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 		if (err) {
398 			dev_warn(pf->dev,
399 				 "AF not responding to VF%d messages\n", vf);
400 			/* restore PF mbase and exit */
401 			dst_mdev->mbase = pf->mbox.bbuf_base;
402 			mutex_unlock(&pf->mbox.lock);
403 			return err;
404 		}
405 		/* At this point, all the VF messages sent to AF are acked
406 		 * with proper responses and responses are copied to VF
407 		 * mailbox hence raise interrupt to VF.
408 		 */
409 		req_hdr = (struct mbox_hdr *)(dst_mdev->mbase +
410 					      dst_mbox->mbox.rx_start);
411 		req_hdr->num_msgs = num_msgs;
412 
413 		otx2_forward_msg_pfvf(dst_mdev, &pf->mbox_pfvf[0].mbox,
414 				      pf->mbox.bbuf_base, vf);
415 		mutex_unlock(&pf->mbox.lock);
416 	} else if (dir == MBOX_DIR_PFVF_UP) {
417 		src_mdev = &src_mbox->dev[0];
418 		mbox_hdr = src_mbox->hwbase + src_mbox->rx_start;
419 		req_hdr = (struct mbox_hdr *)(src_mdev->mbase +
420 					      src_mbox->rx_start);
421 		req_hdr->num_msgs = num_msgs;
422 
423 		dst_mbox = &pf->mbox_pfvf[0];
424 		dst_size = dst_mbox->mbox_up.tx_size -
425 				ALIGN(sizeof(*mbox_hdr), MBOX_MSG_ALIGN);
426 		/* Check if msgs fit into destination area */
427 		if (mbox_hdr->msg_size > dst_size)
428 			return -EINVAL;
429 
430 		dst_mdev = &dst_mbox->mbox_up.dev[vf];
431 		dst_mdev->mbase = src_mdev->mbase;
432 		dst_mdev->msg_size = mbox_hdr->msg_size;
433 		dst_mdev->num_msgs = mbox_hdr->num_msgs;
434 		err = otx2_sync_mbox_up_msg(dst_mbox, vf);
435 		if (err) {
436 			dev_warn(pf->dev,
437 				 "VF%d is not responding to mailbox\n", vf);
438 			return err;
439 		}
440 	} else if (dir == MBOX_DIR_VFPF_UP) {
441 		req_hdr = (struct mbox_hdr *)(src_mbox->dev[0].mbase +
442 					      src_mbox->rx_start);
443 		req_hdr->num_msgs = num_msgs;
444 		otx2_forward_msg_pfvf(&pf->mbox_pfvf->mbox_up.dev[vf],
445 				      &pf->mbox.mbox_up,
446 				      pf->mbox_pfvf[vf].bbuf_base,
447 				      0);
448 	}
449 
450 	return 0;
451 }
452 
453 static void otx2_pfvf_mbox_handler(struct work_struct *work)
454 {
455 	struct mbox_msghdr *msg = NULL;
456 	int offset, vf_idx, id, err;
457 	struct otx2_mbox_dev *mdev;
458 	struct mbox_hdr *req_hdr;
459 	struct otx2_mbox *mbox;
460 	struct mbox *vf_mbox;
461 	struct otx2_nic *pf;
462 
463 	vf_mbox = container_of(work, struct mbox, mbox_wrk);
464 	pf = vf_mbox->pfvf;
465 	vf_idx = vf_mbox - pf->mbox_pfvf;
466 
467 	mbox = &pf->mbox_pfvf[0].mbox;
468 	mdev = &mbox->dev[vf_idx];
469 	req_hdr = (struct mbox_hdr *)(mdev->mbase + mbox->rx_start);
470 
471 	offset = ALIGN(sizeof(*req_hdr), MBOX_MSG_ALIGN);
472 
473 	for (id = 0; id < vf_mbox->num_msgs; id++) {
474 		msg = (struct mbox_msghdr *)(mdev->mbase + mbox->rx_start +
475 					     offset);
476 
477 		if (msg->sig != OTX2_MBOX_REQ_SIG)
478 			goto inval_msg;
479 
480 		/* Set VF's number in each of the msg */
481 		msg->pcifunc &= RVU_PFVF_FUNC_MASK;
482 		msg->pcifunc |= (vf_idx + 1) & RVU_PFVF_FUNC_MASK;
483 		offset = msg->next_msgoff;
484 	}
485 	err = otx2_forward_vf_mbox_msgs(pf, mbox, MBOX_DIR_PFAF, vf_idx,
486 					vf_mbox->num_msgs);
487 	if (err)
488 		goto inval_msg;
489 	return;
490 
491 inval_msg:
492 	otx2_reply_invalid_msg(mbox, vf_idx, 0, msg->id);
493 	otx2_mbox_msg_send(mbox, vf_idx);
494 }
495 
496 static void otx2_pfvf_mbox_up_handler(struct work_struct *work)
497 {
498 	struct mbox *vf_mbox = container_of(work, struct mbox, mbox_up_wrk);
499 	struct otx2_nic *pf = vf_mbox->pfvf;
500 	struct otx2_mbox_dev *mdev;
501 	int offset, id, vf_idx = 0;
502 	struct mbox_hdr *rsp_hdr;
503 	struct mbox_msghdr *msg;
504 	struct otx2_mbox *mbox;
505 
506 	vf_idx = vf_mbox - pf->mbox_pfvf;
507 	mbox = &pf->mbox_pfvf[0].mbox_up;
508 	mdev = &mbox->dev[vf_idx];
509 
510 	rsp_hdr = (struct mbox_hdr *)(mdev->mbase + mbox->rx_start);
511 	offset = mbox->rx_start + ALIGN(sizeof(*rsp_hdr), MBOX_MSG_ALIGN);
512 
513 	for (id = 0; id < vf_mbox->up_num_msgs; id++) {
514 		msg = mdev->mbase + offset;
515 
516 		if (msg->id >= MBOX_MSG_MAX) {
517 			dev_err(pf->dev,
518 				"Mbox msg with unknown ID 0x%x\n", msg->id);
519 			goto end;
520 		}
521 
522 		if (msg->sig != OTX2_MBOX_RSP_SIG) {
523 			dev_err(pf->dev,
524 				"Mbox msg with wrong signature %x, ID 0x%x\n",
525 				msg->sig, msg->id);
526 			goto end;
527 		}
528 
529 		switch (msg->id) {
530 		case MBOX_MSG_CGX_LINK_EVENT:
531 			break;
532 		default:
533 			if (msg->rc)
534 				dev_err(pf->dev,
535 					"Mbox msg response has err %d, ID 0x%x\n",
536 					msg->rc, msg->id);
537 			break;
538 		}
539 
540 end:
541 		offset = mbox->rx_start + msg->next_msgoff;
542 		if (mdev->msgs_acked == (vf_mbox->up_num_msgs - 1))
543 			__otx2_mbox_reset(mbox, 0);
544 		mdev->msgs_acked++;
545 	}
546 }
547 
548 static irqreturn_t otx2_pfvf_mbox_intr_handler(int irq, void *pf_irq)
549 {
550 	struct otx2_nic *pf = (struct otx2_nic *)(pf_irq);
551 	int vfs = pf->total_vfs;
552 	struct mbox *mbox;
553 	u64 intr;
554 
555 	mbox = pf->mbox_pfvf;
556 	/* Handle VF interrupts */
557 	if (vfs > 64) {
558 		intr = otx2_read64(pf, RVU_PF_VFPF_MBOX_INTX(1));
559 		otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(1), intr);
560 		otx2_queue_work(mbox, pf->mbox_pfvf_wq, 64, vfs, intr,
561 				TYPE_PFVF);
562 		vfs -= 64;
563 	}
564 
565 	intr = otx2_read64(pf, RVU_PF_VFPF_MBOX_INTX(0));
566 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(0), intr);
567 
568 	otx2_queue_work(mbox, pf->mbox_pfvf_wq, 0, vfs, intr, TYPE_PFVF);
569 
570 	trace_otx2_msg_interrupt(mbox->mbox.pdev, "VF(s) to PF", intr);
571 
572 	return IRQ_HANDLED;
573 }
574 
575 static int otx2_pfvf_mbox_init(struct otx2_nic *pf, int numvfs)
576 {
577 	void __iomem *hwbase;
578 	struct mbox *mbox;
579 	int err, vf;
580 	u64 base;
581 
582 	if (!numvfs)
583 		return -EINVAL;
584 
585 	pf->mbox_pfvf = devm_kcalloc(&pf->pdev->dev, numvfs,
586 				     sizeof(struct mbox), GFP_KERNEL);
587 	if (!pf->mbox_pfvf)
588 		return -ENOMEM;
589 
590 	pf->mbox_pfvf_wq = alloc_workqueue("otx2_pfvf_mailbox",
591 					   WQ_UNBOUND | WQ_HIGHPRI |
592 					   WQ_MEM_RECLAIM, 1);
593 	if (!pf->mbox_pfvf_wq)
594 		return -ENOMEM;
595 
596 	/* On CN10K platform, PF <-> VF mailbox region follows after
597 	 * PF <-> AF mailbox region.
598 	 */
599 	if (test_bit(CN10K_MBOX, &pf->hw.cap_flag))
600 		base = pci_resource_start(pf->pdev, PCI_MBOX_BAR_NUM) +
601 		       MBOX_SIZE;
602 	else
603 		base = readq((void __iomem *)((u64)pf->reg_base +
604 					      RVU_PF_VF_BAR4_ADDR));
605 
606 	hwbase = ioremap_wc(base, MBOX_SIZE * pf->total_vfs);
607 	if (!hwbase) {
608 		err = -ENOMEM;
609 		goto free_wq;
610 	}
611 
612 	mbox = &pf->mbox_pfvf[0];
613 	err = otx2_mbox_init(&mbox->mbox, hwbase, pf->pdev, pf->reg_base,
614 			     MBOX_DIR_PFVF, numvfs);
615 	if (err)
616 		goto free_iomem;
617 
618 	err = otx2_mbox_init(&mbox->mbox_up, hwbase, pf->pdev, pf->reg_base,
619 			     MBOX_DIR_PFVF_UP, numvfs);
620 	if (err)
621 		goto free_iomem;
622 
623 	for (vf = 0; vf < numvfs; vf++) {
624 		mbox->pfvf = pf;
625 		INIT_WORK(&mbox->mbox_wrk, otx2_pfvf_mbox_handler);
626 		INIT_WORK(&mbox->mbox_up_wrk, otx2_pfvf_mbox_up_handler);
627 		mbox++;
628 	}
629 
630 	return 0;
631 
632 free_iomem:
633 	if (hwbase)
634 		iounmap(hwbase);
635 free_wq:
636 	destroy_workqueue(pf->mbox_pfvf_wq);
637 	return err;
638 }
639 
640 static void otx2_pfvf_mbox_destroy(struct otx2_nic *pf)
641 {
642 	struct mbox *mbox = &pf->mbox_pfvf[0];
643 
644 	if (!mbox)
645 		return;
646 
647 	if (pf->mbox_pfvf_wq) {
648 		destroy_workqueue(pf->mbox_pfvf_wq);
649 		pf->mbox_pfvf_wq = NULL;
650 	}
651 
652 	if (mbox->mbox.hwbase)
653 		iounmap(mbox->mbox.hwbase);
654 
655 	otx2_mbox_destroy(&mbox->mbox);
656 }
657 
658 static void otx2_enable_pfvf_mbox_intr(struct otx2_nic *pf, int numvfs)
659 {
660 	/* Clear PF <=> VF mailbox IRQ */
661 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(0), ~0ull);
662 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(1), ~0ull);
663 
664 	/* Enable PF <=> VF mailbox IRQ */
665 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INT_ENA_W1SX(0), INTR_MASK(numvfs));
666 	if (numvfs > 64) {
667 		numvfs -= 64;
668 		otx2_write64(pf, RVU_PF_VFPF_MBOX_INT_ENA_W1SX(1),
669 			     INTR_MASK(numvfs));
670 	}
671 }
672 
673 static void otx2_disable_pfvf_mbox_intr(struct otx2_nic *pf, int numvfs)
674 {
675 	int vector;
676 
677 	/* Disable PF <=> VF mailbox IRQ */
678 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INT_ENA_W1CX(0), ~0ull);
679 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INT_ENA_W1CX(1), ~0ull);
680 
681 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(0), ~0ull);
682 	vector = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFPF_MBOX0);
683 	free_irq(vector, pf);
684 
685 	if (numvfs > 64) {
686 		otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(1), ~0ull);
687 		vector = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFPF_MBOX1);
688 		free_irq(vector, pf);
689 	}
690 }
691 
692 static int otx2_register_pfvf_mbox_intr(struct otx2_nic *pf, int numvfs)
693 {
694 	struct otx2_hw *hw = &pf->hw;
695 	char *irq_name;
696 	int err;
697 
698 	/* Register MBOX0 interrupt handler */
699 	irq_name = &hw->irq_name[RVU_PF_INT_VEC_VFPF_MBOX0 * NAME_SIZE];
700 	if (pf->pcifunc)
701 		snprintf(irq_name, NAME_SIZE,
702 			 "RVUPF%d_VF Mbox0", rvu_get_pf(pf->pcifunc));
703 	else
704 		snprintf(irq_name, NAME_SIZE, "RVUPF_VF Mbox0");
705 	err = request_irq(pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFPF_MBOX0),
706 			  otx2_pfvf_mbox_intr_handler, 0, irq_name, pf);
707 	if (err) {
708 		dev_err(pf->dev,
709 			"RVUPF: IRQ registration failed for PFVF mbox0 irq\n");
710 		return err;
711 	}
712 
713 	if (numvfs > 64) {
714 		/* Register MBOX1 interrupt handler */
715 		irq_name = &hw->irq_name[RVU_PF_INT_VEC_VFPF_MBOX1 * NAME_SIZE];
716 		if (pf->pcifunc)
717 			snprintf(irq_name, NAME_SIZE,
718 				 "RVUPF%d_VF Mbox1", rvu_get_pf(pf->pcifunc));
719 		else
720 			snprintf(irq_name, NAME_SIZE, "RVUPF_VF Mbox1");
721 		err = request_irq(pci_irq_vector(pf->pdev,
722 						 RVU_PF_INT_VEC_VFPF_MBOX1),
723 						 otx2_pfvf_mbox_intr_handler,
724 						 0, irq_name, pf);
725 		if (err) {
726 			dev_err(pf->dev,
727 				"RVUPF: IRQ registration failed for PFVF mbox1 irq\n");
728 			return err;
729 		}
730 	}
731 
732 	otx2_enable_pfvf_mbox_intr(pf, numvfs);
733 
734 	return 0;
735 }
736 
737 static void otx2_process_pfaf_mbox_msg(struct otx2_nic *pf,
738 				       struct mbox_msghdr *msg)
739 {
740 	int devid;
741 
742 	if (msg->id >= MBOX_MSG_MAX) {
743 		dev_err(pf->dev,
744 			"Mbox msg with unknown ID 0x%x\n", msg->id);
745 		return;
746 	}
747 
748 	if (msg->sig != OTX2_MBOX_RSP_SIG) {
749 		dev_err(pf->dev,
750 			"Mbox msg with wrong signature %x, ID 0x%x\n",
751 			 msg->sig, msg->id);
752 		return;
753 	}
754 
755 	/* message response heading VF */
756 	devid = msg->pcifunc & RVU_PFVF_FUNC_MASK;
757 	if (devid) {
758 		struct otx2_vf_config *config = &pf->vf_configs[devid - 1];
759 		struct delayed_work *dwork;
760 
761 		switch (msg->id) {
762 		case MBOX_MSG_NIX_LF_START_RX:
763 			config->intf_down = false;
764 			dwork = &config->link_event_work;
765 			schedule_delayed_work(dwork, msecs_to_jiffies(100));
766 			break;
767 		case MBOX_MSG_NIX_LF_STOP_RX:
768 			config->intf_down = true;
769 			break;
770 		}
771 
772 		return;
773 	}
774 
775 	switch (msg->id) {
776 	case MBOX_MSG_READY:
777 		pf->pcifunc = msg->pcifunc;
778 		break;
779 	case MBOX_MSG_MSIX_OFFSET:
780 		mbox_handler_msix_offset(pf, (struct msix_offset_rsp *)msg);
781 		break;
782 	case MBOX_MSG_NPA_LF_ALLOC:
783 		mbox_handler_npa_lf_alloc(pf, (struct npa_lf_alloc_rsp *)msg);
784 		break;
785 	case MBOX_MSG_NIX_LF_ALLOC:
786 		mbox_handler_nix_lf_alloc(pf, (struct nix_lf_alloc_rsp *)msg);
787 		break;
788 	case MBOX_MSG_NIX_TXSCH_ALLOC:
789 		mbox_handler_nix_txsch_alloc(pf,
790 					     (struct nix_txsch_alloc_rsp *)msg);
791 		break;
792 	case MBOX_MSG_NIX_BP_ENABLE:
793 		mbox_handler_nix_bp_enable(pf, (struct nix_bp_cfg_rsp *)msg);
794 		break;
795 	case MBOX_MSG_CGX_STATS:
796 		mbox_handler_cgx_stats(pf, (struct cgx_stats_rsp *)msg);
797 		break;
798 	case MBOX_MSG_CGX_FEC_STATS:
799 		mbox_handler_cgx_fec_stats(pf, (struct cgx_fec_stats_rsp *)msg);
800 		break;
801 	default:
802 		if (msg->rc)
803 			dev_err(pf->dev,
804 				"Mbox msg response has err %d, ID 0x%x\n",
805 				msg->rc, msg->id);
806 		break;
807 	}
808 }
809 
810 static void otx2_pfaf_mbox_handler(struct work_struct *work)
811 {
812 	struct otx2_mbox_dev *mdev;
813 	struct mbox_hdr *rsp_hdr;
814 	struct mbox_msghdr *msg;
815 	struct otx2_mbox *mbox;
816 	struct mbox *af_mbox;
817 	struct otx2_nic *pf;
818 	int offset, id;
819 
820 	af_mbox = container_of(work, struct mbox, mbox_wrk);
821 	mbox = &af_mbox->mbox;
822 	mdev = &mbox->dev[0];
823 	rsp_hdr = (struct mbox_hdr *)(mdev->mbase + mbox->rx_start);
824 
825 	offset = mbox->rx_start + ALIGN(sizeof(*rsp_hdr), MBOX_MSG_ALIGN);
826 	pf = af_mbox->pfvf;
827 
828 	for (id = 0; id < af_mbox->num_msgs; id++) {
829 		msg = (struct mbox_msghdr *)(mdev->mbase + offset);
830 		otx2_process_pfaf_mbox_msg(pf, msg);
831 		offset = mbox->rx_start + msg->next_msgoff;
832 		if (mdev->msgs_acked == (af_mbox->num_msgs - 1))
833 			__otx2_mbox_reset(mbox, 0);
834 		mdev->msgs_acked++;
835 	}
836 
837 }
838 
839 static void otx2_handle_link_event(struct otx2_nic *pf)
840 {
841 	struct cgx_link_user_info *linfo = &pf->linfo;
842 	struct net_device *netdev = pf->netdev;
843 
844 	pr_info("%s NIC Link is %s %d Mbps %s duplex\n", netdev->name,
845 		linfo->link_up ? "UP" : "DOWN", linfo->speed,
846 		linfo->full_duplex ? "Full" : "Half");
847 	if (linfo->link_up) {
848 		netif_carrier_on(netdev);
849 		netif_tx_start_all_queues(netdev);
850 	} else {
851 		netif_tx_stop_all_queues(netdev);
852 		netif_carrier_off(netdev);
853 	}
854 }
855 
856 int otx2_mbox_up_handler_cgx_link_event(struct otx2_nic *pf,
857 					struct cgx_link_info_msg *msg,
858 					struct msg_rsp *rsp)
859 {
860 	int i;
861 
862 	/* Copy the link info sent by AF */
863 	pf->linfo = msg->link_info;
864 
865 	/* notify VFs about link event */
866 	for (i = 0; i < pci_num_vf(pf->pdev); i++) {
867 		struct otx2_vf_config *config = &pf->vf_configs[i];
868 		struct delayed_work *dwork = &config->link_event_work;
869 
870 		if (config->intf_down)
871 			continue;
872 
873 		schedule_delayed_work(dwork, msecs_to_jiffies(100));
874 	}
875 
876 	/* interface has not been fully configured yet */
877 	if (pf->flags & OTX2_FLAG_INTF_DOWN)
878 		return 0;
879 
880 	otx2_handle_link_event(pf);
881 	return 0;
882 }
883 
884 static int otx2_process_mbox_msg_up(struct otx2_nic *pf,
885 				    struct mbox_msghdr *req)
886 {
887 	/* Check if valid, if not reply with a invalid msg */
888 	if (req->sig != OTX2_MBOX_REQ_SIG) {
889 		otx2_reply_invalid_msg(&pf->mbox.mbox_up, 0, 0, req->id);
890 		return -ENODEV;
891 	}
892 
893 	switch (req->id) {
894 #define M(_name, _id, _fn_name, _req_type, _rsp_type)			\
895 	case _id: {							\
896 		struct _rsp_type *rsp;					\
897 		int err;						\
898 									\
899 		rsp = (struct _rsp_type *)otx2_mbox_alloc_msg(		\
900 			&pf->mbox.mbox_up, 0,				\
901 			sizeof(struct _rsp_type));			\
902 		if (!rsp)						\
903 			return -ENOMEM;					\
904 									\
905 		rsp->hdr.id = _id;					\
906 		rsp->hdr.sig = OTX2_MBOX_RSP_SIG;			\
907 		rsp->hdr.pcifunc = 0;					\
908 		rsp->hdr.rc = 0;					\
909 									\
910 		err = otx2_mbox_up_handler_ ## _fn_name(		\
911 			pf, (struct _req_type *)req, rsp);		\
912 		return err;						\
913 	}
914 MBOX_UP_CGX_MESSAGES
915 #undef M
916 		break;
917 	default:
918 		otx2_reply_invalid_msg(&pf->mbox.mbox_up, 0, 0, req->id);
919 		return -ENODEV;
920 	}
921 	return 0;
922 }
923 
924 static void otx2_pfaf_mbox_up_handler(struct work_struct *work)
925 {
926 	struct mbox *af_mbox = container_of(work, struct mbox, mbox_up_wrk);
927 	struct otx2_mbox *mbox = &af_mbox->mbox_up;
928 	struct otx2_mbox_dev *mdev = &mbox->dev[0];
929 	struct otx2_nic *pf = af_mbox->pfvf;
930 	int offset, id, devid = 0;
931 	struct mbox_hdr *rsp_hdr;
932 	struct mbox_msghdr *msg;
933 
934 	rsp_hdr = (struct mbox_hdr *)(mdev->mbase + mbox->rx_start);
935 
936 	offset = mbox->rx_start + ALIGN(sizeof(*rsp_hdr), MBOX_MSG_ALIGN);
937 
938 	for (id = 0; id < af_mbox->up_num_msgs; id++) {
939 		msg = (struct mbox_msghdr *)(mdev->mbase + offset);
940 
941 		devid = msg->pcifunc & RVU_PFVF_FUNC_MASK;
942 		/* Skip processing VF's messages */
943 		if (!devid)
944 			otx2_process_mbox_msg_up(pf, msg);
945 		offset = mbox->rx_start + msg->next_msgoff;
946 	}
947 	if (devid) {
948 		otx2_forward_vf_mbox_msgs(pf, &pf->mbox.mbox_up,
949 					  MBOX_DIR_PFVF_UP, devid - 1,
950 					  af_mbox->up_num_msgs);
951 		return;
952 	}
953 
954 	otx2_mbox_msg_send(mbox, 0);
955 }
956 
957 static irqreturn_t otx2_pfaf_mbox_intr_handler(int irq, void *pf_irq)
958 {
959 	struct otx2_nic *pf = (struct otx2_nic *)pf_irq;
960 	struct mbox *mbox;
961 
962 	/* Clear the IRQ */
963 	otx2_write64(pf, RVU_PF_INT, BIT_ULL(0));
964 
965 	mbox = &pf->mbox;
966 
967 	trace_otx2_msg_interrupt(mbox->mbox.pdev, "AF to PF", BIT_ULL(0));
968 
969 	otx2_queue_work(mbox, pf->mbox_wq, 0, 1, 1, TYPE_PFAF);
970 
971 	return IRQ_HANDLED;
972 }
973 
974 static void otx2_disable_mbox_intr(struct otx2_nic *pf)
975 {
976 	int vector = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_AFPF_MBOX);
977 
978 	/* Disable AF => PF mailbox IRQ */
979 	otx2_write64(pf, RVU_PF_INT_ENA_W1C, BIT_ULL(0));
980 	free_irq(vector, pf);
981 }
982 
983 static int otx2_register_mbox_intr(struct otx2_nic *pf, bool probe_af)
984 {
985 	struct otx2_hw *hw = &pf->hw;
986 	struct msg_req *req;
987 	char *irq_name;
988 	int err;
989 
990 	/* Register mailbox interrupt handler */
991 	irq_name = &hw->irq_name[RVU_PF_INT_VEC_AFPF_MBOX * NAME_SIZE];
992 	snprintf(irq_name, NAME_SIZE, "RVUPFAF Mbox");
993 	err = request_irq(pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_AFPF_MBOX),
994 			  otx2_pfaf_mbox_intr_handler, 0, irq_name, pf);
995 	if (err) {
996 		dev_err(pf->dev,
997 			"RVUPF: IRQ registration failed for PFAF mbox irq\n");
998 		return err;
999 	}
1000 
1001 	/* Enable mailbox interrupt for msgs coming from AF.
1002 	 * First clear to avoid spurious interrupts, if any.
1003 	 */
1004 	otx2_write64(pf, RVU_PF_INT, BIT_ULL(0));
1005 	otx2_write64(pf, RVU_PF_INT_ENA_W1S, BIT_ULL(0));
1006 
1007 	if (!probe_af)
1008 		return 0;
1009 
1010 	/* Check mailbox communication with AF */
1011 	req = otx2_mbox_alloc_msg_ready(&pf->mbox);
1012 	if (!req) {
1013 		otx2_disable_mbox_intr(pf);
1014 		return -ENOMEM;
1015 	}
1016 	err = otx2_sync_mbox_msg(&pf->mbox);
1017 	if (err) {
1018 		dev_warn(pf->dev,
1019 			 "AF not responding to mailbox, deferring probe\n");
1020 		otx2_disable_mbox_intr(pf);
1021 		return -EPROBE_DEFER;
1022 	}
1023 
1024 	return 0;
1025 }
1026 
1027 static void otx2_pfaf_mbox_destroy(struct otx2_nic *pf)
1028 {
1029 	struct mbox *mbox = &pf->mbox;
1030 
1031 	if (pf->mbox_wq) {
1032 		destroy_workqueue(pf->mbox_wq);
1033 		pf->mbox_wq = NULL;
1034 	}
1035 
1036 	if (mbox->mbox.hwbase)
1037 		iounmap((void __iomem *)mbox->mbox.hwbase);
1038 
1039 	otx2_mbox_destroy(&mbox->mbox);
1040 	otx2_mbox_destroy(&mbox->mbox_up);
1041 }
1042 
1043 static int otx2_pfaf_mbox_init(struct otx2_nic *pf)
1044 {
1045 	struct mbox *mbox = &pf->mbox;
1046 	void __iomem *hwbase;
1047 	int err;
1048 
1049 	mbox->pfvf = pf;
1050 	pf->mbox_wq = alloc_workqueue("otx2_pfaf_mailbox",
1051 				      WQ_UNBOUND | WQ_HIGHPRI |
1052 				      WQ_MEM_RECLAIM, 1);
1053 	if (!pf->mbox_wq)
1054 		return -ENOMEM;
1055 
1056 	/* Mailbox is a reserved memory (in RAM) region shared between
1057 	 * admin function (i.e AF) and this PF, shouldn't be mapped as
1058 	 * device memory to allow unaligned accesses.
1059 	 */
1060 	hwbase = ioremap_wc(pci_resource_start(pf->pdev, PCI_MBOX_BAR_NUM),
1061 			    MBOX_SIZE);
1062 	if (!hwbase) {
1063 		dev_err(pf->dev, "Unable to map PFAF mailbox region\n");
1064 		err = -ENOMEM;
1065 		goto exit;
1066 	}
1067 
1068 	err = otx2_mbox_init(&mbox->mbox, hwbase, pf->pdev, pf->reg_base,
1069 			     MBOX_DIR_PFAF, 1);
1070 	if (err)
1071 		goto exit;
1072 
1073 	err = otx2_mbox_init(&mbox->mbox_up, hwbase, pf->pdev, pf->reg_base,
1074 			     MBOX_DIR_PFAF_UP, 1);
1075 	if (err)
1076 		goto exit;
1077 
1078 	err = otx2_mbox_bbuf_init(mbox, pf->pdev);
1079 	if (err)
1080 		goto exit;
1081 
1082 	INIT_WORK(&mbox->mbox_wrk, otx2_pfaf_mbox_handler);
1083 	INIT_WORK(&mbox->mbox_up_wrk, otx2_pfaf_mbox_up_handler);
1084 	mutex_init(&mbox->lock);
1085 
1086 	return 0;
1087 exit:
1088 	otx2_pfaf_mbox_destroy(pf);
1089 	return err;
1090 }
1091 
1092 static int otx2_cgx_config_linkevents(struct otx2_nic *pf, bool enable)
1093 {
1094 	struct msg_req *msg;
1095 	int err;
1096 
1097 	mutex_lock(&pf->mbox.lock);
1098 	if (enable)
1099 		msg = otx2_mbox_alloc_msg_cgx_start_linkevents(&pf->mbox);
1100 	else
1101 		msg = otx2_mbox_alloc_msg_cgx_stop_linkevents(&pf->mbox);
1102 
1103 	if (!msg) {
1104 		mutex_unlock(&pf->mbox.lock);
1105 		return -ENOMEM;
1106 	}
1107 
1108 	err = otx2_sync_mbox_msg(&pf->mbox);
1109 	mutex_unlock(&pf->mbox.lock);
1110 	return err;
1111 }
1112 
1113 static int otx2_cgx_config_loopback(struct otx2_nic *pf, bool enable)
1114 {
1115 	struct msg_req *msg;
1116 	int err;
1117 
1118 	if (enable && bitmap_weight(&pf->flow_cfg->dmacflt_bmap,
1119 				    pf->flow_cfg->dmacflt_max_flows))
1120 		netdev_warn(pf->netdev,
1121 			    "CGX/RPM internal loopback might not work as DMAC filters are active\n");
1122 
1123 	mutex_lock(&pf->mbox.lock);
1124 	if (enable)
1125 		msg = otx2_mbox_alloc_msg_cgx_intlbk_enable(&pf->mbox);
1126 	else
1127 		msg = otx2_mbox_alloc_msg_cgx_intlbk_disable(&pf->mbox);
1128 
1129 	if (!msg) {
1130 		mutex_unlock(&pf->mbox.lock);
1131 		return -ENOMEM;
1132 	}
1133 
1134 	err = otx2_sync_mbox_msg(&pf->mbox);
1135 	mutex_unlock(&pf->mbox.lock);
1136 	return err;
1137 }
1138 
1139 int otx2_set_real_num_queues(struct net_device *netdev,
1140 			     int tx_queues, int rx_queues)
1141 {
1142 	int err;
1143 
1144 	err = netif_set_real_num_tx_queues(netdev, tx_queues);
1145 	if (err) {
1146 		netdev_err(netdev,
1147 			   "Failed to set no of Tx queues: %d\n", tx_queues);
1148 		return err;
1149 	}
1150 
1151 	err = netif_set_real_num_rx_queues(netdev, rx_queues);
1152 	if (err)
1153 		netdev_err(netdev,
1154 			   "Failed to set no of Rx queues: %d\n", rx_queues);
1155 	return err;
1156 }
1157 EXPORT_SYMBOL(otx2_set_real_num_queues);
1158 
1159 static irqreturn_t otx2_q_intr_handler(int irq, void *data)
1160 {
1161 	struct otx2_nic *pf = data;
1162 	u64 val, *ptr;
1163 	u64 qidx = 0;
1164 
1165 	/* CQ */
1166 	for (qidx = 0; qidx < pf->qset.cq_cnt; qidx++) {
1167 		ptr = otx2_get_regaddr(pf, NIX_LF_CQ_OP_INT);
1168 		val = otx2_atomic64_add((qidx << 44), ptr);
1169 
1170 		otx2_write64(pf, NIX_LF_CQ_OP_INT, (qidx << 44) |
1171 			     (val & NIX_CQERRINT_BITS));
1172 		if (!(val & (NIX_CQERRINT_BITS | BIT_ULL(42))))
1173 			continue;
1174 
1175 		if (val & BIT_ULL(42)) {
1176 			netdev_err(pf->netdev, "CQ%lld: error reading NIX_LF_CQ_OP_INT, NIX_LF_ERR_INT 0x%llx\n",
1177 				   qidx, otx2_read64(pf, NIX_LF_ERR_INT));
1178 		} else {
1179 			if (val & BIT_ULL(NIX_CQERRINT_DOOR_ERR))
1180 				netdev_err(pf->netdev, "CQ%lld: Doorbell error",
1181 					   qidx);
1182 			if (val & BIT_ULL(NIX_CQERRINT_CQE_FAULT))
1183 				netdev_err(pf->netdev, "CQ%lld: Memory fault on CQE write to LLC/DRAM",
1184 					   qidx);
1185 		}
1186 
1187 		schedule_work(&pf->reset_task);
1188 	}
1189 
1190 	/* SQ */
1191 	for (qidx = 0; qidx < pf->hw.tot_tx_queues; qidx++) {
1192 		ptr = otx2_get_regaddr(pf, NIX_LF_SQ_OP_INT);
1193 		val = otx2_atomic64_add((qidx << 44), ptr);
1194 		otx2_write64(pf, NIX_LF_SQ_OP_INT, (qidx << 44) |
1195 			     (val & NIX_SQINT_BITS));
1196 
1197 		if (!(val & (NIX_SQINT_BITS | BIT_ULL(42))))
1198 			continue;
1199 
1200 		if (val & BIT_ULL(42)) {
1201 			netdev_err(pf->netdev, "SQ%lld: error reading NIX_LF_SQ_OP_INT, NIX_LF_ERR_INT 0x%llx\n",
1202 				   qidx, otx2_read64(pf, NIX_LF_ERR_INT));
1203 		} else {
1204 			if (val & BIT_ULL(NIX_SQINT_LMT_ERR)) {
1205 				netdev_err(pf->netdev, "SQ%lld: LMT store error NIX_LF_SQ_OP_ERR_DBG:0x%llx",
1206 					   qidx,
1207 					   otx2_read64(pf,
1208 						       NIX_LF_SQ_OP_ERR_DBG));
1209 				otx2_write64(pf, NIX_LF_SQ_OP_ERR_DBG,
1210 					     BIT_ULL(44));
1211 			}
1212 			if (val & BIT_ULL(NIX_SQINT_MNQ_ERR)) {
1213 				netdev_err(pf->netdev, "SQ%lld: Meta-descriptor enqueue error NIX_LF_MNQ_ERR_DGB:0x%llx\n",
1214 					   qidx,
1215 					   otx2_read64(pf, NIX_LF_MNQ_ERR_DBG));
1216 				otx2_write64(pf, NIX_LF_MNQ_ERR_DBG,
1217 					     BIT_ULL(44));
1218 			}
1219 			if (val & BIT_ULL(NIX_SQINT_SEND_ERR)) {
1220 				netdev_err(pf->netdev, "SQ%lld: Send error, NIX_LF_SEND_ERR_DBG 0x%llx",
1221 					   qidx,
1222 					   otx2_read64(pf,
1223 						       NIX_LF_SEND_ERR_DBG));
1224 				otx2_write64(pf, NIX_LF_SEND_ERR_DBG,
1225 					     BIT_ULL(44));
1226 			}
1227 			if (val & BIT_ULL(NIX_SQINT_SQB_ALLOC_FAIL))
1228 				netdev_err(pf->netdev, "SQ%lld: SQB allocation failed",
1229 					   qidx);
1230 		}
1231 
1232 		schedule_work(&pf->reset_task);
1233 	}
1234 
1235 	return IRQ_HANDLED;
1236 }
1237 
1238 static irqreturn_t otx2_cq_intr_handler(int irq, void *cq_irq)
1239 {
1240 	struct otx2_cq_poll *cq_poll = (struct otx2_cq_poll *)cq_irq;
1241 	struct otx2_nic *pf = (struct otx2_nic *)cq_poll->dev;
1242 	int qidx = cq_poll->cint_idx;
1243 
1244 	/* Disable interrupts.
1245 	 *
1246 	 * Completion interrupts behave in a level-triggered interrupt
1247 	 * fashion, and hence have to be cleared only after it is serviced.
1248 	 */
1249 	otx2_write64(pf, NIX_LF_CINTX_ENA_W1C(qidx), BIT_ULL(0));
1250 
1251 	/* Schedule NAPI */
1252 	napi_schedule_irqoff(&cq_poll->napi);
1253 
1254 	return IRQ_HANDLED;
1255 }
1256 
1257 static void otx2_disable_napi(struct otx2_nic *pf)
1258 {
1259 	struct otx2_qset *qset = &pf->qset;
1260 	struct otx2_cq_poll *cq_poll;
1261 	int qidx;
1262 
1263 	for (qidx = 0; qidx < pf->hw.cint_cnt; qidx++) {
1264 		cq_poll = &qset->napi[qidx];
1265 		napi_disable(&cq_poll->napi);
1266 		netif_napi_del(&cq_poll->napi);
1267 	}
1268 }
1269 
1270 static void otx2_free_cq_res(struct otx2_nic *pf)
1271 {
1272 	struct otx2_qset *qset = &pf->qset;
1273 	struct otx2_cq_queue *cq;
1274 	int qidx;
1275 
1276 	/* Disable CQs */
1277 	otx2_ctx_disable(&pf->mbox, NIX_AQ_CTYPE_CQ, false);
1278 	for (qidx = 0; qidx < qset->cq_cnt; qidx++) {
1279 		cq = &qset->cq[qidx];
1280 		qmem_free(pf->dev, cq->cqe);
1281 	}
1282 }
1283 
1284 static void otx2_free_sq_res(struct otx2_nic *pf)
1285 {
1286 	struct otx2_qset *qset = &pf->qset;
1287 	struct otx2_snd_queue *sq;
1288 	int qidx;
1289 
1290 	/* Disable SQs */
1291 	otx2_ctx_disable(&pf->mbox, NIX_AQ_CTYPE_SQ, false);
1292 	/* Free SQB pointers */
1293 	otx2_sq_free_sqbs(pf);
1294 	for (qidx = 0; qidx < pf->hw.tot_tx_queues; qidx++) {
1295 		sq = &qset->sq[qidx];
1296 		qmem_free(pf->dev, sq->sqe);
1297 		qmem_free(pf->dev, sq->tso_hdrs);
1298 		kfree(sq->sg);
1299 		kfree(sq->sqb_ptrs);
1300 	}
1301 }
1302 
1303 static int otx2_get_rbuf_size(struct otx2_nic *pf, int mtu)
1304 {
1305 	int frame_size;
1306 	int total_size;
1307 	int rbuf_size;
1308 
1309 	/* The data transferred by NIX to memory consists of actual packet
1310 	 * plus additional data which has timestamp and/or EDSA/HIGIG2
1311 	 * headers if interface is configured in corresponding modes.
1312 	 * NIX transfers entire data using 6 segments/buffers and writes
1313 	 * a CQE_RX descriptor with those segment addresses. First segment
1314 	 * has additional data prepended to packet. Also software omits a
1315 	 * headroom of 128 bytes and sizeof(struct skb_shared_info) in
1316 	 * each segment. Hence the total size of memory needed
1317 	 * to receive a packet with 'mtu' is:
1318 	 * frame size =  mtu + additional data;
1319 	 * memory = frame_size + (headroom + struct skb_shared_info size) * 6;
1320 	 * each receive buffer size = memory / 6;
1321 	 */
1322 	frame_size = mtu + OTX2_ETH_HLEN + OTX2_HW_TIMESTAMP_LEN;
1323 	total_size = frame_size + (OTX2_HEAD_ROOM +
1324 		     OTX2_DATA_ALIGN(sizeof(struct skb_shared_info))) * 6;
1325 	rbuf_size = total_size / 6;
1326 
1327 	return ALIGN(rbuf_size, 2048);
1328 }
1329 
1330 static int otx2_init_hw_resources(struct otx2_nic *pf)
1331 {
1332 	struct nix_lf_free_req *free_req;
1333 	struct mbox *mbox = &pf->mbox;
1334 	struct otx2_hw *hw = &pf->hw;
1335 	struct msg_req *req;
1336 	int err = 0, lvl;
1337 
1338 	/* Set required NPA LF's pool counts
1339 	 * Auras and Pools are used in a 1:1 mapping,
1340 	 * so, aura count = pool count.
1341 	 */
1342 	hw->rqpool_cnt = hw->rx_queues;
1343 	hw->sqpool_cnt = hw->tot_tx_queues;
1344 	hw->pool_cnt = hw->rqpool_cnt + hw->sqpool_cnt;
1345 
1346 	pf->max_frs = pf->netdev->mtu + OTX2_ETH_HLEN + OTX2_HW_TIMESTAMP_LEN;
1347 
1348 	pf->rbsize = otx2_get_rbuf_size(pf, pf->netdev->mtu);
1349 
1350 	mutex_lock(&mbox->lock);
1351 	/* NPA init */
1352 	err = otx2_config_npa(pf);
1353 	if (err)
1354 		goto exit;
1355 
1356 	/* NIX init */
1357 	err = otx2_config_nix(pf);
1358 	if (err)
1359 		goto err_free_npa_lf;
1360 
1361 	/* Enable backpressure */
1362 	otx2_nix_config_bp(pf, true);
1363 
1364 	/* Init Auras and pools used by NIX RQ, for free buffer ptrs */
1365 	err = otx2_rq_aura_pool_init(pf);
1366 	if (err) {
1367 		mutex_unlock(&mbox->lock);
1368 		goto err_free_nix_lf;
1369 	}
1370 	/* Init Auras and pools used by NIX SQ, for queueing SQEs */
1371 	err = otx2_sq_aura_pool_init(pf);
1372 	if (err) {
1373 		mutex_unlock(&mbox->lock);
1374 		goto err_free_rq_ptrs;
1375 	}
1376 
1377 	err = otx2_txsch_alloc(pf);
1378 	if (err) {
1379 		mutex_unlock(&mbox->lock);
1380 		goto err_free_sq_ptrs;
1381 	}
1382 
1383 	err = otx2_config_nix_queues(pf);
1384 	if (err) {
1385 		mutex_unlock(&mbox->lock);
1386 		goto err_free_txsch;
1387 	}
1388 	for (lvl = 0; lvl < NIX_TXSCH_LVL_CNT; lvl++) {
1389 		err = otx2_txschq_config(pf, lvl);
1390 		if (err) {
1391 			mutex_unlock(&mbox->lock);
1392 			goto err_free_nix_queues;
1393 		}
1394 	}
1395 	mutex_unlock(&mbox->lock);
1396 	return err;
1397 
1398 err_free_nix_queues:
1399 	otx2_free_sq_res(pf);
1400 	otx2_free_cq_res(pf);
1401 	otx2_ctx_disable(mbox, NIX_AQ_CTYPE_RQ, false);
1402 err_free_txsch:
1403 	if (otx2_txschq_stop(pf))
1404 		dev_err(pf->dev, "%s failed to stop TX schedulers\n", __func__);
1405 err_free_sq_ptrs:
1406 	otx2_sq_free_sqbs(pf);
1407 err_free_rq_ptrs:
1408 	otx2_free_aura_ptr(pf, AURA_NIX_RQ);
1409 	otx2_ctx_disable(mbox, NPA_AQ_CTYPE_POOL, true);
1410 	otx2_ctx_disable(mbox, NPA_AQ_CTYPE_AURA, true);
1411 	otx2_aura_pool_free(pf);
1412 err_free_nix_lf:
1413 	mutex_lock(&mbox->lock);
1414 	free_req = otx2_mbox_alloc_msg_nix_lf_free(mbox);
1415 	if (free_req) {
1416 		free_req->flags = NIX_LF_DISABLE_FLOWS;
1417 		if (otx2_sync_mbox_msg(mbox))
1418 			dev_err(pf->dev, "%s failed to free nixlf\n", __func__);
1419 	}
1420 err_free_npa_lf:
1421 	/* Reset NPA LF */
1422 	req = otx2_mbox_alloc_msg_npa_lf_free(mbox);
1423 	if (req) {
1424 		if (otx2_sync_mbox_msg(mbox))
1425 			dev_err(pf->dev, "%s failed to free npalf\n", __func__);
1426 	}
1427 exit:
1428 	mutex_unlock(&mbox->lock);
1429 	return err;
1430 }
1431 
1432 static void otx2_free_hw_resources(struct otx2_nic *pf)
1433 {
1434 	struct otx2_qset *qset = &pf->qset;
1435 	struct nix_lf_free_req *free_req;
1436 	struct mbox *mbox = &pf->mbox;
1437 	struct otx2_cq_queue *cq;
1438 	struct msg_req *req;
1439 	int qidx, err;
1440 
1441 	/* Ensure all SQE are processed */
1442 	otx2_sqb_flush(pf);
1443 
1444 	/* Stop transmission */
1445 	err = otx2_txschq_stop(pf);
1446 	if (err)
1447 		dev_err(pf->dev, "RVUPF: Failed to stop/free TX schedulers\n");
1448 
1449 	mutex_lock(&mbox->lock);
1450 	/* Disable backpressure */
1451 	if (!(pf->pcifunc & RVU_PFVF_FUNC_MASK))
1452 		otx2_nix_config_bp(pf, false);
1453 	mutex_unlock(&mbox->lock);
1454 
1455 	/* Disable RQs */
1456 	otx2_ctx_disable(mbox, NIX_AQ_CTYPE_RQ, false);
1457 
1458 	/*Dequeue all CQEs */
1459 	for (qidx = 0; qidx < qset->cq_cnt; qidx++) {
1460 		cq = &qset->cq[qidx];
1461 		if (cq->cq_type == CQ_RX)
1462 			otx2_cleanup_rx_cqes(pf, cq);
1463 		else
1464 			otx2_cleanup_tx_cqes(pf, cq);
1465 	}
1466 
1467 	otx2_free_sq_res(pf);
1468 
1469 	/* Free RQ buffer pointers*/
1470 	otx2_free_aura_ptr(pf, AURA_NIX_RQ);
1471 
1472 	otx2_free_cq_res(pf);
1473 
1474 	/* Free all ingress bandwidth profiles allocated */
1475 	cn10k_free_all_ipolicers(pf);
1476 
1477 	mutex_lock(&mbox->lock);
1478 	/* Reset NIX LF */
1479 	free_req = otx2_mbox_alloc_msg_nix_lf_free(mbox);
1480 	if (free_req) {
1481 		free_req->flags = NIX_LF_DISABLE_FLOWS;
1482 		if (!(pf->flags & OTX2_FLAG_PF_SHUTDOWN))
1483 			free_req->flags |= NIX_LF_DONT_FREE_TX_VTAG;
1484 		if (otx2_sync_mbox_msg(mbox))
1485 			dev_err(pf->dev, "%s failed to free nixlf\n", __func__);
1486 	}
1487 	mutex_unlock(&mbox->lock);
1488 
1489 	/* Disable NPA Pool and Aura hw context */
1490 	otx2_ctx_disable(mbox, NPA_AQ_CTYPE_POOL, true);
1491 	otx2_ctx_disable(mbox, NPA_AQ_CTYPE_AURA, true);
1492 	otx2_aura_pool_free(pf);
1493 
1494 	mutex_lock(&mbox->lock);
1495 	/* Reset NPA LF */
1496 	req = otx2_mbox_alloc_msg_npa_lf_free(mbox);
1497 	if (req) {
1498 		if (otx2_sync_mbox_msg(mbox))
1499 			dev_err(pf->dev, "%s failed to free npalf\n", __func__);
1500 	}
1501 	mutex_unlock(&mbox->lock);
1502 }
1503 
1504 static void otx2_do_set_rx_mode(struct otx2_nic *pf)
1505 {
1506 	struct net_device *netdev = pf->netdev;
1507 	struct nix_rx_mode *req;
1508 	bool promisc = false;
1509 
1510 	if (!(netdev->flags & IFF_UP))
1511 		return;
1512 
1513 	if ((netdev->flags & IFF_PROMISC) ||
1514 	    (netdev_uc_count(netdev) > OTX2_MAX_UNICAST_FLOWS)) {
1515 		promisc = true;
1516 	}
1517 
1518 	/* Write unicast address to mcam entries or del from mcam */
1519 	if (!promisc && netdev->priv_flags & IFF_UNICAST_FLT)
1520 		__dev_uc_sync(netdev, otx2_add_macfilter, otx2_del_macfilter);
1521 
1522 	mutex_lock(&pf->mbox.lock);
1523 	req = otx2_mbox_alloc_msg_nix_set_rx_mode(&pf->mbox);
1524 	if (!req) {
1525 		mutex_unlock(&pf->mbox.lock);
1526 		return;
1527 	}
1528 
1529 	req->mode = NIX_RX_MODE_UCAST;
1530 
1531 	if (promisc)
1532 		req->mode |= NIX_RX_MODE_PROMISC;
1533 	if (netdev->flags & (IFF_ALLMULTI | IFF_MULTICAST))
1534 		req->mode |= NIX_RX_MODE_ALLMULTI;
1535 
1536 	req->mode |= NIX_RX_MODE_USE_MCE;
1537 
1538 	otx2_sync_mbox_msg(&pf->mbox);
1539 	mutex_unlock(&pf->mbox.lock);
1540 }
1541 
1542 int otx2_open(struct net_device *netdev)
1543 {
1544 	struct otx2_nic *pf = netdev_priv(netdev);
1545 	struct otx2_cq_poll *cq_poll = NULL;
1546 	struct otx2_qset *qset = &pf->qset;
1547 	int err = 0, qidx, vec;
1548 	char *irq_name;
1549 
1550 	netif_carrier_off(netdev);
1551 
1552 	pf->qset.cq_cnt = pf->hw.rx_queues + pf->hw.tot_tx_queues;
1553 	/* RQ and SQs are mapped to different CQs,
1554 	 * so find out max CQ IRQs (i.e CINTs) needed.
1555 	 */
1556 	pf->hw.cint_cnt = max(pf->hw.rx_queues, pf->hw.tx_queues);
1557 	qset->napi = kcalloc(pf->hw.cint_cnt, sizeof(*cq_poll), GFP_KERNEL);
1558 	if (!qset->napi)
1559 		return -ENOMEM;
1560 
1561 	/* CQ size of RQ */
1562 	qset->rqe_cnt = qset->rqe_cnt ? qset->rqe_cnt : Q_COUNT(Q_SIZE_256);
1563 	/* CQ size of SQ */
1564 	qset->sqe_cnt = qset->sqe_cnt ? qset->sqe_cnt : Q_COUNT(Q_SIZE_4K);
1565 
1566 	err = -ENOMEM;
1567 	qset->cq = kcalloc(pf->qset.cq_cnt,
1568 			   sizeof(struct otx2_cq_queue), GFP_KERNEL);
1569 	if (!qset->cq)
1570 		goto err_free_mem;
1571 
1572 	qset->sq = kcalloc(pf->hw.tot_tx_queues,
1573 			   sizeof(struct otx2_snd_queue), GFP_KERNEL);
1574 	if (!qset->sq)
1575 		goto err_free_mem;
1576 
1577 	qset->rq = kcalloc(pf->hw.rx_queues,
1578 			   sizeof(struct otx2_rcv_queue), GFP_KERNEL);
1579 	if (!qset->rq)
1580 		goto err_free_mem;
1581 
1582 	err = otx2_init_hw_resources(pf);
1583 	if (err)
1584 		goto err_free_mem;
1585 
1586 	/* Register NAPI handler */
1587 	for (qidx = 0; qidx < pf->hw.cint_cnt; qidx++) {
1588 		cq_poll = &qset->napi[qidx];
1589 		cq_poll->cint_idx = qidx;
1590 		/* RQ0 & SQ0 are mapped to CINT0 and so on..
1591 		 * 'cq_ids[0]' points to RQ's CQ and
1592 		 * 'cq_ids[1]' points to SQ's CQ and
1593 		 * 'cq_ids[2]' points to XDP's CQ and
1594 		 */
1595 		cq_poll->cq_ids[CQ_RX] =
1596 			(qidx <  pf->hw.rx_queues) ? qidx : CINT_INVALID_CQ;
1597 		cq_poll->cq_ids[CQ_TX] = (qidx < pf->hw.tx_queues) ?
1598 				      qidx + pf->hw.rx_queues : CINT_INVALID_CQ;
1599 		if (pf->xdp_prog)
1600 			cq_poll->cq_ids[CQ_XDP] = (qidx < pf->hw.xdp_queues) ?
1601 						  (qidx + pf->hw.rx_queues +
1602 						  pf->hw.tx_queues) :
1603 						  CINT_INVALID_CQ;
1604 		else
1605 			cq_poll->cq_ids[CQ_XDP] = CINT_INVALID_CQ;
1606 
1607 		cq_poll->dev = (void *)pf;
1608 		netif_napi_add(netdev, &cq_poll->napi,
1609 			       otx2_napi_handler, NAPI_POLL_WEIGHT);
1610 		napi_enable(&cq_poll->napi);
1611 	}
1612 
1613 	/* Set maximum frame size allowed in HW */
1614 	err = otx2_hw_set_mtu(pf, netdev->mtu);
1615 	if (err)
1616 		goto err_disable_napi;
1617 
1618 	/* Setup segmentation algorithms, if failed, clear offload capability */
1619 	otx2_setup_segmentation(pf);
1620 
1621 	/* Initialize RSS */
1622 	err = otx2_rss_init(pf);
1623 	if (err)
1624 		goto err_disable_napi;
1625 
1626 	/* Register Queue IRQ handlers */
1627 	vec = pf->hw.nix_msixoff + NIX_LF_QINT_VEC_START;
1628 	irq_name = &pf->hw.irq_name[vec * NAME_SIZE];
1629 
1630 	snprintf(irq_name, NAME_SIZE, "%s-qerr", pf->netdev->name);
1631 
1632 	err = request_irq(pci_irq_vector(pf->pdev, vec),
1633 			  otx2_q_intr_handler, 0, irq_name, pf);
1634 	if (err) {
1635 		dev_err(pf->dev,
1636 			"RVUPF%d: IRQ registration failed for QERR\n",
1637 			rvu_get_pf(pf->pcifunc));
1638 		goto err_disable_napi;
1639 	}
1640 
1641 	/* Enable QINT IRQ */
1642 	otx2_write64(pf, NIX_LF_QINTX_ENA_W1S(0), BIT_ULL(0));
1643 
1644 	/* Register CQ IRQ handlers */
1645 	vec = pf->hw.nix_msixoff + NIX_LF_CINT_VEC_START;
1646 	for (qidx = 0; qidx < pf->hw.cint_cnt; qidx++) {
1647 		irq_name = &pf->hw.irq_name[vec * NAME_SIZE];
1648 
1649 		snprintf(irq_name, NAME_SIZE, "%s-rxtx-%d", pf->netdev->name,
1650 			 qidx);
1651 
1652 		err = request_irq(pci_irq_vector(pf->pdev, vec),
1653 				  otx2_cq_intr_handler, 0, irq_name,
1654 				  &qset->napi[qidx]);
1655 		if (err) {
1656 			dev_err(pf->dev,
1657 				"RVUPF%d: IRQ registration failed for CQ%d\n",
1658 				rvu_get_pf(pf->pcifunc), qidx);
1659 			goto err_free_cints;
1660 		}
1661 		vec++;
1662 
1663 		otx2_config_irq_coalescing(pf, qidx);
1664 
1665 		/* Enable CQ IRQ */
1666 		otx2_write64(pf, NIX_LF_CINTX_INT(qidx), BIT_ULL(0));
1667 		otx2_write64(pf, NIX_LF_CINTX_ENA_W1S(qidx), BIT_ULL(0));
1668 	}
1669 
1670 	otx2_set_cints_affinity(pf);
1671 
1672 	if (pf->flags & OTX2_FLAG_RX_VLAN_SUPPORT)
1673 		otx2_enable_rxvlan(pf, true);
1674 
1675 	/* When reinitializing enable time stamping if it is enabled before */
1676 	if (pf->flags & OTX2_FLAG_TX_TSTAMP_ENABLED) {
1677 		pf->flags &= ~OTX2_FLAG_TX_TSTAMP_ENABLED;
1678 		otx2_config_hw_tx_tstamp(pf, true);
1679 	}
1680 	if (pf->flags & OTX2_FLAG_RX_TSTAMP_ENABLED) {
1681 		pf->flags &= ~OTX2_FLAG_RX_TSTAMP_ENABLED;
1682 		otx2_config_hw_rx_tstamp(pf, true);
1683 	}
1684 
1685 	pf->flags &= ~OTX2_FLAG_INTF_DOWN;
1686 	/* 'intf_down' may be checked on any cpu */
1687 	smp_wmb();
1688 
1689 	/* we have already received link status notification */
1690 	if (pf->linfo.link_up && !(pf->pcifunc & RVU_PFVF_FUNC_MASK))
1691 		otx2_handle_link_event(pf);
1692 
1693 	/* Restore pause frame settings */
1694 	otx2_config_pause_frm(pf);
1695 
1696 	/* Install DMAC Filters */
1697 	if (pf->flags & OTX2_FLAG_DMACFLTR_SUPPORT)
1698 		otx2_dmacflt_reinstall_flows(pf);
1699 
1700 	err = otx2_rxtx_enable(pf, true);
1701 	if (err)
1702 		goto err_tx_stop_queues;
1703 
1704 	otx2_do_set_rx_mode(pf);
1705 
1706 	return 0;
1707 
1708 err_tx_stop_queues:
1709 	netif_tx_stop_all_queues(netdev);
1710 	netif_carrier_off(netdev);
1711 	pf->flags |= OTX2_FLAG_INTF_DOWN;
1712 err_free_cints:
1713 	otx2_free_cints(pf, qidx);
1714 	vec = pci_irq_vector(pf->pdev,
1715 			     pf->hw.nix_msixoff + NIX_LF_QINT_VEC_START);
1716 	otx2_write64(pf, NIX_LF_QINTX_ENA_W1C(0), BIT_ULL(0));
1717 	synchronize_irq(vec);
1718 	free_irq(vec, pf);
1719 err_disable_napi:
1720 	otx2_disable_napi(pf);
1721 	otx2_free_hw_resources(pf);
1722 err_free_mem:
1723 	kfree(qset->sq);
1724 	kfree(qset->cq);
1725 	kfree(qset->rq);
1726 	kfree(qset->napi);
1727 	return err;
1728 }
1729 EXPORT_SYMBOL(otx2_open);
1730 
1731 int otx2_stop(struct net_device *netdev)
1732 {
1733 	struct otx2_nic *pf = netdev_priv(netdev);
1734 	struct otx2_cq_poll *cq_poll = NULL;
1735 	struct otx2_qset *qset = &pf->qset;
1736 	struct otx2_rss_info *rss;
1737 	int qidx, vec, wrk;
1738 
1739 	/* If the DOWN flag is set resources are already freed */
1740 	if (pf->flags & OTX2_FLAG_INTF_DOWN)
1741 		return 0;
1742 
1743 	netif_carrier_off(netdev);
1744 	netif_tx_stop_all_queues(netdev);
1745 
1746 	pf->flags |= OTX2_FLAG_INTF_DOWN;
1747 	/* 'intf_down' may be checked on any cpu */
1748 	smp_wmb();
1749 
1750 	/* First stop packet Rx/Tx */
1751 	otx2_rxtx_enable(pf, false);
1752 
1753 	/* Clear RSS enable flag */
1754 	rss = &pf->hw.rss_info;
1755 	rss->enable = false;
1756 
1757 	/* Cleanup Queue IRQ */
1758 	vec = pci_irq_vector(pf->pdev,
1759 			     pf->hw.nix_msixoff + NIX_LF_QINT_VEC_START);
1760 	otx2_write64(pf, NIX_LF_QINTX_ENA_W1C(0), BIT_ULL(0));
1761 	synchronize_irq(vec);
1762 	free_irq(vec, pf);
1763 
1764 	/* Cleanup CQ NAPI and IRQ */
1765 	vec = pf->hw.nix_msixoff + NIX_LF_CINT_VEC_START;
1766 	for (qidx = 0; qidx < pf->hw.cint_cnt; qidx++) {
1767 		/* Disable interrupt */
1768 		otx2_write64(pf, NIX_LF_CINTX_ENA_W1C(qidx), BIT_ULL(0));
1769 
1770 		synchronize_irq(pci_irq_vector(pf->pdev, vec));
1771 
1772 		cq_poll = &qset->napi[qidx];
1773 		napi_synchronize(&cq_poll->napi);
1774 		vec++;
1775 	}
1776 
1777 	netif_tx_disable(netdev);
1778 
1779 	otx2_free_hw_resources(pf);
1780 	otx2_free_cints(pf, pf->hw.cint_cnt);
1781 	otx2_disable_napi(pf);
1782 
1783 	for (qidx = 0; qidx < netdev->num_tx_queues; qidx++)
1784 		netdev_tx_reset_queue(netdev_get_tx_queue(netdev, qidx));
1785 
1786 	for (wrk = 0; wrk < pf->qset.cq_cnt; wrk++)
1787 		cancel_delayed_work_sync(&pf->refill_wrk[wrk].pool_refill_work);
1788 	devm_kfree(pf->dev, pf->refill_wrk);
1789 
1790 	kfree(qset->sq);
1791 	kfree(qset->cq);
1792 	kfree(qset->rq);
1793 	kfree(qset->napi);
1794 	/* Do not clear RQ/SQ ringsize settings */
1795 	memset((void *)qset + offsetof(struct otx2_qset, sqe_cnt), 0,
1796 	       sizeof(*qset) - offsetof(struct otx2_qset, sqe_cnt));
1797 	return 0;
1798 }
1799 EXPORT_SYMBOL(otx2_stop);
1800 
1801 static netdev_tx_t otx2_xmit(struct sk_buff *skb, struct net_device *netdev)
1802 {
1803 	struct otx2_nic *pf = netdev_priv(netdev);
1804 	int qidx = skb_get_queue_mapping(skb);
1805 	struct otx2_snd_queue *sq;
1806 	struct netdev_queue *txq;
1807 
1808 	/* Check for minimum and maximum packet length */
1809 	if (skb->len <= ETH_HLEN ||
1810 	    (!skb_shinfo(skb)->gso_size && skb->len > pf->max_frs)) {
1811 		dev_kfree_skb(skb);
1812 		return NETDEV_TX_OK;
1813 	}
1814 
1815 	sq = &pf->qset.sq[qidx];
1816 	txq = netdev_get_tx_queue(netdev, qidx);
1817 
1818 	if (!otx2_sq_append_skb(netdev, sq, skb, qidx)) {
1819 		netif_tx_stop_queue(txq);
1820 
1821 		/* Check again, incase SQBs got freed up */
1822 		smp_mb();
1823 		if (((sq->num_sqbs - *sq->aura_fc_addr) * sq->sqe_per_sqb)
1824 							> sq->sqe_thresh)
1825 			netif_tx_wake_queue(txq);
1826 
1827 		return NETDEV_TX_BUSY;
1828 	}
1829 
1830 	return NETDEV_TX_OK;
1831 }
1832 
1833 static netdev_features_t otx2_fix_features(struct net_device *dev,
1834 					   netdev_features_t features)
1835 {
1836 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
1837 		features |= NETIF_F_HW_VLAN_STAG_RX;
1838 	else
1839 		features &= ~NETIF_F_HW_VLAN_STAG_RX;
1840 
1841 	return features;
1842 }
1843 
1844 static void otx2_set_rx_mode(struct net_device *netdev)
1845 {
1846 	struct otx2_nic *pf = netdev_priv(netdev);
1847 
1848 	queue_work(pf->otx2_wq, &pf->rx_mode_work);
1849 }
1850 
1851 static void otx2_rx_mode_wrk_handler(struct work_struct *work)
1852 {
1853 	struct otx2_nic *pf = container_of(work, struct otx2_nic, rx_mode_work);
1854 
1855 	otx2_do_set_rx_mode(pf);
1856 }
1857 
1858 static int otx2_set_features(struct net_device *netdev,
1859 			     netdev_features_t features)
1860 {
1861 	netdev_features_t changed = features ^ netdev->features;
1862 	bool ntuple = !!(features & NETIF_F_NTUPLE);
1863 	struct otx2_nic *pf = netdev_priv(netdev);
1864 	bool tc = !!(features & NETIF_F_HW_TC);
1865 
1866 	if ((changed & NETIF_F_LOOPBACK) && netif_running(netdev))
1867 		return otx2_cgx_config_loopback(pf,
1868 						features & NETIF_F_LOOPBACK);
1869 
1870 	if ((changed & NETIF_F_HW_VLAN_CTAG_RX) && netif_running(netdev))
1871 		return otx2_enable_rxvlan(pf,
1872 					  features & NETIF_F_HW_VLAN_CTAG_RX);
1873 
1874 	if ((changed & NETIF_F_NTUPLE) && !ntuple)
1875 		otx2_destroy_ntuple_flows(pf);
1876 
1877 	if ((changed & NETIF_F_NTUPLE) && ntuple) {
1878 		if (!pf->flow_cfg->max_flows) {
1879 			netdev_err(netdev,
1880 				   "Can't enable NTUPLE, MCAM entries not allocated\n");
1881 			return -EINVAL;
1882 		}
1883 	}
1884 
1885 	if ((changed & NETIF_F_HW_TC) && tc) {
1886 		if (!pf->flow_cfg->max_flows) {
1887 			netdev_err(netdev,
1888 				   "Can't enable TC, MCAM entries not allocated\n");
1889 			return -EINVAL;
1890 		}
1891 	}
1892 
1893 	if ((changed & NETIF_F_HW_TC) && !tc &&
1894 	    pf->flow_cfg && pf->flow_cfg->nr_flows) {
1895 		netdev_err(netdev, "Can't disable TC hardware offload while flows are active\n");
1896 		return -EBUSY;
1897 	}
1898 
1899 	if ((changed & NETIF_F_NTUPLE) && ntuple &&
1900 	    (netdev->features & NETIF_F_HW_TC) && !(changed & NETIF_F_HW_TC)) {
1901 		netdev_err(netdev,
1902 			   "Can't enable NTUPLE when TC is active, disable TC and retry\n");
1903 		return -EINVAL;
1904 	}
1905 
1906 	if ((changed & NETIF_F_HW_TC) && tc &&
1907 	    (netdev->features & NETIF_F_NTUPLE) && !(changed & NETIF_F_NTUPLE)) {
1908 		netdev_err(netdev,
1909 			   "Can't enable TC when NTUPLE is active, disable NTUPLE and retry\n");
1910 		return -EINVAL;
1911 	}
1912 
1913 	return 0;
1914 }
1915 
1916 static void otx2_reset_task(struct work_struct *work)
1917 {
1918 	struct otx2_nic *pf = container_of(work, struct otx2_nic, reset_task);
1919 
1920 	if (!netif_running(pf->netdev))
1921 		return;
1922 
1923 	rtnl_lock();
1924 	otx2_stop(pf->netdev);
1925 	pf->reset_count++;
1926 	otx2_open(pf->netdev);
1927 	netif_trans_update(pf->netdev);
1928 	rtnl_unlock();
1929 }
1930 
1931 static int otx2_config_hw_rx_tstamp(struct otx2_nic *pfvf, bool enable)
1932 {
1933 	struct msg_req *req;
1934 	int err;
1935 
1936 	if (pfvf->flags & OTX2_FLAG_RX_TSTAMP_ENABLED && enable)
1937 		return 0;
1938 
1939 	mutex_lock(&pfvf->mbox.lock);
1940 	if (enable)
1941 		req = otx2_mbox_alloc_msg_cgx_ptp_rx_enable(&pfvf->mbox);
1942 	else
1943 		req = otx2_mbox_alloc_msg_cgx_ptp_rx_disable(&pfvf->mbox);
1944 	if (!req) {
1945 		mutex_unlock(&pfvf->mbox.lock);
1946 		return -ENOMEM;
1947 	}
1948 
1949 	err = otx2_sync_mbox_msg(&pfvf->mbox);
1950 	if (err) {
1951 		mutex_unlock(&pfvf->mbox.lock);
1952 		return err;
1953 	}
1954 
1955 	mutex_unlock(&pfvf->mbox.lock);
1956 	if (enable)
1957 		pfvf->flags |= OTX2_FLAG_RX_TSTAMP_ENABLED;
1958 	else
1959 		pfvf->flags &= ~OTX2_FLAG_RX_TSTAMP_ENABLED;
1960 	return 0;
1961 }
1962 
1963 static int otx2_config_hw_tx_tstamp(struct otx2_nic *pfvf, bool enable)
1964 {
1965 	struct msg_req *req;
1966 	int err;
1967 
1968 	if (pfvf->flags & OTX2_FLAG_TX_TSTAMP_ENABLED && enable)
1969 		return 0;
1970 
1971 	mutex_lock(&pfvf->mbox.lock);
1972 	if (enable)
1973 		req = otx2_mbox_alloc_msg_nix_lf_ptp_tx_enable(&pfvf->mbox);
1974 	else
1975 		req = otx2_mbox_alloc_msg_nix_lf_ptp_tx_disable(&pfvf->mbox);
1976 	if (!req) {
1977 		mutex_unlock(&pfvf->mbox.lock);
1978 		return -ENOMEM;
1979 	}
1980 
1981 	err = otx2_sync_mbox_msg(&pfvf->mbox);
1982 	if (err) {
1983 		mutex_unlock(&pfvf->mbox.lock);
1984 		return err;
1985 	}
1986 
1987 	mutex_unlock(&pfvf->mbox.lock);
1988 	if (enable)
1989 		pfvf->flags |= OTX2_FLAG_TX_TSTAMP_ENABLED;
1990 	else
1991 		pfvf->flags &= ~OTX2_FLAG_TX_TSTAMP_ENABLED;
1992 	return 0;
1993 }
1994 
1995 int otx2_config_hwtstamp(struct net_device *netdev, struct ifreq *ifr)
1996 {
1997 	struct otx2_nic *pfvf = netdev_priv(netdev);
1998 	struct hwtstamp_config config;
1999 
2000 	if (!pfvf->ptp)
2001 		return -ENODEV;
2002 
2003 	if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
2004 		return -EFAULT;
2005 
2006 	/* reserved for future extensions */
2007 	if (config.flags)
2008 		return -EINVAL;
2009 
2010 	switch (config.tx_type) {
2011 	case HWTSTAMP_TX_OFF:
2012 		otx2_config_hw_tx_tstamp(pfvf, false);
2013 		break;
2014 	case HWTSTAMP_TX_ON:
2015 		otx2_config_hw_tx_tstamp(pfvf, true);
2016 		break;
2017 	default:
2018 		return -ERANGE;
2019 	}
2020 
2021 	switch (config.rx_filter) {
2022 	case HWTSTAMP_FILTER_NONE:
2023 		otx2_config_hw_rx_tstamp(pfvf, false);
2024 		break;
2025 	case HWTSTAMP_FILTER_ALL:
2026 	case HWTSTAMP_FILTER_SOME:
2027 	case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
2028 	case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
2029 	case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
2030 	case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
2031 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
2032 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
2033 	case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
2034 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
2035 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
2036 	case HWTSTAMP_FILTER_PTP_V2_EVENT:
2037 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
2038 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
2039 		otx2_config_hw_rx_tstamp(pfvf, true);
2040 		config.rx_filter = HWTSTAMP_FILTER_ALL;
2041 		break;
2042 	default:
2043 		return -ERANGE;
2044 	}
2045 
2046 	memcpy(&pfvf->tstamp, &config, sizeof(config));
2047 
2048 	return copy_to_user(ifr->ifr_data, &config,
2049 			    sizeof(config)) ? -EFAULT : 0;
2050 }
2051 EXPORT_SYMBOL(otx2_config_hwtstamp);
2052 
2053 int otx2_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
2054 {
2055 	struct otx2_nic *pfvf = netdev_priv(netdev);
2056 	struct hwtstamp_config *cfg = &pfvf->tstamp;
2057 
2058 	switch (cmd) {
2059 	case SIOCSHWTSTAMP:
2060 		return otx2_config_hwtstamp(netdev, req);
2061 	case SIOCGHWTSTAMP:
2062 		return copy_to_user(req->ifr_data, cfg,
2063 				    sizeof(*cfg)) ? -EFAULT : 0;
2064 	default:
2065 		return -EOPNOTSUPP;
2066 	}
2067 }
2068 EXPORT_SYMBOL(otx2_ioctl);
2069 
2070 static int otx2_do_set_vf_mac(struct otx2_nic *pf, int vf, const u8 *mac)
2071 {
2072 	struct npc_install_flow_req *req;
2073 	int err;
2074 
2075 	mutex_lock(&pf->mbox.lock);
2076 	req = otx2_mbox_alloc_msg_npc_install_flow(&pf->mbox);
2077 	if (!req) {
2078 		err = -ENOMEM;
2079 		goto out;
2080 	}
2081 
2082 	ether_addr_copy(req->packet.dmac, mac);
2083 	eth_broadcast_addr((u8 *)&req->mask.dmac);
2084 	req->features = BIT_ULL(NPC_DMAC);
2085 	req->channel = pf->hw.rx_chan_base;
2086 	req->intf = NIX_INTF_RX;
2087 	req->default_rule = 1;
2088 	req->append = 1;
2089 	req->vf = vf + 1;
2090 	req->op = NIX_RX_ACTION_DEFAULT;
2091 
2092 	err = otx2_sync_mbox_msg(&pf->mbox);
2093 out:
2094 	mutex_unlock(&pf->mbox.lock);
2095 	return err;
2096 }
2097 
2098 static int otx2_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
2099 {
2100 	struct otx2_nic *pf = netdev_priv(netdev);
2101 	struct pci_dev *pdev = pf->pdev;
2102 	struct otx2_vf_config *config;
2103 	int ret;
2104 
2105 	if (!netif_running(netdev))
2106 		return -EAGAIN;
2107 
2108 	if (vf >= pf->total_vfs)
2109 		return -EINVAL;
2110 
2111 	if (!is_valid_ether_addr(mac))
2112 		return -EINVAL;
2113 
2114 	config = &pf->vf_configs[vf];
2115 	ether_addr_copy(config->mac, mac);
2116 
2117 	ret = otx2_do_set_vf_mac(pf, vf, mac);
2118 	if (ret == 0)
2119 		dev_info(&pdev->dev,
2120 			 "Load/Reload VF driver\n");
2121 
2122 	return ret;
2123 }
2124 
2125 static int otx2_do_set_vf_vlan(struct otx2_nic *pf, int vf, u16 vlan, u8 qos,
2126 			       __be16 proto)
2127 {
2128 	struct otx2_flow_config *flow_cfg = pf->flow_cfg;
2129 	struct nix_vtag_config_rsp *vtag_rsp;
2130 	struct npc_delete_flow_req *del_req;
2131 	struct nix_vtag_config *vtag_req;
2132 	struct npc_install_flow_req *req;
2133 	struct otx2_vf_config *config;
2134 	int err = 0;
2135 	u32 idx;
2136 
2137 	config = &pf->vf_configs[vf];
2138 
2139 	if (!vlan && !config->vlan)
2140 		goto out;
2141 
2142 	mutex_lock(&pf->mbox.lock);
2143 
2144 	/* free old tx vtag entry */
2145 	if (config->vlan) {
2146 		vtag_req = otx2_mbox_alloc_msg_nix_vtag_cfg(&pf->mbox);
2147 		if (!vtag_req) {
2148 			err = -ENOMEM;
2149 			goto out;
2150 		}
2151 		vtag_req->cfg_type = 0;
2152 		vtag_req->tx.free_vtag0 = 1;
2153 		vtag_req->tx.vtag0_idx = config->tx_vtag_idx;
2154 
2155 		err = otx2_sync_mbox_msg(&pf->mbox);
2156 		if (err)
2157 			goto out;
2158 	}
2159 
2160 	if (!vlan && config->vlan) {
2161 		/* rx */
2162 		del_req = otx2_mbox_alloc_msg_npc_delete_flow(&pf->mbox);
2163 		if (!del_req) {
2164 			err = -ENOMEM;
2165 			goto out;
2166 		}
2167 		idx = ((vf * OTX2_PER_VF_VLAN_FLOWS) + OTX2_VF_VLAN_RX_INDEX);
2168 		del_req->entry =
2169 			flow_cfg->def_ent[flow_cfg->vf_vlan_offset + idx];
2170 		err = otx2_sync_mbox_msg(&pf->mbox);
2171 		if (err)
2172 			goto out;
2173 
2174 		/* tx */
2175 		del_req = otx2_mbox_alloc_msg_npc_delete_flow(&pf->mbox);
2176 		if (!del_req) {
2177 			err = -ENOMEM;
2178 			goto out;
2179 		}
2180 		idx = ((vf * OTX2_PER_VF_VLAN_FLOWS) + OTX2_VF_VLAN_TX_INDEX);
2181 		del_req->entry =
2182 			flow_cfg->def_ent[flow_cfg->vf_vlan_offset + idx];
2183 		err = otx2_sync_mbox_msg(&pf->mbox);
2184 
2185 		goto out;
2186 	}
2187 
2188 	/* rx */
2189 	req = otx2_mbox_alloc_msg_npc_install_flow(&pf->mbox);
2190 	if (!req) {
2191 		err = -ENOMEM;
2192 		goto out;
2193 	}
2194 
2195 	idx = ((vf * OTX2_PER_VF_VLAN_FLOWS) + OTX2_VF_VLAN_RX_INDEX);
2196 	req->entry = flow_cfg->def_ent[flow_cfg->vf_vlan_offset + idx];
2197 	req->packet.vlan_tci = htons(vlan);
2198 	req->mask.vlan_tci = htons(VLAN_VID_MASK);
2199 	/* af fills the destination mac addr */
2200 	eth_broadcast_addr((u8 *)&req->mask.dmac);
2201 	req->features = BIT_ULL(NPC_OUTER_VID) | BIT_ULL(NPC_DMAC);
2202 	req->channel = pf->hw.rx_chan_base;
2203 	req->intf = NIX_INTF_RX;
2204 	req->vf = vf + 1;
2205 	req->op = NIX_RX_ACTION_DEFAULT;
2206 	req->vtag0_valid = true;
2207 	req->vtag0_type = NIX_AF_LFX_RX_VTAG_TYPE7;
2208 	req->set_cntr = 1;
2209 
2210 	err = otx2_sync_mbox_msg(&pf->mbox);
2211 	if (err)
2212 		goto out;
2213 
2214 	/* tx */
2215 	vtag_req = otx2_mbox_alloc_msg_nix_vtag_cfg(&pf->mbox);
2216 	if (!vtag_req) {
2217 		err = -ENOMEM;
2218 		goto out;
2219 	}
2220 
2221 	/* configure tx vtag params */
2222 	vtag_req->vtag_size = VTAGSIZE_T4;
2223 	vtag_req->cfg_type = 0; /* tx vlan cfg */
2224 	vtag_req->tx.cfg_vtag0 = 1;
2225 	vtag_req->tx.vtag0 = ((u64)ntohs(proto) << 16) | vlan;
2226 
2227 	err = otx2_sync_mbox_msg(&pf->mbox);
2228 	if (err)
2229 		goto out;
2230 
2231 	vtag_rsp = (struct nix_vtag_config_rsp *)otx2_mbox_get_rsp
2232 			(&pf->mbox.mbox, 0, &vtag_req->hdr);
2233 	if (IS_ERR(vtag_rsp)) {
2234 		err = PTR_ERR(vtag_rsp);
2235 		goto out;
2236 	}
2237 	config->tx_vtag_idx = vtag_rsp->vtag0_idx;
2238 
2239 	req = otx2_mbox_alloc_msg_npc_install_flow(&pf->mbox);
2240 	if (!req) {
2241 		err = -ENOMEM;
2242 		goto out;
2243 	}
2244 
2245 	eth_zero_addr((u8 *)&req->mask.dmac);
2246 	idx = ((vf * OTX2_PER_VF_VLAN_FLOWS) + OTX2_VF_VLAN_TX_INDEX);
2247 	req->entry = flow_cfg->def_ent[flow_cfg->vf_vlan_offset + idx];
2248 	req->features = BIT_ULL(NPC_DMAC);
2249 	req->channel = pf->hw.tx_chan_base;
2250 	req->intf = NIX_INTF_TX;
2251 	req->vf = vf + 1;
2252 	req->op = NIX_TX_ACTIONOP_UCAST_DEFAULT;
2253 	req->vtag0_def = vtag_rsp->vtag0_idx;
2254 	req->vtag0_op = VTAG_INSERT;
2255 	req->set_cntr = 1;
2256 
2257 	err = otx2_sync_mbox_msg(&pf->mbox);
2258 out:
2259 	config->vlan = vlan;
2260 	mutex_unlock(&pf->mbox.lock);
2261 	return err;
2262 }
2263 
2264 static int otx2_set_vf_vlan(struct net_device *netdev, int vf, u16 vlan, u8 qos,
2265 			    __be16 proto)
2266 {
2267 	struct otx2_nic *pf = netdev_priv(netdev);
2268 	struct pci_dev *pdev = pf->pdev;
2269 
2270 	if (!netif_running(netdev))
2271 		return -EAGAIN;
2272 
2273 	if (vf >= pci_num_vf(pdev))
2274 		return -EINVAL;
2275 
2276 	/* qos is currently unsupported */
2277 	if (vlan >= VLAN_N_VID || qos)
2278 		return -EINVAL;
2279 
2280 	if (proto != htons(ETH_P_8021Q))
2281 		return -EPROTONOSUPPORT;
2282 
2283 	if (!(pf->flags & OTX2_FLAG_VF_VLAN_SUPPORT))
2284 		return -EOPNOTSUPP;
2285 
2286 	return otx2_do_set_vf_vlan(pf, vf, vlan, qos, proto);
2287 }
2288 
2289 static int otx2_get_vf_config(struct net_device *netdev, int vf,
2290 			      struct ifla_vf_info *ivi)
2291 {
2292 	struct otx2_nic *pf = netdev_priv(netdev);
2293 	struct pci_dev *pdev = pf->pdev;
2294 	struct otx2_vf_config *config;
2295 
2296 	if (!netif_running(netdev))
2297 		return -EAGAIN;
2298 
2299 	if (vf >= pci_num_vf(pdev))
2300 		return -EINVAL;
2301 
2302 	config = &pf->vf_configs[vf];
2303 	ivi->vf = vf;
2304 	ether_addr_copy(ivi->mac, config->mac);
2305 	ivi->vlan = config->vlan;
2306 	ivi->trusted = config->trusted;
2307 
2308 	return 0;
2309 }
2310 
2311 static int otx2_xdp_xmit_tx(struct otx2_nic *pf, struct xdp_frame *xdpf,
2312 			    int qidx)
2313 {
2314 	struct page *page;
2315 	u64 dma_addr;
2316 	int err = 0;
2317 
2318 	dma_addr = otx2_dma_map_page(pf, virt_to_page(xdpf->data),
2319 				     offset_in_page(xdpf->data), xdpf->len,
2320 				     DMA_TO_DEVICE);
2321 	if (dma_mapping_error(pf->dev, dma_addr))
2322 		return -ENOMEM;
2323 
2324 	err = otx2_xdp_sq_append_pkt(pf, dma_addr, xdpf->len, qidx);
2325 	if (!err) {
2326 		otx2_dma_unmap_page(pf, dma_addr, xdpf->len, DMA_TO_DEVICE);
2327 		page = virt_to_page(xdpf->data);
2328 		put_page(page);
2329 		return -ENOMEM;
2330 	}
2331 	return 0;
2332 }
2333 
2334 static int otx2_xdp_xmit(struct net_device *netdev, int n,
2335 			 struct xdp_frame **frames, u32 flags)
2336 {
2337 	struct otx2_nic *pf = netdev_priv(netdev);
2338 	int qidx = smp_processor_id();
2339 	struct otx2_snd_queue *sq;
2340 	int drops = 0, i;
2341 
2342 	if (!netif_running(netdev))
2343 		return -ENETDOWN;
2344 
2345 	qidx += pf->hw.tx_queues;
2346 	sq = pf->xdp_prog ? &pf->qset.sq[qidx] : NULL;
2347 
2348 	/* Abort xmit if xdp queue is not */
2349 	if (unlikely(!sq))
2350 		return -ENXIO;
2351 
2352 	if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK))
2353 		return -EINVAL;
2354 
2355 	for (i = 0; i < n; i++) {
2356 		struct xdp_frame *xdpf = frames[i];
2357 		int err;
2358 
2359 		err = otx2_xdp_xmit_tx(pf, xdpf, qidx);
2360 		if (err)
2361 			drops++;
2362 	}
2363 	return n - drops;
2364 }
2365 
2366 static int otx2_xdp_setup(struct otx2_nic *pf, struct bpf_prog *prog)
2367 {
2368 	struct net_device *dev = pf->netdev;
2369 	bool if_up = netif_running(pf->netdev);
2370 	struct bpf_prog *old_prog;
2371 
2372 	if (prog && dev->mtu > MAX_XDP_MTU) {
2373 		netdev_warn(dev, "Jumbo frames not yet supported with XDP\n");
2374 		return -EOPNOTSUPP;
2375 	}
2376 
2377 	if (if_up)
2378 		otx2_stop(pf->netdev);
2379 
2380 	old_prog = xchg(&pf->xdp_prog, prog);
2381 
2382 	if (old_prog)
2383 		bpf_prog_put(old_prog);
2384 
2385 	if (pf->xdp_prog)
2386 		bpf_prog_add(pf->xdp_prog, pf->hw.rx_queues - 1);
2387 
2388 	/* Network stack and XDP shared same rx queues.
2389 	 * Use separate tx queues for XDP and network stack.
2390 	 */
2391 	if (pf->xdp_prog)
2392 		pf->hw.xdp_queues = pf->hw.rx_queues;
2393 	else
2394 		pf->hw.xdp_queues = 0;
2395 
2396 	pf->hw.tot_tx_queues += pf->hw.xdp_queues;
2397 
2398 	if (if_up)
2399 		otx2_open(pf->netdev);
2400 
2401 	return 0;
2402 }
2403 
2404 static int otx2_xdp(struct net_device *netdev, struct netdev_bpf *xdp)
2405 {
2406 	struct otx2_nic *pf = netdev_priv(netdev);
2407 
2408 	switch (xdp->command) {
2409 	case XDP_SETUP_PROG:
2410 		return otx2_xdp_setup(pf, xdp->prog);
2411 	default:
2412 		return -EINVAL;
2413 	}
2414 }
2415 
2416 static int otx2_set_vf_permissions(struct otx2_nic *pf, int vf,
2417 				   int req_perm)
2418 {
2419 	struct set_vf_perm *req;
2420 	int rc;
2421 
2422 	mutex_lock(&pf->mbox.lock);
2423 	req = otx2_mbox_alloc_msg_set_vf_perm(&pf->mbox);
2424 	if (!req) {
2425 		rc = -ENOMEM;
2426 		goto out;
2427 	}
2428 
2429 	/* Let AF reset VF permissions as sriov is disabled */
2430 	if (req_perm == OTX2_RESET_VF_PERM) {
2431 		req->flags |= RESET_VF_PERM;
2432 	} else if (req_perm == OTX2_TRUSTED_VF) {
2433 		if (pf->vf_configs[vf].trusted)
2434 			req->flags |= VF_TRUSTED;
2435 	}
2436 
2437 	req->vf = vf;
2438 	rc = otx2_sync_mbox_msg(&pf->mbox);
2439 out:
2440 	mutex_unlock(&pf->mbox.lock);
2441 	return rc;
2442 }
2443 
2444 static int otx2_ndo_set_vf_trust(struct net_device *netdev, int vf,
2445 				 bool enable)
2446 {
2447 	struct otx2_nic *pf = netdev_priv(netdev);
2448 	struct pci_dev *pdev = pf->pdev;
2449 	int rc;
2450 
2451 	if (vf >= pci_num_vf(pdev))
2452 		return -EINVAL;
2453 
2454 	if (pf->vf_configs[vf].trusted == enable)
2455 		return 0;
2456 
2457 	pf->vf_configs[vf].trusted = enable;
2458 	rc = otx2_set_vf_permissions(pf, vf, OTX2_TRUSTED_VF);
2459 
2460 	if (rc)
2461 		pf->vf_configs[vf].trusted = !enable;
2462 	else
2463 		netdev_info(pf->netdev, "VF %d is %strusted\n",
2464 			    vf, enable ? "" : "not ");
2465 	return rc;
2466 }
2467 
2468 static const struct net_device_ops otx2_netdev_ops = {
2469 	.ndo_open		= otx2_open,
2470 	.ndo_stop		= otx2_stop,
2471 	.ndo_start_xmit		= otx2_xmit,
2472 	.ndo_fix_features	= otx2_fix_features,
2473 	.ndo_set_mac_address    = otx2_set_mac_address,
2474 	.ndo_change_mtu		= otx2_change_mtu,
2475 	.ndo_set_rx_mode	= otx2_set_rx_mode,
2476 	.ndo_set_features	= otx2_set_features,
2477 	.ndo_tx_timeout		= otx2_tx_timeout,
2478 	.ndo_get_stats64	= otx2_get_stats64,
2479 	.ndo_eth_ioctl		= otx2_ioctl,
2480 	.ndo_set_vf_mac		= otx2_set_vf_mac,
2481 	.ndo_set_vf_vlan	= otx2_set_vf_vlan,
2482 	.ndo_get_vf_config	= otx2_get_vf_config,
2483 	.ndo_bpf		= otx2_xdp,
2484 	.ndo_xdp_xmit           = otx2_xdp_xmit,
2485 	.ndo_setup_tc		= otx2_setup_tc,
2486 	.ndo_set_vf_trust	= otx2_ndo_set_vf_trust,
2487 };
2488 
2489 static int otx2_wq_init(struct otx2_nic *pf)
2490 {
2491 	pf->otx2_wq = create_singlethread_workqueue("otx2_wq");
2492 	if (!pf->otx2_wq)
2493 		return -ENOMEM;
2494 
2495 	INIT_WORK(&pf->rx_mode_work, otx2_rx_mode_wrk_handler);
2496 	INIT_WORK(&pf->reset_task, otx2_reset_task);
2497 	return 0;
2498 }
2499 
2500 static int otx2_check_pf_usable(struct otx2_nic *nic)
2501 {
2502 	u64 rev;
2503 
2504 	rev = otx2_read64(nic, RVU_PF_BLOCK_ADDRX_DISC(BLKADDR_RVUM));
2505 	rev = (rev >> 12) & 0xFF;
2506 	/* Check if AF has setup revision for RVUM block,
2507 	 * otherwise this driver probe should be deferred
2508 	 * until AF driver comes up.
2509 	 */
2510 	if (!rev) {
2511 		dev_warn(nic->dev,
2512 			 "AF is not initialized, deferring probe\n");
2513 		return -EPROBE_DEFER;
2514 	}
2515 	return 0;
2516 }
2517 
2518 static int otx2_realloc_msix_vectors(struct otx2_nic *pf)
2519 {
2520 	struct otx2_hw *hw = &pf->hw;
2521 	int num_vec, err;
2522 
2523 	/* NPA interrupts are inot registered, so alloc only
2524 	 * upto NIX vector offset.
2525 	 */
2526 	num_vec = hw->nix_msixoff;
2527 	num_vec += NIX_LF_CINT_VEC_START + hw->max_queues;
2528 
2529 	otx2_disable_mbox_intr(pf);
2530 	pci_free_irq_vectors(hw->pdev);
2531 	err = pci_alloc_irq_vectors(hw->pdev, num_vec, num_vec, PCI_IRQ_MSIX);
2532 	if (err < 0) {
2533 		dev_err(pf->dev, "%s: Failed to realloc %d IRQ vectors\n",
2534 			__func__, num_vec);
2535 		return err;
2536 	}
2537 
2538 	return otx2_register_mbox_intr(pf, false);
2539 }
2540 
2541 static int otx2_sriov_vfcfg_init(struct otx2_nic *pf)
2542 {
2543 	int i;
2544 
2545 	pf->vf_configs = devm_kcalloc(pf->dev, pf->total_vfs,
2546 				      sizeof(struct otx2_vf_config),
2547 				      GFP_KERNEL);
2548 	if (!pf->vf_configs)
2549 		return -ENOMEM;
2550 
2551 	for (i = 0; i < pf->total_vfs; i++) {
2552 		pf->vf_configs[i].pf = pf;
2553 		pf->vf_configs[i].intf_down = true;
2554 		pf->vf_configs[i].trusted = false;
2555 		INIT_DELAYED_WORK(&pf->vf_configs[i].link_event_work,
2556 				  otx2_vf_link_event_task);
2557 	}
2558 
2559 	return 0;
2560 }
2561 
2562 static void otx2_sriov_vfcfg_cleanup(struct otx2_nic *pf)
2563 {
2564 	int i;
2565 
2566 	if (!pf->vf_configs)
2567 		return;
2568 
2569 	for (i = 0; i < pf->total_vfs; i++) {
2570 		cancel_delayed_work_sync(&pf->vf_configs[i].link_event_work);
2571 		otx2_set_vf_permissions(pf, i, OTX2_RESET_VF_PERM);
2572 	}
2573 }
2574 
2575 static int otx2_probe(struct pci_dev *pdev, const struct pci_device_id *id)
2576 {
2577 	struct device *dev = &pdev->dev;
2578 	struct net_device *netdev;
2579 	struct otx2_nic *pf;
2580 	struct otx2_hw *hw;
2581 	int err, qcount;
2582 	int num_vec;
2583 
2584 	err = pcim_enable_device(pdev);
2585 	if (err) {
2586 		dev_err(dev, "Failed to enable PCI device\n");
2587 		return err;
2588 	}
2589 
2590 	err = pci_request_regions(pdev, DRV_NAME);
2591 	if (err) {
2592 		dev_err(dev, "PCI request regions failed 0x%x\n", err);
2593 		return err;
2594 	}
2595 
2596 	err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(48));
2597 	if (err) {
2598 		dev_err(dev, "DMA mask config failed, abort\n");
2599 		goto err_release_regions;
2600 	}
2601 
2602 	pci_set_master(pdev);
2603 
2604 	/* Set number of queues */
2605 	qcount = min_t(int, num_online_cpus(), OTX2_MAX_CQ_CNT);
2606 
2607 	netdev = alloc_etherdev_mqs(sizeof(*pf), qcount, qcount);
2608 	if (!netdev) {
2609 		err = -ENOMEM;
2610 		goto err_release_regions;
2611 	}
2612 
2613 	pci_set_drvdata(pdev, netdev);
2614 	SET_NETDEV_DEV(netdev, &pdev->dev);
2615 	pf = netdev_priv(netdev);
2616 	pf->netdev = netdev;
2617 	pf->pdev = pdev;
2618 	pf->dev = dev;
2619 	pf->total_vfs = pci_sriov_get_totalvfs(pdev);
2620 	pf->flags |= OTX2_FLAG_INTF_DOWN;
2621 
2622 	hw = &pf->hw;
2623 	hw->pdev = pdev;
2624 	hw->rx_queues = qcount;
2625 	hw->tx_queues = qcount;
2626 	hw->tot_tx_queues = qcount;
2627 	hw->max_queues = qcount;
2628 
2629 	num_vec = pci_msix_vec_count(pdev);
2630 	hw->irq_name = devm_kmalloc_array(&hw->pdev->dev, num_vec, NAME_SIZE,
2631 					  GFP_KERNEL);
2632 	if (!hw->irq_name) {
2633 		err = -ENOMEM;
2634 		goto err_free_netdev;
2635 	}
2636 
2637 	hw->affinity_mask = devm_kcalloc(&hw->pdev->dev, num_vec,
2638 					 sizeof(cpumask_var_t), GFP_KERNEL);
2639 	if (!hw->affinity_mask) {
2640 		err = -ENOMEM;
2641 		goto err_free_netdev;
2642 	}
2643 
2644 	/* Map CSRs */
2645 	pf->reg_base = pcim_iomap(pdev, PCI_CFG_REG_BAR_NUM, 0);
2646 	if (!pf->reg_base) {
2647 		dev_err(dev, "Unable to map physical function CSRs, aborting\n");
2648 		err = -ENOMEM;
2649 		goto err_free_netdev;
2650 	}
2651 
2652 	err = otx2_check_pf_usable(pf);
2653 	if (err)
2654 		goto err_free_netdev;
2655 
2656 	err = pci_alloc_irq_vectors(hw->pdev, RVU_PF_INT_VEC_CNT,
2657 				    RVU_PF_INT_VEC_CNT, PCI_IRQ_MSIX);
2658 	if (err < 0) {
2659 		dev_err(dev, "%s: Failed to alloc %d IRQ vectors\n",
2660 			__func__, num_vec);
2661 		goto err_free_netdev;
2662 	}
2663 
2664 	otx2_setup_dev_hw_settings(pf);
2665 
2666 	/* Init PF <=> AF mailbox stuff */
2667 	err = otx2_pfaf_mbox_init(pf);
2668 	if (err)
2669 		goto err_free_irq_vectors;
2670 
2671 	/* Register mailbox interrupt */
2672 	err = otx2_register_mbox_intr(pf, true);
2673 	if (err)
2674 		goto err_mbox_destroy;
2675 
2676 	/* Request AF to attach NPA and NIX LFs to this PF.
2677 	 * NIX and NPA LFs are needed for this PF to function as a NIC.
2678 	 */
2679 	err = otx2_attach_npa_nix(pf);
2680 	if (err)
2681 		goto err_disable_mbox_intr;
2682 
2683 	err = otx2_realloc_msix_vectors(pf);
2684 	if (err)
2685 		goto err_detach_rsrc;
2686 
2687 	err = otx2_set_real_num_queues(netdev, hw->tx_queues, hw->rx_queues);
2688 	if (err)
2689 		goto err_detach_rsrc;
2690 
2691 	err = cn10k_lmtst_init(pf);
2692 	if (err)
2693 		goto err_detach_rsrc;
2694 
2695 	/* Assign default mac address */
2696 	otx2_get_mac_from_af(netdev);
2697 
2698 	/* Don't check for error.  Proceed without ptp */
2699 	otx2_ptp_init(pf);
2700 
2701 	/* NPA's pool is a stack to which SW frees buffer pointers via Aura.
2702 	 * HW allocates buffer pointer from stack and uses it for DMA'ing
2703 	 * ingress packet. In some scenarios HW can free back allocated buffer
2704 	 * pointers to pool. This makes it impossible for SW to maintain a
2705 	 * parallel list where physical addresses of buffer pointers (IOVAs)
2706 	 * given to HW can be saved for later reference.
2707 	 *
2708 	 * So the only way to convert Rx packet's buffer address is to use
2709 	 * IOMMU's iova_to_phys() handler which translates the address by
2710 	 * walking through the translation tables.
2711 	 */
2712 	pf->iommu_domain = iommu_get_domain_for_dev(dev);
2713 
2714 	netdev->hw_features = (NETIF_F_RXCSUM | NETIF_F_IP_CSUM |
2715 			       NETIF_F_IPV6_CSUM | NETIF_F_RXHASH |
2716 			       NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
2717 			       NETIF_F_GSO_UDP_L4);
2718 	netdev->features |= netdev->hw_features;
2719 
2720 	err = otx2_mcam_flow_init(pf);
2721 	if (err)
2722 		goto err_ptp_destroy;
2723 
2724 	if (pf->flags & OTX2_FLAG_NTUPLE_SUPPORT)
2725 		netdev->hw_features |= NETIF_F_NTUPLE;
2726 
2727 	if (pf->flags & OTX2_FLAG_UCAST_FLTR_SUPPORT)
2728 		netdev->priv_flags |= IFF_UNICAST_FLT;
2729 
2730 	/* Support TSO on tag interface */
2731 	netdev->vlan_features |= netdev->features;
2732 	netdev->hw_features  |= NETIF_F_HW_VLAN_CTAG_TX |
2733 				NETIF_F_HW_VLAN_STAG_TX;
2734 	if (pf->flags & OTX2_FLAG_RX_VLAN_SUPPORT)
2735 		netdev->hw_features |= NETIF_F_HW_VLAN_CTAG_RX |
2736 				       NETIF_F_HW_VLAN_STAG_RX;
2737 	netdev->features |= netdev->hw_features;
2738 
2739 	/* HW supports tc offload but mutually exclusive with n-tuple filters */
2740 	if (pf->flags & OTX2_FLAG_TC_FLOWER_SUPPORT)
2741 		netdev->hw_features |= NETIF_F_HW_TC;
2742 
2743 	netdev->hw_features |= NETIF_F_LOOPBACK | NETIF_F_RXALL;
2744 
2745 	netdev->gso_max_segs = OTX2_MAX_GSO_SEGS;
2746 	netdev->watchdog_timeo = OTX2_TX_TIMEOUT;
2747 
2748 	netdev->netdev_ops = &otx2_netdev_ops;
2749 
2750 	netdev->min_mtu = OTX2_MIN_MTU;
2751 	netdev->max_mtu = otx2_get_max_mtu(pf);
2752 
2753 	err = register_netdev(netdev);
2754 	if (err) {
2755 		dev_err(dev, "Failed to register netdevice\n");
2756 		goto err_del_mcam_entries;
2757 	}
2758 
2759 	err = otx2_wq_init(pf);
2760 	if (err)
2761 		goto err_unreg_netdev;
2762 
2763 	otx2_set_ethtool_ops(netdev);
2764 
2765 	err = otx2_init_tc(pf);
2766 	if (err)
2767 		goto err_mcam_flow_del;
2768 
2769 	err = otx2_register_dl(pf);
2770 	if (err)
2771 		goto err_mcam_flow_del;
2772 
2773 	/* Initialize SR-IOV resources */
2774 	err = otx2_sriov_vfcfg_init(pf);
2775 	if (err)
2776 		goto err_pf_sriov_init;
2777 
2778 	/* Enable link notifications */
2779 	otx2_cgx_config_linkevents(pf, true);
2780 
2781 	/* Enable pause frames by default */
2782 	pf->flags |= OTX2_FLAG_RX_PAUSE_ENABLED;
2783 	pf->flags |= OTX2_FLAG_TX_PAUSE_ENABLED;
2784 
2785 	return 0;
2786 
2787 err_pf_sriov_init:
2788 	otx2_shutdown_tc(pf);
2789 err_mcam_flow_del:
2790 	otx2_mcam_flow_del(pf);
2791 err_unreg_netdev:
2792 	unregister_netdev(netdev);
2793 err_del_mcam_entries:
2794 	otx2_mcam_flow_del(pf);
2795 err_ptp_destroy:
2796 	otx2_ptp_destroy(pf);
2797 err_detach_rsrc:
2798 	if (pf->hw.lmt_info)
2799 		free_percpu(pf->hw.lmt_info);
2800 	if (test_bit(CN10K_LMTST, &pf->hw.cap_flag))
2801 		qmem_free(pf->dev, pf->dync_lmt);
2802 	otx2_detach_resources(&pf->mbox);
2803 err_disable_mbox_intr:
2804 	otx2_disable_mbox_intr(pf);
2805 err_mbox_destroy:
2806 	otx2_pfaf_mbox_destroy(pf);
2807 err_free_irq_vectors:
2808 	pci_free_irq_vectors(hw->pdev);
2809 err_free_netdev:
2810 	pci_set_drvdata(pdev, NULL);
2811 	free_netdev(netdev);
2812 err_release_regions:
2813 	pci_release_regions(pdev);
2814 	return err;
2815 }
2816 
2817 static void otx2_vf_link_event_task(struct work_struct *work)
2818 {
2819 	struct otx2_vf_config *config;
2820 	struct cgx_link_info_msg *req;
2821 	struct mbox_msghdr *msghdr;
2822 	struct otx2_nic *pf;
2823 	int vf_idx;
2824 
2825 	config = container_of(work, struct otx2_vf_config,
2826 			      link_event_work.work);
2827 	vf_idx = config - config->pf->vf_configs;
2828 	pf = config->pf;
2829 
2830 	msghdr = otx2_mbox_alloc_msg_rsp(&pf->mbox_pfvf[0].mbox_up, vf_idx,
2831 					 sizeof(*req), sizeof(struct msg_rsp));
2832 	if (!msghdr) {
2833 		dev_err(pf->dev, "Failed to create VF%d link event\n", vf_idx);
2834 		return;
2835 	}
2836 
2837 	req = (struct cgx_link_info_msg *)msghdr;
2838 	req->hdr.id = MBOX_MSG_CGX_LINK_EVENT;
2839 	req->hdr.sig = OTX2_MBOX_REQ_SIG;
2840 	memcpy(&req->link_info, &pf->linfo, sizeof(req->link_info));
2841 
2842 	otx2_sync_mbox_up_msg(&pf->mbox_pfvf[0], vf_idx);
2843 }
2844 
2845 static int otx2_sriov_enable(struct pci_dev *pdev, int numvfs)
2846 {
2847 	struct net_device *netdev = pci_get_drvdata(pdev);
2848 	struct otx2_nic *pf = netdev_priv(netdev);
2849 	int ret;
2850 
2851 	/* Init PF <=> VF mailbox stuff */
2852 	ret = otx2_pfvf_mbox_init(pf, numvfs);
2853 	if (ret)
2854 		return ret;
2855 
2856 	ret = otx2_register_pfvf_mbox_intr(pf, numvfs);
2857 	if (ret)
2858 		goto free_mbox;
2859 
2860 	ret = otx2_pf_flr_init(pf, numvfs);
2861 	if (ret)
2862 		goto free_intr;
2863 
2864 	ret = otx2_register_flr_me_intr(pf, numvfs);
2865 	if (ret)
2866 		goto free_flr;
2867 
2868 	ret = pci_enable_sriov(pdev, numvfs);
2869 	if (ret)
2870 		goto free_flr_intr;
2871 
2872 	return numvfs;
2873 free_flr_intr:
2874 	otx2_disable_flr_me_intr(pf);
2875 free_flr:
2876 	otx2_flr_wq_destroy(pf);
2877 free_intr:
2878 	otx2_disable_pfvf_mbox_intr(pf, numvfs);
2879 free_mbox:
2880 	otx2_pfvf_mbox_destroy(pf);
2881 	return ret;
2882 }
2883 
2884 static int otx2_sriov_disable(struct pci_dev *pdev)
2885 {
2886 	struct net_device *netdev = pci_get_drvdata(pdev);
2887 	struct otx2_nic *pf = netdev_priv(netdev);
2888 	int numvfs = pci_num_vf(pdev);
2889 
2890 	if (!numvfs)
2891 		return 0;
2892 
2893 	pci_disable_sriov(pdev);
2894 
2895 	otx2_disable_flr_me_intr(pf);
2896 	otx2_flr_wq_destroy(pf);
2897 	otx2_disable_pfvf_mbox_intr(pf, numvfs);
2898 	otx2_pfvf_mbox_destroy(pf);
2899 
2900 	return 0;
2901 }
2902 
2903 static int otx2_sriov_configure(struct pci_dev *pdev, int numvfs)
2904 {
2905 	if (numvfs == 0)
2906 		return otx2_sriov_disable(pdev);
2907 	else
2908 		return otx2_sriov_enable(pdev, numvfs);
2909 }
2910 
2911 static void otx2_remove(struct pci_dev *pdev)
2912 {
2913 	struct net_device *netdev = pci_get_drvdata(pdev);
2914 	struct otx2_nic *pf;
2915 
2916 	if (!netdev)
2917 		return;
2918 
2919 	pf = netdev_priv(netdev);
2920 
2921 	pf->flags |= OTX2_FLAG_PF_SHUTDOWN;
2922 
2923 	if (pf->flags & OTX2_FLAG_TX_TSTAMP_ENABLED)
2924 		otx2_config_hw_tx_tstamp(pf, false);
2925 	if (pf->flags & OTX2_FLAG_RX_TSTAMP_ENABLED)
2926 		otx2_config_hw_rx_tstamp(pf, false);
2927 
2928 	cancel_work_sync(&pf->reset_task);
2929 	/* Disable link notifications */
2930 	otx2_cgx_config_linkevents(pf, false);
2931 
2932 	otx2_unregister_dl(pf);
2933 	unregister_netdev(netdev);
2934 	otx2_sriov_disable(pf->pdev);
2935 	otx2_sriov_vfcfg_cleanup(pf);
2936 	if (pf->otx2_wq)
2937 		destroy_workqueue(pf->otx2_wq);
2938 
2939 	otx2_ptp_destroy(pf);
2940 	otx2_mcam_flow_del(pf);
2941 	otx2_shutdown_tc(pf);
2942 	otx2_detach_resources(&pf->mbox);
2943 	if (pf->hw.lmt_info)
2944 		free_percpu(pf->hw.lmt_info);
2945 	if (test_bit(CN10K_LMTST, &pf->hw.cap_flag))
2946 		qmem_free(pf->dev, pf->dync_lmt);
2947 	otx2_disable_mbox_intr(pf);
2948 	otx2_pfaf_mbox_destroy(pf);
2949 	pci_free_irq_vectors(pf->pdev);
2950 	pci_set_drvdata(pdev, NULL);
2951 	free_netdev(netdev);
2952 
2953 	pci_release_regions(pdev);
2954 }
2955 
2956 static struct pci_driver otx2_pf_driver = {
2957 	.name = DRV_NAME,
2958 	.id_table = otx2_pf_id_table,
2959 	.probe = otx2_probe,
2960 	.shutdown = otx2_remove,
2961 	.remove = otx2_remove,
2962 	.sriov_configure = otx2_sriov_configure
2963 };
2964 
2965 static int __init otx2_rvupf_init_module(void)
2966 {
2967 	pr_info("%s: %s\n", DRV_NAME, DRV_STRING);
2968 
2969 	return pci_register_driver(&otx2_pf_driver);
2970 }
2971 
2972 static void __exit otx2_rvupf_cleanup_module(void)
2973 {
2974 	pci_unregister_driver(&otx2_pf_driver);
2975 }
2976 
2977 module_init(otx2_rvupf_init_module);
2978 module_exit(otx2_rvupf_cleanup_module);
2979