* a11/ original length
*/
+#include <linux/linkage.h>
#include <variant/core.h>
-
-#ifdef __XTENSA_EB__
-#define ALIGN(R, W0, W1) src R, W0, W1
-#define SSA8(R) ssa8b R
-#else
-#define ALIGN(R, W0, W1) src R, W1, W0
-#define SSA8(R) ssa8l R
-#endif
-
-/* Load or store instructions that may cause exceptions use the EX macro. */
-
-#define EX(insn,reg1,reg2,offset,handler) \
-9: insn reg1, reg2, offset; \
- .section __ex_table, "a"; \
- .word 9b, handler; \
- .previous
-
+#include <asm/asmmacro.h>
.text
- .align 4
- .global __xtensa_copy_user
- .type __xtensa_copy_user,@function
-__xtensa_copy_user:
+ENTRY(__xtensa_copy_user)
+
entry sp, 16 # minimal stack frame
# a2/ dst, a3/ src, a4/ len
mov a5, a2 # copy dst so that a2 is return value
# per iteration
movi a8, 3 # if source is also aligned,
bnone a3, a8, .Laligned # then use word copy
- SSA8( a3) # set shift amount from byte offset
+ __ssa8 a3 # set shift amount from byte offset
bnez a4, .Lsrcunaligned
movi a2, 0 # return success for len==0
retw
bltui a4, 7, .Lbytecopy # do short copies byte by byte
# copy 1 byte
- EX(l8ui, a6, a3, 0, fixup)
+EX(10f) l8ui a6, a3, 0
addi a3, a3, 1
- EX(s8i, a6, a5, 0, fixup)
+EX(10f) s8i a6, a5, 0
addi a5, a5, 1
addi a4, a4, -1
bbci.l a5, 1, .Ldstaligned # if dst is now aligned, then
.Ldst2mod4: # dst 16-bit aligned
# copy 2 bytes
bltui a4, 6, .Lbytecopy # do short copies byte by byte
- EX(l8ui, a6, a3, 0, fixup)
- EX(l8ui, a7, a3, 1, fixup)
+EX(10f) l8ui a6, a3, 0
+EX(10f) l8ui a7, a3, 1
addi a3, a3, 2
- EX(s8i, a6, a5, 0, fixup)
- EX(s8i, a7, a5, 1, fixup)
+EX(10f) s8i a6, a5, 0
+EX(10f) s8i a7, a5, 1
addi a5, a5, 2
addi a4, a4, -2
j .Ldstaligned # dst is now aligned, return to main algorithm
add a7, a3, a4 # a7 = end address for source
#endif /* !XCHAL_HAVE_LOOPS */
.Lnextbyte:
- EX(l8ui, a6, a3, 0, fixup)
+EX(10f) l8ui a6, a3, 0
addi a3, a3, 1
- EX(s8i, a6, a5, 0, fixup)
+EX(10f) s8i a6, a5, 0
addi a5, a5, 1
#if !XCHAL_HAVE_LOOPS
blt a3, a7, .Lnextbyte
add a8, a8, a3 # a8 = end of last 16B source chunk
#endif /* !XCHAL_HAVE_LOOPS */
.Loop1:
- EX(l32i, a6, a3, 0, fixup)
- EX(l32i, a7, a3, 4, fixup)
- EX(s32i, a6, a5, 0, fixup)
- EX(l32i, a6, a3, 8, fixup)
- EX(s32i, a7, a5, 4, fixup)
- EX(l32i, a7, a3, 12, fixup)
- EX(s32i, a6, a5, 8, fixup)
+EX(10f) l32i a6, a3, 0
+EX(10f) l32i a7, a3, 4
+EX(10f) s32i a6, a5, 0
+EX(10f) l32i a6, a3, 8
+EX(10f) s32i a7, a5, 4
+EX(10f) l32i a7, a3, 12
+EX(10f) s32i a6, a5, 8
addi a3, a3, 16
- EX(s32i, a7, a5, 12, fixup)
+EX(10f) s32i a7, a5, 12
addi a5, a5, 16
#if !XCHAL_HAVE_LOOPS
blt a3, a8, .Loop1
.Loop1done:
bbci.l a4, 3, .L2
# copy 8 bytes
- EX(l32i, a6, a3, 0, fixup)
- EX(l32i, a7, a3, 4, fixup)
+EX(10f) l32i a6, a3, 0
+EX(10f) l32i a7, a3, 4
addi a3, a3, 8
- EX(s32i, a6, a5, 0, fixup)
- EX(s32i, a7, a5, 4, fixup)
+EX(10f) s32i a6, a5, 0
+EX(10f) s32i a7, a5, 4
addi a5, a5, 8
.L2:
bbci.l a4, 2, .L3
# copy 4 bytes
- EX(l32i, a6, a3, 0, fixup)
+EX(10f) l32i a6, a3, 0
addi a3, a3, 4
- EX(s32i, a6, a5, 0, fixup)
+EX(10f) s32i a6, a5, 0
addi a5, a5, 4
.L3:
bbci.l a4, 1, .L4
# copy 2 bytes
- EX(l16ui, a6, a3, 0, fixup)
+EX(10f) l16ui a6, a3, 0
addi a3, a3, 2
- EX(s16i, a6, a5, 0, fixup)
+EX(10f) s16i a6, a5, 0
addi a5, a5, 2
.L4:
bbci.l a4, 0, .L5
# copy 1 byte
- EX(l8ui, a6, a3, 0, fixup)
- EX(s8i, a6, a5, 0, fixup)
+EX(10f) l8ui a6, a3, 0
+EX(10f) s8i a6, a5, 0
.L5:
movi a2, 0 # return success for len bytes copied
retw
# copy 16 bytes per iteration for word-aligned dst and unaligned src
and a10, a3, a8 # save unalignment offset for below
sub a3, a3, a10 # align a3 (to avoid sim warnings only; not needed for hardware)
- EX(l32i, a6, a3, 0, fixup) # load first word
+EX(10f) l32i a6, a3, 0 # load first word
#if XCHAL_HAVE_LOOPS
loopnez a7, .Loop2done
#else /* !XCHAL_HAVE_LOOPS */
add a12, a12, a3 # a12 = end of last 16B source chunk
#endif /* !XCHAL_HAVE_LOOPS */
.Loop2:
- EX(l32i, a7, a3, 4, fixup)
- EX(l32i, a8, a3, 8, fixup)
- ALIGN( a6, a6, a7)
- EX(s32i, a6, a5, 0, fixup)
- EX(l32i, a9, a3, 12, fixup)
- ALIGN( a7, a7, a8)
- EX(s32i, a7, a5, 4, fixup)
- EX(l32i, a6, a3, 16, fixup)
- ALIGN( a8, a8, a9)
- EX(s32i, a8, a5, 8, fixup)
+EX(10f) l32i a7, a3, 4
+EX(10f) l32i a8, a3, 8
+ __src_b a6, a6, a7
+EX(10f) s32i a6, a5, 0
+EX(10f) l32i a9, a3, 12
+ __src_b a7, a7, a8
+EX(10f) s32i a7, a5, 4
+EX(10f) l32i a6, a3, 16
+ __src_b a8, a8, a9
+EX(10f) s32i a8, a5, 8
addi a3, a3, 16
- ALIGN( a9, a9, a6)
- EX(s32i, a9, a5, 12, fixup)
+ __src_b a9, a9, a6
+EX(10f) s32i a9, a5, 12
addi a5, a5, 16
#if !XCHAL_HAVE_LOOPS
blt a3, a12, .Loop2
.Loop2done:
bbci.l a4, 3, .L12
# copy 8 bytes
- EX(l32i, a7, a3, 4, fixup)
- EX(l32i, a8, a3, 8, fixup)
- ALIGN( a6, a6, a7)
- EX(s32i, a6, a5, 0, fixup)
+EX(10f) l32i a7, a3, 4
+EX(10f) l32i a8, a3, 8
+ __src_b a6, a6, a7
+EX(10f) s32i a6, a5, 0
addi a3, a3, 8
- ALIGN( a7, a7, a8)
- EX(s32i, a7, a5, 4, fixup)
+ __src_b a7, a7, a8
+EX(10f) s32i a7, a5, 4
addi a5, a5, 8
mov a6, a8
.L12:
bbci.l a4, 2, .L13
# copy 4 bytes
- EX(l32i, a7, a3, 4, fixup)
+EX(10f) l32i a7, a3, 4
addi a3, a3, 4
- ALIGN( a6, a6, a7)
- EX(s32i, a6, a5, 0, fixup)
+ __src_b a6, a6, a7
+EX(10f) s32i a6, a5, 0
addi a5, a5, 4
mov a6, a7
.L13:
add a3, a3, a10 # readjust a3 with correct misalignment
bbci.l a4, 1, .L14
# copy 2 bytes
- EX(l8ui, a6, a3, 0, fixup)
- EX(l8ui, a7, a3, 1, fixup)
+EX(10f) l8ui a6, a3, 0
+EX(10f) l8ui a7, a3, 1
addi a3, a3, 2
- EX(s8i, a6, a5, 0, fixup)
- EX(s8i, a7, a5, 1, fixup)
+EX(10f) s8i a6, a5, 0
+EX(10f) s8i a7, a5, 1
addi a5, a5, 2
.L14:
bbci.l a4, 0, .L15
# copy 1 byte
- EX(l8ui, a6, a3, 0, fixup)
- EX(s8i, a6, a5, 0, fixup)
+EX(10f) l8ui a6, a3, 0
+EX(10f) s8i a6, a5, 0
.L15:
movi a2, 0 # return success for len bytes copied
retw
+ENDPROC(__xtensa_copy_user)
.section .fixup, "ax"
.align 4
*/
-fixup:
+10:
sub a2, a5, a2 /* a2 <-- bytes copied */
sub a2, a11, a2 /* a2 <-- bytes not copied */
retw
git.samba.org / sfrench / cifs-2.6.git / blobdiff