Date: Mon, 10 Mar 2014 09:45:20 -0400 From: John Baldwin <jhb@freebsd.org> To: freebsd-arch@freebsd.org Cc: freebsd-arm <freebsd-arm@freebsd.org>, Ian Lepore <ian@freebsd.org> Subject: Re: option NEW_PCIB Message-ID: <201403100945.20298.jhb@freebsd.org> In-Reply-To: <58AB4C66-4267-414D-80D4-B97FF86A94A5@bsdimp.com> References: <1394200335.1149.370.camel@revolution.hippie.lan> <58AB4C66-4267-414D-80D4-B97FF86A94A5@bsdimp.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Friday, March 07, 2014 9:38:33 am Warner Losh wrote: >=20 > On Mar 7, 2014, at 6:52 AM, Ian Lepore <ian@FreeBSD.org> wrote: >=20 > > Every architecture has "option NEW_PCIB" in its conf/DEFAULTS except arm > > and mips. Is that on purpose? What are the implications of adding it? > > Or maybe more importantly, what are the implications of it not being > > there? >=20 > This is John Baldwin=92s option for his reworked PCI bridge code. He did = that as > a fallback in case he really messed up something. It introduces renumberi= ng > of busses that don=92t already have numbers assigned. It should be enable= d on > ARM, but the required resource isn=92t defined on arm, and some of the ot= her > required glue doesn=92t seem to be implemented for arm yet, which is why = things > are the way they are at the moment. I think John intends for the option t= o go > away, and everything it covers will be =91standard=92. Yes. I just added a page on the wiki about NEW_PCIB explaining the changes each platform needs for it in a bit more detail on Friday: https://wiki.freebsd.org/NEW_PCIB I have posted patches in the past to arm@ to handle step 2 in the NEW_PCIB base requirements for arm@ but haven't been able to get folks to test them. I just recently made a new pass through sys/arm in a p4 tree to refresh thi= s. I haven't even compiled these yet, but you can find the patch here: http://people.freebsd.org/~jhb/patches/arm_activate2.patch I don't know how best to think about fixing i80321_pci to work with NEW_PCI= B. It has some hack that I don't fully understand. I think it uses an alternate mapping of the same resource range to use a different base address for the mapping. Longer term I think the bus_map_resource() think I suggest at the bottom is how to handle that, but even then there would still need to be a way to know which base address a given resource wanted to use. It may be that we need to implement that differently (bus-specific rman flag?) =2D-=20 John Baldwin
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201403100945.20298.jhb>