Return-Path: Received: from mail-qt0-f193.google.com ([209.85.216.193]:45082 "EHLO mail-qt0-f193.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932686AbeGIUZ5 (ORCPT ); Mon, 9 Jul 2018 16:25:57 -0400 Received: by mail-qt0-f193.google.com with SMTP id y5-v6so16498725qti.12 for ; Mon, 09 Jul 2018 13:25:57 -0700 (PDT) Date: Mon, 9 Jul 2018 17:25:52 -0300 From: Ernesto =?utf-8?Q?A=2E_Fern=C3=A1ndez?= To: linux-fsdevel@vger.kernel.org Cc: Andrew Morton , Anatoly Trosinenko Subject: [PATCH] hfsplus: avoid deadlock on file truncation Message-ID: <20180709202549.auxwkb6memlegb4a@eaf> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline Content-Transfer-Encoding: 8bit Sender: linux-fsdevel-owner@vger.kernel.org List-ID: After an extent is removed from the extent tree, the corresponding bits are also cleared from the block allocation file. This is currently done without releasing the tree lock. The problem is that the allocation file has extents of its own; if it is fragmented enough, some of them may be in the extent tree as well, and hfsplus_get_block() will try to take the lock again. To avoid deadlock, only hold the extent tree lock during the actual tree operations. Reported-by: Anatoly Trosinenko Signed-off-by: Ernesto A. Fernández --- fs/hfsplus/extents.c | 18 ++++++++++++++---- 1 file changed, 14 insertions(+), 4 deletions(-) diff --git a/fs/hfsplus/extents.c b/fs/hfsplus/extents.c index e8770935ce6d..8e0f59767694 100644 --- a/fs/hfsplus/extents.c +++ b/fs/hfsplus/extents.c @@ -336,6 +336,9 @@ static int hfsplus_free_extents(struct super_block *sb, int i; int err = 0; + /* Mapping the allocation file may lock the extent tree */ + WARN_ON(mutex_is_locked(&HFSPLUS_SB(sb)->ext_tree->tree_lock)); + hfsplus_dump_extent(extent); for (i = 0; i < 8; extent++, i++) { count = be32_to_cpu(extent->block_count); @@ -415,11 +418,13 @@ int hfsplus_free_fork(struct super_block *sb, u32 cnid, if (res) break; start = be32_to_cpu(fd.key->ext.start_block); - hfsplus_free_extents(sb, ext_entry, - total_blocks - start, - total_blocks); hfs_brec_remove(&fd); + + mutex_unlock(&fd.tree->tree_lock); + hfsplus_free_extents(sb, ext_entry, total_blocks - start, + total_blocks); total_blocks = start; + mutex_lock(&fd.tree->tree_lock); } while (total_blocks > blocks); hfs_find_exit(&fd); @@ -576,15 +581,20 @@ void hfsplus_file_truncate(struct inode *inode) } while (1) { if (alloc_cnt == hip->first_blocks) { + mutex_unlock(&fd.tree->tree_lock); hfsplus_free_extents(sb, hip->first_extents, alloc_cnt, alloc_cnt - blk_cnt); hfsplus_dump_extent(hip->first_extents); hip->first_blocks = blk_cnt; + mutex_lock(&fd.tree->tree_lock); break; } res = __hfsplus_ext_cache_extent(&fd, inode, alloc_cnt); if (res) break; + hfs_brec_remove(&fd); + + mutex_unlock(&fd.tree->tree_lock); start = hip->cached_start; hfsplus_free_extents(sb, hip->cached_extents, alloc_cnt - start, alloc_cnt - blk_cnt); @@ -596,7 +606,7 @@ void hfsplus_file_truncate(struct inode *inode) alloc_cnt = start; hip->cached_start = hip->cached_blocks = 0; hip->extent_state &= ~(HFSPLUS_EXT_DIRTY | HFSPLUS_EXT_NEW); - hfs_brec_remove(&fd); + mutex_lock(&fd.tree->tree_lock); } hfs_find_exit(&fd); -- 2.11.0