On 27/12/18 17:59, Linus Torvalds wrote: > So the issue seems to be that syzbot is simply not useful enough. It's > output is too rough for people to take it seriously. You see how the > report by Wei Wu then got traction, because Wei took a syzbot report > and added some human background and distilled it down to not be > "here's a big dump of random information". We do take it seriously. Usually the reports are relatively easy to distill and fix, but when new random multi-threaded use-after-free comes, doing the bisection in syzkaller might not work because they are not deterministic in how much it takes to reproduce them. So the only way to process them is "look at when it started to happen and stare at 150 commits until you find the culprit", which is of course time consuming even though the syzkaller script usually gives a clue of which commit to look at. I agree with Linus that the report is more or less useless except for trivial bugs, but I'm not sure what can be done to improve it. I do use it for trivial bugs, and at the very least, having many different reports obviously means "use-after-free" or "dangling pointer" or some other kind of memory corruption. I try to prioritize those, but theory and practice are different. Paolo
