Am 28. September 2026 18:05:11 UTC schrieb Peter Maydell
<[email protected]>:
>On Mon, 21 Sept 2026 at 22:47, Bernhard Beschow <[email protected]> wrote:
>>
>> When using mcimx7d-sabre with newer Linux versions such as 7.0, the
>> machine fails to boot. This happens due to memory transaction failures
>> since the address space is not fully covered with unimp devices. Fix
>> booting by ignoring memory transaction failures in the machine.
>>
>> cc: qemu-stable
>> Signed-off-by: Bernhard Beschow <[email protected]>
>> ---
>> hw/arm/mcimx7d-sabre.c | 1 +
>> 1 file changed, 1 insertion(+)
>>
>> diff --git a/hw/arm/mcimx7d-sabre.c b/hw/arm/mcimx7d-sabre.c
>> index 65fdb19c06..8d223fca6c 100644
>> --- a/hw/arm/mcimx7d-sabre.c
>> +++ b/hw/arm/mcimx7d-sabre.c
>> @@ -83,6 +83,7 @@ static void mcimx7d_sabre_machine_init(MachineClass *mc)
>> mc->max_cpus = FSL_IMX7_NUM_CPUS;
>> mc->default_ram_id = "mcimx7d-sabre.ram";
>> mc->auto_create_sdcard = true;
>> + mc->ignore_memory_transaction_failures = true;
>> }
>
>This flag is in QEMU strictly to allow us to preserve
>backwards compatibility of behaviour with the older QEMU
>versions which pre-dated the suport of transaction failures
>(and where therefore we expected that if we just enabled it
>globally that we'd break previously-working user setups
>because we were missing some devices). It should not be
>newly set for any board model that isn't already setting it
>(and in an ideal world we would eliminate it from all those
>old board models, except that's a lot of work for mostly
>ancient hardly used machine types).
>
>Instead the fix here is to identify which devices we need
>to provide as at least unimplemented-device stubs, and add
>them to the board/SoC.
Okay, then let's drop this patch from the series. The functional test should
still succeed.
Best regards,
Bernhard
>
>thanks
>-- PMM
