On Tue, Oct 1, 2024 at 2:12 AM Mikulas Patocka <mpatocka@xxxxxxxxxx> wrote: > > On Mon, 30 Sep 2024, Will Drewry wrote: > > > On Mon, Sep 30, 2024 at 1:07 PM Sami Tolvanen <samitolvanen@xxxxxxxxxx> wrote: > > > > > > While I agree that this sounds good in principle, devices that use the > > > restart feature typically need to pass the reboot reason to a PMU, for > > > example, and it looks like Android devices depend on reboot notifiers > > > for this. Here's a recent Pixel implementation that checks for the > > > "dm-verity device corrupted" command as an example: > > > > > > https://android.googlesource.com/kernel/google-modules/power/reset/+/refs/heads/android-gs-comet-6.1-android15/pixel-zuma-reboot.c#81 > > > > Good catch! (It's a shame it's using text as the API!) I totally > > spaced on that, and depending on where the corruption is, you mind end > > up in a boot loop. > > If I add that 'reboot-on-eio' flag, should it also restart the kernel with > kernel_restart("dm-verity device corrupted")? Or, should it use a > different string? I think it could be a different string since there are no existing users of this flag and the restart isn't actually caused by corruption, but I don't have strong feelings about this. Sami