Hi Todd,
It is LWG. MgtWG will also be involved.
>
> I am a member of IBTA however I have not noticed this discussion on
the IBTA
> working groups. Which working group have you engaged with this
proposal?
>
> Todd Rimmer
___
openib-general mailing li
On Tue, 2006-05-30 at 18:05, Rimmer, Todd wrote:
> > Eitan Wrote:
> > > As Roland suggest, before implementing a non-standard approach, IBTA
> > should be
> > > engaged to define an appropriate extension to the standard. Such
> > extensions would
> > > need to be carefully defined to avoid breakin
On Tue, May 30, 2006 at 10:09:36PM +0300, Sasha Khapyorsky wrote:
> > XML style syntax is provided for the policy file.
>
> Why XML? It is not too much readable and writable (by human) format.
It is human readable and very portable.
An example is here:
http://svn.gnumonks.org/trunk/mmio_t
>On Tue, 2006-05-30 at 17:33, Sean Hefty wrote:
>> Hal Rosenstock wrote:
>> > Is client reregister handled properly by the multicast module ?
>>
>> Can you clarify what you mean by this? Are you asking about re-sending join
>> requests based on some event?
>
>Yes; when the SM sends a Set PortInfo
Hi,
M E R ? D i A
A M B / E N
P R O Z ^ C
V ? A G R A
C ? A L i S
L E V ? T R A
S O M ^
X ^ N A X
V A L / U M
http://www.roflasikasumon.com
Suddenly on the path ahead appeared some white deer, a hind and fawns as snowy white as the hart had been dark. They glimmered in the shadows.
> Eitan Wrote:
> > As Roland suggest, before implementing a non-standard approach, IBTA
> should be
> > engaged to define an appropriate extension to the standard. Such
> extensions would
> > need to be carefully defined to avoid breaking existing applications
> and fabrics.
> [EZ] You are welcom
Hal Rosenstock voltaire.com> writes:
[snip]
> How would you call ib_mad_port_close ? It's not exposed to userspace.
I'm gonna hack the module and export this symbol. I don't need to run this code
on a vanilla kernel, it is being used in a closed environment.
Thanks Hal,
z
春菜まいちゃんの無修正・裏ＤＶＤや、
いまや痴女優の女王となった立花理子ちゃんの裏ＤＶＤなど、
有名女優の裏モノを各種取り揃えております！
http://masu2.pink-no1.net
裏・無修正DVDのネットショップ
http://masu2.pink-no1.net
まずはその品揃えと、価格の安さに驚いてくださいヽ(^_^)
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman
On Tue, 2006-05-30 at 17:33, Sean Hefty wrote:
> Hal Rosenstock wrote:
> > Is client reregister handled properly by the multicast module ?
>
> Can you clarify what you mean by this? Are you asking about re-sending join
> requests based on some event?
Yes; when the SM sends a Set PortInfo with C
Hal Rosenstock wrote:
Is client reregister handled properly by the multicast module ?
Can you clarify what you mean by this? Are you asking about re-sending join
requests based on some event?
- Sean
___
openib-general mailing list
openib-general@o
On Tue, 2006-05-30 at 13:44, Sean Hefty wrote:
> Hal Rosenstock wrote:
> > Send-only joins is another case. These are full member joins (JoinState
> > 1) to groups which are not yet created so they fail.
>
> I see the problem, and checked in a fix. I forgot to record the last join
> operation t
Michael S. Tsirkin wrote:
I'm still looking at isolating this failure. I'd like to understand the new
code better, however. What prevents ipoib_mcast_leave and later
ipoib_mcast_free from being called on an mcast that has an outstanding query?
We used to have a completion to signal that but it
On 22:43 Tue 30 May , Eitan Zahavi wrote:
> > >
> > > XML style syntax is provided for the policy file.
> >
> > Why XML? It is not too much readable and writable (by human) format.
> [EZ] Well, I agree with you but already got so many requests for XML
> that I could not resists. Maybe we could
Mike> Should iser patches have linux-scsi ccd on them in the
Mike> future? And should they go through the scsi maintainer
Mike> normally (I understand they cannot now since James does not
Mike> have all the infinniband bits)? I am really just trying to
Mike> avoid any coordinata
Quoting r. Fabian Tillier <[EMAIL PROTECTED]>:
> Subject: Re: [openib-general] Re: CMA backlog
>
> On 5/30/06, Michael S. Tsirkin <[EMAIL PROTECTED]> wrote:
> >Quoting r. Sean Hefty <[EMAIL PROTECTED]>:
> >> >Currently, this is not something that can be implemented by ULP on top
> >of
> >> >CMA,
Should iser patches have linux-scsi ccd on them in the future? And
should they go through the scsi maintainer normally (I understand they
cannot now since James does not have all the infinniband bits)? I am
really just trying to avoid any coordinatation issues that come about by
having core iscsi a
Hi Sasha,
Thanks for your comments.
Please see my comments inside
> > 3. Supported Policy
> >
> >
> > The QoS policy supported by this proposal is divided into 4 sub
sections:
> >
> > * Node Group: a set of HCAs, Routers or Switches that share the same
settings.
> > A node gr
Hi Hal,
Please see my responses inside
Eitan
> >
> > RFC: OpenFabrics Enhancements for QoS Support
> > ===
> >
> > Authors: . Eitan Zahavi <[EMAIL PROTECTED]>
> > Date: May 2006.
> > Revision: 0.1
> >
> > Table of conten
On Tue, 2006-05-30 at 11:02, yipee wrote:
> Hal Rosenstock voltaire.com> writes:
>
> > On Tue, 2006-05-30 at 05:35, yipee wrote:
> > > Can I use ib_mad_port_close() (mad.c) to close qp0 & qp1
> >
> > Yes, that would close QP0/1 on each port. Just unloading the ib_mad
> > module will have that ef
Hi Todd,
> While using the Service ID is an interesting idea, the problem is the
Service ID values
> are not well defined by IBTA. Rather each endpoint is permitted to
define its own,
> potentially transient set of Service ID values. The Service ID values
are discovered via
> Service Records in
Descubra os segredos de Dan Brown, autor do maior sucesso literário e
cinematográfico dos últimos tempos. http://codigodavinc.webcindario.com/codigodavinci.exe
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman/listinfo
On 5/30/06, Michael S. Tsirkin <[EMAIL PROTECTED]> wrote:
Quoting r. Sean Hefty <[EMAIL PROTECTED]>:
> >Currently, this is not something that can be implemented by ULP on top of
> >CMA, because returning error from REQ will result in reject rather than REQ
> >drop.
>
> A generic ULP could handle
Hi Eitan,
First comments...
On 17:53 Tue 30 May , Eitan Zahavi wrote:
>
> 3. Supported Policy
>
>
> The QoS policy supported by this proposal is divided into 4 sub sections:
>
> * Node Group: a set of HCAs, Routers or Switches that share the same
> settings.
> A nod
Quoting r. Sean Hefty <[EMAIL PROTECTED]>:
> Subject: Re: ipoib use of multicast module on trunk causes kernel oops on
> 2.6.16
>
> Michael S. Tsirkin wrote:
> >>It's not completely trivial to reproduce. I tried loading and
> >>unloading ib_ipoib a few times, then I tried to load ib_ipoib and
>
Quoting r. Sean Hefty <[EMAIL PROTECTED]>:
> Subject: RE: CMA backlog
>
> >This approach would affect all ULPs, however. For example, no SDP
> >imlementation
> >that I know of retries after a REJ - so this approach won't be interoperable.
> >And AFAIK SDP spec already interprets reject as connect
Quoting r. Sean Hefty <[EMAIL PROTECTED]>:
> >Currently, this is not something that can be implemented by ULP on top of
> >CMA, because returning error from REQ will result in reject rather than REQ
> >drop.
>
> A generic ULP could handle this by making use of the private data, and
> retrying
> r
>This approach would affect all ULPs, however. For example, no SDP imlementation
>that I know of retries after a REJ - so this approach won't be interoperable.
>And AFAIK SDP spec already interprets reject as connection refused.
>There's no provision I cansee in SDP spec for retries on specific
>re
Quoting r. Sean Hefty <[EMAIL PROTECTED]>:
> Subject: RE: CMA backlog
>
> I think that there are some issues that would need to be worked out, but in
> general I'm in favor of trying to do something here.
>
> >Currently, this is not something that can be implemented by ULP on top of
> >CMA, becau
[EMAIL PROTECTED] wrote:
> Hello, Sean!
> I am looking at implementing the listen backlog parameter correctly.
> Here's what this does in TCP: TCP counts the number of
> connect requests at the specific local socket that were not
> yet accepted by accept().
> Once this number exceeds the backlog sp
I think that there are some issues that would need to be worked out, but in
general I'm in favor of trying to do something here.
>Currently, this is not something that can be implemented by ULP on top of
>CMA, because returning error from REQ will result in reject rather than REQ
>drop.
A generic
High-level feedback:
- An IB fabric could be used for a single ULP and still require
QoS. The issue is how to differentiate flows on a given shared
element within the fabric.
- QoS controls must be dynamic. The document references initialization as
the time when decisions are made but obviously
Hello, Sean!
I am looking at implementing the listen backlog parameter correctly.
Here's what this does in TCP: TCP counts the number of connect requests
at the specific local socket that were not yet accepted by accept().
Once this number exceeds the backlog specified in listen, new SYN
packets wi
On Tue, 2006-05-30 at 10:53, Eitan Zahavi wrote:
> To: OPENIB
> Subject: QoS RFC - Resend using a friendly mailer
> --text follows this line--
> Hi All
>
> Please find the attached RFC describing how QoS policy support could be
> implemented in the OpenFabrics stack.
> Your comments are welcome
Hal Rosenstock wrote:
Send-only joins is another case. These are full member joins (JoinState
1) to groups which are not yet created so they fail.
I see the problem, and checked in a fix. I forgot to record the last join
operation that was initiated, so that it could be failed on an error. T
Hi Paul,
On Tue, 2006-05-30 at 11:06, Paul wrote:
> Hi All,
> I will be working on this as time permits this week.
> Unfortunately my employer is not crazy about giving out remote access,
> so I will have to be your hands on this. If you want me to do
> something just tell me what it is. I kn
On Tue, 2006-05-30 at 10:58, Roland Dreier wrote:
> > [EZ] Well, a ULP that uses CMA will have it handled by CMA...
> > But an old SM implementation that does not support this kind of
> > PathRecord extension will probably choke on the new fields once their
> > component mask bits are set.
>
I queued this for 2.6.18.
___
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
Don> I also tried this. I didn't see any output on my terminal.
Don> Where does all this "copious output" go?
Into the kernel log.
- R.
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman/listinfo/openib-general
> Eitan wrote:
> 2.2. The SM analyzes the provided policy to see if it is
> realizable and performs
> the necessary fabric setup. The SM may continuously monitor
> the policy and adapt
> to changes in it. Part of this policy defines the default
> QoS-Level of each
> partition. The SA is bein
Roland,
[EMAIL PROTECTED] wrote on 05/30/2006
07:46:08 AM:
> Don> It is rather difficult for me to debug this
sort of hang,
> Don> since I telecommute from Tucson and the machines
are located
> Don> in Phoenix. Anyone have any suggestions?
>
> cat /proc//wchan for the process in qu
Don,
On Tue, 2006-05-30 at 10:55, [EMAIL PROTECTED] wrote:
> Hal,
>
> With your patch to OpenSM, I think everything is ok on the local node.
That patch with one minor change (elimination of the CL_ASSERT) will be
part of the upcoming RC6.
> The remote node is definitely having some problems,
Tziporet> We got the bug report from a customer. We did not
Tziporet> succeeded to reproduce the failure here, since we don't
Tziporet> have such strong machines (4 dual core CPU) as the
Tziporet> customer has.
I do have 4-socket dual core systems (although my lab has no power
righ
On Tue, 2006-05-30 at 11:06, Eitan Zahavi wrote:
> > -Original Message-
> > From: Hal Rosenstock [mailto:[EMAIL PROTECTED]
> > Sent: Tuesday, May 30, 2006 5:50 PM
> > To: Eitan Zahavi
> > Cc: Roland Dreier; openib-general@openib.org; Nimrod Gindi; Aviram
> Gutman;
> > Sasha Khapyorsky; [EMA
Roland Dreier wrote:
By the way, did you get a chance to test the AH leak fix to see if it
really fixes your leak? That would make me feel better about asking
Linus to pull it into 2.6.17.
- R.
We got the bug report from a customer. We did not succeeded to reproduce
the failure here, since
Hi All, I will be working on this as time permits this week. Unfortunately my employer is not crazy about giving out remote access, so I will have to be your hands on this. If you want me to do something just tell me what it is. I know its a pain I have been there myself.
Regards.On 5/30/06, [E
>
> > [EZ] Well, a ULP that uses CMA will have it handled by CMA...
> > But an old SM implementation that does not support this kind of
> > PathRecord extension will probably choke on the new fields once
their
> > component mask bits are set.
> > You could however query once for each Client-R
Hal Rosenstock voltaire.com> writes:
> On Tue, 2006-05-30 at 05:35, yipee wrote:
> > Can I use ib_mad_port_close() (mad.c) to close qp0 & qp1
>
> Yes, that would close QP0/1 on each port. Just unloading the ib_mad
> module will have that effect.
Would this unloading cause the port status to dow
> -Original Message-
> From: Hal Rosenstock [mailto:[EMAIL PROTECTED]
> Sent: Tuesday, May 30, 2006 5:50 PM
> To: Eitan Zahavi
> Cc: Roland Dreier; openib-general@openib.org; Nimrod Gindi; Aviram
Gutman;
> Sasha Khapyorsky; [EMAIL PROTECTED]; Vu Pham; Roland Dreier
> Subject: RE: QoS RFC
>
On Tue, 2006-05-30 at 10:51, Eitan Zahavi wrote:
> Hi Roland,
> >
> > This is OK but it's sort of a pain to have to query SA ClassPortInfo
> > all the time. Do you have a plan for how to make this transparent to
> ULPs?
> [EZ] Well, a ULP that uses CMA will have it handled by CMA...
> But an old
> [EZ] Well, a ULP that uses CMA will have it handled by CMA...
> But an old SM implementation that does not support this kind of
> PathRecord extension will probably choke on the new fields once their
> component mask bits are set.
> You could however query once for each Client-Reregister ev
Hi Roland,
> BTW I think these changes to PathRecord and MultiPathRecord need to be
> standardized through IBTA before we implement it in Linux, to avoid a
> non-standard implementation proliferating everywhere.
These extensions are already being discussed in IBTA LWG.
It will take some time bef
Hal,
With your patch to OpenSM, I think everything is ok
on the local node. The remote node is definitely having some problems,
resulting in not responding to the MAD packets. I have entered a
separate message on the problems with the "ib0" interface on
that machine.
>
> On Fri, 2006-05-26 at
To: OPENIB
Subject: QoS RFC - Resend using a friendly mailer
--text follows this line--
Hi All
Please find the attached RFC describing how QoS policy support could be
implemented in the OpenFabrics stack.
Your comments are welcome.
Eitan
RFC: OpenFabrics Enhancements for QoS Sup
Hi Roland,
>
> This is OK but it's sort of a pain to have to query SA ClassPortInfo
> all the time. Do you have a plan for how to make this transparent to
ULPs?
[EZ] Well, a ULP that uses CMA will have it handled by CMA...
But an old SM implementation that does not support this kind of
PathRecor
Don> It is rather difficult for me to debug this sort of hang,
Don> since I telecommute from Tucson and the machines are located
Don> in Phoenix. Anyone have any suggestions?
cat /proc//wchan for the process in question. "echo t >
/proc/sysrq-trigger"
will produce copious output tha
BTW I think these changes to PathRecord and MultiPathRecord need to be
standardized through IBTA before we implement it in Linux, to avoid a
non-standard implementation proliferating everywhere.
___
openib-general mailing list
openib-general@openib.org
ht
> Service-ID:
> * For PathRecord: use the first 2 reserved fields whicg are 32bits each
> (component masks 0x1 and 0x2). Component mask 1 should be used to
> refer to the
> merged Service-ID field
> A new capability bit should describe the SM QoS support in the SA class
> port
> in
Michael,
> > The ib_mthca module now initializes
correctly on
> > both EM64T machines. I noticed some discussion between
you and Roland about
> > making the parameter "fw_cmd_doorbell=0" the default.
Did this
> > occur in RC5?
>
> Yes, we changed fw_cmd_doorbell to 0 by default for now beca
By the way, did you get a chance to test the AH leak fix to see if it
really fixes your leak? That would make me feel better about asking
Linus to pull it into 2.6.17.
- R.
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mai
Michael> 2.6.18 then?
Yes, definitely.
- 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
> +struct port_info {
> +__be64 mkey;
> +__be64 gid_prefix;
> +__be16 lid;
> +__be16 sm_lid;
> +__be32 cap_mask;
> +__be16 diag_code;
> +__be16 mkey_lease_period;
> +u8 local_port_num;
> +u8 link_width_enabled;
> +u8 link_widt
Quoting r. Roland Dreier <[EMAIL PROTECTED]>:
> Subject: Re: ia64: kernel unaligned access in ipoib
>
> I don't really see this as 2.6.17 material -- it's slightly annoying
> but at this point in the cycle it's probably not worth the risk.
2.6.18 then?
--
MST
OK, I added both patches on top of what I have queued.
- 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
Michael> The following fixed the issue for us, pls review. Can
Michael> this go into 2.6.17? If yes, I think it's prudent to let
Michael> it run for another night before pushing it out, since we
Michael> had to touch a lot of lines here. We'll do that and let
Michael> you know
Leonid> Roland, Aren't you going to apply these patches? This is
Leonid> a standard feature, and we need it rather urgently.
Sorry, I had let these patches fall off of my queue.
So why is this an urgent feature? A good description of your use case
would make it easier for me to write th
On Tue, 2006-05-30 at 05:35, yipee wrote:
> Hi,
>
> Can I use ib_mad_port_close() (mad.c) to close qp0 & qp1
Yes, that would close QP0/1 on each port. Just unloading the ib_mad
module will have that effect. All agents and managers (using ib_mad)
should be stopped. I'm presuming you don't want the
Looks good. Thanks for fixing it.
Eitan
> -Original Message-
> From: Hal Rosenstock [mailto:[EMAIL PROTECTED]
> Sent: Tuesday, May 30, 2006 3:12 PM
> To: openib-general@openib.org
> Cc: Eitan Zahavi
> Subject: [PATCH] OpenSM: Remove unicast and multicast dump files
relativeto
> dump_files
OpenSM: Remove unicast and multicast dump files relative to
dump_files_dir
Signed-off-by: Hal Rosenstock <[EMAIL PROTECTED]>
Index: opensm/osm_mcast_mgr.c
===
--- opensm/osm_mcast_mgr.c (revision 7535)
+++ opensm/osm_mcast_mgr.c
This fix isn't critical but it aligns the code to be as in the buddy
iscsi_tcp driver - Or.
removed redundant check of the conn stop_stage from iser's conn_bind transport
func
Signed-off-by: Or Gerlitz <[EMAIL PROTECTED]>
diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c
b/drivers/infiniba
Thanks for Mike Christie for pointing this out - Or.
a block driver is not allowed to use GFP_KERNEL allocations on its I/O code
path since the allocation might require I/O (eg to pageout other memory),
resulting in either deadlock or tightloop.
move I/O path (queuecommand) allocations to be done
Roland,
Further to our discussions I'm sending the fixed patch.
This patch implements the module parameters allowing the user to change the
HCA initialization values.
I left only needed parameters and added the parameter validation.
Now the set of the parameters is closest to the profile parame
Roland,
Aren't you going to apply these patches?
This is a standard feature, and we need it rather urgently.
Regards,
Leonid
On 5/9/06, Leonid Arsh <[EMAIL PROTECTED]> wrote:
Roland,
I'm reposting the Client Reregister event support patch for the kernel
space.
The patch defines
Quoting r. Roland Dreier <[EMAIL PROTECTED]>:
> Subject: Re: ia64: kernel unaligned access in ipoib
>
> Michael> We've written up a patch with Jack - do you want us to
> Michael> test it or prefer to re-write it yourself?
>
> Go ahead and test it -- I replied before I saw your patch.
The
Hi,
Can I use ib_mad_port_close() (mad.c) to close qp0 & qp1 amd reopen them from a
userspace application? That way I can handle all mads without any kernel
intervention?
Thanks,
x
___
openib-general mailing list
openib-general@openib.org
http://openi
Thanks,
It was a problem with my git installation. I reinstall and everything is
OK now.
Tziporet
-Original Message-
From: Roland Dreier [mailto:[EMAIL PROTECTED]
Sent: Monday, May 29, 2006 7:11 PM
To: Tziporet Koren
Cc: openib
Subject: Re: problems with git
> git clone
git://www.kerne
D m ea s r H o om v e O h wn a er,
Your c x re l di m t doesn't matter to us!
If you OV v VN r j ea p l e j st i at n e and want I j MM e EDI c ATE c t as y h to s p pen y d ANY
way you like, or simply wish to L k OW b ER your monthly pa p ym x ent r s
by a third or more, here are the d v ea v
76 matches
Mail list logo
