Date: Tue, 21 Feb 2017 20:55:42 -0700 From: Ian Lepore <ian@freebsd.org> To: freebsd-arm@freebsd.org, Tony Hain <tony@tndh.net>, Thomas Laus <lausts@laus.org> Subject: Re: Fatal kernel abort Message-ID: <1487735742.73144.138.camel@freebsd.org> In-Reply-To: <1487711291.73144.131.camel@freebsd.org> References: <155401d28ba9$ccdc28c0$66947a40$@tndh.net> <CAFHCsPWm6fOh1phFfYF%2BjJO5mDV0=mNPd1EtHgDnrp=DBwHekw@mail.gmail.com> <157701d28bda$dd004700$9700d500$@tndh.net> <1487638318.73144.116.camel@freebsd.org> <159c01d28c18$bfbb90d0$3f32b270$@tndh.net> <1487692868.73144.118.camel@freebsd.org> <7d331926-08e2-9699-aaa9-538d40ee29dc@mail.laus.org> <1487711291.73144.131.camel@freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, 2017-02-21 at 14:08 -0700, Ian Lepore wrote: > On Tue, 2017-02-21 at 15:37 -0500, Thomas Laus wrote: > > > > > > > > [... another failure report] > > > > Tom > > > That's a good data point, thanks. > > Mid-January, huh? I'm suddenly remembering that I made changes to > beaglebone sdcard handling around that time... I made the card-detect > pins work correctly so that you could insert an sd card after the > system booted and have it recognized. Maybe there was some > unexpected > side effect to that. > > Hmm, and also around the same time, the u-boot port was switched to a > newer version and we started using different dtb files, and I've just > noticed that was another instance of this readonly file system error > reported back then. I actually flagged that mail to look into it > further, then forgot to follow up. > > It suddenly occurs to me that along with the change to card-detect > gpio > pins, I also made the code honor write-protect pins (write protect > switches don't exist on microsd cards). I wonder if something in the > changed dtb files triggered a bug with my gpio write-protect sensing > changes? I'll have a look and see what I can fix. > Yep, it turned out I caused this with my changes to support gpio pins for card detect and write protect. The fdt data for beaglebone doesn't list any pins for write protect and doesn't have the "wp-disable" property (which is a bug in the fdt data, because eMMC and micro-sd cannot be write protected). The old driver logic treated that as "not protected" and a side effect of the way the new code works effectively reversed that into saying everything is write protected. I just commited a fix for it as r314071, sorry for taking so long to realize all of this was fallout from my change. -- Ian
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1487735742.73144.138.camel>