2008-11-26 18:51:37

by Harvey Harrison

[permalink] [raw]
Subject: [PATCH] sh: move to the asm-generic version

Add the ability to override pieces of the unaligned API from
an arch.

Signed-off-by: Harvey Harrison <[email protected]>
---
Paul, I noticed you added specific support for the sh4a recently, I've got some
patches in -mm that unify the unaligned access handling. To avoid messing this up
could I get this patch in through akpm, there's no need to change what's in your tree
now, but when the asm-generic version gets in tree, this is how I'd like SH
to get wired up.

Patch on top of next-20081126 w/ unaligned patches from -mm applied already.

Akpm, this is the necessary pieces to fixup the -mm patch:
unaligned-use-generic-implementation-on-packed-struct-arches.patch
which will conflict heavily with the recently added sh4a support Paul added,
just drop the SH parts from that patch, and add this patch.

arch/sh/include/asm/unaligned-sh4a.h | 258 ----------------------------------
arch/sh/include/asm/unaligned.h | 81 +++++++++--
include/asm-generic/unaligned.h | 48 +++++--
3 files changed, 103 insertions(+), 284 deletions(-)

diff --git a/arch/sh/include/asm/unaligned-sh4a.h b/arch/sh/include/asm/unaligned-sh4a.h
deleted file mode 100644
index d8f8977..0000000
--- a/arch/sh/include/asm/unaligned-sh4a.h
+++ /dev/null
@@ -1,258 +0,0 @@
-#ifndef __ASM_SH_UNALIGNED_SH4A_H
-#define __ASM_SH_UNALIGNED_SH4A_H
-
-/*
- * SH-4A has support for unaligned 32-bit loads, and 32-bit loads only.
- * Support for 16 and 64-bit accesses are done through shifting and
- * masking relative to the endianness. Unaligned stores are not supported
- * by the instruction encoding, so these continue to use the packed
- * struct.
- *
- * The same note as with the movli.l/movco.l pair applies here, as long
- * as the load is gauranteed to be inlined, nothing else will hook in to
- * r0 and we get the return value for free.
- *
- * NOTE: Due to the fact we require r0 encoding, care should be taken to
- * avoid mixing these heavily with other r0 consumers, such as the atomic
- * ops. Failure to adhere to this can result in the compiler running out
- * of spill registers and blowing up when building at low optimization
- * levels. See http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34777.
- */
-#include <linux/types.h>
-#include <asm/byteorder.h>
-
-static __always_inline u32 __get_unaligned_cpu32(const u8 *p)
-{
- unsigned long unaligned;
-
- __asm__ __volatile__ (
- "movua.l @%1, %0\n\t"
- : "=z" (unaligned)
- : "r" (p)
- );
-
- return unaligned;
-}
-
-struct __una_u16 { u16 x __attribute__((packed)); };
-struct __una_u32 { u32 x __attribute__((packed)); };
-struct __una_u64 { u64 x __attribute__((packed)); };
-
-static inline u16 __get_unaligned_cpu16(const u8 *p)
-{
-#ifdef __LITTLE_ENDIAN
- return __get_unaligned_cpu32(p) & 0xffff;
-#else
- return __get_unaligned_cpu32(p) >> 16;
-#endif
-}
-
-/*
- * Even though movua.l supports auto-increment on the read side, it can
- * only store to r0 due to instruction encoding constraints, so just let
- * the compiler sort it out on its own.
- */
-static inline u64 __get_unaligned_cpu64(const u8 *p)
-{
-#ifdef __LITTLE_ENDIAN
- return (u64)__get_unaligned_cpu32(p + 4) << 32 |
- __get_unaligned_cpu32(p);
-#else
- return (u64)__get_unaligned_cpu32(p) << 32 |
- __get_unaligned_cpu32(p + 4);
-#endif
-}
-
-static inline u16 get_unaligned_le16(const void *p)
-{
- return le16_to_cpu(__get_unaligned_cpu16(p));
-}
-
-static inline u32 get_unaligned_le32(const void *p)
-{
- return le32_to_cpu(__get_unaligned_cpu32(p));
-}
-
-static inline u64 get_unaligned_le64(const void *p)
-{
- return le64_to_cpu(__get_unaligned_cpu64(p));
-}
-
-static inline u16 get_unaligned_be16(const void *p)
-{
- return be16_to_cpu(__get_unaligned_cpu16(p));
-}
-
-static inline u32 get_unaligned_be32(const void *p)
-{
- return be32_to_cpu(__get_unaligned_cpu32(p));
-}
-
-static inline u64 get_unaligned_be64(const void *p)
-{
- return be64_to_cpu(__get_unaligned_cpu64(p));
-}
-
-static inline void __put_le16_noalign(u8 *p, u16 val)
-{
- *p++ = val;
- *p++ = val >> 8;
-}
-
-static inline void __put_le32_noalign(u8 *p, u32 val)
-{
- __put_le16_noalign(p, val);
- __put_le16_noalign(p + 2, val >> 16);
-}
-
-static inline void __put_le64_noalign(u8 *p, u64 val)
-{
- __put_le32_noalign(p, val);
- __put_le32_noalign(p + 4, val >> 32);
-}
-
-static inline void __put_be16_noalign(u8 *p, u16 val)
-{
- *p++ = val >> 8;
- *p++ = val;
-}
-
-static inline void __put_be32_noalign(u8 *p, u32 val)
-{
- __put_be16_noalign(p, val >> 16);
- __put_be16_noalign(p + 2, val);
-}
-
-static inline void __put_be64_noalign(u8 *p, u64 val)
-{
- __put_be32_noalign(p, val >> 32);
- __put_be32_noalign(p + 4, val);
-}
-
-static inline void put_unaligned_le16(u16 val, void *p)
-{
-#ifdef __LITTLE_ENDIAN
- ((struct __una_u16 *)p)->x = val;
-#else
- __put_le16_noalign(p, val);
-#endif
-}
-
-static inline void put_unaligned_le32(u32 val, void *p)
-{
-#ifdef __LITTLE_ENDIAN
- ((struct __una_u32 *)p)->x = val;
-#else
- __put_le32_noalign(p, val);
-#endif
-}
-
-static inline void put_unaligned_le64(u64 val, void *p)
-{
-#ifdef __LITTLE_ENDIAN
- ((struct __una_u64 *)p)->x = val;
-#else
- __put_le64_noalign(p, val);
-#endif
-}
-
-static inline void put_unaligned_be16(u16 val, void *p)
-{
-#ifdef __BIG_ENDIAN
- ((struct __una_u16 *)p)->x = val;
-#else
- __put_be16_noalign(p, val);
-#endif
-}
-
-static inline void put_unaligned_be32(u32 val, void *p)
-{
-#ifdef __BIG_ENDIAN
- ((struct __una_u32 *)p)->x = val;
-#else
- __put_be32_noalign(p, val);
-#endif
-}
-
-static inline void put_unaligned_be64(u64 val, void *p)
-{
-#ifdef __BIG_ENDIAN
- ((struct __una_u64 *)p)->x = val;
-#else
- __put_be64_noalign(p, val);
-#endif
-}
-
-/*
- * Cause a link-time error if we try an unaligned access other than
- * 1,2,4 or 8 bytes long
- */
-extern void __bad_unaligned_access_size(void);
-
-#define __get_unaligned_le(ptr) ((__force typeof(*(ptr)))({ \
- __builtin_choose_expr(sizeof(*(ptr)) == 1, *(ptr), \
- __builtin_choose_expr(sizeof(*(ptr)) == 2, get_unaligned_le16((ptr)), \
- __builtin_choose_expr(sizeof(*(ptr)) == 4, get_unaligned_le32((ptr)), \
- __builtin_choose_expr(sizeof(*(ptr)) == 8, get_unaligned_le64((ptr)), \
- __bad_unaligned_access_size())))); \
- }))
-
-#define __get_unaligned_be(ptr) ((__force typeof(*(ptr)))({ \
- __builtin_choose_expr(sizeof(*(ptr)) == 1, *(ptr), \
- __builtin_choose_expr(sizeof(*(ptr)) == 2, get_unaligned_be16((ptr)), \
- __builtin_choose_expr(sizeof(*(ptr)) == 4, get_unaligned_be32((ptr)), \
- __builtin_choose_expr(sizeof(*(ptr)) == 8, get_unaligned_be64((ptr)), \
- __bad_unaligned_access_size())))); \
- }))
-
-#define __put_unaligned_le(val, ptr) ({ \
- void *__gu_p = (ptr); \
- switch (sizeof(*(ptr))) { \
- case 1: \
- *(u8 *)__gu_p = (__force u8)(val); \
- break; \
- case 2: \
- put_unaligned_le16((__force u16)(val), __gu_p); \
- break; \
- case 4: \
- put_unaligned_le32((__force u32)(val), __gu_p); \
- break; \
- case 8: \
- put_unaligned_le64((__force u64)(val), __gu_p); \
- break; \
- default: \
- __bad_unaligned_access_size(); \
- break; \
- } \
- (void)0; })
-
-#define __put_unaligned_be(val, ptr) ({ \
- void *__gu_p = (ptr); \
- switch (sizeof(*(ptr))) { \
- case 1: \
- *(u8 *)__gu_p = (__force u8)(val); \
- break; \
- case 2: \
- put_unaligned_be16((__force u16)(val), __gu_p); \
- break; \
- case 4: \
- put_unaligned_be32((__force u32)(val), __gu_p); \
- break; \
- case 8: \
- put_unaligned_be64((__force u64)(val), __gu_p); \
- break; \
- default: \
- __bad_unaligned_access_size(); \
- break; \
- } \
- (void)0; })
-
-#ifdef __LITTLE_ENDIAN
-# define get_unaligned __get_unaligned_le
-# define put_unaligned __put_unaligned_le
-#else
-# define get_unaligned __get_unaligned_be
-# define put_unaligned __put_unaligned_be
-#endif
-
-#endif /* __ASM_SH_UNALIGNED_SH4A_H */
diff --git a/arch/sh/include/asm/unaligned.h b/arch/sh/include/asm/unaligned.h
index 8c0ad5e..b97ff74 100644
--- a/arch/sh/include/asm/unaligned.h
+++ b/arch/sh/include/asm/unaligned.h
@@ -1,24 +1,77 @@
#ifndef _ASM_SH_UNALIGNED_H
#define _ASM_SH_UNALIGNED_H

+#include <linux/types.h>
+#include <asm/byteorder.h>
+
#ifdef CONFIG_CPU_SH4A
-/* SH-4A can handle unaligned loads in a relatively neutered fashion. */
-#include <asm/unaligned-sh4a.h>
+/*
+ * SH-4A has support for unaligned 32-bit loads, and 32-bit loads only.
+ * Support for 16 and 64-bit accesses are done through shifting and
+ * masking relative to the endianness. Unaligned stores are not supported
+ * by the instruction encoding, so these continue to use the packed
+ * struct.
+ *
+ * The same note as with the movli.l/movco.l pair applies here, as long
+ * as the load is gauranteed to be inlined, nothing else will hook in to
+ * r0 and we get the return value for free.
+ *
+ * NOTE: Due to the fact we require r0 encoding, care should be taken to
+ * avoid mixing these heavily with other r0 consumers, such as the atomic
+ * ops. Failure to adhere to this can result in the compiler running out
+ * of spill registers and blowing up when building at low optimization
+ * levels. See http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34777.
+ */
+static inline u32 __sh4a_unaligned32(const u8 *p)
+{
+ unsigned long unaligned;
+
+ __asm__ __volatile__ (
+ "movua.l @%1, %0\n\t"
+ : "=z" (unaligned)
+ : "r" (p)
+ );
+
+ return unaligned;
+}
+
+static inline u16 __sh4a_unaligned16(const u8 *p)
+{
+#ifdef __LITTLE_ENDIAN
+ return __sh4a_unaligned32(p) & 0xffff;
#else
-/* Otherwise, SH can't handle unaligned accesses. */
-#ifdef __LITTLE_ENDIAN__
-# include <linux/unaligned/le_struct.h>
-# include <linux/unaligned/be_byteshift.h>
-# include <linux/unaligned/generic.h>
-# define get_unaligned __get_unaligned_le
-# define put_unaligned __put_unaligned_le
+ return __sh4a_unaligned32(p) >> 16;
+#endif
+}
+
+/*
+ * Even though movua.l supports auto-increment on the read side, it can
+ * only store to r0 due to instruction encoding constraints, so just let
+ * the compiler sort it out on its own.
+ */
+static inline u64 __sh4a_unaligned64(const u8 *p)
+{
+#ifdef __LITTLE_ENDIAN
+ return (u64)__sh4a_unaligned32(p + 4) << 32 |
+ __sh4a_unaligned32(p);
#else
-# include <linux/unaligned/be_struct.h>
-# include <linux/unaligned/le_byteshift.h>
-# include <linux/unaligned/generic.h>
-# define get_unaligned __get_unaligned_be
-# define put_unaligned __put_unaligned_be
+ return (u64)__sh4a_unaligned32(p) << 32 |
+ __sh4a_unaligned32(p + 4);
#endif
+}
+
+#ifdef __LITTLE_ENDIAN
+# define __arch_get_le16_noalign __sh4a_unaligned16
+# define __arch_get_le32_noalign __sh4a_unaligned32
+# define __arch_get_le64_noalign __sh4a_unaligned64
+#else
+# define __arch_get_be16_noalign __sh4a_unaligned16
+# define __arch_get_be32_noalign __sh4a_unaligned32
+# define __arch_get_be64_noalign __sh4a_unaligned64
#endif

+#endif /* CONFIG_CPU_SH4A */
+
+#include <asm-generic/unaligned.h>
+
#endif /* _ASM_SH_UNALIGNED_H */
diff --git a/include/asm-generic/unaligned.h b/include/asm-generic/unaligned.h
index 5a055f7..642e157 100644
--- a/include/asm-generic/unaligned.h
+++ b/include/asm-generic/unaligned.h
@@ -118,55 +118,79 @@ static inline u64 __get_be64_noalign(const u8 *p)

static inline u16 load_le16_noalign(const __le16 *p)
{
-#ifdef __LITTLE_ENDIAN
- return ((__force const struct __una_u16 *)p)->x;
+#ifdef __arch_load_le16_noalign
+ return __arch_get_le16_noalign((void *)p);
#else
+# ifdef __LITTLE_ENDIAN
+ return ((__force const struct __una_u16 *)p)->x;
+# else
return __get_le16_noalign((__force const u8 *)p);
+# endif
#endif
}

static inline u32 load_le32_noalign(const __le32 *p)
{
-#ifdef __LITTLE_ENDIAN
- return ((__force const struct __una_u32 *)p)->x;
+#ifdef __arch_load_le32_noalign
+ return __arch_get_le32_noalign((void *)p);
#else
+# ifdef __LITTLE_ENDIAN
+ return ((__force const struct __una_u32 *)p)->x;
+# else
return __get_le32_noalign((__force const u8 *)p);
+# endif
#endif
}

static inline u64 load_le64_noalign(const __le64 *p)
{
-#ifdef __LITTLE_ENDIAN
- return ((__force const struct __una_u64 *)p)->x;
+#ifdef __arch_load_le64_noalign
+ return __arch_get_le64_noalign((void *)p);
#else
+# ifdef __LITTLE_ENDIAN
+ return ((__force const struct __una_u64 *)p)->x;
+# else
return __get_le64_noalign((__force const u8 *)p);
+# endif
#endif
}

static inline u16 load_be16_noalign(const __be16 *p)
{
-#ifdef __BIG_ENDIAN
- return ((__force const struct __una_u16 *)p)->x;
+#ifdef __arch_load_be16_noalign
+ return __arch_get_be16_noalign((void *)p);
#else
+# ifdef __BIG_ENDIAN
+ return ((__force const struct __una_u16 *)p)->x;
+# else
return __get_be16_noalign((__force const u8 *)p);
+# endif
#endif
}

static inline u32 load_be32_noalign(const __be32 *p)
{
-#ifdef __BIG_ENDIAN
- return ((__force const struct __una_u32 *)p)->x;
+#ifdef __arch_load_be32_noalign
+ return __arch_get_be32_noalign((void *)p);
#else
+# ifdef __BIG_ENDIAN
+ return ((__force const struct __una_u32 *)p)->x;
+# else
return __get_be32_noalign((__force const u8 *)p);
+# endif
#endif
}

static inline u64 load_be64_noalign(const __be64 *p)
{
-#ifdef __BIG_ENDIAN
- return ((__force const struct __una_u64 *)p)->x;
+#ifdef __arch_load_be64_noalign
+ return __arch_get_be64_noalign((void *)p);
#else
+# ifdef __BIG_ENDIAN
+ return ((__force const struct __una_u64 *)p)->x;
+# else
return __get_be64_noalign((__force const u8 *)p);
+# endif
#endif
}

--
1.6.0.4.1044.g77718