From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mail-pj2-f3.google.com (mail-pj2-f3.google.com [74.125.227.131]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id 015D32D3A75 for ; Tue, 30 Jun 2026 04:59:26 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=74.125.227.131 ARC-Seal:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1782795568; cv=none; b=kKJqFBReclbMhUuEYTBvdfifLWTS5hBBMgC14FJ8m6HzeNuuL3UMBo23Tn5MCIyZdy9D2y+FvqBt3Bg0srBtmB/maHM7LTDKnLXfuxWMBOAKmz5zWC59jCTVGpu9ADuWd87mmEXbnSat9o8RlzmiTd69UV+Nd1gTEZbw8f6PVJc= ARC-Message-Signature:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1782795568; c=relaxed/simple; bh=/ueZ6U8ojmHyNy1GMbcQHwxHy92rFTUoNJj4acpYfdo=; h=From:To:Cc:Subject:Date:Message-Id:MIME-Version; b=chYjnVpjetltF6NmCzCRssYG4ZGEq12meafkibigJVFUBQ4l8RgL96UcMzGp6H4ldmngk64nepIlk+lZdHttZv/IRBIHeB5+09c8wkcC1mB4mCow/1lbY3rWSWnPrA+FmtAP7K7Hu+DbEmxRC1l+0cjCO4ogyJDtmVN5rnIBPGA= ARC-Authentication-Results:i=1; smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=gmail.com; spf=pass smtp.mailfrom=gmail.com; dkim=pass (2048-bit key) header.d=gmail.com header.i=@gmail.com header.b=YkY8lAyT; arc=none smtp.client-ip=74.125.227.131 Authentication-Results: smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=gmail.com Authentication-Results: smtp.subspace.kernel.org; spf=pass smtp.mailfrom=gmail.com Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=gmail.com header.i=@gmail.com header.b="YkY8lAyT" Received: by mail-pj2-f3.google.com with SMTP id d9443c01a7336-2c9dc2dafcaso765295ad.1 for ; Mon, 29 Jun 2026 21:59:26 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20251104; t=1782795566; x=1783400366; darn=vger.kernel.org; h=content-transfer-encoding:mime-version:message-id:date:subject:cc :to:from:from:to:cc:subject:date:message-id:reply-to; bh=T0nHTHMTaczDjmeESFhp8Z52IMkOcrljdDCN0lb0WWY=; b=YkY8lAyTtmjuK0eb51hOVpjKYNRSX8fU6Bz+vPJb3VGhN7I/eKj3EN+ttWmZ+hj8QW ZT5Ny5UJvyO8fnk42pBJ+vE/zBEwhmgPaFKqYnocXkBvSw/JVkzHrapuXbw2r30kQ1xv k6G+kOYHYj3th5czDXi9z4XVXO4f7csKeAOY6yF9p/emqPDgtfgqCukvmgs4abeecXbh 47K/ospQWZCvEqI/X3At8k8xnLpWtDEfkgjD3BgYwZ5W4RGGwKao3vB6OxhI3irqC7Xc YM63CZgYr+3whwEedji2Ki4Npyg7uCfMbi+F09vftchZGJLG5RMkRQMRHQOnyfbZP/NW t8HQ== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20251104; t=1782795566; x=1783400366; 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=T0nHTHMTaczDjmeESFhp8Z52IMkOcrljdDCN0lb0WWY=; b=NgJUGhstFAwf6V+CUDqbALoak6wDHptS6wabrKnv0ez075jiUrzJUW+m8ZhxhDqh+Q W8G7Et9ZUgsZyfH7nYjV0UBXNSm72CuDcFG5MCoPuP+spk8TLTdf80FIZV/5SfI036YH MOn7DwfXk7b6lrSgF1mvC/FMCYMWOnzZSkDCJ6h57fMZZNmcUvIZLCaEtL5uJDE/S28G KBL8KkMRQ0MMjl9sN+OkD990AH/tnaBSz/fQ+5VA9zT7SOIkNU8qX5MmhUMyYuAuZP+L o+KarIkMIbKtnvFyOGsILqZq8AKLJI/i0Mzw7dRjZ29UT8ijTg2AJTjZar2cWJIW58Mk B+Yw== X-Forwarded-Encrypted: i=1; AHgh+RqmB5Xu38m6ZzKbJdtXd+is57qGnOXHArI+5rageVyuMbyWqJyt/6KGvZvroKrOBik7GntG14AyM3rkgRc=@vger.kernel.org X-Gm-Message-State: AOJu0YyXnrnGZr0Ltkpa4Kpgd/aL4/p1hpw1CXv4oofEEUr+lIKUrVnH F0kaOi4yi2ZbvNEnlKF961LoKjbnfEi75E0UY8/e5efnS3hW3OKGvXMKad8Z8m/hX/c= X-Gm-Gg: AfdE7ck6+0qz3bVD86iWr1JowTbKGBJXwOvUTc9TyWZgbdXZRY6Diw1EGFhuKCDakGD kVbA+Ul4vaV4xn/g9Bi7DjxQNCFPyakCGEHFgr6R9gyNky7SJUlVpiIEZMARK1Icgl6Vv0+SDBv HOj9CvnUrUskG/tSiUmdReR43vNjikVo4FmFRuU/ZcGUkWTIfGjF6sjDHrrQcg5Zcxr8SMRMXCJ DVV3tA/8bNUYz9ZyBswA7yQbQ3tREsby+pWQNoNiJO0nRL/KxDT2T/hsYCoYIOcH9EbUMEl21Pp LD7CgqYgUB6vriEi0mS7HyK8WMhHomtDPbp1Bif7QAKD/dKol6vA8TknLVDMPvnis5Zlz0R9kxG NDkrxXy2r33/PAMHWvDgJtGMM2doPt4IH9hnpsSHNIGVQkElVlhhZbaV22zjnF01fgLoOoPKgMe WCLEJu/qhPxw9Y1pjq8xdO/1UAwoWRYqEDBEAaaLPcNd3K/1trByGEQsUOape23haWOmE= X-Received: by 2002:a17:903:fa3:b0:2ca:2390:ff13 with SMTP id d9443c01a7336-2ca4488914bmr2471015ad.5.1782795566227; Mon, 29 Jun 2026 21:59:26 -0700 (PDT) Received: from localhost ([111.228.63.84]) by smtp.gmail.com with ESMTPSA id d9443c01a7336-2ca37c82becsm5470465ad.27.2026.06.29.21.59.23 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Mon, 29 Jun 2026 21:59:25 -0700 (PDT) From: Zhang Cen To: Jaegeuk Kim , Chao Yu Cc: linux-f2fs-devel@lists.sourceforge.net, linux-kernel@vger.kernel.org Subject: [PATCH v6] f2fs: protect gc task pointer during teardown Date: Tue, 30 Jun 2026 12:59:19 +0800 Message-Id: <20260630045919.1671833-1-rollkingzzc@gmail.com> X-Mailer: git-send-email 2.34.1 Precedence: bulk X-Mailing-List: linux-kernel@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Transfer-Encoding: 8bit This patch is based on the preceding patch "f2fs: embed f2fs_gc_kthread in f2fs_sb_info", which gives the GC waitqueues and thread tunables superblock lifetime. With that container lifetime fixed, f2fs_gc_task remains the running-state marker and the task_struct pointer that needs separate lifetime protection. During shutdown, f2fs_stop_gc_thread() stops the GC task and wakes GC_MERGE foreground waiters while f2fs_balance_fs() can concurrently queue on gc_th->fggc_wq. The preceding embedded-gc_thread patch keeps that waitqueue storage alive for the superblock lifetime; this patch protects the remaining f2fs_gc_task pointer by publishing and detaching it under gc_task_lock, and by taking a task_struct reference for sysfs users that dereference the task. The buggy scenario involves two paths, with each column showing the order within that path: foreground f2fs_balance_fs() caller: shutdown path: 1. observes no checkpoint error 1. sets CP_ERROR_FLAG 2. snapshots sbi->gc_thread 2. enters f2fs_stop_gc_thread() 3. queues on gc_th->fggc_wq 3. stops gc_th->f2fs_gc_task 4. wakes gc_wait_queue_head 4. wakes gc_th->fggc_wq 5. sleeps for foreground GC 5. frees gc_th in the old layout 6. finish_wait() touches fggc_wq GC_MERGE does not keep independent work_struct items that shutdown can cancel. Its pending foreground GC requests are waitqueue waiters. Drain them by withdrawing the GC task pointer, stopping the task, waking gc_th->fggc_wq, and leaving each waiter to remove its own wait entry with finish_wait(). Add gc_task_lock to struct f2fs_gc_kthread and use it to publish the GC task only after the new kthread has been created and its nice value has been set. The start path uses kthread_create() so the task is not woken until after f2fs_gc_task is visible to waiters. The stop path detaches f2fs_gc_task under the same lock before kthread_stop(), so later readers see that no new foreground GC work should be handed to the background thread. f2fs_balance_fs() also rechecks both f2fs_cp_error() and f2fs_gc_task after prepare_to_wait(). If shutdown is visible or the GC task has already been withdrawn, the caller removes its wait entry without waking the GC thread or sleeping for new foreground GC work. Thus shutdown drains already queued waiters and stops accepting new foreground GC work once shutdown is visible to the caller. Task pointer users are protected separately from the embedded container lifetime. A sysfs critical_task_priority store now snapshots the GC task under gc_task_lock and holds a task_struct reference while calling set_user_nice(). Boolean running-state checks that do not dereference the task_struct continue to use READ_ONCE(). One observed report was: BUG: KASAN: slab-use-after-free in finish_wait+0x276/0x290 Write of size 8 at addr ffff8881150819b8 by task dd/802 The buggy address belongs to the object at ffff888115081900 which belongs to the cache kmalloc-256 of size 256 The buggy address is located 184 bytes inside of freed 256-byte region Call trace: finish_wait() f2fs_balance_fs() f2fs_write_single_data_page() f2fs_write_cache_pages() __f2fs_write_data_pages() do_writepages() filemap_fdatawrite_wbc() __filemap_fdatawrite_range() file_write_and_wait_range() f2fs_do_sync_file() f2fs_sync_file() do_fsync() Freed by task stack: kfree() f2fs_stop_gc_thread() f2fs_do_shutdown() f2fs_shutdown() fs_bdev_mark_dead() Fixes: 5911d2d1d1a3 ("f2fs: introduce gc_merge mount option") Assisted-by: Codex:gpt-5.5 Signed-off-by: Zhang Cen --- Based on: - [PATCH] f2fs: embed f2fs_gc_kthread in f2fs_sb_info fs/f2fs/f2fs.h | 37 +++++++++++++++++++++++++++++++++++++ fs/f2fs/gc.c | 12 ++++++++---- fs/f2fs/segment.c | 2 +- fs/f2fs/super.c | 1 + fs/f2fs/sysfs.c | 6 ++++-- 5 files changed, 51 insertions(+), 7 deletions(-) diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h index 9ab196f65643..931d37f64bb3 100644 --- a/fs/f2fs/f2fs.h +++ b/fs/f2fs/f2fs.h @@ -17,8 +17,10 @@ #include #include #include +#include #include #include +#include #include #include #include @@ -430,6 +432,7 @@ struct ckpt_req_control { struct f2fs_gc_kthread { struct task_struct *f2fs_gc_task; + spinlock_t gc_task_lock; /* protects f2fs_gc_task */ wait_queue_head_t gc_wait_queue_head; /* for gc sleep time */ @@ -455,6 +458,40 @@ struct f2fs_gc_kthread { unsigned int boost_gc_greedy; }; +static inline struct task_struct *f2fs_get_gc_task(struct f2fs_gc_kthread *gc_th) +{ + struct task_struct *task; + + spin_lock(&gc_th->gc_task_lock); + task = READ_ONCE(gc_th->f2fs_gc_task); + if (task) + get_task_struct(task); + spin_unlock(&gc_th->gc_task_lock); + + return task; +} + +static inline void f2fs_set_gc_task(struct f2fs_gc_kthread *gc_th, + struct task_struct *task) +{ + spin_lock(&gc_th->gc_task_lock); + WRITE_ONCE(gc_th->f2fs_gc_task, task); + spin_unlock(&gc_th->gc_task_lock); +} + +static inline struct task_struct * +f2fs_detach_gc_task(struct f2fs_gc_kthread *gc_th) +{ + struct task_struct *task; + + spin_lock(&gc_th->gc_task_lock); + task = READ_ONCE(gc_th->f2fs_gc_task); + WRITE_ONCE(gc_th->f2fs_gc_task, NULL); + spin_unlock(&gc_th->gc_task_lock); + + return task; +} + /* a time threshold that checkpoint was blocked for, unit: ms */ #define CP_LONG_LATENCY_THRESHOLD 5000 diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c index 5a98754354b6..c26fcaffa986 100644 --- a/fs/f2fs/gc.c +++ b/fs/f2fs/gc.c @@ -197,6 +197,9 @@ int f2fs_start_gc_thread(struct f2fs_sb_info *sbi) struct task_struct *task; dev_t dev = sbi->sb->s_bdev->bd_dev; + if (READ_ONCE(gc_th->f2fs_gc_task)) + return 0; + gc_th->urgent_sleep_time = DEF_GC_THREAD_URGENT_SLEEP_TIME; gc_th->valid_thresh_ratio = DEF_GC_THREAD_VALID_THRESH_RATIO; gc_th->boost_gc_multiple = BOOST_GC_MULTIPLE; @@ -218,16 +221,17 @@ int f2fs_start_gc_thread(struct f2fs_sb_info *sbi) gc_th->gc_wake = false; - task = kthread_run(gc_thread_func, sbi, "f2fs_gc-%u:%u", - MAJOR(dev), MINOR(dev)); + task = kthread_create(gc_thread_func, sbi, "f2fs_gc-%u:%u", + MAJOR(dev), MINOR(dev)); if (IS_ERR(task)) { int err = PTR_ERR(task); return err; } - WRITE_ONCE(gc_th->f2fs_gc_task, task); set_user_nice(task, PRIO_TO_NICE(sbi->critical_task_priority)); + f2fs_set_gc_task(gc_th, task); + wake_up_process(task); return 0; } @@ -236,7 +240,7 @@ void f2fs_stop_gc_thread(struct f2fs_sb_info *sbi) struct f2fs_gc_kthread *gc_th = &sbi->gc_thread; struct task_struct *task; - task = xchg(&gc_th->f2fs_gc_task, NULL); + task = f2fs_detach_gc_task(gc_th); if (!task) return; kthread_stop(task); diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c index 877b015e24cd..aeda8351398f 100644 --- a/fs/f2fs/segment.c +++ b/fs/f2fs/segment.c @@ -451,7 +451,7 @@ void f2fs_balance_fs(struct f2fs_sb_info *sbi, bool need) prepare_to_wait(&gc_th->fggc_wq, &wait, TASK_UNINTERRUPTIBLE); - if (READ_ONCE(gc_th->f2fs_gc_task)) { + if (!f2fs_cp_error(sbi) && READ_ONCE(gc_th->f2fs_gc_task)) { wake_up(&gc_th->gc_wait_queue_head); io_schedule(); } diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c index ec44c5be8ca0..ed1ce1266ce4 100644 --- a/fs/f2fs/super.c +++ b/fs/f2fs/super.c @@ -4993,6 +4993,7 @@ static int f2fs_fill_super(struct super_block *sb, struct fs_context *fc) spin_lock_init(&sbi->inode_lock[i]); } mutex_init(&sbi->flush_lock); + spin_lock_init(&sbi->gc_thread.gc_task_lock); init_waitqueue_head(&sbi->gc_thread.gc_wait_queue_head); init_waitqueue_head(&sbi->gc_thread.fggc_wq); diff --git a/fs/f2fs/sysfs.c b/fs/f2fs/sysfs.c index 705563dfc560..87aa348f2a2b 100644 --- a/fs/f2fs/sysfs.c +++ b/fs/f2fs/sysfs.c @@ -996,9 +996,11 @@ static ssize_t __sbi_store(struct f2fs_attr *a, nice = PRIO_TO_NICE(sbi->critical_task_priority); if (sbi->cprc_info.f2fs_issue_ckpt) set_user_nice(sbi->cprc_info.f2fs_issue_ckpt, nice); - gc_task = READ_ONCE(sbi->gc_thread.f2fs_gc_task); - if (gc_task) + gc_task = f2fs_get_gc_task(&sbi->gc_thread); + if (gc_task) { set_user_nice(gc_task, nice); + put_task_struct(gc_task); + } return count; } -- 2.43.0