radeon_ttm_bo_destroy() is attempting to access the resource object to
update memory counters. However, the resource object is already freed when
ttm calls this function via the destroy callback. This causes an oops when
a bo is freed:
BUG: kernel NULL pointer dereference, address: 0000000000000010
RIP: 0010:radeon_ttm_bo_destroy+0x2c/0x100 [radeon]
Call Trace:
radeon_bo_unref+0x1a/0x30 [radeon]
radeon_gem_object_free+0x33/0x50 [radeon]
drm_gem_object_release_handle+0x69/0x70 [drm]
drm_gem_handle_delete+0x62/0xa0 [drm]
? drm_mode_destroy_dumb+0x40/0x40 [drm]
drm_ioctl_kernel+0xb2/0xf0 [drm]
drm_ioctl+0x30a/0x3c0 [drm]
? drm_mode_destroy_dumb+0x40/0x40 [drm]
radeon_drm_ioctl+0x49/0x80 [radeon]
__x64_sys_ioctl+0x8e/0xd0
Avoid the issue by updating the counters in the delete_mem_notify callback
instead. Also, fix memory statistic updating in radeon_bo_move() to
identify the source type correctly. The source type needs to be saved
before the move, because the moved from object may be altered by the move.
Fixes: bfa3357ef9 ("drm/ttm: allocate resource object instead of embedding it v2")
Signed-off-by: Mikel Rychliski <mikel@mikelr.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210624045121.15643-1-mikel@mikelr.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: Christian König <christian.koenig@amd.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210430092508.60710-4-christian.koenig@amd.com
Based on an idea from Dave, but cleaned up a bit.
We had multiple fields for essentially the same thing.
Now bo->base.size is the original size of the BO in
arbitrary units, usually bytes.
bo->mem.num_pages is the size in number of pages in the
resource domain of bo->mem.mem_type.
v2: use the GEM object size instead of the BO size
v3: fix printks in some places
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com> (v1)
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/406831/
We already implemented the fault handler ourself,
just open code what is necessary here.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/392322/
This name better reflects what the object does. I didn't rename
all the pointers it seemed too messy.
Signed-off-by: Dave Airlie <airlied@redhat.com>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200804025632.3868079-60-airlied@gmail.com
Calculate GPU offset in radeon_bo_gpu_offset without depending on
bo->offset.
Signed-off-by: Nirmoy Das <nirmoy.das@amd.com>
Reviewed-and-tested-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/372935/
Be more consistent with the naming of the other DMA-buf objects.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/323401/
Ensure that the driver can listen to evictions even when they don't take the
path through ttm_bo_driver::move.
This is crucial for amdgpu, which relies on an eviction counter to skip
re-binding page tables when possible.
Signed-off-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Not used any more.
Reviewed-by: Sinclair Yeh <syeh@vmware.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Remove the function radeon_bo_fbdev_mmap() that is not used anywhere.
This was partially found by using a static code analysis program called cppcheck.
Signed-off-by: Rickard Strandqvist <rickard_strandqvist@spectrumdigital.se>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This allows us to add the real execution fence as shared.
v2: fix typo
Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Adds an extra argument to radeon_bo_create, which is only used in radeon_prime.c.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
To be consistent with radeon_bo_unref, needed in the following patch.
Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
v2: fix rebase onto drm-fixes
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Userspace should set the first 4 bits of drm_radeon_cs_reloc::flags to
a number from 0 to 15. The higher the number, the higher the priority,
which means a buffer with a higher number will be validated sooner.
The old behavior is preserved: Buffers used for write are prioritized over
read-only buffers if the userspace doesn't set the number.
v2: add buffers to buckets directly, then concatenate them
v3: use a stable sort
Signed-off-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
The statistics are:
- VRAM usage in bytes
- GTT usage in bytes
- number of bytes moved by TTM
The last one is actually a counter, so you need to sample it before and after
command submission and take the difference.
This is useful for finding performance bottlenecks. Userspace queries are
also added.
v2: use atomic64_t
Signed-off-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Use the new vma-manager infrastructure. This doesn't change any
implementation details as the vma-offset-manager is nearly copied 1-to-1
from TTM.
The vm_lock is moved into the offset manager so we can drop it from TTM.
During lookup, we use the vma locking helpers to take a reference to the
found object.
In all other scenarios, locking stays the same as before. We always
guarantee that drm_vma_offset_remove() is called only during destruction.
Hence, helpers like drm_vma_node_offset_addr() are always safe as long as
the node has a valid offset.
This also drops the addr_space_offset member as it is a copy of vm_start
in vma_node objects. Use the accessor functions instead.
v4:
- remove vm_lock
- use drm_vma_offset_lock_lookup() to protect lookup (instead of vm_lock)
Cc: Dave Airlie <airlied@redhat.com>
Cc: Ben Skeggs <bskeggs@redhat.com>
Cc: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Cc: Martin Peres <martin.peres@labri.fr>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@gmail.com>
There are cases where we need more than 4k alignment. No
functional change with this commit.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Try to use lockdep_assert_held or other alternatives where possible.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This commit converts the source of the val_seq counter to
the ww_mutex api. The reservation objects are converted later,
because there is still a lockdep splat in nouveau that has to
resolved first.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Just everything needed to decode videos using UVD.
v6: just all the bugfixes and support for R7xx-SI merged in one patch
v7: UVD_CGC_GATE is a write only register, lockup detection fix
v8: split out VRAM fallback changes, remove support for RV770,
add support for HEMLOCK, add buffer sizes checks
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The no_wait param isn't used anywhere, and actually isn't
very usefull at all.
Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
It doesn't really belong into the object functions,
also rename it to avoid collisions with struct radeon_bo_va.
Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
This adds prime->fd and fd->prime support to radeon.
It passes the sg object to ttm and then populates
the gart entries using it.
Compile tested only.
v2: stub kmap + use new helpers + add reimporting
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Define the interface without modifying the allocation
algorithm in any way.
v2: rebase on top of fence new uint64 patch
v3: add ring to debugfs output
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Christian König <deathsimple@vodafone.de>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Instead of offset + size keep start and end offset directly.
Signed-off-by: Christian König <deathsimple@vodafone.de>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Instead of hacking the calculation multiple times.
Signed-off-by: Christian König <deathsimple@vodafone.de>
Signed-off-by: Dave Airlie <airlied@redhat.com>
The hardware only takes 27 bits for the offset, so larger offsets are
truncated, and the hardware cursor shows random bits other than the intended
ones.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=46796
Cc: stable@vger.kernel.org
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Virtual address space are per drm client (opener of /dev/drm).
Client are in charge of virtual address space, they need to
map bo into it by calling DRM_RADEON_GEM_VA ioctl.
First 16M of virtual address space is reserved by the kernel.
Once using 2 level page table we should be able to have a small
vram memory footprint for each pt (there would be one pt for all
gart, one for all vram and then one first level for each virtual
address space).
Plan include using the sub allocator for a common vm page table
area and using memcpy to copy vm page table in & out. Or use
a gart object and copy things in & out using dma.
v2: agd5f fixes:
- Add vram base offset for vram pages. The GPU physical address of a
vram page is FB_OFFSET + page offset. FB_OFFSET is 0 on discrete
cards and the physical bus address of the stolen memory on
integrated chips.
- VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR covers all vmid's >= 1
v3: agd5f:
- integrate with the semaphore/multi-ring stuff
v4:
- rebase on top ttm dma & multi-ring stuff
- userspace is now in charge of the address space
- no more specific cs vm ioctl, instead cs ioctl has a new
chunk
v5:
- properly handle mem == NULL case from move_notify callback
- fix the vm cleanup path
v6:
- fix update of page table to only happen on valid mem placement
v7:
- add tlb flush for each vm context
- add flags to define mapping property (readable, writeable, snooped)
- make ring id implicit from ib->fence->ring, up to each asic callback
to then do ring specific scheduling if vm ib scheduling function
v8:
- add query for ib limit and kernel reserved virtual space
- rename vm->size to max_pfn (maximum number of page)
- update gem_va ioctl to also allow unmap operation
- bump kernel version to allow userspace to query for vm support
v9:
- rebuild page table only when bind and incrementaly depending
on bo referenced by cs and that have been moved
- allow virtual address space to grow
- use sa allocator for vram page table
- return invalid when querying vm limit on non cayman GPU
- dump vm fault register on lockup
v10: agd5f:
- Move the vm schedule_ib callback to a standalone function, remove
the callback and use the existing ib_execute callback for VM IBs.
v11:
- rebase on top of lastest Linus
v12: agd5f:
- remove spurious backslash
- set IB vm_id to 0 in radeon_ib_get()
v13: agd5f:
- fix handling of RADEON_CHUNK_ID_FLAGS
v14:
- fix va destruction
- fix suspend resume
- forbid bo to have several different va in same vm
v15:
- rebase
v16:
- cleanup left over of vm init/fini
v17: agd5f:
- cs checker
v18: agd5f:
- reworks the CS ioctl to better support multiple rings and
VM. Rather than adding a new chunk id for VM, just re-use the
IB chunk id and add a new flags for VM mode. Also define additional
dwords for the flags chunk id to define the what ring we want to use
(gfx, compute, uvd, etc.) and the priority.
v19:
- fix cs fini in weird case of no ib
- semi working flush fix for ni
- rebase on top of sa allocator changes
v20: agd5f:
- further CS ioctl cleanups from Christian's comments
v21: agd5f:
- integrate CS checker improvements
v22: agd5f:
- final cleanups for release, only allow VM CS on cayman
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Somewhat specializaed sub-allocator designed to perform sub-allocation
for command buffer not only for current cs ioctl but for future command
submission ioctl as well. Patch also convert current ib pool to use
the sub allocator. Idea is that ib poll buffer can be share with other
command buffer submission not having 64K granularity.
v2 Harmonize pool handling and add suspend/resume callback to pin/unpin
sa bo (tested on rv280, rv370, r420, rv515, rv610, rv710, redwood, cayman,
rs480, rs690, rs880)
v3 Simplify allocator
v4 Fix radeon_ib_get error path to properly free fence
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This reverts commit d3ed74027f.
Further upstream discussion between Thomas and Marek decided this needed
more work and driver specifics. So revert before it goes upstream.
Signed-off-by: Dave Airlie <airlied@redhat.com>
With this patch I'm only about 50k larger with DRM debugging
enables (why is that enabled by default?!?), and slightly
smaller without.
[airlied: moved r100.c additions to radeon_ring.c]
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
The new DRM_RADEON_GEM_WAIT ioctl combines GEM_WAIT_IDLE and GEM_BUSY (there
is a NO_WAIT flag to get the latter) with USAGE_READ and USAGE_WRITE flags
to take advantage of the new ttm_bo_wait changes.
Also bump the DRM version.
Signed-off-by: Marek Olšák <maraeo@gmail.com>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Sometimes we want to know whether a buffer is busy and wait for it (bo_wait).
However, sometimes it would be more useful to be able to query whether
a buffer is busy and being either read or written, and wait until it's stopped
being either read or written. The point of this is to be able to avoid
unnecessary waiting, e.g. if a GPU has written something to a buffer and is now
reading that buffer, and a CPU wants to map that buffer for read, it needs to
only wait for the last write. If there were no write, there wouldn't be any
waiting needed.
This, or course, requires user space drivers to send read/write flags
with each relocation (like we have read/write domains in radeon, so we can
actually use those for something useful now).
Now how this patch works:
The read/write flags should passed to ttm_validate_buffer. TTM maintains
separate sync objects of the last read and write for each buffer, in addition
to the sync object of the last use of a buffer. ttm_bo_wait then operates
with one the sync objects.
Signed-off-by: Marek Olšák <maraeo@gmail.com>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Unconditionally initialize the drm gem object - it's not
worth the trouble not to for the few kernel objects.
This patch only changes the place of the drm gem object,
access is still done via pointers.
v2: Uncoditionally align the size in radeon_bo_create. At
least the r600/evergreen blit code didn't to this, angering
the paranoid gem code.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Rather than re-implementing in the Radeon driver,
Use the execbuf / cs / pushbuf utilities that comes with TTM.
This comes with an even greater benefit now that many spinlocks have been
optimized away...
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
The bo lock used only to protect the bo sync object members, and since it
is a per bo lock, fencing a buffer list will see a lot of locks and unlocks.
Replace it with a per-device lock that protects the sync object members on
*all* bos. Reading and setting these members will always be very quick, so
the risc of heavy lock contention is microscopic. Note that waiting for
sync objects will always take place outside of this lock.
The bo device fence lock will eventually be replaced with a seqlock /
rcu mechanism so we can determine that a bo is idle under a
rcu / read seqlock.
However this change will allow us to batch fencing and unreserving of
buffers with a minimal amount of locking.
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Jerome Glisse <j.glisse@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Christian König