Skip to contentSkip to conversation

arm64: mm: Fix operands of clz in __flush_dcache_all

PATCH New Anup Patel 2 participants 2 messages May 14, 2013 arm64

lore
Share & tools
lore

Anup Patel

author
#0 lore raw

PATCH 1/1 arm64: mm: Fix operands of clz in __flush_dcache_all

From: Sukanto Ghosh <[email protected]>

The format of the lower 32-bits of the 64-bit operand to 'dc cisw' is unchanged from ARMv7 architecture and the upper bits are RES0. This implies that the 'way' field of the operand of 'dc cisw' occupies the bit-positions [31 .. (32-A)]. Due to the use of 64-bit extended operands to 'clz', the existing implementation of __flush_dcache_all is incorrectly placing the 'way' field in the bit-positions [63 .. (64-A)].

---
 arch/arm64/mm/cache.S |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/arm64/mm/cache.S +1 -1
index abe69b8..48a3860 100644
@@ -52,7 +52,7 @@ loop1:
 	add	x2, x2, #4			// add 4 (line length offset)
 	mov	x4, #0x3ff
 	and	x4, x4, x1, lsr #3		// find maximum number on the way size
-	clz	x5, x4				// find bit position of way size increment
+	clz	w5, w4				// find bit position of way size increment
 	mov	x7, #0x7fff
 	and	x7, x7, x1, lsr #13		// extract max number of the index size
 loop2:
Signature · 1 line
1.7.9.5
maintainer
#1 lore raw

PATCH 1/1 arm64: mm: Fix operands of clz in __flush_dcache_all

Quoted 20 lines · Anup Patel

From: Sukanto Ghosh <[email protected]>

The format of the lower 32-bits of the 64-bit operand to 'dc cisw' is unchanged from ARMv7 architecture and the upper bits are RES0. This implies that the 'way' field of the operand of 'dc cisw' occupies the bit-positions [31 .. (32-A)]. Due to the use of 64-bit extended operands to 'clz', the existing implementation of __flush_dcache_all is incorrectly placing the 'way' field in the bit-positions [63 .. (64-A)].

---
 arch/arm64/mm/cache.S |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/arm64/mm/cache.S +0 -0
index abe69b8..48a3860 100644

arch/arm64/mm/cache.S loop1: · line 58

Quoted 8 lines
arch/arm64/mm/cache.S +1 -1
@@ -52,7 +52,7 @@ loop1:
 	add	x2, x2, #4			// add 4 (line length offset)
 	mov	x4, #0x3ff
 	and	x4, x4, x1, lsr #3		// find maximum number on the way size
-	clz	x5, x4				// find bit position of way size increment
+	clz	w5, w4				// find bit position of way size increment
 	mov	x7, #0x7fff
 	and	x7, x7, x1, lsr #13		// extract max number of the index size
 loop2:

Good catch. I'll queue it for 3.10-rc2.

Thanks.

Signature · 1 line
Catalin