On Tue, Sep 22, 2015 at 12:59 AM, Doug Ledford wrote:
> Here's the only matrix of IP checksumming that makes sense:
>
> 1) UD over IB (because it is one of the supported IPoIB types)
> 2) RC over IB (same as above)
> 3) Raw ETH over Eth (because IP over Eth makes sense and is a common
> type of p
Hi Sagi,
On 9/20/15 2:36 AM, Sagi Grimberg wrote:
Hi Santosh,
Nice to see this consolidaton happening. I too don't have access to
iWARP hardware for RDS test but will use this series and convert our WIP
IB fastreg code and see how it goes.
I'm very pleased to hear about this WIP. Please feel
> On Sep 20, 2015, at 3:52 AM, Sagi Grimberg wrote:
>
> On 9/17/2015 11:44 PM, Chuck Lever wrote:
>> The rb_send_bufs and rb_recv_bufs arrays are used to implement a
>> pair of stacks for keeping track of free rpcrdma_req and rpcrdma_rep
>> structs. Replace those arrays with free lists.
>>
>> T
> On Sep 21, 2015, at 3:33 AM, Devesh Sharma
> wrote:
>
> On Fri, Sep 18, 2015 at 2:15 AM, Chuck Lever wrote:
>> Pre-allocate extra send and receive Work Requests needed to handle
>> backchannel receives and sends.
>>
>> The transport doesn't know how many extra WRs to pre-allocate until
>> t
> On Sep 20, 2015, at 5:40 AM, Sagi Grimberg wrote:
>
> On 9/17/2015 11:46 PM, Chuck Lever wrote:
>> To support backward direction calls, I'm going to add an
>> svc_rdma_get_context() call in the client RDMA transport.
>>
>> Called from ->buf_alloc(), we can't sleep waiting for memory.
>> So ad
On Mon, Sep 21, 2015 at 01:03:52PM -0400, Doug Ledford wrote:
> On 09/21/2015 12:48 PM, Greg Kroah-Hartman wrote:
>
> > But, that's already not happening, as is obvious by my inbox.
>
> Yeah, I see that.
>
> > So, how about you forward on what you have so far to me, and I'll keep
> > these. Oth
On 09/21/2015 05:41 PM, Or Gerlitz wrote:
> On Fri, Sep 18, 2015 at 5:46 AM, Doug Ledford wrote:
>> On 09/16/2015 11:56 AM, Bodong Wang wrote:
>>> Signed-off-by: Bodong Wang
>>> ---
>>> drivers/infiniband/hw/mlx4/main.c | 3 +++
>>> 1 file changed, 3 insertions(+)
>>>
>>> diff --git a/drivers/in
On Fri, Sep 18, 2015 at 5:46 AM, Doug Ledford wrote:
> On 09/16/2015 11:56 AM, Bodong Wang wrote:
>> Signed-off-by: Bodong Wang
>> ---
>> drivers/infiniband/hw/mlx4/main.c | 3 +++
>> 1 file changed, 3 insertions(+)
>>
>> diff --git a/drivers/infiniband/hw/mlx4/main.c
>> b/drivers/infiniband/hw
This patch gets rid of struct ib_device_attr and cleans up drivers nicely.
It goes on top of my send_wr cleanups and the memory registration udpates
from Sagi.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majord...@vger.kernel.org
More major
On 9/21/2015 11:03 PM, Or Gerlitz wrote:
On Mon, Sep 21, 2015 at 11:03 PM, Or Gerlitz wrote:
On Mon, Sep 21, 2015 at 7:41 PM, Sagi Grimberg wrote:
Changes from v0:
- Replace xprtrdma patch to Chuck's one
- Fixed typo in iser modparam description
The change log of patch #3 needs isn't phrase
On 9/21/2015 8:03 PM, Chuck Lever wrote:
On Sep 21, 2015, at 9:41 AM, Sagi Grimberg wrote:
From: Chuck Lever
The core API has changed so that devices that do not have a global
DMA lkey automatically create an mr, per-PD, and make that lkey
available. The global DMA lkey interface is going a
On Mon, Sep 21, 2015 at 11:03 PM, Or Gerlitz wrote:
> On Mon, Sep 21, 2015 at 7:41 PM, Sagi Grimberg wrote:
>> Changes from v0:
>> - Replace xprtrdma patch to Chuck's one
>> - Fixed typo in iser modparam description
>
> The change log of patch #3 needs isn't phrased clearly, this is quick
> feedb
On Mon, Sep 21, 2015 at 7:41 PM, Sagi Grimberg wrote:
> Changes from v0:
> - Replace xprtrdma patch to Chuck's one
> - Fixed typo in iser modparam description
The change log of patch #3 needs isn't phrased clearly, this is quick
feedback I provided you for V0 and you didn't apply here, please do.
Unsignaled send WRs can get flushed as part of normal unmount, so don't
log them as warnings.
Signed-off-by: Steve Wise
---
net/sunrpc/xprtrdma/frwr_ops.c |7 +--
1 files changed, 5 insertions(+), 2 deletions(-)
diff --git a/net/sunrpc/xprtrdma/frwr_ops.c b/net/sunrpc/xprtrdma/frwr_ops
Otherwise a FRMR completion can cause a touch-after-free crash.
In xprt_rdma_destroy(), call rpcrdma_buffer_destroy() only after calling
rpcrdma_ep_destroy().
In rpcrdma_ep_destroy(), disconnect the cm_id first which should flush the
qp, then drain the cqs, then destroy the qp, and finally destro
The server rdma_read_chunk_lcl() and rdma_read_chunk_frmr() functions
were not taking into account the initial page_offset when determining
the rdma read length. This resulted in a read who's starting address
and length exceeded the base/bounds of the frmr.
Most work loads don't tickle this bug a
> On Sep 21, 2015, at 9:41 AM, Sagi Grimberg wrote:
>
> From: Chuck Lever
>
> The core API has changed so that devices that do not have a global
> DMA lkey automatically create an mr, per-PD, and make that lkey
> available. The global DMA lkey interface is going away in favor of
> the per-PD D
On 09/21/2015 12:48 PM, Greg Kroah-Hartman wrote:
> But, that's already not happening, as is obvious by my inbox.
Yeah, I see that.
> So, how about you forward on what you have so far to me, and I'll keep
> these. Otherwise you will end up with nasty merge issues very quickly
> as people will c
On Mon, Sep 21, 2015 at 11:42:28AM -0400, Doug Ledford wrote:
> On 09/18/2015 11:01 PM, Greg Kroah-Hartman wrote:
> > On Fri, Sep 18, 2015 at 11:51:09AM -0400, Doug Ledford wrote:
> >> On 09/16/2015 02:22 AM, Dan Carpenter wrote:
> >>> __get_txreq() returns an ERR_PTR() but this checks for NULL so
On 9/21/2015 7:39 PM, Jason Gunthorpe wrote:
On Sun, Sep 20, 2015 at 12:52:45PM +0300, Sagi Grimberg wrote:
This module parameter forces memory registration even for
a continuous memory region. It is true by default as sending
an all-physical rkey with remote permissions might be insecure.
Sign
This module parameter forces memory registration even for
a continuous memory region. It is true by default as sending
an all-physical rkey with remote permissions might be insecure.
Signed-off-by: Sagi Grimberg
---
drivers/infiniband/ulp/iser/iscsi_iser.c | 5 +
drivers/infiniband/ulp/ise
From: Chuck Lever
The core API has changed so that devices that do not have a global
DMA lkey automatically create an mr, per-PD, and make that lkey
available. The global DMA lkey interface is going away in favor of
the per-PD DMA lkey.
The per-PD DMA lkey is always available. Convert xprtrdma t
The Connect-IB device has a specific issue with memory registration using
the reserved lkey (device global_dma_lkey). This caused user-space memory
registration which usually uses cached pre-registered memory keys to fail
due to a device access error during registration. kernel-space memory
registr
Since mlx5 driver cannot rely on registration using the
reserved lkey (global_dma_lkey) it used to allocate a private
physical address lkey for each allocated pd.
Commit 96249d70dd70 ("IB/core: Guarantee that a local_dma_lkey
is available") just does it in the core layer so we can go ahead
and use
ConnectIB has some known issues with memory registration
using the local_dma_lkey (SEND, RDMA, RECV seems to work ok).
Thus don't expose support for it (and remove device->local_dma_lkey
setting).
since commit 96249d70dd70 ("IB/core: Guarantee that a local_dma_lkey
is available") addressed that by
On Sun, Sep 20, 2015 at 12:52:43PM +0300, Sagi Grimberg wrote:
> The Connect-IB device has a specific issue with memory registration using
> the reserved lkey (device global_dma_lkey). This caused user-space memory
> registration which usually uses cached pre-registered memory keys to fail
> due to
On Sun, Sep 20, 2015 at 12:52:45PM +0300, Sagi Grimberg wrote:
> This module parameter forces memory registration even for
> a continuous memory region. It is true by default as sending
> an all-physical rkey with remote permissions might be insecure.
>
> Signed-off-by: Sagi Grimberg
> drivers/i
On 9/20/2015 1:37 AM, Sagi Grimberg wrote:
On 9/20/2015 2:04 AM, Santosh Shilimkar wrote:
This series addresses RDS connection bottlenecks on massive workloads and
improve the RDMA performance almost by 3X. RDS TCP also gets a small gain
of about 12%.
RDS is being used in massive systems with h
> On Sep 21, 2015, at 1:51 AM, Devesh Sharma
> wrote:
>
> On Sun, Sep 20, 2015 at 4:05 PM, Sagi Grimberg
> wrote:
It is possible that in a given poll_cq
call you end up getting on 1 completion, the other completion is
delayed due to some reason.
>>>
>>>
>>> If a CQE is allowe
On 09/18/2015 11:01 PM, Greg Kroah-Hartman wrote:
> On Fri, Sep 18, 2015 at 11:51:09AM -0400, Doug Ledford wrote:
>> On 09/16/2015 02:22 AM, Dan Carpenter wrote:
>>> __get_txreq() returns an ERR_PTR() but this checks for NULL so it would
>>> oops on failure.
>>>
>>> Signed-off-by: Dan Carpenter
>>
Fixed warnings found by checkpatch.pl:
Please use a blank line after function/struct/union/enum declarations
FILE: drivers/staging/rdma/amso1100/c2_mq.c:158:
FILE: drivers/staging/rdma/hfi1/file_ops.c:2069:
FILE: drivers/staging/rdma/hfi1/sdma.c:744:
FILE: drivers/staging/rdma/hfi1/verbs.c
Fixes the following sparse warning:
drivers/staging/rdma/hfi1/device.c:127:12:
warning: symbol 'class_name_user' was not declared. Should it be static?
Signed-off-by: Geliang Tang
---
drivers/staging/rdma/hfi1/device.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drive
> Subject: [PATCH] staging/rdma/hfi1: do not use u8 to store a 32-bit integer
>
Thanks for the patch!
Acked-by: Mike Marciniszyn
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.ker
> Subject: [PATCH] IB/hfi1: class_name_user() should be static
>
> Fixes the following sparse warning:
> drivers/staging/rdma/hfi1/device.c:127:12:
> warning: symbol 'class_name_user' was not declared. Should it be static?
>
> Signed-off-by: Geliang Tang
> ---
Thanks for the patch!
Acked-b
On Thu, Aug 20, 2015 at 5:34 PM, Eran Ben Elisha wrote:
> Hi Doug,
>
> This patch-set adds a new implementation for multicast loopback prevention
> for
> mlx4 driver. The current implementation is very limited, especially if link
> layer is Ethernet. The new implementation is based on HW featur
The function can return negative value.
The problem has been detected using proposed semantic patch
scripts/coccinelle/tests/unsigned_lesser_than_zero.cocci [1].
[1]: http://permalink.gmane.org/gmane.linux.kernel/2038576
Signed-off-by: Andrzej Hajda
---
drivers/staging/rdma/ehca/ehca_cq.c | 13
Andrzej Hajda wrote:
> Semantic patch finds comparisons of types:
> unsigned < 0
> unsigned >= 0
> The former is always false, the latter is always true.
> Such comparisons are useless, so theoretically they could be
> safely removed, but their presence quite often indicates bugs.
Or som
> Subject: [PATCH] IB/hfi1: use kvfree() in sdma.c
>
> Use kvfree() instead of open-coding it.
>
> Signed-off-by: Geliang Tang
> ---
> drivers/staging/rdma/hfi1/sdma.c | 5 +
> 1 file changed, 1 insertion(+), 4 deletions(-)
Thanks for the patch.
Acked-by: Mike Marciniszyn
--
To unsubscri
If the lookup of a listening ID failed for an AF_IB request, the code
would try to call dev_put() on a NULL net_dev.
Fixes: be688195bd08 ("IB/cma: Fix net_dev reference leak with failed
requests")
Reported-by: Dan Carpenter
Signed-off-by: Haggai Eran
---
drivers/infiniband/core/cma.c | 2 +-
1
Use kvfree() instead of open-coding it.
Signed-off-by: Geliang Tang
---
drivers/staging/rdma/hfi1/sdma.c | 5 +
1 file changed, 1 insertion(+), 4 deletions(-)
diff --git a/drivers/staging/rdma/hfi1/sdma.c b/drivers/staging/rdma/hfi1/sdma.c
index aecd1a7..9c02a3d 100644
--- a/drivers/staging
On Fri, Sep 18, 2015 at 2:15 AM, Chuck Lever wrote:
> Pre-allocate extra send and receive Work Requests needed to handle
> backchannel receives and sends.
>
> The transport doesn't know how many extra WRs to pre-allocate until
> the xprt_setup_backchannel() call, but that's long after the WRs are
Looks good.
On Fri, Sep 18, 2015 at 2:15 AM, Chuck Lever wrote:
> xprtrdma's backward direction send and receive buffers are the same
> size as the forechannel's inline threshold, and must be pre-
> registered.
>
> The consumer has no control over which receive buffer the adapter
> chooses to cat
Looks good.
On Fri, Sep 18, 2015 at 2:14 AM, Chuck Lever wrote:
> Clean up: The error cases in rpcrdma_reply_handler() almost never
> execute. Ensure the compiler places them out of the hot path.
>
> No behavior change expected.
>
> Signed-off-by: Chuck Lever
> ---
> net/sunrpc/xprtrdma/rpc_rdm
Looks good, will test this ocrdma and update you.
On Fri, Sep 18, 2015 at 2:14 AM, Chuck Lever wrote:
> The core API has changed so that devices that do not have a global
> DMA lkey automatically create an mr, per-PD, and make that lkey
> available. The global DMA lkey interface is going away in
Looks Good.
On Fri, Sep 18, 2015 at 2:14 AM, Chuck Lever wrote:
> After adding a swapfile on an NFS/RDMA mount and removing the
> normal swap partition, I was able to push the NFS client well
> into swap without any issue.
>
> I forgot to swapoff the NFS file before rebooting. This pinned
> the N
On Sun, Sep 20, 2015 at 4:05 PM, Sagi Grimberg wrote:
>>> It is possible that in a given poll_cq
>>> call you end up getting on 1 completion, the other completion is
>>> delayed due to some reason.
>>
>>
>> If a CQE is allowed to be delayed, how does polling
>> again guarantee that the consumer ca
On Fri, Sep 18, 2015 at 7:49 PM, Chuck Lever wrote:
> Hi Devesh-
>
>
> On Sep 18, 2015, at 2:52 AM, Devesh Sharma
> wrote:
>
>> On Fri, Sep 18, 2015 at 2:14 AM, Chuck Lever wrote:
>>>
>>> Commit 8301a2c047cc ("xprtrdma: Limit work done by completion
>>> handler") was supposed to prevent xprtrdm
47 matches
Mail list logo
