Possible runqemu release note for Nanbield

Possible runqemu release note for Nanbield

[Mark Hatle]

While porting meta-xilinx to nanbield I ran into a difference with the way 
runqemu works.  I figure this might be something that should be a release note, 
but unfortunately I'm not sure I understand how to explain it well.  Below are 
my findings and maybe we can word-smith this into a proper release note.

In Mickledore and before 'runqemu' when run like:

runqemu nographic

would automatically add "-serial mon:stdio" after the QB_OPT_APPEND values as 
necessary.  (It might have also added "-serial null" in some cases, but after 
"-serial mon:stdio".). After the commit:

https://git.yoctoproject.org/poky/commit/?h=nanbield&id=8c4bb95c1c472f9eb5235b848e57febcac53aa59

This has changed to counting the number of serial ports and adding "-serial 
null" if the QB_OPT_APPEND only defined a single serial port, and if two or more 
ports are specified avoiding automatically adding -serial mon:stdio.  This can 
break some existing QEMU based configurations by changing the way the serial 
ports are automatically determined when runqemu is executed.

It may be necessary to compare a prior working (Mickledore) executing and 
adjusting QB_OPT_APPEND to include the correct ordering and number of serial 
ports of the emulated machine.



Hopefully Richard or someone can verify what I wrong about is accurate, and 
maybe adjust it to better explain how a user should adjust their BSP integration.

Thanks!

--Mark

Re: [docs] Possible runqemu release note for Nanbield

[Michael Opdenacker]

Hi Mark

On 25.10.23 at 22:58, Mark Hatle wrote:
> While porting meta-xilinx to nanbield I ran into a difference with the 
> way runqemu works.  I figure this might be something that should be a 
> release note, but unfortunately I'm not sure I understand how to 
> explain it well.  Below are my findings and maybe we can word-smith 
> this into a proper release note.
>
> In Mickledore and before 'runqemu' when run like:
>
> runqemu nographic
>
> would automatically add "-serial mon:stdio" after the QB_OPT_APPEND 
> values as necessary.  (It might have also added "-serial null" in some 
> cases, but after "-serial mon:stdio".). After the commit:
>
> https://git.yoctoproject.org/poky/commit/?h=nanbield&id=8c4bb95c1c472f9eb5235b848e57febcac53aa59 
>
>
> This has changed to counting the number of serial ports and adding 
> "-serial null" if the QB_OPT_APPEND only defined a single serial port, 
> and if two or more ports are specified avoiding automatically adding 
> -serial mon:stdio.  This can break some existing QEMU based 
> configurations by changing the way the serial ports are automatically 
> determined when runqemu is executed.
>
> It may be necessary to compare a prior working (Mickledore) executing 
> and adjusting QB_OPT_APPEND to include the correct ordering and number 
> of serial ports of the emulated machine.
>
>
>
> Hopefully Richard or someone can verify what I wrong about is 
> accurate, and maybe adjust it to better explain how a user should 
> adjust their BSP integration.


Thanks for sharing this!
I'll write something from your explanations and I we'll see whether 
people have things to correct.
Cheers
Michael.

-- 
Michael Opdenacker, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com

Re: [docs] Possible runqemu release note for Nanbield

[Mark Hatle]

After following up with Richard, maybe this can be simplified?

For runqemu and automated (testimage) testing the Yocto Project requires a 
minimum of 2 working serial ports in Linux.  As of (commit below) this is now 
enforced by the code.

The stuff below is one _possible_ way of resolving this.

On 10/26/23 10:59 AM, Michael Opdenacker wrote:
> Hi Mark
> 
> On 25.10.23 at 22:58, Mark Hatle wrote:
>> While porting meta-xilinx to nanbield I ran into a difference with the
>> way runqemu works.  I figure this might be something that should be a
>> release note, but unfortunately I'm not sure I understand how to
>> explain it well.  Below are my findings and maybe we can word-smith
>> this into a proper release note.
>>
>> In Mickledore and before 'runqemu' when run like:
>>
>> runqemu nographic
>>
>> would automatically add "-serial mon:stdio" after the QB_OPT_APPEND
>> values as necessary.  (It might have also added "-serial null" in some
>> cases, but after "-serial mon:stdio".). After the commit:
>>
>> https://git.yoctoproject.org/poky/commit/?h=nanbield&id=8c4bb95c1c472f9eb5235b848e57febcac53aa59
>>
>>
>> This has changed to counting the number of serial ports and adding
>> "-serial null" if the QB_OPT_APPEND only defined a single serial port,
>> and if two or more ports are specified avoiding automatically adding
>> -serial mon:stdio.  This can break some existing QEMU based
>> configurations by changing the way the serial ports are automatically
>> determined when runqemu is executed.
>>
>> It may be necessary to compare a prior working (Mickledore) executing
>> and adjusting QB_OPT_APPEND to include the correct ordering and number
>> of serial ports of the emulated machine.
>>
>>
>>
>> Hopefully Richard or someone can verify what I wrong about is
>> accurate, and maybe adjust it to better explain how a user should
>> adjust their BSP integration.
> 
> 
> Thanks for sharing this!
> I'll write something from your explanations and I we'll see whether
> people have things to correct.
> Cheers
> Michael.
>