On Thu, Apr 07, 2022 at 02:15:29PM +0800, WeitaoWang-oc@xxxxxxxxxxx wrote: > On 2022/4/7 00:20, Alan Stern wrote: > > On Wed, Apr 06, 2022 at 10:38:28AM +0800, WeitaoWang-oc@xxxxxxxxxxx wrote: > > > On 2022/4/6 00:02, Alan Stern wrote: > > > > In fact, the resume kernel doesn't call ehci_resume at all. Here's what > > > > it does: > > > > > > > > The resume kernel boots; > > > > > > > > If your patch causes STS_PCD to be set at this point, the flag > > > > should get cleared shortly afterward by ehci_irq; > > > > > > > > ehci-hcd goes into runtime suspend; > > > > > > > > The kernel reads the system image that was stored earlier when > > > > hibernation began; > > > > > > > > After the image is loaded, the system goes into the freeze > > > > state (this does not call any routines in ehci-hcd); > > > On this phase, pci_pm_freeze will be called for pci device. In this > > > function, pm_runtime_resume will be called to resume already > > > runtime-suspend devices. which will cause ehci_resume to be called. > > > Thus STS_PCD flag will be set in ehci_resume function. > > > > Aha! I was missing that piece of information, thanks. > > > > But this still doesn't explain why check_root_hub_suspended is failing. > > That routine checks the HCD_RH_RUNNING bit, which gets set in > > hcd_bus_resume. hcd_bus_resume gets called as part of resuming the root > > hub, and in ehci-hcd this happens when ehci_irq sees that STS_PCD is set > > and calls usb_hcd_resume_root_hub. That routine queues a wakeup request > > on the pm_wq work queue, which is then supposed to run hcd_resume_work > > to actually restart the root hub. > > > > But pm_wq is a freezable work queue! While the system is in the freeze > > state, the work queue isn't running. This means that the root hub > > should remain suspended until the end of the freeze phase, and so the > > call to check_root_hub_suspended should succeed. > > > > Can you check to see what's really happening on your system? Something > > must be wrong with my analysis, but I can't tell what it is. I'm still > > puzzled. > > > > Alan Stern > Your analysis is right, my test platform's kernel version is not the > latest, this kernel not call freeze_kernel_threads on software_resume > function. > (https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/kernel/power/hibernate.c?h=v5.18-rc1&id=2351f8d295ed63393190e39c2f7c1fee1a80578f) > So pm_wq is active and can handle root hub power events. > Update my kernel to fix the issue in the url above, system hibernation > test was successful with our patch(not clear STS_PCD bit). > Thanks for your clarification. Great! I'm glad we sorted that out. So check_root_hub_suspended doesn't need any changes, and the patch you already submitted takes care of everything. Alan Stern
