[Issue]: [KFD/SDMA0] Intermittent GPU page fault in hipMemcpyAsync from file-backed mmap on gfx1100 + gfx1101 (Linux 7.3, ROCm 7.2.4)
まだ誰も着手していません。
評価
- 難易度
- 5/5
- 見積もり時間
- 1週間以上
- 初心者へのやさしさ
- 30/100
- issue の種類
- バグ
- 明瞭さ
- 明確に書かれている
- 活発さ
- 活発
調査の方向性
Start with the standalone reproducer at tools/rocm-mmap-sdma-repro/rocm_mmap_sdma_repro.cpp and its README, then review the KFD/SVM and HIP copy paths for how file-backed mmap pages are handled by hipMemcpyAsync. Reproduce using the provided command and compare direct mmap transfers with --mode malloc; done means identifying why valid mapped pages can trigger the reported SDMA0 fault and confirming the finding with a focused test.
索引モデルが issue の本文から書いたものです。
説明
Problem Description
Summary
A standalone HIP C++ program intermittently crashes with an SDMA0 GPU page fault while copying from a read-only, file-backed mmap using hipMemcpyAsync on a two-discrete-GPU RDNA3 workstation.
Importantly, this reproduces immediately after a clean reboot, before running any llama.cpp or ggml process. On October 25, 2026, the first standalone execution completed 250 iterations. The second identical execution SIGABRTed after the first iteration on both GPUs. Subsequent identical commands, including traced runs, sometimes complete 250 iterations. The problem is intermittent and timing-sensitive.
Kernel logs show Faulty UTCL2 client ID: SDMA0 (0xd) and the faulting address is inside the valid file-backed source mapping.
System
- Distribution: CachyOS (Arch-based rolling release)
- Kernel: 7.3.0-5-cachyos, in-tree
amdgpu, withamdgpu.runpm=0 - ROCm userspace: 7.2.4 (
hip-runtime-amd 7.2.4-1.1,hsa-rocr 7.2.4-1.1) - HIP: 7.2.53211-3d9ef42, HIP runtime
rocclr; HSA runtime 1.18 - CPU: AMD Ryzen 9 7950X, 64 GiB RAM
- GPU 0: AMD Radeon RX 7900 XTX (
gfx1100, PCI 0000:03:00.0, vendor/device 1002:744c) - GPU 1: AMD Radeon RX 7800 XT (
gfx1101, PCI 0000:14:00.0, vendor/device 1002:747e)
Standalone HIP reproducer
No llama.cpp, ggml, inference, model loading logic, or HIP compute kernels are necessary. The program repeatedly copies source chunks from a large file-backed mapping to GPU buffers using HIP.
/opt/rocm/bin/hipcc -O2 -std=c++17 -pthread rocm_mmap_sdma_repro.cpp -o rocm_mmap_sdma_repro
./rocm_mmap_sdma_repro \
--model-file /path/to/gguf-blob \
--file-offset 0x7c2331d40 \
--map whole \
--source-span-mib 1024 \
--pattern random \
--seed 20261025 \
--devices 0,1 \
--iterations 250 \
--mode direct \
--sync each
Source file: a publicly available GGUF shard of size 38,842,920,384 bytes (ukisai/Swift-1.5-Qwen3.8-Flash-Next-GSQ-RCO-GGUF, IQ3_S). The program's --file-offset and chunk layout above are the ones used for the reported reproductions. Whether a generic file of equivalent size also reproduces has not been tested.
Actual behavior — fresh boot, second run
CONFIG mode=direct sync=each pattern=random seed=20261025 iterations=250 devices=2
mapping=model-file-backed map_scope=whole file_offset=0x7c2331d40
map_bytes=38842920384 source_span=1073741824 src=0x7f40beb31d40 chunks=59
READY dev=0 name=AMD Radeon RX 7900 XTX dst=0x7f366a600000
READY dev=1 name=AMD Radeon RX 7800 XT dst=0x7f367dc00000
ROUND PASS dev=0 iteration=1/250 mode=direct sync=each
ROUND PASS dev=1 iteration=1/250 mode=direct sync=each
Memory access fault by GPU node-1 (Agent handle: ...) on address 0x7f40f65d6000.
Reason: Page not present or supervisor privilege.
[process terminated with SIGABRT]
0x7f40f65d6000 is 933,905,088 bytes above the src=0x7f40beb31d40 file-backed source base, within the selected 1 GiB source span.
Kernel journal from the same fresh-boot crash (redacted)
amdgpu 0000:03:00.0: [gfxhub] page fault (src_id:0 ring:24 vmid:8 pasid:62)
amdgpu 0000:03:00.0: Process rocm_mmap_sdma_ pid 4954
amdgpu 0000:03:00.0: in page starting at address 0x00007f40f65d6000 from client 10
amdgpu 0000:03:00.0: GCVM_L2_PROTECTION_FAULT_STATUS:0x00801A31
amdgpu 0000:03:00.0: Faulty UTCL2 client ID: SDMA0 (0xd)
amdgpu 0000:03:00.0: MORE_FAULTS: 0x1
amdgpu 0000:03:00.0: WALKER_ERROR: 0x0
amdgpu 0000:03:00.0: PERMISSION_FAULTS: 0x3
amdgpu 0000:03:00.0: MAPPING_ERROR: 0x0
amdgpu 0000:03:00.0: RW: 0x0
An earlier-boot fully instrumented run also faulted. Its last activity showed a fault address within the source range of a previous copy for which COPY SYNC END had already been printed. This observation does not prove that specific completed copy caused the fault; asynchronous fault reporting is possible. The earlier-boot detailed trace and fresh-boot kernel log are not from the same run.
Control experiments
| Test | Outcome |
|---|---|
| Direct source from file-backed mmap | Intermittent GPU fault, including immediately after fresh boot |
Standalone --mode malloc (CPU-copy source bytes to ordinary malloc before HIP transfer) |
Prior run completed 250 iterations |
| malloc staging in instrumented llama.cpp | Prior runs completed |
| malloc staging but pass original mmap pointer to HIP | Fault reproduced |
| CPU pre-touch of original file mapping | Did not prevent llama.cpp crash |
| Disable context checkpoints or force ggml graph re-reservation | Did not prevent llama.cpp crash; not needed for standalone repro |
HSA_ENABLE_SDMA=0, HSA_ENABLE_PEER_SDMA=0 |
Did not prevent llama.cpp crash; not tested in standalone |
Prior closely related issue
ROCm/rocm-systems#4817 reported a very similar hipMemcpyAsync fault and was closed after the KFD fix "drm/amdkfd: Use exclusive bounds for SVM split alignment checks". However, this standalone reproducer faults on Linux 7.3 with the in-tree driver. This may be a different failure or a remaining uncovered edge case. We have not established identical root cause.
Expected behavior
H2D copies from a valid file-backed mmap should complete normally or return a documented HIP error — not cause a GPU VM page fault and SIGABRT.
Request
Could KFD/SVM and HIP copy-path maintainers investigate the lifetime/permissions of mmap-backed host pages supplied to hipMemcpyAsync? An independently reproducible HIP-only example and a clean-boot AMDGPU SDMA0 kernel fault are available. More focused hardware tests can be run if needed.
Attachments: Privacy-minimized logs with a fresh-boot crash, matched kernel faults, an earlier failing detailed trace, and a successful-run summary. rocm-bugreport-20261025-sanitized.zip
Operating System
CachyOS Linux (Arch)
CPU
AMD Ryzen 9 7950X 16-Core Processor
GPU
AMD Radeon RX 7900 XTX + AMD Radeon RX 7800 XT
ROCm Version
7.2.4-1.1 + in-kernel amdgpu driver (Linux 7.3.0-5-cachyos)
Installation Method
Package manager
Installed ROCm Packages / Versions
Installed ROCm packages / versions
amdgpu_top 0.11.5-1.1
comgr 2:7.2.4-2.1
hip-runtime-amd 7.2.4-1.1
hipblas 7.2.4-1.1
hipblas-common 7.2.4-1
hipblaslt 7.2.4-1
hipcub 7.2.4-1.1
hipfft 7.2.4-1.1
hiprand 7.2.4-1.1
hipsolver 7.2.4-1.1
hipsparse 7.2.4-1.1
hsa-rocr 7.2.4-1.1
linux-firmware-amdgpu 1:20261019-1
miopen-hip 7.2.4-1
rocm-cmake 7.2.4-1
rocm-core 7.2.4-1.1
rocm-device-libs 2:7.2.4-2.1
rocm-hip-libraries 7.2.4-1
rocm-hip-runtime 7.2.4-1
rocm-hip-sdk 7.2.4-1
rocm-language-runtime 7.2.4-1
rocm-llvm 2:7.2.4-2.1
rocm-opencl-runtime 7.2.4-1
rocm-smi-lib 7.2.0-2.1
rocminfo 7.2.4-1.1
xf86-video-amdgpu 25.0.0-1.1
ROCm Component
clr
Steps to Reproduce
-
Use a Linux system with two AMD GPUs (tested with RX 7900 XTX + RX 7800 XT), ROCm 7.2.4 and the in-tree amdgpu kernel driver.
-
Clone the repository containing the standalone HIP reproducer:
git clone -b diagnostics/rocm-checkpoint-fault https://github.com/inf9144/llama.cpp.git -
Enter
llama.cpp/tools/rocm-mmap-sdma-reproand compile:/opt/rocm/bin/hipcc -O2 -std=c++17 -pthread rocm_mmap_sdma_repro.cpp -o rocm_mmap_sdma_repro -
Download the IQ3_S GGUF model shard
Swift-Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00001-of-00002.gguffromukisai/Swift-1.5-Qwen3.8-Flash-Next-GSQ-RCO-GGUFon Hugging Face. -
Run:
./rocm_mmap_sdma_repro \ --model-file /path/to/model-shard.gguf \ --file-offset 0x7c2331d40 \ --map whole \ --source-span-mib 1024 \ --pattern random \ --seed 20261025 \ --devices 0,1 \ --iterations 250 \ --mode direct \ --sync each -
Repeat the identical command if the first run succeeds. The issue is intermittent. After a clean reboot, with no prior llama.cpp execution, the first invocation passed all 250 iterations while the second invocation crashed after the first completed iteration on each GPU.
Observed failure:
Memory access fault by GPU node-1. Reason: Page not present or supervisor privilege.
The process terminates with SIGABRT.
The kernel log identifies SDMA0 as the faulting client, with the fault address inside the valid file-backed mmap region.
Control test: Running the same reproducer with --mode malloc (copying the mmap source bytes into anonymous malloc memory before passing them to HIP) successfully completed 250 iterations in our earlier tests.
The reproducer has no llama.cpp or ggml runtime dependencies. It uses HIP directly.
(Optional for Linux users) Output of rocminfo --support
rocminfo --support output
ROCk module is loaded
=====================
HSA System Attributes
=====================
Runtime Version: 1.18
Runtime Ext Version: 1.15
System Timestamp Freq.: 1000.000000MHz
Sig. Max Wait Duration: 18446744073709551615 (0xFFFFFFFFFFFFFFFF) (timestamp count)
Machine Model: LARGE
System Endianness: LITTLE
Mwaitx: DISABLED
XNACK enabled: NO
DMAbuf Support: YES
VMM Support: YES
==========
HSA Agents
==========
*******
Agent 1
*******
Name: AMD Ryzen 9 7950X 16-Core Processor
Uuid: CPU-XX
Marketing Name: AMD Ryzen 9 7950X 16-Core Processor
Vendor Name: CPU
Feature: None specified
Profile: FULL_PROFILE
Float Round Mode: NEAR
Max Queue Number: 0(0x0)
Queue Min Size: 0(0x0)
Queue Max Size: 0(0x0)
Queue Type: MULTI
Node: 0
Device Type: CPU
Cache Info:
L1: 32768(0x8000) KB
Chip ID: 0(0x0)
ASIC Revision: 0(0x0)
Cacheline Size: 64(0x40)
Max Clock Freq. (MHz): 5883
BDFID: 0
Internal Node ID: 0
Compute Unit: 32
SIMDs per CU: 0
Shader Engines: 0
Shader Arrs. per Eng.: 0
WatchPts on Addr. Ranges:1
Memory Properties:
Features: None
Pool Info:
Pool 1
Segment: GLOBAL; FLAGS: FINE GRAINED
Size: 64901976(0x3de5358) KB
Allocatable: TRUE
Alloc Granule: 4KB
Alloc Recommended Granule:4KB
Alloc Alignment: 4KB
Accessible by all: TRUE
Pool 2
Segment: GLOBAL; FLAGS: EXTENDED FINE GRAINED
Size: 64901976(0x3de5358) KB
Allocatable: TRUE
Alloc Granule: 4KB
Alloc Recommended Granule:4KB
Alloc Alignment: 4KB
Accessible by all: TRUE
Pool 3
Segment: GLOBAL; FLAGS: KERNARG, FINE GRAINED
Size: 64901976(0x3de5358) KB
Allocatable: TRUE
Alloc Granule: 4KB
Alloc Recommended Granule:4KB
Alloc Alignment: 4KB
Accessible by all: TRUE
Pool 4
Segment: GLOBAL; FLAGS: COARSE GRAINED
Size: 64901976(0x3de5358) KB
Allocatable: TRUE
Alloc Granule: 4KB
Alloc Recommended Granule:4KB
Alloc Alignment: 4KB
Accessible by all: TRUE
ISA Info:
*******
Agent 2
*******
Name: gfx1100
Uuid: GPU-cba54b6c77092a25
Marketing Name: AMD Radeon RX 7900 XTX
Vendor Name: AMD
Feature: KERNEL_DISPATCH
Profile: BASE_PROFILE
Float Round Mode: NEAR
Max Queue Number: 128(0x80)
Queue Min Size: 64(0x40)
Queue Max Size: 131072(0x20000)
Queue Type: MULTI
Node: 1
Device Type: GPU
Cache Info:
L1: 32(0x20) KB
L2: 6144(0x1800) KB
L3: 98304(0x18000) KB
Chip ID: 29772(0x744c)
ASIC Revision: 0(0x0)
Cacheline Size: 128(0x80)
Max Clock Freq. (MHz): 2431
BDFID: 768
Internal Node ID: 1
Compute Unit: 96
SIMDs per CU: 2
Shader Engines: 6
Shader Arrs. per Eng.: 2
WatchPts on Addr. Ranges:4
Coherent Host Access: FALSE
Memory Properties:
Features: KERNEL_DISPATCH
Fast F16 Operation: TRUE
Wavefront Size: 32(0x20)
Workgroup Max Size: 1024(0x400)
Workgroup Max Size per Dimension:
x 1024(0x400)
y 1024(0x400)
z 1024(0x400)
Max Waves Per CU: 32(0x20)
Max Work-item Per CU: 1024(0x400)
Grid Max Size: 4294967295(0xffffffff)
Grid Max Size per Dimension:
x 2147483647(0x7fffffff)
y 65535(0xffff)
z 65535(0xffff)
Max fbarriers/Workgrp: 32
Packet Processor uCode:: 632
SDMA engine uCode:: 27
IOMMU Support:: None
Pool Info:
Pool 1
Segment: GLOBAL; FLAGS: COARSE GRAINED
Size: 25149440(0x17fc000) KB
Allocatable: TRUE
Alloc Granule: 4KB
Alloc Recommended Granule:2048KB
Alloc Alignment: 4KB
Accessible by all: FALSE
Pool 2
Segment: GLOBAL; FLAGS: EXTENDED FINE GRAINED
Size: 25149440(0x17fc000) KB
Allocatable: TRUE
Alloc Granule: 4KB
Alloc Recommended Granule:2048KB
Alloc Alignment: 4KB
Accessible by all: FALSE
Pool 3
Segment: GROUP
Size: 64(0x40) KB
Allocatable: FALSE
Alloc Granule: 0KB
Alloc Recommended Granule:0KB
Alloc Alignment: 0KB
Accessible by all: FALSE
ISA Info:
ISA 1
Name: amdgcn-amd-amdhsa--gfx1100
Machine Models: HSA_MACHINE_MODEL_LARGE
Profiles: HSA_PROFILE_BASE
Default Rounding Mode: NEAR
Default Rounding Mode: NEAR
Fast f16: TRUE
Workgroup Max Size: 1024(0x400)
Workgroup Max Size per Dimension:
x 1024(0x400)
y 1024(0x400)
z 1024(0x400)
Grid Max Size: 4294967295(0xffffffff)
Grid Max Size per Dimension:
x 2147483647(0x7fffffff)
y 65535(0xffff)
z 65535(0xffff)
FBarrier Max Size: 32
ISA 2
Name: amdgcn-amd-amdhsa--gfx11-generic
Machine Models: HSA_MACHINE_MODEL_LARGE
Profiles: HSA_PROFILE_BASE
Default Rounding Mode: NEAR
Default Rounding Mode: NEAR
Fast f16: TRUE
Workgroup Max Size: 1024(0x400)
Workgroup Max Size per Dimension:
x 1024(0x400)
y 1024(0x400)
z 1024(0x400)
Grid Max Size: 4294967295(0xffffffff)
Grid Max Size per Dimension:
x 2147483647(0x7fffffff)
y 65535(0xffff)
z 65535(0xffff)
FBarrier Max Size: 32
*******
Agent 3
*******
Name: gfx1101
Uuid: GPU-043328b1d7050147
Marketing Name: AMD Radeon RX 7800 XT
Vendor Name: AMD
Feature: KERNEL_DISPATCH
Profile: BASE_PROFILE
Float Round Mode: NEAR
Max Queue Number: 128(0x80)
Queue Min Size: 64(0x40)
Queue Max Size: 131072(0x20000)
Queue Type: MULTI
Node: 2
Device Type: GPU
Cache Info:
L1: 32(0x20) KB
L2: 4096(0x1000) KB
L3: 65536(0x10000) KB
Chip ID: 29822(0x747e)
ASIC Revision: 0(0x0)
Cacheline Size: 128(0x80)
Max Clock Freq. (MHz): 2254
BDFID: 5120
Internal Node ID: 2
Compute Unit: 60
SIMDs per CU: 2
Shader Engines: 3
Shader Arrs. per Eng.: 2
WatchPts on Addr. Ranges:4
Coherent Host Access: FALSE
Memory Properties:
Features: KERNEL_DISPATCH
Fast F16 Operation: TRUE
Wavefront Size: 32(0x20)
Workgroup Max Size: 1024(0x400)
Workgroup Max Size per Dimension:
x 1024(0x400)
y 1024(0x400)
z 1024(0x400)
Max Waves Per CU: 32(0x20)
Max Work-item Per CU: 1024(0x400)
Grid Max Size: 4294967295(0xffffffff)
Grid Max Size per Dimension:
x 2147483647(0x7fffffff)
y 65535(0xffff)
z 65535(0xffff)
Max fbarriers/Workgrp: 32
Packet Processor uCode:: 742
SDMA engine uCode:: 29
IOMMU Support:: None
Pool Info:
Pool 1
Segment: GLOBAL; FLAGS: COARSE GRAINED
Size: 16760832(0xffc000) KB
Allocatable: TRUE
Alloc Granule: 4KB
Alloc Recommended Granule:2048KB
Alloc Alignment: 4KB
Accessible by all: FALSE
Pool 2
Segment: GLOBAL; FLAGS: EXTENDED FINE GRAINED
Size: 16760832(0xffc000) KB
Allocatable: TRUE
Alloc Granule: 4KB
Alloc Recommended Granule:2048KB
Alloc Alignment: 4KB
Accessible by all: FALSE
Pool 3
Segment: GROUP
Size: 64(0x40) KB
Allocatable: FALSE
Alloc Granule: 0KB
Alloc Recommended Granule:0KB
Alloc Alignment: 0KB
Accessible by all: FALSE
ISA Info:
ISA 1
Name: amdgcn-amd-amdhsa--gfx1101
Machine Models: HSA_MACHINE_MODEL_LARGE
Profiles: HSA_PROFILE_BASE
Default Rounding Mode: NEAR
Default Rounding Mode: NEAR
Fast f16: TRUE
Workgroup Max Size: 1024(0x400)
Workgroup Max Size per Dimension:
x 1024(0x400)
y 1024(0x400)
z 1024(0x400)
Grid Max Size: 4294967295(0xffffffff)
Grid Max Size per Dimension:
x 2147483647(0x7fffffff)
y 65535(0xffff)
z 65535(0xffff)
FBarrier Max Size: 32
ISA 2
Name: amdgcn-amd-amdhsa--gfx11-generic
Machine Models: HSA_MACHINE_MODEL_LARGE
Profiles: HSA_PROFILE_BASE
Default Rounding Mode: NEAR
Default Rounding Mode: NEAR
Fast f16: TRUE
Workgroup Max Size: 1024(0x400)
Workgroup Max Size per Dimension:
x 1024(0x400)
y 1024(0x400)
z 1024(0x400)
Grid Max Size: 4294967295(0xffffffff)
Grid Max Size per Dimension:
x 2147483647(0x7fffffff)
y 65535(0xffff)
z 65535(0xffff)
FBarrier Max Size: 32
*******
Agent 4
*******
Name: gfx1036
Uuid: GPU-XX
Marketing Name: AMD Ryzen 9 7950X 16-Core Processor
Vendor Name: AMD
Feature: KERNEL_DISPATCH
Profile: BASE_PROFILE
Float Round Mode: NEAR
Max Queue Number: 128(0x80)
Queue Min Size: 64(0x40)
Queue Max Size: 131072(0x20000)
Queue Type: MULTI
Node: 3
Device Type: GPU
Cache Info:
L1: 16(0x10) KB
L2: 256(0x100) KB
Chip ID: 5710(0x164e)
ASIC Revision: 1(0x1)
Cacheline Size: 128(0x80)
Max Clock Freq. (MHz): 2200
BDFID: 6400
Internal Node ID: 3
Compute Unit: 2
SIMDs per CU: 2
Shader Engines: 1
Shader Arrs. per Eng.: 1
WatchPts on Addr. Ranges:4
Coherent Host Access: FALSE
Memory Properties: APU
Features: KERNEL_DISPATCH
Fast F16 Operation: TRUE
Wavefront Size: 32(0x20)
Workgroup Max Size: 1024(0x400)
Workgroup Max Size per Dimension:
x 1024(0x400)
y 1024(0x400)
z 1024(0x400)
Max Waves Per CU: 32(0x20)
Max Work-item Per CU: 1024(0x400)
Grid Max Size: 4294967295(0xffffffff)
Grid Max Size per Dimension:
x 2147483647(0x7fffffff)
y 65535(0xffff)
z 65535(0xffff)
Max fbarriers/Workgrp: 32
Packet Processor uCode:: 28
SDMA engine uCode:: 9
IOMMU Support:: None
Pool Info:
Pool 1
Segment: GLOBAL; FLAGS: COARSE GRAINED
Size: 32450988(0x1ef29ac) KB
Allocatable: TRUE
Alloc Granule: 4KB
Alloc Recommended Granule:2048KB
Alloc Alignment: 4KB
Accessible by all: FALSE
Pool 2
Segment: GLOBAL; FLAGS: EXTENDED FINE GRAINED
Size: 32450988(0x1ef29ac) KB
Allocatable: TRUE
Alloc Granule: 4KB
Alloc Recommended Granule:2048KB
Alloc Alignment: 4KB
Accessible by all: FALSE
Pool 3
Segment: GROUP
Size: 64(0x40) KB
Allocatable: FALSE
Alloc Granule: 0KB
Alloc Recommended Granule:0KB
Alloc Alignment: 0KB
Accessible by all: FALSE
ISA Info:
ISA 1
Name: amdgcn-amd-amdhsa--gfx1036
Machine Models: HSA_MACHINE_MODEL_LARGE
Profiles: HSA_PROFILE_BASE
Default Rounding Mode: NEAR
Default Rounding Mode: NEAR
Fast f16: TRUE
Workgroup Max Size: 1024(0x400)
Workgroup Max Size per Dimension:
x 1024(0x400)
y 1024(0x400)
z 1024(0x400)
Grid Max Size: 4294967295(0xffffffff)
Grid Max Size per Dimension:
x 2147483647(0x7fffffff)
y 65535(0xffff)
z 65535(0xffff)
FBarrier Max Size: 32
ISA 2
Name: amdgcn-amd-amdhsa--gfx10-3-generic
Machine Models: HSA_MACHINE_MODEL_LARGE
Profiles: HSA_PROFILE_BASE
Default Rounding Mode: NEAR
Default Rounding Mode: NEAR
Fast f16: TRUE
Workgroup Max Size: 1024(0x400)
Workgroup Max Size per Dimension:
x 1024(0x400)
y 1024(0x400)
z 1024(0x400)
Grid Max Size: 4294967295(0xffffffff)
Grid Max Size per Dimension:
x 2147483647(0x7fffffff)
y 65535(0xffff)
z 65535(0xffff)
FBarrier Max Size: 32
*** Done ***
Additional Information
rocm-bugreport-20261025-sanitized.zip
Attached are sanitized diagnostic logs, including:
- A failing standalone HIP trace from an earlier boot.
- A successful trace showing the intermittent behavior.
- AMDGPU kernel logs from the fresh-boot failure, identifying SDMA0 as the faulting client.
- System and ROCm version information.
The failing trace and the fresh-boot kernel logs originate from different runs.
Potentially related to ROCm/rocm-systems#4817, which was closed following a KFD SVM mapping fix. This issue still reproduces with the in-tree amdgpu driver on Linux 7.3.0.
The original unredacted system logs have deliberately not been attached for privacy reasons.
- 主要言語
- C
- スター
- 470
- フォーク
- 150
- PR マージ指標
- 30日以内にマージされた PR はありません
環境構築
このプロジェクトには開発コンテナ、Dockerfile、コントリビューションガイドがありません。まず README を読み、一般的な手順ははじめてのコントリビューションガイドを参照してください。
はじめの一歩
- issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
- 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
- リポジトリをフォークし、ブランチを切って変更します。
- issue 番号を参照したプルリクエストを送ります。
ROCm/amdgpu のほかの issue
-
難易度 2/5 1〜3時間 初心者へのやさしさ 74/100
-
難易度 4/5 3〜5日 初心者へのやさしさ 25/100
-
難易度 5/5 1週間以上 初心者へのやさしさ 8/100
-
amdgpu unified RAS: GFX correctable errors aggregated by MP5 firmware are dropped on MI450 (no GFX error count, no CPER)対応中かも @tpatward-amd が 19 日前に担当しました。 オープン
難易度 4/5 3〜5日 初心者へのやさしさ 28/100
-
難易度 4/5 3〜5日 初心者へのやさしさ 48/100
似ている issue
-
vim.filetype.match() prefixes a URL buffer name with the cwd before passing it to filetype functionsオープン
難易度 2/5 1〜3時間 初心者へのやさしさ 73/100
メンテナーはふだん 1 日以内に返信
-
compilation errorオープン
難易度 2/5 1〜3時間 初心者へのやさしさ 68/100
COVESA/dlt-daemon#925 · リアクション 1 件 ·
-
stream_curl: retry_count never resets after a successful reconnect, so long sequential streams die after 5 drops対応中かも このイシューにリンクされたプルリクエストがオープン中、またはマージ済みです。 オープン
難易度 2/5 1〜3時間 初心者へのやさしさ 70/100
mpv-player/mpv#18728 ·
メンテナーはふだん 1 日以内に返信
-
followup Prio 📌
難易度 2/5 1〜3時間 初心者へのやさしさ 66/100
メンテナーはふだん 1 日以内に返信
-
難易度 2/5 1〜3時間 初心者へのやさしさ 72/100
cyrusimap/cyrus-imapd#6640 ·
メンテナーはふだん 2 日以内に返信