From mboxrd@z Thu Jan 1 00:00:00 1970 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 53FFE79C9 for ; Thu, 12 Jan 2023 14:36:18 +0000 (UTC) Received: by smtp.kernel.org (Postfix) with ESMTPSA id CE893C433D2; Thu, 12 Jan 2023 14:36:17 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=linuxfoundation.org; s=korg; t=1673534178; bh=FdQ9fvcTciS/WM+YFLgI4Fv7lwPF8crNV4iZRxtLE6U=; h=From:To:Cc:Subject:Date:In-Reply-To:References:From; b=jbJBKRO7+oCMXYovTIPwHbC8ad8KgaTGQIC0N7adxemu8xrfGEqV6zg6FyYx+zEi6 LrfwZHaHPTisKjMQTyIZznD1zcY5ToEaMbao0ZMgv+3JNe0d+vctgxY35WJ3GeC3K8 mS8Qo7nRa2s/aV2sNH3X8ezfGlgFnysVXuQL+LUo= From: Greg Kroah-Hartman To: stable@vger.kernel.org Cc: Greg Kroah-Hartman , patches@lists.linux.dev, Jan Kara , Theodore Tso , Sasha Levin , Ritesh Harjani Subject: [PATCH 5.10 721/783] ext4: fix race when reusing xattr blocks Date: Thu, 12 Jan 2023 14:57:17 +0100 Message-Id: <20230112135557.781641387@linuxfoundation.org> X-Mailer: git-send-email 2.39.0 In-Reply-To: <20230112135524.143670746@linuxfoundation.org> References: <20230112135524.143670746@linuxfoundation.org> User-Agent: quilt/0.67 Precedence: bulk X-Mailing-List: patches@lists.linux.dev List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit From: Jan Kara [ Upstream commit 65f8b80053a1b2fd602daa6814e62d6fa90e5e9b ] When ext4_xattr_block_set() decides to remove xattr block the following race can happen: CPU1 CPU2 ext4_xattr_block_set() ext4_xattr_release_block() new_bh = ext4_xattr_block_cache_find() lock_buffer(bh); ref = le32_to_cpu(BHDR(bh)->h_refcount); if (ref == 1) { ... mb_cache_entry_delete(); unlock_buffer(bh); ext4_free_blocks(); ... ext4_forget(..., bh, ...); jbd2_journal_revoke(..., bh); ext4_journal_get_write_access(..., new_bh, ...) do_get_write_access() jbd2_journal_cancel_revoke(..., new_bh); Later the code in ext4_xattr_block_set() finds out the block got freed and cancels reusal of the block but the revoke stays canceled and so in case of block reuse and journal replay the filesystem can get corrupted. If the race works out slightly differently, we can also hit assertions in the jbd2 code. Fix the problem by making sure that once matching mbcache entry is found, code dropping the last xattr block reference (or trying to modify xattr block in place) waits until the mbcache entry reference is dropped. This way code trying to reuse xattr block is protected from someone trying to drop the last reference to xattr block. Reported-and-tested-by: Ritesh Harjani CC: stable@vger.kernel.org Fixes: 82939d7999df ("ext4: convert to mbcache2") Signed-off-by: Jan Kara Link: https://lore.kernel.org/r/20220712105436.32204-5-jack@suse.cz Signed-off-by: Theodore Ts'o Stable-dep-of: a44e84a9b776 ("ext4: fix deadlock due to mbcache entry corruption") Signed-off-by: Sasha Levin --- fs/ext4/xattr.c | 67 +++++++++++++++++++++++++++++++++---------------- 1 file changed, 45 insertions(+), 22 deletions(-) diff --git a/fs/ext4/xattr.c b/fs/ext4/xattr.c index 9d5ccc90eb63..35251afdf770 100644 --- a/fs/ext4/xattr.c +++ b/fs/ext4/xattr.c @@ -439,9 +439,16 @@ static int ext4_xattr_inode_iget(struct inode *parent, unsigned long ea_ino, /* Remove entry from mbcache when EA inode is getting evicted */ void ext4_evict_ea_inode(struct inode *inode) { - if (EA_INODE_CACHE(inode)) - mb_cache_entry_delete(EA_INODE_CACHE(inode), - ext4_xattr_inode_get_hash(inode), inode->i_ino); + struct mb_cache_entry *oe; + + if (!EA_INODE_CACHE(inode)) + return; + /* Wait for entry to get unused so that we can remove it */ + while ((oe = mb_cache_entry_delete_or_get(EA_INODE_CACHE(inode), + ext4_xattr_inode_get_hash(inode), inode->i_ino))) { + mb_cache_entry_wait_unused(oe); + mb_cache_entry_put(EA_INODE_CACHE(inode), oe); + } } static int @@ -1223,6 +1230,7 @@ ext4_xattr_release_block(handle_t *handle, struct inode *inode, if (error) goto out; +retry_ref: lock_buffer(bh); hash = le32_to_cpu(BHDR(bh)->h_hash); ref = le32_to_cpu(BHDR(bh)->h_refcount); @@ -1232,9 +1240,18 @@ ext4_xattr_release_block(handle_t *handle, struct inode *inode, * This must happen under buffer lock for * ext4_xattr_block_set() to reliably detect freed block */ - if (ea_block_cache) - mb_cache_entry_delete(ea_block_cache, hash, - bh->b_blocknr); + if (ea_block_cache) { + struct mb_cache_entry *oe; + + oe = mb_cache_entry_delete_or_get(ea_block_cache, hash, + bh->b_blocknr); + if (oe) { + unlock_buffer(bh); + mb_cache_entry_wait_unused(oe); + mb_cache_entry_put(ea_block_cache, oe); + goto retry_ref; + } + } get_bh(bh); unlock_buffer(bh); @@ -1862,9 +1879,20 @@ ext4_xattr_block_set(handle_t *handle, struct inode *inode, * ext4_xattr_block_set() to reliably detect modified * block */ - if (ea_block_cache) - mb_cache_entry_delete(ea_block_cache, hash, - bs->bh->b_blocknr); + if (ea_block_cache) { + struct mb_cache_entry *oe; + + oe = mb_cache_entry_delete_or_get(ea_block_cache, + hash, bs->bh->b_blocknr); + if (oe) { + /* + * Xattr block is getting reused. Leave + * it alone. + */ + mb_cache_entry_put(ea_block_cache, oe); + goto clone_block; + } + } ea_bdebug(bs->bh, "modifying in-place"); error = ext4_xattr_set_entry(i, s, handle, inode, true /* is_block */); @@ -1880,6 +1908,7 @@ ext4_xattr_block_set(handle_t *handle, struct inode *inode, goto cleanup; goto inserted; } +clone_block: unlock_buffer(bs->bh); ea_bdebug(bs->bh, "cloning"); s->base = kmemdup(BHDR(bs->bh), bs->bh->b_size, GFP_NOFS); @@ -1985,18 +2014,13 @@ ext4_xattr_block_set(handle_t *handle, struct inode *inode, lock_buffer(new_bh); /* * We have to be careful about races with - * freeing, rehashing or adding references to - * xattr block. Once we hold buffer lock xattr - * block's state is stable so we can check - * whether the block got freed / rehashed or - * not. Since we unhash mbcache entry under - * buffer lock when freeing / rehashing xattr - * block, checking whether entry is still - * hashed is reliable. Same rules hold for - * e_reusable handling. + * adding references to xattr block. Once we + * hold buffer lock xattr block's state is + * stable so we can check the additional + * reference fits. */ - if (hlist_bl_unhashed(&ce->e_hash_list) || - !ce->e_reusable) { + ref = le32_to_cpu(BHDR(new_bh)->h_refcount) + 1; + if (ref > EXT4_XATTR_REFCOUNT_MAX) { /* * Undo everything and check mbcache * again. @@ -2011,9 +2035,8 @@ ext4_xattr_block_set(handle_t *handle, struct inode *inode, new_bh = NULL; goto inserted; } - ref = le32_to_cpu(BHDR(new_bh)->h_refcount) + 1; BHDR(new_bh)->h_refcount = cpu_to_le32(ref); - if (ref >= EXT4_XATTR_REFCOUNT_MAX) + if (ref == EXT4_XATTR_REFCOUNT_MAX) ce->e_reusable = 0; ea_bdebug(new_bh, "reusing; refcount now=%d", ref); -- 2.35.1