I have a basic question about the CM. It looks like the gen2 CM doesn't implement the CM state machine as defined in the IB spec. It doesn't perform retransmissions, handle timeouts, etc. Is the current CM API intended as the final API, or is this just an intermediate step on the way to implementin
On Tue, Mar 15, 2005 at 02:25:50PM -0800, Woodruff, Robert J wrote:
>
> Roland> I just spent a little time creating a new "ibv" module for
> NetPIPE
> >that runs on top of the userspace verbs I've been developing on the
> >roland-uverbs branch.
>
> Cool. this will be very useful. Any idea if/wh
Quoting r. Roland Dreier <[EMAIL PROTECTED]>:
> Subject: Re: [openib-general] [PATCH] set lkey in mthca mpt entry
>
> Michael> lkey does not seem to be set in the mpt entry. does this
> Michael> look right?
>
> You would know better but my docs say that the lkey field should be
> set to 0
Thanks, applied.
- R.
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman/listinfo/openib-general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
On Tue, 2005-03-15 at 18:11, Bob Woodruff wrote:
> Any idea when the user-mode CM support will show up ?
I think it should be there in about a couple of weeks.
-- Hal
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman/
>It might be reasonable to wait until the APIs are a little more frozen
>and the support has landed on the OpenIB trunk (as I said, userspace
>verbs are still only on the roland-uverbs branch). I would estimate a
>time frame on the order of weeks for that to happen.
> - R.
Good point, probably
Robert> Cool. this will be very useful. Any idea if/when the
Robert> netpipe folks will release a version of netpipe that has
Robert> this patch included ?
That's up to the netpipe folks. Posting the patch was the first
contact I've made beyond downloading the source yesterday.
It mi
Roland> I just spent a little time creating a new "ibv" module for
NetPIPE
>that runs on top of the userspace verbs I've been developing on the
>roland-uverbs branch.
Cool. this will be very useful. Any idea if/when the netpipe folks will
release a version of netpipe that has this patch include
Hi Nitin,
On Tue, 2005-03-15 at 16:15, Nitin Hande wrote:
> This is cool, I have got Solaris IPoIB happily working with the
> OpenSM now. It plumbs, pings and snoops on 0x pkey.
Great. That's good news. I'll work on a real fix for this now.
> On other hand, on my linux node, if I try to use
I just spent a little time creating a new "ibv" module for NetPIPE
that runs on top of the userspace verbs I've been developing on the
roland-uverbs branch. This is pretty much a straight port of the
current Mellanox VAPI "ib" module, with the main changes coming from
the fact that OpenIB doesn't
Nitin> On other hand, on my linux node, if I try to use 8001
Nitin> partition and configure IB interface with IP addr (same
Nitin> time while ib0 is using 0x pkey), I get the following
Nitin> error, you may want to investigate that
I think this is probably an OpenSM issue (
Hal,
On Fri, 2005-03-04 at 12:53, Hal Rosenstock wrote:
> Hi again Nitin,
>
> Finally got a chance to work on this. I have a workaround for you for
> now. Real patch later... Let me know if this does the trick for you. It
> did for me.
>
> -- Hal
>
> Index: osm_sa_mcmember_record.c
> ==
Using in_atomic() to decide between GFP_KERNEL and GFP_ATOMIC is not
safe (it doesn't work if CONFIG_PREEMPT=n). Change to just always
allocating with GFP_ATOMIC, since we don't know if we can sleep or not.
Signed-off-by: Roland Dreier <[EMAIL PROTECTED]>
--- linux-export.orig/drivers/infiniban
Hal Rosenstock wrote:
On Mon, 2005-03-14 at 19:22, Sean Hefty wrote:
This patch updates the MAD API to help provide support for the RMPP
implementation and clients. Notable changes:
Wouldn't this change also impact ib_user_mad.h and user_mad.c ?
I don't think that they effect those files directl
Michael> Hi, Roland! Looking at uverbs kernel module, I notice
Michael> that in some instances it passes some parameters from
Michael> userspace directly to ib core, without verifying their
Michael> sanity.
Michael> One example of this is qp attributes in create and modify
On Mon, 2005-03-14 at 19:22, Sean Hefty wrote:
> This patch updates the MAD API to help provide support for the RMPP
> implementation and clients. Notable changes:
Wouldn't this change also impact ib_user_mad.h and user_mad.c ?
-- Hal
___
openib-gene
At 05:35 PM 3/14/2005, Caitlin Bestler wrote:
> -Original Message-
> From: Troy Benjegerdes
[
mailto:[EMAIL PROTECTED]]
> Sent: Monday, March 14, 2005 5:06 PM
> To: Caitlin Bestler
> Cc: openib-general@openib.org
> Subject: Re: [openib-general] Getting rid of pinned memory
requirement
On Fri, 2005-03-11 at 19:15 -0600, Troy Benjegerdes wrote:
> I have in my office a shiny new kernel.org 2.6.11 64 bit kernel running
> on my Mac G5, with the drivers/infiniband modules loaded.
>
> What do I need to do to verify this all works?
Do you have the IB card plugged into an IB switch? I
Hal> Hi Roland, Just ran across this reminder:
Hal> Should user_mad.c be updated for the following: /* XXX remove
Hal> once 2.6.11 is released */
Yep, I'd apply that patch for sure.
- R.
___
openib-general mailing list
openib-general@openi
Michael> lkey does not seem to be set in the mpt entry. does this
Michael> look right?
You would know better but my docs say that the lkey field should be
set to 0 for SW2HW_MPT and is only used to refer to the original
region for memory windows.
- R.
lkey does not seem to be set in the mpt entry. does this look right?
Signed-off-by: Michael S. Tsirkin <[EMAIL PROTECTED]>
Index: hw/mthca/mthca_mr.c
===
--- hw/mthca/mthca_mr.c (revision 1983)
+++ hw/mthca/mthca_mr.c (working copy)
I have updated mstflint in the openib repository.
revision 1990 fixes a crash and cleans up progress reporting
in flash error recovery process.
Tested on x86/ia64/i686.
--
MST - Michael S. Tsirkin
___
openib-general mailing list
openib-general@openib.or
On Tue, Mar 15, 2005 at 08:23:52AM +0200, Michael S. Tsirkin wrote:
> Hi, Roland!
> Looking at uverbs kernel module, I notice that in some instances
> it passes some parameters from userspace directly to ib core, without
> verifying their sanity.
>
> One example of this is qp attributes in create
Hi Roland,
Just ran across this reminder:
Should user_mad.c be updated for the following:
/* XXX remove once 2.6.11 is released */
Thanks.
-- Hal
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman/listinfo/openib-gene
Hi, Does openSM supports non-fat tree (irregular such
as graph) topologies? Abhijeet
[shaharf]
Yes. OpenSM supports any type of mesh (connected graph).
___
openib-general mailing list
openib-general@openib.org
http://open
25 matches
Mail list logo