stable.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
To: linux-kernel@vger.kernel.org
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	stable@vger.kernel.org, Alaa Hleihel <alaa@nvidia.com>,
	Leon Romanovsky <leonro@nvidia.com>,
	Jason Gunthorpe <jgg@nvidia.com>, Sasha Levin <sashal@kernel.org>
Subject: [PATCH 5.15 04/42] RDMA/mlx5: Fix releasing unallocated memory in dereg MR flow
Date: Wed, 15 Dec 2021 18:20:45 +0100	[thread overview]
Message-ID: <20211215172026.789963312@linuxfoundation.org> (raw)
In-Reply-To: <20211215172026.641863587@linuxfoundation.org>

From: Alaa Hleihel <alaa@nvidia.com>

[ Upstream commit f0ae4afe3d35e67db042c58a52909e06262b740f ]

For the case of IB_MR_TYPE_DM the mr does doesn't have a umem, even though
it is a user MR. This causes function mlx5_free_priv_descs() to think that
it is a kernel MR, leading to wrongly accessing mr->descs that will get
wrong values in the union which leads to attempt to release resources that
were not allocated in the first place.

For example:
 DMA-API: mlx5_core 0000:08:00.1: device driver tries to free DMA memory it has not allocated [device address=0x0000000000000000] [size=0 bytes]
 WARNING: CPU: 8 PID: 1021 at kernel/dma/debug.c:961 check_unmap+0x54f/0x8b0
 RIP: 0010:check_unmap+0x54f/0x8b0
 Call Trace:
  debug_dma_unmap_page+0x57/0x60
  mlx5_free_priv_descs+0x57/0x70 [mlx5_ib]
  mlx5_ib_dereg_mr+0x1fb/0x3d0 [mlx5_ib]
  ib_dereg_mr_user+0x60/0x140 [ib_core]
  uverbs_destroy_uobject+0x59/0x210 [ib_uverbs]
  uobj_destroy+0x3f/0x80 [ib_uverbs]
  ib_uverbs_cmd_verbs+0x435/0xd10 [ib_uverbs]
  ? uverbs_finalize_object+0x50/0x50 [ib_uverbs]
  ? lock_acquire+0xc4/0x2e0
  ? lock_acquired+0x12/0x380
  ? lock_acquire+0xc4/0x2e0
  ? lock_acquire+0xc4/0x2e0
  ? ib_uverbs_ioctl+0x7c/0x140 [ib_uverbs]
  ? lock_release+0x28a/0x400
  ib_uverbs_ioctl+0xc0/0x140 [ib_uverbs]
  ? ib_uverbs_ioctl+0x7c/0x140 [ib_uverbs]
  __x64_sys_ioctl+0x7f/0xb0
  do_syscall_64+0x38/0x90

Fix it by reorganizing the dereg flow and mlx5_ib_mr structure:
 - Move the ib_umem field into the user MRs structure in the union as it's
   applicable only there.
 - Function mlx5_ib_dereg_mr() will now call mlx5_free_priv_descs() only
   in case there isn't udata, which indicates that this isn't a user MR.

Fixes: f18ec4223117 ("RDMA/mlx5: Use a union inside mlx5_ib_mr")
Link: https://lore.kernel.org/r/66bb1dd253c1fd7ceaa9fc411061eefa457b86fb.1637581144.git.leonro@nvidia.com
Signed-off-by: Alaa Hleihel <alaa@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/infiniband/hw/mlx5/mlx5_ib.h |  6 +++---
 drivers/infiniband/hw/mlx5/mr.c      | 26 ++++++++++++--------------
 2 files changed, 15 insertions(+), 17 deletions(-)

diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index bf20a388eabe1..6204ae2caef58 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -641,7 +641,6 @@ struct mlx5_ib_mr {
 
 	/* User MR data */
 	struct mlx5_cache_ent *cache_ent;
-	struct ib_umem *umem;
 
 	/* This is zero'd when the MR is allocated */
 	union {
@@ -653,7 +652,7 @@ struct mlx5_ib_mr {
 			struct list_head list;
 		};
 
-		/* Used only by kernel MRs (umem == NULL) */
+		/* Used only by kernel MRs */
 		struct {
 			void *descs;
 			void *descs_alloc;
@@ -675,8 +674,9 @@ struct mlx5_ib_mr {
 			int data_length;
 		};
 
-		/* Used only by User MRs (umem != NULL) */
+		/* Used only by User MRs */
 		struct {
+			struct ib_umem *umem;
 			unsigned int page_shift;
 			/* Current access_flags */
 			int access_flags;
diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c
index 22e2f4d79743d..69b2ce4c292ae 100644
--- a/drivers/infiniband/hw/mlx5/mr.c
+++ b/drivers/infiniband/hw/mlx5/mr.c
@@ -1911,19 +1911,18 @@ mlx5_alloc_priv_descs(struct ib_device *device,
 	return ret;
 }
 
-static void
-mlx5_free_priv_descs(struct mlx5_ib_mr *mr)
+static void mlx5_free_priv_descs(struct mlx5_ib_mr *mr)
 {
-	if (!mr->umem && 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);
+	struct mlx5_ib_dev *dev = to_mdev(mr->ibmr.device);
+	int size = mr->max_descs * mr->desc_size;
 
-		dma_unmap_single(&dev->mdev->pdev->dev, mr->desc_map, size,
-				 DMA_TO_DEVICE);
-		kfree(mr->descs_alloc);
-		mr->descs = NULL;
-	}
+	if (!mr->descs)
+		return;
+
+	dma_unmap_single(&dev->mdev->pdev->dev, mr->desc_map, size,
+			 DMA_TO_DEVICE);
+	kfree(mr->descs_alloc);
+	mr->descs = NULL;
 }
 
 int mlx5_ib_dereg_mr(struct ib_mr *ibmr, struct ib_udata *udata)
@@ -1999,7 +1998,8 @@ int mlx5_ib_dereg_mr(struct ib_mr *ibmr, struct ib_udata *udata)
 	if (mr->cache_ent) {
 		mlx5_mr_cache_free(dev, mr);
 	} else {
-		mlx5_free_priv_descs(mr);
+		if (!udata)
+			mlx5_free_priv_descs(mr);
 		kfree(mr);
 	}
 	return 0;
@@ -2086,7 +2086,6 @@ static struct mlx5_ib_mr *mlx5_ib_alloc_pi_mr(struct ib_pd *pd,
 	if (err)
 		goto err_free_in;
 
-	mr->umem = NULL;
 	kfree(in);
 
 	return mr;
@@ -2213,7 +2212,6 @@ static struct ib_mr *__mlx5_ib_alloc_mr(struct ib_pd *pd,
 	}
 
 	mr->ibmr.device = pd->device;
-	mr->umem = NULL;
 
 	switch (mr_type) {
 	case IB_MR_TYPE_MEM_REG:
-- 
2.33.0




  parent reply	other threads:[~2021-12-15 17:23 UTC|newest]

Thread overview: 52+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-12-15 17:20 [PATCH 5.15 00/42] 5.15.9-rc1 review Greg Kroah-Hartman
2021-12-15 17:20 ` [PATCH 5.15 01/42] nfc: fix segfault in nfc_genl_dump_devices_done Greg Kroah-Hartman
2021-12-15 17:20 ` [PATCH 5.15 02/42] hwmon: (corsair-psu) fix plain integer used as NULL pointer Greg Kroah-Hartman
2021-12-15 17:20 ` [PATCH 5.15 03/42] RDMA: Fix use-after-free in rxe_queue_cleanup Greg Kroah-Hartman
2021-12-15 17:20 ` Greg Kroah-Hartman [this message]
2022-01-01 10:56   ` [PATCH 5.15 04/42] RDMA/mlx5: Fix releasing unallocated memory in dereg MR flow Thorsten Leemhuis
2022-01-07  5:57     ` Thorsten Leemhuis
2022-01-07 10:57       ` Greg Kroah-Hartman
2021-12-15 17:20 ` [PATCH 5.15 05/42] mtd: rawnand: Fix nand_erase_op delay Greg Kroah-Hartman
2021-12-15 17:20 ` [PATCH 5.15 06/42] mtd: rawnand: Fix nand_choose_best_timings() on unsupported interface Greg Kroah-Hartman
2021-12-15 17:20 ` [PATCH 5.15 07/42] inet: use #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING consistently Greg Kroah-Hartman
2021-12-15 17:20 ` [PATCH 5.15 08/42] dt-bindings: media: nxp,imx7-mipi-csi2: Drop bad if/then schema Greg Kroah-Hartman
2021-12-15 17:20 ` [PATCH 5.15 09/42] clk: qcom: sm6125-gcc: Swap ops of ice and apps on sdcc1 Greg Kroah-Hartman
2021-12-15 17:20 ` [PATCH 5.15 10/42] perf bpf_skel: Do not use typedef to avoid error on old clang Greg Kroah-Hartman
2021-12-15 17:20 ` [PATCH 5.15 11/42] netfs: Fix lockdep warning from taking sb_writers whilst holding mmap_lock Greg Kroah-Hartman
2021-12-15 17:20 ` [PATCH 5.15 12/42] RDMA/irdma: Fix a user-after-free in add_pble_prm Greg Kroah-Hartman
2021-12-15 17:20 ` [PATCH 5.15 13/42] RDMA/irdma: Fix a potential memory allocation issue in irdma_prm_add_pble_mem() Greg Kroah-Hartman
2021-12-15 17:20 ` [PATCH 5.15 14/42] RDMA/irdma: Report correct WC errors Greg Kroah-Hartman
2021-12-15 17:20 ` [PATCH 5.15 15/42] RDMA/irdma: Dont arm the CQ more than two times if no CE for this CQ Greg Kroah-Hartman
2021-12-15 17:20 ` [PATCH 5.15 16/42] ice: fix FDIR init missing when reset VF Greg Kroah-Hartman
2021-12-15 17:20 ` [PATCH 5.15 17/42] vmxnet3: fix minimum vectors alloc issue Greg Kroah-Hartman
2021-12-15 17:20 ` [PATCH 5.15 18/42] i2c: virtio: fix completion handling Greg Kroah-Hartman
2021-12-15 17:21 ` [PATCH 5.15 19/42] drm/msm: Fix null ptr access msm_ioctl_gem_submit() Greg Kroah-Hartman
2021-12-15 17:21 ` [PATCH 5.15 20/42] drm/msm/a6xx: Fix uinitialized use of gpu_scid Greg Kroah-Hartman
2021-12-15 17:21 ` [PATCH 5.15 21/42] drm/msm/dsi: set default num_data_lanes Greg Kroah-Hartman
2021-12-15 17:21 ` [PATCH 5.15 22/42] drm/msm/dp: Avoid unpowered AUX xfers that caused crashes Greg Kroah-Hartman
2021-12-15 17:21 ` [PATCH 5.15 23/42] KVM: arm64: Save PSTATE early on exit Greg Kroah-Hartman
2021-12-15 17:21 ` [PATCH 5.15 24/42] s390/test_unwind: use raw opcode instead of invalid instruction Greg Kroah-Hartman
2021-12-15 17:21 ` [PATCH 5.15 25/42] Revert "tty: serial: fsl_lpuart: drop earlycon entry for i.MX8QXP" Greg Kroah-Hartman
2021-12-15 17:21 ` [PATCH 5.15 26/42] net/mlx4_en: Update reported link modes for 1/10G Greg Kroah-Hartman
2021-12-15 17:21 ` [PATCH 5.15 27/42] loop: Use pr_warn_once() for loop_control_remove() warning Greg Kroah-Hartman
2021-12-15 17:21 ` [PATCH 5.15 28/42] ALSA: hda: Add Intel DG2 PCI ID and HDMI codec vid Greg Kroah-Hartman
2021-12-15 17:21 ` [PATCH 5.15 29/42] ALSA: hda/hdmi: fix HDA codec entry table order for ADL-P Greg Kroah-Hartman
2021-12-15 17:21 ` [PATCH 5.15 30/42] parisc/agp: Annotate parisc agp init functions with __init Greg Kroah-Hartman
2021-12-15 17:21 ` [PATCH 5.15 31/42] i2c: rk3x: Handle a spurious start completion interrupt flag Greg Kroah-Hartman
2021-12-15 17:21 ` [PATCH 5.15 32/42] net: netlink: af_netlink: Prevent empty skb by adding a check on len Greg Kroah-Hartman
2021-12-15 17:21 ` [PATCH 5.15 33/42] drm/amdgpu: cancel the correct hrtimer on exit Greg Kroah-Hartman
2021-12-15 17:21 ` [PATCH 5.15 34/42] drm/amdgpu: check atomic flag to differeniate with legacy path Greg Kroah-Hartman
2021-12-15 17:21 ` [PATCH 5.15 35/42] drm/amd/display: Fix for the no Audio bug with Tiled Displays Greg Kroah-Hartman
2021-12-15 17:21 ` [PATCH 5.15 36/42] drm/amdkfd: fix double free mem structure Greg Kroah-Hartman
2021-12-15 17:21 ` [PATCH 5.15 37/42] drm/amd/display: add connector type check for CRC source set Greg Kroah-Hartman
2021-12-15 17:21 ` [PATCH 5.15 38/42] drm/amdkfd: process_info lock not needed for svm Greg Kroah-Hartman
2021-12-15 17:21 ` [PATCH 5.15 39/42] tracing: Fix a kmemleak false positive in tracing_map Greg Kroah-Hartman
2021-12-15 17:21 ` [PATCH 5.15 40/42] staging: most: dim2: use device release method Greg Kroah-Hartman
2021-12-15 17:21 ` [PATCH 5.15 41/42] fuse: make sure reclaim doesnt write the inode Greg Kroah-Hartman
2021-12-15 17:21 ` [PATCH 5.15 42/42] perf inject: Fix itrace space allowed for new attributes Greg Kroah-Hartman
2021-12-15 20:01 ` [PATCH 5.15 00/42] 5.15.9-rc1 review Jon Hunter
2021-12-15 21:51 ` Shuah Khan
2021-12-15 23:03 ` Fox Chen
2021-12-15 23:46 ` Florian Fainelli
2021-12-16  3:08 ` Naresh Kamboju
2021-12-16 18:08 ` Guenter Roeck

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20211215172026.789963312@linuxfoundation.org \
    --to=gregkh@linuxfoundation.org \
    --cc=alaa@nvidia.com \
    --cc=jgg@nvidia.com \
    --cc=leonro@nvidia.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=sashal@kernel.org \
    --cc=stable@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).