Comment 217 for bug 1671360

Created attachment 255151
kernel config

I'm trying to run ubuntu on Ryzen 1700x with Gigabyte GA-AB350-gaming-3 motherboard,
and it has a load of problems, starting with not being able to boot normally.

During normal boot, on 16.10 as well as 17.04 beta:
system doesn't boot normally, hangs with a lot of "unexpected irq trap at vector 07"
messages displayed.

Following advice from various places, I've tried to:

1. add "acpi=off" to boot params

That helps, allowing me to boot into recovery mode, though it leaves me with system seeing only one core, is really slow and still only boots in recovery mode.

2. Compile own kernel using 4.11.rc1 and disabling cpu freq governor and cpu handling in acpi settings. Boot with "quiet loglevel=3" option.

That gets me even further - system sees all cores now. Still only recovery mode though,
but its enough to get info for this bug report.

Some observed problems:

1. dmesg reports *a lot* of messages like this all the time:

[ 163.362068] ->handle_irq(): ffffffff87a7e090,
[ 163.362081] bad_chained_irq+0x0/0x40
[ 163.362089] ->handle_irq(): ffffffff87a7e090,
[ 163.362090] amd_gpio_irq_handler+0x0/0x200
[ 163.362090] ->irq_data.chip(): ffffffff88587e20,
[ 163.362090] ioapic_ir_chip+0x0/0x120
[ 163.362090] ->action(): ffffffff884601c0
[ 163.362091] IRQ_NOPROBE set
[ 163.362099] ->handle_irq(): ffffffff87a7e090,
[ 163.362099] amd_gpio_irq_handler+0x0/0x200
[ 163.362100] ->irq_data.chip(): ffffffff88587e20,
[ 163.362100] ioapic_ir_chip+0x0/0x120
[ 163.362101] ->action(): ffffffff884601c0

I've tried to redirect dmesg to a file, stopped after a short while, it generated 400M of those.

2. Systemd cannot start journald. Perhaps because it cannot cope with amount of kernel logs?

3. Looking at pci, I've noticed something called AMDI0040 (/sys/bus/acpi/devices/AMDI0040, path=_SB_.EMMC), among AMDI0010, AMDI0020, AMDI0030. Those however are mentioned in kernel source, kernel and google are completely silent about AMDI0040.

Phoronix tested ryzen using different motherboard, and it worked better (though not well),
so I suspect it is an issue with motherboard.