This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH] x86-64: Properly align La_x86_64_retval to VEC_SIZE [BZ #22715]
- From: "H.J. Lu" <hjl dot tools at gmail dot com>
- To: GNU C Library <libc-alpha at sourceware dot org>
- Date: Wed, 17 Jan 2018 04:00:35 -0800
- Subject: Re: [PATCH] x86-64: Properly align La_x86_64_retval to VEC_SIZE [BZ #22715]
- Authentication-results: sourceware.org; auth=none
- References: <20180116041818.GA235600@intel.com>
On Mon, Jan 15, 2018 at 8:18 PM, H.J. Lu <hongjiu.lu@intel.com> wrote:
> _dl_runtime_profile calls _dl_call_pltexit, passing a pointer to
> La_x86_64_retval which is allocated on stack. The lrv_vector0
> field in La_x86_64_retval must be aligned to size of vector register.
> When allocating stack space for La_x86_64_retval, we need to make sure
> that the address of La_x86_64_retval + RV_VECTOR0_OFFSET is aligned to
> VEC_SIZE. This patch checks the alignment of the lrv_vector0 field
> and pads the stack space if needed.
>
> Tested with x32 and x86-64 on SSE4, AVX and AVX512 machines.
>
> OK for master?
>
> H.J.
> ---
> [BZ #22715]
> * sysdeps/x86_64/dl-trampoline.h (_dl_runtime_profile): Properly
> align La_x86_64_retval to VEC_SIZE.
> ---
> sysdeps/x86_64/dl-trampoline.h | 12 ++++++++++--
> 1 file changed, 10 insertions(+), 2 deletions(-)
>
> diff --git a/sysdeps/x86_64/dl-trampoline.h b/sysdeps/x86_64/dl-trampoline.h
> index 15edf49674..298cfb3d99 100644
> --- a/sysdeps/x86_64/dl-trampoline.h
> +++ b/sysdeps/x86_64/dl-trampoline.h
> @@ -440,8 +440,16 @@ _dl_runtime_profile:
> # ifdef RESTORE_AVX
> /* sizeof(La_x86_64_retval). Need extra space for 2 SSE
> registers to detect if xmm0/xmm1 registers are changed
> - by audit module. */
> - sub $(LRV_SIZE + XMM_SIZE*2), %RSP_LP
> + by audit module. Since rsp is aligned to VEC_SIZE, we
> + need to make sure that the address of La_x86_64_retval +
> + LRV_VECTOR0_OFFSET is aligned to VEC_SIZE. */
> +# define LRV_SPACE (LRV_SIZE + XMM_SIZE*2)
> +# define LRV_MISALIGNED ((LRV_SIZE + LRV_VECTOR0_OFFSET) & (VEC_SIZE - 1))
> +# if LRV_MISALIGNED == 0
> + sub $LRV_SPACE, %RSP_LP
> +# else
> + sub $(LRV_SPACE + VEC_SIZE - LRV_MISALIGNED), %RSP_LP
> +# endif
> # else
> sub $LRV_SIZE, %RSP_LP # sizeof(La_x86_64_retval)
> # endif
> --
> 2.14.3
>
I am checking it in.
--
H.J.