[Top] [All Lists]

Re: [Lse-tech] Re: A common layer for Accounting packages

To: Marcelo Tosatti <marcelo.tosatti@xxxxxxxxxxxx>
Subject: Re: [Lse-tech] Re: A common layer for Accounting packages
From: Kaigai Kohei <kaigai@xxxxxxxxxxxxx>
Date: Mon, 28 Feb 2005 10:59:06 +0900
Cc: Andrew Morton <akpm@xxxxxxxx>, davem@xxxxxxxxxx, jlan@xxxxxxx, lse-tech@xxxxxxxxxxxxxxxxxxxxx, linux-kernel@xxxxxxxxxxxxxxx, netdev@xxxxxxxxxxx
In-reply-to: <20050227140355.GA23055@xxxxxxxxxx>
References: <42168D9E.1010900@xxxxxxx> <20050218171610.757ba9c9.akpm@xxxxxxxx> <421993A2.4020308@xxxxxxxxxxxxx> <421B955A.9060000@xxxxxxx> <421C2B99.2040600@xxxxxxxxxxxxx> <421CEC38.7010008@xxxxxxx> <421EB299.4010906@xxxxxxxxxxxxx> <20050224212839.7953167c.akpm@xxxxxxxx> <20050227094949.GA22439@xxxxxxxxxx> <4221E548.4000008@xxxxxxxxxxxxx> <20050227140355.GA23055@xxxxxxxxxx>
Sender: netdev-bounce@xxxxxxxxxxx
User-agent: Mozilla Thunderbird 1.0 (Windows/20041206)

Marcelo Tosatti wrote:
> Yep, the netlink people should be able to help - they known what would be
> required for not sending messages in case there is no listener registered.
> Maybe its already possible? I have never used netlink myself.

If we notify the fork/exec/exit-events to user-space directly as you said,
I don't think some hackings on netlink is necessary.
For example, such packets is sent only when /proc/sys/.../process_grouping is 
and user-side daemon set this value, and unset when daemon will exit.
It's not necessary to take too seriously.

>>And, why can't netlink packets send always?
>>If there are fork/exec/exit hooks, and they call CSA or other
>>process-grouping modules,
>>then those modules will decide whether packets for interaction with the
>>daemon should be
>>sent or not.
> The netlink data will be sent to userspace at fork/exec/exit hooks - one wants
> to avoid that if there are no listeners, so setups which dont want to run the
> accounting daemon dont pay the cost of building and sending the information
> through netlink.
> Thats what Andrew asked for if I understand correctly.

Does it mean "netlink packets shouled be sent to userspace unconditionally." ?
I have advocated steadfastly that fork/exec/exit hooks is necessary to support
process-grouping and to account per process-grouping.
It intend to be decided whether packets should be sent or not by hooked 
in my understanding.
Is it also one of the implementations whether using netlink-socket or not ?

>>In most considerable case, CSA's kernel-loadable-module using such hooks
>>will not be loaded
>>when no accounting daemon is running. Adversely, this module must be loaded
>>when accounting
>>daemon needs CSA's netlink packets.
>>Thus, it is only necessary to refer flag valiable and to execute
>>when no-accounting daemon is running.
> That would be one hack, although it is uglier than the pure netlink
> selection.

No, I can't agree this opinion.
It means netlink-packets will be sent unconditionally when fork/exec/exit occur.
Nobady can decide which packet is sent user-space, I think.

In addition, the definition of process grouping is lightweight in many cases.
For example, CpuSet can define own process-group by one increment-operation.

I think it's not impossible to implement it in userspace, but it's not 
An implementation as a kernel loadable module is reasonable and enough tiny.

>>In my estimation, we must pay additional cost for an increment-operation,
>>an decrement-op,
>>an comparison-op and an conditional jump-op. It's enough lightweight, I
>>For example:
>>If CSA's module isn't loaded, 'privates_for_grouping' will be empty.
>>inline int on_fork_hook(task_struct *parent, task_struct *newtask){
>>  rcu_read_lock();
>>  if( !list_empty(&parent->privates_for_grouping) ){
>>    ..<Calling to any process grouping module>..;
>>  }
>>  rcu_read_unlock();
> Andrew has been talking about sending data over netlink to implement the
> accounting at userspace, so this piece of code is out of the game, no?

Indeed, I'm not opposed to implement the accounting in userspace and
using netlink-socket for kernel-daemon communication. But definition
of process-grouping based on any grouping policy should be done
in kernel space at reasonability viewpoint.

Linux Promotion Center, NEC
KaiGai Kohei <kaigai@xxxxxxxxxxxxx>

<Prev in Thread] Current Thread [Next in Thread>