xref: /openbmc/linux/drivers/net/ethernet/ibm/ibmvnic.c (revision 7af6fbdd)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /**************************************************************************/
3 /*                                                                        */
4 /*  IBM System i and System p Virtual NIC Device Driver                   */
5 /*  Copyright (C) 2014 IBM Corp.                                          */
6 /*  Santiago Leon (santi_leon@yahoo.com)                                  */
7 /*  Thomas Falcon (tlfalcon@linux.vnet.ibm.com)                           */
8 /*  John Allen (jallen@linux.vnet.ibm.com)                                */
9 /*                                                                        */
10 /*                                                                        */
11 /* This module contains the implementation of a virtual ethernet device   */
12 /* for use with IBM i/p Series LPAR Linux. It utilizes the logical LAN    */
13 /* option of the RS/6000 Platform Architecture to interface with virtual  */
14 /* ethernet NICs that are presented to the partition by the hypervisor.   */
15 /*									   */
16 /* Messages are passed between the VNIC driver and the VNIC server using  */
17 /* Command/Response Queues (CRQs) and sub CRQs (sCRQs). CRQs are used to  */
18 /* issue and receive commands that initiate communication with the server */
19 /* on driver initialization. Sub CRQs (sCRQs) are similar to CRQs, but    */
20 /* are used by the driver to notify the server that a packet is           */
21 /* ready for transmission or that a buffer has been added to receive a    */
22 /* packet. Subsequently, sCRQs are used by the server to notify the       */
23 /* driver that a packet transmission has been completed or that a packet  */
24 /* has been received and placed in a waiting buffer.                      */
25 /*                                                                        */
26 /* In lieu of a more conventional "on-the-fly" DMA mapping strategy in    */
27 /* which skbs are DMA mapped and immediately unmapped when the transmit   */
28 /* or receive has been completed, the VNIC driver is required to use      */
29 /* "long term mapping". This entails that large, continuous DMA mapped    */
30 /* buffers are allocated on driver initialization and these buffers are   */
31 /* then continuously reused to pass skbs to and from the VNIC server.     */
32 /*                                                                        */
33 /**************************************************************************/
34 
35 #include <linux/module.h>
36 #include <linux/moduleparam.h>
37 #include <linux/types.h>
38 #include <linux/errno.h>
39 #include <linux/completion.h>
40 #include <linux/ioport.h>
41 #include <linux/dma-mapping.h>
42 #include <linux/kernel.h>
43 #include <linux/netdevice.h>
44 #include <linux/etherdevice.h>
45 #include <linux/skbuff.h>
46 #include <linux/init.h>
47 #include <linux/delay.h>
48 #include <linux/mm.h>
49 #include <linux/ethtool.h>
50 #include <linux/proc_fs.h>
51 #include <linux/if_arp.h>
52 #include <linux/in.h>
53 #include <linux/ip.h>
54 #include <linux/ipv6.h>
55 #include <linux/irq.h>
56 #include <linux/kthread.h>
57 #include <linux/seq_file.h>
58 #include <linux/interrupt.h>
59 #include <net/net_namespace.h>
60 #include <asm/hvcall.h>
61 #include <linux/atomic.h>
62 #include <asm/vio.h>
63 #include <asm/iommu.h>
64 #include <linux/uaccess.h>
65 #include <asm/firmware.h>
66 #include <linux/workqueue.h>
67 #include <linux/if_vlan.h>
68 #include <linux/utsname.h>
69 
70 #include "ibmvnic.h"
71 
72 static const char ibmvnic_driver_name[] = "ibmvnic";
73 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
74 
75 MODULE_AUTHOR("Santiago Leon");
76 MODULE_DESCRIPTION("IBM System i/p Virtual NIC Driver");
77 MODULE_LICENSE("GPL");
78 MODULE_VERSION(IBMVNIC_DRIVER_VERSION);
79 
80 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
81 static int ibmvnic_remove(struct vio_dev *);
82 static void release_sub_crqs(struct ibmvnic_adapter *, bool);
83 static int ibmvnic_reset_crq(struct ibmvnic_adapter *);
84 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *);
85 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *);
86 static int ibmvnic_send_crq(struct ibmvnic_adapter *, union ibmvnic_crq *);
87 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
88 		       union sub_crq *sub_crq);
89 static int send_subcrq_indirect(struct ibmvnic_adapter *, u64, u64, u64);
90 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
91 static int enable_scrq_irq(struct ibmvnic_adapter *,
92 			   struct ibmvnic_sub_crq_queue *);
93 static int disable_scrq_irq(struct ibmvnic_adapter *,
94 			    struct ibmvnic_sub_crq_queue *);
95 static int pending_scrq(struct ibmvnic_adapter *,
96 			struct ibmvnic_sub_crq_queue *);
97 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
98 					struct ibmvnic_sub_crq_queue *);
99 static int ibmvnic_poll(struct napi_struct *napi, int data);
100 static void send_map_query(struct ibmvnic_adapter *adapter);
101 static int send_request_map(struct ibmvnic_adapter *, dma_addr_t, __be32, u8);
102 static int send_request_unmap(struct ibmvnic_adapter *, u8);
103 static int send_login(struct ibmvnic_adapter *adapter);
104 static void send_cap_queries(struct ibmvnic_adapter *adapter);
105 static int init_sub_crqs(struct ibmvnic_adapter *);
106 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
107 static int ibmvnic_init(struct ibmvnic_adapter *);
108 static int ibmvnic_reset_init(struct ibmvnic_adapter *);
109 static void release_crq_queue(struct ibmvnic_adapter *);
110 static int __ibmvnic_set_mac(struct net_device *, u8 *);
111 static int init_crq_queue(struct ibmvnic_adapter *adapter);
112 static int send_query_phys_parms(struct ibmvnic_adapter *adapter);
113 
114 struct ibmvnic_stat {
115 	char name[ETH_GSTRING_LEN];
116 	int offset;
117 };
118 
119 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
120 			     offsetof(struct ibmvnic_statistics, stat))
121 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
122 
123 static const struct ibmvnic_stat ibmvnic_stats[] = {
124 	{"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
125 	{"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
126 	{"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
127 	{"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
128 	{"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
129 	{"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
130 	{"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
131 	{"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
132 	{"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
133 	{"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
134 	{"align_errors", IBMVNIC_STAT_OFF(align_errors)},
135 	{"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
136 	{"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
137 	{"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
138 	{"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
139 	{"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
140 	{"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
141 	{"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
142 	{"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
143 	{"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
144 	{"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
145 	{"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
146 };
147 
148 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
149 			  unsigned long length, unsigned long *number,
150 			  unsigned long *irq)
151 {
152 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
153 	long rc;
154 
155 	rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
156 	*number = retbuf[0];
157 	*irq = retbuf[1];
158 
159 	return rc;
160 }
161 
162 /**
163  * ibmvnic_wait_for_completion - Check device state and wait for completion
164  * @adapter: private device data
165  * @comp_done: completion structure to wait for
166  * @timeout: time to wait in milliseconds
167  *
168  * Wait for a completion signal or until the timeout limit is reached
169  * while checking that the device is still active.
170  */
171 static int ibmvnic_wait_for_completion(struct ibmvnic_adapter *adapter,
172 				       struct completion *comp_done,
173 				       unsigned long timeout)
174 {
175 	struct net_device *netdev;
176 	unsigned long div_timeout;
177 	u8 retry;
178 
179 	netdev = adapter->netdev;
180 	retry = 5;
181 	div_timeout = msecs_to_jiffies(timeout / retry);
182 	while (true) {
183 		if (!adapter->crq.active) {
184 			netdev_err(netdev, "Device down!\n");
185 			return -ENODEV;
186 		}
187 		if (!retry--)
188 			break;
189 		if (wait_for_completion_timeout(comp_done, div_timeout))
190 			return 0;
191 	}
192 	netdev_err(netdev, "Operation timed out.\n");
193 	return -ETIMEDOUT;
194 }
195 
196 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
197 				struct ibmvnic_long_term_buff *ltb, int size)
198 {
199 	struct device *dev = &adapter->vdev->dev;
200 	int rc;
201 
202 	ltb->size = size;
203 	ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
204 				       GFP_KERNEL);
205 
206 	if (!ltb->buff) {
207 		dev_err(dev, "Couldn't alloc long term buffer\n");
208 		return -ENOMEM;
209 	}
210 	ltb->map_id = adapter->map_id;
211 	adapter->map_id++;
212 
213 	mutex_lock(&adapter->fw_lock);
214 	adapter->fw_done_rc = 0;
215 	reinit_completion(&adapter->fw_done);
216 	rc = send_request_map(adapter, ltb->addr,
217 			      ltb->size, ltb->map_id);
218 	if (rc) {
219 		dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
220 		mutex_unlock(&adapter->fw_lock);
221 		return rc;
222 	}
223 
224 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
225 	if (rc) {
226 		dev_err(dev,
227 			"Long term map request aborted or timed out,rc = %d\n",
228 			rc);
229 		dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
230 		mutex_unlock(&adapter->fw_lock);
231 		return rc;
232 	}
233 
234 	if (adapter->fw_done_rc) {
235 		dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
236 			adapter->fw_done_rc);
237 		dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
238 		mutex_unlock(&adapter->fw_lock);
239 		return -1;
240 	}
241 	mutex_unlock(&adapter->fw_lock);
242 	return 0;
243 }
244 
245 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
246 				struct ibmvnic_long_term_buff *ltb)
247 {
248 	struct device *dev = &adapter->vdev->dev;
249 
250 	if (!ltb->buff)
251 		return;
252 
253 	if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
254 	    adapter->reset_reason != VNIC_RESET_MOBILITY)
255 		send_request_unmap(adapter, ltb->map_id);
256 	dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
257 }
258 
259 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
260 				struct ibmvnic_long_term_buff *ltb)
261 {
262 	struct device *dev = &adapter->vdev->dev;
263 	int rc;
264 
265 	memset(ltb->buff, 0, ltb->size);
266 
267 	mutex_lock(&adapter->fw_lock);
268 	adapter->fw_done_rc = 0;
269 
270 	reinit_completion(&adapter->fw_done);
271 	rc = send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
272 	if (rc) {
273 		mutex_unlock(&adapter->fw_lock);
274 		return rc;
275 	}
276 
277 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
278 	if (rc) {
279 		dev_info(dev,
280 			 "Reset failed, long term map request timed out or aborted\n");
281 		mutex_unlock(&adapter->fw_lock);
282 		return rc;
283 	}
284 
285 	if (adapter->fw_done_rc) {
286 		dev_info(dev,
287 			 "Reset failed, attempting to free and reallocate buffer\n");
288 		free_long_term_buff(adapter, ltb);
289 		mutex_unlock(&adapter->fw_lock);
290 		return alloc_long_term_buff(adapter, ltb, ltb->size);
291 	}
292 	mutex_unlock(&adapter->fw_lock);
293 	return 0;
294 }
295 
296 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
297 {
298 	int i;
299 
300 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
301 	     i++)
302 		adapter->rx_pool[i].active = 0;
303 }
304 
305 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
306 			      struct ibmvnic_rx_pool *pool)
307 {
308 	int count = pool->size - atomic_read(&pool->available);
309 	struct device *dev = &adapter->vdev->dev;
310 	int buffers_added = 0;
311 	unsigned long lpar_rc;
312 	union sub_crq sub_crq;
313 	struct sk_buff *skb;
314 	unsigned int offset;
315 	dma_addr_t dma_addr;
316 	unsigned char *dst;
317 	u64 *handle_array;
318 	int shift = 0;
319 	int index;
320 	int i;
321 
322 	if (!pool->active)
323 		return;
324 
325 	handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
326 				      be32_to_cpu(adapter->login_rsp_buf->
327 				      off_rxadd_subcrqs));
328 
329 	for (i = 0; i < count; ++i) {
330 		skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
331 		if (!skb) {
332 			dev_err(dev, "Couldn't replenish rx buff\n");
333 			adapter->replenish_no_mem++;
334 			break;
335 		}
336 
337 		index = pool->free_map[pool->next_free];
338 
339 		if (pool->rx_buff[index].skb)
340 			dev_err(dev, "Inconsistent free_map!\n");
341 
342 		/* Copy the skb to the long term mapped DMA buffer */
343 		offset = index * pool->buff_size;
344 		dst = pool->long_term_buff.buff + offset;
345 		memset(dst, 0, pool->buff_size);
346 		dma_addr = pool->long_term_buff.addr + offset;
347 		pool->rx_buff[index].data = dst;
348 
349 		pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
350 		pool->rx_buff[index].dma = dma_addr;
351 		pool->rx_buff[index].skb = skb;
352 		pool->rx_buff[index].pool_index = pool->index;
353 		pool->rx_buff[index].size = pool->buff_size;
354 
355 		memset(&sub_crq, 0, sizeof(sub_crq));
356 		sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
357 		sub_crq.rx_add.correlator =
358 		    cpu_to_be64((u64)&pool->rx_buff[index]);
359 		sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
360 		sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
361 
362 		/* The length field of the sCRQ is defined to be 24 bits so the
363 		 * buffer size needs to be left shifted by a byte before it is
364 		 * converted to big endian to prevent the last byte from being
365 		 * truncated.
366 		 */
367 #ifdef __LITTLE_ENDIAN__
368 		shift = 8;
369 #endif
370 		sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
371 
372 		lpar_rc = send_subcrq(adapter, handle_array[pool->index],
373 				      &sub_crq);
374 		if (lpar_rc != H_SUCCESS)
375 			goto failure;
376 
377 		buffers_added++;
378 		adapter->replenish_add_buff_success++;
379 		pool->next_free = (pool->next_free + 1) % pool->size;
380 	}
381 	atomic_add(buffers_added, &pool->available);
382 	return;
383 
384 failure:
385 	if (lpar_rc != H_PARAMETER && lpar_rc != H_CLOSED)
386 		dev_err_ratelimited(dev, "rx: replenish packet buffer failed\n");
387 	pool->free_map[pool->next_free] = index;
388 	pool->rx_buff[index].skb = NULL;
389 
390 	dev_kfree_skb_any(skb);
391 	adapter->replenish_add_buff_failure++;
392 	atomic_add(buffers_added, &pool->available);
393 
394 	if (lpar_rc == H_CLOSED || adapter->failover_pending) {
395 		/* Disable buffer pool replenishment and report carrier off if
396 		 * queue is closed or pending failover.
397 		 * Firmware guarantees that a signal will be sent to the
398 		 * driver, triggering a reset.
399 		 */
400 		deactivate_rx_pools(adapter);
401 		netif_carrier_off(adapter->netdev);
402 	}
403 }
404 
405 static void replenish_pools(struct ibmvnic_adapter *adapter)
406 {
407 	int i;
408 
409 	adapter->replenish_task_cycles++;
410 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
411 	     i++) {
412 		if (adapter->rx_pool[i].active)
413 			replenish_rx_pool(adapter, &adapter->rx_pool[i]);
414 	}
415 }
416 
417 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
418 {
419 	kfree(adapter->tx_stats_buffers);
420 	kfree(adapter->rx_stats_buffers);
421 	adapter->tx_stats_buffers = NULL;
422 	adapter->rx_stats_buffers = NULL;
423 }
424 
425 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
426 {
427 	adapter->tx_stats_buffers =
428 				kcalloc(IBMVNIC_MAX_QUEUES,
429 					sizeof(struct ibmvnic_tx_queue_stats),
430 					GFP_KERNEL);
431 	if (!adapter->tx_stats_buffers)
432 		return -ENOMEM;
433 
434 	adapter->rx_stats_buffers =
435 				kcalloc(IBMVNIC_MAX_QUEUES,
436 					sizeof(struct ibmvnic_rx_queue_stats),
437 					GFP_KERNEL);
438 	if (!adapter->rx_stats_buffers)
439 		return -ENOMEM;
440 
441 	return 0;
442 }
443 
444 static void release_stats_token(struct ibmvnic_adapter *adapter)
445 {
446 	struct device *dev = &adapter->vdev->dev;
447 
448 	if (!adapter->stats_token)
449 		return;
450 
451 	dma_unmap_single(dev, adapter->stats_token,
452 			 sizeof(struct ibmvnic_statistics),
453 			 DMA_FROM_DEVICE);
454 	adapter->stats_token = 0;
455 }
456 
457 static int init_stats_token(struct ibmvnic_adapter *adapter)
458 {
459 	struct device *dev = &adapter->vdev->dev;
460 	dma_addr_t stok;
461 
462 	stok = dma_map_single(dev, &adapter->stats,
463 			      sizeof(struct ibmvnic_statistics),
464 			      DMA_FROM_DEVICE);
465 	if (dma_mapping_error(dev, stok)) {
466 		dev_err(dev, "Couldn't map stats buffer\n");
467 		return -1;
468 	}
469 
470 	adapter->stats_token = stok;
471 	netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
472 	return 0;
473 }
474 
475 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
476 {
477 	struct ibmvnic_rx_pool *rx_pool;
478 	int rx_scrqs;
479 	int i, j, rc;
480 	u64 *size_array;
481 
482 	if (!adapter->rx_pool)
483 		return -1;
484 
485 	size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
486 		be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
487 
488 	rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
489 	for (i = 0; i < rx_scrqs; i++) {
490 		rx_pool = &adapter->rx_pool[i];
491 
492 		netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
493 
494 		if (rx_pool->buff_size != be64_to_cpu(size_array[i])) {
495 			free_long_term_buff(adapter, &rx_pool->long_term_buff);
496 			rx_pool->buff_size = be64_to_cpu(size_array[i]);
497 			rc = alloc_long_term_buff(adapter,
498 						  &rx_pool->long_term_buff,
499 						  rx_pool->size *
500 						  rx_pool->buff_size);
501 		} else {
502 			rc = reset_long_term_buff(adapter,
503 						  &rx_pool->long_term_buff);
504 		}
505 
506 		if (rc)
507 			return rc;
508 
509 		for (j = 0; j < rx_pool->size; j++)
510 			rx_pool->free_map[j] = j;
511 
512 		memset(rx_pool->rx_buff, 0,
513 		       rx_pool->size * sizeof(struct ibmvnic_rx_buff));
514 
515 		atomic_set(&rx_pool->available, 0);
516 		rx_pool->next_alloc = 0;
517 		rx_pool->next_free = 0;
518 		rx_pool->active = 1;
519 	}
520 
521 	return 0;
522 }
523 
524 static void release_rx_pools(struct ibmvnic_adapter *adapter)
525 {
526 	struct ibmvnic_rx_pool *rx_pool;
527 	int i, j;
528 
529 	if (!adapter->rx_pool)
530 		return;
531 
532 	for (i = 0; i < adapter->num_active_rx_pools; i++) {
533 		rx_pool = &adapter->rx_pool[i];
534 
535 		netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
536 
537 		kfree(rx_pool->free_map);
538 		free_long_term_buff(adapter, &rx_pool->long_term_buff);
539 
540 		if (!rx_pool->rx_buff)
541 			continue;
542 
543 		for (j = 0; j < rx_pool->size; j++) {
544 			if (rx_pool->rx_buff[j].skb) {
545 				dev_kfree_skb_any(rx_pool->rx_buff[j].skb);
546 				rx_pool->rx_buff[j].skb = NULL;
547 			}
548 		}
549 
550 		kfree(rx_pool->rx_buff);
551 	}
552 
553 	kfree(adapter->rx_pool);
554 	adapter->rx_pool = NULL;
555 	adapter->num_active_rx_pools = 0;
556 }
557 
558 static int init_rx_pools(struct net_device *netdev)
559 {
560 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
561 	struct device *dev = &adapter->vdev->dev;
562 	struct ibmvnic_rx_pool *rx_pool;
563 	int rxadd_subcrqs;
564 	u64 *size_array;
565 	int i, j;
566 
567 	rxadd_subcrqs =
568 		be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
569 	size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
570 		be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
571 
572 	adapter->rx_pool = kcalloc(rxadd_subcrqs,
573 				   sizeof(struct ibmvnic_rx_pool),
574 				   GFP_KERNEL);
575 	if (!adapter->rx_pool) {
576 		dev_err(dev, "Failed to allocate rx pools\n");
577 		return -1;
578 	}
579 
580 	adapter->num_active_rx_pools = rxadd_subcrqs;
581 
582 	for (i = 0; i < rxadd_subcrqs; i++) {
583 		rx_pool = &adapter->rx_pool[i];
584 
585 		netdev_dbg(adapter->netdev,
586 			   "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
587 			   i, adapter->req_rx_add_entries_per_subcrq,
588 			   be64_to_cpu(size_array[i]));
589 
590 		rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
591 		rx_pool->index = i;
592 		rx_pool->buff_size = be64_to_cpu(size_array[i]);
593 		rx_pool->active = 1;
594 
595 		rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
596 					    GFP_KERNEL);
597 		if (!rx_pool->free_map) {
598 			release_rx_pools(adapter);
599 			return -1;
600 		}
601 
602 		rx_pool->rx_buff = kcalloc(rx_pool->size,
603 					   sizeof(struct ibmvnic_rx_buff),
604 					   GFP_KERNEL);
605 		if (!rx_pool->rx_buff) {
606 			dev_err(dev, "Couldn't alloc rx buffers\n");
607 			release_rx_pools(adapter);
608 			return -1;
609 		}
610 
611 		if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
612 					 rx_pool->size * rx_pool->buff_size)) {
613 			release_rx_pools(adapter);
614 			return -1;
615 		}
616 
617 		for (j = 0; j < rx_pool->size; ++j)
618 			rx_pool->free_map[j] = j;
619 
620 		atomic_set(&rx_pool->available, 0);
621 		rx_pool->next_alloc = 0;
622 		rx_pool->next_free = 0;
623 	}
624 
625 	return 0;
626 }
627 
628 static int reset_one_tx_pool(struct ibmvnic_adapter *adapter,
629 			     struct ibmvnic_tx_pool *tx_pool)
630 {
631 	int rc, i;
632 
633 	rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
634 	if (rc)
635 		return rc;
636 
637 	memset(tx_pool->tx_buff, 0,
638 	       tx_pool->num_buffers *
639 	       sizeof(struct ibmvnic_tx_buff));
640 
641 	for (i = 0; i < tx_pool->num_buffers; i++)
642 		tx_pool->free_map[i] = i;
643 
644 	tx_pool->consumer_index = 0;
645 	tx_pool->producer_index = 0;
646 
647 	return 0;
648 }
649 
650 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
651 {
652 	int tx_scrqs;
653 	int i, rc;
654 
655 	if (!adapter->tx_pool)
656 		return -1;
657 
658 	tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
659 	for (i = 0; i < tx_scrqs; i++) {
660 		rc = reset_one_tx_pool(adapter, &adapter->tso_pool[i]);
661 		if (rc)
662 			return rc;
663 		rc = reset_one_tx_pool(adapter, &adapter->tx_pool[i]);
664 		if (rc)
665 			return rc;
666 	}
667 
668 	return 0;
669 }
670 
671 static void release_vpd_data(struct ibmvnic_adapter *adapter)
672 {
673 	if (!adapter->vpd)
674 		return;
675 
676 	kfree(adapter->vpd->buff);
677 	kfree(adapter->vpd);
678 
679 	adapter->vpd = NULL;
680 }
681 
682 static void release_one_tx_pool(struct ibmvnic_adapter *adapter,
683 				struct ibmvnic_tx_pool *tx_pool)
684 {
685 	kfree(tx_pool->tx_buff);
686 	kfree(tx_pool->free_map);
687 	free_long_term_buff(adapter, &tx_pool->long_term_buff);
688 }
689 
690 static void release_tx_pools(struct ibmvnic_adapter *adapter)
691 {
692 	int i;
693 
694 	if (!adapter->tx_pool)
695 		return;
696 
697 	for (i = 0; i < adapter->num_active_tx_pools; i++) {
698 		release_one_tx_pool(adapter, &adapter->tx_pool[i]);
699 		release_one_tx_pool(adapter, &adapter->tso_pool[i]);
700 	}
701 
702 	kfree(adapter->tx_pool);
703 	adapter->tx_pool = NULL;
704 	kfree(adapter->tso_pool);
705 	adapter->tso_pool = NULL;
706 	adapter->num_active_tx_pools = 0;
707 }
708 
709 static int init_one_tx_pool(struct net_device *netdev,
710 			    struct ibmvnic_tx_pool *tx_pool,
711 			    int num_entries, int buf_size)
712 {
713 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
714 	int i;
715 
716 	tx_pool->tx_buff = kcalloc(num_entries,
717 				   sizeof(struct ibmvnic_tx_buff),
718 				   GFP_KERNEL);
719 	if (!tx_pool->tx_buff)
720 		return -1;
721 
722 	if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
723 				 num_entries * buf_size))
724 		return -1;
725 
726 	tx_pool->free_map = kcalloc(num_entries, sizeof(int), GFP_KERNEL);
727 	if (!tx_pool->free_map)
728 		return -1;
729 
730 	for (i = 0; i < num_entries; i++)
731 		tx_pool->free_map[i] = i;
732 
733 	tx_pool->consumer_index = 0;
734 	tx_pool->producer_index = 0;
735 	tx_pool->num_buffers = num_entries;
736 	tx_pool->buf_size = buf_size;
737 
738 	return 0;
739 }
740 
741 static int init_tx_pools(struct net_device *netdev)
742 {
743 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
744 	int tx_subcrqs;
745 	int i, rc;
746 
747 	tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
748 	adapter->tx_pool = kcalloc(tx_subcrqs,
749 				   sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
750 	if (!adapter->tx_pool)
751 		return -1;
752 
753 	adapter->tso_pool = kcalloc(tx_subcrqs,
754 				    sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
755 	if (!adapter->tso_pool)
756 		return -1;
757 
758 	adapter->num_active_tx_pools = tx_subcrqs;
759 
760 	for (i = 0; i < tx_subcrqs; i++) {
761 		rc = init_one_tx_pool(netdev, &adapter->tx_pool[i],
762 				      adapter->req_tx_entries_per_subcrq,
763 				      adapter->req_mtu + VLAN_HLEN);
764 		if (rc) {
765 			release_tx_pools(adapter);
766 			return rc;
767 		}
768 
769 		rc = init_one_tx_pool(netdev, &adapter->tso_pool[i],
770 				      IBMVNIC_TSO_BUFS,
771 				      IBMVNIC_TSO_BUF_SZ);
772 		if (rc) {
773 			release_tx_pools(adapter);
774 			return rc;
775 		}
776 	}
777 
778 	return 0;
779 }
780 
781 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
782 {
783 	int i;
784 
785 	if (adapter->napi_enabled)
786 		return;
787 
788 	for (i = 0; i < adapter->req_rx_queues; i++)
789 		napi_enable(&adapter->napi[i]);
790 
791 	adapter->napi_enabled = true;
792 }
793 
794 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
795 {
796 	int i;
797 
798 	if (!adapter->napi_enabled)
799 		return;
800 
801 	for (i = 0; i < adapter->req_rx_queues; i++) {
802 		netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
803 		napi_disable(&adapter->napi[i]);
804 	}
805 
806 	adapter->napi_enabled = false;
807 }
808 
809 static int init_napi(struct ibmvnic_adapter *adapter)
810 {
811 	int i;
812 
813 	adapter->napi = kcalloc(adapter->req_rx_queues,
814 				sizeof(struct napi_struct), GFP_KERNEL);
815 	if (!adapter->napi)
816 		return -ENOMEM;
817 
818 	for (i = 0; i < adapter->req_rx_queues; i++) {
819 		netdev_dbg(adapter->netdev, "Adding napi[%d]\n", i);
820 		netif_napi_add(adapter->netdev, &adapter->napi[i],
821 			       ibmvnic_poll, NAPI_POLL_WEIGHT);
822 	}
823 
824 	adapter->num_active_rx_napi = adapter->req_rx_queues;
825 	return 0;
826 }
827 
828 static void release_napi(struct ibmvnic_adapter *adapter)
829 {
830 	int i;
831 
832 	if (!adapter->napi)
833 		return;
834 
835 	for (i = 0; i < adapter->num_active_rx_napi; i++) {
836 		netdev_dbg(adapter->netdev, "Releasing napi[%d]\n", i);
837 		netif_napi_del(&adapter->napi[i]);
838 	}
839 
840 	kfree(adapter->napi);
841 	adapter->napi = NULL;
842 	adapter->num_active_rx_napi = 0;
843 	adapter->napi_enabled = false;
844 }
845 
846 static int ibmvnic_login(struct net_device *netdev)
847 {
848 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
849 	unsigned long timeout = msecs_to_jiffies(30000);
850 	int retry_count = 0;
851 	int retries = 10;
852 	bool retry;
853 	int rc;
854 
855 	do {
856 		retry = false;
857 		if (retry_count > retries) {
858 			netdev_warn(netdev, "Login attempts exceeded\n");
859 			return -1;
860 		}
861 
862 		adapter->init_done_rc = 0;
863 		reinit_completion(&adapter->init_done);
864 		rc = send_login(adapter);
865 		if (rc) {
866 			netdev_warn(netdev, "Unable to login\n");
867 			return rc;
868 		}
869 
870 		if (!wait_for_completion_timeout(&adapter->init_done,
871 						 timeout)) {
872 			netdev_warn(netdev, "Login timed out, retrying...\n");
873 			retry = true;
874 			adapter->init_done_rc = 0;
875 			retry_count++;
876 			continue;
877 		}
878 
879 		if (adapter->init_done_rc == ABORTED) {
880 			netdev_warn(netdev, "Login aborted, retrying...\n");
881 			retry = true;
882 			adapter->init_done_rc = 0;
883 			retry_count++;
884 			/* FW or device may be busy, so
885 			 * wait a bit before retrying login
886 			 */
887 			msleep(500);
888 		} else if (adapter->init_done_rc == PARTIALSUCCESS) {
889 			retry_count++;
890 			release_sub_crqs(adapter, 1);
891 
892 			retry = true;
893 			netdev_dbg(netdev,
894 				   "Received partial success, retrying...\n");
895 			adapter->init_done_rc = 0;
896 			reinit_completion(&adapter->init_done);
897 			send_cap_queries(adapter);
898 			if (!wait_for_completion_timeout(&adapter->init_done,
899 							 timeout)) {
900 				netdev_warn(netdev,
901 					    "Capabilities query timed out\n");
902 				return -1;
903 			}
904 
905 			rc = init_sub_crqs(adapter);
906 			if (rc) {
907 				netdev_warn(netdev,
908 					    "SCRQ initialization failed\n");
909 				return -1;
910 			}
911 
912 			rc = init_sub_crq_irqs(adapter);
913 			if (rc) {
914 				netdev_warn(netdev,
915 					    "SCRQ irq initialization failed\n");
916 				return -1;
917 			}
918 		} else if (adapter->init_done_rc) {
919 			netdev_warn(netdev, "Adapter login failed\n");
920 			return -1;
921 		}
922 	} while (retry);
923 
924 	__ibmvnic_set_mac(netdev, adapter->mac_addr);
925 
926 	return 0;
927 }
928 
929 static void release_login_buffer(struct ibmvnic_adapter *adapter)
930 {
931 	kfree(adapter->login_buf);
932 	adapter->login_buf = NULL;
933 }
934 
935 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
936 {
937 	kfree(adapter->login_rsp_buf);
938 	adapter->login_rsp_buf = NULL;
939 }
940 
941 static void release_resources(struct ibmvnic_adapter *adapter)
942 {
943 	release_vpd_data(adapter);
944 
945 	release_tx_pools(adapter);
946 	release_rx_pools(adapter);
947 
948 	release_napi(adapter);
949 	release_login_rsp_buffer(adapter);
950 }
951 
952 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
953 {
954 	struct net_device *netdev = adapter->netdev;
955 	unsigned long timeout = msecs_to_jiffies(30000);
956 	union ibmvnic_crq crq;
957 	bool resend;
958 	int rc;
959 
960 	netdev_dbg(netdev, "setting link state %d\n", link_state);
961 
962 	memset(&crq, 0, sizeof(crq));
963 	crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
964 	crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
965 	crq.logical_link_state.link_state = link_state;
966 
967 	do {
968 		resend = false;
969 
970 		reinit_completion(&adapter->init_done);
971 		rc = ibmvnic_send_crq(adapter, &crq);
972 		if (rc) {
973 			netdev_err(netdev, "Failed to set link state\n");
974 			return rc;
975 		}
976 
977 		if (!wait_for_completion_timeout(&adapter->init_done,
978 						 timeout)) {
979 			netdev_err(netdev, "timeout setting link state\n");
980 			return -1;
981 		}
982 
983 		if (adapter->init_done_rc == 1) {
984 			/* Partuial success, delay and re-send */
985 			mdelay(1000);
986 			resend = true;
987 		} else if (adapter->init_done_rc) {
988 			netdev_warn(netdev, "Unable to set link state, rc=%d\n",
989 				    adapter->init_done_rc);
990 			return adapter->init_done_rc;
991 		}
992 	} while (resend);
993 
994 	return 0;
995 }
996 
997 static int set_real_num_queues(struct net_device *netdev)
998 {
999 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1000 	int rc;
1001 
1002 	netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
1003 		   adapter->req_tx_queues, adapter->req_rx_queues);
1004 
1005 	rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
1006 	if (rc) {
1007 		netdev_err(netdev, "failed to set the number of tx queues\n");
1008 		return rc;
1009 	}
1010 
1011 	rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
1012 	if (rc)
1013 		netdev_err(netdev, "failed to set the number of rx queues\n");
1014 
1015 	return rc;
1016 }
1017 
1018 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
1019 {
1020 	struct device *dev = &adapter->vdev->dev;
1021 	union ibmvnic_crq crq;
1022 	int len = 0;
1023 	int rc;
1024 
1025 	if (adapter->vpd->buff)
1026 		len = adapter->vpd->len;
1027 
1028 	mutex_lock(&adapter->fw_lock);
1029 	adapter->fw_done_rc = 0;
1030 	reinit_completion(&adapter->fw_done);
1031 
1032 	crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
1033 	crq.get_vpd_size.cmd = GET_VPD_SIZE;
1034 	rc = ibmvnic_send_crq(adapter, &crq);
1035 	if (rc) {
1036 		mutex_unlock(&adapter->fw_lock);
1037 		return rc;
1038 	}
1039 
1040 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1041 	if (rc) {
1042 		dev_err(dev, "Could not retrieve VPD size, rc = %d\n", rc);
1043 		mutex_unlock(&adapter->fw_lock);
1044 		return rc;
1045 	}
1046 	mutex_unlock(&adapter->fw_lock);
1047 
1048 	if (!adapter->vpd->len)
1049 		return -ENODATA;
1050 
1051 	if (!adapter->vpd->buff)
1052 		adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
1053 	else if (adapter->vpd->len != len)
1054 		adapter->vpd->buff =
1055 			krealloc(adapter->vpd->buff,
1056 				 adapter->vpd->len, GFP_KERNEL);
1057 
1058 	if (!adapter->vpd->buff) {
1059 		dev_err(dev, "Could allocate VPD buffer\n");
1060 		return -ENOMEM;
1061 	}
1062 
1063 	adapter->vpd->dma_addr =
1064 		dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
1065 			       DMA_FROM_DEVICE);
1066 	if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
1067 		dev_err(dev, "Could not map VPD buffer\n");
1068 		kfree(adapter->vpd->buff);
1069 		adapter->vpd->buff = NULL;
1070 		return -ENOMEM;
1071 	}
1072 
1073 	mutex_lock(&adapter->fw_lock);
1074 	adapter->fw_done_rc = 0;
1075 	reinit_completion(&adapter->fw_done);
1076 
1077 	crq.get_vpd.first = IBMVNIC_CRQ_CMD;
1078 	crq.get_vpd.cmd = GET_VPD;
1079 	crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
1080 	crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
1081 	rc = ibmvnic_send_crq(adapter, &crq);
1082 	if (rc) {
1083 		kfree(adapter->vpd->buff);
1084 		adapter->vpd->buff = NULL;
1085 		mutex_unlock(&adapter->fw_lock);
1086 		return rc;
1087 	}
1088 
1089 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1090 	if (rc) {
1091 		dev_err(dev, "Unable to retrieve VPD, rc = %d\n", rc);
1092 		kfree(adapter->vpd->buff);
1093 		adapter->vpd->buff = NULL;
1094 		mutex_unlock(&adapter->fw_lock);
1095 		return rc;
1096 	}
1097 
1098 	mutex_unlock(&adapter->fw_lock);
1099 	return 0;
1100 }
1101 
1102 static int init_resources(struct ibmvnic_adapter *adapter)
1103 {
1104 	struct net_device *netdev = adapter->netdev;
1105 	int rc;
1106 
1107 	rc = set_real_num_queues(netdev);
1108 	if (rc)
1109 		return rc;
1110 
1111 	adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
1112 	if (!adapter->vpd)
1113 		return -ENOMEM;
1114 
1115 	/* Vital Product Data (VPD) */
1116 	rc = ibmvnic_get_vpd(adapter);
1117 	if (rc) {
1118 		netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
1119 		return rc;
1120 	}
1121 
1122 	adapter->map_id = 1;
1123 
1124 	rc = init_napi(adapter);
1125 	if (rc)
1126 		return rc;
1127 
1128 	send_map_query(adapter);
1129 
1130 	rc = init_rx_pools(netdev);
1131 	if (rc)
1132 		return rc;
1133 
1134 	rc = init_tx_pools(netdev);
1135 	return rc;
1136 }
1137 
1138 static int __ibmvnic_open(struct net_device *netdev)
1139 {
1140 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1141 	enum vnic_state prev_state = adapter->state;
1142 	int i, rc;
1143 
1144 	adapter->state = VNIC_OPENING;
1145 	replenish_pools(adapter);
1146 	ibmvnic_napi_enable(adapter);
1147 
1148 	/* We're ready to receive frames, enable the sub-crq interrupts and
1149 	 * set the logical link state to up
1150 	 */
1151 	for (i = 0; i < adapter->req_rx_queues; i++) {
1152 		netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1153 		if (prev_state == VNIC_CLOSED)
1154 			enable_irq(adapter->rx_scrq[i]->irq);
1155 		enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1156 	}
1157 
1158 	for (i = 0; i < adapter->req_tx_queues; i++) {
1159 		netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1160 		if (prev_state == VNIC_CLOSED)
1161 			enable_irq(adapter->tx_scrq[i]->irq);
1162 		enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1163 	}
1164 
1165 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1166 	if (rc) {
1167 		for (i = 0; i < adapter->req_rx_queues; i++)
1168 			napi_disable(&adapter->napi[i]);
1169 		release_resources(adapter);
1170 		return rc;
1171 	}
1172 
1173 	netif_tx_start_all_queues(netdev);
1174 
1175 	if (prev_state == VNIC_CLOSED) {
1176 		for (i = 0; i < adapter->req_rx_queues; i++)
1177 			napi_schedule(&adapter->napi[i]);
1178 	}
1179 
1180 	adapter->state = VNIC_OPEN;
1181 	return rc;
1182 }
1183 
1184 static int ibmvnic_open(struct net_device *netdev)
1185 {
1186 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1187 	int rc;
1188 
1189 	/* If device failover is pending, just set device state and return.
1190 	 * Device operation will be handled by reset routine.
1191 	 */
1192 	if (adapter->failover_pending) {
1193 		adapter->state = VNIC_OPEN;
1194 		return 0;
1195 	}
1196 
1197 	if (adapter->state != VNIC_CLOSED) {
1198 		rc = ibmvnic_login(netdev);
1199 		if (rc)
1200 			return rc;
1201 
1202 		rc = init_resources(adapter);
1203 		if (rc) {
1204 			netdev_err(netdev, "failed to initialize resources\n");
1205 			release_resources(adapter);
1206 			return rc;
1207 		}
1208 	}
1209 
1210 	rc = __ibmvnic_open(netdev);
1211 
1212 	return rc;
1213 }
1214 
1215 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1216 {
1217 	struct ibmvnic_rx_pool *rx_pool;
1218 	struct ibmvnic_rx_buff *rx_buff;
1219 	u64 rx_entries;
1220 	int rx_scrqs;
1221 	int i, j;
1222 
1223 	if (!adapter->rx_pool)
1224 		return;
1225 
1226 	rx_scrqs = adapter->num_active_rx_pools;
1227 	rx_entries = adapter->req_rx_add_entries_per_subcrq;
1228 
1229 	/* Free any remaining skbs in the rx buffer pools */
1230 	for (i = 0; i < rx_scrqs; i++) {
1231 		rx_pool = &adapter->rx_pool[i];
1232 		if (!rx_pool || !rx_pool->rx_buff)
1233 			continue;
1234 
1235 		netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1236 		for (j = 0; j < rx_entries; j++) {
1237 			rx_buff = &rx_pool->rx_buff[j];
1238 			if (rx_buff && rx_buff->skb) {
1239 				dev_kfree_skb_any(rx_buff->skb);
1240 				rx_buff->skb = NULL;
1241 			}
1242 		}
1243 	}
1244 }
1245 
1246 static void clean_one_tx_pool(struct ibmvnic_adapter *adapter,
1247 			      struct ibmvnic_tx_pool *tx_pool)
1248 {
1249 	struct ibmvnic_tx_buff *tx_buff;
1250 	u64 tx_entries;
1251 	int i;
1252 
1253 	if (!tx_pool || !tx_pool->tx_buff)
1254 		return;
1255 
1256 	tx_entries = tx_pool->num_buffers;
1257 
1258 	for (i = 0; i < tx_entries; i++) {
1259 		tx_buff = &tx_pool->tx_buff[i];
1260 		if (tx_buff && tx_buff->skb) {
1261 			dev_kfree_skb_any(tx_buff->skb);
1262 			tx_buff->skb = NULL;
1263 		}
1264 	}
1265 }
1266 
1267 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1268 {
1269 	int tx_scrqs;
1270 	int i;
1271 
1272 	if (!adapter->tx_pool || !adapter->tso_pool)
1273 		return;
1274 
1275 	tx_scrqs = adapter->num_active_tx_pools;
1276 
1277 	/* Free any remaining skbs in the tx buffer pools */
1278 	for (i = 0; i < tx_scrqs; i++) {
1279 		netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1280 		clean_one_tx_pool(adapter, &adapter->tx_pool[i]);
1281 		clean_one_tx_pool(adapter, &adapter->tso_pool[i]);
1282 	}
1283 }
1284 
1285 static void ibmvnic_disable_irqs(struct ibmvnic_adapter *adapter)
1286 {
1287 	struct net_device *netdev = adapter->netdev;
1288 	int i;
1289 
1290 	if (adapter->tx_scrq) {
1291 		for (i = 0; i < adapter->req_tx_queues; i++)
1292 			if (adapter->tx_scrq[i]->irq) {
1293 				netdev_dbg(netdev,
1294 					   "Disabling tx_scrq[%d] irq\n", i);
1295 				disable_scrq_irq(adapter, adapter->tx_scrq[i]);
1296 				disable_irq(adapter->tx_scrq[i]->irq);
1297 			}
1298 	}
1299 
1300 	if (adapter->rx_scrq) {
1301 		for (i = 0; i < adapter->req_rx_queues; i++) {
1302 			if (adapter->rx_scrq[i]->irq) {
1303 				netdev_dbg(netdev,
1304 					   "Disabling rx_scrq[%d] irq\n", i);
1305 				disable_scrq_irq(adapter, adapter->rx_scrq[i]);
1306 				disable_irq(adapter->rx_scrq[i]->irq);
1307 			}
1308 		}
1309 	}
1310 }
1311 
1312 static void ibmvnic_cleanup(struct net_device *netdev)
1313 {
1314 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1315 
1316 	/* ensure that transmissions are stopped if called by do_reset */
1317 	if (test_bit(0, &adapter->resetting))
1318 		netif_tx_disable(netdev);
1319 	else
1320 		netif_tx_stop_all_queues(netdev);
1321 
1322 	ibmvnic_napi_disable(adapter);
1323 	ibmvnic_disable_irqs(adapter);
1324 
1325 	clean_rx_pools(adapter);
1326 	clean_tx_pools(adapter);
1327 }
1328 
1329 static int __ibmvnic_close(struct net_device *netdev)
1330 {
1331 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1332 	int rc = 0;
1333 
1334 	adapter->state = VNIC_CLOSING;
1335 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1336 	if (rc)
1337 		return rc;
1338 	adapter->state = VNIC_CLOSED;
1339 	return 0;
1340 }
1341 
1342 static int ibmvnic_close(struct net_device *netdev)
1343 {
1344 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1345 	int rc;
1346 
1347 	/* If device failover is pending, just set device state and return.
1348 	 * Device operation will be handled by reset routine.
1349 	 */
1350 	if (adapter->failover_pending) {
1351 		adapter->state = VNIC_CLOSED;
1352 		return 0;
1353 	}
1354 
1355 	rc = __ibmvnic_close(netdev);
1356 	ibmvnic_cleanup(netdev);
1357 
1358 	return rc;
1359 }
1360 
1361 /**
1362  * build_hdr_data - creates L2/L3/L4 header data buffer
1363  * @hdr_field - bitfield determining needed headers
1364  * @skb - socket buffer
1365  * @hdr_len - array of header lengths
1366  * @tot_len - total length of data
1367  *
1368  * Reads hdr_field to determine which headers are needed by firmware.
1369  * Builds a buffer containing these headers.  Saves individual header
1370  * lengths and total buffer length to be used to build descriptors.
1371  */
1372 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1373 			  int *hdr_len, u8 *hdr_data)
1374 {
1375 	int len = 0;
1376 	u8 *hdr;
1377 
1378 	if (skb_vlan_tagged(skb) && !skb_vlan_tag_present(skb))
1379 		hdr_len[0] = sizeof(struct vlan_ethhdr);
1380 	else
1381 		hdr_len[0] = sizeof(struct ethhdr);
1382 
1383 	if (skb->protocol == htons(ETH_P_IP)) {
1384 		hdr_len[1] = ip_hdr(skb)->ihl * 4;
1385 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1386 			hdr_len[2] = tcp_hdrlen(skb);
1387 		else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1388 			hdr_len[2] = sizeof(struct udphdr);
1389 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1390 		hdr_len[1] = sizeof(struct ipv6hdr);
1391 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1392 			hdr_len[2] = tcp_hdrlen(skb);
1393 		else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1394 			hdr_len[2] = sizeof(struct udphdr);
1395 	} else if (skb->protocol == htons(ETH_P_ARP)) {
1396 		hdr_len[1] = arp_hdr_len(skb->dev);
1397 		hdr_len[2] = 0;
1398 	}
1399 
1400 	memset(hdr_data, 0, 120);
1401 	if ((hdr_field >> 6) & 1) {
1402 		hdr = skb_mac_header(skb);
1403 		memcpy(hdr_data, hdr, hdr_len[0]);
1404 		len += hdr_len[0];
1405 	}
1406 
1407 	if ((hdr_field >> 5) & 1) {
1408 		hdr = skb_network_header(skb);
1409 		memcpy(hdr_data + len, hdr, hdr_len[1]);
1410 		len += hdr_len[1];
1411 	}
1412 
1413 	if ((hdr_field >> 4) & 1) {
1414 		hdr = skb_transport_header(skb);
1415 		memcpy(hdr_data + len, hdr, hdr_len[2]);
1416 		len += hdr_len[2];
1417 	}
1418 	return len;
1419 }
1420 
1421 /**
1422  * create_hdr_descs - create header and header extension descriptors
1423  * @hdr_field - bitfield determining needed headers
1424  * @data - buffer containing header data
1425  * @len - length of data buffer
1426  * @hdr_len - array of individual header lengths
1427  * @scrq_arr - descriptor array
1428  *
1429  * Creates header and, if needed, header extension descriptors and
1430  * places them in a descriptor array, scrq_arr
1431  */
1432 
1433 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1434 			    union sub_crq *scrq_arr)
1435 {
1436 	union sub_crq hdr_desc;
1437 	int tmp_len = len;
1438 	int num_descs = 0;
1439 	u8 *data, *cur;
1440 	int tmp;
1441 
1442 	while (tmp_len > 0) {
1443 		cur = hdr_data + len - tmp_len;
1444 
1445 		memset(&hdr_desc, 0, sizeof(hdr_desc));
1446 		if (cur != hdr_data) {
1447 			data = hdr_desc.hdr_ext.data;
1448 			tmp = tmp_len > 29 ? 29 : tmp_len;
1449 			hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1450 			hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1451 			hdr_desc.hdr_ext.len = tmp;
1452 		} else {
1453 			data = hdr_desc.hdr.data;
1454 			tmp = tmp_len > 24 ? 24 : tmp_len;
1455 			hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1456 			hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1457 			hdr_desc.hdr.len = tmp;
1458 			hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1459 			hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1460 			hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1461 			hdr_desc.hdr.flag = hdr_field << 1;
1462 		}
1463 		memcpy(data, cur, tmp);
1464 		tmp_len -= tmp;
1465 		*scrq_arr = hdr_desc;
1466 		scrq_arr++;
1467 		num_descs++;
1468 	}
1469 
1470 	return num_descs;
1471 }
1472 
1473 /**
1474  * build_hdr_descs_arr - build a header descriptor array
1475  * @skb - socket buffer
1476  * @num_entries - number of descriptors to be sent
1477  * @subcrq - first TX descriptor
1478  * @hdr_field - bit field determining which headers will be sent
1479  *
1480  * This function will build a TX descriptor array with applicable
1481  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1482  */
1483 
1484 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1485 				int *num_entries, u8 hdr_field)
1486 {
1487 	int hdr_len[3] = {0, 0, 0};
1488 	int tot_len;
1489 	u8 *hdr_data = txbuff->hdr_data;
1490 
1491 	tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1492 				 txbuff->hdr_data);
1493 	*num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1494 			 txbuff->indir_arr + 1);
1495 }
1496 
1497 static int ibmvnic_xmit_workarounds(struct sk_buff *skb,
1498 				    struct net_device *netdev)
1499 {
1500 	/* For some backing devices, mishandling of small packets
1501 	 * can result in a loss of connection or TX stall. Device
1502 	 * architects recommend that no packet should be smaller
1503 	 * than the minimum MTU value provided to the driver, so
1504 	 * pad any packets to that length
1505 	 */
1506 	if (skb->len < netdev->min_mtu)
1507 		return skb_put_padto(skb, netdev->min_mtu);
1508 
1509 	return 0;
1510 }
1511 
1512 static netdev_tx_t ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1513 {
1514 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1515 	int queue_num = skb_get_queue_mapping(skb);
1516 	u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1517 	struct device *dev = &adapter->vdev->dev;
1518 	struct ibmvnic_tx_buff *tx_buff = NULL;
1519 	struct ibmvnic_sub_crq_queue *tx_scrq;
1520 	struct ibmvnic_tx_pool *tx_pool;
1521 	unsigned int tx_send_failed = 0;
1522 	unsigned int tx_map_failed = 0;
1523 	unsigned int tx_dropped = 0;
1524 	unsigned int tx_packets = 0;
1525 	unsigned int tx_bytes = 0;
1526 	dma_addr_t data_dma_addr;
1527 	struct netdev_queue *txq;
1528 	unsigned long lpar_rc;
1529 	union sub_crq tx_crq;
1530 	unsigned int offset;
1531 	int num_entries = 1;
1532 	unsigned char *dst;
1533 	u64 *handle_array;
1534 	int index = 0;
1535 	u8 proto = 0;
1536 	netdev_tx_t ret = NETDEV_TX_OK;
1537 
1538 	if (test_bit(0, &adapter->resetting)) {
1539 		if (!netif_subqueue_stopped(netdev, skb))
1540 			netif_stop_subqueue(netdev, queue_num);
1541 		dev_kfree_skb_any(skb);
1542 
1543 		tx_send_failed++;
1544 		tx_dropped++;
1545 		ret = NETDEV_TX_OK;
1546 		goto out;
1547 	}
1548 
1549 	if (ibmvnic_xmit_workarounds(skb, netdev)) {
1550 		tx_dropped++;
1551 		tx_send_failed++;
1552 		ret = NETDEV_TX_OK;
1553 		goto out;
1554 	}
1555 	if (skb_is_gso(skb))
1556 		tx_pool = &adapter->tso_pool[queue_num];
1557 	else
1558 		tx_pool = &adapter->tx_pool[queue_num];
1559 
1560 	tx_scrq = adapter->tx_scrq[queue_num];
1561 	txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1562 	handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1563 		be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1564 
1565 	index = tx_pool->free_map[tx_pool->consumer_index];
1566 
1567 	if (index == IBMVNIC_INVALID_MAP) {
1568 		dev_kfree_skb_any(skb);
1569 		tx_send_failed++;
1570 		tx_dropped++;
1571 		ret = NETDEV_TX_OK;
1572 		goto out;
1573 	}
1574 
1575 	tx_pool->free_map[tx_pool->consumer_index] = IBMVNIC_INVALID_MAP;
1576 
1577 	offset = index * tx_pool->buf_size;
1578 	dst = tx_pool->long_term_buff.buff + offset;
1579 	memset(dst, 0, tx_pool->buf_size);
1580 	data_dma_addr = tx_pool->long_term_buff.addr + offset;
1581 
1582 	if (skb_shinfo(skb)->nr_frags) {
1583 		int cur, i;
1584 
1585 		/* Copy the head */
1586 		skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1587 		cur = skb_headlen(skb);
1588 
1589 		/* Copy the frags */
1590 		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1591 			const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1592 
1593 			memcpy(dst + cur,
1594 			       page_address(skb_frag_page(frag)) +
1595 			       skb_frag_off(frag), skb_frag_size(frag));
1596 			cur += skb_frag_size(frag);
1597 		}
1598 	} else {
1599 		skb_copy_from_linear_data(skb, dst, skb->len);
1600 	}
1601 
1602 	tx_pool->consumer_index =
1603 	    (tx_pool->consumer_index + 1) % tx_pool->num_buffers;
1604 
1605 	tx_buff = &tx_pool->tx_buff[index];
1606 	tx_buff->skb = skb;
1607 	tx_buff->data_dma[0] = data_dma_addr;
1608 	tx_buff->data_len[0] = skb->len;
1609 	tx_buff->index = index;
1610 	tx_buff->pool_index = queue_num;
1611 	tx_buff->last_frag = true;
1612 
1613 	memset(&tx_crq, 0, sizeof(tx_crq));
1614 	tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1615 	tx_crq.v1.type = IBMVNIC_TX_DESC;
1616 	tx_crq.v1.n_crq_elem = 1;
1617 	tx_crq.v1.n_sge = 1;
1618 	tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1619 
1620 	if (skb_is_gso(skb))
1621 		tx_crq.v1.correlator =
1622 			cpu_to_be32(index | IBMVNIC_TSO_POOL_MASK);
1623 	else
1624 		tx_crq.v1.correlator = cpu_to_be32(index);
1625 	tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1626 	tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1627 	tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1628 
1629 	if (adapter->vlan_header_insertion && skb_vlan_tag_present(skb)) {
1630 		tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1631 		tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1632 	}
1633 
1634 	if (skb->protocol == htons(ETH_P_IP)) {
1635 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1636 		proto = ip_hdr(skb)->protocol;
1637 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1638 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1639 		proto = ipv6_hdr(skb)->nexthdr;
1640 	}
1641 
1642 	if (proto == IPPROTO_TCP)
1643 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1644 	else if (proto == IPPROTO_UDP)
1645 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1646 
1647 	if (skb->ip_summed == CHECKSUM_PARTIAL) {
1648 		tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1649 		hdrs += 2;
1650 	}
1651 	if (skb_is_gso(skb)) {
1652 		tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1653 		tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1654 		hdrs += 2;
1655 	}
1656 	/* determine if l2/3/4 headers are sent to firmware */
1657 	if ((*hdrs >> 7) & 1) {
1658 		build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1659 		tx_crq.v1.n_crq_elem = num_entries;
1660 		tx_buff->num_entries = num_entries;
1661 		tx_buff->indir_arr[0] = tx_crq;
1662 		tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1663 						    sizeof(tx_buff->indir_arr),
1664 						    DMA_TO_DEVICE);
1665 		if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1666 			dev_kfree_skb_any(skb);
1667 			tx_buff->skb = NULL;
1668 			if (!firmware_has_feature(FW_FEATURE_CMO))
1669 				dev_err(dev, "tx: unable to map descriptor array\n");
1670 			tx_map_failed++;
1671 			tx_dropped++;
1672 			ret = NETDEV_TX_OK;
1673 			goto tx_err_out;
1674 		}
1675 		lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1676 					       (u64)tx_buff->indir_dma,
1677 					       (u64)num_entries);
1678 		dma_unmap_single(dev, tx_buff->indir_dma,
1679 				 sizeof(tx_buff->indir_arr), DMA_TO_DEVICE);
1680 	} else {
1681 		tx_buff->num_entries = num_entries;
1682 		lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1683 				      &tx_crq);
1684 	}
1685 	if (lpar_rc != H_SUCCESS) {
1686 		if (lpar_rc != H_CLOSED && lpar_rc != H_PARAMETER)
1687 			dev_err_ratelimited(dev, "tx: send failed\n");
1688 		dev_kfree_skb_any(skb);
1689 		tx_buff->skb = NULL;
1690 
1691 		if (lpar_rc == H_CLOSED || adapter->failover_pending) {
1692 			/* Disable TX and report carrier off if queue is closed
1693 			 * or pending failover.
1694 			 * Firmware guarantees that a signal will be sent to the
1695 			 * driver, triggering a reset or some other action.
1696 			 */
1697 			netif_tx_stop_all_queues(netdev);
1698 			netif_carrier_off(netdev);
1699 		}
1700 
1701 		tx_send_failed++;
1702 		tx_dropped++;
1703 		ret = NETDEV_TX_OK;
1704 		goto tx_err_out;
1705 	}
1706 
1707 	if (atomic_add_return(num_entries, &tx_scrq->used)
1708 					>= adapter->req_tx_entries_per_subcrq) {
1709 		netdev_dbg(netdev, "Stopping queue %d\n", queue_num);
1710 		netif_stop_subqueue(netdev, queue_num);
1711 	}
1712 
1713 	tx_packets++;
1714 	tx_bytes += skb->len;
1715 	txq->trans_start = jiffies;
1716 	ret = NETDEV_TX_OK;
1717 	goto out;
1718 
1719 tx_err_out:
1720 	/* roll back consumer index and map array*/
1721 	if (tx_pool->consumer_index == 0)
1722 		tx_pool->consumer_index =
1723 			tx_pool->num_buffers - 1;
1724 	else
1725 		tx_pool->consumer_index--;
1726 	tx_pool->free_map[tx_pool->consumer_index] = index;
1727 out:
1728 	netdev->stats.tx_dropped += tx_dropped;
1729 	netdev->stats.tx_bytes += tx_bytes;
1730 	netdev->stats.tx_packets += tx_packets;
1731 	adapter->tx_send_failed += tx_send_failed;
1732 	adapter->tx_map_failed += tx_map_failed;
1733 	adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1734 	adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1735 	adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1736 
1737 	return ret;
1738 }
1739 
1740 static void ibmvnic_set_multi(struct net_device *netdev)
1741 {
1742 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1743 	struct netdev_hw_addr *ha;
1744 	union ibmvnic_crq crq;
1745 
1746 	memset(&crq, 0, sizeof(crq));
1747 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
1748 	crq.request_capability.cmd = REQUEST_CAPABILITY;
1749 
1750 	if (netdev->flags & IFF_PROMISC) {
1751 		if (!adapter->promisc_supported)
1752 			return;
1753 	} else {
1754 		if (netdev->flags & IFF_ALLMULTI) {
1755 			/* Accept all multicast */
1756 			memset(&crq, 0, sizeof(crq));
1757 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1758 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1759 			crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1760 			ibmvnic_send_crq(adapter, &crq);
1761 		} else if (netdev_mc_empty(netdev)) {
1762 			/* Reject all multicast */
1763 			memset(&crq, 0, sizeof(crq));
1764 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1765 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1766 			crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1767 			ibmvnic_send_crq(adapter, &crq);
1768 		} else {
1769 			/* Accept one or more multicast(s) */
1770 			netdev_for_each_mc_addr(ha, netdev) {
1771 				memset(&crq, 0, sizeof(crq));
1772 				crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1773 				crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1774 				crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1775 				ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1776 						ha->addr);
1777 				ibmvnic_send_crq(adapter, &crq);
1778 			}
1779 		}
1780 	}
1781 }
1782 
1783 static int __ibmvnic_set_mac(struct net_device *netdev, u8 *dev_addr)
1784 {
1785 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1786 	union ibmvnic_crq crq;
1787 	int rc;
1788 
1789 	if (!is_valid_ether_addr(dev_addr)) {
1790 		rc = -EADDRNOTAVAIL;
1791 		goto err;
1792 	}
1793 
1794 	memset(&crq, 0, sizeof(crq));
1795 	crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1796 	crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1797 	ether_addr_copy(&crq.change_mac_addr.mac_addr[0], dev_addr);
1798 
1799 	mutex_lock(&adapter->fw_lock);
1800 	adapter->fw_done_rc = 0;
1801 	reinit_completion(&adapter->fw_done);
1802 
1803 	rc = ibmvnic_send_crq(adapter, &crq);
1804 	if (rc) {
1805 		rc = -EIO;
1806 		mutex_unlock(&adapter->fw_lock);
1807 		goto err;
1808 	}
1809 
1810 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1811 	/* netdev->dev_addr is changed in handle_change_mac_rsp function */
1812 	if (rc || adapter->fw_done_rc) {
1813 		rc = -EIO;
1814 		mutex_unlock(&adapter->fw_lock);
1815 		goto err;
1816 	}
1817 	mutex_unlock(&adapter->fw_lock);
1818 	return 0;
1819 err:
1820 	ether_addr_copy(adapter->mac_addr, netdev->dev_addr);
1821 	return rc;
1822 }
1823 
1824 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1825 {
1826 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1827 	struct sockaddr *addr = p;
1828 	int rc;
1829 
1830 	rc = 0;
1831 	ether_addr_copy(adapter->mac_addr, addr->sa_data);
1832 	if (adapter->state != VNIC_PROBED)
1833 		rc = __ibmvnic_set_mac(netdev, addr->sa_data);
1834 
1835 	return rc;
1836 }
1837 
1838 /**
1839  * do_change_param_reset returns zero if we are able to keep processing reset
1840  * events, or non-zero if we hit a fatal error and must halt.
1841  */
1842 static int do_change_param_reset(struct ibmvnic_adapter *adapter,
1843 				 struct ibmvnic_rwi *rwi,
1844 				 u32 reset_state)
1845 {
1846 	struct net_device *netdev = adapter->netdev;
1847 	int i, rc;
1848 
1849 	netdev_dbg(adapter->netdev, "Change param resetting driver (%d)\n",
1850 		   rwi->reset_reason);
1851 
1852 	netif_carrier_off(netdev);
1853 	adapter->reset_reason = rwi->reset_reason;
1854 
1855 	ibmvnic_cleanup(netdev);
1856 
1857 	if (reset_state == VNIC_OPEN) {
1858 		rc = __ibmvnic_close(netdev);
1859 		if (rc)
1860 			return rc;
1861 	}
1862 
1863 	release_resources(adapter);
1864 	release_sub_crqs(adapter, 1);
1865 	release_crq_queue(adapter);
1866 
1867 	adapter->state = VNIC_PROBED;
1868 
1869 	rc = init_crq_queue(adapter);
1870 
1871 	if (rc) {
1872 		netdev_err(adapter->netdev,
1873 			   "Couldn't initialize crq. rc=%d\n", rc);
1874 		return rc;
1875 	}
1876 
1877 	rc = ibmvnic_reset_init(adapter);
1878 	if (rc)
1879 		return IBMVNIC_INIT_FAILED;
1880 
1881 	/* If the adapter was in PROBE state prior to the reset,
1882 	 * exit here.
1883 	 */
1884 	if (reset_state == VNIC_PROBED)
1885 		return 0;
1886 
1887 	rc = ibmvnic_login(netdev);
1888 	if (rc) {
1889 		adapter->state = reset_state;
1890 		return rc;
1891 	}
1892 
1893 	rc = init_resources(adapter);
1894 	if (rc)
1895 		return rc;
1896 
1897 	ibmvnic_disable_irqs(adapter);
1898 
1899 	adapter->state = VNIC_CLOSED;
1900 
1901 	if (reset_state == VNIC_CLOSED)
1902 		return 0;
1903 
1904 	rc = __ibmvnic_open(netdev);
1905 	if (rc)
1906 		return IBMVNIC_OPEN_FAILED;
1907 
1908 	/* refresh device's multicast list */
1909 	ibmvnic_set_multi(netdev);
1910 
1911 	/* kick napi */
1912 	for (i = 0; i < adapter->req_rx_queues; i++)
1913 		napi_schedule(&adapter->napi[i]);
1914 
1915 	return 0;
1916 }
1917 
1918 /**
1919  * do_reset returns zero if we are able to keep processing reset events, or
1920  * non-zero if we hit a fatal error and must halt.
1921  */
1922 static int do_reset(struct ibmvnic_adapter *adapter,
1923 		    struct ibmvnic_rwi *rwi, u32 reset_state)
1924 {
1925 	u64 old_num_rx_queues, old_num_tx_queues;
1926 	u64 old_num_rx_slots, old_num_tx_slots;
1927 	struct net_device *netdev = adapter->netdev;
1928 	int i, rc;
1929 
1930 	netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1931 		   rwi->reset_reason);
1932 
1933 	rtnl_lock();
1934 
1935 	netif_carrier_off(netdev);
1936 	adapter->reset_reason = rwi->reset_reason;
1937 
1938 	old_num_rx_queues = adapter->req_rx_queues;
1939 	old_num_tx_queues = adapter->req_tx_queues;
1940 	old_num_rx_slots = adapter->req_rx_add_entries_per_subcrq;
1941 	old_num_tx_slots = adapter->req_tx_entries_per_subcrq;
1942 
1943 	ibmvnic_cleanup(netdev);
1944 
1945 	if (reset_state == VNIC_OPEN &&
1946 	    adapter->reset_reason != VNIC_RESET_MOBILITY &&
1947 	    adapter->reset_reason != VNIC_RESET_FAILOVER) {
1948 		adapter->state = VNIC_CLOSING;
1949 
1950 		/* Release the RTNL lock before link state change and
1951 		 * re-acquire after the link state change to allow
1952 		 * linkwatch_event to grab the RTNL lock and run during
1953 		 * a reset.
1954 		 */
1955 		rtnl_unlock();
1956 		rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1957 		rtnl_lock();
1958 		if (rc)
1959 			goto out;
1960 
1961 		if (adapter->state != VNIC_CLOSING) {
1962 			rc = -1;
1963 			goto out;
1964 		}
1965 
1966 		adapter->state = VNIC_CLOSED;
1967 	}
1968 
1969 	if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1970 		/* remove the closed state so when we call open it appears
1971 		 * we are coming from the probed state.
1972 		 */
1973 		adapter->state = VNIC_PROBED;
1974 
1975 		if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
1976 			rc = ibmvnic_reenable_crq_queue(adapter);
1977 			release_sub_crqs(adapter, 1);
1978 		} else {
1979 			rc = ibmvnic_reset_crq(adapter);
1980 			if (rc == H_CLOSED || rc == H_SUCCESS) {
1981 				rc = vio_enable_interrupts(adapter->vdev);
1982 				if (rc)
1983 					netdev_err(adapter->netdev,
1984 						   "Reset failed to enable interrupts. rc=%d\n",
1985 						   rc);
1986 			}
1987 		}
1988 
1989 		if (rc) {
1990 			netdev_err(adapter->netdev,
1991 				   "Reset couldn't initialize crq. rc=%d\n", rc);
1992 			goto out;
1993 		}
1994 
1995 		rc = ibmvnic_reset_init(adapter);
1996 		if (rc) {
1997 			rc = IBMVNIC_INIT_FAILED;
1998 			goto out;
1999 		}
2000 
2001 		/* If the adapter was in PROBE state prior to the reset,
2002 		 * exit here.
2003 		 */
2004 		if (reset_state == VNIC_PROBED) {
2005 			rc = 0;
2006 			goto out;
2007 		}
2008 
2009 		rc = ibmvnic_login(netdev);
2010 		if (rc) {
2011 			adapter->state = reset_state;
2012 			goto out;
2013 		}
2014 
2015 		if (adapter->req_rx_queues != old_num_rx_queues ||
2016 		    adapter->req_tx_queues != old_num_tx_queues ||
2017 		    adapter->req_rx_add_entries_per_subcrq !=
2018 		    old_num_rx_slots ||
2019 		    adapter->req_tx_entries_per_subcrq !=
2020 		    old_num_tx_slots ||
2021 		    !adapter->rx_pool ||
2022 		    !adapter->tso_pool ||
2023 		    !adapter->tx_pool) {
2024 			release_rx_pools(adapter);
2025 			release_tx_pools(adapter);
2026 			release_napi(adapter);
2027 			release_vpd_data(adapter);
2028 
2029 			rc = init_resources(adapter);
2030 			if (rc)
2031 				goto out;
2032 
2033 		} else {
2034 			rc = reset_tx_pools(adapter);
2035 			if (rc) {
2036 				netdev_dbg(adapter->netdev, "reset tx pools failed (%d)\n",
2037 						rc);
2038 				goto out;
2039 			}
2040 
2041 			rc = reset_rx_pools(adapter);
2042 			if (rc) {
2043 				netdev_dbg(adapter->netdev, "reset rx pools failed (%d)\n",
2044 						rc);
2045 				goto out;
2046 			}
2047 		}
2048 		ibmvnic_disable_irqs(adapter);
2049 	}
2050 	adapter->state = VNIC_CLOSED;
2051 
2052 	if (reset_state == VNIC_CLOSED) {
2053 		rc = 0;
2054 		goto out;
2055 	}
2056 
2057 	rc = __ibmvnic_open(netdev);
2058 	if (rc) {
2059 		rc = IBMVNIC_OPEN_FAILED;
2060 		goto out;
2061 	}
2062 
2063 	/* refresh device's multicast list */
2064 	ibmvnic_set_multi(netdev);
2065 
2066 	/* kick napi */
2067 	for (i = 0; i < adapter->req_rx_queues; i++)
2068 		napi_schedule(&adapter->napi[i]);
2069 
2070 	if (adapter->reset_reason != VNIC_RESET_FAILOVER)
2071 		call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, netdev);
2072 
2073 	rc = 0;
2074 
2075 out:
2076 	rtnl_unlock();
2077 
2078 	return rc;
2079 }
2080 
2081 static int do_hard_reset(struct ibmvnic_adapter *adapter,
2082 			 struct ibmvnic_rwi *rwi, u32 reset_state)
2083 {
2084 	struct net_device *netdev = adapter->netdev;
2085 	int rc;
2086 
2087 	netdev_dbg(adapter->netdev, "Hard resetting driver (%d)\n",
2088 		   rwi->reset_reason);
2089 
2090 	netif_carrier_off(netdev);
2091 	adapter->reset_reason = rwi->reset_reason;
2092 
2093 	ibmvnic_cleanup(netdev);
2094 	release_resources(adapter);
2095 	release_sub_crqs(adapter, 0);
2096 	release_crq_queue(adapter);
2097 
2098 	/* remove the closed state so when we call open it appears
2099 	 * we are coming from the probed state.
2100 	 */
2101 	adapter->state = VNIC_PROBED;
2102 
2103 	reinit_completion(&adapter->init_done);
2104 	rc = init_crq_queue(adapter);
2105 	if (rc) {
2106 		netdev_err(adapter->netdev,
2107 			   "Couldn't initialize crq. rc=%d\n", rc);
2108 		return rc;
2109 	}
2110 
2111 	rc = ibmvnic_init(adapter);
2112 	if (rc)
2113 		return rc;
2114 
2115 	/* If the adapter was in PROBE state prior to the reset,
2116 	 * exit here.
2117 	 */
2118 	if (reset_state == VNIC_PROBED)
2119 		return 0;
2120 
2121 	rc = ibmvnic_login(netdev);
2122 	if (rc) {
2123 		adapter->state = VNIC_PROBED;
2124 		return 0;
2125 	}
2126 
2127 	rc = init_resources(adapter);
2128 	if (rc)
2129 		return rc;
2130 
2131 	ibmvnic_disable_irqs(adapter);
2132 	adapter->state = VNIC_CLOSED;
2133 
2134 	if (reset_state == VNIC_CLOSED)
2135 		return 0;
2136 
2137 	rc = __ibmvnic_open(netdev);
2138 	if (rc)
2139 		return IBMVNIC_OPEN_FAILED;
2140 
2141 	return 0;
2142 }
2143 
2144 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
2145 {
2146 	struct ibmvnic_rwi *rwi;
2147 	unsigned long flags;
2148 
2149 	spin_lock_irqsave(&adapter->rwi_lock, flags);
2150 
2151 	if (!list_empty(&adapter->rwi_list)) {
2152 		rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
2153 				       list);
2154 		list_del(&rwi->list);
2155 	} else {
2156 		rwi = NULL;
2157 	}
2158 
2159 	spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2160 	return rwi;
2161 }
2162 
2163 static void free_all_rwi(struct ibmvnic_adapter *adapter)
2164 {
2165 	struct ibmvnic_rwi *rwi;
2166 
2167 	rwi = get_next_rwi(adapter);
2168 	while (rwi) {
2169 		kfree(rwi);
2170 		rwi = get_next_rwi(adapter);
2171 	}
2172 }
2173 
2174 static void __ibmvnic_reset(struct work_struct *work)
2175 {
2176 	struct ibmvnic_rwi *rwi;
2177 	struct ibmvnic_adapter *adapter;
2178 	bool saved_state = false;
2179 	unsigned long flags;
2180 	u32 reset_state;
2181 	int rc = 0;
2182 
2183 	adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
2184 
2185 	if (test_and_set_bit_lock(0, &adapter->resetting)) {
2186 		schedule_delayed_work(&adapter->ibmvnic_delayed_reset,
2187 				      IBMVNIC_RESET_DELAY);
2188 		return;
2189 	}
2190 
2191 	rwi = get_next_rwi(adapter);
2192 	while (rwi) {
2193 		spin_lock_irqsave(&adapter->state_lock, flags);
2194 
2195 		if (adapter->state == VNIC_REMOVING ||
2196 		    adapter->state == VNIC_REMOVED) {
2197 			spin_unlock_irqrestore(&adapter->state_lock, flags);
2198 			kfree(rwi);
2199 			rc = EBUSY;
2200 			break;
2201 		}
2202 
2203 		if (!saved_state) {
2204 			reset_state = adapter->state;
2205 			adapter->state = VNIC_RESETTING;
2206 			saved_state = true;
2207 		}
2208 		spin_unlock_irqrestore(&adapter->state_lock, flags);
2209 
2210 		if (rwi->reset_reason == VNIC_RESET_CHANGE_PARAM) {
2211 			/* CHANGE_PARAM requestor holds rtnl_lock */
2212 			rc = do_change_param_reset(adapter, rwi, reset_state);
2213 		} else if (adapter->force_reset_recovery) {
2214 			/* Transport event occurred during previous reset */
2215 			if (adapter->wait_for_reset) {
2216 				/* Previous was CHANGE_PARAM; caller locked */
2217 				adapter->force_reset_recovery = false;
2218 				rc = do_hard_reset(adapter, rwi, reset_state);
2219 			} else {
2220 				rtnl_lock();
2221 				adapter->force_reset_recovery = false;
2222 				rc = do_hard_reset(adapter, rwi, reset_state);
2223 				rtnl_unlock();
2224 			}
2225 		} else if (!(rwi->reset_reason == VNIC_RESET_FATAL &&
2226 				adapter->from_passive_init)) {
2227 			rc = do_reset(adapter, rwi, reset_state);
2228 		}
2229 		kfree(rwi);
2230 		if (rc == IBMVNIC_OPEN_FAILED) {
2231 			if (list_empty(&adapter->rwi_list))
2232 				adapter->state = VNIC_CLOSED;
2233 			else
2234 				adapter->state = reset_state;
2235 			rc = 0;
2236 		} else if (rc && rc != IBMVNIC_INIT_FAILED &&
2237 		    !adapter->force_reset_recovery)
2238 			break;
2239 
2240 		rwi = get_next_rwi(adapter);
2241 
2242 		if (rwi && (rwi->reset_reason == VNIC_RESET_FAILOVER ||
2243 			    rwi->reset_reason == VNIC_RESET_MOBILITY))
2244 			adapter->force_reset_recovery = true;
2245 	}
2246 
2247 	if (adapter->wait_for_reset) {
2248 		adapter->reset_done_rc = rc;
2249 		complete(&adapter->reset_done);
2250 	}
2251 
2252 	if (rc) {
2253 		netdev_dbg(adapter->netdev, "Reset failed\n");
2254 		free_all_rwi(adapter);
2255 	}
2256 
2257 	clear_bit_unlock(0, &adapter->resetting);
2258 }
2259 
2260 static void __ibmvnic_delayed_reset(struct work_struct *work)
2261 {
2262 	struct ibmvnic_adapter *adapter;
2263 
2264 	adapter = container_of(work, struct ibmvnic_adapter,
2265 			       ibmvnic_delayed_reset.work);
2266 	__ibmvnic_reset(&adapter->ibmvnic_reset);
2267 }
2268 
2269 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
2270 			 enum ibmvnic_reset_reason reason)
2271 {
2272 	struct list_head *entry, *tmp_entry;
2273 	struct ibmvnic_rwi *rwi, *tmp;
2274 	struct net_device *netdev = adapter->netdev;
2275 	unsigned long flags;
2276 	int ret;
2277 
2278 	if (adapter->state == VNIC_REMOVING ||
2279 	    adapter->state == VNIC_REMOVED ||
2280 	    adapter->failover_pending) {
2281 		ret = EBUSY;
2282 		netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
2283 		goto err;
2284 	}
2285 
2286 	if (adapter->state == VNIC_PROBING) {
2287 		netdev_warn(netdev, "Adapter reset during probe\n");
2288 		ret = adapter->init_done_rc = EAGAIN;
2289 		goto err;
2290 	}
2291 
2292 	spin_lock_irqsave(&adapter->rwi_lock, flags);
2293 
2294 	list_for_each(entry, &adapter->rwi_list) {
2295 		tmp = list_entry(entry, struct ibmvnic_rwi, list);
2296 		if (tmp->reset_reason == reason) {
2297 			netdev_dbg(netdev, "Skipping matching reset\n");
2298 			spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2299 			ret = EBUSY;
2300 			goto err;
2301 		}
2302 	}
2303 
2304 	rwi = kzalloc(sizeof(*rwi), GFP_ATOMIC);
2305 	if (!rwi) {
2306 		spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2307 		ibmvnic_close(netdev);
2308 		ret = ENOMEM;
2309 		goto err;
2310 	}
2311 	/* if we just received a transport event,
2312 	 * flush reset queue and process this reset
2313 	 */
2314 	if (adapter->force_reset_recovery && !list_empty(&adapter->rwi_list)) {
2315 		list_for_each_safe(entry, tmp_entry, &adapter->rwi_list)
2316 			list_del(entry);
2317 	}
2318 	rwi->reset_reason = reason;
2319 	list_add_tail(&rwi->list, &adapter->rwi_list);
2320 	spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2321 	netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
2322 	schedule_work(&adapter->ibmvnic_reset);
2323 
2324 	return 0;
2325 err:
2326 	return -ret;
2327 }
2328 
2329 static void ibmvnic_tx_timeout(struct net_device *dev, unsigned int txqueue)
2330 {
2331 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2332 
2333 	ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
2334 }
2335 
2336 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
2337 				  struct ibmvnic_rx_buff *rx_buff)
2338 {
2339 	struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
2340 
2341 	rx_buff->skb = NULL;
2342 
2343 	pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
2344 	pool->next_alloc = (pool->next_alloc + 1) % pool->size;
2345 
2346 	atomic_dec(&pool->available);
2347 }
2348 
2349 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2350 {
2351 	struct net_device *netdev = napi->dev;
2352 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2353 	int scrq_num = (int)(napi - adapter->napi);
2354 	int frames_processed = 0;
2355 
2356 restart_poll:
2357 	while (frames_processed < budget) {
2358 		struct sk_buff *skb;
2359 		struct ibmvnic_rx_buff *rx_buff;
2360 		union sub_crq *next;
2361 		u32 length;
2362 		u16 offset;
2363 		u8 flags = 0;
2364 
2365 		if (unlikely(test_bit(0, &adapter->resetting) &&
2366 			     adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2367 			enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2368 			napi_complete_done(napi, frames_processed);
2369 			return frames_processed;
2370 		}
2371 
2372 		if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
2373 			break;
2374 		next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
2375 		rx_buff =
2376 		    (struct ibmvnic_rx_buff *)be64_to_cpu(next->
2377 							  rx_comp.correlator);
2378 		/* do error checking */
2379 		if (next->rx_comp.rc) {
2380 			netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2381 				   be16_to_cpu(next->rx_comp.rc));
2382 			/* free the entry */
2383 			next->rx_comp.first = 0;
2384 			dev_kfree_skb_any(rx_buff->skb);
2385 			remove_buff_from_pool(adapter, rx_buff);
2386 			continue;
2387 		} else if (!rx_buff->skb) {
2388 			/* free the entry */
2389 			next->rx_comp.first = 0;
2390 			remove_buff_from_pool(adapter, rx_buff);
2391 			continue;
2392 		}
2393 
2394 		length = be32_to_cpu(next->rx_comp.len);
2395 		offset = be16_to_cpu(next->rx_comp.off_frame_data);
2396 		flags = next->rx_comp.flags;
2397 		skb = rx_buff->skb;
2398 		skb_copy_to_linear_data(skb, rx_buff->data + offset,
2399 					length);
2400 
2401 		/* VLAN Header has been stripped by the system firmware and
2402 		 * needs to be inserted by the driver
2403 		 */
2404 		if (adapter->rx_vlan_header_insertion &&
2405 		    (flags & IBMVNIC_VLAN_STRIPPED))
2406 			__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2407 					       ntohs(next->rx_comp.vlan_tci));
2408 
2409 		/* free the entry */
2410 		next->rx_comp.first = 0;
2411 		remove_buff_from_pool(adapter, rx_buff);
2412 
2413 		skb_put(skb, length);
2414 		skb->protocol = eth_type_trans(skb, netdev);
2415 		skb_record_rx_queue(skb, scrq_num);
2416 
2417 		if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2418 		    flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2419 			skb->ip_summed = CHECKSUM_UNNECESSARY;
2420 		}
2421 
2422 		length = skb->len;
2423 		napi_gro_receive(napi, skb); /* send it up */
2424 		netdev->stats.rx_packets++;
2425 		netdev->stats.rx_bytes += length;
2426 		adapter->rx_stats_buffers[scrq_num].packets++;
2427 		adapter->rx_stats_buffers[scrq_num].bytes += length;
2428 		frames_processed++;
2429 	}
2430 
2431 	if (adapter->state != VNIC_CLOSING)
2432 		replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2433 
2434 	if (frames_processed < budget) {
2435 		enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2436 		napi_complete_done(napi, frames_processed);
2437 		if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
2438 		    napi_reschedule(napi)) {
2439 			disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2440 			goto restart_poll;
2441 		}
2442 	}
2443 	return frames_processed;
2444 }
2445 
2446 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2447 {
2448 	int rc, ret;
2449 
2450 	adapter->fallback.mtu = adapter->req_mtu;
2451 	adapter->fallback.rx_queues = adapter->req_rx_queues;
2452 	adapter->fallback.tx_queues = adapter->req_tx_queues;
2453 	adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2454 	adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2455 
2456 	reinit_completion(&adapter->reset_done);
2457 	adapter->wait_for_reset = true;
2458 	rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2459 
2460 	if (rc) {
2461 		ret = rc;
2462 		goto out;
2463 	}
2464 	rc = ibmvnic_wait_for_completion(adapter, &adapter->reset_done, 60000);
2465 	if (rc) {
2466 		ret = -ENODEV;
2467 		goto out;
2468 	}
2469 
2470 	ret = 0;
2471 	if (adapter->reset_done_rc) {
2472 		ret = -EIO;
2473 		adapter->desired.mtu = adapter->fallback.mtu;
2474 		adapter->desired.rx_queues = adapter->fallback.rx_queues;
2475 		adapter->desired.tx_queues = adapter->fallback.tx_queues;
2476 		adapter->desired.rx_entries = adapter->fallback.rx_entries;
2477 		adapter->desired.tx_entries = adapter->fallback.tx_entries;
2478 
2479 		reinit_completion(&adapter->reset_done);
2480 		adapter->wait_for_reset = true;
2481 		rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2482 		if (rc) {
2483 			ret = rc;
2484 			goto out;
2485 		}
2486 		rc = ibmvnic_wait_for_completion(adapter, &adapter->reset_done,
2487 						 60000);
2488 		if (rc) {
2489 			ret = -ENODEV;
2490 			goto out;
2491 		}
2492 	}
2493 out:
2494 	adapter->wait_for_reset = false;
2495 
2496 	return ret;
2497 }
2498 
2499 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2500 {
2501 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2502 
2503 	adapter->desired.mtu = new_mtu + ETH_HLEN;
2504 
2505 	return wait_for_reset(adapter);
2506 }
2507 
2508 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2509 						struct net_device *dev,
2510 						netdev_features_t features)
2511 {
2512 	/* Some backing hardware adapters can not
2513 	 * handle packets with a MSS less than 224
2514 	 * or with only one segment.
2515 	 */
2516 	if (skb_is_gso(skb)) {
2517 		if (skb_shinfo(skb)->gso_size < 224 ||
2518 		    skb_shinfo(skb)->gso_segs == 1)
2519 			features &= ~NETIF_F_GSO_MASK;
2520 	}
2521 
2522 	return features;
2523 }
2524 
2525 static const struct net_device_ops ibmvnic_netdev_ops = {
2526 	.ndo_open		= ibmvnic_open,
2527 	.ndo_stop		= ibmvnic_close,
2528 	.ndo_start_xmit		= ibmvnic_xmit,
2529 	.ndo_set_rx_mode	= ibmvnic_set_multi,
2530 	.ndo_set_mac_address	= ibmvnic_set_mac,
2531 	.ndo_validate_addr	= eth_validate_addr,
2532 	.ndo_tx_timeout		= ibmvnic_tx_timeout,
2533 	.ndo_change_mtu		= ibmvnic_change_mtu,
2534 	.ndo_features_check     = ibmvnic_features_check,
2535 };
2536 
2537 /* ethtool functions */
2538 
2539 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2540 				      struct ethtool_link_ksettings *cmd)
2541 {
2542 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2543 	int rc;
2544 
2545 	rc = send_query_phys_parms(adapter);
2546 	if (rc) {
2547 		adapter->speed = SPEED_UNKNOWN;
2548 		adapter->duplex = DUPLEX_UNKNOWN;
2549 	}
2550 	cmd->base.speed = adapter->speed;
2551 	cmd->base.duplex = adapter->duplex;
2552 	cmd->base.port = PORT_FIBRE;
2553 	cmd->base.phy_address = 0;
2554 	cmd->base.autoneg = AUTONEG_ENABLE;
2555 
2556 	return 0;
2557 }
2558 
2559 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2560 				struct ethtool_drvinfo *info)
2561 {
2562 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2563 
2564 	strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2565 	strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2566 	strlcpy(info->fw_version, adapter->fw_version,
2567 		sizeof(info->fw_version));
2568 }
2569 
2570 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2571 {
2572 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2573 
2574 	return adapter->msg_enable;
2575 }
2576 
2577 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2578 {
2579 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2580 
2581 	adapter->msg_enable = data;
2582 }
2583 
2584 static u32 ibmvnic_get_link(struct net_device *netdev)
2585 {
2586 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2587 
2588 	/* Don't need to send a query because we request a logical link up at
2589 	 * init and then we wait for link state indications
2590 	 */
2591 	return adapter->logical_link_state;
2592 }
2593 
2594 static void ibmvnic_get_ringparam(struct net_device *netdev,
2595 				  struct ethtool_ringparam *ring)
2596 {
2597 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2598 
2599 	if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2600 		ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2601 		ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2602 	} else {
2603 		ring->rx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2604 		ring->tx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2605 	}
2606 	ring->rx_mini_max_pending = 0;
2607 	ring->rx_jumbo_max_pending = 0;
2608 	ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2609 	ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2610 	ring->rx_mini_pending = 0;
2611 	ring->rx_jumbo_pending = 0;
2612 }
2613 
2614 static int ibmvnic_set_ringparam(struct net_device *netdev,
2615 				 struct ethtool_ringparam *ring)
2616 {
2617 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2618 	int ret;
2619 
2620 	ret = 0;
2621 	adapter->desired.rx_entries = ring->rx_pending;
2622 	adapter->desired.tx_entries = ring->tx_pending;
2623 
2624 	ret = wait_for_reset(adapter);
2625 
2626 	if (!ret &&
2627 	    (adapter->req_rx_add_entries_per_subcrq != ring->rx_pending ||
2628 	     adapter->req_tx_entries_per_subcrq != ring->tx_pending))
2629 		netdev_info(netdev,
2630 			    "Could not match full ringsize request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2631 			    ring->rx_pending, ring->tx_pending,
2632 			    adapter->req_rx_add_entries_per_subcrq,
2633 			    adapter->req_tx_entries_per_subcrq);
2634 	return ret;
2635 }
2636 
2637 static void ibmvnic_get_channels(struct net_device *netdev,
2638 				 struct ethtool_channels *channels)
2639 {
2640 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2641 
2642 	if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2643 		channels->max_rx = adapter->max_rx_queues;
2644 		channels->max_tx = adapter->max_tx_queues;
2645 	} else {
2646 		channels->max_rx = IBMVNIC_MAX_QUEUES;
2647 		channels->max_tx = IBMVNIC_MAX_QUEUES;
2648 	}
2649 
2650 	channels->max_other = 0;
2651 	channels->max_combined = 0;
2652 	channels->rx_count = adapter->req_rx_queues;
2653 	channels->tx_count = adapter->req_tx_queues;
2654 	channels->other_count = 0;
2655 	channels->combined_count = 0;
2656 }
2657 
2658 static int ibmvnic_set_channels(struct net_device *netdev,
2659 				struct ethtool_channels *channels)
2660 {
2661 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2662 	int ret;
2663 
2664 	ret = 0;
2665 	adapter->desired.rx_queues = channels->rx_count;
2666 	adapter->desired.tx_queues = channels->tx_count;
2667 
2668 	ret = wait_for_reset(adapter);
2669 
2670 	if (!ret &&
2671 	    (adapter->req_rx_queues != channels->rx_count ||
2672 	     adapter->req_tx_queues != channels->tx_count))
2673 		netdev_info(netdev,
2674 			    "Could not match full channels request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2675 			    channels->rx_count, channels->tx_count,
2676 			    adapter->req_rx_queues, adapter->req_tx_queues);
2677 	return ret;
2678 
2679 }
2680 
2681 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2682 {
2683 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2684 	int i;
2685 
2686 	switch (stringset) {
2687 	case ETH_SS_STATS:
2688 		for (i = 0; i < ARRAY_SIZE(ibmvnic_stats);
2689 				i++, data += ETH_GSTRING_LEN)
2690 			memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2691 
2692 		for (i = 0; i < adapter->req_tx_queues; i++) {
2693 			snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2694 			data += ETH_GSTRING_LEN;
2695 
2696 			snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2697 			data += ETH_GSTRING_LEN;
2698 
2699 			snprintf(data, ETH_GSTRING_LEN,
2700 				 "tx%d_dropped_packets", i);
2701 			data += ETH_GSTRING_LEN;
2702 		}
2703 
2704 		for (i = 0; i < adapter->req_rx_queues; i++) {
2705 			snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2706 			data += ETH_GSTRING_LEN;
2707 
2708 			snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2709 			data += ETH_GSTRING_LEN;
2710 
2711 			snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2712 			data += ETH_GSTRING_LEN;
2713 		}
2714 		break;
2715 
2716 	case ETH_SS_PRIV_FLAGS:
2717 		for (i = 0; i < ARRAY_SIZE(ibmvnic_priv_flags); i++)
2718 			strcpy(data + i * ETH_GSTRING_LEN,
2719 			       ibmvnic_priv_flags[i]);
2720 		break;
2721 	default:
2722 		return;
2723 	}
2724 }
2725 
2726 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2727 {
2728 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2729 
2730 	switch (sset) {
2731 	case ETH_SS_STATS:
2732 		return ARRAY_SIZE(ibmvnic_stats) +
2733 		       adapter->req_tx_queues * NUM_TX_STATS +
2734 		       adapter->req_rx_queues * NUM_RX_STATS;
2735 	case ETH_SS_PRIV_FLAGS:
2736 		return ARRAY_SIZE(ibmvnic_priv_flags);
2737 	default:
2738 		return -EOPNOTSUPP;
2739 	}
2740 }
2741 
2742 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2743 				      struct ethtool_stats *stats, u64 *data)
2744 {
2745 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2746 	union ibmvnic_crq crq;
2747 	int i, j;
2748 	int rc;
2749 
2750 	memset(&crq, 0, sizeof(crq));
2751 	crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2752 	crq.request_statistics.cmd = REQUEST_STATISTICS;
2753 	crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2754 	crq.request_statistics.len =
2755 	    cpu_to_be32(sizeof(struct ibmvnic_statistics));
2756 
2757 	/* Wait for data to be written */
2758 	reinit_completion(&adapter->stats_done);
2759 	rc = ibmvnic_send_crq(adapter, &crq);
2760 	if (rc)
2761 		return;
2762 	rc = ibmvnic_wait_for_completion(adapter, &adapter->stats_done, 10000);
2763 	if (rc)
2764 		return;
2765 
2766 	for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2767 		data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2768 						ibmvnic_stats[i].offset));
2769 
2770 	for (j = 0; j < adapter->req_tx_queues; j++) {
2771 		data[i] = adapter->tx_stats_buffers[j].packets;
2772 		i++;
2773 		data[i] = adapter->tx_stats_buffers[j].bytes;
2774 		i++;
2775 		data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2776 		i++;
2777 	}
2778 
2779 	for (j = 0; j < adapter->req_rx_queues; j++) {
2780 		data[i] = adapter->rx_stats_buffers[j].packets;
2781 		i++;
2782 		data[i] = adapter->rx_stats_buffers[j].bytes;
2783 		i++;
2784 		data[i] = adapter->rx_stats_buffers[j].interrupts;
2785 		i++;
2786 	}
2787 }
2788 
2789 static u32 ibmvnic_get_priv_flags(struct net_device *netdev)
2790 {
2791 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2792 
2793 	return adapter->priv_flags;
2794 }
2795 
2796 static int ibmvnic_set_priv_flags(struct net_device *netdev, u32 flags)
2797 {
2798 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2799 	bool which_maxes = !!(flags & IBMVNIC_USE_SERVER_MAXES);
2800 
2801 	if (which_maxes)
2802 		adapter->priv_flags |= IBMVNIC_USE_SERVER_MAXES;
2803 	else
2804 		adapter->priv_flags &= ~IBMVNIC_USE_SERVER_MAXES;
2805 
2806 	return 0;
2807 }
2808 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2809 	.get_drvinfo		= ibmvnic_get_drvinfo,
2810 	.get_msglevel		= ibmvnic_get_msglevel,
2811 	.set_msglevel		= ibmvnic_set_msglevel,
2812 	.get_link		= ibmvnic_get_link,
2813 	.get_ringparam		= ibmvnic_get_ringparam,
2814 	.set_ringparam		= ibmvnic_set_ringparam,
2815 	.get_channels		= ibmvnic_get_channels,
2816 	.set_channels		= ibmvnic_set_channels,
2817 	.get_strings            = ibmvnic_get_strings,
2818 	.get_sset_count         = ibmvnic_get_sset_count,
2819 	.get_ethtool_stats	= ibmvnic_get_ethtool_stats,
2820 	.get_link_ksettings	= ibmvnic_get_link_ksettings,
2821 	.get_priv_flags		= ibmvnic_get_priv_flags,
2822 	.set_priv_flags		= ibmvnic_set_priv_flags,
2823 };
2824 
2825 /* Routines for managing CRQs/sCRQs  */
2826 
2827 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2828 				   struct ibmvnic_sub_crq_queue *scrq)
2829 {
2830 	int rc;
2831 
2832 	if (scrq->irq) {
2833 		free_irq(scrq->irq, scrq);
2834 		irq_dispose_mapping(scrq->irq);
2835 		scrq->irq = 0;
2836 	}
2837 
2838 	memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2839 	atomic_set(&scrq->used, 0);
2840 	scrq->cur = 0;
2841 
2842 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2843 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2844 	return rc;
2845 }
2846 
2847 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2848 {
2849 	int i, rc;
2850 
2851 	for (i = 0; i < adapter->req_tx_queues; i++) {
2852 		netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2853 		rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2854 		if (rc)
2855 			return rc;
2856 	}
2857 
2858 	for (i = 0; i < adapter->req_rx_queues; i++) {
2859 		netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2860 		rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2861 		if (rc)
2862 			return rc;
2863 	}
2864 
2865 	return rc;
2866 }
2867 
2868 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2869 				  struct ibmvnic_sub_crq_queue *scrq,
2870 				  bool do_h_free)
2871 {
2872 	struct device *dev = &adapter->vdev->dev;
2873 	long rc;
2874 
2875 	netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2876 
2877 	if (do_h_free) {
2878 		/* Close the sub-crqs */
2879 		do {
2880 			rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2881 						adapter->vdev->unit_address,
2882 						scrq->crq_num);
2883 		} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2884 
2885 		if (rc) {
2886 			netdev_err(adapter->netdev,
2887 				   "Failed to release sub-CRQ %16lx, rc = %ld\n",
2888 				   scrq->crq_num, rc);
2889 		}
2890 	}
2891 
2892 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2893 			 DMA_BIDIRECTIONAL);
2894 	free_pages((unsigned long)scrq->msgs, 2);
2895 	kfree(scrq);
2896 }
2897 
2898 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2899 							*adapter)
2900 {
2901 	struct device *dev = &adapter->vdev->dev;
2902 	struct ibmvnic_sub_crq_queue *scrq;
2903 	int rc;
2904 
2905 	scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2906 	if (!scrq)
2907 		return NULL;
2908 
2909 	scrq->msgs =
2910 		(union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2911 	if (!scrq->msgs) {
2912 		dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2913 		goto zero_page_failed;
2914 	}
2915 
2916 	scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2917 					 DMA_BIDIRECTIONAL);
2918 	if (dma_mapping_error(dev, scrq->msg_token)) {
2919 		dev_warn(dev, "Couldn't map crq queue messages page\n");
2920 		goto map_failed;
2921 	}
2922 
2923 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2924 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2925 
2926 	if (rc == H_RESOURCE)
2927 		rc = ibmvnic_reset_crq(adapter);
2928 
2929 	if (rc == H_CLOSED) {
2930 		dev_warn(dev, "Partner adapter not ready, waiting.\n");
2931 	} else if (rc) {
2932 		dev_warn(dev, "Error %d registering sub-crq\n", rc);
2933 		goto reg_failed;
2934 	}
2935 
2936 	scrq->adapter = adapter;
2937 	scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2938 	spin_lock_init(&scrq->lock);
2939 
2940 	netdev_dbg(adapter->netdev,
2941 		   "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2942 		   scrq->crq_num, scrq->hw_irq, scrq->irq);
2943 
2944 	return scrq;
2945 
2946 reg_failed:
2947 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2948 			 DMA_BIDIRECTIONAL);
2949 map_failed:
2950 	free_pages((unsigned long)scrq->msgs, 2);
2951 zero_page_failed:
2952 	kfree(scrq);
2953 
2954 	return NULL;
2955 }
2956 
2957 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
2958 {
2959 	int i;
2960 
2961 	if (adapter->tx_scrq) {
2962 		for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
2963 			if (!adapter->tx_scrq[i])
2964 				continue;
2965 
2966 			netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2967 				   i);
2968 			if (adapter->tx_scrq[i]->irq) {
2969 				free_irq(adapter->tx_scrq[i]->irq,
2970 					 adapter->tx_scrq[i]);
2971 				irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2972 				adapter->tx_scrq[i]->irq = 0;
2973 			}
2974 
2975 			release_sub_crq_queue(adapter, adapter->tx_scrq[i],
2976 					      do_h_free);
2977 		}
2978 
2979 		kfree(adapter->tx_scrq);
2980 		adapter->tx_scrq = NULL;
2981 		adapter->num_active_tx_scrqs = 0;
2982 	}
2983 
2984 	if (adapter->rx_scrq) {
2985 		for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
2986 			if (!adapter->rx_scrq[i])
2987 				continue;
2988 
2989 			netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2990 				   i);
2991 			if (adapter->rx_scrq[i]->irq) {
2992 				free_irq(adapter->rx_scrq[i]->irq,
2993 					 adapter->rx_scrq[i]);
2994 				irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2995 				adapter->rx_scrq[i]->irq = 0;
2996 			}
2997 
2998 			release_sub_crq_queue(adapter, adapter->rx_scrq[i],
2999 					      do_h_free);
3000 		}
3001 
3002 		kfree(adapter->rx_scrq);
3003 		adapter->rx_scrq = NULL;
3004 		adapter->num_active_rx_scrqs = 0;
3005 	}
3006 }
3007 
3008 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
3009 			    struct ibmvnic_sub_crq_queue *scrq)
3010 {
3011 	struct device *dev = &adapter->vdev->dev;
3012 	unsigned long rc;
3013 
3014 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
3015 				H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
3016 	if (rc)
3017 		dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
3018 			scrq->hw_irq, rc);
3019 	return rc;
3020 }
3021 
3022 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
3023 			   struct ibmvnic_sub_crq_queue *scrq)
3024 {
3025 	struct device *dev = &adapter->vdev->dev;
3026 	unsigned long rc;
3027 
3028 	if (scrq->hw_irq > 0x100000000ULL) {
3029 		dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
3030 		return 1;
3031 	}
3032 
3033 	if (test_bit(0, &adapter->resetting) &&
3034 	    adapter->reset_reason == VNIC_RESET_MOBILITY) {
3035 		u64 val = (0xff000000) | scrq->hw_irq;
3036 
3037 		rc = plpar_hcall_norets(H_EOI, val);
3038 		/* H_EOI would fail with rc = H_FUNCTION when running
3039 		 * in XIVE mode which is expected, but not an error.
3040 		 */
3041 		if (rc && (rc != H_FUNCTION))
3042 			dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
3043 				val, rc);
3044 	}
3045 
3046 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
3047 				H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
3048 	if (rc)
3049 		dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
3050 			scrq->hw_irq, rc);
3051 	return rc;
3052 }
3053 
3054 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
3055 			       struct ibmvnic_sub_crq_queue *scrq)
3056 {
3057 	struct device *dev = &adapter->vdev->dev;
3058 	struct ibmvnic_tx_pool *tx_pool;
3059 	struct ibmvnic_tx_buff *txbuff;
3060 	union sub_crq *next;
3061 	int index;
3062 	int i, j;
3063 
3064 restart_loop:
3065 	while (pending_scrq(adapter, scrq)) {
3066 		unsigned int pool = scrq->pool_index;
3067 		int num_entries = 0;
3068 
3069 		next = ibmvnic_next_scrq(adapter, scrq);
3070 		for (i = 0; i < next->tx_comp.num_comps; i++) {
3071 			if (next->tx_comp.rcs[i]) {
3072 				dev_err(dev, "tx error %x\n",
3073 					next->tx_comp.rcs[i]);
3074 				continue;
3075 			}
3076 			index = be32_to_cpu(next->tx_comp.correlators[i]);
3077 			if (index & IBMVNIC_TSO_POOL_MASK) {
3078 				tx_pool = &adapter->tso_pool[pool];
3079 				index &= ~IBMVNIC_TSO_POOL_MASK;
3080 			} else {
3081 				tx_pool = &adapter->tx_pool[pool];
3082 			}
3083 
3084 			txbuff = &tx_pool->tx_buff[index];
3085 
3086 			for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
3087 				if (!txbuff->data_dma[j])
3088 					continue;
3089 
3090 				txbuff->data_dma[j] = 0;
3091 			}
3092 
3093 			if (txbuff->last_frag) {
3094 				dev_kfree_skb_any(txbuff->skb);
3095 				txbuff->skb = NULL;
3096 			}
3097 
3098 			num_entries += txbuff->num_entries;
3099 
3100 			tx_pool->free_map[tx_pool->producer_index] = index;
3101 			tx_pool->producer_index =
3102 				(tx_pool->producer_index + 1) %
3103 					tx_pool->num_buffers;
3104 		}
3105 		/* remove tx_comp scrq*/
3106 		next->tx_comp.first = 0;
3107 
3108 		if (atomic_sub_return(num_entries, &scrq->used) <=
3109 		    (adapter->req_tx_entries_per_subcrq / 2) &&
3110 		    __netif_subqueue_stopped(adapter->netdev,
3111 					     scrq->pool_index)) {
3112 			netif_wake_subqueue(adapter->netdev, scrq->pool_index);
3113 			netdev_dbg(adapter->netdev, "Started queue %d\n",
3114 				   scrq->pool_index);
3115 		}
3116 	}
3117 
3118 	enable_scrq_irq(adapter, scrq);
3119 
3120 	if (pending_scrq(adapter, scrq)) {
3121 		disable_scrq_irq(adapter, scrq);
3122 		goto restart_loop;
3123 	}
3124 
3125 	return 0;
3126 }
3127 
3128 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
3129 {
3130 	struct ibmvnic_sub_crq_queue *scrq = instance;
3131 	struct ibmvnic_adapter *adapter = scrq->adapter;
3132 
3133 	disable_scrq_irq(adapter, scrq);
3134 	ibmvnic_complete_tx(adapter, scrq);
3135 
3136 	return IRQ_HANDLED;
3137 }
3138 
3139 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
3140 {
3141 	struct ibmvnic_sub_crq_queue *scrq = instance;
3142 	struct ibmvnic_adapter *adapter = scrq->adapter;
3143 
3144 	/* When booting a kdump kernel we can hit pending interrupts
3145 	 * prior to completing driver initialization.
3146 	 */
3147 	if (unlikely(adapter->state != VNIC_OPEN))
3148 		return IRQ_NONE;
3149 
3150 	adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
3151 
3152 	if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
3153 		disable_scrq_irq(adapter, scrq);
3154 		__napi_schedule(&adapter->napi[scrq->scrq_num]);
3155 	}
3156 
3157 	return IRQ_HANDLED;
3158 }
3159 
3160 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
3161 {
3162 	struct device *dev = &adapter->vdev->dev;
3163 	struct ibmvnic_sub_crq_queue *scrq;
3164 	int i = 0, j = 0;
3165 	int rc = 0;
3166 
3167 	for (i = 0; i < adapter->req_tx_queues; i++) {
3168 		netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
3169 			   i);
3170 		scrq = adapter->tx_scrq[i];
3171 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
3172 
3173 		if (!scrq->irq) {
3174 			rc = -EINVAL;
3175 			dev_err(dev, "Error mapping irq\n");
3176 			goto req_tx_irq_failed;
3177 		}
3178 
3179 		snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-tx%d",
3180 			 adapter->vdev->unit_address, i);
3181 		rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
3182 				 0, scrq->name, scrq);
3183 
3184 		if (rc) {
3185 			dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
3186 				scrq->irq, rc);
3187 			irq_dispose_mapping(scrq->irq);
3188 			goto req_tx_irq_failed;
3189 		}
3190 	}
3191 
3192 	for (i = 0; i < adapter->req_rx_queues; i++) {
3193 		netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
3194 			   i);
3195 		scrq = adapter->rx_scrq[i];
3196 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
3197 		if (!scrq->irq) {
3198 			rc = -EINVAL;
3199 			dev_err(dev, "Error mapping irq\n");
3200 			goto req_rx_irq_failed;
3201 		}
3202 		snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-rx%d",
3203 			 adapter->vdev->unit_address, i);
3204 		rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
3205 				 0, scrq->name, scrq);
3206 		if (rc) {
3207 			dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
3208 				scrq->irq, rc);
3209 			irq_dispose_mapping(scrq->irq);
3210 			goto req_rx_irq_failed;
3211 		}
3212 	}
3213 	return rc;
3214 
3215 req_rx_irq_failed:
3216 	for (j = 0; j < i; j++) {
3217 		free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
3218 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
3219 	}
3220 	i = adapter->req_tx_queues;
3221 req_tx_irq_failed:
3222 	for (j = 0; j < i; j++) {
3223 		free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
3224 		irq_dispose_mapping(adapter->tx_scrq[j]->irq);
3225 	}
3226 	release_sub_crqs(adapter, 1);
3227 	return rc;
3228 }
3229 
3230 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
3231 {
3232 	struct device *dev = &adapter->vdev->dev;
3233 	struct ibmvnic_sub_crq_queue **allqueues;
3234 	int registered_queues = 0;
3235 	int total_queues;
3236 	int more = 0;
3237 	int i;
3238 
3239 	total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
3240 
3241 	allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
3242 	if (!allqueues)
3243 		return -1;
3244 
3245 	for (i = 0; i < total_queues; i++) {
3246 		allqueues[i] = init_sub_crq_queue(adapter);
3247 		if (!allqueues[i]) {
3248 			dev_warn(dev, "Couldn't allocate all sub-crqs\n");
3249 			break;
3250 		}
3251 		registered_queues++;
3252 	}
3253 
3254 	/* Make sure we were able to register the minimum number of queues */
3255 	if (registered_queues <
3256 	    adapter->min_tx_queues + adapter->min_rx_queues) {
3257 		dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
3258 		goto tx_failed;
3259 	}
3260 
3261 	/* Distribute the failed allocated queues*/
3262 	for (i = 0; i < total_queues - registered_queues + more ; i++) {
3263 		netdev_dbg(adapter->netdev, "Reducing number of queues\n");
3264 		switch (i % 3) {
3265 		case 0:
3266 			if (adapter->req_rx_queues > adapter->min_rx_queues)
3267 				adapter->req_rx_queues--;
3268 			else
3269 				more++;
3270 			break;
3271 		case 1:
3272 			if (adapter->req_tx_queues > adapter->min_tx_queues)
3273 				adapter->req_tx_queues--;
3274 			else
3275 				more++;
3276 			break;
3277 		}
3278 	}
3279 
3280 	adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
3281 				   sizeof(*adapter->tx_scrq), GFP_KERNEL);
3282 	if (!adapter->tx_scrq)
3283 		goto tx_failed;
3284 
3285 	for (i = 0; i < adapter->req_tx_queues; i++) {
3286 		adapter->tx_scrq[i] = allqueues[i];
3287 		adapter->tx_scrq[i]->pool_index = i;
3288 		adapter->num_active_tx_scrqs++;
3289 	}
3290 
3291 	adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
3292 				   sizeof(*adapter->rx_scrq), GFP_KERNEL);
3293 	if (!adapter->rx_scrq)
3294 		goto rx_failed;
3295 
3296 	for (i = 0; i < adapter->req_rx_queues; i++) {
3297 		adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
3298 		adapter->rx_scrq[i]->scrq_num = i;
3299 		adapter->num_active_rx_scrqs++;
3300 	}
3301 
3302 	kfree(allqueues);
3303 	return 0;
3304 
3305 rx_failed:
3306 	kfree(adapter->tx_scrq);
3307 	adapter->tx_scrq = NULL;
3308 tx_failed:
3309 	for (i = 0; i < registered_queues; i++)
3310 		release_sub_crq_queue(adapter, allqueues[i], 1);
3311 	kfree(allqueues);
3312 	return -1;
3313 }
3314 
3315 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
3316 {
3317 	struct device *dev = &adapter->vdev->dev;
3318 	union ibmvnic_crq crq;
3319 	int max_entries;
3320 
3321 	if (!retry) {
3322 		/* Sub-CRQ entries are 32 byte long */
3323 		int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
3324 
3325 		if (adapter->min_tx_entries_per_subcrq > entries_page ||
3326 		    adapter->min_rx_add_entries_per_subcrq > entries_page) {
3327 			dev_err(dev, "Fatal, invalid entries per sub-crq\n");
3328 			return;
3329 		}
3330 
3331 		if (adapter->desired.mtu)
3332 			adapter->req_mtu = adapter->desired.mtu;
3333 		else
3334 			adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
3335 
3336 		if (!adapter->desired.tx_entries)
3337 			adapter->desired.tx_entries =
3338 					adapter->max_tx_entries_per_subcrq;
3339 		if (!adapter->desired.rx_entries)
3340 			adapter->desired.rx_entries =
3341 					adapter->max_rx_add_entries_per_subcrq;
3342 
3343 		max_entries = IBMVNIC_MAX_LTB_SIZE /
3344 			      (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
3345 
3346 		if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3347 			adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
3348 			adapter->desired.tx_entries = max_entries;
3349 		}
3350 
3351 		if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3352 			adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
3353 			adapter->desired.rx_entries = max_entries;
3354 		}
3355 
3356 		if (adapter->desired.tx_entries)
3357 			adapter->req_tx_entries_per_subcrq =
3358 					adapter->desired.tx_entries;
3359 		else
3360 			adapter->req_tx_entries_per_subcrq =
3361 					adapter->max_tx_entries_per_subcrq;
3362 
3363 		if (adapter->desired.rx_entries)
3364 			adapter->req_rx_add_entries_per_subcrq =
3365 					adapter->desired.rx_entries;
3366 		else
3367 			adapter->req_rx_add_entries_per_subcrq =
3368 					adapter->max_rx_add_entries_per_subcrq;
3369 
3370 		if (adapter->desired.tx_queues)
3371 			adapter->req_tx_queues =
3372 					adapter->desired.tx_queues;
3373 		else
3374 			adapter->req_tx_queues =
3375 					adapter->opt_tx_comp_sub_queues;
3376 
3377 		if (adapter->desired.rx_queues)
3378 			adapter->req_rx_queues =
3379 					adapter->desired.rx_queues;
3380 		else
3381 			adapter->req_rx_queues =
3382 					adapter->opt_rx_comp_queues;
3383 
3384 		adapter->req_rx_add_queues = adapter->max_rx_add_queues;
3385 	}
3386 
3387 	memset(&crq, 0, sizeof(crq));
3388 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
3389 	crq.request_capability.cmd = REQUEST_CAPABILITY;
3390 
3391 	crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
3392 	crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
3393 	atomic_inc(&adapter->running_cap_crqs);
3394 	ibmvnic_send_crq(adapter, &crq);
3395 
3396 	crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
3397 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
3398 	atomic_inc(&adapter->running_cap_crqs);
3399 	ibmvnic_send_crq(adapter, &crq);
3400 
3401 	crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
3402 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
3403 	atomic_inc(&adapter->running_cap_crqs);
3404 	ibmvnic_send_crq(adapter, &crq);
3405 
3406 	crq.request_capability.capability =
3407 	    cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3408 	crq.request_capability.number =
3409 	    cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3410 	atomic_inc(&adapter->running_cap_crqs);
3411 	ibmvnic_send_crq(adapter, &crq);
3412 
3413 	crq.request_capability.capability =
3414 	    cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3415 	crq.request_capability.number =
3416 	    cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3417 	atomic_inc(&adapter->running_cap_crqs);
3418 	ibmvnic_send_crq(adapter, &crq);
3419 
3420 	crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3421 	crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3422 	atomic_inc(&adapter->running_cap_crqs);
3423 	ibmvnic_send_crq(adapter, &crq);
3424 
3425 	if (adapter->netdev->flags & IFF_PROMISC) {
3426 		if (adapter->promisc_supported) {
3427 			crq.request_capability.capability =
3428 			    cpu_to_be16(PROMISC_REQUESTED);
3429 			crq.request_capability.number = cpu_to_be64(1);
3430 			atomic_inc(&adapter->running_cap_crqs);
3431 			ibmvnic_send_crq(adapter, &crq);
3432 		}
3433 	} else {
3434 		crq.request_capability.capability =
3435 		    cpu_to_be16(PROMISC_REQUESTED);
3436 		crq.request_capability.number = cpu_to_be64(0);
3437 		atomic_inc(&adapter->running_cap_crqs);
3438 		ibmvnic_send_crq(adapter, &crq);
3439 	}
3440 }
3441 
3442 static int pending_scrq(struct ibmvnic_adapter *adapter,
3443 			struct ibmvnic_sub_crq_queue *scrq)
3444 {
3445 	union sub_crq *entry = &scrq->msgs[scrq->cur];
3446 
3447 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
3448 		return 1;
3449 	else
3450 		return 0;
3451 }
3452 
3453 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3454 					struct ibmvnic_sub_crq_queue *scrq)
3455 {
3456 	union sub_crq *entry;
3457 	unsigned long flags;
3458 
3459 	spin_lock_irqsave(&scrq->lock, flags);
3460 	entry = &scrq->msgs[scrq->cur];
3461 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3462 		if (++scrq->cur == scrq->size)
3463 			scrq->cur = 0;
3464 	} else {
3465 		entry = NULL;
3466 	}
3467 	spin_unlock_irqrestore(&scrq->lock, flags);
3468 
3469 	return entry;
3470 }
3471 
3472 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3473 {
3474 	struct ibmvnic_crq_queue *queue = &adapter->crq;
3475 	union ibmvnic_crq *crq;
3476 
3477 	crq = &queue->msgs[queue->cur];
3478 	if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3479 		if (++queue->cur == queue->size)
3480 			queue->cur = 0;
3481 	} else {
3482 		crq = NULL;
3483 	}
3484 
3485 	return crq;
3486 }
3487 
3488 static void print_subcrq_error(struct device *dev, int rc, const char *func)
3489 {
3490 	switch (rc) {
3491 	case H_PARAMETER:
3492 		dev_warn_ratelimited(dev,
3493 				     "%s failed: Send request is malformed or adapter failover pending. (rc=%d)\n",
3494 				     func, rc);
3495 		break;
3496 	case H_CLOSED:
3497 		dev_warn_ratelimited(dev,
3498 				     "%s failed: Backing queue closed. Adapter is down or failover pending. (rc=%d)\n",
3499 				     func, rc);
3500 		break;
3501 	default:
3502 		dev_err_ratelimited(dev, "%s failed: (rc=%d)\n", func, rc);
3503 		break;
3504 	}
3505 }
3506 
3507 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
3508 		       union sub_crq *sub_crq)
3509 {
3510 	unsigned int ua = adapter->vdev->unit_address;
3511 	struct device *dev = &adapter->vdev->dev;
3512 	u64 *u64_crq = (u64 *)sub_crq;
3513 	int rc;
3514 
3515 	netdev_dbg(adapter->netdev,
3516 		   "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
3517 		   (unsigned long int)cpu_to_be64(remote_handle),
3518 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
3519 		   (unsigned long int)cpu_to_be64(u64_crq[1]),
3520 		   (unsigned long int)cpu_to_be64(u64_crq[2]),
3521 		   (unsigned long int)cpu_to_be64(u64_crq[3]));
3522 
3523 	/* Make sure the hypervisor sees the complete request */
3524 	mb();
3525 
3526 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
3527 				cpu_to_be64(remote_handle),
3528 				cpu_to_be64(u64_crq[0]),
3529 				cpu_to_be64(u64_crq[1]),
3530 				cpu_to_be64(u64_crq[2]),
3531 				cpu_to_be64(u64_crq[3]));
3532 
3533 	if (rc)
3534 		print_subcrq_error(dev, rc, __func__);
3535 
3536 	return rc;
3537 }
3538 
3539 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3540 				u64 remote_handle, u64 ioba, u64 num_entries)
3541 {
3542 	unsigned int ua = adapter->vdev->unit_address;
3543 	struct device *dev = &adapter->vdev->dev;
3544 	int rc;
3545 
3546 	/* Make sure the hypervisor sees the complete request */
3547 	mb();
3548 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3549 				cpu_to_be64(remote_handle),
3550 				ioba, num_entries);
3551 
3552 	if (rc)
3553 		print_subcrq_error(dev, rc, __func__);
3554 
3555 	return rc;
3556 }
3557 
3558 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3559 			    union ibmvnic_crq *crq)
3560 {
3561 	unsigned int ua = adapter->vdev->unit_address;
3562 	struct device *dev = &adapter->vdev->dev;
3563 	u64 *u64_crq = (u64 *)crq;
3564 	int rc;
3565 
3566 	netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3567 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
3568 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
3569 
3570 	if (!adapter->crq.active &&
3571 	    crq->generic.first != IBMVNIC_CRQ_INIT_CMD) {
3572 		dev_warn(dev, "Invalid request detected while CRQ is inactive, possible device state change during reset\n");
3573 		return -EINVAL;
3574 	}
3575 
3576 	/* Make sure the hypervisor sees the complete request */
3577 	mb();
3578 
3579 	rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3580 				cpu_to_be64(u64_crq[0]),
3581 				cpu_to_be64(u64_crq[1]));
3582 
3583 	if (rc) {
3584 		if (rc == H_CLOSED) {
3585 			dev_warn(dev, "CRQ Queue closed\n");
3586 			if (test_bit(0, &adapter->resetting))
3587 				ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3588 		}
3589 
3590 		dev_warn(dev, "Send error (rc=%d)\n", rc);
3591 	}
3592 
3593 	return rc;
3594 }
3595 
3596 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3597 {
3598 	union ibmvnic_crq crq;
3599 
3600 	memset(&crq, 0, sizeof(crq));
3601 	crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3602 	crq.generic.cmd = IBMVNIC_CRQ_INIT;
3603 	netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3604 
3605 	return ibmvnic_send_crq(adapter, &crq);
3606 }
3607 
3608 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3609 {
3610 	union ibmvnic_crq crq;
3611 
3612 	memset(&crq, 0, sizeof(crq));
3613 	crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3614 	crq.version_exchange.cmd = VERSION_EXCHANGE;
3615 	crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3616 
3617 	return ibmvnic_send_crq(adapter, &crq);
3618 }
3619 
3620 struct vnic_login_client_data {
3621 	u8	type;
3622 	__be16	len;
3623 	char	name[];
3624 } __packed;
3625 
3626 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3627 {
3628 	int len;
3629 
3630 	/* Calculate the amount of buffer space needed for the
3631 	 * vnic client data in the login buffer. There are four entries,
3632 	 * OS name, LPAR name, device name, and a null last entry.
3633 	 */
3634 	len = 4 * sizeof(struct vnic_login_client_data);
3635 	len += 6; /* "Linux" plus NULL */
3636 	len += strlen(utsname()->nodename) + 1;
3637 	len += strlen(adapter->netdev->name) + 1;
3638 
3639 	return len;
3640 }
3641 
3642 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3643 				 struct vnic_login_client_data *vlcd)
3644 {
3645 	const char *os_name = "Linux";
3646 	int len;
3647 
3648 	/* Type 1 - LPAR OS */
3649 	vlcd->type = 1;
3650 	len = strlen(os_name) + 1;
3651 	vlcd->len = cpu_to_be16(len);
3652 	strncpy(vlcd->name, os_name, len);
3653 	vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3654 
3655 	/* Type 2 - LPAR name */
3656 	vlcd->type = 2;
3657 	len = strlen(utsname()->nodename) + 1;
3658 	vlcd->len = cpu_to_be16(len);
3659 	strncpy(vlcd->name, utsname()->nodename, len);
3660 	vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3661 
3662 	/* Type 3 - device name */
3663 	vlcd->type = 3;
3664 	len = strlen(adapter->netdev->name) + 1;
3665 	vlcd->len = cpu_to_be16(len);
3666 	strncpy(vlcd->name, adapter->netdev->name, len);
3667 }
3668 
3669 static int send_login(struct ibmvnic_adapter *adapter)
3670 {
3671 	struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3672 	struct ibmvnic_login_buffer *login_buffer;
3673 	struct device *dev = &adapter->vdev->dev;
3674 	dma_addr_t rsp_buffer_token;
3675 	dma_addr_t buffer_token;
3676 	size_t rsp_buffer_size;
3677 	union ibmvnic_crq crq;
3678 	size_t buffer_size;
3679 	__be64 *tx_list_p;
3680 	__be64 *rx_list_p;
3681 	int client_data_len;
3682 	struct vnic_login_client_data *vlcd;
3683 	int i;
3684 
3685 	if (!adapter->tx_scrq || !adapter->rx_scrq) {
3686 		netdev_err(adapter->netdev,
3687 			   "RX or TX queues are not allocated, device login failed\n");
3688 		return -1;
3689 	}
3690 
3691 	release_login_rsp_buffer(adapter);
3692 	client_data_len = vnic_client_data_len(adapter);
3693 
3694 	buffer_size =
3695 	    sizeof(struct ibmvnic_login_buffer) +
3696 	    sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3697 	    client_data_len;
3698 
3699 	login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3700 	if (!login_buffer)
3701 		goto buf_alloc_failed;
3702 
3703 	buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3704 				      DMA_TO_DEVICE);
3705 	if (dma_mapping_error(dev, buffer_token)) {
3706 		dev_err(dev, "Couldn't map login buffer\n");
3707 		goto buf_map_failed;
3708 	}
3709 
3710 	rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3711 			  sizeof(u64) * adapter->req_tx_queues +
3712 			  sizeof(u64) * adapter->req_rx_queues +
3713 			  sizeof(u64) * adapter->req_rx_queues +
3714 			  sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3715 
3716 	login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3717 	if (!login_rsp_buffer)
3718 		goto buf_rsp_alloc_failed;
3719 
3720 	rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3721 					  rsp_buffer_size, DMA_FROM_DEVICE);
3722 	if (dma_mapping_error(dev, rsp_buffer_token)) {
3723 		dev_err(dev, "Couldn't map login rsp buffer\n");
3724 		goto buf_rsp_map_failed;
3725 	}
3726 
3727 	adapter->login_buf = login_buffer;
3728 	adapter->login_buf_token = buffer_token;
3729 	adapter->login_buf_sz = buffer_size;
3730 	adapter->login_rsp_buf = login_rsp_buffer;
3731 	adapter->login_rsp_buf_token = rsp_buffer_token;
3732 	adapter->login_rsp_buf_sz = rsp_buffer_size;
3733 
3734 	login_buffer->len = cpu_to_be32(buffer_size);
3735 	login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3736 	login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3737 	login_buffer->off_txcomp_subcrqs =
3738 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3739 	login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3740 	login_buffer->off_rxcomp_subcrqs =
3741 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3742 			sizeof(u64) * adapter->req_tx_queues);
3743 	login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3744 	login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3745 
3746 	tx_list_p = (__be64 *)((char *)login_buffer +
3747 				      sizeof(struct ibmvnic_login_buffer));
3748 	rx_list_p = (__be64 *)((char *)login_buffer +
3749 				      sizeof(struct ibmvnic_login_buffer) +
3750 				      sizeof(u64) * adapter->req_tx_queues);
3751 
3752 	for (i = 0; i < adapter->req_tx_queues; i++) {
3753 		if (adapter->tx_scrq[i]) {
3754 			tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3755 						   crq_num);
3756 		}
3757 	}
3758 
3759 	for (i = 0; i < adapter->req_rx_queues; i++) {
3760 		if (adapter->rx_scrq[i]) {
3761 			rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3762 						   crq_num);
3763 		}
3764 	}
3765 
3766 	/* Insert vNIC login client data */
3767 	vlcd = (struct vnic_login_client_data *)
3768 		((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3769 	login_buffer->client_data_offset =
3770 			cpu_to_be32((char *)vlcd - (char *)login_buffer);
3771 	login_buffer->client_data_len = cpu_to_be32(client_data_len);
3772 
3773 	vnic_add_client_data(adapter, vlcd);
3774 
3775 	netdev_dbg(adapter->netdev, "Login Buffer:\n");
3776 	for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3777 		netdev_dbg(adapter->netdev, "%016lx\n",
3778 			   ((unsigned long int *)(adapter->login_buf))[i]);
3779 	}
3780 
3781 	memset(&crq, 0, sizeof(crq));
3782 	crq.login.first = IBMVNIC_CRQ_CMD;
3783 	crq.login.cmd = LOGIN;
3784 	crq.login.ioba = cpu_to_be32(buffer_token);
3785 	crq.login.len = cpu_to_be32(buffer_size);
3786 	ibmvnic_send_crq(adapter, &crq);
3787 
3788 	return 0;
3789 
3790 buf_rsp_map_failed:
3791 	kfree(login_rsp_buffer);
3792 buf_rsp_alloc_failed:
3793 	dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3794 buf_map_failed:
3795 	kfree(login_buffer);
3796 buf_alloc_failed:
3797 	return -1;
3798 }
3799 
3800 static int send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3801 			    u32 len, u8 map_id)
3802 {
3803 	union ibmvnic_crq crq;
3804 
3805 	memset(&crq, 0, sizeof(crq));
3806 	crq.request_map.first = IBMVNIC_CRQ_CMD;
3807 	crq.request_map.cmd = REQUEST_MAP;
3808 	crq.request_map.map_id = map_id;
3809 	crq.request_map.ioba = cpu_to_be32(addr);
3810 	crq.request_map.len = cpu_to_be32(len);
3811 	return ibmvnic_send_crq(adapter, &crq);
3812 }
3813 
3814 static int send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3815 {
3816 	union ibmvnic_crq crq;
3817 
3818 	memset(&crq, 0, sizeof(crq));
3819 	crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3820 	crq.request_unmap.cmd = REQUEST_UNMAP;
3821 	crq.request_unmap.map_id = map_id;
3822 	return ibmvnic_send_crq(adapter, &crq);
3823 }
3824 
3825 static void send_map_query(struct ibmvnic_adapter *adapter)
3826 {
3827 	union ibmvnic_crq crq;
3828 
3829 	memset(&crq, 0, sizeof(crq));
3830 	crq.query_map.first = IBMVNIC_CRQ_CMD;
3831 	crq.query_map.cmd = QUERY_MAP;
3832 	ibmvnic_send_crq(adapter, &crq);
3833 }
3834 
3835 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3836 static void send_cap_queries(struct ibmvnic_adapter *adapter)
3837 {
3838 	union ibmvnic_crq crq;
3839 
3840 	atomic_set(&adapter->running_cap_crqs, 0);
3841 	memset(&crq, 0, sizeof(crq));
3842 	crq.query_capability.first = IBMVNIC_CRQ_CMD;
3843 	crq.query_capability.cmd = QUERY_CAPABILITY;
3844 
3845 	crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3846 	atomic_inc(&adapter->running_cap_crqs);
3847 	ibmvnic_send_crq(adapter, &crq);
3848 
3849 	crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3850 	atomic_inc(&adapter->running_cap_crqs);
3851 	ibmvnic_send_crq(adapter, &crq);
3852 
3853 	crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3854 	atomic_inc(&adapter->running_cap_crqs);
3855 	ibmvnic_send_crq(adapter, &crq);
3856 
3857 	crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3858 	atomic_inc(&adapter->running_cap_crqs);
3859 	ibmvnic_send_crq(adapter, &crq);
3860 
3861 	crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3862 	atomic_inc(&adapter->running_cap_crqs);
3863 	ibmvnic_send_crq(adapter, &crq);
3864 
3865 	crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3866 	atomic_inc(&adapter->running_cap_crqs);
3867 	ibmvnic_send_crq(adapter, &crq);
3868 
3869 	crq.query_capability.capability =
3870 	    cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3871 	atomic_inc(&adapter->running_cap_crqs);
3872 	ibmvnic_send_crq(adapter, &crq);
3873 
3874 	crq.query_capability.capability =
3875 	    cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3876 	atomic_inc(&adapter->running_cap_crqs);
3877 	ibmvnic_send_crq(adapter, &crq);
3878 
3879 	crq.query_capability.capability =
3880 	    cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3881 	atomic_inc(&adapter->running_cap_crqs);
3882 	ibmvnic_send_crq(adapter, &crq);
3883 
3884 	crq.query_capability.capability =
3885 	    cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3886 	atomic_inc(&adapter->running_cap_crqs);
3887 	ibmvnic_send_crq(adapter, &crq);
3888 
3889 	crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3890 	atomic_inc(&adapter->running_cap_crqs);
3891 	ibmvnic_send_crq(adapter, &crq);
3892 
3893 	crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3894 	atomic_inc(&adapter->running_cap_crqs);
3895 	ibmvnic_send_crq(adapter, &crq);
3896 
3897 	crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3898 	atomic_inc(&adapter->running_cap_crqs);
3899 	ibmvnic_send_crq(adapter, &crq);
3900 
3901 	crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3902 	atomic_inc(&adapter->running_cap_crqs);
3903 	ibmvnic_send_crq(adapter, &crq);
3904 
3905 	crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3906 	atomic_inc(&adapter->running_cap_crqs);
3907 	ibmvnic_send_crq(adapter, &crq);
3908 
3909 	crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3910 	atomic_inc(&adapter->running_cap_crqs);
3911 	ibmvnic_send_crq(adapter, &crq);
3912 
3913 	crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3914 	atomic_inc(&adapter->running_cap_crqs);
3915 	ibmvnic_send_crq(adapter, &crq);
3916 
3917 	crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3918 	atomic_inc(&adapter->running_cap_crqs);
3919 	ibmvnic_send_crq(adapter, &crq);
3920 
3921 	crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3922 	atomic_inc(&adapter->running_cap_crqs);
3923 	ibmvnic_send_crq(adapter, &crq);
3924 
3925 	crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3926 	atomic_inc(&adapter->running_cap_crqs);
3927 	ibmvnic_send_crq(adapter, &crq);
3928 
3929 	crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3930 	atomic_inc(&adapter->running_cap_crqs);
3931 	ibmvnic_send_crq(adapter, &crq);
3932 
3933 	crq.query_capability.capability =
3934 			cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3935 	atomic_inc(&adapter->running_cap_crqs);
3936 	ibmvnic_send_crq(adapter, &crq);
3937 
3938 	crq.query_capability.capability =
3939 			cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3940 	atomic_inc(&adapter->running_cap_crqs);
3941 	ibmvnic_send_crq(adapter, &crq);
3942 
3943 	crq.query_capability.capability =
3944 			cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3945 	atomic_inc(&adapter->running_cap_crqs);
3946 	ibmvnic_send_crq(adapter, &crq);
3947 
3948 	crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3949 	atomic_inc(&adapter->running_cap_crqs);
3950 	ibmvnic_send_crq(adapter, &crq);
3951 }
3952 
3953 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
3954 				struct ibmvnic_adapter *adapter)
3955 {
3956 	struct device *dev = &adapter->vdev->dev;
3957 
3958 	if (crq->get_vpd_size_rsp.rc.code) {
3959 		dev_err(dev, "Error retrieving VPD size, rc=%x\n",
3960 			crq->get_vpd_size_rsp.rc.code);
3961 		complete(&adapter->fw_done);
3962 		return;
3963 	}
3964 
3965 	adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
3966 	complete(&adapter->fw_done);
3967 }
3968 
3969 static void handle_vpd_rsp(union ibmvnic_crq *crq,
3970 			   struct ibmvnic_adapter *adapter)
3971 {
3972 	struct device *dev = &adapter->vdev->dev;
3973 	unsigned char *substr = NULL;
3974 	u8 fw_level_len = 0;
3975 
3976 	memset(adapter->fw_version, 0, 32);
3977 
3978 	dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
3979 			 DMA_FROM_DEVICE);
3980 
3981 	if (crq->get_vpd_rsp.rc.code) {
3982 		dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
3983 			crq->get_vpd_rsp.rc.code);
3984 		goto complete;
3985 	}
3986 
3987 	/* get the position of the firmware version info
3988 	 * located after the ASCII 'RM' substring in the buffer
3989 	 */
3990 	substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
3991 	if (!substr) {
3992 		dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
3993 		goto complete;
3994 	}
3995 
3996 	/* get length of firmware level ASCII substring */
3997 	if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
3998 		fw_level_len = *(substr + 2);
3999 	} else {
4000 		dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
4001 		goto complete;
4002 	}
4003 
4004 	/* copy firmware version string from vpd into adapter */
4005 	if ((substr + 3 + fw_level_len) <
4006 	    (adapter->vpd->buff + adapter->vpd->len)) {
4007 		strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
4008 	} else {
4009 		dev_info(dev, "FW substr extrapolated VPD buff\n");
4010 	}
4011 
4012 complete:
4013 	if (adapter->fw_version[0] == '\0')
4014 		strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
4015 	complete(&adapter->fw_done);
4016 }
4017 
4018 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
4019 {
4020 	struct device *dev = &adapter->vdev->dev;
4021 	struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
4022 	netdev_features_t old_hw_features = 0;
4023 	union ibmvnic_crq crq;
4024 	int i;
4025 
4026 	dma_unmap_single(dev, adapter->ip_offload_tok,
4027 			 sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
4028 
4029 	netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
4030 	for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
4031 		netdev_dbg(adapter->netdev, "%016lx\n",
4032 			   ((unsigned long int *)(buf))[i]);
4033 
4034 	netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
4035 	netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
4036 	netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
4037 		   buf->tcp_ipv4_chksum);
4038 	netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
4039 		   buf->tcp_ipv6_chksum);
4040 	netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
4041 		   buf->udp_ipv4_chksum);
4042 	netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
4043 		   buf->udp_ipv6_chksum);
4044 	netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
4045 		   buf->large_tx_ipv4);
4046 	netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
4047 		   buf->large_tx_ipv6);
4048 	netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
4049 		   buf->large_rx_ipv4);
4050 	netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
4051 		   buf->large_rx_ipv6);
4052 	netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
4053 		   buf->max_ipv4_header_size);
4054 	netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
4055 		   buf->max_ipv6_header_size);
4056 	netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
4057 		   buf->max_tcp_header_size);
4058 	netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
4059 		   buf->max_udp_header_size);
4060 	netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
4061 		   buf->max_large_tx_size);
4062 	netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
4063 		   buf->max_large_rx_size);
4064 	netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
4065 		   buf->ipv6_extension_header);
4066 	netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
4067 		   buf->tcp_pseudosum_req);
4068 	netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
4069 		   buf->num_ipv6_ext_headers);
4070 	netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
4071 		   buf->off_ipv6_ext_headers);
4072 
4073 	adapter->ip_offload_ctrl_tok =
4074 	    dma_map_single(dev, &adapter->ip_offload_ctrl,
4075 			   sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
4076 
4077 	if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
4078 		dev_err(dev, "Couldn't map ip offload control buffer\n");
4079 		return;
4080 	}
4081 
4082 	adapter->ip_offload_ctrl.len =
4083 	    cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
4084 	adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
4085 	adapter->ip_offload_ctrl.ipv4_chksum = buf->ipv4_chksum;
4086 	adapter->ip_offload_ctrl.ipv6_chksum = buf->ipv6_chksum;
4087 	adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
4088 	adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
4089 	adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
4090 	adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
4091 	adapter->ip_offload_ctrl.large_tx_ipv4 = buf->large_tx_ipv4;
4092 	adapter->ip_offload_ctrl.large_tx_ipv6 = buf->large_tx_ipv6;
4093 
4094 	/* large_rx disabled for now, additional features needed */
4095 	adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
4096 	adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
4097 
4098 	if (adapter->state != VNIC_PROBING) {
4099 		old_hw_features = adapter->netdev->hw_features;
4100 		adapter->netdev->hw_features = 0;
4101 	}
4102 
4103 	adapter->netdev->hw_features = NETIF_F_SG | NETIF_F_GSO | NETIF_F_GRO;
4104 
4105 	if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
4106 		adapter->netdev->hw_features |= NETIF_F_IP_CSUM;
4107 
4108 	if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
4109 		adapter->netdev->hw_features |= NETIF_F_IPV6_CSUM;
4110 
4111 	if ((adapter->netdev->features &
4112 	    (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
4113 		adapter->netdev->hw_features |= NETIF_F_RXCSUM;
4114 
4115 	if (buf->large_tx_ipv4)
4116 		adapter->netdev->hw_features |= NETIF_F_TSO;
4117 	if (buf->large_tx_ipv6)
4118 		adapter->netdev->hw_features |= NETIF_F_TSO6;
4119 
4120 	if (adapter->state == VNIC_PROBING) {
4121 		adapter->netdev->features |= adapter->netdev->hw_features;
4122 	} else if (old_hw_features != adapter->netdev->hw_features) {
4123 		netdev_features_t tmp = 0;
4124 
4125 		/* disable features no longer supported */
4126 		adapter->netdev->features &= adapter->netdev->hw_features;
4127 		/* turn on features now supported if previously enabled */
4128 		tmp = (old_hw_features ^ adapter->netdev->hw_features) &
4129 			adapter->netdev->hw_features;
4130 		adapter->netdev->features |=
4131 				tmp & adapter->netdev->wanted_features;
4132 	}
4133 
4134 	memset(&crq, 0, sizeof(crq));
4135 	crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
4136 	crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
4137 	crq.control_ip_offload.len =
4138 	    cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
4139 	crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
4140 	ibmvnic_send_crq(adapter, &crq);
4141 }
4142 
4143 static const char *ibmvnic_fw_err_cause(u16 cause)
4144 {
4145 	switch (cause) {
4146 	case ADAPTER_PROBLEM:
4147 		return "adapter problem";
4148 	case BUS_PROBLEM:
4149 		return "bus problem";
4150 	case FW_PROBLEM:
4151 		return "firmware problem";
4152 	case DD_PROBLEM:
4153 		return "device driver problem";
4154 	case EEH_RECOVERY:
4155 		return "EEH recovery";
4156 	case FW_UPDATED:
4157 		return "firmware updated";
4158 	case LOW_MEMORY:
4159 		return "low Memory";
4160 	default:
4161 		return "unknown";
4162 	}
4163 }
4164 
4165 static void handle_error_indication(union ibmvnic_crq *crq,
4166 				    struct ibmvnic_adapter *adapter)
4167 {
4168 	struct device *dev = &adapter->vdev->dev;
4169 	u16 cause;
4170 
4171 	cause = be16_to_cpu(crq->error_indication.error_cause);
4172 
4173 	dev_warn_ratelimited(dev,
4174 			     "Firmware reports %serror, cause: %s. Starting recovery...\n",
4175 			     crq->error_indication.flags
4176 				& IBMVNIC_FATAL_ERROR ? "FATAL " : "",
4177 			     ibmvnic_fw_err_cause(cause));
4178 
4179 	if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
4180 		ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4181 	else
4182 		ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
4183 }
4184 
4185 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
4186 				 struct ibmvnic_adapter *adapter)
4187 {
4188 	struct net_device *netdev = adapter->netdev;
4189 	struct device *dev = &adapter->vdev->dev;
4190 	long rc;
4191 
4192 	rc = crq->change_mac_addr_rsp.rc.code;
4193 	if (rc) {
4194 		dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
4195 		goto out;
4196 	}
4197 	ether_addr_copy(netdev->dev_addr,
4198 			&crq->change_mac_addr_rsp.mac_addr[0]);
4199 out:
4200 	complete(&adapter->fw_done);
4201 	return rc;
4202 }
4203 
4204 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
4205 				   struct ibmvnic_adapter *adapter)
4206 {
4207 	struct device *dev = &adapter->vdev->dev;
4208 	u64 *req_value;
4209 	char *name;
4210 
4211 	atomic_dec(&adapter->running_cap_crqs);
4212 	switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
4213 	case REQ_TX_QUEUES:
4214 		req_value = &adapter->req_tx_queues;
4215 		name = "tx";
4216 		break;
4217 	case REQ_RX_QUEUES:
4218 		req_value = &adapter->req_rx_queues;
4219 		name = "rx";
4220 		break;
4221 	case REQ_RX_ADD_QUEUES:
4222 		req_value = &adapter->req_rx_add_queues;
4223 		name = "rx_add";
4224 		break;
4225 	case REQ_TX_ENTRIES_PER_SUBCRQ:
4226 		req_value = &adapter->req_tx_entries_per_subcrq;
4227 		name = "tx_entries_per_subcrq";
4228 		break;
4229 	case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
4230 		req_value = &adapter->req_rx_add_entries_per_subcrq;
4231 		name = "rx_add_entries_per_subcrq";
4232 		break;
4233 	case REQ_MTU:
4234 		req_value = &adapter->req_mtu;
4235 		name = "mtu";
4236 		break;
4237 	case PROMISC_REQUESTED:
4238 		req_value = &adapter->promisc;
4239 		name = "promisc";
4240 		break;
4241 	default:
4242 		dev_err(dev, "Got invalid cap request rsp %d\n",
4243 			crq->request_capability.capability);
4244 		return;
4245 	}
4246 
4247 	switch (crq->request_capability_rsp.rc.code) {
4248 	case SUCCESS:
4249 		break;
4250 	case PARTIALSUCCESS:
4251 		dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
4252 			 *req_value,
4253 			 (long int)be64_to_cpu(crq->request_capability_rsp.
4254 					       number), name);
4255 
4256 		if (be16_to_cpu(crq->request_capability_rsp.capability) ==
4257 		    REQ_MTU) {
4258 			pr_err("mtu of %llu is not supported. Reverting.\n",
4259 			       *req_value);
4260 			*req_value = adapter->fallback.mtu;
4261 		} else {
4262 			*req_value =
4263 				be64_to_cpu(crq->request_capability_rsp.number);
4264 		}
4265 
4266 		ibmvnic_send_req_caps(adapter, 1);
4267 		return;
4268 	default:
4269 		dev_err(dev, "Error %d in request cap rsp\n",
4270 			crq->request_capability_rsp.rc.code);
4271 		return;
4272 	}
4273 
4274 	/* Done receiving requested capabilities, query IP offload support */
4275 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
4276 		union ibmvnic_crq newcrq;
4277 		int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
4278 		struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
4279 		    &adapter->ip_offload_buf;
4280 
4281 		adapter->wait_capability = false;
4282 		adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
4283 							 buf_sz,
4284 							 DMA_FROM_DEVICE);
4285 
4286 		if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
4287 			if (!firmware_has_feature(FW_FEATURE_CMO))
4288 				dev_err(dev, "Couldn't map offload buffer\n");
4289 			return;
4290 		}
4291 
4292 		memset(&newcrq, 0, sizeof(newcrq));
4293 		newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
4294 		newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
4295 		newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
4296 		newcrq.query_ip_offload.ioba =
4297 		    cpu_to_be32(adapter->ip_offload_tok);
4298 
4299 		ibmvnic_send_crq(adapter, &newcrq);
4300 	}
4301 }
4302 
4303 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
4304 			    struct ibmvnic_adapter *adapter)
4305 {
4306 	struct device *dev = &adapter->vdev->dev;
4307 	struct net_device *netdev = adapter->netdev;
4308 	struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
4309 	struct ibmvnic_login_buffer *login = adapter->login_buf;
4310 	int i;
4311 
4312 	dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
4313 			 DMA_TO_DEVICE);
4314 	dma_unmap_single(dev, adapter->login_rsp_buf_token,
4315 			 adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
4316 
4317 	/* If the number of queues requested can't be allocated by the
4318 	 * server, the login response will return with code 1. We will need
4319 	 * to resend the login buffer with fewer queues requested.
4320 	 */
4321 	if (login_rsp_crq->generic.rc.code) {
4322 		adapter->init_done_rc = login_rsp_crq->generic.rc.code;
4323 		complete(&adapter->init_done);
4324 		return 0;
4325 	}
4326 
4327 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
4328 
4329 	netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
4330 	for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
4331 		netdev_dbg(adapter->netdev, "%016lx\n",
4332 			   ((unsigned long int *)(adapter->login_rsp_buf))[i]);
4333 	}
4334 
4335 	/* Sanity checks */
4336 	if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
4337 	    (be32_to_cpu(login->num_rxcomp_subcrqs) *
4338 	     adapter->req_rx_add_queues !=
4339 	     be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
4340 		dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
4341 		ibmvnic_remove(adapter->vdev);
4342 		return -EIO;
4343 	}
4344 	release_login_buffer(adapter);
4345 	complete(&adapter->init_done);
4346 
4347 	return 0;
4348 }
4349 
4350 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
4351 				     struct ibmvnic_adapter *adapter)
4352 {
4353 	struct device *dev = &adapter->vdev->dev;
4354 	long rc;
4355 
4356 	rc = crq->request_unmap_rsp.rc.code;
4357 	if (rc)
4358 		dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4359 }
4360 
4361 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4362 				 struct ibmvnic_adapter *adapter)
4363 {
4364 	struct net_device *netdev = adapter->netdev;
4365 	struct device *dev = &adapter->vdev->dev;
4366 	long rc;
4367 
4368 	rc = crq->query_map_rsp.rc.code;
4369 	if (rc) {
4370 		dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4371 		return;
4372 	}
4373 	netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4374 		   crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4375 		   crq->query_map_rsp.free_pages);
4376 }
4377 
4378 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4379 				 struct ibmvnic_adapter *adapter)
4380 {
4381 	struct net_device *netdev = adapter->netdev;
4382 	struct device *dev = &adapter->vdev->dev;
4383 	long rc;
4384 
4385 	atomic_dec(&adapter->running_cap_crqs);
4386 	netdev_dbg(netdev, "Outstanding queries: %d\n",
4387 		   atomic_read(&adapter->running_cap_crqs));
4388 	rc = crq->query_capability.rc.code;
4389 	if (rc) {
4390 		dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4391 		goto out;
4392 	}
4393 
4394 	switch (be16_to_cpu(crq->query_capability.capability)) {
4395 	case MIN_TX_QUEUES:
4396 		adapter->min_tx_queues =
4397 		    be64_to_cpu(crq->query_capability.number);
4398 		netdev_dbg(netdev, "min_tx_queues = %lld\n",
4399 			   adapter->min_tx_queues);
4400 		break;
4401 	case MIN_RX_QUEUES:
4402 		adapter->min_rx_queues =
4403 		    be64_to_cpu(crq->query_capability.number);
4404 		netdev_dbg(netdev, "min_rx_queues = %lld\n",
4405 			   adapter->min_rx_queues);
4406 		break;
4407 	case MIN_RX_ADD_QUEUES:
4408 		adapter->min_rx_add_queues =
4409 		    be64_to_cpu(crq->query_capability.number);
4410 		netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4411 			   adapter->min_rx_add_queues);
4412 		break;
4413 	case MAX_TX_QUEUES:
4414 		adapter->max_tx_queues =
4415 		    be64_to_cpu(crq->query_capability.number);
4416 		netdev_dbg(netdev, "max_tx_queues = %lld\n",
4417 			   adapter->max_tx_queues);
4418 		break;
4419 	case MAX_RX_QUEUES:
4420 		adapter->max_rx_queues =
4421 		    be64_to_cpu(crq->query_capability.number);
4422 		netdev_dbg(netdev, "max_rx_queues = %lld\n",
4423 			   adapter->max_rx_queues);
4424 		break;
4425 	case MAX_RX_ADD_QUEUES:
4426 		adapter->max_rx_add_queues =
4427 		    be64_to_cpu(crq->query_capability.number);
4428 		netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4429 			   adapter->max_rx_add_queues);
4430 		break;
4431 	case MIN_TX_ENTRIES_PER_SUBCRQ:
4432 		adapter->min_tx_entries_per_subcrq =
4433 		    be64_to_cpu(crq->query_capability.number);
4434 		netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4435 			   adapter->min_tx_entries_per_subcrq);
4436 		break;
4437 	case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4438 		adapter->min_rx_add_entries_per_subcrq =
4439 		    be64_to_cpu(crq->query_capability.number);
4440 		netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4441 			   adapter->min_rx_add_entries_per_subcrq);
4442 		break;
4443 	case MAX_TX_ENTRIES_PER_SUBCRQ:
4444 		adapter->max_tx_entries_per_subcrq =
4445 		    be64_to_cpu(crq->query_capability.number);
4446 		netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4447 			   adapter->max_tx_entries_per_subcrq);
4448 		break;
4449 	case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4450 		adapter->max_rx_add_entries_per_subcrq =
4451 		    be64_to_cpu(crq->query_capability.number);
4452 		netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4453 			   adapter->max_rx_add_entries_per_subcrq);
4454 		break;
4455 	case TCP_IP_OFFLOAD:
4456 		adapter->tcp_ip_offload =
4457 		    be64_to_cpu(crq->query_capability.number);
4458 		netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4459 			   adapter->tcp_ip_offload);
4460 		break;
4461 	case PROMISC_SUPPORTED:
4462 		adapter->promisc_supported =
4463 		    be64_to_cpu(crq->query_capability.number);
4464 		netdev_dbg(netdev, "promisc_supported = %lld\n",
4465 			   adapter->promisc_supported);
4466 		break;
4467 	case MIN_MTU:
4468 		adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4469 		netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4470 		netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4471 		break;
4472 	case MAX_MTU:
4473 		adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4474 		netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4475 		netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4476 		break;
4477 	case MAX_MULTICAST_FILTERS:
4478 		adapter->max_multicast_filters =
4479 		    be64_to_cpu(crq->query_capability.number);
4480 		netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4481 			   adapter->max_multicast_filters);
4482 		break;
4483 	case VLAN_HEADER_INSERTION:
4484 		adapter->vlan_header_insertion =
4485 		    be64_to_cpu(crq->query_capability.number);
4486 		if (adapter->vlan_header_insertion)
4487 			netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4488 		netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4489 			   adapter->vlan_header_insertion);
4490 		break;
4491 	case RX_VLAN_HEADER_INSERTION:
4492 		adapter->rx_vlan_header_insertion =
4493 		    be64_to_cpu(crq->query_capability.number);
4494 		netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4495 			   adapter->rx_vlan_header_insertion);
4496 		break;
4497 	case MAX_TX_SG_ENTRIES:
4498 		adapter->max_tx_sg_entries =
4499 		    be64_to_cpu(crq->query_capability.number);
4500 		netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4501 			   adapter->max_tx_sg_entries);
4502 		break;
4503 	case RX_SG_SUPPORTED:
4504 		adapter->rx_sg_supported =
4505 		    be64_to_cpu(crq->query_capability.number);
4506 		netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4507 			   adapter->rx_sg_supported);
4508 		break;
4509 	case OPT_TX_COMP_SUB_QUEUES:
4510 		adapter->opt_tx_comp_sub_queues =
4511 		    be64_to_cpu(crq->query_capability.number);
4512 		netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4513 			   adapter->opt_tx_comp_sub_queues);
4514 		break;
4515 	case OPT_RX_COMP_QUEUES:
4516 		adapter->opt_rx_comp_queues =
4517 		    be64_to_cpu(crq->query_capability.number);
4518 		netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4519 			   adapter->opt_rx_comp_queues);
4520 		break;
4521 	case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4522 		adapter->opt_rx_bufadd_q_per_rx_comp_q =
4523 		    be64_to_cpu(crq->query_capability.number);
4524 		netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4525 			   adapter->opt_rx_bufadd_q_per_rx_comp_q);
4526 		break;
4527 	case OPT_TX_ENTRIES_PER_SUBCRQ:
4528 		adapter->opt_tx_entries_per_subcrq =
4529 		    be64_to_cpu(crq->query_capability.number);
4530 		netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4531 			   adapter->opt_tx_entries_per_subcrq);
4532 		break;
4533 	case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4534 		adapter->opt_rxba_entries_per_subcrq =
4535 		    be64_to_cpu(crq->query_capability.number);
4536 		netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4537 			   adapter->opt_rxba_entries_per_subcrq);
4538 		break;
4539 	case TX_RX_DESC_REQ:
4540 		adapter->tx_rx_desc_req = crq->query_capability.number;
4541 		netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4542 			   adapter->tx_rx_desc_req);
4543 		break;
4544 
4545 	default:
4546 		netdev_err(netdev, "Got invalid cap rsp %d\n",
4547 			   crq->query_capability.capability);
4548 	}
4549 
4550 out:
4551 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
4552 		adapter->wait_capability = false;
4553 		ibmvnic_send_req_caps(adapter, 0);
4554 	}
4555 }
4556 
4557 static int send_query_phys_parms(struct ibmvnic_adapter *adapter)
4558 {
4559 	union ibmvnic_crq crq;
4560 	int rc;
4561 
4562 	memset(&crq, 0, sizeof(crq));
4563 	crq.query_phys_parms.first = IBMVNIC_CRQ_CMD;
4564 	crq.query_phys_parms.cmd = QUERY_PHYS_PARMS;
4565 
4566 	mutex_lock(&adapter->fw_lock);
4567 	adapter->fw_done_rc = 0;
4568 	reinit_completion(&adapter->fw_done);
4569 
4570 	rc = ibmvnic_send_crq(adapter, &crq);
4571 	if (rc) {
4572 		mutex_unlock(&adapter->fw_lock);
4573 		return rc;
4574 	}
4575 
4576 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
4577 	if (rc) {
4578 		mutex_unlock(&adapter->fw_lock);
4579 		return rc;
4580 	}
4581 
4582 	mutex_unlock(&adapter->fw_lock);
4583 	return adapter->fw_done_rc ? -EIO : 0;
4584 }
4585 
4586 static int handle_query_phys_parms_rsp(union ibmvnic_crq *crq,
4587 				       struct ibmvnic_adapter *adapter)
4588 {
4589 	struct net_device *netdev = adapter->netdev;
4590 	int rc;
4591 	__be32 rspeed = cpu_to_be32(crq->query_phys_parms_rsp.speed);
4592 
4593 	rc = crq->query_phys_parms_rsp.rc.code;
4594 	if (rc) {
4595 		netdev_err(netdev, "Error %d in QUERY_PHYS_PARMS\n", rc);
4596 		return rc;
4597 	}
4598 	switch (rspeed) {
4599 	case IBMVNIC_10MBPS:
4600 		adapter->speed = SPEED_10;
4601 		break;
4602 	case IBMVNIC_100MBPS:
4603 		adapter->speed = SPEED_100;
4604 		break;
4605 	case IBMVNIC_1GBPS:
4606 		adapter->speed = SPEED_1000;
4607 		break;
4608 	case IBMVNIC_10GBP:
4609 		adapter->speed = SPEED_10000;
4610 		break;
4611 	case IBMVNIC_25GBPS:
4612 		adapter->speed = SPEED_25000;
4613 		break;
4614 	case IBMVNIC_40GBPS:
4615 		adapter->speed = SPEED_40000;
4616 		break;
4617 	case IBMVNIC_50GBPS:
4618 		adapter->speed = SPEED_50000;
4619 		break;
4620 	case IBMVNIC_100GBPS:
4621 		adapter->speed = SPEED_100000;
4622 		break;
4623 	default:
4624 		if (netif_carrier_ok(netdev))
4625 			netdev_warn(netdev, "Unknown speed 0x%08x\n", rspeed);
4626 		adapter->speed = SPEED_UNKNOWN;
4627 	}
4628 	if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_FULL_DUPLEX)
4629 		adapter->duplex = DUPLEX_FULL;
4630 	else if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_HALF_DUPLEX)
4631 		adapter->duplex = DUPLEX_HALF;
4632 	else
4633 		adapter->duplex = DUPLEX_UNKNOWN;
4634 
4635 	return rc;
4636 }
4637 
4638 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4639 			       struct ibmvnic_adapter *adapter)
4640 {
4641 	struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4642 	struct net_device *netdev = adapter->netdev;
4643 	struct device *dev = &adapter->vdev->dev;
4644 	u64 *u64_crq = (u64 *)crq;
4645 	long rc;
4646 
4647 	netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4648 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
4649 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
4650 	switch (gen_crq->first) {
4651 	case IBMVNIC_CRQ_INIT_RSP:
4652 		switch (gen_crq->cmd) {
4653 		case IBMVNIC_CRQ_INIT:
4654 			dev_info(dev, "Partner initialized\n");
4655 			adapter->from_passive_init = true;
4656 			adapter->failover_pending = false;
4657 			if (!completion_done(&adapter->init_done)) {
4658 				complete(&adapter->init_done);
4659 				adapter->init_done_rc = -EIO;
4660 			}
4661 			ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4662 			break;
4663 		case IBMVNIC_CRQ_INIT_COMPLETE:
4664 			dev_info(dev, "Partner initialization complete\n");
4665 			adapter->crq.active = true;
4666 			send_version_xchg(adapter);
4667 			break;
4668 		default:
4669 			dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4670 		}
4671 		return;
4672 	case IBMVNIC_CRQ_XPORT_EVENT:
4673 		netif_carrier_off(netdev);
4674 		adapter->crq.active = false;
4675 		/* terminate any thread waiting for a response
4676 		 * from the device
4677 		 */
4678 		if (!completion_done(&adapter->fw_done)) {
4679 			adapter->fw_done_rc = -EIO;
4680 			complete(&adapter->fw_done);
4681 		}
4682 		if (!completion_done(&adapter->stats_done))
4683 			complete(&adapter->stats_done);
4684 		if (test_bit(0, &adapter->resetting))
4685 			adapter->force_reset_recovery = true;
4686 		if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4687 			dev_info(dev, "Migrated, re-enabling adapter\n");
4688 			ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4689 		} else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4690 			dev_info(dev, "Backing device failover detected\n");
4691 			adapter->failover_pending = true;
4692 		} else {
4693 			/* The adapter lost the connection */
4694 			dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4695 				gen_crq->cmd);
4696 			ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4697 		}
4698 		return;
4699 	case IBMVNIC_CRQ_CMD_RSP:
4700 		break;
4701 	default:
4702 		dev_err(dev, "Got an invalid msg type 0x%02x\n",
4703 			gen_crq->first);
4704 		return;
4705 	}
4706 
4707 	switch (gen_crq->cmd) {
4708 	case VERSION_EXCHANGE_RSP:
4709 		rc = crq->version_exchange_rsp.rc.code;
4710 		if (rc) {
4711 			dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4712 			break;
4713 		}
4714 		ibmvnic_version =
4715 			    be16_to_cpu(crq->version_exchange_rsp.version);
4716 		dev_info(dev, "Partner protocol version is %d\n",
4717 			 ibmvnic_version);
4718 		send_cap_queries(adapter);
4719 		break;
4720 	case QUERY_CAPABILITY_RSP:
4721 		handle_query_cap_rsp(crq, adapter);
4722 		break;
4723 	case QUERY_MAP_RSP:
4724 		handle_query_map_rsp(crq, adapter);
4725 		break;
4726 	case REQUEST_MAP_RSP:
4727 		adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4728 		complete(&adapter->fw_done);
4729 		break;
4730 	case REQUEST_UNMAP_RSP:
4731 		handle_request_unmap_rsp(crq, adapter);
4732 		break;
4733 	case REQUEST_CAPABILITY_RSP:
4734 		handle_request_cap_rsp(crq, adapter);
4735 		break;
4736 	case LOGIN_RSP:
4737 		netdev_dbg(netdev, "Got Login Response\n");
4738 		handle_login_rsp(crq, adapter);
4739 		break;
4740 	case LOGICAL_LINK_STATE_RSP:
4741 		netdev_dbg(netdev,
4742 			   "Got Logical Link State Response, state: %d rc: %d\n",
4743 			   crq->logical_link_state_rsp.link_state,
4744 			   crq->logical_link_state_rsp.rc.code);
4745 		adapter->logical_link_state =
4746 		    crq->logical_link_state_rsp.link_state;
4747 		adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4748 		complete(&adapter->init_done);
4749 		break;
4750 	case LINK_STATE_INDICATION:
4751 		netdev_dbg(netdev, "Got Logical Link State Indication\n");
4752 		adapter->phys_link_state =
4753 		    crq->link_state_indication.phys_link_state;
4754 		adapter->logical_link_state =
4755 		    crq->link_state_indication.logical_link_state;
4756 		if (adapter->phys_link_state && adapter->logical_link_state)
4757 			netif_carrier_on(netdev);
4758 		else
4759 			netif_carrier_off(netdev);
4760 		break;
4761 	case CHANGE_MAC_ADDR_RSP:
4762 		netdev_dbg(netdev, "Got MAC address change Response\n");
4763 		adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4764 		break;
4765 	case ERROR_INDICATION:
4766 		netdev_dbg(netdev, "Got Error Indication\n");
4767 		handle_error_indication(crq, adapter);
4768 		break;
4769 	case REQUEST_STATISTICS_RSP:
4770 		netdev_dbg(netdev, "Got Statistics Response\n");
4771 		complete(&adapter->stats_done);
4772 		break;
4773 	case QUERY_IP_OFFLOAD_RSP:
4774 		netdev_dbg(netdev, "Got Query IP offload Response\n");
4775 		handle_query_ip_offload_rsp(adapter);
4776 		break;
4777 	case MULTICAST_CTRL_RSP:
4778 		netdev_dbg(netdev, "Got multicast control Response\n");
4779 		break;
4780 	case CONTROL_IP_OFFLOAD_RSP:
4781 		netdev_dbg(netdev, "Got Control IP offload Response\n");
4782 		dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4783 				 sizeof(adapter->ip_offload_ctrl),
4784 				 DMA_TO_DEVICE);
4785 		complete(&adapter->init_done);
4786 		break;
4787 	case COLLECT_FW_TRACE_RSP:
4788 		netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4789 		complete(&adapter->fw_done);
4790 		break;
4791 	case GET_VPD_SIZE_RSP:
4792 		handle_vpd_size_rsp(crq, adapter);
4793 		break;
4794 	case GET_VPD_RSP:
4795 		handle_vpd_rsp(crq, adapter);
4796 		break;
4797 	case QUERY_PHYS_PARMS_RSP:
4798 		adapter->fw_done_rc = handle_query_phys_parms_rsp(crq, adapter);
4799 		complete(&adapter->fw_done);
4800 		break;
4801 	default:
4802 		netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4803 			   gen_crq->cmd);
4804 	}
4805 }
4806 
4807 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4808 {
4809 	struct ibmvnic_adapter *adapter = instance;
4810 
4811 	tasklet_schedule(&adapter->tasklet);
4812 	return IRQ_HANDLED;
4813 }
4814 
4815 static void ibmvnic_tasklet(void *data)
4816 {
4817 	struct ibmvnic_adapter *adapter = data;
4818 	struct ibmvnic_crq_queue *queue = &adapter->crq;
4819 	union ibmvnic_crq *crq;
4820 	unsigned long flags;
4821 	bool done = false;
4822 
4823 	spin_lock_irqsave(&queue->lock, flags);
4824 	while (!done) {
4825 		/* Pull all the valid messages off the CRQ */
4826 		while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4827 			ibmvnic_handle_crq(crq, adapter);
4828 			crq->generic.first = 0;
4829 		}
4830 
4831 		/* remain in tasklet until all
4832 		 * capabilities responses are received
4833 		 */
4834 		if (!adapter->wait_capability)
4835 			done = true;
4836 	}
4837 	/* if capabilities CRQ's were sent in this tasklet, the following
4838 	 * tasklet must wait until all responses are received
4839 	 */
4840 	if (atomic_read(&adapter->running_cap_crqs) != 0)
4841 		adapter->wait_capability = true;
4842 	spin_unlock_irqrestore(&queue->lock, flags);
4843 }
4844 
4845 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4846 {
4847 	struct vio_dev *vdev = adapter->vdev;
4848 	int rc;
4849 
4850 	do {
4851 		rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4852 	} while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4853 
4854 	if (rc)
4855 		dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4856 
4857 	return rc;
4858 }
4859 
4860 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4861 {
4862 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4863 	struct device *dev = &adapter->vdev->dev;
4864 	struct vio_dev *vdev = adapter->vdev;
4865 	int rc;
4866 
4867 	/* Close the CRQ */
4868 	do {
4869 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4870 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4871 
4872 	/* Clean out the queue */
4873 	memset(crq->msgs, 0, PAGE_SIZE);
4874 	crq->cur = 0;
4875 	crq->active = false;
4876 
4877 	/* And re-open it again */
4878 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4879 				crq->msg_token, PAGE_SIZE);
4880 
4881 	if (rc == H_CLOSED)
4882 		/* Adapter is good, but other end is not ready */
4883 		dev_warn(dev, "Partner adapter not ready\n");
4884 	else if (rc != 0)
4885 		dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4886 
4887 	return rc;
4888 }
4889 
4890 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4891 {
4892 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4893 	struct vio_dev *vdev = adapter->vdev;
4894 	long rc;
4895 
4896 	if (!crq->msgs)
4897 		return;
4898 
4899 	netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4900 	free_irq(vdev->irq, adapter);
4901 	tasklet_kill(&adapter->tasklet);
4902 	do {
4903 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4904 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4905 
4906 	dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4907 			 DMA_BIDIRECTIONAL);
4908 	free_page((unsigned long)crq->msgs);
4909 	crq->msgs = NULL;
4910 	crq->active = false;
4911 }
4912 
4913 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4914 {
4915 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4916 	struct device *dev = &adapter->vdev->dev;
4917 	struct vio_dev *vdev = adapter->vdev;
4918 	int rc, retrc = -ENOMEM;
4919 
4920 	if (crq->msgs)
4921 		return 0;
4922 
4923 	crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4924 	/* Should we allocate more than one page? */
4925 
4926 	if (!crq->msgs)
4927 		return -ENOMEM;
4928 
4929 	crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4930 	crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4931 					DMA_BIDIRECTIONAL);
4932 	if (dma_mapping_error(dev, crq->msg_token))
4933 		goto map_failed;
4934 
4935 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4936 				crq->msg_token, PAGE_SIZE);
4937 
4938 	if (rc == H_RESOURCE)
4939 		/* maybe kexecing and resource is busy. try a reset */
4940 		rc = ibmvnic_reset_crq(adapter);
4941 	retrc = rc;
4942 
4943 	if (rc == H_CLOSED) {
4944 		dev_warn(dev, "Partner adapter not ready\n");
4945 	} else if (rc) {
4946 		dev_warn(dev, "Error %d opening adapter\n", rc);
4947 		goto reg_crq_failed;
4948 	}
4949 
4950 	retrc = 0;
4951 
4952 	tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
4953 		     (unsigned long)adapter);
4954 
4955 	netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
4956 	snprintf(crq->name, sizeof(crq->name), "ibmvnic-%x",
4957 		 adapter->vdev->unit_address);
4958 	rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, crq->name, adapter);
4959 	if (rc) {
4960 		dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
4961 			vdev->irq, rc);
4962 		goto req_irq_failed;
4963 	}
4964 
4965 	rc = vio_enable_interrupts(vdev);
4966 	if (rc) {
4967 		dev_err(dev, "Error %d enabling interrupts\n", rc);
4968 		goto req_irq_failed;
4969 	}
4970 
4971 	crq->cur = 0;
4972 	spin_lock_init(&crq->lock);
4973 
4974 	return retrc;
4975 
4976 req_irq_failed:
4977 	tasklet_kill(&adapter->tasklet);
4978 	do {
4979 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4980 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4981 reg_crq_failed:
4982 	dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
4983 map_failed:
4984 	free_page((unsigned long)crq->msgs);
4985 	crq->msgs = NULL;
4986 	return retrc;
4987 }
4988 
4989 static int ibmvnic_reset_init(struct ibmvnic_adapter *adapter)
4990 {
4991 	struct device *dev = &adapter->vdev->dev;
4992 	unsigned long timeout = msecs_to_jiffies(30000);
4993 	u64 old_num_rx_queues, old_num_tx_queues;
4994 	int rc;
4995 
4996 	adapter->from_passive_init = false;
4997 
4998 	old_num_rx_queues = adapter->req_rx_queues;
4999 	old_num_tx_queues = adapter->req_tx_queues;
5000 
5001 	reinit_completion(&adapter->init_done);
5002 	adapter->init_done_rc = 0;
5003 	ibmvnic_send_crq_init(adapter);
5004 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
5005 		dev_err(dev, "Initialization sequence timed out\n");
5006 		return -1;
5007 	}
5008 
5009 	if (adapter->init_done_rc) {
5010 		release_crq_queue(adapter);
5011 		return adapter->init_done_rc;
5012 	}
5013 
5014 	if (adapter->from_passive_init) {
5015 		adapter->state = VNIC_OPEN;
5016 		adapter->from_passive_init = false;
5017 		return -1;
5018 	}
5019 
5020 	if (test_bit(0, &adapter->resetting) && !adapter->wait_for_reset &&
5021 	    adapter->reset_reason != VNIC_RESET_MOBILITY) {
5022 		if (adapter->req_rx_queues != old_num_rx_queues ||
5023 		    adapter->req_tx_queues != old_num_tx_queues) {
5024 			release_sub_crqs(adapter, 0);
5025 			rc = init_sub_crqs(adapter);
5026 		} else {
5027 			rc = reset_sub_crq_queues(adapter);
5028 		}
5029 	} else {
5030 		rc = init_sub_crqs(adapter);
5031 	}
5032 
5033 	if (rc) {
5034 		dev_err(dev, "Initialization of sub crqs failed\n");
5035 		release_crq_queue(adapter);
5036 		return rc;
5037 	}
5038 
5039 	rc = init_sub_crq_irqs(adapter);
5040 	if (rc) {
5041 		dev_err(dev, "Failed to initialize sub crq irqs\n");
5042 		release_crq_queue(adapter);
5043 	}
5044 
5045 	return rc;
5046 }
5047 
5048 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
5049 {
5050 	struct device *dev = &adapter->vdev->dev;
5051 	unsigned long timeout = msecs_to_jiffies(30000);
5052 	int rc;
5053 
5054 	adapter->from_passive_init = false;
5055 
5056 	adapter->init_done_rc = 0;
5057 	ibmvnic_send_crq_init(adapter);
5058 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
5059 		dev_err(dev, "Initialization sequence timed out\n");
5060 		return -1;
5061 	}
5062 
5063 	if (adapter->init_done_rc) {
5064 		release_crq_queue(adapter);
5065 		return adapter->init_done_rc;
5066 	}
5067 
5068 	if (adapter->from_passive_init) {
5069 		adapter->state = VNIC_OPEN;
5070 		adapter->from_passive_init = false;
5071 		return -1;
5072 	}
5073 
5074 	rc = init_sub_crqs(adapter);
5075 	if (rc) {
5076 		dev_err(dev, "Initialization of sub crqs failed\n");
5077 		release_crq_queue(adapter);
5078 		return rc;
5079 	}
5080 
5081 	rc = init_sub_crq_irqs(adapter);
5082 	if (rc) {
5083 		dev_err(dev, "Failed to initialize sub crq irqs\n");
5084 		release_crq_queue(adapter);
5085 	}
5086 
5087 	return rc;
5088 }
5089 
5090 static struct device_attribute dev_attr_failover;
5091 
5092 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
5093 {
5094 	struct ibmvnic_adapter *adapter;
5095 	struct net_device *netdev;
5096 	unsigned char *mac_addr_p;
5097 	int rc;
5098 
5099 	dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
5100 		dev->unit_address);
5101 
5102 	mac_addr_p = (unsigned char *)vio_get_attribute(dev,
5103 							VETH_MAC_ADDR, NULL);
5104 	if (!mac_addr_p) {
5105 		dev_err(&dev->dev,
5106 			"(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
5107 			__FILE__, __LINE__);
5108 		return 0;
5109 	}
5110 
5111 	netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
5112 				   IBMVNIC_MAX_QUEUES);
5113 	if (!netdev)
5114 		return -ENOMEM;
5115 
5116 	adapter = netdev_priv(netdev);
5117 	adapter->state = VNIC_PROBING;
5118 	dev_set_drvdata(&dev->dev, netdev);
5119 	adapter->vdev = dev;
5120 	adapter->netdev = netdev;
5121 
5122 	ether_addr_copy(adapter->mac_addr, mac_addr_p);
5123 	ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
5124 	netdev->irq = dev->irq;
5125 	netdev->netdev_ops = &ibmvnic_netdev_ops;
5126 	netdev->ethtool_ops = &ibmvnic_ethtool_ops;
5127 	SET_NETDEV_DEV(netdev, &dev->dev);
5128 
5129 	spin_lock_init(&adapter->stats_lock);
5130 
5131 	INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
5132 	INIT_DELAYED_WORK(&adapter->ibmvnic_delayed_reset,
5133 			  __ibmvnic_delayed_reset);
5134 	INIT_LIST_HEAD(&adapter->rwi_list);
5135 	spin_lock_init(&adapter->rwi_lock);
5136 	spin_lock_init(&adapter->state_lock);
5137 	mutex_init(&adapter->fw_lock);
5138 	init_completion(&adapter->init_done);
5139 	init_completion(&adapter->fw_done);
5140 	init_completion(&adapter->reset_done);
5141 	init_completion(&adapter->stats_done);
5142 	clear_bit(0, &adapter->resetting);
5143 
5144 	do {
5145 		rc = init_crq_queue(adapter);
5146 		if (rc) {
5147 			dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
5148 				rc);
5149 			goto ibmvnic_init_fail;
5150 		}
5151 
5152 		rc = ibmvnic_init(adapter);
5153 		if (rc && rc != EAGAIN)
5154 			goto ibmvnic_init_fail;
5155 	} while (rc == EAGAIN);
5156 
5157 	rc = init_stats_buffers(adapter);
5158 	if (rc)
5159 		goto ibmvnic_init_fail;
5160 
5161 	rc = init_stats_token(adapter);
5162 	if (rc)
5163 		goto ibmvnic_stats_fail;
5164 
5165 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
5166 	netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
5167 	netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
5168 
5169 	rc = device_create_file(&dev->dev, &dev_attr_failover);
5170 	if (rc)
5171 		goto ibmvnic_dev_file_err;
5172 
5173 	netif_carrier_off(netdev);
5174 	rc = register_netdev(netdev);
5175 	if (rc) {
5176 		dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
5177 		goto ibmvnic_register_fail;
5178 	}
5179 	dev_info(&dev->dev, "ibmvnic registered\n");
5180 
5181 	adapter->state = VNIC_PROBED;
5182 
5183 	adapter->wait_for_reset = false;
5184 
5185 	return 0;
5186 
5187 ibmvnic_register_fail:
5188 	device_remove_file(&dev->dev, &dev_attr_failover);
5189 
5190 ibmvnic_dev_file_err:
5191 	release_stats_token(adapter);
5192 
5193 ibmvnic_stats_fail:
5194 	release_stats_buffers(adapter);
5195 
5196 ibmvnic_init_fail:
5197 	release_sub_crqs(adapter, 1);
5198 	release_crq_queue(adapter);
5199 	mutex_destroy(&adapter->fw_lock);
5200 	free_netdev(netdev);
5201 
5202 	return rc;
5203 }
5204 
5205 static int ibmvnic_remove(struct vio_dev *dev)
5206 {
5207 	struct net_device *netdev = dev_get_drvdata(&dev->dev);
5208 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5209 	unsigned long flags;
5210 
5211 	spin_lock_irqsave(&adapter->state_lock, flags);
5212 	if (adapter->state == VNIC_RESETTING) {
5213 		spin_unlock_irqrestore(&adapter->state_lock, flags);
5214 		return -EBUSY;
5215 	}
5216 
5217 	adapter->state = VNIC_REMOVING;
5218 	spin_unlock_irqrestore(&adapter->state_lock, flags);
5219 
5220 	flush_work(&adapter->ibmvnic_reset);
5221 	flush_delayed_work(&adapter->ibmvnic_delayed_reset);
5222 
5223 	rtnl_lock();
5224 	unregister_netdevice(netdev);
5225 
5226 	release_resources(adapter);
5227 	release_sub_crqs(adapter, 1);
5228 	release_crq_queue(adapter);
5229 
5230 	release_stats_token(adapter);
5231 	release_stats_buffers(adapter);
5232 
5233 	adapter->state = VNIC_REMOVED;
5234 
5235 	rtnl_unlock();
5236 	mutex_destroy(&adapter->fw_lock);
5237 	device_remove_file(&dev->dev, &dev_attr_failover);
5238 	free_netdev(netdev);
5239 	dev_set_drvdata(&dev->dev, NULL);
5240 
5241 	return 0;
5242 }
5243 
5244 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
5245 			      const char *buf, size_t count)
5246 {
5247 	struct net_device *netdev = dev_get_drvdata(dev);
5248 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5249 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
5250 	__be64 session_token;
5251 	long rc;
5252 
5253 	if (!sysfs_streq(buf, "1"))
5254 		return -EINVAL;
5255 
5256 	rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
5257 			 H_GET_SESSION_TOKEN, 0, 0, 0);
5258 	if (rc) {
5259 		netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
5260 			   rc);
5261 		return -EINVAL;
5262 	}
5263 
5264 	session_token = (__be64)retbuf[0];
5265 	netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
5266 		   be64_to_cpu(session_token));
5267 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
5268 				H_SESSION_ERR_DETECTED, session_token, 0, 0);
5269 	if (rc) {
5270 		netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
5271 			   rc);
5272 		return -EINVAL;
5273 	}
5274 
5275 	return count;
5276 }
5277 
5278 static DEVICE_ATTR_WO(failover);
5279 
5280 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
5281 {
5282 	struct net_device *netdev = dev_get_drvdata(&vdev->dev);
5283 	struct ibmvnic_adapter *adapter;
5284 	struct iommu_table *tbl;
5285 	unsigned long ret = 0;
5286 	int i;
5287 
5288 	tbl = get_iommu_table_base(&vdev->dev);
5289 
5290 	/* netdev inits at probe time along with the structures we need below*/
5291 	if (!netdev)
5292 		return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
5293 
5294 	adapter = netdev_priv(netdev);
5295 
5296 	ret += PAGE_SIZE; /* the crq message queue */
5297 	ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
5298 
5299 	for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
5300 		ret += 4 * PAGE_SIZE; /* the scrq message queue */
5301 
5302 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
5303 	     i++)
5304 		ret += adapter->rx_pool[i].size *
5305 		    IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
5306 
5307 	return ret;
5308 }
5309 
5310 static int ibmvnic_resume(struct device *dev)
5311 {
5312 	struct net_device *netdev = dev_get_drvdata(dev);
5313 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5314 
5315 	if (adapter->state != VNIC_OPEN)
5316 		return 0;
5317 
5318 	tasklet_schedule(&adapter->tasklet);
5319 
5320 	return 0;
5321 }
5322 
5323 static const struct vio_device_id ibmvnic_device_table[] = {
5324 	{"network", "IBM,vnic"},
5325 	{"", "" }
5326 };
5327 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
5328 
5329 static const struct dev_pm_ops ibmvnic_pm_ops = {
5330 	.resume = ibmvnic_resume
5331 };
5332 
5333 static struct vio_driver ibmvnic_driver = {
5334 	.id_table       = ibmvnic_device_table,
5335 	.probe          = ibmvnic_probe,
5336 	.remove         = ibmvnic_remove,
5337 	.get_desired_dma = ibmvnic_get_desired_dma,
5338 	.name		= ibmvnic_driver_name,
5339 	.pm		= &ibmvnic_pm_ops,
5340 };
5341 
5342 /* module functions */
5343 static int __init ibmvnic_module_init(void)
5344 {
5345 	pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
5346 		IBMVNIC_DRIVER_VERSION);
5347 
5348 	return vio_register_driver(&ibmvnic_driver);
5349 }
5350 
5351 static void __exit ibmvnic_module_exit(void)
5352 {
5353 	vio_unregister_driver(&ibmvnic_driver);
5354 }
5355 
5356 module_init(ibmvnic_module_init);
5357 module_exit(ibmvnic_module_exit);
5358