xfs
[Top] [All Lists]

Re: CVS 2.3 tree

To: Andi Kleen <ak@xxxxxxx>, linux-xfs@xxxxxxxxxxx
Subject: Re: CVS 2.3 tree
From: Russell Cattelan <cattelan@xxxxxxxxxxx>
Date: Fri, 02 Jun 2000 14:09:52 -0500
References: <3936AC63.E38A4E6F@thebarn.com> <20000601225718.A12013@gruyere.muc.suse.de> <3936D18D.174640C@thebarn.com> <20000602175713.A25566@gruyere.muc.suse.de>
Sender: owner-linux-xfs@xxxxxxxxxxx
Andi Kleen wrote:

> On Thu, Jun 01, 2000 at 04:11:41PM -0500, Russell Cattelan wrote:
> > >
> > > to do that for the ``TAKE'' messages on the list? Then you could just
> > > click on a change to look at it.
> > >
> > > -Andi
> >
> > BTW did switching compilers help?
>
> Yes, it seems switching compilers did help.
>
> Do you know which function is miscompiled ? [it may be useful to write a gcc
> bugreport]
>
> -Andi

Look at about line 3826 in the function
__pb_set_blocks_mp

 bh->b_blocknr = (long)mp->pbm_bn +
   (mp->pbm_delta >> inode->i_sb->s_blocksize_bits);

This line -->   bh->b_blocknr >>= block_bits;

Putting printk's in will show that b_blocknr to be correct.
Taking them out and looking at the values with kdb will show the wrong values.

One way arond this is to call a function that does the shift.

I didn't spend to much time looking at the asm code to try and understand what
is going wrong.
If  you spot something obvious let me know.


-Russell


<Prev in Thread] Current Thread [Next in Thread>