1 // SPDX-License-Identifier: GPL-2.0
2 /* Marvell RVU Physical Function ethernet driver
3  *
4  * Copyright (C) 2020 Marvell.
5  *
6  */
7 
8 #include <linux/module.h>
9 #include <linux/interrupt.h>
10 #include <linux/pci.h>
11 #include <linux/etherdevice.h>
12 #include <linux/of.h>
13 #include <linux/if_vlan.h>
14 #include <linux/iommu.h>
15 #include <net/ip.h>
16 #include <linux/bpf.h>
17 #include <linux/bpf_trace.h>
18 
19 #include "otx2_reg.h"
20 #include "otx2_common.h"
21 #include "otx2_txrx.h"
22 #include "otx2_struct.h"
23 #include "otx2_ptp.h"
24 #include "cn10k.h"
25 #include <rvu_trace.h>
26 
27 #define DRV_NAME	"rvu_nicpf"
28 #define DRV_STRING	"Marvell RVU NIC Physical Function Driver"
29 
30 /* Supported devices */
31 static const struct pci_device_id otx2_pf_id_table[] = {
32 	{ PCI_DEVICE(PCI_VENDOR_ID_CAVIUM, PCI_DEVID_OCTEONTX2_RVU_PF) },
33 	{ 0, }  /* end of table */
34 };
35 
36 MODULE_AUTHOR("Sunil Goutham <sgoutham@marvell.com>");
37 MODULE_DESCRIPTION(DRV_STRING);
38 MODULE_LICENSE("GPL v2");
39 MODULE_DEVICE_TABLE(pci, otx2_pf_id_table);
40 
41 static void otx2_vf_link_event_task(struct work_struct *work);
42 
43 enum {
44 	TYPE_PFAF,
45 	TYPE_PFVF,
46 };
47 
48 static int otx2_config_hw_tx_tstamp(struct otx2_nic *pfvf, bool enable);
49 static int otx2_config_hw_rx_tstamp(struct otx2_nic *pfvf, bool enable);
50 
51 static int otx2_change_mtu(struct net_device *netdev, int new_mtu)
52 {
53 	struct otx2_nic *pf = netdev_priv(netdev);
54 	bool if_up = netif_running(netdev);
55 	int err = 0;
56 
57 	if (pf->xdp_prog && new_mtu > MAX_XDP_MTU) {
58 		netdev_warn(netdev, "Jumbo frames not yet supported with XDP, current MTU %d.\n",
59 			    netdev->mtu);
60 		return -EINVAL;
61 	}
62 	if (if_up)
63 		otx2_stop(netdev);
64 
65 	netdev_info(netdev, "Changing MTU from %d to %d\n",
66 		    netdev->mtu, new_mtu);
67 	netdev->mtu = new_mtu;
68 
69 	if (if_up)
70 		err = otx2_open(netdev);
71 
72 	return err;
73 }
74 
75 static void otx2_disable_flr_me_intr(struct otx2_nic *pf)
76 {
77 	int irq, vfs = pf->total_vfs;
78 
79 	/* Disable VFs ME interrupts */
80 	otx2_write64(pf, RVU_PF_VFME_INT_ENA_W1CX(0), INTR_MASK(vfs));
81 	irq = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFME0);
82 	free_irq(irq, pf);
83 
84 	/* Disable VFs FLR interrupts */
85 	otx2_write64(pf, RVU_PF_VFFLR_INT_ENA_W1CX(0), INTR_MASK(vfs));
86 	irq = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFFLR0);
87 	free_irq(irq, pf);
88 
89 	if (vfs <= 64)
90 		return;
91 
92 	otx2_write64(pf, RVU_PF_VFME_INT_ENA_W1CX(1), INTR_MASK(vfs - 64));
93 	irq = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFME1);
94 	free_irq(irq, pf);
95 
96 	otx2_write64(pf, RVU_PF_VFFLR_INT_ENA_W1CX(1), INTR_MASK(vfs - 64));
97 	irq = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFFLR1);
98 	free_irq(irq, pf);
99 }
100 
101 static void otx2_flr_wq_destroy(struct otx2_nic *pf)
102 {
103 	if (!pf->flr_wq)
104 		return;
105 	destroy_workqueue(pf->flr_wq);
106 	pf->flr_wq = NULL;
107 	devm_kfree(pf->dev, pf->flr_wrk);
108 }
109 
110 static void otx2_flr_handler(struct work_struct *work)
111 {
112 	struct flr_work *flrwork = container_of(work, struct flr_work, work);
113 	struct otx2_nic *pf = flrwork->pf;
114 	struct mbox *mbox = &pf->mbox;
115 	struct msg_req *req;
116 	int vf, reg = 0;
117 
118 	vf = flrwork - pf->flr_wrk;
119 
120 	mutex_lock(&mbox->lock);
121 	req = otx2_mbox_alloc_msg_vf_flr(mbox);
122 	if (!req) {
123 		mutex_unlock(&mbox->lock);
124 		return;
125 	}
126 	req->hdr.pcifunc &= RVU_PFVF_FUNC_MASK;
127 	req->hdr.pcifunc |= (vf + 1) & RVU_PFVF_FUNC_MASK;
128 
129 	if (!otx2_sync_mbox_msg(&pf->mbox)) {
130 		if (vf >= 64) {
131 			reg = 1;
132 			vf = vf - 64;
133 		}
134 		/* clear transcation pending bit */
135 		otx2_write64(pf, RVU_PF_VFTRPENDX(reg), BIT_ULL(vf));
136 		otx2_write64(pf, RVU_PF_VFFLR_INT_ENA_W1SX(reg), BIT_ULL(vf));
137 	}
138 
139 	mutex_unlock(&mbox->lock);
140 }
141 
142 static irqreturn_t otx2_pf_flr_intr_handler(int irq, void *pf_irq)
143 {
144 	struct otx2_nic *pf = (struct otx2_nic *)pf_irq;
145 	int reg, dev, vf, start_vf, num_reg = 1;
146 	u64 intr;
147 
148 	if (pf->total_vfs > 64)
149 		num_reg = 2;
150 
151 	for (reg = 0; reg < num_reg; reg++) {
152 		intr = otx2_read64(pf, RVU_PF_VFFLR_INTX(reg));
153 		if (!intr)
154 			continue;
155 		start_vf = 64 * reg;
156 		for (vf = 0; vf < 64; vf++) {
157 			if (!(intr & BIT_ULL(vf)))
158 				continue;
159 			dev = vf + start_vf;
160 			queue_work(pf->flr_wq, &pf->flr_wrk[dev].work);
161 			/* Clear interrupt */
162 			otx2_write64(pf, RVU_PF_VFFLR_INTX(reg), BIT_ULL(vf));
163 			/* Disable the interrupt */
164 			otx2_write64(pf, RVU_PF_VFFLR_INT_ENA_W1CX(reg),
165 				     BIT_ULL(vf));
166 		}
167 	}
168 	return IRQ_HANDLED;
169 }
170 
171 static irqreturn_t otx2_pf_me_intr_handler(int irq, void *pf_irq)
172 {
173 	struct otx2_nic *pf = (struct otx2_nic *)pf_irq;
174 	int vf, reg, num_reg = 1;
175 	u64 intr;
176 
177 	if (pf->total_vfs > 64)
178 		num_reg = 2;
179 
180 	for (reg = 0; reg < num_reg; reg++) {
181 		intr = otx2_read64(pf, RVU_PF_VFME_INTX(reg));
182 		if (!intr)
183 			continue;
184 		for (vf = 0; vf < 64; vf++) {
185 			if (!(intr & BIT_ULL(vf)))
186 				continue;
187 			/* clear trpend bit */
188 			otx2_write64(pf, RVU_PF_VFTRPENDX(reg), BIT_ULL(vf));
189 			/* clear interrupt */
190 			otx2_write64(pf, RVU_PF_VFME_INTX(reg), BIT_ULL(vf));
191 		}
192 	}
193 	return IRQ_HANDLED;
194 }
195 
196 static int otx2_register_flr_me_intr(struct otx2_nic *pf, int numvfs)
197 {
198 	struct otx2_hw *hw = &pf->hw;
199 	char *irq_name;
200 	int ret;
201 
202 	/* Register ME interrupt handler*/
203 	irq_name = &hw->irq_name[RVU_PF_INT_VEC_VFME0 * NAME_SIZE];
204 	snprintf(irq_name, NAME_SIZE, "RVUPF%d_ME0", rvu_get_pf(pf->pcifunc));
205 	ret = request_irq(pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFME0),
206 			  otx2_pf_me_intr_handler, 0, irq_name, pf);
207 	if (ret) {
208 		dev_err(pf->dev,
209 			"RVUPF: IRQ registration failed for ME0\n");
210 	}
211 
212 	/* Register FLR interrupt handler */
213 	irq_name = &hw->irq_name[RVU_PF_INT_VEC_VFFLR0 * NAME_SIZE];
214 	snprintf(irq_name, NAME_SIZE, "RVUPF%d_FLR0", rvu_get_pf(pf->pcifunc));
215 	ret = request_irq(pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFFLR0),
216 			  otx2_pf_flr_intr_handler, 0, irq_name, pf);
217 	if (ret) {
218 		dev_err(pf->dev,
219 			"RVUPF: IRQ registration failed for FLR0\n");
220 		return ret;
221 	}
222 
223 	if (numvfs > 64) {
224 		irq_name = &hw->irq_name[RVU_PF_INT_VEC_VFME1 * NAME_SIZE];
225 		snprintf(irq_name, NAME_SIZE, "RVUPF%d_ME1",
226 			 rvu_get_pf(pf->pcifunc));
227 		ret = request_irq(pci_irq_vector
228 				  (pf->pdev, RVU_PF_INT_VEC_VFME1),
229 				  otx2_pf_me_intr_handler, 0, irq_name, pf);
230 		if (ret) {
231 			dev_err(pf->dev,
232 				"RVUPF: IRQ registration failed for ME1\n");
233 		}
234 		irq_name = &hw->irq_name[RVU_PF_INT_VEC_VFFLR1 * NAME_SIZE];
235 		snprintf(irq_name, NAME_SIZE, "RVUPF%d_FLR1",
236 			 rvu_get_pf(pf->pcifunc));
237 		ret = request_irq(pci_irq_vector
238 				  (pf->pdev, RVU_PF_INT_VEC_VFFLR1),
239 				  otx2_pf_flr_intr_handler, 0, irq_name, pf);
240 		if (ret) {
241 			dev_err(pf->dev,
242 				"RVUPF: IRQ registration failed for FLR1\n");
243 			return ret;
244 		}
245 	}
246 
247 	/* Enable ME interrupt for all VFs*/
248 	otx2_write64(pf, RVU_PF_VFME_INTX(0), INTR_MASK(numvfs));
249 	otx2_write64(pf, RVU_PF_VFME_INT_ENA_W1SX(0), INTR_MASK(numvfs));
250 
251 	/* Enable FLR interrupt for all VFs*/
252 	otx2_write64(pf, RVU_PF_VFFLR_INTX(0), INTR_MASK(numvfs));
253 	otx2_write64(pf, RVU_PF_VFFLR_INT_ENA_W1SX(0), INTR_MASK(numvfs));
254 
255 	if (numvfs > 64) {
256 		numvfs -= 64;
257 
258 		otx2_write64(pf, RVU_PF_VFME_INTX(1), INTR_MASK(numvfs));
259 		otx2_write64(pf, RVU_PF_VFME_INT_ENA_W1SX(1),
260 			     INTR_MASK(numvfs));
261 
262 		otx2_write64(pf, RVU_PF_VFFLR_INTX(1), INTR_MASK(numvfs));
263 		otx2_write64(pf, RVU_PF_VFFLR_INT_ENA_W1SX(1),
264 			     INTR_MASK(numvfs));
265 	}
266 	return 0;
267 }
268 
269 static int otx2_pf_flr_init(struct otx2_nic *pf, int num_vfs)
270 {
271 	int vf;
272 
273 	pf->flr_wq = alloc_workqueue("otx2_pf_flr_wq",
274 				     WQ_UNBOUND | WQ_HIGHPRI, 1);
275 	if (!pf->flr_wq)
276 		return -ENOMEM;
277 
278 	pf->flr_wrk = devm_kcalloc(pf->dev, num_vfs,
279 				   sizeof(struct flr_work), GFP_KERNEL);
280 	if (!pf->flr_wrk) {
281 		destroy_workqueue(pf->flr_wq);
282 		return -ENOMEM;
283 	}
284 
285 	for (vf = 0; vf < num_vfs; vf++) {
286 		pf->flr_wrk[vf].pf = pf;
287 		INIT_WORK(&pf->flr_wrk[vf].work, otx2_flr_handler);
288 	}
289 
290 	return 0;
291 }
292 
293 static void otx2_queue_work(struct mbox *mw, struct workqueue_struct *mbox_wq,
294 			    int first, int mdevs, u64 intr, int type)
295 {
296 	struct otx2_mbox_dev *mdev;
297 	struct otx2_mbox *mbox;
298 	struct mbox_hdr *hdr;
299 	int i;
300 
301 	for (i = first; i < mdevs; i++) {
302 		/* start from 0 */
303 		if (!(intr & BIT_ULL(i - first)))
304 			continue;
305 
306 		mbox = &mw->mbox;
307 		mdev = &mbox->dev[i];
308 		if (type == TYPE_PFAF)
309 			otx2_sync_mbox_bbuf(mbox, i);
310 		hdr = mdev->mbase + mbox->rx_start;
311 		/* The hdr->num_msgs is set to zero immediately in the interrupt
312 		 * handler to  ensure that it holds a correct value next time
313 		 * when the interrupt handler is called.
314 		 * pf->mbox.num_msgs holds the data for use in pfaf_mbox_handler
315 		 * pf>mbox.up_num_msgs holds the data for use in
316 		 * pfaf_mbox_up_handler.
317 		 */
318 		if (hdr->num_msgs) {
319 			mw[i].num_msgs = hdr->num_msgs;
320 			hdr->num_msgs = 0;
321 			if (type == TYPE_PFAF)
322 				memset(mbox->hwbase + mbox->rx_start, 0,
323 				       ALIGN(sizeof(struct mbox_hdr),
324 					     sizeof(u64)));
325 
326 			queue_work(mbox_wq, &mw[i].mbox_wrk);
327 		}
328 
329 		mbox = &mw->mbox_up;
330 		mdev = &mbox->dev[i];
331 		if (type == TYPE_PFAF)
332 			otx2_sync_mbox_bbuf(mbox, i);
333 		hdr = mdev->mbase + mbox->rx_start;
334 		if (hdr->num_msgs) {
335 			mw[i].up_num_msgs = hdr->num_msgs;
336 			hdr->num_msgs = 0;
337 			if (type == TYPE_PFAF)
338 				memset(mbox->hwbase + mbox->rx_start, 0,
339 				       ALIGN(sizeof(struct mbox_hdr),
340 					     sizeof(u64)));
341 
342 			queue_work(mbox_wq, &mw[i].mbox_up_wrk);
343 		}
344 	}
345 }
346 
347 static void otx2_forward_msg_pfvf(struct otx2_mbox_dev *mdev,
348 				  struct otx2_mbox *pfvf_mbox, void *bbuf_base,
349 				  int devid)
350 {
351 	struct otx2_mbox_dev *src_mdev = mdev;
352 	int offset;
353 
354 	/* Msgs are already copied, trigger VF's mbox irq */
355 	smp_wmb();
356 
357 	offset = pfvf_mbox->trigger | (devid << pfvf_mbox->tr_shift);
358 	writeq(1, (void __iomem *)pfvf_mbox->reg_base + offset);
359 
360 	/* Restore VF's mbox bounce buffer region address */
361 	src_mdev->mbase = bbuf_base;
362 }
363 
364 static int otx2_forward_vf_mbox_msgs(struct otx2_nic *pf,
365 				     struct otx2_mbox *src_mbox,
366 				     int dir, int vf, int num_msgs)
367 {
368 	struct otx2_mbox_dev *src_mdev, *dst_mdev;
369 	struct mbox_hdr *mbox_hdr;
370 	struct mbox_hdr *req_hdr;
371 	struct mbox *dst_mbox;
372 	int dst_size, err;
373 
374 	if (dir == MBOX_DIR_PFAF) {
375 		/* Set VF's mailbox memory as PF's bounce buffer memory, so
376 		 * that explicit copying of VF's msgs to PF=>AF mbox region
377 		 * and AF=>PF responses to VF's mbox region can be avoided.
378 		 */
379 		src_mdev = &src_mbox->dev[vf];
380 		mbox_hdr = src_mbox->hwbase +
381 				src_mbox->rx_start + (vf * MBOX_SIZE);
382 
383 		dst_mbox = &pf->mbox;
384 		dst_size = dst_mbox->mbox.tx_size -
385 				ALIGN(sizeof(*mbox_hdr), MBOX_MSG_ALIGN);
386 		/* Check if msgs fit into destination area and has valid size */
387 		if (mbox_hdr->msg_size > dst_size || !mbox_hdr->msg_size)
388 			return -EINVAL;
389 
390 		dst_mdev = &dst_mbox->mbox.dev[0];
391 
392 		mutex_lock(&pf->mbox.lock);
393 		dst_mdev->mbase = src_mdev->mbase;
394 		dst_mdev->msg_size = mbox_hdr->msg_size;
395 		dst_mdev->num_msgs = num_msgs;
396 		err = otx2_sync_mbox_msg(dst_mbox);
397 		/* Error code -EIO indicate there is a communication failure
398 		 * to the AF. Rest of the error codes indicate that AF processed
399 		 * VF messages and set the error codes in response messages
400 		 * (if any) so simply forward responses to VF.
401 		 */
402 		if (err == -EIO) {
403 			dev_warn(pf->dev,
404 				 "AF not responding to VF%d messages\n", vf);
405 			/* restore PF mbase and exit */
406 			dst_mdev->mbase = pf->mbox.bbuf_base;
407 			mutex_unlock(&pf->mbox.lock);
408 			return err;
409 		}
410 		/* At this point, all the VF messages sent to AF are acked
411 		 * with proper responses and responses are copied to VF
412 		 * mailbox hence raise interrupt to VF.
413 		 */
414 		req_hdr = (struct mbox_hdr *)(dst_mdev->mbase +
415 					      dst_mbox->mbox.rx_start);
416 		req_hdr->num_msgs = num_msgs;
417 
418 		otx2_forward_msg_pfvf(dst_mdev, &pf->mbox_pfvf[0].mbox,
419 				      pf->mbox.bbuf_base, vf);
420 		mutex_unlock(&pf->mbox.lock);
421 	} else if (dir == MBOX_DIR_PFVF_UP) {
422 		src_mdev = &src_mbox->dev[0];
423 		mbox_hdr = src_mbox->hwbase + src_mbox->rx_start;
424 		req_hdr = (struct mbox_hdr *)(src_mdev->mbase +
425 					      src_mbox->rx_start);
426 		req_hdr->num_msgs = num_msgs;
427 
428 		dst_mbox = &pf->mbox_pfvf[0];
429 		dst_size = dst_mbox->mbox_up.tx_size -
430 				ALIGN(sizeof(*mbox_hdr), MBOX_MSG_ALIGN);
431 		/* Check if msgs fit into destination area */
432 		if (mbox_hdr->msg_size > dst_size)
433 			return -EINVAL;
434 
435 		dst_mdev = &dst_mbox->mbox_up.dev[vf];
436 		dst_mdev->mbase = src_mdev->mbase;
437 		dst_mdev->msg_size = mbox_hdr->msg_size;
438 		dst_mdev->num_msgs = mbox_hdr->num_msgs;
439 		err = otx2_sync_mbox_up_msg(dst_mbox, vf);
440 		if (err) {
441 			dev_warn(pf->dev,
442 				 "VF%d is not responding to mailbox\n", vf);
443 			return err;
444 		}
445 	} else if (dir == MBOX_DIR_VFPF_UP) {
446 		req_hdr = (struct mbox_hdr *)(src_mbox->dev[0].mbase +
447 					      src_mbox->rx_start);
448 		req_hdr->num_msgs = num_msgs;
449 		otx2_forward_msg_pfvf(&pf->mbox_pfvf->mbox_up.dev[vf],
450 				      &pf->mbox.mbox_up,
451 				      pf->mbox_pfvf[vf].bbuf_base,
452 				      0);
453 	}
454 
455 	return 0;
456 }
457 
458 static void otx2_pfvf_mbox_handler(struct work_struct *work)
459 {
460 	struct mbox_msghdr *msg = NULL;
461 	int offset, vf_idx, id, err;
462 	struct otx2_mbox_dev *mdev;
463 	struct mbox_hdr *req_hdr;
464 	struct otx2_mbox *mbox;
465 	struct mbox *vf_mbox;
466 	struct otx2_nic *pf;
467 
468 	vf_mbox = container_of(work, struct mbox, mbox_wrk);
469 	pf = vf_mbox->pfvf;
470 	vf_idx = vf_mbox - pf->mbox_pfvf;
471 
472 	mbox = &pf->mbox_pfvf[0].mbox;
473 	mdev = &mbox->dev[vf_idx];
474 	req_hdr = (struct mbox_hdr *)(mdev->mbase + mbox->rx_start);
475 
476 	offset = ALIGN(sizeof(*req_hdr), MBOX_MSG_ALIGN);
477 
478 	for (id = 0; id < vf_mbox->num_msgs; id++) {
479 		msg = (struct mbox_msghdr *)(mdev->mbase + mbox->rx_start +
480 					     offset);
481 
482 		if (msg->sig != OTX2_MBOX_REQ_SIG)
483 			goto inval_msg;
484 
485 		/* Set VF's number in each of the msg */
486 		msg->pcifunc &= RVU_PFVF_FUNC_MASK;
487 		msg->pcifunc |= (vf_idx + 1) & RVU_PFVF_FUNC_MASK;
488 		offset = msg->next_msgoff;
489 	}
490 	err = otx2_forward_vf_mbox_msgs(pf, mbox, MBOX_DIR_PFAF, vf_idx,
491 					vf_mbox->num_msgs);
492 	if (err)
493 		goto inval_msg;
494 	return;
495 
496 inval_msg:
497 	otx2_reply_invalid_msg(mbox, vf_idx, 0, msg->id);
498 	otx2_mbox_msg_send(mbox, vf_idx);
499 }
500 
501 static void otx2_pfvf_mbox_up_handler(struct work_struct *work)
502 {
503 	struct mbox *vf_mbox = container_of(work, struct mbox, mbox_up_wrk);
504 	struct otx2_nic *pf = vf_mbox->pfvf;
505 	struct otx2_mbox_dev *mdev;
506 	int offset, id, vf_idx = 0;
507 	struct mbox_hdr *rsp_hdr;
508 	struct mbox_msghdr *msg;
509 	struct otx2_mbox *mbox;
510 
511 	vf_idx = vf_mbox - pf->mbox_pfvf;
512 	mbox = &pf->mbox_pfvf[0].mbox_up;
513 	mdev = &mbox->dev[vf_idx];
514 
515 	rsp_hdr = (struct mbox_hdr *)(mdev->mbase + mbox->rx_start);
516 	offset = mbox->rx_start + ALIGN(sizeof(*rsp_hdr), MBOX_MSG_ALIGN);
517 
518 	for (id = 0; id < vf_mbox->up_num_msgs; id++) {
519 		msg = mdev->mbase + offset;
520 
521 		if (msg->id >= MBOX_MSG_MAX) {
522 			dev_err(pf->dev,
523 				"Mbox msg with unknown ID 0x%x\n", msg->id);
524 			goto end;
525 		}
526 
527 		if (msg->sig != OTX2_MBOX_RSP_SIG) {
528 			dev_err(pf->dev,
529 				"Mbox msg with wrong signature %x, ID 0x%x\n",
530 				msg->sig, msg->id);
531 			goto end;
532 		}
533 
534 		switch (msg->id) {
535 		case MBOX_MSG_CGX_LINK_EVENT:
536 			break;
537 		default:
538 			if (msg->rc)
539 				dev_err(pf->dev,
540 					"Mbox msg response has err %d, ID 0x%x\n",
541 					msg->rc, msg->id);
542 			break;
543 		}
544 
545 end:
546 		offset = mbox->rx_start + msg->next_msgoff;
547 		if (mdev->msgs_acked == (vf_mbox->up_num_msgs - 1))
548 			__otx2_mbox_reset(mbox, 0);
549 		mdev->msgs_acked++;
550 	}
551 }
552 
553 static irqreturn_t otx2_pfvf_mbox_intr_handler(int irq, void *pf_irq)
554 {
555 	struct otx2_nic *pf = (struct otx2_nic *)(pf_irq);
556 	int vfs = pf->total_vfs;
557 	struct mbox *mbox;
558 	u64 intr;
559 
560 	mbox = pf->mbox_pfvf;
561 	/* Handle VF interrupts */
562 	if (vfs > 64) {
563 		intr = otx2_read64(pf, RVU_PF_VFPF_MBOX_INTX(1));
564 		otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(1), intr);
565 		otx2_queue_work(mbox, pf->mbox_pfvf_wq, 64, vfs, intr,
566 				TYPE_PFVF);
567 		vfs -= 64;
568 	}
569 
570 	intr = otx2_read64(pf, RVU_PF_VFPF_MBOX_INTX(0));
571 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(0), intr);
572 
573 	otx2_queue_work(mbox, pf->mbox_pfvf_wq, 0, vfs, intr, TYPE_PFVF);
574 
575 	trace_otx2_msg_interrupt(mbox->mbox.pdev, "VF(s) to PF", intr);
576 
577 	return IRQ_HANDLED;
578 }
579 
580 static int otx2_pfvf_mbox_init(struct otx2_nic *pf, int numvfs)
581 {
582 	void __iomem *hwbase;
583 	struct mbox *mbox;
584 	int err, vf;
585 	u64 base;
586 
587 	if (!numvfs)
588 		return -EINVAL;
589 
590 	pf->mbox_pfvf = devm_kcalloc(&pf->pdev->dev, numvfs,
591 				     sizeof(struct mbox), GFP_KERNEL);
592 	if (!pf->mbox_pfvf)
593 		return -ENOMEM;
594 
595 	pf->mbox_pfvf_wq = alloc_workqueue("otx2_pfvf_mailbox",
596 					   WQ_UNBOUND | WQ_HIGHPRI |
597 					   WQ_MEM_RECLAIM, 1);
598 	if (!pf->mbox_pfvf_wq)
599 		return -ENOMEM;
600 
601 	/* On CN10K platform, PF <-> VF mailbox region follows after
602 	 * PF <-> AF mailbox region.
603 	 */
604 	if (test_bit(CN10K_MBOX, &pf->hw.cap_flag))
605 		base = pci_resource_start(pf->pdev, PCI_MBOX_BAR_NUM) +
606 		       MBOX_SIZE;
607 	else
608 		base = readq((void __iomem *)((u64)pf->reg_base +
609 					      RVU_PF_VF_BAR4_ADDR));
610 
611 	hwbase = ioremap_wc(base, MBOX_SIZE * pf->total_vfs);
612 	if (!hwbase) {
613 		err = -ENOMEM;
614 		goto free_wq;
615 	}
616 
617 	mbox = &pf->mbox_pfvf[0];
618 	err = otx2_mbox_init(&mbox->mbox, hwbase, pf->pdev, pf->reg_base,
619 			     MBOX_DIR_PFVF, numvfs);
620 	if (err)
621 		goto free_iomem;
622 
623 	err = otx2_mbox_init(&mbox->mbox_up, hwbase, pf->pdev, pf->reg_base,
624 			     MBOX_DIR_PFVF_UP, numvfs);
625 	if (err)
626 		goto free_iomem;
627 
628 	for (vf = 0; vf < numvfs; vf++) {
629 		mbox->pfvf = pf;
630 		INIT_WORK(&mbox->mbox_wrk, otx2_pfvf_mbox_handler);
631 		INIT_WORK(&mbox->mbox_up_wrk, otx2_pfvf_mbox_up_handler);
632 		mbox++;
633 	}
634 
635 	return 0;
636 
637 free_iomem:
638 	if (hwbase)
639 		iounmap(hwbase);
640 free_wq:
641 	destroy_workqueue(pf->mbox_pfvf_wq);
642 	return err;
643 }
644 
645 static void otx2_pfvf_mbox_destroy(struct otx2_nic *pf)
646 {
647 	struct mbox *mbox = &pf->mbox_pfvf[0];
648 
649 	if (!mbox)
650 		return;
651 
652 	if (pf->mbox_pfvf_wq) {
653 		destroy_workqueue(pf->mbox_pfvf_wq);
654 		pf->mbox_pfvf_wq = NULL;
655 	}
656 
657 	if (mbox->mbox.hwbase)
658 		iounmap(mbox->mbox.hwbase);
659 
660 	otx2_mbox_destroy(&mbox->mbox);
661 }
662 
663 static void otx2_enable_pfvf_mbox_intr(struct otx2_nic *pf, int numvfs)
664 {
665 	/* Clear PF <=> VF mailbox IRQ */
666 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(0), ~0ull);
667 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(1), ~0ull);
668 
669 	/* Enable PF <=> VF mailbox IRQ */
670 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INT_ENA_W1SX(0), INTR_MASK(numvfs));
671 	if (numvfs > 64) {
672 		numvfs -= 64;
673 		otx2_write64(pf, RVU_PF_VFPF_MBOX_INT_ENA_W1SX(1),
674 			     INTR_MASK(numvfs));
675 	}
676 }
677 
678 static void otx2_disable_pfvf_mbox_intr(struct otx2_nic *pf, int numvfs)
679 {
680 	int vector;
681 
682 	/* Disable PF <=> VF mailbox IRQ */
683 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INT_ENA_W1CX(0), ~0ull);
684 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INT_ENA_W1CX(1), ~0ull);
685 
686 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(0), ~0ull);
687 	vector = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFPF_MBOX0);
688 	free_irq(vector, pf);
689 
690 	if (numvfs > 64) {
691 		otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(1), ~0ull);
692 		vector = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFPF_MBOX1);
693 		free_irq(vector, pf);
694 	}
695 }
696 
697 static int otx2_register_pfvf_mbox_intr(struct otx2_nic *pf, int numvfs)
698 {
699 	struct otx2_hw *hw = &pf->hw;
700 	char *irq_name;
701 	int err;
702 
703 	/* Register MBOX0 interrupt handler */
704 	irq_name = &hw->irq_name[RVU_PF_INT_VEC_VFPF_MBOX0 * NAME_SIZE];
705 	if (pf->pcifunc)
706 		snprintf(irq_name, NAME_SIZE,
707 			 "RVUPF%d_VF Mbox0", rvu_get_pf(pf->pcifunc));
708 	else
709 		snprintf(irq_name, NAME_SIZE, "RVUPF_VF Mbox0");
710 	err = request_irq(pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFPF_MBOX0),
711 			  otx2_pfvf_mbox_intr_handler, 0, irq_name, pf);
712 	if (err) {
713 		dev_err(pf->dev,
714 			"RVUPF: IRQ registration failed for PFVF mbox0 irq\n");
715 		return err;
716 	}
717 
718 	if (numvfs > 64) {
719 		/* Register MBOX1 interrupt handler */
720 		irq_name = &hw->irq_name[RVU_PF_INT_VEC_VFPF_MBOX1 * NAME_SIZE];
721 		if (pf->pcifunc)
722 			snprintf(irq_name, NAME_SIZE,
723 				 "RVUPF%d_VF Mbox1", rvu_get_pf(pf->pcifunc));
724 		else
725 			snprintf(irq_name, NAME_SIZE, "RVUPF_VF Mbox1");
726 		err = request_irq(pci_irq_vector(pf->pdev,
727 						 RVU_PF_INT_VEC_VFPF_MBOX1),
728 						 otx2_pfvf_mbox_intr_handler,
729 						 0, irq_name, pf);
730 		if (err) {
731 			dev_err(pf->dev,
732 				"RVUPF: IRQ registration failed for PFVF mbox1 irq\n");
733 			return err;
734 		}
735 	}
736 
737 	otx2_enable_pfvf_mbox_intr(pf, numvfs);
738 
739 	return 0;
740 }
741 
742 static void otx2_process_pfaf_mbox_msg(struct otx2_nic *pf,
743 				       struct mbox_msghdr *msg)
744 {
745 	int devid;
746 
747 	if (msg->id >= MBOX_MSG_MAX) {
748 		dev_err(pf->dev,
749 			"Mbox msg with unknown ID 0x%x\n", msg->id);
750 		return;
751 	}
752 
753 	if (msg->sig != OTX2_MBOX_RSP_SIG) {
754 		dev_err(pf->dev,
755 			"Mbox msg with wrong signature %x, ID 0x%x\n",
756 			 msg->sig, msg->id);
757 		return;
758 	}
759 
760 	/* message response heading VF */
761 	devid = msg->pcifunc & RVU_PFVF_FUNC_MASK;
762 	if (devid) {
763 		struct otx2_vf_config *config = &pf->vf_configs[devid - 1];
764 		struct delayed_work *dwork;
765 
766 		switch (msg->id) {
767 		case MBOX_MSG_NIX_LF_START_RX:
768 			config->intf_down = false;
769 			dwork = &config->link_event_work;
770 			schedule_delayed_work(dwork, msecs_to_jiffies(100));
771 			break;
772 		case MBOX_MSG_NIX_LF_STOP_RX:
773 			config->intf_down = true;
774 			break;
775 		}
776 
777 		return;
778 	}
779 
780 	switch (msg->id) {
781 	case MBOX_MSG_READY:
782 		pf->pcifunc = msg->pcifunc;
783 		break;
784 	case MBOX_MSG_MSIX_OFFSET:
785 		mbox_handler_msix_offset(pf, (struct msix_offset_rsp *)msg);
786 		break;
787 	case MBOX_MSG_NPA_LF_ALLOC:
788 		mbox_handler_npa_lf_alloc(pf, (struct npa_lf_alloc_rsp *)msg);
789 		break;
790 	case MBOX_MSG_NIX_LF_ALLOC:
791 		mbox_handler_nix_lf_alloc(pf, (struct nix_lf_alloc_rsp *)msg);
792 		break;
793 	case MBOX_MSG_NIX_TXSCH_ALLOC:
794 		mbox_handler_nix_txsch_alloc(pf,
795 					     (struct nix_txsch_alloc_rsp *)msg);
796 		break;
797 	case MBOX_MSG_NIX_BP_ENABLE:
798 		mbox_handler_nix_bp_enable(pf, (struct nix_bp_cfg_rsp *)msg);
799 		break;
800 	case MBOX_MSG_CGX_STATS:
801 		mbox_handler_cgx_stats(pf, (struct cgx_stats_rsp *)msg);
802 		break;
803 	case MBOX_MSG_CGX_FEC_STATS:
804 		mbox_handler_cgx_fec_stats(pf, (struct cgx_fec_stats_rsp *)msg);
805 		break;
806 	default:
807 		if (msg->rc)
808 			dev_err(pf->dev,
809 				"Mbox msg response has err %d, ID 0x%x\n",
810 				msg->rc, msg->id);
811 		break;
812 	}
813 }
814 
815 static void otx2_pfaf_mbox_handler(struct work_struct *work)
816 {
817 	struct otx2_mbox_dev *mdev;
818 	struct mbox_hdr *rsp_hdr;
819 	struct mbox_msghdr *msg;
820 	struct otx2_mbox *mbox;
821 	struct mbox *af_mbox;
822 	struct otx2_nic *pf;
823 	int offset, id;
824 
825 	af_mbox = container_of(work, struct mbox, mbox_wrk);
826 	mbox = &af_mbox->mbox;
827 	mdev = &mbox->dev[0];
828 	rsp_hdr = (struct mbox_hdr *)(mdev->mbase + mbox->rx_start);
829 
830 	offset = mbox->rx_start + ALIGN(sizeof(*rsp_hdr), MBOX_MSG_ALIGN);
831 	pf = af_mbox->pfvf;
832 
833 	for (id = 0; id < af_mbox->num_msgs; id++) {
834 		msg = (struct mbox_msghdr *)(mdev->mbase + offset);
835 		otx2_process_pfaf_mbox_msg(pf, msg);
836 		offset = mbox->rx_start + msg->next_msgoff;
837 		if (mdev->msgs_acked == (af_mbox->num_msgs - 1))
838 			__otx2_mbox_reset(mbox, 0);
839 		mdev->msgs_acked++;
840 	}
841 
842 }
843 
844 static void otx2_handle_link_event(struct otx2_nic *pf)
845 {
846 	struct cgx_link_user_info *linfo = &pf->linfo;
847 	struct net_device *netdev = pf->netdev;
848 
849 	pr_info("%s NIC Link is %s %d Mbps %s duplex\n", netdev->name,
850 		linfo->link_up ? "UP" : "DOWN", linfo->speed,
851 		linfo->full_duplex ? "Full" : "Half");
852 	if (linfo->link_up) {
853 		netif_carrier_on(netdev);
854 		netif_tx_start_all_queues(netdev);
855 	} else {
856 		netif_tx_stop_all_queues(netdev);
857 		netif_carrier_off(netdev);
858 	}
859 }
860 
861 int otx2_mbox_up_handler_cgx_link_event(struct otx2_nic *pf,
862 					struct cgx_link_info_msg *msg,
863 					struct msg_rsp *rsp)
864 {
865 	int i;
866 
867 	/* Copy the link info sent by AF */
868 	pf->linfo = msg->link_info;
869 
870 	/* notify VFs about link event */
871 	for (i = 0; i < pci_num_vf(pf->pdev); i++) {
872 		struct otx2_vf_config *config = &pf->vf_configs[i];
873 		struct delayed_work *dwork = &config->link_event_work;
874 
875 		if (config->intf_down)
876 			continue;
877 
878 		schedule_delayed_work(dwork, msecs_to_jiffies(100));
879 	}
880 
881 	/* interface has not been fully configured yet */
882 	if (pf->flags & OTX2_FLAG_INTF_DOWN)
883 		return 0;
884 
885 	otx2_handle_link_event(pf);
886 	return 0;
887 }
888 
889 static int otx2_process_mbox_msg_up(struct otx2_nic *pf,
890 				    struct mbox_msghdr *req)
891 {
892 	/* Check if valid, if not reply with a invalid msg */
893 	if (req->sig != OTX2_MBOX_REQ_SIG) {
894 		otx2_reply_invalid_msg(&pf->mbox.mbox_up, 0, 0, req->id);
895 		return -ENODEV;
896 	}
897 
898 	switch (req->id) {
899 #define M(_name, _id, _fn_name, _req_type, _rsp_type)			\
900 	case _id: {							\
901 		struct _rsp_type *rsp;					\
902 		int err;						\
903 									\
904 		rsp = (struct _rsp_type *)otx2_mbox_alloc_msg(		\
905 			&pf->mbox.mbox_up, 0,				\
906 			sizeof(struct _rsp_type));			\
907 		if (!rsp)						\
908 			return -ENOMEM;					\
909 									\
910 		rsp->hdr.id = _id;					\
911 		rsp->hdr.sig = OTX2_MBOX_RSP_SIG;			\
912 		rsp->hdr.pcifunc = 0;					\
913 		rsp->hdr.rc = 0;					\
914 									\
915 		err = otx2_mbox_up_handler_ ## _fn_name(		\
916 			pf, (struct _req_type *)req, rsp);		\
917 		return err;						\
918 	}
919 MBOX_UP_CGX_MESSAGES
920 #undef M
921 		break;
922 	default:
923 		otx2_reply_invalid_msg(&pf->mbox.mbox_up, 0, 0, req->id);
924 		return -ENODEV;
925 	}
926 	return 0;
927 }
928 
929 static void otx2_pfaf_mbox_up_handler(struct work_struct *work)
930 {
931 	struct mbox *af_mbox = container_of(work, struct mbox, mbox_up_wrk);
932 	struct otx2_mbox *mbox = &af_mbox->mbox_up;
933 	struct otx2_mbox_dev *mdev = &mbox->dev[0];
934 	struct otx2_nic *pf = af_mbox->pfvf;
935 	int offset, id, devid = 0;
936 	struct mbox_hdr *rsp_hdr;
937 	struct mbox_msghdr *msg;
938 
939 	rsp_hdr = (struct mbox_hdr *)(mdev->mbase + mbox->rx_start);
940 
941 	offset = mbox->rx_start + ALIGN(sizeof(*rsp_hdr), MBOX_MSG_ALIGN);
942 
943 	for (id = 0; id < af_mbox->up_num_msgs; id++) {
944 		msg = (struct mbox_msghdr *)(mdev->mbase + offset);
945 
946 		devid = msg->pcifunc & RVU_PFVF_FUNC_MASK;
947 		/* Skip processing VF's messages */
948 		if (!devid)
949 			otx2_process_mbox_msg_up(pf, msg);
950 		offset = mbox->rx_start + msg->next_msgoff;
951 	}
952 	if (devid) {
953 		otx2_forward_vf_mbox_msgs(pf, &pf->mbox.mbox_up,
954 					  MBOX_DIR_PFVF_UP, devid - 1,
955 					  af_mbox->up_num_msgs);
956 		return;
957 	}
958 
959 	otx2_mbox_msg_send(mbox, 0);
960 }
961 
962 static irqreturn_t otx2_pfaf_mbox_intr_handler(int irq, void *pf_irq)
963 {
964 	struct otx2_nic *pf = (struct otx2_nic *)pf_irq;
965 	struct mbox *mbox;
966 
967 	/* Clear the IRQ */
968 	otx2_write64(pf, RVU_PF_INT, BIT_ULL(0));
969 
970 	mbox = &pf->mbox;
971 
972 	trace_otx2_msg_interrupt(mbox->mbox.pdev, "AF to PF", BIT_ULL(0));
973 
974 	otx2_queue_work(mbox, pf->mbox_wq, 0, 1, 1, TYPE_PFAF);
975 
976 	return IRQ_HANDLED;
977 }
978 
979 static void otx2_disable_mbox_intr(struct otx2_nic *pf)
980 {
981 	int vector = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_AFPF_MBOX);
982 
983 	/* Disable AF => PF mailbox IRQ */
984 	otx2_write64(pf, RVU_PF_INT_ENA_W1C, BIT_ULL(0));
985 	free_irq(vector, pf);
986 }
987 
988 static int otx2_register_mbox_intr(struct otx2_nic *pf, bool probe_af)
989 {
990 	struct otx2_hw *hw = &pf->hw;
991 	struct msg_req *req;
992 	char *irq_name;
993 	int err;
994 
995 	/* Register mailbox interrupt handler */
996 	irq_name = &hw->irq_name[RVU_PF_INT_VEC_AFPF_MBOX * NAME_SIZE];
997 	snprintf(irq_name, NAME_SIZE, "RVUPFAF Mbox");
998 	err = request_irq(pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_AFPF_MBOX),
999 			  otx2_pfaf_mbox_intr_handler, 0, irq_name, pf);
1000 	if (err) {
1001 		dev_err(pf->dev,
1002 			"RVUPF: IRQ registration failed for PFAF mbox irq\n");
1003 		return err;
1004 	}
1005 
1006 	/* Enable mailbox interrupt for msgs coming from AF.
1007 	 * First clear to avoid spurious interrupts, if any.
1008 	 */
1009 	otx2_write64(pf, RVU_PF_INT, BIT_ULL(0));
1010 	otx2_write64(pf, RVU_PF_INT_ENA_W1S, BIT_ULL(0));
1011 
1012 	if (!probe_af)
1013 		return 0;
1014 
1015 	/* Check mailbox communication with AF */
1016 	req = otx2_mbox_alloc_msg_ready(&pf->mbox);
1017 	if (!req) {
1018 		otx2_disable_mbox_intr(pf);
1019 		return -ENOMEM;
1020 	}
1021 	err = otx2_sync_mbox_msg(&pf->mbox);
1022 	if (err) {
1023 		dev_warn(pf->dev,
1024 			 "AF not responding to mailbox, deferring probe\n");
1025 		otx2_disable_mbox_intr(pf);
1026 		return -EPROBE_DEFER;
1027 	}
1028 
1029 	return 0;
1030 }
1031 
1032 static void otx2_pfaf_mbox_destroy(struct otx2_nic *pf)
1033 {
1034 	struct mbox *mbox = &pf->mbox;
1035 
1036 	if (pf->mbox_wq) {
1037 		destroy_workqueue(pf->mbox_wq);
1038 		pf->mbox_wq = NULL;
1039 	}
1040 
1041 	if (mbox->mbox.hwbase)
1042 		iounmap((void __iomem *)mbox->mbox.hwbase);
1043 
1044 	otx2_mbox_destroy(&mbox->mbox);
1045 	otx2_mbox_destroy(&mbox->mbox_up);
1046 }
1047 
1048 static int otx2_pfaf_mbox_init(struct otx2_nic *pf)
1049 {
1050 	struct mbox *mbox = &pf->mbox;
1051 	void __iomem *hwbase;
1052 	int err;
1053 
1054 	mbox->pfvf = pf;
1055 	pf->mbox_wq = alloc_workqueue("otx2_pfaf_mailbox",
1056 				      WQ_UNBOUND | WQ_HIGHPRI |
1057 				      WQ_MEM_RECLAIM, 1);
1058 	if (!pf->mbox_wq)
1059 		return -ENOMEM;
1060 
1061 	/* Mailbox is a reserved memory (in RAM) region shared between
1062 	 * admin function (i.e AF) and this PF, shouldn't be mapped as
1063 	 * device memory to allow unaligned accesses.
1064 	 */
1065 	hwbase = ioremap_wc(pci_resource_start(pf->pdev, PCI_MBOX_BAR_NUM),
1066 			    MBOX_SIZE);
1067 	if (!hwbase) {
1068 		dev_err(pf->dev, "Unable to map PFAF mailbox region\n");
1069 		err = -ENOMEM;
1070 		goto exit;
1071 	}
1072 
1073 	err = otx2_mbox_init(&mbox->mbox, hwbase, pf->pdev, pf->reg_base,
1074 			     MBOX_DIR_PFAF, 1);
1075 	if (err)
1076 		goto exit;
1077 
1078 	err = otx2_mbox_init(&mbox->mbox_up, hwbase, pf->pdev, pf->reg_base,
1079 			     MBOX_DIR_PFAF_UP, 1);
1080 	if (err)
1081 		goto exit;
1082 
1083 	err = otx2_mbox_bbuf_init(mbox, pf->pdev);
1084 	if (err)
1085 		goto exit;
1086 
1087 	INIT_WORK(&mbox->mbox_wrk, otx2_pfaf_mbox_handler);
1088 	INIT_WORK(&mbox->mbox_up_wrk, otx2_pfaf_mbox_up_handler);
1089 	mutex_init(&mbox->lock);
1090 
1091 	return 0;
1092 exit:
1093 	otx2_pfaf_mbox_destroy(pf);
1094 	return err;
1095 }
1096 
1097 static int otx2_cgx_config_linkevents(struct otx2_nic *pf, bool enable)
1098 {
1099 	struct msg_req *msg;
1100 	int err;
1101 
1102 	mutex_lock(&pf->mbox.lock);
1103 	if (enable)
1104 		msg = otx2_mbox_alloc_msg_cgx_start_linkevents(&pf->mbox);
1105 	else
1106 		msg = otx2_mbox_alloc_msg_cgx_stop_linkevents(&pf->mbox);
1107 
1108 	if (!msg) {
1109 		mutex_unlock(&pf->mbox.lock);
1110 		return -ENOMEM;
1111 	}
1112 
1113 	err = otx2_sync_mbox_msg(&pf->mbox);
1114 	mutex_unlock(&pf->mbox.lock);
1115 	return err;
1116 }
1117 
1118 static int otx2_cgx_config_loopback(struct otx2_nic *pf, bool enable)
1119 {
1120 	struct msg_req *msg;
1121 	int err;
1122 
1123 	if (enable && bitmap_weight(&pf->flow_cfg->dmacflt_bmap,
1124 				    pf->flow_cfg->dmacflt_max_flows))
1125 		netdev_warn(pf->netdev,
1126 			    "CGX/RPM internal loopback might not work as DMAC filters are active\n");
1127 
1128 	mutex_lock(&pf->mbox.lock);
1129 	if (enable)
1130 		msg = otx2_mbox_alloc_msg_cgx_intlbk_enable(&pf->mbox);
1131 	else
1132 		msg = otx2_mbox_alloc_msg_cgx_intlbk_disable(&pf->mbox);
1133 
1134 	if (!msg) {
1135 		mutex_unlock(&pf->mbox.lock);
1136 		return -ENOMEM;
1137 	}
1138 
1139 	err = otx2_sync_mbox_msg(&pf->mbox);
1140 	mutex_unlock(&pf->mbox.lock);
1141 	return err;
1142 }
1143 
1144 int otx2_set_real_num_queues(struct net_device *netdev,
1145 			     int tx_queues, int rx_queues)
1146 {
1147 	int err;
1148 
1149 	err = netif_set_real_num_tx_queues(netdev, tx_queues);
1150 	if (err) {
1151 		netdev_err(netdev,
1152 			   "Failed to set no of Tx queues: %d\n", tx_queues);
1153 		return err;
1154 	}
1155 
1156 	err = netif_set_real_num_rx_queues(netdev, rx_queues);
1157 	if (err)
1158 		netdev_err(netdev,
1159 			   "Failed to set no of Rx queues: %d\n", rx_queues);
1160 	return err;
1161 }
1162 EXPORT_SYMBOL(otx2_set_real_num_queues);
1163 
1164 static irqreturn_t otx2_q_intr_handler(int irq, void *data)
1165 {
1166 	struct otx2_nic *pf = data;
1167 	u64 val, *ptr;
1168 	u64 qidx = 0;
1169 
1170 	/* CQ */
1171 	for (qidx = 0; qidx < pf->qset.cq_cnt; qidx++) {
1172 		ptr = otx2_get_regaddr(pf, NIX_LF_CQ_OP_INT);
1173 		val = otx2_atomic64_add((qidx << 44), ptr);
1174 
1175 		otx2_write64(pf, NIX_LF_CQ_OP_INT, (qidx << 44) |
1176 			     (val & NIX_CQERRINT_BITS));
1177 		if (!(val & (NIX_CQERRINT_BITS | BIT_ULL(42))))
1178 			continue;
1179 
1180 		if (val & BIT_ULL(42)) {
1181 			netdev_err(pf->netdev, "CQ%lld: error reading NIX_LF_CQ_OP_INT, NIX_LF_ERR_INT 0x%llx\n",
1182 				   qidx, otx2_read64(pf, NIX_LF_ERR_INT));
1183 		} else {
1184 			if (val & BIT_ULL(NIX_CQERRINT_DOOR_ERR))
1185 				netdev_err(pf->netdev, "CQ%lld: Doorbell error",
1186 					   qidx);
1187 			if (val & BIT_ULL(NIX_CQERRINT_CQE_FAULT))
1188 				netdev_err(pf->netdev, "CQ%lld: Memory fault on CQE write to LLC/DRAM",
1189 					   qidx);
1190 		}
1191 
1192 		schedule_work(&pf->reset_task);
1193 	}
1194 
1195 	/* SQ */
1196 	for (qidx = 0; qidx < pf->hw.tot_tx_queues; qidx++) {
1197 		ptr = otx2_get_regaddr(pf, NIX_LF_SQ_OP_INT);
1198 		val = otx2_atomic64_add((qidx << 44), ptr);
1199 		otx2_write64(pf, NIX_LF_SQ_OP_INT, (qidx << 44) |
1200 			     (val & NIX_SQINT_BITS));
1201 
1202 		if (!(val & (NIX_SQINT_BITS | BIT_ULL(42))))
1203 			continue;
1204 
1205 		if (val & BIT_ULL(42)) {
1206 			netdev_err(pf->netdev, "SQ%lld: error reading NIX_LF_SQ_OP_INT, NIX_LF_ERR_INT 0x%llx\n",
1207 				   qidx, otx2_read64(pf, NIX_LF_ERR_INT));
1208 		} else {
1209 			if (val & BIT_ULL(NIX_SQINT_LMT_ERR)) {
1210 				netdev_err(pf->netdev, "SQ%lld: LMT store error NIX_LF_SQ_OP_ERR_DBG:0x%llx",
1211 					   qidx,
1212 					   otx2_read64(pf,
1213 						       NIX_LF_SQ_OP_ERR_DBG));
1214 				otx2_write64(pf, NIX_LF_SQ_OP_ERR_DBG,
1215 					     BIT_ULL(44));
1216 			}
1217 			if (val & BIT_ULL(NIX_SQINT_MNQ_ERR)) {
1218 				netdev_err(pf->netdev, "SQ%lld: Meta-descriptor enqueue error NIX_LF_MNQ_ERR_DGB:0x%llx\n",
1219 					   qidx,
1220 					   otx2_read64(pf, NIX_LF_MNQ_ERR_DBG));
1221 				otx2_write64(pf, NIX_LF_MNQ_ERR_DBG,
1222 					     BIT_ULL(44));
1223 			}
1224 			if (val & BIT_ULL(NIX_SQINT_SEND_ERR)) {
1225 				netdev_err(pf->netdev, "SQ%lld: Send error, NIX_LF_SEND_ERR_DBG 0x%llx",
1226 					   qidx,
1227 					   otx2_read64(pf,
1228 						       NIX_LF_SEND_ERR_DBG));
1229 				otx2_write64(pf, NIX_LF_SEND_ERR_DBG,
1230 					     BIT_ULL(44));
1231 			}
1232 			if (val & BIT_ULL(NIX_SQINT_SQB_ALLOC_FAIL))
1233 				netdev_err(pf->netdev, "SQ%lld: SQB allocation failed",
1234 					   qidx);
1235 		}
1236 
1237 		schedule_work(&pf->reset_task);
1238 	}
1239 
1240 	return IRQ_HANDLED;
1241 }
1242 
1243 static irqreturn_t otx2_cq_intr_handler(int irq, void *cq_irq)
1244 {
1245 	struct otx2_cq_poll *cq_poll = (struct otx2_cq_poll *)cq_irq;
1246 	struct otx2_nic *pf = (struct otx2_nic *)cq_poll->dev;
1247 	int qidx = cq_poll->cint_idx;
1248 
1249 	/* Disable interrupts.
1250 	 *
1251 	 * Completion interrupts behave in a level-triggered interrupt
1252 	 * fashion, and hence have to be cleared only after it is serviced.
1253 	 */
1254 	otx2_write64(pf, NIX_LF_CINTX_ENA_W1C(qidx), BIT_ULL(0));
1255 
1256 	/* Schedule NAPI */
1257 	napi_schedule_irqoff(&cq_poll->napi);
1258 
1259 	return IRQ_HANDLED;
1260 }
1261 
1262 static void otx2_disable_napi(struct otx2_nic *pf)
1263 {
1264 	struct otx2_qset *qset = &pf->qset;
1265 	struct otx2_cq_poll *cq_poll;
1266 	int qidx;
1267 
1268 	for (qidx = 0; qidx < pf->hw.cint_cnt; qidx++) {
1269 		cq_poll = &qset->napi[qidx];
1270 		napi_disable(&cq_poll->napi);
1271 		netif_napi_del(&cq_poll->napi);
1272 	}
1273 }
1274 
1275 static void otx2_free_cq_res(struct otx2_nic *pf)
1276 {
1277 	struct otx2_qset *qset = &pf->qset;
1278 	struct otx2_cq_queue *cq;
1279 	int qidx;
1280 
1281 	/* Disable CQs */
1282 	otx2_ctx_disable(&pf->mbox, NIX_AQ_CTYPE_CQ, false);
1283 	for (qidx = 0; qidx < qset->cq_cnt; qidx++) {
1284 		cq = &qset->cq[qidx];
1285 		qmem_free(pf->dev, cq->cqe);
1286 	}
1287 }
1288 
1289 static void otx2_free_sq_res(struct otx2_nic *pf)
1290 {
1291 	struct otx2_qset *qset = &pf->qset;
1292 	struct otx2_snd_queue *sq;
1293 	int qidx;
1294 
1295 	/* Disable SQs */
1296 	otx2_ctx_disable(&pf->mbox, NIX_AQ_CTYPE_SQ, false);
1297 	/* Free SQB pointers */
1298 	otx2_sq_free_sqbs(pf);
1299 	for (qidx = 0; qidx < pf->hw.tot_tx_queues; qidx++) {
1300 		sq = &qset->sq[qidx];
1301 		qmem_free(pf->dev, sq->sqe);
1302 		qmem_free(pf->dev, sq->tso_hdrs);
1303 		kfree(sq->sg);
1304 		kfree(sq->sqb_ptrs);
1305 	}
1306 }
1307 
1308 static int otx2_get_rbuf_size(struct otx2_nic *pf, int mtu)
1309 {
1310 	int frame_size;
1311 	int total_size;
1312 	int rbuf_size;
1313 
1314 	if (pf->hw.rbuf_len)
1315 		return ALIGN(pf->hw.rbuf_len, OTX2_ALIGN) + OTX2_HEAD_ROOM;
1316 
1317 	/* The data transferred by NIX to memory consists of actual packet
1318 	 * plus additional data which has timestamp and/or EDSA/HIGIG2
1319 	 * headers if interface is configured in corresponding modes.
1320 	 * NIX transfers entire data using 6 segments/buffers and writes
1321 	 * a CQE_RX descriptor with those segment addresses. First segment
1322 	 * has additional data prepended to packet. Also software omits a
1323 	 * headroom of 128 bytes in each segment. Hence the total size of
1324 	 * memory needed to receive a packet with 'mtu' is:
1325 	 * frame size =  mtu + additional data;
1326 	 * memory = frame_size + headroom * 6;
1327 	 * each receive buffer size = memory / 6;
1328 	 */
1329 	frame_size = mtu + OTX2_ETH_HLEN + OTX2_HW_TIMESTAMP_LEN;
1330 	total_size = frame_size + OTX2_HEAD_ROOM * 6;
1331 	rbuf_size = total_size / 6;
1332 
1333 	return ALIGN(rbuf_size, 2048);
1334 }
1335 
1336 static int otx2_init_hw_resources(struct otx2_nic *pf)
1337 {
1338 	struct nix_lf_free_req *free_req;
1339 	struct mbox *mbox = &pf->mbox;
1340 	struct otx2_hw *hw = &pf->hw;
1341 	struct msg_req *req;
1342 	int err = 0, lvl;
1343 
1344 	/* Set required NPA LF's pool counts
1345 	 * Auras and Pools are used in a 1:1 mapping,
1346 	 * so, aura count = pool count.
1347 	 */
1348 	hw->rqpool_cnt = hw->rx_queues;
1349 	hw->sqpool_cnt = hw->tot_tx_queues;
1350 	hw->pool_cnt = hw->rqpool_cnt + hw->sqpool_cnt;
1351 
1352 	/* Maximum hardware supported transmit length */
1353 	pf->tx_max_pktlen = pf->netdev->max_mtu + OTX2_ETH_HLEN;
1354 
1355 	pf->rbsize = otx2_get_rbuf_size(pf, pf->netdev->mtu);
1356 
1357 	mutex_lock(&mbox->lock);
1358 	/* NPA init */
1359 	err = otx2_config_npa(pf);
1360 	if (err)
1361 		goto exit;
1362 
1363 	/* NIX init */
1364 	err = otx2_config_nix(pf);
1365 	if (err)
1366 		goto err_free_npa_lf;
1367 
1368 	/* Enable backpressure */
1369 	otx2_nix_config_bp(pf, true);
1370 
1371 	/* Init Auras and pools used by NIX RQ, for free buffer ptrs */
1372 	err = otx2_rq_aura_pool_init(pf);
1373 	if (err) {
1374 		mutex_unlock(&mbox->lock);
1375 		goto err_free_nix_lf;
1376 	}
1377 	/* Init Auras and pools used by NIX SQ, for queueing SQEs */
1378 	err = otx2_sq_aura_pool_init(pf);
1379 	if (err) {
1380 		mutex_unlock(&mbox->lock);
1381 		goto err_free_rq_ptrs;
1382 	}
1383 
1384 	err = otx2_txsch_alloc(pf);
1385 	if (err) {
1386 		mutex_unlock(&mbox->lock);
1387 		goto err_free_sq_ptrs;
1388 	}
1389 
1390 	err = otx2_config_nix_queues(pf);
1391 	if (err) {
1392 		mutex_unlock(&mbox->lock);
1393 		goto err_free_txsch;
1394 	}
1395 	for (lvl = 0; lvl < NIX_TXSCH_LVL_CNT; lvl++) {
1396 		err = otx2_txschq_config(pf, lvl);
1397 		if (err) {
1398 			mutex_unlock(&mbox->lock);
1399 			goto err_free_nix_queues;
1400 		}
1401 	}
1402 	mutex_unlock(&mbox->lock);
1403 	return err;
1404 
1405 err_free_nix_queues:
1406 	otx2_free_sq_res(pf);
1407 	otx2_free_cq_res(pf);
1408 	otx2_ctx_disable(mbox, NIX_AQ_CTYPE_RQ, false);
1409 err_free_txsch:
1410 	if (otx2_txschq_stop(pf))
1411 		dev_err(pf->dev, "%s failed to stop TX schedulers\n", __func__);
1412 err_free_sq_ptrs:
1413 	otx2_sq_free_sqbs(pf);
1414 err_free_rq_ptrs:
1415 	otx2_free_aura_ptr(pf, AURA_NIX_RQ);
1416 	otx2_ctx_disable(mbox, NPA_AQ_CTYPE_POOL, true);
1417 	otx2_ctx_disable(mbox, NPA_AQ_CTYPE_AURA, true);
1418 	otx2_aura_pool_free(pf);
1419 err_free_nix_lf:
1420 	mutex_lock(&mbox->lock);
1421 	free_req = otx2_mbox_alloc_msg_nix_lf_free(mbox);
1422 	if (free_req) {
1423 		free_req->flags = NIX_LF_DISABLE_FLOWS;
1424 		if (otx2_sync_mbox_msg(mbox))
1425 			dev_err(pf->dev, "%s failed to free nixlf\n", __func__);
1426 	}
1427 err_free_npa_lf:
1428 	/* Reset NPA LF */
1429 	req = otx2_mbox_alloc_msg_npa_lf_free(mbox);
1430 	if (req) {
1431 		if (otx2_sync_mbox_msg(mbox))
1432 			dev_err(pf->dev, "%s failed to free npalf\n", __func__);
1433 	}
1434 exit:
1435 	mutex_unlock(&mbox->lock);
1436 	return err;
1437 }
1438 
1439 static void otx2_free_hw_resources(struct otx2_nic *pf)
1440 {
1441 	struct otx2_qset *qset = &pf->qset;
1442 	struct nix_lf_free_req *free_req;
1443 	struct mbox *mbox = &pf->mbox;
1444 	struct otx2_cq_queue *cq;
1445 	struct msg_req *req;
1446 	int qidx, err;
1447 
1448 	/* Ensure all SQE are processed */
1449 	otx2_sqb_flush(pf);
1450 
1451 	/* Stop transmission */
1452 	err = otx2_txschq_stop(pf);
1453 	if (err)
1454 		dev_err(pf->dev, "RVUPF: Failed to stop/free TX schedulers\n");
1455 
1456 	mutex_lock(&mbox->lock);
1457 	/* Disable backpressure */
1458 	if (!(pf->pcifunc & RVU_PFVF_FUNC_MASK))
1459 		otx2_nix_config_bp(pf, false);
1460 	mutex_unlock(&mbox->lock);
1461 
1462 	/* Disable RQs */
1463 	otx2_ctx_disable(mbox, NIX_AQ_CTYPE_RQ, false);
1464 
1465 	/*Dequeue all CQEs */
1466 	for (qidx = 0; qidx < qset->cq_cnt; qidx++) {
1467 		cq = &qset->cq[qidx];
1468 		if (cq->cq_type == CQ_RX)
1469 			otx2_cleanup_rx_cqes(pf, cq);
1470 		else
1471 			otx2_cleanup_tx_cqes(pf, cq);
1472 	}
1473 
1474 	otx2_free_sq_res(pf);
1475 
1476 	/* Free RQ buffer pointers*/
1477 	otx2_free_aura_ptr(pf, AURA_NIX_RQ);
1478 
1479 	otx2_free_cq_res(pf);
1480 
1481 	/* Free all ingress bandwidth profiles allocated */
1482 	cn10k_free_all_ipolicers(pf);
1483 
1484 	mutex_lock(&mbox->lock);
1485 	/* Reset NIX LF */
1486 	free_req = otx2_mbox_alloc_msg_nix_lf_free(mbox);
1487 	if (free_req) {
1488 		free_req->flags = NIX_LF_DISABLE_FLOWS;
1489 		if (!(pf->flags & OTX2_FLAG_PF_SHUTDOWN))
1490 			free_req->flags |= NIX_LF_DONT_FREE_TX_VTAG;
1491 		if (otx2_sync_mbox_msg(mbox))
1492 			dev_err(pf->dev, "%s failed to free nixlf\n", __func__);
1493 	}
1494 	mutex_unlock(&mbox->lock);
1495 
1496 	/* Disable NPA Pool and Aura hw context */
1497 	otx2_ctx_disable(mbox, NPA_AQ_CTYPE_POOL, true);
1498 	otx2_ctx_disable(mbox, NPA_AQ_CTYPE_AURA, true);
1499 	otx2_aura_pool_free(pf);
1500 
1501 	mutex_lock(&mbox->lock);
1502 	/* Reset NPA LF */
1503 	req = otx2_mbox_alloc_msg_npa_lf_free(mbox);
1504 	if (req) {
1505 		if (otx2_sync_mbox_msg(mbox))
1506 			dev_err(pf->dev, "%s failed to free npalf\n", __func__);
1507 	}
1508 	mutex_unlock(&mbox->lock);
1509 }
1510 
1511 static void otx2_do_set_rx_mode(struct otx2_nic *pf)
1512 {
1513 	struct net_device *netdev = pf->netdev;
1514 	struct nix_rx_mode *req;
1515 	bool promisc = false;
1516 
1517 	if (!(netdev->flags & IFF_UP))
1518 		return;
1519 
1520 	if ((netdev->flags & IFF_PROMISC) ||
1521 	    (netdev_uc_count(netdev) > OTX2_MAX_UNICAST_FLOWS)) {
1522 		promisc = true;
1523 	}
1524 
1525 	/* Write unicast address to mcam entries or del from mcam */
1526 	if (!promisc && netdev->priv_flags & IFF_UNICAST_FLT)
1527 		__dev_uc_sync(netdev, otx2_add_macfilter, otx2_del_macfilter);
1528 
1529 	mutex_lock(&pf->mbox.lock);
1530 	req = otx2_mbox_alloc_msg_nix_set_rx_mode(&pf->mbox);
1531 	if (!req) {
1532 		mutex_unlock(&pf->mbox.lock);
1533 		return;
1534 	}
1535 
1536 	req->mode = NIX_RX_MODE_UCAST;
1537 
1538 	if (promisc)
1539 		req->mode |= NIX_RX_MODE_PROMISC;
1540 	if (netdev->flags & (IFF_ALLMULTI | IFF_MULTICAST))
1541 		req->mode |= NIX_RX_MODE_ALLMULTI;
1542 
1543 	req->mode |= NIX_RX_MODE_USE_MCE;
1544 
1545 	otx2_sync_mbox_msg(&pf->mbox);
1546 	mutex_unlock(&pf->mbox.lock);
1547 }
1548 
1549 int otx2_open(struct net_device *netdev)
1550 {
1551 	struct otx2_nic *pf = netdev_priv(netdev);
1552 	struct otx2_cq_poll *cq_poll = NULL;
1553 	struct otx2_qset *qset = &pf->qset;
1554 	int err = 0, qidx, vec;
1555 	char *irq_name;
1556 
1557 	netif_carrier_off(netdev);
1558 
1559 	pf->qset.cq_cnt = pf->hw.rx_queues + pf->hw.tot_tx_queues;
1560 	/* RQ and SQs are mapped to different CQs,
1561 	 * so find out max CQ IRQs (i.e CINTs) needed.
1562 	 */
1563 	pf->hw.cint_cnt = max(pf->hw.rx_queues, pf->hw.tx_queues);
1564 	qset->napi = kcalloc(pf->hw.cint_cnt, sizeof(*cq_poll), GFP_KERNEL);
1565 	if (!qset->napi)
1566 		return -ENOMEM;
1567 
1568 	/* CQ size of RQ */
1569 	qset->rqe_cnt = qset->rqe_cnt ? qset->rqe_cnt : Q_COUNT(Q_SIZE_256);
1570 	/* CQ size of SQ */
1571 	qset->sqe_cnt = qset->sqe_cnt ? qset->sqe_cnt : Q_COUNT(Q_SIZE_4K);
1572 
1573 	err = -ENOMEM;
1574 	qset->cq = kcalloc(pf->qset.cq_cnt,
1575 			   sizeof(struct otx2_cq_queue), GFP_KERNEL);
1576 	if (!qset->cq)
1577 		goto err_free_mem;
1578 
1579 	qset->sq = kcalloc(pf->hw.tot_tx_queues,
1580 			   sizeof(struct otx2_snd_queue), GFP_KERNEL);
1581 	if (!qset->sq)
1582 		goto err_free_mem;
1583 
1584 	qset->rq = kcalloc(pf->hw.rx_queues,
1585 			   sizeof(struct otx2_rcv_queue), GFP_KERNEL);
1586 	if (!qset->rq)
1587 		goto err_free_mem;
1588 
1589 	err = otx2_init_hw_resources(pf);
1590 	if (err)
1591 		goto err_free_mem;
1592 
1593 	/* Register NAPI handler */
1594 	for (qidx = 0; qidx < pf->hw.cint_cnt; qidx++) {
1595 		cq_poll = &qset->napi[qidx];
1596 		cq_poll->cint_idx = qidx;
1597 		/* RQ0 & SQ0 are mapped to CINT0 and so on..
1598 		 * 'cq_ids[0]' points to RQ's CQ and
1599 		 * 'cq_ids[1]' points to SQ's CQ and
1600 		 * 'cq_ids[2]' points to XDP's CQ and
1601 		 */
1602 		cq_poll->cq_ids[CQ_RX] =
1603 			(qidx <  pf->hw.rx_queues) ? qidx : CINT_INVALID_CQ;
1604 		cq_poll->cq_ids[CQ_TX] = (qidx < pf->hw.tx_queues) ?
1605 				      qidx + pf->hw.rx_queues : CINT_INVALID_CQ;
1606 		if (pf->xdp_prog)
1607 			cq_poll->cq_ids[CQ_XDP] = (qidx < pf->hw.xdp_queues) ?
1608 						  (qidx + pf->hw.rx_queues +
1609 						  pf->hw.tx_queues) :
1610 						  CINT_INVALID_CQ;
1611 		else
1612 			cq_poll->cq_ids[CQ_XDP] = CINT_INVALID_CQ;
1613 
1614 		cq_poll->dev = (void *)pf;
1615 		netif_napi_add(netdev, &cq_poll->napi,
1616 			       otx2_napi_handler, NAPI_POLL_WEIGHT);
1617 		napi_enable(&cq_poll->napi);
1618 	}
1619 
1620 	/* Set maximum frame size allowed in HW */
1621 	err = otx2_hw_set_mtu(pf, netdev->mtu);
1622 	if (err)
1623 		goto err_disable_napi;
1624 
1625 	/* Setup segmentation algorithms, if failed, clear offload capability */
1626 	otx2_setup_segmentation(pf);
1627 
1628 	/* Initialize RSS */
1629 	err = otx2_rss_init(pf);
1630 	if (err)
1631 		goto err_disable_napi;
1632 
1633 	/* Register Queue IRQ handlers */
1634 	vec = pf->hw.nix_msixoff + NIX_LF_QINT_VEC_START;
1635 	irq_name = &pf->hw.irq_name[vec * NAME_SIZE];
1636 
1637 	snprintf(irq_name, NAME_SIZE, "%s-qerr", pf->netdev->name);
1638 
1639 	err = request_irq(pci_irq_vector(pf->pdev, vec),
1640 			  otx2_q_intr_handler, 0, irq_name, pf);
1641 	if (err) {
1642 		dev_err(pf->dev,
1643 			"RVUPF%d: IRQ registration failed for QERR\n",
1644 			rvu_get_pf(pf->pcifunc));
1645 		goto err_disable_napi;
1646 	}
1647 
1648 	/* Enable QINT IRQ */
1649 	otx2_write64(pf, NIX_LF_QINTX_ENA_W1S(0), BIT_ULL(0));
1650 
1651 	/* Register CQ IRQ handlers */
1652 	vec = pf->hw.nix_msixoff + NIX_LF_CINT_VEC_START;
1653 	for (qidx = 0; qidx < pf->hw.cint_cnt; qidx++) {
1654 		irq_name = &pf->hw.irq_name[vec * NAME_SIZE];
1655 
1656 		snprintf(irq_name, NAME_SIZE, "%s-rxtx-%d", pf->netdev->name,
1657 			 qidx);
1658 
1659 		err = request_irq(pci_irq_vector(pf->pdev, vec),
1660 				  otx2_cq_intr_handler, 0, irq_name,
1661 				  &qset->napi[qidx]);
1662 		if (err) {
1663 			dev_err(pf->dev,
1664 				"RVUPF%d: IRQ registration failed for CQ%d\n",
1665 				rvu_get_pf(pf->pcifunc), qidx);
1666 			goto err_free_cints;
1667 		}
1668 		vec++;
1669 
1670 		otx2_config_irq_coalescing(pf, qidx);
1671 
1672 		/* Enable CQ IRQ */
1673 		otx2_write64(pf, NIX_LF_CINTX_INT(qidx), BIT_ULL(0));
1674 		otx2_write64(pf, NIX_LF_CINTX_ENA_W1S(qidx), BIT_ULL(0));
1675 	}
1676 
1677 	otx2_set_cints_affinity(pf);
1678 
1679 	if (pf->flags & OTX2_FLAG_RX_VLAN_SUPPORT)
1680 		otx2_enable_rxvlan(pf, true);
1681 
1682 	/* When reinitializing enable time stamping if it is enabled before */
1683 	if (pf->flags & OTX2_FLAG_TX_TSTAMP_ENABLED) {
1684 		pf->flags &= ~OTX2_FLAG_TX_TSTAMP_ENABLED;
1685 		otx2_config_hw_tx_tstamp(pf, true);
1686 	}
1687 	if (pf->flags & OTX2_FLAG_RX_TSTAMP_ENABLED) {
1688 		pf->flags &= ~OTX2_FLAG_RX_TSTAMP_ENABLED;
1689 		otx2_config_hw_rx_tstamp(pf, true);
1690 	}
1691 
1692 	pf->flags &= ~OTX2_FLAG_INTF_DOWN;
1693 	/* 'intf_down' may be checked on any cpu */
1694 	smp_wmb();
1695 
1696 	/* we have already received link status notification */
1697 	if (pf->linfo.link_up && !(pf->pcifunc & RVU_PFVF_FUNC_MASK))
1698 		otx2_handle_link_event(pf);
1699 
1700 	/* Install DMAC Filters */
1701 	if (pf->flags & OTX2_FLAG_DMACFLTR_SUPPORT)
1702 		otx2_dmacflt_reinstall_flows(pf);
1703 
1704 	err = otx2_rxtx_enable(pf, true);
1705 	if (err)
1706 		goto err_tx_stop_queues;
1707 
1708 	otx2_do_set_rx_mode(pf);
1709 
1710 	return 0;
1711 
1712 err_tx_stop_queues:
1713 	netif_tx_stop_all_queues(netdev);
1714 	netif_carrier_off(netdev);
1715 	pf->flags |= OTX2_FLAG_INTF_DOWN;
1716 err_free_cints:
1717 	otx2_free_cints(pf, qidx);
1718 	vec = pci_irq_vector(pf->pdev,
1719 			     pf->hw.nix_msixoff + NIX_LF_QINT_VEC_START);
1720 	otx2_write64(pf, NIX_LF_QINTX_ENA_W1C(0), BIT_ULL(0));
1721 	synchronize_irq(vec);
1722 	free_irq(vec, pf);
1723 err_disable_napi:
1724 	otx2_disable_napi(pf);
1725 	otx2_free_hw_resources(pf);
1726 err_free_mem:
1727 	kfree(qset->sq);
1728 	kfree(qset->cq);
1729 	kfree(qset->rq);
1730 	kfree(qset->napi);
1731 	return err;
1732 }
1733 EXPORT_SYMBOL(otx2_open);
1734 
1735 int otx2_stop(struct net_device *netdev)
1736 {
1737 	struct otx2_nic *pf = netdev_priv(netdev);
1738 	struct otx2_cq_poll *cq_poll = NULL;
1739 	struct otx2_qset *qset = &pf->qset;
1740 	struct otx2_rss_info *rss;
1741 	int qidx, vec, wrk;
1742 
1743 	/* If the DOWN flag is set resources are already freed */
1744 	if (pf->flags & OTX2_FLAG_INTF_DOWN)
1745 		return 0;
1746 
1747 	netif_carrier_off(netdev);
1748 	netif_tx_stop_all_queues(netdev);
1749 
1750 	pf->flags |= OTX2_FLAG_INTF_DOWN;
1751 	/* 'intf_down' may be checked on any cpu */
1752 	smp_wmb();
1753 
1754 	/* First stop packet Rx/Tx */
1755 	otx2_rxtx_enable(pf, false);
1756 
1757 	/* Clear RSS enable flag */
1758 	rss = &pf->hw.rss_info;
1759 	rss->enable = false;
1760 
1761 	/* Cleanup Queue IRQ */
1762 	vec = pci_irq_vector(pf->pdev,
1763 			     pf->hw.nix_msixoff + NIX_LF_QINT_VEC_START);
1764 	otx2_write64(pf, NIX_LF_QINTX_ENA_W1C(0), BIT_ULL(0));
1765 	synchronize_irq(vec);
1766 	free_irq(vec, pf);
1767 
1768 	/* Cleanup CQ NAPI and IRQ */
1769 	vec = pf->hw.nix_msixoff + NIX_LF_CINT_VEC_START;
1770 	for (qidx = 0; qidx < pf->hw.cint_cnt; qidx++) {
1771 		/* Disable interrupt */
1772 		otx2_write64(pf, NIX_LF_CINTX_ENA_W1C(qidx), BIT_ULL(0));
1773 
1774 		synchronize_irq(pci_irq_vector(pf->pdev, vec));
1775 
1776 		cq_poll = &qset->napi[qidx];
1777 		napi_synchronize(&cq_poll->napi);
1778 		vec++;
1779 	}
1780 
1781 	netif_tx_disable(netdev);
1782 
1783 	otx2_free_hw_resources(pf);
1784 	otx2_free_cints(pf, pf->hw.cint_cnt);
1785 	otx2_disable_napi(pf);
1786 
1787 	for (qidx = 0; qidx < netdev->num_tx_queues; qidx++)
1788 		netdev_tx_reset_queue(netdev_get_tx_queue(netdev, qidx));
1789 
1790 	for (wrk = 0; wrk < pf->qset.cq_cnt; wrk++)
1791 		cancel_delayed_work_sync(&pf->refill_wrk[wrk].pool_refill_work);
1792 	devm_kfree(pf->dev, pf->refill_wrk);
1793 
1794 	kfree(qset->sq);
1795 	kfree(qset->cq);
1796 	kfree(qset->rq);
1797 	kfree(qset->napi);
1798 	/* Do not clear RQ/SQ ringsize settings */
1799 	memset((void *)qset + offsetof(struct otx2_qset, sqe_cnt), 0,
1800 	       sizeof(*qset) - offsetof(struct otx2_qset, sqe_cnt));
1801 	return 0;
1802 }
1803 EXPORT_SYMBOL(otx2_stop);
1804 
1805 static netdev_tx_t otx2_xmit(struct sk_buff *skb, struct net_device *netdev)
1806 {
1807 	struct otx2_nic *pf = netdev_priv(netdev);
1808 	int qidx = skb_get_queue_mapping(skb);
1809 	struct otx2_snd_queue *sq;
1810 	struct netdev_queue *txq;
1811 
1812 	/* Check for minimum and maximum packet length */
1813 	if (skb->len <= ETH_HLEN ||
1814 	    (!skb_shinfo(skb)->gso_size && skb->len > pf->tx_max_pktlen)) {
1815 		dev_kfree_skb(skb);
1816 		return NETDEV_TX_OK;
1817 	}
1818 
1819 	sq = &pf->qset.sq[qidx];
1820 	txq = netdev_get_tx_queue(netdev, qidx);
1821 
1822 	if (!otx2_sq_append_skb(netdev, sq, skb, qidx)) {
1823 		netif_tx_stop_queue(txq);
1824 
1825 		/* Check again, incase SQBs got freed up */
1826 		smp_mb();
1827 		if (((sq->num_sqbs - *sq->aura_fc_addr) * sq->sqe_per_sqb)
1828 							> sq->sqe_thresh)
1829 			netif_tx_wake_queue(txq);
1830 
1831 		return NETDEV_TX_BUSY;
1832 	}
1833 
1834 	return NETDEV_TX_OK;
1835 }
1836 
1837 static netdev_features_t otx2_fix_features(struct net_device *dev,
1838 					   netdev_features_t features)
1839 {
1840 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
1841 		features |= NETIF_F_HW_VLAN_STAG_RX;
1842 	else
1843 		features &= ~NETIF_F_HW_VLAN_STAG_RX;
1844 
1845 	return features;
1846 }
1847 
1848 static void otx2_set_rx_mode(struct net_device *netdev)
1849 {
1850 	struct otx2_nic *pf = netdev_priv(netdev);
1851 
1852 	queue_work(pf->otx2_wq, &pf->rx_mode_work);
1853 }
1854 
1855 static void otx2_rx_mode_wrk_handler(struct work_struct *work)
1856 {
1857 	struct otx2_nic *pf = container_of(work, struct otx2_nic, rx_mode_work);
1858 
1859 	otx2_do_set_rx_mode(pf);
1860 }
1861 
1862 static int otx2_set_features(struct net_device *netdev,
1863 			     netdev_features_t features)
1864 {
1865 	netdev_features_t changed = features ^ netdev->features;
1866 	struct otx2_nic *pf = netdev_priv(netdev);
1867 
1868 	if ((changed & NETIF_F_LOOPBACK) && netif_running(netdev))
1869 		return otx2_cgx_config_loopback(pf,
1870 						features & NETIF_F_LOOPBACK);
1871 
1872 	if ((changed & NETIF_F_HW_VLAN_CTAG_RX) && netif_running(netdev))
1873 		return otx2_enable_rxvlan(pf,
1874 					  features & NETIF_F_HW_VLAN_CTAG_RX);
1875 
1876 	return otx2_handle_ntuple_tc_features(netdev, features);
1877 }
1878 
1879 static void otx2_reset_task(struct work_struct *work)
1880 {
1881 	struct otx2_nic *pf = container_of(work, struct otx2_nic, reset_task);
1882 
1883 	if (!netif_running(pf->netdev))
1884 		return;
1885 
1886 	rtnl_lock();
1887 	otx2_stop(pf->netdev);
1888 	pf->reset_count++;
1889 	otx2_open(pf->netdev);
1890 	netif_trans_update(pf->netdev);
1891 	rtnl_unlock();
1892 }
1893 
1894 static int otx2_config_hw_rx_tstamp(struct otx2_nic *pfvf, bool enable)
1895 {
1896 	struct msg_req *req;
1897 	int err;
1898 
1899 	if (pfvf->flags & OTX2_FLAG_RX_TSTAMP_ENABLED && enable)
1900 		return 0;
1901 
1902 	mutex_lock(&pfvf->mbox.lock);
1903 	if (enable)
1904 		req = otx2_mbox_alloc_msg_cgx_ptp_rx_enable(&pfvf->mbox);
1905 	else
1906 		req = otx2_mbox_alloc_msg_cgx_ptp_rx_disable(&pfvf->mbox);
1907 	if (!req) {
1908 		mutex_unlock(&pfvf->mbox.lock);
1909 		return -ENOMEM;
1910 	}
1911 
1912 	err = otx2_sync_mbox_msg(&pfvf->mbox);
1913 	if (err) {
1914 		mutex_unlock(&pfvf->mbox.lock);
1915 		return err;
1916 	}
1917 
1918 	mutex_unlock(&pfvf->mbox.lock);
1919 	if (enable)
1920 		pfvf->flags |= OTX2_FLAG_RX_TSTAMP_ENABLED;
1921 	else
1922 		pfvf->flags &= ~OTX2_FLAG_RX_TSTAMP_ENABLED;
1923 	return 0;
1924 }
1925 
1926 static int otx2_config_hw_tx_tstamp(struct otx2_nic *pfvf, bool enable)
1927 {
1928 	struct msg_req *req;
1929 	int err;
1930 
1931 	if (pfvf->flags & OTX2_FLAG_TX_TSTAMP_ENABLED && enable)
1932 		return 0;
1933 
1934 	mutex_lock(&pfvf->mbox.lock);
1935 	if (enable)
1936 		req = otx2_mbox_alloc_msg_nix_lf_ptp_tx_enable(&pfvf->mbox);
1937 	else
1938 		req = otx2_mbox_alloc_msg_nix_lf_ptp_tx_disable(&pfvf->mbox);
1939 	if (!req) {
1940 		mutex_unlock(&pfvf->mbox.lock);
1941 		return -ENOMEM;
1942 	}
1943 
1944 	err = otx2_sync_mbox_msg(&pfvf->mbox);
1945 	if (err) {
1946 		mutex_unlock(&pfvf->mbox.lock);
1947 		return err;
1948 	}
1949 
1950 	mutex_unlock(&pfvf->mbox.lock);
1951 	if (enable)
1952 		pfvf->flags |= OTX2_FLAG_TX_TSTAMP_ENABLED;
1953 	else
1954 		pfvf->flags &= ~OTX2_FLAG_TX_TSTAMP_ENABLED;
1955 	return 0;
1956 }
1957 
1958 int otx2_config_hwtstamp(struct net_device *netdev, struct ifreq *ifr)
1959 {
1960 	struct otx2_nic *pfvf = netdev_priv(netdev);
1961 	struct hwtstamp_config config;
1962 
1963 	if (!pfvf->ptp)
1964 		return -ENODEV;
1965 
1966 	if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
1967 		return -EFAULT;
1968 
1969 	switch (config.tx_type) {
1970 	case HWTSTAMP_TX_OFF:
1971 		otx2_config_hw_tx_tstamp(pfvf, false);
1972 		break;
1973 	case HWTSTAMP_TX_ON:
1974 		otx2_config_hw_tx_tstamp(pfvf, true);
1975 		break;
1976 	default:
1977 		return -ERANGE;
1978 	}
1979 
1980 	switch (config.rx_filter) {
1981 	case HWTSTAMP_FILTER_NONE:
1982 		otx2_config_hw_rx_tstamp(pfvf, false);
1983 		break;
1984 	case HWTSTAMP_FILTER_ALL:
1985 	case HWTSTAMP_FILTER_SOME:
1986 	case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
1987 	case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
1988 	case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
1989 	case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
1990 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
1991 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
1992 	case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
1993 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
1994 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
1995 	case HWTSTAMP_FILTER_PTP_V2_EVENT:
1996 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
1997 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
1998 		otx2_config_hw_rx_tstamp(pfvf, true);
1999 		config.rx_filter = HWTSTAMP_FILTER_ALL;
2000 		break;
2001 	default:
2002 		return -ERANGE;
2003 	}
2004 
2005 	memcpy(&pfvf->tstamp, &config, sizeof(config));
2006 
2007 	return copy_to_user(ifr->ifr_data, &config,
2008 			    sizeof(config)) ? -EFAULT : 0;
2009 }
2010 EXPORT_SYMBOL(otx2_config_hwtstamp);
2011 
2012 int otx2_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
2013 {
2014 	struct otx2_nic *pfvf = netdev_priv(netdev);
2015 	struct hwtstamp_config *cfg = &pfvf->tstamp;
2016 
2017 	switch (cmd) {
2018 	case SIOCSHWTSTAMP:
2019 		return otx2_config_hwtstamp(netdev, req);
2020 	case SIOCGHWTSTAMP:
2021 		return copy_to_user(req->ifr_data, cfg,
2022 				    sizeof(*cfg)) ? -EFAULT : 0;
2023 	default:
2024 		return -EOPNOTSUPP;
2025 	}
2026 }
2027 EXPORT_SYMBOL(otx2_ioctl);
2028 
2029 static int otx2_do_set_vf_mac(struct otx2_nic *pf, int vf, const u8 *mac)
2030 {
2031 	struct npc_install_flow_req *req;
2032 	int err;
2033 
2034 	mutex_lock(&pf->mbox.lock);
2035 	req = otx2_mbox_alloc_msg_npc_install_flow(&pf->mbox);
2036 	if (!req) {
2037 		err = -ENOMEM;
2038 		goto out;
2039 	}
2040 
2041 	ether_addr_copy(req->packet.dmac, mac);
2042 	eth_broadcast_addr((u8 *)&req->mask.dmac);
2043 	req->features = BIT_ULL(NPC_DMAC);
2044 	req->channel = pf->hw.rx_chan_base;
2045 	req->intf = NIX_INTF_RX;
2046 	req->default_rule = 1;
2047 	req->append = 1;
2048 	req->vf = vf + 1;
2049 	req->op = NIX_RX_ACTION_DEFAULT;
2050 
2051 	err = otx2_sync_mbox_msg(&pf->mbox);
2052 out:
2053 	mutex_unlock(&pf->mbox.lock);
2054 	return err;
2055 }
2056 
2057 static int otx2_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
2058 {
2059 	struct otx2_nic *pf = netdev_priv(netdev);
2060 	struct pci_dev *pdev = pf->pdev;
2061 	struct otx2_vf_config *config;
2062 	int ret;
2063 
2064 	if (!netif_running(netdev))
2065 		return -EAGAIN;
2066 
2067 	if (vf >= pf->total_vfs)
2068 		return -EINVAL;
2069 
2070 	if (!is_valid_ether_addr(mac))
2071 		return -EINVAL;
2072 
2073 	config = &pf->vf_configs[vf];
2074 	ether_addr_copy(config->mac, mac);
2075 
2076 	ret = otx2_do_set_vf_mac(pf, vf, mac);
2077 	if (ret == 0)
2078 		dev_info(&pdev->dev,
2079 			 "Load/Reload VF driver\n");
2080 
2081 	return ret;
2082 }
2083 
2084 static int otx2_do_set_vf_vlan(struct otx2_nic *pf, int vf, u16 vlan, u8 qos,
2085 			       __be16 proto)
2086 {
2087 	struct otx2_flow_config *flow_cfg = pf->flow_cfg;
2088 	struct nix_vtag_config_rsp *vtag_rsp;
2089 	struct npc_delete_flow_req *del_req;
2090 	struct nix_vtag_config *vtag_req;
2091 	struct npc_install_flow_req *req;
2092 	struct otx2_vf_config *config;
2093 	int err = 0;
2094 	u32 idx;
2095 
2096 	config = &pf->vf_configs[vf];
2097 
2098 	if (!vlan && !config->vlan)
2099 		goto out;
2100 
2101 	mutex_lock(&pf->mbox.lock);
2102 
2103 	/* free old tx vtag entry */
2104 	if (config->vlan) {
2105 		vtag_req = otx2_mbox_alloc_msg_nix_vtag_cfg(&pf->mbox);
2106 		if (!vtag_req) {
2107 			err = -ENOMEM;
2108 			goto out;
2109 		}
2110 		vtag_req->cfg_type = 0;
2111 		vtag_req->tx.free_vtag0 = 1;
2112 		vtag_req->tx.vtag0_idx = config->tx_vtag_idx;
2113 
2114 		err = otx2_sync_mbox_msg(&pf->mbox);
2115 		if (err)
2116 			goto out;
2117 	}
2118 
2119 	if (!vlan && config->vlan) {
2120 		/* rx */
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_RX_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 		if (err)
2131 			goto out;
2132 
2133 		/* tx */
2134 		del_req = otx2_mbox_alloc_msg_npc_delete_flow(&pf->mbox);
2135 		if (!del_req) {
2136 			err = -ENOMEM;
2137 			goto out;
2138 		}
2139 		idx = ((vf * OTX2_PER_VF_VLAN_FLOWS) + OTX2_VF_VLAN_TX_INDEX);
2140 		del_req->entry =
2141 			flow_cfg->def_ent[flow_cfg->vf_vlan_offset + idx];
2142 		err = otx2_sync_mbox_msg(&pf->mbox);
2143 
2144 		goto out;
2145 	}
2146 
2147 	/* rx */
2148 	req = otx2_mbox_alloc_msg_npc_install_flow(&pf->mbox);
2149 	if (!req) {
2150 		err = -ENOMEM;
2151 		goto out;
2152 	}
2153 
2154 	idx = ((vf * OTX2_PER_VF_VLAN_FLOWS) + OTX2_VF_VLAN_RX_INDEX);
2155 	req->entry = flow_cfg->def_ent[flow_cfg->vf_vlan_offset + idx];
2156 	req->packet.vlan_tci = htons(vlan);
2157 	req->mask.vlan_tci = htons(VLAN_VID_MASK);
2158 	/* af fills the destination mac addr */
2159 	eth_broadcast_addr((u8 *)&req->mask.dmac);
2160 	req->features = BIT_ULL(NPC_OUTER_VID) | BIT_ULL(NPC_DMAC);
2161 	req->channel = pf->hw.rx_chan_base;
2162 	req->intf = NIX_INTF_RX;
2163 	req->vf = vf + 1;
2164 	req->op = NIX_RX_ACTION_DEFAULT;
2165 	req->vtag0_valid = true;
2166 	req->vtag0_type = NIX_AF_LFX_RX_VTAG_TYPE7;
2167 	req->set_cntr = 1;
2168 
2169 	err = otx2_sync_mbox_msg(&pf->mbox);
2170 	if (err)
2171 		goto out;
2172 
2173 	/* tx */
2174 	vtag_req = otx2_mbox_alloc_msg_nix_vtag_cfg(&pf->mbox);
2175 	if (!vtag_req) {
2176 		err = -ENOMEM;
2177 		goto out;
2178 	}
2179 
2180 	/* configure tx vtag params */
2181 	vtag_req->vtag_size = VTAGSIZE_T4;
2182 	vtag_req->cfg_type = 0; /* tx vlan cfg */
2183 	vtag_req->tx.cfg_vtag0 = 1;
2184 	vtag_req->tx.vtag0 = ((u64)ntohs(proto) << 16) | vlan;
2185 
2186 	err = otx2_sync_mbox_msg(&pf->mbox);
2187 	if (err)
2188 		goto out;
2189 
2190 	vtag_rsp = (struct nix_vtag_config_rsp *)otx2_mbox_get_rsp
2191 			(&pf->mbox.mbox, 0, &vtag_req->hdr);
2192 	if (IS_ERR(vtag_rsp)) {
2193 		err = PTR_ERR(vtag_rsp);
2194 		goto out;
2195 	}
2196 	config->tx_vtag_idx = vtag_rsp->vtag0_idx;
2197 
2198 	req = otx2_mbox_alloc_msg_npc_install_flow(&pf->mbox);
2199 	if (!req) {
2200 		err = -ENOMEM;
2201 		goto out;
2202 	}
2203 
2204 	eth_zero_addr((u8 *)&req->mask.dmac);
2205 	idx = ((vf * OTX2_PER_VF_VLAN_FLOWS) + OTX2_VF_VLAN_TX_INDEX);
2206 	req->entry = flow_cfg->def_ent[flow_cfg->vf_vlan_offset + idx];
2207 	req->features = BIT_ULL(NPC_DMAC);
2208 	req->channel = pf->hw.tx_chan_base;
2209 	req->intf = NIX_INTF_TX;
2210 	req->vf = vf + 1;
2211 	req->op = NIX_TX_ACTIONOP_UCAST_DEFAULT;
2212 	req->vtag0_def = vtag_rsp->vtag0_idx;
2213 	req->vtag0_op = VTAG_INSERT;
2214 	req->set_cntr = 1;
2215 
2216 	err = otx2_sync_mbox_msg(&pf->mbox);
2217 out:
2218 	config->vlan = vlan;
2219 	mutex_unlock(&pf->mbox.lock);
2220 	return err;
2221 }
2222 
2223 static int otx2_set_vf_vlan(struct net_device *netdev, int vf, u16 vlan, u8 qos,
2224 			    __be16 proto)
2225 {
2226 	struct otx2_nic *pf = netdev_priv(netdev);
2227 	struct pci_dev *pdev = pf->pdev;
2228 
2229 	if (!netif_running(netdev))
2230 		return -EAGAIN;
2231 
2232 	if (vf >= pci_num_vf(pdev))
2233 		return -EINVAL;
2234 
2235 	/* qos is currently unsupported */
2236 	if (vlan >= VLAN_N_VID || qos)
2237 		return -EINVAL;
2238 
2239 	if (proto != htons(ETH_P_8021Q))
2240 		return -EPROTONOSUPPORT;
2241 
2242 	if (!(pf->flags & OTX2_FLAG_VF_VLAN_SUPPORT))
2243 		return -EOPNOTSUPP;
2244 
2245 	return otx2_do_set_vf_vlan(pf, vf, vlan, qos, proto);
2246 }
2247 
2248 static int otx2_get_vf_config(struct net_device *netdev, int vf,
2249 			      struct ifla_vf_info *ivi)
2250 {
2251 	struct otx2_nic *pf = netdev_priv(netdev);
2252 	struct pci_dev *pdev = pf->pdev;
2253 	struct otx2_vf_config *config;
2254 
2255 	if (!netif_running(netdev))
2256 		return -EAGAIN;
2257 
2258 	if (vf >= pci_num_vf(pdev))
2259 		return -EINVAL;
2260 
2261 	config = &pf->vf_configs[vf];
2262 	ivi->vf = vf;
2263 	ether_addr_copy(ivi->mac, config->mac);
2264 	ivi->vlan = config->vlan;
2265 	ivi->trusted = config->trusted;
2266 
2267 	return 0;
2268 }
2269 
2270 static int otx2_xdp_xmit_tx(struct otx2_nic *pf, struct xdp_frame *xdpf,
2271 			    int qidx)
2272 {
2273 	struct page *page;
2274 	u64 dma_addr;
2275 	int err = 0;
2276 
2277 	dma_addr = otx2_dma_map_page(pf, virt_to_page(xdpf->data),
2278 				     offset_in_page(xdpf->data), xdpf->len,
2279 				     DMA_TO_DEVICE);
2280 	if (dma_mapping_error(pf->dev, dma_addr))
2281 		return -ENOMEM;
2282 
2283 	err = otx2_xdp_sq_append_pkt(pf, dma_addr, xdpf->len, qidx);
2284 	if (!err) {
2285 		otx2_dma_unmap_page(pf, dma_addr, xdpf->len, DMA_TO_DEVICE);
2286 		page = virt_to_page(xdpf->data);
2287 		put_page(page);
2288 		return -ENOMEM;
2289 	}
2290 	return 0;
2291 }
2292 
2293 static int otx2_xdp_xmit(struct net_device *netdev, int n,
2294 			 struct xdp_frame **frames, u32 flags)
2295 {
2296 	struct otx2_nic *pf = netdev_priv(netdev);
2297 	int qidx = smp_processor_id();
2298 	struct otx2_snd_queue *sq;
2299 	int drops = 0, i;
2300 
2301 	if (!netif_running(netdev))
2302 		return -ENETDOWN;
2303 
2304 	qidx += pf->hw.tx_queues;
2305 	sq = pf->xdp_prog ? &pf->qset.sq[qidx] : NULL;
2306 
2307 	/* Abort xmit if xdp queue is not */
2308 	if (unlikely(!sq))
2309 		return -ENXIO;
2310 
2311 	if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK))
2312 		return -EINVAL;
2313 
2314 	for (i = 0; i < n; i++) {
2315 		struct xdp_frame *xdpf = frames[i];
2316 		int err;
2317 
2318 		err = otx2_xdp_xmit_tx(pf, xdpf, qidx);
2319 		if (err)
2320 			drops++;
2321 	}
2322 	return n - drops;
2323 }
2324 
2325 static int otx2_xdp_setup(struct otx2_nic *pf, struct bpf_prog *prog)
2326 {
2327 	struct net_device *dev = pf->netdev;
2328 	bool if_up = netif_running(pf->netdev);
2329 	struct bpf_prog *old_prog;
2330 
2331 	if (prog && dev->mtu > MAX_XDP_MTU) {
2332 		netdev_warn(dev, "Jumbo frames not yet supported with XDP\n");
2333 		return -EOPNOTSUPP;
2334 	}
2335 
2336 	if (if_up)
2337 		otx2_stop(pf->netdev);
2338 
2339 	old_prog = xchg(&pf->xdp_prog, prog);
2340 
2341 	if (old_prog)
2342 		bpf_prog_put(old_prog);
2343 
2344 	if (pf->xdp_prog)
2345 		bpf_prog_add(pf->xdp_prog, pf->hw.rx_queues - 1);
2346 
2347 	/* Network stack and XDP shared same rx queues.
2348 	 * Use separate tx queues for XDP and network stack.
2349 	 */
2350 	if (pf->xdp_prog)
2351 		pf->hw.xdp_queues = pf->hw.rx_queues;
2352 	else
2353 		pf->hw.xdp_queues = 0;
2354 
2355 	pf->hw.tot_tx_queues += pf->hw.xdp_queues;
2356 
2357 	if (if_up)
2358 		otx2_open(pf->netdev);
2359 
2360 	return 0;
2361 }
2362 
2363 static int otx2_xdp(struct net_device *netdev, struct netdev_bpf *xdp)
2364 {
2365 	struct otx2_nic *pf = netdev_priv(netdev);
2366 
2367 	switch (xdp->command) {
2368 	case XDP_SETUP_PROG:
2369 		return otx2_xdp_setup(pf, xdp->prog);
2370 	default:
2371 		return -EINVAL;
2372 	}
2373 }
2374 
2375 static int otx2_set_vf_permissions(struct otx2_nic *pf, int vf,
2376 				   int req_perm)
2377 {
2378 	struct set_vf_perm *req;
2379 	int rc;
2380 
2381 	mutex_lock(&pf->mbox.lock);
2382 	req = otx2_mbox_alloc_msg_set_vf_perm(&pf->mbox);
2383 	if (!req) {
2384 		rc = -ENOMEM;
2385 		goto out;
2386 	}
2387 
2388 	/* Let AF reset VF permissions as sriov is disabled */
2389 	if (req_perm == OTX2_RESET_VF_PERM) {
2390 		req->flags |= RESET_VF_PERM;
2391 	} else if (req_perm == OTX2_TRUSTED_VF) {
2392 		if (pf->vf_configs[vf].trusted)
2393 			req->flags |= VF_TRUSTED;
2394 	}
2395 
2396 	req->vf = vf;
2397 	rc = otx2_sync_mbox_msg(&pf->mbox);
2398 out:
2399 	mutex_unlock(&pf->mbox.lock);
2400 	return rc;
2401 }
2402 
2403 static int otx2_ndo_set_vf_trust(struct net_device *netdev, int vf,
2404 				 bool enable)
2405 {
2406 	struct otx2_nic *pf = netdev_priv(netdev);
2407 	struct pci_dev *pdev = pf->pdev;
2408 	int rc;
2409 
2410 	if (vf >= pci_num_vf(pdev))
2411 		return -EINVAL;
2412 
2413 	if (pf->vf_configs[vf].trusted == enable)
2414 		return 0;
2415 
2416 	pf->vf_configs[vf].trusted = enable;
2417 	rc = otx2_set_vf_permissions(pf, vf, OTX2_TRUSTED_VF);
2418 
2419 	if (rc)
2420 		pf->vf_configs[vf].trusted = !enable;
2421 	else
2422 		netdev_info(pf->netdev, "VF %d is %strusted\n",
2423 			    vf, enable ? "" : "not ");
2424 	return rc;
2425 }
2426 
2427 static const struct net_device_ops otx2_netdev_ops = {
2428 	.ndo_open		= otx2_open,
2429 	.ndo_stop		= otx2_stop,
2430 	.ndo_start_xmit		= otx2_xmit,
2431 	.ndo_fix_features	= otx2_fix_features,
2432 	.ndo_set_mac_address    = otx2_set_mac_address,
2433 	.ndo_change_mtu		= otx2_change_mtu,
2434 	.ndo_set_rx_mode	= otx2_set_rx_mode,
2435 	.ndo_set_features	= otx2_set_features,
2436 	.ndo_tx_timeout		= otx2_tx_timeout,
2437 	.ndo_get_stats64	= otx2_get_stats64,
2438 	.ndo_eth_ioctl		= otx2_ioctl,
2439 	.ndo_set_vf_mac		= otx2_set_vf_mac,
2440 	.ndo_set_vf_vlan	= otx2_set_vf_vlan,
2441 	.ndo_get_vf_config	= otx2_get_vf_config,
2442 	.ndo_bpf		= otx2_xdp,
2443 	.ndo_xdp_xmit           = otx2_xdp_xmit,
2444 	.ndo_setup_tc		= otx2_setup_tc,
2445 	.ndo_set_vf_trust	= otx2_ndo_set_vf_trust,
2446 };
2447 
2448 static int otx2_wq_init(struct otx2_nic *pf)
2449 {
2450 	pf->otx2_wq = create_singlethread_workqueue("otx2_wq");
2451 	if (!pf->otx2_wq)
2452 		return -ENOMEM;
2453 
2454 	INIT_WORK(&pf->rx_mode_work, otx2_rx_mode_wrk_handler);
2455 	INIT_WORK(&pf->reset_task, otx2_reset_task);
2456 	return 0;
2457 }
2458 
2459 static int otx2_check_pf_usable(struct otx2_nic *nic)
2460 {
2461 	u64 rev;
2462 
2463 	rev = otx2_read64(nic, RVU_PF_BLOCK_ADDRX_DISC(BLKADDR_RVUM));
2464 	rev = (rev >> 12) & 0xFF;
2465 	/* Check if AF has setup revision for RVUM block,
2466 	 * otherwise this driver probe should be deferred
2467 	 * until AF driver comes up.
2468 	 */
2469 	if (!rev) {
2470 		dev_warn(nic->dev,
2471 			 "AF is not initialized, deferring probe\n");
2472 		return -EPROBE_DEFER;
2473 	}
2474 	return 0;
2475 }
2476 
2477 static int otx2_realloc_msix_vectors(struct otx2_nic *pf)
2478 {
2479 	struct otx2_hw *hw = &pf->hw;
2480 	int num_vec, err;
2481 
2482 	/* NPA interrupts are inot registered, so alloc only
2483 	 * upto NIX vector offset.
2484 	 */
2485 	num_vec = hw->nix_msixoff;
2486 	num_vec += NIX_LF_CINT_VEC_START + hw->max_queues;
2487 
2488 	otx2_disable_mbox_intr(pf);
2489 	pci_free_irq_vectors(hw->pdev);
2490 	err = pci_alloc_irq_vectors(hw->pdev, num_vec, num_vec, PCI_IRQ_MSIX);
2491 	if (err < 0) {
2492 		dev_err(pf->dev, "%s: Failed to realloc %d IRQ vectors\n",
2493 			__func__, num_vec);
2494 		return err;
2495 	}
2496 
2497 	return otx2_register_mbox_intr(pf, false);
2498 }
2499 
2500 static int otx2_sriov_vfcfg_init(struct otx2_nic *pf)
2501 {
2502 	int i;
2503 
2504 	pf->vf_configs = devm_kcalloc(pf->dev, pf->total_vfs,
2505 				      sizeof(struct otx2_vf_config),
2506 				      GFP_KERNEL);
2507 	if (!pf->vf_configs)
2508 		return -ENOMEM;
2509 
2510 	for (i = 0; i < pf->total_vfs; i++) {
2511 		pf->vf_configs[i].pf = pf;
2512 		pf->vf_configs[i].intf_down = true;
2513 		pf->vf_configs[i].trusted = false;
2514 		INIT_DELAYED_WORK(&pf->vf_configs[i].link_event_work,
2515 				  otx2_vf_link_event_task);
2516 	}
2517 
2518 	return 0;
2519 }
2520 
2521 static void otx2_sriov_vfcfg_cleanup(struct otx2_nic *pf)
2522 {
2523 	int i;
2524 
2525 	if (!pf->vf_configs)
2526 		return;
2527 
2528 	for (i = 0; i < pf->total_vfs; i++) {
2529 		cancel_delayed_work_sync(&pf->vf_configs[i].link_event_work);
2530 		otx2_set_vf_permissions(pf, i, OTX2_RESET_VF_PERM);
2531 	}
2532 }
2533 
2534 static int otx2_probe(struct pci_dev *pdev, const struct pci_device_id *id)
2535 {
2536 	struct device *dev = &pdev->dev;
2537 	struct net_device *netdev;
2538 	struct otx2_nic *pf;
2539 	struct otx2_hw *hw;
2540 	int err, qcount;
2541 	int num_vec;
2542 
2543 	err = pcim_enable_device(pdev);
2544 	if (err) {
2545 		dev_err(dev, "Failed to enable PCI device\n");
2546 		return err;
2547 	}
2548 
2549 	err = pci_request_regions(pdev, DRV_NAME);
2550 	if (err) {
2551 		dev_err(dev, "PCI request regions failed 0x%x\n", err);
2552 		return err;
2553 	}
2554 
2555 	err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(48));
2556 	if (err) {
2557 		dev_err(dev, "DMA mask config failed, abort\n");
2558 		goto err_release_regions;
2559 	}
2560 
2561 	pci_set_master(pdev);
2562 
2563 	/* Set number of queues */
2564 	qcount = min_t(int, num_online_cpus(), OTX2_MAX_CQ_CNT);
2565 
2566 	netdev = alloc_etherdev_mqs(sizeof(*pf), qcount, qcount);
2567 	if (!netdev) {
2568 		err = -ENOMEM;
2569 		goto err_release_regions;
2570 	}
2571 
2572 	pci_set_drvdata(pdev, netdev);
2573 	SET_NETDEV_DEV(netdev, &pdev->dev);
2574 	pf = netdev_priv(netdev);
2575 	pf->netdev = netdev;
2576 	pf->pdev = pdev;
2577 	pf->dev = dev;
2578 	pf->total_vfs = pci_sriov_get_totalvfs(pdev);
2579 	pf->flags |= OTX2_FLAG_INTF_DOWN;
2580 
2581 	hw = &pf->hw;
2582 	hw->pdev = pdev;
2583 	hw->rx_queues = qcount;
2584 	hw->tx_queues = qcount;
2585 	hw->tot_tx_queues = qcount;
2586 	hw->max_queues = qcount;
2587 	hw->rbuf_len = OTX2_DEFAULT_RBUF_LEN;
2588 	/* Use CQE of 128 byte descriptor size by default */
2589 	hw->xqe_size = 128;
2590 
2591 	num_vec = pci_msix_vec_count(pdev);
2592 	hw->irq_name = devm_kmalloc_array(&hw->pdev->dev, num_vec, NAME_SIZE,
2593 					  GFP_KERNEL);
2594 	if (!hw->irq_name) {
2595 		err = -ENOMEM;
2596 		goto err_free_netdev;
2597 	}
2598 
2599 	hw->affinity_mask = devm_kcalloc(&hw->pdev->dev, num_vec,
2600 					 sizeof(cpumask_var_t), GFP_KERNEL);
2601 	if (!hw->affinity_mask) {
2602 		err = -ENOMEM;
2603 		goto err_free_netdev;
2604 	}
2605 
2606 	/* Map CSRs */
2607 	pf->reg_base = pcim_iomap(pdev, PCI_CFG_REG_BAR_NUM, 0);
2608 	if (!pf->reg_base) {
2609 		dev_err(dev, "Unable to map physical function CSRs, aborting\n");
2610 		err = -ENOMEM;
2611 		goto err_free_netdev;
2612 	}
2613 
2614 	err = otx2_check_pf_usable(pf);
2615 	if (err)
2616 		goto err_free_netdev;
2617 
2618 	err = pci_alloc_irq_vectors(hw->pdev, RVU_PF_INT_VEC_CNT,
2619 				    RVU_PF_INT_VEC_CNT, PCI_IRQ_MSIX);
2620 	if (err < 0) {
2621 		dev_err(dev, "%s: Failed to alloc %d IRQ vectors\n",
2622 			__func__, num_vec);
2623 		goto err_free_netdev;
2624 	}
2625 
2626 	otx2_setup_dev_hw_settings(pf);
2627 
2628 	/* Init PF <=> AF mailbox stuff */
2629 	err = otx2_pfaf_mbox_init(pf);
2630 	if (err)
2631 		goto err_free_irq_vectors;
2632 
2633 	/* Register mailbox interrupt */
2634 	err = otx2_register_mbox_intr(pf, true);
2635 	if (err)
2636 		goto err_mbox_destroy;
2637 
2638 	/* Request AF to attach NPA and NIX LFs to this PF.
2639 	 * NIX and NPA LFs are needed for this PF to function as a NIC.
2640 	 */
2641 	err = otx2_attach_npa_nix(pf);
2642 	if (err)
2643 		goto err_disable_mbox_intr;
2644 
2645 	err = otx2_realloc_msix_vectors(pf);
2646 	if (err)
2647 		goto err_detach_rsrc;
2648 
2649 	err = otx2_set_real_num_queues(netdev, hw->tx_queues, hw->rx_queues);
2650 	if (err)
2651 		goto err_detach_rsrc;
2652 
2653 	err = cn10k_lmtst_init(pf);
2654 	if (err)
2655 		goto err_detach_rsrc;
2656 
2657 	/* Assign default mac address */
2658 	otx2_get_mac_from_af(netdev);
2659 
2660 	/* Don't check for error.  Proceed without ptp */
2661 	otx2_ptp_init(pf);
2662 
2663 	/* NPA's pool is a stack to which SW frees buffer pointers via Aura.
2664 	 * HW allocates buffer pointer from stack and uses it for DMA'ing
2665 	 * ingress packet. In some scenarios HW can free back allocated buffer
2666 	 * pointers to pool. This makes it impossible for SW to maintain a
2667 	 * parallel list where physical addresses of buffer pointers (IOVAs)
2668 	 * given to HW can be saved for later reference.
2669 	 *
2670 	 * So the only way to convert Rx packet's buffer address is to use
2671 	 * IOMMU's iova_to_phys() handler which translates the address by
2672 	 * walking through the translation tables.
2673 	 */
2674 	pf->iommu_domain = iommu_get_domain_for_dev(dev);
2675 
2676 	netdev->hw_features = (NETIF_F_RXCSUM | NETIF_F_IP_CSUM |
2677 			       NETIF_F_IPV6_CSUM | NETIF_F_RXHASH |
2678 			       NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
2679 			       NETIF_F_GSO_UDP_L4);
2680 	netdev->features |= netdev->hw_features;
2681 
2682 	err = otx2_mcam_flow_init(pf);
2683 	if (err)
2684 		goto err_ptp_destroy;
2685 
2686 	if (pf->flags & OTX2_FLAG_NTUPLE_SUPPORT)
2687 		netdev->hw_features |= NETIF_F_NTUPLE;
2688 
2689 	if (pf->flags & OTX2_FLAG_UCAST_FLTR_SUPPORT)
2690 		netdev->priv_flags |= IFF_UNICAST_FLT;
2691 
2692 	/* Support TSO on tag interface */
2693 	netdev->vlan_features |= netdev->features;
2694 	netdev->hw_features  |= NETIF_F_HW_VLAN_CTAG_TX |
2695 				NETIF_F_HW_VLAN_STAG_TX;
2696 	if (pf->flags & OTX2_FLAG_RX_VLAN_SUPPORT)
2697 		netdev->hw_features |= NETIF_F_HW_VLAN_CTAG_RX |
2698 				       NETIF_F_HW_VLAN_STAG_RX;
2699 	netdev->features |= netdev->hw_features;
2700 
2701 	/* HW supports tc offload but mutually exclusive with n-tuple filters */
2702 	if (pf->flags & OTX2_FLAG_TC_FLOWER_SUPPORT)
2703 		netdev->hw_features |= NETIF_F_HW_TC;
2704 
2705 	netdev->hw_features |= NETIF_F_LOOPBACK | NETIF_F_RXALL;
2706 
2707 	netif_set_gso_max_segs(netdev, OTX2_MAX_GSO_SEGS);
2708 	netdev->watchdog_timeo = OTX2_TX_TIMEOUT;
2709 
2710 	netdev->netdev_ops = &otx2_netdev_ops;
2711 
2712 	netdev->min_mtu = OTX2_MIN_MTU;
2713 	netdev->max_mtu = otx2_get_max_mtu(pf);
2714 
2715 	err = register_netdev(netdev);
2716 	if (err) {
2717 		dev_err(dev, "Failed to register netdevice\n");
2718 		goto err_del_mcam_entries;
2719 	}
2720 
2721 	err = otx2_wq_init(pf);
2722 	if (err)
2723 		goto err_unreg_netdev;
2724 
2725 	otx2_set_ethtool_ops(netdev);
2726 
2727 	err = otx2_init_tc(pf);
2728 	if (err)
2729 		goto err_mcam_flow_del;
2730 
2731 	err = otx2_register_dl(pf);
2732 	if (err)
2733 		goto err_mcam_flow_del;
2734 
2735 	/* Initialize SR-IOV resources */
2736 	err = otx2_sriov_vfcfg_init(pf);
2737 	if (err)
2738 		goto err_pf_sriov_init;
2739 
2740 	/* Enable link notifications */
2741 	otx2_cgx_config_linkevents(pf, true);
2742 
2743 #ifdef CONFIG_DCB
2744 	err = otx2_dcbnl_set_ops(netdev);
2745 	if (err)
2746 		goto err_pf_sriov_init;
2747 #endif
2748 
2749 	return 0;
2750 
2751 err_pf_sriov_init:
2752 	otx2_shutdown_tc(pf);
2753 err_mcam_flow_del:
2754 	otx2_mcam_flow_del(pf);
2755 err_unreg_netdev:
2756 	unregister_netdev(netdev);
2757 err_del_mcam_entries:
2758 	otx2_mcam_flow_del(pf);
2759 err_ptp_destroy:
2760 	otx2_ptp_destroy(pf);
2761 err_detach_rsrc:
2762 	if (pf->hw.lmt_info)
2763 		free_percpu(pf->hw.lmt_info);
2764 	if (test_bit(CN10K_LMTST, &pf->hw.cap_flag))
2765 		qmem_free(pf->dev, pf->dync_lmt);
2766 	otx2_detach_resources(&pf->mbox);
2767 err_disable_mbox_intr:
2768 	otx2_disable_mbox_intr(pf);
2769 err_mbox_destroy:
2770 	otx2_pfaf_mbox_destroy(pf);
2771 err_free_irq_vectors:
2772 	pci_free_irq_vectors(hw->pdev);
2773 err_free_netdev:
2774 	pci_set_drvdata(pdev, NULL);
2775 	free_netdev(netdev);
2776 err_release_regions:
2777 	pci_release_regions(pdev);
2778 	return err;
2779 }
2780 
2781 static void otx2_vf_link_event_task(struct work_struct *work)
2782 {
2783 	struct otx2_vf_config *config;
2784 	struct cgx_link_info_msg *req;
2785 	struct mbox_msghdr *msghdr;
2786 	struct otx2_nic *pf;
2787 	int vf_idx;
2788 
2789 	config = container_of(work, struct otx2_vf_config,
2790 			      link_event_work.work);
2791 	vf_idx = config - config->pf->vf_configs;
2792 	pf = config->pf;
2793 
2794 	msghdr = otx2_mbox_alloc_msg_rsp(&pf->mbox_pfvf[0].mbox_up, vf_idx,
2795 					 sizeof(*req), sizeof(struct msg_rsp));
2796 	if (!msghdr) {
2797 		dev_err(pf->dev, "Failed to create VF%d link event\n", vf_idx);
2798 		return;
2799 	}
2800 
2801 	req = (struct cgx_link_info_msg *)msghdr;
2802 	req->hdr.id = MBOX_MSG_CGX_LINK_EVENT;
2803 	req->hdr.sig = OTX2_MBOX_REQ_SIG;
2804 	memcpy(&req->link_info, &pf->linfo, sizeof(req->link_info));
2805 
2806 	otx2_sync_mbox_up_msg(&pf->mbox_pfvf[0], vf_idx);
2807 }
2808 
2809 static int otx2_sriov_enable(struct pci_dev *pdev, int numvfs)
2810 {
2811 	struct net_device *netdev = pci_get_drvdata(pdev);
2812 	struct otx2_nic *pf = netdev_priv(netdev);
2813 	int ret;
2814 
2815 	/* Init PF <=> VF mailbox stuff */
2816 	ret = otx2_pfvf_mbox_init(pf, numvfs);
2817 	if (ret)
2818 		return ret;
2819 
2820 	ret = otx2_register_pfvf_mbox_intr(pf, numvfs);
2821 	if (ret)
2822 		goto free_mbox;
2823 
2824 	ret = otx2_pf_flr_init(pf, numvfs);
2825 	if (ret)
2826 		goto free_intr;
2827 
2828 	ret = otx2_register_flr_me_intr(pf, numvfs);
2829 	if (ret)
2830 		goto free_flr;
2831 
2832 	ret = pci_enable_sriov(pdev, numvfs);
2833 	if (ret)
2834 		goto free_flr_intr;
2835 
2836 	return numvfs;
2837 free_flr_intr:
2838 	otx2_disable_flr_me_intr(pf);
2839 free_flr:
2840 	otx2_flr_wq_destroy(pf);
2841 free_intr:
2842 	otx2_disable_pfvf_mbox_intr(pf, numvfs);
2843 free_mbox:
2844 	otx2_pfvf_mbox_destroy(pf);
2845 	return ret;
2846 }
2847 
2848 static int otx2_sriov_disable(struct pci_dev *pdev)
2849 {
2850 	struct net_device *netdev = pci_get_drvdata(pdev);
2851 	struct otx2_nic *pf = netdev_priv(netdev);
2852 	int numvfs = pci_num_vf(pdev);
2853 
2854 	if (!numvfs)
2855 		return 0;
2856 
2857 	pci_disable_sriov(pdev);
2858 
2859 	otx2_disable_flr_me_intr(pf);
2860 	otx2_flr_wq_destroy(pf);
2861 	otx2_disable_pfvf_mbox_intr(pf, numvfs);
2862 	otx2_pfvf_mbox_destroy(pf);
2863 
2864 	return 0;
2865 }
2866 
2867 static int otx2_sriov_configure(struct pci_dev *pdev, int numvfs)
2868 {
2869 	if (numvfs == 0)
2870 		return otx2_sriov_disable(pdev);
2871 	else
2872 		return otx2_sriov_enable(pdev, numvfs);
2873 }
2874 
2875 static void otx2_remove(struct pci_dev *pdev)
2876 {
2877 	struct net_device *netdev = pci_get_drvdata(pdev);
2878 	struct otx2_nic *pf;
2879 
2880 	if (!netdev)
2881 		return;
2882 
2883 	pf = netdev_priv(netdev);
2884 
2885 	pf->flags |= OTX2_FLAG_PF_SHUTDOWN;
2886 
2887 	if (pf->flags & OTX2_FLAG_TX_TSTAMP_ENABLED)
2888 		otx2_config_hw_tx_tstamp(pf, false);
2889 	if (pf->flags & OTX2_FLAG_RX_TSTAMP_ENABLED)
2890 		otx2_config_hw_rx_tstamp(pf, false);
2891 
2892 	/* Disable 802.3x pause frames */
2893 	if (pf->flags & OTX2_FLAG_RX_PAUSE_ENABLED ||
2894 	    (pf->flags & OTX2_FLAG_TX_PAUSE_ENABLED)) {
2895 		pf->flags &= ~OTX2_FLAG_RX_PAUSE_ENABLED;
2896 		pf->flags &= ~OTX2_FLAG_TX_PAUSE_ENABLED;
2897 		otx2_config_pause_frm(pf);
2898 	}
2899 
2900 #ifdef CONFIG_DCB
2901 	/* Disable PFC config */
2902 	if (pf->pfc_en) {
2903 		pf->pfc_en = 0;
2904 		otx2_config_priority_flow_ctrl(pf);
2905 	}
2906 #endif
2907 	cancel_work_sync(&pf->reset_task);
2908 	/* Disable link notifications */
2909 	otx2_cgx_config_linkevents(pf, false);
2910 
2911 	otx2_unregister_dl(pf);
2912 	unregister_netdev(netdev);
2913 	otx2_sriov_disable(pf->pdev);
2914 	otx2_sriov_vfcfg_cleanup(pf);
2915 	if (pf->otx2_wq)
2916 		destroy_workqueue(pf->otx2_wq);
2917 
2918 	otx2_ptp_destroy(pf);
2919 	otx2_mcam_flow_del(pf);
2920 	otx2_shutdown_tc(pf);
2921 	otx2_detach_resources(&pf->mbox);
2922 	if (pf->hw.lmt_info)
2923 		free_percpu(pf->hw.lmt_info);
2924 	if (test_bit(CN10K_LMTST, &pf->hw.cap_flag))
2925 		qmem_free(pf->dev, pf->dync_lmt);
2926 	otx2_disable_mbox_intr(pf);
2927 	otx2_pfaf_mbox_destroy(pf);
2928 	pci_free_irq_vectors(pf->pdev);
2929 	pci_set_drvdata(pdev, NULL);
2930 	free_netdev(netdev);
2931 
2932 	pci_release_regions(pdev);
2933 }
2934 
2935 static struct pci_driver otx2_pf_driver = {
2936 	.name = DRV_NAME,
2937 	.id_table = otx2_pf_id_table,
2938 	.probe = otx2_probe,
2939 	.shutdown = otx2_remove,
2940 	.remove = otx2_remove,
2941 	.sriov_configure = otx2_sriov_configure
2942 };
2943 
2944 static int __init otx2_rvupf_init_module(void)
2945 {
2946 	pr_info("%s: %s\n", DRV_NAME, DRV_STRING);
2947 
2948 	return pci_register_driver(&otx2_pf_driver);
2949 }
2950 
2951 static void __exit otx2_rvupf_cleanup_module(void)
2952 {
2953 	pci_unregister_driver(&otx2_pf_driver);
2954 }
2955 
2956 module_init(otx2_rvupf_init_module);
2957 module_exit(otx2_rvupf_cleanup_module);
2958