This adds support to discover the  buffer objects that belong to a
process being checkpointed. The data corresponding to these buffer
objects is returned to user space plugin running under criu master
context which then stores this info to recreate these buffer objects
during a restore operation.
Reviewed-by: NFelix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: NDavid Yat Sin <david.yatsin@amd.com>
Signed-off-by: NRajneesh Bhardwaj <rajneesh.bhardwaj@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

We ran into the problem that clearing really larger buffer (60GiB) caused an
SDMA timeout.

Restructure the function to use the dst window instead of mapping the whole
buffer into the GART and then fill only 2MiB/256MiB chunks at a time.

v2: rebase on restructured window map.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NFelix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Instead of limiting the size before we call the mapping
function let the function itself limit the size.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NFelix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

That should never happen, but make sure that we only warn instead of
crash.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NFelix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

We probably never trigger this, but the logic inside the check is
inverted.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NFelix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Debug VRAM access through SDMA has several broken parts resulting in
silent MMIO fallback.

BO kernel creation takes the location of the cpu addr pointer, not
the pointer itself for address kmap.

drm_dev_enter return true on success so change access check.

The source BO is reserved but not pinned so find the address using the
cursor offset relative to its memory domain start.
Signed-off-by: NJonathan Kim <jonathan.kim@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

That's just a leftover from old radeon days and was preventing CS and GART
bindings before the hardware was initialized. But nowdays that is
perfectly valid.

The only thing we need to warn about are GART binding before the table
is even allocated.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NGuchun Chen <guchun.chen@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Some suggested cleanups to declutter ttm when doing debug VRAM access over
SDMA.
Signed-off-by: NJonathan Kim <jonathan.kim@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

For better performance during VRAM access for debugged processes, do
read/write copies over SDMA.

In order to fulfill post mortem debugging on a broken device, fallback to
stable MMIO access when gpu recovery is disabled or when job submission
time outs are set to max.  Failed SDMA access should automatically fall
back to MMIO access.

Use a pre-allocated GTT bounce buffer pre-mapped into GART to avoid
page-table updates and TLB flushes on access.
Signed-off-by: NJonathan Kim <jonathan.kim@amd.com>
Reviewed-by: NFelix Kuehling <felix.kuehling@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Patch: 3efb17ae7e92 ("drm/amdgpu: Call amdgpu_device_unmap_mmio() if device
is unplugged to prevent crash in GPU initialization failure") makes call to
amdgpu_device_unmap_mmio() conditioned on device unplugged. This patch unmaps
MMIO mappings even when device is not unplugged.

v2: Add condition of drm_dev_enter() to deleted unmaps in patch
"drm/amdgpu: Unmap all MMIO mappings"
Signed-off-by: NLeslie Shi <Yuliang.Shi@amd.com>
Reviewed-by: NAndrey Grodzovsky <andrey.grodzovsky@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Partially revert commit 5f319c5c.

First of all this is illegal use of RCU to call dma_fence_enable_sw_signaling()
since we don't hold a reference to the fence in question and can crash badly.

Then the code doesn't seem to have the intended effect since only the
exclusive fence is handled, but the KFD fences are always added as shared fence.

Only keep the handling to throw away the content of SVM BOs.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211122123926.385017-1-christian.koenig@amd.comReviewed-by: NFelix Kuehling <Felix.Kuehling@amd.com>

amdgpu_ttm_backend_bind() only needed for TTM_PL_TT
and AMDGPU_PL_PREEMPT.
Signed-off-by: NNirmoy Das <nirmoy.das@amd.com>
Acked-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

The userptr can be unmapped by application and still registered to
driver, restore userptr work return user pages will get -EFAULT bad
address error. Pretend this error as succeed. GPU access this userptr
will have VM fault later, it is better than application soft hangs with
stalled user mode queues.

v2: squash in warning fix (Alex)
Signed-off-by: NPhilip Yang <Philip.Yang@amd.com>
Reviewed-by: NFelix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

In order to better track where in the kernel the dma-buf code is used,
put the symbols in the namespace DMA_BUF and modify all users of the
symbols to properly import the namespace to not break the build at the
same time.

Now the output of modinfo shows the use of these symbols, making it
easier to watch for users over time:

$ modinfo drivers/misc/fastrpc.ko | grep import
import_ns:      DMA_BUF

Cc: "Pan, Xinhui" <Xinhui.Pan@amd.com>
Cc: David Airlie <airlied@linux.ie>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Maxime Ripard <mripard@kernel.org>
Cc: Thomas Zimmermann <tzimmermann@suse.de>
Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
Cc: dri-devel@lists.freedesktop.org
Acked-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: NChristian König <christian.koenig@amd.com>
Acked-by: NArnd Bergmann <arnd@arndb.de>
Acked-by: NSumit Semwal <sumit.semwal@linaro.org>
Acked-by: NAlex Deucher <alexander.deucher@amd.com>
Link: https://lore.kernel.org/r/20211010124628.17691-1-gregkh@linuxfoundation.orgSigned-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Simplifying the code a bit.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20211005113742.1101-13-christian.koenig@amd.com

When creating unregistered new svm range to recover retry fault, avoid
new svm range to overlap with ranges or userptr ranges managed by TTM,
otherwise svm migration will trigger TTM or userptr eviction, to evict
user queues unexpectedly.

Change helper amdgpu_ttm_tt_affect_userptr to return userptr which is
inside the range. Add helper svm_range_check_vm_userptr to scan all
userptr of the vm, and return overlap userptr bo start, last.
Signed-off-by: NPhilip Yang <Philip.Yang@amd.com>
Reviewed-by: NFelix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Unify BO evicting functionality for possible memory
types in amdgpu_ttm.c.
Signed-off-by: NNirmoy Das <nirmoy.das@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

It covers more than just ttm_bo_type_sg usage, like with say dma-buf,
since one other user is userptr in amdgpu, and in the future we might
have some more. Hence EXTERNAL is likely a more suitable name.

v2(Christian):
  - Rename these to TTM_TT_FLAGS_*
  - Fix up all the holes in the flag values
Suggested-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NMatthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Christian König <christian.koenig@amd.com>
Acked-by: NChristian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210929132629.353541-1-matthew.auld@intel.comSigned-off-by: NChristian König <christian.koenig@amd.com>

Now that setting page->index shouldn't be needed anymore, we are just
left with setting page->mapping, and here it looks like amdgpu is the
only user, where pointing the page->mapping at the dev_mapping is used
to verify that the pages do indeed belong to the device, if userspace
later tries to touch them.

v2(Christian):
  - Drop the functions altogether and just inline modifying
    the page->mapping
Signed-off-by: NMatthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Christian König <christian.koenig@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210927114114.152310-3-matthew.auld@intel.comSigned-off-by: NChristian König <christian.koenig@amd.com>

The memory backing old_mem is already freed at that point, move the
check a bit more up.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Fixes: bfa3357e ("drm/ttm: allocate resource object instead of embedding it v2")
Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/1699Acked-by: NNirmoy Das <nirmoy.das@amd.com>
Reviewed-by: NMichel Dänzer <mdaenzer@redhat.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

The memory backing old_mem is already freed at that point, move the
check a bit more up.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Fixes: bfa3357e ("drm/ttm: allocate resource object instead of embedding it v2")
Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/1699Acked-by: NNirmoy Das <nirmoy.das@amd.com>
Reviewed-by: NMichel Dänzer <mdaenzer@redhat.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Move the functionality into ttm_tt_fini and ttm_bo_tt_destroy instead.

We don't need this any more since we removed the unbind from the destroy
code paths in the drivers.

Also add a warning to ttm_tt_fini() if we try to fini a still populated TT
object.

v2: instead of reverting the patch move the functionality to different
places.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NThomas Hellström <thomas.hellstrom@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210728130552.2074-5-christian.koenig@amd.com

Doing this in amdgpu_ttm_backend_destroy() is to late.

It turned out that this is not a good idea at all because it leaves pointers
to freed up system memory pages in the GART tables of the drivers.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NThomas Hellström <thomas.hellstrom@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210728130552.2074-2-christian.koenig@amd.com

drm/amdgpu: drop redundant null-pointer checks in amdgpu_ttm_tt_populate() and amdgpu_ttm_tt_unpopulate()

The varialbe gtt in the function amdgpu_ttm_tt_populate() and
amdgpu_ttm_tt_unpopulate() is guaranteed to be not NULL in the context.
Thus the null-pointer checks are redundant and can be dropped.
Reported-by: NTOTE Robot <oslab@tsinghua.edu.cn>
Reviewed-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NTuo Li <islituo@gmail.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

1. using vram aper to access vram if possible
2. avoid MM_INDEX/MM_DATA is not working when mmio protect feature is
enabled.
Signed-off-by: NKevin Wang <kevin1.wang@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

using exiting function to replace duplicate code blocks in
amdgpu_ttm_vram_write().
Signed-off-by: NKevin Wang <kevin1.wang@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

The parameter is used in the dev_private_owner to decide if device
pages in the range require to be migrated back to system memory, based
if they are or not in the same memory domain.
In this case, this reference could come from the same memory domain
with devices connected to the same hive.
Signed-off-by: NAlex Sierra <alex.sierra@amd.com>
Reviewed-by: NFelix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

The ttm caching flags (ttm_cached, ttm_write_combined etc) are
used to determine a buffer object's mapping attributes in both
CPU page table and GPU page table (when that buffer is also
accessed by GPU). Currently the ttm caching flags are set in
function amdgpu_ttm_io_mem_reserve which is called during
DRM_AMDGPU_GEM_MMAP ioctl. This has a problem since the GPU
mapping of the buffer object (ioctl DRM_AMDGPU_GEM_VA) can
happen earlier than the mmap time, thus the GPU page table
update code can't pick up the right ttm caching flags to
decide the right GPU page table attributes.

This patch moves the ttm caching flags setting to function
amdgpu_vram_mgr_new - this function is called during the
first step of a buffer object create (eg, DRM_AMDGPU_GEM_CREATE)
so the later both CPU and GPU mapping function calls will
pick up this flag for CPU/GPU page table set up.

v2: rebase (Alex)
Signed-off-by: NOak Zeng <Oak.Zeng@amd.com>
Suggested-by: NChristian Koenig <Christian.Koenig@amd.com>
Reviewed-by: NChristian Koenig <Christian.Koenig@amd.com>
Reviewed-by: NFeifei Xu <Feifei.Xu@amd.com>
Tested-by: NPo Huang <Po.Huang@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Use vma_lookup() to find the VMA at a specific address.  As vma_lookup()
will return NULL if the address is not within any VMA, the start address
no longer needs to be validated.

Link: https://lkml.kernel.org/r/20210521174745.2219620-14-Liam.Howlett@Oracle.comSigned-off-by: NLiam R. Howlett <Liam.Howlett@Oracle.com>
Reviewed-by: NLaurent Dufour <ldufour@linux.ibm.com>
Acked-by: NDavid Hildenbrand <david@redhat.com>
Acked-by: NDavidlohr Bueso <dbueso@suse.de>
Acked-by: NAlex Deucher <alexander.deucher@amd.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Return early for non-TTM_PL_TT BOs so that we don't pass
wrong pointer to amdgpu_gtt_mgr_has_gart_addr() which assumes
ttm_resource argument to be TTM_PL_TT type BO's.

v3: remove extra braces.
v2: merge if-conditions.
Signed-off-by: NNirmoy Das <nirmoy.das@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NChristian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210629114413.3371-1-nirmoy.das@amd.com

When we run out of GTT we should still be able to evict VRAM->SYSTEM
with a bounce bufferdrm/amdgpu: always allow evicting to SYSTEM domain
Signed-off-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAndrey Grodzovsky <andrey.grodzovsky@amd.com>
Acked-by: NNirmoy Das <nirmoy.das@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210622162339.761651-3-andrey.grodzovsky@amd.com

Currently, we have a limitted GTT memory size and need a bounce buffer
when doing buffer migration between VRAM and SYSTEM domain.

The problem is under GTT memory pressure we can't do buffer migration
between VRAM and SYSTEM domain. But in some cases we really need that.
Eespecially when validating a VRAM backing store BO which resides in
SYSTEM domain.

v2: still account temporary GTT allocations
v3 (chk): revert to the simpler change for now
Signed-off-by: NLang Yu <Lang.Yu@amd.com>
Signed-off-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAndrey Grodzovsky <andrey.grodzovsky@amd.com>
Acked-by: NNirmoy Das <nirmoy.das@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210622162339.761651-2-andrey.grodzovsky@amd.com

Pagelist is no long used in amdgpu_gart_bind. Remove it.
Signed-off-by: NYifan Zhang <yifan1.zhang@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Because TTM do page counting on userptr BOs which is actually not
needed. To avoid that, lets set TTM_PAGE_FLAG_SG after tt_create and
before tt_populate.
Signed-off-by: Nxinhui pan <xinhui.pan@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

We are calling the eviction_valuable driver callback at eviction time to
determine whether we actually can evict a buffer object.
The upcoming i915 TTM backend needs the same functionality for swapout,
and that might actually be beneficial to other drivers as well.

Add an eviction_valuable call also in the swapout path. Try to keep the
current behaviour for all drivers by returning true if the buffer object
is already in the TTM_PL_SYSTEM placement. We change behaviour for the
case where a buffer object is in a TT backed placement when swapped out,
in which case the drivers normal eviction_valuable path is run.
Reviewed-by: NMaarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Christian König <christian.koenig@amd.com>
Signed-off-by: NThomas Hellström <thomas.hellstrom@linux.intel.com>
Acked-by: NChristian König <christian.koenig@amd.com>
Link: https://lore.kernel.org/r/20210602083818.241793-8-thomas.hellstrom@linux.intel.com
Link: https://patchwork.freedesktop.org/patch/msgid/20210602083818.241793-8-thomas.hellstrom@linux.intel.com

When the comment needs to state explicitly that this is doesn't get a reference
to the object then the function is named rather badly.

Rename the function and use it in even more places.

v2: use dma_resv_shared_list as new name
Signed-off-by: NChristian König <christian.koenig@amd.com>
Acked-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20210602111714.212426-5-christian.koenig@amd.com

Some ASICs such as Yellow Carp needs to reserve a region of video memory
to avoid access from driver. So this patch is to introduce a stolen
reserved buffer to protect specific buffer region.

v2: free this buffer in amdgpu_ttm_fini.
Signed-off-by: NHuang Rui <ray.huang@amd.com>
Acked-and-Tested-by: NAaron Liu <aaron.liu@amd.com>
Acked-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Start with the range manager to make the resource object the base
class for the allocated nodes.

While at it cleanup a lot of the code around that.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NMatthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210602100914.46246-2-christian.koenig@amd.com

To improve the handling we want the establish the resource object as base
class for the backend allocations.

v2: add missing error handling
Signed-off-by: NChristian König <christian.koenig@amd.com>
Acked-by: NThomas Hellström <thomas.hellstrom@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210602100914.46246-1-christian.koenig@amd.com

When we want to decouble resource management from buffer management we need to
be able to handle resources separately.

Add a resource pointer and rename bo->mem so that all code needs to
change to access the pointer instead.

No functional change.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NMatthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210430092508.60710-4-christian.koenig@amd.com