From 49f4138346b3cec2706adff02658fe27ceb1e46f Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Fri, 26 Nov 2010 13:42:47 +0100 Subject: printk: Fix wake_up_klogd() vs cpu hotplug wake_up_klogd() may get called from preemptible context but uses __raw_get_cpu_var() to write to a per cpu variable. If it gets preempted between getting the address and writing to it, the cpu in question could be offline if the process gets scheduled back and hence writes to the per cpu data of an offline cpu. This buggy behaviour was introduced with fa33507a "printk: robustify printk, fix #2" which was supposed to fix a "using smp_processor_id() in preemptible" warning. Let's use this_cpu_write() instead which disables preemption and makes sure that the outlined scenario cannot happen. Signed-off-by: Heiko Carstens Acked-by: Eric Dumazet Signed-off-by: Peter Zijlstra LKML-Reference: <20101126124247.GC7023@osiris.boeblingen.de.ibm.com> Signed-off-by: Ingo Molnar --- kernel/printk.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/kernel/printk.c b/kernel/printk.c index 9a2264fc42c..cf7588e93f6 100644 --- a/kernel/printk.c +++ b/kernel/printk.c @@ -1088,7 +1088,7 @@ int printk_needs_cpu(int cpu) void wake_up_klogd(void) { if (waitqueue_active(&log_wait)) - __raw_get_cpu_var(printk_pending) = 1; + this_cpu_write(printk_pending, 1); } /** -- cgit v1.2.3 From 61ab25447ad6334a74e32f60efb135a3467223f8 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Fri, 26 Nov 2010 13:00:59 +0100 Subject: nohz: Fix printk_needs_cpu() return value on offline cpus This patch fixes a hang observed with 2.6.32 kernels where timers got enqueued on offline cpus. printk_needs_cpu() may return 1 if called on offline cpus. When a cpu gets offlined it schedules the idle process which, before killing its own cpu, will call tick_nohz_stop_sched_tick(). That function in turn will call printk_needs_cpu() in order to check if the local tick can be disabled. On offline cpus this function should naturally return 0 since regardless if the tick gets disabled or not the cpu will be dead short after. That is besides the fact that __cpu_disable() should already have made sure that no interrupts on the offlined cpu will be delivered anyway. In this case it prevents tick_nohz_stop_sched_tick() to call select_nohz_load_balancer(). No idea if that really is a problem. However what made me debug this is that on 2.6.32 the function get_nohz_load_balancer() is used within __mod_timer() to select a cpu on which a timer gets enqueued. If printk_needs_cpu() returns 1 then the nohz_load_balancer cpu doesn't get updated when a cpu gets offlined. It may contain the cpu number of an offline cpu. In turn timers get enqueued on an offline cpu and not very surprisingly they never expire and cause system hangs. This has been observed 2.6.32 kernels. On current kernels __mod_timer() uses get_nohz_timer_target() which doesn't have that problem. However there might be other problems because of the too early exit tick_nohz_stop_sched_tick() in case a cpu goes offline. Easiest way to fix this is just to test if the current cpu is offline and call printk_tick() directly which clears the condition. Alternatively I tried a cpu hotplug notifier which would clear the condition, however between calling the notifier function and printk_needs_cpu() something could have called printk() again and the problem is back again. This seems to be the safest fix. Signed-off-by: Heiko Carstens Signed-off-by: Peter Zijlstra Cc: stable@kernel.org LKML-Reference: <20101126120235.406766476@de.ibm.com> Signed-off-by: Ingo Molnar --- kernel/printk.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/kernel/printk.c b/kernel/printk.c index cf7588e93f6..a23315dc449 100644 --- a/kernel/printk.c +++ b/kernel/printk.c @@ -1082,6 +1082,8 @@ void printk_tick(void) int printk_needs_cpu(int cpu) { + if (unlikely(cpu_is_offline(cpu))) + printk_tick(); return per_cpu(printk_pending, cpu); } -- cgit v1.2.3 From e7a3481c0246c8e45e79c629efd63b168e91fcda Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 25 Oct 2010 16:53:46 -0700 Subject: x86/pvclock: Zero last_value on resume If the guest domain has been suspend/resumed or migrated, then the system clock backing the pvclock clocksource may revert to a smaller value (ie, can be non-monotonic across the migration/save-restore). Make sure we zero last_value in that case so that the domain continues to see clock updates. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Ingo Molnar --- arch/x86/include/asm/pvclock.h | 1 + arch/x86/kernel/pvclock.c | 5 +++++ arch/x86/xen/time.c | 2 ++ 3 files changed, 8 insertions(+) diff --git a/arch/x86/include/asm/pvclock.h b/arch/x86/include/asm/pvclock.h index 7f7e577a0e3..31d84acc151 100644 --- a/arch/x86/include/asm/pvclock.h +++ b/arch/x86/include/asm/pvclock.h @@ -11,6 +11,7 @@ unsigned long pvclock_tsc_khz(struct pvclock_vcpu_time_info *src); void pvclock_read_wallclock(struct pvclock_wall_clock *wall, struct pvclock_vcpu_time_info *vcpu, struct timespec *ts); +void pvclock_resume(void); /* * Scale a 64-bit delta by scaling and multiplying by a 32-bit fraction, diff --git a/arch/x86/kernel/pvclock.c b/arch/x86/kernel/pvclock.c index 008b91eefa1..42eb3300dfc 100644 --- a/arch/x86/kernel/pvclock.c +++ b/arch/x86/kernel/pvclock.c @@ -83,6 +83,11 @@ unsigned long pvclock_tsc_khz(struct pvclock_vcpu_time_info *src) static atomic64_t last_value = ATOMIC64_INIT(0); +void pvclock_resume(void) +{ + atomic64_set(&last_value, 0); +} + cycle_t pvclock_clocksource_read(struct pvclock_vcpu_time_info *src) { struct pvclock_shadow_time shadow; diff --git a/arch/x86/xen/time.c b/arch/x86/xen/time.c index b2bb5aa3b05..5da5e53fb94 100644 --- a/arch/x86/xen/time.c +++ b/arch/x86/xen/time.c @@ -426,6 +426,8 @@ void xen_timer_resume(void) { int cpu; + pvclock_resume(); + if (xen_clockevent != &xen_vcpuop_clockevent) return; -- cgit v1.2.3 From 8a9533123f43f2cdb3eb601c17ff2ad336882eff Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 29 Nov 2010 12:44:15 -0200 Subject: perf symbols: Fix kallsyms kernel/module map splitting On ARM, module addresss space is ahead of kernel space, so the module symbols are handled before kernel symbol in dso__split_kallsyms, then was causing one map to be created for each kernel symbol. Reported-by: Ming Lei Tested-by: Ming Lei Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Mike Galbraith Cc: Ming Lei Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Tom Zanussi LKML-Reference: <20101124144540.GB15875@ghostprotocols.net> Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/symbol.c | 16 +++++++++++----- 1 file changed, 11 insertions(+), 5 deletions(-) diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 0500895a45a..2af4d7d2237 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -532,7 +532,7 @@ static int dso__split_kallsyms(struct dso *self, struct map *map, struct machine *machine = kmaps->machine; struct map *curr_map = map; struct symbol *pos; - int count = 0; + int count = 0, moved = 0; struct rb_root *root = &self->symbols[map->type]; struct rb_node *next = rb_first(root); int kernel_range = 0; @@ -590,6 +590,11 @@ static int dso__split_kallsyms(struct dso *self, struct map *map, char dso_name[PATH_MAX]; struct dso *dso; + if (count == 0) { + curr_map = map; + goto filter_symbol; + } + if (self->kernel == DSO_TYPE_GUEST_KERNEL) snprintf(dso_name, sizeof(dso_name), "[guest.kernel].%d", @@ -615,7 +620,7 @@ static int dso__split_kallsyms(struct dso *self, struct map *map, map_groups__insert(kmaps, curr_map); ++kernel_range; } - +filter_symbol: if (filter && filter(curr_map, pos)) { discard_symbol: rb_erase(&pos->rb_node, root); symbol__delete(pos); @@ -623,8 +628,9 @@ discard_symbol: rb_erase(&pos->rb_node, root); if (curr_map != map) { rb_erase(&pos->rb_node, root); symbols__insert(&curr_map->dso->symbols[curr_map->type], pos); - } - count++; + ++moved; + } else + ++count; } } @@ -634,7 +640,7 @@ discard_symbol: rb_erase(&pos->rb_node, root); dso__set_loaded(curr_map->dso, curr_map->type); } - return count; + return count + moved; } int dso__load_kallsyms(struct dso *self, const char *filename, -- cgit v1.2.3 From d214afbd81405d4da2c5745fe867e6b313fd4178 Mon Sep 17 00:00:00 2001 From: Ming Lei Date: Thu, 25 Nov 2010 19:27:25 +0800 Subject: perf symbols: Figure out start address of kernel map from kallsyms On ARM, module symbol start address is ahead of kernel symbol start address, so we can't suppose that the start address of kernel map always is zero, otherwise may cause incorrect .start and .end of kernel map (caused by fixup) when there are modules loaded, then map_groups__find may return incorrect map for symbol query. This patch always figures out the start address of kernel map from /proc/kallsyms if the file is available, so fix the issues on ARM for module loaded case. This patch fixes the following issues on ARM when modules are loaded: - vmlinux symbol can't be found by kallsyms maps doing 'perf test' - module symbols are parsed mistakenlly when doing 'perf top'/'perf report' Cc: Ian Munsie Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: Tom Zanussi LKML-Reference: <20101125192725.62d31b42@tom-lei> Signed-off-by: Ming Lei Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/symbol.c | 43 ++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 42 insertions(+), 1 deletion(-) diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 2af4d7d2237..d628c8d1cf5 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -2131,14 +2131,55 @@ static struct dso *machine__create_kernel(struct machine *self) return kernel; } +struct process_args { + u64 start; +}; + +static int symbol__in_kernel(void *arg, const char *name, + char type __used, u64 start) +{ + struct process_args *args = arg; + + if (strchr(name, '[')) + return 0; + + args->start = start; + return 1; +} + +/* Figure out the start address of kernel map from /proc/kallsyms */ +static u64 machine__get_kernel_start_addr(struct machine *machine) +{ + const char *filename; + char path[PATH_MAX]; + struct process_args args; + + if (machine__is_host(machine)) { + filename = "/proc/kallsyms"; + } else { + if (machine__is_default_guest(machine)) + filename = (char *)symbol_conf.default_guest_kallsyms; + else { + sprintf(path, "%s/proc/kallsyms", machine->root_dir); + filename = path; + } + } + + if (kallsyms__parse(filename, &args, symbol__in_kernel) <= 0) + return 0; + + return args.start; +} + int __machine__create_kernel_maps(struct machine *self, struct dso *kernel) { enum map_type type; + u64 start = machine__get_kernel_start_addr(self); for (type = 0; type < MAP__NR_TYPES; ++type) { struct kmap *kmap; - self->vmlinux_maps[type] = map__new2(0, kernel, type); + self->vmlinux_maps[type] = map__new2(start, kernel, type); if (self->vmlinux_maps[type] == NULL) return -1; -- cgit v1.2.3 From 60e677373be9c0bf7c9a22937601d5a40e51c042 Mon Sep 17 00:00:00 2001 From: Franck Bui-Huu Date: Tue, 30 Nov 2010 12:50:14 +0100 Subject: perf header: Don't assume there's no attr info if no sample ids is provided This primarily fixes perf-report, which didn't report the correct type of event if perf-record was called to record one event different from 'cycles': $ perf record -e instructions true [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.007 MB perf.data (~295 samples) ] $ perf report | head -n1 # Events: 7 cycles LPU-Reference: Signed-off-by: Franck Bui-Huu --- tools/perf/util/header.c | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-) diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index d7e67b167ea..64a85bafde6 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -946,11 +946,16 @@ perf_header__find_attr(u64 id, struct perf_header *header) /* * We set id to -1 if the data file doesn't contain sample - * ids. Check for this and avoid walking through the entire - * list of ids which may be large. + * ids. This can happen when the data file contains one type + * of event and in that case, the header can still store the + * event attribute information. Check for this and avoid + * walking through the entire list of ids which may be large. */ - if (id == -1ULL) + if (id == -1ULL) { + if (header->attrs > 0) + return &header->attr[0]->attr; return NULL; + } for (i = 0; i < header->attrs; i++) { struct perf_header_attr *attr = header->attr[i]; -- cgit v1.2.3 From 18483b81ee7e70ee68d4b18be618be5cfcc0b290 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 6 Dec 2010 15:13:38 -0200 Subject: perf record: Fix eternal wait for stillborn child When execvp fails to find the specified command on the path we won't get SIGCHLD, so send a SIGUSR1 and exit right away. Current situation would require a SIGINT performed by the user and would produce meaningless summary. Now: [acme@emilia linux]$ ./foo -bash: ./foo: No such file or directory [acme@emilia linux]$ perf record ./foo ./foo: No such file or directory [acme@emilia linux]$ Acked-by: Thomas Gleixner Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Tom Zanussi Cc: Thomas Gleixner LKML-Reference: Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index e2c2de201ee..564491fa18b 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -197,7 +197,7 @@ static void sig_atexit(void) if (child_pid > 0) kill(child_pid, SIGTERM); - if (signr == -1) + if (signr == -1 || signr == SIGUSR1) return; signal(signr, SIG_DFL); @@ -515,6 +515,7 @@ static int __cmd_record(int argc, const char **argv) atexit(sig_atexit); signal(SIGCHLD, sig_handler); signal(SIGINT, sig_handler); + signal(SIGUSR1, sig_handler); if (forks && (pipe(child_ready_pipe) < 0 || pipe(go_pipe) < 0)) { perror("failed to create pipes"); @@ -606,6 +607,7 @@ static int __cmd_record(int argc, const char **argv) execvp(argv[0], (char **)argv); perror(argv[0]); + kill(getppid(), SIGUSR1); exit(-1); } @@ -762,7 +764,7 @@ static int __cmd_record(int argc, const char **argv) } } - if (quiet) + if (quiet || signr == SIGUSR1) return 0; fprintf(stderr, "[ perf record: Woken up %ld times to write data ]\n", waking); -- cgit v1.2.3