drivers/net/ethernet/ibm/ibmvnic.c

   1 /**************************************************************************/
   2 /*                                                                        */
   3 /*  IBM System i and System p Virtual NIC Device Driver                   */
   4 /*  Copyright (C) 2014 IBM Corp.                                          */
   5 /*  Santiago Leon (santi_leon@yahoo.com)                                  */
   6 /*  Thomas Falcon (tlfalcon@linux.vnet.ibm.com)                           */
   7 /*  John Allen (jallen@linux.vnet.ibm.com)                                */
   8 /*                                                                        */
   9 /*  This program is free software; you can redistribute it and/or modify  */
  10 /*  it under the terms of the GNU General Public License as published by  */
  11 /*  the Free Software Foundation; either version 2 of the License, or     */
  12 /*  (at your option) any later version.                                   */
  13 /*                                                                        */
  14 /*  This program is distributed in the hope that it will be useful,       */
  15 /*  but WITHOUT ANY WARRANTY; without even the implied warranty of        */
  16 /*  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the         */
  17 /*  GNU General Public License for more details.                          */
  18 /*                                                                        */
  19 /*  You should have received a copy of the GNU General Public License     */
  20 /*  along with this program.                                              */
  21 /*                                                                        */
  22 /* This module contains the implementation of a virtual ethernet device   */
  23 /* for use with IBM i/p Series LPAR Linux. It utilizes the logical LAN    */
  24 /* option of the RS/6000 Platform Architecture to interface with virtual  */
  25 /* ethernet NICs that are presented to the partition by the hypervisor.   */
  26 /*                                                                         */
  27 /* Messages are passed between the VNIC driver and the VNIC server using  */
  28 /* Command/Response Queues (CRQs) and sub CRQs (sCRQs). CRQs are used to  */
  29 /* issue and receive commands that initiate communication with the server */
  30 /* on driver initialization. Sub CRQs (sCRQs) are similar to CRQs, but    */
  31 /* are used by the driver to notify the server that a packet is           */
  32 /* ready for transmission or that a buffer has been added to receive a    */
  33 /* packet. Subsequently, sCRQs are used by the server to notify the       */
  34 /* driver that a packet transmission has been completed or that a packet  */
  35 /* has been received and placed in a waiting buffer.                      */
  36 /*                                                                        */
  37 /* In lieu of a more conventional "on-the-fly" DMA mapping strategy in    */
  38 /* which skbs are DMA mapped and immediately unmapped when the transmit   */
  39 /* or receive has been completed, the VNIC driver is required to use      */
  40 /* "long term mapping". This entails that large, continuous DMA mapped    */
  41 /* buffers are allocated on driver initialization and these buffers are   */
  42 /* then continuously reused to pass skbs to and from the VNIC server.     */
  43 /*                                                                        */
  44 /**************************************************************************/
  45
  46 #include <linux/module.h>
  47 #include <linux/moduleparam.h>
  48 #include <linux/types.h>
  49 #include <linux/errno.h>
  50 #include <linux/completion.h>
  51 #include <linux/ioport.h>
  52 #include <linux/dma-mapping.h>
  53 #include <linux/kernel.h>
  54 #include <linux/netdevice.h>
  55 #include <linux/etherdevice.h>
  56 #include <linux/skbuff.h>
  57 #include <linux/init.h>
  58 #include <linux/delay.h>
  59 #include <linux/mm.h>
  60 #include <linux/ethtool.h>
  61 #include <linux/proc_fs.h>
  62 #include <linux/if_arp.h>
  63 #include <linux/in.h>
  64 #include <linux/ip.h>
  65 #include <linux/ipv6.h>
  66 #include <linux/irq.h>
  67 #include <linux/kthread.h>
  68 #include <linux/seq_file.h>
  69 #include <linux/interrupt.h>
  70 #include <net/net_namespace.h>
  71 #include <asm/hvcall.h>
  72 #include <linux/atomic.h>
  73 #include <asm/vio.h>
  74 #include <asm/iommu.h>
  75 #include <linux/uaccess.h>
  76 #include <asm/firmware.h>
  77 #include <linux/workqueue.h>
  78 #include <linux/if_vlan.h>
  79 #include <linux/utsname.h>
  80
  81 #include "ibmvnic.h"
  82
  83 static const char ibmvnic_driver_name[] = "ibmvnic";
  84 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
  85
  86 MODULE_AUTHOR("Santiago Leon");
  87 MODULE_DESCRIPTION("IBM System i/p Virtual NIC Driver");
  88 MODULE_LICENSE("GPL");
  89 MODULE_VERSION(IBMVNIC_DRIVER_VERSION);
  90
  91 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
  92 static int ibmvnic_remove(struct vio_dev *);
  93 static void release_sub_crqs(struct ibmvnic_adapter *, bool);
  94 static int ibmvnic_reset_crq(struct ibmvnic_adapter *);
  95 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *);
  96 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *);
  97 static int ibmvnic_send_crq(struct ibmvnic_adapter *, union ibmvnic_crq *);
  98 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
  99                        union sub_crq *sub_crq);
 100 static int send_subcrq_indirect(struct ibmvnic_adapter *, u64, u64, u64);
 101 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
 102 static int enable_scrq_irq(struct ibmvnic_adapter *,
 103                            struct ibmvnic_sub_crq_queue *);
 104 static int disable_scrq_irq(struct ibmvnic_adapter *,
 105                             struct ibmvnic_sub_crq_queue *);
 106 static int pending_scrq(struct ibmvnic_adapter *,
 107                         struct ibmvnic_sub_crq_queue *);
 108 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
 109                                         struct ibmvnic_sub_crq_queue *);
 110 static int ibmvnic_poll(struct napi_struct *napi, int data);
 111 static void send_map_query(struct ibmvnic_adapter *adapter);
 112 static void send_request_map(struct ibmvnic_adapter *, dma_addr_t, __be32, u8);
 113 static void send_request_unmap(struct ibmvnic_adapter *, u8);
 114 static int send_login(struct ibmvnic_adapter *adapter);
 115 static void send_cap_queries(struct ibmvnic_adapter *adapter);
 116 static int init_sub_crqs(struct ibmvnic_adapter *);
 117 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
 118 static int ibmvnic_init(struct ibmvnic_adapter *);
 119 static void release_crq_queue(struct ibmvnic_adapter *);
 120 static int __ibmvnic_set_mac(struct net_device *netdev, struct sockaddr *p);
 121 static int init_crq_queue(struct ibmvnic_adapter *adapter);
 122
 123 struct ibmvnic_stat {
 124         char name[ETH_GSTRING_LEN];
 125         int offset;
 126 };
 127
 128 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
 129                              offsetof(struct ibmvnic_statistics, stat))
 130 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
 131
 132 static const struct ibmvnic_stat ibmvnic_stats[] = {
 133         {"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
 134         {"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
 135         {"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
 136         {"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
 137         {"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
 138         {"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
 139         {"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
 140         {"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
 141         {"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
 142         {"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
 143         {"align_errors", IBMVNIC_STAT_OFF(align_errors)},
 144         {"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
 145         {"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
 146         {"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
 147         {"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
 148         {"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
 149         {"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
 150         {"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
 151         {"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
 152         {"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
 153         {"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
 154         {"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
 155 };
 156
 157 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
 158                           unsigned long length, unsigned long *number,
 159                           unsigned long *irq)
 160 {
 161         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
 162         long rc;
 163
 164         rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
 165         *number = retbuf[0];
 166         *irq = retbuf[1];
 167
 168         return rc;
 169 }
 170
 171 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
 172                                 struct ibmvnic_long_term_buff *ltb, int size)
 173 {
 174         struct device *dev = &adapter->vdev->dev;
 175
 176         ltb->size = size;
 177         ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
 178                                        GFP_KERNEL);
 179
 180         if (!ltb->buff) {
 181                 dev_err(dev, "Couldn't alloc long term buffer\n");
 182                 return -ENOMEM;
 183         }
 184         ltb->map_id = adapter->map_id;
 185         adapter->map_id++;
 186
 187         init_completion(&adapter->fw_done);
 188         send_request_map(adapter, ltb->addr,
 189                          ltb->size, ltb->map_id);
 190         wait_for_completion(&adapter->fw_done);
 191
 192         if (adapter->fw_done_rc) {
 193                 dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
 194                         adapter->fw_done_rc);
 195                 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 196                 return -1;
 197         }
 198         return 0;
 199 }
 200
 201 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
 202                                 struct ibmvnic_long_term_buff *ltb)
 203 {
 204         struct device *dev = &adapter->vdev->dev;
 205
 206         if (!ltb->buff)
 207                 return;
 208
 209         if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
 210             adapter->reset_reason != VNIC_RESET_MOBILITY)
 211                 send_request_unmap(adapter, ltb->map_id);
 212         dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 213 }
 214
 215 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
 216                                 struct ibmvnic_long_term_buff *ltb)
 217 {
 218         memset(ltb->buff, 0, ltb->size);
 219
 220         init_completion(&adapter->fw_done);
 221         send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
 222         wait_for_completion(&adapter->fw_done);
 223
 224         if (adapter->fw_done_rc) {
 225                 dev_info(&adapter->vdev->dev,
 226                          "Reset failed, attempting to free and reallocate buffer\n");
 227                 free_long_term_buff(adapter, ltb);
 228                 return alloc_long_term_buff(adapter, ltb, ltb->size);
 229         }
 230         return 0;
 231 }
 232
 233 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
 234 {
 235         int i;
 236
 237         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 238              i++)
 239                 adapter->rx_pool[i].active = 0;
 240 }
 241
 242 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
 243                               struct ibmvnic_rx_pool *pool)
 244 {
 245         int count = pool->size - atomic_read(&pool->available);
 246         struct device *dev = &adapter->vdev->dev;
 247         int buffers_added = 0;
 248         unsigned long lpar_rc;
 249         union sub_crq sub_crq;
 250         struct sk_buff *skb;
 251         unsigned int offset;
 252         dma_addr_t dma_addr;
 253         unsigned char *dst;
 254         u64 *handle_array;
 255         int shift = 0;
 256         int index;
 257         int i;
 258
 259         if (!pool->active)
 260                 return;
 261
 262         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 263                                       be32_to_cpu(adapter->login_rsp_buf->
 264                                       off_rxadd_subcrqs));
 265
 266         for (i = 0; i < count; ++i) {
 267                 skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
 268                 if (!skb) {
 269                         dev_err(dev, "Couldn't replenish rx buff\n");
 270                         adapter->replenish_no_mem++;
 271                         break;
 272                 }
 273
 274                 index = pool->free_map[pool->next_free];
 275
 276                 if (pool->rx_buff[index].skb)
 277                         dev_err(dev, "Inconsistent free_map!\n");
 278
 279                 /* Copy the skb to the long term mapped DMA buffer */
 280                 offset = index * pool->buff_size;
 281                 dst = pool->long_term_buff.buff + offset;
 282                 memset(dst, 0, pool->buff_size);
 283                 dma_addr = pool->long_term_buff.addr + offset;
 284                 pool->rx_buff[index].data = dst;
 285
 286                 pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
 287                 pool->rx_buff[index].dma = dma_addr;
 288                 pool->rx_buff[index].skb = skb;
 289                 pool->rx_buff[index].pool_index = pool->index;
 290                 pool->rx_buff[index].size = pool->buff_size;
 291
 292                 memset(&sub_crq, 0, sizeof(sub_crq));
 293                 sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
 294                 sub_crq.rx_add.correlator =
 295                     cpu_to_be64((u64)&pool->rx_buff[index]);
 296                 sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
 297                 sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
 298
 299                 /* The length field of the sCRQ is defined to be 24 bits so the
 300                  * buffer size needs to be left shifted by a byte before it is
 301                  * converted to big endian to prevent the last byte from being
 302                  * truncated.
 303                  */
 304 #ifdef __LITTLE_ENDIAN__
 305                 shift = 8;
 306 #endif
 307                 sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
 308
 309                 lpar_rc = send_subcrq(adapter, handle_array[pool->index],
 310                                       &sub_crq);
 311                 if (lpar_rc != H_SUCCESS)
 312                         goto failure;
 313
 314                 buffers_added++;
 315                 adapter->replenish_add_buff_success++;
 316                 pool->next_free = (pool->next_free + 1) % pool->size;
 317         }
 318         atomic_add(buffers_added, &pool->available);
 319         return;
 320
 321 failure:
 322         dev_info(dev, "replenish pools failure\n");
 323         pool->free_map[pool->next_free] = index;
 324         pool->rx_buff[index].skb = NULL;
 325
 326         dev_kfree_skb_any(skb);
 327         adapter->replenish_add_buff_failure++;
 328         atomic_add(buffers_added, &pool->available);
 329
 330         if (lpar_rc == H_CLOSED || adapter->failover_pending) {
 331                 /* Disable buffer pool replenishment and report carrier off if
 332                  * queue is closed or pending failover.
 333                  * Firmware guarantees that a signal will be sent to the
 334                  * driver, triggering a reset.
 335                  */
 336                 deactivate_rx_pools(adapter);
 337                 netif_carrier_off(adapter->netdev);
 338         }
 339 }
 340
 341 static void replenish_pools(struct ibmvnic_adapter *adapter)
 342 {
 343         int i;
 344
 345         adapter->replenish_task_cycles++;
 346         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 347              i++) {
 348                 if (adapter->rx_pool[i].active)
 349                         replenish_rx_pool(adapter, &adapter->rx_pool[i]);
 350         }
 351 }
 352
 353 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
 354 {
 355         kfree(adapter->tx_stats_buffers);
 356         kfree(adapter->rx_stats_buffers);
 357         adapter->tx_stats_buffers = NULL;
 358         adapter->rx_stats_buffers = NULL;
 359 }
 360
 361 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
 362 {
 363         adapter->tx_stats_buffers =
 364                                 kcalloc(IBMVNIC_MAX_QUEUES,
 365                                         sizeof(struct ibmvnic_tx_queue_stats),
 366                                         GFP_KERNEL);
 367         if (!adapter->tx_stats_buffers)
 368                 return -ENOMEM;
 369
 370         adapter->rx_stats_buffers =
 371                                 kcalloc(IBMVNIC_MAX_QUEUES,
 372                                         sizeof(struct ibmvnic_rx_queue_stats),
 373                                         GFP_KERNEL);
 374         if (!adapter->rx_stats_buffers)
 375                 return -ENOMEM;
 376
 377         return 0;
 378 }
 379
 380 static void release_stats_token(struct ibmvnic_adapter *adapter)
 381 {
 382         struct device *dev = &adapter->vdev->dev;
 383
 384         if (!adapter->stats_token)
 385                 return;
 386
 387         dma_unmap_single(dev, adapter->stats_token,
 388                          sizeof(struct ibmvnic_statistics),
 389                          DMA_FROM_DEVICE);
 390         adapter->stats_token = 0;
 391 }
 392
 393 static int init_stats_token(struct ibmvnic_adapter *adapter)
 394 {
 395         struct device *dev = &adapter->vdev->dev;
 396         dma_addr_t stok;
 397
 398         stok = dma_map_single(dev, &adapter->stats,
 399                               sizeof(struct ibmvnic_statistics),
 400                               DMA_FROM_DEVICE);
 401         if (dma_mapping_error(dev, stok)) {
 402                 dev_err(dev, "Couldn't map stats buffer\n");
 403                 return -1;
 404         }
 405
 406         adapter->stats_token = stok;
 407         netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
 408         return 0;
 409 }
 410
 411 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
 412 {
 413         struct ibmvnic_rx_pool *rx_pool;
 414         int rx_scrqs;
 415         int i, j, rc;
 416         u64 *size_array;
 417
 418         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 419                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
 420
 421         rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 422         for (i = 0; i < rx_scrqs; i++) {
 423                 rx_pool = &adapter->rx_pool[i];
 424
 425                 netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
 426
 427                 if (rx_pool->buff_size != be64_to_cpu(size_array[i])) {
 428                         free_long_term_buff(adapter, &rx_pool->long_term_buff);
 429                         rx_pool->buff_size = be64_to_cpu(size_array[i]);
 430                         alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
 431                                              rx_pool->size *
 432                                              rx_pool->buff_size);
 433                 } else {
 434                         rc = reset_long_term_buff(adapter,
 435                                                   &rx_pool->long_term_buff);
 436                 }
 437
 438                 if (rc)
 439                         return rc;
 440
 441                 for (j = 0; j < rx_pool->size; j++)
 442                         rx_pool->free_map[j] = j;
 443
 444                 memset(rx_pool->rx_buff, 0,
 445                        rx_pool->size * sizeof(struct ibmvnic_rx_buff));
 446
 447                 atomic_set(&rx_pool->available, 0);
 448                 rx_pool->next_alloc = 0;
 449                 rx_pool->next_free = 0;
 450                 rx_pool->active = 1;
 451         }
 452
 453         return 0;
 454 }
 455
 456 static void release_rx_pools(struct ibmvnic_adapter *adapter)
 457 {
 458         struct ibmvnic_rx_pool *rx_pool;
 459         int i, j;
 460
 461         if (!adapter->rx_pool)
 462                 return;
 463
 464         for (i = 0; i < adapter->num_active_rx_pools; i++) {
 465                 rx_pool = &adapter->rx_pool[i];
 466
 467                 netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
 468
 469                 kfree(rx_pool->free_map);
 470                 free_long_term_buff(adapter, &rx_pool->long_term_buff);
 471
 472                 if (!rx_pool->rx_buff)
 473                         continue;
 474
 475                 for (j = 0; j < rx_pool->size; j++) {
 476                         if (rx_pool->rx_buff[j].skb) {
 477                                 dev_kfree_skb_any(rx_pool->rx_buff[i].skb);
 478                                 rx_pool->rx_buff[i].skb = NULL;
 479                         }
 480                 }
 481
 482                 kfree(rx_pool->rx_buff);
 483         }
 484
 485         kfree(adapter->rx_pool);
 486         adapter->rx_pool = NULL;
 487         adapter->num_active_rx_pools = 0;
 488 }
 489
 490 static int init_rx_pools(struct net_device *netdev)
 491 {
 492         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 493         struct device *dev = &adapter->vdev->dev;
 494         struct ibmvnic_rx_pool *rx_pool;
 495         int rxadd_subcrqs;
 496         u64 *size_array;
 497         int i, j;
 498
 499         rxadd_subcrqs =
 500                 be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 501         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 502                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
 503
 504         adapter->rx_pool = kcalloc(rxadd_subcrqs,
 505                                    sizeof(struct ibmvnic_rx_pool),
 506                                    GFP_KERNEL);
 507         if (!adapter->rx_pool) {
 508                 dev_err(dev, "Failed to allocate rx pools\n");
 509                 return -1;
 510         }
 511
 512         adapter->num_active_rx_pools = rxadd_subcrqs;
 513
 514         for (i = 0; i < rxadd_subcrqs; i++) {
 515                 rx_pool = &adapter->rx_pool[i];
 516
 517                 netdev_dbg(adapter->netdev,
 518                            "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
 519                            i, adapter->req_rx_add_entries_per_subcrq,
 520                            be64_to_cpu(size_array[i]));
 521
 522                 rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
 523                 rx_pool->index = i;
 524                 rx_pool->buff_size = be64_to_cpu(size_array[i]);
 525                 rx_pool->active = 1;
 526
 527                 rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
 528                                             GFP_KERNEL);
 529                 if (!rx_pool->free_map) {
 530                         release_rx_pools(adapter);
 531                         return -1;
 532                 }
 533
 534                 rx_pool->rx_buff = kcalloc(rx_pool->size,
 535                                            sizeof(struct ibmvnic_rx_buff),
 536                                            GFP_KERNEL);
 537                 if (!rx_pool->rx_buff) {
 538                         dev_err(dev, "Couldn't alloc rx buffers\n");
 539                         release_rx_pools(adapter);
 540                         return -1;
 541                 }
 542
 543                 if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
 544                                          rx_pool->size * rx_pool->buff_size)) {
 545                         release_rx_pools(adapter);
 546                         return -1;
 547                 }
 548
 549                 for (j = 0; j < rx_pool->size; ++j)
 550                         rx_pool->free_map[j] = j;
 551
 552                 atomic_set(&rx_pool->available, 0);
 553                 rx_pool->next_alloc = 0;
 554                 rx_pool->next_free = 0;
 555         }
 556
 557         return 0;
 558 }
 559
 560 static int reset_one_tx_pool(struct ibmvnic_adapter *adapter,
 561                              struct ibmvnic_tx_pool *tx_pool)
 562 {
 563         int rc, i;
 564
 565         rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
 566         if (rc)
 567                 return rc;
 568
 569         memset(tx_pool->tx_buff, 0,
 570                tx_pool->num_buffers *
 571                sizeof(struct ibmvnic_tx_buff));
 572
 573         for (i = 0; i < tx_pool->num_buffers; i++)
 574                 tx_pool->free_map[i] = i;
 575
 576         tx_pool->consumer_index = 0;
 577         tx_pool->producer_index = 0;
 578
 579         return 0;
 580 }
 581
 582 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
 583 {
 584         int tx_scrqs;
 585         int i, rc;
 586
 587         tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 588         for (i = 0; i < tx_scrqs; i++) {
 589                 rc = reset_one_tx_pool(adapter, &adapter->tso_pool[i]);
 590                 if (rc)
 591                         return rc;
 592                 rc = reset_one_tx_pool(adapter, &adapter->tx_pool[i]);
 593                 if (rc)
 594                         return rc;
 595         }
 596
 597         return 0;
 598 }
 599
 600 static void release_vpd_data(struct ibmvnic_adapter *adapter)
 601 {
 602         if (!adapter->vpd)
 603                 return;
 604
 605         kfree(adapter->vpd->buff);
 606         kfree(adapter->vpd);
 607
 608         adapter->vpd = NULL;
 609 }
 610
 611 static void release_one_tx_pool(struct ibmvnic_adapter *adapter,
 612                                 struct ibmvnic_tx_pool *tx_pool)
 613 {
 614         kfree(tx_pool->tx_buff);
 615         kfree(tx_pool->free_map);
 616         free_long_term_buff(adapter, &tx_pool->long_term_buff);
 617 }
 618
 619 static void release_tx_pools(struct ibmvnic_adapter *adapter)
 620 {
 621         int i;
 622
 623         if (!adapter->tx_pool)
 624                 return;
 625
 626         for (i = 0; i < adapter->num_active_tx_pools; i++) {
 627                 release_one_tx_pool(adapter, &adapter->tx_pool[i]);
 628                 release_one_tx_pool(adapter, &adapter->tso_pool[i]);
 629         }
 630
 631         kfree(adapter->tx_pool);
 632         adapter->tx_pool = NULL;
 633         kfree(adapter->tso_pool);
 634         adapter->tso_pool = NULL;
 635         adapter->num_active_tx_pools = 0;
 636 }
 637
 638 static int init_one_tx_pool(struct net_device *netdev,
 639                             struct ibmvnic_tx_pool *tx_pool,
 640                             int num_entries, int buf_size)
 641 {
 642         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 643         int i;
 644
 645         tx_pool->tx_buff = kcalloc(num_entries,
 646                                    sizeof(struct ibmvnic_tx_buff),
 647                                    GFP_KERNEL);
 648         if (!tx_pool->tx_buff)
 649                 return -1;
 650
 651         if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
 652                                  num_entries * buf_size))
 653                 return -1;
 654
 655         tx_pool->free_map = kcalloc(num_entries, sizeof(int), GFP_KERNEL);
 656         if (!tx_pool->free_map)
 657                 return -1;
 658
 659         for (i = 0; i < num_entries; i++)
 660                 tx_pool->free_map[i] = i;
 661
 662         tx_pool->consumer_index = 0;
 663         tx_pool->producer_index = 0;
 664         tx_pool->num_buffers = num_entries;
 665         tx_pool->buf_size = buf_size;
 666
 667         return 0;
 668 }
 669
 670 static int init_tx_pools(struct net_device *netdev)
 671 {
 672         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 673         int tx_subcrqs;
 674         int i, rc;
 675
 676         tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 677         adapter->tx_pool = kcalloc(tx_subcrqs,
 678                                    sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 679         if (!adapter->tx_pool)
 680                 return -1;
 681
 682         adapter->tso_pool = kcalloc(tx_subcrqs,
 683                                     sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 684         if (!adapter->tso_pool)
 685                 return -1;
 686
 687         adapter->num_active_tx_pools = tx_subcrqs;
 688
 689         for (i = 0; i < tx_subcrqs; i++) {
 690                 rc = init_one_tx_pool(netdev, &adapter->tx_pool[i],
 691                                       adapter->req_tx_entries_per_subcrq,
 692                                       adapter->req_mtu + VLAN_HLEN);
 693                 if (rc) {
 694                         release_tx_pools(adapter);
 695                         return rc;
 696                 }
 697
 698                 init_one_tx_pool(netdev, &adapter->tso_pool[i],
 699                                  IBMVNIC_TSO_BUFS,
 700                                  IBMVNIC_TSO_BUF_SZ);
 701                 if (rc) {
 702                         release_tx_pools(adapter);
 703                         return rc;
 704                 }
 705         }
 706
 707         return 0;
 708 }
 709
 710 static void release_error_buffers(struct ibmvnic_adapter *adapter)
 711 {
 712         struct device *dev = &adapter->vdev->dev;
 713         struct ibmvnic_error_buff *error_buff, *tmp;
 714         unsigned long flags;
 715
 716         spin_lock_irqsave(&adapter->error_list_lock, flags);
 717         list_for_each_entry_safe(error_buff, tmp, &adapter->errors, list) {
 718                 list_del(&error_buff->list);
 719                 dma_unmap_single(dev, error_buff->dma, error_buff->len,
 720                                  DMA_FROM_DEVICE);
 721                 kfree(error_buff->buff);
 722                 kfree(error_buff);
 723         }
 724         spin_unlock_irqrestore(&adapter->error_list_lock, flags);
 725 }
 726
 727 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
 728 {
 729         int i;
 730
 731         if (adapter->napi_enabled)
 732                 return;
 733
 734         for (i = 0; i < adapter->req_rx_queues; i++)
 735                 napi_enable(&adapter->napi[i]);
 736
 737         adapter->napi_enabled = true;
 738 }
 739
 740 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
 741 {
 742         int i;
 743
 744         if (!adapter->napi_enabled)
 745                 return;
 746
 747         for (i = 0; i < adapter->req_rx_queues; i++) {
 748                 netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
 749                 napi_disable(&adapter->napi[i]);
 750         }
 751
 752         adapter->napi_enabled = false;
 753 }
 754
 755 static int init_napi(struct ibmvnic_adapter *adapter)
 756 {
 757         int i;
 758
 759         adapter->napi = kcalloc(adapter->req_rx_queues,
 760                                 sizeof(struct napi_struct), GFP_KERNEL);
 761         if (!adapter->napi)
 762                 return -ENOMEM;
 763
 764         for (i = 0; i < adapter->req_rx_queues; i++) {
 765                 netdev_dbg(adapter->netdev, "Adding napi[%d]\n", i);
 766                 netif_napi_add(adapter->netdev, &adapter->napi[i],
 767                                ibmvnic_poll, NAPI_POLL_WEIGHT);
 768         }
 769
 770         adapter->num_active_rx_napi = adapter->req_rx_queues;
 771         return 0;
 772 }
 773
 774 static void release_napi(struct ibmvnic_adapter *adapter)
 775 {
 776         int i;
 777
 778         if (!adapter->napi)
 779                 return;
 780
 781         for (i = 0; i < adapter->num_active_rx_napi; i++) {
 782                 if (&adapter->napi[i]) {
 783                         netdev_dbg(adapter->netdev,
 784                                    "Releasing napi[%d]\n", i);
 785                         netif_napi_del(&adapter->napi[i]);
 786                 }
 787         }
 788
 789         kfree(adapter->napi);
 790         adapter->napi = NULL;
 791         adapter->num_active_rx_napi = 0;
 792         adapter->napi_enabled = false;
 793 }
 794
 795 static int ibmvnic_login(struct net_device *netdev)
 796 {
 797         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 798         unsigned long timeout = msecs_to_jiffies(30000);
 799         int retry_count = 0;
 800         int rc;
 801
 802         do {
 803                 if (retry_count > IBMVNIC_MAX_QUEUES) {
 804                         netdev_warn(netdev, "Login attempts exceeded\n");
 805                         return -1;
 806                 }
 807
 808                 adapter->init_done_rc = 0;
 809                 reinit_completion(&adapter->init_done);
 810                 rc = send_login(adapter);
 811                 if (rc) {
 812                         netdev_warn(netdev, "Unable to login\n");
 813                         return rc;
 814                 }
 815
 816                 if (!wait_for_completion_timeout(&adapter->init_done,
 817                                                  timeout)) {
 818                         netdev_warn(netdev, "Login timed out\n");
 819                         return -1;
 820                 }
 821
 822                 if (adapter->init_done_rc == PARTIALSUCCESS) {
 823                         retry_count++;
 824                         release_sub_crqs(adapter, 1);
 825
 826                         adapter->init_done_rc = 0;
 827                         reinit_completion(&adapter->init_done);
 828                         send_cap_queries(adapter);
 829                         if (!wait_for_completion_timeout(&adapter->init_done,
 830                                                          timeout)) {
 831                                 netdev_warn(netdev,
 832                                             "Capabilities query timed out\n");
 833                                 return -1;
 834                         }
 835
 836                         rc = init_sub_crqs(adapter);
 837                         if (rc) {
 838                                 netdev_warn(netdev,
 839                                             "SCRQ initialization failed\n");
 840                                 return -1;
 841                         }
 842
 843                         rc = init_sub_crq_irqs(adapter);
 844                         if (rc) {
 845                                 netdev_warn(netdev,
 846                                             "SCRQ irq initialization failed\n");
 847                                 return -1;
 848                         }
 849                 } else if (adapter->init_done_rc) {
 850                         netdev_warn(netdev, "Adapter login failed\n");
 851                         return -1;
 852                 }
 853         } while (adapter->init_done_rc == PARTIALSUCCESS);
 854
 855         /* handle pending MAC address changes after successful login */
 856         if (adapter->mac_change_pending) {
 857                 __ibmvnic_set_mac(netdev, &adapter->desired.mac);
 858                 adapter->mac_change_pending = false;
 859         }
 860
 861         return 0;
 862 }
 863
 864 static void release_login_buffer(struct ibmvnic_adapter *adapter)
 865 {
 866         kfree(adapter->login_buf);
 867         adapter->login_buf = NULL;
 868 }
 869
 870 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
 871 {
 872         kfree(adapter->login_rsp_buf);
 873         adapter->login_rsp_buf = NULL;
 874 }
 875
 876 static void release_resources(struct ibmvnic_adapter *adapter)
 877 {
 878         release_vpd_data(adapter);
 879
 880         release_tx_pools(adapter);
 881         release_rx_pools(adapter);
 882
 883         release_error_buffers(adapter);
 884         release_napi(adapter);
 885         release_login_rsp_buffer(adapter);
 886 }
 887
 888 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
 889 {
 890         struct net_device *netdev = adapter->netdev;
 891         unsigned long timeout = msecs_to_jiffies(30000);
 892         union ibmvnic_crq crq;
 893         bool resend;
 894         int rc;
 895
 896         netdev_dbg(netdev, "setting link state %d\n", link_state);
 897
 898         memset(&crq, 0, sizeof(crq));
 899         crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
 900         crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
 901         crq.logical_link_state.link_state = link_state;
 902
 903         do {
 904                 resend = false;
 905
 906                 reinit_completion(&adapter->init_done);
 907                 rc = ibmvnic_send_crq(adapter, &crq);
 908                 if (rc) {
 909                         netdev_err(netdev, "Failed to set link state\n");
 910                         return rc;
 911                 }
 912
 913                 if (!wait_for_completion_timeout(&adapter->init_done,
 914                                                  timeout)) {
 915                         netdev_err(netdev, "timeout setting link state\n");
 916                         return -1;
 917                 }
 918
 919                 if (adapter->init_done_rc == 1) {
 920                         /* Partuial success, delay and re-send */
 921                         mdelay(1000);
 922                         resend = true;
 923                 }
 924         } while (resend);
 925
 926         return 0;
 927 }
 928
 929 static int set_real_num_queues(struct net_device *netdev)
 930 {
 931         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 932         int rc;
 933
 934         netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
 935                    adapter->req_tx_queues, adapter->req_rx_queues);
 936
 937         rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
 938         if (rc) {
 939                 netdev_err(netdev, "failed to set the number of tx queues\n");
 940                 return rc;
 941         }
 942
 943         rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
 944         if (rc)
 945                 netdev_err(netdev, "failed to set the number of rx queues\n");
 946
 947         return rc;
 948 }
 949
 950 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
 951 {
 952         struct device *dev = &adapter->vdev->dev;
 953         union ibmvnic_crq crq;
 954         int len = 0;
 955
 956         if (adapter->vpd->buff)
 957                 len = adapter->vpd->len;
 958
 959         init_completion(&adapter->fw_done);
 960         crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
 961         crq.get_vpd_size.cmd = GET_VPD_SIZE;
 962         ibmvnic_send_crq(adapter, &crq);
 963         wait_for_completion(&adapter->fw_done);
 964
 965         if (!adapter->vpd->len)
 966                 return -ENODATA;
 967
 968         if (!adapter->vpd->buff)
 969                 adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
 970         else if (adapter->vpd->len != len)
 971                 adapter->vpd->buff =
 972                         krealloc(adapter->vpd->buff,
 973                                  adapter->vpd->len, GFP_KERNEL);
 974
 975         if (!adapter->vpd->buff) {
 976                 dev_err(dev, "Could allocate VPD buffer\n");
 977                 return -ENOMEM;
 978         }
 979
 980         adapter->vpd->dma_addr =
 981                 dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
 982                                DMA_FROM_DEVICE);
 983         if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
 984                 dev_err(dev, "Could not map VPD buffer\n");
 985                 kfree(adapter->vpd->buff);
 986                 adapter->vpd->buff = NULL;
 987                 return -ENOMEM;
 988         }
 989
 990         reinit_completion(&adapter->fw_done);
 991         crq.get_vpd.first = IBMVNIC_CRQ_CMD;
 992         crq.get_vpd.cmd = GET_VPD;
 993         crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
 994         crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
 995         ibmvnic_send_crq(adapter, &crq);
 996         wait_for_completion(&adapter->fw_done);
 997
 998         return 0;
 999 }
1000
1001 static int init_resources(struct ibmvnic_adapter *adapter)
1002 {
1003         struct net_device *netdev = adapter->netdev;
1004         int rc;
1005
1006         rc = set_real_num_queues(netdev);
1007         if (rc)
1008                 return rc;
1009
1010         adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
1011         if (!adapter->vpd)
1012                 return -ENOMEM;
1013
1014         /* Vital Product Data (VPD) */
1015         rc = ibmvnic_get_vpd(adapter);
1016         if (rc) {
1017                 netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
1018                 return rc;
1019         }
1020
1021         adapter->map_id = 1;
1022
1023         rc = init_napi(adapter);
1024         if (rc)
1025                 return rc;
1026
1027         send_map_query(adapter);
1028
1029         rc = init_rx_pools(netdev);
1030         if (rc)
1031                 return rc;
1032
1033         rc = init_tx_pools(netdev);
1034         return rc;
1035 }
1036
1037 static int __ibmvnic_open(struct net_device *netdev)
1038 {
1039         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1040         enum vnic_state prev_state = adapter->state;
1041         int i, rc;
1042
1043         adapter->state = VNIC_OPENING;
1044         replenish_pools(adapter);
1045         ibmvnic_napi_enable(adapter);
1046
1047         /* We're ready to receive frames, enable the sub-crq interrupts and
1048          * set the logical link state to up
1049          */
1050         for (i = 0; i < adapter->req_rx_queues; i++) {
1051                 netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1052                 if (prev_state == VNIC_CLOSED)
1053                         enable_irq(adapter->rx_scrq[i]->irq);
1054                 enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1055         }
1056
1057         for (i = 0; i < adapter->req_tx_queues; i++) {
1058                 netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1059                 if (prev_state == VNIC_CLOSED)
1060                         enable_irq(adapter->tx_scrq[i]->irq);
1061                 enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1062         }
1063
1064         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1065         if (rc) {
1066                 for (i = 0; i < adapter->req_rx_queues; i++)
1067                         napi_disable(&adapter->napi[i]);
1068                 release_resources(adapter);
1069                 return rc;
1070         }
1071
1072         netif_tx_start_all_queues(netdev);
1073
1074         if (prev_state == VNIC_CLOSED) {
1075                 for (i = 0; i < adapter->req_rx_queues; i++)
1076                         napi_schedule(&adapter->napi[i]);
1077         }
1078
1079         adapter->state = VNIC_OPEN;
1080         return rc;
1081 }
1082
1083 static int ibmvnic_open(struct net_device *netdev)
1084 {
1085         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1086         int rc;
1087
1088         /* If device failover is pending, just set device state and return.
1089          * Device operation will be handled by reset routine.
1090          */
1091         if (adapter->failover_pending) {
1092                 adapter->state = VNIC_OPEN;
1093                 return 0;
1094         }
1095
1096         mutex_lock(&adapter->reset_lock);
1097
1098         if (adapter->state != VNIC_CLOSED) {
1099                 rc = ibmvnic_login(netdev);
1100                 if (rc) {
1101                         mutex_unlock(&adapter->reset_lock);
1102                         return rc;
1103                 }
1104
1105                 rc = init_resources(adapter);
1106                 if (rc) {
1107                         netdev_err(netdev, "failed to initialize resources\n");
1108                         release_resources(adapter);
1109                         mutex_unlock(&adapter->reset_lock);
1110                         return rc;
1111                 }
1112         }
1113
1114         rc = __ibmvnic_open(netdev);
1115         netif_carrier_on(netdev);
1116
1117         mutex_unlock(&adapter->reset_lock);
1118
1119         return rc;
1120 }
1121
1122 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1123 {
1124         struct ibmvnic_rx_pool *rx_pool;
1125         struct ibmvnic_rx_buff *rx_buff;
1126         u64 rx_entries;
1127         int rx_scrqs;
1128         int i, j;
1129
1130         if (!adapter->rx_pool)
1131                 return;
1132
1133         rx_scrqs = adapter->num_active_rx_pools;
1134         rx_entries = adapter->req_rx_add_entries_per_subcrq;
1135
1136         /* Free any remaining skbs in the rx buffer pools */
1137         for (i = 0; i < rx_scrqs; i++) {
1138                 rx_pool = &adapter->rx_pool[i];
1139                 if (!rx_pool || !rx_pool->rx_buff)
1140                         continue;
1141
1142                 netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1143                 for (j = 0; j < rx_entries; j++) {
1144                         rx_buff = &rx_pool->rx_buff[j];
1145                         if (rx_buff && rx_buff->skb) {
1146                                 dev_kfree_skb_any(rx_buff->skb);
1147                                 rx_buff->skb = NULL;
1148                         }
1149                 }
1150         }
1151 }
1152
1153 static void clean_one_tx_pool(struct ibmvnic_adapter *adapter,
1154                               struct ibmvnic_tx_pool *tx_pool)
1155 {
1156         struct ibmvnic_tx_buff *tx_buff;
1157         u64 tx_entries;
1158         int i;
1159
1160         if (!tx_pool || !tx_pool->tx_buff)
1161                 return;
1162
1163         tx_entries = tx_pool->num_buffers;
1164
1165         for (i = 0; i < tx_entries; i++) {
1166                 tx_buff = &tx_pool->tx_buff[i];
1167                 if (tx_buff && tx_buff->skb) {
1168                         dev_kfree_skb_any(tx_buff->skb);
1169                         tx_buff->skb = NULL;
1170                 }
1171         }
1172 }
1173
1174 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1175 {
1176         int tx_scrqs;
1177         int i;
1178
1179         if (!adapter->tx_pool || !adapter->tso_pool)
1180                 return;
1181
1182         tx_scrqs = adapter->num_active_tx_pools;
1183
1184         /* Free any remaining skbs in the tx buffer pools */
1185         for (i = 0; i < tx_scrqs; i++) {
1186                 netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1187                 clean_one_tx_pool(adapter, &adapter->tx_pool[i]);
1188                 clean_one_tx_pool(adapter, &adapter->tso_pool[i]);
1189         }
1190 }
1191
1192 static void ibmvnic_disable_irqs(struct ibmvnic_adapter *adapter)
1193 {
1194         struct net_device *netdev = adapter->netdev;
1195         int i;
1196
1197         if (adapter->tx_scrq) {
1198                 for (i = 0; i < adapter->req_tx_queues; i++)
1199                         if (adapter->tx_scrq[i]->irq) {
1200                                 netdev_dbg(netdev,
1201                                            "Disabling tx_scrq[%d] irq\n", i);
1202                                 disable_scrq_irq(adapter, adapter->tx_scrq[i]);
1203                                 disable_irq(adapter->tx_scrq[i]->irq);
1204                         }
1205         }
1206
1207         if (adapter->rx_scrq) {
1208                 for (i = 0; i < adapter->req_rx_queues; i++) {
1209                         if (adapter->rx_scrq[i]->irq) {
1210                                 netdev_dbg(netdev,
1211                                            "Disabling rx_scrq[%d] irq\n", i);
1212                                 disable_scrq_irq(adapter, adapter->rx_scrq[i]);
1213                                 disable_irq(adapter->rx_scrq[i]->irq);
1214                         }
1215                 }
1216         }
1217 }
1218
1219 static void ibmvnic_cleanup(struct net_device *netdev)
1220 {
1221         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1222
1223         /* ensure that transmissions are stopped if called by do_reset */
1224         if (adapter->resetting)
1225                 netif_tx_disable(netdev);
1226         else
1227                 netif_tx_stop_all_queues(netdev);
1228
1229         ibmvnic_napi_disable(adapter);
1230         ibmvnic_disable_irqs(adapter);
1231
1232         clean_rx_pools(adapter);
1233         clean_tx_pools(adapter);
1234 }
1235
1236 static int __ibmvnic_close(struct net_device *netdev)
1237 {
1238         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1239         int rc = 0;
1240
1241         adapter->state = VNIC_CLOSING;
1242         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1243         if (rc)
1244                 return rc;
1245         adapter->state = VNIC_CLOSED;
1246         return 0;
1247 }
1248
1249 static int ibmvnic_close(struct net_device *netdev)
1250 {
1251         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1252         int rc;
1253
1254         /* If device failover is pending, just set device state and return.
1255          * Device operation will be handled by reset routine.
1256          */
1257         if (adapter->failover_pending) {
1258                 adapter->state = VNIC_CLOSED;
1259                 return 0;
1260         }
1261
1262         mutex_lock(&adapter->reset_lock);
1263         rc = __ibmvnic_close(netdev);
1264         ibmvnic_cleanup(netdev);
1265         mutex_unlock(&adapter->reset_lock);
1266
1267         return rc;
1268 }
1269
1270 /**
1271  * build_hdr_data - creates L2/L3/L4 header data buffer
1272  * @hdr_field - bitfield determining needed headers
1273  * @skb - socket buffer
1274  * @hdr_len - array of header lengths
1275  * @tot_len - total length of data
1276  *
1277  * Reads hdr_field to determine which headers are needed by firmware.
1278  * Builds a buffer containing these headers.  Saves individual header
1279  * lengths and total buffer length to be used to build descriptors.
1280  */
1281 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1282                           int *hdr_len, u8 *hdr_data)
1283 {
1284         int len = 0;
1285         u8 *hdr;
1286
1287         if (skb_vlan_tagged(skb) && !skb_vlan_tag_present(skb))
1288                 hdr_len[0] = sizeof(struct vlan_ethhdr);
1289         else
1290                 hdr_len[0] = sizeof(struct ethhdr);
1291
1292         if (skb->protocol == htons(ETH_P_IP)) {
1293                 hdr_len[1] = ip_hdr(skb)->ihl * 4;
1294                 if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1295                         hdr_len[2] = tcp_hdrlen(skb);
1296                 else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1297                         hdr_len[2] = sizeof(struct udphdr);
1298         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1299                 hdr_len[1] = sizeof(struct ipv6hdr);
1300                 if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1301                         hdr_len[2] = tcp_hdrlen(skb);
1302                 else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1303                         hdr_len[2] = sizeof(struct udphdr);
1304         } else if (skb->protocol == htons(ETH_P_ARP)) {
1305                 hdr_len[1] = arp_hdr_len(skb->dev);
1306                 hdr_len[2] = 0;
1307         }
1308
1309         memset(hdr_data, 0, 120);
1310         if ((hdr_field >> 6) & 1) {
1311                 hdr = skb_mac_header(skb);
1312                 memcpy(hdr_data, hdr, hdr_len[0]);
1313                 len += hdr_len[0];
1314         }
1315
1316         if ((hdr_field >> 5) & 1) {
1317                 hdr = skb_network_header(skb);
1318                 memcpy(hdr_data + len, hdr, hdr_len[1]);
1319                 len += hdr_len[1];
1320         }
1321
1322         if ((hdr_field >> 4) & 1) {
1323                 hdr = skb_transport_header(skb);
1324                 memcpy(hdr_data + len, hdr, hdr_len[2]);
1325                 len += hdr_len[2];
1326         }
1327         return len;
1328 }
1329
1330 /**
1331  * create_hdr_descs - create header and header extension descriptors
1332  * @hdr_field - bitfield determining needed headers
1333  * @data - buffer containing header data
1334  * @len - length of data buffer
1335  * @hdr_len - array of individual header lengths
1336  * @scrq_arr - descriptor array
1337  *
1338  * Creates header and, if needed, header extension descriptors and
1339  * places them in a descriptor array, scrq_arr
1340  */
1341
1342 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1343                             union sub_crq *scrq_arr)
1344 {
1345         union sub_crq hdr_desc;
1346         int tmp_len = len;
1347         int num_descs = 0;
1348         u8 *data, *cur;
1349         int tmp;
1350
1351         while (tmp_len > 0) {
1352                 cur = hdr_data + len - tmp_len;
1353
1354                 memset(&hdr_desc, 0, sizeof(hdr_desc));
1355                 if (cur != hdr_data) {
1356                         data = hdr_desc.hdr_ext.data;
1357                         tmp = tmp_len > 29 ? 29 : tmp_len;
1358                         hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1359                         hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1360                         hdr_desc.hdr_ext.len = tmp;
1361                 } else {
1362                         data = hdr_desc.hdr.data;
1363                         tmp = tmp_len > 24 ? 24 : tmp_len;
1364                         hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1365                         hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1366                         hdr_desc.hdr.len = tmp;
1367                         hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1368                         hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1369                         hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1370                         hdr_desc.hdr.flag = hdr_field << 1;
1371                 }
1372                 memcpy(data, cur, tmp);
1373                 tmp_len -= tmp;
1374                 *scrq_arr = hdr_desc;
1375                 scrq_arr++;
1376                 num_descs++;
1377         }
1378
1379         return num_descs;
1380 }
1381
1382 /**
1383  * build_hdr_descs_arr - build a header descriptor array
1384  * @skb - socket buffer
1385  * @num_entries - number of descriptors to be sent
1386  * @subcrq - first TX descriptor
1387  * @hdr_field - bit field determining which headers will be sent
1388  *
1389  * This function will build a TX descriptor array with applicable
1390  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1391  */
1392
1393 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1394                                 int *num_entries, u8 hdr_field)
1395 {
1396         int hdr_len[3] = {0, 0, 0};
1397         int tot_len;
1398         u8 *hdr_data = txbuff->hdr_data;
1399
1400         tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1401                                  txbuff->hdr_data);
1402         *num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1403                          txbuff->indir_arr + 1);
1404 }
1405
1406 static int ibmvnic_xmit_workarounds(struct sk_buff *skb,
1407                                     struct net_device *netdev)
1408 {
1409         /* For some backing devices, mishandling of small packets
1410          * can result in a loss of connection or TX stall. Device
1411          * architects recommend that no packet should be smaller
1412          * than the minimum MTU value provided to the driver, so
1413          * pad any packets to that length
1414          */
1415         if (skb->len < netdev->min_mtu)
1416                 return skb_put_padto(skb, netdev->min_mtu);
1417
1418         return 0;
1419 }
1420
1421 static int ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1422 {
1423         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1424         int queue_num = skb_get_queue_mapping(skb);
1425         u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1426         struct device *dev = &adapter->vdev->dev;
1427         struct ibmvnic_tx_buff *tx_buff = NULL;
1428         struct ibmvnic_sub_crq_queue *tx_scrq;
1429         struct ibmvnic_tx_pool *tx_pool;
1430         unsigned int tx_send_failed = 0;
1431         unsigned int tx_map_failed = 0;
1432         unsigned int tx_dropped = 0;
1433         unsigned int tx_packets = 0;
1434         unsigned int tx_bytes = 0;
1435         dma_addr_t data_dma_addr;
1436         struct netdev_queue *txq;
1437         unsigned long lpar_rc;
1438         union sub_crq tx_crq;
1439         unsigned int offset;
1440         int num_entries = 1;
1441         unsigned char *dst;
1442         u64 *handle_array;
1443         int index = 0;
1444         u8 proto = 0;
1445         int ret = 0;
1446
1447         if (adapter->resetting) {
1448                 if (!netif_subqueue_stopped(netdev, skb))
1449                         netif_stop_subqueue(netdev, queue_num);
1450                 dev_kfree_skb_any(skb);
1451
1452                 tx_send_failed++;
1453                 tx_dropped++;
1454                 ret = NETDEV_TX_OK;
1455                 goto out;
1456         }
1457
1458         if (ibmvnic_xmit_workarounds(skb, netdev)) {
1459                 tx_dropped++;
1460                 tx_send_failed++;
1461                 ret = NETDEV_TX_OK;
1462                 goto out;
1463         }
1464         if (skb_is_gso(skb))
1465                 tx_pool = &adapter->tso_pool[queue_num];
1466         else
1467                 tx_pool = &adapter->tx_pool[queue_num];
1468
1469         tx_scrq = adapter->tx_scrq[queue_num];
1470         txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1471         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1472                 be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1473
1474         index = tx_pool->free_map[tx_pool->consumer_index];
1475
1476         if (index == IBMVNIC_INVALID_MAP) {
1477                 dev_kfree_skb_any(skb);
1478                 tx_send_failed++;
1479                 tx_dropped++;
1480                 ret = NETDEV_TX_OK;
1481                 goto out;
1482         }
1483
1484         tx_pool->free_map[tx_pool->consumer_index] = IBMVNIC_INVALID_MAP;
1485
1486         offset = index * tx_pool->buf_size;
1487         dst = tx_pool->long_term_buff.buff + offset;
1488         memset(dst, 0, tx_pool->buf_size);
1489         data_dma_addr = tx_pool->long_term_buff.addr + offset;
1490
1491         if (skb_shinfo(skb)->nr_frags) {
1492                 int cur, i;
1493
1494                 /* Copy the head */
1495                 skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1496                 cur = skb_headlen(skb);
1497
1498                 /* Copy the frags */
1499                 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1500                         const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1501
1502                         memcpy(dst + cur,
1503                                page_address(skb_frag_page(frag)) +
1504                                frag->page_offset, skb_frag_size(frag));
1505                         cur += skb_frag_size(frag);
1506                 }
1507         } else {
1508                 skb_copy_from_linear_data(skb, dst, skb->len);
1509         }
1510
1511         tx_pool->consumer_index =
1512             (tx_pool->consumer_index + 1) % tx_pool->num_buffers;
1513
1514         tx_buff = &tx_pool->tx_buff[index];
1515         tx_buff->skb = skb;
1516         tx_buff->data_dma[0] = data_dma_addr;
1517         tx_buff->data_len[0] = skb->len;
1518         tx_buff->index = index;
1519         tx_buff->pool_index = queue_num;
1520         tx_buff->last_frag = true;
1521
1522         memset(&tx_crq, 0, sizeof(tx_crq));
1523         tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1524         tx_crq.v1.type = IBMVNIC_TX_DESC;
1525         tx_crq.v1.n_crq_elem = 1;
1526         tx_crq.v1.n_sge = 1;
1527         tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1528
1529         if (skb_is_gso(skb))
1530                 tx_crq.v1.correlator =
1531                         cpu_to_be32(index | IBMVNIC_TSO_POOL_MASK);
1532         else
1533                 tx_crq.v1.correlator = cpu_to_be32(index);
1534         tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1535         tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1536         tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1537
1538         if (adapter->vlan_header_insertion) {
1539                 tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1540                 tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1541         }
1542
1543         if (skb->protocol == htons(ETH_P_IP)) {
1544                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1545                 proto = ip_hdr(skb)->protocol;
1546         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1547                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1548                 proto = ipv6_hdr(skb)->nexthdr;
1549         }
1550
1551         if (proto == IPPROTO_TCP)
1552                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1553         else if (proto == IPPROTO_UDP)
1554                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1555
1556         if (skb->ip_summed == CHECKSUM_PARTIAL) {
1557                 tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1558                 hdrs += 2;
1559         }
1560         if (skb_is_gso(skb)) {
1561                 tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1562                 tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1563                 hdrs += 2;
1564         }
1565         /* determine if l2/3/4 headers are sent to firmware */
1566         if ((*hdrs >> 7) & 1) {
1567                 build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1568                 tx_crq.v1.n_crq_elem = num_entries;
1569                 tx_buff->num_entries = num_entries;
1570                 tx_buff->indir_arr[0] = tx_crq;
1571                 tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1572                                                     sizeof(tx_buff->indir_arr),
1573                                                     DMA_TO_DEVICE);
1574                 if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1575                         dev_kfree_skb_any(skb);
1576                         tx_buff->skb = NULL;
1577                         if (!firmware_has_feature(FW_FEATURE_CMO))
1578                                 dev_err(dev, "tx: unable to map descriptor array\n");
1579                         tx_map_failed++;
1580                         tx_dropped++;
1581                         ret = NETDEV_TX_OK;
1582                         goto tx_err_out;
1583                 }
1584                 lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1585                                                (u64)tx_buff->indir_dma,
1586                                                (u64)num_entries);
1587         } else {
1588                 tx_buff->num_entries = num_entries;
1589                 lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1590                                       &tx_crq);
1591         }
1592         if (lpar_rc != H_SUCCESS) {
1593                 dev_err(dev, "tx failed with code %ld\n", lpar_rc);
1594                 dev_kfree_skb_any(skb);
1595                 tx_buff->skb = NULL;
1596
1597                 if (lpar_rc == H_CLOSED || adapter->failover_pending) {
1598                         /* Disable TX and report carrier off if queue is closed
1599                          * or pending failover.
1600                          * Firmware guarantees that a signal will be sent to the
1601                          * driver, triggering a reset or some other action.
1602                          */
1603                         netif_tx_stop_all_queues(netdev);
1604                         netif_carrier_off(netdev);
1605                 }
1606
1607                 tx_send_failed++;
1608                 tx_dropped++;
1609                 ret = NETDEV_TX_OK;
1610                 goto tx_err_out;
1611         }
1612
1613         if (atomic_add_return(num_entries, &tx_scrq->used)
1614                                         >= adapter->req_tx_entries_per_subcrq) {
1615                 netdev_dbg(netdev, "Stopping queue %d\n", queue_num);
1616                 netif_stop_subqueue(netdev, queue_num);
1617         }
1618
1619         tx_packets++;
1620         tx_bytes += skb->len;
1621         txq->trans_start = jiffies;
1622         ret = NETDEV_TX_OK;
1623         goto out;
1624
1625 tx_err_out:
1626         /* roll back consumer index and map array*/
1627         if (tx_pool->consumer_index == 0)
1628                 tx_pool->consumer_index =
1629                         tx_pool->num_buffers - 1;
1630         else
1631                 tx_pool->consumer_index--;
1632         tx_pool->free_map[tx_pool->consumer_index] = index;
1633 out:
1634         netdev->stats.tx_dropped += tx_dropped;
1635         netdev->stats.tx_bytes += tx_bytes;
1636         netdev->stats.tx_packets += tx_packets;
1637         adapter->tx_send_failed += tx_send_failed;
1638         adapter->tx_map_failed += tx_map_failed;
1639         adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1640         adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1641         adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1642
1643         return ret;
1644 }
1645
1646 static void ibmvnic_set_multi(struct net_device *netdev)
1647 {
1648         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1649         struct netdev_hw_addr *ha;
1650         union ibmvnic_crq crq;
1651
1652         memset(&crq, 0, sizeof(crq));
1653         crq.request_capability.first = IBMVNIC_CRQ_CMD;
1654         crq.request_capability.cmd = REQUEST_CAPABILITY;
1655
1656         if (netdev->flags & IFF_PROMISC) {
1657                 if (!adapter->promisc_supported)
1658                         return;
1659         } else {
1660                 if (netdev->flags & IFF_ALLMULTI) {
1661                         /* Accept all multicast */
1662                         memset(&crq, 0, sizeof(crq));
1663                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1664                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1665                         crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1666                         ibmvnic_send_crq(adapter, &crq);
1667                 } else if (netdev_mc_empty(netdev)) {
1668                         /* Reject all multicast */
1669                         memset(&crq, 0, sizeof(crq));
1670                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1671                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1672                         crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1673                         ibmvnic_send_crq(adapter, &crq);
1674                 } else {
1675                         /* Accept one or more multicast(s) */
1676                         netdev_for_each_mc_addr(ha, netdev) {
1677                                 memset(&crq, 0, sizeof(crq));
1678                                 crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1679                                 crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1680                                 crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1681                                 ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1682                                                 ha->addr);
1683                                 ibmvnic_send_crq(adapter, &crq);
1684                         }
1685                 }
1686         }
1687 }
1688
1689 static int __ibmvnic_set_mac(struct net_device *netdev, struct sockaddr *p)
1690 {
1691         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1692         struct sockaddr *addr = p;
1693         union ibmvnic_crq crq;
1694
1695         if (!is_valid_ether_addr(addr->sa_data))
1696                 return -EADDRNOTAVAIL;
1697
1698         memset(&crq, 0, sizeof(crq));
1699         crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1700         crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1701         ether_addr_copy(&crq.change_mac_addr.mac_addr[0], addr->sa_data);
1702
1703         init_completion(&adapter->fw_done);
1704         ibmvnic_send_crq(adapter, &crq);
1705         wait_for_completion(&adapter->fw_done);
1706         /* netdev->dev_addr is changed in handle_change_mac_rsp function */
1707         return adapter->fw_done_rc ? -EIO : 0;
1708 }
1709
1710 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1711 {
1712         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1713         struct sockaddr *addr = p;
1714         int rc;
1715
1716         if (adapter->state == VNIC_PROBED) {
1717                 memcpy(&adapter->desired.mac, addr, sizeof(struct sockaddr));
1718                 adapter->mac_change_pending = true;
1719                 return 0;
1720         }
1721
1722         rc = __ibmvnic_set_mac(netdev, addr);
1723
1724         return rc;
1725 }
1726
1727 /**
1728  * do_reset returns zero if we are able to keep processing reset events, or
1729  * non-zero if we hit a fatal error and must halt.
1730  */
1731 static int do_reset(struct ibmvnic_adapter *adapter,
1732                     struct ibmvnic_rwi *rwi, u32 reset_state)
1733 {
1734         u64 old_num_rx_queues, old_num_tx_queues;
1735         struct net_device *netdev = adapter->netdev;
1736         int i, rc;
1737
1738         netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1739                    rwi->reset_reason);
1740
1741         netif_carrier_off(netdev);
1742         adapter->reset_reason = rwi->reset_reason;
1743
1744         old_num_rx_queues = adapter->req_rx_queues;
1745         old_num_tx_queues = adapter->req_tx_queues;
1746
1747         ibmvnic_cleanup(netdev);
1748
1749         if (adapter->reset_reason != VNIC_RESET_MOBILITY &&
1750             adapter->reset_reason != VNIC_RESET_FAILOVER) {
1751                 rc = __ibmvnic_close(netdev);
1752                 if (rc)
1753                         return rc;
1754         }
1755
1756         if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1757             adapter->wait_for_reset) {
1758                 release_resources(adapter);
1759                 release_sub_crqs(adapter, 1);
1760                 release_crq_queue(adapter);
1761         }
1762
1763         if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1764                 /* remove the closed state so when we call open it appears
1765                  * we are coming from the probed state.
1766                  */
1767                 adapter->state = VNIC_PROBED;
1768
1769                 if (adapter->wait_for_reset) {
1770                         rc = init_crq_queue(adapter);
1771                 } else if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
1772                         rc = ibmvnic_reenable_crq_queue(adapter);
1773                         release_sub_crqs(adapter, 1);
1774                 } else {
1775                         rc = ibmvnic_reset_crq(adapter);
1776                         if (!rc)
1777                                 rc = vio_enable_interrupts(adapter->vdev);
1778                 }
1779
1780                 if (rc) {
1781                         netdev_err(adapter->netdev,
1782                                    "Couldn't initialize crq. rc=%d\n", rc);
1783                         return rc;
1784                 }
1785
1786                 rc = ibmvnic_init(adapter);
1787                 if (rc)
1788                         return IBMVNIC_INIT_FAILED;
1789
1790                 /* If the adapter was in PROBE state prior to the reset,
1791                  * exit here.
1792                  */
1793                 if (reset_state == VNIC_PROBED)
1794                         return 0;
1795
1796                 rc = ibmvnic_login(netdev);
1797                 if (rc) {
1798                         adapter->state = VNIC_PROBED;
1799                         return 0;
1800                 }
1801
1802                 if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1803                     adapter->wait_for_reset) {
1804                         rc = init_resources(adapter);
1805                         if (rc)
1806                                 return rc;
1807                 } else if (adapter->req_rx_queues != old_num_rx_queues ||
1808                            adapter->req_tx_queues != old_num_tx_queues) {
1809                         adapter->map_id = 1;
1810                         release_rx_pools(adapter);
1811                         release_tx_pools(adapter);
1812                         init_rx_pools(netdev);
1813                         init_tx_pools(netdev);
1814
1815                         release_napi(adapter);
1816                         init_napi(adapter);
1817                 } else {
1818                         rc = reset_tx_pools(adapter);
1819                         if (rc)
1820                                 return rc;
1821
1822                         rc = reset_rx_pools(adapter);
1823                         if (rc)
1824                                 return rc;
1825                 }
1826                 ibmvnic_disable_irqs(adapter);
1827         }
1828         adapter->state = VNIC_CLOSED;
1829
1830         if (reset_state == VNIC_CLOSED)
1831                 return 0;
1832
1833         rc = __ibmvnic_open(netdev);
1834         if (rc) {
1835                 if (list_empty(&adapter->rwi_list))
1836                         adapter->state = VNIC_CLOSED;
1837                 else
1838                         adapter->state = reset_state;
1839
1840                 return 0;
1841         }
1842
1843         /* kick napi */
1844         for (i = 0; i < adapter->req_rx_queues; i++)
1845                 napi_schedule(&adapter->napi[i]);
1846
1847         if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
1848             adapter->reset_reason != VNIC_RESET_CHANGE_PARAM)
1849                 netdev_notify_peers(netdev);
1850
1851         netif_carrier_on(netdev);
1852
1853         return 0;
1854 }
1855
1856 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
1857 {
1858         struct ibmvnic_rwi *rwi;
1859
1860         mutex_lock(&adapter->rwi_lock);
1861
1862         if (!list_empty(&adapter->rwi_list)) {
1863                 rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
1864                                        list);
1865                 list_del(&rwi->list);
1866         } else {
1867                 rwi = NULL;
1868         }
1869
1870         mutex_unlock(&adapter->rwi_lock);
1871         return rwi;
1872 }
1873
1874 static void free_all_rwi(struct ibmvnic_adapter *adapter)
1875 {
1876         struct ibmvnic_rwi *rwi;
1877
1878         rwi = get_next_rwi(adapter);
1879         while (rwi) {
1880                 kfree(rwi);
1881                 rwi = get_next_rwi(adapter);
1882         }
1883 }
1884
1885 static void __ibmvnic_reset(struct work_struct *work)
1886 {
1887         struct ibmvnic_rwi *rwi;
1888         struct ibmvnic_adapter *adapter;
1889         struct net_device *netdev;
1890         u32 reset_state;
1891         int rc = 0;
1892
1893         adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
1894         netdev = adapter->netdev;
1895
1896         mutex_lock(&adapter->reset_lock);
1897         adapter->resetting = true;
1898         reset_state = adapter->state;
1899
1900         rwi = get_next_rwi(adapter);
1901         while (rwi) {
1902                 rc = do_reset(adapter, rwi, reset_state);
1903                 kfree(rwi);
1904                 if (rc && rc != IBMVNIC_INIT_FAILED)
1905                         break;
1906
1907                 rwi = get_next_rwi(adapter);
1908         }
1909
1910         if (adapter->wait_for_reset) {
1911                 adapter->wait_for_reset = false;
1912                 adapter->reset_done_rc = rc;
1913                 complete(&adapter->reset_done);
1914         }
1915
1916         if (rc) {
1917                 netdev_dbg(adapter->netdev, "Reset failed\n");
1918                 free_all_rwi(adapter);
1919                 mutex_unlock(&adapter->reset_lock);
1920                 return;
1921         }
1922
1923         adapter->resetting = false;
1924         mutex_unlock(&adapter->reset_lock);
1925 }
1926
1927 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
1928                          enum ibmvnic_reset_reason reason)
1929 {
1930         struct ibmvnic_rwi *rwi, *tmp;
1931         struct net_device *netdev = adapter->netdev;
1932         struct list_head *entry;
1933         int ret;
1934
1935         if (adapter->state == VNIC_REMOVING ||
1936             adapter->state == VNIC_REMOVED ||
1937             adapter->failover_pending) {
1938                 ret = EBUSY;
1939                 netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
1940                 goto err;
1941         }
1942
1943         if (adapter->state == VNIC_PROBING) {
1944                 netdev_warn(netdev, "Adapter reset during probe\n");
1945                 ret = adapter->init_done_rc = EAGAIN;
1946                 goto err;
1947         }
1948
1949         mutex_lock(&adapter->rwi_lock);
1950
1951         list_for_each(entry, &adapter->rwi_list) {
1952                 tmp = list_entry(entry, struct ibmvnic_rwi, list);
1953                 if (tmp->reset_reason == reason) {
1954                         netdev_dbg(netdev, "Skipping matching reset\n");
1955                         mutex_unlock(&adapter->rwi_lock);
1956                         ret = EBUSY;
1957                         goto err;
1958                 }
1959         }
1960
1961         rwi = kzalloc(sizeof(*rwi), GFP_KERNEL);
1962         if (!rwi) {
1963                 mutex_unlock(&adapter->rwi_lock);
1964                 ibmvnic_close(netdev);
1965                 ret = ENOMEM;
1966                 goto err;
1967         }
1968
1969         rwi->reset_reason = reason;
1970         list_add_tail(&rwi->list, &adapter->rwi_list);
1971         mutex_unlock(&adapter->rwi_lock);
1972
1973         netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
1974         schedule_work(&adapter->ibmvnic_reset);
1975
1976         return 0;
1977 err:
1978         if (adapter->wait_for_reset)
1979                 adapter->wait_for_reset = false;
1980         return -ret;
1981 }
1982
1983 static void ibmvnic_tx_timeout(struct net_device *dev)
1984 {
1985         struct ibmvnic_adapter *adapter = netdev_priv(dev);
1986
1987         ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
1988 }
1989
1990 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
1991                                   struct ibmvnic_rx_buff *rx_buff)
1992 {
1993         struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
1994
1995         rx_buff->skb = NULL;
1996
1997         pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
1998         pool->next_alloc = (pool->next_alloc + 1) % pool->size;
1999
2000         atomic_dec(&pool->available);
2001 }
2002
2003 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2004 {
2005         struct net_device *netdev = napi->dev;
2006         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2007         int scrq_num = (int)(napi - adapter->napi);
2008         int frames_processed = 0;
2009
2010 restart_poll:
2011         while (frames_processed < budget) {
2012                 struct sk_buff *skb;
2013                 struct ibmvnic_rx_buff *rx_buff;
2014                 union sub_crq *next;
2015                 u32 length;
2016                 u16 offset;
2017                 u8 flags = 0;
2018
2019                 if (unlikely(adapter->resetting &&
2020                              adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2021                         enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2022                         napi_complete_done(napi, frames_processed);
2023                         return frames_processed;
2024                 }
2025
2026                 if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
2027                         break;
2028                 next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
2029                 rx_buff =
2030                     (struct ibmvnic_rx_buff *)be64_to_cpu(next->
2031                                                           rx_comp.correlator);
2032                 /* do error checking */
2033                 if (next->rx_comp.rc) {
2034                         netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2035                                    be16_to_cpu(next->rx_comp.rc));
2036                         /* free the entry */
2037                         next->rx_comp.first = 0;
2038                         dev_kfree_skb_any(rx_buff->skb);
2039                         remove_buff_from_pool(adapter, rx_buff);
2040                         continue;
2041                 } else if (!rx_buff->skb) {
2042                         /* free the entry */
2043                         next->rx_comp.first = 0;
2044                         remove_buff_from_pool(adapter, rx_buff);
2045                         continue;
2046                 }
2047
2048                 length = be32_to_cpu(next->rx_comp.len);
2049                 offset = be16_to_cpu(next->rx_comp.off_frame_data);
2050                 flags = next->rx_comp.flags;
2051                 skb = rx_buff->skb;
2052                 skb_copy_to_linear_data(skb, rx_buff->data + offset,
2053                                         length);
2054
2055                 /* VLAN Header has been stripped by the system firmware and
2056                  * needs to be inserted by the driver
2057                  */
2058                 if (adapter->rx_vlan_header_insertion &&
2059                     (flags & IBMVNIC_VLAN_STRIPPED))
2060                         __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2061                                                ntohs(next->rx_comp.vlan_tci));
2062
2063                 /* free the entry */
2064                 next->rx_comp.first = 0;
2065                 remove_buff_from_pool(adapter, rx_buff);
2066
2067                 skb_put(skb, length);
2068                 skb->protocol = eth_type_trans(skb, netdev);
2069                 skb_record_rx_queue(skb, scrq_num);
2070
2071                 if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2072                     flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2073                         skb->ip_summed = CHECKSUM_UNNECESSARY;
2074                 }
2075
2076                 length = skb->len;
2077                 napi_gro_receive(napi, skb); /* send it up */
2078                 netdev->stats.rx_packets++;
2079                 netdev->stats.rx_bytes += length;
2080                 adapter->rx_stats_buffers[scrq_num].packets++;
2081                 adapter->rx_stats_buffers[scrq_num].bytes += length;
2082                 frames_processed++;
2083         }
2084
2085         if (adapter->state != VNIC_CLOSING)
2086                 replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2087
2088         if (frames_processed < budget) {
2089                 enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2090                 napi_complete_done(napi, frames_processed);
2091                 if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
2092                     napi_reschedule(napi)) {
2093                         disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2094                         goto restart_poll;
2095                 }
2096         }
2097         return frames_processed;
2098 }
2099
2100 #ifdef CONFIG_NET_POLL_CONTROLLER
2101 static void ibmvnic_netpoll_controller(struct net_device *dev)
2102 {
2103         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2104         int i;
2105
2106         replenish_pools(netdev_priv(dev));
2107         for (i = 0; i < adapter->req_rx_queues; i++)
2108                 ibmvnic_interrupt_rx(adapter->rx_scrq[i]->irq,
2109                                      adapter->rx_scrq[i]);
2110 }
2111 #endif
2112
2113 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2114 {
2115         int rc, ret;
2116
2117         adapter->fallback.mtu = adapter->req_mtu;
2118         adapter->fallback.rx_queues = adapter->req_rx_queues;
2119         adapter->fallback.tx_queues = adapter->req_tx_queues;
2120         adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2121         adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2122
2123         init_completion(&adapter->reset_done);
2124         adapter->wait_for_reset = true;
2125         rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2126         if (rc)
2127                 return rc;
2128         wait_for_completion(&adapter->reset_done);
2129
2130         ret = 0;
2131         if (adapter->reset_done_rc) {
2132                 ret = -EIO;
2133                 adapter->desired.mtu = adapter->fallback.mtu;
2134                 adapter->desired.rx_queues = adapter->fallback.rx_queues;
2135                 adapter->desired.tx_queues = adapter->fallback.tx_queues;
2136                 adapter->desired.rx_entries = adapter->fallback.rx_entries;
2137                 adapter->desired.tx_entries = adapter->fallback.tx_entries;
2138
2139                 init_completion(&adapter->reset_done);
2140                 adapter->wait_for_reset = true;
2141                 rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2142                 if (rc)
2143                         return ret;
2144                 wait_for_completion(&adapter->reset_done);
2145         }
2146         adapter->wait_for_reset = false;
2147
2148         return ret;
2149 }
2150
2151 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2152 {
2153         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2154
2155         adapter->desired.mtu = new_mtu + ETH_HLEN;
2156
2157         return wait_for_reset(adapter);
2158 }
2159
2160 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2161                                                 struct net_device *dev,
2162                                                 netdev_features_t features)
2163 {
2164         /* Some backing hardware adapters can not
2165          * handle packets with a MSS less than 224
2166          * or with only one segment.
2167          */
2168         if (skb_is_gso(skb)) {
2169                 if (skb_shinfo(skb)->gso_size < 224 ||
2170                     skb_shinfo(skb)->gso_segs == 1)
2171                         features &= ~NETIF_F_GSO_MASK;
2172         }
2173
2174         return features;
2175 }
2176
2177 static const struct net_device_ops ibmvnic_netdev_ops = {
2178         .ndo_open               = ibmvnic_open,
2179         .ndo_stop               = ibmvnic_close,
2180         .ndo_start_xmit         = ibmvnic_xmit,
2181         .ndo_set_rx_mode        = ibmvnic_set_multi,
2182         .ndo_set_mac_address    = ibmvnic_set_mac,
2183         .ndo_validate_addr      = eth_validate_addr,
2184         .ndo_tx_timeout         = ibmvnic_tx_timeout,
2185 #ifdef CONFIG_NET_POLL_CONTROLLER
2186         .ndo_poll_controller    = ibmvnic_netpoll_controller,
2187 #endif
2188         .ndo_change_mtu         = ibmvnic_change_mtu,
2189         .ndo_features_check     = ibmvnic_features_check,
2190 };
2191
2192 /* ethtool functions */
2193
2194 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2195                                       struct ethtool_link_ksettings *cmd)
2196 {
2197         u32 supported, advertising;
2198
2199         supported = (SUPPORTED_1000baseT_Full | SUPPORTED_Autoneg |
2200                           SUPPORTED_FIBRE);
2201         advertising = (ADVERTISED_1000baseT_Full | ADVERTISED_Autoneg |
2202                             ADVERTISED_FIBRE);
2203         cmd->base.speed = SPEED_1000;
2204         cmd->base.duplex = DUPLEX_FULL;
2205         cmd->base.port = PORT_FIBRE;
2206         cmd->base.phy_address = 0;
2207         cmd->base.autoneg = AUTONEG_ENABLE;
2208
2209         ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.supported,
2210                                                 supported);
2211         ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.advertising,
2212                                                 advertising);
2213
2214         return 0;
2215 }
2216
2217 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2218                                 struct ethtool_drvinfo *info)
2219 {
2220         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2221
2222         strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2223         strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2224         strlcpy(info->fw_version, adapter->fw_version,
2225                 sizeof(info->fw_version));
2226 }
2227
2228 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2229 {
2230         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2231
2232         return adapter->msg_enable;
2233 }
2234
2235 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2236 {
2237         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2238
2239         adapter->msg_enable = data;
2240 }
2241
2242 static u32 ibmvnic_get_link(struct net_device *netdev)
2243 {
2244         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2245
2246         /* Don't need to send a query because we request a logical link up at
2247          * init and then we wait for link state indications
2248          */
2249         return adapter->logical_link_state;
2250 }
2251
2252 static void ibmvnic_get_ringparam(struct net_device *netdev,
2253                                   struct ethtool_ringparam *ring)
2254 {
2255         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2256
2257         ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2258         ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2259         ring->rx_mini_max_pending = 0;
2260         ring->rx_jumbo_max_pending = 0;
2261         ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2262         ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2263         ring->rx_mini_pending = 0;
2264         ring->rx_jumbo_pending = 0;
2265 }
2266
2267 static int ibmvnic_set_ringparam(struct net_device *netdev,
2268                                  struct ethtool_ringparam *ring)
2269 {
2270         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2271
2272         if (ring->rx_pending > adapter->max_rx_add_entries_per_subcrq  ||
2273             ring->tx_pending > adapter->max_tx_entries_per_subcrq) {
2274                 netdev_err(netdev, "Invalid request.\n");
2275                 netdev_err(netdev, "Max tx buffers = %llu\n",
2276                            adapter->max_rx_add_entries_per_subcrq);
2277                 netdev_err(netdev, "Max rx buffers = %llu\n",
2278                            adapter->max_tx_entries_per_subcrq);
2279                 return -EINVAL;
2280         }
2281
2282         adapter->desired.rx_entries = ring->rx_pending;
2283         adapter->desired.tx_entries = ring->tx_pending;
2284
2285         return wait_for_reset(adapter);
2286 }
2287
2288 static void ibmvnic_get_channels(struct net_device *netdev,
2289                                  struct ethtool_channels *channels)
2290 {
2291         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2292
2293         channels->max_rx = adapter->max_rx_queues;
2294         channels->max_tx = adapter->max_tx_queues;
2295         channels->max_other = 0;
2296         channels->max_combined = 0;
2297         channels->rx_count = adapter->req_rx_queues;
2298         channels->tx_count = adapter->req_tx_queues;
2299         channels->other_count = 0;
2300         channels->combined_count = 0;
2301 }
2302
2303 static int ibmvnic_set_channels(struct net_device *netdev,
2304                                 struct ethtool_channels *channels)
2305 {
2306         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2307
2308         adapter->desired.rx_queues = channels->rx_count;
2309         adapter->desired.tx_queues = channels->tx_count;
2310
2311         return wait_for_reset(adapter);
2312 }
2313
2314 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2315 {
2316         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2317         int i;
2318
2319         if (stringset != ETH_SS_STATS)
2320                 return;
2321
2322         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++, data += ETH_GSTRING_LEN)
2323                 memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2324
2325         for (i = 0; i < adapter->req_tx_queues; i++) {
2326                 snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2327                 data += ETH_GSTRING_LEN;
2328
2329                 snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2330                 data += ETH_GSTRING_LEN;
2331
2332                 snprintf(data, ETH_GSTRING_LEN, "tx%d_dropped_packets", i);
2333                 data += ETH_GSTRING_LEN;
2334         }
2335
2336         for (i = 0; i < adapter->req_rx_queues; i++) {
2337                 snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2338                 data += ETH_GSTRING_LEN;
2339
2340                 snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2341                 data += ETH_GSTRING_LEN;
2342
2343                 snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2344                 data += ETH_GSTRING_LEN;
2345         }
2346 }
2347
2348 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2349 {
2350         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2351
2352         switch (sset) {
2353         case ETH_SS_STATS:
2354                 return ARRAY_SIZE(ibmvnic_stats) +
2355                        adapter->req_tx_queues * NUM_TX_STATS +
2356                        adapter->req_rx_queues * NUM_RX_STATS;
2357         default:
2358                 return -EOPNOTSUPP;
2359         }
2360 }
2361
2362 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2363                                       struct ethtool_stats *stats, u64 *data)
2364 {
2365         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2366         union ibmvnic_crq crq;
2367         int i, j;
2368
2369         memset(&crq, 0, sizeof(crq));
2370         crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2371         crq.request_statistics.cmd = REQUEST_STATISTICS;
2372         crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2373         crq.request_statistics.len =
2374             cpu_to_be32(sizeof(struct ibmvnic_statistics));
2375
2376         /* Wait for data to be written */
2377         init_completion(&adapter->stats_done);
2378         ibmvnic_send_crq(adapter, &crq);
2379         wait_for_completion(&adapter->stats_done);
2380
2381         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2382                 data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2383                                                 ibmvnic_stats[i].offset));
2384
2385         for (j = 0; j < adapter->req_tx_queues; j++) {
2386                 data[i] = adapter->tx_stats_buffers[j].packets;
2387                 i++;
2388                 data[i] = adapter->tx_stats_buffers[j].bytes;
2389                 i++;
2390                 data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2391                 i++;
2392         }
2393
2394         for (j = 0; j < adapter->req_rx_queues; j++) {
2395                 data[i] = adapter->rx_stats_buffers[j].packets;
2396                 i++;
2397                 data[i] = adapter->rx_stats_buffers[j].bytes;
2398                 i++;
2399                 data[i] = adapter->rx_stats_buffers[j].interrupts;
2400                 i++;
2401         }
2402 }
2403
2404 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2405         .get_drvinfo            = ibmvnic_get_drvinfo,
2406         .get_msglevel           = ibmvnic_get_msglevel,
2407         .set_msglevel           = ibmvnic_set_msglevel,
2408         .get_link               = ibmvnic_get_link,
2409         .get_ringparam          = ibmvnic_get_ringparam,
2410         .set_ringparam          = ibmvnic_set_ringparam,
2411         .get_channels           = ibmvnic_get_channels,
2412         .set_channels           = ibmvnic_set_channels,
2413         .get_strings            = ibmvnic_get_strings,
2414         .get_sset_count         = ibmvnic_get_sset_count,
2415         .get_ethtool_stats      = ibmvnic_get_ethtool_stats,
2416         .get_link_ksettings     = ibmvnic_get_link_ksettings,
2417 };
2418
2419 /* Routines for managing CRQs/sCRQs  */
2420
2421 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2422                                    struct ibmvnic_sub_crq_queue *scrq)
2423 {
2424         int rc;
2425
2426         if (scrq->irq) {
2427                 free_irq(scrq->irq, scrq);
2428                 irq_dispose_mapping(scrq->irq);
2429                 scrq->irq = 0;
2430         }
2431
2432         memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2433         atomic_set(&scrq->used, 0);
2434         scrq->cur = 0;
2435
2436         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2437                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2438         return rc;
2439 }
2440
2441 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2442 {
2443         int i, rc;
2444
2445         for (i = 0; i < adapter->req_tx_queues; i++) {
2446                 netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2447                 rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2448                 if (rc)
2449                         return rc;
2450         }
2451
2452         for (i = 0; i < adapter->req_rx_queues; i++) {
2453                 netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2454                 rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2455                 if (rc)
2456                         return rc;
2457         }
2458
2459         return rc;
2460 }
2461
2462 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2463                                   struct ibmvnic_sub_crq_queue *scrq,
2464                                   bool do_h_free)
2465 {
2466         struct device *dev = &adapter->vdev->dev;
2467         long rc;
2468
2469         netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2470
2471         if (do_h_free) {
2472                 /* Close the sub-crqs */
2473                 do {
2474                         rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2475                                                 adapter->vdev->unit_address,
2476                                                 scrq->crq_num);
2477                 } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2478
2479                 if (rc) {
2480                         netdev_err(adapter->netdev,
2481                                    "Failed to release sub-CRQ %16lx, rc = %ld\n",
2482                                    scrq->crq_num, rc);
2483                 }
2484         }
2485
2486         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2487                          DMA_BIDIRECTIONAL);
2488         free_pages((unsigned long)scrq->msgs, 2);
2489         kfree(scrq);
2490 }
2491
2492 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2493                                                         *adapter)
2494 {
2495         struct device *dev = &adapter->vdev->dev;
2496         struct ibmvnic_sub_crq_queue *scrq;
2497         int rc;
2498
2499         scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2500         if (!scrq)
2501                 return NULL;
2502
2503         scrq->msgs =
2504                 (union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2505         if (!scrq->msgs) {
2506                 dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2507                 goto zero_page_failed;
2508         }
2509
2510         scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2511                                          DMA_BIDIRECTIONAL);
2512         if (dma_mapping_error(dev, scrq->msg_token)) {
2513                 dev_warn(dev, "Couldn't map crq queue messages page\n");
2514                 goto map_failed;
2515         }
2516
2517         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2518                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2519
2520         if (rc == H_RESOURCE)
2521                 rc = ibmvnic_reset_crq(adapter);
2522
2523         if (rc == H_CLOSED) {
2524                 dev_warn(dev, "Partner adapter not ready, waiting.\n");
2525         } else if (rc) {
2526                 dev_warn(dev, "Error %d registering sub-crq\n", rc);
2527                 goto reg_failed;
2528         }
2529
2530         scrq->adapter = adapter;
2531         scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2532         spin_lock_init(&scrq->lock);
2533
2534         netdev_dbg(adapter->netdev,
2535                    "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2536                    scrq->crq_num, scrq->hw_irq, scrq->irq);
2537
2538         return scrq;
2539
2540 reg_failed:
2541         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2542                          DMA_BIDIRECTIONAL);
2543 map_failed:
2544         free_pages((unsigned long)scrq->msgs, 2);
2545 zero_page_failed:
2546         kfree(scrq);
2547
2548         return NULL;
2549 }
2550
2551 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
2552 {
2553         int i;
2554
2555         if (adapter->tx_scrq) {
2556                 for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
2557                         if (!adapter->tx_scrq[i])
2558                                 continue;
2559
2560                         netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2561                                    i);
2562                         if (adapter->tx_scrq[i]->irq) {
2563                                 free_irq(adapter->tx_scrq[i]->irq,
2564                                          adapter->tx_scrq[i]);
2565                                 irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2566                                 adapter->tx_scrq[i]->irq = 0;
2567                         }
2568
2569                         release_sub_crq_queue(adapter, adapter->tx_scrq[i],
2570                                               do_h_free);
2571                 }
2572
2573                 kfree(adapter->tx_scrq);
2574                 adapter->tx_scrq = NULL;
2575                 adapter->num_active_tx_scrqs = 0;
2576         }
2577
2578         if (adapter->rx_scrq) {
2579                 for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
2580                         if (!adapter->rx_scrq[i])
2581                                 continue;
2582
2583                         netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2584                                    i);
2585                         if (adapter->rx_scrq[i]->irq) {
2586                                 free_irq(adapter->rx_scrq[i]->irq,
2587                                          adapter->rx_scrq[i]);
2588                                 irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2589                                 adapter->rx_scrq[i]->irq = 0;
2590                         }
2591
2592                         release_sub_crq_queue(adapter, adapter->rx_scrq[i],
2593                                               do_h_free);
2594                 }
2595
2596                 kfree(adapter->rx_scrq);
2597                 adapter->rx_scrq = NULL;
2598                 adapter->num_active_rx_scrqs = 0;
2599         }
2600 }
2601
2602 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
2603                             struct ibmvnic_sub_crq_queue *scrq)
2604 {
2605         struct device *dev = &adapter->vdev->dev;
2606         unsigned long rc;
2607
2608         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2609                                 H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2610         if (rc)
2611                 dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
2612                         scrq->hw_irq, rc);
2613         return rc;
2614 }
2615
2616 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
2617                            struct ibmvnic_sub_crq_queue *scrq)
2618 {
2619         struct device *dev = &adapter->vdev->dev;
2620         unsigned long rc;
2621         u64 val;
2622
2623         if (scrq->hw_irq > 0x100000000ULL) {
2624                 dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
2625                 return 1;
2626         }
2627
2628         val = (0xff000000) | scrq->hw_irq;
2629         rc = plpar_hcall_norets(H_EOI, val);
2630         if (rc)
2631                 dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
2632                         val, rc);
2633
2634         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2635                                 H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2636         if (rc)
2637                 dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
2638                         scrq->hw_irq, rc);
2639         return rc;
2640 }
2641
2642 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
2643                                struct ibmvnic_sub_crq_queue *scrq)
2644 {
2645         struct device *dev = &adapter->vdev->dev;
2646         struct ibmvnic_tx_pool *tx_pool;
2647         struct ibmvnic_tx_buff *txbuff;
2648         union sub_crq *next;
2649         int index;
2650         int i, j;
2651         u8 *first;
2652
2653 restart_loop:
2654         while (pending_scrq(adapter, scrq)) {
2655                 unsigned int pool = scrq->pool_index;
2656                 int num_entries = 0;
2657
2658                 next = ibmvnic_next_scrq(adapter, scrq);
2659                 for (i = 0; i < next->tx_comp.num_comps; i++) {
2660                         if (next->tx_comp.rcs[i]) {
2661                                 dev_err(dev, "tx error %x\n",
2662                                         next->tx_comp.rcs[i]);
2663                                 continue;
2664                         }
2665                         index = be32_to_cpu(next->tx_comp.correlators[i]);
2666                         if (index & IBMVNIC_TSO_POOL_MASK) {
2667                                 tx_pool = &adapter->tso_pool[pool];
2668                                 index &= ~IBMVNIC_TSO_POOL_MASK;
2669                         } else {
2670                                 tx_pool = &adapter->tx_pool[pool];
2671                         }
2672
2673                         txbuff = &tx_pool->tx_buff[index];
2674
2675                         for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
2676                                 if (!txbuff->data_dma[j])
2677                                         continue;
2678
2679                                 txbuff->data_dma[j] = 0;
2680                         }
2681                         /* if sub_crq was sent indirectly */
2682                         first = &txbuff->indir_arr[0].generic.first;
2683                         if (*first == IBMVNIC_CRQ_CMD) {
2684                                 dma_unmap_single(dev, txbuff->indir_dma,
2685                                                  sizeof(txbuff->indir_arr),
2686                                                  DMA_TO_DEVICE);
2687                                 *first = 0;
2688                         }
2689
2690                         if (txbuff->last_frag) {
2691                                 dev_kfree_skb_any(txbuff->skb);
2692                                 txbuff->skb = NULL;
2693                         }
2694
2695                         num_entries += txbuff->num_entries;
2696
2697                         tx_pool->free_map[tx_pool->producer_index] = index;
2698                         tx_pool->producer_index =
2699                                 (tx_pool->producer_index + 1) %
2700                                         tx_pool->num_buffers;
2701                 }
2702                 /* remove tx_comp scrq*/
2703                 next->tx_comp.first = 0;
2704
2705                 if (atomic_sub_return(num_entries, &scrq->used) <=
2706                     (adapter->req_tx_entries_per_subcrq / 2) &&
2707                     __netif_subqueue_stopped(adapter->netdev,
2708                                              scrq->pool_index)) {
2709                         netif_wake_subqueue(adapter->netdev, scrq->pool_index);
2710                         netdev_dbg(adapter->netdev, "Started queue %d\n",
2711                                    scrq->pool_index);
2712                 }
2713         }
2714
2715         enable_scrq_irq(adapter, scrq);
2716
2717         if (pending_scrq(adapter, scrq)) {
2718                 disable_scrq_irq(adapter, scrq);
2719                 goto restart_loop;
2720         }
2721
2722         return 0;
2723 }
2724
2725 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
2726 {
2727         struct ibmvnic_sub_crq_queue *scrq = instance;
2728         struct ibmvnic_adapter *adapter = scrq->adapter;
2729
2730         disable_scrq_irq(adapter, scrq);
2731         ibmvnic_complete_tx(adapter, scrq);
2732
2733         return IRQ_HANDLED;
2734 }
2735
2736 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
2737 {
2738         struct ibmvnic_sub_crq_queue *scrq = instance;
2739         struct ibmvnic_adapter *adapter = scrq->adapter;
2740
2741         /* When booting a kdump kernel we can hit pending interrupts
2742          * prior to completing driver initialization.
2743          */
2744         if (unlikely(adapter->state != VNIC_OPEN))
2745                 return IRQ_NONE;
2746
2747         adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
2748
2749         if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
2750                 disable_scrq_irq(adapter, scrq);
2751                 __napi_schedule(&adapter->napi[scrq->scrq_num]);
2752         }
2753
2754         return IRQ_HANDLED;
2755 }
2756
2757 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
2758 {
2759         struct device *dev = &adapter->vdev->dev;
2760         struct ibmvnic_sub_crq_queue *scrq;
2761         int i = 0, j = 0;
2762         int rc = 0;
2763
2764         for (i = 0; i < adapter->req_tx_queues; i++) {
2765                 netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
2766                            i);
2767                 scrq = adapter->tx_scrq[i];
2768                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2769
2770                 if (!scrq->irq) {
2771                         rc = -EINVAL;
2772                         dev_err(dev, "Error mapping irq\n");
2773                         goto req_tx_irq_failed;
2774                 }
2775
2776                 rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
2777                                  0, "ibmvnic_tx", scrq);
2778
2779                 if (rc) {
2780                         dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
2781                                 scrq->irq, rc);
2782                         irq_dispose_mapping(scrq->irq);
2783                         goto req_tx_irq_failed;
2784                 }
2785         }
2786
2787         for (i = 0; i < adapter->req_rx_queues; i++) {
2788                 netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
2789                            i);
2790                 scrq = adapter->rx_scrq[i];
2791                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2792                 if (!scrq->irq) {
2793                         rc = -EINVAL;
2794                         dev_err(dev, "Error mapping irq\n");
2795                         goto req_rx_irq_failed;
2796                 }
2797                 rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
2798                                  0, "ibmvnic_rx", scrq);
2799                 if (rc) {
2800                         dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
2801                                 scrq->irq, rc);
2802                         irq_dispose_mapping(scrq->irq);
2803                         goto req_rx_irq_failed;
2804                 }
2805         }
2806         return rc;
2807
2808 req_rx_irq_failed:
2809         for (j = 0; j < i; j++) {
2810                 free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
2811                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2812         }
2813         i = adapter->req_tx_queues;
2814 req_tx_irq_failed:
2815         for (j = 0; j < i; j++) {
2816                 free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
2817                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2818         }
2819         release_sub_crqs(adapter, 1);
2820         return rc;
2821 }
2822
2823 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
2824 {
2825         struct device *dev = &adapter->vdev->dev;
2826         struct ibmvnic_sub_crq_queue **allqueues;
2827         int registered_queues = 0;
2828         int total_queues;
2829         int more = 0;
2830         int i;
2831
2832         total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
2833
2834         allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
2835         if (!allqueues)
2836                 return -1;
2837
2838         for (i = 0; i < total_queues; i++) {
2839                 allqueues[i] = init_sub_crq_queue(adapter);
2840                 if (!allqueues[i]) {
2841                         dev_warn(dev, "Couldn't allocate all sub-crqs\n");
2842                         break;
2843                 }
2844                 registered_queues++;
2845         }
2846
2847         /* Make sure we were able to register the minimum number of queues */
2848         if (registered_queues <
2849             adapter->min_tx_queues + adapter->min_rx_queues) {
2850                 dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
2851                 goto tx_failed;
2852         }
2853
2854         /* Distribute the failed allocated queues*/
2855         for (i = 0; i < total_queues - registered_queues + more ; i++) {
2856                 netdev_dbg(adapter->netdev, "Reducing number of queues\n");
2857                 switch (i % 3) {
2858                 case 0:
2859                         if (adapter->req_rx_queues > adapter->min_rx_queues)
2860                                 adapter->req_rx_queues--;
2861                         else
2862                                 more++;
2863                         break;
2864                 case 1:
2865                         if (adapter->req_tx_queues > adapter->min_tx_queues)
2866                                 adapter->req_tx_queues--;
2867                         else
2868                                 more++;
2869                         break;
2870                 }
2871         }
2872
2873         adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
2874                                    sizeof(*adapter->tx_scrq), GFP_KERNEL);
2875         if (!adapter->tx_scrq)
2876                 goto tx_failed;
2877
2878         for (i = 0; i < adapter->req_tx_queues; i++) {
2879                 adapter->tx_scrq[i] = allqueues[i];
2880                 adapter->tx_scrq[i]->pool_index = i;
2881                 adapter->num_active_tx_scrqs++;
2882         }
2883
2884         adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
2885                                    sizeof(*adapter->rx_scrq), GFP_KERNEL);
2886         if (!adapter->rx_scrq)
2887                 goto rx_failed;
2888
2889         for (i = 0; i < adapter->req_rx_queues; i++) {
2890                 adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
2891                 adapter->rx_scrq[i]->scrq_num = i;
2892                 adapter->num_active_rx_scrqs++;
2893         }
2894
2895         kfree(allqueues);
2896         return 0;
2897
2898 rx_failed:
2899         kfree(adapter->tx_scrq);
2900         adapter->tx_scrq = NULL;
2901 tx_failed:
2902         for (i = 0; i < registered_queues; i++)
2903                 release_sub_crq_queue(adapter, allqueues[i], 1);
2904         kfree(allqueues);
2905         return -1;
2906 }
2907
2908 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
2909 {
2910         struct device *dev = &adapter->vdev->dev;
2911         union ibmvnic_crq crq;
2912         int max_entries;
2913
2914         if (!retry) {
2915                 /* Sub-CRQ entries are 32 byte long */
2916                 int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
2917
2918                 if (adapter->min_tx_entries_per_subcrq > entries_page ||
2919                     adapter->min_rx_add_entries_per_subcrq > entries_page) {
2920                         dev_err(dev, "Fatal, invalid entries per sub-crq\n");
2921                         return;
2922                 }
2923
2924                 if (adapter->desired.mtu)
2925                         adapter->req_mtu = adapter->desired.mtu;
2926                 else
2927                         adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
2928
2929                 if (!adapter->desired.tx_entries)
2930                         adapter->desired.tx_entries =
2931                                         adapter->max_tx_entries_per_subcrq;
2932                 if (!adapter->desired.rx_entries)
2933                         adapter->desired.rx_entries =
2934                                         adapter->max_rx_add_entries_per_subcrq;
2935
2936                 max_entries = IBMVNIC_MAX_LTB_SIZE /
2937                               (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
2938
2939                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
2940                         adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
2941                         adapter->desired.tx_entries = max_entries;
2942                 }
2943
2944                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
2945                         adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
2946                         adapter->desired.rx_entries = max_entries;
2947                 }
2948
2949                 if (adapter->desired.tx_entries)
2950                         adapter->req_tx_entries_per_subcrq =
2951                                         adapter->desired.tx_entries;
2952                 else
2953                         adapter->req_tx_entries_per_subcrq =
2954                                         adapter->max_tx_entries_per_subcrq;
2955
2956                 if (adapter->desired.rx_entries)
2957                         adapter->req_rx_add_entries_per_subcrq =
2958                                         adapter->desired.rx_entries;
2959                 else
2960                         adapter->req_rx_add_entries_per_subcrq =
2961                                         adapter->max_rx_add_entries_per_subcrq;
2962
2963                 if (adapter->desired.tx_queues)
2964                         adapter->req_tx_queues =
2965                                         adapter->desired.tx_queues;
2966                 else
2967                         adapter->req_tx_queues =
2968                                         adapter->opt_tx_comp_sub_queues;
2969
2970                 if (adapter->desired.rx_queues)
2971                         adapter->req_rx_queues =
2972                                         adapter->desired.rx_queues;
2973                 else
2974                         adapter->req_rx_queues =
2975                                         adapter->opt_rx_comp_queues;
2976
2977                 adapter->req_rx_add_queues = adapter->max_rx_add_queues;
2978         }
2979
2980         memset(&crq, 0, sizeof(crq));
2981         crq.request_capability.first = IBMVNIC_CRQ_CMD;
2982         crq.request_capability.cmd = REQUEST_CAPABILITY;
2983
2984         crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
2985         crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
2986         atomic_inc(&adapter->running_cap_crqs);
2987         ibmvnic_send_crq(adapter, &crq);
2988
2989         crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
2990         crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
2991         atomic_inc(&adapter->running_cap_crqs);
2992         ibmvnic_send_crq(adapter, &crq);
2993
2994         crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
2995         crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
2996         atomic_inc(&adapter->running_cap_crqs);
2997         ibmvnic_send_crq(adapter, &crq);
2998
2999         crq.request_capability.capability =
3000             cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3001         crq.request_capability.number =
3002             cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3003         atomic_inc(&adapter->running_cap_crqs);
3004         ibmvnic_send_crq(adapter, &crq);
3005
3006         crq.request_capability.capability =
3007             cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3008         crq.request_capability.number =
3009             cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3010         atomic_inc(&adapter->running_cap_crqs);
3011         ibmvnic_send_crq(adapter, &crq);
3012
3013         crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3014         crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3015         atomic_inc(&adapter->running_cap_crqs);
3016         ibmvnic_send_crq(adapter, &crq);
3017
3018         if (adapter->netdev->flags & IFF_PROMISC) {
3019                 if (adapter->promisc_supported) {
3020                         crq.request_capability.capability =
3021                             cpu_to_be16(PROMISC_REQUESTED);
3022                         crq.request_capability.number = cpu_to_be64(1);
3023                         atomic_inc(&adapter->running_cap_crqs);
3024                         ibmvnic_send_crq(adapter, &crq);
3025                 }
3026         } else {
3027                 crq.request_capability.capability =
3028                     cpu_to_be16(PROMISC_REQUESTED);
3029                 crq.request_capability.number = cpu_to_be64(0);
3030                 atomic_inc(&adapter->running_cap_crqs);
3031                 ibmvnic_send_crq(adapter, &crq);
3032         }
3033 }
3034
3035 static int pending_scrq(struct ibmvnic_adapter *adapter,
3036                         struct ibmvnic_sub_crq_queue *scrq)
3037 {
3038         union sub_crq *entry = &scrq->msgs[scrq->cur];
3039
3040         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
3041                 return 1;
3042         else
3043                 return 0;
3044 }
3045
3046 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3047                                         struct ibmvnic_sub_crq_queue *scrq)
3048 {
3049         union sub_crq *entry;
3050         unsigned long flags;
3051
3052         spin_lock_irqsave(&scrq->lock, flags);
3053         entry = &scrq->msgs[scrq->cur];
3054         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3055                 if (++scrq->cur == scrq->size)
3056                         scrq->cur = 0;
3057         } else {
3058                 entry = NULL;
3059         }
3060         spin_unlock_irqrestore(&scrq->lock, flags);
3061
3062         return entry;
3063 }
3064
3065 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3066 {
3067         struct ibmvnic_crq_queue *queue = &adapter->crq;
3068         union ibmvnic_crq *crq;
3069
3070         crq = &queue->msgs[queue->cur];
3071         if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3072                 if (++queue->cur == queue->size)
3073                         queue->cur = 0;
3074         } else {
3075                 crq = NULL;
3076         }
3077
3078         return crq;
3079 }
3080
3081 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
3082                        union sub_crq *sub_crq)
3083 {
3084         unsigned int ua = adapter->vdev->unit_address;
3085         struct device *dev = &adapter->vdev->dev;
3086         u64 *u64_crq = (u64 *)sub_crq;
3087         int rc;
3088
3089         netdev_dbg(adapter->netdev,
3090                    "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
3091                    (unsigned long int)cpu_to_be64(remote_handle),
3092                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3093                    (unsigned long int)cpu_to_be64(u64_crq[1]),
3094                    (unsigned long int)cpu_to_be64(u64_crq[2]),
3095                    (unsigned long int)cpu_to_be64(u64_crq[3]));
3096
3097         /* Make sure the hypervisor sees the complete request */
3098         mb();
3099
3100         rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
3101                                 cpu_to_be64(remote_handle),
3102                                 cpu_to_be64(u64_crq[0]),
3103                                 cpu_to_be64(u64_crq[1]),
3104                                 cpu_to_be64(u64_crq[2]),
3105                                 cpu_to_be64(u64_crq[3]));
3106
3107         if (rc) {
3108                 if (rc == H_CLOSED)
3109                         dev_warn(dev, "CRQ Queue closed\n");
3110                 dev_err(dev, "Send error (rc=%d)\n", rc);
3111         }
3112
3113         return rc;
3114 }
3115
3116 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3117                                 u64 remote_handle, u64 ioba, u64 num_entries)
3118 {
3119         unsigned int ua = adapter->vdev->unit_address;
3120         struct device *dev = &adapter->vdev->dev;
3121         int rc;
3122
3123         /* Make sure the hypervisor sees the complete request */
3124         mb();
3125         rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3126                                 cpu_to_be64(remote_handle),
3127                                 ioba, num_entries);
3128
3129         if (rc) {
3130                 if (rc == H_CLOSED)
3131                         dev_warn(dev, "CRQ Queue closed\n");
3132                 dev_err(dev, "Send (indirect) error (rc=%d)\n", rc);
3133         }
3134
3135         return rc;
3136 }
3137
3138 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3139                             union ibmvnic_crq *crq)
3140 {
3141         unsigned int ua = adapter->vdev->unit_address;
3142         struct device *dev = &adapter->vdev->dev;
3143         u64 *u64_crq = (u64 *)crq;
3144         int rc;
3145
3146         netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3147                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3148                    (unsigned long int)cpu_to_be64(u64_crq[1]));
3149
3150         if (!adapter->crq.active &&
3151             crq->generic.first != IBMVNIC_CRQ_INIT_CMD) {
3152                 dev_warn(dev, "Invalid request detected while CRQ is inactive, possible device state change during reset\n");
3153                 return -EINVAL;
3154         }
3155
3156         /* Make sure the hypervisor sees the complete request */
3157         mb();
3158
3159         rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3160                                 cpu_to_be64(u64_crq[0]),
3161                                 cpu_to_be64(u64_crq[1]));
3162
3163         if (rc) {
3164                 if (rc == H_CLOSED) {
3165                         dev_warn(dev, "CRQ Queue closed\n");
3166                         if (adapter->resetting)
3167                                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3168                 }
3169
3170                 dev_warn(dev, "Send error (rc=%d)\n", rc);
3171         }
3172
3173         return rc;
3174 }
3175
3176 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3177 {
3178         union ibmvnic_crq crq;
3179
3180         memset(&crq, 0, sizeof(crq));
3181         crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3182         crq.generic.cmd = IBMVNIC_CRQ_INIT;
3183         netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3184
3185         return ibmvnic_send_crq(adapter, &crq);
3186 }
3187
3188 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3189 {
3190         union ibmvnic_crq crq;
3191
3192         memset(&crq, 0, sizeof(crq));
3193         crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3194         crq.version_exchange.cmd = VERSION_EXCHANGE;
3195         crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3196
3197         return ibmvnic_send_crq(adapter, &crq);
3198 }
3199
3200 struct vnic_login_client_data {
3201         u8      type;
3202         __be16  len;
3203         char    name[];
3204 } __packed;
3205
3206 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3207 {
3208         int len;
3209
3210         /* Calculate the amount of buffer space needed for the
3211          * vnic client data in the login buffer. There are four entries,
3212          * OS name, LPAR name, device name, and a null last entry.
3213          */
3214         len = 4 * sizeof(struct vnic_login_client_data);
3215         len += 6; /* "Linux" plus NULL */
3216         len += strlen(utsname()->nodename) + 1;
3217         len += strlen(adapter->netdev->name) + 1;
3218
3219         return len;
3220 }
3221
3222 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3223                                  struct vnic_login_client_data *vlcd)
3224 {
3225         const char *os_name = "Linux";
3226         int len;
3227
3228         /* Type 1 - LPAR OS */
3229         vlcd->type = 1;
3230         len = strlen(os_name) + 1;
3231         vlcd->len = cpu_to_be16(len);
3232         strncpy(vlcd->name, os_name, len);
3233         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3234
3235         /* Type 2 - LPAR name */
3236         vlcd->type = 2;
3237         len = strlen(utsname()->nodename) + 1;
3238         vlcd->len = cpu_to_be16(len);
3239         strncpy(vlcd->name, utsname()->nodename, len);
3240         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3241
3242         /* Type 3 - device name */
3243         vlcd->type = 3;
3244         len = strlen(adapter->netdev->name) + 1;
3245         vlcd->len = cpu_to_be16(len);
3246         strncpy(vlcd->name, adapter->netdev->name, len);
3247 }
3248
3249 static int send_login(struct ibmvnic_adapter *adapter)
3250 {
3251         struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3252         struct ibmvnic_login_buffer *login_buffer;
3253         struct device *dev = &adapter->vdev->dev;
3254         dma_addr_t rsp_buffer_token;
3255         dma_addr_t buffer_token;
3256         size_t rsp_buffer_size;
3257         union ibmvnic_crq crq;
3258         size_t buffer_size;
3259         __be64 *tx_list_p;
3260         __be64 *rx_list_p;
3261         int client_data_len;
3262         struct vnic_login_client_data *vlcd;
3263         int i;
3264
3265         if (!adapter->tx_scrq || !adapter->rx_scrq) {
3266                 netdev_err(adapter->netdev,
3267                            "RX or TX queues are not allocated, device login failed\n");
3268                 return -1;
3269         }
3270
3271         release_login_rsp_buffer(adapter);
3272         client_data_len = vnic_client_data_len(adapter);
3273
3274         buffer_size =
3275             sizeof(struct ibmvnic_login_buffer) +
3276             sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3277             client_data_len;
3278
3279         login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3280         if (!login_buffer)
3281                 goto buf_alloc_failed;
3282
3283         buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3284                                       DMA_TO_DEVICE);
3285         if (dma_mapping_error(dev, buffer_token)) {
3286                 dev_err(dev, "Couldn't map login buffer\n");
3287                 goto buf_map_failed;
3288         }
3289
3290         rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3291                           sizeof(u64) * adapter->req_tx_queues +
3292                           sizeof(u64) * adapter->req_rx_queues +
3293                           sizeof(u64) * adapter->req_rx_queues +
3294                           sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3295
3296         login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3297         if (!login_rsp_buffer)
3298                 goto buf_rsp_alloc_failed;
3299
3300         rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3301                                           rsp_buffer_size, DMA_FROM_DEVICE);
3302         if (dma_mapping_error(dev, rsp_buffer_token)) {
3303                 dev_err(dev, "Couldn't map login rsp buffer\n");
3304                 goto buf_rsp_map_failed;
3305         }
3306
3307         adapter->login_buf = login_buffer;
3308         adapter->login_buf_token = buffer_token;
3309         adapter->login_buf_sz = buffer_size;
3310         adapter->login_rsp_buf = login_rsp_buffer;
3311         adapter->login_rsp_buf_token = rsp_buffer_token;
3312         adapter->login_rsp_buf_sz = rsp_buffer_size;
3313
3314         login_buffer->len = cpu_to_be32(buffer_size);
3315         login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3316         login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3317         login_buffer->off_txcomp_subcrqs =
3318             cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3319         login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3320         login_buffer->off_rxcomp_subcrqs =
3321             cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3322                         sizeof(u64) * adapter->req_tx_queues);
3323         login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3324         login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3325
3326         tx_list_p = (__be64 *)((char *)login_buffer +
3327                                       sizeof(struct ibmvnic_login_buffer));
3328         rx_list_p = (__be64 *)((char *)login_buffer +
3329                                       sizeof(struct ibmvnic_login_buffer) +
3330                                       sizeof(u64) * adapter->req_tx_queues);
3331
3332         for (i = 0; i < adapter->req_tx_queues; i++) {
3333                 if (adapter->tx_scrq[i]) {
3334                         tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3335                                                    crq_num);
3336                 }
3337         }
3338
3339         for (i = 0; i < adapter->req_rx_queues; i++) {
3340                 if (adapter->rx_scrq[i]) {
3341                         rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3342                                                    crq_num);
3343                 }
3344         }
3345
3346         /* Insert vNIC login client data */
3347         vlcd = (struct vnic_login_client_data *)
3348                 ((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3349         login_buffer->client_data_offset =
3350                         cpu_to_be32((char *)vlcd - (char *)login_buffer);
3351         login_buffer->client_data_len = cpu_to_be32(client_data_len);
3352
3353         vnic_add_client_data(adapter, vlcd);
3354
3355         netdev_dbg(adapter->netdev, "Login Buffer:\n");
3356         for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3357                 netdev_dbg(adapter->netdev, "%016lx\n",
3358                            ((unsigned long int *)(adapter->login_buf))[i]);
3359         }
3360
3361         memset(&crq, 0, sizeof(crq));
3362         crq.login.first = IBMVNIC_CRQ_CMD;
3363         crq.login.cmd = LOGIN;
3364         crq.login.ioba = cpu_to_be32(buffer_token);
3365         crq.login.len = cpu_to_be32(buffer_size);
3366         ibmvnic_send_crq(adapter, &crq);
3367
3368         return 0;
3369
3370 buf_rsp_map_failed:
3371         kfree(login_rsp_buffer);
3372 buf_rsp_alloc_failed:
3373         dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3374 buf_map_failed:
3375         kfree(login_buffer);
3376 buf_alloc_failed:
3377         return -1;
3378 }
3379
3380 static void send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3381                              u32 len, u8 map_id)
3382 {
3383         union ibmvnic_crq crq;
3384
3385         memset(&crq, 0, sizeof(crq));
3386         crq.request_map.first = IBMVNIC_CRQ_CMD;
3387         crq.request_map.cmd = REQUEST_MAP;
3388         crq.request_map.map_id = map_id;
3389         crq.request_map.ioba = cpu_to_be32(addr);
3390         crq.request_map.len = cpu_to_be32(len);
3391         ibmvnic_send_crq(adapter, &crq);
3392 }
3393
3394 static void send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3395 {
3396         union ibmvnic_crq crq;
3397
3398         memset(&crq, 0, sizeof(crq));
3399         crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3400         crq.request_unmap.cmd = REQUEST_UNMAP;
3401         crq.request_unmap.map_id = map_id;
3402         ibmvnic_send_crq(adapter, &crq);
3403 }
3404
3405 static void send_map_query(struct ibmvnic_adapter *adapter)
3406 {
3407         union ibmvnic_crq crq;
3408
3409         memset(&crq, 0, sizeof(crq));
3410         crq.query_map.first = IBMVNIC_CRQ_CMD;
3411         crq.query_map.cmd = QUERY_MAP;
3412         ibmvnic_send_crq(adapter, &crq);
3413 }
3414
3415 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3416 static void send_cap_queries(struct ibmvnic_adapter *adapter)
3417 {
3418         union ibmvnic_crq crq;
3419
3420         atomic_set(&adapter->running_cap_crqs, 0);
3421         memset(&crq, 0, sizeof(crq));
3422         crq.query_capability.first = IBMVNIC_CRQ_CMD;
3423         crq.query_capability.cmd = QUERY_CAPABILITY;
3424
3425         crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3426         atomic_inc(&adapter->running_cap_crqs);
3427         ibmvnic_send_crq(adapter, &crq);
3428
3429         crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3430         atomic_inc(&adapter->running_cap_crqs);
3431         ibmvnic_send_crq(adapter, &crq);
3432
3433         crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3434         atomic_inc(&adapter->running_cap_crqs);
3435         ibmvnic_send_crq(adapter, &crq);
3436
3437         crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3438         atomic_inc(&adapter->running_cap_crqs);
3439         ibmvnic_send_crq(adapter, &crq);
3440
3441         crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3442         atomic_inc(&adapter->running_cap_crqs);
3443         ibmvnic_send_crq(adapter, &crq);
3444
3445         crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3446         atomic_inc(&adapter->running_cap_crqs);
3447         ibmvnic_send_crq(adapter, &crq);
3448
3449         crq.query_capability.capability =
3450             cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3451         atomic_inc(&adapter->running_cap_crqs);
3452         ibmvnic_send_crq(adapter, &crq);
3453
3454         crq.query_capability.capability =
3455             cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3456         atomic_inc(&adapter->running_cap_crqs);
3457         ibmvnic_send_crq(adapter, &crq);
3458
3459         crq.query_capability.capability =
3460             cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3461         atomic_inc(&adapter->running_cap_crqs);
3462         ibmvnic_send_crq(adapter, &crq);
3463
3464         crq.query_capability.capability =
3465             cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3466         atomic_inc(&adapter->running_cap_crqs);
3467         ibmvnic_send_crq(adapter, &crq);
3468
3469         crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3470         atomic_inc(&adapter->running_cap_crqs);
3471         ibmvnic_send_crq(adapter, &crq);
3472
3473         crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3474         atomic_inc(&adapter->running_cap_crqs);
3475         ibmvnic_send_crq(adapter, &crq);
3476
3477         crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3478         atomic_inc(&adapter->running_cap_crqs);
3479         ibmvnic_send_crq(adapter, &crq);
3480
3481         crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3482         atomic_inc(&adapter->running_cap_crqs);
3483         ibmvnic_send_crq(adapter, &crq);
3484
3485         crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3486         atomic_inc(&adapter->running_cap_crqs);
3487         ibmvnic_send_crq(adapter, &crq);
3488
3489         crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3490         atomic_inc(&adapter->running_cap_crqs);
3491         ibmvnic_send_crq(adapter, &crq);
3492
3493         crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3494         atomic_inc(&adapter->running_cap_crqs);
3495         ibmvnic_send_crq(adapter, &crq);
3496
3497         crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3498         atomic_inc(&adapter->running_cap_crqs);
3499         ibmvnic_send_crq(adapter, &crq);
3500
3501         crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3502         atomic_inc(&adapter->running_cap_crqs);
3503         ibmvnic_send_crq(adapter, &crq);
3504
3505         crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3506         atomic_inc(&adapter->running_cap_crqs);
3507         ibmvnic_send_crq(adapter, &crq);
3508
3509         crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3510         atomic_inc(&adapter->running_cap_crqs);
3511         ibmvnic_send_crq(adapter, &crq);
3512
3513         crq.query_capability.capability =
3514                         cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3515         atomic_inc(&adapter->running_cap_crqs);
3516         ibmvnic_send_crq(adapter, &crq);
3517
3518         crq.query_capability.capability =
3519                         cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3520         atomic_inc(&adapter->running_cap_crqs);
3521         ibmvnic_send_crq(adapter, &crq);
3522
3523         crq.query_capability.capability =
3524                         cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3525         atomic_inc(&adapter->running_cap_crqs);
3526         ibmvnic_send_crq(adapter, &crq);
3527
3528         crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3529         atomic_inc(&adapter->running_cap_crqs);
3530         ibmvnic_send_crq(adapter, &crq);
3531 }
3532
3533 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
3534                                 struct ibmvnic_adapter *adapter)
3535 {
3536         struct device *dev = &adapter->vdev->dev;
3537
3538         if (crq->get_vpd_size_rsp.rc.code) {
3539                 dev_err(dev, "Error retrieving VPD size, rc=%x\n",
3540                         crq->get_vpd_size_rsp.rc.code);
3541                 complete(&adapter->fw_done);
3542                 return;
3543         }
3544
3545         adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
3546         complete(&adapter->fw_done);
3547 }
3548
3549 static void handle_vpd_rsp(union ibmvnic_crq *crq,
3550                            struct ibmvnic_adapter *adapter)
3551 {
3552         struct device *dev = &adapter->vdev->dev;
3553         unsigned char *substr = NULL;
3554         u8 fw_level_len = 0;
3555
3556         memset(adapter->fw_version, 0, 32);
3557
3558         dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
3559                          DMA_FROM_DEVICE);
3560
3561         if (crq->get_vpd_rsp.rc.code) {
3562                 dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
3563                         crq->get_vpd_rsp.rc.code);
3564                 goto complete;
3565         }
3566
3567         /* get the position of the firmware version info
3568          * located after the ASCII 'RM' substring in the buffer
3569          */
3570         substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
3571         if (!substr) {
3572                 dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
3573                 goto complete;
3574         }
3575
3576         /* get length of firmware level ASCII substring */
3577         if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
3578                 fw_level_len = *(substr + 2);
3579         } else {
3580                 dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
3581                 goto complete;
3582         }
3583
3584         /* copy firmware version string from vpd into adapter */
3585         if ((substr + 3 + fw_level_len) <
3586             (adapter->vpd->buff + adapter->vpd->len)) {
3587                 strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
3588         } else {
3589                 dev_info(dev, "FW substr extrapolated VPD buff\n");
3590         }
3591
3592 complete:
3593         if (adapter->fw_version[0] == '\0')
3594                 strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
3595         complete(&adapter->fw_done);
3596 }
3597
3598 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
3599 {
3600         struct device *dev = &adapter->vdev->dev;
3601         struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
3602         union ibmvnic_crq crq;
3603         int i;
3604
3605         dma_unmap_single(dev, adapter->ip_offload_tok,
3606                          sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
3607
3608         netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
3609         for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
3610                 netdev_dbg(adapter->netdev, "%016lx\n",
3611                            ((unsigned long int *)(buf))[i]);
3612
3613         netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
3614         netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
3615         netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
3616                    buf->tcp_ipv4_chksum);
3617         netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
3618                    buf->tcp_ipv6_chksum);
3619         netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
3620                    buf->udp_ipv4_chksum);
3621         netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
3622                    buf->udp_ipv6_chksum);
3623         netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
3624                    buf->large_tx_ipv4);
3625         netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
3626                    buf->large_tx_ipv6);
3627         netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
3628                    buf->large_rx_ipv4);
3629         netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
3630                    buf->large_rx_ipv6);
3631         netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
3632                    buf->max_ipv4_header_size);
3633         netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
3634                    buf->max_ipv6_header_size);
3635         netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
3636                    buf->max_tcp_header_size);
3637         netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
3638                    buf->max_udp_header_size);
3639         netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
3640                    buf->max_large_tx_size);
3641         netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
3642                    buf->max_large_rx_size);
3643         netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
3644                    buf->ipv6_extension_header);
3645         netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
3646                    buf->tcp_pseudosum_req);
3647         netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
3648                    buf->num_ipv6_ext_headers);
3649         netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
3650                    buf->off_ipv6_ext_headers);
3651
3652         adapter->ip_offload_ctrl_tok =
3653             dma_map_single(dev, &adapter->ip_offload_ctrl,
3654                            sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
3655
3656         if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
3657                 dev_err(dev, "Couldn't map ip offload control buffer\n");
3658                 return;
3659         }
3660
3661         adapter->ip_offload_ctrl.len =
3662             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3663         adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
3664         adapter->ip_offload_ctrl.ipv4_chksum = buf->ipv4_chksum;
3665         adapter->ip_offload_ctrl.ipv6_chksum = buf->ipv6_chksum;
3666         adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
3667         adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
3668         adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
3669         adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
3670         adapter->ip_offload_ctrl.large_tx_ipv4 = buf->large_tx_ipv4;
3671         adapter->ip_offload_ctrl.large_tx_ipv6 = buf->large_tx_ipv6;
3672
3673         /* large_rx disabled for now, additional features needed */
3674         adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
3675         adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
3676
3677         adapter->netdev->features = NETIF_F_SG | NETIF_F_GSO;
3678
3679         if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
3680                 adapter->netdev->features |= NETIF_F_IP_CSUM;
3681
3682         if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
3683                 adapter->netdev->features |= NETIF_F_IPV6_CSUM;
3684
3685         if ((adapter->netdev->features &
3686             (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
3687                 adapter->netdev->features |= NETIF_F_RXCSUM;
3688
3689         if (buf->large_tx_ipv4)
3690                 adapter->netdev->features |= NETIF_F_TSO;
3691         if (buf->large_tx_ipv6)
3692                 adapter->netdev->features |= NETIF_F_TSO6;
3693
3694         adapter->netdev->hw_features |= adapter->netdev->features;
3695
3696         memset(&crq, 0, sizeof(crq));
3697         crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
3698         crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
3699         crq.control_ip_offload.len =
3700             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3701         crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
3702         ibmvnic_send_crq(adapter, &crq);
3703 }
3704
3705 static void handle_error_info_rsp(union ibmvnic_crq *crq,
3706                                   struct ibmvnic_adapter *adapter)
3707 {
3708         struct device *dev = &adapter->vdev->dev;
3709         struct ibmvnic_error_buff *error_buff, *tmp;
3710         unsigned long flags;
3711         bool found = false;
3712         int i;
3713
3714         if (!crq->request_error_rsp.rc.code) {
3715                 dev_info(dev, "Request Error Rsp returned with rc=%x\n",
3716                          crq->request_error_rsp.rc.code);
3717                 return;
3718         }
3719
3720         spin_lock_irqsave(&adapter->error_list_lock, flags);
3721         list_for_each_entry_safe(error_buff, tmp, &adapter->errors, list)
3722                 if (error_buff->error_id == crq->request_error_rsp.error_id) {
3723                         found = true;
3724                         list_del(&error_buff->list);
3725                         break;
3726                 }
3727         spin_unlock_irqrestore(&adapter->error_list_lock, flags);
3728
3729         if (!found) {
3730                 dev_err(dev, "Couldn't find error id %x\n",
3731                         be32_to_cpu(crq->request_error_rsp.error_id));
3732                 return;
3733         }
3734
3735         dev_err(dev, "Detailed info for error id %x:",
3736                 be32_to_cpu(crq->request_error_rsp.error_id));
3737
3738         for (i = 0; i < error_buff->len; i++) {
3739                 pr_cont("%02x", (int)error_buff->buff[i]);
3740                 if (i % 8 == 7)
3741                         pr_cont(" ");
3742         }
3743         pr_cont("\n");
3744
3745         dma_unmap_single(dev, error_buff->dma, error_buff->len,
3746                          DMA_FROM_DEVICE);
3747         kfree(error_buff->buff);
3748         kfree(error_buff);
3749 }
3750
3751 static void request_error_information(struct ibmvnic_adapter *adapter,
3752                                       union ibmvnic_crq *err_crq)
3753 {
3754         struct device *dev = &adapter->vdev->dev;
3755         struct net_device *netdev = adapter->netdev;
3756         struct ibmvnic_error_buff *error_buff;
3757         unsigned long timeout = msecs_to_jiffies(30000);
3758         union ibmvnic_crq crq;
3759         unsigned long flags;
3760         int rc, detail_len;
3761
3762         error_buff = kmalloc(sizeof(*error_buff), GFP_ATOMIC);
3763         if (!error_buff)
3764                 return;
3765
3766         detail_len = be32_to_cpu(err_crq->error_indication.detail_error_sz);
3767         error_buff->buff = kmalloc(detail_len, GFP_ATOMIC);
3768         if (!error_buff->buff) {
3769                 kfree(error_buff);
3770                 return;
3771         }
3772
3773         error_buff->dma = dma_map_single(dev, error_buff->buff, detail_len,
3774                                          DMA_FROM_DEVICE);
3775         if (dma_mapping_error(dev, error_buff->dma)) {
3776                 netdev_err(netdev, "Couldn't map error buffer\n");
3777                 kfree(error_buff->buff);
3778                 kfree(error_buff);
3779                 return;
3780         }
3781
3782         error_buff->len = detail_len;
3783         error_buff->error_id = err_crq->error_indication.error_id;
3784
3785         spin_lock_irqsave(&adapter->error_list_lock, flags);
3786         list_add_tail(&error_buff->list, &adapter->errors);
3787         spin_unlock_irqrestore(&adapter->error_list_lock, flags);
3788
3789         memset(&crq, 0, sizeof(crq));
3790         crq.request_error_info.first = IBMVNIC_CRQ_CMD;
3791         crq.request_error_info.cmd = REQUEST_ERROR_INFO;
3792         crq.request_error_info.ioba = cpu_to_be32(error_buff->dma);
3793         crq.request_error_info.len = cpu_to_be32(detail_len);
3794         crq.request_error_info.error_id = err_crq->error_indication.error_id;
3795
3796         rc = ibmvnic_send_crq(adapter, &crq);
3797         if (rc) {
3798                 netdev_err(netdev, "failed to request error information\n");
3799                 goto err_info_fail;
3800         }
3801
3802         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
3803                 netdev_err(netdev, "timeout waiting for error information\n");
3804                 goto err_info_fail;
3805         }
3806
3807         return;
3808
3809 err_info_fail:
3810         spin_lock_irqsave(&adapter->error_list_lock, flags);
3811         list_del(&error_buff->list);
3812         spin_unlock_irqrestore(&adapter->error_list_lock, flags);
3813
3814         kfree(error_buff->buff);
3815         kfree(error_buff);
3816 }
3817
3818 static void handle_error_indication(union ibmvnic_crq *crq,
3819                                     struct ibmvnic_adapter *adapter)
3820 {
3821         struct device *dev = &adapter->vdev->dev;
3822
3823         dev_err(dev, "Firmware reports %serror id %x, cause %d\n",
3824                 crq->error_indication.flags
3825                         & IBMVNIC_FATAL_ERROR ? "FATAL " : "",
3826                 be32_to_cpu(crq->error_indication.error_id),
3827                 be16_to_cpu(crq->error_indication.error_cause));
3828
3829         if (be32_to_cpu(crq->error_indication.error_id))
3830                 request_error_information(adapter, crq);
3831
3832         if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
3833                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3834         else
3835                 ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
3836 }
3837
3838 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
3839                                  struct ibmvnic_adapter *adapter)
3840 {
3841         struct net_device *netdev = adapter->netdev;
3842         struct device *dev = &adapter->vdev->dev;
3843         long rc;
3844
3845         rc = crq->change_mac_addr_rsp.rc.code;
3846         if (rc) {
3847                 dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
3848                 goto out;
3849         }
3850         memcpy(netdev->dev_addr, &crq->change_mac_addr_rsp.mac_addr[0],
3851                ETH_ALEN);
3852 out:
3853         complete(&adapter->fw_done);
3854         return rc;
3855 }
3856
3857 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
3858                                    struct ibmvnic_adapter *adapter)
3859 {
3860         struct device *dev = &adapter->vdev->dev;
3861         u64 *req_value;
3862         char *name;
3863
3864         atomic_dec(&adapter->running_cap_crqs);
3865         switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
3866         case REQ_TX_QUEUES:
3867                 req_value = &adapter->req_tx_queues;
3868                 name = "tx";
3869                 break;
3870         case REQ_RX_QUEUES:
3871                 req_value = &adapter->req_rx_queues;
3872                 name = "rx";
3873                 break;
3874         case REQ_RX_ADD_QUEUES:
3875                 req_value = &adapter->req_rx_add_queues;
3876                 name = "rx_add";
3877                 break;
3878         case REQ_TX_ENTRIES_PER_SUBCRQ:
3879                 req_value = &adapter->req_tx_entries_per_subcrq;
3880                 name = "tx_entries_per_subcrq";
3881                 break;
3882         case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
3883                 req_value = &adapter->req_rx_add_entries_per_subcrq;
3884                 name = "rx_add_entries_per_subcrq";
3885                 break;
3886         case REQ_MTU:
3887                 req_value = &adapter->req_mtu;
3888                 name = "mtu";
3889                 break;
3890         case PROMISC_REQUESTED:
3891                 req_value = &adapter->promisc;
3892                 name = "promisc";
3893                 break;
3894         default:
3895                 dev_err(dev, "Got invalid cap request rsp %d\n",
3896                         crq->request_capability.capability);
3897                 return;
3898         }
3899
3900         switch (crq->request_capability_rsp.rc.code) {
3901         case SUCCESS:
3902                 break;
3903         case PARTIALSUCCESS:
3904                 dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
3905                          *req_value,
3906                          (long int)be64_to_cpu(crq->request_capability_rsp.
3907                                                number), name);
3908
3909                 if (be16_to_cpu(crq->request_capability_rsp.capability) ==
3910                     REQ_MTU) {
3911                         pr_err("mtu of %llu is not supported. Reverting.\n",
3912                                *req_value);
3913                         *req_value = adapter->fallback.mtu;
3914                 } else {
3915                         *req_value =
3916                                 be64_to_cpu(crq->request_capability_rsp.number);
3917                 }
3918
3919                 ibmvnic_send_req_caps(adapter, 1);
3920                 return;
3921         default:
3922                 dev_err(dev, "Error %d in request cap rsp\n",
3923                         crq->request_capability_rsp.rc.code);
3924                 return;
3925         }
3926
3927         /* Done receiving requested capabilities, query IP offload support */
3928         if (atomic_read(&adapter->running_cap_crqs) == 0) {
3929                 union ibmvnic_crq newcrq;
3930                 int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
3931                 struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
3932                     &adapter->ip_offload_buf;
3933
3934                 adapter->wait_capability = false;
3935                 adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
3936                                                          buf_sz,
3937                                                          DMA_FROM_DEVICE);
3938
3939                 if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
3940                         if (!firmware_has_feature(FW_FEATURE_CMO))
3941                                 dev_err(dev, "Couldn't map offload buffer\n");
3942                         return;
3943                 }
3944
3945                 memset(&newcrq, 0, sizeof(newcrq));
3946                 newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
3947                 newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
3948                 newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
3949                 newcrq.query_ip_offload.ioba =
3950                     cpu_to_be32(adapter->ip_offload_tok);
3951
3952                 ibmvnic_send_crq(adapter, &newcrq);
3953         }
3954 }
3955
3956 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
3957                             struct ibmvnic_adapter *adapter)
3958 {
3959         struct device *dev = &adapter->vdev->dev;
3960         struct net_device *netdev = adapter->netdev;
3961         struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
3962         struct ibmvnic_login_buffer *login = adapter->login_buf;
3963         int i;
3964
3965         dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
3966                          DMA_TO_DEVICE);
3967         dma_unmap_single(dev, adapter->login_rsp_buf_token,
3968                          adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
3969
3970         /* If the number of queues requested can't be allocated by the
3971          * server, the login response will return with code 1. We will need
3972          * to resend the login buffer with fewer queues requested.
3973          */
3974         if (login_rsp_crq->generic.rc.code) {
3975                 adapter->init_done_rc = login_rsp_crq->generic.rc.code;
3976                 complete(&adapter->init_done);
3977                 return 0;
3978         }
3979
3980         netdev->mtu = adapter->req_mtu - ETH_HLEN;
3981
3982         netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
3983         for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
3984                 netdev_dbg(adapter->netdev, "%016lx\n",
3985                            ((unsigned long int *)(adapter->login_rsp_buf))[i]);
3986         }
3987
3988         /* Sanity checks */
3989         if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
3990             (be32_to_cpu(login->num_rxcomp_subcrqs) *
3991              adapter->req_rx_add_queues !=
3992              be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
3993                 dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
3994                 ibmvnic_remove(adapter->vdev);
3995                 return -EIO;
3996         }
3997         release_login_buffer(adapter);
3998         complete(&adapter->init_done);
3999
4000         return 0;
4001 }
4002
4003 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
4004                                      struct ibmvnic_adapter *adapter)
4005 {
4006         struct device *dev = &adapter->vdev->dev;
4007         long rc;
4008
4009         rc = crq->request_unmap_rsp.rc.code;
4010         if (rc)
4011                 dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4012 }
4013
4014 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4015                                  struct ibmvnic_adapter *adapter)
4016 {
4017         struct net_device *netdev = adapter->netdev;
4018         struct device *dev = &adapter->vdev->dev;
4019         long rc;
4020
4021         rc = crq->query_map_rsp.rc.code;
4022         if (rc) {
4023                 dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4024                 return;
4025         }
4026         netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4027                    crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4028                    crq->query_map_rsp.free_pages);
4029 }
4030
4031 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4032                                  struct ibmvnic_adapter *adapter)
4033 {
4034         struct net_device *netdev = adapter->netdev;
4035         struct device *dev = &adapter->vdev->dev;
4036         long rc;
4037
4038         atomic_dec(&adapter->running_cap_crqs);
4039         netdev_dbg(netdev, "Outstanding queries: %d\n",
4040                    atomic_read(&adapter->running_cap_crqs));
4041         rc = crq->query_capability.rc.code;
4042         if (rc) {
4043                 dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4044                 goto out;
4045         }
4046
4047         switch (be16_to_cpu(crq->query_capability.capability)) {
4048         case MIN_TX_QUEUES:
4049                 adapter->min_tx_queues =
4050                     be64_to_cpu(crq->query_capability.number);
4051                 netdev_dbg(netdev, "min_tx_queues = %lld\n",
4052                            adapter->min_tx_queues);
4053                 break;
4054         case MIN_RX_QUEUES:
4055                 adapter->min_rx_queues =
4056                     be64_to_cpu(crq->query_capability.number);
4057                 netdev_dbg(netdev, "min_rx_queues = %lld\n",
4058                            adapter->min_rx_queues);
4059                 break;
4060         case MIN_RX_ADD_QUEUES:
4061                 adapter->min_rx_add_queues =
4062                     be64_to_cpu(crq->query_capability.number);
4063                 netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4064                            adapter->min_rx_add_queues);
4065                 break;
4066         case MAX_TX_QUEUES:
4067                 adapter->max_tx_queues =
4068                     be64_to_cpu(crq->query_capability.number);
4069                 netdev_dbg(netdev, "max_tx_queues = %lld\n",
4070                            adapter->max_tx_queues);
4071                 break;
4072         case MAX_RX_QUEUES:
4073                 adapter->max_rx_queues =
4074                     be64_to_cpu(crq->query_capability.number);
4075                 netdev_dbg(netdev, "max_rx_queues = %lld\n",
4076                            adapter->max_rx_queues);
4077                 break;
4078         case MAX_RX_ADD_QUEUES:
4079                 adapter->max_rx_add_queues =
4080                     be64_to_cpu(crq->query_capability.number);
4081                 netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4082                            adapter->max_rx_add_queues);
4083                 break;
4084         case MIN_TX_ENTRIES_PER_SUBCRQ:
4085                 adapter->min_tx_entries_per_subcrq =
4086                     be64_to_cpu(crq->query_capability.number);
4087                 netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4088                            adapter->min_tx_entries_per_subcrq);
4089                 break;
4090         case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4091                 adapter->min_rx_add_entries_per_subcrq =
4092                     be64_to_cpu(crq->query_capability.number);
4093                 netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4094                            adapter->min_rx_add_entries_per_subcrq);
4095                 break;
4096         case MAX_TX_ENTRIES_PER_SUBCRQ:
4097                 adapter->max_tx_entries_per_subcrq =
4098                     be64_to_cpu(crq->query_capability.number);
4099                 netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4100                            adapter->max_tx_entries_per_subcrq);
4101                 break;
4102         case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4103                 adapter->max_rx_add_entries_per_subcrq =
4104                     be64_to_cpu(crq->query_capability.number);
4105                 netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4106                            adapter->max_rx_add_entries_per_subcrq);
4107                 break;
4108         case TCP_IP_OFFLOAD:
4109                 adapter->tcp_ip_offload =
4110                     be64_to_cpu(crq->query_capability.number);
4111                 netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4112                            adapter->tcp_ip_offload);
4113                 break;
4114         case PROMISC_SUPPORTED:
4115                 adapter->promisc_supported =
4116                     be64_to_cpu(crq->query_capability.number);
4117                 netdev_dbg(netdev, "promisc_supported = %lld\n",
4118                            adapter->promisc_supported);
4119                 break;
4120         case MIN_MTU:
4121                 adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4122                 netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4123                 netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4124                 break;
4125         case MAX_MTU:
4126                 adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4127                 netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4128                 netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4129                 break;
4130         case MAX_MULTICAST_FILTERS:
4131                 adapter->max_multicast_filters =
4132                     be64_to_cpu(crq->query_capability.number);
4133                 netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4134                            adapter->max_multicast_filters);
4135                 break;
4136         case VLAN_HEADER_INSERTION:
4137                 adapter->vlan_header_insertion =
4138                     be64_to_cpu(crq->query_capability.number);
4139                 if (adapter->vlan_header_insertion)
4140                         netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4141                 netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4142                            adapter->vlan_header_insertion);
4143                 break;
4144         case RX_VLAN_HEADER_INSERTION:
4145                 adapter->rx_vlan_header_insertion =
4146                     be64_to_cpu(crq->query_capability.number);
4147                 netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4148                            adapter->rx_vlan_header_insertion);
4149                 break;
4150         case MAX_TX_SG_ENTRIES:
4151                 adapter->max_tx_sg_entries =
4152                     be64_to_cpu(crq->query_capability.number);
4153                 netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4154                            adapter->max_tx_sg_entries);
4155                 break;
4156         case RX_SG_SUPPORTED:
4157                 adapter->rx_sg_supported =
4158                     be64_to_cpu(crq->query_capability.number);
4159                 netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4160                            adapter->rx_sg_supported);
4161                 break;
4162         case OPT_TX_COMP_SUB_QUEUES:
4163                 adapter->opt_tx_comp_sub_queues =
4164                     be64_to_cpu(crq->query_capability.number);
4165                 netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4166                            adapter->opt_tx_comp_sub_queues);
4167                 break;
4168         case OPT_RX_COMP_QUEUES:
4169                 adapter->opt_rx_comp_queues =
4170                     be64_to_cpu(crq->query_capability.number);
4171                 netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4172                            adapter->opt_rx_comp_queues);
4173                 break;
4174         case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4175                 adapter->opt_rx_bufadd_q_per_rx_comp_q =
4176                     be64_to_cpu(crq->query_capability.number);
4177                 netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4178                            adapter->opt_rx_bufadd_q_per_rx_comp_q);
4179                 break;
4180         case OPT_TX_ENTRIES_PER_SUBCRQ:
4181                 adapter->opt_tx_entries_per_subcrq =
4182                     be64_to_cpu(crq->query_capability.number);
4183                 netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4184                            adapter->opt_tx_entries_per_subcrq);
4185                 break;
4186         case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4187                 adapter->opt_rxba_entries_per_subcrq =
4188                     be64_to_cpu(crq->query_capability.number);
4189                 netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4190                            adapter->opt_rxba_entries_per_subcrq);
4191                 break;
4192         case TX_RX_DESC_REQ:
4193                 adapter->tx_rx_desc_req = crq->query_capability.number;
4194                 netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4195                            adapter->tx_rx_desc_req);
4196                 break;
4197
4198         default:
4199                 netdev_err(netdev, "Got invalid cap rsp %d\n",
4200                            crq->query_capability.capability);
4201         }
4202
4203 out:
4204         if (atomic_read(&adapter->running_cap_crqs) == 0) {
4205                 adapter->wait_capability = false;
4206                 ibmvnic_send_req_caps(adapter, 0);
4207         }
4208 }
4209
4210 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4211                                struct ibmvnic_adapter *adapter)
4212 {
4213         struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4214         struct net_device *netdev = adapter->netdev;
4215         struct device *dev = &adapter->vdev->dev;
4216         u64 *u64_crq = (u64 *)crq;
4217         long rc;
4218
4219         netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4220                    (unsigned long int)cpu_to_be64(u64_crq[0]),
4221                    (unsigned long int)cpu_to_be64(u64_crq[1]));
4222         switch (gen_crq->first) {
4223         case IBMVNIC_CRQ_INIT_RSP:
4224                 switch (gen_crq->cmd) {
4225                 case IBMVNIC_CRQ_INIT:
4226                         dev_info(dev, "Partner initialized\n");
4227                         adapter->from_passive_init = true;
4228                         adapter->failover_pending = false;
4229                         complete(&adapter->init_done);
4230                         ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4231                         break;
4232                 case IBMVNIC_CRQ_INIT_COMPLETE:
4233                         dev_info(dev, "Partner initialization complete\n");
4234                         adapter->crq.active = true;
4235                         send_version_xchg(adapter);
4236                         break;
4237                 default:
4238                         dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4239                 }
4240                 return;
4241         case IBMVNIC_CRQ_XPORT_EVENT:
4242                 netif_carrier_off(netdev);
4243                 adapter->crq.active = false;
4244                 if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4245                         dev_info(dev, "Migrated, re-enabling adapter\n");
4246                         ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4247                 } else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4248                         dev_info(dev, "Backing device failover detected\n");
4249                         adapter->failover_pending = true;
4250                 } else {
4251                         /* The adapter lost the connection */
4252                         dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4253                                 gen_crq->cmd);
4254                         ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4255                 }
4256                 return;
4257         case IBMVNIC_CRQ_CMD_RSP:
4258                 break;
4259         default:
4260                 dev_err(dev, "Got an invalid msg type 0x%02x\n",
4261                         gen_crq->first);
4262                 return;
4263         }
4264
4265         switch (gen_crq->cmd) {
4266         case VERSION_EXCHANGE_RSP:
4267                 rc = crq->version_exchange_rsp.rc.code;
4268                 if (rc) {
4269                         dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4270                         break;
4271                 }
4272                 dev_info(dev, "Partner protocol version is %d\n",
4273                          crq->version_exchange_rsp.version);
4274                 if (be16_to_cpu(crq->version_exchange_rsp.version) <
4275                     ibmvnic_version)
4276                         ibmvnic_version =
4277                             be16_to_cpu(crq->version_exchange_rsp.version);
4278                 send_cap_queries(adapter);
4279                 break;
4280         case QUERY_CAPABILITY_RSP:
4281                 handle_query_cap_rsp(crq, adapter);
4282                 break;
4283         case QUERY_MAP_RSP:
4284                 handle_query_map_rsp(crq, adapter);
4285                 break;
4286         case REQUEST_MAP_RSP:
4287                 adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4288                 complete(&adapter->fw_done);
4289                 break;
4290         case REQUEST_UNMAP_RSP:
4291                 handle_request_unmap_rsp(crq, adapter);
4292                 break;
4293         case REQUEST_CAPABILITY_RSP:
4294                 handle_request_cap_rsp(crq, adapter);
4295                 break;
4296         case LOGIN_RSP:
4297                 netdev_dbg(netdev, "Got Login Response\n");
4298                 handle_login_rsp(crq, adapter);
4299                 break;
4300         case LOGICAL_LINK_STATE_RSP:
4301                 netdev_dbg(netdev,
4302                            "Got Logical Link State Response, state: %d rc: %d\n",
4303                            crq->logical_link_state_rsp.link_state,
4304                            crq->logical_link_state_rsp.rc.code);
4305                 adapter->logical_link_state =
4306                     crq->logical_link_state_rsp.link_state;
4307                 adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4308                 complete(&adapter->init_done);
4309                 break;
4310         case LINK_STATE_INDICATION:
4311                 netdev_dbg(netdev, "Got Logical Link State Indication\n");
4312                 adapter->phys_link_state =
4313                     crq->link_state_indication.phys_link_state;
4314                 adapter->logical_link_state =
4315                     crq->link_state_indication.logical_link_state;
4316                 break;
4317         case CHANGE_MAC_ADDR_RSP:
4318                 netdev_dbg(netdev, "Got MAC address change Response\n");
4319                 adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4320                 break;
4321         case ERROR_INDICATION:
4322                 netdev_dbg(netdev, "Got Error Indication\n");
4323                 handle_error_indication(crq, adapter);
4324                 break;
4325         case REQUEST_ERROR_RSP:
4326                 netdev_dbg(netdev, "Got Error Detail Response\n");
4327                 handle_error_info_rsp(crq, adapter);
4328                 break;
4329         case REQUEST_STATISTICS_RSP:
4330                 netdev_dbg(netdev, "Got Statistics Response\n");
4331                 complete(&adapter->stats_done);
4332                 break;
4333         case QUERY_IP_OFFLOAD_RSP:
4334                 netdev_dbg(netdev, "Got Query IP offload Response\n");
4335                 handle_query_ip_offload_rsp(adapter);
4336                 break;
4337         case MULTICAST_CTRL_RSP:
4338                 netdev_dbg(netdev, "Got multicast control Response\n");
4339                 break;
4340         case CONTROL_IP_OFFLOAD_RSP:
4341                 netdev_dbg(netdev, "Got Control IP offload Response\n");
4342                 dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4343                                  sizeof(adapter->ip_offload_ctrl),
4344                                  DMA_TO_DEVICE);
4345                 complete(&adapter->init_done);
4346                 break;
4347         case COLLECT_FW_TRACE_RSP:
4348                 netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4349                 complete(&adapter->fw_done);
4350                 break;
4351         case GET_VPD_SIZE_RSP:
4352                 handle_vpd_size_rsp(crq, adapter);
4353                 break;
4354         case GET_VPD_RSP:
4355                 handle_vpd_rsp(crq, adapter);
4356                 break;
4357         default:
4358                 netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4359                            gen_crq->cmd);
4360         }
4361 }
4362
4363 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4364 {
4365         struct ibmvnic_adapter *adapter = instance;
4366
4367         tasklet_schedule(&adapter->tasklet);
4368         return IRQ_HANDLED;
4369 }
4370
4371 static void ibmvnic_tasklet(void *data)
4372 {
4373         struct ibmvnic_adapter *adapter = data;
4374         struct ibmvnic_crq_queue *queue = &adapter->crq;
4375         union ibmvnic_crq *crq;
4376         unsigned long flags;
4377         bool done = false;
4378
4379         spin_lock_irqsave(&queue->lock, flags);
4380         while (!done) {
4381                 /* Pull all the valid messages off the CRQ */
4382                 while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4383                         ibmvnic_handle_crq(crq, adapter);
4384                         crq->generic.first = 0;
4385                 }
4386
4387                 /* remain in tasklet until all
4388                  * capabilities responses are received
4389                  */
4390                 if (!adapter->wait_capability)
4391                         done = true;
4392         }
4393         /* if capabilities CRQ's were sent in this tasklet, the following
4394          * tasklet must wait until all responses are received
4395          */
4396         if (atomic_read(&adapter->running_cap_crqs) != 0)
4397                 adapter->wait_capability = true;
4398         spin_unlock_irqrestore(&queue->lock, flags);
4399 }
4400
4401 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4402 {
4403         struct vio_dev *vdev = adapter->vdev;
4404         int rc;
4405
4406         do {
4407                 rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4408         } while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4409
4410         if (rc)
4411                 dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4412
4413         return rc;
4414 }
4415
4416 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4417 {
4418         struct ibmvnic_crq_queue *crq = &adapter->crq;
4419         struct device *dev = &adapter->vdev->dev;
4420         struct vio_dev *vdev = adapter->vdev;
4421         int rc;
4422
4423         /* Close the CRQ */
4424         do {
4425                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4426         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4427
4428         /* Clean out the queue */
4429         memset(crq->msgs, 0, PAGE_SIZE);
4430         crq->cur = 0;
4431         crq->active = false;
4432
4433         /* And re-open it again */
4434         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4435                                 crq->msg_token, PAGE_SIZE);
4436
4437         if (rc == H_CLOSED)
4438                 /* Adapter is good, but other end is not ready */
4439                 dev_warn(dev, "Partner adapter not ready\n");
4440         else if (rc != 0)
4441                 dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4442
4443         return rc;
4444 }
4445
4446 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4447 {
4448         struct ibmvnic_crq_queue *crq = &adapter->crq;
4449         struct vio_dev *vdev = adapter->vdev;
4450         long rc;
4451
4452         if (!crq->msgs)
4453                 return;
4454
4455         netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4456         free_irq(vdev->irq, adapter);
4457         tasklet_kill(&adapter->tasklet);
4458         do {
4459                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4460         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4461
4462         dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4463                          DMA_BIDIRECTIONAL);
4464         free_page((unsigned long)crq->msgs);
4465         crq->msgs = NULL;
4466         crq->active = false;
4467 }
4468
4469 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4470 {
4471         struct ibmvnic_crq_queue *crq = &adapter->crq;
4472         struct device *dev = &adapter->vdev->dev;
4473         struct vio_dev *vdev = adapter->vdev;
4474         int rc, retrc = -ENOMEM;
4475
4476         if (crq->msgs)
4477                 return 0;
4478
4479         crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4480         /* Should we allocate more than one page? */
4481
4482         if (!crq->msgs)
4483                 return -ENOMEM;
4484
4485         crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4486         crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4487                                         DMA_BIDIRECTIONAL);
4488         if (dma_mapping_error(dev, crq->msg_token))
4489                 goto map_failed;
4490
4491         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4492                                 crq->msg_token, PAGE_SIZE);
4493
4494         if (rc == H_RESOURCE)
4495                 /* maybe kexecing and resource is busy. try a reset */
4496                 rc = ibmvnic_reset_crq(adapter);
4497         retrc = rc;
4498
4499         if (rc == H_CLOSED) {
4500                 dev_warn(dev, "Partner adapter not ready\n");
4501         } else if (rc) {
4502                 dev_warn(dev, "Error %d opening adapter\n", rc);
4503                 goto reg_crq_failed;
4504         }
4505
4506         retrc = 0;
4507
4508         tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
4509                      (unsigned long)adapter);
4510
4511         netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
4512         rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, IBMVNIC_NAME,
4513                          adapter);
4514         if (rc) {
4515                 dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
4516                         vdev->irq, rc);
4517                 goto req_irq_failed;
4518         }
4519
4520         rc = vio_enable_interrupts(vdev);
4521         if (rc) {
4522                 dev_err(dev, "Error %d enabling interrupts\n", rc);
4523                 goto req_irq_failed;
4524         }
4525
4526         crq->cur = 0;
4527         spin_lock_init(&crq->lock);
4528
4529         return retrc;
4530
4531 req_irq_failed:
4532         tasklet_kill(&adapter->tasklet);
4533         do {
4534                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4535         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4536 reg_crq_failed:
4537         dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
4538 map_failed:
4539         free_page((unsigned long)crq->msgs);
4540         crq->msgs = NULL;
4541         return retrc;
4542 }
4543
4544 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
4545 {
4546         struct device *dev = &adapter->vdev->dev;
4547         unsigned long timeout = msecs_to_jiffies(30000);
4548         u64 old_num_rx_queues, old_num_tx_queues;
4549         int rc;
4550
4551         adapter->from_passive_init = false;
4552
4553         old_num_rx_queues = adapter->req_rx_queues;
4554         old_num_tx_queues = adapter->req_tx_queues;
4555
4556         init_completion(&adapter->init_done);
4557         adapter->init_done_rc = 0;
4558         ibmvnic_send_crq_init(adapter);
4559         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4560                 dev_err(dev, "Initialization sequence timed out\n");
4561                 return -1;
4562         }
4563
4564         if (adapter->init_done_rc) {
4565                 release_crq_queue(adapter);
4566                 return adapter->init_done_rc;
4567         }
4568
4569         if (adapter->from_passive_init) {
4570                 adapter->state = VNIC_OPEN;
4571                 adapter->from_passive_init = false;
4572                 return -1;
4573         }
4574
4575         if (adapter->resetting && !adapter->wait_for_reset &&
4576             adapter->reset_reason != VNIC_RESET_MOBILITY) {
4577                 if (adapter->req_rx_queues != old_num_rx_queues ||
4578                     adapter->req_tx_queues != old_num_tx_queues) {
4579                         release_sub_crqs(adapter, 0);
4580                         rc = init_sub_crqs(adapter);
4581                 } else {
4582                         rc = reset_sub_crq_queues(adapter);
4583                 }
4584         } else {
4585                 rc = init_sub_crqs(adapter);
4586         }
4587
4588         if (rc) {
4589                 dev_err(dev, "Initialization of sub crqs failed\n");
4590                 release_crq_queue(adapter);
4591                 return rc;
4592         }
4593
4594         rc = init_sub_crq_irqs(adapter);
4595         if (rc) {
4596                 dev_err(dev, "Failed to initialize sub crq irqs\n");
4597                 release_crq_queue(adapter);
4598         }
4599
4600         return rc;
4601 }
4602
4603 static struct device_attribute dev_attr_failover;
4604
4605 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
4606 {
4607         struct ibmvnic_adapter *adapter;
4608         struct net_device *netdev;
4609         unsigned char *mac_addr_p;
4610         int rc;
4611
4612         dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
4613                 dev->unit_address);
4614
4615         mac_addr_p = (unsigned char *)vio_get_attribute(dev,
4616                                                         VETH_MAC_ADDR, NULL);
4617         if (!mac_addr_p) {
4618                 dev_err(&dev->dev,
4619                         "(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
4620                         __FILE__, __LINE__);
4621                 return 0;
4622         }
4623
4624         netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
4625                                    IBMVNIC_MAX_QUEUES);
4626         if (!netdev)
4627                 return -ENOMEM;
4628
4629         adapter = netdev_priv(netdev);
4630         adapter->state = VNIC_PROBING;
4631         dev_set_drvdata(&dev->dev, netdev);
4632         adapter->vdev = dev;
4633         adapter->netdev = netdev;
4634
4635         ether_addr_copy(adapter->mac_addr, mac_addr_p);
4636         ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
4637         netdev->irq = dev->irq;
4638         netdev->netdev_ops = &ibmvnic_netdev_ops;
4639         netdev->ethtool_ops = &ibmvnic_ethtool_ops;
4640         SET_NETDEV_DEV(netdev, &dev->dev);
4641
4642         spin_lock_init(&adapter->stats_lock);
4643
4644         INIT_LIST_HEAD(&adapter->errors);
4645         spin_lock_init(&adapter->error_list_lock);
4646
4647         INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
4648         INIT_LIST_HEAD(&adapter->rwi_list);
4649         mutex_init(&adapter->reset_lock);
4650         mutex_init(&adapter->rwi_lock);
4651         adapter->resetting = false;
4652
4653         adapter->mac_change_pending = false;
4654
4655         do {
4656                 rc = init_crq_queue(adapter);
4657                 if (rc) {
4658                         dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
4659                                 rc);
4660                         goto ibmvnic_init_fail;
4661                 }
4662
4663                 rc = ibmvnic_init(adapter);
4664                 if (rc && rc != EAGAIN)
4665                         goto ibmvnic_init_fail;
4666         } while (rc == EAGAIN);
4667
4668         rc = init_stats_buffers(adapter);
4669         if (rc)
4670                 goto ibmvnic_init_fail;
4671
4672         rc = init_stats_token(adapter);
4673         if (rc)
4674                 goto ibmvnic_stats_fail;
4675
4676         netdev->mtu = adapter->req_mtu - ETH_HLEN;
4677         netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4678         netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4679
4680         rc = device_create_file(&dev->dev, &dev_attr_failover);
4681         if (rc)
4682                 goto ibmvnic_dev_file_err;
4683
4684         netif_carrier_off(netdev);
4685         rc = register_netdev(netdev);
4686         if (rc) {
4687                 dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
4688                 goto ibmvnic_register_fail;
4689         }
4690         dev_info(&dev->dev, "ibmvnic registered\n");
4691
4692         adapter->state = VNIC_PROBED;
4693
4694         adapter->wait_for_reset = false;
4695
4696         return 0;
4697
4698 ibmvnic_register_fail:
4699         device_remove_file(&dev->dev, &dev_attr_failover);
4700
4701 ibmvnic_dev_file_err:
4702         release_stats_token(adapter);
4703
4704 ibmvnic_stats_fail:
4705         release_stats_buffers(adapter);
4706
4707 ibmvnic_init_fail:
4708         release_sub_crqs(adapter, 1);
4709         release_crq_queue(adapter);
4710         free_netdev(netdev);
4711
4712         return rc;
4713 }
4714
4715 static int ibmvnic_remove(struct vio_dev *dev)
4716 {
4717         struct net_device *netdev = dev_get_drvdata(&dev->dev);
4718         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4719
4720         adapter->state = VNIC_REMOVING;
4721         unregister_netdev(netdev);
4722         mutex_lock(&adapter->reset_lock);
4723
4724         release_resources(adapter);
4725         release_sub_crqs(adapter, 1);
4726         release_crq_queue(adapter);
4727
4728         release_stats_token(adapter);
4729         release_stats_buffers(adapter);
4730
4731         adapter->state = VNIC_REMOVED;
4732
4733         mutex_unlock(&adapter->reset_lock);
4734         device_remove_file(&dev->dev, &dev_attr_failover);
4735         free_netdev(netdev);
4736         dev_set_drvdata(&dev->dev, NULL);
4737
4738         return 0;
4739 }
4740
4741 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
4742                               const char *buf, size_t count)
4743 {
4744         struct net_device *netdev = dev_get_drvdata(dev);
4745         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4746         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
4747         __be64 session_token;
4748         long rc;
4749
4750         if (!sysfs_streq(buf, "1"))
4751                 return -EINVAL;
4752
4753         rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
4754                          H_GET_SESSION_TOKEN, 0, 0, 0);
4755         if (rc) {
4756                 netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
4757                            rc);
4758                 return -EINVAL;
4759         }
4760
4761         session_token = (__be64)retbuf[0];
4762         netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
4763                    be64_to_cpu(session_token));
4764         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
4765                                 H_SESSION_ERR_DETECTED, session_token, 0, 0);
4766         if (rc) {
4767                 netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
4768                            rc);
4769                 return -EINVAL;
4770         }
4771
4772         return count;
4773 }
4774
4775 static DEVICE_ATTR_WO(failover);
4776
4777 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
4778 {
4779         struct net_device *netdev = dev_get_drvdata(&vdev->dev);
4780         struct ibmvnic_adapter *adapter;
4781         struct iommu_table *tbl;
4782         unsigned long ret = 0;
4783         int i;
4784
4785         tbl = get_iommu_table_base(&vdev->dev);
4786
4787         /* netdev inits at probe time along with the structures we need below*/
4788         if (!netdev)
4789                 return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
4790
4791         adapter = netdev_priv(netdev);
4792
4793         ret += PAGE_SIZE; /* the crq message queue */
4794         ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
4795
4796         for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
4797                 ret += 4 * PAGE_SIZE; /* the scrq message queue */
4798
4799         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4800              i++)
4801                 ret += adapter->rx_pool[i].size *
4802                     IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
4803
4804         return ret;
4805 }
4806
4807 static int ibmvnic_resume(struct device *dev)
4808 {
4809         struct net_device *netdev = dev_get_drvdata(dev);
4810         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4811
4812         if (adapter->state != VNIC_OPEN)
4813                 return 0;
4814
4815         tasklet_schedule(&adapter->tasklet);
4816
4817         return 0;
4818 }
4819
4820 static const struct vio_device_id ibmvnic_device_table[] = {
4821         {"network", "IBM,vnic"},
4822         {"", "" }
4823 };
4824 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
4825
4826 static const struct dev_pm_ops ibmvnic_pm_ops = {
4827         .resume = ibmvnic_resume
4828 };
4829
4830 static struct vio_driver ibmvnic_driver = {
4831         .id_table       = ibmvnic_device_table,
4832         .probe          = ibmvnic_probe,
4833         .remove         = ibmvnic_remove,
4834         .get_desired_dma = ibmvnic_get_desired_dma,
4835         .name           = ibmvnic_driver_name,
4836         .pm             = &ibmvnic_pm_ops,
4837 };
4838
4839 /* module functions */
4840 static int __init ibmvnic_module_init(void)
4841 {
4842         pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
4843                 IBMVNIC_DRIVER_VERSION);
4844
4845         return vio_register_driver(&ibmvnic_driver);
4846 }
4847
4848 static void __exit ibmvnic_module_exit(void)
4849 {
4850         vio_unregister_driver(&ibmvnic_driver);
4851 }
4852
4853 module_init(ibmvnic_module_init);
4854 module_exit(ibmvnic_module_exit);