From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from gabe.freedesktop.org (gabe.freedesktop.org [131.252.210.177]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.lore.kernel.org (Postfix) with ESMTPS id 4B9B0CCD1BC for ; Thu, 23 Oct 2025 14:30:52 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id B73BF10E909; Thu, 23 Oct 2025 14:30:51 +0000 (UTC) Authentication-Results: gabe.freedesktop.org; dkim=pass (2048-bit key; unprotected) header.d=kernel.org header.i=@kernel.org header.b="hcQIrwxt"; dkim-atps=neutral Received: from sea.source.kernel.org (sea.source.kernel.org [172.234.252.31]) by gabe.freedesktop.org (Postfix) with ESMTPS id 2326D10E90A for ; Thu, 23 Oct 2025 14:30:48 +0000 (UTC) Received: from smtp.kernel.org (transwarp.subspace.kernel.org [100.75.92.58]) by sea.source.kernel.org (Postfix) with ESMTP id F25BF4AEDD; Thu, 23 Oct 2025 14:30:47 +0000 (UTC) Received: by smtp.kernel.org (Postfix) with ESMTPSA id 6C273C4CEF7; Thu, 23 Oct 2025 14:30:45 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=kernel.org; s=k20201202; t=1761229847; bh=lRRm98sjHaVsvhoTag39RzQU6GJrENsxxPSr2a2wy8g=; h=From:To:Cc:Subject:Date:In-Reply-To:References:From; b=hcQIrwxtBlYYrXZf3XUxjvVdQdMgos2eH0AI+i52u4J5Lc9drlw1/M/N5EFNIckEu CVfKOeQgIfvCNx3KK+L4fdwqZlMQNTCWM3c3pPoUk9vPgvroVwz91kH7kLRdvJbAS3 ptr2DoR3CqFoWgTTgroKBzqt2rTsd2QOyDWxrccVjPe9iRHSxxwfPJtbfWEqwy/Ouy RLGbf3buKk4oKus8R0brAGh9sec+F6ShfsbDjKjXwYBSIIkiRRfJsA4Wpcyo7tMV16 BoFzWAm3duA3IWR8/mmhs6toB++A937UsZYTtSwwsdAU1XB3YPm0tNe60+8TDbq3Kz F5Vd/WB9OvGuQ== From: Philipp Stanner To: Matthew Brost , Danilo Krummrich , Philipp Stanner , =?UTF-8?q?Christian=20K=C3=B6nig?= , David Airlie , Simona Vetter , Tvrtko Ursulin Cc: dri-devel@lists.freedesktop.org, linux-kernel@vger.kernel.org, linux-media@vger.kernel.org Subject: [PATCH v2 3/4] drm/sched: Add TODO entry for resubmitting jobs Date: Thu, 23 Oct 2025 16:30:30 +0200 Message-ID: <20251023143031.149496-5-phasta@kernel.org> X-Mailer: git-send-email 2.49.0 In-Reply-To: <20251023143031.149496-2-phasta@kernel.org> References: <20251023143031.149496-2-phasta@kernel.org> MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit X-BeenThere: dri-devel@lists.freedesktop.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Direct Rendering Infrastructure - Development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dri-devel-bounces@lists.freedesktop.org Sender: "dri-devel" Add the issue of a successor of drm_sched_resubmit_jobs() missing to the TODO file. Signed-off-by: Philipp Stanner --- drivers/gpu/drm/scheduler/TODO | 26 ++++++++++++++++++++++++++ 1 file changed, 26 insertions(+) diff --git a/drivers/gpu/drm/scheduler/TODO b/drivers/gpu/drm/scheduler/TODO index 79044adb7d01..713dd62c58da 100644 --- a/drivers/gpu/drm/scheduler/TODO +++ b/drivers/gpu/drm/scheduler/TODO @@ -10,3 +10,29 @@ - Tasks: 1. Read the example entry. 2. Remove the entry once solved (never in this case) + +* GPU job resubmits + - Difficulty: hard + - Contact: + - Christian König + - Philipp Stanner + - Description: + drm_sched_resubmit_jobs() is deprecated. Main reason being that it leads to + reinitializing dma_fences. See that function's docu for details. The better + approach for valid resubmissions by amdgpu and Xe is (apparently) to figure + out which job (and, through association: which entity) caused the hang. Then, + the job's buffer data, together with all other jobs' buffer data currently + in the same hardware ring, must be invalidated. This can for example be done + by overwriting it. + amdgpu currently determines which jobs are in the ring and need to be + overwritten by keeping copies of the job. Xe obtains that information by + directly accessing drm_sched's pending_list. + - Tasks: + 1. implement scheduler functionality through which + the driver can obtain the information which *broken* jobs are currently in + the hardware ring. + 2. Such infrastructure would then typically be used in + drm_sched_backend_ops.timedout_job(). Document that. + 3. Port a driver as first user. + 3. Document the new alternative in the docu of deprecated + drm_sched_resubmit_jobs(). -- 2.49.0