In several places we do special handling for some crash types.
Currently we compare report title with magic strings,
which is error-prone. Add explicit Type to reports.
We are seeing some flakes during bisection and image testing.
Hard to tell what's the root cause because they are episodic.
But using non-preemptible VMs for bisection and image testing
looks good on all fronts.
Update #501
Running bisection using an older repro failed with:
<stdin>: In function ‘syz_mount_image.constprop’:
<stdin>:298:3: error: argument 1 null where non-null expected [-Werror=nonnull]
In file included from <stdin>:26:0:
/usr/include/x86_64-linux-gnu/sys/stat.h:320:12: note: in a call to function ‘mkdir’ declared here
extern int mkdir (const char *__path, __mode_t __mode)
Let's be safe and ignore warnings during repro/bisect.
Everything that runs during tests still has all warnings.
Update #501
Currently we truncate output up to rep.EndPos after unexpected reboot.
But report sets EndPos to the _last_ report in output,
so if there are any other errors they are all skipped after
truncation to EndPos. Truncate just one line instead.
Since we can be building very old revisions for bisection here,
make the build as permissive as possible.
Newer compilers tend to produce more warnings also kernel headers may be broken, e.g.:
ebtables.h:197:19: error: invalid conversion from ‘void*’ to ‘ebt_entry_target*’
Update #501
Currently we always pass cover=false during instance testing
and as the result covereage is not tested. This can result in
broken images. Pass the actual value of coverage for testing.
Currently we only support canExit flag.
However there are actually 3 separate conditions:
- program can exit normally
- program can timeout (e.g. fuzzer test or runtest can't)
- program can exit with error (e.g. C test can)
Allow to specify these 3 conditions separately.
We currently have this list in multiple places (somewhat diverged).
Specify this "overcommit" property in VM implementations.
In particular, we also want to allow overcommit for "vmm" type.
Update #712
The tool is run as:
$ syz-runtest -config manager.config
This runs all programs from sys/*/test/* in different modes
on actual VMs and checks results.
Fixes#603
mgrconfig was used only by syz-manager initially,
but now it's used by a dozen of packages and it's
weird to import from under a binary dir.
pkg/ is much more reasonable dir for a widely used
helper package.
Only akaros needs OS, because the rest assume host OS.
But speciying OS for all OSes breaks patch testing on syzbot
because old execprog does not have os flag.
Currently all (linux-specific) suppressions are hardcoded in mgrconfig.
This is very wrong. Move them to pkg/report and allow to specify per OS.
Add gvisor-specific suppressions.
This required a bit of refactoring. Introduce mgrconfig.KernelObj finally.
Make report.NewReporter and vm.Create accept mgrconfig directly
instead of passing it as multiple scattered args.
Remove tools/syz-parse and it always did the same as tools/syz-symbolize.
Simplify global vars in syz-manager/cover.go.
Create reporter eagerly in manager. Use sort.Slice more.
Overall -90 lines removed.
Strictly saying, we may not get the connection when
the fuzzer process exits. The accepting goroutine
may have not been scheduled yet.
For the connection for up to 10 seconds.
Currently kernel build failures are insanely verbose
(contain full kernel build output) and there is no
way to separate short descriptions from full output.
Make it possible.
Also try to extract failure root cause froom build log.
Use this in pkg/bisect to not pollute log on build failures.
Update #501
Only check that syzkaller path is in GOPATH if we are going to build it.
syz-ci image testing does not have syzkaller path in GOPATH,
but it also does not build syzkaller.
