From 39fe05e58c5e448601ce46e6b03900d5bf31c4b0 Mon Sep 17 00:00:00 2001 From: Shaohua Li Date: Wed, 12 Aug 2009 11:16:12 +0800 Subject: x86, hpet: Disable per-cpu hpet timer if ARAT is supported If CPU support always running local APIC timer, per-cpu hpet timer could be disabled, which is useless and wasteful in such case. Let's leave the timers to others. The effect is that we reserve less timers. Signed-off-by: Shaohua Li Cc: venkatesh.pallipadi@intel.com LKML-Reference: <20090812031612.GA10062@sli10-desk.sh.intel.com> Signed-off-by: Ingo Molnar --- arch/x86/kernel/hpet.c | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/hpet.c b/arch/x86/kernel/hpet.c index dedc2bddf7a..5969e1078fc 100644 --- a/arch/x86/kernel/hpet.c +++ b/arch/x86/kernel/hpet.c @@ -584,6 +584,8 @@ static void hpet_msi_capability_lookup(unsigned int start_timer) unsigned int num_timers_used = 0; int i; + if (boot_cpu_has(X86_FEATURE_ARAT)) + return; id = hpet_readl(HPET_ID); num_timers = ((id & HPET_ID_NUMBER) >> HPET_ID_NUMBER_SHIFT); @@ -872,10 +874,8 @@ int __init hpet_enable(void) if (id & HPET_ID_LEGSUP) { hpet_legacy_clockevent_register(); - hpet_msi_capability_lookup(2); return 1; } - hpet_msi_capability_lookup(0); return 0; out_nohpet: @@ -908,9 +908,17 @@ static __init int hpet_late_init(void) if (!hpet_virt_address) return -ENODEV; + if (hpet_readl(HPET_ID) & HPET_ID_LEGSUP) + hpet_msi_capability_lookup(2); + else + hpet_msi_capability_lookup(0); + hpet_reserve_platform_timers(hpet_readl(HPET_ID)); hpet_print_config(); + if (boot_cpu_has(X86_FEATURE_ARAT)) + return 0; + for_each_online_cpu(cpu) { hpet_cpuhp_notify(NULL, CPU_ONLINE, (void *)(long)cpu); } -- cgit v1.2.3 From 5946fa3d5cdeb846a647a1900026af9f8b08c8b5 Mon Sep 17 00:00:00 2001 From: Jan Beulich Date: Wed, 19 Aug 2009 08:44:24 +0100 Subject: x86, hpet: Simplify the HPET code On 64-bits, using unsigned long when unsigned int suffices needlessly creates larger code (due to the need for REX prefixes), and most of the logic in hpet.c really doesn't need 64-bit operations. At once this avoids the need for a couple of type casts. Signed-off-by: Jan Beulich Cc: Shaohua Li Cc: Venkatesh Pallipadi LKML-Reference: <4A8BC9780200007800010832@vpn.id2.novell.com> Signed-off-by: Ingo Molnar --- arch/x86/include/asm/hpet.h | 2 +- arch/x86/kernel/hpet.c | 45 +++++++++++++++++++++++---------------------- 2 files changed, 24 insertions(+), 23 deletions(-) (limited to 'arch') diff --git a/arch/x86/include/asm/hpet.h b/arch/x86/include/asm/hpet.h index 1c22cb05ad6..65847c578b7 100644 --- a/arch/x86/include/asm/hpet.h +++ b/arch/x86/include/asm/hpet.h @@ -69,7 +69,7 @@ extern int hpet_force_user; extern int is_hpet_enabled(void); extern int hpet_enable(void); extern void hpet_disable(void); -extern unsigned long hpet_readl(unsigned long a); +extern unsigned int hpet_readl(unsigned int a); extern void force_hpet_resume(void); extern void hpet_msi_unmask(unsigned int irq); diff --git a/arch/x86/kernel/hpet.c b/arch/x86/kernel/hpet.c index 5969e1078fc..ba575f0f2e3 100644 --- a/arch/x86/kernel/hpet.c +++ b/arch/x86/kernel/hpet.c @@ -47,12 +47,12 @@ struct hpet_dev { char name[10]; }; -unsigned long hpet_readl(unsigned long a) +inline unsigned int hpet_readl(unsigned int a) { return readl(hpet_virt_address + a); } -static inline void hpet_writel(unsigned long d, unsigned long a) +static inline void hpet_writel(unsigned int d, unsigned int a) { writel(d, hpet_virt_address + a); } @@ -167,7 +167,7 @@ do { \ static void hpet_reserve_msi_timers(struct hpet_data *hd); -static void hpet_reserve_platform_timers(unsigned long id) +static void hpet_reserve_platform_timers(unsigned int id) { struct hpet __iomem *hpet = hpet_virt_address; struct hpet_timer __iomem *timer = &hpet->hpet_timers[2]; @@ -205,7 +205,7 @@ static void hpet_reserve_platform_timers(unsigned long id) } #else -static void hpet_reserve_platform_timers(unsigned long id) { } +static void hpet_reserve_platform_timers(unsigned int id) { } #endif /* @@ -246,7 +246,7 @@ static void hpet_reset_counter(void) static void hpet_start_counter(void) { - unsigned long cfg = hpet_readl(HPET_CFG); + unsigned int cfg = hpet_readl(HPET_CFG); cfg |= HPET_CFG_ENABLE; hpet_writel(cfg, HPET_CFG); } @@ -271,7 +271,7 @@ static void hpet_resume_counter(void) static void hpet_enable_legacy_int(void) { - unsigned long cfg = hpet_readl(HPET_CFG); + unsigned int cfg = hpet_readl(HPET_CFG); cfg |= HPET_CFG_LEGACY; hpet_writel(cfg, HPET_CFG); @@ -314,7 +314,7 @@ static int hpet_setup_msi_irq(unsigned int irq); static void hpet_set_mode(enum clock_event_mode mode, struct clock_event_device *evt, int timer) { - unsigned long cfg, cmp, now; + unsigned int cfg, cmp, now; uint64_t delta; switch (mode) { @@ -323,7 +323,7 @@ static void hpet_set_mode(enum clock_event_mode mode, delta = ((uint64_t)(NSEC_PER_SEC/HZ)) * evt->mult; delta >>= evt->shift; now = hpet_readl(HPET_COUNTER); - cmp = now + (unsigned long) delta; + cmp = now + (unsigned int) delta; cfg = hpet_readl(HPET_Tn_CFG(timer)); /* Make sure we use edge triggered interrupts */ cfg &= ~HPET_TN_LEVEL; @@ -339,7 +339,7 @@ static void hpet_set_mode(enum clock_event_mode mode, * (See AMD-8111 HyperTransport I/O Hub Data Sheet, * Publication # 24674) */ - hpet_writel((unsigned long) delta, HPET_Tn_CMP(timer)); + hpet_writel((unsigned int) delta, HPET_Tn_CMP(timer)); hpet_start_counter(); hpet_print_config(); break; @@ -387,9 +387,9 @@ static int hpet_next_event(unsigned long delta, * what we wrote hit the chip before we compare it to the * counter. */ - WARN_ON_ONCE((u32)hpet_readl(HPET_Tn_CMP(timer)) != cnt); + WARN_ON_ONCE(hpet_readl(HPET_Tn_CMP(timer)) != cnt); - return (s32)((u32)hpet_readl(HPET_COUNTER) - cnt) >= 0 ? -ETIME : 0; + return (s32)(hpet_readl(HPET_COUNTER) - cnt) >= 0 ? -ETIME : 0; } static void hpet_legacy_set_mode(enum clock_event_mode mode, @@ -415,7 +415,7 @@ static struct hpet_dev *hpet_devs; void hpet_msi_unmask(unsigned int irq) { struct hpet_dev *hdev = get_irq_data(irq); - unsigned long cfg; + unsigned int cfg; /* unmask it */ cfg = hpet_readl(HPET_Tn_CFG(hdev->num)); @@ -425,7 +425,7 @@ void hpet_msi_unmask(unsigned int irq) void hpet_msi_mask(unsigned int irq) { - unsigned long cfg; + unsigned int cfg; struct hpet_dev *hdev = get_irq_data(irq); /* mask it */ @@ -600,7 +600,7 @@ static void hpet_msi_capability_lookup(unsigned int start_timer) for (i = start_timer; i < num_timers - RESERVE_TIMERS; i++) { struct hpet_dev *hdev = &hpet_devs[num_timers_used]; - unsigned long cfg = hpet_readl(HPET_Tn_CFG(i)); + unsigned int cfg = hpet_readl(HPET_Tn_CFG(i)); /* Only consider HPET timer with MSI support */ if (!(cfg & HPET_TN_FSB_CAP)) @@ -815,7 +815,7 @@ static int hpet_clocksource_register(void) */ int __init hpet_enable(void) { - unsigned long id; + unsigned int id; int i; if (!is_hpet_capable()) @@ -933,7 +933,7 @@ fs_initcall(hpet_late_init); void hpet_disable(void) { if (is_hpet_capable()) { - unsigned long cfg = hpet_readl(HPET_CFG); + unsigned int cfg = hpet_readl(HPET_CFG); if (hpet_legacy_int_enabled) { cfg &= ~HPET_CFG_LEGACY; @@ -973,8 +973,8 @@ static int hpet_prev_update_sec; static struct rtc_time hpet_alarm_time; static unsigned long hpet_pie_count; static u32 hpet_t1_cmp; -static unsigned long hpet_default_delta; -static unsigned long hpet_pie_delta; +static u32 hpet_default_delta; +static u32 hpet_pie_delta; static unsigned long hpet_pie_limit; static rtc_irq_handler irq_handler; @@ -1025,7 +1025,8 @@ EXPORT_SYMBOL_GPL(hpet_unregister_irq_handler); */ int hpet_rtc_timer_init(void) { - unsigned long cfg, cnt, delta, flags; + unsigned int cfg, cnt, delta; + unsigned long flags; if (!is_hpet_enabled()) return 0; @@ -1035,7 +1036,7 @@ int hpet_rtc_timer_init(void) clc = (uint64_t) hpet_clockevent.mult * NSEC_PER_SEC; clc >>= hpet_clockevent.shift + DEFAULT_RTC_SHIFT; - hpet_default_delta = (unsigned long) clc; + hpet_default_delta = clc; } if (!(hpet_rtc_flags & RTC_PIE) || hpet_pie_limit) @@ -1121,7 +1122,7 @@ int hpet_set_periodic_freq(unsigned long freq) clc = (uint64_t) hpet_clockevent.mult * NSEC_PER_SEC; do_div(clc, freq); clc >>= hpet_clockevent.shift; - hpet_pie_delta = (unsigned long) clc; + hpet_pie_delta = clc; } return 1; } @@ -1135,7 +1136,7 @@ EXPORT_SYMBOL_GPL(hpet_rtc_dropped_irq); static void hpet_rtc_timer_reinit(void) { - unsigned long cfg, delta; + unsigned int cfg, delta; int lost_ints = -1; if (unlikely(!hpet_rtc_flags)) { -- cgit v1.2.3 From c8bc6f3c806f1fcbfdbf0b1ff6c52dba59192d3b Mon Sep 17 00:00:00 2001 From: Suresh Siddha Date: Tue, 4 Aug 2009 12:07:09 -0700 Subject: x86: arch specific support for remapping HPET MSIs x86 arch support for remapping HPET MSI's by associating the HPET timer block with the interrupt-remapping HW unit and setting up appropriate irq_chip Signed-off-by: Suresh Siddha Cc: Venkatesh Pallipadi Cc: David Woodhouse Cc: Jesse Barnes Cc: Jay Fenlason LKML-Reference: <20090804190729.630510000@intel.com> Signed-off-by: Thomas Gleixner --- arch/x86/include/asm/hpet.h | 5 +++-- arch/x86/kernel/acpi/boot.c | 1 + arch/x86/kernel/apic/io_apic.c | 49 +++++++++++++++++++++++++++++++++++------- arch/x86/kernel/hpet.c | 3 ++- 4 files changed, 47 insertions(+), 11 deletions(-) (limited to 'arch') diff --git a/arch/x86/include/asm/hpet.h b/arch/x86/include/asm/hpet.h index 65847c578b7..5d89fd2a369 100644 --- a/arch/x86/include/asm/hpet.h +++ b/arch/x86/include/asm/hpet.h @@ -65,6 +65,7 @@ /* hpet memory map physical address */ extern unsigned long hpet_address; extern unsigned long force_hpet_address; +extern u8 hpet_blockid; extern int hpet_force_user; extern int is_hpet_enabled(void); extern int hpet_enable(void); @@ -78,9 +79,9 @@ extern void hpet_msi_write(unsigned int irq, struct msi_msg *msg); extern void hpet_msi_read(unsigned int irq, struct msi_msg *msg); #ifdef CONFIG_PCI_MSI -extern int arch_setup_hpet_msi(unsigned int irq); +extern int arch_setup_hpet_msi(unsigned int irq, unsigned int id); #else -static inline int arch_setup_hpet_msi(unsigned int irq) +static inline int arch_setup_hpet_msi(unsigned int irq, unsigned int id) { return -EINVAL; } diff --git a/arch/x86/kernel/acpi/boot.c b/arch/x86/kernel/acpi/boot.c index 6b8ca3a0285..eae642b0f34 100644 --- a/arch/x86/kernel/acpi/boot.c +++ b/arch/x86/kernel/acpi/boot.c @@ -624,6 +624,7 @@ static int __init acpi_parse_hpet(struct acpi_table_header *table) } hpet_address = hpet_tbl->address.address; + hpet_blockid = hpet_tbl->sequence; /* * Some broken BIOSes advertise HPET at 0x0. We really do not diff --git a/arch/x86/kernel/apic/io_apic.c b/arch/x86/kernel/apic/io_apic.c index d2ed6c5ddc8..d9c6f14d3b3 100644 --- a/arch/x86/kernel/apic/io_apic.c +++ b/arch/x86/kernel/apic/io_apic.c @@ -3254,7 +3254,8 @@ void destroy_irq(unsigned int irq) * MSI message composition */ #ifdef CONFIG_PCI_MSI -static int msi_compose_msg(struct pci_dev *pdev, unsigned int irq, struct msi_msg *msg) +static int msi_compose_msg(struct pci_dev *pdev, unsigned int irq, + struct msi_msg *msg, u8 hpet_id) { struct irq_cfg *cfg; int err; @@ -3288,7 +3289,10 @@ static int msi_compose_msg(struct pci_dev *pdev, unsigned int irq, struct msi_ms irte.dest_id = IRTE_DEST(dest); /* Set source-id of interrupt request */ - set_msi_sid(&irte, pdev); + if (pdev) + set_msi_sid(&irte, pdev); + else + set_hpet_sid(&irte, hpet_id); modify_irte(irq, &irte); @@ -3453,7 +3457,7 @@ static int setup_msi_irq(struct pci_dev *dev, struct msi_desc *msidesc, int irq) int ret; struct msi_msg msg; - ret = msi_compose_msg(dev, irq, &msg); + ret = msi_compose_msg(dev, irq, &msg, -1); if (ret < 0) return ret; @@ -3586,7 +3590,7 @@ int arch_setup_dmar_msi(unsigned int irq) int ret; struct msi_msg msg; - ret = msi_compose_msg(NULL, irq, &msg); + ret = msi_compose_msg(NULL, irq, &msg, -1); if (ret < 0) return ret; dmar_msi_write(irq, &msg); @@ -3626,6 +3630,19 @@ static int hpet_msi_set_affinity(unsigned int irq, const struct cpumask *mask) #endif /* CONFIG_SMP */ +static struct irq_chip ir_hpet_msi_type = { + .name = "IR-HPET_MSI", + .unmask = hpet_msi_unmask, + .mask = hpet_msi_mask, +#ifdef CONFIG_INTR_REMAP + .ack = ir_ack_apic_edge, +#ifdef CONFIG_SMP + .set_affinity = ir_set_msi_irq_affinity, +#endif +#endif + .retrigger = ioapic_retrigger_irq, +}; + static struct irq_chip hpet_msi_type = { .name = "HPET_MSI", .unmask = hpet_msi_unmask, @@ -3637,20 +3654,36 @@ static struct irq_chip hpet_msi_type = { .retrigger = ioapic_retrigger_irq, }; -int arch_setup_hpet_msi(unsigned int irq) +int arch_setup_hpet_msi(unsigned int irq, unsigned int id) { int ret; struct msi_msg msg; struct irq_desc *desc = irq_to_desc(irq); - ret = msi_compose_msg(NULL, irq, &msg); + if (intr_remapping_enabled) { + struct intel_iommu *iommu = map_hpet_to_ir(id); + int index; + + if (!iommu) + return -1; + + index = alloc_irte(iommu, irq, 1); + if (index < 0) + return -1; + } + + ret = msi_compose_msg(NULL, irq, &msg, id); if (ret < 0) return ret; hpet_msi_write(irq, &msg); desc->status |= IRQ_MOVE_PCNTXT; - set_irq_chip_and_handler_name(irq, &hpet_msi_type, handle_edge_irq, - "edge"); + if (irq_remapped(irq)) + set_irq_chip_and_handler_name(irq, &ir_hpet_msi_type, + handle_edge_irq, "edge"); + else + set_irq_chip_and_handler_name(irq, &hpet_msi_type, + handle_edge_irq, "edge"); return 0; } diff --git a/arch/x86/kernel/hpet.c b/arch/x86/kernel/hpet.c index ba575f0f2e3..7f024ff47d1 100644 --- a/arch/x86/kernel/hpet.c +++ b/arch/x86/kernel/hpet.c @@ -33,6 +33,7 @@ * HPET address is set in acpi/boot.c, when an ACPI entry exists */ unsigned long hpet_address; +u8 hpet_blockid; /* OS timer block num */ #ifdef CONFIG_PCI_MSI static unsigned long hpet_num_timers; #endif @@ -467,7 +468,7 @@ static int hpet_msi_next_event(unsigned long delta, static int hpet_setup_msi_irq(unsigned int irq) { - if (arch_setup_hpet_msi(irq)) { + if (arch_setup_hpet_msi(irq, hpet_blockid)) { destroy_irq(irq); return -EINVAL; } -- cgit v1.2.3 From 654d364e26c797e8a5f9e2a1393607e6ca0106eb Mon Sep 17 00:00:00 2001 From: Paul Mundt Date: Wed, 9 Sep 2009 14:04:06 +0900 Subject: sh: sh4_flush_cache_mm() optimizations. The i-cache flush in the case of VM_EXEC was added way back when as a sanity measure, and in practice we only care about evicting aliases from the d-cache. As a result, it's possible to drop the i-cache flush completely here. After careful profiling it's also come up that all of the work associated with hunting down aliases and doing ranged flushing ends up generating more overhead than simply blasting away the entire dcache, particularly if there are many mm's that need to be iterated over. As a result of that, just move back to flush_dcache_all() in these cases, which restores the old behaviour, and vastly simplifies the path. Additionally, on platforms without aliases at all, this can simply be nopped out. Presently we have the alias check in the SH-4 specific version, but this is true for all of the platforms, so move the check up to a generic location. This cuts down quite a bit on superfluous cacheop IPIs. Signed-off-by: Paul Mundt --- arch/sh/mm/cache-sh4.c | 124 ++----------------------------------------------- arch/sh/mm/cache.c | 6 +++ 2 files changed, 10 insertions(+), 120 deletions(-) (limited to 'arch') diff --git a/arch/sh/mm/cache-sh4.c b/arch/sh/mm/cache-sh4.c index b2453bbef4c..a5c339bca8a 100644 --- a/arch/sh/mm/cache-sh4.c +++ b/arch/sh/mm/cache-sh4.c @@ -170,89 +170,13 @@ static void sh4_flush_cache_all(void *unused) flush_icache_all(); } -static void __flush_cache_mm(struct mm_struct *mm, unsigned long start, - unsigned long end) -{ - unsigned long d = 0, p = start & PAGE_MASK; - unsigned long alias_mask = boot_cpu_data.dcache.alias_mask; - unsigned long n_aliases = boot_cpu_data.dcache.n_aliases; - unsigned long select_bit; - unsigned long all_aliases_mask; - unsigned long addr_offset; - pgd_t *dir; - pmd_t *pmd; - pud_t *pud; - pte_t *pte; - int i; - - dir = pgd_offset(mm, p); - pud = pud_offset(dir, p); - pmd = pmd_offset(pud, p); - end = PAGE_ALIGN(end); - - all_aliases_mask = (1 << n_aliases) - 1; - - do { - if (pmd_none(*pmd) || unlikely(pmd_bad(*pmd))) { - p &= PMD_MASK; - p += PMD_SIZE; - pmd++; - - continue; - } - - pte = pte_offset_kernel(pmd, p); - - do { - unsigned long phys; - pte_t entry = *pte; - - if (!(pte_val(entry) & _PAGE_PRESENT)) { - pte++; - p += PAGE_SIZE; - continue; - } - - phys = pte_val(entry) & PTE_PHYS_MASK; - - if ((p ^ phys) & alias_mask) { - d |= 1 << ((p & alias_mask) >> PAGE_SHIFT); - d |= 1 << ((phys & alias_mask) >> PAGE_SHIFT); - - if (d == all_aliases_mask) - goto loop_exit; - } - - pte++; - p += PAGE_SIZE; - } while (p < end && ((unsigned long)pte & ~PAGE_MASK)); - pmd++; - } while (p < end); - -loop_exit: - addr_offset = 0; - select_bit = 1; - - for (i = 0; i < n_aliases; i++) { - if (d & select_bit) { - (*__flush_dcache_segment_fn)(addr_offset, PAGE_SIZE); - wmb(); - } - - select_bit <<= 1; - addr_offset += PAGE_SIZE; - } -} - /* * Note : (RPC) since the caches are physically tagged, the only point * of flush_cache_mm for SH-4 is to get rid of aliases from the * D-cache. The assumption elsewhere, e.g. flush_cache_range, is that * lines can stay resident so long as the virtual address they were * accessed with (hence cache set) is in accord with the physical - * address (i.e. tag). It's no different here. So I reckon we don't - * need to flush the I-cache, since aliases don't matter for that. We - * should try that. + * address (i.e. tag). It's no different here. * * Caller takes mm->mmap_sem. */ @@ -263,33 +187,7 @@ static void sh4_flush_cache_mm(void *arg) if (cpu_context(smp_processor_id(), mm) == NO_CONTEXT) return; - /* - * If cache is only 4k-per-way, there are never any 'aliases'. Since - * the cache is physically tagged, the data can just be left in there. - */ - if (boot_cpu_data.dcache.n_aliases == 0) - return; - - /* - * Don't bother groveling around the dcache for the VMA ranges - * if there are too many PTEs to make it worthwhile. - */ - if (mm->nr_ptes >= MAX_DCACHE_PAGES) - flush_dcache_all(); - else { - struct vm_area_struct *vma; - - /* - * In this case there are reasonably sized ranges to flush, - * iterate through the VMA list and take care of any aliases. - */ - for (vma = mm->mmap; vma; vma = vma->vm_next) - __flush_cache_mm(mm, vma->vm_start, vma->vm_end); - } - - /* Only touch the icache if one of the VMAs has VM_EXEC set. */ - if (mm->exec_vm) - flush_icache_all(); + flush_dcache_all(); } /* @@ -372,24 +270,10 @@ static void sh4_flush_cache_range(void *args) if (boot_cpu_data.dcache.n_aliases == 0) return; - /* - * Don't bother with the lookup and alias check if we have a - * wide range to cover, just blow away the dcache in its - * entirety instead. -- PFM. - */ - if (((end - start) >> PAGE_SHIFT) >= MAX_DCACHE_PAGES) - flush_dcache_all(); - else - __flush_cache_mm(vma->vm_mm, start, end); + flush_dcache_all(); - if (vma->vm_flags & VM_EXEC) { - /* - * TODO: Is this required??? Need to look at how I-cache - * coherency is assured when new programs are loaded to see if - * this matters. - */ + if (vma->vm_flags & VM_EXEC) flush_icache_all(); - } } /** diff --git a/arch/sh/mm/cache.c b/arch/sh/mm/cache.c index 35c37b7f717..4aa92605453 100644 --- a/arch/sh/mm/cache.c +++ b/arch/sh/mm/cache.c @@ -164,11 +164,17 @@ void flush_cache_all(void) void flush_cache_mm(struct mm_struct *mm) { + if (boot_cpu_data.dcache.n_aliases == 0) + return; + cacheop_on_each_cpu(local_flush_cache_mm, mm, 1); } void flush_cache_dup_mm(struct mm_struct *mm) { + if (boot_cpu_data.dcache.n_aliases == 0) + return; + cacheop_on_each_cpu(local_flush_cache_dup_mm, mm, 1); } -- cgit v1.2.3 From 31c9efde786252112cc3d04a1ed3513b6ec63a7b Mon Sep 17 00:00:00 2001 From: Paul Mundt Date: Wed, 9 Sep 2009 14:10:28 +0900 Subject: sh: Kill off broken PHYSADDR() usage in sh4_flush_dcache_page(). PHYSADDR() runs in to issues in 32-bit mode when we do not have the legacy P1/P2 areas mapped, as such, we need to use page_to_phys() directly, which also happens to do the right thing in legacy 29-bit mode. Signed-off-by: Paul Mundt --- arch/sh/mm/cache-sh4.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'arch') diff --git a/arch/sh/mm/cache-sh4.c b/arch/sh/mm/cache-sh4.c index a5c339bca8a..f0999606686 100644 --- a/arch/sh/mm/cache-sh4.c +++ b/arch/sh/mm/cache-sh4.c @@ -123,12 +123,12 @@ static void sh4_flush_dcache_page(void *arg) else #endif { - unsigned long phys = PHYSADDR(page_address(page)); + unsigned long phys = page_to_phys(page); unsigned long addr = CACHE_OC_ADDRESS_ARRAY; int i, n; /* Loop all the D-cache */ - n = boot_cpu_data.dcache.n_aliases; + n = boot_cpu_data.dcache.way_incr >> 12; for (i = 0; i < n; i++, addr += 4096) flush_cache_4096(addr, phys); } -- cgit v1.2.3 From bd6df57481b329dfeeb4889068848ee4f4761561 Mon Sep 17 00:00:00 2001 From: Paul Mundt Date: Wed, 9 Sep 2009 14:22:15 +0900 Subject: sh: Kill off segment-based d-cache flushing on SH-4. This kills off the unrolled segment based flushers on SH-4 and switches over to a generic unrolled approach derived from the writethrough segment flusher. Signed-off-by: Paul Mundt --- arch/sh/mm/cache-sh4.c | 291 ++++--------------------------------------------- 1 file changed, 20 insertions(+), 271 deletions(-) (limited to 'arch') diff --git a/arch/sh/mm/cache-sh4.c b/arch/sh/mm/cache-sh4.c index f0999606686..92b7d947db9 100644 --- a/arch/sh/mm/cache-sh4.c +++ b/arch/sh/mm/cache-sh4.c @@ -29,14 +29,6 @@ static void __flush_cache_4096(unsigned long addr, unsigned long phys, unsigned long exec_offset); -/* - * This is initialised here to ensure that it is not placed in the BSS. If - * that were to happen, note that cache_init gets called before the BSS is - * cleared, so this would get nulled out which would be hopeless. - */ -static void (*__flush_dcache_segment_fn)(unsigned long, unsigned long) = - (void (*)(unsigned long, unsigned long))0xdeadbeef; - /* * Write back the range of D-cache, and purge the I-cache. * @@ -158,10 +150,27 @@ static void __uses_jump_to_uncached flush_icache_all(void) local_irq_restore(flags); } -static inline void flush_dcache_all(void) +static void flush_dcache_all(void) { - (*__flush_dcache_segment_fn)(0UL, boot_cpu_data.dcache.way_size); - wmb(); + unsigned long addr, end_addr, entry_offset; + + end_addr = CACHE_OC_ADDRESS_ARRAY + + (current_cpu_data.dcache.sets << + current_cpu_data.dcache.entry_shift) * + current_cpu_data.dcache.ways; + + entry_offset = 1 << current_cpu_data.dcache.entry_shift; + + for (addr = CACHE_OC_ADDRESS_ARRAY; addr < end_addr; ) { + __raw_writel(0, addr); addr += entry_offset; + __raw_writel(0, addr); addr += entry_offset; + __raw_writel(0, addr); addr += entry_offset; + __raw_writel(0, addr); addr += entry_offset; + __raw_writel(0, addr); addr += entry_offset; + __raw_writel(0, addr); addr += entry_offset; + __raw_writel(0, addr); addr += entry_offset; + __raw_writel(0, addr); addr += entry_offset; + } } static void sh4_flush_cache_all(void *unused) @@ -347,245 +356,6 @@ static void __flush_cache_4096(unsigned long addr, unsigned long phys, } while (--way_count != 0); } -/* - * Break the 1, 2 and 4 way variants of this out into separate functions to - * avoid nearly all the overhead of having the conditional stuff in the function - * bodies (+ the 1 and 2 way cases avoid saving any registers too). - * - * We want to eliminate unnecessary bus transactions, so this code uses - * a non-obvious technique. - * - * Loop over a cache way sized block of, one cache line at a time. For each - * line, use movca.a to cause the current cache line contents to be written - * back, but without reading anything from main memory. However this has the - * side effect that the cache is now caching that memory location. So follow - * this with a cache invalidate to mark the cache line invalid. And do all - * this with interrupts disabled, to avoid the cache line being accidently - * evicted while it is holding garbage. - * - * This also breaks in a number of circumstances: - * - if there are modifications to the region of memory just above - * empty_zero_page (for example because a breakpoint has been placed - * there), then these can be lost. - * - * This is because the the memory address which the cache temporarily - * caches in the above description is empty_zero_page. So the - * movca.l hits the cache (it is assumed that it misses, or at least - * isn't dirty), modifies the line and then invalidates it, losing the - * required change. - * - * - If caches are disabled or configured in write-through mode, then - * the movca.l writes garbage directly into memory. - */ -static void __flush_dcache_segment_writethrough(unsigned long start, - unsigned long extent_per_way) -{ - unsigned long addr; - int i; - - addr = CACHE_OC_ADDRESS_ARRAY | (start & cpu_data->dcache.entry_mask); - - while (extent_per_way) { - for (i = 0; i < cpu_data->dcache.ways; i++) - __raw_writel(0, addr + cpu_data->dcache.way_incr * i); - - addr += cpu_data->dcache.linesz; - extent_per_way -= cpu_data->dcache.linesz; - } -} - -static void __flush_dcache_segment_1way(unsigned long start, - unsigned long extent_per_way) -{ - unsigned long orig_sr, sr_with_bl; - unsigned long base_addr; - unsigned long way_incr, linesz, way_size; - struct cache_info *dcache; - register unsigned long a0, a0e; - - asm volatile("stc sr, %0" : "=r" (orig_sr)); - sr_with_bl = orig_sr | (1<<28); - base_addr = ((unsigned long)&empty_zero_page[0]); - - /* - * The previous code aligned base_addr to 16k, i.e. the way_size of all - * existing SH-4 D-caches. Whilst I don't see a need to have this - * aligned to any better than the cache line size (which it will be - * anyway by construction), let's align it to at least the way_size of - * any existing or conceivable SH-4 D-cache. -- RPC - */ - base_addr = ((base_addr >> 16) << 16); - base_addr |= start; - - dcache = &boot_cpu_data.dcache; - linesz = dcache->linesz; - way_incr = dcache->way_incr; - way_size = dcache->way_size; - - a0 = base_addr; - a0e = base_addr + extent_per_way; - do { - asm volatile("ldc %0, sr" : : "r" (sr_with_bl)); - asm volatile("movca.l r0, @%0\n\t" - "ocbi @%0" : : "r" (a0)); - a0 += linesz; - asm volatile("movca.l r0, @%0\n\t" - "ocbi @%0" : : "r" (a0)); - a0 += linesz; - asm volatile("movca.l r0, @%0\n\t" - "ocbi @%0" : : "r" (a0)); - a0 += linesz; - asm volatile("movca.l r0, @%0\n\t" - "ocbi @%0" : : "r" (a0)); - asm volatile("ldc %0, sr" : : "r" (orig_sr)); - a0 += linesz; - } while (a0 < a0e); -} - -static void __flush_dcache_segment_2way(unsigned long start, - unsigned long extent_per_way) -{ - unsigned long orig_sr, sr_with_bl; - unsigned long base_addr; - unsigned long way_incr, linesz, way_size; - struct cache_info *dcache; - register unsigned long a0, a1, a0e; - - asm volatile("stc sr, %0" : "=r" (orig_sr)); - sr_with_bl = orig_sr | (1<<28); - base_addr = ((unsigned long)&empty_zero_page[0]); - - /* See comment under 1-way above */ - base_addr = ((base_addr >> 16) << 16); - base_addr |= start; - - dcache = &boot_cpu_data.dcache; - linesz = dcache->linesz; - way_incr = dcache->way_incr; - way_size = dcache->way_size; - - a0 = base_addr; - a1 = a0 + way_incr; - a0e = base_addr + extent_per_way; - do { - asm volatile("ldc %0, sr" : : "r" (sr_with_bl)); - asm volatile("movca.l r0, @%0\n\t" - "movca.l r0, @%1\n\t" - "ocbi @%0\n\t" - "ocbi @%1" : : - "r" (a0), "r" (a1)); - a0 += linesz; - a1 += linesz; - asm volatile("movca.l r0, @%0\n\t" - "movca.l r0, @%1\n\t" - "ocbi @%0\n\t" - "ocbi @%1" : : - "r" (a0), "r" (a1)); - a0 += linesz; - a1 += linesz; - asm volatile("movca.l r0, @%0\n\t" - "movca.l r0, @%1\n\t" - "ocbi @%0\n\t" - "ocbi @%1" : : - "r" (a0), "r" (a1)); - a0 += linesz; - a1 += linesz; - asm volatile("movca.l r0, @%0\n\t" - "movca.l r0, @%1\n\t" - "ocbi @%0\n\t" - "ocbi @%1" : : - "r" (a0), "r" (a1)); - asm volatile("ldc %0, sr" : : "r" (orig_sr)); - a0 += linesz; - a1 += linesz; - } while (a0 < a0e); -} - -static void __flush_dcache_segment_4way(unsigned long start, - unsigned long extent_per_way) -{ - unsigned long orig_sr, sr_with_bl; - unsigned long base_addr; - unsigned long way_incr, linesz, way_size; - struct cache_info *dcache; - register unsigned long a0, a1, a2, a3, a0e; - - asm volatile("stc sr, %0" : "=r" (orig_sr)); - sr_with_bl = orig_sr | (1<<28); - base_addr = ((unsigned long)&empty_zero_page[0]); - - /* See comment under 1-way above */ - base_addr = ((base_addr >> 16) << 16); - base_addr |= start; - - dcache = &boot_cpu_data.dcache; - linesz = dcache->linesz; - way_incr = dcache->way_incr; - way_size = dcache->way_size; - - a0 = base_addr; - a1 = a0 + way_incr; - a2 = a1 + way_incr; - a3 = a2 + way_incr; - a0e = base_addr + extent_per_way; - do { - asm volatile("ldc %0, sr" : : "r" (sr_with_bl)); - asm volatile("movca.l r0, @%0\n\t" - "movca.l r0, @%1\n\t" - "movca.l r0, @%2\n\t" - "movca.l r0, @%3\n\t" - "ocbi @%0\n\t" - "ocbi @%1\n\t" - "ocbi @%2\n\t" - "ocbi @%3\n\t" : : - "r" (a0), "r" (a1), "r" (a2), "r" (a3)); - a0 += linesz; - a1 += linesz; - a2 += linesz; - a3 += linesz; - asm volatile("movca.l r0, @%0\n\t" - "movca.l r0, @%1\n\t" - "movca.l r0, @%2\n\t" - "movca.l r0, @%3\n\t" - "ocbi @%0\n\t" - "ocbi @%1\n\t" - "ocbi @%2\n\t" - "ocbi @%3\n\t" : : - "r" (a0), "r" (a1), "r" (a2), "r" (a3)); - a0 += linesz; - a1 += linesz; - a2 += linesz; - a3 += linesz; - asm volatile("movca.l r0, @%0\n\t" - "movca.l r0, @%1\n\t" - "movca.l r0, @%2\n\t" - "movca.l r0, @%3\n\t" - "ocbi @%0\n\t" - "ocbi @%1\n\t" - "ocbi @%2\n\t" - "ocbi @%3\n\t" : : - "r" (a0), "r" (a1), "r" (a2), "r" (a3)); - a0 += linesz; - a1 += linesz; - a2 += linesz; - a3 += linesz; - asm volatile("movca.l r0, @%0\n\t" - "movca.l r0, @%1\n\t" - "movca.l r0, @%2\n\t" - "movca.l r0, @%3\n\t" - "ocbi @%0\n\t" - "ocbi @%1\n\t" - "ocbi @%2\n\t" - "ocbi @%3\n\t" : : - "r" (a0), "r" (a1), "r" (a2), "r" (a3)); - asm volatile("ldc %0, sr" : : "r" (orig_sr)); - a0 += linesz; - a1 += linesz; - a2 += linesz; - a3 += linesz; - } while (a0 < a0e); -} - extern void __weak sh4__flush_region_init(void); /* @@ -593,32 +363,11 @@ extern void __weak sh4__flush_region_init(void); */ void __init sh4_cache_init(void) { - unsigned int wt_enabled = !!(__raw_readl(CCR) & CCR_CACHE_WT); - printk("PVR=%08x CVR=%08x PRR=%08x\n", ctrl_inl(CCN_PVR), ctrl_inl(CCN_CVR), ctrl_inl(CCN_PRR)); - if (wt_enabled) - __flush_dcache_segment_fn = __flush_dcache_segment_writethrough; - else { - switch (boot_cpu_data.dcache.ways) { - case 1: - __flush_dcache_segment_fn = __flush_dcache_segment_1way; - break; - case 2: - __flush_dcache_segment_fn = __flush_dcache_segment_2way; - break; - case 4: - __flush_dcache_segment_fn = __flush_dcache_segment_4way; - break; - default: - panic("unknown number of cache ways\n"); - break; - } - } - local_flush_icache_range = sh4_flush_icache_range; local_flush_dcache_page = sh4_flush_dcache_page; local_flush_cache_all = sh4_flush_cache_all; -- cgit v1.2.3 From deaef20e9789d93c06d2d3b5ffc99939814802ca Mon Sep 17 00:00:00 2001 From: Paul Mundt Date: Wed, 9 Sep 2009 16:06:39 +0900 Subject: sh: Rework sh4_flush_cache_page() for coherent kmap mapping. This builds on top of the MIPS r4k code that does roughly the same thing. This permits the use of kmap_coherent() for mapped pages with dirty dcache lines and falls back on kmap_atomic() otherwise. This also fixes up a problem with the alias check and defers to shm_align_mask directly. Signed-off-by: Paul Mundt --- arch/sh/mm/cache-sh4.c | 75 ++++++++++++++++++++++++++++++++------------------ 1 file changed, 48 insertions(+), 27 deletions(-) (limited to 'arch') diff --git a/arch/sh/mm/cache-sh4.c b/arch/sh/mm/cache-sh4.c index 92b7d947db9..e3fbd99b323 100644 --- a/arch/sh/mm/cache-sh4.c +++ b/arch/sh/mm/cache-sh4.c @@ -2,7 +2,7 @@ * arch/sh/mm/cache-sh4.c * * Copyright (C) 1999, 2000, 2002 Niibe Yutaka - * Copyright (C) 2001 - 2007 Paul Mundt + * Copyright (C) 2001 - 2009 Paul Mundt * Copyright (C) 2003 Richard Curnow * Copyright (c) 2007 STMicroelectronics (R&D) Ltd. * @@ -15,6 +15,8 @@ #include #include #include +#include +#include #include #include @@ -23,7 +25,6 @@ * flushing. Anything exceeding this will simply flush the dcache in its * entirety. */ -#define MAX_DCACHE_PAGES 64 /* XXX: Tune for ways */ #define MAX_ICACHE_PAGES 32 static void __flush_cache_4096(unsigned long addr, unsigned long phys, @@ -209,44 +210,64 @@ static void sh4_flush_cache_page(void *args) { struct flusher_data *data = args; struct vm_area_struct *vma; + struct page *page; unsigned long address, pfn, phys; - unsigned int alias_mask; + int map_coherent = 0; + pgd_t *pgd; + pud_t *pud; + pmd_t *pmd; + pte_t *pte; + void *vaddr; vma = data->vma; address = data->addr1; pfn = data->addr2; phys = pfn << PAGE_SHIFT; + page = pfn_to_page(pfn); if (cpu_context(smp_processor_id(), vma->vm_mm) == NO_CONTEXT) return; - alias_mask = boot_cpu_data.dcache.alias_mask; - - /* We only need to flush D-cache when we have alias */ - if ((address^phys) & alias_mask) { - /* Loop 4K of the D-cache */ - flush_cache_4096( - CACHE_OC_ADDRESS_ARRAY | (address & alias_mask), - phys); - /* Loop another 4K of the D-cache */ - flush_cache_4096( - CACHE_OC_ADDRESS_ARRAY | (phys & alias_mask), - phys); - } + address &= PAGE_MASK; + pgd = pgd_offset(vma->vm_mm, address); + pud = pud_offset(pgd, address); + pmd = pmd_offset(pud, address); + pte = pte_offset_kernel(pmd, address); + + /* If the page isn't present, there is nothing to do here. */ + if (!(pte_val(*pte) & _PAGE_PRESENT)) + return; - alias_mask = boot_cpu_data.icache.alias_mask; - if (vma->vm_flags & VM_EXEC) { + if ((vma->vm_mm == current->active_mm)) + vaddr = NULL; + else { /* - * Evict entries from the portion of the cache from which code - * may have been executed at this address (virtual). There's - * no need to evict from the portion corresponding to the - * physical address as for the D-cache, because we know the - * kernel has never executed the code through its identity - * translation. + * Use kmap_coherent or kmap_atomic to do flushes for + * another ASID than the current one. */ - flush_cache_4096( - CACHE_IC_ADDRESS_ARRAY | (address & alias_mask), - phys); + map_coherent = (current_cpu_data.dcache.n_aliases && + !test_bit(PG_dcache_dirty, &page->flags) && + page_mapped(page)); + if (map_coherent) + vaddr = kmap_coherent(page, address); + else + vaddr = kmap_atomic(page, KM_USER0); + + address = (unsigned long)vaddr; + } + + if (pages_do_alias(address, phys)) + flush_cache_4096(CACHE_OC_ADDRESS_ARRAY | + (address & shm_align_mask), phys); + + if (vma->vm_flags & VM_EXEC) + flush_icache_all(); + + if (vaddr) { + if (map_coherent) + kunmap_coherent(vaddr); + else + kunmap_atomic(vaddr, KM_USER0); } } -- cgit v1.2.3 From c4845a4b2288a9e5d96a0558e474809028c8aff3 Mon Sep 17 00:00:00 2001 From: Paul Mundt Date: Wed, 9 Sep 2009 17:13:07 +0900 Subject: sh: Fix up redundant cache flushing for PAGE_SIZE > 4k. If PAGE_SIZE is presently over 4k we do a lot of extra flushing given that we purge the cache 4k at a time. Make it explicitly 4k per iteration, rather than iterating for PAGE_SIZE before looping over again. Signed-off-by: Paul Mundt --- arch/sh/mm/cache-sh4.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/sh/mm/cache-sh4.c b/arch/sh/mm/cache-sh4.c index e3fbd99b323..8362d312ad9 100644 --- a/arch/sh/mm/cache-sh4.c +++ b/arch/sh/mm/cache-sh4.c @@ -357,7 +357,7 @@ static void __flush_cache_4096(unsigned long addr, unsigned long phys, * pointless nead-of-loop check for 0 iterations. */ do { - ea = base_addr + PAGE_SIZE; + ea = base_addr + 4096; a = base_addr; p = phys; -- cgit v1.2.3 From f9e2bdfdbb4c9da13422b349227be8c7b41dbd44 Mon Sep 17 00:00:00 2001 From: Paul Mundt Date: Wed, 9 Sep 2009 17:14:19 +0900 Subject: sh: Factor in cpu id for selection of cache colour fixmap. In the SMP VIPT case the page copy/clear ops still perform colouring, care needs to be taken that CPUs don't end up stepping on each other, so we give them a bit of room to work with. At the same time, we reduce the worst-case colouring given that these pages are always consumed. Signed-off-by: Paul Mundt --- arch/sh/include/asm/fixmap.h | 6 +++--- arch/sh/mm/kmap.c | 4 +++- 2 files changed, 6 insertions(+), 4 deletions(-) (limited to 'arch') diff --git a/arch/sh/include/asm/fixmap.h b/arch/sh/include/asm/fixmap.h index 721fcc4d5e9..76c5a3099cb 100644 --- a/arch/sh/include/asm/fixmap.h +++ b/arch/sh/include/asm/fixmap.h @@ -14,9 +14,9 @@ #define _ASM_FIXMAP_H #include +#include #include #ifdef CONFIG_HIGHMEM -#include #include #endif @@ -46,9 +46,9 @@ * fix-mapped? */ enum fixed_addresses { -#define FIX_N_COLOURS 16 +#define FIX_N_COLOURS 8 FIX_CMAP_BEGIN, - FIX_CMAP_END = FIX_CMAP_BEGIN + FIX_N_COLOURS, + FIX_CMAP_END = FIX_CMAP_BEGIN + (FIX_N_COLOURS * NR_CPUS), FIX_UNCACHED, #ifdef CONFIG_HIGHMEM FIX_KMAP_BEGIN, /* reserved pte's for temporary kernel mappings */ diff --git a/arch/sh/mm/kmap.c b/arch/sh/mm/kmap.c index 16e01b5fed0..15d74ea4209 100644 --- a/arch/sh/mm/kmap.c +++ b/arch/sh/mm/kmap.c @@ -39,7 +39,9 @@ void *kmap_coherent(struct page *page, unsigned long addr) pagefault_disable(); idx = FIX_CMAP_END - - ((addr & current_cpu_data.dcache.alias_mask) >> PAGE_SHIFT); + (((addr >> PAGE_SHIFT) & (FIX_N_COLOURS - 1)) + + (FIX_N_COLOURS * smp_processor_id())); + vaddr = __fix_to_virt(idx); BUG_ON(!pte_none(*(kmap_coherent_pte - idx))); -- cgit v1.2.3 From a6f05a6a0a1713d5b019f096799d49226807d3df Mon Sep 17 00:00:00 2001 From: "H. Peter Anvin" Date: Thu, 8 Oct 2009 18:02:54 -0700 Subject: x86-64: make compat_start_thread() match start_thread() For no real good reason, compat_start_thread() was embedded inline in whereas the native start_thread() lives in process_*.c. Move compat_start_thread() to process_64.c, remove gratuitious differences, and fix a few items which mostly look like bit rot. In particular, compat_start_thread() didn't do free_thread_xstate(), which means it was hanging on to the xstate store area even when it was not needed. It was also not setting old_rsp, but it looks like that generally shouldn't matter for a 32-bit process. Note: compat_start_thread *has* to be a macro, since it is tested with start_thread_ia32() as the out of line function name. Signed-off-by: H. Peter Anvin Acked-by: Suresh Siddha --- arch/x86/include/asm/elf.h | 20 ++------------------ arch/x86/kernel/process_64.c | 23 ++++++++++++++++++++++- 2 files changed, 24 insertions(+), 19 deletions(-) (limited to 'arch') diff --git a/arch/x86/include/asm/elf.h b/arch/x86/include/asm/elf.h index 456a304b817..8a024babe5e 100644 --- a/arch/x86/include/asm/elf.h +++ b/arch/x86/include/asm/elf.h @@ -157,19 +157,6 @@ do { \ #define compat_elf_check_arch(x) elf_check_arch_ia32(x) -static inline void start_ia32_thread(struct pt_regs *regs, u32 ip, u32 sp) -{ - loadsegment(fs, 0); - loadsegment(ds, __USER32_DS); - loadsegment(es, __USER32_DS); - load_gs_index(0); - regs->ip = ip; - regs->sp = sp; - regs->flags = X86_EFLAGS_IF; - regs->cs = __USER32_CS; - regs->ss = __USER32_DS; -} - static inline void elf_common_init(struct thread_struct *t, struct pt_regs *regs, const u16 ds) { @@ -191,11 +178,8 @@ do { \ #define COMPAT_ELF_PLAT_INIT(regs, load_addr) \ elf_common_init(¤t->thread, regs, __USER_DS) -#define compat_start_thread(regs, ip, sp) \ -do { \ - start_ia32_thread(regs, ip, sp); \ - set_fs(USER_DS); \ -} while (0) +void start_thread_ia32(struct pt_regs *regs, u32 new_ip, u32 new_sp); +#define compat_start_thread start_thread_ia32 #define COMPAT_SET_PERSONALITY(ex) \ do { \ diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c index ad535b68317..7cf0a6b6d4b 100644 --- a/arch/x86/kernel/process_64.c +++ b/arch/x86/kernel/process_64.c @@ -356,7 +356,7 @@ start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp) percpu_write(old_rsp, new_sp); regs->cs = __USER_CS; regs->ss = __USER_DS; - regs->flags = 0x200; + regs->flags = X86_EFLAGS_IF; set_fs(USER_DS); /* * Free the old FP and other extended state @@ -365,6 +365,27 @@ start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp) } EXPORT_SYMBOL_GPL(start_thread); +#ifdef CONFIG_IA32_EMULATION +void start_thread_ia32(struct pt_regs *regs, u32 new_ip, u32 new_sp) +{ + loadsegment(fs, 0); + loadsegment(ds, __USER32_DS); + loadsegment(es, __USER32_DS); + load_gs_index(0); + regs->ip = new_ip; + regs->sp = new_sp; + percpu_write(old_rsp, new_sp); + regs->cs = __USER32_CS; + regs->ss = __USER32_DS; + regs->flags = X86_EFLAGS_IF; + set_fs(USER_DS); + /* + * Free the old FP and other extended state + */ + free_thread_xstate(current); +} +#endif + /* * switch_to(x,y) should switch tasks from x to y. * -- cgit v1.2.3 From e634d8fc792c66c3d4ff45518c04848c1e28f221 Mon Sep 17 00:00:00 2001 From: "H. Peter Anvin" Date: Fri, 9 Oct 2009 15:56:53 -0700 Subject: x86-64: merge the standard and compat start_thread() functions The only thing left that differs between the standard and compat start_thread functions is the actual segment numbers and the prototype, so have a single common function which contains the guts and two very small wrappers. Signed-off-by: H. Peter Anvin Acked-by: Suresh Siddha --- arch/x86/kernel/process_64.c | 39 +++++++++++++++++---------------------- 1 file changed, 17 insertions(+), 22 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c index 7cf0a6b6d4b..eb261c582a4 100644 --- a/arch/x86/kernel/process_64.c +++ b/arch/x86/kernel/process_64.c @@ -344,18 +344,20 @@ out: return err; } -void -start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp) +static void +start_thread_common(struct pt_regs *regs, unsigned long new_ip, + unsigned long new_sp, + unsigned int _cs, unsigned int _ss, unsigned int _ds) { loadsegment(fs, 0); - loadsegment(es, 0); - loadsegment(ds, 0); + loadsegment(es, _ds); + loadsegment(ds, _ds); load_gs_index(0); regs->ip = new_ip; regs->sp = new_sp; percpu_write(old_rsp, new_sp); - regs->cs = __USER_CS; - regs->ss = __USER_DS; + regs->cs = _cs; + regs->ss = _ss; regs->flags = X86_EFLAGS_IF; set_fs(USER_DS); /* @@ -363,26 +365,19 @@ start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp) */ free_thread_xstate(current); } -EXPORT_SYMBOL_GPL(start_thread); + +void +start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp) +{ + start_thread_common(regs, new_ip, new_sp, + __USER_CS, __USER_DS, 0); +} #ifdef CONFIG_IA32_EMULATION void start_thread_ia32(struct pt_regs *regs, u32 new_ip, u32 new_sp) { - loadsegment(fs, 0); - loadsegment(ds, __USER32_DS); - loadsegment(es, __USER32_DS); - load_gs_index(0); - regs->ip = new_ip; - regs->sp = new_sp; - percpu_write(old_rsp, new_sp); - regs->cs = __USER32_CS; - regs->ss = __USER32_DS; - regs->flags = X86_EFLAGS_IF; - set_fs(USER_DS); - /* - * Free the old FP and other extended state - */ - free_thread_xstate(current); + start_thread_common(regs, new_ip, new_sp, + __USER32_CS, __USER32_DS, __USER32_DS); } #endif -- cgit v1.2.3 From a469f627c15de2af392be23508e6094d7268e2b7 Mon Sep 17 00:00:00 2001 From: Guennadi Liakhovetski Date: Sat, 3 Oct 2009 11:21:30 +0000 Subject: SH: add support for the RJ54N1CB0C camera for the kfr2r09 platform Signed-off-by: Guennadi Liakhovetski Signed-off-by: Paul Mundt --- arch/sh/boards/mach-kfr2r09/setup.c | 139 ++++++++++++++++++++++++++++++++++++ 1 file changed, 139 insertions(+) (limited to 'arch') diff --git a/arch/sh/boards/mach-kfr2r09/setup.c b/arch/sh/boards/mach-kfr2r09/setup.c index c08d33fe210..ce01d6a953b 100644 --- a/arch/sh/boards/mach-kfr2r09/setup.c +++ b/arch/sh/boards/mach-kfr2r09/setup.c @@ -18,6 +18,8 @@ #include #include #include +#include +#include #include