tmakin/ack-tegra
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

drm/xe: Fix early wedge on GuC load failure

Browse Source 
commit a39d082c3553d35b4fe5585e1e2fb221c130cae8 upstream.

When the GuC fails to load we declare the device wedged. However, the
very first GuC load attempt on GT0 (from xe_gt_init_hwconfig) is done
before the GT1 GuC objects are initialized, so things go bad when the
wedge code attempts to cleanup GT1. To fix this, check the initialization
status in the functions called during wedge.

Fixes: 7dbe8af13c ("drm/xe: Wedge the entire device")
Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Jonathan Cavitt <jonathan.cavitt@intel.com>
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Cc: Zhanjun Dong <zhanjun.dong@intel.com>
Cc: stable@vger.kernel.org # v6.12+: 1e1981b16bb1: drm/xe: Fix taking invalid lock on wedge
Cc: stable@vger.kernel.org # v6.12+
Reviewed-by: Jonathan Cavitt <jonathan.cavitt@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://lore.kernel.org/r/20250611214453.1159846-2-daniele.ceraolospurio@intel.com
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
(cherry picked from commit 0b93b7dcd9eb888a6ac7546560877705d4ad61bf)
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This commit is contained in:
Daniele Ceraolo Spurio
2025-06-11 14:44:54 -07:00
committed by Greg Kroah-Hartman
parent a6d81b2d70
commit e5e1996780
4 changed files with 21 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
drivers/gpu/drm/xe/xe_gt_tlb_invalidation.c
View File

@@ -137,6 +137,14 @@ void xe_gt_tlb_invalidation_reset(struct xe_gt *gt)
struct xe_gt_tlb_invalidation_fence *fence, *next;
int pending_seqno;
/*
* we can get here before the CTs are even initialized if we're wedging
* very early, in which case there are not going to be any pending
* fences so we can bail immediately.
*/
if (!xe_guc_ct_initialized(&gt->uc.guc.ct))
return;
/*
* CT channel is already disabled at this point. No new TLB requests can
* appear.

7
drivers/gpu/drm/xe/xe_guc_ct.c
View File

@@ -454,6 +454,9 @@ void xe_guc_ct_disable(struct xe_guc_ct *ct)
*/
void xe_guc_ct_stop(struct xe_guc_ct *ct)
{
if (!xe_guc_ct_initialized(ct))
return;
xe_guc_ct_set_state(ct, XE_GUC_CT_STATE_STOPPED);
stop_g2h_handler(ct);
}
@@ -638,7 +641,7 @@ static int __guc_ct_send_locked(struct xe_guc_ct *ct, const u32 *action,
u16 seqno;
int ret;
xe_gt_assert(gt, ct->state != XE_GUC_CT_STATE_NOT_INITIALIZED);
xe_gt_assert(gt, xe_guc_ct_initialized(ct));
xe_gt_assert(gt, !g2h_len || !g2h_fence);
xe_gt_assert(gt, !num_g2h || !g2h_fence);
xe_gt_assert(gt, !g2h_len || num_g2h);
@@ -1209,7 +1212,7 @@ static int g2h_read(struct xe_guc_ct *ct, u32 *msg, bool fast_path)
u32 action;
u32 *hxg;
xe_gt_assert(gt, ct->state != XE_GUC_CT_STATE_NOT_INITIALIZED);
xe_gt_assert(gt, xe_guc_ct_initialized(ct));
lockdep_assert_held(&ct->fast_lock);
if (ct->state == XE_GUC_CT_STATE_DISABLED)

5
drivers/gpu/drm/xe/xe_guc_ct.h
View File

@@ -23,6 +23,11 @@ void xe_guc_ct_snapshot_print(struct xe_guc_ct_snapshot *snapshot,
void xe_guc_ct_snapshot_free(struct xe_guc_ct_snapshot *snapshot);
void xe_guc_ct_print(struct xe_guc_ct *ct, struct drm_printer *p, bool atomic);
static inline bool xe_guc_ct_initialized(struct xe_guc_ct *ct)
{
return ct->state != XE_GUC_CT_STATE_NOT_INITIALIZED;
}
static inline bool xe_guc_ct_enabled(struct xe_guc_ct *ct)
{
return ct->state == XE_GUC_CT_STATE_ENABLED;

3
drivers/gpu/drm/xe/xe_guc_submit.c
View File

@@ -1722,6 +1722,9 @@ int xe_guc_submit_reset_prepare(struct xe_guc *guc)
{
int ret;
if (!guc->submission_state.initialized)
return 0;
/*
* Using an atomic here rather than submission_state.lock as this
* function can be called while holding the CT lock (engine reset