svn commit: r351514 - head/sys/dev/pci

[Alexander Motin]

On 26.08.2019 18:00, Conrad Meyer wrote:
> On Mon, Aug 26, 2019 at 10:27 AM Alexander Motin <mav at freebsd.org> wrote:
>>
>> Author: mav
>> Date: Mon Aug 26 17:27:36 2019
>> New Revision: 351514
>> ...
>> -               device_printf(dev,
>> -                   "HotPlug Command Timed Out - forcing detach\n");
>> -               sc->flags &= ~(PCIB_HOTPLUG_CMD_PENDING | PCIB_DETACH_PENDING);
>> -               sc->flags |= PCIB_DETACHING;
>> -               pcib_pcie_hotplug_update(sc, 0, 0, true);
>> +               device_printf(dev, "HotPlug Command Timed Out\n");
>> +               sc->flags &= ~PCIB_HOTPLUG_CMD_PENDING;
> 
> If this case is expected (and it seems like it is, at least in
> relatively common hardware), and the kernel isn't taking an action
> based on it anymore (formerly, it would then cause detach), is there
> any reason to print a message at all?  What information should a user
> glean from the message, or what action should be taken?  It seems like
> Nothing and None, respectively.  So perhaps it is better to not print
> anything in this case?  Or maybe mask it behind bootverbose.

While you may be right, for now I would leave it visible just to show
that something went wrong, in case somebody wish to diagnose some real
issue that may have same symptoms.  It is just one line per event, so it
should not be too annoying.  Most people will never see it, since PCIe
hot-plug is still very rare.  On the mentioned Supermicro board ACPI
actually blocks hot-plug support, since FreeBSD does not announce ASPM
capability.  This is the next real issue we need to think about.
Cosmetics will follow later.

-- 
Alexander Motin