lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <46C54FB8.7050504@bull.net>
Date:	Fri, 17 Aug 2007 09:35:20 +0200
From:	Laurent Vivier <Laurent.Vivier@...l.net>
To:	Rusty Russell <rusty@...tcorp.com.au>
Cc:	kvm-devel <kvm-devel@...ts.sourceforge.net>,
	Ingo Molnar <mingo@...e.hu>,
	virtualization <virtualization@...ts.linux-foundation.org>,
	linux-kernel <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH/RFC 3/4]Introduce "account modifiers" mechanism

Rusty Russell wrote:
> On Thu, 2007-08-16 at 17:58 +0200, Laurent Vivier wrote:
>> [PATCH 3/3] introduce "account modifiers" mechanism in the kernel allowing a
>> module to modify the collected accounting for a given task. This implementation
>> is based on the "preempt_notifier". "account_system_time()" and
>> "account_user_time()" can call functions registered by a module to modify the
>> cputime value.
>>
>> Signed-off-by: Laurent Vivier <Laurent.Vivier@...l.net>
> 
> 
> Hi Laurent,

Hi Rusty,
how are your puppies ?
And thank you for your comment.

> 	This seems a little like overkill.  Why not just add an
> "account_guest_time" which subtracts the given amount of time from
> system time (if available) and adds it to guest time?  Then kvm (and
> lguest) should just need to call this at the right times.

We can. I did something like this before.
By doing like that, I think there is a major issue: system time can be
decreasing (as we substract a value from it), and thus we can have negative
value in a tool like top. It's why I play with the cputime to add to system time
and not directly with the system time.

BUT I'm very open, my only goal is be able to compute guest time, "how" is not
very important...

what we can do:

- keep PATCHES 1 and 2, because we need to store guest time for cpu and tasks.
It is very generic.

- remove PATCH 3, and add in task_struct a "ktime vtime" where we accumulate
guest time (by calling something like guest_enter() and guest_exit() from the
virtualization engine), and when in account_system_time() we have cputime >
vtime we substrate vtime from cputime and add vtime to user time and guest time.
But doing like this we freeze in kernel/sched.c the link between system time,
user time and guest time (i.e. system time = system time - vtime, user time =
user time + vtime and guest time = guest time + vtime).

- modify PATCH 4 to use new PATCH 3.

Do you agree ? Anybody doesn't agree ?

Laurent
-- 
------------- Laurent.Vivier@...l.net  --------------
          "Software is hard" - Donald Knuth


Download attachment "signature.asc" of type "application/pgp-signature" (190 bytes)

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ