xfs
[Top] [All Lists]

Re: Linux 2.4.4

To: linux-xfs@xxxxxxxxxxx
Subject: Re: Linux 2.4.4
From: Russell Cattelan <cattelan@xxxxxxxxxxx>
Date: Mon, 30 Apr 2001 16:10:20 -0400
References: <4.3.2.7.2.20010428140403.04271ec0@xxxxxxxxxxxxx> <3AEAE40A.8E940486@xxxxxxxxxxx> <4.3.2.7.2.20010428221754.04278bd8@xxxxxxxxxxxxx> <20010428230127.A1753@xxxxxxxxxxxxxxxxxxx> <20010430181324.A22446@xxxxxxxxxxxxxxxxxx> <3AED9741.57322D14@xxxxxxxxxxx>
Sender: owner-linux-xfs@xxxxxxxxxxx
Russell Cattelan wrote:

>
>
> >
> >  Yes, ofcourse we know we have to pay for living on the bleeding edge.
> > Anyway, is XFS backported to 2.2.x? If not, then it is not a good
> > solution for me, as it means I can not access my /home partition any
> > more, guess why.
> >
> > I think Russel said something that it is compiling. So, what are the
> > experience with it? Will the CVS tree bumped to it somewhen?
>
> It's compiled and booted but it t'aint running very well.
> I've managed to get an XFS fs mounted but it crashed shortly after
> running dbench 128.
>

Ok after thinking about the problem for a bit I realized what happened.
The kiobuf allocation routines switched from using slab allocated memory to vm
allocated
memory... (I'm not sure why this was done) but since XFS calls alloc_kiobuf from
an interrupt
thread, and since vmalloc cannot be called from and interrupt thread it was
deadlocking a  processor.

I've reverted the kiobuf allocation to using slab allocated memory and things
seem to be running
just fine.
I need to look at this some more since kiobufs now have buffer_heads allocated
along with the kiobuf
since XFS does not use these bh's it's wasting slab memory.

If anybody really really wants to try 2.4.4 +xfs let me know; I'll generate a
quick patch.


>
> --
> Russell Cattelan
> --
> Digital Elves inc. -- Currently on loan to SGI
> Linux XFS core developer.

--
Russell Cattelan
--
Digital Elves inc. -- Currently on loan to SGI
Linux XFS core developer.




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