Commit Graph

51 Commits

Author SHA1 Message Date
Christian König
765e7fbf08 drm/amdgpu: rename all rbo variable to abo v2
Just to cleanup some radeon leftovers.

sed -i "s/rbo/abo/g" drivers/gpu/drm/amd/amdgpu/*.c
sed -i "s/rbo/abo/g" drivers/gpu/drm/amd/amdgpu/*.h

v2: rebased

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>
2016-09-28 16:16:20 -04:00
Christian König
1927ffc0c1 drm/amdgpu: remove unused member from struct amdgpu_bo
Not used in a while.

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>
2016-09-28 16:16:20 -04:00
Christian König
bb990bb092 drm/amdgpu: add a custom GTT memory manager v2
Only allocate address space when we really need it.

v2: fix a typo, add correct function description,
    stop leaking the node in the error case.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2016-09-28 16:16:19 -04:00
Baoyou Xie
9f31a0b032 drm/amdgpu: clean function declarations in amdgpu_ttm.c up
We get 2 warnings when building kernel with W=1:

drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c:985:5: warning: no previous prototype for 'amdgpu_ttm_init' [-Wmissing-prototypes]
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c:1092:6: warning: no previous prototype for 'amdgpu_ttm_fini' [-Wmissing-prototypes]

In fact, both functions are declared in
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c, but should be declared in
a header file, thus can be recognized in other file.

So this patch moves the declarations into
drivers/gpu/drm/amd/amdgpu/amdgpu.h.

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Baoyou Xie <baoyou.xie@linaro.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2016-09-16 15:53:04 -04:00
Christian König
9702d40dd3 drm/amdgpu: add AMDGPU_BO_INVALID_OFFSET
An invalid offset to be used by custom VRAM/GGT manager when we
don't have an offset for an BO.

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>
2016-09-14 15:10:44 -04:00
Flora Cui
441f90ecf5 drm/amdgpu: update gart_pin_size if bo is in GTT
Signed-off-by: Flora Cui <Flora.Cui@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2016-09-14 15:10:31 -04:00
Christian König
c855e25090 drm/amdgpu: bind GTT on demand
We don't really need the GTT table any more most of the time. So bind it
only on demand.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2016-09-14 15:10:30 -04:00
Junwei Zhang
aa1d562e64 drm/amdgpu: add a new helper to free a BO in kernel allocations
Free the BO allocated by amdgpu_bo_create_kernel()

Signed-off-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2016-09-14 15:10:26 -04:00
Christian König
6369f6f133 drm/amdgpu: stop adding dummy entry in amdgpu_ttm_placement_init
AMDGPU_GEM_CREATE_NO_CPU_ACCESS and AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED are
obviously mutual exclusive. So stop adding a dummy entry without effect when
both are specified.

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>
2016-08-22 13:47:24 -04:00
Christian König
faceaf6a73 drm/amdgpu: cleanup amdgpu_ttm_placement_init
Make it more clear what this function does. No intendet functional change.

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>
2016-08-22 13:47:23 -04:00
Chunming Zhou
0c4e7fa561 drm/amdgpu: link all shadow bo V2
V2:
1. use mutex instead of spinlock for shadow list, since its process could
sleep.
2. move list_del to bo destroy phase.

Signed-off-by: Chunming Zhou <David1.Zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2016-08-22 13:47:20 -04:00
Chunming Zhou
20f4eff1c8 drm/amdgpu: sync bo and shadow V3
Use shadow flag to judge which direction to sync.
V2:
Don't need bo pin, so remove it.

V3:
1. Split to two functions, one is backup_to_shadow, another is
restore_from_shadow.
2. Clean up previous shadow direction difinitions.

Signed-off-by: Chunming Zhou <David1.Zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2016-08-22 13:47:18 -04:00
Chunming Zhou
3ad81f167e drm/amdgpu: add need backup function V2
V2:
add checking if need backup in amdgpu_bo_create.

Signed-off-by: Chunming Zhou <David1.Zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2016-08-19 12:33:22 -04:00
Flora Cui
27798e07ed drm/amdgpu: use domain's gpu_offset for start addr
Signed-off-by: Flora Cui <Flora.Cui@amd.com>
Reviewed-by: Ken Wang <Qingqing.Wang@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2016-08-19 12:32:38 -04:00
Flora Cui
32ab75f076 drm/amdgpu: update gart_pin_size only if the bo is pined to GTT
Signed-off-by: Flora Cui <Flora.Cui@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2016-08-19 12:31:36 -04:00
Flora Cui
408778e8e9 drm/amdgpu: check domain sanity in amdgpu_bo_pin_restricted()
abort if the bo is pined to other domain already

Signed-off-by: Flora Cui <Flora.Cui@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2016-08-19 12:31:19 -04:00
Christian König
6681c5eba1 drm/amdgpu: fix coding style in amdgpu_object.c
Just a few 80 chars problems.

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>
2016-08-16 10:44:55 -04:00
Chunming Zhou
e7893c4bd3 drm/amdgpu: add shadow bo support V2
shadow bo is the shadow of a bo, which is always in GTT,
which can be used to backup the original bo.
V2:
reference shadow parent, shadow bo will be freed by who allocted him.

Signed-off-by: Chunming Zhou <David1.Zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2016-08-16 10:40:01 -04:00
Flora Cui
4fea83ff0f drm/amdgpu: expose AMDGPU_GEM_CREATE_VRAM_CLEARED to user space
V2: fix the return value for fill failure and validate bo before
filling data

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Flora Cui <Flora.Cui@amd.com>
Reviewed-by: Chunming Zhou <David1.Zhou@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2016-08-08 11:33:01 -04:00
Christian König
cdb7e8f273 drm/amdgpu: add more warning to amdgpu_bo_offset
Warn when we try to get the address and the BO isn't locked or reserved.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2016-08-08 11:32:17 -04:00
Christian König
7c204889de drm/amdgpu: add new helper for in kernel allocations
We often allocate, pin and map things at the same time in the kernel.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2016-08-08 11:32:15 -04:00
David Mao
15da301d88 drm/amd/amdgpu : adding new tracepoints to track memory information.
- adding amdgpu_cs_bo_status to track total size and
   total entry count of bo for each submission.
 - adding amdgpu_ttm_bo_move to track the bo eviction
   including the size of bo and the location before/after the move

Signed-off-by: David Mao <David.Mao@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2016-07-07 14:51:29 -04:00
Dave Airlie
0092d3edcb drm/amdgpu: set metadata pointer to NULL after freeing.
Without this there was a double free of the metadata,
which ended up freeing the fd table for me here, and taking
out the machine more often than not.

I reproduced with X.org + modesetting DDX + latest llvm/mesa,
also required using dri3.

Cc: stable@vger.kernel.org
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2016-05-03 11:53:44 -04:00
Chunming Zhou
e131b914c1 drm/amdgpu: add invisible pin size statistic
Signed-off-by: Chunming Zhou <David1.Zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2016-04-06 14:42:59 -04:00
Alex Deucher
1f8628c764 drm/amdgpu: print vram type rather than just DDR
We have the info, so use it rather than reporting just DDR.

Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2016-04-01 10:08:37 -04:00
Michel Dänzer
104ece9757 drm/amdgpu: Don't move pinned BOs
The purpose of pinning is to prevent a buffer from moving.

Reviewed-by: Christian König <christian.koenig@amd.com>
Tested-by: Rex Zhu <Rex.Zhu@amd.com>
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2016-03-28 11:55:38 -04:00
Christian König
587f3c70aa drm/amdgpu: always wait before kmap a BO
When a BO is currently moving we otherwise would blindly
access the new location without checking.

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>
2016-03-14 13:43:09 -04:00
Christian König
418aa0c296 drm/amdgpu: cleanup gem init/finit
Remove the double housekeeping and use something sane to
forcefuly delete BOs on unload.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
2016-02-16 17:46:24 -05:00
Christian König
cc325d1913 drm/amdgpu: check userptrs mm earlier
Instead of when we try to bind it check the usermm when
we try to use it in the IOCTLs.

Signed-off-by: Christian König <christian.koenig@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
2016-02-10 14:17:16 -05:00
Christian König
1ea863fd73 drm/amdgpu: keep the prefered/allowed domains in the BO
Stop copying that to the bo list entry, it doesn't change anyway.

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>
2016-02-10 14:16:47 -05:00
Oded Gabbay
a187f17f0e drm/amdgpu: mask out WC from BO on unsupported arches
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
2016-02-02 10:09:11 -05:00
Christian König
78d0e182b6 drm/amdgpu: fix amdgpu_bo_pin_restricted VRAM placing v2
We could pin BOs into invisible VRAM otherwise.

v2: make logic more readable as suggested by Michel

Cc: stable@vger.kernel.org
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com> (v1)
Reviewed-by: Rex Zhu <Rex.Zhu@amd.com> (v1)
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
2016-01-22 10:47:48 -05:00
Christian König
82b9c55b1e drm/amdgpu: fix VM page table reference counting
We use the reservation object of the page directory for the page tables as
well, because of this the page directory should be freed last. Ensure that
by keeping a reference from the page tables to the directory.

Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2015-11-30 14:41:33 -05:00
Chunming Zhou
95d7918323 drm/amdgpu: add TOPDOWN flag to the whole vram
need to decrease visible vram usage by default.

Signed-off-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: monk.liu <monk.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
2015-10-07 23:48:22 -04:00
Andrzej Hajda
71affda522 drm/amdgpu: use kmemdup rather than duplicating its implementation
The patch was generated using fixed coccinelle semantic patch
scripts/coccinelle/api/memdup.cocci [1].

[1]: http://permalink.gmane.org/gmane.linux.kernel/2014320

Signed-off-by: Andrzej Hajda <a.hajda@samsung.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2015-09-23 17:23:41 -04:00
Christian König
72d7668b5b drm/amdgpu: export reservation_object from dmabuf to ttm (v2)
Adds an extra argument to amdgpu_bo_create, which is only used in amdgpu_prime.c.

Port of radeon commit 831b6966a6.

v2: fix up kfd.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Jammy Zhou <Jammy.Zhou@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
2015-09-23 17:23:34 -04:00
Alex Deucher
cace5dce55 drm/amdgpu: use top down allocation for non-CPU accessible vram
Should help avoid fragmentation of vram due to CPU access
requirements.

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2015-09-03 10:29:32 -04:00
Daniel Vetter
db4448f78d drm/amdgpu: Don't take dev->struct_mutex in bo_force_delete
It really doesn't protect anything which doesn't have other locks
already. Also this is run from driver unload code so not much need for
locks anyway.

Same changes as for radeon really.

Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: "Christian König" <christian.koenig@amd.com>
Reviewed-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2015-08-17 16:51:13 -04:00
Christian König
c548b345d1 drm/amdgpu: remove VI hw bug workaround v3
The workaround simply doesn't work because VM mappings
are controlled by userspace not the kernel.

Additional to that this is just a performance problem
which happens if you have holes in your VM mapping.

v2: adjust virtual addr alignment as well.
v3: fix trivial warning

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Monk Liu  <monk.liu@amd.com> (v1)
Reviewed-by: Jammy Zhou <Jammy.Zhou@amd.com> (v2)
2015-08-17 16:51:11 -04:00
Chunming Zhou
e40a31159b drm/amdgpu: use kernel fence diretly in amdgpu_bo_fence
Signed-off-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Christian K?nig <christian.koenig@amd.com>
Reviewed-by: Jammy Zhou <jammy.zhou@amd.com>
2015-08-17 16:50:50 -04:00
Alex Deucher
d07f5c4c23 drm/amdgpu: remove VM workaround for Fiji
The bug is fixed in fiji.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2015-08-17 16:50:29 -04:00
Jammy Zhou
2f7d10b393 drm/amdgpu: merge amdgpu_family.h into amd_shared.h (v2)
Make the definitions common for all driver components

v2: fix kfd

Signed-off-by: Jammy Zhou <Jammy.Zhou@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
2015-08-17 16:50:21 -04:00
Christian König
9298e52f8b drm/amdgpu: fix user ptr race condition
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: monk liu <monk.liu@amd.com>
2015-06-05 10:59:32 -04:00
Christian König
e176fe176d drm/amdgpu: remove mclk_lock
Not needed any more.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
2015-06-03 21:03:58 -04:00
Chunming Zhou
7e5a547f64 drm/amdgpu: implement the allocation range (v3)
Pass a ttm_placement pointer to amdgpu_bo_create_restricted
add min_offset to amdgpu_bo_pin_restricted.  This makes it
easier to allocate memory with address restrictions.  With
this patch we can also enable 2-ended allocation again.

v2: fix rebase conflicts
v3: memset placements before using

Reviewed-by: Jammy Zhou <Jammy.Zhou@amd.com>
Signed-off-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
2015-06-03 21:03:52 -04:00
Christian König
5fb1941d0c drm/amdgpu: port fault_reserve_notify changes from radeon
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
2015-06-03 21:03:49 -04:00
Marek Olšák
fbd76d59ef drm/amdgpu: rework tiling flags
Signed-off-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
2015-06-03 21:03:46 -04:00
Christian König
271c812566 drm/amdgpu: enforce AMDGPU_GEM_CREATE_NO_CPU_ACCESS
Deny user and kernel mapping if we said we never want to do so.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Jammy Zhou <Jammy.Zhou@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
2015-06-03 21:03:43 -04:00
Jammy Zhou
886712881d drm/amdgpu: remove AMDGPU_GEM_CREATE_CPU_GTT_UC
This flag isn't used by user mode drivers, remove it to avoid
confusion. And rename GTT_WC to GTT_USWC to make it clear.

Signed-off-by: Jammy Zhou <Jammy.Zhou@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
2015-06-03 21:03:24 -04:00
Alex Deucher
2d8bd23a05 drm/amdgpu: drop ttm two ended allocation
amdgpu_bo_create() calls amdgpu_ttm_placement_from_domain()
before ttm_bo_init() is called.  amdgpu_ttm_placement_from_domain()
uses the ttm bo size to determine when to select top down
allocation but since the ttm bo is not initialized yet the
check is always false.  It only took affect when buffers
were validated later.  It also seemed to regress suspend
and resume on some systems possibly due to it not
taking affect in amdgpu_bo_create().

amdgpu_bo_create() and amdgpu_ttm_placement_from_domain()
need to be reworked substantially for this to be optimally
effective.  Re-enable it at that point.

Ported from radeon commit:
a239118a24

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
2015-06-03 21:03:20 -04:00