Open Bug 1270130 Opened 5 years ago Updated 3 years ago
Driver restarts without process crash go unreported
<The_8472> do you track driver restarts that don't crash the process? i've had a few recently that did lead to black e10s tabs but the content process didn't crash <nical> The_8472: we add annotations to the crash report if a crash occurs for this session. We don't get any information reported automatically if the process does not crash I had to manually kill the content process in that case and I can't reproduce this issue right now, but I think it might be useful to track driver restarts since they can lead to buggy behavior.
We do currently track this with telemetry; we don't get a particular information from each user, but we get cumulative information as to how many sessions contain a driver reset (currently ~0.18%). However, people that do get driver resets, get a lot of them, and Windows 10 has been particularly disproportionately bad. We understand Microsoft and Intel are working on making improving the situation. Also - next time this happens, check if we note it in about:support (the graphics section?) So, we do track it, but we don't currently do anything drastic about it (e.g., we don't force a process crash.) Eventually, if we move into a separate graphics process type of setup, we would probably do exactly that - restart it, so that we get a clean sheet after the driver reset. But we do not want to do that until the graphics and content processes are separate.
Just to confirm, is it tracked in e10s child processes too? I was running with > 1 content processes and only one was affected, so I assume it was something that happened in the child.
The 8472, does it occur if dom.ipc.processCount is 1?
I haven't tried and it's not what this bug is about. I just want to ensure that it shows up in some telemetry so the incidence rate doesn't get any worse, it's kinda tolerable at the moment.
(In reply to The 8472 from comment #4) > I haven't tried and it's not what this bug is about. I just want to ensure > that it shows up in some telemetry so the incidence rate doesn't get any > worse, it's kinda tolerable at the moment. In the words of Yoda, tracked it is - http://people.mozilla.org/~danderson/moz-gfx-telemetry/www/#view=tdrs Probably nothing actionable on this right now, but let's keep it open for future reference.
Good enough for me.
3 years ago
Priority: -- → P3
You need to log in before you can comment on or make changes to this bug.