From 4595f9620cda8a1e973588e743cf5f8436dd20c6 Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Sat, 10 Jan 2009 21:58:09 -0800 Subject: x86: change flush_tlb_others to take a const struct cpumask Impact: reduce stack usage, use new cpumask API. This is made a little more tricky by uv_flush_tlb_others which actually alters its argument, for an IPI to be sent to the remaining cpus in the mask. I solve this by allocating a cpumask_var_t for this case and falling back to IPI should this fail. To eliminate temporaries in the caller, all flush_tlb_others implementations now do the this-cpu-elimination step themselves. Note also the curious "cpus_or(f->flush_cpumask, cpumask, f->flush_cpumask)" which has been there since pre-git and yet f->flush_cpumask is always zero at this point. Signed-off-by: Rusty Russell Signed-off-by: Mike Travis --- arch/x86/xen/enlighten.c | 31 ++++++++++++------------------- 1 file changed, 12 insertions(+), 19 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index bea215230b2..965539ec425 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -634,35 +634,27 @@ static void xen_flush_tlb_single(unsigned long addr) preempt_enable(); } -static void xen_flush_tlb_others(const cpumask_t *cpus, struct mm_struct *mm, - unsigned long va) +static void xen_flush_tlb_others(const struct cpumask *cpus, + struct mm_struct *mm, unsigned long va) { struct { struct mmuext_op op; - cpumask_t mask; + DECLARE_BITMAP(mask, NR_CPUS); } *args; - cpumask_t cpumask = *cpus; struct multicall_space mcs; - /* - * A couple of (to be removed) sanity checks: - * - * - current CPU must not be in mask - * - mask must exist :) - */ - BUG_ON(cpus_empty(cpumask)); - BUG_ON(cpu_isset(smp_processor_id(), cpumask)); + BUG_ON(cpumask_empty(cpus)); BUG_ON(!mm); - /* If a CPU which we ran on has gone down, OK. */ - cpus_and(cpumask, cpumask, cpu_online_map); - if (cpus_empty(cpumask)) - return; - mcs = xen_mc_entry(sizeof(*args)); args = mcs.args; - args->mask = cpumask; - args->op.arg2.vcpumask = &args->mask; + args->op.arg2.vcpumask = to_cpumask(args->mask); + + /* Remove us, and any offline CPUS. */ + cpumask_and(to_cpumask(args->mask), cpus, cpu_online_mask); + cpumask_clear_cpu(smp_processor_id(), to_cpumask(args->mask)); + if (unlikely(cpumask_empty(to_cpumask(args->mask)))) + goto issue; if (va == TLB_FLUSH_ALL) { args->op.cmd = MMUEXT_TLB_FLUSH_MULTI; @@ -673,6 +665,7 @@ static void xen_flush_tlb_others(const cpumask_t *cpus, struct mm_struct *mm, MULTI_mmuext_op(mcs.mc, &args->op, 1, NULL, DOMID_SELF); +issue: xen_mc_issue(PARAVIRT_LAZY_MMU); } -- cgit v1.2.3-70-g09d2 From 1a51e3a0aed18767cf2762e95456ecfeb0bca5e6 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Tue, 13 Jan 2009 20:41:35 +0900 Subject: x86: fold pda into percpu area on SMP [ Based on original patch from Christoph Lameter and Mike Travis. ] Currently pdas and percpu areas are allocated separately. %gs points to local pda and percpu area can be reached using pda->data_offset. This patch folds pda into percpu area. Due to strange gcc requirement, pda needs to be at the beginning of the percpu area so that pda->stack_canary is at %gs:40. To achieve this, a new percpu output section macro - PERCPU_VADDR_PREALLOC() - is added and used to reserve pda sized chunk at the start of the percpu area. After this change, for boot cpu, %gs first points to pda in the data.init area and later during setup_per_cpu_areas() gets updated to point to the actual pda. This means that setup_per_cpu_areas() need to reload %gs for CPU0 while clearing pda area for other cpus as cpu0 already has modified it when control reaches setup_per_cpu_areas(). This patch also removes now unnecessary get_local_pda() and its call sites. A lot of this patch is taken from Mike Travis' "x86_64: Fold pda into per cpu area" patch. Signed-off-by: Tejun Heo Signed-off-by: Ingo Molnar --- arch/x86/include/asm/percpu.h | 8 +++ arch/x86/include/asm/smp.h | 2 - arch/x86/kernel/asm-offsets_64.c | 1 + arch/x86/kernel/cpu/common.c | 6 +-- arch/x86/kernel/head64.c | 8 ++- arch/x86/kernel/head_64.S | 15 ++++-- arch/x86/kernel/setup_percpu.c | 107 ++++++++++++++++---------------------- arch/x86/kernel/smpboot.c | 60 +-------------------- arch/x86/kernel/vmlinux_64.lds.S | 6 ++- arch/x86/xen/smp.c | 10 ---- include/asm-generic/vmlinux.lds.h | 25 ++++++++- 11 files changed, 104 insertions(+), 144 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/include/asm/percpu.h b/arch/x86/include/asm/percpu.h index df644f3e53e..0ed77cf33f7 100644 --- a/arch/x86/include/asm/percpu.h +++ b/arch/x86/include/asm/percpu.h @@ -1,6 +1,14 @@ #ifndef _ASM_X86_PERCPU_H #define _ASM_X86_PERCPU_H +#ifndef __ASSEMBLY__ +#ifdef CONFIG_X86_64 +extern void load_pda_offset(int cpu); +#else +static inline void load_pda_offset(int cpu) { } +#endif +#endif + #ifdef CONFIG_X86_64 #include diff --git a/arch/x86/include/asm/smp.h b/arch/x86/include/asm/smp.h index a8cea7b0943..127415402ea 100644 --- a/arch/x86/include/asm/smp.h +++ b/arch/x86/include/asm/smp.h @@ -19,8 +19,6 @@ #include #include -extern int __cpuinit get_local_pda(int cpu); - extern int smp_num_siblings; extern unsigned int num_processors; diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c index 1d41d3f1edb..f8d1b047ef4 100644 --- a/arch/x86/kernel/asm-offsets_64.c +++ b/arch/x86/kernel/asm-offsets_64.c @@ -56,6 +56,7 @@ int main(void) ENTRY(cpunumber); ENTRY(irqstackptr); ENTRY(data_offset); + DEFINE(pda_size, sizeof(struct x8664_pda)); BLANK(); #undef ENTRY #ifdef CONFIG_PARAVIRT diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index c116c599326..7041acdf557 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -893,10 +893,8 @@ void __cpuinit pda_init(int cpu) /* Setup up data that may be needed in __get_free_pages early */ loadsegment(fs, 0); loadsegment(gs, 0); - /* Memory clobbers used to order PDA accessed */ - mb(); - wrmsrl(MSR_GS_BASE, pda); - mb(); + + load_pda_offset(cpu); pda->cpunumber = cpu; pda->irqcount = -1; diff --git a/arch/x86/kernel/head64.c b/arch/x86/kernel/head64.c index 462d0beccb6..1a311293f73 100644 --- a/arch/x86/kernel/head64.c +++ b/arch/x86/kernel/head64.c @@ -26,12 +26,18 @@ #include #include -/* boot cpu pda, referenced by head_64.S to initialize %gs for boot CPU */ +#ifndef CONFIG_SMP +/* boot cpu pda, referenced by head_64.S to initialize %gs on UP */ struct x8664_pda _boot_cpu_pda; +#endif void __init x86_64_init_pda(void) { +#ifdef CONFIG_SMP + cpu_pda(0) = (void *)__per_cpu_load; +#else cpu_pda(0) = &_boot_cpu_pda; +#endif cpu_pda(0)->data_offset = (unsigned long)(__per_cpu_load - __per_cpu_start); pda_init(0); diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S index 2f0ab008988..7a995d0e9f7 100644 --- a/arch/x86/kernel/head_64.S +++ b/arch/x86/kernel/head_64.S @@ -245,10 +245,13 @@ ENTRY(secondary_startup_64) /* Set up %gs. * - * %gs should point to the pda. For initial boot, make %gs point - * to the _boot_cpu_pda in data section. For a secondary CPU, - * initial_gs should be set to its pda address before the CPU runs - * this code. + * On SMP, %gs should point to the per-cpu area. For initial + * boot, make %gs point to the init data section. For a + * secondary CPU,initial_gs should be set to its pda address + * before the CPU runs this code. + * + * On UP, initial_gs points to _boot_cpu_pda and doesn't + * change. */ movl $MSR_GS_BASE,%ecx movq initial_gs(%rip),%rax @@ -278,7 +281,11 @@ ENTRY(secondary_startup_64) ENTRY(initial_code) .quad x86_64_start_kernel ENTRY(initial_gs) +#ifdef CONFIG_SMP + .quad __per_cpu_load +#else .quad _boot_cpu_pda +#endif __FINITDATA ENTRY(stack_start) diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index 73ab01b297c..63d46280227 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -13,6 +13,7 @@ #include #include #include +#include #include #ifdef CONFIG_DEBUG_PER_CPU_MAPS @@ -65,6 +66,36 @@ static void __init setup_node_to_cpumask_map(void); static inline void setup_node_to_cpumask_map(void) { } #endif +#ifdef CONFIG_X86_64 +void __cpuinit load_pda_offset(int cpu) +{ + /* Memory clobbers used to order pda/percpu accesses */ + mb(); + wrmsrl(MSR_GS_BASE, cpu_pda(cpu)); + mb(); +} + +#endif /* CONFIG_SMP && CONFIG_X86_64 */ + +#ifdef CONFIG_X86_64 + +/* correctly size the local cpu masks */ +static void setup_cpu_local_masks(void) +{ + alloc_bootmem_cpumask_var(&cpu_initialized_mask); + alloc_bootmem_cpumask_var(&cpu_callin_mask); + alloc_bootmem_cpumask_var(&cpu_callout_mask); + alloc_bootmem_cpumask_var(&cpu_sibling_setup_mask); +} + +#else /* CONFIG_X86_32 */ + +static inline void setup_cpu_local_masks(void) +{ +} + +#endif /* CONFIG_X86_32 */ + #ifdef CONFIG_HAVE_SETUP_PER_CPU_AREA /* * Copy data used in early init routines from the initial arrays to the @@ -101,63 +132,7 @@ static void __init setup_per_cpu_maps(void) */ unsigned long __per_cpu_offset[NR_CPUS] __read_mostly; EXPORT_SYMBOL(__per_cpu_offset); -static inline void setup_cpu_pda_map(void) { } - -#elif !defined(CONFIG_SMP) -static inline void setup_cpu_pda_map(void) { } - -#else /* CONFIG_SMP && CONFIG_X86_64 */ - -/* - * Allocate cpu_pda pointer table and array via alloc_bootmem. - */ -static void __init setup_cpu_pda_map(void) -{ - char *pda; - unsigned long size; - int cpu; - - size = roundup(sizeof(struct x8664_pda), cache_line_size()); - - /* allocate cpu_pda array and pointer table */ - { - unsigned long asize = size * (nr_cpu_ids - 1); - - pda = alloc_bootmem(asize); - } - - /* initialize pointer table to static pda's */ - for_each_possible_cpu(cpu) { - if (cpu == 0) { - /* leave boot cpu pda in place */ - continue; - } - cpu_pda(cpu) = (struct x8664_pda *)pda; - cpu_pda(cpu)->in_bootmem = 1; - pda += size; - } -} - -#endif /* CONFIG_SMP && CONFIG_X86_64 */ - -#ifdef CONFIG_X86_64 - -/* correctly size the local cpu masks */ -static void setup_cpu_local_masks(void) -{ - alloc_bootmem_cpumask_var(&cpu_initialized_mask); - alloc_bootmem_cpumask_var(&cpu_callin_mask); - alloc_bootmem_cpumask_var(&cpu_callout_mask); - alloc_bootmem_cpumask_var(&cpu_sibling_setup_mask); -} - -#else /* CONFIG_X86_32 */ - -static inline void setup_cpu_local_masks(void) -{ -} - -#endif /* CONFIG_X86_32 */ +#endif /* * Great future plan: @@ -171,9 +146,6 @@ void __init setup_per_cpu_areas(void) int cpu; unsigned long align = 1; - /* Setup cpu_pda map */ - setup_cpu_pda_map(); - /* Copy section for each CPU (we discard the original) */ old_size = PERCPU_ENOUGH_ROOM; align = max_t(unsigned long, PAGE_SIZE, align); @@ -204,8 +176,21 @@ void __init setup_per_cpu_areas(void) cpu, node, __pa(ptr)); } #endif - per_cpu_offset(cpu) = ptr - __per_cpu_start; + memcpy(ptr, __per_cpu_load, __per_cpu_end - __per_cpu_start); +#ifdef CONFIG_X86_64 + cpu_pda(cpu) = (void *)ptr; + + /* + * CPU0 modified pda in the init data area, reload pda + * offset for CPU0 and clear the area for others. + */ + if (cpu == 0) + load_pda_offset(0); + else + memset(cpu_pda(cpu), 0, sizeof(*cpu_pda(cpu))); +#endif + per_cpu_offset(cpu) = ptr - __per_cpu_start; DBG("PERCPU: cpu %4d %p\n", cpu, ptr); } diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c index 70d846628bb..f2f77ca494d 100644 --- a/arch/x86/kernel/smpboot.c +++ b/arch/x86/kernel/smpboot.c @@ -744,52 +744,6 @@ static void __cpuinit do_fork_idle(struct work_struct *work) complete(&c_idle->done); } -#ifdef CONFIG_X86_64 - -/* __ref because it's safe to call free_bootmem when after_bootmem == 0. */ -static void __ref free_bootmem_pda(struct x8664_pda *oldpda) -{ - if (!after_bootmem) - free_bootmem((unsigned long)oldpda, sizeof(*oldpda)); -} - -/* - * Allocate node local memory for the AP pda. - * - * Must be called after the _cpu_pda pointer table is initialized. - */ -int __cpuinit get_local_pda(int cpu) -{ - struct x8664_pda *oldpda, *newpda; - unsigned long size = sizeof(struct x8664_pda); - int node = cpu_to_node(cpu); - - if (cpu_pda(cpu) && !cpu_pda(cpu)->in_bootmem) - return 0; - - oldpda = cpu_pda(cpu); - newpda = kmalloc_node(size, GFP_ATOMIC, node); - if (!newpda) { - printk(KERN_ERR "Could not allocate node local PDA " - "for CPU %d on node %d\n", cpu, node); - - if (oldpda) - return 0; /* have a usable pda */ - else - return -1; - } - - if (oldpda) { - memcpy(newpda, oldpda, size); - free_bootmem_pda(oldpda); - } - - newpda->in_bootmem = 0; - cpu_pda(cpu) = newpda; - return 0; -} -#endif /* CONFIG_X86_64 */ - static int __cpuinit do_boot_cpu(int apicid, int cpu) /* * NOTE - on most systems this is a PHYSICAL apic ID, but on multiquad @@ -807,16 +761,6 @@ static int __cpuinit do_boot_cpu(int apicid, int cpu) }; INIT_WORK(&c_idle.work, do_fork_idle); -#ifdef CONFIG_X86_64 - /* Allocate node local memory for AP pdas */ - if (cpu > 0) { - boot_error = get_local_pda(cpu); - if (boot_error) - goto restore_state; - /* if can't get pda memory, can't start cpu */ - } -#endif - alternatives_smp_switch(1); c_idle.idle = get_idle_for_cpu(cpu); @@ -931,9 +875,7 @@ do_rest: inquire_remote_apic(apicid); } } -#ifdef CONFIG_X86_64 -restore_state: -#endif + if (boot_error) { /* Try to put things back the way they were before ... */ numa_remove_cpu(cpu); /* was set by numa_add_cpu */ diff --git a/arch/x86/kernel/vmlinux_64.lds.S b/arch/x86/kernel/vmlinux_64.lds.S index f50280db0df..962f21f1d4d 100644 --- a/arch/x86/kernel/vmlinux_64.lds.S +++ b/arch/x86/kernel/vmlinux_64.lds.S @@ -5,6 +5,7 @@ #define LOAD_OFFSET __START_KERNEL_map #include +#include #include #undef i386 /* in case the preprocessor is a 32bit one */ @@ -215,10 +216,11 @@ SECTIONS /* * percpu offsets are zero-based on SMP. PERCPU_VADDR() changes the * output PHDR, so the next output section - __data_nosave - should - * switch it back to data.init. + * switch it back to data.init. Also, pda should be at the head of + * percpu area. Preallocate it. */ . = ALIGN(PAGE_SIZE); - PERCPU_VADDR(0, :percpu) + PERCPU_VADDR_PREALLOC(0, :percpu, pda_size) #else PERCPU(PAGE_SIZE) #endif diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c index c44e2069c7c..83fa4236477 100644 --- a/arch/x86/xen/smp.c +++ b/arch/x86/xen/smp.c @@ -283,16 +283,6 @@ static int __cpuinit xen_cpu_up(unsigned int cpu) struct task_struct *idle = idle_task(cpu); int rc; -#ifdef CONFIG_X86_64 - /* Allocate node local memory for AP pdas */ - WARN_ON(cpu == 0); - if (cpu > 0) { - rc = get_local_pda(cpu); - if (rc) - return rc; - } -#endif - #ifdef CONFIG_X86_32 init_gdt(cpu); per_cpu(current_task, cpu) = idle; diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h index fc2f55f2dcd..e53319cf29c 100644 --- a/include/asm-generic/vmlinux.lds.h +++ b/include/asm-generic/vmlinux.lds.h @@ -441,9 +441,10 @@ . = __per_cpu_load + SIZEOF(.data.percpu); /** - * PERCPU_VADDR - define output section for percpu area + * PERCPU_VADDR_PREALLOC - define output section for percpu area with prealloc * @vaddr: explicit base address (optional) * @phdr: destination PHDR (optional) + * @prealloc: the size of prealloc area * * Macro which expands to output section for percpu area. If @vaddr * is not blank, it specifies explicit base address and all percpu @@ -455,11 +456,33 @@ * section in the linker script will go there too. @phdr should have * a leading colon. * + * If @prealloc is non-zero, the specified number of bytes will be + * reserved at the start of percpu area. As the prealloc area is + * likely to break alignment, this macro puts areas in increasing + * alignment order. + * * This macro defines three symbols, __per_cpu_load, __per_cpu_start * and __per_cpu_end. The first one is the vaddr of loaded percpu * init data. __per_cpu_start equals @vaddr and __per_cpu_end is the * end offset. */ +#define PERCPU_VADDR_PREALLOC(vaddr, segment, prealloc) \ + PERCPU_PROLOG(vaddr) \ + . += prealloc; \ + *(.data.percpu) \ + *(.data.percpu.shared_aligned) \ + *(.data.percpu.page_aligned) \ + PERCPU_EPILOG(segment) + +/** + * PERCPU_VADDR - define output section for percpu area + * @vaddr: explicit base address (optional) + * @phdr: destination PHDR (optional) + * + * Macro which expands to output section for percpu area. Mostly + * identical to PERCPU_VADDR_PREALLOC(@vaddr, @phdr, 0) other than + * using slighly different layout. + */ #define PERCPU_VADDR(vaddr, phdr) \ PERCPU_PROLOG(vaddr) \ *(.data.percpu.page_aligned) \ -- cgit v1.2.3-70-g09d2 From 004aa322f855a765741d9437a98dd8fe2e4f32a6 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Tue, 13 Jan 2009 20:41:35 +0900 Subject: x86: misc clean up after the percpu update Do the following cleanups: * kill x86_64_init_pda() which now is equivalent to pda_init() * use per_cpu_offset() instead of cpu_pda() when initializing initial_gs Signed-off-by: Tejun Heo Signed-off-by: Ingo Molnar --- arch/x86/include/asm/setup.h | 1 - arch/x86/kernel/acpi/sleep.c | 2 +- arch/x86/kernel/head64.c | 7 +------ arch/x86/kernel/smpboot.c | 2 +- arch/x86/xen/enlighten.c | 2 +- 5 files changed, 4 insertions(+), 10 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/include/asm/setup.h b/arch/x86/include/asm/setup.h index ebe858cdc8a..536949749bc 100644 --- a/arch/x86/include/asm/setup.h +++ b/arch/x86/include/asm/setup.h @@ -100,7 +100,6 @@ extern unsigned long init_pg_tables_start; extern unsigned long init_pg_tables_end; #else -void __init x86_64_init_pda(void); void __init x86_64_start_kernel(char *real_mode); void __init x86_64_start_reservations(char *real_mode_data); diff --git a/arch/x86/kernel/acpi/sleep.c b/arch/x86/kernel/acpi/sleep.c index 9ff67f8dc2c..4abff454c55 100644 --- a/arch/x86/kernel/acpi/sleep.c +++ b/arch/x86/kernel/acpi/sleep.c @@ -101,7 +101,7 @@ int acpi_save_state_mem(void) stack_start.sp = temp_stack + sizeof(temp_stack); early_gdt_descr.address = (unsigned long)get_cpu_gdt_table(smp_processor_id()); - initial_gs = (unsigned long)cpu_pda(smp_processor_id()); + initial_gs = per_cpu_offset(smp_processor_id()); #endif initial_code = (unsigned long)wakeup_long64; saved_magic = 0x123456789abcdef0; diff --git a/arch/x86/kernel/head64.c b/arch/x86/kernel/head64.c index 71b6f6ec96a..af67d3227ea 100644 --- a/arch/x86/kernel/head64.c +++ b/arch/x86/kernel/head64.c @@ -26,11 +26,6 @@ #include #include -void __init x86_64_init_pda(void) -{ - pda_init(0); -} - static void __init zap_identity_mappings(void) { pgd_t *pgd = pgd_offset_k(0UL); @@ -96,7 +91,7 @@ void __init x86_64_start_kernel(char * real_mode_data) if (console_loglevel == 10) early_printk("Kernel alive\n"); - x86_64_init_pda(); + pda_init(0); x86_64_start_reservations(real_mode_data); } diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c index f2f77ca494d..2f0e0f1090f 100644 --- a/arch/x86/kernel/smpboot.c +++ b/arch/x86/kernel/smpboot.c @@ -798,7 +798,7 @@ do_rest: #else cpu_pda(cpu)->pcurrent = c_idle.idle; clear_tsk_thread_flag(c_idle.idle, TIF_FORK); - initial_gs = (unsigned long)cpu_pda(cpu); + initial_gs = per_cpu_offset(cpu); #endif early_gdt_descr.address = (unsigned long)get_cpu_gdt_table(cpu); initial_code = (unsigned long)start_secondary; diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 965539ec425..312414ef936 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1645,7 +1645,7 @@ asmlinkage void __init xen_start_kernel(void) #ifdef CONFIG_X86_64 /* Disable until direct per-cpu data access. */ have_vcpu_info_placement = 0; - x86_64_init_pda(); + pda_init(0); #endif xen_smp_init(); -- cgit v1.2.3-70-g09d2 From 6dbde3530850d4d8bfc1b6bd4006d92786a2787f Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Thu, 15 Jan 2009 22:15:53 +0900 Subject: percpu: add optimized generic percpu accessors It is an optimization and a cleanup, and adds the following new generic percpu methods: percpu_read() percpu_write() percpu_add() percpu_sub() percpu_and() percpu_or() percpu_xor() and implements support for them on x86. (other architectures will fall back to a default implementation) The advantage is that for example to read a local percpu variable, instead of this sequence: return __get_cpu_var(var); ffffffff8102ca2b: 48 8b 14 fd 80 09 74 mov -0x7e8bf680(,%rdi,8),%rdx ffffffff8102ca32: 81 ffffffff8102ca33: 48 c7 c0 d8 59 00 00 mov $0x59d8,%rax ffffffff8102ca3a: 48 8b 04 10 mov (%rax,%rdx,1),%rax We can get a single instruction by using the optimized variants: return percpu_read(var); ffffffff8102ca3f: 65 48 8b 05 91 8f fd mov %gs:0x7efd8f91(%rip),%rax I also cleaned up the x86-specific APIs and made the x86 code use these new generic percpu primitives. tj: * fixed generic percpu_sub() definition as Roel Kluin pointed out * added percpu_and() for completeness's sake * made generic percpu ops atomic against preemption Signed-off-by: Ingo Molnar Signed-off-by: Tejun Heo --- arch/x86/include/asm/current.h | 2 +- arch/x86/include/asm/irq_regs_32.h | 4 +-- arch/x86/include/asm/mmu_context_32.h | 12 ++++---- arch/x86/include/asm/pda.h | 10 +++---- arch/x86/include/asm/percpu.h | 24 ++++++++-------- arch/x86/include/asm/smp.h | 2 +- arch/x86/kernel/process_32.c | 2 +- arch/x86/kernel/tlb_32.c | 10 +++---- arch/x86/mach-voyager/voyager_smp.c | 4 +-- arch/x86/xen/enlighten.c | 14 +++++----- arch/x86/xen/irq.c | 8 +++--- arch/x86/xen/mmu.c | 2 +- arch/x86/xen/multicalls.h | 2 +- arch/x86/xen/smp.c | 2 +- include/asm-generic/percpu.h | 52 +++++++++++++++++++++++++++++++++++ 15 files changed, 102 insertions(+), 48 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/include/asm/current.h b/arch/x86/include/asm/current.h index 0930b4f8d67..0728480f5c5 100644 --- a/arch/x86/include/asm/current.h +++ b/arch/x86/include/asm/current.h @@ -10,7 +10,7 @@ struct task_struct; DECLARE_PER_CPU(struct task_struct *, current_task); static __always_inline struct task_struct *get_current(void) { - return x86_read_percpu(current_task); + return percpu_read(current_task); } #else /* X86_32 */ diff --git a/arch/x86/include/asm/irq_regs_32.h b/arch/x86/include/asm/irq_regs_32.h index 86afd747345..d7ed33ee94e 100644 --- a/arch/x86/include/asm/irq_regs_32.h +++ b/arch/x86/include/asm/irq_regs_32.h @@ -15,7 +15,7 @@ DECLARE_PER_CPU(struct pt_regs *, irq_regs); static inline struct pt_regs *get_irq_regs(void) { - return x86_read_percpu(irq_regs); + return percpu_read(irq_regs); } static inline struct pt_regs *set_irq_regs(struct pt_regs *new_regs) @@ -23,7 +23,7 @@ static inline struct pt_regs *set_irq_regs(struct pt_regs *new_regs) struct pt_regs *old_regs; old_regs = get_irq_regs(); - x86_write_percpu(irq_regs, new_regs); + percpu_write(irq_regs, new_regs); return old_regs; } diff --git a/arch/x86/include/asm/mmu_context_32.h b/arch/x86/include/asm/mmu_context_32.h index 7e98ce1d2c0..08b53454f83 100644 --- a/arch/x86/include/asm/mmu_context_32.h +++ b/arch/x86/include/asm/mmu_context_32.h @@ -4,8 +4,8 @@ static inline void enter_lazy_tlb(struct mm_struct *mm, struct task_struct *tsk) { #ifdef CONFIG_SMP - if (x86_read_percpu(cpu_tlbstate.state) == TLBSTATE_OK) - x86_write_percpu(cpu_tlbstate.state, TLBSTATE_LAZY); + if (percpu_read(cpu_tlbstate.state) == TLBSTATE_OK) + percpu_write(cpu_tlbstate.state, TLBSTATE_LAZY); #endif } @@ -19,8 +19,8 @@ static inline void switch_mm(struct mm_struct *prev, /* stop flush ipis for the previous mm */ cpu_clear(cpu, prev->cpu_vm_mask); #ifdef CONFIG_SMP - x86_write_percpu(cpu_tlbstate.state, TLBSTATE_OK); - x86_write_percpu(cpu_tlbstate.active_mm, next); + percpu_write(cpu_tlbstate.state, TLBSTATE_OK); + percpu_write(cpu_tlbstate.active_mm, next); #endif cpu_set(cpu, next->cpu_vm_mask); @@ -35,8 +35,8 @@ static inline void switch_mm(struct mm_struct *prev, } #ifdef CONFIG_SMP else { - x86_write_percpu(cpu_tlbstate.state, TLBSTATE_OK); - BUG_ON(x86_read_percpu(cpu_tlbstate.active_mm) != next); + percpu_write(cpu_tlbstate.state, TLBSTATE_OK); + BUG_ON(percpu_read(cpu_tlbstate.active_mm) != next); if (!cpu_test_and_set(cpu, next->cpu_vm_mask)) { /* We were in lazy tlb mode and leave_mm disabled diff --git a/arch/x86/include/asm/pda.h b/arch/x86/include/asm/pda.h index e3d3a081d79..47f274fe695 100644 --- a/arch/x86/include/asm/pda.h +++ b/arch/x86/include/asm/pda.h @@ -45,11 +45,11 @@ extern void pda_init(int); #define cpu_pda(cpu) (&per_cpu(__pda, cpu)) -#define read_pda(field) x86_read_percpu(__pda.field) -#define write_pda(field, val) x86_write_percpu(__pda.field, val) -#define add_pda(field, val) x86_add_percpu(__pda.field, val) -#define sub_pda(field, val) x86_sub_percpu(__pda.field, val) -#define or_pda(field, val) x86_or_percpu(__pda.field, val) +#define read_pda(field) percpu_read(__pda.field) +#define write_pda(field, val) percpu_write(__pda.field, val) +#define add_pda(field, val) percpu_add(__pda.field, val) +#define sub_pda(field, val) percpu_sub(__pda.field, val) +#define or_pda(field, val) percpu_or(__pda.field, val) /* This is not atomic against other CPUs -- CPU preemption needs to be off */ #define test_and_clear_bit_pda(bit, field) \ diff --git a/arch/x86/include/asm/percpu.h b/arch/x86/include/asm/percpu.h index 328b31a429d..03aa4b00a1c 100644 --- a/arch/x86/include/asm/percpu.h +++ b/arch/x86/include/asm/percpu.h @@ -40,16 +40,11 @@ #ifdef CONFIG_SMP #define __percpu_seg_str "%%"__stringify(__percpu_seg)":" -#define __my_cpu_offset x86_read_percpu(this_cpu_off) +#define __my_cpu_offset percpu_read(this_cpu_off) #else #define __percpu_seg_str #endif -#include - -/* We can use this directly for local CPU (faster). */ -DECLARE_PER_CPU(unsigned long, this_cpu_off); - /* For arch-specific code, we can use direct single-insn ops (they * don't give an lvalue though). */ extern void __bad_percpu_size(void); @@ -115,11 +110,13 @@ do { \ ret__; \ }) -#define x86_read_percpu(var) percpu_from_op("mov", per_cpu__##var) -#define x86_write_percpu(var, val) percpu_to_op("mov", per_cpu__##var, val) -#define x86_add_percpu(var, val) percpu_to_op("add", per_cpu__##var, val) -#define x86_sub_percpu(var, val) percpu_to_op("sub", per_cpu__##var, val) -#define x86_or_percpu(var, val) percpu_to_op("or", per_cpu__##var, val) +#define percpu_read(var) percpu_from_op("mov", per_cpu__##var) +#define percpu_write(var, val) percpu_to_op("mov", per_cpu__##var, val) +#define percpu_add(var, val) percpu_to_op("add", per_cpu__##var, val) +#define percpu_sub(var, val) percpu_to_op("sub", per_cpu__##var, val) +#define percpu_and(var, val) percpu_to_op("and", per_cpu__##var, val) +#define percpu_or(var, val) percpu_to_op("or", per_cpu__##var, val) +#define percpu_xor(var, val) percpu_to_op("xor", per_cpu__##var, val) /* This is not atomic against other CPUs -- CPU preemption needs to be off */ #define x86_test_and_clear_bit_percpu(bit, var) \ @@ -131,6 +128,11 @@ do { \ old__; \ }) +#include + +/* We can use this directly for local CPU (faster). */ +DECLARE_PER_CPU(unsigned long, this_cpu_off); + #ifdef CONFIG_X86_64 extern void load_pda_offset(int cpu); #else diff --git a/arch/x86/include/asm/smp.h b/arch/x86/include/asm/smp.h index 127415402ea..c7bbbbe65d3 100644 --- a/arch/x86/include/asm/smp.h +++ b/arch/x86/include/asm/smp.h @@ -160,7 +160,7 @@ extern unsigned disabled_cpus __cpuinitdata; * from the initial startup. We map APIC_BASE very early in page_setup(), * so this is correct in the x86 case. */ -#define raw_smp_processor_id() (x86_read_percpu(cpu_number)) +#define raw_smp_processor_id() (percpu_read(cpu_number)) extern int safe_smp_processor_id(void); #elif defined(CONFIG_X86_64_SMP) diff --git a/arch/x86/kernel/process_32.c b/arch/x86/kernel/process_32.c index a546f55c77b..77d546817d9 100644 --- a/arch/x86/kernel/process_32.c +++ b/arch/x86/kernel/process_32.c @@ -591,7 +591,7 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p) if (prev->gs | next->gs) loadsegment(gs, next->gs); - x86_write_percpu(current_task, next_p); + percpu_write(current_task, next_p); return prev_p; } diff --git a/arch/x86/kernel/tlb_32.c b/arch/x86/kernel/tlb_32.c index ec53818f4e3..e65449d0f7d 100644 --- a/arch/x86/kernel/tlb_32.c +++ b/arch/x86/kernel/tlb_32.c @@ -34,8 +34,8 @@ static DEFINE_SPINLOCK(tlbstate_lock); */ void leave_mm(int cpu) { - BUG_ON(x86_read_percpu(cpu_tlbstate.state) == TLBSTATE_OK); - cpu_clear(cpu, x86_read_percpu(cpu_tlbstate.active_mm)->cpu_vm_mask); + BUG_ON(percpu_read(cpu_tlbstate.state) == TLBSTATE_OK); + cpu_clear(cpu, percpu_read(cpu_tlbstate.active_mm)->cpu_vm_mask); load_cr3(swapper_pg_dir); } EXPORT_SYMBOL_GPL(leave_mm); @@ -103,8 +103,8 @@ void smp_invalidate_interrupt(struct pt_regs *regs) * BUG(); */ - if (flush_mm == x86_read_percpu(cpu_tlbstate.active_mm)) { - if (x86_read_percpu(cpu_tlbstate.state) == TLBSTATE_OK) { + if (flush_mm == percpu_read(cpu_tlbstate.active_mm)) { + if (percpu_read(cpu_tlbstate.state) == TLBSTATE_OK) { if (flush_va == TLB_FLUSH_ALL) local_flush_tlb(); else @@ -222,7 +222,7 @@ static void do_flush_tlb_all(void *info) unsigned long cpu = smp_processor_id(); __flush_tlb_all(); - if (x86_read_percpu(cpu_tlbstate.state) == TLBSTATE_LAZY) + if (percpu_read(cpu_tlbstate.state) == TLBSTATE_LAZY) leave_mm(cpu); } diff --git a/arch/x86/mach-voyager/voyager_smp.c b/arch/x86/mach-voyager/voyager_smp.c index 1a48368acb0..96f15b09a4c 100644 --- a/arch/x86/mach-voyager/voyager_smp.c +++ b/arch/x86/mach-voyager/voyager_smp.c @@ -402,7 +402,7 @@ void __init find_smp_config(void) VOYAGER_SUS_IN_CONTROL_PORT); current_thread_info()->cpu = boot_cpu_id; - x86_write_percpu(cpu_number, boot_cpu_id); + percpu_write(cpu_number, boot_cpu_id); } /* @@ -1782,7 +1782,7 @@ static void __init voyager_smp_cpus_done(unsigned int max_cpus) void __init smp_setup_processor_id(void) { current_thread_info()->cpu = hard_smp_processor_id(); - x86_write_percpu(cpu_number, hard_smp_processor_id()); + percpu_write(cpu_number, hard_smp_processor_id()); } static void voyager_send_call_func(cpumask_t callmask) diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 312414ef936..75b94139e1f 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -695,17 +695,17 @@ static void xen_write_cr0(unsigned long cr0) static void xen_write_cr2(unsigned long cr2) { - x86_read_percpu(xen_vcpu)->arch.cr2 = cr2; + percpu_read(xen_vcpu)->arch.cr2 = cr2; } static unsigned long xen_read_cr2(void) { - return x86_read_percpu(xen_vcpu)->arch.cr2; + return percpu_read(xen_vcpu)->arch.cr2; } static unsigned long xen_read_cr2_direct(void) { - return x86_read_percpu(xen_vcpu_info.arch.cr2); + return percpu_read(xen_vcpu_info.arch.cr2); } static void xen_write_cr4(unsigned long cr4) @@ -718,12 +718,12 @@ static void xen_write_cr4(unsigned long cr4) static unsigned long xen_read_cr3(void) { - return x86_read_percpu(xen_cr3); + return percpu_read(xen_cr3); } static void set_current_cr3(void *v) { - x86_write_percpu(xen_current_cr3, (unsigned long)v); + percpu_write(xen_current_cr3, (unsigned long)v); } static void __xen_write_cr3(bool kernel, unsigned long cr3) @@ -748,7 +748,7 @@ static void __xen_write_cr3(bool kernel, unsigned long cr3) MULTI_mmuext_op(mcs.mc, op, 1, NULL, DOMID_SELF); if (kernel) { - x86_write_percpu(xen_cr3, cr3); + percpu_write(xen_cr3, cr3); /* Update xen_current_cr3 once the batch has actually been submitted. */ @@ -764,7 +764,7 @@ static void xen_write_cr3(unsigned long cr3) /* Update while interrupts are disabled, so its atomic with respect to ipis */ - x86_write_percpu(xen_cr3, cr3); + percpu_write(xen_cr3, cr3); __xen_write_cr3(true, cr3); diff --git a/arch/x86/xen/irq.c b/arch/x86/xen/irq.c index bb042608c60..2e8271431e1 100644 --- a/arch/x86/xen/irq.c +++ b/arch/x86/xen/irq.c @@ -39,7 +39,7 @@ static unsigned long xen_save_fl(void) struct vcpu_info *vcpu; unsigned long flags; - vcpu = x86_read_percpu(xen_vcpu); + vcpu = percpu_read(xen_vcpu); /* flag has opposite sense of mask */ flags = !vcpu->evtchn_upcall_mask; @@ -62,7 +62,7 @@ static void xen_restore_fl(unsigned long flags) make sure we're don't switch CPUs between getting the vcpu pointer and updating the mask. */ preempt_disable(); - vcpu = x86_read_percpu(xen_vcpu); + vcpu = percpu_read(xen_vcpu); vcpu->evtchn_upcall_mask = flags; preempt_enable_no_resched(); @@ -83,7 +83,7 @@ static void xen_irq_disable(void) make sure we're don't switch CPUs between getting the vcpu pointer and updating the mask. */ preempt_disable(); - x86_read_percpu(xen_vcpu)->evtchn_upcall_mask = 1; + percpu_read(xen_vcpu)->evtchn_upcall_mask = 1; preempt_enable_no_resched(); } @@ -96,7 +96,7 @@ static void xen_irq_enable(void) the caller is confused and is trying to re-enable interrupts on an indeterminate processor. */ - vcpu = x86_read_percpu(xen_vcpu); + vcpu = percpu_read(xen_vcpu); vcpu->evtchn_upcall_mask = 0; /* Doesn't matter if we get preempted here, because any diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 503c240e26c..7bc7852cc5c 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -1074,7 +1074,7 @@ static void drop_other_mm_ref(void *info) /* If this cpu still has a stale cr3 reference, then make sure it has been flushed. */ - if (x86_read_percpu(xen_current_cr3) == __pa(mm->pgd)) { + if (percpu_read(xen_current_cr3) == __pa(mm->pgd)) { load_cr3(swapper_pg_dir); arch_flush_lazy_cpu_mode(); } diff --git a/arch/x86/xen/multicalls.h b/arch/x86/xen/multicalls.h index 85893824161..e786fa7f261 100644 --- a/arch/x86/xen/multicalls.h +++ b/arch/x86/xen/multicalls.h @@ -39,7 +39,7 @@ static inline void xen_mc_issue(unsigned mode) xen_mc_flush(); /* restore flags saved in xen_mc_batch */ - local_irq_restore(x86_read_percpu(xen_mc_irq_flags)); + local_irq_restore(percpu_read(xen_mc_irq_flags)); } /* Set up a callback to be called when the current batch is flushed */ diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c index 83fa4236477..3bfd6dd0b47 100644 --- a/arch/x86/xen/smp.c +++ b/arch/x86/xen/smp.c @@ -78,7 +78,7 @@ static __cpuinit void cpu_bringup(void) xen_setup_cpu_clockevents(); cpu_set(cpu, cpu_online_map); - x86_write_percpu(cpu_state, CPU_ONLINE); + percpu_write(cpu_state, CPU_ONLINE); wmb(); /* We can take interrupts now: we're officially "up". */ diff --git a/include/asm-generic/percpu.h b/include/asm-generic/percpu.h index b0e63c672eb..00f45ff081a 100644 --- a/include/asm-generic/percpu.h +++ b/include/asm-generic/percpu.h @@ -80,4 +80,56 @@ extern void setup_per_cpu_areas(void); #define DECLARE_PER_CPU(type, name) extern PER_CPU_ATTRIBUTES \ __typeof__(type) per_cpu_var(name) +/* + * Optional methods for optimized non-lvalue per-cpu variable access. + * + * @var can be a percpu variable or a field of it and its size should + * equal char, int or long. percpu_read() evaluates to a lvalue and + * all others to void. + * + * These operations are guaranteed to be atomic w.r.t. preemption. + * The generic versions use plain get/put_cpu_var(). Archs are + * encouraged to implement single-instruction alternatives which don't + * require preemption protection. + */ +#ifndef percpu_read +# define percpu_read(var) \ + ({ \ + typeof(per_cpu_var(var)) __tmp_var__; \ + __tmp_var__ = get_cpu_var(var); \ + put_cpu_var(var); \ + __tmp_var__; \ + }) +#endif + +#define __percpu_generic_to_op(var, val, op) \ +do { \ + get_cpu_var(var) op val; \ + put_cpu_var(var); \ +} while (0) + +#ifndef percpu_write +# define percpu_write(var, val) __percpu_generic_to_op(var, (val), =) +#endif + +#ifndef percpu_add +# define percpu_add(var, val) __percpu_generic_to_op(var, (val), +=) +#endif + +#ifndef percpu_sub +# define percpu_sub(var, val) __percpu_generic_to_op(var, (val), -=) +#endif + +#ifndef percpu_and +# define percpu_and(var, val) __percpu_generic_to_op(var, (val), &=) +#endif + +#ifndef percpu_or +# define percpu_or(var, val) __percpu_generic_to_op(var, (val), |=) +#endif + +#ifndef percpu_xor +# define percpu_xor(var, val) __percpu_generic_to_op(var, (val), ^=) +#endif + #endif /* _ASM_GENERIC_PERCPU_H_ */ -- cgit v1.2.3-70-g09d2 From 1b437c8c73a36daa471dd54a63c426d72af5723d Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Mon, 19 Jan 2009 00:38:57 +0900 Subject: x86-64: Move irq stats from PDA to per-cpu and consolidate with 32-bit. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/include/asm/hardirq_64.h | 24 +++++++++++++++++++----- arch/x86/include/asm/pda.h | 10 ---------- arch/x86/kernel/irq.c | 6 +----- arch/x86/kernel/irq_64.c | 3 +++ arch/x86/kernel/nmi.c | 10 +--------- arch/x86/xen/smp.c | 18 +++--------------- 6 files changed, 27 insertions(+), 44 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/include/asm/hardirq_64.h b/arch/x86/include/asm/hardirq_64.h index b5a6b5d5670..a65bab20f6c 100644 --- a/arch/x86/include/asm/hardirq_64.h +++ b/arch/x86/include/asm/hardirq_64.h @@ -3,22 +3,36 @@ #include #include -#include #include +typedef struct { + unsigned int __softirq_pending; + unsigned int __nmi_count; /* arch dependent */ + unsigned int apic_timer_irqs; /* arch dependent */ + unsigned int irq0_irqs; + unsigned int irq_resched_count; + unsigned int irq_call_count; + unsigned int irq_tlb_count; + unsigned int irq_thermal_count; + unsigned int irq_spurious_count; + unsigned int irq_threshold_count; +} ____cacheline_aligned irq_cpustat_t; + +DECLARE_PER_CPU(irq_cpustat_t, irq_stat); + /* We can have at most NR_VECTORS irqs routed to a cpu at a time */ #define MAX_HARDIRQS_PER_CPU NR_VECTORS #define __ARCH_IRQ_STAT 1 -#define inc_irq_stat(member) add_pda(member, 1) +#define inc_irq_stat(member) percpu_add(irq_stat.member, 1) -#define local_softirq_pending() read_pda(__softirq_pending) +#define local_softirq_pending() percpu_read(irq_stat.__softirq_pending) #define __ARCH_SET_SOFTIRQ_PENDING 1 -#define set_softirq_pending(x) write_pda(__softirq_pending, (x)) -#define or_softirq_pending(x) or_pda(__softirq_pending, (x)) +#define set_softirq_pending(x) percpu_write(irq_stat.__softirq_pending, (x)) +#define or_softirq_pending(x) percpu_or(irq_stat.__softirq_pending, (x)) extern void ack_bad_irq(unsigned int irq); diff --git a/arch/x86/include/asm/pda.h b/arch/x86/include/asm/pda.h index 47f274fe695..69a40757e21 100644 --- a/arch/x86/include/asm/pda.h +++ b/arch/x86/include/asm/pda.h @@ -25,19 +25,9 @@ struct x8664_pda { char *irqstackptr; short nodenumber; /* number of current node (32k max) */ short in_bootmem; /* pda lives in bootmem */ - unsigned int __softirq_pending; - unsigned int __nmi_count; /* number of NMI on this CPUs */ short mmu_state; short isidle; struct mm_struct *active_mm; - unsigned apic_timer_irqs; - unsigned irq0_irqs; - unsigned irq_resched_count; - unsigned irq_call_count; - unsigned irq_tlb_count; - unsigned irq_thermal_count; - unsigned irq_threshold_count; - unsigned irq_spurious_count; } ____cacheline_aligned_in_smp; DECLARE_PER_CPU(struct x8664_pda, __pda); diff --git a/arch/x86/kernel/irq.c b/arch/x86/kernel/irq.c index 3973e2df7f8..8b30d0c2512 100644 --- a/arch/x86/kernel/irq.c +++ b/arch/x86/kernel/irq.c @@ -36,11 +36,7 @@ void ack_bad_irq(unsigned int irq) #endif } -#ifdef CONFIG_X86_32 -# define irq_stats(x) (&per_cpu(irq_stat, x)) -#else -# define irq_stats(x) cpu_pda(x) -#endif +#define irq_stats(x) (&per_cpu(irq_stat, x)) /* * /proc/interrupts printing: */ diff --git a/arch/x86/kernel/irq_64.c b/arch/x86/kernel/irq_64.c index 0b21cb1ea11..1db05247b47 100644 --- a/arch/x86/kernel/irq_64.c +++ b/arch/x86/kernel/irq_64.c @@ -19,6 +19,9 @@ #include #include +DEFINE_PER_CPU_SHARED_ALIGNED(irq_cpustat_t, irq_stat); +EXPORT_PER_CPU_SYMBOL(irq_stat); + /* * Probabilistic stack overflow check: * diff --git a/arch/x86/kernel/nmi.c b/arch/x86/kernel/nmi.c index 7228979f1e7..23b6d9e6e4f 100644 --- a/arch/x86/kernel/nmi.c +++ b/arch/x86/kernel/nmi.c @@ -61,11 +61,7 @@ static int endflag __initdata; static inline unsigned int get_nmi_count(int cpu) { -#ifdef CONFIG_X86_64 - return cpu_pda(cpu)->__nmi_count; -#else - return nmi_count(cpu); -#endif + return per_cpu(irq_stat, cpu).__nmi_count; } static inline int mce_in_progress(void) @@ -82,12 +78,8 @@ static inline int mce_in_progress(void) */ static inline unsigned int get_timer_irqs(int cpu) { -#ifdef CONFIG_X86_64 - return read_pda(apic_timer_irqs) + read_pda(irq0_irqs); -#else return per_cpu(irq_stat, cpu).apic_timer_irqs + per_cpu(irq_stat, cpu).irq0_irqs; -#endif } #ifdef CONFIG_SMP diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c index 3bfd6dd0b47..9ff3b0999cf 100644 --- a/arch/x86/xen/smp.c +++ b/arch/x86/xen/smp.c @@ -50,11 +50,7 @@ static irqreturn_t xen_call_function_single_interrupt(int irq, void *dev_id); */ static irqreturn_t xen_reschedule_interrupt(int irq, void *dev_id) { -#ifdef CONFIG_X86_32 - __get_cpu_var(irq_stat).irq_resched_count++; -#else - add_pda(irq_resched_count, 1); -#endif + inc_irq_stat(irq_resched_count); return IRQ_HANDLED; } @@ -435,11 +431,7 @@ static irqreturn_t xen_call_function_interrupt(int irq, void *dev_id) { irq_enter(); generic_smp_call_function_interrupt(); -#ifdef CONFIG_X86_32 - __get_cpu_var(irq_stat).irq_call_count++; -#else - add_pda(irq_call_count, 1); -#endif + inc_irq_stat(irq_call_count); irq_exit(); return IRQ_HANDLED; @@ -449,11 +441,7 @@ static irqreturn_t xen_call_function_single_interrupt(int irq, void *dev_id) { irq_enter(); generic_smp_call_function_single_interrupt(); -#ifdef CONFIG_X86_32 - __get_cpu_var(irq_stat).irq_call_count++; -#else - add_pda(irq_call_count, 1); -#endif + inc_irq_stat(irq_call_count); irq_exit(); return IRQ_HANDLED; -- cgit v1.2.3-70-g09d2 From 9eb912d1aa6b8106e06a73ea6702ec3dab0d6a1a Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Mon, 19 Jan 2009 00:38:57 +0900 Subject: x86-64: Move TLB state from PDA to per-cpu and consolidate with 32-bit. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/include/asm/mmu_context_64.h | 16 +++++++--------- arch/x86/include/asm/pda.h | 2 -- arch/x86/include/asm/tlbflush.h | 7 ++----- arch/x86/kernel/cpu/common.c | 2 -- arch/x86/kernel/tlb_32.c | 12 ++---------- arch/x86/kernel/tlb_64.c | 13 ++++++++----- arch/x86/xen/mmu.c | 6 +----- 7 files changed, 20 insertions(+), 38 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/include/asm/mmu_context_64.h b/arch/x86/include/asm/mmu_context_64.h index 677d36e9540..c4572505ab3 100644 --- a/arch/x86/include/asm/mmu_context_64.h +++ b/arch/x86/include/asm/mmu_context_64.h @@ -1,13 +1,11 @@ #ifndef _ASM_X86_MMU_CONTEXT_64_H #define _ASM_X86_MMU_CONTEXT_64_H -#include - static inline void enter_lazy_tlb(struct mm_struct *mm, struct task_struct *tsk) { #ifdef CONFIG_SMP - if (read_pda(mmu_state) == TLBSTATE_OK) - write_pda(mmu_state, TLBSTATE_LAZY); + if (percpu_read(cpu_tlbstate.state) == TLBSTATE_OK) + percpu_write(cpu_tlbstate.state, TLBSTATE_LAZY); #endif } @@ -19,8 +17,8 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next, /* stop flush ipis for the previous mm */ cpu_clear(cpu, prev->cpu_vm_mask); #ifdef CONFIG_SMP - write_pda(mmu_state, TLBSTATE_OK); - write_pda(active_mm, next); + percpu_write(cpu_tlbstate.state, TLBSTATE_OK); + percpu_write(cpu_tlbstate.active_mm, next); #endif cpu_set(cpu, next->cpu_vm_mask); load_cr3(next->pgd); @@ -30,9 +28,9 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next, } #ifdef CONFIG_SMP else { - write_pda(mmu_state, TLBSTATE_OK); - if (read_pda(active_mm) != next) - BUG(); + percpu_write(cpu_tlbstate.state, TLBSTATE_OK); + BUG_ON(percpu_read(cpu_tlbstate.active_mm) != next); + if (!cpu_test_and_set(cpu, next->cpu_vm_mask)) { /* We were in lazy tlb mode and leave_mm disabled * tlb flush IPI delivery. We must reload CR3 diff --git a/arch/x86/include/asm/pda.h b/arch/x86/include/asm/pda.h index 69a40757e21..8ee835ed10e 100644 --- a/arch/x86/include/asm/pda.h +++ b/arch/x86/include/asm/pda.h @@ -25,9 +25,7 @@ struct x8664_pda { char *irqstackptr; short nodenumber; /* number of current node (32k max) */ short in_bootmem; /* pda lives in bootmem */ - short mmu_state; short isidle; - struct mm_struct *active_mm; } ____cacheline_aligned_in_smp; DECLARE_PER_CPU(struct x8664_pda, __pda); diff --git a/arch/x86/include/asm/tlbflush.h b/arch/x86/include/asm/tlbflush.h index 17feaa9c7e7..d3539f998f8 100644 --- a/arch/x86/include/asm/tlbflush.h +++ b/arch/x86/include/asm/tlbflush.h @@ -148,20 +148,17 @@ void native_flush_tlb_others(const struct cpumask *cpumask, #define TLBSTATE_OK 1 #define TLBSTATE_LAZY 2 -#ifdef CONFIG_X86_32 struct tlb_state { struct mm_struct *active_mm; int state; - char __cacheline_padding[L1_CACHE_BYTES-8]; }; DECLARE_PER_CPU(struct tlb_state, cpu_tlbstate); -void reset_lazy_tlbstate(void); -#else static inline void reset_lazy_tlbstate(void) { + percpu_write(cpu_tlbstate.state, 0); + percpu_write(cpu_tlbstate.active_mm, &init_mm); } -#endif #endif /* SMP */ diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index c49498d4083..3d0cc6f1711 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -897,8 +897,6 @@ void __cpuinit pda_init(int cpu) pda->irqcount = -1; pda->kernelstack = (unsigned long)stack_thread_info() - PDA_STACKOFFSET + THREAD_SIZE; - pda->active_mm = &init_mm; - pda->mmu_state = 0; if (cpu == 0) { /* others are initialized in smpboot.c */ diff --git a/arch/x86/kernel/tlb_32.c b/arch/x86/kernel/tlb_32.c index e65449d0f7d..abf0808d6fc 100644 --- a/arch/x86/kernel/tlb_32.c +++ b/arch/x86/kernel/tlb_32.c @@ -4,8 +4,8 @@ #include -DEFINE_PER_CPU(struct tlb_state, cpu_tlbstate) - ____cacheline_aligned = { &init_mm, 0, }; +DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate) + = { &init_mm, 0, }; /* must come after the send_IPI functions above for inlining */ #include @@ -231,14 +231,6 @@ void flush_tlb_all(void) on_each_cpu(do_flush_tlb_all, NULL, 1); } -void reset_lazy_tlbstate(void) -{ - int cpu = raw_smp_processor_id(); - - per_cpu(cpu_tlbstate, cpu).state = 0; - per_cpu(cpu_tlbstate, cpu).active_mm = &init_mm; -} - static int init_flush_cpumask(void) { alloc_cpumask_var(&flush_cpumask, GFP_KERNEL); diff --git a/arch/x86/kernel/tlb_64.c b/arch/x86/kernel/tlb_64.c index 7f4141d3b66..e64a32c4882 100644 --- a/arch/x86/kernel/tlb_64.c +++ b/arch/x86/kernel/tlb_64.c @@ -18,6 +18,9 @@ #include #include +DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate) + = { &init_mm, 0, }; + #include /* * Smarter SMP flushing macros. @@ -62,9 +65,9 @@ static DEFINE_PER_CPU(union smp_flush_state, flush_state); */ void leave_mm(int cpu) { - if (read_pda(mmu_state) == TLBSTATE_OK) + if (percpu_read(cpu_tlbstate.state) == TLBSTATE_OK) BUG(); - cpu_clear(cpu, read_pda(active_mm)->cpu_vm_mask); + cpu_clear(cpu, percpu_read(cpu_tlbstate.active_mm)->cpu_vm_mask); load_cr3(swapper_pg_dir); } EXPORT_SYMBOL_GPL(leave_mm); @@ -142,8 +145,8 @@ asmlinkage void smp_invalidate_interrupt(struct pt_regs *regs) * BUG(); */ - if (f->flush_mm == read_pda(active_mm)) { - if (read_pda(mmu_state) == TLBSTATE_OK) { + if (f->flush_mm == percpu_read(cpu_tlbstate.active_mm)) { + if (percpu_read(cpu_tlbstate.state) == TLBSTATE_OK) { if (f->flush_va == TLB_FLUSH_ALL) local_flush_tlb(); else @@ -281,7 +284,7 @@ static void do_flush_tlb_all(void *info) unsigned long cpu = smp_processor_id(); __flush_tlb_all(); - if (read_pda(mmu_state) == TLBSTATE_LAZY) + if (percpu_read(cpu_tlbstate.state) == TLBSTATE_LAZY) leave_mm(cpu); } diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 7bc7852cc5c..98cb9869eb2 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -1063,11 +1063,7 @@ static void drop_other_mm_ref(void *info) struct mm_struct *mm = info; struct mm_struct *active_mm; -#ifdef CONFIG_X86_64 - active_mm = read_pda(active_mm); -#else - active_mm = __get_cpu_var(cpu_tlbstate).active_mm; -#endif + active_mm = percpu_read(cpu_tlbstate.active_mm); if (active_mm == mm) leave_mm(smp_processor_id()); -- cgit v1.2.3-70-g09d2 From c6f5e0acd5d12ee23f701f15889872e67b47caa6 Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Mon, 19 Jan 2009 00:38:58 +0900 Subject: x86-64: Move current task from PDA to per-cpu and consolidate with 32-bit. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/include/asm/current.h | 24 +++--------------------- arch/x86/include/asm/pda.h | 4 ++-- arch/x86/include/asm/system.h | 4 ++-- arch/x86/kernel/asm-offsets_64.c | 1 - arch/x86/kernel/cpu/common.c | 5 +---- arch/x86/kernel/dumpstack_64.c | 2 +- arch/x86/kernel/process_64.c | 5 ++++- arch/x86/kernel/smpboot.c | 3 +-- arch/x86/xen/smp.c | 3 +-- 9 files changed, 15 insertions(+), 36 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/include/asm/current.h b/arch/x86/include/asm/current.h index 0728480f5c5..c68c361697e 100644 --- a/arch/x86/include/asm/current.h +++ b/arch/x86/include/asm/current.h @@ -1,39 +1,21 @@ #ifndef _ASM_X86_CURRENT_H #define _ASM_X86_CURRENT_H -#ifdef CONFIG_X86_32 #include #include +#ifndef __ASSEMBLY__ struct task_struct; DECLARE_PER_CPU(struct task_struct *, current_task); -static __always_inline struct task_struct *get_current(void) -{ - return percpu_read(current_task); -} - -#else /* X86_32 */ - -#ifndef __ASSEMBLY__ -#include - -struct task_struct; static __always_inline struct task_struct *get_current(void) { - return read_pda(pcurrent); + return percpu_read(current_task); } -#else /* __ASSEMBLY__ */ - -#include -#define GET_CURRENT(reg) movq %gs:(pda_pcurrent),reg +#define current get_current() #endif /* __ASSEMBLY__ */ -#endif /* X86_32 */ - -#define current get_current() - #endif /* _ASM_X86_CURRENT_H */ diff --git a/arch/x86/include/asm/pda.h b/arch/x86/include/asm/pda.h index 668d5a5b6f7..7209302d922 100644 --- a/arch/x86/include/asm/pda.h +++ b/arch/x86/include/asm/pda.h @@ -11,8 +11,8 @@ /* Per processor datastructure. %gs points to it while the kernel runs */ struct x8664_pda { - struct task_struct *pcurrent; /* 0 Current process */ - unsigned long dummy; + unsigned long unused1; + unsigned long unused2; unsigned long kernelstack; /* 16 top of kernel stack for current */ unsigned long oldrsp; /* 24 user rsp for system call */ int irqcount; /* 32 Irq nesting counter. Starts -1 */ diff --git a/arch/x86/include/asm/system.h b/arch/x86/include/asm/system.h index 8e626ea33a1..4399aac680e 100644 --- a/arch/x86/include/asm/system.h +++ b/arch/x86/include/asm/system.h @@ -94,7 +94,7 @@ do { \ "call __switch_to\n\t" \ ".globl thread_return\n" \ "thread_return:\n\t" \ - "movq %%gs:%P[pda_pcurrent],%%rsi\n\t" \ + "movq "__percpu_seg_str"%P[current_task],%%rsi\n\t" \ "movq %P[thread_info](%%rsi),%%r8\n\t" \ LOCK_PREFIX "btr %[tif_fork],%P[ti_flags](%%r8)\n\t" \ "movq %%rax,%%rdi\n\t" \ @@ -106,7 +106,7 @@ do { \ [ti_flags] "i" (offsetof(struct thread_info, flags)), \ [tif_fork] "i" (TIF_FORK), \ [thread_info] "i" (offsetof(struct task_struct, stack)), \ - [pda_pcurrent] "i" (offsetof(struct x8664_pda, pcurrent)) \ + [current_task] "m" (per_cpu_var(current_task)) \ : "memory", "cc" __EXTRA_CLOBBER) #endif diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c index cae6697c099..4f7a210e1e5 100644 --- a/arch/x86/kernel/asm-offsets_64.c +++ b/arch/x86/kernel/asm-offsets_64.c @@ -51,7 +51,6 @@ int main(void) #define ENTRY(entry) DEFINE(pda_ ## entry, offsetof(struct x8664_pda, entry)) ENTRY(kernelstack); ENTRY(oldrsp); - ENTRY(pcurrent); ENTRY(irqcount); DEFINE(pda_size, sizeof(struct x8664_pda)); BLANK(); diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index 4221e920886..b50e38d1688 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -903,10 +903,7 @@ void __cpuinit pda_init(int cpu) pda->kernelstack = (unsigned long)stack_thread_info() - PDA_STACKOFFSET + THREAD_SIZE; - if (cpu == 0) { - /* others are initialized in smpboot.c */ - pda->pcurrent = &init_task; - } else { + if (cpu != 0) { if (pda->nodenumber == 0 && cpu_to_node(cpu) != NUMA_NO_NODE) pda->nodenumber = cpu_to_node(cpu); } diff --git a/arch/x86/kernel/dumpstack_64.c b/arch/x86/kernel/dumpstack_64.c index 28e26a4315d..d35db5993fd 100644 --- a/arch/x86/kernel/dumpstack_64.c +++ b/arch/x86/kernel/dumpstack_64.c @@ -242,7 +242,7 @@ void show_registers(struct pt_regs *regs) int i; unsigned long sp; const int cpu = smp_processor_id(); - struct task_struct *cur = cpu_pda(cpu)->pcurrent; + struct task_struct *cur = current; sp = regs->sp; printk("CPU %d ", cpu); diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c index 416fb9282f4..e00c31a4b3c 100644 --- a/arch/x86/kernel/process_64.c +++ b/arch/x86/kernel/process_64.c @@ -57,6 +57,9 @@ asmlinkage extern void ret_from_fork(void); +DEFINE_PER_CPU(struct task_struct *, current_task) = &init_task; +EXPORT_PER_CPU_SYMBOL(current_task); + unsigned long kernel_thread_flags = CLONE_VM | CLONE_UNTRACED; static ATOMIC_NOTIFIER_HEAD(idle_notifier); @@ -615,7 +618,7 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p) */ prev->usersp = read_pda(oldrsp); write_pda(oldrsp, next->usersp); - write_pda(pcurrent, next_p); + percpu_write(current_task, next_p); write_pda(kernelstack, (unsigned long)task_stack_page(next_p) + diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c index 2f0e0f1090f..5854be0fb80 100644 --- a/arch/x86/kernel/smpboot.c +++ b/arch/x86/kernel/smpboot.c @@ -790,13 +790,12 @@ static int __cpuinit do_boot_cpu(int apicid, int cpu) set_idle_for_cpu(cpu, c_idle.idle); do_rest: -#ifdef CONFIG_X86_32 per_cpu(current_task, cpu) = c_idle.idle; +#ifdef CONFIG_X86_32 init_gdt(cpu); /* Stack for startup_32 can be just as for start_secondary onwards */ irq_ctx_init(cpu); #else - cpu_pda(cpu)->pcurrent = c_idle.idle; clear_tsk_thread_flag(c_idle.idle, TIF_FORK); initial_gs = per_cpu_offset(cpu); #endif diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c index 9ff3b0999cf..72c2eb9b64c 100644 --- a/arch/x86/xen/smp.c +++ b/arch/x86/xen/smp.c @@ -279,12 +279,11 @@ static int __cpuinit xen_cpu_up(unsigned int cpu) struct task_struct *idle = idle_task(cpu); int rc; + per_cpu(current_task, cpu) = idle; #ifdef CONFIG_X86_32 init_gdt(cpu); - per_cpu(current_task, cpu) = idle; irq_ctx_init(cpu); #else - cpu_pda(cpu)->pcurrent = idle; clear_tsk_thread_flag(idle, TIF_FORK); #endif xen_setup_timer(cpu); -- cgit v1.2.3-70-g09d2 From 9af45651f1f7c89942e016a1a00a7ebddfa727f8 Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Mon, 19 Jan 2009 00:38:58 +0900 Subject: x86-64: Move kernelstack from PDA to per-cpu. Also clean up PER_CPU_VAR usage in xen-asm_64.S tj: * remove now unused stack_thread_info() * s/kernelstack/kernel_stack/ * added FIXME comment in xen-asm_64.S Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/ia32/ia32entry.S | 8 ++++---- arch/x86/include/asm/pda.h | 4 +--- arch/x86/include/asm/thread_info.h | 20 ++++++++------------ arch/x86/kernel/asm-offsets_64.c | 1 - arch/x86/kernel/cpu/common.c | 6 ++++-- arch/x86/kernel/entry_64.S | 4 ++-- arch/x86/kernel/process_64.c | 4 ++-- arch/x86/kernel/smpboot.c | 3 +++ arch/x86/xen/xen-asm_64.S | 23 +++++++++++------------ 9 files changed, 35 insertions(+), 38 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/ia32/ia32entry.S b/arch/x86/ia32/ia32entry.S index 256b00b6189..9c79b247700 100644 --- a/arch/x86/ia32/ia32entry.S +++ b/arch/x86/ia32/ia32entry.S @@ -112,8 +112,8 @@ ENTRY(ia32_sysenter_target) CFI_DEF_CFA rsp,0 CFI_REGISTER rsp,rbp SWAPGS_UNSAFE_STACK - movq %gs:pda_kernelstack, %rsp - addq $(PDA_STACKOFFSET),%rsp + movq PER_CPU_VAR(kernel_stack), %rsp + addq $(KERNEL_STACK_OFFSET),%rsp /* * No need to follow this irqs on/off section: the syscall * disabled irqs, here we enable it straight after entry: @@ -273,13 +273,13 @@ ENDPROC(ia32_sysenter_target) ENTRY(ia32_cstar_target) CFI_STARTPROC32 simple CFI_SIGNAL_FRAME - CFI_DEF_CFA rsp,PDA_STACKOFFSET + CFI_DEF_CFA rsp,KERNEL_STACK_OFFSET CFI_REGISTER rip,rcx /*CFI_REGISTER rflags,r11*/ SWAPGS_UNSAFE_STACK movl %esp,%r8d CFI_REGISTER rsp,r8 - movq %gs:pda_kernelstack,%rsp + movq PER_CPU_VAR(kernel_stack),%rsp /* * No need to follow this irqs on/off section: the syscall * disabled irqs and here we enable it straight after entry: diff --git a/arch/x86/include/asm/pda.h b/arch/x86/include/asm/pda.h index 7209302d922..4d28ffba6e1 100644 --- a/arch/x86/include/asm/pda.h +++ b/arch/x86/include/asm/pda.h @@ -13,7 +13,7 @@ struct x8664_pda { unsigned long unused1; unsigned long unused2; - unsigned long kernelstack; /* 16 top of kernel stack for current */ + unsigned long unused3; unsigned long oldrsp; /* 24 user rsp for system call */ int irqcount; /* 32 Irq nesting counter. Starts -1 */ unsigned int unused6; /* 36 was cpunumber */ @@ -44,6 +44,4 @@ extern void pda_init(int); #endif -#define PDA_STACKOFFSET (5*8) - #endif /* _ASM_X86_PDA_H */ diff --git a/arch/x86/include/asm/thread_info.h b/arch/x86/include/asm/thread_info.h index 98789647baa..b46f8ca007b 100644 --- a/arch/x86/include/asm/thread_info.h +++ b/arch/x86/include/asm/thread_info.h @@ -194,25 +194,21 @@ static inline struct thread_info *current_thread_info(void) #else /* X86_32 */ -#include +#include +#define KERNEL_STACK_OFFSET (5*8) /* * macros/functions for gaining access to the thread information structure * preempt_count needs to be 1 initially, until the scheduler is functional. */ #ifndef __ASSEMBLY__ -static inline struct thread_info *current_thread_info(void) -{ - struct thread_info *ti; - ti = (void *)(read_pda(kernelstack) + PDA_STACKOFFSET - THREAD_SIZE); - return ti; -} +DECLARE_PER_CPU(unsigned long, kernel_stack); -/* do not use in interrupt context */ -static inline struct thread_info *stack_thread_info(void) +static inline struct thread_info *current_thread_info(void) { struct thread_info *ti; - asm("andq %%rsp,%0; " : "=r" (ti) : "0" (~(THREAD_SIZE - 1))); + ti = (void *)(percpu_read(kernel_stack) + + KERNEL_STACK_OFFSET - THREAD_SIZE); return ti; } @@ -220,8 +216,8 @@ static inline struct thread_info *stack_thread_info(void) /* how to get the thread information struct from ASM */ #define GET_THREAD_INFO(reg) \ - movq %gs:pda_kernelstack,reg ; \ - subq $(THREAD_SIZE-PDA_STACKOFFSET),reg + movq PER_CPU_VAR(kernel_stack),reg ; \ + subq $(THREAD_SIZE-KERNEL_STACK_OFFSET),reg #endif diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c index 4f7a210e1e5..cafff5f4a03 100644 --- a/arch/x86/kernel/asm-offsets_64.c +++ b/arch/x86/kernel/asm-offsets_64.c @@ -49,7 +49,6 @@ int main(void) BLANK(); #undef ENTRY #define ENTRY(entry) DEFINE(pda_ ## entry, offsetof(struct x8664_pda, entry)) - ENTRY(kernelstack); ENTRY(oldrsp); ENTRY(irqcount); DEFINE(pda_size, sizeof(struct x8664_pda)); diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index b50e38d1688..06b6290088f 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -889,6 +889,10 @@ DEFINE_PER_CPU(char *, irq_stack_ptr) = per_cpu_var(irq_stack) + IRQ_STACK_SIZE - 64; #endif +DEFINE_PER_CPU(unsigned long, kernel_stack) = + (unsigned long)&init_thread_union - KERNEL_STACK_OFFSET + THREAD_SIZE; +EXPORT_PER_CPU_SYMBOL(kernel_stack); + void __cpuinit pda_init(int cpu) { struct x8664_pda *pda = cpu_pda(cpu); @@ -900,8 +904,6 @@ void __cpuinit pda_init(int cpu) load_pda_offset(cpu); pda->irqcount = -1; - pda->kernelstack = (unsigned long)stack_thread_info() - - PDA_STACKOFFSET + THREAD_SIZE; if (cpu != 0) { if (pda->nodenumber == 0 && cpu_to_node(cpu) != NUMA_NO_NODE) diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S index d22677a6643..0dd45859a7a 100644 --- a/arch/x86/kernel/entry_64.S +++ b/arch/x86/kernel/entry_64.S @@ -468,7 +468,7 @@ END(ret_from_fork) ENTRY(system_call) CFI_STARTPROC simple CFI_SIGNAL_FRAME - CFI_DEF_CFA rsp,PDA_STACKOFFSET + CFI_DEF_CFA rsp,KERNEL_STACK_OFFSET CFI_REGISTER rip,rcx /*CFI_REGISTER rflags,r11*/ SWAPGS_UNSAFE_STACK @@ -480,7 +480,7 @@ ENTRY(system_call) ENTRY(system_call_after_swapgs) movq %rsp,%gs:pda_oldrsp - movq %gs:pda_kernelstack,%rsp + movq PER_CPU_VAR(kernel_stack),%rsp /* * No need to follow this irqs off/on section - it's straight * and short: diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c index e00c31a4b3c..6c5f5760210 100644 --- a/arch/x86/kernel/process_64.c +++ b/arch/x86/kernel/process_64.c @@ -620,9 +620,9 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p) write_pda(oldrsp, next->usersp); percpu_write(current_task, next_p); - write_pda(kernelstack, + percpu_write(kernel_stack, (unsigned long)task_stack_page(next_p) + - THREAD_SIZE - PDA_STACKOFFSET); + THREAD_SIZE - KERNEL_STACK_OFFSET); #ifdef CONFIG_CC_STACKPROTECTOR write_pda(stack_canary, next_p->stack_canary); /* diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c index 5854be0fb80..869b98840fd 100644 --- a/arch/x86/kernel/smpboot.c +++ b/arch/x86/kernel/smpboot.c @@ -798,6 +798,9 @@ do_rest: #else clear_tsk_thread_flag(c_idle.idle, TIF_FORK); initial_gs = per_cpu_offset(cpu); + per_cpu(kernel_stack, cpu) = + (unsigned long)task_stack_page(c_idle.idle) - + KERNEL_STACK_OFFSET + THREAD_SIZE; #endif early_gdt_descr.address = (unsigned long)get_cpu_gdt_table(cpu); initial_code = (unsigned long)start_secondary; diff --git a/arch/x86/xen/xen-asm_64.S b/arch/x86/xen/xen-asm_64.S index 05794c566e8..5a23e899367 100644 --- a/arch/x86/xen/xen-asm_64.S +++ b/arch/x86/xen/xen-asm_64.S @@ -17,6 +17,7 @@ #include #include #include +#include #include @@ -28,12 +29,10 @@ #if 1 /* - x86-64 does not yet support direct access to percpu variables - via a segment override, so we just need to make sure this code - never gets used + FIXME: x86_64 now can support direct access to percpu variables + via a segment override. Update xen accordingly. */ #define BUG ud2a -#define PER_CPU_VAR(var, off) 0xdeadbeef #endif /* @@ -45,14 +44,14 @@ ENTRY(xen_irq_enable_direct) BUG /* Unmask events */ - movb $0, PER_CPU_VAR(xen_vcpu_info, XEN_vcpu_info_mask) + movb $0, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_mask /* Preempt here doesn't matter because that will deal with any pending interrupts. The pending check may end up being run on the wrong CPU, but that doesn't hurt. */ /* Test for pending */ - testb $0xff, PER_CPU_VAR(xen_vcpu_info, XEN_vcpu_info_pending) + testb $0xff, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_pending jz 1f 2: call check_events @@ -69,7 +68,7 @@ ENDPATCH(xen_irq_enable_direct) ENTRY(xen_irq_disable_direct) BUG - movb $1, PER_CPU_VAR(xen_vcpu_info, XEN_vcpu_info_mask) + movb $1, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_mask ENDPATCH(xen_irq_disable_direct) ret ENDPROC(xen_irq_disable_direct) @@ -87,7 +86,7 @@ ENDPATCH(xen_irq_disable_direct) ENTRY(xen_save_fl_direct) BUG - testb $0xff, PER_CPU_VAR(xen_vcpu_info, XEN_vcpu_info_mask) + testb $0xff, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_mask setz %ah addb %ah,%ah ENDPATCH(xen_save_fl_direct) @@ -107,13 +106,13 @@ ENTRY(xen_restore_fl_direct) BUG testb $X86_EFLAGS_IF>>8, %ah - setz PER_CPU_VAR(xen_vcpu_info, XEN_vcpu_info_mask) + setz PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_mask /* Preempt here doesn't matter because that will deal with any pending interrupts. The pending check may end up being run on the wrong CPU, but that doesn't hurt. */ /* check for unmasked and pending */ - cmpw $0x0001, PER_CPU_VAR(xen_vcpu_info, XEN_vcpu_info_pending) + cmpw $0x0001, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_pending jz 1f 2: call check_events 1: @@ -196,7 +195,7 @@ ENTRY(xen_sysret64) /* We're already on the usermode stack at this point, but still with the kernel gs, so we can easily switch back */ movq %rsp, %gs:pda_oldrsp - movq %gs:pda_kernelstack,%rsp + movq PER_CPU_VAR(kernel_stack),%rsp pushq $__USER_DS pushq %gs:pda_oldrsp @@ -213,7 +212,7 @@ ENTRY(xen_sysret32) /* We're already on the usermode stack at this point, but still with the kernel gs, so we can easily switch back */ movq %rsp, %gs:pda_oldrsp - movq %gs:pda_kernelstack, %rsp + movq PER_CPU_VAR(kernel_stack), %rsp pushq $__USER32_DS pushq %gs:pda_oldrsp -- cgit v1.2.3-70-g09d2 From 3d1e42a7cf945e289d6ba26159aa0e2b0645401b Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Mon, 19 Jan 2009 00:38:58 +0900 Subject: x86-64: Move oldrsp from PDA to per-cpu. tj: * in asm-offsets_64.c, pda.h inclusion shouldn't be removed as pda is still referenced in the file * s/oldrsp/old_rsp/ Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/include/asm/pda.h | 2 +- arch/x86/kernel/asm-offsets_64.c | 1 - arch/x86/kernel/entry_64.S | 10 +++++----- arch/x86/kernel/process_64.c | 8 +++++--- arch/x86/xen/xen-asm_64.S | 8 ++++---- 5 files changed, 15 insertions(+), 14 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/include/asm/pda.h b/arch/x86/include/asm/pda.h index 4d28ffba6e1..ae23deb9955 100644 --- a/arch/x86/include/asm/pda.h +++ b/arch/x86/include/asm/pda.h @@ -14,7 +14,7 @@ struct x8664_pda { unsigned long unused1; unsigned long unused2; unsigned long unused3; - unsigned long oldrsp; /* 24 user rsp for system call */ + unsigned long unused4; int irqcount; /* 32 Irq nesting counter. Starts -1 */ unsigned int unused6; /* 36 was cpunumber */ #ifdef CONFIG_CC_STACKPROTECTOR diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c index cafff5f4a03..afda6deb851 100644 --- a/arch/x86/kernel/asm-offsets_64.c +++ b/arch/x86/kernel/asm-offsets_64.c @@ -49,7 +49,6 @@ int main(void) BLANK(); #undef ENTRY #define ENTRY(entry) DEFINE(pda_ ## entry, offsetof(struct x8664_pda, entry)) - ENTRY(oldrsp); ENTRY(irqcount); DEFINE(pda_size, sizeof(struct x8664_pda)); BLANK(); diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S index 0dd45859a7a..7c27da407da 100644 --- a/arch/x86/kernel/entry_64.S +++ b/arch/x86/kernel/entry_64.S @@ -210,7 +210,7 @@ ENTRY(native_usergs_sysret64) /* %rsp:at FRAMEEND */ .macro FIXUP_TOP_OF_STACK tmp offset=0 - movq %gs:pda_oldrsp,\tmp + movq PER_CPU_VAR(old_rsp),\tmp movq \tmp,RSP+\offset(%rsp) movq $__USER_DS,SS+\offset(%rsp) movq $__USER_CS,CS+\offset(%rsp) @@ -221,7 +221,7 @@ ENTRY(native_usergs_sysret64) .macro RESTORE_TOP_OF_STACK tmp offset=0 movq RSP+\offset(%rsp),\tmp - movq \tmp,%gs:pda_oldrsp + movq \tmp,PER_CPU_VAR(old_rsp) movq EFLAGS+\offset(%rsp),\tmp movq \tmp,R11+\offset(%rsp) .endm @@ -479,7 +479,7 @@ ENTRY(system_call) */ ENTRY(system_call_after_swapgs) - movq %rsp,%gs:pda_oldrsp + movq %rsp,PER_CPU_VAR(old_rsp) movq PER_CPU_VAR(kernel_stack),%rsp /* * No need to follow this irqs off/on section - it's straight @@ -523,7 +523,7 @@ sysret_check: CFI_REGISTER rip,rcx RESTORE_ARGS 0,-ARG_SKIP,1 /*CFI_REGISTER rflags,r11*/ - movq %gs:pda_oldrsp, %rsp + movq PER_CPU_VAR(old_rsp), %rsp USERGS_SYSRET64 CFI_RESTORE_STATE @@ -833,7 +833,7 @@ common_interrupt: XCPT_FRAME addq $-0x80,(%rsp) /* Adjust vector to [-256,-1] range */ interrupt do_IRQ - /* 0(%rsp): oldrsp-ARGOFFSET */ + /* 0(%rsp): old_rsp-ARGOFFSET */ ret_from_intr: DISABLE_INTERRUPTS(CLBR_NONE) TRACE_IRQS_OFF diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c index 6c5f5760210..48012891892 100644 --- a/arch/x86/kernel/process_64.c +++ b/arch/x86/kernel/process_64.c @@ -60,6 +60,8 @@ asmlinkage extern void ret_from_fork(void); DEFINE_PER_CPU(struct task_struct *, current_task) = &init_task; EXPORT_PER_CPU_SYMBOL(current_task); +DEFINE_PER_CPU(unsigned long, old_rsp); + unsigned long kernel_thread_flags = CLONE_VM | CLONE_UNTRACED; static ATOMIC_NOTIFIER_HEAD(idle_notifier); @@ -395,7 +397,7 @@ start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp) load_gs_index(0); regs->ip = new_ip; regs->sp = new_sp; - write_pda(oldrsp, new_sp); + percpu_write(old_rsp, new_sp); regs->cs = __USER_CS; regs->ss = __USER_DS; regs->flags = 0x200; @@ -616,8 +618,8 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p) /* * Switch the PDA and FPU contexts. */ - prev->usersp = read_pda(oldrsp); - write_pda(oldrsp, next->usersp); + prev->usersp = percpu_read(old_rsp); + percpu_write(old_rsp, next->usersp); percpu_write(current_task, next_p); percpu_write(kernel_stack, diff --git a/arch/x86/xen/xen-asm_64.S b/arch/x86/xen/xen-asm_64.S index 5a23e899367..d6fc51f4ce8 100644 --- a/arch/x86/xen/xen-asm_64.S +++ b/arch/x86/xen/xen-asm_64.S @@ -194,11 +194,11 @@ RELOC(xen_sysexit, 1b+1) ENTRY(xen_sysret64) /* We're already on the usermode stack at this point, but still with the kernel gs, so we can easily switch back */ - movq %rsp, %gs:pda_oldrsp + movq %rsp, PER_CPU_VAR(old_rsp) movq PER_CPU_VAR(kernel_stack),%rsp pushq $__USER_DS - pushq %gs:pda_oldrsp + pushq PER_CPU_VAR(old_rsp) pushq %r11 pushq $__USER_CS pushq %rcx @@ -211,11 +211,11 @@ RELOC(xen_sysret64, 1b+1) ENTRY(xen_sysret32) /* We're already on the usermode stack at this point, but still with the kernel gs, so we can easily switch back */ - movq %rsp, %gs:pda_oldrsp + movq %rsp, PER_CPU_VAR(old_rsp) movq PER_CPU_VAR(kernel_stack), %rsp pushq $__USER32_DS - pushq %gs:pda_oldrsp + pushq PER_CPU_VAR(old_rsp) pushq %r11 pushq $__USER32_CS pushq %rcx -- cgit v1.2.3-70-g09d2 From 8ce031972b40da58c268caba8c5ea3c0856d7131 Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Mon, 19 Jan 2009 12:21:27 +0900 Subject: x86: remove pda_init() Impact: cleanup Copy the code to cpu_init() to satisfy the requirement that the cpu be reinitialized. Remove all other calls, since the segments are already initialized in head_64.S. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/include/asm/pda.h | 1 - arch/x86/kernel/cpu/common.c | 15 +++------------ arch/x86/kernel/head64.c | 2 -- arch/x86/xen/enlighten.c | 1 - 4 files changed, 3 insertions(+), 16 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/include/asm/pda.h b/arch/x86/include/asm/pda.h index 4a8c9d382c9..b473e952439 100644 --- a/arch/x86/include/asm/pda.h +++ b/arch/x86/include/asm/pda.h @@ -24,7 +24,6 @@ struct x8664_pda { } ____cacheline_aligned_in_smp; DECLARE_PER_CPU(struct x8664_pda, __pda); -extern void pda_init(int); #define cpu_pda(cpu) (&per_cpu(__pda, cpu)) diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index 7976a6a0f65..f83a4d6160f 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -895,15 +895,6 @@ EXPORT_PER_CPU_SYMBOL(kernel_stack); DEFINE_PER_CPU(unsigned int, irq_count) = -1; -void __cpuinit pda_init(int cpu) -{ - /* Setup up data that may be needed in __get_free_pages early */ - loadsegment(fs, 0); - loadsegment(gs, 0); - - load_pda_offset(cpu); -} - static DEFINE_PER_CPU_PAGE_ALIGNED(char, exception_stacks [(N_EXCEPTION_STACKS - 1) * EXCEPTION_STKSZ + DEBUG_STKSZ]) __aligned(PAGE_SIZE); @@ -967,9 +958,9 @@ void __cpuinit cpu_init(void) struct task_struct *me; int i; - /* CPU 0 is initialised in head64.c */ - if (cpu != 0) - pda_init(cpu); + loadsegment(fs, 0); + loadsegment(gs, 0); + load_pda_offset(cpu); #ifdef CONFIG_NUMA if (cpu != 0 && percpu_read(node_number) == 0 && diff --git a/arch/x86/kernel/head64.c b/arch/x86/kernel/head64.c index af67d3227ea..f5b27224769 100644 --- a/arch/x86/kernel/head64.c +++ b/arch/x86/kernel/head64.c @@ -91,8 +91,6 @@ void __init x86_64_start_kernel(char * real_mode_data) if (console_loglevel == 10) early_printk("Kernel alive\n"); - pda_init(0); - x86_64_start_reservations(real_mode_data); } diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 75b94139e1f..bef941f6145 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1645,7 +1645,6 @@ asmlinkage void __init xen_start_kernel(void) #ifdef CONFIG_X86_64 /* Disable until direct per-cpu data access. */ have_vcpu_info_placement = 0; - pda_init(0); #endif xen_smp_init(); -- cgit v1.2.3-70-g09d2 From ab897d2013128f470240a541b31cf5e636984e71 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Thu, 22 Jan 2009 14:24:16 -0800 Subject: x86/pvops: remove pte_flags pvop pte_flags() was introduced as a new pvop in order to extract just the flags portion of a pte, which is a potentially cheaper operation than extracting the page number as well. It turns out this operation is not needed, because simply using a mask to extract the flags from a pte is sufficient for all current users. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Ingo Molnar --- arch/x86/include/asm/page.h | 3 +-- arch/x86/include/asm/paravirt.h | 18 ------------------ arch/x86/kernel/paravirt.c | 1 - arch/x86/xen/enlighten.c | 1 - 4 files changed, 1 insertion(+), 22 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/include/asm/page.h b/arch/x86/include/asm/page.h index e9873a2e869..6b9810859da 100644 --- a/arch/x86/include/asm/page.h +++ b/arch/x86/include/asm/page.h @@ -147,7 +147,7 @@ static inline pteval_t native_pte_val(pte_t pte) return pte.pte; } -static inline pteval_t native_pte_flags(pte_t pte) +static inline pteval_t pte_flags(pte_t pte) { return native_pte_val(pte) & PTE_FLAGS_MASK; } @@ -173,7 +173,6 @@ static inline pteval_t native_pte_flags(pte_t pte) #endif #define pte_val(x) native_pte_val(x) -#define pte_flags(x) native_pte_flags(x) #define __pte(x) native_make_pte(x) #endif /* CONFIG_PARAVIRT */ diff --git a/arch/x86/include/asm/paravirt.h b/arch/x86/include/asm/paravirt.h index ba3e2ff6aed..e25c410f3d8 100644 --- a/arch/x86/include/asm/paravirt.h +++ b/arch/x86/include/asm/paravirt.h @@ -279,7 +279,6 @@ struct pv_mmu_ops { pte_t *ptep, pte_t pte); pteval_t (*pte_val)(pte_t); - pteval_t (*pte_flags)(pte_t); pte_t (*make_pte)(pteval_t pte); pgdval_t (*pgd_val)(pgd_t); @@ -1084,23 +1083,6 @@ static inline pteval_t pte_val(pte_t pte) return ret; } -static inline pteval_t pte_flags(pte_t pte) -{ - pteval_t ret; - - if (sizeof(pteval_t) > sizeof(long)) - ret = PVOP_CALL2(pteval_t, pv_mmu_ops.pte_flags, - pte.pte, (u64)pte.pte >> 32); - else - ret = PVOP_CALL1(pteval_t, pv_mmu_ops.pte_flags, - pte.pte); - -#ifdef CONFIG_PARAVIRT_DEBUG - BUG_ON(ret & PTE_PFN_MASK); -#endif - return ret; -} - static inline pgd_t __pgd(pgdval_t val) { pgdval_t ret; diff --git a/arch/x86/kernel/paravirt.c b/arch/x86/kernel/paravirt.c index e4c8fb60887..202514be592 100644 --- a/arch/x86/kernel/paravirt.c +++ b/arch/x86/kernel/paravirt.c @@ -435,7 +435,6 @@ struct pv_mmu_ops pv_mmu_ops = { #endif /* PAGETABLE_LEVELS >= 3 */ .pte_val = native_pte_val, - .pte_flags = native_pte_flags, .pgd_val = native_pgd_val, .make_pte = native_make_pte, diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index bea215230b2..6f1bb71aa13 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1314,7 +1314,6 @@ static const struct pv_mmu_ops xen_mmu_ops __initdata = { .ptep_modify_prot_commit = __ptep_modify_prot_commit, .pte_val = xen_pte_val, - .pte_flags = native_pte_flags, .pgd_val = xen_pgd_val, .make_pte = xen_make_pte, -- cgit v1.2.3-70-g09d2 From 99d0000f710f3432182761f65f9658f1cf0bf455 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Fri, 23 Jan 2009 11:09:15 +0100 Subject: x86, xen: fix hardirq.h merge fallout Impact: build fix This build error: arch/x86/xen/suspend.c:22: error: implicit declaration of function 'fix_to_virt' arch/x86/xen/suspend.c:22: error: 'FIX_PARAVIRT_BOOTMAP' undeclared (first use in this function) arch/x86/xen/suspend.c:22: error: (Each undeclared identifier is reported only once arch/x86/xen/suspend.c:22: error: for each function it appears in.) triggers because the hardirq.h unification removed an implicit fixmap.h include - on which arch/x86/xen/suspend.c depended. Add the fixmap.h include explicitly. Signed-off-by: Ingo Molnar --- arch/x86/xen/suspend.c | 1 + 1 file changed, 1 insertion(+) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/suspend.c b/arch/x86/xen/suspend.c index 212ffe012b7..95be7b43472 100644 --- a/arch/x86/xen/suspend.c +++ b/arch/x86/xen/suspend.c @@ -6,6 +6,7 @@ #include #include +#include #include "xen-ops.h" #include "mmu.h" -- cgit v1.2.3-70-g09d2 From b2d2f4312b117a6cc647c8521e2643a88771f757 Mon Sep 17 00:00:00 2001 From: Brian Gerst Date: Tue, 27 Jan 2009 12:56:48 +0900 Subject: x86: initialize per-cpu GDT segment in per-cpu setup Impact: cleanup Rename init_gdt() to setup_percpu_segment(), and move it to setup_percpu.c. Signed-off-by: Brian Gerst Signed-off-by: Tejun Heo --- arch/x86/include/asm/processor.h | 1 - arch/x86/kernel/Makefile | 3 +-- arch/x86/kernel/setup_percpu.c | 14 ++++++++++++++ arch/x86/kernel/smpboot.c | 4 ---- arch/x86/kernel/smpcommon.c | 25 ------------------------- arch/x86/mach-voyager/voyager_smp.c | 2 -- arch/x86/xen/smp.c | 1 - 7 files changed, 15 insertions(+), 35 deletions(-) delete mode 100644 arch/x86/kernel/smpcommon.c (limited to 'arch/x86/xen') diff --git a/arch/x86/include/asm/processor.h b/arch/x86/include/asm/processor.h index 48676b943b9..32c30b02b51 100644 --- a/arch/x86/include/asm/processor.h +++ b/arch/x86/include/asm/processor.h @@ -778,7 +778,6 @@ extern struct desc_ptr early_gdt_descr; extern void cpu_set_gdt(int); extern void switch_to_new_gdt(void); extern void cpu_init(void); -extern void init_gdt(int cpu); static inline unsigned long get_debugctlmsr(void) { diff --git a/arch/x86/kernel/Makefile b/arch/x86/kernel/Makefile index 73de055c29c..37fa30bada1 100644 --- a/arch/x86/kernel/Makefile +++ b/arch/x86/kernel/Makefile @@ -60,8 +60,7 @@ obj-$(CONFIG_APM) += apm.o obj-$(CONFIG_X86_SMP) += smp.o obj-$(CONFIG_X86_SMP) += smpboot.o tsc_sync.o ipi.o obj-$(CONFIG_SMP) += setup_percpu.o -obj-$(CONFIG_X86_32_SMP) += smpcommon.o -obj-$(CONFIG_X86_64_SMP) += tsc_sync.o smpcommon.o +obj-$(CONFIG_X86_64_SMP) += tsc_sync.o obj-$(CONFIG_X86_TRAMPOLINE) += trampoline_$(BITS).o obj-$(CONFIG_X86_MPPARSE) += mpparse.o obj-$(CONFIG_X86_LOCAL_APIC) += apic.o nmi.o diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index 599dc1cc1da..bcca3a7b374 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -40,6 +40,19 @@ unsigned long __per_cpu_offset[NR_CPUS] __read_mostly = { }; EXPORT_SYMBOL(__per_cpu_offset); +static inline void setup_percpu_segment(int cpu) +{ +#ifdef CONFIG_X86_32 + struct desc_struct gdt; + + pack_descriptor(&gdt, per_cpu_offset(cpu), 0xFFFFF, + 0x2 | DESCTYPE_S, 0x8); + gdt.s = 1; + write_gdt_entry(get_cpu_gdt_table(cpu), + GDT_ENTRY_PERCPU, &gdt, DESCTYPE_S); +#endif +} + /* * Great future plan: * Declare PDA itself and support (irqstack,tss,pgd) as per cpu data. @@ -81,6 +94,7 @@ void __init setup_per_cpu_areas(void) per_cpu_offset(cpu) = ptr - __per_cpu_start; per_cpu(this_cpu_off, cpu) = per_cpu_offset(cpu); per_cpu(cpu_number, cpu) = cpu; + setup_percpu_segment(cpu); /* * Copy data used in early init routines from the initial arrays to the * per cpu data areas. These arrays then become expendable and the diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c index def770b57b5..f9dbcff4354 100644 --- a/arch/x86/kernel/smpboot.c +++ b/arch/x86/kernel/smpboot.c @@ -793,7 +793,6 @@ static int __cpuinit do_boot_cpu(int apicid, int cpu) do_rest: per_cpu(current_task, cpu) = c_idle.idle; #ifdef CONFIG_X86_32 - init_gdt(cpu); /* Stack for startup_32 can be just as for start_secondary onwards */ irq_ctx_init(cpu); #else @@ -1186,9 +1185,6 @@ out: void __init native_smp_prepare_boot_cpu(void) { int me = smp_processor_id(); -#ifdef CONFIG_X86_32 - init_gdt(me); -#endif switch_to_new_gdt(); /* already set me in cpu_online_mask in boot_cpu_init() */ cpumask_set_cpu(me, cpu_callout_mask); diff --git a/arch/x86/kernel/smpcommon.c b/arch/x86/kernel/smpcommon.c deleted file mode 100644 index 5ec29a1a846..00000000000 --- a/arch/x86/kernel/smpcommon.c +++ /dev/null @@ -1,25 +0,0 @@ -/* - * SMP stuff which is common to all sub-architectures. - */ -#include -#include -#include - -#ifdef CONFIG_X86_32 -/* - * Initialize the CPU's GDT. This is either the boot CPU doing itself - * (still using the master per-cpu area), or a CPU doing it for a - * secondary which will soon come up. - */ -__cpuinit void init_gdt(int cpu) -{ - struct desc_struct gdt; - - pack_descriptor(&gdt, __per_cpu_offset[cpu], 0xFFFFF, - 0x2 | DESCTYPE_S, 0x8); - gdt.s = 1; - - write_gdt_entry(get_cpu_gdt_table(cpu), - GDT_ENTRY_PERCPU, &gdt, DESCTYPE_S); -} -#endif diff --git a/arch/x86/mach-voyager/voyager_smp.c b/arch/x86/mach-voyager/voyager_smp.c index dd82f2052f3..331cd6d5648 100644 --- a/arch/x86/mach-voyager/voyager_smp.c +++ b/arch/x86/mach-voyager/voyager_smp.c @@ -530,7 +530,6 @@ static void __init do_boot_cpu(__u8 cpu) /* init_tasks (in sched.c) is indexed logically */ stack_start.sp = (void *)idle->thread.sp; - init_gdt(cpu); per_cpu(current_task, cpu) = idle; early_gdt_descr.address = (unsigned long)get_cpu_gdt_table(cpu); irq_ctx_init(cpu); @@ -1747,7 +1746,6 @@ static void __init voyager_smp_prepare_cpus(unsigned int max_cpus) static void __cpuinit voyager_smp_prepare_boot_cpu(void) { - init_gdt(smp_processor_id()); switch_to_new_gdt(); cpu_set(smp_processor_id(), cpu_online_map); diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c index 72c2eb9b64c..7735e3dd359 100644 --- a/arch/x86/xen/smp.c +++ b/arch/x86/xen/smp.c @@ -281,7 +281,6 @@ static int __cpuinit xen_cpu_up(unsigned int cpu) per_cpu(current_task, cpu) = idle; #ifdef CONFIG_X86_32 - init_gdt(cpu); irq_ctx_init(cpu); #else clear_tsk_thread_flag(idle, TIF_FORK); -- cgit v1.2.3-70-g09d2 From 319f3ba52c71630865b10ac3b99dd020440d681d Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Wed, 28 Jan 2009 14:35:01 -0800 Subject: xen: move remaining mmu-related stuff into mmu.c Impact: Cleanup Move remaining mmu-related stuff into mmu.c. A general cleanup, and lay the groundwork for later patches. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: H. Peter Anvin --- arch/x86/xen/enlighten.c | 731 +---------------------------------------------- arch/x86/xen/mmu.c | 700 +++++++++++++++++++++++++++++++++++++++++++++ arch/x86/xen/mmu.h | 3 + arch/x86/xen/xen-ops.h | 8 + 4 files changed, 712 insertions(+), 730 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 6b3f7eef57e..0cd2a165f17 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -61,35 +61,6 @@ DEFINE_PER_CPU(struct vcpu_info, xen_vcpu_info); enum xen_domain_type xen_domain_type = XEN_NATIVE; EXPORT_SYMBOL_GPL(xen_domain_type); -/* - * Identity map, in addition to plain kernel map. This needs to be - * large enough to allocate page table pages to allocate the rest. - * Each page can map 2MB. - */ -static pte_t level1_ident_pgt[PTRS_PER_PTE * 4] __page_aligned_bss; - -#ifdef CONFIG_X86_64 -/* l3 pud for userspace vsyscall mapping */ -static pud_t level3_user_vsyscall[PTRS_PER_PUD] __page_aligned_bss; -#endif /* CONFIG_X86_64 */ - -/* - * Note about cr3 (pagetable base) values: - * - * xen_cr3 contains the current logical cr3 value; it contains the - * last set cr3. This may not be the current effective cr3, because - * its update may be being lazily deferred. However, a vcpu looking - * at its own cr3 can use this value knowing that it everything will - * be self-consistent. - * - * xen_current_cr3 contains the actual vcpu cr3; it is set once the - * hypercall to set the vcpu cr3 is complete (so it may be a little - * out of date, but it will never be set early). If one vcpu is - * looking at another vcpu's cr3 value, it should use this variable. - */ -DEFINE_PER_CPU(unsigned long, xen_cr3); /* cr3 stored as physaddr */ -DEFINE_PER_CPU(unsigned long, xen_current_cr3); /* actual vcpu cr3 */ - struct start_info *xen_start_info; EXPORT_SYMBOL_GPL(xen_start_info); @@ -237,7 +208,7 @@ static unsigned long xen_get_debugreg(int reg) return HYPERVISOR_get_debugreg(reg); } -static void xen_leave_lazy(void) +void xen_leave_lazy(void) { paravirt_leave_lazy(paravirt_get_lazy_mode()); xen_mc_flush(); @@ -598,76 +569,6 @@ static struct apic_ops xen_basic_apic_ops = { #endif -static void xen_flush_tlb(void) -{ - struct mmuext_op *op; - struct multicall_space mcs; - - preempt_disable(); - - mcs = xen_mc_entry(sizeof(*op)); - - op = mcs.args; - op->cmd = MMUEXT_TLB_FLUSH_LOCAL; - MULTI_mmuext_op(mcs.mc, op, 1, NULL, DOMID_SELF); - - xen_mc_issue(PARAVIRT_LAZY_MMU); - - preempt_enable(); -} - -static void xen_flush_tlb_single(unsigned long addr) -{ - struct mmuext_op *op; - struct multicall_space mcs; - - preempt_disable(); - - mcs = xen_mc_entry(sizeof(*op)); - op = mcs.args; - op->cmd = MMUEXT_INVLPG_LOCAL; - op->arg1.linear_addr = addr & PAGE_MASK; - MULTI_mmuext_op(mcs.mc, op, 1, NULL, DOMID_SELF); - - xen_mc_issue(PARAVIRT_LAZY_MMU); - - preempt_enable(); -} - -static void xen_flush_tlb_others(const struct cpumask *cpus, - struct mm_struct *mm, unsigned long va) -{ - struct { - struct mmuext_op op; - DECLARE_BITMAP(mask, NR_CPUS); - } *args; - struct multicall_space mcs; - - BUG_ON(cpumask_empty(cpus)); - BUG_ON(!mm); - - mcs = xen_mc_entry(sizeof(*args)); - args = mcs.args; - args->op.arg2.vcpumask = to_cpumask(args->mask); - - /* Remove us, and any offline CPUS. */ - cpumask_and(to_cpumask(args->mask), cpus, cpu_online_mask); - cpumask_clear_cpu(smp_processor_id(), to_cpumask(args->mask)); - if (unlikely(cpumask_empty(to_cpumask(args->mask)))) - goto issue; - - if (va == TLB_FLUSH_ALL) { - args->op.cmd = MMUEXT_TLB_FLUSH_MULTI; - } else { - args->op.cmd = MMUEXT_INVLPG_MULTI; - args->op.arg1.linear_addr = va; - } - - MULTI_mmuext_op(mcs.mc, &args->op, 1, NULL, DOMID_SELF); - -issue: - xen_mc_issue(PARAVIRT_LAZY_MMU); -} static void xen_clts(void) { @@ -693,21 +594,6 @@ static void xen_write_cr0(unsigned long cr0) xen_mc_issue(PARAVIRT_LAZY_CPU); } -static void xen_write_cr2(unsigned long cr2) -{ - percpu_read(xen_vcpu)->arch.cr2 = cr2; -} - -static unsigned long xen_read_cr2(void) -{ - return percpu_read(xen_vcpu)->arch.cr2; -} - -static unsigned long xen_read_cr2_direct(void) -{ - return percpu_read(xen_vcpu_info.arch.cr2); -} - static void xen_write_cr4(unsigned long cr4) { cr4 &= ~X86_CR4_PGE; @@ -716,71 +602,6 @@ static void xen_write_cr4(unsigned long cr4) native_write_cr4(cr4); } -static unsigned long xen_read_cr3(void) -{ - return percpu_read(xen_cr3); -} - -static void set_current_cr3(void *v) -{ - percpu_write(xen_current_cr3, (unsigned long)v); -} - -static void __xen_write_cr3(bool kernel, unsigned long cr3) -{ - struct mmuext_op *op; - struct multicall_space mcs; - unsigned long mfn; - - if (cr3) - mfn = pfn_to_mfn(PFN_DOWN(cr3)); - else - mfn = 0; - - WARN_ON(mfn == 0 && kernel); - - mcs = __xen_mc_entry(sizeof(*op)); - - op = mcs.args; - op->cmd = kernel ? MMUEXT_NEW_BASEPTR : MMUEXT_NEW_USER_BASEPTR; - op->arg1.mfn = mfn; - - MULTI_mmuext_op(mcs.mc, op, 1, NULL, DOMID_SELF); - - if (kernel) { - percpu_write(xen_cr3, cr3); - - /* Update xen_current_cr3 once the batch has actually - been submitted. */ - xen_mc_callback(set_current_cr3, (void *)cr3); - } -} - -static void xen_write_cr3(unsigned long cr3) -{ - BUG_ON(preemptible()); - - xen_mc_batch(); /* disables interrupts */ - - /* Update while interrupts are disabled, so its atomic with - respect to ipis */ - percpu_write(xen_cr3, cr3); - - __xen_write_cr3(true, cr3); - -#ifdef CONFIG_X86_64 - { - pgd_t *user_pgd = xen_get_user_pgd(__va(cr3)); - if (user_pgd) - __xen_write_cr3(false, __pa(user_pgd)); - else - __xen_write_cr3(false, 0); - } -#endif - - xen_mc_issue(PARAVIRT_LAZY_CPU); /* interrupts restored */ -} - static int xen_write_msr_safe(unsigned int msr, unsigned low, unsigned high) { int ret; @@ -822,185 +643,6 @@ static int xen_write_msr_safe(unsigned int msr, unsigned low, unsigned high) return ret; } -/* Early in boot, while setting up the initial pagetable, assume - everything is pinned. */ -static __init void xen_alloc_pte_init(struct mm_struct *mm, unsigned long pfn) -{ -#ifdef CONFIG_FLATMEM - BUG_ON(mem_map); /* should only be used early */ -#endif - make_lowmem_page_readonly(__va(PFN_PHYS(pfn))); -} - -/* Early release_pte assumes that all pts are pinned, since there's - only init_mm and anything attached to that is pinned. */ -static void xen_release_pte_init(unsigned long pfn) -{ - make_lowmem_page_readwrite(__va(PFN_PHYS(pfn))); -} - -static void pin_pagetable_pfn(unsigned cmd, unsigned long pfn) -{ - struct mmuext_op op; - op.cmd = cmd; - op.arg1.mfn = pfn_to_mfn(pfn); - if (HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF)) - BUG(); -} - -/* This needs to make sure the new pte page is pinned iff its being - attached to a pinned pagetable. */ -static void xen_alloc_ptpage(struct mm_struct *mm, unsigned long pfn, unsigned level) -{ - struct page *page = pfn_to_page(pfn); - - if (PagePinned(virt_to_page(mm->pgd))) { - SetPagePinned(page); - - vm_unmap_aliases(); - if (!PageHighMem(page)) { - make_lowmem_page_readonly(__va(PFN_PHYS((unsigned long)pfn))); - if (level == PT_PTE && USE_SPLIT_PTLOCKS) - pin_pagetable_pfn(MMUEXT_PIN_L1_TABLE, pfn); - } else { - /* make sure there are no stray mappings of - this page */ - kmap_flush_unused(); - } - } -} - -static void xen_alloc_pte(struct mm_struct *mm, unsigned long pfn) -{ - xen_alloc_ptpage(mm, pfn, PT_PTE); -} - -static void xen_alloc_pmd(struct mm_struct *mm, unsigned long pfn) -{ - xen_alloc_ptpage(mm, pfn, PT_PMD); -} - -static int xen_pgd_alloc(struct mm_struct *mm) -{ - pgd_t *pgd = mm->pgd; - int ret = 0; - - BUG_ON(PagePinned(virt_to_page(pgd))); - -#ifdef CONFIG_X86_64 - { - struct page *page = virt_to_page(pgd); - pgd_t *user_pgd; - - BUG_ON(page->private != 0); - - ret = -ENOMEM; - - user_pgd = (pgd_t *)__get_free_page(GFP_KERNEL | __GFP_ZERO); - page->private = (unsigned long)user_pgd; - - if (user_pgd != NULL) { - user_pgd[pgd_index(VSYSCALL_START)] = - __pgd(__pa(level3_user_vsyscall) | _PAGE_TABLE); - ret = 0; - } - - BUG_ON(PagePinned(virt_to_page(xen_get_user_pgd(pgd)))); - } -#endif - - return ret; -} - -static void xen_pgd_free(struct mm_struct *mm, pgd_t *pgd) -{ -#ifdef CONFIG_X86_64 - pgd_t *user_pgd = xen_get_user_pgd(pgd); - - if (user_pgd) - free_page((unsigned long)user_pgd); -#endif -} - -/* This should never happen until we're OK to use struct page */ -static void xen_release_ptpage(unsigned long pfn, unsigned level) -{ - struct page *page = pfn_to_page(pfn); - - if (PagePinned(page)) { - if (!PageHighMem(page)) { - if (level == PT_PTE && USE_SPLIT_PTLOCKS) - pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, pfn); - make_lowmem_page_readwrite(__va(PFN_PHYS(pfn))); - } - ClearPagePinned(page); - } -} - -static void xen_release_pte(unsigned long pfn) -{ - xen_release_ptpage(pfn, PT_PTE); -} - -static void xen_release_pmd(unsigned long pfn) -{ - xen_release_ptpage(pfn, PT_PMD); -} - -#if PAGETABLE_LEVELS == 4 -static void xen_alloc_pud(struct mm_struct *mm, unsigned long pfn) -{ - xen_alloc_ptpage(mm, pfn, PT_PUD); -} - -static void xen_release_pud(unsigned long pfn) -{ - xen_release_ptpage(pfn, PT_PUD); -} -#endif - -#ifdef CONFIG_HIGHPTE -static void *xen_kmap_atomic_pte(struct page *page, enum km_type type) -{ - pgprot_t prot = PAGE_KERNEL; - - if (PagePinned(page)) - prot = PAGE_KERNEL_RO; - - if (0 && PageHighMem(page)) - printk("mapping highpte %lx type %d prot %s\n", - page_to_pfn(page), type, - (unsigned long)pgprot_val(prot) & _PAGE_RW ? "WRITE" : "READ"); - - return kmap_atomic_prot(page, type, prot); -} -#endif - -#ifdef CONFIG_X86_32 -static __init pte_t mask_rw_pte(pte_t *ptep, pte_t pte) -{ - /* If there's an existing pte, then don't allow _PAGE_RW to be set */ - if (pte_val_ma(*ptep) & _PAGE_PRESENT) - pte = __pte_ma(((pte_val_ma(*ptep) & _PAGE_RW) | ~_PAGE_RW) & - pte_val_ma(pte)); - - return pte; -} - -/* Init-time set_pte while constructing initial pagetables, which - doesn't allow RO pagetable pages to be remapped RW */ -static __init void xen_set_pte_init(pte_t *ptep, pte_t pte) -{ - pte = mask_rw_pte(ptep, pte); - - xen_set_pte(ptep, pte); -} -#endif - -static __init void xen_pagetable_setup_start(pgd_t *base) -{ -} - void xen_setup_shared_info(void) { if (!xen_feature(XENFEAT_auto_translated_physmap)) { @@ -1021,37 +663,6 @@ void xen_setup_shared_info(void) xen_setup_mfn_list_list(); } -static __init void xen_pagetable_setup_done(pgd_t *base) -{ - xen_setup_shared_info(); -} - -static __init void xen_post_allocator_init(void) -{ - pv_mmu_ops.set_pte = xen_set_pte; - pv_mmu_ops.set_pmd = xen_set_pmd; - pv_mmu_ops.set_pud = xen_set_pud; -#if PAGETABLE_LEVELS == 4 - pv_mmu_ops.set_pgd = xen_set_pgd; -#endif - - /* This will work as long as patching hasn't happened yet - (which it hasn't) */ - pv_mmu_ops.alloc_pte = xen_alloc_pte; - pv_mmu_ops.alloc_pmd = xen_alloc_pmd; - pv_mmu_ops.release_pte = xen_release_pte; - pv_mmu_ops.release_pmd = xen_release_pmd; -#if PAGETABLE_LEVELS == 4 - pv_mmu_ops.alloc_pud = xen_alloc_pud; - pv_mmu_ops.release_pud = xen_release_pud; -#endif - -#ifdef CONFIG_X86_64 - SetPagePinned(virt_to_page(level3_user_vsyscall)); -#endif - xen_mark_init_mm_pinned(); -} - /* This is called once we have the cpu_possible_map */ void xen_setup_vcpu_info_placement(void) { @@ -1126,49 +737,6 @@ static unsigned xen_patch(u8 type, u16 clobbers, void *insnbuf, return ret; } -static void xen_set_fixmap(unsigned idx, unsigned long phys, pgprot_t prot) -{ - pte_t pte; - - phys >>= PAGE_SHIFT; - - switch (idx) { - case FIX_BTMAP_END ... FIX_BTMAP_BEGIN: -#ifdef CONFIG_X86_F00F_BUG - case FIX_F00F_IDT: -#endif -#ifdef CONFIG_X86_32 - case FIX_WP_TEST: - case FIX_VDSO: -# ifdef CONFIG_HIGHMEM - case FIX_KMAP_BEGIN ... FIX_KMAP_END: -# endif -#else - case VSYSCALL_LAST_PAGE ... VSYSCALL_FIRST_PAGE: -#endif -#ifdef CONFIG_X86_LOCAL_APIC - case FIX_APIC_BASE: /* maps dummy local APIC */ -#endif - pte = pfn_pte(phys, prot); - break; - - default: - pte = mfn_pte(phys, prot); - break; - } - - __native_set_fixmap(idx, pte); - -#ifdef CONFIG_X86_64 - /* Replicate changes to map the vsyscall page into the user - pagetable vsyscall mapping. */ - if (idx >= VSYSCALL_LAST_PAGE && idx <= VSYSCALL_FIRST_PAGE) { - unsigned long vaddr = __fix_to_virt(idx); - set_pte_vaddr_pud(level3_user_vsyscall, vaddr, pte); - } -#endif -} - static const struct pv_info xen_info __initdata = { .paravirt_enabled = 1, .shared_kernel_pmd = 0, @@ -1264,86 +832,6 @@ static const struct pv_apic_ops xen_apic_ops __initdata = { #endif }; -static const struct pv_mmu_ops xen_mmu_ops __initdata = { - .pagetable_setup_start = xen_pagetable_setup_start, - .pagetable_setup_done = xen_pagetable_setup_done, - - .read_cr2 = xen_read_cr2, - .write_cr2 = xen_write_cr2, - - .read_cr3 = xen_read_cr3, - .write_cr3 = xen_write_cr3, - - .flush_tlb_user = xen_flush_tlb, - .flush_tlb_kernel = xen_flush_tlb, - .flush_tlb_single = xen_flush_tlb_single, - .flush_tlb_others = xen_flush_tlb_others, - - .pte_update = paravirt_nop, - .pte_update_defer = paravirt_nop, - - .pgd_alloc = xen_pgd_alloc, - .pgd_free = xen_pgd_free, - - .alloc_pte = xen_alloc_pte_init, - .release_pte = xen_release_pte_init, - .alloc_pmd = xen_alloc_pte_init, - .alloc_pmd_clone = paravirt_nop, - .release_pmd = xen_release_pte_init, - -#ifdef CONFIG_HIGHPTE - .kmap_atomic_pte = xen_kmap_atomic_pte, -#endif - -#ifdef CONFIG_X86_64 - .set_pte = xen_set_pte, -#else - .set_pte = xen_set_pte_init, -#endif - .set_pte_at = xen_set_pte_at, - .set_pmd = xen_set_pmd_hyper, - - .ptep_modify_prot_start = __ptep_modify_prot_start, - .ptep_modify_prot_commit = __ptep_modify_prot_commit, - - .pte_val = xen_pte_val, - .pgd_val = xen_pgd_val, - - .make_pte = xen_make_pte, - .make_pgd = xen_make_pgd, - -#ifdef CONFIG_X86_PAE - .set_pte_atomic = xen_set_pte_atomic, - .set_pte_present = xen_set_pte_at, - .pte_clear = xen_pte_clear, - .pmd_clear = xen_pmd_clear, -#endif /* CONFIG_X86_PAE */ - .set_pud = xen_set_pud_hyper, - - .make_pmd = xen_make_pmd, - .pmd_val = xen_pmd_val, - -#if PAGETABLE_LEVELS == 4 - .pud_val = xen_pud_val, - .make_pud = xen_make_pud, - .set_pgd = xen_set_pgd_hyper, - - .alloc_pud = xen_alloc_pte_init, - .release_pud = xen_release_pte_init, -#endif /* PAGETABLE_LEVELS == 4 */ - - .activate_mm = xen_activate_mm, - .dup_mmap = xen_dup_mmap, - .exit_mmap = xen_exit_mmap, - - .lazy_mode = { - .enter = paravirt_enter_lazy_mmu, - .leave = xen_leave_lazy, - }, - - .set_fixmap = xen_set_fixmap, -}; - static void xen_reboot(int reason) { struct sched_shutdown r = { .reason = reason }; @@ -1386,223 +874,6 @@ static const struct machine_ops __initdata xen_machine_ops = { }; -static void __init xen_reserve_top(void) -{ -#ifdef CONFIG_X86_32 - unsigned long top = HYPERVISOR_VIRT_START; - struct xen_platform_parameters pp; - - if (HYPERVISOR_xen_version(XENVER_platform_parameters, &pp) == 0) - top = pp.virt_start; - - reserve_top_address(-top); -#endif /* CONFIG_X86_32 */ -} - -/* - * Like __va(), but returns address in the kernel mapping (which is - * all we have until the physical memory mapping has been set up. - */ -static void *__ka(phys_addr_t paddr) -{ -#ifdef CONFIG_X86_64 - return (void *)(paddr + __START_KERNEL_map); -#else - return __va(paddr); -#endif -} - -/* Convert a machine address to physical address */ -static unsigned long m2p(phys_addr_t maddr) -{ - phys_addr_t paddr; - - maddr &= PTE_PFN_MASK; - paddr = mfn_to_pfn(maddr >> PAGE_SHIFT) << PAGE_SHIFT; - - return paddr; -} - -/* Convert a machine address to kernel virtual */ -static void *m2v(phys_addr_t maddr) -{ - return __ka(m2p(maddr)); -} - -static void set_page_prot(void *addr, pgprot_t prot) -{ - unsigned long pfn = __pa(addr) >> PAGE_SHIFT; - pte_t pte = pfn_pte(pfn, prot); - - if (HYPERVISOR_update_va_mapping((unsigned long)addr, pte, 0)) - BUG(); -} - -static __init void xen_map_identity_early(pmd_t *pmd, unsigned long max_pfn) -{ - unsigned pmdidx, pteidx; - unsigned ident_pte; - unsigned long pfn; - - ident_pte = 0; - pfn = 0; - for (pmdidx = 0; pmdidx < PTRS_PER_PMD && pfn < max_pfn; pmdidx++) { - pte_t *pte_page; - - /* Reuse or allocate a page of ptes */ - if (pmd_present(pmd[pmdidx])) - pte_page = m2v(pmd[pmdidx].pmd); - else { - /* Check for free pte pages */ - if (ident_pte == ARRAY_SIZE(level1_ident_pgt)) - break; - - pte_page = &level1_ident_pgt[ident_pte]; - ident_pte += PTRS_PER_PTE; - - pmd[pmdidx] = __pmd(__pa(pte_page) | _PAGE_TABLE); - } - - /* Install mappings */ - for (pteidx = 0; pteidx < PTRS_PER_PTE; pteidx++, pfn++) { - pte_t pte; - - if (pfn > max_pfn_mapped) - max_pfn_mapped = pfn; - - if (!pte_none(pte_page[pteidx])) - continue; - - pte = pfn_pte(pfn, PAGE_KERNEL_EXEC); - pte_page[pteidx] = pte; - } - } - - for (pteidx = 0; pteidx < ident_pte; pteidx += PTRS_PER_PTE) - set_page_prot(&level1_ident_pgt[pteidx], PAGE_KERNEL_RO); - - set_page_prot(pmd, PAGE_KERNEL_RO); -} - -#ifdef CONFIG_X86_64 -static void convert_pfn_mfn(void *v) -{ - pte_t *pte = v; - int i; - - /* All levels are converted the same way, so just treat them - as ptes. */ - for (i = 0; i < PTRS_PER_PTE; i++) - pte[i] = xen_make_pte(pte[i].pte); -} - -/* - * Set up the inital kernel pagetable. - * - * We can construct this by grafting the Xen provided pagetable into - * head_64.S's preconstructed pagetables. We copy the Xen L2's into - * level2_ident_pgt, level2_kernel_pgt and level2_fixmap_pgt. This - * means that only the kernel has a physical mapping to start with - - * but that's enough to get __va working. We need to fill in the rest - * of the physical mapping once some sort of allocator has been set - * up. - */ -static __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd, - unsigned long max_pfn) -{ - pud_t *l3; - pmd_t *l2; - - /* Zap identity mapping */ - init_level4_pgt[0] = __pgd(0); - - /* Pre-constructed entries are in pfn, so convert to mfn */ - convert_pfn_mfn(init_level4_pgt); - convert_pfn_mfn(level3_ident_pgt); - convert_pfn_mfn(level3_kernel_pgt); - - l3 = m2v(pgd[pgd_index(__START_KERNEL_map)].pgd); - l2 = m2v(l3[pud_index(__START_KERNEL_map)].pud); - - memcpy(level2_ident_pgt, l2, sizeof(pmd_t) * PTRS_PER_PMD); - memcpy(level2_kernel_pgt, l2, sizeof(pmd_t) * PTRS_PER_PMD); - - l3 = m2v(pgd[pgd_index(__START_KERNEL_map + PMD_SIZE)].pgd); - l2 = m2v(l3[pud_index(__START_KERNEL_map + PMD_SIZE)].pud); - memcpy(level2_fixmap_pgt, l2, sizeof(pmd_t) * PTRS_PER_PMD); - - /* Set up identity map */ - xen_map_identity_early(level2_ident_pgt, max_pfn); - - /* Make pagetable pieces RO */ - set_page_prot(init_level4_pgt, PAGE_KERNEL_RO); - set_page_prot(level3_ident_pgt, PAGE_KERNEL_RO); - set_page_prot(level3_kernel_pgt, PAGE_KERNEL_RO); - set_page_prot(level3_user_vsyscall, PAGE_KERNEL_RO); - set_page_prot(level2_kernel_pgt, PAGE_KERNEL_RO); - set_page_prot(level2_fixmap_pgt, PAGE_KERNEL_RO); - - /* Pin down new L4 */ - pin_pagetable_pfn(MMUEXT_PIN_L4_TABLE, - PFN_DOWN(__pa_symbol(init_level4_pgt))); - - /* Unpin Xen-provided one */ - pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, PFN_DOWN(__pa(pgd))); - - /* Switch over */ - pgd = init_level4_pgt; - - /* - * At this stage there can be no user pgd, and no page - * structure to attach it to, so make sure we just set kernel - * pgd. - */ - xen_mc_batch(); - __xen_write_cr3(true, __pa(pgd)); - xen_mc_issue(PARAVIRT_LAZY_CPU); - - reserve_early(__pa(xen_start_info->pt_base), - __pa(xen_start_info->pt_base + - xen_start_info->nr_pt_frames * PAGE_SIZE), - "XEN PAGETABLES"); - - return pgd; -} -#else /* !CONFIG_X86_64 */ -static pmd_t level2_kernel_pgt[PTRS_PER_PMD] __page_aligned_bss; - -static __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd, - unsigned long max_pfn) -{ - pmd_t *kernel_pmd; - - init_pg_tables_start = __pa(pgd); - init_pg_tables_end = __pa(pgd) + xen_start_info->nr_pt_frames*PAGE_SIZE; - max_pfn_mapped = PFN_DOWN(init_pg_tables_end + 512*1024); - - kernel_pmd = m2v(pgd[KERNEL_PGD_BOUNDARY].pgd); - memcpy(level2_kernel_pgt, kernel_pmd, sizeof(pmd_t) * PTRS_PER_PMD); - - xen_map_identity_early(level2_kernel_pgt, max_pfn); - - memcpy(swapper_pg_dir, pgd, sizeof(pgd_t) * PTRS_PER_PGD); - set_pgd(&swapper_pg_dir[KERNEL_PGD_BOUNDARY], - __pgd(__pa(level2_kernel_pgt) | _PAGE_PRESENT)); - - set_page_prot(level2_kernel_pgt, PAGE_KERNEL_RO); - set_page_prot(swapper_pg_dir, PAGE_KERNEL_RO); - set_page_prot(empty_zero_page, PAGE_KERNEL_RO); - - pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, PFN_DOWN(__pa(pgd))); - - xen_write_cr3(__pa(swapper_pg_dir)); - - pin_pagetable_pfn(MMUEXT_PIN_L3_TABLE, PFN_DOWN(__pa(swapper_pg_dir))); - - return swapper_pg_dir; -} -#endif /* CONFIG_X86_64 */ - /* First C function to be called on Xen boot */ asmlinkage void __init xen_start_kernel(void) { diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 98cb9869eb2..94e452c0b00 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -47,6 +47,7 @@ #include #include #include +#include #include #include @@ -55,6 +56,8 @@ #include #include +#include +#include #include "multicalls.h" #include "mmu.h" @@ -114,6 +117,37 @@ static inline void check_zero(void) #endif /* CONFIG_XEN_DEBUG_FS */ + +/* + * Identity map, in addition to plain kernel map. This needs to be + * large enough to allocate page table pages to allocate the rest. + * Each page can map 2MB. + */ +static pte_t level1_ident_pgt[PTRS_PER_PTE * 4] __page_aligned_bss; + +#ifdef CONFIG_X86_64 +/* l3 pud for userspace vsyscall mapping */ +static pud_t level3_user_vsyscall[PTRS_PER_PUD] __page_aligned_bss; +#endif /* CONFIG_X86_64 */ + +/* + * Note about cr3 (pagetable base) values: + * + * xen_cr3 contains the current logical cr3 value; it contains the + * last set cr3. This may not be the current effective cr3, because + * its update may be being lazily deferred. However, a vcpu looking + * at its own cr3 can use this value knowing that it everything will + * be self-consistent. + * + * xen_current_cr3 contains the actual vcpu cr3; it is set once the + * hypercall to set the vcpu cr3 is complete (so it may be a little + * out of date, but it will never be set early). If one vcpu is + * looking at another vcpu's cr3 value, it should use this variable. + */ +DEFINE_PER_CPU(unsigned long, xen_cr3); /* cr3 stored as physaddr */ +DEFINE_PER_CPU(unsigned long, xen_current_cr3); /* actual vcpu cr3 */ + + /* * Just beyond the highest usermode address. STACK_TOP_MAX has a * redzone above it, so round it up to a PGD boundary. @@ -1152,6 +1186,672 @@ void xen_exit_mmap(struct mm_struct *mm) spin_unlock(&mm->page_table_lock); } +static __init void xen_pagetable_setup_start(pgd_t *base) +{ +} + +static __init void xen_pagetable_setup_done(pgd_t *base) +{ + xen_setup_shared_info(); +} + +static void xen_write_cr2(unsigned long cr2) +{ + percpu_read(xen_vcpu)->arch.cr2 = cr2; +} + +static unsigned long xen_read_cr2(void) +{ + return percpu_read(xen_vcpu)->arch.cr2; +} + +unsigned long xen_read_cr2_direct(void) +{ + return percpu_read(xen_vcpu_info.arch.cr2); +} + +static void xen_flush_tlb(void) +{ + struct mmuext_op *op; + struct multicall_space mcs; + + preempt_disable(); + + mcs = xen_mc_entry(sizeof(*op)); + + op = mcs.args; + op->cmd = MMUEXT_TLB_FLUSH_LOCAL; + MULTI_mmuext_op(mcs.mc, op, 1, NULL, DOMID_SELF); + + xen_mc_issue(PARAVIRT_LAZY_MMU); + + preempt_enable(); +} + +static void xen_flush_tlb_single(unsigned long addr) +{ + struct mmuext_op *op; + struct multicall_space mcs; + + preempt_disable(); + + mcs = xen_mc_entry(sizeof(*op)); + op = mcs.args; + op->cmd = MMUEXT_INVLPG_LOCAL; + op->arg1.linear_addr = addr & PAGE_MASK; + MULTI_mmuext_op(mcs.mc, op, 1, NULL, DOMID_SELF); + + xen_mc_issue(PARAVIRT_LAZY_MMU); + + preempt_enable(); +} + +static void xen_flush_tlb_others(const struct cpumask *cpus, + struct mm_struct *mm, unsigned long va) +{ + struct { + struct mmuext_op op; + DECLARE_BITMAP(mask, NR_CPUS); + } *args; + struct multicall_space mcs; + + BUG_ON(cpumask_empty(cpus)); + BUG_ON(!mm); + + mcs = xen_mc_entry(sizeof(*args)); + args = mcs.args; + args->op.arg2.vcpumask = to_cpumask(args->mask); + + /* Remove us, and any offline CPUS. */ + cpumask_and(to_cpumask(args->mask), cpus, cpu_online_mask); + cpumask_clear_cpu(smp_processor_id(), to_cpumask(args->mask)); + if (unlikely(cpumask_empty(to_cpumask(args->mask)))) + goto issue; + + if (va == TLB_FLUSH_ALL) { + args->op.cmd = MMUEXT_TLB_FLUSH_MULTI; + } else { + args->op.cmd = MMUEXT_INVLPG_MULTI; + args->op.arg1.linear_addr = va; + } + + MULTI_mmuext_op(mcs.mc, &args->op, 1, NULL, DOMID_SELF); + +issue: + xen_mc_issue(PARAVIRT_LAZY_MMU); +} + +static unsigned long xen_read_cr3(void) +{ + return percpu_read(xen_cr3); +} + +static void set_current_cr3(void *v) +{ + percpu_write(xen_current_cr3, (unsigned long)v); +} + +static void __xen_write_cr3(bool kernel, unsigned long cr3) +{ + struct mmuext_op *op; + struct multicall_space mcs; + unsigned long mfn; + + if (cr3) + mfn = pfn_to_mfn(PFN_DOWN(cr3)); + else + mfn = 0; + + WARN_ON(mfn == 0 && kernel); + + mcs = __xen_mc_entry(sizeof(*op)); + + op = mcs.args; + op->cmd = kernel ? MMUEXT_NEW_BASEPTR : MMUEXT_NEW_USER_BASEPTR; + op->arg1.mfn = mfn; + + MULTI_mmuext_op(mcs.mc, op, 1, NULL, DOMID_SELF); + + if (kernel) { + percpu_write(xen_cr3, cr3); + + /* Update xen_current_cr3 once the batch has actually + been submitted. */ + xen_mc_callback(set_current_cr3, (void *)cr3); + } +} + +static void xen_write_cr3(unsigned long cr3) +{ + BUG_ON(preemptible()); + + xen_mc_batch(); /* disables interrupts */ + + /* Update while interrupts are disabled, so its atomic with + respect to ipis */ + percpu_write(xen_cr3, cr3); + + __xen_write_cr3(true, cr3); + +#ifdef CONFIG_X86_64 + { + pgd_t *user_pgd = xen_get_user_pgd(__va(cr3)); + if (user_pgd) + __xen_write_cr3(false, __pa(user_pgd)); + else + __xen_write_cr3(false, 0); + } +#endif + + xen_mc_issue(PARAVIRT_LAZY_CPU); /* interrupts restored */ +} + +static int xen_pgd_alloc(struct mm_struct *mm) +{ + pgd_t *pgd = mm->pgd; + int ret = 0; + + BUG_ON(PagePinned(virt_to_page(pgd))); + +#ifdef CONFIG_X86_64 + { + struct page *page = virt_to_page(pgd); + pgd_t *user_pgd; + + BUG_ON(page->private != 0); + + ret = -ENOMEM; + + user_pgd = (pgd_t *)__get_free_page(GFP_KERNEL | __GFP_ZERO); + page->private = (unsigned long)user_pgd; + + if (user_pgd != NULL) { + user_pgd[pgd_index(VSYSCALL_START)] = + __pgd(__pa(level3_user_vsyscall) | _PAGE_TABLE); + ret = 0; + } + + BUG_ON(PagePinned(virt_to_page(xen_get_user_pgd(pgd)))); + } +#endif + + return ret; +} + +static void xen_pgd_free(struct mm_struct *mm, pgd_t *pgd) +{ +#ifdef CONFIG_X86_64 + pgd_t *user_pgd = xen_get_user_pgd(pgd); + + if (user_pgd) + free_page((unsigned long)user_pgd); +#endif +} + + +/* Early in boot, while setting up the initial pagetable, assume + everything is pinned. */ +static __init void xen_alloc_pte_init(struct mm_struct *mm, unsigned long pfn) +{ +#ifdef CONFIG_FLATMEM + BUG_ON(mem_map); /* should only be used early */ +#endif + make_lowmem_page_readonly(__va(PFN_PHYS(pfn))); +} + +/* Early release_pte assumes that all pts are pinned, since there's + only init_mm and anything attached to that is pinned. */ +static void xen_release_pte_init(unsigned long pfn) +{ + make_lowmem_page_readwrite(__va(PFN_PHYS(pfn))); +} + +static void pin_pagetable_pfn(unsigned cmd, unsigned long pfn) +{ + struct mmuext_op op; + op.cmd = cmd; + op.arg1.mfn = pfn_to_mfn(pfn); + if (HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF)) + BUG(); +} + +/* This needs to make sure the new pte page is pinned iff its being + attached to a pinned pagetable. */ +static void xen_alloc_ptpage(struct mm_struct *mm, unsigned long pfn, unsigned level) +{ + struct page *page = pfn_to_page(pfn); + + if (PagePinned(virt_to_page(mm->pgd))) { + SetPagePinned(page); + + vm_unmap_aliases(); + if (!PageHighMem(page)) { + make_lowmem_page_readonly(__va(PFN_PHYS((unsigned long)pfn))); + if (level == PT_PTE && USE_SPLIT_PTLOCKS) + pin_pagetable_pfn(MMUEXT_PIN_L1_TABLE, pfn); + } else { + /* make sure there are no stray mappings of + this page */ + kmap_flush_unused(); + } + } +} + +static void xen_alloc_pte(struct mm_struct *mm, unsigned long pfn) +{ + xen_alloc_ptpage(mm, pfn, PT_PTE); +} + +static void xen_alloc_pmd(struct mm_struct *mm, unsigned long pfn) +{ + xen_alloc_ptpage(mm, pfn, PT_PMD); +} + +/* This should never happen until we're OK to use struct page */ +static void xen_release_ptpage(unsigned long pfn, unsigned level) +{ + struct page *page = pfn_to_page(pfn); + + if (PagePinned(page)) { + if (!PageHighMem(page)) { + if (level == PT_PTE && USE_SPLIT_PTLOCKS) + pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, pfn); + make_lowmem_page_readwrite(__va(PFN_PHYS(pfn))); + } + ClearPagePinned(page); + } +} + +static void xen_release_pte(unsigned long pfn) +{ + xen_release_ptpage(pfn, PT_PTE); +} + +static void xen_release_pmd(unsigned long pfn) +{ + xen_release_ptpage(pfn, PT_PMD); +} + +#if PAGETABLE_LEVELS == 4 +static void xen_alloc_pud(struct mm_struct *mm, unsigned long pfn) +{ + xen_alloc_ptpage(mm, pfn, PT_PUD); +} + +static void xen_release_pud(unsigned long pfn) +{ + xen_release_ptpage(pfn, PT_PUD); +} +#endif + +void __init xen_reserve_top(void) +{ +#ifdef CONFIG_X86_32 + unsigned long top = HYPERVISOR_VIRT_START; + struct xen_platform_parameters pp; + + if (HYPERVISOR_xen_version(XENVER_platform_parameters, &pp) == 0) + top = pp.virt_start; + + reserve_top_address(-top); +#endif /* CONFIG_X86_32 */ +} + +/* + * Like __va(), but returns address in the kernel mapping (which is + * all we have until the physical memory mapping has been set up. + */ +static void *__ka(phys_addr_t paddr) +{ +#ifdef CONFIG_X86_64 + return (void *)(paddr + __START_KERNEL_map); +#else + return __va(paddr); +#endif +} + +/* Convert a machine address to physical address */ +static unsigned long m2p(phys_addr_t maddr) +{ + phys_addr_t paddr; + + maddr &= PTE_PFN_MASK; + paddr = mfn_to_pfn(maddr >> PAGE_SHIFT) << PAGE_SHIFT; + + return paddr; +} + +/* Convert a machine address to kernel virtual */ +static void *m2v(phys_addr_t maddr) +{ + return __ka(m2p(maddr)); +} + +static void set_page_prot(void *addr, pgprot_t prot) +{ + unsigned long pfn = __pa(addr) >> PAGE_SHIFT; + pte_t pte = pfn_pte(pfn, prot); + + if (HYPERVISOR_update_va_mapping((unsigned long)addr, pte, 0)) + BUG(); +} + +static __init void xen_map_identity_early(pmd_t *pmd, unsigned long max_pfn) +{ + unsigned pmdidx, pteidx; + unsigned ident_pte; + unsigned long pfn; + + ident_pte = 0; + pfn = 0; + for (pmdidx = 0; pmdidx < PTRS_PER_PMD && pfn < max_pfn; pmdidx++) { + pte_t *pte_page; + + /* Reuse or allocate a page of ptes */ + if (pmd_present(pmd[pmdidx])) + pte_page = m2v(pmd[pmdidx].pmd); + else { + /* Check for free pte pages */ + if (ident_pte == ARRAY_SIZE(level1_ident_pgt)) + break; + + pte_page = &level1_ident_pgt[ident_pte]; + ident_pte += PTRS_PER_PTE; + + pmd[pmdidx] = __pmd(__pa(pte_page) | _PAGE_TABLE); + } + + /* Install mappings */ + for (pteidx = 0; pteidx < PTRS_PER_PTE; pteidx++, pfn++) { + pte_t pte; + + if (pfn > max_pfn_mapped) + max_pfn_mapped = pfn; + + if (!pte_none(pte_page[pteidx])) + continue; + + pte = pfn_pte(pfn, PAGE_KERNEL_EXEC); + pte_page[pteidx] = pte; + } + } + + for (pteidx = 0; pteidx < ident_pte; pteidx += PTRS_PER_PTE) + set_page_prot(&level1_ident_pgt[pteidx], PAGE_KERNEL_RO); + + set_page_prot(pmd, PAGE_KERNEL_RO); +} + +#ifdef CONFIG_X86_64 +static void convert_pfn_mfn(void *v) +{ + pte_t *pte = v; + int i; + + /* All levels are converted the same way, so just treat them + as ptes. */ + for (i = 0; i < PTRS_PER_PTE; i++) + pte[i] = xen_make_pte(pte[i].pte); +} + +/* + * Set up the inital kernel pagetable. + * + * We can construct this by grafting the Xen provided pagetable into + * head_64.S's preconstructed pagetables. We copy the Xen L2's into + * level2_ident_pgt, level2_kernel_pgt and level2_fixmap_pgt. This + * means that only the kernel has a physical mapping to start with - + * but that's enough to get __va working. We need to fill in the rest + * of the physical mapping once some sort of allocator has been set + * up. + */ +__init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd, + unsigned long max_pfn) +{ + pud_t *l3; + pmd_t *l2; + + /* Zap identity mapping */ + init_level4_pgt[0] = __pgd(0); + + /* Pre-constructed entries are in pfn, so convert to mfn */ + convert_pfn_mfn(init_level4_pgt); + convert_pfn_mfn(level3_ident_pgt); + convert_pfn_mfn(level3_kernel_pgt); + + l3 = m2v(pgd[pgd_index(__START_KERNEL_map)].pgd); + l2 = m2v(l3[pud_index(__START_KERNEL_map)].pud); + + memcpy(level2_ident_pgt, l2, sizeof(pmd_t) * PTRS_PER_PMD); + memcpy(level2_kernel_pgt, l2, sizeof(pmd_t) * PTRS_PER_PMD); + + l3 = m2v(pgd[pgd_index(__START_KERNEL_map + PMD_SIZE)].pgd); + l2 = m2v(l3[pud_index(__START_KERNEL_map + PMD_SIZE)].pud); + memcpy(level2_fixmap_pgt, l2, sizeof(pmd_t) * PTRS_PER_PMD); + + /* Set up identity map */ + xen_map_identity_early(level2_ident_pgt, max_pfn); + + /* Make pagetable pieces RO */ + set_page_prot(init_level4_pgt, PAGE_KERNEL_RO); + set_page_prot(level3_ident_pgt, PAGE_KERNEL_RO); + set_page_prot(level3_kernel_pgt, PAGE_KERNEL_RO); + set_page_prot(level3_user_vsyscall, PAGE_KERNEL_RO); + set_page_prot(level2_kernel_pgt, PAGE_KERNEL_RO); + set_page_prot(level2_fixmap_pgt, PAGE_KERNEL_RO); + + /* Pin down new L4 */ + pin_pagetable_pfn(MMUEXT_PIN_L4_TABLE, + PFN_DOWN(__pa_symbol(init_level4_pgt))); + + /* Unpin Xen-provided one */ + pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, PFN_DOWN(__pa(pgd))); + + /* Switch over */ + pgd = init_level4_pgt; + + /* + * At this stage there can be no user pgd, and no page + * structure to attach it to, so make sure we just set kernel + * pgd. + */ + xen_mc_batch(); + __xen_write_cr3(true, __pa(pgd)); + xen_mc_issue(PARAVIRT_LAZY_CPU); + + reserve_early(__pa(xen_start_info->pt_base), + __pa(xen_start_info->pt_base + + xen_start_info->nr_pt_frames * PAGE_SIZE), + "XEN PAGETABLES"); + + return pgd; +} +#else /* !CONFIG_X86_64 */ +static pmd_t level2_kernel_pgt[PTRS_PER_PMD] __page_aligned_bss; + +__init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd, + unsigned long max_pfn) +{ + pmd_t *kernel_pmd; + + init_pg_tables_start = __pa(pgd); + init_pg_tables_end = __pa(pgd) + xen_start_info->nr_pt_frames*PAGE_SIZE; + max_pfn_mapped = PFN_DOWN(init_pg_tables_end + 512*1024); + + kernel_pmd = m2v(pgd[KERNEL_PGD_BOUNDARY].pgd); + memcpy(level2_kernel_pgt, kernel_pmd, sizeof(pmd_t) * PTRS_PER_PMD); + + xen_map_identity_early(level2_kernel_pgt, max_pfn); + + memcpy(swapper_pg_dir, pgd, sizeof(pgd_t) * PTRS_PER_PGD); + set_pgd(&swapper_pg_dir[KERNEL_PGD_BOUNDARY], + __pgd(__pa(level2_kernel_pgt) | _PAGE_PRESENT)); + + set_page_prot(level2_kernel_pgt, PAGE_KERNEL_RO); + set_page_prot(swapper_pg_dir, PAGE_KERNEL_RO); + set_page_prot(empty_zero_page, PAGE_KERNEL_RO); + + pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, PFN_DOWN(__pa(pgd))); + + xen_write_cr3(__pa(swapper_pg_dir)); + + pin_pagetable_pfn(MMUEXT_PIN_L3_TABLE, PFN_DOWN(__pa(swapper_pg_dir))); + + return swapper_pg_dir; +} +#endif /* CONFIG_X86_64 */ + +static void xen_set_fixmap(unsigned idx, unsigned long phys, pgprot_t prot) +{ + pte_t pte; + + phys >>= PAGE_SHIFT; + + switch (idx) { + case FIX_BTMAP_END ... FIX_BTMAP_BEGIN: +#ifdef CONFIG_X86_F00F_BUG + case FIX_F00F_IDT: +#endif +#ifdef CONFIG_X86_32 + case FIX_WP_TEST: + case FIX_VDSO: +# ifdef CONFIG_HIGHMEM + case FIX_KMAP_BEGIN ... FIX_KMAP_END: +# endif +#else + case VSYSCALL_LAST_PAGE ... VSYSCALL_FIRST_PAGE: +#endif +#ifdef CONFIG_X86_LOCAL_APIC + case FIX_APIC_BASE: /* maps dummy local APIC */ +#endif + pte = pfn_pte(phys, prot); + break; + + default: + pte = mfn_pte(phys, prot); + break; + } + + __native_set_fixmap(idx, pte); + +#ifdef CONFIG_X86_64 + /* Replicate changes to map the vsyscall page into the user + pagetable vsyscall mapping. */ + if (idx >= VSYSCALL_LAST_PAGE && idx <= VSYSCALL_FIRST_PAGE) { + unsigned long vaddr = __fix_to_virt(idx); + set_pte_vaddr_pud(level3_user_vsyscall, vaddr, pte); + } +#endif +} + +__init void xen_post_allocator_init(void) +{ + pv_mmu_ops.set_pte = xen_set_pte; + pv_mmu_ops.set_pmd = xen_set_pmd; + pv_mmu_ops.set_pud = xen_set_pud; +#if PAGETABLE_LEVELS == 4 + pv_mmu_ops.set_pgd = xen_set_pgd; +#endif + + /* This will work as long as patching hasn't happened yet + (which it hasn't) */ + pv_mmu_ops.alloc_pte = xen_alloc_pte; + pv_mmu_ops.alloc_pmd = xen_alloc_pmd; + pv_mmu_ops.release_pte = xen_release_pte; + pv_mmu_ops.release_pmd = xen_release_pmd; +#if PAGETABLE_LEVELS == 4 + pv_mmu_ops.alloc_pud = xen_alloc_pud; + pv_mmu_ops.release_pud = xen_release_pud; +#endif + +#ifdef CONFIG_X86_64 + SetPagePinned(virt_to_page(level3_user_vsyscall)); +#endif + xen_mark_init_mm_pinned(); +} + + +const struct pv_mmu_ops xen_mmu_ops __initdata = { + .pagetable_setup_start = xen_pagetable_setup_start, + .pagetable_setup_done = xen_pagetable_setup_done, + + .read_cr2 = xen_read_cr2, + .write_cr2 = xen_write_cr2, + + .read_cr3 = xen_read_cr3, + .write_cr3 = xen_write_cr3, + + .flush_tlb_user = xen_flush_tlb, + .flush_tlb_kernel = xen_flush_tlb, + .flush_tlb_single = xen_flush_tlb_single, + .flush_tlb_others = xen_flush_tlb_others, + + .pte_update = paravirt_nop, + .pte_update_defer = paravirt_nop, + + .pgd_alloc = xen_pgd_alloc, + .pgd_free = xen_pgd_free, + + .alloc_pte = xen_alloc_pte_init, + .release_pte = xen_release_pte_init, + .alloc_pmd = xen_alloc_pte_init, + .alloc_pmd_clone = paravirt_nop, + .release_pmd = xen_release_pte_init, + +#ifdef CONFIG_HIGHPTE + .kmap_atomic_pte = xen_kmap_atomic_pte, +#endif + +#ifdef CONFIG_X86_64 + .set_pte = xen_set_pte, +#else + .set_pte = xen_set_pte_init, +#endif + .set_pte_at = xen_set_pte_at, + .set_pmd = xen_set_pmd_hyper, + + .ptep_modify_prot_start = __ptep_modify_prot_start, + .ptep_modify_prot_commit = __ptep_modify_prot_commit, + + .pte_val = xen_pte_val, + .pgd_val = xen_pgd_val, + + .make_pte = xen_make_pte, + .make_pgd = xen_make_pgd, + +#ifdef CONFIG_X86_PAE + .set_pte_atomic = xen_set_pte_atomic, + .set_pte_present = xen_set_pte_at, + .pte_clear = xen_pte_clear, + .pmd_clear = xen_pmd_clear, +#endif /* CONFIG_X86_PAE */ + .set_pud = xen_set_pud_hyper, + + .make_pmd = xen_make_pmd, + .pmd_val = xen_pmd_val, + +#if PAGETABLE_LEVELS == 4 + .pud_val = xen_pud_val, + .make_pud = xen_make_pud, + .set_pgd = xen_set_pgd_hyper, + + .alloc_pud = xen_alloc_pte_init, + .release_pud = xen_release_pte_init, +#endif /* PAGETABLE_LEVELS == 4 */ + + .activate_mm = xen_activate_mm, + .dup_mmap = xen_dup_mmap, + .exit_mmap = xen_exit_mmap, + + .lazy_mode = { + .enter = paravirt_enter_lazy_mmu, + .leave = xen_leave_lazy, + }, + + .set_fixmap = xen_set_fixmap, +}; + + #ifdef CONFIG_XEN_DEBUG_FS static struct dentry *d_mmu_debug; diff --git a/arch/x86/xen/mmu.h b/arch/x86/xen/mmu.h index 98d71659da5..24d1b44a337 100644 --- a/arch/x86/xen/mmu.h +++ b/arch/x86/xen/mmu.h @@ -54,4 +54,7 @@ pte_t xen_ptep_modify_prot_start(struct mm_struct *mm, unsigned long addr, pte_t void xen_ptep_modify_prot_commit(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pte); +unsigned long xen_read_cr2_direct(void); + +extern const struct pv_mmu_ops xen_mmu_ops; #endif /* _XEN_MMU_H */ diff --git a/arch/x86/xen/xen-ops.h b/arch/x86/xen/xen-ops.h index c1f8faf0a2c..11913fc94c1 100644 --- a/arch/x86/xen/xen-ops.h +++ b/arch/x86/xen/xen-ops.h @@ -13,6 +13,7 @@ extern const char xen_failsafe_callback[]; struct trap_info; void xen_copy_trap_info(struct trap_info *traps); +DECLARE_PER_CPU(struct vcpu_info, xen_vcpu_info); DECLARE_PER_CPU(unsigned long, xen_cr3); DECLARE_PER_CPU(unsigned long, xen_current_cr3); @@ -22,6 +23,13 @@ extern struct shared_info *HYPERVISOR_shared_info; void xen_setup_mfn_list_list(void); void xen_setup_shared_info(void); +void xen_setup_machphys_mapping(void); +pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd, unsigned long max_pfn); +void xen_ident_map_ISA(void); +void xen_reserve_top(void); + +void xen_leave_lazy(void); +void xen_post_allocator_init(void); char * __init xen_memory_setup(void); void __init xen_arch_setup(void); -- cgit v1.2.3-70-g09d2 From ecb93d1ccd0aac63f03be2db3cac3fa974716f4c Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Wed, 28 Jan 2009 14:35:05 -0800 Subject: x86/paravirt: add register-saving thunks to reduce caller register pressure Impact: Optimization One of the problems with inserting a pile of C calls where previously there were none is that the register pressure is greatly increased. The C calling convention says that the caller must expect a certain set of registers may be trashed by the callee, and that the callee can use those registers without restriction. This includes the function argument registers, and several others. This patch seeks to alleviate this pressure by introducing wrapper thunks that will do the register saving/restoring, so that the callsite doesn't need to worry about it, but the callee function can be conventional compiler-generated code. In many cases (particularly performance-sensitive cases) the callee will be in assembler anyway, and need not use the compiler's calling convention. Standard calling convention is: arguments return scratch x86-32 eax edx ecx eax ? x86-64 rdi rsi rdx rcx rax r8 r9 r10 r11 The thunk preserves all argument and scratch registers. The return register is not preserved, and is available as a scratch register for unwrapped callee code (and of course the return value). Wrapped function pointers are themselves wrapped in a struct paravirt_callee_save structure, in order to get some warning from the compiler when functions with mismatched calling conventions are used. The most common paravirt ops, both statically and dynamically, are interrupt enable/disable/save/restore, so handle them first. This is particularly easy since their calls are handled specially anyway. XXX Deal with VMI. What's their calling convention? Signed-off-by: H. Peter Anvin --- arch/x86/include/asm/paravirt.h | 126 +++++++++++++++++++++++++++++++--------- arch/x86/kernel/paravirt.c | 8 +-- arch/x86/kernel/vsmp_64.c | 12 ++-- arch/x86/lguest/boot.c | 13 +++-- arch/x86/xen/enlighten.c | 8 +-- arch/x86/xen/irq.c | 14 +++-- 6 files changed, 132 insertions(+), 49 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/include/asm/paravirt.h b/arch/x86/include/asm/paravirt.h index dcce961262b..f9107b88631 100644 --- a/arch/x86/include/asm/paravirt.h +++ b/arch/x86/include/asm/paravirt.h @@ -17,6 +17,10 @@ #ifdef CONFIG_X86_32 /* CLBR_ANY should match all regs platform has. For i386, that's just it */ #define CLBR_ANY ((1 << 4) - 1) + +#define CLBR_ARG_REGS (CLBR_EAX | CLBR_EDX | CLBR_ECX) +#define CLBR_RET_REG (CLBR_EAX) +#define CLBR_SCRATCH (0) #else #define CLBR_RAX CLBR_EAX #define CLBR_RCX CLBR_ECX @@ -27,16 +31,19 @@ #define CLBR_R9 (1 << 6) #define CLBR_R10 (1 << 7) #define CLBR_R11 (1 << 8) + #define CLBR_ANY ((1 << 9) - 1) #define CLBR_ARG_REGS (CLBR_RDI | CLBR_RSI | CLBR_RDX | \ CLBR_RCX | CLBR_R8 | CLBR_R9) -#define CLBR_RET_REG (CLBR_RAX | CLBR_RDX) +#define CLBR_RET_REG (CLBR_RAX) #define CLBR_SCRATCH (CLBR_R10 | CLBR_R11) #include #endif /* X86_64 */ +#define CLBR_CALLEE_SAVE ((CLBR_ARG_REGS | CLBR_SCRATCH) & ~CLBR_RET_REG) + #ifndef __ASSEMBLY__ #include #include @@ -50,6 +57,14 @@ struct tss_struct; struct mm_struct; struct desc_struct; +/* + * Wrapper type for pointers to code which uses the non-standard + * calling convention. See PV_CALL_SAVE_REGS_THUNK below. + */ +struct paravirt_callee_save { + void *func; +}; + /* general info */ struct pv_info { unsigned int kernel_rpl; @@ -199,11 +214,15 @@ struct pv_irq_ops { * expected to use X86_EFLAGS_IF; all other bits * returned from save_fl are undefined, and may be ignored by * restore_fl. + * + * NOTE: These functions callers expect the callee to preserve + * more registers than the standard C calling convention. */ - unsigned long (*save_fl)(void); - void (*restore_fl)(unsigned long); - void (*irq_disable)(void); - void (*irq_enable)(void); + struct paravirt_callee_save save_fl; + struct paravirt_callee_save restore_fl; + struct paravirt_callee_save irq_disable; + struct paravirt_callee_save irq_enable; + void (*safe_halt)(void); void (*halt)(void); @@ -1437,12 +1456,37 @@ extern struct paravirt_patch_site __parainstructions[], __parainstructions_end[]; #ifdef CONFIG_X86_32 -#define PV_SAVE_REGS "pushl %%ecx; pushl %%edx;" -#define PV_RESTORE_REGS "popl %%edx; popl %%ecx" +#define PV_SAVE_REGS "pushl %ecx; pushl %edx;" +#define PV_RESTORE_REGS "popl %edx; popl %ecx;" + +/* save and restore all caller-save registers, except return value */ +#define PV_SAVE_ALL_CALLER_REGS PV_SAVE_REGS +#define PV_RESTORE_ALL_CALLER_REGS PV_RESTORE_REGS + #define PV_FLAGS_ARG "0" #define PV_EXTRA_CLOBBERS #define PV_VEXTRA_CLOBBERS #else +/* save and restore all caller-save registers, except return value */ +#define PV_SAVE_ALL_CALLER_REGS \ + "push %rcx;" \ + "push %rdx;" \ + "push %rsi;" \ + "push %rdi;" \ + "push %r8;" \ + "push %r9;" \ + "push %r10;" \ + "push %r11;" +#define PV_RESTORE_ALL_CALLER_REGS \ + "pop %r11;" \ + "pop %r10;" \ + "pop %r9;" \ + "pop %r8;" \ + "pop %rdi;" \ + "pop %rsi;" \ + "pop %rdx;" \ + "pop %rcx;" + /* We save some registers, but all of them, that's too much. We clobber all * caller saved registers but the argument parameter */ #define PV_SAVE_REGS "pushq %%rdi;" @@ -1452,52 +1496,76 @@ extern struct paravirt_patch_site __parainstructions[], #define PV_FLAGS_ARG "D" #endif +/* + * Generate a thunk around a function which saves all caller-save + * registers except for the return value. This allows C functions to + * be called from assembler code where fewer than normal registers are + * available. It may also help code generation around calls from C + * code if the common case doesn't use many registers. + * + * When a callee is wrapped in a thunk, the caller can assume that all + * arg regs and all scratch registers are preserved across the + * call. The return value in rax/eax will not be saved, even for void + * functions. + */ +#define PV_CALLEE_SAVE_REGS_THUNK(func) \ + extern typeof(func) __raw_callee_save_##func; \ + static void *__##func##__ __used = func; \ + \ + asm(".pushsection .text;" \ + "__raw_callee_save_" #func ": " \ + PV_SAVE_ALL_CALLER_REGS \ + "call " #func ";" \ + PV_RESTORE_ALL_CALLER_REGS \ + "ret;" \ + ".popsection") + +/* Get a reference to a callee-save function */ +#define PV_CALLEE_SAVE(func) \ + ((struct paravirt_callee_save) { __raw_callee_save_##func }) + +/* Promise that "func" already uses the right calling convention */ +#define __PV_IS_CALLEE_SAVE(func) \ + ((struct paravirt_callee_save) { func }) + static inline unsigned long __raw_local_save_flags(void) { unsigned long f; - asm volatile(paravirt_alt(PV_SAVE_REGS - PARAVIRT_CALL - PV_RESTORE_REGS) + asm volatile(paravirt_alt(PARAVIRT_CALL) : "=a"(f) : paravirt_type(pv_irq_ops.save_fl), paravirt_clobber(CLBR_EAX) - : "memory", "cc" PV_VEXTRA_CLOBBERS); + : "memory", "cc"); return f; } static inline void raw_local_irq_restore(unsigned long f) { - asm volatile(paravirt_alt(PV_SAVE_REGS - PARAVIRT_CALL - PV_RESTORE_REGS) + asm volatile(paravirt_alt(PARAVIRT_CALL) : "=a"(f) : PV_FLAGS_ARG(f), paravirt_type(pv_irq_ops.restore_fl), paravirt_clobber(CLBR_EAX) - : "memory", "cc" PV_EXTRA_CLOBBERS); + : "memory", "cc"); } static inline void raw_local_irq_disable(void) { - asm volatile(paravirt_alt(PV_SAVE_REGS - PARAVIRT_CALL - PV_RESTORE_REGS) + asm volatile(paravirt_alt(PARAVIRT_CALL) : : paravirt_type(pv_irq_ops.irq_disable), paravirt_clobber(CLBR_EAX) - : "memory", "eax", "cc" PV_EXTRA_CLOBBERS); + : "memory", "eax", "cc"); } static inline void raw_local_irq_enable(void) { - asm volatile(paravirt_alt(PV_SAVE_REGS - PARAVIRT_CALL - PV_RESTORE_REGS) + asm volatile(paravirt_alt(PARAVIRT_CALL) : : paravirt_type(pv_irq_ops.irq_enable), paravirt_clobber(CLBR_EAX) - : "memory", "eax", "cc" PV_EXTRA_CLOBBERS); + : "memory", "eax", "cc"); } static inline unsigned long __raw_local_irq_save(void) @@ -1541,9 +1609,9 @@ static inline unsigned long __raw_local_irq_save(void) #define COND_PUSH(set, mask, reg) \ - .if ((~set) & mask); push %reg; .endif + .if ((~(set)) & mask); push %reg; .endif #define COND_POP(set, mask, reg) \ - .if ((~set) & mask); pop %reg; .endif + .if ((~(set)) & mask); pop %reg; .endif #ifdef CONFIG_X86_64 @@ -1594,15 +1662,15 @@ static inline unsigned long __raw_local_irq_save(void) #define DISABLE_INTERRUPTS(clobbers) \ PARA_SITE(PARA_PATCH(pv_irq_ops, PV_IRQ_irq_disable), clobbers, \ - PV_SAVE_REGS(clobbers); \ + PV_SAVE_REGS(clobbers | CLBR_CALLEE_SAVE); \ call PARA_INDIRECT(pv_irq_ops+PV_IRQ_irq_disable); \ - PV_RESTORE_REGS(clobbers);) + PV_RESTORE_REGS(clobbers | CLBR_CALLEE_SAVE);) #define ENABLE_INTERRUPTS(clobbers) \ PARA_SITE(PARA_PATCH(pv_irq_ops, PV_IRQ_irq_enable), clobbers, \ - PV_SAVE_REGS(clobbers); \ + PV_SAVE_REGS(clobbers | CLBR_CALLEE_SAVE); \ call PARA_INDIRECT(pv_irq_ops+PV_IRQ_irq_enable); \ - PV_RESTORE_REGS(clobbers);) + PV_RESTORE_REGS(clobbers | CLBR_CALLEE_SAVE);) #define USERGS_SYSRET32 \ PARA_SITE(PARA_PATCH(pv_cpu_ops, PV_CPU_usergs_sysret32), \ diff --git a/arch/x86/kernel/paravirt.c b/arch/x86/kernel/paravirt.c index dd25e2b1593..8adb6b5aa42 100644 --- a/arch/x86/kernel/paravirt.c +++ b/arch/x86/kernel/paravirt.c @@ -310,10 +310,10 @@ struct pv_time_ops pv_time_ops = { struct pv_irq_ops pv_irq_ops = { .init_IRQ = native_init_IRQ, - .save_fl = native_save_fl, - .restore_fl = native_restore_fl, - .irq_disable = native_irq_disable, - .irq_enable = native_irq_enable, + .save_fl = __PV_IS_CALLEE_SAVE(native_save_fl), + .restore_fl = __PV_IS_CALLEE_SAVE(native_restore_fl), + .irq_disable = __PV_IS_CALLEE_SAVE(native_irq_disable), + .irq_enable = __PV_IS_CALLEE_SAVE(native_irq_enable), .safe_halt = native_safe_halt, .halt = native_halt, #ifdef CONFIG_X86_64 diff --git a/arch/x86/kernel/vsmp_64.c b/arch/x86/kernel/vsmp_64.c index a688f3bfaec..c609205df59 100644 --- a/arch/x86/kernel/vsmp_64.c +++ b/arch/x86/kernel/vsmp_64.c @@ -37,6 +37,7 @@ static unsigned long vsmp_save_fl(void) flags &= ~X86_EFLAGS_IF; return flags; } +PV_CALLEE_SAVE_REGS_THUNK(vsmp_save_fl); static void vsmp_restore_fl(unsigned long flags) { @@ -46,6 +47,7 @@ static void vsmp_restore_fl(unsigned long flags) flags |= X86_EFLAGS_AC; native_restore_fl(flags); } +PV_CALLEE_SAVE_REGS_THUNK(vsmp_restore_fl); static void vsmp_irq_disable(void) { @@ -53,6 +55,7 @@ static void vsmp_irq_disable(void) native_restore_fl((flags & ~X86_EFLAGS_IF) | X86_EFLAGS_AC); } +PV_CALLEE_SAVE_REGS_THUNK(vsmp_irq_disable); static void vsmp_irq_enable(void) { @@ -60,6 +63,7 @@ static void vsmp_irq_enable(void) native_restore_fl((flags | X86_EFLAGS_IF) & (~X86_EFLAGS_AC)); } +PV_CALLEE_SAVE_REGS_THUNK(vsmp_irq_enable); static unsigned __init_or_module vsmp_patch(u8 type, u16 clobbers, void *ibuf, unsigned long addr, unsigned len) @@ -90,10 +94,10 @@ static void __init set_vsmp_pv_ops(void) cap, ctl); if (cap & ctl & (1 << 4)) { /* Setup irq ops and turn on vSMP IRQ fastpath handling */ - pv_irq_ops.irq_disable = vsmp_irq_disable; - pv_irq_ops.irq_enable = vsmp_irq_enable; - pv_irq_ops.save_fl = vsmp_save_fl; - pv_irq_ops.restore_fl = vsmp_restore_fl; + pv_irq_ops.irq_disable = PV_CALLEE_SAVE(vsmp_irq_disable); + pv_irq_ops.irq_enable = PV_CALLEE_SAVE(vsmp_irq_enable); + pv_irq_ops.save_fl = PV_CALLEE_SAVE(vsmp_save_fl); + pv_irq_ops.restore_fl = PV_CALLEE_SAVE(vsmp_restore_fl); pv_init_ops.patch = vsmp_patch; ctl &= ~(1 << 4); diff --git a/arch/x86/lguest/boot.c b/arch/x86/lguest/boot.c index 92f1c6f3e19..19e33b6cd59 100644 --- a/arch/x86/lguest/boot.c +++ b/arch/x86/lguest/boot.c @@ -173,24 +173,29 @@ static unsigned long save_fl(void) { return lguest_data.irq_enabled; } +PV_CALLEE_SAVE_REGS_THUNK(save_fl); /* restore_flags() just sets the flags back to the value given. */ static void restore_fl(unsigned long flags) { lguest_data.irq_enabled = flags; } +PV_CALLEE_SAVE_REGS_THUNK(restore_fl); /* Interrupts go off... */ static void irq_disable(void) { lguest_data.irq_enabled = 0; } +PV_CALLEE_SAVE_REGS_THUNK(irq_disable); /* Interrupts go on... */ static void irq_enable(void) { lguest_data.irq_enabled = X86_EFLAGS_IF; } +PV_CALLEE_SAVE_REGS_THUNK(irq_enable); + /*:*/ /*M:003 Note that we don't check for outstanding interrupts when we re-enable * them (or when we unmask an interrupt). This seems to work for the moment, @@ -984,10 +989,10 @@ __init void lguest_init(void) /* interrupt-related operations */ pv_irq_ops.init_IRQ = lguest_init_IRQ; - pv_irq_ops.save_fl = save_fl; - pv_irq_ops.restore_fl = restore_fl; - pv_irq_ops.irq_disable = irq_disable; - pv_irq_ops.irq_enable = irq_enable; + pv_irq_ops.save_fl = PV_CALLEE_SAVE(save_fl); + pv_irq_ops.restore_fl = PV_CALLEE_SAVE(restore_fl); + pv_irq_ops.irq_disable = PV_CALLEE_SAVE(irq_disable); + pv_irq_ops.irq_enable = PV_CALLEE_SAVE(irq_enable); pv_irq_ops.safe_halt = lguest_safe_halt; /* init-time operations */ diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 0cd2a165f17..ff6d530ccc7 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -676,10 +676,10 @@ void xen_setup_vcpu_info_placement(void) if (have_vcpu_info_placement) { printk(KERN_INFO "Xen: using vcpu_info placement\n"); - pv_irq_ops.save_fl = xen_save_fl_direct; - pv_irq_ops.restore_fl = xen_restore_fl_direct; - pv_irq_ops.irq_disable = xen_irq_disable_direct; - pv_irq_ops.irq_enable = xen_irq_enable_direct; + pv_irq_ops.save_fl = __PV_IS_CALLEE_SAVE(xen_save_fl_direct); + pv_irq_ops.restore_fl = __PV_IS_CALLEE_SAVE(xen_restore_fl_direct); + pv_irq_ops.irq_disable = __PV_IS_CALLEE_SAVE(xen_irq_disable_direct); + pv_irq_ops.irq_enable = __PV_IS_CALLEE_SAVE(xen_irq_enable_direct); pv_mmu_ops.read_cr2 = xen_read_cr2_direct; } } diff --git a/arch/x86/xen/irq.c b/arch/x86/xen/irq.c index 2e8271431e1..5a070900ad3 100644 --- a/arch/x86/xen/irq.c +++ b/arch/x86/xen/irq.c @@ -50,6 +50,7 @@ static unsigned long xen_save_fl(void) */ return (-flags) & X86_EFLAGS_IF; } +PV_CALLEE_SAVE_REGS_THUNK(xen_save_fl); static void xen_restore_fl(unsigned long flags) { @@ -76,6 +77,7 @@ static void xen_restore_fl(unsigned long flags) xen_force_evtchn_callback(); } } +PV_CALLEE_SAVE_REGS_THUNK(xen_restore_fl); static void xen_irq_disable(void) { @@ -86,6 +88,7 @@ static void xen_irq_disable(void) percpu_read(xen_vcpu)->evtchn_upcall_mask = 1; preempt_enable_no_resched(); } +PV_CALLEE_SAVE_REGS_THUNK(xen_irq_disable); static void xen_irq_enable(void) { @@ -106,6 +109,7 @@ static void xen_irq_enable(void) if (unlikely(vcpu->evtchn_upcall_pending)) xen_force_evtchn_callback(); } +PV_CALLEE_SAVE_REGS_THUNK(xen_irq_enable); static void xen_safe_halt(void) { @@ -124,10 +128,12 @@ static void xen_halt(void) static const struct pv_irq_ops xen_irq_ops __initdata = { .init_IRQ = __xen_init_IRQ, - .save_fl = xen_save_fl, - .restore_fl = xen_restore_fl, - .irq_disable = xen_irq_disable, - .irq_enable = xen_irq_enable, + + .save_fl = PV_CALLEE_SAVE(xen_save_fl), + .restore_fl = PV_CALLEE_SAVE(xen_restore_fl), + .irq_disable = PV_CALLEE_SAVE(xen_irq_disable), + .irq_enable = PV_CALLEE_SAVE(xen_irq_enable), + .safe_halt = xen_safe_halt, .halt = xen_halt, #ifdef CONFIG_X86_64 -- cgit v1.2.3-70-g09d2 From da5de7c22eb705be709a57e486e7475a6969b994 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Wed, 28 Jan 2009 14:35:07 -0800 Subject: x86/paravirt: use callee-saved convention for pte_val/make_pte/etc Impact: Optimization In the native case, pte_val, make_pte, etc are all just identity functions, so there's no need to clobber a lot of registers over them. (This changes the 32-bit callee-save calling convention to return both EAX and EDX so functions can return 64-bit values.) Signed-off-by: Jeremy Fitzhardinge Signed-off-by: H. Peter Anvin --- arch/x86/include/asm/paravirt.h | 78 ++++++++++++++++++++--------------------- arch/x86/kernel/paravirt.c | 53 +++++++--------------------- arch/x86/xen/mmu.c | 24 ++++++++----- 3 files changed, 67 insertions(+), 88 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/include/asm/paravirt.h b/arch/x86/include/asm/paravirt.h index beb10ecdbe6..2d098b78bc1 100644 --- a/arch/x86/include/asm/paravirt.h +++ b/arch/x86/include/asm/paravirt.h @@ -19,7 +19,7 @@ #define CLBR_ANY ((1 << 4) - 1) #define CLBR_ARG_REGS (CLBR_EAX | CLBR_EDX | CLBR_ECX) -#define CLBR_RET_REG (CLBR_EAX) +#define CLBR_RET_REG (CLBR_EAX | CLBR_EDX) #define CLBR_SCRATCH (0) #else #define CLBR_RAX CLBR_EAX @@ -308,11 +308,11 @@ struct pv_mmu_ops { void (*ptep_modify_prot_commit)(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pte); - pteval_t (*pte_val)(pte_t); - pte_t (*make_pte)(pteval_t pte); + struct paravirt_callee_save pte_val; + struct paravirt_callee_save make_pte; - pgdval_t (*pgd_val)(pgd_t); - pgd_t (*make_pgd)(pgdval_t pgd); + struct paravirt_callee_save pgd_val; + struct paravirt_callee_save make_pgd; #if PAGETABLE_LEVELS >= 3 #ifdef CONFIG_X86_PAE @@ -327,12 +327,12 @@ struct pv_mmu_ops { void (*set_pud)(pud_t *pudp, pud_t pudval); - pmdval_t (*pmd_val)(pmd_t); - pmd_t (*make_pmd)(pmdval_t pmd); + struct paravirt_callee_save pmd_val; + struct paravirt_callee_save make_pmd; #if PAGETABLE_LEVELS == 4 - pudval_t (*pud_val)(pud_t); - pud_t (*make_pud)(pudval_t pud); + struct paravirt_callee_save pud_val; + struct paravirt_callee_save make_pud; void (*set_pgd)(pgd_t *pudp, pgd_t pgdval); #endif /* PAGETABLE_LEVELS == 4 */ @@ -1155,13 +1155,13 @@ static inline pte_t __pte(pteval_t val) pteval_t ret; if (sizeof(pteval_t) > sizeof(long)) - ret = PVOP_CALL2(pteval_t, - pv_mmu_ops.make_pte, - val, (u64)val >> 32); + ret = PVOP_CALLEE2(pteval_t, + pv_mmu_ops.make_pte, + val, (u64)val >> 32); else - ret = PVOP_CALL1(pteval_t, - pv_mmu_ops.make_pte, - val); + ret = PVOP_CALLEE1(pteval_t, + pv_mmu_ops.make_pte, + val); return (pte_t) { .pte = ret }; } @@ -1171,11 +1171,11 @@ static inline pteval_t pte_val(pte_t pte) pteval_t ret; if (sizeof(pteval_t) > sizeof(long)) - ret = PVOP_CALL2(pteval_t, pv_mmu_ops.pte_val, - pte.pte, (u64)pte.pte >> 32); + ret = PVOP_CALLEE2(pteval_t, pv_mmu_ops.pte_val, + pte.pte, (u64)pte.pte >> 32); else - ret = PVOP_CALL1(pteval_t, pv_mmu_ops.pte_val, - pte.pte); + ret = PVOP_CALLEE1(pteval_t, pv_mmu_ops.pte_val, + pte.pte); return ret; } @@ -1185,11 +1185,11 @@ static inline pgd_t __pgd(pgdval_t val) pgdval_t ret; if (sizeof(pgdval_t) > sizeof(long)) - ret = PVOP_CALL2(pgdval_t, pv_mmu_ops.make_pgd, - val, (u64)val >> 32); + ret = PVOP_CALLEE2(pgdval_t, pv_mmu_ops.make_pgd, + val, (u64)val >> 32); else - ret = PVOP_CALL1(pgdval_t, pv_mmu_ops.make_pgd, - val); + ret = PVOP_CALLEE1(pgdval_t, pv_mmu_ops.make_pgd, + val); return (pgd_t) { ret }; } @@ -1199,11 +1199,11 @@ static inline pgdval_t pgd_val(pgd_t pgd) pgdval_t ret; if (sizeof(pgdval_t) > sizeof(long)) - ret = PVOP_CALL2(pgdval_t, pv_mmu_ops.pgd_val, - pgd.pgd, (u64)pgd.pgd >> 32); + ret = PVOP_CALLEE2(pgdval_t, pv_mmu_ops.pgd_val, + pgd.pgd, (u64)pgd.pgd >> 32); else - ret = PVOP_CALL1(pgdval_t, pv_mmu_ops.pgd_val, - pgd.pgd); + ret = PVOP_CALLEE1(pgdval_t, pv_mmu_ops.pgd_val, + pgd.pgd); return ret; } @@ -1267,11 +1267,11 @@ static inline pmd_t __pmd(pmdval_t val) pmdval_t ret; if (sizeof(pmdval_t) > sizeof(long)) - ret = PVOP_CALL2(pmdval_t, pv_mmu_ops.make_pmd, - val, (u64)val >> 32); + ret = PVOP_CALLEE2(pmdval_t, pv_mmu_ops.make_pmd, + val, (u64)val >> 32); else - ret = PVOP_CALL1(pmdval_t, pv_mmu_ops.make_pmd, - val); + ret = PVOP_CALLEE1(pmdval_t, pv_mmu_ops.make_pmd, + val); return (pmd_t) { ret }; } @@ -1281,11 +1281,11 @@ static inline pmdval_t pmd_val(pmd_t pmd) pmdval_t ret; if (sizeof(pmdval_t) > sizeof(long)) - ret = PVOP_CALL2(pmdval_t, pv_mmu_ops.pmd_val, - pmd.pmd, (u64)pmd.pmd >> 32); + ret = PVOP_CALLEE2(pmdval_t, pv_mmu_ops.pmd_val, + pmd.pmd, (u64)pmd.pmd >> 32); else - ret = PVOP_CALL1(pmdval_t, pv_mmu_ops.pmd_val, - pmd.pmd); + ret = PVOP_CALLEE1(pmdval_t, pv_mmu_ops.pmd_val, + pmd.pmd); return ret; } @@ -1307,11 +1307,11 @@ static inline pud_t __pud(pudval_t val) pudval_t ret; if (sizeof(pudval_t) > sizeof(long)) - ret = PVOP_CALL2(pudval_t, pv_mmu_ops.make_pud, - val, (u64)val >> 32); + ret = PVOP_CALLEE2(pudval_t, pv_mmu_ops.make_pud, + val, (u64)val >> 32); else - ret = PVOP_CALL1(pudval_t, pv_mmu_ops.make_pud, - val); + ret = PVOP_CALLEE1(pudval_t, pv_mmu_ops.make_pud, + val); return (pud_t) { ret }; } diff --git a/arch/x86/kernel/paravirt.c b/arch/x86/kernel/paravirt.c index 8adb6b5aa42..cea11c8e304 100644 --- a/arch/x86/kernel/paravirt.c +++ b/arch/x86/kernel/paravirt.c @@ -391,43 +391,12 @@ struct pv_apic_ops pv_apic_ops = { #endif }; -typedef pte_t make_pte_t(pteval_t); -typedef pmd_t make_pmd_t(pmdval_t); -typedef pud_t make_pud_t(pudval_t); -typedef pgd_t make_pgd_t(pgdval_t); - -typedef pteval_t pte_val_t(pte_t); -typedef pmdval_t pmd_val_t(pmd_t); -typedef pudval_t pud_val_t(pud_t); -typedef pgdval_t pgd_val_t(pgd_t); - - #if defined(CONFIG_X86_32) && !defined(CONFIG_X86_PAE) /* 32-bit pagetable entries */ -#define paravirt_native_make_pte (make_pte_t *)_paravirt_ident_32 -#define paravirt_native_pte_val (pte_val_t *)_paravirt_ident_32 - -#define paravirt_native_make_pmd (make_pmd_t *)_paravirt_ident_32 -#define paravirt_native_pmd_val (pmd_val_t *)_paravirt_ident_32 - -#define paravirt_native_make_pud (make_pud_t *)_paravirt_ident_32 -#define paravirt_native_pud_val (pud_val_t *)_paravirt_ident_32 - -#define paravirt_native_make_pgd (make_pgd_t *)_paravirt_ident_32 -#define paravirt_native_pgd_val (pgd_val_t *)_paravirt_ident_32 +#define PTE_IDENT __PV_IS_CALLEE_SAVE(_paravirt_ident_32) #else /* 64-bit pagetable entries */ -#define paravirt_native_make_pte (make_pte_t *)_paravirt_ident_64 -#define paravirt_native_pte_val (pte_val_t *)_paravirt_ident_64 - -#define paravirt_native_make_pmd (make_pmd_t *)_paravirt_ident_64 -#define paravirt_native_pmd_val (pmd_val_t *)_paravirt_ident_64 - -#define paravirt_native_make_pud (make_pud_t *)_paravirt_ident_64 -#define paravirt_native_pud_val (pud_val_t *)_paravirt_ident_64 - -#define paravirt_native_make_pgd (make_pgd_t *)_paravirt_ident_64 -#define paravirt_native_pgd_val (pgd_val_t *)_paravirt_ident_64 +#define PTE_IDENT __PV_IS_CALLEE_SAVE(_paravirt_ident_64) #endif struct pv_mmu_ops pv_mmu_ops = { @@ -481,21 +450,23 @@ struct pv_mmu_ops pv_mmu_ops = { .pmd_clear = native_pmd_clear, #endif .set_pud = native_set_pud, - .pmd_val = paravirt_native_pmd_val, - .make_pmd = paravirt_native_make_pmd, + + .pmd_val = PTE_IDENT, + .make_pmd = PTE_IDENT, #if PAGETABLE_LEVELS == 4 - .pud_val = paravirt_native_pud_val, - .make_pud = paravirt_native_make_pud, + .pud_val = PTE_IDENT, + .make_pud = PTE_IDENT, + .set_pgd = native_set_pgd, #endif #endif /* PAGETABLE_LEVELS >= 3 */ - .pte_val = paravirt_native_pte_val, - .pgd_val = paravirt_native_pgd_val, + .pte_val = PTE_IDENT, + .pgd_val = PTE_IDENT, - .make_pte = paravirt_native_make_pte, - .make_pgd = paravirt_native_make_pgd, + .make_pte = PTE_IDENT, + .make_pgd = PTE_IDENT, .dup_mmap = paravirt_nop, .exit_mmap = paravirt_nop, diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 94e452c0b00..5e41f7fc6cf 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -492,28 +492,33 @@ pteval_t xen_pte_val(pte_t pte) { return pte_mfn_to_pfn(pte.pte); } +PV_CALLEE_SAVE_REGS_THUNK(xen_pte_val); pgdval_t xen_pgd_val(pgd_t pgd) { return pte_mfn_to_pfn(pgd.pgd); } +PV_CALLEE_SAVE_REGS_THUNK(xen_pgd_val); pte_t xen_make_pte(pteval_t pte) { pte = pte_pfn_to_mfn(pte); return native_make_pte(pte); } +PV_CALLEE_SAVE_REGS_THUNK(xen_make_pte); pgd_t xen_make_pgd(pgdval_t pgd) { pgd = pte_pfn_to_mfn(pgd); return native_make_pgd(pgd); } +PV_CALLEE_SAVE_REGS_THUNK(xen_make_pgd); pmdval_t xen_pmd_val(pmd_t pmd) { return pte_mfn_to_pfn(pmd.pmd); } +PV_CALLEE_SAVE_REGS_THUNK(xen_pmd_val); void xen_set_pud_hyper(pud_t *ptr, pud_t val) { @@ -590,12 +595,14 @@ pmd_t xen_make_pmd(pmdval_t pmd) pmd = pte_pfn_to_mfn(pmd); return native_make_pmd(pmd); } +PV_CALLEE_SAVE_REGS_THUNK(xen_make_pmd); #if PAGETABLE_LEVELS == 4 pudval_t xen_pud_val(pud_t pud) { return pte_mfn_to_pfn(pud.pud); } +PV_CALLEE_SAVE_REGS_THUNK(xen_pud_val); pud_t xen_make_pud(pudval_t pud) { @@ -603,6 +610,7 @@ pud_t xen_make_pud(pudval_t pud) return native_make_pud(pud); } +PV_CALLEE_SAVE_REGS_THUNK(xen_make_pud); pgd_t *xen_get_user_pgd(pgd_t *pgd) { @@ -1813,11 +1821,11 @@ const struct pv_mmu_ops xen_mmu_ops __initdata = { .ptep_modify_prot_start = __ptep_modify_prot_start, .ptep_modify_prot_commit = __ptep_modify_prot_commit, - .pte_val = xen_pte_val, - .pgd_val = xen_pgd_val, + .pte_val = PV_CALLEE_SAVE(xen_pte_val), + .pgd_val = PV_CALLEE_SAVE(xen_pgd_val), - .make_pte = xen_make_pte, - .make_pgd = xen_make_pgd, + .make_pte = PV_CALLEE_SAVE(xen_make_pte), + .make_pgd = PV_CALLEE_SAVE(xen_make_pgd), #ifdef CONFIG_X86_PAE .set_pte_atomic = xen_set_pte_atomic, @@ -1827,12 +1835,12 @@ const struct pv_mmu_ops xen_mmu_ops __initdata = { #endif /* CONFIG_X86_PAE */ .set_pud = xen_set_pud_hyper, - .make_pmd = xen_make_pmd, - .pmd_val = xen_pmd_val, + .make_pmd = PV_CALLEE_SAVE(xen_make_pmd), + .pmd_val = PV_CALLEE_SAVE(xen_pmd_val), #if PAGETABLE_LEVELS == 4 - .pud_val = xen_pud_val, - .make_pud = xen_make_pud, + .pud_val = PV_CALLEE_SAVE(xen_pud_val), + .make_pud = PV_CALLEE_SAVE(xen_make_pud), .set_pgd = xen_set_pgd_hyper, .alloc_pud = xen_alloc_pte_init, -- cgit v1.2.3-70-g09d2 From 795f99b61d20c34cb04d17d8906b32f745a635ec Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Fri, 30 Jan 2009 17:47:54 +0900 Subject: xen: setup percpu data pointers Impact: fix xen booting We need to access percpu data fairly early, so set up the percpu registers as soon as possible. We only need to load the appropriate segment register. We already have a GDT, but its hard to change it early because we need to manipulate the pagetable to do so, and that hasn't been set up yet. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Tejun Heo --- arch/x86/xen/enlighten.c | 3 +++ arch/x86/xen/smp.c | 5 ++++- 2 files changed, 7 insertions(+), 1 deletion(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index bef941f6145..fe19c88a502 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1647,6 +1647,9 @@ asmlinkage void __init xen_start_kernel(void) have_vcpu_info_placement = 0; #endif + /* setup percpu state */ + load_percpu_segment(0); + xen_smp_init(); /* Get mfn list */ diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c index 7735e3dd359..88d5d5ec6be 100644 --- a/arch/x86/xen/smp.c +++ b/arch/x86/xen/smp.c @@ -170,7 +170,8 @@ static void __init xen_smp_prepare_boot_cpu(void) /* We've switched to the "real" per-cpu gdt, so make sure the old memory can be recycled */ - make_lowmem_page_readwrite(&per_cpu_var(gdt_page)); + make_lowmem_page_readwrite(__per_cpu_load + + (unsigned long)&per_cpu_var(gdt_page)); xen_setup_vcpu_info_placement(); } @@ -235,6 +236,8 @@ cpu_initialize_context(unsigned int cpu, struct task_struct *idle) ctxt->user_regs.ss = __KERNEL_DS; #ifdef CONFIG_X86_32 ctxt->user_regs.fs = __KERNEL_PERCPU; +#else + ctxt->gs_base_kernel = per_cpu_offset(cpu); #endif ctxt->user_regs.eip = (unsigned long)cpu_bringup_and_idle; ctxt->user_regs.eflags = 0x1000; /* IOPL_RING1 */ -- cgit v1.2.3-70-g09d2 From 1f4f931501e9270c156d05ee76b7b872de486304 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 2 Feb 2009 13:58:06 -0800 Subject: xen: fix 32-bit build resulting from mmu move Moving the mmu code from enlighten.c to mmu.c inadvertently broke the 32-bit build. Fix it. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: H. Peter Anvin --- arch/x86/xen/mmu.c | 37 +++++++++++++++++++++++++++++++++++++ 1 file changed, 37 insertions(+) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 5e41f7fc6cf..d2e8ed1aff3 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -1396,6 +1396,43 @@ static void xen_pgd_free(struct mm_struct *mm, pgd_t *pgd) #endif } +#ifdef CONFIG_HIGHPTE +static void *xen_kmap_atomic_pte(struct page *page, enum km_type type) +{ + pgprot_t prot = PAGE_KERNEL; + + if (PagePinned(page)) + prot = PAGE_KERNEL_RO; + + if (0 && PageHighMem(page)) + printk("mapping highpte %lx type %d prot %s\n", + page_to_pfn(page), type, + (unsigned long)pgprot_val(prot) & _PAGE_RW ? "WRITE" : "READ"); + + return kmap_atomic_prot(page, type, prot); +} +#endif + +#ifdef CONFIG_X86_32 +static __init pte_t mask_rw_pte(pte_t *ptep, pte_t pte) +{ + /* If there's an existing pte, then don't allow _PAGE_RW to be set */ + if (pte_val_ma(*ptep) & _PAGE_PRESENT) + pte = __pte_ma(((pte_val_ma(*ptep) & _PAGE_RW) | ~_PAGE_RW) & + pte_val_ma(pte)); + + return pte; +} + +/* Init-time set_pte while constructing initial pagetables, which + doesn't allow RO pagetable pages to be remapped RW */ +static __init void xen_set_pte_init(pte_t *ptep, pte_t pte) +{ + pte = mask_rw_pte(ptep, pte); + + xen_set_pte(ptep, pte); +} +#endif /* Early in boot, while setting up the initial pagetable, assume everything is pinned. */ -- cgit v1.2.3-70-g09d2 From 383414322b3b3ced0cbc146801e0cc6c60a6c5f4 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 2 Feb 2009 13:55:31 -0800 Subject: xen: setup percpu data pointers We need to access percpu data fairly early, so set up the percpu registers as soon as possible. We only need to load the appropriate segment register. We already have a GDT, but its hard to change it early because we need to manipulate the pagetable to do so, and that hasn't been set up yet. Also, set the kernel stack when bringing up secondary CPUs. If we don't they all end up sharing the same stack... Signed-off-by: Jeremy Fitzhardinge Signed-off-by: H. Peter Anvin --- arch/x86/xen/enlighten.c | 15 ++++++++++++++- arch/x86/xen/smp.c | 6 ++++-- arch/x86/xen/xen-ops.h | 2 ++ 3 files changed, 20 insertions(+), 3 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index cd022c43dfb..aed7ceeb4b6 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -66,6 +66,8 @@ EXPORT_SYMBOL_GPL(xen_start_info); struct shared_info xen_dummy_shared_info; +void *xen_initial_gdt; + /* * Point at some empty memory to start with. We map the real shared_info * page as soon as fixmap is up and running. @@ -917,8 +919,19 @@ asmlinkage void __init xen_start_kernel(void) have_vcpu_info_placement = 0; #endif - /* setup percpu state */ +#ifdef CONFIG_X86_64 + /* + * Setup percpu state. We only need to do this for 64-bit + * because 32-bit already has %fs set properly. + */ load_percpu_segment(0); +#endif + /* + * The only reliable way to retain the initial address of the + * percpu gdt_page is to remember it here, so we can go and + * mark it RW later, when the initial percpu area is freed. + */ + xen_initial_gdt = &per_cpu(gdt_page, 0); xen_smp_init(); diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c index 88d5d5ec6be..035582ae815 100644 --- a/arch/x86/xen/smp.c +++ b/arch/x86/xen/smp.c @@ -170,8 +170,7 @@ static void __init xen_smp_prepare_boot_cpu(void) /* We've switched to the "real" per-cpu gdt, so make sure the old memory can be recycled */ - make_lowmem_page_readwrite(__per_cpu_load + - (unsigned long)&per_cpu_var(gdt_page)); + make_lowmem_page_readwrite(xen_initial_gdt); xen_setup_vcpu_info_placement(); } @@ -287,6 +286,9 @@ static int __cpuinit xen_cpu_up(unsigned int cpu) irq_ctx_init(cpu); #else clear_tsk_thread_flag(idle, TIF_FORK); + per_cpu(kernel_stack, cpu) = + (unsigned long)task_stack_page(idle) - + KERNEL_STACK_OFFSET + THREAD_SIZE; #endif xen_setup_timer(cpu); xen_init_lock_cpu(cpu); diff --git a/arch/x86/xen/xen-ops.h b/arch/x86/xen/xen-ops.h index 11913fc94c1..2f5ef2632ea 100644 --- a/arch/x86/xen/xen-ops.h +++ b/arch/x86/xen/xen-ops.h @@ -10,6 +10,8 @@ extern const char xen_hypervisor_callback[]; extern const char xen_failsafe_callback[]; +extern void *xen_initial_gdt; + struct trap_info; void xen_copy_trap_info(struct trap_info *traps); -- cgit v1.2.3-70-g09d2 From 5393744b71ce797f1b1546fafaed127fc50c2b61 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 2 Feb 2009 13:55:42 -0800 Subject: xen: make direct versions of irq_enable/disable/save/restore to common code Now that x86-64 has directly accessible percpu variables, it can also implement the direct versions of these operations, which operate on a vcpu_info structure directly embedded in the percpu area. In fact, the 64-bit versions are more or less identical, and so can be shared. The only two differences are: 1. xen_restore_fl_direct takes its argument in eax on 32-bit, and rdi on 64-bit. Unfortunately it isn't possible to directly refer to the 2nd lsb of rdi directly (as you can with %ah), so the code isn't quite as dense. 2. check_events needs to variants to save different registers. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: H. Peter Anvin --- arch/x86/xen/Makefile | 3 +- arch/x86/xen/xen-asm.S | 140 ++++++++++++++++++++++++++++++++++++++++++++++ arch/x86/xen/xen-asm.h | 12 ++++ arch/x86/xen/xen-asm_32.S | 111 ++++-------------------------------- arch/x86/xen/xen-asm_64.S | 134 +------------------------------------------- 5 files changed, 169 insertions(+), 231 deletions(-) create mode 100644 arch/x86/xen/xen-asm.S create mode 100644 arch/x86/xen/xen-asm.h (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/Makefile b/arch/x86/xen/Makefile index 6dcefba7836..3b767d03fd6 100644 --- a/arch/x86/xen/Makefile +++ b/arch/x86/xen/Makefile @@ -6,7 +6,8 @@ CFLAGS_REMOVE_irq.o = -pg endif obj-y := enlighten.o setup.o multicalls.o mmu.o irq.o \ - time.o xen-asm_$(BITS).o grant-table.o suspend.o + time.o xen-asm.o xen-asm_$(BITS).o \ + grant-table.o suspend.o obj-$(CONFIG_SMP) += smp.o spinlock.o obj-$(CONFIG_XEN_DEBUG_FS) += debugfs.o \ No newline at end of file diff --git a/arch/x86/xen/xen-asm.S b/arch/x86/xen/xen-asm.S new file mode 100644 index 00000000000..4c6f9679913 --- /dev/null +++ b/arch/x86/xen/xen-asm.S @@ -0,0 +1,140 @@ +/* + Asm versions of Xen pv-ops, suitable for either direct use or inlining. + The inline versions are the same as the direct-use versions, with the + pre- and post-amble chopped off. + + This code is encoded for size rather than absolute efficiency, + with a view to being able to inline as much as possible. + + We only bother with direct forms (ie, vcpu in percpu data) of + the operations here; the indirect forms are better handled in + C, since they're generally too large to inline anyway. + */ + +#include +#include +#include + +#include "xen-asm.h" + +/* + Enable events. This clears the event mask and tests the pending + event status with one and operation. If there are pending + events, then enter the hypervisor to get them handled. + */ +ENTRY(xen_irq_enable_direct) + /* Unmask events */ + movb $0, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_mask + + /* Preempt here doesn't matter because that will deal with + any pending interrupts. The pending check may end up being + run on the wrong CPU, but that doesn't hurt. */ + + /* Test for pending */ + testb $0xff, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_pending + jz 1f + +2: call check_events +1: +ENDPATCH(xen_irq_enable_direct) + ret + ENDPROC(xen_irq_enable_direct) + RELOC(xen_irq_enable_direct, 2b+1) + + +/* + Disabling events is simply a matter of making the event mask + non-zero. + */ +ENTRY(xen_irq_disable_direct) + movb $1, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_mask +ENDPATCH(xen_irq_disable_direct) + ret + ENDPROC(xen_irq_disable_direct) + RELOC(xen_irq_disable_direct, 0) + +/* + (xen_)save_fl is used to get the current interrupt enable status. + Callers expect the status to be in X86_EFLAGS_IF, and other bits + may be set in the return value. We take advantage of this by + making sure that X86_EFLAGS_IF has the right value (and other bits + in that byte are 0), but other bits in the return value are + undefined. We need to toggle the state of the bit, because + Xen and x86 use opposite senses (mask vs enable). + */ +ENTRY(xen_save_fl_direct) + testb $0xff, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_mask + setz %ah + addb %ah,%ah +ENDPATCH(xen_save_fl_direct) + ret + ENDPROC(xen_save_fl_direct) + RELOC(xen_save_fl_direct, 0) + + +/* + In principle the caller should be passing us a value return + from xen_save_fl_direct, but for robustness sake we test only + the X86_EFLAGS_IF flag rather than the whole byte. After + setting the interrupt mask state, it checks for unmasked + pending events and enters the hypervisor to get them delivered + if so. + */ +ENTRY(xen_restore_fl_direct) +#ifdef CONFIG_X86_64 + testw $X86_EFLAGS_IF, %di +#else + testb $X86_EFLAGS_IF>>8, %ah +#endif + setz PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_mask + /* Preempt here doesn't matter because that will deal with + any pending interrupts. The pending check may end up being + run on the wrong CPU, but that doesn't hurt. */ + + /* check for unmasked and pending */ + cmpw $0x0001, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_pending + jz 1f +2: call check_events +1: +ENDPATCH(xen_restore_fl_direct) + ret + ENDPROC(xen_restore_fl_direct) + RELOC(xen_restore_fl_direct, 2b+1) + + +/* + Force an event check by making a hypercall, + but preserve regs before making the call. + */ +check_events: +#ifdef CONFIG_X86_32 + push %eax + push %ecx + push %edx + call xen_force_evtchn_callback + pop %edx + pop %ecx + pop %eax +#else + push %rax + push %rcx + push %rdx + push %rsi + push %rdi + push %r8 + push %r9 + push %r10 + push %r11 + call xen_force_evtchn_callback + pop %r11 + pop %r10 + pop %r9 + pop %r8 + pop %rdi + pop %rsi + pop %rdx + pop %rcx + pop %rax +#endif + ret + diff --git a/arch/x86/xen/xen-asm.h b/arch/x86/xen/xen-asm.h new file mode 100644 index 00000000000..465276467a4 --- /dev/null +++ b/arch/x86/xen/xen-asm.h @@ -0,0 +1,12 @@ +#ifndef _XEN_XEN_ASM_H +#define _XEN_XEN_ASM_H + +#include + +#define RELOC(x, v) .globl x##_reloc; x##_reloc=v +#define ENDPATCH(x) .globl x##_end; x##_end=. + +/* Pseudo-flag used for virtual NMI, which we don't implement yet */ +#define XEN_EFLAGS_NMI 0x80000000 + +#endif diff --git a/arch/x86/xen/xen-asm_32.S b/arch/x86/xen/xen-asm_32.S index 42786f59d9c..082d173caaf 100644 --- a/arch/x86/xen/xen-asm_32.S +++ b/arch/x86/xen/xen-asm_32.S @@ -11,101 +11,28 @@ generally too large to inline anyway. */ -#include - -#include +//#include #include -#include #include #include #include -#define RELOC(x, v) .globl x##_reloc; x##_reloc=v -#define ENDPATCH(x) .globl x##_end; x##_end=. - -/* Pseudo-flag used for virtual NMI, which we don't implement yet */ -#define XEN_EFLAGS_NMI 0x80000000 - -/* - Enable events. This clears the event mask and tests the pending - event status with one and operation. If there are pending - events, then enter the hypervisor to get them handled. - */ -ENTRY(xen_irq_enable_direct) - /* Unmask events */ - movb $0, PER_CPU_VAR(xen_vcpu_info)+XEN_vcpu_info_mask - - /* Preempt here doesn't matter because that will deal with - any pending interrupts. The pending check may end up being - run on the wrong CPU, but that doesn't hurt. */ - - /* Test for pending */ - testb $0xff, PER_CPU_VAR(xen_vcpu_info)+XEN_vcpu_info_pending - jz 1f - -2: call check_events -1: -ENDPATCH(xen_irq_enable_direct) - ret - ENDPROC(xen_irq_enable_direct) - RELOC(xen_irq_enable_direct, 2b+1) - +#include "xen-asm.h" /* - Disabling events is simply a matter of making the event mask - non-zero. - */ -ENTRY(xen_irq_disable_direct) - movb $1, PER_CPU_VAR(xen_vcpu_info)+XEN_vcpu_info_mask -ENDPATCH(xen_irq_disable_direct) - ret - ENDPROC(xen_irq_disable_direct) - RELOC(xen_irq_disable_direct, 0) - -/* - (xen_)save_fl is used to get the current interrupt enable status. - Callers expect the status to be in X86_EFLAGS_IF, and other bits - may be set in the return value. We take advantage of this by - making sure that X86_EFLAGS_IF has the right value (and other bits - in that byte are 0), but other bits in the return value are - undefined. We need to toggle the state of the bit, because - Xen and x86 use opposite senses (mask vs enable). - */ -ENTRY(xen_save_fl_direct) - testb $0xff, PER_CPU_VAR(xen_vcpu_info)+XEN_vcpu_info_mask - setz %ah - addb %ah,%ah -ENDPATCH(xen_save_fl_direct) - ret - ENDPROC(xen_save_fl_direct) - RELOC(xen_save_fl_direct, 0) - - -/* - In principle the caller should be passing us a value return - from xen_save_fl_direct, but for robustness sake we test only - the X86_EFLAGS_IF flag rather than the whole byte. After - setting the interrupt mask state, it checks for unmasked - pending events and enters the hypervisor to get them delivered - if so. + Force an event check by making a hypercall, + but preserve regs before making the call. */ -ENTRY(xen_restore_fl_direct) - testb $X86_EFLAGS_IF>>8, %ah - setz PER_CPU_VAR(xen_vcpu_info)+XEN_vcpu_info_mask - /* Preempt here doesn't matter because that will deal with - any pending interrupts. The pending check may end up being - run on the wrong CPU, but that doesn't hurt. */ - - /* check for unmasked and pending */ - cmpw $0x0001, PER_CPU_VAR(xen_vcpu_info)+XEN_vcpu_info_pending - jz 1f -2: call check_events -1: -ENDPATCH(xen_restore_fl_direct) +check_events: + push %eax + push %ecx + push %edx + call xen_force_evtchn_callback + pop %edx + pop %ecx + pop %eax ret - ENDPROC(xen_restore_fl_direct) - RELOC(xen_restore_fl_direct, 2b+1) /* We can't use sysexit directly, because we're not running in ring0. @@ -289,17 +216,3 @@ ENTRY(xen_iret_crit_fixup) lea 4(%edi),%esp /* point esp to new frame */ 2: jmp xen_do_upcall - -/* - Force an event check by making a hypercall, - but preserve regs before making the call. - */ -check_events: - push %eax - push %ecx - push %edx - call xen_force_evtchn_callback - pop %edx - pop %ecx - pop %eax - ret diff --git a/arch/x86/xen/xen-asm_64.S b/arch/x86/xen/xen-asm_64.S index d6fc51f4ce8..d205a283efe 100644 --- a/arch/x86/xen/xen-asm_64.S +++ b/arch/x86/xen/xen-asm_64.S @@ -11,142 +11,14 @@ generally too large to inline anyway. */ -#include - -#include -#include #include -#include #include +#include +#include #include -#define RELOC(x, v) .globl x##_reloc; x##_reloc=v -#define ENDPATCH(x) .globl x##_end; x##_end=. - -/* Pseudo-flag used for virtual NMI, which we don't implement yet */ -#define XEN_EFLAGS_NMI 0x80000000 - -#if 1 -/* - FIXME: x86_64 now can support direct access to percpu variables - via a segment override. Update xen accordingly. - */ -#define BUG ud2a -#endif - -/* - Enable events. This clears the event mask and tests the pending - event status with one and operation. If there are pending - events, then enter the hypervisor to get them handled. - */ -ENTRY(xen_irq_enable_direct) - BUG - - /* Unmask events */ - movb $0, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_mask - - /* Preempt here doesn't matter because that will deal with - any pending interrupts. The pending check may end up being - run on the wrong CPU, but that doesn't hurt. */ - - /* Test for pending */ - testb $0xff, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_pending - jz 1f - -2: call check_events -1: -ENDPATCH(xen_irq_enable_direct) - ret - ENDPROC(xen_irq_enable_direct) - RELOC(xen_irq_enable_direct, 2b+1) - -/* - Disabling events is simply a matter of making the event mask - non-zero. - */ -ENTRY(xen_irq_disable_direct) - BUG - - movb $1, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_mask -ENDPATCH(xen_irq_disable_direct) - ret - ENDPROC(xen_irq_disable_direct) - RELOC(xen_irq_disable_direct, 0) - -/* - (xen_)save_fl is used to get the current interrupt enable status. - Callers expect the status to be in X86_EFLAGS_IF, and other bits - may be set in the return value. We take advantage of this by - making sure that X86_EFLAGS_IF has the right value (and other bits - in that byte are 0), but other bits in the return value are - undefined. We need to toggle the state of the bit, because - Xen and x86 use opposite senses (mask vs enable). - */ -ENTRY(xen_save_fl_direct) - BUG - - testb $0xff, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_mask - setz %ah - addb %ah,%ah -ENDPATCH(xen_save_fl_direct) - ret - ENDPROC(xen_save_fl_direct) - RELOC(xen_save_fl_direct, 0) - -/* - In principle the caller should be passing us a value return - from xen_save_fl_direct, but for robustness sake we test only - the X86_EFLAGS_IF flag rather than the whole byte. After - setting the interrupt mask state, it checks for unmasked - pending events and enters the hypervisor to get them delivered - if so. - */ -ENTRY(xen_restore_fl_direct) - BUG - - testb $X86_EFLAGS_IF>>8, %ah - setz PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_mask - /* Preempt here doesn't matter because that will deal with - any pending interrupts. The pending check may end up being - run on the wrong CPU, but that doesn't hurt. */ - - /* check for unmasked and pending */ - cmpw $0x0001, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_pending - jz 1f -2: call check_events -1: -ENDPATCH(xen_restore_fl_direct) - ret - ENDPROC(xen_restore_fl_direct) - RELOC(xen_restore_fl_direct, 2b+1) - - -/* - Force an event check by making a hypercall, - but preserve regs before making the call. - */ -check_events: - push %rax - push %rcx - push %rdx - push %rsi - push %rdi - push %r8 - push %r9 - push %r10 - push %r11 - call xen_force_evtchn_callback - pop %r11 - pop %r10 - pop %r9 - pop %r8 - pop %rdi - pop %rsi - pop %rdx - pop %rcx - pop %rax - ret +#include "xen-asm.h" ENTRY(xen_adjust_exception_frame) mov 8+0(%rsp),%rcx -- cgit v1.2.3-70-g09d2 From e4d0407185cdbdcfd99fc23bde2e5454bbc46329 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 2 Feb 2009 13:55:54 -0800 Subject: xen: use direct ops on 64-bit Enable the use of the direct vcpu-access operations on 64-bit. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: H. Peter Anvin --- arch/x86/xen/enlighten.c | 14 +------------- 1 file changed, 1 insertion(+), 13 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index aed7ceeb4b6..37230342c2c 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -87,14 +87,7 @@ struct shared_info *HYPERVISOR_shared_info = (void *)&xen_dummy_shared_info; * * 0: not available, 1: available */ -static int have_vcpu_info_placement = -#ifdef CONFIG_X86_32 - 1 -#else - 0 -#endif - ; - +static int have_vcpu_info_placement = 1; static void xen_vcpu_setup(int cpu) { @@ -914,11 +907,6 @@ asmlinkage void __init xen_start_kernel(void) machine_ops = xen_machine_ops; -#ifdef CONFIG_X86_64 - /* Disable until direct per-cpu data access. */ - have_vcpu_info_placement = 0; -#endif - #ifdef CONFIG_X86_64 /* * Setup percpu state. We only need to do this for 64-bit -- cgit v1.2.3-70-g09d2 From 130ace11a9dc682541336d1fe5cb3bc7771a149e Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Fri, 6 Feb 2009 00:57:48 +0900 Subject: x86: style cleanups for xen assemblies Make the following style cleanups: * drop unnecessary //#include from xen-asm_32.S * compulsive adding of space after comma * reformat multiline comments Signed-off-by: Tejun Heo Signed-off-by: Ingo Molnar --- arch/x86/xen/xen-asm.S | 78 +++++++-------- arch/x86/xen/xen-asm_32.S | 238 ++++++++++++++++++++++++---------------------- arch/x86/xen/xen-asm_64.S | 107 +++++++++++---------- 3 files changed, 219 insertions(+), 204 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/xen-asm.S b/arch/x86/xen/xen-asm.S index 4c6f9679913..79d7362ad6d 100644 --- a/arch/x86/xen/xen-asm.S +++ b/arch/x86/xen/xen-asm.S @@ -1,14 +1,14 @@ /* - Asm versions of Xen pv-ops, suitable for either direct use or inlining. - The inline versions are the same as the direct-use versions, with the - pre- and post-amble chopped off. - - This code is encoded for size rather than absolute efficiency, - with a view to being able to inline as much as possible. - - We only bother with direct forms (ie, vcpu in percpu data) of - the operations here; the indirect forms are better handled in - C, since they're generally too large to inline anyway. + * Asm versions of Xen pv-ops, suitable for either direct use or + * inlining. The inline versions are the same as the direct-use + * versions, with the pre- and post-amble chopped off. + * + * This code is encoded for size rather than absolute efficiency, with + * a view to being able to inline as much as possible. + * + * We only bother with direct forms (ie, vcpu in percpu data) of the + * operations here; the indirect forms are better handled in C, since + * they're generally too large to inline anyway. */ #include @@ -18,17 +18,19 @@ #include "xen-asm.h" /* - Enable events. This clears the event mask and tests the pending - event status with one and operation. If there are pending - events, then enter the hypervisor to get them handled. + * Enable events. This clears the event mask and tests the pending + * event status with one and operation. If there are pending events, + * then enter the hypervisor to get them handled. */ ENTRY(xen_irq_enable_direct) /* Unmask events */ movb $0, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_mask - /* Preempt here doesn't matter because that will deal with - any pending interrupts. The pending check may end up being - run on the wrong CPU, but that doesn't hurt. */ + /* + * Preempt here doesn't matter because that will deal with any + * pending interrupts. The pending check may end up being run + * on the wrong CPU, but that doesn't hurt. + */ /* Test for pending */ testb $0xff, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_pending @@ -43,8 +45,8 @@ ENDPATCH(xen_irq_enable_direct) /* - Disabling events is simply a matter of making the event mask - non-zero. + * Disabling events is simply a matter of making the event mask + * non-zero. */ ENTRY(xen_irq_disable_direct) movb $1, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_mask @@ -54,18 +56,18 @@ ENDPATCH(xen_irq_disable_direct) RELOC(xen_irq_disable_direct, 0) /* - (xen_)save_fl is used to get the current interrupt enable status. - Callers expect the status to be in X86_EFLAGS_IF, and other bits - may be set in the return value. We take advantage of this by - making sure that X86_EFLAGS_IF has the right value (and other bits - in that byte are 0), but other bits in the return value are - undefined. We need to toggle the state of the bit, because - Xen and x86 use opposite senses (mask vs enable). + * (xen_)save_fl is used to get the current interrupt enable status. + * Callers expect the status to be in X86_EFLAGS_IF, and other bits + * may be set in the return value. We take advantage of this by + * making sure that X86_EFLAGS_IF has the right value (and other bits + * in that byte are 0), but other bits in the return value are + * undefined. We need to toggle the state of the bit, because Xen and + * x86 use opposite senses (mask vs enable). */ ENTRY(xen_save_fl_direct) testb $0xff, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_mask setz %ah - addb %ah,%ah + addb %ah, %ah ENDPATCH(xen_save_fl_direct) ret ENDPROC(xen_save_fl_direct) @@ -73,12 +75,11 @@ ENDPATCH(xen_save_fl_direct) /* - In principle the caller should be passing us a value return - from xen_save_fl_direct, but for robustness sake we test only - the X86_EFLAGS_IF flag rather than the whole byte. After - setting the interrupt mask state, it checks for unmasked - pending events and enters the hypervisor to get them delivered - if so. + * In principle the caller should be passing us a value return from + * xen_save_fl_direct, but for robustness sake we test only the + * X86_EFLAGS_IF flag rather than the whole byte. After setting the + * interrupt mask state, it checks for unmasked pending events and + * enters the hypervisor to get them delivered if so. */ ENTRY(xen_restore_fl_direct) #ifdef CONFIG_X86_64 @@ -87,9 +88,11 @@ ENTRY(xen_restore_fl_direct) testb $X86_EFLAGS_IF>>8, %ah #endif setz PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_mask - /* Preempt here doesn't matter because that will deal with - any pending interrupts. The pending check may end up being - run on the wrong CPU, but that doesn't hurt. */ + /* + * Preempt here doesn't matter because that will deal with any + * pending interrupts. The pending check may end up being run + * on the wrong CPU, but that doesn't hurt. + */ /* check for unmasked and pending */ cmpw $0x0001, PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_pending @@ -103,8 +106,8 @@ ENDPATCH(xen_restore_fl_direct) /* - Force an event check by making a hypercall, - but preserve regs before making the call. + * Force an event check by making a hypercall, but preserve regs + * before making the call. */ check_events: #ifdef CONFIG_X86_32 @@ -137,4 +140,3 @@ check_events: pop %rax #endif ret - diff --git a/arch/x86/xen/xen-asm_32.S b/arch/x86/xen/xen-asm_32.S index 082d173caaf..88e15deb8b8 100644 --- a/arch/x86/xen/xen-asm_32.S +++ b/arch/x86/xen/xen-asm_32.S @@ -1,17 +1,16 @@ /* - Asm versions of Xen pv-ops, suitable for either direct use or inlining. - The inline versions are the same as the direct-use versions, with the - pre- and post-amble chopped off. - - This code is encoded for size rather than absolute efficiency, - with a view to being able to inline as much as possible. - - We only bother with direct forms (ie, vcpu in pda) of the operations - here; the indirect forms are better handled in C, since they're - generally too large to inline anyway. + * Asm versions of Xen pv-ops, suitable for either direct use or + * inlining. The inline versions are the same as the direct-use + * versions, with the pre- and post-amble chopped off. + * + * This code is encoded for size rather than absolute efficiency, with + * a view to being able to inline as much as possible. + * + * We only bother with direct forms (ie, vcpu in pda) of the + * operations here; the indirect forms are better handled in C, since + * they're generally too large to inline anyway. */ -//#include #include #include #include @@ -21,8 +20,8 @@ #include "xen-asm.h" /* - Force an event check by making a hypercall, - but preserve regs before making the call. + * Force an event check by making a hypercall, but preserve regs + * before making the call. */ check_events: push %eax @@ -35,10 +34,10 @@ check_events: ret /* - We can't use sysexit directly, because we're not running in ring0. - But we can easily fake it up using iret. Assuming xen_sysexit - is jumped to with a standard stack frame, we can just strip it - back to a standard iret frame and use iret. + * We can't use sysexit directly, because we're not running in ring0. + * But we can easily fake it up using iret. Assuming xen_sysexit is + * jumped to with a standard stack frame, we can just strip it back to + * a standard iret frame and use iret. */ ENTRY(xen_sysexit) movl PT_EAX(%esp), %eax /* Shouldn't be necessary? */ @@ -49,33 +48,31 @@ ENTRY(xen_sysexit) ENDPROC(xen_sysexit) /* - This is run where a normal iret would be run, with the same stack setup: - 8: eflags - 4: cs - esp-> 0: eip - - This attempts to make sure that any pending events are dealt - with on return to usermode, but there is a small window in - which an event can happen just before entering usermode. If - the nested interrupt ends up setting one of the TIF_WORK_MASK - pending work flags, they will not be tested again before - returning to usermode. This means that a process can end up - with pending work, which will be unprocessed until the process - enters and leaves the kernel again, which could be an - unbounded amount of time. This means that a pending signal or - reschedule event could be indefinitely delayed. - - The fix is to notice a nested interrupt in the critical - window, and if one occurs, then fold the nested interrupt into - the current interrupt stack frame, and re-process it - iteratively rather than recursively. This means that it will - exit via the normal path, and all pending work will be dealt - with appropriately. - - Because the nested interrupt handler needs to deal with the - current stack state in whatever form its in, we keep things - simple by only using a single register which is pushed/popped - on the stack. + * This is run where a normal iret would be run, with the same stack setup: + * 8: eflags + * 4: cs + * esp-> 0: eip + * + * This attempts to make sure that any pending events are dealt with + * on return to usermode, but there is a small window in which an + * event can happen just before entering usermode. If the nested + * interrupt ends up setting one of the TIF_WORK_MASK pending work + * flags, they will not be tested again before returning to + * usermode. This means that a process can end up with pending work, + * which will be unprocessed until the process enters and leaves the + * kernel again, which could be an unbounded amount of time. This + * means that a pending signal or reschedule event could be + * indefinitely delayed. + * + * The fix is to notice a nested interrupt in the critical window, and + * if one occurs, then fold the nested interrupt into the current + * interrupt stack frame, and re-process it iteratively rather than + * recursively. This means that it will exit via the normal path, and + * all pending work will be dealt with appropriately. + * + * Because the nested interrupt handler needs to deal with the current + * stack state in whatever form its in, we keep things simple by only + * using a single register which is pushed/popped on the stack. */ ENTRY(xen_iret) /* test eflags for special cases */ @@ -85,13 +82,15 @@ ENTRY(xen_iret) push %eax ESP_OFFSET=4 # bytes pushed onto stack - /* Store vcpu_info pointer for easy access. Do it this - way to avoid having to reload %fs */ + /* + * Store vcpu_info pointer for easy access. Do it this way to + * avoid having to reload %fs + */ #ifdef CONFIG_SMP GET_THREAD_INFO(%eax) - movl TI_cpu(%eax),%eax - movl __per_cpu_offset(,%eax,4),%eax - mov per_cpu__xen_vcpu(%eax),%eax + movl TI_cpu(%eax), %eax + movl __per_cpu_offset(,%eax,4), %eax + mov per_cpu__xen_vcpu(%eax), %eax #else movl per_cpu__xen_vcpu, %eax #endif @@ -99,37 +98,46 @@ ENTRY(xen_iret) /* check IF state we're restoring */ testb $X86_EFLAGS_IF>>8, 8+1+ESP_OFFSET(%esp) - /* Maybe enable events. Once this happens we could get a - recursive event, so the critical region starts immediately - afterwards. However, if that happens we don't end up - resuming the code, so we don't have to be worried about - being preempted to another CPU. */ + /* + * Maybe enable events. Once this happens we could get a + * recursive event, so the critical region starts immediately + * afterwards. However, if that happens we don't end up + * resuming the code, so we don't have to be worried about + * being preempted to another CPU. + */ setz XEN_vcpu_info_mask(%eax) xen_iret_start_crit: /* check for unmasked and pending */ cmpw $0x0001, XEN_vcpu_info_pending(%eax) - /* If there's something pending, mask events again so we - can jump back into xen_hypervisor_callback */ + /* + * If there's something pending, mask events again so we can + * jump back into xen_hypervisor_callback + */ sete XEN_vcpu_info_mask(%eax) popl %eax - /* From this point on the registers are restored and the stack - updated, so we don't need to worry about it if we're preempted */ + /* + * From this point on the registers are restored and the stack + * updated, so we don't need to worry about it if we're + * preempted + */ iret_restore_end: - /* Jump to hypervisor_callback after fixing up the stack. - Events are masked, so jumping out of the critical - region is OK. */ + /* + * Jump to hypervisor_callback after fixing up the stack. + * Events are masked, so jumping out of the critical region is + * OK. + */ je xen_hypervisor_callback 1: iret xen_iret_end_crit: -.section __ex_table,"a" +.section __ex_table, "a" .align 4 - .long 1b,iret_exc + .long 1b, iret_exc .previous hyper_iret: @@ -139,55 +147,55 @@ hyper_iret: .globl xen_iret_start_crit, xen_iret_end_crit /* - This is called by xen_hypervisor_callback in entry.S when it sees - that the EIP at the time of interrupt was between xen_iret_start_crit - and xen_iret_end_crit. We're passed the EIP in %eax so we can do - a more refined determination of what to do. - - The stack format at this point is: - ---------------- - ss : (ss/esp may be present if we came from usermode) - esp : - eflags } outer exception info - cs } - eip } - ---------------- <- edi (copy dest) - eax : outer eax if it hasn't been restored - ---------------- - eflags } nested exception info - cs } (no ss/esp because we're nested - eip } from the same ring) - orig_eax }<- esi (copy src) - - - - - - - - - - fs } - es } - ds } SAVE_ALL state - eax } - : : - ebx }<- esp - ---------------- - - In order to deliver the nested exception properly, we need to shift - everything from the return addr up to the error code so it - sits just under the outer exception info. This means that when we - handle the exception, we do it in the context of the outer exception - rather than starting a new one. - - The only caveat is that if the outer eax hasn't been - restored yet (ie, it's still on stack), we need to insert - its value into the SAVE_ALL state before going on, since - it's usermode state which we eventually need to restore. + * This is called by xen_hypervisor_callback in entry.S when it sees + * that the EIP at the time of interrupt was between + * xen_iret_start_crit and xen_iret_end_crit. We're passed the EIP in + * %eax so we can do a more refined determination of what to do. + * + * The stack format at this point is: + * ---------------- + * ss : (ss/esp may be present if we came from usermode) + * esp : + * eflags } outer exception info + * cs } + * eip } + * ---------------- <- edi (copy dest) + * eax : outer eax if it hasn't been restored + * ---------------- + * eflags } nested exception info + * cs } (no ss/esp because we're nested + * eip } from the same ring) + * orig_eax }<- esi (copy src) + * - - - - - - - - + * fs } + * es } + * ds } SAVE_ALL state + * eax } + * : : + * ebx }<- esp + * ---------------- + * + * In order to deliver the nested exception properly, we need to shift + * everything from the return addr up to the error code so it sits + * just under the outer exception info. This means that when we + * handle the exception, we do it in the context of the outer + * exception rather than starting a new one. + * + * The only caveat is that if the outer eax hasn't been restored yet + * (ie, it's still on stack), we need to insert its value into the + * SAVE_ALL state before going on, since it's usermode state which we + * eventually need to restore. */ ENTRY(xen_iret_crit_fixup) /* - Paranoia: Make sure we're really coming from kernel space. - One could imagine a case where userspace jumps into the - critical range address, but just before the CPU delivers a GP, - it decides to deliver an interrupt instead. Unlikely? - Definitely. Easy to avoid? Yes. The Intel documents - explicitly say that the reported EIP for a bad jump is the - jump instruction itself, not the destination, but some virtual - environments get this wrong. + * Paranoia: Make sure we're really coming from kernel space. + * One could imagine a case where userspace jumps into the + * critical range address, but just before the CPU delivers a + * GP, it decides to deliver an interrupt instead. Unlikely? + * Definitely. Easy to avoid? Yes. The Intel documents + * explicitly say that the reported EIP for a bad jump is the + * jump instruction itself, not the destination, but some + * virtual environments get this wrong. */ movl PT_CS(%esp), %ecx andl $SEGMENT_RPL_MASK, %ecx @@ -197,15 +205,17 @@ ENTRY(xen_iret_crit_fixup) lea PT_ORIG_EAX(%esp), %esi lea PT_EFLAGS(%esp), %edi - /* If eip is before iret_restore_end then stack - hasn't been restored yet. */ + /* + * If eip is before iret_restore_end then stack + * hasn't been restored yet. + */ cmp $iret_restore_end, %eax jae 1f - movl 0+4(%edi),%eax /* copy EAX (just above top of frame) */ + movl 0+4(%edi), %eax /* copy EAX (just above top of frame) */ movl %eax, PT_EAX(%esp) - lea ESP_OFFSET(%edi),%edi /* move dest up over saved regs */ + lea ESP_OFFSET(%edi), %edi /* move dest up over saved regs */ /* set up the copy */ 1: std @@ -213,6 +223,6 @@ ENTRY(xen_iret_crit_fixup) rep movsl cld - lea 4(%edi),%esp /* point esp to new frame */ + lea 4(%edi), %esp /* point esp to new frame */ 2: jmp xen_do_upcall diff --git a/arch/x86/xen/xen-asm_64.S b/arch/x86/xen/xen-asm_64.S index d205a283efe..02f496a8dba 100644 --- a/arch/x86/xen/xen-asm_64.S +++ b/arch/x86/xen/xen-asm_64.S @@ -1,14 +1,14 @@ /* - Asm versions of Xen pv-ops, suitable for either direct use or inlining. - The inline versions are the same as the direct-use versions, with the - pre- and post-amble chopped off. - - This code is encoded for size rather than absolute efficiency, - with a view to being able to inline as much as possible. - - We only bother with direct forms (ie, vcpu in pda) of the operations - here; the indirect forms are better handled in C, since they're - generally too large to inline anyway. + * Asm versions of Xen pv-ops, suitable for either direct use or + * inlining. The inline versions are the same as the direct-use + * versions, with the pre- and post-amble chopped off. + * + * This code is encoded for size rather than absolute efficiency, with + * a view to being able to inline as much as possible. + * + * We only bother with direct forms (ie, vcpu in pda) of the + * operations here; the indirect forms are better handled in C, since + * they're generally too large to inline anyway. */ #include @@ -21,25 +21,25 @@ #include "xen-asm.h" ENTRY(xen_adjust_exception_frame) - mov 8+0(%rsp),%rcx - mov 8+8(%rsp),%r11 + mov 8+0(%rsp), %rcx + mov 8+8(%rsp), %r11 ret $16 hypercall_iret = hypercall_page + __HYPERVISOR_iret * 32 /* - Xen64 iret frame: - - ss - rsp - rflags - cs - rip <-- standard iret frame - - flags - - rcx } - r11 }<-- pushed by hypercall page -rsp -> rax } + * Xen64 iret frame: + * + * ss + * rsp + * rflags + * cs + * rip <-- standard iret frame + * + * flags + * + * rcx } + * r11 }<-- pushed by hypercall page + * rsp->rax } */ ENTRY(xen_iret) pushq $0 @@ -48,8 +48,8 @@ ENDPATCH(xen_iret) RELOC(xen_iret, 1b+1) /* - sysexit is not used for 64-bit processes, so it's - only ever used to return to 32-bit compat userspace. + * sysexit is not used for 64-bit processes, so it's only ever used to + * return to 32-bit compat userspace. */ ENTRY(xen_sysexit) pushq $__USER32_DS @@ -64,10 +64,12 @@ ENDPATCH(xen_sysexit) RELOC(xen_sysexit, 1b+1) ENTRY(xen_sysret64) - /* We're already on the usermode stack at this point, but still - with the kernel gs, so we can easily switch back */ + /* + * We're already on the usermode stack at this point, but + * still with the kernel gs, so we can easily switch back + */ movq %rsp, PER_CPU_VAR(old_rsp) - movq PER_CPU_VAR(kernel_stack),%rsp + movq PER_CPU_VAR(kernel_stack), %rsp pushq $__USER_DS pushq PER_CPU_VAR(old_rsp) @@ -81,8 +83,10 @@ ENDPATCH(xen_sysret64) RELOC(xen_sysret64, 1b+1) ENTRY(xen_sysret32) - /* We're already on the usermode stack at this point, but still - with the kernel gs, so we can easily switch back */ + /* + * We're already on the usermode stack at this point, but + * still with the kernel gs, so we can easily switch back + */ movq %rsp, PER_CPU_VAR(old_rsp) movq PER_CPU_VAR(kernel_stack), %rsp @@ -98,28 +102,27 @@ ENDPATCH(xen_sysret32) RELOC(xen_sysret32, 1b+1) /* - Xen handles syscall callbacks much like ordinary exceptions, - which means we have: - - kernel gs - - kernel rsp - - an iret-like stack frame on the stack (including rcx and r11): - ss - rsp - rflags - cs - rip - r11 - rsp-> rcx - - In all the entrypoints, we undo all that to make it look - like a CPU-generated syscall/sysenter and jump to the normal - entrypoint. + * Xen handles syscall callbacks much like ordinary exceptions, which + * means we have: + * - kernel gs + * - kernel rsp + * - an iret-like stack frame on the stack (including rcx and r11): + * ss + * rsp + * rflags + * cs + * rip + * r11 + * rsp->rcx + * + * In all the entrypoints, we undo all that to make it look like a + * CPU-generated syscall/sysenter and jump to the normal entrypoint. */ .macro undo_xen_syscall - mov 0*8(%rsp),%rcx - mov 1*8(%rsp),%r11 - mov 5*8(%rsp),%rsp + mov 0*8(%rsp), %rcx + mov 1*8(%rsp), %r11 + mov 5*8(%rsp), %rsp .endm /* Normal 64-bit system call target */ @@ -146,7 +149,7 @@ ENDPROC(xen_sysenter_target) ENTRY(xen_syscall32_target) ENTRY(xen_sysenter_target) - lea 16(%rsp), %rsp /* strip %rcx,%r11 */ + lea 16(%rsp), %rsp /* strip %rcx, %r11 */ mov $-ENOSYS, %rax pushq $VGCF_in_syscall jmp hypercall_iret -- cgit v1.2.3-70-g09d2 From 792dc4f6cdacf50d3f2b93756d282fc04ee34bd5 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Fri, 6 Feb 2009 14:09:43 -0800 Subject: xen: use our own eventchannel->irq path Rather than overloading vectors for event channels, take full responsibility for mapping an event channel to irq directly. With this patch Xen has its own irq allocator. When the kernel gets an event channel upcall, it maps the event channel number to an irq and injects it into the normal interrupt path. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Ingo Molnar --- arch/x86/include/asm/xen/events.h | 6 ------ arch/x86/xen/irq.c | 17 +---------------- drivers/xen/events.c | 22 ++++++++++++++++++++-- 3 files changed, 21 insertions(+), 24 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/include/asm/xen/events.h b/arch/x86/include/asm/xen/events.h index 19144184983..1df35417c41 100644 --- a/arch/x86/include/asm/xen/events.h +++ b/arch/x86/include/asm/xen/events.h @@ -15,10 +15,4 @@ static inline int xen_irqs_disabled(struct pt_regs *regs) return raw_irqs_disabled_flags(regs->flags); } -static inline void xen_do_IRQ(int irq, struct pt_regs *regs) -{ - regs->orig_ax = ~irq; - do_IRQ(regs); -} - #endif /* _ASM_X86_XEN_EVENTS_H */ diff --git a/arch/x86/xen/irq.c b/arch/x86/xen/irq.c index 5a070900ad3..cfd17799bd6 100644 --- a/arch/x86/xen/irq.c +++ b/arch/x86/xen/irq.c @@ -19,21 +19,6 @@ void xen_force_evtchn_callback(void) (void)HYPERVISOR_xen_version(0, NULL); } -static void __init __xen_init_IRQ(void) -{ - int i; - - /* Create identity vector->irq map */ - for(i = 0; i < NR_VECTORS; i++) { - int cpu; - - for_each_possible_cpu(cpu) - per_cpu(vector_irq, cpu)[i] = i; - } - - xen_init_IRQ(); -} - static unsigned long xen_save_fl(void) { struct vcpu_info *vcpu; @@ -127,7 +112,7 @@ static void xen_halt(void) } static const struct pv_irq_ops xen_irq_ops __initdata = { - .init_IRQ = __xen_init_IRQ, + .init_IRQ = xen_init_IRQ, .save_fl = PV_CALLEE_SAVE(xen_save_fl), .restore_fl = PV_CALLEE_SAVE(xen_restore_fl), diff --git a/drivers/xen/events.c b/drivers/xen/events.c index 7c3705479ea..2c8d710713f 100644 --- a/drivers/xen/events.c +++ b/drivers/xen/events.c @@ -30,6 +30,7 @@ #include #include +#include #include #include #include @@ -517,6 +518,24 @@ irqreturn_t xen_debug_interrupt(int irq, void *dev_id) } +static void xen_do_irq(unsigned irq, struct pt_regs *regs) +{ + struct pt_regs *old_regs = set_irq_regs(regs); + + if (WARN_ON(irq == -1)) + return; + + exit_idle(); + irq_enter(); + + //printk("cpu %d handling irq %d\n", smp_processor_id(), info->irq); + handle_irq(irq, regs); + + irq_exit(); + + set_irq_regs(old_regs); +} + /* * Search the CPUs pending events bitmasks. For each one found, map * the event number to an irq, and feed it into do_IRQ() for @@ -557,8 +576,7 @@ void xen_evtchn_do_upcall(struct pt_regs *regs) int port = (word_idx * BITS_PER_LONG) + bit_idx; int irq = evtchn_to_irq[port]; - if (irq != -1) - xen_do_IRQ(irq, regs); + xen_do_irq(irq, regs); } } -- cgit v1.2.3-70-g09d2 From ccbeed3a05908d201b47b6c3dd1a373138bba566 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 9 Feb 2009 22:17:40 +0900 Subject: x86: make lazy %gs optional on x86_32 Impact: pt_regs changed, lazy gs handling made optional, add slight overhead to SAVE_ALL, simplifies error_code path a bit On x86_32, %gs hasn't been used by kernel and handled lazily. pt_regs doesn't have place for it and gs is saved/loaded only when necessary. In preparation for stack protector support, this patch makes lazy %gs handling optional by doing the followings. * Add CONFIG_X86_32_LAZY_GS and place for gs in pt_regs. * Save and restore %gs along with other registers in entry_32.S unless LAZY_GS. Note that this unfortunately adds "pushl $0" on SAVE_ALL even when LAZY_GS. However, it adds no overhead to common exit path and simplifies entry path with error code. * Define different user_gs accessors depending on LAZY_GS and add lazy_save_gs() and lazy_load_gs() which are noop if !LAZY_GS. The lazy_*_gs() ops are used to save, load and clear %gs lazily. * Define ELF_CORE_COPY_KERNEL_REGS() which always read %gs directly. xen and lguest changes need to be verified. Signed-off-by: Tejun Heo Cc: Jeremy Fitzhardinge Cc: Rusty Russell Signed-off-by: Ingo Molnar --- arch/x86/Kconfig | 4 ++ arch/x86/include/asm/elf.h | 15 ++++- arch/x86/include/asm/mmu_context.h | 2 +- arch/x86/include/asm/ptrace.h | 4 +- arch/x86/include/asm/system.h | 12 +++- arch/x86/kernel/asm-offsets_32.c | 1 + arch/x86/kernel/entry_32.S | 132 +++++++++++++++++++++++++++++++------ arch/x86/kernel/process_32.c | 4 +- arch/x86/kernel/ptrace.c | 5 +- arch/x86/lguest/boot.c | 2 +- arch/x86/xen/enlighten.c | 17 ++--- 11 files changed, 158 insertions(+), 40 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig index 5c8e353c112..5bcdede71ba 100644 --- a/arch/x86/Kconfig +++ b/arch/x86/Kconfig @@ -207,6 +207,10 @@ config X86_TRAMPOLINE depends on X86_SMP || (X86_VOYAGER && SMP) || (64BIT && ACPI_SLEEP) default y +config X86_32_LAZY_GS + def_bool y + depends on X86_32 + config KTIME_SCALAR def_bool X86_32 source "init/Kconfig" diff --git a/arch/x86/include/asm/elf.h b/arch/x86/include/asm/elf.h index 39b0aac1675..83c1bc8d2e8 100644 --- a/arch/x86/include/asm/elf.h +++ b/arch/x86/include/asm/elf.h @@ -112,7 +112,7 @@ extern unsigned int vdso_enabled; * now struct_user_regs, they are different) */ -#define ELF_CORE_COPY_REGS(pr_reg, regs) \ +#define ELF_CORE_COPY_REGS_COMMON(pr_reg, regs) \ do { \ pr_reg[0] = regs->bx; \ pr_reg[1] = regs->cx; \ @@ -124,7 +124,6 @@ do { \ pr_reg[7] = regs->ds & 0xffff; \ pr_reg[8] = regs->es & 0xffff; \ pr_reg[9] = regs->fs & 0xffff; \ - pr_reg[10] = get_user_gs(regs); \ pr_reg[11] = regs->orig_ax; \ pr_reg[12] = regs->ip; \ pr_reg[13] = regs->cs & 0xffff; \ @@ -133,6 +132,18 @@ do { \ pr_reg[16] = regs->ss & 0xffff; \ } while (0); +#define ELF_CORE_COPY_REGS(pr_reg, regs) \ +do { \ + ELF_CORE_COPY_REGS_COMMON(pr_reg, regs);\ + pr_reg[10] = get_user_gs(regs); \ +} while (0); + +#define ELF_CORE_COPY_KERNEL_REGS(pr_reg, regs) \ +do { \ + ELF_CORE_COPY_REGS_COMMON(pr_reg, regs);\ + savesegment(gs, pr_reg[10]); \ +} while (0); + #define ELF_PLATFORM (utsname()->machine) #define set_personality_64bit() do { } while (0) diff --git a/arch/x86/include/asm/mmu_context.h b/arch/x86/include/asm/mmu_context.h index 4955165682c..f923203dc39 100644 --- a/arch/x86/include/asm/mmu_context.h +++ b/arch/x86/include/asm/mmu_context.h @@ -79,7 +79,7 @@ do { \ #ifdef CONFIG_X86_32 #define deactivate_mm(tsk, mm) \ do { \ - set_user_gs(task_pt_regs(tsk), 0); \ + lazy_load_gs(0); \ } while (0) #else #define deactivate_mm(tsk, mm) \ diff --git a/arch/x86/include/asm/ptrace.h b/arch/x86/include/asm/ptrace.h index 6d34d954c22..e304b66abee 100644 --- a/arch/x86/include/asm/ptrace.h +++ b/arch/x86/include/asm/ptrace.h @@ -28,7 +28,7 @@ struct pt_regs { int xds; int xes; int xfs; - /* int gs; */ + int xgs; long orig_eax; long eip; int xcs; @@ -50,7 +50,7 @@ struct pt_regs { unsigned long ds; unsigned long es; unsigned long fs; - /* int gs; */ + unsigned long gs; unsigned long orig_ax; unsigned long ip; unsigned long cs; diff --git a/arch/x86/include/asm/system.h b/arch/x86/include/asm/system.h index 70c74b8db87..79b98e5b96f 100644 --- a/arch/x86/include/asm/system.h +++ b/arch/x86/include/asm/system.h @@ -186,10 +186,20 @@ extern void native_load_gs_index(unsigned); * x86_32 user gs accessors. */ #ifdef CONFIG_X86_32 +#ifdef CONFIG_X86_32_LAZY_GS #define get_user_gs(regs) (u16)({unsigned long v; savesegment(gs, v); v;}) #define set_user_gs(regs, v) loadsegment(gs, (unsigned long)(v)) #define task_user_gs(tsk) ((tsk)->thread.gs) -#endif +#define lazy_save_gs(v) savesegment(gs, (v)) +#define lazy_load_gs(v) loadsegment(gs, (v)) +#else /* X86_32_LAZY_GS */ +#define get_user_gs(regs) (u16)((regs)->gs) +#define set_user_gs(regs, v) do { (regs)->gs = (v); } while (0) +#define task_user_gs(tsk) (task_pt_regs(tsk)->gs) +#define lazy_save_gs(v) do { } while (0) +#define lazy_load_gs(v) do { } while (0) +#endif /* X86_32_LAZY_GS */ +#endif /* X86_32 */ static inline unsigned long get_limit(unsigned long segment) { diff --git a/arch/x86/kernel/asm-offsets_32.c b/arch/x86/kernel/asm-offsets_32.c index ee4df08feee..fbf2f33e308 100644 --- a/arch/x86/kernel/asm-offsets_32.c +++ b/arch/x86/kernel/asm-offsets_32.c @@ -75,6 +75,7 @@ void foo(void) OFFSET(PT_DS, pt_regs, ds); OFFSET(PT_ES, pt_regs, es); OFFSET(PT_FS, pt_regs, fs); + OFFSET(PT_GS, pt_regs, gs); OFFSET(PT_ORIG_EAX, pt_regs, orig_ax); OFFSET(PT_EIP, pt_regs, ip); OFFSET(PT_CS, pt_regs, cs); diff --git a/arch/x86/kernel/entry_32.S b/arch/x86/kernel/entry_32.S index c461925d3b6..82e6868bee4 100644 --- a/arch/x86/kernel/entry_32.S +++ b/arch/x86/kernel/entry_32.S @@ -30,12 +30,13 @@ * 1C(%esp) - %ds * 20(%esp) - %es * 24(%esp) - %fs - * 28(%esp) - orig_eax - * 2C(%esp) - %eip - * 30(%esp) - %cs - * 34(%esp) - %eflags - * 38(%esp) - %oldesp - * 3C(%esp) - %oldss + * 28(%esp) - %gs saved iff !CONFIG_X86_32_LAZY_GS + * 2C(%esp) - orig_eax + * 30(%esp) - %eip + * 34(%esp) - %cs + * 38(%esp) - %eflags + * 3C(%esp) - %oldesp + * 40(%esp) - %oldss * * "current" is in register %ebx during any slow entries. */ @@ -101,8 +102,99 @@ #define resume_userspace_sig resume_userspace #endif +/* + * User gs save/restore + * + * %gs is used for userland TLS and kernel only uses it for stack + * canary which is required to be at %gs:20 by gcc. Read the comment + * at the top of stackprotector.h for more info. + * + * Local labels 98 and 99 are used. + */ +#ifdef CONFIG_X86_32_LAZY_GS + + /* unfortunately push/pop can't be no-op */ +.macro PUSH_GS + pushl $0 + CFI_ADJUST_CFA_OFFSET 4 +.endm +.macro POP_GS pop=0 + addl $(4 + \pop), %esp + CFI_ADJUST_CFA_OFFSET -(4 + \pop) +.endm +.macro POP_GS_EX +.endm + + /* all the rest are no-op */ +.macro PTGS_TO_GS +.endm +.macro PTGS_TO_GS_EX +.endm +.macro GS_TO_REG reg +.endm +.macro REG_TO_PTGS reg +.endm +.macro SET_KERNEL_GS reg +.endm + +#else /* CONFIG_X86_32_LAZY_GS */ + +.macro PUSH_GS + pushl %gs + CFI_ADJUST_CFA_OFFSET 4 + /*CFI_REL_OFFSET gs, 0*/ +.endm + +.macro POP_GS pop=0 +98: popl %gs + CFI_ADJUST_CFA_OFFSET -4 + /*CFI_RESTORE gs*/ + .if \pop <> 0 + add $\pop, %esp + CFI_ADJUST_CFA_OFFSET -\pop + .endif +.endm +.macro POP_GS_EX +.pushsection .fixup, "ax" +99: movl $0, (%esp) + jmp 98b +.section __ex_table, "a" + .align 4 + .long 98b, 99b +.popsection +.endm + +.macro PTGS_TO_GS +98: mov PT_GS(%esp), %gs +.endm +.macro PTGS_TO_GS_EX +.pushsection .fixup, "ax" +99: movl $0, PT_GS(%esp) + jmp 98b +.section __ex_table, "a" + .align 4 + .long 98b, 99b +.popsection +.endm + +.macro GS_TO_REG reg + movl %gs, \reg + /*CFI_REGISTER gs, \reg*/ +.endm +.macro REG_TO_PTGS reg + movl \reg, PT_GS(%esp) + /*CFI_REL_OFFSET gs, PT_GS*/ +.endm +.macro SET_KERNEL_GS reg + xorl \reg, \reg + movl \reg, %gs +.endm + +#endif /* CONFIG_X86_32_LAZY_GS */ + .macro SAVE_ALL cld + PUSH_GS pushl %fs CFI_ADJUST_CFA_OFFSET 4 /*CFI_REL_OFFSET fs, 0;*/ @@ -138,6 +230,7 @@ movl %edx, %es movl $(__KERNEL_PERCPU), %edx movl %edx, %fs + SET_KERNEL_GS %edx .endm .macro RESTORE_INT_REGS @@ -164,7 +257,7 @@ CFI_RESTORE eax .endm -.macro RESTORE_REGS +.macro RESTORE_REGS pop=0 RESTORE_INT_REGS 1: popl %ds CFI_ADJUST_CFA_OFFSET -4 @@ -175,6 +268,7 @@ 3: popl %fs CFI_ADJUST_CFA_OFFSET -4 /*CFI_RESTORE fs;*/ + POP_GS \pop .pushsection .fixup, "ax" 4: movl $0, (%esp) jmp 1b @@ -188,6 +282,7 @@ .long 2b, 5b .long 3b, 6b .popsection + POP_GS_EX .endm .macro RING0_INT_FRAME @@ -368,6 +463,7 @@ sysenter_exit: xorl %ebp,%ebp TRACE_IRQS_ON 1: mov PT_FS(%esp), %fs + PTGS_TO_GS ENABLE_INTERRUPTS_SYSEXIT #ifdef CONFIG_AUDITSYSCALL @@ -416,6 +512,7 @@ sysexit_audit: .align 4 .long 1b,2b .popsection + PTGS_TO_GS_EX ENDPROC(ia32_sysenter_target) # system call handler stub @@ -458,8 +555,7 @@ restore_all: restore_nocheck: TRACE_IRQS_IRET restore_nocheck_notrace: - RESTORE_REGS - addl $4, %esp # skip orig_eax/error_code + RESTORE_REGS 4 # skip orig_eax/error_code CFI_ADJUST_CFA_OFFSET -4 irq_return: INTERRUPT_RETURN @@ -1078,7 +1174,10 @@ ENTRY(page_fault) CFI_ADJUST_CFA_OFFSET 4 ALIGN error_code: - /* the function address is in %fs's slot on the stack */ + /* the function address is in %gs's slot on the stack */ + pushl %fs + CFI_ADJUST_CFA_OFFSET 4 + /*CFI_REL_OFFSET fs, 0*/ pushl %es CFI_ADJUST_CFA_OFFSET 4 /*CFI_REL_OFFSET es, 0*/ @@ -1107,20 +1206,15 @@ error_code: CFI_ADJUST_CFA_OFFSET 4 CFI_REL_OFFSET ebx, 0 cld - pushl %fs - CFI_ADJUST_CFA_OFFSET 4 - /*CFI_REL_OFFSET fs, 0*/ movl $(__KERNEL_PERCPU), %ecx movl %ecx, %fs UNWIND_ESPFIX_STACK - popl %ecx - CFI_ADJUST_CFA_OFFSET -4 - /*CFI_REGISTER es, ecx*/ - movl PT_FS(%esp), %edi # get the function address + GS_TO_REG %ecx + movl PT_GS(%esp), %edi # get the function address movl PT_ORIG_EAX(%esp), %edx # get the error code movl $-1, PT_ORIG_EAX(%esp) # no syscall to restart - mov %ecx, PT_FS(%esp) - /*CFI_REL_OFFSET fs, ES*/ + REG_TO_PTGS %ecx + SET_KERNEL_GS %ecx movl $(__USER_DS), %ecx movl %ecx, %ds movl %ecx, %es diff --git a/arch/x86/kernel/process_32.c b/arch/x86/kernel/process_32.c index d58a340e1be..86122fa2a1b 100644 --- a/arch/x86/kernel/process_32.c +++ b/arch/x86/kernel/process_32.c @@ -539,7 +539,7 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p) * used %fs or %gs (it does not today), or if the kernel is * running inside of a hypervisor layer. */ - savesegment(gs, prev->gs); + lazy_save_gs(prev->gs); /* * Load the per-thread Thread-Local Storage descriptor. @@ -585,7 +585,7 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p) * Restore %gs if needed (which is common) */ if (prev->gs | next->gs) - loadsegment(gs, next->gs); + lazy_load_gs(next->gs); percpu_write(current_task, next_p); diff --git a/arch/x86/kernel/ptrace.c b/arch/x86/kernel/ptrace.c index 508b6b57d0c..7ec39ab37a2 100644 --- a/arch/x86/kernel/ptrace.c +++ b/arch/x86/kernel/ptrace.c @@ -75,10 +75,7 @@ static inline bool invalid_selector(u16 value) static unsigned long *pt_regs_access(struct pt_regs *regs, unsigned long regno) { BUILD_BUG_ON(offsetof(struct pt_regs, bx) != 0); - regno >>= 2; - if (regno > FS) - --regno; - return ®s->bx + regno; + return ®s->bx + (regno >> 2); } static u16 get_segment_reg(struct task_struct *task, unsigned long offset) diff --git a/arch/x86/lguest/boot.c b/arch/x86/lguest/boot.c index 19e33b6cd59..da2e314f61b 100644 --- a/arch/x86/lguest/boot.c +++ b/arch/x86/lguest/boot.c @@ -283,7 +283,7 @@ static void lguest_load_tls(struct thread_struct *t, unsigned int cpu) /* There's one problem which normal hardware doesn't have: the Host * can't handle us removing entries we're currently using. So we clear * the GS register here: if it's needed it'll be reloaded anyway. */ - loadsegment(gs, 0); + lazy_load_gs(0); lazy_hcall(LHCALL_LOAD_TLS, __pa(&t->tls_array), cpu, 0); } diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 37230342c2c..95ff6a0e942 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -323,13 +323,14 @@ static void load_TLS_descriptor(struct thread_struct *t, static void xen_load_tls(struct thread_struct *t, unsigned int cpu) { /* - * XXX sleazy hack: If we're being called in a lazy-cpu zone, - * it means we're in a context switch, and %gs has just been - * saved. This means we can zero it out to prevent faults on - * exit from the hypervisor if the next process has no %gs. - * Either way, it has been saved, and the new value will get - * loaded properly. This will go away as soon as Xen has been - * modified to not save/restore %gs for normal hypercalls. + * XXX sleazy hack: If we're being called in a lazy-cpu zone + * and lazy gs handling is enabled, it means we're in a + * context switch, and %gs has just been saved. This means we + * can zero it out to prevent faults on exit from the + * hypervisor if the next process has no %gs. Either way, it + * has been saved, and the new value will get loaded properly. + * This will go away as soon as Xen has been modified to not + * save/restore %gs for normal hypercalls. * * On x86_64, this hack is not used for %gs, because gs points * to KERNEL_GS_BASE (and uses it for PDA references), so we @@ -341,7 +342,7 @@ static void xen_load_tls(struct thread_struct *t, unsigned int cpu) */ if (paravirt_get_lazy_mode() == PARAVIRT_LAZY_CPU) { #ifdef CONFIG_X86_32 - loadsegment(gs, 0); + lazy_load_gs(0); #else loadsegment(fs, 0); #endif -- cgit v1.2.3-70-g09d2 From 694aa960608d2976666d850bd4ef78053bbd0c84 Mon Sep 17 00:00:00 2001 From: Ian Campbell Date: Thu, 12 Feb 2009 16:25:42 -0800 Subject: xen: fix xen_flush_tlb_others The commit commit 4595f9620cda8a1e973588e743cf5f8436dd20c6 Author: Rusty Russell Date: Sat Jan 10 21:58:09 2009 -0800 x86: change flush_tlb_others to take a const struct cpumask causes xen_flush_tlb_others to allocate a multicall and then issue it without initializing it in the case where the cpumask is empty, leading to: [ 8.354898] 1 multicall(s) failed: cpu 1 [ 8.354921] Pid: 2213, comm: bootclean Not tainted 2.6.29-rc3-x86_32p-xenU-tip #135 [ 8.354937] Call Trace: [ 8.354955] [] xen_mc_flush+0x133/0x1b0 [ 8.354971] [] ? xen_force_evtchn_callback+0x1a/0x30 [ 8.354988] [] xen_flush_tlb_others+0xb0/0xd0 [ 8.355003] [] flush_tlb_page+0x53/0xa0 [ 8.355018] [] do_wp_page+0x2a0/0x7c0 [ 8.355034] [] ? notify_remote_via_irq+0x3a/0x70 [ 8.355049] [] handle_mm_fault+0x7b0/0xa50 [ 8.355065] [] ? wake_up_new_task+0x8e/0xb0 [ 8.355079] [] ? do_fork+0xe5/0x320 [ 8.355095] [] do_page_fault+0xe9/0x240 [ 8.355109] [] ? do_page_fault+0x0/0x240 [ 8.355125] [] error_code+0x72/0x78 [ 8.355139] call 1/1: op=2863311530 arg=[aaaaaaaa] result=-38 xen_flush_tlb_others+0x41/0xd0 Since empty cpumasks are rare and undoing an xen_mc_entry() is tricky just issue such requests normally. Signed-off-by: Ian Campbell Signed-off-by: Ingo Molnar --- arch/x86/xen/mmu.c | 3 --- 1 file changed, 3 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index d2e8ed1aff3..319bd40a57c 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -1273,8 +1273,6 @@ static void xen_flush_tlb_others(const struct cpumask *cpus, /* Remove us, and any offline CPUS. */ cpumask_and(to_cpumask(args->mask), cpus, cpu_online_mask); cpumask_clear_cpu(smp_processor_id(), to_cpumask(args->mask)); - if (unlikely(cpumask_empty(to_cpumask(args->mask)))) - goto issue; if (va == TLB_FLUSH_ALL) { args->op.cmd = MMUEXT_TLB_FLUSH_MULTI; @@ -1285,7 +1283,6 @@ static void xen_flush_tlb_others(const struct cpumask *cpus, MULTI_mmuext_op(mcs.mc, &args->op, 1, NULL, DOMID_SELF); -issue: xen_mc_issue(PARAVIRT_LAZY_MMU); } -- cgit v1.2.3-70-g09d2 From b93d51dc62a41b5c2d6f32a0870709dc0cc55545 Mon Sep 17 00:00:00 2001 From: Ian Campbell Date: Fri, 6 Feb 2009 13:35:57 -0800 Subject: x86, xen: record and display initiator of each multicall when debugging Store the caller for each multicall so we can report it on failure. Signed-off-by: Ian Campbell Signed-off-by: Ingo Molnar --- arch/x86/xen/multicalls.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/multicalls.c b/arch/x86/xen/multicalls.c index c738644b543..82f2513e975 100644 --- a/arch/x86/xen/multicalls.c +++ b/arch/x86/xen/multicalls.c @@ -39,6 +39,7 @@ struct mc_buffer { struct multicall_entry entries[MC_BATCH]; #if MC_DEBUG struct multicall_entry debug[MC_BATCH]; + void *caller[MC_BATCH]; #endif unsigned char args[MC_ARGS]; struct callback { @@ -154,11 +155,12 @@ void xen_mc_flush(void) ret, smp_processor_id()); dump_stack(); for (i = 0; i < b->mcidx; i++) { - printk(KERN_DEBUG " call %2d/%d: op=%lu arg=[%lx] result=%ld\n", + printk(KERN_DEBUG " call %2d/%d: op=%lu arg=[%lx] result=%ld\t%pF\n", i+1, b->mcidx, b->debug[i].op, b->debug[i].args[0], - b->entries[i].result); + b->entries[i].result, + b->caller[i]); } } #endif @@ -197,6 +199,9 @@ struct multicall_space __xen_mc_entry(size_t args) } ret.mc = &b->entries[b->mcidx]; +#ifdef MC_DEBUG + b->caller[b->mcidx] = __builtin_return_address(0); +#endif b->mcidx++; ret.args = &b->args[argidx]; b->argidx = argidx + args; -- cgit v1.2.3-70-g09d2 From 3d39e9d07b576ee72f2c94cfad9e618fe21763b2 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Fri, 6 Feb 2009 13:38:51 -0800 Subject: x86, xen: degrade BUG to WARN when multicall fails If one of the components of a multicall fails, WARN rather than BUG, to help with debugging. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Ingo Molnar --- arch/x86/xen/multicalls.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/multicalls.c b/arch/x86/xen/multicalls.c index 82f2513e975..6cffd5532b9 100644 --- a/arch/x86/xen/multicalls.c +++ b/arch/x86/xen/multicalls.c @@ -179,7 +179,7 @@ void xen_mc_flush(void) } b->cbidx = 0; - BUG_ON(ret); + WARN_ON(ret); } struct multicall_space __xen_mc_entry(size_t args) -- cgit v1.2.3-70-g09d2 From c99608637eac8834d830496c462c054137772122 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Fri, 6 Feb 2009 13:38:56 -0800 Subject: x86, xen: do multicall callbacks with interrupts disabled We can't call the callbacks after enabling interrupts, as we may get a nested multicall call, which would cause a great deal of havok. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Ingo Molnar --- arch/x86/xen/multicalls.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/multicalls.c b/arch/x86/xen/multicalls.c index 6cffd5532b9..8bff7e7c290 100644 --- a/arch/x86/xen/multicalls.c +++ b/arch/x86/xen/multicalls.c @@ -170,8 +170,6 @@ void xen_mc_flush(void) } else BUG_ON(b->argidx != 0); - local_irq_restore(flags); - for (i = 0; i < b->cbidx; i++) { struct callback *cb = &b->callbacks[i]; @@ -179,6 +177,8 @@ void xen_mc_flush(void) } b->cbidx = 0; + local_irq_restore(flags); + WARN_ON(ret); } -- cgit v1.2.3-70-g09d2