[BUG, regression] Dereferencing of NULL pointer in radeon_mn_unregister()
Jason Gunthorpe
jgg at mellanox.com
Mon Sep 2 06:19:00 UTC 2019
On Sun, Sep 01, 2019 at 05:48:36PM +0200, Petr Cvek wrote:
> > Is just adding a
> >
> > if (!rmn)
> > retrun
> >
> > To the top of radeon_mn_unregister enough to fix it?
>
> Yeah it seems to work. A further test with minetest works too.
Okay, I added this patch to the hmm tree, thanks.
>From 829394d77e3026e08e7879fb37f14c90de7b0fd8 Mon Sep 17 00:00:00 2001
From: Jason Gunthorpe <jgg at mellanox.com>
Date: Mon, 2 Sep 2019 03:01:03 -0300
Subject: [PATCH] drm/radeon: guard against calling an unpaired
radeon_mn_unregister()
This check was accidently deleted in the below commit. There are cases
where the driver will call unregister even though it hasn't registered
anything.
CPU 0 Unable to handle kernel paging request at virtual address 0000001c, epc == 808de6d4, ra == 804d32ec
Call Trace:
[<808de6d4>] mutex_lock+0x8/0x44
[<804d32ec>] radeon_mn_unregister+0x3c/0xb0
[<8041583c>] radeon_gem_object_free+0x18/0x2c
[<803a451c>] drm_gem_object_release_handle+0x74/0xac
[<803a45d0>] drm_gem_handle_delete+0x7c/0x128
[<803a5bf4>] drm_ioctl_kernel+0xb0/0x108
[<803a5e74>] drm_ioctl+0x200/0x3a8
[<803e07b4>] radeon_drm_ioctl+0x54/0xc0
[<801214dc>] do_vfs_ioctl+0x4e8/0x81c
[<80121864>] ksys_ioctl+0x54/0xb0
[<8001100c>] syscall_common+0x34/0x58
Link: https://lore.kernel.org/r/2fc7ef14-e89a-1f2d-381d-1c9b05da02d3@gmail.com
Fixes: 534e5f84b7a9 ("drm/radeon: use mmu_notifier_get/put for struct radeon_mn")
Reported-by: Petr Cvek <petrcvekcz at gmail.com>
Signed-off-by: Jason Gunthorpe <jgg at mellanox.com>
---
drivers/gpu/drm/radeon/radeon_mn.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/gpu/drm/radeon/radeon_mn.c b/drivers/gpu/drm/radeon/radeon_mn.c
index fc8254273a800b..1ee20d528a7c24 100644
--- a/drivers/gpu/drm/radeon/radeon_mn.c
+++ b/drivers/gpu/drm/radeon/radeon_mn.c
@@ -234,6 +234,9 @@ void radeon_mn_unregister(struct radeon_bo *bo)
struct radeon_mn *rmn = bo->mn;
struct list_head *head;
+ if (!rmn)
+ return;
+
mutex_lock(&rmn->lock);
/* save the next list entry for later */
head = bo->mn_list.next;
--
2.23.0
More information about the amd-gfx
mailing list