kernel: support gcc-optimized inlining on all architectures

Optimized inlining was disabled by default when gcc 4 was still
relatively new. By now, all gcc versions handle this well and there
seems to be no real reason to keep it x86-only.

Signed-off-by: Felix Fietkau <nbd@nbd.name>
(backported from 1e8882585c6f4a5e7f5e2b4f18cd550aafa6f81d)
(rebased patches)
Signed-off-by: Jo-Philipp Wich <jo@mein.io>

5 files changed, 143 insertions, 1 deletions

diff --git a/target/linux/generic/config-4.14 b/target/linux/generic/config-4.14
index 96916c0e43..34773958f5 100644
--- a/target/linux/generic/config-4.14
+++ b/target/linux/generic/config-4.14
@@ -3312,6 +3312,7 @@ CONFIG_NMI_LOG_BUF_SHIFT=13
 # CONFIG_OPROFILE is not set
 # CONFIG_OPROFILE_EVENT_MULTIPLEX is not set
 # CONFIG_OPT3001 is not set
+# CONFIG_OPTIMIZE_INLINING is not set
 # CONFIG_ORANGEFS_FS is not set
 # CONFIG_ORION_WATCHDOG is not set
 # CONFIG_OSF_PARTITION is not set
diff --git a/target/linux/generic/config-4.9 b/target/linux/generic/config-4.9
index 53955d9ae7..5e2b25f31e 100644
--- a/target/linux/generic/config-4.9
+++ b/target/linux/generic/config-4.9
@@ -3117,6 +3117,7 @@ CONFIG_NMI_LOG_BUF_SHIFT=13
 # CONFIG_OPROFILE is not set
 # CONFIG_OPROFILE_EVENT_MULTIPLEX is not set
 # CONFIG_OPT3001 is not set
+# CONFIG_OPTIMIZE_INLINING is not set
 # CONFIG_ORANGEFS_FS is not set
 # CONFIG_ORION_WATCHDOG is not set
 # CONFIG_OSF_PARTITION is not set
diff --git a/target/linux/generic/pending-4.14/220-optimize_inlining.patch b/target/linux/generic/pending-4.14/220-optimize_inlining.patch
new file mode 100644
index 0000000000..dca1dcb204
--- /dev/null
+++ b/target/linux/generic/pending-4.14/220-optimize_inlining.patch
@@ -0,0 +1,70 @@
+--- a/arch/x86/Kconfig.debug
++++ b/arch/x86/Kconfig.debug
+@@ -284,20 +284,6 @@ 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/lib/Kconfig.debug
++++ b/lib/Kconfig.debug
+@@ -137,6 +137,20 @@ endmenu # "printk and dmesg options"
+ 
+ menu "Compile-time checks and compiler options"
+ 
++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_INFO
+ 	bool "Compile the kernel with debug info"
+ 	depends on DEBUG_KERNEL && !COMPILE_TEST
+--- a/arch/x86/Kconfig
++++ b/arch/x86/Kconfig
+@@ -296,9 +296,6 @@ 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-gcc.h
++++ b/include/linux/compiler-gcc.h
+@@ -90,8 +90,7 @@
+  * 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) || (__GNUC__ < 4)
++#if !defined(CONFIG_OPTIMIZE_INLINING) || (__GNUC__ < 4)
+ #define inline \
+ 	inline __attribute__((always_inline, unused)) notrace __gnu_inline
+ #else
diff --git a/target/linux/generic/pending-4.9/220-optimize_inlining.patch b/target/linux/generic/pending-4.9/220-optimize_inlining.patch
new file mode 100644
index 0000000000..75c04ea15c
--- /dev/null
+++ b/target/linux/generic/pending-4.9/220-optimize_inlining.patch
@@ -0,0 +1,70 @@
+--- a/arch/x86/Kconfig.debug
++++ b/arch/x86/Kconfig.debug
+@@ -287,20 +287,6 @@ 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/lib/Kconfig.debug
++++ b/lib/Kconfig.debug
+@@ -117,6 +117,20 @@ endmenu # "printk and dmesg options"
+ 
+ menu "Compile-time checks and compiler options"
+ 
++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_INFO
+ 	bool "Compile the kernel with debug info"
+ 	depends on DEBUG_KERNEL && !COMPILE_TEST
+--- a/arch/x86/Kconfig
++++ b/arch/x86/Kconfig
+@@ -273,9 +273,6 @@ 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-gcc.h
++++ b/include/linux/compiler-gcc.h
+@@ -89,8 +89,7 @@
+  * 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) || (__GNUC__ < 4)
++#if !defined(CONFIG_OPTIMIZE_INLINING) || (__GNUC__ < 4)
+ #define inline \
+ 	inline __attribute__((always_inline, unused)) notrace __gnu_inline
+ #else
diff --git a/target/linux/generic/pending-4.9/890-uart_optional_sysrq.patch b/target/linux/generic/pending-4.9/890-uart_optional_sysrq.patch
index ff0824d343..d80246db99 100644
--- a/target/linux/generic/pending-4.9/890-uart_optional_sysrq.patch
+++ b/target/linux/generic/pending-4.9/890-uart_optional_sysrq.patch
@@ -26,7 +26,7 @@ Signed-off-by: Felix Fietkau <nbd@nbd.name>
  {
 --- a/lib/Kconfig.debug
 +++ b/lib/Kconfig.debug
-@@ -396,6 +396,11 @@ config MAGIC_SYSRQ_DEFAULT_ENABLE
+@@ -410,6 +410,11 @@ config MAGIC_SYSRQ_DEFAULT_ENABLE
  	  This may be set to 1 or 0 to enable or disable them all, or
  	  to a bitmask as described in Documentation/sysrq.txt.