: make sure the buffer head members are zeroed out before using them.

On Mon, Jan 26, 2009 at 6:03 PM, Manish Katiyar <mkatiyar@gmail.com> wrote:
> On Mon, Jan 26, 2009 at 5:59 PM, Jan Kara <jack@suse.cz> wrote:
>>> On Tue, Jan 20, 2009 at 10:36 PM, Manish Katiyar <mkatiyar@gmail.com> wrote:
>>> > ext2_quota_read doesn't bzeroes tmp_bh before calling ext2_get_block()
>>> > where we access the b_size of it. Since it is a local variable it
>>> > might contain some garbage. Make sure it is filled with zero before
>>> > passing.
>>>
>>> Hi Ted/mingming,
>>>
>>> Any feedback on this ??
>>  Ops, sorry. I wanted to reply but first I wanted to research more
>> whether we can set b_size to 0 and then forgot about it. Looking into
>> other code (e.g. in fs/mpage.c or fs/buffer.c) I think it would be
>> better to set b_size to sb->s_blocksize and be done with that. Mapping
>> code does not need anything else set to a deterministic value so using
>> memset is a bit overkill.
>
> Thanks Jan for your comments. Yes memset is an overkill. I did it just
> because other users of ext2_get_block were doing the same way. Will
> rework the patch with setting b_size as blocksize and send again.

Here is the updated patch. compile tested.

Signed-off-by : Manish Katiyar <mkatiyar@gmail.com>
---
 fs/ext2/super.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

 			return err;
@@ -1367,6 +1368,7 @@ static ssize_t ext2_quota_write(struct
super_block *sb, int type,
 				sb->s_blocksize - offset : towrite;

 		tmp_bh.b_state = 0;
+		tmp_bh.b_size = sb->s_blocksize;
 		err = ext2_get_block(inode, blk, &tmp_bh, 1);
 		if (err < 0)
 			goto out;

On Mon 26-01-09 18:18:19, Manish Katiyar wrote:
> On Mon, Jan 26, 2009 at 6:03 PM, Manish Katiyar <mkatiyar@gmail.com> wrote:
> > On Mon, Jan 26, 2009 at 5:59 PM, Jan Kara <jack@suse.cz> wrote:
> >>> On Tue, Jan 20, 2009 at 10:36 PM, Manish Katiyar <mkatiyar@gmail.com> wrote:
> >>> > ext2_quota_read doesn't bzeroes tmp_bh before calling ext2_get_block()
> >>> > where we access the b_size of it. Since it is a local variable it
> >>> > might contain some garbage. Make sure it is filled with zero before
> >>> > passing.
> >>>
> >>> Hi Ted/mingming,
> >>>
> >>> Any feedback on this ??
> >>  Ops, sorry. I wanted to reply but first I wanted to research more
> >> whether we can set b_size to 0 and then forgot about it. Looking into
> >> other code (e.g. in fs/mpage.c or fs/buffer.c) I think it would be
> >> better to set b_size to sb->s_blocksize and be done with that. Mapping
> >> code does not need anything else set to a deterministic value so using
> >> memset is a bit overkill.
> >
> > Thanks Jan for your comments. Yes memset is an overkill. I did it just
> > because other users of ext2_get_block were doing the same way. Will
> > rework the patch with setting b_size as blocksize and send again.
> 
> Here is the updated patch. compile tested.
> 
> Signed-off-by : Manish Katiyar <mkatiyar@gmail.com>
  Yes, now it looks fine. Thanks for the fix.

  Acked-by: Jan Kara <jack@suse.cz>

									Honza
> ---
>  fs/ext2/super.c |    2 ++
>  1 files changed, 2 insertions(+), 0 deletions(-)
> 
> diff --git a/fs/ext2/super.c b/fs/ext2/super.c
> index da8bdea..2a1c0c6 100644
> --- a/fs/ext2/super.c
> +++ b/fs/ext2/super.c
> @@ -1328,6 +1328,7 @@ static ssize_t ext2_quota_read(struct
> super_block *sb, int type, char *data,
>  				sb->s_blocksize - offset : toread;
> 
>  		tmp_bh.b_state = 0;
> +		tmp_bh.b_size = sb->s_blocksize;
>  		err = ext2_get_block(inode, blk, &tmp_bh, 0);
>  		if (err < 0)
>  			return err;
> @@ -1367,6 +1368,7 @@ static ssize_t ext2_quota_write(struct
> super_block *sb, int type,
>  				sb->s_blocksize - offset : towrite;
> 
>  		tmp_bh.b_state = 0;
> +		tmp_bh.b_size = sb->s_blocksize;
>  		err = ext2_get_block(inode, blk, &tmp_bh, 1);
>  		if (err < 0)
>  			goto out;
> -- 
> 1.5.4.3