Håkon Bugge wrote:
I used the information at
www.openfabrics.org/git/?p=ofed_1_2_5/libibverbs.git;a=summary
which states the owner to be Vlad. May be that confused me. I'll send a
copy to Roland
Roland's user space git trees are all hosted @ kernel.org
the libibverbs one is
On Jan 24, 2010, at 6:26 , Or Gerlitz wrote:
attaching a debugger is typically helpful to see where a program talking
directly to the hardware hangs. If it happens on the slow pass, strace can be
useful as well. Did you take a look on the actual values set for this qp,
that it as
Håkon Bugge wrote:
The capabilities in qp_init_attr used as input to ibv_create_qp() are:
max_send_sge = 1, max_recv_sge = 1, max_inline_data = 928
Upon return the capabilities are modified to the following
max_send_sge = 32, max_recv_sge = 1, max_inline_data = 928
Note decreasing the size of
Signed-off-by: Hal Rosenstock hal.rosenst...@gmail.com
---
diff --git a/opensm/opensm/osm_qos.c b/opensm/opensm/osm_qos.c
index 6bbbfa2..afa454f 100644
--- a/opensm/opensm/osm_qos.c
+++ b/opensm/opensm/osm_qos.c
@@ -196,7 +196,7 @@ static ib_api_status_t sl2vl_update_table(osm_sm_t * sm,
Signed-off-by: Hal Rosenstock hal.rosenst...@gmail.com
---
diff --git a/infiniband-diags/src/ibqueryerrors.c
b/infiniband-diags/src/ibqueryerrors.c
index 04842d5..0b320ec 100644
--- a/infiniband-diags/src/ibqueryerrors.c
+++ b/infiniband-diags/src/ibqueryerrors.c
@@ -270,7 +270,7 @@ static void
On 07:25 Sat 23 Jan , Hal Rosenstock wrote:
It does. There's a requirement that SL2VL mapping is required when VLCap
1.
Correct, I found it in o7-4. Basically we can check both (now after
switch/endport separation in the code this should be easier), but I
would prefer to
Hi Ira,
On 09:09 Mon 25 Jan , Ira Weiny wrote:
I did not pursue this further because all this code is changed with your
algorithm. So I did not clean this patch up further. As I said above this is
only test code to show where the differences I found lie.
Ok, so do you wish to not
On 07:26 Sat 23 Jan , Hal Rosenstock wrote:
Yes, and how is this different from the original patch?
The original patch didn't do this but missed accounting for the switch
port 0 configuration.
I cannot understand. What do you mean?
Remember, that there is a different configuration
On 17:54 Thu 21 Jan , Yevgeny Kliteynik wrote:
Since osm_sa_db_file_dump() doesn't have to actually dump
SA DB, reflecting this in the function return values.
Function returns 0 if the SA DB was actually dumped,
0 if there was no need to dump the SA DB and 0 if
some error occurred.
Hi everyone,
I finally found time to finish going over the base XRC patches. I fixed
the userspace can destroy an XRCD handle while still having QPs/SRQs
referring to it problem with a reference count, rather than the more
complicated method used by the original patches.
I think the next step
From: Jack Morgenstein ja...@dev.mellanox.co.il
Add support for associating an XRC domain to an inode so that XRC
domains can be shared between processes. We keep a per-device RB tree
of XRCDs (indexed by inode) in the uverbs module, and use struct
ib_xrcd's usecnt member to reference count
When we add support for sharing XRC domains among multiple processes,
we will need to keep track of which XRCD each QP and SRQ is associated
with, and keep a per-userspace-context reference count for each XRCD.
Userspace QPs already have a struct ib_uqp_object associated with
them, so we can put
From: Jack Morgenstein ja...@dev.mellanox.co.il
Add support for core userspace XRC operations (alloc/dealloc XRC
domain, create XRC SRQ), including adding an ABI for marshalling
requests and responses.
Signed-off-by: Jack Morgenstein ja...@dev.mellanox.co.il
Signed-off-by: Roland Dreier
From: Jack Morgenstein ja...@dev.mellanox.co.il
Add the core implementation for XRC (eXtended reliable connected)
transport. XRC provides better scalability by allowing senders to
specify which shared receive queue (SRQ) should be used to receive a
message, which essentially allows one transport
enum {
MLX4_FLAG_MSI_X = 1 0,
MLX4_FLAG_OLD_PORT_CMDS = 1 1,
+MLX4_FLAG_MASTER= 1 2,
+MLX4_FLAG_SLAVE = 1 3,
+MLX4_FLAG_SRIOV = 1 4,
};
I'm a bit confused by these flags. As far as I can tell, the SRIOV flag
will be set
Hi Sasha,
On Mon, Jan 25, 2010 at 12:46 PM, Sasha Khapyorsky sas...@voltaire.com wrote:
Hi Hal,
On 08:29 Mon 25 Jan , Hal Rosenstock wrote:
when using optimized programming and it's supported by the switch
Signed-off-by: Hal Rosenstock hal.rosenst...@gmail.com
---
diff --git
On Mon, Jan 25, 2010 at 10:20 AM, Sasha Khapyorsky sas...@voltaire.com wrote:
On 07:25 Sat 23 Jan , Hal Rosenstock wrote:
It does. There's a requirement that SL2VL mapping is required when VLCap
1.
Correct, I found it in o7-4. Basically we can check both (now after
when using optimized programming and it's supported by the switch
Signed-off-by: Hal Rosenstock hal.rosenst...@gmail.com
---
Changes since v1:
Add in update of vlarb tables for port 0 when optimized but no SL to VL
tables needed to be configured on that port
diff --git a/opensm/opensm/osm_qos.c
Hello,
I'm pretty unfamiliar with Infiniband, so apologies for the
stupid question.
Is there a limit on how many IB devices a system might support?
My colleague points out the following enum in uverbs_main.c:
enum {
IB_UVERBS_MAJOR = 231,
My colleague points out the following enum in uverbs_main.c:
enum {
IB_UVERBS_MAJOR = 231,
IB_UVERBS_BASE_MINOR = 192,
IB_UVERBS_MAX_DEVICES = 32
};
Experimentally, we've determined that on a system where we
plugged in
* Roland Dreier rdre...@cisco.com:
My colleague points out the following enum in uverbs_main.c:
enum {
IB_UVERBS_MAJOR = 231,
IB_UVERBS_BASE_MINOR = 192,
IB_UVERBS_MAX_DEVICES = 32
};
Experimentally, we've determined that
21 matches
Mail list logo