Re: [PATCH] zero out i_blocks in get_pipe_inode

Re: [PATCH] zero out i_blocks in get_pipe_inode

[Andrew Morton]

"Martin J. Bligh" <mbligh@aracnet.com> wrote:
>
>  +	inode->i_blocks = 0;

alloc_inode() already did that.

Re: [PATCH] zero out i_blocks in get_pipe_inode

[Andi Kleen]

On Tue, 11 Nov 2003 14:48:42 -0800
Andrew Morton <akpm@osdl.org> wrote:

> "Martin J. Bligh" <mbligh@aracnet.com> wrote:
> >
> >  +	inode->i_blocks = 0;
> 
> alloc_inode() already did that.

It didn't in 2.4 I think. But it may be worth auditing if all fields are covered.

-Andi

[PATCH] zero out i_blocks in get_pipe_inode

[Martin J. Bligh]

Fished from the 2.4 SuSE tree, which I'm trawling through.
This seems like a perfectly reasonable thing to do to me,
there was some discussion earlier on IRC. Some extracts from
the conversation were "returning random old data from the kernel 
is always a bug", "any userland code that trips on that one is 
broken", and "iirc it broke postfix". Nobody seemed to think it
was actively evil, and it seems to fix a bug ;-)

diff -purN -X /home/mbligh/.diff.exclude virgin/fs/pipe.c pipe_init/fs/pipe.c
--- virgin/fs/pipe.c	2003-10-14 15:50:30.000000000 -0700
+++ pipe_init/fs/pipe.c	2003-11-11 12:13:32.000000000 -0800
@@ -527,6 +527,7 @@ static struct inode * get_pipe_inode(voi
 	inode->i_gid = current->fsgid;
 	inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
 	inode->i_blksize = PAGE_SIZE;
+	inode->i_blocks = 0;
 	return inode;
 
 fail_iput:

Re: [PATCH] zero out i_blocks in get_pipe_inode

[Martin J. Bligh]

>>  +	inode->i_blocks = 0;
> 
> alloc_inode() already did that.

Sigh ... yes, you're right - sorry.

get_pipe_inode -> new_inode -> alloc_inode. definitely covered.

I shall crawl back under my regularly scheduled stone ;-)

M.