Re: [Devel] [PATCH RH7 0/2]

[Pavel Tikhomirov]

And please don't forget vX tag to easier distinguish patches between 
versions.

On 1/13/20 11:25 AM, Valeriy Vdovin wrote:
> This patch addresses the problem of varying views on start time
> of a resumed process between container and host. From a container
> point of view the resumed process is the same old process with
> the same old start time, which hasn't changed after suspend/resume.
>  From a host point of view the resumed process is just a new one
> with a new start_time. Both the host and the container want to
> see start time that is only relevant to their context.
> 
> To support that we should teach the output of of /proc/pid/stat
> to show container-related value from container context and host-
> related value from host context.
> 
> The patch introduces new field to task_struct that stores container-
> related start time value. It will be seen in /proc/pid/stat in case
> if it was requested by a task, that's in ve cgroup.
> Setting this field can be done with prctl only from host or a
> container task with pseudosuper flag set.
> 
> At suspend this value is dumped by criu and at restore it's set back
> from a dump by criu.
> 
> Valeriy Vdovin (1):
>ve/proc: Added separate start time field to task_struct to show in
>  container
> 
>   fs/proc/array.c| 16 
>   include/linux/sched.h  |  5 +
>   include/uapi/linux/prctl.h |  7 +++
>   kernel/fork.c  | 15 +++
>   kernel/sys.c   | 23 +++
>   kernel/ve/ve.c | 22 ++
>   6 files changed, 76 insertions(+), 12 deletions(-)
> 
> (2):
>dump/restore: Maintain proper start_time param from
>   criu/cr-dump.c | 49 
>   criu/cr-restore.c  | 68 
> ++
>   criu/include/crtools.h | 31 +++
>   images/core.proto  |  2 ++
>   4 files changed, 123 insertions(+), 27 deletions(-)
> 

-- 
Best regards, Tikhomirov Pavel
Software Developer, Virtuozzo.

___
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel

Re: [Devel] [PATCH RH7 0/2]

[Pavel Tikhomirov]

Empty commit message header. Please remain the same name across 
versions, e.g. "Start time of a task inside a container"

On 1/13/20 11:25 AM, Valeriy Vdovin wrote:
> This patch addresses the problem of varying views on start time
> of a resumed process between container and host. From a container
> point of view the resumed process is the same old process with
> the same old start time, which hasn't changed after suspend/resume.
>  From a host point of view the resumed process is just a new one
> with a new start_time. Both the host and the container want to
> see start time that is only relevant to their context.
> 
> To support that we should teach the output of of /proc/pid/stat
> to show container-related value from container context and host-
> related value from host context.
> 
> The patch introduces new field to task_struct that stores container-
> related start time value. It will be seen in /proc/pid/stat in case
> if it was requested by a task, that's in ve cgroup.
> Setting this field can be done with prctl only from host or a
> container task with pseudosuper flag set.
> 
> At suspend this value is dumped by criu and at restore it's set back
> from a dump by criu.
> 
> Valeriy Vdovin (1):
>ve/proc: Added separate start time field to task_struct to show in
>  container
> 
>   fs/proc/array.c| 16 
>   include/linux/sched.h  |  5 +
>   include/uapi/linux/prctl.h |  7 +++
>   kernel/fork.c  | 15 +++
>   kernel/sys.c   | 23 +++
>   kernel/ve/ve.c | 22 ++
>   6 files changed, 76 insertions(+), 12 deletions(-)
> 
> (2):
>dump/restore: Maintain proper start_time param from
>   criu/cr-dump.c | 49 
>   criu/cr-restore.c  | 68 
> ++
>   criu/include/crtools.h | 31 +++
>   images/core.proto  |  2 ++
>   4 files changed, 123 insertions(+), 27 deletions(-)
> 

-- 
Best regards, Tikhomirov Pavel
Software Developer, Virtuozzo.

___
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel

Re: [Devel] [PATCH rh7 0/2] bc: rework "numproc" ubc limit to be based on pids cgroup

[Andrey Ryabinin]

On 06/25/2018 06:24 PM, Konstantin Khorenko wrote:
> It's time to rework our home brew ubc "numproc" limit to be based on stock 
> pids
> cgroup.
> 
> Note: the semantic will change a bit: now we limit number of possible threads
> in a Container while old ubc "numproc" limited number of tasks.
> 
> Note1: vzctl update is required.
> 
> https://jira.sw.ru/browse/PSBM-86044
> 
> Konstantin Khorenko (2):
>   bc: roll back beancounters-based numer of processes accounting
>   bc: implement ubc "numproc" limit based on pids cgroup
> 
>  include/bc/beancounter.h | 10 ++
>  include/bc/misc.h|  2 --
>  kernel/bc/beancounter.c  | 43 +--
>  kernel/bc/misc.c | 10 --
>  kernel/bc/proc.c |  2 ++
>  kernel/cgroup_pids.c | 38 ++
>  kernel/exit.c|  1 -
>  kernel/fork.c|  7 +--
>  8 files changed, 92 insertions(+), 21 deletions(-)
> 

Reviewed-by: Andrey Ryabinin 
___
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel

Re: [Devel] [PATCH rh7 0/2] ploop: improve the performance of submit_alloc path

[Konstantin Khorenko]

Dima, please review.

--
Best regards,

Konstantin Khorenko,
Virtuozzo Linux Kernel Team

On 03/14/2016 12:10 AM, Maxim Patlasov wrote:

Instead of using pagecache_write_begin/pagecache_write_end method, it is
beneficial to preallocate space, then write data directly to block-device,
then convert uninitialized extents (ext4) explicitly.

The following series implements:
  - a preparation patch replcacing BLOCK_UNINIT with em->uninit
  - a patch implementing performance improvement

https://jira.sw.ru/browse/PSBM-22381

Signed-off-by: Maxim Patlasov 
---

Maxim Patlasov (2):
   ploop: replace BLOCK_UNINIT
   ploop: use FALLOC_FL_CONVERT_UNWRITTEN in io_direct


  drivers/block/ploop/dev.c   |7 +
  drivers/block/ploop/io_direct.c |   52 ++-
  drivers/block/ploop/io_direct_map.c |6 ++--
  drivers/block/ploop/io_direct_map.h |4 +--
  include/linux/ploop/ploop.h |6 
  5 files changed, 62 insertions(+), 13 deletions(-)

--
Signature


___
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel

Re: [Devel] [PATCH rh7 0/2] Implement UB_DCACHESIZE stats

[Konstantin Khorenko]

Kirill, please review the patch set.

--
Best regards,

Konstantin Khorenko,
Virtuozzo Linux Kernel Team

On 05/22/2015 02:08 PM, Vladimir Davydov wrote:
 See individual patches for more details.
 
 Related to https://jira.sw.ru/browse/PSBM-20089
 
 Vladimir Davydov (2):
   memcg: account dcache size
   memcg: sync UB_DCACHESIZE
 
  include/linux/memcontrol.h |3 ---
  kernel/bc/beancounter.c|3 ++-
  kernel/bc/vm_pages.c   |1 -
  mm/memcontrol.c|   54 
 +---
  mm/slab.h  |8 ---
  5 files changed, 58 insertions(+), 11 deletions(-)
 
___
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel