Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   Max Filippov <jcmvbkbc@gmail.com>
To:     linux-xtensa@linux-xtensa.org
Cc:     Chris Zankel <chris@zankel.net>, linux-kernel@vger.kernel.org,
        Max Filippov <jcmvbkbc@gmail.com>
Subject: [PATCH 03/10] xtensa: don't use a12 in strncpy_user
Date:   Mon, 18 Oct 2021 22:17:09 -0700
Message-Id: <20211019051716.4173-4-jcmvbkbc@gmail.com>
In-Reply-To: <20211019051716.4173-1-jcmvbkbc@gmail.com>
References: <20211019051716.4173-1-jcmvbkbc@gmail.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

a12 is callee-saved register in xtensa call0 ABI, so a function must not
change it. a10 is not used in this function at all, use it instead of
a12 to avoid saving/restoring it.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
---
 arch/xtensa/lib/strncpy_user.S | 17 ++++++++---------
 1 file changed, 8 insertions(+), 9 deletions(-)

diff --git a/arch/xtensa/lib/strncpy_user.S b/arch/xtensa/lib/strncpy_user.S
index 4faf46fe3f38..0731912227d3 100644
--- a/arch/xtensa/lib/strncpy_user.S
+++ b/arch/xtensa/lib/strncpy_user.S
@@ -45,7 +45,6 @@
 #   a9/ tmp
 #   a10/ tmp
 #   a11/ dst
-#   a12/ tmp
 
 .text
 ENTRY(__strncpy_user)
@@ -61,7 +60,7 @@ ENTRY(__strncpy_user)
 	bbsi.l	a3, 0, .Lsrc1mod2 # if only  8-bit aligned
 	bbsi.l	a3, 1, .Lsrc2mod4 # if only 16-bit aligned
 .Lsrcaligned:	# return here when src is word-aligned
-	srli	a12, a4, 2	# number of loop iterations with 4B per loop
+	srli	a10, a4, 2	# number of loop iterations with 4B per loop
 	movi	a9, 3
 	bnone	a11, a9, .Laligned
 	j	.Ldstunaligned
@@ -102,11 +101,11 @@ EX(10f)	s8i	a9, a11, 0		# store byte 0
 	.byte	0		# (0 mod 4 alignment for LBEG)
 .Laligned:
 #if XCHAL_HAVE_LOOPS
-	loopnez	a12, .Loop1done
+	loopnez	a10, .Loop1done
 #else
-	beqz	a12, .Loop1done
-	slli	a12, a12, 2
-	add	a12, a12, a11	# a12 = end of last 4B chunck
+	beqz	a10, .Loop1done
+	slli	a10, a10, 2
+	add	a10, a10, a11	# a10 = end of last 4B chunck
 #endif
 .Loop1:
 EX(11f)	l32i	a9, a3, 0		# get word from src
@@ -118,7 +117,7 @@ EX(10f)	s32i	a9, a11, 0		# store word to dst
 	bnone	a9, a8, .Lz3		# if byte 3 is zero
 	addi	a11, a11, 4		# advance dst pointer
 #if !XCHAL_HAVE_LOOPS
-	blt	a11, a12, .Loop1
+	blt	a11, a10, .Loop1
 #endif
 
 .Loop1done:
@@ -185,7 +184,7 @@ EX(10f)	s8i	a9, a11, 2
 	loopnez	a4, .Lunalignedend
 #else
 	beqz	a4, .Lunalignedend
-	add	a12, a11, a4		# a12 = ending address
+	add	a10, a11, a4		# a10 = ending address
 #endif /* XCHAL_HAVE_LOOPS */
 .Lnextbyte:
 EX(11f)	l8ui	a9, a3, 0
@@ -194,7 +193,7 @@ EX(10f)	s8i	a9, a11, 0
 	beqz	a9, .Lunalignedend
 	addi	a11, a11, 1
 #if !XCHAL_HAVE_LOOPS
-	blt	a11, a12, .Lnextbyte
+	blt	a11, a10, .Lnextbyte
 #endif
 
 .Lunalignedend:
-- 
2.20.1