From: Mingming Cao Subject: Re: [EXT4 set 5][PATCH 1/1] expand inode i_extra_isize to support features in larger inode Date: Mon, 16 Jul 2007 16:52:03 -0700 Message-ID: <1184629924.3836.5.camel@localhost.localdomain> References: <1183275482.4010.133.camel@localhost.localdomain> <20070710163247.5c8bfa3f.akpm@linux-foundation.org> <20070713020529.1486491f.akpm@linux-foundation.org> Reply-To: cmm@us.ibm.com Mime-Version: 1.0 Content-Type: text/plain Content-Transfer-Encoding: 7bit Cc: linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org, linux-ext4@vger.kernel.org, Andy Whitcroft To: Andrew Morton Return-path: Received: from e4.ny.us.ibm.com ([32.97.182.144]:46435 "EHLO e4.ny.us.ibm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1761598AbXGPXwD (ORCPT ); Mon, 16 Jul 2007 19:52:03 -0400 In-Reply-To: <20070713020529.1486491f.akpm@linux-foundation.org> Sender: linux-ext4-owner@vger.kernel.org List-Id: linux-ext4.vger.kernel.org On Fri, 2007-07-13 at 02:05 -0700, Andrew Morton wrote: > On Tue, 10 Jul 2007 16:32:47 -0700 Andrew Morton wrote: > > > > + brelse(bh); > > > + up_write(&EXT4_I(inode)->xattr_sem); > > > + return error; > > > +} > > > + > > > > We're doing GFP_KERNEL memory allocations while holding xattr_sem. This > > can cause the VM to reenter the filesystem, perhaps taking i_mutex and/or > > i_truncate_sem and/or journal_start() (I forget whether this still > > happens). Have we checked whether this can occur and if so, whether we are > > OK from a lock ranking POV? Bear in mind that journalled-data mode is more > > complex in this regard. > > I notice that everyone carefully avoided addressing this ;) I am not sure why we need GFP_KERNEL flag here. I think we should use GFP_NOFS instead. The following patch use the GFP_NOFS flag, as well as fixing memory leak issue introduced by the ext4 expand inode extra isize patch. Fixing memory allocation issue with expand inode extra isize patch. - use GFP_NOFS instead of GFP_KERNEL flag for memory allocation - use kzalloc instead of kmalloc - fix memory leak in the success case, at the end of while loop. Signed-off-by: Mingming Cao --- fs/ext4/xattr.c | 16 ++++++++++++---- 1 file changed, 12 insertions(+), 4 deletions(-) Index: linux-2.6.22/fs/ext4/xattr.c =================================================================== --- linux-2.6.22.orig/fs/ext4/xattr.c 2007-07-16 16:12:18.000000000 -0700 +++ linux-2.6.22/fs/ext4/xattr.c 2007-07-16 16:35:59.000000000 -0700 @@ -1204,8 +1204,8 @@ retry: unsigned int shift_bytes; /* No. of bytes to shift EAs by? */ unsigned int min_total_size = ~0U; - is = kzalloc(sizeof(struct ext4_xattr_ibody_find), GFP_KERNEL); - bs = kzalloc(sizeof(struct ext4_xattr_block_find), GFP_KERNEL); + is = kzalloc(sizeof(struct ext4_xattr_ibody_find), GFP_NOFS); + bs = kzalloc(sizeof(struct ext4_xattr_block_find), GFP_NOFS); if (!is || !bs) { error = -ENOMEM; goto cleanup; @@ -1251,8 +1251,8 @@ retry: size = le32_to_cpu(entry->e_value_size); entry_size = EXT4_XATTR_LEN(entry->e_name_len); i.name_index = entry->e_name_index, - buffer = kmalloc(EXT4_XATTR_SIZE(size), GFP_KERNEL); - b_entry_name = kmalloc(entry->e_name_len + 1, GFP_KERNEL); + buffer = kzalloc(EXT4_XATTR_SIZE(size), GFP_NOFS); + b_entry_name = kzalloc(entry->e_name_len + 1, GFP_NOFS); if (!buffer || !b_entry_name) { error = -ENOMEM; goto cleanup; @@ -1302,7 +1302,15 @@ retry: error = ext4_xattr_block_set(handle, inode, &i, bs); if (error) goto cleanup; + kfree(b_entry_name); + kfree(buffer); + brelse(is->iloc.bh); + kfree(is); + kfree(bs); + brelse(bh); } + up_write(&EXT4_I(inode)->xattr_sem); + return 0; cleanup: kfree(b_entry_name);