Re: [PATCH 0/10] Make bsd process accounting work in pid namespaces

[Pavel Emelyanov <xemul@openvz.org>]

Andrew Morton wrote:
> On Thu, 15 May 2008 14:52:46 +0400 Pavel Emelyanov <xemul@openvz.org> wrote:
> 
>> fter I fixed access to task->tgid in kernel/acct.c, Oleg 
>> pointed out some bad side effects with this accounting vs pid 
>> namespaces interaction. I.e. when some task in pid namespace 
>> sets this accounting up, this blocks all the others from doing 
>> the same. Restricting this to init namespace only could help, 
>> but didn't look as a grace solution.
>>
>> So here is the approach to make this accounting work with pid
>> namespaces properly.
> 
> We still have this silliness in acct_file_reopen():
> 
> 		spin_unlock(&acct_lock);
> 		do_acct_process(acct, old_ns, old_acct);
> 		filp_close(old_acct, NULL);
> 		spin_lock(&acct_lock);
> 
> Was this lock-dropping safe before?
> 
> Is it safe now?

Yup :)

> To work this out, we'd need to know what acct_lock protects.  What does
> acct_lock protect?

It protected the acct_glbls state before this set. This re-lock was OK 
before wrt. acct_glbls members and still such now.

After the set it still does so and (!) also protects the list of those
bsd_acct_struct-s, that have a file opened.

So, as far as the list is concerned. There are two places, that walk this
list under this lock and call acct_file_reopen - the acct_auto_close and the
acct_auto_close_mnt - both perform "goto restart" to start the iteration
from the very beginning, so this re-lock doesn't affect this list integrity.
List modifications (add/del) always happen under this lock.


Thanks,
Pavel