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