On Monday 22 May 2006 10:01, Robert Watson wrote: > On Mon, 22 May 2006, Sergey Matveychuk wrote: > >> We have also discussed requiring that all network device drivers > >> in the system for 7.0 be able to operate MPSAFE. This runs into > >> some issues due to non-MPSAFEty of some key device driver > >> subsystems, including Firewire, USB, and the TTY subsystem. > >> Poul-Henning is in the process of working on MPSAFE tty support, > >> and I hope that people will step up to the plate to work on the > >> remaining subsystems in the near future. That way we will be > >> able to remove the compatibility code, and continue cleanup and > >> optimization work. > > > > Looks like there are no developers right now who could rewrite > > USB stack to be MPSAFE. So all USB-Ethernet are broken. > > Yes, hence this being something we are discussing, as opposed to > implementing. In order for us to get to a world entirely without > Giant-compatibility for network device drivers, we need to address > the MPSAFEty of those major subsystems. The goal of this message > was to make sure we don't start growing new major subsystems and > network device drivers -- i.e., that we're into the realm of > monotonically decreasing numbers of non-MPSAFE network device > drivers. Specifically, I don't want to see any more PCI network > device drivers that require Giant! > > Per discussion at the developer summit, a significant rewrite of > the USB code is needed to make it MPSAFE. Volunteers welcome. I'm pretty sure HPS USB stack is MPSAFE, or at least that's the goal.
-- Anish Mistry
pgpRn4VwXC3Kf.pgp
Description: PGP signature
