aboutsummaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
AgeCommit message (Collapse)AuthorFilesLines
2017-05-24drm/amdgpu: fix NULL pointer panic of emit_gds_switchChunming Zhou1-1/+1
[ 338.384770] BUG: unable to handle kernel NULL pointer dereference at (null) [ 338.384817] IP: [< (null)>] (null) [ 338.385505] RIP: 0010:[<0000000000000000>] [< (null)>] (null) [ 338.385950] Call Trace: [ 338.385993] [<ffffffffa05d2313>] ? amdgpu_vm_flush+0x283/0x400 [amdgpu] [ 338.386025] [<ffffffff811818d3>] ? printk+0x4d/0x4f [ 338.386074] [<ffffffffa05d4906>] amdgpu_ib_schedule+0x4a6/0x4d0 [amdgpu] [ 338.386140] [<ffffffffa0673e54>] amdgpu_job_run+0x64/0x180 [amdgpu] [ 338.386203] [<ffffffffa0672e09>] amd_sched_main+0x2e9/0x4a0 [amdgpu] [ 338.386232] [<ffffffff810bfce0>] ? prepare_to_wait_event+0x110/0x110 [ 338.386295] [<ffffffffa0672b20>] ? amd_sched_select_entity+0xe0/0xe0 [amdgpu] [ 338.386327] [<ffffffff8109b423>] kthread+0xd3/0xf0 [ 338.386349] [<ffffffff8109b350>] ? kthread_park+0x60/0x60 [ 338.386376] [<ffffffff817e1ee5>] ret_from_fork+0x25/0x30 [ 338.386401] Code: Bad RIP value. [ 338.386420] RIP [< (null)>] (null) [ 338.386443] RSP <ffffc90001bd7d40> [ 338.386458] CR2: 0000000000000000 [ 338.398508] ---[ end trace 4c66fcdc74b9a0a2 ]--- Signed-off-by: Chunming Zhou <[email protected]> Reviewed-by: Christian König <[email protected]> Reviewed-by: Alex Deucher <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-05-24drm/amdgpu: fix fundamental suspend/resume issueChristian König1-1/+21
Reinitializing the VM manager during suspend/resume is a very very bad idea since all the VMs are still active and kicking. This can lead to random VM faults after resume when new processes become the same client ID assigned. Signed-off-by: Christian König <[email protected]> Reviewed-by: Alex Deucher <[email protected]> Signed-off-by: Alex Deucher <[email protected]> Cc: [email protected]
2017-05-10drm/amdgpu: fix dependency issueChunming Zhou1-1/+1
The problem is that executing the jobs in the right order doesn't give you the right result because consecutive jobs executed on the same engine are pipelined. In other words job B does it buffer read before job A has written it's result. Signed-off-by: Chunming Zhou <[email protected]> Reviewed-by: Christian König <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-04-28drm/amdgpu: PRT support for gfx9 (v3)Zhang, Jerry1-0/+6
Fix PRT handling on gfx9 v2: unify PRT bit for all ASICs v3: move PRT flag checking in amdgpu_vm_bo_split_mapping() Signed-off-by: Junwei Zhang <[email protected]> Acked-by: David Zhou <[email protected]> Reviewed-by: Christian König <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-04-28drm/amdgpu: fix amdgpu_vm_clear_freed v2Christian König1-2/+3
Use amdgpu_vm_bo_update_mapping() instead of amdgpu_vm_bo_split_mapping() here. We don't want any flags set in the cleared areas and splitting shouldn't be necessary. v2: fix typo in commit message Signed-off-by: Christian König <[email protected]> Reviewed-by: Alex Deucher <[email protected]> Reviewed-by: Junwei Zhang <[email protected]> Signed-off-by: Alex Deucher <[email protected]> Cc: [email protected]
2017-04-28drm/amdgpu: trace vm hub during flush as well v2Christian König1-1/+1
Trace on which hub we are doing the flush. v2: fix typo in commit message Signed-off-by: Christian König <[email protected]> Reviewed-by: Alex Deucher <[email protected]> Reviewed-by: Andres Rodriguez <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-04-28drm/amdgpu: trace the vmhub in grab_id as wellChristian König1-1/+1
Trace on which VMHUB we assigned an VMID. Signed-off-by: Christian König <[email protected]> Reviewed-by: Alex Deucher <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-04-28drm/amdgpu: allow concurrent VM flushesChristian König1-22/+28
Enable concurrent VM flushes for Vega10. Signed-off-by: Christian König <[email protected]> Acked-by: Alex Deucher <[email protected]> Reviewed-by: Andres Rodriguez <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-04-28drm/amdgpu: invalidate only the currently needed VMHUB v2Christian König1-2/+1
Drop invalidating both hubs from each engine. v2: don't use hardcoded values Signed-off-by: Christian König <[email protected]> Reviewed-by: Alex Deucher <[email protected]> Reviewed-by: Andres Rodriguez <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-04-28drm/amdgpu: split VMID management by VMHUBChristian König1-32/+50
This way GFX and MM won't fight for VMIDs any more. Initially disabled since we need to stop flushing all HUBS at the same time as well. Signed-off-by: Christian König <[email protected]> Reviewed-by: Andres Rodriguez <[email protected]> Acked-by: Alex Deucher <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-04-28drm/amdgpu: drop VMID per ring trackingChristian König1-14/+3
David suggested this a long time ago, instead of checking each ring just walk over all the VMIDs in reverse LRU order. Signed-off-by: Christian König <[email protected]> Reviewed-by: Andres Rodriguez <[email protected]> Reviewed-by: Alex Deucher <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-04-07drm/amdgpu: set vm size and block size by individual gmc by default (v3)Junwei Zhang1-0/+38
By default, the value is set by individual gmc. if a specific value is input, it overrides the global value for all v2: create helper funcs v3: update gmc9 APU's num_level athough it may be updated in the future. Signed-off-by: Junwei Zhang <[email protected]> Reviewed-by: Christian König <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-04-06drm/amdgpu: fix vm size and block size for VMPT (v5)Zhang, Jerry1-10/+12
Set reasonable defaults per family. v2: set both of them in gmc v3: move vm size and block size in vm manager v4: squash in warning fix from Alex Xie v5: squash in min() warning fix Signed-off-by: Junwei Zhang <[email protected]> Reviewed-by: Alex Deucher <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-04-04drm/amdgpu: move adjust_mc_addr into amdgpu_gart_funcsChristian König1-2/+2
We should probably rename amdgpu_gart_funcs sooner or later. Reviewed-by: Junwei Zhang <[email protected]> Signed-off-by: Christian König <[email protected]> Reviewed-by: Alex Deucher <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-04-04drm/amdgpu: cleanup logic in amdgpu_vm_flushChristian König1-10/+10
Remove some of the extra checks where they don't hurt us. Reviewed-by: Junwei Zhang <[email protected]> Signed-off-by: Christian König <[email protected]> Reviewed-by: Alex Deucher <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-04-04drm/amdgpu: cleanup coding style in amdgpu_vm_flushChristian König1-43/+45
Abort early if there is nothing todo and correctly indent the "if"s. Reviewed-by: Junwei Zhang <[email protected]> Signed-off-by: Christian König <[email protected]> Reviewed-by: Alex Deucher <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-04-04drm/amdgpu: coding style of amdgpu_vm_is_gpu_resetChristian König1-7/+14
The name is a bit confusing and the extra "? true : false" is superflous. Additional to that remove setting the reset counter directly after checking it. Reviewed-by: Junwei Zhang <[email protected]> Signed-off-by: Christian König <[email protected]> Reviewed-by: Alex Deucher <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-04-04drm/amdgpu: remove VMID first trackingChristian König1-4/+0
Not used any more. Reviewed-by: Junwei Zhang <[email protected]> Signed-off-by: Christian König <[email protected]> Reviewed-by: Alex Deucher <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-04-04drm/amdgpu: Fix a NULL deref in amdgpu_vm_add_prt_cb()Dan Carpenter1-1/+1
We accidentally dereference "cb" if the kmalloc() fails. Fixes: 451bc8eb8fe6 ("drm/amdgpu: fix PRT teardown on VM fini v3") Reviewed-by: Christian König <[email protected]> Reviewed-by: Harry Wentland <[email protected]> Signed-off-by: Dan Carpenter <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-04-04drm/amdgpu: use a 64bit interval tree for VM management v2Christian König1-44/+49
This only makes a difference for 32-bit systems. The idea is to have a fixed virtual address space size with 4-level page tables and to minimize differences between 32 and 64-bit systems. v2: Update commit message. Signed-off-by: Christian König <[email protected]> Reviewed-by: Felix Kuehling <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: Fix multi-level page table bugs for large BOs v3Felix Kuehling1-6/+18
Fix the start/end address calculation for address ranges that span multiple page directories in amdgpu_vm_alloc_levels. Add error messages if page tables aren't found. Otherwise the page table update would just fail silently. v2: * Change WARN_ON to WARN_ON_ONCE * Move masking of high address bits to caller * Add range-check for "from" and "to" v3: * Replace WARN_ON_ONCE in get_pt with pr_err in caller Signed-off-by: Felix Kuehling <[email protected]> Reviewed-by: Christian König <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: Make max_pfn 64-bitFelix Kuehling1-2/+2
With 4-level page tables the maximum VM size is 256TB. That's 64G pages, which can't be represented in 32-bit. Signed-off-by: Felix Kuehling <[email protected]> Reviewed-by: Chunming Zhou <[email protected]> Reviewed-by: Junwei Zhang <[email protected]> Reviewed-by: Christian König <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu:changes in gfx DMAframe scheme (v2)Monk Liu1-30/+47
1) Adapt to vulkan: Now use double SWITCH BUFFER to replace the 128 nops w/a, because when vulkan introduced, umd can insert 7 ~ 16 IBs per submit which makes 256 DW size cannot hold the whole DMAframe (if we still insert those 128 nops), CP team suggests use double SWITCH_BUFFERs, instead of tricky 128 NOPs w/a. 2) To fix the CE VM fault issue when MCBP introduced: Need one more COND_EXEC wrapping IB part (original one us for VM switch part). this change can fix vm fault issue caused by below scenario without this change: >CE passed original COND_EXEC (no MCBP issued this moment), proceed as normal. >DE catch up to this COND_EXEC, but this time MCBP issued, thus DE treats all following packages as NOP. The following VM switch packages now looks just as NOP to DE, so DE dosen't do VM flush at all. >Now CE proceeds to the first IBc, and triggers VM fault, because DE didn't do VM flush for this DMAframe. 3) change estimated alloc size for gfx9. with new DMAframe scheme, we need modify emit_frame_size for gfx9 4) No need to insert 128 nops after gfx8 vm flush anymore because there was double SWITCH_BUFFER append to vm flush, and for gfx7 we already use double SWITCH_BUFFER following after vm_flush so no change needed for it. 5) Change emit_frame_size for gfx8 v2: squash in BUG removal from Monk Signed-off-by: Monk Liu <[email protected]> Acked-by: Alex Deucher <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: add alloc/free for multi level PDs V2Christian König1-69/+106
Allocate and free page directories on demand. V2: a. clear entries allocation b. fix entries index calculation c. need alloc sub level even parent bo was allocated Signed-off-by: Christian König <[email protected]> (v1) Reviewed-by: Alex Deucher <[email protected]> (v1) Signed-off-by: Chunming Zhou <[email protected]> (v2) Acked-by: Alex Deucher <[email protected]> (v2) Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: handle multi level PD during PT updatesChristian König1-5/+34
Not the best solution, but good enough for now. Signed-off-by: Christian König <[email protected]> Reviewed-by: Alex Deucher <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: handle multi level PD updates V2Christian König1-33/+64
Update all levels of the page directory. V2: a. sub level pdes always are written to incorrect place. b. sub levels need to update regardless of parent updates. Signed-off-by: Christian König <[email protected]> (V1) Reviewed-by: Alex Deucher <[email protected]> (V1) Signed-off-by: Chunming Zhou <[email protected]> (V2) Acked-by: Alex Deucher <[email protected]> (V2) Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: handle multi level PD in the LRUChristian König1-9/+27
Move all levels to the end after command submission. Signed-off-by: Christian König <[email protected]> Reviewed-by: Alex Deucher <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: handle multi level PD during validationChristian König1-15/+42
All page directory levels should be in place after this. Signed-off-by: Christian König <[email protected]> Reviewed-by: Alex Deucher <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: handle multi level PD size calculation (v2)Christian König1-12/+22
Allows us to get the size for all levels as well. v2: agd: fix warning Signed-off-by: Christian König <[email protected]> Reviewed-by: Alex Deucher <[email protected]> Reviewed-by: Junwei Zhang <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: generalize page table levelChristian König1-44/+44
No functional change, but the base for multi level page tables. Signed-off-by: Christian König <[email protected]> Reviewed-by: Alex Deucher <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: add the VM pointer to the amdgpu_pte_update_params as wellChristian König1-10/+11
This way we save passing it through the different functions. Signed-off-by: Christian König <[email protected]> Reviewed-by: Alex Deucher <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: rename page_directory_fence to last_dir_updateChristian König1-4/+4
Decribes better what this is used for. Signed-off-by: Christian König <[email protected]> Reviewed-by: Alex Deucher <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: Add GMC 9.0 support (v2)Alex Xie1-6/+22
On SOC-15 parts, the GMC (Graphics Memory Controller) consists of two hubs: GFX (graphics and compute) and MM (sdma, uvd, vce). v2: drop sdma from Makefile, fix duplicate return statement. Signed-off-by: Alex Xie <[email protected]> Reviewed-by: Alex Deucher <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: handle PTE MTYPE in amdgpu_vm_bo_split_mappingAlex Xie1-0/+3
Reviewed-by: Christian König <[email protected]> Signed-off-by: Alex Xie <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: handle PTE EXEC in amdgpu_vm_bo_split_mappingAlex Xie1-0/+3
Reviewed-by: Christian König <[email protected]> Signed-off-by: Alex Xie <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: add optional fence out-parameter to amdgpu_vm_clear_freedNicolai Hähnle1-6/+15
We will add the fence to freed buffer objects in a later commit, to ensure that the underlying memory can only be re-used after all references in page tables have been cleared. Signed-off-by: Nicolai Hähnle <[email protected]> Reviewed-by: Chunming Zhou <[email protected]> Reviewed-by: Junwei Zhang <[email protected]> Reviewed-by: Christian König <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: fix duplicated codeChunming Zhou1-4/+0
it could come from branch merge. Signed-off-by: Chunming Zhou <[email protected]> Reviewed-by: Junwei Zhang <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: fix before and after mapping judgement for replace mappingJunwei Zhang1-4/+6
If the before mapping is 1 page size, so its start and last will be same. Thus below condition will become false, then to free the before mapping. > if (before->it.start != before->it.last) But in this case, we need the before mapping of 1 page size. So does after mapping. Signed-off-by: Junwei Zhang <[email protected]> Reviewed-by: Christian König <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: add a VM mapping replace operation v2Christian König1-0/+64
Add a new operation to replace mappings in a VM with a new one. v2: Fix Jerry's comment, separate out clear operation. Signed-off-by: Christian König <[email protected]> Reviewed-by: Junwei Zhang <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: implement AMDGPU_VA_OP_CLEAR v2Christian König1-0/+99
A new VM operation to remove all mappings in a range. v2: limit unmapped area as noted by Jerry Signed-off-by: Christian König <[email protected]> Reviewed-by: Junwei Zhang <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: separate page table allocation from mappingChristian König1-61/+75
This makes it easier to implement a replace operation. Signed-off-by: Christian König <[email protected]> Reviewed-by: Junwei Zhang <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: make set_prt callback optional and fix error handlingChristian König1-14/+15
PRT support is completely implemented now and we left it turned on accidentially in the error path. Signed-off-by: Christian König <[email protected]> Reviewed-by: Junwei Zhang <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: expand pte flags to uint64_tChunming Zhou1-9/+9
Necessary for new asics. Signed-off-by: Chunming Zhou <[email protected]> Reviewed-by: Christian König <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: fix PRT teardown on VM fini v3Christian König1-25/+81
v2: new approach fixing this by registering a fence callback for all users of the VM on teardown v3: agd: rebase Signed-off-by: Christian König <[email protected]> Reviewed-by: Nicolai Hähnle <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: add OOM fallback on PRT teardown (v2)Christian König1-6/+22
Don't assume kmalloc will always succeed. v2: agd: rebase Signed-off-by: Christian König <[email protected]> Reviewed-by: Nicolai Hähnle <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: minor PRT turnoff fix (v2)Christian König1-7/+5
When two VMs stop using PRT support at the same time we might not disable it in the right order otherwise. v2: agd: rebase Signed-off-by: Christian König <[email protected]> Reviewed-by: Nicolai Hähnle <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: add basic PRT support (v2)Christian König1-11/+90
Future hardware generations can handle PRT flags on a per page basis, but current hardware can only turn it on globally. Add the basic handling for both, a global callback to enable/disable triggered by setting a per mapping flag. v2: agd: rebase fixes Signed-off-by: Christian König <[email protected]> Reviewed-by: Nicolai Hähnle <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-03-29drm/amdgpu: add support for BO_VAs without BO v2Christian König1-6/+14
For PRT support we need mappings which aren't backed by any memory. v2: fix parameter checking Signed-off-by: Christian König <[email protected]> Reviewed-by: Nicolai Hähnle <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2017-01-27drm/amdgpu: fix amdgpu_bo_va_mapping flagsChristian König1-1/+1
They are 64bit not 32 for a while now. Signed-off-by: Christian König <[email protected]> Reviewed-by: Nicolai Hähnle <[email protected]> Reviewed-by: Alex Deucher <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
2016-11-23drm/amdgpu: use AMDGPU_GEM_CREATE_VRAM_CLEARED for VM PD/PTs (v2)Christian König1-95/+4
Doesn't make much sense to have the same functionality twice. v2: rebase on dma_fence renaming Signed-off-by: Christian König <[email protected]> Reviewed-by: Alex Deucher <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
Powered by cgit, Gruvboxed by Blaster4385.