From 7317d7b87edb41a9135e30be1ec3f7ef817c53dd Mon Sep 17 00:00:00 2001 From: Nick Piggin Date: Tue, 30 Sep 2008 20:50:27 +1000 Subject: sched: improve preempt debugging This patch helped me out with a problem I recently had.... Basically, when the kernel lock is held, then preempt_count underflow does not get detected until it is released which may be a long time (and arbitrarily, eg at different points it may be rescheduled). If the bkl is released at schedule, the resulting output is actually fairly cryptic... With any other lock that elevates preempt_count, it is illegal to schedule under it (which would get found pretty quickly). bkl allows scheduling with preempt_count elevated, which makes underflows hard to debug. Signed-off-by: Ingo Molnar --- kernel/sched.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index 98890807375b..ec3bd1f398b3 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -4305,7 +4305,7 @@ void __kprobes sub_preempt_count(int val) /* * Underflow? */ - if (DEBUG_LOCKS_WARN_ON(val > preempt_count())) + if (DEBUG_LOCKS_WARN_ON(val > preempt_count() - (!!kernel_locked()))) return; /* * Is the spinlock portion underflowing? -- cgit v1.2.3 From eab172294d5e24464f332dd8e94a57a9819c81c4 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Wed, 29 Oct 2008 17:03:22 +0800 Subject: sched: cleanup for alloc_rt/fair_sched_group() Impact: cleanup Remove checking parent == NULL. It won't be NULLL, because we dynamically create sub task_group only, and sub task_group always has its parent. (root task_group is statically defined) Also replace kmalloc_node(GFP_ZERO) with kzalloc_node(). Signed-off-by: Li Zefan Signed-off-by: Ingo Molnar --- kernel/sched.c | 26 ++++++++++++-------------- 1 file changed, 12 insertions(+), 14 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index e8819bc6f462..7dd6c860773b 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -8472,7 +8472,7 @@ static int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent) { struct cfs_rq *cfs_rq; - struct sched_entity *se, *parent_se; + struct sched_entity *se; struct rq *rq; int i; @@ -8488,18 +8488,17 @@ int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent) for_each_possible_cpu(i) { rq = cpu_rq(i); - cfs_rq = kmalloc_node(sizeof(struct cfs_rq), - GFP_KERNEL|__GFP_ZERO, cpu_to_node(i)); + cfs_rq = kzalloc_node(sizeof(struct cfs_rq), + GFP_KERNEL, cpu_to_node(i)); if (!cfs_rq) goto err; - se = kmalloc_node(sizeof(struct sched_entity), - GFP_KERNEL|__GFP_ZERO, cpu_to_node(i)); + se = kzalloc_node(sizeof(struct sched_entity), + GFP_KERNEL, cpu_to_node(i)); if (!se) goto err; - parent_se = parent ? parent->se[i] : NULL; - init_tg_cfs_entry(tg, cfs_rq, se, i, 0, parent_se); + init_tg_cfs_entry(tg, cfs_rq, se, i, 0, parent->se[i]); } return 1; @@ -8560,7 +8559,7 @@ static int alloc_rt_sched_group(struct task_group *tg, struct task_group *parent) { struct rt_rq *rt_rq; - struct sched_rt_entity *rt_se, *parent_se; + struct sched_rt_entity *rt_se; struct rq *rq; int i; @@ -8577,18 +8576,17 @@ int alloc_rt_sched_group(struct task_group *tg, struct task_group *parent) for_each_possible_cpu(i) { rq = cpu_rq(i); - rt_rq = kmalloc_node(sizeof(struct rt_rq), - GFP_KERNEL|__GFP_ZERO, cpu_to_node(i)); + rt_rq = kzalloc_node(sizeof(struct rt_rq), + GFP_KERNEL, cpu_to_node(i)); if (!rt_rq) goto err; - rt_se = kmalloc_node(sizeof(struct sched_rt_entity), - GFP_KERNEL|__GFP_ZERO, cpu_to_node(i)); + rt_se = kzalloc_node(sizeof(struct sched_rt_entity), + GFP_KERNEL, cpu_to_node(i)); if (!rt_se) goto err; - parent_se = parent ? parent->rt_se[i] : NULL; - init_tg_rt_entry(tg, rt_rq, rt_se, i, 0, parent_se); + init_tg_rt_entry(tg, rt_rq, rt_se, i, 0, parent->rt_se[i]); } return 1; -- cgit v1.2.3 From 34f3a814eef8069a24e5b3ebcf27aba9dabac2ea Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Thu, 30 Oct 2008 15:23:32 +0800 Subject: sched: switch sched_features to seqfile Impact: cleanup So handling of sched_features read is simplified. Signed-off-by: Li Zefan Acked-by: Peter Zijlstra Signed-off-by: Ingo Molnar --- kernel/sched.c | 52 ++++++++++++++++------------------------------------ 1 file changed, 16 insertions(+), 36 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index 7dd6c860773b..5419df9cc5c4 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -703,45 +703,18 @@ static __read_mostly char *sched_feat_names[] = { #undef SCHED_FEAT -static int sched_feat_open(struct inode *inode, struct file *filp) +static int sched_feat_show(struct seq_file *m, void *v) { - filp->private_data = inode->i_private; - return 0; -} - -static ssize_t -sched_feat_read(struct file *filp, char __user *ubuf, - size_t cnt, loff_t *ppos) -{ - char *buf; - int r = 0; - int len = 0; int i; for (i = 0; sched_feat_names[i]; i++) { - len += strlen(sched_feat_names[i]); - len += 4; + if (!(sysctl_sched_features & (1UL << i))) + seq_puts(m, "NO_"); + seq_printf(m, "%s ", sched_feat_names[i]); } + seq_puts(m, "\n"); - buf = kmalloc(len + 2, GFP_KERNEL); - if (!buf) - return -ENOMEM; - - for (i = 0; sched_feat_names[i]; i++) { - if (sysctl_sched_features & (1UL << i)) - r += sprintf(buf + r, "%s ", sched_feat_names[i]); - else - r += sprintf(buf + r, "NO_%s ", sched_feat_names[i]); - } - - r += sprintf(buf + r, "\n"); - WARN_ON(r >= len + 2); - - r = simple_read_from_buffer(ubuf, cnt, ppos, buf, r); - - kfree(buf); - - return r; + return 0; } static ssize_t @@ -786,10 +759,17 @@ sched_feat_write(struct file *filp, const char __user *ubuf, return cnt; } +static int sched_feat_open(struct inode *inode, struct file *filp) +{ + return single_open(filp, sched_feat_show, NULL); +} + static struct file_operations sched_feat_fops = { - .open = sched_feat_open, - .read = sched_feat_read, - .write = sched_feat_write, + .open = sched_feat_open, + .write = sched_feat_write, + .read = seq_read, + .llseek = seq_lseek, + .release = single_release, }; static __init int sched_init_debug(void) -- cgit v1.2.3 From eefd796a8e831408ce17e633d73d70430748c47a Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 4 Nov 2008 16:15:37 +0800 Subject: sched debug: remove sd_level_to_string() Impact: cleanup Just use the newly introduced sd->name. Signed-off-by: Li Zefan Acked-by: Peter Zijlstra Signed-off-by: Ingo Molnar --- kernel/sched.c | 25 +------------------------ 1 file changed, 1 insertion(+), 24 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index 5419df9cc5c4..7ac59bae87d2 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -6602,28 +6602,6 @@ early_initcall(migration_init); #ifdef CONFIG_SCHED_DEBUG -static inline const char *sd_level_to_string(enum sched_domain_level lvl) -{ - switch (lvl) { - case SD_LV_NONE: - return "NONE"; - case SD_LV_SIBLING: - return "SIBLING"; - case SD_LV_MC: - return "MC"; - case SD_LV_CPU: - return "CPU"; - case SD_LV_NODE: - return "NODE"; - case SD_LV_ALLNODES: - return "ALLNODES"; - case SD_LV_MAX: - return "MAX"; - - } - return "MAX"; -} - static int sched_domain_debug_one(struct sched_domain *sd, int cpu, int level, cpumask_t *groupmask) { @@ -6643,8 +6621,7 @@ static int sched_domain_debug_one(struct sched_domain *sd, int cpu, int level, return -1; } - printk(KERN_CONT "span %s level %s\n", - str, sd_level_to_string(sd->level)); + printk(KERN_CONT "span %s level %s\n", str, sd->name); if (!cpu_isset(cpu, sd->span)) { printk(KERN_ERR "ERROR: domain->span does not contain " -- cgit v1.2.3 From a17e2260926f681a0eb983c1e3cb859ba2064bce Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 4 Nov 2008 16:19:13 +0800 Subject: sched: remove redundant call to unregister_sched_domain_sysctl() Impact: cleanup The sysctl has been unregistered by partition_sched_domains(). Signed-off-by: Li Zefan Acked-by: Peter Zijlstra Signed-off-by: Ingo Molnar --- kernel/sched.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index 7ac59bae87d2..3cb94fad33ca 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -7691,8 +7691,6 @@ static void detach_destroy_domains(const cpumask_t *cpu_map) cpumask_t tmpmask; int i; - unregister_sched_domain_sysctl(); - for_each_cpu_mask_nr(i, *cpu_map) cpu_attach_domain(NULL, &def_root_domain, i); synchronize_sched(); -- cgit v1.2.3 From faa2f98f856e89d1afb6e4a91707284d242e816e Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 4 Nov 2008 16:20:23 +0800 Subject: sched: add sanity check in partition_sched_domains() Impact: cleanup, add debug check It's wrong to make dattr_new = NULL if doms_new == NULL, it introduces memory leak if dattr_new != NULL. Fortunately dattr_new is always NULL in this case. So remove the code and add a sanity check. Signed-off-by: Li Zefan Acked-by: Peter Zijlstra Signed-off-by: Ingo Molnar --- kernel/sched.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index 3cb94fad33ca..213cad5e50aa 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -7767,7 +7767,7 @@ match1: ndoms_cur = 0; doms_new = &fallback_doms; cpus_andnot(doms_new[0], cpu_online_map, cpu_isolated_map); - dattr_new = NULL; + WARN_ON_ONCE(dattr_new); } /* Build new domains */ -- cgit v1.2.3 From cf7f8690e864c6fe11e77202dd847fa60f483418 Mon Sep 17 00:00:00 2001 From: Sripathi Kodi Date: Wed, 5 Nov 2008 18:57:14 +0530 Subject: sched, lockdep: inline double_unlock_balance() We have a test case which measures the variation in the amount of time needed to perform a fixed amount of work on the preempt_rt kernel. We started seeing deterioration in it's performance recently. The test should never take more than 10 microseconds, but we started 5-10% failure rate. Using elimination method, we traced the problem to commit 1b12bbc747560ea68bcc132c3d05699e52271da0 (lockdep: re-annotate scheduler runqueues). When LOCKDEP is disabled, this patch only adds an additional function call to double_unlock_balance(). Hence I inlined double_unlock_balance() and the problem went away. Here is a patch to make this change. Signed-off-by: Sripathi Kodi Acked-by: Peter Zijlstra Signed-off-by: Ingo Molnar --- kernel/sched.c | 2 +- kernel/sched_rt.c | 3 ++- 2 files changed, 3 insertions(+), 2 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index e8819bc6f462..ad10d0aae1d7 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -2825,7 +2825,7 @@ static int double_lock_balance(struct rq *this_rq, struct rq *busiest) return ret; } -static void double_unlock_balance(struct rq *this_rq, struct rq *busiest) +static inline void double_unlock_balance(struct rq *this_rq, struct rq *busiest) __releases(busiest->lock) { spin_unlock(&busiest->lock); diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c index c7963d5d0625..2bdd44423599 100644 --- a/kernel/sched_rt.c +++ b/kernel/sched_rt.c @@ -910,7 +910,8 @@ static void put_prev_task_rt(struct rq *rq, struct task_struct *p) #define RT_MAX_TRIES 3 static int double_lock_balance(struct rq *this_rq, struct rq *busiest); -static void double_unlock_balance(struct rq *this_rq, struct rq *busiest); +static inline void double_unlock_balance(struct rq *this_rq, + struct rq *busiest); static void deactivate_task(struct rq *rq, struct task_struct *p, int sleep); -- cgit v1.2.3 From 6d21cd62516a9697cb7ec33cc52e6b814fb65a13 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Fri, 7 Nov 2008 17:03:18 +0800 Subject: sched: clean up SCHED_CPUMASK_ALLOC Impact: cleanup The #if/#endif is ugly. Change SCHED_CPUMASK_ALLOC and SCHED_CPUMASK_FREE to static inline functions. Signed-off-by: Li Zefan Acked-by: Peter Zijlstra Signed-off-by: Ingo Molnar --- kernel/sched.c | 29 ++++++++++++++++++----------- 1 file changed, 18 insertions(+), 11 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index b24e57a10f6f..59db86c915f9 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -7279,13 +7279,21 @@ struct allmasks { }; #if NR_CPUS > 128 -#define SCHED_CPUMASK_ALLOC 1 -#define SCHED_CPUMASK_FREE(v) kfree(v) -#define SCHED_CPUMASK_DECLARE(v) struct allmasks *v +#define SCHED_CPUMASK_DECLARE(v) struct allmasks *v +static inline void sched_cpumask_alloc(struct allmasks **masks) +{ + *masks = kmalloc(sizeof(**masks), GFP_KERNEL); +} +static inline void sched_cpumask_free(struct allmasks *masks) +{ + kfree(masks); +} #else -#define SCHED_CPUMASK_ALLOC 0 -#define SCHED_CPUMASK_FREE(v) -#define SCHED_CPUMASK_DECLARE(v) struct allmasks _v, *v = &_v +#define SCHED_CPUMASK_DECLARE(v) struct allmasks _v, *v = &_v +static inline void sched_cpumask_alloc(struct allmasks **masks) +{ } +static inline void sched_cpumask_free(struct allmasks *masks) +{ } #endif #define SCHED_CPUMASK_VAR(v, a) cpumask_t *v = (cpumask_t *) \ @@ -7361,9 +7369,8 @@ static int __build_sched_domains(const cpumask_t *cpu_map, return -ENOMEM; } -#if SCHED_CPUMASK_ALLOC /* get space for all scratch cpumask variables */ - allmasks = kmalloc(sizeof(*allmasks), GFP_KERNEL); + sched_cpumask_alloc(&allmasks); if (!allmasks) { printk(KERN_WARNING "Cannot alloc cpumask array\n"); kfree(rd); @@ -7372,7 +7379,7 @@ static int __build_sched_domains(const cpumask_t *cpu_map, #endif return -ENOMEM; } -#endif + tmpmask = (cpumask_t *)allmasks; @@ -7626,13 +7633,13 @@ static int __build_sched_domains(const cpumask_t *cpu_map, cpu_attach_domain(sd, rd, i); } - SCHED_CPUMASK_FREE((void *)allmasks); + sched_cpumask_free(allmasks); return 0; #ifdef CONFIG_NUMA error: free_sched_groups(cpu_map, tmpmask); - SCHED_CPUMASK_FREE((void *)allmasks); + sched_cpumask_free(allmasks); kfree(rd); return -ENOMEM; #endif -- cgit v1.2.3 From 934352f214b3251eb0793c1209d346595a661d80 Mon Sep 17 00:00:00 2001 From: Bharata B Rao Date: Mon, 10 Nov 2008 20:41:13 +0530 Subject: sched: add hierarchical accounting to cpu accounting controller Impact: improve CPU time accounting of tasks under the cpu accounting controller Add hierarchical accounting to cpu accounting controller and include cpuacct documentation. Currently, while charging the task's cputime to its accounting group, the accounting group hierarchy isn't updated. This patch charges the cputime of a task to its accounting group and all its parent accounting groups. Reported-by: Srivatsa Vaddagiri Signed-off-by: Bharata B Rao Reviewed-by: Paul Menage Acked-by: Balbir Singh Acked-by: Peter Zijlstra Signed-off-by: Ingo Molnar --- Documentation/controllers/cpuacct.txt | 32 ++++++++++++++++++++++++++++++++ kernel/sched.c | 12 +++++++++--- 2 files changed, 41 insertions(+), 3 deletions(-) create mode 100644 Documentation/controllers/cpuacct.txt (limited to 'kernel/sched.c') diff --git a/Documentation/controllers/cpuacct.txt b/Documentation/controllers/cpuacct.txt new file mode 100644 index 000000000000..bb775fbe43d7 --- /dev/null +++ b/Documentation/controllers/cpuacct.txt @@ -0,0 +1,32 @@ +CPU Accounting Controller +------------------------- + +The CPU accounting controller is used to group tasks using cgroups and +account the CPU usage of these groups of tasks. + +The CPU accounting controller supports multi-hierarchy groups. An accounting +group accumulates the CPU usage of all of its child groups and the tasks +directly present in its group. + +Accounting groups can be created by first mounting the cgroup filesystem. + +# mkdir /cgroups +# mount -t cgroup -ocpuacct none /cgroups + +With the above step, the initial or the parent accounting group +becomes visible at /cgroups. At bootup, this group includes all the +tasks in the system. /cgroups/tasks lists the tasks in this cgroup. +/cgroups/cpuacct.usage gives the CPU time (in nanoseconds) obtained by +this group which is essentially the CPU time obtained by all the tasks +in the system. + +New accounting groups can be created under the parent group /cgroups. + +# cd /cgroups +# mkdir g1 +# echo $$ > g1 + +The above steps create a new group g1 and move the current shell +process (bash) into it. CPU time consumed by this bash and its children +can be obtained from g1/cpuacct.usage and the same is accumulated in +/cgroups/cpuacct.usage also. diff --git a/kernel/sched.c b/kernel/sched.c index 59db86c915f9..ebaf432365f6 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -9196,11 +9196,12 @@ struct cgroup_subsys cpu_cgroup_subsys = { * (balbir@in.ibm.com). */ -/* track cpu usage of a group of tasks */ +/* track cpu usage of a group of tasks and its child groups */ struct cpuacct { struct cgroup_subsys_state css; /* cpuusage holds pointer to a u64-type object on every cpu */ u64 *cpuusage; + struct cpuacct *parent; }; struct cgroup_subsys cpuacct_subsys; @@ -9234,6 +9235,9 @@ static struct cgroup_subsys_state *cpuacct_create( return ERR_PTR(-ENOMEM); } + if (cgrp->parent) + ca->parent = cgroup_ca(cgrp->parent); + return &ca->css; } @@ -9313,14 +9317,16 @@ static int cpuacct_populate(struct cgroup_subsys *ss, struct cgroup *cgrp) static void cpuacct_charge(struct task_struct *tsk, u64 cputime) { struct cpuacct *ca; + int cpu; if (!cpuacct_subsys.active) return; + cpu = task_cpu(tsk); ca = task_ca(tsk); - if (ca) { - u64 *cpuusage = percpu_ptr(ca->cpuusage, task_cpu(tsk)); + for (; ca; ca = ca->parent) { + u64 *cpuusage = percpu_ptr(ca->cpuusage, cpu); *cpuusage += cputime; } } -- cgit v1.2.3 From 7e066fb870fcd1025ec3ba7bbde5d541094f4ce1 Mon Sep 17 00:00:00 2001 From: Mathieu Desnoyers Date: Fri, 14 Nov 2008 17:47:47 -0500 Subject: tracepoints: add DECLARE_TRACE() and DEFINE_TRACE() Impact: API *CHANGE*. Must update all tracepoint users. Add DEFINE_TRACE() to tracepoints to let them declare the tracepoint structure in a single spot for all the kernel. It helps reducing memory consumption, especially when declaring a lot of tracepoints, e.g. for kmalloc tracing. *API CHANGE WARNING*: now, DECLARE_TRACE() must be used in headers for tracepoint declarations rather than DEFINE_TRACE(). This is the sane way to do it. The name previously used was misleading. Updates scheduler instrumentation to follow this API change. Signed-off-by: Mathieu Desnoyers Signed-off-by: Ingo Molnar --- Documentation/tracepoints.txt | 7 ++++++- include/asm-generic/vmlinux.lds.h | 1 + include/linux/tracepoint.h | 35 +++++++++++++++++++++++---------- include/trace/sched.h | 24 +++++++++++----------- kernel/exit.c | 4 ++++ kernel/fork.c | 2 ++ kernel/kthread.c | 3 +++ kernel/sched.c | 6 ++++++ kernel/signal.c | 2 ++ samples/tracepoints/tp-samples-trace.h | 4 ++-- samples/tracepoints/tracepoint-sample.c | 3 +++ 11 files changed, 66 insertions(+), 25 deletions(-) (limited to 'kernel/sched.c') diff --git a/Documentation/tracepoints.txt b/Documentation/tracepoints.txt index 5d354e167494..e8ad47b437f3 100644 --- a/Documentation/tracepoints.txt +++ b/Documentation/tracepoints.txt @@ -42,7 +42,7 @@ In include/trace/subsys.h : #include -DEFINE_TRACE(subsys_eventname, +DECLARE_TRACE(subsys_eventname, TPPTOTO(int firstarg, struct task_struct *p), TPARGS(firstarg, p)); @@ -50,6 +50,8 @@ In subsys/file.c (where the tracing statement must be added) : #include +DEFINE_TRACE(subsys_eventname); + void somefct(void) { ... @@ -86,6 +88,9 @@ to limit collisions. Tracepoint names are global to the kernel: they are considered as being the same whether they are in the core kernel image or in modules. +If the tracepoint has to be used in kernel modules, an +EXPORT_TRACEPOINT_SYMBOL_GPL() or EXPORT_TRACEPOINT_SYMBOL() can be used to +export the defined tracepoints. * Probe / tracepoint example diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h index a5e4ed9baec8..3b46ae464933 100644 --- a/include/asm-generic/vmlinux.lds.h +++ b/include/asm-generic/vmlinux.lds.h @@ -71,6 +71,7 @@ VMLINUX_SYMBOL(__start___markers) = .; \ *(__markers) \ VMLINUX_SYMBOL(__stop___markers) = .; \ + . = ALIGN(32); \ VMLINUX_SYMBOL(__start___tracepoints) = .; \ *(__tracepoints) \ VMLINUX_SYMBOL(__stop___tracepoints) = .; \ diff --git a/include/linux/tracepoint.h b/include/linux/tracepoint.h index 7e9b42aeae0e..757005458366 100644 --- a/include/linux/tracepoint.h +++ b/include/linux/tracepoint.h @@ -24,8 +24,12 @@ struct tracepoint { const char *name; /* Tracepoint name */ int state; /* State. */ void **funcs; -} __attribute__((aligned(8))); - +} __attribute__((aligned(32))); /* + * Aligned on 32 bytes because it is + * globally visible and gcc happily + * align these on the structure size. + * Keep in sync with vmlinux.lds.h. + */ #define TPPROTO(args...) args #define TPARGS(args...) args @@ -55,15 +59,10 @@ struct tracepoint { * not add unwanted padding between the beginning of the section and the * structure. Force alignment to the same alignment as the section start. */ -#define DEFINE_TRACE(name, proto, args) \ +#define DECLARE_TRACE(name, proto, args) \ + extern struct tracepoint __tracepoint_##name; \ static inline void trace_##name(proto) \ { \ - static const char __tpstrtab_##name[] \ - __attribute__((section("__tracepoints_strings"))) \ - = #name; \ - static struct tracepoint __tracepoint_##name \ - __attribute__((section("__tracepoints"), aligned(8))) = \ - { __tpstrtab_##name, 0, NULL }; \ if (unlikely(__tracepoint_##name.state)) \ __DO_TRACE(&__tracepoint_##name, \ TPPROTO(proto), TPARGS(args)); \ @@ -77,11 +76,23 @@ struct tracepoint { return tracepoint_probe_unregister(#name, (void *)probe);\ } +#define DEFINE_TRACE(name) \ + static const char __tpstrtab_##name[] \ + __attribute__((section("__tracepoints_strings"))) = #name; \ + struct tracepoint __tracepoint_##name \ + __attribute__((section("__tracepoints"), aligned(32))) = \ + { __tpstrtab_##name, 0, NULL } + +#define EXPORT_TRACEPOINT_SYMBOL_GPL(name) \ + EXPORT_SYMBOL_GPL(__tracepoint_##name) +#define EXPORT_TRACEPOINT_SYMBOL(name) \ + EXPORT_SYMBOL(__tracepoint_##name) + extern void tracepoint_update_probe_range(struct tracepoint *begin, struct tracepoint *end); #else /* !CONFIG_TRACEPOINTS */ -#define DEFINE_TRACE(name, proto, args) \ +#define DECLARE_TRACE(name, proto, args) \ static inline void _do_trace_##name(struct tracepoint *tp, proto) \ { } \ static inline void trace_##name(proto) \ @@ -95,6 +106,10 @@ extern void tracepoint_update_probe_range(struct tracepoint *begin, return -ENOSYS; \ } +#define DEFINE_TRACE(name) +#define EXPORT_TRACEPOINT_SYMBOL_GPL(name) +#define EXPORT_TRACEPOINT_SYMBOL(name) + static inline void tracepoint_update_probe_range(struct tracepoint *begin, struct tracepoint *end) { } diff --git a/include/trace/sched.h b/include/trace/sched.h index ad47369d01b5..9b2854abf7e2 100644 --- a/include/trace/sched.h +++ b/include/trace/sched.h @@ -4,52 +4,52 @@ #include #include -DEFINE_TRACE(sched_kthread_stop, +DECLARE_TRACE(sched_kthread_stop, TPPROTO(struct task_struct *t), TPARGS(t)); -DEFINE_TRACE(sched_kthread_stop_ret, +DECLARE_TRACE(sched_kthread_stop_ret, TPPROTO(int ret), TPARGS(ret)); -DEFINE_TRACE(sched_wait_task, +DECLARE_TRACE(sched_wait_task, TPPROTO(struct rq *rq, struct task_struct *p), TPARGS(rq, p)); -DEFINE_TRACE(sched_wakeup, +DECLARE_TRACE(sched_wakeup, TPPROTO(struct rq *rq, struct task_struct *p), TPARGS(rq, p)); -DEFINE_TRACE(sched_wakeup_new, +DECLARE_TRACE(sched_wakeup_new, TPPROTO(struct rq *rq, struct task_struct *p), TPARGS(rq, p)); -DEFINE_TRACE(sched_switch, +DECLARE_TRACE(sched_switch, TPPROTO(struct rq *rq, struct task_struct *prev, struct task_struct *next), TPARGS(rq, prev, next)); -DEFINE_TRACE(sched_migrate_task, +DECLARE_TRACE(sched_migrate_task, TPPROTO(struct rq *rq, struct task_struct *p, int dest_cpu), TPARGS(rq, p, dest_cpu)); -DEFINE_TRACE(sched_process_free, +DECLARE_TRACE(sched_process_free, TPPROTO(struct task_struct *p), TPARGS(p)); -DEFINE_TRACE(sched_process_exit, +DECLARE_TRACE(sched_process_exit, TPPROTO(struct task_struct *p), TPARGS(p)); -DEFINE_TRACE(sched_process_wait, +DECLARE_TRACE(sched_process_wait, TPPROTO(struct pid *pid), TPARGS(pid)); -DEFINE_TRACE(sched_process_fork, +DECLARE_TRACE(sched_process_fork, TPPROTO(struct task_struct *parent, struct task_struct *child), TPARGS(parent, child)); -DEFINE_TRACE(sched_signal_send, +DECLARE_TRACE(sched_signal_send, TPPROTO(int sig, struct task_struct *p), TPARGS(sig, p)); diff --git a/kernel/exit.c b/kernel/exit.c index ae2b92be5fae..f995d2418668 100644 --- a/kernel/exit.c +++ b/kernel/exit.c @@ -54,6 +54,10 @@ #include #include +DEFINE_TRACE(sched_process_free); +DEFINE_TRACE(sched_process_exit); +DEFINE_TRACE(sched_process_wait); + static void exit_mm(struct task_struct * tsk); static inline int task_detached(struct task_struct *p) diff --git a/kernel/fork.c b/kernel/fork.c index f6083561dfe0..0837d0deee5f 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -79,6 +79,8 @@ DEFINE_PER_CPU(unsigned long, process_counts) = 0; __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */ +DEFINE_TRACE(sched_process_fork); + int nr_processes(void) { int cpu; diff --git a/kernel/kthread.c b/kernel/kthread.c index 8e7a7ce3ed0a..4fbc456f393d 100644 --- a/kernel/kthread.c +++ b/kernel/kthread.c @@ -21,6 +21,9 @@ static DEFINE_SPINLOCK(kthread_create_lock); static LIST_HEAD(kthread_create_list); struct task_struct *kthreadd_task; +DEFINE_TRACE(sched_kthread_stop); +DEFINE_TRACE(sched_kthread_stop_ret); + struct kthread_create_info { /* Information passed to kthread() from kthreadd. */ diff --git a/kernel/sched.c b/kernel/sched.c index 50a21f964679..327f91c63c99 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -118,6 +118,12 @@ */ #define RUNTIME_INF ((u64)~0ULL) +DEFINE_TRACE(sched_wait_task); +DEFINE_TRACE(sched_wakeup); +DEFINE_TRACE(sched_wakeup_new); +DEFINE_TRACE(sched_switch); +DEFINE_TRACE(sched_migrate_task); + #ifdef CONFIG_SMP /* * Divide a load by a sched group cpu_power : (load / sg->__cpu_power) diff --git a/kernel/signal.c b/kernel/signal.c index 4530fc654455..e9afe63da24b 100644 --- a/kernel/signal.c +++ b/kernel/signal.c @@ -41,6 +41,8 @@ static struct kmem_cache *sigqueue_cachep; +DEFINE_TRACE(sched_signal_send); + static void __user *sig_handler(struct task_struct *t, int sig) { return t->sighand->action[sig - 1].sa.sa_handler; diff --git a/samples/tracepoints/tp-samples-trace.h b/samples/tracepoints/tp-samples-trace.h index 0216b55bd640..01724e04c556 100644 --- a/samples/tracepoints/tp-samples-trace.h +++ b/samples/tracepoints/tp-samples-trace.h @@ -4,10 +4,10 @@ #include /* for struct inode and struct file */ #include -DEFINE_TRACE(subsys_event, +DECLARE_TRACE(subsys_event, TPPROTO(struct inode *inode, struct file *file), TPARGS(inode, file)); -DEFINE_TRACE(subsys_eventb, +DECLARE_TRACE(subsys_eventb, TPPROTO(void), TPARGS()); #endif diff --git a/samples/tracepoints/tracepoint-sample.c b/samples/tracepoints/tracepoint-sample.c index 4ae4b7fcc043..00d169792a3e 100644 --- a/samples/tracepoints/tracepoint-sample.c +++ b/samples/tracepoints/tracepoint-sample.c @@ -13,6 +13,9 @@ #include #include "tp-samples-trace.h" +DEFINE_TRACE(subsys_event); +DEFINE_TRACE(subsys_eventb); + struct proc_dir_entry *pentry_example; static int my_open(struct inode *inode, struct file *file) -- cgit v1.2.3 From 74fcd524e808975dd546dac847119f1995a7c622 Mon Sep 17 00:00:00 2001 From: Oleg Nesterov Date: Mon, 17 Nov 2008 15:39:52 +0100 Subject: account_steal_time: kill the unneeded account_group_system_time() Impact: remove unnecessary accounting call I don't actually understand account_steal_time() and I failed to find the commit which added account_group_system_time(), but this looks bogus. In any case rq->idle must be single-threaded, so it can't have ->totals. Signed-off-by: Oleg Nesterov Signed-off-by: Ingo Molnar --- kernel/sched.c | 1 - 1 file changed, 1 deletion(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index c94baf2969e7..b388c9b243e9 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -4202,7 +4202,6 @@ void account_steal_time(struct task_struct *p, cputime_t steal) if (p == rq->idle) { p->stime = cputime_add(p->stime, steal); - account_group_system_time(p, steal); if (atomic_read(&rq->nr_iowait) > 0) cpustat->iowait = cputime64_add(cpustat->iowait, tmp); else -- cgit v1.2.3 From ec4e0e2fe018992d980910db901637c814575914 Mon Sep 17 00:00:00 2001 From: Ken Chen Date: Tue, 18 Nov 2008 22:41:57 -0800 Subject: sched: fix inconsistency when redistribute per-cpu tg->cfs_rq shares Impact: make load-balancing more consistent In the update_shares() path leading to tg_shares_up(), the calculation of per-cpu cfs_rq shares is rather erratic even under moderate task wake up rate. The problem is that the per-cpu tg->cfs_rq load weight used in the sd_rq_weight aggregation and actual redistribution of the cfs_rq->shares are collected at different time. Under moderate system load, we've seen quite a bit of variation on the cfs_rq->shares and ultimately wildly affects sched_entity's load weight. This patch caches the result of initial per-cpu load weight when doing the sum calculation, and then pass it down to update_group_shares_cpu() for redistributing per-cpu cfs_rq shares. This allows consistent total cfs_rq shares across all CPUs. It also simplifies the rounding and zero load weight check. Signed-off-by: Ken Chen Acked-by: Peter Zijlstra Signed-off-by: Ingo Molnar --- kernel/sched.c | 41 +++++++++++++++-------------------------- 1 file changed, 15 insertions(+), 26 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index a4c156d9a4a5..93bfb086e60f 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -1453,27 +1453,13 @@ static void update_group_shares_cpu(struct task_group *tg, int cpu, unsigned long sd_shares, unsigned long sd_rq_weight) { - int boost = 0; unsigned long shares; unsigned long rq_weight; if (!tg->se[cpu]) return; - rq_weight = tg->cfs_rq[cpu]->load.weight; - - /* - * If there are currently no tasks on the cpu pretend there is one of - * average load so that when a new task gets to run here it will not - * get delayed by group starvation. - */ - if (!rq_weight) { - boost = 1; - rq_weight = NICE_0_LOAD; - } - - if (unlikely(rq_weight > sd_rq_weight)) - rq_weight = sd_rq_weight; + rq_weight = tg->cfs_rq[cpu]->rq_weight; /* * \Sum shares * rq_weight @@ -1481,7 +1467,7 @@ update_group_shares_cpu(struct task_group *tg, int cpu, * \Sum rq_weight * */ - shares = (sd_shares * rq_weight) / (sd_rq_weight + 1); + shares = (sd_shares * rq_weight) / sd_rq_weight; shares = clamp_t(unsigned long, shares, MIN_SHARES, MAX_SHARES); if (abs(shares - tg->se[cpu]->load.weight) > @@ -1490,11 +1476,7 @@ update_group_shares_cpu(struct task_group *tg, int cpu, unsigned long flags; spin_lock_irqsave(&rq->lock, flags); - /* - * record the actual number of shares, not the boosted amount. - */ - tg->cfs_rq[cpu]->shares = boost ? 0 : shares; - tg->cfs_rq[cpu]->rq_weight = rq_weight; + tg->cfs_rq[cpu]->shares = shares; __set_se_shares(tg->se[cpu], shares); spin_unlock_irqrestore(&rq->lock, flags); @@ -1508,13 +1490,23 @@ update_group_shares_cpu(struct task_group *tg, int cpu, */ static int tg_shares_up(struct task_group *tg, void *data) { - unsigned long rq_weight = 0; + unsigned long weight, rq_weight = 0; unsigned long shares = 0; struct sched_domain *sd = data; int i; for_each_cpu_mask(i, sd->span) { - rq_weight += tg->cfs_rq[i]->load.weight; + /* + * If there are currently no tasks on the cpu pretend there + * is one of average load so that when a new task gets to + * run here it will not get delayed by group starvation. + */ + weight = tg->cfs_rq[i]->load.weight; + if (!weight) + weight = NICE_0_LOAD; + + tg->cfs_rq[i]->rq_weight = weight; + rq_weight += weight; shares += tg->cfs_rq[i]->shares; } @@ -1524,9 +1516,6 @@ static int tg_shares_up(struct task_group *tg, void *data) if (!sd->parent || !(sd->parent->flags & SD_LOAD_BALANCE)) shares = tg->shares; - if (!rq_weight) - rq_weight = cpus_weight(sd->span) * NICE_0_LOAD; - for_each_cpu_mask(i, sd->span) update_group_shares_cpu(tg, i, shares, rq_weight); -- cgit v1.2.3 From 957ad0166e9f76a8561dafa5e14ef5bd3f5e9a3b Mon Sep 17 00:00:00 2001 From: Vegard Nossum Date: Fri, 21 Nov 2008 01:30:36 +0100 Subject: sched: update comment for move_task_off_dead_cpu Impact: cleanup This commit: commit f7b4cddcc5aca03e80e357360c9424dfba5056c2 Author: Oleg Nesterov Date: Tue Oct 16 23:30:56 2007 -0700 do CPU_DEAD migrating under read_lock(tasklist) instead of write_lock_irq(ta Currently move_task_off_dead_cpu() is called under write_lock_irq(tasklist). This means it can't use task_lock() which is needed to improve migrating to take task's ->cpuset into account. Change the code to call move_task_off_dead_cpu() with irqs enabled, and change migrate_live_tasks() to use read_lock(tasklist). ...forgot to update the comment in front of move_task_off_dead_cpu. Reference: http://lkml.org/lkml/2008/6/23/135 Signed-off-by: Vegard Nossum Signed-off-by: Ingo Molnar --- kernel/sched.c | 1 - 1 file changed, 1 deletion(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index 93bfb086e60f..a6085d5166dd 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -6094,7 +6094,6 @@ static int __migrate_task_irq(struct task_struct *p, int src_cpu, int dest_cpu) /* * Figure out where task on dead CPU should go, use force if necessary. - * NOTE: interrupts should be disabled by the caller */ static void move_task_off_dead_cpu(int dead_cpu, struct task_struct *p) { -- cgit v1.2.3 From f201ae2356c74bcae130b2177b3dca903ea98071 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Sun, 23 Nov 2008 06:22:56 +0100 Subject: tracing/function-return-tracer: store return stack into task_struct and allocate it dynamically Impact: use deeper function tracing depth safely Some tests showed that function return tracing needed a more deeper depth of function calls. But it could be unsafe to store these return addresses to the stack. So these arrays will now be allocated dynamically into task_struct of current only when the tracer is activated. Typical scheme when tracer is activated: - allocate a return stack for each task in global list. - fork: allocate the return stack for the newly created task - exit: free return stack of current - idle init: same as fork I chose a default depth of 50. I don't have overruns anymore. Signed-off-by: Frederic Weisbecker Signed-off-by: Ingo Molnar --- arch/x86/include/asm/ftrace.h | 1 - arch/x86/include/asm/thread_info.h | 29 ------------ arch/x86/kernel/ftrace.c | 29 ++++++------ include/linux/ftrace.h | 5 ++ include/linux/sched.h | 23 +++++---- kernel/exit.c | 5 +- kernel/fork.c | 4 ++ kernel/sched.c | 3 ++ kernel/trace/ftrace.c | 96 +++++++++++++++++++++++++++++++++++++- 9 files changed, 137 insertions(+), 58 deletions(-) (limited to 'kernel/sched.c') diff --git a/arch/x86/include/asm/ftrace.h b/arch/x86/include/asm/ftrace.h index 2bb43b433e07..754a3e082f94 100644 --- a/arch/x86/include/asm/ftrace.h +++ b/arch/x86/include/asm/ftrace.h @@ -29,7 +29,6 @@ struct dyn_arch_ftrace { #endif /* CONFIG_FUNCTION_TRACER */ #ifdef CONFIG_FUNCTION_RET_TRACER -#define FTRACE_RET_STACK_SIZE 20 #ifndef __ASSEMBLY__ diff --git a/arch/x86/include/asm/thread_info.h b/arch/x86/include/asm/thread_info.h index e90e81ef6ab9..0921b4018c11 100644 --- a/arch/x86/include/asm/thread_info.h +++ b/arch/x86/include/asm/thread_info.h @@ -40,36 +40,8 @@ struct thread_info { */ __u8 supervisor_stack[0]; #endif - -#ifdef CONFIG_FUNCTION_RET_TRACER - /* Index of current stored adress in ret_stack */ - int curr_ret_stack; - /* Stack of return addresses for return function tracing */ - struct ftrace_ret_stack ret_stack[FTRACE_RET_STACK_SIZE]; - /* - * Number of functions that haven't been traced - * because of depth overrun. - */ - atomic_t trace_overrun; -#endif }; -#ifdef CONFIG_FUNCTION_RET_TRACER -#define INIT_THREAD_INFO(tsk) \ -{ \ - .task = &tsk, \ - .exec_domain = &default_exec_domain, \ - .flags = 0, \ - .cpu = 0, \ - .preempt_count = 1, \ - .addr_limit = KERNEL_DS, \ - .restart_block = { \ - .fn = do_no_restart_syscall, \ - }, \ - .curr_ret_stack = -1,\ - .trace_overrun = ATOMIC_INIT(0) \ -} -#else #define INIT_THREAD_INFO(tsk) \ { \ .task = &tsk, \ @@ -82,7 +54,6 @@ struct thread_info { .fn = do_no_restart_syscall, \ }, \ } -#endif #define init_thread_info (init_thread_union.thread_info) #define init_stack (init_thread_union.stack) diff --git a/arch/x86/kernel/ftrace.c b/arch/x86/kernel/ftrace.c index 356bb1eb6e9a..bb137f7297ed 100644 --- a/arch/x86/kernel/ftrace.c +++ b/arch/x86/kernel/ftrace.c @@ -350,19 +350,21 @@ static int push_return_trace(unsigned long ret, unsigned long long time, unsigned long func) { int index; - struct thread_info *ti = current_thread_info(); + + if (!current->ret_stack) + return -EBUSY; /* The return trace stack is full */ - if (ti->curr_ret_stack == FTRACE_RET_STACK_SIZE - 1) { - atomic_inc(&ti->trace_overrun); + if (current->curr_ret_stack == FTRACE_RETFUNC_DEPTH - 1) { + atomic_inc(¤t->trace_overrun); return -EBUSY; } - index = ++ti->curr_ret_stack; + index = ++current->curr_ret_stack; barrier(); - ti->ret_stack[index].ret = ret; - ti->ret_stack[index].func = func; - ti->ret_stack[index].calltime = time; + current->ret_stack[index].ret = ret; + current->ret_stack[index].func = func; + current->ret_stack[index].calltime = time; return 0; } @@ -373,13 +375,12 @@ static void pop_return_trace(unsigned long *ret, unsigned long long *time, { int index; - struct thread_info *ti = current_thread_info(); - index = ti->curr_ret_stack; - *ret = ti->ret_stack[index].ret; - *func = ti->ret_stack[index].func; - *time = ti->ret_stack[index].calltime; - *overrun = atomic_read(&ti->trace_overrun); - ti->curr_ret_stack--; + index = current->curr_ret_stack; + *ret = current->ret_stack[index].ret; + *func = current->ret_stack[index].func; + *time = current->ret_stack[index].calltime; + *overrun = atomic_read(¤t->trace_overrun); + current->curr_ret_stack--; } /* diff --git a/include/linux/ftrace.h b/include/linux/ftrace.h index f7ba4ea5e128..2ba259b2defa 100644 --- a/include/linux/ftrace.h +++ b/include/linux/ftrace.h @@ -323,6 +323,8 @@ struct ftrace_retfunc { }; #ifdef CONFIG_FUNCTION_RET_TRACER +#define FTRACE_RETFUNC_DEPTH 50 +#define FTRACE_RETSTACK_ALLOC_SIZE 32 /* Type of a callback handler of tracing return function */ typedef void (*trace_function_return_t)(struct ftrace_retfunc *); @@ -330,6 +332,9 @@ extern int register_ftrace_return(trace_function_return_t func); /* The current handler in use */ extern trace_function_return_t ftrace_function_return; extern void unregister_ftrace_return(void); + +extern void ftrace_retfunc_init_task(struct task_struct *t); +extern void ftrace_retfunc_exit_task(struct task_struct *t); #endif #endif /* _LINUX_FTRACE_H */ diff --git a/include/linux/sched.h b/include/linux/sched.h index c8e0db464206..bee1e93c95ad 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1352,6 +1352,17 @@ struct task_struct { unsigned long default_timer_slack_ns; struct list_head *scm_work_list; +#ifdef CONFIG_FUNCTION_RET_TRACER + /* Index of current stored adress in ret_stack */ + int curr_ret_stack; + /* Stack of return addresses for return function tracing */ + struct ftrace_ret_stack *ret_stack; + /* + * Number of functions that haven't been traced + * because of depth overrun. + */ + atomic_t trace_overrun; +#endif }; /* @@ -2006,18 +2017,6 @@ static inline void setup_thread_stack(struct task_struct *p, struct task_struct { *task_thread_info(p) = *task_thread_info(org); task_thread_info(p)->task = p; - -#ifdef CONFIG_FUNCTION_RET_TRACER - /* - * When fork() creates a child process, this function is called. - * But the child task may not inherit the return adresses traced - * by the return function tracer because it will directly execute - * in userspace and will not return to kernel functions its parent - * used. - */ - task_thread_info(p)->curr_ret_stack = -1; - atomic_set(&task_thread_info(p)->trace_overrun, 0); -#endif } static inline unsigned long *end_of_stack(struct task_struct *p) diff --git a/kernel/exit.c b/kernel/exit.c index 35c8ec2ba03a..b9d446329da1 100644 --- a/kernel/exit.c +++ b/kernel/exit.c @@ -47,6 +47,7 @@ #include #include #include +#include #include #include @@ -1127,7 +1128,9 @@ NORET_TYPE void do_exit(long code) preempt_disable(); /* causes final put_task_struct in finish_task_switch(). */ tsk->state = TASK_DEAD; - +#ifdef CONFIG_FUNCTION_RET_TRACER + ftrace_retfunc_exit_task(tsk); +#endif schedule(); BUG(); /* Avoid "noreturn function does return". */ diff --git a/kernel/fork.c b/kernel/fork.c index ac62f43ee430..d1eb30e69ccc 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -47,6 +47,7 @@ #include #include #include +#include #include #include #include @@ -1269,6 +1270,9 @@ static struct task_struct *copy_process(unsigned long clone_flags, total_forks++; spin_unlock(¤t->sighand->siglock); write_unlock_irq(&tasklist_lock); +#ifdef CONFIG_FUNCTION_RET_TRACER + ftrace_retfunc_init_task(p); +#endif proc_fork_connector(p); cgroup_post_fork(p); return p; diff --git a/kernel/sched.c b/kernel/sched.c index 4de56108c86f..fb17205950de 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -5901,6 +5901,9 @@ void __cpuinit init_idle(struct task_struct *idle, int cpu) * The idle tasks have their own, simple scheduling class: */ idle->sched_class = &idle_sched_class; +#ifdef CONFIG_FUNCTION_RET_TRACER + ftrace_retfunc_init_task(idle); +#endif } /* diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c index f212da486689..90d99fb02ae4 100644 --- a/kernel/trace/ftrace.c +++ b/kernel/trace/ftrace.c @@ -1498,10 +1498,77 @@ ftrace_enable_sysctl(struct ctl_table *table, int write, #ifdef CONFIG_FUNCTION_RET_TRACER +static atomic_t ftrace_retfunc_active; + /* The callback that hooks the return of a function */ trace_function_return_t ftrace_function_return = (trace_function_return_t)ftrace_stub; + +/* Try to assign a return stack array on FTRACE_RETSTACK_ALLOC_SIZE tasks. */ +static int alloc_retstack_tasklist(struct ftrace_ret_stack **ret_stack_list) +{ + int i; + int ret = 0; + unsigned long flags; + int start = 0, end = FTRACE_RETSTACK_ALLOC_SIZE; + struct task_struct *g, *t; + + for (i = 0; i < FTRACE_RETSTACK_ALLOC_SIZE; i++) { + ret_stack_list[i] = kmalloc(FTRACE_RETFUNC_DEPTH + * sizeof(struct ftrace_ret_stack), + GFP_KERNEL); + if (!ret_stack_list[i]) { + start = 0; + end = i; + ret = -ENOMEM; + goto free; + } + } + + read_lock_irqsave(&tasklist_lock, flags); + do_each_thread(g, t) { + if (start == end) { + ret = -EAGAIN; + goto unlock; + } + + if (t->ret_stack == NULL) { + t->ret_stack = ret_stack_list[start++]; + t->curr_ret_stack = -1; + atomic_set(&t->trace_overrun, 0); + } + } while_each_thread(g, t); + +unlock: + read_unlock_irqrestore(&tasklist_lock, flags); +free: + for (i = start; i < end; i++) + kfree(ret_stack_list[i]); + return ret; +} + +/* Allocate a return stack for each task */ +static int start_return_tracing(void) +{ + struct ftrace_ret_stack **ret_stack_list; + int ret; + + ret_stack_list = kmalloc(FTRACE_RETSTACK_ALLOC_SIZE * + sizeof(struct ftrace_ret_stack *), + GFP_KERNEL); + + if (!ret_stack_list) + return -ENOMEM; + + do { + ret = alloc_retstack_tasklist(ret_stack_list); + } while (ret == -EAGAIN); + + kfree(ret_stack_list); + return ret; +} + int register_ftrace_return(trace_function_return_t func) { int ret = 0; @@ -1516,7 +1583,12 @@ int register_ftrace_return(trace_function_return_t func) ret = -EBUSY; goto out; } - + atomic_inc(&ftrace_retfunc_active); + ret = start_return_tracing(); + if (ret) { + atomic_dec(&ftrace_retfunc_active); + goto out; + } ftrace_tracing_type = FTRACE_TYPE_RETURN; ftrace_function_return = func; ftrace_startup(); @@ -1530,6 +1602,7 @@ void unregister_ftrace_return(void) { mutex_lock(&ftrace_sysctl_lock); + atomic_dec(&ftrace_retfunc_active); ftrace_function_return = (trace_function_return_t)ftrace_stub; ftrace_shutdown(); /* Restore normal tracing type */ @@ -1537,6 +1610,27 @@ void unregister_ftrace_return(void) mutex_unlock(&ftrace_sysctl_lock); } + +/* Allocate a return stack for newly created task */ +void ftrace_retfunc_init_task(struct task_struct *t) +{ + if (atomic_read(&ftrace_retfunc_active)) { + t->ret_stack = kmalloc(FTRACE_RETFUNC_DEPTH + * sizeof(struct ftrace_ret_stack), + GFP_KERNEL); + if (!t->ret_stack) + return; + t->curr_ret_stack = -1; + atomic_set(&t->trace_overrun, 0); + } else + t->ret_stack = NULL; +} + +void ftrace_retfunc_exit_task(struct task_struct *t) +{ + kfree(t->ret_stack); + t->ret_stack = NULL; +} #endif -- cgit v1.2.3 From 82f60f0bc854aada696f27d863c03bef91f1509d Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Sun, 23 Nov 2008 09:18:56 +0100 Subject: tracing/function-return-tracer: clean up task start/exit callbacks Impact: cleanup Eliminate #ifdefs in core code by using empty inline functions. Signed-off-by: Ingo Molnar --- include/linux/ftrace.h | 3 +++ kernel/exit.c | 2 -- kernel/fork.c | 2 -- kernel/sched.c | 2 -- 4 files changed, 3 insertions(+), 6 deletions(-) (limited to 'kernel/sched.c') diff --git a/include/linux/ftrace.h b/include/linux/ftrace.h index 2ba259b2defa..938ca1942641 100644 --- a/include/linux/ftrace.h +++ b/include/linux/ftrace.h @@ -335,6 +335,9 @@ extern void unregister_ftrace_return(void); extern void ftrace_retfunc_init_task(struct task_struct *t); extern void ftrace_retfunc_exit_task(struct task_struct *t); +#else +static inline void ftrace_retfunc_init_task(struct task_struct *t) { } +static inline void ftrace_retfunc_exit_task(struct task_struct *t) { } #endif #endif /* _LINUX_FTRACE_H */ diff --git a/kernel/exit.c b/kernel/exit.c index b9d446329da1..ef04d03b3286 100644 --- a/kernel/exit.c +++ b/kernel/exit.c @@ -1128,9 +1128,7 @@ NORET_TYPE void do_exit(long code) preempt_disable(); /* causes final put_task_struct in finish_task_switch(). */ tsk->state = TASK_DEAD; -#ifdef CONFIG_FUNCTION_RET_TRACER ftrace_retfunc_exit_task(tsk); -#endif schedule(); BUG(); /* Avoid "noreturn function does return". */ diff --git a/kernel/fork.c b/kernel/fork.c index d1eb30e69ccc..fbf4a4c0a628 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -1270,9 +1270,7 @@ static struct task_struct *copy_process(unsigned long clone_flags, total_forks++; spin_unlock(¤t->sighand->siglock); write_unlock_irq(&tasklist_lock); -#ifdef CONFIG_FUNCTION_RET_TRACER ftrace_retfunc_init_task(p); -#endif proc_fork_connector(p); cgroup_post_fork(p); return p; diff --git a/kernel/sched.c b/kernel/sched.c index fb17205950de..388d9db044ab 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -5901,9 +5901,7 @@ void __cpuinit init_idle(struct task_struct *idle, int cpu) * The idle tasks have their own, simple scheduling class: */ idle->sched_class = &idle_sched_class; -#ifdef CONFIG_FUNCTION_RET_TRACER ftrace_retfunc_init_task(idle); -#endif } /* -- cgit v1.2.3 From ea6f18ed5a1531caf678374f30a0990c9e6742f3 Mon Sep 17 00:00:00 2001 From: Mike Travis Date: Tue, 25 Nov 2008 02:35:02 +1030 Subject: sched: reduce stack size requirements in kernel/sched.c Impact: cleanup * use node_to_cpumask_ptr in place of node_to_cpumask to reduce stack requirements in sched.c Signed-off-by: Mike Travis Signed-off-by: Rusty Russell Signed-off-by: Ingo Molnar --- kernel/sched.c | 13 +++++++------ 1 file changed, 7 insertions(+), 6 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index bb827651558e..dd22cec499b8 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -6110,8 +6110,9 @@ static void move_task_off_dead_cpu(int dead_cpu, struct task_struct *p) do { /* On same node? */ - mask = node_to_cpumask(cpu_to_node(dead_cpu)); - cpus_and(mask, mask, p->cpus_allowed); + node_to_cpumask_ptr(pnodemask, cpu_to_node(dead_cpu)); + + cpus_and(mask, *pnodemask, p->cpus_allowed); dest_cpu = any_online_cpu(mask); /* On any allowed CPU? */ @@ -7098,9 +7099,9 @@ static int cpu_to_allnodes_group(int cpu, const cpumask_t *cpu_map, struct sched_group **sg, cpumask_t *nodemask) { int group; + node_to_cpumask_ptr(pnodemask, cpu_to_node(cpu)); - *nodemask = node_to_cpumask(cpu_to_node(cpu)); - cpus_and(*nodemask, *nodemask, *cpu_map); + cpus_and(*nodemask, *pnodemask, *cpu_map); group = first_cpu(*nodemask); if (sg) @@ -7150,9 +7151,9 @@ static void free_sched_groups(const cpumask_t *cpu_map, cpumask_t *nodemask) for (i = 0; i < nr_node_ids; i++) { struct sched_group *oldsg, *sg = sched_group_nodes[i]; + node_to_cpumask_ptr(pnodemask, i); - *nodemask = node_to_cpumask(i); - cpus_and(*nodemask, *nodemask, *cpu_map); + cpus_and(*nodemask, *pnodemask, *cpu_map); if (cpus_empty(*nodemask)) continue; -- cgit v1.2.3 From abcd083a1a658d2bc1f7fced02632bfe03918002 Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Tue, 25 Nov 2008 02:35:02 +1030 Subject: sched: convert sched.c from for_each_cpu_mask to for_each_cpu. Impact: trivial API conversion This is a simple conversion, but note that for_each_cpu() terminates with i >= nr_cpu_ids, not i == NR_CPUS like for_each_cpu_mask() did. I don't convert all of them: sd->span changes in a later patch, so change those iterators there rather than here. Signed-off-by: Rusty Russell Signed-off-by: Ingo Molnar --- kernel/sched.c | 36 ++++++++++++++++++------------------ 1 file changed, 18 insertions(+), 18 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index dd22cec499b8..e59978eead17 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -2061,7 +2061,7 @@ find_idlest_group(struct sched_domain *sd, struct task_struct *p, int this_cpu) /* Tally up the load of all CPUs in the group */ avg_load = 0; - for_each_cpu_mask_nr(i, group->cpumask) { + for_each_cpu(i, &group->cpumask) { /* Bias balancing toward cpus of our domain */ if (local_group) load = source_load(i, load_idx); @@ -2103,7 +2103,7 @@ find_idlest_cpu(struct sched_group *group, struct task_struct *p, int this_cpu, /* Traverse only the allowed CPUs */ cpus_and(*tmp, group->cpumask, p->cpus_allowed); - for_each_cpu_mask_nr(i, *tmp) { + for_each_cpu(i, tmp) { load = weighted_cpuload(i); if (load < min_load || (load == min_load && i == this_cpu)) { @@ -3121,7 +3121,7 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, max_cpu_load = 0; min_cpu_load = ~0UL; - for_each_cpu_mask_nr(i, group->cpumask) { + for_each_cpu(i, &group->cpumask) { struct rq *rq; if (!cpu_isset(i, *cpus)) @@ -3400,7 +3400,7 @@ find_busiest_queue(struct sched_group *group, enum cpu_idle_type idle, unsigned long max_load = 0; int i; - for_each_cpu_mask_nr(i, group->cpumask) { + for_each_cpu(i, &group->cpumask) { unsigned long wl; if (!cpu_isset(i, *cpus)) @@ -3942,7 +3942,7 @@ static void run_rebalance_domains(struct softirq_action *h) int balance_cpu; cpu_clear(this_cpu, cpus); - for_each_cpu_mask_nr(balance_cpu, cpus) { + for_each_cpu(balance_cpu, &cpus) { /* * If this cpu gets work to do, stop the load balancing * work being done for other cpus. Next load @@ -6906,7 +6906,7 @@ init_sched_build_groups(const cpumask_t *span, const cpumask_t *cpu_map, cpus_clear(*covered); - for_each_cpu_mask_nr(i, *span) { + for_each_cpu(i, span) { struct sched_group *sg; int group = group_fn(i, cpu_map, &sg, tmpmask); int j; @@ -6917,7 +6917,7 @@ init_sched_build_groups(const cpumask_t *span, const cpumask_t *cpu_map, cpus_clear(sg->cpumask); sg->__cpu_power = 0; - for_each_cpu_mask_nr(j, *span) { + for_each_cpu(j, span) { if (group_fn(j, cpu_map, NULL, tmpmask) != group) continue; @@ -7117,7 +7117,7 @@ static void init_numa_sched_groups_power(struct sched_group *group_head) if (!sg) return; do { - for_each_cpu_mask_nr(j, sg->cpumask) { + for_each_cpu(j, &sg->cpumask) { struct sched_domain *sd; sd = &per_cpu(phys_domains, j); @@ -7142,7 +7142,7 @@ static void free_sched_groups(const cpumask_t *cpu_map, cpumask_t *nodemask) { int cpu, i; - for_each_cpu_mask_nr(cpu, *cpu_map) { + for_each_cpu(cpu, cpu_map) { struct sched_group **sched_group_nodes = sched_group_nodes_bycpu[cpu]; @@ -7396,7 +7396,7 @@ static int __build_sched_domains(const cpumask_t *cpu_map, /* * Set up domains for cpus specified by the cpu_map. */ - for_each_cpu_mask_nr(i, *cpu_map) { + for_each_cpu(i, cpu_map) { struct sched_domain *sd = NULL, *p; SCHED_CPUMASK_VAR(nodemask, allmasks); @@ -7463,7 +7463,7 @@ static int __build_sched_domains(const cpumask_t *cpu_map, #ifdef CONFIG_SCHED_SMT /* Set up CPU (sibling) groups */ - for_each_cpu_mask_nr(i, *cpu_map) { + for_each_cpu(i, cpu_map) { SCHED_CPUMASK_VAR(this_sibling_map, allmasks); SCHED_CPUMASK_VAR(send_covered, allmasks); @@ -7480,7 +7480,7 @@ static int __build_sched_domains(const cpumask_t *cpu_map, #ifdef CONFIG_SCHED_MC /* Set up multi-core groups */ - for_each_cpu_mask_nr(i, *cpu_map) { + for_each_cpu(i, cpu_map) { SCHED_CPUMASK_VAR(this_core_map, allmasks); SCHED_CPUMASK_VAR(send_covered, allmasks); @@ -7547,7 +7547,7 @@ static int __build_sched_domains(const cpumask_t *cpu_map, goto error; } sched_group_nodes[i] = sg; - for_each_cpu_mask_nr(j, *nodemask) { + for_each_cpu(j, nodemask) { struct sched_domain *sd; sd = &per_cpu(node_domains, j); @@ -7593,21 +7593,21 @@ static int __build_sched_domains(const cpumask_t *cpu_map, /* Calculate CPU power for physical packages and nodes */ #ifdef CONFIG_SCHED_SMT - for_each_cpu_mask_nr(i, *cpu_map) { + for_each_cpu(i, cpu_map) { struct sched_domain *sd = &per_cpu(cpu_domains, i); init_sched_groups_power(i, sd); } #endif #ifdef CONFIG_SCHED_MC - for_each_cpu_mask_nr(i, *cpu_map) { + for_each_cpu(i, cpu_map) { struct sched_domain *sd = &per_cpu(core_domains, i); init_sched_groups_power(i, sd); } #endif - for_each_cpu_mask_nr(i, *cpu_map) { + for_each_cpu(i, cpu_map) { struct sched_domain *sd = &per_cpu(phys_domains, i); init_sched_groups_power(i, sd); @@ -7627,7 +7627,7 @@ static int __build_sched_domains(const cpumask_t *cpu_map, #endif /* Attach the domains */ - for_each_cpu_mask_nr(i, *cpu_map) { + for_each_cpu(i, cpu_map) { struct sched_domain *sd; #ifdef CONFIG_SCHED_SMT sd = &per_cpu(cpu_domains, i); @@ -7709,7 +7709,7 @@ static void detach_destroy_domains(const cpumask_t *cpu_map) cpumask_t tmpmask; int i; - for_each_cpu_mask_nr(i, *cpu_map) + for_each_cpu(i, cpu_map) cpu_attach_domain(NULL, &def_root_domain, i); synchronize_sched(); arch_destroy_sched_domains(cpu_map, &tmpmask); -- cgit v1.2.3 From 3404c8d97c2d3eb87b1bf4aadad957bfb5235b14 Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Tue, 25 Nov 2008 02:35:03 +1030 Subject: sched: get rid of boutique sched.c allocations, use cpumask_var_t. Impact: use new general API Using lots of allocs rather than one big alloc is less efficient, but who cares for this setup function? Signed-off-by: Rusty Russell Signed-off-by: Mike Travis Acked-by: Ingo Molnar Signed-off-by: Ingo Molnar --- kernel/sched.c | 139 +++++++++++++++++++++++---------------------------------- 1 file changed, 55 insertions(+), 84 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index e59978eead17..0dc9d5752d68 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -7263,48 +7263,6 @@ SD_INIT_FUNC(CPU) SD_INIT_FUNC(MC) #endif -/* - * To minimize stack usage kmalloc room for cpumasks and share the - * space as the usage in build_sched_domains() dictates. Used only - * if the amount of space is significant. - */ -struct allmasks { - cpumask_t tmpmask; /* make this one first */ - union { - cpumask_t nodemask; - cpumask_t this_sibling_map; - cpumask_t this_core_map; - }; - cpumask_t send_covered; - -#ifdef CONFIG_NUMA - cpumask_t domainspan; - cpumask_t covered; - cpumask_t notcovered; -#endif -}; - -#if NR_CPUS > 128 -#define SCHED_CPUMASK_DECLARE(v) struct allmasks *v -static inline void sched_cpumask_alloc(struct allmasks **masks) -{ - *masks = kmalloc(sizeof(**masks), GFP_KERNEL); -} -static inline void sched_cpumask_free(struct allmasks *masks) -{ - kfree(masks); -} -#else -#define SCHED_CPUMASK_DECLARE(v) struct allmasks _v, *v = &_v -static inline void sched_cpumask_alloc(struct allmasks **masks) -{ } -static inline void sched_cpumask_free(struct allmasks *masks) -{ } -#endif - -#define SCHED_CPUMASK_VAR(v, a) cpumask_t *v = (cpumask_t *) \ - ((unsigned long)(a) + offsetof(struct allmasks, v)) - static int default_relax_domain_level = -1; static int __init setup_relax_domain_level(char *str) @@ -7347,14 +7305,35 @@ static void set_domain_attribute(struct sched_domain *sd, static int __build_sched_domains(const cpumask_t *cpu_map, struct sched_domain_attr *attr) { - int i; + int i, err = -ENOMEM; struct root_domain *rd; - SCHED_CPUMASK_DECLARE(allmasks); - cpumask_t *tmpmask; + cpumask_var_t nodemask, this_sibling_map, this_core_map, send_covered, + tmpmask; #ifdef CONFIG_NUMA + cpumask_var_t domainspan, covered, notcovered; struct sched_group **sched_group_nodes = NULL; int sd_allnodes = 0; + if (!alloc_cpumask_var(&domainspan, GFP_KERNEL)) + goto out; + if (!alloc_cpumask_var(&covered, GFP_KERNEL)) + goto free_domainspan; + if (!alloc_cpumask_var(¬covered, GFP_KERNEL)) + goto free_covered; +#endif + + if (!alloc_cpumask_var(&nodemask, GFP_KERNEL)) + goto free_notcovered; + if (!alloc_cpumask_var(&this_sibling_map, GFP_KERNEL)) + goto free_nodemask; + if (!alloc_cpumask_var(&this_core_map, GFP_KERNEL)) + goto free_this_sibling_map; + if (!alloc_cpumask_var(&send_covered, GFP_KERNEL)) + goto free_this_core_map; + if (!alloc_cpumask_var(&tmpmask, GFP_KERNEL)) + goto free_send_covered; + +#ifdef CONFIG_NUMA /* * Allocate the per-node list of sched groups */ @@ -7362,33 +7341,16 @@ static int __build_sched_domains(const cpumask_t *cpu_map, GFP_KERNEL); if (!sched_group_nodes) { printk(KERN_WARNING "Can not alloc sched group node list\n"); - return -ENOMEM; + goto free_tmpmask; } #endif rd = alloc_rootdomain(); if (!rd) { printk(KERN_WARNING "Cannot alloc root domain\n"); -#ifdef CONFIG_NUMA - kfree(sched_group_nodes); -#endif - return -ENOMEM; + goto free_sched_groups; } - /* get space for all scratch cpumask variables */ - sched_cpumask_alloc(&allmasks); - if (!allmasks) { - printk(KERN_WARNING "Cannot alloc cpumask array\n"); - kfree(rd); -#ifdef CONFIG_NUMA - kfree(sched_group_nodes); -#endif - return -ENOMEM; - } - - tmpmask = (cpumask_t *)allmasks; - - #ifdef CONFIG_NUMA sched_group_nodes_bycpu[first_cpu(*cpu_map)] = sched_group_nodes; #endif @@ -7398,7 +7360,6 @@ static int __build_sched_domains(const cpumask_t *cpu_map, */ for_each_cpu(i, cpu_map) { struct sched_domain *sd = NULL, *p; - SCHED_CPUMASK_VAR(nodemask, allmasks); *nodemask = node_to_cpumask(cpu_to_node(i)); cpus_and(*nodemask, *nodemask, *cpu_map); @@ -7464,9 +7425,6 @@ static int __build_sched_domains(const cpumask_t *cpu_map, #ifdef CONFIG_SCHED_SMT /* Set up CPU (sibling) groups */ for_each_cpu(i, cpu_map) { - SCHED_CPUMASK_VAR(this_sibling_map, allmasks); - SCHED_CPUMASK_VAR(send_covered, allmasks); - *this_sibling_map = per_cpu(cpu_sibling_map, i); cpus_and(*this_sibling_map, *this_sibling_map, *cpu_map); if (i != first_cpu(*this_sibling_map)) @@ -7481,9 +7439,6 @@ static int __build_sched_domains(const cpumask_t *cpu_map, #ifdef CONFIG_SCHED_MC /* Set up multi-core groups */ for_each_cpu(i, cpu_map) { - SCHED_CPUMASK_VAR(this_core_map, allmasks); - SCHED_CPUMASK_VAR(send_covered, allmasks); - *this_core_map = cpu_coregroup_map(i); cpus_and(*this_core_map, *this_core_map, *cpu_map); if (i != first_cpu(*this_core_map)) @@ -7497,9 +7452,6 @@ static int __build_sched_domains(const cpumask_t *cpu_map, /* Set up physical groups */ for (i = 0; i < nr_node_ids; i++) { - SCHED_CPUMASK_VAR(nodemask, allmasks); - SCHED_CPUMASK_VAR(send_covered, allmasks); - *nodemask = node_to_cpumask(i); cpus_and(*nodemask, *nodemask, *cpu_map); if (cpus_empty(*nodemask)) @@ -7513,8 +7465,6 @@ static int __build_sched_domains(const cpumask_t *cpu_map, #ifdef CONFIG_NUMA /* Set up node groups */ if (sd_allnodes) { - SCHED_CPUMASK_VAR(send_covered, allmasks); - init_sched_build_groups(cpu_map, cpu_map, &cpu_to_allnodes_group, send_covered, tmpmask); @@ -7523,9 +7473,6 @@ static int __build_sched_domains(const cpumask_t *cpu_map, for (i = 0; i < nr_node_ids; i++) { /* Set up node groups */ struct sched_group *sg, *prev; - SCHED_CPUMASK_VAR(nodemask, allmasks); - SCHED_CPUMASK_VAR(domainspan, allmasks); - SCHED_CPUMASK_VAR(covered, allmasks); int j; *nodemask = node_to_cpumask(i); @@ -7560,7 +7507,6 @@ static int __build_sched_domains(const cpumask_t *cpu_map, prev = sg; for (j = 0; j < nr_node_ids; j++) { - SCHED_CPUMASK_VAR(notcovered, allmasks); int n = (i + j) % nr_node_ids; node_to_cpumask_ptr(pnodemask, n); @@ -7639,15 +7585,40 @@ static int __build_sched_domains(const cpumask_t *cpu_map, cpu_attach_domain(sd, rd, i); } - sched_cpumask_free(allmasks); - return 0; + err = 0; + +free_tmpmask: + free_cpumask_var(tmpmask); +free_send_covered: + free_cpumask_var(send_covered); +free_this_core_map: + free_cpumask_var(this_core_map); +free_this_sibling_map: + free_cpumask_var(this_sibling_map); +free_nodemask: + free_cpumask_var(nodemask); +free_notcovered: +#ifdef CONFIG_NUMA + free_cpumask_var(notcovered); +free_covered: + free_cpumask_var(covered); +free_domainspan: + free_cpumask_var(domainspan); +out: +#endif + return err; + +free_sched_groups: +#ifdef CONFIG_NUMA + kfree(sched_group_nodes); +#endif + goto free_tmpmask; #ifdef CONFIG_NUMA error: free_sched_groups(cpu_map, tmpmask); - sched_cpumask_free(allmasks); kfree(rd); - return -ENOMEM; + goto free_tmpmask; #endif } -- cgit v1.2.3 From 1e5ce4f4a755ee498bd9217dae26143afa0d8f31 Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Tue, 25 Nov 2008 02:35:03 +1030 Subject: sched: remove any_online_cpu() Impact: use new API any_online_cpu() is a good name, but it takes a cpumask_t, not a pointer. There are several places where any_online_cpu() doesn't really want a mask arg at all. Replace all callers with cpumask_any() and cpumask_any_and(). Signed-off-by: Rusty Russell Signed-off-by: Mike Travis Signed-off-by: Ingo Molnar --- kernel/sched.c | 14 ++++++++------ 1 file changed, 8 insertions(+), 6 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index 0dc9d5752d68..a2de33d05340 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -5964,7 +5964,7 @@ int set_cpus_allowed_ptr(struct task_struct *p, const cpumask_t *new_mask) if (cpu_isset(task_cpu(p), *new_mask)) goto out; - if (migrate_task(p, any_online_cpu(*new_mask), &req)) { + if (migrate_task(p, cpumask_any_and(cpu_online_mask, new_mask), &req)) { /* Need help from migration thread: drop lock and wait. */ task_rq_unlock(rq, &flags); wake_up_process(rq->migration_thread); @@ -6113,11 +6113,12 @@ static void move_task_off_dead_cpu(int dead_cpu, struct task_struct *p) node_to_cpumask_ptr(pnodemask, cpu_to_node(dead_cpu)); cpus_and(mask, *pnodemask, p->cpus_allowed); - dest_cpu = any_online_cpu(mask); + dest_cpu = cpumask_any_and(cpu_online_mask, &mask); /* On any allowed CPU? */ if (dest_cpu >= nr_cpu_ids) - dest_cpu = any_online_cpu(p->cpus_allowed); + dest_cpu = cpumask_any_and(cpu_online_mask, + &p->cpus_allowed); /* No more Mr. Nice Guy. */ if (dest_cpu >= nr_cpu_ids) { @@ -6133,7 +6134,8 @@ static void move_task_off_dead_cpu(int dead_cpu, struct task_struct *p) */ rq = task_rq_lock(p, &flags); p->cpus_allowed = cpus_allowed; - dest_cpu = any_online_cpu(p->cpus_allowed); + dest_cpu = cpumask_any_and(cpu_online_mask, + &p->cpus_allowed); task_rq_unlock(rq, &flags); /* @@ -6159,7 +6161,7 @@ static void move_task_off_dead_cpu(int dead_cpu, struct task_struct *p) */ static void migrate_nr_uninterruptible(struct rq *rq_src) { - struct rq *rq_dest = cpu_rq(any_online_cpu(*CPU_MASK_ALL_PTR)); + struct rq *rq_dest = cpu_rq(cpumask_any(cpu_online_mask)); unsigned long flags; local_irq_save(flags); @@ -6524,7 +6526,7 @@ migration_call(struct notifier_block *nfb, unsigned long action, void *hcpu) break; /* Unbind it from offline cpu so it can run. Fall thru. */ kthread_bind(cpu_rq(cpu)->migration_thread, - any_online_cpu(cpu_online_map)); + cpumask_any(cpu_online_mask)); kthread_stop(cpu_rq(cpu)->migration_thread); cpu_rq(cpu)->migration_thread = NULL; break; -- cgit v1.2.3 From 758b2cdc6f6a22c702bd8f2344382fb1270b2161 Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Tue, 25 Nov 2008 02:35:04 +1030 Subject: sched: wrap sched_group and sched_domain cpumask accesses. Impact: trivial wrap of member accesses This eases the transition in the next patch. We also get rid of a temporary cpumask in find_idlest_cpu() thanks to for_each_cpu_and, and sched_balance_self() due to getting weight before setting sd to NULL. Signed-off-by: Rusty Russell Signed-off-by: Ingo Molnar --- include/linux/sched.h | 10 +++++ kernel/sched.c | 114 ++++++++++++++++++++++++-------------------------- kernel/sched_fair.c | 10 ++--- kernel/sched_rt.c | 3 +- kernel/sched_stats.h | 3 +- 5 files changed, 73 insertions(+), 67 deletions(-) (limited to 'kernel/sched.c') diff --git a/include/linux/sched.h b/include/linux/sched.h index 4ce5c603c51a..2b95aa9f779b 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -786,6 +786,11 @@ struct sched_group { u32 reciprocal_cpu_power; }; +static inline struct cpumask *sched_group_cpus(struct sched_group *sg) +{ + return &sg->cpumask; +} + enum sched_domain_level { SD_LV_NONE = 0, SD_LV_SIBLING, @@ -866,6 +871,11 @@ struct sched_domain { #endif }; +static inline struct cpumask *sched_domain_span(struct sched_domain *sd) +{ + return &sd->span; +} + extern void partition_sched_domains(int ndoms_new, cpumask_t *doms_new, struct sched_domain_attr *dattr_new); extern int arch_reinit_sched_domains(void); diff --git a/kernel/sched.c b/kernel/sched.c index a2de33d05340..575f38acf4da 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -1501,7 +1501,7 @@ static int tg_shares_up(struct task_group *tg, void *data) struct sched_domain *sd = data; int i; - for_each_cpu_mask(i, sd->span) { + for_each_cpu(i, sched_domain_span(sd)) { /* * If there are currently no tasks on the cpu pretend there * is one of average load so that when a new task gets to @@ -1522,7 +1522,7 @@ static int tg_shares_up(struct task_group *tg, void *data) if (!sd->parent || !(sd->parent->flags & SD_LOAD_BALANCE)) shares = tg->shares; - for_each_cpu_mask(i, sd->span) + for_each_cpu(i, sched_domain_span(sd)) update_group_shares_cpu(tg, i, shares, rq_weight); return 0; @@ -2053,15 +2053,17 @@ find_idlest_group(struct sched_domain *sd, struct task_struct *p, int this_cpu) int i; /* Skip over this group if it has no CPUs allowed */ - if (!cpus_intersects(group->cpumask, p->cpus_allowed)) + if (!cpumask_intersects(sched_group_cpus(group), + &p->cpus_allowed)) continue; - local_group = cpu_isset(this_cpu, group->cpumask); + local_group = cpumask_test_cpu(this_cpu, + sched_group_cpus(group)); /* Tally up the load of all CPUs in the group */ avg_load = 0; - for_each_cpu(i, &group->cpumask) { + for_each_cpu(i, sched_group_cpus(group)) { /* Bias balancing toward cpus of our domain */ if (local_group) load = source_load(i, load_idx); @@ -2093,17 +2095,14 @@ find_idlest_group(struct sched_domain *sd, struct task_struct *p, int this_cpu) * find_idlest_cpu - find the idlest cpu among the cpus in group. */ static int -find_idlest_cpu(struct sched_group *group, struct task_struct *p, int this_cpu, - cpumask_t *tmp) +find_idlest_cpu(struct sched_group *group, struct task_struct *p, int this_cpu) { unsigned long load, min_load = ULONG_MAX; int idlest = -1; int i; /* Traverse only the allowed CPUs */ - cpus_and(*tmp, group->cpumask, p->cpus_allowed); - - for_each_cpu(i, tmp) { + for_each_cpu_and(i, sched_group_cpus(group), &p->cpus_allowed) { load = weighted_cpuload(i); if (load < min_load || (load == min_load && i == this_cpu)) { @@ -2145,7 +2144,6 @@ static int sched_balance_self(int cpu, int flag) update_shares(sd); while (sd) { - cpumask_t span, tmpmask; struct sched_group *group; int new_cpu, weight; @@ -2154,14 +2152,13 @@ static int sched_balance_self(int cpu, int flag) continue; } - span = sd->span; group = find_idlest_group(sd, t, cpu); if (!group) { sd = sd->child; continue; } - new_cpu = find_idlest_cpu(group, t, cpu, &tmpmask); + new_cpu = find_idlest_cpu(group, t, cpu); if (new_cpu == -1 || new_cpu == cpu) { /* Now try balancing at a lower domain level of cpu */ sd = sd->child; @@ -2170,10 +2167,10 @@ static int sched_balance_self(int cpu, int flag) /* Now try balancing at a lower domain level of new_cpu */ cpu = new_cpu; + weight = cpumask_weight(sched_domain_span(sd)); sd = NULL; - weight = cpus_weight(span); for_each_domain(cpu, tmp) { - if (weight <= cpus_weight(tmp->span)) + if (weight <= cpumask_weight(sched_domain_span(tmp))) break; if (tmp->flags & flag) sd = tmp; @@ -2218,7 +2215,7 @@ static int try_to_wake_up(struct task_struct *p, unsigned int state, int sync) cpu = task_cpu(p); for_each_domain(this_cpu, sd) { - if (cpu_isset(cpu, sd->span)) { + if (cpumask_test_cpu(cpu, sched_domain_span(sd))) { update_shares(sd); break; } @@ -2266,7 +2263,7 @@ static int try_to_wake_up(struct task_struct *p, unsigned int state, int sync) else { struct sched_domain *sd; for_each_domain(this_cpu, sd) { - if (cpu_isset(cpu, sd->span)) { + if (cpumask_test_cpu(cpu, sched_domain_span(sd))) { schedstat_inc(sd, ttwu_wake_remote); break; } @@ -3109,10 +3106,11 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, unsigned long sum_avg_load_per_task; unsigned long avg_load_per_task; - local_group = cpu_isset(this_cpu, group->cpumask); + local_group = cpumask_test_cpu(this_cpu, + sched_group_cpus(group)); if (local_group) - balance_cpu = first_cpu(group->cpumask); + balance_cpu = cpumask_first(sched_group_cpus(group)); /* Tally up the load of all CPUs in the group */ sum_weighted_load = sum_nr_running = avg_load = 0; @@ -3121,13 +3119,8 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, max_cpu_load = 0; min_cpu_load = ~0UL; - for_each_cpu(i, &group->cpumask) { - struct rq *rq; - - if (!cpu_isset(i, *cpus)) - continue; - - rq = cpu_rq(i); + for_each_cpu_and(i, sched_group_cpus(group), cpus) { + struct rq *rq = cpu_rq(i); if (*sd_idle && rq->nr_running) *sd_idle = 0; @@ -3238,8 +3231,8 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, */ if ((sum_nr_running < min_nr_running) || (sum_nr_running == min_nr_running && - first_cpu(group->cpumask) < - first_cpu(group_min->cpumask))) { + cpumask_first(sched_group_cpus(group)) < + cpumask_first(sched_group_cpus(group_min)))) { group_min = group; min_nr_running = sum_nr_running; min_load_per_task = sum_weighted_load / @@ -3254,8 +3247,8 @@ find_busiest_group(struct sched_domain *sd, int this_cpu, if (sum_nr_running <= group_capacity - 1) { if (sum_nr_running > leader_nr_running || (sum_nr_running == leader_nr_running && - first_cpu(group->cpumask) > - first_cpu(group_leader->cpumask))) { + cpumask_first(sched_group_cpus(group)) > + cpumask_first(sched_group_cpus(group_leader)))) { group_leader = group; leader_nr_running = sum_nr_running; } @@ -3400,7 +3393,7 @@ find_busiest_queue(struct sched_group *group, enum cpu_idle_type idle, unsigned long max_load = 0; int i; - for_each_cpu(i, &group->cpumask) { + for_each_cpu(i, sched_group_cpus(group)) { unsigned long wl; if (!cpu_isset(i, *cpus)) @@ -3746,7 +3739,7 @@ static void active_load_balance(struct rq *busiest_rq, int busiest_cpu) /* Search for an sd spanning us and the target CPU. */ for_each_domain(target_cpu, sd) { if ((sd->flags & SD_LOAD_BALANCE) && - cpu_isset(busiest_cpu, sd->span)) + cpumask_test_cpu(busiest_cpu, sched_domain_span(sd))) break; } @@ -6618,7 +6611,7 @@ static int sched_domain_debug_one(struct sched_domain *sd, int cpu, int level, struct sched_group *group = sd->groups; char str[256]; - cpulist_scnprintf(str, sizeof(str), sd->span); + cpulist_scnprintf(str, sizeof(str), *sched_domain_span(sd)); cpus_clear(*groupmask); printk(KERN_DEBUG "%*s domain %d: ", level, "", level); @@ -6633,11 +6626,11 @@ static int sched_domain_debug_one(struct sched_domain *sd, int cpu, int level, printk(KERN_CONT "span %s level %s\n", str, sd->name); - if (!cpu_isset(cpu, sd->span)) { + if (!cpumask_test_cpu(cpu, sched_domain_span(sd))) { printk(KERN_ERR "ERROR: domain->span does not contain " "CPU%d\n", cpu); } - if (!cpu_isset(cpu, group->cpumask)) { + if (!cpumask_test_cpu(cpu, sched_group_cpus(group))) { printk(KERN_ERR "ERROR: domain->groups does not contain" " CPU%d\n", cpu); } @@ -6657,31 +6650,32 @@ static int sched_domain_debug_one(struct sched_domain *sd, int cpu, int level, break; } - if (!cpus_weight(group->cpumask)) { + if (!cpumask_weight(sched_group_cpus(group))) { printk(KERN_CONT "\n"); printk(KERN_ERR "ERROR: empty group\n"); break; } - if (cpus_intersects(*groupmask, group->cpumask)) { + if (cpumask_intersects(groupmask, sched_group_cpus(group))) { printk(KERN_CONT "\n"); printk(KERN_ERR "ERROR: repeated CPUs\n"); break; } - cpus_or(*groupmask, *groupmask, group->cpumask); + cpumask_or(groupmask, groupmask, sched_group_cpus(group)); - cpulist_scnprintf(str, sizeof(str), group->cpumask); + cpulist_scnprintf(str, sizeof(str), *sched_group_cpus(group)); printk(KERN_CONT " %s", str); group = group->next; } while (group != sd->groups); printk(KERN_CONT "\n"); - if (!cpus_equal(sd->span, *groupmask)) + if (!cpumask_equal(sched_domain_span(sd), groupmask)) printk(KERN_ERR "ERROR: groups don't span domain->span\n"); - if (sd->parent && !cpus_subset(*groupmask, sd->parent->span)) + if (sd->parent && + !cpumask_subset(groupmask, sched_domain_span(sd->parent))) printk(KERN_ERR "ERROR: parent span is not a superset " "of domain->span\n"); return 0; @@ -6721,7 +6715,7 @@ static void sched_domain_debug(struct sched_domain *sd, int cpu) static int sd_degenerate(struct sched_domain *sd) { - if (cpus_weight(sd->span) == 1) + if (cpumask_weight(sched_domain_span(sd)) == 1) return 1; /* Following flags need at least 2 groups */ @@ -6752,7 +6746,7 @@ sd_parent_degenerate(struct sched_domain *sd, struct sched_domain *parent) if (sd_degenerate(parent)) return 1; - if (!cpus_equal(sd->span, parent->span)) + if (!cpumask_equal(sched_domain_span(sd), sched_domain_span(parent))) return 0; /* Does parent contain flags not in child? */ @@ -6913,10 +6907,10 @@ init_sched_build_groups(const cpumask_t *span, const cpumask_t *cpu_map, int group = group_fn(i, cpu_map, &sg, tmpmask); int j; - if (cpu_isset(i, *covered)) + if (cpumask_test_cpu(i, covered)) continue; - cpus_clear(sg->cpumask); + cpumask_clear(sched_group_cpus(sg)); sg->__cpu_power = 0; for_each_cpu(j, span) { @@ -6924,7 +6918,7 @@ init_sched_build_groups(const cpumask_t *span, const cpumask_t *cpu_map, continue; cpu_set(j, *covered); - cpu_set(j, sg->cpumask); + cpumask_set_cpu(j, sched_group_cpus(sg)); } if (!first) first = sg; @@ -7119,11 +7113,11 @@ static void init_numa_sched_groups_power(struct sched_group *group_head) if (!sg) return; do { - for_each_cpu(j, &sg->cpumask) { + for_each_cpu(j, sched_group_cpus(sg)) { struct sched_domain *sd; sd = &per_cpu(phys_domains, j); - if (j != first_cpu(sd->groups->cpumask)) { + if (j != cpumask_first(sched_group_cpus(sd->groups))) { /* * Only add "power" once for each * physical package. @@ -7200,7 +7194,7 @@ static void init_sched_groups_power(int cpu, struct sched_domain *sd) WARN_ON(!sd || !sd->groups); - if (cpu != first_cpu(sd->groups->cpumask)) + if (cpu != cpumask_first(sched_group_cpus(sd->groups))) return; child = sd->child; @@ -7372,7 +7366,7 @@ static int __build_sched_domains(const cpumask_t *cpu_map, sd = &per_cpu(allnodes_domains, i); SD_INIT(sd, ALLNODES); set_domain_attribute(sd, attr); - sd->span = *cpu_map; + cpumask_copy(sched_domain_span(sd), cpu_map); cpu_to_allnodes_group(i, cpu_map, &sd->groups, tmpmask); p = sd; sd_allnodes = 1; @@ -7382,18 +7376,19 @@ static int __build_sched_domains(const cpumask_t *cpu_map, sd = &per_cpu(node_domains, i); SD_INIT(sd, NODE); set_domain_attribute(sd, attr); - sched_domain_node_span(cpu_to_node(i), &sd->span); + sched_domain_node_span(cpu_to_node(i), sched_domain_span(sd)); sd->parent = p; if (p) p->child = sd; - cpus_and(sd->span, sd->span, *cpu_map); + cpumask_and(sched_domain_span(sd), + sched_domain_span(sd), cpu_map); #endif p = sd; sd = &per_cpu(phys_domains, i); SD_INIT(sd, CPU); set_domain_attribute(sd, attr); - sd->span = *nodemask; + cpumask_copy(sched_domain_span(sd), nodemask); sd->parent = p; if (p) p->child = sd; @@ -7404,8 +7399,9 @@ static int __build_sched_domains(const cpumask_t *cpu_map, sd = &per_cpu(core_domains, i); SD_INIT(sd, MC); set_domain_attribute(sd, attr); - sd->span = cpu_coregroup_map(i); - cpus_and(sd->span, sd->span, *cpu_map); + *sched_domain_span(sd) = cpu_coregroup_map(i); + cpumask_and(sched_domain_span(sd), + sched_domain_span(sd), cpu_map); sd->parent = p; p->child = sd; cpu_to_core_group(i, cpu_map, &sd->groups, tmpmask); @@ -7416,8 +7412,8 @@ static int __build_sched_domains(const cpumask_t *cpu_map, sd = &per_cpu(cpu_domains, i); SD_INIT(sd, SIBLING); set_domain_attribute(sd, attr); - sd->span = per_cpu(cpu_sibling_map, i); - cpus_and(sd->span, sd->span, *cpu_map); + cpumask_and(sched_domain_span(sd), + &per_cpu(cpu_sibling_map, i), cpu_map); sd->parent = p; p->child = sd; cpu_to_cpu_group(i, cpu_map, &sd->groups, tmpmask); @@ -7503,7 +7499,7 @@ static int __build_sched_domains(const cpumask_t *cpu_map, sd->groups = sg; } sg->__cpu_power = 0; - sg->cpumask = *nodemask; + cpumask_copy(sched_group_cpus(sg), nodemask); sg->next = sg; cpus_or(*covered, *covered, *nodemask); prev = sg; @@ -7530,7 +7526,7 @@ static int __build_sched_domains(const cpumask_t *cpu_map, goto error; } sg->__cpu_power = 0; - sg->cpumask = *tmpmask; + cpumask_copy(sched_group_cpus(sg), tmpmask); sg->next = prev->next; cpus_or(*covered, *covered, *tmpmask); prev->next = sg; diff --git a/kernel/sched_fair.c b/kernel/sched_fair.c index 98345e45b059..bba00402ed90 100644 --- a/kernel/sched_fair.c +++ b/kernel/sched_fair.c @@ -1024,7 +1024,6 @@ static void yield_task_fair(struct rq *rq) #if defined(ARCH_HAS_SCHED_WAKE_IDLE) static int wake_idle(int cpu, struct task_struct *p) { - cpumask_t tmp; struct sched_domain *sd; int i; @@ -1044,10 +1043,9 @@ static int wake_idle(int cpu, struct task_struct *p) if ((sd->flags & SD_WAKE_IDLE) || ((sd->flags & SD_WAKE_IDLE_FAR) && !task_hot(p, task_rq(p)->clock, sd))) { - cpus_and(tmp, sd->span, p->cpus_allowed); - cpus_and(tmp, tmp, cpu_active_map); - for_each_cpu_mask_nr(i, tmp) { - if (idle_cpu(i)) { + for_each_cpu_and(i, sched_domain_span(sd), + &p->cpus_allowed) { + if (cpu_active(i) && idle_cpu(i)) { if (i != task_cpu(p)) { schedstat_inc(p, se.nr_wakeups_idle); @@ -1240,7 +1238,7 @@ static int select_task_rq_fair(struct task_struct *p, int sync) * this_cpu and prev_cpu are present in: */ for_each_domain(this_cpu, sd) { - if (cpu_isset(prev_cpu, sd->span)) { + if (cpumask_test_cpu(prev_cpu, sched_domain_span(sd))) { this_sd = sd; break; } diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c index 2bdd44423599..4cd813abc23a 100644 --- a/kernel/sched_rt.c +++ b/kernel/sched_rt.c @@ -1017,7 +1017,8 @@ static int find_lowest_rq(struct task_struct *task) cpumask_t domain_mask; int best_cpu; - cpus_and(domain_mask, sd->span, *lowest_mask); + cpumask_and(&domain_mask, sched_domain_span(sd), + lowest_mask); best_cpu = pick_optimal_cpu(this_cpu, &domain_mask); diff --git a/kernel/sched_stats.h b/kernel/sched_stats.h index 7dbf72a2b02c..ce340835d055 100644 --- a/kernel/sched_stats.h +++ b/kernel/sched_stats.h @@ -42,7 +42,8 @@ static int show_schedstat(struct seq_file *seq, void *v) for_each_domain(cpu, sd) { enum cpu_idle_type itype; - cpumask_scnprintf(mask_str, mask_len, sd->span); + cpumask_scnprintf(mask_str, mask_len, + *sched_domain_span(sd)); seq_printf(seq, "domain%d %s", dcount++, mask_str); for (itype = CPU_IDLE; itype < CPU_MAX_IDLE_TYPES; itype++) { -- cgit v1.2.3 From 6c99e9ad47d9c082bd096f42fb49e397b05d58a8 Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Tue, 25 Nov 2008 02:35:04 +1030 Subject: sched: convert struct sched_group/sched_domain cpumask_ts to variable bitmaps Impact: (future) size reduction for large NR_CPUS. We move the 'cpumask' member of sched_group to the end, so when we kmalloc it we can do a minimal allocation: saves space for small nr_cpu_ids but big CONFIG_NR_CPUS. Similar trick for 'span' in sched_domain. This isn't quite as good as converting to a cpumask_var_t, as some sched_groups are actually static, but it's safer: we don't have to figure out where to call alloc_cpumask_var/free_cpumask_var. Signed-off-by: Rusty Russell Signed-off-by: Ingo Molnar --- include/linux/sched.h | 11 +++++---- kernel/sched.c | 65 ++++++++++++++++++++++++++++++++------------------- 2 files changed, 48 insertions(+), 28 deletions(-) (limited to 'kernel/sched.c') diff --git a/include/linux/sched.h b/include/linux/sched.h index 2b95aa9f779b..c5be6c6bc741 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -771,7 +771,6 @@ enum cpu_idle_type { struct sched_group { struct sched_group *next; /* Must be a circular list */ - cpumask_t cpumask; /* * CPU power of this group, SCHED_LOAD_SCALE being max power for a @@ -784,11 +783,13 @@ struct sched_group { * (see include/linux/reciprocal_div.h) */ u32 reciprocal_cpu_power; + + unsigned long cpumask[]; }; static inline struct cpumask *sched_group_cpus(struct sched_group *sg) { - return &sg->cpumask; + return to_cpumask(sg->cpumask); } enum sched_domain_level { @@ -814,7 +815,6 @@ struct sched_domain { struct sched_domain *parent; /* top domain must be null terminated */ struct sched_domain *child; /* bottom domain must be null terminated */ struct sched_group *groups; /* the balancing groups of the domain */ - cpumask_t span; /* span of all CPUs in this domain */ unsigned long min_interval; /* Minimum balance interval ms */ unsigned long max_interval; /* Maximum balance interval ms */ unsigned int busy_factor; /* less balancing by factor if busy */ @@ -869,11 +869,14 @@ struct sched_domain { #ifdef CONFIG_SCHED_DEBUG char *name; #endif + + /* span of all CPUs in this domain */ + unsigned long span[]; }; static inline struct cpumask *sched_domain_span(struct sched_domain *sd) { - return &sd->span; + return to_cpumask(sd->span); } extern void partition_sched_domains(int ndoms_new, cpumask_t *doms_new, diff --git a/kernel/sched.c b/kernel/sched.c index 575f38acf4da..6b9606a6cabf 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -7005,19 +7005,34 @@ static void sched_domain_node_span(int node, cpumask_t *span) int sched_smt_power_savings = 0, sched_mc_power_savings = 0; +/* + * The cpus mask in sched_group and sched_domain hangs off the end. + * FIXME: use cpumask_var_t or dynamic percpu alloc to avoid wasting space + * for nr_cpu_ids < CONFIG_NR_CPUS. + */ +struct static_sched_group { + struct sched_group sg; + DECLARE_BITMAP(cpus, CONFIG_NR_CPUS); +}; + +struct static_sched_domain { + struct sched_domain sd; + DECLARE_BITMAP(span, CONFIG_NR_CPUS); +}; + /* * SMT sched-domains: */ #ifdef CONFIG_SCHED_SMT -static DEFINE_PER_CPU(struct sched_domain, cpu_domains); -static DEFINE_PER_CPU(struct sched_group, sched_group_cpus); +static DEFINE_PER_CPU(struct static_sched_domain, cpu_domains); +static DEFINE_PER_CPU(struct static_sched_group, sched_group_cpus); static int cpu_to_cpu_group(int cpu, const cpumask_t *cpu_map, struct sched_group **sg, cpumask_t *unused) { if (sg) - *sg = &per_cpu(sched_group_cpus, cpu); + *sg = &per_cpu(sched_group_cpus, cpu).sg; return cpu; } #endif /* CONFIG_SCHED_SMT */ @@ -7026,8 +7041,8 @@ cpu_to_cpu_group(int cpu, const cpumask_t *cpu_map, struct sched_group **sg, * multi-core sched-domains: */ #ifdef CONFIG_SCHED_MC -static DEFINE_PER_CPU(struct sched_domain, core_domains); -static DEFINE_PER_CPU(struct sched_group, sched_group_core); +static DEFINE_PER_CPU(struct static_sched_domain, core_domains); +static DEFINE_PER_CPU(struct static_sched_group, sched_group_core); #endif /* CONFIG_SCHED_MC */ #if defined(CONFIG_SCHED_MC) && defined(CONFIG_SCHED_SMT) @@ -7041,7 +7056,7 @@ cpu_to_core_group(int cpu, const cpumask_t *cpu_map, struct sched_group **sg, cpus_and(*mask, *mask, *cpu_map); group = first_cpu(*mask); if (sg) - *sg = &per_cpu(sched_group_core, group); + *sg = &per_cpu(sched_group_core, group).sg; return group; } #elif defined(CONFIG_SCHED_MC) @@ -7050,13 +7065,13 @@ cpu_to_core_group(int cpu, const cpumask_t *cpu_map, struct sched_group **sg, cpumask_t *unused) { if (sg) - *sg = &per_cpu(sched_group_core, cpu); + *sg = &per_cpu(sched_group_core, cpu).sg; return cpu; } #endif -static DEFINE_PER_CPU(struct sched_domain, phys_domains); -static DEFINE_PER_CPU(struct sched_group, sched_group_phys); +static DEFINE_PER_CPU(struct static_sched_domain, phys_domains); +static DEFINE_PER_CPU(struct static_sched_group, sched_group_phys); static int cpu_to_phys_group(int cpu, const cpumask_t *cpu_map, struct sched_group **sg, @@ -7075,7 +7090,7 @@ cpu_to_phys_group(int cpu, const cpumask_t *cpu_map, struct sched_group **sg, group = cpu; #endif if (sg) - *sg = &per_cpu(sched_group_phys, group); + *sg = &per_cpu(sched_group_phys, group).sg; return group; } @@ -7089,7 +7104,7 @@ static DEFINE_PER_CPU(struct sched_domain, node_domains); static struct sched_group ***sched_group_nodes_bycpu; static DEFINE_PER_CPU(struct sched_domain, allnodes_domains); -static DEFINE_PER_CPU(struct sched_group, sched_group_allnodes); +static DEFINE_PER_CPU(struct static_sched_group, sched_group_allnodes); static int cpu_to_allnodes_group(int cpu, const cpumask_t *cpu_map, struct sched_group **sg, cpumask_t *nodemask) @@ -7101,7 +7116,7 @@ static int cpu_to_allnodes_group(int cpu, const cpumask_t *cpu_map, group = first_cpu(*nodemask); if (sg) - *sg = &per_cpu(sched_group_allnodes, group); + *sg = &per_cpu(sched_group_allnodes, group).sg; return group; } @@ -7116,7 +7131,7 @@ static void init_numa_sched_groups_power(struct sched_group *group_head) for_each_cpu(j, sched_group_cpus(sg)) { struct sched_domain *sd; - sd = &per_cpu(phys_domains, j); + sd = &per_cpu(phys_domains, j).sd; if (j != cpumask_first(sched_group_cpus(sd->groups))) { /* * Only add "power" once for each @@ -7385,7 +7400,7 @@ static int __build_sched_domains(const cpumask_t *cpu_map, #endif p = sd; - sd = &per_cpu(phys_domains, i); + sd = &per_cpu(phys_domains, i).sd; SD_INIT(sd, CPU); set_domain_attribute(sd, attr); cpumask_copy(sched_domain_span(sd), nodemask); @@ -7396,7 +7411,7 @@ static int __build_sched_domains(const cpumask_t *cpu_map, #ifdef CONFIG_SCHED_MC p = sd; - sd = &per_cpu(core_domains, i); + sd = &per_cpu(core_domains, i).sd; SD_INIT(sd, MC); set_domain_attribute(sd, attr); *sched_domain_span(sd) = cpu_coregroup_map(i); @@ -7409,7 +7424,7 @@ static int __build_sched_domains(const cpumask_t *cpu_map, #ifdef CONFIG_SCHED_SMT p = sd; - sd = &per_cpu(cpu_domains, i); + sd = &per_cpu(cpu_domains, i).sd; SD_INIT(sd, SIBLING); set_domain_attribute(sd, attr); cpumask_and(sched_domain_span(sd), @@ -7485,7 +7500,8 @@ static int __build_sched_domains(const cpumask_t *cpu_map, sched_domain_node_span(i, domainspan); cpus_and(*domainspan, *domainspan, *cpu_map); - sg = kmalloc_node(sizeof(struct sched_group), GFP_KERNEL, i); + sg = kmalloc_node(sizeof(struct sched_group) + cpumask_size(), + GFP_KERNEL, i); if (!sg) { printk(KERN_WARNING "Can not alloc domain group for " "node %d\n", i); @@ -7518,7 +7534,8 @@ static int __build_sched_domains(const cpumask_t *cpu_map, if (cpus_empty(*tmpmask)) continue; - sg = kmalloc_node(sizeof(struct sched_group), + sg = kmalloc_node(sizeof(struct sched_group) + + cpumask_size(), GFP_KERNEL, i); if (!sg) { printk(KERN_WARNING @@ -7538,21 +7555,21 @@ static int __build_sched_domains(const cpumask_t *cpu_map, /* Calculate CPU power for physical packages and nodes */ #ifdef CONFIG_SCHED_SMT for_each_cpu(i, cpu_map) { - struct sched_domain *sd = &per_cpu(cpu_domains, i); + struct sched_domain *sd = &per_cpu(cpu_domains, i).sd; init_sched_groups_power(i, sd); } #endif #ifdef CONFIG_SCHED_MC for_each_cpu(i, cpu_map) { - struct sched_domain *sd = &per_cpu(core_domains, i); + struct sched_domain *sd = &per_cpu(core_domains, i).sd; init_sched_groups_power(i, sd); } #endif for_each_cpu(i, cpu_map) { - struct sched_domain *sd = &per_cpu(phys_domains, i); + struct sched_domain *sd = &per_cpu(phys_domains, i).sd; init_sched_groups_power(i, sd); } @@ -7574,11 +7591,11 @@ static int __build_sched_domains(const cpumask_t *cpu_map, for_each_cpu(i, cpu_map) { struct sched_domain *sd; #ifdef CONFIG_SCHED_SMT - sd = &per_cpu(cpu_domains, i); + sd = &per_cpu(cpu_domains, i).sd; #elif defined(CONFIG_SCHED_MC) - sd = &per_cpu(core_domains, i); + sd = &per_cpu(core_domains, i).sd; #else - sd = &per_cpu(phys_domains, i); + sd = &per_cpu(phys_domains, i).sd; #endif cpu_attach_domain(sd, rd, i); } -- cgit v1.2.3 From 6a7b3dc3440f7b5a9b67594af01ed562cdeb41e4 Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Tue, 25 Nov 2008 02:35:04 +1030 Subject: sched: convert nohz_cpu_mask to cpumask_var_t. Impact: (future) size reduction for large NR_CPUS. Dynamically allocating cpumasks (when CONFIG_CPUMASK_OFFSTACK) saves space for small nr_cpu_ids but big CONFIG_NR_CPUS. cpumask_var_t is just a struct cpumask for !CONFIG_CPUMASK_OFFSTACK. Signed-off-by: Rusty Russell Signed-off-by: Ingo Molnar --- include/linux/sched.h | 2 +- kernel/rcuclassic.c | 2 +- kernel/sched.c | 7 +++++-- kernel/time/tick-sched.c | 10 +++++----- 4 files changed, 12 insertions(+), 9 deletions(-) (limited to 'kernel/sched.c') diff --git a/include/linux/sched.h b/include/linux/sched.h index c5be6c6bc741..1e33e2cb7f8c 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -249,7 +249,7 @@ extern void init_idle_bootup_task(struct task_struct *idle); extern int runqueue_is_locked(void); extern void task_rq_unlock_wait(struct task_struct *p); -extern cpumask_t nohz_cpu_mask; +extern cpumask_var_t nohz_cpu_mask; #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ) extern int select_nohz_load_balancer(int cpu); #else diff --git a/kernel/rcuclassic.c b/kernel/rcuclassic.c index e503a002f330..c03ca3e61919 100644 --- a/kernel/rcuclassic.c +++ b/kernel/rcuclassic.c @@ -393,7 +393,7 @@ static void rcu_start_batch(struct rcu_ctrlblk *rcp) * unnecessarily. */ smp_mb(); - cpus_andnot(rcp->cpumask, cpu_online_map, nohz_cpu_mask); + cpumask_andnot(&rcp->cpumask, cpu_online_mask, nohz_cpu_mask); rcp->signaled = 0; } diff --git a/kernel/sched.c b/kernel/sched.c index 6b9606a6cabf..2723d7a4a42d 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -5870,9 +5870,9 @@ void __cpuinit init_idle(struct task_struct *idle, int cpu) * indicates which cpus entered this state. This is used * in the rcu update to wait only for active cpus. For system * which do not switch off the HZ timer nohz_cpu_mask should - * always be CPU_MASK_NONE. + * always be CPU_BITS_NONE. */ -cpumask_t nohz_cpu_mask = CPU_MASK_NONE; +cpumask_var_t nohz_cpu_mask; /* * Increase the granularity value when there are more CPUs, @@ -8274,6 +8274,9 @@ void __init sched_init(void) */ current->sched_class = &fair_sched_class; + /* Allocate the nohz_cpu_mask if CONFIG_CPUMASK_OFFSTACK */ + alloc_bootmem_cpumask_var(&nohz_cpu_mask); + scheduler_running = 1; } diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c index 342fc9ccab46..70f872c71f4e 100644 --- a/kernel/time/tick-sched.c +++ b/kernel/time/tick-sched.c @@ -144,7 +144,7 @@ void tick_nohz_update_jiffies(void) if (!ts->tick_stopped) return; - cpu_clear(cpu, nohz_cpu_mask); + cpumask_clear_cpu(cpu, nohz_cpu_mask); now = ktime_get(); ts->idle_waketime = now; @@ -283,7 +283,7 @@ void tick_nohz_stop_sched_tick(int inidle) if ((long)delta_jiffies >= 1) { if (delta_jiffies > 1) - cpu_set(cpu, nohz_cpu_mask); + cpumask_set_cpu(cpu, nohz_cpu_mask); /* * nohz_stop_sched_tick can be called several times before * the nohz_restart_sched_tick is called. This happens when @@ -296,7 +296,7 @@ void tick_nohz_stop_sched_tick(int inidle) /* * sched tick not stopped! */ - cpu_clear(cpu, nohz_cpu_mask); + cpumask_clear_cpu(cpu, nohz_cpu_mask); goto out; } @@ -354,7 +354,7 @@ void tick_nohz_stop_sched_tick(int inidle) * softirq. */ tick_do_update_jiffies64(ktime_get()); - cpu_clear(cpu, nohz_cpu_mask); + cpumask_clear_cpu(cpu, nohz_cpu_mask); } raise_softirq_irqoff(TIMER_SOFTIRQ); out: @@ -432,7 +432,7 @@ void tick_nohz_restart_sched_tick(void) select_nohz_load_balancer(0); now = ktime_get(); tick_do_update_jiffies64(now); - cpu_clear(cpu, nohz_cpu_mask); + cpumask_clear_cpu(cpu, nohz_cpu_mask); /* * We stopped the tick in idle. Update process times would miss the -- cgit v1.2.3 From c6c4927b22a3514c6660f0e72c78716226bd3cc8 Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Tue, 25 Nov 2008 02:35:05 +1030 Subject: sched: convert struct root_domain to cpumask_var_t. Impact: (future) size reduction for large NR_CPUS. Dynamically allocating cpumasks (when CONFIG_CPUMASK_OFFSTACK) saves space for small nr_cpu_ids but big CONFIG_NR_CPUS. cpumask_var_t is just a struct cpumask for !CONFIG_CPUMASK_OFFSTACK. def_root_domain is static, and so its masks are initialized with alloc_bootmem_cpumask_var. After that, alloc_cpumask_var is used. Signed-off-by: Rusty Russell Signed-off-by: Ingo Molnar --- kernel/sched.c | 69 ++++++++++++++++++++++++++++++++++++++++--------------- kernel/sched_rt.c | 26 ++++++++++----------- 2 files changed, 64 insertions(+), 31 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index 2723d7a4a42d..93309c3034de 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -487,14 +487,14 @@ struct rt_rq { */ struct root_domain { atomic_t refcount; - cpumask_t span; - cpumask_t online; + cpumask_var_t span; + cpumask_var_t online; /* * The "RT overload" flag: it gets set if a CPU has more than * one runnable RT task. */ - cpumask_t rto_mask; + cpumask_var_t rto_mask; atomic_t rto_count; #ifdef CONFIG_SMP struct cpupri cpupri; @@ -6444,7 +6444,7 @@ static void set_rq_online(struct rq *rq) if (!rq->online) { const struct sched_class *class; - cpu_set(rq->cpu, rq->rd->online); + cpumask_set_cpu(rq->cpu, rq->rd->online); rq->online = 1; for_each_class(class) { @@ -6464,7 +6464,7 @@ static void set_rq_offline(struct rq *rq) class->rq_offline(rq); } - cpu_clear(rq->cpu, rq->rd->online); + cpumask_clear_cpu(rq->cpu, rq->rd->online); rq->online = 0; } } @@ -6505,7 +6505,7 @@ migration_call(struct notifier_block *nfb, unsigned long action, void *hcpu) rq = cpu_rq(cpu); spin_lock_irqsave(&rq->lock, flags); if (rq->rd) { - BUG_ON(!cpu_isset(cpu, rq->rd->span)); + BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span)); set_rq_online(rq); } @@ -6567,7 +6567,7 @@ migration_call(struct notifier_block *nfb, unsigned long action, void *hcpu) rq = cpu_rq(cpu); spin_lock_irqsave(&rq->lock, flags); if (rq->rd) { - BUG_ON(!cpu_isset(cpu, rq->rd->span)); + BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span)); set_rq_offline(rq); } spin_unlock_irqrestore(&rq->lock, flags); @@ -6768,6 +6768,14 @@ sd_parent_degenerate(struct sched_domain *sd, struct sched_domain *parent) return 1; } +static void free_rootdomain(struct root_domain *rd) +{ + free_cpumask_var(rd->rto_mask); + free_cpumask_var(rd->online); + free_cpumask_var(rd->span); + kfree(rd); +} + static void rq_attach_root(struct rq *rq, struct root_domain *rd) { unsigned long flags; @@ -6777,38 +6785,60 @@ static void rq_attach_root(struct rq *rq, struct root_domain *rd) if (rq->rd) { struct root_domain *old_rd = rq->rd; - if (cpu_isset(rq->cpu, old_rd->online)) + if (cpumask_test_cpu(rq->cpu, old_rd->online)) set_rq_offline(rq); - cpu_clear(rq->cpu, old_rd->span); + cpumask_clear_cpu(rq->cpu, old_rd->span); if (atomic_dec_and_test(&old_rd->refcount)) - kfree(old_rd); + free_rootdomain(old_rd); } atomic_inc(&rd->refcount); rq->rd = rd; - cpu_set(rq->cpu, rd->span); - if (cpu_isset(rq->cpu, cpu_online_map)) + cpumask_set_cpu(rq->cpu, rd->span); + if (cpumask_test_cpu(rq->cpu, cpu_online_mask)) set_rq_online(rq); spin_unlock_irqrestore(&rq->lock, flags); } -static void init_rootdomain(struct root_domain *rd) +static int init_rootdomain(struct root_domain *rd, bool bootmem) { memset(rd, 0, sizeof(*rd)); - cpus_clear(rd->span); - cpus_clear(rd->online); + if (bootmem) { + alloc_bootmem_cpumask_var(&def_root_domain.span); + alloc_bootmem_cpumask_var(&def_root_domain.online); + alloc_bootmem_cpumask_var(&def_root_domain.rto_mask); + cpupri_init(&rd->cpupri); + return 0; + } + + if (!alloc_cpumask_var(&rd->span, GFP_KERNEL)) + goto free_rd; + if (!alloc_cpumask_var(&rd->online, GFP_KERNEL)) + goto free_span; + if (!alloc_cpumask_var(&rd->rto_mask, GFP_KERNEL)) + goto free_online; cpupri_init(&rd->cpupri); + return 0; + +free_online: + free_cpumask_var(rd->online); +free_span: + free_cpumask_var(rd->span); +free_rd: + kfree(rd); + return -ENOMEM; } static void init_defrootdomain(void) { - init_rootdomain(&def_root_domain); + init_rootdomain(&def_root_domain, true); + atomic_set(&def_root_domain.refcount, 1); } @@ -6820,7 +6850,10 @@ static struct root_domain *alloc_rootdomain(void) if (!rd) return NULL; - init_rootdomain(rd); + if (init_rootdomain(rd, false) != 0) { + kfree(rd); + return NULL; + } return rd; } @@ -7632,7 +7665,7 @@ free_sched_groups: #ifdef CONFIG_NUMA error: free_sched_groups(cpu_map, tmpmask); - kfree(rd); + free_rootdomain(rd); goto free_tmpmask; #endif } diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c index 4cd813abc23a..820fc422c6df 100644 --- a/kernel/sched_rt.c +++ b/kernel/sched_rt.c @@ -15,7 +15,7 @@ static inline void rt_set_overload(struct rq *rq) if (!rq->online) return; - cpu_set(rq->cpu, rq->rd->rto_mask); + cpumask_set_cpu(rq->cpu, rq->rd->rto_mask); /* * Make sure the mask is visible before we set * the overload count. That is checked to determine @@ -34,7 +34,7 @@ static inline void rt_clear_overload(struct rq *rq) /* the order here really doesn't matter */ atomic_dec(&rq->rd->rto_count); - cpu_clear(rq->cpu, rq->rd->rto_mask); + cpumask_clear_cpu(rq->cpu, rq->rd->rto_mask); } static void update_rt_migration(struct rq *rq) @@ -139,14 +139,14 @@ static int rt_se_boosted(struct sched_rt_entity *rt_se) } #ifdef CONFIG_SMP -static inline cpumask_t sched_rt_period_mask(void) +static inline const struct cpumask *sched_rt_period_mask(void) { return cpu_rq(smp_processor_id())->rd->span; } #else -static inline cpumask_t sched_rt_period_mask(void) +static inline const struct cpumask *sched_rt_period_mask(void) { - return cpu_online_map; + return cpu_online_mask; } #endif @@ -212,9 +212,9 @@ static inline int rt_rq_throttled(struct rt_rq *rt_rq) return rt_rq->rt_throttled; } -static inline cpumask_t sched_rt_period_mask(void) +static inline const struct cpumask *sched_rt_period_mask(void) { - return cpu_online_map; + return cpu_online_mask; } static inline @@ -241,11 +241,11 @@ static int do_balance_runtime(struct rt_rq *rt_rq) int i, weight, more = 0; u64 rt_period; - weight = cpus_weight(rd->span); + weight = cpumask_weight(rd->span); spin_lock(&rt_b->rt_runtime_lock); rt_period = ktime_to_ns(rt_b->rt_period); - for_each_cpu_mask_nr(i, rd->span) { + for_each_cpu(i, rd->span) { struct rt_rq *iter = sched_rt_period_rt_rq(rt_b, i); s64 diff; @@ -324,7 +324,7 @@ static void __disable_runtime(struct rq *rq) /* * Greedy reclaim, take back as much as we can. */ - for_each_cpu_mask(i, rd->span) { + for_each_cpu(i, rd->span) { struct rt_rq *iter = sched_rt_period_rt_rq(rt_b, i); s64 diff; @@ -429,13 +429,13 @@ static inline int balance_runtime(struct rt_rq *rt_rq) static int do_sched_rt_period_timer(struct rt_bandwidth *rt_b, int overrun) { int i, idle = 1; - cpumask_t span; + const struct cpumask *span; if (!rt_bandwidth_enabled() || rt_b->rt_runtime == RUNTIME_INF) return 1; span = sched_rt_period_mask(); - for_each_cpu_mask(i, span) { + for_each_cpu(i, span) { int enqueue = 0; struct rt_rq *rt_rq = sched_rt_period_rt_rq(rt_b, i); struct rq *rq = rq_of_rt_rq(rt_rq); @@ -1181,7 +1181,7 @@ static int pull_rt_task(struct rq *this_rq) next = pick_next_task_rt(this_rq); - for_each_cpu_mask_nr(cpu, this_rq->rd->rto_mask) { + for_each_cpu(cpu, this_rq->rd->rto_mask) { if (this_cpu == cpu) continue; -- cgit v1.2.3 From 7d1e6a9b95e3edeac91888bc683ae62f18519432 Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Tue, 25 Nov 2008 02:35:09 +1030 Subject: sched: convert nohz struct to cpumask_var_t. Impact: (future) size reduction for large NR_CPUS. Dynamically allocating cpumasks (when CONFIG_CPUMASK_OFFSTACK) saves space for small nr_cpu_ids but big CONFIG_NR_CPUS. cpumask_var_t is just a struct cpumask for !CONFIG_CPUMASK_OFFSTACK. Signed-off-by: Rusty Russell Signed-off-by: Ingo Molnar --- kernel/sched.c | 29 ++++++++++++++++------------- 1 file changed, 16 insertions(+), 13 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index 93309c3034de..2f8ea99df16a 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -3758,10 +3758,9 @@ static void active_load_balance(struct rq *busiest_rq, int busiest_cpu) #ifdef CONFIG_NO_HZ static struct { atomic_t load_balancer; - cpumask_t cpu_mask; + cpumask_var_t cpu_mask; } nohz ____cacheline_aligned = { .load_balancer = ATOMIC_INIT(-1), - .cpu_mask = CPU_MASK_NONE, }; /* @@ -3789,7 +3788,7 @@ int select_nohz_load_balancer(int stop_tick) int cpu = smp_processor_id(); if (stop_tick) { - cpu_set(cpu, nohz.cpu_mask); + cpumask_set_cpu(cpu, nohz.cpu_mask); cpu_rq(cpu)->in_nohz_recently = 1; /* @@ -3803,7 +3802,7 @@ int select_nohz_load_balancer(int stop_tick) } /* time for ilb owner also to sleep */ - if (cpus_weight(nohz.cpu_mask) == num_online_cpus()) { + if (cpumask_weight(nohz.cpu_mask) == num_online_cpus()) { if (atomic_read(&nohz.load_balancer) == cpu) atomic_set(&nohz.load_balancer, -1); return 0; @@ -3816,10 +3815,10 @@ int select_nohz_load_balancer(int stop_tick) } else if (atomic_read(&nohz.load_balancer) == cpu) return 1; } else { - if (!cpu_isset(cpu, nohz.cpu_mask)) + if (!cpumask_test_cpu(cpu, nohz.cpu_mask)) return 0; - cpu_clear(cpu, nohz.cpu_mask); + cpumask_clear_cpu(cpu, nohz.cpu_mask); if (atomic_read(&nohz.load_balancer) == cpu) if (atomic_cmpxchg(&nohz.load_balancer, cpu, -1) != cpu) @@ -3930,12 +3929,13 @@ static void run_rebalance_domains(struct softirq_action *h) */ if (this_rq->idle_at_tick && atomic_read(&nohz.load_balancer) == this_cpu) { - cpumask_t cpus = nohz.cpu_mask; struct rq *rq; int balance_cpu; - cpu_clear(this_cpu, cpus); - for_each_cpu(balance_cpu, &cpus) { + for_each_cpu(balance_cpu, nohz.cpu_mask) { + if (balance_cpu == this_cpu) + continue; + /* * If this cpu gets work to do, stop the load balancing * work being done for other cpus. Next load @@ -3973,7 +3973,7 @@ static inline void trigger_load_balance(struct rq *rq, int cpu) rq->in_nohz_recently = 0; if (atomic_read(&nohz.load_balancer) == cpu) { - cpu_clear(cpu, nohz.cpu_mask); + cpumask_clear_cpu(cpu, nohz.cpu_mask); atomic_set(&nohz.load_balancer, -1); } @@ -3986,7 +3986,7 @@ static inline void trigger_load_balance(struct rq *rq, int cpu) * TBD: Traverse the sched domains and nominate * the nearest cpu in the nohz.cpu_mask. */ - int ilb = first_cpu(nohz.cpu_mask); + int ilb = cpumask_first(nohz.cpu_mask); if (ilb < nr_cpu_ids) resched_cpu(ilb); @@ -3998,7 +3998,7 @@ static inline void trigger_load_balance(struct rq *rq, int cpu) * cpus with ticks stopped, is it time for that to stop? */ if (rq->idle_at_tick && atomic_read(&nohz.load_balancer) == cpu && - cpus_weight(nohz.cpu_mask) == num_online_cpus()) { + cpumask_weight(nohz.cpu_mask) == num_online_cpus()) { resched_cpu(cpu); return; } @@ -4008,7 +4008,7 @@ static inline void trigger_load_balance(struct rq *rq, int cpu) * someone else, then no need raise the SCHED_SOFTIRQ */ if (rq->idle_at_tick && atomic_read(&nohz.load_balancer) != cpu && - cpu_isset(cpu, nohz.cpu_mask)) + cpumask_test_cpu(cpu, nohz.cpu_mask)) return; #endif if (time_after_eq(jiffies, rq->next_balance)) @@ -8309,6 +8309,9 @@ void __init sched_init(void) /* Allocate the nohz_cpu_mask if CONFIG_CPUMASK_OFFSTACK */ alloc_bootmem_cpumask_var(&nohz_cpu_mask); +#ifdef CONFIG_NO_HZ + alloc_bootmem_cpumask_var(&nohz.cpu_mask); +#endif scheduler_running = 1; } -- cgit v1.2.3 From 4d2732c63e0c05cfef2a74868d08eace922dfc3e Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Tue, 25 Nov 2008 02:35:10 +1030 Subject: sched: convert idle_balance() to cpumask_var_t. Impact: stack usage reduction Dynamically allocating cpumasks (when CONFIG_CPUMASK_OFFSTACK) saves space in the stack. cpumask_var_t is just a struct cpumask for !CONFIG_CPUMASK_OFFSTACK. Signed-off-by: Rusty Russell Signed-off-by: Ingo Molnar --- kernel/sched.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index 2f8ea99df16a..154a95fcea7e 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -3676,7 +3676,10 @@ static void idle_balance(int this_cpu, struct rq *this_rq) struct sched_domain *sd; int pulled_task = -1; unsigned long next_balance = jiffies + HZ; - cpumask_t tmpmask; + cpumask_var_t tmpmask; + + if (!alloc_cpumask_var(&tmpmask, GFP_ATOMIC)) + return; for_each_domain(this_cpu, sd) { unsigned long interval; @@ -3687,7 +3690,7 @@ static void idle_balance(int this_cpu, struct rq *this_rq) if (sd->flags & SD_BALANCE_NEWIDLE) /* If we've pulled tasks over stop searching: */ pulled_task = load_balance_newidle(this_cpu, this_rq, - sd, &tmpmask); + sd, tmpmask); interval = msecs_to_jiffies(sd->balance_interval); if (time_after(next_balance, sd->last_balance + interval)) @@ -3702,6 +3705,7 @@ static void idle_balance(int this_cpu, struct rq *this_rq) */ this_rq->next_balance = next_balance; } + free_cpumask_var(tmpmask); } /* -- cgit v1.2.3 From a0e902452da16b79d7c9230630ed8a595d14fa85 Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Tue, 25 Nov 2008 02:35:11 +1030 Subject: sched: convert rebalance_domains() to cpumask_var_t. Impact: stack usage reduction Dynamically allocating cpumasks (when CONFIG_CPUMASK_OFFSTACK) saves space in the stack. cpumask_var_t is just a struct cpumask for !CONFIG_CPUMASK_OFFSTACK. Signed-off-by: Rusty Russell Signed-off-by: Ingo Molnar --- kernel/sched.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index 154a95fcea7e..67383e7f1ccd 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -3850,7 +3850,11 @@ static void rebalance_domains(int cpu, enum cpu_idle_type idle) unsigned long next_balance = jiffies + 60*HZ; int update_next_balance = 0; int need_serialize; - cpumask_t tmp; + cpumask_var_t tmp; + + /* Fails alloc? Rebalancing probably not a priority right now. */ + if (!alloc_cpumask_var(&tmp, GFP_ATOMIC)) + return; for_each_domain(cpu, sd) { if (!(sd->flags & SD_LOAD_BALANCE)) @@ -3875,7 +3879,7 @@ static void rebalance_domains(int cpu, enum cpu_idle_type idle) } if (time_after_eq(jiffies, sd->last_balance + interval)) { - if (load_balance(cpu, rq, sd, idle, &balance, &tmp)) { + if (load_balance(cpu, rq, sd, idle, &balance, tmp)) { /* * We've pulled tasks over so either we're no * longer idle, or one of our SMT siblings is @@ -3909,6 +3913,8 @@ out: */ if (likely(update_next_balance)) rq->next_balance = next_balance; + + free_cpumask_var(tmp); } /* -- cgit v1.2.3 From f17c860760927c2a8e41a021eab3317e4415e962 Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Tue, 25 Nov 2008 02:35:11 +1030 Subject: sched: convert sys_sched_getaffinity() to cpumask_var_t. Impact: stack usage reduction Dynamically allocating cpumasks (when CONFIG_CPUMASK_OFFSTACK) saves space in the stack. cpumask_var_t is just a struct cpumask for !CONFIG_CPUMASK_OFFSTACK. Some jiggling here to make sure we always exit at the bottom (so we hit the free_cpumask_var there). Signed-off-by: Rusty Russell Signed-off-by: Ingo Molnar --- kernel/sched.c | 21 +++++++++++++-------- 1 file changed, 13 insertions(+), 8 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index 67383e7f1ccd..6deff24349b6 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -5499,19 +5499,24 @@ asmlinkage long sys_sched_getaffinity(pid_t pid, unsigned int len, unsigned long __user *user_mask_ptr) { int ret; - cpumask_t mask; + cpumask_var_t mask; - if (len < sizeof(cpumask_t)) + if (len < cpumask_size()) return -EINVAL; - ret = sched_getaffinity(pid, &mask); - if (ret < 0) - return ret; + if (!alloc_cpumask_var(&mask, GFP_KERNEL)) + return -ENOMEM; - if (copy_to_user(user_mask_ptr, &mask, sizeof(cpumask_t))) - return -EFAULT; + ret = sched_getaffinity(pid, mask); + if (ret == 0) { + if (copy_to_user(user_mask_ptr, mask, cpumask_size())) + ret = -EFAULT; + else + ret = cpumask_size(); + } + free_cpumask_var(mask); - return sizeof(cpumask_t); + return ret; } /** -- cgit v1.2.3 From e76bd8d9850c2296a7e8e24c9dce9b5e6b55fe2f Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Tue, 25 Nov 2008 02:35:11 +1030 Subject: sched: avoid stack var in move_task_off_dead_cpu Impact: stack usage reduction With some care, we can avoid needing a temporary cpumask (we can't really allocate here, since we can't fail). This version calls cpuset_cpus_allowed_locked() with the task_rq_lock held. I'm fairly sure this works, but there might be a deadlock hiding. And of course, we can't get rid of the last cpumask on stack until we can use cpumask_of_node instead of node_to_cpumask. Signed-off-by: Rusty Russell Signed-off-by: Ingo Molnar --- kernel/sched.c | 78 +++++++++++++++++++++++++++------------------------------- 1 file changed, 36 insertions(+), 42 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index 6deff24349b6..f7dee2029e4d 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -6112,52 +6112,46 @@ static int __migrate_task_irq(struct task_struct *p, int src_cpu, int dest_cpu) static void move_task_off_dead_cpu(int dead_cpu, struct task_struct *p) { unsigned long flags; - cpumask_t mask; struct rq *rq; int dest_cpu; + /* FIXME: Use cpumask_of_node here. */ + cpumask_t _nodemask = node_to_cpumask(cpu_to_node(dead_cpu)); + const struct cpumask *nodemask = &_nodemask; + +again: + /* Look for allowed, online CPU in same node. */ + for_each_cpu_and(dest_cpu, nodemask, cpu_online_mask) + if (cpumask_test_cpu(dest_cpu, &p->cpus_allowed)) + goto move; + + /* Any allowed, online CPU? */ + dest_cpu = cpumask_any_and(&p->cpus_allowed, cpu_online_mask); + if (dest_cpu < nr_cpu_ids) + goto move; + + /* No more Mr. Nice Guy. */ + if (dest_cpu >= nr_cpu_ids) { + rq = task_rq_lock(p, &flags); + cpuset_cpus_allowed_locked(p, &p->cpus_allowed); + dest_cpu = cpumask_any_and(cpu_online_mask, &p->cpus_allowed); + task_rq_unlock(rq, &flags); - do { - /* On same node? */ - node_to_cpumask_ptr(pnodemask, cpu_to_node(dead_cpu)); - - cpus_and(mask, *pnodemask, p->cpus_allowed); - dest_cpu = cpumask_any_and(cpu_online_mask, &mask); - - /* On any allowed CPU? */ - if (dest_cpu >= nr_cpu_ids) - dest_cpu = cpumask_any_and(cpu_online_mask, - &p->cpus_allowed); - - /* No more Mr. Nice Guy. */ - if (dest_cpu >= nr_cpu_ids) { - cpumask_t cpus_allowed; - - cpuset_cpus_allowed_locked(p, &cpus_allowed); - /* - * Try to stay on the same cpuset, where the - * current cpuset may be a subset of all cpus. - * The cpuset_cpus_allowed_locked() variant of - * cpuset_cpus_allowed() will not block. It must be - * called within calls to cpuset_lock/cpuset_unlock. - */ - rq = task_rq_lock(p, &flags); - p->cpus_allowed = cpus_allowed; - dest_cpu = cpumask_any_and(cpu_online_mask, - &p->cpus_allowed); - task_rq_unlock(rq, &flags); - - /* - * Don't tell them about moving exiting tasks or - * kernel threads (both mm NULL), since they never - * leave kernel. - */ - if (p->mm && printk_ratelimit()) { - printk(KERN_INFO "process %d (%s) no " - "longer affine to cpu%d\n", - task_pid_nr(p), p->comm, dead_cpu); - } + /* + * Don't tell them about moving exiting tasks or + * kernel threads (both mm NULL), since they never + * leave kernel. + */ + if (p->mm && printk_ratelimit()) { + printk(KERN_INFO "process %d (%s) no " + "longer affine to cpu%d\n", + task_pid_nr(p), p->comm, dead_cpu); } - } while (!__migrate_task_irq(p, dead_cpu, dest_cpu)); + } + +move: + /* It can have affinity changed while we were choosing. */ + if (unlikely(!__migrate_task_irq(p, dead_cpu, dest_cpu))) + goto again; } /* -- cgit v1.2.3 From 5a16f3d30ca4e3f166d691220c003066a14e32b5 Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Tue, 25 Nov 2008 02:35:11 +1030 Subject: sched: convert struct (sys_)sched_setaffinity() to cpumask_var_t. Impact: stack usage reduction Dynamically allocating cpumasks (when CONFIG_CPUMASK_OFFSTACK) saves space on the stack. cpumask_var_t is just a struct cpumask for !CONFIG_CPUMASK_OFFSTACK. Note the removal of the initializer of new_mask: since the first thing we did was "cpus_and(new_mask, new_mask, cpus_allowed)" I just changed that to "cpumask_and(new_mask, in_mask, cpus_allowed);". Signed-off-by: Rusty Russell Signed-off-by: Ingo Molnar --- kernel/sched.c | 40 +++++++++++++++++++++++++++------------- 1 file changed, 27 insertions(+), 13 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index f7dee2029e4d..2d4ff91e0c97 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -5378,8 +5378,7 @@ out_unlock: long sched_setaffinity(pid_t pid, const cpumask_t *in_mask) { - cpumask_t cpus_allowed; - cpumask_t new_mask = *in_mask; + cpumask_var_t cpus_allowed, new_mask; struct task_struct *p; int retval; @@ -5401,6 +5400,14 @@ long sched_setaffinity(pid_t pid, const cpumask_t *in_mask) get_task_struct(p); read_unlock(&tasklist_lock); + if (!alloc_cpumask_var(&cpus_allowed, GFP_KERNEL)) { + retval = -ENOMEM; + goto out_put_task; + } + if (!alloc_cpumask_var(&new_mask, GFP_KERNEL)) { + retval = -ENOMEM; + goto out_free_cpus_allowed; + } retval = -EPERM; if ((current->euid != p->euid) && (current->euid != p->uid) && !capable(CAP_SYS_NICE)) @@ -5410,24 +5417,28 @@ long sched_setaffinity(pid_t pid, const cpumask_t *in_mask) if (retval) goto out_unlock; - cpuset_cpus_allowed(p, &cpus_allowed); - cpus_and(new_mask, new_mask, cpus_allowed); + cpuset_cpus_allowed(p, cpus_allowed); + cpumask_and(new_mask, in_mask, cpus_allowed); again: - retval = set_cpus_allowed_ptr(p, &new_mask); + retval = set_cpus_allowed_ptr(p, new_mask); if (!retval) { - cpuset_cpus_allowed(p, &cpus_allowed); - if (!cpus_subset(new_mask, cpus_allowed)) { + cpuset_cpus_allowed(p, cpus_allowed); + if (!cpumask_subset(new_mask, cpus_allowed)) { /* * We must have raced with a concurrent cpuset * update. Just reset the cpus_allowed to the * cpuset's cpus_allowed */ - new_mask = cpus_allowed; + cpumask_copy(new_mask, cpus_allowed); goto again; } } out_unlock: + free_cpumask_var(new_mask); +out_free_cpus_allowed: + free_cpumask_var(cpus_allowed); +out_put_task: put_task_struct(p); put_online_cpus(); return retval; @@ -5453,14 +5464,17 @@ static int get_user_cpu_mask(unsigned long __user *user_mask_ptr, unsigned len, asmlinkage long sys_sched_setaffinity(pid_t pid, unsigned int len, unsigned long __user *user_mask_ptr) { - cpumask_t new_mask; + cpumask_var_t new_mask; int retval; - retval = get_user_cpu_mask(user_mask_ptr, len, &new_mask); - if (retval) - return retval; + if (!alloc_cpumask_var(&new_mask, GFP_KERNEL)) + return -ENOMEM; - return sched_setaffinity(pid, &new_mask); + retval = get_user_cpu_mask(user_mask_ptr, len, new_mask); + if (retval == 0) + retval = sched_setaffinity(pid, new_mask); + free_cpumask_var(new_mask); + return retval; } long sched_getaffinity(pid_t pid, cpumask_t *mask) -- cgit v1.2.3 From d5dd3db1dce73cdd5c45c5a3498c51bd21b8864b Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Tue, 25 Nov 2008 02:35:12 +1030 Subject: sched: convert sched_domain_debug to cpumask_var_t. Impact: stack usage reduction Dynamically allocating cpumasks (when CONFIG_CPUMASK_OFFSTACK) saves stack space. cpumask_var_t is just a struct cpumask for !CONFIG_CPUMASK_OFFSTACK. In this case, we always alloced, but we don't need to any more. Signed-off-by: Rusty Russell Signed-off-by: Ingo Molnar --- kernel/sched.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index 2d4ff91e0c97..24012c2a8892 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -6706,7 +6706,7 @@ static int sched_domain_debug_one(struct sched_domain *sd, int cpu, int level, static void sched_domain_debug(struct sched_domain *sd, int cpu) { - cpumask_t *groupmask; + cpumask_var_t groupmask; int level = 0; if (!sd) { @@ -6716,8 +6716,7 @@ static void sched_domain_debug(struct sched_domain *sd, int cpu) printk(KERN_DEBUG "CPU%d attaching sched-domain:\n", cpu); - groupmask = kmalloc(sizeof(cpumask_t), GFP_KERNEL); - if (!groupmask) { + if (!alloc_cpumask_var(&groupmask, GFP_KERNEL)) { printk(KERN_DEBUG "Cannot load-balance (out of memory)\n"); return; } @@ -6730,7 +6729,7 @@ static void sched_domain_debug(struct sched_domain *sd, int cpu) if (!sd) break; } - kfree(groupmask); + free_cpumask_var(groupmask); } #else /* !CONFIG_SCHED_DEBUG */ # define sched_domain_debug(sd, cpu) do { } while (0) -- cgit v1.2.3 From dcc30a35f71bcf51f1e9b336dc5e41923071509a Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Tue, 25 Nov 2008 02:35:12 +1030 Subject: sched: convert cpu_isolated_map to cpumask_var_t. Impact: stack usage reduction, (future) size reduction, cleanup Dynamically allocating cpumasks (when CONFIG_CPUMASK_OFFSTACK) saves space for small nr_cpu_ids but big CONFIG_NR_CPUS. cpumask_var_t is just a struct cpumask for !CONFIG_CPUMASK_OFFSTACK. We can also use cpulist_parse() instead of doing it manually in isolated_cpu_setup. Signed-off-by: Rusty Russell Signed-off-by: Ingo Molnar --- kernel/sched.c | 31 ++++++++++++++----------------- 1 file changed, 14 insertions(+), 17 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index 24012c2a8892..526618fe4a78 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -6917,19 +6917,12 @@ cpu_attach_domain(struct sched_domain *sd, struct root_domain *rd, int cpu) } /* cpus with isolated domains */ -static cpumask_t cpu_isolated_map = CPU_MASK_NONE; +static cpumask_var_t cpu_isolated_map; /* Setup the mask of cpus configured for isolated domains */ static int __init isolated_cpu_setup(char *str) { - static int __initdata ints[NR_CPUS]; - int i; - - str = get_options(str, ARRAY_SIZE(ints), ints); - cpus_clear(cpu_isolated_map); - for (i = 1; i <= ints[0]; i++) - if (ints[i] < NR_CPUS) - cpu_set(ints[i], cpu_isolated_map); + cpulist_parse(str, *cpu_isolated_map); return 1; } @@ -7727,7 +7720,7 @@ static int arch_init_sched_domains(const cpumask_t *cpu_map) doms_cur = kmalloc(sizeof(cpumask_t), GFP_KERNEL); if (!doms_cur) doms_cur = &fallback_doms; - cpus_andnot(*doms_cur, *cpu_map, cpu_isolated_map); + cpumask_andnot(doms_cur, cpu_map, cpu_isolated_map); dattr_cur = NULL; err = build_sched_domains(doms_cur); register_sched_domain_sysctl(); @@ -7826,7 +7819,7 @@ match1: if (doms_new == NULL) { ndoms_cur = 0; doms_new = &fallback_doms; - cpus_andnot(doms_new[0], cpu_online_map, cpu_isolated_map); + cpumask_andnot(&doms_new[0], cpu_online_mask, cpu_isolated_map); WARN_ON_ONCE(dattr_new); } @@ -7985,7 +7978,9 @@ static int update_runtime(struct notifier_block *nfb, void __init sched_init_smp(void) { - cpumask_t non_isolated_cpus; + cpumask_var_t non_isolated_cpus; + + alloc_cpumask_var(&non_isolated_cpus, GFP_KERNEL); #if defined(CONFIG_NUMA) sched_group_nodes_bycpu = kzalloc(nr_cpu_ids * sizeof(void **), @@ -7994,10 +7989,10 @@ void __init sched_init_smp(void) #endif get_online_cpus(); mutex_lock(&sched_domains_mutex); - arch_init_sched_domains(&cpu_online_map); - cpus_andnot(non_isolated_cpus, cpu_possible_map, cpu_isolated_map); - if (cpus_empty(non_isolated_cpus)) - cpu_set(smp_processor_id(), non_isolated_cpus); + arch_init_sched_domains(cpu_online_mask); + cpumask_andnot(non_isolated_cpus, cpu_possible_mask, cpu_isolated_map); + if (cpumask_empty(non_isolated_cpus)) + cpumask_set_cpu(smp_processor_id(), non_isolated_cpus); mutex_unlock(&sched_domains_mutex); put_online_cpus(); @@ -8012,9 +8007,10 @@ void __init sched_init_smp(void) init_hrtick(); /* Move init over to a non-isolated CPU */ - if (set_cpus_allowed_ptr(current, &non_isolated_cpus) < 0) + if (set_cpus_allowed_ptr(current, non_isolated_cpus) < 0) BUG(); sched_init_granularity(); + free_cpumask_var(non_isolated_cpus); } #else void __init sched_init_smp(void) @@ -8334,6 +8330,7 @@ void __init sched_init(void) #ifdef CONFIG_NO_HZ alloc_bootmem_cpumask_var(&nohz.cpu_mask); #endif + alloc_bootmem_cpumask_var(&cpu_isolated_map); scheduler_running = 1; } -- cgit v1.2.3 From 4212823fb459eacc8098dd420bb68ebb9917989d Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Tue, 25 Nov 2008 02:35:12 +1030 Subject: sched: convert falback_doms to cpumask_var_t. Impact: (future) size reduction for large NR_CPUS. Dynamically allocating cpumasks (when CONFIG_CPUMASK_OFFSTACK) saves space for small nr_cpu_ids but big CONFIG_NR_CPUS. cpumask_var_t is just a struct cpumask for !CONFIG_CPUMASK_OFFSTACK. Signed-off-by: Rusty Russell Signed-off-by: Ingo Molnar --- kernel/sched.c | 14 ++++++++------ 1 file changed, 8 insertions(+), 6 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index 526618fe4a78..42588ad93b25 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -7697,10 +7697,10 @@ static struct sched_domain_attr *dattr_cur; /* * Special case: If a kmalloc of a doms_cur partition (array of - * cpumask_t) fails, then fallback to a single sched domain, - * as determined by the single cpumask_t fallback_doms. + * cpumask) fails, then fallback to a single sched domain, + * as determined by the single cpumask fallback_doms. */ -static cpumask_t fallback_doms; +static cpumask_var_t fallback_doms; void __attribute__((weak)) arch_update_cpu_topology(void) { @@ -7719,7 +7719,7 @@ static int arch_init_sched_domains(const cpumask_t *cpu_map) ndoms_cur = 1; doms_cur = kmalloc(sizeof(cpumask_t), GFP_KERNEL); if (!doms_cur) - doms_cur = &fallback_doms; + doms_cur = fallback_doms; cpumask_andnot(doms_cur, cpu_map, cpu_isolated_map); dattr_cur = NULL; err = build_sched_domains(doms_cur); @@ -7818,7 +7818,7 @@ match1: if (doms_new == NULL) { ndoms_cur = 0; - doms_new = &fallback_doms; + doms_new = fallback_doms; cpumask_andnot(&doms_new[0], cpu_online_mask, cpu_isolated_map); WARN_ON_ONCE(dattr_new); } @@ -7838,7 +7838,7 @@ match2: } /* Remember the new sched domains */ - if (doms_cur != &fallback_doms) + if (doms_cur != fallback_doms) kfree(doms_cur); kfree(dattr_cur); /* kfree(NULL) is safe */ doms_cur = doms_new; @@ -8011,6 +8011,8 @@ void __init sched_init_smp(void) BUG(); sched_init_granularity(); free_cpumask_var(non_isolated_cpus); + + alloc_cpumask_var(&fallback_doms, GFP_KERNEL); } #else void __init sched_init_smp(void) -- cgit v1.2.3 From 68e74568fbe5854952355e942acca51f138096d9 Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Tue, 25 Nov 2008 02:35:13 +1030 Subject: sched: convert struct cpupri_vec cpumask_var_t. Impact: stack usage reduction, (future) size reduction for large NR_CPUS. Dynamically allocating cpumasks (when CONFIG_CPUMASK_OFFSTACK) saves space for small nr_cpu_ids but big CONFIG_NR_CPUS. The fact cpupro_init is called both before and after the slab is available makes for an ugly parameter unfortunately. We also use cpumask_any_and to get rid of a temporary in cpupri_find. Signed-off-by: Rusty Russell Signed-off-by: Ingo Molnar --- kernel/sched.c | 9 +++++++-- kernel/sched_cpupri.c | 39 ++++++++++++++++++++++++++++----------- kernel/sched_cpupri.h | 5 +++-- 3 files changed, 38 insertions(+), 15 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index 42588ad93b25..94fa333c1e7c 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -6792,6 +6792,8 @@ sd_parent_degenerate(struct sched_domain *sd, struct sched_domain *parent) static void free_rootdomain(struct root_domain *rd) { + cpupri_cleanup(&rd->cpupri); + free_cpumask_var(rd->rto_mask); free_cpumask_var(rd->online); free_cpumask_var(rd->span); @@ -6834,7 +6836,7 @@ static int init_rootdomain(struct root_domain *rd, bool bootmem) alloc_bootmem_cpumask_var(&def_root_domain.span); alloc_bootmem_cpumask_var(&def_root_domain.online); alloc_bootmem_cpumask_var(&def_root_domain.rto_mask); - cpupri_init(&rd->cpupri); + cpupri_init(&rd->cpupri, true); return 0; } @@ -6845,9 +6847,12 @@ static int init_rootdomain(struct root_domain *rd, bool bootmem) if (!alloc_cpumask_var(&rd->rto_mask, GFP_KERNEL)) goto free_online; - cpupri_init(&rd->cpupri); + if (cpupri_init(&rd->cpupri, false) != 0) + goto free_rto_mask; return 0; +free_rto_mask: + free_cpumask_var(rd->rto_mask); free_online: free_cpumask_var(rd->online); free_span: diff --git a/kernel/sched_cpupri.c b/kernel/sched_cpupri.c index 52154fefab7e..018b7be1db2e 100644 --- a/kernel/sched_cpupri.c +++ b/kernel/sched_cpupri.c @@ -67,24 +67,21 @@ static int convert_prio(int prio) * Returns: (int)bool - CPUs were found */ int cpupri_find(struct cpupri *cp, struct task_struct *p, - cpumask_t *lowest_mask) + struct cpumask *lowest_mask) { int idx = 0; int task_pri = convert_prio(p->prio); for_each_cpupri_active(cp->pri_active, idx) { struct cpupri_vec *vec = &cp->pri_to_cpu[idx]; - cpumask_t mask; if (idx >= task_pri) break; - cpus_and(mask, p->cpus_allowed, vec->mask); - - if (cpus_empty(mask)) + if (cpumask_any_and(&p->cpus_allowed, vec->mask) >= nr_cpu_ids) continue; - *lowest_mask = mask; + cpumask_and(lowest_mask, &p->cpus_allowed, vec->mask); return 1; } @@ -126,7 +123,7 @@ void cpupri_set(struct cpupri *cp, int cpu, int newpri) vec->count--; if (!vec->count) clear_bit(oldpri, cp->pri_active); - cpu_clear(cpu, vec->mask); + cpumask_clear_cpu(cpu, vec->mask); spin_unlock_irqrestore(&vec->lock, flags); } @@ -136,7 +133,7 @@ void cpupri_set(struct cpupri *cp, int cpu, int newpri) spin_lock_irqsave(&vec->lock, flags); - cpu_set(cpu, vec->mask); + cpumask_set_cpu(cpu, vec->mask); vec->count++; if (vec->count == 1) set_bit(newpri, cp->pri_active); @@ -150,10 +147,11 @@ void cpupri_set(struct cpupri *cp, int cpu, int newpri) /** * cpupri_init - initialize the cpupri structure * @cp: The cpupri context + * @bootmem: true if allocations need to use bootmem * - * Returns: (void) + * Returns: -ENOMEM if memory fails. */ -void cpupri_init(struct cpupri *cp) +int cpupri_init(struct cpupri *cp, bool bootmem) { int i; @@ -164,11 +162,30 @@ void cpupri_init(struct cpupri *cp) spin_lock_init(&vec->lock); vec->count = 0; - cpus_clear(vec->mask); + if (bootmem) + alloc_bootmem_cpumask_var(&vec->mask); + else if (!alloc_cpumask_var(&vec->mask, GFP_KERNEL)) + goto cleanup; } for_each_possible_cpu(i) cp->cpu_to_pri[i] = CPUPRI_INVALID; + return 0; + +cleanup: + for (i--; i >= 0; i--) + free_cpumask_var(cp->pri_to_cpu[i].mask); + return -ENOMEM; } +/** + * cpupri_cleanup - clean up the cpupri structure + * @cp: The cpupri context + */ +void cpupri_cleanup(struct cpupri *cp) +{ + int i; + for (i = 0; i < CPUPRI_NR_PRIORITIES; i++) + free_cpumask_var(cp->pri_to_cpu[i].mask); +} diff --git a/kernel/sched_cpupri.h b/kernel/sched_cpupri.h index f25811b0f931..642a94ef8a0a 100644 --- a/kernel/sched_cpupri.h +++ b/kernel/sched_cpupri.h @@ -14,7 +14,7 @@ struct cpupri_vec { spinlock_t lock; int count; - cpumask_t mask; + cpumask_var_t mask; }; struct cpupri { @@ -27,7 +27,8 @@ struct cpupri { int cpupri_find(struct cpupri *cp, struct task_struct *p, cpumask_t *lowest_mask); void cpupri_set(struct cpupri *cp, int cpu, int pri); -void cpupri_init(struct cpupri *cp); +int cpupri_init(struct cpupri *cp, bool bootmem); +void cpupri_cleanup(struct cpupri *cp); #else #define cpupri_set(cp, cpu, pri) do { } while (0) #define cpupri_init() do { } while (0) -- cgit v1.2.3 From 0e3900e6d3b04c44737ebc505604dcd8ed30e354 Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Tue, 25 Nov 2008 02:35:13 +1030 Subject: sched: convert local_cpu_mask to cpumask_var_t. Impact: (future) size reduction for large NR_CPUS. Dynamically allocating cpumasks (when CONFIG_CPUMASK_OFFSTACK) saves space for small nr_cpu_ids but big CONFIG_NR_CPUS. cpumask_var_t is just a struct cpumask for !CONFIG_CPUMASK_OFFSTACK. Signed-off-by: Rusty Russell Signed-off-by: Ingo Molnar --- kernel/sched.c | 1 + kernel/sched_rt.c | 13 +++++++++++-- 2 files changed, 12 insertions(+), 2 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index 94fa333c1e7c..f2be61870030 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -8018,6 +8018,7 @@ void __init sched_init_smp(void) free_cpumask_var(non_isolated_cpus); alloc_cpumask_var(&fallback_doms, GFP_KERNEL); + init_sched_rt_class(); } #else void __init sched_init_smp(void) diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c index 1fa13624293e..1f0e99d1a8ce 100644 --- a/kernel/sched_rt.c +++ b/kernel/sched_rt.c @@ -962,7 +962,7 @@ static struct task_struct *pick_next_highest_task_rt(struct rq *rq, int cpu) return next; } -static DEFINE_PER_CPU(cpumask_t, local_cpu_mask); +static DEFINE_PER_CPU(cpumask_var_t, local_cpu_mask); static inline int pick_optimal_cpu(int this_cpu, cpumask_t *mask) { @@ -982,7 +982,7 @@ static inline int pick_optimal_cpu(int this_cpu, cpumask_t *mask) static int find_lowest_rq(struct task_struct *task) { struct sched_domain *sd; - cpumask_t *lowest_mask = &__get_cpu_var(local_cpu_mask); + cpumask_t *lowest_mask = __get_cpu_var(local_cpu_mask); int this_cpu = smp_processor_id(); int cpu = task_cpu(task); @@ -1551,3 +1551,12 @@ static void print_rt_stats(struct seq_file *m, int cpu) rcu_read_unlock(); } #endif /* CONFIG_SCHED_DEBUG */ + +/* Note that this is never called for !SMP, but that's OK. */ +static inline void init_sched_rt_class(void) +{ + unsigned int i; + + for_each_possible_cpu(i) + alloc_cpumask_var(&per_cpu(local_cpu_mask, i), GFP_KERNEL); +} -- cgit v1.2.3 From 96f874e26428ab5d2db681c100210c254775e154 Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Tue, 25 Nov 2008 02:35:14 +1030 Subject: sched: convert remaining old-style cpumask operators Impact: Trivial API conversion NR_CPUS -> nr_cpu_ids cpumask_t -> struct cpumask sizeof(cpumask_t) -> cpumask_size() cpumask_a = cpumask_b -> cpumask_copy(&cpumask_a, &cpumask_b) cpu_set() -> cpumask_set_cpu() first_cpu() -> cpumask_first() cpumask_of_cpu() -> cpumask_of() cpus_* -> cpumask_* There are some FIXMEs where we all archs to complete infrastructure (patches have been sent): cpu_coregroup_map -> cpu_coregroup_mask node_to_cpumask* -> cpumask_of_node There is also one FIXME where we pass an array of cpumasks to partition_sched_domains(): this implies knowing the definition of 'struct cpumask' and the size of a cpumask. This will be fixed in a future patch. Signed-off-by: Rusty Russell Signed-off-by: Ingo Molnar --- include/linux/sched.h | 16 ++-- kernel/sched.c | 212 +++++++++++++++++++++++++++----------------------- kernel/sched_fair.c | 4 +- kernel/sched_rt.c | 18 ++--- 4 files changed, 132 insertions(+), 118 deletions(-) (limited to 'kernel/sched.c') diff --git a/include/linux/sched.h b/include/linux/sched.h index 1e33e2cb7f8c..4b7b0187374c 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -879,7 +879,7 @@ static inline struct cpumask *sched_domain_span(struct sched_domain *sd) return to_cpumask(sd->span); } -extern void partition_sched_domains(int ndoms_new, cpumask_t *doms_new, +extern void partition_sched_domains(int ndoms_new, struct cpumask *doms_new, struct sched_domain_attr *dattr_new); extern int arch_reinit_sched_domains(void); @@ -888,7 +888,7 @@ extern int arch_reinit_sched_domains(void); struct sched_domain_attr; static inline void -partition_sched_domains(int ndoms_new, cpumask_t *doms_new, +partition_sched_domains(int ndoms_new, struct cpumask *doms_new, struct sched_domain_attr *dattr_new) { } @@ -970,7 +970,7 @@ struct sched_class { void (*task_wake_up) (struct rq *this_rq, struct task_struct *task); void (*set_cpus_allowed)(struct task_struct *p, - const cpumask_t *newmask); + const struct cpumask *newmask); void (*rq_online)(struct rq *rq); void (*rq_offline)(struct rq *rq); @@ -1612,12 +1612,12 @@ extern cputime_t task_gtime(struct task_struct *p); #ifdef CONFIG_SMP extern int set_cpus_allowed_ptr(struct task_struct *p, - const cpumask_t *new_mask); + const struct cpumask *new_mask); #else static inline int set_cpus_allowed_ptr(struct task_struct *p, - const cpumask_t *new_mask) + const struct cpumask *new_mask) { - if (!cpu_isset(0, *new_mask)) + if (!cpumask_test_cpu(0, new_mask)) return -EINVAL; return 0; } @@ -2230,8 +2230,8 @@ __trace_special(void *__tr, void *__data, } #endif -extern long sched_setaffinity(pid_t pid, const cpumask_t *new_mask); -extern long sched_getaffinity(pid_t pid, cpumask_t *mask); +extern long sched_setaffinity(pid_t pid, const struct cpumask *new_mask); +extern long sched_getaffinity(pid_t pid, struct cpumask *mask); extern int sched_mc_power_savings, sched_smt_power_savings; diff --git a/kernel/sched.c b/kernel/sched.c index f2be61870030..eba6a156d334 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -2829,7 +2829,7 @@ static void sched_migrate_task(struct task_struct *p, int dest_cpu) struct rq *rq; rq = task_rq_lock(p, &flags); - if (!cpu_isset(dest_cpu, p->cpus_allowed) + if (!cpumask_test_cpu(dest_cpu, &p->cpus_allowed) || unlikely(!cpu_active(dest_cpu))) goto out; @@ -2895,7 +2895,7 @@ int can_migrate_task(struct task_struct *p, struct rq *rq, int this_cpu, * 2) cannot be migrated to this CPU due to cpus_allowed, or * 3) are cache-hot on their current CPU. */ - if (!cpu_isset(this_cpu, p->cpus_allowed)) { + if (!cpumask_test_cpu(this_cpu, &p->cpus_allowed)) { schedstat_inc(p, se.nr_failed_migrations_affine); return 0; } @@ -3070,7 +3070,7 @@ static int move_one_task(struct rq *this_rq, int this_cpu, struct rq *busiest, static struct sched_group * find_busiest_group(struct sched_domain *sd, int this_cpu, unsigned long *imbalance, enum cpu_idle_type idle, - int *sd_idle, const cpumask_t *cpus, int *balance) + int *sd_idle, const struct cpumask *cpus, int *balance) { struct sched_group *busiest = NULL, *this = NULL, *group = sd->groups; unsigned long max_load, avg_load, total_load, this_load, total_pwr; @@ -3387,7 +3387,7 @@ ret: */ static struct rq * find_busiest_queue(struct sched_group *group, enum cpu_idle_type idle, - unsigned long imbalance, const cpumask_t *cpus) + unsigned long imbalance, const struct cpumask *cpus) { struct rq *busiest = NULL, *rq; unsigned long max_load = 0; @@ -3396,7 +3396,7 @@ find_busiest_queue(struct sched_group *group, enum cpu_idle_type idle, for_each_cpu(i, sched_group_cpus(group)) { unsigned long wl; - if (!cpu_isset(i, *cpus)) + if (!cpumask_test_cpu(i, cpus)) continue; rq = cpu_rq(i); @@ -3426,7 +3426,7 @@ find_busiest_queue(struct sched_group *group, enum cpu_idle_type idle, */ static int load_balance(int this_cpu, struct rq *this_rq, struct sched_domain *sd, enum cpu_idle_type idle, - int *balance, cpumask_t *cpus) + int *balance, struct cpumask *cpus) { int ld_moved, all_pinned = 0, active_balance = 0, sd_idle = 0; struct sched_group *group; @@ -3434,7 +3434,7 @@ static int load_balance(int this_cpu, struct rq *this_rq, struct rq *busiest; unsigned long flags; - cpus_setall(*cpus); + cpumask_setall(cpus); /* * When power savings policy is enabled for the parent domain, idle @@ -3494,8 +3494,8 @@ redo: /* All tasks on this runqueue were pinned by CPU affinity */ if (unlikely(all_pinned)) { - cpu_clear(cpu_of(busiest), *cpus); - if (!cpus_empty(*cpus)) + cpumask_clear_cpu(cpu_of(busiest), cpus); + if (!cpumask_empty(cpus)) goto redo; goto out_balanced; } @@ -3512,7 +3512,8 @@ redo: /* don't kick the migration_thread, if the curr * task on busiest cpu can't be moved to this_cpu */ - if (!cpu_isset(this_cpu, busiest->curr->cpus_allowed)) { + if (!cpumask_test_cpu(this_cpu, + &busiest->curr->cpus_allowed)) { spin_unlock_irqrestore(&busiest->lock, flags); all_pinned = 1; goto out_one_pinned; @@ -3587,7 +3588,7 @@ out: */ static int load_balance_newidle(int this_cpu, struct rq *this_rq, struct sched_domain *sd, - cpumask_t *cpus) + struct cpumask *cpus) { struct sched_group *group; struct rq *busiest = NULL; @@ -3596,7 +3597,7 @@ load_balance_newidle(int this_cpu, struct rq *this_rq, struct sched_domain *sd, int sd_idle = 0; int all_pinned = 0; - cpus_setall(*cpus); + cpumask_setall(cpus); /* * When power savings policy is enabled for the parent domain, idle @@ -3640,8 +3641,8 @@ redo: double_unlock_balance(this_rq, busiest); if (unlikely(all_pinned)) { - cpu_clear(cpu_of(busiest), *cpus); - if (!cpus_empty(*cpus)) + cpumask_clear_cpu(cpu_of(busiest), cpus); + if (!cpumask_empty(cpus)) goto redo; } } @@ -5376,7 +5377,7 @@ out_unlock: return retval; } -long sched_setaffinity(pid_t pid, const cpumask_t *in_mask) +long sched_setaffinity(pid_t pid, const struct cpumask *in_mask) { cpumask_var_t cpus_allowed, new_mask; struct task_struct *p; @@ -5445,13 +5446,13 @@ out_put_task: } static int get_user_cpu_mask(unsigned long __user *user_mask_ptr, unsigned len, - cpumask_t *new_mask) + struct cpumask *new_mask) { - if (len < sizeof(cpumask_t)) { - memset(new_mask, 0, sizeof(cpumask_t)); - } else if (len > sizeof(cpumask_t)) { - len = sizeof(cpumask_t); - } + if (len < cpumask_size()) + cpumask_clear(new_mask); + else if (len > cpumask_size()) + len = cpumask_size(); + return copy_from_user(new_mask, user_mask_ptr, len) ? -EFAULT : 0; } @@ -5477,7 +5478,7 @@ asmlinkage long sys_sched_setaffinity(pid_t pid, unsigned int len, return retval; } -long sched_getaffinity(pid_t pid, cpumask_t *mask) +long sched_getaffinity(pid_t pid, struct cpumask *mask) { struct task_struct *p; int retval; @@ -5494,7 +5495,7 @@ long sched_getaffinity(pid_t pid, cpumask_t *mask) if (retval) goto out_unlock; - cpus_and(*mask, p->cpus_allowed, cpu_online_map); + cpumask_and(mask, &p->cpus_allowed, cpu_online_mask); out_unlock: read_unlock(&tasklist_lock); @@ -5872,7 +5873,7 @@ void __cpuinit init_idle(struct task_struct *idle, int cpu) idle->se.exec_start = sched_clock(); idle->prio = idle->normal_prio = MAX_PRIO; - idle->cpus_allowed = cpumask_of_cpu(cpu); + cpumask_copy(&idle->cpus_allowed, cpumask_of(cpu)); __set_task_cpu(idle, cpu); rq->curr = rq->idle = idle; @@ -5956,7 +5957,7 @@ static inline void sched_init_granularity(void) * task must not exit() & deallocate itself prematurely. The * call is not atomic; no spinlocks may be held. */ -int set_cpus_allowed_ptr(struct task_struct *p, const cpumask_t *new_mask) +int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask) { struct migration_req req; unsigned long flags; @@ -5964,13 +5965,13 @@ int set_cpus_allowed_ptr(struct task_struct *p, const cpumask_t *new_mask) int ret = 0; rq = task_rq_lock(p, &flags); - if (!cpus_intersects(*new_mask, cpu_online_map)) { + if (!cpumask_intersects(new_mask, cpu_online_mask)) { ret = -EINVAL; goto out; } if (unlikely((p->flags & PF_THREAD_BOUND) && p != current && - !cpus_equal(p->cpus_allowed, *new_mask))) { + !cpumask_equal(&p->cpus_allowed, new_mask))) { ret = -EINVAL; goto out; } @@ -5978,12 +5979,12 @@ int set_cpus_allowed_ptr(struct task_struct *p, const cpumask_t *new_mask) if (p->sched_class->set_cpus_allowed) p->sched_class->set_cpus_allowed(p, new_mask); else { - p->cpus_allowed = *new_mask; - p->rt.nr_cpus_allowed = cpus_weight(*new_mask); + cpumask_copy(&p->cpus_allowed, new_mask); + p->rt.nr_cpus_allowed = cpumask_weight(new_mask); } /* Can the task run on the task's current CPU? If so, we're done */ - if (cpu_isset(task_cpu(p), *new_mask)) + if (cpumask_test_cpu(task_cpu(p), new_mask)) goto out; if (migrate_task(p, cpumask_any_and(cpu_online_mask, new_mask), &req)) { @@ -6028,7 +6029,7 @@ static int __migrate_task(struct task_struct *p, int src_cpu, int dest_cpu) if (task_cpu(p) != src_cpu) goto done; /* Affinity changed (again). */ - if (!cpu_isset(dest_cpu, p->cpus_allowed)) + if (!cpumask_test_cpu(dest_cpu, &p->cpus_allowed)) goto fail; on_rq = p->se.on_rq; @@ -6629,13 +6630,13 @@ early_initcall(migration_init); #ifdef CONFIG_SCHED_DEBUG static int sched_domain_debug_one(struct sched_domain *sd, int cpu, int level, - cpumask_t *groupmask) + struct cpumask *groupmask) { struct sched_group *group = sd->groups; char str[256]; cpulist_scnprintf(str, sizeof(str), *sched_domain_span(sd)); - cpus_clear(*groupmask); + cpumask_clear(groupmask); printk(KERN_DEBUG "%*s domain %d: ", level, "", level); @@ -6936,24 +6937,25 @@ __setup("isolcpus=", isolated_cpu_setup); /* * init_sched_build_groups takes the cpumask we wish to span, and a pointer * to a function which identifies what group(along with sched group) a CPU - * belongs to. The return value of group_fn must be a >= 0 and < NR_CPUS - * (due to the fact that we keep track of groups covered with a cpumask_t). + * belongs to. The return value of group_fn must be a >= 0 and < nr_cpu_ids + * (due to the fact that we keep track of groups covered with a struct cpumask). * * init_sched_build_groups will build a circular linked list of the groups * covered by the given span, and will set each group's ->cpumask correctly, * and ->cpu_power to 0. */ static void -init_sched_build_groups(const cpumask_t *span, const cpumask_t *cpu_map, - int (*group_fn)(int cpu, const cpumask_t *cpu_map, +init_sched_build_groups(const struct cpumask *span, + const struct cpumask *cpu_map, + int (*group_fn)(int cpu, const struct cpumask *cpu_map, struct sched_group **sg, - cpumask_t *tmpmask), - cpumask_t *covered, cpumask_t *tmpmask) + struct cpumask *tmpmask), + struct cpumask *covered, struct cpumask *tmpmask) { struct sched_group *first = NULL, *last = NULL; int i; - cpus_clear(*covered); + cpumask_clear(covered); for_each_cpu(i, span) { struct sched_group *sg; @@ -6970,7 +6972,7 @@ init_sched_build_groups(const cpumask_t *span, const cpumask_t *cpu_map, if (group_fn(j, cpu_map, NULL, tmpmask) != group) continue; - cpu_set(j, *covered); + cpumask_set_cpu(j, covered); cpumask_set_cpu(j, sched_group_cpus(sg)); } if (!first) @@ -7035,9 +7037,10 @@ static int find_next_best_node(int node, nodemask_t *used_nodes) * should be one that prevents unnecessary balancing, but also spreads tasks * out optimally. */ -static void sched_domain_node_span(int node, cpumask_t *span) +static void sched_domain_node_span(int node, struct cpumask *span) { nodemask_t used_nodes; + /* FIXME: use cpumask_of_node() */ node_to_cpumask_ptr(nodemask, node); int i; @@ -7081,8 +7084,8 @@ static DEFINE_PER_CPU(struct static_sched_domain, cpu_domains); static DEFINE_PER_CPU(struct static_sched_group, sched_group_cpus); static int -cpu_to_cpu_group(int cpu, const cpumask_t *cpu_map, struct sched_group **sg, - cpumask_t *unused) +cpu_to_cpu_group(int cpu, const struct cpumask *cpu_map, + struct sched_group **sg, struct cpumask *unused) { if (sg) *sg = &per_cpu(sched_group_cpus, cpu).sg; @@ -7100,22 +7103,21 @@ static DEFINE_PER_CPU(struct static_sched_group, sched_group_core); #if defined(CONFIG_SCHED_MC) && defined(CONFIG_SCHED_SMT) static int -cpu_to_core_group(int cpu, const cpumask_t *cpu_map, struct sched_group **sg, - cpumask_t *mask) +cpu_to_core_group(int cpu, const struct cpumask *cpu_map, + struct sched_group **sg, struct cpumask *mask) { int group; - *mask = per_cpu(cpu_sibling_map, cpu); - cpus_and(*mask, *mask, *cpu_map); - group = first_cpu(*mask); + cpumask_and(mask, &per_cpu(cpu_sibling_map, cpu), cpu_map); + group = cpumask_first(mask); if (sg) *sg = &per_cpu(sched_group_core, group).sg; return group; } #elif defined(CONFIG_SCHED_MC) static int -cpu_to_core_group(int cpu, const cpumask_t *cpu_map, struct sched_group **sg, - cpumask_t *unused) +cpu_to_core_group(int cpu, const struct cpumask *cpu_map, + struct sched_group **sg, struct cpumask *unused) { if (sg) *sg = &per_cpu(sched_group_core, cpu).sg; @@ -7127,18 +7129,18 @@ static DEFINE_PER_CPU(struct static_sched_domain, phys_domains); static DEFINE_PER_CPU(struct static_sched_group, sched_group_phys); static int -cpu_to_phys_group(int cpu, const cpumask_t *cpu_map, struct sched_group **sg, - cpumask_t *mask) +cpu_to_phys_group(int cpu, const struct cpumask *cpu_map, + struct sched_group **sg, struct cpumask *mask) { int group; #ifdef CONFIG_SCHED_MC + /* FIXME: Use cpu_coregroup_mask. */ *mask = cpu_coregroup_map(cpu); cpus_and(*mask, *mask, *cpu_map); - group = first_cpu(*mask); + group = cpumask_first(mask); #elif defined(CONFIG_SCHED_SMT) - *mask = per_cpu(cpu_sibling_map, cpu); - cpus_and(*mask, *mask, *cpu_map); - group = first_cpu(*mask); + cpumask_and(mask, &per_cpu(cpu_sibling_map, cpu), cpu_map); + group = cpumask_first(mask); #else group = cpu; #endif @@ -7159,14 +7161,16 @@ static struct sched_group ***sched_group_nodes_bycpu; static DEFINE_PER_CPU(struct sched_domain, allnodes_domains); static DEFINE_PER_CPU(struct static_sched_group, sched_group_allnodes); -static int cpu_to_allnodes_group(int cpu, const cpumask_t *cpu_map, - struct sched_group **sg, cpumask_t *nodemask) +static int cpu_to_allnodes_group(int cpu, const struct cpumask *cpu_map, + struct sched_group **sg, + struct cpumask *nodemask) { int group; + /* FIXME: use cpumask_of_node */ node_to_cpumask_ptr(pnodemask, cpu_to_node(cpu)); - cpus_and(*nodemask, *pnodemask, *cpu_map); - group = first_cpu(*nodemask); + cpumask_and(nodemask, pnodemask, cpu_map); + group = cpumask_first(nodemask); if (sg) *sg = &per_cpu(sched_group_allnodes, group).sg; @@ -7202,7 +7206,8 @@ static void init_numa_sched_groups_power(struct sched_group *group_head) #ifdef CONFIG_NUMA /* Free memory allocated for various sched_group structures */ -static void free_sched_groups(const cpumask_t *cpu_map, cpumask_t *nodemask) +static void free_sched_groups(const struct cpumask *cpu_map, + struct cpumask *nodemask) { int cpu, i; @@ -7215,10 +7220,11 @@ static void free_sched_groups(const cpumask_t *cpu_map, cpumask_t *nodemask) for (i = 0; i < nr_node_ids; i++) { struct sched_group *oldsg, *sg = sched_group_nodes[i]; + /* FIXME: Use cpumask_of_node */ node_to_cpumask_ptr(pnodemask, i); cpus_and(*nodemask, *pnodemask, *cpu_map); - if (cpus_empty(*nodemask)) + if (cpumask_empty(nodemask)) continue; if (sg == NULL) @@ -7236,7 +7242,8 @@ next_sg: } } #else /* !CONFIG_NUMA */ -static void free_sched_groups(const cpumask_t *cpu_map, cpumask_t *nodemask) +static void free_sched_groups(const struct cpumask *cpu_map, + struct cpumask *nodemask) { } #endif /* CONFIG_NUMA */ @@ -7366,7 +7373,7 @@ static void set_domain_attribute(struct sched_domain *sd, * Build sched domains for a given set of cpus and attach the sched domains * to the individual cpus */ -static int __build_sched_domains(const cpumask_t *cpu_map, +static int __build_sched_domains(const struct cpumask *cpu_map, struct sched_domain_attr *attr) { int i, err = -ENOMEM; @@ -7416,7 +7423,7 @@ static int __build_sched_domains(const cpumask_t *cpu_map, } #ifdef CONFIG_NUMA - sched_group_nodes_bycpu[first_cpu(*cpu_map)] = sched_group_nodes; + sched_group_nodes_bycpu[cpumask_first(cpu_map)] = sched_group_nodes; #endif /* @@ -7425,12 +7432,13 @@ static int __build_sched_domains(const cpumask_t *cpu_map, for_each_cpu(i, cpu_map) { struct sched_domain *sd = NULL, *p; + /* FIXME: use cpumask_of_node */ *nodemask = node_to_cpumask(cpu_to_node(i)); cpus_and(*nodemask, *nodemask, *cpu_map); #ifdef CONFIG_NUMA - if (cpus_weight(*cpu_map) > - SD_NODES_PER_DOMAIN*cpus_weight(*nodemask)) { + if (cpumask_weight(cpu_map) > + SD_NODES_PER_DOMAIN*cpumask_weight(nodemask)) { sd = &per_cpu(allnodes_domains, i); SD_INIT(sd, ALLNODES); set_domain_attribute(sd, attr); @@ -7491,9 +7499,9 @@ static int __build_sched_domains(const cpumask_t *cpu_map, #ifdef CONFIG_SCHED_SMT /* Set up CPU (sibling) groups */ for_each_cpu(i, cpu_map) { - *this_sibling_map = per_cpu(cpu_sibling_map, i); - cpus_and(*this_sibling_map, *this_sibling_map, *cpu_map); - if (i != first_cpu(*this_sibling_map)) + cpumask_and(this_sibling_map, + &per_cpu(cpu_sibling_map, i), cpu_map); + if (i != cpumask_first(this_sibling_map)) continue; init_sched_build_groups(this_sibling_map, cpu_map, @@ -7505,9 +7513,10 @@ static int __build_sched_domains(const cpumask_t *cpu_map, #ifdef CONFIG_SCHED_MC /* Set up multi-core groups */ for_each_cpu(i, cpu_map) { + /* FIXME: Use cpu_coregroup_mask */ *this_core_map = cpu_coregroup_map(i); cpus_and(*this_core_map, *this_core_map, *cpu_map); - if (i != first_cpu(*this_core_map)) + if (i != cpumask_first(this_core_map)) continue; init_sched_build_groups(this_core_map, cpu_map, @@ -7518,9 +7527,10 @@ static int __build_sched_domains(const cpumask_t *cpu_map, /* Set up physical groups */ for (i = 0; i < nr_node_ids; i++) { + /* FIXME: Use cpumask_of_node */ *nodemask = node_to_cpumask(i); cpus_and(*nodemask, *nodemask, *cpu_map); - if (cpus_empty(*nodemask)) + if (cpumask_empty(nodemask)) continue; init_sched_build_groups(nodemask, cpu_map, @@ -7541,17 +7551,18 @@ static int __build_sched_domains(const cpumask_t *cpu_map, struct sched_group *sg, *prev; int j; + /* FIXME: Use cpumask_of_node */ *nodemask = node_to_cpumask(i); - cpus_clear(*covered); + cpumask_clear(covered); cpus_and(*nodemask, *nodemask, *cpu_map); - if (cpus_empty(*nodemask)) { + if (cpumask_empty(nodemask)) { sched_group_nodes[i] = NULL; continue; } sched_domain_node_span(i, domainspan); - cpus_and(*domainspan, *domainspan, *cpu_map); + cpumask_and(domainspan, domainspan, cpu_map); sg = kmalloc_node(sizeof(struct sched_group) + cpumask_size(), GFP_KERNEL, i); @@ -7570,21 +7581,22 @@ static int __build_sched_domains(const cpumask_t *cpu_map, sg->__cpu_power = 0; cpumask_copy(sched_group_cpus(sg), nodemask); sg->next = sg; - cpus_or(*covered, *covered, *nodemask); + cpumask_or(covered, covered, nodemask); prev = sg; for (j = 0; j < nr_node_ids; j++) { int n = (i + j) % nr_node_ids; + /* FIXME: Use cpumask_of_node */ node_to_cpumask_ptr(pnodemask, n); - cpus_complement(*notcovered, *covered); - cpus_and(*tmpmask, *notcovered, *cpu_map); - cpus_and(*tmpmask, *tmpmask, *domainspan); - if (cpus_empty(*tmpmask)) + cpumask_complement(notcovered, covered); + cpumask_and(tmpmask, notcovered, cpu_map); + cpumask_and(tmpmask, tmpmask, domainspan); + if (cpumask_empty(tmpmask)) break; - cpus_and(*tmpmask, *tmpmask, *pnodemask); - if (cpus_empty(*tmpmask)) + cpumask_and(tmpmask, tmpmask, pnodemask); + if (cpumask_empty(tmpmask)) continue; sg = kmalloc_node(sizeof(struct sched_group) + @@ -7598,7 +7610,7 @@ static int __build_sched_domains(const cpumask_t *cpu_map, sg->__cpu_power = 0; cpumask_copy(sched_group_cpus(sg), tmpmask); sg->next = prev->next; - cpus_or(*covered, *covered, *tmpmask); + cpumask_or(covered, covered, tmpmask); prev->next = sg; prev = sg; } @@ -7634,7 +7646,7 @@ static int __build_sched_domains(const cpumask_t *cpu_map, if (sd_allnodes) { struct sched_group *sg; - cpu_to_allnodes_group(first_cpu(*cpu_map), cpu_map, &sg, + cpu_to_allnodes_group(cpumask_first(cpu_map), cpu_map, &sg, tmpmask); init_numa_sched_groups_power(sg); } @@ -7690,12 +7702,12 @@ error: #endif } -static int build_sched_domains(const cpumask_t *cpu_map) +static int build_sched_domains(const struct cpumask *cpu_map) { return __build_sched_domains(cpu_map, NULL); } -static cpumask_t *doms_cur; /* current sched domains */ +static struct cpumask *doms_cur; /* current sched domains */ static int ndoms_cur; /* number of sched domains in 'doms_cur' */ static struct sched_domain_attr *dattr_cur; /* attribues of custom domains in 'doms_cur' */ @@ -7716,13 +7728,13 @@ void __attribute__((weak)) arch_update_cpu_topology(void) * For now this just excludes isolated cpus, but could be used to * exclude other special cases in the future. */ -static int arch_init_sched_domains(const cpumask_t *cpu_map) +static int arch_init_sched_domains(const struct cpumask *cpu_map) { int err; arch_update_cpu_topology(); ndoms_cur = 1; - doms_cur = kmalloc(sizeof(cpumask_t), GFP_KERNEL); + doms_cur = kmalloc(cpumask_size(), GFP_KERNEL); if (!doms_cur) doms_cur = fallback_doms; cpumask_andnot(doms_cur, cpu_map, cpu_isolated_map); @@ -7733,8 +7745,8 @@ static int arch_init_sched_domains(const cpumask_t *cpu_map) return err; } -static void arch_destroy_sched_domains(const cpumask_t *cpu_map, - cpumask_t *tmpmask) +static void arch_destroy_sched_domains(const struct cpumask *cpu_map, + struct cpumask *tmpmask) { free_sched_groups(cpu_map, tmpmask); } @@ -7743,15 +7755,16 @@ static void arch_destroy_sched_domains(const cpumask_t *cpu_map, * Detach sched domains from a group of cpus specified in cpu_map * These cpus will now be attached to the NULL domain */ -static void detach_destroy_domains(const cpumask_t *cpu_map) +static void detach_destroy_domains(const struct cpumask *cpu_map) { - cpumask_t tmpmask; + /* Save because hotplug lock held. */ + static DECLARE_BITMAP(tmpmask, CONFIG_NR_CPUS); int i; for_each_cpu(i, cpu_map) cpu_attach_domain(NULL, &def_root_domain, i); synchronize_sched(); - arch_destroy_sched_domains(cpu_map, &tmpmask); + arch_destroy_sched_domains(cpu_map, to_cpumask(tmpmask)); } /* handle null as "default" */ @@ -7776,7 +7789,7 @@ static int dattrs_equal(struct sched_domain_attr *cur, int idx_cur, * doms_new[] to the current sched domain partitioning, doms_cur[]. * It destroys each deleted domain and builds each new domain. * - * 'doms_new' is an array of cpumask_t's of length 'ndoms_new'. + * 'doms_new' is an array of cpumask's of length 'ndoms_new'. * The masks don't intersect (don't overlap.) We should setup one * sched domain for each mask. CPUs not in any of the cpumasks will * not be load balanced. If the same cpumask appears both in the @@ -7790,13 +7803,14 @@ static int dattrs_equal(struct sched_domain_attr *cur, int idx_cur, * the single partition 'fallback_doms', it also forces the domains * to be rebuilt. * - * If doms_new == NULL it will be replaced with cpu_online_map. + * If doms_new == NULL it will be replaced with cpu_online_mask. * ndoms_new == 0 is a special case for destroying existing domains, * and it will not create the default domain. * * Call with hotplug lock held */ -void partition_sched_domains(int ndoms_new, cpumask_t *doms_new, +/* FIXME: Change to struct cpumask *doms_new[] */ +void partition_sched_domains(int ndoms_new, struct cpumask *doms_new, struct sched_domain_attr *dattr_new) { int i, j, n; @@ -7811,7 +7825,7 @@ void partition_sched_domains(int ndoms_new, cpumask_t *doms_new, /* Destroy deleted domains */ for (i = 0; i < ndoms_cur; i++) { for (j = 0; j < n; j++) { - if (cpus_equal(doms_cur[i], doms_new[j]) + if (cpumask_equal(&doms_cur[i], &doms_new[j]) && dattrs_equal(dattr_cur, i, dattr_new, j)) goto match1; } @@ -7831,7 +7845,7 @@ match1: /* Build new domains */ for (i = 0; i < ndoms_new; i++) { for (j = 0; j < ndoms_cur; j++) { - if (cpus_equal(doms_new[i], doms_cur[j]) + if (cpumask_equal(&doms_new[i], &doms_cur[j]) && dattrs_equal(dattr_new, i, dattr_cur, j)) goto match2; } diff --git a/kernel/sched_fair.c b/kernel/sched_fair.c index bba00402ed90..08ffffd4a410 100644 --- a/kernel/sched_fair.c +++ b/kernel/sched_fair.c @@ -1017,7 +1017,7 @@ static void yield_task_fair(struct rq *rq) * search starts with cpus closest then further out as needed, * so we always favor a closer, idle cpu. * Domains may include CPUs that are not usable for migration, - * hence we need to mask them out (cpu_active_map) + * hence we need to mask them out (cpu_active_mask) * * Returns the CPU we should wake onto. */ @@ -1244,7 +1244,7 @@ static int select_task_rq_fair(struct task_struct *p, int sync) } } - if (unlikely(!cpu_isset(this_cpu, p->cpus_allowed))) + if (unlikely(!cpumask_test_cpu(this_cpu, &p->cpus_allowed))) goto out; /* diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c index 1f0e99d1a8ce..fb3964579a8a 100644 --- a/kernel/sched_rt.c +++ b/kernel/sched_rt.c @@ -923,7 +923,7 @@ static void deactivate_task(struct rq *rq, struct task_struct *p, int sleep); static int pick_rt_task(struct rq *rq, struct task_struct *p, int cpu) { if (!task_running(rq, p) && - (cpu < 0 || cpu_isset(cpu, p->cpus_allowed)) && + (cpu < 0 || cpumask_test_cpu(cpu, &p->cpus_allowed)) && (p->rt.nr_cpus_allowed > 1)) return 1; return 0; @@ -982,7 +982,7 @@ static inline int pick_optimal_cpu(int this_cpu, cpumask_t *mask) static int find_lowest_rq(struct task_struct *task) { struct sched_domain *sd; - cpumask_t *lowest_mask = __get_cpu_var(local_cpu_mask); + struct cpumask *lowest_mask = __get_cpu_var(local_cpu_mask); int this_cpu = smp_processor_id(); int cpu = task_cpu(task); @@ -997,7 +997,7 @@ static int find_lowest_rq(struct task_struct *task) * I guess we might want to change cpupri_find() to ignore those * in the first place. */ - cpus_and(*lowest_mask, *lowest_mask, cpu_active_map); + cpumask_and(lowest_mask, lowest_mask, cpu_active_mask); /* * At this point we have built a mask of cpus representing the @@ -1007,7 +1007,7 @@ static int find_lowest_rq(struct task_struct *task) * We prioritize the last cpu that the task executed on since * it is most likely cache-hot in that location. */ - if (cpu_isset(cpu, *lowest_mask)) + if (cpumask_test_cpu(cpu, lowest_mask)) return cpu; /* @@ -1064,8 +1064,8 @@ static struct rq *find_lock_lowest_rq(struct task_struct *task, struct rq *rq) * Also make sure that it wasn't scheduled on its rq. */ if (unlikely(task_rq(task) != rq || - !cpu_isset(lowest_rq->cpu, - task->cpus_allowed) || + !cpumask_test_cpu(lowest_rq->cpu, + &task->cpus_allowed) || task_running(rq, task) || !task->se.on_rq)) { @@ -1315,9 +1315,9 @@ move_one_task_rt(struct rq *this_rq, int this_cpu, struct rq *busiest, } static void set_cpus_allowed_rt(struct task_struct *p, - const cpumask_t *new_mask) + const struct cpumask *new_mask) { - int weight = cpus_weight(*new_mask); + int weight = cpumask_weight(new_mask); BUG_ON(!rt_task(p)); @@ -1338,7 +1338,7 @@ static void set_cpus_allowed_rt(struct task_struct *p, update_rt_migration(rq); } - p->cpus_allowed = *new_mask; + cpumask_copy(&p->cpus_allowed, new_mask); p->rt.nr_cpus_allowed = weight; } -- cgit v1.2.3 From fb52607afcd0629776f1dc9e657647ceae81dd50 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Tue, 25 Nov 2008 21:07:04 +0100 Subject: tracing/function-return-tracer: change the name into function-graph-tracer Impact: cleanup This patch changes the name of the "return function tracer" into function-graph-tracer which is a more suitable name for a tracing which makes one able to retrieve the ordered call stack during the code flow. Signed-off-by: Frederic Weisbecker Acked-by: Steven Rostedt Signed-off-by: Ingo Molnar --- arch/x86/Kconfig | 2 +- arch/x86/include/asm/ftrace.h | 4 +- arch/x86/kernel/Makefile | 4 +- arch/x86/kernel/entry_32.S | 12 ++--- arch/x86/kernel/ftrace.c | 12 ++--- include/linux/ftrace.h | 24 ++++----- include/linux/ftrace_irq.h | 2 +- include/linux/sched.h | 2 +- kernel/Makefile | 2 +- kernel/fork.c | 4 +- kernel/sched.c | 2 +- kernel/trace/Kconfig | 19 ++++--- kernel/trace/Makefile | 2 +- kernel/trace/ftrace.c | 26 +++++----- kernel/trace/trace.c | 18 +++---- kernel/trace/trace.h | 12 ++--- kernel/trace/trace_functions_graph.c | 98 ++++++++++++++++++++++++++++++++++++ 17 files changed, 173 insertions(+), 72 deletions(-) create mode 100644 kernel/trace/trace_functions_graph.c (limited to 'kernel/sched.c') diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig index e49a4fd718fe..0842b1127684 100644 --- a/arch/x86/Kconfig +++ b/arch/x86/Kconfig @@ -29,7 +29,7 @@ config X86 select HAVE_FTRACE_MCOUNT_RECORD select HAVE_DYNAMIC_FTRACE select HAVE_FUNCTION_TRACER - select HAVE_FUNCTION_RET_TRACER if X86_32 + select HAVE_FUNCTION_GRAPH_TRACER if X86_32 select HAVE_FUNCTION_TRACE_MCOUNT_TEST select HAVE_KVM if ((X86_32 && !X86_VOYAGER && !X86_VISWS && !X86_NUMAQ) || X86_64) select HAVE_ARCH_KGDB if !X86_VOYAGER diff --git a/arch/x86/include/asm/ftrace.h b/arch/x86/include/asm/ftrace.h index 754a3e082f94..7e61b4ceb9a4 100644 --- a/arch/x86/include/asm/ftrace.h +++ b/arch/x86/include/asm/ftrace.h @@ -28,7 +28,7 @@ struct dyn_arch_ftrace { #endif /* __ASSEMBLY__ */ #endif /* CONFIG_FUNCTION_TRACER */ -#ifdef CONFIG_FUNCTION_RET_TRACER +#ifdef CONFIG_FUNCTION_GRAPH_TRACER #ifndef __ASSEMBLY__ @@ -51,6 +51,6 @@ struct ftrace_ret_stack { extern void return_to_handler(void); #endif /* __ASSEMBLY__ */ -#endif /* CONFIG_FUNCTION_RET_TRACER */ +#endif /* CONFIG_FUNCTION_GRAPH_TRACER */ #endif /* _ASM_X86_FTRACE_H */ diff --git a/arch/x86/kernel/Makefile b/arch/x86/kernel/Makefile index af2bc36ca1c4..64939a0c3986 100644 --- a/arch/x86/kernel/Makefile +++ b/arch/x86/kernel/Makefile @@ -14,7 +14,7 @@ CFLAGS_REMOVE_paravirt-spinlocks.o = -pg CFLAGS_REMOVE_ftrace.o = -pg endif -ifdef CONFIG_FUNCTION_RET_TRACER +ifdef CONFIG_FUNCTION_GRAPH_TRACER # Don't trace __switch_to() but let it for function tracer CFLAGS_REMOVE_process_32.o = -pg endif @@ -70,7 +70,7 @@ obj-$(CONFIG_X86_LOCAL_APIC) += apic.o nmi.o obj-$(CONFIG_X86_IO_APIC) += io_apic.o obj-$(CONFIG_X86_REBOOTFIXUPS) += reboot_fixups_32.o obj-$(CONFIG_DYNAMIC_FTRACE) += ftrace.o -obj-$(CONFIG_FUNCTION_RET_TRACER) += ftrace.o +obj-$(CONFIG_FUNCTION_GRAPH_TRACER) += ftrace.o obj-$(CONFIG_KEXEC) += machine_kexec_$(BITS).o obj-$(CONFIG_KEXEC) += relocate_kernel_$(BITS).o crash.o obj-$(CONFIG_CRASH_DUMP) += crash_dump_$(BITS).o diff --git a/arch/x86/kernel/entry_32.S b/arch/x86/kernel/entry_32.S index 74defe21ba42..2b1f0f081a6b 100644 --- a/arch/x86/kernel/entry_32.S +++ b/arch/x86/kernel/entry_32.S @@ -1188,9 +1188,9 @@ ENTRY(mcount) cmpl $ftrace_stub, ftrace_trace_function jnz trace -#ifdef CONFIG_FUNCTION_RET_TRACER - cmpl $ftrace_stub, ftrace_function_return - jnz ftrace_return_caller +#ifdef CONFIG_FUNCTION_GRAPH_TRACER + cmpl $ftrace_stub, ftrace_graph_function + jnz ftrace_graph_caller #endif .globl ftrace_stub ftrace_stub: @@ -1215,8 +1215,8 @@ END(mcount) #endif /* CONFIG_DYNAMIC_FTRACE */ #endif /* CONFIG_FUNCTION_TRACER */ -#ifdef CONFIG_FUNCTION_RET_TRACER -ENTRY(ftrace_return_caller) +#ifdef CONFIG_FUNCTION_GRAPH_TRACER +ENTRY(ftrace_graph_caller) cmpl $0, function_trace_stop jne ftrace_stub @@ -1230,7 +1230,7 @@ ENTRY(ftrace_return_caller) popl %ecx popl %eax ret -END(ftrace_return_caller) +END(ftrace_graph_caller) .globl return_to_handler return_to_handler: diff --git a/arch/x86/kernel/ftrace.c b/arch/x86/kernel/ftrace.c index bb137f7297ed..3595a4c14aba 100644 --- a/arch/x86/kernel/ftrace.c +++ b/arch/x86/kernel/ftrace.c @@ -323,7 +323,7 @@ int __init ftrace_dyn_arch_init(void *data) } #endif -#ifdef CONFIG_FUNCTION_RET_TRACER +#ifdef CONFIG_FUNCTION_GRAPH_TRACER #ifndef CONFIG_DYNAMIC_FTRACE @@ -389,11 +389,11 @@ static void pop_return_trace(unsigned long *ret, unsigned long long *time, */ unsigned long ftrace_return_to_handler(void) { - struct ftrace_retfunc trace; + struct ftrace_graph_ret trace; pop_return_trace(&trace.ret, &trace.calltime, &trace.func, &trace.overrun); trace.rettime = cpu_clock(raw_smp_processor_id()); - ftrace_function_return(&trace); + ftrace_graph_function(&trace); return trace.ret; } @@ -440,12 +440,12 @@ void prepare_ftrace_return(unsigned long *parent, unsigned long self_addr) ); if (WARN_ON(faulted)) { - unregister_ftrace_return(); + unregister_ftrace_graph(); return; } if (WARN_ON(!__kernel_text_address(old))) { - unregister_ftrace_return(); + unregister_ftrace_graph(); *parent = old; return; } @@ -456,4 +456,4 @@ void prepare_ftrace_return(unsigned long *parent, unsigned long self_addr) *parent = old; } -#endif /* CONFIG_FUNCTION_RET_TRACER */ +#endif /* CONFIG_FUNCTION_GRAPH_TRACER */ diff --git a/include/linux/ftrace.h b/include/linux/ftrace.h index 7854d87b97b2..b4ac734ad8d6 100644 --- a/include/linux/ftrace.h +++ b/include/linux/ftrace.h @@ -115,8 +115,8 @@ extern int ftrace_update_ftrace_func(ftrace_func_t func); extern void ftrace_caller(void); extern void ftrace_call(void); extern void mcount_call(void); -#ifdef CONFIG_FUNCTION_RET_TRACER -extern void ftrace_return_caller(void); +#ifdef CONFIG_FUNCTION_GRAPH_TRACER +extern void ftrace_graph_caller(void); #endif /** @@ -315,7 +315,7 @@ ftrace_init_module(struct module *mod, /* * Structure that defines a return function trace. */ -struct ftrace_retfunc { +struct ftrace_graph_ret { unsigned long ret; /* Return address */ unsigned long func; /* Current function */ unsigned long long calltime; @@ -324,22 +324,22 @@ struct ftrace_retfunc { unsigned long overrun; }; -#ifdef CONFIG_FUNCTION_RET_TRACER +#ifdef CONFIG_FUNCTION_GRAPH_TRACER #define FTRACE_RETFUNC_DEPTH 50 #define FTRACE_RETSTACK_ALLOC_SIZE 32 /* Type of a callback handler of tracing return function */ -typedef void (*trace_function_return_t)(struct ftrace_retfunc *); +typedef void (*trace_function_graph_t)(struct ftrace_graph_ret *); -extern int register_ftrace_return(trace_function_return_t func); +extern int register_ftrace_graph(trace_function_graph_t func); /* The current handler in use */ -extern trace_function_return_t ftrace_function_return; -extern void unregister_ftrace_return(void); +extern trace_function_graph_t ftrace_graph_function; +extern void unregister_ftrace_graph(void); -extern void ftrace_retfunc_init_task(struct task_struct *t); -extern void ftrace_retfunc_exit_task(struct task_struct *t); +extern void ftrace_graph_init_task(struct task_struct *t); +extern void ftrace_graph_exit_task(struct task_struct *t); #else -static inline void ftrace_retfunc_init_task(struct task_struct *t) { } -static inline void ftrace_retfunc_exit_task(struct task_struct *t) { } +static inline void ftrace_graph_init_task(struct task_struct *t) { } +static inline void ftrace_graph_exit_task(struct task_struct *t) { } #endif #endif /* _LINUX_FTRACE_H */ diff --git a/include/linux/ftrace_irq.h b/include/linux/ftrace_irq.h index 0b4df55d7a74..366a054d0b05 100644 --- a/include/linux/ftrace_irq.h +++ b/include/linux/ftrace_irq.h @@ -2,7 +2,7 @@ #define _LINUX_FTRACE_IRQ_H -#if defined(CONFIG_DYNAMIC_FTRACE) || defined(CONFIG_FUNCTION_RET_TRACER) +#if defined(CONFIG_DYNAMIC_FTRACE) || defined(CONFIG_FUNCTION_GRAPH_TRACER) extern void ftrace_nmi_enter(void); extern void ftrace_nmi_exit(void); #else diff --git a/include/linux/sched.h b/include/linux/sched.h index d02a0ca70ee9..7ad48f2a2758 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1365,7 +1365,7 @@ struct task_struct { unsigned long default_timer_slack_ns; struct list_head *scm_work_list; -#ifdef CONFIG_FUNCTION_RET_TRACER +#ifdef CONFIG_FUNCTION_GRAPH_TRACER /* Index of current stored adress in ret_stack */ int curr_ret_stack; /* Stack of return addresses for return function tracing */ diff --git a/kernel/Makefile b/kernel/Makefile index 03a45e7e87b7..703cf3b7389c 100644 --- a/kernel/Makefile +++ b/kernel/Makefile @@ -21,7 +21,7 @@ CFLAGS_REMOVE_cgroup-debug.o = -pg CFLAGS_REMOVE_sched_clock.o = -pg CFLAGS_REMOVE_sched.o = -pg endif -ifdef CONFIG_FUNCTION_RET_TRACER +ifdef CONFIG_FUNCTION_GRAPH_TRACER CFLAGS_REMOVE_extable.o = -pg # For __kernel_text_address() CFLAGS_REMOVE_module.o = -pg # For __module_text_address() endif diff --git a/kernel/fork.c b/kernel/fork.c index d6e1a3205f62..5f82a999c032 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -140,7 +140,7 @@ void free_task(struct task_struct *tsk) prop_local_destroy_single(&tsk->dirties); free_thread_info(tsk->stack); rt_mutex_debug_task_free(tsk); - ftrace_retfunc_exit_task(tsk); + ftrace_graph_exit_task(tsk); free_task_struct(tsk); } EXPORT_SYMBOL(free_task); @@ -1271,7 +1271,7 @@ static struct task_struct *copy_process(unsigned long clone_flags, total_forks++; spin_unlock(¤t->sighand->siglock); write_unlock_irq(&tasklist_lock); - ftrace_retfunc_init_task(p); + ftrace_graph_init_task(p); proc_fork_connector(p); cgroup_post_fork(p); return p; diff --git a/kernel/sched.c b/kernel/sched.c index 388d9db044ab..52490bf6b884 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -5901,7 +5901,7 @@ void __cpuinit init_idle(struct task_struct *idle, int cpu) * The idle tasks have their own, simple scheduling class: */ idle->sched_class = &idle_sched_class; - ftrace_retfunc_init_task(idle); + ftrace_graph_init_task(idle); } /* diff --git a/kernel/trace/Kconfig b/kernel/trace/Kconfig index 620feadff67a..eb9b901e0777 100644 --- a/kernel/trace/Kconfig +++ b/kernel/trace/Kconfig @@ -12,7 +12,7 @@ config NOP_TRACER config HAVE_FUNCTION_TRACER bool -config HAVE_FUNCTION_RET_TRACER +config HAVE_FUNCTION_GRAPH_TRACER bool config HAVE_FUNCTION_TRACE_MCOUNT_TEST @@ -63,15 +63,18 @@ config FUNCTION_TRACER (the bootup default), then the overhead of the instructions is very small and not measurable even in micro-benchmarks. -config FUNCTION_RET_TRACER - bool "Kernel Function return Tracer" - depends on HAVE_FUNCTION_RET_TRACER +config FUNCTION_GRAPH_TRACER + bool "Kernel Function Graph Tracer" + depends on HAVE_FUNCTION_GRAPH_TRACER depends on FUNCTION_TRACER help - Enable the kernel to trace a function at its return. - It's first purpose is to trace the duration of functions. - This is done by setting the current return address on the thread - info structure of the current task. + Enable the kernel to trace a function at both its return + and its entry. + It's first purpose is to trace the duration of functions and + draw a call graph for each thread with some informations like + the return value. + This is done by setting the current return address on the current + task structure into a stack of calls. config IRQSOFF_TRACER bool "Interrupts-off Latency Tracer" diff --git a/kernel/trace/Makefile b/kernel/trace/Makefile index cef4bcb4e822..08c5fe6ddc09 100644 --- a/kernel/trace/Makefile +++ b/kernel/trace/Makefile @@ -29,7 +29,7 @@ obj-$(CONFIG_NOP_TRACER) += trace_nop.o obj-$(CONFIG_STACK_TRACER) += trace_stack.o obj-$(CONFIG_MMIOTRACE) += trace_mmiotrace.o obj-$(CONFIG_BOOT_TRACER) += trace_boot.o -obj-$(CONFIG_FUNCTION_RET_TRACER) += trace_functions_return.o +obj-$(CONFIG_FUNCTION_GRAPH_TRACER) += trace_functions_graph.o obj-$(CONFIG_TRACE_BRANCH_PROFILING) += trace_branch.o obj-$(CONFIG_BTS_TRACER) += trace_bts.o diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c index 53042f118f23..9e19976af727 100644 --- a/kernel/trace/ftrace.c +++ b/kernel/trace/ftrace.c @@ -395,11 +395,11 @@ __ftrace_replace_code(struct dyn_ftrace *rec, int enable) unsigned long ip, fl; unsigned long ftrace_addr; -#ifdef CONFIG_FUNCTION_RET_TRACER +#ifdef CONFIG_FUNCTION_GRAPH_TRACER if (ftrace_tracing_type == FTRACE_TYPE_ENTER) ftrace_addr = (unsigned long)ftrace_caller; else - ftrace_addr = (unsigned long)ftrace_return_caller; + ftrace_addr = (unsigned long)ftrace_graph_caller; #else ftrace_addr = (unsigned long)ftrace_caller; #endif @@ -1496,13 +1496,13 @@ ftrace_enable_sysctl(struct ctl_table *table, int write, return ret; } -#ifdef CONFIG_FUNCTION_RET_TRACER +#ifdef CONFIG_FUNCTION_GRAPH_TRACER static atomic_t ftrace_retfunc_active; /* The callback that hooks the return of a function */ -trace_function_return_t ftrace_function_return = - (trace_function_return_t)ftrace_stub; +trace_function_graph_t ftrace_graph_function = + (trace_function_graph_t)ftrace_stub; /* Try to assign a return stack array on FTRACE_RETSTACK_ALLOC_SIZE tasks. */ @@ -1549,7 +1549,7 @@ free: } /* Allocate a return stack for each task */ -static int start_return_tracing(void) +static int start_graph_tracing(void) { struct ftrace_ret_stack **ret_stack_list; int ret; @@ -1569,7 +1569,7 @@ static int start_return_tracing(void) return ret; } -int register_ftrace_return(trace_function_return_t func) +int register_ftrace_graph(trace_function_graph_t func) { int ret = 0; @@ -1584,13 +1584,13 @@ int register_ftrace_return(trace_function_return_t func) goto out; } atomic_inc(&ftrace_retfunc_active); - ret = start_return_tracing(); + ret = start_graph_tracing(); if (ret) { atomic_dec(&ftrace_retfunc_active); goto out; } ftrace_tracing_type = FTRACE_TYPE_RETURN; - ftrace_function_return = func; + ftrace_graph_function = func; ftrace_startup(); out: @@ -1598,12 +1598,12 @@ out: return ret; } -void unregister_ftrace_return(void) +void unregister_ftrace_graph(void) { mutex_lock(&ftrace_sysctl_lock); atomic_dec(&ftrace_retfunc_active); - ftrace_function_return = (trace_function_return_t)ftrace_stub; + ftrace_graph_function = (trace_function_graph_t)ftrace_stub; ftrace_shutdown(); /* Restore normal tracing type */ ftrace_tracing_type = FTRACE_TYPE_ENTER; @@ -1612,7 +1612,7 @@ void unregister_ftrace_return(void) } /* Allocate a return stack for newly created task */ -void ftrace_retfunc_init_task(struct task_struct *t) +void ftrace_graph_init_task(struct task_struct *t) { if (atomic_read(&ftrace_retfunc_active)) { t->ret_stack = kmalloc(FTRACE_RETFUNC_DEPTH @@ -1626,7 +1626,7 @@ void ftrace_retfunc_init_task(struct task_struct *t) t->ret_stack = NULL; } -void ftrace_retfunc_exit_task(struct task_struct *t) +void ftrace_graph_exit_task(struct task_struct *t) { struct ftrace_ret_stack *ret_stack = t->ret_stack; diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c index 8df8fdd69c95..f21ab2c68fd4 100644 --- a/kernel/trace/trace.c +++ b/kernel/trace/trace.c @@ -878,15 +878,15 @@ trace_function(struct trace_array *tr, struct trace_array_cpu *data, ring_buffer_unlock_commit(tr->buffer, event, irq_flags); } -#ifdef CONFIG_FUNCTION_RET_TRACER -static void __trace_function_return(struct trace_array *tr, +#ifdef CONFIG_FUNCTION_GRAPH_TRACER +static void __trace_function_graph(struct trace_array *tr, struct trace_array_cpu *data, - struct ftrace_retfunc *trace, + struct ftrace_graph_ret *trace, unsigned long flags, int pc) { struct ring_buffer_event *event; - struct ftrace_ret_entry *entry; + struct ftrace_graph_entry *entry; unsigned long irq_flags; if (unlikely(local_read(&__get_cpu_var(ftrace_cpu_disabled)))) @@ -1177,8 +1177,8 @@ function_trace_call(unsigned long ip, unsigned long parent_ip) local_irq_restore(flags); } -#ifdef CONFIG_FUNCTION_RET_TRACER -void trace_function_return(struct ftrace_retfunc *trace) +#ifdef CONFIG_FUNCTION_GRAPH_TRACER +void trace_function_graph(struct ftrace_graph_ret *trace) { struct trace_array *tr = &global_trace; struct trace_array_cpu *data; @@ -1193,12 +1193,12 @@ void trace_function_return(struct ftrace_retfunc *trace) disabled = atomic_inc_return(&data->disabled); if (likely(disabled == 1)) { pc = preempt_count(); - __trace_function_return(tr, data, trace, flags, pc); + __trace_function_graph(tr, data, trace, flags, pc); } atomic_dec(&data->disabled); raw_local_irq_restore(flags); } -#endif /* CONFIG_FUNCTION_RET_TRACER */ +#endif /* CONFIG_FUNCTION_GRAPH_TRACER */ static struct ftrace_ops trace_ops __read_mostly = { @@ -2001,7 +2001,7 @@ static enum print_line_t print_trace_fmt(struct trace_iterator *iter) break; } case TRACE_FN_RET: { - return print_return_function(iter); + return print_graph_function(iter); break; } case TRACE_BRANCH: { diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h index 3abd645e8af2..72b5ef868765 100644 --- a/kernel/trace/trace.h +++ b/kernel/trace/trace.h @@ -57,7 +57,7 @@ struct ftrace_entry { }; /* Function return entry */ -struct ftrace_ret_entry { +struct ftrace_graph_entry { struct trace_entry ent; unsigned long ip; unsigned long parent_ip; @@ -264,7 +264,7 @@ extern void __ftrace_bad_type(void); IF_ASSIGN(var, ent, struct trace_boot_call, TRACE_BOOT_CALL);\ IF_ASSIGN(var, ent, struct trace_boot_ret, TRACE_BOOT_RET);\ IF_ASSIGN(var, ent, struct trace_branch, TRACE_BRANCH); \ - IF_ASSIGN(var, ent, struct ftrace_ret_entry, TRACE_FN_RET);\ + IF_ASSIGN(var, ent, struct ftrace_graph_entry, TRACE_FN_RET);\ IF_ASSIGN(var, ent, struct bts_entry, TRACE_BTS);\ __ftrace_bad_type(); \ } while (0) @@ -398,7 +398,7 @@ void trace_function(struct trace_array *tr, unsigned long parent_ip, unsigned long flags, int pc); void -trace_function_return(struct ftrace_retfunc *trace); +trace_function_graph(struct ftrace_graph_ret *trace); void trace_bts(struct trace_array *tr, unsigned long from, @@ -489,11 +489,11 @@ extern int trace_vprintk(unsigned long ip, const char *fmt, va_list args); extern unsigned long trace_flags; /* Standard output formatting function used for function return traces */ -#ifdef CONFIG_FUNCTION_RET_TRACER -extern enum print_line_t print_return_function(struct trace_iterator *iter); +#ifdef CONFIG_FUNCTION_GRAPH_TRACER +extern enum print_line_t print_graph_function(struct trace_iterator *iter); #else static inline enum print_line_t -print_return_function(struct trace_iterator *iter) +print_graph_function(struct trace_iterator *iter) { return TRACE_TYPE_UNHANDLED; } diff --git a/kernel/trace/trace_functions_graph.c b/kernel/trace/trace_functions_graph.c new file mode 100644 index 000000000000..f5bad4624d2b --- /dev/null +++ b/kernel/trace/trace_functions_graph.c @@ -0,0 +1,98 @@ +/* + * + * Function graph tracer. + * Copyright (c) 2008 Frederic Weisbecker + * Mostly borrowed from function tracer which + * is Copyright (c) Steven Rostedt + * + */ +#include +#include +#include +#include + +#include "trace.h" + + +#define TRACE_GRAPH_PRINT_OVERRUN 0x1 +static struct tracer_opt trace_opts[] = { + /* Display overruns or not */ + { TRACER_OPT(overrun, TRACE_GRAPH_PRINT_OVERRUN) }, + { } /* Empty entry */ +}; + +static struct tracer_flags tracer_flags = { + .val = 0, /* Don't display overruns by default */ + .opts = trace_opts +}; + + +static int graph_trace_init(struct trace_array *tr) +{ + int cpu; + for_each_online_cpu(cpu) + tracing_reset(tr, cpu); + + return register_ftrace_graph(&trace_function_graph); +} + +static void graph_trace_reset(struct trace_array *tr) +{ + unregister_ftrace_graph(); +} + + +enum print_line_t +print_graph_function(struct trace_iterator *iter) +{ + struct trace_seq *s = &iter->seq; + struct trace_entry *entry = iter->ent; + struct ftrace_graph_entry *field; + int ret; + + if (entry->type == TRACE_FN_RET) { + trace_assign_type(field, entry); + ret = trace_seq_printf(s, "%pF -> ", (void *)field->parent_ip); + if (!ret) + return TRACE_TYPE_PARTIAL_LINE; + + ret = seq_print_ip_sym(s, field->ip, + trace_flags & TRACE_ITER_SYM_MASK); + if (!ret) + return TRACE_TYPE_PARTIAL_LINE; + + ret = trace_seq_printf(s, " (%llu ns)", + field->rettime - field->calltime); + if (!ret) + return TRACE_TYPE_PARTIAL_LINE; + + if (tracer_flags.val & TRACE_GRAPH_PRINT_OVERRUN) { + ret = trace_seq_printf(s, " (Overruns: %lu)", + field->overrun); + if (!ret) + return TRACE_TYPE_PARTIAL_LINE; + } + + ret = trace_seq_printf(s, "\n"); + if (!ret) + return TRACE_TYPE_PARTIAL_LINE; + + return TRACE_TYPE_HANDLED; + } + return TRACE_TYPE_UNHANDLED; +} + +static struct tracer graph_trace __read_mostly = { + .name = "function-graph", + .init = graph_trace_init, + .reset = graph_trace_reset, + .print_line = print_graph_function, + .flags = &tracer_flags, +}; + +static __init int init_graph_trace(void) +{ + return register_tracer(&graph_trace); +} + +device_initcall(init_graph_trace); -- cgit v1.2.3 From bf4d83f66476086c6b50dc52aac00d71ad70494e Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Tue, 25 Nov 2008 09:57:51 +1030 Subject: sched: convert nohz struct to cpumask_var_t, fix Impact: build fix Fix the !CONFIG_SMP case. Signed-off-by: Rusty Russell Acked-by: Mike Travis Signed-off-by: Ingo Molnar --- kernel/sched.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index eba6a156d334..1aa840a9f585 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -8349,10 +8349,12 @@ void __init sched_init(void) /* Allocate the nohz_cpu_mask if CONFIG_CPUMASK_OFFSTACK */ alloc_bootmem_cpumask_var(&nohz_cpu_mask); +#ifdef CONFIG_SMP #ifdef CONFIG_NO_HZ alloc_bootmem_cpumask_var(&nohz.cpu_mask); #endif alloc_bootmem_cpumask_var(&cpu_isolated_map); +#endif /* SMP */ scheduler_running = 1; } -- cgit v1.2.3 From 1224e376f2a7e3c7ab19ef37099a78597978a696 Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Tue, 25 Nov 2008 09:59:20 +1030 Subject: sched: avoid stack var in move_task_off_dead_cpu, fix Impact: locking fix We can't call cpuset_cpus_allowed_locked() with the rq lock held. However, the rq lock merely protects us from (1) cpu_online_mask changing and (2) someone else changing p->cpus_allowed. The first can't happen because we're being called from a cpu hotplug notifier. The second doesn't really matter: we are forcing the task off a CPU it was affine to, so we're not doing very well anyway. So we remove the rq lock from this path, and all is good. Signed-off-by: Rusty Russell Acked-by: Mike Travis Signed-off-by: Ingo Molnar --- kernel/sched.c | 4 ---- 1 file changed, 4 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index 1aa840a9f585..3f5bfdc3d94d 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -6126,8 +6126,6 @@ static int __migrate_task_irq(struct task_struct *p, int src_cpu, int dest_cpu) */ static void move_task_off_dead_cpu(int dead_cpu, struct task_struct *p) { - unsigned long flags; - struct rq *rq; int dest_cpu; /* FIXME: Use cpumask_of_node here. */ cpumask_t _nodemask = node_to_cpumask(cpu_to_node(dead_cpu)); @@ -6146,10 +6144,8 @@ again: /* No more Mr. Nice Guy. */ if (dest_cpu >= nr_cpu_ids) { - rq = task_rq_lock(p, &flags); cpuset_cpus_allowed_locked(p, &p->cpus_allowed); dest_cpu = cpumask_any_and(cpu_online_mask, &p->cpus_allowed); - task_rq_unlock(rq, &flags); /* * Don't tell them about moving exiting tasks or -- cgit v1.2.3 From 70574a996fc7a70c5586eb56bd92a544eccf18b6 Mon Sep 17 00:00:00 2001 From: Alexey Dobriyan Date: Fri, 28 Nov 2008 22:08:00 +0300 Subject: sched: move double_unlock_balance() higher Move double_lock_balance()/double_unlock_balance() higher to fix the following with gcc-3.4.6: CC kernel/sched.o In file included from kernel/sched.c:1605: kernel/sched_rt.c: In function `find_lock_lowest_rq': kernel/sched_rt.c:914: sorry, unimplemented: inlining failed in call to 'double_unlock_balance': function body not available kernel/sched_rt.c:1077: sorry, unimplemented: called from here make[2]: *** [kernel/sched.o] Error 1 Signed-off-by: Alexey Dobriyan Signed-off-by: Ingo Molnar --- kernel/sched.c | 67 +++++++++++++++++++++++++++---------------------------- kernel/sched_rt.c | 4 ---- 2 files changed, 33 insertions(+), 38 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index 3d1ee429219b..6a99703e0eb0 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -1581,6 +1581,39 @@ static inline void update_shares_locked(struct rq *rq, struct sched_domain *sd) #endif +/* + * double_lock_balance - lock the busiest runqueue, this_rq is locked already. + */ +static int double_lock_balance(struct rq *this_rq, struct rq *busiest) + __releases(this_rq->lock) + __acquires(busiest->lock) + __acquires(this_rq->lock) +{ + int ret = 0; + + if (unlikely(!irqs_disabled())) { + /* printk() doesn't work good under rq->lock */ + spin_unlock(&this_rq->lock); + BUG_ON(1); + } + if (unlikely(!spin_trylock(&busiest->lock))) { + if (busiest < this_rq) { + spin_unlock(&this_rq->lock); + spin_lock(&busiest->lock); + spin_lock_nested(&this_rq->lock, SINGLE_DEPTH_NESTING); + ret = 1; + } else + spin_lock_nested(&busiest->lock, SINGLE_DEPTH_NESTING); + } + return ret; +} + +static inline void double_unlock_balance(struct rq *this_rq, struct rq *busiest) + __releases(busiest->lock) +{ + spin_unlock(&busiest->lock); + lock_set_subclass(&this_rq->lock.dep_map, 0, _RET_IP_); +} #endif #ifdef CONFIG_FAIR_GROUP_SCHED @@ -2780,40 +2813,6 @@ static void double_rq_unlock(struct rq *rq1, struct rq *rq2) __release(rq2->lock); } -/* - * double_lock_balance - lock the busiest runqueue, this_rq is locked already. - */ -static int double_lock_balance(struct rq *this_rq, struct rq *busiest) - __releases(this_rq->lock) - __acquires(busiest->lock) - __acquires(this_rq->lock) -{ - int ret = 0; - - if (unlikely(!irqs_disabled())) { - /* printk() doesn't work good under rq->lock */ - spin_unlock(&this_rq->lock); - BUG_ON(1); - } - if (unlikely(!spin_trylock(&busiest->lock))) { - if (busiest < this_rq) { - spin_unlock(&this_rq->lock); - spin_lock(&busiest->lock); - spin_lock_nested(&this_rq->lock, SINGLE_DEPTH_NESTING); - ret = 1; - } else - spin_lock_nested(&busiest->lock, SINGLE_DEPTH_NESTING); - } - return ret; -} - -static inline void double_unlock_balance(struct rq *this_rq, struct rq *busiest) - __releases(busiest->lock) -{ - spin_unlock(&busiest->lock); - lock_set_subclass(&this_rq->lock.dep_map, 0, _RET_IP_); -} - /* * If dest_cpu is allowed for this process, migrate the task to it. * This is accomplished by forcing the cpu_allowed mask to only diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c index 2bdd44423599..587a16e2a8f5 100644 --- a/kernel/sched_rt.c +++ b/kernel/sched_rt.c @@ -909,10 +909,6 @@ static void put_prev_task_rt(struct rq *rq, struct task_struct *p) /* Only try algorithms three times */ #define RT_MAX_TRIES 3 -static int double_lock_balance(struct rq *this_rq, struct rq *busiest); -static inline void double_unlock_balance(struct rq *this_rq, - struct rq *busiest); - static void deactivate_task(struct rq *rq, struct task_struct *p, int sleep); static int pick_rt_task(struct rq *rq, struct task_struct *p, int cpu) -- cgit v1.2.3 From 6c415b9234a8c71f290e5d4fddc467f103f32719 Mon Sep 17 00:00:00 2001 From: Arun R Bharadwaj Date: Mon, 1 Dec 2008 20:49:05 +0530 Subject: sched: add uid information to sched_debug for CONFIG_USER_SCHED Impact: extend information in /proc/sched_debug This patch adds uid information in sched_debug for CONFIG_USER_SCHED Signed-off-by: Arun R Bharadwaj Acked-by: Peter Zijlstra Signed-off-by: Ingo Molnar --- include/linux/sched.h | 1 + kernel/sched.c | 10 ++++++++++ kernel/sched_debug.c | 6 +++++- kernel/user.c | 2 ++ 4 files changed, 18 insertions(+), 1 deletion(-) (limited to 'kernel/sched.c') diff --git a/include/linux/sched.h b/include/linux/sched.h index 7a69c4d224ee..d8733f07d80b 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -2218,6 +2218,7 @@ extern void normalize_rt_tasks(void); extern struct task_group init_task_group; #ifdef CONFIG_USER_SCHED extern struct task_group root_task_group; +extern void set_tg_uid(struct user_struct *user); #endif extern struct task_group *sched_create_group(struct task_group *parent); diff --git a/kernel/sched.c b/kernel/sched.c index 6a99703e0eb0..4c7388ef5be7 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -261,6 +261,10 @@ struct task_group { struct cgroup_subsys_state css; #endif +#ifdef CONFIG_USER_SCHED + uid_t uid; +#endif + #ifdef CONFIG_FAIR_GROUP_SCHED /* schedulable entities of this group on each cpu */ struct sched_entity **se; @@ -286,6 +290,12 @@ struct task_group { #ifdef CONFIG_USER_SCHED +/* Helper function to pass uid information to create_sched_user() */ +void set_tg_uid(struct user_struct *user) +{ + user->tg->uid = user->uid; +} + /* * Root task group. * Every UID task group (including init_task_group aka UID-0) will diff --git a/kernel/sched_debug.c b/kernel/sched_debug.c index baf2f17af462..4293cfa9681d 100644 --- a/kernel/sched_debug.c +++ b/kernel/sched_debug.c @@ -160,10 +160,14 @@ void print_cfs_rq(struct seq_file *m, int cpu, struct cfs_rq *cfs_rq) cgroup_path(tg->css.cgroup, path, sizeof(path)); SEQ_printf(m, "\ncfs_rq[%d]:%s\n", cpu, path); +#elif defined(CONFIG_USER_SCHED) && defined(CONFIG_FAIR_GROUP_SCHED) + { + uid_t uid = cfs_rq->tg->uid; + SEQ_printf(m, "\ncfs_rq[%d] for UID: %u\n", cpu, uid); + } #else SEQ_printf(m, "\ncfs_rq[%d]:\n", cpu); #endif - SEQ_printf(m, " .%-30s: %Ld.%06ld\n", "exec_clock", SPLIT_NS(cfs_rq->exec_clock)); diff --git a/kernel/user.c b/kernel/user.c index 39d6159fae43..cec2224bc9f5 100644 --- a/kernel/user.c +++ b/kernel/user.c @@ -101,6 +101,8 @@ static int sched_create_user(struct user_struct *up) if (IS_ERR(up->tg)) rc = -ENOMEM; + set_tg_uid(up); + return rc; } -- cgit v1.2.3 From 5436499e6098759c2340f8b906ea52f993dc4efb Mon Sep 17 00:00:00 2001 From: Ken Chen Date: Sun, 7 Dec 2008 18:47:37 -0800 Subject: sched: fix sd_parent_degenerate on non-numa smp machine Impact: optimize the sched domains tree some more The addition of SD_SERIALIZE flag added to SD_NODE_INIT prevented top level dummy numa sched_domain to be properly degenerated on non-numa smp machine. The reason is that in sd_parent_degenerate(), it found that the child and parent does not have comon sched_domain flags due to SD_SERIALIZE. However, for non-numa smp box, the top level is a dummy with a single sched_group. Filter out SD_SERIALIZE if it is on non-numa machine to properly degenerate top level node sched_domain. this will cut back some of the sd domain walk in the load balancer code. Signed-off-by: Ken Chen Acked-by: Peter Zijlstra Signed-off-by: Ingo Molnar --- kernel/sched.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index 152828239ef0..74498c840f93 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -6768,6 +6768,8 @@ sd_parent_degenerate(struct sched_domain *sd, struct sched_domain *parent) SD_BALANCE_EXEC | SD_SHARE_CPUPOWER | SD_SHARE_PKG_RESOURCES); + if (nr_node_ids == 1) + pflags &= ~SD_SERIALIZE; } if (~cflags & pflags) return 0; -- cgit v1.2.3 From efbe027e95dc13ac343b6130948418d7ead7ddf1 Mon Sep 17 00:00:00 2001 From: Vaidyanathan Srinivasan Date: Mon, 8 Dec 2008 20:52:49 +0530 Subject: sched: idle_balance() does not call load_balance_newidle() Impact: fix SD_BALANCE_NEWIDLEand broaden its use load_balance_newidle() does not get called if SD_BALANCE_NEWIDLE is set at higher level domain (3-CPU) and not in low level domain (2-MC). pulled_task is initialised to -1 and checked for non-zero which is always true if the lowest level sched_domain does not have SD_BALANCE_NEWIDLE flag set. Signed-off-by: Vaidyanathan Srinivasan Signed-off-by: Ingo Molnar --- kernel/sched.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index 74498c840f93..bb9c6384d077 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -3685,7 +3685,7 @@ out_balanced: static void idle_balance(int this_cpu, struct rq *this_rq) { struct sched_domain *sd; - int pulled_task = -1; + int pulled_task = 0; unsigned long next_balance = jiffies + HZ; cpumask_t tmpmask; -- cgit v1.2.3 From ee79d1bdb6a10499e53f80b1e8d14110215178ba Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Tue, 9 Dec 2008 18:49:50 +0100 Subject: sched: let arch_update_cpu_topology indicate if topology changed Change arch_update_cpu_topology so it returns 1 if the cpu topology changed and 0 if it didn't change. This will be useful for the next patch which adds a call to this function in partition_sched_domains. Signed-off-by: Heiko Carstens Signed-off-by: Ingo Molnar --- arch/s390/kernel/topology.c | 5 +++-- include/linux/topology.h | 2 +- kernel/sched.c | 8 +++++++- 3 files changed, 11 insertions(+), 4 deletions(-) (limited to 'kernel/sched.c') diff --git a/arch/s390/kernel/topology.c b/arch/s390/kernel/topology.c index a947899dcba1..bf96f1b5c6ec 100644 --- a/arch/s390/kernel/topology.c +++ b/arch/s390/kernel/topology.c @@ -212,7 +212,7 @@ static void update_cpu_core_map(void) cpu_core_map[cpu] = cpu_coregroup_map(cpu); } -void arch_update_cpu_topology(void) +int arch_update_cpu_topology(void) { struct tl_info *info = tl_info; struct sys_device *sysdev; @@ -221,7 +221,7 @@ void arch_update_cpu_topology(void) if (!machine_has_topology) { update_cpu_core_map(); topology_update_polarization_simple(); - return; + return 0; } stsi(info, 15, 1, 2); tl_to_cores(info); @@ -230,6 +230,7 @@ void arch_update_cpu_topology(void) sysdev = get_cpu_sysdev(cpu); kobject_uevent(&sysdev->kobj, KOBJ_CHANGE); } + return 1; } static void topology_work_fn(struct work_struct *work) diff --git a/include/linux/topology.h b/include/linux/topology.h index 117f1b7405cf..0c5b5ac36d8e 100644 --- a/include/linux/topology.h +++ b/include/linux/topology.h @@ -49,7 +49,7 @@ for_each_online_node(node) \ if (nr_cpus_node(node)) -void arch_update_cpu_topology(void); +int arch_update_cpu_topology(void); /* Conform to ACPI 2.0 SLIT distance definitions */ #define LOCAL_DISTANCE 10 diff --git a/kernel/sched.c b/kernel/sched.c index ef212da928e8..fcfbbd9dbd60 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -7675,8 +7675,14 @@ static struct sched_domain_attr *dattr_cur; */ static cpumask_t fallback_doms; -void __attribute__((weak)) arch_update_cpu_topology(void) +/* + * arch_update_cpu_topology lets virtualized architectures update the + * cpu core maps. It is supposed to return 1 if the topology changed + * or 0 if it stayed the same. + */ +int __attribute__((weak)) arch_update_cpu_topology(void) { + return 0; } /* -- cgit v1.2.3 From d65bd5ecb2bd166cea4952a59b7e16cc3ad6ef6c Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Tue, 9 Dec 2008 18:49:51 +0100 Subject: sched: add missing arch_update_cpu_topology() call arch_reinit_sched_domains() used to call arch_update_cpu_topology() via arch_init_sched_domains(). This call got lost with e761b7725234276a802322549cee5255305a0930 ("cpu hotplug, sched: Introduce cpu_active_map and redo sched domain managment (take 2)". So we might end up with outdated and missing cpus in the cpu core maps (architecture used to call arch_reinit_sched_domains if cpu topology changed). This adds a call to arch_update_cpu_topology in partition_sched_domains which gets called whenever scheduling domains get updated. Which is what is supposed to happen when cpu topology changes. Signed-off-by: Heiko Carstens Signed-off-by: Ingo Molnar --- kernel/sched.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) (limited to 'kernel/sched.c') diff --git a/kernel/sched.c b/kernel/sched.c index fcfbbd9dbd60..ad7b93be5691 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -7774,17 +7774,21 @@ void partition_sched_domains(int ndoms_new, cpumask_t *doms_new, struct sched_domain_attr *dattr_new) { int i, j, n; + int new_topology; mutex_lock(&sched_domains_mutex); /* always unregister in case we don't destroy any domains */ unregister_sched_domain_sysctl(); + /* Let architecture update cpu core mappings. */ + new_topology = arch_update_cpu_topology(); + n = doms_new ? ndoms_new : 0; /* Destroy deleted domains */ for (i = 0; i < ndoms_cur; i++) { - for (j = 0; j < n; j++) { + for (j = 0; j < n && !new_topology; j++) { if (cpus_equal(doms_cur[i], doms_new[j]) && dattrs_equal(dattr_cur, i, dattr_new, j)) goto match1; @@ -7804,7 +7808,7 @@ match1: /* Build new domains */ for (i = 0; i < ndoms_new; i++) { - for (j = 0; j < ndoms_cur; j++) { + for (j = 0; j < ndoms_cur && !new_topology; j++) { if (cpus_equal(doms_new[i], doms_cur[j]) && dattrs_equal(dattr_new, i, dattr_cur, j)) goto match2; -- cgit v1.2.3