Hi Helge,
please drop this patch -- it is not needed on master, and I should have
checked that before posting.
The abort reproduces on 10.2.1, but not on master: since
9097087147 ("linux-user: make syscall emulation interruptible")
a #GP raised while a syscall is being emulated no longer longjmps out of
cpu_exec(), so a frame with zeroed CS/SS already ends up as SIGSEGV in the
guest and qemu survives. I checked this on a clean master build
(8967b2fe1b): the same frame exits 0 with the SIGSEGV handler running,
while a v10.2.1 build aborts with the cpu == current_cpu assertion.
The same holds for the v1 of this series -- please disregard both.
So the fix for 10.2.x is a backport of 9097087147, not this patch, and for
master there is nothing to fix.
Sorry for the noise, and thanks for catching the name.
Takano
