xref: /linux/drivers/net/ethernet/marvell/octeontx2/nic/otx2_pf.c (revision 8e07e0e3964ca4e23ce7b68e2096fe660a888942)
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 		if (intr)
570 			trace_otx2_msg_interrupt(mbox->mbox.pdev, "VF(s) to PF", intr);
571 		vfs = 64;
572 	}
573 
574 	intr = otx2_read64(pf, RVU_PF_VFPF_MBOX_INTX(0));
575 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(0), intr);
576 
577 	otx2_queue_work(mbox, pf->mbox_pfvf_wq, 0, vfs, intr, TYPE_PFVF);
578 
579 	if (intr)
580 		trace_otx2_msg_interrupt(mbox->mbox.pdev, "VF(s) to PF", intr);
581 
582 	return IRQ_HANDLED;
583 }
584 
585 static int otx2_pfvf_mbox_init(struct otx2_nic *pf, int numvfs)
586 {
587 	void __iomem *hwbase;
588 	struct mbox *mbox;
589 	int err, vf;
590 	u64 base;
591 
592 	if (!numvfs)
593 		return -EINVAL;
594 
595 	pf->mbox_pfvf = devm_kcalloc(&pf->pdev->dev, numvfs,
596 				     sizeof(struct mbox), GFP_KERNEL);
597 	if (!pf->mbox_pfvf)
598 		return -ENOMEM;
599 
600 	pf->mbox_pfvf_wq = alloc_ordered_workqueue("otx2_pfvf_mailbox",
601 						   WQ_HIGHPRI | WQ_MEM_RECLAIM);
602 	if (!pf->mbox_pfvf_wq)
603 		return -ENOMEM;
604 
605 	/* On CN10K platform, PF <-> VF mailbox region follows after
606 	 * PF <-> AF mailbox region.
607 	 */
608 	if (test_bit(CN10K_MBOX, &pf->hw.cap_flag))
609 		base = pci_resource_start(pf->pdev, PCI_MBOX_BAR_NUM) +
610 		       MBOX_SIZE;
611 	else
612 		base = readq((void __iomem *)((u64)pf->reg_base +
613 					      RVU_PF_VF_BAR4_ADDR));
614 
615 	hwbase = ioremap_wc(base, MBOX_SIZE * pf->total_vfs);
616 	if (!hwbase) {
617 		err = -ENOMEM;
618 		goto free_wq;
619 	}
620 
621 	mbox = &pf->mbox_pfvf[0];
622 	err = otx2_mbox_init(&mbox->mbox, hwbase, pf->pdev, pf->reg_base,
623 			     MBOX_DIR_PFVF, numvfs);
624 	if (err)
625 		goto free_iomem;
626 
627 	err = otx2_mbox_init(&mbox->mbox_up, hwbase, pf->pdev, pf->reg_base,
628 			     MBOX_DIR_PFVF_UP, numvfs);
629 	if (err)
630 		goto free_iomem;
631 
632 	for (vf = 0; vf < numvfs; vf++) {
633 		mbox->pfvf = pf;
634 		INIT_WORK(&mbox->mbox_wrk, otx2_pfvf_mbox_handler);
635 		INIT_WORK(&mbox->mbox_up_wrk, otx2_pfvf_mbox_up_handler);
636 		mbox++;
637 	}
638 
639 	return 0;
640 
641 free_iomem:
642 	if (hwbase)
643 		iounmap(hwbase);
644 free_wq:
645 	destroy_workqueue(pf->mbox_pfvf_wq);
646 	return err;
647 }
648 
649 static void otx2_pfvf_mbox_destroy(struct otx2_nic *pf)
650 {
651 	struct mbox *mbox = &pf->mbox_pfvf[0];
652 
653 	if (!mbox)
654 		return;
655 
656 	if (pf->mbox_pfvf_wq) {
657 		destroy_workqueue(pf->mbox_pfvf_wq);
658 		pf->mbox_pfvf_wq = NULL;
659 	}
660 
661 	if (mbox->mbox.hwbase)
662 		iounmap(mbox->mbox.hwbase);
663 
664 	otx2_mbox_destroy(&mbox->mbox);
665 }
666 
667 static void otx2_enable_pfvf_mbox_intr(struct otx2_nic *pf, int numvfs)
668 {
669 	/* Clear PF <=> VF mailbox IRQ */
670 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(0), ~0ull);
671 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(1), ~0ull);
672 
673 	/* Enable PF <=> VF mailbox IRQ */
674 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INT_ENA_W1SX(0), INTR_MASK(numvfs));
675 	if (numvfs > 64) {
676 		numvfs -= 64;
677 		otx2_write64(pf, RVU_PF_VFPF_MBOX_INT_ENA_W1SX(1),
678 			     INTR_MASK(numvfs));
679 	}
680 }
681 
682 static void otx2_disable_pfvf_mbox_intr(struct otx2_nic *pf, int numvfs)
683 {
684 	int vector;
685 
686 	/* Disable PF <=> VF mailbox IRQ */
687 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INT_ENA_W1CX(0), ~0ull);
688 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INT_ENA_W1CX(1), ~0ull);
689 
690 	otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(0), ~0ull);
691 	vector = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFPF_MBOX0);
692 	free_irq(vector, pf);
693 
694 	if (numvfs > 64) {
695 		otx2_write64(pf, RVU_PF_VFPF_MBOX_INTX(1), ~0ull);
696 		vector = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFPF_MBOX1);
697 		free_irq(vector, pf);
698 	}
699 }
700 
701 static int otx2_register_pfvf_mbox_intr(struct otx2_nic *pf, int numvfs)
702 {
703 	struct otx2_hw *hw = &pf->hw;
704 	char *irq_name;
705 	int err;
706 
707 	/* Register MBOX0 interrupt handler */
708 	irq_name = &hw->irq_name[RVU_PF_INT_VEC_VFPF_MBOX0 * NAME_SIZE];
709 	if (pf->pcifunc)
710 		snprintf(irq_name, NAME_SIZE,
711 			 "RVUPF%d_VF Mbox0", rvu_get_pf(pf->pcifunc));
712 	else
713 		snprintf(irq_name, NAME_SIZE, "RVUPF_VF Mbox0");
714 	err = request_irq(pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_VFPF_MBOX0),
715 			  otx2_pfvf_mbox_intr_handler, 0, irq_name, pf);
716 	if (err) {
717 		dev_err(pf->dev,
718 			"RVUPF: IRQ registration failed for PFVF mbox0 irq\n");
719 		return err;
720 	}
721 
722 	if (numvfs > 64) {
723 		/* Register MBOX1 interrupt handler */
724 		irq_name = &hw->irq_name[RVU_PF_INT_VEC_VFPF_MBOX1 * NAME_SIZE];
725 		if (pf->pcifunc)
726 			snprintf(irq_name, NAME_SIZE,
727 				 "RVUPF%d_VF Mbox1", rvu_get_pf(pf->pcifunc));
728 		else
729 			snprintf(irq_name, NAME_SIZE, "RVUPF_VF Mbox1");
730 		err = request_irq(pci_irq_vector(pf->pdev,
731 						 RVU_PF_INT_VEC_VFPF_MBOX1),
732 						 otx2_pfvf_mbox_intr_handler,
733 						 0, irq_name, pf);
734 		if (err) {
735 			dev_err(pf->dev,
736 				"RVUPF: IRQ registration failed for PFVF mbox1 irq\n");
737 			return err;
738 		}
739 	}
740 
741 	otx2_enable_pfvf_mbox_intr(pf, numvfs);
742 
743 	return 0;
744 }
745 
746 static void otx2_process_pfaf_mbox_msg(struct otx2_nic *pf,
747 				       struct mbox_msghdr *msg)
748 {
749 	int devid;
750 
751 	if (msg->id >= MBOX_MSG_MAX) {
752 		dev_err(pf->dev,
753 			"Mbox msg with unknown ID 0x%x\n", msg->id);
754 		return;
755 	}
756 
757 	if (msg->sig != OTX2_MBOX_RSP_SIG) {
758 		dev_err(pf->dev,
759 			"Mbox msg with wrong signature %x, ID 0x%x\n",
760 			 msg->sig, msg->id);
761 		return;
762 	}
763 
764 	/* message response heading VF */
765 	devid = msg->pcifunc & RVU_PFVF_FUNC_MASK;
766 	if (devid) {
767 		struct otx2_vf_config *config = &pf->vf_configs[devid - 1];
768 		struct delayed_work *dwork;
769 
770 		switch (msg->id) {
771 		case MBOX_MSG_NIX_LF_START_RX:
772 			config->intf_down = false;
773 			dwork = &config->link_event_work;
774 			schedule_delayed_work(dwork, msecs_to_jiffies(100));
775 			break;
776 		case MBOX_MSG_NIX_LF_STOP_RX:
777 			config->intf_down = true;
778 			break;
779 		}
780 
781 		return;
782 	}
783 
784 	switch (msg->id) {
785 	case MBOX_MSG_READY:
786 		pf->pcifunc = msg->pcifunc;
787 		break;
788 	case MBOX_MSG_MSIX_OFFSET:
789 		mbox_handler_msix_offset(pf, (struct msix_offset_rsp *)msg);
790 		break;
791 	case MBOX_MSG_NPA_LF_ALLOC:
792 		mbox_handler_npa_lf_alloc(pf, (struct npa_lf_alloc_rsp *)msg);
793 		break;
794 	case MBOX_MSG_NIX_LF_ALLOC:
795 		mbox_handler_nix_lf_alloc(pf, (struct nix_lf_alloc_rsp *)msg);
796 		break;
797 	case MBOX_MSG_NIX_BP_ENABLE:
798 		mbox_handler_nix_bp_enable(pf, (struct nix_bp_cfg_rsp *)msg);
799 		break;
800 	case MBOX_MSG_CGX_STATS:
801 		mbox_handler_cgx_stats(pf, (struct cgx_stats_rsp *)msg);
802 		break;
803 	case MBOX_MSG_CGX_FEC_STATS:
804 		mbox_handler_cgx_fec_stats(pf, (struct cgx_fec_stats_rsp *)msg);
805 		break;
806 	default:
807 		if (msg->rc)
808 			dev_err(pf->dev,
809 				"Mbox msg response has err %d, ID 0x%x\n",
810 				msg->rc, msg->id);
811 		break;
812 	}
813 }
814 
815 static void otx2_pfaf_mbox_handler(struct work_struct *work)
816 {
817 	struct otx2_mbox_dev *mdev;
818 	struct mbox_hdr *rsp_hdr;
819 	struct mbox_msghdr *msg;
820 	struct otx2_mbox *mbox;
821 	struct mbox *af_mbox;
822 	struct otx2_nic *pf;
823 	int offset, id;
824 
825 	af_mbox = container_of(work, struct mbox, mbox_wrk);
826 	mbox = &af_mbox->mbox;
827 	mdev = &mbox->dev[0];
828 	rsp_hdr = (struct mbox_hdr *)(mdev->mbase + mbox->rx_start);
829 
830 	offset = mbox->rx_start + ALIGN(sizeof(*rsp_hdr), MBOX_MSG_ALIGN);
831 	pf = af_mbox->pfvf;
832 
833 	for (id = 0; id < af_mbox->num_msgs; id++) {
834 		msg = (struct mbox_msghdr *)(mdev->mbase + offset);
835 		otx2_process_pfaf_mbox_msg(pf, msg);
836 		offset = mbox->rx_start + msg->next_msgoff;
837 		if (mdev->msgs_acked == (af_mbox->num_msgs - 1))
838 			__otx2_mbox_reset(mbox, 0);
839 		mdev->msgs_acked++;
840 	}
841 
842 }
843 
844 static void otx2_handle_link_event(struct otx2_nic *pf)
845 {
846 	struct cgx_link_user_info *linfo = &pf->linfo;
847 	struct net_device *netdev = pf->netdev;
848 
849 	pr_info("%s NIC Link is %s %d Mbps %s duplex\n", netdev->name,
850 		linfo->link_up ? "UP" : "DOWN", linfo->speed,
851 		linfo->full_duplex ? "Full" : "Half");
852 	if (linfo->link_up) {
853 		netif_carrier_on(netdev);
854 		netif_tx_start_all_queues(netdev);
855 	} else {
856 		netif_tx_stop_all_queues(netdev);
857 		netif_carrier_off(netdev);
858 	}
859 }
860 
861 int otx2_mbox_up_handler_mcs_intr_notify(struct otx2_nic *pf,
862 					 struct mcs_intr_info *event,
863 					 struct msg_rsp *rsp)
864 {
865 	cn10k_handle_mcs_event(pf, event);
866 
867 	return 0;
868 }
869 
870 int otx2_mbox_up_handler_cgx_link_event(struct otx2_nic *pf,
871 					struct cgx_link_info_msg *msg,
872 					struct msg_rsp *rsp)
873 {
874 	int i;
875 
876 	/* Copy the link info sent by AF */
877 	pf->linfo = msg->link_info;
878 
879 	/* notify VFs about link event */
880 	for (i = 0; i < pci_num_vf(pf->pdev); i++) {
881 		struct otx2_vf_config *config = &pf->vf_configs[i];
882 		struct delayed_work *dwork = &config->link_event_work;
883 
884 		if (config->intf_down)
885 			continue;
886 
887 		schedule_delayed_work(dwork, msecs_to_jiffies(100));
888 	}
889 
890 	/* interface has not been fully configured yet */
891 	if (pf->flags & OTX2_FLAG_INTF_DOWN)
892 		return 0;
893 
894 	otx2_handle_link_event(pf);
895 	return 0;
896 }
897 
898 static int otx2_process_mbox_msg_up(struct otx2_nic *pf,
899 				    struct mbox_msghdr *req)
900 {
901 	/* Check if valid, if not reply with a invalid msg */
902 	if (req->sig != OTX2_MBOX_REQ_SIG) {
903 		otx2_reply_invalid_msg(&pf->mbox.mbox_up, 0, 0, req->id);
904 		return -ENODEV;
905 	}
906 
907 	switch (req->id) {
908 #define M(_name, _id, _fn_name, _req_type, _rsp_type)			\
909 	case _id: {							\
910 		struct _rsp_type *rsp;					\
911 		int err;						\
912 									\
913 		rsp = (struct _rsp_type *)otx2_mbox_alloc_msg(		\
914 			&pf->mbox.mbox_up, 0,				\
915 			sizeof(struct _rsp_type));			\
916 		if (!rsp)						\
917 			return -ENOMEM;					\
918 									\
919 		rsp->hdr.id = _id;					\
920 		rsp->hdr.sig = OTX2_MBOX_RSP_SIG;			\
921 		rsp->hdr.pcifunc = 0;					\
922 		rsp->hdr.rc = 0;					\
923 									\
924 		err = otx2_mbox_up_handler_ ## _fn_name(		\
925 			pf, (struct _req_type *)req, rsp);		\
926 		return err;						\
927 	}
928 MBOX_UP_CGX_MESSAGES
929 MBOX_UP_MCS_MESSAGES
930 #undef M
931 		break;
932 	default:
933 		otx2_reply_invalid_msg(&pf->mbox.mbox_up, 0, 0, req->id);
934 		return -ENODEV;
935 	}
936 	return 0;
937 }
938 
939 static void otx2_pfaf_mbox_up_handler(struct work_struct *work)
940 {
941 	struct mbox *af_mbox = container_of(work, struct mbox, mbox_up_wrk);
942 	struct otx2_mbox *mbox = &af_mbox->mbox_up;
943 	struct otx2_mbox_dev *mdev = &mbox->dev[0];
944 	struct otx2_nic *pf = af_mbox->pfvf;
945 	int offset, id, devid = 0;
946 	struct mbox_hdr *rsp_hdr;
947 	struct mbox_msghdr *msg;
948 
949 	rsp_hdr = (struct mbox_hdr *)(mdev->mbase + mbox->rx_start);
950 
951 	offset = mbox->rx_start + ALIGN(sizeof(*rsp_hdr), MBOX_MSG_ALIGN);
952 
953 	for (id = 0; id < af_mbox->up_num_msgs; id++) {
954 		msg = (struct mbox_msghdr *)(mdev->mbase + offset);
955 
956 		devid = msg->pcifunc & RVU_PFVF_FUNC_MASK;
957 		/* Skip processing VF's messages */
958 		if (!devid)
959 			otx2_process_mbox_msg_up(pf, msg);
960 		offset = mbox->rx_start + msg->next_msgoff;
961 	}
962 	if (devid) {
963 		otx2_forward_vf_mbox_msgs(pf, &pf->mbox.mbox_up,
964 					  MBOX_DIR_PFVF_UP, devid - 1,
965 					  af_mbox->up_num_msgs);
966 		return;
967 	}
968 
969 	otx2_mbox_msg_send(mbox, 0);
970 }
971 
972 static irqreturn_t otx2_pfaf_mbox_intr_handler(int irq, void *pf_irq)
973 {
974 	struct otx2_nic *pf = (struct otx2_nic *)pf_irq;
975 	struct mbox *mbox;
976 
977 	/* Clear the IRQ */
978 	otx2_write64(pf, RVU_PF_INT, BIT_ULL(0));
979 
980 	mbox = &pf->mbox;
981 
982 	trace_otx2_msg_interrupt(mbox->mbox.pdev, "AF to PF", BIT_ULL(0));
983 
984 	otx2_queue_work(mbox, pf->mbox_wq, 0, 1, 1, TYPE_PFAF);
985 
986 	return IRQ_HANDLED;
987 }
988 
989 static void otx2_disable_mbox_intr(struct otx2_nic *pf)
990 {
991 	int vector = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_AFPF_MBOX);
992 
993 	/* Disable AF => PF mailbox IRQ */
994 	otx2_write64(pf, RVU_PF_INT_ENA_W1C, BIT_ULL(0));
995 	free_irq(vector, pf);
996 }
997 
998 static int otx2_register_mbox_intr(struct otx2_nic *pf, bool probe_af)
999 {
1000 	struct otx2_hw *hw = &pf->hw;
1001 	struct msg_req *req;
1002 	char *irq_name;
1003 	int err;
1004 
1005 	/* Register mailbox interrupt handler */
1006 	irq_name = &hw->irq_name[RVU_PF_INT_VEC_AFPF_MBOX * NAME_SIZE];
1007 	snprintf(irq_name, NAME_SIZE, "RVUPFAF Mbox");
1008 	err = request_irq(pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_AFPF_MBOX),
1009 			  otx2_pfaf_mbox_intr_handler, 0, irq_name, pf);
1010 	if (err) {
1011 		dev_err(pf->dev,
1012 			"RVUPF: IRQ registration failed for PFAF mbox irq\n");
1013 		return err;
1014 	}
1015 
1016 	/* Enable mailbox interrupt for msgs coming from AF.
1017 	 * First clear to avoid spurious interrupts, if any.
1018 	 */
1019 	otx2_write64(pf, RVU_PF_INT, BIT_ULL(0));
1020 	otx2_write64(pf, RVU_PF_INT_ENA_W1S, BIT_ULL(0));
1021 
1022 	if (!probe_af)
1023 		return 0;
1024 
1025 	/* Check mailbox communication with AF */
1026 	req = otx2_mbox_alloc_msg_ready(&pf->mbox);
1027 	if (!req) {
1028 		otx2_disable_mbox_intr(pf);
1029 		return -ENOMEM;
1030 	}
1031 	err = otx2_sync_mbox_msg(&pf->mbox);
1032 	if (err) {
1033 		dev_warn(pf->dev,
1034 			 "AF not responding to mailbox, deferring probe\n");
1035 		otx2_disable_mbox_intr(pf);
1036 		return -EPROBE_DEFER;
1037 	}
1038 
1039 	return 0;
1040 }
1041 
1042 static void otx2_pfaf_mbox_destroy(struct otx2_nic *pf)
1043 {
1044 	struct mbox *mbox = &pf->mbox;
1045 
1046 	if (pf->mbox_wq) {
1047 		destroy_workqueue(pf->mbox_wq);
1048 		pf->mbox_wq = NULL;
1049 	}
1050 
1051 	if (mbox->mbox.hwbase)
1052 		iounmap((void __iomem *)mbox->mbox.hwbase);
1053 
1054 	otx2_mbox_destroy(&mbox->mbox);
1055 	otx2_mbox_destroy(&mbox->mbox_up);
1056 }
1057 
1058 static int otx2_pfaf_mbox_init(struct otx2_nic *pf)
1059 {
1060 	struct mbox *mbox = &pf->mbox;
1061 	void __iomem *hwbase;
1062 	int err;
1063 
1064 	mbox->pfvf = pf;
1065 	pf->mbox_wq = alloc_ordered_workqueue("otx2_pfaf_mailbox",
1066 					      WQ_HIGHPRI | WQ_MEM_RECLAIM);
1067 	if (!pf->mbox_wq)
1068 		return -ENOMEM;
1069 
1070 	/* Mailbox is a reserved memory (in RAM) region shared between
1071 	 * admin function (i.e AF) and this PF, shouldn't be mapped as
1072 	 * device memory to allow unaligned accesses.
1073 	 */
1074 	hwbase = ioremap_wc(pci_resource_start(pf->pdev, PCI_MBOX_BAR_NUM),
1075 			    MBOX_SIZE);
1076 	if (!hwbase) {
1077 		dev_err(pf->dev, "Unable to map PFAF mailbox region\n");
1078 		err = -ENOMEM;
1079 		goto exit;
1080 	}
1081 
1082 	err = otx2_mbox_init(&mbox->mbox, hwbase, pf->pdev, pf->reg_base,
1083 			     MBOX_DIR_PFAF, 1);
1084 	if (err)
1085 		goto exit;
1086 
1087 	err = otx2_mbox_init(&mbox->mbox_up, hwbase, pf->pdev, pf->reg_base,
1088 			     MBOX_DIR_PFAF_UP, 1);
1089 	if (err)
1090 		goto exit;
1091 
1092 	err = otx2_mbox_bbuf_init(mbox, pf->pdev);
1093 	if (err)
1094 		goto exit;
1095 
1096 	INIT_WORK(&mbox->mbox_wrk, otx2_pfaf_mbox_handler);
1097 	INIT_WORK(&mbox->mbox_up_wrk, otx2_pfaf_mbox_up_handler);
1098 	mutex_init(&mbox->lock);
1099 
1100 	return 0;
1101 exit:
1102 	otx2_pfaf_mbox_destroy(pf);
1103 	return err;
1104 }
1105 
1106 static int otx2_cgx_config_linkevents(struct otx2_nic *pf, bool enable)
1107 {
1108 	struct msg_req *msg;
1109 	int err;
1110 
1111 	mutex_lock(&pf->mbox.lock);
1112 	if (enable)
1113 		msg = otx2_mbox_alloc_msg_cgx_start_linkevents(&pf->mbox);
1114 	else
1115 		msg = otx2_mbox_alloc_msg_cgx_stop_linkevents(&pf->mbox);
1116 
1117 	if (!msg) {
1118 		mutex_unlock(&pf->mbox.lock);
1119 		return -ENOMEM;
1120 	}
1121 
1122 	err = otx2_sync_mbox_msg(&pf->mbox);
1123 	mutex_unlock(&pf->mbox.lock);
1124 	return err;
1125 }
1126 
1127 static int otx2_cgx_config_loopback(struct otx2_nic *pf, bool enable)
1128 {
1129 	struct msg_req *msg;
1130 	int err;
1131 
1132 	if (enable && !bitmap_empty(pf->flow_cfg->dmacflt_bmap,
1133 				    pf->flow_cfg->dmacflt_max_flows))
1134 		netdev_warn(pf->netdev,
1135 			    "CGX/RPM internal loopback might not work as DMAC filters are active\n");
1136 
1137 	mutex_lock(&pf->mbox.lock);
1138 	if (enable)
1139 		msg = otx2_mbox_alloc_msg_cgx_intlbk_enable(&pf->mbox);
1140 	else
1141 		msg = otx2_mbox_alloc_msg_cgx_intlbk_disable(&pf->mbox);
1142 
1143 	if (!msg) {
1144 		mutex_unlock(&pf->mbox.lock);
1145 		return -ENOMEM;
1146 	}
1147 
1148 	err = otx2_sync_mbox_msg(&pf->mbox);
1149 	mutex_unlock(&pf->mbox.lock);
1150 	return err;
1151 }
1152 
1153 int otx2_set_real_num_queues(struct net_device *netdev,
1154 			     int tx_queues, int rx_queues)
1155 {
1156 	int err;
1157 
1158 	err = netif_set_real_num_tx_queues(netdev, tx_queues);
1159 	if (err) {
1160 		netdev_err(netdev,
1161 			   "Failed to set no of Tx queues: %d\n", tx_queues);
1162 		return err;
1163 	}
1164 
1165 	err = netif_set_real_num_rx_queues(netdev, rx_queues);
1166 	if (err)
1167 		netdev_err(netdev,
1168 			   "Failed to set no of Rx queues: %d\n", rx_queues);
1169 	return err;
1170 }
1171 EXPORT_SYMBOL(otx2_set_real_num_queues);
1172 
1173 static char *nix_sqoperr_e_str[NIX_SQOPERR_MAX] = {
1174 	"NIX_SQOPERR_OOR",
1175 	"NIX_SQOPERR_CTX_FAULT",
1176 	"NIX_SQOPERR_CTX_POISON",
1177 	"NIX_SQOPERR_DISABLED",
1178 	"NIX_SQOPERR_SIZE_ERR",
1179 	"NIX_SQOPERR_OFLOW",
1180 	"NIX_SQOPERR_SQB_NULL",
1181 	"NIX_SQOPERR_SQB_FAULT",
1182 	"NIX_SQOPERR_SQE_SZ_ZERO",
1183 };
1184 
1185 static char *nix_mnqerr_e_str[NIX_MNQERR_MAX] = {
1186 	"NIX_MNQERR_SQ_CTX_FAULT",
1187 	"NIX_MNQERR_SQ_CTX_POISON",
1188 	"NIX_MNQERR_SQB_FAULT",
1189 	"NIX_MNQERR_SQB_POISON",
1190 	"NIX_MNQERR_TOTAL_ERR",
1191 	"NIX_MNQERR_LSO_ERR",
1192 	"NIX_MNQERR_CQ_QUERY_ERR",
1193 	"NIX_MNQERR_MAX_SQE_SIZE_ERR",
1194 	"NIX_MNQERR_MAXLEN_ERR",
1195 	"NIX_MNQERR_SQE_SIZEM1_ZERO",
1196 };
1197 
1198 static char *nix_snd_status_e_str[NIX_SND_STATUS_MAX] =  {
1199 	[NIX_SND_STATUS_GOOD] = "NIX_SND_STATUS_GOOD",
1200 	[NIX_SND_STATUS_SQ_CTX_FAULT] = "NIX_SND_STATUS_SQ_CTX_FAULT",
1201 	[NIX_SND_STATUS_SQ_CTX_POISON] = "NIX_SND_STATUS_SQ_CTX_POISON",
1202 	[NIX_SND_STATUS_SQB_FAULT] = "NIX_SND_STATUS_SQB_FAULT",
1203 	[NIX_SND_STATUS_SQB_POISON] = "NIX_SND_STATUS_SQB_POISON",
1204 	[NIX_SND_STATUS_HDR_ERR] = "NIX_SND_STATUS_HDR_ERR",
1205 	[NIX_SND_STATUS_EXT_ERR] = "NIX_SND_STATUS_EXT_ERR",
1206 	[NIX_SND_STATUS_JUMP_FAULT] = "NIX_SND_STATUS_JUMP_FAULT",
1207 	[NIX_SND_STATUS_JUMP_POISON] = "NIX_SND_STATUS_JUMP_POISON",
1208 	[NIX_SND_STATUS_CRC_ERR] = "NIX_SND_STATUS_CRC_ERR",
1209 	[NIX_SND_STATUS_IMM_ERR] = "NIX_SND_STATUS_IMM_ERR",
1210 	[NIX_SND_STATUS_SG_ERR] = "NIX_SND_STATUS_SG_ERR",
1211 	[NIX_SND_STATUS_MEM_ERR] = "NIX_SND_STATUS_MEM_ERR",
1212 	[NIX_SND_STATUS_INVALID_SUBDC] = "NIX_SND_STATUS_INVALID_SUBDC",
1213 	[NIX_SND_STATUS_SUBDC_ORDER_ERR] = "NIX_SND_STATUS_SUBDC_ORDER_ERR",
1214 	[NIX_SND_STATUS_DATA_FAULT] = "NIX_SND_STATUS_DATA_FAULT",
1215 	[NIX_SND_STATUS_DATA_POISON] = "NIX_SND_STATUS_DATA_POISON",
1216 	[NIX_SND_STATUS_NPC_DROP_ACTION] = "NIX_SND_STATUS_NPC_DROP_ACTION",
1217 	[NIX_SND_STATUS_LOCK_VIOL] = "NIX_SND_STATUS_LOCK_VIOL",
1218 	[NIX_SND_STATUS_NPC_UCAST_CHAN_ERR] = "NIX_SND_STAT_NPC_UCAST_CHAN_ERR",
1219 	[NIX_SND_STATUS_NPC_MCAST_CHAN_ERR] = "NIX_SND_STAT_NPC_MCAST_CHAN_ERR",
1220 	[NIX_SND_STATUS_NPC_MCAST_ABORT] = "NIX_SND_STATUS_NPC_MCAST_ABORT",
1221 	[NIX_SND_STATUS_NPC_VTAG_PTR_ERR] = "NIX_SND_STATUS_NPC_VTAG_PTR_ERR",
1222 	[NIX_SND_STATUS_NPC_VTAG_SIZE_ERR] = "NIX_SND_STATUS_NPC_VTAG_SIZE_ERR",
1223 	[NIX_SND_STATUS_SEND_MEM_FAULT] = "NIX_SND_STATUS_SEND_MEM_FAULT",
1224 	[NIX_SND_STATUS_SEND_STATS_ERR] = "NIX_SND_STATUS_SEND_STATS_ERR",
1225 };
1226 
1227 static irqreturn_t otx2_q_intr_handler(int irq, void *data)
1228 {
1229 	struct otx2_nic *pf = data;
1230 	struct otx2_snd_queue *sq;
1231 	u64 val, *ptr;
1232 	u64 qidx = 0;
1233 
1234 	/* CQ */
1235 	for (qidx = 0; qidx < pf->qset.cq_cnt; qidx++) {
1236 		ptr = otx2_get_regaddr(pf, NIX_LF_CQ_OP_INT);
1237 		val = otx2_atomic64_add((qidx << 44), ptr);
1238 
1239 		otx2_write64(pf, NIX_LF_CQ_OP_INT, (qidx << 44) |
1240 			     (val & NIX_CQERRINT_BITS));
1241 		if (!(val & (NIX_CQERRINT_BITS | BIT_ULL(42))))
1242 			continue;
1243 
1244 		if (val & BIT_ULL(42)) {
1245 			netdev_err(pf->netdev,
1246 				   "CQ%lld: error reading NIX_LF_CQ_OP_INT, NIX_LF_ERR_INT 0x%llx\n",
1247 				   qidx, otx2_read64(pf, NIX_LF_ERR_INT));
1248 		} else {
1249 			if (val & BIT_ULL(NIX_CQERRINT_DOOR_ERR))
1250 				netdev_err(pf->netdev, "CQ%lld: Doorbell error",
1251 					   qidx);
1252 			if (val & BIT_ULL(NIX_CQERRINT_CQE_FAULT))
1253 				netdev_err(pf->netdev,
1254 					   "CQ%lld: Memory fault on CQE write to LLC/DRAM",
1255 					   qidx);
1256 		}
1257 
1258 		schedule_work(&pf->reset_task);
1259 	}
1260 
1261 	/* SQ */
1262 	for (qidx = 0; qidx < otx2_get_total_tx_queues(pf); qidx++) {
1263 		u64 sq_op_err_dbg, mnq_err_dbg, snd_err_dbg;
1264 		u8 sq_op_err_code, mnq_err_code, snd_err_code;
1265 
1266 		sq = &pf->qset.sq[qidx];
1267 		if (!sq->sqb_ptrs)
1268 			continue;
1269 
1270 		/* Below debug registers captures first errors corresponding to
1271 		 * those registers. We don't have to check against SQ qid as
1272 		 * these are fatal errors.
1273 		 */
1274 
1275 		ptr = otx2_get_regaddr(pf, NIX_LF_SQ_OP_INT);
1276 		val = otx2_atomic64_add((qidx << 44), ptr);
1277 		otx2_write64(pf, NIX_LF_SQ_OP_INT, (qidx << 44) |
1278 			     (val & NIX_SQINT_BITS));
1279 
1280 		if (val & BIT_ULL(42)) {
1281 			netdev_err(pf->netdev,
1282 				   "SQ%lld: error reading NIX_LF_SQ_OP_INT, NIX_LF_ERR_INT 0x%llx\n",
1283 				   qidx, otx2_read64(pf, NIX_LF_ERR_INT));
1284 			goto done;
1285 		}
1286 
1287 		sq_op_err_dbg = otx2_read64(pf, NIX_LF_SQ_OP_ERR_DBG);
1288 		if (!(sq_op_err_dbg & BIT(44)))
1289 			goto chk_mnq_err_dbg;
1290 
1291 		sq_op_err_code = FIELD_GET(GENMASK(7, 0), sq_op_err_dbg);
1292 		netdev_err(pf->netdev,
1293 			   "SQ%lld: NIX_LF_SQ_OP_ERR_DBG(0x%llx)  err=%s(%#x)\n",
1294 			   qidx, sq_op_err_dbg,
1295 			   nix_sqoperr_e_str[sq_op_err_code],
1296 			   sq_op_err_code);
1297 
1298 		otx2_write64(pf, NIX_LF_SQ_OP_ERR_DBG, BIT_ULL(44));
1299 
1300 		if (sq_op_err_code == NIX_SQOPERR_SQB_NULL)
1301 			goto chk_mnq_err_dbg;
1302 
1303 		/* Err is not NIX_SQOPERR_SQB_NULL, call aq function to read SQ structure.
1304 		 * TODO: But we are in irq context. How to call mbox functions which does sleep
1305 		 */
1306 
1307 chk_mnq_err_dbg:
1308 		mnq_err_dbg = otx2_read64(pf, NIX_LF_MNQ_ERR_DBG);
1309 		if (!(mnq_err_dbg & BIT(44)))
1310 			goto chk_snd_err_dbg;
1311 
1312 		mnq_err_code = FIELD_GET(GENMASK(7, 0), mnq_err_dbg);
1313 		netdev_err(pf->netdev,
1314 			   "SQ%lld: NIX_LF_MNQ_ERR_DBG(0x%llx)  err=%s(%#x)\n",
1315 			   qidx, mnq_err_dbg,  nix_mnqerr_e_str[mnq_err_code],
1316 			   mnq_err_code);
1317 		otx2_write64(pf, NIX_LF_MNQ_ERR_DBG, BIT_ULL(44));
1318 
1319 chk_snd_err_dbg:
1320 		snd_err_dbg = otx2_read64(pf, NIX_LF_SEND_ERR_DBG);
1321 		if (snd_err_dbg & BIT(44)) {
1322 			snd_err_code = FIELD_GET(GENMASK(7, 0), snd_err_dbg);
1323 			netdev_err(pf->netdev,
1324 				   "SQ%lld: NIX_LF_SND_ERR_DBG:0x%llx err=%s(%#x)\n",
1325 				   qidx, snd_err_dbg,
1326 				   nix_snd_status_e_str[snd_err_code],
1327 				   snd_err_code);
1328 			otx2_write64(pf, NIX_LF_SEND_ERR_DBG, BIT_ULL(44));
1329 		}
1330 
1331 done:
1332 		/* Print values and reset */
1333 		if (val & BIT_ULL(NIX_SQINT_SQB_ALLOC_FAIL))
1334 			netdev_err(pf->netdev, "SQ%lld: SQB allocation failed",
1335 				   qidx);
1336 
1337 		schedule_work(&pf->reset_task);
1338 	}
1339 
1340 	return IRQ_HANDLED;
1341 }
1342 
1343 static irqreturn_t otx2_cq_intr_handler(int irq, void *cq_irq)
1344 {
1345 	struct otx2_cq_poll *cq_poll = (struct otx2_cq_poll *)cq_irq;
1346 	struct otx2_nic *pf = (struct otx2_nic *)cq_poll->dev;
1347 	int qidx = cq_poll->cint_idx;
1348 
1349 	/* Disable interrupts.
1350 	 *
1351 	 * Completion interrupts behave in a level-triggered interrupt
1352 	 * fashion, and hence have to be cleared only after it is serviced.
1353 	 */
1354 	otx2_write64(pf, NIX_LF_CINTX_ENA_W1C(qidx), BIT_ULL(0));
1355 
1356 	/* Schedule NAPI */
1357 	pf->napi_events++;
1358 	napi_schedule_irqoff(&cq_poll->napi);
1359 
1360 	return IRQ_HANDLED;
1361 }
1362 
1363 static void otx2_disable_napi(struct otx2_nic *pf)
1364 {
1365 	struct otx2_qset *qset = &pf->qset;
1366 	struct otx2_cq_poll *cq_poll;
1367 	int qidx;
1368 
1369 	for (qidx = 0; qidx < pf->hw.cint_cnt; qidx++) {
1370 		cq_poll = &qset->napi[qidx];
1371 		cancel_work_sync(&cq_poll->dim.work);
1372 		napi_disable(&cq_poll->napi);
1373 		netif_napi_del(&cq_poll->napi);
1374 	}
1375 }
1376 
1377 static void otx2_free_cq_res(struct otx2_nic *pf)
1378 {
1379 	struct otx2_qset *qset = &pf->qset;
1380 	struct otx2_cq_queue *cq;
1381 	int qidx;
1382 
1383 	/* Disable CQs */
1384 	otx2_ctx_disable(&pf->mbox, NIX_AQ_CTYPE_CQ, false);
1385 	for (qidx = 0; qidx < qset->cq_cnt; qidx++) {
1386 		cq = &qset->cq[qidx];
1387 		qmem_free(pf->dev, cq->cqe);
1388 	}
1389 }
1390 
1391 static void otx2_free_sq_res(struct otx2_nic *pf)
1392 {
1393 	struct otx2_qset *qset = &pf->qset;
1394 	struct otx2_snd_queue *sq;
1395 	int qidx;
1396 
1397 	/* Disable SQs */
1398 	otx2_ctx_disable(&pf->mbox, NIX_AQ_CTYPE_SQ, false);
1399 	/* Free SQB pointers */
1400 	otx2_sq_free_sqbs(pf);
1401 	for (qidx = 0; qidx < otx2_get_total_tx_queues(pf); qidx++) {
1402 		sq = &qset->sq[qidx];
1403 		/* Skip freeing Qos queues if they are not initialized */
1404 		if (!sq->sqe)
1405 			continue;
1406 		qmem_free(pf->dev, sq->sqe);
1407 		qmem_free(pf->dev, sq->tso_hdrs);
1408 		kfree(sq->sg);
1409 		kfree(sq->sqb_ptrs);
1410 	}
1411 }
1412 
1413 static int otx2_get_rbuf_size(struct otx2_nic *pf, int mtu)
1414 {
1415 	int frame_size;
1416 	int total_size;
1417 	int rbuf_size;
1418 
1419 	if (pf->hw.rbuf_len)
1420 		return ALIGN(pf->hw.rbuf_len, OTX2_ALIGN) + OTX2_HEAD_ROOM;
1421 
1422 	/* The data transferred by NIX to memory consists of actual packet
1423 	 * plus additional data which has timestamp and/or EDSA/HIGIG2
1424 	 * headers if interface is configured in corresponding modes.
1425 	 * NIX transfers entire data using 6 segments/buffers and writes
1426 	 * a CQE_RX descriptor with those segment addresses. First segment
1427 	 * has additional data prepended to packet. Also software omits a
1428 	 * headroom of 128 bytes in each segment. Hence the total size of
1429 	 * memory needed to receive a packet with 'mtu' is:
1430 	 * frame size =  mtu + additional data;
1431 	 * memory = frame_size + headroom * 6;
1432 	 * each receive buffer size = memory / 6;
1433 	 */
1434 	frame_size = mtu + OTX2_ETH_HLEN + OTX2_HW_TIMESTAMP_LEN;
1435 	total_size = frame_size + OTX2_HEAD_ROOM * 6;
1436 	rbuf_size = total_size / 6;
1437 
1438 	return ALIGN(rbuf_size, 2048);
1439 }
1440 
1441 static int otx2_init_hw_resources(struct otx2_nic *pf)
1442 {
1443 	struct nix_lf_free_req *free_req;
1444 	struct mbox *mbox = &pf->mbox;
1445 	struct otx2_hw *hw = &pf->hw;
1446 	struct msg_req *req;
1447 	int err = 0, lvl;
1448 
1449 	/* Set required NPA LF's pool counts
1450 	 * Auras and Pools are used in a 1:1 mapping,
1451 	 * so, aura count = pool count.
1452 	 */
1453 	hw->rqpool_cnt = hw->rx_queues;
1454 	hw->sqpool_cnt = otx2_get_total_tx_queues(pf);
1455 	hw->pool_cnt = hw->rqpool_cnt + hw->sqpool_cnt;
1456 
1457 	/* Maximum hardware supported transmit length */
1458 	pf->tx_max_pktlen = pf->netdev->max_mtu + OTX2_ETH_HLEN;
1459 
1460 	pf->rbsize = otx2_get_rbuf_size(pf, pf->netdev->mtu);
1461 
1462 	mutex_lock(&mbox->lock);
1463 	/* NPA init */
1464 	err = otx2_config_npa(pf);
1465 	if (err)
1466 		goto exit;
1467 
1468 	/* NIX init */
1469 	err = otx2_config_nix(pf);
1470 	if (err)
1471 		goto err_free_npa_lf;
1472 
1473 	/* Enable backpressure for CGX mapped PF/VFs */
1474 	if (!is_otx2_lbkvf(pf->pdev))
1475 		otx2_nix_config_bp(pf, true);
1476 
1477 	/* Init Auras and pools used by NIX RQ, for free buffer ptrs */
1478 	err = otx2_rq_aura_pool_init(pf);
1479 	if (err) {
1480 		mutex_unlock(&mbox->lock);
1481 		goto err_free_nix_lf;
1482 	}
1483 	/* Init Auras and pools used by NIX SQ, for queueing SQEs */
1484 	err = otx2_sq_aura_pool_init(pf);
1485 	if (err) {
1486 		mutex_unlock(&mbox->lock);
1487 		goto err_free_rq_ptrs;
1488 	}
1489 
1490 	err = otx2_txsch_alloc(pf);
1491 	if (err) {
1492 		mutex_unlock(&mbox->lock);
1493 		goto err_free_sq_ptrs;
1494 	}
1495 
1496 #ifdef CONFIG_DCB
1497 	if (pf->pfc_en) {
1498 		err = otx2_pfc_txschq_alloc(pf);
1499 		if (err) {
1500 			mutex_unlock(&mbox->lock);
1501 			goto err_free_sq_ptrs;
1502 		}
1503 	}
1504 #endif
1505 
1506 	err = otx2_config_nix_queues(pf);
1507 	if (err) {
1508 		mutex_unlock(&mbox->lock);
1509 		goto err_free_txsch;
1510 	}
1511 
1512 	for (lvl = 0; lvl < NIX_TXSCH_LVL_CNT; lvl++) {
1513 		err = otx2_txschq_config(pf, lvl, 0, false);
1514 		if (err) {
1515 			mutex_unlock(&mbox->lock);
1516 			goto err_free_nix_queues;
1517 		}
1518 	}
1519 
1520 #ifdef CONFIG_DCB
1521 	if (pf->pfc_en) {
1522 		err = otx2_pfc_txschq_config(pf);
1523 		if (err) {
1524 			mutex_unlock(&mbox->lock);
1525 			goto err_free_nix_queues;
1526 		}
1527 	}
1528 #endif
1529 
1530 	mutex_unlock(&mbox->lock);
1531 	return err;
1532 
1533 err_free_nix_queues:
1534 	otx2_free_sq_res(pf);
1535 	otx2_free_cq_res(pf);
1536 	otx2_ctx_disable(mbox, NIX_AQ_CTYPE_RQ, false);
1537 err_free_txsch:
1538 	otx2_txschq_stop(pf);
1539 err_free_sq_ptrs:
1540 	otx2_sq_free_sqbs(pf);
1541 err_free_rq_ptrs:
1542 	otx2_free_aura_ptr(pf, AURA_NIX_RQ);
1543 	otx2_ctx_disable(mbox, NPA_AQ_CTYPE_POOL, true);
1544 	otx2_ctx_disable(mbox, NPA_AQ_CTYPE_AURA, true);
1545 	otx2_aura_pool_free(pf);
1546 err_free_nix_lf:
1547 	mutex_lock(&mbox->lock);
1548 	free_req = otx2_mbox_alloc_msg_nix_lf_free(mbox);
1549 	if (free_req) {
1550 		free_req->flags = NIX_LF_DISABLE_FLOWS;
1551 		if (otx2_sync_mbox_msg(mbox))
1552 			dev_err(pf->dev, "%s failed to free nixlf\n", __func__);
1553 	}
1554 err_free_npa_lf:
1555 	/* Reset NPA LF */
1556 	req = otx2_mbox_alloc_msg_npa_lf_free(mbox);
1557 	if (req) {
1558 		if (otx2_sync_mbox_msg(mbox))
1559 			dev_err(pf->dev, "%s failed to free npalf\n", __func__);
1560 	}
1561 exit:
1562 	mutex_unlock(&mbox->lock);
1563 	return err;
1564 }
1565 
1566 static void otx2_free_hw_resources(struct otx2_nic *pf)
1567 {
1568 	struct otx2_qset *qset = &pf->qset;
1569 	struct nix_lf_free_req *free_req;
1570 	struct mbox *mbox = &pf->mbox;
1571 	struct otx2_cq_queue *cq;
1572 	struct otx2_pool *pool;
1573 	struct msg_req *req;
1574 	int pool_id;
1575 	int qidx;
1576 
1577 	/* Ensure all SQE are processed */
1578 	otx2_sqb_flush(pf);
1579 
1580 	/* Stop transmission */
1581 	otx2_txschq_stop(pf);
1582 
1583 #ifdef CONFIG_DCB
1584 	if (pf->pfc_en)
1585 		otx2_pfc_txschq_stop(pf);
1586 #endif
1587 
1588 	otx2_clean_qos_queues(pf);
1589 
1590 	mutex_lock(&mbox->lock);
1591 	/* Disable backpressure */
1592 	if (!(pf->pcifunc & RVU_PFVF_FUNC_MASK))
1593 		otx2_nix_config_bp(pf, false);
1594 	mutex_unlock(&mbox->lock);
1595 
1596 	/* Disable RQs */
1597 	otx2_ctx_disable(mbox, NIX_AQ_CTYPE_RQ, false);
1598 
1599 	/*Dequeue all CQEs */
1600 	for (qidx = 0; qidx < qset->cq_cnt; qidx++) {
1601 		cq = &qset->cq[qidx];
1602 		if (cq->cq_type == CQ_RX)
1603 			otx2_cleanup_rx_cqes(pf, cq, qidx);
1604 		else
1605 			otx2_cleanup_tx_cqes(pf, cq);
1606 	}
1607 	otx2_free_pending_sqe(pf);
1608 
1609 	otx2_free_sq_res(pf);
1610 
1611 	/* Free RQ buffer pointers*/
1612 	otx2_free_aura_ptr(pf, AURA_NIX_RQ);
1613 
1614 	for (qidx = 0; qidx < pf->hw.rx_queues; qidx++) {
1615 		pool_id = otx2_get_pool_idx(pf, AURA_NIX_RQ, qidx);
1616 		pool = &pf->qset.pool[pool_id];
1617 		page_pool_destroy(pool->page_pool);
1618 		pool->page_pool = NULL;
1619 	}
1620 
1621 	otx2_free_cq_res(pf);
1622 
1623 	/* Free all ingress bandwidth profiles allocated */
1624 	cn10k_free_all_ipolicers(pf);
1625 
1626 	mutex_lock(&mbox->lock);
1627 	/* Reset NIX LF */
1628 	free_req = otx2_mbox_alloc_msg_nix_lf_free(mbox);
1629 	if (free_req) {
1630 		free_req->flags = NIX_LF_DISABLE_FLOWS;
1631 		if (!(pf->flags & OTX2_FLAG_PF_SHUTDOWN))
1632 			free_req->flags |= NIX_LF_DONT_FREE_TX_VTAG;
1633 		if (otx2_sync_mbox_msg(mbox))
1634 			dev_err(pf->dev, "%s failed to free nixlf\n", __func__);
1635 	}
1636 	mutex_unlock(&mbox->lock);
1637 
1638 	/* Disable NPA Pool and Aura hw context */
1639 	otx2_ctx_disable(mbox, NPA_AQ_CTYPE_POOL, true);
1640 	otx2_ctx_disable(mbox, NPA_AQ_CTYPE_AURA, true);
1641 	otx2_aura_pool_free(pf);
1642 
1643 	mutex_lock(&mbox->lock);
1644 	/* Reset NPA LF */
1645 	req = otx2_mbox_alloc_msg_npa_lf_free(mbox);
1646 	if (req) {
1647 		if (otx2_sync_mbox_msg(mbox))
1648 			dev_err(pf->dev, "%s failed to free npalf\n", __func__);
1649 	}
1650 	mutex_unlock(&mbox->lock);
1651 }
1652 
1653 static bool otx2_promisc_use_mce_list(struct otx2_nic *pfvf)
1654 {
1655 	int vf;
1656 
1657 	/* The AF driver will determine whether to allow the VF netdev or not */
1658 	if (is_otx2_vf(pfvf->pcifunc))
1659 		return true;
1660 
1661 	/* check if there are any trusted VFs associated with the PF netdev */
1662 	for (vf = 0; vf < pci_num_vf(pfvf->pdev); vf++)
1663 		if (pfvf->vf_configs[vf].trusted)
1664 			return true;
1665 	return false;
1666 }
1667 
1668 static void otx2_do_set_rx_mode(struct otx2_nic *pf)
1669 {
1670 	struct net_device *netdev = pf->netdev;
1671 	struct nix_rx_mode *req;
1672 	bool promisc = false;
1673 
1674 	if (!(netdev->flags & IFF_UP))
1675 		return;
1676 
1677 	if ((netdev->flags & IFF_PROMISC) ||
1678 	    (netdev_uc_count(netdev) > OTX2_MAX_UNICAST_FLOWS)) {
1679 		promisc = true;
1680 	}
1681 
1682 	/* Write unicast address to mcam entries or del from mcam */
1683 	if (!promisc && netdev->priv_flags & IFF_UNICAST_FLT)
1684 		__dev_uc_sync(netdev, otx2_add_macfilter, otx2_del_macfilter);
1685 
1686 	mutex_lock(&pf->mbox.lock);
1687 	req = otx2_mbox_alloc_msg_nix_set_rx_mode(&pf->mbox);
1688 	if (!req) {
1689 		mutex_unlock(&pf->mbox.lock);
1690 		return;
1691 	}
1692 
1693 	req->mode = NIX_RX_MODE_UCAST;
1694 
1695 	if (promisc)
1696 		req->mode |= NIX_RX_MODE_PROMISC;
1697 	if (netdev->flags & (IFF_ALLMULTI | IFF_MULTICAST))
1698 		req->mode |= NIX_RX_MODE_ALLMULTI;
1699 
1700 	if (otx2_promisc_use_mce_list(pf))
1701 		req->mode |= NIX_RX_MODE_USE_MCE;
1702 
1703 	otx2_sync_mbox_msg(&pf->mbox);
1704 	mutex_unlock(&pf->mbox.lock);
1705 }
1706 
1707 static void otx2_set_irq_coalesce(struct otx2_nic *pfvf)
1708 {
1709 	int cint;
1710 
1711 	for (cint = 0; cint < pfvf->hw.cint_cnt; cint++)
1712 		otx2_config_irq_coalescing(pfvf, cint);
1713 }
1714 
1715 static void otx2_dim_work(struct work_struct *w)
1716 {
1717 	struct dim_cq_moder cur_moder;
1718 	struct otx2_cq_poll *cq_poll;
1719 	struct otx2_nic *pfvf;
1720 	struct dim *dim;
1721 
1722 	dim = container_of(w, struct dim, work);
1723 	cur_moder = net_dim_get_rx_moderation(dim->mode, dim->profile_ix);
1724 	cq_poll = container_of(dim, struct otx2_cq_poll, dim);
1725 	pfvf = (struct otx2_nic *)cq_poll->dev;
1726 	pfvf->hw.cq_time_wait = (cur_moder.usec > CQ_TIMER_THRESH_MAX) ?
1727 		CQ_TIMER_THRESH_MAX : cur_moder.usec;
1728 	pfvf->hw.cq_ecount_wait = (cur_moder.pkts > NAPI_POLL_WEIGHT) ?
1729 		NAPI_POLL_WEIGHT : cur_moder.pkts;
1730 	otx2_set_irq_coalesce(pfvf);
1731 	dim->state = DIM_START_MEASURE;
1732 }
1733 
1734 int otx2_open(struct net_device *netdev)
1735 {
1736 	struct otx2_nic *pf = netdev_priv(netdev);
1737 	struct otx2_cq_poll *cq_poll = NULL;
1738 	struct otx2_qset *qset = &pf->qset;
1739 	int err = 0, qidx, vec;
1740 	char *irq_name;
1741 
1742 	netif_carrier_off(netdev);
1743 
1744 	/* RQ and SQs are mapped to different CQs,
1745 	 * so find out max CQ IRQs (i.e CINTs) needed.
1746 	 */
1747 	pf->hw.cint_cnt = max3(pf->hw.rx_queues, pf->hw.tx_queues,
1748 			       pf->hw.tc_tx_queues);
1749 
1750 	pf->qset.cq_cnt = pf->hw.rx_queues + otx2_get_total_tx_queues(pf);
1751 
1752 	qset->napi = kcalloc(pf->hw.cint_cnt, sizeof(*cq_poll), GFP_KERNEL);
1753 	if (!qset->napi)
1754 		return -ENOMEM;
1755 
1756 	/* CQ size of RQ */
1757 	qset->rqe_cnt = qset->rqe_cnt ? qset->rqe_cnt : Q_COUNT(Q_SIZE_256);
1758 	/* CQ size of SQ */
1759 	qset->sqe_cnt = qset->sqe_cnt ? qset->sqe_cnt : Q_COUNT(Q_SIZE_4K);
1760 
1761 	err = -ENOMEM;
1762 	qset->cq = kcalloc(pf->qset.cq_cnt,
1763 			   sizeof(struct otx2_cq_queue), GFP_KERNEL);
1764 	if (!qset->cq)
1765 		goto err_free_mem;
1766 
1767 	qset->sq = kcalloc(otx2_get_total_tx_queues(pf),
1768 			   sizeof(struct otx2_snd_queue), GFP_KERNEL);
1769 	if (!qset->sq)
1770 		goto err_free_mem;
1771 
1772 	qset->rq = kcalloc(pf->hw.rx_queues,
1773 			   sizeof(struct otx2_rcv_queue), GFP_KERNEL);
1774 	if (!qset->rq)
1775 		goto err_free_mem;
1776 
1777 	err = otx2_init_hw_resources(pf);
1778 	if (err)
1779 		goto err_free_mem;
1780 
1781 	/* Register NAPI handler */
1782 	for (qidx = 0; qidx < pf->hw.cint_cnt; qidx++) {
1783 		cq_poll = &qset->napi[qidx];
1784 		cq_poll->cint_idx = qidx;
1785 		/* RQ0 & SQ0 are mapped to CINT0 and so on..
1786 		 * 'cq_ids[0]' points to RQ's CQ and
1787 		 * 'cq_ids[1]' points to SQ's CQ and
1788 		 * 'cq_ids[2]' points to XDP's CQ and
1789 		 */
1790 		cq_poll->cq_ids[CQ_RX] =
1791 			(qidx <  pf->hw.rx_queues) ? qidx : CINT_INVALID_CQ;
1792 		cq_poll->cq_ids[CQ_TX] = (qidx < pf->hw.tx_queues) ?
1793 				      qidx + pf->hw.rx_queues : CINT_INVALID_CQ;
1794 		if (pf->xdp_prog)
1795 			cq_poll->cq_ids[CQ_XDP] = (qidx < pf->hw.xdp_queues) ?
1796 						  (qidx + pf->hw.rx_queues +
1797 						  pf->hw.tx_queues) :
1798 						  CINT_INVALID_CQ;
1799 		else
1800 			cq_poll->cq_ids[CQ_XDP] = CINT_INVALID_CQ;
1801 
1802 		cq_poll->cq_ids[CQ_QOS] = (qidx < pf->hw.tc_tx_queues) ?
1803 					  (qidx + pf->hw.rx_queues +
1804 					   pf->hw.non_qos_queues) :
1805 					  CINT_INVALID_CQ;
1806 
1807 		cq_poll->dev = (void *)pf;
1808 		cq_poll->dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_CQE;
1809 		INIT_WORK(&cq_poll->dim.work, otx2_dim_work);
1810 		netif_napi_add(netdev, &cq_poll->napi, otx2_napi_handler);
1811 		napi_enable(&cq_poll->napi);
1812 	}
1813 
1814 	/* Set maximum frame size allowed in HW */
1815 	err = otx2_hw_set_mtu(pf, netdev->mtu);
1816 	if (err)
1817 		goto err_disable_napi;
1818 
1819 	/* Setup segmentation algorithms, if failed, clear offload capability */
1820 	otx2_setup_segmentation(pf);
1821 
1822 	/* Initialize RSS */
1823 	err = otx2_rss_init(pf);
1824 	if (err)
1825 		goto err_disable_napi;
1826 
1827 	/* Register Queue IRQ handlers */
1828 	vec = pf->hw.nix_msixoff + NIX_LF_QINT_VEC_START;
1829 	irq_name = &pf->hw.irq_name[vec * NAME_SIZE];
1830 
1831 	snprintf(irq_name, NAME_SIZE, "%s-qerr", pf->netdev->name);
1832 
1833 	err = request_irq(pci_irq_vector(pf->pdev, vec),
1834 			  otx2_q_intr_handler, 0, irq_name, pf);
1835 	if (err) {
1836 		dev_err(pf->dev,
1837 			"RVUPF%d: IRQ registration failed for QERR\n",
1838 			rvu_get_pf(pf->pcifunc));
1839 		goto err_disable_napi;
1840 	}
1841 
1842 	/* Enable QINT IRQ */
1843 	otx2_write64(pf, NIX_LF_QINTX_ENA_W1S(0), BIT_ULL(0));
1844 
1845 	/* Register CQ IRQ handlers */
1846 	vec = pf->hw.nix_msixoff + NIX_LF_CINT_VEC_START;
1847 	for (qidx = 0; qidx < pf->hw.cint_cnt; qidx++) {
1848 		irq_name = &pf->hw.irq_name[vec * NAME_SIZE];
1849 
1850 		snprintf(irq_name, NAME_SIZE, "%s-rxtx-%d", pf->netdev->name,
1851 			 qidx);
1852 
1853 		err = request_irq(pci_irq_vector(pf->pdev, vec),
1854 				  otx2_cq_intr_handler, 0, irq_name,
1855 				  &qset->napi[qidx]);
1856 		if (err) {
1857 			dev_err(pf->dev,
1858 				"RVUPF%d: IRQ registration failed for CQ%d\n",
1859 				rvu_get_pf(pf->pcifunc), qidx);
1860 			goto err_free_cints;
1861 		}
1862 		vec++;
1863 
1864 		otx2_config_irq_coalescing(pf, qidx);
1865 
1866 		/* Enable CQ IRQ */
1867 		otx2_write64(pf, NIX_LF_CINTX_INT(qidx), BIT_ULL(0));
1868 		otx2_write64(pf, NIX_LF_CINTX_ENA_W1S(qidx), BIT_ULL(0));
1869 	}
1870 
1871 	otx2_set_cints_affinity(pf);
1872 
1873 	if (pf->flags & OTX2_FLAG_RX_VLAN_SUPPORT)
1874 		otx2_enable_rxvlan(pf, true);
1875 
1876 	/* When reinitializing enable time stamping if it is enabled before */
1877 	if (pf->flags & OTX2_FLAG_TX_TSTAMP_ENABLED) {
1878 		pf->flags &= ~OTX2_FLAG_TX_TSTAMP_ENABLED;
1879 		otx2_config_hw_tx_tstamp(pf, true);
1880 	}
1881 	if (pf->flags & OTX2_FLAG_RX_TSTAMP_ENABLED) {
1882 		pf->flags &= ~OTX2_FLAG_RX_TSTAMP_ENABLED;
1883 		otx2_config_hw_rx_tstamp(pf, true);
1884 	}
1885 
1886 	pf->flags &= ~OTX2_FLAG_INTF_DOWN;
1887 	/* 'intf_down' may be checked on any cpu */
1888 	smp_wmb();
1889 
1890 	/* Enable QoS configuration before starting tx queues */
1891 	otx2_qos_config_txschq(pf);
1892 
1893 	/* we have already received link status notification */
1894 	if (pf->linfo.link_up && !(pf->pcifunc & RVU_PFVF_FUNC_MASK))
1895 		otx2_handle_link_event(pf);
1896 
1897 	/* Install DMAC Filters */
1898 	if (pf->flags & OTX2_FLAG_DMACFLTR_SUPPORT)
1899 		otx2_dmacflt_reinstall_flows(pf);
1900 
1901 	otx2_tc_apply_ingress_police_rules(pf);
1902 
1903 	err = otx2_rxtx_enable(pf, true);
1904 	/* If a mbox communication error happens at this point then interface
1905 	 * will end up in a state such that it is in down state but hardware
1906 	 * mcam entries are enabled to receive the packets. Hence disable the
1907 	 * packet I/O.
1908 	 */
1909 	if (err == EIO)
1910 		goto err_disable_rxtx;
1911 	else if (err)
1912 		goto err_tx_stop_queues;
1913 
1914 	otx2_do_set_rx_mode(pf);
1915 
1916 	return 0;
1917 
1918 err_disable_rxtx:
1919 	otx2_rxtx_enable(pf, false);
1920 err_tx_stop_queues:
1921 	netif_tx_stop_all_queues(netdev);
1922 	netif_carrier_off(netdev);
1923 	pf->flags |= OTX2_FLAG_INTF_DOWN;
1924 err_free_cints:
1925 	otx2_free_cints(pf, qidx);
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 err_disable_napi:
1931 	otx2_disable_napi(pf);
1932 	otx2_free_hw_resources(pf);
1933 err_free_mem:
1934 	kfree(qset->sq);
1935 	kfree(qset->cq);
1936 	kfree(qset->rq);
1937 	kfree(qset->napi);
1938 	return err;
1939 }
1940 EXPORT_SYMBOL(otx2_open);
1941 
1942 int otx2_stop(struct net_device *netdev)
1943 {
1944 	struct otx2_nic *pf = netdev_priv(netdev);
1945 	struct otx2_cq_poll *cq_poll = NULL;
1946 	struct otx2_qset *qset = &pf->qset;
1947 	struct otx2_rss_info *rss;
1948 	int qidx, vec, wrk;
1949 
1950 	/* If the DOWN flag is set resources are already freed */
1951 	if (pf->flags & OTX2_FLAG_INTF_DOWN)
1952 		return 0;
1953 
1954 	netif_carrier_off(netdev);
1955 	netif_tx_stop_all_queues(netdev);
1956 
1957 	pf->flags |= OTX2_FLAG_INTF_DOWN;
1958 	/* 'intf_down' may be checked on any cpu */
1959 	smp_wmb();
1960 
1961 	/* First stop packet Rx/Tx */
1962 	otx2_rxtx_enable(pf, false);
1963 
1964 	/* Clear RSS enable flag */
1965 	rss = &pf->hw.rss_info;
1966 	rss->enable = false;
1967 	if (!netif_is_rxfh_configured(netdev))
1968 		kfree(rss->rss_ctx[DEFAULT_RSS_CONTEXT_GROUP]);
1969 
1970 	/* Cleanup Queue IRQ */
1971 	vec = pci_irq_vector(pf->pdev,
1972 			     pf->hw.nix_msixoff + NIX_LF_QINT_VEC_START);
1973 	otx2_write64(pf, NIX_LF_QINTX_ENA_W1C(0), BIT_ULL(0));
1974 	free_irq(vec, pf);
1975 
1976 	/* Cleanup CQ NAPI and IRQ */
1977 	vec = pf->hw.nix_msixoff + NIX_LF_CINT_VEC_START;
1978 	for (qidx = 0; qidx < pf->hw.cint_cnt; qidx++) {
1979 		/* Disable interrupt */
1980 		otx2_write64(pf, NIX_LF_CINTX_ENA_W1C(qidx), BIT_ULL(0));
1981 
1982 		synchronize_irq(pci_irq_vector(pf->pdev, vec));
1983 
1984 		cq_poll = &qset->napi[qidx];
1985 		napi_synchronize(&cq_poll->napi);
1986 		vec++;
1987 	}
1988 
1989 	netif_tx_disable(netdev);
1990 
1991 	for (wrk = 0; wrk < pf->qset.cq_cnt; wrk++)
1992 		cancel_delayed_work_sync(&pf->refill_wrk[wrk].pool_refill_work);
1993 	devm_kfree(pf->dev, pf->refill_wrk);
1994 
1995 	otx2_free_hw_resources(pf);
1996 	otx2_free_cints(pf, pf->hw.cint_cnt);
1997 	otx2_disable_napi(pf);
1998 
1999 	for (qidx = 0; qidx < netdev->num_tx_queues; qidx++)
2000 		netdev_tx_reset_queue(netdev_get_tx_queue(netdev, qidx));
2001 
2002 
2003 	kfree(qset->sq);
2004 	kfree(qset->cq);
2005 	kfree(qset->rq);
2006 	kfree(qset->napi);
2007 	/* Do not clear RQ/SQ ringsize settings */
2008 	memset_startat(qset, 0, sqe_cnt);
2009 	return 0;
2010 }
2011 EXPORT_SYMBOL(otx2_stop);
2012 
2013 static netdev_tx_t otx2_xmit(struct sk_buff *skb, struct net_device *netdev)
2014 {
2015 	struct otx2_nic *pf = netdev_priv(netdev);
2016 	int qidx = skb_get_queue_mapping(skb);
2017 	struct otx2_snd_queue *sq;
2018 	struct netdev_queue *txq;
2019 	int sq_idx;
2020 
2021 	/* XDP SQs are not mapped with TXQs
2022 	 * advance qid to derive correct sq mapped with QOS
2023 	 */
2024 	sq_idx = (qidx >= pf->hw.tx_queues) ? (qidx + pf->hw.xdp_queues) : qidx;
2025 
2026 	/* Check for minimum and maximum packet length */
2027 	if (skb->len <= ETH_HLEN ||
2028 	    (!skb_shinfo(skb)->gso_size && skb->len > pf->tx_max_pktlen)) {
2029 		dev_kfree_skb(skb);
2030 		return NETDEV_TX_OK;
2031 	}
2032 
2033 	sq = &pf->qset.sq[sq_idx];
2034 	txq = netdev_get_tx_queue(netdev, qidx);
2035 
2036 	if (!otx2_sq_append_skb(netdev, sq, skb, qidx)) {
2037 		netif_tx_stop_queue(txq);
2038 
2039 		/* Check again, incase SQBs got freed up */
2040 		smp_mb();
2041 		if (((sq->num_sqbs - *sq->aura_fc_addr) * sq->sqe_per_sqb)
2042 							> sq->sqe_thresh)
2043 			netif_tx_wake_queue(txq);
2044 
2045 		return NETDEV_TX_BUSY;
2046 	}
2047 
2048 	return NETDEV_TX_OK;
2049 }
2050 
2051 static int otx2_qos_select_htb_queue(struct otx2_nic *pf, struct sk_buff *skb,
2052 				     u16 htb_maj_id)
2053 {
2054 	u16 classid;
2055 
2056 	if ((TC_H_MAJ(skb->priority) >> 16) == htb_maj_id)
2057 		classid = TC_H_MIN(skb->priority);
2058 	else
2059 		classid = READ_ONCE(pf->qos.defcls);
2060 
2061 	if (!classid)
2062 		return 0;
2063 
2064 	return otx2_get_txq_by_classid(pf, classid);
2065 }
2066 
2067 u16 otx2_select_queue(struct net_device *netdev, struct sk_buff *skb,
2068 		      struct net_device *sb_dev)
2069 {
2070 	struct otx2_nic *pf = netdev_priv(netdev);
2071 	bool qos_enabled;
2072 #ifdef CONFIG_DCB
2073 	u8 vlan_prio;
2074 #endif
2075 	int txq;
2076 
2077 	qos_enabled = netdev->real_num_tx_queues > pf->hw.tx_queues;
2078 	if (unlikely(qos_enabled)) {
2079 		/* This smp_load_acquire() pairs with smp_store_release() in
2080 		 * otx2_qos_root_add() called from htb offload root creation
2081 		 */
2082 		u16 htb_maj_id = smp_load_acquire(&pf->qos.maj_id);
2083 
2084 		if (unlikely(htb_maj_id)) {
2085 			txq = otx2_qos_select_htb_queue(pf, skb, htb_maj_id);
2086 			if (txq > 0)
2087 				return txq;
2088 			goto process_pfc;
2089 		}
2090 	}
2091 
2092 process_pfc:
2093 #ifdef CONFIG_DCB
2094 	if (!skb_vlan_tag_present(skb))
2095 		goto pick_tx;
2096 
2097 	vlan_prio = skb->vlan_tci >> 13;
2098 	if ((vlan_prio > pf->hw.tx_queues - 1) ||
2099 	    !pf->pfc_alloc_status[vlan_prio])
2100 		goto pick_tx;
2101 
2102 	return vlan_prio;
2103 
2104 pick_tx:
2105 #endif
2106 	txq = netdev_pick_tx(netdev, skb, NULL);
2107 	if (unlikely(qos_enabled))
2108 		return txq % pf->hw.tx_queues;
2109 
2110 	return txq;
2111 }
2112 EXPORT_SYMBOL(otx2_select_queue);
2113 
2114 static netdev_features_t otx2_fix_features(struct net_device *dev,
2115 					   netdev_features_t features)
2116 {
2117 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
2118 		features |= NETIF_F_HW_VLAN_STAG_RX;
2119 	else
2120 		features &= ~NETIF_F_HW_VLAN_STAG_RX;
2121 
2122 	return features;
2123 }
2124 
2125 static void otx2_set_rx_mode(struct net_device *netdev)
2126 {
2127 	struct otx2_nic *pf = netdev_priv(netdev);
2128 
2129 	queue_work(pf->otx2_wq, &pf->rx_mode_work);
2130 }
2131 
2132 static void otx2_rx_mode_wrk_handler(struct work_struct *work)
2133 {
2134 	struct otx2_nic *pf = container_of(work, struct otx2_nic, rx_mode_work);
2135 
2136 	otx2_do_set_rx_mode(pf);
2137 }
2138 
2139 static int otx2_set_features(struct net_device *netdev,
2140 			     netdev_features_t features)
2141 {
2142 	netdev_features_t changed = features ^ netdev->features;
2143 	struct otx2_nic *pf = netdev_priv(netdev);
2144 
2145 	if ((changed & NETIF_F_LOOPBACK) && netif_running(netdev))
2146 		return otx2_cgx_config_loopback(pf,
2147 						features & NETIF_F_LOOPBACK);
2148 
2149 	if ((changed & NETIF_F_HW_VLAN_CTAG_RX) && netif_running(netdev))
2150 		return otx2_enable_rxvlan(pf,
2151 					  features & NETIF_F_HW_VLAN_CTAG_RX);
2152 
2153 	return otx2_handle_ntuple_tc_features(netdev, features);
2154 }
2155 
2156 static void otx2_reset_task(struct work_struct *work)
2157 {
2158 	struct otx2_nic *pf = container_of(work, struct otx2_nic, reset_task);
2159 
2160 	if (!netif_running(pf->netdev))
2161 		return;
2162 
2163 	rtnl_lock();
2164 	otx2_stop(pf->netdev);
2165 	pf->reset_count++;
2166 	otx2_open(pf->netdev);
2167 	netif_trans_update(pf->netdev);
2168 	rtnl_unlock();
2169 }
2170 
2171 static int otx2_config_hw_rx_tstamp(struct otx2_nic *pfvf, bool enable)
2172 {
2173 	struct msg_req *req;
2174 	int err;
2175 
2176 	if (pfvf->flags & OTX2_FLAG_RX_TSTAMP_ENABLED && enable)
2177 		return 0;
2178 
2179 	mutex_lock(&pfvf->mbox.lock);
2180 	if (enable)
2181 		req = otx2_mbox_alloc_msg_cgx_ptp_rx_enable(&pfvf->mbox);
2182 	else
2183 		req = otx2_mbox_alloc_msg_cgx_ptp_rx_disable(&pfvf->mbox);
2184 	if (!req) {
2185 		mutex_unlock(&pfvf->mbox.lock);
2186 		return -ENOMEM;
2187 	}
2188 
2189 	err = otx2_sync_mbox_msg(&pfvf->mbox);
2190 	if (err) {
2191 		mutex_unlock(&pfvf->mbox.lock);
2192 		return err;
2193 	}
2194 
2195 	mutex_unlock(&pfvf->mbox.lock);
2196 	if (enable)
2197 		pfvf->flags |= OTX2_FLAG_RX_TSTAMP_ENABLED;
2198 	else
2199 		pfvf->flags &= ~OTX2_FLAG_RX_TSTAMP_ENABLED;
2200 	return 0;
2201 }
2202 
2203 static int otx2_config_hw_tx_tstamp(struct otx2_nic *pfvf, bool enable)
2204 {
2205 	struct msg_req *req;
2206 	int err;
2207 
2208 	if (pfvf->flags & OTX2_FLAG_TX_TSTAMP_ENABLED && enable)
2209 		return 0;
2210 
2211 	mutex_lock(&pfvf->mbox.lock);
2212 	if (enable)
2213 		req = otx2_mbox_alloc_msg_nix_lf_ptp_tx_enable(&pfvf->mbox);
2214 	else
2215 		req = otx2_mbox_alloc_msg_nix_lf_ptp_tx_disable(&pfvf->mbox);
2216 	if (!req) {
2217 		mutex_unlock(&pfvf->mbox.lock);
2218 		return -ENOMEM;
2219 	}
2220 
2221 	err = otx2_sync_mbox_msg(&pfvf->mbox);
2222 	if (err) {
2223 		mutex_unlock(&pfvf->mbox.lock);
2224 		return err;
2225 	}
2226 
2227 	mutex_unlock(&pfvf->mbox.lock);
2228 	if (enable)
2229 		pfvf->flags |= OTX2_FLAG_TX_TSTAMP_ENABLED;
2230 	else
2231 		pfvf->flags &= ~OTX2_FLAG_TX_TSTAMP_ENABLED;
2232 	return 0;
2233 }
2234 
2235 int otx2_config_hwtstamp(struct net_device *netdev, struct ifreq *ifr)
2236 {
2237 	struct otx2_nic *pfvf = netdev_priv(netdev);
2238 	struct hwtstamp_config config;
2239 
2240 	if (!pfvf->ptp)
2241 		return -ENODEV;
2242 
2243 	if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
2244 		return -EFAULT;
2245 
2246 	switch (config.tx_type) {
2247 	case HWTSTAMP_TX_OFF:
2248 		if (pfvf->flags & OTX2_FLAG_PTP_ONESTEP_SYNC)
2249 			pfvf->flags &= ~OTX2_FLAG_PTP_ONESTEP_SYNC;
2250 
2251 		cancel_delayed_work(&pfvf->ptp->synctstamp_work);
2252 		otx2_config_hw_tx_tstamp(pfvf, false);
2253 		break;
2254 	case HWTSTAMP_TX_ONESTEP_SYNC:
2255 		if (!test_bit(CN10K_PTP_ONESTEP, &pfvf->hw.cap_flag))
2256 			return -ERANGE;
2257 		pfvf->flags |= OTX2_FLAG_PTP_ONESTEP_SYNC;
2258 		schedule_delayed_work(&pfvf->ptp->synctstamp_work,
2259 				      msecs_to_jiffies(500));
2260 		fallthrough;
2261 	case HWTSTAMP_TX_ON:
2262 		otx2_config_hw_tx_tstamp(pfvf, true);
2263 		break;
2264 	default:
2265 		return -ERANGE;
2266 	}
2267 
2268 	switch (config.rx_filter) {
2269 	case HWTSTAMP_FILTER_NONE:
2270 		otx2_config_hw_rx_tstamp(pfvf, false);
2271 		break;
2272 	case HWTSTAMP_FILTER_ALL:
2273 	case HWTSTAMP_FILTER_SOME:
2274 	case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
2275 	case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
2276 	case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
2277 	case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
2278 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
2279 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
2280 	case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
2281 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
2282 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
2283 	case HWTSTAMP_FILTER_PTP_V2_EVENT:
2284 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
2285 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
2286 		otx2_config_hw_rx_tstamp(pfvf, true);
2287 		config.rx_filter = HWTSTAMP_FILTER_ALL;
2288 		break;
2289 	default:
2290 		return -ERANGE;
2291 	}
2292 
2293 	memcpy(&pfvf->tstamp, &config, sizeof(config));
2294 
2295 	return copy_to_user(ifr->ifr_data, &config,
2296 			    sizeof(config)) ? -EFAULT : 0;
2297 }
2298 EXPORT_SYMBOL(otx2_config_hwtstamp);
2299 
2300 int otx2_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
2301 {
2302 	struct otx2_nic *pfvf = netdev_priv(netdev);
2303 	struct hwtstamp_config *cfg = &pfvf->tstamp;
2304 
2305 	switch (cmd) {
2306 	case SIOCSHWTSTAMP:
2307 		return otx2_config_hwtstamp(netdev, req);
2308 	case SIOCGHWTSTAMP:
2309 		return copy_to_user(req->ifr_data, cfg,
2310 				    sizeof(*cfg)) ? -EFAULT : 0;
2311 	default:
2312 		return -EOPNOTSUPP;
2313 	}
2314 }
2315 EXPORT_SYMBOL(otx2_ioctl);
2316 
2317 static int otx2_do_set_vf_mac(struct otx2_nic *pf, int vf, const u8 *mac)
2318 {
2319 	struct npc_install_flow_req *req;
2320 	int err;
2321 
2322 	mutex_lock(&pf->mbox.lock);
2323 	req = otx2_mbox_alloc_msg_npc_install_flow(&pf->mbox);
2324 	if (!req) {
2325 		err = -ENOMEM;
2326 		goto out;
2327 	}
2328 
2329 	ether_addr_copy(req->packet.dmac, mac);
2330 	eth_broadcast_addr((u8 *)&req->mask.dmac);
2331 	req->features = BIT_ULL(NPC_DMAC);
2332 	req->channel = pf->hw.rx_chan_base;
2333 	req->intf = NIX_INTF_RX;
2334 	req->default_rule = 1;
2335 	req->append = 1;
2336 	req->vf = vf + 1;
2337 	req->op = NIX_RX_ACTION_DEFAULT;
2338 
2339 	err = otx2_sync_mbox_msg(&pf->mbox);
2340 out:
2341 	mutex_unlock(&pf->mbox.lock);
2342 	return err;
2343 }
2344 
2345 static int otx2_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
2346 {
2347 	struct otx2_nic *pf = netdev_priv(netdev);
2348 	struct pci_dev *pdev = pf->pdev;
2349 	struct otx2_vf_config *config;
2350 	int ret;
2351 
2352 	if (!netif_running(netdev))
2353 		return -EAGAIN;
2354 
2355 	if (vf >= pf->total_vfs)
2356 		return -EINVAL;
2357 
2358 	if (!is_valid_ether_addr(mac))
2359 		return -EINVAL;
2360 
2361 	config = &pf->vf_configs[vf];
2362 	ether_addr_copy(config->mac, mac);
2363 
2364 	ret = otx2_do_set_vf_mac(pf, vf, mac);
2365 	if (ret == 0)
2366 		dev_info(&pdev->dev,
2367 			 "Load/Reload VF driver\n");
2368 
2369 	return ret;
2370 }
2371 
2372 static int otx2_do_set_vf_vlan(struct otx2_nic *pf, int vf, u16 vlan, u8 qos,
2373 			       __be16 proto)
2374 {
2375 	struct otx2_flow_config *flow_cfg = pf->flow_cfg;
2376 	struct nix_vtag_config_rsp *vtag_rsp;
2377 	struct npc_delete_flow_req *del_req;
2378 	struct nix_vtag_config *vtag_req;
2379 	struct npc_install_flow_req *req;
2380 	struct otx2_vf_config *config;
2381 	int err = 0;
2382 	u32 idx;
2383 
2384 	config = &pf->vf_configs[vf];
2385 
2386 	if (!vlan && !config->vlan)
2387 		goto out;
2388 
2389 	mutex_lock(&pf->mbox.lock);
2390 
2391 	/* free old tx vtag entry */
2392 	if (config->vlan) {
2393 		vtag_req = otx2_mbox_alloc_msg_nix_vtag_cfg(&pf->mbox);
2394 		if (!vtag_req) {
2395 			err = -ENOMEM;
2396 			goto out;
2397 		}
2398 		vtag_req->cfg_type = 0;
2399 		vtag_req->tx.free_vtag0 = 1;
2400 		vtag_req->tx.vtag0_idx = config->tx_vtag_idx;
2401 
2402 		err = otx2_sync_mbox_msg(&pf->mbox);
2403 		if (err)
2404 			goto out;
2405 	}
2406 
2407 	if (!vlan && config->vlan) {
2408 		/* rx */
2409 		del_req = otx2_mbox_alloc_msg_npc_delete_flow(&pf->mbox);
2410 		if (!del_req) {
2411 			err = -ENOMEM;
2412 			goto out;
2413 		}
2414 		idx = ((vf * OTX2_PER_VF_VLAN_FLOWS) + OTX2_VF_VLAN_RX_INDEX);
2415 		del_req->entry =
2416 			flow_cfg->def_ent[flow_cfg->vf_vlan_offset + idx];
2417 		err = otx2_sync_mbox_msg(&pf->mbox);
2418 		if (err)
2419 			goto out;
2420 
2421 		/* tx */
2422 		del_req = otx2_mbox_alloc_msg_npc_delete_flow(&pf->mbox);
2423 		if (!del_req) {
2424 			err = -ENOMEM;
2425 			goto out;
2426 		}
2427 		idx = ((vf * OTX2_PER_VF_VLAN_FLOWS) + OTX2_VF_VLAN_TX_INDEX);
2428 		del_req->entry =
2429 			flow_cfg->def_ent[flow_cfg->vf_vlan_offset + idx];
2430 		err = otx2_sync_mbox_msg(&pf->mbox);
2431 
2432 		goto out;
2433 	}
2434 
2435 	/* rx */
2436 	req = otx2_mbox_alloc_msg_npc_install_flow(&pf->mbox);
2437 	if (!req) {
2438 		err = -ENOMEM;
2439 		goto out;
2440 	}
2441 
2442 	idx = ((vf * OTX2_PER_VF_VLAN_FLOWS) + OTX2_VF_VLAN_RX_INDEX);
2443 	req->entry = flow_cfg->def_ent[flow_cfg->vf_vlan_offset + idx];
2444 	req->packet.vlan_tci = htons(vlan);
2445 	req->mask.vlan_tci = htons(VLAN_VID_MASK);
2446 	/* af fills the destination mac addr */
2447 	eth_broadcast_addr((u8 *)&req->mask.dmac);
2448 	req->features = BIT_ULL(NPC_OUTER_VID) | BIT_ULL(NPC_DMAC);
2449 	req->channel = pf->hw.rx_chan_base;
2450 	req->intf = NIX_INTF_RX;
2451 	req->vf = vf + 1;
2452 	req->op = NIX_RX_ACTION_DEFAULT;
2453 	req->vtag0_valid = true;
2454 	req->vtag0_type = NIX_AF_LFX_RX_VTAG_TYPE7;
2455 	req->set_cntr = 1;
2456 
2457 	err = otx2_sync_mbox_msg(&pf->mbox);
2458 	if (err)
2459 		goto out;
2460 
2461 	/* tx */
2462 	vtag_req = otx2_mbox_alloc_msg_nix_vtag_cfg(&pf->mbox);
2463 	if (!vtag_req) {
2464 		err = -ENOMEM;
2465 		goto out;
2466 	}
2467 
2468 	/* configure tx vtag params */
2469 	vtag_req->vtag_size = VTAGSIZE_T4;
2470 	vtag_req->cfg_type = 0; /* tx vlan cfg */
2471 	vtag_req->tx.cfg_vtag0 = 1;
2472 	vtag_req->tx.vtag0 = ((u64)ntohs(proto) << 16) | vlan;
2473 
2474 	err = otx2_sync_mbox_msg(&pf->mbox);
2475 	if (err)
2476 		goto out;
2477 
2478 	vtag_rsp = (struct nix_vtag_config_rsp *)otx2_mbox_get_rsp
2479 			(&pf->mbox.mbox, 0, &vtag_req->hdr);
2480 	if (IS_ERR(vtag_rsp)) {
2481 		err = PTR_ERR(vtag_rsp);
2482 		goto out;
2483 	}
2484 	config->tx_vtag_idx = vtag_rsp->vtag0_idx;
2485 
2486 	req = otx2_mbox_alloc_msg_npc_install_flow(&pf->mbox);
2487 	if (!req) {
2488 		err = -ENOMEM;
2489 		goto out;
2490 	}
2491 
2492 	eth_zero_addr((u8 *)&req->mask.dmac);
2493 	idx = ((vf * OTX2_PER_VF_VLAN_FLOWS) + OTX2_VF_VLAN_TX_INDEX);
2494 	req->entry = flow_cfg->def_ent[flow_cfg->vf_vlan_offset + idx];
2495 	req->features = BIT_ULL(NPC_DMAC);
2496 	req->channel = pf->hw.tx_chan_base;
2497 	req->intf = NIX_INTF_TX;
2498 	req->vf = vf + 1;
2499 	req->op = NIX_TX_ACTIONOP_UCAST_DEFAULT;
2500 	req->vtag0_def = vtag_rsp->vtag0_idx;
2501 	req->vtag0_op = VTAG_INSERT;
2502 	req->set_cntr = 1;
2503 
2504 	err = otx2_sync_mbox_msg(&pf->mbox);
2505 out:
2506 	config->vlan = vlan;
2507 	mutex_unlock(&pf->mbox.lock);
2508 	return err;
2509 }
2510 
2511 static int otx2_set_vf_vlan(struct net_device *netdev, int vf, u16 vlan, u8 qos,
2512 			    __be16 proto)
2513 {
2514 	struct otx2_nic *pf = netdev_priv(netdev);
2515 	struct pci_dev *pdev = pf->pdev;
2516 
2517 	if (!netif_running(netdev))
2518 		return -EAGAIN;
2519 
2520 	if (vf >= pci_num_vf(pdev))
2521 		return -EINVAL;
2522 
2523 	/* qos is currently unsupported */
2524 	if (vlan >= VLAN_N_VID || qos)
2525 		return -EINVAL;
2526 
2527 	if (proto != htons(ETH_P_8021Q))
2528 		return -EPROTONOSUPPORT;
2529 
2530 	if (!(pf->flags & OTX2_FLAG_VF_VLAN_SUPPORT))
2531 		return -EOPNOTSUPP;
2532 
2533 	return otx2_do_set_vf_vlan(pf, vf, vlan, qos, proto);
2534 }
2535 
2536 static int otx2_get_vf_config(struct net_device *netdev, int vf,
2537 			      struct ifla_vf_info *ivi)
2538 {
2539 	struct otx2_nic *pf = netdev_priv(netdev);
2540 	struct pci_dev *pdev = pf->pdev;
2541 	struct otx2_vf_config *config;
2542 
2543 	if (!netif_running(netdev))
2544 		return -EAGAIN;
2545 
2546 	if (vf >= pci_num_vf(pdev))
2547 		return -EINVAL;
2548 
2549 	config = &pf->vf_configs[vf];
2550 	ivi->vf = vf;
2551 	ether_addr_copy(ivi->mac, config->mac);
2552 	ivi->vlan = config->vlan;
2553 	ivi->trusted = config->trusted;
2554 
2555 	return 0;
2556 }
2557 
2558 static int otx2_xdp_xmit_tx(struct otx2_nic *pf, struct xdp_frame *xdpf,
2559 			    int qidx)
2560 {
2561 	struct page *page;
2562 	u64 dma_addr;
2563 	int err = 0;
2564 
2565 	dma_addr = otx2_dma_map_page(pf, virt_to_page(xdpf->data),
2566 				     offset_in_page(xdpf->data), xdpf->len,
2567 				     DMA_TO_DEVICE);
2568 	if (dma_mapping_error(pf->dev, dma_addr))
2569 		return -ENOMEM;
2570 
2571 	err = otx2_xdp_sq_append_pkt(pf, dma_addr, xdpf->len, qidx);
2572 	if (!err) {
2573 		otx2_dma_unmap_page(pf, dma_addr, xdpf->len, DMA_TO_DEVICE);
2574 		page = virt_to_page(xdpf->data);
2575 		put_page(page);
2576 		return -ENOMEM;
2577 	}
2578 	return 0;
2579 }
2580 
2581 static int otx2_xdp_xmit(struct net_device *netdev, int n,
2582 			 struct xdp_frame **frames, u32 flags)
2583 {
2584 	struct otx2_nic *pf = netdev_priv(netdev);
2585 	int qidx = smp_processor_id();
2586 	struct otx2_snd_queue *sq;
2587 	int drops = 0, i;
2588 
2589 	if (!netif_running(netdev))
2590 		return -ENETDOWN;
2591 
2592 	qidx += pf->hw.tx_queues;
2593 	sq = pf->xdp_prog ? &pf->qset.sq[qidx] : NULL;
2594 
2595 	/* Abort xmit if xdp queue is not */
2596 	if (unlikely(!sq))
2597 		return -ENXIO;
2598 
2599 	if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK))
2600 		return -EINVAL;
2601 
2602 	for (i = 0; i < n; i++) {
2603 		struct xdp_frame *xdpf = frames[i];
2604 		int err;
2605 
2606 		err = otx2_xdp_xmit_tx(pf, xdpf, qidx);
2607 		if (err)
2608 			drops++;
2609 	}
2610 	return n - drops;
2611 }
2612 
2613 static int otx2_xdp_setup(struct otx2_nic *pf, struct bpf_prog *prog)
2614 {
2615 	struct net_device *dev = pf->netdev;
2616 	bool if_up = netif_running(pf->netdev);
2617 	struct bpf_prog *old_prog;
2618 
2619 	if (prog && dev->mtu > MAX_XDP_MTU) {
2620 		netdev_warn(dev, "Jumbo frames not yet supported with XDP\n");
2621 		return -EOPNOTSUPP;
2622 	}
2623 
2624 	if (if_up)
2625 		otx2_stop(pf->netdev);
2626 
2627 	old_prog = xchg(&pf->xdp_prog, prog);
2628 
2629 	if (old_prog)
2630 		bpf_prog_put(old_prog);
2631 
2632 	if (pf->xdp_prog)
2633 		bpf_prog_add(pf->xdp_prog, pf->hw.rx_queues - 1);
2634 
2635 	/* Network stack and XDP shared same rx queues.
2636 	 * Use separate tx queues for XDP and network stack.
2637 	 */
2638 	if (pf->xdp_prog) {
2639 		pf->hw.xdp_queues = pf->hw.rx_queues;
2640 		xdp_features_set_redirect_target(dev, false);
2641 	} else {
2642 		pf->hw.xdp_queues = 0;
2643 		xdp_features_clear_redirect_target(dev);
2644 	}
2645 
2646 	pf->hw.non_qos_queues += pf->hw.xdp_queues;
2647 
2648 	if (if_up)
2649 		otx2_open(pf->netdev);
2650 
2651 	return 0;
2652 }
2653 
2654 static int otx2_xdp(struct net_device *netdev, struct netdev_bpf *xdp)
2655 {
2656 	struct otx2_nic *pf = netdev_priv(netdev);
2657 
2658 	switch (xdp->command) {
2659 	case XDP_SETUP_PROG:
2660 		return otx2_xdp_setup(pf, xdp->prog);
2661 	default:
2662 		return -EINVAL;
2663 	}
2664 }
2665 
2666 static int otx2_set_vf_permissions(struct otx2_nic *pf, int vf,
2667 				   int req_perm)
2668 {
2669 	struct set_vf_perm *req;
2670 	int rc;
2671 
2672 	mutex_lock(&pf->mbox.lock);
2673 	req = otx2_mbox_alloc_msg_set_vf_perm(&pf->mbox);
2674 	if (!req) {
2675 		rc = -ENOMEM;
2676 		goto out;
2677 	}
2678 
2679 	/* Let AF reset VF permissions as sriov is disabled */
2680 	if (req_perm == OTX2_RESET_VF_PERM) {
2681 		req->flags |= RESET_VF_PERM;
2682 	} else if (req_perm == OTX2_TRUSTED_VF) {
2683 		if (pf->vf_configs[vf].trusted)
2684 			req->flags |= VF_TRUSTED;
2685 	}
2686 
2687 	req->vf = vf;
2688 	rc = otx2_sync_mbox_msg(&pf->mbox);
2689 out:
2690 	mutex_unlock(&pf->mbox.lock);
2691 	return rc;
2692 }
2693 
2694 static int otx2_ndo_set_vf_trust(struct net_device *netdev, int vf,
2695 				 bool enable)
2696 {
2697 	struct otx2_nic *pf = netdev_priv(netdev);
2698 	struct pci_dev *pdev = pf->pdev;
2699 	int rc;
2700 
2701 	if (vf >= pci_num_vf(pdev))
2702 		return -EINVAL;
2703 
2704 	if (pf->vf_configs[vf].trusted == enable)
2705 		return 0;
2706 
2707 	pf->vf_configs[vf].trusted = enable;
2708 	rc = otx2_set_vf_permissions(pf, vf, OTX2_TRUSTED_VF);
2709 
2710 	if (rc) {
2711 		pf->vf_configs[vf].trusted = !enable;
2712 	} else {
2713 		netdev_info(pf->netdev, "VF %d is %strusted\n",
2714 			    vf, enable ? "" : "not ");
2715 		otx2_set_rx_mode(netdev);
2716 	}
2717 
2718 	return rc;
2719 }
2720 
2721 static const struct net_device_ops otx2_netdev_ops = {
2722 	.ndo_open		= otx2_open,
2723 	.ndo_stop		= otx2_stop,
2724 	.ndo_start_xmit		= otx2_xmit,
2725 	.ndo_select_queue	= otx2_select_queue,
2726 	.ndo_fix_features	= otx2_fix_features,
2727 	.ndo_set_mac_address    = otx2_set_mac_address,
2728 	.ndo_change_mtu		= otx2_change_mtu,
2729 	.ndo_set_rx_mode	= otx2_set_rx_mode,
2730 	.ndo_set_features	= otx2_set_features,
2731 	.ndo_tx_timeout		= otx2_tx_timeout,
2732 	.ndo_get_stats64	= otx2_get_stats64,
2733 	.ndo_eth_ioctl		= otx2_ioctl,
2734 	.ndo_set_vf_mac		= otx2_set_vf_mac,
2735 	.ndo_set_vf_vlan	= otx2_set_vf_vlan,
2736 	.ndo_get_vf_config	= otx2_get_vf_config,
2737 	.ndo_bpf		= otx2_xdp,
2738 	.ndo_xdp_xmit           = otx2_xdp_xmit,
2739 	.ndo_setup_tc		= otx2_setup_tc,
2740 	.ndo_set_vf_trust	= otx2_ndo_set_vf_trust,
2741 };
2742 
2743 static int otx2_wq_init(struct otx2_nic *pf)
2744 {
2745 	pf->otx2_wq = create_singlethread_workqueue("otx2_wq");
2746 	if (!pf->otx2_wq)
2747 		return -ENOMEM;
2748 
2749 	INIT_WORK(&pf->rx_mode_work, otx2_rx_mode_wrk_handler);
2750 	INIT_WORK(&pf->reset_task, otx2_reset_task);
2751 	return 0;
2752 }
2753 
2754 static int otx2_check_pf_usable(struct otx2_nic *nic)
2755 {
2756 	u64 rev;
2757 
2758 	rev = otx2_read64(nic, RVU_PF_BLOCK_ADDRX_DISC(BLKADDR_RVUM));
2759 	rev = (rev >> 12) & 0xFF;
2760 	/* Check if AF has setup revision for RVUM block,
2761 	 * otherwise this driver probe should be deferred
2762 	 * until AF driver comes up.
2763 	 */
2764 	if (!rev) {
2765 		dev_warn(nic->dev,
2766 			 "AF is not initialized, deferring probe\n");
2767 		return -EPROBE_DEFER;
2768 	}
2769 	return 0;
2770 }
2771 
2772 static int otx2_realloc_msix_vectors(struct otx2_nic *pf)
2773 {
2774 	struct otx2_hw *hw = &pf->hw;
2775 	int num_vec, err;
2776 
2777 	/* NPA interrupts are inot registered, so alloc only
2778 	 * upto NIX vector offset.
2779 	 */
2780 	num_vec = hw->nix_msixoff;
2781 	num_vec += NIX_LF_CINT_VEC_START + hw->max_queues;
2782 
2783 	otx2_disable_mbox_intr(pf);
2784 	pci_free_irq_vectors(hw->pdev);
2785 	err = pci_alloc_irq_vectors(hw->pdev, num_vec, num_vec, PCI_IRQ_MSIX);
2786 	if (err < 0) {
2787 		dev_err(pf->dev, "%s: Failed to realloc %d IRQ vectors\n",
2788 			__func__, num_vec);
2789 		return err;
2790 	}
2791 
2792 	return otx2_register_mbox_intr(pf, false);
2793 }
2794 
2795 static int otx2_sriov_vfcfg_init(struct otx2_nic *pf)
2796 {
2797 	int i;
2798 
2799 	pf->vf_configs = devm_kcalloc(pf->dev, pf->total_vfs,
2800 				      sizeof(struct otx2_vf_config),
2801 				      GFP_KERNEL);
2802 	if (!pf->vf_configs)
2803 		return -ENOMEM;
2804 
2805 	for (i = 0; i < pf->total_vfs; i++) {
2806 		pf->vf_configs[i].pf = pf;
2807 		pf->vf_configs[i].intf_down = true;
2808 		pf->vf_configs[i].trusted = false;
2809 		INIT_DELAYED_WORK(&pf->vf_configs[i].link_event_work,
2810 				  otx2_vf_link_event_task);
2811 	}
2812 
2813 	return 0;
2814 }
2815 
2816 static void otx2_sriov_vfcfg_cleanup(struct otx2_nic *pf)
2817 {
2818 	int i;
2819 
2820 	if (!pf->vf_configs)
2821 		return;
2822 
2823 	for (i = 0; i < pf->total_vfs; i++) {
2824 		cancel_delayed_work_sync(&pf->vf_configs[i].link_event_work);
2825 		otx2_set_vf_permissions(pf, i, OTX2_RESET_VF_PERM);
2826 	}
2827 }
2828 
2829 static int otx2_probe(struct pci_dev *pdev, const struct pci_device_id *id)
2830 {
2831 	struct device *dev = &pdev->dev;
2832 	int err, qcount, qos_txqs;
2833 	struct net_device *netdev;
2834 	struct otx2_nic *pf;
2835 	struct otx2_hw *hw;
2836 	int num_vec;
2837 
2838 	err = pcim_enable_device(pdev);
2839 	if (err) {
2840 		dev_err(dev, "Failed to enable PCI device\n");
2841 		return err;
2842 	}
2843 
2844 	err = pci_request_regions(pdev, DRV_NAME);
2845 	if (err) {
2846 		dev_err(dev, "PCI request regions failed 0x%x\n", err);
2847 		return err;
2848 	}
2849 
2850 	err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(48));
2851 	if (err) {
2852 		dev_err(dev, "DMA mask config failed, abort\n");
2853 		goto err_release_regions;
2854 	}
2855 
2856 	pci_set_master(pdev);
2857 
2858 	/* Set number of queues */
2859 	qcount = min_t(int, num_online_cpus(), OTX2_MAX_CQ_CNT);
2860 	qos_txqs = min_t(int, qcount, OTX2_QOS_MAX_LEAF_NODES);
2861 
2862 	netdev = alloc_etherdev_mqs(sizeof(*pf), qcount + qos_txqs, qcount);
2863 	if (!netdev) {
2864 		err = -ENOMEM;
2865 		goto err_release_regions;
2866 	}
2867 
2868 	pci_set_drvdata(pdev, netdev);
2869 	SET_NETDEV_DEV(netdev, &pdev->dev);
2870 	pf = netdev_priv(netdev);
2871 	pf->netdev = netdev;
2872 	pf->pdev = pdev;
2873 	pf->dev = dev;
2874 	pf->total_vfs = pci_sriov_get_totalvfs(pdev);
2875 	pf->flags |= OTX2_FLAG_INTF_DOWN;
2876 
2877 	hw = &pf->hw;
2878 	hw->pdev = pdev;
2879 	hw->rx_queues = qcount;
2880 	hw->tx_queues = qcount;
2881 	hw->non_qos_queues = qcount;
2882 	hw->max_queues = qcount;
2883 	hw->rbuf_len = OTX2_DEFAULT_RBUF_LEN;
2884 	/* Use CQE of 128 byte descriptor size by default */
2885 	hw->xqe_size = 128;
2886 
2887 	num_vec = pci_msix_vec_count(pdev);
2888 	hw->irq_name = devm_kmalloc_array(&hw->pdev->dev, num_vec, NAME_SIZE,
2889 					  GFP_KERNEL);
2890 	if (!hw->irq_name) {
2891 		err = -ENOMEM;
2892 		goto err_free_netdev;
2893 	}
2894 
2895 	hw->affinity_mask = devm_kcalloc(&hw->pdev->dev, num_vec,
2896 					 sizeof(cpumask_var_t), GFP_KERNEL);
2897 	if (!hw->affinity_mask) {
2898 		err = -ENOMEM;
2899 		goto err_free_netdev;
2900 	}
2901 
2902 	/* Map CSRs */
2903 	pf->reg_base = pcim_iomap(pdev, PCI_CFG_REG_BAR_NUM, 0);
2904 	if (!pf->reg_base) {
2905 		dev_err(dev, "Unable to map physical function CSRs, aborting\n");
2906 		err = -ENOMEM;
2907 		goto err_free_netdev;
2908 	}
2909 
2910 	err = otx2_check_pf_usable(pf);
2911 	if (err)
2912 		goto err_free_netdev;
2913 
2914 	err = pci_alloc_irq_vectors(hw->pdev, RVU_PF_INT_VEC_CNT,
2915 				    RVU_PF_INT_VEC_CNT, PCI_IRQ_MSIX);
2916 	if (err < 0) {
2917 		dev_err(dev, "%s: Failed to alloc %d IRQ vectors\n",
2918 			__func__, num_vec);
2919 		goto err_free_netdev;
2920 	}
2921 
2922 	otx2_setup_dev_hw_settings(pf);
2923 
2924 	/* Init PF <=> AF mailbox stuff */
2925 	err = otx2_pfaf_mbox_init(pf);
2926 	if (err)
2927 		goto err_free_irq_vectors;
2928 
2929 	/* Register mailbox interrupt */
2930 	err = otx2_register_mbox_intr(pf, true);
2931 	if (err)
2932 		goto err_mbox_destroy;
2933 
2934 	/* Request AF to attach NPA and NIX LFs to this PF.
2935 	 * NIX and NPA LFs are needed for this PF to function as a NIC.
2936 	 */
2937 	err = otx2_attach_npa_nix(pf);
2938 	if (err)
2939 		goto err_disable_mbox_intr;
2940 
2941 	err = otx2_realloc_msix_vectors(pf);
2942 	if (err)
2943 		goto err_detach_rsrc;
2944 
2945 	err = otx2_set_real_num_queues(netdev, hw->tx_queues, hw->rx_queues);
2946 	if (err)
2947 		goto err_detach_rsrc;
2948 
2949 	err = cn10k_lmtst_init(pf);
2950 	if (err)
2951 		goto err_detach_rsrc;
2952 
2953 	/* Assign default mac address */
2954 	otx2_get_mac_from_af(netdev);
2955 
2956 	/* Don't check for error.  Proceed without ptp */
2957 	otx2_ptp_init(pf);
2958 
2959 	/* NPA's pool is a stack to which SW frees buffer pointers via Aura.
2960 	 * HW allocates buffer pointer from stack and uses it for DMA'ing
2961 	 * ingress packet. In some scenarios HW can free back allocated buffer
2962 	 * pointers to pool. This makes it impossible for SW to maintain a
2963 	 * parallel list where physical addresses of buffer pointers (IOVAs)
2964 	 * given to HW can be saved for later reference.
2965 	 *
2966 	 * So the only way to convert Rx packet's buffer address is to use
2967 	 * IOMMU's iova_to_phys() handler which translates the address by
2968 	 * walking through the translation tables.
2969 	 */
2970 	pf->iommu_domain = iommu_get_domain_for_dev(dev);
2971 
2972 	netdev->hw_features = (NETIF_F_RXCSUM | NETIF_F_IP_CSUM |
2973 			       NETIF_F_IPV6_CSUM | NETIF_F_RXHASH |
2974 			       NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
2975 			       NETIF_F_GSO_UDP_L4);
2976 	netdev->features |= netdev->hw_features;
2977 
2978 	err = otx2_mcam_flow_init(pf);
2979 	if (err)
2980 		goto err_ptp_destroy;
2981 
2982 	err = cn10k_mcs_init(pf);
2983 	if (err)
2984 		goto err_del_mcam_entries;
2985 
2986 	if (pf->flags & OTX2_FLAG_NTUPLE_SUPPORT)
2987 		netdev->hw_features |= NETIF_F_NTUPLE;
2988 
2989 	if (pf->flags & OTX2_FLAG_UCAST_FLTR_SUPPORT)
2990 		netdev->priv_flags |= IFF_UNICAST_FLT;
2991 
2992 	/* Support TSO on tag interface */
2993 	netdev->vlan_features |= netdev->features;
2994 	netdev->hw_features  |= NETIF_F_HW_VLAN_CTAG_TX |
2995 				NETIF_F_HW_VLAN_STAG_TX;
2996 	if (pf->flags & OTX2_FLAG_RX_VLAN_SUPPORT)
2997 		netdev->hw_features |= NETIF_F_HW_VLAN_CTAG_RX |
2998 				       NETIF_F_HW_VLAN_STAG_RX;
2999 	netdev->features |= netdev->hw_features;
3000 
3001 	/* HW supports tc offload but mutually exclusive with n-tuple filters */
3002 	if (pf->flags & OTX2_FLAG_TC_FLOWER_SUPPORT)
3003 		netdev->hw_features |= NETIF_F_HW_TC;
3004 
3005 	netdev->hw_features |= NETIF_F_LOOPBACK | NETIF_F_RXALL;
3006 
3007 	netif_set_tso_max_segs(netdev, OTX2_MAX_GSO_SEGS);
3008 	netdev->watchdog_timeo = OTX2_TX_TIMEOUT;
3009 
3010 	netdev->netdev_ops = &otx2_netdev_ops;
3011 	netdev->xdp_features = NETDEV_XDP_ACT_BASIC | NETDEV_XDP_ACT_REDIRECT;
3012 
3013 	netdev->min_mtu = OTX2_MIN_MTU;
3014 	netdev->max_mtu = otx2_get_max_mtu(pf);
3015 
3016 	err = register_netdev(netdev);
3017 	if (err) {
3018 		dev_err(dev, "Failed to register netdevice\n");
3019 		goto err_mcs_free;
3020 	}
3021 
3022 	err = otx2_wq_init(pf);
3023 	if (err)
3024 		goto err_unreg_netdev;
3025 
3026 	otx2_set_ethtool_ops(netdev);
3027 
3028 	err = otx2_init_tc(pf);
3029 	if (err)
3030 		goto err_mcam_flow_del;
3031 
3032 	err = otx2_register_dl(pf);
3033 	if (err)
3034 		goto err_mcam_flow_del;
3035 
3036 	/* Initialize SR-IOV resources */
3037 	err = otx2_sriov_vfcfg_init(pf);
3038 	if (err)
3039 		goto err_pf_sriov_init;
3040 
3041 	/* Enable link notifications */
3042 	otx2_cgx_config_linkevents(pf, true);
3043 
3044 #ifdef CONFIG_DCB
3045 	err = otx2_dcbnl_set_ops(netdev);
3046 	if (err)
3047 		goto err_pf_sriov_init;
3048 #endif
3049 
3050 	otx2_qos_init(pf, qos_txqs);
3051 
3052 	return 0;
3053 
3054 err_pf_sriov_init:
3055 	otx2_shutdown_tc(pf);
3056 err_mcam_flow_del:
3057 	otx2_mcam_flow_del(pf);
3058 err_unreg_netdev:
3059 	unregister_netdev(netdev);
3060 err_mcs_free:
3061 	cn10k_mcs_free(pf);
3062 err_del_mcam_entries:
3063 	otx2_mcam_flow_del(pf);
3064 err_ptp_destroy:
3065 	otx2_ptp_destroy(pf);
3066 err_detach_rsrc:
3067 	if (pf->hw.lmt_info)
3068 		free_percpu(pf->hw.lmt_info);
3069 	if (test_bit(CN10K_LMTST, &pf->hw.cap_flag))
3070 		qmem_free(pf->dev, pf->dync_lmt);
3071 	otx2_detach_resources(&pf->mbox);
3072 err_disable_mbox_intr:
3073 	otx2_disable_mbox_intr(pf);
3074 err_mbox_destroy:
3075 	otx2_pfaf_mbox_destroy(pf);
3076 err_free_irq_vectors:
3077 	pci_free_irq_vectors(hw->pdev);
3078 err_free_netdev:
3079 	pci_set_drvdata(pdev, NULL);
3080 	free_netdev(netdev);
3081 err_release_regions:
3082 	pci_release_regions(pdev);
3083 	return err;
3084 }
3085 
3086 static void otx2_vf_link_event_task(struct work_struct *work)
3087 {
3088 	struct otx2_vf_config *config;
3089 	struct cgx_link_info_msg *req;
3090 	struct mbox_msghdr *msghdr;
3091 	struct otx2_nic *pf;
3092 	int vf_idx;
3093 
3094 	config = container_of(work, struct otx2_vf_config,
3095 			      link_event_work.work);
3096 	vf_idx = config - config->pf->vf_configs;
3097 	pf = config->pf;
3098 
3099 	msghdr = otx2_mbox_alloc_msg_rsp(&pf->mbox_pfvf[0].mbox_up, vf_idx,
3100 					 sizeof(*req), sizeof(struct msg_rsp));
3101 	if (!msghdr) {
3102 		dev_err(pf->dev, "Failed to create VF%d link event\n", vf_idx);
3103 		return;
3104 	}
3105 
3106 	req = (struct cgx_link_info_msg *)msghdr;
3107 	req->hdr.id = MBOX_MSG_CGX_LINK_EVENT;
3108 	req->hdr.sig = OTX2_MBOX_REQ_SIG;
3109 	memcpy(&req->link_info, &pf->linfo, sizeof(req->link_info));
3110 
3111 	otx2_sync_mbox_up_msg(&pf->mbox_pfvf[0], vf_idx);
3112 }
3113 
3114 static int otx2_sriov_enable(struct pci_dev *pdev, int numvfs)
3115 {
3116 	struct net_device *netdev = pci_get_drvdata(pdev);
3117 	struct otx2_nic *pf = netdev_priv(netdev);
3118 	int ret;
3119 
3120 	/* Init PF <=> VF mailbox stuff */
3121 	ret = otx2_pfvf_mbox_init(pf, numvfs);
3122 	if (ret)
3123 		return ret;
3124 
3125 	ret = otx2_register_pfvf_mbox_intr(pf, numvfs);
3126 	if (ret)
3127 		goto free_mbox;
3128 
3129 	ret = otx2_pf_flr_init(pf, numvfs);
3130 	if (ret)
3131 		goto free_intr;
3132 
3133 	ret = otx2_register_flr_me_intr(pf, numvfs);
3134 	if (ret)
3135 		goto free_flr;
3136 
3137 	ret = pci_enable_sriov(pdev, numvfs);
3138 	if (ret)
3139 		goto free_flr_intr;
3140 
3141 	return numvfs;
3142 free_flr_intr:
3143 	otx2_disable_flr_me_intr(pf);
3144 free_flr:
3145 	otx2_flr_wq_destroy(pf);
3146 free_intr:
3147 	otx2_disable_pfvf_mbox_intr(pf, numvfs);
3148 free_mbox:
3149 	otx2_pfvf_mbox_destroy(pf);
3150 	return ret;
3151 }
3152 
3153 static int otx2_sriov_disable(struct pci_dev *pdev)
3154 {
3155 	struct net_device *netdev = pci_get_drvdata(pdev);
3156 	struct otx2_nic *pf = netdev_priv(netdev);
3157 	int numvfs = pci_num_vf(pdev);
3158 
3159 	if (!numvfs)
3160 		return 0;
3161 
3162 	pci_disable_sriov(pdev);
3163 
3164 	otx2_disable_flr_me_intr(pf);
3165 	otx2_flr_wq_destroy(pf);
3166 	otx2_disable_pfvf_mbox_intr(pf, numvfs);
3167 	otx2_pfvf_mbox_destroy(pf);
3168 
3169 	return 0;
3170 }
3171 
3172 static int otx2_sriov_configure(struct pci_dev *pdev, int numvfs)
3173 {
3174 	if (numvfs == 0)
3175 		return otx2_sriov_disable(pdev);
3176 	else
3177 		return otx2_sriov_enable(pdev, numvfs);
3178 }
3179 
3180 static void otx2_remove(struct pci_dev *pdev)
3181 {
3182 	struct net_device *netdev = pci_get_drvdata(pdev);
3183 	struct otx2_nic *pf;
3184 
3185 	if (!netdev)
3186 		return;
3187 
3188 	pf = netdev_priv(netdev);
3189 
3190 	pf->flags |= OTX2_FLAG_PF_SHUTDOWN;
3191 
3192 	if (pf->flags & OTX2_FLAG_TX_TSTAMP_ENABLED)
3193 		otx2_config_hw_tx_tstamp(pf, false);
3194 	if (pf->flags & OTX2_FLAG_RX_TSTAMP_ENABLED)
3195 		otx2_config_hw_rx_tstamp(pf, false);
3196 
3197 	/* Disable 802.3x pause frames */
3198 	if (pf->flags & OTX2_FLAG_RX_PAUSE_ENABLED ||
3199 	    (pf->flags & OTX2_FLAG_TX_PAUSE_ENABLED)) {
3200 		pf->flags &= ~OTX2_FLAG_RX_PAUSE_ENABLED;
3201 		pf->flags &= ~OTX2_FLAG_TX_PAUSE_ENABLED;
3202 		otx2_config_pause_frm(pf);
3203 	}
3204 
3205 #ifdef CONFIG_DCB
3206 	/* Disable PFC config */
3207 	if (pf->pfc_en) {
3208 		pf->pfc_en = 0;
3209 		otx2_config_priority_flow_ctrl(pf);
3210 	}
3211 #endif
3212 	cancel_work_sync(&pf->reset_task);
3213 	/* Disable link notifications */
3214 	otx2_cgx_config_linkevents(pf, false);
3215 
3216 	otx2_unregister_dl(pf);
3217 	unregister_netdev(netdev);
3218 	cn10k_mcs_free(pf);
3219 	otx2_sriov_disable(pf->pdev);
3220 	otx2_sriov_vfcfg_cleanup(pf);
3221 	if (pf->otx2_wq)
3222 		destroy_workqueue(pf->otx2_wq);
3223 
3224 	otx2_ptp_destroy(pf);
3225 	otx2_mcam_flow_del(pf);
3226 	otx2_shutdown_tc(pf);
3227 	otx2_shutdown_qos(pf);
3228 	otx2_detach_resources(&pf->mbox);
3229 	if (pf->hw.lmt_info)
3230 		free_percpu(pf->hw.lmt_info);
3231 	if (test_bit(CN10K_LMTST, &pf->hw.cap_flag))
3232 		qmem_free(pf->dev, pf->dync_lmt);
3233 	otx2_disable_mbox_intr(pf);
3234 	otx2_pfaf_mbox_destroy(pf);
3235 	pci_free_irq_vectors(pf->pdev);
3236 	pci_set_drvdata(pdev, NULL);
3237 	free_netdev(netdev);
3238 
3239 	pci_release_regions(pdev);
3240 }
3241 
3242 static struct pci_driver otx2_pf_driver = {
3243 	.name = DRV_NAME,
3244 	.id_table = otx2_pf_id_table,
3245 	.probe = otx2_probe,
3246 	.shutdown = otx2_remove,
3247 	.remove = otx2_remove,
3248 	.sriov_configure = otx2_sriov_configure
3249 };
3250 
3251 static int __init otx2_rvupf_init_module(void)
3252 {
3253 	pr_info("%s: %s\n", DRV_NAME, DRV_STRING);
3254 
3255 	return pci_register_driver(&otx2_pf_driver);
3256 }
3257 
3258 static void __exit otx2_rvupf_cleanup_module(void)
3259 {
3260 	pci_unregister_driver(&otx2_pf_driver);
3261 }
3262 
3263 module_init(otx2_rvupf_init_module);
3264 module_exit(otx2_rvupf_cleanup_module);
3265