summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorTejun Heo <tj@kernel.org>2011-01-25 14:26:50 +0100
committerTejun Heo <tj@kernel.org>2011-01-25 14:26:50 +0100
commit19df0c2fef010e94e90df514aaf4e73f6b80145c (patch)
tree4b0b9c10622aead0d8b658cca6c49090149a91a8
parentc723fdab8aa728dc2bf0da6a0de8bb9c3f588d84 (diff)
downloadlinux-stable-19df0c2fef010e94e90df514aaf4e73f6b80145c.tar.gz
linux-stable-19df0c2fef010e94e90df514aaf4e73f6b80145c.tar.bz2
linux-stable-19df0c2fef010e94e90df514aaf4e73f6b80145c.zip
percpu: align percpu readmostly subsection to cacheline
Currently percpu readmostly subsection may share cachelines with other percpu subsections which may result in unnecessary cacheline bounce and performance degradation. This patch adds @cacheline parameter to PERCPU() and PERCPU_VADDR() linker macros, makes each arch linker scripts specify its cacheline size and use it to align percpu subsections. This is based on Shaohua's x86 only patch. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Shaohua Li <shaohua.li@intel.com>
-rw-r--r--arch/alpha/kernel/vmlinux.lds.S2
-rw-r--r--arch/arm/kernel/vmlinux.lds.S2
-rw-r--r--arch/blackfin/kernel/vmlinux.lds.S2
-rw-r--r--arch/cris/kernel/vmlinux.lds.S2
-rw-r--r--arch/frv/kernel/vmlinux.lds.S2
-rw-r--r--arch/ia64/kernel/vmlinux.lds.S2
-rw-r--r--arch/m32r/kernel/vmlinux.lds.S2
-rw-r--r--arch/mips/kernel/vmlinux.lds.S2
-rw-r--r--arch/mn10300/kernel/vmlinux.lds.S2
-rw-r--r--arch/parisc/kernel/vmlinux.lds.S2
-rw-r--r--arch/powerpc/kernel/vmlinux.lds.S2
-rw-r--r--arch/s390/kernel/vmlinux.lds.S2
-rw-r--r--arch/sh/kernel/vmlinux.lds.S2
-rw-r--r--arch/sparc/kernel/vmlinux.lds.S2
-rw-r--r--arch/tile/kernel/vmlinux.lds.S2
-rw-r--r--arch/um/include/asm/common.lds.S2
-rw-r--r--arch/x86/kernel/vmlinux.lds.S4
-rw-r--r--arch/xtensa/kernel/vmlinux.lds.S2
-rw-r--r--include/asm-generic/vmlinux.lds.h35
19 files changed, 41 insertions, 32 deletions
diff --git a/arch/alpha/kernel/vmlinux.lds.S b/arch/alpha/kernel/vmlinux.lds.S
index 003ef4c02585..173518f8c8bb 100644
--- a/arch/alpha/kernel/vmlinux.lds.S
+++ b/arch/alpha/kernel/vmlinux.lds.S
@@ -38,7 +38,7 @@ SECTIONS
__init_begin = ALIGN(PAGE_SIZE);
INIT_TEXT_SECTION(PAGE_SIZE)
INIT_DATA_SECTION(16)
- PERCPU(PAGE_SIZE)
+ PERCPU(64, PAGE_SIZE)
/* Align to THREAD_SIZE rather than PAGE_SIZE here so any padding page
needed for the THREAD_SIZE aligned init_task gets freed after init */
. = ALIGN(THREAD_SIZE);
diff --git a/arch/arm/kernel/vmlinux.lds.S b/arch/arm/kernel/vmlinux.lds.S
index 86b66f3f2031..cf78a03bf810 100644
--- a/arch/arm/kernel/vmlinux.lds.S
+++ b/arch/arm/kernel/vmlinux.lds.S
@@ -70,7 +70,7 @@ SECTIONS
#endif
}
- PERCPU(PAGE_SIZE)
+ PERCPU(32, PAGE_SIZE)
#ifndef CONFIG_XIP_KERNEL
. = ALIGN(PAGE_SIZE);
diff --git a/arch/blackfin/kernel/vmlinux.lds.S b/arch/blackfin/kernel/vmlinux.lds.S
index 4122678529c0..c40d07f708e8 100644
--- a/arch/blackfin/kernel/vmlinux.lds.S
+++ b/arch/blackfin/kernel/vmlinux.lds.S
@@ -136,7 +136,7 @@ SECTIONS
. = ALIGN(16);
INIT_DATA_SECTION(16)
- PERCPU(4)
+ PERCPU(32, 4)
.exit.data :
{
diff --git a/arch/cris/kernel/vmlinux.lds.S b/arch/cris/kernel/vmlinux.lds.S
index 442218980db0..c62e1346f47c 100644
--- a/arch/cris/kernel/vmlinux.lds.S
+++ b/arch/cris/kernel/vmlinux.lds.S
@@ -107,7 +107,7 @@ SECTIONS
#endif
__vmlinux_end = .; /* Last address of the physical file. */
#ifdef CONFIG_ETRAX_ARCH_V32
- PERCPU(PAGE_SIZE)
+ PERCPU(32, PAGE_SIZE)
.init.ramfs : {
INIT_RAM_FS
diff --git a/arch/frv/kernel/vmlinux.lds.S b/arch/frv/kernel/vmlinux.lds.S
index 8b973f3cc90e..0daae8af5787 100644
--- a/arch/frv/kernel/vmlinux.lds.S
+++ b/arch/frv/kernel/vmlinux.lds.S
@@ -37,7 +37,7 @@ SECTIONS
_einittext = .;
INIT_DATA_SECTION(8)
- PERCPU(4096)
+ PERCPU(L1_CACHE_BYTES, 4096)
. = ALIGN(PAGE_SIZE);
__init_end = .;
diff --git a/arch/ia64/kernel/vmlinux.lds.S b/arch/ia64/kernel/vmlinux.lds.S
index 5a4d044dcb1c..787de4a77d82 100644
--- a/arch/ia64/kernel/vmlinux.lds.S
+++ b/arch/ia64/kernel/vmlinux.lds.S
@@ -198,7 +198,7 @@ SECTIONS {
/* Per-cpu data: */
. = ALIGN(PERCPU_PAGE_SIZE);
- PERCPU_VADDR(PERCPU_ADDR, :percpu)
+ PERCPU_VADDR(SMP_CACHE_BYTES, PERCPU_ADDR, :percpu)
__phys_per_cpu_start = __per_cpu_load;
/*
* ensure percpu data fits
diff --git a/arch/m32r/kernel/vmlinux.lds.S b/arch/m32r/kernel/vmlinux.lds.S
index 7da94eaa082b..c194d64cdbb9 100644
--- a/arch/m32r/kernel/vmlinux.lds.S
+++ b/arch/m32r/kernel/vmlinux.lds.S
@@ -53,7 +53,7 @@ SECTIONS
__init_begin = .;
INIT_TEXT_SECTION(PAGE_SIZE)
INIT_DATA_SECTION(16)
- PERCPU(PAGE_SIZE)
+ PERCPU(32, PAGE_SIZE)
. = ALIGN(PAGE_SIZE);
__init_end = .;
/* freed after init ends here */
diff --git a/arch/mips/kernel/vmlinux.lds.S b/arch/mips/kernel/vmlinux.lds.S
index 570607b376b5..832afbb87588 100644
--- a/arch/mips/kernel/vmlinux.lds.S
+++ b/arch/mips/kernel/vmlinux.lds.S
@@ -115,7 +115,7 @@ SECTIONS
EXIT_DATA
}
- PERCPU(PAGE_SIZE)
+ PERCPU(1 << CONFIG_MIPS_L1_CACHE_SHIFT, PAGE_SIZE)
. = ALIGN(PAGE_SIZE);
__init_end = .;
/* freed after init ends here */
diff --git a/arch/mn10300/kernel/vmlinux.lds.S b/arch/mn10300/kernel/vmlinux.lds.S
index febbeee7f2f5..968bcd2cb022 100644
--- a/arch/mn10300/kernel/vmlinux.lds.S
+++ b/arch/mn10300/kernel/vmlinux.lds.S
@@ -70,7 +70,7 @@ SECTIONS
.exit.text : { EXIT_TEXT; }
.exit.data : { EXIT_DATA; }
- PERCPU(PAGE_SIZE)
+ PERCPU(32, PAGE_SIZE)
. = ALIGN(PAGE_SIZE);
__init_end = .;
/* freed after init ends here */
diff --git a/arch/parisc/kernel/vmlinux.lds.S b/arch/parisc/kernel/vmlinux.lds.S
index d64a6bbec2aa..8f1e4efd143e 100644
--- a/arch/parisc/kernel/vmlinux.lds.S
+++ b/arch/parisc/kernel/vmlinux.lds.S
@@ -145,7 +145,7 @@ SECTIONS
EXIT_DATA
}
- PERCPU(PAGE_SIZE)
+ PERCPU(L1_CACHE_BYTES, PAGE_SIZE)
. = ALIGN(PAGE_SIZE);
__init_end = .;
/* freed after init ends here */
diff --git a/arch/powerpc/kernel/vmlinux.lds.S b/arch/powerpc/kernel/vmlinux.lds.S
index 8a0deefac08d..b9150f07d266 100644
--- a/arch/powerpc/kernel/vmlinux.lds.S
+++ b/arch/powerpc/kernel/vmlinux.lds.S
@@ -160,7 +160,7 @@ SECTIONS
INIT_RAM_FS
}
- PERCPU(PAGE_SIZE)
+ PERCPU(L1_CACHE_BYTES, PAGE_SIZE)
. = ALIGN(8);
.machine.desc : AT(ADDR(.machine.desc) - LOAD_OFFSET) {
diff --git a/arch/s390/kernel/vmlinux.lds.S b/arch/s390/kernel/vmlinux.lds.S
index a68ac10213b2..1bc18cdb525b 100644
--- a/arch/s390/kernel/vmlinux.lds.S
+++ b/arch/s390/kernel/vmlinux.lds.S
@@ -77,7 +77,7 @@ SECTIONS
. = ALIGN(PAGE_SIZE);
INIT_DATA_SECTION(0x100)
- PERCPU(PAGE_SIZE)
+ PERCPU(0x100, PAGE_SIZE)
. = ALIGN(PAGE_SIZE);
__init_end = .; /* freed after init ends here */
diff --git a/arch/sh/kernel/vmlinux.lds.S b/arch/sh/kernel/vmlinux.lds.S
index 7f8a709c3ada..af4d46187a79 100644
--- a/arch/sh/kernel/vmlinux.lds.S
+++ b/arch/sh/kernel/vmlinux.lds.S
@@ -66,7 +66,7 @@ SECTIONS
__machvec_end = .;
}
- PERCPU(PAGE_SIZE)
+ PERCPU(L1_CACHE_BYTES, PAGE_SIZE)
/*
* .exit.text is discarded at runtime, not link time, to deal with
diff --git a/arch/sparc/kernel/vmlinux.lds.S b/arch/sparc/kernel/vmlinux.lds.S
index 0c1e6783657f..92b557afe535 100644
--- a/arch/sparc/kernel/vmlinux.lds.S
+++ b/arch/sparc/kernel/vmlinux.lds.S
@@ -108,7 +108,7 @@ SECTIONS
__sun4v_2insn_patch_end = .;
}
- PERCPU(PAGE_SIZE)
+ PERCPU(SMP_CACHE_BYTES, PAGE_SIZE)
. = ALIGN(PAGE_SIZE);
__init_end = .;
diff --git a/arch/tile/kernel/vmlinux.lds.S b/arch/tile/kernel/vmlinux.lds.S
index 25fdc0c1839a..c6ce378e0678 100644
--- a/arch/tile/kernel/vmlinux.lds.S
+++ b/arch/tile/kernel/vmlinux.lds.S
@@ -63,7 +63,7 @@ SECTIONS
*(.init.page)
} :data =0
INIT_DATA_SECTION(16)
- PERCPU(PAGE_SIZE)
+ PERCPU(L2_CACHE_BYTES, PAGE_SIZE)
. = ALIGN(PAGE_SIZE);
VMLINUX_SYMBOL(_einitdata) = .;
diff --git a/arch/um/include/asm/common.lds.S b/arch/um/include/asm/common.lds.S
index ac55b9efa1ce..34bede8aad4a 100644
--- a/arch/um/include/asm/common.lds.S
+++ b/arch/um/include/asm/common.lds.S
@@ -42,7 +42,7 @@
INIT_SETUP(0)
}
- PERCPU(32)
+ PERCPU(32, 32)
.initcall.init : {
INIT_CALLS
diff --git a/arch/x86/kernel/vmlinux.lds.S b/arch/x86/kernel/vmlinux.lds.S
index bf4700755184..cef446f8ac78 100644
--- a/arch/x86/kernel/vmlinux.lds.S
+++ b/arch/x86/kernel/vmlinux.lds.S
@@ -230,7 +230,7 @@ SECTIONS
* output PHDR, so the next output section - .init.text - should
* start another segment - init.
*/
- PERCPU_VADDR(0, :percpu)
+ PERCPU_VADDR(INTERNODE_CACHE_BYTES, 0, :percpu)
#endif
INIT_TEXT_SECTION(PAGE_SIZE)
@@ -305,7 +305,7 @@ SECTIONS
}
#if !defined(CONFIG_X86_64) || !defined(CONFIG_SMP)
- PERCPU(THREAD_SIZE)
+ PERCPU(INTERNODE_CACHE_BYTES, THREAD_SIZE)
#endif
. = ALIGN(PAGE_SIZE);
diff --git a/arch/xtensa/kernel/vmlinux.lds.S b/arch/xtensa/kernel/vmlinux.lds.S
index 9b526154c9ba..a2820065927e 100644
--- a/arch/xtensa/kernel/vmlinux.lds.S
+++ b/arch/xtensa/kernel/vmlinux.lds.S
@@ -155,7 +155,7 @@ SECTIONS
INIT_RAM_FS
}
- PERCPU(PAGE_SIZE)
+ PERCPU(XCHAL_ICACHE_LINESIZE, PAGE_SIZE)
/* We need this dummy segment here */
diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h
index 6ebb81030d2d..439df587c12c 100644
--- a/include/asm-generic/vmlinux.lds.h
+++ b/include/asm-generic/vmlinux.lds.h
@@ -15,7 +15,7 @@
* HEAD_TEXT_SECTION
* INIT_TEXT_SECTION(PAGE_SIZE)
* INIT_DATA_SECTION(...)
- * PERCPU(PAGE_SIZE)
+ * PERCPU(CACHELINE_SIZE, PAGE_SIZE)
* __init_end = .;
*
* _stext = .;
@@ -683,13 +683,18 @@
/**
* PERCPU_VADDR - define output section for percpu area
+ * @cacheline: cacheline size
* @vaddr: explicit base address (optional)
* @phdr: destination PHDR (optional)
*
- * Macro which expands to output section for percpu area. If @vaddr
- * is not blank, it specifies explicit base address and all percpu
- * symbols will be offset from the given address. If blank, @vaddr
- * always equals @laddr + LOAD_OFFSET.
+ * Macro which expands to output section for percpu area.
+ *
+ * @cacheline is used to align subsections to avoid false cacheline
+ * sharing between subsections for different purposes.
+ *
+ * If @vaddr is not blank, it specifies explicit base address and all
+ * percpu symbols will be offset from the given address. If blank,
+ * @vaddr always equals @laddr + LOAD_OFFSET.
*
* @phdr defines the output PHDR to use if not blank. Be warned that
* output PHDR is sticky. If @phdr is specified, the next output
@@ -700,7 +705,7 @@
* If there is no need to put the percpu section at a predetermined
* address, use PERCPU().
*/
-#define PERCPU_VADDR(vaddr, phdr) \
+#define PERCPU_VADDR(cacheline, vaddr, phdr) \
VMLINUX_SYMBOL(__per_cpu_load) = .; \
.data..percpu vaddr : AT(VMLINUX_SYMBOL(__per_cpu_load) \
- LOAD_OFFSET) { \
@@ -708,7 +713,9 @@
*(.data..percpu..first) \
. = ALIGN(PAGE_SIZE); \
*(.data..percpu..page_aligned) \
+ . = ALIGN(cacheline); \
*(.data..percpu..readmostly) \
+ . = ALIGN(cacheline); \
*(.data..percpu) \
*(.data..percpu..shared_aligned) \
VMLINUX_SYMBOL(__per_cpu_end) = .; \
@@ -717,18 +724,18 @@
/**
* PERCPU - define output section for percpu area, simple version
+ * @cacheline: cacheline size
* @align: required alignment
*
- * Align to @align and outputs output section for percpu area. This
- * macro doesn't maniuplate @vaddr or @phdr and __per_cpu_load and
+ * Align to @align and outputs output section for percpu area. This macro
+ * doesn't manipulate @vaddr or @phdr and __per_cpu_load and
* __per_cpu_start will be identical.
*
- * This macro is equivalent to ALIGN(align); PERCPU_VADDR( , ) except
- * that __per_cpu_load is defined as a relative symbol against
- * .data..percpu which is required for relocatable x86_32
- * configuration.
+ * This macro is equivalent to ALIGN(@align); PERCPU_VADDR(@cacheline,,)
+ * except that __per_cpu_load is defined as a relative symbol against
+ * .data..percpu which is required for relocatable x86_32 configuration.
*/
-#define PERCPU(align) \
+#define PERCPU(cacheline, align) \
. = ALIGN(align); \
.data..percpu : AT(ADDR(.data..percpu) - LOAD_OFFSET) { \
VMLINUX_SYMBOL(__per_cpu_load) = .; \
@@ -736,7 +743,9 @@
*(.data..percpu..first) \
. = ALIGN(PAGE_SIZE); \
*(.data..percpu..page_aligned) \
+ . = ALIGN(cacheline); \
*(.data..percpu..readmostly) \
+ . = ALIGN(cacheline); \
*(.data..percpu) \
*(.data..percpu..shared_aligned) \
VMLINUX_SYMBOL(__per_cpu_end) = .; \