Re: [PATCH] hw/acpi: Set memory regions to native endian as a work around

[BALATON Zoltan]

On Tue, 7 Mar 2023, Paolo Bonzini wrote:
> On 3/6/23 23:56, Paolo Bonzini wrote:
> > On 2/21/23 13:55, BALATON Zoltan wrote:
> > > To get that menu with Shut Down, first Quit the installer then again right 
> > > click or click on background first to get the menu of the Ambient desktop. 
> > > I also see an error from the firmware at the beginning:
> > > Initializing KBD...00000012    FAILED.
> > > when it's broken and it says Done without the hex number when it works. 
> > > (Two other FAILED messages about clock chip is normal as we don't emulate 
> > > that but all others should be green.)
> >
> > Ok, I've reproduced it.  The mouse is a bit flaky but using the keyboard 
> > for everything except right clicking works better.
>
> Now the OS doesn't boot anymore, it doesn't get to the point where it 
> initializes the VGA.

Sorry, that is a known breakage from a recent series that's not yet fixed. 
This patch should fix it:

http://patchew.org/QEMU/cover.1677628524.git.balaton@eik.bme.hu/cdfb3c5a42e505450f6803124f27856434c5b298.1677628524.git.balaton@eik.bme.hu/

or another proposed altenative is here:

20230304114043.121024-1-shentey@gmail.com/20230304114043.121024-2-shentey@gmail.com/">http://patchew.org/QEMU/20230304114043.121024-1-shentey@gmail.com/20230304114043.121024-2-shentey@gmail.com/

Either of the above should work.

> I got some quick logs with .impl.min_access_size to 1, 
> to understand what the firmware (but not the OS) does.  With this at least I 
> could confirm that your patch is wrong:
>
> cnt 1 1 write 80
> evt 3 1 write 1         // enable timer
> evt 0 2 read
> evt 0 2 write 1         // just writes again the same value
> evt 1 1 write
> evt 1 1 write 0
>
> Since you have both 1-size and 2-size writes, and the 2-byte reads/writes are 
> done with byte swapping instructions lhbrx and sthbrx, your patch would cause 
> 0x100 to be read and written on the third and fourth lines.
>
> Likewise, any 4-byte read of the timer port would be byte swapped.
>
> The solution is a patch similar to mine, applied to both evt and cnt; while 
> perhaps tmr can be left as is and only accept 4-byte reads, I don't know. 
> I'll try to come up with something that can be tested at least with the 
> firmware.

I'm not sure I follow what you mean so I'd need a patch to see then I can 
test it with the clients I run on pegasos2.

Regards,
BALATON Zoltan