From: Matthew Brost <matthew.brost@intel.com>
To: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: <intel-xe@lists.freedesktop.org>
Subject: Re: [PATCH v7 11/32] drm/xe: Move GGTT lock init to alloc
Date: Tue, 7 Oct 2025 05:53:08 -0700 [thread overview]
Message-ID: <aOUNNFAx2in/HIwo@lstrano-desk.jf.intel.com> (raw)
In-Reply-To: <f5d2328f-8069-406b-bda8-1d98803fa568@intel.com>
On Tue, Oct 07, 2025 at 01:47:27PM +0200, Michal Wajdeczko wrote:
>
>
> On 10/7/2025 1:26 PM, Matthew Brost wrote:
> > The GGTT lock is needed very early during GT initialization for a VF;
> > move the GGTT lock initialization to the allocation phase.
> >
> > Signed-off-by: Matthew Brost <matthew.brost@intel.com>
> > ---
> > drivers/gpu/drm/xe/xe_ggtt.c | 36 ++++++++++++++++++++++--------------
> > 1 file changed, 22 insertions(+), 14 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/xe/xe_ggtt.c b/drivers/gpu/drm/xe/xe_ggtt.c
> > index 7fdd0a97a628..f7717057cd14 100644
> > --- a/drivers/gpu/drm/xe/xe_ggtt.c
> > +++ b/drivers/gpu/drm/xe/xe_ggtt.c
> > @@ -159,6 +159,16 @@ static void xe_ggtt_clear(struct xe_ggtt *ggtt, u64 start, u64 size)
> > }
> > }
> >
> > +static void primelockdep(struct xe_ggtt *ggtt)
> > +{
> > + if (!IS_ENABLED(CONFIG_LOCKDEP))
> > + return;
> > +
> > + fs_reclaim_acquire(GFP_KERNEL);
> > + might_lock(&ggtt->lock);
> > + fs_reclaim_release(GFP_KERNEL);
> > +}
> > +
> > /**
> > * xe_ggtt_alloc - Allocate a GGTT for a given &xe_tile
> > * @tile: &xe_tile
> > @@ -170,8 +180,19 @@ static void xe_ggtt_clear(struct xe_ggtt *ggtt, u64 start, u64 size)
> > struct xe_ggtt *xe_ggtt_alloc(struct xe_tile *tile)
> > {
> > struct xe_ggtt *ggtt = drmm_kzalloc(&tile_to_xe(tile)->drm, sizeof(*ggtt), GFP_KERNEL);
> > - if (ggtt)
> > +
> > + if (ggtt) {
> > + int err;
> > +
> > ggtt->tile = tile;
> > +
> > + err = drmm_mutex_init(&tile_to_xe(tile)->drm, &ggtt->lock);
> > + if (err)
> > + ggtt = NULL;
> > + else
> > + primelockdep(ggtt);
> > + }
> > +
>
> all this looks over-complicated and against the rule to exit early on error, so maybe:
>
> struct xe_device *xe = tile_to_xe(tile);
> struct xe_ggtt *ggtt;
>
> ggtt = drmm_kzalloc(&xe->drm, sizeof(*ggtt), GFP_KERNEL);
> if (!ggtt)
> return NULL;
> if (drmm_mutex_init(&xe->drm, &ggtt->lock))
> return NULL;
> if (IS_ENABLED(CONFIG_LOCKDEP))
> primelockdep(ggtt);
>
> ggtt->tile = tile;
>
Sure, I have CI issue so I need to respin this series now.
Matt
> > return ggtt;
> > }
> >
> > @@ -180,7 +201,6 @@ static void ggtt_fini_early(struct drm_device *drm, void *arg)
> > struct xe_ggtt *ggtt = arg;
> >
> > destroy_workqueue(ggtt->wq);
> > - mutex_destroy(&ggtt->lock);
> > drm_mm_takedown(&ggtt->mm);
> > }
> >
> > @@ -198,16 +218,6 @@ void xe_ggtt_might_lock(struct xe_ggtt *ggtt)
> > }
> > #endif
> >
> > -static void primelockdep(struct xe_ggtt *ggtt)
> > -{
> > - if (!IS_ENABLED(CONFIG_LOCKDEP))
> > - return;
> > -
> > - fs_reclaim_acquire(GFP_KERNEL);
> > - might_lock(&ggtt->lock);
> > - fs_reclaim_release(GFP_KERNEL);
> > -}
> > -
> > static const struct xe_ggtt_pt_ops xelp_pt_ops = {
> > .pte_encode_flags = xelp_ggtt_pte_flags,
> > .ggtt_set_pte = xe_ggtt_set_pte,
> > @@ -227,8 +237,6 @@ static void __xe_ggtt_init_early(struct xe_ggtt *ggtt, u32 reserved)
> > {
> > drm_mm_init(&ggtt->mm, reserved,
> > ggtt->size - reserved);
> > - mutex_init(&ggtt->lock);
> > - primelockdep(ggtt);
> > }
> >
> > int xe_ggtt_init_kunit(struct xe_ggtt *ggtt, u32 reserved, u32 size)
>
next prev parent reply other threads:[~2025-10-07 12:53 UTC|newest]
Thread overview: 42+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-10-07 11:26 [PATCH v7 00/32] VF migration redesign Matthew Brost
2025-10-07 11:26 ` [PATCH v7 01/32] drm/xe: Add NULL checks to scratch LRC allocation Matthew Brost
2025-10-07 11:26 ` [PATCH v7 02/32] drm/xe: Save off position in ring in which a job was programmed Matthew Brost
2025-10-07 11:26 ` [PATCH v7 03/32] drm/xe/guc: Track pending-enable source in submission state Matthew Brost
2025-10-07 11:26 ` [PATCH v7 04/32] drm/xe: Track LR jobs in DRM scheduler pending list Matthew Brost
2025-10-07 11:26 ` [PATCH v7 05/32] drm/xe: Don't change LRC ring head on job resubmission Matthew Brost
2025-10-07 12:29 ` Matthew Auld
2025-10-07 13:00 ` Matthew Brost
2025-10-07 11:26 ` [PATCH v7 06/32] drm/xe: Make LRC W/A scratch buffer usage consistent Matthew Brost
2025-10-07 11:26 ` [PATCH v7 07/32] drm/xe/vf: Add xe_gt_recovery_pending helper Matthew Brost
2025-10-07 11:26 ` [PATCH v7 08/32] drm/xe/vf: Make VF recovery run on per-GT worker Matthew Brost
2025-10-07 11:26 ` [PATCH v7 09/32] drm/xe/vf: Abort H2G sends during VF post-migration recovery Matthew Brost
2025-10-07 11:26 ` [PATCH v7 10/32] drm/xe/vf: Remove memory allocations from VF post migration recovery Matthew Brost
2025-10-07 11:26 ` [PATCH v7 11/32] drm/xe: Move GGTT lock init to alloc Matthew Brost
2025-10-07 11:47 ` Michal Wajdeczko
2025-10-07 12:53 ` Matthew Brost [this message]
2025-10-07 11:26 ` [PATCH v7 12/32] drm/xe/vf: Close multi-GT GGTT shift race Matthew Brost
2025-10-07 13:03 ` Michal Wajdeczko
2025-10-07 11:26 ` [PATCH v7 13/32] drm/xe/vf: Teardown VF post migration worker on driver unload Matthew Brost
2025-10-07 11:26 ` [PATCH v7 14/32] drm/xe/vf: Don't allow GT reset to be queued during VF post migration recovery Matthew Brost
2025-10-07 11:26 ` [PATCH v7 15/32] drm/xe/vf: Wakeup in GuC backend on " Matthew Brost
2025-10-07 11:26 ` [PATCH v7 16/32] drm/xe/vf: Avoid indefinite blocking in preempt rebind worker for VFs supporting migration Matthew Brost
2025-10-07 11:26 ` [PATCH v7 17/32] drm/xe/vf: Use GUC_HXG_TYPE_EVENT for GuC context register Matthew Brost
2025-10-07 11:26 ` [PATCH v7 18/32] drm/xe/vf: Flush and stop CTs in VF post migration recovery Matthew Brost
2025-10-07 11:26 ` [PATCH v7 19/32] drm/xe/vf: Reset TLB invalidations during " Matthew Brost
2025-10-07 11:26 ` [PATCH v7 20/32] drm/xe/vf: Kickstart after resfix in " Matthew Brost
2025-10-07 11:26 ` [PATCH v7 21/32] drm/xe: Add CTB_H2G_BUFFER_OFFSET define Matthew Brost
2025-10-07 11:26 ` [PATCH v7 22/32] drm/xe/vf: Start CTs before resfix VF post migration recovery Matthew Brost
2025-10-07 11:26 ` [PATCH v7 23/32] drm/xe/vf: Abort VF post migration recovery on failure Matthew Brost
2025-10-07 11:26 ` [PATCH v7 24/32] drm/xe/vf: Replay GuC submission state on pause / unpause Matthew Brost
2025-10-07 11:26 ` [PATCH v7 25/32] drm/xe: Move queue init before LRC creation Matthew Brost
2025-10-07 11:26 ` [PATCH v7 26/32] drm/xe/vf: Add debug prints for GuC replaying state during VF recovery Matthew Brost
2025-10-07 11:26 ` [PATCH v7 27/32] drm/xe/vf: Workaround for race condition in GuC firmware during VF pause Matthew Brost
2025-10-07 11:26 ` [PATCH v7 28/32] drm/xe: Use PPGTT addresses for TLB invalidation to avoid GGTT fixups Matthew Brost
2025-10-07 11:26 ` [PATCH v7 29/32] drm/xe/vf: Use primary GT ordered work queue on media GT on PTL VF Matthew Brost
2025-10-07 11:26 ` [PATCH v7 30/32] drm/xe/vf: Ensure media GT VF recovery runs after primary GT on PTL Matthew Brost
2025-10-07 11:26 ` [PATCH v7 31/32] drm/xe/vf: Rebase CCS save/restore BB GGTT addresses Matthew Brost
2025-10-07 11:26 ` [PATCH v7 32/32] drm/xe/guc: Increase wait timeout to 2sec after BUSY reply from GuC Matthew Brost
2025-10-07 11:38 ` ✗ CI.checkpatch: warning for VF migration redesign (rev7) Patchwork
2025-10-07 11:39 ` ✓ CI.KUnit: success " Patchwork
2025-10-07 12:26 ` ✗ Xe.CI.BAT: failure " Patchwork
2025-10-07 14:23 ` ✗ 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=aOUNNFAx2in/HIwo@lstrano-desk.jf.intel.com \
--to=matthew.brost@intel.com \
--cc=intel-xe@lists.freedesktop.org \
--cc=michal.wajdeczko@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