xtensa: abstract 'entry' and 'retw' in assembly code
Provide abi_entry, abi_entry_default, abi_ret and abi_ret_default macros
that allocate aligned stack frame in windowed and call0 ABIs.
Provide XTENSA_SPILL_STACK_RESERVE macro that specifies required stack
frame size when register spilling is involved.
Replace all uses of 'entry' and 'retw' with the above macros.
This makes most of the xtensa assembly code ready for XEA3 and call0 ABI.
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
diff --git a/arch/xtensa/lib/checksum.S b/arch/xtensa/lib/checksum.S
index c6e73b1..4cb9ca5 100644
--- a/arch/xtensa/lib/checksum.S
+++ b/arch/xtensa/lib/checksum.S
@@ -43,7 +43,7 @@
* Experiments with Ethernet and SLIP connections show that buf
* is aligned on either a 2-byte or 4-byte boundary.
*/
- entry sp, 32
+ abi_entry_default
extui a5, a2, 0, 2
bnez a5, 8f /* branch if 2-byte aligned */
/* Fall-through on common case, 4-byte alignment */
@@ -107,7 +107,7 @@
ONES_ADD(a4, a6)
7:
mov a2, a4
- retw
+ abi_ret_default
/* uncommon case, buf is 2-byte aligned */
8:
@@ -195,7 +195,7 @@
ENTRY(csum_partial_copy_generic)
- entry sp, 32
+ abi_entry_default
mov a12, a3
mov a11, a4
or a10, a2, a3
@@ -316,7 +316,7 @@
ONES_ADD(a5, a9)
8:
mov a2, a5
- retw
+ abi_ret_default
5:
/* Control branch to here when either src or dst is odd. We
@@ -383,12 +383,12 @@
blt a12, a11, .Leloop
#endif
2:
- retw
+ abi_ret_default
11:
movi a2, -EFAULT
s32i a2, a7, 0 /* dst_err_ptr */
movi a2, 0
- retw
+ abi_ret_default
.previous
diff --git a/arch/xtensa/lib/memcopy.S b/arch/xtensa/lib/memcopy.S
index efecfd7..582d817 100644
--- a/arch/xtensa/lib/memcopy.S
+++ b/arch/xtensa/lib/memcopy.S
@@ -79,7 +79,7 @@
bne a3, a7, .Lnextbyte # continue loop if $a3:src != $a7:src_end
#endif /* !XCHAL_HAVE_LOOPS */
.Lbytecopydone:
- retw
+ abi_ret_default
/*
* Destination is unaligned
@@ -112,7 +112,7 @@
ENTRY(__memcpy)
WEAK(memcpy)
- entry sp, 16 # minimal stack frame
+ abi_entry_default
# a2/ dst, a3/ src, a4/ len
mov a5, a2 # copy dst so that a2 is return value
.Lcommon:
@@ -161,7 +161,7 @@
bbsi.l a4, 2, .L3
bbsi.l a4, 1, .L4
bbsi.l a4, 0, .L5
- retw
+ abi_ret_default
.L3:
# copy 4 bytes
l32i a6, a3, 0
@@ -170,7 +170,7 @@
addi a5, a5, 4
bbsi.l a4, 1, .L4
bbsi.l a4, 0, .L5
- retw
+ abi_ret_default
.L4:
# copy 2 bytes
l16ui a6, a3, 0
@@ -178,12 +178,12 @@
s16i a6, a5, 0
addi a5, a5, 2
bbsi.l a4, 0, .L5
- retw
+ abi_ret_default
.L5:
# copy 1 byte
l8ui a6, a3, 0
s8i a6, a5, 0
- retw
+ abi_ret_default
/*
* Destination is aligned, Source is unaligned
@@ -255,7 +255,7 @@
#endif
bbsi.l a4, 1, .L14
bbsi.l a4, 0, .L15
-.Ldone: retw
+.Ldone: abi_ret_default
.L14:
# copy 2 bytes
l8ui a6, a3, 0
@@ -265,12 +265,12 @@
s8i a7, a5, 1
addi a5, a5, 2
bbsi.l a4, 0, .L15
- retw
+ abi_ret_default
.L15:
# copy 1 byte
l8ui a6, a3, 0
s8i a6, a5, 0
- retw
+ abi_ret_default
ENDPROC(__memcpy)
@@ -280,7 +280,7 @@
ENTRY(bcopy)
- entry sp, 16 # minimal stack frame
+ abi_entry_default
# a2=src, a3=dst, a4=len
mov a5, a3
mov a3, a2
@@ -346,7 +346,7 @@
# $a3:src != $a7:src_start
#endif /* !XCHAL_HAVE_LOOPS */
.Lbackbytecopydone:
- retw
+ abi_ret_default
/*
* Destination is unaligned
@@ -380,7 +380,7 @@
ENTRY(__memmove)
WEAK(memmove)
- entry sp, 16 # minimal stack frame
+ abi_entry_default
# a2/ dst, a3/ src, a4/ len
mov a5, a2 # copy dst so that a2 is return value
.Lmovecommon:
@@ -435,7 +435,7 @@
bbsi.l a4, 2, .Lback3
bbsi.l a4, 1, .Lback4
bbsi.l a4, 0, .Lback5
- retw
+ abi_ret_default
.Lback3:
# copy 4 bytes
addi a3, a3, -4
@@ -444,7 +444,7 @@
s32i a6, a5, 0
bbsi.l a4, 1, .Lback4
bbsi.l a4, 0, .Lback5
- retw
+ abi_ret_default
.Lback4:
# copy 2 bytes
addi a3, a3, -2
@@ -452,14 +452,14 @@
addi a5, a5, -2
s16i a6, a5, 0
bbsi.l a4, 0, .Lback5
- retw
+ abi_ret_default
.Lback5:
# copy 1 byte
addi a3, a3, -1
l8ui a6, a3, 0
addi a5, a5, -1
s8i a6, a5, 0
- retw
+ abi_ret_default
/*
* Destination is aligned, Source is unaligned
@@ -531,7 +531,7 @@
bbsi.l a4, 1, .Lback14
bbsi.l a4, 0, .Lback15
.Lbackdone:
- retw
+ abi_ret_default
.Lback14:
# copy 2 bytes
addi a3, a3, -2
@@ -541,13 +541,13 @@
s8i a6, a5, 0
s8i a7, a5, 1
bbsi.l a4, 0, .Lback15
- retw
+ abi_ret_default
.Lback15:
# copy 1 byte
addi a3, a3, -1
addi a5, a5, -1
l8ui a6, a3, 0
s8i a6, a5, 0
- retw
+ abi_ret_default
ENDPROC(__memmove)
diff --git a/arch/xtensa/lib/memset.S b/arch/xtensa/lib/memset.S
index 8632eac..59b1524 100644
--- a/arch/xtensa/lib/memset.S
+++ b/arch/xtensa/lib/memset.S
@@ -34,7 +34,7 @@
ENTRY(__memset)
WEAK(memset)
- entry sp, 16 # minimal stack frame
+ abi_entry_default
# a2/ dst, a3/ c, a4/ length
extui a3, a3, 0, 8 # mask to just 8 bits
slli a7, a3, 8 # duplicate character in all bytes of word
@@ -48,7 +48,7 @@
srli a7, a4, 4 # number of loop iterations with 16B
# per iteration
bnez a4, .Laligned
- retw
+ abi_ret_default
/*
* Destination is word-aligned.
@@ -95,7 +95,7 @@
EX(10f) s8i a3, a5, 0
.L5:
.Lret1:
- retw
+ abi_ret_default
/*
* Destination is unaligned
@@ -139,7 +139,7 @@
blt a5, a6, .Lbyteloop
#endif /* !XCHAL_HAVE_LOOPS */
.Lbytesetdone:
- retw
+ abi_ret_default
ENDPROC(__memset)
@@ -150,4 +150,4 @@
10:
movi a2, 0
- retw
+ abi_ret_default
diff --git a/arch/xtensa/lib/strncpy_user.S b/arch/xtensa/lib/strncpy_user.S
index c4c6c85..4faf46f 100644
--- a/arch/xtensa/lib/strncpy_user.S
+++ b/arch/xtensa/lib/strncpy_user.S
@@ -50,7 +50,7 @@
.text
ENTRY(__strncpy_user)
- entry sp, 16 # minimal stack frame
+ abi_entry_default
# a2/ dst, a3/ src, a4/ len
mov a11, a2 # leave dst in return value register
beqz a4, .Lret # if len is zero
@@ -93,7 +93,7 @@
bnez a4, .Lsrcaligned # if len is nonzero
.Lret:
sub a2, a11, a2 # compute strlen
- retw
+ abi_ret_default
/*
* dst is word-aligned, src is word-aligned
@@ -148,14 +148,14 @@
.Lz3: # byte 3 is zero
addi a11, a11, 3 # advance dst pointer
sub a2, a11, a2 # compute strlen
- retw
+ abi_ret_default
.Lz0: # byte 0 is zero
#ifdef __XTENSA_EB__
movi a9, 0
#endif /* __XTENSA_EB__ */
EX(10f) s8i a9, a11, 0
sub a2, a11, a2 # compute strlen
- retw
+ abi_ret_default
.Lz1: # byte 1 is zero
#ifdef __XTENSA_EB__
extui a9, a9, 16, 16
@@ -163,7 +163,7 @@
EX(10f) s16i a9, a11, 0
addi a11, a11, 1 # advance dst pointer
sub a2, a11, a2 # compute strlen
- retw
+ abi_ret_default
.Lz2: # byte 2 is zero
#ifdef __XTENSA_EB__
extui a9, a9, 16, 16
@@ -173,7 +173,7 @@
EX(10f) s8i a9, a11, 2
addi a11, a11, 2 # advance dst pointer
sub a2, a11, a2 # compute strlen
- retw
+ abi_ret_default
.align 4 # 1 mod 4 alignment for LOOPNEZ
.byte 0 # (0 mod 4 alignment for LBEG)
@@ -199,7 +199,7 @@
.Lunalignedend:
sub a2, a11, a2 # compute strlen
- retw
+ abi_ret_default
ENDPROC(__strncpy_user)
@@ -214,4 +214,4 @@
10:
11:
movi a2, -EFAULT
- retw
+ abi_ret_default
diff --git a/arch/xtensa/lib/strnlen_user.S b/arch/xtensa/lib/strnlen_user.S
index 1f2ca2b..3d391dca 100644
--- a/arch/xtensa/lib/strnlen_user.S
+++ b/arch/xtensa/lib/strnlen_user.S
@@ -45,7 +45,7 @@
.text
ENTRY(__strnlen_user)
- entry sp, 16 # minimal stack frame
+ abi_entry_default
# a2/ s, a3/ len
addi a4, a2, -4 # because we overincrement at the end;
# we compensate with load offsets of 4
@@ -96,7 +96,7 @@
addi a4, a4, 1 # advance string pointer
.L101:
sub a2, a4, a2 # compute length
- retw
+ abi_ret_default
# NOTE that in several places below, we point to the byte just after
# the zero byte in order to include the NULL terminator in the count.
@@ -106,15 +106,15 @@
.Lz0: # byte 0 is zero
addi a4, a4, 1 # point just beyond zero byte
sub a2, a4, a2 # subtract to get length
- retw
+ abi_ret_default
.Lz1: # byte 1 is zero
addi a4, a4, 1+1 # point just beyond zero byte
sub a2, a4, a2 # subtract to get length
- retw
+ abi_ret_default
.Lz2: # byte 2 is zero
addi a4, a4, 2+1 # point just beyond zero byte
sub a2, a4, a2 # subtract to get length
- retw
+ abi_ret_default
.L1mod2: # address is odd
EX(10f) l8ui a9, a4, 4 # get byte 0
@@ -130,7 +130,7 @@
# byte 3 is zero
addi a4, a4, 3+1 # point just beyond zero byte
sub a2, a4, a2 # subtract to get length
- retw
+ abi_ret_default
ENDPROC(__strnlen_user)
@@ -138,4 +138,4 @@
.align 4
10:
movi a2, 0
- retw
+ abi_ret_default
diff --git a/arch/xtensa/lib/usercopy.S b/arch/xtensa/lib/usercopy.S
index 228607e..a0aa404 100644
--- a/arch/xtensa/lib/usercopy.S
+++ b/arch/xtensa/lib/usercopy.S
@@ -60,7 +60,7 @@
.text
ENTRY(__xtensa_copy_user)
- entry sp, 16 # minimal stack frame
+ abi_entry_default
# a2/ dst, a3/ src, a4/ len
mov a5, a2 # copy dst so that a2 is return value
mov a11, a4 # preserve original len for error case
@@ -75,7 +75,7 @@
__ssa8 a3 # set shift amount from byte offset
bnez a4, .Lsrcunaligned
movi a2, 0 # return success for len==0
- retw
+ abi_ret_default
/*
* Destination is unaligned
@@ -127,7 +127,7 @@
#endif /* !XCHAL_HAVE_LOOPS */
.Lbytecopydone:
movi a2, 0 # return success for len bytes copied
- retw
+ abi_ret_default
/*
* Destination and source are word-aligned.
@@ -187,7 +187,7 @@
EX(10f) s8i a6, a5, 0
.L5:
movi a2, 0 # return success for len bytes copied
- retw
+ abi_ret_default
/*
* Destination is aligned, Source is unaligned
@@ -264,7 +264,7 @@
EX(10f) s8i a6, a5, 0
.L15:
movi a2, 0 # return success for len bytes copied
- retw
+ abi_ret_default
ENDPROC(__xtensa_copy_user)
@@ -281,4 +281,4 @@
10:
sub a2, a5, a2 /* a2 <-- bytes copied */
sub a2, a11, a2 /* a2 <-- bytes not copied */
- retw
+ abi_ret_default