linux/arch/arm/lib/copy_page.S
Russell King 8478132a87 Revert "arm: move exports to definitions"
This reverts commit 4dd1837d75.

Moving the exports for assembly code into the assembly files breaks
KSYM trimming, but also breaks modversions.

While fixing the KSYM trimming is trivial, fixing modversions brings
us to a technically worse position that we had prior to the above
change:

- We end up with the prototype definitions divorsed from everything
  else, which means that adding or removing assembly level ksyms
  become more fragile:
  * if adding a new assembly ksyms export, a missed prototype in
    asm-prototypes.h results in a successful build if no module in
    the selected configuration makes use of the symbol.
  * when removing a ksyms export, asm-prototypes.h will get forgotten,
    with armksyms.c, you'll get a build error if you forget to touch
    the file.

- We end up with the same amount of include files and prototypes,
  they're just in a header file instead of a .c file with their
  exports.

As for lines of code, we don't get much of a size reduction:
 (original commit)
 47 files changed, 131 insertions(+), 208 deletions(-)
 (fix for ksyms trimming)
 7 files changed, 18 insertions(+), 5 deletions(-)
 (two fixes for modversions)
 1 file changed, 34 insertions(+)
 3 files changed, 7 insertions(+), 2 deletions(-)
which results in a net total of only 25 lines deleted.

As there does not seem to be much benefit from this change of approach,
revert the change.

Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
2016-11-23 10:00:03 +00:00

48 lines
1.3 KiB
ArmAsm

/*
* linux/arch/arm/lib/copypage.S
*
* Copyright (C) 1995-1999 Russell King
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2 as
* published by the Free Software Foundation.
*
* ASM optimised string functions
*/
#include <linux/linkage.h>
#include <asm/assembler.h>
#include <asm/asm-offsets.h>
#include <asm/cache.h>
#define COPY_COUNT (PAGE_SZ / (2 * L1_CACHE_BYTES) PLD( -1 ))
.text
.align 5
/*
* StrongARM optimised copy_page routine
* now 1.78bytes/cycle, was 1.60 bytes/cycle (50MHz bus -> 89MB/s)
* Note that we probably achieve closer to the 100MB/s target with
* the core clock switching.
*/
ENTRY(copy_page)
stmfd sp!, {r4, lr} @ 2
PLD( pld [r1, #0] )
PLD( pld [r1, #L1_CACHE_BYTES] )
mov r2, #COPY_COUNT @ 1
ldmia r1!, {r3, r4, ip, lr} @ 4+1
1: PLD( pld [r1, #2 * L1_CACHE_BYTES])
PLD( pld [r1, #3 * L1_CACHE_BYTES])
2:
.rept (2 * L1_CACHE_BYTES / 16 - 1)
stmia r0!, {r3, r4, ip, lr} @ 4
ldmia r1!, {r3, r4, ip, lr} @ 4
.endr
subs r2, r2, #1 @ 1
stmia r0!, {r3, r4, ip, lr} @ 4
ldmgtia r1!, {r3, r4, ip, lr} @ 4
bgt 1b @ 1
PLD( ldmeqia r1!, {r3, r4, ip, lr} )
PLD( beq 2b )
ldmfd sp!, {r4, pc} @ 3
ENDPROC(copy_page)