X-Git-Url: http://git.droids-corp.org/?a=blobdiff_plain;f=drivers%2Fnet%2Fmlx4%2Fmlx4.c;h=21762ccfbe4fd7bcbe485f9c1d58d79c32e05372;hb=63c2f23c852a;hp=05923e953e3fc1eba49d457979385602c1e19843;hpb=8ca27e24e27267888aa12926c4a4bd73313fbadd;p=dpdk.git diff --git a/drivers/net/mlx4/mlx4.c b/drivers/net/mlx4/mlx4.c index 05923e953e..21762ccfbe 100644 --- a/drivers/net/mlx4/mlx4.c +++ b/drivers/net/mlx4/mlx4.c @@ -31,11 +31,6 @@ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ -/* - * Known limitations: - * - RSS hash key and options cannot be modified. - */ - /* System headers. */ #include #include @@ -64,15 +59,13 @@ #include #include #include -#include -#include -#include #include #include #include #include #include #include +#include /* Generated configuration header. */ #include "mlx4_autoconf.h" @@ -80,32 +73,7 @@ /* PMD headers. */ #include "mlx4.h" #include "mlx4_flow.h" - -/* Convenience macros for accessing mbuf fields. */ -#define NEXT(m) ((m)->next) -#define DATA_LEN(m) ((m)->data_len) -#define PKT_LEN(m) ((m)->pkt_len) -#define DATA_OFF(m) ((m)->data_off) -#define SET_DATA_OFF(m, o) ((m)->data_off = (o)) -#define NB_SEGS(m) ((m)->nb_segs) -#define PORT(m) ((m)->port) - -/* Work Request ID data type (64 bit). */ -typedef union { - struct { - uint32_t id; - uint16_t offset; - } data; - uint64_t raw; -} wr_id_t; - -#define WR_ID(o) (((wr_id_t *)&(o))->data) - -/* Transpose flags. Useful to convert IBV to DPDK flags. */ -#define TRANSPOSE(val, from, to) \ - (((from) >= (to)) ? \ - (((val) & (from)) / ((from) / (to))) : \ - (((val) & (from)) * ((to) / (from)))) +#include "mlx4_utils.h" /** Configuration structure for device arguments. */ struct mlx4_conf { @@ -133,29 +101,6 @@ priv_rx_intr_vec_enable(struct priv *priv); static void priv_rx_intr_vec_disable(struct priv *priv); -/** - * Lock private structure to protect it from concurrent access in the - * control path. - * - * @param priv - * Pointer to private structure. - */ -void priv_lock(struct priv *priv) -{ - rte_spinlock_lock(&priv->lock); -} - -/** - * Unlock private structure. - * - * @param priv - * Pointer to private structure. - */ -void priv_unlock(struct priv *priv) -{ - rte_spinlock_unlock(&priv->lock); -} - /* Allocate a buffer on the stack and fill it with a printf format string. */ #define MKSTR(name, ...) \ char name[snprintf(NULL, 0, __VA_ARGS__) + 1]; \ @@ -171,7 +116,7 @@ void priv_unlock(struct priv *priv) * Interface name output buffer. * * @return - * 0 on success, -1 on failure and errno is set. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int priv_get_ifname(const struct priv *priv, char (*ifname)[IF_NAMESIZE]) @@ -186,8 +131,10 @@ priv_get_ifname(const struct priv *priv, char (*ifname)[IF_NAMESIZE]) MKSTR(path, "%s/device/net", priv->ctx->device->ibdev_path); dir = opendir(path); - if (dir == NULL) - return -1; + if (dir == NULL) { + rte_errno = errno; + return -rte_errno; + } } while ((dent = readdir(dir)) != NULL) { char *name = dent->d_name; @@ -237,8 +184,10 @@ try_dev_id: snprintf(match, sizeof(match), "%s", name); } closedir(dir); - if (match[0] == '\0') - return -1; + if (match[0] == '\0') { + rte_errno = ENODEV; + return -rte_errno; + } strncpy(*ifname, match, sizeof(*ifname)); return 0; } @@ -256,7 +205,8 @@ try_dev_id: * Buffer size. * * @return - * 0 on success, -1 on failure and errno is set. + * Number of bytes read on success, negative errno value otherwise and + * rte_errno is set. */ static int priv_sysfs_read(const struct priv *priv, const char *entry, @@ -265,25 +215,27 @@ priv_sysfs_read(const struct priv *priv, const char *entry, char ifname[IF_NAMESIZE]; FILE *file; int ret; - int err; - if (priv_get_ifname(priv, &ifname)) - return -1; + ret = priv_get_ifname(priv, &ifname); + if (ret) + return ret; MKSTR(path, "%s/device/net/%s/%s", priv->ctx->device->ibdev_path, ifname, entry); file = fopen(path, "rb"); - if (file == NULL) - return -1; + if (file == NULL) { + rte_errno = errno; + return -rte_errno; + } ret = fread(buf, 1, size, file); - err = errno; - if (((size_t)ret < size) && (ferror(file))) - ret = -1; - else + if ((size_t)ret < size && ferror(file)) { + rte_errno = EIO; + ret = -rte_errno; + } else { ret = size; + } fclose(file); - errno = err; return ret; } @@ -300,7 +252,8 @@ priv_sysfs_read(const struct priv *priv, const char *entry, * Buffer size. * * @return - * 0 on success, -1 on failure and errno is set. + * Number of bytes written on success, negative errno value otherwise and + * rte_errno is set. */ static int priv_sysfs_write(const struct priv *priv, const char *entry, @@ -309,25 +262,27 @@ priv_sysfs_write(const struct priv *priv, const char *entry, char ifname[IF_NAMESIZE]; FILE *file; int ret; - int err; - if (priv_get_ifname(priv, &ifname)) - return -1; + ret = priv_get_ifname(priv, &ifname); + if (ret) + return ret; MKSTR(path, "%s/device/net/%s/%s", priv->ctx->device->ibdev_path, ifname, entry); file = fopen(path, "wb"); - if (file == NULL) - return -1; + if (file == NULL) { + rte_errno = errno; + return -rte_errno; + } ret = fwrite(buf, 1, size, file); - err = errno; - if (((size_t)ret < size) || (ferror(file))) - ret = -1; - else + if ((size_t)ret < size || ferror(file)) { + rte_errno = EIO; + ret = -rte_errno; + } else { ret = size; + } fclose(file); - errno = err; return ret; } @@ -342,7 +297,7 @@ priv_sysfs_write(const struct priv *priv, const char *entry, * Value output buffer. * * @return - * 0 on success, -1 on failure and errno is set. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int priv_get_sysfs_ulong(struct priv *priv, const char *name, unsigned long *value) @@ -352,18 +307,19 @@ priv_get_sysfs_ulong(struct priv *priv, const char *name, unsigned long *value) char value_str[32]; ret = priv_sysfs_read(priv, name, value_str, (sizeof(value_str) - 1)); - if (ret == -1) { + if (ret < 0) { DEBUG("cannot read %s value from sysfs: %s", - name, strerror(errno)); - return -1; + name, strerror(rte_errno)); + return ret; } value_str[ret] = '\0'; errno = 0; value_ret = strtoul(value_str, NULL, 0); if (errno) { + rte_errno = errno; DEBUG("invalid %s value `%s': %s", name, value_str, - strerror(errno)); - return -1; + strerror(rte_errno)); + return -rte_errno; } *value = value_ret; return 0; @@ -380,7 +336,7 @@ priv_get_sysfs_ulong(struct priv *priv, const char *name, unsigned long *value) * Value to set. * * @return - * 0 on success, -1 on failure and errno is set. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int priv_set_sysfs_ulong(struct priv *priv, const char *name, unsigned long value) @@ -389,10 +345,10 @@ priv_set_sysfs_ulong(struct priv *priv, const char *name, unsigned long value) MKSTR(value_str, "%lu", value); ret = priv_sysfs_write(priv, name, value_str, (sizeof(value_str) - 1)); - if (ret == -1) { + if (ret < 0) { DEBUG("cannot write %s `%s' (%lu) to sysfs: %s", - name, value_str, value, strerror(errno)); - return -1; + name, value_str, value, strerror(rte_errno)); + return ret; } return 0; } @@ -408,18 +364,23 @@ priv_set_sysfs_ulong(struct priv *priv, const char *name, unsigned long value) * Interface request structure output buffer. * * @return - * 0 on success, -1 on failure and errno is set. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int priv_ifreq(const struct priv *priv, int req, struct ifreq *ifr) { int sock = socket(PF_INET, SOCK_DGRAM, IPPROTO_IP); - int ret = -1; + int ret; - if (sock == -1) - return ret; - if (priv_get_ifname(priv, &ifr->ifr_name) == 0) - ret = ioctl(sock, req, ifr); + if (sock == -1) { + rte_errno = errno; + return -rte_errno; + } + ret = priv_get_ifname(priv, &ifr->ifr_name); + if (!ret && ioctl(sock, req, ifr) == -1) { + rte_errno = errno; + ret = -rte_errno; + } close(sock); return ret; } @@ -433,42 +394,49 @@ priv_ifreq(const struct priv *priv, int req, struct ifreq *ifr) * MTU value output buffer. * * @return - * 0 on success, -1 on failure and errno is set. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int priv_get_mtu(struct priv *priv, uint16_t *mtu) { - unsigned long ulong_mtu; + unsigned long ulong_mtu = 0; + int ret = priv_get_sysfs_ulong(priv, "mtu", &ulong_mtu); - if (priv_get_sysfs_ulong(priv, "mtu", &ulong_mtu) == -1) - return -1; + if (ret) + return ret; *mtu = ulong_mtu; return 0; } /** - * Set device MTU. + * DPDK callback to change the MTU. * * @param priv - * Pointer to private structure. + * Pointer to Ethernet device structure. * @param mtu * MTU value to set. * * @return - * 0 on success, -1 on failure and errno is set. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int -priv_set_mtu(struct priv *priv, uint16_t mtu) +mlx4_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu) { + struct priv *priv = dev->data->dev_private; uint16_t new_mtu; + int ret = priv_set_sysfs_ulong(priv, "mtu", mtu); - if (priv_set_sysfs_ulong(priv, "mtu", mtu) || - priv_get_mtu(priv, &new_mtu)) - return -1; - if (new_mtu == mtu) + if (ret) + return ret; + ret = priv_get_mtu(priv, &new_mtu); + if (ret) + return ret; + if (new_mtu == mtu) { + priv->mtu = mtu; return 0; - errno = EINVAL; - return -1; + } + rte_errno = EINVAL; + return -rte_errno; } /** @@ -482,15 +450,16 @@ priv_set_mtu(struct priv *priv, uint16_t mtu) * Bitmask for flags to modify. * * @return - * 0 on success, -1 on failure and errno is set. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int priv_set_flags(struct priv *priv, unsigned int keep, unsigned int flags) { - unsigned long tmp; + unsigned long tmp = 0; + int ret = priv_get_sysfs_ulong(priv, "flags", &tmp); - if (priv_get_sysfs_ulong(priv, "flags", &tmp) == -1) - return -1; + if (ret) + return ret; tmp &= keep; tmp |= (flags & (~keep)); return priv_set_sysfs_ulong(priv, "flags", tmp); @@ -507,10 +476,8 @@ txq_cleanup(struct txq *txq); static int rxq_setup(struct rte_eth_dev *dev, struct rxq *rxq, uint16_t desc, - unsigned int socket, int inactive, - const struct rte_eth_rxconf *conf, - struct rte_mempool *mp, int children_n, - struct rxq *rxq_parent); + unsigned int socket, const struct rte_eth_rxconf *conf, + struct rte_mempool *mp); static void rxq_cleanup(struct rxq *rxq); @@ -519,102 +486,22 @@ static void priv_mac_addr_del(struct priv *priv); /** - * Create RSS parent queue. - * - * The new parent is inserted in front of the list in the private structure. - * - * @param priv - * Pointer to private structure. - * @param queues - * Queues indices array, if NULL use all Rx queues. - * @param children_n - * The number of entries in queues[]. - * - * @return - * Pointer to a parent rxq structure, NULL on failure. - */ -struct rxq * -priv_parent_create(struct priv *priv, - uint16_t queues[], - uint16_t children_n) -{ - int ret; - uint16_t i; - struct rxq *parent; - - parent = rte_zmalloc("parent queue", - sizeof(*parent), - RTE_CACHE_LINE_SIZE); - if (!parent) { - ERROR("cannot allocate memory for RSS parent queue"); - return NULL; - } - ret = rxq_setup(priv->dev, parent, 0, 0, 0, - NULL, NULL, children_n, NULL); - if (ret) { - rte_free(parent); - return NULL; - } - parent->rss.queues_n = children_n; - if (queues) { - for (i = 0; i < children_n; ++i) - parent->rss.queues[i] = queues[i]; - } else { - /* the default RSS ring case */ - assert(priv->rxqs_n == children_n); - for (i = 0; i < priv->rxqs_n; ++i) - parent->rss.queues[i] = i; - } - LIST_INSERT_HEAD(&priv->parents, parent, next); - return parent; -} - -/** - * Clean up RX queue parent structure. - * - * @param parent - * RX queue parent structure. - */ -void -rxq_parent_cleanup(struct rxq *parent) -{ - LIST_REMOVE(parent, next); - rxq_cleanup(parent); - rte_free(parent); -} - -/** - * Clean up parent structures from the parent list. - * - * @param priv - * Pointer to private structure. - */ -static void -priv_parent_list_cleanup(struct priv *priv) -{ - while (!LIST_EMPTY(&priv->parents)) - rxq_parent_cleanup(LIST_FIRST(&priv->parents)); -} - -/** - * Ethernet device configuration. + * DPDK callback for Ethernet device configuration. * * Prepare the driver for a given number of TX and RX queues. - * Allocate parent RSS queue when several RX queues are requested. * * @param dev * Pointer to Ethernet device structure. * * @return - * 0 on success, errno value on failure. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int -dev_configure(struct rte_eth_dev *dev) +mlx4_dev_configure(struct rte_eth_dev *dev) { struct priv *priv = dev->data->dev_private; unsigned int rxqs_n = dev->data->nb_rx_queues; unsigned int txqs_n = dev->data->nb_tx_queues; - unsigned int tmp; priv->rxqs = (void *)dev->data->rx_queues; priv->txqs = (void *)dev->data->tx_queues; @@ -623,83 +510,12 @@ dev_configure(struct rte_eth_dev *dev) (void *)dev, priv->txqs_n, txqs_n); priv->txqs_n = txqs_n; } - if (rxqs_n == priv->rxqs_n) - return 0; - if (!rte_is_power_of_2(rxqs_n) && !priv->isolated) { - unsigned n_active; - - n_active = rte_align32pow2(rxqs_n + 1) >> 1; - WARN("%p: number of RX queues must be a power" - " of 2: %u queues among %u will be active", - (void *)dev, n_active, rxqs_n); - } - - INFO("%p: RX queues number update: %u -> %u", - (void *)dev, priv->rxqs_n, rxqs_n); - /* If RSS is enabled, disable it first. */ - if (priv->rss) { - unsigned int i; - - /* Only if there are no remaining child RX queues. */ - for (i = 0; (i != priv->rxqs_n); ++i) - if ((*priv->rxqs)[i] != NULL) - return EINVAL; - priv_mac_addr_del(priv); - priv_parent_list_cleanup(priv); - priv->rss = 0; - priv->rxqs_n = 0; - } - if (rxqs_n <= 1) { - /* Nothing else to do. */ + if (rxqs_n != priv->rxqs_n) { + INFO("%p: Rx queues number update: %u -> %u", + (void *)dev, priv->rxqs_n, rxqs_n); priv->rxqs_n = rxqs_n; - return 0; - } - /* Allocate a new RSS parent queue if supported by hardware. */ - if (!priv->hw_rss) { - ERROR("%p: only a single RX queue can be configured when" - " hardware doesn't support RSS", - (void *)dev); - return EINVAL; - } - /* Fail if hardware doesn't support that many RSS queues. */ - if (rxqs_n >= priv->max_rss_tbl_sz) { - ERROR("%p: only %u RX queues can be configured for RSS", - (void *)dev, priv->max_rss_tbl_sz); - return EINVAL; } - priv->rss = 1; - tmp = priv->rxqs_n; - priv->rxqs_n = rxqs_n; - if (priv->isolated) - return 0; - if (priv_parent_create(priv, NULL, priv->rxqs_n)) - return 0; - /* Failure, rollback. */ - priv->rss = 0; - priv->rxqs_n = tmp; - return ENOMEM; -} - -/** - * DPDK callback for Ethernet device configuration. - * - * @param dev - * Pointer to Ethernet device structure. - * - * @return - * 0 on success, negative errno value on failure. - */ -static int -mlx4_dev_configure(struct rte_eth_dev *dev) -{ - struct priv *priv = dev->data->dev_private; - int ret; - - priv_lock(priv); - ret = dev_configure(dev); - assert(ret >= 0); - priv_unlock(priv); - return -ret; + return 0; } static uint16_t mlx4_tx_burst(void *, struct rte_mbuf **, uint16_t); @@ -716,7 +532,7 @@ static uint16_t removed_rx_burst(void *, struct rte_mbuf **, uint16_t); * Number of elements to allocate. * * @return - * 0 on success, errno value on failure. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int txq_alloc_elts(struct txq *txq, unsigned int elts_n) @@ -724,26 +540,13 @@ txq_alloc_elts(struct txq *txq, unsigned int elts_n) unsigned int i; struct txq_elt (*elts)[elts_n] = rte_calloc_socket("TXQ", 1, sizeof(*elts), 0, txq->socket); - linear_t (*elts_linear)[elts_n] = - rte_calloc_socket("TXQ", 1, sizeof(*elts_linear), 0, - txq->socket); - struct ibv_mr *mr_linear = NULL; int ret = 0; - if ((elts == NULL) || (elts_linear == NULL)) { + if (elts == NULL) { ERROR("%p: can't allocate packets array", (void *)txq); ret = ENOMEM; goto error; } - mr_linear = - ibv_reg_mr(txq->priv->pd, elts_linear, sizeof(*elts_linear), - IBV_ACCESS_LOCAL_WRITE); - if (mr_linear == NULL) { - ERROR("%p: unable to configure MR, ibv_reg_mr() failed", - (void *)txq); - ret = EINVAL; - goto error; - } for (i = 0; (i != elts_n); ++i) { struct txq_elt *elt = &(*elts)[i]; @@ -755,26 +558,22 @@ txq_alloc_elts(struct txq *txq, unsigned int elts_n) txq->elts_head = 0; txq->elts_tail = 0; txq->elts_comp = 0; - /* Request send completion every MLX4_PMD_TX_PER_COMP_REQ packets or - * at least 4 times per ring. */ + /* + * Request send completion every MLX4_PMD_TX_PER_COMP_REQ packets or + * at least 4 times per ring. + */ txq->elts_comp_cd_init = ((MLX4_PMD_TX_PER_COMP_REQ < (elts_n / 4)) ? MLX4_PMD_TX_PER_COMP_REQ : (elts_n / 4)); txq->elts_comp_cd = txq->elts_comp_cd_init; - txq->elts_linear = elts_linear; - txq->mr_linear = mr_linear; assert(ret == 0); return 0; error: - if (mr_linear != NULL) - claim_zero(ibv_dereg_mr(mr_linear)); - - rte_free(elts_linear); rte_free(elts); - DEBUG("%p: failed, freed everything", (void *)txq); assert(ret > 0); - return ret; + rte_errno = ret; + return -rte_errno; } /** @@ -790,8 +589,6 @@ txq_free_elts(struct txq *txq) unsigned int elts_head = txq->elts_head; unsigned int elts_tail = txq->elts_tail; struct txq_elt (*elts)[elts_n] = txq->elts; - linear_t (*elts_linear)[elts_n] = txq->elts_linear; - struct ibv_mr *mr_linear = txq->mr_linear; DEBUG("%p: freeing WRs", (void *)txq); txq->elts_n = 0; @@ -801,12 +598,6 @@ txq_free_elts(struct txq *txq) txq->elts_comp_cd = 0; txq->elts_comp_cd_init = 0; txq->elts = NULL; - txq->elts_linear = NULL; - txq->mr_linear = NULL; - if (mr_linear != NULL) - claim_zero(ibv_dereg_mr(mr_linear)); - - rte_free(elts_linear); if (elts == NULL) return; while (elts_tail != elts_head) { @@ -824,7 +615,6 @@ txq_free_elts(struct txq *txq) rte_free(elts); } - /** * Clean up a TX queue. * @@ -836,49 +626,15 @@ txq_free_elts(struct txq *txq) static void txq_cleanup(struct txq *txq) { - struct ibv_exp_release_intf_params params; size_t i; DEBUG("cleaning up %p", (void *)txq); txq_free_elts(txq); - if (txq->if_qp != NULL) { - assert(txq->priv != NULL); - assert(txq->priv->ctx != NULL); - assert(txq->qp != NULL); - params = (struct ibv_exp_release_intf_params){ - .comp_mask = 0, - }; - claim_zero(ibv_exp_release_intf(txq->priv->ctx, - txq->if_qp, - ¶ms)); - } - if (txq->if_cq != NULL) { - assert(txq->priv != NULL); - assert(txq->priv->ctx != NULL); - assert(txq->cq != NULL); - params = (struct ibv_exp_release_intf_params){ - .comp_mask = 0, - }; - claim_zero(ibv_exp_release_intf(txq->priv->ctx, - txq->if_cq, - ¶ms)); - } if (txq->qp != NULL) claim_zero(ibv_destroy_qp(txq->qp)); if (txq->cq != NULL) claim_zero(ibv_destroy_cq(txq->cq)); - if (txq->rd != NULL) { - struct ibv_exp_destroy_res_domain_attr attr = { - .comp_mask = 0, - }; - - assert(txq->priv != NULL); - assert(txq->priv->ctx != NULL); - claim_zero(ibv_exp_destroy_res_domain(txq->priv->ctx, - txq->rd, - &attr)); - } - for (i = 0; (i != elemof(txq->mp2mr)); ++i) { + for (i = 0; (i != RTE_DIM(txq->mp2mr)); ++i) { if (txq->mp2mr[i].mp == NULL) break; assert(txq->mp2mr[i].mr != NULL); @@ -907,11 +663,12 @@ txq_complete(struct txq *txq) unsigned int elts_comp = txq->elts_comp; unsigned int elts_tail = txq->elts_tail; const unsigned int elts_n = txq->elts_n; + struct ibv_wc wcs[elts_comp]; int wcs_n; if (unlikely(elts_comp == 0)) return 0; - wcs_n = txq->if_cq->poll_cnt(txq->cq, elts_comp); + wcs_n = ibv_poll_cq(txq->cq, elts_comp, wcs); if (unlikely(wcs_n == 0)) return 0; if (unlikely(wcs_n < 0)) { @@ -948,7 +705,6 @@ static void mlx4_check_mempool_cb(struct rte_mempool *mp, (void)mp; (void)mem_idx; - /* It already failed, skip the next chunks. */ if (data->ret != 0) return; @@ -992,7 +748,6 @@ static int mlx4_check_mempool(struct rte_mempool *mp, uintptr_t *start, rte_mempool_mem_iter(mp, mlx4_check_mempool_cb, &data); *start = (uintptr_t)data.start; *end = (uintptr_t)data.end; - return data.ret; } @@ -1009,7 +764,7 @@ static struct ibv_mr *mlx4_mp2mr(struct ibv_pd *, struct rte_mempool *) * Pointer to memory pool. * * @return - * Memory region pointer, NULL in case of error. + * Memory region pointer, NULL in case of error and rte_errno is set. */ static struct ibv_mr * mlx4_mp2mr(struct ibv_pd *pd, struct rte_mempool *mp) @@ -1018,13 +773,14 @@ mlx4_mp2mr(struct ibv_pd *pd, struct rte_mempool *mp) uintptr_t start; uintptr_t end; unsigned int i; + struct ibv_mr *mr; if (mlx4_check_mempool(mp, &start, &end) != 0) { + rte_errno = EINVAL; ERROR("mempool %p: not virtually contiguous", (void *)mp); return NULL; } - DEBUG("mempool %p area start=%p end=%p size=%zu", (void *)mp, (void *)start, (void *)end, (size_t)(end - start)); @@ -1042,10 +798,13 @@ mlx4_mp2mr(struct ibv_pd *pd, struct rte_mempool *mp) DEBUG("mempool %p using start=%p end=%p size=%zu for MR", (void *)mp, (void *)start, (void *)end, (size_t)(end - start)); - return ibv_reg_mr(pd, - (void *)start, - end - start, - IBV_ACCESS_LOCAL_WRITE); + mr = ibv_reg_mr(pd, + (void *)start, + end - start, + IBV_ACCESS_LOCAL_WRITE); + if (!mr) + rte_errno = errno ? errno : EINVAL; + return mr; } /** @@ -1085,7 +844,7 @@ txq_mp2mr(struct txq *txq, struct rte_mempool *mp) unsigned int i; struct ibv_mr *mr; - for (i = 0; (i != elemof(txq->mp2mr)); ++i) { + for (i = 0; (i != RTE_DIM(txq->mp2mr)); ++i) { if (unlikely(txq->mp2mr[i].mp == NULL)) { /* Unknown MP, add a new MR for it. */ break; @@ -1105,7 +864,7 @@ txq_mp2mr(struct txq *txq, struct rte_mempool *mp) (void *)txq); return (uint32_t)-1; } - if (unlikely(i == elemof(txq->mp2mr))) { + if (unlikely(i == RTE_DIM(txq->mp2mr))) { /* Table is full, remove oldest entry. */ DEBUG("%p: MR <-> MP table full, dropping oldest entry.", (void *)txq); @@ -1148,8 +907,10 @@ txq_mp2mr_mbuf_check(struct rte_mempool *mp, void *arg, void *obj, struct txq_mp2mr_mbuf_check_data *data = arg; struct rte_mbuf *buf = obj; - /* Check whether mbuf structure fits element size and whether mempool - * pointer is valid. */ + /* + * Check whether mbuf structure fits element size and whether mempool + * pointer is valid. + */ if (sizeof(*buf) > mp->elt_size || buf->pool != mp) data->ret = -1; } @@ -1178,152 +939,6 @@ txq_mp2mr_iter(struct rte_mempool *mp, void *arg) txq_mp2mr(txq, mp); } -/** - * Copy scattered mbuf contents to a single linear buffer. - * - * @param[out] linear - * Linear output buffer. - * @param[in] buf - * Scattered input buffer. - * - * @return - * Number of bytes copied to the output buffer or 0 if not large enough. - */ -static unsigned int -linearize_mbuf(linear_t *linear, struct rte_mbuf *buf) -{ - unsigned int size = 0; - unsigned int offset; - - do { - unsigned int len = DATA_LEN(buf); - - offset = size; - size += len; - if (unlikely(size > sizeof(*linear))) - return 0; - memcpy(&(*linear)[offset], - rte_pktmbuf_mtod(buf, uint8_t *), - len); - buf = NEXT(buf); - } while (buf != NULL); - return size; -} - -/** - * Handle scattered buffers for mlx4_tx_burst(). - * - * @param txq - * TX queue structure. - * @param segs - * Number of segments in buf. - * @param elt - * TX queue element to fill. - * @param[in] buf - * Buffer to process. - * @param elts_head - * Index of the linear buffer to use if necessary (normally txq->elts_head). - * @param[out] sges - * Array filled with SGEs on success. - * - * @return - * A structure containing the processed packet size in bytes and the - * number of SGEs. Both fields are set to (unsigned int)-1 in case of - * failure. - */ -static struct tx_burst_sg_ret { - unsigned int length; - unsigned int num; -} -tx_burst_sg(struct txq *txq, unsigned int segs, struct txq_elt *elt, - struct rte_mbuf *buf, unsigned int elts_head, - struct ibv_sge (*sges)[MLX4_PMD_SGE_WR_N]) -{ - unsigned int sent_size = 0; - unsigned int j; - int linearize = 0; - - /* When there are too many segments, extra segments are - * linearized in the last SGE. */ - if (unlikely(segs > elemof(*sges))) { - segs = (elemof(*sges) - 1); - linearize = 1; - } - /* Update element. */ - elt->buf = buf; - /* Register segments as SGEs. */ - for (j = 0; (j != segs); ++j) { - struct ibv_sge *sge = &(*sges)[j]; - uint32_t lkey; - - /* Retrieve Memory Region key for this memory pool. */ - lkey = txq_mp2mr(txq, txq_mb2mp(buf)); - if (unlikely(lkey == (uint32_t)-1)) { - /* MR does not exist. */ - DEBUG("%p: unable to get MP <-> MR association", - (void *)txq); - /* Clean up TX element. */ - elt->buf = NULL; - goto stop; - } - /* Update SGE. */ - sge->addr = rte_pktmbuf_mtod(buf, uintptr_t); - if (txq->priv->vf) - rte_prefetch0((volatile void *) - (uintptr_t)sge->addr); - sge->length = DATA_LEN(buf); - sge->lkey = lkey; - sent_size += sge->length; - buf = NEXT(buf); - } - /* If buf is not NULL here and is not going to be linearized, - * nb_segs is not valid. */ - assert(j == segs); - assert((buf == NULL) || (linearize)); - /* Linearize extra segments. */ - if (linearize) { - struct ibv_sge *sge = &(*sges)[segs]; - linear_t *linear = &(*txq->elts_linear)[elts_head]; - unsigned int size = linearize_mbuf(linear, buf); - - assert(segs == (elemof(*sges) - 1)); - if (size == 0) { - /* Invalid packet. */ - DEBUG("%p: packet too large to be linearized.", - (void *)txq); - /* Clean up TX element. */ - elt->buf = NULL; - goto stop; - } - /* If MLX4_PMD_SGE_WR_N is 1, free mbuf immediately. */ - if (elemof(*sges) == 1) { - do { - struct rte_mbuf *next = NEXT(buf); - - rte_pktmbuf_free_seg(buf); - buf = next; - } while (buf != NULL); - elt->buf = NULL; - } - /* Update SGE. */ - sge->addr = (uintptr_t)&(*linear)[0]; - sge->length = size; - sge->lkey = txq->mr_linear->lkey; - sent_size += size; - /* Include last segment. */ - segs++; - } - return (struct tx_burst_sg_ret){ - .length = sent_size, - .num = segs, - }; -stop: - return (struct tx_burst_sg_ret){ - .length = -1, - .num = -1, - }; -} - /** * DPDK callback for TX. * @@ -1341,6 +956,9 @@ static uint16_t mlx4_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n) { struct txq *txq = (struct txq *)dpdk_txq; + struct ibv_send_wr *wr_head = NULL; + struct ibv_send_wr **wr_next = &wr_head; + struct ibv_send_wr *wr_bad = NULL; unsigned int elts_head = txq->elts_head; const unsigned int elts_n = txq->elts_n; unsigned int elts_comp_cd = txq->elts_comp_cd; @@ -1368,7 +986,8 @@ mlx4_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n) (((elts_head + 1) == elts_n) ? 0 : elts_head + 1); struct txq_elt *elt_next = &(*txq->elts)[elts_head_next]; struct txq_elt *elt = &(*txq->elts)[elts_head]; - unsigned int segs = NB_SEGS(buf); + struct ibv_send_wr *wr = &elt->wr; + unsigned int segs = buf->nb_segs; unsigned int sent_size = 0; uint32_t send_flags = 0; @@ -1382,7 +1001,7 @@ mlx4_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n) #endif /* Faster than rte_pktmbuf_free(). */ do { - struct rte_mbuf *next = NEXT(tmp); + struct rte_mbuf *next = tmp->next; rte_pktmbuf_free_seg(tmp); tmp = next; @@ -1392,27 +1011,17 @@ mlx4_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n) if (unlikely(--elts_comp_cd == 0)) { elts_comp_cd = txq->elts_comp_cd_init; ++elts_comp; - send_flags |= IBV_EXP_QP_BURST_SIGNALED; - } - /* Should we enable HW CKSUM offload */ - if (buf->ol_flags & - (PKT_TX_IP_CKSUM | PKT_TX_TCP_CKSUM | PKT_TX_UDP_CKSUM)) { - send_flags |= IBV_EXP_QP_BURST_IP_CSUM; - /* HW does not support checksum offloads at arbitrary - * offsets but automatically recognizes the packet - * type. For inner L3/L4 checksums, only VXLAN (UDP) - * tunnels are currently supported. */ - if (RTE_ETH_IS_TUNNEL_PKT(buf->packet_type)) - send_flags |= IBV_EXP_QP_BURST_TUNNEL; + send_flags |= IBV_SEND_SIGNALED; } if (likely(segs == 1)) { + struct ibv_sge *sge = &elt->sge; uintptr_t addr; uint32_t length; uint32_t lkey; /* Retrieve buffer information. */ addr = rte_pktmbuf_mtod(buf, uintptr_t); - length = DATA_LEN(buf); + length = buf->data_len; /* Retrieve Memory Region key for this memory pool. */ lkey = txq_mp2mr(txq, txq_mb2mp(buf)); if (unlikely(lkey == (uint32_t)-1)) { @@ -1429,45 +1038,26 @@ mlx4_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n) rte_prefetch0((volatile void *) (uintptr_t)addr); RTE_MBUF_PREFETCH_TO_FREE(elt_next->buf); - /* Put packet into send queue. */ - if (length <= txq->max_inline) - err = txq->if_qp->send_pending_inline - (txq->qp, - (void *)addr, - length, - send_flags); - else - err = txq->if_qp->send_pending - (txq->qp, - addr, - length, - lkey, - send_flags); - if (unlikely(err)) - goto stop; + sge->addr = addr; + sge->length = length; + sge->lkey = lkey; sent_size += length; } else { - struct ibv_sge sges[MLX4_PMD_SGE_WR_N]; - struct tx_burst_sg_ret ret; - - ret = tx_burst_sg(txq, segs, elt, buf, elts_head, - &sges); - if (ret.length == (unsigned int)-1) - goto stop; - RTE_MBUF_PREFETCH_TO_FREE(elt_next->buf); - /* Put SG list into send queue. */ - err = txq->if_qp->send_pending_sg_list - (txq->qp, - sges, - ret.num, - send_flags); - if (unlikely(err)) - goto stop; - sent_size += ret.length; + err = -1; + goto stop; } + if (sent_size <= txq->max_inline) + send_flags |= IBV_SEND_INLINE; elts_head = elts_head_next; /* Increment sent bytes counter. */ txq->stats.obytes += sent_size; + /* Set up WR. */ + wr->sg_list = &elt->sge; + wr->num_sge = segs; + wr->opcode = IBV_WR_SEND; + wr->send_flags = send_flags; + *wr_next = wr; + wr_next = &wr->next; } stop: /* Take a shortcut if nothing must be sent. */ @@ -1476,12 +1066,37 @@ stop: /* Increment sent packets counter. */ txq->stats.opackets += i; /* Ring QP doorbell. */ - err = txq->if_qp->send_flush(txq->qp); + *wr_next = NULL; + assert(wr_head); + err = ibv_post_send(txq->qp, wr_head, &wr_bad); if (unlikely(err)) { - /* A nonzero value is not supposed to be returned. - * Nothing can be done about it. */ - DEBUG("%p: send_flush() failed with error %d", - (void *)txq, err); + uint64_t obytes = 0; + uint64_t opackets = 0; + + /* Rewind bad WRs. */ + while (wr_bad != NULL) { + int j; + + /* Force completion request if one was lost. */ + if (wr_bad->send_flags & IBV_SEND_SIGNALED) { + elts_comp_cd = 1; + --elts_comp; + } + ++opackets; + for (j = 0; j < wr_bad->num_sge; ++j) + obytes += wr_bad->sg_list[j].length; + elts_head = (elts_head ? elts_head : elts_n) - 1; + wr_bad = wr_bad->next; + } + txq->stats.opackets -= opackets; + txq->stats.obytes -= obytes; + i -= opackets; + DEBUG("%p: ibv_post_send() failed, %" PRIu64 " packets" + " (%" PRIu64 " bytes) rejected: %s", + (void *)txq, + opackets, + obytes, + (err <= -1) ? "Internal error" : strerror(err)); } txq->elts_head = elts_head; txq->elts_comp += elts_comp; @@ -1504,7 +1119,7 @@ stop: * Thresholds parameters. * * @return - * 0 on success, errno value on failure. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int txq_setup(struct rte_eth_dev *dev, struct txq *txq, uint16_t desc, @@ -1516,54 +1131,34 @@ txq_setup(struct rte_eth_dev *dev, struct txq *txq, uint16_t desc, .socket = socket }; union { - struct ibv_exp_query_intf_params params; - struct ibv_exp_qp_init_attr init; - struct ibv_exp_res_domain_init_attr rd; - struct ibv_exp_cq_init_attr cq; - struct ibv_exp_qp_attr mod; + struct ibv_qp_init_attr init; + struct ibv_qp_attr mod; } attr; - enum ibv_exp_query_intf_status status; - int ret = 0; + int ret; (void)conf; /* Thresholds configuration (ignored). */ - if (priv == NULL) - return EINVAL; - if ((desc == 0) || (desc % MLX4_PMD_SGE_WR_N)) { - ERROR("%p: invalid number of TX descriptors (must be a" - " multiple of %d)", (void *)dev, MLX4_PMD_SGE_WR_N); - return EINVAL; + if (priv == NULL) { + rte_errno = EINVAL; + goto error; } - desc /= MLX4_PMD_SGE_WR_N; - /* MRs will be registered in mp2mr[] later. */ - attr.rd = (struct ibv_exp_res_domain_init_attr){ - .comp_mask = (IBV_EXP_RES_DOMAIN_THREAD_MODEL | - IBV_EXP_RES_DOMAIN_MSG_MODEL), - .thread_model = IBV_EXP_THREAD_SINGLE, - .msg_model = IBV_EXP_MSG_HIGH_BW, - }; - tmpl.rd = ibv_exp_create_res_domain(priv->ctx, &attr.rd); - if (tmpl.rd == NULL) { - ret = ENOMEM; - ERROR("%p: RD creation failure: %s", - (void *)dev, strerror(ret)); + if (desc == 0) { + rte_errno = EINVAL; + ERROR("%p: invalid number of Tx descriptors", (void *)dev); goto error; } - attr.cq = (struct ibv_exp_cq_init_attr){ - .comp_mask = IBV_EXP_CQ_INIT_ATTR_RES_DOMAIN, - .res_domain = tmpl.rd, - }; - tmpl.cq = ibv_exp_create_cq(priv->ctx, desc, NULL, NULL, 0, &attr.cq); + /* MRs will be registered in mp2mr[] later. */ + tmpl.cq = ibv_create_cq(priv->ctx, desc, NULL, NULL, 0); if (tmpl.cq == NULL) { - ret = ENOMEM; + rte_errno = ENOMEM; ERROR("%p: CQ creation failure: %s", - (void *)dev, strerror(ret)); + (void *)dev, strerror(rte_errno)); goto error; } DEBUG("priv->device_attr.max_qp_wr is %d", priv->device_attr.max_qp_wr); DEBUG("priv->device_attr.max_sge is %d", priv->device_attr.max_sge); - attr.init = (struct ibv_exp_qp_init_attr){ + attr.init = (struct ibv_qp_init_attr){ /* CQ to be associated with the send queue. */ .send_cq = tmpl.cq, /* CQ to be associated with the receive queue. */ @@ -1574,92 +1169,61 @@ txq_setup(struct rte_eth_dev *dev, struct txq *txq, uint16_t desc, priv->device_attr.max_qp_wr : desc), /* Max number of scatter/gather elements in a WR. */ - .max_send_sge = ((priv->device_attr.max_sge < - MLX4_PMD_SGE_WR_N) ? - priv->device_attr.max_sge : - MLX4_PMD_SGE_WR_N), + .max_send_sge = 1, .max_inline_data = MLX4_PMD_MAX_INLINE, }, .qp_type = IBV_QPT_RAW_PACKET, - /* Do *NOT* enable this, completions events are managed per - * TX burst. */ + /* + * Do *NOT* enable this, completions events are managed per + * Tx burst. + */ .sq_sig_all = 0, - .pd = priv->pd, - .res_domain = tmpl.rd, - .comp_mask = (IBV_EXP_QP_INIT_ATTR_PD | - IBV_EXP_QP_INIT_ATTR_RES_DOMAIN), }; - tmpl.qp = ibv_exp_create_qp(priv->ctx, &attr.init); + tmpl.qp = ibv_create_qp(priv->pd, &attr.init); if (tmpl.qp == NULL) { - ret = (errno ? errno : EINVAL); + rte_errno = errno ? errno : EINVAL; ERROR("%p: QP creation failure: %s", - (void *)dev, strerror(ret)); + (void *)dev, strerror(rte_errno)); goto error; } /* ibv_create_qp() updates this value. */ tmpl.max_inline = attr.init.cap.max_inline_data; - attr.mod = (struct ibv_exp_qp_attr){ + attr.mod = (struct ibv_qp_attr){ /* Move the QP to this state. */ .qp_state = IBV_QPS_INIT, /* Primary port number. */ .port_num = priv->port }; - ret = ibv_exp_modify_qp(tmpl.qp, &attr.mod, - (IBV_EXP_QP_STATE | IBV_EXP_QP_PORT)); + ret = ibv_modify_qp(tmpl.qp, &attr.mod, IBV_QP_STATE | IBV_QP_PORT); if (ret) { + rte_errno = ret; ERROR("%p: QP state to IBV_QPS_INIT failed: %s", - (void *)dev, strerror(ret)); + (void *)dev, strerror(rte_errno)); goto error; } ret = txq_alloc_elts(&tmpl, desc); if (ret) { + rte_errno = ret; ERROR("%p: TXQ allocation failed: %s", - (void *)dev, strerror(ret)); + (void *)dev, strerror(rte_errno)); goto error; } - attr.mod = (struct ibv_exp_qp_attr){ + attr.mod = (struct ibv_qp_attr){ .qp_state = IBV_QPS_RTR }; - ret = ibv_exp_modify_qp(tmpl.qp, &attr.mod, IBV_EXP_QP_STATE); + ret = ibv_modify_qp(tmpl.qp, &attr.mod, IBV_QP_STATE); if (ret) { + rte_errno = ret; ERROR("%p: QP state to IBV_QPS_RTR failed: %s", - (void *)dev, strerror(ret)); + (void *)dev, strerror(rte_errno)); goto error; } attr.mod.qp_state = IBV_QPS_RTS; - ret = ibv_exp_modify_qp(tmpl.qp, &attr.mod, IBV_EXP_QP_STATE); + ret = ibv_modify_qp(tmpl.qp, &attr.mod, IBV_QP_STATE); if (ret) { + rte_errno = ret; ERROR("%p: QP state to IBV_QPS_RTS failed: %s", - (void *)dev, strerror(ret)); - goto error; - } - attr.params = (struct ibv_exp_query_intf_params){ - .intf_scope = IBV_EXP_INTF_GLOBAL, - .intf = IBV_EXP_INTF_CQ, - .obj = tmpl.cq, - }; - tmpl.if_cq = ibv_exp_query_intf(priv->ctx, &attr.params, &status); - if (tmpl.if_cq == NULL) { - ERROR("%p: CQ interface family query failed with status %d", - (void *)dev, status); - goto error; - } - attr.params = (struct ibv_exp_query_intf_params){ - .intf_scope = IBV_EXP_INTF_GLOBAL, - .intf = IBV_EXP_INTF_QP_BURST, - .obj = tmpl.qp, -#ifdef HAVE_EXP_QP_BURST_CREATE_DISABLE_ETH_LOOPBACK - /* MC loopback must be disabled when not using a VF. */ - .family_flags = - (!priv->vf ? - IBV_EXP_QP_BURST_CREATE_DISABLE_ETH_LOOPBACK : - 0), -#endif - }; - tmpl.if_qp = ibv_exp_query_intf(priv->ctx, &attr.params, &status); - if (tmpl.if_qp == NULL) { - ERROR("%p: QP interface family query failed with status %d", - (void *)dev, status); + (void *)dev, strerror(rte_errno)); goto error; } /* Clean up txq in case we're reinitializing it. */ @@ -1669,12 +1233,13 @@ txq_setup(struct rte_eth_dev *dev, struct txq *txq, uint16_t desc, DEBUG("%p: txq updated with %p", (void *)txq, (void *)&tmpl); /* Pre-register known mempools. */ rte_mempool_walk(txq_mp2mr_iter, txq); - assert(ret == 0); return 0; error: + ret = rte_errno; txq_cleanup(&tmpl); - assert(ret > 0); - return ret; + rte_errno = ret; + assert(rte_errno > 0); + return -rte_errno; } /** @@ -1692,7 +1257,7 @@ error: * Thresholds parameters. * * @return - * 0 on success, negative errno value on failure. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int mlx4_tx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc, @@ -1702,31 +1267,30 @@ mlx4_tx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc, struct txq *txq = (*priv->txqs)[idx]; int ret; - priv_lock(priv); DEBUG("%p: configuring queue %u for %u descriptors", (void *)dev, idx, desc); if (idx >= priv->txqs_n) { + rte_errno = EOVERFLOW; ERROR("%p: queue index out of range (%u >= %u)", (void *)dev, idx, priv->txqs_n); - priv_unlock(priv); - return -EOVERFLOW; + return -rte_errno; } if (txq != NULL) { DEBUG("%p: reusing already allocated queue index %u (%p)", (void *)dev, idx, (void *)txq); if (priv->started) { - priv_unlock(priv); - return -EEXIST; + rte_errno = EEXIST; + return -rte_errno; } (*priv->txqs)[idx] = NULL; txq_cleanup(txq); } else { txq = rte_calloc_socket("TXQ", 1, sizeof(*txq), 0, socket); if (txq == NULL) { + rte_errno = ENOMEM; ERROR("%p: unable to allocate queue index %u", (void *)dev, idx); - priv_unlock(priv); - return -ENOMEM; + return -rte_errno; } } ret = txq_setup(dev, txq, desc, socket, conf); @@ -1740,8 +1304,7 @@ mlx4_tx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc, /* Update send callback. */ dev->tx_pkt_burst = mlx4_tx_burst; } - priv_unlock(priv); - return -ret; + return ret; } /** @@ -1760,7 +1323,6 @@ mlx4_tx_queue_release(void *dpdk_txq) if (txq == NULL) return; priv = txq->priv; - priv_lock(priv); for (i = 0; (i != priv->txqs_n); ++i) if ((*priv->txqs)[i] == txq) { DEBUG("%p: removing TX queue %p from list", @@ -1770,217 +1332,52 @@ mlx4_tx_queue_release(void *dpdk_txq) } txq_cleanup(txq); rte_free(txq); - priv_unlock(priv); } /* RX queues handling. */ /** - * Allocate RX queue elements with scattered packets support. + * Allocate RX queue elements. * * @param rxq * Pointer to RX queue structure. * @param elts_n * Number of elements to allocate. - * @param[in] pool - * If not NULL, fetch buffers from this array instead of allocating them - * with rte_pktmbuf_alloc(). * * @return - * 0 on success, errno value on failure. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int -rxq_alloc_elts_sp(struct rxq *rxq, unsigned int elts_n, - struct rte_mbuf **pool) +rxq_alloc_elts(struct rxq *rxq, unsigned int elts_n) { unsigned int i; - struct rxq_elt_sp (*elts)[elts_n] = + struct rxq_elt (*elts)[elts_n] = rte_calloc_socket("RXQ elements", 1, sizeof(*elts), 0, rxq->socket); - int ret = 0; if (elts == NULL) { + rte_errno = ENOMEM; ERROR("%p: can't allocate packets array", (void *)rxq); - ret = ENOMEM; goto error; } /* For each WR (packet). */ for (i = 0; (i != elts_n); ++i) { - unsigned int j; - struct rxq_elt_sp *elt = &(*elts)[i]; + struct rxq_elt *elt = &(*elts)[i]; struct ibv_recv_wr *wr = &elt->wr; - struct ibv_sge (*sges)[(elemof(elt->sges))] = &elt->sges; + struct ibv_sge *sge = &(*elts)[i].sge; + struct rte_mbuf *buf = rte_pktmbuf_alloc(rxq->mp); - /* These two arrays must have the same size. */ - assert(elemof(elt->sges) == elemof(elt->bufs)); - /* Configure WR. */ - wr->wr_id = i; - wr->next = &(*elts)[(i + 1)].wr; - wr->sg_list = &(*sges)[0]; - wr->num_sge = elemof(*sges); - /* For each SGE (segment). */ - for (j = 0; (j != elemof(elt->bufs)); ++j) { - struct ibv_sge *sge = &(*sges)[j]; - struct rte_mbuf *buf; - - if (pool != NULL) { - buf = *(pool++); - assert(buf != NULL); - rte_pktmbuf_reset(buf); - } else - buf = rte_pktmbuf_alloc(rxq->mp); - if (buf == NULL) { - assert(pool == NULL); - ERROR("%p: empty mbuf pool", (void *)rxq); - ret = ENOMEM; - goto error; - } - elt->bufs[j] = buf; - /* Headroom is reserved by rte_pktmbuf_alloc(). */ - assert(DATA_OFF(buf) == RTE_PKTMBUF_HEADROOM); - /* Buffer is supposed to be empty. */ - assert(rte_pktmbuf_data_len(buf) == 0); - assert(rte_pktmbuf_pkt_len(buf) == 0); - /* sge->addr must be able to store a pointer. */ - assert(sizeof(sge->addr) >= sizeof(uintptr_t)); - if (j == 0) { - /* The first SGE keeps its headroom. */ - sge->addr = rte_pktmbuf_mtod(buf, uintptr_t); - sge->length = (buf->buf_len - - RTE_PKTMBUF_HEADROOM); - } else { - /* Subsequent SGEs lose theirs. */ - assert(DATA_OFF(buf) == RTE_PKTMBUF_HEADROOM); - SET_DATA_OFF(buf, 0); - sge->addr = (uintptr_t)buf->buf_addr; - sge->length = buf->buf_len; - } - sge->lkey = rxq->mr->lkey; - /* Redundant check for tailroom. */ - assert(sge->length == rte_pktmbuf_tailroom(buf)); - } - } - /* The last WR pointer must be NULL. */ - (*elts)[(i - 1)].wr.next = NULL; - DEBUG("%p: allocated and configured %u WRs (%zu segments)", - (void *)rxq, elts_n, (elts_n * elemof((*elts)[0].sges))); - rxq->elts_n = elts_n; - rxq->elts_head = 0; - rxq->elts.sp = elts; - assert(ret == 0); - return 0; -error: - if (elts != NULL) { - assert(pool == NULL); - for (i = 0; (i != elemof(*elts)); ++i) { - unsigned int j; - struct rxq_elt_sp *elt = &(*elts)[i]; - - for (j = 0; (j != elemof(elt->bufs)); ++j) { - struct rte_mbuf *buf = elt->bufs[j]; - - if (buf != NULL) - rte_pktmbuf_free_seg(buf); - } - } - rte_free(elts); - } - DEBUG("%p: failed, freed everything", (void *)rxq); - assert(ret > 0); - return ret; -} - -/** - * Free RX queue elements with scattered packets support. - * - * @param rxq - * Pointer to RX queue structure. - */ -static void -rxq_free_elts_sp(struct rxq *rxq) -{ - unsigned int i; - unsigned int elts_n = rxq->elts_n; - struct rxq_elt_sp (*elts)[elts_n] = rxq->elts.sp; - - DEBUG("%p: freeing WRs", (void *)rxq); - rxq->elts_n = 0; - rxq->elts.sp = NULL; - if (elts == NULL) - return; - for (i = 0; (i != elemof(*elts)); ++i) { - unsigned int j; - struct rxq_elt_sp *elt = &(*elts)[i]; - - for (j = 0; (j != elemof(elt->bufs)); ++j) { - struct rte_mbuf *buf = elt->bufs[j]; - - if (buf != NULL) - rte_pktmbuf_free_seg(buf); - } - } - rte_free(elts); -} - -/** - * Allocate RX queue elements. - * - * @param rxq - * Pointer to RX queue structure. - * @param elts_n - * Number of elements to allocate. - * @param[in] pool - * If not NULL, fetch buffers from this array instead of allocating them - * with rte_pktmbuf_alloc(). - * - * @return - * 0 on success, errno value on failure. - */ -static int -rxq_alloc_elts(struct rxq *rxq, unsigned int elts_n, struct rte_mbuf **pool) -{ - unsigned int i; - struct rxq_elt (*elts)[elts_n] = - rte_calloc_socket("RXQ elements", 1, sizeof(*elts), 0, - rxq->socket); - int ret = 0; - - if (elts == NULL) { - ERROR("%p: can't allocate packets array", (void *)rxq); - ret = ENOMEM; - goto error; - } - /* For each WR (packet). */ - for (i = 0; (i != elts_n); ++i) { - struct rxq_elt *elt = &(*elts)[i]; - struct ibv_recv_wr *wr = &elt->wr; - struct ibv_sge *sge = &(*elts)[i].sge; - struct rte_mbuf *buf; - - if (pool != NULL) { - buf = *(pool++); - assert(buf != NULL); - rte_pktmbuf_reset(buf); - } else - buf = rte_pktmbuf_alloc(rxq->mp); if (buf == NULL) { - assert(pool == NULL); + rte_errno = ENOMEM; ERROR("%p: empty mbuf pool", (void *)rxq); - ret = ENOMEM; goto error; } - /* Configure WR. Work request ID contains its own index in - * the elts array and the offset between SGE buffer header and - * its data. */ - WR_ID(wr->wr_id).id = i; - WR_ID(wr->wr_id).offset = - (((uintptr_t)buf->buf_addr + RTE_PKTMBUF_HEADROOM) - - (uintptr_t)buf); + elt->buf = buf; wr->next = &(*elts)[(i + 1)].wr; wr->sg_list = sge; wr->num_sge = 1; /* Headroom is reserved by rte_pktmbuf_alloc(). */ - assert(DATA_OFF(buf) == RTE_PKTMBUF_HEADROOM); + assert(buf->data_off == RTE_PKTMBUF_HEADROOM); /* Buffer is supposed to be empty. */ assert(rte_pktmbuf_data_len(buf) == 0); assert(rte_pktmbuf_pkt_len(buf) == 0); @@ -1993,18 +1390,6 @@ rxq_alloc_elts(struct rxq *rxq, unsigned int elts_n, struct rte_mbuf **pool) sge->lkey = rxq->mr->lkey; /* Redundant check for tailroom. */ assert(sge->length == rte_pktmbuf_tailroom(buf)); - /* Make sure elts index and SGE mbuf pointer can be deduced - * from WR ID. */ - if ((WR_ID(wr->wr_id).id != i) || - ((void *)((uintptr_t)sge->addr - - WR_ID(wr->wr_id).offset) != buf)) { - ERROR("%p: cannot store index and offset in WR ID", - (void *)rxq); - sge->addr = 0; - rte_pktmbuf_free(buf); - ret = EOVERFLOW; - goto error; - } } /* The last WR pointer must be NULL. */ (*elts)[(i - 1)].wr.next = NULL; @@ -2012,28 +1397,17 @@ rxq_alloc_elts(struct rxq *rxq, unsigned int elts_n, struct rte_mbuf **pool) (void *)rxq, elts_n); rxq->elts_n = elts_n; rxq->elts_head = 0; - rxq->elts.no_sp = elts; - assert(ret == 0); + rxq->elts = elts; return 0; error: if (elts != NULL) { - assert(pool == NULL); - for (i = 0; (i != elemof(*elts)); ++i) { - struct rxq_elt *elt = &(*elts)[i]; - struct rte_mbuf *buf; - - if (elt->sge.addr == 0) - continue; - assert(WR_ID(elt->wr.wr_id).id == i); - buf = (void *)((uintptr_t)elt->sge.addr - - WR_ID(elt->wr.wr_id).offset); - rte_pktmbuf_free_seg(buf); - } + for (i = 0; (i != RTE_DIM(*elts)); ++i) + rte_pktmbuf_free_seg((*elts)[i].buf); rte_free(elts); } DEBUG("%p: failed, freed everything", (void *)rxq); - assert(ret > 0); - return ret; + assert(rte_errno > 0); + return -rte_errno; } /** @@ -2047,24 +1421,15 @@ rxq_free_elts(struct rxq *rxq) { unsigned int i; unsigned int elts_n = rxq->elts_n; - struct rxq_elt (*elts)[elts_n] = rxq->elts.no_sp; + struct rxq_elt (*elts)[elts_n] = rxq->elts; DEBUG("%p: freeing WRs", (void *)rxq); rxq->elts_n = 0; - rxq->elts.no_sp = NULL; + rxq->elts = NULL; if (elts == NULL) return; - for (i = 0; (i != elemof(*elts)); ++i) { - struct rxq_elt *elt = &(*elts)[i]; - struct rte_mbuf *buf; - - if (elt->sge.addr == 0) - continue; - assert(WR_ID(elt->wr.wr_id).id == i); - buf = (void *)((uintptr_t)elt->sge.addr - - WR_ID(elt->wr.wr_id).offset); - rte_pktmbuf_free_seg(buf); - } + for (i = 0; (i != RTE_DIM(*elts)); ++i) + rte_pktmbuf_free_seg((*elts)[i].buf); rte_free(elts); } @@ -2093,14 +1458,13 @@ priv_mac_addr_del(struct priv *priv) /** * Register a MAC address. * - * In RSS mode, the MAC address is registered in the parent queue, - * otherwise it is registered in queue 0. + * The MAC address is registered in queue 0. * * @param priv * Pointer to private structure. * * @return - * 0 on success, errno value on failure. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int priv_mac_addr_add(struct priv *priv) @@ -2114,9 +1478,7 @@ priv_mac_addr_add(struct priv *priv) return 0; if (priv->isolated) return 0; - if (priv->rss) - rxq = LIST_FIRST(&priv->parents); - else if (*priv->rxqs && (*priv->rxqs)[0]) + if (*priv->rxqs && (*priv->rxqs)[0]) rxq = (*priv->rxqs)[0]; else return 0; @@ -2160,16 +1522,12 @@ priv_mac_addr_add(struct priv *priv) (void *)priv, (*mac)[0], (*mac)[1], (*mac)[2], (*mac)[3], (*mac)[4], (*mac)[5]); /* Create related flow. */ - errno = 0; flow = ibv_create_flow(rxq->qp, attr); if (flow == NULL) { - /* It's not clear whether errno is always set in this case. */ + rte_errno = errno ? errno : EINVAL; ERROR("%p: flow configuration failed, errno=%d: %s", - (void *)rxq, errno, - (errno ? strerror(errno) : "Unknown error")); - if (errno) - return errno; - return EINVAL; + (void *)rxq, rte_errno, strerror(errno)); + return -rte_errno; } assert(priv->mac_flow == NULL); priv->mac_flow = flow; @@ -2187,138 +1545,23 @@ priv_mac_addr_add(struct priv *priv) static void rxq_cleanup(struct rxq *rxq) { - struct ibv_exp_release_intf_params params; - DEBUG("cleaning up %p", (void *)rxq); - if (rxq->sp) - rxq_free_elts_sp(rxq); - else - rxq_free_elts(rxq); - if (rxq->if_qp != NULL) { - assert(rxq->priv != NULL); - assert(rxq->priv->ctx != NULL); - assert(rxq->qp != NULL); - params = (struct ibv_exp_release_intf_params){ - .comp_mask = 0, - }; - claim_zero(ibv_exp_release_intf(rxq->priv->ctx, - rxq->if_qp, - ¶ms)); - } - if (rxq->if_cq != NULL) { - assert(rxq->priv != NULL); - assert(rxq->priv->ctx != NULL); - assert(rxq->cq != NULL); - params = (struct ibv_exp_release_intf_params){ - .comp_mask = 0, - }; - claim_zero(ibv_exp_release_intf(rxq->priv->ctx, - rxq->if_cq, - ¶ms)); - } + rxq_free_elts(rxq); if (rxq->qp != NULL) claim_zero(ibv_destroy_qp(rxq->qp)); if (rxq->cq != NULL) claim_zero(ibv_destroy_cq(rxq->cq)); if (rxq->channel != NULL) claim_zero(ibv_destroy_comp_channel(rxq->channel)); - if (rxq->rd != NULL) { - struct ibv_exp_destroy_res_domain_attr attr = { - .comp_mask = 0, - }; - - assert(rxq->priv != NULL); - assert(rxq->priv->ctx != NULL); - claim_zero(ibv_exp_destroy_res_domain(rxq->priv->ctx, - rxq->rd, - &attr)); - } if (rxq->mr != NULL) claim_zero(ibv_dereg_mr(rxq->mr)); memset(rxq, 0, sizeof(*rxq)); } /** - * Translate RX completion flags to packet type. - * - * @param flags - * RX completion flags returned by poll_length_flags(). - * - * @note: fix mlx4_dev_supported_ptypes_get() if any change here. - * - * @return - * Packet type for struct rte_mbuf. - */ -static inline uint32_t -rxq_cq_to_pkt_type(uint32_t flags) -{ - uint32_t pkt_type; - - if (flags & IBV_EXP_CQ_RX_TUNNEL_PACKET) - pkt_type = - TRANSPOSE(flags, - IBV_EXP_CQ_RX_OUTER_IPV4_PACKET, - RTE_PTYPE_L3_IPV4_EXT_UNKNOWN) | - TRANSPOSE(flags, - IBV_EXP_CQ_RX_OUTER_IPV6_PACKET, - RTE_PTYPE_L3_IPV6_EXT_UNKNOWN) | - TRANSPOSE(flags, - IBV_EXP_CQ_RX_IPV4_PACKET, - RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN) | - TRANSPOSE(flags, - IBV_EXP_CQ_RX_IPV6_PACKET, - RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN); - else - pkt_type = - TRANSPOSE(flags, - IBV_EXP_CQ_RX_IPV4_PACKET, - RTE_PTYPE_L3_IPV4_EXT_UNKNOWN) | - TRANSPOSE(flags, - IBV_EXP_CQ_RX_IPV6_PACKET, - RTE_PTYPE_L3_IPV6_EXT_UNKNOWN); - return pkt_type; -} - -/** - * Translate RX completion flags to offload flags. - * - * @param[in] rxq - * Pointer to RX queue structure. - * @param flags - * RX completion flags returned by poll_length_flags(). + * DPDK callback for RX. * - * @return - * Offload flags (ol_flags) for struct rte_mbuf. - */ -static inline uint32_t -rxq_cq_to_ol_flags(const struct rxq *rxq, uint32_t flags) -{ - uint32_t ol_flags = 0; - - if (rxq->csum) - ol_flags |= - TRANSPOSE(flags, - IBV_EXP_CQ_RX_IP_CSUM_OK, - PKT_RX_IP_CKSUM_GOOD) | - TRANSPOSE(flags, - IBV_EXP_CQ_RX_TCP_UDP_CSUM_OK, - PKT_RX_L4_CKSUM_GOOD); - if ((flags & IBV_EXP_CQ_RX_TUNNEL_PACKET) && (rxq->csum_l2tun)) - ol_flags |= - TRANSPOSE(flags, - IBV_EXP_CQ_RX_OUTER_IP_CSUM_OK, - PKT_RX_IP_CKSUM_GOOD) | - TRANSPOSE(flags, - IBV_EXP_CQ_RX_OUTER_TCP_UDP_CSUM_OK, - PKT_RX_L4_CKSUM_GOOD); - return ol_flags; -} - -static uint16_t -mlx4_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n); - -/** - * DPDK callback for RX with scattered packets support. + * The following function doesn't manage scattered packets. * * @param dpdk_rxq * Generic pointer to RX queue structure. @@ -2331,250 +1574,39 @@ mlx4_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n); * Number of packets successfully received (<= pkts_n). */ static uint16_t -mlx4_rx_burst_sp(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n) +mlx4_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n) { struct rxq *rxq = (struct rxq *)dpdk_rxq; - struct rxq_elt_sp (*elts)[rxq->elts_n] = rxq->elts.sp; + struct rxq_elt (*elts)[rxq->elts_n] = rxq->elts; const unsigned int elts_n = rxq->elts_n; unsigned int elts_head = rxq->elts_head; - struct ibv_recv_wr head; - struct ibv_recv_wr **next = &head.next; - struct ibv_recv_wr *bad_wr; + struct ibv_wc wcs[pkts_n]; + struct ibv_recv_wr *wr_head = NULL; + struct ibv_recv_wr **wr_next = &wr_head; + struct ibv_recv_wr *wr_bad = NULL; unsigned int i; unsigned int pkts_ret = 0; int ret; - if (unlikely(!rxq->sp)) - return mlx4_rx_burst(dpdk_rxq, pkts, pkts_n); - if (unlikely(elts == NULL)) /* See RTE_DEV_CMD_SET_MTU. */ + ret = ibv_poll_cq(rxq->cq, pkts_n, wcs); + if (unlikely(ret == 0)) return 0; - for (i = 0; (i != pkts_n); ++i) { - struct rxq_elt_sp *elt = &(*elts)[elts_head]; - struct ibv_recv_wr *wr = &elt->wr; - uint64_t wr_id = wr->wr_id; - unsigned int len; - unsigned int pkt_buf_len; - struct rte_mbuf *pkt_buf = NULL; /* Buffer returned in pkts. */ - struct rte_mbuf **pkt_buf_next = &pkt_buf; - unsigned int seg_headroom = RTE_PKTMBUF_HEADROOM; - unsigned int j = 0; - uint32_t flags; - - /* Sanity checks. */ -#ifdef NDEBUG - (void)wr_id; -#endif - assert(wr_id < rxq->elts_n); - assert(wr->sg_list == elt->sges); - assert(wr->num_sge == elemof(elt->sges)); - assert(elts_head < rxq->elts_n); - assert(rxq->elts_head < rxq->elts_n); - ret = rxq->if_cq->poll_length_flags(rxq->cq, NULL, NULL, - &flags); - if (unlikely(ret < 0)) { - struct ibv_wc wc; - int wcs_n; - - DEBUG("rxq=%p, poll_length() failed (ret=%d)", - (void *)rxq, ret); - /* ibv_poll_cq() must be used in case of failure. */ - wcs_n = ibv_poll_cq(rxq->cq, 1, &wc); - if (unlikely(wcs_n == 0)) - break; - if (unlikely(wcs_n < 0)) { - DEBUG("rxq=%p, ibv_poll_cq() failed (wcs_n=%d)", - (void *)rxq, wcs_n); - break; - } - assert(wcs_n == 1); - if (unlikely(wc.status != IBV_WC_SUCCESS)) { - /* Whatever, just repost the offending WR. */ - DEBUG("rxq=%p, wr_id=%" PRIu64 ": bad work" - " completion status (%d): %s", - (void *)rxq, wc.wr_id, wc.status, - ibv_wc_status_str(wc.status)); - /* Increment dropped packets counter. */ - ++rxq->stats.idropped; - /* Link completed WRs together for repost. */ - *next = wr; - next = &wr->next; - goto repost; - } - ret = wc.byte_len; - } - if (ret == 0) - break; - len = ret; - pkt_buf_len = len; - /* Link completed WRs together for repost. */ - *next = wr; - next = &wr->next; - /* - * Replace spent segments with new ones, concatenate and - * return them as pkt_buf. - */ - while (1) { - struct ibv_sge *sge = &elt->sges[j]; - struct rte_mbuf *seg = elt->bufs[j]; - struct rte_mbuf *rep; - unsigned int seg_tailroom; - - /* - * Fetch initial bytes of packet descriptor into a - * cacheline while allocating rep. - */ - rte_prefetch0(seg); - rep = rte_mbuf_raw_alloc(rxq->mp); - if (unlikely(rep == NULL)) { - /* - * Unable to allocate a replacement mbuf, - * repost WR. - */ - DEBUG("rxq=%p, wr_id=%" PRIu64 ":" - " can't allocate a new mbuf", - (void *)rxq, wr_id); - if (pkt_buf != NULL) { - *pkt_buf_next = NULL; - rte_pktmbuf_free(pkt_buf); - } - /* Increase out of memory counters. */ - ++rxq->stats.rx_nombuf; - ++rxq->priv->dev->data->rx_mbuf_alloc_failed; - goto repost; - } -#ifndef NDEBUG - /* Poison user-modifiable fields in rep. */ - NEXT(rep) = (void *)((uintptr_t)-1); - SET_DATA_OFF(rep, 0xdead); - DATA_LEN(rep) = 0xd00d; - PKT_LEN(rep) = 0xdeadd00d; - NB_SEGS(rep) = 0x2a; - PORT(rep) = 0x2a; - rep->ol_flags = -1; - /* - * Clear special flags in mbuf to avoid - * crashing while freeing. - */ - rep->ol_flags &= - ~(uint64_t)(IND_ATTACHED_MBUF | - CTRL_MBUF_FLAG); -#endif - assert(rep->buf_len == seg->buf_len); - /* Reconfigure sge to use rep instead of seg. */ - assert(sge->lkey == rxq->mr->lkey); - sge->addr = ((uintptr_t)rep->buf_addr + seg_headroom); - elt->bufs[j] = rep; - ++j; - /* Update pkt_buf if it's the first segment, or link - * seg to the previous one and update pkt_buf_next. */ - *pkt_buf_next = seg; - pkt_buf_next = &NEXT(seg); - /* Update seg information. */ - seg_tailroom = (seg->buf_len - seg_headroom); - assert(sge->length == seg_tailroom); - SET_DATA_OFF(seg, seg_headroom); - if (likely(len <= seg_tailroom)) { - /* Last segment. */ - DATA_LEN(seg) = len; - PKT_LEN(seg) = len; - /* Sanity check. */ - assert(rte_pktmbuf_headroom(seg) == - seg_headroom); - assert(rte_pktmbuf_tailroom(seg) == - (seg_tailroom - len)); - break; - } - DATA_LEN(seg) = seg_tailroom; - PKT_LEN(seg) = seg_tailroom; - /* Sanity check. */ - assert(rte_pktmbuf_headroom(seg) == seg_headroom); - assert(rte_pktmbuf_tailroom(seg) == 0); - /* Fix len and clear headroom for next segments. */ - len -= seg_tailroom; - seg_headroom = 0; - } - /* Update head and tail segments. */ - *pkt_buf_next = NULL; - assert(pkt_buf != NULL); - assert(j != 0); - NB_SEGS(pkt_buf) = j; - PORT(pkt_buf) = rxq->port_id; - PKT_LEN(pkt_buf) = pkt_buf_len; - pkt_buf->packet_type = rxq_cq_to_pkt_type(flags); - pkt_buf->ol_flags = rxq_cq_to_ol_flags(rxq, flags); - - /* Return packet. */ - *(pkts++) = pkt_buf; - ++pkts_ret; - /* Increase bytes counter. */ - rxq->stats.ibytes += pkt_buf_len; -repost: - if (++elts_head >= elts_n) - elts_head = 0; - continue; - } - if (unlikely(i == 0)) + if (unlikely(ret < 0)) { + DEBUG("rxq=%p, ibv_poll_cq() failed (wc_n=%d)", + (void *)rxq, ret); return 0; - *next = NULL; - /* Repost WRs. */ - ret = ibv_post_recv(rxq->qp, head.next, &bad_wr); - if (unlikely(ret)) { - /* Inability to repost WRs is fatal. */ - DEBUG("%p: ibv_post_recv(): failed for WR %p: %s", - (void *)rxq->priv, - (void *)bad_wr, - strerror(ret)); - abort(); } - rxq->elts_head = elts_head; - /* Increase packets counter. */ - rxq->stats.ipackets += pkts_ret; - return pkts_ret; -} - -/** - * DPDK callback for RX. - * - * The following function is the same as mlx4_rx_burst_sp(), except it doesn't - * manage scattered packets. Improves performance when MRU is lower than the - * size of the first segment. - * - * @param dpdk_rxq - * Generic pointer to RX queue structure. - * @param[out] pkts - * Array to store received packets. - * @param pkts_n - * Maximum number of packets in array. - * - * @return - * Number of packets successfully received (<= pkts_n). - */ -static uint16_t -mlx4_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n) -{ - struct rxq *rxq = (struct rxq *)dpdk_rxq; - struct rxq_elt (*elts)[rxq->elts_n] = rxq->elts.no_sp; - const unsigned int elts_n = rxq->elts_n; - unsigned int elts_head = rxq->elts_head; - struct ibv_sge sges[pkts_n]; - unsigned int i; - unsigned int pkts_ret = 0; - int ret; - - if (unlikely(rxq->sp)) - return mlx4_rx_burst_sp(dpdk_rxq, pkts, pkts_n); - for (i = 0; (i != pkts_n); ++i) { + assert(ret <= (int)pkts_n); + /* For each work completion. */ + for (i = 0; i != (unsigned int)ret; ++i) { + struct ibv_wc *wc = &wcs[i]; struct rxq_elt *elt = &(*elts)[elts_head]; struct ibv_recv_wr *wr = &elt->wr; - uint64_t wr_id = wr->wr_id; - unsigned int len; - struct rte_mbuf *seg = (void *)((uintptr_t)elt->sge.addr - - WR_ID(wr_id).offset); + uint32_t len = wc->byte_len; + struct rte_mbuf *seg = elt->buf; struct rte_mbuf *rep; - uint32_t flags; /* Sanity checks. */ - assert(WR_ID(wr_id).id < rxq->elts_n); assert(wr->sg_list == &elt->sge); assert(wr->num_sge == 1); assert(elts_head < rxq->elts_n); @@ -2585,79 +1617,44 @@ mlx4_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n) */ rte_mbuf_prefetch_part1(seg); rte_mbuf_prefetch_part2(seg); - ret = rxq->if_cq->poll_length_flags(rxq->cq, NULL, NULL, - &flags); - if (unlikely(ret < 0)) { - struct ibv_wc wc; - int wcs_n; - - DEBUG("rxq=%p, poll_length() failed (ret=%d)", - (void *)rxq, ret); - /* ibv_poll_cq() must be used in case of failure. */ - wcs_n = ibv_poll_cq(rxq->cq, 1, &wc); - if (unlikely(wcs_n == 0)) - break; - if (unlikely(wcs_n < 0)) { - DEBUG("rxq=%p, ibv_poll_cq() failed (wcs_n=%d)", - (void *)rxq, wcs_n); - break; - } - assert(wcs_n == 1); - if (unlikely(wc.status != IBV_WC_SUCCESS)) { - /* Whatever, just repost the offending WR. */ - DEBUG("rxq=%p, wr_id=%" PRIu64 ": bad work" - " completion status (%d): %s", - (void *)rxq, wc.wr_id, wc.status, - ibv_wc_status_str(wc.status)); - /* Increment dropped packets counter. */ - ++rxq->stats.idropped; - /* Add SGE to array for repost. */ - sges[i] = elt->sge; - goto repost; - } - ret = wc.byte_len; + /* Link completed WRs together for repost. */ + *wr_next = wr; + wr_next = &wr->next; + if (unlikely(wc->status != IBV_WC_SUCCESS)) { + /* Whatever, just repost the offending WR. */ + DEBUG("rxq=%p: bad work completion status (%d): %s", + (void *)rxq, wc->status, + ibv_wc_status_str(wc->status)); + /* Increment dropped packets counter. */ + ++rxq->stats.idropped; + goto repost; } - if (ret == 0) - break; - len = ret; rep = rte_mbuf_raw_alloc(rxq->mp); if (unlikely(rep == NULL)) { /* * Unable to allocate a replacement mbuf, * repost WR. */ - DEBUG("rxq=%p, wr_id=%" PRIu32 ":" - " can't allocate a new mbuf", - (void *)rxq, WR_ID(wr_id).id); + DEBUG("rxq=%p: can't allocate a new mbuf", + (void *)rxq); /* Increase out of memory counters. */ ++rxq->stats.rx_nombuf; ++rxq->priv->dev->data->rx_mbuf_alloc_failed; - /* Add SGE to array for repost. */ - sges[i] = elt->sge; goto repost; } - /* Reconfigure sge to use rep instead of seg. */ elt->sge.addr = (uintptr_t)rep->buf_addr + RTE_PKTMBUF_HEADROOM; assert(elt->sge.lkey == rxq->mr->lkey); - WR_ID(wr->wr_id).offset = - (((uintptr_t)rep->buf_addr + RTE_PKTMBUF_HEADROOM) - - (uintptr_t)rep); - assert(WR_ID(wr->wr_id).id == WR_ID(wr_id).id); - - /* Add SGE to array for repost. */ - sges[i] = elt->sge; - + elt->buf = rep; /* Update seg information. */ - SET_DATA_OFF(seg, RTE_PKTMBUF_HEADROOM); - NB_SEGS(seg) = 1; - PORT(seg) = rxq->port_id; - NEXT(seg) = NULL; - PKT_LEN(seg) = len; - DATA_LEN(seg) = len; - seg->packet_type = rxq_cq_to_pkt_type(flags); - seg->ol_flags = rxq_cq_to_ol_flags(rxq, flags); - + seg->data_off = RTE_PKTMBUF_HEADROOM; + seg->nb_segs = 1; + seg->port = rxq->port_id; + seg->next = NULL; + seg->pkt_len = len; + seg->data_len = len; + seg->packet_type = 0; + seg->ol_flags = 0; /* Return packet. */ *(pkts++) = seg; ++pkts_ret; @@ -2671,7 +1668,9 @@ repost: if (unlikely(i == 0)) return 0; /* Repost WRs. */ - ret = rxq->if_qp->recv_burst(rxq->qp, sges, i); + *wr_next = NULL; + assert(wr_head); + ret = ibv_post_recv(rxq->qp, wr_head, &wr_bad); if (unlikely(ret)) { /* Inability to repost WRs is fatal. */ DEBUG("%p: recv_burst(): failed (ret=%d)", @@ -2679,390 +1678,50 @@ repost: ret); abort(); } - rxq->elts_head = elts_head; - /* Increase packets counter. */ - rxq->stats.ipackets += pkts_ret; - return pkts_ret; -} - -/** - * Allocate a Queue Pair. - * Optionally setup inline receive if supported. - * - * @param priv - * Pointer to private structure. - * @param cq - * Completion queue to associate with QP. - * @param desc - * Number of descriptors in QP (hint only). - * - * @return - * QP pointer or NULL in case of error. - */ -static struct ibv_qp * -rxq_setup_qp(struct priv *priv, struct ibv_cq *cq, uint16_t desc, - struct ibv_exp_res_domain *rd) -{ - struct ibv_exp_qp_init_attr attr = { - /* CQ to be associated with the send queue. */ - .send_cq = cq, - /* CQ to be associated with the receive queue. */ - .recv_cq = cq, - .cap = { - /* Max number of outstanding WRs. */ - .max_recv_wr = ((priv->device_attr.max_qp_wr < desc) ? - priv->device_attr.max_qp_wr : - desc), - /* Max number of scatter/gather elements in a WR. */ - .max_recv_sge = ((priv->device_attr.max_sge < - MLX4_PMD_SGE_WR_N) ? - priv->device_attr.max_sge : - MLX4_PMD_SGE_WR_N), - }, - .qp_type = IBV_QPT_RAW_PACKET, - .comp_mask = (IBV_EXP_QP_INIT_ATTR_PD | - IBV_EXP_QP_INIT_ATTR_RES_DOMAIN), - .pd = priv->pd, - .res_domain = rd, - }; - - attr.max_inl_recv = priv->inl_recv_size; - attr.comp_mask |= IBV_EXP_QP_INIT_ATTR_INL_RECV; - return ibv_exp_create_qp(priv->ctx, &attr); -} - -/** - * Allocate a RSS Queue Pair. - * Optionally setup inline receive if supported. - * - * @param priv - * Pointer to private structure. - * @param cq - * Completion queue to associate with QP. - * @param desc - * Number of descriptors in QP (hint only). - * @param children_n - * If nonzero, a number of children for parent QP and zero for a child. - * @param rxq_parent - * Pointer for a parent in a child case, NULL otherwise. - * - * @return - * QP pointer or NULL in case of error. - */ -static struct ibv_qp * -rxq_setup_qp_rss(struct priv *priv, struct ibv_cq *cq, uint16_t desc, - int children_n, struct ibv_exp_res_domain *rd, - struct rxq *rxq_parent) -{ - struct ibv_exp_qp_init_attr attr = { - /* CQ to be associated with the send queue. */ - .send_cq = cq, - /* CQ to be associated with the receive queue. */ - .recv_cq = cq, - .cap = { - /* Max number of outstanding WRs. */ - .max_recv_wr = ((priv->device_attr.max_qp_wr < desc) ? - priv->device_attr.max_qp_wr : - desc), - /* Max number of scatter/gather elements in a WR. */ - .max_recv_sge = ((priv->device_attr.max_sge < - MLX4_PMD_SGE_WR_N) ? - priv->device_attr.max_sge : - MLX4_PMD_SGE_WR_N), - }, - .qp_type = IBV_QPT_RAW_PACKET, - .comp_mask = (IBV_EXP_QP_INIT_ATTR_PD | - IBV_EXP_QP_INIT_ATTR_RES_DOMAIN | - IBV_EXP_QP_INIT_ATTR_QPG), - .pd = priv->pd, - .res_domain = rd, - }; - - attr.max_inl_recv = priv->inl_recv_size, - attr.comp_mask |= IBV_EXP_QP_INIT_ATTR_INL_RECV; - if (children_n > 0) { - attr.qpg.qpg_type = IBV_EXP_QPG_PARENT; - /* TSS isn't necessary. */ - attr.qpg.parent_attrib.tss_child_count = 0; - attr.qpg.parent_attrib.rss_child_count = - rte_align32pow2(children_n + 1) >> 1; - DEBUG("initializing parent RSS queue"); - } else { - attr.qpg.qpg_type = IBV_EXP_QPG_CHILD_RX; - attr.qpg.qpg_parent = rxq_parent->qp; - DEBUG("initializing child RSS queue"); - } - return ibv_exp_create_qp(priv->ctx, &attr); -} - -/** - * Reconfigure a RX queue with new parameters. - * - * rxq_rehash() does not allocate mbufs, which, if not done from the right - * thread (such as a control thread), may corrupt the pool. - * In case of failure, the queue is left untouched. - * - * @param dev - * Pointer to Ethernet device structure. - * @param rxq - * RX queue pointer. - * - * @return - * 0 on success, errno value on failure. - */ -static int -rxq_rehash(struct rte_eth_dev *dev, struct rxq *rxq) -{ - struct priv *priv = rxq->priv; - struct rxq tmpl = *rxq; - unsigned int mbuf_n; - unsigned int desc_n; - struct rte_mbuf **pool; - unsigned int i, k; - struct ibv_exp_qp_attr mod; - struct ibv_recv_wr *bad_wr; - unsigned int mb_len; - int err; - - mb_len = rte_pktmbuf_data_room_size(rxq->mp); - DEBUG("%p: rehashing queue %p", (void *)dev, (void *)rxq); - /* Number of descriptors and mbufs currently allocated. */ - desc_n = (tmpl.elts_n * (tmpl.sp ? MLX4_PMD_SGE_WR_N : 1)); - mbuf_n = desc_n; - /* Toggle RX checksum offload if hardware supports it. */ - if (priv->hw_csum) { - tmpl.csum = !!dev->data->dev_conf.rxmode.hw_ip_checksum; - rxq->csum = tmpl.csum; - } - if (priv->hw_csum_l2tun) { - tmpl.csum_l2tun = !!dev->data->dev_conf.rxmode.hw_ip_checksum; - rxq->csum_l2tun = tmpl.csum_l2tun; - } - /* Enable scattered packets support for this queue if necessary. */ - assert(mb_len >= RTE_PKTMBUF_HEADROOM); - if (dev->data->dev_conf.rxmode.enable_scatter && - (dev->data->dev_conf.rxmode.max_rx_pkt_len > - (mb_len - RTE_PKTMBUF_HEADROOM))) { - tmpl.sp = 1; - desc_n /= MLX4_PMD_SGE_WR_N; - } else - tmpl.sp = 0; - DEBUG("%p: %s scattered packets support (%u WRs)", - (void *)dev, (tmpl.sp ? "enabling" : "disabling"), desc_n); - /* If scatter mode is the same as before, nothing to do. */ - if (tmpl.sp == rxq->sp) { - DEBUG("%p: nothing to do", (void *)dev); - return 0; - } - /* From now on, any failure will render the queue unusable. - * Reinitialize QP. */ - if (!tmpl.qp) - goto skip_init; - mod = (struct ibv_exp_qp_attr){ .qp_state = IBV_QPS_RESET }; - err = ibv_exp_modify_qp(tmpl.qp, &mod, IBV_EXP_QP_STATE); - if (err) { - ERROR("%p: cannot reset QP: %s", (void *)dev, strerror(err)); - assert(err > 0); - return err; - } - mod = (struct ibv_exp_qp_attr){ - /* Move the QP to this state. */ - .qp_state = IBV_QPS_INIT, - /* Primary port number. */ - .port_num = priv->port - }; - err = ibv_exp_modify_qp(tmpl.qp, &mod, - (IBV_EXP_QP_STATE | - IBV_EXP_QP_PORT)); - if (err) { - ERROR("%p: QP state to IBV_QPS_INIT failed: %s", - (void *)dev, strerror(err)); - assert(err > 0); - return err; - }; -skip_init: - err = ibv_resize_cq(tmpl.cq, desc_n); - if (err) { - ERROR("%p: cannot resize CQ: %s", (void *)dev, strerror(err)); - assert(err > 0); - return err; - } - /* Allocate pool. */ - pool = rte_malloc(__func__, (mbuf_n * sizeof(*pool)), 0); - if (pool == NULL) { - ERROR("%p: cannot allocate memory", (void *)dev); - return ENOBUFS; - } - /* Snatch mbufs from original queue. */ - k = 0; - if (rxq->sp) { - struct rxq_elt_sp (*elts)[rxq->elts_n] = rxq->elts.sp; - - for (i = 0; (i != elemof(*elts)); ++i) { - struct rxq_elt_sp *elt = &(*elts)[i]; - unsigned int j; - - for (j = 0; (j != elemof(elt->bufs)); ++j) { - assert(elt->bufs[j] != NULL); - pool[k++] = elt->bufs[j]; - } - } - } else { - struct rxq_elt (*elts)[rxq->elts_n] = rxq->elts.no_sp; - - for (i = 0; (i != elemof(*elts)); ++i) { - struct rxq_elt *elt = &(*elts)[i]; - struct rte_mbuf *buf = (void *) - ((uintptr_t)elt->sge.addr - - WR_ID(elt->wr.wr_id).offset); - - assert(WR_ID(elt->wr.wr_id).id == i); - pool[k++] = buf; - } - } - assert(k == mbuf_n); - tmpl.elts_n = 0; - tmpl.elts.sp = NULL; - assert((void *)&tmpl.elts.sp == (void *)&tmpl.elts.no_sp); - err = ((tmpl.sp) ? - rxq_alloc_elts_sp(&tmpl, desc_n, pool) : - rxq_alloc_elts(&tmpl, desc_n, pool)); - if (err) { - ERROR("%p: cannot reallocate WRs, aborting", (void *)dev); - rte_free(pool); - assert(err > 0); - return err; - } - assert(tmpl.elts_n == desc_n); - assert(tmpl.elts.sp != NULL); - rte_free(pool); - /* Clean up original data. */ - rxq->elts_n = 0; - rte_free(rxq->elts.sp); - rxq->elts.sp = NULL; - if (!tmpl.qp) - goto skip_rtr; - /* Post WRs. */ - err = ibv_post_recv(tmpl.qp, - (tmpl.sp ? - &(*tmpl.elts.sp)[0].wr : - &(*tmpl.elts.no_sp)[0].wr), - &bad_wr); - if (err) { - ERROR("%p: ibv_post_recv() failed for WR %p: %s", - (void *)dev, - (void *)bad_wr, - strerror(err)); - goto skip_rtr; - } - mod = (struct ibv_exp_qp_attr){ - .qp_state = IBV_QPS_RTR - }; - err = ibv_exp_modify_qp(tmpl.qp, &mod, IBV_EXP_QP_STATE); - if (err) - ERROR("%p: QP state to IBV_QPS_RTR failed: %s", - (void *)dev, strerror(err)); -skip_rtr: - *rxq = tmpl; - assert(err >= 0); - return err; + rxq->elts_head = elts_head; + /* Increase packets counter. */ + rxq->stats.ipackets += pkts_ret; + return pkts_ret; } /** - * Create verbs QP resources associated with a rxq. + * Allocate a Queue Pair. + * Optionally setup inline receive if supported. * - * @param rxq - * Pointer to RX queue structure. + * @param priv + * Pointer to private structure. + * @param cq + * Completion queue to associate with QP. * @param desc - * Number of descriptors to configure in queue. - * @param inactive - * If true, the queue is disabled because its index is higher or - * equal to the real number of queues, which must be a power of 2. - * @param children_n - * The number of children in a parent case, zero for a child. - * @param rxq_parent - * The pointer to a parent RX structure for a child in RSS case, - * NULL for parent. + * Number of descriptors in QP (hint only). * * @return - * 0 on success, errno value on failure. + * QP pointer or NULL in case of error and rte_errno is set. */ -int -rxq_create_qp(struct rxq *rxq, - uint16_t desc, - int inactive, - int children_n, - struct rxq *rxq_parent) +static struct ibv_qp * +rxq_setup_qp(struct priv *priv, struct ibv_cq *cq, uint16_t desc) { - int ret; - struct ibv_exp_qp_attr mod; - struct ibv_exp_query_intf_params params; - enum ibv_exp_query_intf_status status; - struct ibv_recv_wr *bad_wr; - int parent = (children_n > 0); - struct priv *priv = rxq->priv; - - if (priv->rss && !inactive && (rxq_parent || parent)) - rxq->qp = rxq_setup_qp_rss(priv, rxq->cq, desc, - children_n, rxq->rd, - rxq_parent); - else - rxq->qp = rxq_setup_qp(priv, rxq->cq, desc, rxq->rd); - if (rxq->qp == NULL) { - ret = (errno ? errno : EINVAL); - ERROR("QP creation failure: %s", - strerror(ret)); - return ret; - } - mod = (struct ibv_exp_qp_attr){ - /* Move the QP to this state. */ - .qp_state = IBV_QPS_INIT, - /* Primary port number. */ - .port_num = priv->port - }; - ret = ibv_exp_modify_qp(rxq->qp, &mod, - (IBV_EXP_QP_STATE | - (parent ? IBV_EXP_QP_GROUP_RSS : 0) | - IBV_EXP_QP_PORT)); - if (ret) { - ERROR("QP state to IBV_QPS_INIT failed: %s", - strerror(ret)); - return ret; - } - if (!parent) { - ret = ibv_post_recv(rxq->qp, - (rxq->sp ? - &(*rxq->elts.sp)[0].wr : - &(*rxq->elts.no_sp)[0].wr), - &bad_wr); - if (ret) { - ERROR("ibv_post_recv() failed for WR %p: %s", - (void *)bad_wr, - strerror(ret)); - return ret; - } - } - mod = (struct ibv_exp_qp_attr){ - .qp_state = IBV_QPS_RTR - }; - ret = ibv_exp_modify_qp(rxq->qp, &mod, IBV_EXP_QP_STATE); - if (ret) { - ERROR("QP state to IBV_QPS_RTR failed: %s", - strerror(ret)); - return ret; - } - params = (struct ibv_exp_query_intf_params){ - .intf_scope = IBV_EXP_INTF_GLOBAL, - .intf = IBV_EXP_INTF_QP_BURST, - .obj = rxq->qp, + struct ibv_qp *qp; + struct ibv_qp_init_attr attr = { + /* CQ to be associated with the send queue. */ + .send_cq = cq, + /* CQ to be associated with the receive queue. */ + .recv_cq = cq, + .cap = { + /* Max number of outstanding WRs. */ + .max_recv_wr = ((priv->device_attr.max_qp_wr < desc) ? + priv->device_attr.max_qp_wr : + desc), + /* Max number of scatter/gather elements in a WR. */ + .max_recv_sge = 1, + }, + .qp_type = IBV_QPT_RAW_PACKET, }; - rxq->if_qp = ibv_exp_query_intf(priv->ctx, ¶ms, &status); - if (rxq->if_qp == NULL) { - ERROR("QP interface family query failed with status %d", - status); - return errno; - } - return 0; + + qp = ibv_create_qp(priv->pd, &attr); + if (!qp) + rte_errno = errno ? errno : EINVAL; + return qp; } /** @@ -3076,28 +1735,18 @@ rxq_create_qp(struct rxq *rxq, * Number of descriptors to configure in queue. * @param socket * NUMA socket on which memory must be allocated. - * @param inactive - * If true, the queue is disabled because its index is higher or - * equal to the real number of queues, which must be a power of 2. * @param[in] conf * Thresholds parameters. * @param mp * Memory pool for buffer allocations. - * @param children_n - * The number of children in a parent case, zero for a child. - * @param rxq_parent - * The pointer to a parent RX structure (or NULL) in a child case, - * NULL for parent. * * @return - * 0 on success, errno value on failure. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int rxq_setup(struct rte_eth_dev *dev, struct rxq *rxq, uint16_t desc, - unsigned int socket, int inactive, - const struct rte_eth_rxconf *conf, - struct rte_mempool *mp, int children_n, - struct rxq *rxq_parent) + unsigned int socket, const struct rte_eth_rxconf *conf, + struct rte_mempool *mp) { struct priv *priv = dev->data->dev_private; struct rxq tmpl = { @@ -3105,47 +1754,27 @@ rxq_setup(struct rte_eth_dev *dev, struct rxq *rxq, uint16_t desc, .mp = mp, .socket = socket }; - union { - struct ibv_exp_query_intf_params params; - struct ibv_exp_cq_init_attr cq; - struct ibv_exp_res_domain_init_attr rd; - } attr; - enum ibv_exp_query_intf_status status; + struct ibv_qp_attr mod; + struct ibv_recv_wr *bad_wr; unsigned int mb_len; - int ret = 0; - int parent = (children_n > 0); + int ret; (void)conf; /* Thresholds configuration (ignored). */ - /* - * If this is a parent queue, hardware must support RSS and - * RSS must be enabled. - */ - assert((!parent) || ((priv->hw_rss) && (priv->rss))); - if (parent) { - /* Even if unused, ibv_create_cq() requires at least one - * descriptor. */ - desc = 1; - goto skip_mr; - } mb_len = rte_pktmbuf_data_room_size(mp); - if ((desc == 0) || (desc % MLX4_PMD_SGE_WR_N)) { - ERROR("%p: invalid number of RX descriptors (must be a" - " multiple of %d)", (void *)dev, MLX4_PMD_SGE_WR_N); - return EINVAL; + if (desc == 0) { + rte_errno = EINVAL; + ERROR("%p: invalid number of Rx descriptors", (void *)dev); + goto error; } - /* Toggle RX checksum offload if hardware supports it. */ - if (priv->hw_csum) - tmpl.csum = !!dev->data->dev_conf.rxmode.hw_ip_checksum; - if (priv->hw_csum_l2tun) - tmpl.csum_l2tun = !!dev->data->dev_conf.rxmode.hw_ip_checksum; /* Enable scattered packets support for this queue if necessary. */ assert(mb_len >= RTE_PKTMBUF_HEADROOM); if (dev->data->dev_conf.rxmode.max_rx_pkt_len <= (mb_len - RTE_PKTMBUF_HEADROOM)) { - tmpl.sp = 0; + ; } else if (dev->data->dev_conf.rxmode.enable_scatter) { - tmpl.sp = 1; - desc /= MLX4_PMD_SGE_WR_N; + WARN("%p: scattered mode has been requested but is" + " not supported, this may lead to packet loss", + (void *)dev); } else { WARN("%p: the requested maximum Rx packet size (%u) is" " larger than a single mbuf (%u) and scattered" @@ -3154,101 +1783,94 @@ rxq_setup(struct rte_eth_dev *dev, struct rxq *rxq, uint16_t desc, dev->data->dev_conf.rxmode.max_rx_pkt_len, mb_len - RTE_PKTMBUF_HEADROOM); } - DEBUG("%p: %s scattered packets support (%u WRs)", - (void *)dev, (tmpl.sp ? "enabling" : "disabling"), desc); /* Use the entire RX mempool as the memory region. */ tmpl.mr = mlx4_mp2mr(priv->pd, mp); if (tmpl.mr == NULL) { - ret = EINVAL; + rte_errno = EINVAL; ERROR("%p: MR creation failure: %s", - (void *)dev, strerror(ret)); - goto error; - } -skip_mr: - attr.rd = (struct ibv_exp_res_domain_init_attr){ - .comp_mask = (IBV_EXP_RES_DOMAIN_THREAD_MODEL | - IBV_EXP_RES_DOMAIN_MSG_MODEL), - .thread_model = IBV_EXP_THREAD_SINGLE, - .msg_model = IBV_EXP_MSG_HIGH_BW, - }; - tmpl.rd = ibv_exp_create_res_domain(priv->ctx, &attr.rd); - if (tmpl.rd == NULL) { - ret = ENOMEM; - ERROR("%p: RD creation failure: %s", - (void *)dev, strerror(ret)); + (void *)dev, strerror(rte_errno)); goto error; } if (dev->data->dev_conf.intr_conf.rxq) { tmpl.channel = ibv_create_comp_channel(priv->ctx); if (tmpl.channel == NULL) { - ret = ENOMEM; + rte_errno = ENOMEM; ERROR("%p: Rx interrupt completion channel creation" " failure: %s", - (void *)dev, strerror(ret)); + (void *)dev, strerror(rte_errno)); goto error; } } - attr.cq = (struct ibv_exp_cq_init_attr){ - .comp_mask = IBV_EXP_CQ_INIT_ATTR_RES_DOMAIN, - .res_domain = tmpl.rd, - }; - tmpl.cq = ibv_exp_create_cq(priv->ctx, desc, NULL, tmpl.channel, 0, - &attr.cq); + tmpl.cq = ibv_create_cq(priv->ctx, desc, NULL, tmpl.channel, 0); if (tmpl.cq == NULL) { - ret = ENOMEM; + rte_errno = ENOMEM; ERROR("%p: CQ creation failure: %s", - (void *)dev, strerror(ret)); + (void *)dev, strerror(rte_errno)); goto error; } DEBUG("priv->device_attr.max_qp_wr is %d", priv->device_attr.max_qp_wr); DEBUG("priv->device_attr.max_sge is %d", priv->device_attr.max_sge); - /* Allocate descriptors for RX queues, except for the RSS parent. */ - if (parent) - goto skip_alloc; - if (tmpl.sp) - ret = rxq_alloc_elts_sp(&tmpl, desc, NULL); - else - ret = rxq_alloc_elts(&tmpl, desc, NULL); + tmpl.qp = rxq_setup_qp(priv, tmpl.cq, desc); + if (tmpl.qp == NULL) { + ERROR("%p: QP creation failure: %s", + (void *)dev, strerror(rte_errno)); + goto error; + } + mod = (struct ibv_qp_attr){ + /* Move the QP to this state. */ + .qp_state = IBV_QPS_INIT, + /* Primary port number. */ + .port_num = priv->port + }; + ret = ibv_modify_qp(tmpl.qp, &mod, IBV_QP_STATE | IBV_QP_PORT); + if (ret) { + rte_errno = ret; + ERROR("%p: QP state to IBV_QPS_INIT failed: %s", + (void *)dev, strerror(rte_errno)); + goto error; + } + ret = rxq_alloc_elts(&tmpl, desc); if (ret) { ERROR("%p: RXQ allocation failed: %s", - (void *)dev, strerror(ret)); - return ret; + (void *)dev, strerror(rte_errno)); + goto error; } -skip_alloc: - if (parent || rxq_parent || !priv->rss) { - ret = rxq_create_qp(&tmpl, desc, inactive, - children_n, rxq_parent); - if (ret) - goto error; + ret = ibv_post_recv(tmpl.qp, &(*tmpl.elts)[0].wr, &bad_wr); + if (ret) { + rte_errno = ret; + ERROR("%p: ibv_post_recv() failed for WR %p: %s", + (void *)dev, + (void *)bad_wr, + strerror(rte_errno)); + goto error; } - /* Save port ID. */ - tmpl.port_id = dev->data->port_id; - DEBUG("%p: RTE port ID: %u", (void *)rxq, tmpl.port_id); - attr.params = (struct ibv_exp_query_intf_params){ - .intf_scope = IBV_EXP_INTF_GLOBAL, - .intf = IBV_EXP_INTF_CQ, - .obj = tmpl.cq, + mod = (struct ibv_qp_attr){ + .qp_state = IBV_QPS_RTR }; - tmpl.if_cq = ibv_exp_query_intf(priv->ctx, &attr.params, &status); - if (tmpl.if_cq == NULL) { - ret = EINVAL; - ERROR("%p: CQ interface family query failed with status %d", - (void *)dev, status); + ret = ibv_modify_qp(tmpl.qp, &mod, IBV_QP_STATE); + if (ret) { + rte_errno = ret; + ERROR("%p: QP state to IBV_QPS_RTR failed: %s", + (void *)dev, strerror(rte_errno)); goto error; } + /* Save port ID. */ + tmpl.port_id = dev->data->port_id; + DEBUG("%p: RTE port ID: %u", (void *)rxq, tmpl.port_id); /* Clean up rxq in case we're reinitializing it. */ DEBUG("%p: cleaning-up old rxq just in case", (void *)rxq); rxq_cleanup(rxq); *rxq = tmpl; DEBUG("%p: rxq updated with %p", (void *)rxq, (void *)&tmpl); - assert(ret == 0); return 0; error: + ret = rte_errno; rxq_cleanup(&tmpl); - assert(ret > 0); - return ret; + rte_errno = ret; + assert(rte_errno > 0); + return -rte_errno; } /** @@ -3268,34 +1890,31 @@ error: * Memory pool for buffer allocations. * * @return - * 0 on success, negative errno value on failure. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int mlx4_rx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc, unsigned int socket, const struct rte_eth_rxconf *conf, struct rte_mempool *mp) { - struct rxq *parent; struct priv *priv = dev->data->dev_private; struct rxq *rxq = (*priv->rxqs)[idx]; - int inactive = 0; int ret; - priv_lock(priv); DEBUG("%p: configuring queue %u for %u descriptors", (void *)dev, idx, desc); if (idx >= priv->rxqs_n) { + rte_errno = EOVERFLOW; ERROR("%p: queue index out of range (%u >= %u)", (void *)dev, idx, priv->rxqs_n); - priv_unlock(priv); - return -EOVERFLOW; + return -rte_errno; } if (rxq != NULL) { DEBUG("%p: reusing already allocated queue index %u (%p)", (void *)dev, idx, (void *)rxq); if (priv->started) { - priv_unlock(priv); - return -EEXIST; + rte_errno = EEXIST; + return -rte_errno; } (*priv->rxqs)[idx] = NULL; if (idx == 0) @@ -3304,22 +1923,13 @@ mlx4_rx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc, } else { rxq = rte_calloc_socket("RXQ", 1, sizeof(*rxq), 0, socket); if (rxq == NULL) { + rte_errno = ENOMEM; ERROR("%p: unable to allocate queue index %u", (void *)dev, idx); - priv_unlock(priv); - return -ENOMEM; + return -rte_errno; } } - if (priv->rss && !priv->isolated) { - /* The list consists of the single default one. */ - parent = LIST_FIRST(&priv->parents); - if (idx >= rte_align32pow2(priv->rxqs_n + 1) >> 1) - inactive = 1; - } else { - parent = NULL; - } - ret = rxq_setup(dev, rxq, desc, socket, - inactive, conf, mp, 0, parent); + ret = rxq_setup(dev, rxq, desc, socket, conf, mp); if (ret) rte_free(rxq); else { @@ -3328,13 +1938,9 @@ mlx4_rx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc, (void *)dev, (void *)rxq); (*priv->rxqs)[idx] = rxq; /* Update receive callback. */ - if (rxq->sp) - dev->rx_pkt_burst = mlx4_rx_burst_sp; - else - dev->rx_pkt_burst = mlx4_rx_burst; + dev->rx_pkt_burst = mlx4_rx_burst; } - priv_unlock(priv); - return -ret; + return ret; } /** @@ -3353,7 +1959,6 @@ mlx4_rx_queue_release(void *dpdk_rxq) if (rxq == NULL) return; priv = rxq->priv; - priv_lock(priv); for (i = 0; (i != priv->rxqs_n); ++i) if ((*priv->rxqs)[i] == rxq) { DEBUG("%p: removing RX queue %p from list", @@ -3365,7 +1970,6 @@ mlx4_rx_queue_release(void *dpdk_rxq) } rxq_cleanup(rxq); rte_free(rxq); - priv_unlock(priv); } static int @@ -3386,7 +1990,7 @@ priv_dev_link_interrupt_handler_install(struct priv *, struct rte_eth_dev *); * Pointer to Ethernet device structure. * * @return - * 0 on success, negative errno value on failure. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int mlx4_dev_start(struct rte_eth_dev *dev) @@ -3394,11 +1998,8 @@ mlx4_dev_start(struct rte_eth_dev *dev) struct priv *priv = dev->data->dev_private; int ret; - priv_lock(priv); - if (priv->started) { - priv_unlock(priv); + if (priv->started) return 0; - } DEBUG("%p: attaching configured flows to all RX queues", (void *)dev); priv->started = 1; ret = priv_mac_addr_add(priv); @@ -3428,14 +2029,12 @@ mlx4_dev_start(struct rte_eth_dev *dev) (void *)dev, strerror(ret)); goto err; } - priv_unlock(priv); return 0; err: /* Rollback. */ priv_mac_addr_del(priv); priv->started = 0; - priv_unlock(priv); - return -ret; + return ret; } /** @@ -3451,16 +2050,12 @@ mlx4_dev_stop(struct rte_eth_dev *dev) { struct priv *priv = dev->data->dev_private; - priv_lock(priv); - if (!priv->started) { - priv_unlock(priv); + if (!priv->started) return; - } DEBUG("%p: detaching flows from all RX queues", (void *)dev); priv->started = 0; mlx4_priv_flow_stop(priv); priv_mac_addr_del(priv); - priv_unlock(priv); } /** @@ -3540,14 +2135,15 @@ mlx4_dev_close(struct rte_eth_dev *dev) if (priv == NULL) return; - priv_lock(priv); DEBUG("%p: closing device \"%s\"", (void *)dev, ((priv->ctx != NULL) ? priv->ctx->device->name : "")); priv_mac_addr_del(priv); - /* Prevent crashes when queues are still in use. This is unfortunately + /* + * Prevent crashes when queues are still in use. This is unfortunately * still required for DPDK 1.3 because some programs (such as testpmd) - * never release them before closing the device. */ + * never release them before closing the device. + */ dev->rx_pkt_burst = removed_rx_burst; dev->tx_pkt_burst = removed_tx_burst; if (priv->rxqs != NULL) { @@ -3578,8 +2174,6 @@ mlx4_dev_close(struct rte_eth_dev *dev) priv->txqs_n = 0; priv->txqs = NULL; } - if (priv->rss) - priv_parent_list_cleanup(priv); if (priv->pd != NULL) { assert(priv->ctx != NULL); claim_zero(ibv_dealloc_pd(priv->pd)); @@ -3589,7 +2183,6 @@ mlx4_dev_close(struct rte_eth_dev *dev) priv_dev_removal_interrupt_handler_uninstall(priv, dev); priv_dev_link_interrupt_handler_uninstall(priv, dev); priv_rx_intr_vec_disable(priv); - priv_unlock(priv); memset(priv, 0, sizeof(*priv)); } @@ -3602,30 +2195,19 @@ mlx4_dev_close(struct rte_eth_dev *dev) * Nonzero for link up, otherwise link down. * * @return - * 0 on success, errno value on failure. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int priv_set_link(struct priv *priv, int up) { struct rte_eth_dev *dev = priv->dev; int err; - unsigned int i; if (up) { err = priv_set_flags(priv, ~IFF_UP, IFF_UP); if (err) return err; - for (i = 0; i < priv->rxqs_n; i++) - if ((*priv->rxqs)[i]->sp) - break; - /* Check if an sp queue exists. - * Note: Some old frames might be received. - */ - if (i == priv->rxqs_n) - dev->rx_pkt_burst = mlx4_rx_burst; - else - dev->rx_pkt_burst = mlx4_rx_burst_sp; - dev->tx_pkt_burst = mlx4_tx_burst; + dev->rx_pkt_burst = mlx4_rx_burst; } else { err = priv_set_flags(priv, ~IFF_UP, ~IFF_UP); if (err) @@ -3643,18 +2225,14 @@ priv_set_link(struct priv *priv, int up) * Pointer to Ethernet device structure. * * @return - * 0 on success, errno value on failure. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int mlx4_set_link_down(struct rte_eth_dev *dev) { struct priv *priv = dev->data->dev_private; - int err; - priv_lock(priv); - err = priv_set_link(priv, 0); - priv_unlock(priv); - return err; + return priv_set_link(priv, 0); } /** @@ -3664,19 +2242,16 @@ mlx4_set_link_down(struct rte_eth_dev *dev) * Pointer to Ethernet device structure. * * @return - * 0 on success, errno value on failure. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int mlx4_set_link_up(struct rte_eth_dev *dev) { struct priv *priv = dev->data->dev_private; - int err; - priv_lock(priv); - err = priv_set_link(priv, 1); - priv_unlock(priv); - return err; + return priv_set_link(priv, 1); } + /** * DPDK callback to get information about the device. * @@ -3693,10 +2268,8 @@ mlx4_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *info) char ifname[IF_NAMESIZE]; info->pci_dev = RTE_ETH_DEV_TO_PCI(dev); - if (priv == NULL) return; - priv_lock(priv); /* FIXME: we should ask the device for these values. */ info->min_rx_bufsize = 32; info->max_rx_pktlen = 65536; @@ -3713,18 +2286,8 @@ mlx4_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *info) info->max_tx_queues = max; /* Last array entry is reserved for broadcast. */ info->max_mac_addrs = 1; - info->rx_offload_capa = - (priv->hw_csum ? - (DEV_RX_OFFLOAD_IPV4_CKSUM | - DEV_RX_OFFLOAD_UDP_CKSUM | - DEV_RX_OFFLOAD_TCP_CKSUM) : - 0); - info->tx_offload_capa = - (priv->hw_csum ? - (DEV_TX_OFFLOAD_IPV4_CKSUM | - DEV_TX_OFFLOAD_UDP_CKSUM | - DEV_TX_OFFLOAD_TCP_CKSUM) : - 0); + info->rx_offload_capa = 0; + info->tx_offload_capa = 0; if (priv_get_ifname(priv, &ifname) == 0) info->if_index = if_nametoindex(ifname); info->speed_capa = @@ -3733,25 +2296,6 @@ mlx4_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *info) ETH_LINK_SPEED_20G | ETH_LINK_SPEED_40G | ETH_LINK_SPEED_56G; - priv_unlock(priv); -} - -static const uint32_t * -mlx4_dev_supported_ptypes_get(struct rte_eth_dev *dev) -{ - static const uint32_t ptypes[] = { - /* refers to rxq_cq_to_pkt_type() */ - RTE_PTYPE_L3_IPV4, - RTE_PTYPE_L3_IPV6, - RTE_PTYPE_INNER_L3_IPV4, - RTE_PTYPE_INNER_L3_IPV6, - RTE_PTYPE_UNKNOWN - }; - - if (dev->rx_pkt_burst == mlx4_rx_burst || - dev->rx_pkt_burst == mlx4_rx_burst_sp) - return ptypes; - return NULL; } /** @@ -3772,7 +2316,6 @@ mlx4_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats) if (priv == NULL) return; - priv_lock(priv); /* Add software counters. */ for (i = 0; (i != priv->rxqs_n); ++i) { struct rxq *rxq = (*priv->rxqs)[i]; @@ -3807,7 +2350,6 @@ mlx4_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats) tmp.oerrors += txq->stats.odropped; } *stats = tmp; - priv_unlock(priv); } /** @@ -3825,7 +2367,6 @@ mlx4_stats_reset(struct rte_eth_dev *dev) if (priv == NULL) return; - priv_lock(priv); for (i = 0; (i != priv->rxqs_n); ++i) { if ((*priv->rxqs)[i] == NULL) continue; @@ -3840,7 +2381,6 @@ mlx4_stats_reset(struct rte_eth_dev *dev) (*priv->txqs)[i]->stats = (struct mlx4_txq_stats){ .idx = idx }; } - priv_unlock(priv); } /** @@ -3850,6 +2390,9 @@ mlx4_stats_reset(struct rte_eth_dev *dev) * Pointer to Ethernet device structure. * @param wait_to_complete * Wait for request completion (ignored). + * + * @return + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int mlx4_link_update(struct rte_eth_dev *dev, int wait_to_complete) @@ -3862,14 +2405,14 @@ mlx4_link_update(struct rte_eth_dev *dev, int wait_to_complete) struct rte_eth_link dev_link; int link_speed = 0; - /* priv_lock() is not taken to allow concurrent calls. */ - - if (priv == NULL) - return -EINVAL; + if (priv == NULL) { + rte_errno = EINVAL; + return -rte_errno; + } (void)wait_to_complete; if (priv_ifreq(priv, SIOCGIFFLAGS, &ifr)) { - WARN("ioctl(SIOCGIFFLAGS) failed: %s", strerror(errno)); - return -1; + WARN("ioctl(SIOCGIFFLAGS) failed: %s", strerror(rte_errno)); + return -rte_errno; } memset(&dev_link, 0, sizeof(dev_link)); dev_link.link_status = ((ifr.ifr_flags & IFF_UP) && @@ -3877,8 +2420,8 @@ mlx4_link_update(struct rte_eth_dev *dev, int wait_to_complete) ifr.ifr_data = (void *)&edata; if (priv_ifreq(priv, SIOCETHTOOL, &ifr)) { WARN("ioctl(SIOCETHTOOL, ETHTOOL_GSET) failed: %s", - strerror(errno)); - return -1; + strerror(rte_errno)); + return -rte_errno; } link_speed = ethtool_cmd_speed(&edata); if (link_speed == -1) @@ -3889,94 +2432,8 @@ mlx4_link_update(struct rte_eth_dev *dev, int wait_to_complete) ETH_LINK_HALF_DUPLEX : ETH_LINK_FULL_DUPLEX); dev_link.link_autoneg = !(dev->data->dev_conf.link_speeds & ETH_LINK_SPEED_FIXED); - if (memcmp(&dev_link, &dev->data->dev_link, sizeof(dev_link))) { - /* Link status changed. */ - dev->data->dev_link = dev_link; - return 0; - } - /* Link status is still the same. */ - return -1; -} - -/** - * DPDK callback to change the MTU. - * - * Setting the MTU affects hardware MRU (packets larger than the MTU cannot be - * received). Use this as a hint to enable/disable scattered packets support - * and improve performance when not needed. - * Since failure is not an option, reconfiguring queues on the fly is not - * recommended. - * - * @param dev - * Pointer to Ethernet device structure. - * @param in_mtu - * New MTU. - * - * @return - * 0 on success, negative errno value on failure. - */ -static int -mlx4_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu) -{ - struct priv *priv = dev->data->dev_private; - int ret = 0; - unsigned int i; - uint16_t (*rx_func)(void *, struct rte_mbuf **, uint16_t) = - mlx4_rx_burst; - - priv_lock(priv); - /* Set kernel interface MTU first. */ - if (priv_set_mtu(priv, mtu)) { - ret = errno; - WARN("cannot set port %u MTU to %u: %s", priv->port, mtu, - strerror(ret)); - goto out; - } else - DEBUG("adapter port %u MTU set to %u", priv->port, mtu); - priv->mtu = mtu; - /* Remove MAC flow. */ - priv_mac_addr_del(priv); - /* Temporarily replace RX handler with a fake one, assuming it has not - * been copied elsewhere. */ - dev->rx_pkt_burst = removed_rx_burst; - /* Make sure everyone has left mlx4_rx_burst() and uses - * removed_rx_burst() instead. */ - rte_wmb(); - usleep(1000); - /* Reconfigure each RX queue. */ - for (i = 0; (i != priv->rxqs_n); ++i) { - struct rxq *rxq = (*priv->rxqs)[i]; - unsigned int max_frame_len; - - if (rxq == NULL) - continue; - /* Calculate new maximum frame length according to MTU. */ - max_frame_len = (priv->mtu + ETHER_HDR_LEN + - (ETHER_MAX_VLAN_FRAME_LEN - ETHER_MAX_LEN)); - /* Provide new values to rxq_setup(). */ - dev->data->dev_conf.rxmode.jumbo_frame = - (max_frame_len > ETHER_MAX_LEN); - dev->data->dev_conf.rxmode.max_rx_pkt_len = max_frame_len; - ret = rxq_rehash(dev, rxq); - if (ret) { - /* Force SP RX if that queue requires it and abort. */ - if (rxq->sp) - rx_func = mlx4_rx_burst_sp; - break; - } - /* Scattered burst function takes priority. */ - if (rxq->sp) - rx_func = mlx4_rx_burst_sp; - } - /* Burst functions can now be called again. */ - rte_wmb(); - dev->rx_pkt_burst = rx_func; - /* Restore MAC flow. */ - ret = priv_mac_addr_add(priv); -out: - priv_unlock(priv); - assert(ret >= 0); - return -ret; + dev->data->dev_link = dev_link; + return 0; } /** @@ -3988,7 +2445,7 @@ out: * Flow control output buffer. * * @return - * 0 on success, negative errno value on failure. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int mlx4_dev_get_flow_ctrl(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf) @@ -4001,15 +2458,13 @@ mlx4_dev_get_flow_ctrl(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf) int ret; ifr.ifr_data = (void *)ðpause; - priv_lock(priv); if (priv_ifreq(priv, SIOCETHTOOL, &ifr)) { - ret = errno; + ret = rte_errno; WARN("ioctl(SIOCETHTOOL, ETHTOOL_GPAUSEPARAM)" " failed: %s", - strerror(ret)); + strerror(rte_errno)); goto out; } - fc_conf->autoneg = ethpause.autoneg; if (ethpause.rx_pause && ethpause.tx_pause) fc_conf->mode = RTE_FC_FULL; @@ -4020,9 +2475,7 @@ mlx4_dev_get_flow_ctrl(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf) else fc_conf->mode = RTE_FC_NONE; ret = 0; - out: - priv_unlock(priv); assert(ret >= 0); return -ret; } @@ -4036,7 +2489,7 @@ out: * Flow control parameters. * * @return - * 0 on success, negative errno value on failure. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int mlx4_dev_set_flow_ctrl(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf) @@ -4055,25 +2508,20 @@ mlx4_dev_set_flow_ctrl(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf) ethpause.rx_pause = 1; else ethpause.rx_pause = 0; - if (((fc_conf->mode & RTE_FC_FULL) == RTE_FC_FULL) || (fc_conf->mode & RTE_FC_TX_PAUSE)) ethpause.tx_pause = 1; else ethpause.tx_pause = 0; - - priv_lock(priv); if (priv_ifreq(priv, SIOCETHTOOL, &ifr)) { - ret = errno; + ret = rte_errno; WARN("ioctl(SIOCETHTOOL, ETHTOOL_SPAUSEPARAM)" " failed: %s", - strerror(ret)); + strerror(rte_errno)); goto out; } ret = 0; - out: - priv_unlock(priv); assert(ret >= 0); return -ret; } @@ -4100,7 +2548,7 @@ const struct rte_flow_ops mlx4_flow_ops = { * Pointer to operation-specific structure. * * @return - * 0 on success, negative errno value on failure. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int mlx4_dev_filter_ctrl(struct rte_eth_dev *dev, @@ -4108,12 +2556,10 @@ mlx4_dev_filter_ctrl(struct rte_eth_dev *dev, enum rte_filter_op filter_op, void *arg) { - int ret = EINVAL; - switch (filter_type) { case RTE_ETH_FILTER_GENERIC: if (filter_op != RTE_ETH_FILTER_GET) - return -EINVAL; + break; *(const void **)arg = &mlx4_flow_ops; return 0; default: @@ -4121,7 +2567,8 @@ mlx4_dev_filter_ctrl(struct rte_eth_dev *dev, (void *)dev, filter_type); break; } - return -ret; + rte_errno = ENOTSUP; + return -rte_errno; } static const struct eth_dev_ops mlx4_dev_ops = { @@ -4135,7 +2582,6 @@ static const struct eth_dev_ops mlx4_dev_ops = { .stats_get = mlx4_stats_get, .stats_reset = mlx4_stats_reset, .dev_infos_get = mlx4_dev_infos_get, - .dev_supported_ptypes_get = mlx4_dev_supported_ptypes_get, .rx_queue_setup = mlx4_rx_queue_setup, .tx_queue_setup = mlx4_tx_queue_setup, .rx_queue_release = mlx4_rx_queue_release, @@ -4157,7 +2603,7 @@ static const struct eth_dev_ops mlx4_dev_ops = { * PCI bus address output buffer. * * @return - * 0 on success, -1 on failure and errno is set. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int mlx4_ibv_device_to_pci_addr(const struct ibv_device *device, @@ -4168,8 +2614,10 @@ mlx4_ibv_device_to_pci_addr(const struct ibv_device *device, MKSTR(path, "%s/device/uevent", device->ibdev_path); file = fopen(path, "rb"); - if (file == NULL) - return -1; + if (file == NULL) { + rte_errno = errno; + return -rte_errno; + } while (fgets(line, sizeof(line), file) == line) { size_t len = strlen(line); int ret; @@ -4207,38 +2655,20 @@ mlx4_ibv_device_to_pci_addr(const struct ibv_device *device, * MAC address output buffer. * * @return - * 0 on success, -1 on failure and errno is set. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int priv_get_mac(struct priv *priv, uint8_t (*mac)[ETHER_ADDR_LEN]) { struct ifreq request; + int ret = priv_ifreq(priv, SIOCGIFHWADDR, &request); - if (priv_ifreq(priv, SIOCGIFHWADDR, &request)) - return -1; + if (ret) + return ret; memcpy(mac, request.ifr_hwaddr.sa_data, ETHER_ADDR_LEN); return 0; } -/** - * Retrieve integer value from environment variable. - * - * @param[in] name - * Environment variable name. - * - * @return - * Integer value, 0 if the variable is not set. - */ -static int -mlx4_getenv_int(const char *name) -{ - const char *val = getenv(name); - - if (val == NULL) - return 0; - return atoi(val); -} - static void mlx4_dev_link_status_handler(void *); static void @@ -4317,14 +2747,12 @@ mlx4_dev_link_status_handler(void *arg) uint32_t events; int ret; - priv_lock(priv); assert(priv->pending_alarm == 1); priv->pending_alarm = 0; ret = priv_dev_status_handler(priv, dev, &events); - priv_unlock(priv); if (ret > 0 && events & (1 << RTE_ETH_EVENT_INTR_LSC)) - _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL, - NULL); + _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, + NULL, NULL); } /** @@ -4344,9 +2772,7 @@ mlx4_dev_interrupt_handler(void *cb_arg) uint32_t ev; int i; - priv_lock(priv); ret = priv_dev_status_handler(priv, dev, &ev); - priv_unlock(priv); if (ret > 0) { for (i = RTE_ETH_EVENT_UNKNOWN; i < RTE_ETH_EVENT_MAX; @@ -4371,8 +2797,9 @@ mlx4_dev_interrupt_handler(void *cb_arg) * Pointer to private structure. * @param dev * Pointer to the rte_eth_dev structure. + * * @return - * 0 on success, negative errno value on failure. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int priv_dev_interrupt_handler_uninstall(struct priv *priv, struct rte_eth_dev *dev) @@ -4386,14 +2813,11 @@ priv_dev_interrupt_handler_uninstall(struct priv *priv, struct rte_eth_dev *dev) mlx4_dev_interrupt_handler, dev); if (ret < 0) { - ERROR("rte_intr_callback_unregister failed with %d" - "%s%s%s", ret, - (errno ? " (errno: " : ""), - (errno ? strerror(errno) : ""), - (errno ? ")" : "")); + rte_errno = ret; + ERROR("rte_intr_callback_unregister failed with %d %s", + ret, strerror(rte_errno)); } - priv->intr_handle.fd = 0; - priv->intr_handle.type = RTE_INTR_HANDLE_UNKNOWN; + priv->intr_handle.fd = -1; return ret; } @@ -4404,8 +2828,9 @@ priv_dev_interrupt_handler_uninstall(struct priv *priv, struct rte_eth_dev *dev) * Pointer to private structure. * @param dev * Pointer to the rte_eth_dev structure. + * * @return - * 0 on success, negative errno value on failure. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int priv_dev_interrupt_handler_install(struct priv *priv, @@ -4414,8 +2839,9 @@ priv_dev_interrupt_handler_install(struct priv *priv, int flags; int rc; - /* Check whether the interrupt handler has already been installed - * for either type of interrupt + /* + * Check whether the interrupt handler has already been installed + * for either type of interrupt. */ if (priv->intr_conf.lsc && priv->intr_conf.rmv && @@ -4425,20 +2851,22 @@ priv_dev_interrupt_handler_install(struct priv *priv, flags = fcntl(priv->ctx->async_fd, F_GETFL); rc = fcntl(priv->ctx->async_fd, F_SETFL, flags | O_NONBLOCK); if (rc < 0) { + rte_errno = errno ? errno : EINVAL; INFO("failed to change file descriptor async event queue"); dev->data->dev_conf.intr_conf.lsc = 0; dev->data->dev_conf.intr_conf.rmv = 0; - return -errno; + return -rte_errno; } else { priv->intr_handle.fd = priv->ctx->async_fd; - priv->intr_handle.type = RTE_INTR_HANDLE_EXT; rc = rte_intr_callback_register(&priv->intr_handle, mlx4_dev_interrupt_handler, dev); if (rc) { + rte_errno = -rc; ERROR("rte_intr_callback_register failed " - " (errno: %s)", strerror(errno)); - return rc; + " (rte_errno: %s)", strerror(rte_errno)); + priv->intr_handle.fd = -1; + return -rte_errno; } } return 0; @@ -4451,8 +2879,9 @@ priv_dev_interrupt_handler_install(struct priv *priv, * Pointer to private structure. * @param dev * Pointer to the rte_eth_dev structure. + * * @return - * 0 on success, negative value on error. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int priv_dev_removal_interrupt_handler_uninstall(struct priv *priv, @@ -4472,8 +2901,9 @@ priv_dev_removal_interrupt_handler_uninstall(struct priv *priv, * Pointer to private structure. * @param dev * Pointer to the rte_eth_dev structure. + * * @return - * 0 on success, negative value on error, + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int priv_dev_link_interrupt_handler_uninstall(struct priv *priv, @@ -4491,7 +2921,7 @@ priv_dev_link_interrupt_handler_uninstall(struct priv *priv, if (rte_eal_alarm_cancel(mlx4_dev_link_status_handler, dev)) { ERROR("rte_eal_alarm_cancel failed " - " (errno: %s)", strerror(rte_errno)); + " (rte_errno: %s)", strerror(rte_errno)); return -rte_errno; } priv->pending_alarm = 0; @@ -4505,8 +2935,9 @@ priv_dev_link_interrupt_handler_uninstall(struct priv *priv, * Pointer to private structure. * @param dev * Pointer to the rte_eth_dev structure. + * * @return - * 0 on success, negative value on error. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int priv_dev_link_interrupt_handler_install(struct priv *priv, @@ -4530,8 +2961,9 @@ priv_dev_link_interrupt_handler_install(struct priv *priv, * Pointer to private structure. * @param dev * Pointer to the rte_eth_dev structure. + * * @return - * 0 on success, negative value on error. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int priv_dev_removal_interrupt_handler_install(struct priv *priv, @@ -4555,7 +2987,7 @@ priv_dev_removal_interrupt_handler_install(struct priv *priv, * Pointer to private structure. * * @return - * 0 on success, negative on failure. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int priv_rx_intr_vec_enable(struct priv *priv) @@ -4564,18 +2996,18 @@ priv_rx_intr_vec_enable(struct priv *priv) unsigned int rxqs_n = priv->rxqs_n; unsigned int n = RTE_MIN(rxqs_n, (uint32_t)RTE_MAX_RXTX_INTR_VEC_ID); unsigned int count = 0; - struct rte_intr_handle *intr_handle = priv->dev->intr_handle; + struct rte_intr_handle *intr_handle = &priv->intr_handle; if (!priv->dev->data->dev_conf.intr_conf.rxq) return 0; priv_rx_intr_vec_disable(priv); intr_handle->intr_vec = malloc(sizeof(intr_handle->intr_vec[rxqs_n])); if (intr_handle->intr_vec == NULL) { + rte_errno = ENOMEM; ERROR("failed to allocate memory for interrupt vector," " Rx interrupts will not be supported"); - return -ENOMEM; + return -rte_errno; } - intr_handle->type = RTE_INTR_HANDLE_EXT; for (i = 0; i != n; ++i) { struct rxq *rxq = (*priv->rxqs)[i]; int fd; @@ -4591,20 +3023,22 @@ priv_rx_intr_vec_enable(struct priv *priv) continue; } if (count >= RTE_MAX_RXTX_INTR_VEC_ID) { + rte_errno = E2BIG; ERROR("too many Rx queues for interrupt vector size" " (%d), Rx interrupts cannot be enabled", RTE_MAX_RXTX_INTR_VEC_ID); priv_rx_intr_vec_disable(priv); - return -1; + return -rte_errno; } fd = rxq->channel->fd; flags = fcntl(fd, F_GETFL); rc = fcntl(fd, F_SETFL, flags | O_NONBLOCK); if (rc < 0) { + rte_errno = errno; ERROR("failed to make Rx interrupt file descriptor" " %d non-blocking for queue index %d", fd, i); priv_rx_intr_vec_disable(priv); - return rc; + return -rte_errno; } intr_handle->intr_vec[i] = RTE_INTR_VEC_RXTX_OFFSET + count; intr_handle->efds[count] = fd; @@ -4626,7 +3060,7 @@ priv_rx_intr_vec_enable(struct priv *priv) static void priv_rx_intr_vec_disable(struct priv *priv) { - struct rte_intr_handle *intr_handle = priv->dev->intr_handle; + struct rte_intr_handle *intr_handle = &priv->intr_handle; rte_intr_free_epoll_fd(intr_handle); free(intr_handle->intr_vec); @@ -4643,7 +3077,7 @@ priv_rx_intr_vec_disable(struct priv *priv) * Rx queue index. * * @return - * 0 on success, negative on failure. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int mlx4_rx_intr_enable(struct rte_eth_dev *dev, uint16_t idx) @@ -4656,8 +3090,10 @@ mlx4_rx_intr_enable(struct rte_eth_dev *dev, uint16_t idx) ret = EINVAL; else ret = ibv_req_notify_cq(rxq->cq, 0); - if (ret) + if (ret) { + rte_errno = ret; WARN("unable to arm interrupt on rx queue %d", idx); + } return -ret; } @@ -4670,7 +3106,7 @@ mlx4_rx_intr_enable(struct rte_eth_dev *dev, uint16_t idx) * Rx queue index. * * @return - * 0 on success, negative on failure. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int mlx4_rx_intr_disable(struct rte_eth_dev *dev, uint16_t idx) @@ -4688,11 +3124,13 @@ mlx4_rx_intr_disable(struct rte_eth_dev *dev, uint16_t idx) if (ret || ev_cq != rxq->cq) ret = EINVAL; } - if (ret) + if (ret) { + rte_errno = ret; WARN("unable to disable interrupt on rx queue %d", idx); - else + } else { ibv_ack_cq_events(rxq->cq, 1); + } return -ret; } @@ -4707,7 +3145,7 @@ mlx4_rx_intr_disable(struct rte_eth_dev *dev, uint16_t idx) * Shared configuration data. * * @return - * 0 on success, negative errno value on failure. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int mlx4_arg_parse(const char *key, const char *val, struct mlx4_conf *conf) @@ -4717,8 +3155,9 @@ mlx4_arg_parse(const char *key, const char *val, struct mlx4_conf *conf) errno = 0; tmp = strtoul(val, NULL, 0); if (errno) { + rte_errno = errno; WARN("%s: \"%s\" is not a valid integer", key, val); - return -errno; + return -rte_errno; } if (strcmp(MLX4_PMD_PORT_KVARG, key) == 0) { uint32_t ports = rte_log2_u32(conf->ports.present); @@ -4729,13 +3168,15 @@ mlx4_arg_parse(const char *key, const char *val, struct mlx4_conf *conf) return -EINVAL; } if (!(conf->ports.present & (1 << tmp))) { + rte_errno = EINVAL; ERROR("invalid port index %lu", tmp); - return -EINVAL; + return -rte_errno; } conf->ports.enabled |= 1 << tmp; } else { + rte_errno = EINVAL; WARN("%s: unknown parameter", key); - return -EINVAL; + return -rte_errno; } return 0; } @@ -4747,7 +3188,7 @@ mlx4_arg_parse(const char *key, const char *val, struct mlx4_conf *conf) * Device arguments structure. * * @return - * 0 on success, negative errno value on failure. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int mlx4_args(struct rte_devargs *devargs, struct mlx4_conf *conf) @@ -4761,8 +3202,9 @@ mlx4_args(struct rte_devargs *devargs, struct mlx4_conf *conf) return 0; kvlist = rte_kvargs_parse(devargs->args, pmd_mlx4_init_params); if (kvlist == NULL) { + rte_errno = EINVAL; ERROR("failed to parse kvargs"); - return -EINVAL; + return -rte_errno; } /* Process parameters. */ for (i = 0; pmd_mlx4_init_params[i]; ++i) { @@ -4798,7 +3240,7 @@ static struct rte_pci_driver mlx4_driver; * PCI device information. * * @return - * 0 on success, negative errno value on failure. + * 0 on success, negative errno value otherwise and rte_errno is set. */ static int mlx4_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev) @@ -4816,13 +3258,13 @@ mlx4_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev) (void)pci_drv; assert(pci_drv == &mlx4_driver); - list = ibv_get_device_list(&i); if (list == NULL) { - assert(errno); - if (errno == ENOSYS) + rte_errno = errno; + assert(rte_errno); + if (rte_errno == ENOSYS) ERROR("cannot list devices, is ib_uverbs loaded?"); - return -errno; + return -rte_errno; } assert(i >= 0); /* @@ -4853,28 +3295,29 @@ mlx4_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev) ibv_free_device_list(list); switch (err) { case 0: + rte_errno = ENODEV; ERROR("cannot access device, is mlx4_ib loaded?"); - return -ENODEV; + return -rte_errno; case EINVAL: + rte_errno = EINVAL; ERROR("cannot use device, are drivers up to date?"); - return -EINVAL; + return -rte_errno; } assert(err > 0); - return -err; + rte_errno = err; + return -rte_errno; } ibv_dev = list[i]; - DEBUG("device opened"); if (ibv_query_device(attr_ctx, &device_attr)) { - err = ENODEV; + rte_errno = ENODEV; goto error; } INFO("%u port(s) detected", device_attr.phys_port_cnt); - conf.ports.present |= (UINT64_C(1) << device_attr.phys_port_cnt) - 1; if (mlx4_args(pci_dev->device.devargs, &conf)) { ERROR("failed to process device arguments"); - err = EINVAL; + rte_errno = EINVAL; goto error; } /* Use all ports when none are defined */ @@ -4887,139 +3330,60 @@ mlx4_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev) struct ibv_pd *pd = NULL; struct priv *priv = NULL; struct rte_eth_dev *eth_dev = NULL; - struct ibv_exp_device_attr exp_device_attr; struct ether_addr mac; /* If port is not enabled, skip. */ if (!(conf.ports.enabled & (1 << i))) continue; - exp_device_attr.comp_mask = IBV_EXP_DEVICE_ATTR_EXP_CAP_FLAGS; - exp_device_attr.comp_mask |= IBV_EXP_DEVICE_ATTR_RSS_TBL_SZ; - DEBUG("using port %u", port); - ctx = ibv_open_device(ibv_dev); if (ctx == NULL) { - err = ENODEV; + rte_errno = ENODEV; goto port_error; } - /* Check port status. */ err = ibv_query_port(ctx, port, &port_attr); if (err) { - ERROR("port query failed: %s", strerror(err)); - err = ENODEV; + rte_errno = err; + ERROR("port query failed: %s", strerror(rte_errno)); goto port_error; } - if (port_attr.link_layer != IBV_LINK_LAYER_ETHERNET) { + rte_errno = ENOTSUP; ERROR("port %d is not configured in Ethernet mode", port); - err = EINVAL; goto port_error; } - if (port_attr.state != IBV_PORT_ACTIVE) DEBUG("port %d is not active: \"%s\" (%d)", port, ibv_port_state_str(port_attr.state), port_attr.state); - /* Allocate protection domain. */ pd = ibv_alloc_pd(ctx); if (pd == NULL) { + rte_errno = ENOMEM; ERROR("PD allocation failure"); - err = ENOMEM; goto port_error; } - /* from rte_ethdev.c */ priv = rte_zmalloc("ethdev private structure", sizeof(*priv), RTE_CACHE_LINE_SIZE); if (priv == NULL) { + rte_errno = ENOMEM; ERROR("priv allocation failure"); - err = ENOMEM; goto port_error; } - priv->ctx = ctx; priv->device_attr = device_attr; priv->port = port; priv->pd = pd; priv->mtu = ETHER_MTU; - if (ibv_exp_query_device(ctx, &exp_device_attr)) { - ERROR("ibv_exp_query_device() failed"); - err = ENODEV; - goto port_error; - } - if ((exp_device_attr.exp_device_cap_flags & - IBV_EXP_DEVICE_QPG) && - (exp_device_attr.exp_device_cap_flags & - IBV_EXP_DEVICE_UD_RSS) && - (exp_device_attr.comp_mask & - IBV_EXP_DEVICE_ATTR_RSS_TBL_SZ) && - (exp_device_attr.max_rss_tbl_sz > 0)) { - priv->hw_qpg = 1; - priv->hw_rss = 1; - priv->max_rss_tbl_sz = exp_device_attr.max_rss_tbl_sz; - } else { - priv->hw_qpg = 0; - priv->hw_rss = 0; - priv->max_rss_tbl_sz = 0; - } - priv->hw_tss = !!(exp_device_attr.exp_device_cap_flags & - IBV_EXP_DEVICE_UD_TSS); - DEBUG("device flags: %s%s%s", - (priv->hw_qpg ? "IBV_DEVICE_QPG " : ""), - (priv->hw_tss ? "IBV_DEVICE_TSS " : ""), - (priv->hw_rss ? "IBV_DEVICE_RSS " : "")); - if (priv->hw_rss) - DEBUG("maximum RSS indirection table size: %u", - exp_device_attr.max_rss_tbl_sz); - - priv->hw_csum = - ((exp_device_attr.exp_device_cap_flags & - IBV_EXP_DEVICE_RX_CSUM_TCP_UDP_PKT) && - (exp_device_attr.exp_device_cap_flags & - IBV_EXP_DEVICE_RX_CSUM_IP_PKT)); - DEBUG("checksum offloading is %ssupported", - (priv->hw_csum ? "" : "not ")); - - priv->hw_csum_l2tun = !!(exp_device_attr.exp_device_cap_flags & - IBV_EXP_DEVICE_VXLAN_SUPPORT); - DEBUG("L2 tunnel checksum offloads are %ssupported", - (priv->hw_csum_l2tun ? "" : "not ")); - - priv->inl_recv_size = mlx4_getenv_int("MLX4_INLINE_RECV_SIZE"); - - if (priv->inl_recv_size) { - exp_device_attr.comp_mask = - IBV_EXP_DEVICE_ATTR_INLINE_RECV_SZ; - if (ibv_exp_query_device(ctx, &exp_device_attr)) { - INFO("Couldn't query device for inline-receive" - " capabilities."); - priv->inl_recv_size = 0; - } else { - if ((unsigned)exp_device_attr.inline_recv_sz < - priv->inl_recv_size) { - INFO("Max inline-receive (%d) <" - " requested inline-receive (%u)", - exp_device_attr.inline_recv_sz, - priv->inl_recv_size); - priv->inl_recv_size = - exp_device_attr.inline_recv_sz; - } - } - INFO("Set inline receive size to %u", - priv->inl_recv_size); - } - priv->vf = vf; /* Configure the first MAC address by default. */ if (priv_get_mac(priv, &mac.addr_bytes)) { ERROR("cannot get MAC address, is mlx4_en loaded?" - " (errno: %s)", strerror(errno)); - err = ENODEV; + " (rte_errno: %s)", strerror(rte_errno)); goto port_error; } INFO("port %u MAC address is %02x:%02x:%02x:%02x:%02x:%02x", @@ -5045,7 +3409,6 @@ mlx4_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev) /* Get actual MTU if possible. */ priv_get_mtu(priv, &priv->mtu); DEBUG("port %u MTU is %u", priv->port, priv->mtu); - /* from rte_ethdev.c */ { char name[RTE_ETH_NAME_MAX_LEN]; @@ -5056,31 +3419,35 @@ mlx4_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev) } if (eth_dev == NULL) { ERROR("can not allocate rte ethdev"); - err = ENOMEM; + rte_errno = ENOMEM; goto port_error; } - eth_dev->data->dev_private = priv; eth_dev->data->mac_addrs = &priv->mac; eth_dev->device = &pci_dev->device; - rte_eth_copy_pci_info(eth_dev, pci_dev); - eth_dev->device->driver = &mlx4_driver.driver; - + /* Initialize local interrupt handle for current port. */ + priv->intr_handle = (struct rte_intr_handle){ + .fd = -1, + .type = RTE_INTR_HANDLE_EXT, + }; /* - * Copy and override interrupt handle to prevent it from - * being shared between all ethdev instances of a given PCI - * device. This is required to properly handle Rx interrupts - * on all ports. + * Override ethdev interrupt handle pointer with private + * handle instead of that of the parent PCI device used by + * default. This prevents it from being shared between all + * ports of the same PCI device since each of them is + * associated its own Verbs context. + * + * Rx interrupts in particular require this as the PMD has + * no control over the registration of queue interrupts + * besides setting up eth_dev->intr_handle, the rest is + * handled by rte_intr_rx_ctl(). */ - priv->intr_handle_dev = *eth_dev->intr_handle; - eth_dev->intr_handle = &priv->intr_handle_dev; - + eth_dev->intr_handle = &priv->intr_handle; priv->dev = eth_dev; eth_dev->dev_ops = &mlx4_dev_ops; eth_dev->data->dev_flags |= RTE_ETH_DEV_DETACHABLE; - /* Bring Ethernet device up. */ DEBUG("forcing Ethernet interface up"); priv_set_flags(priv, ~IFF_UP, IFF_UP); @@ -5088,7 +3455,6 @@ mlx4_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev) if (eth_dev->data->dev_flags & RTE_ETH_DEV_INTR_LSC) mlx4_link_update(eth_dev, 0); continue; - port_error: rte_free(priv); if (pd) @@ -5101,21 +3467,19 @@ port_error: } if (i == device_attr.phys_port_cnt) return 0; - /* * XXX if something went wrong in the loop above, there is a resource * leak (ctx, pd, priv, dpdk ethdev) but we can do nothing about it as * long as the dpdk does not provide a way to deallocate a ethdev and a * way to enumerate the registered ethdevs to free the previous ones. */ - error: if (attr_ctx) claim_zero(ibv_close_device(attr_ctx)); if (list) ibv_free_device_list(list); - assert(err >= 0); - return -err; + assert(rte_errno >= 0); + return -rte_errno; } static const struct rte_pci_id mlx4_pci_id_map[] = { @@ -5153,7 +3517,6 @@ RTE_INIT(rte_mlx4_pmd_init); static void rte_mlx4_pmd_init(void) { - RTE_BUILD_BUG_ON(sizeof(wr_id_t) != sizeof(uint64_t)); /* * RDMAV_HUGEPAGES_SAFE tells ibv_fork_init() we intend to use * huge pages. Calling ibv_fork_init() during init allows