diff options
Diffstat (limited to 'arch/arc/include')
| -rw-r--r-- | arch/arc/include/asm/Kbuild | 1 | ||||
| -rw-r--r-- | arch/arc/include/asm/cache.h | 8 | ||||
| -rw-r--r-- | arch/arc/include/asm/irq.h | 4 | ||||
| -rw-r--r-- | arch/arc/include/asm/irqflags.h | 22 | ||||
| -rw-r--r-- | arch/arc/include/asm/mmu.h | 2 | ||||
| -rw-r--r-- | arch/arc/include/asm/mmu_context.h | 61 | ||||
| -rw-r--r-- | arch/arc/include/asm/setup.h | 2 | ||||
| -rw-r--r-- | arch/arc/include/asm/smp.h | 2 | ||||
| -rw-r--r-- | arch/arc/include/asm/tlbflush.h | 11 | ||||
| -rw-r--r-- | arch/arc/include/asm/unaligned.h | 3 |
10 files changed, 84 insertions, 32 deletions
diff --git a/arch/arc/include/asm/Kbuild b/arch/arc/include/asm/Kbuild index d8dd660..5943f7f 100644 --- a/arch/arc/include/asm/Kbuild +++ b/arch/arc/include/asm/Kbuild @@ -46,3 +46,4 @@ generic-y += ucontext.h generic-y += user.h generic-y += vga.h generic-y += xor.h +generic-y += preempt.h diff --git a/arch/arc/include/asm/cache.h b/arch/arc/include/asm/cache.h index e4abdaa..2fd3162 100644 --- a/arch/arc/include/asm/cache.h +++ b/arch/arc/include/asm/cache.h @@ -17,13 +17,7 @@ #endif #define L1_CACHE_BYTES (1 << L1_CACHE_SHIFT) - -/* For a rare case where customers have differently config I/D */ -#define ARC_ICACHE_LINE_LEN L1_CACHE_BYTES -#define ARC_DCACHE_LINE_LEN L1_CACHE_BYTES - -#define ICACHE_LINE_MASK (~(ARC_ICACHE_LINE_LEN - 1)) -#define DCACHE_LINE_MASK (~(ARC_DCACHE_LINE_LEN - 1)) +#define CACHE_LINE_MASK (~(L1_CACHE_BYTES - 1)) /* * ARC700 doesn't cache any access in top 256M. diff --git a/arch/arc/include/asm/irq.h b/arch/arc/include/asm/irq.h index c0a7210..291a70d 100644 --- a/arch/arc/include/asm/irq.h +++ b/arch/arc/include/asm/irq.h @@ -18,8 +18,8 @@ #include <asm-generic/irq.h> -extern void __init arc_init_IRQ(void); -extern int __init get_hw_config_num_irq(void); +extern void arc_init_IRQ(void); +extern int get_hw_config_num_irq(void); void arc_local_timer_setup(unsigned int cpu); diff --git a/arch/arc/include/asm/irqflags.h b/arch/arc/include/asm/irqflags.h index b68b53f..cb7efc2 100644 --- a/arch/arc/include/asm/irqflags.h +++ b/arch/arc/include/asm/irqflags.h @@ -151,16 +151,38 @@ static inline void arch_unmask_irq(unsigned int irq) #else +#ifdef CONFIG_TRACE_IRQFLAGS + +.macro TRACE_ASM_IRQ_DISABLE + bl trace_hardirqs_off +.endm + +.macro TRACE_ASM_IRQ_ENABLE + bl trace_hardirqs_on +.endm + +#else + +.macro TRACE_ASM_IRQ_DISABLE +.endm + +.macro TRACE_ASM_IRQ_ENABLE +.endm + +#endif + .macro IRQ_DISABLE scratch lr \scratch, [status32] bic \scratch, \scratch, (STATUS_E1_MASK | STATUS_E2_MASK) flag \scratch + TRACE_ASM_IRQ_DISABLE .endm .macro IRQ_ENABLE scratch lr \scratch, [status32] or \scratch, \scratch, (STATUS_E1_MASK | STATUS_E2_MASK) flag \scratch + TRACE_ASM_IRQ_ENABLE .endm #endif /* __ASSEMBLY__ */ diff --git a/arch/arc/include/asm/mmu.h b/arch/arc/include/asm/mmu.h index c2663b3..8c84ae9 100644 --- a/arch/arc/include/asm/mmu.h +++ b/arch/arc/include/asm/mmu.h @@ -48,7 +48,7 @@ #ifndef __ASSEMBLY__ typedef struct { - unsigned long asid; /* 8 bit MMU PID + Generation cycle */ + unsigned long asid[NR_CPUS]; /* 8 bit MMU PID + Generation cycle */ } mm_context_t; #ifdef CONFIG_ARC_DBG_TLB_PARANOIA diff --git a/arch/arc/include/asm/mmu_context.h b/arch/arc/include/asm/mmu_context.h index 43a1b51..1fd467e 100644 --- a/arch/arc/include/asm/mmu_context.h +++ b/arch/arc/include/asm/mmu_context.h @@ -30,13 +30,13 @@ * "Fast Context Switch" i.e. no TLB flush on ctxt-switch * * Linux assigns each task a unique ASID. A simple round-robin allocation - * of H/w ASID is done using software tracker @asid_cache. + * of H/w ASID is done using software tracker @asid_cpu. * When it reaches max 255, the allocation cycle starts afresh by flushing * the entire TLB and wrapping ASID back to zero. * * A new allocation cycle, post rollover, could potentially reassign an ASID * to a different task. Thus the rule is to refresh the ASID in a new cycle. - * The 32 bit @asid_cache (and mm->asid) have 8 bits MMU PID and rest 24 bits + * The 32 bit @asid_cpu (and mm->asid) have 8 bits MMU PID and rest 24 bits * serve as cycle/generation indicator and natural 32 bit unsigned math * automagically increments the generation when lower 8 bits rollover. */ @@ -47,9 +47,11 @@ #define MM_CTXT_FIRST_CYCLE (MM_CTXT_ASID_MASK + 1) #define MM_CTXT_NO_ASID 0UL -#define hw_pid(mm) (mm->context.asid & MM_CTXT_ASID_MASK) +#define asid_mm(mm, cpu) mm->context.asid[cpu] +#define hw_pid(mm, cpu) (asid_mm(mm, cpu) & MM_CTXT_ASID_MASK) -extern unsigned int asid_cache; +DECLARE_PER_CPU(unsigned int, asid_cache); +#define asid_cpu(cpu) per_cpu(asid_cache, cpu) /* * Get a new ASID if task doesn't have a valid one (unalloc or from prev cycle) @@ -57,6 +59,7 @@ extern unsigned int asid_cache; */ static inline void get_new_mmu_context(struct mm_struct *mm) { + const unsigned int cpu = smp_processor_id(); unsigned long flags; local_irq_save(flags); @@ -71,28 +74,28 @@ static inline void get_new_mmu_context(struct mm_struct *mm) * first need to destroy the context, setting it to invalid * value. */ - if (!((mm->context.asid ^ asid_cache) & MM_CTXT_CYCLE_MASK)) + if (!((asid_mm(mm, cpu) ^ asid_cpu(cpu)) & MM_CTXT_CYCLE_MASK)) goto set_hw; /* move to new ASID and handle rollover */ - if (unlikely(!(++asid_cache & MM_CTXT_ASID_MASK))) { + if (unlikely(!(++asid_cpu(cpu) & MM_CTXT_ASID_MASK))) { - flush_tlb_all(); + local_flush_tlb_all(); /* * Above checke for rollover of 8 bit ASID in 32 bit container. * If the container itself wrapped around, set it to a non zero * "generation" to distinguish from no context */ - if (!asid_cache) - asid_cache = MM_CTXT_FIRST_CYCLE; + if (!asid_cpu(cpu)) + asid_cpu(cpu) = MM_CTXT_FIRST_CYCLE; } /* Assign new ASID to tsk */ - mm->context.asid = asid_cache; + asid_mm(mm, cpu) = asid_cpu(cpu); set_hw: - write_aux_reg(ARC_REG_PID, hw_pid(mm) | MMU_ENABLE); + write_aux_reg(ARC_REG_PID, hw_pid(mm, cpu) | MMU_ENABLE); local_irq_restore(flags); } @@ -104,16 +107,45 @@ set_hw: static inline int init_new_context(struct task_struct *tsk, struct mm_struct *mm) { - mm->context.asid = MM_CTXT_NO_ASID; + int i; + + for_each_possible_cpu(i) + asid_mm(mm, i) = MM_CTXT_NO_ASID; + return 0; } +static inline void destroy_context(struct mm_struct *mm) +{ + unsigned long flags; + + /* Needed to elide CONFIG_DEBUG_PREEMPT warning */ + local_irq_save(flags); + asid_mm(mm, smp_processor_id()) = MM_CTXT_NO_ASID; + local_irq_restore(flags); +} + /* Prepare the MMU for task: setup PID reg with allocated ASID If task doesn't have an ASID (never alloc or stolen, get a new ASID) */ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next, struct task_struct *tsk) { + const int cpu = smp_processor_id(); + + /* + * Note that the mm_cpumask is "aggregating" only, we don't clear it + * for the switched-out task, unlike some other arches. + * It is used to enlist cpus for sending TLB flush IPIs and not sending + * it to CPUs where a task once ran-on, could cause stale TLB entry + * re-use, specially for a multi-threaded task. + * e.g. T1 runs on C1, migrates to C3. T2 running on C2 munmaps. + * For a non-aggregating mm_cpumask, IPI not sent C1, and if T1 + * were to re-migrate to C1, it could access the unmapped region + * via any existing stale TLB entries. + */ + cpumask_set_cpu(cpu, mm_cpumask(next)); + #ifndef CONFIG_SMP /* PGD cached in MMU reg to avoid 3 mem lookups: task->mm->pgd */ write_aux_reg(ARC_REG_SCRATCH_DATA0, next->pgd); @@ -131,11 +163,6 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next, */ #define activate_mm(prev, next) switch_mm(prev, next, NULL) -static inline void destroy_context(struct mm_struct *mm) -{ - mm->context.asid = MM_CTXT_NO_ASID; -} - /* it seemed that deactivate_mm( ) is a reasonable place to do book-keeping * for retiring-mm. However destroy_context( ) still needs to do that because * between mm_release( ) = >deactive_mm( ) and diff --git a/arch/arc/include/asm/setup.h b/arch/arc/include/asm/setup.h index 229e506..e10f8ce 100644 --- a/arch/arc/include/asm/setup.h +++ b/arch/arc/include/asm/setup.h @@ -31,7 +31,7 @@ struct cpuinfo_data { extern int root_mountflags, end_mem; extern int running_on_hw; -void __init setup_processor(void); +void setup_processor(void); void __init setup_arch_memory(void); #endif /* __ASMARC_SETUP_H */ diff --git a/arch/arc/include/asm/smp.h b/arch/arc/include/asm/smp.h index c4fb211..eefc29f 100644 --- a/arch/arc/include/asm/smp.h +++ b/arch/arc/include/asm/smp.h @@ -30,7 +30,7 @@ extern void arch_send_call_function_ipi_mask(const struct cpumask *mask); * APIs provided by arch SMP code to rest of arch code */ extern void __init smp_init_cpus(void); -extern void __init first_lines_of_secondary(void); +extern void first_lines_of_secondary(void); extern const char *arc_platform_smp_cpuinfo(void); /* diff --git a/arch/arc/include/asm/tlbflush.h b/arch/arc/include/asm/tlbflush.h index b2f9bc7..71c7b2e 100644 --- a/arch/arc/include/asm/tlbflush.h +++ b/arch/arc/include/asm/tlbflush.h @@ -18,11 +18,18 @@ void local_flush_tlb_kernel_range(unsigned long start, unsigned long end); void local_flush_tlb_range(struct vm_area_struct *vma, unsigned long start, unsigned long end); -/* XXX: Revisit for SMP */ +#ifndef CONFIG_SMP #define flush_tlb_range(vma, s, e) local_flush_tlb_range(vma, s, e) #define flush_tlb_page(vma, page) local_flush_tlb_page(vma, page) #define flush_tlb_kernel_range(s, e) local_flush_tlb_kernel_range(s, e) #define flush_tlb_all() local_flush_tlb_all() #define flush_tlb_mm(mm) local_flush_tlb_mm(mm) - +#else +extern void flush_tlb_range(struct vm_area_struct *vma, unsigned long start, + unsigned long end); +extern void flush_tlb_page(struct vm_area_struct *vma, unsigned long page); +extern void flush_tlb_kernel_range(unsigned long start, unsigned long end); +extern void flush_tlb_all(void); +extern void flush_tlb_mm(struct mm_struct *mm); +#endif /* CONFIG_SMP */ #endif diff --git a/arch/arc/include/asm/unaligned.h b/arch/arc/include/asm/unaligned.h index 60702f3..3e5f071 100644 --- a/arch/arc/include/asm/unaligned.h +++ b/arch/arc/include/asm/unaligned.h @@ -22,7 +22,8 @@ static inline int misaligned_fixup(unsigned long address, struct pt_regs *regs, struct callee_regs *cregs) { - return 0; + /* Not fixed */ + return 1; } #endif |
