Re: [VERY RFC PATCH 2/2] hurd: Make it possible to call memcpy very early

[H.J. Lu]

On Thu, Apr 20, 2023 at 11:43 AM Sergey Bugaev <bugaevc@gmail.com> wrote:
>
> Normally, in static builds, the first code that runs is _start, in e.g.
> sysdeps/x86_64/start.S, which quickly calls __libc_start_main, passing
> it the argv etc. Among the first things __libc_start_main does is
> initializing the tunables (based on env), then CPU features, and then
> calls _dl_relocate_static_pie (). Specifically, this runs ifunc
> resolvers to pick, based on the CPU features discovered earlier, the
> most suitable implementation of "string" functions such as memcpy.
>
> Before that point, calling memcpy (or other ifunc-resolved functions)
> will not work.
>
> In the Hurd port, things are more complex. In order to get argv/env for
> our process, glibc normally needs to do an RPC to the exec server,
> unless our args/env are already located on the stack (which is what
> happens to bootstrap processes spawned by GNU Mach). Fetching our
> argv/env from the exec server has to be done before the call to
> __libc_start_main, since we need to know what our argv/env are to pass
> them to __libc_start_main.
>
> On the other hand, the implementation of the RPC (and other initial
> setup needed on the Hurd before __libc_start_main can be run) is not
> very trivial. In particular, it may (and on x86_64, will) use memcpy.
> But as described above, calling memcpy before __libc_start_main can not
> work, since the GOT entry for it is not yet initialized at that point.
>
> Work around this by pre-filling the GOT entry with the baseline version
> of memcpy, __memcpy_sse2_unaligned. This makes it possible for early
> calls to memcpy to just work. Once _dl_relocate_static_pie () is called,
> the baseline version will get replaced with the most suitable one, and
> that's what subsequent calls of memcpy are going to call.
>
> Also, apply the same treatment to __stpncpy, which can also be used by
> the RPCs (see mig_strncpy.c), and is an ifunc-resolved function on both
> x86_64 and i386.
>
> Tested on x86_64-gnu (!).
>
> Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
> ---
>
> Please tell me:
>
> * if the approach is at all sane
> * if there's a better way to do this without hardcoding
>   "__memcpy_sse2_unaligned"
> * are the GOT entries for indirect functions supposed to be statically
>   initialized to anything (in the binary)? if yes, why? if not, why is
>   PROGBITS and not NOBITS?
> * am I doing all this _GLOBAL_OFFSET_TABLE_, @GOT, @GOTOFF, @GOTPCREL
>   correctly?
> * should there be a !PIC version as well? does the GOT exist under
>   !PIC (to access indirect functions), and if it does then how do I
>   access it? it would seem gcc just generates a direct $function even
>   for indirect functions in this case.
>
>  sysdeps/mach/hurd/i386/static-start.S   | 7 +++++++
>  sysdeps/mach/hurd/x86_64/static-start.S | 8 ++++++++
>  2 files changed, 15 insertions(+)
>
> diff --git a/sysdeps/mach/hurd/i386/static-start.S 
> b/sysdeps/mach/hurd/i386/static-start.S
> index c5d12645..1b1ae559 100644
> --- a/sysdeps/mach/hurd/i386/static-start.S
> +++ b/sysdeps/mach/hurd/i386/static-start.S
> @@ -19,6 +19,13 @@
>         .text
>         .globl _start
>  _start:
> +#ifdef PIC
> +       call __x86.get_pc_thunk.bx
> +       addl $_GLOBAL_OFFSET_TABLE_, %ebx
> +       leal __stpncpy_ia32@GOTOFF(%ebx), %eax
> +       movl %eax, __stpncpy@GOT(%ebx)
> +#endif
> +
>         call _hurd_stack_setup
>         xorl %edx, %edx
>         jmp _start1
> diff --git a/sysdeps/mach/hurd/x86_64/static-start.S 
> b/sysdeps/mach/hurd/x86_64/static-start.S
> index 982d3d52..81b3c0ac 100644
> --- a/sysdeps/mach/hurd/x86_64/static-start.S
> +++ b/sysdeps/mach/hurd/x86_64/static-start.S
> @@ -19,6 +19,14 @@
>         .text
>         .globl _start
>  _start:
> +
> +#ifdef PIC
> +       leaq __memcpy_sse2_unaligned(%rip), %rax
> +       movq %rax, memcpy@GOTPCREL(%rip)
> +       leaq __stpncpy_sse2_unaligned(%rip), %rax
> +       movq %rax, __stpncpy@GOTPCREL(%rip)
> +#endif
> +
>         call _hurd_stack_setup
>         xorq %rdx, %rdx
>         jmp _start1
> --
> 2.40.0
>

Doesn't it disable IFUNC for memcpy and stpncpy?

-- 
H.J.