Fix strcpy_chk and stpcpy_chk performance.

  Hi, as I wrote in previous patches a performance of checked strcpy and
stpcpy is terrible as these don't use sse2 and are around four times
slower that strcpy and stpcpy now.

As this bug shows that these functions are not performance sensitive I
decided just to improve generic implementation instead for easier
maintainance.

I could improve performance more if we allow to write past bound and
then fail as we could just call stpcpy and check if it exceed bounds or
not.

Other arches could also have bad performance so sobebody should finally
read benchtests there to see problems.

	* debug/strcpy_chk.c: Improve performance.
	* debug/stpcpy_chk.c: Likewise.
	* sysdeps/x86_64/strcpy_chk.S: Remove.
	* sysdeps/x86_64/stpcpy_chk.S: Remove.
  

This seems reasonable to me, of course I can't approve it.

It would not be OK for these functions to write beyond the bounds of
the destination even if they trap afterward, that would make debugging
significantly more difficult.

On Wed, Aug 12, 2015 at 2:42 AM, Ondřej Bílka <neleai@seznam.cz> wrote:
> Hi, as I wrote in previous patches a performance of checked strcpy and
> stpcpy is terrible as these don't use sse2 and are around four times
> slower that strcpy and stpcpy now.
>
> As this bug shows that these functions are not performance sensitive I
> decided just to improve generic implementation instead for easier
> maintainance.
>
> I could improve performance more if we allow to write past bound and
> then fail as we could just call stpcpy and check if it exceed bounds or
> not.
>
> Other arches could also have bad performance so sobebody should finally
> read benchtests there to see problems.
>
>         * debug/strcpy_chk.c: Improve performance.
>         * debug/stpcpy_chk.c: Likewise.
>         * sysdeps/x86_64/strcpy_chk.S: Remove.
>         * sysdeps/x86_64/stpcpy_chk.S: Remove.
>
> diff --git a/debug/stpcpy_chk.c b/debug/stpcpy_chk.c
> index 91c5031..69476a2 100644
> --- a/debug/stpcpy_chk.c
> +++ b/debug/stpcpy_chk.c
> @@ -29,16 +29,9 @@ __stpcpy_chk (dest, src, destlen)
>       const char *src;
>       size_t destlen;
>  {
> -  char *d = dest;
> -  const char *s = src;
> -
> -  do
> -    {
> -      if (__glibc_unlikely (destlen-- == 0))
> -       __chk_fail ();
> -      *d++ = *s;
> -    }
> -  while (*s++ != '\0');
> -
> -  return d - 1;
> +  size_t len = strlen (src);
> +  if (len > destlen)
> +    __chk_fail ();
> +
> +  return memcpy (dest, src, len + 1) + len;
>  }
> diff --git a/debug/strcpy_chk.c b/debug/strcpy_chk.c
> index 91bf0dd..92eb69d2 100644
> --- a/debug/strcpy_chk.c
> +++ b/debug/strcpy_chk.c
> @@ -28,40 +28,9 @@ __strcpy_chk (dest, src, destlen)
>       const char *src;
>       size_t destlen;
>  {
> -  char c;
> -  char *s = (char *) src;
> -  const ptrdiff_t off = dest - s;
> -
> -  while (__builtin_expect (destlen >= 4, 0))
> -    {
> -      c = s[0];
> -      s[off] = c;
> -      if (c == '\0')
> -        return dest;
> -      c = s[1];
> -      s[off + 1] = c;
> -      if (c == '\0')
> -        return dest;
> -      c = s[2];
> -      s[off + 2] = c;
> -      if (c == '\0')
> -        return dest;
> -      c = s[3];
> -      s[off + 3] = c;
> -      if (c == '\0')
> -        return dest;
> -      destlen -= 4;
> -      s += 4;
> -    }
> -
> -  do
> -    {
> -      if (__glibc_unlikely (destlen-- == 0))
> -        __chk_fail ();
> -      c = *s;
> -      *(s++ + off) = c;
> -    }
> -  while (c != '\0');
> -
> -  return dest;
> +  size_t len = strlen (src);
> +  if (len > destlen)
> +    __chk_fail ();
> +
> +  return memcpy (dest, src, len + 1);
>  }
> diff --git a/sysdeps/x86_64/stpcpy_chk.S b/sysdeps/x86_64/stpcpy_chk.S
> deleted file mode 100644
> index 905e8d7..0000000
> --- a/sysdeps/x86_64/stpcpy_chk.S
> +++ /dev/null
> @@ -1,3 +0,0 @@
> -#define USE_AS_STPCPY_CHK
> -#define STRCPY_CHK __stpcpy_chk
> -#include <sysdeps/x86_64/strcpy_chk.S>
> diff --git a/sysdeps/x86_64/strcpy_chk.S b/sysdeps/x86_64/strcpy_chk.S
> deleted file mode 100644
> index 24e51c6..0000000
> --- a/sysdeps/x86_64/strcpy_chk.S
> +++ /dev/null
> @@ -1,208 +0,0 @@
> -/* strcpy/stpcpy checking implementation for x86-64.
> -   Copyright (C) 2002-2015 Free Software Foundation, Inc.
> -   This file is part of the GNU C Library.
> -   Contributed by Andreas Jaeger <aj@suse.de>, 2002.
> -   Adopted into checking version by Jakub Jelinek <jakub@redhat.com>.
> -
> -   The GNU C Library is free software; you can redistribute it and/or
> -   modify it under the terms of the GNU Lesser General Public
> -   License as published by the Free Software Foundation; either
> -   version 2.1 of the License, or (at your option) any later version.
> -
> -   The GNU C Library is distributed in the hope that it will be useful,
> -   but WITHOUT ANY WARRANTY; without even the implied warranty of
> -   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
> -   Lesser General Public License for more details.
> -
> -   You should have received a copy of the GNU Lesser General Public
> -   License along with the GNU C Library; if not, see
> -   <http://www.gnu.org/licenses/>.  */
> -
> -#include <sysdep.h>
> -#include "asm-syntax.h"
> -
> -#ifndef USE_AS_STPCPY_CHK
> -# define STRCPY_CHK __strcpy_chk
> -#endif
> -
> -       .text
> -ENTRY (STRCPY_CHK)
> -       movq    %rsi, %rcx      /* Source register. */
> -       andl    $7, %ecx        /* mask alignment bits */
> -#ifndef USE_AS_STPCPY_CHK
> -       movq    %rdi, %r10      /* Duplicate destination pointer.  */
> -#endif
> -       jz 5f                   /* aligned => start loop */
> -
> -       cmpq    $8, %rdx        /* Check if only few bytes left in
> -                                  destination.  */
> -       jb      50f
> -
> -       subq    $8, %rcx        /* We need to align to 8 bytes.  */
> -       addq    %rcx, %rdx      /* Subtract count of stored bytes
> -                                  in the cycle below from destlen.  */
> -
> -       /* Search the first bytes directly.  */
> -0:
> -       movb    (%rsi), %al     /* Fetch a byte */
> -       testb   %al, %al        /* Is it NUL? */
> -       movb    %al, (%rdi)     /* Store it */
> -       jz      4f              /* If it was NUL, done! */
> -       incq    %rsi
> -       incq    %rdi
> -       incl    %ecx
> -       jnz     0b
> -
> -5:
> -       movq $0xfefefefefefefeff,%r8
> -       cmpq    $32, %rdx       /* Are there enough bytes in destination
> -                                  for the next unrolled round?  */
> -       jb      60f             /* If not, avoid the unrolled loop.  */
> -
> -       /* Now the sources is aligned.  Unfortunatly we cannot force
> -          to have both source and destination aligned, so ignore the
> -          alignment of the destination.  */
> -       .p2align 4
> -1:
> -       /* 1st unroll.  */
> -       movq    (%rsi), %rax    /* Read double word (8 bytes).  */
> -       addq    $8, %rsi        /* Adjust pointer for next word.  */
> -       movq    %rax, %r9       /* Save a copy for NUL finding.  */
> -       addq    %r8, %r9        /* add the magic value to the word.  We get
> -                                  carry bits reported for each byte which
> -                                  is *not* 0 */
> -       jnc     3f              /* highest byte is NUL => return pointer */
> -       xorq    %rax, %r9       /* (word+magic)^word */
> -       orq     %r8, %r9        /* set all non-carry bits */
> -       incq    %r9             /* add 1: if one carry bit was *not* set
> -                                  the addition will not result in 0.  */
> -
> -       jnz     3f              /* found NUL => return pointer */
> -
> -       movq    %rax, (%rdi)    /* Write value to destination.  */
> -       addq    $8, %rdi        /* Adjust pointer.  */
> -
> -       /* 2nd unroll.  */
> -       movq    (%rsi), %rax    /* Read double word (8 bytes).  */
> -       addq    $8, %rsi        /* Adjust pointer for next word.  */
> -       movq    %rax, %r9       /* Save a copy for NUL finding.  */
> -       addq    %r8, %r9        /* add the magic value to the word.  We get
> -                                  carry bits reported for each byte which
> -                                  is *not* 0 */
> -       jnc     3f              /* highest byte is NUL => return pointer */
> -       xorq    %rax, %r9       /* (word+magic)^word */
> -       orq     %r8, %r9        /* set all non-carry bits */
> -       incq    %r9             /* add 1: if one carry bit was *not* set
> -                                  the addition will not result in 0.  */
> -
> -       jnz     3f              /* found NUL => return pointer */
> -
> -       movq    %rax, (%rdi)    /* Write value to destination.  */
> -       addq    $8, %rdi        /* Adjust pointer.  */
> -
> -       /* 3rd unroll.  */
> -       movq    (%rsi), %rax    /* Read double word (8 bytes).  */
> -       addq    $8, %rsi        /* Adjust pointer for next word.  */
> -       movq    %rax, %r9       /* Save a copy for NUL finding.  */
> -       addq    %r8, %r9        /* add the magic value to the word.  We get
> -                                  carry bits reported for each byte which
> -                                  is *not* 0 */
> -       jnc     3f              /* highest byte is NUL => return pointer */
> -       xorq    %rax, %r9       /* (word+magic)^word */
> -       orq     %r8, %r9        /* set all non-carry bits */
> -       incq    %r9             /* add 1: if one carry bit was *not* set
> -                                  the addition will not result in 0.  */
> -
> -       jnz     3f              /* found NUL => return pointer */
> -
> -       movq    %rax, (%rdi)    /* Write value to destination.  */
> -       addq    $8, %rdi        /* Adjust pointer.  */
> -
> -       /* 4th unroll.  */
> -       movq    (%rsi), %rax    /* Read double word (8 bytes).  */
> -       addq    $8, %rsi        /* Adjust pointer for next word.  */
> -       movq    %rax, %r9       /* Save a copy for NUL finding.  */
> -       addq    %r8, %r9        /* add the magic value to the word.  We get
> -                                  carry bits reported for each byte which
> -                                  is *not* 0 */
> -       jnc     3f              /* highest byte is NUL => return pointer */
> -       xorq    %rax, %r9       /* (word+magic)^word */
> -       orq     %r8, %r9        /* set all non-carry bits */
> -       incq    %r9             /* add 1: if one carry bit was *not* set
> -                                  the addition will not result in 0.  */
> -
> -       jnz     3f              /* found NUL => return pointer */
> -
> -       subq    $32, %rdx       /* Adjust destlen.  */
> -       movq    %rax, (%rdi)    /* Write value to destination.  */
> -       addq    $8, %rdi        /* Adjust pointer.  */
> -       cmpq    $32, %rdx       /* Are there enough bytes in destination
> -                                  for the next unrolled round?  */
> -       jae     1b              /* Next iteration.  */
> -
> -60:
> -       cmpq    $8, %rdx        /* Are there enough bytes in destination
> -                                  for the next unrolled round?  */
> -       jb      50f             /* Now, copy and check byte by byte.  */
> -
> -       movq    (%rsi), %rax    /* Read double word (8 bytes).  */
> -       addq    $8, %rsi        /* Adjust pointer for next word.  */
> -       movq    %rax, %r9       /* Save a copy for NUL finding.  */
> -       addq    %r8, %r9        /* add the magic value to the word.  We get
> -                                  carry bits reported for each byte which
> -                                  is *not* 0 */
> -       jnc     3f              /* highest byte is NUL => return pointer */
> -       xorq    %rax, %r9       /* (word+magic)^word */
> -       orq     %r8, %r9        /* set all non-carry bits */
> -       incq    %r9             /* add 1: if one carry bit was *not* set
> -                                  the addition will not result in 0.  */
> -
> -       jnz     3f              /* found NUL => return pointer */
> -
> -       subq    $8, %rdx        /* Adjust destlen.  */
> -       movq    %rax, (%rdi)    /* Write value to destination.  */
> -       addq    $8, %rdi        /* Adjust pointer.  */
> -       jmp     60b             /* Next iteration.  */
> -
> -       /* Do the last few bytes. %rax contains the value to write.
> -          The loop is unrolled twice.  */
> -       .p2align 4
> -3:
> -       /* Note that stpcpy needs to return with the value of the NUL
> -          byte.  */
> -       movb    %al, (%rdi)     /* 1st byte.  */
> -       testb   %al, %al        /* Is it NUL.  */
> -       jz      4f              /* yes, finish.  */
> -       incq    %rdi            /* Increment destination.  */
> -       movb    %ah, (%rdi)     /* 2nd byte.  */
> -       testb   %ah, %ah        /* Is it NUL?.  */
> -       jz      4f              /* yes, finish.  */
> -       incq    %rdi            /* Increment destination.  */
> -       shrq    $16, %rax       /* Shift...  */
> -       jmp     3b              /* and look at next two bytes in %rax.  */
> -
> -51:
> -       /* Search the bytes directly, checking for overflows.  */
> -       incq    %rsi
> -       incq    %rdi
> -       decq    %rdx
> -       jz      HIDDEN_JUMPTARGET (__chk_fail)
> -52:
> -       movb    (%rsi), %al     /* Fetch a byte */
> -       testb   %al, %al        /* Is it NUL? */
> -       movb    %al, (%rdi)     /* Store it */
> -       jnz     51b             /* If it was NUL, done! */
> -4:
> -#ifdef USE_AS_STPCPY_CHK
> -       movq    %rdi, %rax      /* Destination is return value.  */
> -#else
> -       movq    %r10, %rax      /* Source is return value.  */
> -#endif
> -       retq
> -
> -50:
> -       testq   %rdx, %rdx
> -       jnz     52b
> -       jmp     HIDDEN_JUMPTARGET (__chk_fail)
> -
> -END (STRCPY_CHK)
>
  

ping
On Wed, Aug 12, 2015 at 08:42:13AM +0200, Ondřej Bílka wrote:
> Hi, as I wrote in previous patches a performance of checked strcpy and
> stpcpy is terrible as these don't use sse2 and are around four times
> slower that strcpy and stpcpy now.
> 
> As this bug shows that these functions are not performance sensitive I
> decided just to improve generic implementation instead for easier
> maintainance.
> 
> I could improve performance more if we allow to write past bound and
> then fail as we could just call stpcpy and check if it exceed bounds or
> not.
> 
> Other arches could also have bad performance so sobebody should finally
> read benchtests there to see problems.
> 
> 	* debug/strcpy_chk.c: Improve performance.
> 	* debug/stpcpy_chk.c: Likewise.
> 	* sysdeps/x86_64/strcpy_chk.S: Remove.
> 	* sysdeps/x86_64/stpcpy_chk.S: Remove.
> 
> diff --git a/debug/stpcpy_chk.c b/debug/stpcpy_chk.c
> index 91c5031..69476a2 100644
> --- a/debug/stpcpy_chk.c
> +++ b/debug/stpcpy_chk.c
> @@ -29,16 +29,9 @@ __stpcpy_chk (dest, src, destlen)
>       const char *src;
>       size_t destlen;
>  {
> -  char *d = dest;
> -  const char *s = src;
> -
> -  do
> -    {
> -      if (__glibc_unlikely (destlen-- == 0))
> -	__chk_fail ();
> -      *d++ = *s;
> -    }
> -  while (*s++ != '\0');
> -
> -  return d - 1;
> +  size_t len = strlen (src);
> +  if (len > destlen)
> +    __chk_fail ();
> +    
> +  return memcpy (dest, src, len + 1) + len;
>  }
> diff --git a/debug/strcpy_chk.c b/debug/strcpy_chk.c
> index 91bf0dd..92eb69d2 100644
> --- a/debug/strcpy_chk.c
> +++ b/debug/strcpy_chk.c
> @@ -28,40 +28,9 @@ __strcpy_chk (dest, src, destlen)
>       const char *src;
>       size_t destlen;
>  {
> -  char c;
> -  char *s = (char *) src;
> -  const ptrdiff_t off = dest - s;
> -
> -  while (__builtin_expect (destlen >= 4, 0))
> -    {
> -      c = s[0];
> -      s[off] = c;
> -      if (c == '\0')
> -        return dest;
> -      c = s[1];
> -      s[off + 1] = c;
> -      if (c == '\0')
> -        return dest;
> -      c = s[2];
> -      s[off + 2] = c;
> -      if (c == '\0')
> -        return dest;
> -      c = s[3];
> -      s[off + 3] = c;
> -      if (c == '\0')
> -        return dest;
> -      destlen -= 4;
> -      s += 4;
> -    }
> -
> -  do
> -    {
> -      if (__glibc_unlikely (destlen-- == 0))
> -        __chk_fail ();
> -      c = *s;
> -      *(s++ + off) = c;
> -    }
> -  while (c != '\0');
> -
> -  return dest;
> +  size_t len = strlen (src);
> +  if (len > destlen)
> +    __chk_fail ();
> +    
> +  return memcpy (dest, src, len + 1);
>  }
> diff --git a/sysdeps/x86_64/stpcpy_chk.S b/sysdeps/x86_64/stpcpy_chk.S
> deleted file mode 100644
> index 905e8d7..0000000
> --- a/sysdeps/x86_64/stpcpy_chk.S
> +++ /dev/null
> @@ -1,3 +0,0 @@
> -#define USE_AS_STPCPY_CHK
> -#define STRCPY_CHK __stpcpy_chk
> -#include <sysdeps/x86_64/strcpy_chk.S>
> diff --git a/sysdeps/x86_64/strcpy_chk.S b/sysdeps/x86_64/strcpy_chk.S
> deleted file mode 100644
> index 24e51c6..0000000
> --- a/sysdeps/x86_64/strcpy_chk.S
> +++ /dev/null
> @@ -1,208 +0,0 @@
> -/* strcpy/stpcpy checking implementation for x86-64.
> -   Copyright (C) 2002-2015 Free Software Foundation, Inc.
> -   This file is part of the GNU C Library.
> -   Contributed by Andreas Jaeger <aj@suse.de>, 2002.
> -   Adopted into checking version by Jakub Jelinek <jakub@redhat.com>.
> -
> -   The GNU C Library is free software; you can redistribute it and/or
> -   modify it under the terms of the GNU Lesser General Public
> -   License as published by the Free Software Foundation; either
> -   version 2.1 of the License, or (at your option) any later version.
> -
> -   The GNU C Library is distributed in the hope that it will be useful,
> -   but WITHOUT ANY WARRANTY; without even the implied warranty of
> -   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
> -   Lesser General Public License for more details.
> -
> -   You should have received a copy of the GNU Lesser General Public
> -   License along with the GNU C Library; if not, see
> -   <http://www.gnu.org/licenses/>.  */
> -
> -#include <sysdep.h>
> -#include "asm-syntax.h"
> -
> -#ifndef USE_AS_STPCPY_CHK
> -# define STRCPY_CHK __strcpy_chk
> -#endif
> -
> -	.text
> -ENTRY (STRCPY_CHK)
> -	movq	%rsi, %rcx	/* Source register. */
> -	andl	$7, %ecx	/* mask alignment bits */
> -#ifndef USE_AS_STPCPY_CHK
> -	movq	%rdi, %r10	/* Duplicate destination pointer.  */
> -#endif
> -	jz 5f			/* aligned => start loop */
> -
> -	cmpq	$8, %rdx	/* Check if only few bytes left in
> -				   destination.  */
> -	jb	50f
> -
> -	subq	$8, %rcx	/* We need to align to 8 bytes.  */
> -	addq	%rcx, %rdx	/* Subtract count of stored bytes
> -				   in the cycle below from destlen.  */
> -
> -	/* Search the first bytes directly.  */
> -0:
> -	movb	(%rsi), %al	/* Fetch a byte */
> -	testb	%al, %al	/* Is it NUL? */
> -	movb	%al, (%rdi)	/* Store it */
> -	jz	4f		/* If it was NUL, done! */
> -	incq	%rsi
> -	incq	%rdi
> -	incl	%ecx
> -	jnz	0b
> -
> -5:
> -	movq $0xfefefefefefefeff,%r8
> -	cmpq	$32, %rdx	/* Are there enough bytes in destination
> -				   for the next unrolled round?  */
> -	jb	60f		/* If not, avoid the unrolled loop.  */
> -
> -	/* Now the sources is aligned.  Unfortunatly we cannot force
> -	   to have both source and destination aligned, so ignore the
> -	   alignment of the destination.  */
> -	.p2align 4
> -1:
> -	/* 1st unroll.  */
> -	movq	(%rsi), %rax	/* Read double word (8 bytes).  */
> -	addq	$8, %rsi	/* Adjust pointer for next word.  */
> -	movq	%rax, %r9	/* Save a copy for NUL finding.  */
> -	addq	%r8, %r9	/* add the magic value to the word.  We get
> -				   carry bits reported for each byte which
> -				   is *not* 0 */
> -	jnc	3f		/* highest byte is NUL => return pointer */
> -	xorq	%rax, %r9	/* (word+magic)^word */
> -	orq	%r8, %r9	/* set all non-carry bits */
> -	incq	%r9		/* add 1: if one carry bit was *not* set
> -				   the addition will not result in 0.  */
> -
> -	jnz	3f		/* found NUL => return pointer */
> -
> -	movq	%rax, (%rdi)	/* Write value to destination.  */
> -	addq	$8, %rdi	/* Adjust pointer.  */
> -
> -	/* 2nd unroll.  */
> -	movq	(%rsi), %rax	/* Read double word (8 bytes).  */
> -	addq	$8, %rsi	/* Adjust pointer for next word.  */
> -	movq	%rax, %r9	/* Save a copy for NUL finding.  */
> -	addq	%r8, %r9	/* add the magic value to the word.  We get
> -				   carry bits reported for each byte which
> -				   is *not* 0 */
> -	jnc	3f		/* highest byte is NUL => return pointer */
> -	xorq	%rax, %r9	/* (word+magic)^word */
> -	orq	%r8, %r9	/* set all non-carry bits */
> -	incq	%r9		/* add 1: if one carry bit was *not* set
> -				   the addition will not result in 0.  */
> -
> -	jnz	3f		/* found NUL => return pointer */
> -
> -	movq	%rax, (%rdi)	/* Write value to destination.  */
> -	addq	$8, %rdi	/* Adjust pointer.  */
> -
> -	/* 3rd unroll.  */
> -	movq	(%rsi), %rax	/* Read double word (8 bytes).  */
> -	addq	$8, %rsi	/* Adjust pointer for next word.  */
> -	movq	%rax, %r9	/* Save a copy for NUL finding.  */
> -	addq	%r8, %r9	/* add the magic value to the word.  We get
> -				   carry bits reported for each byte which
> -				   is *not* 0 */
> -	jnc	3f		/* highest byte is NUL => return pointer */
> -	xorq	%rax, %r9	/* (word+magic)^word */
> -	orq	%r8, %r9	/* set all non-carry bits */
> -	incq	%r9		/* add 1: if one carry bit was *not* set
> -				   the addition will not result in 0.  */
> -
> -	jnz	3f		/* found NUL => return pointer */
> -
> -	movq	%rax, (%rdi)	/* Write value to destination.  */
> -	addq	$8, %rdi	/* Adjust pointer.  */
> -
> -	/* 4th unroll.  */
> -	movq	(%rsi), %rax	/* Read double word (8 bytes).  */
> -	addq	$8, %rsi	/* Adjust pointer for next word.  */
> -	movq	%rax, %r9	/* Save a copy for NUL finding.  */
> -	addq	%r8, %r9	/* add the magic value to the word.  We get
> -				   carry bits reported for each byte which
> -				   is *not* 0 */
> -	jnc	3f		/* highest byte is NUL => return pointer */
> -	xorq	%rax, %r9	/* (word+magic)^word */
> -	orq	%r8, %r9	/* set all non-carry bits */
> -	incq	%r9		/* add 1: if one carry bit was *not* set
> -				   the addition will not result in 0.  */
> -
> -	jnz	3f		/* found NUL => return pointer */
> -
> -	subq	$32, %rdx	/* Adjust destlen.  */
> -	movq	%rax, (%rdi)	/* Write value to destination.  */
> -	addq	$8, %rdi	/* Adjust pointer.  */
> -	cmpq	$32, %rdx	/* Are there enough bytes in destination
> -				   for the next unrolled round?  */
> -	jae	1b		/* Next iteration.  */
> -
> -60:
> -	cmpq	$8, %rdx	/* Are there enough bytes in destination
> -				   for the next unrolled round?  */
> -	jb	50f		/* Now, copy and check byte by byte.  */
> -
> -	movq	(%rsi), %rax	/* Read double word (8 bytes).  */
> -	addq	$8, %rsi	/* Adjust pointer for next word.  */
> -	movq	%rax, %r9	/* Save a copy for NUL finding.  */
> -	addq	%r8, %r9	/* add the magic value to the word.  We get
> -				   carry bits reported for each byte which
> -				   is *not* 0 */
> -	jnc	3f		/* highest byte is NUL => return pointer */
> -	xorq	%rax, %r9	/* (word+magic)^word */
> -	orq	%r8, %r9	/* set all non-carry bits */
> -	incq	%r9		/* add 1: if one carry bit was *not* set
> -				   the addition will not result in 0.  */
> -
> -	jnz	3f		/* found NUL => return pointer */
> -
> -	subq	$8, %rdx	/* Adjust destlen.  */
> -	movq	%rax, (%rdi)	/* Write value to destination.  */
> -	addq	$8, %rdi	/* Adjust pointer.  */
> -	jmp	60b		/* Next iteration.  */
> -
> -	/* Do the last few bytes. %rax contains the value to write.
> -	   The loop is unrolled twice.  */
> -	.p2align 4
> -3:
> -	/* Note that stpcpy needs to return with the value of the NUL
> -	   byte.  */
> -	movb	%al, (%rdi)	/* 1st byte.  */
> -	testb	%al, %al	/* Is it NUL.  */
> -	jz	4f		/* yes, finish.  */
> -	incq	%rdi		/* Increment destination.  */
> -	movb	%ah, (%rdi)	/* 2nd byte.  */
> -	testb	%ah, %ah	/* Is it NUL?.  */
> -	jz	4f		/* yes, finish.  */
> -	incq	%rdi		/* Increment destination.  */
> -	shrq	$16, %rax	/* Shift...  */
> -	jmp	3b		/* and look at next two bytes in %rax.  */
> -
> -51:
> -	/* Search the bytes directly, checking for overflows.  */
> -	incq	%rsi
> -	incq	%rdi
> -	decq	%rdx
> -	jz	HIDDEN_JUMPTARGET (__chk_fail)
> -52:
> -	movb	(%rsi), %al	/* Fetch a byte */
> -	testb	%al, %al	/* Is it NUL? */
> -	movb	%al, (%rdi)	/* Store it */
> -	jnz	51b		/* If it was NUL, done! */
> -4:
> -#ifdef USE_AS_STPCPY_CHK
> -	movq	%rdi, %rax	/* Destination is return value.  */
> -#else
> -	movq	%r10, %rax	/* Source is return value.  */
> -#endif
> -	retq
> -
> -50:
> -	testq	%rdx, %rdx
> -	jnz	52b
> -	jmp	HIDDEN_JUMPTARGET (__chk_fail)
> -
> -END (STRCPY_CHK)
  

LGTM with some nits below:

On 19-08-2015 03:15, Ondřej Bílka wrote:
> ping
> On Wed, Aug 12, 2015 at 08:42:13AM +0200, Ondřej Bílka wrote:
>> Hi, as I wrote in previous patches a performance of checked strcpy and
>> stpcpy is terrible as these don't use sse2 and are around four times
>> slower that strcpy and stpcpy now.
>>
>> As this bug shows that these functions are not performance sensitive I
>> decided just to improve generic implementation instead for easier
>> maintainance.
>>
>> I could improve performance more if we allow to write past bound and
>> then fail as we could just call stpcpy and check if it exceed bounds or
>> not.
>>
>> Other arches could also have bad performance so sobebody should finally
>> read benchtests there to see problems.
>>
>> 	* debug/strcpy_chk.c: Improve performance.
>> 	* debug/stpcpy_chk.c: Likewise.

I think you should add the (__stpcpy_chk) to indicate which function
you are refering.

>> 	* sysdeps/x86_64/strcpy_chk.S: Remove.
>> 	* sysdeps/x86_64/stpcpy_chk.S: Remove.
>>
>> diff --git a/debug/stpcpy_chk.c b/debug/stpcpy_chk.c
>> index 91c5031..69476a2 100644
>> --- a/debug/stpcpy_chk.c
>> +++ b/debug/stpcpy_chk.c
>> @@ -29,16 +29,9 @@ __stpcpy_chk (dest, src, destlen)
>>       const char *src;
>>       size_t destlen;
>>  {

Could you also get rid of the k&r function prototype (same for strcpy_chk) ?

>> -  char *d = dest;
>> -  const char *s = src;
>> -
>> -  do
>> -    {
>> -      if (__glibc_unlikely (destlen-- == 0))
>> -	__chk_fail ();
>> -      *d++ = *s;
>> -    }
>> -  while (*s++ != '\0');
>> -
>> -  return d - 1;
>> +  size_t len = strlen (src);
>> +  if (len > destlen)
>> +    __chk_fail ();
>> +    
>> +  return memcpy (dest, src, len + 1) + len;
>>  }
>> diff --git a/debug/strcpy_chk.c b/debug/strcpy_chk.c
>> index 91bf0dd..92eb69d2 100644
>> --- a/debug/strcpy_chk.c
>> +++ b/debug/strcpy_chk.c
>> @@ -28,40 +28,9 @@ __strcpy_chk (dest, src, destlen)
>>       const char *src;
>>       size_t destlen;
>>  {
>> -  char c;
>> -  char *s = (char *) src;
>> -  const ptrdiff_t off = dest - s;
>> -
>> -  while (__builtin_expect (destlen >= 4, 0))
>> -    {
>> -      c = s[0];
>> -      s[off] = c;
>> -      if (c == '\0')
>> -        return dest;
>> -      c = s[1];
>> -      s[off + 1] = c;
>> -      if (c == '\0')
>> -        return dest;
>> -      c = s[2];
>> -      s[off + 2] = c;
>> -      if (c == '\0')
>> -        return dest;
>> -      c = s[3];
>> -      s[off + 3] = c;
>> -      if (c == '\0')
>> -        return dest;
>> -      destlen -= 4;
>> -      s += 4;
>> -    }
>> -
>> -  do
>> -    {
>> -      if (__glibc_unlikely (destlen-- == 0))
>> -        __chk_fail ();
>> -      c = *s;
>> -      *(s++ + off) = c;
>> -    }
>> -  while (c != '\0');
>> -
>> -  return dest;
>> +  size_t len = strlen (src);
>> +  if (len > destlen)
>> +    __chk_fail ();
>> +    
>> +  return memcpy (dest, src, len + 1);
>>  }
>> diff --git a/sysdeps/x86_64/stpcpy_chk.S b/sysdeps/x86_64/stpcpy_chk.S
>> deleted file mode 100644
>> index 905e8d7..0000000
>> --- a/sysdeps/x86_64/stpcpy_chk.S
>> +++ /dev/null
>> @@ -1,3 +0,0 @@
>> -#define USE_AS_STPCPY_CHK
>> -#define STRCPY_CHK __stpcpy_chk
>> -#include <sysdeps/x86_64/strcpy_chk.S>
>> diff --git a/sysdeps/x86_64/strcpy_chk.S b/sysdeps/x86_64/strcpy_chk.S
>> deleted file mode 100644
>> index 24e51c6..0000000
>> --- a/sysdeps/x86_64/strcpy_chk.S
>> +++ /dev/null
>> @@ -1,208 +0,0 @@
>> -/* strcpy/stpcpy checking implementation for x86-64.
>> -   Copyright (C) 2002-2015 Free Software Foundation, Inc.
>> -   This file is part of the GNU C Library.
>> -   Contributed by Andreas Jaeger <aj@suse.de>, 2002.
>> -   Adopted into checking version by Jakub Jelinek <jakub@redhat.com>.
>> -
>> -   The GNU C Library is free software; you can redistribute it and/or
>> -   modify it under the terms of the GNU Lesser General Public
>> -   License as published by the Free Software Foundation; either
>> -   version 2.1 of the License, or (at your option) any later version.
>> -
>> -   The GNU C Library is distributed in the hope that it will be useful,
>> -   but WITHOUT ANY WARRANTY; without even the implied warranty of
>> -   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
>> -   Lesser General Public License for more details.
>> -
>> -   You should have received a copy of the GNU Lesser General Public
>> -   License along with the GNU C Library; if not, see
>> -   <http://www.gnu.org/licenses/>.  */
>> -
>> -#include <sysdep.h>
>> -#include "asm-syntax.h"
>> -
>> -#ifndef USE_AS_STPCPY_CHK
>> -# define STRCPY_CHK __strcpy_chk
>> -#endif
>> -
>> -	.text
>> -ENTRY (STRCPY_CHK)
>> -	movq	%rsi, %rcx	/* Source register. */
>> -	andl	$7, %ecx	/* mask alignment bits */
>> -#ifndef USE_AS_STPCPY_CHK
>> -	movq	%rdi, %r10	/* Duplicate destination pointer.  */
>> -#endif
>> -	jz 5f			/* aligned => start loop */
>> -
>> -	cmpq	$8, %rdx	/* Check if only few bytes left in
>> -				   destination.  */
>> -	jb	50f
>> -
>> -	subq	$8, %rcx	/* We need to align to 8 bytes.  */
>> -	addq	%rcx, %rdx	/* Subtract count of stored bytes
>> -				   in the cycle below from destlen.  */
>> -
>> -	/* Search the first bytes directly.  */
>> -0:
>> -	movb	(%rsi), %al	/* Fetch a byte */
>> -	testb	%al, %al	/* Is it NUL? */
>> -	movb	%al, (%rdi)	/* Store it */
>> -	jz	4f		/* If it was NUL, done! */
>> -	incq	%rsi
>> -	incq	%rdi
>> -	incl	%ecx
>> -	jnz	0b
>> -
>> -5:
>> -	movq $0xfefefefefefefeff,%r8
>> -	cmpq	$32, %rdx	/* Are there enough bytes in destination
>> -				   for the next unrolled round?  */
>> -	jb	60f		/* If not, avoid the unrolled loop.  */
>> -
>> -	/* Now the sources is aligned.  Unfortunatly we cannot force
>> -	   to have both source and destination aligned, so ignore the
>> -	   alignment of the destination.  */
>> -	.p2align 4
>> -1:
>> -	/* 1st unroll.  */
>> -	movq	(%rsi), %rax	/* Read double word (8 bytes).  */
>> -	addq	$8, %rsi	/* Adjust pointer for next word.  */
>> -	movq	%rax, %r9	/* Save a copy for NUL finding.  */
>> -	addq	%r8, %r9	/* add the magic value to the word.  We get
>> -				   carry bits reported for each byte which
>> -				   is *not* 0 */
>> -	jnc	3f		/* highest byte is NUL => return pointer */
>> -	xorq	%rax, %r9	/* (word+magic)^word */
>> -	orq	%r8, %r9	/* set all non-carry bits */
>> -	incq	%r9		/* add 1: if one carry bit was *not* set
>> -				   the addition will not result in 0.  */
>> -
>> -	jnz	3f		/* found NUL => return pointer */
>> -
>> -	movq	%rax, (%rdi)	/* Write value to destination.  */
>> -	addq	$8, %rdi	/* Adjust pointer.  */
>> -
>> -	/* 2nd unroll.  */
>> -	movq	(%rsi), %rax	/* Read double word (8 bytes).  */
>> -	addq	$8, %rsi	/* Adjust pointer for next word.  */
>> -	movq	%rax, %r9	/* Save a copy for NUL finding.  */
>> -	addq	%r8, %r9	/* add the magic value to the word.  We get
>> -				   carry bits reported for each byte which
>> -				   is *not* 0 */
>> -	jnc	3f		/* highest byte is NUL => return pointer */
>> -	xorq	%rax, %r9	/* (word+magic)^word */
>> -	orq	%r8, %r9	/* set all non-carry bits */
>> -	incq	%r9		/* add 1: if one carry bit was *not* set
>> -				   the addition will not result in 0.  */
>> -
>> -	jnz	3f		/* found NUL => return pointer */
>> -
>> -	movq	%rax, (%rdi)	/* Write value to destination.  */
>> -	addq	$8, %rdi	/* Adjust pointer.  */
>> -
>> -	/* 3rd unroll.  */
>> -	movq	(%rsi), %rax	/* Read double word (8 bytes).  */
>> -	addq	$8, %rsi	/* Adjust pointer for next word.  */
>> -	movq	%rax, %r9	/* Save a copy for NUL finding.  */
>> -	addq	%r8, %r9	/* add the magic value to the word.  We get
>> -				   carry bits reported for each byte which
>> -				   is *not* 0 */
>> -	jnc	3f		/* highest byte is NUL => return pointer */
>> -	xorq	%rax, %r9	/* (word+magic)^word */
>> -	orq	%r8, %r9	/* set all non-carry bits */
>> -	incq	%r9		/* add 1: if one carry bit was *not* set
>> -				   the addition will not result in 0.  */
>> -
>> -	jnz	3f		/* found NUL => return pointer */
>> -
>> -	movq	%rax, (%rdi)	/* Write value to destination.  */
>> -	addq	$8, %rdi	/* Adjust pointer.  */
>> -
>> -	/* 4th unroll.  */
>> -	movq	(%rsi), %rax	/* Read double word (8 bytes).  */
>> -	addq	$8, %rsi	/* Adjust pointer for next word.  */
>> -	movq	%rax, %r9	/* Save a copy for NUL finding.  */
>> -	addq	%r8, %r9	/* add the magic value to the word.  We get
>> -				   carry bits reported for each byte which
>> -				   is *not* 0 */
>> -	jnc	3f		/* highest byte is NUL => return pointer */
>> -	xorq	%rax, %r9	/* (word+magic)^word */
>> -	orq	%r8, %r9	/* set all non-carry bits */
>> -	incq	%r9		/* add 1: if one carry bit was *not* set
>> -				   the addition will not result in 0.  */
>> -
>> -	jnz	3f		/* found NUL => return pointer */
>> -
>> -	subq	$32, %rdx	/* Adjust destlen.  */
>> -	movq	%rax, (%rdi)	/* Write value to destination.  */
>> -	addq	$8, %rdi	/* Adjust pointer.  */
>> -	cmpq	$32, %rdx	/* Are there enough bytes in destination
>> -				   for the next unrolled round?  */
>> -	jae	1b		/* Next iteration.  */
>> -
>> -60:
>> -	cmpq	$8, %rdx	/* Are there enough bytes in destination
>> -				   for the next unrolled round?  */
>> -	jb	50f		/* Now, copy and check byte by byte.  */
>> -
>> -	movq	(%rsi), %rax	/* Read double word (8 bytes).  */
>> -	addq	$8, %rsi	/* Adjust pointer for next word.  */
>> -	movq	%rax, %r9	/* Save a copy for NUL finding.  */
>> -	addq	%r8, %r9	/* add the magic value to the word.  We get
>> -				   carry bits reported for each byte which
>> -				   is *not* 0 */
>> -	jnc	3f		/* highest byte is NUL => return pointer */
>> -	xorq	%rax, %r9	/* (word+magic)^word */
>> -	orq	%r8, %r9	/* set all non-carry bits */
>> -	incq	%r9		/* add 1: if one carry bit was *not* set
>> -				   the addition will not result in 0.  */
>> -
>> -	jnz	3f		/* found NUL => return pointer */
>> -
>> -	subq	$8, %rdx	/* Adjust destlen.  */
>> -	movq	%rax, (%rdi)	/* Write value to destination.  */
>> -	addq	$8, %rdi	/* Adjust pointer.  */
>> -	jmp	60b		/* Next iteration.  */
>> -
>> -	/* Do the last few bytes. %rax contains the value to write.
>> -	   The loop is unrolled twice.  */
>> -	.p2align 4
>> -3:
>> -	/* Note that stpcpy needs to return with the value of the NUL
>> -	   byte.  */
>> -	movb	%al, (%rdi)	/* 1st byte.  */
>> -	testb	%al, %al	/* Is it NUL.  */
>> -	jz	4f		/* yes, finish.  */
>> -	incq	%rdi		/* Increment destination.  */
>> -	movb	%ah, (%rdi)	/* 2nd byte.  */
>> -	testb	%ah, %ah	/* Is it NUL?.  */
>> -	jz	4f		/* yes, finish.  */
>> -	incq	%rdi		/* Increment destination.  */
>> -	shrq	$16, %rax	/* Shift...  */
>> -	jmp	3b		/* and look at next two bytes in %rax.  */
>> -
>> -51:
>> -	/* Search the bytes directly, checking for overflows.  */
>> -	incq	%rsi
>> -	incq	%rdi
>> -	decq	%rdx
>> -	jz	HIDDEN_JUMPTARGET (__chk_fail)
>> -52:
>> -	movb	(%rsi), %al	/* Fetch a byte */
>> -	testb	%al, %al	/* Is it NUL? */
>> -	movb	%al, (%rdi)	/* Store it */
>> -	jnz	51b		/* If it was NUL, done! */
>> -4:
>> -#ifdef USE_AS_STPCPY_CHK
>> -	movq	%rdi, %rax	/* Destination is return value.  */
>> -#else
>> -	movq	%r10, %rax	/* Source is return value.  */
>> -#endif
>> -	retq
>> -
>> -50:
>> -	testq	%rdx, %rdx
>> -	jnz	52b
>> -	jmp	HIDDEN_JUMPTARGET (__chk_fail)
>> -
>> -END (STRCPY_CHK)
>
  

On Wed, 19 Aug 2015, Adhemerval Zanella wrote:

> Could you also get rid of the k&r function prototype (same for strcpy_chk) ?

I'm not really convinced it's good to mix such fixes with changes for 
other things (at least it shouldn't be required).

Ondřej had some automation for fixing such old-style definitions, see e.g. 
<https://sourceware.org/ml/libc-alpha/2013-06/msg00270.html>.  I think it 
would be good to try to get such changes in (eventually adding 
-Wold-style-definition to the warning options, possibly with an 
intermediate state of -Wold-style-definition 
-Wno-error=old-style-definition while architecture maintainers clean up 
their architectures) - starting with straightforward cases in files that 
do not use assertions (so changes to line numbers don't affect comparisons 
of generated code) and for which the changes do not affect the installed 
stripped shared libraries at all, then moving to other cases for which 
more careful review is needed.