Sean> Hmm... it just occurred to me that I don't know who
Sean> originally submitted these patches. I would guess you and
Sean> Hal based on the $Id.
This diff between svn and the kernel is actually because svn is
missing some changes. Your patch partially reverts the following
diff
On 3/3/06, Bryan O'Sullivan <[EMAIL PROTECTED]> wrote:
>
> The proviso here is that I'm just not going to have time to build binary
> kernel RPMs, so any kernel bits that I do package will have to build
> cleanly as modules against unpatched kernels. I know SDP does; I don't
> know about iSER.
>
On Sat, Mar 04, 2006 at 10:24:28AM +0530, Damaru wrote:
> May I know what is OpenSM and why is it needed ?
https://openib.org/tiki/tiki-index.php?page=OpenSM
grant
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman/listi
Add support for sending and receiving large RMPP transfers.
Current code supports transfers only as large as a single memory allocation.
This patch uses linked list of memory buffers when sending and receiving
data to avoid needing contiguous pages for larger transfers.
Signed-off-by: Sean Hefty <
Fix synchronization around accessing event list.
Signed-off-by: Sean Hefty <[EMAIL PROTECTED]>
---
drivers/infiniband/core/ucm.c |2 ++
1 files changed, 2 insertions(+), 0 deletions(-)
2b2a9f64d1bdba30894c575641c85ec798adca1c
diff --git a/drivers/infiniband/core/ucm.c b/drivers/infiniband/
Hmm... it just occurred to me that I don't know who originally submitted these
patches.
I would guess you and Hal based on the $Id.
- Sean
---
drivers/infiniband/core/agent.c |5 +
drivers/infiniband/core/ud_header.c |3 +--
drivers/infiniband/core/verbs.c |3 +--
3 fil
>If you have changes that you would like to get into 2.6.17, the
>for-2.6.17 branch is the base that makes it easiest for me to merge
>patches into my queue. That branch is what I will ask Linus to pull
>from when 2.6.17 opens up.
I did a diff between your for-2.6.17 branch and svn. The differen
Hal,
On 03 Mar 2006 12:07:16 -0500, Hal Rosenstock <[EMAIL PROTECTED]> wrote:
> [ snip ]
> and use IPoIB underneath iSCSI. Is there an open source iSCSI target ?
>
> Once there is an open source iSER target, you will be able to do this
> with iSER too.
I read an old anouncement where, I think, i
Fabian,
On 3/3/06, Fabian Tillier <[EMAIL PROTECTED]> wrote:
> On 3/3/06, Damaru <[EMAIL PROTECTED]> wrote:
> > Hello All,
> >
> > I am planning to setup an iSCSI target and an iSCSI initiator via
> > infiniband in point to point mode. is it possible with two linux
> > machines to connect in p2p
On Fri, 2006-03-03 at 21:35, Hal Rosenstock wrote:
> On Fri, 2006-03-03 at 21:28, Jean-Christophe Hugly wrote:
> > I have not started to look for faulty mutexes, yet. Where the fixes
> > recently proposed in that area committed as of 5594 ?
>
> No. The fixes were related to IPC (signals) and anoth
On Fri, 2006-03-03 at 21:28, Jean-Christophe Hugly wrote:
> I have not started to look for faulty mutexes, yet. Where the fixes
> recently proposed in that area committed as of 5594 ?
No. The fixes were related to IPC (signals) and another on atomics which
had a faulty mutex in it. If you could ap
Hi Guys,
I have been having trouble with gen2's osm for a while. I finally
isolated the faulty behaviour to one easy test case:
run osm somewhere.
then one whatever workstation has an HCA connected to the same subnet,
do this:
i=1
while true; do
modprobe -r ib_mthca
sleep 3
From: Roland Dreier <[EMAIL PROTECTED]>
Date: Wed, 01 Feb 2006 17:28:10 -0800
> Dave, if you want to merge this directly, that's fine. Or I'm fine
> with merging this through the IB tree if you'd prefer (if you want me
> to do that, let me know if you think it's 2.6.16 material).
Applied to net-
Linus, please pull from
master.kernel.org:/pub/scm/linux/kernel/git/roland/infiniband.git for-linus
This tree is also available from kernel.org mirrors at:
git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband.git
for-linus
The pull will get the following change, which fixes
>I'm getting the following error when trying to compile svn 5606 with a
>2.6.16-rc5 kernel.
Looks like you're picking up the standard 2.6.16-rc5 include files.
- Sean
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman/
I'm getting the following error when trying to compile svn 5606 with a
2.6.16-rc5 kernel.
CC [M] drivers/infiniband/core/addr.o
In file included from drivers/infiniband/core/addr.c:43:
drivers/infiniband/include/rdma/ib_addr.h:45: error: field `dev_type'
has incomplete type
drivers/infiniband/
Sean> I can resubmit the patches if necessary.
Yes, can you please send out the patches again, with descriptive
subjects for each patch and with the ip_dev_find() re-export split
into its own patch? That would make it easier for me to pull into my
git tree.
Thanks,
Roland
_
Grant> IA64 .config options allows 4k, 8K, 16K, 64K. Default page
Grant> size is 16k. SLES9 versions for SGI/Altix use 64K as the
Grant> default.
Yes, but fortunately no ia64 machines have an IBM GX bus, so you don't
have to worry about the ehca driver ;)
- R.
_
Bob> I looked at this briefly yesterday and it looks like the
Bob> Scsi_Host struct is in the srp_target_port structure in a
Bob> member called scsi_host, so if I read it right then this
Bob> patch would fix the problem. However, if Roland or one of the
Bob> SRP experts could re
Hi Yael,
On Thu, 2006-02-23 at 03:13, Yael Kalka wrote:
> Hi Hal,
>
> I have done some work on osmtest/osmt_multicast.c.
> There was a problem that in the middle of the test and in the end
> there were checks that the only multicast groups that exist are the
> ones discovered in the begining of
I wanted to make doubly sure that this didn't get lost in the patch series, but
ip_dev_find() is re-exported. The use is shown below.
- Sean
>+int rdma_translate_ip(struct sockaddr *addr, struct rdma_dev_addr *dev_addr)
>+{
>+ struct net_device *dev;
>+ u32 ip = ((struct sockaddr_in *)
OpenIB Members
OpenIB and the InfiniBand Trade
Association are proud to jointly sponsor the upcoming InfiniBand Plugfest to be
held March 13 - 17 at the University
of New Hampshire Interoperability Lab
in Durham, NH.
Information on event registration is
available at http://openib.or
On Fri, 3 Mar 2006, Caitlin Bestler wrote:
> Does anyone have a scenario where an application would rely on
> hardware based end to end credits for some QPs and not others
> running in the same application?
If it is a module parameter, it is a system wide configuration, not
per application.
I
[EMAIL PROTECTED] wrote:
> Hello!
> As was recently discussed on this list, infiniband implements
> an optional hardware end to end credits mechanism, with
> credits encoded in the AETH field in an ACK packet.
>
> The result is that an application might see performance
> degradation even if it alw
On Thu, 2 Mar 2006, Michael S. Tsirkin wrote:
mst> I am, therefore, considering one of the following options:
mst> - Add a flag rq_e2e_flow_control to create qp, make e2e flow control
mst> disabled by default
mst> - Add a flag disable_rq_e2e_flow_control to create qp, make e2e flwo
control
ms
On Fri, Mar 03, 2006 at 07:57:24AM -0800, Roland Dreier wrote:
> Heiko> Hello Christoph, I only did kmalloc(PAGE_SIZE, GFP_KERNEL)
> Heiko> because we need some amount of memory which is 4k
> Heiko> aligned. I'm not sure, but I think if I use kmalloc(SIZE,
> Heiko> GFP_KERNEL) and S
On Fri, 2006-03-03 at 09:39 -0800, Sean Hefty wrote:
> Having parts of release 1.0 be true release quality, and other parts of it be
> alpha/beta quality is confusing. There's nothing to prevent a user from
> getting
> SDP or driver X directly from svn if they want to run some sort of technolo
Bryan O'Sullivan wrote:
Basically, I am reluctant to completely drop things like SDP support and
drivers that are not in upstream kernels. In every one of these cases,
I'd prefer to make the sources available with the caveat that they are
"technology previews" where binary compatibility with fut
Using this setup, you'll be getting IPoIB performance out of the iSCSI P2P.
While decent, it is nowhere near the performance you can get by really
using IB capabilities, namely RDMA.
With IPOIB the itarget initiator will still do all the IP processing
as well as extra data copies.
Dan
On 03 Mar
Hi Heiko,
On Fri, 2006-03-03 at 09:00, Heiko J Schick wrote:
> Hallo Hal,
>
> we can only modify a QP when the port is active. The port goes
> active when you create an AQP1
You are referring to how the eHCA works here, right ?
> and the communication with the
> SM and the firmware was complet
On Fri, 2006-03-03 at 12:03, Fabian Tillier wrote:
> On 3/3/06, Damaru <[EMAIL PROTECTED]> wrote:
> > Hello All,
> >
> > I am planning to setup an iSCSI target and an iSCSI initiator via
> > infiniband in point to point mode. is it possible with two linux
> > machines to connect in p2p ? what hardw
Ira wrote,
>scsi_host.h:
> void scsi_scan_target(struct Scsi_Host *shost, unsigned int channel,
> unsigned int id, unsigned int lun, int rescan)
>Indeed it looks like shost_gendev (struct device) is wrong.
>Did I miss something?
>Thanks,
>Ira
I looked at this briefly y
On 3/3/06, Damaru <[EMAIL PROTECTED]> wrote:
> Hello All,
>
> I am planning to setup an iSCSI target and an iSCSI initiator via
> infiniband in point to point mode. is it possible with two linux
> machines to connect in p2p ? what hardware and software do I need?
Yes, it is possible. You'll need
Heiko> Hello Christoph, I only did kmalloc(PAGE_SIZE, GFP_KERNEL)
Heiko> because we need some amount of memory which is 4k
Heiko> aligned. I'm not sure, but I think if I use kmalloc(SIZE,
Heiko> GFP_KERNEL) and SIZE is smaller as 4k the memory will not
Heiko> be 4k aligned?
If
On Thu, 2006-03-02 at 09:47 -0800, Sean Hefty wrote:
> >Since kernel level components are released through normal kernel.org
> >mainstream
> >process, this directory linux-kernel/ should be removed to avoid he chance of
> >confusion.
>
> I think that there's disagreement on releasing kernel compo
On Thu, 2006-03-02 at 18:31 -0600, Steve Wise wrote:
> Do we want to setup components for iWARP core and the amso driver? Or
> at least a component for the iwarp branch? If so, I'll use this to
> track bugs and fixes for the iwarp stuff...
>
> Tom or I can be owners of these components.
Sure.
On Fri, 2006-03-03 at 10:01, Devesh Sharma wrote:
> why It will not OK in future, Do you have some point about that?
It depends on how many multicast LIDs a switch supports. The
architecture allows for 16K - 1 which is larger than 255. A switch can
implement any number up to that.
-- Hal
>
> On
why It will not OK in future, Do you have some point about that?On 03 Mar 2006 08:27:49 -0500, Hal Rosenstock <[EMAIL PROTECTED]
> wrote:On Fri, 2006-03-03 at 07:55, Devesh Sharma wrote:> Hi Hal,> Thanks for replying.
>> On 03 Mar 2006 06:29:50 -0500, Hal Rosenstock <[EMAIL PROTECTED]>> wrote:>
Hello Hal!
Yes, will fix that. Good catch. Thanks!
Mit freundlichen Gruessen/Kind Regards
Hoang-Nam Nguyen
Hal Rosenstock
<[EMAIL PROTECTED]
On Fri, 2006-03-03 at 09:13, Heiko J Schick wrote:
> Hello Hal,
>
> we don't have a fw_ver which is available for our adapter.
> The be compatible with the OpenIB stack we copy the hw_ver
> into the fw_her. The hw_ver alredy has a device file.
So there is no way to get the real firmware version ?
Hello Christoph,
I only did kmalloc(PAGE_SIZE, GFP_KERNEL) because we need
some amount of memory which is 4k aligned. I'm not sure,
but I think if I use kmalloc(SIZE, GFP_KERNEL) and SIZE
is smaller as 4k the memory will not be 4k aligned?
If that is the case we can change kmalloc(PAGE_SIZE, ...
Hello Hal,
we don't have a fw_ver which is available for our adapter.
The be compatible with the OpenIB stack we copy the hw_ver
into the fw_her. The hw_ver alredy has a device file.
If it is necessary, I can add a file for the fw_ver too, but
this will include the same data as is in the hw_ver
Hallo Hal,
we can only modify a QP when the port is active. The port goes
active when you create an AQP1 and the communication with the
SM and the firmware was completed. This can take a while.
We've included the wait, because ib_mad creates an AQP1. After
the AQP1 was created sucessfully, ib_ma
I am trying to use VAPI in kernel space and have a question about the registration of physical memory."11.2.8.3 REGISTER PHYSICAL MEMORY REGION" of IB spec. 1.2 saysActual size of the PBL resources allocated. The actual size of the
PBL resources allocated shall be greater than or equal to the sizeo
On Fri, 2006-03-03 at 07:55, Devesh Sharma wrote:
> Hi Hal,
> Thanks for replying.
>
> On 03 Mar 2006 06:29:50 -0500, Hal Rosenstock <[EMAIL PROTECTED]>
> wrote:
> Hi Devesh,
>
> On Thu, 2006-03-02 at 23:31, Devesh Sharma wrote:
> > On 02 Mar 2006 08:35:04 -0500, H
On Thu, 2006-03-02 at 03:38, Heiko J Schick wrote:
> -- linux-2.6.16-rc4-orig/drivers/infiniband/hw/ehca/ehca_sqp.c
> 1970-01-01 01:00:00.0 +0100
> +++ linux-2.6.16-rc4/drivers/infiniband/hw/ehca/ehca_sqp.c2006-03-01
> 09:23:12.0 +0100
> +extern int ehca_port_act_time
On Thu, 2006-03-02 at 03:37, Heiko J Schick wrote:
> --- linux-2.6.16-rc4-orig/drivers/infiniband/hw/ehca/ehca_hca.c
> 1970-01-01 01:00:00.0 +0100
> +++ linux-2.6.16-rc4/drivers/infiniband/hw/ehca/ehca_hca.c2006-02-28
> 08:41:04.0 +0100
> @@ -0,0 +1,317 @@
> +/*
> + *
On Thu, 2006-03-02 at 03:37, Heiko J Schick wrote:
> --- linux-2.6.16-rc4-orig/drivers/infiniband/hw/ehca/ehca_mcast.c
> 1970-01-01 01:00:00.0 +0100
> +++ linux-2.6.16-rc4/drivers/infiniband/hw/ehca/ehca_mcast.c 2006-02-28
> 12:41:24.0 +0100
> +#define MAX_MC_LID 0xFFFE
> +
Hi Hal,
Thanks for replying.On 03 Mar 2006 06:29:50 -0500, Hal Rosenstock <[EMAIL PROTECTED]> wrote:
Hi Devesh,On Thu, 2006-03-02 at 23:31, Devesh Sharma wrote:> On 02 Mar 2006 08:35:04 -0500, Hal Rosenstock <[EMAIL PROTECTED]>> wrote:> Hi Devesh,
>> On Thu, 2006-03-02 at 08:03, Dev
Hello All,
I am planning to setup an iSCSI target and an iSCSI initiator via
infiniband in point to point mode. is it possible with two linux
machines to connect in p2p ? what hardware and software do I need?
Thanks in advance,
Damaru
___
openib-general
Hi Devesh,
On Thu, 2006-03-02 at 23:31, Devesh Sharma wrote:
> On 02 Mar 2006 08:35:04 -0500, Hal Rosenstock <[EMAIL PROTECTED]>
> wrote:
> Hi Devesh,
>
> On Thu, 2006-03-02 at 08:03, Devesh Sharma wrote:
> > Hi All,
> > I have another query regarding Opens
> > + rblock = kmalloc(PAGE_SIZE, GFP_KERNEL);
> > + if (rblock == NULL) {
> > + EDEB_ERR(4, "Cannot allocate rblock memory.");
> > + ret = -ENOMEM;
> > + goto num_ports0;
> > + }
> > +
> > + memset(rblock, 0, PAGE_SIZE);
>
> Use kzalloc instead (this appears
52 matches
Mail list logo
