RaspberryPi 3B segfaults on 15.1-RELEASE

V

VulcanRidr

Running into an issue with segfaults ever since upgrading to 15.1 on a RPi3B (1GB RAM). Did not notice it before, but after upgrading it to 15.1, I did start seeing them afterwards. Been seeing messages like the following:

Code: 
koraga kernel: pid 9807 (pkg), jid 0, uid 0: exited on signal 11 (core dumped)
koraga kernel: pid 13163 (cron), jid 0, uid 0: exited on signal 11 (no core dump - sugid process denied by kern.sugid_coredump)

I have gotten them in processes including bash, cron, dhcpd, named, netstat, ntpd, pkg, ps, sudo, tmux, top, vi. All were standard pkg-installed packages. Now I am not a programmer, so I am not sure where to even begin to start looking. I know that signal 11 is generally something trying to access out of range memory, or it can also be a hardware problem. As I said, I noticed it happening more often since upgrading to 15.1.

Any suggestions on how to proceed?

Thanks,
--vr
 
bakul said:
See https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=296240
This is due to a chip bug. There is a new libc.so.7 as part of patch 2 that should fix this. You can probably run freebsd-update on a different arm machine and use the -b option to target the sdcard for your pi3 (mounted on this other machine) but I haven't tried this before....
Click to expand...
Thanks bakul! For the record, my system is on 15.1-p2. Here is the freebsd-version info and the sha256 sum of libc.so.7:

Code: 
[root@koraga ~]# freebsd-version -u
15.1-RELEASE-p2
[root@koraga ~]# sha256 /lib/libc.so.7
SHA256 (/lib/libc.so.7) = 8f14ca787dd853988f829aad291c7f4834a073d8398ceef0217433a07
92bcffe

Is this the right version? because i am getting several segfaults/signal 11s. I was running top in an ssh session to watch the freebsd-update (there were no updates to the system, btw), and top segfaulted on me.

And unfortunately, I do not have another pi...
 
Hm... I too see some segfaults again (I upgraded that pi3b to -p2 last night). sha256 of libc.so.7 matches with yours. Running my test program (see the attachment in comment 2 of the above bug report), I see it dying in the same place. Suggest adding your segfaults report to the above bug report. I will see if I can cross build a working libc for this patch.....
 
bakul said:
Hm... I too see some segfaults again (I upgraded that pi3b to -p2 last night). sha256 of libc.so.7 matches with yours. Running my test program (see the attachment in comment 2 of the above bug report), I see it dying in the same place. Suggest adding your segfaults report to the above bug report. I will see if I can cross build a working libc for this patch.....
Click to expand...
Interesting...I'm wondering if, this being a pi 3b, it is worth continuing to fight with it. Also, (again not a programmer) wondering why this issue just cropped up with 15.1...I've had this pi on my network doing DNS resolution since at least 13.0...

So how do I generate the segfault reports?
 
VulcanRidr said:
wondering why this issue just cropped up with 15.1...
Click to expand...
Given the nature of erratum 843419 on cortex a53 it depends on, roughly, having certain code at certain offset for the bug to trigger. We got lucky so far. clang/llvm know not to generate such code, given appropriate compiler option.

Just add the text of your original message in this thread to the bug report.
 
You must log in or register to reply here.
Back
Top