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/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 ---- 5 files changed, 27 insertions(+), 73 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/Kconfig b/arch/x86/xen/Kconfig index 2e641be2737..525b108411b 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 c8a56e457d6..a05b7721bc8 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 126766d43ae..07c2653ec33 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); } diff --git a/arch/x86/xen/mmu.h b/arch/x86/xen/mmu.h index b5e189b1519..5fe961caffd 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 288d587ce73..2ab5f42f34d 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 */ -- cgit v1.2.3 From 7b1333aa4cb546ddeb9c05098a53d9a777623a05 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 26 May 2008 23:31:01 +0100 Subject: xen: use hypercall rather than clts Xen will trap and emulate clts, but its better to use a hypercall. Also, xenner doesn't handle clts. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Thomas Gleixner --- arch/x86/xen/enlighten.c | 28 ++++++++++++++++++++++++++-- 1 file changed, 26 insertions(+), 2 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index a05b7721bc8..446f4cd649e 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -607,6 +607,30 @@ static void xen_flush_tlb_others(const cpumask_t *cpus, struct mm_struct *mm, xen_mc_issue(PARAVIRT_LAZY_MMU); } +static void xen_clts(void) +{ + struct multicall_space mcs; + + mcs = xen_mc_entry(0); + + MULTI_fpu_taskswitch(mcs.mc, 0); + + xen_mc_issue(PARAVIRT_LAZY_CPU); +} + +static void xen_write_cr0(unsigned long cr0) +{ + struct multicall_space mcs; + + /* Only pay attention to cr0.TS; everything else is + ignored. */ + mcs = xen_mc_entry(0); + + MULTI_fpu_taskswitch(mcs.mc, (cr0 & X86_CR0_TS) != 0); + + xen_mc_issue(PARAVIRT_LAZY_CPU); +} + static void xen_write_cr2(unsigned long cr2) { x86_read_percpu(xen_vcpu)->arch.cr2 = cr2; @@ -978,10 +1002,10 @@ static const struct pv_cpu_ops xen_cpu_ops __initdata = { .set_debugreg = xen_set_debugreg, .get_debugreg = xen_get_debugreg, - .clts = native_clts, + .clts = xen_clts, .read_cr0 = native_read_cr0, - .write_cr0 = native_write_cr0, + .write_cr0 = xen_write_cr0, .read_cr4 = native_read_cr4, .read_cr4_safe = native_read_cr4_safe, -- cgit v1.2.3 From 349c709f42453707f74bece0d9d35ee5b3842893 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 26 May 2008 23:31:02 +0100 Subject: xen: use new sched_op Use the new sched_op hypercall, mainly because xenner doesn't support the old one. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Thomas Gleixner --- arch/x86/xen/enlighten.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 446f4cd649e..35ddaf50180 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -254,7 +254,7 @@ static void xen_irq_enable(void) static void xen_safe_halt(void) { /* Blocking includes an implicit local_irq_enable(). */ - if (HYPERVISOR_sched_op(SCHEDOP_block, 0) != 0) + if (HYPERVISOR_sched_op(SCHEDOP_block, NULL) != 0) BUG(); } @@ -1138,11 +1138,13 @@ static const struct smp_ops xen_smp_ops __initdata = { static void xen_reboot(int reason) { + struct sched_shutdown r = { .reason = reason }; + #ifdef CONFIG_SMP smp_send_stop(); #endif - if (HYPERVISOR_sched_op(SCHEDOP_shutdown, reason)) + if (HYPERVISOR_sched_op(SCHEDOP_shutdown, &r)) BUG(); } -- cgit v1.2.3 From 2956a3511c8c5dccb1d4739ead17c7c3c23a24b7 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 26 May 2008 23:31:04 +0100 Subject: xen: allow some cr4 updates The guest can legitimately change things like cr4.OSFXSR and OSXMMEXCPT, so let it. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Thomas Gleixner --- arch/x86/xen/enlighten.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 35ddaf50180..f6876485ee7 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -648,8 +648,10 @@ static unsigned long xen_read_cr2_direct(void) static void xen_write_cr4(unsigned long cr4) { - /* Just ignore cr4 changes; Xen doesn't allow us to do - anything anyway. */ + cr4 &= ~X86_CR4_PGE; + cr4 &= ~X86_CR4_PSE; + + native_write_cr4(cr4); } static unsigned long xen_read_cr3(void) -- cgit v1.2.3 From 239d1fc04ed0b58d638096b12a7f6d50269d30c9 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 26 May 2008 23:31:05 +0100 Subject: xen: don't worry about preempt during xen_irq_enable() When enabling interrupts, we don't need to worry about preemption, because we either enter with interrupts disabled - so no preemption - or the caller is confused and is re-enabling interrupts on some indeterminate processor. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Thomas Gleixner --- arch/x86/xen/enlighten.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index f6876485ee7..4a372b71239 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -235,13 +235,13 @@ static void xen_irq_enable(void) { struct vcpu_info *vcpu; - /* There's a one instruction preempt window here. We need to - make sure we're don't switch CPUs between getting the vcpu - pointer and updating the mask. */ - preempt_disable(); + /* We don't need to worry about being preempted here, since + either a) interrupts are disabled, so no preemption, or b) + the caller is confused and is trying to re-enable interrupts + on an indeterminate processor. */ + vcpu = x86_read_percpu(xen_vcpu); vcpu->evtchn_upcall_mask = 0; - preempt_enable_no_resched(); /* Doesn't matter if we get preempted here, because any pending event will get dealt with anyway. */ -- cgit v1.2.3 From a15af1c9ea2750a9ff01e51615c45950bad8221b Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 26 May 2008 23:31:06 +0100 Subject: x86/paravirt: add pte_flags to just get pte flags Add pte_flags() to extract the flags from a pte. This is a special case of pte_val() which is only guaranteed to return the pte's flags correctly; the page number may be corrupted or missing. The intent is to allow paravirt implementations to return pte flags without having to do any translation of the page number (most notably, Xen). Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Thomas Gleixner --- arch/x86/xen/enlighten.c | 1 + 1 file changed, 1 insertion(+) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 4a372b71239..1b4b5fa498b 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1101,6 +1101,7 @@ static const struct pv_mmu_ops xen_mmu_ops __initdata = { .set_pmd = xen_set_pmd, .pte_val = xen_pte_val, + .pte_flags = native_pte_val, .pgd_val = xen_pgd_val, .make_pte = xen_make_pte, -- cgit v1.2.3 From 9e124fe16ff24746d6de5a2ad685266d7bce0e08 Mon Sep 17 00:00:00 2001 From: Markus Armbruster Date: Mon, 26 May 2008 23:31:07 +0100 Subject: xen: Enable console tty by default in domU if it's not a dummy Without console= arguments on the kernel command line, the first console to register becomes enabled and the preferred console (the one behind /dev/console). This is normally tty (assuming CONFIG_VT_CONSOLE is enabled, which it commonly is). This is okay as long tty is a useful console. But unless we have the PV framebuffer, and it is enabled for this domain, tty0 in domU is merely a dummy. In that case, we want the preferred console to be the Xen console hvc0, and we want it without having to fiddle with the kernel command line. Commit b8c2d3dfbc117dff26058fbac316b8acfc2cb5f7 did that for us. Since we now have the PV framebuffer, we want to enable and prefer tty again, but only when PVFB is enabled. But even then we still want to enable the Xen console as well. Problem: when tty registers, we can't yet know whether the PVFB is enabled. By the time we can know (xenstore is up), the console setup game is over. Solution: enable console tty by default, but keep hvc as the preferred console. Change the preferred console to tty when PVFB probes successfully, unless we've been given console kernel parameters. Signed-off-by: Markus Armbruster Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Thomas Gleixner --- arch/x86/xen/enlighten.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 1b4b5fa498b..6cfb708408e 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1256,8 +1256,10 @@ asmlinkage void __init xen_start_kernel(void) ? __pa(xen_start_info->mod_start) : 0; boot_params.hdr.ramdisk_size = xen_start_info->mod_len; - if (!is_initial_xendomain()) + if (!is_initial_xendomain()) { + add_preferred_console("tty", 0, NULL); add_preferred_console("hvc", 0, NULL); + } /* Start the world */ start_kernel(); -- cgit v1.2.3 From 83abc70a4c6e306f4c1672e25884322f797e4fcb Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 26 May 2008 23:31:12 +0100 Subject: xen: make earlyprintk=xen work again For some perverse reason, if you call add_preferred_console() it prevents setup_early_printk() from successfully enabling the boot console - unless you make it a preferred console too... Also, make xenboot console output distinct from normal console output, since it gets repeated when the console handover happens, and the duplicated output is confusing without disambiguation. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Thomas Gleixner Cc: Markus Armbruster Cc: Gerd Hoffmann --- arch/x86/xen/enlighten.c | 1 + 1 file changed, 1 insertion(+) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 6cfb708408e..5c0635a8bff 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1257,6 +1257,7 @@ asmlinkage void __init xen_start_kernel(void) boot_params.hdr.ramdisk_size = xen_start_info->mod_len; if (!is_initial_xendomain()) { + add_preferred_console("xenboot", 0, NULL); add_preferred_console("tty", 0, NULL); add_preferred_console("hvc", 0, NULL); } -- cgit v1.2.3 From ec9b2065d4d3b797604c09a569083dd9ff951b1b Mon Sep 17 00:00:00 2001 From: Isaku Yamahata Date: Mon, 26 May 2008 23:31:13 +0100 Subject: xen: Move manage.c to drivers/xen for ia64/xen support move arch/x86/xen/manage.c under drivers/xen/to share codes with x86 and ia64. ia64/xen also uses manage.c Signed-off-by: Isaku Yamahata Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Thomas Gleixner --- arch/x86/xen/Makefile | 2 +- arch/x86/xen/manage.c | 143 -------------------------------------------------- 2 files changed, 1 insertion(+), 144 deletions(-) delete mode 100644 arch/x86/xen/manage.c (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/Makefile b/arch/x86/xen/Makefile index 3d8df981d5f..40b119b6b10 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 manage.o xen-asm.o grant-table.o + time.o xen-asm.o grant-table.o obj-$(CONFIG_SMP) += smp.o diff --git a/arch/x86/xen/manage.c b/arch/x86/xen/manage.c deleted file mode 100644 index aa7af9e6abc..00000000000 --- a/arch/x86/xen/manage.c +++ /dev/null @@ -1,143 +0,0 @@ -/* - * Handle extern requests for shutdown, reboot and sysrq - */ -#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 - -/* Ignore multiple shutdown requests. */ -static int shutting_down = SHUTDOWN_INVALID; - -static void shutdown_handler(struct xenbus_watch *watch, - const char **vec, unsigned int len) -{ - char *str; - struct xenbus_transaction xbt; - int err; - - if (shutting_down != SHUTDOWN_INVALID) - return; - - again: - err = xenbus_transaction_start(&xbt); - if (err) - return; - - str = (char *)xenbus_read(xbt, "control", "shutdown", NULL); - /* Ignore read errors and empty reads. */ - if (XENBUS_IS_ERR_READ(str)) { - xenbus_transaction_end(xbt, 1); - return; - } - - xenbus_write(xbt, "control", "shutdown", ""); - - err = xenbus_transaction_end(xbt, 0); - if (err == -EAGAIN) { - kfree(str); - goto again; - } - - if (strcmp(str, "poweroff") == 0 || - strcmp(str, "halt") == 0) - orderly_poweroff(false); - else if (strcmp(str, "reboot") == 0) - ctrl_alt_del(); - else { - printk(KERN_INFO "Ignoring shutdown request: %s\n", str); - shutting_down = SHUTDOWN_INVALID; - } - - kfree(str); -} - -static void sysrq_handler(struct xenbus_watch *watch, const char **vec, - unsigned int len) -{ - char sysrq_key = '\0'; - struct xenbus_transaction xbt; - int err; - - again: - err = xenbus_transaction_start(&xbt); - if (err) - return; - if (!xenbus_scanf(xbt, "control", "sysrq", "%c", &sysrq_key)) { - printk(KERN_ERR "Unable to read sysrq code in " - "control/sysrq\n"); - xenbus_transaction_end(xbt, 1); - return; - } - - if (sysrq_key != '\0') - xenbus_printf(xbt, "control", "sysrq", "%c", '\0'); - - err = xenbus_transaction_end(xbt, 0); - if (err == -EAGAIN) - goto again; - - if (sysrq_key != '\0') - handle_sysrq(sysrq_key, NULL); -} - -static struct xenbus_watch shutdown_watch = { - .node = "control/shutdown", - .callback = shutdown_handler -}; - -static struct xenbus_watch sysrq_watch = { - .node = "control/sysrq", - .callback = sysrq_handler -}; - -static int setup_shutdown_watcher(void) -{ - int err; - - err = register_xenbus_watch(&shutdown_watch); - if (err) { - printk(KERN_ERR "Failed to set shutdown watcher\n"); - return err; - } - - err = register_xenbus_watch(&sysrq_watch); - if (err) { - printk(KERN_ERR "Failed to set sysrq watcher\n"); - return err; - } - - return 0; -} - -static int shutdown_event(struct notifier_block *notifier, - unsigned long event, - void *data) -{ - setup_shutdown_watcher(); - return NOTIFY_DONE; -} - -static int __init setup_shutdown_event(void) -{ - static struct notifier_block xenstore_notifier = { - .notifier_call = shutdown_event - }; - register_xenstore_notifier(&xenstore_notifier); - - return 0; -} - -subsys_initcall(setup_shutdown_event); -- cgit v1.2.3 From 38bb5ab4179572f4d24d3ca7188172a31ca51a69 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 26 May 2008 23:31:16 +0100 Subject: xen: count resched interrupts properly Make sure resched interrupts appear in /proc/interrupts in the proper place. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Thomas Gleixner --- arch/x86/xen/smp.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c index 94e69000f98..74ab8968c52 100644 --- a/arch/x86/xen/smp.c +++ b/arch/x86/xen/smp.c @@ -65,6 +65,12 @@ static struct call_data_struct *call_data; */ static irqreturn_t xen_reschedule_interrupt(int irq, void *dev_id) { +#ifdef CONFIG_X86_32 + __get_cpu_var(irq_stat).irq_resched_count++; +#else + add_pda(irq_resched_count, 1); +#endif + return IRQ_HANDLED; } -- 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 ++ 4 files changed, 88 insertions(+), 3 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 5c0635a8bff..73d3c84a349 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 07c2653ec33..c3b27dec6f0 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 82517e4a752..37f8f0b8f74 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 f1063ae0803..7bdc8c5c924 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); -- 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 +++ 3 files changed, 25 insertions(+), 13 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/Kconfig b/arch/x86/xen/Kconfig index 525b108411b..d0f1c7c5dc3 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 c3b27dec6f0..644232aa7bf 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 37f8f0b8f74..488447878a9 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); -- 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') diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 644232aa7bf..da7b45b0506 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 a0d695c821544947342a2d372ec4108bc813b979 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 26 May 2008 23:31:21 +0100 Subject: xen: make dummy_shared_info non-static Rename dummy_shared_info to xen_dummy_shared_info and make it non-static, in anticipation of users outside of enlighten.c Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Thomas Gleixner --- arch/x86/xen/enlighten.c | 6 +++--- arch/x86/xen/xen-ops.h | 1 + 2 files changed, 4 insertions(+), 3 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 73d3c84a349..d9faaa26c10 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -75,13 +75,13 @@ DEFINE_PER_CPU(unsigned long, xen_current_cr3); /* actual vcpu cr3 */ struct start_info *xen_start_info; EXPORT_SYMBOL_GPL(xen_start_info); -static /* __initdata */ struct shared_info dummy_shared_info; +struct shared_info xen_dummy_shared_info; /* * Point at some empty memory to start with. We map the real shared_info * page as soon as fixmap is up and running. */ -struct shared_info *HYPERVISOR_shared_info = (void *)&dummy_shared_info; +struct shared_info *HYPERVISOR_shared_info = (void *)&xen_dummy_shared_info; /* * Flag to determine whether vcpu info placement is available on all @@ -104,7 +104,7 @@ static void __init xen_vcpu_setup(int cpu) int err; struct vcpu_info *vcpup; - BUG_ON(HYPERVISOR_shared_info == &dummy_shared_info); + BUG_ON(HYPERVISOR_shared_info == &xen_dummy_shared_info); per_cpu(xen_vcpu, cpu) = &HYPERVISOR_shared_info->vcpu_info[cpu]; if (!have_vcpu_info_placement) diff --git a/arch/x86/xen/xen-ops.h b/arch/x86/xen/xen-ops.h index 7bdc8c5c924..826d6e4c3da 100644 --- a/arch/x86/xen/xen-ops.h +++ b/arch/x86/xen/xen-ops.h @@ -15,6 +15,7 @@ DECLARE_PER_CPU(unsigned long, xen_cr3); DECLARE_PER_CPU(unsigned long, xen_current_cr3); extern struct start_info *xen_start_info; +extern struct shared_info xen_dummy_shared_info; extern struct shared_info *HYPERVISOR_shared_info; char * __init xen_memory_setup(void); -- 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') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index d9faaa26c10..ce67dc8f36a 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 da7b45b0506..4740cda3656 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 826d6e4c3da..a1bc89a8f16 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 ++++ 7 files changed, 105 insertions(+), 5 deletions(-) create mode 100644 arch/x86/xen/suspend.c (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/Makefile b/arch/x86/xen/Makefile index 40b119b6b10..2ba2d164913 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 ce67dc8f36a..b94f63ac228 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 4740cda3656..e95955968ba 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 74ab8968c52..d2e3c20127d 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 00000000000..7620a16fe53 --- /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 c39e1a5aa24..0bef256e5f2 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 a1bc89a8f16..a0503acad66 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 */ -- cgit v1.2.3 From 359cdd3f866b6219a6729e313faf2221397f3278 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 26 May 2008 23:31:28 +0100 Subject: xen: maintain clock offset over save/restore Hook into the device model to make sure that timekeeping's resume handler is called. This deals with our clocksource's non-monotonicity over the save/restore. Explicitly call clock_has_changed() to make sure that all the timers get retriggered properly. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Thomas Gleixner --- arch/x86/xen/time.c | 8 -------- 1 file changed, 8 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/time.c b/arch/x86/xen/time.c index 0bef256e5f2..c39e1a5aa24 100644 --- a/arch/x86/xen/time.c +++ b/arch/x86/xen/time.c @@ -572,14 +572,6 @@ 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(); -- 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') diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index e95955968ba..eef3b5c6e2f 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 f0d43100f13be0fa5bf52741d7084bb27f00e621 Mon Sep 17 00:00:00 2001 From: Yinghai Lu Date: Thu, 29 May 2008 12:56:36 -0700 Subject: x86: extend e820 early_res support 32bit -fix #3 introduce init_pg_table_start, so xen PV could specify the value. Signed-off-by: Yinghai Lu Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index c8a56e457d6..ccc7b84ddab 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1211,6 +1211,7 @@ asmlinkage void __init xen_start_kernel(void) pgd = (pgd_t *)xen_start_info->pt_base; + init_pg_tables_start = __pa(pgd); init_pg_tables_end = __pa(pgd) + xen_start_info->nr_pt_frames*PAGE_SIZE; init_mm.pgd = pgd; /* use the Xen pagetables to start */ @@ -1246,5 +1247,5 @@ asmlinkage void __init xen_start_kernel(void) add_preferred_console("hvc", 0, NULL); /* Start the world */ - start_kernel(); + i386_start_kernel(); } -- 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') diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index eef3b5c6e2f..17f374eb1fa 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') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index b94f63ac228..ed9f04b3836 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 17f374eb1fa..4fa0934db92 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 5fe961caffd..e3dd09e25c6 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 9c7a794209f8a91f47697c3be20597eb60531e6d Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Sat, 31 May 2008 01:33:02 +0100 Subject: xen: restore vcpu_info mapping If we're using vcpu_info mapping, then make sure its restored on all processors before relasing them from stop_machine. The only complication is that if this fails, we can't continue because we've already made assumptions that the mapping is available (baked in calls to the _direct versions of the functions, for example). Fortunately this can only happen with a 32-bit hypervisor, which may possibly run out of mapping space. On a 64-bit hypervisor, this is a non-issue. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 30 +++++++++++++++++++++++++++++- arch/x86/xen/suspend.c | 4 +++- arch/x86/xen/xen-ops.h | 1 + 3 files changed, 33 insertions(+), 2 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index ed9f04b3836..8e6152e6ed8 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -98,7 +98,7 @@ struct shared_info *HYPERVISOR_shared_info = (void *)&xen_dummy_shared_info; */ static int have_vcpu_info_placement = 1; -static void __init xen_vcpu_setup(int cpu) +static void xen_vcpu_setup(int cpu) { struct vcpu_register_vcpu_info info; int err; @@ -136,6 +136,34 @@ static void __init xen_vcpu_setup(int cpu) } } +/* + * On restore, set the vcpu placement up again. + * If it fails, then we're in a bad state, since + * we can't back out from using it... + */ +void xen_vcpu_restore(void) +{ + if (have_vcpu_info_placement) { + int cpu; + + for_each_online_cpu(cpu) { + bool other_cpu = (cpu != smp_processor_id()); + + if (other_cpu && + HYPERVISOR_vcpu_op(VCPUOP_down, cpu, NULL)) + BUG(); + + xen_vcpu_setup(cpu); + + if (other_cpu && + HYPERVISOR_vcpu_op(VCPUOP_up, cpu, NULL)) + BUG(); + } + + BUG_ON(!have_vcpu_info_placement); + } +} + static void __init xen_banner(void) { printk(KERN_INFO "Booting paravirtualized kernel on %s\n", diff --git a/arch/x86/xen/suspend.c b/arch/x86/xen/suspend.c index 7620a16fe53..7ab10f6b39a 100644 --- a/arch/x86/xen/suspend.c +++ b/arch/x86/xen/suspend.c @@ -26,6 +26,8 @@ void xen_pre_suspend(void) void xen_post_suspend(int suspend_cancelled) { + xen_setup_shared_info(); + if (suspend_cancelled) { xen_start_info->store_mfn = pfn_to_mfn(xen_start_info->store_mfn); @@ -35,8 +37,8 @@ void xen_post_suspend(int suspend_cancelled) #ifdef CONFIG_SMP xen_cpu_initialized_map = cpu_online_map; #endif + xen_vcpu_restore(); } - xen_setup_shared_info(); } diff --git a/arch/x86/xen/xen-ops.h b/arch/x86/xen/xen-ops.h index a0503acad66..a457e03e1b2 100644 --- a/arch/x86/xen/xen-ops.h +++ b/arch/x86/xen/xen-ops.h @@ -26,6 +26,7 @@ char * __init xen_memory_setup(void); void __init xen_arch_setup(void); void __init xen_init_IRQ(void); void xen_enable_sysenter(void); +void xen_vcpu_restore(void); void __init xen_build_dynamic_phys_to_machine(void); -- cgit v1.2.3 From d07af1f0e3a3e378074fc36322dd7b0e72d9a3e2 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Sat, 31 May 2008 01:33:03 +0100 Subject: xen: resume timers on all vcpus On resume, the vcpu timer modes will not be restored. The timer infrastructure doesn't do this for us, since it assumes the cpus are offline. We can just poke the other vcpus into the right mode directly though. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Ingo Molnar --- arch/x86/xen/suspend.c | 1 + arch/x86/xen/time.c | 13 +++++++++++++ arch/x86/xen/xen-ops.h | 1 + 3 files changed, 15 insertions(+) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/suspend.c b/arch/x86/xen/suspend.c index 7ab10f6b39a..251669a932d 100644 --- a/arch/x86/xen/suspend.c +++ b/arch/x86/xen/suspend.c @@ -38,6 +38,7 @@ void xen_post_suspend(int suspend_cancelled) xen_cpu_initialized_map = cpu_online_map; #endif xen_vcpu_restore(); + xen_timer_resume(); } } diff --git a/arch/x86/xen/time.c b/arch/x86/xen/time.c index c39e1a5aa24..ea137fb660c 100644 --- a/arch/x86/xen/time.c +++ b/arch/x86/xen/time.c @@ -572,6 +572,19 @@ void xen_setup_cpu_clockevents(void) clockevents_register_device(&__get_cpu_var(xen_clock_events)); } +void xen_timer_resume(void) +{ + int cpu; + + if (xen_clockevent != &xen_vcpuop_clockevent) + return; + + for_each_online_cpu(cpu) { + if (HYPERVISOR_vcpu_op(VCPUOP_stop_periodic_timer, cpu, NULL)) + BUG(); + } +} + __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 a457e03e1b2..9a055592a30 100644 --- a/arch/x86/xen/xen-ops.h +++ b/arch/x86/xen/xen-ops.h @@ -37,6 +37,7 @@ void __init xen_time_init(void); unsigned long xen_get_wallclock(void); int xen_set_wallclock(unsigned long time); unsigned long long xen_sched_clock(void); +void xen_timer_resume(void); irqreturn_t xen_debug_interrupt(int irq, void *dev_id); -- cgit v1.2.3 From 7e0edc1bc343231029084761ebf59e522902eb49 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Sat, 31 May 2008 01:33:04 +0100 Subject: xen: add new Xen elfnote types and use them appropriately Define recently added XEN_ELFNOTEs, and use them appropriately. Most significantly, this enables domain checkpointing (xm save -c). Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Ingo Molnar --- arch/x86/xen/xen-head.S | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/xen-head.S b/arch/x86/xen/xen-head.S index 2ab5f42f34d..ef6c9e005f9 100644 --- a/arch/x86/xen/xen-head.S +++ b/arch/x86/xen/xen-head.S @@ -7,6 +7,7 @@ #include #include #include +#include __INIT ENTRY(startup_xen) @@ -32,5 +33,9 @@ ENTRY(hypercall_page) ELFNOTE(Xen, XEN_ELFNOTE_FEATURES, .asciz "!writable_page_tables|pae_pgdir_above_4gb") ELFNOTE(Xen, XEN_ELFNOTE_PAE_MODE, .asciz "yes") ELFNOTE(Xen, XEN_ELFNOTE_LOADER, .asciz "generic") + ELFNOTE(Xen, XEN_ELFNOTE_L1_MFN_VALID, + .quad _PAGE_PRESENT; .quad _PAGE_PRESENT) + ELFNOTE(Xen, XEN_ELFNOTE_SUSPEND_CANCEL, .long 1) + ELFNOTE(Xen, XEN_ELFNOTE_HV_START_LOW, .long __HYPERVISOR_VIRT_START) #endif /*CONFIG_XEN */ -- 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') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 8e6152e6ed8..73fb0c4c150 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 7c9935858f9..f0078d9a5b7 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') diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index f0078d9a5b7..8132aa8c5d4 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 ef6c9e005f9..7c0cf6320a0 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 aeaaa59c7e15dcfaaf57ce069ef81683067d575d Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 17 Jun 2008 11:42:01 -0700 Subject: x86/paravirt/xen: add set_fixmap pv_mmu_ops Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Juan Quintela Signed-off-by: Eduardo Habkost Signed-off-by: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 29 +++++++++++++++++++++++++++++ 1 file changed, 29 insertions(+) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index c8a56e457d6..0ad8a64a2e0 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -960,6 +960,33 @@ static unsigned xen_patch(u8 type, u16 clobbers, void *insnbuf, return ret; } +static void xen_set_fixmap(unsigned idx, unsigned long phys, pgprot_t prot) +{ + pte_t pte; + + phys >>= PAGE_SHIFT; + + switch (idx) { + case FIX_BTMAP_END ... FIX_BTMAP_BEGIN: +#ifdef CONFIG_X86_F00F_BUG + case FIX_F00F_IDT: +#endif + case FIX_WP_TEST: + case FIX_VDSO: +#ifdef CONFIG_X86_LOCAL_APIC + case FIX_APIC_BASE: /* maps dummy local APIC */ +#endif + pte = pfn_pte(phys, prot); + break; + + default: + pte = mfn_pte(phys, prot); + break; + } + + __native_set_fixmap(idx, pte); +} + static const struct pv_info xen_info __initdata = { .paravirt_enabled = 1, .shared_kernel_pmd = 0, @@ -1112,6 +1139,8 @@ static const struct pv_mmu_ops xen_mmu_ops __initdata = { .enter = paravirt_enter_lazy_mmu, .leave = xen_leave_lazy, }, + + .set_fixmap = xen_set_fixmap, }; #ifdef CONFIG_SMP -- cgit v1.2.3 From 08b882c627aeeeb3cfd3c4354f0d360d7949549d Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 16 Jun 2008 04:30:01 -0700 Subject: paravirt: add hooks for ptep_modify_prot_start/commit This patch adds paravirt-ops hooks in pv_mmu_ops for ptep_modify_prot_start and ptep_modify_prot_commit. This allows the hypervisor-specific backends to implement these in some more efficient way. Signed-off-by: Jeremy Fitzhardinge Acked-by: Linus Torvalds Acked-by: Hugh Dickins Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 73fb0c4c150..0b7553cbc52 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1138,6 +1138,9 @@ static const struct pv_mmu_ops xen_mmu_ops __initdata = { .set_pte_at = xen_set_pte_at, .set_pmd = xen_set_pmd_hyper, + .ptep_modify_prot_start = __ptep_modify_prot_start, + .ptep_modify_prot_commit = __ptep_modify_prot_commit, + .pte_val = xen_pte_val, .pte_flags = native_pte_val, .pgd_val = xen_pgd_val, -- 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 ++++ 3 files changed, 35 insertions(+), 3 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 0b7553cbc52..bd74229081c 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 8132aa8c5d4..846dad7d54a 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 e3dd09e25c6..297bf9f5b8b 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 */ -- 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') diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 846dad7d54a..f6b8225c2a0 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 5791eb2e375..3c63c4da7ed 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 8bae996d99a..85893824161 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 d0be6bdea103b8d04c8a3495538b7c0011ae4129 Mon Sep 17 00:00:00 2001 From: Yinghai Lu Date: Sun, 15 Jun 2008 18:58:51 -0700 Subject: x86: rename two e820 related functions rename update_memory_range to e820_update_range rename add_memory_region to e820_add_region to make it more clear that they are about e820 map operations. Signed-off-by: Yinghai Lu Signed-off-by: Ingo Molnar --- arch/x86/xen/setup.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/setup.c b/arch/x86/xen/setup.c index 82517e4a752..9001c9df04d 100644 --- a/arch/x86/xen/setup.c +++ b/arch/x86/xen/setup.c @@ -39,8 +39,8 @@ char * __init xen_memory_setup(void) unsigned long max_pfn = xen_start_info->nr_pages; e820.nr_map = 0; - add_memory_region(0, LOWMEMSIZE(), E820_RAM); - add_memory_region(HIGH_MEMORY, PFN_PHYS(max_pfn)-HIGH_MEMORY, E820_RAM); + e820_add_region(0, LOWMEMSIZE(), E820_RAM); + e820_add_region(HIGH_MEMORY, PFN_PHYS(max_pfn)-HIGH_MEMORY, E820_RAM); return "Xen"; } -- cgit v1.2.3 From be5bf9fa1c327fa6fd6e7ba44665437dd558dfe3 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 16 Jun 2008 14:54:46 -0700 Subject: xen: reserve Xen-specific memory in e820 map Signed-off-by: Jeremy Fitzhardinge Cc: Yinghai Lu Cc: the arch/x86 maintainers Signed-off-by: Ingo Molnar Cc: Yinghai Lu Signed-off-by: Ingo Molnar --- arch/x86/xen/setup.c | 13 +++++++++++++ 1 file changed, 13 insertions(+) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/setup.c b/arch/x86/xen/setup.c index a2957580320..dc2ca8ad360 100644 --- a/arch/x86/xen/setup.c +++ b/arch/x86/xen/setup.c @@ -40,9 +40,22 @@ char * __init xen_memory_setup(void) max_pfn = min(MAX_DOMAIN_PAGES, max_pfn); e820.nr_map = 0; + e820_add_region(0, LOWMEMSIZE(), E820_RAM); e820_add_region(HIGH_MEMORY, PFN_PHYS(max_pfn)-HIGH_MEMORY, E820_RAM); + /* + * Reserve Xen bits: + * - mfn_list + * - xen_start_info + * See comment above "struct start_info" in + */ + e820_add_region(__pa(xen_start_info->mfn_list), + xen_start_info->pt_base - xen_start_info->mfn_list, + E820_RESERVED); + + sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map); + return "Xen"; } -- cgit v1.2.3 From b792c755907cffceab393585b626ef2553c38538 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 16 Jun 2008 14:54:49 -0700 Subject: xen: reserve ISA space in e820 map [ TODO: release the underlying memory back to Xen. ] Signed-off-by: Jeremy Fitzhardinge Cc: Yinghai Lu Cc: the arch/x86 maintainers Signed-off-by: Ingo Molnar Cc: Yinghai Lu Cc: Ian Campbell Signed-off-by: Ingo Molnar --- arch/x86/xen/setup.c | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/setup.c b/arch/x86/xen/setup.c index dc2ca8ad360..e0a39595bde 100644 --- a/arch/x86/xen/setup.c +++ b/arch/x86/xen/setup.c @@ -13,6 +13,7 @@ #include #include #include +#include #include #include @@ -41,8 +42,15 @@ char * __init xen_memory_setup(void) e820.nr_map = 0; - e820_add_region(0, LOWMEMSIZE(), E820_RAM); - e820_add_region(HIGH_MEMORY, PFN_PHYS(max_pfn)-HIGH_MEMORY, E820_RAM); + e820_add_region(0, PFN_PHYS(max_pfn), E820_RAM); + + /* + * Even though this is normal, usable memory under Xen, reserve + * ISA memory anyway because too many things think they can poke + * about in there. + */ + e820_add_region(ISA_START_ADDRESS, ISA_END_ADDRESS - ISA_START_ADDRESS, + E820_RESERVED); /* * Reserve Xen bits: -- cgit v1.2.3 From 88a6846c70ad6bf33a545d554ace801d69e8a1a5 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 16 Jun 2008 14:54:51 -0700 Subject: xen: set max_pfn_mapped Signed-off-by: Jeremy Fitzhardinge Cc: Yinghai Lu Cc: the arch/x86 maintainers Signed-off-by: Ingo Molnar Cc: Yinghai Lu Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 1 + 1 file changed, 1 insertion(+) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 316623cf411..76ad1efaf09 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1304,6 +1304,7 @@ asmlinkage void __init xen_start_kernel(void) init_pg_tables_start = __pa(pgd); init_pg_tables_end = __pa(pgd) + xen_start_info->nr_pt_frames*PAGE_SIZE; + max_pfn_mapped = (init_pg_tables_end + 512*1024) >> PAGE_SHIFT; init_mm.pgd = pgd; /* use the Xen pagetables to start */ -- cgit v1.2.3 From eba0045ff87bab465d3c80c289f3bf709c1800f5 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Wed, 25 Jun 2008 00:19:12 -0400 Subject: x86/paravirt: add a pgd_alloc/free hooks MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Add hooks which are called at pgd_alloc/free time. The pgd_alloc hook may return an error code, which if non-zero, causes the pgd allocation to be failed. The hooks may be used to allocate/free auxillary per-pgd information. also fix: > * Ingo Molnar wrote: > > include/asm/pgalloc.h: In function ‘paravirt_pgd_free': > include/asm/pgalloc.h:14: error: parameter name omitted > arch/x86/kernel/entry_64.S: In file included from > arch/x86/kernel/traps_64.c:51:include/asm/pgalloc.h: In function ‘paravirt_pgd_free': > include/asm/pgalloc.h:14: error: parameter name omitted Signed-off-by: Jeremy Fitzhardinge Cc: xen-devel Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 76ad1efaf09..d62f14e2070 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -45,6 +45,7 @@ #include #include #include +#include #include "xen-ops.h" #include "mmu.h" @@ -1153,6 +1154,9 @@ static const struct pv_mmu_ops xen_mmu_ops __initdata = { .pte_update = paravirt_nop, .pte_update_defer = paravirt_nop, + .pgd_alloc = __paravirt_pgd_alloc, + .pgd_free = paravirt_nop, + .alloc_pte = xen_alloc_pte_init, .release_pte = xen_release_pte_init, .alloc_pmd = xen_alloc_pte_init, -- cgit v1.2.3 From d75cd22fdd5f7d203fb60014d426942df33dd9a6 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Wed, 25 Jun 2008 00:19:26 -0400 Subject: x86/paravirt: split sysret and sysexit Don't conflate sysret and sysexit; they're different instructions with different semantics, and may be in use at the same time (at least within the same kernel, depending on whether its an Intel or AMD system). sysexit - just return to userspace, does no register restoration of any kind; must explicitly atomically enable interrupts. sysret - reloads flags from r11, so no need to explicitly enable interrupts on 64-bit, responsible for restoring usermode %gs Signed-off-by: Jeremy Fitzhardinge Cc: xen-devel Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index d62f14e2070..119c88fa769 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1089,7 +1089,7 @@ static const struct pv_cpu_ops xen_cpu_ops __initdata = { .read_pmc = native_read_pmc, .iret = xen_iret, - .irq_enable_syscall_ret = xen_sysexit, + .irq_enable_sysexit = xen_sysexit, .load_tr_desc = paravirt_nop, .set_ldt = xen_set_ldt, -- cgit v1.2.3 From fab58420ac0007a452b540cfb07923225ea4f48d Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Wed, 25 Jun 2008 00:19:31 -0400 Subject: x86/paravirt, 64-bit: add adjust_exception_frame 64-bit Xen pushes a couple of extra words onto an exception frame. Add a hook to deal with them. Signed-off-by: Jeremy Fitzhardinge Cc: xen-devel Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 119c88fa769..3b980831602 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1123,6 +1123,9 @@ static const struct pv_irq_ops xen_irq_ops __initdata = { .irq_enable = xen_irq_enable, .safe_halt = xen_safe_halt, .halt = xen_halt, +#ifdef CONFIG_X86_64 + .adjust_exception_frame = paravirt_nop, +#endif }; static const struct pv_apic_ops xen_apic_ops __initdata = { -- cgit v1.2.3 From e93ef949fd9a3f237aedfb8e64414b28980530b8 Mon Sep 17 00:00:00 2001 From: Alok Kataria Date: Tue, 1 Jul 2008 11:43:36 -0700 Subject: x86: rename paravirtualized TSC functions Rename the paravirtualized calculate_cpu_khz to calibrate_tsc. In all cases, we actually calibrate_tsc and use that as the cpu_khz value. Signed-off-by: Alok N Kataria Signed-off-by: Dan Hecht Cc: Dan Hecht Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 2 +- arch/x86/xen/time.c | 4 ++-- arch/x86/xen/xen-ops.h | 2 +- 3 files changed, 4 insertions(+), 4 deletions(-) (limited to 'arch/x86/xen') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 3b980831602..dcd4e51f2f1 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1062,7 +1062,7 @@ static const struct pv_time_ops xen_time_ops __initdata = { .set_wallclock = xen_set_wallclock, .get_wallclock = xen_get_wallclock, - .get_cpu_khz = xen_cpu_khz, + .get_tsc_khz = xen_tsc_khz, .sched_clock = xen_sched_clock, }; diff --git a/arch/x86/xen/time.c b/arch/x86/xen/time.c index 64f0038b955..685b77470fc 100644 --- a/arch/x86/xen/time.c +++ b/arch/x86/xen/time.c @@ -197,8 +197,8 @@ unsigned long long xen_sched_clock(void) } -/* Get the CPU speed from Xen */ -unsigned long xen_cpu_khz(void) +/* Get the TSC speed from Xen */ +unsigned long xen_tsc_khz(void) { u64 xen_khz = 1000000ULL << 32; const struct pvclock_vcpu_time_info *info = diff --git a/arch/x86/xen/xen-ops.h b/arch/x86/xen/xen-ops.h index 9a055592a30..d852ddbb344 100644 --- a/arch/x86/xen/xen-ops.h +++ b/arch/x86/xen/xen-ops.h @@ -32,7 +32,7 @@ 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); +unsigned long xen_tsc_khz(void); void __init xen_time_init(void); unsigned long xen_get_wallclock(void); int xen_set_wallclock(unsigned long time); -- cgit v1.2.3