pkg/ifaceprobe: note error only when res.Err is not null

[jiangenj]

---

Before sending a pull request, please review Contribution Guidelines:
https://github.com/google/syzkaller/blob/master/docs/contributing.md

---

[a-nogikh]

What are the cases when res.Status != queue.Success but res.Err is nil?

[jiangenj]

This is one kind of error:

[FATAL] interface probing failed: failed to execute prog: %!w(<nil>) (Crashed)
r0 = openat(0xffffffffffffff9c, &(0x7f0000000080)='/sys/devices/pci0000:00/0000:00:01.1/ata1/host0/target0:0:0/0:0:0:0/block/sda/queue/scheduler', 0x1, 0x0)
write(r0, &(0x7f0000000100)=' ', 0x1)

exit status 1

[a-nogikh]

The fact that the kernel crashed when the tool was probing an interface does seem to me like a valid case for reporting an error. Otherwise we will get different descriptions each run depending on what programs were not properly executed.

Cc @dvyukov who wrote this code.

[dvyukov]

I don't know what's the best answer here. Both failing the tool and ignoring the kernel bug bad. Ideally of course kernel does not crash on such simple actions like opening a file.

One recommendation when running the tool is to add this to the manager config:

	"cmdline": "panic_on_warn=0"
	"ignores": [".*"]

And maybe configuring the kernel w/o most debugging configs (it only needs KCOV).
Joey, does this help in your case?

[jiangenj]

Hi @dvyukov ,
Tried

	"cmdline": "panic_on_warn=0"
	"ignores": [".*"]

and disabled CONFIG_KASAN, still see this

2025/05/16 16:04:12 VM 9: crash: no output from test machine
2025/05/16 16:04:12 VM 11: crash: no output from test machine
2025/05/16 16:04:12 VM 17: crash: no output from test machine
2025/05/16 16:04:18 exec total=366811 (545/sec) pending=17 reproducing=0
2025/05/16 16:04:19 VM 5: crash: no output from test machine
2025/05/16 16:04:20 done file 61000
2025/05/16 16:04:20 VM 14: crash: no output from test machine
2025/05/16 16:04:21 VM 19: crash: no output from test machine
2025/05/16 16:04:21 VM 1: crash: no output from test machine
2025/05/16 16:04:22 VM 18: crash: no output from test machine
2025/05/16 16:04:22 VM 4: crash: no output from test machine
2025/05/16 16:04:23 [FATAL] interface probing failed: failed to execute prog: %!w(<nil>) (Crashed)
r0 = openat(0xffffffffffffff9c, &(0x7f0000000040)='/sys/fs/cgroup/system.slice/memory.max', 0x1, 0x0)
write(r0, &(0x7f0000000080)=' ', 0x1)

can you share your .config and manager.cfg, I'm currently using dashboard/config/linux/upstream-apparmor-kasan.config without CONFIG_KASAN.

I also tried defconfig, however it will enter into rescue mode (no network) when starting bullseye.img.

[dvyukov]

I did not record exact kernel revision/config, etc. But also used something similar to dashboard/config/linux/upstream-apparmor-kasan.config.

If looks like your kernel/VM is somehow mis-configured based on very frequent "no output from test machine". Do you run qemu w/o KVM by any chance? Does syz-manager work fine with this exact config?