From d97505baea64d93538b16baf14ce7b8c1fbad746 Mon Sep 17 00:00:00 2001 From: Yishai Hadas Date: Sun, 19 Jan 2025 14:36:13 +0200 Subject: [PATCH 01/15] RDMA/mlx5: Fix the recovery flow of the UMR QP This patch addresses an issue in the recovery flow of the UMR QP, ensuring tasks do not get stuck, as highlighted by the call trace [1]. During recovery, before transitioning the QP to the RESET state, the software must wait for all outstanding WRs to complete. Failing to do so can cause the firmware to skip sending some flushed CQEs with errors and simply discard them upon the RESET, as per the IB specification. This race condition can result in lost CQEs and tasks becoming stuck. To resolve this, the patch sends a final WR which serves only as a barrier before moving the QP state to RESET. Once a CQE is received for that final WR, it guarantees that no outstanding WRs remain, making it safe to transition the QP to RESET and subsequently back to RTS, restoring proper functionality. Note: For the barrier WR, we simply reuse the failed and ready WR. Since the QP is in an error state, it will only receive IB_WC_WR_FLUSH_ERR. However, as it serves only as a barrier we don't care about its status. [1] INFO: task rdma_resource_l:1922 blocked for more than 120 seconds. Tainted: G W 6.12.0-rc7+ #1626 "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. task:rdma_resource_l state:D stack:0 pid:1922 tgid:1922 ppid:1369 flags:0x00004004 Call Trace: __schedule+0x420/0xd30 schedule+0x47/0x130 schedule_timeout+0x280/0x300 ? mark_held_locks+0x48/0x80 ? lockdep_hardirqs_on_prepare+0xe5/0x1a0 wait_for_completion+0x75/0x130 mlx5r_umr_post_send_wait+0x3c2/0x5b0 [mlx5_ib] ? __pfx_mlx5r_umr_done+0x10/0x10 [mlx5_ib] mlx5r_umr_revoke_mr+0x93/0xc0 [mlx5_ib] __mlx5_ib_dereg_mr+0x299/0x520 [mlx5_ib] ? _raw_spin_unlock_irq+0x24/0x40 ? wait_for_completion+0xfe/0x130 ? rdma_restrack_put+0x63/0xe0 [ib_core] ib_dereg_mr_user+0x5f/0x120 [ib_core] ? lock_release+0xc6/0x280 destroy_hw_idr_uobject+0x1d/0x60 [ib_uverbs] uverbs_destroy_uobject+0x58/0x1d0 [ib_uverbs] uobj_destroy+0x3f/0x70 [ib_uverbs] ib_uverbs_cmd_verbs+0x3e4/0xbb0 [ib_uverbs] ? __pfx_uverbs_destroy_def_handler+0x10/0x10 [ib_uverbs] ? __lock_acquire+0x64e/0x2080 ? mark_held_locks+0x48/0x80 ? find_held_lock+0x2d/0xa0 ? lock_acquire+0xc1/0x2f0 ? ib_uverbs_ioctl+0xcb/0x170 [ib_uverbs] ? __fget_files+0xc3/0x1b0 ib_uverbs_ioctl+0xe7/0x170 [ib_uverbs] ? ib_uverbs_ioctl+0xcb/0x170 [ib_uverbs] __x64_sys_ioctl+0x1b0/0xa70 do_syscall_64+0x6b/0x140 entry_SYSCALL_64_after_hwframe+0x76/0x7e RIP: 0033:0x7f99c918b17b RSP: 002b:00007ffc766d0468 EFLAGS: 00000246 ORIG_RAX: 0000000000000010 RAX: ffffffffffffffda RBX: 00007ffc766d0578 RCX: 00007f99c918b17b RDX: 00007ffc766d0560 RSI: 00000000c0181b01 RDI: 0000000000000003 RBP: 00007ffc766d0540 R08: 00007f99c8f99010 R09: 000000000000bd7e R10: 00007f99c94c1c70 R11: 0000000000000246 R12: 00007ffc766d0530 R13: 000000000000001c R14: 0000000040246a80 R15: 0000000000000000 Fixes: 158e71bb69e3 ("RDMA/mlx5: Add a umr recovery flow") Signed-off-by: Yishai Hadas Reviewed-by: Michael Guralnik Link: https://patch.msgid.link/27b51b92ec42dfb09d8096fcbd51878f397ce6ec.1737290141.git.leon@kernel.org Signed-off-by: Leon Romanovsky --- drivers/infiniband/hw/mlx5/umr.c | 83 +++++++++++++++++++++----------- 1 file changed, 56 insertions(+), 27 deletions(-) diff --git a/drivers/infiniband/hw/mlx5/umr.c b/drivers/infiniband/hw/mlx5/umr.c index 887fd6fa3ba9..793f3c5c4d01 100644 --- a/drivers/infiniband/hw/mlx5/umr.c +++ b/drivers/infiniband/hw/mlx5/umr.c @@ -231,30 +231,6 @@ void mlx5r_umr_cleanup(struct mlx5_ib_dev *dev) ib_dealloc_pd(dev->umrc.pd); } -static int mlx5r_umr_recover(struct mlx5_ib_dev *dev) -{ - struct umr_common *umrc = &dev->umrc; - struct ib_qp_attr attr; - int err; - - attr.qp_state = IB_QPS_RESET; - err = ib_modify_qp(umrc->qp, &attr, IB_QP_STATE); - if (err) { - mlx5_ib_dbg(dev, "Couldn't modify UMR QP\n"); - goto err; - } - - err = mlx5r_umr_qp_rst2rts(dev, umrc->qp); - if (err) - goto err; - - umrc->state = MLX5_UMR_STATE_ACTIVE; - return 0; - -err: - umrc->state = MLX5_UMR_STATE_ERR; - return err; -} static int mlx5r_umr_post_send(struct ib_qp *ibqp, u32 mkey, struct ib_cqe *cqe, struct mlx5r_umr_wqe *wqe, bool with_data) @@ -302,6 +278,61 @@ static int mlx5r_umr_post_send(struct ib_qp *ibqp, u32 mkey, struct ib_cqe *cqe, return err; } +static int mlx5r_umr_recover(struct mlx5_ib_dev *dev, u32 mkey, + struct mlx5r_umr_context *umr_context, + struct mlx5r_umr_wqe *wqe, bool with_data) +{ + struct umr_common *umrc = &dev->umrc; + struct ib_qp_attr attr; + int err; + + mutex_lock(&umrc->lock); + /* Preventing any further WRs to be sent now */ + if (umrc->state != MLX5_UMR_STATE_RECOVER) { + mlx5_ib_warn(dev, "UMR recovery encountered an unexpected state=%d\n", + umrc->state); + umrc->state = MLX5_UMR_STATE_RECOVER; + } + mutex_unlock(&umrc->lock); + + /* Sending a final/barrier WR (the failed one) and wait for its completion. + * This will ensure that all the previous WRs got a completion before + * we set the QP state to RESET. + */ + err = mlx5r_umr_post_send(umrc->qp, mkey, &umr_context->cqe, wqe, + with_data); + if (err) { + mlx5_ib_warn(dev, "UMR recovery post send failed, err %d\n", err); + goto err; + } + + /* Since the QP is in an error state, it will only receive + * IB_WC_WR_FLUSH_ERR. However, as it serves only as a barrier + * we don't care about its status. + */ + wait_for_completion(&umr_context->done); + + attr.qp_state = IB_QPS_RESET; + err = ib_modify_qp(umrc->qp, &attr, IB_QP_STATE); + if (err) { + mlx5_ib_warn(dev, "Couldn't modify UMR QP to RESET, err=%d\n", err); + goto err; + } + + err = mlx5r_umr_qp_rst2rts(dev, umrc->qp); + if (err) { + mlx5_ib_warn(dev, "Couldn't modify UMR QP to RTS, err=%d\n", err); + goto err; + } + + umrc->state = MLX5_UMR_STATE_ACTIVE; + return 0; + +err: + umrc->state = MLX5_UMR_STATE_ERR; + return err; +} + static void mlx5r_umr_done(struct ib_cq *cq, struct ib_wc *wc) { struct mlx5_ib_umr_context *context = @@ -366,9 +397,7 @@ static int mlx5r_umr_post_send_wait(struct mlx5_ib_dev *dev, u32 mkey, mlx5_ib_warn(dev, "reg umr failed (%u). Trying to recover and resubmit the flushed WQEs, mkey = %u\n", umr_context.status, mkey); - mutex_lock(&umrc->lock); - err = mlx5r_umr_recover(dev); - mutex_unlock(&umrc->lock); + err = mlx5r_umr_recover(dev, mkey, &umr_context, wqe, with_data); if (err) mlx5_ib_warn(dev, "couldn't recover UMR, err %d\n", err); From 12d044770e12c4205fa69535b4fa8a9981fea98f Mon Sep 17 00:00:00 2001 From: Mark Zhang Date: Sun, 19 Jan 2025 14:39:46 +0200 Subject: [PATCH 02/15] IB/mlx5: Set and get correct qp_num for a DCT QP When a DCT QP is created on an active lag, it's dctc.port is assigned in a round-robin way, which is from 1 to dev->lag_port. In this case when querying this QP, we may get qp_attr.port_num > 2. Fix this by setting qp->port when modifying a DCT QP, and read port_num from qp->port instead of dctc.port when querying it. Fixes: 7c4b1ab9f167 ("IB/mlx5: Add DCT RoCE LAG support") Signed-off-by: Mark Zhang Reviewed-by: Maher Sanalla Link: https://patch.msgid.link/94c76bf0adbea997f87ffa27674e0a7118ad92a9.1737290358.git.leon@kernel.org Signed-off-by: Leon Romanovsky --- drivers/infiniband/hw/mlx5/qp.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c index a43eba9d3572..08d22db8dca9 100644 --- a/drivers/infiniband/hw/mlx5/qp.c +++ b/drivers/infiniband/hw/mlx5/qp.c @@ -4579,6 +4579,8 @@ static int mlx5_ib_modify_dct(struct ib_qp *ibqp, struct ib_qp_attr *attr, set_id = mlx5_ib_get_counters_id(dev, attr->port_num - 1); MLX5_SET(dctc, dctc, counter_set_id, set_id); + + qp->port = attr->port_num; } else if (cur_state == IB_QPS_INIT && new_state == IB_QPS_RTR) { struct mlx5_ib_modify_qp_resp resp = {}; u32 out[MLX5_ST_SZ_DW(create_dct_out)] = {}; @@ -5074,7 +5076,7 @@ static int mlx5_ib_dct_query_qp(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *mqp, } if (qp_attr_mask & IB_QP_PORT) - qp_attr->port_num = MLX5_GET(dctc, dctc, port); + qp_attr->port_num = mqp->port; if (qp_attr_mask & IB_QP_MIN_RNR_TIMER) qp_attr->min_rnr_timer = MLX5_GET(dctc, dctc, min_rnr_nak); if (qp_attr_mask & IB_QP_AV) { From cc668a11e6ac8adb0e016711080d3f314722cc91 Mon Sep 17 00:00:00 2001 From: Yishai Hadas Date: Mon, 3 Feb 2025 14:50:59 +0200 Subject: [PATCH 03/15] RDMA/mlx5: Fix a race for DMABUF MR which can lead to CQE with error This patch addresses a potential race condition for a DMABUF MR that can result in a CQE with an error on the UMR QP. During the __mlx5_ib_dereg_mr() flow, the following sequence of calls occurs: mlx5_revoke_mr() mlx5r_umr_revoke_mr() mlx5r_umr_post_send_wait() At this point, the lkey is freed from the hardware's perspective. However, concurrently, mlx5_ib_dmabuf_invalidate_cb() might be triggered by another task attempting to invalidate the MR having that freed lkey. Since the lkey has already been freed, this can lead to a CQE error, causing the UMR QP to enter an error state. To resolve this race condition, the dma_resv_lock() which was hold as part of the mlx5_ib_dmabuf_invalidate_cb() is now also acquired as part of the mlx5_revoke_mr() scope. Upon a successful revoke, we set umem_dmabuf->private which points to that MR to NULL, preventing any further invalidation attempts on its lkey. Fixes: e6fb246ccafb ("RDMA/mlx5: Consolidate MR destruction to mlx5_ib_dereg_mr()") Signed-off-by: Yishai Hadas Reviewed-by: Artemy Kovalyov Link: https://patch.msgid.link/70617067abbfaa0c816a2544c922e7f4346def58.1738587016.git.leon@kernel.org Signed-off-by: Leon Romanovsky --- drivers/infiniband/hw/mlx5/mr.c | 13 ++++++++++++- 1 file changed, 12 insertions(+), 1 deletion(-) diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c index bb02b6adbf2c..0a3cbb14e183 100644 --- a/drivers/infiniband/hw/mlx5/mr.c +++ b/drivers/infiniband/hw/mlx5/mr.c @@ -1550,7 +1550,7 @@ static void mlx5_ib_dmabuf_invalidate_cb(struct dma_buf_attachment *attach) dma_resv_assert_held(umem_dmabuf->attach->dmabuf->resv); - if (!umem_dmabuf->sgt) + if (!umem_dmabuf->sgt || !mr) return; mlx5r_umr_update_mr_pas(mr, MLX5_IB_UPD_XLT_ZAP); @@ -2022,11 +2022,16 @@ static int mlx5_revoke_mr(struct mlx5_ib_mr *mr) struct mlx5_ib_dev *dev = to_mdev(mr->ibmr.device); struct mlx5_cache_ent *ent = mr->mmkey.cache_ent; bool is_odp = is_odp_mr(mr); + bool is_odp_dma_buf = is_dmabuf_mr(mr) && + !to_ib_umem_dmabuf(mr->umem)->pinned; int ret = 0; if (is_odp) mutex_lock(&to_ib_umem_odp(mr->umem)->umem_mutex); + if (is_odp_dma_buf) + dma_resv_lock(to_ib_umem_dmabuf(mr->umem)->attach->dmabuf->resv, NULL); + if (mr->mmkey.cacheable && !mlx5r_umr_revoke_mr(mr) && !cache_ent_find_and_store(dev, mr)) { ent = mr->mmkey.cache_ent; /* upon storing to a clean temp entry - schedule its cleanup */ @@ -2054,6 +2059,12 @@ static int mlx5_revoke_mr(struct mlx5_ib_mr *mr) mutex_unlock(&to_ib_umem_odp(mr->umem)->umem_mutex); } + if (is_odp_dma_buf) { + if (!ret) + to_ib_umem_dmabuf(mr->umem)->private = NULL; + dma_resv_unlock(to_ib_umem_dmabuf(mr->umem)->attach->dmabuf->resv); + } + return ret; } From abc7b3f1f056d69a8f11d6dceecc0c9549ace770 Mon Sep 17 00:00:00 2001 From: Yishai Hadas Date: Mon, 3 Feb 2025 14:51:43 +0200 Subject: [PATCH 04/15] RDMA/mlx5: Fix a WARN during dereg_mr for DM type Memory regions (MR) of type DM (device memory) do not have an associated umem. In the __mlx5_ib_dereg_mr() -> mlx5_free_priv_descs() flow, the code incorrectly takes the wrong branch, attempting to call dma_unmap_single() on a DMA address that is not mapped. This results in a WARN [1], as shown below. The issue is resolved by properly accounting for the DM type and ensuring the correct branch is selected in mlx5_free_priv_descs(). [1] WARNING: CPU: 12 PID: 1346 at drivers/iommu/dma-iommu.c:1230 iommu_dma_unmap_page+0x79/0x90 Modules linked in: ip6table_mangle ip6table_nat ip6table_filter ip6_tables iptable_mangle xt_conntrack xt_MASQUERADE nf_conntrack_netlink nfnetlink xt_addrtype iptable_nat nf_nat br_netfilter rpcsec_gss_krb5 auth_rpcgss oid_registry ovelay rpcrdma rdma_ucm ib_iser libiscsi scsi_transport_iscsi ib_umad rdma_cm ib_ipoib iw_cm ib_cm mlx5_ib ib_uverbs ib_core fuse mlx5_core CPU: 12 UID: 0 PID: 1346 Comm: ibv_rc_pingpong Not tainted 6.12.0-rc7+ #1631 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014 RIP: 0010:iommu_dma_unmap_page+0x79/0x90 Code: 2b 49 3b 29 72 26 49 3b 69 08 73 20 4d 89 f0 44 89 e9 4c 89 e2 48 89 ee 48 89 df 5b 5d 41 5c 41 5d 41 5e 41 5f e9 07 b8 88 ff <0f> 0b 5b 5d 41 5c 41 5d 41 5e 41 5f c3 cc cc cc cc 66 0f 1f 44 00 RSP: 0018:ffffc90001913a10 EFLAGS: 00010246 RAX: 0000000000000000 RBX: ffff88810194b0a8 RCX: 0000000000000000 RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000001 RBP: ffff88810194b0a8 R08: 0000000000000000 R09: 0000000000000000 R10: 0000000000000001 R11: 0000000000000000 R12: 0000000000000000 R13: 0000000000000001 R14: 0000000000000000 R15: 0000000000000000 FS: 00007f537abdd740(0000) GS:ffff88885fb00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00007f537aeb8000 CR3: 000000010c248001 CR4: 0000000000372eb0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Call Trace: ? __warn+0x84/0x190 ? iommu_dma_unmap_page+0x79/0x90 ? report_bug+0xf8/0x1c0 ? handle_bug+0x55/0x90 ? exc_invalid_op+0x13/0x60 ? asm_exc_invalid_op+0x16/0x20 ? iommu_dma_unmap_page+0x79/0x90 dma_unmap_page_attrs+0xe6/0x290 mlx5_free_priv_descs+0xb0/0xe0 [mlx5_ib] __mlx5_ib_dereg_mr+0x37e/0x520 [mlx5_ib] ? _raw_spin_unlock_irq+0x24/0x40 ? wait_for_completion+0xfe/0x130 ? rdma_restrack_put+0x63/0xe0 [ib_core] ib_dereg_mr_user+0x5f/0x120 [ib_core] ? lock_release+0xc6/0x280 destroy_hw_idr_uobject+0x1d/0x60 [ib_uverbs] uverbs_destroy_uobject+0x58/0x1d0 [ib_uverbs] uobj_destroy+0x3f/0x70 [ib_uverbs] ib_uverbs_cmd_verbs+0x3e4/0xbb0 [ib_uverbs] ? __pfx_uverbs_destroy_def_handler+0x10/0x10 [ib_uverbs] ? lock_acquire+0xc1/0x2f0 ? ib_uverbs_ioctl+0xcb/0x170 [ib_uverbs] ? ib_uverbs_ioctl+0x116/0x170 [ib_uverbs] ? lock_release+0xc6/0x280 ib_uverbs_ioctl+0xe7/0x170 [ib_uverbs] ? ib_uverbs_ioctl+0xcb/0x170 [ib_uverbs] __x64_sys_ioctl+0x1b0/0xa70 do_syscall_64+0x6b/0x140 entry_SYSCALL_64_after_hwframe+0x76/0x7e RIP: 0033:0x7f537adaf17b Code: 0f 1e fa 48 8b 05 1d ad 0c 00 64 c7 00 26 00 00 00 48 c7 c0 ff ff ff ff c3 66 0f 1f 44 00 00 f3 0f 1e fa b8 10 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d ed ac 0c 00 f7 d8 64 89 01 48 RSP: 002b:00007ffff218f0b8 EFLAGS: 00000246 ORIG_RAX: 0000000000000010 RAX: ffffffffffffffda RBX: 00007ffff218f1d8 RCX: 00007f537adaf17b RDX: 00007ffff218f1c0 RSI: 00000000c0181b01 RDI: 0000000000000003 RBP: 00007ffff218f1a0 R08: 00007f537aa8d010 R09: 0000561ee2e4f270 R10: 00007f537aace3a8 R11: 0000000000000246 R12: 00007ffff218f190 R13: 000000000000001c R14: 0000561ee2e4d7c0 R15: 00007ffff218f450 Fixes: f18ec4223117 ("RDMA/mlx5: Use a union inside mlx5_ib_mr") Signed-off-by: Yishai Hadas Link: https://patch.msgid.link/2039c22cfc3df02378747ba4d623a558b53fc263.1738587076.git.leon@kernel.org Signed-off-by: Leon Romanovsky --- drivers/infiniband/hw/mlx5/mr.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c index 0a3cbb14e183..753faa9ad06a 100644 --- a/drivers/infiniband/hw/mlx5/mr.c +++ b/drivers/infiniband/hw/mlx5/mr.c @@ -1935,7 +1935,8 @@ mlx5_alloc_priv_descs(struct ib_device *device, static void mlx5_free_priv_descs(struct mlx5_ib_mr *mr) { - if (!mr->umem && !mr->data_direct && mr->descs) { + if (!mr->umem && !mr->data_direct && + mr->ibmr.type != IB_MR_TYPE_DM && mr->descs) { struct ib_device *device = mr->ibmr.device; int size = mr->max_descs * mr->desc_size; struct mlx5_ib_dev *dev = to_mdev(device); From 29b7bb98234cc287cebef9bccf638c2e3f39be71 Mon Sep 17 00:00:00 2001 From: Konstantin Taranov Date: Wed, 5 Feb 2025 02:30:05 -0800 Subject: [PATCH 05/15] RDMA/mana_ib: Allocate PAGE aligned doorbell index Allocate a PAGE aligned doorbell index to ensure each process gets a separate PAGE sized doorbell area space remapped to it in mana_ib_mmap Fixes: 0266a177631d ("RDMA/mana_ib: Add a driver for Microsoft Azure Network Adapter") Signed-off-by: Shiraz Saleem Signed-off-by: Konstantin Taranov Link: https://patch.msgid.link/1738751405-15041-1-git-send-email-kotaranov@linux.microsoft.com Reviewed-by: Long Li Signed-off-by: Leon Romanovsky --- drivers/infiniband/hw/mana/main.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/infiniband/hw/mana/main.c b/drivers/infiniband/hw/mana/main.c index 67c2d43135a8..457cea6d9909 100644 --- a/drivers/infiniband/hw/mana/main.c +++ b/drivers/infiniband/hw/mana/main.c @@ -174,7 +174,7 @@ static int mana_gd_allocate_doorbell_page(struct gdma_context *gc, req.resource_type = GDMA_RESOURCE_DOORBELL_PAGE; req.num_resources = 1; - req.alignment = 1; + req.alignment = PAGE_SIZE / MANA_PAGE_SIZE; /* Have GDMA start searching from 0 */ req.allocated_resources = 0; From ee9d1619ef6e4a3412a13788256cb8c3e5efbe3d Mon Sep 17 00:00:00 2001 From: Long Li Date: Fri, 7 Feb 2025 13:46:28 -0800 Subject: [PATCH 06/15] MAINTAINERS: update maintainer for Microsoft MANA RDMA driver Ajay is no longer working on the MANA RDMA driver. Konstantin Taranov has made significant contributions to implementing RC QP in both kernel and user-mode. He will take the responsibility of fixing bugs, reviewing patches and developing new features for MANA RDMA driver. Signed-off-by: Long Li Link: https://patch.msgid.link/1738964792-21140-1-git-send-email-longli@linuxonhyperv.com Signed-off-by: Leon Romanovsky --- MAINTAINERS | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/MAINTAINERS b/MAINTAINERS index 896a307fa065..c51503268a02 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -15689,7 +15689,7 @@ F: include/uapi/linux/cciss*.h MICROSOFT MANA RDMA DRIVER M: Long Li -M: Ajay Sharma +M: Konstantin Taranov L: linux-rdma@vger.kernel.org S: Supported F: drivers/infiniband/hw/mana/ From 9747c0c7791d4a5a62018a0c9c563dd2e6f6c1c0 Mon Sep 17 00:00:00 2001 From: Junxian Huang Date: Sat, 8 Feb 2025 18:59:30 +0800 Subject: [PATCH 07/15] RDMA/hns: Fix mbox timing out by adding retry mechanism If a QP is modified to error state and a flush CQE process is triggered, the subsequent QP destruction mbox can still be successfully posted but will be blocked in HW until the flush CQE process finishes. This causes further mbox posting timeouts in driver. The blocking time is related to QP depth. Considering an extreme case where SQ depth and RQ depth are both 32K, the blocking time can reach about 135ms. This patch adds a retry mechanism for mbox posting. For each try, FW waits 15ms for HW to complete the previous mbox, otherwise return a timeout error code to driver. Counting other time consumption in FW, set 8 tries for mbox posting and a 5ms time gap before each retry to increase to a sufficient timeout limit. Fixes: 0425e3e6e0c7 ("RDMA/hns: Support flush cqe for hip08 in kernel space") Signed-off-by: Junxian Huang Link: https://patch.msgid.link/20250208105930.522796-1-huangjunxian6@hisilicon.com Signed-off-by: Leon Romanovsky --- drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 64 ++++++++++++++++------ drivers/infiniband/hw/hns/hns_roce_hw_v2.h | 2 + 2 files changed, 50 insertions(+), 16 deletions(-) diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c index dded339802b3..160e8927d364 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c @@ -1286,10 +1286,8 @@ static u32 hns_roce_cmdq_tx_timeout(u16 opcode, u32 tx_timeout) return tx_timeout; } -static void hns_roce_wait_csq_done(struct hns_roce_dev *hr_dev, u16 opcode) +static void hns_roce_wait_csq_done(struct hns_roce_dev *hr_dev, u32 tx_timeout) { - struct hns_roce_v2_priv *priv = hr_dev->priv; - u32 tx_timeout = hns_roce_cmdq_tx_timeout(opcode, priv->cmq.tx_timeout); u32 timeout = 0; do { @@ -1299,8 +1297,9 @@ static void hns_roce_wait_csq_done(struct hns_roce_dev *hr_dev, u16 opcode) } while (++timeout < tx_timeout); } -static int __hns_roce_cmq_send(struct hns_roce_dev *hr_dev, - struct hns_roce_cmq_desc *desc, int num) +static int __hns_roce_cmq_send_one(struct hns_roce_dev *hr_dev, + struct hns_roce_cmq_desc *desc, + int num, u32 tx_timeout) { struct hns_roce_v2_priv *priv = hr_dev->priv; struct hns_roce_v2_cmq_ring *csq = &priv->cmq.csq; @@ -1309,8 +1308,6 @@ static int __hns_roce_cmq_send(struct hns_roce_dev *hr_dev, int ret; int i; - spin_lock_bh(&csq->lock); - tail = csq->head; for (i = 0; i < num; i++) { @@ -1324,22 +1321,17 @@ static int __hns_roce_cmq_send(struct hns_roce_dev *hr_dev, atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_CMDS_CNT]); - hns_roce_wait_csq_done(hr_dev, le16_to_cpu(desc->opcode)); + hns_roce_wait_csq_done(hr_dev, tx_timeout); if (hns_roce_cmq_csq_done(hr_dev)) { ret = 0; for (i = 0; i < num; i++) { /* check the result of hardware write back */ - desc[i] = csq->desc[tail++]; + desc_ret = le16_to_cpu(csq->desc[tail++].retval); if (tail == csq->desc_num) tail = 0; - - desc_ret = le16_to_cpu(desc[i].retval); if (likely(desc_ret == CMD_EXEC_SUCCESS)) continue; - dev_err_ratelimited(hr_dev->dev, - "Cmdq IO error, opcode = 0x%x, return = 0x%x.\n", - desc->opcode, desc_ret); ret = hns_roce_cmd_err_convert_errno(desc_ret); } } else { @@ -1354,14 +1346,54 @@ static int __hns_roce_cmq_send(struct hns_roce_dev *hr_dev, ret = -EAGAIN; } - spin_unlock_bh(&csq->lock); - if (ret) atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_CMDS_ERR_CNT]); return ret; } +static int __hns_roce_cmq_send(struct hns_roce_dev *hr_dev, + struct hns_roce_cmq_desc *desc, int num) +{ + struct hns_roce_v2_priv *priv = hr_dev->priv; + struct hns_roce_v2_cmq_ring *csq = &priv->cmq.csq; + u16 opcode = le16_to_cpu(desc->opcode); + u32 tx_timeout = hns_roce_cmdq_tx_timeout(opcode, priv->cmq.tx_timeout); + u8 try_cnt = HNS_ROCE_OPC_POST_MB_TRY_CNT; + u32 rsv_tail; + int ret; + int i; + + while (try_cnt) { + try_cnt--; + + spin_lock_bh(&csq->lock); + rsv_tail = csq->head; + ret = __hns_roce_cmq_send_one(hr_dev, desc, num, tx_timeout); + if (opcode == HNS_ROCE_OPC_POST_MB && ret == -ETIME && + try_cnt) { + spin_unlock_bh(&csq->lock); + mdelay(HNS_ROCE_OPC_POST_MB_RETRY_GAP_MSEC); + continue; + } + + for (i = 0; i < num; i++) { + desc[i] = csq->desc[rsv_tail++]; + if (rsv_tail == csq->desc_num) + rsv_tail = 0; + } + spin_unlock_bh(&csq->lock); + break; + } + + if (ret) + dev_err_ratelimited(hr_dev->dev, + "Cmdq IO error, opcode = 0x%x, return = %d.\n", + opcode, ret); + + return ret; +} + static int hns_roce_cmq_send(struct hns_roce_dev *hr_dev, struct hns_roce_cmq_desc *desc, int num) { diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.h b/drivers/infiniband/hw/hns/hns_roce_hw_v2.h index cbdbc9edbce6..91a5665465ff 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.h +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.h @@ -230,6 +230,8 @@ enum hns_roce_opcode_type { }; #define HNS_ROCE_OPC_POST_MB_TIMEOUT 35000 +#define HNS_ROCE_OPC_POST_MB_TRY_CNT 8 +#define HNS_ROCE_OPC_POST_MB_RETRY_GAP_MSEC 5 struct hns_roce_cmdq_tx_timeout_map { u16 opcode; u32 tx_timeout; From a27c6f46dcec8f697cbf15c8a10f8534c7b8a2c3 Mon Sep 17 00:00:00 2001 From: Kalesh AP Date: Tue, 4 Feb 2025 00:21:22 -0800 Subject: [PATCH 08/15] RDMA/bnxt_re: Fix an issue in bnxt_re_async_notifier In the bnxt_re_async_notifier() callback, the way driver retrieves rdev pointer is wrong. The rdev pointer should be parsed from adev pointer as while registering with the L2 for ULP, driver uses the aux device pointer for the handle. Fixes: 7fea32784068 ("RDMA/bnxt_re: Add Async event handling support") Signed-off-by: Kalesh AP Signed-off-by: Selvin Xavier Link: https://patch.msgid.link/1738657285-23968-2-git-send-email-selvin.xavier@broadcom.com Signed-off-by: Leon Romanovsky --- drivers/infiniband/hw/bnxt_re/main.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/drivers/infiniband/hw/bnxt_re/main.c b/drivers/infiniband/hw/bnxt_re/main.c index e9e4da4dd576..c4c3d67c42cc 100644 --- a/drivers/infiniband/hw/bnxt_re/main.c +++ b/drivers/infiniband/hw/bnxt_re/main.c @@ -396,11 +396,16 @@ static void bnxt_re_dcb_wq_task(struct work_struct *work) static void bnxt_re_async_notifier(void *handle, struct hwrm_async_event_cmpl *cmpl) { - struct bnxt_re_dev *rdev = (struct bnxt_re_dev *)handle; + struct bnxt_re_en_dev_info *en_info = auxiliary_get_drvdata(handle); struct bnxt_re_dcb_work *dcb_work; + struct bnxt_re_dev *rdev; u32 data1, data2; u16 event_id; + rdev = en_info->rdev; + if (!rdev) + return; + event_id = le16_to_cpu(cmpl->event_id); data1 = le32_to_cpu(cmpl->event_data1); data2 = le32_to_cpu(cmpl->event_data2); From f0df225d12fcb049429fb5bf5122afe143c2dd15 Mon Sep 17 00:00:00 2001 From: Kalesh AP Date: Tue, 4 Feb 2025 00:21:23 -0800 Subject: [PATCH 09/15] RDMA/bnxt_re: Add sanity checks on rdev validity There is a possibility that ulp_irq_stop and ulp_irq_start callbacks will be called when the device is in detached state. This can cause a crash due to NULL pointer dereference as the rdev is already freed. Fixes: cc5b9b48d447 ("RDMA/bnxt_re: Recover the device when FW error is detected") Signed-off-by: Kalesh AP Signed-off-by: Selvin Xavier Link: https://patch.msgid.link/1738657285-23968-3-git-send-email-selvin.xavier@broadcom.com Signed-off-by: Leon Romanovsky --- drivers/infiniband/hw/bnxt_re/main.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/drivers/infiniband/hw/bnxt_re/main.c b/drivers/infiniband/hw/bnxt_re/main.c index c4c3d67c42cc..89ac5c21ca7a 100644 --- a/drivers/infiniband/hw/bnxt_re/main.c +++ b/drivers/infiniband/hw/bnxt_re/main.c @@ -438,6 +438,8 @@ static void bnxt_re_stop_irq(void *handle, bool reset) int indx; rdev = en_info->rdev; + if (!rdev) + return; rcfw = &rdev->rcfw; if (reset) { @@ -466,6 +468,8 @@ static void bnxt_re_start_irq(void *handle, struct bnxt_msix_entry *ent) int indx, rc; rdev = en_info->rdev; + if (!rdev) + return; msix_ent = rdev->nqr->msix_entries; rcfw = &rdev->rcfw; if (!ent) { @@ -2438,6 +2442,7 @@ static int bnxt_re_suspend(struct auxiliary_device *adev, pm_message_t state) ibdev_info(&rdev->ibdev, "%s: L2 driver notified to stop en_state 0x%lx", __func__, en_dev->en_state); bnxt_re_remove_device(rdev, BNXT_RE_PRE_RECOVERY_REMOVE, adev); + bnxt_re_update_en_info_rdev(NULL, en_info, adev); mutex_unlock(&bnxt_re_mutex); return 0; From e2f105277411c4ebacd00d4ae1a57f693ba7d22d Mon Sep 17 00:00:00 2001 From: Kalesh AP Date: Tue, 4 Feb 2025 00:21:24 -0800 Subject: [PATCH 10/15] RDMA/bnxt_re: Fix issue in the unload path The cited comment removed the netdev notifier register call from the driver. But, it did not remove the cleanup code from the unload path. As a result, driver unload is not clean and resulted in undesired behaviour. Fixes: d3b15fcc4201 ("RDMA/bnxt_re: Remove deliver net device event") Signed-off-by: Kalesh AP Signed-off-by: Selvin Xavier Link: https://patch.msgid.link/1738657285-23968-4-git-send-email-selvin.xavier@broadcom.com Signed-off-by: Leon Romanovsky --- drivers/infiniband/hw/bnxt_re/bnxt_re.h | 1 - drivers/infiniband/hw/bnxt_re/main.c | 10 ---------- 2 files changed, 11 deletions(-) diff --git a/drivers/infiniband/hw/bnxt_re/bnxt_re.h b/drivers/infiniband/hw/bnxt_re/bnxt_re.h index b91a85a491d0..3721446c6ba4 100644 --- a/drivers/infiniband/hw/bnxt_re/bnxt_re.h +++ b/drivers/infiniband/hw/bnxt_re/bnxt_re.h @@ -187,7 +187,6 @@ struct bnxt_re_dev { #define BNXT_RE_FLAG_ISSUE_ROCE_STATS 29 struct net_device *netdev; struct auxiliary_device *adev; - struct notifier_block nb; unsigned int version, major, minor; struct bnxt_qplib_chip_ctx *chip_ctx; struct bnxt_en_dev *en_dev; diff --git a/drivers/infiniband/hw/bnxt_re/main.c b/drivers/infiniband/hw/bnxt_re/main.c index 89ac5c21ca7a..a94c8c5387d9 100644 --- a/drivers/infiniband/hw/bnxt_re/main.c +++ b/drivers/infiniband/hw/bnxt_re/main.c @@ -1359,7 +1359,6 @@ static struct bnxt_re_dev *bnxt_re_dev_add(struct auxiliary_device *adev, return NULL; } /* Default values */ - rdev->nb.notifier_call = NULL; rdev->netdev = en_dev->net; rdev->en_dev = en_dev; rdev->adev = adev; @@ -2354,15 +2353,6 @@ static int bnxt_re_add_device(struct auxiliary_device *adev, u8 op_type) static void bnxt_re_remove_device(struct bnxt_re_dev *rdev, u8 op_type, struct auxiliary_device *aux_dev) { - if (rdev->nb.notifier_call) { - unregister_netdevice_notifier(&rdev->nb); - rdev->nb.notifier_call = NULL; - } else { - /* If notifier is null, we should have already done a - * clean up before coming here. - */ - return; - } bnxt_re_setup_cc(rdev, false); ib_unregister_device(&rdev->ibdev); bnxt_re_dev_uninit(rdev, op_type); From 8238c7bd84209c8216b1381ab0dbe6db9e203769 Mon Sep 17 00:00:00 2001 From: Selvin Xavier Date: Tue, 4 Feb 2025 00:21:25 -0800 Subject: [PATCH 11/15] RDMA/bnxt_re: Fix the statistics for Gen P7 VF Gen P7 VF support the extended stats and is prevented by a VF check. Fix the check to issue the FW command for GenP7 VFs also. Fixes: 1801d87b3598 ("RDMA/bnxt_re: Support new 5760X P7 devices") Signed-off-by: Selvin Xavier Link: https://patch.msgid.link/1738657285-23968-5-git-send-email-selvin.xavier@broadcom.com Signed-off-by: Leon Romanovsky --- drivers/infiniband/hw/bnxt_re/hw_counters.c | 4 ++-- drivers/infiniband/hw/bnxt_re/qplib_res.h | 8 ++++++++ 2 files changed, 10 insertions(+), 2 deletions(-) diff --git a/drivers/infiniband/hw/bnxt_re/hw_counters.c b/drivers/infiniband/hw/bnxt_re/hw_counters.c index 3ac47f4e6122..f039aefcaf67 100644 --- a/drivers/infiniband/hw/bnxt_re/hw_counters.c +++ b/drivers/infiniband/hw/bnxt_re/hw_counters.c @@ -348,8 +348,8 @@ int bnxt_re_ib_get_hw_stats(struct ib_device *ibdev, goto done; } bnxt_re_copy_err_stats(rdev, stats, err_s); - if (_is_ext_stats_supported(rdev->dev_attr->dev_cap_flags) && - !rdev->is_virtfn) { + if (bnxt_ext_stats_supported(rdev->chip_ctx, rdev->dev_attr->dev_cap_flags, + rdev->is_virtfn)) { rc = bnxt_re_get_ext_stat(rdev, stats); if (rc) { clear_bit(BNXT_RE_FLAG_ISSUE_ROCE_STATS, diff --git a/drivers/infiniband/hw/bnxt_re/qplib_res.h b/drivers/infiniband/hw/bnxt_re/qplib_res.h index be5d907a036b..711990232de1 100644 --- a/drivers/infiniband/hw/bnxt_re/qplib_res.h +++ b/drivers/infiniband/hw/bnxt_re/qplib_res.h @@ -547,6 +547,14 @@ static inline bool _is_ext_stats_supported(u16 dev_cap_flags) CREQ_QUERY_FUNC_RESP_SB_EXT_STATS; } +static inline int bnxt_ext_stats_supported(struct bnxt_qplib_chip_ctx *ctx, + u16 flags, bool virtfn) +{ + /* ext stats supported if cap flag is set AND is a PF OR a Thor2 VF */ + return (_is_ext_stats_supported(flags) && + ((virtfn && bnxt_qplib_is_chip_gen_p7(ctx)) || (!virtfn))); +} + static inline bool _is_hw_retx_supported(u16 dev_cap_flags) { return dev_cap_flags & From 3d8c6f26893d55fab218ad086719de1fc9bb86ba Mon Sep 17 00:00:00 2001 From: Yishai Hadas Date: Mon, 10 Feb 2025 13:31:11 +0200 Subject: [PATCH 12/15] RDMA/mlx5: Fix implicit ODP hang on parent deregistration Fix the destroy_unused_implicit_child_mr() to prevent hanging during parent deregistration as of below [1]. Upon entering destroy_unused_implicit_child_mr(), the reference count for the implicit MR parent is incremented using: refcount_inc_not_zero(). A corresponding decrement must be performed if free_implicit_child_mr_work() is not called. The code has been updated to properly manage the reference count that was incremented. [1] INFO: task python3:2157 blocked for more than 120 seconds. Not tainted 6.12.0-rc7+ #1633 "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. task:python3 state:D stack:0 pid:2157 tgid:2157 ppid:1685 flags:0x00000000 Call Trace: __schedule+0x420/0xd30 schedule+0x47/0x130 __mlx5_ib_dereg_mr+0x379/0x5d0 [mlx5_ib] ? __pfx_autoremove_wake_function+0x10/0x10 ib_dereg_mr_user+0x5f/0x120 [ib_core] ? lock_release+0xc6/0x280 destroy_hw_idr_uobject+0x1d/0x60 [ib_uverbs] uverbs_destroy_uobject+0x58/0x1d0 [ib_uverbs] uobj_destroy+0x3f/0x70 [ib_uverbs] ib_uverbs_cmd_verbs+0x3e4/0xbb0 [ib_uverbs] ? __pfx_uverbs_destroy_def_handler+0x10/0x10 [ib_uverbs] ? lock_acquire+0xc1/0x2f0 ? ib_uverbs_ioctl+0xcb/0x170 [ib_uverbs] ? ib_uverbs_ioctl+0x116/0x170 [ib_uverbs] ? lock_release+0xc6/0x280 ib_uverbs_ioctl+0xe7/0x170 [ib_uverbs] ? ib_uverbs_ioctl+0xcb/0x170 [ib_uverbs] __x64_sys_ioctl+0x1b0/0xa70 ? kmem_cache_free+0x221/0x400 do_syscall_64+0x6b/0x140 entry_SYSCALL_64_after_hwframe+0x76/0x7e RIP: 0033:0x7f20f21f017b RSP: 002b:00007ffcfc4a77c8 EFLAGS: 00000246 ORIG_RAX: 0000000000000010 RAX: ffffffffffffffda RBX: 00007ffcfc4a78d8 RCX: 00007f20f21f017b RDX: 00007ffcfc4a78c0 RSI: 00000000c0181b01 RDI: 0000000000000003 RBP: 00007ffcfc4a78a0 R08: 000056147d125190 R09: 00007f20f1f14c60 R10: 0000000000000001 R11: 0000000000000246 R12: 00007ffcfc4a7890 R13: 000000000000001c R14: 000056147d100fc0 R15: 00007f20e365c9d0 Fixes: d3d930411ce3 ("RDMA/mlx5: Fix implicit ODP use after free") Signed-off-by: Yishai Hadas Reviewed-by: Artemy Kovalyov Link: https://patch.msgid.link/80f2fcd19952dfa7d9981d93fd6359b4471f8278.1739186929.git.leon@kernel.org Signed-off-by: Leon Romanovsky --- drivers/infiniband/hw/mlx5/odp.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/infiniband/hw/mlx5/odp.c b/drivers/infiniband/hw/mlx5/odp.c index f1e23583e6c0..e77c9280c07e 100644 --- a/drivers/infiniband/hw/mlx5/odp.c +++ b/drivers/infiniband/hw/mlx5/odp.c @@ -242,6 +242,7 @@ static void destroy_unused_implicit_child_mr(struct mlx5_ib_mr *mr) if (__xa_cmpxchg(&imr->implicit_children, idx, mr, NULL, GFP_KERNEL) != mr) { xa_unlock(&imr->implicit_children); + mlx5r_deref_odp_mkey(&imr->mmkey); return; } From c534ffda781f44a1c6ac25ef6e0e444da38ca8af Mon Sep 17 00:00:00 2001 From: Patrisious Haddad Date: Mon, 10 Feb 2025 13:32:39 +0200 Subject: [PATCH 13/15] RDMA/mlx5: Fix AH static rate parsing Previously static rate wasn't translated according to our PRM but simply used the 4 lower bytes. Correctly translate static rate value passed in AH creation attribute according to our PRM expected values. In addition change 800GB mapping to zero, which is the PRM specified value. Fixes: e126ba97dba9 ("mlx5: Add driver for Mellanox Connect-IB adapters") Signed-off-by: Patrisious Haddad Reviewed-by: Maor Gottlieb Link: https://patch.msgid.link/18ef4cc5396caf80728341eb74738cd777596f60.1739187089.git.leon@kernel.org Signed-off-by: Leon Romanovsky --- drivers/infiniband/hw/mlx5/ah.c | 3 ++- drivers/infiniband/hw/mlx5/qp.c | 6 +++--- drivers/infiniband/hw/mlx5/qp.h | 1 + 3 files changed, 6 insertions(+), 4 deletions(-) diff --git a/drivers/infiniband/hw/mlx5/ah.c b/drivers/infiniband/hw/mlx5/ah.c index 505bc47fd575..99036afb3aef 100644 --- a/drivers/infiniband/hw/mlx5/ah.c +++ b/drivers/infiniband/hw/mlx5/ah.c @@ -67,7 +67,8 @@ static void create_ib_ah(struct mlx5_ib_dev *dev, struct mlx5_ib_ah *ah, ah->av.tclass = grh->traffic_class; } - ah->av.stat_rate_sl = (rdma_ah_get_static_rate(ah_attr) << 4); + ah->av.stat_rate_sl = + (mlx5r_ib_rate(dev, rdma_ah_get_static_rate(ah_attr)) << 4); if (ah_attr->type == RDMA_AH_ATTR_TYPE_ROCE) { if (init_attr->xmit_slave) diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c index 08d22db8dca9..88724d15705d 100644 --- a/drivers/infiniband/hw/mlx5/qp.c +++ b/drivers/infiniband/hw/mlx5/qp.c @@ -3447,11 +3447,11 @@ static int ib_to_mlx5_rate_map(u8 rate) return 0; } -static int ib_rate_to_mlx5(struct mlx5_ib_dev *dev, u8 rate) +int mlx5r_ib_rate(struct mlx5_ib_dev *dev, u8 rate) { u32 stat_rate_support; - if (rate == IB_RATE_PORT_CURRENT) + if (rate == IB_RATE_PORT_CURRENT || rate == IB_RATE_800_GBPS) return 0; if (rate < IB_RATE_2_5_GBPS || rate > IB_RATE_800_GBPS) @@ -3596,7 +3596,7 @@ static int mlx5_set_path(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp, sizeof(grh->dgid.raw)); } - err = ib_rate_to_mlx5(dev, rdma_ah_get_static_rate(ah)); + err = mlx5r_ib_rate(dev, rdma_ah_get_static_rate(ah)); if (err < 0) return err; MLX5_SET(ads, path, stat_rate, err); diff --git a/drivers/infiniband/hw/mlx5/qp.h b/drivers/infiniband/hw/mlx5/qp.h index b6ee7c3ee1ca..2530e7730635 100644 --- a/drivers/infiniband/hw/mlx5/qp.h +++ b/drivers/infiniband/hw/mlx5/qp.h @@ -56,4 +56,5 @@ int mlx5_core_xrcd_dealloc(struct mlx5_ib_dev *dev, u32 xrcdn); int mlx5_ib_qp_set_counter(struct ib_qp *qp, struct rdma_counter *counter); int mlx5_ib_qp_event_init(void); void mlx5_ib_qp_event_cleanup(void); +int mlx5r_ib_rate(struct mlx5_ib_dev *dev, u8 rate); #endif /* _MLX5_IB_QP_H */ From e1a0bdbdfdf08428f0ede5ae49c7f4139ac73ef5 Mon Sep 17 00:00:00 2001 From: Patrisious Haddad Date: Thu, 20 Feb 2025 08:47:10 +0200 Subject: [PATCH 14/15] RDMA/mlx5: Fix bind QP error cleanup flow When there is a failure during bind QP, the cleanup flow destroys the counter regardless if it is the one that created it or not, which is problematic since if it isn't the one that created it, that counter could still be in use. Fix that by destroying the counter only if it was created during this call. Fixes: 45842fc627c7 ("IB/mlx5: Support statistic q counter configuration") Signed-off-by: Patrisious Haddad Reviewed-by: Mark Zhang Link: https://patch.msgid.link/25dfefddb0ebefa668c32e06a94d84e3216257cf.1740033937.git.leon@kernel.org Reviewed-by: Zhu Yanjun Signed-off-by: Leon Romanovsky --- drivers/infiniband/hw/mlx5/counters.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) diff --git a/drivers/infiniband/hw/mlx5/counters.c b/drivers/infiniband/hw/mlx5/counters.c index 4f6c1968a2ee..81cfa74147a1 100644 --- a/drivers/infiniband/hw/mlx5/counters.c +++ b/drivers/infiniband/hw/mlx5/counters.c @@ -546,6 +546,7 @@ static int mlx5_ib_counter_bind_qp(struct rdma_counter *counter, struct ib_qp *qp) { struct mlx5_ib_dev *dev = to_mdev(qp->device); + bool new = false; int err; if (!counter->id) { @@ -560,6 +561,7 @@ static int mlx5_ib_counter_bind_qp(struct rdma_counter *counter, return err; counter->id = MLX5_GET(alloc_q_counter_out, out, counter_set_id); + new = true; } err = mlx5_ib_qp_set_counter(qp, counter); @@ -569,8 +571,10 @@ static int mlx5_ib_counter_bind_qp(struct rdma_counter *counter, return 0; fail_set_counter: - mlx5_ib_counter_dealloc(counter); - counter->id = 0; + if (new) { + mlx5_ib_counter_dealloc(counter); + counter->id = 0; + } return err; } From b66535356a4834a234f99e16a97eb51f2c6c5a7d Mon Sep 17 00:00:00 2001 From: Kashyap Desai Date: Sat, 22 Feb 2025 07:20:21 -0800 Subject: [PATCH 15/15] RDMA/bnxt_re: Fix the page details for the srq created by kernel consumers While using nvme target with use_srq on, below kernel panic is noticed. [ 549.698111] bnxt_en 0000:41:00.0 enp65s0np0: FEC autoneg off encoding: Clause 91 RS(544,514) [ 566.393619] Oops: divide error: 0000 [#1] PREEMPT SMP NOPTI .. [ 566.393799] [ 566.393807] ? __die_body+0x1a/0x60 [ 566.393823] ? die+0x38/0x60 [ 566.393835] ? do_trap+0xe4/0x110 [ 566.393847] ? bnxt_qplib_alloc_init_hwq+0x1d4/0x580 [bnxt_re] [ 566.393867] ? bnxt_qplib_alloc_init_hwq+0x1d4/0x580 [bnxt_re] [ 566.393881] ? do_error_trap+0x7c/0x120 [ 566.393890] ? bnxt_qplib_alloc_init_hwq+0x1d4/0x580 [bnxt_re] [ 566.393911] ? exc_divide_error+0x34/0x50 [ 566.393923] ? bnxt_qplib_alloc_init_hwq+0x1d4/0x580 [bnxt_re] [ 566.393939] ? asm_exc_divide_error+0x16/0x20 [ 566.393966] ? bnxt_qplib_alloc_init_hwq+0x1d4/0x580 [bnxt_re] [ 566.393997] bnxt_qplib_create_srq+0xc9/0x340 [bnxt_re] [ 566.394040] bnxt_re_create_srq+0x335/0x3b0 [bnxt_re] [ 566.394057] ? srso_return_thunk+0x5/0x5f [ 566.394068] ? __init_swait_queue_head+0x4a/0x60 [ 566.394090] ib_create_srq_user+0xa7/0x150 [ib_core] [ 566.394147] nvmet_rdma_queue_connect+0x7d0/0xbe0 [nvmet_rdma] [ 566.394174] ? lock_release+0x22c/0x3f0 [ 566.394187] ? srso_return_thunk+0x5/0x5f Page size and shift info is set only for the user space SRQs. Set page size and page shift for kernel space SRQs also. Fixes: 0c4dcd602817 ("RDMA/bnxt_re: Refactor hardware queue memory allocation") Signed-off-by: Kashyap Desai Signed-off-by: Selvin Xavier Link: https://patch.msgid.link/1740237621-29291-1-git-send-email-selvin.xavier@broadcom.com Signed-off-by: Leon Romanovsky --- drivers/infiniband/hw/bnxt_re/ib_verbs.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.c b/drivers/infiniband/hw/bnxt_re/ib_verbs.c index 2de101d6e825..6f5db32082dd 100644 --- a/drivers/infiniband/hw/bnxt_re/ib_verbs.c +++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.c @@ -1870,6 +1870,8 @@ int bnxt_re_create_srq(struct ib_srq *ib_srq, srq->qplib_srq.threshold = srq_init_attr->attr.srq_limit; srq->srq_limit = srq_init_attr->attr.srq_limit; srq->qplib_srq.eventq_hw_ring_id = rdev->nqr->nq[0].ring_id; + srq->qplib_srq.sg_info.pgsize = PAGE_SIZE; + srq->qplib_srq.sg_info.pgshft = PAGE_SHIFT; nq = &rdev->nqr->nq[0]; if (udata) {