[OpenIndiana-discuss] Unbootable hipster upgrade with Nvidia USB investigation

[Gary Mills]

The symptoms I've related before.  A hipster upgrade or a recent DVD
image fails to boot.  At the same time, my USB keyboard and mouse stop
working.  I've filed this bug report:

    https://www.illumos.org/issues/8684

These are some of the telltale console messages:

    WARNING: /pci at 0,0/pci1043,81c0 at b,1 (ehci0): No SOF interrupts have been received, this USB EHCI controller is unusable
    WARNING: /pci at 0,0/pci1043,81c0 at b (ohci0): No SOF interrupts have been received, this USB OHCI controller is unusable

This system has an ASUS M2NPV-VM motherboard, with an AMD Athlon(tm)
64 X2 Dual Core CPU, and NVIDIA Corporation MCP51 USB ports.  It's
running the 1401 BIOS version.  The hipster-20160421 BE boots and runs
nicely.  It's only when I upgrade to a 2017 version of hipster or boot
a recent hipster DVD that the problem appears.

Articles that I've read say that Linux had the same problem with the
same USB controller.  The recommended solution is to add this boot
option:

    acpi_use_timer_override=1

On Linux, the immediate cause is HPET (high precision event timer)
support in the kernel.  Apparently, some BIOS versions contained an
incorrect HPET table.  The result was that interrupts appeared on the
wrong line on the motherboard.  This table error was corrected in
later BIOS versions.  Then, that boot option is no longer required.

What's significant here is that a timer error can affect the USB
controller.  The same thing could happen in illumos, even though the
kernels are quite different.

I did notice some curious lines in the messages logs for my current
OS, hipster-20160421:

    Sep 22 17:01:34 amd acpica: [ID 429496 kern.notice] ACPI: HPET c7ef9f80 00038 (v1 Nvidia ASUSACPI 42302E31 AWRD 00000098)
    Sep 22 17:01:34 amd pcplusmp: [ID 801116 kern.info] NOTICE: ACPI HPET table query failed

Has anyone else seen these?  I assume that the first line indicates
that the HPET table exists in the BIOS, but that the second line
indicates that the table cannot be used.  The second line came from
the acpi_probe() function in this file:

    usr/src/uts/i86pc/io/mp_platform_common.c

This new information leaves me with some questions.  Why did the query
fail?  I need more debugging information to tell that.  What happens
with this table when I attempt to boot the upgraded hipster?  I need
more debugging for that too.  I suppose I should dump the HPET table,
just to see what it contains.

My next adventure will be to rig up a serial console, and use that to
collect more debugging information.  In the meantime, any guidance or
insights would be welcome.


-- 
-Gary Mills-		-refurb-		-Winnipeg, Manitoba, Canada-