From: Jan Maslak <jan.maslak@intel.com>
To: intel-xe@lists.freedesktop.org
Cc: matthew.auld@intel.com, matthew.brost@intel.com,
thomas.hellstrom@linux.intel.com, rodrigo.vivi@intel.com,
Jan Maslak <jan.maslak@intel.com>
Subject: [PATCH v3 0/4] drm/xe/xe_migrate: Fix memory corruption with unaligned dst in MEM_COPY
Date: Wed, 9 Sep 2026 08:18:41 +0200 [thread overview]
Message-ID: <20260909061845.4048047-1-jan.maslak@intel.com> (raw)
MEM_COPY_CMD in MATRIX_COPY mode produces incorrect results when the
destination address is not aligned to the pitch. This manifests as
memory corruption during xe_migrate_access_memory() reads from VRAM.
The probable cause appears to be a hardware requirement that the
destination address must be aligned to the pitch when using MATRIX_COPY
mode.
This was observed on BMG during xe_migrate_access_memory() calls where
the destination offset within a page was not aligned to the computed
pitch value.
Patch 1 adds the minimal workaround by selecting a pitch that is aligned
to both the copy length and the destination address.
Patch 2 fixes an independent page tracking bug in
xe_migrate_access_memory().
Patch 3 adds a follow-up optimization to reduce the number of linear
copy jobs for unaligned destinations.
Patch 4 adds a KUnit live test that exercises xe_migrate_access_memory()
with various unaligned destination offsets to prevent regressions.
Tested on BMG (Battlemage).
v2:
- Split the original xe_migrate change into a minimal workaround,
a standalone page tracking fix, and a follow-up optimization
per Matthew Auld's review feedback
v3:
- Used a helper function and added an additional comment in patch 3
per Matthew Auld's review feedback
Christoph Manszewski (2):
drm/xe/xe_migrate: Align MEM_COPY pitch with destination address
drm/xe/xe_migrate: Fix page tracking in access_memory
Jan Maslak (2):
drm/xe/xe_migrate: Optimize unaligned access_memory copies
drm/xe/tests: Add xe_migrate_access_memory subtest
drivers/gpu/drm/xe/tests/xe_migrate.c | 278 ++++++++++++++++++++++++++
drivers/gpu/drm/xe/xe_migrate.c | 176 ++++++++++++++--
2 files changed, 435 insertions(+), 19 deletions(-)
--
2.43.0
next reply other threads:[~2026-09-09 6:19 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-09-09 6:18 Jan Maslak [this message]
2026-09-09 6:18 ` [PATCH v3 1/4] drm/xe/xe_migrate: Align MEM_COPY pitch with destination address Jan Maslak
2026-09-09 6:32 ` sashiko-bot
2026-09-09 6:18 ` [PATCH v3 2/4] drm/xe/xe_migrate: Fix page tracking in access_memory Jan Maslak
2026-09-09 6:30 ` sashiko-bot
2026-09-09 6:18 ` [PATCH v3 3/4] drm/xe/xe_migrate: Optimize unaligned access_memory copies Jan Maslak
2026-09-09 6:31 ` sashiko-bot
2026-09-09 6:18 ` [PATCH v3 4/4] drm/xe/tests: Add xe_migrate_access_memory subtest Jan Maslak
2026-09-09 6:28 ` sashiko-bot
2026-09-09 6:25 ` ✗ CI.checkpatch: warning for drm/xe/xe_migrate: Fix memory corruption with unaligned dst in MEM_COPY (rev3) Patchwork
2026-09-09 6:27 ` ✓ CI.KUnit: success " Patchwork
2026-09-09 7:04 ` ✗ Xe.CI.BAT: failure " Patchwork
2026-09-09 11:36 ` ✗ Xe.CI.FULL: " Patchwork
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20260909061845.4048047-1-jan.maslak@intel.com \
--to=jan.maslak@intel.com \
--cc=intel-xe@lists.freedesktop.org \
--cc=matthew.auld@intel.com \
--cc=matthew.brost@intel.com \
--cc=rodrigo.vivi@intel.com \
--cc=thomas.hellstrom@linux.intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox