On Thu, Jan 05, 2023 at 10:49:53AM +0800, Huacai Chen wrote:
> On Thu, Jan 5, 2023 at 2:37 AM Bjorn Helgaas <helgaas@xxxxxxxxxx> wrote:
> > On Tue, Jan 03, 2023 at 03:34:01PM +0800, Huacai Chen wrote:
> > > cc27b735ad3a7557 ("PCI/portdrv: Turn off PCIe services during shutdown")
> > > causes poweroff/reboot failure on systems with LS7A chipset. We found
> > > that if we remove "pci_command &= ~PCI_COMMAND_MASTER" in do_pci_disable
> > > _device(), it can work well. The hardware engineer says that the root
> > > cause is that CPU is still accessing PCIe devices while poweroff/reboot,
> >
> > Did you ever figure out what these CPU accesses are? If we call the
> > Root Port .shutdown() method, and later access a downstream device,
> > that seems like a problem in itself. At least, we should understand
> > exactly *why* we access that downstream device.
>
> Maybe I failed to get the key point, but from my point of view, the
> root cause is clear in previous discussions:
> https://lore.kernel.org/linux-pci/CAAhV-H5uT+wDRkVbW_o1hG2u0rtv6FFABTymL1VdjMMD_UEN+Q@xxxxxxxxxxxxxx/
> https://lore.kernel.org/linux-pci/20220617113708.GA1177054@bhelgaas/
> https://lore.kernel.org/linux-pci/CAAhV-H6raQnXZ4ZZRq19cugew26wXYONctcFO0392gZ00LC6bw@xxxxxxxxxxxxxx/
That's great, but the root cause should be summarized here in the
commit log.
> > To be clear, cc27b735ad3a does not cause the failure. IIUC, the cause
> > is:
>
> cc27b735ad3a is not a bug, we refer to it just because we observe
> problems after it.
Right. But you said "cc27b735ad3a ... causes failure," which is not
quite true. cc27b735ad3a may *expose* an LS7A hardware defect that
previously didn't cause a problem, but I don't want to blame
cc27b735ad3a for that hardware issue.
Bjorn
