On Thu, Sep 27, 2007 at 01:25:48PM -0600, Eric W. Biederman wrote:
>
> I still need to look at the code in detail but I have some concerns
> I want to inject into this conversation of future sysfs architecture.
>
> - If we want to carefully limit sysfs from going to wild code review
> is clearl
Pavel Emelianov [EMAIL PROTECTED] wrote:
| Some time ago Sukadev noticed that the vmlinux size has
Cedric pointed it out to me first :-)
| grown 5Kb due to merged pid namespaces. One of the big
| problems with it was fat inline functions. The other thing
| was noticed by Matt - the checks for tas
Simplify the memory controller and resource counter I/O routines
This patch strips out some I/O boilerplate from resource counters and
the memory controller. It also adds locking to the resource counter
reads and writes, and forbids writes to the root memory cgroup's limit
file.
cgroup_write_uin
On 10/4/07, Balbir Singh <[EMAIL PROTECTED]> wrote:
>
> Yes, either that way or add a strategy function, that would take
> the string input from the user and convert it to unsigned long long
> value. I am ok with either approach.
>
OK, new version of the patch sent in a separate mail.
Paul
__
| > | > One solution I was thinking of was to possibly queue pending blocked
| > | > signals to a container init seperately and then requeue them on the
| > | > normal queue when signals are unblocked. Its definitely not an easier
| > | > solution, but might be less intrusive than the "signal from
Found-by: Hugh Dickins <[EMAIL PROTECTED]>
mem_cgroup_charge() in unuse_pte() is called under a lock, the pte_lock. That's
clearly incorrect, since we pass GFP_KERNEL to mem_cgroup_charge() for
allocation of page_cgroup.
This patch release the lock and reacquires the lock after the call to
mem_
Paul Menage wrote:
> On 10/4/07, Balbir Singh <[EMAIL PROTECTED]> wrote:
>> Paul Menage wrote:
>>> On 10/4/07, Balbir Singh <[EMAIL PROTECTED]> wrote:
Forbidding writing to the root resource counter is a policy decision
I am unable to make up my mind about. It sounds right, but unless
>>>
On 10/4/07, Balbir Singh <[EMAIL PROTECTED]> wrote:
> Paul Menage wrote:
> > On 10/4/07, Balbir Singh <[EMAIL PROTECTED]> wrote:
> >> Forbidding writing to the root resource counter is a policy decision
> >> I am unable to make up my mind about. It sounds right, but unless
> >> we have a notion of
YAMAMOTO Takashi wrote:
>> hi,
>>
>> i implemented some statistics for your memory controller.
>>
>> it's tested with 2.6.23-rc2-mm2 + memory controller v7.
>> i think it can be applied to 2.6.23-rc4-mm1 as well.
>>
>> YAMOMOTO Takshi
>>
>> todo: something like nr_active/inactive in /proc/vmstat.
>
Paul Menage wrote:
> On 10/4/07, Balbir Singh <[EMAIL PROTECTED]> wrote:
>> Forbidding writing to the root resource counter is a policy decision
>> I am unable to make up my mind about. It sounds right, but unless
>> we have a notion of unlimited resources, I am a bit concerned about
>> taking away
On 10/4/07, Balbir Singh <[EMAIL PROTECTED]> wrote:
>
> Forbidding writing to the root resource counter is a policy decision
> I am unable to make up my mind about. It sounds right, but unless
> we have a notion of unlimited resources, I am a bit concerned about
> taking away this flexibility.
One
Paul Menage wrote:
> Hi Balbir,
>
> Any thoughts on this patch?
>
Hi, Paul,
I remember seeing this patch, sorry for not responding earlier
> Cheers,
>
> Paul
>
> On 9/25/07, Paul Menage <[EMAIL PROTECTED]> wrote:
>> Simplify the memory controller and resource counter I/O routines
>>
>> This
Hi Balbir,
Any thoughts on this patch?
Cheers,
Paul
On 9/25/07, Paul Menage <[EMAIL PROTECTED]> wrote:
> Simplify the memory controller and resource counter I/O routines
>
> This patch strips out some I/O boilerplate from resource counters and
> the memory controller. It also adds locking to th
On Thu, Oct 04, 2007 at 12:54:17PM +0400, Pavel Emelyanov wrote:
> Matt Mackall wrote:
> > On Wed, Oct 03, 2007 at 06:20:43PM +0400, Pavel Emelyanov wrote:
> >> Just make the __pid_nr() etc functions that expect the argument
> >> to always be not NULL.
> >>
> >> Signed-off-by: Pavel Emelyanov <[EMA
Paul M wrote:
> I didn't notice any performance hit on a pure allocate/free memory
> benchmark relative to non-cgroup cpusets.
Good.
--
I won't rest till it's the best ...
Programmer, Linux Scalability
Paul Jackson <[EMAIL PROTECTED]> 1.925.6
Paul M wrote:
> It's two constant-indexed dereferences *in total*, compared to a
> single constant-indexed dereference in the pre-cgroup case.
Ok - the C expression is longer and I didn't realize how
little difference it made in the end (the executing code.)
Good - thanks.
--
From: Sukadev Bhattiprolu <[EMAIL PROTECTED]>
Subject: [PATCH] Rename is_cgroup_init()
is_container_init() was accidentally renamed to is_cgroup_init() when
renaming "container" to "control group". This patch restores the
original name.
Signed-off-by: Sukadev Bhattiprolu <[EMAIL PROTECTED]>
Acked
Eric W. Biederman wrote:
> Daniel Lezcano <[EMAIL PROTECTED]> writes:
>> Yes, it will work.
>>
>> Do we want to be inside a network namespace and to use a socket belonging to
>> another network namespace ? If yes, then my remark is irrelevant.
>
> Yes we do.
>
Shall we close all fd sockets w
On 10/4/07, Pavel Emelyanov <[EMAIL PROTECTED]> wrote:
> This one is responsible for initializing the RES_CFT_MAX files
> properly and register them inside the container.
>
> The caller must provide the cgroup, the cgroup_subsys, the
> RES_CFT_MAX * sizeof(cftype) chunk of zeroed memory, the
> unit
On 10/4/07, Pavel Emelyanov <[EMAIL PROTECTED]> wrote:
> This is just to reduce the code amount in the future.
>
> Signed-off-by: Pavel Emelyanov <[EMAIL PROTECTED]>
Acked-by: Paul Menage <[EMAIL PROTECTED]>
>
> ---
>
> diff --git a/include/linux/cgroup.h b/include/linux/cgroup.h
> index 8747932.
On 10/4/07, Paul Jackson <[EMAIL PROTECTED]> wrote:
> Paul M,
>
> This snippet from the memory allocation hot path worries me a bit.
>
> Once per memory page allocation, we go through here, needing to peak inside
> the current tasks cpuset to see if it has changed (it's 'mems_generation'
> value do
On 10/4/07, Kir Kolyshkin <[EMAIL PROTECTED]> wrote:
>
> 3. Put units into a separate new files named "units" (or, well,
> "measurement_units" (or even "measured_in") if you are fan of long
> descriptive names). So, "cat units" will show us "bytes" or "items" or
> "pages"...
I'd vote for this opti
On Thu, Oct 04, 2007 at 05:54:11PM +0400, Pavel Emelyanov wrote:
> With the net namespaces many code leaved the __init section,
> thus making the kernel occupy more memory than it did before.
> Since we have a config option that prohibits the namespace
> creation, the functions that initialize/fina
With the net namespaces many code leaved the __init section,
thus making the kernel occupy more memory than it did before.
Since we have a config option that prohibits the namespace
creation, the functions that initialize/finalize some netns
stuff are simply not needed and can be freed after the bo
Quoting Cedric Le Goater ([EMAIL PROTECTED]):
> [EMAIL PROTECTED] wrote:
> > Eric W. Biederman [EMAIL PROTECTED] wrote:
> > | [EMAIL PROTECTED] writes:
> > |
> > | > Cedric Le Goater [EMAIL PROTECTED] wrote:
> > | > | > I think you and Eric (and I) are disagreeing about those
> > limitations.
> >
[EMAIL PROTECTED] wrote:
> Eric W. Biederman [EMAIL PROTECTED] wrote:
> | [EMAIL PROTECTED] writes:
> |
> | > Cedric Le Goater [EMAIL PROTECTED] wrote:
> | > | > I think you and Eric (and I) are disagreeing about those limitations.
> | > | > You take it for granted that a sibling pidns is off limi
Eric W. Biederman [EMAIL PROTECTED] wrote:
| [EMAIL PROTECTED] writes:
|
| > Cedric Le Goater [EMAIL PROTECTED] wrote:
| > | > I think you and Eric (and I) are disagreeing about those limitations.
| > | > You take it for granted that a sibling pidns is off limits for signals.
| > | > But the signa
[EMAIL PROTECTED] writes:
> Cedric Le Goater [EMAIL PROTECTED] wrote:
> | > I think you and Eric (and I) are disagreeing about those limitations.
> | > You take it for granted that a sibling pidns is off limits for signals.
> | > But the signal wasn't sent using a pid, but using a file (in SIGIO
>
Cedric Le Goater <[EMAIL PROTECTED]> writes:
> [
> I have big fingers this morning and I managed to send this email
> while typing it ... see below for the end. I should be awake now :)
> ]
>
>> The really challenging case to handle here is what happens if we are
>> signaling to someone in
> I think you and Eric (and I) are disagreeing about those limitations.
> You take it for granted that a sibling pidns is off limits for signals.
> But the signal wasn't sent using a pid, but using a file (in SIGIO
> case). So since the fs was shared, the signal should be sent. An
> event happene
Cedric Le Goater [EMAIL PROTECTED] wrote:
| > I think you and Eric (and I) are disagreeing about those limitations.
| > You take it for granted that a sibling pidns is off limits for signals.
| > But the signal wasn't sent using a pid, but using a file (in SIGIO
| > case). So since the fs was shar
Quoting Cedric Le Goater ([EMAIL PROTECTED]):
> [EMAIL PROTECTED] wrote:
> > Cedric Le Goater [EMAIL PROTECTED] wrote:
> > |
> > | >> however, we have an issue with the signal notification in __do_notify()
> > | >> we could kill a process in a different pid namespace.
> > | >
> > | > So I took a
Pavel Emelyanov wrote:
<...skipped...>
+static char * units_names[RES_UNITS_MAX][RES_CFT_MAX] = {
+ [RES_UNITS_BYTES] = {
+ "usage_in_bytes",
+ "limit_in_bytes",
+ "failcnt",
+ },
+ [RES_UNITS_ITEMS] = {
+ "usage",
+
Paul M,
This snippet from the memory allocation hot path worries me a bit.
Once per memory page allocation, we go through here, needing to peak inside
the current tasks cpuset to see if it has changed (it's 'mems_generation'
value doesn't match the last seen value we have stashed in the task stru
Note, that the controller code dealing with the cftype files
for resource counters becomes much shorter and won't have to
be changed in the future.
Signed-off-by: Pavel Emelyanov <[EMAIL PROTECTED]>
---
diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index 1b8bf24..2e62d24 100644
--- a/mm/memcon
This is just to reduce the code amount in the future.
Signed-off-by: Pavel Emelyanov <[EMAIL PROTECTED]>
---
diff --git a/include/linux/cgroup.h b/include/linux/cgroup.h
index 8747932..0635004 100644
--- a/include/linux/cgroup.h
+++ b/include/linux/cgroup.h
@@ -178,6 +178,15 @@ struct css_set {
This one is responsible for initializing the RES_CFT_MAX files
properly and register them inside the container.
The caller must provide the cgroup, the cgroup_subsys, the
RES_CFT_MAX * sizeof(cftype) chunk of zeroed memory, the
units of measure and the read and write callbacks.
Right now I made n
Changes from previous version
* made the names configurable
* fixed race between res_counter_populate and reading
any of these files (memset to zero could spoof the
pointers)
Right now we have only one controller in -mm tree - the memory
one - and it initializes all its files manually. After I
Randy Dunlap wrote:
> On Wed, 03 Oct 2007 18:19:01 +0400 Pavel Emelyanov wrote:
>
>> This is a pid which is attached to tasks when they detach
>> their pids. This is done in detach_pid() and transfer_pid().
>> The pid_alive() check is changed to reflect this fact.
>>
>> Signed-off-by: Pavel Emelya
Matt Mackall wrote:
> On Wed, Oct 03, 2007 at 06:20:43PM +0400, Pavel Emelyanov wrote:
>> Just make the __pid_nr() etc functions that expect the argument
>> to always be not NULL.
>>
>> Signed-off-by: Pavel Emelyanov <[EMAIL PROTECTED]>
>
>> static inline pid_t pid_nr(struct pid *pid)
>> {
>>
40 matches
Mail list logo