bug-hurd
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: tmpfs status


From: Maksym Planeta
Subject: Re: tmpfs status
Date: Thu, 29 Mar 2012 21:29:58 +0300
User-agent: Gnus/5.13 (Gnus v5.13) Emacs/23.3 (gnu/linux)

Thomas Schwinge <address@hidden> writes:

> Hi!
>
> On Wed, 28 Mar 2012 23:41:57 +0300, Maksym Planeta <address@hidden> wrote:
>> Thomas Schwinge <address@hidden> writes:
>> > On Wed, 28 Mar 2012 22:40:22 +0300, Maksym Planeta <address@hidden> wrote:
>> >> Samuel Thibault <address@hidden> writes:
>> >> 
>> >> > Thomas Schwinge, le Mon 26 Mar 2012 22:24:55 +0200, a écrit :
>> >> >> > 1. 
>> >> >> > http://git.savannah.gnu.org/cgit/hurd/hurd.git/log/?h=mplaneta/tmpfs/master
>> >> >> >  
>> >> >> 
>> >> >> Someone to review the patches...  :-/
>> >> >
>> >> > Just a few comments.
>> >> >
>> >> >> Fix auto-terminating of tmpfs due to idle.
>> >> >
>> >> > Err, really?  I'd rather not see my tmpfs fly away just because I've
>> >> > left it dormant for some time.
>> >> 
>> >> I meant that now it doesn't auto-terminate.
>> >
>> > Yes.  But that should be possible to achieve very much simpler: just put
>> > ``thread_timeout = server_timeout = 0;'' before diskfs_spawn_first_thread
>> > is called.  (Or am I missing something?)
>> >
>> 
>> Variables thread_timeout and server_timeout declared in
>> ../libdiskfs/init-first.c and are static. 
>
> Indeed.  :-) Options: remove static, add two declarations to diskfs.h (I
> don't really like that); add a new diskfs_spawn_first_thread_no_timeout
> function (I don't really like that either); break the ABI (usually not
> preferable either...), add the two timeout values to the existing
> diskfs_spawn_first_thread, and add #defines for the default values in
> diskfs.h.
>

Than, why just not to remove "static" and do like you proposed before:

> just put ``thread_timeout = server_timeout = 0;'' before
> diskfs_spawn_first_thread is called. 

?

> The latter one is not as problematic as it sounds, as
> diskfs_spawn_first_thread is really only invoked from tmpfs: all other
> diskfs-based file systems go by using
> libdiskfs/init-main.c:diskfs_init_main, which handles libstore setup and
> then invokes diskfs_spawn_first_thread.  So apart from tmpfs this would
> just be an internal ABI change (and libdiskfs.so and tmpfs would even be
> updated at the same time).
>
> I don't think any other libdiskfs-based translators would benefit from
> being able to configure their timeout values -- all but tmpfs serialize
> their state to the backing store, from where they can restore it after a
> timeout has terminated the translator.
>

Or, if only tmpfs would benefit from ABI change, than why not to left
all changes within tmpfs code, like I did?

Regards,
Maksym Planeta.



reply via email to

[Prev in Thread] Current Thread [Next in Thread]