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