Discussion:
Mandatory PFIL_HOOKS
(too old to reply)
Mindaugas Rasiukevicius
2013-06-09 01:19:00 UTC
Permalink
Hi,

Unless anyone objects, I would like to make packet filter hooks mandatory
and therefore remove PFIL_HOOKS option. They are small, often used and not
worth the #ifdef mess.

Thanks.
--
Mindaugas

--
Posted automagically by a mail2news gateway at muc.de e.V.
Please direct questions, flames, donations, etc. to news-***@muc.de
Christos Zoulas
2013-06-09 03:02:08 UTC
Permalink
Post by Mindaugas Rasiukevicius
Hi,
Unless anyone objects, I would like to make packet filter hooks mandatory
and therefore remove PFIL_HOOKS option. They are small, often used and not
worth the #ifdef mess.
I don't object :-)

christos


--
Posted automagically by a mail2news gateway at muc.de e.V.
Please direct questions, flames, donations, etc. to news-***@muc.de
Masao Uebayashi
2013-06-10 09:52:53 UTC
Permalink
Could you put the pfil hook code in if (__predict_false()) ... ? ISTR
that this made a little change when benchmarking packet forwarding.

On Sat, Jun 8, 2013 at 9:19 PM, Mindaugas Rasiukevicius
Post by Mindaugas Rasiukevicius
Hi,
Unless anyone objects, I would like to make packet filter hooks mandatory
and therefore remove PFIL_HOOKS option. They are small, often used and not
worth the #ifdef mess.
Thanks.
--
Mindaugas
--
Posted automagically by a mail2news gateway at muc.de e.V.
Please direct questions, flames, donations, etc. to news-***@muc.de
Mindaugas Rasiukevicius
2013-06-10 19:58:50 UTC
Permalink
Post by Masao Uebayashi
Could you put the pfil hook code in if (__predict_false()) ... ? ISTR
that this made a little change when benchmarking packet forwarding.
You want to avoid function call overhead? I will add a check. However,
I think the __predict_false() part is not fair. :)
--
Mindaugas

--
Posted automagically by a mail2news gateway at muc.de e.V.
Please direct questions, flames, donations, etc. to news-***@muc.de
Joerg Sonnenberger
2013-06-10 20:14:14 UTC
Permalink
Post by Mindaugas Rasiukevicius
Post by Masao Uebayashi
Could you put the pfil hook code in if (__predict_false()) ... ? ISTR
that this made a little change when benchmarking packet forwarding.
You want to avoid function call overhead? I will add a check. However,
I think the __predict_false() part is not fair. :)
In fact, it can seriously hurt. Please don't.

Joerg

--
Posted automagically by a mail2news gateway at muc.de e.V.
Please direct questions, flames, donations, etc. to news-***@muc.de
Greg Troxel
2013-06-10 20:18:06 UTC
Permalink
Post by Mindaugas Rasiukevicius
Unless anyone objects, I would like to make packet filter hooks mandatory
and therefore remove PFIL_HOOKS option. They are small, often used and not
worth the #ifdef mess.
I am not objecting, but I wonder if anyone can shed light on

what fraction of machines/kernels are in the following categories:
don't have PFIL_HOOKS because they care about performance
don't have PFIL_HOOKS and don't care
dot have PFIL_HOOKS and don't need them
do have PFIL_HOOKS and do need them

for a machine which is just forwarding, and doesn't need firewalling
or anything else, how does having PFIL_HOOKS with no actual hooks
change performance?

Loading...