linux/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c, branch v5.10

drm/amdgpu: Fix size calculation when init onchip memory

2020-11-25T05:04:38Z

Size is page count here. Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/1372 Reviewed-by: Christian König Signed-off-by: xinhui pan Signed-off-by: Alex Deucher (cherry picked from commit d836917da7e5ca9b33ef4d499972f1feeb519e00) [airlied: from drm-next] Signed-off-by: Dave Airlie

Merge tag 'drm-misc-next-2020-09-21' of git://anongit.freedesktop.org/drm/drm-misc into drm-next

2020-09-22T23:52:24Z

drm-misc-next for 5.10: UAPI Changes: Cross-subsystem Changes: - virtio: Merged a PR for patches that will affect drm/virtio Core Changes: - dev: More devm_drm convertions and removal of drm_dev_init - atomic: Split out drm_atomic_helper_calc_timestamping_constants of drm_atomic_helper_update_legacy_modeset_state - ttm: More rework Driver Changes: - i915: selftests improvements - panfrost: support for Amlogic SoC - vc4: one fix - tree-wide: conversions to devm_drm_dev_alloc, - ast: simplifications of the atomic modesetting code - panfrost: multiple fixes - vc4: multiple fixes Signed-off-by: Dave Airlie From: Maxime Ripard Link: https://patchwork.freedesktop.org/patch/msgid/20200921152956.2gxnsdgxmwhvjyut@gilmour.lan

drm/amdgpu: prevent double kfree ttm->sg

2020-09-17T21:56:38Z

Set ttm->sg to NULL after kfree, to avoid memory corruption backtrace: [ 420.932812] kernel BUG at /build/linux-do9eLF/linux-4.15.0/mm/slub.c:295! [ 420.934182] invalid opcode: 0000 [#1] SMP NOPTI [ 420.935445] Modules linked in: xt_conntrack ipt_MASQUERADE [ 420.951332] Hardware name: Dell Inc. PowerEdge R7525/0PYVT1, BIOS 1.5.4 07/09/2020 [ 420.952887] RIP: 0010:__slab_free+0x180/0x2d0 [ 420.954419] RSP: 0018:ffffbe426291fa60 EFLAGS: 00010246 [ 420.955963] RAX: ffff9e29263e9c30 RBX: ffff9e29263e9c30 RCX: 000000018100004b [ 420.957512] RDX: ffff9e29263e9c30 RSI: fffff3d33e98fa40 RDI: ffff9e297e407a80 [ 420.959055] RBP: ffffbe426291fb00 R08: 0000000000000001 R09: ffffffffc0d39ade [ 420.960587] R10: ffffbe426291fb20 R11: ffff9e49ffdd4000 R12: ffff9e297e407a80 [ 420.962105] R13: fffff3d33e98fa40 R14: ffff9e29263e9c30 R15: ffff9e2954464fd8 [ 420.963611] FS: 00007fa2ea097780(0000) GS:ffff9e297e840000(0000) knlGS:0000000000000000 [ 420.965144] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 420.966663] CR2: 00007f16bfffefb8 CR3: 0000001ff0c62000 CR4: 0000000000340ee0 [ 420.968193] Call Trace: [ 420.969703] ? __page_cache_release+0x3c/0x220 [ 420.971294] ? amdgpu_ttm_tt_unpopulate+0x5e/0x80 [amdgpu] [ 420.972789] kfree+0x168/0x180 [ 420.974353] ? amdgpu_ttm_tt_set_user_pages+0x64/0xc0 [amdgpu] [ 420.975850] ? kfree+0x168/0x180 [ 420.977403] amdgpu_ttm_tt_unpopulate+0x5e/0x80 [amdgpu] [ 420.978888] ttm_tt_unpopulate.part.10+0x53/0x60 [amdttm] [ 420.980357] ttm_tt_destroy.part.11+0x4f/0x60 [amdttm] [ 420.981814] ttm_tt_destroy+0x13/0x20 [amdttm] [ 420.983273] ttm_bo_cleanup_memtype_use+0x36/0x80 [amdttm] [ 420.984725] ttm_bo_release+0x1c9/0x360 [amdttm] [ 420.986167] amdttm_bo_put+0x24/0x30 [amdttm] [ 420.987663] amdgpu_bo_unref+0x1e/0x30 [amdgpu] [ 420.989165] amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu+0x9ca/0xb10 [amdgpu] [ 420.990666] kfd_ioctl_alloc_memory_of_gpu+0xef/0x2c0 [amdgpu] Signed-off-by: Philip Yang Reviewed-by: Felix Kuehling Reviewed-by: Christian König Signed-off-by: Alex Deucher

drm/ttm: drop special pipeline accel cleanup function.

2020-09-17T20:23:06Z

The two accel cleanup paths were mostly the same once refactored. Just pass a bool to say if the evictions are to be pipelined. Signed-off-by: Dave Airlie Reviewed-by: Christian König Link: https://patchwork.freedesktop.org/patch/msgid/20200917064132.148521-2-airlied@gmail.com

drm/ttm/drivers: call the bind function directly.

2020-09-17T20:16:03Z

Now the bind functions have all the protection explicitly the drivers can just call them directly, and the api can be unexported Reviewed-by: Christian König Signed-off-by: Dave Airlie Link: https://patchwork.freedesktop.org/patch/msgid/20200917043040.146575-5-airlied@gmail.com

drm/ttm: move unbind into the tt destroy.

2020-09-17T20:15:24Z

This moves unbind into the driver side on destroy paths. Reviewed-by: Christian König Signed-off-by: Dave Airlie Link: https://patchwork.freedesktop.org/patch/msgid/20200917043040.146575-4-airlied@gmail.com

drm/ttm: flip tt destroy ordering.

2020-09-17T20:14:41Z

Call the driver first and have it call the common code cleanup. This is useful later to fix unbind. Reviewed-by: Christian König Signed-off-by: Dave Airlie Link: https://patchwork.freedesktop.org/patch/msgid/20200917043040.146575-3-airlied@gmail.com

drm/ttm: protect against reentrant bind in the drivers

2020-09-17T20:14:00Z

This moves the generic tracking into the drivers and protects against reentrancy in the drivers. It fixes up radeon and agp to be able to query the bound status as that is required. Reviewed-by: Christian König Signed-off-by: Dave Airlie Link: https://patchwork.freedesktop.org/patch/msgid/20200917043040.146575-2-airlied@gmail.com

drm/ttm: move ttm binding/unbinding out of ttm_tt paths.

2020-09-15T23:35:30Z

Move these up to the bo level, moving ttm_tt to just being backing store. Next step is to move the bound flag out. Reviewed-by: Christian König Signed-off-by: Dave Airlie Link: https://patchwork.freedesktop.org/patch/msgid/20200915024007.67163-6-airlied@gmail.com

drm/ttm: split populate out from binding.

2020-09-15T23:34:54Z

Drivers have to call populate themselves now before binding. Reviewed-by: Christian König Signed-off-by: Dave Airlie Link: https://patchwork.freedesktop.org/patch/msgid/20200915024007.67163-5-airlied@gmail.com