openwrt/target/linux/generic/pending-4.19/220-optimize_inlining.patch
Hauke Mehrtens 6dac1c0a9b kernel: Activate CONFIG_OPTIMIZE_INLINING
This will reduce the size of the kernel if CONFIG_CC_OPTIMIZE_FOR_SIZE is
set like for all targets with small_flash feature flag.
I haven't seen any changes for an ARM64 target which optimizes the
kernel for speed instead.

On the ath79/tiny target the uncompressed kernel size was reduced by
3.2% and the compressed kernel size by 2.1%

kernel size with CONFIG_OPTIMIZE_INLINING=n
4346412 build_dir/target-mips_24kc_musl/linux-ath79_tiny/vmlinux
1391169 build_dir/target-mips_24kc_musl/linux-ath79_tiny/tplink_tl-wr941-v4-kernel.bin

Kernel size with CONFIG_OPTIMIZE_INLINING=y
4212396 build_dir/target-mips_24kc_musl/linux-ath79_tiny/vmlinux
1362051 build_dir/target-mips_24kc_musl/linux-ath79_tiny/tplink_tl-wr941-v4-kernel.bin

This change is currently pending for kernel 5.2 and already in
linux-next, this updates our patch to match the upstream version.

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
2019-06-16 18:40:02 +02:00

258 lines
8.6 KiB
Diff

--- a/arch/arm/include/asm/hardirq.h
+++ b/arch/arm/include/asm/hardirq.h
@@ -6,6 +6,7 @@
#include <linux/threads.h>
#include <asm/irq.h>
+/* number of IPIS _not_ including IPI_CPU_BACKTRACE */
#define NR_IPI 7
typedef struct {
--- a/arch/arm/kernel/atags.h
+++ b/arch/arm/kernel/atags.h
@@ -5,7 +5,7 @@ void convert_to_tag_list(struct tag *tag
const struct machine_desc *setup_machine_tags(phys_addr_t __atags_pointer,
unsigned int machine_nr);
#else
-static inline const struct machine_desc *
+static inline const struct machine_desc * __init __noreturn
setup_machine_tags(phys_addr_t __atags_pointer, unsigned int machine_nr)
{
early_print("no ATAGS support: can't continue\n");
--- a/arch/arm/kernel/smp.c
+++ b/arch/arm/kernel/smp.c
@@ -76,6 +76,10 @@ enum ipi_msg_type {
IPI_CPU_STOP,
IPI_IRQ_WORK,
IPI_COMPLETION,
+ /*
+ * CPU_BACKTRACE is special and not included in NR_IPI
+ * or tracable with trace_ipi_*
+ */
IPI_CPU_BACKTRACE,
/*
* SGI8-15 can be reserved by secure firmware, and thus may
@@ -803,7 +807,7 @@ core_initcall(register_cpufreq_notifier)
static void raise_nmi(cpumask_t *mask)
{
- smp_cross_call(mask, IPI_CPU_BACKTRACE);
+ __smp_cross_call(mask, IPI_CPU_BACKTRACE);
}
void arch_trigger_cpumask_backtrace(const cpumask_t *mask, bool exclude_self)
--- a/arch/arm64/include/asm/cpufeature.h
+++ b/arch/arm64/include/asm/cpufeature.h
@@ -365,7 +365,7 @@ static inline bool cpu_have_feature(unsi
}
/* System capability check for constant caps */
-static inline bool __cpus_have_const_cap(int num)
+static __always_inline bool __cpus_have_const_cap(int num)
{
if (num >= ARM64_NCAPS)
return false;
@@ -379,7 +379,7 @@ static inline bool cpus_have_cap(unsigne
return test_bit(num, cpu_hwcaps);
}
-static inline bool cpus_have_const_cap(int num)
+static __always_inline bool cpus_have_const_cap(int num)
{
if (static_branch_likely(&arm64_const_caps_ready))
return __cpus_have_const_cap(num);
--- a/arch/mips/include/asm/bitops.h
+++ b/arch/mips/include/asm/bitops.h
@@ -463,7 +463,7 @@ static inline void __clear_bit_unlock(un
* Return the bit position (0..63) of the most significant 1 bit in a word
* Returns -1 if no 1 bit exists
*/
-static inline unsigned long __fls(unsigned long word)
+static __always_inline unsigned long __fls(unsigned long word)
{
int num;
@@ -529,7 +529,7 @@ static inline unsigned long __fls(unsign
* Returns 0..SZLONG-1
* Undefined if no bit exists, so code should check against 0 first.
*/
-static inline unsigned long __ffs(unsigned long word)
+static __always_inline unsigned long __ffs(unsigned long word)
{
return __fls(word & -word);
}
--- a/arch/mips/kernel/cpu-bugs64.c
+++ b/arch/mips/kernel/cpu-bugs64.c
@@ -42,8 +42,8 @@ static inline void align_mod(const int a
: GCC_IMM_ASM() (align), GCC_IMM_ASM() (mod));
}
-static inline void mult_sh_align_mod(long *v1, long *v2, long *w,
- const int align, const int mod)
+static __always_inline void mult_sh_align_mod(long *v1, long *v2, long *w,
+ const int align, const int mod)
{
unsigned long flags;
int m1, m2;
--- a/arch/powerpc/kernel/prom_init.c
+++ b/arch/powerpc/kernel/prom_init.c
@@ -498,14 +498,14 @@ static int __init prom_next_node(phandle
}
}
-static inline int prom_getprop(phandle node, const char *pname,
- void *value, size_t valuelen)
+static inline int __init prom_getprop(phandle node, const char *pname,
+ void *value, size_t valuelen)
{
return call_prom("getprop", 4, 1, node, ADDR(pname),
(u32)(unsigned long) value, (u32) valuelen);
}
-static inline int prom_getproplen(phandle node, const char *pname)
+static inline int __init prom_getproplen(phandle node, const char *pname)
{
return call_prom("getproplen", 2, 1, node, ADDR(pname));
}
--- a/arch/powerpc/mm/tlb-radix.c
+++ b/arch/powerpc/mm/tlb-radix.c
@@ -90,8 +90,8 @@ void radix__tlbiel_all(unsigned int acti
asm volatile(PPC_INVALIDATE_ERAT "; isync" : : :"memory");
}
-static inline void __tlbiel_pid(unsigned long pid, int set,
- unsigned long ric)
+static __always_inline void __tlbiel_pid(unsigned long pid, int set,
+ unsigned long ric)
{
unsigned long rb,rs,prs,r;
@@ -106,7 +106,7 @@ static inline void __tlbiel_pid(unsigned
trace_tlbie(0, 1, rb, rs, ric, prs, r);
}
-static inline void __tlbie_pid(unsigned long pid, unsigned long ric)
+static __always_inline void __tlbie_pid(unsigned long pid, unsigned long ric)
{
unsigned long rb,rs,prs,r;
@@ -136,7 +136,7 @@ static inline void __tlbiel_lpid(unsigne
trace_tlbie(lpid, 1, rb, rs, ric, prs, r);
}
-static inline void __tlbie_lpid(unsigned long lpid, unsigned long ric)
+static __always_inline void __tlbie_lpid(unsigned long lpid, unsigned long ric)
{
unsigned long rb,rs,prs,r;
@@ -239,7 +239,7 @@ static inline void fixup_tlbie_lpid(unsi
/*
* We use 128 set in radix mode and 256 set in hpt mode.
*/
-static inline void _tlbiel_pid(unsigned long pid, unsigned long ric)
+static __always_inline void _tlbiel_pid(unsigned long pid, unsigned long ric)
{
int set;
@@ -918,7 +918,7 @@ void radix__tlb_flush(struct mmu_gather
tlb->need_flush_all = 0;
}
-static inline void __radix__flush_tlb_range_psize(struct mm_struct *mm,
+static __always_inline void __radix__flush_tlb_range_psize(struct mm_struct *mm,
unsigned long start, unsigned long end,
int psize, bool also_pwc)
{
--- a/arch/s390/include/asm/cpacf.h
+++ b/arch/s390/include/asm/cpacf.h
@@ -202,7 +202,7 @@ static inline int __cpacf_check_opcode(u
}
}
-static inline int cpacf_query(unsigned int opcode, cpacf_mask_t *mask)
+static __always_inline int cpacf_query(unsigned int opcode, cpacf_mask_t *mask)
{
if (__cpacf_check_opcode(opcode)) {
__cpacf_query(opcode, mask);
--- a/arch/x86/Kconfig.debug
+++ b/arch/x86/Kconfig.debug
@@ -276,20 +276,6 @@ config CPA_DEBUG
---help---
Do change_page_attr() self-tests every 30 seconds.
-config OPTIMIZE_INLINING
- bool "Allow gcc to uninline functions marked 'inline'"
- ---help---
- This option determines if the kernel forces gcc to inline the functions
- developers have marked 'inline'. Doing so takes away freedom from gcc to
- do what it thinks is best, which is desirable for the gcc 3.x series of
- compilers. The gcc 4.x series have a rewritten inlining algorithm and
- enabling this option will generate a smaller kernel there. Hopefully
- this algorithm is so good that allowing gcc 4.x and above to make the
- decision will become the default in the future. Until then this option
- is there to test gcc for this.
-
- If unsure, say N.
-
config DEBUG_ENTRY
bool "Debug low-level entry code"
depends on DEBUG_KERNEL
--- a/drivers/mtd/nand/raw/vf610_nfc.c
+++ b/drivers/mtd/nand/raw/vf610_nfc.c
@@ -373,7 +373,7 @@ static int vf610_nfc_cmd(struct nand_chi
{
const struct nand_op_instr *instr;
struct vf610_nfc *nfc = chip_to_nfc(chip);
- int op_id = -1, trfr_sz = 0, offset;
+ int op_id = -1, trfr_sz = 0, offset = 0;
u32 col = 0, row = 0, cmd1 = 0, cmd2 = 0, code = 0;
bool force8bit = false;
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -309,6 +309,20 @@ config HEADERS_CHECK
exported to $(INSTALL_HDR_PATH) (usually 'usr/include' in
your build tree), to make sure they're suitable.
+config OPTIMIZE_INLINING
+ bool "Allow compiler to uninline functions marked 'inline'"
+ help
+ This option determines if the kernel forces gcc to inline the functions
+ developers have marked 'inline'. Doing so takes away freedom from gcc to
+ do what it thinks is best, which is desirable for the gcc 3.x series of
+ compilers. The gcc 4.x series have a rewritten inlining algorithm and
+ enabling this option will generate a smaller kernel there. Hopefully
+ this algorithm is so good that allowing gcc 4.x and above to make the
+ decision will become the default in the future. Until then this option
+ is there to test gcc for this.
+
+ If unsure, say N.
+
config DEBUG_SECTION_MISMATCH
bool "Enable full Section mismatch analysis"
help
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -306,9 +306,6 @@ config ZONE_DMA32
config AUDIT_ARCH
def_bool y if X86_64
-config ARCH_SUPPORTS_OPTIMIZED_INLINING
- def_bool y
-
config ARCH_SUPPORTS_DEBUG_PAGEALLOC
def_bool y
--- a/include/linux/compiler_types.h
+++ b/include/linux/compiler_types.h
@@ -268,8 +268,7 @@ struct ftrace_likely_data {
* of extern inline functions at link time.
* A lot of inline functions can cause havoc with function tracing.
*/
-#if !defined(CONFIG_ARCH_SUPPORTS_OPTIMIZED_INLINING) || \
- !defined(CONFIG_OPTIMIZE_INLINING)
+#if !defined(CONFIG_OPTIMIZE_INLINING)
#define inline \
inline __attribute__((always_inline, unused)) notrace __gnu_inline
#else