Date: Wed, 17 Feb 2016 10:19:43 -0800 From: John Baldwin <jhb@freebsd.org> To: freebsd-arch@freebsd.org Cc: Poul-Henning Kamp <phk@phk.freebsd.dk>, Warner Losh <imp@bsdimp.com> Subject: Re: Starting APs earlier during boot Message-ID: <7708748.E0vxJ7C8cf@ralph.baldwin.cx> In-Reply-To: <41169.1455702411@critter.freebsd.dk> References: <1730061.8Ii36ORVKt@ralph.baldwin.cx> <CANCZdfqRiEb=fEV1fiE8E9Lr=KYPxDcs5jS2iDW-OowwgoFL3Q@mail.gmail.com> <41169.1455702411@critter.freebsd.dk>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wednesday, February 17, 2016 09:46:51 AM Poul-Henning Kamp wrote: > -------- > In message <CANCZdfqRiEb=fEV1fiE8E9Lr=KYPxDcs5jS2iDW-OowwgoFL3Q@mail.gmail.com>, Warner Losh writes: > > >> what is the goal? cleaner code? faster boot? > > > >Two goals were in his original email. > > And I hope that in the longer term we also aim to configure I/O > in parallel ? I'm a bit leery of doing this fully parallel. In particular, users currently depend on the behavior of deterministic names in new-bus (so em0 is always em0 and not sometimes em1). OTOH, I think that we could eventually allow drivers to start doing some of the background scans sooner and only harvest the results at the interrupt config hooks instead of starting the scans and timers at the interrupt config hook (and this is a step towards that). From what I understand, most of our boot time start up delay isn't the new-bus device probe but userland startup. Nevertheless, I think the changes I've proposed here are a prerequisite for even thinking about possibly making device probe more parallel. -- John Baldwin
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?7708748.E0vxJ7C8cf>