05.07.2021 05:21, Peter Chen пишет: > On 21-07-03 20:22:38, Dmitry Osipenko wrote: >> 03.07.2021 14:08, Peter Chen пишет: >>> On 21-07-02 02:43:10, Dmitry Osipenko wrote: >>>> The HNP work can be re-scheduled while it's still in-fly. This results in >>>> re-initialization of the busy work, resetting the hrtimer's list node of >>>> the work and crashing kernel with null dereference within kernel/timer >>>> once work's timer is expired. It's very easy to trigger this problem by >>>> re-plugging USB cable quickly. Initialize HNP work only once to fix this >>>> trouble. >>> >>> Fully OTG compliance support has not maintained for years, what's the use case you >>> still want to use? >> >> I don't have any use case for it, but I had CONFIG_USB_OTG_FSM=y and it >> was crashing kernel badly. The OTG works perfectly fine without the FSM. > > You could add below at your dts to disable OTG FSM: > hnp-disable > srp-disable > adp-disable > > Since there are no users for OTG FSM, it hasn't maintained for years, > I am not sure if it still works OK. If I remember correctly, the VBUS > will be off if you enable HNP, and the device at the host port will be > disconnected, that's may not your expectation. > Since OTG FSM is known to be in a bad shape, could you please make a patch to remove it? I hope it's not enabled by default in a distro kernels.. oh no, CONFIG_USB_OTG_FSM=y at least in ArchLinux [1]. [1] https://archlinuxarm.org/packages/armv7h/linux-armv7/files/config I think we should fix that hrtimer bug, beackport the fix into stable kernels and then remove OTG FSM. Does this sound good to you?
