> Date: Thu, 13 Feb 2014 17:12:09 -0500 > From: Ted Unangst <[email protected]> > Cc: [email protected] > X-XS4ALL-DNSBL-Checked: mxdrop102.xs4all.nl checked 208.82.130.146 against > DNS blacklists > X-CNFS-Analysis: v=2.1 cv=Wak1NSRX c=1 sm=0 tr=0 > a=iS0VH5W3/nZARN5H4QS42w==:117 a=iS0VH5W3/nZARN5H4QS42w==:17 > a=OxEAv6DlAAAA:8 a=_lmKfE_oAAAA:8 a=dfy0HCNRudsA:10 a=YPbdu89bTecA:10 > a=IkcTkHD0fZMA:10 a=djBvByLjHukA:10 a=8_S25J89exnB1W3R318A:9 > a=QEXdDO2ut3YA:10 > X-Virus-Scanned: by XS4ALL Virus Scanner > X-XS4ALL-Spam-Score: 0.1 () RDNS_NONE, SPF_HELO_PASS, SPF_PASS > X-XS4ALL-Spam: NO > Envelope-To: [email protected] > > On Thu, Feb 13, 2014 at 22:52, Mark Kettenis wrote: > > Turns out we are pretty close. The only problem is that X still wants > > to probe the pci bus. We allow access to /dev/pci0 if > > machdep.allowaperture is 0, but libpciaccess uses O_RDWR to open it. > > The diff below chages this such that it tries O_RDONLY if O_RDWR fails. > > Is there a reason not to use O_RDONLY in the first place? I'm > wondering if this will make it harder to diagnose a later failure > where some machines have writable fds and some don't.
"legacy" X drivers need to be able to poke pci config space for which you need write permission.
