On Sat, Dec 06, 2008 at 08:23:44PM -0700, M. Warner Losh wrote: > In message: <[EMAIL PROTECTED]> > Kostik Belousov <[EMAIL PROTECTED]> writes: > : On Fri, Dec 05, 2008 at 11:46:00PM +0100, Roman Divacky wrote: > : > On Fri, Dec 05, 2008 at 08:50:24PM +0000, Konstantin Belousov wrote: > : > > Author: kib > : > > Date: Fri Dec 5 20:50:24 2008 > : > > New Revision: 185647 > : > > URL: http://svn.freebsd.org/changeset/base/185647 > : > > > : > > Log: > : > > Several threads in a process may do vfork() simultaneously. Then, all > : > > parent threads sleep on the parent' struct proc until corresponding > : > > child releases the vmspace. Each sleep is interlocked with proc mutex > of > : > > the child, that triggers assertion in the sleepq_add(). The assertion > : > > requires that at any time, all simultaneous sleepers for the channel > use > : > > the same interlock. > : > > > : > > Silent the assertion by using conditional variable allocated in the > : > > child. Broadcast the variable event on exec() and exit(). > : > > > : > > Since struct proc * sleep wait channel is overloaded for several > : > > unrelated events, I was unable to remove wakeups from the places where > : > > cv_broadcast() is added, except exec(). > : > > : > are there any differences (performance etc.) in using condition variables > : > instead of sleep/wakeup? > : > : I do not think that there is any measurable difference. On the other > : hand, the patch makes struct proc bigger by int + pointer. This shall > : not be a problem too. > : > : Would I been able to convert _all_ uses of the struct proc * wait channel > : to cond vars operation, this may be measurable on some loads, since it > : would exclude spurious wakeups. > > Is that a measurable good difference, or a measurable bad difference?
I expect this could be measurable good difference, i.e. such patch might improve performance on some loads by eliminating false wakeups.
pgpLJhOY3kc2C.pgp
Description: PGP signature
