Intel-XE Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: Nitin Gote <nitin.r.gote@intel.com>
To: intel-xe@lists.freedesktop.org
Cc: himal.prasad.ghimiray@intel.com, matthew.brost@intel.com,
	thomas.hellstrom@intel.com, Nitin Gote <nitin.r.gote@intel.com>
Subject: [PATCH v1 3/5] drm/xe: add drm_exec helper to atomically lock VM (+ optional BO) with BOOKKEEP retry
Date: Thu, 18 Sep 2025 19:55:27 +0530	[thread overview]
Message-ID: <20250918142529.608432-4-nitin.r.gote@intel.com> (raw)
In-Reply-To: <20250918142529.608432-1-nitin.r.gote@intel.com>

Add xe_vm_exec_lock_vm_and_bo(), a drm_exec-based helper that atomically
acquires the VM reservation and an optional BO reservation. When requested
(wait_bookkeep) the helper verifies DMA_RESV_USAGE_BOOKKEEP is clear and,
if not, releases the exec locks, waits outside the reservation and retries.

This centralises the drm_exec + BOOKKEEP wait/retry logic used by vm_bind
paths, avoiding duplicated code and preventing sleeps while holding resv
locks. Caller must call drm_exec_fini(exec) on success and must not
re-acquire vm->lock while exec holds the reservations.

Cc: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
Signed-off-by: Nitin Gote <nitin.r.gote@intel.com>
---
 drivers/gpu/drm/xe/xe_vm.c | 53 ++++++++++++++++++++++++++++++++++++++
 1 file changed, 53 insertions(+)

diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c
index 0cacab20ff85..fae88c4c981e 100644
--- a/drivers/gpu/drm/xe/xe_vm.c
+++ b/drivers/gpu/drm/xe/xe_vm.c
@@ -3496,6 +3496,59 @@ static int xe_vm_bind_ioctl_validate_bo(struct xe_device *xe, struct xe_bo *bo,
 	return 0;
 }
 
+/* Helper:
+ * Atomically acquire VM (+ optional BO) reservations using drm_exec.
+ * If wait_bookkeep is true, ensure VM BOOKKEEP is clear before returning:
+ *  - if BOOKKEEP fences exist, release exec locks, wait outside and retry.
+ *
+ * On success returns 0 with *exec initialized and the reservations held.
+ * Caller MUST call drm_exec_fini(exec) to release reservations.
+ */
+static int xe_vm_exec_lock_vm_and_bo(struct xe_vm *vm, struct xe_bo *bo,
+				     struct drm_exec *exec, bool intr,
+				     bool wait_bookkeep)
+{
+	int err = 0, wait;
+
+retry:
+	drm_exec_init(exec, intr ? DRM_EXEC_INTERRUPTIBLE_WAIT : 0, 0);
+	drm_exec_until_all_locked(exec) {
+		err = drm_exec_lock_obj(exec, xe_vm_obj(vm));
+		drm_exec_retry_on_contention(exec);
+		if (err)
+			break;
+
+		if (bo)	{
+			err = drm_exec_lock_obj(exec, &bo->ttm.base);
+			drm_exec_retry_on_contention(exec);
+			if (err)
+				break;
+		}
+	}
+
+	if (err) {
+		/* Ensure exec cleaned up if partially initialized */
+		drm_exec_fini(exec);
+		return err;
+	}
+
+	if (wait_bookkeep) {
+		/* If BOOKKEEP fences exist, release exec and wait outside, then retry */
+		if (!dma_resv_test_signaled(xe_vm_resv(vm), DMA_RESV_USAGE_BOOKKEEP)) {
+			drm_exec_fini(exec);
+			wait = dma_resv_wait_timeout(xe_vm_resv(vm),
+						     DMA_RESV_USAGE_BOOKKEEP,
+						     true, MAX_SCHEDULE_TIMEOUT);
+			if (wait <= 0)
+				return -ETIME;
+			goto retry;
+		}
+	}
+
+	/* Success: exec holds the reservations */
+	return 0;
+}
+
 int xe_vm_bind_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
 {
 	struct xe_device *xe = to_xe_device(dev);
-- 
2.25.1


  parent reply	other threads:[~2025-09-18 13:56 UTC|newest]

Thread overview: 14+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-09-18 14:25 [PATCH v1 0/5] drm/xe: add VM_BIND DECOMPRESS support and on‑demand decompression Nitin Gote
2025-09-18 14:25 ` [PATCH v1 1/5] drm/xe: add VM_BIND DECOMPRESS uapi flag Nitin Gote
2025-09-23 17:35   ` Matthew Brost
2025-09-18 14:25 ` [PATCH v1 2/5] drm/xe: add xe_migrate_resolve wrapper and is_vram_resolve support Nitin Gote
2025-09-18 14:25 ` Nitin Gote [this message]
2025-09-18 14:25 ` [PATCH v1 4/5] drm/xe: implement VM_BIND decompression in vm_bind_ioctl Nitin Gote
2025-09-18 16:13   ` Matthew Auld
2025-09-22  7:57     ` Gote, Nitin R
2025-09-23 17:26   ` Matthew Brost
2025-09-18 14:25 ` [PATCH v1 5/5] drm/xe: perform on-demand decompression on VMA pagefaults Nitin Gote
2025-09-23 17:29   ` Matthew Brost
2025-09-18 15:12 ` ✓ CI.KUnit: success for drm/xe: add VM_BIND DECOMPRESS support and on‑demand decompression Patchwork
2025-09-18 15:57 ` ✓ Xe.CI.BAT: " Patchwork
2025-09-19  0:37 ` ✓ 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=20250918142529.608432-4-nitin.r.gote@intel.com \
    --to=nitin.r.gote@intel.com \
    --cc=himal.prasad.ghimiray@intel.com \
    --cc=intel-xe@lists.freedesktop.org \
    --cc=matthew.brost@intel.com \
    --cc=thomas.hellstrom@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