select() timeouts on large installations

[Cornelia Huck]

2005/8/23, Artem Kachitchkine <Artem.Kachitchkin at sun.com>:

> It could also be due to coldplug event serialization in hald/linux2, in
> which case adding some concurrency would help, e.g. all volumes on a
> disk probed in parallel.

I guess not on that particular machine (of all those devices, only two
disks and the network device were online; offline devices only show up
in sysfs but don't have e.g. an associated block device).

I agree some kind of asynchronous probing could speed up detection
in the general case - maybe even several devices in parallel?

Cornelia (still trying to figure out what takes so much time)