On 08/03/2017 10:40 AM, Philippe Mathieu-Daudé wrote: > On 07/26/2017 05:02 AM, Alex Bennée wrote: >> >> Philippe Mathieu-Daudé <f4...@amsat.org> writes: >> >>> Hi Peter, >>> >>> I think this patch belongs to 2.10, as there is no time to fix TCI + >>> MTTCG. Should I RESEND it alone with "for 2.10" subject? >>> >>> One other option might be disable TCI if MTTCG enabled, but there is >>> no ./configure option for MTTCG while there is for TCI. >>> >>> Regards, >>> >>> Phil. >>> >>> On 06/28/2017 10:02 PM, Philippe Mathieu-Daudé wrote: >>>> TCI + MTTCG cause strange errors... >>>> >>>> $ arm-softmmu/qemu-system-arm -machine raspi2 -cpu cortex-a7 -smp 4 >>>> -accel tcg,thread=multi -kernel kernel7.img >>>> qemu-system-arm: Guest expects a stronger memory ordering than the host >>>> provides >>>> This may cause strange/hard to debug errors >>>> Segmentation fault (core dumped) >> >> So this isn't TCI enabling MTTCG by accident - this is the user forcing >> it when it wouldn't otherwise be enabled. Hence the scary warning... > > Hmmm this is the user enabling TCI, I'm not sure about "forcing it". > Now MTTCG is enforced by the ./configure and there is no option to disable it.
Er, no. If you write "-accel tcg" you are not forcing mttcg on. If you write "-accel tcg,thread=multi" you are forcing it on. If you write "-accel tcg,thread=single" you are forcing it off. MTTCG is only enabled by default if TCG_TARGET_DEFAULT_MO is defined, and in the case of TCI, it is not. r~
