From 3843fc2575e3389f4f0ad0420a720240a5746a5d Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Fri, 9 May 2008 12:05:57 +0100 Subject: xen: remove support for non-PAE 32-bit Non-PAE operation has been deprecated in Xen for a while, and is rarely tested or used. xen-unstable has now officially dropped non-PAE support. Since Xen/pvops' non-PAE support has also been broken for a while, we may as well completely drop it altogether. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Ingo Molnar Signed-off-by: Thomas Gleixner --- arch/x86/xen/mmu.c | 19 ++----------------- 1 file changed, 2 insertions(+), 17 deletions(-) (limited to 'arch/x86/xen/mmu.c') diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 126766d43aea..07c2653ec335 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -222,7 +222,7 @@ pmdval_t xen_pmd_val(pmd_t pmd) ret = machine_to_phys(XMADDR(ret)).paddr | _PAGE_PRESENT; return ret; } -#ifdef CONFIG_X86_PAE + void xen_set_pud(pud_t *ptr, pud_t val) { struct multicall_space mcs; @@ -272,12 +272,6 @@ pmd_t xen_make_pmd(pmdval_t pmd) return native_make_pmd(pmd); } -#else /* !PAE */ -void xen_set_pte(pte_t *ptep, pte_t pte) -{ - *ptep = pte; -} -#endif /* CONFIG_X86_PAE */ /* (Yet another) pagetable walker. This one is intended for pinning a @@ -430,8 +424,6 @@ static int pin_page(struct page *page, enum pt_level level) read-only, and can be pinned. */ void xen_pgd_pin(pgd_t *pgd) { - unsigned level; - xen_mc_batch(); if (pgd_walk(pgd, pin_page, TASK_SIZE)) { @@ -441,14 +433,7 @@ void xen_pgd_pin(pgd_t *pgd) xen_mc_batch(); } -#ifdef CONFIG_X86_PAE - level = MMUEXT_PIN_L3_TABLE; -#else - level = MMUEXT_PIN_L2_TABLE; -#endif - - xen_do_pin(level, PFN_DOWN(__pa(pgd))); - + xen_do_pin(MMUEXT_PIN_L3_TABLE, PFN_DOWN(__pa(pgd))); xen_mc_issue(0); } -- cgit v1.2.3 From de067814d6b69030d0030e1c5b3dbaf0385aae41 Mon Sep 17 00:00:00 2001 From: Jan Beulich Date: Thu, 15 May 2008 13:24:52 +0100 Subject: x86/xen: fix arbitrary_virt_to_machine() While I realize that the function isn't currently being used, I still think an obvious mistake like this should be corrected. Signed-off-by: Jan Beulich Acked-by: Jeremy Fitzhardinge Signed-off-by: Ingo Molnar Signed-off-by: Thomas Gleixner --- arch/x86/xen/mmu.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/x86/xen/mmu.c') diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 126766d43aea..3525ef523a74 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -60,7 +60,7 @@ xmaddr_t arbitrary_virt_to_machine(unsigned long address) { unsigned int level; pte_t *pte = lookup_address(address, &level); - unsigned offset = address & PAGE_MASK; + unsigned offset = address & ~PAGE_MASK; BUG_ON(pte == NULL); -- cgit v1.2.3 From d451bb7aa852627bdf7be7937dc3d9d9f261b235 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 26 May 2008 23:31:18 +0100 Subject: xen: make phys_to_machine structure dynamic We now support the use of memory hotplug, so the physical to machine page mapping structure must be dynamic. This is implemented as a two-level radix tree structure, which allows us to efficiently incrementally allocate memory for the p2m table as new pages are added. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Thomas Gleixner --- arch/x86/xen/enlighten.c | 2 +- arch/x86/xen/mmu.c | 85 ++++++++++++++++++++++++++++++++++++++++++++++ arch/x86/xen/setup.c | 2 -- arch/x86/xen/xen-ops.h | 2 ++ include/asm-x86/xen/page.h | 20 ++++------- 5 files changed, 94 insertions(+), 17 deletions(-) (limited to 'arch/x86/xen/mmu.c') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 5c0635a8bffd..73d3c84a3495 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1221,7 +1221,7 @@ asmlinkage void __init xen_start_kernel(void) /* Get mfn list */ if (!xen_feature(XENFEAT_auto_translated_physmap)) - phys_to_machine_mapping = (unsigned long *)xen_start_info->mfn_list; + xen_build_dynamic_phys_to_machine(); pgd = (pgd_t *)xen_start_info->pt_base; diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 07c2653ec335..c3b27dec6f03 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -56,6 +56,91 @@ #include "multicalls.h" #include "mmu.h" +/* + * This should probably be a config option. On 32-bit, it costs 1 + * page/gig of memory; on 64-bit its 2 pages/gig. If we want it to be + * completely unbounded we can add another level to the p2m structure. + */ +#define MAX_GUEST_PAGES (16ull * 1024*1024*1024 / PAGE_SIZE) +#define P2M_ENTRIES_PER_PAGE (PAGE_SIZE / sizeof(unsigned long)) + +static unsigned long *p2m_top[MAX_GUEST_PAGES / P2M_ENTRIES_PER_PAGE]; + +static inline unsigned p2m_top_index(unsigned long pfn) +{ + BUG_ON(pfn >= MAX_GUEST_PAGES); + return pfn / P2M_ENTRIES_PER_PAGE; +} + +static inline unsigned p2m_index(unsigned long pfn) +{ + return pfn % P2M_ENTRIES_PER_PAGE; +} + +void __init xen_build_dynamic_phys_to_machine(void) +{ + unsigned pfn; + unsigned long *mfn_list = (unsigned long *)xen_start_info->mfn_list; + + BUG_ON(xen_start_info->nr_pages >= MAX_GUEST_PAGES); + + for(pfn = 0; + pfn < xen_start_info->nr_pages; + pfn += P2M_ENTRIES_PER_PAGE) { + unsigned topidx = p2m_top_index(pfn); + + p2m_top[topidx] = &mfn_list[pfn]; + } +} + +unsigned long get_phys_to_machine(unsigned long pfn) +{ + unsigned topidx, idx; + + topidx = p2m_top_index(pfn); + if (p2m_top[topidx] == NULL) + return INVALID_P2M_ENTRY; + + idx = p2m_index(pfn); + return p2m_top[topidx][idx]; +} + +static void alloc_p2m(unsigned long **pp) +{ + unsigned long *p; + unsigned i; + + p = (void *)__get_free_page(GFP_KERNEL | __GFP_NOFAIL); + BUG_ON(p == NULL); + + for(i = 0; i < P2M_ENTRIES_PER_PAGE; i++) + p[i] = INVALID_P2M_ENTRY; + + if (cmpxchg(pp, NULL, p) != NULL) + free_page((unsigned long)p); +} + +void set_phys_to_machine(unsigned long pfn, unsigned long mfn) +{ + unsigned topidx, idx; + + if (unlikely(xen_feature(XENFEAT_auto_translated_physmap))) { + BUG_ON(pfn != mfn && mfn != INVALID_P2M_ENTRY); + return; + } + + topidx = p2m_top_index(pfn); + if (p2m_top[topidx] == NULL) { + /* no need to allocate a page to store an invalid entry */ + if (mfn == INVALID_P2M_ENTRY) + return; + alloc_p2m(&p2m_top[topidx]); + } + + idx = p2m_index(pfn); + p2m_top[topidx][idx] = mfn; +} + xmaddr_t arbitrary_virt_to_machine(unsigned long address) { unsigned int level; diff --git a/arch/x86/xen/setup.c b/arch/x86/xen/setup.c index 82517e4a752a..37f8f0b8f743 100644 --- a/arch/x86/xen/setup.c +++ b/arch/x86/xen/setup.c @@ -27,8 +27,6 @@ extern const char xen_hypervisor_callback[]; extern const char xen_failsafe_callback[]; -unsigned long *phys_to_machine_mapping; -EXPORT_SYMBOL(phys_to_machine_mapping); /** * machine_specific_memory_setup - Hook for machine specific memory setup. diff --git a/arch/x86/xen/xen-ops.h b/arch/x86/xen/xen-ops.h index f1063ae08037..7bdc8c5c9244 100644 --- a/arch/x86/xen/xen-ops.h +++ b/arch/x86/xen/xen-ops.h @@ -22,6 +22,8 @@ void __init xen_arch_setup(void); void __init xen_init_IRQ(void); void xen_enable_sysenter(void); +void __init xen_build_dynamic_phys_to_machine(void); + void xen_setup_timer(int cpu); void xen_setup_cpu_clockevents(void); unsigned long xen_cpu_khz(void); diff --git a/include/asm-x86/xen/page.h b/include/asm-x86/xen/page.h index e11f24038b1d..293344f8102e 100644 --- a/include/asm-x86/xen/page.h +++ b/include/asm-x86/xen/page.h @@ -26,15 +26,15 @@ typedef struct xpaddr { #define FOREIGN_FRAME_BIT (1UL<<31) #define FOREIGN_FRAME(m) ((m) | FOREIGN_FRAME_BIT) -extern unsigned long *phys_to_machine_mapping; +extern unsigned long get_phys_to_machine(unsigned long pfn); +extern void set_phys_to_machine(unsigned long pfn, unsigned long mfn); static inline unsigned long pfn_to_mfn(unsigned long pfn) { if (xen_feature(XENFEAT_auto_translated_physmap)) return pfn; - return phys_to_machine_mapping[(unsigned int)(pfn)] & - ~FOREIGN_FRAME_BIT; + return get_phys_to_machine(pfn) & ~FOREIGN_FRAME_BIT; } static inline int phys_to_machine_mapping_valid(unsigned long pfn) @@ -42,7 +42,7 @@ static inline int phys_to_machine_mapping_valid(unsigned long pfn) if (xen_feature(XENFEAT_auto_translated_physmap)) return 1; - return (phys_to_machine_mapping[pfn] != INVALID_P2M_ENTRY); + return get_phys_to_machine(pfn) != INVALID_P2M_ENTRY; } static inline unsigned long mfn_to_pfn(unsigned long mfn) @@ -106,20 +106,12 @@ static inline unsigned long mfn_to_local_pfn(unsigned long mfn) unsigned long pfn = mfn_to_pfn(mfn); if ((pfn < max_mapnr) && !xen_feature(XENFEAT_auto_translated_physmap) - && (phys_to_machine_mapping[pfn] != mfn)) + && (get_phys_to_machine(pfn) != mfn)) return max_mapnr; /* force !pfn_valid() */ + /* XXX fixme; not true with sparsemem */ return pfn; } -static inline void set_phys_to_machine(unsigned long pfn, unsigned long mfn) -{ - if (xen_feature(XENFEAT_auto_translated_physmap)) { - BUG_ON(pfn != mfn && mfn != INVALID_P2M_ENTRY); - return; - } - phys_to_machine_mapping[pfn] = mfn; -} - /* VIRT <-> MACHINE conversion */ #define virt_to_machine(v) (phys_to_machine(XPADDR(__pa(v)))) #define virt_to_mfn(v) (pfn_to_mfn(PFN_DOWN(__pa(v)))) -- cgit v1.2.3 From 8006ec3e911f93d702e1d4a4e387e244ab434924 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 26 May 2008 23:31:19 +0100 Subject: xen: add configurable max domain size Add a config option to set the max size of a Xen domain. This is used to scale the size of the physical-to-machine array; it ends up using around 1 page/GByte, so there's no reason to be very restrictive. For a 32-bit guest, the default value of 8GB is probably sufficient; there's not much point in giving a 32-bit machine much more memory than that. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Thomas Gleixner --- arch/x86/xen/Kconfig | 10 ++++++++++ arch/x86/xen/mmu.c | 25 ++++++++++++------------- arch/x86/xen/setup.c | 3 +++ include/asm-x86/xen/page.h | 5 +++++ 4 files changed, 30 insertions(+), 13 deletions(-) (limited to 'arch/x86/xen/mmu.c') diff --git a/arch/x86/xen/Kconfig b/arch/x86/xen/Kconfig index 525b108411bd..d0f1c7c5dc3d 100644 --- a/arch/x86/xen/Kconfig +++ b/arch/x86/xen/Kconfig @@ -11,3 +11,13 @@ config XEN This is the Linux Xen port. Enabling this will allow the kernel to boot in a paravirtualized environment under the Xen hypervisor. + +config XEN_MAX_DOMAIN_MEMORY + int "Maximum allowed size of a domain in gigabytes" + default 8 + depends on XEN + help + The pseudo-physical to machine address array is sized + according to the maximum possible memory size of a Xen + domain. This array uses 1 page per gigabyte, so there's no + need to be too stingy here. \ No newline at end of file diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index c3b27dec6f03..644232aa7bfb 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -56,19 +56,13 @@ #include "multicalls.h" #include "mmu.h" -/* - * This should probably be a config option. On 32-bit, it costs 1 - * page/gig of memory; on 64-bit its 2 pages/gig. If we want it to be - * completely unbounded we can add another level to the p2m structure. - */ -#define MAX_GUEST_PAGES (16ull * 1024*1024*1024 / PAGE_SIZE) #define P2M_ENTRIES_PER_PAGE (PAGE_SIZE / sizeof(unsigned long)) -static unsigned long *p2m_top[MAX_GUEST_PAGES / P2M_ENTRIES_PER_PAGE]; +static unsigned long *p2m_top[MAX_DOMAIN_PAGES / P2M_ENTRIES_PER_PAGE]; static inline unsigned p2m_top_index(unsigned long pfn) { - BUG_ON(pfn >= MAX_GUEST_PAGES); + BUG_ON(pfn >= MAX_DOMAIN_PAGES); return pfn / P2M_ENTRIES_PER_PAGE; } @@ -81,12 +75,9 @@ void __init xen_build_dynamic_phys_to_machine(void) { unsigned pfn; unsigned long *mfn_list = (unsigned long *)xen_start_info->mfn_list; + unsigned long max_pfn = min(MAX_DOMAIN_PAGES, xen_start_info->nr_pages); - BUG_ON(xen_start_info->nr_pages >= MAX_GUEST_PAGES); - - for(pfn = 0; - pfn < xen_start_info->nr_pages; - pfn += P2M_ENTRIES_PER_PAGE) { + for(pfn = 0; pfn < max_pfn; pfn += P2M_ENTRIES_PER_PAGE) { unsigned topidx = p2m_top_index(pfn); p2m_top[topidx] = &mfn_list[pfn]; @@ -97,6 +88,9 @@ unsigned long get_phys_to_machine(unsigned long pfn) { unsigned topidx, idx; + if (unlikely(pfn >= MAX_DOMAIN_PAGES)) + return INVALID_P2M_ENTRY; + topidx = p2m_top_index(pfn); if (p2m_top[topidx] == NULL) return INVALID_P2M_ENTRY; @@ -129,6 +123,11 @@ void set_phys_to_machine(unsigned long pfn, unsigned long mfn) return; } + if (unlikely(pfn >= MAX_DOMAIN_PAGES)) { + BUG_ON(mfn != INVALID_P2M_ENTRY); + return; + } + topidx = p2m_top_index(pfn); if (p2m_top[topidx] == NULL) { /* no need to allocate a page to store an invalid entry */ diff --git a/arch/x86/xen/setup.c b/arch/x86/xen/setup.c index 37f8f0b8f743..488447878a9d 100644 --- a/arch/x86/xen/setup.c +++ b/arch/x86/xen/setup.c @@ -16,6 +16,7 @@ #include #include +#include #include #include #include @@ -36,6 +37,8 @@ char * __init xen_memory_setup(void) { unsigned long max_pfn = xen_start_info->nr_pages; + max_pfn = min(MAX_DOMAIN_PAGES, max_pfn); + e820.nr_map = 0; add_memory_region(0, LOWMEMSIZE(), E820_RAM); add_memory_region(HIGH_MEMORY, PFN_PHYS(max_pfn)-HIGH_MEMORY, E820_RAM); diff --git a/include/asm-x86/xen/page.h b/include/asm-x86/xen/page.h index 293344f8102e..377c04591c15 100644 --- a/include/asm-x86/xen/page.h +++ b/include/asm-x86/xen/page.h @@ -26,6 +26,11 @@ typedef struct xpaddr { #define FOREIGN_FRAME_BIT (1UL<<31) #define FOREIGN_FRAME(m) ((m) | FOREIGN_FRAME_BIT) +/* Maximum amount of memory we can handle in a domain in pages */ +#define MAX_DOMAIN_PAGES \ + ((unsigned long)((u64)CONFIG_XEN_MAX_DOMAIN_MEMORY * 1024 * 1024 * 1024 / PAGE_SIZE)) + + extern unsigned long get_phys_to_machine(unsigned long pfn); extern void set_phys_to_machine(unsigned long pfn, unsigned long mfn); -- cgit v1.2.3 From cf0923ea295ba08ae656ef04164a43cb6553ba99 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 26 May 2008 23:31:20 +0100 Subject: xen: efficiently support a holey p2m table When using sparsemem and memory hotplug, the kernel's pseudo-physical address space can be discontigious. Previously this was dealt with by having the upper parts of the radix tree stubbed off. Unfortunately, this is incompatible with save/restore, which requires a complete p2m table. The solution is to have a special distinguished all-invalid p2m leaf page, which we can point all the hole areas at. This allows the tools to see a complete p2m table, but it only costs a page for all memory holes. It also simplifies the code since it removes a few special cases. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Thomas Gleixner --- arch/x86/xen/mmu.c | 18 ++++++++++++------ 1 file changed, 12 insertions(+), 6 deletions(-) (limited to 'arch/x86/xen/mmu.c') diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 644232aa7bfb..da7b45b05066 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -57,8 +57,17 @@ #include "mmu.h" #define P2M_ENTRIES_PER_PAGE (PAGE_SIZE / sizeof(unsigned long)) +#define TOP_ENTRIES (MAX_DOMAIN_PAGES / P2M_ENTRIES_PER_PAGE) -static unsigned long *p2m_top[MAX_DOMAIN_PAGES / P2M_ENTRIES_PER_PAGE]; +/* Placeholder for holes in the address space */ +static unsigned long p2m_missing[P2M_ENTRIES_PER_PAGE] + __attribute__((section(".data.page_aligned"))) = + { [ 0 ... P2M_ENTRIES_PER_PAGE-1 ] = ~0UL }; + + /* Array of pointers to pages containing p2m entries */ +static unsigned long *p2m_top[TOP_ENTRIES] + __attribute__((section(".data.page_aligned"))) = + { [ 0 ... TOP_ENTRIES - 1] = &p2m_missing[0] }; static inline unsigned p2m_top_index(unsigned long pfn) { @@ -92,9 +101,6 @@ unsigned long get_phys_to_machine(unsigned long pfn) return INVALID_P2M_ENTRY; topidx = p2m_top_index(pfn); - if (p2m_top[topidx] == NULL) - return INVALID_P2M_ENTRY; - idx = p2m_index(pfn); return p2m_top[topidx][idx]; } @@ -110,7 +116,7 @@ static void alloc_p2m(unsigned long **pp) for(i = 0; i < P2M_ENTRIES_PER_PAGE; i++) p[i] = INVALID_P2M_ENTRY; - if (cmpxchg(pp, NULL, p) != NULL) + if (cmpxchg(pp, p2m_missing, p) != p2m_missing) free_page((unsigned long)p); } @@ -129,7 +135,7 @@ void set_phys_to_machine(unsigned long pfn, unsigned long mfn) } topidx = p2m_top_index(pfn); - if (p2m_top[topidx] == NULL) { + if (p2m_top[topidx] == p2m_missing) { /* no need to allocate a page to store an invalid entry */ if (mfn == INVALID_P2M_ENTRY) return; -- cgit v1.2.3 From d5edbc1f75420935b1ec7e65df10c8f81cea82de Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 26 May 2008 23:31:22 +0100 Subject: xen: add p2m mfn_list_list When saving a domain, the Xen tools need to remap all our mfns to portable pfns. In order to remap our p2m table, it needs to know where all its pages are, so maintain the references to the p2m table for it to use. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Thomas Gleixner --- arch/x86/xen/enlighten.c | 2 ++ arch/x86/xen/mmu.c | 39 ++++++++++++++++++++++++++++++++++++--- arch/x86/xen/xen-ops.h | 2 ++ 3 files changed, 40 insertions(+), 3 deletions(-) (limited to 'arch/x86/xen/mmu.c') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index d9faaa26c104..ce67dc8f36af 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -880,6 +880,8 @@ static __init void setup_shared_info(void) /* In UP this is as good a place as any to set up shared info */ xen_setup_vcpu_info_placement(); #endif + + xen_setup_mfn_list_list(); } static __init void xen_pagetable_setup_done(pgd_t *base) diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index da7b45b05066..4740cda36563 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -69,6 +69,13 @@ static unsigned long *p2m_top[TOP_ENTRIES] __attribute__((section(".data.page_aligned"))) = { [ 0 ... TOP_ENTRIES - 1] = &p2m_missing[0] }; +/* Arrays of p2m arrays expressed in mfns used for save/restore */ +static unsigned long p2m_top_mfn[TOP_ENTRIES] + __attribute__((section(".bss.page_aligned"))); + +static unsigned long p2m_top_mfn_list[TOP_ENTRIES / P2M_ENTRIES_PER_PAGE] + __attribute__((section(".bss.page_aligned"))); + static inline unsigned p2m_top_index(unsigned long pfn) { BUG_ON(pfn >= MAX_DOMAIN_PAGES); @@ -80,11 +87,35 @@ static inline unsigned p2m_index(unsigned long pfn) return pfn % P2M_ENTRIES_PER_PAGE; } +/* Build the parallel p2m_top_mfn structures */ +void xen_setup_mfn_list_list(void) +{ + unsigned pfn, idx; + + for(pfn = 0; pfn < MAX_DOMAIN_PAGES; pfn += P2M_ENTRIES_PER_PAGE) { + unsigned topidx = p2m_top_index(pfn); + + p2m_top_mfn[topidx] = virt_to_mfn(p2m_top[topidx]); + } + + for(idx = 0; idx < ARRAY_SIZE(p2m_top_mfn_list); idx++) { + unsigned topidx = idx * P2M_ENTRIES_PER_PAGE; + p2m_top_mfn_list[idx] = virt_to_mfn(&p2m_top_mfn[topidx]); + } + + BUG_ON(HYPERVISOR_shared_info == &xen_dummy_shared_info); + + HYPERVISOR_shared_info->arch.pfn_to_mfn_frame_list_list = + virt_to_mfn(p2m_top_mfn_list); + HYPERVISOR_shared_info->arch.max_pfn = xen_start_info->nr_pages; +} + +/* Set up p2m_top to point to the domain-builder provided p2m pages */ void __init xen_build_dynamic_phys_to_machine(void) { - unsigned pfn; unsigned long *mfn_list = (unsigned long *)xen_start_info->mfn_list; unsigned long max_pfn = min(MAX_DOMAIN_PAGES, xen_start_info->nr_pages); + unsigned pfn; for(pfn = 0; pfn < max_pfn; pfn += P2M_ENTRIES_PER_PAGE) { unsigned topidx = p2m_top_index(pfn); @@ -105,7 +136,7 @@ unsigned long get_phys_to_machine(unsigned long pfn) return p2m_top[topidx][idx]; } -static void alloc_p2m(unsigned long **pp) +static void alloc_p2m(unsigned long **pp, unsigned long *mfnp) { unsigned long *p; unsigned i; @@ -118,6 +149,8 @@ static void alloc_p2m(unsigned long **pp) if (cmpxchg(pp, p2m_missing, p) != p2m_missing) free_page((unsigned long)p); + else + *mfnp = virt_to_mfn(p); } void set_phys_to_machine(unsigned long pfn, unsigned long mfn) @@ -139,7 +172,7 @@ void set_phys_to_machine(unsigned long pfn, unsigned long mfn) /* no need to allocate a page to store an invalid entry */ if (mfn == INVALID_P2M_ENTRY) return; - alloc_p2m(&p2m_top[topidx]); + alloc_p2m(&p2m_top[topidx], &p2m_top_mfn[topidx]); } idx = p2m_index(pfn); diff --git a/arch/x86/xen/xen-ops.h b/arch/x86/xen/xen-ops.h index 826d6e4c3dac..a1bc89a8f169 100644 --- a/arch/x86/xen/xen-ops.h +++ b/arch/x86/xen/xen-ops.h @@ -18,6 +18,8 @@ extern struct start_info *xen_start_info; extern struct shared_info xen_dummy_shared_info; extern struct shared_info *HYPERVISOR_shared_info; +void xen_setup_mfn_list_list(void); + char * __init xen_memory_setup(void); void __init xen_arch_setup(void); void __init xen_init_IRQ(void); -- cgit v1.2.3 From 0e91398f2a5d4eb6b07df8115917d0d1cf3e9b58 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 26 May 2008 23:31:27 +0100 Subject: xen: implement save/restore This patch implements Xen save/restore and migration. Saving is triggered via xenbus, which is polled in drivers/xen/manage.c. When a suspend request comes in, the kernel prepares itself for saving by: 1 - Freeze all processes. This is primarily to prevent any partially-completed pagetable updates from confusing the suspend process. If CONFIG_PREEMPT isn't defined, then this isn't necessary. 2 - Suspend xenbus and other devices 3 - Stop_machine, to make sure all the other vcpus are quiescent. The Xen tools require the domain to run its save off vcpu0. 4 - Within the stop_machine state, it pins any unpinned pgds (under construction or destruction), performs canonicalizes various other pieces of state (mostly converting mfns to pfns), and finally 5 - Suspend the domain Restore reverses the steps used to save the domain, ending when all the frozen processes are thawed. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Thomas Gleixner --- arch/x86/xen/Makefile | 2 +- arch/x86/xen/enlighten.c | 6 +-- arch/x86/xen/mmu.c | 46 +++++++++++++++++ arch/x86/xen/smp.c | 2 +- arch/x86/xen/suspend.c | 42 +++++++++++++++ arch/x86/xen/time.c | 8 +++ arch/x86/xen/xen-ops.h | 4 ++ drivers/xen/events.c | 83 +++++++++++++++++++++++++++++ drivers/xen/grant-table.c | 4 +- drivers/xen/manage.c | 126 ++++++++++++++++++++++++++++++++++++++++----- include/linux/page-flags.h | 1 + include/xen/events.h | 3 ++ include/xen/grant_table.h | 3 ++ include/xen/xen-ops.h | 9 ++++ 14 files changed, 318 insertions(+), 21 deletions(-) create mode 100644 arch/x86/xen/suspend.c (limited to 'arch/x86/xen/mmu.c') diff --git a/arch/x86/xen/Makefile b/arch/x86/xen/Makefile index 40b119b6b103..2ba2d1649131 100644 --- a/arch/x86/xen/Makefile +++ b/arch/x86/xen/Makefile @@ -1,4 +1,4 @@ obj-y := enlighten.o setup.o multicalls.o mmu.o \ - time.o xen-asm.o grant-table.o + time.o xen-asm.o grant-table.o suspend.o obj-$(CONFIG_SMP) += smp.o diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index ce67dc8f36af..b94f63ac228b 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -857,7 +857,7 @@ static __init void xen_pagetable_setup_start(pgd_t *base) PFN_DOWN(__pa(xen_start_info->pt_base))); } -static __init void setup_shared_info(void) +void xen_setup_shared_info(void) { if (!xen_feature(XENFEAT_auto_translated_physmap)) { unsigned long addr = fix_to_virt(FIX_PARAVIRT_BOOTMAP); @@ -894,7 +894,7 @@ static __init void xen_pagetable_setup_done(pgd_t *base) pv_mmu_ops.release_pmd = xen_release_pmd; pv_mmu_ops.set_pte = xen_set_pte; - setup_shared_info(); + xen_setup_shared_info(); /* Actually pin the pagetable down, but we can't set PG_pinned yet because the page structures don't exist yet. */ @@ -902,7 +902,7 @@ static __init void xen_pagetable_setup_done(pgd_t *base) } /* This is called once we have the cpu_possible_map */ -void __init xen_setup_vcpu_info_placement(void) +void xen_setup_vcpu_info_placement(void) { int cpu; diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 4740cda36563..e95955968ba3 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -560,6 +560,29 @@ void xen_pgd_pin(pgd_t *pgd) xen_mc_issue(0); } +/* + * On save, we need to pin all pagetables to make sure they get their + * mfns turned into pfns. Search the list for any unpinned pgds and pin + * them (unpinned pgds are not currently in use, probably because the + * process is under construction or destruction). + */ +void xen_mm_pin_all(void) +{ + unsigned long flags; + struct page *page; + + spin_lock_irqsave(&pgd_lock, flags); + + list_for_each_entry(page, &pgd_list, lru) { + if (!PagePinned(page)) { + xen_pgd_pin((pgd_t *)page_address(page)); + SetPageSavePinned(page); + } + } + + spin_unlock_irqrestore(&pgd_lock, flags); +} + /* The init_mm pagetable is really pinned as soon as its created, but that's before we have page structures to store the bits. So do all the book-keeping now. */ @@ -617,6 +640,29 @@ static void xen_pgd_unpin(pgd_t *pgd) xen_mc_issue(0); } +/* + * On resume, undo any pinning done at save, so that the rest of the + * kernel doesn't see any unexpected pinned pagetables. + */ +void xen_mm_unpin_all(void) +{ + unsigned long flags; + struct page *page; + + spin_lock_irqsave(&pgd_lock, flags); + + list_for_each_entry(page, &pgd_list, lru) { + if (PageSavePinned(page)) { + BUG_ON(!PagePinned(page)); + printk("unpinning pinned %p\n", page_address(page)); + xen_pgd_unpin((pgd_t *)page_address(page)); + ClearPageSavePinned(page); + } + } + + spin_unlock_irqrestore(&pgd_lock, flags); +} + void xen_activate_mm(struct mm_struct *prev, struct mm_struct *next) { spin_lock(&next->page_table_lock); diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c index 74ab8968c525..d2e3c20127d7 100644 --- a/arch/x86/xen/smp.c +++ b/arch/x86/xen/smp.c @@ -35,7 +35,7 @@ #include "xen-ops.h" #include "mmu.h" -static cpumask_t xen_cpu_initialized_map; +cpumask_t xen_cpu_initialized_map; static DEFINE_PER_CPU(int, resched_irq) = -1; static DEFINE_PER_CPU(int, callfunc_irq) = -1; static DEFINE_PER_CPU(int, debug_irq) = -1; diff --git a/arch/x86/xen/suspend.c b/arch/x86/xen/suspend.c new file mode 100644 index 000000000000..7620a16fe535 --- /dev/null +++ b/arch/x86/xen/suspend.c @@ -0,0 +1,42 @@ +#include + +#include +#include +#include + +#include +#include + +#include "xen-ops.h" +#include "mmu.h" + +void xen_pre_suspend(void) +{ + xen_start_info->store_mfn = mfn_to_pfn(xen_start_info->store_mfn); + xen_start_info->console.domU.mfn = + mfn_to_pfn(xen_start_info->console.domU.mfn); + + BUG_ON(!irqs_disabled()); + + HYPERVISOR_shared_info = &xen_dummy_shared_info; + if (HYPERVISOR_update_va_mapping(fix_to_virt(FIX_PARAVIRT_BOOTMAP), + __pte_ma(0), 0)) + BUG(); +} + +void xen_post_suspend(int suspend_cancelled) +{ + if (suspend_cancelled) { + xen_start_info->store_mfn = + pfn_to_mfn(xen_start_info->store_mfn); + xen_start_info->console.domU.mfn = + pfn_to_mfn(xen_start_info->console.domU.mfn); + } else { +#ifdef CONFIG_SMP + xen_cpu_initialized_map = cpu_online_map; +#endif + } + + xen_setup_shared_info(); +} + diff --git a/arch/x86/xen/time.c b/arch/x86/xen/time.c index c39e1a5aa241..0bef256e5f2d 100644 --- a/arch/x86/xen/time.c +++ b/arch/x86/xen/time.c @@ -572,6 +572,14 @@ void xen_setup_cpu_clockevents(void) clockevents_register_device(&__get_cpu_var(xen_clock_events)); } +void xen_time_suspend(void) +{ +} + +void xen_time_resume(void) +{ +} + __init void xen_time_init(void) { int cpu = smp_processor_id(); diff --git a/arch/x86/xen/xen-ops.h b/arch/x86/xen/xen-ops.h index a1bc89a8f169..a0503acad664 100644 --- a/arch/x86/xen/xen-ops.h +++ b/arch/x86/xen/xen-ops.h @@ -9,6 +9,7 @@ extern const char xen_hypervisor_callback[]; extern const char xen_failsafe_callback[]; +struct trap_info; void xen_copy_trap_info(struct trap_info *traps); DECLARE_PER_CPU(unsigned long, xen_cr3); @@ -19,6 +20,7 @@ extern struct shared_info xen_dummy_shared_info; extern struct shared_info *HYPERVISOR_shared_info; void xen_setup_mfn_list_list(void); +void xen_setup_shared_info(void); char * __init xen_memory_setup(void); void __init xen_arch_setup(void); @@ -59,6 +61,8 @@ int xen_smp_call_function_single(int cpu, void (*func) (void *info), void *info, int xen_smp_call_function_mask(cpumask_t mask, void (*func)(void *), void *info, int wait); +extern cpumask_t xen_cpu_initialized_map; + /* Declare an asm function, along with symbols needed to make it inlineable */ diff --git a/drivers/xen/events.c b/drivers/xen/events.c index 70375a690761..73d78dc9b875 100644 --- a/drivers/xen/events.c +++ b/drivers/xen/events.c @@ -674,6 +674,89 @@ static int retrigger_dynirq(unsigned int irq) return ret; } +static void restore_cpu_virqs(unsigned int cpu) +{ + struct evtchn_bind_virq bind_virq; + int virq, irq, evtchn; + + for (virq = 0; virq < NR_VIRQS; virq++) { + if ((irq = per_cpu(virq_to_irq, cpu)[virq]) == -1) + continue; + + BUG_ON(irq_info[irq].type != IRQT_VIRQ); + BUG_ON(irq_info[irq].index != virq); + + /* Get a new binding from Xen. */ + bind_virq.virq = virq; + bind_virq.vcpu = cpu; + if (HYPERVISOR_event_channel_op(EVTCHNOP_bind_virq, + &bind_virq) != 0) + BUG(); + evtchn = bind_virq.port; + + /* Record the new mapping. */ + evtchn_to_irq[evtchn] = irq; + irq_info[irq] = mk_irq_info(IRQT_VIRQ, virq, evtchn); + bind_evtchn_to_cpu(evtchn, cpu); + + /* Ready for use. */ + unmask_evtchn(evtchn); + } +} + +static void restore_cpu_ipis(unsigned int cpu) +{ + struct evtchn_bind_ipi bind_ipi; + int ipi, irq, evtchn; + + for (ipi = 0; ipi < XEN_NR_IPIS; ipi++) { + if ((irq = per_cpu(ipi_to_irq, cpu)[ipi]) == -1) + continue; + + BUG_ON(irq_info[irq].type != IRQT_IPI); + BUG_ON(irq_info[irq].index != ipi); + + /* Get a new binding from Xen. */ + bind_ipi.vcpu = cpu; + if (HYPERVISOR_event_channel_op(EVTCHNOP_bind_ipi, + &bind_ipi) != 0) + BUG(); + evtchn = bind_ipi.port; + + /* Record the new mapping. */ + evtchn_to_irq[evtchn] = irq; + irq_info[irq] = mk_irq_info(IRQT_IPI, ipi, evtchn); + bind_evtchn_to_cpu(evtchn, cpu); + + /* Ready for use. */ + unmask_evtchn(evtchn); + + } +} + +void xen_irq_resume(void) +{ + unsigned int cpu, irq, evtchn; + + init_evtchn_cpu_bindings(); + + /* New event-channel space is not 'live' yet. */ + for (evtchn = 0; evtchn < NR_EVENT_CHANNELS; evtchn++) + mask_evtchn(evtchn); + + /* No IRQ <-> event-channel mappings. */ + for (irq = 0; irq < NR_IRQS; irq++) + irq_info[irq].evtchn = 0; /* zap event-channel binding */ + + for (evtchn = 0; evtchn < NR_EVENT_CHANNELS; evtchn++) + evtchn_to_irq[evtchn] = -1; + + for_each_possible_cpu(cpu) { + restore_cpu_virqs(cpu); + restore_cpu_ipis(cpu); + } +} + static struct irq_chip xen_dynamic_chip __read_mostly = { .name = "xen-dyn", .mask = disable_dynirq, diff --git a/drivers/xen/grant-table.c b/drivers/xen/grant-table.c index 52b6b41b909d..e9e11168616a 100644 --- a/drivers/xen/grant-table.c +++ b/drivers/xen/grant-table.c @@ -471,14 +471,14 @@ static int gnttab_map(unsigned int start_idx, unsigned int end_idx) return 0; } -static int gnttab_resume(void) +int gnttab_resume(void) { if (max_nr_grant_frames() < nr_grant_frames) return -ENOSYS; return gnttab_map(0, nr_grant_frames - 1); } -static int gnttab_suspend(void) +int gnttab_suspend(void) { arch_gnttab_unmap_shared(shared, nr_grant_frames); return 0; diff --git a/drivers/xen/manage.c b/drivers/xen/manage.c index aa7af9e6abc0..ba85fa2cff33 100644 --- a/drivers/xen/manage.c +++ b/drivers/xen/manage.c @@ -5,21 +5,113 @@ #include #include #include +#include +#include #include - -#define SHUTDOWN_INVALID -1 -#define SHUTDOWN_POWEROFF 0 -#define SHUTDOWN_SUSPEND 2 -/* Code 3 is SHUTDOWN_CRASH, which we don't use because the domain can only - * report a crash, not be instructed to crash! - * HALT is the same as POWEROFF, as far as we're concerned. The tools use - * the distinction when we return the reason code to them. - */ -#define SHUTDOWN_HALT 4 +#include +#include +#include +#include + +#include +#include + +enum shutdown_state { + SHUTDOWN_INVALID = -1, + SHUTDOWN_POWEROFF = 0, + SHUTDOWN_SUSPEND = 2, + /* Code 3 is SHUTDOWN_CRASH, which we don't use because the domain can only + report a crash, not be instructed to crash! + HALT is the same as POWEROFF, as far as we're concerned. The tools use + the distinction when we return the reason code to them. */ + SHUTDOWN_HALT = 4, +}; /* Ignore multiple shutdown requests. */ -static int shutting_down = SHUTDOWN_INVALID; +static enum shutdown_state shutting_down = SHUTDOWN_INVALID; + +static int xen_suspend(void *data) +{ + int *cancelled = data; + + BUG_ON(!irqs_disabled()); + + load_cr3(swapper_pg_dir); + + xen_mm_pin_all(); + gnttab_suspend(); + xen_time_suspend(); + xen_pre_suspend(); + + /* + * This hypercall returns 1 if suspend was cancelled + * or the domain was merely checkpointed, and 0 if it + * is resuming in a new domain. + */ + *cancelled = HYPERVISOR_suspend(virt_to_mfn(xen_start_info)); + + xen_post_suspend(*cancelled); + xen_time_resume(); + gnttab_resume(); + xen_mm_unpin_all(); + + if (!*cancelled) { + xen_irq_resume(); + xen_console_resume(); + } + + return 0; +} + +static void do_suspend(void) +{ + int err; + int cancelled = 1; + + shutting_down = SHUTDOWN_SUSPEND; + +#ifdef CONFIG_PREEMPT + /* If the kernel is preemptible, we need to freeze all the processes + to prevent them from being in the middle of a pagetable update + during suspend. */ + err = freeze_processes(); + if (err) { + printk(KERN_ERR "xen suspend: freeze failed %d\n", err); + return; + } +#endif + + err = device_suspend(PMSG_SUSPEND); + if (err) { + printk(KERN_ERR "xen suspend: device_suspend %d\n", err); + goto out; + } + + printk("suspending xenbus...\n"); + /* XXX use normal device tree? */ + xenbus_suspend(); + + err = stop_machine_run(xen_suspend, &cancelled, 0); + if (err) { + printk(KERN_ERR "failed to start xen_suspend: %d\n", err); + goto out; + } + + if (!cancelled) + xenbus_resume(); + else + xenbus_suspend_cancel(); + + device_resume(); + + +out: +#ifdef CONFIG_PREEMPT + thaw_processes(); +#endif + shutting_down = SHUTDOWN_INVALID; +} static void shutdown_handler(struct xenbus_watch *watch, const char **vec, unsigned int len) @@ -52,11 +144,17 @@ static void shutdown_handler(struct xenbus_watch *watch, } if (strcmp(str, "poweroff") == 0 || - strcmp(str, "halt") == 0) + strcmp(str, "halt") == 0) { + shutting_down = SHUTDOWN_POWEROFF; orderly_poweroff(false); - else if (strcmp(str, "reboot") == 0) + } else if (strcmp(str, "reboot") == 0) { + shutting_down = SHUTDOWN_POWEROFF; /* ? */ ctrl_alt_del(); - else { +#ifdef CONFIG_PM_SLEEP + } else if (strcmp(str, "suspend") == 0) { + do_suspend(); +#endif + } else { printk(KERN_INFO "Ignoring shutdown request: %s\n", str); shutting_down = SHUTDOWN_INVALID; } diff --git a/include/linux/page-flags.h b/include/linux/page-flags.h index 590cff32415d..02955a1c3d76 100644 --- a/include/linux/page-flags.h +++ b/include/linux/page-flags.h @@ -157,6 +157,7 @@ PAGEFLAG(Active, active) __CLEARPAGEFLAG(Active, active) __PAGEFLAG(Slab, slab) PAGEFLAG(Checked, owner_priv_1) /* Used by some filesystems */ PAGEFLAG(Pinned, owner_priv_1) TESTSCFLAG(Pinned, owner_priv_1) /* Xen */ +PAGEFLAG(SavePinned, dirty); /* Xen */ PAGEFLAG(Reserved, reserved) __CLEARPAGEFLAG(Reserved, reserved) PAGEFLAG(Private, private) __CLEARPAGEFLAG(Private, private) __SETPAGEFLAG(Private, private) diff --git a/include/xen/events.h b/include/xen/events.h index a82ec0c45c38..67c4436554a9 100644 --- a/include/xen/events.h +++ b/include/xen/events.h @@ -41,4 +41,7 @@ static inline void notify_remote_via_evtchn(int port) } extern void notify_remote_via_irq(int irq); + +extern void xen_irq_resume(void); + #endif /* _XEN_EVENTS_H */ diff --git a/include/xen/grant_table.h b/include/xen/grant_table.h index 466204846121..a40f1cd91be1 100644 --- a/include/xen/grant_table.h +++ b/include/xen/grant_table.h @@ -51,6 +51,9 @@ struct gnttab_free_callback { u16 count; }; +int gnttab_suspend(void); +int gnttab_resume(void); + int gnttab_grant_foreign_access(domid_t domid, unsigned long frame, int readonly); diff --git a/include/xen/xen-ops.h b/include/xen/xen-ops.h index 10ddfe0142d0..5d7a6db54a8c 100644 --- a/include/xen/xen-ops.h +++ b/include/xen/xen-ops.h @@ -5,4 +5,13 @@ DECLARE_PER_CPU(struct vcpu_info *, xen_vcpu); +void xen_pre_suspend(void); +void xen_post_suspend(int suspend_cancelled); + +void xen_mm_pin_all(void); +void xen_mm_unpin_all(void); + +void xen_time_suspend(void); +void xen_time_resume(void); + #endif /* INCLUDE_XEN_OPS_H */ -- cgit v1.2.3 From b20aeccd6ad42ccb6be1b3d1d32618ddd2b31bf0 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Wed, 28 May 2008 14:24:38 +0200 Subject: xen: fix early bootup crash on native hardware -tip tree auto-testing found the following early bootup hang: --------------> get_memcfg_from_srat: assigning address to rsdp RSD PTR v0 [Nvidia] BUG: Int 14: CR2 ffd00040 EDI 8092fbfe ESI ffd00040 EBP 80b0aee8 ESP 80b0aed0 EBX 000f76f0 EDX 0000000e ECX 00000003 EAX ffd00040 err 00000000 EIP 802c055a CS 00000060 flg 00010006 Stack: ffd00040 80bc78d0 80b0af6c 80b1dbfe 8093d8ba 00000008 80b42810 80b4ddb4 80b42842 00000000 80b0af1c 801079c8 808e724e 00000000 80b42871 802c0531 00000100 00000000 0003fff0 80b0af40 80129999 00040100 00040100 00000000 Pid: 0, comm: swapper Not tainted 2.6.26-rc4-sched-devel.git #570 [<802c055a>] ? strncmp+0x11/0x25 [<80b1dbfe>] ? get_memcfg_from_srat+0xb4/0x568 [<801079c8>] ? mcount_call+0x5/0x9 [<802c0531>] ? strcmp+0xa/0x22 [<80129999>] ? printk+0x38/0x3a [<80129999>] ? printk+0x38/0x3a [<8011b122>] ? memory_present+0x66/0x6f [<80b216b4>] ? setup_memory+0x13/0x40c [<80b16b47>] ? propagate_e820_map+0x80/0x97 [<80b1622a>] ? setup_arch+0x248/0x477 [<80129999>] ? printk+0x38/0x3a [<80b11759>] ? start_kernel+0x6e/0x2eb [<80b110fc>] ? i386_start_kernel+0xeb/0xf2 ======================= <------ with this config: http://redhat.com/~mingo/misc/config-Wed_May_28_01_33_33_CEST_2008.bad The thing is, the crash makes little sense at first sight. We crash on a benign-looking printk. The code around it got changed in -tip but checking those topic branches individually did not reproduce the bug. Bisection led to this commit: | d5edbc1f75420935b1ec7e65df10c8f81cea82de is first bad commit | commit d5edbc1f75420935b1ec7e65df10c8f81cea82de | Author: Jeremy Fitzhardinge | Date: Mon May 26 23:31:22 2008 +0100 | | xen: add p2m mfn_list_list Which is somewhat surprising, as on native hardware Xen client side should have little to no side-effects. After some head scratching, it turns out the following happened: randconfig enabled the following Xen options: CONFIG_XEN=y CONFIG_XEN_MAX_DOMAIN_MEMORY=8 # CONFIG_XEN_BLKDEV_FRONTEND is not set # CONFIG_XEN_NETDEV_FRONTEND is not set CONFIG_HVC_XEN=y # CONFIG_XEN_BALLOON is not set which activated this piece of code in arch/x86/xen/mmu.c: > @@ -69,6 +69,13 @@ > __attribute__((section(".data.page_aligned"))) = > { [ 0 ... TOP_ENTRIES - 1] = &p2m_missing[0] }; > > +/* Arrays of p2m arrays expressed in mfns used for save/restore */ > +static unsigned long p2m_top_mfn[TOP_ENTRIES] > + __attribute__((section(".bss.page_aligned"))); > + > +static unsigned long p2m_top_mfn_list[TOP_ENTRIES / P2M_ENTRIES_PER_PAGE] > + __attribute__((section(".bss.page_aligned"))); The problem is, you must only put variables into .bss.page_aligned that have a _size_ that is _exactly_ page aligned. In this case the size of p2m_top_mfn_list is not page aligned: 80b8d000 b p2m_top_mfn 80b8f000 b p2m_top_mfn_list 80b8f008 b softirq_stack 80b97008 b hardirq_stack 80b9f008 b bm_pte So all subsequent variables get unaligned which, depending on luck, breaks the kernel in various funny ways. In this case what killed the kernel first was the misaligned bootmap pte page, resulting in that creative crash above. Anyway, this was a fun bug to track down :-) I think the moral is that .bss.page_aligned is a dangerous construct in its current form, and the symptoms of breakage are very non-trivial, so i think we need build-time checks to make sure all symbols in .bss.page_aligned are truly page aligned. The Xen fix below gets the kernel booting again. Signed-off-by: Ingo Molnar --- arch/x86/xen/mmu.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'arch/x86/xen/mmu.c') diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index e95955968ba3..eef3b5c6e2fe 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -73,7 +73,8 @@ static unsigned long *p2m_top[TOP_ENTRIES] static unsigned long p2m_top_mfn[TOP_ENTRIES] __attribute__((section(".bss.page_aligned"))); -static unsigned long p2m_top_mfn_list[TOP_ENTRIES / P2M_ENTRIES_PER_PAGE] +static unsigned long p2m_top_mfn_list[ + PAGE_ALIGN(TOP_ENTRIES / P2M_ENTRIES_PER_PAGE)] __attribute__((section(".bss.page_aligned"))); static inline unsigned p2m_top_index(unsigned long pfn) -- cgit v1.2.3 From 15ce60056b24a65b65e28de973a9fd8ac0750a2f Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Mon, 2 Jun 2008 13:20:11 +0200 Subject: xen: export get_phys_to_machine -tip testing found the following xen-console symbols trouble: ERROR: "get_phys_to_machine" [drivers/video/xen-fbfront.ko] undefined! ERROR: "get_phys_to_machine" [drivers/net/xen-netfront.ko] undefined! ERROR: "get_phys_to_machine" [drivers/input/xen-kbdfront.ko] undefined! with: http://redhat.com/~mingo/misc/config-Mon_Jun__2_12_25_13_CEST_2008.bad --- arch/x86/xen/mmu.c | 1 + 1 file changed, 1 insertion(+) (limited to 'arch/x86/xen/mmu.c') diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index eef3b5c6e2fe..17f374eb1faa 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -136,6 +136,7 @@ unsigned long get_phys_to_machine(unsigned long pfn) idx = p2m_index(pfn); return p2m_top[topidx][idx]; } +EXPORT_SYMBOL_GPL(get_phys_to_machine); static void alloc_p2m(unsigned long **pp, unsigned long *mfnp) { -- cgit v1.2.3 From e2426cf85f8db5891fb5831323d2d0c176c4dadc Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Sat, 31 May 2008 01:24:27 +0100 Subject: xen: avoid hypercalls when updating unpinned pud/pmd When operating on an unpinned pagetable (ie, one under construction or destruction), it isn't necessary to use a hypercall to update a pud/pmd entry. Jan Beulich observed that a similar optimisation avoided many thousands of hypercalls while doing a kernel build. One tricky part is that early in the kernel boot there's no page structure, so we can't check to see if the page is pinned. In that case, we just always use the hypercall. Signed-off-by: Jeremy Fitzhardinge Cc: Jan Beulich Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 14 +++++++++++--- arch/x86/xen/mmu.c | 39 +++++++++++++++++++++++++++++++++++---- arch/x86/xen/mmu.h | 8 ++++---- 3 files changed, 50 insertions(+), 11 deletions(-) (limited to 'arch/x86/xen/mmu.c') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index b94f63ac228b..ed9f04b3836d 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -901,6 +901,14 @@ static __init void xen_pagetable_setup_done(pgd_t *base) pin_pagetable_pfn(MMUEXT_PIN_L3_TABLE, PFN_DOWN(__pa(base))); } +static __init void xen_post_allocator_init(void) +{ + pv_mmu_ops.set_pmd = xen_set_pmd; + pv_mmu_ops.set_pud = xen_set_pud; + + xen_mark_init_mm_pinned(); +} + /* This is called once we have the cpu_possible_map */ void xen_setup_vcpu_info_placement(void) { @@ -988,7 +996,7 @@ static const struct pv_init_ops xen_init_ops __initdata = { .banner = xen_banner, .memory_setup = xen_memory_setup, .arch_setup = xen_arch_setup, - .post_allocator_init = xen_mark_init_mm_pinned, + .post_allocator_init = xen_post_allocator_init, }; static const struct pv_time_ops xen_time_ops __initdata = { @@ -1100,7 +1108,7 @@ static const struct pv_mmu_ops xen_mmu_ops __initdata = { .set_pte = NULL, /* see xen_pagetable_setup_* */ .set_pte_at = xen_set_pte_at, - .set_pmd = xen_set_pmd, + .set_pmd = xen_set_pmd_hyper, .pte_val = xen_pte_val, .pte_flags = native_pte_val, @@ -1111,7 +1119,7 @@ static const struct pv_mmu_ops xen_mmu_ops __initdata = { .set_pte_atomic = xen_set_pte_atomic, .set_pte_present = xen_set_pte_at, - .set_pud = xen_set_pud, + .set_pud = xen_set_pud_hyper, .pte_clear = xen_pte_clear, .pmd_clear = xen_pmd_clear, diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 17f374eb1faa..4fa0934db925 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -223,7 +223,14 @@ void make_lowmem_page_readwrite(void *vaddr) } -void xen_set_pmd(pmd_t *ptr, pmd_t val) +static bool page_pinned(void *ptr) +{ + struct page *page = virt_to_page(ptr); + + return PagePinned(page); +} + +void xen_set_pmd_hyper(pmd_t *ptr, pmd_t val) { struct multicall_space mcs; struct mmu_update *u; @@ -241,6 +248,18 @@ void xen_set_pmd(pmd_t *ptr, pmd_t val) preempt_enable(); } +void xen_set_pmd(pmd_t *ptr, pmd_t val) +{ + /* If page is not pinned, we can just update the entry + directly */ + if (!page_pinned(ptr)) { + *ptr = val; + return; + } + + xen_set_pmd_hyper(ptr, val); +} + /* * Associate a virtual page frame with a given physical page frame * and protection flags for that frame. @@ -348,7 +367,7 @@ pmdval_t xen_pmd_val(pmd_t pmd) return ret; } -void xen_set_pud(pud_t *ptr, pud_t val) +void xen_set_pud_hyper(pud_t *ptr, pud_t val) { struct multicall_space mcs; struct mmu_update *u; @@ -366,6 +385,18 @@ void xen_set_pud(pud_t *ptr, pud_t val) preempt_enable(); } +void xen_set_pud(pud_t *ptr, pud_t val) +{ + /* If page is not pinned, we can just update the entry + directly */ + if (!page_pinned(ptr)) { + *ptr = val; + return; + } + + xen_set_pud_hyper(ptr, val); +} + void xen_set_pte(pte_t *ptep, pte_t pte) { ptep->pte_high = pte.pte_high; @@ -387,7 +418,7 @@ void xen_pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep) void xen_pmd_clear(pmd_t *pmdp) { - xen_set_pmd(pmdp, __pmd(0)); + set_pmd(pmdp, __pmd(0)); } pmd_t xen_make_pmd(pmdval_t pmd) @@ -758,7 +789,7 @@ void xen_exit_mmap(struct mm_struct *mm) spin_lock(&mm->page_table_lock); /* pgd may not be pinned in the error exit path of execve */ - if (PagePinned(virt_to_page(mm->pgd))) + if (page_pinned(mm->pgd)) xen_pgd_unpin(mm->pgd); spin_unlock(&mm->page_table_lock); diff --git a/arch/x86/xen/mmu.h b/arch/x86/xen/mmu.h index 5fe961caffd4..e3dd09e25c63 100644 --- a/arch/x86/xen/mmu.h +++ b/arch/x86/xen/mmu.h @@ -25,10 +25,6 @@ enum pt_level { void set_pte_mfn(unsigned long vaddr, unsigned long pfn, pgprot_t flags); -void xen_set_pte(pte_t *ptep, pte_t pteval); -void xen_set_pte_at(struct mm_struct *mm, unsigned long addr, - pte_t *ptep, pte_t pteval); -void xen_set_pmd(pmd_t *pmdp, pmd_t pmdval); void xen_activate_mm(struct mm_struct *prev, struct mm_struct *next); void xen_dup_mmap(struct mm_struct *oldmm, struct mm_struct *mm); @@ -45,10 +41,14 @@ pte_t xen_make_pte(pteval_t); pmd_t xen_make_pmd(pmdval_t); pgd_t xen_make_pgd(pgdval_t); +void xen_set_pte(pte_t *ptep, pte_t pteval); void xen_set_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pteval); void xen_set_pte_atomic(pte_t *ptep, pte_t pte); +void xen_set_pmd(pmd_t *pmdp, pmd_t pmdval); void xen_set_pud(pud_t *ptr, pud_t val); +void xen_set_pmd_hyper(pmd_t *pmdp, pmd_t pmdval); +void xen_set_pud_hyper(pud_t *ptr, pud_t val); void xen_pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep); void xen_pmd_clear(pmd_t *pmdp); -- cgit v1.2.3 From eb179e443deb0a5c81a62b4c157124a4b7ff1813 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 16 Jun 2008 15:01:53 -0700 Subject: xen: mask unwanted pte bits in __supported_pte_mask [ Stable: this isn't a bugfix in itself, but it's a pre-requiste for "xen: don't drop NX bit" ] Signed-off-by: Jeremy Fitzhardinge Cc: Stable Kernel Cc: the arch/x86 maintainers Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 5 +++++ arch/x86/xen/mmu.c | 4 +--- 2 files changed, 6 insertions(+), 3 deletions(-) (limited to 'arch/x86/xen/mmu.c') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 8e6152e6ed88..73fb0c4c150a 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1280,6 +1280,11 @@ asmlinkage void __init xen_start_kernel(void) if (xen_feature(XENFEAT_supervisor_mode_kernel)) pv_info.kernel_rpl = 0; + /* Prevent unwanted bits from being set in PTEs. */ + __supported_pte_mask &= ~_PAGE_GLOBAL; + if (!is_initial_xendomain()) + __supported_pte_mask &= ~(_PAGE_PWT | _PAGE_PCD); + /* set the limit of our address space */ xen_reserve_top(); diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 7c9935858f92..f0078d9a5b78 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -343,10 +343,8 @@ pgdval_t xen_pgd_val(pgd_t pgd) pte_t xen_make_pte(pteval_t pte) { - if (pte & _PAGE_PRESENT) { + if (pte & _PAGE_PRESENT) pte = phys_to_machine(XPADDR(pte)).maddr; - pte &= ~(_PAGE_PCD | _PAGE_PWT); - } return (pte_t){ .pte = pte }; } -- cgit v1.2.3 From a987b16cc6123af2c9414032701bab5f73c54c89 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 16 Jun 2008 15:01:56 -0700 Subject: xen: don't drop NX bit Because NX is now enforced properly, we must put the hypercall page into the .text segment so that it is executable. Signed-off-by: Jeremy Fitzhardinge Cc: Stable Kernel Cc: the arch/x86 maintainers Signed-off-by: Ingo Molnar --- arch/x86/xen/mmu.c | 54 +++++++++++++++++++++++++++---------------------- arch/x86/xen/xen-head.S | 2 +- 2 files changed, 31 insertions(+), 25 deletions(-) (limited to 'arch/x86/xen/mmu.c') diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index f0078d9a5b78..8132aa8c5d49 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -323,46 +323,54 @@ out: preempt_enable(); } -pteval_t xen_pte_val(pte_t pte) +/* Assume pteval_t is equivalent to all the other *val_t types. */ +static pteval_t pte_mfn_to_pfn(pteval_t val) { - pteval_t ret = pte.pte; + if (val & _PAGE_PRESENT) { + unsigned long mfn = (val & PTE_MASK) >> PAGE_SHIFT; + pteval_t flags = val & ~PTE_MASK; + val = (mfn_to_pfn(mfn) << PAGE_SHIFT) | flags; + } - if (ret & _PAGE_PRESENT) - ret = machine_to_phys(XMADDR(ret)).paddr | _PAGE_PRESENT; + return val; +} - return ret; +static pteval_t pte_pfn_to_mfn(pteval_t val) +{ + if (val & _PAGE_PRESENT) { + unsigned long pfn = (val & PTE_MASK) >> PAGE_SHIFT; + pteval_t flags = val & ~PTE_MASK; + val = (pfn_to_mfn(pfn) << PAGE_SHIFT) | flags; + } + + return val; +} + +pteval_t xen_pte_val(pte_t pte) +{ + return pte_mfn_to_pfn(pte.pte); } pgdval_t xen_pgd_val(pgd_t pgd) { - pgdval_t ret = pgd.pgd; - if (ret & _PAGE_PRESENT) - ret = machine_to_phys(XMADDR(ret)).paddr | _PAGE_PRESENT; - return ret; + return pte_mfn_to_pfn(pgd.pgd); } pte_t xen_make_pte(pteval_t pte) { - if (pte & _PAGE_PRESENT) - pte = phys_to_machine(XPADDR(pte)).maddr; - - return (pte_t){ .pte = pte }; + pte = pte_pfn_to_mfn(pte); + return native_make_pte(pte); } pgd_t xen_make_pgd(pgdval_t pgd) { - if (pgd & _PAGE_PRESENT) - pgd = phys_to_machine(XPADDR(pgd)).maddr; - - return (pgd_t){ pgd }; + pgd = pte_pfn_to_mfn(pgd); + return native_make_pgd(pgd); } pmdval_t xen_pmd_val(pmd_t pmd) { - pmdval_t ret = native_pmd_val(pmd); - if (ret & _PAGE_PRESENT) - ret = machine_to_phys(XMADDR(ret)).paddr | _PAGE_PRESENT; - return ret; + return pte_mfn_to_pfn(pmd.pmd); } void xen_set_pud_hyper(pud_t *ptr, pud_t val) @@ -421,9 +429,7 @@ void xen_pmd_clear(pmd_t *pmdp) pmd_t xen_make_pmd(pmdval_t pmd) { - if (pmd & _PAGE_PRESENT) - pmd = phys_to_machine(XPADDR(pmd)).maddr; - + pmd = pte_pfn_to_mfn(pmd); return native_make_pmd(pmd); } diff --git a/arch/x86/xen/xen-head.S b/arch/x86/xen/xen-head.S index ef6c9e005f90..7c0cf6320a0a 100644 --- a/arch/x86/xen/xen-head.S +++ b/arch/x86/xen/xen-head.S @@ -18,7 +18,7 @@ ENTRY(startup_xen) __FINIT -.pushsection .bss.page_aligned +.pushsection .text .align PAGE_SIZE_asm ENTRY(hypercall_page) .skip 0x1000 -- cgit v1.2.3 From 05345b0f006ac226d0d25d48fcb2d792ac44a071 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 16 Jun 2008 15:01:53 -0700 Subject: xen: mask unwanted pte bits in __supported_pte_mask [ Stable: this isn't a bugfix in itself, but it's a pre-requiste for "xen: don't drop NX bit" ] Signed-off-by: Jeremy Fitzhardinge Cc: Stable Kernel Cc: the arch/x86 maintainers Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 5 +++++ arch/x86/xen/mmu.c | 4 +--- 2 files changed, 6 insertions(+), 3 deletions(-) (limited to 'arch/x86/xen/mmu.c') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index c8a56e457d61..c048de34d6a1 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1228,6 +1228,11 @@ asmlinkage void __init xen_start_kernel(void) if (xen_feature(XENFEAT_supervisor_mode_kernel)) pv_info.kernel_rpl = 0; + /* Prevent unwanted bits from being set in PTEs. */ + __supported_pte_mask &= ~_PAGE_GLOBAL; + if (!is_initial_xendomain()) + __supported_pte_mask &= ~(_PAGE_PWT | _PAGE_PCD); + /* set the limit of our address space */ xen_reserve_top(); diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 3525ef523a74..3f2a67fe6ad6 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -199,10 +199,8 @@ pgdval_t xen_pgd_val(pgd_t pgd) pte_t xen_make_pte(pteval_t pte) { - if (pte & _PAGE_PRESENT) { + if (pte & _PAGE_PRESENT) pte = phys_to_machine(XPADDR(pte)).maddr; - pte &= ~(_PAGE_PCD | _PAGE_PWT); - } return (pte_t){ .pte = pte }; } -- cgit v1.2.3 From ebb9cfe20fe167f29960a5e913193a684fac50bf Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 16 Jun 2008 15:01:56 -0700 Subject: xen: don't drop NX bit Because NX is now enforced properly, we must put the hypercall page into the .text segment so that it is executable. Signed-off-by: Jeremy Fitzhardinge Cc: Stable Kernel Cc: the arch/x86 maintainers Signed-off-by: Ingo Molnar --- arch/x86/xen/mmu.c | 54 +++++++++++++++++++++++++++---------------------- arch/x86/xen/xen-head.S | 2 +- 2 files changed, 31 insertions(+), 25 deletions(-) (limited to 'arch/x86/xen/mmu.c') diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 3f2a67fe6ad6..265601d5a6ae 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -179,46 +179,54 @@ out: preempt_enable(); } -pteval_t xen_pte_val(pte_t pte) +/* Assume pteval_t is equivalent to all the other *val_t types. */ +static pteval_t pte_mfn_to_pfn(pteval_t val) { - pteval_t ret = pte.pte; + if (val & _PAGE_PRESENT) { + unsigned long mfn = (val & PTE_MASK) >> PAGE_SHIFT; + pteval_t flags = val & ~PTE_MASK; + val = (mfn_to_pfn(mfn) << PAGE_SHIFT) | flags; + } - if (ret & _PAGE_PRESENT) - ret = machine_to_phys(XMADDR(ret)).paddr | _PAGE_PRESENT; + return val; +} - return ret; +static pteval_t pte_pfn_to_mfn(pteval_t val) +{ + if (val & _PAGE_PRESENT) { + unsigned long pfn = (val & PTE_MASK) >> PAGE_SHIFT; + pteval_t flags = val & ~PTE_MASK; + val = (pfn_to_mfn(pfn) << PAGE_SHIFT) | flags; + } + + return val; +} + +pteval_t xen_pte_val(pte_t pte) +{ + return pte_mfn_to_pfn(pte.pte); } pgdval_t xen_pgd_val(pgd_t pgd) { - pgdval_t ret = pgd.pgd; - if (ret & _PAGE_PRESENT) - ret = machine_to_phys(XMADDR(ret)).paddr | _PAGE_PRESENT; - return ret; + return pte_mfn_to_pfn(pgd.pgd); } pte_t xen_make_pte(pteval_t pte) { - if (pte & _PAGE_PRESENT) - pte = phys_to_machine(XPADDR(pte)).maddr; - - return (pte_t){ .pte = pte }; + pte = pte_pfn_to_mfn(pte); + return native_make_pte(pte); } pgd_t xen_make_pgd(pgdval_t pgd) { - if (pgd & _PAGE_PRESENT) - pgd = phys_to_machine(XPADDR(pgd)).maddr; - - return (pgd_t){ pgd }; + pgd = pte_pfn_to_mfn(pgd); + return native_make_pgd(pgd); } pmdval_t xen_pmd_val(pmd_t pmd) { - pmdval_t ret = native_pmd_val(pmd); - if (ret & _PAGE_PRESENT) - ret = machine_to_phys(XMADDR(ret)).paddr | _PAGE_PRESENT; - return ret; + return pte_mfn_to_pfn(pmd.pmd); } #ifdef CONFIG_X86_PAE void xen_set_pud(pud_t *ptr, pud_t val) @@ -265,9 +273,7 @@ void xen_pmd_clear(pmd_t *pmdp) pmd_t xen_make_pmd(pmdval_t pmd) { - if (pmd & _PAGE_PRESENT) - pmd = phys_to_machine(XPADDR(pmd)).maddr; - + pmd = pte_pfn_to_mfn(pmd); return native_make_pmd(pmd); } #else /* !PAE */ diff --git a/arch/x86/xen/xen-head.S b/arch/x86/xen/xen-head.S index 288d587ce73c..3175e973fd0d 100644 --- a/arch/x86/xen/xen-head.S +++ b/arch/x86/xen/xen-head.S @@ -17,7 +17,7 @@ ENTRY(startup_xen) __FINIT -.pushsection .bss.page_aligned +.pushsection .text .align PAGE_SIZE_asm ENTRY(hypercall_page) .skip 0x1000 -- cgit v1.2.3 From 28499143933f19b28008a556ed59255d6009391a Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Fri, 9 May 2008 12:05:57 +0100 Subject: xen: remove support for non-PAE 32-bit Non-PAE operation has been deprecated in Xen for a while, and is rarely tested or used. xen-unstable has now officially dropped non-PAE support. Since Xen/pvops' non-PAE support has also been broken for a while, we may as well completely drop it altogether. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Ingo Molnar Signed-off-by: Thomas Gleixner Signed-off-by: Ingo Molnar --- arch/x86/xen/Kconfig | 2 +- arch/x86/xen/enlighten.c | 51 ++++++++++++++++------------------------------ arch/x86/xen/mmu.c | 19 ++--------------- arch/x86/xen/mmu.h | 24 ++++++---------------- arch/x86/xen/xen-head.S | 4 ---- include/asm-x86/xen/page.h | 4 ---- 6 files changed, 27 insertions(+), 77 deletions(-) (limited to 'arch/x86/xen/mmu.c') diff --git a/arch/x86/xen/Kconfig b/arch/x86/xen/Kconfig index 2e641be2737e..525b108411bd 100644 --- a/arch/x86/xen/Kconfig +++ b/arch/x86/xen/Kconfig @@ -6,7 +6,7 @@ config XEN bool "Xen guest support" select PARAVIRT depends on X86_32 - depends on X86_CMPXCHG && X86_TSC && !(X86_VISWS || X86_VOYAGER) + depends on X86_CMPXCHG && X86_TSC && X86_PAE && !(X86_VISWS || X86_VOYAGER) help This is the Linux Xen port. Enabling this will allow the kernel to boot in a paravirtualized environment under the diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index c048de34d6a1..f09c1c69c37a 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -785,38 +785,35 @@ static __init void xen_set_pte_init(pte_t *ptep, pte_t pte) static __init void xen_pagetable_setup_start(pgd_t *base) { pgd_t *xen_pgd = (pgd_t *)xen_start_info->pt_base; + int i; /* special set_pte for pagetable initialization */ pv_mmu_ops.set_pte = xen_set_pte_init; init_mm.pgd = base; /* - * copy top-level of Xen-supplied pagetable into place. For - * !PAE we can use this as-is, but for PAE it is a stand-in - * while we copy the pmd pages. + * copy top-level of Xen-supplied pagetable into place. This + * is a stand-in while we copy the pmd pages. */ memcpy(base, xen_pgd, PTRS_PER_PGD * sizeof(pgd_t)); - if (PTRS_PER_PMD > 1) { - int i; - /* - * For PAE, need to allocate new pmds, rather than - * share Xen's, since Xen doesn't like pmd's being - * shared between address spaces. - */ - for (i = 0; i < PTRS_PER_PGD; i++) { - if (pgd_val_ma(xen_pgd[i]) & _PAGE_PRESENT) { - pmd_t *pmd = (pmd_t *)alloc_bootmem_low_pages(PAGE_SIZE); + /* + * For PAE, need to allocate new pmds, rather than + * share Xen's, since Xen doesn't like pmd's being + * shared between address spaces. + */ + for (i = 0; i < PTRS_PER_PGD; i++) { + if (pgd_val_ma(xen_pgd[i]) & _PAGE_PRESENT) { + pmd_t *pmd = (pmd_t *)alloc_bootmem_low_pages(PAGE_SIZE); - memcpy(pmd, (void *)pgd_page_vaddr(xen_pgd[i]), - PAGE_SIZE); + memcpy(pmd, (void *)pgd_page_vaddr(xen_pgd[i]), + PAGE_SIZE); - make_lowmem_page_readonly(pmd); + make_lowmem_page_readonly(pmd); - set_pgd(&base[i], __pgd(1 + __pa(pmd))); - } else - pgd_clear(&base[i]); - } + set_pgd(&base[i], __pgd(1 + __pa(pmd))); + } else + pgd_clear(&base[i]); } /* make sure zero_page is mapped RO so we can use it in pagetables */ @@ -873,17 +870,7 @@ static __init void xen_pagetable_setup_done(pgd_t *base) /* Actually pin the pagetable down, but we can't set PG_pinned yet because the page structures don't exist yet. */ - { - unsigned level; - -#ifdef CONFIG_X86_PAE - level = MMUEXT_PIN_L3_TABLE; -#else - level = MMUEXT_PIN_L2_TABLE; -#endif - - pin_pagetable_pfn(level, PFN_DOWN(__pa(base))); - } + pin_pagetable_pfn(MMUEXT_PIN_L3_TABLE, PFN_DOWN(__pa(base))); } /* This is called once we have the cpu_possible_map */ @@ -1093,7 +1080,6 @@ static const struct pv_mmu_ops xen_mmu_ops __initdata = { .make_pte = xen_make_pte, .make_pgd = xen_make_pgd, -#ifdef CONFIG_X86_PAE .set_pte_atomic = xen_set_pte_atomic, .set_pte_present = xen_set_pte_at, .set_pud = xen_set_pud, @@ -1102,7 +1088,6 @@ static const struct pv_mmu_ops xen_mmu_ops __initdata = { .make_pmd = xen_make_pmd, .pmd_val = xen_pmd_val, -#endif /* PAE */ .activate_mm = xen_activate_mm, .dup_mmap = xen_dup_mmap, diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 265601d5a6ae..df40bf74ea75 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -228,7 +228,7 @@ pmdval_t xen_pmd_val(pmd_t pmd) { return pte_mfn_to_pfn(pmd.pmd); } -#ifdef CONFIG_X86_PAE + void xen_set_pud(pud_t *ptr, pud_t val) { struct multicall_space mcs; @@ -276,12 +276,6 @@ pmd_t xen_make_pmd(pmdval_t pmd) pmd = pte_pfn_to_mfn(pmd); return native_make_pmd(pmd); } -#else /* !PAE */ -void xen_set_pte(pte_t *ptep, pte_t pte) -{ - *ptep = pte; -} -#endif /* CONFIG_X86_PAE */ /* (Yet another) pagetable walker. This one is intended for pinning a @@ -434,8 +428,6 @@ static int pin_page(struct page *page, enum pt_level level) read-only, and can be pinned. */ void xen_pgd_pin(pgd_t *pgd) { - unsigned level; - xen_mc_batch(); if (pgd_walk(pgd, pin_page, TASK_SIZE)) { @@ -445,14 +437,7 @@ void xen_pgd_pin(pgd_t *pgd) xen_mc_batch(); } -#ifdef CONFIG_X86_PAE - level = MMUEXT_PIN_L3_TABLE; -#else - level = MMUEXT_PIN_L2_TABLE; -#endif - - xen_do_pin(level, PFN_DOWN(__pa(pgd))); - + xen_do_pin(MMUEXT_PIN_L3_TABLE, PFN_DOWN(__pa(pgd))); xen_mc_issue(0); } diff --git a/arch/x86/xen/mmu.h b/arch/x86/xen/mmu.h index b5e189b1519d..5fe961caffd4 100644 --- a/arch/x86/xen/mmu.h +++ b/arch/x86/xen/mmu.h @@ -37,14 +37,13 @@ void xen_exit_mmap(struct mm_struct *mm); void xen_pgd_pin(pgd_t *pgd); //void xen_pgd_unpin(pgd_t *pgd); -#ifdef CONFIG_X86_PAE -unsigned long long xen_pte_val(pte_t); -unsigned long long xen_pmd_val(pmd_t); -unsigned long long xen_pgd_val(pgd_t); +pteval_t xen_pte_val(pte_t); +pmdval_t xen_pmd_val(pmd_t); +pgdval_t xen_pgd_val(pgd_t); -pte_t xen_make_pte(unsigned long long); -pmd_t xen_make_pmd(unsigned long long); -pgd_t xen_make_pgd(unsigned long long); +pte_t xen_make_pte(pteval_t); +pmd_t xen_make_pmd(pmdval_t); +pgd_t xen_make_pgd(pgdval_t); void xen_set_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pteval); @@ -53,15 +52,4 @@ void xen_set_pud(pud_t *ptr, pud_t val); void xen_pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep); void xen_pmd_clear(pmd_t *pmdp); - -#else -unsigned long xen_pte_val(pte_t); -unsigned long xen_pmd_val(pmd_t); -unsigned long xen_pgd_val(pgd_t); - -pte_t xen_make_pte(unsigned long); -pmd_t xen_make_pmd(unsigned long); -pgd_t xen_make_pgd(unsigned long); -#endif - #endif /* _XEN_MMU_H */ diff --git a/arch/x86/xen/xen-head.S b/arch/x86/xen/xen-head.S index 3175e973fd0d..6ec3b4f7719b 100644 --- a/arch/x86/xen/xen-head.S +++ b/arch/x86/xen/xen-head.S @@ -30,11 +30,7 @@ ENTRY(hypercall_page) ELFNOTE(Xen, XEN_ELFNOTE_ENTRY, .long startup_xen) ELFNOTE(Xen, XEN_ELFNOTE_HYPERCALL_PAGE, .long hypercall_page) ELFNOTE(Xen, XEN_ELFNOTE_FEATURES, .asciz "!writable_page_tables|pae_pgdir_above_4gb") -#ifdef CONFIG_X86_PAE ELFNOTE(Xen, XEN_ELFNOTE_PAE_MODE, .asciz "yes") -#else - ELFNOTE(Xen, XEN_ELFNOTE_PAE_MODE, .asciz "no") -#endif ELFNOTE(Xen, XEN_ELFNOTE_LOADER, .asciz "generic") #endif /*CONFIG_XEN */ diff --git a/include/asm-x86/xen/page.h b/include/asm-x86/xen/page.h index baf3a4dce28c..e11f24038b1d 100644 --- a/include/asm-x86/xen/page.h +++ b/include/asm-x86/xen/page.h @@ -150,13 +150,9 @@ static inline pte_t __pte_ma(pteval_t x) return (pte_t) { .pte = x }; } -#ifdef CONFIG_X86_PAE #define pmd_val_ma(v) ((v).pmd) #define pud_val_ma(v) ((v).pgd.pgd) #define __pmd_ma(x) ((pmd_t) { (x) } ) -#else /* !X86_PAE */ -#define pmd_val_ma(v) ((v).pud.pgd.pgd) -#endif /* CONFIG_X86_PAE */ #define pgd_val_ma(x) ((x).pgd) -- cgit v1.2.3 From e57778a1e30470c9f5b79e370511b9af29b59c48 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 16 Jun 2008 04:30:02 -0700 Subject: xen: implement ptep_modify_prot_start/commit Xen has a pte update function which will update a pte while preserving its accessed and dirty bits. This means that ptep_modify_prot_start() can be implemented as a simple read of the pte value. The hardware may update the pte in the meantime, but ptep_modify_prot_commit() updates it while preserving any changes that may have happened in the meantime. The updates in ptep_modify_prot_commit() are batched if we're currently in lazy mmu mode. The mmu_update hypercall can take a batch of updates to perform, but this code doesn't make particular use of that feature, in favour of using generic multicall batching to get them all into the hypervisor. The net effect of this is that each mprotect pte update turns from two expensive trap-and-emulate faults into they hypervisor into a single hypercall whose cost is amortized in a batched multicall. Signed-off-by: Jeremy Fitzhardinge Acked-by: Linus Torvalds Acked-by: Hugh Dickins Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 13 ++++++++++--- arch/x86/xen/mmu.c | 21 +++++++++++++++++++++ arch/x86/xen/mmu.h | 4 ++++ include/xen/interface/features.h | 3 +++ include/xen/interface/xen.h | 9 +++++++-- 5 files changed, 45 insertions(+), 5 deletions(-) (limited to 'arch/x86/xen/mmu.c') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 0b7553cbc529..bd74229081c3 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -168,7 +168,9 @@ static void __init xen_banner(void) { printk(KERN_INFO "Booting paravirtualized kernel on %s\n", pv_info.name); - printk(KERN_INFO "Hypervisor signature: %s\n", xen_start_info->magic); + printk(KERN_INFO "Hypervisor signature: %s%s\n", + xen_start_info->magic, + xen_feature(XENFEAT_mmu_pt_update_preserve_ad) ? " (preserve-AD)" : ""); } static void xen_cpuid(unsigned int *ax, unsigned int *bx, @@ -1243,6 +1245,8 @@ asmlinkage void __init xen_start_kernel(void) 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; @@ -1252,14 +1256,17 @@ asmlinkage void __init xen_start_kernel(void) pv_apic_ops = xen_apic_ops; pv_mmu_ops = xen_mmu_ops; + 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; + } + machine_ops = xen_machine_ops; #ifdef CONFIG_SMP smp_ops = xen_smp_ops; #endif - xen_setup_features(); - /* Get mfn list */ if (!xen_feature(XENFEAT_auto_translated_physmap)) xen_build_dynamic_phys_to_machine(); diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 8132aa8c5d49..846dad7d54a5 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -323,6 +323,27 @@ out: preempt_enable(); } +pte_t xen_ptep_modify_prot_start(struct mm_struct *mm, unsigned long addr, pte_t *ptep) +{ + /* Just return the pte as-is. We preserve the bits on commit */ + return *ptep; +} + +void xen_ptep_modify_prot_commit(struct mm_struct *mm, unsigned long addr, + pte_t *ptep, pte_t pte) +{ + struct multicall_space mcs; + struct mmu_update *u; + + mcs = xen_mc_entry(sizeof(*u)); + u = mcs.args; + u->ptr = virt_to_machine(ptep).maddr | MMU_PT_UPDATE_PRESERVE_AD; + u->val = pte_val_ma(pte); + MULTI_mmu_update(mcs.mc, u, 1, NULL, DOMID_SELF); + + xen_mc_issue(PARAVIRT_LAZY_MMU); +} + /* Assume pteval_t is equivalent to all the other *val_t types. */ static pteval_t pte_mfn_to_pfn(pteval_t val) { diff --git a/arch/x86/xen/mmu.h b/arch/x86/xen/mmu.h index e3dd09e25c63..297bf9f5b8bc 100644 --- a/arch/x86/xen/mmu.h +++ b/arch/x86/xen/mmu.h @@ -52,4 +52,8 @@ void xen_set_pud_hyper(pud_t *ptr, pud_t val); void xen_pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep); void xen_pmd_clear(pmd_t *pmdp); +pte_t xen_ptep_modify_prot_start(struct mm_struct *mm, unsigned long addr, pte_t *ptep); +void xen_ptep_modify_prot_commit(struct mm_struct *mm, unsigned long addr, + pte_t *ptep, pte_t pte); + #endif /* _XEN_MMU_H */ diff --git a/include/xen/interface/features.h b/include/xen/interface/features.h index d73228d16488..f51b6413b054 100644 --- a/include/xen/interface/features.h +++ b/include/xen/interface/features.h @@ -38,6 +38,9 @@ */ #define XENFEAT_pae_pgdir_above_4gb 4 +/* x86: Does this Xen host support the MMU_PT_UPDATE_PRESERVE_AD hypercall? */ +#define XENFEAT_mmu_pt_update_preserve_ad 5 + #define XENFEAT_NR_SUBMAPS 1 #endif /* __XEN_PUBLIC_FEATURES_H__ */ diff --git a/include/xen/interface/xen.h b/include/xen/interface/xen.h index 819a0331cda9..2befa3e2f1bc 100644 --- a/include/xen/interface/xen.h +++ b/include/xen/interface/xen.h @@ -114,9 +114,14 @@ * ptr[:2] -- Machine address within the frame whose mapping to modify. * The frame must belong to the FD, if one is specified. * val -- Value to write into the mapping entry. + * + * ptr[1:0] == MMU_PT_UPDATE_PRESERVE_AD: + * As MMU_NORMAL_PT_UPDATE above, but A/D bits currently in the PTE are ORed + * with those in @val. */ -#define MMU_NORMAL_PT_UPDATE 0 /* checked '*ptr = val'. ptr is MA. */ -#define MMU_MACHPHYS_UPDATE 1 /* ptr = MA of frame to modify entry for */ +#define MMU_NORMAL_PT_UPDATE 0 /* checked '*ptr = val'. ptr is MA. */ +#define MMU_MACHPHYS_UPDATE 1 /* ptr = MA of frame to modify entry for */ +#define MMU_PT_UPDATE_PRESERVE_AD 2 /* atomically: *ptr = val | (*ptr&(A|D)) */ /* * MMU EXTENDED OPERATIONS -- cgit v1.2.3 From 400d34944c4ad82a817c06e570bc93b1114aa596 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 16 Jun 2008 04:30:03 -0700 Subject: xen: add mechanism to extend existing multicalls Some Xen hypercalls accept an array of operations to work on. In general this is because its more efficient for the hypercall to the work all at once rather than as separate hypercalls (even batched as a multicall). This patch adds a mechanism (xen_mc_extend_args()) to allocate more argument space to the last-issued multicall, in order to extend its argument list. The user of this mechanism is xen/mmu.c, which uses it to extend the args array of mmu_update. This is particularly valuable when doing the update for a large mprotect, which goes via ptep_modify_prot_commit(), but it also manages to batch updates to pgd/pmds as well. Signed-off-by: Jeremy Fitzhardinge Acked-by: Linus Torvalds Acked-by: Hugh Dickins Signed-off-by: Ingo Molnar --- arch/x86/xen/mmu.c | 55 ++++++++++++++++++++++++++++++----------------- arch/x86/xen/multicalls.c | 40 ++++++++++++++++++++++++++++------ arch/x86/xen/multicalls.h | 12 +++++++++++ 3 files changed, 81 insertions(+), 26 deletions(-) (limited to 'arch/x86/xen/mmu.c') diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 846dad7d54a5..f6b8225c2a0b 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -230,18 +230,35 @@ static bool page_pinned(void *ptr) return PagePinned(page); } -void xen_set_pmd_hyper(pmd_t *ptr, pmd_t val) +static void extend_mmu_update(const struct mmu_update *update) { struct multicall_space mcs; struct mmu_update *u; - preempt_disable(); + mcs = xen_mc_extend_args(__HYPERVISOR_mmu_update, sizeof(*u)); + + if (mcs.mc != NULL) + mcs.mc->args[1]++; + else { + mcs = __xen_mc_entry(sizeof(*u)); + MULTI_mmu_update(mcs.mc, mcs.args, 1, NULL, DOMID_SELF); + } - mcs = xen_mc_entry(sizeof(*u)); u = mcs.args; - u->ptr = virt_to_machine(ptr).maddr; - u->val = pmd_val_ma(val); - MULTI_mmu_update(mcs.mc, u, 1, NULL, DOMID_SELF); + *u = *update; +} + +void xen_set_pmd_hyper(pmd_t *ptr, pmd_t val) +{ + struct mmu_update u; + + preempt_disable(); + + xen_mc_batch(); + + u.ptr = virt_to_machine(ptr).maddr; + u.val = pmd_val_ma(val); + extend_mmu_update(&u); xen_mc_issue(PARAVIRT_LAZY_MMU); @@ -332,14 +349,13 @@ pte_t xen_ptep_modify_prot_start(struct mm_struct *mm, unsigned long addr, pte_t void xen_ptep_modify_prot_commit(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pte) { - struct multicall_space mcs; - struct mmu_update *u; + struct mmu_update u; - mcs = xen_mc_entry(sizeof(*u)); - u = mcs.args; - u->ptr = virt_to_machine(ptep).maddr | MMU_PT_UPDATE_PRESERVE_AD; - u->val = pte_val_ma(pte); - MULTI_mmu_update(mcs.mc, u, 1, NULL, DOMID_SELF); + xen_mc_batch(); + + u.ptr = virt_to_machine(ptep).maddr | MMU_PT_UPDATE_PRESERVE_AD; + u.val = pte_val_ma(pte); + extend_mmu_update(&u); xen_mc_issue(PARAVIRT_LAZY_MMU); } @@ -396,16 +412,15 @@ pmdval_t xen_pmd_val(pmd_t pmd) void xen_set_pud_hyper(pud_t *ptr, pud_t val) { - struct multicall_space mcs; - struct mmu_update *u; + struct mmu_update u; preempt_disable(); - mcs = xen_mc_entry(sizeof(*u)); - u = mcs.args; - u->ptr = virt_to_machine(ptr).maddr; - u->val = pud_val_ma(val); - MULTI_mmu_update(mcs.mc, u, 1, NULL, DOMID_SELF); + xen_mc_batch(); + + u.ptr = virt_to_machine(ptr).maddr; + u.val = pud_val_ma(val); + extend_mmu_update(&u); xen_mc_issue(PARAVIRT_LAZY_MMU); diff --git a/arch/x86/xen/multicalls.c b/arch/x86/xen/multicalls.c index 5791eb2e3750..3c63c4da7ed1 100644 --- a/arch/x86/xen/multicalls.c +++ b/arch/x86/xen/multicalls.c @@ -29,14 +29,14 @@ #define MC_DEBUG 1 #define MC_BATCH 32 -#define MC_ARGS (MC_BATCH * 16 / sizeof(u64)) +#define MC_ARGS (MC_BATCH * 16) struct mc_buffer { struct multicall_entry entries[MC_BATCH]; #if MC_DEBUG struct multicall_entry debug[MC_BATCH]; #endif - u64 args[MC_ARGS]; + unsigned char args[MC_ARGS]; struct callback { void (*fn)(void *); void *data; @@ -107,20 +107,48 @@ struct multicall_space __xen_mc_entry(size_t args) { struct mc_buffer *b = &__get_cpu_var(mc_buffer); struct multicall_space ret; - unsigned argspace = (args + sizeof(u64) - 1) / sizeof(u64); + unsigned argidx = roundup(b->argidx, sizeof(u64)); BUG_ON(preemptible()); - BUG_ON(argspace > MC_ARGS); + BUG_ON(b->argidx > MC_ARGS); if (b->mcidx == MC_BATCH || - (b->argidx + argspace) > MC_ARGS) + (argidx + args) > MC_ARGS) { xen_mc_flush(); + argidx = roundup(b->argidx, sizeof(u64)); + } ret.mc = &b->entries[b->mcidx]; b->mcidx++; + ret.args = &b->args[argidx]; + b->argidx = argidx + args; + + BUG_ON(b->argidx > MC_ARGS); + return ret; +} + +struct multicall_space xen_mc_extend_args(unsigned long op, size_t size) +{ + struct mc_buffer *b = &__get_cpu_var(mc_buffer); + struct multicall_space ret = { NULL, NULL }; + + BUG_ON(preemptible()); + BUG_ON(b->argidx > MC_ARGS); + + if (b->mcidx == 0) + return ret; + + if (b->entries[b->mcidx - 1].op != op) + return ret; + + if ((b->argidx + size) > MC_ARGS) + return ret; + + ret.mc = &b->entries[b->mcidx - 1]; ret.args = &b->args[b->argidx]; - b->argidx += argspace; + b->argidx += size; + BUG_ON(b->argidx > MC_ARGS); return ret; } diff --git a/arch/x86/xen/multicalls.h b/arch/x86/xen/multicalls.h index 8bae996d99a3..858938241616 100644 --- a/arch/x86/xen/multicalls.h +++ b/arch/x86/xen/multicalls.h @@ -45,4 +45,16 @@ static inline void xen_mc_issue(unsigned mode) /* Set up a callback to be called when the current batch is flushed */ void xen_mc_callback(void (*fn)(void *), void *data); +/* + * Try to extend the arguments of the previous multicall command. The + * previous command's op must match. If it does, then it attempts to + * extend the argument space allocated to the multicall entry by + * arg_size bytes. + * + * The returned multicall_space will return with mc pointing to the + * command on success, or NULL on failure, and args pointing to the + * newly allocated space. + */ +struct multicall_space xen_mc_extend_args(unsigned long op, size_t arg_size); + #endif /* _XEN_MULTICALLS_H */ -- cgit v1.2.3 From d8355aca23863be659ec5b7e0393cfbfa91ec221 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Thu, 3 Jul 2008 22:10:18 -0700 Subject: xen: fix address truncation in pte mfn<->pfn conversion When converting the page number in a pte/pmd/pud/pgd between machine and pseudo-physical addresses, the converted result was being truncated at 32-bits. This caused failures on machines with more than 4G of physical memory. Signed-off-by: Jeremy Fitzhardinge Cc: "Christopher S. Aker" Cc: Ian Campbell Signed-off-by: Ingo Molnar --- arch/x86/xen/mmu.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'arch/x86/xen/mmu.c') diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index df40bf74ea75..4e527e7893a8 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -185,7 +185,7 @@ static pteval_t pte_mfn_to_pfn(pteval_t val) if (val & _PAGE_PRESENT) { unsigned long mfn = (val & PTE_MASK) >> PAGE_SHIFT; pteval_t flags = val & ~PTE_MASK; - val = (mfn_to_pfn(mfn) << PAGE_SHIFT) | flags; + val = ((pteval_t)mfn_to_pfn(mfn) << PAGE_SHIFT) | flags; } return val; @@ -196,7 +196,7 @@ static pteval_t pte_pfn_to_mfn(pteval_t val) if (val & _PAGE_PRESENT) { unsigned long pfn = (val & PTE_MASK) >> PAGE_SHIFT; pteval_t flags = val & ~PTE_MASK; - val = (pfn_to_mfn(pfn) << PAGE_SHIFT) | flags; + val = ((pteval_t)pfn_to_mfn(pfn) << PAGE_SHIFT) | flags; } return val; -- cgit v1.2.3