x86: Copy the FPU/XSAVE state from the creating thread to new threads.

[jhb]

Summary

POSIX states that new threads created via pthread_create() should
inherit the "floating point environment" from the creating thread.

[jhb]

I tested this on amd64 using kib@'s test program from D29024.

[kib]

sys/i386/i386/vm_machdep.c

434

Why not call npxexit() there, same as for amd64?

466

This should be fine but unrelated?

[jhb]

sys/i386/i386/vm_machdep.c

434

Why not call npxexit() there, same as for amd64?

This matches cpu_fork() for i386 which does not use npxexit(). If we'd rather fix i386 to use npxexit() can I throw that into this series.

466

This should be fine but unrelated?

It's just part of the 'update pcb before copy logic' which now matches cpu_fork(). Since I had to add the FPU save, I figured this should mirror cpu_fork() at the same time.