[9front] vac crash in libthread threadexitsall()

[9front] vac crash in libthread threadexitsall()

[james palmer]

i found a weird vac crash.
it traps in libthread's threadexitsall():
vac 420: suicide: sys: trap: general protection violation pc=0x2170aa

here's a stack trace:

acid: lstk()
threadexitsall(exitstr=0x404bf9)+0xcd /sys/src/libthread/exit.c:50
	mypid=0x1000001a4
	npid=0x1
	pid=0x76e3a8d
	i=0x76e3a8d00000000
threadmain(argc=0xfefefefe00000017,argv=0x7fffffffee40)+0x688 /sys/src/cmd/vac/vac.c:286
	blocksize=0x2000
	stdinname=0x0
	printstats=0x0
	fsdiff=0x0
	diffvac=0x0
	_argc=0x32fefefe00000034
	_args=0x426f38
	fdiff=0x0
	outfd=0x200000000003
	oldpath=0xfefefefefefefefe
	tm=0x1400000000
	yyyy=0x426f380031323032
	fp=0x426f38
	f=0x85db80
	mmdd=0xfe00000034313230
	n=0xfefefefe00000000
	i=0x216b1700000017
	fd=0xfefefefefefefefe
	d=0x8c8470
	j=0x17fefefefe
	s=0xfefefefefefefefe
	qid=0x15f1f
mainlauncher()+0x18 /sys/src/libthread/main.c:59
launcheramd64(arg=0x421f68,f=0x216aff)+0x10 /sys/src/libthread/amd64.c:11
0xfefefefefefefefe ?file?:0

Re: [9front] vac crash in libthread threadexitsall()

[cinap_lenrek]

you need to take a look at the memory image.

obviously, p is nil or garbage (you havnet told us the address
it faults on), but this should not happen. 

it is most likely that a memory corruption has broken the 
Proc* structures or at least the next pointer in one
of the proc's or the global _threadpq.

is there a way to reproduce it?

maybe you can make a memory snapshot if the process with
snap(4) when it happens again.

--
cinap

Re: [9front] vac crash in libthread threadexitsall()

[james palmer]

[-- Attachment #1: Type: text/plain, Size: 228 bytes --]

Quoth cinap_lenrek@felloff.net:
> 
> maybe you can make a memory snapshot if the process with
> snap(4) when it happens again.
> 

i didn't kill it so i just made a snapshot.
i hope it isn't too large of an attachment.

- james

[-- Attachment #2.1: Type: text/plain, Size: 337 bytes --]

from postmaster@1ess:
The following attachment had content that we can't
prove to be harmless.  To avoid possible automatic
execution, we changed the content headers.
The original header was:

	Content-Disposition: attachment;filename="vac.snap"
	Content-Type: application/octet-stream; name="vac.snap"
	Content-Transfer-Encoding: BASE64

[-- Attachment #2.2: vac.snap.suspect --]
[-- Type: application/octet-stream, Size: 1548342 bytes --]

Re: [9front] vac crash in libthread threadexitsall()

[Noam Preil]

>
> is there a way to reproduce it?
>
> maybe you can make a memory snapshot if the process with
> snap(4) when it happens again.

This happens consistently if the backup has a vac file in it with merge, 
i believe.

Re: [9front] vac crash in libthread threadexitsall()

[james palmer]

Quoth noam@pixelhero.dev:
> This happens consistently if the backup has a vac file in it with merge, 
> i believe.

thats weird, i'm not using merge, i'm using -a.
maybe it's something to do with adding files to an existing archive?

- james