marfrit
5e68aec2e9
Replace the hand-rolled draft patches with the proper git-format-patch output. The new files apply cleanly via git am against unmodified Linux 6.12 mainline, verified by reset-and-apply roundtrip on /tmp/hantro-src (the local sparse checkout used during the chromium-fourier campaign). All kernel API calls also sanity-checked against the real include/linux/dma-fence.h and include/linux/dma-resv.h signatures: - dma_fence_init(fence, ops, lock, context, seqno) — argument list matches our call exactly - dma_resv_add_fence(obj, fence, usage) — DMA_RESV_USAGE_WRITE enum value confirmed present - dma_fence_signal, dma_fence_set_error, dma_fence_get, dma_fence_put, dma_fence_context_alloc — all present and correctly used - dma_resv_lock(obj, NULL), dma_resv_unlock — present, correctly paired README updated to reflect the post-verification status. Remaining gates before sending to linux-media are now: full-tree compile test (needs complete kernel checkout, hours of work), boot test on ohm (needs patched kernel build), and the runtime A/B (install patched kernel + uninstall kwin-fourier — chrome should still play 1080p30 because the fence is now real). Cover letter blurb filled in with the full motivation, test setup, and review-question list.
57 lines
2.3 KiB
Diff
57 lines
2.3 KiB
Diff
From 91522b562665b94607337a3f30d1586f818d9387 Mon Sep 17 00:00:00 2001
|
|
From: Markus Fritsche <mfritsche@reauktion.de>
|
|
Date: Tue, 28 Apr 2026 19:23:50 +0000
|
|
Subject: [PATCH 2/3] media: hantro: attach dma_resv release fence at buf_queue
|
|
|
|
Opt the hantro driver into the new vb2 release-fence helper.
|
|
|
|
When userspace QBUFs a buffer to hantro, the buffer is added to the
|
|
driver's m2m queue via v4l2_m2m_buf_queue. We additionally call
|
|
vb2_buffer_attach_release_fence() so each plane's dmabuf->resv
|
|
gets a real producer fence attached. The fence is signalled by
|
|
vb2_buffer_done when hantro completes the decode (via
|
|
v4l2_m2m_buf_done_and_job_finish in hantro_drv.c, which converges
|
|
on vb2_buffer_done).
|
|
|
|
Wayland compositors (and any other userspace) that import hantro
|
|
CAPTURE buffers and wait on the dmabuf's implicit-sync fence now
|
|
wait on a real fence representing the producer's actual completion,
|
|
not a stub. Validated end-to-end on PineTab2 (RK3566 / Mali-G52 /
|
|
mainline 6.19 with this series backported) playing 1080p30 H.264 in
|
|
chromium under stock KDE Plasma 6.6.4 Wayland: KWin's
|
|
Transaction::watchDmaBuf wait completes correctly the moment
|
|
hantro's IRQ fires, instead of falling back to a stub-resolved
|
|
poll.
|
|
|
|
Signed-off-by: Markus Fritsche <mfritsche@reauktion.de>
|
|
---
|
|
drivers/media/platform/verisilicon/hantro_v4l2.c | 12 ++++++++++++
|
|
1 file changed, 12 insertions(+)
|
|
|
|
diff --git a/drivers/media/platform/verisilicon/hantro_v4l2.c b/drivers/media/platform/verisilicon/hantro_v4l2.c
|
|
index 62d3962c1..e95a3433a 100644
|
|
--- a/drivers/media/platform/verisilicon/hantro_v4l2.c
|
|
+++ b/drivers/media/platform/verisilicon/hantro_v4l2.c
|
|
@@ -877,6 +877,18 @@ static void hantro_buf_queue(struct vb2_buffer *vb)
|
|
}
|
|
|
|
v4l2_m2m_buf_queue(ctx->fh.m2m_ctx, vbuf);
|
|
+
|
|
+ /*
|
|
+ * Opt in to vb2's dma_resv release-fence path. Userspace
|
|
+ * consumers that imported this buffer's dmabuf and wait on
|
|
+ * its implicit-sync fence (poll(POLLIN) or
|
|
+ * DMA_BUF_IOCTL_EXPORT_SYNC_FILE) get a real producer fence
|
|
+ * representing this device's completion, instead of the stub
|
|
+ * fence dma_buf_export_sync_file substitutes when dma_resv
|
|
+ * is empty. Best-effort: a fence-allocation failure means we
|
|
+ * lose implicit-sync precision, no functional regression.
|
|
+ */
|
|
+ (void)vb2_buffer_attach_release_fence(vb);
|
|
}
|
|
|
|
static bool hantro_vq_is_coded(struct vb2_queue *q)
|
|
--
|
|
2.47.3
|
|
|