syzkaller: testing failed: failed to run ["go" "test" "-short" "./..."]: exit status 1 (11)

[syzbot]

Hello,

syzbot found the following crash on:

HEAD commit:
git tree: https://github.com/google/syzkaller.git master
console output: https://syzkaller.appspot.com/x/log.txt?x=101c2c6c100000
dashboard link: https://syzkaller.appspot.com/bug?extid=33ec745deb35b45ccd0c
compiler: go version go1.13.9 openbsd/amd64

Unfortunately, I don't have any reproducer for this crash yet.

IMPORTANT: if you fix the bug, please add the following tag to the commit:
Reported-by: syzbot+33ec74...@syzkaller.appspotmail.com



---
This bug is generated by a bot. It may contain errors.
See https://goo.gl/tpsmEJ for more information about syzbot.
syzbot engineers can be reached at syzk...@googlegroups.com.

syzbot will keep track of this bug report. See:
https://goo.gl/tpsmEJ#status for how to communicate with syzbot.

[Dmitry Vyukov]

Reposting on the bug report itself as well:

On Thu, May 28, 2020 at 8:13 AM Anton Lindqvist <an...@basename.se> wrote:
>
> Hi,
> Looks like `go test' timeouts on OpenBSD[1][2]. Anyone care to take a
> look?
>
> Thanks,
> Anton
>
> [1] https://syzkaller.appspot.com/bug?id=fa01c6fd4323ecfd78220f6648e4e1ff3f3ced0d
> [2] https://syzkaller.appspot.com/text?tag=CrashLog&x=17770b7e100000

Hi Anton,

I looked at the instance, but I don't see what we can fix quickly.
First, I checked older similar bugs from Jan-Feb and they are about
different things. For this one we have only 2 occurrences (each 3x
times) on May 28 and May 5. So it does not happen every other time.

I checked the performance of the machine. While it's 2-3x slower than
my laptop when not loader too much:

$ go test -short ./pkg/compiler/
ok github.com/google/syzkaller/pkg/compiler 1.593s
$ go test -short ./prog
ok github.com/google/syzkaller/prog 7.948s

vs

# go test -short ./pkg/compiler/
ok github.com/google/syzkaller/pkg/compiler 4.457s
# go test -short ./prog
ok github.com/google/syzkaller/prog 16.282s

The difference does not seem to be too radical. That's not worse than
what we have on github CI.

prog tests are far from exceeding the 600s limit.

My current hypothesis is that that testing happened concurrently with
a kernel build.

Well, now looking at this more closely, we probably can fix it
relatively easily by shoving kernelBuildSem somewhere in
syz-ci/updater.go. This should serialize it with all kernel builds.

[Dmitry Vyukov]

I've pushed https://github.com/google/syzkaller/commit/c7192a2f9a8eb9f99abe0b2878d4756d8bcc7b53
Let's see if it will help.

#syz invalid

[Anton Lindqvist]

... and we're back. Thanks for the fix!