On 7/8/2015 12:42 PM, Jack Wang wrote:
We're using MLX OFED 2.4-1.0.4 together on top of 3.18.14.
So this list is for upstream things.. still, let's see
We hit bug below spontaneously, our test trigger this bug around 1 in 5 times.
and what is your test if I may ask?!
HCA 'mlx4_0'
CA
-Original Message-
From: Jason Gunthorpe [mailto:jguntho...@obsidianresearch.com]
Sent: Friday, July 03, 2015 5:38 PM
To: Wan, Kaike
Cc: linux-rdma@vger.kernel.org; Fleck, John; Weiny, Ira
Subject: Re: [PATCH v7 4/4] IB/sa: Route SA pathrecord query through netlink
On Tue, Jun 30,
On 7/8/2015 1:20 PM, 'Christoph Hellwig' wrote:
On Wed, Jul 08, 2015 at 01:05:28PM +0300, Sagi Grimberg wrote:
If we agree to consolidate on a single MR allocation API, I don't see
how this wrapper is moving us forward. But if you guys prefer to have it
than I don't have a hard objection.
On Wed, Jul 08, 2015 at 12:26:56PM +0530, Devesh Sharma wrote:
We (Emulex/Avago) were lobbied by the Open-Fabrics Alliance (OFA) to
change the licensing from just GPLv2 to a dual GPLv2/BSD license.
They would prefer the elements in the OFED stack all be dual licensed.
We're trying to move to
On Tue, Jul 07, 2015 at 09:05:15AM -0500, Steve Wise wrote:
I took the feedback from Christoph and Jason to mean I should remove
ib_get_dma_mr() entirely and pull its guts into
rdma_get_dma_mr(), and change all the users of ib_get_dma_mr() to use
rdma_get_dma_mr(). So the net result isn't a
On 7/8/2015 3:47 PM, Jack Wang wrote:
static void mlx4_ib_cq_comp(struct mlx4_cq *cq)
47 {
48 struct ib_cq *ibcq = to_mibcq(cq)-ibcq;
49 ibcq-comp_handler(ibcq, ibcq-cq_context);
50 }
Looks like cq use-after-free? I have no idea where.
see if you have in the code base you're using (why not
Hi Or,
We're testing our rdma kernel module, the tests is load module, create
RDMA connection, do some traffic, and unload module.
No mlx4_en involved, in fact we disable mlx4_en in kernel build,
because we don't need that.
I did some debug with gdb:
(gdb)list *mlx4_test_interrupts+0x84a
0xb0ea
Thanks for your time.
Looks the last one is missing in OFED 2.4 driver, I just checked the
history of mainline
commit bf1bac5b7882daa41249f85fbc97828f0597de5c
Author: Eli Cohen e...@dev.mellanox.co.il
Date: Thu Oct 23 15:57:27 2014 +0300
net/mlx4_core: Call synchronize_irq() before
Destroy multcast_idr on moudle exit, reclaiming the allocated memory.
This was detected by the following semantic patch (written by Luis Rodriguez
mcg...@suse.com)
SmPL
@ defines_module_init @
declarer name module_init, module_exit;
declarer name DEFINE_IDR;
identifier init;
@@
Destroy ocrdma_dev_id IDR on module exit, reclaiming the allocated memory.
This was detected by the following semantic patch (written by Luis Rodriguez
mcg...@suse.com)
SmPL
@ defines_module_init @
declarer name module_init, module_exit;
declarer name DEFINE_IDR;
identifier init;
@@
On Wed, Jul 8, 2015 at 5:07 PM, Jack Wang xjtu...@gmail.com wrote:
Looks the last one is missing in OFED 2.4 driver, I just checked the
history of mainline
commit bf1bac5b7882daa41249f85fbc97828f0597de5c
Author: Eli Cohen e...@dev.mellanox.co.il
Date: Thu Oct 23 15:57:27 2014 +0300
I am still not clear if all of us agree that we need it.
Sean and Steve had some disclaimers...
A single entry point doesn't help a whole lot if the app must deal with
different behavior based on how the API is used. We have a single entry point
for post_send, for example, and that makes
On Wed, Jul 08, 2015 at 04:15:00PM -0400, Doug Ledford wrote:
On 07/08/2015 04:02 PM, Christoph Hellwig wrote:
So how about someone tells OFED to stop trying to enforce this BS?
Unfortunately, simply not enforcing a bylaw of a multi-company
organization isn't really a valid option, you
On Wed, Jul 08, 2015 at 01:08:42PM -0600, Jason Gunthorpe wrote:
Then, what is left is all remote MRs and maybe it will be clearer what
to do about them then...
From looking at that for a while the APIs needed seem pretty simple
to me from a consumer perspective:
struct rdma_mr
On Mon, Jun 22, 2015 at 03:42:31PM +0300, Haggai Eran wrote:
+/**
+ * ib_get_net_dev_by_params() - Return the appropriate net_dev
+ * for a received CM request
+ * @dev: An RDMA device on which the request has been received.
+ * @port:Port number on the RDMA device.
+ * @pkey:The
On 07/08/2015 04:25 PM, Christoph Hellwig wrote:
On Wed, Jul 08, 2015 at 04:15:00PM -0400, Doug Ledford wrote:
On 07/08/2015 04:02 PM, Christoph Hellwig wrote:
So how about someone tells OFED to stop trying to enforce this BS?
Unfortunately, simply not enforcing a bylaw of a multi-company
On 07/08/2015 11:23 AM, Johannes Thumshirn wrote:
Destroy ocrdma_dev_id IDR on module exit, reclaiming the allocated memory.
Thanks, applied.
--
Doug Ledford dledf...@redhat.com
GPG KeyID: 0E572FDD
signature.asc
Description: OpenPGP digital signature
On 07/08/2015 12:23 AM, Maninder Singh wrote:
Hello,
+ for (i = 0; i dev-caps.num_ports; i++)
+ kfree(dm[i]);
goto out;
}
}
--
1.7.9.5
If you are going to change this, you might as well make it 100%
On 07/03/2015 11:38 AM, Weiny, Ira wrote:
Christoph,
Apologies, I misspoke in my response to you. There was a study of the code
and
we thought it was reasonable to post. However, in retrospect we should have
used more due diligence. We're going back to seek explicit consent from key
So how about someone tells OFED to stop trying to enforce this BS?
This just confirms my byass that Open-Fabrics Alliance are a bunch of
idiots making life hard, similar to all their horrible OFED driver
distributions that crated a total mess for everyone involved.
--
To unsubscribe from this
On 07/08/2015 04:02 PM, Christoph Hellwig wrote:
So how about someone tells OFED to stop trying to enforce this BS?
Unfortunately, simply not enforcing a bylaw of a multi-company
organization isn't really a valid option, you should know that. You
have to work to change the bylaw, which usually
On Mon, Jun 22, 2015 at 03:42:30PM +0300, Haggai Eran wrote:
An ib_client callback that is called with the lists_rwsem locked only for
read is protected from changes to the IB client lists, but not from
ib_unregister_device() freeing its client data. This is because
ib_unregister_device() will
anything wrong with the umem services provided by the IB core that
requires this implementation? what?
The current level of the API is mismatched with the PSM SDMA.
The ib_umem api:
- maps an SG list which isn't required by PSM since DMA mapping is done by the
low level SDMA
- the mapping
On 07/08/2015 04:01 AM, Christoph Hellwig wrote:
On Wed, Jul 08, 2015 at 12:26:56PM +0530, Devesh Sharma wrote:
We (Emulex/Avago) were lobbied by the Open-Fabrics Alliance (OFA) to
change the licensing from just GPLv2 to a dual GPLv2/BSD license.
They would prefer the elements in the OFED
On Wed, Jul 08, 2015 at 01:32:05PM -0700, 'Christoph Hellwig' wrote:
/* updates *sg if the SG couldn't be fully registered due to offsets */
int rdma_register_sg(struct rdma_mr *mr, struct scatterlist **sg,
u32 *pkey, u32 *offset, u32 *len);
plus an enum dma_data_direction
On 07/08/2015 11:21 AM, Johannes Thumshirn wrote:
Destroy multcast_idr on moudle exit, reclaiming the allocated memory.
Thanks, applied.
--
Doug Ledford dledf...@redhat.com
GPG KeyID: 0E572FDD
signature.asc
Description: OpenPGP digital signature
netlink is a reasonable low speed format to use for this kind of
serialization,
either via the common mux or via your own char device.
A couple of follow-up netlink questions.
1. I assume you are talking about generic netlink vs. say the RDMA netlink.
The generic netlink handles dynamic
On Wed, Jul 08, 2015 at 02:29:10PM -0600, Jason Gunthorpe wrote:
On Mon, Jun 22, 2015 at 03:42:30PM +0300, Haggai Eran wrote:
An ib_client callback that is called with the lists_rwsem locked only for
read is protected from changes to the IB client lists, but not from
ib_unregister_device()
On 07/08/2015 12:13 AM, Maninder Singh wrote:
o If allocation of dm fails, no need to free it.
o Free only allocated items.
I've taken the patch, but I reworked your commit message. The v1
version of the patch had a more correct commit message, but it could
have used a little rework as well.
so what's the role of the char-device? why should a low-level driver which is
part of the upstream RDMA stack contain a char-device?
Or.
Contains additional char devices:
- PSM character device
- diagnostic character devices
The nature of the hardware requires both of these additional
On Fri, Jul 3, 2015 at 9:08 PM, Weiny, Ira ira.we...@intel.com wrote:
Christoph,
Apologies, I misspoke in my response to you. There was a study of the code
and
we thought it was reasonable to post. However, in retrospect we should have
used more due diligence. We're going back to seek
Hi Christoph,
On Fri, Jul 3, 2015 at 9:22 PM, Christoph Hellwig h...@infradead.org wrote:
On Fri, Jul 03, 2015 at 03:38:55PM +, Weiny, Ira wrote:
Christoph,
Apologies, I misspoke in my response to you. There was a study of the
code and
we thought it was reasonable to post.
On 7/8/2015 12:36 AM, Jason Gunthorpe wrote:
On Tue, Jul 07, 2015 at 07:27:47PM +0300, Sagi Grimberg wrote:
Doesn't it look odd to you?
Sure, but the oddness is that rdma_device_access_flags exists at all,
not the wrapper. The wrapper is what we want the API to look like,
I don't
In mlx4_main.c:do_slave_init(), the function is supposed to queue up
each work struct. However, it checks to make sure the sriov support
isn't going down first. When it is going down, it doesn't queue up the
work struct, which results in us leaking the work struct at the end of
the function. As
On Wed, Jul 08, 2015 at 05:38:05PM -0400, Tom Talpey wrote:
On 7/8/2015 3:08 PM, Jason Gunthorpe wrote:
The MR stuff was never really designed, the HW people provided some
capability and the SW side just raw exposed it, thoughtlessly.
Jason, I don't disagree that the API can be improved. I
On Wed, Jul 08, 2015 at 01:32:05PM -0700, 'Christoph Hellwig' wrote:
On Wed, Jul 08, 2015 at 01:08:42PM -0600, Jason Gunthorpe wrote:
Then, what is left is all remote MRs and maybe it will be clearer what
to do about them then...
From looking at that for a while the APIs needed seem pretty
On 06/29/2015 09:57 AM, Hal Rosenstock wrote:
Persuant to Liran's comments on node_type on linux-rdma
mailing list:
In an effort to reform the RDMA core and ULPs to minimize use of
node_type in struct ib_device, an additional bit is added to
struct ib_device for is_switch (IB switch). This
On 07/07/2015 10:45 AM, Haggai Eran wrote:
Hi,
These two patches fix a couple of lockdep warnings I ran into, in IPoIB and
RDMA CM.
Regards,
Haggai
Haggai Eran (2):
IB/ucma: Fix lockdep warning in ucma_lock_files
IB/ipoib: Prevent lockdep warning in __ipoib_ib_dev_flush
Ummm.. Could we get some more descriptions as to what this code is for?
The next set with contain a great deal more background info in the cover letter.
Do we have a new OmniPath protocol here as well or is it IB? Which
standards are followed?
This will be covered in the additional
On 06/25/2015 12:04 PM, ira.we...@intel.com wrote:
From: Ira Weiny ira.we...@intel.com
The define OPA_LID_PERMISSIVE is big endian and was compared to cpu value
opa_drslid.
0-day build caught this while building with the OPA (hfi1) driver which was
recently sent to the list.
Fixes:
On Wed, Jul 08, 2015 at 10:32:44PM +, Marciniszyn, Mike wrote:
This sysfs entries are used by PSM2 to form packets from user space.
You didn't explain what's SC and what's SC-to-VL and why PSM2 can't talk to
the SM to query that.
Like IB the SL to SC to VL maps are available via SMA
On 07/02/2015 01:52 PM, Tatyana Nikolova wrote:
Fix for incorrect recording of the MAC address
Signed-off-by: Tatyana Nikolova tatyana.e.nikol...@intel.com
---
drivers/infiniband/hw/nes/nes_hw.c |2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
diff --git
On Mon, Jun 22, 2015 at 03:42:32PM +0300, Haggai Eran wrote:
+ if (net_dev) {
+ ipoib_warn(priv, matching net_dev found: %s\n,
+net_dev-name);
Is that a debug print?
+ default:
+ dev_warn(dev-dev, duplicate IP
On 07/02/2015 01:49 PM, Tatyana Nikolova wrote:
Neighbor resolution doesn't work without this fix
Signed-off-by: Tatyana Nikolova tatyana.e.nikol...@intel.com
Thanks, applied.
--
Doug Ledford dledf...@redhat.com
GPG KeyID: 0E572FDD
signature.asc
Description: OpenPGP
This sysfs entries are used by PSM2 to form packets from user space.
You didn't explain what's SC and what's SC-to-VL and why PSM2 can't talk to
the SM to query that.
SC stands for Service Channel and is a Fabric wide concept.
These tables are used to as to perform the following mapping:
On 07/02/2015 01:47 PM, Tatyana Nikolova wrote:
Fixes to allow clients to make remove mapping requests, after
they have provided the user space service with the mapping
information, they are using when the service is restarted.
1) Adding IWPM_REG_VALID, IWPM_REG_INCOMPL and IWPM_REG_UNDEF
On Wed, Jul 08, 2015 at 03:33:03PM -0400, Doug Ledford wrote:
I am not a lawyer, but this has been explained to me on numerous
occasions, so I relay the layman's interpretation here:
No, you don't always need everyone's approval. There are contributions
that are not legally copyright
On 7/8/2015 3:08 PM, Jason Gunthorpe wrote:
The MR stuff was never really designed, the HW people provided some
capability and the SW side just raw exposed it, thoughtlessly.
Jason, I don't disagree that the API can be improved. I have some
responses to your statements below though.
Why is
On Wed, Jul 08, 2015 at 10:29:56AM +0300, Sagi Grimberg wrote:
Sure, but the oddness is that rdma_device_access_flags exists at all,
not the wrapper. The wrapper is what we want the API to look like,
I don't necessarily agree. The API we'd want is a single API at all
the call sites to all
49 matches
Mail list logo
