Re: [PATCH] sysctl: add support for poll()

[Lucas De Marchi]

* Kay Sievers <kay.sievers@xxxxxxxx> [2011-06-13 18:05:51 +0200]:

> On Thu, 2011-06-09 at 15:16 +0200, Kay Sievers wrote:
> > On Thu, Jun 9, 2011 at 00:17, Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> wrote:
> 
> > > We already have several pollable procfs files, such as
> > > fs/proc/base.c:mounts_poll() and I think drivers/md has one.  I do
> > > think that any work in this area should end up with those custom
> > > make-procfs-pollable hacks being identified and removed.
> > 
> > For these files we can probably move the event counter into the
> > seq_file structure, and get rid of the dance to kmalloc it and assign
> > it to seq_file->private. That might simplify the logic a bit.
> > 
> > [Adding Neil, to get his opinion of moving 'event' so seq_file and get
> > rid of the malloc dance]
> 
> I guess, we could do something like this, which looks quite a bit
> simpler by moving the poll event counter into the dynamically allocated
> seq_file structure itself, instead of having private structures
> allocated on top to just carry the counter (patch is just
> compile-tested).
> 
> Thanks,
> Kay
> 
> ---
>  drivers/md/md.c               |   26 ++++++++------------------
>  fs/namespace.c                |    4 ++--
>  fs/proc/base.c                |    2 +-
>  include/linux/mnt_namespace.h |    1 -
>  include/linux/seq_file.h      |    1 +
>  mm/swapfile.c                 |   29 ++++++++---------------------
>  6 files changed, 20 insertions(+), 43 deletions(-)

I've successfully tested this on top of 3.0-rc3 for /proc/mounts and
/proc/swaps. I've booted with systemd (that depends on mounts and swaps
being pollable) and created a small test using epoll.


Lucas De Marchi
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/