x86: Configure out compilation of Simple Boot Flag support
RFC PATCH 1/1 x86: Configure out compilation of Simple Boot Flag support
Hi,
The Simple Boot Flag specification support (in arch/x86/kernel/bootflag.c) is not really needed on embedded platforms, so it can probably be compiled out. Zwane Mwaikambo did a patch a few months (years ?) ago to configure out the compilation of that file, and I've updated the patch to recent versions of the kernel.
However, after writing the patch, I discovered that the space savings are not that interesting. When Zwane wrote the patch:
text data bss dec hex filename
1645 8 0 1653 675 arch/i386/kernel/bootflag.o
And now:
text data bss dec hex filename
310 8 0 318 13e arch/x86/kernel/bootflag.oWhich means that on the whole kernel, the savings are really small:
text data bss dec hex filename
1081314 121464 94208 1296986 13ca5a vmlinux.before 1081106 121460 94208 1296774 13c986 vmlinux.after
-208 -4 0 -212 -D4 +/-
Is it interesting to add such a patch to the kernel ?
I've enclosed the patch in that mail. Maybe it can considered as a clean-up patch, because of the move to arch/x86/kernel/bootflag.c of the acpi_parse_sbf() function ?
(Note: the arch/x86/kernel/acpi/boot.c file calls acpi_table_parse(ACPI_SIG_BOOT, acpi_parse_sbf) twice, once in acpi_boot_init() and once in acpi_boot_table_init(). Both functions are called from setup_arch(), either in the 32bits and 64bits versions. Can't we remove one of the call to acpi_table_parse() ? The patch below assumes that the call to acpi_table_parse() has been removed from acpi_boot_init(), but I can of course change that.)
Thanks for your comments,
Thomas
---
Allow to disable the compilation of Simple Boot Flag specification support, which allows Linux to tell the BIOS various boot options.
Basically, the patch adds a BOOTFLAG configuration option, on which the compilation of arch/x86/kernel/bootflag.c depends. The patch also changes the way sbf_port is initialized by the ACPI initialization routines. Instead of having sbf_port exported, we rather export the sbf_acpi_parse() function (formerly known as acpi_parse_buf() and implemented in the ACPI initialization code).
Result:
text data bss dec hex filename
1081314 121464 94208 1296986 13ca5a vmlinux.before 1081106 121460 94208 1296774 13c986 vmlinux.after
-208 -4 0 -212 -D4 +/-
This patch is part of the Linux Tiny project and is based on previous work done by Zwane Mwaikambo <[email protected]>.
--- arch/x86/Kconfig | 8 ++++++++ arch/x86/kernel/Makefile | 3 ++- arch/x86/kernel/acpi/boot.c | 17 +---------------- arch/x86/kernel/bootflag.c | 18 +++++++++++++++++- include/linux/acpi.h | 10 +++++++++- 5 files changed, 37 insertions(+), 19 deletions(-)
arch/x86/Kconfig +8 -0
@@ -439,6 +439,14 @@ affected by entries in the DMI blacklist. Required by PNP BIOS code. +config BOOTFLAG + default y + bool "Enable Simple Boot Flag specification support" if EMBEDDED + depends on ACPI + help + Enable support for the Simple Boot Flag specification, that + can be use by Linux to communicate boot options to the BIOS. + config GART_IOMMU bool "GART IOMMU support" if EMBEDDED default y
arch/x86/kernel/Makefile +2 -1
@@ -15,7 +15,7 @@ obj-$(CONFIG_X86_32) += sys_i386_32.o i386_ksyms_32.o obj-$(CONFIG_X86_64) += sys_x86_64.o x8664_ksyms_64.o obj-$(CONFIG_X86_64) += syscall_64.o vsyscall_64.o setup64.o -obj-y += pci-dma_$(BITS).o bootflag.o e820_$(BITS).o +obj-y += pci-dma_$(BITS).o e820_$(BITS).o obj-y += quirks.o i8237.o topology.o kdebugfs.o obj-y += alternative.o i8253.o obj-$(CONFIG_X86_64) += pci-nommu_64.o bugs_64.o
@@ -58,6 +58,7 @@ obj-$(CONFIG_ACPI_SRAT) += srat_32.o obj-$(CONFIG_EFI) += efi.o efi_$(BITS).o efi_stub_$(BITS).o obj-$(CONFIG_DOUBLEFAULT) += doublefault_32.o +obj-$(CONFIG_BOOTFLAG) += bootflag.o obj-$(CONFIG_VM86) += vm86_32.o obj-$(CONFIG_EARLY_PRINTK) += early_printk.o
arch/x86/kernel/acpi/boot.c +1 -16
@@ -584,21 +584,6 @@EXPORT_SYMBOL (acpi_unregister_ioapic); -static int __init acpi_parse_sbf (struct acpi_table_header *table) -{ -struct acpi_table_boot *sb; - - sb = (struct acpi_table_boot *)table; -if (!sb) { - printk(KERN_WARNING PREFIX "Unable to map SBF\n"); -return -ENODEV ; - } - - sbf_port = sb->cmos_index;/* Save CMOS port */ - -return 0 ; -} -#ifdef CONFIG_HPET_TIMER #include <asm/hpet.h>
@@ -1170,7 +1155,7 @@return error; } -acpi_table_parse (ACPI_SIG_BOOT , acpi_parse_sbf); +acpi_table_parse (ACPI_SIG_BOOT , sbf_acpi_parse);/* * blacklist may disable ACPI entirely
arch/x86/kernel/bootflag.c +17 -1
@@ -18,7 +18,7 @@#define SBF_DIAG (1<<2)#define SBF_PARITY (1<<7) -int sbf_port__initdata =-1 ;/* set via acpi_boot_init() */ +static int sbf_port__initdata =-1 ;static int __init parity (u8 v) {
@@ -76,6 +76,22 @@return 1 ; } +/* called via acpi_boot_init() */ +int __init sbf_acpi_parse (struct acpi_table_header *table) +{ +struct acpi_table_boot *sb; + + sb = (struct acpi_table_boot *)table; +if (!sb) { + printk(KERN_WARNING PREFIX "Unable to map SBF\n"); +return -ENODEV ; + } + + sbf_port = sb->cmos_index;/* Save CMOS port */ + +return 0 ; +} +static int __init sbf_init (void ) {u8 v;
include/linux/acpi.h +9 -1
@@ -119,7 +119,15 @@extern struct acpi_mcfg_allocation *pci_mmcfg_config;extern int pci_mmcfg_config_num; -extern int sbf_port; +#ifdef CONFIG_BOOTFLAG +int sbf_acpi_parse (struct acpi_table_header *table); +#else +static inline int sbf_acpi_parse (struct acpi_table_header *table) +{ +return -ENODEV ; +} +#endif +extern unsigned long acpi_realmode_flags;int acpi_register_gsi (u32 gsi,int triggering,int polarity);
Signature · 4 lines
Thomas Petazzoni, Free Electrons Free Embedded Linux Training Materials on http://free-electrons.com/training (More than 1500 pages!)
arch/x86/kernel/acpi/boot.c line 1158
Quoted 2 lines · Thomas Petazzoni
arch/x86/kernel/acpi/boot.c +1
-1-acpi_table_parse (ACPI_SIG_BOOT , acpi_parse_sbf); +acpi_table_parse (ACPI_SIG_BOOT , sbf_acpi_parse);
no strong feelings if the ACPI guys ack it - but i suspect the renaming of the function from acpi_parse_sbf to sbf_acpi_parse was unnecessary?
Ingo
Quoted 3 lines · Ingo Molnar
no strong feelings if the ACPI guys ack it - but i suspect the renaming of the function from
acpi_parse_sbftosbf_acpi_parsewas unnecessary?
I rename acpi_parse_sbf() to sbf_acpi_parse() because the function was moved from arch/x86/kernel/acpi/boot.c, where pretty much all functions are named acpi_*() to the file arch/x86/kernel/bootflag.c, where all the functions are named sbf_*(). However, I don't mind reverting that renaming, I just thought it was cleaner, but I have no strong opinion on it.
Any thoughts about the duplicate call toacpi_table_parse(ACPI_SIG_BOOT, acpi_parse_sbf) ?
Thanks for your comments,
Thomas
Signature · 4 lines
Thomas Petazzoni, Free Electrons Free Embedded Linux Training Materials on http://free-electrons.com/training (More than 1500 pages!)