From mboxrd@z Thu Jan 1 00:00:00 1970 From: Marco Stornelli Date: Sat, 06 Apr 2013 17:00:07 +0200 Subject: [Cluster-devel] [PATCH 3/4] fsfreeze: manage kill signal when sb_start_pagefault is called In-Reply-To: <20130406132028.GD28744@parisc-linux.org> References: <515FF380.5020406@gmail.com> <20130406132028.GD28744@parisc-linux.org> Message-ID: <51603877.7070904@gmail.com> List-Id: To: cluster-devel.redhat.com MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Il 06/04/2013 15:20, Matthew Wilcox ha scritto: > On Sat, Apr 06, 2013 at 12:05:52PM +0200, Marco Stornelli wrote: >> In every place where sb_start_pagefault was called now we must manage >> the error code and return VM_FAULT_RETRY. > > Erm ... in patch 1/4: > > static inline void sb_start_pagefault(struct super_block *sb) > { > - __sb_start_write(sb, SB_FREEZE_PAGEFAULT, true); > + __sb_start_write_wait(sb, SB_FREEZE_PAGEFAULT, false); > } > >> >> - sb_start_pagefault(inode->i_sb); >> + ret = sb_start_pagefault(inode->i_sb); >> + if (ret) >> + return VM_FAULT_RETRY; >> ret = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE); > > Does the compiler not warn that you're assigning void to 'ret'? Or was > there some other SNAFU sending these patches? > I'm sorry, my fault :) As I said in 00 these patches are completely *not* tested, it was only a "quick coding & review" to understand if someone can see any problem to this kind of implementation, since I touched several points in the kernel. So there is still on-going work and I need to do several tests. Maybe I had to add the RFC tag, sorry again. Marco