xref: /linux/drivers/net/ethernet/realtek/rtase/rtase_main.c (revision 7f71507851fc7764b36a3221839607d3a45c2025)
1 // SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause
2 /*
3  *  rtase is the Linux device driver released for Realtek Automotive Switch
4  *  controllers with PCI-Express interface.
5  *
6  *  Copyright(c) 2024 Realtek Semiconductor Corp.
7  *
8  *  Below is a simplified block diagram of the chip and its relevant interfaces.
9  *
10  *               *************************
11  *               *                       *
12  *               *  CPU network device   *
13  *               *                       *
14  *               *   +-------------+     *
15  *               *   |  PCIE Host  |     *
16  *               ***********++************
17  *                          ||
18  *                         PCIE
19  *                          ||
20  *      ********************++**********************
21  *      *            | PCIE Endpoint |             *
22  *      *            +---------------+             *
23  *      *                | GMAC |                  *
24  *      *                +--++--+  Realtek         *
25  *      *                   ||     RTL90xx Series  *
26  *      *                   ||                     *
27  *      *     +-------------++----------------+    *
28  *      *     |           | MAC |             |    *
29  *      *     |           +-----+             |    *
30  *      *     |                               |    *
31  *      *     |     Ethernet Switch Core      |    *
32  *      *     |                               |    *
33  *      *     |   +-----+           +-----+   |    *
34  *      *     |   | MAC |...........| MAC |   |    *
35  *      *     +---+-----+-----------+-----+---+    *
36  *      *         | PHY |...........| PHY |        *
37  *      *         +--++-+           +--++-+        *
38  *      *************||****************||***********
39  *
40  *  The block of the Realtek RTL90xx series is our entire chip architecture,
41  *  the GMAC is connected to the switch core, and there is no PHY in between.
42  *  In addition, this driver is mainly used to control GMAC, but does not
43  *  control the switch core, so it is not the same as DSA. Linux only plays
44  *  the role of a normal leaf node in this model.
45  */
46 
47 #include <linux/crc32.h>
48 #include <linux/dma-mapping.h>
49 #include <linux/etherdevice.h>
50 #include <linux/if_vlan.h>
51 #include <linux/in.h>
52 #include <linux/init.h>
53 #include <linux/interrupt.h>
54 #include <linux/io.h>
55 #include <linux/iopoll.h>
56 #include <linux/ip.h>
57 #include <linux/ipv6.h>
58 #include <linux/mdio.h>
59 #include <linux/module.h>
60 #include <linux/netdevice.h>
61 #include <linux/pci.h>
62 #include <linux/pm_runtime.h>
63 #include <linux/prefetch.h>
64 #include <linux/rtnetlink.h>
65 #include <linux/tcp.h>
66 #include <asm/irq.h>
67 #include <net/ip6_checksum.h>
68 #include <net/netdev_queues.h>
69 #include <net/page_pool/helpers.h>
70 #include <net/pkt_cls.h>
71 
72 #include "rtase.h"
73 
74 #define RTK_OPTS1_DEBUG_VALUE 0x0BADBEEF
75 #define RTK_MAGIC_NUMBER      0x0BADBADBADBADBAD
76 
77 static const struct pci_device_id rtase_pci_tbl[] = {
78 	{PCI_VDEVICE(REALTEK, 0x906A)},
79 	{}
80 };
81 
82 MODULE_DEVICE_TABLE(pci, rtase_pci_tbl);
83 
84 MODULE_AUTHOR("Realtek ARD Software Team");
85 MODULE_DESCRIPTION("Network Driver for the PCIe interface of Realtek Automotive Ethernet Switch");
86 MODULE_LICENSE("Dual BSD/GPL");
87 
88 struct rtase_counters {
89 	__le64 tx_packets;
90 	__le64 rx_packets;
91 	__le64 tx_errors;
92 	__le32 rx_errors;
93 	__le16 rx_missed;
94 	__le16 align_errors;
95 	__le32 tx_one_collision;
96 	__le32 tx_multi_collision;
97 	__le64 rx_unicast;
98 	__le64 rx_broadcast;
99 	__le32 rx_multicast;
100 	__le16 tx_aborted;
101 	__le16 tx_underrun;
102 } __packed;
103 
104 static void rtase_w8(const struct rtase_private *tp, u16 reg, u8 val8)
105 {
106 	writeb(val8, tp->mmio_addr + reg);
107 }
108 
109 static void rtase_w16(const struct rtase_private *tp, u16 reg, u16 val16)
110 {
111 	writew(val16, tp->mmio_addr + reg);
112 }
113 
114 static void rtase_w32(const struct rtase_private *tp, u16 reg, u32 val32)
115 {
116 	writel(val32, tp->mmio_addr + reg);
117 }
118 
119 static u8 rtase_r8(const struct rtase_private *tp, u16 reg)
120 {
121 	return readb(tp->mmio_addr + reg);
122 }
123 
124 static u16 rtase_r16(const struct rtase_private *tp, u16 reg)
125 {
126 	return readw(tp->mmio_addr + reg);
127 }
128 
129 static u32 rtase_r32(const struct rtase_private *tp, u16 reg)
130 {
131 	return readl(tp->mmio_addr + reg);
132 }
133 
134 static void rtase_free_desc(struct rtase_private *tp)
135 {
136 	struct pci_dev *pdev = tp->pdev;
137 	u32 i;
138 
139 	for (i = 0; i < tp->func_tx_queue_num; i++) {
140 		if (!tp->tx_ring[i].desc)
141 			continue;
142 
143 		dma_free_coherent(&pdev->dev, RTASE_TX_RING_DESC_SIZE,
144 				  tp->tx_ring[i].desc,
145 				  tp->tx_ring[i].phy_addr);
146 		tp->tx_ring[i].desc = NULL;
147 	}
148 
149 	for (i = 0; i < tp->func_rx_queue_num; i++) {
150 		if (!tp->rx_ring[i].desc)
151 			continue;
152 
153 		dma_free_coherent(&pdev->dev, RTASE_RX_RING_DESC_SIZE,
154 				  tp->rx_ring[i].desc,
155 				  tp->rx_ring[i].phy_addr);
156 		tp->rx_ring[i].desc = NULL;
157 	}
158 }
159 
160 static int rtase_alloc_desc(struct rtase_private *tp)
161 {
162 	struct pci_dev *pdev = tp->pdev;
163 	u32 i;
164 
165 	/* rx and tx descriptors needs 256 bytes alignment.
166 	 * dma_alloc_coherent provides more.
167 	 */
168 	for (i = 0; i < tp->func_tx_queue_num; i++) {
169 		tp->tx_ring[i].desc =
170 				dma_alloc_coherent(&pdev->dev,
171 						   RTASE_TX_RING_DESC_SIZE,
172 						   &tp->tx_ring[i].phy_addr,
173 						   GFP_KERNEL);
174 		if (!tp->tx_ring[i].desc)
175 			goto err_out;
176 	}
177 
178 	for (i = 0; i < tp->func_rx_queue_num; i++) {
179 		tp->rx_ring[i].desc =
180 				dma_alloc_coherent(&pdev->dev,
181 						   RTASE_RX_RING_DESC_SIZE,
182 						   &tp->rx_ring[i].phy_addr,
183 						   GFP_KERNEL);
184 		if (!tp->rx_ring[i].desc)
185 			goto err_out;
186 	}
187 
188 	return 0;
189 
190 err_out:
191 	rtase_free_desc(tp);
192 	return -ENOMEM;
193 }
194 
195 static void rtase_unmap_tx_skb(struct pci_dev *pdev, u32 len,
196 			       struct rtase_tx_desc *desc)
197 {
198 	dma_unmap_single(&pdev->dev, le64_to_cpu(desc->addr), len,
199 			 DMA_TO_DEVICE);
200 	desc->opts1 = cpu_to_le32(RTK_OPTS1_DEBUG_VALUE);
201 	desc->opts2 = 0x00;
202 	desc->addr = cpu_to_le64(RTK_MAGIC_NUMBER);
203 }
204 
205 static void rtase_tx_clear_range(struct rtase_ring *ring, u32 start, u32 n)
206 {
207 	struct rtase_tx_desc *desc_base = ring->desc;
208 	struct rtase_private *tp = ring->ivec->tp;
209 	u32 i;
210 
211 	for (i = 0; i < n; i++) {
212 		u32 entry = (start + i) % RTASE_NUM_DESC;
213 		struct rtase_tx_desc *desc = desc_base + entry;
214 		u32 len = ring->mis.len[entry];
215 		struct sk_buff *skb;
216 
217 		if (len == 0)
218 			continue;
219 
220 		rtase_unmap_tx_skb(tp->pdev, len, desc);
221 		ring->mis.len[entry] = 0;
222 		skb = ring->skbuff[entry];
223 		if (!skb)
224 			continue;
225 
226 		tp->stats.tx_dropped++;
227 		dev_kfree_skb_any(skb);
228 		ring->skbuff[entry] = NULL;
229 	}
230 }
231 
232 static void rtase_tx_clear(struct rtase_private *tp)
233 {
234 	struct rtase_ring *ring;
235 	u16 i;
236 
237 	for (i = 0; i < tp->func_tx_queue_num; i++) {
238 		ring = &tp->tx_ring[i];
239 		rtase_tx_clear_range(ring, ring->dirty_idx, RTASE_NUM_DESC);
240 		ring->cur_idx = 0;
241 		ring->dirty_idx = 0;
242 	}
243 }
244 
245 static void rtase_mark_to_asic(union rtase_rx_desc *desc, u32 rx_buf_sz)
246 {
247 	u32 eor = le32_to_cpu(desc->desc_cmd.opts1) & RTASE_RING_END;
248 
249 	desc->desc_status.opts2 = 0;
250 	/* force memory writes to complete before releasing descriptor */
251 	dma_wmb();
252 	WRITE_ONCE(desc->desc_cmd.opts1,
253 		   cpu_to_le32(RTASE_DESC_OWN | eor | rx_buf_sz));
254 }
255 
256 static u32 rtase_tx_avail(struct rtase_ring *ring)
257 {
258 	return READ_ONCE(ring->dirty_idx) + RTASE_NUM_DESC -
259 	       READ_ONCE(ring->cur_idx);
260 }
261 
262 static int tx_handler(struct rtase_ring *ring, int budget)
263 {
264 	const struct rtase_private *tp = ring->ivec->tp;
265 	struct net_device *dev = tp->dev;
266 	u32 dirty_tx, tx_left;
267 	u32 bytes_compl = 0;
268 	u32 pkts_compl = 0;
269 	int workdone = 0;
270 
271 	dirty_tx = ring->dirty_idx;
272 	tx_left = READ_ONCE(ring->cur_idx) - dirty_tx;
273 
274 	while (tx_left > 0) {
275 		u32 entry = dirty_tx % RTASE_NUM_DESC;
276 		struct rtase_tx_desc *desc = ring->desc +
277 				       sizeof(struct rtase_tx_desc) * entry;
278 		u32 status;
279 
280 		status = le32_to_cpu(desc->opts1);
281 
282 		if (status & RTASE_DESC_OWN)
283 			break;
284 
285 		rtase_unmap_tx_skb(tp->pdev, ring->mis.len[entry], desc);
286 		ring->mis.len[entry] = 0;
287 		if (ring->skbuff[entry]) {
288 			pkts_compl++;
289 			bytes_compl += ring->skbuff[entry]->len;
290 			napi_consume_skb(ring->skbuff[entry], budget);
291 			ring->skbuff[entry] = NULL;
292 		}
293 
294 		dirty_tx++;
295 		tx_left--;
296 		workdone++;
297 
298 		if (workdone == RTASE_TX_BUDGET_DEFAULT)
299 			break;
300 	}
301 
302 	if (ring->dirty_idx != dirty_tx) {
303 		dev_sw_netstats_tx_add(dev, pkts_compl, bytes_compl);
304 		WRITE_ONCE(ring->dirty_idx, dirty_tx);
305 
306 		netif_subqueue_completed_wake(dev, ring->index, pkts_compl,
307 					      bytes_compl,
308 					      rtase_tx_avail(ring),
309 					      RTASE_TX_START_THRS);
310 
311 		if (ring->cur_idx != dirty_tx)
312 			rtase_w8(tp, RTASE_TPPOLL, BIT(ring->index));
313 	}
314 
315 	return 0;
316 }
317 
318 static void rtase_tx_desc_init(struct rtase_private *tp, u16 idx)
319 {
320 	struct rtase_ring *ring = &tp->tx_ring[idx];
321 	struct rtase_tx_desc *desc;
322 	u32 i;
323 
324 	memset(ring->desc, 0x0, RTASE_TX_RING_DESC_SIZE);
325 	memset(ring->skbuff, 0x0, sizeof(ring->skbuff));
326 	ring->cur_idx = 0;
327 	ring->dirty_idx = 0;
328 	ring->index = idx;
329 	ring->alloc_fail = 0;
330 
331 	for (i = 0; i < RTASE_NUM_DESC; i++) {
332 		ring->mis.len[i] = 0;
333 		if ((RTASE_NUM_DESC - 1) == i) {
334 			desc = ring->desc + sizeof(struct rtase_tx_desc) * i;
335 			desc->opts1 = cpu_to_le32(RTASE_RING_END);
336 		}
337 	}
338 
339 	ring->ring_handler = tx_handler;
340 	if (idx < 4) {
341 		ring->ivec = &tp->int_vector[idx];
342 		list_add_tail(&ring->ring_entry,
343 			      &tp->int_vector[idx].ring_list);
344 	} else {
345 		ring->ivec = &tp->int_vector[0];
346 		list_add_tail(&ring->ring_entry, &tp->int_vector[0].ring_list);
347 	}
348 }
349 
350 static void rtase_map_to_asic(union rtase_rx_desc *desc, dma_addr_t mapping,
351 			      u32 rx_buf_sz)
352 {
353 	desc->desc_cmd.addr = cpu_to_le64(mapping);
354 
355 	rtase_mark_to_asic(desc, rx_buf_sz);
356 }
357 
358 static void rtase_make_unusable_by_asic(union rtase_rx_desc *desc)
359 {
360 	desc->desc_cmd.addr = cpu_to_le64(RTK_MAGIC_NUMBER);
361 	desc->desc_cmd.opts1 &= ~cpu_to_le32(RTASE_DESC_OWN | RSVD_MASK);
362 }
363 
364 static int rtase_alloc_rx_data_buf(struct rtase_ring *ring,
365 				   void **p_data_buf,
366 				   union rtase_rx_desc *desc,
367 				   dma_addr_t *rx_phy_addr)
368 {
369 	struct rtase_int_vector *ivec = ring->ivec;
370 	const struct rtase_private *tp = ivec->tp;
371 	dma_addr_t mapping;
372 	struct page *page;
373 
374 	page = page_pool_dev_alloc_pages(tp->page_pool);
375 	if (!page) {
376 		ring->alloc_fail++;
377 		goto err_out;
378 	}
379 
380 	*p_data_buf = page_address(page);
381 	mapping = page_pool_get_dma_addr(page);
382 	*rx_phy_addr = mapping;
383 	rtase_map_to_asic(desc, mapping, tp->rx_buf_sz);
384 
385 	return 0;
386 
387 err_out:
388 	rtase_make_unusable_by_asic(desc);
389 
390 	return -ENOMEM;
391 }
392 
393 static u32 rtase_rx_ring_fill(struct rtase_ring *ring, u32 ring_start,
394 			      u32 ring_end)
395 {
396 	union rtase_rx_desc *desc_base = ring->desc;
397 	u32 cur;
398 
399 	for (cur = ring_start; ring_end - cur > 0; cur++) {
400 		u32 i = cur % RTASE_NUM_DESC;
401 		union rtase_rx_desc *desc = desc_base + i;
402 		int ret;
403 
404 		if (ring->data_buf[i])
405 			continue;
406 
407 		ret = rtase_alloc_rx_data_buf(ring, &ring->data_buf[i], desc,
408 					      &ring->mis.data_phy_addr[i]);
409 		if (ret)
410 			break;
411 	}
412 
413 	return cur - ring_start;
414 }
415 
416 static void rtase_mark_as_last_descriptor(union rtase_rx_desc *desc)
417 {
418 	desc->desc_cmd.opts1 |= cpu_to_le32(RTASE_RING_END);
419 }
420 
421 static void rtase_rx_ring_clear(struct page_pool *page_pool,
422 				struct rtase_ring *ring)
423 {
424 	union rtase_rx_desc *desc;
425 	struct page *page;
426 	u32 i;
427 
428 	for (i = 0; i < RTASE_NUM_DESC; i++) {
429 		desc = ring->desc + sizeof(union rtase_rx_desc) * i;
430 		page = virt_to_head_page(ring->data_buf[i]);
431 
432 		if (ring->data_buf[i])
433 			page_pool_put_full_page(page_pool, page, true);
434 
435 		rtase_make_unusable_by_asic(desc);
436 	}
437 }
438 
439 static int rtase_fragmented_frame(u32 status)
440 {
441 	return (status & (RTASE_RX_FIRST_FRAG | RTASE_RX_LAST_FRAG)) !=
442 	       (RTASE_RX_FIRST_FRAG | RTASE_RX_LAST_FRAG);
443 }
444 
445 static void rtase_rx_csum(const struct rtase_private *tp, struct sk_buff *skb,
446 			  const union rtase_rx_desc *desc)
447 {
448 	u32 opts2 = le32_to_cpu(desc->desc_status.opts2);
449 
450 	/* rx csum offload */
451 	if (((opts2 & RTASE_RX_V4F) && !(opts2 & RTASE_RX_IPF)) ||
452 	    (opts2 & RTASE_RX_V6F)) {
453 		if (((opts2 & RTASE_RX_TCPT) && !(opts2 & RTASE_RX_TCPF)) ||
454 		    ((opts2 & RTASE_RX_UDPT) && !(opts2 & RTASE_RX_UDPF)))
455 			skb->ip_summed = CHECKSUM_UNNECESSARY;
456 		else
457 			skb->ip_summed = CHECKSUM_NONE;
458 	} else {
459 		skb->ip_summed = CHECKSUM_NONE;
460 	}
461 }
462 
463 static void rtase_rx_vlan_skb(union rtase_rx_desc *desc, struct sk_buff *skb)
464 {
465 	u32 opts2 = le32_to_cpu(desc->desc_status.opts2);
466 
467 	if (!(opts2 & RTASE_RX_VLAN_TAG))
468 		return;
469 
470 	__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
471 			       swab16(opts2 & RTASE_VLAN_TAG_MASK));
472 }
473 
474 static void rtase_rx_skb(const struct rtase_ring *ring, struct sk_buff *skb)
475 {
476 	struct rtase_int_vector *ivec = ring->ivec;
477 
478 	napi_gro_receive(&ivec->napi, skb);
479 }
480 
481 static int rx_handler(struct rtase_ring *ring, int budget)
482 {
483 	union rtase_rx_desc *desc_base = ring->desc;
484 	u32 pkt_size, cur_rx, delta, entry, status;
485 	struct rtase_private *tp = ring->ivec->tp;
486 	struct net_device *dev = tp->dev;
487 	union rtase_rx_desc *desc;
488 	struct sk_buff *skb;
489 	int workdone = 0;
490 
491 	cur_rx = ring->cur_idx;
492 	entry = cur_rx % RTASE_NUM_DESC;
493 	desc = &desc_base[entry];
494 
495 	while (workdone < budget) {
496 		status = le32_to_cpu(desc->desc_status.opts1);
497 
498 		if (status & RTASE_DESC_OWN)
499 			break;
500 
501 		/* This barrier is needed to keep us from reading
502 		 * any other fields out of the rx descriptor until
503 		 * we know the status of RTASE_DESC_OWN
504 		 */
505 		dma_rmb();
506 
507 		if (unlikely(status & RTASE_RX_RES)) {
508 			if (net_ratelimit())
509 				netdev_warn(dev, "Rx ERROR. status = %08x\n",
510 					    status);
511 
512 			tp->stats.rx_errors++;
513 
514 			if (status & (RTASE_RX_RWT | RTASE_RX_RUNT))
515 				tp->stats.rx_length_errors++;
516 
517 			if (status & RTASE_RX_CRC)
518 				tp->stats.rx_crc_errors++;
519 
520 			if (dev->features & NETIF_F_RXALL)
521 				goto process_pkt;
522 
523 			rtase_mark_to_asic(desc, tp->rx_buf_sz);
524 			goto skip_process_pkt;
525 		}
526 
527 process_pkt:
528 		pkt_size = status & RTASE_RX_PKT_SIZE_MASK;
529 		if (likely(!(dev->features & NETIF_F_RXFCS)))
530 			pkt_size -= ETH_FCS_LEN;
531 
532 		/* The driver does not support incoming fragmented frames.
533 		 * They are seen as a symptom of over-mtu sized frames.
534 		 */
535 		if (unlikely(rtase_fragmented_frame(status))) {
536 			tp->stats.rx_dropped++;
537 			tp->stats.rx_length_errors++;
538 			rtase_mark_to_asic(desc, tp->rx_buf_sz);
539 			goto skip_process_pkt;
540 		}
541 
542 		dma_sync_single_for_cpu(&tp->pdev->dev,
543 					ring->mis.data_phy_addr[entry],
544 					tp->rx_buf_sz, DMA_FROM_DEVICE);
545 
546 		skb = build_skb(ring->data_buf[entry], PAGE_SIZE);
547 		if (!skb) {
548 			tp->stats.rx_dropped++;
549 			rtase_mark_to_asic(desc, tp->rx_buf_sz);
550 			goto skip_process_pkt;
551 		}
552 		ring->data_buf[entry] = NULL;
553 
554 		if (dev->features & NETIF_F_RXCSUM)
555 			rtase_rx_csum(tp, skb, desc);
556 
557 		skb_put(skb, pkt_size);
558 		skb_mark_for_recycle(skb);
559 		skb->protocol = eth_type_trans(skb, dev);
560 
561 		if (skb->pkt_type == PACKET_MULTICAST)
562 			tp->stats.multicast++;
563 
564 		rtase_rx_vlan_skb(desc, skb);
565 		rtase_rx_skb(ring, skb);
566 
567 		dev_sw_netstats_rx_add(dev, pkt_size);
568 
569 skip_process_pkt:
570 		workdone++;
571 		cur_rx++;
572 		entry = cur_rx % RTASE_NUM_DESC;
573 		desc = ring->desc + sizeof(union rtase_rx_desc) * entry;
574 	}
575 
576 	ring->cur_idx = cur_rx;
577 	delta = rtase_rx_ring_fill(ring, ring->dirty_idx, ring->cur_idx);
578 	ring->dirty_idx += delta;
579 
580 	return workdone;
581 }
582 
583 static void rtase_rx_desc_init(struct rtase_private *tp, u16 idx)
584 {
585 	struct rtase_ring *ring = &tp->rx_ring[idx];
586 	u16 i;
587 
588 	memset(ring->desc, 0x0, RTASE_RX_RING_DESC_SIZE);
589 	memset(ring->data_buf, 0x0, sizeof(ring->data_buf));
590 	ring->cur_idx = 0;
591 	ring->dirty_idx = 0;
592 	ring->index = idx;
593 	ring->alloc_fail = 0;
594 
595 	for (i = 0; i < RTASE_NUM_DESC; i++)
596 		ring->mis.data_phy_addr[i] = 0;
597 
598 	ring->ring_handler = rx_handler;
599 	ring->ivec = &tp->int_vector[idx];
600 	list_add_tail(&ring->ring_entry, &tp->int_vector[idx].ring_list);
601 }
602 
603 static void rtase_rx_clear(struct rtase_private *tp)
604 {
605 	u32 i;
606 
607 	for (i = 0; i < tp->func_rx_queue_num; i++)
608 		rtase_rx_ring_clear(tp->page_pool, &tp->rx_ring[i]);
609 
610 	page_pool_destroy(tp->page_pool);
611 	tp->page_pool = NULL;
612 }
613 
614 static int rtase_init_ring(const struct net_device *dev)
615 {
616 	struct rtase_private *tp = netdev_priv(dev);
617 	struct page_pool_params pp_params = { 0 };
618 	struct page_pool *page_pool;
619 	u32 num;
620 	u16 i;
621 
622 	pp_params.flags = PP_FLAG_DMA_MAP | PP_FLAG_DMA_SYNC_DEV;
623 	pp_params.order = 0;
624 	pp_params.pool_size = RTASE_NUM_DESC * tp->func_rx_queue_num;
625 	pp_params.nid = dev_to_node(&tp->pdev->dev);
626 	pp_params.dev = &tp->pdev->dev;
627 	pp_params.dma_dir = DMA_FROM_DEVICE;
628 	pp_params.max_len = PAGE_SIZE;
629 	pp_params.offset = 0;
630 
631 	page_pool = page_pool_create(&pp_params);
632 	if (IS_ERR(page_pool)) {
633 		netdev_err(tp->dev, "failed to create page pool\n");
634 		return -ENOMEM;
635 	}
636 
637 	tp->page_pool = page_pool;
638 
639 	for (i = 0; i < tp->func_tx_queue_num; i++)
640 		rtase_tx_desc_init(tp, i);
641 
642 	for (i = 0; i < tp->func_rx_queue_num; i++) {
643 		rtase_rx_desc_init(tp, i);
644 
645 		num = rtase_rx_ring_fill(&tp->rx_ring[i], 0, RTASE_NUM_DESC);
646 		if (num != RTASE_NUM_DESC)
647 			goto err_out;
648 
649 		rtase_mark_as_last_descriptor(tp->rx_ring[i].desc +
650 					      sizeof(union rtase_rx_desc) *
651 					      (RTASE_NUM_DESC - 1));
652 	}
653 
654 	return 0;
655 
656 err_out:
657 	rtase_rx_clear(tp);
658 	return -ENOMEM;
659 }
660 
661 static void rtase_interrupt_mitigation(const struct rtase_private *tp)
662 {
663 	u32 i;
664 
665 	for (i = 0; i < tp->func_tx_queue_num; i++)
666 		rtase_w16(tp, RTASE_INT_MITI_TX + i * 2, tp->tx_int_mit);
667 
668 	for (i = 0; i < tp->func_rx_queue_num; i++)
669 		rtase_w16(tp, RTASE_INT_MITI_RX + i * 2, tp->rx_int_mit);
670 }
671 
672 static void rtase_tally_counter_addr_fill(const struct rtase_private *tp)
673 {
674 	rtase_w32(tp, RTASE_DTCCR4, upper_32_bits(tp->tally_paddr));
675 	rtase_w32(tp, RTASE_DTCCR0, lower_32_bits(tp->tally_paddr));
676 }
677 
678 static void rtase_tally_counter_clear(const struct rtase_private *tp)
679 {
680 	u32 cmd = lower_32_bits(tp->tally_paddr);
681 
682 	rtase_w32(tp, RTASE_DTCCR4, upper_32_bits(tp->tally_paddr));
683 	rtase_w32(tp, RTASE_DTCCR0, cmd | RTASE_COUNTER_RESET);
684 }
685 
686 static void rtase_desc_addr_fill(const struct rtase_private *tp)
687 {
688 	const struct rtase_ring *ring;
689 	u16 i, cmd, val;
690 	int err;
691 
692 	for (i = 0; i < tp->func_tx_queue_num; i++) {
693 		ring = &tp->tx_ring[i];
694 
695 		rtase_w32(tp, RTASE_TX_DESC_ADDR0,
696 			  lower_32_bits(ring->phy_addr));
697 		rtase_w32(tp, RTASE_TX_DESC_ADDR4,
698 			  upper_32_bits(ring->phy_addr));
699 
700 		cmd = i | RTASE_TX_DESC_CMD_WE | RTASE_TX_DESC_CMD_CS;
701 		rtase_w16(tp, RTASE_TX_DESC_COMMAND, cmd);
702 
703 		err = read_poll_timeout(rtase_r16, val,
704 					!(val & RTASE_TX_DESC_CMD_CS), 10,
705 					1000, false, tp,
706 					RTASE_TX_DESC_COMMAND);
707 
708 		if (err == -ETIMEDOUT)
709 			netdev_err(tp->dev,
710 				   "error occurred in fill tx descriptor\n");
711 	}
712 
713 	for (i = 0; i < tp->func_rx_queue_num; i++) {
714 		ring = &tp->rx_ring[i];
715 
716 		if (i == 0) {
717 			rtase_w32(tp, RTASE_Q0_RX_DESC_ADDR0,
718 				  lower_32_bits(ring->phy_addr));
719 			rtase_w32(tp, RTASE_Q0_RX_DESC_ADDR4,
720 				  upper_32_bits(ring->phy_addr));
721 		} else {
722 			rtase_w32(tp, (RTASE_Q1_RX_DESC_ADDR0 + ((i - 1) * 8)),
723 				  lower_32_bits(ring->phy_addr));
724 			rtase_w32(tp, (RTASE_Q1_RX_DESC_ADDR4 + ((i - 1) * 8)),
725 				  upper_32_bits(ring->phy_addr));
726 		}
727 	}
728 }
729 
730 static void rtase_hw_set_features(const struct net_device *dev,
731 				  netdev_features_t features)
732 {
733 	const struct rtase_private *tp = netdev_priv(dev);
734 	u16 rx_config, val;
735 
736 	rx_config = rtase_r16(tp, RTASE_RX_CONFIG_0);
737 	if (features & NETIF_F_RXALL)
738 		rx_config |= (RTASE_ACCEPT_ERR | RTASE_ACCEPT_RUNT);
739 	else
740 		rx_config &= ~(RTASE_ACCEPT_ERR | RTASE_ACCEPT_RUNT);
741 
742 	rtase_w16(tp, RTASE_RX_CONFIG_0, rx_config);
743 
744 	val = rtase_r16(tp, RTASE_CPLUS_CMD);
745 	if (features & NETIF_F_RXCSUM)
746 		rtase_w16(tp, RTASE_CPLUS_CMD, val | RTASE_RX_CHKSUM);
747 	else
748 		rtase_w16(tp, RTASE_CPLUS_CMD, val & ~RTASE_RX_CHKSUM);
749 
750 	rx_config = rtase_r16(tp, RTASE_RX_CONFIG_1);
751 	if (dev->features & NETIF_F_HW_VLAN_CTAG_RX)
752 		rx_config |= (RTASE_INNER_VLAN_DETAG_EN |
753 			      RTASE_OUTER_VLAN_DETAG_EN);
754 	else
755 		rx_config &= ~(RTASE_INNER_VLAN_DETAG_EN |
756 			       RTASE_OUTER_VLAN_DETAG_EN);
757 
758 	rtase_w16(tp, RTASE_RX_CONFIG_1, rx_config);
759 }
760 
761 static void rtase_hw_set_rx_packet_filter(struct net_device *dev)
762 {
763 	u32 mc_filter[2] = { 0xFFFFFFFF, 0xFFFFFFFF };
764 	struct rtase_private *tp = netdev_priv(dev);
765 	u16 rx_mode;
766 
767 	rx_mode = rtase_r16(tp, RTASE_RX_CONFIG_0) & ~RTASE_ACCEPT_MASK;
768 	rx_mode |= RTASE_ACCEPT_BROADCAST | RTASE_ACCEPT_MYPHYS;
769 
770 	if (dev->flags & IFF_PROMISC) {
771 		rx_mode |= RTASE_ACCEPT_MULTICAST | RTASE_ACCEPT_ALLPHYS;
772 	} else if (dev->flags & IFF_ALLMULTI) {
773 		rx_mode |= RTASE_ACCEPT_MULTICAST;
774 	} else {
775 		struct netdev_hw_addr *hw_addr;
776 
777 		mc_filter[0] = 0;
778 		mc_filter[1] = 0;
779 
780 		netdev_for_each_mc_addr(hw_addr, dev) {
781 			u32 bit_nr = eth_hw_addr_crc(hw_addr);
782 			u32 idx = u32_get_bits(bit_nr, BIT(31));
783 			u32 bit = u32_get_bits(bit_nr,
784 					       RTASE_MULTICAST_FILTER_MASK);
785 
786 			mc_filter[idx] |= BIT(bit);
787 			rx_mode |= RTASE_ACCEPT_MULTICAST;
788 		}
789 	}
790 
791 	if (dev->features & NETIF_F_RXALL)
792 		rx_mode |= RTASE_ACCEPT_ERR | RTASE_ACCEPT_RUNT;
793 
794 	rtase_w32(tp, RTASE_MAR0, swab32(mc_filter[1]));
795 	rtase_w32(tp, RTASE_MAR1, swab32(mc_filter[0]));
796 	rtase_w16(tp, RTASE_RX_CONFIG_0, rx_mode);
797 }
798 
799 static void rtase_irq_dis_and_clear(const struct rtase_private *tp)
800 {
801 	const struct rtase_int_vector *ivec = &tp->int_vector[0];
802 	u32 val1;
803 	u16 val2;
804 	u8 i;
805 
806 	rtase_w32(tp, ivec->imr_addr, 0);
807 	val1 = rtase_r32(tp, ivec->isr_addr);
808 	rtase_w32(tp, ivec->isr_addr, val1);
809 
810 	for (i = 1; i < tp->int_nums; i++) {
811 		ivec = &tp->int_vector[i];
812 		rtase_w16(tp, ivec->imr_addr, 0);
813 		val2 = rtase_r16(tp, ivec->isr_addr);
814 		rtase_w16(tp, ivec->isr_addr, val2);
815 	}
816 }
817 
818 static void rtase_poll_timeout(const struct rtase_private *tp, u32 cond,
819 			       u32 sleep_us, u64 timeout_us, u16 reg)
820 {
821 	int err;
822 	u8 val;
823 
824 	err = read_poll_timeout(rtase_r8, val, val & cond, sleep_us,
825 				timeout_us, false, tp, reg);
826 
827 	if (err == -ETIMEDOUT)
828 		netdev_err(tp->dev, "poll reg 0x00%x timeout\n", reg);
829 }
830 
831 static void rtase_nic_reset(const struct net_device *dev)
832 {
833 	const struct rtase_private *tp = netdev_priv(dev);
834 	u16 rx_config;
835 	u8 val;
836 
837 	rx_config = rtase_r16(tp, RTASE_RX_CONFIG_0);
838 	rtase_w16(tp, RTASE_RX_CONFIG_0, rx_config & ~RTASE_ACCEPT_MASK);
839 
840 	val = rtase_r8(tp, RTASE_MISC);
841 	rtase_w8(tp, RTASE_MISC, val | RTASE_RX_DV_GATE_EN);
842 
843 	val = rtase_r8(tp, RTASE_CHIP_CMD);
844 	rtase_w8(tp, RTASE_CHIP_CMD, val | RTASE_STOP_REQ);
845 	mdelay(2);
846 
847 	rtase_poll_timeout(tp, RTASE_STOP_REQ_DONE, 100, 150000,
848 			   RTASE_CHIP_CMD);
849 
850 	rtase_poll_timeout(tp, RTASE_TX_FIFO_EMPTY, 100, 100000,
851 			   RTASE_FIFOR);
852 
853 	rtase_poll_timeout(tp, RTASE_RX_FIFO_EMPTY, 100, 100000,
854 			   RTASE_FIFOR);
855 
856 	val = rtase_r8(tp, RTASE_CHIP_CMD);
857 	rtase_w8(tp, RTASE_CHIP_CMD, val & ~(RTASE_TE | RTASE_RE));
858 	val = rtase_r8(tp, RTASE_CHIP_CMD);
859 	rtase_w8(tp, RTASE_CHIP_CMD, val & ~RTASE_STOP_REQ);
860 
861 	rtase_w16(tp, RTASE_RX_CONFIG_0, rx_config);
862 }
863 
864 static void rtase_hw_reset(const struct net_device *dev)
865 {
866 	const struct rtase_private *tp = netdev_priv(dev);
867 
868 	rtase_irq_dis_and_clear(tp);
869 
870 	rtase_nic_reset(dev);
871 }
872 
873 static void rtase_set_rx_queue(const struct rtase_private *tp)
874 {
875 	u16 reg_data;
876 
877 	reg_data = rtase_r16(tp, RTASE_FCR);
878 	switch (tp->func_rx_queue_num) {
879 	case 1:
880 		u16p_replace_bits(&reg_data, 0x1, RTASE_FCR_RXQ_MASK);
881 		break;
882 	case 2:
883 		u16p_replace_bits(&reg_data, 0x2, RTASE_FCR_RXQ_MASK);
884 		break;
885 	case 4:
886 		u16p_replace_bits(&reg_data, 0x3, RTASE_FCR_RXQ_MASK);
887 		break;
888 	}
889 	rtase_w16(tp, RTASE_FCR, reg_data);
890 }
891 
892 static void rtase_set_tx_queue(const struct rtase_private *tp)
893 {
894 	u16 reg_data;
895 
896 	reg_data = rtase_r16(tp, RTASE_TX_CONFIG_1);
897 	switch (tp->tx_queue_ctrl) {
898 	case 1:
899 		u16p_replace_bits(&reg_data, 0x0, RTASE_TC_MODE_MASK);
900 		break;
901 	case 2:
902 		u16p_replace_bits(&reg_data, 0x1, RTASE_TC_MODE_MASK);
903 		break;
904 	case 3:
905 	case 4:
906 		u16p_replace_bits(&reg_data, 0x2, RTASE_TC_MODE_MASK);
907 		break;
908 	default:
909 		u16p_replace_bits(&reg_data, 0x3, RTASE_TC_MODE_MASK);
910 		break;
911 	}
912 	rtase_w16(tp, RTASE_TX_CONFIG_1, reg_data);
913 }
914 
915 static void rtase_hw_config(struct net_device *dev)
916 {
917 	const struct rtase_private *tp = netdev_priv(dev);
918 	u32 reg_data32;
919 	u16 reg_data16;
920 
921 	rtase_hw_reset(dev);
922 
923 	/* set rx dma burst */
924 	reg_data16 = rtase_r16(tp, RTASE_RX_CONFIG_0);
925 	reg_data16 &= ~(RTASE_RX_SINGLE_TAG | RTASE_RX_SINGLE_FETCH);
926 	u16p_replace_bits(&reg_data16, RTASE_RX_DMA_BURST_256,
927 			  RTASE_RX_MX_DMA_MASK);
928 	rtase_w16(tp, RTASE_RX_CONFIG_0, reg_data16);
929 
930 	/* new rx descritpor */
931 	reg_data16 = rtase_r16(tp, RTASE_RX_CONFIG_1);
932 	reg_data16 |= RTASE_RX_NEW_DESC_FORMAT_EN | RTASE_PCIE_NEW_FLOW;
933 	u16p_replace_bits(&reg_data16, 0xF, RTASE_RX_MAX_FETCH_DESC_MASK);
934 	rtase_w16(tp, RTASE_RX_CONFIG_1, reg_data16);
935 
936 	rtase_set_rx_queue(tp);
937 
938 	rtase_interrupt_mitigation(tp);
939 
940 	/* set tx dma burst size and interframe gap time */
941 	reg_data32 = rtase_r32(tp, RTASE_TX_CONFIG_0);
942 	u32p_replace_bits(&reg_data32, RTASE_TX_DMA_BURST_UNLIMITED,
943 			  RTASE_TX_DMA_MASK);
944 	u32p_replace_bits(&reg_data32, RTASE_INTERFRAMEGAP,
945 			  RTASE_TX_INTER_FRAME_GAP_MASK);
946 	rtase_w32(tp, RTASE_TX_CONFIG_0, reg_data32);
947 
948 	/* new tx descriptor */
949 	reg_data16 = rtase_r16(tp, RTASE_TFUN_CTRL);
950 	rtase_w16(tp, RTASE_TFUN_CTRL, reg_data16 |
951 		  RTASE_TX_NEW_DESC_FORMAT_EN);
952 
953 	/* tx fetch desc number */
954 	rtase_w8(tp, RTASE_TDFNR, 0x10);
955 
956 	/* tag num select */
957 	reg_data16 = rtase_r16(tp, RTASE_MTPS);
958 	u16p_replace_bits(&reg_data16, 0x4, RTASE_TAG_NUM_SEL_MASK);
959 	rtase_w16(tp, RTASE_MTPS, reg_data16);
960 
961 	rtase_set_tx_queue(tp);
962 
963 	rtase_w16(tp, RTASE_TOKSEL, 0x5555);
964 
965 	rtase_tally_counter_addr_fill(tp);
966 	rtase_desc_addr_fill(tp);
967 	rtase_hw_set_features(dev, dev->features);
968 
969 	/* enable flow control */
970 	reg_data16 = rtase_r16(tp, RTASE_CPLUS_CMD);
971 	reg_data16 |= (RTASE_FORCE_TXFLOW_EN | RTASE_FORCE_RXFLOW_EN);
972 	rtase_w16(tp, RTASE_CPLUS_CMD, reg_data16);
973 	/* set near fifo threshold - rx missed issue. */
974 	rtase_w16(tp, RTASE_RFIFONFULL, 0x190);
975 
976 	rtase_w16(tp, RTASE_RMS, tp->rx_buf_sz);
977 
978 	rtase_hw_set_rx_packet_filter(dev);
979 }
980 
981 static void rtase_nic_enable(const struct net_device *dev)
982 {
983 	const struct rtase_private *tp = netdev_priv(dev);
984 	u16 rcr = rtase_r16(tp, RTASE_RX_CONFIG_1);
985 	u8 val;
986 
987 	rtase_w16(tp, RTASE_RX_CONFIG_1, rcr & ~RTASE_PCIE_RELOAD_EN);
988 	rtase_w16(tp, RTASE_RX_CONFIG_1, rcr | RTASE_PCIE_RELOAD_EN);
989 
990 	val = rtase_r8(tp, RTASE_CHIP_CMD);
991 	rtase_w8(tp, RTASE_CHIP_CMD, val | RTASE_TE | RTASE_RE);
992 
993 	val = rtase_r8(tp, RTASE_MISC);
994 	rtase_w8(tp, RTASE_MISC, val & ~RTASE_RX_DV_GATE_EN);
995 }
996 
997 static void rtase_enable_hw_interrupt(const struct rtase_private *tp)
998 {
999 	const struct rtase_int_vector *ivec = &tp->int_vector[0];
1000 	u32 i;
1001 
1002 	rtase_w32(tp, ivec->imr_addr, ivec->imr);
1003 
1004 	for (i = 1; i < tp->int_nums; i++) {
1005 		ivec = &tp->int_vector[i];
1006 		rtase_w16(tp, ivec->imr_addr, ivec->imr);
1007 	}
1008 }
1009 
1010 static void rtase_hw_start(const struct net_device *dev)
1011 {
1012 	const struct rtase_private *tp = netdev_priv(dev);
1013 
1014 	rtase_nic_enable(dev);
1015 	rtase_enable_hw_interrupt(tp);
1016 }
1017 
1018 /*  the interrupt handler does RXQ0 and TXQ0, TXQ4~7 interrutp status
1019  */
1020 static irqreturn_t rtase_interrupt(int irq, void *dev_instance)
1021 {
1022 	const struct rtase_private *tp;
1023 	struct rtase_int_vector *ivec;
1024 	u32 status;
1025 
1026 	ivec = dev_instance;
1027 	tp = ivec->tp;
1028 	status = rtase_r32(tp, ivec->isr_addr);
1029 
1030 	rtase_w32(tp, ivec->imr_addr, 0x0);
1031 	rtase_w32(tp, ivec->isr_addr, status & ~RTASE_FOVW);
1032 
1033 	if (napi_schedule_prep(&ivec->napi))
1034 		__napi_schedule(&ivec->napi);
1035 
1036 	return IRQ_HANDLED;
1037 }
1038 
1039 /*  the interrupt handler does RXQ1&TXQ1 or RXQ2&TXQ2 or RXQ3&TXQ3 interrupt
1040  *  status according to interrupt vector
1041  */
1042 static irqreturn_t rtase_q_interrupt(int irq, void *dev_instance)
1043 {
1044 	const struct rtase_private *tp;
1045 	struct rtase_int_vector *ivec;
1046 	u16 status;
1047 
1048 	ivec = dev_instance;
1049 	tp = ivec->tp;
1050 	status = rtase_r16(tp, ivec->isr_addr);
1051 
1052 	rtase_w16(tp, ivec->imr_addr, 0x0);
1053 	rtase_w16(tp, ivec->isr_addr, status);
1054 
1055 	if (napi_schedule_prep(&ivec->napi))
1056 		__napi_schedule(&ivec->napi);
1057 
1058 	return IRQ_HANDLED;
1059 }
1060 
1061 static int rtase_poll(struct napi_struct *napi, int budget)
1062 {
1063 	const struct rtase_int_vector *ivec;
1064 	const struct rtase_private *tp;
1065 	struct rtase_ring *ring;
1066 	int total_workdone = 0;
1067 
1068 	ivec = container_of(napi, struct rtase_int_vector, napi);
1069 	tp = ivec->tp;
1070 
1071 	list_for_each_entry(ring, &ivec->ring_list, ring_entry)
1072 		total_workdone += ring->ring_handler(ring, budget);
1073 
1074 	if (total_workdone >= budget)
1075 		return budget;
1076 
1077 	if (napi_complete_done(napi, total_workdone)) {
1078 		if (!ivec->index)
1079 			rtase_w32(tp, ivec->imr_addr, ivec->imr);
1080 		else
1081 			rtase_w16(tp, ivec->imr_addr, ivec->imr);
1082 	}
1083 
1084 	return total_workdone;
1085 }
1086 
1087 static int rtase_open(struct net_device *dev)
1088 {
1089 	struct rtase_private *tp = netdev_priv(dev);
1090 	const struct pci_dev *pdev = tp->pdev;
1091 	struct rtase_int_vector *ivec;
1092 	u16 i = 0, j;
1093 	int ret;
1094 
1095 	ivec = &tp->int_vector[0];
1096 	tp->rx_buf_sz = RTASE_RX_BUF_SIZE;
1097 
1098 	ret = rtase_alloc_desc(tp);
1099 	if (ret)
1100 		return ret;
1101 
1102 	ret = rtase_init_ring(dev);
1103 	if (ret)
1104 		goto err_free_all_allocated_mem;
1105 
1106 	rtase_hw_config(dev);
1107 
1108 	if (tp->sw_flag & RTASE_SWF_MSIX_ENABLED) {
1109 		ret = request_irq(ivec->irq, rtase_interrupt, 0,
1110 				  dev->name, ivec);
1111 		if (ret)
1112 			goto err_free_all_allocated_irq;
1113 
1114 		/* request other interrupts to handle multiqueue */
1115 		for (i = 1; i < tp->int_nums; i++) {
1116 			ivec = &tp->int_vector[i];
1117 			snprintf(ivec->name, sizeof(ivec->name), "%s_int%i",
1118 				 tp->dev->name, i);
1119 			ret = request_irq(ivec->irq, rtase_q_interrupt, 0,
1120 					  ivec->name, ivec);
1121 			if (ret)
1122 				goto err_free_all_allocated_irq;
1123 		}
1124 	} else {
1125 		ret = request_irq(pdev->irq, rtase_interrupt, 0, dev->name,
1126 				  ivec);
1127 		if (ret)
1128 			goto err_free_all_allocated_mem;
1129 	}
1130 
1131 	rtase_hw_start(dev);
1132 
1133 	for (i = 0; i < tp->int_nums; i++) {
1134 		ivec = &tp->int_vector[i];
1135 		napi_enable(&ivec->napi);
1136 	}
1137 
1138 	netif_carrier_on(dev);
1139 	netif_wake_queue(dev);
1140 
1141 	return 0;
1142 
1143 err_free_all_allocated_irq:
1144 	for (j = 0; j < i; j++)
1145 		free_irq(tp->int_vector[j].irq, &tp->int_vector[j]);
1146 
1147 err_free_all_allocated_mem:
1148 	rtase_free_desc(tp);
1149 
1150 	return ret;
1151 }
1152 
1153 static void rtase_down(struct net_device *dev)
1154 {
1155 	struct rtase_private *tp = netdev_priv(dev);
1156 	struct rtase_int_vector *ivec;
1157 	struct rtase_ring *ring, *tmp;
1158 	u32 i;
1159 
1160 	for (i = 0; i < tp->int_nums; i++) {
1161 		ivec = &tp->int_vector[i];
1162 		napi_disable(&ivec->napi);
1163 		list_for_each_entry_safe(ring, tmp, &ivec->ring_list,
1164 					 ring_entry)
1165 			list_del(&ring->ring_entry);
1166 	}
1167 
1168 	netif_tx_disable(dev);
1169 
1170 	netif_carrier_off(dev);
1171 
1172 	rtase_hw_reset(dev);
1173 
1174 	rtase_tx_clear(tp);
1175 
1176 	rtase_rx_clear(tp);
1177 }
1178 
1179 static int rtase_close(struct net_device *dev)
1180 {
1181 	struct rtase_private *tp = netdev_priv(dev);
1182 	const struct pci_dev *pdev = tp->pdev;
1183 	u32 i;
1184 
1185 	rtase_down(dev);
1186 
1187 	if (tp->sw_flag & RTASE_SWF_MSIX_ENABLED) {
1188 		for (i = 0; i < tp->int_nums; i++)
1189 			free_irq(tp->int_vector[i].irq, &tp->int_vector[i]);
1190 
1191 	} else {
1192 		free_irq(pdev->irq, &tp->int_vector[0]);
1193 	}
1194 
1195 	rtase_free_desc(tp);
1196 
1197 	return 0;
1198 }
1199 
1200 static u32 rtase_tx_vlan_tag(const struct rtase_private *tp,
1201 			     const struct sk_buff *skb)
1202 {
1203 	return (skb_vlan_tag_present(skb)) ?
1204 		(RTASE_TX_VLAN_TAG | swab16(skb_vlan_tag_get(skb))) : 0x00;
1205 }
1206 
1207 static u32 rtase_tx_csum(struct sk_buff *skb, const struct net_device *dev)
1208 {
1209 	u32 csum_cmd = 0;
1210 	u8 ip_protocol;
1211 
1212 	switch (vlan_get_protocol(skb)) {
1213 	case htons(ETH_P_IP):
1214 		csum_cmd = RTASE_TX_IPCS_C;
1215 		ip_protocol = ip_hdr(skb)->protocol;
1216 		break;
1217 
1218 	case htons(ETH_P_IPV6):
1219 		csum_cmd = RTASE_TX_IPV6F_C;
1220 		ip_protocol = ipv6_hdr(skb)->nexthdr;
1221 		break;
1222 
1223 	default:
1224 		ip_protocol = IPPROTO_RAW;
1225 		break;
1226 	}
1227 
1228 	if (ip_protocol == IPPROTO_TCP)
1229 		csum_cmd |= RTASE_TX_TCPCS_C;
1230 	else if (ip_protocol == IPPROTO_UDP)
1231 		csum_cmd |= RTASE_TX_UDPCS_C;
1232 
1233 	csum_cmd |= u32_encode_bits(skb_transport_offset(skb),
1234 				    RTASE_TCPHO_MASK);
1235 
1236 	return csum_cmd;
1237 }
1238 
1239 static int rtase_xmit_frags(struct rtase_ring *ring, struct sk_buff *skb,
1240 			    u32 opts1, u32 opts2)
1241 {
1242 	const struct skb_shared_info *info = skb_shinfo(skb);
1243 	const struct rtase_private *tp = ring->ivec->tp;
1244 	const u8 nr_frags = info->nr_frags;
1245 	struct rtase_tx_desc *txd = NULL;
1246 	u32 cur_frag, entry;
1247 
1248 	entry = ring->cur_idx;
1249 	for (cur_frag = 0; cur_frag < nr_frags; cur_frag++) {
1250 		const skb_frag_t *frag = &info->frags[cur_frag];
1251 		dma_addr_t mapping;
1252 		u32 status, len;
1253 		void *addr;
1254 
1255 		entry = (entry + 1) % RTASE_NUM_DESC;
1256 
1257 		txd = ring->desc + sizeof(struct rtase_tx_desc) * entry;
1258 		len = skb_frag_size(frag);
1259 		addr = skb_frag_address(frag);
1260 		mapping = dma_map_single(&tp->pdev->dev, addr, len,
1261 					 DMA_TO_DEVICE);
1262 
1263 		if (unlikely(dma_mapping_error(&tp->pdev->dev, mapping))) {
1264 			if (unlikely(net_ratelimit()))
1265 				netdev_err(tp->dev,
1266 					   "Failed to map TX fragments DMA!\n");
1267 
1268 			goto err_out;
1269 		}
1270 
1271 		if (((entry + 1) % RTASE_NUM_DESC) == 0)
1272 			status = (opts1 | len | RTASE_RING_END);
1273 		else
1274 			status = opts1 | len;
1275 
1276 		if (cur_frag == (nr_frags - 1)) {
1277 			ring->skbuff[entry] = skb;
1278 			status |= RTASE_TX_LAST_FRAG;
1279 		}
1280 
1281 		ring->mis.len[entry] = len;
1282 		txd->addr = cpu_to_le64(mapping);
1283 		txd->opts2 = cpu_to_le32(opts2);
1284 
1285 		/* make sure the operating fields have been updated */
1286 		dma_wmb();
1287 		txd->opts1 = cpu_to_le32(status);
1288 	}
1289 
1290 	return cur_frag;
1291 
1292 err_out:
1293 	rtase_tx_clear_range(ring, ring->cur_idx + 1, cur_frag);
1294 	return -EIO;
1295 }
1296 
1297 static netdev_tx_t rtase_start_xmit(struct sk_buff *skb,
1298 				    struct net_device *dev)
1299 {
1300 	struct skb_shared_info *shinfo = skb_shinfo(skb);
1301 	struct rtase_private *tp = netdev_priv(dev);
1302 	u32 q_idx, entry, len, opts1, opts2;
1303 	struct netdev_queue *tx_queue;
1304 	bool stop_queue, door_bell;
1305 	u32 mss = shinfo->gso_size;
1306 	struct rtase_tx_desc *txd;
1307 	struct rtase_ring *ring;
1308 	dma_addr_t mapping;
1309 	int frags;
1310 
1311 	/* multiqueues */
1312 	q_idx = skb_get_queue_mapping(skb);
1313 	ring = &tp->tx_ring[q_idx];
1314 	tx_queue = netdev_get_tx_queue(dev, q_idx);
1315 
1316 	if (unlikely(!rtase_tx_avail(ring))) {
1317 		if (net_ratelimit())
1318 			netdev_err(dev,
1319 				   "BUG! Tx Ring full when queue awake!\n");
1320 
1321 		netif_stop_queue(dev);
1322 		return NETDEV_TX_BUSY;
1323 	}
1324 
1325 	entry = ring->cur_idx % RTASE_NUM_DESC;
1326 	txd = ring->desc + sizeof(struct rtase_tx_desc) * entry;
1327 
1328 	opts1 = RTASE_DESC_OWN;
1329 	opts2 = rtase_tx_vlan_tag(tp, skb);
1330 
1331 	/* tcp segmentation offload (or tcp large send) */
1332 	if (mss) {
1333 		if (shinfo->gso_type & SKB_GSO_TCPV4) {
1334 			opts1 |= RTASE_GIANT_SEND_V4;
1335 		} else if (shinfo->gso_type & SKB_GSO_TCPV6) {
1336 			if (skb_cow_head(skb, 0))
1337 				goto err_dma_0;
1338 
1339 			tcp_v6_gso_csum_prep(skb);
1340 			opts1 |= RTASE_GIANT_SEND_V6;
1341 		} else {
1342 			WARN_ON_ONCE(1);
1343 		}
1344 
1345 		opts1 |= u32_encode_bits(skb_transport_offset(skb),
1346 					 RTASE_TCPHO_MASK);
1347 		opts2 |= u32_encode_bits(mss, RTASE_MSS_MASK);
1348 	} else if (skb->ip_summed == CHECKSUM_PARTIAL) {
1349 		opts2 |= rtase_tx_csum(skb, dev);
1350 	}
1351 
1352 	frags = rtase_xmit_frags(ring, skb, opts1, opts2);
1353 	if (unlikely(frags < 0))
1354 		goto err_dma_0;
1355 
1356 	if (frags) {
1357 		len = skb_headlen(skb);
1358 		opts1 |= RTASE_TX_FIRST_FRAG;
1359 	} else {
1360 		len = skb->len;
1361 		ring->skbuff[entry] = skb;
1362 		opts1 |= RTASE_TX_FIRST_FRAG | RTASE_TX_LAST_FRAG;
1363 	}
1364 
1365 	if (((entry + 1) % RTASE_NUM_DESC) == 0)
1366 		opts1 |= (len | RTASE_RING_END);
1367 	else
1368 		opts1 |= len;
1369 
1370 	mapping = dma_map_single(&tp->pdev->dev, skb->data, len,
1371 				 DMA_TO_DEVICE);
1372 
1373 	if (unlikely(dma_mapping_error(&tp->pdev->dev, mapping))) {
1374 		if (unlikely(net_ratelimit()))
1375 			netdev_err(dev, "Failed to map TX DMA!\n");
1376 
1377 		goto err_dma_1;
1378 	}
1379 
1380 	ring->mis.len[entry] = len;
1381 	txd->addr = cpu_to_le64(mapping);
1382 	txd->opts2 = cpu_to_le32(opts2);
1383 	txd->opts1 = cpu_to_le32(opts1 & ~RTASE_DESC_OWN);
1384 
1385 	/* make sure the operating fields have been updated */
1386 	dma_wmb();
1387 
1388 	door_bell = __netdev_tx_sent_queue(tx_queue, skb->len,
1389 					   netdev_xmit_more());
1390 
1391 	txd->opts1 = cpu_to_le32(opts1);
1392 
1393 	skb_tx_timestamp(skb);
1394 
1395 	/* tx needs to see descriptor changes before updated cur_idx */
1396 	smp_wmb();
1397 
1398 	WRITE_ONCE(ring->cur_idx, ring->cur_idx + frags + 1);
1399 
1400 	stop_queue = !netif_subqueue_maybe_stop(dev, ring->index,
1401 						rtase_tx_avail(ring),
1402 						RTASE_TX_STOP_THRS,
1403 						RTASE_TX_START_THRS);
1404 
1405 	if (door_bell || stop_queue)
1406 		rtase_w8(tp, RTASE_TPPOLL, BIT(ring->index));
1407 
1408 	return NETDEV_TX_OK;
1409 
1410 err_dma_1:
1411 	ring->skbuff[entry] = NULL;
1412 	rtase_tx_clear_range(ring, ring->cur_idx + 1, frags);
1413 
1414 err_dma_0:
1415 	tp->stats.tx_dropped++;
1416 	dev_kfree_skb_any(skb);
1417 	return NETDEV_TX_OK;
1418 }
1419 
1420 static void rtase_set_rx_mode(struct net_device *dev)
1421 {
1422 	rtase_hw_set_rx_packet_filter(dev);
1423 }
1424 
1425 static void rtase_enable_eem_write(const struct rtase_private *tp)
1426 {
1427 	u8 val;
1428 
1429 	val = rtase_r8(tp, RTASE_EEM);
1430 	rtase_w8(tp, RTASE_EEM, val | RTASE_EEM_UNLOCK);
1431 }
1432 
1433 static void rtase_disable_eem_write(const struct rtase_private *tp)
1434 {
1435 	u8 val;
1436 
1437 	val = rtase_r8(tp, RTASE_EEM);
1438 	rtase_w8(tp, RTASE_EEM, val & ~RTASE_EEM_UNLOCK);
1439 }
1440 
1441 static void rtase_rar_set(const struct rtase_private *tp, const u8 *addr)
1442 {
1443 	u32 rar_low, rar_high;
1444 
1445 	rar_low = (u32)addr[0] | ((u32)addr[1] << 8) |
1446 		  ((u32)addr[2] << 16) | ((u32)addr[3] << 24);
1447 
1448 	rar_high = (u32)addr[4] | ((u32)addr[5] << 8);
1449 
1450 	rtase_enable_eem_write(tp);
1451 	rtase_w32(tp, RTASE_MAC0, rar_low);
1452 	rtase_w32(tp, RTASE_MAC4, rar_high);
1453 	rtase_disable_eem_write(tp);
1454 	rtase_w16(tp, RTASE_LBK_CTRL, RTASE_LBK_ATLD | RTASE_LBK_CLR);
1455 }
1456 
1457 static int rtase_set_mac_address(struct net_device *dev, void *p)
1458 {
1459 	struct rtase_private *tp = netdev_priv(dev);
1460 	int ret;
1461 
1462 	ret = eth_mac_addr(dev, p);
1463 	if (ret)
1464 		return ret;
1465 
1466 	rtase_rar_set(tp, dev->dev_addr);
1467 
1468 	return 0;
1469 }
1470 
1471 static int rtase_change_mtu(struct net_device *dev, int new_mtu)
1472 {
1473 	dev->mtu = new_mtu;
1474 
1475 	netdev_update_features(dev);
1476 
1477 	return 0;
1478 }
1479 
1480 static void rtase_wait_for_quiescence(const struct net_device *dev)
1481 {
1482 	struct rtase_private *tp = netdev_priv(dev);
1483 	struct rtase_int_vector *ivec;
1484 	u32 i;
1485 
1486 	for (i = 0; i < tp->int_nums; i++) {
1487 		ivec = &tp->int_vector[i];
1488 		synchronize_irq(ivec->irq);
1489 		/* wait for any pending NAPI task to complete */
1490 		napi_disable(&ivec->napi);
1491 	}
1492 
1493 	rtase_irq_dis_and_clear(tp);
1494 
1495 	for (i = 0; i < tp->int_nums; i++) {
1496 		ivec = &tp->int_vector[i];
1497 		napi_enable(&ivec->napi);
1498 	}
1499 }
1500 
1501 static void rtase_sw_reset(struct net_device *dev)
1502 {
1503 	struct rtase_private *tp = netdev_priv(dev);
1504 	int ret;
1505 
1506 	netif_stop_queue(dev);
1507 	netif_carrier_off(dev);
1508 	rtase_hw_reset(dev);
1509 
1510 	/* let's wait a bit while any (async) irq lands on */
1511 	rtase_wait_for_quiescence(dev);
1512 	rtase_tx_clear(tp);
1513 	rtase_rx_clear(tp);
1514 
1515 	ret = rtase_init_ring(dev);
1516 	if (ret) {
1517 		netdev_err(dev, "unable to init ring\n");
1518 		rtase_free_desc(tp);
1519 		return;
1520 	}
1521 
1522 	rtase_hw_config(dev);
1523 	/* always link, so start to transmit & receive */
1524 	rtase_hw_start(dev);
1525 
1526 	netif_carrier_on(dev);
1527 	netif_wake_queue(dev);
1528 }
1529 
1530 static void rtase_dump_tally_counter(const struct rtase_private *tp)
1531 {
1532 	dma_addr_t paddr = tp->tally_paddr;
1533 	u32 cmd = lower_32_bits(paddr);
1534 	u32 val;
1535 	int err;
1536 
1537 	rtase_w32(tp, RTASE_DTCCR4, upper_32_bits(paddr));
1538 	rtase_w32(tp, RTASE_DTCCR0, cmd);
1539 	rtase_w32(tp, RTASE_DTCCR0, cmd | RTASE_COUNTER_DUMP);
1540 
1541 	err = read_poll_timeout(rtase_r32, val, !(val & RTASE_COUNTER_DUMP),
1542 				10, 250, false, tp, RTASE_DTCCR0);
1543 
1544 	if (err == -ETIMEDOUT)
1545 		netdev_err(tp->dev, "error occurred in dump tally counter\n");
1546 }
1547 
1548 static void rtase_dump_state(const struct net_device *dev)
1549 {
1550 	const struct rtase_private *tp = netdev_priv(dev);
1551 	int max_reg_size = RTASE_PCI_REGS_SIZE;
1552 	const struct rtase_counters *counters;
1553 	const struct rtase_ring *ring;
1554 	u32 dword_rd;
1555 	int n = 0;
1556 
1557 	ring = &tp->tx_ring[0];
1558 	netdev_err(dev, "Tx descriptor info:\n");
1559 	netdev_err(dev, "Tx curIdx = 0x%x\n", ring->cur_idx);
1560 	netdev_err(dev, "Tx dirtyIdx = 0x%x\n", ring->dirty_idx);
1561 	netdev_err(dev, "Tx phyAddr = %pad\n", &ring->phy_addr);
1562 
1563 	ring = &tp->rx_ring[0];
1564 	netdev_err(dev, "Rx descriptor info:\n");
1565 	netdev_err(dev, "Rx curIdx = 0x%x\n", ring->cur_idx);
1566 	netdev_err(dev, "Rx dirtyIdx = 0x%x\n", ring->dirty_idx);
1567 	netdev_err(dev, "Rx phyAddr = %pad\n", &ring->phy_addr);
1568 
1569 	netdev_err(dev, "Device Registers:\n");
1570 	netdev_err(dev, "Chip Command = 0x%02x\n",
1571 		   rtase_r8(tp, RTASE_CHIP_CMD));
1572 	netdev_err(dev, "IMR = %08x\n", rtase_r32(tp, RTASE_IMR0));
1573 	netdev_err(dev, "ISR = %08x\n", rtase_r32(tp, RTASE_ISR0));
1574 	netdev_err(dev, "Boot Ctrl Reg(0xE004) = %04x\n",
1575 		   rtase_r16(tp, RTASE_BOOT_CTL));
1576 	netdev_err(dev, "EPHY ISR(0xE014) = %04x\n",
1577 		   rtase_r16(tp, RTASE_EPHY_ISR));
1578 	netdev_err(dev, "EPHY IMR(0xE016) = %04x\n",
1579 		   rtase_r16(tp, RTASE_EPHY_IMR));
1580 	netdev_err(dev, "CLKSW SET REG(0xE018) = %04x\n",
1581 		   rtase_r16(tp, RTASE_CLKSW_SET));
1582 
1583 	netdev_err(dev, "Dump PCI Registers:\n");
1584 
1585 	while (n < max_reg_size) {
1586 		if ((n % RTASE_DWORD_MOD) == 0)
1587 			netdev_err(tp->dev, "0x%03x:\n", n);
1588 
1589 		pci_read_config_dword(tp->pdev, n, &dword_rd);
1590 		netdev_err(tp->dev, "%08x\n", dword_rd);
1591 		n += 4;
1592 	}
1593 
1594 	netdev_err(dev, "Dump tally counter:\n");
1595 	counters = tp->tally_vaddr;
1596 	rtase_dump_tally_counter(tp);
1597 
1598 	netdev_err(dev, "tx_packets %lld\n",
1599 		   le64_to_cpu(counters->tx_packets));
1600 	netdev_err(dev, "rx_packets %lld\n",
1601 		   le64_to_cpu(counters->rx_packets));
1602 	netdev_err(dev, "tx_errors %lld\n",
1603 		   le64_to_cpu(counters->tx_errors));
1604 	netdev_err(dev, "rx_errors %d\n",
1605 		   le32_to_cpu(counters->rx_errors));
1606 	netdev_err(dev, "rx_missed %d\n",
1607 		   le16_to_cpu(counters->rx_missed));
1608 	netdev_err(dev, "align_errors %d\n",
1609 		   le16_to_cpu(counters->align_errors));
1610 	netdev_err(dev, "tx_one_collision %d\n",
1611 		   le32_to_cpu(counters->tx_one_collision));
1612 	netdev_err(dev, "tx_multi_collision %d\n",
1613 		   le32_to_cpu(counters->tx_multi_collision));
1614 	netdev_err(dev, "rx_unicast %lld\n",
1615 		   le64_to_cpu(counters->rx_unicast));
1616 	netdev_err(dev, "rx_broadcast %lld\n",
1617 		   le64_to_cpu(counters->rx_broadcast));
1618 	netdev_err(dev, "rx_multicast %d\n",
1619 		   le32_to_cpu(counters->rx_multicast));
1620 	netdev_err(dev, "tx_aborted %d\n",
1621 		   le16_to_cpu(counters->tx_aborted));
1622 	netdev_err(dev, "tx_underrun %d\n",
1623 		   le16_to_cpu(counters->tx_underrun));
1624 }
1625 
1626 static void rtase_tx_timeout(struct net_device *dev, unsigned int txqueue)
1627 {
1628 	rtase_dump_state(dev);
1629 	rtase_sw_reset(dev);
1630 }
1631 
1632 static void rtase_get_stats64(struct net_device *dev,
1633 			      struct rtnl_link_stats64 *stats)
1634 {
1635 	const struct rtase_private *tp = netdev_priv(dev);
1636 	const struct rtase_counters *counters;
1637 
1638 	counters = tp->tally_vaddr;
1639 
1640 	dev_fetch_sw_netstats(stats, dev->tstats);
1641 
1642 	/* fetch additional counter values missing in stats collected by driver
1643 	 * from tally counter
1644 	 */
1645 	rtase_dump_tally_counter(tp);
1646 	stats->rx_errors = tp->stats.rx_errors;
1647 	stats->tx_errors = le64_to_cpu(counters->tx_errors);
1648 	stats->rx_dropped = tp->stats.rx_dropped;
1649 	stats->tx_dropped = tp->stats.tx_dropped;
1650 	stats->multicast = tp->stats.multicast;
1651 	stats->rx_length_errors = tp->stats.rx_length_errors;
1652 }
1653 
1654 static netdev_features_t rtase_fix_features(struct net_device *dev,
1655 					    netdev_features_t features)
1656 {
1657 	netdev_features_t features_fix = features;
1658 
1659 	/* not support TSO for jumbo frames */
1660 	if (dev->mtu > ETH_DATA_LEN)
1661 		features_fix &= ~NETIF_F_ALL_TSO;
1662 
1663 	return features_fix;
1664 }
1665 
1666 static int rtase_set_features(struct net_device *dev,
1667 			      netdev_features_t features)
1668 {
1669 	netdev_features_t features_set = features;
1670 
1671 	features_set &= NETIF_F_RXALL | NETIF_F_RXCSUM |
1672 			NETIF_F_HW_VLAN_CTAG_RX;
1673 
1674 	if (features_set ^ dev->features)
1675 		rtase_hw_set_features(dev, features_set);
1676 
1677 	return 0;
1678 }
1679 
1680 static const struct net_device_ops rtase_netdev_ops = {
1681 	.ndo_open = rtase_open,
1682 	.ndo_stop = rtase_close,
1683 	.ndo_start_xmit = rtase_start_xmit,
1684 	.ndo_set_rx_mode = rtase_set_rx_mode,
1685 	.ndo_set_mac_address = rtase_set_mac_address,
1686 	.ndo_change_mtu = rtase_change_mtu,
1687 	.ndo_tx_timeout = rtase_tx_timeout,
1688 	.ndo_get_stats64 = rtase_get_stats64,
1689 	.ndo_fix_features = rtase_fix_features,
1690 	.ndo_set_features = rtase_set_features,
1691 };
1692 
1693 static void rtase_get_mac_address(struct net_device *dev)
1694 {
1695 	struct rtase_private *tp = netdev_priv(dev);
1696 	u8 mac_addr[ETH_ALEN] __aligned(2) = {};
1697 	u32 i;
1698 
1699 	for (i = 0; i < ETH_ALEN; i++)
1700 		mac_addr[i] = rtase_r8(tp, RTASE_MAC0 + i);
1701 
1702 	if (!is_valid_ether_addr(mac_addr)) {
1703 		eth_hw_addr_random(dev);
1704 		netdev_warn(dev, "Random ether addr %pM\n", dev->dev_addr);
1705 	} else {
1706 		eth_hw_addr_set(dev, mac_addr);
1707 		ether_addr_copy(dev->perm_addr, dev->dev_addr);
1708 	}
1709 
1710 	rtase_rar_set(tp, dev->dev_addr);
1711 }
1712 
1713 static int rtase_get_settings(struct net_device *dev,
1714 			      struct ethtool_link_ksettings *cmd)
1715 {
1716 	u32 supported = SUPPORTED_MII | SUPPORTED_Pause | SUPPORTED_Asym_Pause;
1717 	const struct rtase_private *tp = netdev_priv(dev);
1718 
1719 	ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.supported,
1720 						supported);
1721 
1722 	switch (tp->hw_ver) {
1723 	case RTASE_HW_VER_906X_7XA:
1724 	case RTASE_HW_VER_906X_7XC:
1725 		cmd->base.speed = SPEED_5000;
1726 		break;
1727 	case RTASE_HW_VER_907XD_V1:
1728 		cmd->base.speed = SPEED_10000;
1729 		break;
1730 	}
1731 
1732 	cmd->base.duplex = DUPLEX_FULL;
1733 	cmd->base.port = PORT_MII;
1734 	cmd->base.autoneg = AUTONEG_DISABLE;
1735 
1736 	return 0;
1737 }
1738 
1739 static void rtase_get_pauseparam(struct net_device *dev,
1740 				 struct ethtool_pauseparam *pause)
1741 {
1742 	const struct rtase_private *tp = netdev_priv(dev);
1743 	u16 value = rtase_r16(tp, RTASE_CPLUS_CMD);
1744 
1745 	pause->autoneg = AUTONEG_DISABLE;
1746 	pause->tx_pause = !!(value & RTASE_FORCE_TXFLOW_EN);
1747 	pause->rx_pause = !!(value & RTASE_FORCE_RXFLOW_EN);
1748 }
1749 
1750 static int rtase_set_pauseparam(struct net_device *dev,
1751 				struct ethtool_pauseparam *pause)
1752 {
1753 	const struct rtase_private *tp = netdev_priv(dev);
1754 	u16 value = rtase_r16(tp, RTASE_CPLUS_CMD);
1755 
1756 	if (pause->autoneg)
1757 		return -EOPNOTSUPP;
1758 
1759 	value &= ~(RTASE_FORCE_TXFLOW_EN | RTASE_FORCE_RXFLOW_EN);
1760 
1761 	if (pause->tx_pause)
1762 		value |= RTASE_FORCE_TXFLOW_EN;
1763 
1764 	if (pause->rx_pause)
1765 		value |= RTASE_FORCE_RXFLOW_EN;
1766 
1767 	rtase_w16(tp, RTASE_CPLUS_CMD, value);
1768 	return 0;
1769 }
1770 
1771 static void rtase_get_eth_mac_stats(struct net_device *dev,
1772 				    struct ethtool_eth_mac_stats *stats)
1773 {
1774 	struct rtase_private *tp = netdev_priv(dev);
1775 	const struct rtase_counters *counters;
1776 
1777 	counters = tp->tally_vaddr;
1778 
1779 	rtase_dump_tally_counter(tp);
1780 
1781 	stats->FramesTransmittedOK = le64_to_cpu(counters->tx_packets);
1782 	stats->FramesReceivedOK = le64_to_cpu(counters->rx_packets);
1783 	stats->FramesLostDueToIntMACXmitError =
1784 		le64_to_cpu(counters->tx_errors);
1785 	stats->BroadcastFramesReceivedOK = le64_to_cpu(counters->rx_broadcast);
1786 }
1787 
1788 static const struct ethtool_ops rtase_ethtool_ops = {
1789 	.get_link = ethtool_op_get_link,
1790 	.get_link_ksettings = rtase_get_settings,
1791 	.get_pauseparam = rtase_get_pauseparam,
1792 	.set_pauseparam = rtase_set_pauseparam,
1793 	.get_eth_mac_stats = rtase_get_eth_mac_stats,
1794 	.get_ts_info = ethtool_op_get_ts_info,
1795 };
1796 
1797 static void rtase_init_netdev_ops(struct net_device *dev)
1798 {
1799 	dev->netdev_ops = &rtase_netdev_ops;
1800 	dev->ethtool_ops = &rtase_ethtool_ops;
1801 }
1802 
1803 static void rtase_reset_interrupt(struct pci_dev *pdev,
1804 				  const struct rtase_private *tp)
1805 {
1806 	if (tp->sw_flag & RTASE_SWF_MSIX_ENABLED)
1807 		pci_disable_msix(pdev);
1808 	else
1809 		pci_disable_msi(pdev);
1810 }
1811 
1812 static int rtase_alloc_msix(struct pci_dev *pdev, struct rtase_private *tp)
1813 {
1814 	int ret, irq;
1815 	u16 i;
1816 
1817 	memset(tp->msix_entry, 0x0, RTASE_NUM_MSIX *
1818 	       sizeof(struct msix_entry));
1819 
1820 	for (i = 0; i < RTASE_NUM_MSIX; i++)
1821 		tp->msix_entry[i].entry = i;
1822 
1823 	ret = pci_enable_msix_exact(pdev, tp->msix_entry, tp->int_nums);
1824 
1825 	if (ret)
1826 		return ret;
1827 
1828 	for (i = 0; i < tp->int_nums; i++) {
1829 		irq = pci_irq_vector(pdev, i);
1830 		if (!irq) {
1831 			pci_disable_msix(pdev);
1832 			return irq;
1833 		}
1834 
1835 		tp->int_vector[i].irq = irq;
1836 	}
1837 
1838 	return 0;
1839 }
1840 
1841 static int rtase_alloc_interrupt(struct pci_dev *pdev,
1842 				 struct rtase_private *tp)
1843 {
1844 	int ret;
1845 
1846 	ret = rtase_alloc_msix(pdev, tp);
1847 	if (ret) {
1848 		ret = pci_enable_msi(pdev);
1849 		if (ret) {
1850 			dev_err(&pdev->dev,
1851 				"unable to alloc interrupt.(MSI)\n");
1852 			return ret;
1853 		}
1854 
1855 		tp->sw_flag |= RTASE_SWF_MSI_ENABLED;
1856 	} else {
1857 		tp->sw_flag |= RTASE_SWF_MSIX_ENABLED;
1858 	}
1859 
1860 	return 0;
1861 }
1862 
1863 static void rtase_init_hardware(const struct rtase_private *tp)
1864 {
1865 	u16 i;
1866 
1867 	for (i = 0; i < RTASE_VLAN_FILTER_ENTRY_NUM; i++)
1868 		rtase_w32(tp, RTASE_VLAN_ENTRY_0 + i * 4, 0);
1869 }
1870 
1871 static void rtase_init_int_vector(struct rtase_private *tp)
1872 {
1873 	u16 i;
1874 
1875 	/* interrupt vector 0 */
1876 	tp->int_vector[0].tp = tp;
1877 	tp->int_vector[0].index = 0;
1878 	tp->int_vector[0].imr_addr = RTASE_IMR0;
1879 	tp->int_vector[0].isr_addr = RTASE_ISR0;
1880 	tp->int_vector[0].imr = RTASE_ROK | RTASE_RDU | RTASE_TOK |
1881 				RTASE_TOK4 | RTASE_TOK5 | RTASE_TOK6 |
1882 				RTASE_TOK7;
1883 	tp->int_vector[0].poll = rtase_poll;
1884 
1885 	memset(tp->int_vector[0].name, 0x0, sizeof(tp->int_vector[0].name));
1886 	INIT_LIST_HEAD(&tp->int_vector[0].ring_list);
1887 
1888 	netif_napi_add(tp->dev, &tp->int_vector[0].napi,
1889 		       tp->int_vector[0].poll);
1890 
1891 	/* interrupt vector 1 ~ 3 */
1892 	for (i = 1; i < tp->int_nums; i++) {
1893 		tp->int_vector[i].tp = tp;
1894 		tp->int_vector[i].index = i;
1895 		tp->int_vector[i].imr_addr = RTASE_IMR1 + (i - 1) * 4;
1896 		tp->int_vector[i].isr_addr = RTASE_ISR1 + (i - 1) * 4;
1897 		tp->int_vector[i].imr = RTASE_Q_ROK | RTASE_Q_RDU |
1898 					RTASE_Q_TOK;
1899 		tp->int_vector[i].poll = rtase_poll;
1900 
1901 		memset(tp->int_vector[i].name, 0x0,
1902 		       sizeof(tp->int_vector[0].name));
1903 		INIT_LIST_HEAD(&tp->int_vector[i].ring_list);
1904 
1905 		netif_napi_add(tp->dev, &tp->int_vector[i].napi,
1906 			       tp->int_vector[i].poll);
1907 	}
1908 }
1909 
1910 static u16 rtase_calc_time_mitigation(u32 time_us)
1911 {
1912 	u8 msb, time_count, time_unit;
1913 	u16 int_miti;
1914 
1915 	time_us = min_t(int, time_us, RTASE_MITI_MAX_TIME);
1916 
1917 	msb = fls(time_us);
1918 	if (msb >= RTASE_MITI_COUNT_BIT_NUM) {
1919 		time_unit = msb - RTASE_MITI_COUNT_BIT_NUM;
1920 		time_count = time_us >> (msb - RTASE_MITI_COUNT_BIT_NUM);
1921 	} else {
1922 		time_unit = 0;
1923 		time_count = time_us;
1924 	}
1925 
1926 	int_miti = u16_encode_bits(time_count, RTASE_MITI_TIME_COUNT_MASK) |
1927 		   u16_encode_bits(time_unit, RTASE_MITI_TIME_UNIT_MASK);
1928 
1929 	return int_miti;
1930 }
1931 
1932 static u16 rtase_calc_packet_num_mitigation(u16 pkt_num)
1933 {
1934 	u8 msb, pkt_num_count, pkt_num_unit;
1935 	u16 int_miti;
1936 
1937 	pkt_num = min_t(int, pkt_num, RTASE_MITI_MAX_PKT_NUM);
1938 
1939 	if (pkt_num > 60) {
1940 		pkt_num_unit = RTASE_MITI_MAX_PKT_NUM_IDX;
1941 		pkt_num_count = pkt_num / RTASE_MITI_MAX_PKT_NUM_UNIT;
1942 	} else {
1943 		msb = fls(pkt_num);
1944 		if (msb >= RTASE_MITI_COUNT_BIT_NUM) {
1945 			pkt_num_unit = msb - RTASE_MITI_COUNT_BIT_NUM;
1946 			pkt_num_count = pkt_num >> (msb -
1947 						    RTASE_MITI_COUNT_BIT_NUM);
1948 		} else {
1949 			pkt_num_unit = 0;
1950 			pkt_num_count = pkt_num;
1951 		}
1952 	}
1953 
1954 	int_miti = u16_encode_bits(pkt_num_count,
1955 				   RTASE_MITI_PKT_NUM_COUNT_MASK) |
1956 		   u16_encode_bits(pkt_num_unit,
1957 				   RTASE_MITI_PKT_NUM_UNIT_MASK);
1958 
1959 	return int_miti;
1960 }
1961 
1962 static void rtase_init_software_variable(struct pci_dev *pdev,
1963 					 struct rtase_private *tp)
1964 {
1965 	u16 int_miti;
1966 
1967 	tp->tx_queue_ctrl = RTASE_TXQ_CTRL;
1968 	tp->func_tx_queue_num = RTASE_FUNC_TXQ_NUM;
1969 	tp->func_rx_queue_num = RTASE_FUNC_RXQ_NUM;
1970 	tp->int_nums = RTASE_INTERRUPT_NUM;
1971 
1972 	int_miti = rtase_calc_time_mitigation(RTASE_MITI_DEFAULT_TIME) |
1973 		   rtase_calc_packet_num_mitigation(RTASE_MITI_DEFAULT_PKT_NUM);
1974 	tp->tx_int_mit = int_miti;
1975 	tp->rx_int_mit = int_miti;
1976 
1977 	tp->sw_flag = 0;
1978 
1979 	rtase_init_int_vector(tp);
1980 
1981 	/* MTU range: 60 - hw-specific max */
1982 	tp->dev->min_mtu = ETH_ZLEN;
1983 	tp->dev->max_mtu = RTASE_MAX_JUMBO_SIZE;
1984 }
1985 
1986 static int rtase_check_mac_version_valid(struct rtase_private *tp)
1987 {
1988 	int ret = -ENODEV;
1989 
1990 	tp->hw_ver = rtase_r32(tp, RTASE_TX_CONFIG_0) & RTASE_HW_VER_MASK;
1991 
1992 	switch (tp->hw_ver) {
1993 	case RTASE_HW_VER_906X_7XA:
1994 	case RTASE_HW_VER_906X_7XC:
1995 	case RTASE_HW_VER_907XD_V1:
1996 		ret = 0;
1997 		break;
1998 	}
1999 
2000 	return ret;
2001 }
2002 
2003 static int rtase_init_board(struct pci_dev *pdev, struct net_device **dev_out,
2004 			    void __iomem **ioaddr_out)
2005 {
2006 	struct net_device *dev;
2007 	void __iomem *ioaddr;
2008 	int ret = -ENOMEM;
2009 
2010 	/* dev zeroed in alloc_etherdev */
2011 	dev = alloc_etherdev_mq(sizeof(struct rtase_private),
2012 				RTASE_FUNC_TXQ_NUM);
2013 	if (!dev)
2014 		goto err_out;
2015 
2016 	SET_NETDEV_DEV(dev, &pdev->dev);
2017 
2018 	ret = pci_enable_device(pdev);
2019 	if (ret < 0)
2020 		goto err_out_free_dev;
2021 
2022 	/* make sure PCI base addr 1 is MMIO */
2023 	if (!(pci_resource_flags(pdev, 2) & IORESOURCE_MEM)) {
2024 		ret = -ENODEV;
2025 		goto err_out_disable;
2026 	}
2027 
2028 	/* check for weird/broken PCI region reporting */
2029 	if (pci_resource_len(pdev, 2) < RTASE_REGS_SIZE) {
2030 		ret = -ENODEV;
2031 		goto err_out_disable;
2032 	}
2033 
2034 	ret = pci_request_regions(pdev, KBUILD_MODNAME);
2035 	if (ret < 0)
2036 		goto err_out_disable;
2037 
2038 	ret = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
2039 	if (ret) {
2040 		dev_err(&pdev->dev, "no usable dma addressing method\n");
2041 		goto err_out_free_res;
2042 	}
2043 
2044 	pci_set_master(pdev);
2045 
2046 	/* ioremap MMIO region */
2047 	ioaddr = ioremap(pci_resource_start(pdev, 2),
2048 			 pci_resource_len(pdev, 2));
2049 	if (!ioaddr) {
2050 		ret = -EIO;
2051 		goto err_out_free_res;
2052 	}
2053 
2054 	*ioaddr_out = ioaddr;
2055 	*dev_out = dev;
2056 
2057 	return ret;
2058 
2059 err_out_free_res:
2060 	pci_release_regions(pdev);
2061 
2062 err_out_disable:
2063 	pci_disable_device(pdev);
2064 
2065 err_out_free_dev:
2066 	free_netdev(dev);
2067 
2068 err_out:
2069 	*ioaddr_out = NULL;
2070 	*dev_out = NULL;
2071 
2072 	return ret;
2073 }
2074 
2075 static void rtase_release_board(struct pci_dev *pdev, struct net_device *dev,
2076 				void __iomem *ioaddr)
2077 {
2078 	const struct rtase_private *tp = netdev_priv(dev);
2079 
2080 	rtase_rar_set(tp, tp->dev->perm_addr);
2081 	iounmap(ioaddr);
2082 
2083 	if (tp->sw_flag & RTASE_SWF_MSIX_ENABLED)
2084 		pci_disable_msix(pdev);
2085 	else
2086 		pci_disable_msi(pdev);
2087 
2088 	pci_release_regions(pdev);
2089 	pci_disable_device(pdev);
2090 	free_netdev(dev);
2091 }
2092 
2093 static int rtase_init_one(struct pci_dev *pdev,
2094 			  const struct pci_device_id *ent)
2095 {
2096 	struct net_device *dev = NULL;
2097 	struct rtase_int_vector *ivec;
2098 	void __iomem *ioaddr = NULL;
2099 	struct rtase_private *tp;
2100 	int ret, i;
2101 
2102 	if (!pdev->is_physfn && pdev->is_virtfn) {
2103 		dev_err(&pdev->dev,
2104 			"This module does not support a virtual function.");
2105 		return -EINVAL;
2106 	}
2107 
2108 	dev_dbg(&pdev->dev, "Automotive Switch Ethernet driver loaded\n");
2109 
2110 	ret = rtase_init_board(pdev, &dev, &ioaddr);
2111 	if (ret != 0)
2112 		return ret;
2113 
2114 	tp = netdev_priv(dev);
2115 	tp->mmio_addr = ioaddr;
2116 	tp->dev = dev;
2117 	tp->pdev = pdev;
2118 
2119 	/* identify chip attached to board */
2120 	ret = rtase_check_mac_version_valid(tp);
2121 	if (ret != 0) {
2122 		dev_err(&pdev->dev,
2123 			"unknown chip version: 0x%08x, contact rtase maintainers (see MAINTAINERS file)\n",
2124 			tp->hw_ver);
2125 		goto err_out_release_board;
2126 	}
2127 
2128 	rtase_init_software_variable(pdev, tp);
2129 	rtase_init_hardware(tp);
2130 
2131 	ret = rtase_alloc_interrupt(pdev, tp);
2132 	if (ret < 0) {
2133 		dev_err(&pdev->dev, "unable to alloc MSIX/MSI\n");
2134 		goto err_out_del_napi;
2135 	}
2136 
2137 	rtase_init_netdev_ops(dev);
2138 
2139 	dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS;
2140 
2141 	dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX |
2142 			 NETIF_F_IP_CSUM | NETIF_F_HIGHDMA |
2143 			 NETIF_F_RXCSUM | NETIF_F_SG |
2144 			 NETIF_F_TSO | NETIF_F_IPV6_CSUM |
2145 			 NETIF_F_TSO6;
2146 
2147 	dev->hw_features = NETIF_F_SG | NETIF_F_IP_CSUM |
2148 			   NETIF_F_TSO | NETIF_F_RXCSUM |
2149 			   NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX |
2150 			   NETIF_F_RXALL | NETIF_F_RXFCS |
2151 			   NETIF_F_IPV6_CSUM | NETIF_F_TSO6;
2152 
2153 	dev->vlan_features = NETIF_F_SG | NETIF_F_IP_CSUM | NETIF_F_TSO |
2154 			     NETIF_F_HIGHDMA;
2155 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2156 	netif_set_tso_max_size(dev, RTASE_LSO_64K);
2157 	netif_set_tso_max_segs(dev, RTASE_NIC_MAX_PHYS_BUF_COUNT_LSO2);
2158 
2159 	rtase_get_mac_address(dev);
2160 
2161 	tp->tally_vaddr = dma_alloc_coherent(&pdev->dev,
2162 					     sizeof(*tp->tally_vaddr),
2163 					     &tp->tally_paddr,
2164 					     GFP_KERNEL);
2165 	if (!tp->tally_vaddr) {
2166 		ret = -ENOMEM;
2167 		goto err_out_free_dma;
2168 	}
2169 
2170 	rtase_tally_counter_clear(tp);
2171 
2172 	pci_set_drvdata(pdev, dev);
2173 
2174 	netif_carrier_off(dev);
2175 
2176 	ret = register_netdev(dev);
2177 	if (ret != 0)
2178 		goto err_out_free_dma;
2179 
2180 	netdev_dbg(dev, "%pM, IRQ %d\n", dev->dev_addr, dev->irq);
2181 
2182 	return 0;
2183 
2184 err_out_free_dma:
2185 	if (tp->tally_vaddr) {
2186 		dma_free_coherent(&pdev->dev,
2187 				  sizeof(*tp->tally_vaddr),
2188 				  tp->tally_vaddr,
2189 				  tp->tally_paddr);
2190 
2191 		tp->tally_vaddr = NULL;
2192 	}
2193 
2194 err_out_del_napi:
2195 	for (i = 0; i < tp->int_nums; i++) {
2196 		ivec = &tp->int_vector[i];
2197 		netif_napi_del(&ivec->napi);
2198 	}
2199 
2200 err_out_release_board:
2201 	rtase_release_board(pdev, dev, ioaddr);
2202 
2203 	return ret;
2204 }
2205 
2206 static void rtase_remove_one(struct pci_dev *pdev)
2207 {
2208 	struct net_device *dev = pci_get_drvdata(pdev);
2209 	struct rtase_private *tp = netdev_priv(dev);
2210 	struct rtase_int_vector *ivec;
2211 	u32 i;
2212 
2213 	unregister_netdev(dev);
2214 
2215 	for (i = 0; i < tp->int_nums; i++) {
2216 		ivec = &tp->int_vector[i];
2217 		netif_napi_del(&ivec->napi);
2218 	}
2219 
2220 	rtase_reset_interrupt(pdev, tp);
2221 	if (tp->tally_vaddr) {
2222 		dma_free_coherent(&pdev->dev,
2223 				  sizeof(*tp->tally_vaddr),
2224 				  tp->tally_vaddr,
2225 				  tp->tally_paddr);
2226 		tp->tally_vaddr = NULL;
2227 	}
2228 
2229 	rtase_release_board(pdev, dev, tp->mmio_addr);
2230 	pci_set_drvdata(pdev, NULL);
2231 }
2232 
2233 static void rtase_shutdown(struct pci_dev *pdev)
2234 {
2235 	struct net_device *dev = pci_get_drvdata(pdev);
2236 	const struct rtase_private *tp;
2237 
2238 	tp = netdev_priv(dev);
2239 
2240 	if (netif_running(dev))
2241 		rtase_close(dev);
2242 
2243 	rtase_reset_interrupt(pdev, tp);
2244 }
2245 
2246 static int rtase_suspend(struct device *device)
2247 {
2248 	struct net_device *dev = dev_get_drvdata(device);
2249 
2250 	if (netif_running(dev)) {
2251 		netif_device_detach(dev);
2252 		rtase_hw_reset(dev);
2253 	}
2254 
2255 	return 0;
2256 }
2257 
2258 static int rtase_resume(struct device *device)
2259 {
2260 	struct net_device *dev = dev_get_drvdata(device);
2261 	struct rtase_private *tp = netdev_priv(dev);
2262 	int ret;
2263 
2264 	/* restore last modified mac address */
2265 	rtase_rar_set(tp, dev->dev_addr);
2266 
2267 	if (!netif_running(dev))
2268 		goto out;
2269 
2270 	rtase_wait_for_quiescence(dev);
2271 
2272 	rtase_tx_clear(tp);
2273 	rtase_rx_clear(tp);
2274 
2275 	ret = rtase_init_ring(dev);
2276 	if (ret) {
2277 		netdev_err(dev, "unable to init ring\n");
2278 		rtase_free_desc(tp);
2279 		return -ENOMEM;
2280 	}
2281 
2282 	rtase_hw_config(dev);
2283 	/* always link, so start to transmit & receive */
2284 	rtase_hw_start(dev);
2285 
2286 	netif_device_attach(dev);
2287 out:
2288 
2289 	return 0;
2290 }
2291 
2292 static const struct dev_pm_ops rtase_pm_ops = {
2293 	SYSTEM_SLEEP_PM_OPS(rtase_suspend, rtase_resume)
2294 };
2295 
2296 static struct pci_driver rtase_pci_driver = {
2297 	.name = KBUILD_MODNAME,
2298 	.id_table = rtase_pci_tbl,
2299 	.probe = rtase_init_one,
2300 	.remove = rtase_remove_one,
2301 	.shutdown = rtase_shutdown,
2302 	.driver.pm = pm_ptr(&rtase_pm_ops),
2303 };
2304 
2305 module_pci_driver(rtase_pci_driver);
2306