Configuration should be cached IMO when using -s: heuristics shows that the amount of rules is less than amount of device nodes. And if we want to do things uniformly (which is a powerful way to shrink the code) the configuration should be cached even if mdev is used as hotplug helper.
If such a behavior differs too much from mdev concepts we should start a fork of mdev which would accumulate the features vanilla mdev refuses to have. -- Vladimir 2008/3/29, Denys Vlasenko <[EMAIL PROTECTED]>: > On Saturday 29 March 2008 05:32, Mike Frysinger wrote: > > On Thursday 27 March 2008, [EMAIL PROTECTED] wrote: > > > Author: vda > > > Date: 2008-03-27 15:45:44 -0700 (Thu, 27 Mar 2008) > > > New Revision: 21522 > > > > > > Log: > > > mdev: plug a few memory and fd leaks; simplify code a bit > > > > as long as the resulting code is smaller, OK, otherwise this is a bug imo. > > > mdev is not meant to be executed via the shell, only by the kernel via the > > hotplug mechanism. that means any memory it may leak is irrelevant as > mdev > > is a very short lived process. > > mdev -s executes make_device() function repeatedly for each device. > And make_device() contained this: > > if (ENABLE_FEATURE_CLEAN_UP) > fclose(fp); > > Although in many cases mdev -s will successfully iterate over all > devices before it runs out of fds, there's no way to be sure about it. > -- > vda > _______________________________________________ > busybox-cvs mailing list > [EMAIL PROTECTED] > http://busybox.net/cgi-bin/mailman/listinfo/busybox-cvs > _______________________________________________ busybox mailing list [email protected] http://busybox.net/cgi-bin/mailman/listinfo/busybox
