Hi all,

the attached patch (relative to 2.6.8-rc2-mm2) restricts the number of 
concurrently running hotplug  / kmod processes. It's an update to the 
patch posted earlier, adapted to suggestions by Andrew Morton (THX for 
that, Andrew).

The current design of call_usermodehelper has the disadvantage that 
always a kernel thread is started without checking whether enough 
resources are available. This makes it possible to run any machine out 
of memory by just generating enough hotplug processes.

The patch introduces a semaphore khelper_sem, which is initialised with 
the maximum number of concurrently running hotplug processes. When the 
semaphore count hits zero, the process is either blocked (if its an 
synchronous process) or delayed by returning control to the calling 
function prior to call down(). In either case we have to wait for the 
helper thread to finish, as we have to up() the semaphore and free up 
the process information.

Note that we have to copy the process information in case the hotplug 
event gets delayed, as originally the process information is allocated 
locally to the calling function, which is lost after control is returned.

Please apply.

Cheers,

Hannes
-- 
Dr. Hannes Reinecke			hare@suse.de
SuSE Linux AG				S390 & zSeries
Maxfeldstraße 5				+49 911 74053 688
90409 Nürnberg				http://www.suse.de