Re: udev queue error

[Kay Sievers <kay.sievers@xxxxxxxx>]

On Wed, Jun 1, 2011 at 06:42, Tejun Heo <htejun@xxxxxxxxx> wrote:
> I _think_ what we really need is a timer based throttling. Âe.g. one
> token is generated every, say, three seconds and can accumulate upto 5
> and each media changed event checking consumes a token and should wait
> if there's no token left. ÂDunno how difficult it owuld be to
> implement this tho.

The problem is that currently every udev event run causes three new
kernel events. The whole thing is constantly triggered by the (still
present) userspace polling once a second. Udev will need to forcefully
drop events from the event queue every second to break this loop --
means constant throttling without any real idle time. Sure, such
ratelimit we should probably have, but it does not really solve our
problem properly, I guess. constantly dropping events from udev that
the kernel send us shouldn't be the expected behavior.

Maybe we can find a clean way to disable the event generation during
the time udev runs the event handler? We do a similar thing with
inotify watches too, so that udev does not generate events it has
caused itself during the event run.

Kay
--
To unsubscribe from this list: send the line "unsubscribe linux-hotplug" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html