_txsch;
1377 	}
1378 	for (lvl = 0; lvl < NIX_TXSCH_LVL_CNT; lvl++) {
1379 		err = otx2_txschq_config(pf, lvl);
1380 		if (err) {
1381 			mutex_unlock(&mbox->lock);
1382 			goto err_free_nix_queues;
1383 		}
1384 	}
1385 	mutex_unlock(&mbox->lock);
1386 	return err;
1387 
1388 err_free_nix_queues:
1389 	otx2_free_sq_res(pf);
1390 	otx2_free_cq_res(pf);
1391 	otx2_ctx_disable(mbox, NIX_AQ_CTYPE_RQ, false);
1392 err_free_txsch:
1393 	if (otx2_txschq_stop(pf))
1394 		dev_err(pf->dev, "%s failed to stop TX schedulers\n", __func__);
1395 err_free_sq_ptrs:
1396 	otx2_sq_free_sqbs(pf);
1397 err_free_rq_ptrs:
1398 	otx2_free_aura_ptr(pf, AURA_NIX_RQ);
1399 	otx2_ctx_disable(mbox, NPA_AQ_CTYPE_POOL, true);
1400 	otx2_ctx_disable(mbox, NPA_AQ_CTYPE_AURA, true);
1401 	otx2_aura_pool_free(pf);
1402 err_free_nix_lf:
1403 	mutex_lock(&mbox->lock);
1404 	free_req = otx2_mbox_alloc_msg_nix_lf_free(mbox);
1405 	if (free_req) {
1406 		free_req->flags = NIX_LF_DISABLE_FLOWS;
1407 		if (otx2_sync_mbox_msg(mbox))
1408 			dev_err(pf->dev, "%s failed to free nixlf\n", __func__);
1409 	}
1410 err_free_npa_lf:
1411 	/* Reset NPA LF */
1412 	req = otx2_mbox_alloc_msg_npa_lf_free(mbox);
1413 	if (req) {
1414 		if (otx2_sync_mbox_msg(mbox))
1415 			dev_err(pf->dev, "%s failed to free npalf\n", __func__);
1416 	}
1417 exit:
1418 	mutex_unlock(&mbox->lock);
1419 	return err;
1420 }
1421 
1422 static void otx2_free_hw_resources(struct otx2_nic *pf)
1423 {
1424 	struct otx2_qset *qset = &pf->qset;
1425 	struct nix_lf_free_req *free_req;
1426 	struct mbox *mbox = &pf->mbox;
1427 	struct otx2_cq_queue *cq;
1428 	struct msg_req *req;
1429 	int qidx, err;
1430 
1431 	/* Ensure all SQE are processed */
1432 	otx2_sqb_flush(pf);
1433 
1434 	/* Stop transmission */
1435 	err = otx2_txschq_stop(pf);
1436 	if (err)
1437 		dev_err(pf->dev, "RVUPF: Failed to stop/free TX schedulers\n");
1438 
1439 	mutex_lock(&mbox->lock);
1440 	/* Disable backpressure */
1441 	if (!(pf->pcifunc & RVU_PFVF_FUNC_MASK))
1442 		otx2_nix_config_bp(pf, false);
1443 	mutex_unlock(&mbox->lock);
1444 
1445 	/* Disable RQs */
1446 	otx2_ctx_disable(mbox, NIX_AQ_CTYPE_RQ, false);
1447 
1448 	/*Dequeue all CQEs */
1449 	for (qidx = 0; qidx < qset->cq_cnt; qidx++) {
1450 		cq = &qset->cq[qidx];
1451 		if (cq->cq_type == CQ_RX)
1452 			otx2_cleanup_rx_cqes(pf, cq);
1453 		else
1454 			otx2_cleanup_tx_cqes(pf, cq);
1455 	}
1456 
1457 	otx2_free_sq_res(pf);
1458 
1459 	/* Free RQ buffer pointers*/
1460 	otx2_free_aura_ptr(pf, AURA_NIX_RQ);
1461 
1462 	otx2_free_cq_res(pf);
1463 
1464 	/* Free all ingress bandwidth profiles allocated */
1465 	cn10k_free_all_ipolicers(pf);
1466 
1467 	mutex_lock(&mbox->lock);
1468 	/* Reset NIX LF */
1469 	free_req = otx2_mbox_alloc_msg_nix_lf_free(mbox);
1470 	if (free_req) {
1471 		free_req->flags = NIX_LF_DISABLE_FLOWS;
1472 		if (!(pf->flags & OTX2_FLAG_PF_SHUTDOWN))
1473 			free_req->flags |= NIX_LF_DONT_FREE_TX_VTAG;
1474 		if (otx2_sync_mbox_msg(mbox))
1475 			dev_err(pf->dev, "%s failed to free nixlf\n", __func__);
1476 	}
1477 	mutex_unlock(&mbox->lock);
1478 
1479 	/* Disable NPA Pool and Aura hw context */
1480 	otx2_ctx_disable(mbox, NPA_AQ_CTYPE_POOL, true);
1481 	otx2_ctx_disable(mbox, NPA_AQ_CTYPE_AURA, true);
1482 	otx2_aura_pool_free(pf);
1483 
1484 	mutex_lock(&mbox->lock);
1485 	/* Reset NPA LF */
1486 	req = otx2_mbox_alloc_msg_npa_lf_free(mbox);
1487 	if (req) {
1488 		if (otx2_sync_mbox_msg(mbox))
1489 			dev_err(pf->dev, "%s failed to free npalf\n", __func__);
1490 	}
1491 	mutex_unlock(&mbox->lock);
1492 }
1493 
1494 int otx2_open(struct net_device *netdev)
1495 {
1496 	struct otx2_nic *pf = netdev_priv(netdev);
1497 	struct otx2_cq_poll *cq_poll = NULL;
1498 	struct otx2_qset *qset = &pf->qset;
1499 	int err = 0, qidx, vec;
1500 	char *irq_name;
1501 
1502 	netif_carrier_off(netdev);
1503 
1504 	pf->qset.cq_cnt = pf->hw.rx_queues + pf->hw.tx_queues;
1505 	/* RQ and SQs are mapped to different CQs,
1506 	 * so find out max CQ IRQs (i.e CINTs) needed.
1507 	 */
1508 	pf->hw.cint_cnt = max(pf->hw.rx_queues, pf->hw.tx_queues);
1509 	qset->napi = kcalloc(pf->hw.cint_cnt, sizeof(*cq_poll), GFP_KERNEL);
1510 	if (!qset->napi)
1511 		return -ENOMEM;
1512 
1513 	/* CQ size of RQ */
1514 	qset->rqe_cnt = qset->rqe_cnt ? qset->rqe_cnt : Q_COUNT(Q_SIZE_256);
1515 	/* CQ size of SQ */
1516 	qset->sqe_cnt = qset->sqe_cnt ? qset->sqe_cnt : Q_COUNT(Q_SIZE_4K);
1517 
1518 	err = -ENOMEM;
1519 	qset->cq = kcalloc(pf->qset.cq_cnt,
1520 			   sizeof(struct otx2_cq_queue), GFP_KERNEL);
1521 	if (!qset->cq)
1522 		goto err_free_mem;
1523 
1524 	qset->sq = kcalloc(pf->hw.tx_queues,
1525 			   sizeof(struct otx2_snd_queue), GFP_KERNEL);
1526 	if (!qset->sq)
1527 		goto err_free_mem;
1528 
1529 	qset->rq = kcalloc(pf->hw.rx_queues,
1530 			   sizeof(struct otx2_rcv_queue), GFP_KERNEL);
1531 	if (!qset->rq)
1532 		goto err_free_mem;
1533 
1534 	if (test_bit(CN10K_LMTST, &pf->hw.cap_flag)) {
1535 		/* Reserve LMT lines for NPA AURA batch free */
1536 		pf->hw.npa_lmt_base = (__force u64 *)pf->hw.lmt_base;
1537 		/* Reserve LMT lines for NIX TX */
1538 		pf->hw.nix_lmt_base = (__force u64 *)((u64)pf->hw.npa_lmt_base +
1539 				      (NIX_LMTID_BASE * LMT_LINE_SIZE));
1540 	}
1541 
1542 	err = otx2_init_hw_resources(pf);
1543 	if (err)
1544 		goto err_free_mem;
1545 
1546 	/* Register NAPI handler */
1547 	for (qidx = 0; qidx < pf->hw.cint_cnt; qidx++) {
1548 		cq_poll = &qset->napi[qidx];
1549 		cq_poll->cint_idx = qidx;
1550 		/* RQ0 & SQ0 are mapped to CINT0 and so on..
1551 		 * 'cq_ids[0]' points to RQ's CQ and
1552 		 * 'cq_ids[1]' points to SQ's CQ and
1553 		 */
1554 		cq_poll->cq_ids[CQ_RX] =
1555 			(qidx <  pf->hw.rx_queues) ? qidx : CINT_INVALID_CQ;
1556 		cq_poll->cq_ids[CQ_TX] = (qidx < pf->hw.tx_queues) ?
1557 				      qidx + pf->hw.rx_queues : CINT_INVALID_CQ;
1558 		cq_poll->dev = (void *)pf;
1559 		netif_napi_add(netdev, &cq_poll->napi,
1560 			       otx2_napi_handler, NAPI_POLL_WEIGHT);
1561 		napi_enable(&cq_poll->napi);
1562 	}
1563 
1564 	/* Set maximum frame size allowed in HW */
1565 	err = otx2_hw_set_mtu(pf, netdev->mtu);
1566 	if (err)
1567 		goto err_disable_napi;
1568 
1569 	/* Setup segmentation algorithms, if failed, clear offload capability */
1570 	otx2_setup_segmentation(pf);
1571 
1572 	/* Initialize RSS */
1573 	err = otx2_rss_init(pf);
1574 	if (err)
1575 		goto err_disable_napi;
1576 
1577 	/* Register Queue IRQ handlers */
1578 	vec = pf->hw.nix_msixoff + NIX_LF_QINT_VEC_START;
1579 	irq_name = &pf->hw.irq_name[vec * NAME_SIZE];
1580 
1581 	snprintf(irq_name, NAME_SIZE, "%s-qerr", pf->netdev->name);
1582 
1583 	err = request_irq(pci_irq_vector(pf->pdev, vec),
1584 			  otx2_q_intr_handler, 0, irq_name, pf);
1585 	if (err) {
1586 		dev_err(pf->dev,
1587 			"RVUPF%d: IRQ registration failed for QERR\n",
1588 			rvu_get_pf(pf->pcifunc));
1589 		goto err_disable_napi;
1590 	}
1591 
1592 	/* Enable QINT IRQ */
1593 	otx2_write64(pf, NIX_LF_QINTX_ENA_W1S(0), BIT_ULL(0));
1594 
1595 	/* Register CQ IRQ handlers */
1596 	vec = pf->hw.nix_msixoff + NIX_LF_CINT_VEC_START;
1597 	for (qidx = 0; qidx < pf->hw.cint_cnt; qidx++) {
1598 		irq_name = &pf->hw.irq_name[vec * NAME_SIZE];
1599 
1600 		snprintf(irq_name, NAME_SIZE, "%s-rxtx-%d", pf->netdev->name,
1601 			 qidx);
1602 
1603 		err = request_irq(pci_irq_vector(pf->pdev, vec),
1604 				  otx2_cq_intr_handler, 0, irq_name,
1605 				  &qset->napi[qidx]);
1606 		if (err) {
1607 			dev_err(pf->dev,
1608 				"RVUPF%d: IRQ registration failed for CQ%d\n",
1609 				rvu_get_pf(pf->pcifunc), qidx);
1610 			goto err_free_cints;
1611 		}
1612 		vec++;
1613 
1614 		otx2_config_irq_coalescing(pf, qidx);
1615 
1616 		/* Enable CQ IRQ */
1617 		otx2_write64(pf, NIX_LF_CINTX_INT(qidx), BIT_ULL(0));
1618 		otx2_write64(pf, NIX_LF_CINTX_ENA_W1S(qidx), BIT_ULL(0));
1619 	}
1620 
1621 	otx2_set_cints_affinity(pf);
1622 
1623 	if (pf->flags & OTX2_FLAG_RX_VLAN_SUPPORT)
1624 		otx2_enable_rxvlan(pf, true);
1625 
1626 	/* When reinitializing enable time stamping if it is enabled before */
1627 	if (pf->flags & OTX2_FLAG_TX_TSTAMP_ENABLED) {
1628 		pf->flags &= ~OTX2_FLAG_TX_TSTAMP_ENABLED;
1629 		otx2_config_hw_tx_tstamp(pf, true);
1630 	}
1631 	if (pf->flags & OTX2_FLAG_RX_TSTAMP_ENABLED) {
1632 		pf->flags &= ~OTX2_FLAG_RX_TSTAMP_ENABLED;
1633 		otx2_config_hw_rx_tstamp(pf, true);
1634 	}
1635 
1636 	pf->flags &= ~OTX2_FLAG_INTF_DOWN;
1637 	/* 'intf_down' may be checked on any cpu */
1638 	smp_wmb();
1639 
1640 	/* we have already received link status notification */
1641 	if (pf->linfo.link_up && !(pf->pcifunc & RVU_PFVF_FUNC_MASK))
1642 		otx2_handle_link_event(pf);
1643 
1644 	/* Restore pause frame settings */
1645 	otx2_config_pause_frm(pf);
1646 
1647 	err = otx2_rxtx_enable(pf, true);
1648 	if (err)
1649 		goto err_tx_stop_queues;
1650 
1651 	return 0;
1652 
1653 err_tx_stop_queues:
1654 	netif_tx_stop_all_queues(netdev);
1655 	netif_carrier_off(netdev);
1656 err_free_cints:
1657 	otx2_free_cints(pf, qidx);
1658 	vec = pci_irq_vector(pf->pdev,
1659 			     pf->hw.nix_msixoff + NIX_LF_QINT_VEC_START);
1660 	otx2_write64(pf, NIX_LF_QINTX_ENA_W1C(0), BIT_ULL(0));
1661 	synchronize_irq(vec);
1662 	free_irq(vec, pf);
1663 err_disable_napi:
1664 	otx2_disable_napi(pf);
1665 	otx2_free_hw_resources(pf);
1666 err_free_mem:
1667 	kfree(qset->sq);
1668 	kfree(qset->cq);
1669 	kfree(qset->rq);
1670 	kfree(qset->napi);
1671 	return err;
1672 }
1673 EXPORT_SYMBOL(otx2_open);
1674 
1675 int otx2_stop(struct net_device *netdev)
1676 {
1677 	struct otx2_nic *pf = netdev_priv(netdev);
1678 	struct otx2_cq_poll *cq_poll = NULL;
1679 	struct otx2_qset *qset = &pf->qset;
1680 	struct otx2_rss_info *rss;
1681 	int qidx, vec, wrk;
1682 
1683 	netif_carrier_off(netdev);
1684 	netif_tx_stop_all_queues(netdev);
1685 
1686 	pf->flags |= OTX2_FLAG_INTF_DOWN;
1687 	/* 'intf_down' may be checked on any cpu */
1688 	smp_wmb();
1689 
1690 	/* First stop packet Rx/Tx */
1691 	otx2_rxtx_enable(pf, false);
1692 
1693 	/* Clear RSS enable flag */
1694 	rss = &pf->hw.rss_info;
1695 	rss->enable = false;
1696 
1697 	/* Cleanup Queue IRQ */
1698 	vec = pci_irq_vector(pf->pdev,
1699 			     pf->hw.nix_msixoff + NIX_LF_QINT_VEC_START);
1700 	otx2_write64(pf, NIX_LF_QINTX_ENA_W1C(0), BIT_ULL(0));
1701 	synchronize_irq(vec);
1702 	free_irq(vec, pf);
1703 
1704 	/* Cleanup CQ NAPI and IRQ */
1705 	vec = pf->hw.nix_msixoff + NIX_LF_CINT_VEC_START;
1706 	for (qidx = 0; qidx < pf->hw.cint_cnt; qidx++) {
1707 		/* Disable interrupt */
1708 		otx2_write64(pf, NIX_LF_CINTX_ENA_W1C(qidx), BIT_ULL(0));
1709 
1710 		synchronize_irq(pci_irq_vector(pf->pdev, vec));
1711 
1712 		cq_poll = &qset->napi[qidx];
1713 		napi_synchronize(&cq_poll->napi);
1714 		vec++;
1715 	}
1716 
1717 	netif_tx_disable(netdev);
1718 
1719 	otx2_free_hw_resources(pf);
1720 	otx2_free_cints(pf, pf->hw.cint_cnt);
1721 	otx2_disable_napi(pf);
1722 
1723 	for (qidx = 0; qidx < netdev->num_tx_queues; qidx++)
1724 		netdev_tx_reset_queue(netdev_get_tx_queue(netdev, qidx));
1725 
1726 	for (wrk = 0; wrk < pf->qset.cq_cnt; wrk++)
1727 		cancel_delayed_work_sync(&pf->refill_wrk[wrk].pool_refill_work);
1728 	devm_kfree(pf->dev, pf->refill_wrk);
1729 
1730 	kfree(qset->sq);
1731 	kfree(qset->cq);
1732 	kfree(qset->rq);
1733 	kfree(qset->napi);
1734 	/* Do not clear RQ/SQ ringsize settings */
1735 	memset((void *)qset + offsetof(struct otx2_qset, sqe_cnt), 0,
1736 	       sizeof(*qset) - offsetof(struct otx2_qset, sqe_cnt));
1737 	return 0;
1738 }
1739 EXPORT_SYMBOL(otx2_stop);
1740 
1741 static netdev_tx_t otx2_xmit(struct sk_buff *skb, struct net_device *netdev)
1742 {
1743 	struct otx2_nic *pf = netdev_priv(netdev);
1744 	int qidx = skb_get_queue_mapping(skb);
1745 	struct otx2_snd_queue *sq;
1746 	struct netdev_queue *txq;
1747 
1748 	/* Check for minimum and maximum packet length */
1749 	if (skb->len <= ETH_HLEN ||
1750 	    (!skb_shinfo(skb)->gso_size && skb->len > pf->max_frs)) {
1751 		dev_kfree_skb(skb);
1752 		return NETDEV_TX_OK;
1753 	}
1754 
1755 	sq = &pf->qset.sq[qidx];
1756 	txq = netdev_get_tx_queue(netdev, qidx);
1757 
1758 	if (!otx2_sq_append_skb(netdev, sq, skb, qidx)) {
1759 		netif_tx_stop_queue(txq);
1760 
1761 		/* Check again, incase SQBs got freed up */
1762 		smp_mb();
1763 		if (((sq->num_sqbs - *sq->aura_fc_addr) * sq->sqe_per_sqb)
1764 							> sq->sqe_thresh)
1765 			netif_tx_wake_queue(txq);
1766 
1767 		return NETDEV_TX_BUSY;
1768 	}
1769 
1770 	return NETDEV_TX_OK;
1771 }
1772 
1773 static netdev_features_t otx2_fix_features(struct net_device *dev,
1774 					   netdev_features_t features)
1775 {
1776 	/* check if n-tuple filters are ON */
1777 	if ((features & NETIF_F_HW_TC) && (dev->features & NETIF_F_NTUPLE)) {
1778 		netdev_info(dev, "Disabling n-tuple filters\n");
1779 		features &= ~NETIF_F_NTUPLE;
1780 	}
1781 
1782 	/* check if tc hw offload is ON */
1783 	if ((features & NETIF_F_NTUPLE) && (dev->features & NETIF_F_HW_TC)) {
1784 		netdev_info(dev, "Disabling TC hardware offload\n");
1785 		features &= ~NETIF_F_HW_TC;
1786 	}
1787 
1788 	return features;
1789 }
1790 
1791 static void otx2_set_rx_mode(struct net_device *netdev)
1792 {
1793 	struct otx2_nic *pf = netdev_priv(netdev);
1794 
1795 	queue_work(pf->otx2_wq, &pf->rx_mode_work);
1796 }
1797 
1798 static void otx2_do_set_rx_mode(struct work_struct *work)
1799 {
1800 	struct otx2_nic *pf = container_of(work, struct otx2_nic, rx_mode_work);
1801 	struct net_device *netdev = pf->netdev;
1802 	struct nix_rx_mode *req;
1803 	bool promisc = false;
1804 
1805 	if (!(netdev->flags & IFF_UP))
1806 		return;
1807 
1808 	if ((netdev->flags & IFF_PROMISC) ||
1809 	    (netdev_uc_count(netdev) > OTX2_MAX_UNICAST_FLOWS)) {
1810 		promisc = true;
1811 	}
1812 
1813 	/* Write unicast address to mcam entries or del from mcam */
1814 	if (!promisc && netdev->priv_flags & IFF_UNICAST_FLT)
1815 		__dev_uc_sync(netdev, otx2_add_macfilter, otx2_del_macfilter);
1816 
1817 	mutex_lock(&pf->mbox.lock);
1818 	req = otx2_mbox_alloc_msg_nix_set_rx_mode(&pf->mbox);
1819 	if (!req) {
1820 		mutex_unlock(&pf->mbox.lock);
1821 		return;
1822 	}
1823 
1824 	req->mode = NIX_RX_MODE_UCAST;
1825 
1826 	if (promisc)
1827 		req->mode |= NIX_RX_MODE_PROMISC;
1828 	if (netdev->flags & (IFF_ALLMULTI | IFF_MULTICAST))
1829 		req->mode |= NIX_RX_MODE_ALLMULTI;
1830 
1831 	req->mode |= NIX_RX_MODE_USE_MCE;
1832 
1833 	otx2_sync_mbox_msg(&pf->mbox);
1834 	mutex_unlock(&pf->mbox.lock);
1835 }
1836 
1837 static int otx2_set_features(struct net_device *netdev,
1838 			     netdev_features_t features)
1839 {
1840 	netdev_features_t changed = features ^ netdev->features;
1841 	bool ntuple = !!(features & NETIF_F_NTUPLE);
1842 	struct otx2_nic *pf = netdev_priv(netdev);
1843 
1844 	if ((changed & NETIF_F_LOOPBACK) && netif_running(netdev))
1845 		return otx2_cgx_config_loopback(pf,
1846 						features & NETIF_F_LOOPBACK);
1847 
1848 	if ((changed & NETIF_F_HW_VLAN_CTAG_RX) && netif_running(netdev))
1849 		return otx2_enable_rxvlan(pf,
1850 					  features & NETIF_F_HW_VLAN_CTAG_RX);
1851 
1852 	if ((changed & NETIF_F_NTUPLE) && !ntuple)
1853 		otx2_destroy_ntuple_flows(pf);
1854 
1855 	if ((netdev->features & NETIF_F_HW_TC) > (features & NETIF_F_HW_TC) &&
1856 	    pf->tc_info.num_entries) {
1857 		netdev_err(netdev, "Can't disable TC hardware offload while flows are active\n");
1858 		return -EBUSY;
1859 	}
1860 
1861 	return 0;
1862 }
1863 
1864 static void otx2_reset_task(struct work_struct *work)
1865 {
1866 	struct otx2_nic *pf = container_of(work, struct otx2_nic, reset_task);
1867 
1868 	if (!netif_running(pf->netdev))
1869 		return;
1870 
1871 	rtnl_lock();
1872 	otx2_stop(pf->netdev);
1873 	pf->reset_count++;
1874 	otx2_open(pf->netdev);
1875 	netif_trans_update(pf->netdev);
1876 	rtnl_unlock();
1877 }
1878 
1879 static int otx2_config_hw_rx_tstamp(struct otx2_nic *pfvf, bool enable)
1880 {
1881 	struct msg_req *req;
1882 	int err;
1883 
1884 	if (pfvf->flags & OTX2_FLAG_RX_TSTAMP_ENABLED && enable)
1885 		return 0;
1886 
1887 	mutex_lock(&pfvf->mbox.lock);
1888 	if (enable)
1889 		req = otx2_mbox_alloc_msg_cgx_ptp_rx_enable(&pfvf->mbox);
1890 	else
1891 		req = otx2_mbox_alloc_msg_cgx_ptp_rx_disable(&pfvf->mbox);
1892 	if (!req) {
1893 		mutex_unlock(&pfvf->mbox.lock);
1894 		return -ENOMEM;
1895 	}
1896 
1897 	err = otx2_sync_mbox_msg(&pfvf->mbox);
1898 	if (err) {
1899 		mutex_unlock(&pfvf->mbox.lock);
1900 		return err;
1901 	}
1902 
1903 	mutex_unlock(&pfvf->mbox.lock);
1904 	if (enable)
1905 		pfvf->flags |= OTX2_FLAG_RX_TSTAMP_ENABLED;
1906 	else
1907 		pfvf->flags &= ~OTX2_FLAG_RX_TSTAMP_ENABLED;
1908 	return 0;
1909 }
1910 
1911 static int otx2_config_hw_tx_tstamp(struct otx2_nic *pfvf, bool enable)
1912 {
1913 	struct msg_req *req;
1914 	int err;
1915 
1916 	if (pfvf->flags & OTX2_FLAG_TX_TSTAMP_ENABLED && enable)
1917 		return 0;
1918 
1919 	mutex_lock(&pfvf->mbox.lock);
1920 	if (enable)
1921 		req = otx2_mbox_alloc_msg_nix_lf_ptp_tx_enable(&pfvf->mbox);
1922 	else
1923 		req = otx2_mbox_alloc_msg_nix_lf_ptp_tx_disable(&pfvf->mbox);
1924 	if (!req) {
1925 		mutex_unlock(&pfvf->mbox.lock);
1926 		return -ENOMEM;
1927 	}
1928 
1929 	err = otx2_sync_mbox_msg(&pfvf->mbox);
1930 	if (err) {
1931 		mutex_unlock(&pfvf->mbox.lock);
1932 		return err;
1933 	}
1934 
1935 	mutex_unlock(&pfvf->mbox.lock);
1936 	if (enable)
1937 		pfvf->flags |= OTX2_FLAG_TX_TSTAMP_ENABLED;
1938 	else
1939 		pfvf->flags &= ~OTX2_FLAG_TX_TSTAMP_ENABLED;
1940 	return 0;
1941 }
1942 
1943 static int otx2_config_hwtstamp(struct net_device *netdev, struct ifreq *ifr)
1944 {
1945 	struct otx2_nic *pfvf = netdev_priv(netdev);
1946 	struct hwtstamp_config config;
1947 
1948 	if (!pfvf->ptp)
1949 		return -ENODEV;
1950 
1951 	if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
1952 		return -EFAULT;
1953 
1954 	/* reserved for future extensions */
1955 	if (config.flags)
1956 		return -EINVAL;
1957 
1958 	switch (config.tx_type) {
1959 	case HWTSTAMP_TX_OFF:
1960 		otx2_config_hw_tx_tstamp(pfvf, false);
1961 		break;
1962 	case HWTSTAMP_TX_ON:
1963 		otx2_config_hw_tx_tstamp(pfvf, true);
1964 		break;
1965 	default:
1966 		return -ERANGE;
1967 	}
1968 
1969 	switch (config.rx_filter) {
1970 	case HWTSTAMP_FILTER_NONE:
1971 		otx2_config_hw_rx_tstamp(pfvf, false);
1972 		break;
1973 	case HWTSTAMP_FILTER_ALL:
1974 	case HWTSTAMP_FILTER_SOME:
1975 	case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
1976 	case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
1977 	case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
1978 	case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
1979 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
1980 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
1981 	case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
1982 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
1983 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
1984 	case HWTSTAMP_FILTER_PTP_V2_EVENT:
1985 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
1986 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
1987 		otx2_config_hw_rx_tstamp(pfvf, true);
1988 		config.rx_filter = HWTSTAMP_FILTER_ALL;
1989 		break;
1990 	default:
1991 		return -ERANGE;
1992 	}
1993 
1994 	memcpy(&pfvf->tstamp, &config, sizeof(config));
1995 
1996 	return copy_to_user(ifr->ifr_data, &config,
1997 			    sizeof(config)) ? -EFAULT : 0;
1998 }
1999 
2000 static int otx2_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
2001 {
2002 	struct otx2_nic *pfvf = netdev_priv(netdev);
2003 	struct hwtstamp_config *cfg = &pfvf->tstamp;
2004 
2005 	switch (cmd) {
2006 	case SIOCSHWTSTAMP:
2007 		return otx2_config_hwtstamp(netdev, req);
2008 	case SIOCGHWTSTAMP:
2009 		return copy_to_user(req->ifr_data, cfg,
2010 				    sizeof(*cfg)) ? -EFAULT : 0;
2011 	default:
2012 		return -EOPNOTSUPP;
2013 	}
2014 }
2015 
2016 static int otx2_do_set_vf_mac(struct otx2_nic *pf, int vf, const u8 *mac)
2017 {
2018 	struct npc_install_flow_req *req;
2019 	int err;
2020 
2021 	mutex_lock(&pf->mbox.lock);
2022 	req = otx2_mbox_alloc_msg_npc_install_flow(&pf->mbox);
2023 	if (!req) {
2024 		err = -ENOMEM;
2025 		goto out;
2026 	}
2027 
2028 	ether_addr_copy(req->packet.dmac, mac);
2029 	eth_broadcast_addr((u8 *)&req->mask.dmac);
2030 	req->features = BIT_ULL(NPC_DMAC);
2031 	req->channel = pf->hw.rx_chan_base;
2032 	req->intf = NIX_INTF_RX;
2033 	req->default_rule = 1;
2034 	req->append = 1;
2035 	req->vf = vf + 1;
2036 	req->op = NIX_RX_ACTION_DEFAULT;
2037 
2038 	err = otx2_sync_mbox_msg(&pf->mbox);
2039 out:
2040 	mutex_unlock(&pf->mbox.lock);
2041 	return err;
2042 }
2043 
2044 static int otx2_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
2045 {
2046 	struct otx2_nic *pf = netdev_priv(netdev);
2047 	struct pci_dev *pdev = pf->pdev;
2048 	struct otx2_vf_config *config;
2049 	int ret;
2050 
2051 	if (!netif_running(netdev))
2052 		return -EAGAIN;
2053 
2054 	if (vf >= pf->total_vfs)
2055 		return -EINVAL;
2056 
2057 	if (!is_valid_ether_addr(mac))
2058 		return -EINVAL;
2059 
2060 	config = &pf->vf_configs[vf];
2061 	ether_addr_copy(config->mac, mac);
2062 
2063 	ret = otx2_do_set_vf_mac(pf, vf, mac);
2064 	if (ret == 0)
2065 		dev_info(&pdev->dev,
2066 			 "Load/Reload VF driver\n");
2067 
2068 	return ret;
2069 }
2070 
2071 static int otx2_do_set_vf_vlan(struct otx2_nic *pf, int vf, u16 vlan, u8 qos,
2072 			       __be16 proto)
2073 {
2074 	struct otx2_flow_config *flow_cfg = pf->flow_cfg;
2075 	struct nix_vtag_config_rsp *vtag_rsp;
2076 	struct npc_delete_flow_req *del_req;
2077 	struct nix_vtag_config *vtag_req;
2078 	struct npc_install_flow_req *req;
2079 	struct otx2_vf_config *config;
2080 	int err = 0;
2081 	u32 idx;
2082 
2083 	config = &pf->vf_configs[vf];
2084 
2085 	if (!vlan && !config->vlan)
2086 		goto out;
2087 
2088 	mutex_lock(&pf->mbox.lock);
2089 
2090 	/* free old tx vtag entry */
2091 	if (config->vlan) {
2092 		vtag_req = otx2_mbox_alloc_msg_nix_vtag_cfg(&pf->mbox);
2093 		if (!vtag_req) {
2094 			err = -ENOMEM;
2095 			goto out;
2096 		}
2097 		vtag_req->cfg_type = 0;
2098 		vtag_req->tx.free_vtag0 = 1;
2099 		vtag_req->tx.vtag0_idx = config->tx_vtag_idx;
2100 
2101 		err = otx2_sync_mbox_msg(&pf->mbox);
2102 		if (err)
2103 			goto out;
2104 	}
2105 
2106 	if (!vlan && config->vlan) {
2107 		/* rx */
2108 		del_req = otx2_mbox_alloc_msg_npc_delete_flow(&pf->mbox);
2109 		if (!del_req) {
2110 			err = -ENOMEM;
2111 			goto out;
2112 		}
2113 		idx = ((vf * OTX2_PER_VF_VLAN_FLOWS) + OTX2_VF_VLAN_RX_INDEX);
2114 		del_req->entry =
2115 			flow_cfg->def_ent[flow_cfg->vf_vlan_offset + idx];
2116 		err = otx2_sync_mbox_msg(&pf->mbox);
2117 		if (err)
2118 			goto out;
2119 
2120 		/* tx */
2121 		del_req = otx2_mbox_alloc_msg_npc_delete_flow(&pf->mbox);
2122 		if (!del_req) {
2123 			err = -ENOMEM;
2124 			goto out;
2125 		}
2126 		idx = ((vf * OTX2_PER_VF_VLAN_FLOWS) + OTX2_VF_VLAN_TX_INDEX);
2127 		del_req->entry =
2128 			flow_cfg->def_ent[flow_cfg->vf_vlan_offset + idx];
2129 		err = otx2_sync_mbox_msg(&pf->mbox);
2130 
2131 		goto out;
2132 	}
2133 
2134 	/* rx */
2135 	req = otx2_mbox_alloc_msg_npc_install_flow(&pf->mbox);
2136 	if (!req) {
2137 		err = -ENOMEM;
2138 		goto out;
2139 	}
2140 
2141 	idx = ((vf * OTX2_PER_VF_VLAN_FLOWS) + OTX2_VF_VLAN_RX_INDEX);
2142 	req->entry = flow_cfg->def_ent[flow_cfg->vf_vlan_offset + idx];
2143 	req->packet.vlan_tci = htons(vlan);
2144 	req->mask.vlan_tci = htons(VLAN_VID_MASK);
2145 	/* af fills the destination mac addr */
2146 	eth_broadcast_addr((u8 *)&req->mask.dmac);
2147 	req->features = BIT_ULL(NPC_OUTER_VID) | BIT_ULL(NPC_DMAC);
2148 	req->channel = pf->hw.rx_chan_base;
2149 	req->intf = NIX_INTF_RX;
2150 	req->vf = vf + 1;
2151 	req->op = NIX_RX_ACTION_DEFAULT;
2152 	req->vtag0_valid = true;
2153 	req->vtag0_type = NIX_AF_LFX_RX_VTAG_TYPE7;
2154 	req->set_cntr = 1;
2155 
2156 	err = otx2_sync_mbox_msg(&pf->mbox);
2157 	if (err)
2158 		goto out;
2159 
2160 	/* tx */
2161 	vtag_req = otx2_mbox_alloc_msg_nix_vtag_cfg(&pf->mbox);
2162 	if (!vtag_req) {
2163 		err = -ENOMEM;
2164 		goto out;
2165 	}
2166 
2167 	/* configure tx vtag params */
2168 	vtag_req->vtag_size = VTAGSIZE_T4;
2169 	vtag_req->cfg_type = 0; /* tx vlan cfg */
2170 	vtag_req->tx.cfg_vtag0 = 1;
2171 	vtag_req->tx.vtag0 = ((u64)ntohs(proto) << 16) | vlan;
2172 
2173 	err = otx2_sync_mbox_msg(&pf->mbox);
2174 	if (err)
2175 		goto out;
2176 
2177 	vtag_rsp = (struct nix_vtag_config_rsp *)otx2_mbox_get_rsp
2178 			(&pf->mbox.mbox, 0, &vtag_req->hdr);
2179 	if (IS_ERR(vtag_rsp)) {
2180 		err = PTR_ERR(vtag_rsp);
2181 		goto out;
2182 	}
2183 	config->tx_vtag_idx = vtag_rsp->vtag0_idx;
2184 
2185 	req = otx2_mbox_alloc_msg_npc_install_flow(&pf->mbox);
2186 	if (!req) {
2187 		err = -ENOMEM;
2188 		goto out;
2189 	}
2190 
2191 	eth_zero_addr((u8 *)&req->mask.dmac);
2192 	idx = ((vf * OTX2_PER_VF_VLAN_FLOWS) + OTX2_VF_VLAN_TX_INDEX);
2193 	req->entry = flow_cfg->def_ent[flow_cfg->vf_vlan_offset + idx];
2194 	req->features = BIT_ULL(NPC_DMAC);
2195 	req->channel = pf->hw.tx_chan_base;
2196 	req->intf = NIX_INTF_TX;
2197 	req->vf = vf + 1;
2198 	req->op = NIX_TX_ACTIONOP_UCAST_DEFAULT;
2199 	req->vtag0_def = vtag_rsp->vtag0_idx;
2200 	req->vtag0_op = VTAG_INSERT;
2201 	req->set_cntr = 1;
2202 
2203 	err = otx2_sync_mbox_msg(&pf->mbox);
2204 out:
2205 	config->vlan = vlan;
2206 	mutex_unlock(&pf->mbox.lock);
2207 	return err;
2208 }
2209 
2210 static int otx2_set_vf_vlan(struct net_device *netdev, int vf, u16 vlan, u8 qos,
2211 			    __be16 proto)
2212 {
2213 	struct otx2_nic *pf = netdev_priv(netdev);
2214 	struct pci_dev *pdev = pf->pdev;
2215 
2216 	if (!netif_running(netdev))
2217 		return -EAGAIN;
2218 
2219 	if (vf >= pci_num_vf(pdev))
2220 		return -EINVAL;
2221 
2222 	/* qos is currently unsupported */
2223 	if (vlan >= VLAN_N_VID || qos)
2224 		return -EINVAL;
2225 
2226 	if (proto != htons(ETH_P_8021Q))
2227 		return -EPROTONOSUPPORT;
2228 
2229 	if (!(pf->flags & OTX2_FLAG_VF_VLAN_SUPPORT))
2230 		return -EOPNOTSUPP;
2231 
2232 	return otx2_do_set_vf_vlan(pf, vf, vlan, qos, proto);
2233 }
2234 
2235 static int otx2_get_vf_config(struct net_device *netdev, int vf,
2236 			      struct ifla_vf_info *ivi)
2237 {
2238 	struct otx2_nic *pf = netdev_priv(netdev);
2239 	struct pci_dev *pdev = pf->pdev;
2240 	struct otx2_vf_config *config;
2241 
2242 	if (!netif_running(netdev))
2243 		return -EAGAIN;
2244 
2245 	if (vf >= pci_num_vf(pdev))
2246 		return -EINVAL;
2247 
2248 	config = &pf->vf_configs[vf];
2249 	ivi->vf = vf;
2250 	ether_addr_copy(ivi->mac, config->mac);
2251 	ivi->vlan = config->vlan;
2252 	ivi->trusted = config->trusted;
2253 
2254 	return 0;
2255 }
2256 
2257 static int otx2_set_vf_permissions(struct otx2_nic *pf, int vf,
2258 				   int req_perm)
2259 {
2260 	struct set_vf_perm *req;
2261 	int rc;
2262 
2263 	mutex_lock(&pf->mbox.lock);
2264 	req = otx2_mbox_alloc_msg_set_vf_perm(&pf->mbox);
2265 	if (!req) {
2266 		rc = -ENOMEM;
2267 		goto out;
2268 	}
2269 
2270 	/* Let AF reset VF permissions as sriov is disabled */
2271 	if (req_perm == OTX2_RESET_VF_PERM) {
2272 		req->flags |= RESET_VF_PERM;
2273 	} else if (req_perm == OTX2_TRUSTED_VF) {
2274 		if (pf->vf_configs[vf].trusted)
2275 			req->flags |= VF_TRUSTED;
2276 	}
2277 
2278 	req->vf = vf;
2279 	rc = otx2_sync_mbox_msg(&pf->mbox);
2280 out:
2281 	mutex_unlock(&pf->mbox.lock);
2282 	return rc;
2283 }
2284 
2285 static int otx2_ndo_set_vf_trust(struct net_device *netdev, int vf,
2286 				 bool enable)
2287 {
2288 	struct otx2_nic *pf = netdev_priv(netdev);
2289 	struct pci_dev *pdev = pf->pdev;
2290 	int rc;
2291 
2292 	if (vf >= pci_num_vf(pdev))
2293 		return -EINVAL;
2294 
2295 	if (pf->vf_configs[vf].trusted == enable)
2296 		return 0;
2297 
2298 	pf->vf_configs[vf].trusted = enable;
2299 	rc = otx2_set_vf_permissions(pf, vf, OTX2_TRUSTED_VF);
2300 
2301 	if (rc)
2302 		pf->vf_configs[vf].trusted = !enable;
2303 	else
2304 		netdev_info(pf->netdev, "VF %d is %strusted\n",
2305 			    vf, enable ? "" : "not ");
2306 	return rc;
2307 }
2308 
2309 static const struct net_device_ops otx2_netdev_ops = {
2310 	.ndo_open		= otx2_open,
2311 	.ndo_stop		= otx2_stop,
2312 	.ndo_start_xmit		= otx2_xmit,
2313 	.ndo_fix_features	= otx2_fix_features,
2314 	.ndo_set_mac_address    = otx2_set_mac_address,
2315 	.ndo_change_mtu		= otx2_change_mtu,
2316 	.ndo_set_rx_mode	= otx2_set_rx_mode,
2317 	.ndo_set_features	= otx2_set_features,
2318 	.ndo_tx_timeout		= otx2_tx_timeout,
2319 	.ndo_get_stats64	= otx2_get_stats64,
2320 	.ndo_do_ioctl		= otx2_ioctl,
2321 	.ndo_set_vf_mac		= otx2_set_vf_mac,
2322 	.ndo_set_vf_vlan	= otx2_set_vf_vlan,
2323 	.ndo_get_vf_config	= otx2_get_vf_config,
2324 	.ndo_setup_tc		= otx2_setup_tc,
2325 	.ndo_set_vf_trust	= otx2_ndo_set_vf_trust,
2326 };
2327 
2328 static int otx2_wq_init(struct otx2_nic *pf)
2329 {
2330 	pf->otx2_wq = create_singlethread_workqueue("otx2_wq");
2331 	if (!pf->otx2_wq)
2332 		return -ENOMEM;
2333 
2334 	INIT_WORK(&pf->rx_mode_work, otx2_do_set_rx_mode);
2335 	INIT_WORK(&pf->reset_task, otx2_reset_task);
2336 	return 0;
2337 }
2338 
2339 static int otx2_check_pf_usable(struct otx2_nic *nic)
2340 {
2341 	u64 rev;
2342 
2343 	rev = otx2_read64(nic, RVU_PF_BLOCK_ADDRX_DISC(BLKADDR_RVUM));
2344 	rev = (rev >> 12) & 0xFF;
2345 	/* Check if AF has setup revision for RVUM block,
2346 	 * otherwise this driver probe should be deferred
2347 	 * until AF driver comes up.
2348 	 */
2349 	if (!rev) {
2350 		dev_warn(nic->dev,
2351 			 "AF is not initialized, deferring probe\n");
2352 		return -EPROBE_DEFER;
2353 	}
2354 	return 0;
2355 }
2356 
2357 static int otx2_realloc_msix_vectors(struct otx2_nic *pf)
2358 {
2359 	struct otx2_hw *hw = &pf->hw;
2360 	int num_vec, err;
2361 
2362 	/* NPA interrupts are inot registered, so alloc only
2363 	 * upto NIX vector offset.
2364 	 */
2365 	num_vec = hw->nix_msixoff;
2366 	num_vec += NIX_LF_CINT_VEC_START + hw->max_queues;
2367 
2368 	otx2_disable_mbox_intr(pf);
2369 	pci_free_irq_vectors(hw->pdev);
2370 	err = pci_alloc_irq_vectors(hw->pdev, num_vec, num_vec, PCI_IRQ_MSIX);
2371 	if (err < 0) {
2372 		dev_err(pf->dev, "%s: Failed to realloc %d IRQ vectors\n",
2373 			__func__, num_vec);
2374 		return err;
2375 	}
2376 
2377 	return otx2_register_mbox_intr(pf, false);
2378 }
2379 
2380 static int otx2_sriov_vfcfg_init(struct otx2_nic *pf)
2381 {
2382 	int i;
2383 
2384 	pf->vf_configs = devm_kcalloc(pf->dev, pf->total_vfs,
2385 				      sizeof(struct otx2_vf_config),
2386 				      GFP_KERNEL);
2387 	if (!pf->vf_configs)
2388 		return -ENOMEM;
2389 
2390 	for (i = 0; i < pf->total_vfs; i++) {
2391 		pf->vf_configs[i].pf = pf;
2392 		pf->vf_configs[i].intf_down = true;
2393 		pf->vf_configs[i].trusted = false;
2394 		INIT_DELAYED_WORK(&pf->vf_configs[i].link_event_work,
2395 				  otx2_vf_link_event_task);
2396 	}
2397 
2398 	return 0;
2399 }
2400 
2401 static void otx2_sriov_vfcfg_cleanup(struct otx2_nic *pf)
2402 {
2403 	int i;
2404 
2405 	if (!pf->vf_configs)
2406 		return;
2407 
2408 	for (i = 0; i < pf->total_vfs; i++) {
2409 		cancel_delayed_work_sync(&pf->vf_configs[i].link_event_work);
2410 		otx2_set_vf_permissions(pf, i, OTX2_RESET_VF_PERM);
2411 	}
2412 }
2413 
2414 static int otx2_probe(struct pci_dev *pdev, const struct pci_device_id *id)
2415 {
2416 	struct device *dev = &pdev->dev;
2417 	struct net_device *netdev;
2418 	struct otx2_nic *pf;
2419 	struct otx2_hw *hw;
2420 	int err, qcount;
2421 	int num_vec;
2422 
2423 	err = pcim_enable_device(pdev);
2424 	if (err) {
2425 		dev_err(dev, "Failed to enable PCI device\n");
2426 		return err;
2427 	}
2428 
2429 	err = pci_request_regions(pdev, DRV_NAME);
2430 	if (err) {
2431 		dev_err(dev, "PCI request regions failed 0x%x\n", err);
2432 		return err;
2433 	}
2434 
2435 	err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(48));
2436 	if (err) {
2437 		dev_err(dev, "DMA mask config failed, abort\n");
2438 		goto err_release_regions;
2439 	}
2440 
2441 	pci_set_master(pdev);
2442 
2443 	/* Set number of queues */
2444 	qcount = min_t(int, num_online_cpus(), OTX2_MAX_CQ_CNT);
2445 
2446 	netdev = alloc_etherdev_mqs(sizeof(*pf), qcount, qcount);
2447 	if (!netdev) {
2448 		err = -ENOMEM;
2449 		goto err_release_regions;
2450 	}
2451 
2452 	pci_set_drvdata(pdev, netdev);
2453 	SET_NETDEV_DEV(netdev, &pdev->dev);
2454 	pf = netdev_priv(netdev);
2455 	pf->netdev = netdev;
2456 	pf->pdev = pdev;
2457 	pf->dev = dev;
2458 	pf->total_vfs = pci_sriov_get_totalvfs(pdev);
2459 	pf->flags |= OTX2_FLAG_INTF_DOWN;
2460 
2461 	hw = &pf->hw;
2462 	hw->pdev = pdev;
2463 	hw->rx_queues = qcount;
2464 	hw->tx_queues = qcount;
2465 	hw->max_queues = qcount;
2466 
2467 	num_vec = pci_msix_vec_count(pdev);
2468 	hw->irq_name = devm_kmalloc_array(&hw->pdev->dev, num_vec, NAME_SIZE,
2469 					  GFP_KERNEL);
2470 	if (!hw->irq_name) {
2471 		err = -ENOMEM;
2472 		goto err_free_netdev;
2473 	}
2474 
2475 	hw->affinity_mask = devm_kcalloc(&hw->pdev->dev, num_vec,
2476 					 sizeof(cpumask_var_t), GFP_KERNEL);
2477 	if (!hw->affinity_mask) {
2478 		err = -ENOMEM;
2479 		goto err_free_netdev;
2480 	}
2481 
2482 	/* Map CSRs */
2483 	pf->reg_base = pcim_iomap(pdev, PCI_CFG_REG_BAR_NUM, 0);
2484 	if (!pf->reg_base) {
2485 		dev_err(dev, "Unable to map physical function CSRs, aborting\n");
2486 		err = -ENOMEM;
2487 		goto err_free_netdev;
2488 	}
2489 
2490 	err = otx2_check_pf_usable(pf);
2491 	if (err)
2492 		goto err_free_netdev;
2493 
2494 	err = pci_alloc_irq_vectors(hw->pdev, RVU_PF_INT_VEC_CNT,
2495 				    RVU_PF_INT_VEC_CNT, PCI_IRQ_MSIX);
2496 	if (err < 0) {
2497 		dev_err(dev, "%s: Failed to alloc %d IRQ vectors\n",
2498 			__func__, num_vec);
2499 		goto err_free_netdev;
2500 	}
2501 
2502 	otx2_setup_dev_hw_settings(pf);
2503 
2504 	/* Init PF <=> AF mailbox stuff */
2505 	err = otx2_pfaf_mbox_init(pf);
2506 	if (err)
2507 		goto err_free_irq_vectors;
2508 
2509 	/* Register mailbox interrupt */
2510 	err = otx2_register_mbox_intr(pf, true);
2511 	if (err)
2512 		goto err_mbox_destroy;
2513 
2514 	/* Request AF to attach NPA and NIX LFs to this PF.
2515 	 * NIX and NPA LFs are needed for this PF to function as a NIC.
2516 	 */
2517 	err = otx2_attach_npa_nix(pf);
2518 	if (err)
2519 		goto err_disable_mbox_intr;
2520 
2521 	err = otx2_realloc_msix_vectors(pf);
2522 	if (err)
2523 		goto err_detach_rsrc;
2524 
2525 	err = otx2_set_real_num_queues(netdev, hw->tx_queues, hw->rx_queues);
2526 	if (err)
2527 		goto err_detach_rsrc;
2528 
2529 	err = cn10k_pf_lmtst_init(pf);
2530 	if (err)
2531 		goto err_detach_rsrc;
2532 
2533 	/* Assign default mac address */
2534 	otx2_get_mac_from_af(netdev);
2535 
2536 	/* Don't check for error.  Proceed without ptp */
2537 	otx2_ptp_init(pf);
2538 
2539 	/* NPA's pool is a stack to which SW frees buffer pointers via Aura.
2540 	 * HW allocates buffer pointer from stack and uses it for DMA'ing
2541 	 * ingress packet. In some scenarios HW can free back allocated buffer
2542 	 * pointers to pool. This makes it impossible for SW to maintain a
2543 	 * parallel list where physical addresses of buffer pointers (IOVAs)
2544 	 * given to HW can be saved for later reference.
2545 	 *
2546 	 * So the only way to convert Rx packet's buffer address is to use
2547 	 * IOMMU's iova_to_phys() handler which translates the address by
2548 	 * walking through the translation tables.
2549 	 */
2550 	pf->iommu_domain = iommu_get_domain_for_dev(dev);
2551 
2552 	netdev->hw_features = (NETIF_F_RXCSUM | NETIF_F_IP_CSUM |
2553 			       NETIF_F_IPV6_CSUM | NETIF_F_RXHASH |
2554 			       NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
2555 			       NETIF_F_GSO_UDP_L4);
2556 	netdev->features |= netdev->hw_features;
2557 
2558 	netdev->hw_features |= NETIF_F_LOOPBACK | NETIF_F_RXALL;
2559 
2560 	err = otx2_mcam_flow_init(pf);
2561 	if (err)
2562 		goto err_ptp_destroy;
2563 
2564 	if (pf->flags & OTX2_FLAG_NTUPLE_SUPPORT)
2565 		netdev->hw_features |= NETIF_F_NTUPLE;
2566 
2567 	if (pf->flags & OTX2_FLAG_UCAST_FLTR_SUPPORT)
2568 		netdev->priv_flags |= IFF_UNICAST_FLT;
2569 
2570 	/* Support TSO on tag interface */
2571 	netdev->vlan_features |= netdev->features;
2572 	netdev->hw_features  |= NETIF_F_HW_VLAN_CTAG_TX |
2573 				NETIF_F_HW_VLAN_STAG_TX;
2574 	if (pf->flags & OTX2_FLAG_RX_VLAN_SUPPORT)
2575 		netdev->hw_features |= NETIF_F_HW_VLAN_CTAG_RX |
2576 				       NETIF_F_HW_VLAN_STAG_RX;
2577 	netdev->features |= netdev->hw_features;
2578 
2579 	/* HW supports tc offload but mutually exclusive with n-tuple filters */
2580 	if (pf->flags & OTX2_FLAG_TC_FLOWER_SUPPORT)
2581 		netdev->hw_features |= NETIF_F_HW_TC;
2582 
2583 	netdev->gso_max_segs = OTX2_MAX_GSO_SEGS;
2584 	netdev->watchdog_timeo = OTX2_TX_TIMEOUT;
2585 
2586 	netdev->netdev_ops = &otx2_netdev_ops;
2587 
2588 	/* MTU range: 64 - 9190 */
2589 	netdev->min_mtu = OTX2_MIN_MTU;
2590 	netdev->max_mtu = otx2_get_max_mtu(pf);
2591 
2592 	err = register_netdev(netdev);
2593 	if (err) {
2594 		dev_err(dev, "Failed to register netdevice\n");
2595 		goto err_del_mcam_entries;
2596 	}
2597 
2598 	err = otx2_wq_init(pf);
2599 	if (err)
2600 		goto err_unreg_netdev;
2601 
2602 	otx2_set_ethtool_ops(netdev);
2603 
2604 	err = otx2_init_tc(pf);
2605 	if (err)
2606 		goto err_mcam_flow_del;
2607 
2608 	/* Initialize SR-IOV resources */
2609 	err = otx2_sriov_vfcfg_init(pf);
2610 	if (err)
2611 		goto err_pf_sriov_init;
2612 
2613 	/* Enable link notifications */
2614 	otx2_cgx_config_linkevents(pf, true);
2615 
2616 	/* Enable pause frames by default */
2617 	pf->flags |= OTX2_FLAG_RX_PAUSE_ENABLED;
2618 	pf->flags |= OTX2_FLAG_TX_PAUSE_ENABLED;
2619 
2620 	return 0;
2621 
2622 err_pf_sriov_init:
2623 	otx2_shutdown_tc(pf);
2624 err_mcam_flow_del:
2625 	otx2_mcam_flow_del(pf);
2626 err_unreg_netdev:
2627 	unregister_netdev(netdev);
2628 err_del_mcam_entries:
2629 	otx2_mcam_flow_del(pf);
2630 err_ptp_destroy:
2631 	otx2_ptp_destroy(pf);
2632 err_detach_rsrc:
2633 	if (hw->lmt_base)
2634 		iounmap(hw->lmt_base);
2635 	otx2_detach_resources(&pf->mbox);
2636 err_disable_mbox_intr:
2637 	otx2_disable_mbox_intr(pf);
2638 err_mbox_destroy:
2639 	otx2_pfaf_mbox_destroy(pf);
2640 err_free_irq_vectors:
2641 	pci_free_irq_vectors(hw->pdev);
2642 err_free_netdev:
2643 	pci_set_drvdata(pdev, NULL);
2644 	free_netdev(netdev);
2645 err_release_regions:
2646 	pci_release_regions(pdev);
2647 	return err;
2648 }
2649 
2650 static void otx2_vf_link_event_task(struct work_struct *work)
2651 {
2652 	struct otx2_vf_config *config;
2653 	struct cgx_link_info_msg *req;
2654 	struct mbox_msghdr *msghdr;
2655 	struct otx2_nic *pf;
2656 	int vf_idx;
2657 
2658 	config = container_of(work, struct otx2_vf_config,
2659 			      link_event_work.work);
2660 	vf_idx = config - config->pf->vf_configs;
2661 	pf = config->pf;
2662 
2663 	msghdr = otx2_mbox_alloc_msg_rsp(&pf->mbox_pfvf[0].mbox_up, vf_idx,
2664 					 sizeof(*req), sizeof(struct msg_rsp));
2665 	if (!msghdr) {
2666 		dev_err(pf->dev, "Failed to create VF%d link event\n", vf_idx);
2667 		return;
2668 	}
2669 
2670 	req = (struct cgx_link_info_msg *)msghdr;
2671 	req->hdr.id = MBOX_MSG_CGX_LINK_EVENT;
2672 	req->hdr.sig = OTX2_MBOX_REQ_SIG;
2673 	memcpy(&req->link_info, &pf->linfo, sizeof(req->link_info));
2674 
2675 	otx2_sync_mbox_up_msg(&pf->mbox_pfvf[0], vf_idx);
2676 }
2677 
2678 static int otx2_sriov_enable(struct pci_dev *pdev, int numvfs)
2679 {
2680 	struct net_device *netdev = pci_get_drvdata(pdev);
2681 	struct otx2_nic *pf = netdev_priv(netdev);
2682 	int ret;
2683 
2684 	/* Init PF <=> VF mailbox stuff */
2685 	ret = otx2_pfvf_mbox_init(pf, numvfs);
2686 	if (ret)
2687 		return ret;
2688 
2689 	ret = otx2_register_pfvf_mbox_intr(pf, numvfs);
2690 	if (ret)
2691 		goto free_mbox;
2692 
2693 	ret = otx2_pf_flr_init(pf, numvfs);
2694 	if (ret)
2695 		goto free_intr;
2696 
2697 	ret = otx2_register_flr_me_intr(pf, numvfs);
2698 	if (ret)
2699 		goto free_flr;
2700 
2701 	ret = pci_enable_sriov(pdev, numvfs);
2702 	if (ret)
2703 		goto free_flr_intr;
2704 
2705 	return numvfs;
2706 free_flr_intr:
2707 	otx2_disable_flr_me_intr(pf);
2708 free_flr:
2709 	otx2_flr_wq_destroy(pf);
2710 free_intr:
2711 	otx2_disable_pfvf_mbox_intr(pf, numvfs);
2712 free_mbox:
2713 	otx2_pfvf_mbox_destroy(pf);
2714 	return ret;
2715 }
2716 
2717 static int otx2_sriov_disable(struct pci_dev *pdev)
2718 {
2719 	struct net_device *netdev = pci_get_drvdata(pdev);
2720 	struct otx2_nic *pf = netdev_priv(netdev);
2721 	int numvfs = pci_num_vf(pdev);
2722 
2723 	if (!numvfs)
2724 		return 0;
2725 
2726 	pci_disable_sriov(pdev);
2727 
2728 	otx2_disable_flr_me_intr(pf);
2729 	otx2_flr_wq_destroy(pf);
2730 	otx2_disable_pfvf_mbox_intr(pf, numvfs);
2731 	otx2_pfvf_mbox_destroy(pf);
2732 
2733 	return 0;
2734 }
2735 
2736 static int otx2_sriov_configure(struct pci_dev *pdev, int numvfs)
2737 {
2738 	if (numvfs == 0)
2739 		return otx2_sriov_disable(pdev);
2740 	else
2741 		return otx2_sriov_enable(pdev, numvfs);
2742 }
2743 
2744 static void otx2_remove(struct pci_dev *pdev)
2745 {
2746 	struct net_device *netdev = pci_get_drvdata(pdev);
2747 	struct otx2_nic *pf;
2748 
2749 	if (!netdev)
2750 		return;
2751 
2752 	pf = netdev_priv(netdev);
2753 
2754 	pf->flags |= OTX2_FLAG_PF_SHUTDOWN;
2755 
2756 	if (pf->flags & OTX2_FLAG_TX_TSTAMP_ENABLED)
2757 		otx2_config_hw_tx_tstamp(pf, false);
2758 	if (pf->flags & OTX2_FLAG_RX_TSTAMP_ENABLED)
2759 		otx2_config_hw_rx_tstamp(pf, false);
2760 
2761 	cancel_work_sync(&pf->reset_task);
2762 	/* Disable link notifications */
2763 	otx2_cgx_config_linkevents(pf, false);
2764 
2765 	unregister_netdev(netdev);
2766 	otx2_sriov_disable(pf->pdev);
2767 	otx2_sriov_vfcfg_cleanup(pf);
2768 	if (pf->otx2_wq)
2769 		destroy_workqueue(pf->otx2_wq);
2770 
2771 	otx2_ptp_destroy(pf);
2772 	otx2_mcam_flow_del(pf);
2773 	otx2_shutdown_tc(pf);
2774 	otx2_detach_resources(&pf->mbox);
2775 	if (pf->hw.lmt_base)
2776 		iounmap(pf->hw.lmt_base);
2777 
2778 	otx2_disable_mbox_intr(pf);
2779 	otx2_pfaf_mbox_destroy(pf);
2780 	pci_free_irq_vectors(pf->pdev);
2781 	pci_set_drvdata(pdev, NULL);
2782 	free_netdev(netdev);
2783 
2784 	pci_release_regions(pdev);
2785 }
2786 
2787 static struct pci_driver otx2_pf_driver = {
2788 	.name = DRV_NAME,
2789 	.id_table = otx2_pf_id_table,
2790 	.probe = otx2_probe,
2791 	.shutdown = otx2_remove,
2792 	.remove = otx2_remove,
2793 	.sriov_configure = otx2_sriov_configure
2794 };
2795 
2796 static int __init otx2_rvupf_init_module(void)
2797 {
2798 	pr_info("%s: %s\n", DRV_NAME, DRV_STRING);
2799 
2800 	return pci_register_driver(&otx2_pf_driver);
2801 }
2802 
2803 static void __exit otx2_rvupf_cleanup_module(void)
2804 {
2805 	pci_unregister_driver(&otx2_pf_driver);
2806 }
2807 
2808 module_init(otx2_rvupf_init_module);
2809 module_exit(otx2_rvupf_cleanup_module);
2810