ments welcome !!
Santosh Shilimkar (12):
RDS: Drop stale iWARP RDMA transport
RDS: Add support for SO_TIMESTAMP for incoming messages
MAINTAINERS: update RDS entry
RDS: IB: Remove the RDS_IB_SEND_OP dependency
RDS: IB: Re-organise ibmr code
RDS: IB: create struct rds_ib_fmr
RDS: IB: mov
Add MR reuse statistics to RDS IB transport.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/ib.h | 2 ++
net/rds/ib_rdma.c | 7 ++-
net/rds/ib_stats.c | 2 ++
3 files changed, 10 insertions(+), 1 deletion(-)
diff --git a/net/rds/ib.h b/net/rds/ib.h
On 2/5/2016 10:13 AM, Masahiro Yamada wrote:
2016-02-06 2:56 GMT+09:00 santosh shilimkar :
On 2/5/2016 9:40 AM, Masahiro Yamada wrote:
KeyStone is Multi-platform. ARM_PATCH_PHYS_VIRT is already select'ed
by ARCH_MULTIPLATFORM.
Signed-off-by: Masahiro Yamada
---
There is also Key
On 2/5/2016 9:40 AM, Masahiro Yamada wrote:
KeyStone is Multi-platform. ARM_PATCH_PHYS_VIRT is already select'ed
by ARCH_MULTIPLATFORM.
Signed-off-by: Masahiro Yamada
---
There is also Keystone config so its nit redundant as you think.
P.S: arch/arm/configs/keystone_defconfig
Regards,
Sant
orii/Santosh, can you ACK/comment on this patch?
Reviewed-by: Grygorii Strashko
Acked-by: Santosh Shilimkar
-by: Santosh Shilimkar
Tested-by: Santosh Shilimkar
Signed-off-by: Sebastian Andrzej Siewior
---
v2:
Test and Ack tag added.
net/rds/page.c | 31 ++-
1 file changed, 6 insertions(+), 25 deletions(-)
diff --git a/net/rds/page.c b/net/rds/page.c
index 9005a2c..5a14e6d
On 12/10/2015 3:38 PM, Rob Herring wrote:
diff --git a/MAINTAINERS b/MAINTAINERS
index 69c8a9c..415b731 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
[..]
@@ -1202,6 +1209,7 @@ M:Santosh Shilimkar
L:linux-arm-ker...@lists.infradead.org (moderated for non-subscribers)
S
On 12/8/2015 1:34 PM, Arnd Bergmann wrote:
On Tuesday 08 December 2015 08:53:53 santosh shilimkar wrote:
On 12/8/2015 7:30 AM, Arnd Bergmann wrote:
The knav_qmss driver is currently broken when CONFIG_LPAE is
set, which is a bit surprising because I'd expect that any serious
users of
On 12/8/2015 7:30 AM, Arnd Bergmann wrote:
The knav_qmss driver is currently broken when CONFIG_LPAE is
set, which is a bit surprising because I'd expect that any serious
users of this platforms would have more than 2GB of RAM and require
LPAE.
Well it has been working with LPAE enable always.
On 11/25/2015 4:52 AM, Quentin Casasnovas wrote:
On Wed, Nov 25, 2015 at 12:21:45PM +, David Laight wrote:
From: Santosh Shilimkar
Sent: 24 November 2015 22:13
...
Sasha's found a NULL pointer dereference in the RDS connection code when
sending a message to an apparently unbound s
uot;)
Cc: David S. Miller
Cc: sta...@vger.kernel.org
Reviewed-by: Vegard Nossum
Reviewed-by: Sasha Levin
Acked-by: Santosh Shilimkar
Signed-off-by: Quentin Casasnovas
---
Dave,
Resending refreshed version from earlier patch which was missed out
for some reason.
https://lkml.org/lkml
On 11/24/2015 10:25 AM, Ben Hutchings wrote:
On Tue, 2015-11-03 at 12:25 +0100, Quentin Casasnovas wrote:
On Fri, Oct 16, 2015 at 10:47:49AM -0700, santosh shilimkar wrote:
On 10/16/2015 8:11 AM, Quentin Casasnovas wrote:
Sasha's found a NULL pointer dereference in the RDS connection
On 11/23/2015 8:54 AM, Murali Karicheri wrote:
Currently kernel crash randomly when K2L EVM is booted without
clk_ignore_unused in the bootargs. This workaround is not needed
on other K2 devices such as K2HK and K2E and with this fix, we can
remove the workaround altogether. netcp driver on K2L u
On 11/20/2015 3:39 PM, Michal Morawiec wrote:
Configure linking RAM for both queue managers also in case
when only linking RAM 0 is specified in device tree.
why ?
Signed-off-by: Michal Morawiec
---
drivers/soc/ti/knav_qmss_queue.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
di
On 11/20/2015 10:46 AM, Marc Zyngier wrote:
On 20/11/15 18:35, Grygorii Strashko wrote:
Hi Santosh,
On 11/20/2015 07:23 PM, santosh shilimkar wrote:
+ Thomas, Marc
On 11/20/2015 5:57 AM, Grygorii Strashko wrote:
Now the System stall is observed on TI AM437x based board
(am437x-gp-evm
regression for MPUIO interrupts")
Reported-by: Aaro Koskinen
Signed-off-by: Grygorii Strashko
---
Acked-by: Santosh Shilimkar
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at htt
ays-on"
Also if the clock-soucre can't tick in the low power states
then that device shouldn't be used as a clock-source.
CC: Arnd Bergmann
Cc: John Stultz
Cc: Felipe Balbi
Cc: Tony Lindgren
Cc: Santosh Shilimkar
Signed-off-by: Grygorii Strashko
---
Changes in v2:
- su
On 11/6/2015 6:46 AM, Carlos Hernandez wrote:
System V IPC is required by some LTP tests.
It is also enabled on omap2plus_defconfig which is the other base
config used for TI platforms.
Signed-off-by: Carlos Hernandez
---
Looks fine. I will pick this up.
Regards,
Santosh
--
To unsubscribe fr
On 11/3/2015 3:25 AM, Quentin Casasnovas wrote:
On Fri, Oct 16, 2015 at 10:47:49AM -0700, santosh shilimkar wrote:
On 10/16/2015 8:11 AM, Quentin Casasnovas wrote:
Sasha's found a NULL pointer dereference in the RDS connection code when
sending a message to an apparently unbound socket.
On 10/30/2015 4:15 AM, David Miller wrote:
From: Santosh Shilimkar
Date: Wed, 28 Oct 2015 15:51:12 -0700
+ &rs->rs_bound_node, ht_parms)) {
*port = rs->rs_bound_port;
the flexibility to grow the hashtbable based
on use and also comes up with inbuilt efficient bucket(chain) handling.
Reviewed-by: David Miller
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
v2: Dropped empty new line from rds_add_bound() (David Miller)
net/rds/af_rds.c
the flexibility to grow the hashtbable based
on use and also comes up with inbuilt efficient bucket(chain) handling.
Reviewed-by: David Miller
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/af_rds.c | 10 -
net/rds/bind.c | 126
fix in stable versions as
well. It seems to be applicable for all v2.6.32+ stable versions.
FWIW,
Acked-by: Santosh Shilimkar
Regards,
Santosh
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordo
On 10/20/2015 2:50 PM, Franklin S Cooper Jr wrote:
Newer Keystone 2 devices support CAN. Enable CAN support as modules.
Signed-off-by: Franklin S Cooper Jr
---
Again. Please shown me the usecases before adding config
option. DTS nodes have to be there.
arch/arm/configs/keystone_defconfig |
On 10/20/2015 2:50 PM, Franklin S Cooper Jr wrote:
Newer Keystone 2 devices support MMC and can boot from MMC. Therefore,
enable MMC support along with MMC OMAP HS whose driver is reused.
Signed-off-by: Franklin S Cooper Jr
---
Please notify once MMC patch is accepted and I will
this one then.
On 10/20/2015 2:50 PM, Franklin S Cooper Jr wrote:
From: Franklin Cooper
Unlike other Keystone 2 devices, newer Keystone 2 SOCs may utilize
pinmuxing which requires PINCTRL to be enabled. Therefore, enable
PINCTRL for all Keystone 2 devices.
Signed-off-by: Franklin S Cooper Jr
---
arch/arm/
On 10/20/2015 2:50 PM, Franklin S Cooper Jr wrote:
From: Lokesh Vutla
Enable omap_hsmmc for Keystone 2 architecture which reuses the HSMMC
IP found on OMAP platforms.
Signed-off-by: Franklin S Cooper Jr
---
drivers/mmc/host/Kconfig | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
dif
On 10/19/2015 10:02 AM, Murali Karicheri wrote:
On 10/19/2015 11:29 AM, santosh shilimkar wrote:
On 10/19/2015 8:28 AM, Murali Karicheri wrote:
On 10/16/2015 10:00 AM, Murali Karicheri wrote:
On 10/15/2015 02:59 PM, Murali Karicheri wrote:
When firmware image for PDSP firmware is absent in
Hi David L,
On 10/14/2015 2:15 PM, Santosh Shilimkar wrote:
From: Santosh Shilimkar
To further improve the RDS connection scalabilty on massive systems
where number of sockets grows into tens of thousands of sockets, there
is a need of larger bind hashtable. Pre-allocated 8K or 16K table is
On 10/18/2015 6:56 PM, David Miller wrote:
From: Santosh Shilimkar
Date: Wed, 14 Oct 2015 14:15:31 -0700
From: Santosh Shilimkar
To further improve the RDS connection scalabilty on massive systems
where number of sockets grows into tens of thousands of sockets, there
is a need of larger
On 10/19/2015 8:28 AM, Murali Karicheri wrote:
On 10/16/2015 10:00 AM, Murali Karicheri wrote:
On 10/15/2015 02:59 PM, Murali Karicheri wrote:
When firmware image for PDSP firmware is absent in the file system
the kernel boot with ramfs/nfs is stuck for 60 seconds being the
the default timeout.
transport like IB.
Reported-by: Sowmini Varadhan
Acked-by: Sowmini Varadhan
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
v2: Dropped the confusing SEND_LL_FULL check from v1
net/rds/send.c | 5 ++---
1 file changed, 2 insertions(+), 3 deletions(-)
diff --git a/net/r
On 10/16/2015 6:45 PM, Sowmini Varadhan wrote:
On (10/16/15 20:26), Santosh Shilimkar wrote:
diff --git a/net/rds/send.c b/net/rds/send.c
+ if (!test_bit(RDS_LL_SEND_FULL, &conn->c_flags))
+ queue_delayed_work(rds_wq, &conn->c_send_w, 0);
A minor note- it
transport like IB.
Reported-by: Sowmini Varadhan
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/send.c | 5 ++---
1 file changed, 2 insertions(+), 3 deletions(-)
diff --git a/net/rds/send.c b/net/rds/send.c
index ee49c25..7a377a1 100644
--- a/net/rds/send.c
+++ b
eating a
connection")
Signed-off-by: Quentin Casasnovas
Reviewed-by: Vegard Nossum
Reviewed-by: Sasha Levin
Cc: Vegard Nossum
Cc: Sasha Levin
Cc: Chien Yen
Cc: Santosh Shilimkar
Cc: David S. Miller
Cc: sta...@vger.kernel.org
---
Looks right. Am glad that we got deference issue as
On 10/15/2015 9:02 AM, Murali Karicheri wrote:
On 10/14/2015 11:41 AM, santosh shilimkar wrote:
10/14/2015 7:17 AM, Murali Karicheri wrote:
This patch series enable accumulator queue support for K2 SoCs.
Accumulator
queues are a type of qmss queue that is monitored by the PDSP firmware
and
From: Santosh Shilimkar
To further improve the RDS connection scalabilty on massive systems
where number of sockets grows into tens of thousands of sockets, there
is a need of larger bind hashtable. Pre-allocated 8K or 16K table is
not very flexible in terms of memory utilisation. The
10/14/2015 7:17 AM, Murali Karicheri wrote:
This patch series enable accumulator queue support for K2 SoCs. Accumulator
queues are a type of qmss queue that is monitored by the PDSP firmware and
accumulated. Host is interrupted by PDSP firmware when packets become
available in a ring buffer share
On 10/13/2015 9:14 AM, Murali Karicheri wrote:
Santosh,
On 10/13/2015 12:01 PM, santosh shilimkar wrote:
On 10/13/2015 6:56 AM, Murali Karicheri wrote:
On 10/12/2015 03:46 PM, Murali Karicheri wrote:
This patch series enable accumulator queue support for K2 SoCs.
Santosh, Arnd,
Could you
On 10/13/2015 6:56 AM, Murali Karicheri wrote:
On 10/12/2015 03:46 PM, Murali Karicheri wrote:
This patch series enable accumulator queue support for K2 SoCs.
Santosh, Arnd,
Could you please review and let me know if there is any comment. If
looks good, could you please merge to v4.4 next br
Nishant,
On 10/7/2015 12:56 PM, Nishanth Menon wrote:
On 10/07/2015 02:37 PM, Arnd Bergmann wrote:
On Wednesday 07 October 2015 14:28:09 Nishanth Menon wrote:
Add ramdisk support to allow for minimal kernel to be supported.
Signed-off-by: Nishanth Menon
I have not seen that in a while. Ca
On 10/7/2015 12:37 PM, Arnd Bergmann wrote:
On Wednesday 07 October 2015 14:28:09 Nishanth Menon wrote:
Add ramdisk support to allow for minimal kernel to be supported.
Signed-off-by: Nishanth Menon
I have not seen that in a while. Can you explain why the normal
initramfs method doesn't wor
sure not to loop
forever inside rds_send_xmit"). The idea is to process as big a
batch as we can but at the same time we don't hold other waiting
processes for send. Hence back-off after the send_batch_count
limit (1024) to avoid soft-lock ups.
Signed-off-by: Santosh Shilimkar
Signed-
f rcu for bind hash-table.
In subsequent patch, we also covert the global lock with per-bucket
lock to reduce the global lock contention.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/af_rds.c | 6 --
net/rds/bind.c | 35 +++--
Just in case we are still handling the QP receive completion while the
rds_ibdev is released, drop the connection instead of crashing the kernel.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/ib_cm.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff
times we are seeing connections stuck with
the LL_SEND_FULL bit getting set and never cleared.
We kick krdsd after any time we see -ENOMEM or -EAGAIN from the
ring allocation code.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/send.c| 10 ++
ne
.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/ib.h | 6 ++-
net/rds/ib_cm.c| 45 --
net/rds/ib_send.c | 110 +
net/rds/ib_stats.c | 1 -
net/rds/send.c | 1 +
5 files
blown IB performance and
am hoping to get that in mainline with help of them.
Santosh Shilimkar (14):
RDS: use kfree_rcu in rds_ib_remove_ipaddr
RDS: make socket bind/release locking scheme simple and more efficient
RDS: fix rds_sock reference bug while doing bind
RDS: Use per-bucket rw
All HCA drivers seems to popullate max_mr caps and few of
them do both max_mr and max_fmr.
Hence update RDS code to make use of max_mr.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/ib.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a
rds_ib_mr already keeps the pool handle which it associates
with. Lets use that instead of round about way of fetching
it from rds_ib_device.
No functional change.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/ib_rdma.c | 5 ++---
1 file changed, 2 insertions
RDS IB mr pool has its own workqueue 'rds_ib_fmr_wq', so we need
to use queue_delayed_work() to kick the work. This was hurting
the performance since pool maintenance was less often triggered
from other path.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
--
One need to take rds socket reference while using it and release it
once done with it. rds_add_bind() code path does not do that so
lets fix it.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/bind.c | 16 +++-
1 file changed, 11 insertions(+), 5
Fix below warning by marking rds_ib_fmr_wq static
net/rds/ib_rdma.c:87:25: warning: symbol 'rds_ib_fmr_wq' was not declared.
Should it be static?
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/ib_rdma.c | 2 +-
1 file changed, 1 insertion(+),
mutual exclusion locks are being performed.
In next patch, send completion is also split which re-uses the poll_cq()
and hence the code is moved to ib_cm.c
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/ib.h | 28 +--
net/rds/ib_cm.c| 70
hich could lead to 8k pull being exhausted, we fall-back
to 1m pool till 8k pool recovers for use.
This helps to at least push ~55 kB/s bidirectional data which
is a nice improvement.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/ib.c
e per-bucket lock to improve the scalability.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/af_rds.c | 2 ++
net/rds/bind.c | 47 ---
net/rds/rds.h| 1 +
3 files changed, 35 insertions(+), 15 deletions(-)
diff --
synchronize_rcu() slowing down un-necessarily the socket shutdown
path. It is used just kfree() the ip addresses in rds_ib_remove_ipaddr()
which is perfect usecase for kfree_rcu();
So lets use that to gain some speedup.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net
On 10/5/2015 10:56 AM, Santosh Shilimkar wrote:
Darn. Header hunk remained in the repo. Resending it.
From 4bebdd7a4d2960b2ff6c40b27156d041ea270765 Mon Sep 17 00:00:00 2001
From: Santosh Shilimkar
Date: Thu, 10 Sep 2015 11:57:14 -0700
Subject: [PATCH v3 05/14] RDS: defer the over_batch work
One need to take rds socket reference while using it and release it
once done with it. rds_add_bind() code path does not do that so
lets fix it.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/bind.c | 16 +++-
1 file changed, 11 insertions(+), 5
sure not to loop
forever inside rds_send_xmit"). The idea is to process as big a
batch as we can but at the same time we don't hold other waiting
processes for send. Hence back-off after the send_batch_count
limit (1024) to avoid soft-lock ups.
Signed-off-by: Santosh Shilimkar
Signed
times we are seeing connections stuck with
the LL_SEND_FULL bit getting set and never cleared.
We kick krdsd after any time we see -ENOMEM or -EAGAIN from the
ring allocation code.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/send.c| 10 ++
ne
synchronize_rcu() slowing down un-necessarily the socket shutdown
path. It is used just kfree() the ip addresses in rds_ib_remove_ipaddr()
which is perfect usecase for kfree_rcu();
So lets use that to gain some speedup.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net
Just in case we are still handling the QP receive completion while the
rds_ibdev is released, drop the connection instead of crashing the kernel.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/ib_cm.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff
e per-bucket lock to improve the scalability.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/af_rds.c | 2 ++
net/rds/bind.c | 47 ---
net/rds/rds.h| 1 +
3 files changed, 35 insertions(+), 15 deletions(-)
diff --
.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/ib.h | 6 ++-
net/rds/ib_cm.c| 45 --
net/rds/ib_send.c | 110 +
net/rds/ib_stats.c | 1 -
net/rds/send.c | 1 +
5 files
Fix below warning by marking rds_ib_fmr_wq static
net/rds/ib_rdma.c:87:25: warning: symbol 'rds_ib_fmr_wq' was not declared.
Should it be static?
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/ib_rdma.c | 2 +-
1 file changed, 1 insertion(+),
f rcu for bind hash-table.
In subsequent patch, we also covert the global lock with per-bucket
lock to reduce the global lock contention.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/af_rds.c | 6 --
net/rds/bind.c | 35 +++--
RDS IB mr pool has its own workqueue 'rds_ib_fmr_wq', so we need
to use queue_delayed_work() to kick the work. This was hurting
the performance since pool maintenance was less often triggered
from other path.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
--
rds_ib_mr already keeps the pool handle which it associates
with. Lets use that instead of round about way of fetching
it from rds_ib_device.
No functional change.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/ib_rdma.c | 5 ++---
1 file changed, 2 insertions
://git.kernel.org/pub/scm/linux/kernel/git/ssantosh/linux.git net/rds/4.3-v3
As a side note, the IB HCA driver I used for testing misses at least 3
important patches in upstream to see the full blown IB performance and
am hoping to get that in mainline with help of them.
Santosh Shilimkar (14)
hich could lead to 8k pull being exhausted, we fall-back
to 1m pool till 8k pool recovers for use.
This helps to at least push ~55 kB/s bidirectional data which
is a nice improvement.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/ib.c
mutual exclusion locks are being performed.
In next patch, send completion is also split which re-uses the poll_cq()
and hence the code is moved to ib_cm.c
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/ib.h | 28 +--
net/rds/ib_cm.c| 70
All HCA drivers seems to popullate max_mr caps and few of
them do both max_mr and max_fmr.
Hence update RDS code to make use of max_mr.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/ib.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a
On 10/5/2015 3:30 AM, David Miller wrote:
From: Santosh Shilimkar
Date: Wed, 30 Sep 2015 13:24:24 -0400
@@ -423,7 +423,9 @@ over_batch:
!list_empty(&conn->c_send_queue)) &&
send_gen == conn->c_send_gen) {
Nishant,
On 9/25/2015 10:38 AM, Nishanth Menon wrote:
On 09/25/2015 11:15 AM, santosh shilimkar wrote:
9/25/2015 9:01 AM, Nishanth Menon wrote:
[..]
Please refresh the series commit messages based on the
discussion so far and repost. Will pick it up then.
Thanks. I will do so (probably
, straightforward perf bug fixes
that align the RDS TCP socket with other parts of the kernel stack.
v2: fix kbuild-test-robot warnings, comments from Sergei Shtylov
and Santosh Shilimkar.
Acked-by: Santosh Shilimkar
--
To unsubscribe from this list: send the line "unsubscribe linux-kerne
synchronize_rcu() slowing down un-necessarily the socket shutdown
path. It is used just kfree() the ip addresses in rds_ib_remove_ipaddr()
which is perfect usecase for kfree_rcu();
So lets use that to gain some speedup.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net
Fix below warning by marking rds_ib_fmr_wq static
net/rds/ib_rdma.c:87:25: warning: symbol 'rds_ib_fmr_wq' was not declared.
Should it be static?
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/ib_rdma.c | 2 +-
1 file changed, 1 insertion(+),
hich could lead to 8k pull being exhausted, we fall-back
to 1m pool till 8k pool recovers for use.
This helps to at least push ~55 kB/s bidirectional data which
is a nice improvement.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/ib.c
rds_ib_mr already keeps the pool handle which it associates
with. Lets use that instead of round about way of fetching
it from rds_ib_device.
No functional change.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/ib_rdma.c | 5 ++---
1 file changed, 2 insertions
One need to take rds socket reference while using it and release it
once done with it. rds_add_bind() code path does not do that so
lets fix it.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/bind.c | 16 +++-
1 file changed, 11 insertions(+), 5
f rcu for bind hash-table.
In subsequent patch, we also covert the global lock with per-bucket
lock to reduce the global lock contention.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/af_rds.c | 6 --
net/rds/bind.c | 35 +++--
case where we reached to
over_batch code for some other reason so just retrying again
before giving up.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/send.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/net/rds/send.c b/net/rds/send.c
ind
RDS IB mr pool has its own workqueue 'rds_ib_fmr_wq', so we need
to use queue_delayed_work() to kick the work. This was hurting
the performance since pool maintenance was less often triggered
from other path.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
--
e per-bucket lock to improve the scalability.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/af_rds.c | 2 ++
net/rds/bind.c | 47 ---
net/rds/rds.h| 1 +
3 files changed, 35 insertions(+), 15 deletions(-)
diff --
times we are seeing connections stuck with
the LL_SEND_FULL bit getting set and never cleared.
We kick krdsd after any time we see -ENOMEM or -EAGAIN from the
ring allocation code.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/send.c| 10 ++
ne
.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/ib.h | 6 ++-
net/rds/ib_cm.c| 45 --
net/rds/ib_send.c | 110 +
net/rds/ib_stats.c | 1 -
net/rds/send.c | 1 +
5 files
From: Santosh Shilimkar
Just in case we are still handling the QP receive completion while the
rds_ibdev is released, drop the connection instead of crashing the kernel.
Signed-off-by: Santosh Shilimkar
---
net/rds/ib_cm.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a
mutual exclusion locks are being performed.
In next patch, send completion is also split which re-uses the poll_cq()
and hence the code is moved to ib_cm.c
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/ib.h | 28 +--
net/rds/ib_cm.c| 70
All HCA drivers seems to popullate max_mr caps and few of
them do both max_mr and max_fmr.
Hence update RDS code to make use of max_mr.
Signed-off-by: Santosh Shilimkar
Signed-off-by: Santosh Shilimkar
---
net/rds/ib.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a
kernel.org/pub/scm/linux/kernel/git/ssantosh/linux.git net/rds/4.3-v2
As a side note, the IB HCA driver I used for testing misses at least 3
important patches in upstream to see the full blown IB performance and
am hoping to get that in mainline with help of them.
Santosh Shilimkar (14):
RDS: use
On 9/30/2015 9:09 AM, Sowmini Varadhan wrote:
On (09/30/15 08:50), santosh shilimkar wrote:
rs_tcp = (struct rds_tcp_connection *)conn->c_transport_data;
- WARN_ON(!rs_tcp || rs_tcp->t_sock);
+ if (rs_tcp->t_sock && inet->inet_sadd
On 9/30/2015 8:58 AM, Sowmini Varadhan wrote:
On (09/30/15 08:50), santosh shilimkar wrote:
minor nit though not a strict rule. Just to be consistent based on
what we are following.
- core RDS patches "RDS:"
- RDS IB patches "RDS: IB:" or "RDS/IB:"
- RDS IW patch
about commit
reference in the change-log. You might want to fix that
for consistency.
Patch itself is fine.
Acked-by: Santosh Shilimkar
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo inf
S TCP performance.
Acked-by: Santosh Shilimkar
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
inet_saddr < inet->inet_daddr) {
+ struct sock *nsk = new_sock->sk;
Any reason you dropped the WARN_ON. Note that till we got commit
74e98eb0 (" RDS: verify the underlying transport exists before creating
a connection") merged, we had an issue. That guards it
file changed, 1 insertion(+), 1 deletion(-)
Acked-by: Santosh Shilimkar
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
+Greg, arm-sock folks
On 9/17/2015 10:56 AM, santosh shilimkar wrote:
On 9/17/2015 10:41 AM, Luis de Bethencourt wrote:
Hello,
These patches add the missing MODULE_DEVICE_TABLE() for OF to export
the information so modules have the correct aliases built-in and
autoloading works correctly.
A
9/25/2015 9:01 AM, Nishanth Menon wrote:
On 09/25/2015 10:18 AM, santosh shilimkar wrote:
On 9/25/2015 7:50 AM, Nishanth Menon wrote:
[...]
But, how about userspace
needing to know which SoC they are on, without needing to depend on
board->soc mapping? How do we help resolve that?
Why
On 9/25/2015 7:50 AM, Nishanth Menon wrote:
On 09/24/2015 10:54 AM, Murali Karicheri wrote:
[...]
ti,omap3 is the family of omap3 devices similar to keystone. ti,omap3450
is required if there is an exceptional treatment required for ti,omap3450.
In keystone case so far there is no case of excep
401 - 500 of 1304 matches
Mail list logo