Unsure the scope of this bug or the proper fix, so bringing it up here
gdbstub.c:gdb_read_register will return 0, and thus E14, when a remote gdb
tries to call a function in the exposed linux kernel. This appears to be
because the caller expects to be able to receive a generic register size by
calling one plus the number of known registers. However,
x86_cpu_gdb_read_register() in target/i386/gdbstub.c will only provide
register sizes for known registers, since they're not always the same.
I'm dealing with this just by shimming 8 whenever gdb_read_register returns
0, but that's obviously not correct. Suggestions?
P.S. call still won't work unless linux is launched with noexec=off
noexec32=off, due to NX.