Applications may try to allocate lots host memory and reaches the mmap
limit (/proc/sys/vm/max_map_count). When Applications fails to allocate
memory and calls hsaKmtFreeMemory to release the memory, Thunk fails to
reduce the maps count so the following hsaKmtAllocMemory calls continue
to fail, which doesn't make sense to the application. This patch checks
the mmap to NORESERVE return value. If it fails and the error number is
ENOMEM, reduce the map count by munmap and map it again immediately.
Change-Id: I127cb479dfd86b199172eef269d59426f23859ea
Signed-off-by: Amber Lin <Amber.Lin@amd.com>
[ROCm/ROCR-Runtime commit: a81b29890c]
Support all fragment sizes up to 2MB by aligning buffers according
to their size.
Change-Id: I82b7ef8be6f1507d941e5c97edb6618adf8c66de
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
[ROCm/ROCR-Runtime commit: 52598cf37e]
Due to max_mmap_count issue, set default of guard page as disabled.
Change-Id: Ic9dfe69b621733e9fac86831b008a122994a67e7
Signed-off-by: Amber Lin <Amber.Lin@amd.com>
[ROCm/ROCR-Runtime commit: 65d680c035]
Use the mapped_device_id_array size when allocating temp_node_id_array
for unmapping queues in fmm_map_to_gpu_nodes. registered_device_id_array
size may be 0. Also, this temporary array is small enough to allocate it
on the stack. Malloc and free are overkill here.
Fix potential memory leak when registering the same device ID array
multiple times.
Change-Id: I83f09fd0925d9de7cf11bf72ba0ebb77273f587d
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
[ROCm/ROCR-Runtime commit: 395ecaa985]
Add pr_debug to all memory APIs and pr_err to some failure cases.
Change-Id: I8b519a1228cc19e6c04118fd87432e7f48f3cbf9
Signed-off-by: Amber Lin <Amber.Lin@amd.com>
[ROCm/ROCR-Runtime commit: 73707766ef]
Simplify fmm_map_to_gpu_nodes code. Also fix a memory leak in this change.
Change-Id: I3487338b78c915de44588d0206bac4c53e728c60
Signed-off-by: Amber Lin <Amber.Lin@amd.com>
[ROCm/ROCR-Runtime commit: f1a5248cf2]
Fall back to older apertures API and old events page size if the new APIs
fail. This allows running on current upstream kernels (with only minor
fixes) on gfx801 and enables testing of further changes during upstreaming.
Change-Id: I9d86d4f576e52fcbb5bc158d80f1bf41261e4e87
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
[ROCm/ROCR-Runtime commit: 78e683acf4]
Non paged system memory is allocated with node id 0. However, since a
gpu node is required for allocating system memory via KFD, the first
dgpu is used. In hsaKmtShareMemory() if system memory use the same
(first) dgpu.
Change-Id: I85789a89a4e4f7888e3826826401ea89ce4d1718
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
[ROCm/ROCR-Runtime commit: 186527d0b7]
Needed for some tiling formats.
Change-Id: Icd460edaa77ccbeb3c98bc74b574ca5517db22af
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
[ROCm/ROCR-Runtime commit: d563e2cb1d]
This should allow us to take advantage of BigK fragments and huge pages
and improve TLB efficiency for VRAM allocations. Huge pages only work
with 4-level page tables (gfx900 and up). BigK fragments work on older
GPUs.
Change-Id: I02e1fbf74de554e16fdaf44e44d03b47df45c3b0
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
[ROCm/ROCR-Runtime commit: c7bd7733e5]
Imported graphics buffers are most likely images. Align them for
tiled image access. 64KB seems to do the trick.
This fixes VM faults with OpenCL graphics interop.
Change-Id: I7f60e205d93fff9407e0d00d3dbb02cc4990b863
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
[ROCm/ROCR-Runtime commit: dc2c52be78]
Libraries normally don't print messages. We use pr_err, pr_warn,
pr_info, and pr_debug to print messages to stderr when prints are
enabled for debugging.
Change-Id: I9caf719343aa618c88e7b500f9737a46702e424a
Signed-off-by: Amber Lin <Amber.Lin@amd.com>
[ROCm/ROCR-Runtime commit: 897c4e2fff]
The KFD implementation has been removed and will not be upstreamed.
This API has been superseded by hsaKmtRegisterGraphicsHandleToNodes.
Change-Id: I5f2d8da3260974618cdb6ea3fdcd77d37b82c9cb
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Amber Lin <Amber.Lin@amd.com>
[ROCm/ROCR-Runtime commit: 374bd89d8c]
Use checkpatch.pl to fix the majority of errors. Some that remain and
will be excluded:
Use of typedefs/externs/volatile/sscanf
Lines over 80 characters
Remaining errors are due to misunderstanding the * symbol with typedefs
Also use this opportunity to spell manageable properly
Change-Id: I0b335e9cb3e1eea38bee27eaa1f582b2c9b09b38
[ROCm/ROCR-Runtime commit: b78e0e152a]
export HSA_CHECK_USERPTR=1 to check user pointers on registration. If
the pointer doesn't point to a valid mapping, there will be a segfault.
Change-Id: I459c0902cbc90338517fbf79678871ebfbe5183b
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
[ROCm/ROCR-Runtime commit: 34ddde0c50]
Guard pages help catch out-of-bounds memory accesses by applications
by generating VM faults (GPU) and segfaults (CPU).
Remove address space reservation from scratch aperture. That address
space is managed by the Thunk client. Guard pages would cause Thunk's
address space management to get out of sync with the client's.
Change-Id: I2e5aee2923a90186358cc7b0e131baf547996df6
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
[ROCm/ROCR-Runtime commit: 11862b9f61]
Unless HSA_USERPTR_FOR_PAGED_MEM is explicitly set, don't use userptr
for all paged memory. This will also allow us to work around some 4.9
issues, and then we can explicitly set HSA_USERPTR_FOR_PAGED_MEM for
all usage once those issues are resolved.
Change-Id: I25ce22b73ae6e93f1567f2318d9d2b47d4a44e69
[ROCm/ROCR-Runtime commit: c991951288]
Use KFD_IOC_ALLOC_MEM_FLAGS_COHERENT when allocating fine-grained
memory and doorbell BOs so that they will be mapped with MTYPE_UC
on GFX9 hardware.
Change-Id: I51adf45b13105f479e6bcdaf54955b467920ee9a
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Shaoyun Liu <Shaoyun.Liu@amd.com>
[ROCm/ROCR-Runtime commit: 8cb89b6926]
This avoids unnecessary evictions and failed restores due to the
munmap of userptr BOs that are just about to be freed.
Change-Id: Icf2f0b73991455556a201c54c05ea7e20af80f47
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
[ROCm/ROCR-Runtime commit: 74ebfca9f0]
If fork() is called, clear all duplicated data that is invalid in the
child process.
Change-Id: I4e27198060db593c630c6337b7071dfbd0d80b83
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
[ROCm/ROCR-Runtime commit: f1f62d863c]
Implement three new APIs for IPC buffer sharing:
-hsaKmtShareMemory()
-hsaKmtRegisterSharedHandle()
-hsaKmtRegisterSharedHandleToNodes()
Add new ioclts necessary for the above APIs.
Change-Id: Ia2b4d0dc91ec64bff959395d11c0536467404792
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
[ROCm/ROCR-Runtime commit: 559e31d6ff]
A memory region is allowed to be registered multiple times when the memory
is specified by a user pointer. If it's registered with the same user
pointer but with different sizes, it's treated as different instances and
multiple VM objects are created with different GPU address.
Change-Id: I49627111bb5db36d18f1133b252fb62a611f06a4
[ROCm/ROCR-Runtime commit: 2a50ebba98]
Allocate doorbells for dGPUs in the SVM aperture and map them for
GPU access. This is necessary to allow GPU-initiated submissions to
user mode queues.
Depends on new doorbell BO allocation flag in KFD.
Change-Id: I0737bef4a4764bb4a66c43846707ead2108f6601
[ROCm/ROCR-Runtime commit: 2e0a6eb371]
Add vm_find_object_by_userptr_range so QueryPointerInfo can find the
object as well when the pointer is not the starting address but it's
inside the memory range. Also rename vm_find_object_xxx functions to
_by_address and _by_address_range to be consistent.
Change-Id: I5c2b3a05b41493e32b7fd9154665bf078b043606
[ROCm/ROCR-Runtime commit: 4911c91389]
Add CPUVM aperture to keep track of memory allocation that is not known
to GPU driver. Together with GPUVM, this patch adds the pointer attributes
support to APU.
Change-Id: If13f9cf01ff8b9f709b99b66661e7505246adf4c
[ROCm/ROCR-Runtime commit: 19f2676ea7]
Add two pointer attributes APIs:
hsaKmtQueryPointerInfo - allow the user to query the memory information
using a pointer. This pointer can point to any address inside the
range known to HSA.
hsaKmtSetMemoryUserData - allow the user to attach data to a pointer to
add memory tracking information. This pointer must match the start
address of a memory allocation or registration.
TODO: This patch implements support on dGPU. Needs to add APU.
Change-Id: I4711809274248434901f0794f50ebfa13a7371a8
[ROCm/ROCR-Runtime commit: 51e4d27c37]
For APUs, use /proc/cpuinfo to get Marketing name.
Change-Id: I4a17516d26a092683f36631032be00ad44f7e7fe
Signed-off-by: Lan Xiao <Lan.Xiao@amd.com>
[ROCm/ROCR-Runtime commit: df593aa076]
HSA thunk API is currently reporting engineering name to MarketingName
and returning NULL when querying for AMDName.
-Change current name reporting from MarketingName to AMDName.
-Use libpci to get MarketingName
Change-Id: I819a6de7b067a2e724a6695e7d800274b83a71f8
Signed-off-by: Lan Xiao <Lan.Xiao@amd.com>
[ROCm/ROCR-Runtime commit: 9cbbf30be7]
Intermediate size was stored in a 32-bit variable. This resulted in
4GB allocations to fail in KFD due to 0 size. Larger allocations
would allocate the wrong amount of memory.
Change-Id: If19dedf64952f1d2edd813793241e12c0362d220
[ROCm/ROCR-Runtime commit: 82b3fad320]
This option may help debug synchronization or coherency issues
involving the GPU caches. It works only on dGPUs, by changing the
cache policy of the GPUVM default aperture to "cohrent", which is
implemented as non-cached on current dGPU hardware.
Change-Id: I544ac9cc5c0cf1fa5c4e30f67aa42b3b5e44ae67
[ROCm/ROCR-Runtime commit: 06d391c6c9]
The Runtime requested this information so they can tell easily
whether a pointer is part of HSA shared address space or not.
Change-Id: If2041ed34031636677d692bc2dc6625634027ed4
[ROCm/ROCR-Runtime commit: 0ed29f5191]
Lstopo doesn't have system memory mappings at low addresses. Make
sure we leave enough GPUVM address space for kernel allocations
(currently only CWSR) before the start of the user-managed SVM
aperture.
Change-Id: Ic197f7bd5a3cfb150a0da2bfdbc848664e7869be
[ROCm/ROCR-Runtime commit: cac0c08496]
Non-canonical GPUVM aperture doesn't exist on dGPUs. Remove comments
and code that say otherwise.
Fix alignment of GPUVM aperture for gfx801. Requires the same workaround
as gfx802. It's not used for anything on gfx801 yet, but will be soon.
Change-Id: I88607fe7b340081cc0715b85f28fdbf5f1bb0ad7
[ROCm/ROCR-Runtime commit: b837c3e7b0]
Use the object size when freeing address space, instead of the
parameter passed in by the caller. The parameter may be incorrect
due to app or runtime bugs, or when the buffers is an AQL ring
buffer with double mapping workaround.
Change-Id: I00bb31d4520ef969a49d6d5ea723e8a33418acc3
[ROCm/ROCR-Runtime commit: 006f3ee41b]
The alignment performed in vm_find_object_by_address isn't sufficient
because it doesn't take into account the offset from the start of the
page.
This fixes a bug where certain unaligned userpointers and sizes fail
to register correctly.
Change-Id: I17872e264467a619f5e1bedb7e1ed3d994a856bf
[ROCm/ROCR-Runtime commit: 8a0161d6bb]
Amber Lin
Felix Kuehling
Harish Kasiviswanathan
Kent Russell
Sean Keely
Lan Xiao
Ben Goz