Our coverage reports have reached an enormous size (up to 500 MB / report). We now serve them as gzipped content, so at least we only transfer ~80MB, but it still requires a very fast computer to work with the page. It seems that we will have to abandon the idea of generating one single html file at...

We do get complaints about reporting filesystem bugs and not including the right mailing list: 1, 2. One more case: 3. This might probably be fixed by proper report parsing, but what about some extra fallback mechanism like "if the reproducer mentions syz_mount_image$ntfs, add the ntfs subsystem to...

In an email from Tetsuo Handa: By the way, a possible improvement on "Patch testing requests:" table. Although the "Patch" link showing diff output after applying proposed patch is OK, I'd like to also see a link to original "#syz test:" mail, for the intent of diff (which would be in patch descript...

OSS-Fuzz has found a bug in this project. Please see https://oss-fuzz.com/testcase?key=5145981141778432 for details and reproducers. This issue is mirrored from https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=51436 and will auto-close if the status changes there. If you have trouble accessing...

Updating the config requires syzkaller to be completely restarted, which can take quite a while with many VMs running. It would be nice to be able to reload the config file whilst syzkaller is still running. If the VM/process counts have increased and other settings have stayed the same, it can jus...

Currently we process release tags one by one, which can be a real problem if there are many tags. We could bisect them instead. The related code is here: syzkaller/pkg/bisect/bisect.go Lines 379 to 389 in 356d821...

Currently network namespaces are shared across multiple tests. This was done for performance reasons. Creation/destruction of net ns is slow + it's globally serialized thus parallel test processes suffer + destruction is asynchronous and we can create them faster than they are destroyed and deplete...

Currently we extract thew function name for the title before symbolization, as the result inlined functions are never considered. Inlining can depend on compiler, arch, potentially config and unrelated code changes (e.g. some function is optimized to be inlinable). Perhaps we need to consider doing...

Running syzkaller with newer kernel versions, I am getting: 2022/09/05 14:15:31 wifi device emulation : failed to parse kernel version (6.0.0-rc4) 2022/09/05 15:25:32 wifi device emulation : failed to parse kernel version (6.0.0-rc4^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@...

In order to re-test a repro, we now build the most up-to-date version of the kernel and run it there. The problem is that over time new options might have become necessary to successfully build and run the kernel in the syzbot environment. The obvious solution is to take not only the branch/repo fro...