Hi, Here are a few more fixes for regressions caused by the MTTCG merge. There is still one regression I'm aware of left to fix (record/replay breakage) but I thought it would be worth getting these posted now for review.
Another fix for graphic artefacts has already been merged via Gerd's graphics tree. The first fix is really just for completeness. It wasn't spotted in the original work and while I'm not aware of a regression attached to it I fixed it while investigating the other bits. The user-exec fix solves the looping assert Laurent found with running LTP tests. Finally the bsd-user fix is a compile tested only fix to ensure mmap_lock and friends do actually work. I wasn't able to complete building on my set-up due to an unrelated optionrom problem but I'm confident the MTTCG regression is addressed. Regards, Alex. Alex Bennée (3): cputlb: ensure tbl_set_dirty1 updates addr_write atomically user-exec: handle synchronous signals from QEMU gracefully bsd-user: align use of mmap_lock to that of linux-user bsd-user/mmap.c | 13 +------------ bsd-user/qemu.h | 2 -- cputlb.c | 8 ++++++++ user-exec.c | 18 +++++++++++++++--- 4 files changed, 24 insertions(+), 17 deletions(-) -- 2.11.0
