From patchwork Sun Nov 17 21:32:05 2024 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Chuck Lever X-Patchwork-Id: 13877966 Received: from smtp.kernel.org (aws-us-west-2-korg-mail-1.web.codeaurora.org [10.30.226.201]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id DE4FD1DDA2D; Sun, 17 Nov 2024 21:32:23 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=10.30.226.201 ARC-Seal: i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1731879144; cv=none; b=NvhcYjulBbQjZ1yqYAEA3xOE8/urNGbGiH6e0YBqIBa9gb56eKpf06KkiMXCODfmwUDY9bj6CVq7Qet/abP6tz0JYCIiGISU9V7nOEjlP8qNU8KQ6vEzzZbK+w4uqUQY2dZHGmUqxpMP2N/CFVSfq/1CZv2DnujgsnS6rKIoCN4= ARC-Message-Signature: i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1731879144; c=relaxed/simple; bh=uhgh93plhiFV1pHFjFAu6nt5SYYT8N+YCq9mpcqsLXk=; h=From:To:Cc:Subject:Date:Message-ID:In-Reply-To:References: MIME-Version; b=ItpIu73t2OlXpHpOpVIKnsP38XQ4kYVnx0nr9hL2EOSGLf7j+UkYdUKL3OmlTKeTXLt1SiUph8CKyQbWzKgjAhYR2NMWRlPXpaSd103u2A5xKEX4XbfUAL2LiOvF/gB1aq1W5zjinogB0yAQc0wMZUi6b3DMje6xzm5qoezlTgI= ARC-Authentication-Results: i=1; smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=kernel.org header.i=@kernel.org header.b=YwLmSCsh; arc=none smtp.client-ip=10.30.226.201 Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=kernel.org header.i=@kernel.org header.b="YwLmSCsh" Received: by smtp.kernel.org (Postfix) with ESMTPSA id 14632C4CED8; Sun, 17 Nov 2024 21:32:23 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=kernel.org; s=k20201202; t=1731879143; bh=uhgh93plhiFV1pHFjFAu6nt5SYYT8N+YCq9mpcqsLXk=; h=From:To:Cc:Subject:Date:In-Reply-To:References:From; b=YwLmSCshoiCMBHQ8SdisMNZY5nlg2cAQVDgmOSPyURaXMQ4aCumx3TQV0KV/+aF0d nY4VP9HmKvlqTs/RX9o2Z1gCH0tU7BiqNXsFXj5+AnT546vfGt5QpnRiK29A8edJoB GwoKdrdVbhYnlwkXxdzVH3PDZ3CIOFJxcLxZhjtoqOVrmnCBWJqV47cbz23wZ8Fhjt 4+Y/fD9zKUvNuY4hTnlGY+ssNQ5Bp0O0X00VooV2Hr7/PAuer8xTbU1F32iEz5F4+x q6Y+uvCsWTU5LKrqqnCg9dgroyUS3FYb8wZG0msfJ4XmRLA9qWnZHFzGFhRfR9kgtV NK7InFdZjg3lg== From: cel@kernel.org To: , , Hugh Dickens Cc: yukuai3@huawei.com, yangerkun@huaweicloud.com, Chuck Lever , stable@vger.kernel.org Subject: [RFC PATCH 1/2] libfs: Return ENOSPC when the directory offset range is exhausted Date: Sun, 17 Nov 2024 16:32:05 -0500 Message-ID: <20241117213206.1636438-2-cel@kernel.org> X-Mailer: git-send-email 2.47.0 In-Reply-To: <20241117213206.1636438-1-cel@kernel.org> References: <20241117213206.1636438-1-cel@kernel.org> Precedence: bulk X-Mailing-List: linux-fsdevel@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 From: Chuck Lever Testing shows that the EBUSY error return from mtree_alloc_cyclic() leaks into user space. The ERRORS section of "man creat(2)" says: > EBUSY O_EXCL was specified in flags and pathname refers > to a block device that is in use by the system > (e.g., it is mounted). ENOSPC is closer to what applications expect in this situation. Note that the normal range of simple directory offset values is 2..2^63, so hitting this error is going to be rare to impossible. Fixes: 6faddda69f62 ("libfs: Add directory operations for stable offsets") Cc: # v6.9+ Signed-off-by: Chuck Lever Reviewed-by: Jeff Layton --- fs/libfs.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/fs/libfs.c b/fs/libfs.c index 46966fd8bcf9..bf67954b525b 100644 --- a/fs/libfs.c +++ b/fs/libfs.c @@ -288,7 +288,9 @@ int simple_offset_add(struct offset_ctx *octx, struct dentry *dentry) ret = mtree_alloc_cyclic(&octx->mt, &offset, dentry, DIR_OFFSET_MIN, LONG_MAX, &octx->next_offset, GFP_KERNEL); - if (ret < 0) + if (unlikely(ret == -EBUSY)) + return -ENOSPC; + if (unlikely(ret < 0)) return ret; offset_set(dentry, offset); From patchwork Sun Nov 17 21:32:06 2024 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Chuck Lever X-Patchwork-Id: 13877967 Received: from smtp.kernel.org (aws-us-west-2-korg-mail-1.web.codeaurora.org [10.30.226.201]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id 083CD1DDA2D for ; Sun, 17 Nov 2024 21:32:24 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=10.30.226.201 ARC-Seal: i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1731879145; cv=none; b=g9H2vlgzGsxl3mXBntGap0LlLS09NWlhpMSZeiEhyBzt6BdgS2aCZIObh+NzOztbNebtVczsri1RxhDZW3qghMlS2G1yeBayAlQxPqfhUYgAEYRvgcnT92AOf0jPVjzC26PSCRRahUz9LXIIWFIAWktOiU160CoKvyjoPl03YnA= ARC-Message-Signature: i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1731879145; c=relaxed/simple; bh=Imws/SMxyuWl9wa5OM4mA83z05SUQjW00CX49uy8e2Q=; h=From:To:Cc:Subject:Date:Message-ID:In-Reply-To:References: MIME-Version; b=ErWBAqvUCtRE7f0De64+Y8rWfU5WruOsJ2LeSrfpe637MqAS9Up9jwUYzEf93+SsstYizeYbEsTXfzokSizkNnDuRisO2H+f4VdVK+yA9lVR1+poEDonHP4OLdjtnWAsOWG1ivSMF1xx8IntZRv11EnxQxNaX26rffVUjetp1Cg= ARC-Authentication-Results: i=1; smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=kernel.org header.i=@kernel.org header.b=CrUEwBWi; arc=none smtp.client-ip=10.30.226.201 Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=kernel.org header.i=@kernel.org header.b="CrUEwBWi" Received: by smtp.kernel.org (Postfix) with ESMTPSA id F02CEC4CED7; Sun, 17 Nov 2024 21:32:23 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=kernel.org; s=k20201202; t=1731879144; bh=Imws/SMxyuWl9wa5OM4mA83z05SUQjW00CX49uy8e2Q=; h=From:To:Cc:Subject:Date:In-Reply-To:References:From; b=CrUEwBWi8qr+8LgPYCTL0ObS1dvFTb6tStBaAmqhwR7WOv75uob1xy+nsGzACW624 983hu1q4sR4cX2Lk83LRRVBgzOtVX/6owVkk0TrNqlzfdYLXdwCM1XCR0Np7GPjuSU Og8beXb5TVOqbhoeokxP7WPfYbCgY+DnXAiqTVGuT6U/dlT6k9cvhSK9ggu9RMZ+xR 8D3o4NLYk2W6dkM3nuWMPPRYJd+pRrgJjMhRsoEJ+oYPBgYwkU9qX/mN0B3VNI4BNK ghueOUjbNtj2HMnqRey5UkJoKe8lBeI+KHwdbhSMMID/1zeihiN8kyRHaZkNYuYjYA s6bhxF0uRx91w== From: cel@kernel.org To: , , Hugh Dickens Cc: yukuai3@huawei.com, yangerkun@huaweicloud.com, Chuck Lever Subject: [RFC PATCH 2/2] libfs: Improve behavior when directory offset values wrap Date: Sun, 17 Nov 2024 16:32:06 -0500 Message-ID: <20241117213206.1636438-3-cel@kernel.org> X-Mailer: git-send-email 2.47.0 In-Reply-To: <20241117213206.1636438-1-cel@kernel.org> References: <20241117213206.1636438-1-cel@kernel.org> Precedence: bulk X-Mailing-List: linux-fsdevel@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 From: Chuck Lever The fix in commit 64a7ce76fb90 ("libfs: fix infinite directory reads for offset dir") introduced a fence in offset_iterate_dir() to stop the loop from returning child entries created after the directory was opened. This comparison relies on the strong ordering of DIR_OFFSET_MIN <= largest child offset <= next_offset to terminate the directory iteration. However, because simple_offset_add() uses mtree_alloc_cyclic() to select each next new directory offset, ctx->next_offset is not always the highest unused offset. Once mtree_alloc_cyclic() allows a new offset value to wrap, ctx->next_offset will be set to a value less than the actual largest child offset. The result is that readdir(3) no longer shows any entries in the directory because their offsets are above ctx->next_offset, which is now a small value. This situation is persistent, and the directory cannot be removed unless all current children are already known and can be explicitly removed by name first. In the current Maple tree implementation, there is no practical way that 63-bit offset values can ever wrap, so this issue is cleverly avoided. But the ordering dependency is not documented via comments or code, making the mechanism somewhat brittle. And it makes the continued use of mtree_alloc_cyclic() somewhat confusing. Further, if commit 64a7ce76fb90 ("libfs: fix infinite directory reads for offset dir") were to be backported to a kernel that still uses xarray to manage simple directory offsets, the directory offset value range is limited to 32-bits, which is small enough to allow a wrap after a few weeks of constant creation of entries in one directory. Therefore, replace the use of ctx->next_offset for fencing new children from appearing in readdir results. A jiffies timestamp marks the end of each opendir epoch. Entries created after this timestamp will not be visible to the file descriptor. I chose jiffies so that the dentry->d_time field can be re-used for storing the entry creation time. The new mechanism has its own corner cases. For instance, I think if jiffies wraps twice while a directory is open, some children might become invisible. On 32-bit systems, the jiffies value wraps every 49 days. Double-wrapping is not a risk on systems with 64-bit jiffies. Unlike with the next_offset-based mechanism, re-opening the directory will make invisible children re-appear. Reported-by: Yu Kuai Closes: https://lore.kernel.org/stable/20241111005242.34654-1-cel@kernel.org/T/#m1c448e5bd4aae3632a09468affcfe1d1594c6a59 Fixes: 64a7ce76fb90 ("libfs: fix infinite directory reads for offset dir") Signed-off-by: Chuck Lever Reviewed-by: Jeff Layton Signed-off-by: Hugh Dickins --- fs/libfs.c | 36 +++++++++++++++++------------------- 1 file changed, 17 insertions(+), 19 deletions(-) diff --git a/fs/libfs.c b/fs/libfs.c index bf67954b525b..862a603fd454 100644 --- a/fs/libfs.c +++ b/fs/libfs.c @@ -294,6 +294,7 @@ int simple_offset_add(struct offset_ctx *octx, struct dentry *dentry) return ret; offset_set(dentry, offset); + WRITE_ONCE(dentry->d_time, jiffies); return 0; } @@ -454,9 +455,7 @@ void simple_offset_destroy(struct offset_ctx *octx) static int offset_dir_open(struct inode *inode, struct file *file) { - struct offset_ctx *ctx = inode->i_op->get_offset_ctx(inode); - - file->private_data = (void *)ctx->next_offset; + file->private_data = (void *)jiffies; return 0; } @@ -473,9 +472,6 @@ static int offset_dir_open(struct inode *inode, struct file *file) */ static loff_t offset_dir_llseek(struct file *file, loff_t offset, int whence) { - struct inode *inode = file->f_inode; - struct offset_ctx *ctx = inode->i_op->get_offset_ctx(inode); - switch (whence) { case SEEK_CUR: offset += file->f_pos; @@ -490,7 +486,8 @@ static loff_t offset_dir_llseek(struct file *file, loff_t offset, int whence) /* In this case, ->private_data is protected by f_pos_lock */ if (!offset) - file->private_data = (void *)ctx->next_offset; + /* Make newer child entries visible */ + file->private_data = (void *)jiffies; return vfs_setpos(file, offset, LONG_MAX); } @@ -521,7 +518,8 @@ static bool offset_dir_emit(struct dir_context *ctx, struct dentry *dentry) inode->i_ino, fs_umode_to_dtype(inode->i_mode)); } -static void offset_iterate_dir(struct inode *inode, struct dir_context *ctx, long last_index) +static void offset_iterate_dir(struct inode *inode, struct dir_context *ctx, + unsigned long fence) { struct offset_ctx *octx = inode->i_op->get_offset_ctx(inode); struct dentry *dentry; @@ -531,14 +529,15 @@ static void offset_iterate_dir(struct inode *inode, struct dir_context *ctx, lon if (!dentry) return; - if (dentry2offset(dentry) >= last_index) { - dput(dentry); - return; - } - - if (!offset_dir_emit(ctx, dentry)) { - dput(dentry); - return; + /* + * Output only child entries created during or before + * the current opendir epoch. + */ + if (time_before_eq(dentry->d_time, fence)) { + if (!offset_dir_emit(ctx, dentry)) { + dput(dentry); + return; + } } ctx->pos = dentry2offset(dentry) + 1; @@ -569,15 +568,14 @@ static void offset_iterate_dir(struct inode *inode, struct dir_context *ctx, lon */ static int offset_readdir(struct file *file, struct dir_context *ctx) { + unsigned long fence = (unsigned long)file->private_data; struct dentry *dir = file->f_path.dentry; - long last_index = (long)file->private_data; lockdep_assert_held(&d_inode(dir)->i_rwsem); if (!dir_emit_dots(file, ctx)) return 0; - - offset_iterate_dir(d_inode(dir), ctx, last_index); + offset_iterate_dir(d_inode(dir), ctx, fence); return 0; }