From 5cd7376200be7b8bab085557ff5876b04bd84191 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Thu, 15 Jan 2009 15:46:08 +0100 Subject: fix: crash: IP: __bitmap_intersects+0x48/0x73 -tip testing found this crash: > [ 35.258515] calling acpi_cpufreq_init+0x0/0x127 @ 1 > [ 35.264127] BUG: unable to handle kernel NULL pointer dereference at (null) > [ 35.267554] IP: [] __bitmap_intersects+0x48/0x73 > [ 35.267554] PGD 0 > [ 35.267554] Oops: 0000 [#1] SMP DEBUG_PAGEALLOC arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c is still broken: there's no allocation of the variable mask, so we pass in an uninitialized cmd.mask field to drv_read(), which then passes it to the scheduler which then crashes ... Switch it over to the much simpler constant-cpumask-pointers approach. Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c | 13 ++++--------- 1 file changed, 4 insertions(+), 9 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c b/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c index 6f11e029e8c..019276717a7 100644 --- a/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c +++ b/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c @@ -145,7 +145,7 @@ typedef union { struct drv_cmd { unsigned int type; - cpumask_var_t mask; + const struct cpumask *mask; drv_addr_union addr; u32 val; }; @@ -235,8 +235,7 @@ static u32 get_cur_val(const struct cpumask *mask) return 0; } - cpumask_copy(cmd.mask, mask); - + cmd.mask = mask; drv_read(&cmd); dprintk("get_cur_val = %u\n", cmd.val); @@ -403,9 +402,6 @@ static int acpi_cpufreq_target(struct cpufreq_policy *policy, return -ENODEV; } - if (unlikely(!alloc_cpumask_var(&cmd.mask, GFP_KERNEL))) - return -ENOMEM; - perf = data->acpi_data; result = cpufreq_frequency_table_target(policy, data->freq_table, @@ -450,9 +446,9 @@ static int acpi_cpufreq_target(struct cpufreq_policy *policy, /* cpufreq holds the hotplug lock, so we are safe from here on */ if (policy->shared_type != CPUFREQ_SHARED_TYPE_ANY) - cpumask_and(cmd.mask, cpu_online_mask, policy->cpus); + cmd.mask = policy->cpus; else - cpumask_copy(cmd.mask, cpumask_of(policy->cpu)); + cmd.mask = cpumask_of(policy->cpu); freqs.old = perf->states[perf->state].core_frequency * 1000; freqs.new = data->freq_table[next_state].frequency; @@ -479,7 +475,6 @@ static int acpi_cpufreq_target(struct cpufreq_policy *policy, perf->state = next_perf_state; out: - free_cpumask_var(cmd.mask); return result; } -- cgit v1.2.3-70-g09d2 From f2a082711905312dc7b6675e913fee0c4689f7ae Mon Sep 17 00:00:00 2001 From: Mike Travis Date: Thu, 15 Jan 2009 09:19:32 -0800 Subject: x86: fix build warning when CONFIG_NUMA not defined. Impact: fix build warning The macro cpu_to_node did not reference it's argument, and instead simply returned a 0. This causes a "unused variable" warning if it's the only reference in a function (show_cache_disable). Replace it with the more correct inline function. Signed-off-by: Mike Travis --- arch/x86/include/asm/topology.h | 17 ++++++++++++++--- 1 file changed, 14 insertions(+), 3 deletions(-) (limited to 'arch') diff --git a/arch/x86/include/asm/topology.h b/arch/x86/include/asm/topology.h index 4e2f2e0aab2..d0c68e29163 100644 --- a/arch/x86/include/asm/topology.h +++ b/arch/x86/include/asm/topology.h @@ -192,9 +192,20 @@ extern int __node_distance(int, int); #else /* !CONFIG_NUMA */ -#define numa_node_id() 0 -#define cpu_to_node(cpu) 0 -#define early_cpu_to_node(cpu) 0 +static inline int numa_node_id(void) +{ + return 0; +} + +static inline int cpu_to_node(int cpu) +{ + return 0; +} + +static inline int early_cpu_to_node(int cpu) +{ + return 0; +} static inline const cpumask_t *cpumask_of_node(int node) { -- cgit v1.2.3-70-g09d2 From c99dbbe9f8f6b3e9383e64710217e873431d1c31 Mon Sep 17 00:00:00 2001 From: Mike Travis Date: Thu, 15 Jan 2009 12:09:44 -0800 Subject: sched: fix warning on ia64 Andrew Morton reported this warning on ia64: kernel/sched.c: In function `sd_init_NODE': kernel/sched.c:7449: warning: comparison of distinct pointer types lacks a cast Using the untyped min() function produces such warnings. Fix: type the constant 32 as unsigned int to match typeof(num_online_cpus). Reported-by: Andrew Morton Signed-off-by: Mike Travis Signed-off-by: Ingo Molnar --- arch/ia64/include/asm/topology.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/ia64/include/asm/topology.h b/arch/ia64/include/asm/topology.h index 32f3af1641c..3193f4417e1 100644 --- a/arch/ia64/include/asm/topology.h +++ b/arch/ia64/include/asm/topology.h @@ -84,7 +84,7 @@ void build_cpu_to_node_map(void); .child = NULL, \ .groups = NULL, \ .min_interval = 8, \ - .max_interval = 8*(min(num_online_cpus(), 32)), \ + .max_interval = 8*(min(num_online_cpus(), 32U)), \ .busy_factor = 64, \ .imbalance_pct = 125, \ .cache_nice_tries = 2, \ -- cgit v1.2.3-70-g09d2 From 6eb714c63ed5bd663627f7dda8c4d5258f3b64ef Mon Sep 17 00:00:00 2001 From: Mike Travis Date: Fri, 16 Jan 2009 15:31:15 -0800 Subject: cpufreq: use work_on_cpu in acpi-cpufreq.c for drv_read and drv_write Impact: use new work_on_cpu function to reduce stack usage Replace the saving of current->cpus_allowed and set_cpus_allowed_ptr() with a work_on_cpu function for drv_read() and drv_write(). Basically converts do_drv_{read,write} into "work_on_cpu" functions that are now called by drv_read and drv_write. Note: This patch basically reverts 50c668d6 which reverted 7503bfba, now that the work_on_cpu() function is more stable. Signed-off-by: Mike Travis Acked-by: Rusty Russell Tested-by: Dieter Ries Tested-by: Maciej Rutecki Cc: Dave Jones Cc: --- arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c | 22 +++++++++------------- 1 file changed, 9 insertions(+), 13 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c b/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c index 019276717a7..4b1c319d30c 100644 --- a/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c +++ b/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c @@ -150,8 +150,9 @@ struct drv_cmd { u32 val; }; -static void do_drv_read(struct drv_cmd *cmd) +static long do_drv_read(void *_cmd) { + struct drv_cmd *cmd = _cmd; u32 h; switch (cmd->type) { @@ -166,10 +167,12 @@ static void do_drv_read(struct drv_cmd *cmd) default: break; } + return 0; } -static void do_drv_write(struct drv_cmd *cmd) +static long do_drv_write(void *_cmd) { + struct drv_cmd *cmd = _cmd; u32 lo, hi; switch (cmd->type) { @@ -186,30 +189,23 @@ static void do_drv_write(struct drv_cmd *cmd) default: break; } + return 0; } static void drv_read(struct drv_cmd *cmd) { - cpumask_t saved_mask = current->cpus_allowed; cmd->val = 0; - set_cpus_allowed_ptr(current, cmd->mask); - do_drv_read(cmd); - set_cpus_allowed_ptr(current, &saved_mask); + work_on_cpu(cpumask_any(cmd->mask), do_drv_read, cmd); } static void drv_write(struct drv_cmd *cmd) { - cpumask_t saved_mask = current->cpus_allowed; unsigned int i; for_each_cpu(i, cmd->mask) { - set_cpus_allowed_ptr(current, cpumask_of(i)); - do_drv_write(cmd); + work_on_cpu(i, do_drv_write, cmd); } - - set_cpus_allowed_ptr(current, &saved_mask); - return; } static u32 get_cur_val(const struct cpumask *mask) @@ -367,7 +363,7 @@ static unsigned int get_cur_freq_on_cpu(unsigned int cpu) return freq; } -static unsigned int check_freqs(const cpumask_t *mask, unsigned int freq, +static unsigned int check_freqs(const struct cpumask *mask, unsigned int freq, struct acpi_cpufreq_data *data) { unsigned int cur_freq; -- cgit v1.2.3-70-g09d2 From cef30b3a84e1c7cbd49987d83c5863c001445842 Mon Sep 17 00:00:00 2001 From: Mike Travis Date: Fri, 16 Jan 2009 15:58:13 -0800 Subject: x86: put trigger in to detect mismatched apic versions. Fire off one message if two apic's discovered with different apic versions. Signed-off-by: Mike Travis --- arch/x86/kernel/apic.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'arch') diff --git a/arch/x86/kernel/apic.c b/arch/x86/kernel/apic.c index 0f830e4f567..db0998641c5 100644 --- a/arch/x86/kernel/apic.c +++ b/arch/x86/kernel/apic.c @@ -1833,6 +1833,11 @@ void __cpuinit generic_processor_info(int apicid, int version) num_processors++; cpu = cpumask_next_zero(-1, cpu_present_mask); + if (version != apic_version[boot_cpu_physical_apicid]) + WARN_ONCE(1, + "ACPI: apic version mismatch, bootcpu: %x cpu %d: %x\n", + apic_version[boot_cpu_physical_apicid], cpu, version); + physid_set(apicid, phys_cpu_present_map); if (apicid == boot_cpu_physical_apicid) { /* -- cgit v1.2.3-70-g09d2 From 5766b842b23c6b40935a5f3bd435b2bcdaff2143 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Tue, 20 Jan 2009 09:13:15 +0100 Subject: x86, cpumask: fix tlb flush race Impact: fix bootup crash The cpumask is now passed in as a reference to mm->cpu_vm_mask, not on the stack - hence it is not constant anymore during the TLB flush. That way it could race and some static sanity checks would trigger: [ 238.154287] ------------[ cut here ]------------ [ 238.156039] kernel BUG at arch/x86/kernel/tlb_32.c:130! [ 238.156039] invalid opcode: 0000 [#1] SMP [ 238.156039] last sysfs file: /sys/class/net/eth2/address [ 238.156039] Modules linked in: [ 238.156039] [ 238.156039] Pid: 6493, comm: ifup-eth Not tainted (2.6.29-rc2-tip #1) P4DC6 [ 238.156039] EIP: 0060:[] EFLAGS: 00010202 CPU: 2 [ 238.156039] EIP is at native_flush_tlb_others+0x35/0x158 [ 238.156039] EAX: c0ef972c EBX: f6143301 ECX: 00000000 EDX: 00000000 [ 238.156039] ESI: f61433a8 EDI: f6143200 EBP: f34f3e00 ESP: f34f3df0 [ 238.156039] DS: 007b ES: 007b FS: 00d8 GS: 0000 SS: 0068 [ 238.156039] Process ifup-eth (pid: 6493, ti=f34f2000 task=f399ab00 task.ti=f34f2000) [ 238.156039] Stack: [ 238.156039] ffffffff f61433a8 ffffffff f6143200 f34f3e18 c0118e9c 00000000 f6143200 [ 238.156039] f61433a8 f5bec738 f34f3e28 c0119435 c2b5b830 f6143200 f34f3e34 c01c2dc3 [ 238.156039] bffd9000 f34f3e60 c01c3051 00000000 ffffffff f34f3e4c 00000000 00000071 [ 238.156039] Call Trace: [ 238.156039] [] ? flush_tlb_others+0x52/0x5b [ 238.156039] [] ? flush_tlb_mm+0x7f/0x8b [ 238.156039] [] ? tlb_finish_mmu+0x2d/0x55 [ 238.156039] [] ? exit_mmap+0x124/0x170 [ 238.156039] [] ? mmput+0x40/0xf5 [ 238.156039] [] ? flush_old_exec+0x640/0x94b [ 238.156039] [] ? fsnotify_access+0x37/0x39 [ 238.156039] [] ? kernel_read+0x39/0x4b [ 238.156039] [] ? load_elf_binary+0x4a1/0x11bb [ 238.156039] [] ? might_fault+0x51/0x9c [ 238.156039] [] ? paravirt_read_tsc+0x20/0x4f [ 238.156039] [] ? native_sched_clock+0x5d/0x60 [ 238.156039] [] ? search_binary_handler+0xab/0x2c4 [ 238.156039] [] ? load_elf_binary+0x0/0x11bb [ 238.156039] [] ? _raw_read_unlock+0x21/0x46 [ 238.156039] [] ? load_elf_binary+0x0/0x11bb [ 238.156039] [] ? search_binary_handler+0xb2/0x2c4 [ 238.156039] [] ? do_execve+0x21c/0x2ee [ 238.156039] [] ? sys_execve+0x51/0x8c [ 238.156039] [] ? sysenter_do_call+0x12/0x43 Fix it by not assuming that the cpumask is constant. Signed-off-by: Ingo Molnar --- arch/x86/kernel/tlb_32.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/tlb_32.c b/arch/x86/kernel/tlb_32.c index ec53818f4e3..d37bbfcb813 100644 --- a/arch/x86/kernel/tlb_32.c +++ b/arch/x86/kernel/tlb_32.c @@ -125,9 +125,8 @@ void native_flush_tlb_others(const struct cpumask *cpumask, struct mm_struct *mm, unsigned long va) { /* - * - mask must exist :) + * mm must exist :) */ - BUG_ON(cpumask_empty(cpumask)); BUG_ON(!mm); /* @@ -138,14 +137,18 @@ void native_flush_tlb_others(const struct cpumask *cpumask, spin_lock(&tlbstate_lock); cpumask_andnot(flush_cpumask, cpumask, cpumask_of(smp_processor_id())); -#ifdef CONFIG_HOTPLUG_CPU - /* If a CPU which we ran on has gone down, OK. */ cpumask_and(flush_cpumask, flush_cpumask, cpu_online_mask); + + /* + * If a task whose mm mask we are looking at has descheduled and + * has cleared its presence from the mask, or if a CPU which we ran + * on has gone down then there might be no flush work left: + */ if (unlikely(cpumask_empty(flush_cpumask))) { spin_unlock(&tlbstate_lock); return; } -#endif + flush_mm = mm; flush_va = va; -- cgit v1.2.3-70-g09d2