diff options
| author | Matthew Brost <matthew.brost@intel.com> | 2026-03-26 14:01:15 -0700 |
|---|---|---|
| committer | Rodrigo Vivi <rodrigo.vivi@intel.com> | 2026-03-30 08:52:20 -0400 |
| commit | 56b7432b7e8e6ae1b289cb405d16db4150ef193b (patch) | |
| tree | 508277c924fd5caff35afe57eee742f5f7bffff9 /drivers/gpu/drm | |
| parent | bce7cd6db2386e7a2b49ad3c09df0beabddfa3c4 (diff) | |
drm/xe: Avoid memory allocations in xe_device_declare_wedged()
xe_device_declare_wedged() runs in the DMA-fence signaling path, where
GFP_KERNEL memory allocations are not allowed. However, registering
xe_device_wedged_fini via drmm_add_action_or_reset() triggers a
GFP_KERNEL allocation.
Fix this by deferring the registration of xe_device_wedged_fini until
late in the driver load sequence. Additionally, drop the wedged PM
reference only if the device is actually wedged in
xe_device_wedged_fini.
Fixes: 452bca0edbd0 ("drm/xe: Don't suspend device upon wedge")
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patch.msgid.link/20260326210116.202585-2-matthew.brost@intel.com
(cherry picked from commit b08ceb443866808b881b12d4183008d214d816c1)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Diffstat (limited to 'drivers/gpu/drm')
| -rw-r--r-- | drivers/gpu/drm/xe/xe_device.c | 27 |
1 files changed, 13 insertions, 14 deletions
diff --git a/drivers/gpu/drm/xe/xe_device.c b/drivers/gpu/drm/xe/xe_device.c index 52ee24e9cd3a..3eb06b27db7e 100644 --- a/drivers/gpu/drm/xe/xe_device.c +++ b/drivers/gpu/drm/xe/xe_device.c @@ -837,6 +837,14 @@ static void detect_preproduction_hw(struct xe_device *xe) } } +static void xe_device_wedged_fini(struct drm_device *drm, void *arg) +{ + struct xe_device *xe = arg; + + if (atomic_read(&xe->wedged.flag)) + xe_pm_runtime_put(xe); +} + int xe_device_probe(struct xe_device *xe) { struct xe_tile *tile; @@ -1013,6 +1021,10 @@ int xe_device_probe(struct xe_device *xe) detect_preproduction_hw(xe); + err = drmm_add_action_or_reset(&xe->drm, xe_device_wedged_fini, xe); + if (err) + goto err_unregister_display; + return devm_add_action_or_reset(xe->drm.dev, xe_device_sanitize, xe); err_unregister_display: @@ -1216,13 +1228,6 @@ u64 xe_device_uncanonicalize_addr(struct xe_device *xe, u64 address) return address & GENMASK_ULL(xe->info.va_bits - 1, 0); } -static void xe_device_wedged_fini(struct drm_device *drm, void *arg) -{ - struct xe_device *xe = arg; - - xe_pm_runtime_put(xe); -} - /** * DOC: Xe Device Wedging * @@ -1300,15 +1305,9 @@ void xe_device_declare_wedged(struct xe_device *xe) return; } - xe_pm_runtime_get_noresume(xe); - - if (drmm_add_action_or_reset(&xe->drm, xe_device_wedged_fini, xe)) { - drm_err(&xe->drm, "Failed to register xe_device_wedged_fini clean-up. Although device is wedged.\n"); - return; - } - if (!atomic_xchg(&xe->wedged.flag, 1)) { xe->needs_flr_on_fini = true; + xe_pm_runtime_get_noresume(xe); drm_err(&xe->drm, "CRITICAL: Xe has declared device %s as wedged.\n" "IOCTLs and executions are blocked. Only a rebind may clear the failure\n" |