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 9BB5DCD342F for ; Fri, 8 May 2026 06:57:32 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id E372610F374; Fri, 8 May 2026 06:57:31 +0000 (UTC) Authentication-Results: gabe.freedesktop.org; dkim=pass (1024-bit key; secure) header.d=quora.org header.i=@quora.org header.b="Fg148Ifz"; dkim-atps=neutral Received: from mail-pl1-f173.google.com (mail-pl1-f173.google.com [209.85.214.173]) by gabe.freedesktop.org (Postfix) with ESMTPS id 9B6E210F369 for ; Fri, 8 May 2026 06:57:30 +0000 (UTC) Received: by mail-pl1-f173.google.com with SMTP id d9443c01a7336-2ba928852a5so11130975ad.1 for ; Thu, 07 May 2026 23:57:30 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=quora.org; s=google; t=1778223450; x=1778828250; darn=lists.freedesktop.org; h=content-transfer-encoding:mime-version:message-id:date:subject:cc :to:from:from:to:cc:subject:date:message-id:reply-to; bh=TetnCK6YoCpHxJceDDOD/QB+oKi7cDlyFmiU7j9vyl4=; b=Fg148Ifz4z82MpRZspkofvPaR5ppXVugAHpVhEEyMQ0W2HG/X9qbj5roP6TGy6cjYR f11vbEJdojEnDwHc7bXo7gW87e7EbCXSs/fvZtKz8qJbnuP0qrAeuu8NCnxvyUdShpEa 0G8hI/MXL81jquvhMBR5Wx/mUcnmMEu4Uzh7E= X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20251104; t=1778223450; x=1778828250; h=content-transfer-encoding:mime-version:message-id:date:subject:cc :to:from:x-gm-gg:x-gm-message-state:from:to:cc:subject:date :message-id:reply-to; bh=TetnCK6YoCpHxJceDDOD/QB+oKi7cDlyFmiU7j9vyl4=; b=a51+q+icNW0d7K+Q5Hnc//Zrx969VF40xMbMS14WnB9CsmLO+p9+tRYZ+in0gvk7ms O8SredP9nhmPwffbdXX2BSntPIU14H24bXfSLl0VIGs3B8gXYI/Lxvojz4lVtXelA3gv eK5BYlCWczw7hMlLenRYt6KFNEdHyDiTAtLws9rLHVKQVHaDxZUlWr8iwcwxSA5p6jox pnnHf1x9H/etJKTkz/xe6Kw4qBwYUwexF8Bgx71/RDqn+PD+ZlEYUo3rlyqFIzthSAjn AT9qf29dlr4jP276Rld3TWhnxKrg1RtCGfcnPZRLVNFzOz1EpIpCHh9dLoniAAP67dVz Te+A== X-Forwarded-Encrypted: i=1; AFNElJ/PDtz/6b+LFOcO/fBEG7j/taZsCSqcn5uE1tONE4LaOw5AYpLvs2tleoLkWZzPvwSu4je43gXr36c=@lists.freedesktop.org X-Gm-Message-State: AOJu0Yw4qogwB/wXqSbnnjVBBjEyY4q0uX7HHHH4phG6XpgeYy48fwWf pq/Yv0kRsH14f2xMfmPJQvygzAA54ccWLfdr8c1A9wEAqY3TqF7kSCGE94nMepw+0KQ= X-Gm-Gg: Acq92OHz/GU+bmpcL0ogGLq5NBzuCFMtZMOBGB/q+CRXBPGwVJ/kFPvYV/d5ZLOVm5z l5N3QnAAMg2v+VeMyC8NhaoHGkbVbBtHGGIFKw2nyi0oev367XfBUzJOKCloH0dwVVwg5aFBuVz c8av22jgXSZK/S1KA5Kcp/Mttey1VO0w/DQJp8EJq2DoPIKP3Iy8aM/LMbgoR8ol54WlAhzFpFr FwFYdeBDYngz2uci8yIGt6CK4nn+y+bkzeFgZ57j6Owzq+gd9dlp2gj+K/JV5zp6F6KaWx4ICth k5d38cmJsgXO6LeqnXVoK2rYkDvW3WPXh4l15WA6i9swvDpmNKm2B+48Pt3LZbA84quVVO51Lgs IHLC+EUnx2QaFCNGDMhcvsnEE3dRfG2wEFa/+8zct0iH6D+ytbf2M1FcIGgam/u1cO/2HEROkp7 5qrqgX1IS0dnvS69Y4FTNaRz4LHBnZmWqiKtXfAYvnD/eonJaTdG8MwF9BWkpWhbq1l5Kge9992 kCXz79GiUgAIwrQg2izJ2wVHBzavP6GmjDaipj5R8P1AwHukOUAtpeUwSHMUqstoUIT6/wQrVnR Qw5ZynFc299VVCgHH94xxiUSxsq/AFa1U9RVgQQBi4t3EULUSKkAuzccb1uRAqq+ X-Received: by 2002:a17:903:3c2b:b0:2b9:ea53:4cfc with SMTP id d9443c01a7336-2ba794b8118mr117896365ad.19.1778223450102; Thu, 07 May 2026 23:57:30 -0700 (PDT) Received: from aegis ([2001:fd8:4d01:3602:5317:86f3:5298:5270]) by smtp.gmail.com with ESMTPSA id d9443c01a7336-2baf1ead938sm10650785ad.73.2026.05.07.23.57.26 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Thu, 07 May 2026 23:57:29 -0700 (PDT) From: Daniel J Blueman To: Rob Clark , Dmitry Baryshkov , Abhinav Kumar , Jessica Zhang , Sean Paul , Marijn Suijten , David Airlie , Simona Vetter , Antonino Maniscalco Cc: Daniel J Blueman , linux-arm-msm@vger.kernel.org, dri-devel@lists.freedesktop.org, freedreno@lists.freedesktop.org, linux-kernel@vger.kernel.org, stable@vger.kernel.org Subject: [PATCH] drm/msm: Fix shrinker deadlock Date: Fri, 8 May 2026 14:57:21 +0800 Message-ID: <20260508065722.18785-1-daniel@quora.org> X-Mailer: git-send-email 2.53.0 MIME-Version: 1.0 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" With PROVE_LOCKING on an Snapdragon X1 and VM reclaim pressure, we see: """ kswapd0/121 is trying to acquire lock: ffff800080ed3800 (reservation_ww_class_acquire){+.+.}-{0:0}, at: msm_gem_shrinker_scan (drivers/gpu/drm/msm/msm_gem_shrinker.c:189) but task is already holding lock: ffffbf4ddb44ca40 (fs_reclaim){+.+.}-{0:0}, at: balance_pgdat (mm/vmscan.c:7236 (discriminator 2)) which lock already depends on the new lock. the existing dependency chain (in reverse order) is: -> #2 (fs_reclaim){+.+.}-{0:0}: lock_acquire (kernel/locking/lockdep.c:5868 kernel/locking/lockdep.c:5825) fs_reclaim_acquire (mm/page_alloc.c:4325 mm/page_alloc.c:4339) dma_resv_lockdep (drivers/dma-buf/dma-resv.c:798) do_one_initcall (init/main.c:1392) kernel_init_freeable (init/main.c:1454 (discriminator 1) init/main.c:1470 (discriminator 1) init/main.c:1490 (discriminator 1) init/main.c:1703 (discriminator 1)) kernel_init (init/main.c:1593) ret_from_fork (arch/arm64/kernel/entry.S:858) -> #1 (reservation_ww_class_mutex){+.+.}-{4:4}: lock_acquire (kernel/locking/lockdep.c:5868 kernel/locking/lockdep.c:5825) dma_resv_lockdep (./include/linux/ww_mutex.h:164 (discriminator 1) drivers/dma-buf/dma-resv.c:791 (discriminator 1)) do_one_initcall (init/main.c:1392) kernel_init_freeable (init/main.c:1454 (discriminator 1) init/main.c:1470 (discriminator 1) init/main.c:1490 (discriminator 1) init/main.c:1703 (discriminator 1)) kernel_init (init/main.c:1593) ret_from_fork (arch/arm64/kernel/entry.S:858) -> #0 (reservation_ww_class_acquire){+.+.}-{0:0}: check_prev_add (kernel/locking/lockdep.c:3165) __lock_acquire (kernel/locking/lockdep.c:3284 kernel/locking/lockdep.c:3908 kernel/locking/lockdep.c:5237) lock_acquire (kernel/locking/lockdep.c:5868 kernel/locking/lockdep.c:5825) drm_gem_lru_scan (./include/linux/ww_mutex.h:163 (discriminator 1) drivers/gpu/drm/drm_gem.c:1681 (discriminator 1)) msm_gem_shrinker_scan (drivers/gpu/drm/msm/msm_gem_shrinker.c:189) do_shrink_slab (mm/shrinker.c:436) shrink_slab (mm/shrinker.c:667) shrink_one (mm/vmscan.c:5071) shrink_node (mm/vmscan.c:5132 mm/vmscan.c:5210 mm/vmscan.c:6198) balance_pgdat (mm/vmscan.c:7052 mm/vmscan.c:7228) kswapd (mm/vmscan.c:7501) kthread (kernel/kthread.c:436) ret_from_fork (arch/arm64/kernel/entry.S:858) other info that might help us debug this: Chain exists of: reservation_ww_class_acquire --> reservation_ww_class_mutex --> fs_reclaim """ kswapd0 holding fs_reclaim calls the MSM shrinker, which calls dma_resv_lock. This in turn acquires fs_reclaim. Fix this deadlock by using dma_resv_trylock() instead, dropping the subsequently unused passed wait-wound lock 'ticket'. Cc: stable@vger.kernel.org Signed-off-by: Daniel J Blueman Fixes: fe4952b5f27c ("drm/msm: Convert vm locking") --- drivers/gpu/drm/msm/msm_gem_shrinker.c | 34 ++++++++++---------------- 1 file changed, 13 insertions(+), 21 deletions(-) diff --git a/drivers/gpu/drm/msm/msm_gem_shrinker.c b/drivers/gpu/drm/msm/msm_gem_shrinker.c index 31fa51a44f86..5320ef57dd90 100644 --- a/drivers/gpu/drm/msm/msm_gem_shrinker.c +++ b/drivers/gpu/drm/msm/msm_gem_shrinker.c @@ -43,8 +43,7 @@ msm_gem_shrinker_count(struct shrinker *shrinker, struct shrink_control *sc) } static bool -with_vm_locks(struct ww_acquire_ctx *ticket, - void (*fn)(struct drm_gem_object *obj), +with_vm_locks(void (*fn)(struct drm_gem_object *obj), struct drm_gem_object *obj) { /* @@ -52,7 +51,7 @@ with_vm_locks(struct ww_acquire_ctx *ticket, * success paths */ struct drm_gpuvm_bo *vm_bo, *last_locked = NULL; - int ret = 0; + bool locked = true; drm_gem_for_each_gpuvm_bo (vm_bo, obj) { struct dma_resv *resv = drm_gpuvm_resv(vm_bo->vm); @@ -60,23 +59,14 @@ with_vm_locks(struct ww_acquire_ctx *ticket, if (resv == obj->resv) continue; - ret = dma_resv_lock(resv, ticket); - /* - * Since we already skip the case when the VM and obj - * share a resv (ie. _NO_SHARE objs), we don't expect - * to hit a double-locking scenario... which the lock - * unwinding cannot really cope with. + * dma_resv_lock can't be used due to acquiring 'ticket' before the + * fs_reclaim lock, which is held in shrinker context */ - WARN_ON(ret == -EALREADY); - - /* - * Don't bother with slow-lock / backoff / retry sequence, - * if we can't get the lock just give up and move on to - * the next object. - */ - if (ret) + if (!dma_resv_trylock(resv)) { + locked = false; goto out_unlock; + } /* * Hold a ref to prevent the vm_bo from being freed @@ -108,7 +98,7 @@ with_vm_locks(struct ww_acquire_ctx *ticket, } } - return ret == 0; + return locked; } static bool @@ -120,7 +110,7 @@ purge(struct drm_gem_object *obj, struct ww_acquire_ctx *ticket) if (msm_gem_active(obj)) return false; - return with_vm_locks(ticket, msm_gem_purge, obj); + return with_vm_locks(msm_gem_purge, obj); } static bool @@ -164,7 +154,6 @@ static unsigned long msm_gem_shrinker_scan(struct shrinker *shrinker, struct shrink_control *sc) { struct msm_drm_private *priv = shrinker->private_data; - struct ww_acquire_ctx ticket; struct { struct drm_gem_lru *lru; bool (*shrink)(struct drm_gem_object *obj, struct ww_acquire_ctx *ticket); @@ -185,11 +174,14 @@ msm_gem_shrinker_scan(struct shrinker *shrinker, struct shrink_control *sc) for (unsigned i = 0; (nr > 0) && (i < ARRAY_SIZE(stages)); i++) { if (!stages[i].cond) continue; + /* + * 'ticket' not needed on trylock paths + */ stages[i].freed = drm_gem_lru_scan(stages[i].lru, nr, &stages[i].remaining, stages[i].shrink, - &ticket); + NULL); nr -= stages[i].freed; freed += stages[i].freed; remaining += stages[i].remaining; -- 2.53.0