Re: [Xen-changelog] Add a /proc interface for setting rx-buffer allocation policy in

Re: [Xen-changelog] Add a /proc interface for setting rx-buffer allocation policy in

[Gerd Knorr]

BitKeeper Bot <riel@surriel.com> writes:

> ChangeSet 1.1159.256.110, 2005/05/03 11:54:00+01:00, kaf24@firebug.cl.cam.ac.uk
> 
> 	Add a /proc interface for setting rx-buffer allocation policy in
> 	the netfront driver.
> 	e.g., 'echo 256 >/proc/xen/net/eth0/rxbuf_min' will avoid unnecessary
> 	packet drops by always allocating the maximum possible number of
> 	buffers at all times, at the expense of extra memory usage when the
> 	interface is idle.

Be prepared to be flamed for that on lkml, using /proc for stuff like
that (i.e. driver specifix settings) is depricated these days.  It
should go to sysfs instead.

I can come up with a patch for that, might take some days though as
I'm busy with some non-xen stuff at the moment.

cheers,

  Gerd

btw: can whoever runs the changelog list (rik?) please add the
repository (2.0, unstable, ...) to to subject line please?  And while
you are at it: -p1 for the diffstat would be cool as well, especially
for one-file commits, there are -- for example -- plenty of page.h
files in the tree.  Thanks.

-- 
#define printk(args...) fprintf(stderr, ## args)

Re: Re: [Xen-changelog] Add a /proc interface for setting rx-buffer allocation policy in

[Keir Fraser]

On 3 May 2005, at 12:31, Gerd Knorr wrote:

> Be prepared to be flamed for that on lkml, using /proc for stuff like
> that (i.e. driver specifix settings) is depricated these days.  It
> should go to sysfs instead.
>
> I can come up with a patch for that, might take some days though as
> I'm busy with some non-xen stuff at the moment.

This should happen when xenbus is implemented -- it can create a 
virtual bus in sysfs and then virtual device info and parameters can 
hang off of that. I think someone from Redhat (possibly Jeremy Katz) 
was going to look into implementing xenbus.

For the time being I don't see any reason for them not to hang out in 
/proc. We probably want xenbus before going into mainline anyway.

  -- Keir