[ENBD] (no subject)

[Peter T. Breuer]

"A month of sundays ago HUDZIA Bertrand wrote:"
> On Fri, 19 Apr 2002 11:47:15 +0200 (MET DST)
> "Peter T. Breuer" <ptb@it.uc3m.es> wrote:
> 
> > "A month of sundays ago HUDZIA Bertrand wrote:"
> > > I still manage to crash my system even with aa's kernel patch.
> > 
> > Turn off the "sync" limit in bdflush, or get more memory! 
> > 
> Does a 
> 	update -s 0 
> Make it ? (it set 0 as time between flush)

I don't know what update does. Oh ... it's one of the old daemons isn't
it? I didn't think it was still in use. I'm fairly sure it isn't. I
think it may start a kernel thread ... kupdated.

No, I just mean you to change the entries in /proc/sys/vm/bdflush
even further, to something like 5/99 insted of the habitual 40/60 or
even the 25/80 that I think the client daemon changes it to.

> > it appears that ioctls were perhaps changing around during the 2.4.*
> > development.
> > 
> > 
> Ok, I've notice that in the source, there is a link :
> 
> 	linux -> linux-2.2.10
> 
> that is done when i "make" enbd.
> Is it normal that it always point to linux-2.2.10, with every kernel source
>  tree i gave in the Makefile ?? (2.4.9, 2.4.18 or 2.4.19-pre6-aa1)

Yes. It's just a historical accident.

> The linux-2.4.18 tree isn't used yet ?

It's used, but it's ONLY for the 2.4.18 kernel (or thereabouts). I.e.
it doesn't contain the accumulated dirtiness for supporting older (or
newer) kernels.

> I've modify the nbd/nbd-client and the linux-2.2.10/drivers/block/nbd.c
> (nothing to change in this one) as you told me but i still have the same
> errors :
> 
> 	nbd-client   763: setdeviceblksize set blksize (1024) failed in kernel

No, you can't have the same error. It can't fail. You can see that from
the code you inserted! If you are interested in seeing why, activate
some of the debug statements (compile with DEBUG=1 or higher).

> 	nbd-client   763: setkernel Failed set blksize 1024 on fd 4: Invalid argument

If you like, I'll boot kernel 2.4.9 and have a look.

Peter