diff options
Diffstat (limited to 'arch/x86')
53 files changed, 1313 insertions, 662 deletions
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig index 738bdc6b0f8..1d9c18aa17e 100644 --- a/arch/x86/Kconfig +++ b/arch/x86/Kconfig @@ -24,6 +24,7 @@ config X86 select HAVE_UNSTABLE_SCHED_CLOCK select HAVE_IDE select HAVE_OPROFILE + select HAVE_PERF_COUNTERS if (!M386 && !M486) select HAVE_IOREMAP_PROT select HAVE_KPROBES select ARCH_WANT_OPTIONAL_GPIOLIB @@ -585,7 +586,6 @@ config GART_IOMMU bool "GART IOMMU support" if EMBEDDED default y select SWIOTLB - select AGP depends on X86_64 && PCI ---help--- Support for full DMA access of devices with 32bit memory access only @@ -742,7 +742,6 @@ config X86_UP_IOAPIC config X86_LOCAL_APIC def_bool y depends on X86_64 || SMP || X86_32_NON_STANDARD || X86_UP_APIC - select HAVE_PERF_COUNTERS if (!M386 && !M486) config X86_IO_APIC def_bool y diff --git a/arch/x86/boot/compressed/Makefile b/arch/x86/boot/compressed/Makefile index e2ff504b4dd..f8ed0658404 100644 --- a/arch/x86/boot/compressed/Makefile +++ b/arch/x86/boot/compressed/Makefile @@ -4,7 +4,7 @@ # create a compressed vmlinux image from the original vmlinux # -targets := vmlinux vmlinux.bin vmlinux.bin.gz vmlinux.bin.bz2 vmlinux.bin.lzma head_$(BITS).o misc.o piggy.o +targets := vmlinux.lds vmlinux vmlinux.bin vmlinux.bin.gz vmlinux.bin.bz2 vmlinux.bin.lzma head_$(BITS).o misc.o piggy.o KBUILD_CFLAGS := -m$(BITS) -D__KERNEL__ $(LINUX_INCLUDE) -O2 KBUILD_CFLAGS += -fno-strict-aliasing -fPIC diff --git a/arch/x86/include/asm/amd_iommu.h b/arch/x86/include/asm/amd_iommu.h index bdf96f119f0..ac95995b7ba 100644 --- a/arch/x86/include/asm/amd_iommu.h +++ b/arch/x86/include/asm/amd_iommu.h @@ -25,6 +25,7 @@ #ifdef CONFIG_AMD_IOMMU extern int amd_iommu_init(void); extern int amd_iommu_init_dma_ops(void); +extern int amd_iommu_init_passthrough(void); extern void amd_iommu_detect(void); extern irqreturn_t amd_iommu_int_handler(int irq, void *data); extern void amd_iommu_flush_all_domains(void); diff --git a/arch/x86/include/asm/amd_iommu_types.h b/arch/x86/include/asm/amd_iommu_types.h index 0c878caaa0a..2a2cc7a78a8 100644 --- a/arch/x86/include/asm/amd_iommu_types.h +++ b/arch/x86/include/asm/amd_iommu_types.h @@ -143,22 +143,29 @@ #define EVT_BUFFER_SIZE 8192 /* 512 entries */ #define EVT_LEN_MASK (0x9ULL << 56) +#define PAGE_MODE_NONE 0x00 #define PAGE_MODE_1_LEVEL 0x01 #define PAGE_MODE_2_LEVEL 0x02 #define PAGE_MODE_3_LEVEL 0x03 - -#define IOMMU_PDE_NL_0 0x000ULL -#define IOMMU_PDE_NL_1 0x200ULL -#define IOMMU_PDE_NL_2 0x400ULL -#define IOMMU_PDE_NL_3 0x600ULL - -#define IOMMU_PTE_L2_INDEX(address) (((address) >> 30) & 0x1ffULL) -#define IOMMU_PTE_L1_INDEX(address) (((address) >> 21) & 0x1ffULL) -#define IOMMU_PTE_L0_INDEX(address) (((address) >> 12) & 0x1ffULL) - -#define IOMMU_MAP_SIZE_L1 (1ULL << 21) -#define IOMMU_MAP_SIZE_L2 (1ULL << 30) -#define IOMMU_MAP_SIZE_L3 (1ULL << 39) +#define PAGE_MODE_4_LEVEL 0x04 +#define PAGE_MODE_5_LEVEL 0x05 +#define PAGE_MODE_6_LEVEL 0x06 + +#define PM_LEVEL_SHIFT(x) (12 + ((x) * 9)) +#define PM_LEVEL_SIZE(x) (((x) < 6) ? \ + ((1ULL << PM_LEVEL_SHIFT((x))) - 1): \ + (0xffffffffffffffffULL)) +#define PM_LEVEL_INDEX(x, a) (((a) >> PM_LEVEL_SHIFT((x))) & 0x1ffULL) +#define PM_LEVEL_ENC(x) (((x) << 9) & 0xe00ULL) +#define PM_LEVEL_PDE(x, a) ((a) | PM_LEVEL_ENC((x)) | \ + IOMMU_PTE_P | IOMMU_PTE_IR | IOMMU_PTE_IW) +#define PM_PTE_LEVEL(pte) (((pte) >> 9) & 0x7ULL) + +#define PM_MAP_4k 0 +#define PM_ADDR_MASK 0x000ffffffffff000ULL +#define PM_MAP_MASK(lvl) (PM_ADDR_MASK & \ + (~((1ULL << (12 + ((lvl) * 9))) - 1))) +#define PM_ALIGNED(lvl, addr) ((PM_MAP_MASK(lvl) & (addr)) == (addr)) #define IOMMU_PTE_P (1ULL << 0) #define IOMMU_PTE_TV (1ULL << 1) @@ -167,11 +174,6 @@ #define IOMMU_PTE_IR (1ULL << 61) #define IOMMU_PTE_IW (1ULL << 62) -#define IOMMU_L1_PDE(address) \ - ((address) | IOMMU_PDE_NL_1 | IOMMU_PTE_P | IOMMU_PTE_IR | IOMMU_PTE_IW) -#define IOMMU_L2_PDE(address) \ - ((address) | IOMMU_PDE_NL_2 | IOMMU_PTE_P | IOMMU_PTE_IR | IOMMU_PTE_IW) - #define IOMMU_PAGE_MASK (((1ULL << 52) - 1) & ~0xfffULL) #define IOMMU_PTE_PRESENT(pte) ((pte) & IOMMU_PTE_P) #define IOMMU_PTE_PAGE(pte) (phys_to_virt((pte) & IOMMU_PAGE_MASK)) @@ -194,11 +196,14 @@ #define PD_DMA_OPS_MASK (1UL << 0) /* domain used for dma_ops */ #define PD_DEFAULT_MASK (1UL << 1) /* domain is a default dma_ops domain for an IOMMU */ +#define PD_PASSTHROUGH_MASK (1UL << 2) /* domain has no page + translation */ + extern bool amd_iommu_dump; #define DUMP_printk(format, arg...) \ do { \ if (amd_iommu_dump) \ - printk(KERN_INFO "AMD IOMMU: " format, ## arg); \ + printk(KERN_INFO "AMD-Vi: " format, ## arg); \ } while(0); /* @@ -226,6 +231,7 @@ struct protection_domain { int mode; /* paging mode (0-6 levels) */ u64 *pt_root; /* page table root pointer */ unsigned long flags; /* flags to find out type of domain */ + bool updated; /* complete domain flush required */ unsigned dev_cnt; /* devices assigned to this domain */ void *priv; /* private data */ }; @@ -337,6 +343,9 @@ struct amd_iommu { /* if one, we need to send a completion wait command */ bool need_sync; + /* becomes true if a command buffer reset is running */ + bool reset_in_progress; + /* default dma_ops domain for that IOMMU */ struct dma_ops_domain *default_dom; }; @@ -457,4 +466,7 @@ static inline void amd_iommu_stats_init(void) { } #endif /* CONFIG_AMD_IOMMU_STATS */ +/* some function prototypes */ +extern void amd_iommu_reset_cmd_buffer(struct amd_iommu *iommu); + #endif /* _ASM_X86_AMD_IOMMU_TYPES_H */ diff --git a/arch/x86/include/asm/efi.h b/arch/x86/include/asm/efi.h index edc90f23e70..8406ed7f992 100644 --- a/arch/x86/include/asm/efi.h +++ b/arch/x86/include/asm/efi.h @@ -33,7 +33,7 @@ extern unsigned long asmlinkage efi_call_phys(void *, ...); #define efi_call_virt6(f, a1, a2, a3, a4, a5, a6) \ efi_call_virt(f, a1, a2, a3, a4, a5, a6) -#define efi_ioremap(addr, size) ioremap_cache(addr, size) +#define efi_ioremap(addr, size, type) ioremap_cache(addr, size) #else /* !CONFIG_X86_32 */ @@ -84,7 +84,8 @@ extern u64 efi_call6(void *fp, u64 arg1, u64 arg2, u64 arg3, efi_call6((void *)(efi.systab->runtime->f), (u64)(a1), (u64)(a2), \ (u64)(a3), (u64)(a4), (u64)(a5), (u64)(a6)) -extern void __iomem *efi_ioremap(unsigned long addr, unsigned long size); +extern void __iomem *efi_ioremap(unsigned long addr, unsigned long size, + u32 type); #endif /* CONFIG_X86_32 */ diff --git a/arch/x86/include/asm/irqflags.h b/arch/x86/include/asm/irqflags.h index 2bdab21f089..c6ccbe7e81a 100644 --- a/arch/x86/include/asm/irqflags.h +++ b/arch/x86/include/asm/irqflags.h @@ -12,9 +12,15 @@ static inline unsigned long native_save_fl(void) { unsigned long flags; + /* + * Note: this needs to be "=r" not "=rm", because we have the + * stack offset from what gcc expects at the time the "pop" is + * executed, and so a memory reference with respect to the stack + * would end up using the wrong address. + */ asm volatile("# __raw_save_flags\n\t" "pushf ; pop %0" - : "=g" (flags) + : "=r" (flags) : /* no input */ : "memory"); diff --git a/arch/x86/include/asm/lguest.h b/arch/x86/include/asm/lguest.h index 313389cd50d..5136dad57cb 100644 --- a/arch/x86/include/asm/lguest.h +++ b/arch/x86/include/asm/lguest.h @@ -17,8 +17,7 @@ /* Pages for switcher itself, then two pages per cpu */ #define TOTAL_SWITCHER_PAGES (SHARED_SWITCHER_PAGES + 2 * nr_cpu_ids) -/* We map at -4M (-2M when PAE is activated) for ease of mapping - * into the guest (one PTE page). */ +/* We map at -4M (-2M for PAE) for ease of mapping (one PTE page). */ #ifdef CONFIG_X86_PAE #define SWITCHER_ADDR 0xFFE00000 #else diff --git a/arch/x86/include/asm/lguest_hcall.h b/arch/x86/include/asm/lguest_hcall.h index 33600a66755..ba0eed8aa1a 100644 --- a/arch/x86/include/asm/lguest_hcall.h +++ b/arch/x86/include/asm/lguest_hcall.h @@ -30,27 +30,27 @@ #include <asm/hw_irq.h> #include <asm/kvm_para.h> -/*G:030 But first, how does our Guest contact the Host to ask for privileged +/*G:030 + * But first, how does our Guest contact the Host to ask for privileged * operations? There are two ways: the direct way is to make a "hypercall", * to make requests of the Host Itself. * - * We use the KVM hypercall mechanism. Seventeen hypercalls are - * available: the hypercall number is put in the %eax register, and the - * arguments (when required) are placed in %ebx, %ecx, %edx and %esi. - * If a return value makes sense, it's returned in %eax. + * We use the KVM hypercall mechanism, though completely different hypercall + * numbers. Seventeen hypercalls are available: the hypercall number is put in + * the %eax register, and the arguments (when required) are placed in %ebx, + * %ecx, %edx and %esi. If a return value makes sense, it's returned in %eax. * * Grossly invalid calls result in Sudden Death at the hands of the vengeful * Host, rather than returning failure. This reflects Winston Churchill's - * definition of a gentleman: "someone who is only rude intentionally". */ -/*:*/ + * definition of a gentleman: "someone who is only rude intentionally". +:*/ /* Can't use our min() macro here: needs to be a constant */ #define LGUEST_IRQS (NR_IRQS < 32 ? NR_IRQS: 32) #define LHCALL_RING_SIZE 64 struct hcall_args { - /* These map directly onto eax, ebx, ecx, edx and esi - * in struct lguest_regs */ + /* These map directly onto eax/ebx/ecx/edx/esi in struct lguest_regs */ unsigned long arg0, arg1, arg2, arg3, arg4; }; diff --git a/arch/x86/include/asm/pgtable.h b/arch/x86/include/asm/pgtable.h index 3cc06e3fceb..16748077559 100644 --- a/arch/x86/include/asm/pgtable.h +++ b/arch/x86/include/asm/pgtable.h @@ -2,6 +2,7 @@ #define _ASM_X86_PGTABLE_H #include <asm/page.h> +#include <asm/e820.h> #include <asm/pgtable_types.h> @@ -269,10 +270,17 @@ static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot) #define canon_pgprot(p) __pgprot(massage_pgprot(p)) -static inline int is_new_memtype_allowed(unsigned long flags, - unsigned long new_flags) +static inline int is_new_memtype_allowed(u64 paddr, unsigned long size, + unsigned long flags, + unsigned long new_flags) { /* + * PAT type is always WB for ISA. So no need to check. + */ + if (is_ISA_range(paddr, paddr + size - 1)) + return 1; + + /* * Certain new memtypes are not allowed with certain * requested memtype: * - request is uncached, return cannot be write-back diff --git a/arch/x86/include/asm/uv/uv_bau.h b/arch/x86/include/asm/uv/uv_bau.h index bddd44f2f0a..80e2984f521 100644 --- a/arch/x86/include/asm/uv/uv_bau.h +++ b/arch/x86/include/asm/uv/uv_bau.h @@ -133,7 +133,7 @@ struct bau_msg_payload { * see table 4.2.3.0.1 in broacast_assist spec. */ struct bau_msg_header { - unsigned int dest_subnodeid:6; /* must be zero */ + unsigned int dest_subnodeid:6; /* must be 0x10, for the LB */ /* bits 5:0 */ unsigned int base_dest_nodeid:15; /* nasid>>1 (pnode) of */ /* bits 20:6 */ /* first bit in node_map */ diff --git a/arch/x86/include/asm/uv/uv_hub.h b/arch/x86/include/asm/uv/uv_hub.h index 341070f7ad5..77a68505419 100644 --- a/arch/x86/include/asm/uv/uv_hub.h +++ b/arch/x86/include/asm/uv/uv_hub.h @@ -175,7 +175,7 @@ DECLARE_PER_CPU(struct uv_hub_info_s, __uv_hub_info); #define UV_GLOBAL_MMR32_PNODE_BITS(p) ((p) << (UV_GLOBAL_MMR32_PNODE_SHIFT)) #define UV_GLOBAL_MMR64_PNODE_BITS(p) \ - ((unsigned long)(UV_PNODE_TO_GNODE(p)) << UV_GLOBAL_MMR64_PNODE_SHIFT) + (((unsigned long)(p)) << UV_GLOBAL_MMR64_PNODE_SHIFT) #define UV_APIC_PNODE_SHIFT 6 @@ -327,6 +327,7 @@ struct uv_blade_info { unsigned short nr_possible_cpus; unsigned short nr_online_cpus; unsigned short pnode; + short memory_nid; }; extern struct uv_blade_info *uv_blade_info; extern short *uv_node_to_blade; @@ -363,6 +364,12 @@ static inline int uv_blade_to_pnode(int bid) return uv_blade_info[bid].pnode; } +/* Nid of memory node on blade. -1 if no blade-local memory */ +static inline int uv_blade_to_memory_nid(int bid) +{ + return uv_blade_info[bid].memory_nid; +} + /* Determine the number of possible cpus on a blade */ static inline int uv_blade_nr_possible_cpus(int bid) { diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c index 6c99f503780..98f230f6a28 100644 --- a/arch/x86/kernel/amd_iommu.c +++ b/arch/x86/kernel/amd_iommu.c @@ -41,9 +41,13 @@ static DEFINE_RWLOCK(amd_iommu_devtable_lock); static LIST_HEAD(iommu_pd_list); static DEFINE_SPINLOCK(iommu_pd_list_lock); -#ifdef CONFIG_IOMMU_API +/* + * Domain for untranslated devices - only allocated + * if iommu=pt passed on kernel cmd line. + */ +static struct protection_domain *pt_domain; + static struct iommu_ops amd_iommu_ops; -#endif /* * general struct to manage commands send to an IOMMU @@ -55,16 +59,16 @@ struct iommu_cmd { static int dma_ops_unity_map(struct dma_ops_domain *dma_dom, struct unity_map_entry *e); static struct dma_ops_domain *find_protection_domain(u16 devid); -static u64* alloc_pte(struct protection_domain *dom, - unsigned long address, u64 - **pte_page, gfp_t gfp); +static u64 *alloc_pte(struct protection_domain *domain, + unsigned long address, int end_lvl, + u64 **pte_page, gfp_t gfp); static void dma_ops_reserve_addresses(struct dma_ops_domain *dom, unsigned long start_page, unsigned int pages); - -#ifndef BUS_NOTIFY_UNBOUND_DRIVER -#define BUS_NOTIFY_UNBOUND_DRIVER 0x0005 -#endif +static void reset_iommu_command_buffer(struct amd_iommu *iommu); +static u64 *fetch_pte(struct protection_domain *domain, + unsigned long address, int map_size); +static void update_domain(struct protection_domain *domain); #ifdef CONFIG_AMD_IOMMU_STATS @@ -138,7 +142,25 @@ static int iommu_has_npcache(struct amd_iommu *iommu) * ****************************************************************************/ -static void iommu_print_event(void *__evt) +static void dump_dte_entry(u16 devid) +{ + int i; + + for (i = 0; i < 8; ++i) + pr_err("AMD-Vi: DTE[%d]: %08x\n", i, + amd_iommu_dev_table[devid].data[i]); +} + +static void dump_command(unsigned long phys_addr) +{ + struct iommu_cmd *cmd = phys_to_virt(phys_addr); + int i; + + for (i = 0; i < 4; ++i) + pr_err("AMD-Vi: CMD[%d]: %08x\n", i, cmd->data[i]); +} + +static void iommu_print_event(struct amd_iommu *iommu, void *__evt) { u32 *event = __evt; int type = (event[1] >> EVENT_TYPE_SHIFT) & EVENT_TYPE_MASK; @@ -147,7 +169,7 @@ static void iommu_print_event(void *__evt) int flags = (event[1] >> EVENT_FLAGS_SHIFT) & EVENT_FLAGS_MASK; u64 address = (u64)(((u64)event[3]) << 32) | event[2]; - printk(KERN_ERR "AMD IOMMU: Event logged ["); + printk(KERN_ERR "AMD-Vi: Event logged ["); switch (type) { case EVENT_TYPE_ILL_DEV: @@ -155,6 +177,7 @@ static void iommu_print_event(void *__evt) "address=0x%016llx flags=0x%04x]\n", PCI_BUS(devid), PCI_SLOT(devid), PCI_FUNC(devid), address, flags); + dump_dte_entry(devid); break; case EVENT_TYPE_IO_FAULT: printk("IO_PAGE_FAULT device=%02x:%02x.%x " @@ -176,6 +199,8 @@ static void iommu_print_event(void *__evt) break; case EVENT_TYPE_ILL_CMD: printk("ILLEGAL_COMMAND_ERROR address=0x%016llx]\n", address); + reset_iommu_command_buffer(iommu); + dump_command(address); break; case EVENT_TYPE_CMD_HARD_ERR: printk("COMMAND_HARDWARE_ERROR address=0x%016llx " @@ -209,7 +234,7 @@ static void iommu_poll_events(struct amd_iommu *iommu) tail = readl(iommu->mmio_base + MMIO_EVT_TAIL_OFFSET); while (head != tail) { - iommu_print_event(iommu->evt_buf + head); + iommu_print_event(iommu, iommu->evt_buf + head); head = (head + EVENT_ENTRY_SIZE) % iommu->evt_buf_size; } @@ -296,8 +321,11 @@ static void __iommu_wait_for_completion(struct amd_iommu *iommu) status &= ~MMIO_STATUS_COM_WAIT_INT_MASK; writel(status, iommu->mmio_base + MMIO_STATUS_OFFSET); - if (unlikely(i == EXIT_LOOP_COUNT)) - panic("AMD IOMMU: Completion wait loop failed\n"); + if (unlikely(i == EXIT_LOOP_COUNT)) { + spin_unlock(&iommu->lock); + reset_iommu_command_buffer(iommu); + spin_lock(&iommu->lock); + } } /* @@ -445,47 +473,78 @@ static void iommu_flush_tlb_pde(struct amd_iommu *iommu, u16 domid) } /* + * This function flushes one domain on one IOMMU + */ +static void flush_domain_on_iommu(struct amd_iommu *iommu, u16 domid) +{ + struct iommu_cmd cmd; + unsigned long flags; + + __iommu_build_inv_iommu_pages(&cmd, CMD_INV_IOMMU_ALL_PAGES_ADDRESS, + domid, 1, 1); + + spin_lock_irqsave(&iommu->lock, flags); + __iommu_queue_command(iommu, &cmd); + __iommu_completion_wait(iommu); + __iommu_wait_for_completion(iommu); + spin_unlock_irqrestore(&iommu->lock, flags); +} + +static void flush_all_domains_on_iommu(struct amd_iommu *iommu) +{ + int i; + + for (i = 1; i < MAX_DOMAIN_ID; ++i) { + if (!test_bit(i, amd_iommu_pd_alloc_bitmap)) + continue; + flush_domain_on_iommu(iommu, i); + } + +} + +/* * This function is used to flush the IO/TLB for a given protection domain * on every IOMMU in the system */ static void iommu_flush_domain(u16 domid) { - unsigned long flags; struct amd_iommu *iommu; - struct iommu_cmd cmd; INC_STATS_COUNTER(domain_flush_all); - __iommu_build_inv_iommu_pages(&cmd, CMD_INV_IOMMU_ALL_PAGES_ADDRESS, - domid, 1, 1); - - for_each_iommu(iommu) { - spin_lock_irqsave(&iommu->lock, flags); - __iommu_queue_command(iommu, &cmd); - __iommu_completion_wait(iommu); - __iommu_wait_for_completion(iommu); - spin_unlock_irqrestore(&iommu->lock, flags); - } + for_each_iommu(iommu) + flush_domain_on_iommu(iommu, domid); } void amd_iommu_flush_all_domains(void) { + struct amd_iommu *iommu; + + for_each_iommu(iommu) + flush_all_domains_on_iommu(iommu); +} + +static void flush_all_devices_for_iommu(struct amd_iommu *iommu) +{ int i; - for (i = 1; i < MAX_DOMAIN_ID; ++i) { - if (!test_bit(i, amd_iommu_pd_alloc_bitmap)) + for (i = 0; i <= amd_iommu_last_bdf; ++i) { + if (iommu != amd_iommu_rlookup_table[i]) continue; - iommu_flush_domain(i); + + iommu_queue_inv_dev_entry(iommu, i); + iommu_completion_wait(iommu); } } -void amd_iommu_flush_all_devices(void) +static void flush_devices_by_domain(struct protection_domain *domain) { struct amd_iommu *iommu; int i; for (i = 0; i <= amd_iommu_last_bdf; ++i) { - if (amd_iommu_pd_table[i] == NULL) + if ((domain == NULL && amd_iommu_pd_table[i] == NULL) || + (amd_iommu_pd_table[i] != domain)) continue; iommu = amd_iommu_rlookup_table[i]; @@ -497,6 +556,27 @@ void amd_iommu_flush_all_devices(void) } } +static void reset_iommu_command_buffer(struct amd_iommu *iommu) +{ + pr_err("AMD-Vi: Resetting IOMMU command buffer\n"); + + if (iommu->reset_in_progress) + panic("AMD-Vi: ILLEGAL_COMMAND_ERROR while resetting command buffer\n"); + + iommu->reset_in_progress = true; + + amd_iommu_reset_cmd_buffer(iommu); + flush_all_devices_for_iommu(iommu); + flush_all_domains_on_iommu(iommu); + + iommu->reset_in_progress = false; +} + +void amd_iommu_flush_all_devices(void) +{ + flush_devices_by_domain(NULL); +} + /**************************************************************************** * * The functions below are used the create the page table mappings for @@ -514,18 +594,21 @@ void amd_iommu_flush_all_devices(void) static int iommu_map_page(struct protection_domain *dom, unsigned long bus_addr, unsigned long phys_addr, - int prot) + int prot, + int map_size) { u64 __pte, *pte; bus_addr = PAGE_ALIGN(bus_addr); phys_addr = PAGE_ALIGN(phys_addr); - /* only support 512GB address spaces for now */ - if (bus_addr > IOMMU_MAP_SIZE_L3 || !(prot & IOMMU_PROT_MASK)) + BUG_ON(!PM_ALIGNED(map_size, bus_addr)); + BUG_ON(!PM_ALIGNED(map_size, phys_addr)); + + if (!(prot & IOMMU_PROT_MASK)) return -EINVAL; - pte = alloc_pte(dom, bus_addr, NULL, GFP_KERNEL); + pte = alloc_pte(dom, bus_addr, map_size, NULL, GFP_KERNEL); if (IOMMU_PTE_PRESENT(*pte)) return -EBUSY; @@ -538,29 +621,18 @@ static int iommu_map_page(struct protection_domain *dom, *pte = __pte; + update_domain(dom); + return 0; } static void iommu_unmap_page(struct protection_domain *dom, - unsigned long bus_addr) + unsigned long bus_addr, int map_size) { - u64 *pte; - - pte = &dom->pt_root[IOMMU_PTE_L2_INDEX(bus_addr)]; - - if (!IOMMU_PTE_PRESENT(*pte)) - return; - - pte = IOMMU_PTE_PAGE(*pte); - pte = &pte[IOMMU_PTE_L1_INDEX(bus_addr)]; + u64 *pte = fetch_pte(dom, bus_addr, map_size); - if (!IOMMU_PTE_PRESENT(*pte)) - return; - - pte = IOMMU_PTE_PAGE(*pte); - pte = &pte[IOMMU_PTE_L1_INDEX(bus_addr)]; - - *pte = 0; + if (pte) + *pte = 0; } /* @@ -615,7 +687,8 @@ static int dma_ops_unity_map(struct dma_ops_domain *dma_dom, for (addr = e->address_start; addr < e->address_end; addr += PAGE_SIZE) { - ret = iommu_map_page(&dma_dom->domain, addr, addr, e->prot); + ret = iommu_map_page(&dma_dom->domain, addr, addr, e->prot, + PM_MAP_4k); if (ret) return ret; /* @@ -670,24 +743,29 @@ static int init_unity_mappings_for_device(struct dma_ops_domain *dma_dom, * This function checks if there is a PTE for a given dma address. If * there is one, it returns the pointer to it. */ -static u64* fetch_pte(struct protection_domain *domain, - unsigned long address) +static u64 *fetch_pte(struct protection_domain *domain, + unsigned long address, int map_size) { + int level; u64 *pte; - pte = &domain->pt_root[IOMMU_PTE_L2_INDEX(address)]; + level = domain->mode - 1; + pte = &domain->pt_root[PM_LEVEL_INDEX(level, address)]; - if (!IOMMU_PTE_PRESENT(*pte)) - return NULL; + while (level > map_size) { + if (!IOMMU_PTE_PRESENT(*pte)) + return NULL; - pte = IOMMU_PTE_PAGE(*pte); - pte = &pte[IOMMU_PTE_L1_INDEX(address)]; + level -= 1; - if (!IOMMU_PTE_PRESENT(*pte)) - return NULL; + pte = IOMMU_PTE_PAGE(*pte); + pte = &pte[PM_LEVEL_INDEX(level, address)]; - pte = IOMMU_PTE_PAGE(*pte); - pte = &pte[IOMMU_PTE_L0_INDEX(address)]; + if ((PM_PTE_LEVEL(*pte) == 0) && level != map_size) { + pte = NULL; + break; + } + } return pte; } @@ -727,7 +805,7 @@ static int alloc_new_range(struct amd_iommu *iommu, u64 *pte, *pte_page; for (i = 0; i < num_ptes; ++i) { - pte = alloc_pte(&dma_dom->domain, address, + pte = alloc_pte(&dma_dom->domain, address, PM_MAP_4k, &pte_page, gfp); if (!pte) goto out_free; @@ -760,16 +838,20 @@ static int alloc_new_range(struct amd_iommu *iommu, for (i = dma_dom->aperture[index]->offset; i < dma_dom->aperture_size; i += PAGE_SIZE) { - u64 *pte = fetch_pte(&dma_dom->domain, i); + u64 *pte = fetch_pte(&dma_dom->domain, i, PM_MAP_4k); if (!pte || !IOMMU_PTE_PRESENT(*pte)) continue; dma_ops_reserve_addresses(dma_dom, i << PAGE_SHIFT, 1); } + update_domain(&dma_dom->domain); + return 0; out_free: + update_domain(&dma_dom->domain); + free_page((unsigned long)dma_dom->aperture[index]->bitmap); kfree(dma_dom->aperture[index]); @@ -1009,7 +1091,7 @@ static struct dma_ops_domain *dma_ops_domain_alloc(struct amd_iommu *iommu) dma_dom->domain.id = domain_id_alloc(); if (dma_dom->domain.id == 0) goto free_dma_dom; - dma_dom->domain.mode = PAGE_MODE_3_LEVEL; + dma_dom->domain.mode = PAGE_MODE_2_LEVEL; dma_dom->domain.pt_root = (void *)get_zeroed_page(GFP_KERNEL); dma_dom->domain.flags = PD_DMA_OPS_MASK; dma_dom->domain.priv = dma_dom; @@ -1063,6 +1145,41 @@ static struct protection_domain *domain_for_device(u16 devid) return dom; } +static void set_dte_entry(u16 devid, struct protection_domain *domain) +{ + u64 pte_root = virt_to_phys(domain->pt_root); + + pte_root |= (domain->mode & DEV_ENTRY_MODE_MASK) + << DEV_ENTRY_MODE_SHIFT; + pte_root |= IOMMU_PTE_IR | IOMMU_PTE_IW | IOMMU_PTE_P | IOMMU_PTE_TV; + + amd_iommu_dev_table[devid].data[2] = domain->id; + amd_iommu_dev_table[devid].data[1] = upper_32_bits(pte_root); + amd_iommu_dev_table[devid].data[0] = lower_32_bits(pte_root); + + amd_iommu_pd_table[devid] = domain; +} + +/* + * If a device is not yet associated with a domain, this function does + * assigns it visible for the hardware + */ +static void __attach_device(struct amd_iommu *iommu, + struct protection_domain *domain, + u16 devid) +{ + /* lock domain */ + spin_lock(&domain->lock); + + /* update DTE entry */ + set_dte_entry(devid, domain); + + domain->dev_cnt += 1; + + /* ready */ + spin_unlock(&domain->lock); +} + /* * If a device is not yet associated with a domain, this function does * assigns it visible for the hardware @@ -1072,27 +1189,16 @@ static void attach_device(struct amd_iommu *iommu, u16 devid) { unsigned long flags; - u64 pte_root = virt_to_phys(domain->pt_root); - - domain->dev_cnt += 1; - - pte_root |= (domain->mode & DEV_ENTRY_MODE_MASK) - << DEV_ENTRY_MODE_SHIFT; - pte_root |= IOMMU_PTE_IR | IOMMU_PTE_IW | IOMMU_PTE_P | IOMMU_PTE_TV; write_lock_irqsave(&amd_iommu_devtable_lock, flags); - amd_iommu_dev_table[devid].data[0] = lower_32_bits(pte_root); - amd_iommu_dev_table[devid].data[1] = upper_32_bits(pte_root); - amd_iommu_dev_table[devid].data[2] = domain->id; - - amd_iommu_pd_table[devid] = domain; + __attach_device(iommu, domain, devid); write_unlock_irqrestore(&amd_iommu_devtable_lock, flags); - /* - * We might boot into a crash-kernel here. The crashed kernel - * left the caches in the IOMMU dirty. So we have to flush - * here to evict all dirty stuff. - */ + /* + * We might boot into a crash-kernel here. The crashed kernel + * left the caches in the IOMMU dirty. So we have to flush + * here to evict all dirty stuff. + */ iommu_queue_inv_dev_entry(iommu, devid); iommu_flush_tlb_pde(iommu, domain->id); } @@ -1119,6 +1225,15 @@ static void __detach_device(struct protection_domain *domain, u16 devid) /* ready */ spin_unlock(&domain->lock); + + /* + * If we run in passthrough mode the device must be assigned to the + * passthrough domain if it is detached from any other domain + */ + if (iommu_pass_through) { + struct amd_iommu *iommu = amd_iommu_rlookup_table[devid]; + __attach_device(iommu, pt_domain, devid); + } } /* @@ -1164,6 +1279,8 @@ static int device_change_notifier(struct notifier_block *nb, case BUS_NOTIFY_UNBOUND_DRIVER: if (!domain) goto out; + if (iommu_pass_through) + break; detach_device(domain, devid); break; case BUS_NOTIFY_ADD_DEVICE: @@ -1292,39 +1409,91 @@ static int get_device_resources(struct device *dev, return 1; } +static void update_device_table(struct protection_domain *domain) +{ + unsigned long flags; + int i; + + for (i = 0; i <= amd_iommu_last_bdf; ++i) { + if (amd_iommu_pd_table[i] != domain) + continue; + write_lock_irqsave(&amd_iommu_devtable_lock, flags); + set_dte_entry(i, domain); + write_unlock_irqrestore(&amd_iommu_devtable_lock, flags); + } +} + +static void update_domain(struct protection_domain *domain) +{ + if (!domain->updated) + return; + + update_device_table(domain); + flush_devices_by_domain(domain); + iommu_flush_domain(domain->id); + + domain->updated = false; +} + /* - * If the pte_page is not yet allocated this function is called + * This function is used to add another level to an IO page table. Adding + * another level increases the size of the address space by 9 bits to a size up + * to 64 bits. */ -static u64* alloc_pte(struct protection_domain *dom, - unsigned long address, u64 **pte_page, gfp_t gfp) +static bool increase_address_space(struct protection_domain *domain, + gfp_t gfp) +{ + u64 *pte; + + if (domain->mode == PAGE_MODE_6_LEVEL) + /* address space already 64 bit large */ + return false; + + pte = (void *)get_zeroed_page(gfp); + if (!pte) + return false; + + *pte = PM_LEVEL_PDE(domain->mode, + virt_to_phys(domain->pt_root)); + domain->pt_root = pte; + domain->mode += 1; + domain->updated = true; + + return true; +} + +static u64 *alloc_pte(struct protection_domain *domain, + unsigned long address, + int end_lvl, + u64 **pte_page, + gfp_t gfp) { u64 *pte, *page; + int level; - pte = &dom->pt_root[IOMMU_PTE_L2_INDEX(address)]; + while (address > PM_LEVEL_SIZE(domain->mode)) + increase_address_space(domain, gfp); - if (!IOMMU_PTE_PRESENT(*pte)) { - page = (u64 *)get_zeroed_page(gfp); - if (!page) - return NULL; - *pte = IOMMU_L2_PDE(virt_to_phys(page)); - } + level = domain->mode - 1; + pte = &domain->pt_root[PM_LEVEL_INDEX(level, address)]; - pte = IOMMU_PTE_PAGE(*pte); - pte = &pte[IOMMU_PTE_L1_INDEX(address)]; + while (level > end_lvl) { + if (!IOMMU_PTE_PRESENT(*pte)) { + page = (u64 *)get_zeroed_page(gfp); + if (!page) + return NULL; + *pte = PM_LEVEL_PDE(level, virt_to_phys(page)); + } - if (!IOMMU_PTE_PRESENT(*pte)) { - page = (u64 *)get_zeroed_page(gfp); - if (!page) - return NULL; - *pte = IOMMU_L1_PDE(virt_to_phys(page)); - } + level -= 1; - pte = IOMMU_PTE_PAGE(*pte); + pte = IOMMU_PTE_PAGE(*pte); - if (pte_page) - *pte_page = pte; + if (pte_page && level == end_lvl) + *pte_page = pte; - pte = &pte[IOMMU_PTE_L0_INDEX(address)]; + pte = &pte[PM_LEVEL_INDEX(level, address)]; + } return pte; } @@ -1344,10 +1513,13 @@ static u64* dma_ops_get_pte(struct dma_ops_domain *dom, pte = aperture->pte_pages[APERTURE_PAGE_INDEX(address)]; if (!pte) { - pte = alloc_pte(&dom->domain, address, &pte_page, GFP_ATOMIC); + pte = alloc_pte(&dom->domain, address, PM_MAP_4k, &pte_page, + GFP_ATOMIC); aperture->pte_pages[APERTURE_PAGE_INDEX(address)] = pte_page; } else - pte += IOMMU_PTE_L0_INDEX(address); + pte += PM_LEVEL_INDEX(0, address); + + update_domain(&dom->domain); return pte; } @@ -1409,7 +1581,7 @@ static void dma_ops_domain_unmap(struct amd_iommu *iommu, if (!pte) return; - pte += IOMMU_PTE_L0_INDEX(address); + pte += PM_LEVEL_INDEX(0, address); WARN_ON(!*pte); @@ -1988,19 +2160,47 @@ static void cleanup_domain(struct protection_domain *domain) write_unlock_irqrestore(&amd_iommu_devtable_lock, flags); } -static int amd_iommu_domain_init(struct iommu_domain *dom) +static void protection_domain_free(struct protection_domain *domain) +{ + if (!domain) + return; + + if (domain->id) + domain_id_free(domain->id); + + kfree(domain); +} + +static struct protection_domain *protection_domain_alloc(void) { struct protection_domain *domain; domain = kzalloc(sizeof(*domain), GFP_KERNEL); if (!domain) - return -ENOMEM; + return NULL; spin_lock_init(&domain->lock); - domain->mode = PAGE_MODE_3_LEVEL; domain->id = domain_id_alloc(); if (!domain->id) + goto out_err; + + return domain; + +out_err: + kfree(domain); + + return NULL; +} + +static int amd_iommu_domain_init(struct iommu_domain *dom) +{ + struct protection_domain *domain; + + domain = protection_domain_alloc(); + if (!domain) goto out_free; + + domain->mode = PAGE_MODE_3_LEVEL; domain->pt_root = (void *)get_zeroed_page(GFP_KERNEL); if (!domain->pt_root) goto out_free; @@ -2010,7 +2210,7 @@ static int amd_iommu_domain_init(struct iommu_domain *dom) return 0; out_free: - kfree(domain); + protection_domain_free(domain); return -ENOMEM; } @@ -2115,7 +2315,7 @@ static int amd_iommu_map_range(struct iommu_domain *dom, paddr &= PAGE_MASK; for (i = 0; i < npages; ++i) { - ret = iommu_map_page(domain, iova, paddr, prot); + ret = iommu_map_page(domain, iova, paddr, prot, PM_MAP_4k); if (ret) return ret; @@ -2136,7 +2336,7 @@ static void amd_iommu_unmap_range(struct iommu_domain *dom, iova &= PAGE_MASK; for (i = 0; i < npages; ++i) { - iommu_unmap_page(domain, iova); + iommu_unmap_page(domain, iova, PM_MAP_4k); iova += PAGE_SIZE; } @@ -2151,21 +2351,9 @@ static phys_addr_t amd_iommu_iova_to_phys(struct iommu_domain *dom, phys_addr_t paddr; u64 *pte; - pte = &domain->pt_root[IOMMU_PTE_L2_INDEX(iova)]; - - if (!IOMMU_PTE_PRESENT(*pte)) - return 0; - - pte = IOMMU_PTE_PAGE(*pte); - pte = &pte[IOMMU_PTE_L1_INDEX(iova)]; - - if (!IOMMU_PTE_PRESENT(*pte)) - return 0; - - pte = IOMMU_PTE_PAGE(*pte); - pte = &pte[IOMMU_PTE_L0_INDEX(iova)]; + pte = fetch_pte(domain, iova, PM_MAP_4k); - if (!IOMMU_PTE_PRESENT(*pte)) + if (!pte || !IOMMU_PTE_PRESENT(*pte)) return 0; paddr = *pte & IOMMU_PAGE_MASK; @@ -2191,3 +2379,46 @@ static struct iommu_ops amd_iommu_ops = { .domain_has_cap = amd_iommu_domain_has_cap, }; +/***************************************************************************** + * + * The next functions do a basic initialization of IOMMU for pass through + * mode + * + * In passthrough mode the IOMMU is initialized and enabled but not used for + * DMA-API translation. + * + *****************************************************************************/ + +int __init amd_iommu_init_passthrough(void) +{ + struct pci_dev *dev = NULL; + u16 devid, devid2; + + /* allocate passthroug domain */ + pt_domain = protection_domain_alloc(); + if (!pt_domain) + return -ENOMEM; + + pt_domain->mode |= PAGE_MODE_NONE; + + while ((dev = pci_get_device(PCI_ANY_ID, PCI_ANY_ID, dev)) != NULL) { + struct amd_iommu *iommu; + + devid = calc_devid(dev->bus->number, dev->devfn); + if (devid > amd_iommu_last_bdf) + continue; + + devid2 = amd_iommu_alias_table[devid]; + + iommu = amd_iommu_rlookup_table[devid2]; + if (!iommu) + continue; + + __attach_device(iommu, pt_domain, devid); + __attach_device(iommu, pt_domain, devid2); + } + + pr_info("AMD-Vi: Initialized for Passthrough Mode\n"); + + return 0; +} diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c index c1b17e97252..b4b61d462dc 100644 --- a/arch/x86/kernel/amd_iommu_init.c +++ b/arch/x86/kernel/amd_iommu_init.c @@ -252,7 +252,7 @@ static void __init iommu_feature_disable(struct amd_iommu *iommu, u8 bit) /* Function to enable the hardware */ static void iommu_enable(struct amd_iommu *iommu) { - printk(KERN_INFO "AMD IOMMU: Enabling IOMMU at %s cap 0x%hx\n", + printk(KERN_INFO "AMD-Vi: Enabling IOMMU at %s cap 0x%hx\n", dev_name(&iommu->dev->dev), iommu->cap_ptr); iommu_feature_enable(iommu, CONTROL_IOMMU_EN); @@ -435,6 +435,20 @@ static u8 * __init alloc_command_buffer(struct amd_iommu *iommu) } /* + * This function resets the command buffer if the IOMMU stopped fetching + * commands from it. + */ +void amd_iommu_reset_cmd_buffer(struct amd_iommu *iommu) +{ + iommu_feature_disable(iommu, CONTROL_CMDBUF_EN); + + writel(0x00, iommu->mmio_base + MMIO_CMD_HEAD_OFFSET); + writel(0x00, iommu->mmio_base + MMIO_CMD_TAIL_OFFSET); + + iommu_feature_enable(iommu, CONTROL_CMDBUF_EN); +} + +/* * This function writes the command buffer address to the hardware and * enables it. */ @@ -450,11 +464,7 @@ static void iommu_enable_command_buffer(struct amd_iommu *iommu) memcpy_toio(iommu->mmio_base + MMIO_CMD_BUF_OFFSET, &entry, sizeof(entry)); - /* set head and tail to zero manually */ - writel(0x00, iommu->mmio_base + MMIO_CMD_HEAD_OFFSET); - writel(0x00, iommu->mmio_base + MMIO_CMD_TAIL_OFFSET); - - iommu_feature_enable(iommu, CONTROL_CMDBUF_EN); + amd_iommu_reset_cmd_buffer(iommu); } static void __init free_command_buffer(struct amd_iommu *iommu) @@ -858,7 +868,7 @@ static int __init init_iommu_all(struct acpi_table_header *table) switch (*p) { case ACPI_IVHD_TYPE: - DUMP_printk("IOMMU: device: %02x:%02x.%01x cap: %04x " + DUMP_printk("device: %02x:%02x.%01x cap: %04x " "seg: %d flags: %01x info %04x\n", PCI_BUS(h->devid), PCI_SLOT(h->devid), PCI_FUNC(h->devid), h->cap_ptr, @@ -902,7 +912,7 @@ static int __init iommu_setup_msi(struct amd_iommu *iommu) r = request_irq(iommu->dev->irq, amd_iommu_int_handler, IRQF_SAMPLE_RANDOM, - "AMD IOMMU", + "AMD-Vi", NULL); if (r) { @@ -1150,7 +1160,7 @@ int __init amd_iommu_init(void) if (no_iommu) { - printk(KERN_INFO "AMD IOMMU disabled by kernel command line\n"); + printk(KERN_INFO "AMD-Vi disabled by kernel command line\n"); return 0; } @@ -1242,22 +1252,28 @@ int __init amd_iommu_init(void) if (ret) goto free; - ret = amd_iommu_init_dma_ops(); + if (iommu_pass_through) + ret = amd_iommu_init_passthrough(); + else + ret = amd_iommu_init_dma_ops(); if (ret) goto free; enable_iommus(); - printk(KERN_INFO "AMD IOMMU: device isolation "); + if (iommu_pass_through) + goto out; + + printk(KERN_INFO "AMD-Vi: device isolation "); if (amd_iommu_isolate) printk("enabled\n"); else printk("disabled\n"); if (amd_iommu_unmap_flush) - printk(KERN_INFO "AMD IOMMU: IO/TLB flush on unmap enabled\n"); + printk(KERN_INFO "AMD-Vi: IO/TLB flush on unmap enabled\n"); else - printk(KERN_INFO "AMD IOMMU: Lazy IO/TLB flushing enabled\n"); + printk(KERN_INFO "AMD-Vi: Lazy IO/TLB flushing enabled\n"); out: return ret; diff --git a/arch/x86/kernel/apic/io_apic.c b/arch/x86/kernel/apic/io_apic.c index 2284a4812b6..d2ed6c5ddc8 100644 --- a/arch/x86/kernel/apic/io_apic.c +++ b/arch/x86/kernel/apic/io_apic.c @@ -3793,6 +3793,9 @@ int arch_enable_uv_irq(char *irq_name, unsigned int irq, int cpu, int mmr_blade, mmr_pnode = uv_blade_to_pnode(mmr_blade); uv_write_global_mmr64(mmr_pnode, mmr_offset, mmr_value); + if (cfg->move_in_progress) + send_cleanup_vector(cfg); + return irq; } diff --git a/arch/x86/kernel/apic/ipi.c b/arch/x86/kernel/apic/ipi.c index dbf5445727a..6ef00ba4c88 100644 --- a/arch/x86/kernel/apic/ipi.c +++ b/arch/x86/kernel/apic/ipi.c @@ -106,6 +106,9 @@ void default_send_IPI_mask_logical(const struct cpumask *cpumask, int vector) unsigned long mask = cpumask_bits(cpumask)[0]; unsigned long flags; + if (WARN_ONCE(!mask, "empty IPI mask")) + return; + local_irq_save(flags); WARN_ON(mask & ~cpumask_bits(cpu_online_mask)[0]); __default_send_IPI_dest_field(mask, vector, apic->dest_logical); diff --git a/arch/x86/kernel/apic/probe_64.c b/arch/x86/kernel/apic/probe_64.c index bc3e880f9b8..fcec2f1d34a 100644 --- a/arch/x86/kernel/apic/probe_64.c +++ b/arch/x86/kernel/apic/probe_64.c @@ -44,6 +44,11 @@ static struct apic *apic_probe[] __initdata = { NULL, }; +static int apicid_phys_pkg_id(int initial_apic_id, int index_msb) +{ + return hard_smp_processor_id() >> index_msb; +} + /* * Check the APIC IDs in bios_cpu_apicid and choose the APIC mode. */ @@ -69,6 +74,11 @@ void __init default_setup_apic_routing(void) printk(KERN_INFO "Setting APIC routing to %s\n", apic->name); } + if (is_vsmp_box()) { + /* need to update phys_pkg_id */ + apic->phys_pkg_id = apicid_phys_pkg_id; + } + /* * Now that apic routing model is selected, configure the * fault handling for intr remapping. diff --git a/arch/x86/kernel/apic/x2apic_cluster.c b/arch/x86/kernel/apic/x2apic_cluster.c index 8e4cbb255c3..a5371ec3677 100644 --- a/arch/x86/kernel/apic/x2apic_cluster.c +++ b/arch/x86/kernel/apic/x2apic_cluster.c @@ -17,11 +17,13 @@ static int x2apic_acpi_madt_oem_check(char *oem_id, char *oem_table_id) return x2apic_enabled(); } -/* Start with all IRQs pointing to boot CPU. IRQ balancing will shift them. */ - +/* + * need to use more than cpu 0, because we need more vectors when + * MSI-X are used. + */ static const struct cpumask *x2apic_target_cpus(void) { - return cpumask_of(0); + return cpu_online_mask; } /* @@ -170,7 +172,7 @@ static unsigned long set_apic_id(unsigned int id) static int x2apic_cluster_phys_pkg_id(int initial_apicid, int index_msb) { - return current_cpu_data.initial_apicid >> index_msb; + return initial_apicid >> index_msb; } static void x2apic_send_IPI_self(int vector) diff --git a/arch/x86/kernel/apic/x2apic_phys.c b/arch/x86/kernel/apic/x2apic_phys.c index a284359627e..a8989aadc99 100644 --- a/arch/x86/kernel/apic/x2apic_phys.c +++ b/arch/x86/kernel/apic/x2apic_phys.c @@ -27,11 +27,13 @@ static int x2apic_acpi_madt_oem_check(char *oem_id, char *oem_table_id) return 0; } -/* Start with all IRQs pointing to boot CPU. IRQ balancing will shift them. */ - +/* + * need to use more than cpu 0, because we need more vectors when + * MSI-X are used. + */ static const struct cpumask *x2apic_target_cpus(void) { - return cpumask_of(0); + return cpu_online_mask; } static void x2apic_vector_allocation_domain(int cpu, struct cpumask *retmask) @@ -162,7 +164,7 @@ static unsigned long set_apic_id(unsigned int id) static int x2apic_phys_pkg_id(int initial_apicid, int index_msb) { - return current_cpu_data.initial_apicid >> index_msb; + return initial_apicid >> index_msb; } static void x2apic_send_IPI_self(int vector) diff --git a/arch/x86/kernel/apic/x2apic_uv_x.c b/arch/x86/kernel/apic/x2apic_uv_x.c index 096d19aea2f..601159374e8 100644 --- a/arch/x86/kernel/apic/x2apic_uv_x.c +++ b/arch/x86/kernel/apic/x2apic_uv_x.c @@ -46,7 +46,7 @@ static int early_get_nodeid(void) return node_id.s.node_id; } -static int uv_acpi_madt_oem_check(char *oem_id, char *oem_table_id) +static int __init uv_acpi_madt_oem_check(char *oem_id, char *oem_table_id) { if (!strcmp(oem_id, "SGI")) { if (!strcmp(oem_table_id, "UVL")) @@ -253,7 +253,7 @@ static void uv_send_IPI_self(int vector) apic_write(APIC_SELF_IPI, vector); } -struct apic apic_x2apic_uv_x = { +struct apic __refdata apic_x2apic_uv_x = { .name = "UV large system", .probe = NULL, @@ -261,7 +261,7 @@ struct apic apic_x2apic_uv_x = { .apic_id_registered = uv_apic_id_registered, .irq_delivery_mode = dest_Fixed, - .irq_dest_mode = 1, /* logical */ + .irq_dest_mode = 0, /* physical */ .target_cpus = uv_target_cpus, .disable_esr = 0, @@ -362,12 +362,6 @@ static __init void get_lowmem_redirect(unsigned long *base, unsigned long *size) BUG(); } -static __init void map_low_mmrs(void) -{ - init_extra_mapping_uc(UV_GLOBAL_MMR32_BASE, UV_GLOBAL_MMR32_SIZE); - init_extra_mapping_uc(UV_LOCAL_MMR_BASE, UV_LOCAL_MMR_SIZE); -} - enum map_type {map_wb, map_uc}; static __init void map_high(char *id, unsigned long base, int shift, @@ -395,26 +389,6 @@ static __init void map_gru_high(int max_pnode) map_high("GRU", gru.s.base, shift, max_pnode, map_wb); } -static __init void map_config_high(int max_pnode) -{ - union uvh_rh_gam_cfg_overlay_config_mmr_u cfg; - int shift = UVH_RH_GAM_CFG_OVERLAY_CONFIG_MMR_BASE_SHFT; - - cfg.v = uv_read_local_mmr(UVH_RH_GAM_CFG_OVERLAY_CONFIG_MMR); - if (cfg.s.enable) - map_high("CONFIG", cfg.s.base, shift, max_pnode, map_uc); -} - -static __init void map_mmr_high(int max_pnode) -{ - union uvh_rh_gam_mmr_overlay_config_mmr_u mmr; - int shift = UVH_RH_GAM_MMR_OVERLAY_CONFIG_MMR_BASE_SHFT; - - mmr.v = uv_read_local_mmr(UVH_RH_GAM_MMR_OVERLAY_CONFIG_MMR); - if (mmr.s.enable) - map_high("MMR", mmr.s.base, shift, max_pnode, map_uc); -} - static __init void map_mmioh_high(int max_pnode) { union uvh_rh_gam_mmioh_overlay_config_mmr_u mmioh; @@ -566,8 +540,6 @@ void __init uv_system_init(void) unsigned long mmr_base, present, paddr; unsigned short pnode_mask; - map_low_mmrs(); - m_n_config.v = uv_read_local_mmr(UVH_SI_ADDR_MAP_CONFIG); m_val = m_n_config.s.m_skt; n_val = m_n_config.s.n_skt; @@ -591,6 +563,8 @@ void __init uv_system_init(void) bytes = sizeof(struct uv_blade_info) * uv_num_possible_blades(); uv_blade_info = kmalloc(bytes, GFP_KERNEL); BUG_ON(!uv_blade_info); + for (blade = 0; blade < uv_num_possible_blades(); blade++) + uv_blade_info[blade].memory_nid = -1; get_lowmem_redirect(&lowmem_redir_base, &lowmem_redir_size); @@ -629,6 +603,9 @@ void __init uv_system_init(void) lcpu = uv_blade_info[blade].nr_possible_cpus; uv_blade_info[blade].nr_possible_cpus++; + /* Any node on the blade, else will contain -1. */ + uv_blade_info[blade].memory_nid = nid; + uv_cpu_hub_info(cpu)->lowmem_remap_base = lowmem_redir_base; uv_cpu_hub_info(cpu)->lowmem_remap_top = lowmem_redir_size; uv_cpu_hub_info(cpu)->m_val = m_val; @@ -662,11 +639,10 @@ void __init uv_system_init(void) pnode = (paddr >> m_val) & pnode_mask; blade = boot_pnode_to_blade(pnode); uv_node_to_blade[nid] = blade; + max_pnode = max(pnode, max_pnode); } map_gru_high(max_pnode); - map_mmr_high(max_pnode); - map_config_high(max_pnode); map_mmioh_high(max_pnode); uv_cpu_init(); diff --git a/arch/x86/kernel/apm_32.c b/arch/x86/kernel/apm_32.c index 79302e9a33a..442b5508893 100644 --- a/arch/x86/kernel/apm_32.c +++ b/arch/x86/kernel/apm_32.c @@ -811,7 +811,7 @@ static int apm_do_idle(void) u8 ret = 0; int idled = 0; int polling; - int err; + int err = 0; polling = !!(current_thread_info()->status & TS_POLLING); if (polling) { diff --git a/arch/x86/kernel/cpu/Makefile b/arch/x86/kernel/cpu/Makefile index 3efcb2b96a1..c1f253dac15 100644 --- a/arch/x86/kernel/cpu/Makefile +++ b/arch/x86/kernel/cpu/Makefile @@ -7,6 +7,10 @@ ifdef CONFIG_FUNCTION_TRACER CFLAGS_REMOVE_common.o = -pg endif +# Make sure load_percpu_segment has no stackprotector +nostackp := $(call cc-option, -fno-stack-protector) +CFLAGS_common.o := $(nostackp) + obj-y := intel_cacheinfo.o addon_cpuid_features.o obj-y += proc.o capflags.o powerflags.o common.o obj-y += vmware.o hypervisor.o diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c index e2485b03f1c..63fddcd082c 100644 --- a/arch/x86/kernel/cpu/amd.c +++ b/arch/x86/kernel/cpu/amd.c @@ -400,6 +400,13 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c) level = cpuid_eax(1); if((level >= 0x0f48 && level < 0x0f50) || level >= 0x0f58) set_cpu_cap(c, X86_FEATURE_REP_GOOD); + + /* + * Some BIOSes incorrectly force this feature, but only K8 + * revision D (model = 0x14) and later actually support it. + */ + if (c->x86_model < 0x14) + clear_cpu_cap(c, X86_FEATURE_LAHF_LM); } if (c->x86 == 0x10 || c->x86 == 0x11) set_cpu_cap(c, X86_FEATURE_REP_GOOD); diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index f1961c07af9..5ce60a88027 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -59,7 +59,30 @@ void __init setup_cpu_local_masks(void) alloc_bootmem_cpumask_var(&cpu_sibling_setup_mask); } -static const struct cpu_dev *this_cpu __cpuinitdata; +static void __cpuinit default_init(struct cpuinfo_x86 *c) +{ +#ifdef CONFIG_X86_64 + display_cacheinfo(c); +#else + /* Not much we can do here... */ + /* Check if at least it has cpuid */ + if (c->cpuid_level == -1) { + /* No cpuid. It must be an ancient CPU */ + if (c->x86 == 4) + strcpy(c->x86_model_id, "486"); + else if (c->x86 == 3) + strcpy(c->x86_model_id, "386"); + } +#endif +} + +static const struct cpu_dev __cpuinitconst default_cpu = { + .c_init = default_init, + .c_vendor = "Unknown", + .c_x86_vendor = X86_VENDOR_UNKNOWN, +}; + +static const struct cpu_dev *this_cpu __cpuinitdata = &default_cpu; DEFINE_PER_CPU_PAGE_ALIGNED(struct gdt_page, gdt_page) = { .gdt = { #ifdef CONFIG_X86_64 @@ -332,29 +355,6 @@ void switch_to_new_gdt(int cpu) static const struct cpu_dev *__cpuinitdata cpu_devs[X86_VENDOR_NUM] = {}; -static void __cpuinit default_init(struct cpuinfo_x86 *c) -{ -#ifdef CONFIG_X86_64 - display_cacheinfo(c); -#else - /* Not much we can do here... */ - /* Check if at least it has cpuid */ - if (c->cpuid_level == -1) { - /* No cpuid. It must be an ancient CPU */ - if (c->x86 == 4) - strcpy(c->x86_model_id, "486"); - else if (c->x86 == 3) - strcpy(c->x86_model_id, "386"); - } -#endif -} - -static const struct cpu_dev __cpuinitconst default_cpu = { - .c_init = default_init, - .c_vendor = "Unknown", - .c_x86_vendor = X86_VENDOR_UNKNOWN, -}; - static void __cpuinit get_model_name(struct cpuinfo_x86 *c) { unsigned int *v; diff --git a/arch/x86/kernel/cpu/mcheck/mce.c b/arch/x86/kernel/cpu/mcheck/mce.c index 1cfb623ce11..01213048f62 100644 --- a/arch/x86/kernel/cpu/mcheck/mce.c +++ b/arch/x86/kernel/cpu/mcheck/mce.c @@ -1226,8 +1226,13 @@ static void mce_init(void) } /* Add per CPU specific workarounds here */ -static void mce_cpu_quirks(struct cpuinfo_x86 *c) +static int mce_cpu_quirks(struct cpuinfo_x86 *c) { + if (c->x86_vendor == X86_VENDOR_UNKNOWN) { + pr_info("MCE: unknown CPU type - not enabling MCE support.\n"); + return -EOPNOTSUPP; + } + /* This should be disabled by the BIOS, but isn't always */ if (c->x86_vendor == X86_VENDOR_AMD) { if (c->x86 == 15 && banks > 4) { @@ -1273,11 +1278,20 @@ static void mce_cpu_quirks(struct cpuinfo_x86 *c) if ((c->x86 > 6 || (c->x86 == 6 && c->x86_model >= 0xe)) && monarch_timeout < 0) monarch_timeout = USEC_PER_SEC; + + /* + * There are also broken BIOSes on some Pentium M and + * earlier systems: + */ + if (c->x86 == 6 && c->x86_model <= 13 && mce_bootlog < 0) + mce_bootlog = 0; } if (monarch_timeout < 0) monarch_timeout = 0; if (mce_bootlog != 0) mce_panic_timeout = 30; + + return 0; } static void __cpuinit mce_ancient_init(struct cpuinfo_x86 *c) @@ -1338,11 +1352,10 @@ void __cpuinit mcheck_init(struct cpuinfo_x86 *c) if (!mce_available(c)) return; - if (mce_cap_init() < 0) { + if (mce_cap_init() < 0 || mce_cpu_quirks(c) < 0) { mce_disabled = 1; return; } - mce_cpu_quirks(c); machine_check_vector = do_machine_check; diff --git a/arch/x86/kernel/cpu/mcheck/therm_throt.c b/arch/x86/kernel/cpu/mcheck/therm_throt.c index bff8dd191dd..5957a93e517 100644 --- a/arch/x86/kernel/cpu/mcheck/therm_throt.c +++ b/arch/x86/kernel/cpu/mcheck/therm_throt.c @@ -36,6 +36,7 @@ static DEFINE_PER_CPU(__u64, next_check) = INITIAL_JIFFIES; static DEFINE_PER_CPU(unsigned long, thermal_throttle_count); +static DEFINE_PER_CPU(bool, thermal_throttle_active); static atomic_t therm_throt_en = ATOMIC_INIT(0); @@ -96,27 +97,33 @@ static int therm_throt_process(int curr) { unsigned int cpu = smp_processor_id(); __u64 tmp_jiffs = get_jiffies_64(); + bool was_throttled = __get_cpu_var(thermal_throttle_active); + bool is_throttled = __get_cpu_var(thermal_throttle_active) = curr; - if (curr) + if (is_throttled) __get_cpu_var(thermal_throttle_count)++; - if (time_before64(tmp_jiffs, __get_cpu_var(next_check))) + if (!(was_throttled ^ is_throttled) && + time_before64(tmp_jiffs, __get_cpu_var(next_check))) return 0; __get_cpu_var(next_check) = tmp_jiffs + CHECK_INTERVAL; /* if we just entered the thermal event */ - if (curr) { + if (is_throttled) { printk(KERN_CRIT "CPU%d: Temperature above threshold, " - "cpu clock throttled (total events = %lu)\n", cpu, - __get_cpu_var(thermal_throttle_count)); + "cpu clock throttled (total events = %lu)\n", + cpu, __get_cpu_var(thermal_throttle_count)); add_taint(TAINT_MACHINE_CHECK); - } else { - printk(KERN_CRIT "CPU%d: Temperature/speed normal\n", cpu); + return 1; + } + if (was_throttled) { + printk(KERN_INFO "CPU%d: Temperature/speed normal\n", cpu); + return 1; } - return 1; + return 0; } #ifdef CONFIG_SYSFS diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c index a7aa8f90095..900332b800f 100644 --- a/arch/x86/kernel/cpu/perf_counter.c +++ b/arch/x86/kernel/cpu/perf_counter.c @@ -55,6 +55,7 @@ struct x86_pmu { int num_counters_fixed; int counter_bits; u64 counter_mask; + int apic; u64 max_period; u64 intel_ctrl; }; @@ -72,8 +73,8 @@ static const u64 p6_perfmon_event_map[] = { [PERF_COUNT_HW_CPU_CYCLES] = 0x0079, [PERF_COUNT_HW_INSTRUCTIONS] = 0x00c0, - [PERF_COUNT_HW_CACHE_REFERENCES] = 0x0000, - [PERF_COUNT_HW_CACHE_MISSES] = 0x0000, + [PERF_COUNT_HW_CACHE_REFERENCES] = 0x0f2e, + [PERF_COUNT_HW_CACHE_MISSES] = 0x012e, [PERF_COUNT_HW_BRANCH_INSTRUCTIONS] = 0x00c4, [PERF_COUNT_HW_BRANCH_MISSES] = 0x00c5, [PERF_COUNT_HW_BUS_CYCLES] = 0x0062, @@ -613,6 +614,7 @@ static DEFINE_MUTEX(pmc_reserve_mutex); static bool reserve_pmc_hardware(void) { +#ifdef CONFIG_X86_LOCAL_APIC int i; if (nmi_watchdog == NMI_LOCAL_APIC) @@ -627,9 +629,11 @@ static bool reserve_pmc_hardware(void) if (!reserve_evntsel_nmi(x86_pmu.eventsel + i)) goto eventsel_fail; } +#endif return true; +#ifdef CONFIG_X86_LOCAL_APIC eventsel_fail: for (i--; i >= 0; i--) release_evntsel_nmi(x86_pmu.eventsel + i); @@ -644,10 +648,12 @@ perfctr_fail: enable_lapic_nmi_watchdog(); return false; +#endif } static void release_pmc_hardware(void) { +#ifdef CONFIG_X86_LOCAL_APIC int i; for (i = 0; i < x86_pmu.num_counters; i++) { @@ -657,6 +663,7 @@ static void release_pmc_hardware(void) if (nmi_watchdog == NMI_LOCAL_APIC) enable_lapic_nmi_watchdog(); +#endif } static void hw_perf_counter_destroy(struct perf_counter *counter) @@ -748,6 +755,15 @@ static int __hw_perf_counter_init(struct perf_counter *counter) hwc->sample_period = x86_pmu.max_period; hwc->last_period = hwc->sample_period; atomic64_set(&hwc->period_left, hwc->sample_period); + } else { + /* + * If we have a PMU initialized but no APIC + * interrupts, we cannot sample hardware + * counters (user-space has to fall back and + * sample via a hrtimer based software counter): + */ + if (!x86_pmu.apic) + return -EOPNOTSUPP; } counter->destroy = hw_perf_counter_destroy; @@ -1449,18 +1465,22 @@ void smp_perf_pending_interrupt(struct pt_regs *regs) void set_perf_counter_pending(void) { +#ifdef CONFIG_X86_LOCAL_APIC apic->send_IPI_self(LOCAL_PENDING_VECTOR); +#endif } void perf_counters_lapic_init(void) { - if (!x86_pmu_initialized()) +#ifdef CONFIG_X86_LOCAL_APIC + if (!x86_pmu.apic || !x86_pmu_initialized()) return; /* * Always use NMI for PMU */ apic_write(APIC_LVTPC, APIC_DM_NMI); +#endif } static int __kprobes @@ -1484,7 +1504,9 @@ perf_counter_nmi_handler(struct notifier_block *self, regs = args->regs; +#ifdef CONFIG_X86_LOCAL_APIC apic_write(APIC_LVTPC, APIC_DM_NMI); +#endif /* * Can't rely on the handled return value to say it was our NMI, two * counters could trigger 'simultaneously' raising two back-to-back NMIs. @@ -1515,6 +1537,7 @@ static struct x86_pmu p6_pmu = { .event_map = p6_pmu_event_map, .raw_event = p6_pmu_raw_event, .max_events = ARRAY_SIZE(p6_perfmon_event_map), + .apic = 1, .max_period = (1ULL << 31) - 1, .version = 0, .num_counters = 2, @@ -1541,6 +1564,7 @@ static struct x86_pmu intel_pmu = { .event_map = intel_pmu_event_map, .raw_event = intel_pmu_raw_event, .max_events = ARRAY_SIZE(intel_perfmon_event_map), + .apic = 1, /* * Intel PMCs cannot be accessed sanely above 32 bit width, * so we install an artificial 1<<31 period regardless of @@ -1564,6 +1588,7 @@ static struct x86_pmu amd_pmu = { .num_counters = 4, .counter_bits = 48, .counter_mask = (1ULL << 48) - 1, + .apic = 1, /* use highest bit to detect overflow */ .max_period = (1ULL << 47) - 1, }; @@ -1589,13 +1614,14 @@ static int p6_pmu_init(void) return -ENODEV; } + x86_pmu = p6_pmu; + if (!cpu_has_apic) { - pr_info("no Local APIC, try rebooting with lapic"); - return -ENODEV; + pr_info("no APIC, boot with the \"lapic\" boot parameter to force-enable it.\n"); + pr_info("no hardware sampling interrupt available.\n"); + x86_pmu.apic = 0; } - x86_pmu = p6_pmu; - return 0; } diff --git a/arch/x86/kernel/efi.c b/arch/x86/kernel/efi.c index 96f7ac0bbf0..fe26ba3e345 100644 --- a/arch/x86/kernel/efi.c +++ b/arch/x86/kernel/efi.c @@ -354,7 +354,7 @@ void __init efi_init(void) */ c16 = tmp = early_ioremap(efi.systab->fw_vendor, 2); if (c16) { - for (i = 0; i < sizeof(vendor) && *c16; ++i) + for (i = 0; i < sizeof(vendor) - 1 && *c16; ++i) vendor[i] = *c16++; vendor[i] = '\0'; } else @@ -512,7 +512,7 @@ void __init efi_enter_virtual_mode(void) && end_pfn <= max_pfn_mapped)) va = __va(md->phys_addr); else - va = efi_ioremap(md->phys_addr, size); + va = efi_ioremap(md->phys_addr, size, md->type); md->virt_addr = (u64) (unsigned long) va; diff --git a/arch/x86/kernel/efi_64.c b/arch/x86/kernel/efi_64.c index 22c3b7828c5..ac0621a7ac3 100644 --- a/arch/x86/kernel/efi_64.c +++ b/arch/x86/kernel/efi_64.c @@ -98,10 +98,14 @@ void __init efi_call_phys_epilog(void) early_runtime_code_mapping_set_exec(0); } -void __iomem *__init efi_ioremap(unsigned long phys_addr, unsigned long size) +void __iomem *__init efi_ioremap(unsigned long phys_addr, unsigned long size, + u32 type) { unsigned long last_map_pfn; + if (type == EFI_MEMORY_MAPPED_IO) + return ioremap(phys_addr, size); + last_map_pfn = init_memory_mapping(phys_addr, phys_addr + size); if ((last_map_pfn << PAGE_SHIFT) < phys_addr + size) return NULL; diff --git a/arch/x86/kernel/head_32.S b/arch/x86/kernel/head_32.S index 8663afb5653..cc827ac9e8d 100644 --- a/arch/x86/kernel/head_32.S +++ b/arch/x86/kernel/head_32.S @@ -261,9 +261,7 @@ page_pde_offset = (__PAGE_OFFSET >> 20); * which will be freed later */ -#ifndef CONFIG_HOTPLUG_CPU -.section .init.text,"ax",@progbits -#endif +__CPUINIT #ifdef CONFIG_SMP ENTRY(startup_32_smp) @@ -602,7 +600,7 @@ ignore_int: #endif iret -.section .cpuinit.data,"wa" + __REFDATA .align 4 ENTRY(initial_code) .long i386_start_kernel diff --git a/arch/x86/kernel/pci-dma.c b/arch/x86/kernel/pci-dma.c index 3c945c0b350..8fb4ce35bea 100644 --- a/arch/x86/kernel/pci-dma.c +++ b/arch/x86/kernel/pci-dma.c @@ -32,7 +32,14 @@ int no_iommu __read_mostly; /* Set this to 1 if there is a HW IOMMU in the system */ int iommu_detected __read_mostly = 0; -int iommu_pass_through; +/* + * This variable becomes 1 if iommu=pt is passed on the kernel command line. + * If this variable is 1, IOMMU implementations do no DMA ranslation for + * devices and allow every device to access to whole physical memory. This is + * useful if a user want to use an IOMMU only for KVM device assignment to + * guests and not for driver dma translation. + */ +int iommu_pass_through __read_mostly; dma_addr_t bad_dma_address __read_mostly = 0; EXPORT_SYMBOL(bad_dma_address); diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c index 994dd6a4a2a..071166a4ba8 100644 --- a/arch/x86/kernel/process.c +++ b/arch/x86/kernel/process.c @@ -519,16 +519,12 @@ static void c1e_idle(void) if (!cpumask_test_cpu(cpu, c1e_mask)) { cpumask_set_cpu(cpu, c1e_mask); /* - * Force broadcast so ACPI can not interfere. Needs - * to run with interrupts enabled as it uses - * smp_function_call. + * Force broadcast so ACPI can not interfere. */ - local_irq_enable(); clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_FORCE, &cpu); printk(KERN_INFO "Switch to broadcast mode on CPU%d\n", cpu); - local_irq_disable(); } clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_ENTER, &cpu); diff --git a/arch/x86/kernel/reboot.c b/arch/x86/kernel/reboot.c index 508e982dd07..a06e8d10184 100644 --- a/arch/x86/kernel/reboot.c +++ b/arch/x86/kernel/reboot.c @@ -3,6 +3,7 @@ #include <linux/init.h> #include <linux/pm.h> #include <linux/efi.h> +#include <linux/dmi.h> #include <acpi/reboot.h> #include <asm/io.h> #include <asm/apic.h> @@ -17,7 +18,6 @@ #include <asm/cpu.h> #ifdef CONFIG_X86_32 -# include <linux/dmi.h> # include <linux/ctype.h> # include <linux/mc146818rtc.h> #else @@ -404,6 +404,46 @@ EXPORT_SYMBOL(machine_real_restart); #endif /* CONFIG_X86_32 */ +/* + * Some Apple MacBook and MacBookPro's needs reboot=p to be able to reboot + */ +static int __init set_pci_reboot(const struct dmi_system_id *d) +{ + if (reboot_type != BOOT_CF9) { + reboot_type = BOOT_CF9; + printk(KERN_INFO "%s series board detected. " + "Selecting PCI-method for reboots.\n", d->ident); + } + return 0; +} + +static struct dmi_system_id __initdata pci_reboot_dmi_table[] = { + { /* Handle problems with rebooting on Apple MacBook5 */ + .callback = set_pci_reboot, + .ident = "Apple MacBook5", + .matches = { + DMI_MATCH(DMI_SYS_VENDOR, "Apple Inc."), + DMI_MATCH(DMI_PRODUCT_NAME, "MacBook5"), + }, + }, + { /* Handle problems with rebooting on Apple MacBookPro5 */ + .callback = set_pci_reboot, + .ident = "Apple MacBookPro5", + .matches = { + DMI_MATCH(DMI_SYS_VENDOR, "Apple Inc."), + DMI_MATCH(DMI_PRODUCT_NAME, "MacBookPro5"), + }, + }, + { } +}; + +static int __init pci_reboot_init(void) +{ + dmi_check_system(pci_reboot_dmi_table); + return 0; +} +core_initcall(pci_reboot_init); + static inline void kb_wait(void) { int i; diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index 29a3eef7cf4..07d81916f21 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c @@ -165,7 +165,7 @@ static ssize_t __init setup_pcpu_lpage(size_t static_size, bool chosen) if (!chosen) { size_t vm_size = VMALLOC_END - VMALLOC_START; - size_t tot_size = num_possible_cpus() * PMD_SIZE; + size_t tot_size = nr_cpu_ids * PMD_SIZE; /* on non-NUMA, embedding is better */ if (!pcpu_need_numa()) @@ -199,7 +199,7 @@ static ssize_t __init setup_pcpu_lpage(size_t static_size, bool chosen) dyn_size = pcpul_size - static_size - PERCPU_FIRST_CHUNK_RESERVE; /* allocate pointer array and alloc large pages */ - map_size = PFN_ALIGN(num_possible_cpus() * sizeof(pcpul_map[0])); + map_size = PFN_ALIGN(nr_cpu_ids * sizeof(pcpul_map[0])); pcpul_map = alloc_bootmem(map_size); for_each_possible_cpu(cpu) { @@ -228,7 +228,7 @@ static ssize_t __init setup_pcpu_lpage(size_t static_size, bool chosen) /* allocate address and map */ pcpul_vm.flags = VM_ALLOC; - pcpul_vm.size = num_possible_cpus() * PMD_SIZE; + pcpul_vm.size = nr_cpu_ids * PMD_SIZE; vm_area_register_early(&pcpul_vm, PMD_SIZE); for_each_possible_cpu(cpu) { @@ -250,8 +250,8 @@ static ssize_t __init setup_pcpu_lpage(size_t static_size, bool chosen) PMD_SIZE, pcpul_vm.addr, NULL); /* sort pcpul_map array for pcpu_lpage_remapped() */ - for (i = 0; i < num_possible_cpus() - 1; i++) - for (j = i + 1; j < num_possible_cpus(); j++) + for (i = 0; i < nr_cpu_ids - 1; i++) + for (j = i + 1; j < nr_cpu_ids; j++) if (pcpul_map[i].ptr > pcpul_map[j].ptr) { struct pcpul_ent tmp = pcpul_map[i]; pcpul_map[i] = pcpul_map[j]; @@ -288,7 +288,7 @@ void *pcpu_lpage_remapped(void *kaddr) { void *pmd_addr = (void *)((unsigned long)kaddr & PMD_MASK); unsigned long offset = (unsigned long)kaddr & ~PMD_MASK; - int left = 0, right = num_possible_cpus() - 1; + int left = 0, right = nr_cpu_ids - 1; int pos; /* pcpul in use at all? */ @@ -377,7 +377,7 @@ static ssize_t __init setup_pcpu_4k(size_t static_size) pcpu4k_nr_static_pages = PFN_UP(static_size); /* unaligned allocations can't be freed, round up to page size */ - pages_size = PFN_ALIGN(pcpu4k_nr_static_pages * num_possible_cpus() + pages_size = PFN_ALIGN(pcpu4k_nr_static_pages * nr_cpu_ids * sizeof(pcpu4k_pages[0])); pcpu4k_pages = alloc_bootmem(pages_size); diff --git a/arch/x86/kernel/tlb_uv.c b/arch/x86/kernel/tlb_uv.c index 8ccabb8a2f6..77b9689f8ed 100644 --- a/arch/x86/kernel/tlb_uv.c +++ b/arch/x86/kernel/tlb_uv.c @@ -744,6 +744,7 @@ uv_activation_descriptor_init(int node, int pnode) * note that base_dest_nodeid is actually a nasid. */ ad2->header.base_dest_nodeid = uv_partition_base_pnode << 1; + ad2->header.dest_subnodeid = 0x10; /* the LB */ ad2->header.command = UV_NET_ENDPOINT_INTD; ad2->header.int_both = 1; /* diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c index 6e1a368d21d..71f4368b357 100644 --- a/arch/x86/kernel/tsc.c +++ b/arch/x86/kernel/tsc.c @@ -275,15 +275,20 @@ static unsigned long pit_calibrate_tsc(u32 latch, unsigned long ms, int loopmin) * use the TSC value at the transitions to calculate a pretty * good value for the TSC frequencty. */ +static inline int pit_verify_msb(unsigned char val) +{ + /* Ignore LSB */ + inb(0x42); + return inb(0x42) == val; +} + static inline int pit_expect_msb(unsigned char val, u64 *tscp, unsigned long *deltap) { int count; u64 tsc = 0; for (count = 0; count < 50000; count++) { - /* Ignore LSB */ - inb(0x42); - if (inb(0x42) != val) + if (!pit_verify_msb(val)) break; tsc = get_cycles(); } @@ -336,8 +341,7 @@ static unsigned long quick_pit_calibrate(void) * to do that is to just read back the 16-bit counter * once from the PIT. */ - inb(0x42); - inb(0x42); + pit_verify_msb(0); if (pit_expect_msb(0xff, &tsc, &d1)) { for (i = 1; i <= MAX_QUICK_PIT_ITERATIONS; i++) { @@ -348,8 +352,19 @@ static unsigned long quick_pit_calibrate(void) * Iterate until the error is less than 500 ppm */ delta -= tsc; - if (d1+d2 < delta >> 11) - goto success; + if (d1+d2 >= delta >> 11) + continue; + + /* + * Check the PIT one more time to verify that + * all TSC reads were stable wrt the PIT. + * + * This also guarantees serialization of the + * last cycle read ('d2') in pit_expect_msb. + */ + if (!pit_verify_msb(0xfe - i)) + break; + goto success; } } printk("Fast TSC calibration failed\n"); diff --git a/arch/x86/kernel/vmi_32.c b/arch/x86/kernel/vmi_32.c index b263423fbe2..95a7289e4b0 100644 --- a/arch/x86/kernel/vmi_32.c +++ b/arch/x86/kernel/vmi_32.c @@ -441,7 +441,7 @@ vmi_startup_ipi_hook(int phys_apicid, unsigned long start_eip, ap.ds = __USER_DS; ap.es = __USER_DS; ap.fs = __KERNEL_PERCPU; - ap.gs = 0; + ap.gs = __KERNEL_STACK_CANARY; ap.eflags = 0; diff --git a/arch/x86/kernel/vmlinux.lds.S b/arch/x86/kernel/vmlinux.lds.S index 59f31d2dd43..9fc178255c0 100644 --- a/arch/x86/kernel/vmlinux.lds.S +++ b/arch/x86/kernel/vmlinux.lds.S @@ -46,11 +46,10 @@ PHDRS { data PT_LOAD FLAGS(7); /* RWE */ #ifdef CONFIG_X86_64 user PT_LOAD FLAGS(7); /* RWE */ - data.init PT_LOAD FLAGS(7); /* RWE */ #ifdef CONFIG_SMP percpu PT_LOAD FLAGS(7); /* RWE */ #endif - data.init2 PT_LOAD FLAGS(7); /* RWE */ + init PT_LOAD FLAGS(7); /* RWE */ #endif note PT_NOTE FLAGS(0); /* ___ */ } @@ -103,65 +102,43 @@ SECTIONS __stop___ex_table = .; } :text = 0x9090 - RODATA + RO_DATA(PAGE_SIZE) /* Data */ - . = ALIGN(PAGE_SIZE); .data : AT(ADDR(.data) - LOAD_OFFSET) { /* Start of data section */ _sdata = .; - DATA_DATA - CONSTRUCTORS - } :data + + /* init_task */ + INIT_TASK_DATA(THREAD_SIZE) #ifdef CONFIG_X86_32 - /* 32 bit has nosave before _edata */ - . = ALIGN(PAGE_SIZE); - .data_nosave : AT(ADDR(.data_nosave) - LOAD_OFFSET) { - __nosave_begin = .; - *(.data.nosave) - . = ALIGN(PAGE_SIZE); - __nosave_end = .; - } + /* 32 bit has nosave before _edata */ + NOSAVE_DATA #endif - . = ALIGN(PAGE_SIZE); - .data.page_aligned : AT(ADDR(.data.page_aligned) - LOAD_OFFSET) { - *(.data.page_aligned) + PAGE_ALIGNED_DATA(PAGE_SIZE) *(.data.idt) - } -#ifdef CONFIG_X86_32 - . = ALIGN(32); -#else - . = ALIGN(PAGE_SIZE); - . = ALIGN(CONFIG_X86_L1_CACHE_BYTES); -#endif - .data.cacheline_aligned : - AT(ADDR(.data.cacheline_aligned) - LOAD_OFFSET) { - *(.data.cacheline_aligned) - } + CACHELINE_ALIGNED_DATA(CONFIG_X86_L1_CACHE_BYTES) - /* rarely changed data like cpu maps */ -#ifdef CONFIG_X86_32 - . = ALIGN(32); -#else - . = ALIGN(CONFIG_X86_INTERNODE_CACHE_BYTES); -#endif - .data.read_mostly : AT(ADDR(.data.read_mostly) - LOAD_OFFSET) { - *(.data.read_mostly) + DATA_DATA + CONSTRUCTORS + + /* rarely changed data like cpu maps */ + READ_MOSTLY_DATA(CONFIG_X86_INTERNODE_CACHE_BYTES) /* End of data section */ _edata = .; - } + } :data #ifdef CONFIG_X86_64 #define VSYSCALL_ADDR (-10*1024*1024) -#define VSYSCALL_PHYS_ADDR ((LOADADDR(.data.read_mostly) + \ - SIZEOF(.data.read_mostly) + 4095) & ~(4095)) -#define VSYSCALL_VIRT_ADDR ((ADDR(.data.read_mostly) + \ - SIZEOF(.data.read_mostly) + 4095) & ~(4095)) +#define VSYSCALL_PHYS_ADDR ((LOADADDR(.data) + SIZEOF(.data) + \ + PAGE_SIZE - 1) & ~(PAGE_SIZE - 1)) +#define VSYSCALL_VIRT_ADDR ((ADDR(.data) + SIZEOF(.data) + \ + PAGE_SIZE - 1) & ~(PAGE_SIZE - 1)) #define VLOAD_OFFSET (VSYSCALL_ADDR - VSYSCALL_PHYS_ADDR) #define VLOAD(x) (ADDR(x) - VLOAD_OFFSET) @@ -227,35 +204,29 @@ SECTIONS #endif /* CONFIG_X86_64 */ - /* init_task */ - . = ALIGN(THREAD_SIZE); - .data.init_task : AT(ADDR(.data.init_task) - LOAD_OFFSET) { - *(.data.init_task) + /* Init code and data - will be freed after init */ + . = ALIGN(PAGE_SIZE); + .init.begin : AT(ADDR(.init.begin) - LOAD_OFFSET) { + __init_begin = .; /* paired with __init_end */ } -#ifdef CONFIG_X86_64 - :data.init -#endif +#if defined(CONFIG_X86_64) && defined(CONFIG_SMP) /* - * smp_locks might be freed after init - * start/end must be page aligned + * percpu offsets are zero-based on SMP. PERCPU_VADDR() changes the + * output PHDR, so the next output section - .init.text - should + * start another segment - init. */ - . = ALIGN(PAGE_SIZE); - .smp_locks : AT(ADDR(.smp_locks) - LOAD_OFFSET) { - __smp_locks = .; - *(.smp_locks) - __smp_locks_end = .; - . = ALIGN(PAGE_SIZE); - } + PERCPU_VADDR(0, :percpu) +#endif - /* Init code and data - will be freed after init */ - . = ALIGN(PAGE_SIZE); .init.text : AT(ADDR(.init.text) - LOAD_OFFSET) { - __init_begin = .; /* paired with __init_end */ _sinittext = .; INIT_TEXT _einittext = .; } +#ifdef CONFIG_X86_64 + :init +#endif .init.data : AT(ADDR(.init.data) - LOAD_OFFSET) { INIT_DATA @@ -326,17 +297,7 @@ SECTIONS } #endif -#if defined(CONFIG_X86_64) && defined(CONFIG_SMP) - /* - * percpu offsets are zero-based on SMP. PERCPU_VADDR() changes the - * output PHDR, so the next output section - __data_nosave - should - * start another section data.init2. Also, pda should be at the head of - * percpu area. Preallocate it and define the percpu offset symbol - * so that it can be accessed as a percpu variable. - */ - . = ALIGN(PAGE_SIZE); - PERCPU_VADDR(0, :percpu) -#else +#if !defined(CONFIG_X86_64) || !defined(CONFIG_SMP) PERCPU(PAGE_SIZE) #endif @@ -347,15 +308,22 @@ SECTIONS __init_end = .; } + /* + * smp_locks might be freed after init + * start/end must be page aligned + */ + . = ALIGN(PAGE_SIZE); + .smp_locks : AT(ADDR(.smp_locks) - LOAD_OFFSET) { + __smp_locks = .; + *(.smp_locks) + __smp_locks_end = .; + . = ALIGN(PAGE_SIZE); + } + #ifdef CONFIG_X86_64 .data_nosave : AT(ADDR(.data_nosave) - LOAD_OFFSET) { - . = ALIGN(PAGE_SIZE); - __nosave_begin = .; - *(.data.nosave) - . = ALIGN(PAGE_SIZE); - __nosave_end = .; - } :data.init2 - /* use another section data.init2, see PERCPU_VADDR() above */ + NOSAVE_DATA + } #endif /* BSS */ @@ -393,8 +361,8 @@ SECTIONS #ifdef CONFIG_X86_32 -ASSERT((_end - LOAD_OFFSET <= KERNEL_IMAGE_SIZE), - "kernel image bigger than KERNEL_IMAGE_SIZE") +. = ASSERT((_end - LOAD_OFFSET <= KERNEL_IMAGE_SIZE), + "kernel image bigger than KERNEL_IMAGE_SIZE"); #else /* * Per-cpu symbols which need to be offset from __per_cpu_load @@ -407,12 +375,12 @@ INIT_PER_CPU(irq_stack_union); /* * Build-time check on the image size: */ -ASSERT((_end - _text <= KERNEL_IMAGE_SIZE), - "kernel image bigger than KERNEL_IMAGE_SIZE") +. = ASSERT((_end - _text <= KERNEL_IMAGE_SIZE), + "kernel image bigger than KERNEL_IMAGE_SIZE"); #ifdef CONFIG_SMP -ASSERT((per_cpu__irq_stack_union == 0), - "irq_stack_union is not at start of per-cpu area"); +. = ASSERT((per_cpu__irq_stack_union == 0), + "irq_stack_union is not at start of per-cpu area"); #endif #endif /* CONFIG_X86_32 */ @@ -420,7 +388,7 @@ ASSERT((per_cpu__irq_stack_union == 0), #ifdef CONFIG_KEXEC #include <asm/kexec.h> -ASSERT(kexec_control_code_size <= KEXEC_CONTROL_CODE_MAX_SIZE, - "kexec control code size is too big") +. = ASSERT(kexec_control_code_size <= KEXEC_CONTROL_CODE_MAX_SIZE, + "kexec control code size is too big"); #endif diff --git a/arch/x86/kvm/i8254.c b/arch/x86/kvm/i8254.c index 4d6f0d293ee..21f68e00524 100644 --- a/arch/x86/kvm/i8254.c +++ b/arch/x86/kvm/i8254.c @@ -104,6 +104,9 @@ static s64 __kpit_elapsed(struct kvm *kvm) ktime_t remaining; struct kvm_kpit_state *ps = &kvm->arch.vpit->pit_state; + if (!ps->pit_timer.period) + return 0; + /* * The Counter does not stop when it reaches zero. In * Modes 0, 1, 4, and 5 the Counter ``wraps around'' to diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c index 7030b5f911b..0ef5bb2b404 100644 --- a/arch/x86/kvm/mmu.c +++ b/arch/x86/kvm/mmu.c @@ -489,16 +489,20 @@ static unsigned long *gfn_to_rmap(struct kvm *kvm, gfn_t gfn, int lpage) * * If rmapp bit zero is one, (then rmap & ~1) points to a struct kvm_rmap_desc * containing more mappings. + * + * Returns the number of rmap entries before the spte was added or zero if + * the spte was not added. + * */ -static void rmap_add(struct kvm_vcpu *vcpu, u64 *spte, gfn_t gfn, int lpage) +static int rmap_add(struct kvm_vcpu *vcpu, u64 *spte, gfn_t gfn, int lpage) { struct kvm_mmu_page *sp; struct kvm_rmap_desc *desc; unsigned long *rmapp; - int i; + int i, count = 0; if (!is_rmap_pte(*spte)) - return; + return count; gfn = unalias_gfn(vcpu->kvm, gfn); sp = page_header(__pa(spte)); sp->gfns[spte - sp->spt] = gfn; @@ -515,8 +519,10 @@ static void rmap_add(struct kvm_vcpu *vcpu, u64 *spte, gfn_t gfn, int lpage) } else { rmap_printk("rmap_add: %p %llx many->many\n", spte, *spte); desc = (struct kvm_rmap_desc *)(*rmapp & ~1ul); - while (desc->shadow_ptes[RMAP_EXT-1] && desc->more) + while (desc->shadow_ptes[RMAP_EXT-1] && desc->more) { desc = desc->more; + count += RMAP_EXT; + } if (desc->shadow_ptes[RMAP_EXT-1]) { desc->more = mmu_alloc_rmap_desc(vcpu); desc = desc->more; @@ -525,6 +531,7 @@ static void rmap_add(struct kvm_vcpu *vcpu, u64 *spte, gfn_t gfn, int lpage) ; desc->shadow_ptes[i] = spte; } + return count; } static void rmap_desc_remove_entry(unsigned long *rmapp, @@ -754,6 +761,19 @@ static int kvm_age_rmapp(struct kvm *kvm, unsigned long *rmapp) return young; } +#define RMAP_RECYCLE_THRESHOLD 1000 + +static void rmap_recycle(struct kvm_vcpu *vcpu, gfn_t gfn, int lpage) +{ + unsigned long *rmapp; + + gfn = unalias_gfn(vcpu->kvm, gfn); + rmapp = gfn_to_rmap(vcpu->kvm, gfn, lpage); + + kvm_unmap_rmapp(vcpu->kvm, rmapp); + kvm_flush_remote_tlbs(vcpu->kvm); +} + int kvm_age_hva(struct kvm *kvm, unsigned long hva) { return kvm_handle_hva(kvm, hva, kvm_age_rmapp); @@ -1407,24 +1427,25 @@ static int kvm_mmu_zap_page(struct kvm *kvm, struct kvm_mmu_page *sp) */ void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages) { + int used_pages; + + used_pages = kvm->arch.n_alloc_mmu_pages - kvm->arch.n_free_mmu_pages; + used_pages = max(0, used_pages); + /* * If we set the number of mmu pages to be smaller be than the * number of actived pages , we must to free some mmu pages before we * change the value */ - if ((kvm->arch.n_alloc_mmu_pages - kvm->arch.n_free_mmu_pages) > - kvm_nr_mmu_pages) { - int n_used_mmu_pages = kvm->arch.n_alloc_mmu_pages - - kvm->arch.n_free_mmu_pages; - - while (n_used_mmu_pages > kvm_nr_mmu_pages) { + if (used_pages > kvm_nr_mmu_pages) { + while (used_pages > kvm_nr_mmu_pages) { struct kvm_mmu_page *page; page = container_of(kvm->arch.active_mmu_pages.prev, struct kvm_mmu_page, link); kvm_mmu_zap_page(kvm, page); - n_used_mmu_pages--; + used_pages--; } kvm->arch.n_free_mmu_pages = 0; } @@ -1740,6 +1761,7 @@ static void mmu_set_spte(struct kvm_vcpu *vcpu, u64 *shadow_pte, { int was_rmapped = 0; int was_writeble = is_writeble_pte(*shadow_pte); + int rmap_count; pgprintk("%s: spte %llx access %x write_fault %d" " user_fault %d gfn %lx\n", @@ -1781,9 +1803,11 @@ static void mmu_set_spte(struct kvm_vcpu *vcpu, u64 *shadow_pte, page_header_update_slot(vcpu->kvm, shadow_pte, gfn); if (!was_rmapped) { - rmap_add(vcpu, shadow_pte, gfn, largepage); + rmap_count = rmap_add(vcpu, shadow_pte, gfn, largepage); if (!is_rmap_pte(*shadow_pte)) kvm_release_pfn_clean(pfn); + if (rmap_count > RMAP_RECYCLE_THRESHOLD) + rmap_recycle(vcpu, gfn, largepage); } else { if (was_writeble) kvm_release_pfn_dirty(pfn); diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c index 71510e07e69..b1f658ad2f0 100644 --- a/arch/x86/kvm/svm.c +++ b/arch/x86/kvm/svm.c @@ -711,6 +711,7 @@ static void svm_vcpu_load(struct kvm_vcpu *vcpu, int cpu) svm->vmcb->control.tsc_offset += delta; vcpu->cpu = cpu; kvm_migrate_timers(vcpu); + svm->asid_generation = 0; } for (i = 0; i < NR_HOST_SAVE_USER_MSRS; i++) @@ -1031,7 +1032,6 @@ static void new_asid(struct vcpu_svm *svm, struct svm_cpu_data *svm_data) svm->vmcb->control.tlb_ctl = TLB_CONTROL_FLUSH_ALL_ASID; } - svm->vcpu.cpu = svm_data->cpu; svm->asid_generation = svm_data->asid_generation; svm->vmcb->control.asid = svm_data->next_asid++; } @@ -2300,8 +2300,8 @@ static void pre_svm_run(struct vcpu_svm *svm) struct svm_cpu_data *svm_data = per_cpu(svm_data, cpu); svm->vmcb->control.tlb_ctl = TLB_CONTROL_DO_NOTHING; - if (svm->vcpu.cpu != cpu || - svm->asid_generation != svm_data->asid_generation) + /* FIXME: handle wraparound of asid_generation */ + if (svm->asid_generation != svm_data->asid_generation) new_asid(svm, svm_data); } diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c index 356a0ce85c6..29f912927a5 100644 --- a/arch/x86/kvm/vmx.c +++ b/arch/x86/kvm/vmx.c @@ -3157,8 +3157,8 @@ static void handle_invalid_guest_state(struct kvm_vcpu *vcpu, struct vcpu_vmx *vmx = to_vmx(vcpu); enum emulation_result err = EMULATE_DONE; - preempt_enable(); local_irq_enable(); + preempt_enable(); while (!guest_state_valid(vcpu)) { err = emulate_instruction(vcpu, kvm_run, 0, 0, 0); @@ -3168,7 +3168,7 @@ static void handle_invalid_guest_state(struct kvm_vcpu *vcpu, if (err != EMULATE_DONE) { kvm_report_emulation_failure(vcpu, "emulation failure"); - return; + break; } if (signal_pending(current)) @@ -3177,8 +3177,8 @@ static void handle_invalid_guest_state(struct kvm_vcpu *vcpu, schedule(); } - local_irq_disable(); preempt_disable(); + local_irq_disable(); vmx->invalid_state_emulation_result = err; } diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index fe5474aec41..3d452901182 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -704,11 +704,48 @@ static bool msr_mtrr_valid(unsigned msr) return false; } +static bool valid_pat_type(unsigned t) +{ + return t < 8 && (1 << t) & 0xf3; /* 0, 1, 4, 5, 6, 7 */ +} + +static bool valid_mtrr_type(unsigned t) +{ + return t < 8 && (1 << t) & 0x73; /* 0, 1, 4, 5, 6 */ +} + +static bool mtrr_valid(struct kvm_vcpu *vcpu, u32 msr, u64 data) +{ + int i; + + if (!msr_mtrr_valid(msr)) + return false; + + if (msr == MSR_IA32_CR_PAT) { + for (i = 0; i < 8; i++) + if (!valid_pat_type((data >> (i * 8)) & 0xff)) + return false; + return true; + } else if (msr == MSR_MTRRdefType) { + if (data & ~0xcff) + return false; + return valid_mtrr_type(data & 0xff); + } else if (msr >= MSR_MTRRfix64K_00000 && msr <= MSR_MTRRfix4K_F8000) { + for (i = 0; i < 8 ; i++) + if (!valid_mtrr_type((data >> (i * 8)) & 0xff)) + return false; + return true; + } + + /* variable MTRRs */ + return valid_mtrr_type(data & 0xff); +} + static int set_msr_mtrr(struct kvm_vcpu *vcpu, u32 msr, u64 data) { u64 *p = (u64 *)&vcpu->arch.mtrr_state.fixed_ranges; - if (!msr_mtrr_valid(msr)) + if (!mtrr_valid(vcpu, msr, data)) return 1; if (msr == MSR_MTRRdefType) { @@ -1079,14 +1116,13 @@ long kvm_arch_dev_ioctl(struct file *filp, if (copy_to_user(user_msr_list, &msr_list, sizeof msr_list)) goto out; r = -E2BIG; - if (n < num_msrs_to_save) + if (n < msr_list.nmsrs) goto out; r = -EFAULT; if (copy_to_user(user_msr_list->indices, &msrs_to_save, num_msrs_to_save * sizeof(u32))) goto out; - if (copy_to_user(user_msr_list->indices - + num_msrs_to_save * sizeof(u32), + if (copy_to_user(user_msr_list->indices + num_msrs_to_save, &emulated_msrs, ARRAY_SIZE(emulated_msrs) * sizeof(u32))) goto out; diff --git a/arch/x86/lguest/boot.c b/arch/x86/lguest/boot.c index f2bf1f73d46..d677fa9ca65 100644 --- a/arch/x86/lguest/boot.c +++ b/arch/x86/lguest/boot.c @@ -22,7 +22,8 @@ * * So how does the kernel know it's a Guest? We'll see that later, but let's * just say that we end up here where we replace the native functions various - * "paravirt" structures with our Guest versions, then boot like normal. :*/ + * "paravirt" structures with our Guest versions, then boot like normal. +:*/ /* * Copyright (C) 2006, Rusty Russell <rusty@rustcorp.com.au> IBM Corporation. @@ -74,7 +75,8 @@ * * The Guest in our tale is a simple creature: identical to the Host but * behaving in simplified but equivalent ways. In particular, the Guest is the - * same kernel as the Host (or at least, built from the same source code). :*/ + * same kernel as the Host (or at least, built from the same source code). +:*/ struct lguest_data lguest_data = { .hcall_status = { [0 ... LHCALL_RING_SIZE-1] = 0xFF }, @@ -85,7 +87,8 @@ struct lguest_data lguest_data = { .syscall_vec = SYSCALL_VECTOR, }; -/*G:037 async_hcall() is pretty simple: I'm quite proud of it really. We have a +/*G:037 + * async_hcall() is pretty simple: I'm quite proud of it really. We have a * ring buffer of stored hypercalls which the Host will run though next time we * do a normal hypercall. Each entry in the ring has 5 slots for the hypercall * arguments, and a "hcall_status" word which is 0 if the call is ready to go, @@ -94,7 +97,8 @@ struct lguest_data lguest_data = { * If we come around to a slot which hasn't been finished, then the table is * full and we just make the hypercall directly. This has the nice side * effect of causing the Host to run all the stored calls in the ring buffer - * which empties it for next time! */ + * which empties it for next time! + */ static void async_hcall(unsigned long call, unsigned long arg1, unsigned long arg2, unsigned long arg3, unsigned long arg4) @@ -103,9 +107,11 @@ static void async_hcall(unsigned long call, unsigned long arg1, static unsigned int next_call; unsigned long flags; - /* Disable interrupts if not already disabled: we don't want an + /* + * Disable interrupts if not already disabled: we don't want an * interrupt handler making a hypercall while we're already doing - * one! */ + * one! + */ local_irq_save(flags); if (lguest_data.hcall_status[next_call] != 0xFF) { /* Table full, so do normal hcall which will flush table. */ @@ -125,8 +131,9 @@ static void async_hcall(unsigned long call, unsigned long arg1, local_irq_restore(flags); } -/*G:035 Notice the lazy_hcall() above, rather than hcall(). This is our first - * real optimization trick! +/*G:035 + * Notice the lazy_hcall() above, rather than hcall(). This is our first real + * optimization trick! * * When lazy_mode is set, it means we're allowed to defer all hypercalls and do * them as a batch when lazy_mode is eventually turned off. Because hypercalls @@ -136,7 +143,8 @@ static void async_hcall(unsigned long call, unsigned long arg1, * lguest_leave_lazy_mode(). * * So, when we're in lazy mode, we call async_hcall() to store the call for - * future processing: */ + * future processing: + */ static void lazy_hcall1(unsigned long call, unsigned long arg1) { @@ -146,6 +154,7 @@ static void lazy_hcall1(unsigned long call, async_hcall(call, arg1, 0, 0, 0); } +/* You can imagine what lazy_hcall2, 3 and 4 look like. :*/ static void lazy_hcall2(unsigned long call, unsigned long arg1, unsigned long arg2) @@ -181,8 +190,10 @@ static void lazy_hcall4(unsigned long call, } #endif -/* When lazy mode is turned off reset the per-cpu lazy mode variable and then - * issue the do-nothing hypercall to flush any stored calls. */ +/*G:036 + * When lazy mode is turned off reset the per-cpu lazy mode variable and then + * issue the do-nothing hypercall to flush any stored calls. +:*/ static void lguest_leave_lazy_mmu_mode(void) { kvm_hypercall0(LHCALL_FLUSH_ASYNC); @@ -208,9 +219,11 @@ static void lguest_end_context_switch(struct task_struct *next) * check there before it tries to deliver an interrupt. */ -/* save_flags() is expected to return the processor state (ie. "flags"). The +/* + * save_flags() is expected to return the processor state (ie. "flags"). The * flags word contains all kind of stuff, but in practice Linux only cares - * about the interrupt flag. Our "save_flags()" just returns that. */ + * about the interrupt flag. Our "save_flags()" just returns that. + */ static unsigned long save_fl(void) { return lguest_data.irq_enabled; @@ -222,13 +235,15 @@ static void irq_disable(void) lguest_data.irq_enabled = 0; } -/* Let's pause a moment. Remember how I said these are called so often? +/* + * Let's pause a moment. Remember how I said these are called so often? * Jeremy Fitzhardinge optimized them so hard early in 2009 that he had to * break some rules. In particular, these functions are assumed to save their * own registers if they need to: normal C functions assume they can trash the * eax register. To use normal C functions, we use * PV_CALLEE_SAVE_REGS_THUNK(), which pushes %eax onto the stack, calls the - * C function, then restores it. */ + * C function, then restores it. + */ PV_CALLEE_SAVE_REGS_THUNK(save_fl); PV_CALLEE_SAVE_REGS_THUNK(irq_disable); /*:*/ @@ -237,18 +252,18 @@ PV_CALLEE_SAVE_REGS_THUNK(irq_disable); extern void lg_irq_enable(void); extern void lg_restore_fl(unsigned long flags); -/*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, - * since interrupts are rare and we'll just get the interrupt on the next timer - * tick, but now we can run with CONFIG_NO_HZ, we should revisit this. One way - * would be to put the "irq_enabled" field in a page by itself, and have the - * Host write-protect it when an interrupt comes in when irqs are disabled. - * There will then be a page fault as soon as interrupts are re-enabled. +/*M:003 + * We could be more efficient in our checking of outstanding interrupts, rather + * than using a branch. One way would be to put the "irq_enabled" field in a + * page by itself, and have the Host write-protect it when an interrupt comes + * in when irqs are disabled. There will then be a page fault as soon as + * interrupts are re-enabled. * * A better method is to implement soft interrupt disable generally for x86: * instead of disabling interrupts, we set a flag. If an interrupt does come * in, we then disable them for real. This is uncommon, so we could simply use - * a hypercall for interrupt control and not worry about efficiency. :*/ + * a hypercall for interrupt control and not worry about efficiency. +:*/ /*G:034 * The Interrupt Descriptor Table (IDT). @@ -261,10 +276,12 @@ extern void lg_restore_fl(unsigned long flags); static void lguest_write_idt_entry(gate_desc *dt, int entrynum, const gate_desc *g) { - /* The gate_desc structure is 8 bytes long: we hand it to the Host in + /* + * The gate_desc structure is 8 bytes long: we hand it to the Host in * two 32-bit chunks. The whole 32-bit kernel used to hand descriptors * around like this; typesafety wasn't a big concern in Linux's early - * years. */ + * years. + */ u32 *desc = (u32 *)g; /* Keep the local copy up to date. */ native_write_idt_entry(dt, entrynum, g); @@ -272,9 +289,11 @@ static void lguest_write_idt_entry(gate_desc *dt, kvm_hypercall3(LHCALL_LOAD_IDT_ENTRY, entrynum, desc[0], desc[1]); } -/* Changing to a different IDT is very rare: we keep the IDT up-to-date every +/* + * Changing to a different IDT is very rare: we keep the IDT up-to-date every * time it is written, so we can simply loop through all entries and tell the - * Host about them. */ + * Host about them. + */ static void lguest_load_idt(const struct desc_ptr *desc) { unsigned int i; @@ -305,9 +324,11 @@ static void lguest_load_gdt(const struct desc_ptr *desc) kvm_hypercall3(LHCALL_LOAD_GDT_ENTRY, i, gdt[i].a, gdt[i].b); } -/* For a single GDT entry which changes, we do the lazy thing: alter our GDT, +/* + * For a single GDT entry which changes, we do the lazy thing: alter our GDT, * then tell the Host to reload the entire thing. This operation is so rare - * that this naive implementation is reasonable. */ + * that this naive implementation is reasonable. + */ static void lguest_write_gdt_entry(struct desc_struct *dt, int entrynum, const void *desc, int type) { @@ -317,29 +338,36 @@ static void lguest_write_gdt_entry(struct desc_struct *dt, int entrynum, dt[entrynum].a, dt[entrynum].b); } -/* OK, I lied. There are three "thread local storage" GDT entries which change +/* + * OK, I lied. There are three "thread local storage" GDT entries which change * on every context switch (these three entries are how glibc implements - * __thread variables). So we have a hypercall specifically for this case. */ + * __thread variables). So we have a hypercall specifically for this case. + */ static void lguest_load_tls(struct thread_struct *t, unsigned int cpu) { - /* There's one problem which normal hardware doesn't have: the Host + /* + * 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. */ + * the GS register here: if it's needed it'll be reloaded anyway. + */ lazy_load_gs(0); lazy_hcall2(LHCALL_LOAD_TLS, __pa(&t->tls_array), cpu); } -/*G:038 That's enough excitement for now, back to ploughing through each of - * the different pv_ops structures (we're about 1/3 of the way through). +/*G:038 + * That's enough excitement for now, back to ploughing through each of the + * different pv_ops structures (we're about 1/3 of the way through). * * This is the Local Descriptor Table, another weird Intel thingy. Linux only * uses this for some strange applications like Wine. We don't do anything - * here, so they'll get an informative and friendly Segmentation Fault. */ + * here, so they'll get an informative and friendly Segmentation Fault. + */ static void lguest_set_ldt(const void *addr, unsigned entries) { } -/* This loads a GDT entry into the "Task Register": that entry points to a +/* + * This loads a GDT entry into the "Task Register": that entry points to a * structure called the Task State Segment. Some comments scattered though the * kernel code indicate that this used for task switching in ages past, along * with blood sacrifice and astrology. @@ -347,19 +375,21 @@ static void lguest_set_ldt(const void *addr, unsigned entries) * Now there's nothing interesting in here that we don't get told elsewhere. * But the native version uses the "ltr" instruction, which makes the Host * complain to the Guest about a Segmentation Fault and it'll oops. So we - * override the native version with a do-nothing version. */ + * override the native version with a do-nothing version. + */ static void lguest_load_tr_desc(void) { } -/* The "cpuid" instruction is a way of querying both the CPU identity +/* + * The "cpuid" instruction is a way of querying both the CPU identity * (manufacturer, model, etc) and its features. It was introduced before the * Pentium in 1993 and keeps getting extended by both Intel, AMD and others. * As you might imagine, after a decade and a half this treatment, it is now a * giant ball of hair. Its entry in the current Intel manual runs to 28 pages. * * This instruction even it has its own Wikipedia entry. The Wikipedia entry - * has been translated into 4 languages. I am not making this up! + * has been translated into 5 languages. I am not making this up! * * We could get funky here and identify ourselves as "GenuineLguest", but * instead we just use the real "cpuid" instruction. Then I pretty much turned @@ -371,7 +401,8 @@ static void lguest_load_tr_desc(void) * Replacing the cpuid so we can turn features off is great for the kernel, but * anyone (including userspace) can just use the raw "cpuid" instruction and * the Host won't even notice since it isn't privileged. So we try not to get - * too worked up about it. */ + * too worked up about it. + */ static void lguest_cpuid(unsigned int *ax, unsigned int *bx, unsigned int *cx, unsigned int *dx) { @@ -379,43 +410,63 @@ static void lguest_cpuid(unsigned int *ax, unsigned int *bx, native_cpuid(ax, bx, cx, dx); switch (function) { - case 0: /* ID and highest CPUID. Futureproof a little by sticking to - * older ones. */ + /* + * CPUID 0 gives the highest legal CPUID number (and the ID string). + * We futureproof our code a little by sticking to known CPUID values. + */ + case 0: if (*ax > 5) *ax = 5; break; - case 1: /* Basic feature request. */ - /* We only allow kernel to see SSE3, CMPXCHG16B and SSSE3 */ + + /* + * CPUID 1 is a basic feature request. + * + * CX: we only allow kernel to see SSE3, CMPXCHG16B and SSSE3 + * DX: SSE, SSE2, FXSR, MMX, CMOV, CMPXCHG8B, TSC, FPU and PAE. + */ + case 1: *cx &= 0x00002201; - /* SSE, SSE2, FXSR, MMX, CMOV, CMPXCHG8B, TSC, FPU, PAE. */ *dx &= 0x07808151; - /* The Host can do a nice optimization if it knows that the + /* + * The Host can do a nice optimization if it knows that the * kernel mappings (addresses above 0xC0000000 or whatever * PAGE_OFFSET is set to) haven't changed. But Linux calls * flush_tlb_user() for both user and kernel mappings unless - * the Page Global Enable (PGE) feature bit is set. */ + * the Page Global Enable (PGE) feature bit is set. + */ *dx |= 0x00002000; - /* We also lie, and say we're family id 5. 6 or greater + /* + * We also lie, and say we're family id 5. 6 or greater * leads to a rdmsr in early_init_intel which we can't handle. - * Family ID is returned as bits 8-12 in ax. */ + * Family ID is returned as bits 8-12 in ax. + */ *ax &= 0xFFFFF0FF; *ax |= 0x00000500; break; + /* + * 0x80000000 returns the highest Extended Function, so we futureproof + * like we do above by limiting it to known fields. + */ case 0x80000000: - /* Futureproof this a little: if they ask how much extended - * processor information there is, limit it to known fields. */ if (*ax > 0x80000008) *ax = 0x80000008; break; + + /* + * PAE systems can mark pages as non-executable. Linux calls this the + * NX bit. Intel calls it XD (eXecute Disable), AMD EVP (Enhanced + * Virus Protection). We just switch turn if off here, since we don't + * support it. + */ case 0x80000001: - /* Here we should fix nx cap depending on host. */ - /* For this version of PAE, we just clear NX bit. */ *dx &= ~(1 << 20); break; } } -/* Intel has four control registers, imaginatively named cr0, cr2, cr3 and cr4. +/* + * Intel has four control registers, imaginatively named cr0, cr2, cr3 and cr4. * I assume there's a cr1, but it hasn't bothered us yet, so we'll not bother * it. The Host needs to know when the Guest wants to change them, so we have * a whole series of functions like read_cr0() and write_cr0(). @@ -430,7 +481,8 @@ static void lguest_cpuid(unsigned int *ax, unsigned int *bx, * name like "FPUTRAP bit" be a little less cryptic? * * We store cr0 locally because the Host never changes it. The Guest sometimes - * wants to read it and we'd prefer not to bother the Host unnecessarily. */ + * wants to read it and we'd prefer not to bother the Host unnecessarily. + */ static unsigned long current_cr0; static void lguest_write_cr0(unsigned long val) { @@ -443,18 +495,22 @@ static unsigned long lguest_read_cr0(void) return current_cr0; } -/* Intel provided a special instruction to clear the TS bit for people too cool +/* + * Intel provided a special instruction to clear the TS bit for people too cool * to use write_cr0() to do it. This "clts" instruction is faster, because all - * the vowels have been optimized out. */ + * the vowels have been optimized out. + */ static void lguest_clts(void) { lazy_hcall1(LHCALL_TS, 0); current_cr0 &= ~X86_CR0_TS; } -/* cr2 is the virtual address of the last page fault, which the Guest only ever +/* + * cr2 is the virtual address of the last page fault, which the Guest only ever * reads. The Host kindly writes this into our "struct lguest_data", so we - * just read it out of there. */ + * just read it out of there. + */ static unsigned long lguest_read_cr2(void) { return lguest_data.cr2; @@ -463,10 +519,12 @@ static unsigned long lguest_read_cr2(void) /* See lguest_set_pte() below. */ static bool cr3_changed = false; -/* cr3 is the current toplevel pagetable page: the principle is the same as +/* + * cr3 is the current toplevel pagetable page: the principle is the same as * cr0. Keep a local copy, and tell the Host when it changes. The only * difference is that our local copy is in lguest_data because the Host needs - * to set it upon our initial hypercall. */ + * to set it upon our initial hypercall. + */ static void lguest_write_cr3(unsigned long cr3) { lguest_data.pgdir = cr3; @@ -511,7 +569,7 @@ static void lguest_write_cr4(unsigned long val) * cr3 ---> +---------+ * | --------->+---------+ * | | | PADDR1 | - * Top-level | | PADDR2 | + * Mid-level | | PADDR2 | * (PMD) page | | | * | | Lower-level | * | | (PTE) page | @@ -531,21 +589,62 @@ static void lguest_write_cr4(unsigned long val) * Index into top Index into second Offset within page * page directory page pagetable page * - * The kernel spends a lot of time changing both the top-level page directory - * and lower-level pagetable pages. The Guest doesn't know physical addresses, - * so while it maintains these page tables exactly like normal, it also needs - * to keep the Host informed whenever it makes a change: the Host will create - * the real page tables based on the Guests'. + * Now, unfortunately, this isn't the whole story: Intel added Physical Address + * Extension (PAE) to allow 32 bit systems to use 64GB of memory (ie. 36 bits). + * These are held in 64-bit page table entries, so we can now only fit 512 + * entries in a page, and the neat three-level tree breaks down. + * + * The result is a four level page table: + * + * cr3 --> [ 4 Upper ] + * [ Level ] + * [ Entries ] + * [(PUD Page)]---> +---------+ + * | --------->+---------+ + * | | | PADDR1 | + * Mid-level | | PADDR2 | + * (PMD) page | | | + * | | Lower-level | + * | | (PTE) page | + * | | | | + * .... .... + * + * + * And the virtual address is decoded as: + * + * 1 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 + * |<-2->|<--- 9 bits ---->|<---- 9 bits --->|<------ 12 bits ------>| + * Index into Index into mid Index into lower Offset within page + * top entries directory page pagetable page + * + * It's too hard to switch between these two formats at runtime, so Linux only + * supports one or the other depending on whether CONFIG_X86_PAE is set. Many + * distributions turn it on, and not just for people with silly amounts of + * memory: the larger PTE entries allow room for the NX bit, which lets the + * kernel disable execution of pages and increase security. + * + * This was a problem for lguest, which couldn't run on these distributions; + * then Matias Zabaljauregui figured it all out and implemented it, and only a + * handful of puppies were crushed in the process! + * + * Back to our point: the kernel spends a lot of time changing both the + * top-level page directory and lower-level pagetable pages. The Guest doesn't + * know physical addresses, so while it maintains these page tables exactly + * like normal, it also needs to keep the Host informed whenever it makes a + * change: the Host will create the real page tables based on the Guests'. */ -/* The Guest calls this to set a second-level entry (pte), ie. to map a page - * into a process' address space. We set the entry then tell the Host the - * toplevel and address this corresponds to. The Guest uses one pagetable per - * process, so we need to tell the Host which one we're changing (mm->pgd). */ +/* + * The Guest calls this after it has set a second-level entry (pte), ie. to map + * a page into a process' address space. Wetell the Host the toplevel and + * address this corresponds to. The Guest uses one pagetable per process, so + * we need to tell the Host which one we're changing (mm->pgd). + */ static void lguest_pte_update(struct mm_struct *mm, unsigned long addr, pte_t *ptep) { #ifdef CONFIG_X86_PAE + /* PAE needs to hand a 64 bit page table entry, so it uses two args. */ lazy_hcall4(LHCALL_SET_PTE, __pa(mm->pgd), addr, ptep->pte_low, ptep->pte_high); #else @@ -553,6 +652,7 @@ static void lguest_pte_update(struct mm_struct *mm, unsigned long addr, #endif } +/* This is the "set and update" combo-meal-deal version. */ static void lguest_set_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pteval) { @@ -560,10 +660,13 @@ static void lguest_set_pte_at(struct mm_struct *mm, unsigned long addr, lguest_pte_update(mm, addr, ptep); } -/* The Guest calls lguest_set_pud to set a top-level entry and lguest_set_pmd +/* + * The Guest calls lguest_set_pud to set a top-level entry and lguest_set_pmd * to set a middle-level entry when PAE is activated. + * * Again, we set the entry then tell the Host which page we changed, - * and the index of the entry we changed. */ + * and the index of the entry we changed. + */ #ifdef CONFIG_X86_PAE static void lguest_set_pud(pud_t *pudp, pud_t pudval) { @@ -582,8 +685,7 @@ static void lguest_set_pmd(pmd_t *pmdp, pmd_t pmdval) } #else -/* The Guest calls lguest_set_pmd to set a top-level entry when PAE is not - * activated. */ +/* The Guest calls lguest_set_pmd to set a top-level entry when !PAE. */ static void lguest_set_pmd(pmd_t *pmdp, pmd_t pmdval) { native_set_pmd(pmdp, pmdval); @@ -592,7 +694,8 @@ static void lguest_set_pmd(pmd_t *pmdp, pmd_t pmdval) } #endif -/* There are a couple of legacy places where the kernel sets a PTE, but we +/* + * There are a couple of legacy places where the kernel sets a PTE, but we * don't know the top level any more. This is useless for us, since we don't * know which pagetable is changing or what address, so we just tell the Host * to forget all of them. Fortunately, this is very rare. @@ -600,7 +703,8 @@ static void lguest_set_pmd(pmd_t *pmdp, pmd_t pmdval) * ... except in early boot when the kernel sets up the initial pagetables, * which makes booting astonishingly slow: 1.83 seconds! So we don't even tell * the Host anything changed until we've done the first page table switch, - * which brings boot back to 0.25 seconds. */ + * which brings boot back to 0.25 seconds. + */ static void lguest_set_pte(pte_t *ptep, pte_t pteval) { native_set_pte(ptep, pteval); @@ -609,6 +713,11 @@ static void lguest_set_pte(pte_t *ptep, pte_t pteval) } #ifdef CONFIG_X86_PAE +/* + * With 64-bit PTE values, we need to be careful setting them: if we set 32 + * bits at a time, the hardware could see a weird half-set entry. These + * versions ensure we update all 64 bits at once. + */ static void lguest_set_pte_atomic(pte_t *ptep, pte_t pte) { native_set_pte_atomic(ptep, pte); @@ -616,19 +725,21 @@ static void lguest_set_pte_atomic(pte_t *ptep, pte_t pte) lazy_hcall1(LHCALL_FLUSH_TLB, 1); } -void lguest_pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep) +static void lguest_pte_clear(struct mm_struct *mm, unsigned long addr, + pte_t *ptep) { native_pte_clear(mm, addr, ptep); lguest_pte_update(mm, addr, ptep); } -void lguest_pmd_clear(pmd_t *pmdp) +static void lguest_pmd_clear(pmd_t *pmdp) { lguest_set_pmd(pmdp, __pmd(0)); } #endif -/* Unfortunately for Lguest, the pv_mmu_ops for page tables were based on +/* + * Unfortunately for Lguest, the pv_mmu_ops for page tables were based on * native page table operations. On native hardware you can set a new page * table entry whenever you want, but if you want to remove one you have to do * a TLB flush (a TLB is a little cache of page table entries kept by the CPU). @@ -637,24 +748,29 @@ void lguest_pmd_clear(pmd_t *pmdp) * called when a valid entry is written, not when it's removed (ie. marked not * present). Instead, this is where we come when the Guest wants to remove a * page table entry: we tell the Host to set that entry to 0 (ie. the present - * bit is zero). */ + * bit is zero). + */ static void lguest_flush_tlb_single(unsigned long addr) { /* Simply set it to zero: if it was not, it will fault back in. */ lazy_hcall3(LHCALL_SET_PTE, lguest_data.pgdir, addr, 0); } -/* This is what happens after the Guest has removed a large number of entries. +/* + * This is what happens after the Guest has removed a large number of entries. * This tells the Host that any of the page table entries for userspace might - * have changed, ie. virtual addresses below PAGE_OFFSET. */ + * have changed, ie. virtual addresses below PAGE_OFFSET. + */ static void lguest_flush_tlb_user(void) { lazy_hcall1(LHCALL_FLUSH_TLB, 0); } -/* This is called when the kernel page tables have changed. That's not very +/* + * This is called when the kernel page tables have changed. That's not very * common (unless the Guest is using highmem, which makes the Guest extremely - * slow), so it's worth separating this from the user flushing above. */ + * slow), so it's worth separating this from the user flushing above. + */ static void lguest_flush_tlb_kernel(void) { lazy_hcall1(LHCALL_FLUSH_TLB, 1); @@ -691,26 +807,38 @@ static struct irq_chip lguest_irq_controller = { .unmask = enable_lguest_irq, }; -/* This sets up the Interrupt Descriptor Table (IDT) entry for each hardware +/* + * This sets up the Interrupt Descriptor Table (IDT) entry for each hardware * interrupt (except 128, which is used for system calls), and then tells the * Linux infrastructure that each interrupt is controlled by our level-based - * lguest interrupt controller. */ + * lguest interrupt controller. + */ static void __init lguest_init_IRQ(void) { unsigned int i; for (i = FIRST_EXTERNAL_VECTOR; i < NR_VECTORS; i++) { - /* Some systems map "vectors" to interrupts weirdly. Lguest has - * a straightforward 1 to 1 mapping, so force that here. */ + /* Some systems map "vectors" to interrupts weirdly. Not us! */ __get_cpu_var(vector_irq)[i] = i - FIRST_EXTERNAL_VECTOR; if (i != SYSCALL_VECTOR) set_intr_gate(i, interrupt[i - FIRST_EXTERNAL_VECTOR]); } - /* This call is required to set up for 4k stacks, where we have - * separate stacks for hard and soft interrupts. */ + + /* + * This call is required to set up for 4k stacks, where we have + * separate stacks for hard and soft interrupts. + */ irq_ctx_init(smp_processor_id()); } +/* + * With CONFIG_SPARSE_IRQ, interrupt descriptors are allocated as-needed, so + * rather than set them in lguest_init_IRQ we are called here every time an + * lguest device needs an interrupt. + * + * FIXME: irq_to_desc_alloc_node() can fail due to lack of memory, we should + * pass that up! + */ void lguest_setup_irq(unsigned int irq) { irq_to_desc_alloc_node(irq, 0); @@ -729,31 +857,39 @@ static unsigned long lguest_get_wallclock(void) return lguest_data.time.tv_sec; } -/* The TSC is an Intel thing called the Time Stamp Counter. The Host tells us +/* + * The TSC is an Intel thing called the Time Stamp Counter. The Host tells us * what speed it runs at, or 0 if it's unusable as a reliable clock source. * This matches what we want here: if we return 0 from this function, the x86 - * TSC clock will give up and not register itself. */ + * TSC clock will give up and not register itself. + */ static unsigned long lguest_tsc_khz(void) { return lguest_data.tsc_khz; } -/* If we can't use the TSC, the kernel falls back to our lower-priority - * "lguest_clock", where we read the time value given to us by the Host. */ +/* + * If we can't use the TSC, the kernel falls back to our lower-priority + * "lguest_clock", where we read the time value given to us by the Host. + */ static cycle_t lguest_clock_read(struct clocksource *cs) { unsigned long sec, nsec; - /* Since the time is in two parts (seconds and nanoseconds), we risk + /* + * Since the time is in two parts (seconds and nanoseconds), we risk * reading it just as it's changing from 99 & 0.999999999 to 100 and 0, * and getting 99 and 0. As Linux tends to come apart under the stress - * of time travel, we must be careful: */ + * of time travel, we must be careful: + */ do { /* First we read the seconds part. */ sec = lguest_data.time.tv_sec; - /* This read memory barrier tells the compiler and the CPU that + /* + * This read memory barrier tells the compiler and the CPU that * this can't be reordered: we have to complete the above - * before going on. */ + * before going on. + */ rmb(); /* Now we read the nanoseconds part. */ nsec = lguest_data.time.tv_nsec; @@ -777,9 +913,11 @@ static struct clocksource lguest_clock = { .flags = CLOCK_SOURCE_IS_CONTINUOUS, }; -/* We also need a "struct clock_event_device": Linux asks us to set it to go +/* + * We also need a "struct clock_event_device": Linux asks us to set it to go * off some time in the future. Actually, James Morris figured all this out, I - * just applied the patch. */ + * just applied the patch. + */ static int lguest_clockevent_set_next_event(unsigned long delta, struct clock_event_device *evt) { @@ -829,8 +967,10 @@ static struct clock_event_device lguest_clockevent = { .max_delta_ns = LG_CLOCK_MAX_DELTA, }; -/* This is the Guest timer interrupt handler (hardware interrupt 0). We just - * call the clockevent infrastructure and it does whatever needs doing. */ +/* + * This is the Guest timer interrupt handler (hardware interrupt 0). We just + * call the clockevent infrastructure and it does whatever needs doing. + */ static void lguest_time_irq(unsigned int irq, struct irq_desc *desc) { unsigned long flags; @@ -841,10 +981,12 @@ static void lguest_time_irq(unsigned int irq, struct irq_desc *desc) local_irq_restore(flags); } -/* At some point in the boot process, we get asked to set up our timing +/* + * At some point in the boot process, we get asked to set up our timing * infrastructure. The kernel doesn't expect timer interrupts before this, but * we cleverly initialized the "blocked_interrupts" field of "struct - * lguest_data" so that timer interrupts were blocked until now. */ + * lguest_data" so that timer interrupts were blocked until now. + */ static void lguest_time_init(void) { /* Set up the timer interrupt (0) to go to our simple timer routine */ @@ -868,14 +1010,16 @@ static void lguest_time_init(void) * to work. They're pretty simple. */ -/* The Guest needs to tell the Host what stack it expects traps to use. For +/* + * The Guest needs to tell the Host what stack it expects traps to use. For * native hardware, this is part of the Task State Segment mentioned above in * lguest_load_tr_desc(), but to help hypervisors there's this special call. * * We tell the Host the segment we want to use (__KERNEL_DS is the kernel data * segment), the privilege level (we're privilege level 1, the Host is 0 and * will not tolerate us trying to use that), the stack pointer, and the number - * of pages in the stack. */ + * of pages in the stack. + */ static void lguest_load_sp0(struct tss_struct *tss, struct thread_struct *thread) { @@ -889,7 +1033,8 @@ static void lguest_set_debugreg(int regno, unsigned long value) /* FIXME: Implement */ } -/* There are times when the kernel wants to make sure that no memory writes are +/* + * There are times when the kernel wants to make sure that no memory writes are * caught in the cache (that they've all reached real hardware devices). This * doesn't matter for the Guest which has virtual hardware. * @@ -903,11 +1048,13 @@ static void lguest_wbinvd(void) { } -/* If the Guest expects to have an Advanced Programmable Interrupt Controller, +/* + * If the Guest expects to have an Advanced Programmable Interrupt Controller, * we play dumb by ignoring writes and returning 0 for reads. So it's no * longer Programmable nor Controlling anything, and I don't think 8 lines of * code qualifies for Advanced. It will also never interrupt anything. It - * does, however, allow us to get through the Linux boot code. */ + * does, however, allow us to get through the Linux boot code. + */ #ifdef CONFIG_X86_LOCAL_APIC static void lguest_apic_write(u32 reg, u32 v) { @@ -956,11 +1103,13 @@ static void lguest_safe_halt(void) kvm_hypercall0(LHCALL_HALT); } -/* The SHUTDOWN hypercall takes a string to describe what's happening, and +/* + * The SHUTDOWN hypercall takes a string to describe what's happening, and * an argument which says whether this to restart (reboot) the Guest or not. * * Note that the Host always prefers that the Guest speak in physical addresses - * rather than virtual addresses, so we use __pa() here. */ + * rather than virtual addresses, so we use __pa() here. + */ static void lguest_power_off(void) { kvm_hypercall2(LHCALL_SHUTDOWN, __pa("Power down"), @@ -991,8 +1140,10 @@ static __init char *lguest_memory_setup(void) * nice to move it back to lguest_init. Patch welcome... */ atomic_notifier_chain_register(&panic_notifier_list, &paniced); - /* The Linux bootloader header contains an "e820" memory map: the - * Launcher populated the first entry with our memory limit. */ + /* + *The Linux bootloader header contains an "e820" memory map: the + * Launcher populated the first entry with our memory limit. + */ e820_add_region(boot_params.e820_map[0].addr, boot_params.e820_map[0].size, boot_params.e820_map[0].type); @@ -1001,16 +1152,17 @@ static __init char *lguest_memory_setup(void) return "LGUEST"; } -/* We will eventually use the virtio console device to produce console output, +/* + * We will eventually use the virtio console device to produce console output, * but before that is set up we use LHCALL_NOTIFY on normal memory to produce - * console output. */ + * console output. + */ static __init int early_put_chars(u32 vtermno, const char *buf, int count) { char scratch[17]; unsigned int len = count; - /* We use a nul-terminated string, so we have to make a copy. Icky, - * huh? */ + /* We use a nul-terminated string, so we make a copy. Icky, huh? */ if (len > sizeof(scratch) - 1) len = sizeof(scratch) - 1; scratch[len] = '\0'; @@ -1021,8 +1173,10 @@ static __init int early_put_chars(u32 vtermno, const char *buf, int count) return len; } -/* Rebooting also tells the Host we're finished, but the RESTART flag tells the - * Launcher to reboot us. */ +/* + * Rebooting also tells the Host we're finished, but the RESTART flag tells the + * Launcher to reboot us. + */ static void lguest_restart(char *reason) { kvm_hypercall2(LHCALL_SHUTDOWN, __pa(reason), LGUEST_SHUTDOWN_RESTART); @@ -1049,7 +1203,8 @@ static void lguest_restart(char *reason) * fit comfortably. * * First we need assembly templates of each of the patchable Guest operations, - * and these are in i386_head.S. */ + * and these are in i386_head.S. + */ /*G:060 We construct a table from the assembler templates: */ static const struct lguest_insns @@ -1060,9 +1215,11 @@ static const struct lguest_insns [PARAVIRT_PATCH(pv_irq_ops.save_fl)] = { lgstart_pushf, lgend_pushf }, }; -/* Now our patch routine is fairly simple (based on the native one in +/* + * Now our patch routine is fairly simple (based on the native one in * paravirt.c). If we have a replacement, we copy it in and return how much of - * the available space we used. */ + * the available space we used. + */ static unsigned lguest_patch(u8 type, u16 clobber, void *ibuf, unsigned long addr, unsigned len) { @@ -1074,8 +1231,7 @@ static unsigned lguest_patch(u8 type, u16 clobber, void *ibuf, insn_len = lguest_insns[type].end - lguest_insns[type].start; - /* Similarly if we can't fit replacement (shouldn't happen, but let's - * be thorough). */ + /* Similarly if it can't fit (doesn't happen, but let's be thorough). */ if (len < insn_len) return paravirt_patch_default(type, clobber, ibuf, addr, len); @@ -1084,22 +1240,28 @@ static unsigned lguest_patch(u8 type, u16 clobber, void *ibuf, return insn_len; } -/*G:029 Once we get to lguest_init(), we know we're a Guest. The various +/*G:029 + * Once we get to lguest_init(), we know we're a Guest. The various * pv_ops structures in the kernel provide points for (almost) every routine we - * have to override to avoid privileged instructions. */ + * have to override to avoid privileged instructions. + */ __init void lguest_init(void) { - /* We're under lguest, paravirt is enabled, and we're running at - * privilege level 1, not 0 as normal. */ + /* We're under lguest. */ pv_info.name = "lguest"; + /* Paravirt is enabled. */ pv_info.paravirt_enabled = 1; + /* We're running at privilege level 1, not 0 as normal. */ pv_info.kernel_rpl = 1; + /* Everyone except Xen runs with this set. */ pv_info.shared_kernel_pmd = 1; - /* We set up all the lguest overrides for sensitive operations. These - * are detailed with the operations themselves. */ + /* + * We set up all the lguest overrides for sensitive operations. These + * are detailed with the operations themselves. + */ - /* interrupt-related operations */ + /* Interrupt-related operations */ pv_irq_ops.init_IRQ = lguest_init_IRQ; pv_irq_ops.save_fl = PV_CALLEE_SAVE(save_fl); pv_irq_ops.restore_fl = __PV_IS_CALLEE_SAVE(lg_restore_fl); @@ -1107,11 +1269,11 @@ __init void lguest_init(void) pv_irq_ops.irq_enable = __PV_IS_CALLEE_SAVE(lg_irq_enable); pv_irq_ops.safe_halt = lguest_safe_halt; - /* init-time operations */ + /* Setup operations */ pv_init_ops.memory_setup = lguest_memory_setup; pv_init_ops.patch = lguest_patch; - /* Intercepts of various cpu instructions */ + /* Intercepts of various CPU instructions */ pv_cpu_ops.load_gdt = lguest_load_gdt; pv_cpu_ops.cpuid = lguest_cpuid; pv_cpu_ops.load_idt = lguest_load_idt; @@ -1132,7 +1294,7 @@ __init void lguest_init(void) pv_cpu_ops.start_context_switch = paravirt_start_context_switch; pv_cpu_ops.end_context_switch = lguest_end_context_switch; - /* pagetable management */ + /* Pagetable management */ pv_mmu_ops.write_cr3 = lguest_write_cr3; pv_mmu_ops.flush_tlb_user = lguest_flush_tlb_user; pv_mmu_ops.flush_tlb_single = lguest_flush_tlb_single; @@ -1154,54 +1316,71 @@ __init void lguest_init(void) pv_mmu_ops.pte_update_defer = lguest_pte_update; #ifdef CONFIG_X86_LOCAL_APIC - /* apic read/write intercepts */ + /* APIC read/write intercepts */ set_lguest_basic_apic_ops(); #endif - /* time operations */ + /* Time operations */ pv_time_ops.get_wallclock = lguest_get_wallclock; pv_time_ops.time_init = lguest_time_init; pv_time_ops.get_tsc_khz = lguest_tsc_khz; - /* Now is a good time to look at the implementations of these functions - * before returning to the rest of lguest_init(). */ + /* + * Now is a good time to look at the implementations of these functions + * before returning to the rest of lguest_init(). + */ - /*G:070 Now we've seen all the paravirt_ops, we return to + /*G:070 + * Now we've seen all the paravirt_ops, we return to * lguest_init() where the rest of the fairly chaotic boot setup - * occurs. */ + * occurs. + */ - /* The stack protector is a weird thing where gcc places a canary + /* + * The stack protector is a weird thing where gcc places a canary * value on the stack and then checks it on return. This file is * compiled with -fno-stack-protector it, so we got this far without * problems. The value of the canary is kept at offset 20 from the * %gs register, so we need to set that up before calling C functions - * in other files. */ + * in other files. + */ setup_stack_canary_segment(0); - /* We could just call load_stack_canary_segment(), but we might as - * call switch_to_new_gdt() which loads the whole table and sets up - * the per-cpu segment descriptor register %fs as well. */ + + /* + * We could just call load_stack_canary_segment(), but we might as well + * call switch_to_new_gdt() which loads the whole table and sets up the + * per-cpu segment descriptor register %fs as well. + */ switch_to_new_gdt(0); - /* As described in head_32.S, we map the first 128M of memory. */ + /* We actually boot with all memory mapped, but let's say 128MB. */ max_pfn_mapped = (128*1024*1024) >> PAGE_SHIFT; - /* The Host<->Guest Switcher lives at the top of our address space, and + /* + * The Host<->Guest Switcher lives at the top of our address space, and * the Host told us how big it is when we made LGUEST_INIT hypercall: - * it put the answer in lguest_data.reserve_mem */ + * it put the answer in lguest_data.reserve_mem + */ reserve_top_address(lguest_data.reserve_mem); - /* If we don't initialize the lock dependency checker now, it crashes - * paravirt_disable_iospace. */ + /* + * If we don't initialize the lock dependency checker now, it crashes + * paravirt_disable_iospace. + */ lockdep_init(); - /* The IDE code spends about 3 seconds probing for disks: if we reserve + /* + * The IDE code spends about 3 seconds probing for disks: if we reserve * all the I/O ports up front it can't get them and so doesn't probe. * Other device drivers are similar (but less severe). This cuts the - * kernel boot time on my machine from 4.1 seconds to 0.45 seconds. */ + * kernel boot time on my machine from 4.1 seconds to 0.45 seconds. + */ paravirt_disable_iospace(); - /* This is messy CPU setup stuff which the native boot code does before - * start_kernel, so we have to do, too: */ + /* + * This is messy CPU setup stuff which the native boot code does before + * start_kernel, so we have to do, too: + */ cpu_detect(&new_cpu_data); /* head.S usually sets up the first capability word, so do it here. */ new_cpu_data.x86_capability[0] = cpuid_edx(1); @@ -1218,22 +1397,28 @@ __init void lguest_init(void) acpi_ht = 0; #endif - /* We set the preferred console to "hvc". This is the "hypervisor + /* + * We set the preferred console to "hvc". This is the "hypervisor * virtual console" driver written by the PowerPC people, which we also - * adapted for lguest's use. */ + * adapted for lguest's use. + */ add_preferred_console("hvc", 0, NULL); /* Register our very early console. */ virtio_cons_early_init(early_put_chars); - /* Last of all, we set the power management poweroff hook to point to + /* + * Last of all, we set the power management poweroff hook to point to * the Guest routine to power off, and the reboot hook to our restart - * routine. */ + * routine. + */ pm_power_off = lguest_power_off; machine_ops.restart = lguest_restart; - /* Now we're set up, call i386_start_kernel() in head32.c and we proceed - * to boot as normal. It never returns. */ + /* + * Now we're set up, call i386_start_kernel() in head32.c and we proceed + * to boot as normal. It never returns. + */ i386_start_kernel(); } /* diff --git a/arch/x86/lguest/i386_head.S b/arch/x86/lguest/i386_head.S index a9c8cfe61cd..27eac0faee4 100644 --- a/arch/x86/lguest/i386_head.S +++ b/arch/x86/lguest/i386_head.S @@ -5,7 +5,8 @@ #include <asm/thread_info.h> #include <asm/processor-flags.h> -/*G:020 Our story starts with the kernel booting into startup_32 in +/*G:020 + * Our story starts with the kernel booting into startup_32 in * arch/x86/kernel/head_32.S. It expects a boot header, which is created by * the bootloader (the Launcher in our case). * @@ -21,11 +22,14 @@ * data without remembering to subtract __PAGE_OFFSET! * * The .section line puts this code in .init.text so it will be discarded after - * boot. */ + * boot. + */ .section .init.text, "ax", @progbits ENTRY(lguest_entry) - /* We make the "initialization" hypercall now to tell the Host about - * us, and also find out where it put our page tables. */ + /* + * We make the "initialization" hypercall now to tell the Host about + * us, and also find out where it put our page tables. + */ movl $LHCALL_LGUEST_INIT, %eax movl $lguest_data - __PAGE_OFFSET, %ebx .byte 0x0f,0x01,0xc1 /* KVM_HYPERCALL */ @@ -33,13 +37,14 @@ ENTRY(lguest_entry) /* Set up the initial stack so we can run C code. */ movl $(init_thread_union+THREAD_SIZE),%esp - /* Jumps are relative, and we're running __PAGE_OFFSET too low at the - * moment. */ + /* Jumps are relative: we're running __PAGE_OFFSET too low. */ jmp lguest_init+__PAGE_OFFSET -/*G:055 We create a macro which puts the assembler code between lgstart_ and - * lgend_ markers. These templates are put in the .text section: they can't be - * discarded after boot as we may need to patch modules, too. */ +/*G:055 + * We create a macro which puts the assembler code between lgstart_ and lgend_ + * markers. These templates are put in the .text section: they can't be + * discarded after boot as we may need to patch modules, too. + */ .text #define LGUEST_PATCH(name, insns...) \ lgstart_##name: insns; lgend_##name:; \ @@ -48,83 +53,103 @@ ENTRY(lguest_entry) LGUEST_PATCH(cli, movl $0, lguest_data+LGUEST_DATA_irq_enabled) LGUEST_PATCH(pushf, movl lguest_data+LGUEST_DATA_irq_enabled, %eax) -/*G:033 But using those wrappers is inefficient (we'll see why that doesn't - * matter for save_fl and irq_disable later). If we write our routines - * carefully in assembler, we can avoid clobbering any registers and avoid - * jumping through the wrapper functions. +/*G:033 + * But using those wrappers is inefficient (we'll see why that doesn't matter + * for save_fl and irq_disable later). If we write our routines carefully in + * assembler, we can avoid clobbering any registers and avoid jumping through + * the wrapper functions. * * I skipped over our first piece of assembler, but this one is worth studying - * in a bit more detail so I'll describe in easy stages. First, the routine - * to enable interrupts: */ + * in a bit more detail so I'll describe in easy stages. First, the routine to + * enable interrupts: + */ ENTRY(lg_irq_enable) - /* The reverse of irq_disable, this sets lguest_data.irq_enabled to - * X86_EFLAGS_IF (ie. "Interrupts enabled"). */ + /* + * The reverse of irq_disable, this sets lguest_data.irq_enabled to + * X86_EFLAGS_IF (ie. "Interrupts enabled"). + */ movl $X86_EFLAGS_IF, lguest_data+LGUEST_DATA_irq_enabled - /* But now we need to check if the Host wants to know: there might have + /* + * But now we need to check if the Host wants to know: there might have * been interrupts waiting to be delivered, in which case it will have * set lguest_data.irq_pending to X86_EFLAGS_IF. If it's not zero, we - * jump to send_interrupts, otherwise we're done. */ + * jump to send_interrupts, otherwise we're done. + */ testl $0, lguest_data+LGUEST_DATA_irq_pending jnz send_interrupts - /* One cool thing about x86 is that you can do many things without using + /* + * One cool thing about x86 is that you can do many things without using * a register. In this case, the normal path hasn't needed to save or - * restore any registers at all! */ + * restore any registers at all! + */ ret send_interrupts: - /* OK, now we need a register: eax is used for the hypercall number, + /* + * OK, now we need a register: eax is used for the hypercall number, * which is LHCALL_SEND_INTERRUPTS. * * We used not to bother with this pending detection at all, which was * much simpler. Sooner or later the Host would realize it had to * send us an interrupt. But that turns out to make performance 7 * times worse on a simple tcp benchmark. So now we do this the hard - * way. */ + * way. + */ pushl %eax movl $LHCALL_SEND_INTERRUPTS, %eax - /* This is a vmcall instruction (same thing that KVM uses). Older + /* + * This is a vmcall instruction (same thing that KVM uses). Older * assembler versions might not know the "vmcall" instruction, so we - * create one manually here. */ + * create one manually here. + */ .byte 0x0f,0x01,0xc1 /* KVM_HYPERCALL */ + /* Put eax back the way we found it. */ popl %eax ret -/* Finally, the "popf" or "restore flags" routine. The %eax register holds the +/* + * Finally, the "popf" or "restore flags" routine. The %eax register holds the * flags (in practice, either X86_EFLAGS_IF or 0): if it's X86_EFLAGS_IF we're - * enabling interrupts again, if it's 0 we're leaving them off. */ + * enabling interrupts again, if it's 0 we're leaving them off. + */ ENTRY(lg_restore_fl) /* This is just "lguest_data.irq_enabled = flags;" */ movl %eax, lguest_data+LGUEST_DATA_irq_enabled - /* Now, if the %eax value has enabled interrupts and + /* + * Now, if the %eax value has enabled interrupts and * lguest_data.irq_pending is set, we want to tell the Host so it can * deliver any outstanding interrupts. Fortunately, both values will * be X86_EFLAGS_IF (ie. 512) in that case, and the "testl" * instruction will AND them together for us. If both are set, we - * jump to send_interrupts. */ + * jump to send_interrupts. + */ testl lguest_data+LGUEST_DATA_irq_pending, %eax jnz send_interrupts /* Again, the normal path has used no extra registers. Clever, huh? */ ret +/*:*/ /* These demark the EIP range where host should never deliver interrupts. */ .global lguest_noirq_start .global lguest_noirq_end -/*M:004 When the Host reflects a trap or injects an interrupt into the Guest, - * it sets the eflags interrupt bit on the stack based on - * lguest_data.irq_enabled, so the Guest iret logic does the right thing when - * restoring it. However, when the Host sets the Guest up for direct traps, - * such as system calls, the processor is the one to push eflags onto the - * stack, and the interrupt bit will be 1 (in reality, interrupts are always - * enabled in the Guest). +/*M:004 + * When the Host reflects a trap or injects an interrupt into the Guest, it + * sets the eflags interrupt bit on the stack based on lguest_data.irq_enabled, + * so the Guest iret logic does the right thing when restoring it. However, + * when the Host sets the Guest up for direct traps, such as system calls, the + * processor is the one to push eflags onto the stack, and the interrupt bit + * will be 1 (in reality, interrupts are always enabled in the Guest). * * This turns out to be harmless: the only trap which should happen under Linux * with interrupts disabled is Page Fault (due to our lazy mapping of vmalloc * regions), which has to be reflected through the Host anyway. If another * trap *does* go off when interrupts are disabled, the Guest will panic, and - * we'll never get to this iret! :*/ + * we'll never get to this iret! +:*/ -/*G:045 There is one final paravirt_op that the Guest implements, and glancing - * at it you can see why I left it to last. It's *cool*! It's in *assembler*! +/*G:045 + * There is one final paravirt_op that the Guest implements, and glancing at it + * you can see why I left it to last. It's *cool*! It's in *assembler*! * * The "iret" instruction is used to return from an interrupt or trap. The * stack looks like this: @@ -148,15 +173,18 @@ ENTRY(lg_restore_fl) * return to userspace or wherever. Our solution to this is to surround the * code with lguest_noirq_start: and lguest_noirq_end: labels. We tell the * Host that it is *never* to interrupt us there, even if interrupts seem to be - * enabled. */ + * enabled. + */ ENTRY(lguest_iret) pushl %eax movl 12(%esp), %eax lguest_noirq_start: - /* Note the %ss: segment prefix here. Normal data accesses use the + /* + * Note the %ss: segment prefix here. Normal data accesses use the * "ds" segment, but that will have already been restored for whatever * we're returning to (such as userspace): we can't trust it. The %ss: - * prefix makes sure we use the stack segment, which is still valid. */ + * prefix makes sure we use the stack segment, which is still valid. + */ movl %eax,%ss:lguest_data+LGUEST_DATA_irq_enabled popl %eax iret diff --git a/arch/x86/lib/msr.c b/arch/x86/lib/msr.c index 1440b9c0547..caa24aca811 100644 --- a/arch/x86/lib/msr.c +++ b/arch/x86/lib/msr.c @@ -89,16 +89,13 @@ void rdmsr_on_cpus(const cpumask_t *mask, u32 msr_no, struct msr *msrs) rv.msrs = msrs; rv.msr_no = msr_no; - preempt_disable(); - /* - * FIXME: handle the CPU we're executing on separately for now until - * smp_call_function_many has been fixed to not skip it. - */ - this_cpu = raw_smp_processor_id(); - smp_call_function_single(this_cpu, __rdmsr_on_cpu, &rv, 1); + this_cpu = get_cpu(); + + if (cpumask_test_cpu(this_cpu, mask)) + __rdmsr_on_cpu(&rv); smp_call_function_many(mask, __rdmsr_on_cpu, &rv, 1); - preempt_enable(); + put_cpu(); } EXPORT_SYMBOL(rdmsr_on_cpus); @@ -121,16 +118,13 @@ void wrmsr_on_cpus(const cpumask_t *mask, u32 msr_no, struct msr *msrs) rv.msrs = msrs; rv.msr_no = msr_no; - preempt_disable(); - /* - * FIXME: handle the CPU we're executing on separately for now until - * smp_call_function_many has been fixed to not skip it. - */ - this_cpu = raw_smp_processor_id(); - smp_call_function_single(this_cpu, __wrmsr_on_cpu, &rv, 1); + this_cpu = get_cpu(); + + if (cpumask_test_cpu(this_cpu, mask)) + __wrmsr_on_cpu(&rv); smp_call_function_many(mask, __wrmsr_on_cpu, &rv, 1); - preempt_enable(); + put_cpu(); } EXPORT_SYMBOL(wrmsr_on_cpus); diff --git a/arch/x86/mm/highmem_32.c b/arch/x86/mm/highmem_32.c index 58f621e8191..2112ed55e7e 100644 --- a/arch/x86/mm/highmem_32.c +++ b/arch/x86/mm/highmem_32.c @@ -103,6 +103,7 @@ EXPORT_SYMBOL(kmap); EXPORT_SYMBOL(kunmap); EXPORT_SYMBOL(kmap_atomic); EXPORT_SYMBOL(kunmap_atomic); +EXPORT_SYMBOL(kmap_atomic_prot); void __init set_highmem_pages_init(void) { diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c index 6176fe8f29e..ea56b8cbb6a 100644 --- a/arch/x86/mm/init_64.c +++ b/arch/x86/mm/init_64.c @@ -796,7 +796,7 @@ int __init reserve_bootmem_generic(unsigned long phys, unsigned long len, return ret; #else - reserve_bootmem(phys, len, BOOTMEM_DEFAULT); + reserve_bootmem(phys, len, flags); #endif if (phys+len <= MAX_DMA_PFN*PAGE_SIZE) { diff --git a/arch/x86/mm/pageattr.c b/arch/x86/mm/pageattr.c index 1b734d7a896..7e600c1962d 100644 --- a/arch/x86/mm/pageattr.c +++ b/arch/x86/mm/pageattr.c @@ -591,9 +591,12 @@ static int __change_page_attr(struct cpa_data *cpa, int primary) unsigned int level; pte_t *kpte, old_pte; - if (cpa->flags & CPA_PAGES_ARRAY) - address = (unsigned long)page_address(cpa->pages[cpa->curpage]); - else if (cpa->flags & CPA_ARRAY) + if (cpa->flags & CPA_PAGES_ARRAY) { + struct page *page = cpa->pages[cpa->curpage]; + if (unlikely(PageHighMem(page))) + return 0; + address = (unsigned long)page_address(page); + } else if (cpa->flags & CPA_ARRAY) address = cpa->vaddr[cpa->curpage]; else address = *cpa->vaddr; @@ -697,9 +700,12 @@ static int cpa_process_alias(struct cpa_data *cpa) * No need to redo, when the primary call touched the direct * mapping already: */ - if (cpa->flags & CPA_PAGES_ARRAY) - vaddr = (unsigned long)page_address(cpa->pages[cpa->curpage]); - else if (cpa->flags & CPA_ARRAY) + if (cpa->flags & CPA_PAGES_ARRAY) { + struct page *page = cpa->pages[cpa->curpage]; + if (unlikely(PageHighMem(page))) + return 0; + vaddr = (unsigned long)page_address(page); + } else if (cpa->flags & CPA_ARRAY) vaddr = cpa->vaddr[cpa->curpage]; else vaddr = *cpa->vaddr; @@ -997,12 +1003,15 @@ EXPORT_SYMBOL(set_memory_array_uc); int _set_memory_wc(unsigned long addr, int numpages) { int ret; + unsigned long addr_copy = addr; + ret = change_page_attr_set(&addr, numpages, __pgprot(_PAGE_CACHE_UC_MINUS), 0); - if (!ret) { - ret = change_page_attr_set(&addr, numpages, - __pgprot(_PAGE_CACHE_WC), 0); + ret = change_page_attr_set_clr(&addr_copy, numpages, + __pgprot(_PAGE_CACHE_WC), + __pgprot(_PAGE_CACHE_MASK), + 0, 0, NULL); } return ret; } @@ -1119,7 +1128,9 @@ int set_pages_array_uc(struct page **pages, int addrinarray) int free_idx; for (i = 0; i < addrinarray; i++) { - start = (unsigned long)page_address(pages[i]); + if (PageHighMem(pages[i])) + continue; + start = page_to_pfn(pages[i]) << PAGE_SHIFT; end = start + PAGE_SIZE; if (reserve_memtype(start, end, _PAGE_CACHE_UC_MINUS, NULL)) goto err_out; @@ -1132,7 +1143,9 @@ int set_pages_array_uc(struct page **pages, int addrinarray) err_out: free_idx = i; for (i = 0; i < free_idx; i++) { - start = (unsigned long)page_address(pages[i]); + if (PageHighMem(pages[i])) + continue; + start = page_to_pfn(pages[i]) << PAGE_SHIFT; end = start + PAGE_SIZE; free_memtype(start, end); } @@ -1161,7 +1174,9 @@ int set_pages_array_wb(struct page **pages, int addrinarray) return retval; for (i = 0; i < addrinarray; i++) { - start = (unsigned long)page_address(pages[i]); + if (PageHighMem(pages[i])) + continue; + start = page_to_pfn(pages[i]) << PAGE_SHIFT; end = start + PAGE_SIZE; free_memtype(start, end); } diff --git a/arch/x86/mm/pat.c b/arch/x86/mm/pat.c index e6718bb2806..352aa9e927e 100644 --- a/arch/x86/mm/pat.c +++ b/arch/x86/mm/pat.c @@ -623,7 +623,8 @@ static int reserve_pfn_range(u64 paddr, unsigned long size, pgprot_t *vma_prot, return ret; if (flags != want_flags) { - if (strict_prot || !is_new_memtype_allowed(want_flags, flags)) { + if (strict_prot || + !is_new_memtype_allowed(paddr, size, want_flags, flags)) { free_memtype(paddr, paddr + size); printk(KERN_ERR "%s:%d map pfn expected mapping type %s" " for %Lx-%Lx, got %s\n", diff --git a/arch/x86/mm/pgtable.c b/arch/x86/mm/pgtable.c index af8f9650058..ed34f5e3599 100644 --- a/arch/x86/mm/pgtable.c +++ b/arch/x86/mm/pgtable.c @@ -329,7 +329,6 @@ void __init reserve_top_address(unsigned long reserve) printk(KERN_INFO "Reserving virtual address space above 0x%08x\n", (int)-reserve); __FIXADDR_TOP = -reserve - PAGE_SIZE; - __VMALLOC_RESERVE += reserve; #endif } diff --git a/arch/x86/mm/tlb.c b/arch/x86/mm/tlb.c index 821e97017e9..c814e144a3f 100644 --- a/arch/x86/mm/tlb.c +++ b/arch/x86/mm/tlb.c @@ -183,18 +183,17 @@ static void flush_tlb_others_ipi(const struct cpumask *cpumask, f->flush_mm = mm; f->flush_va = va; - cpumask_andnot(to_cpumask(f->flush_cpumask), - cpumask, cpumask_of(smp_processor_id())); - - /* - * We have to send the IPI only to - * CPUs affected. - */ - apic->send_IPI_mask(to_cpumask(f->flush_cpumask), - INVALIDATE_TLB_VECTOR_START + sender); + if (cpumask_andnot(to_cpumask(f->flush_cpumask), cpumask, cpumask_of(smp_processor_id()))) { + /* + * We have to send the IPI only to + * CPUs affected. + */ + apic->send_IPI_mask(to_cpumask(f->flush_cpumask), + INVALIDATE_TLB_VECTOR_START + sender); - while (!cpumask_empty(to_cpumask(f->flush_cpumask))) - cpu_relax(); + while (!cpumask_empty(to_cpumask(f->flush_cpumask))) + cpu_relax(); + } f->flush_mm = NULL; f->flush_va = 0; diff --git a/arch/x86/xen/Makefile b/arch/x86/xen/Makefile index 172438f86a0..7410640db17 100644 --- a/arch/x86/xen/Makefile +++ b/arch/x86/xen/Makefile @@ -5,6 +5,10 @@ CFLAGS_REMOVE_time.o = -pg CFLAGS_REMOVE_irq.o = -pg endif +# Make sure early boot has no stackprotector +nostackp := $(call cc-option, -fno-stack-protector) +CFLAGS_enlighten.o := $(nostackp) + obj-y := enlighten.o setup.o multicalls.o mmu.o irq.o \ time.o xen-asm.o xen-asm_$(BITS).o \ grant-table.o suspend.o diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 0a1700a2be9..eb33aaa8415 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -215,6 +215,7 @@ static __init void xen_init_cpuid_mask(void) (1 << X86_FEATURE_ACPI)); /* disable ACPI */ ax = 1; + cx = 0; xen_cpuid(&ax, &bx, &cx, &dx); /* cpuid claims we support xsave; try enabling it to see what happens */ @@ -974,10 +975,6 @@ asmlinkage void __init xen_start_kernel(void) xen_domain_type = XEN_PV_DOMAIN; - BUG_ON(memcmp(xen_start_info->magic, "xen-3", 5) != 0); - - xen_setup_features(); - /* Install Xen paravirt ops */ pv_info = xen_info; pv_init_ops = xen_init_ops; @@ -986,8 +983,15 @@ asmlinkage void __init xen_start_kernel(void) pv_apic_ops = xen_apic_ops; pv_mmu_ops = xen_mmu_ops; - xen_init_irq_ops(); +#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 + xen_init_irq_ops(); xen_init_cpuid_mask(); #ifdef CONFIG_X86_LOCAL_APIC @@ -997,6 +1001,8 @@ asmlinkage void __init xen_start_kernel(void) set_xen_basic_apic_ops(); #endif + xen_setup_features(); + if (xen_feature(XENFEAT_mmu_pt_update_preserve_ad)) { pv_mmu_ops.ptep_modify_prot_start = xen_ptep_modify_prot_start; pv_mmu_ops.ptep_modify_prot_commit = xen_ptep_modify_prot_commit; @@ -1004,13 +1010,6 @@ asmlinkage void __init xen_start_kernel(void) machine_ops = xen_machine_ops; -#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 @@ -1061,6 +1060,7 @@ asmlinkage void __init xen_start_kernel(void) /* set up basic CPUID stuff */ cpu_detect(&new_cpu_data); new_cpu_data.hard_math = 1; + new_cpu_data.wp_works_ok = 1; new_cpu_data.x86_capability[0] = cpuid_edx(1); #endif |