2021-03-20 20:12:49

by Tong Zhang

Subject: [PATCH] drm/radeon: don't evict if not initialized

TTM_PL_VRAM may not initialized at all when calling
radeon_bo_evict_vram(). We need to check before doing eviction.

[ 2.160837] BUG: kernel NULL pointer dereference, address: 0000000000000020
[ 2.161212] #PF: supervisor read access in kernel mode
[ 2.161490] #PF: error_code(0x0000) - not-present page
[ 2.161767] PGD 0 P4D 0
[ 2.163088] RIP: 0010:ttm_resource_manager_evict_all+0x70/0x1c0 [ttm]
[ 2.168506] Call Trace:
[ 2.168641] radeon_bo_evict_vram+0x1c/0x20 [radeon]
[ 2.168936] radeon_device_fini+0x28/0xf9 [radeon]
[ 2.169224] radeon_driver_unload_kms+0x44/0xa0 [radeon]
[ 2.169534] radeon_driver_load_kms+0x174/0x210 [radeon]
[ 2.169843] drm_dev_register+0xd9/0x1c0 [drm]
[ 2.170104] radeon_pci_probe+0x117/0x1a0 [radeon]

Signed-off-by: Tong Zhang <[email protected]>
---
drivers/gpu/drm/radeon/radeon_object.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c
index 9b81786782de..04e9a8118b0e 100644
--- a/drivers/gpu/drm/radeon/radeon_object.c
+++ b/drivers/gpu/drm/radeon/radeon_object.c
@@ -384,7 +384,9 @@ int radeon_bo_evict_vram(struct radeon_device *rdev)
}
#endif
man = ttm_manager_type(bdev, TTM_PL_VRAM);
- return ttm_resource_manager_evict_all(bdev, man);
+ if (man)
+ return ttm_resource_manager_evict_all(bdev, man);
+ return 0;
}

void radeon_bo_force_delete(struct radeon_device *rdev)
--
2.25.1

2021-03-21 13:30:12

by Christian König

Subject: Re: [PATCH] drm/radeon: don't evict if not initialized

Am 20.03.21 um 21:10 schrieb Tong Zhang:
> TTM_PL_VRAM may not initialized at all when calling
> radeon_bo_evict_vram(). We need to check before doing eviction.
>
> [ 2.160837] BUG: kernel NULL pointer dereference, address: 0000000000000020
> [ 2.161212] #PF: supervisor read access in kernel mode
> [ 2.161490] #PF: error_code(0x0000) - not-present page
> [ 2.161767] PGD 0 P4D 0
> [ 2.163088] RIP: 0010:ttm_resource_manager_evict_all+0x70/0x1c0 [ttm]
> [ 2.168506] Call Trace:
> [ 2.168641] radeon_bo_evict_vram+0x1c/0x20 [radeon]
> [ 2.168936] radeon_device_fini+0x28/0xf9 [radeon]
> [ 2.169224] radeon_driver_unload_kms+0x44/0xa0 [radeon]
> [ 2.169534] radeon_driver_load_kms+0x174/0x210 [radeon]
> [ 2.169843] drm_dev_register+0xd9/0x1c0 [drm]
> [ 2.170104] radeon_pci_probe+0x117/0x1a0 [radeon]
>
> Signed-off-by: Tong Zhang <[email protected]>
> ---
> drivers/gpu/drm/radeon/radeon_object.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c
> index 9b81786782de..04e9a8118b0e 100644
> --- a/drivers/gpu/drm/radeon/radeon_object.c
> +++ b/drivers/gpu/drm/radeon/radeon_object.c
> @@ -384,7 +384,9 @@ int radeon_bo_evict_vram(struct radeon_device *rdev)
> }
> #endif
> man = ttm_manager_type(bdev, TTM_PL_VRAM);
> - return ttm_resource_manager_evict_all(bdev, man);
> + if (man)
> + return ttm_resource_manager_evict_all(bdev, man);
> + return 0;

You should probably code this the other way around, e.g.

If (!man)
    return 0;
...

Apart from that looks good to me.

Christian.

> }
>
> void radeon_bo_force_delete(struct radeon_device *rdev)

2021-03-21 16:12:26

by Tong Zhang

Subject: Re: [PATCH] drm/radeon: don't evict if not initialized

Thanks,
Fixed as suggested and sent as v2.
- Tong

On Sun, Mar 21, 2021 at 9:26 AM Christian König
<[email protected]> wrote:
>
>
>
> Am 20.03.21 um 21:10 schrieb Tong Zhang:
> > TTM_PL_VRAM may not initialized at all when calling
> > radeon_bo_evict_vram(). We need to check before doing eviction.
> >
> > [ 2.160837] BUG: kernel NULL pointer dereference, address: 0000000000000020
> > [ 2.161212] #PF: supervisor read access in kernel mode
> > [ 2.161490] #PF: error_code(0x0000) - not-present page
> > [ 2.161767] PGD 0 P4D 0
> > [ 2.163088] RIP: 0010:ttm_resource_manager_evict_all+0x70/0x1c0 [ttm]
> > [ 2.168506] Call Trace:
> > [ 2.168641] radeon_bo_evict_vram+0x1c/0x20 [radeon]
> > [ 2.168936] radeon_device_fini+0x28/0xf9 [radeon]
> > [ 2.169224] radeon_driver_unload_kms+0x44/0xa0 [radeon]
> > [ 2.169534] radeon_driver_load_kms+0x174/0x210 [radeon]
> > [ 2.169843] drm_dev_register+0xd9/0x1c0 [drm]
> > [ 2.170104] radeon_pci_probe+0x117/0x1a0 [radeon]
> >
> > Signed-off-by: Tong Zhang <[email protected]>
> > ---
> > drivers/gpu/drm/radeon/radeon_object.c | 4 +++-
> > 1 file changed, 3 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c
> > index 9b81786782de..04e9a8118b0e 100644
> > --- a/drivers/gpu/drm/radeon/radeon_object.c
> > +++ b/drivers/gpu/drm/radeon/radeon_object.c
> > @@ -384,7 +384,9 @@ int radeon_bo_evict_vram(struct radeon_device *rdev)
> > }
> > #endif
> > man = ttm_manager_type(bdev, TTM_PL_VRAM);
> > - return ttm_resource_manager_evict_all(bdev, man);
> > + if (man)
> > + return ttm_resource_manager_evict_all(bdev, man);
> > + return 0;
>
> You should probably code this the other way around, e.g.
>
> If (!man)
> return 0;
> ...
>
> Apart from that looks good to me.
>
> Christian.
>
> > }
> >
> > void radeon_bo_force_delete(struct radeon_device *rdev)
>