udevadm settle timeout semantics

udevadm settle timeout semantics

[Seewer Philippe]

The manpage for udevadm (version 141) says about the timeout for settle 
and --timeout:

[quote]
udevadm settle [options]
        Watches the udev event queue, and exits if all
        current events are handled.

--timeout=seconds
            Maximum number of seconds to wait for the event
            queue to become empty. The default value is 180
            seconds. A value of 0 will check if the queue
            is empty and always return immediately.
[/quote]


Am I reading this correctly if I assume that udevadm has to wait for an 
event to be "handled" regardless of the timeout value?

Example: dhclient has a default of 60 seconds to try to get an ip 
address until it fails. udevadm settle waits for the whole 60 seconds to 
pass regardless of the timeout parameter.

Is this behaviour as intended?

Thanks for the help,
Philippe
--
To unsubscribe from this list: send the line "unsubscribe initramfs" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: udevadm settle timeout semantics

[Kay Sievers]

On Thu, 2009-06-18 at 13:14 +0200, Seewer Philippe wrote:
> The manpage for udevadm (version 141) says about the timeout for settle 
> and --timeout:
> 
> [quote]
> udevadm settle [options]
>         Watches the udev event queue, and exits if all
>         current events are handled.
> 
> --timeout=seconds
>             Maximum number of seconds to wait for the event
>             queue to become empty. The default value is 180
>             seconds. A value of 0 will check if the queue
>             is empty and always return immediately.
> [/quote]
> 
> 
> Am I reading this correctly if I assume that udevadm has to wait for an 
> event to be "handled" regardless of the timeout value?
> 
> Example: dhclient has a default of 60 seconds to try to get an ip 
> address until it fails. udevadm settle waits for the whole 60 seconds to 
> pass regardless of the timeout parameter.
> 
> Is this behaviour as intended?

The timeout is the maximum to wait:

  $ echo 'RUN+="/bin/sleep 300"' > /etc/udev/rules.d/00.rules
  $ echo add > /sys/class/mem/zero/uevent

  $ time udevadm settle --timeout=3
  udevadm settle - timeout of 3 seconds reached, the event queue contains:
  /sys/devices/virtual/mem/zero (1796)

Not sure, udev 141 maybe had a bug here.

Thanks,
Kay

--
To unsubscribe from this list: send the line "unsubscribe initramfs" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: udevadm settle timeout semantics

[Seewer Philippe]

Kay Sievers wrote:
> On Thu, 2009-06-18 at 13:14 +0200, Seewer Philippe wrote:
>> The manpage for udevadm (version 141) says about the timeout for settle 
>> and --timeout:
>>
>> [quote]
>> udevadm settle [options]
>>         Watches the udev event queue, and exits if all
>>         current events are handled.
>>
>> --timeout=seconds
>>             Maximum number of seconds to wait for the event
>>             queue to become empty. The default value is 180
>>             seconds. A value of 0 will check if the queue
>>             is empty and always return immediately.
>> [/quote]
>>
>>
>> Am I reading this correctly if I assume that udevadm has to wait for an 
>> event to be "handled" regardless of the timeout value?
>>
>> Example: dhclient has a default of 60 seconds to try to get an ip 
>> address until it fails. udevadm settle waits for the whole 60 seconds to 
>> pass regardless of the timeout parameter.
>>
>> Is this behaviour as intended?
> 
> The timeout is the maximum to wait:
> 
>   $ echo 'RUN+="/bin/sleep 300"' > /etc/udev/rules.d/00.rules
>   $ echo add > /sys/class/mem/zero/uevent
> 
>   $ time udevadm settle --timeout=3
>   udevadm settle - timeout of 3 seconds reached, the event queue contains:
>   /sys/devices/virtual/mem/zero (1796)
> 
> Not sure, udev 141 maybe had a bug here.

That is correct.

$ echo 'KERNEL=="zero" RUN+="/bin/sleep 300"' > /etc/udev/rules.d/00.rules
$ echo add > /sys/class/mem/zero/uevent

$ time udevadm settle --timeout=5
real	2m52.687s
user	0m0.004s
sys	0m0.064s

Harald found the bugfix for udevadm which fixed this:
http://article.gmane.org/gmane.linux.hotplug.devel/13952


This creates a problem for network configuration: If dhcp takes
longer than 30 seconds plus the time we allow for mount fall-through,
an emergency-shell will appear while maybe mount is happending in the
background because dhcp succeeded.

Any thoughts?
--
To unsubscribe from this list: send the line "unsubscribe initramfs" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: udevadm settle timeout semantics

[David Dillow]

On Thu, 2009-06-18 at 14:55 +0200, Seewer Philippe wrote:
> This creates a problem for network configuration: If dhcp takes
> longer than 30 seconds plus the time we allow for mount fall-through,
> an emergency-shell will appear while maybe mount is happending in the
> background because dhcp succeeded.
> 
> Any thoughts?

Network booting likely needs to retry DHCP and NFS/netroot mounts if
they timeout during boot. Booting a large cluster can cause these
timeouts to occur if they overload the server and it's a real PITA to
find the machines that didn't boot properly and reboot them.
--
To unsubscribe from this list: send the line "unsubscribe initramfs" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html