[PATCH 2/2] powerpc/64: Use optimized checksum routines on little-endian
From: Paul Mackerras <hidden>
Date: 2016-11-03 05:15:58
Subsystem:
linux for powerpc (32-bit and 64-bit), the rest · Maintainers:
Madhavan Srinivasan, Linus Torvalds
Currently we have optimized hand-coded assembly checksum routines for big-endian 64-bit systems, but for little-endian we use the generic C routines. This modifies the optimized routines to work for little-endian. With this, we no longer need to enable CONFIG_GENERIC_CSUM. This also fixes a couple of comments in checksum_64.S so they accurately reflect what the associated instruction does. Signed-off-by: Paul Mackerras <redacted> --- arch/powerpc/Kconfig | 2 +- arch/powerpc/include/asm/checksum.h | 4 ++++ arch/powerpc/lib/Makefile | 2 -- arch/powerpc/lib/checksum_64.S | 12 ++++++++++-- 4 files changed, 15 insertions(+), 5 deletions(-)
diff --git a/arch/powerpc/Kconfig b/arch/powerpc/Kconfig
index 65fba4c..514e6dd 100644
--- a/arch/powerpc/Kconfig
+++ b/arch/powerpc/Kconfig@@ -164,7 +164,7 @@ config PPC select HAVE_KERNEL_GZIP config GENERIC_CSUM - def_bool CPU_LITTLE_ENDIAN + def_bool n config EARLY_PRINTK bool
diff --git a/arch/powerpc/include/asm/checksum.h b/arch/powerpc/include/asm/checksum.h
index c16c6f8..5f8297b 100644
--- a/arch/powerpc/include/asm/checksum.h
+++ b/arch/powerpc/include/asm/checksum.h@@ -72,7 +72,11 @@ static inline __wsum csum_tcpudp_nofold(__be32 saddr, __be32 daddr, s += (__force u32)saddr; s += (__force u32)daddr; +#ifdef __BIG_ENDIAN s += proto + len; +#else + s += (proto + len) << 8; +#endif return (__force __wsum) from64to32(s); #else __asm__("\n\
diff --git a/arch/powerpc/lib/Makefile b/arch/powerpc/lib/Makefile
index 309361e8..0e649d7 100644
--- a/arch/powerpc/lib/Makefile
+++ b/arch/powerpc/lib/Makefile@@ -21,9 +21,7 @@ obj64-y += copypage_64.o copyuser_64.o usercopy_64.o mem_64.o hweight_64.o \ obj64-$(CONFIG_SMP) += locks.o obj64-$(CONFIG_ALTIVEC) += vmx-helper.o -ifeq ($(CONFIG_GENERIC_CSUM),) obj-y += checksum_$(BITS).o checksum_wrappers.o -endif obj-$(CONFIG_PPC_EMULATE_SSTEP) += sstep.o ldstfp.o
diff --git a/arch/powerpc/lib/checksum_64.S b/arch/powerpc/lib/checksum_64.S
index fd91766..4dd2761 100644
--- a/arch/powerpc/lib/checksum_64.S
+++ b/arch/powerpc/lib/checksum_64.S@@ -36,7 +36,7 @@ _GLOBAL(__csum_partial) * work to calculate the correct checksum, we ignore that case * and take the potential slowdown of unaligned loads. */ - rldicl. r6,r3,64-1,64-2 /* r6 = (r3 & 0x3) >> 1 */ + rldicl. r6,r3,64-1,64-2 /* r6 = (r3 >> 1) & 0x3 */ beq .Lcsum_aligned li r7,4
@@ -168,8 +168,12 @@ _GLOBAL(__csum_partial) beq .Lcsum_finish lbz r6,0(r3) +#ifdef __BIG_ENDIAN sldi r9,r6,8 /* Pad the byte out to 16 bits */ adde r0,r0,r9 +#else + adde r0,r0,r6 +#endif .Lcsum_finish: addze r0,r0 /* add in final carry */
@@ -236,7 +240,7 @@ _GLOBAL(csum_partial_copy_generic) * If the source and destination are relatively unaligned we only * align the source. This keeps things simple. */ - rldicl. r6,r3,64-1,64-2 /* r6 = (r3 & 0x3) >> 1 */ + rldicl. r6,r3,64-1,64-2 /* r6 = (r3 >> 1) & 0x3 */ beq .Lcopy_aligned li r9,4
@@ -398,8 +402,12 @@ dstnr; sth r6,0(r4) beq .Lcopy_finish srcnr; lbz r6,0(r3) +#ifdef __BIG_ENDIAN sldi r9,r6,8 /* Pad the byte out to 16 bits */ adde r0,r0,r9 +#else + adde r0,r0,r6 +#endif dstnr; stb r6,0(r4) .Lcopy_finish:
--
2.10.1