On 18.04.22 23:00, Evan Green wrote: > The documentation for the freeze() method says that it "should quiesce > the device so that it doesn't generate IRQs or DMA". The unspoken > consequence of not doing this is that MSIs aimed at non-boot CPUs may > get fully lost if they're sent during the period where the target CPU is > offline. > > The current behavior of the USB subsystem still allows interrupts to > come in after freeze, both in terms of remote wakeups and HC events > related to things like root plug port activity. This can get controllers > like XHCI, which is very sensitive to lost interrupts, in a wedged > state. This series attempts to fully quiesce interrupts coming from USB > across in a freeze or quiescent state. > > These patches are grouped together because they serve a united purpose, > but are actually independent. They could be merged or reverted > individually. Hi, sorry for being a bit late in this discussion. There was something that I didn't remember immediately. We have a set of quirky devices that need HID_QUIRK_ALWAYS_POLL. They have the nasty firmware bug that, if you suspend them without remote wakeup, they will crash or reset themselves. I am afraid that has an obvious relevance to your cool patches. I am not completely sure how to deal with this. It seems to me that the quirk will need to be shifted from HID to core USB and thaw() needs to be translated into usb_device_reset() + reset_resume() for them, but I am not really sure about the optimal mechanism. Regards Oliver
