diff options
Diffstat (limited to 'drivers/infiniband')
52 files changed, 1273 insertions, 515 deletions
diff --git a/drivers/infiniband/core/addr.c b/drivers/infiniband/core/addr.c index a5b4cf030c11..cb1d2ab13c66 100644 --- a/drivers/infiniband/core/addr.c +++ b/drivers/infiniband/core/addr.c @@ -207,6 +207,22 @@ int rdma_addr_size(struct sockaddr *addr) } EXPORT_SYMBOL(rdma_addr_size); +int rdma_addr_size_in6(struct sockaddr_in6 *addr) +{ + int ret = rdma_addr_size((struct sockaddr *) addr); + + return ret <= sizeof(*addr) ? ret : 0; +} +EXPORT_SYMBOL(rdma_addr_size_in6); + +int rdma_addr_size_kss(struct __kernel_sockaddr_storage *addr) +{ + int ret = rdma_addr_size((struct sockaddr *) addr); + + return ret <= sizeof(*addr) ? ret : 0; +} +EXPORT_SYMBOL(rdma_addr_size_kss); + static struct rdma_addr_client self; void rdma_addr_register_client(struct rdma_addr_client *client) @@ -550,18 +566,13 @@ static int addr_resolve(struct sockaddr *src_in, dst_release(dst); } - if (ndev->flags & IFF_LOOPBACK) { - ret = rdma_translate_ip(dst_in, addr); - /* - * Put the loopback device and get the translated - * device instead. - */ + if (ndev) { + if (ndev->flags & IFF_LOOPBACK) + ret = rdma_translate_ip(dst_in, addr); + else + addr->bound_dev_if = ndev->ifindex; dev_put(ndev); - ndev = dev_get_by_index(addr->net, addr->bound_dev_if); - } else { - addr->bound_dev_if = ndev->ifindex; } - dev_put(ndev); return ret; } @@ -591,6 +602,15 @@ static void process_one_req(struct work_struct *_work) list_del(&req->list); mutex_unlock(&lock); + /* + * Although the work will normally have been canceled by the + * workqueue, it can still be requeued as long as it is on the + * req_list, so it could have been requeued before we grabbed &lock. + * We need to cancel it after it is removed from req_list to really be + * sure it is safe to free. + */ + cancel_delayed_work(&req->work); + req->callback(req->status, (struct sockaddr *)&req->src_addr, req->addr, req->context); put_client(req->client); diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c index e66963ca58bd..6ab1059fed66 100644 --- a/drivers/infiniband/core/cma.c +++ b/drivers/infiniband/core/cma.c @@ -1334,7 +1334,7 @@ static bool validate_ipv6_net_dev(struct net_device *net_dev, IPV6_ADDR_LINKLOCAL; struct rt6_info *rt = rt6_lookup(dev_net(net_dev), &dst_addr->sin6_addr, &src_addr->sin6_addr, net_dev->ifindex, - strict); + NULL, strict); bool ret; if (!rt) @@ -3069,7 +3069,8 @@ static int cma_port_is_unique(struct rdma_bind_list *bind_list, continue; /* different dest port -> unique */ - if (!cma_any_port(cur_daddr) && + if (!cma_any_port(daddr) && + !cma_any_port(cur_daddr) && (dport != cur_dport)) continue; @@ -3080,7 +3081,8 @@ static int cma_port_is_unique(struct rdma_bind_list *bind_list, continue; /* different dst address -> unique */ - if (!cma_any_addr(cur_daddr) && + if (!cma_any_addr(daddr) && + !cma_any_addr(cur_daddr) && cma_addr_cmp(daddr, cur_daddr)) continue; @@ -3378,13 +3380,13 @@ int rdma_bind_addr(struct rdma_cm_id *id, struct sockaddr *addr) } #endif } + daddr = cma_dst_addr(id_priv); + daddr->sa_family = addr->sa_family; + ret = cma_get_port(id_priv); if (ret) goto err2; - daddr = cma_dst_addr(id_priv); - daddr->sa_family = addr->sa_family; - return 0; err2: if (id_priv->cma_dev) @@ -4173,6 +4175,9 @@ int rdma_join_multicast(struct rdma_cm_id *id, struct sockaddr *addr, struct cma_multicast *mc; int ret; + if (!id->device) + return -EINVAL; + id_priv = container_of(id, struct rdma_id_private, id); if (!cma_comp(id_priv, RDMA_CM_ADDR_BOUND) && !cma_comp(id_priv, RDMA_CM_ADDR_RESOLVED)) diff --git a/drivers/infiniband/core/core_priv.h b/drivers/infiniband/core/core_priv.h index c4560d84dfae..25bb178f6074 100644 --- a/drivers/infiniband/core/core_priv.h +++ b/drivers/infiniband/core/core_priv.h @@ -305,16 +305,21 @@ void nldev_exit(void); static inline struct ib_qp *_ib_create_qp(struct ib_device *dev, struct ib_pd *pd, struct ib_qp_init_attr *attr, - struct ib_udata *udata) + struct ib_udata *udata, + struct ib_uobject *uobj) { struct ib_qp *qp; + if (!dev->create_qp) + return ERR_PTR(-EOPNOTSUPP); + qp = dev->create_qp(pd, attr, udata); if (IS_ERR(qp)) return qp; qp->device = dev; qp->pd = pd; + qp->uobject = uobj; /* * We don't track XRC QPs for now, because they don't have PD * and more importantly they are created internaly by driver, diff --git a/drivers/infiniband/core/cq.c b/drivers/infiniband/core/cq.c index bc79ca8215d7..af5ad6a56ae4 100644 --- a/drivers/infiniband/core/cq.c +++ b/drivers/infiniband/core/cq.c @@ -17,6 +17,7 @@ /* # of WCs to poll for with a single call to ib_poll_cq */ #define IB_POLL_BATCH 16 +#define IB_POLL_BATCH_DIRECT 8 /* # of WCs to iterate over before yielding */ #define IB_POLL_BUDGET_IRQ 256 @@ -25,18 +26,18 @@ #define IB_POLL_FLAGS \ (IB_CQ_NEXT_COMP | IB_CQ_REPORT_MISSED_EVENTS) -static int __ib_process_cq(struct ib_cq *cq, int budget, struct ib_wc *poll_wc) +static int __ib_process_cq(struct ib_cq *cq, int budget, struct ib_wc *wcs, + int batch) { int i, n, completed = 0; - struct ib_wc *wcs = poll_wc ? : cq->wc; /* * budget might be (-1) if the caller does not * want to bound this call, thus we need unsigned * minimum here. */ - while ((n = ib_poll_cq(cq, min_t(u32, IB_POLL_BATCH, - budget - completed), wcs)) > 0) { + while ((n = ib_poll_cq(cq, min_t(u32, batch, + budget - completed), wcs)) > 0) { for (i = 0; i < n; i++) { struct ib_wc *wc = &wcs[i]; @@ -48,8 +49,7 @@ static int __ib_process_cq(struct ib_cq *cq, int budget, struct ib_wc *poll_wc) completed += n; - if (n != IB_POLL_BATCH || - (budget != -1 && completed >= budget)) + if (n != batch || (budget != -1 && completed >= budget)) break; } @@ -72,9 +72,9 @@ static int __ib_process_cq(struct ib_cq *cq, int budget, struct ib_wc *poll_wc) */ int ib_process_cq_direct(struct ib_cq *cq, int budget) { - struct ib_wc wcs[IB_POLL_BATCH]; + struct ib_wc wcs[IB_POLL_BATCH_DIRECT]; - return __ib_process_cq(cq, budget, wcs); + return __ib_process_cq(cq, budget, wcs, IB_POLL_BATCH_DIRECT); } EXPORT_SYMBOL(ib_process_cq_direct); @@ -88,7 +88,7 @@ static int ib_poll_handler(struct irq_poll *iop, int budget) struct ib_cq *cq = container_of(iop, struct ib_cq, iop); int completed; - completed = __ib_process_cq(cq, budget, NULL); + completed = __ib_process_cq(cq, budget, cq->wc, IB_POLL_BATCH); if (completed < budget) { irq_poll_complete(&cq->iop); if (ib_req_notify_cq(cq, IB_POLL_FLAGS) > 0) @@ -108,7 +108,8 @@ static void ib_cq_poll_work(struct work_struct *work) struct ib_cq *cq = container_of(work, struct ib_cq, work); int completed; - completed = __ib_process_cq(cq, IB_POLL_BUDGET_WORKQUEUE, NULL); + completed = __ib_process_cq(cq, IB_POLL_BUDGET_WORKQUEUE, cq->wc, + IB_POLL_BATCH); if (completed >= IB_POLL_BUDGET_WORKQUEUE || ib_req_notify_cq(cq, IB_POLL_FLAGS) > 0) queue_work(ib_comp_wq, &cq->work); diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c index e8010e73a1cf..b7459cf524e4 100644 --- a/drivers/infiniband/core/device.c +++ b/drivers/infiniband/core/device.c @@ -290,6 +290,7 @@ void ib_dealloc_device(struct ib_device *device) { WARN_ON(device->reg_state != IB_DEV_UNREGISTERED && device->reg_state != IB_DEV_UNINITIALIZED); + rdma_restrack_clean(&device->res); put_device(&device->dev); } EXPORT_SYMBOL(ib_dealloc_device); @@ -536,14 +537,14 @@ int ib_register_device(struct ib_device *device, ret = device->query_device(device, &device->attrs, &uhw); if (ret) { pr_warn("Couldn't query the device attributes\n"); - goto cache_cleanup; + goto cg_cleanup; } ret = ib_device_register_sysfs(device, port_callback); if (ret) { pr_warn("Couldn't register device %s with driver model\n", device->name); - goto cache_cleanup; + goto cg_cleanup; } device->reg_state = IB_DEV_REGISTERED; @@ -559,6 +560,8 @@ int ib_register_device(struct ib_device *device, mutex_unlock(&device_mutex); return 0; +cg_cleanup: + ib_device_unregister_rdmacg(device); cache_cleanup: ib_cache_cleanup_one(device); ib_cache_release_one(device); @@ -598,8 +601,6 @@ void ib_unregister_device(struct ib_device *device) } up_read(&lists_rwsem); - rdma_restrack_clean(&device->res); - ib_device_unregister_rdmacg(device); ib_device_unregister_sysfs(device); diff --git a/drivers/infiniband/core/rdma_core.c b/drivers/infiniband/core/rdma_core.c index 85b5ee4defa4..d8eead5d106d 100644 --- a/drivers/infiniband/core/rdma_core.c +++ b/drivers/infiniband/core/rdma_core.c @@ -141,7 +141,12 @@ static struct ib_uobject *alloc_uobj(struct ib_ucontext *context, */ uobj->context = context; uobj->type = type; - atomic_set(&uobj->usecnt, 0); + /* + * Allocated objects start out as write locked to deny any other + * syscalls from accessing them until they are committed. See + * rdma_alloc_commit_uobject + */ + atomic_set(&uobj->usecnt, -1); kref_init(&uobj->ref); return uobj; @@ -196,7 +201,15 @@ static struct ib_uobject *lookup_get_idr_uobject(const struct uverbs_obj_type *t goto free; } - uverbs_uobject_get(uobj); + /* + * The idr_find is guaranteed to return a pointer to something that + * isn't freed yet, or NULL, as the free after idr_remove goes through + * kfree_rcu(). However the object may still have been released and + * kfree() could be called at any time. + */ + if (!kref_get_unless_zero(&uobj->ref)) + uobj = ERR_PTR(-ENOENT); + free: rcu_read_unlock(); return uobj; @@ -399,13 +412,13 @@ static int __must_check remove_commit_fd_uobject(struct ib_uobject *uobj, return ret; } -static void lockdep_check(struct ib_uobject *uobj, bool exclusive) +static void assert_uverbs_usecnt(struct ib_uobject *uobj, bool exclusive) { #ifdef CONFIG_LOCKDEP if (exclusive) - WARN_ON(atomic_read(&uobj->usecnt) > 0); + WARN_ON(atomic_read(&uobj->usecnt) != -1); else - WARN_ON(atomic_read(&uobj->usecnt) == -1); + WARN_ON(atomic_read(&uobj->usecnt) <= 0); #endif } @@ -444,7 +457,7 @@ int __must_check rdma_remove_commit_uobject(struct ib_uobject *uobj) WARN(true, "ib_uverbs: Cleanup is running while removing an uobject\n"); return 0; } - lockdep_check(uobj, true); + assert_uverbs_usecnt(uobj, true); ret = _rdma_remove_commit_uobject(uobj, RDMA_REMOVE_DESTROY); up_read(&ucontext->cleanup_rwsem); @@ -474,16 +487,17 @@ int rdma_explicit_destroy(struct ib_uobject *uobject) WARN(true, "ib_uverbs: Cleanup is running while removing an uobject\n"); return 0; } - lockdep_check(uobject, true); + assert_uverbs_usecnt(uobject, true); ret = uobject->type->type_class->remove_commit(uobject, RDMA_REMOVE_DESTROY); if (ret) - return ret; + goto out; uobject->type = &null_obj_type; +out: up_read(&ucontext->cleanup_rwsem); - return 0; + return ret; } static void alloc_commit_idr_uobject(struct ib_uobject *uobj) @@ -527,6 +541,10 @@ int rdma_alloc_commit_uobject(struct ib_uobject *uobj) return ret; } + /* matches atomic_set(-1) in alloc_uobj */ + assert_uverbs_usecnt(uobj, true); + atomic_set(&uobj->usecnt, 0); + uobj->type->type_class->alloc_commit(uobj); up_read(&uobj->context->cleanup_rwsem); @@ -561,7 +579,7 @@ static void lookup_put_fd_uobject(struct ib_uobject *uobj, bool exclusive) void rdma_lookup_put_uobject(struct ib_uobject *uobj, bool exclusive) { - lockdep_check(uobj, exclusive); + assert_uverbs_usecnt(uobj, exclusive); uobj->type->type_class->lookup_put(uobj, exclusive); /* * In order to unlock an object, either decrease its usecnt for diff --git a/drivers/infiniband/core/restrack.c b/drivers/infiniband/core/restrack.c index 857637bf46da..3dbc4e4cca41 100644 --- a/drivers/infiniband/core/restrack.c +++ b/drivers/infiniband/core/restrack.c @@ -7,7 +7,6 @@ #include <rdma/restrack.h> #include <linux/mutex.h> #include <linux/sched/task.h> -#include <linux/uaccess.h> #include <linux/pid_namespace.h> void rdma_restrack_init(struct rdma_restrack_root *res) @@ -63,7 +62,6 @@ static struct ib_device *res_to_dev(struct rdma_restrack_entry *res) { enum rdma_restrack_type type = res->type; struct ib_device *dev; - struct ib_xrcd *xrcd; struct ib_pd *pd; struct ib_cq *cq; struct ib_qp *qp; @@ -81,10 +79,6 @@ static struct ib_device *res_to_dev(struct rdma_restrack_entry *res) qp = container_of(res, struct ib_qp, res); dev = qp->device; break; - case RDMA_RESTRACK_XRCD: - xrcd = container_of(res, struct ib_xrcd, res); - dev = xrcd->device; - break; default: WARN_ONCE(true, "Wrong resource tracking type %u\n", type); return NULL; @@ -93,6 +87,21 @@ static struct ib_device *res_to_dev(struct rdma_restrack_entry *res) return dev; } +static bool res_is_user(struct rdma_restrack_entry *res) +{ + switch (res->type) { + case RDMA_RESTRACK_PD: + return container_of(res, struct ib_pd, res)->uobject; + case RDMA_RESTRACK_CQ: + return container_of(res, struct ib_cq, res)->uobject; + case RDMA_RESTRACK_QP: + return container_of(res, struct ib_qp, res)->uobject; + default: + WARN_ONCE(true, "Wrong resource tracking type %u\n", res->type); + return false; + } +} + void rdma_restrack_add(struct rdma_restrack_entry *res) { struct ib_device *dev = res_to_dev(res); @@ -100,7 +109,7 @@ void rdma_restrack_add(struct rdma_restrack_entry *res) if (!dev) return; - if (!uaccess_kernel()) { + if (res_is_user(res)) { get_task_struct(current); res->task = current; res->kern_name = NULL; diff --git a/drivers/infiniband/core/roce_gid_mgmt.c b/drivers/infiniband/core/roce_gid_mgmt.c index 5a52ec77940a..cc2966380c0c 100644 --- a/drivers/infiniband/core/roce_gid_mgmt.c +++ b/drivers/infiniband/core/roce_gid_mgmt.c @@ -403,10 +403,12 @@ static void enum_all_gids_of_dev_cb(struct ib_device *ib_dev, * our feet */ rtnl_lock(); + down_read(&net_rwsem); for_each_net(net) for_each_netdev(net, ndev) if (is_eth_port_of_netdev(ib_dev, port, rdma_ndev, ndev)) add_netdev_ips(ib_dev, port, rdma_ndev, ndev); + up_read(&net_rwsem); rtnl_unlock(); } diff --git a/drivers/infiniband/core/sa_query.c b/drivers/infiniband/core/sa_query.c index 8cf15d4a8ac4..9f029a1ca5ea 100644 --- a/drivers/infiniband/core/sa_query.c +++ b/drivers/infiniband/core/sa_query.c @@ -1291,10 +1291,9 @@ int ib_init_ah_attr_from_path(struct ib_device *device, u8 port_num, resolved_dev = dev_get_by_index(dev_addr.net, dev_addr.bound_dev_if); - if (resolved_dev->flags & IFF_LOOPBACK) { - dev_put(resolved_dev); - resolved_dev = idev; - dev_hold(resolved_dev); + if (!resolved_dev) { + dev_put(idev); + return -ENODEV; } ndev = ib_get_ndev_from_path(rec); rcu_read_lock(); diff --git a/drivers/infiniband/core/ucma.c b/drivers/infiniband/core/ucma.c index f015f1bf88c9..d933336d7e01 100644 --- a/drivers/infiniband/core/ucma.c +++ b/drivers/infiniband/core/ucma.c @@ -132,7 +132,7 @@ static inline struct ucma_context *_ucma_find_context(int id, ctx = idr_find(&ctx_idr, id); if (!ctx) ctx = ERR_PTR(-ENOENT); - else if (ctx->file != file) + else if (ctx->file != file || !ctx->cm_id) ctx = ERR_PTR(-EINVAL); return ctx; } @@ -456,6 +456,7 @@ static ssize_t ucma_create_id(struct ucma_file *file, const char __user *inbuf, struct rdma_ucm_create_id cmd; struct rdma_ucm_create_id_resp resp; struct ucma_context *ctx; + struct rdma_cm_id *cm_id; enum ib_qp_type qp_type; int ret; @@ -476,10 +477,10 @@ static ssize_t ucma_create_id(struct ucma_file *file, const char __user *inbuf, return -ENOMEM; ctx->uid = cmd.uid; - ctx->cm_id = rdma_create_id(current->nsproxy->net_ns, - ucma_event_handler, ctx, cmd.ps, qp_type); - if (IS_ERR(ctx->cm_id)) { - ret = PTR_ERR(ctx->cm_id); + cm_id = rdma_create_id(current->nsproxy->net_ns, + ucma_event_handler, ctx, cmd.ps, qp_type); + if (IS_ERR(cm_id)) { + ret = PTR_ERR(cm_id); goto err1; } @@ -489,14 +490,19 @@ static ssize_t ucma_create_id(struct ucma_file *file, const char __user *inbuf, ret = -EFAULT; goto err2; } + + ctx->cm_id = cm_id; return 0; err2: - rdma_destroy_id(ctx->cm_id); + rdma_destroy_id(cm_id); err1: mutex_lock(&mut); idr_remove(&ctx_idr, ctx->id); mutex_unlock(&mut); + mutex_lock(&file->mut); + list_del(&ctx->list); + mutex_unlock(&file->mut); kfree(ctx); return ret; } @@ -626,6 +632,9 @@ static ssize_t ucma_bind_ip(struct ucma_file *file, const char __user *inbuf, if (copy_from_user(&cmd, inbuf, sizeof(cmd))) return -EFAULT; + if (!rdma_addr_size_in6(&cmd.addr)) + return -EINVAL; + ctx = ucma_get_ctx(file, cmd.id); if (IS_ERR(ctx)) return PTR_ERR(ctx); @@ -639,22 +648,21 @@ static ssize_t ucma_bind(struct ucma_file *file, const char __user *inbuf, int in_len, int out_len) { struct rdma_ucm_bind cmd; - struct sockaddr *addr; struct ucma_context *ctx; int ret; if (copy_from_user(&cmd, inbuf, sizeof(cmd))) return -EFAULT; - addr = (struct sockaddr *) &cmd.addr; - if (cmd.reserved || !cmd.addr_size || (cmd.addr_size != rdma_addr_size(addr))) + if (cmd.reserved || !cmd.addr_size || + cmd.addr_size != rdma_addr_size_kss(&cmd.addr)) return -EINVAL; ctx = ucma_get_ctx(file, cmd.id); if (IS_ERR(ctx)) return PTR_ERR(ctx); - ret = rdma_bind_addr(ctx->cm_id, addr); + ret = rdma_bind_addr(ctx->cm_id, (struct sockaddr *) &cmd.addr); ucma_put_ctx(ctx); return ret; } @@ -670,13 +678,16 @@ static ssize_t ucma_resolve_ip(struct ucma_file *file, if (copy_from_user(&cmd, inbuf, sizeof(cmd))) return -EFAULT; + if (!rdma_addr_size_in6(&cmd.src_addr) || + !rdma_addr_size_in6(&cmd.dst_addr)) + return -EINVAL; + ctx = ucma_get_ctx(file, cmd.id); if (IS_ERR(ctx)) return PTR_ERR(ctx); ret = rdma_resolve_addr(ctx->cm_id, (struct sockaddr *) &cmd.src_addr, - (struct sockaddr *) &cmd.dst_addr, - cmd.timeout_ms); + (struct sockaddr *) &cmd.dst_addr, cmd.timeout_ms); ucma_put_ctx(ctx); return ret; } @@ -686,24 +697,23 @@ static ssize_t ucma_resolve_addr(struct ucma_file *file, int in_len, int out_len) { struct rdma_ucm_resolve_addr cmd; - struct sockaddr *src, *dst; struct ucma_context *ctx; int ret; if (copy_from_user(&cmd, inbuf, sizeof(cmd))) return -EFAULT; - src = (struct sockaddr *) &cmd.src_addr; - dst = (struct sockaddr *) &cmd.dst_addr; - if (cmd.reserved || (cmd.src_size && (cmd.src_size != rdma_addr_size(src))) || - !cmd.dst_size || (cmd.dst_size != rdma_addr_size(dst))) + if (cmd.reserved || + (cmd.src_size && (cmd.src_size != rdma_addr_size_kss(&cmd.src_addr))) || + !cmd.dst_size || (cmd.dst_size != rdma_addr_size_kss(&cmd.dst_addr))) return -EINVAL; ctx = ucma_get_ctx(file, cmd.id); if (IS_ERR(ctx)) return PTR_ERR(ctx); - ret = rdma_resolve_addr(ctx->cm_id, src, dst, cmd.timeout_ms); + ret = rdma_resolve_addr(ctx->cm_id, (struct sockaddr *) &cmd.src_addr, + (struct sockaddr *) &cmd.dst_addr, cmd.timeout_ms); ucma_put_ctx(ctx); return ret; } @@ -1149,10 +1159,18 @@ static ssize_t ucma_init_qp_attr(struct ucma_file *file, if (copy_from_user(&cmd, inbuf, sizeof(cmd))) return -EFAULT; + if (cmd.qp_state > IB_QPS_ERR) + return -EINVAL; + ctx = ucma_get_ctx(file, cmd.id); if (IS_ERR(ctx)) return PTR_ERR(ctx); + if (!ctx->cm_id->device) { + ret = -EINVAL; + goto out; + } + resp.qp_attr_mask = 0; memset(&qp_attr, 0, sizeof qp_attr); qp_attr.qp_state = cmd.qp_state; @@ -1294,6 +1312,9 @@ static ssize_t ucma_set_option(struct ucma_file *file, const char __user *inbuf, if (IS_ERR(ctx)) return PTR_ERR(ctx); + if (unlikely(cmd.optlen > KMALLOC_MAX_SIZE)) + return -EINVAL; + optval = memdup_user((void __user *) (unsigned long) cmd.optval, cmd.optlen); if (IS_ERR(optval)) { @@ -1315,7 +1336,7 @@ static ssize_t ucma_notify(struct ucma_file *file, const char __user *inbuf, { struct rdma_ucm_notify cmd; struct ucma_context *ctx; - int ret; + int ret = -EINVAL; if (copy_from_user(&cmd, inbuf, sizeof(cmd))) return -EFAULT; @@ -1324,7 +1345,9 @@ static ssize_t ucma_notify(struct ucma_file *file, const char __user *inbuf, if (IS_ERR(ctx)) return PTR_ERR(ctx); - ret = rdma_notify(ctx->cm_id, (enum ib_event_type) cmd.event); + if (ctx->cm_id->device) + ret = rdma_notify(ctx->cm_id, (enum ib_event_type)cmd.event); + ucma_put_ctx(ctx); return ret; } @@ -1343,7 +1366,7 @@ static ssize_t ucma_process_join(struct ucma_file *file, return -ENOSPC; addr = (struct sockaddr *) &cmd->addr; - if (!cmd->addr_size || (cmd->addr_size != rdma_addr_size(addr))) + if (cmd->addr_size != rdma_addr_size(addr)) return -EINVAL; if (cmd->join_flags == RDMA_MC_JOIN_FLAG_FULLMEMBER) @@ -1410,7 +1433,10 @@ static ssize_t ucma_join_ip_multicast(struct ucma_file *file, join_cmd.response = cmd.response; join_cmd.uid = cmd.uid; join_cmd.id = cmd.id; - join_cmd.addr_size = rdma_addr_size((struct sockaddr *) &cmd.addr); + join_cmd.addr_size = rdma_addr_size_in6(&cmd.addr); + if (!join_cmd.addr_size) + return -EINVAL; + join_cmd.join_flags = RDMA_MC_JOIN_FLAG_FULLMEMBER; memcpy(&join_cmd.addr, &cmd.addr, join_cmd.addr_size); @@ -1426,6 +1452,9 @@ static ssize_t ucma_join_multicast(struct ucma_file *file, if (copy_from_user(&cmd, inbuf, sizeof(cmd))) return -EFAULT; + if (!rdma_addr_size_kss(&cmd.addr)) + return -EINVAL; + return ucma_process_join(file, &cmd, out_len); } diff --git a/drivers/infiniband/core/uverbs_cmd.c b/drivers/infiniband/core/uverbs_cmd.c index 256934d1f64f..a148de35df8d 100644 --- a/drivers/infiniband/core/uverbs_cmd.c +++ b/drivers/infiniband/core/uverbs_cmd.c @@ -562,9 +562,10 @@ ssize_t ib_uverbs_open_xrcd(struct ib_uverbs_file *file, if (f.file) fdput(f); + mutex_unlock(&file->device->xrcd_tree_mutex); + uobj_alloc_commit(&obj->uobject); - mutex_unlock(&file->device->xrcd_tree_mutex); return in_len; err_copy: @@ -603,10 +604,8 @@ ssize_t ib_uverbs_close_xrcd(struct ib_uverbs_file *file, uobj = uobj_get_write(uobj_get_type(xrcd), cmd.xrcd_handle, file->ucontext); - if (IS_ERR(uobj)) { - mutex_unlock(&file->device->xrcd_tree_mutex); + if (IS_ERR(uobj)) return PTR_ERR(uobj); - } ret = uobj_remove_commit(uobj); return ret ?: in_len; @@ -979,6 +978,9 @@ static struct ib_ucq_object *create_cq(struct ib_uverbs_file *file, struct ib_uverbs_ex_create_cq_resp resp; struct ib_cq_init_attr attr = {}; + if (!ib_dev->create_cq) + return ERR_PTR(-EOPNOTSUPP); + if (cmd->comp_vector >= file->device->num_comp_vectors) return ERR_PTR(-EINVAL); @@ -1030,14 +1032,14 @@ static struct ib_ucq_object *create_cq(struct ib_uverbs_file *file, resp.response_length = offsetof(typeof(resp), response_length) + sizeof(resp.response_length); + cq->res.type = RDMA_RESTRACK_CQ; + rdma_restrack_add(&cq->res); + ret = cb(file, obj, &resp, ucore, context); if (ret) goto err_cb; uobj_alloc_commit(&obj->uobject); - cq->res.type = RDMA_RESTRACK_CQ; - rdma_restrack_add(&cq->res); - return obj; err_cb: @@ -1518,7 +1520,8 @@ static int create_qp(struct ib_uverbs_file *file, if (cmd->qp_type == IB_QPT_XRC_TGT) qp = ib_create_qp(pd, &attr); else - qp = _ib_create_qp(device, pd, &attr, uhw); + qp = _ib_create_qp(device, pd, &attr, uhw, + &obj->uevent.uobject); if (IS_ERR(qp)) { ret = PTR_ERR(qp); @@ -1550,8 +1553,10 @@ static int create_qp(struct ib_uverbs_file *file, atomic_inc(&attr.srq->usecnt); if (ind_tbl) atomic_inc(&ind_tbl->usecnt); + } else { + /* It is done in _ib_create_qp for other QP types */ + qp->uobject = &obj->uevent.uobject; } - qp->uobject = &obj->uevent.uobject; obj->uevent.uobject.object = qp; @@ -1971,8 +1976,15 @@ static int modify_qp(struct ib_uverbs_file *file, goto release_qp; } + if ((cmd->base.attr_mask & IB_QP_AV) && + !rdma_is_port_valid(qp->device, cmd->base.dest.port_num)) { + ret = -EINVAL; + goto release_qp; + } + if ((cmd->base.attr_mask & IB_QP_ALT_PATH) && - !rdma_is_port_valid(qp->device, cmd->base.alt_port_num)) { + (!rdma_is_port_valid(qp->device, cmd->base.alt_port_num) || + !rdma_is_port_valid(qp->device, cmd->base.alt_dest.port_num))) { ret = -EINVAL; goto release_qp; } @@ -2941,6 +2953,11 @@ int ib_uverbs_ex_create_wq(struct ib_uverbs_file *file, wq_init_attr.create_flags = cmd.create_flags; obj->uevent.events_reported = 0; INIT_LIST_HEAD(&obj->uevent.event_list); + + if (!pd->device->create_wq) { + err = -EOPNOTSUPP; + goto err_put_cq; + } wq = pd->device->create_wq(pd, &wq_init_attr, uhw); if (IS_ERR(wq)) { err = PTR_ERR(wq); @@ -3084,7 +3101,12 @@ int ib_uverbs_ex_modify_wq(struct ib_uverbs_file *file, wq_attr.flags = cmd.flags; wq_attr.flags_mask = cmd.flags_mask; } + if (!wq->device->modify_wq) { + ret = -EOPNOTSUPP; + goto out; + } ret = wq->device->modify_wq(wq, &wq_attr, cmd.attr_mask, uhw); +out: uobj_put_obj_read(wq); return ret; } @@ -3181,6 +3203,11 @@ int ib_uverbs_ex_create_rwq_ind_table(struct ib_uverbs_file *file, init_attr.log_ind_tbl_size = cmd.log_ind_tbl_size; init_attr.ind_tbl = wqs; + + if (!ib_dev->create_rwq_ind_table) { + err = -EOPNOTSUPP; + goto err_uobj; + } rwq_ind_tbl = ib_dev->create_rwq_ind_table(ib_dev, &init_attr, uhw); if (IS_ERR(rwq_ind_tbl)) { @@ -3770,6 +3797,9 @@ int ib_uverbs_ex_query_device(struct ib_uverbs_file *file, struct ib_device_attr attr = {0}; int err; + if (!ib_dev->query_device) + return -EOPNOTSUPP; + if (ucore->inlen < sizeof(cmd)) return -EINVAL; diff --git a/drivers/infiniband/core/uverbs_ioctl.c b/drivers/infiniband/core/uverbs_ioctl.c index d96dc1d17be1..339b85145044 100644 --- a/drivers/infiniband/core/uverbs_ioctl.c +++ b/drivers/infiniband/core/uverbs_ioctl.c @@ -59,6 +59,9 @@ static int uverbs_process_attr(struct ib_device *ibdev, return 0; } + if (test_bit(attr_id, attr_bundle_h->valid_bitmap)) + return -EINVAL; + spec = &attr_spec_bucket->attrs[attr_id]; e = &elements[attr_id]; e->uattr = uattr_ptr; diff --git a/drivers/infiniband/core/uverbs_ioctl_merge.c b/drivers/infiniband/core/uverbs_ioctl_merge.c index 062485f9300d..62e1eb1d2a28 100644 --- a/drivers/infiniband/core/uverbs_ioctl_merge.c +++ b/drivers/infiniband/core/uverbs_ioctl_merge.c @@ -114,6 +114,7 @@ static size_t get_elements_above_id(const void **iters, short min = SHRT_MAX; const void *elem; int i, j, last_stored = -1; + unsigned int equal_min = 0; for_each_element(elem, i, j, elements, num_elements, num_offset, data_offset) { @@ -136,6 +137,10 @@ static size_t get_elements_above_id(const void **iters, */ iters[last_stored == i ? num_iters - 1 : num_iters++] = elem; last_stored = i; + if (min == GET_ID(id)) + equal_min++; + else + equal_min = 1; min = GET_ID(id); } @@ -146,15 +151,10 @@ static size_t get_elements_above_id(const void **iters, * Therefore, we need to clean the beginning of the array to make sure * all ids of final elements are equal to min. */ - for (i = num_iters - 1; i >= 0 && - GET_ID(*(u16 *)(iters[i] + id_offset)) == min; i--) - ; - - num_iters -= i + 1; - memmove(iters, iters + i + 1, sizeof(*iters) * num_iters); + memmove(iters, iters + num_iters - equal_min, sizeof(*iters) * equal_min); *min_id = min; - return num_iters; + return equal_min; } #define find_max_element_entry_id(num_elements, elements, num_objects_fld, \ @@ -322,7 +322,7 @@ static struct uverbs_method_spec *build_method_with_attrs(const struct uverbs_me hash = kzalloc(sizeof(*hash) + ALIGN(sizeof(*hash->attrs) * (attr_max_bucket + 1), sizeof(long)) + - BITS_TO_LONGS(attr_max_bucket) * sizeof(long), + BITS_TO_LONGS(attr_max_bucket + 1) * sizeof(long), GFP_KERNEL); if (!hash) { res = -ENOMEM; @@ -509,7 +509,7 @@ static struct uverbs_object_spec *build_object_with_methods(const struct uverbs_ * first handler which != NULL. This also defines the * set of flags used for this handler. */ - for (i = num_object_defs - 1; + for (i = num_method_defs - 1; i >= 0 && !method_defs[i]->handler; i--) ; hash->methods[min_id++] = method; diff --git a/drivers/infiniband/core/uverbs_main.c b/drivers/infiniband/core/uverbs_main.c index 395a3b091229..b1ca223aa380 100644 --- a/drivers/infiniband/core/uverbs_main.c +++ b/drivers/infiniband/core/uverbs_main.c @@ -650,12 +650,21 @@ static int verify_command_mask(struct ib_device *ib_dev, __u32 command) return -1; } +static bool verify_command_idx(u32 command, bool extended) +{ + if (extended) + return command < ARRAY_SIZE(uverbs_ex_cmd_table); + + return command < ARRAY_SIZE(uverbs_cmd_table); +} + static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf, size_t count, loff_t *pos) { struct ib_uverbs_file *file = filp->private_data; struct ib_device *ib_dev; struct ib_uverbs_cmd_hdr hdr; + bool extended_command; __u32 command; __u32 flags; int srcu_key; @@ -688,6 +697,15 @@ static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf, } command = hdr.command & IB_USER_VERBS_CMD_COMMAND_MASK; + flags = (hdr.command & + IB_USER_VERBS_CMD_FLAGS_MASK) >> IB_USER_VERBS_CMD_FLAGS_SHIFT; + + extended_command = flags & IB_USER_VERBS_CMD_FLAG_EXTENDED; + if (!verify_command_idx(command, extended_command)) { + ret = -EINVAL; + goto out; + } + if (verify_command_mask(ib_dev, command)) { ret = -EOPNOTSUPP; goto out; @@ -699,12 +717,8 @@ static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf, goto out; } - flags = (hdr.command & - IB_USER_VERBS_CMD_FLAGS_MASK) >> IB_USER_VERBS_CMD_FLAGS_SHIFT; - if (!flags) { - if (command >= ARRAY_SIZE(uverbs_cmd_table) || - !uverbs_cmd_table[command]) { + if (!uverbs_cmd_table[command]) { ret = -EINVAL; goto out; } @@ -725,8 +739,7 @@ static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf, struct ib_udata uhw; size_t written_count = count; - if (command >= ARRAY_SIZE(uverbs_ex_cmd_table) || - !uverbs_ex_cmd_table[command]) { + if (!uverbs_ex_cmd_table[command]) { ret = -ENOSYS; goto out; } @@ -942,6 +955,7 @@ static const struct file_operations uverbs_fops = { .llseek = no_llseek, #if IS_ENABLED(CONFIG_INFINIBAND_EXP_USER_ACCESS) .unlocked_ioctl = ib_uverbs_ioctl, + .compat_ioctl = ib_uverbs_ioctl, #endif }; @@ -954,6 +968,7 @@ static const struct file_operations uverbs_mmap_fops = { .llseek = no_llseek, #if IS_ENABLED(CONFIG_INFINIBAND_EXP_USER_ACCESS) .unlocked_ioctl = ib_uverbs_ioctl, + .compat_ioctl = ib_uverbs_ioctl, #endif }; diff --git a/drivers/infiniband/core/uverbs_std_types.c b/drivers/infiniband/core/uverbs_std_types.c index cab0ac3556eb..df1360e6774f 100644 --- a/drivers/infiniband/core/uverbs_std_types.c +++ b/drivers/infiniband/core/uverbs_std_types.c @@ -234,15 +234,18 @@ static void create_udata(struct uverbs_attr_bundle *ctx, uverbs_attr_get(ctx, UVERBS_UHW_OUT); if (!IS_ERR(uhw_in)) { - udata->inbuf = uhw_in->ptr_attr.ptr; udata->inlen = uhw_in->ptr_attr.len; + if (uverbs_attr_ptr_is_inline(uhw_in)) + udata->inbuf = &uhw_in->uattr->data; + else + udata->inbuf = u64_to_user_ptr(uhw_in->ptr_attr.data); } else { udata->inbuf = NULL; udata->inlen = 0; } if (!IS_ERR(uhw_out)) { - udata->outbuf = uhw_out->ptr_attr.ptr; + udata->outbuf = u64_to_user_ptr(uhw_out->ptr_attr.data); udata->outlen = uhw_out->ptr_attr.len; } else { udata->outbuf = NULL; @@ -323,7 +326,8 @@ static int uverbs_create_cq_handler(struct ib_device *ib_dev, cq->res.type = RDMA_RESTRACK_CQ; rdma_restrack_add(&cq->res); - ret = uverbs_copy_to(attrs, CREATE_CQ_RESP_CQE, &cq->cqe); + ret = uverbs_copy_to(attrs, CREATE_CQ_RESP_CQE, &cq->cqe, + sizeof(cq->cqe)); if (ret) goto err_cq; @@ -375,7 +379,7 @@ static int uverbs_destroy_cq_handler(struct ib_device *ib_dev, resp.comp_events_reported = obj->comp_events_reported; resp.async_events_reported = obj->async_events_reported; - return uverbs_copy_to(attrs, DESTROY_CQ_RESP, &resp); + return uverbs_copy_to(attrs, DESTROY_CQ_RESP, &resp, sizeof(resp)); } static DECLARE_UVERBS_METHOD( diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c index 16ebc6372c31..93025d2009b8 100644 --- a/drivers/infiniband/core/verbs.c +++ b/drivers/infiniband/core/verbs.c @@ -887,7 +887,7 @@ struct ib_qp *ib_create_qp(struct ib_pd *pd, if (qp_init_attr->cap.max_rdma_ctxs) rdma_rw_init_qp(device, qp_init_attr); - qp = _ib_create_qp(device, pd, qp_init_attr, NULL); + qp = _ib_create_qp(device, pd, qp_init_attr, NULL, NULL); if (IS_ERR(qp)) return qp; @@ -898,7 +898,6 @@ struct ib_qp *ib_create_qp(struct ib_pd *pd, } qp->real_qp = qp; - qp->uobject = NULL; qp->qp_type = qp_init_attr->qp_type; qp->rwq_ind_tbl = qp_init_attr->rwq_ind_tbl; diff --git a/drivers/infiniband/hw/bnxt_re/bnxt_re.h b/drivers/infiniband/hw/bnxt_re/bnxt_re.h index ca32057e886f..96f76896488d 100644 --- a/drivers/infiniband/hw/bnxt_re/bnxt_re.h +++ b/drivers/infiniband/hw/bnxt_re/bnxt_re.h @@ -57,8 +57,8 @@ #define BNXT_RE_PAGE_SIZE_8M BIT(BNXT_RE_PAGE_SHIFT_8M) #define BNXT_RE_PAGE_SIZE_1G BIT(BNXT_RE_PAGE_SHIFT_1G) -#define BNXT_RE_MAX_MR_SIZE_LOW BIT(BNXT_RE_PAGE_SHIFT_1G) -#define BNXT_RE_MAX_MR_SIZE_HIGH BIT(39) +#define BNXT_RE_MAX_MR_SIZE_LOW BIT_ULL(BNXT_RE_PAGE_SHIFT_1G) +#define BNXT_RE_MAX_MR_SIZE_HIGH BIT_ULL(39) #define BNXT_RE_MAX_MR_SIZE BNXT_RE_MAX_MR_SIZE_HIGH #define BNXT_RE_MAX_QPC_COUNT (64 * 1024) @@ -120,7 +120,6 @@ struct bnxt_re_dev { #define BNXT_RE_FLAG_HAVE_L2_REF 3 #define BNXT_RE_FLAG_RCFW_CHANNEL_EN 4 #define BNXT_RE_FLAG_QOS_WORK_REG 5 -#define BNXT_RE_FLAG_TASK_IN_PROG 6 #define BNXT_RE_FLAG_ISSUE_ROCE_STATS 29 struct net_device *netdev; unsigned int version, major, minor; @@ -158,6 +157,7 @@ struct bnxt_re_dev { atomic_t srq_count; atomic_t mr_count; atomic_t mw_count; + atomic_t sched_count; /* Max of 2 lossless traffic class supported per port */ u16 cosq[2]; diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.c b/drivers/infiniband/hw/bnxt_re/ib_verbs.c index ae9e9ff54826..8301d7e5fa8c 100644 --- a/drivers/infiniband/hw/bnxt_re/ib_verbs.c +++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.c @@ -174,10 +174,8 @@ int bnxt_re_query_device(struct ib_device *ibdev, ib_attr->max_pd = dev_attr->max_pd; ib_attr->max_qp_rd_atom = dev_attr->max_qp_rd_atom; ib_attr->max_qp_init_rd_atom = dev_attr->max_qp_init_rd_atom; - if (dev_attr->is_atomic) { - ib_attr->atomic_cap = IB_ATOMIC_HCA; - ib_attr->masked_atomic_cap = IB_ATOMIC_HCA; - } + ib_attr->atomic_cap = IB_ATOMIC_NONE; + ib_attr->masked_atomic_cap = IB_ATOMIC_NONE; ib_attr->max_ee_rd_atom = 0; ib_attr->max_res_rd_atom = 0; @@ -787,20 +785,51 @@ int bnxt_re_query_ah(struct ib_ah *ib_ah, struct rdma_ah_attr *ah_attr) return 0; } +unsigned long bnxt_re_lock_cqs(struct bnxt_re_qp *qp) + __acquires(&qp->scq->cq_lock) __acquires(&qp->rcq->cq_lock) +{ + unsigned long flags; + + spin_lock_irqsave(&qp->scq->cq_lock, flags); + if (qp->rcq != qp->scq) + spin_lock(&qp->rcq->cq_lock); + else + __acquire(&qp->rcq->cq_lock); + + return flags; +} + +void bnxt_re_unlock_cqs(struct bnxt_re_qp *qp, + unsigned long flags) + __releases(&qp->scq->cq_lock) __releases(&qp->rcq->cq_lock) +{ + if (qp->rcq != qp->scq) + spin_unlock(&qp->rcq->cq_lock); + else + __release(&qp->rcq->cq_lock); + spin_unlock_irqrestore(&qp->scq->cq_lock, flags); +} + /* Queue Pairs */ int bnxt_re_destroy_qp(struct ib_qp *ib_qp) { struct bnxt_re_qp *qp = container_of(ib_qp, struct bnxt_re_qp, ib_qp); struct bnxt_re_dev *rdev = qp->rdev; int rc; + unsigned int flags; bnxt_qplib_flush_cqn_wq(&qp->qplib_qp); - bnxt_qplib_del_flush_qp(&qp->qplib_qp); rc = bnxt_qplib_destroy_qp(&rdev->qplib_res, &qp->qplib_qp); if (rc) { dev_err(rdev_to_dev(rdev), "Failed to destroy HW QP"); return rc; } + + flags = bnxt_re_lock_cqs(qp); + bnxt_qplib_clean_qp(&qp->qplib_qp); + bnxt_re_unlock_cqs(qp, flags); + bnxt_qplib_free_qp_res(&rdev->qplib_res, &qp->qplib_qp); + if (ib_qp->qp_type == IB_QPT_GSI && rdev->qp1_sqp) { rc = bnxt_qplib_destroy_ah(&rdev->qplib_res, &rdev->sqp_ah->qplib_ah); @@ -810,7 +839,7 @@ int bnxt_re_destroy_qp(struct ib_qp *ib_qp) return rc; } - bnxt_qplib_del_flush_qp(&qp->qplib_qp); + bnxt_qplib_clean_qp(&qp->qplib_qp); rc = bnxt_qplib_destroy_qp(&rdev->qplib_res, &rdev->qp1_sqp->qplib_qp); if (rc) { @@ -1069,6 +1098,7 @@ struct ib_qp *bnxt_re_create_qp(struct ib_pd *ib_pd, goto fail; } qp->qplib_qp.scq = &cq->qplib_cq; + qp->scq = cq; } if (qp_init_attr->recv_cq) { @@ -1080,6 +1110,7 @@ struct ib_qp *bnxt_re_create_qp(struct ib_pd *ib_pd, goto fail; } qp->qplib_qp.rcq = &cq->qplib_cq; + qp->rcq = cq; } if (qp_init_attr->srq) { @@ -1185,7 +1216,7 @@ struct ib_qp *bnxt_re_create_qp(struct ib_pd *ib_pd, rc = bnxt_qplib_create_qp(&rdev->qplib_res, &qp->qplib_qp); if (rc) { dev_err(rdev_to_dev(rdev), "Failed to create HW QP"); - goto fail; + goto free_umem; } } @@ -1213,6 +1244,13 @@ struct ib_qp *bnxt_re_create_qp(struct ib_pd *ib_pd, return &qp->ib_qp; qp_destroy: bnxt_qplib_destroy_qp(&rdev->qplib_res, &qp->qplib_qp); +free_umem: + if (udata) { + if (qp->rumem) + ib_umem_release(qp->rumem); + if (qp->sumem) + ib_umem_release(qp->sumem); + } fail: kfree(qp); return ERR_PTR(rc); @@ -1568,6 +1606,7 @@ int bnxt_re_modify_qp(struct ib_qp *ib_qp, struct ib_qp_attr *qp_attr, int status; union ib_gid sgid; struct ib_gid_attr sgid_attr; + unsigned int flags; u8 nw_type; qp->qplib_qp.modify_flags = 0; @@ -1596,14 +1635,18 @@ int bnxt_re_modify_qp(struct ib_qp *ib_qp, struct ib_qp_attr *qp_attr, dev_dbg(rdev_to_dev(rdev), "Move QP = %p to flush list\n", qp); + flags = bnxt_re_lock_cqs(qp); bnxt_qplib_add_flush_qp(&qp->qplib_qp); + bnxt_re_unlock_cqs(qp, flags); } if (!qp->sumem && qp->qplib_qp.state == CMDQ_MODIFY_QP_NEW_STATE_RESET) { dev_dbg(rdev_to_dev(rdev), "Move QP = %p out of flush list\n", qp); - bnxt_qplib_del_flush_qp(&qp->qplib_qp); + flags = bnxt_re_lock_cqs(qp); + bnxt_qplib_clean_qp(&qp->qplib_qp); + bnxt_re_unlock_cqs(qp, flags); } } if (qp_attr_mask & IB_QP_EN_SQD_ASYNC_NOTIFY) { @@ -2189,10 +2232,13 @@ static int bnxt_re_build_inv_wqe(struct ib_send_wr *wr, wqe->type = BNXT_QPLIB_SWQE_TYPE_LOCAL_INV; wqe->local_inv.inv_l_key = wr->ex.invalidate_rkey; + /* Need unconditional fence for local invalidate + * opcode to work as expected. + */ + wqe->flags |= BNXT_QPLIB_SWQE_FLAGS_UC_FENCE; + if (wr->send_flags & IB_SEND_SIGNALED) wqe->flags |= BNXT_QPLIB_SWQE_FLAGS_SIGNAL_COMP; - if (wr->send_flags & IB_SEND_FENCE) - wqe->flags |= BNXT_QPLIB_SWQE_FLAGS_UC_FENCE; if (wr->send_flags & IB_SEND_SOLICITED) wqe->flags |= BNXT_QPLIB_SWQE_FLAGS_SOLICIT_EVENT; @@ -2213,8 +2259,12 @@ static int bnxt_re_build_reg_wqe(struct ib_reg_wr *wr, wqe->frmr.levels = qplib_frpl->hwq.level + 1; wqe->type = BNXT_QPLIB_SWQE_TYPE_REG_MR; - if (wr->wr.send_flags & IB_SEND_FENCE) - wqe->flags |= BNXT_QPLIB_SWQE_FLAGS_UC_FENCE; + /* Need unconditional fence for reg_mr + * opcode to function as expected. + */ + + wqe->flags |= BNXT_QPLIB_SWQE_FLAGS_UC_FENCE; + if (wr->wr.send_flags & IB_SEND_SIGNALED) wqe->flags |= BNXT_QPLIB_SWQE_FLAGS_SIGNAL_COMP; @@ -3548,7 +3598,7 @@ struct ib_mr *bnxt_re_reg_user_mr(struct ib_pd *ib_pd, u64 start, u64 length, int umem_pgs, page_shift, rc; if (length > BNXT_RE_MAX_MR_SIZE) { - dev_err(rdev_to_dev(rdev), "MR Size: %lld > Max supported:%ld\n", + dev_err(rdev_to_dev(rdev), "MR Size: %lld > Max supported:%lld\n", length, BNXT_RE_MAX_MR_SIZE); return ERR_PTR(-ENOMEM); } diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.h b/drivers/infiniband/hw/bnxt_re/ib_verbs.h index 423ebe012f95..e62b7c2c7da6 100644 --- a/drivers/infiniband/hw/bnxt_re/ib_verbs.h +++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.h @@ -89,6 +89,8 @@ struct bnxt_re_qp { /* QP1 */ u32 send_psn; struct ib_ud_header qp1_hdr; + struct bnxt_re_cq *scq; + struct bnxt_re_cq *rcq; }; struct bnxt_re_cq { @@ -220,4 +222,7 @@ struct ib_ucontext *bnxt_re_alloc_ucontext(struct ib_device *ibdev, struct ib_udata *udata); int bnxt_re_dealloc_ucontext(struct ib_ucontext *context); int bnxt_re_mmap(struct ib_ucontext *context, struct vm_area_struct *vma); + +unsigned long bnxt_re_lock_cqs(struct bnxt_re_qp *qp); +void bnxt_re_unlock_cqs(struct bnxt_re_qp *qp, unsigned long flags); #endif /* __BNXT_RE_IB_VERBS_H__ */ diff --git a/drivers/infiniband/hw/bnxt_re/main.c b/drivers/infiniband/hw/bnxt_re/main.c index 508d00a5a106..f6e361750466 100644 --- a/drivers/infiniband/hw/bnxt_re/main.c +++ b/drivers/infiniband/hw/bnxt_re/main.c @@ -656,7 +656,6 @@ static void bnxt_re_dev_remove(struct bnxt_re_dev *rdev) mutex_unlock(&bnxt_re_dev_lock); synchronize_rcu(); - flush_workqueue(bnxt_re_wq); ib_dealloc_device(&rdev->ibdev); /* rdev is gone */ @@ -731,6 +730,13 @@ static int bnxt_re_handle_qp_async_event(struct creq_qp_event *qp_event, struct bnxt_re_qp *qp) { struct ib_event event; + unsigned int flags; + + if (qp->qplib_qp.state == CMDQ_MODIFY_QP_NEW_STATE_ERR) { + flags = bnxt_re_lock_cqs(qp); + bnxt_qplib_add_flush_qp(&qp->qplib_qp); + bnxt_re_unlock_cqs(qp, flags); + } memset(&event, 0, sizeof(event)); if (qp->qplib_qp.srq) { @@ -1417,9 +1423,12 @@ static void bnxt_re_task(struct work_struct *work) switch (re_work->event) { case NETDEV_REGISTER: rc = bnxt_re_ib_reg(rdev); - if (rc) + if (rc) { dev_err(rdev_to_dev(rdev), "Failed to register with IB: %#x", rc); + bnxt_re_remove_one(rdev); + bnxt_re_dev_unreg(rdev); + } break; case NETDEV_UP: bnxt_re_dispatch_event(&rdev->ibdev, NULL, 1, @@ -1441,7 +1450,7 @@ static void bnxt_re_task(struct work_struct *work) break; } smp_mb__before_atomic(); - clear_bit(BNXT_RE_FLAG_TASK_IN_PROG, &rdev->flags); + atomic_dec(&rdev->sched_count); kfree(re_work); } @@ -1503,7 +1512,7 @@ static int bnxt_re_netdev_event(struct notifier_block *notifier, /* netdev notifier will call NETDEV_UNREGISTER again later since * we are still holding the reference to the netdev */ - if (test_bit(BNXT_RE_FLAG_TASK_IN_PROG, &rdev->flags)) + if (atomic_read(&rdev->sched_count) > 0) goto exit; bnxt_re_ib_unreg(rdev, false); bnxt_re_remove_one(rdev); @@ -1523,7 +1532,7 @@ static int bnxt_re_netdev_event(struct notifier_block *notifier, re_work->vlan_dev = (real_dev == netdev ? NULL : netdev); INIT_WORK(&re_work->work, bnxt_re_task); - set_bit(BNXT_RE_FLAG_TASK_IN_PROG, &rdev->flags); + atomic_inc(&rdev->sched_count); queue_work(bnxt_re_wq, &re_work->work); } } @@ -1578,6 +1587,11 @@ static void __exit bnxt_re_mod_exit(void) */ list_for_each_entry_safe_reverse(rdev, next, &to_be_deleted, list) { dev_info(rdev_to_dev(rdev), "Unregistering Device"); + /* + * Flush out any scheduled tasks before destroying the + * resources + */ + flush_workqueue(bnxt_re_wq); bnxt_re_dev_stop(rdev); bnxt_re_ib_unreg(rdev, true); bnxt_re_remove_one(rdev); diff --git a/drivers/infiniband/hw/bnxt_re/qplib_fp.c b/drivers/infiniband/hw/bnxt_re/qplib_fp.c index 1b0e94697fe3..3a78faba8d91 100644 --- a/drivers/infiniband/hw/bnxt_re/qplib_fp.c +++ b/drivers/infiniband/hw/bnxt_re/qplib_fp.c @@ -88,75 +88,35 @@ static void __bnxt_qplib_add_flush_qp(struct bnxt_qplib_qp *qp) } } -void bnxt_qplib_acquire_cq_locks(struct bnxt_qplib_qp *qp, - unsigned long *flags) - __acquires(&qp->scq->hwq.lock) __acquires(&qp->rcq->hwq.lock) +static void bnxt_qplib_acquire_cq_flush_locks(struct bnxt_qplib_qp *qp, + unsigned long *flags) + __acquires(&qp->scq->flush_lock) __acquires(&qp->rcq->flush_lock) { - spin_lock_irqsave(&qp->scq->hwq.lock, *flags); + spin_lock_irqsave(&qp->scq->flush_lock, *flags); if (qp->scq == qp->rcq) - __acquire(&qp->rcq->hwq.lock); + __acquire(&qp->rcq->flush_lock); else - spin_lock(&qp->rcq->hwq.lock); + spin_lock(&qp->rcq->flush_lock); } -void bnxt_qplib_release_cq_locks(struct bnxt_qplib_qp *qp, - unsigned long *flags) - __releases(&qp->scq->hwq.lock) __releases(&qp->rcq->hwq.lock) +static void bnxt_qplib_release_cq_flush_locks(struct bnxt_qplib_qp *qp, + unsigned long *flags) + __releases(&qp->scq->flush_lock) __releases(&qp->rcq->flush_lock) { if (qp->scq == qp->rcq) - __release(&qp->rcq->hwq.lock); + __release(&qp->rcq->flush_lock); else - spin_unlock(&qp->rcq->hwq.lock); - spin_unlock_irqrestore(&qp->scq->hwq.lock, *flags); -} - -static struct bnxt_qplib_cq *bnxt_qplib_find_buddy_cq(struct bnxt_qplib_qp *qp, - struct bnxt_qplib_cq *cq) -{ - struct bnxt_qplib_cq *buddy_cq = NULL; - - if (qp->scq == qp->rcq) - buddy_cq = NULL; - else if (qp->scq == cq) - buddy_cq = qp->rcq; - else - buddy_cq = qp->scq; - return buddy_cq; -} - -static void bnxt_qplib_lock_buddy_cq(struct bnxt_qplib_qp *qp, - struct bnxt_qplib_cq *cq) - __acquires(&buddy_cq->hwq.lock) -{ - struct bnxt_qplib_cq *buddy_cq = NULL; - - buddy_cq = bnxt_qplib_find_buddy_cq(qp, cq); - if (!buddy_cq) - __acquire(&cq->hwq.lock); - else - spin_lock(&buddy_cq->hwq.lock); -} - -static void bnxt_qplib_unlock_buddy_cq(struct bnxt_qplib_qp *qp, - struct bnxt_qplib_cq *cq) - __releases(&buddy_cq->hwq.lock) -{ - struct bnxt_qplib_cq *buddy_cq = NULL; - - buddy_cq = bnxt_qplib_find_buddy_cq(qp, cq); - if (!buddy_cq) - __release(&cq->hwq.lock); - else - spin_unlock(&buddy_cq->hwq.lock); + spin_unlock(&qp->rcq->flush_lock); + spin_unlock_irqrestore(&qp->scq->flush_lock, *flags); } void bnxt_qplib_add_flush_qp(struct bnxt_qplib_qp *qp) { unsigned long flags; - bnxt_qplib_acquire_cq_locks(qp, &flags); + bnxt_qplib_acquire_cq_flush_locks(qp, &flags); __bnxt_qplib_add_flush_qp(qp); - bnxt_qplib_release_cq_locks(qp, &flags); + bnxt_qplib_release_cq_flush_locks(qp, &flags); } static void __bnxt_qplib_del_flush_qp(struct bnxt_qplib_qp *qp) @@ -173,11 +133,11 @@ static void __bnxt_qplib_del_flush_qp(struct bnxt_qplib_qp *qp) } } -void bnxt_qplib_del_flush_qp(struct bnxt_qplib_qp *qp) +void bnxt_qplib_clean_qp(struct bnxt_qplib_qp *qp) { unsigned long flags; - bnxt_qplib_acquire_cq_locks(qp, &flags); + bnxt_qplib_acquire_cq_flush_locks(qp, &flags); __clean_cq(qp->scq, (u64)(unsigned long)qp); qp->sq.hwq.prod = 0; qp->sq.hwq.cons = 0; @@ -186,7 +146,7 @@ void bnxt_qplib_del_flush_qp(struct bnxt_qplib_qp *qp) qp->rq.hwq.cons = 0; __bnxt_qplib_del_flush_qp(qp); - bnxt_qplib_release_cq_locks(qp, &flags); + bnxt_qplib_release_cq_flush_locks(qp, &flags); } static void bnxt_qpn_cqn_sched_task(struct work_struct *work) @@ -283,7 +243,7 @@ static void bnxt_qplib_service_nq(unsigned long data) u32 sw_cons, raw_cons; u16 type; int budget = nq->budget; - u64 q_handle; + uintptr_t q_handle; /* Service the NQ until empty */ raw_cons = hwq->cons; @@ -566,7 +526,7 @@ int bnxt_qplib_create_srq(struct bnxt_qplib_res *res, /* Configure the request */ req.dpi = cpu_to_le32(srq->dpi->dpi); - req.srq_handle = cpu_to_le64(srq); + req.srq_handle = cpu_to_le64((uintptr_t)srq); req.srq_size = cpu_to_le16((u16)srq->hwq.max_elements); pbl = &srq->hwq.pbl[PBL_LVL_0]; @@ -1419,7 +1379,6 @@ int bnxt_qplib_destroy_qp(struct bnxt_qplib_res *res, struct bnxt_qplib_rcfw *rcfw = res->rcfw; struct cmdq_destroy_qp req; struct creq_destroy_qp_resp resp; - unsigned long flags; u16 cmd_flags = 0; int rc; @@ -1437,19 +1396,12 @@ int bnxt_qplib_destroy_qp(struct bnxt_qplib_res *res, return rc; } - /* Must walk the associated CQs to nullified the QP ptr */ - spin_lock_irqsave(&qp->scq->hwq.lock, flags); - - __clean_cq(qp->scq, (u64)(unsigned long)qp); - - if (qp->rcq && qp->rcq != qp->scq) { - spin_lock(&qp->rcq->hwq.lock); - __clean_cq(qp->rcq, (u64)(unsigned long)qp); - spin_unlock(&qp->rcq->hwq.lock); - } - - spin_unlock_irqrestore(&qp->scq->hwq.lock, flags); + return 0; +} +void bnxt_qplib_free_qp_res(struct bnxt_qplib_res *res, + struct bnxt_qplib_qp *qp) +{ bnxt_qplib_free_qp_hdr_buf(res, qp); bnxt_qplib_free_hwq(res->pdev, &qp->sq.hwq); kfree(qp->sq.swq); @@ -1462,7 +1414,6 @@ int bnxt_qplib_destroy_qp(struct bnxt_qplib_res *res, if (qp->orrq.max_elements) bnxt_qplib_free_hwq(res->pdev, &qp->orrq); - return 0; } void *bnxt_qplib_get_qp1_sq_buf(struct bnxt_qplib_qp *qp, @@ -2116,9 +2067,6 @@ void bnxt_qplib_mark_qp_error(void *qp_handle) /* Must block new posting of SQ and RQ */ qp->state = CMDQ_MODIFY_QP_NEW_STATE_ERR; bnxt_qplib_cancel_phantom_processing(qp); - - /* Add qp to flush list of the CQ */ - __bnxt_qplib_add_flush_qp(qp); } /* Note: SQE is valid from sw_sq_cons up to cqe_sq_cons (exclusive) @@ -2294,9 +2242,9 @@ static int bnxt_qplib_cq_process_req(struct bnxt_qplib_cq *cq, sw_sq_cons, cqe->wr_id, cqe->status); cqe++; (*budget)--; - bnxt_qplib_lock_buddy_cq(qp, cq); bnxt_qplib_mark_qp_error(qp); - bnxt_qplib_unlock_buddy_cq(qp, cq); + /* Add qp to flush list of the CQ */ + bnxt_qplib_add_flush_qp(qp); } else { if (swq->flags & SQ_SEND_FLAGS_SIGNAL_COMP) { /* Before we complete, do WA 9060 */ @@ -2412,9 +2360,7 @@ static int bnxt_qplib_cq_process_res_rc(struct bnxt_qplib_cq *cq, if (hwcqe->status != CQ_RES_RC_STATUS_OK) { qp->state = CMDQ_MODIFY_QP_NEW_STATE_ERR; /* Add qp to flush list of the CQ */ - bnxt_qplib_lock_buddy_cq(qp, cq); - __bnxt_qplib_add_flush_qp(qp); - bnxt_qplib_unlock_buddy_cq(qp, cq); + bnxt_qplib_add_flush_qp(qp); } } @@ -2498,9 +2444,7 @@ static int bnxt_qplib_cq_process_res_ud(struct bnxt_qplib_cq *cq, if (hwcqe->status != CQ_RES_RC_STATUS_OK) { qp->state = CMDQ_MODIFY_QP_NEW_STATE_ERR; /* Add qp to flush list of the CQ */ - bnxt_qplib_lock_buddy_cq(qp, cq); - __bnxt_qplib_add_flush_qp(qp); - bnxt_qplib_unlock_buddy_cq(qp, cq); + bnxt_qplib_add_flush_qp(qp); } } done: @@ -2510,11 +2454,9 @@ done: bool bnxt_qplib_is_cq_empty(struct bnxt_qplib_cq *cq) { struct cq_base *hw_cqe, **hw_cqe_ptr; - unsigned long flags; u32 sw_cons, raw_cons; bool rc = true; - spin_lock_irqsave(&cq->hwq.lock, flags); raw_cons = cq->hwq.cons; sw_cons = HWQ_CMP(raw_cons, &cq->hwq); hw_cqe_ptr = (struct cq_base **)cq->hwq.pbl_ptr; @@ -2522,7 +2464,6 @@ bool bnxt_qplib_is_cq_empty(struct bnxt_qplib_cq *cq) /* Check for Valid bit. If the CQE is valid, return false */ rc = !CQE_CMP_VALID(hw_cqe, raw_cons, cq->hwq.max_elements); - spin_unlock_irqrestore(&cq->hwq.lock, flags); return rc; } @@ -2611,9 +2552,7 @@ static int bnxt_qplib_cq_process_res_raweth_qp1(struct bnxt_qplib_cq *cq, if (hwcqe->status != CQ_RES_RC_STATUS_OK) { qp->state = CMDQ_MODIFY_QP_NEW_STATE_ERR; /* Add qp to flush list of the CQ */ - bnxt_qplib_lock_buddy_cq(qp, cq); - __bnxt_qplib_add_flush_qp(qp); - bnxt_qplib_unlock_buddy_cq(qp, cq); + bnxt_qplib_add_flush_qp(qp); } } @@ -2728,9 +2667,7 @@ do_rq: */ /* Add qp to flush list of the CQ */ - bnxt_qplib_lock_buddy_cq(qp, cq); - __bnxt_qplib_add_flush_qp(qp); - bnxt_qplib_unlock_buddy_cq(qp, cq); + bnxt_qplib_add_flush_qp(qp); done: return rc; } @@ -2759,7 +2696,7 @@ int bnxt_qplib_process_flush_list(struct bnxt_qplib_cq *cq, u32 budget = num_cqes; unsigned long flags; - spin_lock_irqsave(&cq->hwq.lock, flags); + spin_lock_irqsave(&cq->flush_lock, flags); list_for_each_entry(qp, &cq->sqf_head, sq_flush) { dev_dbg(&cq->hwq.pdev->dev, "QPLIB: FP: Flushing SQ QP= %p", @@ -2773,7 +2710,7 @@ int bnxt_qplib_process_flush_list(struct bnxt_qplib_cq *cq, qp); __flush_rq(&qp->rq, qp, &cqe, &budget); } - spin_unlock_irqrestore(&cq->hwq.lock, flags); + spin_unlock_irqrestore(&cq->flush_lock, flags); return num_cqes - budget; } @@ -2782,11 +2719,9 @@ int bnxt_qplib_poll_cq(struct bnxt_qplib_cq *cq, struct bnxt_qplib_cqe *cqe, int num_cqes, struct bnxt_qplib_qp **lib_qp) { struct cq_base *hw_cqe, **hw_cqe_ptr; - unsigned long flags; u32 sw_cons, raw_cons; int budget, rc = 0; - spin_lock_irqsave(&cq->hwq.lock, flags); raw_cons = cq->hwq.cons; budget = num_cqes; @@ -2862,20 +2797,15 @@ int bnxt_qplib_poll_cq(struct bnxt_qplib_cq *cq, struct bnxt_qplib_cqe *cqe, bnxt_qplib_arm_cq(cq, DBR_DBR_TYPE_CQ); } exit: - spin_unlock_irqrestore(&cq->hwq.lock, flags); return num_cqes - budget; } void bnxt_qplib_req_notify_cq(struct bnxt_qplib_cq *cq, u32 arm_type) { - unsigned long flags; - - spin_lock_irqsave(&cq->hwq.lock, flags); if (arm_type) bnxt_qplib_arm_cq(cq, arm_type); /* Using cq->arm_state variable to track whether to issue cq handler */ atomic_set(&cq->arm_state, 1); - spin_unlock_irqrestore(&cq->hwq.lock, flags); } void bnxt_qplib_flush_cqn_wq(struct bnxt_qplib_qp *qp) diff --git a/drivers/infiniband/hw/bnxt_re/qplib_fp.h b/drivers/infiniband/hw/bnxt_re/qplib_fp.h index 211b27a8f9e2..ade9f13c0fd1 100644 --- a/drivers/infiniband/hw/bnxt_re/qplib_fp.h +++ b/drivers/infiniband/hw/bnxt_re/qplib_fp.h @@ -389,6 +389,18 @@ struct bnxt_qplib_cq { struct list_head sqf_head, rqf_head; atomic_t arm_state; spinlock_t compl_lock; /* synch CQ handlers */ +/* Locking Notes: + * QP can move to error state from modify_qp, async error event or error + * CQE as part of poll_cq. When QP is moved to error state, it gets added + * to two flush lists, one each for SQ and RQ. + * Each flush list is protected by qplib_cq->flush_lock. Both scq and rcq + * flush_locks should be acquired when QP is moved to error. The control path + * operations(modify_qp and async error events) are synchronized with poll_cq + * using upper level CQ locks (bnxt_re_cq->cq_lock) of both SCQ and RCQ. + * The qplib_cq->flush_lock is required to synchronize two instances of poll_cq + * of the same QP while manipulating the flush list. + */ + spinlock_t flush_lock; /* QP flush management */ }; #define BNXT_QPLIB_MAX_IRRQE_ENTRY_SIZE sizeof(struct xrrq_irrq) @@ -478,6 +490,9 @@ int bnxt_qplib_create_qp(struct bnxt_qplib_res *res, struct bnxt_qplib_qp *qp); int bnxt_qplib_modify_qp(struct bnxt_qplib_res *res, struct bnxt_qplib_qp *qp); int bnxt_qplib_query_qp(struct bnxt_qplib_res *res, struct bnxt_qplib_qp *qp); int bnxt_qplib_destroy_qp(struct bnxt_qplib_res *res, struct bnxt_qplib_qp *qp); +void bnxt_qplib_clean_qp(struct bnxt_qplib_qp *qp); +void bnxt_qplib_free_qp_res(struct bnxt_qplib_res *res, + struct bnxt_qplib_qp *qp); void *bnxt_qplib_get_qp1_sq_buf(struct bnxt_qplib_qp *qp, struct bnxt_qplib_sge *sge); void *bnxt_qplib_get_qp1_rq_buf(struct bnxt_qplib_qp *qp, @@ -500,7 +515,6 @@ void bnxt_qplib_req_notify_cq(struct bnxt_qplib_cq *cq, u32 arm_type); void bnxt_qplib_free_nq(struct bnxt_qplib_nq *nq); int bnxt_qplib_alloc_nq(struct pci_dev *pdev, struct bnxt_qplib_nq *nq); void bnxt_qplib_add_flush_qp(struct bnxt_qplib_qp *qp); -void bnxt_qplib_del_flush_qp(struct bnxt_qplib_qp *qp); void bnxt_qplib_acquire_cq_locks(struct bnxt_qplib_qp *qp, unsigned long *flags); void bnxt_qplib_release_cq_locks(struct bnxt_qplib_qp *qp, diff --git a/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c b/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c index 8329ec6a7946..80027a494730 100644 --- a/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c +++ b/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c @@ -305,9 +305,8 @@ static int bnxt_qplib_process_qp_event(struct bnxt_qplib_rcfw *rcfw, err_event->res_err_state_reason); if (!qp) break; - bnxt_qplib_acquire_cq_locks(qp, &flags); bnxt_qplib_mark_qp_error(qp); - bnxt_qplib_release_cq_locks(qp, &flags); + rcfw->aeq_handler(rcfw, qp_event, qp); break; default: /* Command Response */ @@ -460,7 +459,11 @@ int bnxt_qplib_init_rcfw(struct bnxt_qplib_rcfw *rcfw, int rc; RCFW_CMD_PREP(req, INITIALIZE_FW, cmd_flags); - + /* Supply (log-base-2-of-host-page-size - base-page-shift) + * to bono to adjust the doorbell page sizes. + */ + req.log2_dbr_pg_size = cpu_to_le16(PAGE_SHIFT - + RCFW_DBR_BASE_PAGE_SHIFT); /* * VFs need not setup the HW context area, PF * shall setup this area for VF. Skipping the diff --git a/drivers/infiniband/hw/bnxt_re/qplib_rcfw.h b/drivers/infiniband/hw/bnxt_re/qplib_rcfw.h index 6bee6e3636ea..c7cce2e4185e 100644 --- a/drivers/infiniband/hw/bnxt_re/qplib_rcfw.h +++ b/drivers/infiniband/hw/bnxt_re/qplib_rcfw.h @@ -49,6 +49,7 @@ #define RCFW_COMM_SIZE 0x104 #define RCFW_DBR_PCI_BAR_REGION 2 +#define RCFW_DBR_BASE_PAGE_SHIFT 12 #define RCFW_CMD_PREP(req, CMD, cmd_flags) \ do { \ diff --git a/drivers/infiniband/hw/bnxt_re/qplib_sp.c b/drivers/infiniband/hw/bnxt_re/qplib_sp.c index c015c1861351..ee98e5efef84 100644 --- a/drivers/infiniband/hw/bnxt_re/qplib_sp.c +++ b/drivers/infiniband/hw/bnxt_re/qplib_sp.c @@ -52,18 +52,6 @@ const struct bnxt_qplib_gid bnxt_qplib_gid_zero = {{ 0, 0, 0, 0, 0, 0, 0, 0, /* Device */ -static bool bnxt_qplib_is_atomic_cap(struct bnxt_qplib_rcfw *rcfw) -{ - int rc; - u16 pcie_ctl2; - - rc = pcie_capability_read_word(rcfw->pdev, PCI_EXP_DEVCTL2, - &pcie_ctl2); - if (rc) - return false; - return !!(pcie_ctl2 & PCI_EXP_DEVCTL2_ATOMIC_REQ); -} - static void bnxt_qplib_query_version(struct bnxt_qplib_rcfw *rcfw, char *fw_ver) { @@ -151,7 +139,8 @@ int bnxt_qplib_get_dev_attr(struct bnxt_qplib_rcfw *rcfw, attr->max_pkey = le32_to_cpu(sb->max_pkeys); attr->max_inline_data = le32_to_cpu(sb->max_inline_data); - attr->l2_db_size = (sb->l2_db_space_size + 1) * PAGE_SIZE; + attr->l2_db_size = (sb->l2_db_space_size + 1) * + (0x01 << RCFW_DBR_BASE_PAGE_SHIFT); attr->max_sgid = le32_to_cpu(sb->max_gid); bnxt_qplib_query_version(rcfw, attr->fw_ver); @@ -165,7 +154,7 @@ int bnxt_qplib_get_dev_attr(struct bnxt_qplib_rcfw *rcfw, attr->tqm_alloc_reqs[i * 4 + 3] = *(++tqm_alloc); } - attr->is_atomic = bnxt_qplib_is_atomic_cap(rcfw); + attr->is_atomic = 0; bail: bnxt_qplib_rcfw_free_sbuf(rcfw, sbuf); return rc; diff --git a/drivers/infiniband/hw/bnxt_re/roce_hsi.h b/drivers/infiniband/hw/bnxt_re/roce_hsi.h index 2d7ea096a247..3e5a4f760d0e 100644 --- a/drivers/infiniband/hw/bnxt_re/roce_hsi.h +++ b/drivers/infiniband/hw/bnxt_re/roce_hsi.h @@ -1761,7 +1761,30 @@ struct cmdq_initialize_fw { #define CMDQ_INITIALIZE_FW_TIM_PG_SIZE_PG_2M (0x3UL << 4) #define CMDQ_INITIALIZE_FW_TIM_PG_SIZE_PG_8M (0x4UL << 4) #define CMDQ_INITIALIZE_FW_TIM_PG_SIZE_PG_1G (0x5UL << 4) - __le16 reserved16; + /* This value is (log-base-2-of-DBR-page-size - 12). + * 0 for 4KB. HW supported values are enumerated below. + */ + __le16 log2_dbr_pg_size; + #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_MASK 0xfUL + #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_SFT 0 + #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_4K 0x0UL + #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_8K 0x1UL + #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_16K 0x2UL + #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_32K 0x3UL + #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_64K 0x4UL + #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_128K 0x5UL + #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_256K 0x6UL + #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_512K 0x7UL + #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_1M 0x8UL + #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_2M 0x9UL + #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_4M 0xaUL + #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_8M 0xbUL + #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_16M 0xcUL + #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_32M 0xdUL + #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_64M 0xeUL + #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_128M 0xfUL + #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_LAST \ + CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_128M __le64 qpc_page_dir; __le64 mrw_page_dir; __le64 srq_page_dir; diff --git a/drivers/infiniband/hw/cxgb4/device.c b/drivers/infiniband/hw/cxgb4/device.c index 7a9d0de89d6a..e96771ddc9a7 100644 --- a/drivers/infiniband/hw/cxgb4/device.c +++ b/drivers/infiniband/hw/cxgb4/device.c @@ -1217,6 +1217,7 @@ static int c4iw_uld_state_change(void *handle, enum cxgb4_state new_state) if (ctx->dev) c4iw_remove(ctx); break; + case CXGB4_STATE_FATAL_ERROR: case CXGB4_STATE_START_RECOVERY: pr_info("%s: Fatal Error\n", pci_name(ctx->lldi.pdev)); if (ctx->dev) { diff --git a/drivers/infiniband/hw/hfi1/file_ops.c b/drivers/infiniband/hw/hfi1/file_ops.c index 41fafebe3b0d..da4aa1a95b11 100644 --- a/drivers/infiniband/hw/hfi1/file_ops.c +++ b/drivers/infiniband/hw/hfi1/file_ops.c @@ -1153,7 +1153,7 @@ static int get_ctxt_info(struct hfi1_filedata *fd, unsigned long arg, u32 len) cinfo.sdma_ring_size = fd->cq->nentries; cinfo.rcvegr_size = uctxt->egrbufs.rcvtid_size; - trace_hfi1_ctxt_info(uctxt->dd, uctxt->ctxt, fd->subctxt, cinfo); + trace_hfi1_ctxt_info(uctxt->dd, uctxt->ctxt, fd->subctxt, &cinfo); if (copy_to_user((void __user *)arg, &cinfo, len)) return -EFAULT; diff --git a/drivers/infiniband/hw/hfi1/trace_ctxts.h b/drivers/infiniband/hw/hfi1/trace_ctxts.h index 4eb4cc798035..e00c8a7d559c 100644 --- a/drivers/infiniband/hw/hfi1/trace_ctxts.h +++ b/drivers/infiniband/hw/hfi1/trace_ctxts.h @@ -106,7 +106,7 @@ TRACE_EVENT(hfi1_uctxtdata, TRACE_EVENT(hfi1_ctxt_info, TP_PROTO(struct hfi1_devdata *dd, unsigned int ctxt, unsigned int subctxt, - struct hfi1_ctxt_info cinfo), + struct hfi1_ctxt_info *cinfo), TP_ARGS(dd, ctxt, subctxt, cinfo), TP_STRUCT__entry(DD_DEV_ENTRY(dd) __field(unsigned int, ctxt) @@ -120,11 +120,11 @@ TRACE_EVENT(hfi1_ctxt_info, TP_fast_assign(DD_DEV_ASSIGN(dd); __entry->ctxt = ctxt; __entry->subctxt = subctxt; - __entry->egrtids = cinfo.egrtids; - __entry->rcvhdrq_cnt = cinfo.rcvhdrq_cnt; - __entry->rcvhdrq_size = cinfo.rcvhdrq_entsize; - __entry->sdma_ring_size = cinfo.sdma_ring_size; - __entry->rcvegr_size = cinfo.rcvegr_size; + __entry->egrtids = cinfo->egrtids; + __entry->rcvhdrq_cnt = cinfo->rcvhdrq_cnt; + __entry->rcvhdrq_size = cinfo->rcvhdrq_entsize; + __entry->sdma_ring_size = cinfo->sdma_ring_size; + __entry->rcvegr_size = cinfo->rcvegr_size; ), TP_printk("[%s] ctxt %u:%u " CINFO_FMT, __get_str(dev), diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c index db2ff352d75f..ec638778661c 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c @@ -4383,7 +4383,7 @@ err_dma_alloc_buf: eq->l0_dma = 0; if (mhop_num == 1) - for (i -= i; i >= 0; i--) + for (i -= 1; i >= 0; i--) dma_free_coherent(dev, buf_chk_sz, eq->buf[i], eq->buf_dma[i]); else if (mhop_num == 2) { diff --git a/drivers/infiniband/hw/mlx4/cq.c b/drivers/infiniband/hw/mlx4/cq.c index 9a566ee3ceff..82adc0d1d30e 100644 --- a/drivers/infiniband/hw/mlx4/cq.c +++ b/drivers/infiniband/hw/mlx4/cq.c @@ -601,6 +601,7 @@ static void use_tunnel_data(struct mlx4_ib_qp *qp, struct mlx4_ib_cq *cq, struct wc->dlid_path_bits = 0; if (is_eth) { + wc->slid = 0; wc->vlan_id = be16_to_cpu(hdr->tun.sl_vid); memcpy(&(wc->smac[0]), (char *)&hdr->tun.mac_31_0, 4); memcpy(&(wc->smac[4]), (char *)&hdr->tun.slid_mac_47_32, 2); @@ -851,7 +852,6 @@ repoll: } } - wc->slid = be16_to_cpu(cqe->rlid); g_mlpath_rqpn = be32_to_cpu(cqe->g_mlpath_rqpn); wc->src_qp = g_mlpath_rqpn & 0xffffff; wc->dlid_path_bits = (g_mlpath_rqpn >> 24) & 0x7f; @@ -860,6 +860,7 @@ repoll: wc->wc_flags |= mlx4_ib_ipoib_csum_ok(cqe->status, cqe->checksum) ? IB_WC_IP_CSUM_OK : 0; if (is_eth) { + wc->slid = 0; wc->sl = be16_to_cpu(cqe->sl_vid) >> 13; if (be32_to_cpu(cqe->vlan_my_qpn) & MLX4_CQE_CVLAN_PRESENT_MASK) { @@ -871,6 +872,7 @@ repoll: memcpy(wc->smac, cqe->smac, ETH_ALEN); wc->wc_flags |= (IB_WC_WITH_VLAN | IB_WC_WITH_SMAC); } else { + wc->slid = be16_to_cpu(cqe->rlid); wc->sl = be16_to_cpu(cqe->sl_vid) >> 12; wc->vlan_id = 0xffff; } diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c index 8d2ee9322f2e..5a0e4fc4785a 100644 --- a/drivers/infiniband/hw/mlx4/main.c +++ b/drivers/infiniband/hw/mlx4/main.c @@ -219,8 +219,6 @@ static int mlx4_ib_update_gids_v1_v2(struct gid_entry *gids, gid_tbl[i].version = 2; if (!ipv6_addr_v4mapped((struct in6_addr *)&gids[i].gid)) gid_tbl[i].type = 1; - else - memset(&gid_tbl[i].gid, 0, 12); } } @@ -366,8 +364,13 @@ static int mlx4_ib_del_gid(struct ib_device *device, if (!gids) { ret = -ENOMEM; } else { - for (i = 0; i < MLX4_MAX_PORT_GIDS; i++) - memcpy(&gids[i].gid, &port_gid_table->gids[i].gid, sizeof(union ib_gid)); + for (i = 0; i < MLX4_MAX_PORT_GIDS; i++) { + memcpy(&gids[i].gid, + &port_gid_table->gids[i].gid, + sizeof(union ib_gid)); + gids[i].gid_type = + port_gid_table->gids[i].gid_type; + } } } spin_unlock_bh(&iboe->lock); diff --git a/drivers/infiniband/hw/mlx5/Makefile b/drivers/infiniband/hw/mlx5/Makefile index bc6299697dda..d42b922bede8 100644 --- a/drivers/infiniband/hw/mlx5/Makefile +++ b/drivers/infiniband/hw/mlx5/Makefile @@ -2,3 +2,4 @@ obj-$(CONFIG_MLX5_INFINIBAND) += mlx5_ib.o mlx5_ib-y := main.o cq.o doorbell.o qp.o mem.o srq.o mr.o ah.o mad.o gsi.o ib_virt.o cmd.o cong.o mlx5_ib-$(CONFIG_INFINIBAND_ON_DEMAND_PAGING) += odp.o +mlx5_ib-$(CONFIG_MLX5_ESWITCH) += ib_rep.o diff --git a/drivers/infiniband/hw/mlx5/cq.c b/drivers/infiniband/hw/mlx5/cq.c index 5b974fb97611..77d257ec899b 100644 --- a/drivers/infiniband/hw/mlx5/cq.c +++ b/drivers/infiniband/hw/mlx5/cq.c @@ -64,14 +64,9 @@ static void mlx5_ib_cq_event(struct mlx5_core_cq *mcq, enum mlx5_event type) } } -static void *get_cqe_from_buf(struct mlx5_ib_cq_buf *buf, int n, int size) -{ - return mlx5_buf_offset(&buf->buf, n * size); -} - static void *get_cqe(struct mlx5_ib_cq *cq, int n) { - return get_cqe_from_buf(&cq->buf, n, cq->mcq.cqe_sz); + return mlx5_frag_buf_get_wqe(&cq->buf.fbc, n); } static u8 sw_ownership_bit(int n, int nent) @@ -226,7 +221,6 @@ static void handle_responder(struct ib_wc *wc, struct mlx5_cqe64 *cqe, wc->ex.invalidate_rkey = be32_to_cpu(cqe->imm_inval_pkey); break; } - wc->slid = be16_to_cpu(cqe->slid); wc->src_qp = be32_to_cpu(cqe->flags_rqpn) & 0xffffff; wc->dlid_path_bits = cqe->ml_path; g = (be32_to_cpu(cqe->flags_rqpn) >> 28) & 3; @@ -241,10 +235,12 @@ static void handle_responder(struct ib_wc *wc, struct mlx5_cqe64 *cqe, } if (ll != IB_LINK_LAYER_ETHERNET) { + wc->slid = be16_to_cpu(cqe->slid); wc->sl = (be32_to_cpu(cqe->flags_rqpn) >> 24) & 0xf; return; } + wc->slid = 0; vlan_present = cqe->l4_l3_hdr_type & 0x1; roce_packet_type = (be32_to_cpu(cqe->flags_rqpn) >> 24) & 0x3; if (vlan_present) { @@ -271,14 +267,8 @@ static void handle_responder(struct ib_wc *wc, struct mlx5_cqe64 *cqe, static void dump_cqe(struct mlx5_ib_dev *dev, struct mlx5_err_cqe *cqe) { - __be32 *p = (__be32 *)cqe; - int i; - mlx5_ib_warn(dev, "dump error cqe\n"); - for (i = 0; i < sizeof(*cqe) / 16; i++, p += 4) - pr_info("%08x %08x %08x %08x\n", be32_to_cpu(p[0]), - be32_to_cpu(p[1]), be32_to_cpu(p[2]), - be32_to_cpu(p[3])); + mlx5_dump_err_cqe(dev->mdev, cqe); } static void mlx5_handle_error_cqe(struct mlx5_ib_dev *dev, @@ -403,7 +393,7 @@ static void handle_atomics(struct mlx5_ib_qp *qp, struct mlx5_cqe64 *cqe64, static void free_cq_buf(struct mlx5_ib_dev *dev, struct mlx5_ib_cq_buf *buf) { - mlx5_buf_free(dev->mdev, &buf->buf); + mlx5_frag_buf_free(dev->mdev, &buf->fbc.frag_buf); } static void get_sig_err_item(struct mlx5_sig_err_cqe *cqe, @@ -724,12 +714,25 @@ int mlx5_ib_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags flags) return ret; } -static int alloc_cq_buf(struct mlx5_ib_dev *dev, struct mlx5_ib_cq_buf *buf, - int nent, int cqe_size) +static int alloc_cq_frag_buf(struct mlx5_ib_dev *dev, + struct mlx5_ib_cq_buf *buf, + int nent, + int cqe_size) { + struct mlx5_frag_buf_ctrl *c = &buf->fbc; + struct mlx5_frag_buf *frag_buf = &c->frag_buf; + u32 cqc_buff[MLX5_ST_SZ_DW(cqc)] = {0}; int err; - err = mlx5_buf_alloc(dev->mdev, nent * cqe_size, &buf->buf); + MLX5_SET(cqc, cqc_buff, log_cq_size, ilog2(cqe_size)); + MLX5_SET(cqc, cqc_buff, cqe_sz, (cqe_size == 128) ? 1 : 0); + + mlx5_core_init_cq_frag_buf(&buf->fbc, cqc_buff); + + err = mlx5_frag_buf_alloc_node(dev->mdev, + nent * cqe_size, + frag_buf, + dev->mdev->priv.numa_node); if (err) return err; @@ -862,14 +865,15 @@ static void destroy_cq_user(struct mlx5_ib_cq *cq, struct ib_ucontext *context) ib_umem_release(cq->buf.umem); } -static void init_cq_buf(struct mlx5_ib_cq *cq, struct mlx5_ib_cq_buf *buf) +static void init_cq_frag_buf(struct mlx5_ib_cq *cq, + struct mlx5_ib_cq_buf *buf) { int i; void *cqe; struct mlx5_cqe64 *cqe64; for (i = 0; i < buf->nent; i++) { - cqe = get_cqe_from_buf(buf, i, buf->cqe_size); + cqe = get_cqe(cq, i); cqe64 = buf->cqe_size == 64 ? cqe : cqe + 64; cqe64->op_own = MLX5_CQE_INVALID << 4; } @@ -891,14 +895,15 @@ static int create_cq_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_cq *cq, cq->mcq.arm_db = cq->db.db + 1; cq->mcq.cqe_sz = cqe_size; - err = alloc_cq_buf(dev, &cq->buf, entries, cqe_size); + err = alloc_cq_frag_buf(dev, &cq->buf, entries, cqe_size); if (err) goto err_db; - init_cq_buf(cq, &cq->buf); + init_cq_frag_buf(cq, &cq->buf); *inlen = MLX5_ST_SZ_BYTES(create_cq_in) + - MLX5_FLD_SZ_BYTES(create_cq_in, pas[0]) * cq->buf.buf.npages; + MLX5_FLD_SZ_BYTES(create_cq_in, pas[0]) * + cq->buf.fbc.frag_buf.npages; *cqb = kvzalloc(*inlen, GFP_KERNEL); if (!*cqb) { err = -ENOMEM; @@ -906,11 +911,12 @@ static int create_cq_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_cq *cq, } pas = (__be64 *)MLX5_ADDR_OF(create_cq_in, *cqb, pas); - mlx5_fill_page_array(&cq->buf.buf, pas); + mlx5_fill_page_frag_array(&cq->buf.fbc.frag_buf, pas); cqc = MLX5_ADDR_OF(create_cq_in, *cqb, cq_context); MLX5_SET(cqc, cqc, log_page_size, - cq->buf.buf.page_shift - MLX5_ADAPTER_PAGE_SHIFT); + cq->buf.fbc.frag_buf.page_shift - + MLX5_ADAPTER_PAGE_SHIFT); *index = dev->mdev->priv.uar->index; @@ -1177,7 +1183,12 @@ static int resize_user(struct mlx5_ib_dev *dev, struct mlx5_ib_cq *cq, if (ucmd.reserved0 || ucmd.reserved1) return -EINVAL; - umem = ib_umem_get(context, ucmd.buf_addr, entries * ucmd.cqe_size, + /* check multiplication overflow */ + if (ucmd.cqe_size && SIZE_MAX / ucmd.cqe_size <= entries - 1) + return -EINVAL; + + umem = ib_umem_get(context, ucmd.buf_addr, + (size_t)ucmd.cqe_size * entries, IB_ACCESS_LOCAL_WRITE, 1); if (IS_ERR(umem)) { err = PTR_ERR(umem); @@ -1207,11 +1218,11 @@ static int resize_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_cq *cq, if (!cq->resize_buf) return -ENOMEM; - err = alloc_cq_buf(dev, cq->resize_buf, entries, cqe_size); + err = alloc_cq_frag_buf(dev, cq->resize_buf, entries, cqe_size); if (err) goto ex; - init_cq_buf(cq, cq->resize_buf); + init_cq_frag_buf(cq, cq->resize_buf); return 0; @@ -1256,9 +1267,8 @@ static int copy_resize_cqes(struct mlx5_ib_cq *cq) } while ((scqe64->op_own >> 4) != MLX5_CQE_RESIZE_CQ) { - dcqe = get_cqe_from_buf(cq->resize_buf, - (i + 1) & (cq->resize_buf->nent), - dsize); + dcqe = mlx5_frag_buf_get_wqe(&cq->resize_buf->fbc, + (i + 1) & cq->resize_buf->nent); dcqe64 = dsize == 64 ? dcqe : dcqe + 64; sw_own = sw_ownership_bit(i + 1, cq->resize_buf->nent); memcpy(dcqe, scqe, dsize); @@ -1324,8 +1334,11 @@ int mlx5_ib_resize_cq(struct ib_cq *ibcq, int entries, struct ib_udata *udata) cqe_size = 64; err = resize_kernel(dev, cq, entries, cqe_size); if (!err) { - npas = cq->resize_buf->buf.npages; - page_shift = cq->resize_buf->buf.page_shift; + struct mlx5_frag_buf_ctrl *c; + + c = &cq->resize_buf->fbc; + npas = c->frag_buf.npages; + page_shift = c->frag_buf.page_shift; } } @@ -1346,7 +1359,8 @@ int mlx5_ib_resize_cq(struct ib_cq *ibcq, int entries, struct ib_udata *udata) mlx5_ib_populate_pas(dev, cq->resize_umem, page_shift, pas, 0); else - mlx5_fill_page_array(&cq->resize_buf->buf, pas); + mlx5_fill_page_frag_array(&cq->resize_buf->fbc.frag_buf, + pas); MLX5_SET(modify_cq_in, in, modify_field_select_resize_field_select.resize_field_select.resize_field_select, diff --git a/drivers/infiniband/hw/mlx5/ib_rep.c b/drivers/infiniband/hw/mlx5/ib_rep.c new file mode 100644 index 000000000000..0e04fdddf670 --- /dev/null +++ b/drivers/infiniband/hw/mlx5/ib_rep.c @@ -0,0 +1,192 @@ +/* SPDX-License-Identifier: (GPL-2.0+ OR BSD-3-Clause) */ +/* + * Copyright (c) 2018 Mellanox Technologies. All rights reserved. + */ + +#include "ib_rep.h" + +static const struct mlx5_ib_profile rep_profile = { + STAGE_CREATE(MLX5_IB_STAGE_INIT, + mlx5_ib_stage_init_init, + mlx5_ib_stage_init_cleanup), + STAGE_CREATE(MLX5_IB_STAGE_FLOW_DB, + mlx5_ib_stage_rep_flow_db_init, + NULL), + STAGE_CREATE(MLX5_IB_STAGE_CAPS, + mlx5_ib_stage_caps_init, + NULL), + STAGE_CREATE(MLX5_IB_STAGE_NON_DEFAULT_CB, + mlx5_ib_stage_rep_non_default_cb, + NULL), + STAGE_CREATE(MLX5_IB_STAGE_ROCE, + mlx5_ib_stage_rep_roce_init, + mlx5_ib_stage_rep_roce_cleanup), + STAGE_CREATE(MLX5_IB_STAGE_DEVICE_RESOURCES, + mlx5_ib_stage_dev_res_init, + mlx5_ib_stage_dev_res_cleanup), + STAGE_CREATE(MLX5_IB_STAGE_COUNTERS, + mlx5_ib_stage_counters_init, + mlx5_ib_stage_counters_cleanup), + STAGE_CREATE(MLX5_IB_STAGE_BFREG, + mlx5_ib_stage_bfrag_init, + mlx5_ib_stage_bfrag_cleanup), + STAGE_CREATE(MLX5_IB_STAGE_PRE_IB_REG_UMR, + NULL, + mlx5_ib_stage_pre_ib_reg_umr_cleanup), + STAGE_CREATE(MLX5_IB_STAGE_IB_REG, + mlx5_ib_stage_ib_reg_init, + mlx5_ib_stage_ib_reg_cleanup), + STAGE_CREATE(MLX5_IB_STAGE_POST_IB_REG_UMR, + mlx5_ib_stage_post_ib_reg_umr_init, + NULL), + STAGE_CREATE(MLX5_IB_STAGE_CLASS_ATTR, + mlx5_ib_stage_class_attr_init, + NULL), +}; + +static int +mlx5_ib_nic_rep_load(struct mlx5_core_dev *dev, struct mlx5_eswitch_rep *rep) +{ + return 0; +} + +static void +mlx5_ib_nic_rep_unload(struct mlx5_eswitch_rep *rep) +{ + rep->rep_if[REP_IB].priv = NULL; +} + +static int +mlx5_ib_vport_rep_load(struct mlx5_core_dev *dev, struct mlx5_eswitch_rep *rep) +{ + struct mlx5_ib_dev *ibdev; + + ibdev = (struct mlx5_ib_dev *)ib_alloc_device(sizeof(*ibdev)); + if (!ibdev) + return -ENOMEM; + + ibdev->rep = rep; + ibdev->mdev = dev; + ibdev->num_ports = max(MLX5_CAP_GEN(dev, num_ports), + MLX5_CAP_GEN(dev, num_vhca_ports)); + if (!__mlx5_ib_add(ibdev, &rep_profile)) + return -EINVAL; + + rep->rep_if[REP_IB].priv = ibdev; + + return 0; +} + +static void +mlx5_ib_vport_rep_unload(struct mlx5_eswitch_rep *rep) +{ + struct mlx5_ib_dev *dev; + + if (!rep->rep_if[REP_IB].priv) + return; + + dev = mlx5_ib_rep_to_dev(rep); + __mlx5_ib_remove(dev, dev->profile, MLX5_IB_STAGE_MAX); + rep->rep_if[REP_IB].priv = NULL; +} + +static void *mlx5_ib_vport_get_proto_dev(struct mlx5_eswitch_rep *rep) +{ + return mlx5_ib_rep_to_dev(rep); +} + +static void mlx5_ib_rep_register_vf_vports(struct mlx5_ib_dev *dev) +{ + struct mlx5_eswitch *esw = dev->mdev->priv.eswitch; + int total_vfs = MLX5_TOTAL_VPORTS(dev->mdev); + int vport; + + for (vport = 1; vport < total_vfs; vport++) { + struct mlx5_eswitch_rep_if rep_if = {}; + + rep_if.load = mlx5_ib_vport_rep_load; + rep_if.unload = mlx5_ib_vport_rep_unload; + rep_if.get_proto_dev = mlx5_ib_vport_get_proto_dev; + mlx5_eswitch_register_vport_rep(esw, vport, &rep_if, REP_IB); + } +} + +static void mlx5_ib_rep_unregister_vf_vports(struct mlx5_ib_dev *dev) +{ + struct mlx5_eswitch *esw = dev->mdev->priv.eswitch; + int total_vfs = MLX5_TOTAL_VPORTS(dev->mdev); + int vport; + + for (vport = 1; vport < total_vfs; vport++) + mlx5_eswitch_unregister_vport_rep(esw, vport, REP_IB); +} + +void mlx5_ib_register_vport_reps(struct mlx5_ib_dev *dev) +{ + struct mlx5_eswitch *esw = dev->mdev->priv.eswitch; + struct mlx5_eswitch_rep_if rep_if = {}; + + rep_if.load = mlx5_ib_nic_rep_load; + rep_if.unload = mlx5_ib_nic_rep_unload; + rep_if.get_proto_dev = mlx5_ib_vport_get_proto_dev; + rep_if.priv = dev; + + mlx5_eswitch_register_vport_rep(esw, 0, &rep_if, REP_IB); + + mlx5_ib_rep_register_vf_vports(dev); +} + +void mlx5_ib_unregister_vport_reps(struct mlx5_ib_dev *dev) +{ + struct mlx5_eswitch *esw = dev->mdev->priv.eswitch; + + mlx5_ib_rep_unregister_vf_vports(dev); /* VFs vports */ + mlx5_eswitch_unregister_vport_rep(esw, 0, REP_IB); /* UPLINK PF*/ +} + +u8 mlx5_ib_eswitch_mode(struct mlx5_eswitch *esw) +{ + return mlx5_eswitch_mode(esw); +} + +struct mlx5_ib_dev *mlx5_ib_get_rep_ibdev(struct mlx5_eswitch *esw, + int vport_index) +{ + return mlx5_eswitch_get_proto_dev(esw, vport_index, REP_IB); +} + +struct net_device *mlx5_ib_get_rep_netdev(struct mlx5_eswitch *esw, + int vport_index) +{ + return mlx5_eswitch_get_proto_dev(esw, vport_index, REP_ETH); +} + +struct mlx5_ib_dev *mlx5_ib_get_uplink_ibdev(struct mlx5_eswitch *esw) +{ + return mlx5_eswitch_uplink_get_proto_dev(esw, REP_IB); +} + +struct mlx5_eswitch_rep *mlx5_ib_vport_rep(struct mlx5_eswitch *esw, int vport) +{ + return mlx5_eswitch_vport_rep(esw, vport); +} + +int create_flow_rule_vport_sq(struct mlx5_ib_dev *dev, + struct mlx5_ib_sq *sq) +{ + struct mlx5_flow_handle *flow_rule; + struct mlx5_eswitch *esw = dev->mdev->priv.eswitch; + + if (!dev->rep) + return 0; + + flow_rule = + mlx5_eswitch_add_send_to_vport_rule(esw, + dev->rep->vport, + sq->base.mqp.qpn); + if (IS_ERR(flow_rule)) + return PTR_ERR(flow_rule); + sq->flow_rule = flow_rule; + + return 0; +} diff --git a/drivers/infiniband/hw/mlx5/ib_rep.h b/drivers/infiniband/hw/mlx5/ib_rep.h new file mode 100644 index 000000000000..046fd942fd46 --- /dev/null +++ b/drivers/infiniband/hw/mlx5/ib_rep.h @@ -0,0 +1,72 @@ +/* SPDX-License-Identifier: (GPL-2.0+ OR BSD-3-Clause) */ +/* + * Copyright (c) 2018 Mellanox Technologies. All rights reserved. + */ + +#ifndef __MLX5_IB_REP_H__ +#define __MLX5_IB_REP_H__ + +#include <linux/mlx5/eswitch.h> +#include "mlx5_ib.h" + +#ifdef CONFIG_MLX5_ESWITCH +u8 mlx5_ib_eswitch_mode(struct mlx5_eswitch *esw); +struct mlx5_ib_dev *mlx5_ib_get_rep_ibdev(struct mlx5_eswitch *esw, + int vport_index); +struct mlx5_ib_dev *mlx5_ib_get_uplink_ibdev(struct mlx5_eswitch *esw); +struct mlx5_eswitch_rep *mlx5_ib_vport_rep(struct mlx5_eswitch *esw, + int vport_index); +void mlx5_ib_register_vport_reps(struct mlx5_ib_dev *dev); +void mlx5_ib_unregister_vport_reps(struct mlx5_ib_dev *dev); +int create_flow_rule_vport_sq(struct mlx5_ib_dev *dev, + struct mlx5_ib_sq *sq); +struct net_device *mlx5_ib_get_rep_netdev(struct mlx5_eswitch *esw, + int vport_index); +#else /* CONFIG_MLX5_ESWITCH */ +static inline u8 mlx5_ib_eswitch_mode(struct mlx5_eswitch *esw) +{ + return SRIOV_NONE; +} + +static inline +struct mlx5_ib_dev *mlx5_ib_get_rep_ibdev(struct mlx5_eswitch *esw, + int vport_index) +{ + return NULL; +} + +static inline +struct mlx5_ib_dev *mlx5_ib_get_uplink_ibdev(struct mlx5_eswitch *esw) +{ + return NULL; +} + +static inline +struct mlx5_eswitch_rep *mlx5_ib_vport_rep(struct mlx5_eswitch *esw, + int vport_index) +{ + return NULL; +} + +static inline void mlx5_ib_register_vport_reps(struct mlx5_ib_dev *dev) {} +static inline void mlx5_ib_unregister_vport_reps(struct mlx5_ib_dev *dev) {} +static inline int create_flow_rule_vport_sq(struct mlx5_ib_dev *dev, + struct mlx5_ib_sq *sq) +{ + return 0; +} + +static inline +struct net_device *mlx5_ib_get_rep_netdev(struct mlx5_eswitch *esw, + int vport_index) +{ + return NULL; +} +#endif + +static inline +struct mlx5_ib_dev *mlx5_ib_rep_to_dev(struct mlx5_eswitch_rep *rep) +{ + return (struct mlx5_ib_dev *)rep->rep_if[REP_IB].priv; +} +#endif /* __MLX5_IB_REP_H__ */ diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index 4236c8086820..071fd9a7b919 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -57,7 +57,9 @@ #include <linux/in.h> #include <linux/etherdevice.h> #include "mlx5_ib.h" +#include "ib_rep.h" #include "cmd.h" +#include <linux/mlx5/fs_helpers.h> #define DRIVER_NAME "mlx5_ib" #define DRIVER_VERSION "5.0-0" @@ -130,7 +132,7 @@ static int get_port_state(struct ib_device *ibdev, int ret; memset(&attr, 0, sizeof(attr)); - ret = mlx5_ib_query_port(ibdev, port_num, &attr); + ret = ibdev->query_port(ibdev, port_num, &attr); if (!ret) *state = attr.state; return ret; @@ -154,10 +156,19 @@ static int mlx5_netdev_event(struct notifier_block *this, case NETDEV_REGISTER: case NETDEV_UNREGISTER: write_lock(&roce->netdev_lock); - - if (ndev->dev.parent == &mdev->pdev->dev) - roce->netdev = (event == NETDEV_UNREGISTER) ? + if (ibdev->rep) { + struct mlx5_eswitch *esw = ibdev->mdev->priv.eswitch; + struct net_device *rep_ndev; + + rep_ndev = mlx5_ib_get_rep_netdev(esw, + ibdev->rep->vport); + if (rep_ndev == ndev) + roce->netdev = (event == NETDEV_UNREGISTER) ? NULL : ndev; + } else if (ndev->dev.parent == &ibdev->mdev->pdev->dev) { + roce->netdev = (event == NETDEV_UNREGISTER) ? + NULL : ndev; + } write_unlock(&roce->netdev_lock); break; @@ -245,12 +256,16 @@ struct mlx5_core_dev *mlx5_ib_get_native_port_mdev(struct mlx5_ib_dev *ibdev, struct mlx5_ib_multiport_info *mpi; struct mlx5_ib_port *port; + if (!mlx5_core_mp_enabled(ibdev->mdev) || + ll != IB_LINK_LAYER_ETHERNET) { + if (native_port_num) + *native_port_num = ib_port_num; + return ibdev->mdev; + } + if (native_port_num) *native_port_num = 1; - if (!mlx5_core_mp_enabled(ibdev->mdev) || ll != IB_LINK_LAYER_ETHERNET) - return ibdev->mdev; - port = &ibdev->port[ib_port_num - 1]; if (!port) return NULL; @@ -1268,6 +1283,22 @@ int mlx5_ib_query_port(struct ib_device *ibdev, u8 port, return ret; } +static int mlx5_ib_rep_query_port(struct ib_device *ibdev, u8 port, + struct ib_port_attr *props) +{ + int ret; + + /* Only link layer == ethernet is valid for representors */ + ret = mlx5_query_port_roce(ibdev, port, props); + if (ret || !props) + return ret; + + /* We don't support GIDS */ + props->gid_tbl_len = 0; + + return ret; +} + static int mlx5_ib_query_gid(struct ib_device *ibdev, u8 port, int index, union ib_gid *gid) { @@ -2286,11 +2317,9 @@ static void set_tos(void *outer_c, void *outer_v, u8 mask, u8 val) offsetof(typeof(filter), field) -\ sizeof(filter.field)) -#define IPV4_VERSION 4 -#define IPV6_VERSION 6 static int parse_flow_attr(struct mlx5_core_dev *mdev, u32 *match_c, u32 *match_v, const union ib_flow_spec *ib_spec, - u32 *tag_id, bool *is_drop) + struct mlx5_flow_act *action) { void *misc_params_c = MLX5_ADDR_OF(fte_match_param, match_c, misc_parameters); @@ -2373,7 +2402,7 @@ static int parse_flow_attr(struct mlx5_core_dev *mdev, u32 *match_c, MLX5_SET(fte_match_set_lyr_2_4, headers_c, ip_version, 0xf); MLX5_SET(fte_match_set_lyr_2_4, headers_v, - ip_version, IPV4_VERSION); + ip_version, MLX5_FS_IPV4_VERSION); } else { MLX5_SET(fte_match_set_lyr_2_4, headers_c, ethertype, 0xffff); @@ -2412,7 +2441,7 @@ static int parse_flow_attr(struct mlx5_core_dev *mdev, u32 *match_c, MLX5_SET(fte_match_set_lyr_2_4, headers_c, ip_version, 0xf); MLX5_SET(fte_match_set_lyr_2_4, headers_v, - ip_version, IPV6_VERSION); + ip_version, MLX5_FS_IPV6_VERSION); } else { MLX5_SET(fte_match_set_lyr_2_4, headers_c, ethertype, 0xffff); @@ -2508,13 +2537,14 @@ static int parse_flow_attr(struct mlx5_core_dev *mdev, u32 *match_c, if (ib_spec->flow_tag.tag_id >= BIT(24)) return -EINVAL; - *tag_id = ib_spec->flow_tag.tag_id; + action->flow_tag = ib_spec->flow_tag.tag_id; + action->has_flow_tag = true; break; case IB_FLOW_SPEC_ACTION_DROP: if (FIELDS_NOT_SUPPORTED(ib_spec->drop, LAST_DROP_FIELD)) return -EOPNOTSUPP; - *is_drop = true; + action->action |= MLX5_FLOW_CONTEXT_ACTION_DROP; break; default: return -EINVAL; @@ -2631,7 +2661,7 @@ static int mlx5_ib_destroy_flow(struct ib_flow *flow_id) ibflow); struct mlx5_ib_flow_handler *iter, *tmp; - mutex_lock(&dev->flow_db.lock); + mutex_lock(&dev->flow_db->lock); list_for_each_entry_safe(iter, tmp, &handler->list, list) { mlx5_del_flow_rules(iter->rule); @@ -2642,7 +2672,7 @@ static int mlx5_ib_destroy_flow(struct ib_flow *flow_id) mlx5_del_flow_rules(handler->rule); put_flow_table(dev, handler->prio, true); - mutex_unlock(&dev->flow_db.lock); + mutex_unlock(&dev->flow_db->lock); kfree(handler); @@ -2691,7 +2721,7 @@ static struct mlx5_ib_flow_prio *get_flow_table(struct mlx5_ib_dev *dev, MLX5_FLOW_NAMESPACE_BYPASS); num_entries = MLX5_FS_MAX_ENTRIES; num_groups = MLX5_FS_MAX_TYPES; - prio = &dev->flow_db.prios[priority]; + prio = &dev->flow_db->prios[priority]; } else if (flow_attr->type == IB_FLOW_ATTR_ALL_DEFAULT || flow_attr->type == IB_FLOW_ATTR_MC_DEFAULT) { ns = mlx5_get_flow_namespace(dev->mdev, @@ -2699,7 +2729,7 @@ static struct mlx5_ib_flow_prio *get_flow_table(struct mlx5_ib_dev *dev, build_leftovers_ft_param(&priority, &num_entries, &num_groups); - prio = &dev->flow_db.prios[MLX5_IB_FLOW_LEFTOVERS_PRIO]; + prio = &dev->flow_db->prios[MLX5_IB_FLOW_LEFTOVERS_PRIO]; } else if (flow_attr->type == IB_FLOW_ATTR_SNIFFER) { if (!MLX5_CAP_FLOWTABLE(dev->mdev, allow_sniffer_and_nic_rx_shared_tir)) @@ -2709,7 +2739,7 @@ static struct mlx5_ib_flow_prio *get_flow_table(struct mlx5_ib_dev *dev, MLX5_FLOW_NAMESPACE_SNIFFER_RX : MLX5_FLOW_NAMESPACE_SNIFFER_TX); - prio = &dev->flow_db.sniffer[ft_type]; + prio = &dev->flow_db->sniffer[ft_type]; priority = 0; num_entries = 1; num_groups = 1; @@ -2767,13 +2797,11 @@ static struct mlx5_ib_flow_handler *_create_flow_rule(struct mlx5_ib_dev *dev, { struct mlx5_flow_table *ft = ft_prio->flow_table; struct mlx5_ib_flow_handler *handler; - struct mlx5_flow_act flow_act = {0}; + struct mlx5_flow_act flow_act = {.flow_tag = MLX5_FS_DEFAULT_FLOW_TAG}; struct mlx5_flow_spec *spec; struct mlx5_flow_destination *rule_dst = dst; const void *ib_flow = (const void *)flow_attr + sizeof(*flow_attr); unsigned int spec_index; - u32 flow_tag = MLX5_FS_DEFAULT_FLOW_TAG; - bool is_drop = false; int err = 0; int dest_num = 1; @@ -2792,7 +2820,7 @@ static struct mlx5_ib_flow_handler *_create_flow_rule(struct mlx5_ib_dev *dev, for (spec_index = 0; spec_index < flow_attr->num_of_specs; spec_index++) { err = parse_flow_attr(dev->mdev, spec->match_criteria, spec->match_value, - ib_flow, &flow_tag, &is_drop); + ib_flow, &flow_act); if (err < 0) goto free; @@ -2802,9 +2830,20 @@ static struct mlx5_ib_flow_handler *_create_flow_rule(struct mlx5_ib_dev *dev, if (!flow_is_multicast_only(flow_attr)) set_underlay_qp(dev, spec, underlay_qpn); + if (dev->rep) { + void *misc; + + misc = MLX5_ADDR_OF(fte_match_param, spec->match_value, + misc_parameters); + MLX5_SET(fte_match_set_misc, misc, source_port, + dev->rep->vport); + misc = MLX5_ADDR_OF(fte_match_param, spec->match_criteria, + misc_parameters); + MLX5_SET_TO_ONES(fte_match_set_misc, misc, source_port); + } + spec->match_criteria_enable = get_match_criteria_enable(spec->match_criteria); - if (is_drop) { - flow_act.action = MLX5_FLOW_CONTEXT_ACTION_DROP; + if (flow_act.action & MLX5_FLOW_CONTEXT_ACTION_DROP) { rule_dst = NULL; dest_num = 0; } else { @@ -2812,15 +2851,14 @@ static struct mlx5_ib_flow_handler *_create_flow_rule(struct mlx5_ib_dev *dev, MLX5_FLOW_CONTEXT_ACTION_FWD_NEXT_PRIO; } - if (flow_tag != MLX5_FS_DEFAULT_FLOW_TAG && + if (flow_act.has_flow_tag && (flow_attr->type == IB_FLOW_ATTR_ALL_DEFAULT || flow_attr->type == IB_FLOW_ATTR_MC_DEFAULT)) { mlx5_ib_warn(dev, "Flow tag %u and attribute type %x isn't allowed in leftovers\n", - flow_tag, flow_attr->type); + flow_act.flow_tag, flow_attr->type); err = -EINVAL; goto free; } - flow_act.flow_tag = flow_tag; handler->rule = mlx5_add_flow_rules(ft, spec, &flow_act, rule_dst, dest_num); @@ -2999,7 +3037,7 @@ static struct ib_flow *mlx5_ib_create_flow(struct ib_qp *qp, if (!dst) return ERR_PTR(-ENOMEM); - mutex_lock(&dev->flow_db.lock); + mutex_lock(&dev->flow_db->lock); ft_prio = get_flow_table(dev, flow_attr, MLX5_IB_FT_RX); if (IS_ERR(ft_prio)) { @@ -3048,7 +3086,7 @@ static struct ib_flow *mlx5_ib_create_flow(struct ib_qp *qp, goto destroy_ft; } - mutex_unlock(&dev->flow_db.lock); + mutex_unlock(&dev->flow_db->lock); kfree(dst); return &handler->ibflow; @@ -3058,7 +3096,7 @@ destroy_ft: if (ft_prio_tx) put_flow_table(dev, ft_prio_tx, false); unlock: - mutex_unlock(&dev->flow_db.lock); + mutex_unlock(&dev->flow_db->lock); kfree(dst); kfree(handler); return ERR_PTR(err); @@ -3263,7 +3301,7 @@ static void mlx5_ib_handle_event(struct work_struct *_work) struct mlx5_ib_dev *ibdev; struct ib_event ibev; bool fatal = false; - u8 port = 0; + u8 port = (u8)work->param; if (mlx5_core_is_mp_slave(work->dev)) { ibdev = mlx5_ib_get_ibdev_from_mpi(work->context); @@ -3283,8 +3321,6 @@ static void mlx5_ib_handle_event(struct work_struct *_work) case MLX5_DEV_EVENT_PORT_UP: case MLX5_DEV_EVENT_PORT_DOWN: case MLX5_DEV_EVENT_PORT_INITIALIZED: - port = (u8)work->param; - /* In RoCE, port up/down events are handled in * mlx5_netdev_event(). */ @@ -3298,24 +3334,19 @@ static void mlx5_ib_handle_event(struct work_struct *_work) case MLX5_DEV_EVENT_LID_CHANGE: ibev.event = IB_EVENT_LID_CHANGE; - port = (u8)work->param; break; case MLX5_DEV_EVENT_PKEY_CHANGE: ibev.event = IB_EVENT_PKEY_CHANGE; - port = (u8)work->param; - schedule_work(&ibdev->devr.ports[port - 1].pkey_change_work); break; case MLX5_DEV_EVENT_GUID_CHANGE: ibev.event = IB_EVENT_GID_CHANGE; - port = (u8)work->param; break; case MLX5_DEV_EVENT_CLIENT_REREG: ibev.event = IB_EVENT_CLIENT_REREGISTER; - port = (u8)work->param; break; case MLX5_DEV_EVENT_DELAY_DROP_TIMEOUT: schedule_work(&ibdev->delay_drop.delay_drop_work); @@ -3327,7 +3358,7 @@ static void mlx5_ib_handle_event(struct work_struct *_work) ibev.device = &ibdev->ib_dev; ibev.element.port_num = port; - if (port < 1 || port > ibdev->num_ports) { + if (!rdma_is_port_valid(&ibdev->ib_dev, port)) { mlx5_ib_warn(ibdev, "warning: event on port %d\n", port); goto out; } @@ -3451,9 +3482,12 @@ static void destroy_umrc_res(struct mlx5_ib_dev *dev) if (err) mlx5_ib_warn(dev, "mr cache cleanup failed\n"); - mlx5_ib_destroy_qp(dev->umrc.qp); - ib_free_cq(dev->umrc.cq); - ib_dealloc_pd(dev->umrc.pd); + if (dev->umrc.qp) + mlx5_ib_destroy_qp(dev->umrc.qp); + if (dev->umrc.cq) + ib_free_cq(dev->umrc.cq); + if (dev->umrc.pd) + ib_dealloc_pd(dev->umrc.pd); } enum { @@ -3555,12 +3589,15 @@ static int create_umr_res(struct mlx5_ib_dev *dev) error_4: mlx5_ib_destroy_qp(qp); + dev->umrc.qp = NULL; error_3: ib_free_cq(cq); + dev->umrc.cq = NULL; error_2: ib_dealloc_pd(pd); + dev->umrc.pd = NULL; error_0: kfree(attr); @@ -3772,6 +3809,25 @@ static int mlx5_port_immutable(struct ib_device *ibdev, u8 port_num, return 0; } +static int mlx5_port_rep_immutable(struct ib_device *ibdev, u8 port_num, + struct ib_port_immutable *immutable) +{ + struct ib_port_attr attr; + int err; + + immutable->core_cap_flags = RDMA_CORE_PORT_RAW_PACKET; + + err = ib_query_port(ibdev, port_num, &attr); + if (err) + return err; + + immutable->pkey_tbl_len = attr.pkey_tbl_len; + immutable->gid_tbl_len = attr.gid_tbl_len; + immutable->core_cap_flags = RDMA_CORE_PORT_RAW_PACKET; + + return 0; +} + static void get_dev_fw_str(struct ib_device *ibdev, char *str) { struct mlx5_ib_dev *dev = @@ -3802,7 +3858,7 @@ static int mlx5_eth_lag_init(struct mlx5_ib_dev *dev) goto err_destroy_vport_lag; } - dev->flow_db.lag_demux_ft = ft; + dev->flow_db->lag_demux_ft = ft; return 0; err_destroy_vport_lag: @@ -3814,9 +3870,9 @@ static void mlx5_eth_lag_cleanup(struct mlx5_ib_dev *dev) { struct mlx5_core_dev *mdev = dev->mdev; - if (dev->flow_db.lag_demux_ft) { - mlx5_destroy_flow_table(dev->flow_db.lag_demux_ft); - dev->flow_db.lag_demux_ft = NULL; + if (dev->flow_db->lag_demux_ft) { + mlx5_destroy_flow_table(dev->flow_db->lag_demux_ft); + dev->flow_db->lag_demux_ft = NULL; mlx5_cmd_destroy_vport_lag(mdev); } @@ -3848,14 +3904,10 @@ static int mlx5_enable_eth(struct mlx5_ib_dev *dev, u8 port_num) { int err; - err = mlx5_add_netdev_notifier(dev, port_num); - if (err) - return err; - if (MLX5_CAP_GEN(dev->mdev, roce)) { err = mlx5_nic_vport_enable_roce(dev->mdev); if (err) - goto err_unregister_netdevice_notifier; + return err; } err = mlx5_eth_lag_init(dev); @@ -3868,8 +3920,6 @@ err_disable_roce: if (MLX5_CAP_GEN(dev->mdev, roce)) mlx5_nic_vport_disable_roce(dev->mdev); -err_unregister_netdevice_notifier: - mlx5_remove_netdev_notifier(dev, port_num); return err; } @@ -4503,7 +4553,7 @@ static void mlx5_ib_cleanup_multiport_master(struct mlx5_ib_dev *dev) mlx5_nic_vport_disable_roce(dev->mdev); } -static void mlx5_ib_stage_init_cleanup(struct mlx5_ib_dev *dev) +void mlx5_ib_stage_init_cleanup(struct mlx5_ib_dev *dev) { mlx5_ib_cleanup_multiport_master(dev); #ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING @@ -4512,7 +4562,7 @@ static void mlx5_ib_stage_init_cleanup(struct mlx5_ib_dev *dev) kfree(dev->port); } -static int mlx5_ib_stage_init_init(struct mlx5_ib_dev *dev) +int mlx5_ib_stage_init_init(struct mlx5_ib_dev *dev) { struct mlx5_core_dev *mdev = dev->mdev; const char *name; @@ -4534,8 +4584,6 @@ static int mlx5_ib_stage_init_init(struct mlx5_ib_dev *dev) goto err_free_port; if (!mlx5_core_mp_enabled(mdev)) { - int i; - for (i = 1; i <= dev->num_ports; i++) { err = get_port_caps(dev, i); if (err) @@ -4564,7 +4612,6 @@ static int mlx5_ib_stage_init_init(struct mlx5_ib_dev *dev) dev->mdev->priv.eq_table.num_comp_vectors; dev->ib_dev.dev.parent = &mdev->pdev->dev; - mutex_init(&dev->flow_db.lock); mutex_init(&dev->cap_mask_mutex); INIT_LIST_HEAD(&dev->qp_list); spin_lock_init(&dev->reset_flow_resource_lock); @@ -4585,7 +4632,38 @@ err_free_port: return -ENOMEM; } -static int mlx5_ib_stage_caps_init(struct mlx5_ib_dev *dev) +static int mlx5_ib_stage_flow_db_init(struct mlx5_ib_dev *dev) +{ + dev->flow_db = kzalloc(sizeof(*dev->flow_db), GFP_KERNEL); + + if (!dev->flow_db) + return -ENOMEM; + + mutex_init(&dev->flow_db->lock); + + return 0; +} + +int mlx5_ib_stage_rep_flow_db_init(struct mlx5_ib_dev *dev) +{ + struct mlx5_ib_dev *nic_dev; + + nic_dev = mlx5_ib_get_uplink_ibdev(dev->mdev->priv.eswitch); + + if (!nic_dev) + return -EINVAL; + + dev->flow_db = nic_dev->flow_db; + + return 0; +} + +static void mlx5_ib_stage_flow_db_cleanup(struct mlx5_ib_dev *dev) +{ + kfree(dev->flow_db); +} + +int mlx5_ib_stage_caps_init(struct mlx5_ib_dev *dev) { struct mlx5_core_dev *mdev = dev->mdev; int err; @@ -4626,7 +4704,6 @@ static int mlx5_ib_stage_caps_init(struct mlx5_ib_dev *dev) (1ull << IB_USER_VERBS_EX_CMD_MODIFY_CQ); dev->ib_dev.query_device = mlx5_ib_query_device; - dev->ib_dev.query_port = mlx5_ib_query_port; dev->ib_dev.get_link_layer = mlx5_ib_port_link_layer; dev->ib_dev.query_gid = mlx5_ib_query_gid; dev->ib_dev.add_gid = mlx5_ib_add_gid; @@ -4669,7 +4746,6 @@ static int mlx5_ib_stage_caps_init(struct mlx5_ib_dev *dev) dev->ib_dev.alloc_mr = mlx5_ib_alloc_mr; dev->ib_dev.map_mr_sg = mlx5_ib_map_mr_sg; dev->ib_dev.check_mr_status = mlx5_ib_check_mr_status; - dev->ib_dev.get_port_immutable = mlx5_port_immutable; dev->ib_dev.get_dev_fw_str = get_dev_fw_str; dev->ib_dev.get_vector_affinity = mlx5_ib_get_vector_affinity; if (MLX5_CAP_GEN(mdev, ipoib_enhanced_offloads)) @@ -4720,6 +4796,80 @@ static int mlx5_ib_stage_caps_init(struct mlx5_ib_dev *dev) return 0; } +static int mlx5_ib_stage_non_default_cb(struct mlx5_ib_dev *dev) +{ + dev->ib_dev.get_port_immutable = mlx5_port_immutable; + dev->ib_dev.query_port = mlx5_ib_query_port; + + return 0; +} + +int mlx5_ib_stage_rep_non_default_cb(struct mlx5_ib_dev *dev) +{ + dev->ib_dev.get_port_immutable = mlx5_port_rep_immutable; + dev->ib_dev.query_port = mlx5_ib_rep_query_port; + + return 0; +} + +static int mlx5_ib_stage_common_roce_init(struct mlx5_ib_dev *dev, + u8 port_num) +{ + int i; + + for (i = 0; i < dev->num_ports; i++) { + dev->roce[i].dev = dev; + dev->roce[i].native_port_num = i + 1; + dev->roce[i].last_port_state = IB_PORT_DOWN; + } + + dev->ib_dev.get_netdev = mlx5_ib_get_netdev; + dev->ib_dev.create_wq = mlx5_ib_create_wq; + dev->ib_dev.modify_wq = mlx5_ib_modify_wq; + dev->ib_dev.destroy_wq = mlx5_ib_destroy_wq; + dev->ib_dev.create_rwq_ind_table = mlx5_ib_create_rwq_ind_table; + dev->ib_dev.destroy_rwq_ind_table = mlx5_ib_destroy_rwq_ind_table; + + dev->ib_dev.uverbs_ex_cmd_mask |= + (1ull << IB_USER_VERBS_EX_CMD_CREATE_WQ) | + (1ull << IB_USER_VERBS_EX_CMD_MODIFY_WQ) | + (1ull << IB_USER_VERBS_EX_CMD_DESTROY_WQ) | + (1ull << IB_USER_VERBS_EX_CMD_CREATE_RWQ_IND_TBL) | + (1ull << IB_USER_VERBS_EX_CMD_DESTROY_RWQ_IND_TBL); + + return mlx5_add_netdev_notifier(dev, port_num); +} + +static void mlx5_ib_stage_common_roce_cleanup(struct mlx5_ib_dev *dev) +{ + u8 port_num = mlx5_core_native_port_num(dev->mdev) - 1; + + mlx5_remove_netdev_notifier(dev, port_num); +} + +int mlx5_ib_stage_rep_roce_init(struct mlx5_ib_dev *dev) +{ + struct mlx5_core_dev *mdev = dev->mdev; + enum rdma_link_layer ll; + int port_type_cap; + int err = 0; + u8 port_num; + + port_num = mlx5_core_native_port_num(dev->mdev) - 1; + port_type_cap = MLX5_CAP_GEN(mdev, port_type); + ll = mlx5_port_type_cap_to_rdma_ll(port_type_cap); + + if (ll == IB_LINK_LAYER_ETHERNET) + err = mlx5_ib_stage_common_roce_init(dev, port_num); + + return err; +} + +void mlx5_ib_stage_rep_roce_cleanup(struct mlx5_ib_dev *dev) +{ + mlx5_ib_stage_common_roce_cleanup(dev); +} + static int mlx5_ib_stage_roce_init(struct mlx5_ib_dev *dev) { struct mlx5_core_dev *mdev = dev->mdev; @@ -4727,37 +4877,26 @@ static int mlx5_ib_stage_roce_init(struct mlx5_ib_dev *dev) int port_type_cap; u8 port_num; int err; - int i; port_num = mlx5_core_native_port_num(dev->mdev) - 1; port_type_cap = MLX5_CAP_GEN(mdev, port_type); ll = mlx5_port_type_cap_to_rdma_ll(port_type_cap); if (ll == IB_LINK_LAYER_ETHERNET) { - for (i = 0; i < dev->num_ports; i++) { - dev->roce[i].dev = dev; - dev->roce[i].native_port_num = i + 1; - dev->roce[i].last_port_state = IB_PORT_DOWN; - } + err = mlx5_ib_stage_common_roce_init(dev, port_num); + if (err) + return err; - dev->ib_dev.get_netdev = mlx5_ib_get_netdev; - dev->ib_dev.create_wq = mlx5_ib_create_wq; - dev->ib_dev.modify_wq = mlx5_ib_modify_wq; - dev->ib_dev.destroy_wq = mlx5_ib_destroy_wq; - dev->ib_dev.create_rwq_ind_table = mlx5_ib_create_rwq_ind_table; - dev->ib_dev.destroy_rwq_ind_table = mlx5_ib_destroy_rwq_ind_table; - dev->ib_dev.uverbs_ex_cmd_mask |= - (1ull << IB_USER_VERBS_EX_CMD_CREATE_WQ) | - (1ull << IB_USER_VERBS_EX_CMD_MODIFY_WQ) | - (1ull << IB_USER_VERBS_EX_CMD_DESTROY_WQ) | - (1ull << IB_USER_VERBS_EX_CMD_CREATE_RWQ_IND_TBL) | - (1ull << IB_USER_VERBS_EX_CMD_DESTROY_RWQ_IND_TBL); err = mlx5_enable_eth(dev, port_num); if (err) - return err; + goto cleanup; } return 0; +cleanup: + mlx5_ib_stage_common_roce_cleanup(dev); + + return err; } static void mlx5_ib_stage_roce_cleanup(struct mlx5_ib_dev *dev) @@ -4773,16 +4912,16 @@ static void mlx5_ib_stage_roce_cleanup(struct mlx5_ib_dev *dev) if (ll == IB_LINK_LAYER_ETHERNET) { mlx5_disable_eth(dev); - mlx5_remove_netdev_notifier(dev, port_num); + mlx5_ib_stage_common_roce_cleanup(dev); } } -static int mlx5_ib_stage_dev_res_init(struct mlx5_ib_dev *dev) +int mlx5_ib_stage_dev_res_init(struct mlx5_ib_dev *dev) { return create_dev_resources(&dev->devr); } -static void mlx5_ib_stage_dev_res_cleanup(struct mlx5_ib_dev *dev) +void mlx5_ib_stage_dev_res_cleanup(struct mlx5_ib_dev *dev) { destroy_dev_resources(&dev->devr); } @@ -4794,7 +4933,7 @@ static int mlx5_ib_stage_odp_init(struct mlx5_ib_dev *dev) return mlx5_ib_odp_init_one(dev); } -static int mlx5_ib_stage_counters_init(struct mlx5_ib_dev *dev) +int mlx5_ib_stage_counters_init(struct mlx5_ib_dev *dev) { if (MLX5_CAP_GEN(dev->mdev, max_qp_cnt)) { dev->ib_dev.get_hw_stats = mlx5_ib_get_hw_stats; @@ -4806,7 +4945,7 @@ static int mlx5_ib_stage_counters_init(struct mlx5_ib_dev *dev) return 0; } -static void mlx5_ib_stage_counters_cleanup(struct mlx5_ib_dev *dev) +void mlx5_ib_stage_counters_cleanup(struct mlx5_ib_dev *dev) { if (MLX5_CAP_GEN(dev->mdev, max_qp_cnt)) mlx5_ib_dealloc_counters(dev); @@ -4837,7 +4976,7 @@ static void mlx5_ib_stage_uar_cleanup(struct mlx5_ib_dev *dev) mlx5_put_uars_page(dev->mdev, dev->mdev->priv.uar); } -static int mlx5_ib_stage_bfrag_init(struct mlx5_ib_dev *dev) +int mlx5_ib_stage_bfrag_init(struct mlx5_ib_dev *dev) { int err; @@ -4852,30 +4991,30 @@ static int mlx5_ib_stage_bfrag_init(struct mlx5_ib_dev *dev) return err; } -static void mlx5_ib_stage_bfrag_cleanup(struct mlx5_ib_dev *dev) +void mlx5_ib_stage_bfrag_cleanup(struct mlx5_ib_dev *dev) { mlx5_free_bfreg(dev->mdev, &dev->fp_bfreg); mlx5_free_bfreg(dev->mdev, &dev->bfreg); } -static int mlx5_ib_stage_ib_reg_init(struct mlx5_ib_dev *dev) +int mlx5_ib_stage_ib_reg_init(struct mlx5_ib_dev *dev) { return ib_register_device(&dev->ib_dev, NULL); } -static void mlx5_ib_stage_ib_reg_cleanup(struct mlx5_ib_dev *dev) +void mlx5_ib_stage_pre_ib_reg_umr_cleanup(struct mlx5_ib_dev *dev) { - ib_unregister_device(&dev->ib_dev); + destroy_umrc_res(dev); } -static int mlx5_ib_stage_umr_res_init(struct mlx5_ib_dev *dev) +void mlx5_ib_stage_ib_reg_cleanup(struct mlx5_ib_dev *dev) { - return create_umr_res(dev); + ib_unregister_device(&dev->ib_dev); } -static void mlx5_ib_stage_umr_res_cleanup(struct mlx5_ib_dev *dev) +int mlx5_ib_stage_post_ib_reg_umr_init(struct mlx5_ib_dev *dev) { - destroy_umrc_res(dev); + return create_umr_res(dev); } static int mlx5_ib_stage_delay_drop_init(struct mlx5_ib_dev *dev) @@ -4890,7 +5029,7 @@ static void mlx5_ib_stage_delay_drop_cleanup(struct mlx5_ib_dev *dev) cancel_delay_drop(dev); } -static int mlx5_ib_stage_class_attr_init(struct mlx5_ib_dev *dev) +int mlx5_ib_stage_class_attr_init(struct mlx5_ib_dev *dev) { int err; int i; @@ -4905,9 +5044,21 @@ static int mlx5_ib_stage_class_attr_init(struct mlx5_ib_dev *dev) return 0; } -static void __mlx5_ib_remove(struct mlx5_ib_dev *dev, - const struct mlx5_ib_profile *profile, - int stage) +static int mlx5_ib_stage_rep_reg_init(struct mlx5_ib_dev *dev) +{ + mlx5_ib_register_vport_reps(dev); + + return 0; +} + +static void mlx5_ib_stage_rep_reg_cleanup(struct mlx5_ib_dev *dev) +{ + mlx5_ib_unregister_vport_reps(dev); +} + +void __mlx5_ib_remove(struct mlx5_ib_dev *dev, + const struct mlx5_ib_profile *profile, + int stage) { /* Number of stages to cleanup */ while (stage) { @@ -4921,23 +5072,14 @@ static void __mlx5_ib_remove(struct mlx5_ib_dev *dev, static void *mlx5_ib_add_slave_port(struct mlx5_core_dev *mdev, u8 port_num); -static void *__mlx5_ib_add(struct mlx5_core_dev *mdev, - const struct mlx5_ib_profile *profile) +void *__mlx5_ib_add(struct mlx5_ib_dev *dev, + const struct mlx5_ib_profile *profile) { - struct mlx5_ib_dev *dev; int err; int i; printk_once(KERN_INFO "%s", mlx5_version); - dev = (struct mlx5_ib_dev *)ib_alloc_device(sizeof(*dev)); - if (!dev) - return NULL; - - dev->mdev = mdev; - dev->num_ports = max(MLX5_CAP_GEN(mdev, num_ports), - MLX5_CAP_GEN(mdev, num_vhca_ports)); - for (i = 0; i < MLX5_IB_STAGE_MAX; i++) { if (profile->stage[i].init) { err = profile->stage[i].init(dev); @@ -4961,9 +5103,15 @@ static const struct mlx5_ib_profile pf_profile = { STAGE_CREATE(MLX5_IB_STAGE_INIT, mlx5_ib_stage_init_init, mlx5_ib_stage_init_cleanup), + STAGE_CREATE(MLX5_IB_STAGE_FLOW_DB, + mlx5_ib_stage_flow_db_init, + mlx5_ib_stage_flow_db_cleanup), STAGE_CREATE(MLX5_IB_STAGE_CAPS, mlx5_ib_stage_caps_init, NULL), + STAGE_CREATE(MLX5_IB_STAGE_NON_DEFAULT_CB, + mlx5_ib_stage_non_default_cb, + NULL), STAGE_CREATE(MLX5_IB_STAGE_ROCE, mlx5_ib_stage_roce_init, mlx5_ib_stage_roce_cleanup), @@ -4985,12 +5133,15 @@ static const struct mlx5_ib_profile pf_profile = { STAGE_CREATE(MLX5_IB_STAGE_BFREG, mlx5_ib_stage_bfrag_init, mlx5_ib_stage_bfrag_cleanup), + STAGE_CREATE(MLX5_IB_STAGE_PRE_IB_REG_UMR, + NULL, + mlx5_ib_stage_pre_ib_reg_umr_cleanup), STAGE_CREATE(MLX5_IB_STAGE_IB_REG, mlx5_ib_stage_ib_reg_init, mlx5_ib_stage_ib_reg_cleanup), - STAGE_CREATE(MLX5_IB_STAGE_UMR_RESOURCES, - mlx5_ib_stage_umr_res_init, - mlx5_ib_stage_umr_res_cleanup), + STAGE_CREATE(MLX5_IB_STAGE_POST_IB_REG_UMR, + mlx5_ib_stage_post_ib_reg_umr_init, + NULL), STAGE_CREATE(MLX5_IB_STAGE_DELAY_DROP, mlx5_ib_stage_delay_drop_init, mlx5_ib_stage_delay_drop_cleanup), @@ -4999,6 +5150,51 @@ static const struct mlx5_ib_profile pf_profile = { NULL), }; +static const struct mlx5_ib_profile nic_rep_profile = { + STAGE_CREATE(MLX5_IB_STAGE_INIT, + mlx5_ib_stage_init_init, + mlx5_ib_stage_init_cleanup), + STAGE_CREATE(MLX5_IB_STAGE_FLOW_DB, + mlx5_ib_stage_flow_db_init, + mlx5_ib_stage_flow_db_cleanup), + STAGE_CREATE(MLX5_IB_STAGE_CAPS, + mlx5_ib_stage_caps_init, + NULL), + STAGE_CREATE(MLX5_IB_STAGE_NON_DEFAULT_CB, + mlx5_ib_stage_rep_non_default_cb, + NULL), + STAGE_CREATE(MLX5_IB_STAGE_ROCE, + mlx5_ib_stage_rep_roce_init, + mlx5_ib_stage_rep_roce_cleanup), + STAGE_CREATE(MLX5_IB_STAGE_DEVICE_RESOURCES, + mlx5_ib_stage_dev_res_init, + mlx5_ib_stage_dev_res_cleanup), + STAGE_CREATE(MLX5_IB_STAGE_COUNTERS, + mlx5_ib_stage_counters_init, + mlx5_ib_stage_counters_cleanup), + STAGE_CREATE(MLX5_IB_STAGE_UAR, + mlx5_ib_stage_uar_init, + mlx5_ib_stage_uar_cleanup), + STAGE_CREATE(MLX5_IB_STAGE_BFREG, + mlx5_ib_stage_bfrag_init, + mlx5_ib_stage_bfrag_cleanup), + STAGE_CREATE(MLX5_IB_STAGE_PRE_IB_REG_UMR, + NULL, + mlx5_ib_stage_pre_ib_reg_umr_cleanup), + STAGE_CREATE(MLX5_IB_STAGE_IB_REG, + mlx5_ib_stage_ib_reg_init, + mlx5_ib_stage_ib_reg_cleanup), + STAGE_CREATE(MLX5_IB_STAGE_POST_IB_REG_UMR, + mlx5_ib_stage_post_ib_reg_umr_init, + NULL), + STAGE_CREATE(MLX5_IB_STAGE_CLASS_ATTR, + mlx5_ib_stage_class_attr_init, + NULL), + STAGE_CREATE(MLX5_IB_STAGE_REP_REG, + mlx5_ib_stage_rep_reg_init, + mlx5_ib_stage_rep_reg_cleanup), +}; + static void *mlx5_ib_add_slave_port(struct mlx5_core_dev *mdev, u8 port_num) { struct mlx5_ib_multiport_info *mpi; @@ -5044,8 +5240,11 @@ static void *mlx5_ib_add_slave_port(struct mlx5_core_dev *mdev, u8 port_num) static void *mlx5_ib_add(struct mlx5_core_dev *mdev) { enum rdma_link_layer ll; + struct mlx5_ib_dev *dev; int port_type_cap; + printk_once(KERN_INFO "%s", mlx5_version); + port_type_cap = MLX5_CAP_GEN(mdev, port_type); ll = mlx5_port_type_cap_to_rdma_ll(port_type_cap); @@ -5055,7 +5254,22 @@ static void *mlx5_ib_add(struct mlx5_core_dev *mdev) return mlx5_ib_add_slave_port(mdev, port_num); } - return __mlx5_ib_add(mdev, &pf_profile); + dev = (struct mlx5_ib_dev *)ib_alloc_device(sizeof(*dev)); + if (!dev) + return NULL; + + dev->mdev = mdev; + dev->num_ports = max(MLX5_CAP_GEN(mdev, num_ports), + MLX5_CAP_GEN(mdev, num_vhca_ports)); + + if (MLX5_VPORT_MANAGER(mdev) && + mlx5_ib_eswitch_mode(mdev->priv.eswitch) == SRIOV_OFFLOADS) { + dev->rep = mlx5_ib_vport_rep(mdev->priv.eswitch, 0); + + return __mlx5_ib_add(dev, &nic_rep_profile); + } + + return __mlx5_ib_add(dev, &pf_profile); } static void mlx5_ib_remove(struct mlx5_core_dev *mdev, void *context) diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index 139385129973..c33bf1523d67 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -343,6 +343,7 @@ struct mlx5_ib_sq { struct mlx5_ib_wq *sq; struct mlx5_ib_ubuffer ubuffer; struct mlx5_db *doorbell; + struct mlx5_flow_handle *flow_rule; u32 tisn; u8 state; }; @@ -371,7 +372,7 @@ struct mlx5_ib_qp { struct mlx5_ib_rss_qp rss_qp; struct mlx5_ib_dct dct; }; - struct mlx5_buf buf; + struct mlx5_frag_buf buf; struct mlx5_db db; struct mlx5_ib_wq rq; @@ -413,7 +414,7 @@ struct mlx5_ib_qp { }; struct mlx5_ib_cq_buf { - struct mlx5_buf buf; + struct mlx5_frag_buf_ctrl fbc; struct ib_umem *umem; int cqe_size; int nent; @@ -495,7 +496,7 @@ struct mlx5_ib_wc { struct mlx5_ib_srq { struct ib_srq ibsrq; struct mlx5_core_srq msrq; - struct mlx5_buf buf; + struct mlx5_frag_buf buf; struct mlx5_db db; u64 *wrid; /* protect SRQ hanlding @@ -731,7 +732,9 @@ struct mlx5_ib_delay_drop { enum mlx5_ib_stages { MLX5_IB_STAGE_INIT, + MLX5_IB_STAGE_FLOW_DB, MLX5_IB_STAGE_CAPS, + MLX5_IB_STAGE_NON_DEFAULT_CB, MLX5_IB_STAGE_ROCE, MLX5_IB_STAGE_DEVICE_RESOURCES, MLX5_IB_STAGE_ODP, @@ -739,10 +742,12 @@ enum mlx5_ib_stages { MLX5_IB_STAGE_CONG_DEBUGFS, MLX5_IB_STAGE_UAR, MLX5_IB_STAGE_BFREG, + MLX5_IB_STAGE_PRE_IB_REG_UMR, MLX5_IB_STAGE_IB_REG, - MLX5_IB_STAGE_UMR_RESOURCES, + MLX5_IB_STAGE_POST_IB_REG_UMR, MLX5_IB_STAGE_DELAY_DROP, MLX5_IB_STAGE_CLASS_ATTR, + MLX5_IB_STAGE_REP_REG, MLX5_IB_STAGE_MAX, }; @@ -797,7 +802,7 @@ struct mlx5_ib_dev { struct srcu_struct mr_srcu; u32 null_mkey; #endif - struct mlx5_ib_flow_db flow_db; + struct mlx5_ib_flow_db *flow_db; /* protect resources needed as part of reset flow */ spinlock_t reset_flow_resource_lock; struct list_head qp_list; @@ -807,6 +812,7 @@ struct mlx5_ib_dev { struct mlx5_sq_bfreg fp_bfreg; struct mlx5_ib_delay_drop delay_drop; const struct mlx5_ib_profile *profile; + struct mlx5_eswitch_rep *rep; /* protect the user_td */ struct mutex lb_mutex; @@ -1049,6 +1055,31 @@ static inline void mlx5_odp_populate_klm(struct mlx5_klm *pklm, size_t offset, #endif /* CONFIG_INFINIBAND_ON_DEMAND_PAGING */ +/* Needed for rep profile */ +int mlx5_ib_stage_init_init(struct mlx5_ib_dev *dev); +void mlx5_ib_stage_init_cleanup(struct mlx5_ib_dev *dev); +int mlx5_ib_stage_rep_flow_db_init(struct mlx5_ib_dev *dev); +int mlx5_ib_stage_caps_init(struct mlx5_ib_dev *dev); +int mlx5_ib_stage_rep_non_default_cb(struct mlx5_ib_dev *dev); +int mlx5_ib_stage_rep_roce_init(struct mlx5_ib_dev *dev); +void mlx5_ib_stage_rep_roce_cleanup(struct mlx5_ib_dev *dev); +int mlx5_ib_stage_dev_res_init(struct mlx5_ib_dev *dev); +void mlx5_ib_stage_dev_res_cleanup(struct mlx5_ib_dev *dev); +int mlx5_ib_stage_counters_init(struct mlx5_ib_dev *dev); +void mlx5_ib_stage_counters_cleanup(struct mlx5_ib_dev *dev); +int mlx5_ib_stage_bfrag_init(struct mlx5_ib_dev *dev); +void mlx5_ib_stage_bfrag_cleanup(struct mlx5_ib_dev *dev); +void mlx5_ib_stage_pre_ib_reg_umr_cleanup(struct mlx5_ib_dev *dev); +int mlx5_ib_stage_ib_reg_init(struct mlx5_ib_dev *dev); +void mlx5_ib_stage_ib_reg_cleanup(struct mlx5_ib_dev *dev); +int mlx5_ib_stage_post_ib_reg_umr_init(struct mlx5_ib_dev *dev); +int mlx5_ib_stage_class_attr_init(struct mlx5_ib_dev *dev); +void __mlx5_ib_remove(struct mlx5_ib_dev *dev, + const struct mlx5_ib_profile *profile, + int stage); +void *__mlx5_ib_add(struct mlx5_ib_dev *dev, + const struct mlx5_ib_profile *profile); + int mlx5_ib_get_vf_config(struct ib_device *device, int vf, u8 port, struct ifla_vf_info *info); int mlx5_ib_set_vf_link_state(struct ib_device *device, int vf, diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c index 556e015678de..654bc31bc428 100644 --- a/drivers/infiniband/hw/mlx5/mr.c +++ b/drivers/infiniband/hw/mlx5/mr.c @@ -587,7 +587,7 @@ static void clean_keys(struct mlx5_ib_dev *dev, int c) static void mlx5_mr_cache_debugfs_cleanup(struct mlx5_ib_dev *dev) { - if (!mlx5_debugfs_root) + if (!mlx5_debugfs_root || dev->rep) return; debugfs_remove_recursive(dev->cache.root); @@ -600,7 +600,7 @@ static int mlx5_mr_cache_debugfs_init(struct mlx5_ib_dev *dev) struct mlx5_cache_ent *ent; int i; - if (!mlx5_debugfs_root) + if (!mlx5_debugfs_root || dev->rep) return 0; cache->root = debugfs_create_dir("mr_cache", dev->mdev->priv.dbg_root); @@ -690,6 +690,7 @@ int mlx5_mr_cache_init(struct mlx5_ib_dev *dev) MLX5_IB_UMR_OCTOWORD; ent->access_mode = MLX5_MKC_ACCESS_MODE_MTT; if ((dev->mdev->profile->mask & MLX5_PROF_MASK_MR_CACHE) && + !dev->rep && mlx5_core_is_pf(dev->mdev)) ent->limit = dev->mdev->profile->mr_cache[i].limit; else @@ -739,6 +740,9 @@ int mlx5_mr_cache_cleanup(struct mlx5_ib_dev *dev) { int i; + if (!dev->cache.wq) + return 0; + dev->cache.stopped = 1; flush_workqueue(dev->cache.wq); @@ -838,7 +842,8 @@ static int mr_umem_get(struct ib_pd *pd, u64 start, u64 length, *umem = ib_umem_get(pd->uobject->context, start, length, access_flags, 0); err = PTR_ERR_OR_ZERO(*umem); - if (err < 0) { + if (err) { + *umem = NULL; mlx5_ib_err(dev, "umem get failed (%d)\n", err); return err; } @@ -1415,6 +1420,7 @@ int mlx5_ib_rereg_user_mr(struct ib_mr *ib_mr, int flags, u64 start, if (err) { mlx5_ib_warn(dev, "Failed to rereg UMR\n"); ib_umem_release(mr->umem); + mr->umem = NULL; clean_mr(dev, mr); return err; } @@ -1498,14 +1504,11 @@ static int clean_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr) u32 key = mr->mmkey.key; err = destroy_mkey(dev, mr); - kfree(mr); if (err) { mlx5_ib_warn(dev, "failed to destroy mkey 0x%x (%d)\n", key, err); return err; } - } else { - mlx5_mr_cache_free(dev, mr); } return 0; @@ -1548,6 +1551,11 @@ static int dereg_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr) atomic_sub(npages, &dev->mdev->priv.reg_pages); } + if (!mr->allocated_from_cache) + kfree(mr); + else + mlx5_mr_cache_free(dev, mr); + return 0; } @@ -1816,7 +1824,6 @@ mlx5_ib_sg_to_klms(struct mlx5_ib_mr *mr, mr->ibmr.iova = sg_dma_address(sg) + sg_offset; mr->ibmr.length = 0; - mr->ndescs = sg_nents; for_each_sg(sgl, sg, sg_nents, i) { if (unlikely(i >= mr->max_descs)) @@ -1828,6 +1835,7 @@ mlx5_ib_sg_to_klms(struct mlx5_ib_mr *mr, sg_offset = 0; } + mr->ndescs = i; if (sg_offset_p) *sg_offset_p = sg_offset; diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c index 39d24bf694a8..0d0b0b8dad98 100644 --- a/drivers/infiniband/hw/mlx5/qp.c +++ b/drivers/infiniband/hw/mlx5/qp.c @@ -36,6 +36,7 @@ #include <rdma/ib_user_verbs.h> #include <linux/mlx5/fs.h> #include "mlx5_ib.h" +#include "ib_rep.h" /* not supported currently */ static int wq_signature; @@ -1082,6 +1083,13 @@ static void destroy_raw_packet_qp_tis(struct mlx5_ib_dev *dev, mlx5_core_destroy_tis(dev->mdev, sq->tisn); } +static void destroy_flow_rule_vport_sq(struct mlx5_ib_dev *dev, + struct mlx5_ib_sq *sq) +{ + if (sq->flow_rule) + mlx5_del_flow_rules(sq->flow_rule); +} + static int create_raw_packet_qp_sq(struct mlx5_ib_dev *dev, struct mlx5_ib_sq *sq, void *qpin, struct ib_pd *pd) @@ -1145,8 +1153,15 @@ static int create_raw_packet_qp_sq(struct mlx5_ib_dev *dev, if (err) goto err_umem; + err = create_flow_rule_vport_sq(dev, sq); + if (err) + goto err_flow; + return 0; +err_flow: + mlx5_core_destroy_sq_tracked(dev->mdev, &sq->base.mqp); + err_umem: ib_umem_release(sq->ubuffer.umem); sq->ubuffer.umem = NULL; @@ -1157,11 +1172,12 @@ err_umem: static void destroy_raw_packet_qp_sq(struct mlx5_ib_dev *dev, struct mlx5_ib_sq *sq) { + destroy_flow_rule_vport_sq(dev, sq); mlx5_core_destroy_sq_tracked(dev->mdev, &sq->base.mqp); ib_umem_release(sq->ubuffer.umem); } -static int get_rq_pas_size(void *qpc) +static size_t get_rq_pas_size(void *qpc) { u32 log_page_size = MLX5_GET(qpc, qpc, log_page_size) + 12; u32 log_rq_stride = MLX5_GET(qpc, qpc, log_rq_stride); @@ -1177,7 +1193,8 @@ static int get_rq_pas_size(void *qpc) } static int create_raw_packet_qp_rq(struct mlx5_ib_dev *dev, - struct mlx5_ib_rq *rq, void *qpin) + struct mlx5_ib_rq *rq, void *qpin, + size_t qpinlen) { struct mlx5_ib_qp *mqp = rq->base.container_mibqp; __be64 *pas; @@ -1186,9 +1203,12 @@ static int create_raw_packet_qp_rq(struct mlx5_ib_dev *dev, void *rqc; void *wq; void *qpc = MLX5_ADDR_OF(create_qp_in, qpin, qpc); - int inlen; + size_t rq_pas_size = get_rq_pas_size(qpc); + size_t inlen; int err; - u32 rq_pas_size = get_rq_pas_size(qpc); + + if (qpinlen < rq_pas_size + MLX5_BYTE_OFF(create_qp_in, pas)) + return -EINVAL; inlen = MLX5_ST_SZ_BYTES(create_rq_in) + rq_pas_size; in = kvzalloc(inlen, GFP_KERNEL); @@ -1263,6 +1283,10 @@ static int create_raw_packet_qp_tir(struct mlx5_ib_dev *dev, if (tunnel_offload_en) MLX5_SET(tirc, tirc, tunneled_offload_en, 1); + if (dev->rep) + MLX5_SET(tirc, tirc, self_lb_block, + MLX5_TIRC_SELF_LB_BLOCK_BLOCK_UNICAST_); + err = mlx5_core_create_tir(dev->mdev, in, inlen, &rq->tirn); kvfree(in); @@ -1277,7 +1301,7 @@ static void destroy_raw_packet_qp_tir(struct mlx5_ib_dev *dev, } static int create_raw_packet_qp(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp, - u32 *in, + u32 *in, size_t inlen, struct ib_pd *pd) { struct mlx5_ib_raw_packet_qp *raw_packet_qp = &qp->raw_packet_qp; @@ -1309,7 +1333,7 @@ static int create_raw_packet_qp(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp, rq->flags |= MLX5_IB_RQ_CVLAN_STRIPPING; if (qp->flags & MLX5_IB_QP_PCI_WRITE_END_PADDING) rq->flags |= MLX5_IB_RQ_PCI_WRITE_END_PADDING; - err = create_raw_packet_qp_rq(dev, rq, in); + err = create_raw_packet_qp_rq(dev, rq, in, inlen); if (err) goto err_destroy_sq; @@ -1554,6 +1578,10 @@ static int create_rss_raw_qp_tir(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp, MLX5_SET(rx_hash_field_select, hfso, selected_fields, selected_fields); create_tir: + if (dev->rep) + MLX5_SET(tirc, tirc, self_lb_block, + MLX5_TIRC_SELF_LB_BLOCK_BLOCK_UNICAST_); + err = mlx5_core_create_tir(dev->mdev, in, inlen, &qp->rss_qp.tirn); if (err) @@ -1584,6 +1612,7 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd, u32 uidx = MLX5_IB_DEFAULT_UIDX; struct mlx5_ib_create_qp ucmd; struct mlx5_ib_qp_base *base; + int mlx5_st; void *qpc; u32 *in; int err; @@ -1592,6 +1621,10 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd, spin_lock_init(&qp->sq.lock); spin_lock_init(&qp->rq.lock); + mlx5_st = to_mlx5_st(init_attr->qp_type); + if (mlx5_st < 0) + return -EINVAL; + if (init_attr->rwq_ind_tbl) { if (!udata) return -ENOSYS; @@ -1753,7 +1786,7 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd, qpc = MLX5_ADDR_OF(create_qp_in, in, qpc); - MLX5_SET(qpc, qpc, st, to_mlx5_st(init_attr->qp_type)); + MLX5_SET(qpc, qpc, st, mlx5_st); MLX5_SET(qpc, qpc, pm_state, MLX5_QP_PM_MIGRATED); if (init_attr->qp_type != MLX5_IB_QPT_REG_UMR) @@ -1867,11 +1900,16 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd, } } + if (inlen < 0) { + err = -EINVAL; + goto err; + } + if (init_attr->qp_type == IB_QPT_RAW_PACKET || qp->flags & MLX5_IB_QP_UNDERLAY) { qp->raw_packet_qp.sq.ubuffer.buf_addr = ucmd.sq_buf_addr; raw_packet_qp_copy_info(qp, &qp->raw_packet_qp); - err = create_raw_packet_qp(dev, qp, in, pd); + err = create_raw_packet_qp(dev, qp, in, inlen, pd); } else { err = mlx5_core_create_qp(dev->mdev, &base->mqp, in, inlen); } @@ -2129,7 +2167,6 @@ static struct ib_qp *mlx5_ib_create_dct(struct ib_pd *pd, struct ib_qp_init_attr *attr, struct mlx5_ib_create_qp *ucmd) { - struct mlx5_ib_dev *dev; struct mlx5_ib_qp *qp; int err = 0; u32 uidx = MLX5_IB_DEFAULT_UIDX; @@ -2138,8 +2175,6 @@ static struct ib_qp *mlx5_ib_create_dct(struct ib_pd *pd, if (!attr->srq || !attr->recv_cq) return ERR_PTR(-EINVAL); - dev = to_mdev(pd->device); - err = get_qp_user_index(to_mucontext(pd->uobject->context), ucmd, sizeof(*ucmd), &uidx); if (err) @@ -3095,8 +3130,10 @@ static int __mlx5_ib_modify_qp(struct ib_qp *ibqp, goto out; if (mlx5_cur >= MLX5_QP_NUM_STATE || mlx5_new >= MLX5_QP_NUM_STATE || - !optab[mlx5_cur][mlx5_new]) + !optab[mlx5_cur][mlx5_new]) { + err = -EINVAL; goto out; + } op = optab[mlx5_cur][mlx5_new]; optpar = ib_mask_to_mlx5_opt(attr_mask); @@ -4702,26 +4739,14 @@ static int query_raw_packet_qp_sq_state(struct mlx5_ib_dev *dev, struct mlx5_ib_sq *sq, u8 *sq_state) { - void *out; - void *sqc; - int inlen; int err; - inlen = MLX5_ST_SZ_BYTES(query_sq_out); - out = kvzalloc(inlen, GFP_KERNEL); - if (!out) - return -ENOMEM; - - err = mlx5_core_query_sq(dev->mdev, sq->base.mqp.qpn, out); + err = mlx5_core_query_sq_state(dev->mdev, sq->base.mqp.qpn, sq_state); if (err) goto out; - - sqc = MLX5_ADDR_OF(query_sq_out, out, sq_context); - *sq_state = MLX5_GET(sqc, sqc, state); sq->state = *sq_state; out: - kvfree(out); return err; } diff --git a/drivers/infiniband/hw/mlx5/srq.c b/drivers/infiniband/hw/mlx5/srq.c index 6d5fadad9090..3c7522d025f2 100644 --- a/drivers/infiniband/hw/mlx5/srq.c +++ b/drivers/infiniband/hw/mlx5/srq.c @@ -241,8 +241,8 @@ struct ib_srq *mlx5_ib_create_srq(struct ib_pd *pd, { struct mlx5_ib_dev *dev = to_mdev(pd->device); struct mlx5_ib_srq *srq; - int desc_size; - int buf_size; + size_t desc_size; + size_t buf_size; int err; struct mlx5_srq_attr in = {0}; __u32 max_srq_wqes = 1 << MLX5_CAP_GEN(dev->mdev, log_max_srq_sz); @@ -266,15 +266,18 @@ struct ib_srq *mlx5_ib_create_srq(struct ib_pd *pd, desc_size = sizeof(struct mlx5_wqe_srq_next_seg) + srq->msrq.max_gs * sizeof(struct mlx5_wqe_data_seg); + if (desc_size == 0 || srq->msrq.max_gs > desc_size) + return ERR_PTR(-EINVAL); desc_size = roundup_pow_of_two(desc_size); - desc_size = max_t(int, 32, desc_size); + desc_size = max_t(size_t, 32, desc_size); + if (desc_size < sizeof(struct mlx5_wqe_srq_next_seg)) + return ERR_PTR(-EINVAL); srq->msrq.max_avail_gather = (desc_size - sizeof(struct mlx5_wqe_srq_next_seg)) / sizeof(struct mlx5_wqe_data_seg); srq->msrq.wqe_shift = ilog2(desc_size); buf_size = srq->msrq.max * desc_size; - mlx5_ib_dbg(dev, "desc_size 0x%x, req wr 0x%x, srq size 0x%x, max_gs 0x%x, max_avail_gather 0x%x\n", - desc_size, init_attr->attr.max_wr, srq->msrq.max, srq->msrq.max_gs, - srq->msrq.max_avail_gather); + if (buf_size < desc_size) + return ERR_PTR(-EINVAL); in.type = init_attr->srq_type; if (pd->uobject) diff --git a/drivers/infiniband/hw/qedr/main.c b/drivers/infiniband/hw/qedr/main.c index db4bf97c0e15..f9a645c869ce 100644 --- a/drivers/infiniband/hw/qedr/main.c +++ b/drivers/infiniband/hw/qedr/main.c @@ -90,8 +90,8 @@ static struct net_device *qedr_get_netdev(struct ib_device *dev, u8 port_num) dev_hold(qdev->ndev); /* The HW vendor's device driver must guarantee - * that this function returns NULL before the net device reaches - * NETDEV_UNREGISTER_FINAL state. + * that this function returns NULL before the net device has finished + * NETDEV_UNREGISTER state. */ return qdev->ndev; } @@ -833,7 +833,8 @@ static struct qedr_dev *qedr_add(struct qed_dev *cdev, struct pci_dev *pdev, dev->num_cnq = dev->ops->rdma_get_min_cnq_msix(cdev); if (!dev->num_cnq) { - DP_ERR(dev, "not enough CNQ resources.\n"); + DP_ERR(dev, "Failed. At least one CNQ is required.\n"); + rc = -ENOMEM; goto init_err; } diff --git a/drivers/infiniband/hw/qedr/qedr_hsi_rdma.h b/drivers/infiniband/hw/qedr/qedr_hsi_rdma.h index 78b49002fbd2..b816c80df50b 100644 --- a/drivers/infiniband/hw/qedr/qedr_hsi_rdma.h +++ b/drivers/infiniband/hw/qedr/qedr_hsi_rdma.h @@ -45,7 +45,7 @@ struct rdma_cqe_responder { __le32 imm_data_or_inv_r_Key; __le32 length; __le32 imm_data_hi; - __le16 rq_cons; + __le16 rq_cons_or_srq_id; u8 flags; #define RDMA_CQE_RESPONDER_TOGGLE_BIT_MASK 0x1 #define RDMA_CQE_RESPONDER_TOGGLE_BIT_SHIFT 0 @@ -115,6 +115,7 @@ enum rdma_cqe_requester_status_enum { RDMA_CQE_REQ_STS_RNR_NAK_RETRY_CNT_ERR, RDMA_CQE_REQ_STS_TRANSPORT_RETRY_CNT_ERR, RDMA_CQE_REQ_STS_WORK_REQUEST_FLUSHED_ERR, + RDMA_CQE_REQ_STS_XRC_VOILATION_ERR, MAX_RDMA_CQE_REQUESTER_STATUS_ENUM }; @@ -136,6 +137,7 @@ enum rdma_cqe_type { RDMA_CQE_TYPE_REQUESTER, RDMA_CQE_TYPE_RESPONDER_RQ, RDMA_CQE_TYPE_RESPONDER_SRQ, + RDMA_CQE_TYPE_RESPONDER_XRC_SRQ, RDMA_CQE_TYPE_INVALID, MAX_RDMA_CQE_TYPE }; diff --git a/drivers/infiniband/hw/qedr/qedr_iw_cm.c b/drivers/infiniband/hw/qedr/qedr_iw_cm.c index 478b7317b80a..26dc374787f7 100644 --- a/drivers/infiniband/hw/qedr/qedr_iw_cm.c +++ b/drivers/infiniband/hw/qedr/qedr_iw_cm.c @@ -458,8 +458,7 @@ qedr_addr6_resolve(struct qedr_dev *dev, } return -EINVAL; } - neigh = dst_neigh_lookup(dst, &dst_in); - + neigh = dst_neigh_lookup(dst, &fl6.daddr); if (neigh) { rcu_read_lock(); if (neigh->nud_state & NUD_VALID) { @@ -494,10 +493,14 @@ int qedr_iw_connect(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param) qp = idr_find(&dev->qpidr, conn_param->qpn); - laddr = (struct sockaddr_in *)&cm_id->local_addr; - raddr = (struct sockaddr_in *)&cm_id->remote_addr; - laddr6 = (struct sockaddr_in6 *)&cm_id->local_addr; - raddr6 = (struct sockaddr_in6 *)&cm_id->remote_addr; + laddr = (struct sockaddr_in *)&cm_id->m_local_addr; + raddr = (struct sockaddr_in *)&cm_id->m_remote_addr; + laddr6 = (struct sockaddr_in6 *)&cm_id->m_local_addr; + raddr6 = (struct sockaddr_in6 *)&cm_id->m_remote_addr; + + DP_DEBUG(dev, QEDR_MSG_IWARP, "MAPPED %d %d\n", + ntohs(((struct sockaddr_in *)&cm_id->remote_addr)->sin_port), + ntohs(raddr->sin_port)); DP_DEBUG(dev, QEDR_MSG_IWARP, "Connect source address: %pISpc, remote address: %pISpc\n", @@ -599,8 +602,8 @@ int qedr_iw_create_listen(struct iw_cm_id *cm_id, int backlog) int rc; int i; - laddr = (struct sockaddr_in *)&cm_id->local_addr; - laddr6 = (struct sockaddr_in6 *)&cm_id->local_addr; + laddr = (struct sockaddr_in *)&cm_id->m_local_addr; + laddr6 = (struct sockaddr_in6 *)&cm_id->m_local_addr; DP_DEBUG(dev, QEDR_MSG_IWARP, "Create Listener address: %pISpc\n", &cm_id->local_addr); diff --git a/drivers/infiniband/hw/qedr/verbs.c b/drivers/infiniband/hw/qedr/verbs.c index 53f00dbf313f..f9c3cc71f5c0 100644 --- a/drivers/infiniband/hw/qedr/verbs.c +++ b/drivers/infiniband/hw/qedr/verbs.c @@ -1841,14 +1841,15 @@ static void qedr_reset_qp_hwq_info(struct qedr_qp_hwq_info *qph) static int qedr_update_qp_state(struct qedr_dev *dev, struct qedr_qp *qp, + enum qed_roce_qp_state cur_state, enum qed_roce_qp_state new_state) { int status = 0; - if (new_state == qp->state) + if (new_state == cur_state) return 0; - switch (qp->state) { + switch (cur_state) { case QED_ROCE_QP_STATE_RESET: switch (new_state) { case QED_ROCE_QP_STATE_INIT: @@ -1955,6 +1956,7 @@ int qedr_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, struct qedr_dev *dev = get_qedr_dev(&qp->dev->ibdev); const struct ib_global_route *grh = rdma_ah_read_grh(&attr->ah_attr); enum ib_qp_state old_qp_state, new_qp_state; + enum qed_roce_qp_state cur_state; int rc = 0; DP_DEBUG(dev, QEDR_MSG_QP, @@ -2086,18 +2088,23 @@ int qedr_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, SET_FIELD(qp_params.modify_flags, QED_ROCE_MODIFY_QP_VALID_ACK_TIMEOUT, 1); - qp_params.ack_timeout = attr->timeout; - if (attr->timeout) { - u32 temp; - - temp = 4096 * (1UL << attr->timeout) / 1000 / 1000; - /* FW requires [msec] */ - qp_params.ack_timeout = temp; - } else { - /* Infinite */ + /* The received timeout value is an exponent used like this: + * "12.7.34 LOCAL ACK TIMEOUT + * Value representing the transport (ACK) timeout for use by + * the remote, expressed as: 4.096 * 2^timeout [usec]" + * The FW expects timeout in msec so we need to divide the usec + * result by 1000. We'll approximate 1000~2^10, and 4.096 ~ 2^2, + * so we get: 2^2 * 2^timeout / 2^10 = 2^(timeout - 8). + * The value of zero means infinite so we use a 'max_t' to make + * sure that sub 1 msec values will be configured as 1 msec. + */ + if (attr->timeout) + qp_params.ack_timeout = + 1 << max_t(int, attr->timeout - 8, 0); + else qp_params.ack_timeout = 0; - } } + if (attr_mask & IB_QP_RETRY_CNT) { SET_FIELD(qp_params.modify_flags, QED_ROCE_MODIFY_QP_VALID_RETRY_CNT, 1); @@ -2170,13 +2177,25 @@ int qedr_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, qp->dest_qp_num = attr->dest_qp_num; } + cur_state = qp->state; + + /* Update the QP state before the actual ramrod to prevent a race with + * fast path. Modifying the QP state to error will cause the device to + * flush the CQEs and while polling the flushed CQEs will considered as + * a potential issue if the QP isn't in error state. + */ + if ((attr_mask & IB_QP_STATE) && qp->qp_type != IB_QPT_GSI && + !udata && qp_params.new_state == QED_ROCE_QP_STATE_ERR) + qp->state = QED_ROCE_QP_STATE_ERR; + if (qp->qp_type != IB_QPT_GSI) rc = dev->ops->rdma_modify_qp(dev->rdma_ctx, qp->qed_qp, &qp_params); if (attr_mask & IB_QP_STATE) { if ((qp->qp_type != IB_QPT_GSI) && (!udata)) - rc = qedr_update_qp_state(dev, qp, qp_params.new_state); + rc = qedr_update_qp_state(dev, qp, cur_state, + qp_params.new_state); qp->state = qp_params.new_state; } @@ -3034,6 +3053,11 @@ static int __qedr_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr, switch (wr->opcode) { case IB_WR_SEND_WITH_IMM: + if (unlikely(rdma_protocol_iwarp(&dev->ibdev, 1))) { + rc = -EINVAL; + *bad_wr = wr; + break; + } wqe->req_type = RDMA_SQ_REQ_TYPE_SEND_WITH_IMM; swqe = (struct rdma_sq_send_wqe_1st *)wqe; swqe->wqe_size = 2; @@ -3075,6 +3099,11 @@ static int __qedr_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr, break; case IB_WR_RDMA_WRITE_WITH_IMM: + if (unlikely(rdma_protocol_iwarp(&dev->ibdev, 1))) { + rc = -EINVAL; + *bad_wr = wr; + break; + } wqe->req_type = RDMA_SQ_REQ_TYPE_RDMA_WR_WITH_IMM; rwqe = (struct rdma_sq_rdma_wqe_1st *)wqe; @@ -3685,7 +3714,7 @@ static int process_resp_flush(struct qedr_qp *qp, struct qedr_cq *cq, static void try_consume_resp_cqe(struct qedr_cq *cq, struct qedr_qp *qp, struct rdma_cqe_responder *resp, int *update) { - if (le16_to_cpu(resp->rq_cons) == qp->rq.wqe_cons) { + if (le16_to_cpu(resp->rq_cons_or_srq_id) == qp->rq.wqe_cons) { consume_cqe(cq); *update |= 1; } @@ -3700,7 +3729,7 @@ static int qedr_poll_cq_resp(struct qedr_dev *dev, struct qedr_qp *qp, if (resp->status == RDMA_CQE_RESP_STS_WORK_REQUEST_FLUSHED_ERR) { cnt = process_resp_flush(qp, cq, num_entries, wc, - resp->rq_cons); + resp->rq_cons_or_srq_id); try_consume_resp_cqe(cq, qp, resp, update); } else { cnt = process_resp_one(dev, qp, cq, wc, resp); @@ -3724,7 +3753,7 @@ int qedr_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *wc) { struct qedr_dev *dev = get_qedr_dev(ibcq->device); struct qedr_cq *cq = get_qedr_cq(ibcq); - union rdma_cqe *cqe = cq->latest_cqe; + union rdma_cqe *cqe; u32 old_cons, new_cons; unsigned long flags; int update = 0; @@ -3741,6 +3770,7 @@ int qedr_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *wc) return qedr_gsi_poll_cq(ibcq, num_entries, wc); spin_lock_irqsave(&cq->cq_lock, flags); + cqe = cq->latest_cqe; old_cons = qed_chain_get_cons_idx_u32(&cq->pbl); while (num_entries && is_valid_cqe(cq, cqe)) { struct qedr_qp *qp; diff --git a/drivers/infiniband/hw/usnic/usnic_ib_main.c b/drivers/infiniband/hw/usnic/usnic_ib_main.c index f45e99a938e0..ca5638091b55 100644 --- a/drivers/infiniband/hw/usnic/usnic_ib_main.c +++ b/drivers/infiniband/hw/usnic/usnic_ib_main.c @@ -95,24 +95,6 @@ void usnic_ib_log_vf(struct usnic_ib_vf *vf) } /* Start of netdev section */ -static inline const char *usnic_ib_netdev_event_to_string(unsigned long event) -{ - const char *event2str[] = {"NETDEV_NONE", "NETDEV_UP", "NETDEV_DOWN", - "NETDEV_REBOOT", "NETDEV_CHANGE", - "NETDEV_REGISTER", "NETDEV_UNREGISTER", "NETDEV_CHANGEMTU", - "NETDEV_CHANGEADDR", "NETDEV_GOING_DOWN", "NETDEV_FEAT_CHANGE", - "NETDEV_BONDING_FAILOVER", "NETDEV_PRE_UP", - "NETDEV_PRE_TYPE_CHANGE", "NETDEV_POST_TYPE_CHANGE", - "NETDEV_POST_INT", "NETDEV_UNREGISTER_FINAL", "NETDEV_RELEASE", - "NETDEV_NOTIFY_PEERS", "NETDEV_JOIN" - }; - - if (event >= ARRAY_SIZE(event2str)) - return "UNKNOWN_NETDEV_EVENT"; - else - return event2str[event]; -} - static void usnic_ib_qp_grp_modify_active_to_err(struct usnic_ib_dev *us_ibdev) { struct usnic_ib_ucontext *ctx; @@ -185,7 +167,7 @@ static void usnic_ib_handle_usdev_event(struct usnic_ib_dev *us_ibdev, ib_dispatch_event(&ib_event); } else { usnic_dbg("Ignoring %s on %s\n", - usnic_ib_netdev_event_to_string(event), + netdev_cmd_to_name(event), us_ibdev->ib_dev.name); } break; @@ -222,7 +204,7 @@ static void usnic_ib_handle_usdev_event(struct usnic_ib_dev *us_ibdev, break; default: usnic_dbg("Ignoring event %s on %s", - usnic_ib_netdev_event_to_string(event), + netdev_cmd_to_name(event), us_ibdev->ib_dev.name); } mutex_unlock(&us_ibdev->usdev_lock); @@ -264,7 +246,7 @@ static int usnic_ib_handle_inet_event(struct usnic_ib_dev *us_ibdev, switch (event) { case NETDEV_DOWN: usnic_info("%s via ip notifiers", - usnic_ib_netdev_event_to_string(event)); + netdev_cmd_to_name(event)); usnic_fwd_del_ipaddr(us_ibdev->ufdev); usnic_ib_qp_grp_modify_active_to_err(us_ibdev); ib_event.event = IB_EVENT_GID_CHANGE; @@ -275,7 +257,7 @@ static int usnic_ib_handle_inet_event(struct usnic_ib_dev *us_ibdev, case NETDEV_UP: usnic_fwd_add_ipaddr(us_ibdev->ufdev, ifa->ifa_address); usnic_info("%s via ip notifiers: ip %pI4", - usnic_ib_netdev_event_to_string(event), + netdev_cmd_to_name(event), &us_ibdev->ufdev->inaddr); ib_event.event = IB_EVENT_GID_CHANGE; ib_event.device = &us_ibdev->ib_dev; @@ -284,7 +266,7 @@ static int usnic_ib_handle_inet_event(struct usnic_ib_dev *us_ibdev, break; default: usnic_info("Ignoring event %s on %s", - usnic_ib_netdev_event_to_string(event), + netdev_cmd_to_name(event), us_ibdev->ib_dev.name); } mutex_unlock(&us_ibdev->usdev_lock); diff --git a/drivers/infiniband/hw/usnic/usnic_transport.c b/drivers/infiniband/hw/usnic/usnic_transport.c index de318389a301..67de94343cb4 100644 --- a/drivers/infiniband/hw/usnic/usnic_transport.c +++ b/drivers/infiniband/hw/usnic/usnic_transport.c @@ -174,14 +174,13 @@ void usnic_transport_put_socket(struct socket *sock) int usnic_transport_sock_get_addr(struct socket *sock, int *proto, uint32_t *addr, uint16_t *port) { - int len; int err; struct sockaddr_in sock_addr; err = sock->ops->getname(sock, (struct sockaddr *)&sock_addr, - &len, 0); - if (err) + 0); + if (err < 0) return err; if (sock_addr.sin_family != AF_INET) diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_cq.c b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_cq.c index faa9478c14a6..f95b97646c25 100644 --- a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_cq.c +++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_cq.c @@ -114,6 +114,7 @@ struct ib_cq *pvrdma_create_cq(struct ib_device *ibdev, union pvrdma_cmd_resp rsp; struct pvrdma_cmd_create_cq *cmd = &req.create_cq; struct pvrdma_cmd_create_cq_resp *resp = &rsp.create_cq_resp; + struct pvrdma_create_cq_resp cq_resp = {0}; struct pvrdma_create_cq ucmd; BUILD_BUG_ON(sizeof(struct pvrdma_cqe) != 64); @@ -197,6 +198,7 @@ struct ib_cq *pvrdma_create_cq(struct ib_device *ibdev, cq->ibcq.cqe = resp->cqe; cq->cq_handle = resp->cq_handle; + cq_resp.cqn = resp->cq_handle; spin_lock_irqsave(&dev->cq_tbl_lock, flags); dev->cq_tbl[cq->cq_handle % dev->dsr->caps.max_cq] = cq; spin_unlock_irqrestore(&dev->cq_tbl_lock, flags); @@ -205,7 +207,7 @@ struct ib_cq *pvrdma_create_cq(struct ib_device *ibdev, cq->uar = &(to_vucontext(context)->uar); /* Copy udata back. */ - if (ib_copy_to_udata(udata, &cq->cq_handle, sizeof(__u32))) { + if (ib_copy_to_udata(udata, &cq_resp, sizeof(cq_resp))) { dev_warn(&dev->pdev->dev, "failed to copy back udata\n"); pvrdma_destroy_cq(&cq->ibcq); diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_srq.c b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_srq.c index 5acebb1ef631..af235967a9c2 100644 --- a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_srq.c +++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_srq.c @@ -113,6 +113,7 @@ struct ib_srq *pvrdma_create_srq(struct ib_pd *pd, union pvrdma_cmd_resp rsp; struct pvrdma_cmd_create_srq *cmd = &req.create_srq; struct pvrdma_cmd_create_srq_resp *resp = &rsp.create_srq_resp; + struct pvrdma_create_srq_resp srq_resp = {0}; struct pvrdma_create_srq ucmd; unsigned long flags; int ret; @@ -204,12 +205,13 @@ struct ib_srq *pvrdma_create_srq(struct ib_pd *pd, } srq->srq_handle = resp->srqn; + srq_resp.srqn = resp->srqn; spin_lock_irqsave(&dev->srq_tbl_lock, flags); dev->srq_tbl[srq->srq_handle % dev->dsr->caps.max_srq] = srq; spin_unlock_irqrestore(&dev->srq_tbl_lock, flags); /* Copy udata back. */ - if (ib_copy_to_udata(udata, &srq->srq_handle, sizeof(__u32))) { + if (ib_copy_to_udata(udata, &srq_resp, sizeof(srq_resp))) { dev_warn(&dev->pdev->dev, "failed to copy back udata\n"); pvrdma_destroy_srq(&srq->ibsrq); return ERR_PTR(-EINVAL); diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.c b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.c index 16b96616ef7e..a51463cd2f37 100644 --- a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.c +++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.c @@ -447,6 +447,7 @@ struct ib_pd *pvrdma_alloc_pd(struct ib_device *ibdev, union pvrdma_cmd_resp rsp; struct pvrdma_cmd_create_pd *cmd = &req.create_pd; struct pvrdma_cmd_create_pd_resp *resp = &rsp.create_pd_resp; + struct pvrdma_alloc_pd_resp pd_resp = {0}; int ret; void *ptr; @@ -475,9 +476,10 @@ struct ib_pd *pvrdma_alloc_pd(struct ib_device *ibdev, pd->privileged = !context; pd->pd_handle = resp->pd_handle; pd->pdn = resp->pd_handle; + pd_resp.pdn = resp->pd_handle; if (context) { - if (ib_copy_to_udata(udata, &pd->pdn, sizeof(__u32))) { + if (ib_copy_to_udata(udata, &pd_resp, sizeof(pd_resp))) { dev_warn(&dev->pdev->dev, "failed to copy back protection domain\n"); pvrdma_dealloc_pd(&pd->ibpd); diff --git a/drivers/infiniband/sw/rdmavt/mr.c b/drivers/infiniband/sw/rdmavt/mr.c index 1b2e5362a3ff..cc429b567d0a 100644 --- a/drivers/infiniband/sw/rdmavt/mr.c +++ b/drivers/infiniband/sw/rdmavt/mr.c @@ -489,11 +489,13 @@ static int rvt_check_refs(struct rvt_mregion *mr, const char *t) unsigned long timeout; struct rvt_dev_info *rdi = ib_to_rvt(mr->pd->device); - if (percpu_ref_is_zero(&mr->refcount)) - return 0; - /* avoid dma mr */ - if (mr->lkey) + if (mr->lkey) { + /* avoid dma mr */ rvt_dereg_clean_qps(mr); + /* @mr was indexed on rcu protected @lkey_table */ + synchronize_rcu(); + } + timeout = wait_for_completion_timeout(&mr->comp, 5 * HZ); if (!timeout) { rvt_pr_err(rdi, diff --git a/drivers/infiniband/ulp/ipoib/ipoib_fs.c b/drivers/infiniband/ulp/ipoib/ipoib_fs.c index 11f74cbe6660..ea302b054601 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_fs.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_fs.c @@ -281,8 +281,6 @@ void ipoib_delete_debug_files(struct net_device *dev) { struct ipoib_dev_priv *priv = ipoib_priv(dev); - WARN_ONCE(!priv->mcg_dentry, "null mcg debug file\n"); - WARN_ONCE(!priv->path_dentry, "null path debug file\n"); debugfs_remove(priv->mcg_dentry); debugfs_remove(priv->path_dentry); priv->mcg_dentry = priv->path_dentry = NULL; |