[PATCH] arm64: Better native ptrace support for compat tasks

[will.deacon@arm.com (Will Deacon)]

On Fri, Jul 17, 2015 at 11:49:54AM +0100, Catalin Marinas wrote:
> On Fri, Jul 17, 2015 at 11:26:29AM +0100, Will Deacon wrote:
> > On Tue, Jul 14, 2015 at 04:20:17PM +0100, Catalin Marinas wrote:
> > > The compat ptrace interface allows access to the TLS register, hardware
> > > breakpoints and watchpoints, syscall number. However, a native task
> > > using the native ptrace interface to debug compat tasks (e.g. multi-arch
> > > gdb) only has access to the general and VFP register sets. The compat
> > > ptrace interface cannot be accessed from a native task.
> > > 
> > > This patch adds a new user_aarch32_ptrace_view which contains the TLS,
> > > hardware breakpoint/watchpoint and syscall number regsets in addition to
> > > the existing GPR and VFP regsets. This view is backwards compatible with
> > > the previous kernels. Core dumping of 32-bit tasks and compat ptrace are
> > > not affected since the original user_aarch32_view is preserved.
> > 
> > [...]
> > 
> > > +static const struct user_regset aarch32_ptrace_regsets[] = {
> > > +	[REGSET_GPR] = {
> > > +		.core_note_type = NT_PRSTATUS,
> > > +		.n = COMPAT_ELF_NGREG,
> > > +		.size = sizeof(compat_elf_greg_t),
> > > +		.align = sizeof(compat_elf_greg_t),
> > > +		.get = compat_gpr_get,
> > > +		.set = compat_gpr_set
> > > +	},
> > > +	[REGSET_FPR] = {
> > > +		.core_note_type = NT_ARM_VFP,
> > > +		.n = VFP_STATE_SIZE / sizeof(compat_ulong_t),
> > > +		.size = sizeof(compat_ulong_t),
> > > +		.align = sizeof(compat_ulong_t),
> > > +		.get = compat_vfp_get,
> > > +		.set = compat_vfp_set
> > 
> > 
> > I don't see how this is "backwards compatible with the previous kernels".
> 
> The first two entries of aarch32_ptrace_regsets are the same as the two
> aarch32_regsets entries currently in use before this patch (I just
> dropped "COMPAT" from REGSET_* for consistency with the rest of the
> array).
> 
> > If there is some userspace out there that expects a native view onto the
> > registers of a compat task, then that's no longer the case, right? For
> > example, NT_PRFPREG requests will no longer work afaict.
> 
> But this never worked since task_user_regset_view() always returned
> user_aarch32_view (with the aarch32_regsets) if the ptrace'd task was
> compat. So a native gdb debugging compat task is expected to use
> NT_ARM_VFP before and after this patch.

Aha, sorry, I got confused because you've wired up the native hw-breakpoint
interfaces later on in the regset (which is fine, because I don't think
the parent could've got into the compat hw-breakpoint ptrace requests
since they're not regset-based).

So I think this looks fine, thanks!

Will