[ltp] E575 screen freezing (was:) Re: Dual-boot with Windows 10

[cr]

On Mon, 26 Apr 2021 08:46:02 +0300
Marius Gedminas <marius at gedmin.as> wrote:

> On Mon, Apr 26, 2021 at 12:33:54PM +1200, cr wrote:
> > When Debian (10.9.0 AMD64) first boots up, everything works.   But
> > if I go to run e.g. Synaptic, it draws the first window and then
> > freezes, with it part-drawn. The mouse pointer moves but nothing
> > else works, keyboard, mouseclicks, nothing.   As if it's jammed
> > up. Capslock toggles its light.   No key combinations have any
> > effect. I don't think it's a keyboard issue though.   Only escape
> > is the power button.
> 
> Wait, even Alt+SysRq+S,U,B doesn't do anything?
> 
> Is it enabled in /proc/sys/kernel/sysrq?

No Sysrq key on E575, according to Lenovo Fn+S does the same thing.
And no, Fn+S or Alt+Fn+S had no effect.   
/proc/sys/kernel/sysrq  just says "438"

> 
> > I have ssh'd in.   And that keeps working even when the E575's
> > screen is jammed.   So wifi is still working.  Very handy as I can
> > view lspci, /var/log etc.
> > 
> > Yet, if I first launch LibreOffice Writer instead, which I assume
> > is a memory-hungry app, it works fine, at least for a while.   If I
> > launch a terminal window I can carry on inside it for ages.   So it
> > seems to be something in the window manager that's getting stuck.  
> 
> Sounds like a bug in the graphics driver is causing your GPU to freeze
> maybe?  Any interesting messages in dmesg when the GUI is frozen?
> What sort of video card does the machine have -- Intel, AMD, NVidia?
 
Video controller is AMD Carrizo, video card is Radeon R5, kernel driver
is amdgpu, and apparently it has 2 CPU cores and 4 GPU cores.

> > (Now, something must have timed out, as the screen started to fade,
> > moving the mouse brought it back and the Task Manager window popped
> > up.   Showing CPU usage 12%, memory 396MB of 7441 MB used.   All
> > tasks showing 0% CPU%.   But I still can't scroll it, move it or do
> > anything else with it.)
> 
> (This rules out what would've been my first guess, swapping to death.)

Would have been my first guess too.   (Anecdote - my father's first PC
was a 386-20 with minimum RAM and a 20MB(?) hard drive that someone had
put Win95 and Doublespace on it.   When you went to a new page in Word
you could watch the letters appearing one by one on the screen as the
hard drive thrashed itself to death caching and uncompressing...)
 
> > (Must run memtest but I doubt it's RAM-related, that would be too
> > easy)
> > 
> > I updated the BIOS (first time I ever flashed a BIOS.   Lenovo
> > provide a Windows updater but also, to their credit, a bootable
> > ISO.  I couldn't get it to boot off USB - probably something I did
> > worng making the USB stick - but it did boot off a CD).   But no
> > change to the problem.
> 
> Oh yeah, I'm so happy Lenovo provides BIOS updates via fwupd for their
> more recent machines.
> 
> Marius Gedminas

Progress (of a sort).   The errors were
IOAPIC[4] not in IVRS table
IOAPIC[5] not in IVRS table
No Southbridge IOAPIC found

So (after much Googling) I tried adding (in /etc/default/grub) the boot
parameter  ivrs_oapic[4]=00:14.0    -  no change seen.
Nor for  iommu=soft   or  ivrs_oapic[4]=00:00.2  ivrs_oapic[5]=00:14.0
or  intremap=off  -  none of those had any effect.

However, Linux Mint Debian Edition (LMDE4) runs fine off USB, though its
syslog file shows the exact same IOAPIC errors.

And I found LMDE4 will install to a UEFI partitioning scheme, I just
hadn't clicked the right box when I tried yesterday.   And it has 
NO additional kernel boot parameters in default/grub.   So now I have a
working Debian-derived LMDE installation in /sda3 and Debian in /sda2,
so I should (hopefully) be able to discover what Mint does to make it
work in this system.

I'll post again if I do.

Thanks and regards to all

Chris