From 1b5576e69a5fe168c08a159685ac366316ac9bbc Mon Sep 17 00:00:00 2001 From: Yinghai Lu Date: Fri, 22 Jan 2010 11:21:04 +0800 Subject: x86: Remove BIOS data range from e820 In preparation for moving to the generic page_is_ram(), make explicit what we expect to be reserved and not reserved. Tested-by: Wu Fengguang Signed-off-by: Yinghai Lu LKML-Reference: <20100122033004.335813103@intel.com> Cc: Andrew Morton Signed-off-by: H. Peter Anvin --- arch/x86/kernel/e820.c | 8 ++++++++ arch/x86/kernel/setup.c | 19 ++++++++++++++++++- arch/x86/mm/ioremap.c | 16 ---------------- 3 files changed, 26 insertions(+), 17 deletions(-) (limited to 'arch/x86') diff --git a/arch/x86/kernel/e820.c b/arch/x86/kernel/e820.c index d17d482a04f..230687ba5ba 100644 --- a/arch/x86/kernel/e820.c +++ b/arch/x86/kernel/e820.c @@ -517,11 +517,19 @@ u64 __init e820_remove_range(u64 start, u64 size, unsigned old_type, int checktype) { int i; + u64 end; u64 real_removed_size = 0; if (size > (ULLONG_MAX - start)) size = ULLONG_MAX - start; + end = start + size; + printk(KERN_DEBUG "e820 remove range: %016Lx - %016Lx ", + (unsigned long long) start, + (unsigned long long) end); + e820_print_type(old_type); + printk(KERN_CONT "\n"); + for (i = 0; i < e820.nr_map; i++) { struct e820entry *ei = &e820.map[i]; u64 final_start, final_end; diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c index cdb6a8a506d..f9b1f4e5ab7 100644 --- a/arch/x86/kernel/setup.c +++ b/arch/x86/kernel/setup.c @@ -650,6 +650,23 @@ static struct dmi_system_id __initdata bad_bios_dmi_table[] = { {} }; +static void __init trim_bios_range(void) +{ + /* + * A special case is the first 4Kb of memory; + * This is a BIOS owned area, not kernel ram, but generally + * not listed as such in the E820 table. + */ + e820_update_range(0, PAGE_SIZE, E820_RAM, E820_RESERVED); + /* + * special case: Some BIOSen report the PC BIOS + * area (640->1Mb) as ram even though it is not. + * take them out. + */ + e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1); + sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map); +} + /* * Determine if we were loaded by an EFI loader. If so, then we have also been * passed the efi memmap, systab, etc., so we should use these data structures @@ -813,7 +830,7 @@ void __init setup_arch(char **cmdline_p) insert_resource(&iomem_resource, &data_resource); insert_resource(&iomem_resource, &bss_resource); - + trim_bios_range(); #ifdef CONFIG_X86_32 if (ppro_with_ram_bug()) { e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM, diff --git a/arch/x86/mm/ioremap.c b/arch/x86/mm/ioremap.c index 334e63ca7b2..30e068d6462 100644 --- a/arch/x86/mm/ioremap.c +++ b/arch/x86/mm/ioremap.c @@ -29,22 +29,6 @@ int page_is_ram(unsigned long pagenr) resource_size_t addr, end; int i; - /* - * A special case is the first 4Kb of memory; - * This is a BIOS owned area, not kernel ram, but generally - * not listed as such in the E820 table. - */ - if (pagenr == 0) - return 0; - - /* - * Second special case: Some BIOSen report the PC BIOS - * area (640->1Mb) as ram even though it is not. - */ - if (pagenr >= (BIOS_BEGIN >> PAGE_SHIFT) && - pagenr < (BIOS_END >> PAGE_SHIFT)) - return 0; - for (i = 0; i < e820.nr_map; i++) { /* * Not usable memory: -- cgit v1.2.3 From 13ca0fcaa33f6b1984c4111b6ec5df42689fea6f Mon Sep 17 00:00:00 2001 From: Wu Fengguang Date: Fri, 22 Jan 2010 11:21:05 +0800 Subject: x86: Use the generic page_is_ram() The generic resource based page_is_ram() works better with memory hotplug/hotremove. So switch the x86 e820map based code to it. CC: Andi Kleen CC: KAMEZAWA Hiroyuki CC: Yinghai Lu Signed-off-by: Wu Fengguang LKML-Reference: <20100122033004.470767217@intel.com> Cc: Andrew Morton Signed-off-by: H. Peter Anvin --- arch/x86/include/asm/page_types.h | 1 - arch/x86/mm/ioremap.c | 21 --------------------- 2 files changed, 22 deletions(-) (limited to 'arch/x86') diff --git a/arch/x86/include/asm/page_types.h b/arch/x86/include/asm/page_types.h index 642fe34b36a..a667f24c725 100644 --- a/arch/x86/include/asm/page_types.h +++ b/arch/x86/include/asm/page_types.h @@ -40,7 +40,6 @@ #ifndef __ASSEMBLY__ -extern int page_is_ram(unsigned long pagenr); extern int devmem_is_allowed(unsigned long pagenr); extern unsigned long max_low_pfn_mapped; diff --git a/arch/x86/mm/ioremap.c b/arch/x86/mm/ioremap.c index 30e068d6462..1bf9e08ed73 100644 --- a/arch/x86/mm/ioremap.c +++ b/arch/x86/mm/ioremap.c @@ -24,27 +24,6 @@ #include "physaddr.h" -int page_is_ram(unsigned long pagenr) -{ - resource_size_t addr, end; - int i; - - for (i = 0; i < e820.nr_map; i++) { - /* - * Not usable memory: - */ - if (e820.map[i].type != E820_RAM) - continue; - addr = (e820.map[i].addr + PAGE_SIZE-1) >> PAGE_SHIFT; - end = (e820.map[i].addr + e820.map[i].size) >> PAGE_SHIFT; - - - if ((pagenr >= addr) && (pagenr < end)) - return 1; - } - return 0; -} - /* * Fix up the linear direct mapping of the kernel to avoid cache attribute * conflicts. -- cgit v1.2.3 From 39c662f60c556908faf861ef0430549b1731b891 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Sat, 25 Jul 2009 19:15:48 +0200 Subject: x86: Convert tlbstate_lock to raw_spinlock Signed-off-by: Thomas Gleixner --- arch/x86/mm/tlb.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'arch/x86') diff --git a/arch/x86/mm/tlb.c b/arch/x86/mm/tlb.c index 65b58e4b0b8..426f3a1a64d 100644 --- a/arch/x86/mm/tlb.c +++ b/arch/x86/mm/tlb.c @@ -41,7 +41,7 @@ union smp_flush_state { struct { struct mm_struct *flush_mm; unsigned long flush_va; - spinlock_t tlbstate_lock; + raw_spinlock_t tlbstate_lock; DECLARE_BITMAP(flush_cpumask, NR_CPUS); }; char pad[INTERNODE_CACHE_BYTES]; @@ -181,7 +181,7 @@ static void flush_tlb_others_ipi(const struct cpumask *cpumask, * num_online_cpus() <= NUM_INVALIDATE_TLB_VECTORS, but it is * probably not worth checking this for a cache-hot lock. */ - spin_lock(&f->tlbstate_lock); + raw_spin_lock(&f->tlbstate_lock); f->flush_mm = mm; f->flush_va = va; @@ -199,7 +199,7 @@ static void flush_tlb_others_ipi(const struct cpumask *cpumask, f->flush_mm = NULL; f->flush_va = 0; - spin_unlock(&f->tlbstate_lock); + raw_spin_unlock(&f->tlbstate_lock); } void native_flush_tlb_others(const struct cpumask *cpumask, @@ -223,7 +223,7 @@ static int __cpuinit init_smp_flush(void) int i; for (i = 0; i < ARRAY_SIZE(flush_state); i++) - spin_lock_init(&flush_state[i].tlbstate_lock); + raw_spin_lock_init(&flush_state[i].tlbstate_lock); return 0; } -- cgit v1.2.3 From e808bae2407a087bfd40200a27587898e5a9909d Mon Sep 17 00:00:00 2001 From: Thadeu Lima de Souza Cascardo Date: Tue, 9 Feb 2010 21:38:45 -0200 Subject: x86: Do not reserve brk for DMI if it's not going to be used This will save 64K bytes from memory when loading linux if DMI is disabled, which is good for embedded systems. Signed-off-by: Thadeu Lima de Souza Cascardo LKML-Reference: <1265758732-19320-1-git-send-email-cascardo@holoscopio.com> Signed-off-by: H. Peter Anvin --- arch/x86/kernel/setup.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'arch/x86') diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c index 3499b4fabc9..cb42109a55b 100644 --- a/arch/x86/kernel/setup.c +++ b/arch/x86/kernel/setup.c @@ -121,7 +121,9 @@ unsigned long max_low_pfn_mapped; unsigned long max_pfn_mapped; +#ifdef CONFIG_DMI RESERVE_BRK(dmi_alloc, 65536); +#endif unsigned int boot_cpu_id __read_mostly; -- cgit v1.2.3 From 14315592009c17035cac81f4954d5a1f4d71e489 Mon Sep 17 00:00:00 2001 From: Ian Campbell Date: Wed, 17 Feb 2010 10:38:10 +0000 Subject: x86, mm: Allow highmem user page tables to be disabled at boot time Distros generally (I looked at Debian, RHEL5 and SLES11) seem to enable CONFIG_HIGHPTE for any x86 configuration which has highmem enabled. This means that the overhead applies even to machines which have a fairly modest amount of high memory and which therefore do not really benefit from allocating PTEs in high memory but still pay the price of the additional mapping operations. Running kernbench on a 4G box I found that with CONFIG_HIGHPTE=y but no actual highptes being allocated there was a reduction in system time used from 59.737s to 55.9s. With CONFIG_HIGHPTE=y and highmem PTEs being allocated: Average Optimal load -j 4 Run (std deviation): Elapsed Time 175.396 (0.238914) User Time 515.983 (5.85019) System Time 59.737 (1.26727) Percent CPU 263.8 (71.6796) Context Switches 39989.7 (4672.64) Sleeps 42617.7 (246.307) With CONFIG_HIGHPTE=y but with no highmem PTEs being allocated: Average Optimal load -j 4 Run (std deviation): Elapsed Time 174.278 (0.831968) User Time 515.659 (6.07012) System Time 55.9 (1.07799) Percent CPU 263.8 (71.266) Context Switches 39929.6 (4485.13) Sleeps 42583.7 (373.039) This patch allows the user to control the allocation of PTEs in highmem from the command line ("userpte=nohigh") but retains the status-quo as the default. It is possible that some simple heuristic could be developed which allows auto-tuning of this option however I don't have a sufficiently large machine available to me to perform any particularly meaningful experiments. We could probably handwave up an argument for a threshold at 16G of total RAM. Assuming 768M of lowmem we have 196608 potential lowmem PTE pages. Each page can map 2M of RAM in a PAE-enabled configuration, meaning a maximum of 384G of RAM could potentially be mapped using lowmem PTEs. Even allowing generous factor of 10 to account for other required lowmem allocations, generous slop to account for page sharing (which reduces the total amount of RAM mappable by a given number of PT pages) and other innacuracies in the estimations it would seem that even a 32G machine would not have a particularly pressing need for highmem PTEs. I think 32G could be considered to be at the upper bound of what might be sensible on a 32 bit machine (although I think in practice 64G is still supported). It's seems questionable if HIGHPTE is even a win for any amount of RAM you would sensibly run a 32 bit kernel on rather than going 64 bit. Signed-off-by: Ian Campbell LKML-Reference: <1266403090-20162-1-git-send-email-ian.campbell@citrix.com> Signed-off-by: H. Peter Anvin --- arch/x86/include/asm/pgalloc.h | 5 +++++ arch/x86/mm/pgtable.c | 31 ++++++++++++++++++++++++++----- 2 files changed, 31 insertions(+), 5 deletions(-) (limited to 'arch/x86') diff --git a/arch/x86/include/asm/pgalloc.h b/arch/x86/include/asm/pgalloc.h index 0e8c2a0fd92..271de94c381 100644 --- a/arch/x86/include/asm/pgalloc.h +++ b/arch/x86/include/asm/pgalloc.h @@ -22,6 +22,11 @@ static inline void paravirt_release_pmd(unsigned long pfn) {} static inline void paravirt_release_pud(unsigned long pfn) {} #endif +/* + * Flags to use when allocating a user page table page. + */ +extern gfp_t __userpte_alloc_gfp; + /* * Allocate and free page tables. */ diff --git a/arch/x86/mm/pgtable.c b/arch/x86/mm/pgtable.c index ed34f5e3599..c9ba9deafe8 100644 --- a/arch/x86/mm/pgtable.c +++ b/arch/x86/mm/pgtable.c @@ -6,6 +6,14 @@ #define PGALLOC_GFP GFP_KERNEL | __GFP_NOTRACK | __GFP_REPEAT | __GFP_ZERO +#ifdef CONFIG_HIGHPTE +#define PGALLOC_USER_GFP __GFP_HIGHMEM +#else +#define PGALLOC_USER_GFP 0 +#endif + +gfp_t __userpte_alloc_gfp = PGALLOC_GFP | PGALLOC_USER_GFP; + pte_t *pte_alloc_one_kernel(struct mm_struct *mm, unsigned long address) { return (pte_t *)__get_free_page(PGALLOC_GFP); @@ -15,16 +23,29 @@ pgtable_t pte_alloc_one(struct mm_struct *mm, unsigned long address) { struct page *pte; -#ifdef CONFIG_HIGHPTE - pte = alloc_pages(PGALLOC_GFP | __GFP_HIGHMEM, 0); -#else - pte = alloc_pages(PGALLOC_GFP, 0); -#endif + pte = alloc_pages(__userpte_alloc_gfp, 0); if (pte) pgtable_page_ctor(pte); return pte; } +static int __init setup_userpte(char *arg) +{ + if (!arg) + return -EINVAL; + + /* + * "userpte=nohigh" disables allocation of user pagetables in + * high memory. + */ + if (strcmp(arg, "nohigh") == 0) + __userpte_alloc_gfp &= ~__GFP_HIGHMEM; + else + return -EINVAL; + return 0; +} +early_param("userpte", setup_userpte); + void ___pte_free_tlb(struct mmu_gather *tlb, struct page *pte) { pgtable_page_dtor(pte); -- cgit v1.2.3 From c1fd1b43831fa20c91cdd461342af8edf2e87c2f Mon Sep 17 00:00:00 2001 From: Pekka Enberg Date: Wed, 24 Feb 2010 17:04:47 +0200 Subject: x86, mm: Unify kernel_physical_mapping_init() API This patch changes the 32-bit version of kernel_physical_mapping_init() to return the last mapped address like the 64-bit one so that we can unify the call-site in init_memory_mapping(). Cc: Yinghai Lu Cc: KAMEZAWA Hiroyuki Signed-off-by: Pekka Enberg LKML-Reference: Signed-off-by: H. Peter Anvin --- arch/x86/mm/init.c | 7 ------- arch/x86/mm/init_32.c | 8 +++++--- 2 files changed, 5 insertions(+), 10 deletions(-) (limited to 'arch/x86') diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c index d406c523901..e71c5cbc8f3 100644 --- a/arch/x86/mm/init.c +++ b/arch/x86/mm/init.c @@ -266,16 +266,9 @@ unsigned long __init_refok init_memory_mapping(unsigned long start, if (!after_bootmem) find_early_table_space(end, use_pse, use_gbpages); -#ifdef CONFIG_X86_32 - for (i = 0; i < nr_range; i++) - kernel_physical_mapping_init(mr[i].start, mr[i].end, - mr[i].page_size_mask); - ret = end; -#else /* CONFIG_X86_64 */ for (i = 0; i < nr_range; i++) ret = kernel_physical_mapping_init(mr[i].start, mr[i].end, mr[i].page_size_mask); -#endif #ifdef CONFIG_X86_32 early_ioremap_page_table_range_init(); diff --git a/arch/x86/mm/init_32.c b/arch/x86/mm/init_32.c index 9a0c258a86b..2226f2c70ea 100644 --- a/arch/x86/mm/init_32.c +++ b/arch/x86/mm/init_32.c @@ -241,6 +241,7 @@ kernel_physical_mapping_init(unsigned long start, unsigned long page_size_mask) { int use_pse = page_size_mask == (1<