From a568814a55a0e82bbc7c7b51333d0c38e8fb5520 Mon Sep 17 00:00:00 2001 From: Leon Romanovsky Date: Sun, 9 May 2021 14:39:21 +0300 Subject: [PATCH 01/10] RDMA/siw: Properly check send and receive CQ pointers The check for the NULL of pointer received from container_of() is incorrect by definition as it points to some offset from NULL. Change such check with proper NULL check of SIW QP attributes. Fixes: 303ae1cdfdf7 ("rdma/siw: application interface") Link: https://lore.kernel.org/r/a7535a82925f6f4c1f062abaa294f3ae6e54bdd2.1620560310.git.leonro@nvidia.com Signed-off-by: Leon Romanovsky Reviewed-by: Bernard Metzler Signed-off-by: Jason Gunthorpe --- drivers/infiniband/sw/siw/siw_verbs.c | 9 +++------ 1 file changed, 3 insertions(+), 6 deletions(-) diff --git a/drivers/infiniband/sw/siw/siw_verbs.c b/drivers/infiniband/sw/siw/siw_verbs.c index d2313efb26db..917c8a919f38 100644 --- a/drivers/infiniband/sw/siw/siw_verbs.c +++ b/drivers/infiniband/sw/siw/siw_verbs.c @@ -300,7 +300,6 @@ struct ib_qp *siw_create_qp(struct ib_pd *pd, struct siw_ucontext *uctx = rdma_udata_to_drv_context(udata, struct siw_ucontext, base_ucontext); - struct siw_cq *scq = NULL, *rcq = NULL; unsigned long flags; int num_sqe, num_rqe, rv = 0; size_t length; @@ -343,10 +342,8 @@ struct ib_qp *siw_create_qp(struct ib_pd *pd, rv = -EINVAL; goto err_out; } - scq = to_siw_cq(attrs->send_cq); - rcq = to_siw_cq(attrs->recv_cq); - if (!scq || (!rcq && !attrs->srq)) { + if (!attrs->send_cq || (!attrs->recv_cq && !attrs->srq)) { siw_dbg(base_dev, "send CQ or receive CQ invalid\n"); rv = -EINVAL; goto err_out; @@ -401,8 +398,8 @@ struct ib_qp *siw_create_qp(struct ib_pd *pd, } } qp->pd = pd; - qp->scq = scq; - qp->rcq = rcq; + qp->scq = to_siw_cq(attrs->send_cq); + qp->rcq = to_siw_cq(attrs->recv_cq); if (attrs->srq) { /* From a3d83276d98886879b5bf7b30b7c29882754e4df Mon Sep 17 00:00:00 2001 From: Leon Romanovsky Date: Sun, 9 May 2021 14:41:38 +0300 Subject: [PATCH 02/10] RDMA/siw: Release xarray entry The xarray entry is allocated in siw_qp_add(), but release was missed in case zero-sized SQ was discovered. Fixes: 661f385961f0 ("RDMA/siw: Fix handling of zero-sized Read and Receive Queues.") Link: https://lore.kernel.org/r/f070b59d5a1114d5a4e830346755c2b3f141cde5.1620560472.git.leonro@nvidia.com Signed-off-by: Leon Romanovsky Reviewed-by: Bernard Metzler Signed-off-by: Jason Gunthorpe --- drivers/infiniband/sw/siw/siw_verbs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/infiniband/sw/siw/siw_verbs.c b/drivers/infiniband/sw/siw/siw_verbs.c index 917c8a919f38..3f175f220a22 100644 --- a/drivers/infiniband/sw/siw/siw_verbs.c +++ b/drivers/infiniband/sw/siw/siw_verbs.c @@ -375,7 +375,7 @@ struct ib_qp *siw_create_qp(struct ib_pd *pd, else { /* Zero sized SQ is not supported */ rv = -EINVAL; - goto err_out; + goto err_out_xa; } if (num_rqe) num_rqe = roundup_pow_of_two(num_rqe); From 54d87913f147a983589923c7f651f97de9af5be1 Mon Sep 17 00:00:00 2001 From: Leon Romanovsky Date: Mon, 10 May 2021 17:46:00 +0300 Subject: [PATCH 03/10] RDMA/core: Prevent divide-by-zero error triggered by the user The user_entry_size is supplied by the user and later used as a denominator to calculate number of entries. The zero supplied by the user will trigger the following divide-by-zero error: divide error: 0000 [#1] SMP KASAN PTI CPU: 4 PID: 497 Comm: c_repro Not tainted 5.13.0-rc1+ #281 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014 RIP: 0010:ib_uverbs_handler_UVERBS_METHOD_QUERY_GID_TABLE+0x1b1/0x510 Code: 87 59 03 00 00 e8 9f ab 1e ff 48 8d bd a8 00 00 00 e8 d3 70 41 ff 44 0f b7 b5 a8 00 00 00 e8 86 ab 1e ff 31 d2 4c 89 f0 31 ff <49> f7 f5 48 89 d6 48 89 54 24 10 48 89 04 24 e8 1b ad 1e ff 48 8b RSP: 0018:ffff88810416f828 EFLAGS: 00010246 RAX: 0000000000000008 RBX: 1ffff1102082df09 RCX: ffffffff82183f3d RDX: 0000000000000000 RSI: ffff888105f2da00 RDI: 0000000000000000 RBP: ffff88810416fa98 R08: 0000000000000001 R09: ffffed102082df5f R10: ffff88810416faf7 R11: ffffed102082df5e R12: 0000000000000000 R13: 0000000000000000 R14: 0000000000000008 R15: ffff88810416faf0 FS: 00007f5715efa740(0000) GS:ffff88811a700000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000020000840 CR3: 000000010c2e0001 CR4: 0000000000370ea0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Call Trace: ? ib_uverbs_handler_UVERBS_METHOD_INFO_HANDLES+0x4b0/0x4b0 ib_uverbs_cmd_verbs+0x1546/0x1940 ib_uverbs_ioctl+0x186/0x240 __x64_sys_ioctl+0x38a/0x1220 do_syscall_64+0x3f/0x80 entry_SYSCALL_64_after_hwframe+0x44/0xae Fixes: 9f85cbe50aa0 ("RDMA/uverbs: Expose the new GID query API to user space") Link: https://lore.kernel.org/r/b971cc70a8b240a8b5eda33c99fa0558a0071be2.1620657876.git.leonro@nvidia.com Reviewed-by: Jason Gunthorpe Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/core/uverbs_std_types_device.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/drivers/infiniband/core/uverbs_std_types_device.c b/drivers/infiniband/core/uverbs_std_types_device.c index 9ec6971056fa..a03021d94e11 100644 --- a/drivers/infiniband/core/uverbs_std_types_device.c +++ b/drivers/infiniband/core/uverbs_std_types_device.c @@ -331,6 +331,9 @@ static int UVERBS_HANDLER(UVERBS_METHOD_QUERY_GID_TABLE)( if (ret) return ret; + if (!user_entry_size) + return -EINVAL; + max_entries = uverbs_attr_ptr_get_array_size( attrs, UVERBS_ATTR_QUERY_GID_TABLE_RESP_ENTRIES, user_entry_size); From 67f29896fdc83298eed5a6576ff8f9873f709228 Mon Sep 17 00:00:00 2001 From: Leon Romanovsky Date: Tue, 11 May 2021 10:26:03 +0300 Subject: [PATCH 04/10] RDMA/rxe: Clear all QP fields if creation failed rxe_qp_do_cleanup() relies on valid pointer values in QP for the properly created ones, but in case rxe_qp_from_init() failed it was filled with garbage and caused tot the following error. refcount_t: underflow; use-after-free. WARNING: CPU: 1 PID: 12560 at lib/refcount.c:28 refcount_warn_saturate+0x1d1/0x1e0 lib/refcount.c:28 Modules linked in: CPU: 1 PID: 12560 Comm: syz-executor.4 Not tainted 5.12.0-syzkaller #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 RIP: 0010:refcount_warn_saturate+0x1d1/0x1e0 lib/refcount.c:28 Code: e9 db fe ff ff 48 89 df e8 2c c2 ea fd e9 8a fe ff ff e8 72 6a a7 fd 48 c7 c7 e0 b2 c1 89 c6 05 dc 3a e6 09 01 e8 ee 74 fb 04 <0f> 0b e9 af fe ff ff 0f 1f 84 00 00 00 00 00 41 56 41 55 41 54 55 RSP: 0018:ffffc900097ceba8 EFLAGS: 00010286 RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000000 RDX: 0000000000040000 RSI: ffffffff815bb075 RDI: fffff520012f9d67 RBP: 0000000000000003 R08: 0000000000000000 R09: 0000000000000000 R10: ffffffff815b4eae R11: 0000000000000000 R12: ffff8880322a4800 R13: ffff8880322a4940 R14: ffff888033044e00 R15: 0000000000000000 FS: 00007f6eb2be3700(0000) GS:ffff8880b9d00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00007fdbe5d41000 CR3: 000000001d181000 CR4: 00000000001506e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Call Trace: __refcount_sub_and_test include/linux/refcount.h:283 [inline] __refcount_dec_and_test include/linux/refcount.h:315 [inline] refcount_dec_and_test include/linux/refcount.h:333 [inline] kref_put include/linux/kref.h:64 [inline] rxe_qp_do_cleanup+0x96f/0xaf0 drivers/infiniband/sw/rxe/rxe_qp.c:805 execute_in_process_context+0x37/0x150 kernel/workqueue.c:3327 rxe_elem_release+0x9f/0x180 drivers/infiniband/sw/rxe/rxe_pool.c:391 kref_put include/linux/kref.h:65 [inline] rxe_create_qp+0x2cd/0x310 drivers/infiniband/sw/rxe/rxe_verbs.c:425 _ib_create_qp drivers/infiniband/core/core_priv.h:331 [inline] ib_create_named_qp+0x2ad/0x1370 drivers/infiniband/core/verbs.c:1231 ib_create_qp include/rdma/ib_verbs.h:3644 [inline] create_mad_qp+0x177/0x2d0 drivers/infiniband/core/mad.c:2920 ib_mad_port_open drivers/infiniband/core/mad.c:3001 [inline] ib_mad_init_device+0xd6f/0x1400 drivers/infiniband/core/mad.c:3092 add_client_context+0x405/0x5e0 drivers/infiniband/core/device.c:717 enable_device_and_get+0x1cd/0x3b0 drivers/infiniband/core/device.c:1331 ib_register_device drivers/infiniband/core/device.c:1413 [inline] ib_register_device+0x7c7/0xa50 drivers/infiniband/core/device.c:1365 rxe_register_device+0x3d5/0x4a0 drivers/infiniband/sw/rxe/rxe_verbs.c:1147 rxe_add+0x12fe/0x16d0 drivers/infiniband/sw/rxe/rxe.c:247 rxe_net_add+0x8c/0xe0 drivers/infiniband/sw/rxe/rxe_net.c:503 rxe_newlink drivers/infiniband/sw/rxe/rxe.c:269 [inline] rxe_newlink+0xb7/0xe0 drivers/infiniband/sw/rxe/rxe.c:250 nldev_newlink+0x30e/0x550 drivers/infiniband/core/nldev.c:1555 rdma_nl_rcv_msg+0x36d/0x690 drivers/infiniband/core/netlink.c:195 rdma_nl_rcv_skb drivers/infiniband/core/netlink.c:239 [inline] rdma_nl_rcv+0x2ee/0x430 drivers/infiniband/core/netlink.c:259 netlink_unicast_kernel net/netlink/af_netlink.c:1312 [inline] netlink_unicast+0x533/0x7d0 net/netlink/af_netlink.c:1338 netlink_sendmsg+0x856/0xd90 net/netlink/af_netlink.c:1927 sock_sendmsg_nosec net/socket.c:654 [inline] sock_sendmsg+0xcf/0x120 net/socket.c:674 ____sys_sendmsg+0x6e8/0x810 net/socket.c:2350 ___sys_sendmsg+0xf3/0x170 net/socket.c:2404 __sys_sendmsg+0xe5/0x1b0 net/socket.c:2433 do_syscall_64+0x3a/0xb0 arch/x86/entry/common.c:47 entry_SYSCALL_64_after_hwframe+0x44/0xae Fixes: 8700e3e7c485 ("Soft RoCE driver") Link: https://lore.kernel.org/r/7bf8d548764d406dbbbaf4b574960ebfd5af8387.1620717918.git.leonro@nvidia.com Reported-by: syzbot+36a7f280de4e11c6f04e@syzkaller.appspotmail.com Signed-off-by: Leon Romanovsky Reviewed-by: Zhu Yanjun Signed-off-by: Jason Gunthorpe --- drivers/infiniband/sw/rxe/rxe_qp.c | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/drivers/infiniband/sw/rxe/rxe_qp.c b/drivers/infiniband/sw/rxe/rxe_qp.c index 34ae957a315c..b0f350d674fd 100644 --- a/drivers/infiniband/sw/rxe/rxe_qp.c +++ b/drivers/infiniband/sw/rxe/rxe_qp.c @@ -242,6 +242,7 @@ static int rxe_qp_init_req(struct rxe_dev *rxe, struct rxe_qp *qp, if (err) { vfree(qp->sq.queue->buf); kfree(qp->sq.queue); + qp->sq.queue = NULL; return err; } @@ -295,6 +296,7 @@ static int rxe_qp_init_resp(struct rxe_dev *rxe, struct rxe_qp *qp, if (err) { vfree(qp->rq.queue->buf); kfree(qp->rq.queue); + qp->rq.queue = NULL; return err; } } @@ -355,6 +357,11 @@ int rxe_qp_from_init(struct rxe_dev *rxe, struct rxe_qp *qp, struct rxe_pd *pd, err2: rxe_queue_cleanup(qp->sq.queue); err1: + qp->pd = NULL; + qp->rcq = NULL; + qp->scq = NULL; + qp->srq = NULL; + if (srq) rxe_drop_ref(srq); rxe_drop_ref(scq); From 6863b4d7bf19a54e23fc5838b7e66d954444289d Mon Sep 17 00:00:00 2001 From: Maor Gottlieb Date: Tue, 11 May 2021 08:48:27 +0300 Subject: [PATCH 05/10] RDMA/mlx5: Verify that DM operation is reasonable Fix the complaint from smatch by verifing that the user requested DM operation is not greater than 31. divers/infiniband/hw/mlx5/dm.c:220 mlx5_ib_handler_MLX5_IB_METHOD_DM_MAP_OP_ADDR() error: undefined (user controlled) shift '(((1))) << op' Fixes: cea85fa5dbc2 ("RDMA/mlx5: Add support in MEMIC operations") Link: https://lore.kernel.org/r/458b1d7710c3cf01360c8771893f483665569786.1620711734.git.leonro@nvidia.com Reported-by: Dan Carpenter Signed-off-by: Maor Gottlieb Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/dm.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/drivers/infiniband/hw/mlx5/dm.c b/drivers/infiniband/hw/mlx5/dm.c index 094bf85589db..001d766cf291 100644 --- a/drivers/infiniband/hw/mlx5/dm.c +++ b/drivers/infiniband/hw/mlx5/dm.c @@ -217,6 +217,9 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DM_MAP_OP_ADDR)( if (err) return err; + if (op >= BITS_PER_TYPE(u32)) + return -EOPNOTSUPP; + if (!(MLX5_CAP_DEV_MEM(dev->mdev, memic_operations) & BIT(op))) return -EOPNOTSUPP; From 97f30d324ce6645a4de4ffb71e4ae9b8ca36ff04 Mon Sep 17 00:00:00 2001 From: Maor Gottlieb Date: Tue, 11 May 2021 08:48:29 +0300 Subject: [PATCH 06/10] RDMA/mlx5: Recover from fatal event in dual port mode When there is fatal event on the slave port, the device is marked as not active. We need to mark it as active again when the slave is recovered to regain full functionality. Fixes: d69a24e03659 ("IB/mlx5: Move IB event processing onto a workqueue") Link: https://lore.kernel.org/r/8906754455bb23019ef223c725d2c0d38acfb80b.1620711734.git.leonro@nvidia.com Signed-off-by: Maor Gottlieb Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/main.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index 6d1dd09a4388..644d5d0ac544 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -4419,6 +4419,7 @@ static int mlx5r_mp_probe(struct auxiliary_device *adev, if (bound) { rdma_roce_rescan_device(&dev->ib_dev); + mpi->ibdev->ib_active = true; break; } } From dc07628bd2bbc1da768e265192c28ebd301f509d Mon Sep 17 00:00:00 2001 From: Leon Romanovsky Date: Tue, 11 May 2021 08:48:31 +0300 Subject: [PATCH 07/10] RDMA/rxe: Return CQE error if invalid lkey was supplied RXE is missing update of WQE status in LOCAL_WRITE failures. This caused the following kernel panic if someone sent an atomic operation with an explicitly wrong lkey. [leonro@vm ~]$ mkt test test_atomic_invalid_lkey (tests.test_atomic.AtomicTest) ... WARNING: CPU: 5 PID: 263 at drivers/infiniband/sw/rxe/rxe_comp.c:740 rxe_completer+0x1a6d/0x2e30 [rdma_rxe] Modules linked in: crc32_generic rdma_rxe ip6_udp_tunnel udp_tunnel rdma_ucm rdma_cm ib_umad ib_ipoib iw_cm ib_cm mlx5_ib ib_uverbs ib_core mlx5_core ptp pps_core CPU: 5 PID: 263 Comm: python3 Not tainted 5.13.0-rc1+ #2936 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014 RIP: 0010:rxe_completer+0x1a6d/0x2e30 [rdma_rxe] Code: 03 0f 8e 65 0e 00 00 3b 93 10 06 00 00 0f 84 82 0a 00 00 4c 89 ff 4c 89 44 24 38 e8 2d 74 a9 e1 4c 8b 44 24 38 e9 1c f5 ff ff <0f> 0b e9 0c e8 ff ff b8 05 00 00 00 41 bf 05 00 00 00 e9 ab e7 ff RSP: 0018:ffff8880158af090 EFLAGS: 00010246 RAX: 0000000000000000 RBX: ffff888016a78000 RCX: ffffffffa0cf1652 RDX: 1ffff9200004b442 RSI: 0000000000000004 RDI: ffffc9000025a210 RBP: dffffc0000000000 R08: 00000000ffffffea R09: ffff88801617740b R10: ffffed1002c2ee81 R11: 0000000000000007 R12: ffff88800f3b63e8 R13: ffff888016a78008 R14: ffffc9000025a180 R15: 000000000000000c FS: 00007f88b622a740(0000) GS:ffff88806d540000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00007f88b5a1fa10 CR3: 000000000d848004 CR4: 0000000000370ea0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Call Trace: rxe_do_task+0x130/0x230 [rdma_rxe] rxe_rcv+0xb11/0x1df0 [rdma_rxe] rxe_loopback+0x157/0x1e0 [rdma_rxe] rxe_responder+0x5532/0x7620 [rdma_rxe] rxe_do_task+0x130/0x230 [rdma_rxe] rxe_rcv+0x9c8/0x1df0 [rdma_rxe] rxe_loopback+0x157/0x1e0 [rdma_rxe] rxe_requester+0x1efd/0x58c0 [rdma_rxe] rxe_do_task+0x130/0x230 [rdma_rxe] rxe_post_send+0x998/0x1860 [rdma_rxe] ib_uverbs_post_send+0xd5f/0x1220 [ib_uverbs] ib_uverbs_write+0x847/0xc80 [ib_uverbs] vfs_write+0x1c5/0x840 ksys_write+0x176/0x1d0 do_syscall_64+0x3f/0x80 entry_SYSCALL_64_after_hwframe+0x44/0xae Fixes: 8700e3e7c485 ("Soft RoCE driver") Link: https://lore.kernel.org/r/11e7b553f3a6f5371c6bb3f57c494bb52b88af99.1620711734.git.leonro@nvidia.com Signed-off-by: Leon Romanovsky Acked-by: Zhu Yanjun Signed-off-by: Jason Gunthorpe --- drivers/infiniband/sw/rxe/rxe_comp.c | 16 ++++++++++------ 1 file changed, 10 insertions(+), 6 deletions(-) diff --git a/drivers/infiniband/sw/rxe/rxe_comp.c b/drivers/infiniband/sw/rxe/rxe_comp.c index 2af26737d32d..a6712e373eed 100644 --- a/drivers/infiniband/sw/rxe/rxe_comp.c +++ b/drivers/infiniband/sw/rxe/rxe_comp.c @@ -346,13 +346,15 @@ static inline enum comp_state do_read(struct rxe_qp *qp, ret = copy_data(qp->pd, IB_ACCESS_LOCAL_WRITE, &wqe->dma, payload_addr(pkt), payload_size(pkt), to_mr_obj, NULL); - if (ret) + if (ret) { + wqe->status = IB_WC_LOC_PROT_ERR; return COMPST_ERROR; + } if (wqe->dma.resid == 0 && (pkt->mask & RXE_END_MASK)) return COMPST_COMP_ACK; - else - return COMPST_UPDATE_COMP; + + return COMPST_UPDATE_COMP; } static inline enum comp_state do_atomic(struct rxe_qp *qp, @@ -366,10 +368,12 @@ static inline enum comp_state do_atomic(struct rxe_qp *qp, ret = copy_data(qp->pd, IB_ACCESS_LOCAL_WRITE, &wqe->dma, &atomic_orig, sizeof(u64), to_mr_obj, NULL); - if (ret) + if (ret) { + wqe->status = IB_WC_LOC_PROT_ERR; return COMPST_ERROR; - else - return COMPST_COMP_ACK; + } + + return COMPST_COMP_ACK; } static void make_send_cqe(struct rxe_qp *qp, struct rxe_send_wqe *wqe, From 889d916b6f8a48b8c9489fffcad3b78eedd01a51 Mon Sep 17 00:00:00 2001 From: Shay Drory Date: Tue, 11 May 2021 08:48:28 +0300 Subject: [PATCH 08/10] RDMA/core: Don't access cm_id after its destruction restrack should only be attached to a cm_id while the ID has a valid device pointer. It is set up when the device is first loaded, but not cleared when the device is removed. There is also two copies of the device pointer, one private and one in the public API, and these were left out of sync. Make everything go to NULL together and manipulate restrack right around the device assignments. Found by syzcaller: BUG: KASAN: wild-memory-access in __list_del include/linux/list.h:112 [inline] BUG: KASAN: wild-memory-access in __list_del_entry include/linux/list.h:135 [inline] BUG: KASAN: wild-memory-access in list_del include/linux/list.h:146 [inline] BUG: KASAN: wild-memory-access in cma_cancel_listens drivers/infiniband/core/cma.c:1767 [inline] BUG: KASAN: wild-memory-access in cma_cancel_operation drivers/infiniband/core/cma.c:1795 [inline] BUG: KASAN: wild-memory-access in cma_cancel_operation+0x1f4/0x4b0 drivers/infiniband/core/cma.c:1783 Write of size 8 at addr dead000000000108 by task syz-executor716/334 CPU: 0 PID: 334 Comm: syz-executor716 Not tainted 5.11.0+ #271 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014 Call Trace: __dump_stack lib/dump_stack.c:79 [inline] dump_stack+0xbe/0xf9 lib/dump_stack.c:120 __kasan_report mm/kasan/report.c:400 [inline] kasan_report.cold+0x5f/0xd5 mm/kasan/report.c:413 __list_del include/linux/list.h:112 [inline] __list_del_entry include/linux/list.h:135 [inline] list_del include/linux/list.h:146 [inline] cma_cancel_listens drivers/infiniband/core/cma.c:1767 [inline] cma_cancel_operation drivers/infiniband/core/cma.c:1795 [inline] cma_cancel_operation+0x1f4/0x4b0 drivers/infiniband/core/cma.c:1783 _destroy_id+0x29/0x460 drivers/infiniband/core/cma.c:1862 ucma_close_id+0x36/0x50 drivers/infiniband/core/ucma.c:185 ucma_destroy_private_ctx+0x58d/0x5b0 drivers/infiniband/core/ucma.c:576 ucma_close+0x91/0xd0 drivers/infiniband/core/ucma.c:1797 __fput+0x169/0x540 fs/file_table.c:280 task_work_run+0xb7/0x100 kernel/task_work.c:140 exit_task_work include/linux/task_work.h:30 [inline] do_exit+0x7da/0x17f0 kernel/exit.c:825 do_group_exit+0x9e/0x190 kernel/exit.c:922 __do_sys_exit_group kernel/exit.c:933 [inline] __se_sys_exit_group kernel/exit.c:931 [inline] __x64_sys_exit_group+0x2d/0x30 kernel/exit.c:931 do_syscall_64+0x2d/0x40 arch/x86/entry/common.c:46 entry_SYSCALL_64_after_hwframe+0x44/0xa9 Fixes: 255d0c14b375 ("RDMA/cma: rdma_bind_addr() leaks a cma_dev reference count") Link: https://lore.kernel.org/r/3352ee288fe34f2b44220457a29bfc0548686363.1620711734.git.leonro@nvidia.com Signed-off-by: Shay Drory Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/core/cma.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c index 2b9ffc21cbc4..ab148a696c0c 100644 --- a/drivers/infiniband/core/cma.c +++ b/drivers/infiniband/core/cma.c @@ -473,6 +473,7 @@ static void cma_release_dev(struct rdma_id_private *id_priv) list_del(&id_priv->list); cma_dev_put(id_priv->cma_dev); id_priv->cma_dev = NULL; + id_priv->id.device = NULL; if (id_priv->id.route.addr.dev_addr.sgid_attr) { rdma_put_gid_attr(id_priv->id.route.addr.dev_addr.sgid_attr); id_priv->id.route.addr.dev_addr.sgid_attr = NULL; @@ -1860,6 +1861,7 @@ static void _destroy_id(struct rdma_id_private *id_priv, iw_destroy_cm_id(id_priv->cm_id.iw); } cma_leave_mc_groups(id_priv); + rdma_restrack_del(&id_priv->res); cma_release_dev(id_priv); } @@ -1873,7 +1875,6 @@ static void _destroy_id(struct rdma_id_private *id_priv, kfree(id_priv->id.route.path_rec); put_net(id_priv->id.route.addr.dev_addr.net); - rdma_restrack_del(&id_priv->res); kfree(id_priv); } @@ -3774,7 +3775,7 @@ int rdma_listen(struct rdma_cm_id *id, int backlog) } id_priv->backlog = backlog; - if (id->device) { + if (id_priv->cma_dev) { if (rdma_cap_ib_cm(id->device, 1)) { ret = cma_ib_listen(id_priv); if (ret) From cfa3b797118eda7d68f9ede9b1a0279192aca653 Mon Sep 17 00:00:00 2001 From: Maor Gottlieb Date: Wed, 19 May 2021 11:41:32 +0300 Subject: [PATCH 09/10] RDMA/mlx5: Fix query DCT via DEVX When executing DEVX command to query QP object, we need to take the QP type from the mlx5_ib_qp struct which hold the driver specific QP types as well, such as DC. Fixes: 34613eb1d2ad ("IB/mlx5: Enable modify and query verbs objects via DEVX") Link: https://lore.kernel.org/r/6eee15d63f09bb70787488e0cf96216e2957f5aa.1621413654.git.leonro@nvidia.com Reviewed-by: Yishai Hadas Signed-off-by: Maor Gottlieb Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/hw/mlx5/devx.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) diff --git a/drivers/infiniband/hw/mlx5/devx.c b/drivers/infiniband/hw/mlx5/devx.c index a0b677accd96..eb9b0a2707f8 100644 --- a/drivers/infiniband/hw/mlx5/devx.c +++ b/drivers/infiniband/hw/mlx5/devx.c @@ -630,9 +630,8 @@ static bool devx_is_valid_obj_id(struct uverbs_attr_bundle *attrs, case UVERBS_OBJECT_QP: { struct mlx5_ib_qp *qp = to_mqp(uobj->object); - enum ib_qp_type qp_type = qp->ibqp.qp_type; - if (qp_type == IB_QPT_RAW_PACKET || + if (qp->type == IB_QPT_RAW_PACKET || (qp->flags & IB_QP_CREATE_SOURCE_QPN)) { struct mlx5_ib_raw_packet_qp *raw_packet_qp = &qp->raw_packet_qp; @@ -649,10 +648,9 @@ static bool devx_is_valid_obj_id(struct uverbs_attr_bundle *attrs, sq->tisn) == obj_id); } - if (qp_type == MLX5_IB_QPT_DCT) + if (qp->type == MLX5_IB_QPT_DCT) return get_enc_obj_id(MLX5_CMD_OP_CREATE_DCT, qp->dct.mdct.mqp.qpn) == obj_id; - return get_enc_obj_id(MLX5_CMD_OP_CREATE_QP, qp->ibqp.qp_num) == obj_id; } From 463a3f66473b58d71428a1c3ce69ea52c05440e5 Mon Sep 17 00:00:00 2001 From: Dan Carpenter Date: Fri, 14 May 2021 17:18:10 +0300 Subject: [PATCH 10/10] RDMA/uverbs: Fix a NULL vs IS_ERR() bug The uapi_get_object() function returns error pointers, it never returns NULL. Fixes: 149d3845f4a5 ("RDMA/uverbs: Add a method to introspect handles in a context") Link: https://lore.kernel.org/r/YJ6Got+U7lz+3n9a@mwanda Signed-off-by: Dan Carpenter Reviewed-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/core/uverbs_std_types_device.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/infiniband/core/uverbs_std_types_device.c b/drivers/infiniband/core/uverbs_std_types_device.c index a03021d94e11..049684880ae0 100644 --- a/drivers/infiniband/core/uverbs_std_types_device.c +++ b/drivers/infiniband/core/uverbs_std_types_device.c @@ -117,8 +117,8 @@ static int UVERBS_HANDLER(UVERBS_METHOD_INFO_HANDLES)( return ret; uapi_object = uapi_get_object(attrs->ufile->device->uapi, object_id); - if (!uapi_object) - return -EINVAL; + if (IS_ERR(uapi_object)) + return PTR_ERR(uapi_object); handles = gather_objects_handle(attrs->ufile, uapi_object, attrs, out_len, &total);