xtensa: support call0 ABI

Most changes are mechanical replacement of 'retw' instruction with
'abi_ret' macro, defined to 'retw' or 'ret' according to ABI.
Assembly code that makes calls is duplicated for call0 ABI with changed
register numbers for parameters/return value and call instruction.
'entry' instructions are replaced with 'abi_entry' macro.

More interesting changes:
- non-leaf assembly functions (e.g. _dl_tlsdesc_dynamic,
  _dl_linux_resolve, SYSCALL_ERROR_HANDLER, PSEUDO) now need to preserve
  registers around intermediate calls they make, use temporary stack
  frame for that;
- setjmp/longjmp only need to save and restore return address, stack
  pointer and callee-saved registers in the jmpbuf;
- __clone and syscall functions had hardcoded offsets to parameter
  passed on stack, on call0 ABI they don't need stack frame, so the
  offset is different. Replace these offsets with FRAMESIZE macro.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>

6 files changed, 33 insertions, 33 deletions

diff --git a/libc/string/xtensa/memcpy.S b/libc/string/xtensa/memcpy.S
index 2b23b4113..244205611 100644
--- a/libc/string/xtensa/memcpy.S
+++ b/libc/string/xtensa/memcpy.S
@@ -91,7 +91,7 @@ __memcpy_aux:
 #if !XCHAL_HAVE_LOOPS
 	blt	a3, a7, 1b
 #endif
-2:	retw
+2:	abi_ret
 
 
 /* Destination is unaligned.  */
@@ -181,7 +181,7 @@ ENTRY (memcpy)
 3:	bbsi.l	a4, 2, 4f
 	bbsi.l	a4, 1, 5f
 	bbsi.l	a4, 0, 6f
-	retw
+	abi_ret
 
 	/* Copy 4 bytes.  */
 4:	l32i	a6, a3, 0
@@ -190,7 +190,7 @@ ENTRY (memcpy)
 	addi	a5, a5, 4
 	bbsi.l	a4, 1, 5f
 	bbsi.l	a4, 0, 6f
-	retw
+	abi_ret
 
 	/* Copy 2 bytes.  */
 5:	l16ui	a6, a3, 0
@@ -198,14 +198,14 @@ ENTRY (memcpy)
 	s16i	a6, a5, 0
 	addi	a5, a5, 2
 	bbsi.l	a4, 0, 6f
-	retw
+	abi_ret
 
 	/* Copy 1 byte.  */
 6:	l8ui	a6, a3, 0
 	s8i	a6, a5, 0
 
 .Ldone:
-	retw
+	abi_ret
 
 
 /* Destination is aligned; source is unaligned.  */
@@ -276,7 +276,7 @@ ENTRY (memcpy)
 #endif
 	bbsi.l	a4, 1, 5f
 	bbsi.l	a4, 0, 6f
-	retw
+	abi_ret
 
 	/* Copy 2 bytes.  */
 5:	l8ui	a6, a3, 0
@@ -286,11 +286,11 @@ ENTRY (memcpy)
 	s8i	a7, a5, 1
 	addi	a5, a5, 2
 	bbsi.l	a4, 0, 6f
-	retw
+	abi_ret
 
 	/* Copy 1 byte.  */
 6:	l8ui	a6, a3, 0
 	s8i	a6, a5, 0
-	retw
+	abi_ret
 
 libc_hidden_def (memcpy)
diff --git a/libc/string/xtensa/memset.S b/libc/string/xtensa/memset.S
index 435d6dea0..20bf14c75 100644
--- a/libc/string/xtensa/memset.S
+++ b/libc/string/xtensa/memset.S
@@ -63,7 +63,7 @@ __memset_aux:
 #if !XCHAL_HAVE_LOOPS
 	blt	a5, a6, 1b
 #endif
-2:	retw
+2:	abi_ret
 
 
 /* Destination is unaligned.  */
@@ -159,6 +159,6 @@ ENTRY (memset)
 
 	/* Set 1 byte.  */
 	s8i	a3, a5, 0
-6:	retw
+6:	abi_ret
 
 libc_hidden_def (memset)
diff --git a/libc/string/xtensa/strcmp.S b/libc/string/xtensa/strcmp.S
index 8ed1ed009..2dce590db 100644
--- a/libc/string/xtensa/strcmp.S
+++ b/libc/string/xtensa/strcmp.S
@@ -108,7 +108,7 @@ ENTRY (strcmp)
 #endif
 .Lretdiff:
 	sub	a2, a8, a9
-	retw
+	abi_ret
 
 /* s1 is word-aligned; s2 is word-aligned.
 
@@ -229,7 +229,7 @@ ENTRY (strcmp)
 
 	/* Words are equal; some byte is zero.  */
 .Leq:	movi	a2, 0		/* return equal */
-	retw
+	abi_ret
 
 .Lwne2:	/* Words are not equal.  On big-endian processors, if none of the
 	   bytes are zero, the return value can be determined by a simple
@@ -239,10 +239,10 @@ ENTRY (strcmp)
 	bnall	a10, a7, .Lsomezero
 	bgeu	a8, a9, .Lposreturn
 	movi	a2, -1
-	retw
+	abi_ret
 .Lposreturn:
 	movi	a2, 1
-	retw
+	abi_ret
 .Lsomezero:	/* There is probably some zero byte. */
 #endif /* __XTENSA_EB__ */
 .Lwne:	/* Words are not equal.  */
@@ -263,14 +263,14 @@ ENTRY (strcmp)
 	   byte.  Just subtract words to get the return value.
 	   The high order equal bytes cancel, leaving room for the sign.  */
 	sub	a2, a8, a9
-	retw
+	abi_ret
 
 .Ldiff0:
 	/* Need to make room for the sign, so can't subtract whole words.  */
 	extui	a10, a8, 24, 8
 	extui	a11, a9, 24, 8
 	sub	a2, a10, a11
-	retw
+	abi_ret
 
 #else /* !__XTENSA_EB__ */
 	/* Little-endian is a little more difficult because can't subtract
@@ -281,28 +281,28 @@ ENTRY (strcmp)
 	extui	a10, a8, 24, 8
 	extui	a11, a9, 24, 8
 	sub	a2, a10, a11
-	retw
+	abi_ret
 
 .Ldiff0:
 	/* Byte 0 is different.  */
 	extui	a10, a8, 0, 8
 	extui	a11, a9, 0, 8
 	sub	a2, a10, a11
-	retw
+	abi_ret
 
 .Ldiff1:
 	/* Byte 0 is equal; byte 1 is different.  */
 	extui	a10, a8, 8, 8
 	extui	a11, a9, 8, 8
 	sub	a2, a10, a11
-	retw
+	abi_ret
 
 .Ldiff2:
 	/* Bytes 0-1 are equal; byte 2 is different.  */
 	extui	a10, a8, 16, 8
 	extui	a11, a9, 16, 8
 	sub	a2, a10, a11
-	retw
+	abi_ret
 
 #endif /* !__XTENSA_EB */
 
diff --git a/libc/string/xtensa/strcpy.S b/libc/string/xtensa/strcpy.S
index f9f586896..9f42b34e6 100644
--- a/libc/string/xtensa/strcpy.S
+++ b/libc/string/xtensa/strcpy.S
@@ -68,7 +68,7 @@ ENTRY (strcpy)
 	s8i	a8, a10, 1	/* store byte 0 */
 	addi	a10, a10, 2	/* advance dst pointer */
 	bnez	a8, .Lsrcaligned
-1:	retw
+1:	abi_ret
 
 
 /* dst is word-aligned; src is word-aligned.  */
@@ -102,21 +102,21 @@ ENTRY (strcpy)
 #endif /* !XCHAL_HAVE_LOOPS */
 
 .Lz3:	/* Byte 3 is zero.  */
-	retw
+	abi_ret
 
 .Lz0:	/* Byte 0 is zero.  */
 #ifdef __XTENSA_EB__
 	movi	a8, 0
 #endif
 	s8i	a8, a10, 0
-	retw
+	abi_ret
 
 .Lz1:	/* Byte 1 is zero.  */
 #ifdef __XTENSA_EB__
         extui   a8, a8, 16, 16
 #endif
 	s16i	a8, a10, 0
-	retw
+	abi_ret
 
 .Lz2:	/* Byte 2 is zero.  */
 #ifdef __XTENSA_EB__
@@ -125,7 +125,7 @@ ENTRY (strcpy)
 	s16i	a8, a10, 0
 	movi	a8, 0
 	s8i	a8, a10, 2
-	retw
+	abi_ret
 
 	.align	4
 	/* (2 mod 4) alignment for loop instruction */
@@ -144,6 +144,6 @@ ENTRY (strcpy)
 #else
 	bnez	a8, 1b
 #endif
-2:	retw
+2:	abi_ret
 
 libc_hidden_def (strcpy)
diff --git a/libc/string/xtensa/strlen.S b/libc/string/xtensa/strlen.S
index 86fe11390..e1c98c8f0 100644
--- a/libc/string/xtensa/strlen.S
+++ b/libc/string/xtensa/strlen.S
@@ -59,7 +59,7 @@ ENTRY (strlen)
 	/* Byte 3 is zero.  */
 	addi	a3, a3, 3	/* point to zero byte */
 	sub	a2, a3, a2	/* subtract to get length */
-	retw
+	abi_ret
 
 
 /* String is word-aligned.  */
@@ -88,16 +88,16 @@ ENTRY (strlen)
 
 .Lz0:	/* Byte 0 is zero.  */
 	sub	a2, a3, a2	/* subtract to get length */
-	retw
+	abi_ret
 
 .Lz1:	/* Byte 1 is zero.  */
 	addi	a3, a3, 1	/* point to zero byte */
 	sub	a2, a3, a2	/* subtract to get length */
-	retw
+	abi_ret
 
 .Lz2:	/* Byte 2 is zero.  */
 	addi	a3, a3, 2	/* point to zero byte */
 	sub	a2, a3, a2	/* subtract to get length */
-	retw
+	abi_ret
 
 libc_hidden_def (strlen)
diff --git a/libc/string/xtensa/strncpy.S b/libc/string/xtensa/strncpy.S
index 115d44cf8..aa8db5da1 100644
--- a/libc/string/xtensa/strncpy.S
+++ b/libc/string/xtensa/strncpy.S
@@ -67,7 +67,7 @@ __strncpy_aux:
 	j	.Lfill
 
 .Lret:
-	retw
+	abi_ret
 
 
 ENTRY (strncpy)
@@ -129,7 +129,7 @@ ENTRY (strncpy)
 	addi	a10, a10, 1
 	bnez    a4, .Lfillcleanup
 
-2:	retw
+2:	abi_ret
 
 .Lfill1mod2: /* dst address is odd */
 	s8i	a9, a10, 0	/* store byte 0 */
@@ -235,6 +235,6 @@ ENTRY (strncpy)
 #endif
 2:	j	.Lfill
 
-3:	retw
+3:	abi_ret
 
 libc_hidden_def (strncpy)