ufs write safety

[David Kewley]

Dave Jones wrote on Wednesday 25 May 2005 16:54:
> On Wed, May 25, 2005 at 04:13:50PM -0700, David Kewley wrote:
>  > On Wednesday 25 May 2005 16:07, Dave Jones wrote:
>  > > On Wed, May 25, 2005 at 03:59:16PM -0700, David Kewley wrote:
>  > >  > Thanks *very* much, Dave -- it's good to hear this from someone
>  > >  > highly involved in the kernel.  I'll take your advice.
>  > >  >
>  > >  > I am using UFS and XFS in RHEL4 by rebuilding the kernel with
>  > >  > those filesystems enabled.  The filesystems appear to work fine; I
>  > >  > know others are also using XFS in RHEL4.
>  > >
>  > > beware: XFS can use *lots* of stack space in certain conditions,
>  > > which really doesn't play too nicely with the 4KB stack size.
>  >
>  > Thanks, and acknowledged.  I looked into 4k vs 8k stacks before I
>  > started using XFS on RHEL4.  My conclusion was that I'm safe because I
>  > have x86_64, which has 8k stacks.  Is that reasonable, or are stack
>  > items twice as big on 64- as on 32-bit, so that you still run a risk
>  > with 8k stacks on 64-bit?
>
> Correct, sizeof(long) and  sizeof(pointer) are doubled, so you're still at
> risk of an overflow.

Thanks again, Dave.  I've started a conversation on the linux XFS list, 
because I know this isn't RH's problem.

David