> The Chelsio driver is hogging lots of memory right now for mapping
> PDIDs, QPIDs, CQIDs, and STAG IDs back to their respective kernel
> structures. This is done via an array of pointers, indexed by the ID.
> The critical performance mapping is finding a QP struct from the QPID in
> the pol
> What should be the expected behavior?
> Should this description should be changed or should the low level drivers
> of mthca and ipath need to be changed?
The mask is used as a hint to the low-level driver about which
attributes the consumer cares about. The driver may fill in more
fields, b
Robert Walsh wrote:
> Roland Dreier wrote:
>> > Is there are registration authority for multicast GIDs? Or at
>> least a > safe way of assigning a range of GIDs to a vendor?
>>
>> I don't think so. Perhaps RFC 3307 would be of some use...
>
> Ah - looks exactly like what I was looking for. T
Roland Dreier wrote:
> > Is there are registration authority for multicast GIDs? Or at least a
> > safe way of assigning a range of GIDs to a vendor?
>
> I don't think so. Perhaps RFC 3307 would be of some use...
Ah - looks exactly like what I was looking for. Thanks.
_
> Is there are registration authority for multicast GIDs? Or at least a
> safe way of assigning a range of GIDs to a vendor?
I don't think so. Perhaps RFC 3307 would be of some use...
- R.
___
openib-general mailing list
openib-general@openib.org
hysErr:8
> OverrunErr:..8
> MaxCreditHint:...0
> RoundTrip:...0
Do you have an IB analyzer ?
-- Hal
> Feiyi
>
> -Original Message-
> From: Hal Rosenstock [mailto:[EMAIL PROTECTED]
> Sent:
:...0
RoundTrip:...0
Feiyi
-Original Message-
From: Hal Rosenstock [mailto:[EMAIL PROTECTED]
Sent: Friday, November 03, 2006 3:58 PM
To: Wang, Feiyi
Cc: openib-general@openib.org
Subject: RE: [openib-general] question on QoS support
On Fri, 2006-11-03 at
o:[EMAIL PROTECTED]
> Sent: Friday, November 03, 2006 3:51 PM
> To: Wang, Feiyi
> Cc: openib-general@openib.org
> Subject: RE: [openib-general] question on QoS support
>
> On Fri, 2006-11-03 at 15:43, Wang, Feiyi wrote:
> > The test is done on two hosts, say A and B. A has 4x S
er 03, 2006 3:51 PM
To: Wang, Feiyi
Cc: openib-general@openib.org
Subject: RE: [openib-general] question on QoS support
On Fri, 2006-11-03 at 15:43, Wang, Feiyi wrote:
> The test is done on two hosts, say A and B. A has 4x SDR (run
ib_rdam_bw
> as server), B has 4x DDR (run more than one thread of
On Fri, 2006-11-03 at 15:12, Feiyi Wang wrote:
> In our test at the ORNL - it appears you can "turn off" the traffic by
> giving every VL weight 0.
A weight of 0 indicates to skip that entry.
> As soon as you assign non-zero VL weight,
> the traffic starts to flow, however, VL with more weight d
In our test at the ORNL - it appears you can "turn off" the traffic by
giving every VL weight 0. As soon as you assign non-zero VL weight,
the traffic starts to flow, however, VL with more weight doesn't have
expected preference treatment. In other words, traffic shaping didn't
take place. smpquery
Sean posted 7 patches that include the ucma support.
You'll need those + the one librdmacm patch he posted.
Steve.
On Fri, 2006-11-03 at 13:59 +0530, Krishna Kumar2 wrote:
> Hi,
>
> I installed the 2.6.19-rc3 bits, and when I try to run
> perftest/rdma_bw (with '-c' option), I get the error :
Hi Oliver,
On Thu, 2006-11-02 at 10:20, Oliver wrote:
> Hi, Hal -
>
> > How is this being observed/measured ?
>
> Host A, B, with 4x DDR both connected to Flextronic switch.
> A single process of ibv_read_bw gives about 1415MB /s average
> bandwidth. Two concurrent process report 714.45 MB/s eac
Hi, Hal -
> How is this being observed/measured ?
Host A, B, with 4x DDR both connected to Flextronic switch.
A single process of ibv_read_bw gives about 1415MB /s average
bandwidth. Two concurrent process report 714.45 MB/s each, dead even.
Now if I bump up one process with a different SL, then
On Thu, 2006-11-02 at 09:15, Makia Minich wrote:
> Hal Rosenstock wrote:
> > Makia,
> >
> > On Wed, 2006-11-01 at 17:42, Makia Minich wrote:
> >> It just so happens that we've started looking at this here at ORNL as
> >> well. I had a question about the options. The manpage makes it seem
> >> th
Hal Rosenstock wrote:
> Makia,
>
> On Wed, 2006-11-01 at 17:42, Makia Minich wrote:
>> It just so happens that we've started looking at this here at ORNL as
>> well. I had a question about the options. The manpage makes it seem
>> that you can set these qos options (e.g. qos_high_limit) from the
Makia,
On Wed, 2006-11-01 at 17:42, Makia Minich wrote:
> It just so happens that we've started looking at this here at ORNL as
> well. I had a question about the options. The manpage makes it seem
> that you can set these qos options (e.g. qos_high_limit) from the
> command line, but I haven't
Hi Oliver,
On Wed, 2006-11-01 at 16:52, Oliver wrote:
> Hi, folks -
>
> I am trying to verify and evaluate IB QoS support, running openSM as
> subnet manager. The perftest program is extended to set SL as command
> line options instead of default 0, and by modifying VL arbitration
> tables, I am
On 17:42 Wed 01 Nov , Makia Minich wrote:
> It just so happens that we've started looking at this here at ORNL as
> well. I had a question about the options. The manpage makes it seem
> that you can set these qos options (e.g. qos_high_limit) from the
> command line,
AFAIK there is option -Q
It just so happens that we've started looking at this here at ORNL as
well. I had a question about the options. The manpage makes it seem
that you can set these qos options (e.g. qos_high_limit) from the
command line, but I haven't been overly successful. Is there an example
of this being done?
On 16:52 Wed 01 Nov , Oliver wrote:
> Hi, folks -
>
> I am trying to verify and evaluate IB QoS support, running openSM as
> subnet manager. The perftest program is extended to set SL as command
> line options instead of default 0, and by modifying VL arbitration
> tables, I am expecting to se
> While looking over the ehca driver from the perspective of adding a
> "peek CQ" operation, I noticed some code that looked funny.
>
> In hipz_set_cqx_n0() and hipz_set_cqx_n1(), what is the point of the
> calls to hipz_galpa_load_cq()? The return value is discarded. I see
> that hipz_galpa_load
Hi,One more question. What kind of event mask helps mask the interrupts?thanksharishOn 9/5/06, harish <
[EMAIL PROTECTED]> wrote:Hi All,I tried the following simple experiment and am not able to understand the results:
Calcualted the number of interrupts generated by the infiniband [with little or
> > It gives the page size for the user memory described by the struct.
> > The idea was that if/when someone tries to optimize for huge pages,
> > then the low-level driver can know that a region is using huge pages
> > without having to walk through the page list and search for the
> > minim
Quoting r. Roland Dreier <[EMAIL PROTECTED]>:
> Subject: Re: question: ib_umem page_size
>
> Michael> Roland, could you please clarify what does the page_size
> Michael> field in struct ib_mem do?
>
> It gives the page size for the user memory described by the struct.
> The idea was that
>Cool, I would go for XOR-ing a random value with the **local id** .
>
>Sean, my understanding it can be narrowed for doing so in:
>
>1) cm_alloc_id() after calling idr_get_new_above()
>2) cm_free_id() before calling idr_remove()
>3) cm_get_id() before calling idr_find()
>
>and initializing the ran
Sean Hefty wrote:
> When a new REQ is received, we enter its timewait structure into two
trees: one
> sorted by remote ID, one sorted by remote QPN. If the REQ is new, both would
> succeed, and timewait_info would be NULL. Since timewait_info is not NULL, we
> are dealing with a REQ that re-us
Roland Dreier wrote:
> Sean> If we record a base offset, we can start at any random
> Sean> number. We just need to always add/subtract the base when
> Sean> getting a value from the IDR.
>
> Good point -- or better still, we could XOR in a random bit pattern.
> That way we don't have
Sean> If we record a base offset, we can start at any random
Sean> number. We just need to always add/subtract the base when
Sean> getting a value from the IDR.
Good point -- or better still, we could XOR in a random bit pattern.
That way we don't have to keep straight when to add and
>> If we get here, this means that the REQ was a new REQ and not a
>> duplicate, but the remote_id or remote_qpn is already in use. We need
>> to reject the new REQ as containing stale data.
>
>I don't follow, if we get to the else case its as of cm_get_id()
>returning NULL. This holds when idr_fi
>> Just to emphasize what Sean has pointed out, you are asking how can a CM
>> consumer know that a **local** QPN is not in the timewait state
>> according to the **remote** CM. Since the issue is with the remote CM,
>> it seems to me that pushing down timewait into verbs is not the correct
>> dire
>How about (for the meantime, till this rework is designed && done) going
>to projecting the initial random local id into the range of (say)
>[0-1022] (i think 1023 is prime, if not choose a prime near it) this way
>with very good probability and with very little overhead on memory
>consumption a c
Or> How about (for the meantime, till this rework is designed &&
Or> done) going to projecting the initial random local id into the
Or> range of (say) [0-1022] (i think 1023 is prime, if not choose
Or> a prime near it) this way with very good probability and with
Or> very little
This email appear in the archive, but seems not to be distributed to the
subscribers so i am reposting it.
Or Gerlitz wrote:
> Sean Hefty wrote:
>> Even if we pushed timewait handling under verbs, a user could always
>> get a QP that the remote side thinks is connected. The original
>> connec
This email appear in the archive, but seems not to be distributed to the
subscribers so i am reposting it.
Or Gerlitz wrote:
> Arlin Davis wrote:
>> We are running into connection reject issues (IB_CM_REJ_STALE_CONN)
>> with our application under heavy load and lots of connections.
>>
>> We occa
>>> + } else
>>> + cm_issue_rej(work->port, work->mad_recv_wc,
>>> +IB_CM_REJ_STALE_CONN,
>>> CM_MSG_RESPONSE_REQ,
>>> +NULL, 0);
>>
>>
>> what is this case? there is no entry but there is
Sean Hefty wrote:
> Or Gerlitz wrote:
>> If you don't mind (also related to the patch you have sent Eric of
>> randomizing the initial local cm id) to get into this deeper, can we do
> There's an issue trying to randomize the initial local CM ID. The way
> the IDR works, if you start at a hig
Or Gerlitz wrote:
> If you don't mind (also related to the patch you have sent Eric of
> randomizing the initial local cm id) to get into this deeper, can we do
There's an issue trying to randomize the initial local CM ID. The way the IDR
works, if you start at a high value, then the IDR size
Arlin Davis wrote:
> How can a consumer know for sure that the new QP will not be in a
> timewait state according to the CM?
Given that the QP may have been in use by another process, I don't think that
there's any way for the new owner to know.
> Does it make sense to push the timewait functio
Quoting r. Roland Dreier <[EMAIL PROTECTED]>:
> Subject: Re: question: ib_umem page_size
>
> Michael> Roland, could you please clarify what does the page_size
> Michael> field in struct ib_mem do?
>
> It gives the page size for the user memory described by the struct.
> The idea was that
Michael> Roland, could you please clarify what does the page_size
Michael> field in struct ib_mem do?
It gives the page size for the user memory described by the struct.
The idea was that if/when someone tries to optimize for huge pages,
then the low-level driver can know that a region is
At 12:36 PM 6/5/2006, Talpey, Thomas wrote:
Thanks Parks, this is a very interesting perspective.
I will avoid going into my rant about edge devices for
now, however. :-)
Cool, you can send it direct if you want.
I am not sure what you mean about using SDP "end to end".
I assume you would pe
Thanks Parks, this is a very interesting perspective.
I will avoid going into my rant about edge devices for
now, however. :-)
I am not sure what you mean about using SDP "end to end".
I assume you would perhaps use SDP to these edge nodes,
but this would require terminating the SDP connection and
I consider IPoIB to be Ethernet emulation.
As for apples and oranges, my point exactly.
It is not really about comparisons. Here at LANL we have an
environment where all our new Clusters have to mount our global
parallel file system Panasas. It is ethernet and will be for a while.
Clust
bout the IPoIB bandwidth performance ?
> (Talpey, Thomas)
> 8. Re: Question about the IPoIB bandwidth performance ? (hbchen)
>
>----- Message from "hbchen" <[EMAIL PROTECTED]> on Mon, 05 Jun 2006 09:38:24
>-0600 -
>
>
Tom,
We are in the process of measuring the CPU utilization on our NFS/RDMA
experiments in contrast with regular the NFS, we also intend to include
netperf numbers and will keep you posted with our results as soon as
possible.
Helen
- original Message -
>From [EMAIL PROTECTED] Mon Jun
IL PROTECTED]> on Mon, 05 Jun 2006 09:38:24
-0600 -
To: "Hal Rosenstock" <[EMAIL PROTECTED]>
cc: "OPENIB"
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On
Behalf Of hbchen
Sent: Monday, June 05, 2006 9:12
AM
To: Talpey, Thomas
Cc: openib-general@openib.org
Subject: Re: [openib-general]
Question about the IPoIB bandwidth performance ?
Talpey, Thomas wrote:
At 11:38
At 12:11 PM 6/5/2006, hbchen wrote:
>>Perhaps you are out of CPU.
>>
>>
>Tom,
>I am HB Chen from LANL not the Helen Chen from SNL.
Oops, sorry! I have too many email messages going by. :-)
HB, then.
>I didn't run out of CPU. It is about 70-80 % of CPU utilization.
But, is one CPU at 100%? In
Talpey, Thomas wrote:
At 11:38 AM 6/5/2006, hbchen wrote:
Even with this IB-4X = 8Gb/sec = 1024 MB/sec the IPoIB bandwidth utilization is still very low.
IPoIB=420MB/sec
bandwidth utilization= 420/1024 = 41.01%
Helen, have you
At 11:38 AM 6/5/2006, hbchen wrote:
>Even with this IB-4X = 8Gb/sec = 1024 MB/sec the IPoIB bandwidth utilization
>is still very low.
>>> IPoIB=420MB/sec
>>> bandwidth utilization= 420/1024 = 41.01%
Helen, have you measured the CPU utilizations during these runs?
Perhaps you are out of CPU.
O
Hal Rosenstock wrote:
> On Mon, 2006-06-05 at 11:12, hbchen wrote:
> > Hi,
> > I have a question about the IPoIB bandwidth performance.
> > I did netperf testing using Single GiGE, Myrinet D card, Myrinet 10G
> > ethernet card,
> > and Voltaire Infiniband 4X HCA400Ex (PCI-Express interface).
> >
>
Hal Rosenstock wrote:
On Mon, 2006-06-05 at 11:12, hbchen wrote:
Hi,
I have a question about the IPoIB bandwidth performance.
I did netperf testing using Single GiGE, Myrinet D card, Myrinet 10G
ethernet card,
and Voltaire Infiniband 4X HCA400Ex (PCI-Express interface).
NIC (Jumb
On Mon, 2006-06-05 at 11:12, hbchen wrote:
> Hi,
> I have a question about the IPoIB bandwidth performance.
> I did netperf testing using Single GiGE, Myrinet D card, Myrinet 10G
> ethernet card,
> and Voltaire Infiniband 4X HCA400Ex (PCI-Express interface).
>
>
> NIC (Jumbo enabled) Line bandwid
On Sun, 2006-05-14 at 15:30, Jason Gunthorpe wrote:
> On Sun, May 14, 2006 at 07:40:25AM -0400, Hal Rosenstock wrote:
> > > > Not always true in terms of local subnet (multicast and management MAD
> > > > response exceptions).
> > >
> > > Yes, but these are well specified. Multicast must always ha
On Sun, May 14, 2006 at 07:40:25AM -0400, Hal Rosenstock wrote:
> > > Not always true in terms of local subnet (multicast and management MAD
> > > response exceptions).
> >
> > Yes, but these are well specified. Multicast must always have a GRH.
> > MAD requests are covered under my scenario above
On Fri, 2006-05-12 at 13:55, Sean Hefty wrote:
> Jason Gunthorpe wrote:
> > How about this, how do you see this scenario:
> >
> > 1) Client gets a DGID from 'someplace'
> > 2) Client sends a SA query to resolve the DGID to a Path Record
> > 3) Client configures a QP based on the Path Record
> >
>
On Fri, 2006-05-12 at 13:10, Jason Gunthorpe wrote:
> On Fri, May 12, 2006 at 08:11:17AM -0400, Hal Rosenstock wrote:
>
> > > To allow what Roland is talking about you need an unambiguous
> > > mechanism where the SA can signal to the client that the path
> > > needs a GRH.
> >
> > Ah, you are re
Jason Gunthorpe wrote:
How about this, how do you see this scenario:
1) Client gets a DGID from 'someplace'
2) Client sends a SA query to resolve the DGID to a Path Record
3) Client configures a QP based on the Path Record
Now, the question I'm interested in is this:
During step #3 what test
On Fri, May 12, 2006 at 08:11:17AM -0400, Hal Rosenstock wrote:
> > To allow what Roland is talking about you need an unambiguous
> > mechanism where the SA can signal to the client that the path
> > needs a GRH.
>
> Ah, you are referring to the SA path record response not the request.
Yes.. Tho
On Thu, 2006-05-11 at 13:12, Jason Gunthorpe wrote:
> On Thu, May 11, 2006 at 07:20:19AM -0400, Hal Rosenstock wrote:
>
> > That would be a simpler check but HopLimit is not a required component
> > of PathRecord but I think this may not be sufficient as just because a
> > HopLimit >= 2 doesn't me
On Thu, May 11, 2006 at 10:21:08AM -0700, Sean Hefty wrote:
> Hal Rosenstock wrote:
> >Anytime the send is off the local subnet (as well as multicast), a GRH
> >is required. Also, there is a management response rule for responding
> >when the request contained a GRH that require a GRH (13.5.4.4 p.
On Thu, 2006-05-11 at 13:29, Roland Dreier wrote:
> Sean> We currently use ARP to resolve an IP address to a DGID,
> Sean> which I don't believe will work across a router. Does an
> Sean> app even know enough to be able to get a path record?
>
> I think you're fine. The IB router jus
Sean> We currently use ARP to resolve an IP address to a DGID,
Sean> which I don't believe will work across a router. Does an
Sean> app even know enough to be able to get a path record?
I think you're fine. The IB router just has to handle forwarding
multicasts between two IB subnets
Hal Rosenstock wrote:
Anytime the send is off the local subnet (as well as multicast), a GRH
is required. Also, there is a management response rule for responding
when the request contained a GRH that require a GRH (13.5.4.4 p. 769).
Reading through the responses, I think my problems are worse.
On Thu, May 11, 2006 at 07:20:19AM -0400, Hal Rosenstock wrote:
> That would be a simpler check but HopLimit is not a required component
> of PathRecord but I think this may not be sufficient as just because a
> HopLimit >= 2 doesn't mean that a packet would be forwarded off subnet.
I was thinkin
; From: [EMAIL PROTECTED] [mailto:openib-general-
> [EMAIL PROTECTED] On Behalf Of Hal Rosenstock
> Sent: Thursday, May 11, 2006 2:20 PM
> To: Jason Gunthorpe
> Cc: Roland Dreier; openib-general@openib.org
> Subject: Re: [openib-general] question regarding GRH flag in
ib_ah_attr
>
On Thu, 2006-05-11 at 01:48, Jason Gunthorpe wrote:
> On Wed, May 10, 2006 at 09:56:58PM -0700, Roland Dreier wrote:
> > Hal> Huh ? In this case, aren't the subnet prefixes are required
> > Hal> to be different ?
> >
> > It's kind of a crazy thing to do but I don't see anything in the IB
>
On Thu, 2006-05-11 at 00:56, Roland Dreier wrote:
> Hal> Huh ? In this case, aren't the subnet prefixes are required
> Hal> to be different ?
>
> It's kind of a crazy thing to do but I don't see anything in the IB
> spec that forbids two subnets with the same subnet prefix,
There's errata
On Wed, May 10, 2006 at 09:56:58PM -0700, Roland Dreier wrote:
> Hal> Huh ? In this case, aren't the subnet prefixes are required
> Hal> to be different ?
>
> It's kind of a crazy thing to do but I don't see anything in the IB
> spec that forbids two subnets with the same subnet prefix, or
Hal> What you are describing is similar to a NAT function for IB
Hal> which would need to be supported in the IB edge router to
Hal> that private network.
Why does there have to be any NAT? The router would just have to
replace the DLID the same as it usually does. I don't see why th
Hal> Huh ? In this case, aren't the subnet prefixes are required
Hal> to be different ?
It's kind of a crazy thing to do but I don't see anything in the IB
spec that forbids two subnets with the same subnet prefix, or any
reason why a router couldn't route between them. The SMs would just
On Wed, 2006-05-10 at 21:26, Hal Rosenstock wrote:
> On Wed, 2006-05-10 at 19:44, Roland Dreier wrote:
> > Sean> Does anyone know how the user determines if the grh flag
> > Sean> should be set in the ib_ah_attr when allocating an ib_ah?
> > Sean> Do they do this by examining the GIDs i
On Wed, 2006-05-10 at 19:35, Sean Hefty wrote:
> For context, I'm trying to work backwards from send a message on a UD QP to
> determine what information is needed and how it is obtained.
>
> Does anyone know how the user determines if the grh flag should be set in the
> ib_ah_attr when allocating
On Wed, 2006-05-10 at 19:44, Roland Dreier wrote:
> Sean> Does anyone know how the user determines if the grh flag
> Sean> should be set in the ib_ah_attr when allocating an ib_ah?
> Sean> Do they do this by examining the GIDs in a path record?
>
> Good question. It's always needed fo
On Wed, May 10, 2006 at 04:44:42PM -0700, Roland Dreier wrote:
> Sean> Does anyone know how the user determines if the grh flag
> Sean> should be set in the ib_ah_attr when allocating an ib_ah?
> Sean> Do they do this by examining the GIDs in a path record?
>
> Good question. It's alw
Sean> Does anyone know how the user determines if the grh flag
Sean> should be set in the ib_ah_attr when allocating an ib_ah?
Sean> Do they do this by examining the GIDs in a path record?
Good question. It's always needed for multicast, of course. For
unicast, I guess one could look
On Sat, 8 Apr 2006, Devesh Sharma wrote:
> In your nfs-rdma context what this function is supposed to do?
It should create a memory region for the specified address range. For
the exact semantics, see the IBTA spec's description of the REGISTER
PHYSICAL MEMORY REGION verb (section 11.2.8.3 of
Thanks James for quick reply,
In your nfs-rdma context what this function is supposed to do?
I know that this function returns memory region, but what is the difference from other mr returning functions?why get_dma_mr can't be used?
Devesh
On 4/7/06, James Lentini <[EMAIL PROTECTED]> wrote:
On
On Fri, 7 Apr 2006, Devesh Sharma wrote:
> Hello list,
> In Ib kernel verbs there is a function ib_reg_phys_mr().
> I am not able to trace the call of this verb by any ulp or uverb.
> Who calls this function?
NFS-RDMA uses this function:
http://sourceforge.net/projects/nfs-rdma
> Is this func
Hi list and Roland,
Is this verb (ib_get_dma_mr) is equivalent to the verb explained in the section 11.2.8.1 Allocate L_key?On 3/30/06, Steve Wise
<[EMAIL PROTECTED]> wrote:
On Wed, 2006-03-29 at 20:35 -0800, Roland Dreier wrote:> Devesh> Here I am saying that assigning Key is sufficient Or th
On Wed, 2006-03-29 at 20:35 -0800, Roland Dreier wrote:
> Devesh> Here I am saying that assigning Key is sufficient Or there
> Devesh> are some other specific setps to be taken?
>
> It would depend on the device. You can look at the mthca, ipath and ehca
> drivers' implementation of get_d
yha Ok Thanks for replying Once again.
DeveshOn 3/30/06, Roland Dreier <[EMAIL PROTECTED]> wrote:
Devesh> Here I am saying that assigning Key is sufficient Or thereDevesh> are some other specific setps to be taken?It would depend on the device. You can look at the mthca, ipath and ehcadri
Devesh> Here I am saying that assigning Key is sufficient Or there
Devesh> are some other specific setps to be taken?
It would depend on the device. You can look at the mthca, ipath and ehca
drivers' implementation of get_dma_mr() for examples.
- R.
_
On 3/29/06, Roland Dreier <[EMAIL PROTECTED]> wrote:
Devesh> S/G entry ?scatter gather entryDevesh> What is the size of this region ? is there any limitationDevesh> in providing this size?It must be large enough to cover all DMA (bus) addresses for the device.
Devesh> Finally you me
Devesh> S/G entry ?
scatter gather entry
Devesh> What is the size of this region ? is there any limitation
Devesh> in providing this size?
It must be large enough to cover all DMA (bus) addresses for the device.
Devesh> Finally you mean to say in the implementation of this
D
Thanks to all of you
On 3/27/06, Roland Dreier <[EMAIL PROTECTED]> wrote:
Devesh> Hello all, Please any body explain me about theDevesh> functionality of verbs ib_get_dma_mr()?Actually, the responses you've gotten are not quite right.ib_get_dma_mr() returns a memory region that can be used
On Sun, 26 Mar 2006, Or Gerlitz wrote:
> > I would find calling it rdma_bind_device() confusing.
>
> why? I find it very much unconfusing
I associate the word bind with bind(2). For that reason,
rdma_bind_addr() is a good name because it is the CMA's analog for
bind(2). Since it isn't relat
Devesh> Hello all, Please any body explain me about the
Devesh> functionality of verbs ib_get_dma_mr()?
Actually, the responses you've gotten are not quite right.
ib_get_dma_mr() returns a memory region that can be used for any _bus_
addresses. In other words, if an S/G entry is passed to
James Lentini wrote:
On Thu, 23 Mar 2006, Sean Hefty wrote:
I think that Or is just exploring the idea of synchronously binding
to a local *device* based on a remote address.
This would allow an application to bind, then allocate PDs, CQs,
QPs, etc. up front, rather than deferring resource al
Devesh Sharma wrote:
Please any body explain me about the functionality of verbs
ib_get_dma_mr()?
What is the need of this function?
what a driver implementer is supposed to implement in this function?
This function returns a memory region for all of system memory. See
mthca_provider.c for a
On Thu, 23 Mar 2006, Sean Hefty wrote:
> >What does it mean to bind to a remote address? What functionality
> >would that enable? Spoofing?
>
> I think that Or is just exploring the idea of synchronously binding
> to a local *device* based on a remote address.
>
> This would allow an applica
On Fri, 2006-03-24 at 12:51 +0530, Devesh Sharma wrote:
> Hello all,
>
> Please any body explain me about the functionality of verbs
> ib_get_dma_mr()?
> What is the need of this function?
> what a driver implementer is supposed to implement in this function?
It returns a MR that maps all of phy
>What does it mean to bind to a remote address? What functionality
>would that enable? Spoofing?
I think that Or is just exploring the idea of synchronously binding to a local
*device* based on a remote address. This would allow an application to bind,
then allocate PDs, CQs, QPs, etc. up front,
On Thu, 23 Mar 2006, Sean Hefty wrote:
> >I could not approve my assumptions from looking on the cma/addr
> >code, but if i am correct this opens the door for future
> >enhancement of rdma_bind_addr() to work on non local addresses.
>
> I believe that could be the case.
What does it mean to b
Sean Hefty wrote:
If my understanding is correct, the current code of rdma_bind_addr
assumes you supply it one of three (all are **src** address)
+1 ANY (0.0.0.0) addr
+2 local loopback addr
+3 other local addr
Correct. - Note that currently a valid port number needs to be provided, but
this
Or Gerlitz wrote:
At this point i see an actual need, it just related to some change we
discuss in the open scsi model for iser integration, and i wanted to make
sure that currently creating the IB resources in synchronous manner is
impossible.
Sorry, my fingers are broken today...
I meant to
Devesh Sharma wrote:
In mad.c while calling ib_post_receive() operation
spin_lock_irqsave(&recv_queue->lock, flags);
post = (++recv_queue->count < recv_queue->max_active);
list_add_tail(&mad_priv->header.mad_list.list, &recv_queue->list);
spin_unlock_irqrestore(&recv
Ralph Campbell wrote:
Ralph> When ib_modify_qp() is called with the IB_QP_ACCESS_FLAGS
Ralph> set in the mask, what values should be used in struct
Ralph> ib_qp_attr.qp_access_flags? The IB spec. seems to indicate
Ralph> that RDMA and atomic operations are all enabled or disabled
Jeff> Ditto (I thought those were shmem values / didn't think they
Jeff> had any effect on Open IB). The information that I got was
Jeff> third-hand, which is why I posted here to ask about it. :-)
Jeff> I'll remove them from the FAQ entry -- any other comments?
Well, a normal u
1 - 100 of 175 matches
Mail list logo
