On Friday 11 September 2009, Louis Rilling wrote: > On 11/09/09 12:31 +0200, Arnd Bergmann wrote: > > On Thursday 10 September 2009, Sukadev Bhattiprolu wrote: > > > Since this is a variant of clone() and clone is listed as a PTREGSCALL(), > > > I pass in the pt_regs. > > > > > > arch/x86/kernel/entry_32.S lists clone() under this comment: > > > > > > /* > > > * System calls that need a pt_regs pointer. > > > */ > > > > > > Is there a guideline on what system calls use/need pt_regs ? > > > > You need pt_regs if you access any registers from the user task > > other than the argument registers. In case of clone(), this is > > the user stack pointer. > > AFAICS clone() actually needs all registers (see x86 copy_thread() for > instance). Any variant of clone() will have this requirement. Right, but that part (do_fork) is already handled in common code, which function calls into architecture specific code. The point I was making is that the sys_clone() and sys_clone_with_pids() functions themselves only access the argument registers, the stack pointer and pass those down to do_fork, along with pt_regs. Arnd <>< _______________________________________________ Containers mailing list Containers@xxxxxxxxxxxxxxxxxxxxxxxxxx https://lists.linux-foundation.org/mailman/listinfo/containers
