From: "Michael S. Tsirkin" <mst@mellanox.co.il>
To: Steve Wise <swise@opengridcomputing.com>
Cc: Roland Dreier <rdreier@cisco.com>,
netdev@vger.kernel.org, linux-kernel@vger.kernel.org,
openib-general@openib.org
Subject: Re: [PATCH v4 01/13] Linux RDMA Core Changes
Date: Wed, 3 Jan 2007 16:41:53 +0200 [thread overview]
Message-ID: <20070103144153.GN6019@mellanox.co.il> (raw)
In-Reply-To: <1167834348.4187.3.camel@stevo-desktop>
> > > @@ -1373,7 +1374,7 @@ int ib_peek_cq(struct ib_cq *cq, int wc_
> > > static inline int ib_req_notify_cq(struct ib_cq *cq,
> > > enum ib_cq_notify cq_notify)
> > > {
> > > - return cq->device->req_notify_cq(cq, cq_notify);
> > > + return cq->device->req_notify_cq(cq, cq_notify, NULL);
> > > }
> > >
> > > /**
> >
> > Can't say I like this adding overhead in data path operations (and note this
> > can't be optimized out). And kernel consumers work without passing it in, so it
> > hurts kernel code even for Chelsio. Granted, the cost is small here, but these
> > things do tend to add up.
> >
> > It seems all Chelsio needs is to pass in a consumer index - so, how about a new
> > entry point? Something like void set_cq_udata(struct ib_cq *cq, struct ib_udata *udata)?
> >
>
> Adding a new entry point would hurt chelsio's user mode performance if
> if then requires 2 kernel transitions to rearm the cq.
No, it won't need 2 transitions - just an extra function call,
so it won't hurt performance - it would improve performance.
ib_uverbs_req_notify_cq would call
ib_uverbs_req_notify_cq()
{
ib_set_cq_udata(cq, udata)
ib_req_notify_cq(cq, cmd.solicited_only ?
IB_CQ_SOLICITED : IB_CQ_NEXT_COMP);
}
This way kernel consumers don't incur any overhead,
and in userspace users extra function call is dwarfed
by system call overhead.
> Passing in user data is sort of SOP for these sorts of verbs.
I don't see other examples. Where we did pass extra user data
is in non-data pass verbs such as create QP.
This is most inner tight loop in many ULPs, so we should be very careful
about adding code there - these things do add up.
See recent IRQ API update in kernel.
> How much does passing one more param cost for kernel users?
Donnu. I just reviewed the code.
It really should be up to patch submitter to check the performance
effect of his patch, if there might be any.
--
MST
next prev parent reply other threads:[~2007-01-03 14:43 UTC|newest]
Thread overview: 34+ messages / expand[flat|nested] mbox.gz Atom feed top
2006-12-14 13:52 [PATCH v4 00/13] 2.6.20 Chelsio T3 RDMA Driver Steve Wise
2006-12-14 13:53 ` [PATCH v4 01/13] Linux RDMA Core Changes Steve Wise
2006-12-24 8:49 ` Michael S. Tsirkin
2007-01-03 14:25 ` Steve Wise
2007-01-03 14:41 ` Michael S. Tsirkin [this message]
2007-01-03 14:56 ` Steve Wise
2007-01-03 15:00 ` Michael S. Tsirkin
2007-01-03 15:07 ` Steve Wise
2007-01-03 15:18 ` Michael S. Tsirkin
2007-01-03 19:17 ` Steve Wise
2007-01-03 19:33 ` Michael S. Tsirkin
2007-01-03 20:20 ` Steve Wise
2007-01-03 21:22 ` [openib-general] " Steve Wise
2007-01-04 5:07 ` Michael S. Tsirkin
2007-01-04 14:07 ` Steve Wise
2007-01-04 21:34 ` [openib-general] " Roland Dreier
2007-01-04 21:49 ` Steve Wise
2007-01-05 14:22 ` Steve Wise
2007-01-03 15:02 ` Michael S. Tsirkin
2007-01-03 15:06 ` Steve Wise
2007-01-03 15:10 ` Michael S. Tsirkin
2006-12-14 13:53 ` [PATCH v4 02/13] Device Discovery and ULLD Linkage Steve Wise
2006-12-14 13:54 ` [PATCH v4 03/13] Provider Methods and Data Structures Steve Wise
2006-12-14 13:54 ` [PATCH v4 04/13] Connection Manager Steve Wise
2006-12-14 13:55 ` [PATCH v4 05/13] Queue Pairs Steve Wise
2006-12-14 13:55 ` [PATCH v4 06/13] Completion Queues Steve Wise
2006-12-14 13:56 ` [PATCH v4 07/13] Async Event Handler Steve Wise
2006-12-14 13:56 ` [PATCH v4 08/13] Memory Registration Steve Wise
2006-12-14 13:57 ` [PATCH v4 09/13] Core WQE/CQE Types Steve Wise
2006-12-14 13:57 ` [PATCH v4 10/13] Core HAL Steve Wise
2006-12-14 13:58 ` [PATCH v4 11/13] Core Resource Allocation Steve Wise
2006-12-14 13:58 ` [PATCH v4 12/13] Core Debug functions Steve Wise
2006-12-14 13:59 ` [PATCH v4 13/13] Kconfig/Makefile Steve Wise
2007-01-05 17:32 [openib-general] [PATCH v4 01/13] Linux RDMA Core Changes Felix Marti
2007-01-06 17:34 ` Michael S. Tsirkin
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20070103144153.GN6019@mellanox.co.il \
--to=mst@mellanox.co.il \
--cc=linux-kernel@vger.kernel.org \
--cc=netdev@vger.kernel.org \
--cc=openib-general@openib.org \
--cc=rdreier@cisco.com \
--cc=swise@opengridcomputing.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).