Re: Linux 2.4.10-pre11

[Andrea Arcangeli <andrea@suse.de>]

On Tue, Sep 18, 2001 at 05:31:48AM -0400, Alexander Viro wrote:
> 
> 
> On Mon, 17 Sep 2001, Linus Torvalds wrote:
> 
> > This also merges the blkdev in page cache patch, and that will hopefully
> > make it noticeably easier to do the "do bread() with page cache too", at
> > which point a lot of the current ugly synchronization issues will go away.
> 
> Umm...  Linus, had you actually read through the fs/block_device.c part
> of that?  It's not just ugly as hell, it's (AFAICS) not hard to oops
> if you have several inodes sharing major:minor.  ->bd_inode and its

can you show an exploit? I cannot reproduce any problem here:

root@athlon:/tmp > cp -a /dev/hda hda.1
root@athlon:/tmp > cp -a /dev/hda hda.2
root@athlon:/tmp > cp hda.1 /dev/null & cp hda.2 /dev/null &
[1] 24981
[2] 24982
root@athlon:/tmp > fg
cp hda.2 /dev/null

root@athlon:/tmp > fg
cp hda.1 /dev/null

root@athlon:/tmp > 

> treatment are bogus.  Please, read it through and consider reverting -
> in its current state code is an ugly mess.

what other design solution do you propose rather both inodes sharing the
i_mapping across the different inodes like I did?

I found more handy to just bump the i_count of the first inode and
referencing it from the bd_inode, rather than dynamically allocating the
i_mapping and have a bd_mapping, but if you prefer to dynamically
allocate the i_mapping rather than using the i_data of the fist inode we
can change that of course. Not sure what's the mess in the patch you're
talking about, could you elaborate?

Andrea