Intel-XE Archive on lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH V2 0/3] drm/xe/vram: decay-based reinstatement of soft-offlined VRAM pages
@ 2026-11-12 12:43 Tejas Upadhyay
  2026-11-12 12:43 ` [PATCH V2 1/3] drm/xe/vram: add " Tejas Upadhyay
                   ` (4 more replies)
  0 siblings, 5 replies; 9+ messages in thread
From: Tejas Upadhyay @ 2026-11-12 12:43 UTC (permalink / raw)
  To: intel-xe; +Cc: himal.prasad.ghimiray, matthew.brost, Tejas Upadhyay

The existing bad-page flow soft-offlines a VRAM page on the first HW
report and promotes it to a permanent (hard) retirement on the second
report at the same address. Once soft-offlined, a page stays reserved
forever, which over-punishes a single transient DBE caused by a
one-off environmental disturbance -- the page remains unusable until
the next reboot/probe even if it never faults again.

This series adds a per-VRAM-manager decay timer so a SOFT-offlined
page is automatically reinstated after a configurable window (default
24h) of no further reports at that address, configurable via configfs
(bad_page_decay_secs, 0 disables). HARD-offlined pages -- i.e. pages
that have faulted a second time, or that are reloaded from firmware's
non-volatile permanent list at probe -- are never eligible for decay
and remain permanently retired, as before.

Patch 1 introduces the core feature. Patches 2 and 3 are fixes for two
issues found during internal review before this series was ever
posted:

  - Patch 2 : Pages reloaded from firmware's permanent
    list at boot were being created as SOFT instead of HARD, making
    already-confirmed-bad memory silently eligible for reinstatement
    once the decay window elapsed.

  - Patch 3 : The decay GC worker rescheduled itself
    unconditionally, so once started it never went idle even after
    every tracked page had decayed away or been promoted to HARD.

IGT coverage:
https://patchwork.freedesktop.org/patch/765064/

Tejas Upadhyay (3):
  drm/xe/vram: add decay-based reinstatement of soft-offlined VRAM pages
  drm/xe/vram: don't let firmware-confirmed permanent bad pages decay
    back in
  drm/xe/vram: stop decay worker from rescheduling itself forever

 drivers/gpu/drm/xe/xe_configfs.c           | 55 +++++++++++++++
 drivers/gpu/drm/xe/xe_configfs.h           |  1 +
 drivers/gpu/drm/xe/xe_debugfs.c            |  2 +-
 drivers/gpu/drm/xe/xe_ras.c                |  7 +-
 drivers/gpu/drm/xe/xe_ttm_vram_mgr.c       | 82 ++++++++++++++++++++--
 drivers/gpu/drm/xe/xe_ttm_vram_mgr.h       |  2 +-
 drivers/gpu/drm/xe/xe_ttm_vram_mgr_types.h | 16 +++++
 7 files changed, 157 insertions(+), 8 deletions(-)

-- 
2.52.0


^ permalink raw reply	[flat|nested] 9+ messages in thread

end of thread, other threads:[~2026-11-12 14:21 UTC | newest]

Thread overview: 9+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-12 12:43 [PATCH V2 0/3] drm/xe/vram: decay-based reinstatement of soft-offlined VRAM pages Tejas Upadhyay
2026-11-12 12:43 ` [PATCH V2 1/3] drm/xe/vram: add " Tejas Upadhyay
2026-11-12 13:03   ` sashiko-bot
2026-11-12 12:43 ` [PATCH V2 2/3] drm/xe/vram: don't let firmware-confirmed permanent bad pages decay back in Tejas Upadhyay
2026-11-12 12:54   ` sashiko-bot
2026-11-12 12:43 ` [PATCH V2 3/3] drm/xe/vram: stop decay worker from rescheduling itself forever Tejas Upadhyay
2026-11-12 12:55   ` sashiko-bot
2026-11-12 13:36 ` ✓ CI.KUnit: success for drm/xe/vram: decay-based reinstatement of soft-offlined VRAM pages Patchwork
2026-11-12 14:21 ` ✓ Xe.CI.BAT: " Patchwork

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox