Date: Wed, 29 Mar 2006 11:57:52 -0500 From: John Baldwin <jhb@freebsd.org> To: Scott Long <scottl@samsco.org> Cc: Bruce M Simpson <bms@spc.org>, "Conrad J. Sabatier" <conrads@cox.net>, freebsd-current@freebsd.org Subject: Re: device atpic to be deprecated? Message-ID: <200603291157.54467.jhb@freebsd.org> In-Reply-To: <442AB520.5050505@samsco.org> References: <20060329020527.f8f087a4.conrads@cox.net> <200603290841.50759.jhb@freebsd.org> <442AB520.5050505@samsco.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wednesday 29 March 2006 11:26, Scott Long wrote: > John Baldwin wrote: > > On Wednesday 29 March 2006 03:51 am, Bruce M Simpson wrote: > > > >>On Wed, Mar 29, 2006 at 02:05:27AM -0600, Conrad J. Sabatier wrote: > >> > >>>Is the plan still in effect to abolish this device? > >> > >>To my mind it wouldn't make much sense, given the sheer amount of hardware > >>out there which doesn't have an IOAPIC, then again I'm probably out of > >>touch with the state of interrupt handling in -CURRENT. > > > > > > All amd64 machines (which is where atpic would be removed) have an APIC. > > > > That's kind of like saying that ISA will be removed because there is PCI > =-) Having an APIC doesn't necessarily guarantee that it works. There > have been enough reports of problems on the mailing lists over time that > I think it's a bit premature to declare the ATPIC dead. Also, is the > ATPIC code in amd64 causing problems, holding back progress, or creating > a maintenance burden? I think that once the lapic timer stuff was added almost all of the APIC issues I was aware of went away on amd64 that were fixed by using device atpic instead. Most of the earlier problems were due to chipsets not setting up pin 0 as extint, etc. but all that is no longer relevant when we switched to using the lapic timer and stopped using irq0 and irq8 with APIC. This is the first I've heard since the lapic timer stuff that APIC didn't work on an amd64 box, and device atpic has been off by default in HEAD for quite a while now. If we were able to require APIC on amd64, then we might be able to try out some optimizations and other things I haven't bothered with since they wouldn't be feasible on i386. -- John Baldwin <jhb@FreeBSD.org> <>< http://www.FreeBSD.org/~jhb/ "Power Users Use the Power to Serve" = http://www.FreeBSD.org
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200603291157.54467.jhb>