On Tuesday 10 January 2006 21:53, Bryan O'Sullivan wrote:
> Introduce an x86_64-specific memcpy32 routine. The routine is similar
> to memcpy, but is guaranteed to work in units of 32 bits at a time.
>
> Signed-off-by: Bryan O'Sullivan <[email protected]>
>
> diff -r 2d4af213d9c5 -r b4863171295f arch/x86_64/kernel/x8664_ksyms.c
> --- a/arch/x86_64/kernel/x8664_ksyms.c Tue Jan 10 11:52:46 2006 -0800
> +++ b/arch/x86_64/kernel/x8664_ksyms.c Tue Jan 10 11:52:48 2006 -0800
> @@ -164,6 +164,8 @@
> EXPORT_SYMBOL(memcpy);
> EXPORT_SYMBOL(__memcpy);
>
> +EXPORT_SYMBOL_GPL(memcpy32);
> +
> #ifdef CONFIG_RWSEM_XCHGADD_ALGORITHM
> /* prototypes are wrong, these are assembly with custom calling functions */
> extern void rwsem_down_read_failed_thunk(void);
> diff -r 2d4af213d9c5 -r b4863171295f arch/x86_64/lib/Makefile
> --- a/arch/x86_64/lib/Makefile Tue Jan 10 11:52:46 2006 -0800
> +++ b/arch/x86_64/lib/Makefile Tue Jan 10 11:52:48 2006 -0800
> @@ -9,4 +9,4 @@
> lib-y := csum-partial.o csum-copy.o csum-wrappers.o delay.o \
> usercopy.o getuser.o putuser.o \
> thunk.o clear_page.o copy_page.o bitstr.o bitops.o
> -lib-y += memcpy.o memmove.o memset.o copy_user.o
> +lib-y += memcpy.o memcpy32.o memmove.o memset.o copy_user.o
> diff -r 2d4af213d9c5 -r b4863171295f include/asm-x86_64/string.h
> --- a/include/asm-x86_64/string.h Tue Jan 10 11:52:46 2006 -0800
> +++ b/include/asm-x86_64/string.h Tue Jan 10 11:52:48 2006 -0800
> @@ -45,6 +45,9 @@
> #define __HAVE_ARCH_MEMMOVE
> void * memmove(void * dest,const void *src,size_t count);
>
> +/* copy data, 32 bits at a time */
> +void memcpy32(void *dst, const void *src, size_t count);
> +
> /* Use C out of line version for memcmp */
> #define memcmp __builtin_memcmp
> int memcmp(const void * cs,const void * ct,size_t count);
> diff -r 2d4af213d9c5 -r b4863171295f arch/x86_64/lib/memcpy32.S
> --- /dev/null Thu Jan 1 00:00:00 1970 +0000
> +++ b/arch/x86_64/lib/memcpy32.S Tue Jan 10 11:52:48 2006 -0800
> @@ -0,0 +1,39 @@
> +/*
> + * Copyright 2006 PathScale, Inc. All Rights Reserved.
> + *
> + * This file is free software; you can redistribute it and/or modify
> + * it under the terms of version 2 of the GNU General Public License
> + * as published by the Free Software Foundation.
> + *
> + * This program 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 General Public License for more details.
> + *
> + * You should have received a copy of the GNU General Public License
> + * along with this program; if not, write to the Free Software Foundation,
> + * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA.
> + */
> +
> +/*
> + * Registers used below:
> + * dst - rdi
> + * src - rsi
> + * count - rdx
> + */
> +
> +/**
> + * memcpy32 - copy data, in units of 32 bits at a time
> + * @dst: destination (must be 32-bit aligned)
> + * @src: source (must be 32-bit aligned)
> + * @count: number of 32-bit quantities to copy
> + */
> + .globl memcpy32
> +memcpy32:
> + movl %edx,%ecx
> + shrl $1,%ecx
> + andl $1,%edx
> + rep movsq
> + movl %edx,%ecx
> + rep movsd
> + ret
movsq is not a 32bit move, it's a 64 bit one.
There are three possibilities here:
1) I misunderstand what memcpy32 means (I understand it like "it guarantees
that all accesses will be strictly 32bit")
2) On all current x86_64 hardware each 64bit access from/to
IO mapped addresses is always converted to two 32bit accesses.
3) code is buggy
If it is (1) or (2), consider adding a comment to clear future
reader's confusion.
--
vda
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
[Index of Archives]
[Kernel Newbies]
[Netfilter]
[Bugtraq]
[Photo]
[Stuff]
[Gimp]
[Yosemite News]
[MIPS Linux]
[ARM Linux]
[Linux Security]
[Linux RAID]
[Video 4 Linux]
[Linux for the blind]
[Linux Resources]
