Searched hist:"8 a2bfdcb" (Results 1 – 1 of 1) sorted by relevance
/openbmc/linux/fs/ext4/ |
H A D | xattr.c | c1bb05a6 Mon Feb 20 22:06:18 CST 2012 Eric Sandeen <sandeen@redhat.com> ext4: avoid deadlock on sync-mounted FS w/o journal
Processes hang forever on a sync-mounted ext2 file system that is mounted with the ext4 module (default in Fedora 16).
I can reproduce this reliably by mounting an ext2 partition with "-o sync" and opening a new file an that partition with vim. vim will hang in "D" state forever. The same happens on ext4 without a journal.
I am attaching a small patch here that solves this issue for me. In the sync mounted case without a journal, ext4_handle_dirty_metadata() may call sync_dirty_buffer(), which can't be called with buffer lock held.
Also move mb_cache_entry_release inside lock to avoid race fixed previously by 8a2bfdcb ext[34]: EA block reference count racing fix Note too that ext2 fixed this same problem in 2006 with b2f49033 [PATCH] fix deadlock in ext2
Signed-off-by: Martin.Wilck@ts.fujitsu.com [sandeen@redhat.com: move mb_cache_entry_release before unlock, edit commit msg] Signed-off-by: Eric Sandeen <sandeen@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu> 8a2bfdcb Wed Feb 28 22:13:35 CST 2007 Mingming Cao <cmm@us.ibm.com> [PATCH] ext[34]: EA block reference count racing fix
There are race issues around ext[34] xattr block release code.
ext[34]_xattr_release_block() checks the reference count of xattr block (h_refcount) and frees that xattr block if it is the last one reference it. Unlike ext2, the check of this counter is unprotected by any lock. ext[34]_xattr_release_block() will free the mb_cache entry before freeing that xattr block. There is a small window between the check for the re h_refcount ==1 and the call to mb_cache_entry_free(). During this small window another inode might find this xattr block from the mbcache and reuse it, racing a refcount updates. The xattr block will later be freed by the first inode without notice other inode is still use it. Later if that block is reallocated as a datablock for other file, then more serious problem might happen.
We need put a lock around places checking the refount as well to avoid racing issue. Another place need this kind of protection is in ext3_xattr_block_set(), where it will modify the xattr block content in- the-fly if the refcount is 1 (means it's the only inode reference it).
This will also fix another issue: the xattr block may not get freed at all if no lock is to protect the refcount check at the release time. It is possible that the last two inodes could release the shared xattr block at the same time. But both of them think they are not the last one so only decreased the h_refcount without freeing xattr block at all.
We need to call lock_buffer() after ext3_journal_get_write_access() to avoid deadlock (because the later will call lock_buffer()/unlock_buffer () as well).
Signed-off-by: Mingming Cao <cmm@us.ibm.com> Cc: Andreas Gruenbacher <agruen@suse.de> Cc: <linux-ext4@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|