From: Christophe Leroy <christophe.leroy@xxxxxxxxxx> commit 5486f5bf790b5c664913076c3194b8f916a5c7ad upstream. All functions defined as static inline in net/checksum.h are meant to be inlined for performance reason. But since commit ac7c3e4ff401 ("compiler: enable CONFIG_OPTIMIZE_INLINING forcibly") the compiler is allowed to uninline functions when it wants. Fair enough in the general case, but for tiny performance critical checksum helpers that's counter-productive. The problem mainly arises when selecting CONFIG_CC_OPTIMISE_FOR_SIZE, Those helpers being 'static inline' in header files you suddenly find them duplicated many times in the resulting vmlinux. Here is a typical exemple when building powerpc pmac32_defconfig with CONFIG_CC_OPTIMISE_FOR_SIZE. csum_sub() appears 4 times: c04a23cc <csum_sub>: c04a23cc: 7c 84 20 f8 not r4,r4 c04a23d0: 7c 63 20 14 addc r3,r3,r4 c04a23d4: 7c 63 01 94 addze r3,r3 c04a23d8: 4e 80 00 20 blr ... c04a2ce8: 4b ff f6 e5 bl c04a23cc <csum_sub> ... c04a2d2c: 4b ff f6 a1 bl c04a23cc <csum_sub> ... c04a2d54: 4b ff f6 79 bl c04a23cc <csum_sub> ... c04a754c <csum_sub>: c04a754c: 7c 84 20 f8 not r4,r4 c04a7550: 7c 63 20 14 addc r3,r3,r4 c04a7554: 7c 63 01 94 addze r3,r3 c04a7558: 4e 80 00 20 blr ... c04ac930: 4b ff ac 1d bl c04a754c <csum_sub> ... c04ad264: 4b ff a2 e9 bl c04a754c <csum_sub> ... c04e3b08 <csum_sub>: c04e3b08: 7c 84 20 f8 not r4,r4 c04e3b0c: 7c 63 20 14 addc r3,r3,r4 c04e3b10: 7c 63 01 94 addze r3,r3 c04e3b14: 4e 80 00 20 blr ... c04e5788: 4b ff e3 81 bl c04e3b08 <csum_sub> ... c04e65c8: 4b ff d5 41 bl c04e3b08 <csum_sub> ... c0512d34 <csum_sub>: c0512d34: 7c 84 20 f8 not r4,r4 c0512d38: 7c 63 20 14 addc r3,r3,r4 c0512d3c: 7c 63 01 94 addze r3,r3 c0512d40: 4e 80 00 20 blr ... c0512dfc: 4b ff ff 39 bl c0512d34 <csum_sub> ... c05138bc: 4b ff f4 79 bl c0512d34 <csum_sub> ... Restore the expected behaviour by using __always_inline for all functions defined in net/checksum.h vmlinux size is even reduced by 256 bytes with this patch: text data bss dec hex filename 6980022 2515362 194384 9689768 93daa8 vmlinux.before 6979862 2515266 194384 9689512 93d9a8 vmlinux.now Fixes: ac7c3e4ff401 ("compiler: enable CONFIG_OPTIMIZE_INLINING forcibly") Cc: Masahiro Yamada <yamada.masahiro@xxxxxxxxxxxxx> Cc: Nick Desaulniers <ndesaulniers@xxxxxxxxxx> Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> Signed-off-by: Christophe Leroy <christophe.leroy@xxxxxxxxxx> Signed-off-by: David S. Miller <davem@xxxxxxxxxxxxx> Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx> --- include/net/checksum.h | 41 +++++++++++++++++++++-------------------- 1 file changed, 21 insertions(+), 20 deletions(-) --- a/include/net/checksum.h +++ b/include/net/checksum.h @@ -22,7 +22,7 @@ #include <asm/checksum.h> #ifndef _HAVE_ARCH_COPY_AND_CSUM_FROM_USER -static inline +static __always_inline __wsum csum_and_copy_from_user (const void __user *src, void *dst, int len, __wsum sum, int *err_ptr) { @@ -37,7 +37,7 @@ __wsum csum_and_copy_from_user (const vo #endif #ifndef HAVE_CSUM_COPY_USER -static __inline__ __wsum csum_and_copy_to_user +static __always_inline __wsum csum_and_copy_to_user (const void *src, void __user *dst, int len, __wsum sum, int *err_ptr) { sum = csum_partial(src, len, sum); @@ -54,7 +54,7 @@ static __inline__ __wsum csum_and_copy_t #endif #ifndef HAVE_ARCH_CSUM_ADD -static inline __wsum csum_add(__wsum csum, __wsum addend) +static __always_inline __wsum csum_add(__wsum csum, __wsum addend) { u32 res = (__force u32)csum; res += (__force u32)addend; @@ -62,12 +62,12 @@ static inline __wsum csum_add(__wsum csu } #endif -static inline __wsum csum_sub(__wsum csum, __wsum addend) +static __always_inline __wsum csum_sub(__wsum csum, __wsum addend) { return csum_add(csum, ~addend); } -static inline __sum16 csum16_add(__sum16 csum, __be16 addend) +static __always_inline __sum16 csum16_add(__sum16 csum, __be16 addend) { u16 res = (__force u16)csum; @@ -75,12 +75,12 @@ static inline __sum16 csum16_add(__sum16 return (__force __sum16)(res + (res < (__force u16)addend)); } -static inline __sum16 csum16_sub(__sum16 csum, __be16 addend) +static __always_inline __sum16 csum16_sub(__sum16 csum, __be16 addend) { return csum16_add(csum, ~addend); } -static inline __wsum +static __always_inline __wsum csum_block_add(__wsum csum, __wsum csum2, int offset) { u32 sum = (__force u32)csum2; @@ -92,36 +92,37 @@ csum_block_add(__wsum csum, __wsum csum2 return csum_add(csum, (__force __wsum)sum); } -static inline __wsum +static __always_inline __wsum csum_block_add_ext(__wsum csum, __wsum csum2, int offset, int len) { return csum_block_add(csum, csum2, offset); } -static inline __wsum +static __always_inline __wsum csum_block_sub(__wsum csum, __wsum csum2, int offset) { return csum_block_add(csum, ~csum2, offset); } -static inline __wsum csum_unfold(__sum16 n) +static __always_inline __wsum csum_unfold(__sum16 n) { return (__force __wsum)n; } -static inline __wsum csum_partial_ext(const void *buff, int len, __wsum sum) +static __always_inline +__wsum csum_partial_ext(const void *buff, int len, __wsum sum) { return csum_partial(buff, len, sum); } #define CSUM_MANGLED_0 ((__force __sum16)0xffff) -static inline void csum_replace_by_diff(__sum16 *sum, __wsum diff) +static __always_inline void csum_replace_by_diff(__sum16 *sum, __wsum diff) { *sum = csum_fold(csum_add(diff, ~csum_unfold(*sum))); } -static inline void csum_replace4(__sum16 *sum, __be32 from, __be32 to) +static __always_inline void csum_replace4(__sum16 *sum, __be32 from, __be32 to) { __wsum tmp = csum_sub(~csum_unfold(*sum), (__force __wsum)from); @@ -134,7 +135,7 @@ static inline void csum_replace4(__sum16 * m : old value of a 16bit field * m' : new value of a 16bit field */ -static inline void csum_replace2(__sum16 *sum, __be16 old, __be16 new) +static __always_inline void csum_replace2(__sum16 *sum, __be16 old, __be16 new) { *sum = ~csum16_add(csum16_sub(~(*sum), old), new); } @@ -153,16 +154,16 @@ void inet_proto_csum_replace16(__sum16 * void inet_proto_csum_replace_by_diff(__sum16 *sum, struct sk_buff *skb, __wsum diff, bool pseudohdr); -static inline void inet_proto_csum_replace2(__sum16 *sum, struct sk_buff *skb, - __be16 from, __be16 to, - bool pseudohdr) +static __always_inline +void inet_proto_csum_replace2(__sum16 *sum, struct sk_buff *skb, + __be16 from, __be16 to, bool pseudohdr) { inet_proto_csum_replace4(sum, skb, (__force __be32)from, (__force __be32)to, pseudohdr); } -static inline __wsum remcsum_adjust(void *ptr, __wsum csum, - int start, int offset) +static __always_inline __wsum remcsum_adjust(void *ptr, __wsum csum, + int start, int offset) { __sum16 *psum = (__sum16 *)(ptr + offset); __wsum delta; @@ -178,7 +179,7 @@ static inline __wsum remcsum_adjust(void return delta; } -static inline void remcsum_unadjust(__sum16 *psum, __wsum delta) +static __always_inline void remcsum_unadjust(__sum16 *psum, __wsum delta) { *psum = csum_fold(csum_sub(delta, (__force __wsum)*psum)); } Patches currently in stable-queue which might be from christophe.leroy@xxxxxxxxxx are queue-5.4/net-force-inlining-of-checksum-functions-in-net-checksum.h.patch