rds-stress command with "-D" option failure.

rds-stress command with "-D" option failure.

[David Dai]

[PATCH 1/2] rds-tools user space part of change for rds-stress -D option failure
[PATCH 2/2] rds kernel space part of change for rds-stress -D option failure

Problem descrption:
The rds-stress command with "-D" optoin causes error.

Hardware: IBM Power 7 box.

OS: RHEL 6.7
# uname -a
Linux juno-ioxc1-lp5.austin.ibm.com 2.6.32-548.el6.ppc64 #1 SMP Sat Mar 28 00:59:47 EDT 2015 ppc64 ppc64 ppc64 GNU/Linux

rds-tools level: rds-tools-2.0.6-3.el6

To recreate:
Server side:
# rds-stress -r 10.1.1.136 -D 1K
waiting for incoming connection on 10.1.1.136:4000
accepted connection from 10.1.1.157:48021
negotiated options, tasks will start in 2 seconds
Starting up..sendto() truncated - 1..
tsks   tx/s   rx/s  tx+rx K/s    mbi K/s    mbo K/s tx us/c   rtt us cpu %
child pid 2762 exited with status 1

Client side:
# rds-stress -r 10.1.1.157 -s 10.1.1.136 -D 1K
connecting to 10.1.1.136:4000
negotiated options, tasks will start in 2 seconds
Starting up..sendto() truncated - 1..
tsks   tx/s   rx/s  tx+rx K/s    mbi K/s    mbo K/s tx us/c   rtt us cpu %
   1      0      0       0.00       0.00       0.00    0.00     0.00 -1.00
child pid 2560 exited with status 1
[root@jupiter-ioxc1-lp5 ~]# echo $?
1


After applying the proposed patch, here is the expected output:
# rds-stress -r 10.1.1.136 -D 1k
waiting for incoming connection on 10.1.1.136:4000
accepted connection from 10.1.1.131:37186
negotiated options, tasks will start in 2 seconds
Starting up....
tsks   tx/s   rx/s  tx+rx K/s    mbi K/s    mbo K/s tx us/c   rtt us cpu %
   1   5850   5710   12224.73    5774.04    5774.04    4.72   163.00 -1.00
   1   5874   5701   12239.82    5780.67    5781.67    4.74   162.37 -1.00
   1   5827   5662   12149.04    5737.79    5738.79    4.61   163.82 -1.00
   1   5860   5702   12230.70    5776.86    5776.86    4.64   162.59 -1.00
   1   5840   5668   12169.24    5747.83    5747.83    4.76   163.16 -1.00
   1   5880   5711   12257.07    5789.82    5788.82    4.80   162.02 -1.00
   1   5884   5726   12277.15    5798.80    5798.80    4.75   161.95 -1.00
   1   5823   5668   12151.27    5739.85    5738.85    4.64   163.80 -1.00
   1   5955   5780   12409.27    5860.71    5861.70    4.65   160.82 -1.00
   1   5799   5620   12075.02    5703.83    5702.83    4.85   164.70 -1.00
   1   5841   5705   12209.26    5765.74    5767.74    4.63   163.24 -1.00
   1   5871   5682   12216.68    5770.74    5769.74    4.62   162.50 -1.00
   1   5876   5720   12262.20    5791.74    5791.74    4.77   162.27 -1.00
   1   5861   5695   12219.90    5771.76    5771.76    4.62   162.41 -1.00
   1   5865   5693   12222.11    5772.81    5772.81    4.83   162.53 -1.00
---------------------------------------------
   1   5854   5690   12220.90    5772.20    5772.27    4.71   162.91 -1.00
(average)
[root@jupiter-ioxc1-lp6 /]#

Client side:
# rds-stress -r 10.1.1.131 -s 10.1.1.136 -D 1k
connecting to 10.1.1.136:4000
negotiated options, tasks will start in 2 seconds
Starting up....
tsks   tx/s   rx/s  tx+rx K/s    mbi K/s    mbo K/s tx us/c   rtt us cpu %
   1   5700   5841   12216.19    5770.51    5769.51    5.83   164.57 -1.00
   1   5684   5857   12215.29    5769.08    5770.08    6.01   165.13 -1.00
   1   5655   5820   12146.00    5737.36    5736.36    5.86   165.91 -1.00
   1   5703   5861   12240.11    5781.31    5781.31    5.88   164.41 -1.00
   1   5671   5841   12185.13    5755.34    5755.34    5.98   165.38 -1.00
   1   5710   5880   12267.62    5794.30    5794.30    6.01   164.19 -1.00
   1   5727   5884   12289.88    5804.32    5805.31    5.90   163.71 -1.00
   1   5654   5811   12135.33    5732.32    5731.32    5.82   165.97 -1.00
   1   5769   5944   12397.86    5856.31    5855.31    5.80   162.32 -1.00
   1   5630   5808   12106.92    5718.37    5719.37    6.07   166.76 -1.00
   1   5694   5828   12195.56    5760.29    5759.29    5.87   164.76 -1.00
   1   5671   5858   12203.10    576
Ctrl-C out
#

In addition to the proposed patch, it's recommened to run proposed patch with the following patch which is already pushed upstream:
http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=5c240fa2ab394af1dbde280e00cc038cbc7f0409

Please let us know if this patch is not submitted to the correct place.
And tell us the the the correct email list if you have it.

Signed-off-by: Carol L Soto <clsoto-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>
Tested-by: David Dai <zdai-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 1/2] rds-tools user space part of change for rds-stress -D option failure

[David Dai]

---
 rds-stress.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/rds-stress.c b/rds-stress.c
index 9f0623c..6632d58 100755
--- a/rds-stress.c
+++ b/rds-stress.c
@@ -1400,7 +1400,7 @@ static int recv_message(int fd,
 		return ret;
 	if (ret && ret < sizeof(struct header))
 		die("recvmsg() returned short data: %zd", ret);
-	if (msg.msg_namelen < sizeof(struct sockaddr_in))
+	if (ret && msg.msg_namelen < sizeof(struct sockaddr_in))
 		die("socklen = %d < sizeof(sin) (%zu)\n",
 		    msg.msg_namelen, sizeof(struct sockaddr_in));
 
-- 
1.7.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 1/2] rds-tools user space part of change for rds-stress -D option failure

[santosh shilimkar]

On 2/29/2016 5:47 PM, David Dai wrote:
> ---
>   rds-stress.c |    2 +-
>   1 files changed, 1 insertions(+), 1 deletions(-)
>
> diff --git a/rds-stress.c b/rds-stress.c
> index 9f0623c..6632d58 100755
> --- a/rds-stress.c
> +++ b/rds-stress.c
> @@ -1400,7 +1400,7 @@ static int recv_message(int fd,
>   		return ret;
>   	if (ret && ret < sizeof(struct header))
>   		die("recvmsg() returned short data: %zd", ret);
> -	if (msg.msg_namelen < sizeof(struct sockaddr_in))
> +	if (ret && msg.msg_namelen < sizeof(struct sockaddr_in))
>   		die("socklen = %d < sizeof(sin) (%zu)\n",
>   		    msg.msg_namelen, sizeof(struct sockaddr_in));
>
External rds-tools repo's are really stale. Which sources you
used to create this patch against ?

Regards,
Santosh
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 1/2] rds-tools user space part of change for rds-stress -D option failure

[David Dai]

Hi, Santosh:
   Pardon me, just check the mail box today.
   I used rds-tools-2.0.6-3.el6.src.rpm from RHEL 6,7 GA'ed source rpm. 

Thanks! - David

On Tue, 2016-03-01 at 11:03 -0800, santosh shilimkar wrote:
> On 2/29/2016 5:47 PM, David Dai wrote:
> > ---
> >   rds-stress.c |    2 +-
> >   1 files changed, 1 insertions(+), 1 deletions(-)
> >
> > diff --git a/rds-stress.c b/rds-stress.c
> > index 9f0623c..6632d58 100755
> > --- a/rds-stress.c
> > +++ b/rds-stress.c
> > @@ -1400,7 +1400,7 @@ static int recv_message(int fd,
> >   		return ret;
> >   	if (ret && ret < sizeof(struct header))
> >   		die("recvmsg() returned short data: %zd", ret);
> > -	if (msg.msg_namelen < sizeof(struct sockaddr_in))
> > +	if (ret && msg.msg_namelen < sizeof(struct sockaddr_in))
> >   		die("socklen = %d < sizeof(sin) (%zu)\n",
> >   		    msg.msg_namelen, sizeof(struct sockaddr_in));
> >
> External rds-tools repo's are really stale. Which sources you
> used to create this patch against ?
> 
> Regards,
> Santosh
> 



--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 1/2] rds-tools user space part of change for rds-stress -D option failure

[santosh shilimkar]

On 3/2/2016 4:04 PM, David Dai wrote:
> Hi, Santosh:
>     Pardon me, just check the mail box today.
>     I used rds-tools-2.0.6-3.el6.src.rpm from RHEL 6,7 GA'ed source rpm.
>
No worries. That rpm seems to have stale sources. I will try to get the
external git tree updated in near future.

For now, you can fetch the source RPM from below location. They
should work for you with upstream kernels.

http://public-yum.oracle.com/repo/OracleLinux/OL6/ofed_UEK/x86_64//getPackageSource/rds-tools-2.0.7-1.12.el6.src.rpm

Regards,
Santosh

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 2/2] rds kernel space part of change for rds-stress -D option failure

[David Dai]

---
 net/rds/rdma.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

diff --git a/net/rds/rdma.c b/net/rds/rdma.c
index b08ca60..d603645 100755
--- a/net/rds/rdma.c
+++ b/net/rds/rdma.c
@@ -657,6 +657,7 @@ int rds_cmsg_rdma_args(struct rds_sock *rs, struct rds_message *rm,
 		ret = rds_pin_pages(iov->addr, nr, pages, !op->op_write);
 		if (ret < 0)
 			goto out;
+		ret = 0;	
 
 		rdsdebug("RDS: nr_bytes %u nr %u iov->bytes %llu iov->addr %llx\n",
 			 nr_bytes, nr, iov->bytes, iov->addr);
-- 
1.7.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 2/2] rds kernel space part of change for rds-stress -D option failure

[Yuval Shaia]

Please check commit 1d2e3f39.

On Mon, Feb 29, 2016 at 07:47:26PM -0600, David Dai wrote:
> ---
>  net/rds/rdma.c |    1 +
>  1 files changed, 1 insertions(+), 0 deletions(-)
> 
> diff --git a/net/rds/rdma.c b/net/rds/rdma.c
> index b08ca60..d603645 100755
> --- a/net/rds/rdma.c
> +++ b/net/rds/rdma.c
> @@ -657,6 +657,7 @@ int rds_cmsg_rdma_args(struct rds_sock *rs, struct rds_message *rm,
>  		ret = rds_pin_pages(iov->addr, nr, pages, !op->op_write);
>  		if (ret < 0)
>  			goto out;
> +		ret = 0;	
>  
>  		rdsdebug("RDS: nr_bytes %u nr %u iov->bytes %llu iov->addr %llx\n",
>  			 nr_bytes, nr, iov->bytes, iov->addr);
> -- 
> 1.7.1
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
> the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 2/2] rds kernel space part of change for rds-stress -D option failure

[David Dai]

Hi, Yuval:
   Thanks for the commit number you pointed. 
   I am able to see the commit content. This is the same logic as I
submitted.

Thanks! - David
On Tue, 2016-03-01 at 20:55 +0200, Yuval Shaia wrote:
> Please check commit 1d2e3f39.
> 
> On Mon, Feb 29, 2016 at 07:47:26PM -0600, David Dai wrote:
> > ---
> >  net/rds/rdma.c |    1 +
> >  1 files changed, 1 insertions(+), 0 deletions(-)
> > 
> > diff --git a/net/rds/rdma.c b/net/rds/rdma.c
> > index b08ca60..d603645 100755
> > --- a/net/rds/rdma.c
> > +++ b/net/rds/rdma.c
> > @@ -657,6 +657,7 @@ int rds_cmsg_rdma_args(struct rds_sock *rs, struct rds_message *rm,
> >  		ret = rds_pin_pages(iov->addr, nr, pages, !op->op_write);
> >  		if (ret < 0)
> >  			goto out;
> > +		ret = 0;	
> >  
> >  		rdsdebug("RDS: nr_bytes %u nr %u iov->bytes %llu iov->addr %llx\n",
> >  			 nr_bytes, nr, iov->bytes, iov->addr);
> > -- 
> > 1.7.1
> > 
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
> > the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 



--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 2/2] rds kernel space part of change for rds-stress -D option failure

[santosh shilimkar]

Hi David,

On 2/29/2016 5:47 PM, David Dai wrote:
> ---
>   net/rds/rdma.c |    1 +
>   1 files changed, 1 insertions(+), 0 deletions(-)
>
> diff --git a/net/rds/rdma.c b/net/rds/rdma.c
> index b08ca60..d603645 100755
> --- a/net/rds/rdma.c
> +++ b/net/rds/rdma.c
> @@ -657,6 +657,7 @@ int rds_cmsg_rdma_args(struct rds_sock *rs, struct rds_message *rm,
>   		ret = rds_pin_pages(iov->addr, nr, pages, !op->op_write);
>   		if (ret < 0)
>   			goto out;
> +		ret = 0;	
>
This has been fixes already with many more things. Feel free to
test out the latest Linus's tip and do let me know if you see
any option to run RDMA.

Reagrds,
Santosh
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html