This patch ports:
diff-sched-rework-_proc_stat-output
diff-sched-fix-output-of-vestat-idle
diff-sched-make-allowance-for-vcpu-rate-in-_proc_stat
diff-sched-hide-steal-time-from-inside-CT
diff-sched-cpu.proc.stat-always-count-nr_running-and-co-on-all-cpus
Author: Vladimir Davydov
Email: vdavy...@p
In contrast to RH6, where tg->cpustat was used, now cpu stats are
accounted in the cpuacct cgroup. So zap tg->cpustat and use
cpuacct->cpustat for showing cpu.proc.stat instead. Fortunately cpu and
cpuacct cgroups are always mounted together (even by systemd by
default), so this will work.
Signed-
This reverts commit 6071473d0440fcfd128f3243dfb82d19f6aef668.
The above-mentioned commit dramatically complicates porting of cpu acct
patches from RH6, so revert it. The next patch will fix cpu accounting
once again.
Signed-off-by: Vladimir Davydov
Conflicts:
kernel/sched/core.c
---
dr
Acked-by: Andrew Vagin
I'm agree that we need to remove this function, but I don't know how it
fixes the bug.
On Fri, May 29, 2015 at 04:53:39PM +0300, Kirill Tkhai wrote:
> This is a leftover from earlier versions of PCS,
> and we do not need that functionality in 3.10.
>
> It was the reason o
Packet sockets have incoming queue of packets that is only limited
with per-socket wmem buffer. Strictly speaking we should sum up
all the queues and charge them into kmem once new packet arrives,
but this will result in huge patch. Since there's typically quite
a few of packet sockets in container
So, this one is tricky. Right now most (all but one place) of the
memory allocations in netlink code happen in process context and
are done via kmalloc/slub. Thus they are auto-accounted into kmem.
The single exceptional place is in netlink_alloc_large_skb where
big sending packets are allocated w
For unix sockets there's no such thing as "read buffers" as all
the data is accounted on the send paths. Fortunate enough most
of the stuff is already kmem-auto-charged except one thing --
paged dgram skbs.
Signed-off-by: Pavel Emelyanov
---
net/core/sock.c | 2 +-
1 file changed, 1 insertion(+
Right now UDP outgoing traffic is kmem-auto-charged into cg
kmem. Incoming traffic is not, but it has tcp-like memory
scheduler (but simpler, with just one limit). So here's the
per-cgroup UDP read buffers limiting in the same was as TCP
is done.
Signed-off-by: Pavel Emelyanov
---
include/net/
Kernel limits the total number of orphan TCP sockets in the
system. One container can eat all this limit and make others'
container to suffer from TCP state machine breakage.
Thus here's the per-CT limit on the number of prphans that
doesn't affect the others. The limit is set to be 1/4-th
of the
TCP code already has internal memory management for both -- in
and out traffic. The outgoing packets are also already auto
accounted into kmem (and into cg memory), incoming traffic is
not accounted into kmem. And this management is already per-cg
thanks to Glauber work some time ago.
So TCP mm fi
The previous BC approach was based on two ideas. The first is
that we know the maximum amount of sockets and thair buffers a
container needs. The second was about "poll semantics" i.e.
when a poll() reports "writable", the write/send system call
cannot block on buffer limit hit. To address the latt
Hi,
This is the set that replaces old limit-based network BC code with
simpler approach -- all the socket buffers are just charged into
container's memory.
Such things as TCP memory/window management is already per-CG upstream,
so we'll just re-use this code.
-- Pavel
___
This is a leftover from earlier versions of PCS,
and we do not need that functionality in 3.10.
It was the reason of the kernel panic in 2.6.32:
https://jira.sw.ru/browse/PSBM-33755, in the test
of forced CT destroying.
Also, tcp_v4_kill_ve_sockets() looks as the only
user of synchronize_net(), s
And diff-sched-fix-sched_getaffinity-on-unexisting-tasks-inside-VE,
which is a quickfix for the original patch.
Author: Kirill Tkhai
Email: ktk...@parallels.com
Subject: sched: Return only virtual cpus in sched_getaffinity()
Date: Mon, 28 Apr 2014 12:45:38 +0400
Fill mask using virtual cpus which
Author: Vladimir Davydov
Email: vdavy...@parallels.com
Subject: sched: clear prev entity if curr is dequeued
Date: Fri, 20 Sep 2013 16:55:23 +0400
cfs_rq->prev is used for ctxsw accounting: on put_prev_entity()
cfs_rq->prev is set to curr if curr is on rq, and on set_next_entity()
nr_switches is i
This patch is already upstream, but the feature is disabled by default.
This patch only enables it.
Author: Vladimir Davydov
Email: vdavy...@parallels.com
Subject: sched: Increase SCHED_LOAD_SCALE resolution
Date: Tue, 25 Dec 2012 13:33:21 +0400
Mainstream commit c8b281161dfa4bb5d5be63fb036ce1934
Author: Vladimir Davydov
Email: vdavy...@parallels.com
Subject: sched: initialize runtime to non-zero on cfs bw set
Date: Mon, 21 Jan 2013 11:44:58 +0400
* [sched] running tasks could be throttled and never unthrottled
thus causing random node hangs. (PSBM-17658)
If cfs_rq->runtime_remain
Author: Vladimir Davydov
Email: vdavy...@parallels.com
Subject: sched: do not account iothrottled tasks in loadavg
Date: Mon, 22 Oct 2012 14:27:25 +0400
Changes from v1:
* do not account tasks throttled while doing reclaim too
Signed-off-by: Vladimir Davydov
Acked-by: Konstantin Khlebnikov
===
The commit is pushed to "branch-rh7-3.10.0-123.1.2-ovz" and will appear at
https://src.openvz.org/scm/ovz/vzkernel.git
after rh7-3.10.0-123.1.2.vz7.5.7
-->
commit b103b8fdcd475a79e13c3e8f5325c92da90fd401
Author: Vladimir Davydov
Date: Fri May 29 17:01:34 2015 +0400
ub/netfilter: accoun
The commit is pushed to "branch-rh7-3.10.0-123.1.2-ovz" and will appear at
https://src.openvz.org/scm/ovz/vzkernel.git
after rh7-3.10.0-123.1.2.vz7.5.7
-->
commit e5f41176c3b8653ff237f1fd695d608a176aa46b
Author: Cyrill Gorcunov
Date: Fri May 29 16:50:55 2015 +0400
cgroup: Mangle cgroup
The commit is pushed to "branch-rh7-3.10.0-123.1.2-ovz" and will appear at
https://src.openvz.org/scm/ovz/vzkernel.git
after rh7-3.10.0-123.1.2.vz7.5.7
-->
commit bd28914a36ef98c893dbeb269a0bd4859151936e
Author: Cyrill Gorcunov
Date: Fri May 29 16:50:49 2015 +0400
cgroup: mount -- Disa
Acked-by: Andrew Vagin
On Thu, May 28, 2015 at 05:43:18PM +0300, Vladimir Davydov wrote:
> This patch ports the code accounting netfilter/x_tables to ub
> (UB_NUMXTENT) from RH6.
>
> Related to https://jira.sw.ru/browse/PSBM-20089
>
> Signed-off-by: Vladimir Davydov
> ---
> include/linux/netf
On Fri, May 29, 2015 at 03:07:07PM +0300, Andrew Vagin wrote:
> > +static int recharge_xtables(struct xt_table_info *new, struct
> > xt_table_info *old)
> > +{
> > + struct user_beancounter *ub, *old_ub;
> > + long change;
> > +
> > + ub = new->ub;
> > + old_ub = old->number ? old->ub : ub
On Thu, May 28, 2015 at 05:43:18PM +0300, Vladimir Davydov wrote:
> This patch ports the code accounting netfilter/x_tables to ub
> (UB_NUMXTENT) from RH6.
>
> Related to https://jira.sw.ru/browse/PSBM-20089
>
> Signed-off-by: Vladimir Davydov
> ---
> include/linux/netfilter/x_tables.h | 4 +++
The commit is pushed to "branch-rh7-3.10.0-123.1.2-ovz" and will appear at
https://src.openvz.org/scm/ovz/vzkernel.git
after rh7-3.10.0-123.1.2.vz7.5.7
-->
commit 901bf5dd708537e483a90ebcf62af57239e63702
Author: Vladimir Davydov
Date: Fri May 29 15:49:38 2015 +0400
fairsched: Port diff
The commit is pushed to "branch-rh7-3.10.0-123.1.2-ovz" and will appear at
https://src.openvz.org/scm/ovz/vzkernel.git
after rh7-3.10.0-123.1.2.vz7.5.7
-->
commit 72bcfd69eb80e5ca7624a6c0665494990fd06faf
Author: Vladimir Davydov
Date: Fri May 29 15:45:19 2015 +0400
ve: remove fairsched
Author: Konstantin Khlebnikov
Email: khlebni...@openvz.org
Subject: fairsched: optimize sys_fairsched_mvpr()
Date: Wed, 08 May 2013 13:59:04 +0400
Remove task_list_lock and optimize for current.
https://jira.sw.ru/browse/PCLIN-26766
Signed-off-by: Konstantin Khlebnikov
=
Currently, we try and fail to remove the fairsched node of a UUID-named
container from the kernel on stop:
CT: 956ebfc3-3ca9-44e0-9739-ab8abbe50edc: started
Can't remove fairsched node 1073741823 err=-2
CT: 956ebfc3-3ca9-44e0-9739-ab8abbe50edc: stopped
We should only do that for containers
The commit is pushed to "branch-rh7-3.10.0-123.1.2-ovz" and will appear at
https://src.openvz.org/scm/ovz/vzkernel.git
after rh7-3.10.0-123.1.2.vz7.5.7
-->
commit c238509b21ac68629812f1a27d7c4e95a75a815a
Author: Vladimir Davydov
Date: Fri May 29 14:53:11 2015 +0400
ms/memcg: do not cal
The commit is pushed to "branch-rh7-3.10.0-123.1.2-ovz" and will appear at
https://src.openvz.org/scm/ovz/vzkernel.git
after rh7-3.10.0-123.1.2.vz7.5.7
-->
commit 7b768085e56d53826162aef58ace5f58adc1b4c9
Author: Vladimir Davydov
Date: Fri May 29 14:53:03 2015 +0400
fairsched: fix outpu
В Чт, 28/05/2015 в 18:56 +0300, Andrew Vagin пишет:
> On Wed, May 27, 2015 at 02:32:22PM +0300, Kirill Tkhai wrote:
> > hold_net() doesn't increment net refcounter if NETNS_REFCNT_DEBUG
> > is not defined. In this case inet_twdr_do_twkill_work() may happen
> > after network is destoyed and lead to
On Tue, May 26, 2015 at 06:00:52PM +0300, Cyrill Gorcunov wrote:
> We're bindmounting cgroups for container so if say a container
> is having CTID=200 then @cgroups and @mountinfo output will
> contain /200 as a root. Which makes Docker to lookup for
> appropriate directory inside /sys/fs/cgroup/
>
On Fri, May 29, 2015 at 11:30:00AM +0300, Cyrill Gorcunov wrote:
> On Fri, May 29, 2015 at 11:18:52AM +0300, Vladimir Davydov wrote:
> > Hi Cyrill,
> >
> > On Tue, May 26, 2015 at 06:00:52PM +0300, Cyrill Gorcunov wrote:
> > > We're bindmounting cgroups for container so if say a container
> > > is
The commit is pushed to "branch-rh7-3.10.0-123.1.2-ovz" and will appear at
https://src.openvz.org/scm/ovz/vzkernel.git
after rh7-3.10.0-123.1.2.vz7.5.7
-->
commit 12d5e82b5955aab42daf5b688d2cba662de8111b
Author: Vladimir Davydov
Date: Fri May 29 12:36:01 2015 +0400
ub: ub_dirty_limits:
The commit is pushed to "branch-rh7-3.10.0-123.1.2-ovz" and will appear at
https://src.openvz.org/scm/ovz/vzkernel.git
after rh7-3.10.0-123.1.2.vz7.5.7
-->
commit 2d3d5fbf8c45b0e297ff429f0fbaa15f8e35d647
Author: Vladimir Davydov
Date: Fri May 29 12:35:55 2015 +0400
shmem: fix tmpfs_ram
The commit is pushed to "branch-rh7-3.10.0-123.1.2-ovz" and will appear at
https://src.openvz.org/scm/ovz/vzkernel.git
after rh7-3.10.0-123.1.2.vz7.5.7
-->
commit a67159e9a53cc3328df60c472c3cbb850c6a5fbb
Author: Vladimir Davydov
Date: Fri May 29 12:35:45 2015 +0400
memcg: add function
Hi Cyrill,
On Tue, May 26, 2015 at 06:00:52PM +0300, Cyrill Gorcunov wrote:
> We're bindmounting cgroups for container so if say a container
> is having CTID=200 then @cgroups and @mountinfo output will
> contain /200 as a root. Which makes Docker to lookup for
> appropriate directory inside /sys/
On Tue, May 26, 2015 at 06:00:51PM +0300, Cyrill Gorcunov wrote:
> Even mounting knowing cgroups (ie ones which already known to VE and
> been mounted by vzctl or any other tool for containter sake) is not
> that harmless as it might look like. In particular this introduce
> additional performance
The commit is pushed to "branch-rh7-3.10.0-123.1.2-ovz" and will appear at
https://src.openvz.org/scm/ovz/vzkernel.git
after rh7-3.10.0-123.1.2.vz7.5.7
-->
commit d2e9d1ba7e3acc37c18ae91a11df1fb5bba2972c
Author: Kirill Tkhai
Date: Fri May 29 12:02:00 2015 +0400
ve/kmod: Add rules for n
The commit is pushed to "branch-rh7-3.10.0-123.1.2-ovz" and will appear at
https://src.openvz.org/scm/ovz/vzkernel.git
after rh7-3.10.0-123.1.2.vz7.5.7
-->
commit 5f6dbce004ffb21b500d930b46d2b85287619f6d
Author: Kirill Tkhai
Date: Fri May 29 12:01:52 2015 +0400
ve/kmod: Add rules for a
The commit is pushed to "branch-rh7-3.10.0-123.1.2-ovz" and will appear at
https://src.openvz.org/scm/ovz/vzkernel.git
after rh7-3.10.0-123.1.2.vz7.5.7
-->
commit 4038cd0e029ddee1c3308216bdc5da6c4485656b
Author: Vladimir Davydov
Date: Fri May 29 11:55:46 2015 +0400
ms/memcg: port memor
The commit is pushed to "branch-rh7-3.10.0-123.1.2-ovz" and will appear at
https://src.openvz.org/scm/ovz/vzkernel.git
after rh7-3.10.0-123.1.2.vz7.5.7
-->
commit 24dbf07906c8e288ed18f895f49c36dc850213bb
Author: Vladimir Davydov
Date: Fri May 29 11:55:40 2015 +0400
ms/memcg: use CFTYPE
The commit is pushed to "branch-rh7-3.10.0-123.1.2-ovz" and will appear at
https://src.openvz.org/scm/ovz/vzkernel.git
after rh7-3.10.0-123.1.2.vz7.5.7
-->
commit 9eafb07898336d06d2003d45e8c682837be2ae2e
Author: Vladimir Davydov
Date: Fri May 29 11:55:35 2015 +0400
ms/memcg: rename mem
43 matches
Mail list logo
