xref: /linux/drivers/net/ethernet/marvell/octeontx2/nic/otx2_pf.c (revision 8dd765a5d769c521d73931850d1c8708fbc490cb)
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 #include <linux/bitfield.h>
19 #include <net/page_pool/types.h>
20 
21 #include "otx2_reg.h"
22 #include "otx2_common.h"
23 #include "otx2_txrx.h"
24 #include "otx2_struct.h"
25 #include "otx2_ptp.h"
26 #include "cn10k.h"
27 #include "qos.h"
28 #include <rvu_trace.h>
29 
30 #define DRV_NAME	"rvu_nicpf"
31 #define DRV_STRING	"Marvell RVU NIC Physical Function Driver"
32 
33 /* Supported devices */
34 static const struct pci_device_id otx2_pf_id_table[] = {
35 	{ PCI_DEVICE(PCI_VENDOR_ID_CAVIUM, PCI_DEVID_OCTEONTX2_RVU_PF) },
36 	{ 0, }  /* end of table */
37 };
38 
39 MODULE_AUTHOR("Sunil Goutham <sgoutham@marvell.com>");
40 MODULE_DESCRIPTION(DRV_STRING);
41 MODULE_LICENSE("GPL v2");
42 MODULE_DEVICE_TABLE(pci, otx2_pf_id_table);
43 
44 static void otx2_vf_link_event_task(struct work_struct *work);
45 
46 enum {
47 	TYPE_PFAF,
48 	TYPE_PFVF,
49 };
50 
51 static int otx2_config_hw_tx_tstamp(struct otx2_nic *pfvf, bool enable);
52 static int otx2_config_hw_rx_tstamp(struct otx2_nic *pfvf, bool enable);
53 
54 static int otx2_change_mtu(struct net_device *netdev, int new_mtu)
55 {
56 	struct otx2_nic *pf = netdev_priv(netdev);
57 	bool if_up = netif_running(netdev);
58 	int err = 0;
59 
60 	if (pf->xdp_prog && new_mtu > MAX_XDP_MTU) {
61 		netdev_warn(netdev, "Jumbo frames not yet supported with XDP, current MTU %d.\n",
62 			    netdev->mtu);
63 		return -EINVAL;
64 	}
65 	if (if_up)
66 		otx2_stop(netdev);
67 
68 	netdev_info(netdev, "Changing MTU from %d to %d\n",
69 		    netdev->mtu, new_mtu);
70 	netdev->mtu = new_mtu;
71 
72 	if (if_up)
73 		err = otx2_open(netdev);
74 
75 	return err;
76 }
77 
78 static void otx2_disable_flr_me_intr(struct otx2_nic *pf)
79 {
80 	int irq, vfs = pf->total_vfs;
81 
82 	/* Disable VFs ME interrupts */
83 	otx2_write64(pf, RVU_PF_VFME_INT_ENA_W1CX(0), INTR_MASK(vfs));
84 	irq = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFME0);
85 	free_irq(irq, pf);
86 
87 	/* Disable VFs FLR interrupts */
88 	otx2_write64(pf, RVU_PF_VFFLR_INT_ENA_W1CX(0), INTR_MASK(vfs));
89 	irq = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFFLR0);
90 	free_irq(irq, pf);
91 
92 	if (vfs <= 64)
93 		return;
94 
95 	otx2_write64(pf, RVU_PF_VFME_INT_ENA_W1CX(1), INTR_MASK(vfs - 64));
96 	irq = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFME1);
97 	free_irq(irq, pf);
98 
99 	otx2_write64(pf, RVU_PF_VFFLR_INT_ENA_W1CX(1), INTR_MASK(vfs - 64));
100 	irq = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFFLR1);
101 	free_irq(irq, pf);
102 }
103 
104 static void otx2_flr_wq_destroy(struct otx2_nic *pf)
105 {
106 	if (!pf->flr_wq)
107 		return;
108 	destroy_workqueue(pf->flr_wq);
109 	pf->flr_wq = NULL;
110 	devm_kfree(pf->dev, pf->flr_wrk);
111 }
112 
113 static void otx2_flr_handler(struct work_struct *work)
114 {
115 	struct flr_work *flrwork = container_of(work, struct flr_work, work);
116 	struct otx2_nic *pf = flrwork->pf;
117 	struct mbox *mbox = &pf->mbox;
118 	struct msg_req *req;
119 	int vf, reg = 0;
120 
121 	vf = flrwork - pf->flr_wrk;
122 
123 	mutex_lock(&mbox->lock);
124 	req = otx2_mbox_alloc_msg_vf_flr(mbox);
125 	if (!req) {
126 		mutex_unlock(&mbox->lock);
127 		return;
128 	}
129 	req->hdr.pcifunc &= RVU_PFVF_FUNC_MASK;
130 	req->hdr.pcifunc |= (vf + 1) & RVU_PFVF_FUNC_MASK;
131 
132 	if (!otx2_sync_mbox_msg(&pf->mbox)) {
133 		if (vf >= 64) {
134 			reg = 1;
135 			vf = vf - 64;
136 		}
137 		/* clear transcation pending bit */
138 		otx2_write64(pf, RVU_PF_VFTRPENDX(reg), BIT_ULL(vf));
139 		otx2_write64(pf, RVU_PF_VFFLR_INT_ENA_W1SX(reg), BIT_ULL(vf));
140 	}
141 
142 	mutex_unlock(&mbox->lock);
143 }
144 
145 static irqreturn_t otx2_pf_flr_intr_handler(int irq, void *pf_irq)
146 {
147 	struct otx2_nic *pf = (struct otx2_nic *)pf_irq;
148 	int reg, dev, vf, start_vf, num_reg = 1;
149 	u64 intr;
150 
151 	if (pf->total_vfs > 64)
152 		num_reg = 2;
153 
154 	for (reg = 0; reg < num_reg; reg++) {
155 		intr = otx2_read64(pf, RVU_PF_VFFLR_INTX(reg));
156 		if (!intr)
157 			continue;
158 		start_vf = 64 * reg;
159 		for (vf = 0; vf < 64; vf++) {
160 			if (!(intr & BIT_ULL(vf)))
161 				continue;
162 			dev = vf + start_vf;
163 			queue_work(pf->flr_wq, &pf->flr_wrk[dev].work);
164 			/* Clear interrupt */
165 			otx2_write64(pf, RVU_PF_VFFLR_INTX(reg), BIT_ULL(vf));
166 			/* Disable the interrupt */
167 			otx2_write64(pf, RVU_PF_VFFLR_INT_ENA_W1CX(reg),
168 				     BIT_ULL(vf));
169 		}
170 	}
171 	return IRQ_HANDLED;
172 }
173 
174 static irqreturn_t otx2_pf_me_intr_handler(int irq, void *pf_irq)
175 {
176 	struct otx2_nic *pf = (struct otx2_nic *)pf_irq;
177 	int vf, reg, num_reg = 1;
178 	u64 intr;
179 
180 	if (pf->total_vfs > 64)
181 		num_reg = 2;
182 
183 	for (reg = 0; reg < num_reg; reg++) {
184 		intr = otx2_read64(pf, RVU_PF_VFME_INTX(reg));
185 		if (!intr)
186 			continue;
187 		for (vf = 0; vf < 64; vf++) {
188 			if (!(intr & BIT_ULL(vf)))
189 				continue;
190 			/* clear trpend bit */
191 			otx2_write64(pf, RVU_PF_VFTRPENDX(reg), BIT_ULL(vf));
192 			/* clear interrupt */
193 			otx2_write64(pf, RVU_PF_VFME_INTX(reg), BIT_ULL(vf));
194 		}
195 	}
196 	return IRQ_HANDLED;
197 }
198 
199 static int otx2_register_flr_me_intr(struct otx2_nic *pf, int numvfs)
200 {
201 	struct otx2_hw *hw = &pf->hw;
202 	char *irq_name;
203 	int ret;
204 
205 	/* Register ME interrupt handler*/
206 	irq_name = &hw->irq_name[RVU_PF_INT_VEC_VFME0 * NAME_SIZE];
207 	snprintf(irq_name, NAME_SIZE, "RVUPF%d_ME0", rvu_get_pf(pf->pcifunc));
208 	ret = request_irq(pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFME0),
209 			  otx2_pf_me_intr_handler, 0, irq_name, pf);
210 	if (ret) {
211 		dev_err(pf->dev,
212 			"RVUPF: IRQ registration failed for ME0\n");
213 	}
214 
215 	/* Register FLR interrupt handler */
216 	irq_name = &hw->irq_name[RVU_PF_INT_VEC_VFFLR0 * NAME_SIZE];
217 	snprintf(irq_name, NAME_SIZE, "RVUPF%d_FLR0", rvu_get_pf(pf->pcifunc));
218 	ret = request_irq(pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFFLR0),
219 			  otx2_pf_flr_intr_handler, 0, irq_name, pf);
220 	if (ret) {
221 		dev_err(pf->dev,
222 			"RVUPF: IRQ registration failed for FLR0\n");
223 		return ret;
224 	}
225 
226 	if (numvfs > 64) {
227 		irq_name = &hw->irq_name[RVU_PF_INT_VEC_VFME1 * NAME_SIZE];
228 		snprintf(irq_name, NAME_SIZE, "RVUPF%d_ME1",
229 			 rvu_get_pf(pf->pcifunc));
230 		ret = request_irq(pci_irq_vector
231 				  (pf->pdev, RVU_PF_INT_VEC_VFME1),
232 				  otx2_pf_me_intr_handler, 0, irq_name, pf);
233 		if (ret) {
234 			dev_err(pf->dev,
235 				"RVUPF: IRQ registration failed for ME1\n");
236 		}
237 		irq_name = &hw->irq_name[RVU_PF_INT_VEC_VFFLR1 * NAME_SIZE];
238 		snprintf(irq_name, NAME_SIZE, "RVUPF%d_FLR1",
239 			 rvu_get_pf(pf->pcifunc));
240 		ret = request_irq(pci_irq_vector
241 				  (pf->pdev, RVU_PF_INT_VEC_VFFLR1),
242 				  otx2_pf_flr_intr_handler, 0, irq_name, pf);
243 		if (ret) {
244 			dev_err(pf->dev,
245 				"RVUPF: IRQ registration failed for FLR1\n");
246 			return ret;
247 		}
248 	}
249 
250 	/* Enable ME interrupt for all VFs*/
251 	otx2_write64(pf, RVU_PF_VFME_INTX(0), INTR_MASK(numvfs));
252 	otx2_write64(pf, RVU_PF_VFME_INT_ENA_W1SX(0), INTR_MASK(numvfs));
253 
254 	/* Enable FLR interrupt for all VFs*/
255 	otx2_write64(pf, RVU_PF_VFFLR_INTX(0), INTR_MASK(numvfs));
256 	otx2_write64(pf, RVU_PF_VFFLR_INT_ENA_W1SX(0), INTR_MASK(numvfs));
257 
258 	if (numvfs > 64) {
259 		numvfs -= 64;
260 
261 		otx2_write64(pf, RVU_PF_VFME_INTX(1), INTR_MASK(numvfs));
262 		otx2_write64(pf, RVU_PF_VFME_INT_ENA_W1SX(1),
263 			     INTR_MASK(numvfs));
264 
265 		otx2_write64(pf, RVU_PF_VFFLR_INTX(1), INTR_MASK(numvfs));
266 		otx2_write64(pf, RVU_PF_VFFLR_INT_ENA_W1SX(1),
267 			     INTR_MASK(numvfs));
268 	}
269 	return 0;
270 }
271 
272 static int otx2_pf_flr_init(struct otx2_nic *pf, int num_vfs)
273 {
274 	int vf;
275 
276 	pf->flr_wq = alloc_ordered_workqueue("otx2_pf_flr_wq", WQ_HIGHPRI);
277 	if (!pf->flr_wq)
278 		return -ENOMEM;
279 
280 	pf->flr_wrk = devm_kcalloc(pf->dev, num_vfs,
281 				   sizeof(struct flr_work), GFP_KERNEL);
282 	if (!pf->flr_wrk) {
283 		destroy_workqueue(pf->flr_wq);
284 		return -ENOMEM;
285 	}
286 
287 	for (vf = 0; vf < num_vfs; vf++) {
288 		pf->flr_wrk[vf].pf = pf;
289 		INIT_WORK(&pf->flr_wrk[vf].work, otx2_flr_handler);
290 	}
291 
292 	return 0;
293 }
294 
295 static void otx2_queue_work(struct mbox *mw, struct workqueue_struct *mbox_wq,
296 			    int first, int mdevs, u64 intr, int type)
297 {
298 	struct otx2_mbox_dev *mdev;
299 	struct otx2_mbox *mbox;
300 	struct mbox_hdr *hdr;
301 	int i;
302 
303 	for (i = first; i < mdevs; i++) {
304 		/* start from 0 */
305 		if (!(intr & BIT_ULL(i - first)))
306 			continue;
307 
308 		mbox = &mw->mbox;
309 		mdev = &mbox->dev[i];
310 		if (type == TYPE_PFAF)
311 			otx2_sync_mbox_bbuf(mbox, i);
312 		hdr = mdev->mbase + mbox->rx_start;
313 		/* The hdr->num_msgs is set to zero immediately in the interrupt
314 		 * handler to  ensure that it holds a correct value next time
315 		 * when the interrupt handler is called.
316 		 * pf->mbox.num_msgs holds the data for use in pfaf_mbox_handler
317 		 * pf>mbox.up_num_msgs holds the data for use in
318 		 * pfaf_mbox_up_handler.
319 		 */
320 		if (hdr->num_msgs) {
321 			mw[i].num_msgs = hdr->num_msgs;
322 			hdr->num_msgs = 0;
323 			if (type == TYPE_PFAF)
324 				memset(mbox->hwbase + mbox->rx_start, 0,
325 				       ALIGN(sizeof(struct mbox_hdr),
326 					     sizeof(u64)));
327 
328 			queue_work(mbox_wq, &mw[i].mbox_wrk);
329 		}
330 
331 		mbox = &mw->mbox_up;
332 		mdev = &mbox->dev[i];
333 		if (type == TYPE_PFAF)
334 			otx2_sync_mbox_bbuf(mbox, i);
335 		hdr = mdev->mbase + mbox->rx_start;
336 		if (hdr->num_msgs) {
337 			mw[i].up_num_msgs = hdr->num_msgs;
338 			hdr->num_msgs = 0;
339 			if (type == TYPE_PFAF)
340 				memset(mbox->hwbase + mbox->rx_start, 0,
341 				       ALIGN(sizeof(struct mbox_hdr),
342 					     sizeof(u64)));
343 
344 			queue_work(mbox_wq, &mw[i].mbox_up_wrk);
345 		}
346 	}
347 }
348 
349 static void otx2_forward_msg_pfvf(struct otx2_mbox_dev *mdev,
350 				  struct otx2_mbox *pfvf_mbox, void *bbuf_base,
351 				  int devid)
352 {
353 	struct otx2_mbox_dev *src_mdev = mdev;
354 	int offset;
355 
356 	/* Msgs are already copied, trigger VF's mbox irq */
357 	smp_wmb();
358 
359 	offset = pfvf_mbox->trigger | (devid << pfvf_mbox->tr_shift);
360 	writeq(1, (void __iomem *)pfvf_mbox->reg_base + offset);
361 
362 	/* Restore VF's mbox bounce buffer region address */
363 	src_mdev->mbase = bbuf_base;
364 }
365 
366 static int otx2_forward_vf_mbox_msgs(struct otx2_nic *pf,
367 				     struct otx2_mbox *src_mbox,
368 				     int dir, int vf, int num_msgs)
369 {
370 	struct otx2_mbox_dev *src_mdev, *dst_mdev;
371 	struct mbox_hdr *mbox_hdr;
372 	struct mbox_hdr *req_hdr;
373 	struct mbox *dst_mbox;
374 	int dst_size, err;
375 
376 	if (dir == MBOX_DIR_PFAF) {
377 		/* Set VF's mailbox memory as PF's bounce buffer memory, so
378 		 * that explicit copying of VF's msgs to PF=>AF mbox region
379 		 * and AF=>PF responses to VF's mbox region can be avoided.
380 		 */
381 		src_mdev = &src_mbox->dev[vf];
382 		mbox_hdr = src_mbox->hwbase +
383 				src_mbox->rx_start + (vf * MBOX_SIZE);
384 
385 		dst_mbox = &pf->mbox;
386 		dst_size = dst_mbox->mbox.tx_size -
387 				ALIGN(sizeof(*mbox_hdr), MBOX_MSG_ALIGN);
388 		/* Check if msgs fit into destination area and has valid size */
389 		if (mbox_hdr->msg_size > dst_size || !mbox_hdr->msg_size)
390 			return -EINVAL;
391 
392 		dst_mdev = &dst_mbox->mbox.dev[0];
393 
394 		mutex_lock(&pf->mbox.lock);
395 		dst_mdev->mbase = src_mdev->mbase;
396 		dst_mdev->msg_size = mbox_hdr->msg_size;
397 		dst_mdev->num_msgs = num_msgs;
398 		err = otx2_sync_mbox_msg(dst_mbox);
399 		/* Error code -EIO indicate there is a communication failure
400 		 * to the AF. Rest of the error codes indicate that AF processed
401 		 * VF messages and set the error codes in response messages
402 		 * (if any) so simply forward responses to VF.
403 		 */
404 		if (err == -EIO) {
405 			dev_warn(pf->dev,
406 				 "AF not responding to VF%d messages\n", vf);
407 			/* restore PF mbase and exit */
408 			dst_mdev->mbase = pf->mbox.bbuf_base;
409 			mutex_unlock(&pf->mbox.lock);
410 			return err;
411 		}
412 		/* At this point, all the VF messages sent to AF are acked
413 		 * with proper responses and responses are copied to VF
414 		 * mailbox hence raise interrupt to VF.
415 		 */
416 		req_hdr = (struct mbox_hdr *)(dst_mdev->mbase +
417 					      dst_mbox->mbox.rx_start);
418 		req_hdr->num_msgs = num_msgs;
419 
420 		otx2_forward_msg_pfvf(dst_mdev, &pf->mbox_pfvf[0].mbox,
421 				      pf->mbox.bbuf_base, vf);
422 		mutex_unlock(&pf->mbox.lock);
423 	} else if (dir == MBOX_DIR_PFVF_UP) {
424 		src_mdev = &src_mbox->dev[0];
425 		mbox_hdr = src_mbox->hwbase + src_mbox->rx_start;
426 		req_hdr = (struct mbox_hdr *)(src_mdev->mbase +
427 					      src_mbox->rx_start);
428 		req_hdr->num_msgs = num_msgs;
429 
430 		dst_mbox = &pf->mbox_pfvf[0];
431 		dst_size = dst_mbox->mbox_up.tx_size -
432 				ALIGN(sizeof(*mbox_hdr), MBOX_MSG_ALIGN);
433 		/* Check if msgs fit into destination area */
434 		if (mbox_hdr->msg_size > dst_size)
435 			return -EINVAL;
436 
437 		dst_mdev = &dst_mbox->mbox_up.dev[vf];
438 		dst_mdev->mbase = src_mdev->mbase;
439 		dst_mdev->msg_size = mbox_hdr->msg_size;
440 		dst_mdev->num_msgs = mbox_hdr->num_msgs;
441 		err = otx2_sync_mbox_up_msg(dst_mbox, vf);
442 		if (err) {
443 			dev_warn(pf->dev,
444 				 "VF%d is not responding to mailbox\n", vf);
445 			return err;
446 		}
447 	} else if (dir == MBOX_DIR_VFPF_UP) {
448 		req_hdr = (struct mbox_hdr *)(src_mbox->dev[0].mbase +
449 					      src_mbox->rx_start);
450 		req_hdr->num_msgs = num_msgs;
451 		otx2_forward_msg_pfvf(&pf->mbox_pfvf->mbox_up.dev[vf],
452 				      &pf->mbox.mbox_up,
453 				      pf->mbox_pfvf[vf].bbuf_base,
454 				      0);
455 	}
456 
457 	return 0;
458 }
459 
460 static void otx2_pfvf_mbox_handler(struct work_struct *work)
461 {
462 	struct mbox_msghdr *msg = NULL;
463 	int offset, vf_idx, id, err;
464 	struct otx2_mbox_dev *mdev;
465 	struct mbox_hdr *req_hdr;
466 	struct otx2_mbox *mbox;
467 	struct mbox *vf_mbox;
468 	struct otx2_nic *pf;
469 
470 	vf_mbox = container_of(work, struct mbox, mbox_wrk);
471 	pf = vf_mbox->pfvf;
472 	vf_idx = vf_mbox - pf->mbox_pfvf;
473 
474 	mbox = &pf->mbox_pfvf[0].mbox;
475 	mdev = &mbox->dev[vf_idx];
476 	req_hdr = (struct mbox_hdr *)(mdev->mbase + mbox->rx_start);
477 
478 	offset = ALIGN(sizeof(*req_hdr), MBOX_MSG_ALIGN);
479 
480 	for (id = 0; id < vf_mbox->num_msgs; id++) {
481 		msg = (struct mbox_msghdr *)(mdev->mbase + mbox->rx_start +
482 					     offset);
483 
484 		if (msg->sig != OTX2_MBOX_REQ_SIG)
485 			goto inval_msg;
486 
487 		/* Set VF's number in each of the msg */
488 		msg->pcifunc &= RVU_PFVF_FUNC_MASK;
489 		msg->pcifunc |= (vf_idx + 1) & RVU_PFVF_FUNC_MASK;
490 		offset = msg->next_msgoff;
491 	}
492 	err = otx2_forward_vf_mbox_msgs(pf, mbox, MBOX_DIR_PFAF, vf_idx,
493 					vf_mbox->num_msgs);
494 	if (err)
495 		goto inval_msg;
496 	return;
497 
498 inval_msg:
499 	otx2_reply_invalid_msg(mbox, vf_idx, 0, msg->id);
500 	otx2_mbox_msg_send(mbox, vf_idx);
501 }
502 
503 static void otx2_pfvf_mbox_up_handler(struct work_struct *work)
504 {
505 	struct mbox *vf_mbox = container_of(work, struct mbox, mbox_up_wrk);
506 	struct otx2_nic *pf = vf_mbox->pfvf;
507 	struct otx2_mbox_dev *mdev;
508 	int offset, id, vf_idx = 0;
509 	struct mbox_hdr *rsp_hdr;
510 	struct mbox_msghdr *msg;
511 	struct otx2_mbox *mbox;
512 
513 	vf_idx = vf_mbox - pf->mbox_pfvf;
514 	mbox = &pf->mbox_pfvf[0].mbox_up;
515 	mdev = &mbox->dev[vf_idx];
516 
517 	rsp_hdr = (struct mbox_hdr *)(mdev->mbase + mbox->rx_start);
518 	offset = mbox->rx_start + ALIGN(sizeof(*rsp_hdr), MBOX_MSG_ALIGN);
519 
520 	for (id = 0; id < vf_mbox->up_num_msgs; id++) {
521 		msg = mdev->mbase + offset;
522 
523 		if (msg->id >= MBOX_MSG_MAX) {
524 			dev_err(pf->dev,
525 				"Mbox msg with unknown ID 0x%x\n", msg->id);
526 			goto end;
527 		}
528 
529 		if (msg->sig != OTX2_MBOX_RSP_SIG) {
530 			dev_err(pf->dev,
531 				"Mbox msg with wrong signature %x, ID 0x%x\n",
532 				msg->sig, msg->id);
533 			goto end;
534 		}
535 
536 		switch (msg->id) {
537 		case MBOX_MSG_CGX_LINK_EVENT:
538 			break;
539 		default:
540 			if (msg->rc)
541 				dev_err(pf->dev,
542 					"Mbox msg response has err %d, ID 0x%x\n",
543 					msg->rc, msg->id);
544 			break;
545 		}
546 
547 end:
548 		offset = mbox->rx_start + msg->next_msgoff;
549 		if (mdev->msgs_acked == (vf_mbox->up_num_msgs - 1))
550 			__otx2_mbox_reset(mbox, 0);
551 		mdev->msgs_acked++;
552 	}
553 }
554 
555 static irqreturn_t otx2_pfvf_mbox_intr_handler(int irq, void *pf_irq)
556 {
557 	struct otx2_nic *pf = (struct otx2_nic *)(pf_irq);
558 	int vfs = pf->total_vfs;
559 	struct mbox *mbox;
560 	u64 intr;
561 
562 	mbox = pf->mbox_pfvf;
563 	/* Handle VF interrupts */
564 	if (vfs > 64) {
565 		intr = otx2_read64(pf, RVU_PF_VFPF_MBOX_INTX(1));
566 		otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(1), intr);
567 		otx2_queue_work(mbox, pf->mbox_pfvf_wq, 64, vfs, intr,
568 				TYPE_PFVF);
569 		vfs -= 64;
570 	}
571 
572 	intr = otx2_read64(pf, RVU_PF_VFPF_MBOX_INTX(0));
573 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(0), intr);
574 
575 	otx2_queue_work(mbox, pf->mbox_pfvf_wq, 0, vfs, intr, TYPE_PFVF);
576 
577 	trace_otx2_msg_interrupt(mbox->mbox.pdev, "VF(s) to PF", intr);
578 
579 	return IRQ_HANDLED;
580 }
581 
582 static int otx2_pfvf_mbox_init(struct otx2_nic *pf, int numvfs)
583 {
584 	void __iomem *hwbase;
585 	struct mbox *mbox;
586 	int err, vf;
587 	u64 base;
588 
589 	if (!numvfs)
590 		return -EINVAL;
591 
592 	pf->mbox_pfvf = devm_kcalloc(&pf->pdev->dev, numvfs,
593 				     sizeof(struct mbox), GFP_KERNEL);
594 	if (!pf->mbox_pfvf)
595 		return -ENOMEM;
596 
597 	pf->mbox_pfvf_wq = alloc_ordered_workqueue("otx2_pfvf_mailbox",
598 						   WQ_HIGHPRI | WQ_MEM_RECLAIM);
599 	if (!pf->mbox_pfvf_wq)
600 		return -ENOMEM;
601 
602 	/* On CN10K platform, PF <-> VF mailbox region follows after
603 	 * PF <-> AF mailbox region.
604 	 */
605 	if (test_bit(CN10K_MBOX, &pf->hw.cap_flag))
606 		base = pci_resource_start(pf->pdev, PCI_MBOX_BAR_NUM) +
607 		       MBOX_SIZE;
608 	else
609 		base = readq((void __iomem *)((u64)pf->reg_base +
610 					      RVU_PF_VF_BAR4_ADDR));
611 
612 	hwbase = ioremap_wc(base, MBOX_SIZE * pf->total_vfs);
613 	if (!hwbase) {
614 		err = -ENOMEM;
615 		goto free_wq;
616 	}
617 
618 	mbox = &pf->mbox_pfvf[0];
619 	err = otx2_mbox_init(&mbox->mbox, hwbase, pf->pdev, pf->reg_base,
620 			     MBOX_DIR_PFVF, numvfs);
621 	if (err)
622 		goto free_iomem;
623 
624 	err = otx2_mbox_init(&mbox->mbox_up, hwbase, pf->pdev, pf->reg_base,
625 			     MBOX_DIR_PFVF_UP, numvfs);
626 	if (err)
627 		goto free_iomem;
628 
629 	for (vf = 0; vf < numvfs; vf++) {
630 		mbox->pfvf = pf;
631 		INIT_WORK(&mbox->mbox_wrk, otx2_pfvf_mbox_handler);
632 		INIT_WORK(&mbox->mbox_up_wrk, otx2_pfvf_mbox_up_handler);
633 		mbox++;
634 	}
635 
636 	return 0;
637 
638 free_iomem:
639 	if (hwbase)
640 		iounmap(hwbase);
641 free_wq:
642 	destroy_workqueue(pf->mbox_pfvf_wq);
643 	return err;
644 }
645 
646 static void otx2_pfvf_mbox_destroy(struct otx2_nic *pf)
647 {
648 	struct mbox *mbox = &pf->mbox_pfvf[0];
649 
650 	if (!mbox)
651 		return;
652 
653 	if (pf->mbox_pfvf_wq) {
654 		destroy_workqueue(pf->mbox_pfvf_wq);
655 		pf->mbox_pfvf_wq = NULL;
656 	}
657 
658 	if (mbox->mbox.hwbase)
659 		iounmap(mbox->mbox.hwbase);
660 
661 	otx2_mbox_destroy(&mbox->mbox);
662 }
663 
664 static void otx2_enable_pfvf_mbox_intr(struct otx2_nic *pf, int numvfs)
665 {
666 	/* Clear PF <=> VF mailbox IRQ */
667 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(0), ~0ull);
668 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(1), ~0ull);
669 
670 	/* Enable PF <=> VF mailbox IRQ */
671 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INT_ENA_W1SX(0), INTR_MASK(numvfs));
672 	if (numvfs > 64) {
673 		numvfs -= 64;
674 		otx2_write64(pf, RVU_PF_VFPF_MBOX_INT_ENA_W1SX(1),
675 			     INTR_MASK(numvfs));
676 	}
677 }
678 
679 static void otx2_disable_pfvf_mbox_intr(struct otx2_nic *pf, int numvfs)
680 {
681 	int vector;
682 
683 	/* Disable PF <=> VF mailbox IRQ */
684 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INT_ENA_W1CX(0), ~0ull);
685 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INT_ENA_W1CX(1), ~0ull);
686 
687 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(0), ~0ull);
688 	vector = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFPF_MBOX0);
689 	free_irq(vector, pf);
690 
691 	if (numvfs > 64) {
692 		otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(1), ~0ull);
693 		vector = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFPF_MBOX1);
694 		free_irq(vector, pf);
695 	}
696 }
697 
698 static int otx2_register_pfvf_mbox_intr(struct otx2_nic *pf, int numvfs)
699 {
700 	struct otx2_hw *hw = &pf->hw;
701 	char *irq_name;
702 	int err;
703 
704 	/* Register MBOX0 interrupt handler */
705 	irq_name = &hw->irq_name[RVU_PF_INT_VEC_VFPF_MBOX0 * NAME_SIZE];
706 	if (pf->pcifunc)
707 		snprintf(irq_name, NAME_SIZE,
708 			 "RVUPF%d_VF Mbox0", rvu_get_pf(pf->pcifunc));
709 	else
710 		snprintf(irq_name, NAME_SIZE, "RVUPF_VF Mbox0");
711 	err = request_irq(pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFPF_MBOX0),
712 			  otx2_pfvf_mbox_intr_handler, 0, irq_name, pf);
713 	if (err) {
714 		dev_err(pf->dev,
715 			"RVUPF: IRQ registration failed for PFVF mbox0 irq\n");
716 		return err;
717 	}
718 
719 	if (numvfs > 64) {
720 		/* Register MBOX1 interrupt handler */
721 		irq_name = &hw->irq_name[RVU_PF_INT_VEC_VFPF_MBOX1 * NAME_SIZE];
722 		if (pf->pcifunc)
723 			snprintf(irq_name, NAME_SIZE,
724 				 "RVUPF%d_VF Mbox1", rvu_get_pf(pf->pcifunc));
725 		else
726 			snprintf(irq_name, NAME_SIZE, "RVUPF_VF Mbox1");
727 		err = request_irq(pci_irq_vector(pf->pdev,
728 						 RVU_PF_INT_VEC_VFPF_MBOX1),
729 						 otx2_pfvf_mbox_intr_handler,
730 						 0, irq_name, pf);
731 		if (err) {
732 			dev_err(pf->dev,
733 				"RVUPF: IRQ registration failed for PFVF mbox1 irq\n");
734 			return err;
735 		}
736 	}
737 
738 	otx2_enable_pfvf_mbox_intr(pf, numvfs);
739 
740 	return 0;
741 }
742 
743 static void otx2_process_pfaf_mbox_msg(struct otx2_nic *pf,
744 				       struct mbox_msghdr *msg)
745 {
746 	int devid;
747 
748 	if (msg->id >= MBOX_MSG_MAX) {
749 		dev_err(pf->dev,
750 			"Mbox msg with unknown ID 0x%x\n", msg->id);
751 		return;
752 	}
753 
754 	if (msg->sig != OTX2_MBOX_RSP_SIG) {
755 		dev_err(pf->dev,
756 			"Mbox msg with wrong signature %x, ID 0x%x\n",
757 			 msg->sig, msg->id);
758 		return;
759 	}
760 
761 	/* message response heading VF */
762 	devid = msg->pcifunc & RVU_PFVF_FUNC_MASK;
763 	if (devid) {
764 		struct otx2_vf_config *config = &pf->vf_configs[devid - 1];
765 		struct delayed_work *dwork;
766 
767 		switch (msg->id) {
768 		case MBOX_MSG_NIX_LF_START_RX:
769 			config->intf_down = false;
770 			dwork = &config->link_event_work;
771 			schedule_delayed_work(dwork, msecs_to_jiffies(100));
772 			break;
773 		case MBOX_MSG_NIX_LF_STOP_RX:
774 			config->intf_down = true;
775 			break;
776 		}
777 
778 		return;
779 	}
780 
781 	switch (msg->id) {
782 	case MBOX_MSG_READY:
783 		pf->pcifunc = msg->pcifunc;
784 		break;
785 	case MBOX_MSG_MSIX_OFFSET:
786 		mbox_handler_msix_offset(pf, (struct msix_offset_rsp *)msg);
787 		break;
788 	case MBOX_MSG_NPA_LF_ALLOC:
789 		mbox_handler_npa_lf_alloc(pf, (struct npa_lf_alloc_rsp *)msg);
790 		break;
791 	case MBOX_MSG_NIX_LF_ALLOC:
792 		mbox_handler_nix_lf_alloc(pf, (struct nix_lf_alloc_rsp *)msg);
793 		break;
794 	case MBOX_MSG_NIX_BP_ENABLE:
795 		mbox_handler_nix_bp_enable(pf, (struct nix_bp_cfg_rsp *)msg);
796 		break;
797 	case MBOX_MSG_CGX_STATS:
798 		mbox_handler_cgx_stats(pf, (struct cgx_stats_rsp *)msg);
799 		break;
800 	case MBOX_MSG_CGX_FEC_STATS:
801 		mbox_handler_cgx_fec_stats(pf, (struct cgx_fec_stats_rsp *)msg);
802 		break;
803 	default:
804 		if (msg->rc)
805 			dev_err(pf->dev,
806 				"Mbox msg response has err %d, ID 0x%x\n",
807 				msg->rc, msg->id);
808 		break;
809 	}
810 }
811 
812 static void otx2_pfaf_mbox_handler(struct work_struct *work)
813 {
814 	struct otx2_mbox_dev *mdev;
815 	struct mbox_hdr *rsp_hdr;
816 	struct mbox_msghdr *msg;
817 	struct otx2_mbox *mbox;
818 	struct mbox *af_mbox;
819 	struct otx2_nic *pf;
820 	int offset, id;
821 
822 	af_mbox = container_of(work, struct mbox, mbox_wrk);
823 	mbox = &af_mbox->mbox;
824 	mdev = &mbox->dev[0];
825 	rsp_hdr = (struct mbox_hdr *)(mdev->mbase + mbox->rx_start);
826 
827 	offset = mbox->rx_start + ALIGN(sizeof(*rsp_hdr), MBOX_MSG_ALIGN);
828 	pf = af_mbox->pfvf;
829 
830 	for (id = 0; id < af_mbox->num_msgs; id++) {
831 		msg = (struct mbox_msghdr *)(mdev->mbase + offset);
832 		otx2_process_pfaf_mbox_msg(pf, msg);
833 		offset = mbox->rx_start + msg->next_msgoff;
834 		if (mdev->msgs_acked == (af_mbox->num_msgs - 1))
835 			__otx2_mbox_reset(mbox, 0);
836 		mdev->msgs_acked++;
837 	}
838 
839 }
840 
841 static void otx2_handle_link_event(struct otx2_nic *pf)
842 {
843 	struct cgx_link_user_info *linfo = &pf->linfo;
844 	struct net_device *netdev = pf->netdev;
845 
846 	pr_info("%s NIC Link is %s %d Mbps %s duplex\n", netdev->name,
847 		linfo->link_up ? "UP" : "DOWN", linfo->speed,
848 		linfo->full_duplex ? "Full" : "Half");
849 	if (linfo->link_up) {
850 		netif_carrier_on(netdev);
851 		netif_tx_start_all_queues(netdev);
852 	} else {
853 		netif_tx_stop_all_queues(netdev);
854 		netif_carrier_off(netdev);
855 	}
856 }
857 
858 int otx2_mbox_up_handler_mcs_intr_notify(struct otx2_nic *pf,
859 					 struct mcs_intr_info *event,
860 					 struct msg_rsp *rsp)
861 {
862 	cn10k_handle_mcs_event(pf, event);
863 
864 	return 0;
865 }
866 
867 int otx2_mbox_up_handler_cgx_link_event(struct otx2_nic *pf,
868 					struct cgx_link_info_msg *msg,
869 					struct msg_rsp *rsp)
870 {
871 	int i;
872 
873 	/* Copy the link info sent by AF */
874 	pf->linfo = msg->link_info;
875 
876 	/* notify VFs about link event */
877 	for (i = 0; i < pci_num_vf(pf->pdev); i++) {
878 		struct otx2_vf_config *config = &pf->vf_configs[i];
879 		struct delayed_work *dwork = &config->link_event_work;
880 
881 		if (config->intf_down)
882 			continue;
883 
884 		schedule_delayed_work(dwork, msecs_to_jiffies(100));
885 	}
886 
887 	/* interface has not been fully configured yet */
888 	if (pf->flags & OTX2_FLAG_INTF_DOWN)
889 		return 0;
890 
891 	otx2_handle_link_event(pf);
892 	return 0;
893 }
894 
895 static int otx2_process_mbox_msg_up(struct otx2_nic *pf,
896 				    struct mbox_msghdr *req)
897 {
898 	/* Check if valid, if not reply with a invalid msg */
899 	if (req->sig != OTX2_MBOX_REQ_SIG) {
900 		otx2_reply_invalid_msg(&pf->mbox.mbox_up, 0, 0, req->id);
901 		return -ENODEV;
902 	}
903 
904 	switch (req->id) {
905 #define M(_name, _id, _fn_name, _req_type, _rsp_type)			\
906 	case _id: {							\
907 		struct _rsp_type *rsp;					\
908 		int err;						\
909 									\
910 		rsp = (struct _rsp_type *)otx2_mbox_alloc_msg(		\
911 			&pf->mbox.mbox_up, 0,				\
912 			sizeof(struct _rsp_type));			\
913 		if (!rsp)						\
914 			return -ENOMEM;					\
915 									\
916 		rsp->hdr.id = _id;					\
917 		rsp->hdr.sig = OTX2_MBOX_RSP_SIG;			\
918 		rsp->hdr.pcifunc = 0;					\
919 		rsp->hdr.rc = 0;					\
920 									\
921 		err = otx2_mbox_up_handler_ ## _fn_name(		\
922 			pf, (struct _req_type *)req, rsp);		\
923 		return err;						\
924 	}
925 MBOX_UP_CGX_MESSAGES
926 MBOX_UP_MCS_MESSAGES
927 #undef M
928 		break;
929 	default:
930 		otx2_reply_invalid_msg(&pf->mbox.mbox_up, 0, 0, req->id);
931 		return -ENODEV;
932 	}
933 	return 0;
934 }
935 
936 static void otx2_pfaf_mbox_up_handler(struct work_struct *work)
937 {
938 	struct mbox *af_mbox = container_of(work, struct mbox, mbox_up_wrk);
939 	struct otx2_mbox *mbox = &af_mbox->mbox_up;
940 	struct otx2_mbox_dev *mdev = &mbox->dev[0];
941 	struct otx2_nic *pf = af_mbox->pfvf;
942 	int offset, id, devid = 0;
943 	struct mbox_hdr *rsp_hdr;
944 	struct mbox_msghdr *msg;
945 
946 	rsp_hdr = (struct mbox_hdr *)(mdev->mbase + mbox->rx_start);
947 
948 	offset = mbox->rx_start + ALIGN(sizeof(*rsp_hdr), MBOX_MSG_ALIGN);
949 
950 	for (id = 0; id < af_mbox->up_num_msgs; id++) {
951 		msg = (struct mbox_msghdr *)(mdev->mbase + offset);
952 
953 		devid = msg->pcifunc & RVU_PFVF_FUNC_MASK;
954 		/* Skip processing VF's messages */
955 		if (!devid)
956 			otx2_process_mbox_msg_up(pf, msg);
957 		offset = mbox->rx_start + msg->next_msgoff;
958 	}
959 	if (devid) {
960 		otx2_forward_vf_mbox_msgs(pf, &pf->mbox.mbox_up,
961 					  MBOX_DIR_PFVF_UP, devid - 1,
962 					  af_mbox->up_num_msgs);
963 		return;
964 	}
965 
966 	otx2_mbox_msg_send(mbox, 0);
967 }
968 
969 static irqreturn_t otx2_pfaf_mbox_intr_handler(int irq, void *pf_irq)
970 {
971 	struct otx2_nic *pf = (struct otx2_nic *)pf_irq;
972 	struct mbox *mbox;
973 
974 	/* Clear the IRQ */
975 	otx2_write64(pf, RVU_PF_INT, BIT_ULL(0));
976 
977 	mbox = &pf->mbox;
978 
979 	trace_otx2_msg_interrupt(mbox->mbox.pdev, "AF to PF", BIT_ULL(0));
980 
981 	otx2_queue_work(mbox, pf->mbox_wq, 0, 1, 1, TYPE_PFAF);
982 
983 	return IRQ_HANDLED;
984 }
985 
986 static void otx2_disable_mbox_intr(struct otx2_nic *pf)
987 {
988 	int vector = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_AFPF_MBOX);
989 
990 	/* Disable AF => PF mailbox IRQ */
991 	otx2_write64(pf, RVU_PF_INT_ENA_W1C, BIT_ULL(0));
992 	free_irq(vector, pf);
993 }
994 
995 static int otx2_register_mbox_intr(struct otx2_nic *pf, bool probe_af)
996 {
997 	struct otx2_hw *hw = &pf->hw;
998 	struct msg_req *req;
999 	char *irq_name;
1000 	int err;
1001 
1002 	/* Register mailbox interrupt handler */
1003 	irq_name = &hw->irq_name[RVU_PF_INT_VEC_AFPF_MBOX * NAME_SIZE];
1004 	snprintf(irq_name, NAME_SIZE, "RVUPFAF Mbox");
1005 	err = request_irq(pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_AFPF_MBOX),
1006 			  otx2_pfaf_mbox_intr_handler, 0, irq_name, pf);
1007 	if (err) {
1008 		dev_err(pf->dev,
1009 			"RVUPF: IRQ registration failed for PFAF mbox irq\n");
1010 		return err;
1011 	}
1012 
1013 	/* Enable mailbox interrupt for msgs coming from AF.
1014 	 * First clear to avoid spurious interrupts, if any.
1015 	 */
1016 	otx2_write64(pf, RVU_PF_INT, BIT_ULL(0));
1017 	otx2_write64(pf, RVU_PF_INT_ENA_W1S, BIT_ULL(0));
1018 
1019 	if (!probe_af)
1020 		return 0;
1021 
1022 	/* Check mailbox communication with AF */
1023 	req = otx2_mbox_alloc_msg_ready(&pf->mbox);
1024 	if (!req) {
1025 		otx2_disable_mbox_intr(pf);
1026 		return -ENOMEM;
1027 	}
1028 	err = otx2_sync_mbox_msg(&pf->mbox);
1029 	if (err) {
1030 		dev_warn(pf->dev,
1031 			 "AF not responding to mailbox, deferring probe\n");
1032 		otx2_disable_mbox_intr(pf);
1033 		return -EPROBE_DEFER;
1034 	}
1035 
1036 	return 0;
1037 }
1038 
1039 static void otx2_pfaf_mbox_destroy(struct otx2_nic *pf)
1040 {
1041 	struct mbox *mbox = &pf->mbox;
1042 
1043 	if (pf->mbox_wq) {
1044 		destroy_workqueue(pf->mbox_wq);
1045 		pf->mbox_wq = NULL;
1046 	}
1047 
1048 	if (mbox->mbox.hwbase)
1049 		iounmap((void __iomem *)mbox->mbox.hwbase);
1050 
1051 	otx2_mbox_destroy(&mbox->mbox);
1052 	otx2_mbox_destroy(&mbox->mbox_up);
1053 }
1054 
1055 static int otx2_pfaf_mbox_init(struct otx2_nic *pf)
1056 {
1057 	struct mbox *mbox = &pf->mbox;
1058 	void __iomem *hwbase;
1059 	int err;
1060 
1061 	mbox->pfvf = pf;
1062 	pf->mbox_wq = alloc_ordered_workqueue("otx2_pfaf_mailbox",
1063 					      WQ_HIGHPRI | WQ_MEM_RECLAIM);
1064 	if (!pf->mbox_wq)
1065 		return -ENOMEM;
1066 
1067 	/* Mailbox is a reserved memory (in RAM) region shared between
1068 	 * admin function (i.e AF) and this PF, shouldn't be mapped as
1069 	 * device memory to allow unaligned accesses.
1070 	 */
1071 	hwbase = ioremap_wc(pci_resource_start(pf->pdev, PCI_MBOX_BAR_NUM),
1072 			    MBOX_SIZE);
1073 	if (!hwbase) {
1074 		dev_err(pf->dev, "Unable to map PFAF mailbox region\n");
1075 		err = -ENOMEM;
1076 		goto exit;
1077 	}
1078 
1079 	err = otx2_mbox_init(&mbox->mbox, hwbase, pf->pdev, pf->reg_base,
1080 			     MBOX_DIR_PFAF, 1);
1081 	if (err)
1082 		goto exit;
1083 
1084 	err = otx2_mbox_init(&mbox->mbox_up, hwbase, pf->pdev, pf->reg_base,
1085 			     MBOX_DIR_PFAF_UP, 1);
1086 	if (err)
1087 		goto exit;
1088 
1089 	err = otx2_mbox_bbuf_init(mbox, pf->pdev);
1090 	if (err)
1091 		goto exit;
1092 
1093 	INIT_WORK(&mbox->mbox_wrk, otx2_pfaf_mbox_handler);
1094 	INIT_WORK(&mbox->mbox_up_wrk, otx2_pfaf_mbox_up_handler);
1095 	mutex_init(&mbox->lock);
1096 
1097 	return 0;
1098 exit:
1099 	otx2_pfaf_mbox_destroy(pf);
1100 	return err;
1101 }
1102 
1103 static int otx2_cgx_config_linkevents(struct otx2_nic *pf, bool enable)
1104 {
1105 	struct msg_req *msg;
1106 	int err;
1107 
1108 	mutex_lock(&pf->mbox.lock);
1109 	if (enable)
1110 		msg = otx2_mbox_alloc_msg_cgx_start_linkevents(&pf->mbox);
1111 	else
1112 		msg = otx2_mbox_alloc_msg_cgx_stop_linkevents(&pf->mbox);
1113 
1114 	if (!msg) {
1115 		mutex_unlock(&pf->mbox.lock);
1116 		return -ENOMEM;
1117 	}
1118 
1119 	err = otx2_sync_mbox_msg(&pf->mbox);
1120 	mutex_unlock(&pf->mbox.lock);
1121 	return err;
1122 }
1123 
1124 static int otx2_cgx_config_loopback(struct otx2_nic *pf, bool enable)
1125 {
1126 	struct msg_req *msg;
1127 	int err;
1128 
1129 	if (enable && !bitmap_empty(pf->flow_cfg->dmacflt_bmap,
1130 				    pf->flow_cfg->dmacflt_max_flows))
1131 		netdev_warn(pf->netdev,
1132 			    "CGX/RPM internal loopback might not work as DMAC filters are active\n");
1133 
1134 	mutex_lock(&pf->mbox.lock);
1135 	if (enable)
1136 		msg = otx2_mbox_alloc_msg_cgx_intlbk_enable(&pf->mbox);
1137 	else
1138 		msg = otx2_mbox_alloc_msg_cgx_intlbk_disable(&pf->mbox);
1139 
1140 	if (!msg) {
1141 		mutex_unlock(&pf->mbox.lock);
1142 		return -ENOMEM;
1143 	}
1144 
1145 	err = otx2_sync_mbox_msg(&pf->mbox);
1146 	mutex_unlock(&pf->mbox.lock);
1147 	return err;
1148 }
1149 
1150 int otx2_set_real_num_queues(struct net_device *netdev,
1151 			     int tx_queues, int rx_queues)
1152 {
1153 	int err;
1154 
1155 	err = netif_set_real_num_tx_queues(netdev, tx_queues);
1156 	if (err) {
1157 		netdev_err(netdev,
1158 			   "Failed to set no of Tx queues: %d\n", tx_queues);
1159 		return err;
1160 	}
1161 
1162 	err = netif_set_real_num_rx_queues(netdev, rx_queues);
1163 	if (err)
1164 		netdev_err(netdev,
1165 			   "Failed to set no of Rx queues: %d\n", rx_queues);
1166 	return err;
1167 }
1168 EXPORT_SYMBOL(otx2_set_real_num_queues);
1169 
1170 static char *nix_sqoperr_e_str[NIX_SQOPERR_MAX] = {
1171 	"NIX_SQOPERR_OOR",
1172 	"NIX_SQOPERR_CTX_FAULT",
1173 	"NIX_SQOPERR_CTX_POISON",
1174 	"NIX_SQOPERR_DISABLED",
1175 	"NIX_SQOPERR_SIZE_ERR",
1176 	"NIX_SQOPERR_OFLOW",
1177 	"NIX_SQOPERR_SQB_NULL",
1178 	"NIX_SQOPERR_SQB_FAULT",
1179 	"NIX_SQOPERR_SQE_SZ_ZERO",
1180 };
1181 
1182 static char *nix_mnqerr_e_str[NIX_MNQERR_MAX] = {
1183 	"NIX_MNQERR_SQ_CTX_FAULT",
1184 	"NIX_MNQERR_SQ_CTX_POISON",
1185 	"NIX_MNQERR_SQB_FAULT",
1186 	"NIX_MNQERR_SQB_POISON",
1187 	"NIX_MNQERR_TOTAL_ERR",
1188 	"NIX_MNQERR_LSO_ERR",
1189 	"NIX_MNQERR_CQ_QUERY_ERR",
1190 	"NIX_MNQERR_MAX_SQE_SIZE_ERR",
1191 	"NIX_MNQERR_MAXLEN_ERR",
1192 	"NIX_MNQERR_SQE_SIZEM1_ZERO",
1193 };
1194 
1195 static char *nix_snd_status_e_str[NIX_SND_STATUS_MAX] =  {
1196 	"NIX_SND_STATUS_GOOD",
1197 	"NIX_SND_STATUS_SQ_CTX_FAULT",
1198 	"NIX_SND_STATUS_SQ_CTX_POISON",
1199 	"NIX_SND_STATUS_SQB_FAULT",
1200 	"NIX_SND_STATUS_SQB_POISON",
1201 	"NIX_SND_STATUS_HDR_ERR",
1202 	"NIX_SND_STATUS_EXT_ERR",
1203 	"NIX_SND_STATUS_JUMP_FAULT",
1204 	"NIX_SND_STATUS_JUMP_POISON",
1205 	"NIX_SND_STATUS_CRC_ERR",
1206 	"NIX_SND_STATUS_IMM_ERR",
1207 	"NIX_SND_STATUS_SG_ERR",
1208 	"NIX_SND_STATUS_MEM_ERR",
1209 	"NIX_SND_STATUS_INVALID_SUBDC",
1210 	"NIX_SND_STATUS_SUBDC_ORDER_ERR",
1211 	"NIX_SND_STATUS_DATA_FAULT",
1212 	"NIX_SND_STATUS_DATA_POISON",
1213 	"NIX_SND_STATUS_NPC_DROP_ACTION",
1214 	"NIX_SND_STATUS_LOCK_VIOL",
1215 	"NIX_SND_STATUS_NPC_UCAST_CHAN_ERR",
1216 	"NIX_SND_STATUS_NPC_MCAST_CHAN_ERR",
1217 	"NIX_SND_STATUS_NPC_MCAST_ABORT",
1218 	"NIX_SND_STATUS_NPC_VTAG_PTR_ERR",
1219 	"NIX_SND_STATUS_NPC_VTAG_SIZE_ERR",
1220 	"NIX_SND_STATUS_SEND_STATS_ERR",
1221 };
1222 
1223 static irqreturn_t otx2_q_intr_handler(int irq, void *data)
1224 {
1225 	struct otx2_nic *pf = data;
1226 	struct otx2_snd_queue *sq;
1227 	u64 val, *ptr;
1228 	u64 qidx = 0;
1229 
1230 	/* CQ */
1231 	for (qidx = 0; qidx < pf->qset.cq_cnt; qidx++) {
1232 		ptr = otx2_get_regaddr(pf, NIX_LF_CQ_OP_INT);
1233 		val = otx2_atomic64_add((qidx << 44), ptr);
1234 
1235 		otx2_write64(pf, NIX_LF_CQ_OP_INT, (qidx << 44) |
1236 			     (val & NIX_CQERRINT_BITS));
1237 		if (!(val & (NIX_CQERRINT_BITS | BIT_ULL(42))))
1238 			continue;
1239 
1240 		if (val & BIT_ULL(42)) {
1241 			netdev_err(pf->netdev, "CQ%lld: error reading NIX_LF_CQ_OP_INT, NIX_LF_ERR_INT 0x%llx\n",
1242 				   qidx, otx2_read64(pf, NIX_LF_ERR_INT));
1243 		} else {
1244 			if (val & BIT_ULL(NIX_CQERRINT_DOOR_ERR))
1245 				netdev_err(pf->netdev, "CQ%lld: Doorbell error",
1246 					   qidx);
1247 			if (val & BIT_ULL(NIX_CQERRINT_CQE_FAULT))
1248 				netdev_err(pf->netdev, "CQ%lld: Memory fault on CQE write to LLC/DRAM",
1249 					   qidx);
1250 		}
1251 
1252 		schedule_work(&pf->reset_task);
1253 	}
1254 
1255 	/* SQ */
1256 	for (qidx = 0; qidx < otx2_get_total_tx_queues(pf); qidx++) {
1257 		u64 sq_op_err_dbg, mnq_err_dbg, snd_err_dbg;
1258 		u8 sq_op_err_code, mnq_err_code, snd_err_code;
1259 
1260 		sq = &pf->qset.sq[qidx];
1261 		if (!sq->sqb_ptrs)
1262 			continue;
1263 
1264 		/* Below debug registers captures first errors corresponding to
1265 		 * those registers. We don't have to check against SQ qid as
1266 		 * these are fatal errors.
1267 		 */
1268 
1269 		ptr = otx2_get_regaddr(pf, NIX_LF_SQ_OP_INT);
1270 		val = otx2_atomic64_add((qidx << 44), ptr);
1271 		otx2_write64(pf, NIX_LF_SQ_OP_INT, (qidx << 44) |
1272 			     (val & NIX_SQINT_BITS));
1273 
1274 		if (val & BIT_ULL(42)) {
1275 			netdev_err(pf->netdev, "SQ%lld: error reading NIX_LF_SQ_OP_INT, NIX_LF_ERR_INT 0x%llx\n",
1276 				   qidx, otx2_read64(pf, NIX_LF_ERR_INT));
1277 			goto done;
1278 		}
1279 
1280 		sq_op_err_dbg = otx2_read64(pf, NIX_LF_SQ_OP_ERR_DBG);
1281 		if (!(sq_op_err_dbg & BIT(44)))
1282 			goto chk_mnq_err_dbg;
1283 
1284 		sq_op_err_code = FIELD_GET(GENMASK(7, 0), sq_op_err_dbg);
1285 		netdev_err(pf->netdev, "SQ%lld: NIX_LF_SQ_OP_ERR_DBG(%llx)  err=%s\n",
1286 			   qidx, sq_op_err_dbg, nix_sqoperr_e_str[sq_op_err_code]);
1287 
1288 		otx2_write64(pf, NIX_LF_SQ_OP_ERR_DBG, BIT_ULL(44));
1289 
1290 		if (sq_op_err_code == NIX_SQOPERR_SQB_NULL)
1291 			goto chk_mnq_err_dbg;
1292 
1293 		/* Err is not NIX_SQOPERR_SQB_NULL, call aq function to read SQ structure.
1294 		 * TODO: But we are in irq context. How to call mbox functions which does sleep
1295 		 */
1296 
1297 chk_mnq_err_dbg:
1298 		mnq_err_dbg = otx2_read64(pf, NIX_LF_MNQ_ERR_DBG);
1299 		if (!(mnq_err_dbg & BIT(44)))
1300 			goto chk_snd_err_dbg;
1301 
1302 		mnq_err_code = FIELD_GET(GENMASK(7, 0), mnq_err_dbg);
1303 		netdev_err(pf->netdev, "SQ%lld: NIX_LF_MNQ_ERR_DBG(%llx)  err=%s\n",
1304 			   qidx, mnq_err_dbg,  nix_mnqerr_e_str[mnq_err_code]);
1305 		otx2_write64(pf, NIX_LF_MNQ_ERR_DBG, BIT_ULL(44));
1306 
1307 chk_snd_err_dbg:
1308 		snd_err_dbg = otx2_read64(pf, NIX_LF_SEND_ERR_DBG);
1309 		if (snd_err_dbg & BIT(44)) {
1310 			snd_err_code = FIELD_GET(GENMASK(7, 0), snd_err_dbg);
1311 			netdev_err(pf->netdev, "SQ%lld: NIX_LF_SND_ERR_DBG:0x%llx err=%s\n",
1312 				   qidx, snd_err_dbg, nix_snd_status_e_str[snd_err_code]);
1313 			otx2_write64(pf, NIX_LF_SEND_ERR_DBG, BIT_ULL(44));
1314 		}
1315 
1316 done:
1317 		/* Print values and reset */
1318 		if (val & BIT_ULL(NIX_SQINT_SQB_ALLOC_FAIL))
1319 			netdev_err(pf->netdev, "SQ%lld: SQB allocation failed",
1320 				   qidx);
1321 
1322 		schedule_work(&pf->reset_task);
1323 	}
1324 
1325 	return IRQ_HANDLED;
1326 }
1327 
1328 static irqreturn_t otx2_cq_intr_handler(int irq, void *cq_irq)
1329 {
1330 	struct otx2_cq_poll *cq_poll = (struct otx2_cq_poll *)cq_irq;
1331 	struct otx2_nic *pf = (struct otx2_nic *)cq_poll->dev;
1332 	int qidx = cq_poll->cint_idx;
1333 
1334 	/* Disable interrupts.
1335 	 *
1336 	 * Completion interrupts behave in a level-triggered interrupt
1337 	 * fashion, and hence have to be cleared only after it is serviced.
1338 	 */
1339 	otx2_write64(pf, NIX_LF_CINTX_ENA_W1C(qidx), BIT_ULL(0));
1340 
1341 	/* Schedule NAPI */
1342 	pf->napi_events++;
1343 	napi_schedule_irqoff(&cq_poll->napi);
1344 
1345 	return IRQ_HANDLED;
1346 }
1347 
1348 static void otx2_disable_napi(struct otx2_nic *pf)
1349 {
1350 	struct otx2_qset *qset = &pf->qset;
1351 	struct otx2_cq_poll *cq_poll;
1352 	int qidx;
1353 
1354 	for (qidx = 0; qidx < pf->hw.cint_cnt; qidx++) {
1355 		cq_poll = &qset->napi[qidx];
1356 		cancel_work_sync(&cq_poll->dim.work);
1357 		napi_disable(&cq_poll->napi);
1358 		netif_napi_del(&cq_poll->napi);
1359 	}
1360 }
1361 
1362 static void otx2_free_cq_res(struct otx2_nic *pf)
1363 {
1364 	struct otx2_qset *qset = &pf->qset;
1365 	struct otx2_cq_queue *cq;
1366 	int qidx;
1367 
1368 	/* Disable CQs */
1369 	otx2_ctx_disable(&pf->mbox, NIX_AQ_CTYPE_CQ, false);
1370 	for (qidx = 0; qidx < qset->cq_cnt; qidx++) {
1371 		cq = &qset->cq[qidx];
1372 		qmem_free(pf->dev, cq->cqe);
1373 	}
1374 }
1375 
1376 static void otx2_free_sq_res(struct otx2_nic *pf)
1377 {
1378 	struct otx2_qset *qset = &pf->qset;
1379 	struct otx2_snd_queue *sq;
1380 	int qidx;
1381 
1382 	/* Disable SQs */
1383 	otx2_ctx_disable(&pf->mbox, NIX_AQ_CTYPE_SQ, false);
1384 	/* Free SQB pointers */
1385 	otx2_sq_free_sqbs(pf);
1386 	for (qidx = 0; qidx < otx2_get_total_tx_queues(pf); qidx++) {
1387 		sq = &qset->sq[qidx];
1388 		/* Skip freeing Qos queues if they are not initialized */
1389 		if (!sq->sqe)
1390 			continue;
1391 		qmem_free(pf->dev, sq->sqe);
1392 		qmem_free(pf->dev, sq->tso_hdrs);
1393 		kfree(sq->sg);
1394 		kfree(sq->sqb_ptrs);
1395 	}
1396 }
1397 
1398 static int otx2_get_rbuf_size(struct otx2_nic *pf, int mtu)
1399 {
1400 	int frame_size;
1401 	int total_size;
1402 	int rbuf_size;
1403 
1404 	if (pf->hw.rbuf_len)
1405 		return ALIGN(pf->hw.rbuf_len, OTX2_ALIGN) + OTX2_HEAD_ROOM;
1406 
1407 	/* The data transferred by NIX to memory consists of actual packet
1408 	 * plus additional data which has timestamp and/or EDSA/HIGIG2
1409 	 * headers if interface is configured in corresponding modes.
1410 	 * NIX transfers entire data using 6 segments/buffers and writes
1411 	 * a CQE_RX descriptor with those segment addresses. First segment
1412 	 * has additional data prepended to packet. Also software omits a
1413 	 * headroom of 128 bytes in each segment. Hence the total size of
1414 	 * memory needed to receive a packet with 'mtu' is:
1415 	 * frame size =  mtu + additional data;
1416 	 * memory = frame_size + headroom * 6;
1417 	 * each receive buffer size = memory / 6;
1418 	 */
1419 	frame_size = mtu + OTX2_ETH_HLEN + OTX2_HW_TIMESTAMP_LEN;
1420 	total_size = frame_size + OTX2_HEAD_ROOM * 6;
1421 	rbuf_size = total_size / 6;
1422 
1423 	return ALIGN(rbuf_size, 2048);
1424 }
1425 
1426 static int otx2_init_hw_resources(struct otx2_nic *pf)
1427 {
1428 	struct nix_lf_free_req *free_req;
1429 	struct mbox *mbox = &pf->mbox;
1430 	struct otx2_hw *hw = &pf->hw;
1431 	struct msg_req *req;
1432 	int err = 0, lvl;
1433 
1434 	/* Set required NPA LF's pool counts
1435 	 * Auras and Pools are used in a 1:1 mapping,
1436 	 * so, aura count = pool count.
1437 	 */
1438 	hw->rqpool_cnt = hw->rx_queues;
1439 	hw->sqpool_cnt = otx2_get_total_tx_queues(pf);
1440 	hw->pool_cnt = hw->rqpool_cnt + hw->sqpool_cnt;
1441 
1442 	/* Maximum hardware supported transmit length */
1443 	pf->tx_max_pktlen = pf->netdev->max_mtu + OTX2_ETH_HLEN;
1444 
1445 	pf->rbsize = otx2_get_rbuf_size(pf, pf->netdev->mtu);
1446 
1447 	mutex_lock(&mbox->lock);
1448 	/* NPA init */
1449 	err = otx2_config_npa(pf);
1450 	if (err)
1451 		goto exit;
1452 
1453 	/* NIX init */
1454 	err = otx2_config_nix(pf);
1455 	if (err)
1456 		goto err_free_npa_lf;
1457 
1458 	/* Enable backpressure for CGX mapped PF/VFs */
1459 	if (!is_otx2_lbkvf(pf->pdev))
1460 		otx2_nix_config_bp(pf, true);
1461 
1462 	/* Init Auras and pools used by NIX RQ, for free buffer ptrs */
1463 	err = otx2_rq_aura_pool_init(pf);
1464 	if (err) {
1465 		mutex_unlock(&mbox->lock);
1466 		goto err_free_nix_lf;
1467 	}
1468 	/* Init Auras and pools used by NIX SQ, for queueing SQEs */
1469 	err = otx2_sq_aura_pool_init(pf);
1470 	if (err) {
1471 		mutex_unlock(&mbox->lock);
1472 		goto err_free_rq_ptrs;
1473 	}
1474 
1475 	err = otx2_txsch_alloc(pf);
1476 	if (err) {
1477 		mutex_unlock(&mbox->lock);
1478 		goto err_free_sq_ptrs;
1479 	}
1480 
1481 #ifdef CONFIG_DCB
1482 	if (pf->pfc_en) {
1483 		err = otx2_pfc_txschq_alloc(pf);
1484 		if (err) {
1485 			mutex_unlock(&mbox->lock);
1486 			goto err_free_sq_ptrs;
1487 		}
1488 	}
1489 #endif
1490 
1491 	err = otx2_config_nix_queues(pf);
1492 	if (err) {
1493 		mutex_unlock(&mbox->lock);
1494 		goto err_free_txsch;
1495 	}
1496 
1497 	for (lvl = 0; lvl < NIX_TXSCH_LVL_CNT; lvl++) {
1498 		err = otx2_txschq_config(pf, lvl, 0, false);
1499 		if (err) {
1500 			mutex_unlock(&mbox->lock);
1501 			goto err_free_nix_queues;
1502 		}
1503 	}
1504 
1505 #ifdef CONFIG_DCB
1506 	if (pf->pfc_en) {
1507 		err = otx2_pfc_txschq_config(pf);
1508 		if (err) {
1509 			mutex_unlock(&mbox->lock);
1510 			goto err_free_nix_queues;
1511 		}
1512 	}
1513 #endif
1514 
1515 	mutex_unlock(&mbox->lock);
1516 	return err;
1517 
1518 err_free_nix_queues:
1519 	otx2_free_sq_res(pf);
1520 	otx2_free_cq_res(pf);
1521 	otx2_ctx_disable(mbox, NIX_AQ_CTYPE_RQ, false);
1522 err_free_txsch:
1523 	otx2_txschq_stop(pf);
1524 err_free_sq_ptrs:
1525 	otx2_sq_free_sqbs(pf);
1526 err_free_rq_ptrs:
1527 	otx2_free_aura_ptr(pf, AURA_NIX_RQ);
1528 	otx2_ctx_disable(mbox, NPA_AQ_CTYPE_POOL, true);
1529 	otx2_ctx_disable(mbox, NPA_AQ_CTYPE_AURA, true);
1530 	otx2_aura_pool_free(pf);
1531 err_free_nix_lf:
1532 	mutex_lock(&mbox->lock);
1533 	free_req = otx2_mbox_alloc_msg_nix_lf_free(mbox);
1534 	if (free_req) {
1535 		free_req->flags = NIX_LF_DISABLE_FLOWS;
1536 		if (otx2_sync_mbox_msg(mbox))
1537 			dev_err(pf->dev, "%s failed to free nixlf\n", __func__);
1538 	}
1539 err_free_npa_lf:
1540 	/* Reset NPA LF */
1541 	req = otx2_mbox_alloc_msg_npa_lf_free(mbox);
1542 	if (req) {
1543 		if (otx2_sync_mbox_msg(mbox))
1544 			dev_err(pf->dev, "%s failed to free npalf\n", __func__);
1545 	}
1546 exit:
1547 	mutex_unlock(&mbox->lock);
1548 	return err;
1549 }
1550 
1551 static void otx2_free_hw_resources(struct otx2_nic *pf)
1552 {
1553 	struct otx2_qset *qset = &pf->qset;
1554 	struct nix_lf_free_req *free_req;
1555 	struct mbox *mbox = &pf->mbox;
1556 	struct otx2_cq_queue *cq;
1557 	struct otx2_pool *pool;
1558 	struct msg_req *req;
1559 	int pool_id;
1560 	int qidx;
1561 
1562 	/* Ensure all SQE are processed */
1563 	otx2_sqb_flush(pf);
1564 
1565 	/* Stop transmission */
1566 	otx2_txschq_stop(pf);
1567 
1568 #ifdef CONFIG_DCB
1569 	if (pf->pfc_en)
1570 		otx2_pfc_txschq_stop(pf);
1571 #endif
1572 
1573 	otx2_clean_qos_queues(pf);
1574 
1575 	mutex_lock(&mbox->lock);
1576 	/* Disable backpressure */
1577 	if (!(pf->pcifunc & RVU_PFVF_FUNC_MASK))
1578 		otx2_nix_config_bp(pf, false);
1579 	mutex_unlock(&mbox->lock);
1580 
1581 	/* Disable RQs */
1582 	otx2_ctx_disable(mbox, NIX_AQ_CTYPE_RQ, false);
1583 
1584 	/*Dequeue all CQEs */
1585 	for (qidx = 0; qidx < qset->cq_cnt; qidx++) {
1586 		cq = &qset->cq[qidx];
1587 		if (cq->cq_type == CQ_RX)
1588 			otx2_cleanup_rx_cqes(pf, cq, qidx);
1589 		else
1590 			otx2_cleanup_tx_cqes(pf, cq);
1591 	}
1592 
1593 	otx2_free_sq_res(pf);
1594 
1595 	/* Free RQ buffer pointers*/
1596 	otx2_free_aura_ptr(pf, AURA_NIX_RQ);
1597 
1598 	for (qidx = 0; qidx < pf->hw.rx_queues; qidx++) {
1599 		pool_id = otx2_get_pool_idx(pf, AURA_NIX_RQ, qidx);
1600 		pool = &pf->qset.pool[pool_id];
1601 		page_pool_destroy(pool->page_pool);
1602 		pool->page_pool = NULL;
1603 	}
1604 
1605 	otx2_free_cq_res(pf);
1606 
1607 	/* Free all ingress bandwidth profiles allocated */
1608 	cn10k_free_all_ipolicers(pf);
1609 
1610 	mutex_lock(&mbox->lock);
1611 	/* Reset NIX LF */
1612 	free_req = otx2_mbox_alloc_msg_nix_lf_free(mbox);
1613 	if (free_req) {
1614 		free_req->flags = NIX_LF_DISABLE_FLOWS;
1615 		if (!(pf->flags & OTX2_FLAG_PF_SHUTDOWN))
1616 			free_req->flags |= NIX_LF_DONT_FREE_TX_VTAG;
1617 		if (otx2_sync_mbox_msg(mbox))
1618 			dev_err(pf->dev, "%s failed to free nixlf\n", __func__);
1619 	}
1620 	mutex_unlock(&mbox->lock);
1621 
1622 	/* Disable NPA Pool and Aura hw context */
1623 	otx2_ctx_disable(mbox, NPA_AQ_CTYPE_POOL, true);
1624 	otx2_ctx_disable(mbox, NPA_AQ_CTYPE_AURA, true);
1625 	otx2_aura_pool_free(pf);
1626 
1627 	mutex_lock(&mbox->lock);
1628 	/* Reset NPA LF */
1629 	req = otx2_mbox_alloc_msg_npa_lf_free(mbox);
1630 	if (req) {
1631 		if (otx2_sync_mbox_msg(mbox))
1632 			dev_err(pf->dev, "%s failed to free npalf\n", __func__);
1633 	}
1634 	mutex_unlock(&mbox->lock);
1635 }
1636 
1637 static void otx2_do_set_rx_mode(struct otx2_nic *pf)
1638 {
1639 	struct net_device *netdev = pf->netdev;
1640 	struct nix_rx_mode *req;
1641 	bool promisc = false;
1642 
1643 	if (!(netdev->flags & IFF_UP))
1644 		return;
1645 
1646 	if ((netdev->flags & IFF_PROMISC) ||
1647 	    (netdev_uc_count(netdev) > OTX2_MAX_UNICAST_FLOWS)) {
1648 		promisc = true;
1649 	}
1650 
1651 	/* Write unicast address to mcam entries or del from mcam */
1652 	if (!promisc && netdev->priv_flags & IFF_UNICAST_FLT)
1653 		__dev_uc_sync(netdev, otx2_add_macfilter, otx2_del_macfilter);
1654 
1655 	mutex_lock(&pf->mbox.lock);
1656 	req = otx2_mbox_alloc_msg_nix_set_rx_mode(&pf->mbox);
1657 	if (!req) {
1658 		mutex_unlock(&pf->mbox.lock);
1659 		return;
1660 	}
1661 
1662 	req->mode = NIX_RX_MODE_UCAST;
1663 
1664 	if (promisc)
1665 		req->mode |= NIX_RX_MODE_PROMISC;
1666 	if (netdev->flags & (IFF_ALLMULTI | IFF_MULTICAST))
1667 		req->mode |= NIX_RX_MODE_ALLMULTI;
1668 
1669 	req->mode |= NIX_RX_MODE_USE_MCE;
1670 
1671 	otx2_sync_mbox_msg(&pf->mbox);
1672 	mutex_unlock(&pf->mbox.lock);
1673 }
1674 
1675 static void otx2_dim_work(struct work_struct *w)
1676 {
1677 	struct dim_cq_moder cur_moder;
1678 	struct otx2_cq_poll *cq_poll;
1679 	struct otx2_nic *pfvf;
1680 	struct dim *dim;
1681 
1682 	dim = container_of(w, struct dim, work);
1683 	cur_moder = net_dim_get_rx_moderation(dim->mode, dim->profile_ix);
1684 	cq_poll = container_of(dim, struct otx2_cq_poll, dim);
1685 	pfvf = (struct otx2_nic *)cq_poll->dev;
1686 	pfvf->hw.cq_time_wait = (cur_moder.usec > CQ_TIMER_THRESH_MAX) ?
1687 		CQ_TIMER_THRESH_MAX : cur_moder.usec;
1688 	pfvf->hw.cq_ecount_wait = (cur_moder.pkts > NAPI_POLL_WEIGHT) ?
1689 		NAPI_POLL_WEIGHT : cur_moder.pkts;
1690 	dim->state = DIM_START_MEASURE;
1691 }
1692 
1693 int otx2_open(struct net_device *netdev)
1694 {
1695 	struct otx2_nic *pf = netdev_priv(netdev);
1696 	struct otx2_cq_poll *cq_poll = NULL;
1697 	struct otx2_qset *qset = &pf->qset;
1698 	int err = 0, qidx, vec;
1699 	char *irq_name;
1700 
1701 	netif_carrier_off(netdev);
1702 
1703 	/* RQ and SQs are mapped to different CQs,
1704 	 * so find out max CQ IRQs (i.e CINTs) needed.
1705 	 */
1706 	pf->hw.cint_cnt = max3(pf->hw.rx_queues, pf->hw.tx_queues,
1707 			       pf->hw.tc_tx_queues);
1708 
1709 	pf->qset.cq_cnt = pf->hw.rx_queues + otx2_get_total_tx_queues(pf);
1710 
1711 	qset->napi = kcalloc(pf->hw.cint_cnt, sizeof(*cq_poll), GFP_KERNEL);
1712 	if (!qset->napi)
1713 		return -ENOMEM;
1714 
1715 	/* CQ size of RQ */
1716 	qset->rqe_cnt = qset->rqe_cnt ? qset->rqe_cnt : Q_COUNT(Q_SIZE_256);
1717 	/* CQ size of SQ */
1718 	qset->sqe_cnt = qset->sqe_cnt ? qset->sqe_cnt : Q_COUNT(Q_SIZE_4K);
1719 
1720 	err = -ENOMEM;
1721 	qset->cq = kcalloc(pf->qset.cq_cnt,
1722 			   sizeof(struct otx2_cq_queue), GFP_KERNEL);
1723 	if (!qset->cq)
1724 		goto err_free_mem;
1725 
1726 	qset->sq = kcalloc(otx2_get_total_tx_queues(pf),
1727 			   sizeof(struct otx2_snd_queue), GFP_KERNEL);
1728 	if (!qset->sq)
1729 		goto err_free_mem;
1730 
1731 	qset->rq = kcalloc(pf->hw.rx_queues,
1732 			   sizeof(struct otx2_rcv_queue), GFP_KERNEL);
1733 	if (!qset->rq)
1734 		goto err_free_mem;
1735 
1736 	err = otx2_init_hw_resources(pf);
1737 	if (err)
1738 		goto err_free_mem;
1739 
1740 	/* Register NAPI handler */
1741 	for (qidx = 0; qidx < pf->hw.cint_cnt; qidx++) {
1742 		cq_poll = &qset->napi[qidx];
1743 		cq_poll->cint_idx = qidx;
1744 		/* RQ0 & SQ0 are mapped to CINT0 and so on..
1745 		 * 'cq_ids[0]' points to RQ's CQ and
1746 		 * 'cq_ids[1]' points to SQ's CQ and
1747 		 * 'cq_ids[2]' points to XDP's CQ and
1748 		 */
1749 		cq_poll->cq_ids[CQ_RX] =
1750 			(qidx <  pf->hw.rx_queues) ? qidx : CINT_INVALID_CQ;
1751 		cq_poll->cq_ids[CQ_TX] = (qidx < pf->hw.tx_queues) ?
1752 				      qidx + pf->hw.rx_queues : CINT_INVALID_CQ;
1753 		if (pf->xdp_prog)
1754 			cq_poll->cq_ids[CQ_XDP] = (qidx < pf->hw.xdp_queues) ?
1755 						  (qidx + pf->hw.rx_queues +
1756 						  pf->hw.tx_queues) :
1757 						  CINT_INVALID_CQ;
1758 		else
1759 			cq_poll->cq_ids[CQ_XDP] = CINT_INVALID_CQ;
1760 
1761 		cq_poll->cq_ids[CQ_QOS] = (qidx < pf->hw.tc_tx_queues) ?
1762 					  (qidx + pf->hw.rx_queues +
1763 					   pf->hw.non_qos_queues) :
1764 					  CINT_INVALID_CQ;
1765 
1766 		cq_poll->dev = (void *)pf;
1767 		cq_poll->dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_CQE;
1768 		INIT_WORK(&cq_poll->dim.work, otx2_dim_work);
1769 		netif_napi_add(netdev, &cq_poll->napi, otx2_napi_handler);
1770 		napi_enable(&cq_poll->napi);
1771 	}
1772 
1773 	/* Set maximum frame size allowed in HW */
1774 	err = otx2_hw_set_mtu(pf, netdev->mtu);
1775 	if (err)
1776 		goto err_disable_napi;
1777 
1778 	/* Setup segmentation algorithms, if failed, clear offload capability */
1779 	otx2_setup_segmentation(pf);
1780 
1781 	/* Initialize RSS */
1782 	err = otx2_rss_init(pf);
1783 	if (err)
1784 		goto err_disable_napi;
1785 
1786 	/* Register Queue IRQ handlers */
1787 	vec = pf->hw.nix_msixoff + NIX_LF_QINT_VEC_START;
1788 	irq_name = &pf->hw.irq_name[vec * NAME_SIZE];
1789 
1790 	snprintf(irq_name, NAME_SIZE, "%s-qerr", pf->netdev->name);
1791 
1792 	err = request_irq(pci_irq_vector(pf->pdev, vec),
1793 			  otx2_q_intr_handler, 0, irq_name, pf);
1794 	if (err) {
1795 		dev_err(pf->dev,
1796 			"RVUPF%d: IRQ registration failed for QERR\n",
1797 			rvu_get_pf(pf->pcifunc));
1798 		goto err_disable_napi;
1799 	}
1800 
1801 	/* Enable QINT IRQ */
1802 	otx2_write64(pf, NIX_LF_QINTX_ENA_W1S(0), BIT_ULL(0));
1803 
1804 	/* Register CQ IRQ handlers */
1805 	vec = pf->hw.nix_msixoff + NIX_LF_CINT_VEC_START;
1806 	for (qidx = 0; qidx < pf->hw.cint_cnt; qidx++) {
1807 		irq_name = &pf->hw.irq_name[vec * NAME_SIZE];
1808 
1809 		snprintf(irq_name, NAME_SIZE, "%s-rxtx-%d", pf->netdev->name,
1810 			 qidx);
1811 
1812 		err = request_irq(pci_irq_vector(pf->pdev, vec),
1813 				  otx2_cq_intr_handler, 0, irq_name,
1814 				  &qset->napi[qidx]);
1815 		if (err) {
1816 			dev_err(pf->dev,
1817 				"RVUPF%d: IRQ registration failed for CQ%d\n",
1818 				rvu_get_pf(pf->pcifunc), qidx);
1819 			goto err_free_cints;
1820 		}
1821 		vec++;
1822 
1823 		otx2_config_irq_coalescing(pf, qidx);
1824 
1825 		/* Enable CQ IRQ */
1826 		otx2_write64(pf, NIX_LF_CINTX_INT(qidx), BIT_ULL(0));
1827 		otx2_write64(pf, NIX_LF_CINTX_ENA_W1S(qidx), BIT_ULL(0));
1828 	}
1829 
1830 	otx2_set_cints_affinity(pf);
1831 
1832 	if (pf->flags & OTX2_FLAG_RX_VLAN_SUPPORT)
1833 		otx2_enable_rxvlan(pf, true);
1834 
1835 	/* When reinitializing enable time stamping if it is enabled before */
1836 	if (pf->flags & OTX2_FLAG_TX_TSTAMP_ENABLED) {
1837 		pf->flags &= ~OTX2_FLAG_TX_TSTAMP_ENABLED;
1838 		otx2_config_hw_tx_tstamp(pf, true);
1839 	}
1840 	if (pf->flags & OTX2_FLAG_RX_TSTAMP_ENABLED) {
1841 		pf->flags &= ~OTX2_FLAG_RX_TSTAMP_ENABLED;
1842 		otx2_config_hw_rx_tstamp(pf, true);
1843 	}
1844 
1845 	pf->flags &= ~OTX2_FLAG_INTF_DOWN;
1846 	/* 'intf_down' may be checked on any cpu */
1847 	smp_wmb();
1848 
1849 	/* Enable QoS configuration before starting tx queues */
1850 	otx2_qos_config_txschq(pf);
1851 
1852 	/* we have already received link status notification */
1853 	if (pf->linfo.link_up && !(pf->pcifunc & RVU_PFVF_FUNC_MASK))
1854 		otx2_handle_link_event(pf);
1855 
1856 	/* Install DMAC Filters */
1857 	if (pf->flags & OTX2_FLAG_DMACFLTR_SUPPORT)
1858 		otx2_dmacflt_reinstall_flows(pf);
1859 
1860 	err = otx2_rxtx_enable(pf, true);
1861 	/* If a mbox communication error happens at this point then interface
1862 	 * will end up in a state such that it is in down state but hardware
1863 	 * mcam entries are enabled to receive the packets. Hence disable the
1864 	 * packet I/O.
1865 	 */
1866 	if (err == EIO)
1867 		goto err_disable_rxtx;
1868 	else if (err)
1869 		goto err_tx_stop_queues;
1870 
1871 	otx2_do_set_rx_mode(pf);
1872 
1873 	return 0;
1874 
1875 err_disable_rxtx:
1876 	otx2_rxtx_enable(pf, false);
1877 err_tx_stop_queues:
1878 	netif_tx_stop_all_queues(netdev);
1879 	netif_carrier_off(netdev);
1880 	pf->flags |= OTX2_FLAG_INTF_DOWN;
1881 err_free_cints:
1882 	otx2_free_cints(pf, qidx);
1883 	vec = pci_irq_vector(pf->pdev,
1884 			     pf->hw.nix_msixoff + NIX_LF_QINT_VEC_START);
1885 	otx2_write64(pf, NIX_LF_QINTX_ENA_W1C(0), BIT_ULL(0));
1886 	free_irq(vec, pf);
1887 err_disable_napi:
1888 	otx2_disable_napi(pf);
1889 	otx2_free_hw_resources(pf);
1890 err_free_mem:
1891 	kfree(qset->sq);
1892 	kfree(qset->cq);
1893 	kfree(qset->rq);
1894 	kfree(qset->napi);
1895 	return err;
1896 }
1897 EXPORT_SYMBOL(otx2_open);
1898 
1899 int otx2_stop(struct net_device *netdev)
1900 {
1901 	struct otx2_nic *pf = netdev_priv(netdev);
1902 	struct otx2_cq_poll *cq_poll = NULL;
1903 	struct otx2_qset *qset = &pf->qset;
1904 	struct otx2_rss_info *rss;
1905 	int qidx, vec, wrk;
1906 
1907 	/* If the DOWN flag is set resources are already freed */
1908 	if (pf->flags & OTX2_FLAG_INTF_DOWN)
1909 		return 0;
1910 
1911 	netif_carrier_off(netdev);
1912 	netif_tx_stop_all_queues(netdev);
1913 
1914 	pf->flags |= OTX2_FLAG_INTF_DOWN;
1915 	/* 'intf_down' may be checked on any cpu */
1916 	smp_wmb();
1917 
1918 	/* First stop packet Rx/Tx */
1919 	otx2_rxtx_enable(pf, false);
1920 
1921 	/* Clear RSS enable flag */
1922 	rss = &pf->hw.rss_info;
1923 	rss->enable = false;
1924 
1925 	/* Cleanup Queue IRQ */
1926 	vec = pci_irq_vector(pf->pdev,
1927 			     pf->hw.nix_msixoff + NIX_LF_QINT_VEC_START);
1928 	otx2_write64(pf, NIX_LF_QINTX_ENA_W1C(0), BIT_ULL(0));
1929 	free_irq(vec, pf);
1930 
1931 	/* Cleanup CQ NAPI and IRQ */
1932 	vec = pf->hw.nix_msixoff + NIX_LF_CINT_VEC_START;
1933 	for (qidx = 0; qidx < pf->hw.cint_cnt; qidx++) {
1934 		/* Disable interrupt */
1935 		otx2_write64(pf, NIX_LF_CINTX_ENA_W1C(qidx), BIT_ULL(0));
1936 
1937 		synchronize_irq(pci_irq_vector(pf->pdev, vec));
1938 
1939 		cq_poll = &qset->napi[qidx];
1940 		napi_synchronize(&cq_poll->napi);
1941 		vec++;
1942 	}
1943 
1944 	netif_tx_disable(netdev);
1945 
1946 	for (wrk = 0; wrk < pf->qset.cq_cnt; wrk++)
1947 		cancel_delayed_work_sync(&pf->refill_wrk[wrk].pool_refill_work);
1948 	devm_kfree(pf->dev, pf->refill_wrk);
1949 
1950 	otx2_free_hw_resources(pf);
1951 	otx2_free_cints(pf, pf->hw.cint_cnt);
1952 	otx2_disable_napi(pf);
1953 
1954 	for (qidx = 0; qidx < netdev->num_tx_queues; qidx++)
1955 		netdev_tx_reset_queue(netdev_get_tx_queue(netdev, qidx));
1956 
1957 
1958 	kfree(qset->sq);
1959 	kfree(qset->cq);
1960 	kfree(qset->rq);
1961 	kfree(qset->napi);
1962 	/* Do not clear RQ/SQ ringsize settings */
1963 	memset_startat(qset, 0, sqe_cnt);
1964 	return 0;
1965 }
1966 EXPORT_SYMBOL(otx2_stop);
1967 
1968 static netdev_tx_t otx2_xmit(struct sk_buff *skb, struct net_device *netdev)
1969 {
1970 	struct otx2_nic *pf = netdev_priv(netdev);
1971 	int qidx = skb_get_queue_mapping(skb);
1972 	struct otx2_snd_queue *sq;
1973 	struct netdev_queue *txq;
1974 	int sq_idx;
1975 
1976 	/* XDP SQs are not mapped with TXQs
1977 	 * advance qid to derive correct sq mapped with QOS
1978 	 */
1979 	sq_idx = (qidx >= pf->hw.tx_queues) ? (qidx + pf->hw.xdp_queues) : qidx;
1980 
1981 	/* Check for minimum and maximum packet length */
1982 	if (skb->len <= ETH_HLEN ||
1983 	    (!skb_shinfo(skb)->gso_size && skb->len > pf->tx_max_pktlen)) {
1984 		dev_kfree_skb(skb);
1985 		return NETDEV_TX_OK;
1986 	}
1987 
1988 	sq = &pf->qset.sq[sq_idx];
1989 	txq = netdev_get_tx_queue(netdev, qidx);
1990 
1991 	if (!otx2_sq_append_skb(netdev, sq, skb, qidx)) {
1992 		netif_tx_stop_queue(txq);
1993 
1994 		/* Check again, incase SQBs got freed up */
1995 		smp_mb();
1996 		if (((sq->num_sqbs - *sq->aura_fc_addr) * sq->sqe_per_sqb)
1997 							> sq->sqe_thresh)
1998 			netif_tx_wake_queue(txq);
1999 
2000 		return NETDEV_TX_BUSY;
2001 	}
2002 
2003 	return NETDEV_TX_OK;
2004 }
2005 
2006 static int otx2_qos_select_htb_queue(struct otx2_nic *pf, struct sk_buff *skb,
2007 				     u16 htb_maj_id)
2008 {
2009 	u16 classid;
2010 
2011 	if ((TC_H_MAJ(skb->priority) >> 16) == htb_maj_id)
2012 		classid = TC_H_MIN(skb->priority);
2013 	else
2014 		classid = READ_ONCE(pf->qos.defcls);
2015 
2016 	if (!classid)
2017 		return 0;
2018 
2019 	return otx2_get_txq_by_classid(pf, classid);
2020 }
2021 
2022 u16 otx2_select_queue(struct net_device *netdev, struct sk_buff *skb,
2023 		      struct net_device *sb_dev)
2024 {
2025 	struct otx2_nic *pf = netdev_priv(netdev);
2026 	bool qos_enabled;
2027 #ifdef CONFIG_DCB
2028 	u8 vlan_prio;
2029 #endif
2030 	int txq;
2031 
2032 	qos_enabled = netdev->real_num_tx_queues > pf->hw.tx_queues;
2033 	if (unlikely(qos_enabled)) {
2034 		/* This smp_load_acquire() pairs with smp_store_release() in
2035 		 * otx2_qos_root_add() called from htb offload root creation
2036 		 */
2037 		u16 htb_maj_id = smp_load_acquire(&pf->qos.maj_id);
2038 
2039 		if (unlikely(htb_maj_id)) {
2040 			txq = otx2_qos_select_htb_queue(pf, skb, htb_maj_id);
2041 			if (txq > 0)
2042 				return txq;
2043 			goto process_pfc;
2044 		}
2045 	}
2046 
2047 process_pfc:
2048 #ifdef CONFIG_DCB
2049 	if (!skb_vlan_tag_present(skb))
2050 		goto pick_tx;
2051 
2052 	vlan_prio = skb->vlan_tci >> 13;
2053 	if ((vlan_prio > pf->hw.tx_queues - 1) ||
2054 	    !pf->pfc_alloc_status[vlan_prio])
2055 		goto pick_tx;
2056 
2057 	return vlan_prio;
2058 
2059 pick_tx:
2060 #endif
2061 	txq = netdev_pick_tx(netdev, skb, NULL);
2062 	if (unlikely(qos_enabled))
2063 		return txq % pf->hw.tx_queues;
2064 
2065 	return txq;
2066 }
2067 EXPORT_SYMBOL(otx2_select_queue);
2068 
2069 static netdev_features_t otx2_fix_features(struct net_device *dev,
2070 					   netdev_features_t features)
2071 {
2072 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
2073 		features |= NETIF_F_HW_VLAN_STAG_RX;
2074 	else
2075 		features &= ~NETIF_F_HW_VLAN_STAG_RX;
2076 
2077 	return features;
2078 }
2079 
2080 static void otx2_set_rx_mode(struct net_device *netdev)
2081 {
2082 	struct otx2_nic *pf = netdev_priv(netdev);
2083 
2084 	queue_work(pf->otx2_wq, &pf->rx_mode_work);
2085 }
2086 
2087 static void otx2_rx_mode_wrk_handler(struct work_struct *work)
2088 {
2089 	struct otx2_nic *pf = container_of(work, struct otx2_nic, rx_mode_work);
2090 
2091 	otx2_do_set_rx_mode(pf);
2092 }
2093 
2094 static int otx2_set_features(struct net_device *netdev,
2095 			     netdev_features_t features)
2096 {
2097 	netdev_features_t changed = features ^ netdev->features;
2098 	struct otx2_nic *pf = netdev_priv(netdev);
2099 
2100 	if ((changed & NETIF_F_LOOPBACK) && netif_running(netdev))
2101 		return otx2_cgx_config_loopback(pf,
2102 						features & NETIF_F_LOOPBACK);
2103 
2104 	if ((changed & NETIF_F_HW_VLAN_CTAG_RX) && netif_running(netdev))
2105 		return otx2_enable_rxvlan(pf,
2106 					  features & NETIF_F_HW_VLAN_CTAG_RX);
2107 
2108 	return otx2_handle_ntuple_tc_features(netdev, features);
2109 }
2110 
2111 static void otx2_reset_task(struct work_struct *work)
2112 {
2113 	struct otx2_nic *pf = container_of(work, struct otx2_nic, reset_task);
2114 
2115 	if (!netif_running(pf->netdev))
2116 		return;
2117 
2118 	rtnl_lock();
2119 	otx2_stop(pf->netdev);
2120 	pf->reset_count++;
2121 	otx2_open(pf->netdev);
2122 	netif_trans_update(pf->netdev);
2123 	rtnl_unlock();
2124 }
2125 
2126 static int otx2_config_hw_rx_tstamp(struct otx2_nic *pfvf, bool enable)
2127 {
2128 	struct msg_req *req;
2129 	int err;
2130 
2131 	if (pfvf->flags & OTX2_FLAG_RX_TSTAMP_ENABLED && enable)
2132 		return 0;
2133 
2134 	mutex_lock(&pfvf->mbox.lock);
2135 	if (enable)
2136 		req = otx2_mbox_alloc_msg_cgx_ptp_rx_enable(&pfvf->mbox);
2137 	else
2138 		req = otx2_mbox_alloc_msg_cgx_ptp_rx_disable(&pfvf->mbox);
2139 	if (!req) {
2140 		mutex_unlock(&pfvf->mbox.lock);
2141 		return -ENOMEM;
2142 	}
2143 
2144 	err = otx2_sync_mbox_msg(&pfvf->mbox);
2145 	if (err) {
2146 		mutex_unlock(&pfvf->mbox.lock);
2147 		return err;
2148 	}
2149 
2150 	mutex_unlock(&pfvf->mbox.lock);
2151 	if (enable)
2152 		pfvf->flags |= OTX2_FLAG_RX_TSTAMP_ENABLED;
2153 	else
2154 		pfvf->flags &= ~OTX2_FLAG_RX_TSTAMP_ENABLED;
2155 	return 0;
2156 }
2157 
2158 static int otx2_config_hw_tx_tstamp(struct otx2_nic *pfvf, bool enable)
2159 {
2160 	struct msg_req *req;
2161 	int err;
2162 
2163 	if (pfvf->flags & OTX2_FLAG_TX_TSTAMP_ENABLED && enable)
2164 		return 0;
2165 
2166 	mutex_lock(&pfvf->mbox.lock);
2167 	if (enable)
2168 		req = otx2_mbox_alloc_msg_nix_lf_ptp_tx_enable(&pfvf->mbox);
2169 	else
2170 		req = otx2_mbox_alloc_msg_nix_lf_ptp_tx_disable(&pfvf->mbox);
2171 	if (!req) {
2172 		mutex_unlock(&pfvf->mbox.lock);
2173 		return -ENOMEM;
2174 	}
2175 
2176 	err = otx2_sync_mbox_msg(&pfvf->mbox);
2177 	if (err) {
2178 		mutex_unlock(&pfvf->mbox.lock);
2179 		return err;
2180 	}
2181 
2182 	mutex_unlock(&pfvf->mbox.lock);
2183 	if (enable)
2184 		pfvf->flags |= OTX2_FLAG_TX_TSTAMP_ENABLED;
2185 	else
2186 		pfvf->flags &= ~OTX2_FLAG_TX_TSTAMP_ENABLED;
2187 	return 0;
2188 }
2189 
2190 int otx2_config_hwtstamp(struct net_device *netdev, struct ifreq *ifr)
2191 {
2192 	struct otx2_nic *pfvf = netdev_priv(netdev);
2193 	struct hwtstamp_config config;
2194 
2195 	if (!pfvf->ptp)
2196 		return -ENODEV;
2197 
2198 	if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
2199 		return -EFAULT;
2200 
2201 	switch (config.tx_type) {
2202 	case HWTSTAMP_TX_OFF:
2203 		if (pfvf->flags & OTX2_FLAG_PTP_ONESTEP_SYNC)
2204 			pfvf->flags &= ~OTX2_FLAG_PTP_ONESTEP_SYNC;
2205 
2206 		cancel_delayed_work(&pfvf->ptp->synctstamp_work);
2207 		otx2_config_hw_tx_tstamp(pfvf, false);
2208 		break;
2209 	case HWTSTAMP_TX_ONESTEP_SYNC:
2210 		if (!test_bit(CN10K_PTP_ONESTEP, &pfvf->hw.cap_flag))
2211 			return -ERANGE;
2212 		pfvf->flags |= OTX2_FLAG_PTP_ONESTEP_SYNC;
2213 		schedule_delayed_work(&pfvf->ptp->synctstamp_work,
2214 				      msecs_to_jiffies(500));
2215 		fallthrough;
2216 	case HWTSTAMP_TX_ON:
2217 		otx2_config_hw_tx_tstamp(pfvf, true);
2218 		break;
2219 	default:
2220 		return -ERANGE;
2221 	}
2222 
2223 	switch (config.rx_filter) {
2224 	case HWTSTAMP_FILTER_NONE:
2225 		otx2_config_hw_rx_tstamp(pfvf, false);
2226 		break;
2227 	case HWTSTAMP_FILTER_ALL:
2228 	case HWTSTAMP_FILTER_SOME:
2229 	case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
2230 	case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
2231 	case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
2232 	case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
2233 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
2234 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
2235 	case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
2236 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
2237 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
2238 	case HWTSTAMP_FILTER_PTP_V2_EVENT:
2239 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
2240 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
2241 		otx2_config_hw_rx_tstamp(pfvf, true);
2242 		config.rx_filter = HWTSTAMP_FILTER_ALL;
2243 		break;
2244 	default:
2245 		return -ERANGE;
2246 	}
2247 
2248 	memcpy(&pfvf->tstamp, &config, sizeof(config));
2249 
2250 	return copy_to_user(ifr->ifr_data, &config,
2251 			    sizeof(config)) ? -EFAULT : 0;
2252 }
2253 EXPORT_SYMBOL(otx2_config_hwtstamp);
2254 
2255 int otx2_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
2256 {
2257 	struct otx2_nic *pfvf = netdev_priv(netdev);
2258 	struct hwtstamp_config *cfg = &pfvf->tstamp;
2259 
2260 	switch (cmd) {
2261 	case SIOCSHWTSTAMP:
2262 		return otx2_config_hwtstamp(netdev, req);
2263 	case SIOCGHWTSTAMP:
2264 		return copy_to_user(req->ifr_data, cfg,
2265 				    sizeof(*cfg)) ? -EFAULT : 0;
2266 	default:
2267 		return -EOPNOTSUPP;
2268 	}
2269 }
2270 EXPORT_SYMBOL(otx2_ioctl);
2271 
2272 static int otx2_do_set_vf_mac(struct otx2_nic *pf, int vf, const u8 *mac)
2273 {
2274 	struct npc_install_flow_req *req;
2275 	int err;
2276 
2277 	mutex_lock(&pf->mbox.lock);
2278 	req = otx2_mbox_alloc_msg_npc_install_flow(&pf->mbox);
2279 	if (!req) {
2280 		err = -ENOMEM;
2281 		goto out;
2282 	}
2283 
2284 	ether_addr_copy(req->packet.dmac, mac);
2285 	eth_broadcast_addr((u8 *)&req->mask.dmac);
2286 	req->features = BIT_ULL(NPC_DMAC);
2287 	req->channel = pf->hw.rx_chan_base;
2288 	req->intf = NIX_INTF_RX;
2289 	req->default_rule = 1;
2290 	req->append = 1;
2291 	req->vf = vf + 1;
2292 	req->op = NIX_RX_ACTION_DEFAULT;
2293 
2294 	err = otx2_sync_mbox_msg(&pf->mbox);
2295 out:
2296 	mutex_unlock(&pf->mbox.lock);
2297 	return err;
2298 }
2299 
2300 static int otx2_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
2301 {
2302 	struct otx2_nic *pf = netdev_priv(netdev);
2303 	struct pci_dev *pdev = pf->pdev;
2304 	struct otx2_vf_config *config;
2305 	int ret;
2306 
2307 	if (!netif_running(netdev))
2308 		return -EAGAIN;
2309 
2310 	if (vf >= pf->total_vfs)
2311 		return -EINVAL;
2312 
2313 	if (!is_valid_ether_addr(mac))
2314 		return -EINVAL;
2315 
2316 	config = &pf->vf_configs[vf];
2317 	ether_addr_copy(config->mac, mac);
2318 
2319 	ret = otx2_do_set_vf_mac(pf, vf, mac);
2320 	if (ret == 0)
2321 		dev_info(&pdev->dev,
2322 			 "Load/Reload VF driver\n");
2323 
2324 	return ret;
2325 }
2326 
2327 static int otx2_do_set_vf_vlan(struct otx2_nic *pf, int vf, u16 vlan, u8 qos,
2328 			       __be16 proto)
2329 {
2330 	struct otx2_flow_config *flow_cfg = pf->flow_cfg;
2331 	struct nix_vtag_config_rsp *vtag_rsp;
2332 	struct npc_delete_flow_req *del_req;
2333 	struct nix_vtag_config *vtag_req;
2334 	struct npc_install_flow_req *req;
2335 	struct otx2_vf_config *config;
2336 	int err = 0;
2337 	u32 idx;
2338 
2339 	config = &pf->vf_configs[vf];
2340 
2341 	if (!vlan && !config->vlan)
2342 		goto out;
2343 
2344 	mutex_lock(&pf->mbox.lock);
2345 
2346 	/* free old tx vtag entry */
2347 	if (config->vlan) {
2348 		vtag_req = otx2_mbox_alloc_msg_nix_vtag_cfg(&pf->mbox);
2349 		if (!vtag_req) {
2350 			err = -ENOMEM;
2351 			goto out;
2352 		}
2353 		vtag_req->cfg_type = 0;
2354 		vtag_req->tx.free_vtag0 = 1;
2355 		vtag_req->tx.vtag0_idx = config->tx_vtag_idx;
2356 
2357 		err = otx2_sync_mbox_msg(&pf->mbox);
2358 		if (err)
2359 			goto out;
2360 	}
2361 
2362 	if (!vlan && config->vlan) {
2363 		/* rx */
2364 		del_req = otx2_mbox_alloc_msg_npc_delete_flow(&pf->mbox);
2365 		if (!del_req) {
2366 			err = -ENOMEM;
2367 			goto out;
2368 		}
2369 		idx = ((vf * OTX2_PER_VF_VLAN_FLOWS) + OTX2_VF_VLAN_RX_INDEX);
2370 		del_req->entry =
2371 			flow_cfg->def_ent[flow_cfg->vf_vlan_offset + idx];
2372 		err = otx2_sync_mbox_msg(&pf->mbox);
2373 		if (err)
2374 			goto out;
2375 
2376 		/* tx */
2377 		del_req = otx2_mbox_alloc_msg_npc_delete_flow(&pf->mbox);
2378 		if (!del_req) {
2379 			err = -ENOMEM;
2380 			goto out;
2381 		}
2382 		idx = ((vf * OTX2_PER_VF_VLAN_FLOWS) + OTX2_VF_VLAN_TX_INDEX);
2383 		del_req->entry =
2384 			flow_cfg->def_ent[flow_cfg->vf_vlan_offset + idx];
2385 		err = otx2_sync_mbox_msg(&pf->mbox);
2386 
2387 		goto out;
2388 	}
2389 
2390 	/* rx */
2391 	req = otx2_mbox_alloc_msg_npc_install_flow(&pf->mbox);
2392 	if (!req) {
2393 		err = -ENOMEM;
2394 		goto out;
2395 	}
2396 
2397 	idx = ((vf * OTX2_PER_VF_VLAN_FLOWS) + OTX2_VF_VLAN_RX_INDEX);
2398 	req->entry = flow_cfg->def_ent[flow_cfg->vf_vlan_offset + idx];
2399 	req->packet.vlan_tci = htons(vlan);
2400 	req->mask.vlan_tci = htons(VLAN_VID_MASK);
2401 	/* af fills the destination mac addr */
2402 	eth_broadcast_addr((u8 *)&req->mask.dmac);
2403 	req->features = BIT_ULL(NPC_OUTER_VID) | BIT_ULL(NPC_DMAC);
2404 	req->channel = pf->hw.rx_chan_base;
2405 	req->intf = NIX_INTF_RX;
2406 	req->vf = vf + 1;
2407 	req->op = NIX_RX_ACTION_DEFAULT;
2408 	req->vtag0_valid = true;
2409 	req->vtag0_type = NIX_AF_LFX_RX_VTAG_TYPE7;
2410 	req->set_cntr = 1;
2411 
2412 	err = otx2_sync_mbox_msg(&pf->mbox);
2413 	if (err)
2414 		goto out;
2415 
2416 	/* tx */
2417 	vtag_req = otx2_mbox_alloc_msg_nix_vtag_cfg(&pf->mbox);
2418 	if (!vtag_req) {
2419 		err = -ENOMEM;
2420 		goto out;
2421 	}
2422 
2423 	/* configure tx vtag params */
2424 	vtag_req->vtag_size = VTAGSIZE_T4;
2425 	vtag_req->cfg_type = 0; /* tx vlan cfg */
2426 	vtag_req->tx.cfg_vtag0 = 1;
2427 	vtag_req->tx.vtag0 = ((u64)ntohs(proto) << 16) | vlan;
2428 
2429 	err = otx2_sync_mbox_msg(&pf->mbox);
2430 	if (err)
2431 		goto out;
2432 
2433 	vtag_rsp = (struct nix_vtag_config_rsp *)otx2_mbox_get_rsp
2434 			(&pf->mbox.mbox, 0, &vtag_req->hdr);
2435 	if (IS_ERR(vtag_rsp)) {
2436 		err = PTR_ERR(vtag_rsp);
2437 		goto out;
2438 	}
2439 	config->tx_vtag_idx = vtag_rsp->vtag0_idx;
2440 
2441 	req = otx2_mbox_alloc_msg_npc_install_flow(&pf->mbox);
2442 	if (!req) {
2443 		err = -ENOMEM;
2444 		goto out;
2445 	}
2446 
2447 	eth_zero_addr((u8 *)&req->mask.dmac);
2448 	idx = ((vf * OTX2_PER_VF_VLAN_FLOWS) + OTX2_VF_VLAN_TX_INDEX);
2449 	req->entry = flow_cfg->def_ent[flow_cfg->vf_vlan_offset + idx];
2450 	req->features = BIT_ULL(NPC_DMAC);
2451 	req->channel = pf->hw.tx_chan_base;
2452 	req->intf = NIX_INTF_TX;
2453 	req->vf = vf + 1;
2454 	req->op = NIX_TX_ACTIONOP_UCAST_DEFAULT;
2455 	req->vtag0_def = vtag_rsp->vtag0_idx;
2456 	req->vtag0_op = VTAG_INSERT;
2457 	req->set_cntr = 1;
2458 
2459 	err = otx2_sync_mbox_msg(&pf->mbox);
2460 out:
2461 	config->vlan = vlan;
2462 	mutex_unlock(&pf->mbox.lock);
2463 	return err;
2464 }
2465 
2466 static int otx2_set_vf_vlan(struct net_device *netdev, int vf, u16 vlan, u8 qos,
2467 			    __be16 proto)
2468 {
2469 	struct otx2_nic *pf = netdev_priv(netdev);
2470 	struct pci_dev *pdev = pf->pdev;
2471 
2472 	if (!netif_running(netdev))
2473 		return -EAGAIN;
2474 
2475 	if (vf >= pci_num_vf(pdev))
2476 		return -EINVAL;
2477 
2478 	/* qos is currently unsupported */
2479 	if (vlan >= VLAN_N_VID || qos)
2480 		return -EINVAL;
2481 
2482 	if (proto != htons(ETH_P_8021Q))
2483 		return -EPROTONOSUPPORT;
2484 
2485 	if (!(pf->flags & OTX2_FLAG_VF_VLAN_SUPPORT))
2486 		return -EOPNOTSUPP;
2487 
2488 	return otx2_do_set_vf_vlan(pf, vf, vlan, qos, proto);
2489 }
2490 
2491 static int otx2_get_vf_config(struct net_device *netdev, int vf,
2492 			      struct ifla_vf_info *ivi)
2493 {
2494 	struct otx2_nic *pf = netdev_priv(netdev);
2495 	struct pci_dev *pdev = pf->pdev;
2496 	struct otx2_vf_config *config;
2497 
2498 	if (!netif_running(netdev))
2499 		return -EAGAIN;
2500 
2501 	if (vf >= pci_num_vf(pdev))
2502 		return -EINVAL;
2503 
2504 	config = &pf->vf_configs[vf];
2505 	ivi->vf = vf;
2506 	ether_addr_copy(ivi->mac, config->mac);
2507 	ivi->vlan = config->vlan;
2508 	ivi->trusted = config->trusted;
2509 
2510 	return 0;
2511 }
2512 
2513 static int otx2_xdp_xmit_tx(struct otx2_nic *pf, struct xdp_frame *xdpf,
2514 			    int qidx)
2515 {
2516 	struct page *page;
2517 	u64 dma_addr;
2518 	int err = 0;
2519 
2520 	dma_addr = otx2_dma_map_page(pf, virt_to_page(xdpf->data),
2521 				     offset_in_page(xdpf->data), xdpf->len,
2522 				     DMA_TO_DEVICE);
2523 	if (dma_mapping_error(pf->dev, dma_addr))
2524 		return -ENOMEM;
2525 
2526 	err = otx2_xdp_sq_append_pkt(pf, dma_addr, xdpf->len, qidx);
2527 	if (!err) {
2528 		otx2_dma_unmap_page(pf, dma_addr, xdpf->len, DMA_TO_DEVICE);
2529 		page = virt_to_page(xdpf->data);
2530 		put_page(page);
2531 		return -ENOMEM;
2532 	}
2533 	return 0;
2534 }
2535 
2536 static int otx2_xdp_xmit(struct net_device *netdev, int n,
2537 			 struct xdp_frame **frames, u32 flags)
2538 {
2539 	struct otx2_nic *pf = netdev_priv(netdev);
2540 	int qidx = smp_processor_id();
2541 	struct otx2_snd_queue *sq;
2542 	int drops = 0, i;
2543 
2544 	if (!netif_running(netdev))
2545 		return -ENETDOWN;
2546 
2547 	qidx += pf->hw.tx_queues;
2548 	sq = pf->xdp_prog ? &pf->qset.sq[qidx] : NULL;
2549 
2550 	/* Abort xmit if xdp queue is not */
2551 	if (unlikely(!sq))
2552 		return -ENXIO;
2553 
2554 	if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK))
2555 		return -EINVAL;
2556 
2557 	for (i = 0; i < n; i++) {
2558 		struct xdp_frame *xdpf = frames[i];
2559 		int err;
2560 
2561 		err = otx2_xdp_xmit_tx(pf, xdpf, qidx);
2562 		if (err)
2563 			drops++;
2564 	}
2565 	return n - drops;
2566 }
2567 
2568 static int otx2_xdp_setup(struct otx2_nic *pf, struct bpf_prog *prog)
2569 {
2570 	struct net_device *dev = pf->netdev;
2571 	bool if_up = netif_running(pf->netdev);
2572 	struct bpf_prog *old_prog;
2573 
2574 	if (prog && dev->mtu > MAX_XDP_MTU) {
2575 		netdev_warn(dev, "Jumbo frames not yet supported with XDP\n");
2576 		return -EOPNOTSUPP;
2577 	}
2578 
2579 	if (if_up)
2580 		otx2_stop(pf->netdev);
2581 
2582 	old_prog = xchg(&pf->xdp_prog, prog);
2583 
2584 	if (old_prog)
2585 		bpf_prog_put(old_prog);
2586 
2587 	if (pf->xdp_prog)
2588 		bpf_prog_add(pf->xdp_prog, pf->hw.rx_queues - 1);
2589 
2590 	/* Network stack and XDP shared same rx queues.
2591 	 * Use separate tx queues for XDP and network stack.
2592 	 */
2593 	if (pf->xdp_prog) {
2594 		pf->hw.xdp_queues = pf->hw.rx_queues;
2595 		xdp_features_set_redirect_target(dev, false);
2596 	} else {
2597 		pf->hw.xdp_queues = 0;
2598 		xdp_features_clear_redirect_target(dev);
2599 	}
2600 
2601 	pf->hw.non_qos_queues += pf->hw.xdp_queues;
2602 
2603 	if (if_up)
2604 		otx2_open(pf->netdev);
2605 
2606 	return 0;
2607 }
2608 
2609 static int otx2_xdp(struct net_device *netdev, struct netdev_bpf *xdp)
2610 {
2611 	struct otx2_nic *pf = netdev_priv(netdev);
2612 
2613 	switch (xdp->command) {
2614 	case XDP_SETUP_PROG:
2615 		return otx2_xdp_setup(pf, xdp->prog);
2616 	default:
2617 		return -EINVAL;
2618 	}
2619 }
2620 
2621 static int otx2_set_vf_permissions(struct otx2_nic *pf, int vf,
2622 				   int req_perm)
2623 {
2624 	struct set_vf_perm *req;
2625 	int rc;
2626 
2627 	mutex_lock(&pf->mbox.lock);
2628 	req = otx2_mbox_alloc_msg_set_vf_perm(&pf->mbox);
2629 	if (!req) {
2630 		rc = -ENOMEM;
2631 		goto out;
2632 	}
2633 
2634 	/* Let AF reset VF permissions as sriov is disabled */
2635 	if (req_perm == OTX2_RESET_VF_PERM) {
2636 		req->flags |= RESET_VF_PERM;
2637 	} else if (req_perm == OTX2_TRUSTED_VF) {
2638 		if (pf->vf_configs[vf].trusted)
2639 			req->flags |= VF_TRUSTED;
2640 	}
2641 
2642 	req->vf = vf;
2643 	rc = otx2_sync_mbox_msg(&pf->mbox);
2644 out:
2645 	mutex_unlock(&pf->mbox.lock);
2646 	return rc;
2647 }
2648 
2649 static int otx2_ndo_set_vf_trust(struct net_device *netdev, int vf,
2650 				 bool enable)
2651 {
2652 	struct otx2_nic *pf = netdev_priv(netdev);
2653 	struct pci_dev *pdev = pf->pdev;
2654 	int rc;
2655 
2656 	if (vf >= pci_num_vf(pdev))
2657 		return -EINVAL;
2658 
2659 	if (pf->vf_configs[vf].trusted == enable)
2660 		return 0;
2661 
2662 	pf->vf_configs[vf].trusted = enable;
2663 	rc = otx2_set_vf_permissions(pf, vf, OTX2_TRUSTED_VF);
2664 
2665 	if (rc)
2666 		pf->vf_configs[vf].trusted = !enable;
2667 	else
2668 		netdev_info(pf->netdev, "VF %d is %strusted\n",
2669 			    vf, enable ? "" : "not ");
2670 	return rc;
2671 }
2672 
2673 static const struct net_device_ops otx2_netdev_ops = {
2674 	.ndo_open		= otx2_open,
2675 	.ndo_stop		= otx2_stop,
2676 	.ndo_start_xmit		= otx2_xmit,
2677 	.ndo_select_queue	= otx2_select_queue,
2678 	.ndo_fix_features	= otx2_fix_features,
2679 	.ndo_set_mac_address    = otx2_set_mac_address,
2680 	.ndo_change_mtu		= otx2_change_mtu,
2681 	.ndo_set_rx_mode	= otx2_set_rx_mode,
2682 	.ndo_set_features	= otx2_set_features,
2683 	.ndo_tx_timeout		= otx2_tx_timeout,
2684 	.ndo_get_stats64	= otx2_get_stats64,
2685 	.ndo_eth_ioctl		= otx2_ioctl,
2686 	.ndo_set_vf_mac		= otx2_set_vf_mac,
2687 	.ndo_set_vf_vlan	= otx2_set_vf_vlan,
2688 	.ndo_get_vf_config	= otx2_get_vf_config,
2689 	.ndo_bpf		= otx2_xdp,
2690 	.ndo_xdp_xmit           = otx2_xdp_xmit,
2691 	.ndo_setup_tc		= otx2_setup_tc,
2692 	.ndo_set_vf_trust	= otx2_ndo_set_vf_trust,
2693 };
2694 
2695 static int otx2_wq_init(struct otx2_nic *pf)
2696 {
2697 	pf->otx2_wq = create_singlethread_workqueue("otx2_wq");
2698 	if (!pf->otx2_wq)
2699 		return -ENOMEM;
2700 
2701 	INIT_WORK(&pf->rx_mode_work, otx2_rx_mode_wrk_handler);
2702 	INIT_WORK(&pf->reset_task, otx2_reset_task);
2703 	return 0;
2704 }
2705 
2706 static int otx2_check_pf_usable(struct otx2_nic *nic)
2707 {
2708 	u64 rev;
2709 
2710 	rev = otx2_read64(nic, RVU_PF_BLOCK_ADDRX_DISC(BLKADDR_RVUM));
2711 	rev = (rev >> 12) & 0xFF;
2712 	/* Check if AF has setup revision for RVUM block,
2713 	 * otherwise this driver probe should be deferred
2714 	 * until AF driver comes up.
2715 	 */
2716 	if (!rev) {
2717 		dev_warn(nic->dev,
2718 			 "AF is not initialized, deferring probe\n");
2719 		return -EPROBE_DEFER;
2720 	}
2721 	return 0;
2722 }
2723 
2724 static int otx2_realloc_msix_vectors(struct otx2_nic *pf)
2725 {
2726 	struct otx2_hw *hw = &pf->hw;
2727 	int num_vec, err;
2728 
2729 	/* NPA interrupts are inot registered, so alloc only
2730 	 * upto NIX vector offset.
2731 	 */
2732 	num_vec = hw->nix_msixoff;
2733 	num_vec += NIX_LF_CINT_VEC_START + hw->max_queues;
2734 
2735 	otx2_disable_mbox_intr(pf);
2736 	pci_free_irq_vectors(hw->pdev);
2737 	err = pci_alloc_irq_vectors(hw->pdev, num_vec, num_vec, PCI_IRQ_MSIX);
2738 	if (err < 0) {
2739 		dev_err(pf->dev, "%s: Failed to realloc %d IRQ vectors\n",
2740 			__func__, num_vec);
2741 		return err;
2742 	}
2743 
2744 	return otx2_register_mbox_intr(pf, false);
2745 }
2746 
2747 static int otx2_sriov_vfcfg_init(struct otx2_nic *pf)
2748 {
2749 	int i;
2750 
2751 	pf->vf_configs = devm_kcalloc(pf->dev, pf->total_vfs,
2752 				      sizeof(struct otx2_vf_config),
2753 				      GFP_KERNEL);
2754 	if (!pf->vf_configs)
2755 		return -ENOMEM;
2756 
2757 	for (i = 0; i < pf->total_vfs; i++) {
2758 		pf->vf_configs[i].pf = pf;
2759 		pf->vf_configs[i].intf_down = true;
2760 		pf->vf_configs[i].trusted = false;
2761 		INIT_DELAYED_WORK(&pf->vf_configs[i].link_event_work,
2762 				  otx2_vf_link_event_task);
2763 	}
2764 
2765 	return 0;
2766 }
2767 
2768 static void otx2_sriov_vfcfg_cleanup(struct otx2_nic *pf)
2769 {
2770 	int i;
2771 
2772 	if (!pf->vf_configs)
2773 		return;
2774 
2775 	for (i = 0; i < pf->total_vfs; i++) {
2776 		cancel_delayed_work_sync(&pf->vf_configs[i].link_event_work);
2777 		otx2_set_vf_permissions(pf, i, OTX2_RESET_VF_PERM);
2778 	}
2779 }
2780 
2781 static int otx2_probe(struct pci_dev *pdev, const struct pci_device_id *id)
2782 {
2783 	struct device *dev = &pdev->dev;
2784 	int err, qcount, qos_txqs;
2785 	struct net_device *netdev;
2786 	struct otx2_nic *pf;
2787 	struct otx2_hw *hw;
2788 	int num_vec;
2789 
2790 	err = pcim_enable_device(pdev);
2791 	if (err) {
2792 		dev_err(dev, "Failed to enable PCI device\n");
2793 		return err;
2794 	}
2795 
2796 	err = pci_request_regions(pdev, DRV_NAME);
2797 	if (err) {
2798 		dev_err(dev, "PCI request regions failed 0x%x\n", err);
2799 		return err;
2800 	}
2801 
2802 	err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(48));
2803 	if (err) {
2804 		dev_err(dev, "DMA mask config failed, abort\n");
2805 		goto err_release_regions;
2806 	}
2807 
2808 	pci_set_master(pdev);
2809 
2810 	/* Set number of queues */
2811 	qcount = min_t(int, num_online_cpus(), OTX2_MAX_CQ_CNT);
2812 	qos_txqs = min_t(int, qcount, OTX2_QOS_MAX_LEAF_NODES);
2813 
2814 	netdev = alloc_etherdev_mqs(sizeof(*pf), qcount + qos_txqs, qcount);
2815 	if (!netdev) {
2816 		err = -ENOMEM;
2817 		goto err_release_regions;
2818 	}
2819 
2820 	pci_set_drvdata(pdev, netdev);
2821 	SET_NETDEV_DEV(netdev, &pdev->dev);
2822 	pf = netdev_priv(netdev);
2823 	pf->netdev = netdev;
2824 	pf->pdev = pdev;
2825 	pf->dev = dev;
2826 	pf->total_vfs = pci_sriov_get_totalvfs(pdev);
2827 	pf->flags |= OTX2_FLAG_INTF_DOWN;
2828 
2829 	hw = &pf->hw;
2830 	hw->pdev = pdev;
2831 	hw->rx_queues = qcount;
2832 	hw->tx_queues = qcount;
2833 	hw->non_qos_queues = qcount;
2834 	hw->max_queues = qcount;
2835 	hw->rbuf_len = OTX2_DEFAULT_RBUF_LEN;
2836 	/* Use CQE of 128 byte descriptor size by default */
2837 	hw->xqe_size = 128;
2838 
2839 	num_vec = pci_msix_vec_count(pdev);
2840 	hw->irq_name = devm_kmalloc_array(&hw->pdev->dev, num_vec, NAME_SIZE,
2841 					  GFP_KERNEL);
2842 	if (!hw->irq_name) {
2843 		err = -ENOMEM;
2844 		goto err_free_netdev;
2845 	}
2846 
2847 	hw->affinity_mask = devm_kcalloc(&hw->pdev->dev, num_vec,
2848 					 sizeof(cpumask_var_t), GFP_KERNEL);
2849 	if (!hw->affinity_mask) {
2850 		err = -ENOMEM;
2851 		goto err_free_netdev;
2852 	}
2853 
2854 	/* Map CSRs */
2855 	pf->reg_base = pcim_iomap(pdev, PCI_CFG_REG_BAR_NUM, 0);
2856 	if (!pf->reg_base) {
2857 		dev_err(dev, "Unable to map physical function CSRs, aborting\n");
2858 		err = -ENOMEM;
2859 		goto err_free_netdev;
2860 	}
2861 
2862 	err = otx2_check_pf_usable(pf);
2863 	if (err)
2864 		goto err_free_netdev;
2865 
2866 	err = pci_alloc_irq_vectors(hw->pdev, RVU_PF_INT_VEC_CNT,
2867 				    RVU_PF_INT_VEC_CNT, PCI_IRQ_MSIX);
2868 	if (err < 0) {
2869 		dev_err(dev, "%s: Failed to alloc %d IRQ vectors\n",
2870 			__func__, num_vec);
2871 		goto err_free_netdev;
2872 	}
2873 
2874 	otx2_setup_dev_hw_settings(pf);
2875 
2876 	/* Init PF <=> AF mailbox stuff */
2877 	err = otx2_pfaf_mbox_init(pf);
2878 	if (err)
2879 		goto err_free_irq_vectors;
2880 
2881 	/* Register mailbox interrupt */
2882 	err = otx2_register_mbox_intr(pf, true);
2883 	if (err)
2884 		goto err_mbox_destroy;
2885 
2886 	/* Request AF to attach NPA and NIX LFs to this PF.
2887 	 * NIX and NPA LFs are needed for this PF to function as a NIC.
2888 	 */
2889 	err = otx2_attach_npa_nix(pf);
2890 	if (err)
2891 		goto err_disable_mbox_intr;
2892 
2893 	err = otx2_realloc_msix_vectors(pf);
2894 	if (err)
2895 		goto err_detach_rsrc;
2896 
2897 	err = otx2_set_real_num_queues(netdev, hw->tx_queues, hw->rx_queues);
2898 	if (err)
2899 		goto err_detach_rsrc;
2900 
2901 	err = cn10k_lmtst_init(pf);
2902 	if (err)
2903 		goto err_detach_rsrc;
2904 
2905 	/* Assign default mac address */
2906 	otx2_get_mac_from_af(netdev);
2907 
2908 	/* Don't check for error.  Proceed without ptp */
2909 	otx2_ptp_init(pf);
2910 
2911 	/* NPA's pool is a stack to which SW frees buffer pointers via Aura.
2912 	 * HW allocates buffer pointer from stack and uses it for DMA'ing
2913 	 * ingress packet. In some scenarios HW can free back allocated buffer
2914 	 * pointers to pool. This makes it impossible for SW to maintain a
2915 	 * parallel list where physical addresses of buffer pointers (IOVAs)
2916 	 * given to HW can be saved for later reference.
2917 	 *
2918 	 * So the only way to convert Rx packet's buffer address is to use
2919 	 * IOMMU's iova_to_phys() handler which translates the address by
2920 	 * walking through the translation tables.
2921 	 */
2922 	pf->iommu_domain = iommu_get_domain_for_dev(dev);
2923 
2924 	netdev->hw_features = (NETIF_F_RXCSUM | NETIF_F_IP_CSUM |
2925 			       NETIF_F_IPV6_CSUM | NETIF_F_RXHASH |
2926 			       NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
2927 			       NETIF_F_GSO_UDP_L4);
2928 	netdev->features |= netdev->hw_features;
2929 
2930 	err = otx2_mcam_flow_init(pf);
2931 	if (err)
2932 		goto err_ptp_destroy;
2933 
2934 	err = cn10k_mcs_init(pf);
2935 	if (err)
2936 		goto err_del_mcam_entries;
2937 
2938 	if (pf->flags & OTX2_FLAG_NTUPLE_SUPPORT)
2939 		netdev->hw_features |= NETIF_F_NTUPLE;
2940 
2941 	if (pf->flags & OTX2_FLAG_UCAST_FLTR_SUPPORT)
2942 		netdev->priv_flags |= IFF_UNICAST_FLT;
2943 
2944 	/* Support TSO on tag interface */
2945 	netdev->vlan_features |= netdev->features;
2946 	netdev->hw_features  |= NETIF_F_HW_VLAN_CTAG_TX |
2947 				NETIF_F_HW_VLAN_STAG_TX;
2948 	if (pf->flags & OTX2_FLAG_RX_VLAN_SUPPORT)
2949 		netdev->hw_features |= NETIF_F_HW_VLAN_CTAG_RX |
2950 				       NETIF_F_HW_VLAN_STAG_RX;
2951 	netdev->features |= netdev->hw_features;
2952 
2953 	/* HW supports tc offload but mutually exclusive with n-tuple filters */
2954 	if (pf->flags & OTX2_FLAG_TC_FLOWER_SUPPORT)
2955 		netdev->hw_features |= NETIF_F_HW_TC;
2956 
2957 	netdev->hw_features |= NETIF_F_LOOPBACK | NETIF_F_RXALL;
2958 
2959 	netif_set_tso_max_segs(netdev, OTX2_MAX_GSO_SEGS);
2960 	netdev->watchdog_timeo = OTX2_TX_TIMEOUT;
2961 
2962 	netdev->netdev_ops = &otx2_netdev_ops;
2963 	netdev->xdp_features = NETDEV_XDP_ACT_BASIC | NETDEV_XDP_ACT_REDIRECT;
2964 
2965 	netdev->min_mtu = OTX2_MIN_MTU;
2966 	netdev->max_mtu = otx2_get_max_mtu(pf);
2967 
2968 	err = register_netdev(netdev);
2969 	if (err) {
2970 		dev_err(dev, "Failed to register netdevice\n");
2971 		goto err_mcs_free;
2972 	}
2973 
2974 	err = otx2_wq_init(pf);
2975 	if (err)
2976 		goto err_unreg_netdev;
2977 
2978 	otx2_set_ethtool_ops(netdev);
2979 
2980 	err = otx2_init_tc(pf);
2981 	if (err)
2982 		goto err_mcam_flow_del;
2983 
2984 	err = otx2_register_dl(pf);
2985 	if (err)
2986 		goto err_mcam_flow_del;
2987 
2988 	/* Initialize SR-IOV resources */
2989 	err = otx2_sriov_vfcfg_init(pf);
2990 	if (err)
2991 		goto err_pf_sriov_init;
2992 
2993 	/* Enable link notifications */
2994 	otx2_cgx_config_linkevents(pf, true);
2995 
2996 #ifdef CONFIG_DCB
2997 	err = otx2_dcbnl_set_ops(netdev);
2998 	if (err)
2999 		goto err_pf_sriov_init;
3000 #endif
3001 
3002 	otx2_qos_init(pf, qos_txqs);
3003 
3004 	return 0;
3005 
3006 err_pf_sriov_init:
3007 	otx2_shutdown_tc(pf);
3008 err_mcam_flow_del:
3009 	otx2_mcam_flow_del(pf);
3010 err_unreg_netdev:
3011 	unregister_netdev(netdev);
3012 err_mcs_free:
3013 	cn10k_mcs_free(pf);
3014 err_del_mcam_entries:
3015 	otx2_mcam_flow_del(pf);
3016 err_ptp_destroy:
3017 	otx2_ptp_destroy(pf);
3018 err_detach_rsrc:
3019 	if (pf->hw.lmt_info)
3020 		free_percpu(pf->hw.lmt_info);
3021 	if (test_bit(CN10K_LMTST, &pf->hw.cap_flag))
3022 		qmem_free(pf->dev, pf->dync_lmt);
3023 	otx2_detach_resources(&pf->mbox);
3024 err_disable_mbox_intr:
3025 	otx2_disable_mbox_intr(pf);
3026 err_mbox_destroy:
3027 	otx2_pfaf_mbox_destroy(pf);
3028 err_free_irq_vectors:
3029 	pci_free_irq_vectors(hw->pdev);
3030 err_free_netdev:
3031 	pci_set_drvdata(pdev, NULL);
3032 	free_netdev(netdev);
3033 err_release_regions:
3034 	pci_release_regions(pdev);
3035 	return err;
3036 }
3037 
3038 static void otx2_vf_link_event_task(struct work_struct *work)
3039 {
3040 	struct otx2_vf_config *config;
3041 	struct cgx_link_info_msg *req;
3042 	struct mbox_msghdr *msghdr;
3043 	struct otx2_nic *pf;
3044 	int vf_idx;
3045 
3046 	config = container_of(work, struct otx2_vf_config,
3047 			      link_event_work.work);
3048 	vf_idx = config - config->pf->vf_configs;
3049 	pf = config->pf;
3050 
3051 	msghdr = otx2_mbox_alloc_msg_rsp(&pf->mbox_pfvf[0].mbox_up, vf_idx,
3052 					 sizeof(*req), sizeof(struct msg_rsp));
3053 	if (!msghdr) {
3054 		dev_err(pf->dev, "Failed to create VF%d link event\n", vf_idx);
3055 		return;
3056 	}
3057 
3058 	req = (struct cgx_link_info_msg *)msghdr;
3059 	req->hdr.id = MBOX_MSG_CGX_LINK_EVENT;
3060 	req->hdr.sig = OTX2_MBOX_REQ_SIG;
3061 	memcpy(&req->link_info, &pf->linfo, sizeof(req->link_info));
3062 
3063 	otx2_sync_mbox_up_msg(&pf->mbox_pfvf[0], vf_idx);
3064 }
3065 
3066 static int otx2_sriov_enable(struct pci_dev *pdev, int numvfs)
3067 {
3068 	struct net_device *netdev = pci_get_drvdata(pdev);
3069 	struct otx2_nic *pf = netdev_priv(netdev);
3070 	int ret;
3071 
3072 	/* Init PF <=> VF mailbox stuff */
3073 	ret = otx2_pfvf_mbox_init(pf, numvfs);
3074 	if (ret)
3075 		return ret;
3076 
3077 	ret = otx2_register_pfvf_mbox_intr(pf, numvfs);
3078 	if (ret)
3079 		goto free_mbox;
3080 
3081 	ret = otx2_pf_flr_init(pf, numvfs);
3082 	if (ret)
3083 		goto free_intr;
3084 
3085 	ret = otx2_register_flr_me_intr(pf, numvfs);
3086 	if (ret)
3087 		goto free_flr;
3088 
3089 	ret = pci_enable_sriov(pdev, numvfs);
3090 	if (ret)
3091 		goto free_flr_intr;
3092 
3093 	return numvfs;
3094 free_flr_intr:
3095 	otx2_disable_flr_me_intr(pf);
3096 free_flr:
3097 	otx2_flr_wq_destroy(pf);
3098 free_intr:
3099 	otx2_disable_pfvf_mbox_intr(pf, numvfs);
3100 free_mbox:
3101 	otx2_pfvf_mbox_destroy(pf);
3102 	return ret;
3103 }
3104 
3105 static int otx2_sriov_disable(struct pci_dev *pdev)
3106 {
3107 	struct net_device *netdev = pci_get_drvdata(pdev);
3108 	struct otx2_nic *pf = netdev_priv(netdev);
3109 	int numvfs = pci_num_vf(pdev);
3110 
3111 	if (!numvfs)
3112 		return 0;
3113 
3114 	pci_disable_sriov(pdev);
3115 
3116 	otx2_disable_flr_me_intr(pf);
3117 	otx2_flr_wq_destroy(pf);
3118 	otx2_disable_pfvf_mbox_intr(pf, numvfs);
3119 	otx2_pfvf_mbox_destroy(pf);
3120 
3121 	return 0;
3122 }
3123 
3124 static int otx2_sriov_configure(struct pci_dev *pdev, int numvfs)
3125 {
3126 	if (numvfs == 0)
3127 		return otx2_sriov_disable(pdev);
3128 	else
3129 		return otx2_sriov_enable(pdev, numvfs);
3130 }
3131 
3132 static void otx2_remove(struct pci_dev *pdev)
3133 {
3134 	struct net_device *netdev = pci_get_drvdata(pdev);
3135 	struct otx2_nic *pf;
3136 
3137 	if (!netdev)
3138 		return;
3139 
3140 	pf = netdev_priv(netdev);
3141 
3142 	pf->flags |= OTX2_FLAG_PF_SHUTDOWN;
3143 
3144 	if (pf->flags & OTX2_FLAG_TX_TSTAMP_ENABLED)
3145 		otx2_config_hw_tx_tstamp(pf, false);
3146 	if (pf->flags & OTX2_FLAG_RX_TSTAMP_ENABLED)
3147 		otx2_config_hw_rx_tstamp(pf, false);
3148 
3149 	/* Disable 802.3x pause frames */
3150 	if (pf->flags & OTX2_FLAG_RX_PAUSE_ENABLED ||
3151 	    (pf->flags & OTX2_FLAG_TX_PAUSE_ENABLED)) {
3152 		pf->flags &= ~OTX2_FLAG_RX_PAUSE_ENABLED;
3153 		pf->flags &= ~OTX2_FLAG_TX_PAUSE_ENABLED;
3154 		otx2_config_pause_frm(pf);
3155 	}
3156 
3157 #ifdef CONFIG_DCB
3158 	/* Disable PFC config */
3159 	if (pf->pfc_en) {
3160 		pf->pfc_en = 0;
3161 		otx2_config_priority_flow_ctrl(pf);
3162 	}
3163 #endif
3164 	cancel_work_sync(&pf->reset_task);
3165 	/* Disable link notifications */
3166 	otx2_cgx_config_linkevents(pf, false);
3167 
3168 	otx2_unregister_dl(pf);
3169 	unregister_netdev(netdev);
3170 	cn10k_mcs_free(pf);
3171 	otx2_sriov_disable(pf->pdev);
3172 	otx2_sriov_vfcfg_cleanup(pf);
3173 	if (pf->otx2_wq)
3174 		destroy_workqueue(pf->otx2_wq);
3175 
3176 	otx2_ptp_destroy(pf);
3177 	otx2_mcam_flow_del(pf);
3178 	otx2_shutdown_tc(pf);
3179 	otx2_shutdown_qos(pf);
3180 	otx2_detach_resources(&pf->mbox);
3181 	if (pf->hw.lmt_info)
3182 		free_percpu(pf->hw.lmt_info);
3183 	if (test_bit(CN10K_LMTST, &pf->hw.cap_flag))
3184 		qmem_free(pf->dev, pf->dync_lmt);
3185 	otx2_disable_mbox_intr(pf);
3186 	otx2_pfaf_mbox_destroy(pf);
3187 	pci_free_irq_vectors(pf->pdev);
3188 	pci_set_drvdata(pdev, NULL);
3189 	free_netdev(netdev);
3190 
3191 	pci_release_regions(pdev);
3192 }
3193 
3194 static struct pci_driver otx2_pf_driver = {
3195 	.name = DRV_NAME,
3196 	.id_table = otx2_pf_id_table,
3197 	.probe = otx2_probe,
3198 	.shutdown = otx2_remove,
3199 	.remove = otx2_remove,
3200 	.sriov_configure = otx2_sriov_configure
3201 };
3202 
3203 static int __init otx2_rvupf_init_module(void)
3204 {
3205 	pr_info("%s: %s\n", DRV_NAME, DRV_STRING);
3206 
3207 	return pci_register_driver(&otx2_pf_driver);
3208 }
3209 
3210 static void __exit otx2_rvupf_cleanup_module(void)
3211 {
3212 	pci_unregister_driver(&otx2_pf_driver);
3213 }
3214 
3215 module_init(otx2_rvupf_init_module);
3216 module_exit(otx2_rvupf_cleanup_module);
3217