diff options
author | Huang, Tao <huangtao@rock-chips.com> | 2016-11-04 14:23:25 +0800 |
---|---|---|
committer | Huang, Tao <huangtao@rock-chips.com> | 2016-11-04 14:30:24 +0800 |
commit | f9ae5d202b3953b5d69e860e540a6f53df7015b5 (patch) | |
tree | 92585302aa605d32b2e4782c7b4f700b1a5c2c35 /kernel | |
parent | 1429ad62bef1098f1cec26447cdc12934879a23c (diff) | |
parent | 79df8fa79b6a2aced892ad2b2c9832e7d9bdea6b (diff) |
Merge branch 'linux-linaro-lsk-v4.4-android' of git://git.linaro.org/kernel/linux-linaro-stable.git
* linux-linaro-lsk-v4.4-android: (1362 commits)
Linux 4.4.30
Revert "fix minor infoleak in get_user_ex()"
Revert "x86/mm: Expand the exception table logic to allow new handling options"
Linux 4.4.29
ARM: pxa: pxa_cplds: fix interrupt handling
powerpc/nvram: Fix an incorrect partition merge
mpt3sas: Don't spam logs if logging level is 0
perf symbols: Fixup symbol sizes before picking best ones
perf symbols: Check symbol_conf.allow_aliases for kallsyms loading too
perf hists browser: Fix event group display
clk: divider: Fix clk_divider_round_rate() to use clk_readl()
clk: qoriq: fix a register offset error
s390/con3270: fix insufficient space padding
s390/con3270: fix use of uninitialised data
s390/cio: fix accidental interrupt enabling during resume
x86/mm: Expand the exception table logic to allow new handling options
dmaengine: ipu: remove bogus NO_IRQ reference
power: bq24257: Fix use of uninitialized pointer bq->charger
staging: r8188eu: Fix scheduling while atomic splat
ASoC: dapm: Fix kcontrol creation for output driver widget
...
Diffstat (limited to 'kernel')
47 files changed, 3391 insertions, 770 deletions
diff --git a/kernel/audit.c b/kernel/audit.c index 5ffcbd354a52..34f690b9213a 100644 --- a/kernel/audit.c +++ b/kernel/audit.c @@ -870,6 +870,12 @@ static int audit_receive_msg(struct sk_buff *skb, struct nlmsghdr *nlh) return err; } if (s.mask & AUDIT_STATUS_PID) { + /* NOTE: we are using task_tgid_vnr() below because + * the s.pid value is relative to the namespace + * of the caller; at present this doesn't matter + * much since you can really only run auditd + * from the initial pid namespace, but something + * to keep in mind if this changes */ int new_pid = s.pid; if ((!new_pid) && (task_tgid_vnr(current) != audit_pid)) @@ -1896,7 +1902,7 @@ void audit_log_task_info(struct audit_buffer *ab, struct task_struct *tsk) " euid=%u suid=%u fsuid=%u" " egid=%u sgid=%u fsgid=%u tty=%s ses=%u", task_ppid_nr(tsk), - task_pid_nr(tsk), + task_tgid_nr(tsk), from_kuid(&init_user_ns, audit_get_loginuid(tsk)), from_kuid(&init_user_ns, cred->uid), from_kgid(&init_user_ns, cred->gid), diff --git a/kernel/audit_watch.c b/kernel/audit_watch.c index 656c7e93ac0d..939945a5649c 100644 --- a/kernel/audit_watch.c +++ b/kernel/audit_watch.c @@ -19,6 +19,7 @@ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */ +#include <linux/file.h> #include <linux/kernel.h> #include <linux/audit.h> #include <linux/kthread.h> @@ -544,10 +545,11 @@ int audit_exe_compare(struct task_struct *tsk, struct audit_fsnotify_mark *mark) unsigned long ino; dev_t dev; - rcu_read_lock(); - exe_file = rcu_dereference(tsk->mm->exe_file); + exe_file = get_task_exe_file(tsk); + if (!exe_file) + return 0; ino = exe_file->f_inode->i_ino; dev = exe_file->f_inode->i_sb->s_dev; - rcu_read_unlock(); + fput(exe_file); return audit_mark_compare(mark, ino, dev); } diff --git a/kernel/auditsc.c b/kernel/auditsc.c index b86cc04959de..63f0e495f517 100644 --- a/kernel/auditsc.c +++ b/kernel/auditsc.c @@ -73,6 +73,7 @@ #include <linux/compat.h> #include <linux/ctype.h> #include <linux/string.h> +#include <linux/uaccess.h> #include <uapi/linux/limits.h> #include "audit.h" @@ -82,7 +83,8 @@ #define AUDITSC_SUCCESS 1 #define AUDITSC_FAILURE 2 -/* no execve audit message should be longer than this (userspace limits) */ +/* no execve audit message should be longer than this (userspace limits), + * see the note near the top of audit_log_execve_info() about this value */ #define MAX_EXECVE_AUDIT_LEN 7500 /* max length to print of cmdline/proctitle value during audit */ @@ -456,7 +458,7 @@ static int audit_filter_rules(struct task_struct *tsk, switch (f->type) { case AUDIT_PID: - pid = task_pid_nr(tsk); + pid = task_tgid_nr(tsk); result = audit_comparator(pid, f->op, f->val); break; case AUDIT_PPID: @@ -988,184 +990,178 @@ static int audit_log_pid_context(struct audit_context *context, pid_t pid, return rc; } -/* - * to_send and len_sent accounting are very loose estimates. We aren't - * really worried about a hard cap to MAX_EXECVE_AUDIT_LEN so much as being - * within about 500 bytes (next page boundary) - * - * why snprintf? an int is up to 12 digits long. if we just assumed when - * logging that a[%d]= was going to be 16 characters long we would be wasting - * space in every audit message. In one 7500 byte message we can log up to - * about 1000 min size arguments. That comes down to about 50% waste of space - * if we didn't do the snprintf to find out how long arg_num_len was. - */ -static int audit_log_single_execve_arg(struct audit_context *context, - struct audit_buffer **ab, - int arg_num, - size_t *len_sent, - const char __user *p, - char *buf) +static void audit_log_execve_info(struct audit_context *context, + struct audit_buffer **ab) { - char arg_num_len_buf[12]; - const char __user *tmp_p = p; - /* how many digits are in arg_num? 5 is the length of ' a=""' */ - size_t arg_num_len = snprintf(arg_num_len_buf, 12, "%d", arg_num) + 5; - size_t len, len_left, to_send; - size_t max_execve_audit_len = MAX_EXECVE_AUDIT_LEN; - unsigned int i, has_cntl = 0, too_long = 0; - int ret; - - /* strnlen_user includes the null we don't want to send */ - len_left = len = strnlen_user(p, MAX_ARG_STRLEN) - 1; - - /* - * We just created this mm, if we can't find the strings - * we just copied into it something is _very_ wrong. Similar - * for strings that are too long, we should not have created - * any. - */ - if (WARN_ON_ONCE(len < 0 || len > MAX_ARG_STRLEN - 1)) { - send_sig(SIGKILL, current, 0); - return -1; + long len_max; + long len_rem; + long len_full; + long len_buf; + long len_abuf; + long len_tmp; + bool require_data; + bool encode; + unsigned int iter; + unsigned int arg; + char *buf_head; + char *buf; + const char __user *p = (const char __user *)current->mm->arg_start; + + /* NOTE: this buffer needs to be large enough to hold all the non-arg + * data we put in the audit record for this argument (see the + * code below) ... at this point in time 96 is plenty */ + char abuf[96]; + + /* NOTE: we set MAX_EXECVE_AUDIT_LEN to a rather arbitrary limit, the + * current value of 7500 is not as important as the fact that it + * is less than 8k, a setting of 7500 gives us plenty of wiggle + * room if we go over a little bit in the logging below */ + WARN_ON_ONCE(MAX_EXECVE_AUDIT_LEN > 7500); + len_max = MAX_EXECVE_AUDIT_LEN; + + /* scratch buffer to hold the userspace args */ + buf_head = kmalloc(MAX_EXECVE_AUDIT_LEN + 1, GFP_KERNEL); + if (!buf_head) { + audit_panic("out of memory for argv string"); + return; } + buf = buf_head; - /* walk the whole argument looking for non-ascii chars */ + audit_log_format(*ab, "argc=%d", context->execve.argc); + + len_rem = len_max; + len_buf = 0; + len_full = 0; + require_data = true; + encode = false; + iter = 0; + arg = 0; do { - if (len_left > MAX_EXECVE_AUDIT_LEN) - to_send = MAX_EXECVE_AUDIT_LEN; - else - to_send = len_left; - ret = copy_from_user(buf, tmp_p, to_send); - /* - * There is no reason for this copy to be short. We just - * copied them here, and the mm hasn't been exposed to user- - * space yet. - */ - if (ret) { - WARN_ON(1); - send_sig(SIGKILL, current, 0); - return -1; - } - buf[to_send] = '\0'; - has_cntl = audit_string_contains_control(buf, to_send); - if (has_cntl) { - /* - * hex messages get logged as 2 bytes, so we can only - * send half as much in each message - */ - max_execve_audit_len = MAX_EXECVE_AUDIT_LEN / 2; - break; - } - len_left -= to_send; - tmp_p += to_send; - } while (len_left > 0); - - len_left = len; - - if (len > max_execve_audit_len) - too_long = 1; - - /* rewalk the argument actually logging the message */ - for (i = 0; len_left > 0; i++) { - int room_left; - - if (len_left > max_execve_audit_len) - to_send = max_execve_audit_len; - else - to_send = len_left; - - /* do we have space left to send this argument in this ab? */ - room_left = MAX_EXECVE_AUDIT_LEN - arg_num_len - *len_sent; - if (has_cntl) - room_left -= (to_send * 2); - else - room_left -= to_send; - if (room_left < 0) { - *len_sent = 0; - audit_log_end(*ab); - *ab = audit_log_start(context, GFP_KERNEL, AUDIT_EXECVE); - if (!*ab) - return 0; - } + /* NOTE: we don't ever want to trust this value for anything + * serious, but the audit record format insists we + * provide an argument length for really long arguments, + * e.g. > MAX_EXECVE_AUDIT_LEN, so we have no choice but + * to use strncpy_from_user() to obtain this value for + * recording in the log, although we don't use it + * anywhere here to avoid a double-fetch problem */ + if (len_full == 0) + len_full = strnlen_user(p, MAX_ARG_STRLEN) - 1; + + /* read more data from userspace */ + if (require_data) { + /* can we make more room in the buffer? */ + if (buf != buf_head) { + memmove(buf_head, buf, len_buf); + buf = buf_head; + } + + /* fetch as much as we can of the argument */ + len_tmp = strncpy_from_user(&buf_head[len_buf], p, + len_max - len_buf); + if (len_tmp == -EFAULT) { + /* unable to copy from userspace */ + send_sig(SIGKILL, current, 0); + goto out; + } else if (len_tmp == (len_max - len_buf)) { + /* buffer is not large enough */ + require_data = true; + /* NOTE: if we are going to span multiple + * buffers force the encoding so we stand + * a chance at a sane len_full value and + * consistent record encoding */ + encode = true; + len_full = len_full * 2; + p += len_tmp; + } else { + require_data = false; + if (!encode) + encode = audit_string_contains_control( + buf, len_tmp); + /* try to use a trusted value for len_full */ + if (len_full < len_max) + len_full = (encode ? + len_tmp * 2 : len_tmp); + p += len_tmp + 1; + } + len_buf += len_tmp; + buf_head[len_buf] = '\0'; - /* - * first record needs to say how long the original string was - * so we can be sure nothing was lost. - */ - if ((i == 0) && (too_long)) - audit_log_format(*ab, " a%d_len=%zu", arg_num, - has_cntl ? 2*len : len); - - /* - * normally arguments are small enough to fit and we already - * filled buf above when we checked for control characters - * so don't bother with another copy_from_user - */ - if (len >= max_execve_audit_len) - ret = copy_from_user(buf, p, to_send); - else - ret = 0; - if (ret) { - WARN_ON(1); - send_sig(SIGKILL, current, 0); - return -1; + /* length of the buffer in the audit record? */ + len_abuf = (encode ? len_buf * 2 : len_buf + 2); } - buf[to_send] = '\0'; - - /* actually log it */ - audit_log_format(*ab, " a%d", arg_num); - if (too_long) - audit_log_format(*ab, "[%d]", i); - audit_log_format(*ab, "="); - if (has_cntl) - audit_log_n_hex(*ab, buf, to_send); - else - audit_log_string(*ab, buf); - - p += to_send; - len_left -= to_send; - *len_sent += arg_num_len; - if (has_cntl) - *len_sent += to_send * 2; - else - *len_sent += to_send; - } - /* include the null we didn't log */ - return len + 1; -} -static void audit_log_execve_info(struct audit_context *context, - struct audit_buffer **ab) -{ - int i, len; - size_t len_sent = 0; - const char __user *p; - char *buf; + /* write as much as we can to the audit log */ + if (len_buf > 0) { + /* NOTE: some magic numbers here - basically if we + * can't fit a reasonable amount of data into the + * existing audit buffer, flush it and start with + * a new buffer */ + if ((sizeof(abuf) + 8) > len_rem) { + len_rem = len_max; + audit_log_end(*ab); + *ab = audit_log_start(context, + GFP_KERNEL, AUDIT_EXECVE); + if (!*ab) + goto out; + } - p = (const char __user *)current->mm->arg_start; + /* create the non-arg portion of the arg record */ + len_tmp = 0; + if (require_data || (iter > 0) || + ((len_abuf + sizeof(abuf)) > len_rem)) { + if (iter == 0) { + len_tmp += snprintf(&abuf[len_tmp], + sizeof(abuf) - len_tmp, + " a%d_len=%lu", + arg, len_full); + } + len_tmp += snprintf(&abuf[len_tmp], + sizeof(abuf) - len_tmp, + " a%d[%d]=", arg, iter++); + } else + len_tmp += snprintf(&abuf[len_tmp], + sizeof(abuf) - len_tmp, + " a%d=", arg); + WARN_ON(len_tmp >= sizeof(abuf)); + abuf[sizeof(abuf) - 1] = '\0'; + + /* log the arg in the audit record */ + audit_log_format(*ab, "%s", abuf); + len_rem -= len_tmp; + len_tmp = len_buf; + if (encode) { + if (len_abuf > len_rem) + len_tmp = len_rem / 2; /* encoding */ + audit_log_n_hex(*ab, buf, len_tmp); + len_rem -= len_tmp * 2; + len_abuf -= len_tmp * 2; + } else { + if (len_abuf > len_rem) + len_tmp = len_rem - 2; /* quotes */ + audit_log_n_string(*ab, buf, len_tmp); + len_rem -= len_tmp + 2; + /* don't subtract the "2" because we still need + * to add quotes to the remaining string */ + len_abuf -= len_tmp; + } + len_buf -= len_tmp; + buf += len_tmp; + } - audit_log_format(*ab, "argc=%d", context->execve.argc); + /* ready to move to the next argument? */ + if ((len_buf == 0) && !require_data) { + arg++; + iter = 0; + len_full = 0; + require_data = true; + encode = false; + } + } while (arg < context->execve.argc); - /* - * we need some kernel buffer to hold the userspace args. Just - * allocate one big one rather than allocating one of the right size - * for every single argument inside audit_log_single_execve_arg() - * should be <8k allocation so should be pretty safe. - */ - buf = kmalloc(MAX_EXECVE_AUDIT_LEN + 1, GFP_KERNEL); - if (!buf) { - audit_panic("out of memory for argv string"); - return; - } + /* NOTE: the caller handles the final audit_log_end() call */ - for (i = 0; i < context->execve.argc; i++) { - len = audit_log_single_execve_arg(context, ab, i, - &len_sent, p, buf); - if (len <= 0) - break; - p += len; - } - kfree(buf); +out: + kfree(buf_head); } static void show_special(struct audit_context *context, int *call_panic) @@ -1991,7 +1987,7 @@ static void audit_log_set_loginuid(kuid_t koldloginuid, kuid_t kloginuid, ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_LOGIN); if (!ab) return; - audit_log_format(ab, "pid=%d uid=%u", task_pid_nr(current), uid); + audit_log_format(ab, "pid=%d uid=%u", task_tgid_nr(current), uid); audit_log_task_context(ab); audit_log_format(ab, " old-auid=%u auid=%u old-ses=%u ses=%u res=%d", oldloginuid, loginuid, oldsessionid, sessionid, !rc); @@ -2216,7 +2212,7 @@ void __audit_ptrace(struct task_struct *t) { struct audit_context *context = current->audit_context; - context->target_pid = task_pid_nr(t); + context->target_pid = task_tgid_nr(t); context->target_auid = audit_get_loginuid(t); context->target_uid = task_uid(t); context->target_sessionid = audit_get_sessionid(t); @@ -2241,7 +2237,7 @@ int __audit_signal_info(int sig, struct task_struct *t) if (audit_pid && t->tgid == audit_pid) { if (sig == SIGTERM || sig == SIGHUP || sig == SIGUSR1 || sig == SIGUSR2) { - audit_sig_pid = task_pid_nr(tsk); + audit_sig_pid = task_tgid_nr(tsk); if (uid_valid(tsk->loginuid)) audit_sig_uid = tsk->loginuid; else @@ -2341,7 +2337,7 @@ int __audit_log_bprm_fcaps(struct linux_binprm *bprm, void __audit_log_capset(const struct cred *new, const struct cred *old) { struct audit_context *context = current->audit_context; - context->capset.pid = task_pid_nr(current); + context->capset.pid = task_tgid_nr(current); context->capset.cap.effective = new->cap_effective; context->capset.cap.inheritable = new->cap_effective; context->capset.cap.permitted = new->cap_permitted; @@ -2373,7 +2369,7 @@ static void audit_log_task(struct audit_buffer *ab) from_kgid(&init_user_ns, gid), sessionid); audit_log_task_context(ab); - audit_log_format(ab, " pid=%d comm=", task_pid_nr(current)); + audit_log_format(ab, " pid=%d comm=", task_tgid_nr(current)); audit_log_untrustedstring(ab, get_task_comm(comm, current)); audit_log_d_path_exe(ab, current->mm); } diff --git a/kernel/capability.c b/kernel/capability.c index 45432b54d5c6..00411c82dac5 100644 --- a/kernel/capability.c +++ b/kernel/capability.c @@ -361,6 +361,24 @@ bool has_capability_noaudit(struct task_struct *t, int cap) return has_ns_capability_noaudit(t, &init_user_ns, cap); } +static bool ns_capable_common(struct user_namespace *ns, int cap, bool audit) +{ + int capable; + + if (unlikely(!cap_valid(cap))) { + pr_crit("capable() called with invalid cap=%u\n", cap); + BUG(); + } + + capable = audit ? security_capable(current_cred(), ns, cap) : + security_capable_noaudit(current_cred(), ns, cap); + if (capable == 0) { + current->flags |= PF_SUPERPRIV; + return true; + } + return false; +} + /** * ns_capable - Determine if the current task has a superior capability in effect * @ns: The usernamespace we want the capability in @@ -374,19 +392,27 @@ bool has_capability_noaudit(struct task_struct *t, int cap) */ bool ns_capable(struct user_namespace *ns, int cap) { - if (unlikely(!cap_valid(cap))) { - pr_crit("capable() called with invalid cap=%u\n", cap); - BUG(); - } - - if (security_capable(current_cred(), ns, cap) == 0) { - current->flags |= PF_SUPERPRIV; - return true; - } - return false; + return ns_capable_common(ns, cap, true); } EXPORT_SYMBOL(ns_capable); +/** + * ns_capable_noaudit - Determine if the current task has a superior capability + * (unaudited) in effect + * @ns: The usernamespace we want the capability in + * @cap: The capability to be tested for + * + * Return true if the current task has the given superior capability currently + * available for use, false if not. + * + * This sets PF_SUPERPRIV on the task if the capability is available on the + * assumption that it's about to be used. + */ +bool ns_capable_noaudit(struct user_namespace *ns, int cap) +{ + return ns_capable_common(ns, cap, false); +} +EXPORT_SYMBOL(ns_capable_noaudit); /** * capable - Determine if the current task has a superior capability in effect diff --git a/kernel/cgroup.c b/kernel/cgroup.c index 47851dbc893d..e4552a3cbf41 100644 --- a/kernel/cgroup.c +++ b/kernel/cgroup.c @@ -2671,45 +2671,6 @@ static int cgroup_attach_task(struct cgroup *dst_cgrp, return ret; } -int subsys_cgroup_allow_attach(struct cgroup_taskset *tset) -{ - const struct cred *cred = current_cred(), *tcred; - struct task_struct *task; - struct cgroup_subsys_state *css; - - if (capable(CAP_SYS_NICE)) - return 0; - - cgroup_taskset_for_each(task, css, tset) { - tcred = __task_cred(task); - - if (current != task && !uid_eq(cred->euid, tcred->uid) && - !uid_eq(cred->euid, tcred->suid)) - return -EACCES; - } - - return 0; -} - -static int cgroup_allow_attach(struct cgroup *cgrp, struct cgroup_taskset *tset) -{ - struct cgroup_subsys_state *css; - int i; - int ret; - - for_each_css(css, i, cgrp) { - if (css->ss->allow_attach) { - ret = css->ss->allow_attach(tset); - if (ret) - return ret; - } else { - return -EACCES; - } - } - - return 0; -} - static int cgroup_procs_write_permission(struct task_struct *task, struct cgroup *dst_cgrp, struct kernfs_open_file *of) @@ -2724,24 +2685,9 @@ static int cgroup_procs_write_permission(struct task_struct *task, */ if (!uid_eq(cred->euid, GLOBAL_ROOT_UID) && !uid_eq(cred->euid, tcred->uid) && - !uid_eq(cred->euid, tcred->suid)) { - /* - * if the default permission check fails, give each - * cgroup a chance to extend the permission check - */ - struct cgroup_taskset tset = { - .src_csets = LIST_HEAD_INIT(tset.src_csets), - .dst_csets = LIST_HEAD_INIT(tset.dst_csets), - .csets = &tset.src_csets, - }; - struct css_set *cset; - cset = task_css_set(task); - list_add(&cset->mg_node, &tset.src_csets); - ret = cgroup_allow_attach(dst_cgrp, &tset); - list_del(&tset.src_csets); - if (ret) - ret = -EACCES; - } + !uid_eq(cred->euid, tcred->suid) && + !ns_capable(tcred->user_ns, CAP_SYS_RESOURCE)) + ret = -EACCES; if (!ret && cgroup_on_dfl(dst_cgrp)) { struct super_block *sb = of->file->f_path.dentry->d_sb; @@ -4848,6 +4794,7 @@ static void init_and_link_css(struct cgroup_subsys_state *css, memset(css, 0, sizeof(*css)); css->cgroup = cgrp; css->ss = ss; + css->id = -1; INIT_LIST_HEAD(&css->sibling); INIT_LIST_HEAD(&css->children); css->serial_nr = css_serial_nr_next++; @@ -5379,6 +5326,12 @@ int __init cgroup_init(void) BUG_ON(cgroup_init_cftypes(NULL, cgroup_dfl_base_files)); BUG_ON(cgroup_init_cftypes(NULL, cgroup_legacy_base_files)); + /* + * The latency of the synchronize_sched() is too high for cgroups, + * avoid it at the cost of forcing all readers into the slow path. + */ + rcu_sync_enter_start(&cgroup_threadgroup_rwsem.rss); + mutex_lock(&cgroup_mutex); /* Add init_css_set to the hash table */ diff --git a/kernel/configs/tiny.config b/kernel/configs/tiny.config index c2de56ab0fce..7fa0c4ae6394 100644 --- a/kernel/configs/tiny.config +++ b/kernel/configs/tiny.config @@ -1,4 +1,12 @@ +# CONFIG_CC_OPTIMIZE_FOR_PERFORMANCE is not set CONFIG_CC_OPTIMIZE_FOR_SIZE=y +# CONFIG_KERNEL_GZIP is not set +# CONFIG_KERNEL_BZIP2 is not set +# CONFIG_KERNEL_LZMA is not set CONFIG_KERNEL_XZ=y +# CONFIG_KERNEL_LZO is not set +# CONFIG_KERNEL_LZ4 is not set CONFIG_OPTIMIZE_INLINING=y +# CONFIG_SLAB is not set +# CONFIG_SLUB is not set CONFIG_SLOB=y diff --git a/kernel/cpu.c b/kernel/cpu.c index 9ced7c751648..c8a1751be224 100644 --- a/kernel/cpu.c +++ b/kernel/cpu.c @@ -185,10 +185,17 @@ void cpu_hotplug_disable(void) } EXPORT_SYMBOL_GPL(cpu_hotplug_disable); +static void __cpu_hotplug_enable(void) +{ + if (WARN_ONCE(!cpu_hotplug_disabled, "Unbalanced cpu hotplug enable\n")) + return; + cpu_hotplug_disabled--; +} + void cpu_hotplug_enable(void) { cpu_maps_update_begin(); - WARN_ON(--cpu_hotplug_disabled < 0); + __cpu_hotplug_enable(); cpu_maps_update_done(); } EXPORT_SYMBOL_GPL(cpu_hotplug_enable); @@ -631,7 +638,7 @@ void enable_nonboot_cpus(void) /* Allow everyone to use the CPU hotplug again */ cpu_maps_update_begin(); - WARN_ON(--cpu_hotplug_disabled < 0); + __cpu_hotplug_enable(); if (cpumask_empty(frozen_cpus)) goto out; diff --git a/kernel/cpuset.c b/kernel/cpuset.c index 11eaf14b52c2..3f9db31c5d04 100644 --- a/kernel/cpuset.c +++ b/kernel/cpuset.c @@ -98,6 +98,7 @@ struct cpuset { /* user-configured CPUs and Memory Nodes allow to tasks */ cpumask_var_t cpus_allowed; + cpumask_var_t cpus_requested; nodemask_t mems_allowed; /* effective CPUs and Memory Nodes allow to tasks */ @@ -324,8 +325,7 @@ static struct file_system_type cpuset_fs_type = { /* * Return in pmask the portion of a cpusets's cpus_allowed that * are online. If none are online, walk up the cpuset hierarchy - * until we find one that does have some online cpus. The top - * cpuset always has some cpus online. + * until we find one that does have some online cpus. * * One way or another, we guarantee to return some non-empty subset * of cpu_online_mask. @@ -334,8 +334,20 @@ static struct file_system_type cpuset_fs_type = { */ static void guarantee_online_cpus(struct cpuset *cs, struct cpumask *pmask) { - while (!cpumask_intersects(cs->effective_cpus, cpu_online_mask)) + while (!cpumask_intersects(cs->effective_cpus, cpu_online_mask)) { cs = parent_cs(cs); + if (unlikely(!cs)) { + /* + * The top cpuset doesn't have any online cpu as a + * consequence of a race between cpuset_hotplug_work + * and cpu hotplug notifier. But we know the top + * cpuset's effective_cpus is on its way to to be + * identical to cpu_online_mask. + */ + cpumask_copy(pmask, cpu_online_mask); + return; + } + } cpumask_and(pmask, cs->effective_cpus, cpu_online_mask); } @@ -386,7 +398,7 @@ static void cpuset_update_task_spread_flag(struct cpuset *cs, static int is_cpuset_subset(const struct cpuset *p, const struct cpuset *q) { - return cpumask_subset(p->cpus_allowed, q->cpus_allowed) && + return cpumask_subset(p->cpus_requested, q->cpus_requested) && nodes_subset(p->mems_allowed, q->mems_allowed) && is_cpu_exclusive(p) <= is_cpu_exclusive(q) && is_mem_exclusive(p) <= is_mem_exclusive(q); @@ -486,7 +498,7 @@ static int validate_change(struct cpuset *cur, struct cpuset *trial) cpuset_for_each_child(c, css, par) { if ((is_cpu_exclusive(trial) || is_cpu_exclusive(c)) && c != cur && - cpumask_intersects(trial->cpus_allowed, c->cpus_allowed)) + cpumask_intersects(trial->cpus_requested, c->cpus_requested)) goto out; if ((is_mem_exclusive(trial) || is_mem_exclusive(c)) && c != cur && @@ -945,17 +957,18 @@ static int update_cpumask(struct cpuset *cs, struct cpuset *trialcs, if (!*buf) { cpumask_clear(trialcs->cpus_allowed); } else { - retval = cpulist_parse(buf, trialcs->cpus_allowed); + retval = cpulist_parse(buf, trialcs->cpus_requested); if (retval < 0) return retval; - if (!cpumask_subset(trialcs->cpus_allowed, - top_cpuset.cpus_allowed)) + if (!cpumask_subset(trialcs->cpus_requested, cpu_present_mask)) return -EINVAL; + + cpumask_and(trialcs->cpus_allowed, trialcs->cpus_requested, cpu_active_mask); } /* Nothing to do if the cpus didn't change */ - if (cpumask_equal(cs->cpus_allowed, trialcs->cpus_allowed)) + if (cpumask_equal(cs->cpus_requested, trialcs->cpus_requested)) return 0; retval = validate_change(cs, trialcs); @@ -964,6 +977,7 @@ static int update_cpumask(struct cpuset *cs, struct cpuset *trialcs, spin_lock_irq(&callback_lock); cpumask_copy(cs->cpus_allowed, trialcs->cpus_allowed); + cpumask_copy(cs->cpus_requested, trialcs->cpus_requested); spin_unlock_irq(&callback_lock); /* use trialcs->cpus_allowed as a temp variable */ @@ -1754,7 +1768,7 @@ static int cpuset_common_seq_show(struct seq_file *sf, void *v) switch (type) { case FILE_CPULIST: - seq_printf(sf, "%*pbl\n", cpumask_pr_args(cs->cpus_allowed)); + seq_printf(sf, "%*pbl\n", cpumask_pr_args(cs->cpus_requested)); break; case FILE_MEMLIST: seq_printf(sf, "%*pbl\n", nodemask_pr_args(&cs->mems_allowed)); @@ -1943,11 +1957,14 @@ cpuset_css_alloc(struct cgroup_subsys_state *parent_css) return ERR_PTR(-ENOMEM); if (!alloc_cpumask_var(&cs->cpus_allowed, GFP_KERNEL)) goto free_cs; + if (!alloc_cpumask_var(&cs->cpus_requested, GFP_KERNEL)) + goto free_allowed; if (!alloc_cpumask_var(&cs->effective_cpus, GFP_KERNEL)) - goto free_cpus; + goto free_requested; set_bit(CS_SCHED_LOAD_BALANCE, &cs->flags); cpumask_clear(cs->cpus_allowed); + cpumask_clear(cs->cpus_requested); nodes_clear(cs->mems_allowed); cpumask_clear(cs->effective_cpus); nodes_clear(cs->effective_mems); @@ -1956,7 +1973,9 @@ cpuset_css_alloc(struct cgroup_subsys_state *parent_css) return &cs->css; -free_cpus: +free_requested: + free_cpumask_var(cs->cpus_requested); +free_allowed: free_cpumask_var(cs->cpus_allowed); free_cs: kfree(cs); @@ -2019,6 +2038,7 @@ static int cpuset_css_online(struct cgroup_subsys_state *css) cs->mems_allowed = parent->mems_allowed; cs->effective_mems = parent->mems_allowed; cpumask_copy(cs->cpus_allowed, parent->cpus_allowed); + cpumask_copy(cs->cpus_requested, parent->cpus_requested); cpumask_copy(cs->effective_cpus, parent->cpus_allowed); spin_unlock_irq(&callback_lock); out_unlock: @@ -2053,6 +2073,7 @@ static void cpuset_css_free(struct cgroup_subsys_state *css) free_cpumask_var(cs->effective_cpus); free_cpumask_var(cs->cpus_allowed); + free_cpumask_var(cs->cpus_requested); kfree(cs); } @@ -2074,6 +2095,20 @@ static void cpuset_bind(struct cgroup_subsys_state *root_css) mutex_unlock(&cpuset_mutex); } +/* + * Make sure the new task conform to the current state of its parent, + * which could have been changed by cpuset just after it inherits the + * state from the parent and before it sits on the cgroup's task list. + */ +void cpuset_fork(struct task_struct *task, void *priv) +{ + if (task_css_is_root(task, cpuset_cgrp_id)) + return; + + set_cpus_allowed_ptr(task, ¤t->cpus_allowed); + task->mems_allowed = current->mems_allowed; +} + struct cgroup_subsys cpuset_cgrp_subsys = { .css_alloc = cpuset_css_alloc, .css_online = cpuset_css_online, @@ -2084,6 +2119,7 @@ struct cgroup_subsys cpuset_cgrp_subsys = { .attach = cpuset_attach, .post_attach = cpuset_post_attach, .bind = cpuset_bind, + .fork = cpuset_fork, .legacy_cftypes = files, .early_init = 1, }; @@ -2102,8 +2138,11 @@ int __init cpuset_init(void) BUG(); if (!alloc_cpumask_var(&top_cpuset.effective_cpus, GFP_KERNEL)) BUG(); + if (!alloc_cpumask_var(&top_cpuset.cpus_requested, GFP_KERNEL)) + BUG(); cpumask_setall(top_cpuset.cpus_allowed); + cpumask_setall(top_cpuset.cpus_requested); nodes_setall(top_cpuset.mems_allowed); cpumask_setall(top_cpuset.effective_cpus); nodes_setall(top_cpuset.effective_mems); @@ -2237,7 +2276,7 @@ retry: goto retry; } - cpumask_and(&new_cpus, cs->cpus_allowed, parent_cs(cs)->effective_cpus); + cpumask_and(&new_cpus, cs->cpus_requested, parent_cs(cs)->effective_cpus); nodes_and(new_mems, cs->mems_allowed, parent_cs(cs)->effective_mems); cpus_updated = !cpumask_equal(&new_cpus, cs->effective_cpus); diff --git a/kernel/cred.c b/kernel/cred.c index 71179a09c1d6..ff8606f77d90 100644 --- a/kernel/cred.c +++ b/kernel/cred.c @@ -689,6 +689,8 @@ EXPORT_SYMBOL(set_security_override_from_ctx); */ int set_create_files_as(struct cred *new, struct inode *inode) { + if (!uid_valid(inode->i_uid) || !gid_valid(inode->i_gid)) + return -EINVAL; new->fsuid = inode->i_uid; new->fsgid = inode->i_gid; return security_kernel_create_files_as(new, inode); diff --git a/kernel/events/core.c b/kernel/events/core.c index 662f2b23632f..f4fdaff76f6d 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -1544,12 +1544,33 @@ static int __init perf_workqueue_init(void) core_initcall(perf_workqueue_init); -static inline int pmu_filter_match(struct perf_event *event) +static inline int __pmu_filter_match(struct perf_event *event) { struct pmu *pmu = event->pmu; return pmu->filter_match ? pmu->filter_match(event) : 1; } +/* + * Check whether we should attempt to schedule an event group based on + * PMU-specific filtering. An event group can consist of HW and SW events, + * potentially with a SW leader, so we must check all the filters, to + * determine whether a group is schedulable: + */ +static inline int pmu_filter_match(struct perf_event *event) +{ + struct perf_event *child; + + if (!__pmu_filter_match(event)) + return 0; + + list_for_each_entry(child, &event->sibling_list, group_entry) { + if (!__pmu_filter_match(child)) + return 0; + } + + return 1; +} + static inline int event_filter_match(struct perf_event *event) { diff --git a/kernel/events/uprobes.c b/kernel/events/uprobes.c index 7dad84913abf..7b1b772ab1ce 100644 --- a/kernel/events/uprobes.c +++ b/kernel/events/uprobes.c @@ -171,8 +171,10 @@ static int __replace_page(struct vm_area_struct *vma, unsigned long addr, mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end); err = -EAGAIN; ptep = page_check_address(page, mm, addr, &ptl, 0); - if (!ptep) + if (!ptep) { + mem_cgroup_cancel_charge(kpage, memcg); goto unlock; + } get_page(kpage); page_add_new_anon_rmap(kpage, vma, addr); @@ -199,7 +201,6 @@ static int __replace_page(struct vm_area_struct *vma, unsigned long addr, err = 0; unlock: - mem_cgroup_cancel_charge(kpage, memcg); mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end); unlock_page(page); return err; @@ -1692,8 +1693,7 @@ static int is_trap_at_addr(struct mm_struct *mm, unsigned long vaddr) int result; pagefault_disable(); - result = __copy_from_user_inatomic(&opcode, (void __user*)vaddr, - sizeof(opcode)); + result = __get_user(opcode, (uprobe_opcode_t __user *)vaddr); pagefault_enable(); if (likely(result == 0)) diff --git a/kernel/exit.c b/kernel/exit.c index ffba5df4abd5..62c4bd4abd3a 100644 --- a/kernel/exit.c +++ b/kernel/exit.c @@ -54,6 +54,8 @@ #include <linux/writeback.h> #include <linux/shm.h> +#include "sched/tune.h" + #include <asm/uaccess.h> #include <asm/unistd.h> #include <asm/pgtable.h> @@ -699,6 +701,9 @@ void do_exit(long code) } exit_signals(tsk); /* sets PF_EXITING */ + + schedtune_exit_task(tsk); + /* * tsk->flags are checked in the futex code to protect against * an exiting task cleaning up the robust pi futexes. diff --git a/kernel/fork.c b/kernel/fork.c index 7ec6e9939b2c..18a5cb17035a 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -764,6 +764,29 @@ struct file *get_mm_exe_file(struct mm_struct *mm) EXPORT_SYMBOL(get_mm_exe_file); /** + * get_task_exe_file - acquire a reference to the task's executable file + * + * Returns %NULL if task's mm (if any) has no associated executable file or + * this is a kernel thread with borrowed mm (see the comment above get_task_mm). + * User must release file via fput(). + */ +struct file *get_task_exe_file(struct task_struct *task) +{ + struct file *exe_file = NULL; + struct mm_struct *mm; + + task_lock(task); + mm = task->mm; + if (mm) { + if (!(task->flags & PF_KTHREAD)) + exe_file = get_mm_exe_file(mm); + } + task_unlock(task); + return exe_file; +} +EXPORT_SYMBOL(get_task_exe_file); + +/** * get_task_mm - acquire a reference to the task's mm * * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning @@ -879,14 +902,12 @@ void mm_release(struct task_struct *tsk, struct mm_struct *mm) deactivate_mm(tsk, mm); /* - * If we're exiting normally, clear a user-space tid field if - * requested. We leave this alone when dying by signal, to leave - * the value intact in a core dump, and to save the unnecessary - * trouble, say, a killed vfork parent shouldn't touch this mm. - * Userland only wants this done for a sys_exit. + * Signal userspace if we're not exiting with a core dump + * because we want to leave the value intact for debugging + * purposes. */ if (tsk->clear_child_tid) { - if (!(tsk->flags & PF_SIGNALED) && + if (!(tsk->signal->flags & SIGNAL_GROUP_COREDUMP) && atomic_read(&mm->mm_users) > 1) { /* * We don't check the error code - if userspace has @@ -1370,7 +1391,6 @@ static struct task_struct *copy_process(unsigned long clone_flags, p->real_start_time = ktime_get_boot_ns(); p->io_context = NULL; p->audit_context = NULL; - threadgroup_change_begin(current); cgroup_fork(p); #ifdef CONFIG_NUMA p->mempolicy = mpol_dup(p->mempolicy); @@ -1522,6 +1542,7 @@ static struct task_struct *copy_process(unsigned long clone_flags, INIT_LIST_HEAD(&p->thread_group); p->task_works = NULL; + threadgroup_change_begin(current); /* * Ensure that the cgroup subsystem policies allow the new process to be * forked. It should be noted the the new process's css_set can be changed @@ -1622,6 +1643,7 @@ static struct task_struct *copy_process(unsigned long clone_flags, bad_fork_cancel_cgroup: cgroup_cancel_fork(p, cgrp_ss_priv); bad_fork_free_pid: + threadgroup_change_end(current); if (pid != &init_struct_pid) free_pid(pid); bad_fork_cleanup_io: @@ -1652,7 +1674,6 @@ bad_fork_cleanup_policy: mpol_put(p->mempolicy); bad_fork_cleanup_threadgroup_lock: #endif - threadgroup_change_end(current); delayacct_tsk_free(p); bad_fork_cleanup_count: atomic_dec(&p->cred->user->processes); diff --git a/kernel/futex.c b/kernel/futex.c index 9d8163afd87c..e8af73cc51a7 100644 --- a/kernel/futex.c +++ b/kernel/futex.c @@ -681,7 +681,7 @@ static int get_futex_value_locked(u32 *dest, u32 __user *from) int ret; pagefault_disable(); - ret = __copy_from_user_inatomic(dest, from, sizeof(u32)); + ret = __get_user(*dest, from); pagefault_enable(); return ret ? -EFAULT : 0; diff --git a/kernel/irq/generic-chip.c b/kernel/irq/generic-chip.c index abd286afbd27..a4775f3451b9 100644 --- a/kernel/irq/generic-chip.c +++ b/kernel/irq/generic-chip.c @@ -411,8 +411,29 @@ int irq_map_generic_chip(struct irq_domain *d, unsigned int virq, } EXPORT_SYMBOL_GPL(irq_map_generic_chip); +static void irq_unmap_generic_chip(struct irq_domain *d, unsigned int virq) +{ + struct irq_data *data = irq_domain_get_irq_data(d, virq); + struct irq_domain_chip_generic *dgc = d->gc; + unsigned int hw_irq = data->hwirq; + struct irq_chip_generic *gc; + int irq_idx; + + gc = irq_get_domain_generic_chip(d, hw_irq); + if (!gc) + return; + + irq_idx = hw_irq % dgc->irqs_per_chip; + + clear_bit(irq_idx, &gc->installed); + irq_domain_set_info(d, virq, hw_irq, &no_irq_chip, NULL, NULL, NULL, + NULL); + +} + struct irq_domain_ops irq_generic_chip_ops = { .map = irq_map_generic_chip, + .unmap = irq_unmap_generic_chip, .xlate = irq_domain_xlate_onetwocell, }; EXPORT_SYMBOL_GPL(irq_generic_chip_ops); diff --git a/kernel/irq/msi.c b/kernel/irq/msi.c index 6b0c0b74a2a1..cd6009006510 100644 --- a/kernel/irq/msi.c +++ b/kernel/irq/msi.c @@ -268,7 +268,7 @@ int msi_domain_alloc_irqs(struct irq_domain *domain, struct device *dev, struct msi_domain_ops *ops = info->ops; msi_alloc_info_t arg; struct msi_desc *desc; - int i, ret, virq = -1; + int i, ret, virq; ret = ops->msi_check(domain, info, dev); if (ret == 0) @@ -278,12 +278,8 @@ int msi_domain_alloc_irqs(struct irq_domain *domain, struct device *dev, for_each_msi_entry(desc, dev) { ops->set_desc(&arg, desc); - if (info->flags & MSI_FLAG_IDENTITY_MAP) - virq = (int)ops->get_hwirq(info, &arg); - else - virq = -1; - virq = __irq_domain_alloc_irqs(domain, virq, desc->nvec_used, + virq = __irq_domain_alloc_irqs(domain, -1, desc->nvec_used, dev_to_node(dev), &arg, false); if (virq < 0) { ret = -ENOSPC; @@ -302,11 +298,23 @@ int msi_domain_alloc_irqs(struct irq_domain *domain, struct device *dev, ops->msi_finish(&arg, 0); for_each_msi_entry(desc, dev) { + virq = desc->irq; if (desc->nvec_used == 1) dev_dbg(dev, "irq %d for MSI\n", virq); else dev_dbg(dev, "irq [%d-%d] for MSI\n", virq, virq + desc->nvec_used - 1); + /* + * This flag is set by the PCI layer as we need to activate + * the MSI entries before the PCI layer enables MSI in the + * card. Otherwise the card latches a random msi message. + */ + if (info->flags & MSI_FLAG_ACTIVATE_EARLY) { + struct irq_data *irq_data; + + irq_data = irq_domain_get_irq_data(domain, desc->irq); + irq_domain_activate_irq(irq_data); + } } return 0; diff --git a/kernel/kexec_file.c b/kernel/kexec_file.c index b70ada0028d2..6030efd4a188 100644 --- a/kernel/kexec_file.c +++ b/kernel/kexec_file.c @@ -934,7 +934,10 @@ int kexec_load_purgatory(struct kimage *image, unsigned long min, return 0; out: vfree(pi->sechdrs); + pi->sechdrs = NULL; + vfree(pi->purgatory_buf); + pi->purgatory_buf = NULL; return ret; } diff --git a/kernel/locking/percpu-rwsem.c b/kernel/locking/percpu-rwsem.c index f231e0bb311c..ce182599cf2e 100644 --- a/kernel/locking/percpu-rwsem.c +++ b/kernel/locking/percpu-rwsem.c @@ -8,151 +8,186 @@ #include <linux/sched.h> #include <linux/errno.h> -int __percpu_init_rwsem(struct percpu_rw_semaphore *brw, +int __percpu_init_rwsem(struct percpu_rw_semaphore *sem, const char *name, struct lock_class_key *rwsem_key) { - brw->fast_read_ctr = alloc_percpu(int); - if (unlikely(!brw->fast_read_ctr)) + sem->read_count = alloc_percpu(int); + if (unlikely(!sem->read_count)) return -ENOMEM; /* ->rw_sem represents the whole percpu_rw_semaphore for lockdep */ - __init_rwsem(&brw->rw_sem, name, rwsem_key); - rcu_sync_init(&brw->rss, RCU_SCHED_SYNC); - atomic_set(&brw->slow_read_ctr, 0); - init_waitqueue_head(&brw->write_waitq); + rcu_sync_init(&sem->rss, RCU_SCHED_SYNC); + __init_rwsem(&sem->rw_sem, name, rwsem_key); + init_waitqueue_head(&sem->writer); + sem->readers_block = 0; return 0; } EXPORT_SYMBOL_GPL(__percpu_init_rwsem); -void percpu_free_rwsem(struct percpu_rw_semaphore *brw) +void percpu_free_rwsem(struct percpu_rw_semaphore *sem) { /* * XXX: temporary kludge. The error path in alloc_super() * assumes that percpu_free_rwsem() is safe after kzalloc(). */ - if (!brw->fast_read_ctr) + if (!sem->read_count) return; - rcu_sync_dtor(&brw->rss); - free_percpu(brw->fast_read_ctr); - brw->fast_read_ctr = NULL; /* catch use after free bugs */ + rcu_sync_dtor(&sem->rss); + free_percpu(sem->read_count); + sem->read_count = NULL; /* catch use after free bugs */ } +EXPORT_SYMBOL_GPL(percpu_free_rwsem); -/* - * This is the fast-path for down_read/up_read. If it succeeds we rely - * on the barriers provided by rcu_sync_enter/exit; see the comments in - * percpu_down_write() and percpu_up_write(). - * - * If this helper fails the callers rely on the normal rw_semaphore and - * atomic_dec_and_test(), so in this case we have the necessary barriers. - */ -static bool update_fast_ctr(struct percpu_rw_semaphore *brw, unsigned int val) +int __percpu_down_read(struct percpu_rw_semaphore *sem, int try) { - bool success; + /* + * Due to having preemption disabled the decrement happens on + * the same CPU as the increment, avoiding the + * increment-on-one-CPU-and-decrement-on-another problem. + * + * If the reader misses the writer's assignment of readers_block, then + * the writer is guaranteed to see the reader's increment. + * + * Conversely, any readers that increment their sem->read_count after + * the writer looks are guaranteed to see the readers_block value, + * which in turn means that they are guaranteed to immediately + * decrement their sem->read_count, so that it doesn't matter that the + * writer missed them. + */ - preempt_disable(); - success = rcu_sync_is_idle(&brw->rss); - if (likely(success)) - __this_cpu_add(*brw->fast_read_ctr, val); - preempt_enable(); + smp_mb(); /* A matches D */ - return success; -} + /* + * If !readers_block the critical section starts here, matched by the + * release in percpu_up_write(). + */ + if (likely(!smp_load_acquire(&sem->readers_block))) + return 1; -/* - * Like the normal down_read() this is not recursive, the writer can - * come after the first percpu_down_read() and create the deadlock. - * - * Note: returns with lock_is_held(brw->rw_sem) == T for lockdep, - * percpu_up_read() does rwsem_release(). This pairs with the usage - * of ->rw_sem in percpu_down/up_write(). - */ -void percpu_down_read(struct percpu_rw_semaphore *brw) -{ - might_sleep(); - rwsem_acquire_read(&brw->rw_sem.dep_map, 0, 0, _RET_IP_); + /* + * Per the above comment; we still have preemption disabled and + * will thus decrement on the same CPU as we incremented. + */ + __percpu_up_read(sem); - if (likely(update_fast_ctr(brw, +1))) - return; + if (try) + return 0; - /* Avoid rwsem_acquire_read() and rwsem_release() */ - __down_read(&brw->rw_sem); - atomic_inc(&brw->slow_read_ctr); - __up_read(&brw->rw_sem); -} -EXPORT_SYMBOL_GPL(percpu_down_read); + /* + * We either call schedule() in the wait, or we'll fall through + * and reschedule on the preempt_enable() in percpu_down_read(). + */ + preempt_enable_no_resched(); -int percpu_down_read_trylock(struct percpu_rw_semaphore *brw) -{ - if (unlikely(!update_fast_ctr(brw, +1))) { - if (!__down_read_trylock(&brw->rw_sem)) - return 0; - atomic_inc(&brw->slow_read_ctr); - __up_read(&brw->rw_sem); - } - - rwsem_acquire_read(&brw->rw_sem.dep_map, 0, 1, _RET_IP_); + /* + * Avoid lockdep for the down/up_read() we already have them. + */ + __down_read(&sem->rw_sem); + this_cpu_inc(*sem->read_count); + __up_read(&sem->rw_sem); + + preempt_disable(); return 1; } +EXPORT_SYMBOL_GPL(__percpu_down_read); -void percpu_up_read(struct percpu_rw_semaphore *brw) +void __percpu_up_read(struct percpu_rw_semaphore *sem) { - rwsem_release(&brw->rw_sem.dep_map, 1, _RET_IP_); - - if (likely(update_fast_ctr(brw, -1))) - return; + smp_mb(); /* B matches C */ + /* + * In other words, if they see our decrement (presumably to aggregate + * zero, as that is the only time it matters) they will also see our + * critical section. + */ + __this_cpu_dec(*sem->read_count); - /* false-positive is possible but harmless */ - if (atomic_dec_and_test(&brw->slow_read_ctr)) - wake_up_all(&brw->write_waitq); + /* Prod writer to recheck readers_active */ + wake_up(&sem->writer); } -EXPORT_SYMBOL_GPL(percpu_up_read); +EXPORT_SYMBOL_GPL(__percpu_up_read); + +#define per_cpu_sum(var) \ +({ \ + typeof(var) __sum = 0; \ + int cpu; \ + compiletime_assert_atomic_type(__sum); \ + for_each_possible_cpu(cpu) \ + __sum += per_cpu(var, cpu); \ + __sum; \ +}) -static int clear_fast_ctr(struct percpu_rw_semaphore *brw) +/* + * Return true if the modular sum of the sem->read_count per-CPU variable is + * zero. If this sum is zero, then it is stable due to the fact that if any + * newly arriving readers increment a given counter, they will immediately + * decrement that same counter. + */ +static bool readers_active_check(struct percpu_rw_semaphore *sem) { - unsigned int sum = 0; - int cpu; + if (per_cpu_sum(*sem->read_count) != 0) + return false; + + /* + * If we observed the decrement; ensure we see the entire critical + * section. + */ - for_each_possible_cpu(cpu) { - sum += per_cpu(*brw->fast_read_ctr, cpu); - per_cpu(*brw->fast_read_ctr, cpu) = 0; - } + smp_mb(); /* C matches B */ - return sum; + return true; } -void percpu_down_write(struct percpu_rw_semaphore *brw) +void percpu_down_write(struct percpu_rw_semaphore *sem) { + /* Notify readers to take the slow path. */ + rcu_sync_enter(&sem->rss); + + down_write(&sem->rw_sem); + /* - * Make rcu_sync_is_idle() == F and thus disable the fast-path in - * percpu_down_read() and percpu_up_read(), and wait for gp pass. - * - * The latter synchronises us with the preceding readers which used - * the fast-past, so we can not miss the result of __this_cpu_add() - * or anything else inside their criticial sections. + * Notify new readers to block; up until now, and thus throughout the + * longish rcu_sync_enter() above, new readers could still come in. */ - rcu_sync_enter(&brw->rss); + WRITE_ONCE(sem->readers_block, 1); - /* exclude other writers, and block the new readers completely */ - down_write(&brw->rw_sem); + smp_mb(); /* D matches A */ - /* nobody can use fast_read_ctr, move its sum into slow_read_ctr */ - atomic_add(clear_fast_ctr(brw), &brw->slow_read_ctr); + /* + * If they don't see our writer of readers_block, then we are + * guaranteed to see their sem->read_count increment, and therefore + * will wait for them. + */ - /* wait for all readers to complete their percpu_up_read() */ - wait_event(brw->write_waitq, !atomic_read(&brw->slow_read_ctr)); + /* Wait for all now active readers to complete. */ + wait_event(sem->writer, readers_active_check(sem)); } EXPORT_SYMBOL_GPL(percpu_down_write); -void percpu_up_write(struct percpu_rw_semaphore *brw) +void percpu_up_write(struct percpu_rw_semaphore *sem) { - /* release the lock, but the readers can't use the fast-path */ - up_write(&brw->rw_sem); /* - * Enable the fast-path in percpu_down_read() and percpu_up_read() - * but only after another gp pass; this adds the necessary barrier - * to ensure the reader can't miss the changes done by us. + * Signal the writer is done, no fast path yet. + * + * One reason that we cannot just immediately flip to readers_fast is + * that new readers might fail to see the results of this writer's + * critical section. + * + * Therefore we force it through the slow path which guarantees an + * acquire and thereby guarantees the critical section's consistency. + */ + smp_store_release(&sem->readers_block, 0); + + /* + * Release the write lock, this will allow readers back in the game. + */ + up_write(&sem->rw_sem); + + /* + * Once this completes (at least one RCU-sched grace period hence) the + * reader fast path will be available again. Safe to use outside the + * exclusive write lock because its counting. */ - rcu_sync_exit(&brw->rss); + rcu_sync_exit(&sem->rss); } EXPORT_SYMBOL_GPL(percpu_up_write); diff --git a/kernel/module.c b/kernel/module.c index 0e5c71195f18..b14a4f31221f 100644 --- a/kernel/module.c +++ b/kernel/module.c @@ -2606,13 +2606,18 @@ static inline void kmemleak_load_module(const struct module *mod, #endif #ifdef CONFIG_MODULE_SIG -static int module_sig_check(struct load_info *info) +static int module_sig_check(struct load_info *info, int flags) { int err = -ENOKEY; const unsigned long markerlen = sizeof(MODULE_SIG_STRING) - 1; const void *mod = info->hdr; - if (info->len > markerlen && + /* + * Require flags == 0, as a module with version information + * removed is no longer the module that was signed + */ + if (flags == 0 && + info->len > markerlen && memcmp(mod + info->len - markerlen, MODULE_SIG_STRING, markerlen) == 0) { /* We truncate the module to discard the signature */ info->len -= markerlen; @@ -2631,7 +2636,7 @@ static int module_sig_check(struct load_info *info) return err; } #else /* !CONFIG_MODULE_SIG */ -static int module_sig_check(struct load_info *info) +static int module_sig_check(struct load_info *info, int flags) { return 0; } @@ -3444,7 +3449,7 @@ static int load_module(struct load_info *info, const char __user *uargs, long err; char *after_dashes; - err = module_sig_check(info); + err = module_sig_check(info, flags); if (err) goto free_copy; diff --git a/kernel/panic.c b/kernel/panic.c index 223564d3e1f8..41e2b54f36b5 100644 --- a/kernel/panic.c +++ b/kernel/panic.c @@ -28,9 +28,6 @@ #define PANIC_TIMER_STEP 100 #define PANIC_BLINK_SPD 18 -/* Machine specific panic information string */ -char *mach_panic_string; - int panic_on_oops = CONFIG_PANIC_ON_OOPS_VALUE; static unsigned long tainted_mask; static int pause_on_oops; @@ -415,11 +412,6 @@ late_initcall(init_oops_id); void print_oops_end_marker(void) { init_oops_id(); - - if (mach_panic_string) - printk(KERN_WARNING "Board Information: %s\n", - mach_panic_string); - pr_warn("---[ end trace %016llx ]---\n", (unsigned long long)oops_id); } diff --git a/kernel/power/hibernate.c b/kernel/power/hibernate.c index b7dd5718836e..3124cebaec31 100644 --- a/kernel/power/hibernate.c +++ b/kernel/power/hibernate.c @@ -299,12 +299,12 @@ static int create_image(int platform_mode) save_processor_state(); trace_suspend_resume(TPS("machine_suspend"), PM_EVENT_HIBERNATE, true); error = swsusp_arch_suspend(); + /* Restore control flow magically appears here */ + restore_processor_state(); trace_suspend_resume(TPS("machine_suspend"), PM_EVENT_HIBERNATE, false); if (error) printk(KERN_ERR "PM: Error %d creating hibernation image\n", error); - /* Restore control flow magically appears here */ - restore_processor_state(); if (!in_suspend) events_check_enabled = false; diff --git a/kernel/power/snapshot.c b/kernel/power/snapshot.c index 3a970604308f..f155c62f1f2c 100644 --- a/kernel/power/snapshot.c +++ b/kernel/power/snapshot.c @@ -765,9 +765,9 @@ static bool memory_bm_pfn_present(struct memory_bitmap *bm, unsigned long pfn) */ static bool rtree_next_node(struct memory_bitmap *bm) { - bm->cur.node = list_entry(bm->cur.node->list.next, - struct rtree_node, list); - if (&bm->cur.node->list != &bm->cur.zone->leaves) { + if (!list_is_last(&bm->cur.node->list, &bm->cur.zone->leaves)) { + bm->cur.node = list_entry(bm->cur.node->list.next, + struct rtree_node, list); bm->cur.node_pfn += BM_BITS_PER_BLOCK; bm->cur.node_bit = 0; touch_softlockup_watchdog(); @@ -775,9 +775,9 @@ static bool rtree_next_node(struct memory_bitmap *bm) } /* No more nodes, goto next zone */ - bm->cur.zone = list_entry(bm->cur.zone->list.next, + if (!list_is_last(&bm->cur.zone->list, &bm->zones)) { + bm->cur.zone = list_entry(bm->cur.zone->list.next, struct mem_zone_bm_rtree, list); - if (&bm->cur.zone->list != &bm->zones) { bm->cur.node = list_entry(bm->cur.zone->leaves.next, struct rtree_node, list); bm->cur.node_pfn = 0; diff --git a/kernel/printk/braille.c b/kernel/printk/braille.c index 276762f3a460..d5760c42f042 100644 --- a/kernel/printk/braille.c +++ b/kernel/printk/braille.c @@ -9,10 +9,10 @@ char *_braille_console_setup(char **str, char **brl_options) { - if (!memcmp(*str, "brl,", 4)) { + if (!strncmp(*str, "brl,", 4)) { *brl_options = ""; *str += 4; - } else if (!memcmp(str, "brl=", 4)) { + } else if (!strncmp(*str, "brl=", 4)) { *brl_options = *str + 4; *str = strchr(*brl_options, ','); if (!*str) diff --git a/kernel/rcu/sync.c b/kernel/rcu/sync.c index be922c9f3d37..b49cf3ac2d47 100644 --- a/kernel/rcu/sync.c +++ b/kernel/rcu/sync.c @@ -68,6 +68,7 @@ void rcu_sync_lockdep_assert(struct rcu_sync *rsp) RCU_LOCKDEP_WARN(!gp_ops[rsp->gp_type].held(), "suspicious rcu_sync_is_idle() usage"); } +EXPORT_SYMBOL_GPL(rcu_sync_lockdep_assert); #endif /** @@ -83,6 +84,18 @@ void rcu_sync_init(struct rcu_sync *rsp, enum rcu_sync_type type) } /** + * Must be called after rcu_sync_init() and before first use. + * + * Ensures rcu_sync_is_idle() returns false and rcu_sync_{enter,exit}() + * pairs turn into NO-OPs. + */ +void rcu_sync_enter_start(struct rcu_sync *rsp) +{ + rsp->gp_count++; + rsp->gp_state = GP_PASSED; +} + +/** * rcu_sync_enter() - Force readers onto slowpath * @rsp: Pointer to rcu_sync structure to use for synchronization * diff --git a/kernel/sched/Makefile b/kernel/sched/Makefile index c6a85f813dfd..dc93f1cffc6e 100644 --- a/kernel/sched/Makefile +++ b/kernel/sched/Makefile @@ -14,7 +14,8 @@ endif obj-y += core.o loadavg.o clock.o cputime.o obj-y += idle_task.o fair.o rt.o deadline.o stop_task.o energy.o obj-y += wait.o completion.o idle.o -obj-$(CONFIG_SMP) += cpupri.o cpudeadline.o +obj-$(CONFIG_SMP) += cpupri.o cpudeadline.o energy.o +obj-$(CONFIG_SCHED_WALT) += walt.o obj-$(CONFIG_SCHED_AUTOGROUP) += auto_group.o obj-$(CONFIG_SCHEDSTATS) += stats.o obj-$(CONFIG_SCHED_DEBUG) += debug.o diff --git a/kernel/sched/core.c b/kernel/sched/core.c index a94f6424103c..1df6da0094f0 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -89,6 +89,7 @@ #define CREATE_TRACE_POINTS #include <trace/events/sched.h> +#include "walt.h" DEFINE_MUTEX(sched_domains_mutex); DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues); @@ -287,6 +288,18 @@ int sysctl_sched_rt_runtime = 950000; /* cpus with isolated domains */ cpumask_var_t cpu_isolated_map; +struct rq * +lock_rq_of(struct task_struct *p, unsigned long *flags) +{ + return task_rq_lock(p, flags); +} + +void +unlock_rq_of(struct rq *rq, struct task_struct *p, unsigned long *flags) +{ + task_rq_unlock(rq, p, flags); +} + /* * this_rq_lock - lock this runqueue and disable interrupts. */ @@ -627,7 +640,10 @@ int get_nohz_timer_target(void) rcu_read_lock(); for_each_domain(cpu, sd) { for_each_cpu(i, sched_domain_span(sd)) { - if (!idle_cpu(i) && is_housekeeping_cpu(cpu)) { + if (cpu == i) + continue; + + if (!idle_cpu(i) && is_housekeeping_cpu(i)) { cpu = i; goto unlock; } @@ -1073,7 +1089,9 @@ static struct rq *move_queued_task(struct rq *rq, struct task_struct *p, int new dequeue_task(rq, p, 0); p->on_rq = TASK_ON_RQ_MIGRATING; + double_lock_balance(rq, cpu_rq(new_cpu)); set_task_cpu(p, new_cpu); + double_unlock_balance(rq, cpu_rq(new_cpu)); raw_spin_unlock(&rq->lock); rq = cpu_rq(new_cpu); @@ -1297,6 +1315,8 @@ void set_task_cpu(struct task_struct *p, unsigned int new_cpu) p->sched_class->migrate_task_rq(p); p->se.nr_migrations++; perf_event_task_migrate(p); + + walt_fixup_busy_time(p, new_cpu); } __set_task_cpu(p, new_cpu); @@ -1925,6 +1945,10 @@ try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags) { unsigned long flags; int cpu, success = 0; +#ifdef CONFIG_SMP + struct rq *rq; + u64 wallclock; +#endif /* * If we are going to wake up a thread waiting for CONDITION we @@ -1942,6 +1966,28 @@ try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags) success = 1; /* we're going to change ->state */ cpu = task_cpu(p); + /* + * Ensure we load p->on_rq _after_ p->state, otherwise it would + * be possible to, falsely, observe p->on_rq == 0 and get stuck + * in smp_cond_load_acquire() below. + * + * sched_ttwu_pending() try_to_wake_up() + * [S] p->on_rq = 1; [L] P->state + * UNLOCK rq->lock -----. + * \ + * +--- RMB + * schedule() / + * LOCK rq->lock -----' + * UNLOCK rq->lock + * + * [task p] + * [S] p->state = UNINTERRUPTIBLE [L] p->on_rq + * + * Pairs with the UNLOCK+LOCK on rq->lock from the + * last wakeup of our task and the schedule that got our task + * current. + */ + smp_rmb(); if (p->on_rq && ttwu_remote(p, wake_flags)) goto stat; @@ -1982,6 +2028,14 @@ try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags) */ smp_rmb(); + rq = cpu_rq(task_cpu(p)); + + raw_spin_lock(&rq->lock); + wallclock = walt_ktime_clock(); + walt_update_task_ravg(rq->curr, rq, TASK_UPDATE, wallclock, 0); + walt_update_task_ravg(p, rq, TASK_WAKE, wallclock, 0); + raw_spin_unlock(&rq->lock); + p->sched_contributes_to_load = !!task_contributes_to_load(p); p->state = TASK_WAKING; @@ -1989,10 +2043,12 @@ try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags) p->sched_class->task_waking(p); cpu = select_task_rq(p, p->wake_cpu, SD_BALANCE_WAKE, wake_flags); + if (task_cpu(p) != cpu) { wake_flags |= WF_MIGRATED; set_task_cpu(p, cpu); } + #endif /* CONFIG_SMP */ ttwu_queue(p, cpu); @@ -2041,8 +2097,13 @@ static void try_to_wake_up_local(struct task_struct *p) trace_sched_waking(p); - if (!task_on_rq_queued(p)) + if (!task_on_rq_queued(p)) { + u64 wallclock = walt_ktime_clock(); + + walt_update_task_ravg(rq->curr, rq, TASK_UPDATE, wallclock, 0); + walt_update_task_ravg(p, rq, TASK_WAKE, wallclock, 0); ttwu_activate(rq, p, ENQUEUE_WAKEUP); + } ttwu_do_wakeup(rq, p, 0); ttwu_stat(p, smp_processor_id(), 0); @@ -2108,6 +2169,7 @@ static void __sched_fork(unsigned long clone_flags, struct task_struct *p) p->se.nr_migrations = 0; p->se.vruntime = 0; INIT_LIST_HEAD(&p->se.group_node); + walt_init_new_task_load(p); #ifdef CONFIG_SCHEDSTATS memset(&p->se.statistics, 0, sizeof(p->se.statistics)); @@ -2375,6 +2437,9 @@ void wake_up_new_task(struct task_struct *p) struct rq *rq; raw_spin_lock_irqsave(&p->pi_lock, flags); + + walt_init_new_task_load(p); + /* Initialize new task's runnable average */ init_entity_runnable_average(&p->se); #ifdef CONFIG_SMP @@ -2387,6 +2452,7 @@ void wake_up_new_task(struct task_struct *p) #endif rq = __task_rq_lock(p); + walt_mark_task_starting(p); activate_task(rq, p, ENQUEUE_WAKEUP_NEW); p->on_rq = TASK_ON_RQ_QUEUED; trace_sched_wakeup_new(p); @@ -2768,6 +2834,36 @@ unsigned long nr_iowait_cpu(int cpu) return atomic_read(&this->nr_iowait); } +#ifdef CONFIG_CPU_QUIET +u64 nr_running_integral(unsigned int cpu) +{ + unsigned int seqcnt; + u64 integral; + struct rq *q; + + if (cpu >= nr_cpu_ids) + return 0; + + q = cpu_rq(cpu); + + /* + * Update average to avoid reading stalled value if there were + * no run-queue changes for a long time. On the other hand if + * the changes are happening right now, just read current value + * directly. + */ + + seqcnt = read_seqcount_begin(&q->ave_seqcnt); + integral = do_nr_running_integral(q); + if (read_seqcount_retry(&q->ave_seqcnt, seqcnt)) { + read_seqcount_begin(&q->ave_seqcnt); + integral = q->nr_running_integral; + } + + return integral; +} +#endif + void get_iowait_load(unsigned long *nr_waiters, unsigned long *load) { struct rq *rq = this_rq(); @@ -2855,20 +2951,76 @@ unsigned long long task_sched_runtime(struct task_struct *p) } #ifdef CONFIG_CPU_FREQ_GOV_SCHED -static unsigned long sum_capacity_reqs(unsigned long cfs_cap, - struct sched_capacity_reqs *scr) + +static inline +unsigned long add_capacity_margin(unsigned long cpu_capacity) { - unsigned long total = cfs_cap + scr->rt; + cpu_capacity = cpu_capacity * capacity_margin; + cpu_capacity /= SCHED_CAPACITY_SCALE; + return cpu_capacity; +} - total = total * capacity_margin; - total /= SCHED_CAPACITY_SCALE; - total += scr->dl; - return total; +static inline +unsigned long sum_capacity_reqs(unsigned long cfs_cap, + struct sched_capacity_reqs *scr) +{ + unsigned long total = add_capacity_margin(cfs_cap + scr->rt); + return total += scr->dl; } -static void sched_freq_tick(int cpu) +static void sched_freq_tick_pelt(int cpu) { + unsigned long cpu_utilization = capacity_max; + unsigned long capacity_curr = capacity_curr_of(cpu); struct sched_capacity_reqs *scr; + + scr = &per_cpu(cpu_sched_capacity_reqs, cpu); + if (sum_capacity_reqs(cpu_utilization, scr) < capacity_curr) + return; + + /* + * To make free room for a task that is building up its "real" + * utilization and to harm its performance the least, request + * a jump to a higher OPP as soon as the margin of free capacity + * is impacted (specified by capacity_margin). + */ + set_cfs_cpu_capacity(cpu, true, cpu_utilization); +} + +#ifdef CONFIG_SCHED_WALT +static void sched_freq_tick_walt(int cpu) +{ + unsigned long cpu_utilization = cpu_util(cpu); + unsigned long capacity_curr = capacity_curr_of(cpu); + + if (walt_disabled || !sysctl_sched_use_walt_cpu_util) + return sched_freq_tick_pelt(cpu); + + /* + * Add a margin to the WALT utilization. + * NOTE: WALT tracks a single CPU signal for all the scheduling + * classes, thus this margin is going to be added to the DL class as + * well, which is something we do not do in sched_freq_tick_pelt case. + */ + cpu_utilization = add_capacity_margin(cpu_utilization); + if (cpu_utilization <= capacity_curr) + return; + + /* + * It is likely that the load is growing so we + * keep the added margin in our request as an + * extra boost. + */ + set_cfs_cpu_capacity(cpu, true, cpu_utilization); + +} +#define _sched_freq_tick(cpu) sched_freq_tick_walt(cpu) +#else +#define _sched_freq_tick(cpu) sched_freq_tick_pelt(cpu) +#endif /* CONFIG_SCHED_WALT */ + +static void sched_freq_tick(int cpu) +{ unsigned long capacity_orig, capacity_curr; if (!sched_freq()) @@ -2879,19 +3031,11 @@ static void sched_freq_tick(int cpu) if (capacity_curr == capacity_orig) return; - /* - * To make free room for a task that is building up its "real" - * utilization and to harm its performance the least, request - * a jump to max OPP as soon as the margin of free capacity is - * impacted (specified by capacity_margin). - */ - scr = &per_cpu(cpu_sched_capacity_reqs, cpu); - if (capacity_curr < sum_capacity_reqs(cpu_util(cpu), scr)) - set_cfs_cpu_capacity(cpu, true, capacity_max); + _sched_freq_tick(cpu); } #else static inline void sched_freq_tick(int cpu) { } -#endif +#endif /* CONFIG_CPU_FREQ_GOV_SCHED */ /* * This function gets called by the timer code, with HZ frequency. @@ -2906,9 +3050,12 @@ void scheduler_tick(void) sched_clock_tick(); raw_spin_lock(&rq->lock); + walt_set_window_start(rq); update_rq_clock(rq); curr->sched_class->task_tick(rq, curr, 0); update_cpu_load_active(rq); + walt_update_task_ravg(rq->curr, rq, TASK_UPDATE, + walt_ktime_clock(), 0); calc_global_load_tick(rq); sched_freq_tick(cpu); raw_spin_unlock(&rq->lock); @@ -3147,6 +3294,7 @@ static void __sched notrace __schedule(bool preempt) unsigned long *switch_count; struct rq *rq; int cpu; + u64 wallclock; cpu = smp_processor_id(); rq = cpu_rq(cpu); @@ -3208,6 +3356,9 @@ static void __sched notrace __schedule(bool preempt) update_rq_clock(rq); next = pick_next_task(rq, prev); + wallclock = walt_ktime_clock(); + walt_update_task_ravg(prev, rq, PUT_PREV_TASK, wallclock, 0); + walt_update_task_ravg(next, rq, PICK_NEXT_TASK, wallclock, 0); clear_tsk_need_resched(prev); clear_preempt_need_resched(); rq->clock_skip_update = 0; @@ -4991,14 +5142,16 @@ void show_state_filter(unsigned long state_filter) /* * reset the NMI-timeout, listing all files on a slow * console might take a lot of time: + * Also, reset softlockup watchdogs on all CPUs, because + * another CPU might be blocked waiting for us to process + * an IPI. */ touch_nmi_watchdog(); + touch_all_softlockup_watchdogs(); if (!state_filter || (p->state & state_filter)) sched_show_task(p); } - touch_all_softlockup_watchdogs(); - #ifdef CONFIG_SCHED_DEBUG sysrq_sched_debug_show(); #endif @@ -5032,6 +5185,7 @@ void init_idle(struct task_struct *idle, int cpu) raw_spin_lock(&rq->lock); __sched_fork(0, idle); + idle->state = TASK_RUNNING; idle->se.exec_start = sched_clock(); @@ -5628,6 +5782,9 @@ migration_call(struct notifier_block *nfb, unsigned long action, void *hcpu) switch (action & ~CPU_TASKS_FROZEN) { case CPU_UP_PREPARE: + raw_spin_lock_irqsave(&rq->lock, flags); + walt_set_window_start(rq); + raw_spin_unlock_irqrestore(&rq->lock, flags); rq->calc_load_update = calc_load_update; account_reset_rq(rq); break; @@ -5648,6 +5805,7 @@ migration_call(struct notifier_block *nfb, unsigned long action, void *hcpu) sched_ttwu_pending(); /* Update our root-domain */ raw_spin_lock_irqsave(&rq->lock, flags); + walt_migrate_sync_cpu(cpu); if (rq->rd) { BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span)); set_rq_offline(rq); @@ -7492,6 +7650,7 @@ void __init sched_init_smp(void) { cpumask_var_t non_isolated_cpus; + walt_init_cpu_efficiency(); alloc_cpumask_var(&non_isolated_cpus, GFP_KERNEL); alloc_cpumask_var(&fallback_doms, GFP_KERNEL); @@ -7669,6 +7828,11 @@ void __init sched_init(void) rq->idle_stamp = 0; rq->avg_idle = 2*sysctl_sched_migration_cost; rq->max_idle_balance_cost = sysctl_sched_migration_cost; +#ifdef CONFIG_SCHED_WALT + rq->cur_irqload = 0; + rq->avg_irqload = 0; + rq->irqload_ts = 0; +#endif INIT_LIST_HEAD(&rq->cfs_tasks); @@ -8792,7 +8956,6 @@ struct cgroup_subsys cpu_cgrp_subsys = { .fork = cpu_cgroup_fork, .can_attach = cpu_cgroup_can_attach, .attach = cpu_cgroup_attach, - .allow_attach = subsys_cgroup_allow_attach, .legacy_cftypes = cpu_files, .early_init = 1, }; diff --git a/kernel/sched/cpufreq_sched.c b/kernel/sched/cpufreq_sched.c index e1d208e101ed..f6f9b9b3a4a8 100644 --- a/kernel/sched/cpufreq_sched.c +++ b/kernel/sched/cpufreq_sched.c @@ -19,7 +19,8 @@ #include "sched.h" -#define THROTTLE_NSEC 50000000 /* 50ms default */ +#define THROTTLE_DOWN_NSEC 50000000 /* 50ms default */ +#define THROTTLE_UP_NSEC 500000 /* 500us default */ struct static_key __read_mostly __sched_freq = STATIC_KEY_INIT_FALSE; static bool __read_mostly cpufreq_driver_slow; @@ -33,8 +34,10 @@ DEFINE_PER_CPU(struct sched_capacity_reqs, cpu_sched_capacity_reqs); /** * gov_data - per-policy data internal to the governor - * @throttle: next throttling period expiry. Derived from throttle_nsec - * @throttle_nsec: throttle period length in nanoseconds + * @up_throttle: next throttling period expiry if increasing OPP + * @down_throttle: next throttling period expiry if decreasing OPP + * @up_throttle_nsec: throttle period length in nanoseconds if increasing OPP + * @down_throttle_nsec: throttle period length in nanoseconds if decreasing OPP * @task: worker thread for dvfs transition that may block/sleep * @irq_work: callback used to wake up worker thread * @requested_freq: last frequency requested by the sched governor @@ -48,8 +51,10 @@ DEFINE_PER_CPU(struct sched_capacity_reqs, cpu_sched_capacity_reqs); * call down_write(policy->rwsem). */ struct gov_data { - ktime_t throttle; - unsigned int throttle_nsec; + ktime_t up_throttle; + ktime_t down_throttle; + unsigned int up_throttle_nsec; + unsigned int down_throttle_nsec; struct task_struct *task; struct irq_work irq_work; unsigned int requested_freq; @@ -66,25 +71,29 @@ static void cpufreq_sched_try_driver_target(struct cpufreq_policy *policy, __cpufreq_driver_target(policy, freq, CPUFREQ_RELATION_L); - gd->throttle = ktime_add_ns(ktime_get(), gd->throttle_nsec); + gd->up_throttle = ktime_add_ns(ktime_get(), gd->up_throttle_nsec); + gd->down_throttle = ktime_add_ns(ktime_get(), gd->down_throttle_nsec); up_write(&policy->rwsem); } -static bool finish_last_request(struct gov_data *gd) +static bool finish_last_request(struct gov_data *gd, unsigned int cur_freq) { ktime_t now = ktime_get(); - if (ktime_after(now, gd->throttle)) + ktime_t throttle = gd->requested_freq < cur_freq ? + gd->down_throttle : gd->up_throttle; + + if (ktime_after(now, throttle)) return false; while (1) { - int usec_left = ktime_to_ns(ktime_sub(gd->throttle, now)); + int usec_left = ktime_to_ns(ktime_sub(throttle, now)); usec_left /= NSEC_PER_USEC; trace_cpufreq_sched_throttled(usec_left); usleep_range(usec_left, usec_left + 100); now = ktime_get(); - if (ktime_after(now, gd->throttle)) + if (ktime_after(now, throttle)) return true; } } @@ -122,13 +131,15 @@ static int cpufreq_sched_thread(void *data) new_request = gd->requested_freq; if (new_request == last_request) { set_current_state(TASK_INTERRUPTIBLE); + if (kthread_should_stop()) + break; schedule(); } else { /* * if the frequency thread sleeps while waiting to be * unthrottled, start over to check for a newer request */ - if (finish_last_request(gd)) + if (finish_last_request(gd, policy->cur)) continue; last_request = new_request; cpufreq_sched_try_driver_target(policy, new_request); @@ -190,9 +201,14 @@ static void update_fdomain_capacity_request(int cpu) goto out; freq_new = policy->freq_table[index_new].frequency; + if (freq_new > policy->max) + freq_new = policy->max; + + if (freq_new < policy->min) + freq_new = policy->min; + trace_cpufreq_sched_request_opp(cpu, capacity, freq_new, gd->requested_freq); - if (freq_new == gd->requested_freq) goto out; @@ -246,10 +262,17 @@ static inline void clear_sched_freq(void) static_key_slow_dec(&__sched_freq); } +static struct attribute_group sched_attr_group_gov_pol; +static struct attribute_group *get_sysfs_attr(void) +{ + return &sched_attr_group_gov_pol; +} + static int cpufreq_sched_policy_init(struct cpufreq_policy *policy) { struct gov_data *gd; int cpu; + int rc; for_each_cpu(cpu, policy->cpus) memset(&per_cpu(cpu_sched_capacity_reqs, cpu), 0, @@ -259,12 +282,20 @@ static int cpufreq_sched_policy_init(struct cpufreq_policy *policy) if (!gd) return -ENOMEM; - gd->throttle_nsec = policy->cpuinfo.transition_latency ? + gd->up_throttle_nsec = policy->cpuinfo.transition_latency ? policy->cpuinfo.transition_latency : - THROTTLE_NSEC; + THROTTLE_UP_NSEC; + gd->down_throttle_nsec = THROTTLE_DOWN_NSEC; pr_debug("%s: throttle threshold = %u [ns]\n", - __func__, gd->throttle_nsec); + __func__, gd->up_throttle_nsec); + + rc = sysfs_create_group(get_governor_parent_kobj(policy), get_sysfs_attr()); + if (rc) { + pr_err("%s: couldn't create sysfs attributes: %d\n", __func__, rc); + goto err; + } + policy->governor_data = gd; if (cpufreq_driver_is_slow()) { cpufreq_driver_slow = true; gd->task = kthread_create(cpufreq_sched_thread, policy, @@ -281,12 +312,12 @@ static int cpufreq_sched_policy_init(struct cpufreq_policy *policy) init_irq_work(&gd->irq_work, cpufreq_sched_irq_work); } - policy->governor_data = gd; set_sched_freq(); return 0; err: + policy->governor_data = NULL; kfree(gd); return -ENOMEM; } @@ -301,6 +332,8 @@ static int cpufreq_sched_policy_exit(struct cpufreq_policy *policy) put_task_struct(gd->task); } + sysfs_remove_group(get_governor_parent_kobj(policy), get_sysfs_attr()); + policy->governor_data = NULL; kfree(gd); @@ -317,6 +350,21 @@ static int cpufreq_sched_start(struct cpufreq_policy *policy) return 0; } +static void cpufreq_sched_limits(struct cpufreq_policy *policy) +{ + unsigned int clamp_freq; + struct gov_data *gd = policy->governor_data;; + + pr_debug("limit event for cpu %u: %u - %u kHz, currently %u kHz\n", + policy->cpu, policy->min, policy->max, + policy->cur); + + clamp_freq = clamp(gd->requested_freq, policy->min, policy->max); + + if (policy->cur != clamp_freq) + __cpufreq_driver_target(policy, clamp_freq, CPUFREQ_RELATION_L); +} + static int cpufreq_sched_stop(struct cpufreq_policy *policy) { int cpu; @@ -340,11 +388,95 @@ static int cpufreq_sched_setup(struct cpufreq_policy *policy, case CPUFREQ_GOV_STOP: return cpufreq_sched_stop(policy); case CPUFREQ_GOV_LIMITS: + cpufreq_sched_limits(policy); break; } return 0; } +/* Tunables */ +static ssize_t show_up_throttle_nsec(struct gov_data *gd, char *buf) +{ + return sprintf(buf, "%u\n", gd->up_throttle_nsec); +} + +static ssize_t store_up_throttle_nsec(struct gov_data *gd, + const char *buf, size_t count) +{ + int ret; + long unsigned int val; + + ret = kstrtoul(buf, 0, &val); + if (ret < 0) + return ret; + gd->up_throttle_nsec = val; + return count; +} + +static ssize_t show_down_throttle_nsec(struct gov_data *gd, char *buf) +{ + return sprintf(buf, "%u\n", gd->down_throttle_nsec); +} + +static ssize_t store_down_throttle_nsec(struct gov_data *gd, + const char *buf, size_t count) +{ + int ret; + long unsigned int val; + + ret = kstrtoul(buf, 0, &val); + if (ret < 0) + return ret; + gd->down_throttle_nsec = val; + return count; +} + +/* + * Create show/store routines + * - sys: One governor instance for complete SYSTEM + * - pol: One governor instance per struct cpufreq_policy + */ +#define show_gov_pol_sys(file_name) \ +static ssize_t show_##file_name##_gov_pol \ +(struct cpufreq_policy *policy, char *buf) \ +{ \ + return show_##file_name(policy->governor_data, buf); \ +} + +#define store_gov_pol_sys(file_name) \ +static ssize_t store_##file_name##_gov_pol \ +(struct cpufreq_policy *policy, const char *buf, size_t count) \ +{ \ + return store_##file_name(policy->governor_data, buf, count); \ +} + +#define gov_pol_attr_rw(_name) \ + static struct freq_attr _name##_gov_pol = \ + __ATTR(_name, 0644, show_##_name##_gov_pol, store_##_name##_gov_pol) + +#define show_store_gov_pol_sys(file_name) \ + show_gov_pol_sys(file_name); \ + store_gov_pol_sys(file_name) +#define tunable_handlers(file_name) \ + show_gov_pol_sys(file_name); \ + store_gov_pol_sys(file_name); \ + gov_pol_attr_rw(file_name) + +tunable_handlers(down_throttle_nsec); +tunable_handlers(up_throttle_nsec); + +/* Per policy governor instance */ +static struct attribute *sched_attributes_gov_pol[] = { + &up_throttle_nsec_gov_pol.attr, + &down_throttle_nsec_gov_pol.attr, + NULL, +}; + +static struct attribute_group sched_attr_group_gov_pol = { + .attrs = sched_attributes_gov_pol, + .name = "sched", +}; + #ifndef CONFIG_CPU_FREQ_DEFAULT_GOV_SCHED static #endif diff --git a/kernel/sched/cputime.c b/kernel/sched/cputime.c index f74ea89e77a8..acde1d7c763c 100644 --- a/kernel/sched/cputime.c +++ b/kernel/sched/cputime.c @@ -5,6 +5,7 @@ #include <linux/static_key.h> #include <linux/context_tracking.h> #include "sched.h" +#include "walt.h" #ifdef CONFIG_IRQ_TIME_ACCOUNTING @@ -49,6 +50,10 @@ void irqtime_account_irq(struct task_struct *curr) unsigned long flags; s64 delta; int cpu; +#ifdef CONFIG_SCHED_WALT + u64 wallclock; + bool account = true; +#endif if (!sched_clock_irqtime) return; @@ -56,6 +61,9 @@ void irqtime_account_irq(struct task_struct *curr) local_irq_save(flags); cpu = smp_processor_id(); +#ifdef CONFIG_SCHED_WALT + wallclock = sched_clock_cpu(cpu); +#endif delta = sched_clock_cpu(cpu) - __this_cpu_read(irq_start_time); __this_cpu_add(irq_start_time, delta); @@ -70,8 +78,16 @@ void irqtime_account_irq(struct task_struct *curr) __this_cpu_add(cpu_hardirq_time, delta); else if (in_serving_softirq() && curr != this_cpu_ksoftirqd()) __this_cpu_add(cpu_softirq_time, delta); +#ifdef CONFIG_SCHED_WALT + else + account = false; +#endif irq_time_write_end(); +#ifdef CONFIG_SCHED_WALT + if (account) + walt_account_irqtime(cpu, curr, delta, wallclock); +#endif local_irq_restore(flags); } EXPORT_SYMBOL_GPL(irqtime_account_irq); @@ -600,19 +616,25 @@ static void cputime_adjust(struct task_cputime *curr, stime = curr->stime; utime = curr->utime; - if (utime == 0) { - stime = rtime; + /* + * If either stime or both stime and utime are 0, assume all runtime is + * userspace. Once a task gets some ticks, the monotonicy code at + * 'update' will ensure things converge to the observed ratio. + */ + if (stime == 0) { + utime = rtime; goto update; } - if (stime == 0) { - utime = rtime; + if (utime == 0) { + stime = rtime; goto update; } stime = scale_stime((__force u64)stime, (__force u64)rtime, (__force u64)(stime + utime)); +update: /* * Make sure stime doesn't go backwards; this preserves monotonicity * for utime because rtime is monotonic. @@ -635,7 +657,6 @@ static void cputime_adjust(struct task_cputime *curr, stime = rtime - utime; } -update: prev->stime = stime; prev->utime = utime; out: diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 3217494e326e..d5965cfbd3d7 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -30,11 +30,13 @@ #include <linux/mempolicy.h> #include <linux/migrate.h> #include <linux/task_work.h> +#include <linux/module.h> #include <trace/events/sched.h> #include "sched.h" #include "tune.h" +#include "walt.h" /* * Targeted preemption latency for CPU-bound tasks: @@ -51,6 +53,17 @@ unsigned int sysctl_sched_latency = 6000000ULL; unsigned int normalized_sysctl_sched_latency = 6000000ULL; +unsigned int sysctl_sched_is_big_little = 0; +unsigned int sysctl_sched_sync_hint_enable = 1; +unsigned int sysctl_sched_initial_task_util = 0; +unsigned int sysctl_sched_cstate_aware = 1; + +#ifdef CONFIG_SCHED_WALT +unsigned int sysctl_sched_use_walt_cpu_util = 1; +unsigned int sysctl_sched_use_walt_task_util = 1; +__read_mostly unsigned int sysctl_sched_walt_cpu_high_irqload = + (10 * NSEC_PER_MSEC); +#endif /* * The initial- and re-scaling of tunables is configurable * (default SCHED_TUNABLESCALING_LOG = *(1+ilog(ncpus)) @@ -683,13 +696,13 @@ void init_entity_runnable_average(struct sched_entity *se) sa->period_contrib = 1023; sa->load_avg = scale_load_down(se->load.weight); sa->load_sum = sa->load_avg * LOAD_AVG_MAX; - sa->util_avg = scale_load_down(SCHED_LOAD_SCALE); + sa->util_avg = sched_freq() ? + sysctl_sched_initial_task_util : + scale_load_down(SCHED_LOAD_SCALE); sa->util_sum = sa->util_avg * LOAD_AVG_MAX; /* when this task enqueue'ed, it will contribute to its cfs_rq's load_avg */ } -static inline unsigned long cfs_rq_runnable_load_avg(struct cfs_rq *cfs_rq); -static inline unsigned long cfs_rq_load_avg(struct cfs_rq *cfs_rq); #else void init_entity_runnable_average(struct sched_entity *se) { @@ -1194,8 +1207,6 @@ static void task_numa_assign(struct task_numa_env *env, { if (env->best_task) put_task_struct(env->best_task); - if (p) - get_task_struct(p); env->best_task = p; env->best_imp = imp; @@ -1263,20 +1274,30 @@ static void task_numa_compare(struct task_numa_env *env, long imp = env->p->numa_group ? groupimp : taskimp; long moveimp = imp; int dist = env->dist; + bool assigned = false; rcu_read_lock(); raw_spin_lock_irq(&dst_rq->lock); cur = dst_rq->curr; /* - * No need to move the exiting task, and this ensures that ->curr - * wasn't reaped and thus get_task_struct() in task_numa_assign() - * is safe under RCU read lock. - * Note that rcu_read_lock() itself can't protect from the final - * put_task_struct() after the last schedule(). + * No need to move the exiting task or idle task. */ if ((cur->flags & PF_EXITING) || is_idle_task(cur)) cur = NULL; + else { + /* + * The task_struct must be protected here to protect the + * p->numa_faults access in the task_weight since the + * numa_faults could already be freed in the following path: + * finish_task_switch() + * --> put_task_struct() + * --> __put_task_struct() + * --> task_numa_free() + */ + get_task_struct(cur); + } + raw_spin_unlock_irq(&dst_rq->lock); /* @@ -1360,6 +1381,7 @@ balance: */ if (!load_too_imbalanced(src_load, dst_load, env)) { imp = moveimp - 1; + put_task_struct(cur); cur = NULL; goto assign; } @@ -1385,9 +1407,16 @@ balance: env->dst_cpu = select_idle_sibling(env->p, env->dst_cpu); assign: + assigned = true; task_numa_assign(env, cur, imp); unlock: rcu_read_unlock(); + /* + * The dst_rq->curr isn't assigned. The protection for task_struct is + * finished. + */ + if (cur && !assigned) + put_task_struct(cur); } static void task_numa_find_cpu(struct task_numa_env *env, @@ -4168,8 +4197,14 @@ static inline void hrtick_update(struct rq *rq) } #endif +#ifdef CONFIG_SMP +static bool cpu_overutilized(int cpu); static inline unsigned long boosted_cpu_util(int cpu); +#else +#define boosted_cpu_util(cpu) cpu_util(cpu) +#endif +#ifdef CONFIG_SMP static void update_capacity_of(int cpu) { unsigned long req_cap; @@ -4182,8 +4217,7 @@ static void update_capacity_of(int cpu) req_cap = req_cap * SCHED_CAPACITY_SCALE / capacity_orig_of(cpu); set_cfs_cpu_capacity(cpu, true, req_cap); } - -static bool cpu_overutilized(int cpu); +#endif /* * The enqueue_task method is called before nr_running is @@ -4195,8 +4229,10 @@ enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags) { struct cfs_rq *cfs_rq; struct sched_entity *se = &p->se; +#ifdef CONFIG_SMP int task_new = flags & ENQUEUE_WAKEUP_NEW; int task_wakeup = flags & ENQUEUE_WAKEUP; +#endif for_each_sched_entity(se) { if (se->on_rq) @@ -4213,6 +4249,7 @@ enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags) if (cfs_rq_throttled(cfs_rq)) break; cfs_rq->h_nr_running++; + walt_inc_cfs_cumulative_runnable_avg(cfs_rq, p); flags = ENQUEUE_WAKEUP; } @@ -4220,6 +4257,7 @@ enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags) for_each_sched_entity(se) { cfs_rq = cfs_rq_of(se); cfs_rq->h_nr_running++; + walt_inc_cfs_cumulative_runnable_avg(cfs_rq, p); if (cfs_rq_throttled(cfs_rq)) break; @@ -4228,13 +4266,37 @@ enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags) update_cfs_shares(cfs_rq); } - if (!se) { + if (!se) add_nr_running(rq, 1); + +#ifdef CONFIG_SMP + + /* + * Update SchedTune accounting. + * + * We do it before updating the CPU capacity to ensure the + * boost value of the current task is accounted for in the + * selection of the OPP. + * + * We do it also in the case where we enqueue a throttled task; + * we could argue that a throttled task should not boost a CPU, + * however: + * a) properly implementing CPU boosting considering throttled + * tasks will increase a lot the complexity of the solution + * b) it's not easy to quantify the benefits introduced by + * such a more complex solution. + * Thus, for the time being we go for the simple solution and boost + * also for throttled RQs. + */ + schedtune_enqueue_task(p, cpu_of(rq)); + + if (!se) { + walt_inc_cumulative_runnable_avg(rq, p); if (!task_new && !rq->rd->overutilized && - cpu_overutilized(rq->cpu)) + cpu_overutilized(rq->cpu)) { rq->rd->overutilized = true; - - schedtune_enqueue_task(p, cpu_of(rq)); + trace_sched_overutilized(true); + } /* * We want to potentially trigger a freq switch @@ -4246,6 +4308,8 @@ enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags) if (task_new || task_wakeup) update_capacity_of(cpu_of(rq)); } + +#endif /* CONFIG_SMP */ hrtick_update(rq); } @@ -4275,6 +4339,7 @@ static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags) if (cfs_rq_throttled(cfs_rq)) break; cfs_rq->h_nr_running--; + walt_dec_cfs_cumulative_runnable_avg(cfs_rq, p); /* Don't dequeue parent if it has other entities besides us */ if (cfs_rq->load.weight) { @@ -4295,6 +4360,7 @@ static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags) for_each_sched_entity(se) { cfs_rq = cfs_rq_of(se); cfs_rq->h_nr_running--; + walt_dec_cfs_cumulative_runnable_avg(cfs_rq, p); if (cfs_rq_throttled(cfs_rq)) break; @@ -4303,9 +4369,22 @@ static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags) update_cfs_shares(cfs_rq); } - if (!se) { + if (!se) sub_nr_running(rq, 1); - schedtune_dequeue_task(p, cpu_of(rq)); + +#ifdef CONFIG_SMP + + /* + * Update SchedTune accounting + * + * We do it before updating the CPU capacity to ensure the + * boost value of the current task is accounted for in the + * selection of the OPP. + */ + schedtune_dequeue_task(p, cpu_of(rq)); + + if (!se) { + walt_dec_cumulative_runnable_avg(rq, p); /* * We want to potentially trigger a freq switch @@ -4322,6 +4401,9 @@ static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags) set_cfs_cpu_capacity(cpu_of(rq), false, 0); } } + +#endif /* CONFIG_SMP */ + hrtick_update(rq); } @@ -4660,19 +4742,24 @@ static long effective_load(struct task_group *tg, int cpu, long wl, long wg) return wl; for_each_sched_entity(se) { - long w, W; + struct cfs_rq *cfs_rq = se->my_q; + long W, w = cfs_rq_load_avg(cfs_rq); - tg = se->my_q->tg; + tg = cfs_rq->tg; /* * W = @wg + \Sum rw_j */ - W = wg + calc_tg_weight(tg, se->my_q); + W = wg + atomic_long_read(&tg->load_avg); + + /* Ensure \Sum rw_j >= rw_i */ + W -= cfs_rq->tg_load_avg_contrib; + W += w; /* * w = rw_i + @wl */ - w = cfs_rq_load_avg(se->my_q) + wl; + w += wl; /* * wl = S * s'_i; see (2) @@ -4954,6 +5041,7 @@ static int sched_group_energy(struct energy_env *eenv) } while (sg = sg->next, sg != sd->groups); } next_cpu: + cpumask_clear_cpu(cpu, &visit_cpus); continue; } @@ -4966,44 +5054,6 @@ static inline bool cpu_in_sg(struct sched_group *sg, int cpu) return cpu != -1 && cpumask_test_cpu(cpu, sched_group_cpus(sg)); } -#ifdef CONFIG_SCHED_TUNE -static int energy_diff_evaluate(struct energy_env *eenv) -{ - unsigned int boost; - int nrg_delta; - - /* Return energy diff when boost margin is 0 */ -#ifdef CONFIG_CGROUP_SCHEDTUNE - boost = schedtune_task_boost(eenv->task); -#else - boost = get_sysctl_sched_cfs_boost(); -#endif - if (boost == 0) - return eenv->nrg.diff; - - /* Compute normalized energy diff */ - nrg_delta = schedtune_normalize_energy(eenv->nrg.diff); - eenv->nrg.delta = nrg_delta; - - eenv->payoff = schedtune_accept_deltas( - eenv->nrg.delta, - eenv->cap.delta, - eenv->task); - - /* - * When SchedTune is enabled, the energy_diff() function will return - * the computed energy payoff value. Since the energy_diff() return - * value is expected to be negative by its callers, this evaluation - * function return a negative value each time the evaluation return a - * positive payoff, which is the condition for the acceptance of - * a scheduling decision - */ - return -eenv->payoff; -} -#else /* CONFIG_SCHED_TUNE */ -#define energy_diff_evaluate(eenv) eenv->nrg.diff -#endif - /* * energy_diff(): Estimate the energy impact of changing the utilization * distribution. eenv specifies the change: utilisation amount, source, and @@ -5011,7 +5061,7 @@ static int energy_diff_evaluate(struct energy_env *eenv) * utilization is removed from or added to the system (e.g. task wake-up). If * both are specified, the utilization is migrated. */ -static int energy_diff(struct energy_env *eenv) +static inline int __energy_diff(struct energy_env *eenv) { struct sched_domain *sd; struct sched_group *sg; @@ -5059,9 +5109,86 @@ static int energy_diff(struct energy_env *eenv) eenv->nrg.diff = eenv->nrg.after - eenv->nrg.before; eenv->payoff = 0; - return energy_diff_evaluate(eenv); + trace_sched_energy_diff(eenv->task, + eenv->src_cpu, eenv->dst_cpu, eenv->util_delta, + eenv->nrg.before, eenv->nrg.after, eenv->nrg.diff, + eenv->cap.before, eenv->cap.after, eenv->cap.delta, + eenv->nrg.delta, eenv->payoff); + + return eenv->nrg.diff; } +#ifdef CONFIG_SCHED_TUNE + +struct target_nrg schedtune_target_nrg; + +/* + * System energy normalization + * Returns the normalized value, in the range [0..SCHED_LOAD_SCALE], + * corresponding to the specified energy variation. + */ +static inline int +normalize_energy(int energy_diff) +{ + u32 normalized_nrg; +#ifdef CONFIG_SCHED_DEBUG + int max_delta; + + /* Check for boundaries */ + max_delta = schedtune_target_nrg.max_power; + max_delta -= schedtune_target_nrg.min_power; + WARN_ON(abs(energy_diff) >= max_delta); +#endif + + /* Do scaling using positive numbers to increase the range */ + normalized_nrg = (energy_diff < 0) ? -energy_diff : energy_diff; + + /* Scale by energy magnitude */ + normalized_nrg <<= SCHED_LOAD_SHIFT; + + /* Normalize on max energy for target platform */ + normalized_nrg = reciprocal_divide( + normalized_nrg, schedtune_target_nrg.rdiv); + + return (energy_diff < 0) ? -normalized_nrg : normalized_nrg; +} + +static inline int +energy_diff(struct energy_env *eenv) +{ + int boost = schedtune_task_boost(eenv->task); + int nrg_delta; + + /* Conpute "absolute" energy diff */ + __energy_diff(eenv); + + /* Return energy diff when boost margin is 0 */ + if (boost == 0) + return eenv->nrg.diff; + + /* Compute normalized energy diff */ + nrg_delta = normalize_energy(eenv->nrg.diff); + eenv->nrg.delta = nrg_delta; + + eenv->payoff = schedtune_accept_deltas( + eenv->nrg.delta, + eenv->cap.delta, + eenv->task); + + /* + * When SchedTune is enabled, the energy_diff() function will return + * the computed energy payoff value. Since the energy_diff() return + * value is expected to be negative by its callers, this evaluation + * function return a negative value each time the evaluation return a + * positive payoff, which is the condition for the acceptance of + * a scheduling decision + */ + return -eenv->payoff; +} +#else /* CONFIG_SCHED_TUNE */ +#define energy_diff(eenv) __energy_diff(eenv) +#endif + /* * Detect M:N waker/wakee relationships via a switching-frequency heuristic. * A waker of many should wake a different task than the one last awakened @@ -5155,6 +5282,12 @@ static int wake_affine(struct sched_domain *sd, struct task_struct *p, int sync) static inline unsigned long task_util(struct task_struct *p) { +#ifdef CONFIG_SCHED_WALT + if (!walt_disabled && sysctl_sched_use_walt_task_util) { + unsigned long demand = p->ravg.demand; + return (demand << 10) / walt_ravg_window; + } +#endif return p->se.avg.util_avg; } @@ -5197,22 +5330,25 @@ static bool cpu_overutilized(int cpu) #ifdef CONFIG_SCHED_TUNE -static unsigned long -schedtune_margin(unsigned long signal, unsigned long boost) +static long +schedtune_margin(unsigned long signal, long boost) { - unsigned long long margin = 0; + long long margin = 0; /* * Signal proportional compensation (SPC) * * The Boost (B) value is used to compute a Margin (M) which is * proportional to the complement of the original Signal (S): - * M = B * (SCHED_LOAD_SCALE - S) + * M = B * (SCHED_LOAD_SCALE - S), if B is positive + * M = B * S, if B is negative * The obtained M could be used by the caller to "boost" S. */ - margin = SCHED_LOAD_SCALE - signal; - margin *= boost; - + if (boost >= 0) { + margin = SCHED_LOAD_SCALE - signal; + margin *= boost; + } else + margin = -signal * boost; /* * Fast integer division by constant: * Constant : (C) = 100 @@ -5228,37 +5364,29 @@ schedtune_margin(unsigned long signal, unsigned long boost) margin *= 1311; margin >>= 17; + if (boost < 0) + margin *= -1; return margin; } -static inline unsigned int +static inline int schedtune_cpu_margin(unsigned long util, int cpu) { - unsigned int boost; + int boost = schedtune_cpu_boost(cpu); -#ifdef CONFIG_CGROUP_SCHEDTUNE - boost = schedtune_cpu_boost(cpu); -#else - boost = get_sysctl_sched_cfs_boost(); -#endif if (boost == 0) return 0; return schedtune_margin(util, boost); } -static inline unsigned long +static inline long schedtune_task_margin(struct task_struct *task) { - unsigned int boost; + int boost = schedtune_task_boost(task); unsigned long util; - unsigned long margin; + long margin; -#ifdef CONFIG_CGROUP_SCHEDTUNE - boost = schedtune_task_boost(task); -#else - boost = get_sysctl_sched_cfs_boost(); -#endif if (boost == 0) return 0; @@ -5270,13 +5398,13 @@ schedtune_task_margin(struct task_struct *task) #else /* CONFIG_SCHED_TUNE */ -static inline unsigned int +static inline int schedtune_cpu_margin(unsigned long util, int cpu) { return 0; } -static inline unsigned int +static inline int schedtune_task_margin(struct task_struct *task) { return 0; @@ -5288,7 +5416,7 @@ static inline unsigned long boosted_cpu_util(int cpu) { unsigned long util = cpu_util(cpu); - unsigned long margin = schedtune_cpu_margin(util, cpu); + long margin = schedtune_cpu_margin(util, cpu); trace_sched_boost_cpu(cpu, util, margin); @@ -5299,7 +5427,9 @@ static inline unsigned long boosted_task_util(struct task_struct *task) { unsigned long util = task_util(task); - unsigned long margin = schedtune_task_margin(task); + long margin = schedtune_task_margin(task); + + trace_sched_boost_task(task, util, margin); return util + margin; } @@ -5455,15 +5585,20 @@ static int select_idle_sibling(struct task_struct *p, int target) struct sched_domain *sd; struct sched_group *sg; int i = task_cpu(p); + int best_idle = -1; + int best_idle_cstate = -1; + int best_idle_capacity = INT_MAX; - if (idle_cpu(target)) - return target; + if (!sysctl_sched_cstate_aware) { + if (idle_cpu(target)) + return target; - /* - * If the prevous cpu is cache affine and idle, don't be stupid. - */ - if (i != target && cpus_share_cache(i, target) && idle_cpu(i)) - return i; + /* + * If the prevous cpu is cache affine and idle, don't be stupid. + */ + if (i != target && cpus_share_cache(i, target) && idle_cpu(i)) + return i; + } /* * Otherwise, iterate the domains and find an elegible idle cpu. @@ -5476,30 +5611,165 @@ static int select_idle_sibling(struct task_struct *p, int target) tsk_cpus_allowed(p))) goto next; - for_each_cpu(i, sched_group_cpus(sg)) { - if (i == target || !idle_cpu(i)) - goto next; - } + if (sysctl_sched_cstate_aware) { + for_each_cpu_and(i, tsk_cpus_allowed(p), sched_group_cpus(sg)) { + struct rq *rq = cpu_rq(i); + int idle_idx = idle_get_state_idx(rq); + unsigned long new_usage = boosted_task_util(p); + unsigned long capacity_orig = capacity_orig_of(i); + if (new_usage > capacity_orig || !idle_cpu(i)) + goto next; + + if (i == target && new_usage <= capacity_curr_of(target)) + return target; + + if (best_idle < 0 || (idle_idx < best_idle_cstate && capacity_orig <= best_idle_capacity)) { + best_idle = i; + best_idle_cstate = idle_idx; + best_idle_capacity = capacity_orig; + } + } + } else { + for_each_cpu(i, sched_group_cpus(sg)) { + if (i == target || !idle_cpu(i)) + goto next; + } - target = cpumask_first_and(sched_group_cpus(sg), + target = cpumask_first_and(sched_group_cpus(sg), tsk_cpus_allowed(p)); - goto done; + goto done; + } next: sg = sg->next; } while (sg != sd->groups); } + if (best_idle > 0) + target = best_idle; + done: return target; } -static int energy_aware_wake_cpu(struct task_struct *p, int target) +static inline int find_best_target(struct task_struct *p, bool boosted, bool prefer_idle) +{ + int iter_cpu; + int target_cpu = -1; + int target_util = 0; + int backup_capacity = 0; + int best_idle_cpu = -1; + int best_idle_cstate = INT_MAX; + int backup_cpu = -1; + unsigned long task_util_boosted, new_util; + + task_util_boosted = boosted_task_util(p); + for (iter_cpu = 0; iter_cpu < NR_CPUS; iter_cpu++) { + int cur_capacity; + struct rq *rq; + int idle_idx; + + /* + * Iterate from higher cpus for boosted tasks. + */ + int i = boosted ? NR_CPUS-iter_cpu-1 : iter_cpu; + + if (!cpu_online(i) || !cpumask_test_cpu(i, tsk_cpus_allowed(p))) + continue; + + /* + * p's blocked utilization is still accounted for on prev_cpu + * so prev_cpu will receive a negative bias due to the double + * accounting. However, the blocked utilization may be zero. + */ + new_util = cpu_util(i) + task_util_boosted; + + /* + * Ensure minimum capacity to grant the required boost. + * The target CPU can be already at a capacity level higher + * than the one required to boost the task. + */ + if (new_util > capacity_orig_of(i)) + continue; + +#ifdef CONFIG_SCHED_WALT + if (walt_cpu_high_irqload(i)) + continue; +#endif + /* + * Unconditionally favoring tasks that prefer idle cpus to + * improve latency. + */ + if (idle_cpu(i) && prefer_idle) { + if (best_idle_cpu < 0) + best_idle_cpu = i; + continue; + } + + cur_capacity = capacity_curr_of(i); + rq = cpu_rq(i); + idle_idx = idle_get_state_idx(rq); + + if (new_util < cur_capacity) { + if (cpu_rq(i)->nr_running) { + if (prefer_idle) { + /* Find a target cpu with highest + * utilization. + */ + if (target_util == 0 || + target_util < new_util) { + target_cpu = i; + target_util = new_util; + } + } else { + /* Find a target cpu with lowest + * utilization. + */ + if (target_util == 0 || + target_util > new_util) { + target_cpu = i; + target_util = new_util; + } + } + } else if (!prefer_idle) { + if (best_idle_cpu < 0 || + (sysctl_sched_cstate_aware && + best_idle_cstate > idle_idx)) { + best_idle_cstate = idle_idx; + best_idle_cpu = i; + } + } + } else if (backup_capacity == 0 || + backup_capacity > cur_capacity) { + // Find a backup cpu with least capacity. + backup_capacity = cur_capacity; + backup_cpu = i; + } + } + + if (prefer_idle && best_idle_cpu >= 0) + target_cpu = best_idle_cpu; + else if (target_cpu < 0) + target_cpu = best_idle_cpu >= 0 ? best_idle_cpu : backup_cpu; + + return target_cpu; +} + +static int energy_aware_wake_cpu(struct task_struct *p, int target, int sync) { struct sched_domain *sd; struct sched_group *sg, *sg_target; int target_max_cap = INT_MAX; int target_cpu = task_cpu(p); + unsigned long task_util_boosted, new_util; int i; + if (sysctl_sched_sync_hint_enable && sync) { + int cpu = smp_processor_id(); + cpumask_t search_cpus; + cpumask_and(&search_cpus, tsk_cpus_allowed(p), cpu_online_mask); + if (cpumask_test_cpu(cpu, &search_cpus)) + return cpu; + } + sd = rcu_dereference(per_cpu(sd_ea, task_cpu(p))); if (!sd) @@ -5508,50 +5778,76 @@ static int energy_aware_wake_cpu(struct task_struct *p, int target) sg = sd->groups; sg_target = sg; - /* - * Find group with sufficient capacity. We only get here if no cpu is - * overutilized. We may end up overutilizing a cpu by adding the task, - * but that should not be any worse than select_idle_sibling(). - * load_balance() should sort it out later as we get above the tipping - * point. - */ - do { - /* Assuming all cpus are the same in group */ - int max_cap_cpu = group_first_cpu(sg); + if (sysctl_sched_is_big_little) { /* - * Assume smaller max capacity means more energy-efficient. - * Ideally we should query the energy model for the right - * answer but it easily ends up in an exhaustive search. + * Find group with sufficient capacity. We only get here if no cpu is + * overutilized. We may end up overutilizing a cpu by adding the task, + * but that should not be any worse than select_idle_sibling(). + * load_balance() should sort it out later as we get above the tipping + * point. */ - if (capacity_of(max_cap_cpu) < target_max_cap && - task_fits_max(p, max_cap_cpu)) { - sg_target = sg; - target_max_cap = capacity_of(max_cap_cpu); - } - } while (sg = sg->next, sg != sd->groups); + do { + /* Assuming all cpus are the same in group */ + int max_cap_cpu = group_first_cpu(sg); - /* Find cpu with sufficient capacity */ - for_each_cpu_and(i, tsk_cpus_allowed(p), sched_group_cpus(sg_target)) { - /* - * p's blocked utilization is still accounted for on prev_cpu - * so prev_cpu will receive a negative bias due to the double - * accounting. However, the blocked utilization may be zero. - */ - int new_util = cpu_util(i) + boosted_task_util(p); + /* + * Assume smaller max capacity means more energy-efficient. + * Ideally we should query the energy model for the right + * answer but it easily ends up in an exhaustive search. + */ + if (capacity_of(max_cap_cpu) < target_max_cap && + task_fits_max(p, max_cap_cpu)) { + sg_target = sg; + target_max_cap = capacity_of(max_cap_cpu); + } + } while (sg = sg->next, sg != sd->groups); - if (new_util > capacity_orig_of(i)) - continue; + task_util_boosted = boosted_task_util(p); + /* Find cpu with sufficient capacity */ + for_each_cpu_and(i, tsk_cpus_allowed(p), sched_group_cpus(sg_target)) { + /* + * p's blocked utilization is still accounted for on prev_cpu + * so prev_cpu will receive a negative bias due to the double + * accounting. However, the blocked utilization may be zero. + */ + new_util = cpu_util(i) + task_util_boosted; - if (new_util < capacity_curr_of(i)) { - target_cpu = i; - if (cpu_rq(i)->nr_running) - break; - } + /* + * Ensure minimum capacity to grant the required boost. + * The target CPU can be already at a capacity level higher + * than the one required to boost the task. + */ + if (new_util > capacity_orig_of(i)) + continue; + + if (new_util < capacity_curr_of(i)) { + target_cpu = i; + if (cpu_rq(i)->nr_running) + break; + } - /* cpu has capacity at higher OPP, keep it as fallback */ - if (target_cpu == task_cpu(p)) - target_cpu = i; + /* cpu has capacity at higher OPP, keep it as fallback */ + if (target_cpu == task_cpu(p)) + target_cpu = i; + } + } else { + /* + * Find a cpu with sufficient capacity + */ +#ifdef CONFIG_CGROUP_SCHEDTUNE + bool boosted = schedtune_task_boost(p) > 0; + bool prefer_idle = schedtune_prefer_idle(p) > 0; +#else + bool boosted = 0; + bool prefer_idle = 0; +#endif + int tmp_target = find_best_target(p, boosted, prefer_idle); + if (tmp_target >= 0) { + target_cpu = tmp_target; + if ((boosted || prefer_idle) && idle_cpu(target_cpu)) + return target_cpu; + } } if (target_cpu != task_cpu(p)) { @@ -5628,7 +5924,7 @@ select_task_rq_fair(struct task_struct *p, int prev_cpu, int sd_flag, int wake_f if (!sd) { if (energy_aware() && !cpu_rq(cpu)->rd->overutilized) - new_cpu = energy_aware_wake_cpu(p, prev_cpu); + new_cpu = energy_aware_wake_cpu(p, prev_cpu, sync); else if (sd_flag & SD_BALANCE_WAKE) /* XXX always ? */ new_cpu = select_idle_sibling(p, new_cpu); @@ -5699,6 +5995,8 @@ static void task_dead_fair(struct task_struct *p) { remove_entity_load_avg(&p->se); } +#else +#define task_fits_max(p, cpu) true #endif /* CONFIG_SMP */ static unsigned long @@ -6404,7 +6702,9 @@ static void detach_task(struct task_struct *p, struct lb_env *env) deactivate_task(env->src_rq, p, 0); p->on_rq = TASK_ON_RQ_MIGRATING; + double_lock_balance(env->src_rq, env->dst_rq); set_task_cpu(p, env->dst_cpu); + double_unlock_balance(env->src_rq, env->dst_rq); } /* @@ -6825,7 +7125,10 @@ static void update_cpu_capacity(struct sched_domain *sd, int cpu) mcc->cpu = cpu; #ifdef CONFIG_SCHED_DEBUG raw_spin_unlock_irqrestore(&mcc->lock, flags); - //pr_info("CPU%d: update max cpu_capacity %lu\n", cpu, capacity); +/* + printk_deferred(KERN_INFO "CPU%d: update max cpu_capacity %lu\n", + cpu, capacity); +*/ goto skip_unlock; #endif } @@ -7048,7 +7351,7 @@ static inline void update_sg_lb_stats(struct lb_env *env, bool *overload, bool *overutilized) { unsigned long load; - int i; + int i, nr_running; memset(sgs, 0, sizeof(*sgs)); @@ -7065,7 +7368,8 @@ static inline void update_sg_lb_stats(struct lb_env *env, sgs->group_util += cpu_util(i); sgs->sum_nr_running += rq->cfs.h_nr_running; - if (rq->nr_running > 1) + nr_running = rq->nr_running; + if (nr_running > 1) *overload = true; #ifdef CONFIG_NUMA_BALANCING @@ -7073,7 +7377,10 @@ static inline void update_sg_lb_stats(struct lb_env *env, sgs->nr_preferred_running += rq->nr_preferred_running; #endif sgs->sum_weighted_load += weighted_cpuload(i); - if (idle_cpu(i)) + /* + * No need to call idle_cpu() if nr_running is not 0 + */ + if (!nr_running && idle_cpu(i)) sgs->idle_cpus++; if (cpu_overutilized(i)) { @@ -7279,12 +7586,17 @@ next_group: env->dst_rq->rd->overload = overload; /* Update over-utilization (tipping point, U >= 0) indicator */ - if (env->dst_rq->rd->overutilized != overutilized) + if (env->dst_rq->rd->overutilized != overutilized) { env->dst_rq->rd->overutilized = overutilized; + trace_sched_overutilized(overutilized); + } } else { - if (!env->dst_rq->rd->overutilized && overutilized) + if (!env->dst_rq->rd->overutilized && overutilized) { env->dst_rq->rd->overutilized = true; + trace_sched_overutilized(true); + } } + } /** @@ -8078,6 +8390,7 @@ static int idle_balance(struct rq *this_rq) struct sched_domain *sd; int pulled_task = 0; u64 curr_cost = 0; + long removed_util=0; idle_enter_fair(this_rq); @@ -8101,6 +8414,17 @@ static int idle_balance(struct rq *this_rq) raw_spin_unlock(&this_rq->lock); + /* + * If removed_util_avg is !0 we most probably migrated some task away + * from this_cpu. In this case we might be willing to trigger an OPP + * update, but we want to do so if we don't find anybody else to pull + * here (we will trigger an OPP update with the pulled task's enqueue + * anyway). + * + * Record removed_util before calling update_blocked_averages, and use + * it below (before returning) to see if an OPP update is required. + */ + removed_util = atomic_long_read(&(this_rq->cfs).removed_util_avg); update_blocked_averages(this_cpu); rcu_read_lock(); for_each_domain(this_cpu, sd) { @@ -8165,6 +8489,12 @@ out: if (pulled_task) { idle_exit_fair(this_rq); this_rq->idle_stamp = 0; + } else if (removed_util) { + /* + * No task pulled and someone has been migrated away. + * Good case to trigger an OPP update. + */ + update_capacity_of(this_cpu); } return pulled_task; @@ -8723,10 +9053,15 @@ static void task_tick_fair(struct rq *rq, struct task_struct *curr, int queued) if (static_branch_unlikely(&sched_numa_balancing)) task_tick_numa(rq, curr); - if (!rq->rd->overutilized && cpu_overutilized(task_cpu(curr))) +#ifdef CONFIG_SMP + if (!rq->rd->overutilized && cpu_overutilized(task_cpu(curr))) { rq->rd->overutilized = true; + trace_sched_overutilized(true); + } rq->misfit_task = !task_fits_max(curr, rq->cpu); +#endif + } /* diff --git a/kernel/sched/features.h b/kernel/sched/features.h index b634151ce286..55e461055332 100644 --- a/kernel/sched/features.h +++ b/kernel/sched/features.h @@ -73,4 +73,8 @@ SCHED_FEAT(ATTACH_AGE_LOAD, true) * Energy aware scheduling. Use platform energy model to guide scheduling * decisions optimizing for energy efficiency. */ +#ifdef CONFIG_DEFAULT_USE_ENERGY_AWARE +SCHED_FEAT(ENERGY_AWARE, true) +#else SCHED_FEAT(ENERGY_AWARE, false) +#endif diff --git a/kernel/sched/idle.c b/kernel/sched/idle.c index cbc130efbc5b..917c94abf5bb 100644 --- a/kernel/sched/idle.c +++ b/kernel/sched/idle.c @@ -220,6 +220,7 @@ static void cpu_idle_loop(void) */ __current_set_polling(); + quiet_vmstat(); tick_nohz_idle_enter(); while (!need_resched()) { diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c index 9694204660b7..8a16cba968c4 100644 --- a/kernel/sched/rt.c +++ b/kernel/sched/rt.c @@ -8,6 +8,8 @@ #include <linux/slab.h> #include <linux/irq_work.h> +#include "walt.h" + int sched_rr_timeslice = RR_TIMESLICE; static int do_sched_rt_period_timer(struct rt_bandwidth *rt_b, int overrun); @@ -889,6 +891,51 @@ static inline int rt_se_prio(struct sched_rt_entity *rt_se) return rt_task_of(rt_se)->prio; } +static void dump_throttled_rt_tasks(struct rt_rq *rt_rq) +{ + struct rt_prio_array *array = &rt_rq->active; + struct sched_rt_entity *rt_se; + char buf[500]; + char *pos = buf; + char *end = buf + sizeof(buf); + int idx; + + pos += snprintf(pos, sizeof(buf), + "sched: RT throttling activated for rt_rq %p (cpu %d)\n", + rt_rq, cpu_of(rq_of_rt_rq(rt_rq))); + + if (bitmap_empty(array->bitmap, MAX_RT_PRIO)) + goto out; + + pos += snprintf(pos, end - pos, "potential CPU hogs:\n"); + idx = sched_find_first_bit(array->bitmap); + while (idx < MAX_RT_PRIO) { + list_for_each_entry(rt_se, array->queue + idx, run_list) { + struct task_struct *p; + + if (!rt_entity_is_task(rt_se)) + continue; + + p = rt_task_of(rt_se); + if (pos < end) + pos += snprintf(pos, end - pos, "\t%s (%d)\n", + p->comm, p->pid); + } + idx = find_next_bit(array->bitmap, MAX_RT_PRIO, idx + 1); + } +out: +#ifdef CONFIG_PANIC_ON_RT_THROTTLING + /* + * Use pr_err() in the BUG() case since printk_sched() will + * not get flushed and deadlock is not a concern. + */ + pr_err("%s", buf); + BUG(); +#else + printk_deferred("%s", buf); +#endif +} + static int sched_rt_runtime_exceeded(struct rt_rq *rt_rq) { u64 runtime = sched_rt_runtime(rt_rq); @@ -912,8 +959,14 @@ static int sched_rt_runtime_exceeded(struct rt_rq *rt_rq) * but accrue some time due to boosting. */ if (likely(rt_b->rt_runtime)) { + static bool once = false; + rt_rq->rt_throttled = 1; - printk_deferred_once("sched: RT throttling activated\n"); + + if (!once) { + once = true; + dump_throttled_rt_tasks(rt_rq); + } } else { /* * In case we did anyway, make it go away, @@ -1261,6 +1314,7 @@ enqueue_task_rt(struct rq *rq, struct task_struct *p, int flags) rt_se->timeout = 0; enqueue_rt_entity(rt_se, flags & ENQUEUE_HEAD); + walt_inc_cumulative_runnable_avg(rq, p); if (!task_current(rq, p) && p->nr_cpus_allowed > 1) enqueue_pushable_task(rq, p); @@ -1272,6 +1326,7 @@ static void dequeue_task_rt(struct rq *rq, struct task_struct *p, int flags) update_curr_rt(rq); dequeue_rt_entity(rt_se); + walt_dec_cumulative_runnable_avg(rq, p); dequeue_pushable_task(rq, p); } diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index a537f1864dd0..2f2b959ad244 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -410,6 +410,10 @@ struct cfs_rq { struct list_head leaf_cfs_rq_list; struct task_group *tg; /* group that "owns" this runqueue */ +#ifdef CONFIG_SCHED_WALT + u64 cumulative_runnable_avg; +#endif + #ifdef CONFIG_CFS_BANDWIDTH int runtime_enabled; u64 runtime_expires; @@ -594,6 +598,14 @@ struct rq { #ifdef CONFIG_NO_HZ_FULL unsigned long last_sched_tick; #endif + +#ifdef CONFIG_CPU_QUIET + /* time-based average load */ + u64 nr_last_stamp; + u64 nr_running_integral; + seqcount_t ave_seqcnt; +#endif + /* capture load from *all* tasks on this cpu: */ struct load_weight load; unsigned long nr_load_updates; @@ -655,6 +667,30 @@ struct rq { u64 max_idle_balance_cost; #endif +#ifdef CONFIG_SCHED_WALT + /* + * max_freq = user or thermal defined maximum + * max_possible_freq = maximum supported by hardware + */ + unsigned int cur_freq, max_freq, min_freq, max_possible_freq; + struct cpumask freq_domain_cpumask; + + u64 cumulative_runnable_avg; + int efficiency; /* Differentiate cpus with different IPC capability */ + int load_scale_factor; + int capacity; + int max_possible_capacity; + u64 window_start; + u64 curr_runnable_sum; + u64 prev_runnable_sum; + u64 nt_curr_runnable_sum; + u64 nt_prev_runnable_sum; + u64 cur_irqload; + u64 avg_irqload; + u64 irqload_ts; +#endif /* CONFIG_SCHED_WALT */ + + #ifdef CONFIG_IRQ_TIME_ACCOUNTING u64 prev_irq_time; #endif @@ -1269,6 +1305,7 @@ extern const struct sched_class idle_sched_class; #ifdef CONFIG_SMP +extern void init_max_cpu_capacity(struct max_cpu_capacity *mcc); extern void update_group_capacity(struct sched_domain *sd, int cpu); extern void trigger_load_balance(struct rq *rq); @@ -1351,9 +1388,7 @@ unsigned long to_ratio(u64 period, u64 runtime); extern void init_entity_runnable_average(struct sched_entity *se); -extern void init_max_cpu_capacity(struct max_cpu_capacity *mcc); - -static inline void add_nr_running(struct rq *rq, unsigned count) +static inline void __add_nr_running(struct rq *rq, unsigned count) { unsigned prev_nr = rq->nr_running; @@ -1381,11 +1416,48 @@ static inline void add_nr_running(struct rq *rq, unsigned count) } } -static inline void sub_nr_running(struct rq *rq, unsigned count) +static inline void __sub_nr_running(struct rq *rq, unsigned count) { rq->nr_running -= count; } +#ifdef CONFIG_CPU_QUIET +#define NR_AVE_SCALE(x) ((x) << FSHIFT) +static inline u64 do_nr_running_integral(struct rq *rq) +{ + s64 nr, deltax; + u64 nr_running_integral = rq->nr_running_integral; + + deltax = rq->clock_task - rq->nr_last_stamp; + nr = NR_AVE_SCALE(rq->nr_running); + + nr_running_integral += nr * deltax; + + return nr_running_integral; +} + +static inline void add_nr_running(struct rq *rq, unsigned count) +{ + write_seqcount_begin(&rq->ave_seqcnt); + rq->nr_running_integral = do_nr_running_integral(rq); + rq->nr_last_stamp = rq->clock_task; + __add_nr_running(rq, count); + write_seqcount_end(&rq->ave_seqcnt); +} + +static inline void sub_nr_running(struct rq *rq, unsigned count) +{ + write_seqcount_begin(&rq->ave_seqcnt); + rq->nr_running_integral = do_nr_running_integral(rq); + rq->nr_last_stamp = rq->clock_task; + __sub_nr_running(rq, count); + write_seqcount_end(&rq->ave_seqcnt); +} +#else +#define add_nr_running __add_nr_running +#define sub_nr_running __sub_nr_running +#endif + static inline void rq_last_tick_reset(struct rq *rq) { #ifdef CONFIG_NO_HZ_FULL @@ -1469,6 +1541,10 @@ static inline unsigned long capacity_orig_of(int cpu) return cpu_rq(cpu)->cpu_capacity_orig; } +extern unsigned int sysctl_sched_use_walt_cpu_util; +extern unsigned int walt_ravg_window; +extern unsigned int walt_disabled; + /* * cpu_util returns the amount of capacity of a CPU that is used by CFS * tasks. The unit of the return value must be the one of capacity so we can @@ -1500,6 +1576,12 @@ static inline unsigned long __cpu_util(int cpu, int delta) unsigned long util = cpu_rq(cpu)->cfs.avg.util_avg; unsigned long capacity = capacity_orig_of(cpu); +#ifdef CONFIG_SCHED_WALT + if (!walt_disabled && sysctl_sched_use_walt_cpu_util) { + util = cpu_rq(cpu)->prev_runnable_sum << SCHED_LOAD_SHIFT; + do_div(util, walt_ravg_window); + } +#endif delta += util; if (delta < 0) return 0; @@ -1530,8 +1612,27 @@ void update_cpu_capacity_request(int cpu, bool request); static inline void set_cfs_cpu_capacity(int cpu, bool request, unsigned long capacity) { - if (per_cpu(cpu_sched_capacity_reqs, cpu).cfs != capacity) { - per_cpu(cpu_sched_capacity_reqs, cpu).cfs = capacity; + struct sched_capacity_reqs *scr = &per_cpu(cpu_sched_capacity_reqs, cpu); + +#ifdef CONFIG_SCHED_WALT + if (!walt_disabled && sysctl_sched_use_walt_cpu_util) { + int rtdl = scr->rt + scr->dl; + /* + * WALT tracks the utilization of a CPU considering the load + * generated by all the scheduling classes. + * Since the following call to: + * update_cpu_capacity + * is already adding the RT and DL utilizations let's remove + * these contributions from the WALT signal. + */ + if (capacity > rtdl) + capacity -= rtdl; + else + capacity = 0; + } +#endif + if (scr->cfs != capacity) { + scr->cfs = capacity; update_cpu_capacity_request(cpu, request); } } @@ -1657,6 +1758,9 @@ task_rq_unlock(struct rq *rq, struct task_struct *p, unsigned long *flags) raw_spin_unlock_irqrestore(&p->pi_lock, *flags); } +extern struct rq *lock_rq_of(struct task_struct *p, unsigned long *flags); +extern void unlock_rq_of(struct rq *rq, struct task_struct *p, unsigned long *flags); + #ifdef CONFIG_SMP #ifdef CONFIG_PREEMPT @@ -1729,7 +1833,8 @@ static inline int double_lock_balance(struct rq *this_rq, struct rq *busiest) static inline void double_unlock_balance(struct rq *this_rq, struct rq *busiest) __releases(busiest->lock) { - raw_spin_unlock(&busiest->lock); + if (this_rq != busiest) + raw_spin_unlock(&busiest->lock); lock_set_subclass(&this_rq->lock.dep_map, 0, _RET_IP_); } diff --git a/kernel/sched/stop_task.c b/kernel/sched/stop_task.c index cbc67da10954..61f852d46858 100644 --- a/kernel/sched/stop_task.c +++ b/kernel/sched/stop_task.c @@ -1,4 +1,5 @@ #include "sched.h" +#include "walt.h" /* * stop-task scheduling class. @@ -42,12 +43,14 @@ static void enqueue_task_stop(struct rq *rq, struct task_struct *p, int flags) { add_nr_running(rq, 1); + walt_inc_cumulative_runnable_avg(rq, p); } static void dequeue_task_stop(struct rq *rq, struct task_struct *p, int flags) { sub_nr_running(rq, 1); + walt_dec_cumulative_runnable_avg(rq, p); } static void yield_task_stop(struct rq *rq) diff --git a/kernel/sched/tune.c b/kernel/sched/tune.c index 9d8eeb438140..68a24a044b0a 100644 --- a/kernel/sched/tune.c +++ b/kernel/sched/tune.c @@ -3,24 +3,21 @@ #include <linux/kernel.h> #include <linux/percpu.h> #include <linux/printk.h> -#include <linux/reciprocal_div.h> #include <linux/rcupdate.h> #include <linux/slab.h> #include <trace/events/sched.h> #include "sched.h" +#include "tune.h" + +#ifdef CONFIG_CGROUP_SCHEDTUNE +static bool schedtune_initialized = false; +#endif unsigned int sysctl_sched_cfs_boost __read_mostly; -/* - * System energy normalization constants - */ -static struct target_nrg { - unsigned long min_power; - unsigned long max_power; - struct reciprocal_value rdiv; -} schedtune_target_nrg; +extern struct target_nrg schedtune_target_nrg; /* Performance Boost region (B) threshold params */ static int perf_boost_idx; @@ -41,16 +38,16 @@ struct threshold_params { */ static struct threshold_params threshold_gains[] = { - { 0, 4 }, /* >= 0% */ - { 0, 4 }, /* >= 10% */ - { 1, 4 }, /* >= 20% */ - { 2, 4 }, /* >= 30% */ - { 3, 4 }, /* >= 40% */ - { 4, 3 }, /* >= 50% */ - { 4, 2 }, /* >= 60% */ - { 4, 1 }, /* >= 70% */ - { 4, 0 }, /* >= 80% */ - { 4, 0 } /* >= 90% */ + { 0, 5 }, /* < 10% */ + { 1, 5 }, /* < 20% */ + { 2, 5 }, /* < 30% */ + { 3, 5 }, /* < 40% */ + { 4, 5 }, /* < 50% */ + { 5, 4 }, /* < 60% */ + { 5, 3 }, /* < 70% */ + { 5, 2 }, /* < 80% */ + { 5, 1 }, /* < 90% */ + { 5, 0 } /* <= 100% */ }; static int @@ -58,36 +55,51 @@ __schedtune_accept_deltas(int nrg_delta, int cap_delta, int perf_boost_idx, int perf_constrain_idx) { int payoff = -INT_MAX; + int gain_idx = -1; /* Performance Boost (B) region */ - if (nrg_delta > 0 && cap_delta > 0) { - /* - * Evaluate "Performance Boost" vs "Energy Increase" - * payoff criteria: - * cap_delta / nrg_delta < cap_gain / nrg_gain - * which is: - * nrg_delta * cap_gain > cap_delta * nrg_gain - */ - payoff = nrg_delta * threshold_gains[perf_boost_idx].cap_gain; - payoff -= cap_delta * threshold_gains[perf_boost_idx].nrg_gain; - return payoff; - } - + if (nrg_delta >= 0 && cap_delta > 0) + gain_idx = perf_boost_idx; /* Performance Constraint (C) region */ - if (nrg_delta < 0 && cap_delta < 0) { - /* - * Evaluate "Performance Boost" vs "Energy Increase" - * payoff criteria: - * cap_delta / nrg_delta > cap_gain / nrg_gain - * which is: - * cap_delta * nrg_gain > nrg_delta * cap_gain - */ - payoff = cap_delta * threshold_gains[perf_constrain_idx].nrg_gain; - payoff -= nrg_delta * threshold_gains[perf_constrain_idx].cap_gain; - return payoff; - } + else if (nrg_delta < 0 && cap_delta <= 0) + gain_idx = perf_constrain_idx; /* Default: reject schedule candidate */ + if (gain_idx == -1) + return payoff; + + /* + * Evaluate "Performance Boost" vs "Energy Increase" + * + * - Performance Boost (B) region + * + * Condition: nrg_delta > 0 && cap_delta > 0 + * Payoff criteria: + * cap_gain / nrg_gain < cap_delta / nrg_delta = + * cap_gain * nrg_delta < cap_delta * nrg_gain + * Note that since both nrg_gain and nrg_delta are positive, the + * inequality does not change. Thus: + * + * payoff = (cap_delta * nrg_gain) - (cap_gain * nrg_delta) + * + * - Performance Constraint (C) region + * + * Condition: nrg_delta < 0 && cap_delta < 0 + * payoff criteria: + * cap_gain / nrg_gain > cap_delta / nrg_delta = + * cap_gain * nrg_delta < cap_delta * nrg_gain + * Note that since nrg_gain > 0 while nrg_delta < 0, the + * inequality change. Thus: + * + * payoff = (cap_delta * nrg_gain) - (cap_gain * nrg_delta) + * + * This means that, in case of same positive defined {cap,nrg}_gain + * for both the B and C regions, we can use the same payoff formula + * where a positive value represents the accept condition. + */ + payoff = cap_delta * threshold_gains[gain_idx].nrg_gain; + payoff -= nrg_delta * threshold_gains[gain_idx].cap_gain; + return payoff; } @@ -113,6 +125,10 @@ struct schedtune { /* Performance Constraint (C) region threshold params */ int perf_constrain_idx; + + /* Hint to bias scheduling of tasks on that SchedTune CGroup + * towards idle CPUs */ + int prefer_idle; }; static inline struct schedtune *css_st(struct cgroup_subsys_state *css) @@ -144,6 +160,7 @@ root_schedtune = { .boost = 0, .perf_boost_idx = 0, .perf_constrain_idx = 0, + .prefer_idle = 0, }; int @@ -155,12 +172,16 @@ schedtune_accept_deltas(int nrg_delta, int cap_delta, int perf_constrain_idx; /* Optimal (O) region */ - if (nrg_delta < 0 && cap_delta > 0) + if (nrg_delta < 0 && cap_delta > 0) { + trace_sched_tune_filter(nrg_delta, cap_delta, 0, 0, 1, 0); return INT_MAX; + } /* Suboptimal (S) region */ - if (nrg_delta > 0 && cap_delta < 0) + if (nrg_delta > 0 && cap_delta < 0) { + trace_sched_tune_filter(nrg_delta, cap_delta, 0, 0, -1, 5); return -INT_MAX; + } /* Get task specific perf Boost/Constraints indexes */ rcu_read_lock(); @@ -202,13 +223,16 @@ static struct schedtune *allocated_group[BOOSTGROUPS_COUNT] = { */ struct boost_groups { /* Maximum boost value for all RUNNABLE tasks on a CPU */ - unsigned boost_max; + bool idle; + int boost_max; struct { /* The boost for tasks on that boost group */ - unsigned boost; + int boost; /* Count of RUNNABLE tasks on that boost group */ unsigned tasks; } group[BOOSTGROUPS_COUNT]; + /* CPU's boost group locking */ + raw_spinlock_t lock; }; /* Boost groups affecting each CPU in the system */ @@ -218,7 +242,7 @@ static void schedtune_cpu_update(int cpu) { struct boost_groups *bg; - unsigned boost_max; + int boost_max; int idx; bg = &per_cpu(cpu_boost_groups, cpu); @@ -232,9 +256,13 @@ schedtune_cpu_update(int cpu) */ if (bg->group[idx].tasks == 0) continue; + boost_max = max(boost_max, bg->group[idx].boost); } - + /* Ensures boost_max is non-negative when all cgroup boost values + * are neagtive. Avoids under-accounting of cpu capacity which may cause + * task stacking and frequency spikes.*/ + boost_max = max(boost_max, 0); bg->boost_max = boost_max; } @@ -281,28 +309,24 @@ schedtune_boostgroup_update(int idx, int boost) return 0; } +#define ENQUEUE_TASK 1 +#define DEQUEUE_TASK -1 + static inline void schedtune_tasks_update(struct task_struct *p, int cpu, int idx, int task_count) { - struct boost_groups *bg; - int tasks; - - bg = &per_cpu(cpu_boost_groups, cpu); + struct boost_groups *bg = &per_cpu(cpu_boost_groups, cpu); + int tasks = bg->group[idx].tasks + task_count; /* Update boosted tasks count while avoiding to make it negative */ - if (task_count < 0 && bg->group[idx].tasks <= -task_count) - bg->group[idx].tasks = 0; - else - bg->group[idx].tasks += task_count; - - /* Boost group activation or deactivation on that RQ */ - tasks = bg->group[idx].tasks; - if (tasks == 1 || tasks == 0) - schedtune_cpu_update(cpu); + bg->group[idx].tasks = max(0, tasks); trace_sched_tune_tasks_update(p, cpu, tasks, idx, bg->group[idx].boost, bg->boost_max); + /* Boost group activation or deactivation on that RQ */ + if (tasks == 1 || tasks == 0) + schedtune_cpu_update(cpu); } /* @@ -310,9 +334,14 @@ schedtune_tasks_update(struct task_struct *p, int cpu, int idx, int task_count) */ void schedtune_enqueue_task(struct task_struct *p, int cpu) { + struct boost_groups *bg = &per_cpu(cpu_boost_groups, cpu); + unsigned long irq_flags; struct schedtune *st; int idx; + if (!unlikely(schedtune_initialized)) + return; + /* * When a task is marked PF_EXITING by do_exit() it's going to be * dequeued and enqueued multiple times in the exit path. @@ -322,13 +351,103 @@ void schedtune_enqueue_task(struct task_struct *p, int cpu) if (p->flags & PF_EXITING) return; - /* Get task boost group */ + /* + * Boost group accouting is protected by a per-cpu lock and requires + * interrupt to be disabled to avoid race conditions for example on + * do_exit()::cgroup_exit() and task migration. + */ + raw_spin_lock_irqsave(&bg->lock, irq_flags); rcu_read_lock(); + st = task_schedtune(p); idx = st->idx; + + schedtune_tasks_update(p, cpu, idx, ENQUEUE_TASK); + rcu_read_unlock(); + raw_spin_unlock_irqrestore(&bg->lock, irq_flags); +} + +int schedtune_can_attach(struct cgroup_taskset *tset) +{ + struct task_struct *task; + struct cgroup_subsys_state *css; + struct boost_groups *bg; + unsigned long irq_flags; + unsigned int cpu; + struct rq *rq; + int src_bg; /* Source boost group index */ + int dst_bg; /* Destination boost group index */ + int tasks; + + if (!unlikely(schedtune_initialized)) + return 0; + + + cgroup_taskset_for_each(task, css, tset) { + + /* + * Lock the CPU's RQ the task is enqueued to avoid race + * conditions with migration code while the task is being + * accounted + */ + rq = lock_rq_of(task, &irq_flags); + + if (!task->on_rq) { + unlock_rq_of(rq, task, &irq_flags); + continue; + } + + /* + * Boost group accouting is protected by a per-cpu lock and requires + * interrupt to be disabled to avoid race conditions on... + */ + cpu = cpu_of(rq); + bg = &per_cpu(cpu_boost_groups, cpu); + raw_spin_lock(&bg->lock); + + dst_bg = css_st(css)->idx; + src_bg = task_schedtune(task)->idx; + + /* + * Current task is not changing boostgroup, which can + * happen when the new hierarchy is in use. + */ + if (unlikely(dst_bg == src_bg)) { + raw_spin_unlock(&bg->lock); + unlock_rq_of(rq, task, &irq_flags); + continue; + } + + /* + * This is the case of a RUNNABLE task which is switching its + * current boost group. + */ + + /* Move task from src to dst boost group */ + tasks = bg->group[src_bg].tasks - 1; + bg->group[src_bg].tasks = max(0, tasks); + bg->group[dst_bg].tasks += 1; + + raw_spin_unlock(&bg->lock); + unlock_rq_of(rq, task, &irq_flags); + + /* Update CPU boost group */ + if (bg->group[src_bg].tasks == 0 || bg->group[dst_bg].tasks == 1) + schedtune_cpu_update(task_cpu(task)); + + } - schedtune_tasks_update(p, cpu, idx, 1); + return 0; +} + +void schedtune_cancel_attach(struct cgroup_taskset *tset) +{ + /* This can happen only if SchedTune controller is mounted with + * other hierarchies ane one of them fails. Since usually SchedTune is + * mouted on its own hierarcy, for the time being we do not implement + * a proper rollback mechanism */ + WARN(1, "SchedTune cancel attach not implemented"); } /* @@ -336,26 +455,62 @@ void schedtune_enqueue_task(struct task_struct *p, int cpu) */ void schedtune_dequeue_task(struct task_struct *p, int cpu) { + struct boost_groups *bg = &per_cpu(cpu_boost_groups, cpu); + unsigned long irq_flags; struct schedtune *st; int idx; + if (!unlikely(schedtune_initialized)) + return; + /* * When a task is marked PF_EXITING by do_exit() it's going to be * dequeued and enqueued multiple times in the exit path. * Thus we avoid any further update, since we do not want to change * CPU boosting while the task is exiting. - * The last dequeue will be done by cgroup exit() callback. + * The last dequeue is already enforce by the do_exit() code path + * via schedtune_exit_task(). */ if (p->flags & PF_EXITING) return; - /* Get task boost group */ + /* + * Boost group accouting is protected by a per-cpu lock and requires + * interrupt to be disabled to avoid race conditions on... + */ + raw_spin_lock_irqsave(&bg->lock, irq_flags); rcu_read_lock(); + st = task_schedtune(p); idx = st->idx; + + schedtune_tasks_update(p, cpu, idx, DEQUEUE_TASK); + rcu_read_unlock(); + raw_spin_unlock_irqrestore(&bg->lock, irq_flags); +} + +void schedtune_exit_task(struct task_struct *tsk) +{ + struct schedtune *st; + unsigned long irq_flags; + unsigned int cpu; + struct rq *rq; + int idx; + + if (!unlikely(schedtune_initialized)) + return; + + rq = lock_rq_of(tsk, &irq_flags); + rcu_read_lock(); - schedtune_tasks_update(p, cpu, idx, -1); + cpu = cpu_of(rq); + st = task_schedtune(tsk); + idx = st->idx; + schedtune_tasks_update(tsk, cpu, idx, DEQUEUE_TASK); + + rcu_read_unlock(); + unlock_rq_of(rq, tsk, &irq_flags); } int schedtune_cpu_boost(int cpu) @@ -380,7 +535,39 @@ int schedtune_task_boost(struct task_struct *p) return task_boost; } +int schedtune_prefer_idle(struct task_struct *p) +{ + struct schedtune *st; + int prefer_idle; + + /* Get prefer_idle value */ + rcu_read_lock(); + st = task_schedtune(p); + prefer_idle = st->prefer_idle; + rcu_read_unlock(); + + return prefer_idle; +} + static u64 +prefer_idle_read(struct cgroup_subsys_state *css, struct cftype *cft) +{ + struct schedtune *st = css_st(css); + + return st->prefer_idle; +} + +static int +prefer_idle_write(struct cgroup_subsys_state *css, struct cftype *cft, + u64 prefer_idle) +{ + struct schedtune *st = css_st(css); + st->prefer_idle = prefer_idle; + + return 0; +} + +static s64 boost_read(struct cgroup_subsys_state *css, struct cftype *cft) { struct schedtune *st = css_st(css); @@ -390,16 +577,32 @@ boost_read(struct cgroup_subsys_state *css, struct cftype *cft) static int boost_write(struct cgroup_subsys_state *css, struct cftype *cft, - u64 boost) + s64 boost) { struct schedtune *st = css_st(css); + unsigned threshold_idx; + int boost_pct; - if (boost < 0 || boost > 100) + if (boost < -100 || boost > 100) return -EINVAL; + boost_pct = boost; + + /* + * Update threshold params for Performance Boost (B) + * and Performance Constraint (C) regions. + * The current implementatio uses the same cuts for both + * B and C regions. + */ + threshold_idx = clamp(boost_pct, 0, 99) / 10; + st->perf_boost_idx = threshold_idx; + st->perf_constrain_idx = threshold_idx; st->boost = boost; - if (css == &root_schedtune.css) + if (css == &root_schedtune.css) { sysctl_sched_cfs_boost = boost; + perf_boost_idx = threshold_idx; + perf_constrain_idx = threshold_idx; + } /* Update CPU boost */ schedtune_boostgroup_update(st->idx, st->boost); @@ -412,8 +615,13 @@ boost_write(struct cgroup_subsys_state *css, struct cftype *cft, static struct cftype files[] = { { .name = "boost", - .read_u64 = boost_read, - .write_u64 = boost_write, + .read_s64 = boost_read, + .write_s64 = boost_write, + }, + { + .name = "prefer_idle", + .read_u64 = prefer_idle_read, + .write_u64 = prefer_idle_write, }, { } /* terminate */ }; @@ -437,33 +645,14 @@ schedtune_boostgroup_init(struct schedtune *st) return 0; } -static int -schedtune_init(void) -{ - struct boost_groups *bg; - int cpu; - - /* Initialize the per CPU boost groups */ - for_each_possible_cpu(cpu) { - bg = &per_cpu(cpu_boost_groups, cpu); - memset(bg, 0, sizeof(struct boost_groups)); - } - - pr_info(" schedtune configured to support %d boost groups\n", - BOOSTGROUPS_COUNT); - return 0; -} - static struct cgroup_subsys_state * schedtune_css_alloc(struct cgroup_subsys_state *parent_css) { struct schedtune *st; int idx; - if (!parent_css) { - schedtune_init(); + if (!parent_css) return &root_schedtune.css; - } /* Allow only single level hierachies */ if (parent_css != &root_schedtune.css) { @@ -520,10 +709,30 @@ schedtune_css_free(struct cgroup_subsys_state *css) struct cgroup_subsys schedtune_cgrp_subsys = { .css_alloc = schedtune_css_alloc, .css_free = schedtune_css_free, + .can_attach = schedtune_can_attach, + .cancel_attach = schedtune_cancel_attach, .legacy_cftypes = files, .early_init = 1, }; +static inline void +schedtune_init_cgroups(void) +{ + struct boost_groups *bg; + int cpu; + + /* Initialize the per CPU boost groups */ + for_each_possible_cpu(cpu) { + bg = &per_cpu(cpu_boost_groups, cpu); + memset(bg, 0, sizeof(struct boost_groups)); + } + + pr_info("schedtune: configured to support %d boost groups\n", + BOOSTGROUPS_COUNT); + + schedtune_initialized = true; +} + #else /* CONFIG_CGROUP_SCHEDTUNE */ int @@ -531,12 +740,16 @@ schedtune_accept_deltas(int nrg_delta, int cap_delta, struct task_struct *task) { /* Optimal (O) region */ - if (nrg_delta < 0 && cap_delta > 0) + if (nrg_delta < 0 && cap_delta > 0) { + trace_sched_tune_filter(nrg_delta, cap_delta, 0, 0, 1, 0); return INT_MAX; + } /* Suboptimal (S) region */ - if (nrg_delta > 0 && cap_delta < 0) + if (nrg_delta > 0 && cap_delta < 0) { + trace_sched_tune_filter(nrg_delta, cap_delta, 0, 0, -1, 5); return -INT_MAX; + } return __schedtune_accept_deltas(nrg_delta, cap_delta, perf_boost_idx, perf_constrain_idx); @@ -550,52 +763,29 @@ sysctl_sched_cfs_boost_handler(struct ctl_table *table, int write, loff_t *ppos) { int ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos); + unsigned threshold_idx; + int boost_pct; if (ret || !write) return ret; - /* Performance Boost (B) region threshold params */ - perf_boost_idx = sysctl_sched_cfs_boost; - perf_boost_idx /= 10; + if (sysctl_sched_cfs_boost < -100 || sysctl_sched_cfs_boost > 100) + return -EINVAL; + boost_pct = sysctl_sched_cfs_boost; - /* Performance Constraint (C) region threshold params */ - perf_constrain_idx = 100 - sysctl_sched_cfs_boost; - perf_constrain_idx /= 10; + /* + * Update threshold params for Performance Boost (B) + * and Performance Constraint (C) regions. + * The current implementatio uses the same cuts for both + * B and C regions. + */ + threshold_idx = clamp(boost_pct, 0, 99) / 10; + perf_boost_idx = threshold_idx; + perf_constrain_idx = threshold_idx; return 0; } -/* - * System energy normalization - * Returns the normalized value, in the range [0..SCHED_LOAD_SCALE], - * corresponding to the specified energy variation. - */ -int -schedtune_normalize_energy(int energy_diff) -{ - u32 normalized_nrg; - int max_delta; - -#ifdef CONFIG_SCHED_DEBUG - /* Check for boundaries */ - max_delta = schedtune_target_nrg.max_power; - max_delta -= schedtune_target_nrg.min_power; - WARN_ON(abs(energy_diff) >= max_delta); -#endif - - /* Do scaling using positive numbers to increase the range */ - normalized_nrg = (energy_diff < 0) ? -energy_diff : energy_diff; - - /* Scale by energy magnitude */ - normalized_nrg <<= SCHED_LOAD_SHIFT; - - /* Normalize on max energy for target platform */ - normalized_nrg = reciprocal_divide( - normalized_nrg, schedtune_target_nrg.rdiv); - - return (energy_diff < 0) ? -normalized_nrg : normalized_nrg; -} - #ifdef CONFIG_SCHED_DEBUG static void schedtune_test_nrg(unsigned long delta_pwr) @@ -698,7 +888,7 @@ schedtune_add_cluster_nrg( * that bind the EM to the topology information. */ static int -schedtune_init_late(void) +schedtune_init(void) { struct target_nrg *ste = &schedtune_target_nrg; unsigned long delta_pwr = 0; @@ -738,11 +928,17 @@ schedtune_init_late(void) ste->rdiv.m, ste->rdiv.sh1, ste->rdiv.sh2); schedtune_test_nrg(delta_pwr); + +#ifdef CONFIG_CGROUP_SCHEDTUNE + schedtune_init_cgroups(); +#else + pr_info("schedtune: configured to support global boosting only\n"); +#endif + return 0; nodata: rcu_read_unlock(); return -EINVAL; } -late_initcall(schedtune_init_late); - +postcore_initcall(schedtune_init); diff --git a/kernel/sched/tune.h b/kernel/sched/tune.h index f7273a5d994a..4f6441771e4c 100644 --- a/kernel/sched/tune.h +++ b/kernel/sched/tune.h @@ -1,16 +1,36 @@ #ifdef CONFIG_SCHED_TUNE +#include <linux/reciprocal_div.h> + +/* + * System energy normalization constants + */ +struct target_nrg { + unsigned long min_power; + unsigned long max_power; + struct reciprocal_value rdiv; +}; + #ifdef CONFIG_CGROUP_SCHEDTUNE int schedtune_cpu_boost(int cpu); int schedtune_task_boost(struct task_struct *tsk); +int schedtune_prefer_idle(struct task_struct *tsk); + +void schedtune_exit_task(struct task_struct *tsk); + void schedtune_enqueue_task(struct task_struct *p, int cpu); void schedtune_dequeue_task(struct task_struct *p, int cpu); #else /* CONFIG_CGROUP_SCHEDTUNE */ +#define schedtune_cpu_boost(cpu) get_sysctl_sched_cfs_boost() +#define schedtune_task_boost(tsk) get_sysctl_sched_cfs_boost() + +#define schedtune_exit_task(task) do { } while (0) + #define schedtune_enqueue_task(task, cpu) do { } while (0) #define schedtune_dequeue_task(task, cpu) do { } while (0) @@ -22,10 +42,14 @@ int schedtune_accept_deltas(int nrg_delta, int cap_delta, #else /* CONFIG_SCHED_TUNE */ +#define schedtune_cpu_boost(cpu) 0 +#define schedtune_task_boost(tsk) 0 + +#define schedtune_exit_task(task) do { } while (0) + #define schedtune_enqueue_task(task, cpu) do { } while (0) #define schedtune_dequeue_task(task, cpu) do { } while (0) -#define schedtune_normalize_energy(energy) energy #define schedtune_accept_deltas(nrg_delta, cap_delta, task) nrg_delta #endif /* CONFIG_SCHED_TUNE */ diff --git a/kernel/sched/walt.c b/kernel/sched/walt.c new file mode 100644 index 000000000000..2ffb1680b380 --- /dev/null +++ b/kernel/sched/walt.c @@ -0,0 +1,1168 @@ +/* + * Copyright (c) 2016, The Linux Foundation. All rights reserved. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 and + * only version 2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * + * Window Assisted Load Tracking (WALT) implementation credits: + * Srivatsa Vaddagiri, Steve Muckle, Syed Rameez Mustafa, Joonwoo Park, + * Pavan Kumar Kondeti, Olav Haugan + * + * 2016-03-06: Integration with EAS/refactoring by Vikram Mulukutla + * and Todd Kjos + */ + +#include <linux/syscore_ops.h> +#include <linux/cpufreq.h> +#include <trace/events/sched.h> +#include "sched.h" +#include "walt.h" + +#define WINDOW_STATS_RECENT 0 +#define WINDOW_STATS_MAX 1 +#define WINDOW_STATS_MAX_RECENT_AVG 2 +#define WINDOW_STATS_AVG 3 +#define WINDOW_STATS_INVALID_POLICY 4 + +#define EXITING_TASK_MARKER 0xdeaddead + +static __read_mostly unsigned int walt_ravg_hist_size = 5; +static __read_mostly unsigned int walt_window_stats_policy = + WINDOW_STATS_MAX_RECENT_AVG; +static __read_mostly unsigned int walt_account_wait_time = 1; +static __read_mostly unsigned int walt_freq_account_wait_time = 0; +static __read_mostly unsigned int walt_io_is_busy = 0; + +unsigned int sysctl_sched_walt_init_task_load_pct = 15; + +/* 1 -> use PELT based load stats, 0 -> use window-based load stats */ +unsigned int __read_mostly walt_disabled = 0; + +static unsigned int max_possible_efficiency = 1024; +static unsigned int min_possible_efficiency = 1024; + +/* + * Maximum possible frequency across all cpus. Task demand and cpu + * capacity (cpu_power) metrics are scaled in reference to it. + */ +static unsigned int max_possible_freq = 1; + +/* + * Minimum possible max_freq across all cpus. This will be same as + * max_possible_freq on homogeneous systems and could be different from + * max_possible_freq on heterogenous systems. min_max_freq is used to derive + * capacity (cpu_power) of cpus. + */ +static unsigned int min_max_freq = 1; + +static unsigned int max_capacity = 1024; +static unsigned int min_capacity = 1024; +static unsigned int max_load_scale_factor = 1024; +static unsigned int max_possible_capacity = 1024; + +/* Mask of all CPUs that have max_possible_capacity */ +static cpumask_t mpc_mask = CPU_MASK_ALL; + +/* Window size (in ns) */ +__read_mostly unsigned int walt_ravg_window = 20000000; + +/* Min window size (in ns) = 10ms */ +#define MIN_SCHED_RAVG_WINDOW 10000000 + +/* Max window size (in ns) = 1s */ +#define MAX_SCHED_RAVG_WINDOW 1000000000 + +static unsigned int sync_cpu; +static ktime_t ktime_last; +static bool walt_ktime_suspended; + +static unsigned int task_load(struct task_struct *p) +{ + return p->ravg.demand; +} + +void +walt_inc_cumulative_runnable_avg(struct rq *rq, + struct task_struct *p) +{ + rq->cumulative_runnable_avg += p->ravg.demand; +} + +void +walt_dec_cumulative_runnable_avg(struct rq *rq, + struct task_struct *p) +{ + rq->cumulative_runnable_avg -= p->ravg.demand; + BUG_ON((s64)rq->cumulative_runnable_avg < 0); +} + +static void +fixup_cumulative_runnable_avg(struct rq *rq, + struct task_struct *p, s64 task_load_delta) +{ + rq->cumulative_runnable_avg += task_load_delta; + if ((s64)rq->cumulative_runnable_avg < 0) + panic("cra less than zero: tld: %lld, task_load(p) = %u\n", + task_load_delta, task_load(p)); +} + +u64 walt_ktime_clock(void) +{ + if (unlikely(walt_ktime_suspended)) + return ktime_to_ns(ktime_last); + return ktime_get_ns(); +} + +static void walt_resume(void) +{ + walt_ktime_suspended = false; +} + +static int walt_suspend(void) +{ + ktime_last = ktime_get(); + walt_ktime_suspended = true; + return 0; +} + +static struct syscore_ops walt_syscore_ops = { + .resume = walt_resume, + .suspend = walt_suspend +}; + +static int __init walt_init_ops(void) +{ + register_syscore_ops(&walt_syscore_ops); + return 0; +} +late_initcall(walt_init_ops); + +void walt_inc_cfs_cumulative_runnable_avg(struct cfs_rq *cfs_rq, + struct task_struct *p) +{ + cfs_rq->cumulative_runnable_avg += p->ravg.demand; +} + +void walt_dec_cfs_cumulative_runnable_avg(struct cfs_rq *cfs_rq, + struct task_struct *p) +{ + cfs_rq->cumulative_runnable_avg -= p->ravg.demand; +} + +static int exiting_task(struct task_struct *p) +{ + if (p->flags & PF_EXITING) { + if (p->ravg.sum_history[0] != EXITING_TASK_MARKER) { + p->ravg.sum_history[0] = EXITING_TASK_MARKER; + } + return 1; + } + return 0; +} + +static int __init set_walt_ravg_window(char *str) +{ + get_option(&str, &walt_ravg_window); + + walt_disabled = (walt_ravg_window < MIN_SCHED_RAVG_WINDOW || + walt_ravg_window > MAX_SCHED_RAVG_WINDOW); + return 0; +} + +early_param("walt_ravg_window", set_walt_ravg_window); + +static void +update_window_start(struct rq *rq, u64 wallclock) +{ + s64 delta; + int nr_windows; + + delta = wallclock - rq->window_start; + /* If the MPM global timer is cleared, set delta as 0 to avoid kernel BUG happening */ + if (delta < 0) { + delta = 0; + WARN_ONCE(1, "WALT wallclock appears to have gone backwards or reset\n"); + } + + if (delta < walt_ravg_window) + return; + + nr_windows = div64_u64(delta, walt_ravg_window); + rq->window_start += (u64)nr_windows * (u64)walt_ravg_window; +} + +static u64 scale_exec_time(u64 delta, struct rq *rq) +{ + unsigned int cur_freq = rq->cur_freq; + int sf; + + if (unlikely(cur_freq > max_possible_freq)) + cur_freq = rq->max_possible_freq; + + /* round up div64 */ + delta = div64_u64(delta * cur_freq + max_possible_freq - 1, + max_possible_freq); + + sf = DIV_ROUND_UP(rq->efficiency * 1024, max_possible_efficiency); + + delta *= sf; + delta >>= 10; + + return delta; +} + +static int cpu_is_waiting_on_io(struct rq *rq) +{ + if (!walt_io_is_busy) + return 0; + + return atomic_read(&rq->nr_iowait); +} + +void walt_account_irqtime(int cpu, struct task_struct *curr, + u64 delta, u64 wallclock) +{ + struct rq *rq = cpu_rq(cpu); + unsigned long flags, nr_windows; + u64 cur_jiffies_ts; + + raw_spin_lock_irqsave(&rq->lock, flags); + + /* + * cputime (wallclock) uses sched_clock so use the same here for + * consistency. + */ + delta += sched_clock() - wallclock; + cur_jiffies_ts = get_jiffies_64(); + + if (is_idle_task(curr)) + walt_update_task_ravg(curr, rq, IRQ_UPDATE, walt_ktime_clock(), + delta); + + nr_windows = cur_jiffies_ts - rq->irqload_ts; + + if (nr_windows) { + if (nr_windows < 10) { + /* Decay CPU's irqload by 3/4 for each window. */ + rq->avg_irqload *= (3 * nr_windows); + rq->avg_irqload = div64_u64(rq->avg_irqload, + 4 * nr_windows); + } else { + rq->avg_irqload = 0; + } + rq->avg_irqload += rq->cur_irqload; + rq->cur_irqload = 0; + } + + rq->cur_irqload += delta; + rq->irqload_ts = cur_jiffies_ts; + raw_spin_unlock_irqrestore(&rq->lock, flags); +} + + +#define WALT_HIGH_IRQ_TIMEOUT 3 + +u64 walt_irqload(int cpu) { + struct rq *rq = cpu_rq(cpu); + s64 delta; + delta = get_jiffies_64() - rq->irqload_ts; + + /* + * Current context can be preempted by irq and rq->irqload_ts can be + * updated by irq context so that delta can be negative. + * But this is okay and we can safely return as this means there + * was recent irq occurrence. + */ + + if (delta < WALT_HIGH_IRQ_TIMEOUT) + return rq->avg_irqload; + else + return 0; +} + +int walt_cpu_high_irqload(int cpu) { + return walt_irqload(cpu) >= sysctl_sched_walt_cpu_high_irqload; +} + +static int account_busy_for_cpu_time(struct rq *rq, struct task_struct *p, + u64 irqtime, int event) +{ + if (is_idle_task(p)) { + /* TASK_WAKE && TASK_MIGRATE is not possible on idle task! */ + if (event == PICK_NEXT_TASK) + return 0; + + /* PUT_PREV_TASK, TASK_UPDATE && IRQ_UPDATE are left */ + return irqtime || cpu_is_waiting_on_io(rq); + } + + if (event == TASK_WAKE) + return 0; + + if (event == PUT_PREV_TASK || event == IRQ_UPDATE || + event == TASK_UPDATE) + return 1; + + /* Only TASK_MIGRATE && PICK_NEXT_TASK left */ + return walt_freq_account_wait_time; +} + +/* + * Account cpu activity in its busy time counters (rq->curr/prev_runnable_sum) + */ +static void update_cpu_busy_time(struct task_struct *p, struct rq *rq, + int event, u64 wallclock, u64 irqtime) +{ + int new_window, nr_full_windows = 0; + int p_is_curr_task = (p == rq->curr); + u64 mark_start = p->ravg.mark_start; + u64 window_start = rq->window_start; + u32 window_size = walt_ravg_window; + u64 delta; + + new_window = mark_start < window_start; + if (new_window) { + nr_full_windows = div64_u64((window_start - mark_start), + window_size); + if (p->ravg.active_windows < USHRT_MAX) + p->ravg.active_windows++; + } + + /* Handle per-task window rollover. We don't care about the idle + * task or exiting tasks. */ + if (new_window && !is_idle_task(p) && !exiting_task(p)) { + u32 curr_window = 0; + + if (!nr_full_windows) + curr_window = p->ravg.curr_window; + + p->ravg.prev_window = curr_window; + p->ravg.curr_window = 0; + } + + if (!account_busy_for_cpu_time(rq, p, irqtime, event)) { + /* account_busy_for_cpu_time() = 0, so no update to the + * task's current window needs to be made. This could be + * for example + * + * - a wakeup event on a task within the current + * window (!new_window below, no action required), + * - switching to a new task from idle (PICK_NEXT_TASK) + * in a new window where irqtime is 0 and we aren't + * waiting on IO */ + + if (!new_window) + return; + + /* A new window has started. The RQ demand must be rolled + * over if p is the current task. */ + if (p_is_curr_task) { + u64 prev_sum = 0; + + /* p is either idle task or an exiting task */ + if (!nr_full_windows) { + prev_sum = rq->curr_runnable_sum; + } + + rq->prev_runnable_sum = prev_sum; + rq->curr_runnable_sum = 0; + } + + return; + } + + if (!new_window) { + /* account_busy_for_cpu_time() = 1 so busy time needs + * to be accounted to the current window. No rollover + * since we didn't start a new window. An example of this is + * when a task starts execution and then sleeps within the + * same window. */ + + if (!irqtime || !is_idle_task(p) || cpu_is_waiting_on_io(rq)) + delta = wallclock - mark_start; + else + delta = irqtime; + delta = scale_exec_time(delta, rq); + rq->curr_runnable_sum += delta; + if (!is_idle_task(p) && !exiting_task(p)) + p->ravg.curr_window += delta; + + return; + } + + if (!p_is_curr_task) { + /* account_busy_for_cpu_time() = 1 so busy time needs + * to be accounted to the current window. A new window + * has also started, but p is not the current task, so the + * window is not rolled over - just split up and account + * as necessary into curr and prev. The window is only + * rolled over when a new window is processed for the current + * task. + * + * Irqtime can't be accounted by a task that isn't the + * currently running task. */ + + if (!nr_full_windows) { + /* A full window hasn't elapsed, account partial + * contribution to previous completed window. */ + delta = scale_exec_time(window_start - mark_start, rq); + if (!exiting_task(p)) + p->ravg.prev_window += delta; + } else { + /* Since at least one full window has elapsed, + * the contribution to the previous window is the + * full window (window_size). */ + delta = scale_exec_time(window_size, rq); + if (!exiting_task(p)) + p->ravg.prev_window = delta; + } + rq->prev_runnable_sum += delta; + + /* Account piece of busy time in the current window. */ + delta = scale_exec_time(wallclock - window_start, rq); + rq->curr_runnable_sum += delta; + if (!exiting_task(p)) + p->ravg.curr_window = delta; + + return; + } + + if (!irqtime || !is_idle_task(p) || cpu_is_waiting_on_io(rq)) { + /* account_busy_for_cpu_time() = 1 so busy time needs + * to be accounted to the current window. A new window + * has started and p is the current task so rollover is + * needed. If any of these three above conditions are true + * then this busy time can't be accounted as irqtime. + * + * Busy time for the idle task or exiting tasks need not + * be accounted. + * + * An example of this would be a task that starts execution + * and then sleeps once a new window has begun. */ + + if (!nr_full_windows) { + /* A full window hasn't elapsed, account partial + * contribution to previous completed window. */ + delta = scale_exec_time(window_start - mark_start, rq); + if (!is_idle_task(p) && !exiting_task(p)) + p->ravg.prev_window += delta; + + delta += rq->curr_runnable_sum; + } else { + /* Since at least one full window has elapsed, + * the contribution to the previous window is the + * full window (window_size). */ + delta = scale_exec_time(window_size, rq); + if (!is_idle_task(p) && !exiting_task(p)) + p->ravg.prev_window = delta; + + } + /* + * Rollover for normal runnable sum is done here by overwriting + * the values in prev_runnable_sum and curr_runnable_sum. + * Rollover for new task runnable sum has completed by previous + * if-else statement. + */ + rq->prev_runnable_sum = delta; + + /* Account piece of busy time in the current window. */ + delta = scale_exec_time(wallclock - window_start, rq); + rq->curr_runnable_sum = delta; + if (!is_idle_task(p) && !exiting_task(p)) + p->ravg.curr_window = delta; + + return; + } + + if (irqtime) { + /* account_busy_for_cpu_time() = 1 so busy time needs + * to be accounted to the current window. A new window + * has started and p is the current task so rollover is + * needed. The current task must be the idle task because + * irqtime is not accounted for any other task. + * + * Irqtime will be accounted each time we process IRQ activity + * after a period of idleness, so we know the IRQ busy time + * started at wallclock - irqtime. */ + + BUG_ON(!is_idle_task(p)); + mark_start = wallclock - irqtime; + + /* Roll window over. If IRQ busy time was just in the current + * window then that is all that need be accounted. */ + rq->prev_runnable_sum = rq->curr_runnable_sum; + if (mark_start > window_start) { + rq->curr_runnable_sum = scale_exec_time(irqtime, rq); + return; + } + + /* The IRQ busy time spanned multiple windows. Process the + * busy time preceding the current window start first. */ + delta = window_start - mark_start; + if (delta > window_size) + delta = window_size; + delta = scale_exec_time(delta, rq); + rq->prev_runnable_sum += delta; + + /* Process the remaining IRQ busy time in the current window. */ + delta = wallclock - window_start; + rq->curr_runnable_sum = scale_exec_time(delta, rq); + + return; + } + + BUG(); +} + +static int account_busy_for_task_demand(struct task_struct *p, int event) +{ + /* No need to bother updating task demand for exiting tasks + * or the idle task. */ + if (exiting_task(p) || is_idle_task(p)) + return 0; + + /* When a task is waking up it is completing a segment of non-busy + * time. Likewise, if wait time is not treated as busy time, then + * when a task begins to run or is migrated, it is not running and + * is completing a segment of non-busy time. */ + if (event == TASK_WAKE || (!walt_account_wait_time && + (event == PICK_NEXT_TASK || event == TASK_MIGRATE))) + return 0; + + return 1; +} + +/* + * Called when new window is starting for a task, to record cpu usage over + * recently concluded window(s). Normally 'samples' should be 1. It can be > 1 + * when, say, a real-time task runs without preemption for several windows at a + * stretch. + */ +static void update_history(struct rq *rq, struct task_struct *p, + u32 runtime, int samples, int event) +{ + u32 *hist = &p->ravg.sum_history[0]; + int ridx, widx; + u32 max = 0, avg, demand; + u64 sum = 0; + + /* Ignore windows where task had no activity */ + if (!runtime || is_idle_task(p) || exiting_task(p) || !samples) + goto done; + + /* Push new 'runtime' value onto stack */ + widx = walt_ravg_hist_size - 1; + ridx = widx - samples; + for (; ridx >= 0; --widx, --ridx) { + hist[widx] = hist[ridx]; + sum += hist[widx]; + if (hist[widx] > max) + max = hist[widx]; + } + + for (widx = 0; widx < samples && widx < walt_ravg_hist_size; widx++) { + hist[widx] = runtime; + sum += hist[widx]; + if (hist[widx] > max) + max = hist[widx]; + } + + p->ravg.sum = 0; + + if (walt_window_stats_policy == WINDOW_STATS_RECENT) { + demand = runtime; + } else if (walt_window_stats_policy == WINDOW_STATS_MAX) { + demand = max; + } else { + avg = div64_u64(sum, walt_ravg_hist_size); + if (walt_window_stats_policy == WINDOW_STATS_AVG) + demand = avg; + else + demand = max(avg, runtime); + } + + /* + * A throttled deadline sched class task gets dequeued without + * changing p->on_rq. Since the dequeue decrements hmp stats + * avoid decrementing it here again. + */ + if (task_on_rq_queued(p) && (!task_has_dl_policy(p) || + !p->dl.dl_throttled)) + fixup_cumulative_runnable_avg(rq, p, demand); + + p->ravg.demand = demand; + +done: + trace_walt_update_history(rq, p, runtime, samples, event); + return; +} + +static void add_to_task_demand(struct rq *rq, struct task_struct *p, + u64 delta) +{ + delta = scale_exec_time(delta, rq); + p->ravg.sum += delta; + if (unlikely(p->ravg.sum > walt_ravg_window)) + p->ravg.sum = walt_ravg_window; +} + +/* + * Account cpu demand of task and/or update task's cpu demand history + * + * ms = p->ravg.mark_start; + * wc = wallclock + * ws = rq->window_start + * + * Three possibilities: + * + * a) Task event is contained within one window. + * window_start < mark_start < wallclock + * + * ws ms wc + * | | | + * V V V + * |---------------| + * + * In this case, p->ravg.sum is updated *iff* event is appropriate + * (ex: event == PUT_PREV_TASK) + * + * b) Task event spans two windows. + * mark_start < window_start < wallclock + * + * ms ws wc + * | | | + * V V V + * -----|------------------- + * + * In this case, p->ravg.sum is updated with (ws - ms) *iff* event + * is appropriate, then a new window sample is recorded followed + * by p->ravg.sum being set to (wc - ws) *iff* event is appropriate. + * + * c) Task event spans more than two windows. + * + * ms ws_tmp ws wc + * | | | | + * V V V V + * ---|-------|-------|-------|-------|------ + * | | + * |<------ nr_full_windows ------>| + * + * In this case, p->ravg.sum is updated with (ws_tmp - ms) first *iff* + * event is appropriate, window sample of p->ravg.sum is recorded, + * 'nr_full_window' samples of window_size is also recorded *iff* + * event is appropriate and finally p->ravg.sum is set to (wc - ws) + * *iff* event is appropriate. + * + * IMPORTANT : Leave p->ravg.mark_start unchanged, as update_cpu_busy_time() + * depends on it! + */ +static void update_task_demand(struct task_struct *p, struct rq *rq, + int event, u64 wallclock) +{ + u64 mark_start = p->ravg.mark_start; + u64 delta, window_start = rq->window_start; + int new_window, nr_full_windows; + u32 window_size = walt_ravg_window; + + new_window = mark_start < window_start; + if (!account_busy_for_task_demand(p, event)) { + if (new_window) + /* If the time accounted isn't being accounted as + * busy time, and a new window started, only the + * previous window need be closed out with the + * pre-existing demand. Multiple windows may have + * elapsed, but since empty windows are dropped, + * it is not necessary to account those. */ + update_history(rq, p, p->ravg.sum, 1, event); + return; + } + + if (!new_window) { + /* The simple case - busy time contained within the existing + * window. */ + add_to_task_demand(rq, p, wallclock - mark_start); + return; + } + + /* Busy time spans at least two windows. Temporarily rewind + * window_start to first window boundary after mark_start. */ + delta = window_start - mark_start; + nr_full_windows = div64_u64(delta, window_size); + window_start -= (u64)nr_full_windows * (u64)window_size; + + /* Process (window_start - mark_start) first */ + add_to_task_demand(rq, p, window_start - mark_start); + + /* Push new sample(s) into task's demand history */ + update_history(rq, p, p->ravg.sum, 1, event); + if (nr_full_windows) + update_history(rq, p, scale_exec_time(window_size, rq), + nr_full_windows, event); + + /* Roll window_start back to current to process any remainder + * in current window. */ + window_start += (u64)nr_full_windows * (u64)window_size; + + /* Process (wallclock - window_start) next */ + mark_start = window_start; + add_to_task_demand(rq, p, wallclock - mark_start); +} + +/* Reflect task activity on its demand and cpu's busy time statistics */ +void walt_update_task_ravg(struct task_struct *p, struct rq *rq, + int event, u64 wallclock, u64 irqtime) +{ + if (walt_disabled || !rq->window_start) + return; + + lockdep_assert_held(&rq->lock); + + update_window_start(rq, wallclock); + + if (!p->ravg.mark_start) + goto done; + + update_task_demand(p, rq, event, wallclock); + update_cpu_busy_time(p, rq, event, wallclock, irqtime); + +done: + trace_walt_update_task_ravg(p, rq, event, wallclock, irqtime); + + p->ravg.mark_start = wallclock; +} + +unsigned long __weak arch_get_cpu_efficiency(int cpu) +{ + return SCHED_LOAD_SCALE; +} + +void walt_init_cpu_efficiency(void) +{ + int i, efficiency; + unsigned int max = 0, min = UINT_MAX; + + for_each_possible_cpu(i) { + efficiency = arch_get_cpu_efficiency(i); + cpu_rq(i)->efficiency = efficiency; + + if (efficiency > max) + max = efficiency; + if (efficiency < min) + min = efficiency; + } + + if (max) + max_possible_efficiency = max; + + if (min) + min_possible_efficiency = min; +} + +static void reset_task_stats(struct task_struct *p) +{ + u32 sum = 0; + + if (exiting_task(p)) + sum = EXITING_TASK_MARKER; + + memset(&p->ravg, 0, sizeof(struct ravg)); + /* Retain EXITING_TASK marker */ + p->ravg.sum_history[0] = sum; +} + +void walt_mark_task_starting(struct task_struct *p) +{ + u64 wallclock; + struct rq *rq = task_rq(p); + + if (!rq->window_start) { + reset_task_stats(p); + return; + } + + wallclock = walt_ktime_clock(); + p->ravg.mark_start = wallclock; +} + +void walt_set_window_start(struct rq *rq) +{ + int cpu = cpu_of(rq); + struct rq *sync_rq = cpu_rq(sync_cpu); + + if (rq->window_start) + return; + + if (cpu == sync_cpu) { + rq->window_start = walt_ktime_clock(); + } else { + raw_spin_unlock(&rq->lock); + double_rq_lock(rq, sync_rq); + rq->window_start = cpu_rq(sync_cpu)->window_start; + rq->curr_runnable_sum = rq->prev_runnable_sum = 0; + raw_spin_unlock(&sync_rq->lock); + } + + rq->curr->ravg.mark_start = rq->window_start; +} + +void walt_migrate_sync_cpu(int cpu) +{ + if (cpu == sync_cpu) + sync_cpu = smp_processor_id(); +} + +void walt_fixup_busy_time(struct task_struct *p, int new_cpu) +{ + struct rq *src_rq = task_rq(p); + struct rq *dest_rq = cpu_rq(new_cpu); + u64 wallclock; + + if (!p->on_rq && p->state != TASK_WAKING) + return; + + if (exiting_task(p)) { + return; + } + + if (p->state == TASK_WAKING) + double_rq_lock(src_rq, dest_rq); + + wallclock = walt_ktime_clock(); + + walt_update_task_ravg(task_rq(p)->curr, task_rq(p), + TASK_UPDATE, wallclock, 0); + walt_update_task_ravg(dest_rq->curr, dest_rq, + TASK_UPDATE, wallclock, 0); + + walt_update_task_ravg(p, task_rq(p), TASK_MIGRATE, wallclock, 0); + + if (p->ravg.curr_window) { + src_rq->curr_runnable_sum -= p->ravg.curr_window; + dest_rq->curr_runnable_sum += p->ravg.curr_window; + } + + if (p->ravg.prev_window) { + src_rq->prev_runnable_sum -= p->ravg.prev_window; + dest_rq->prev_runnable_sum += p->ravg.prev_window; + } + + if ((s64)src_rq->prev_runnable_sum < 0) { + src_rq->prev_runnable_sum = 0; + WARN_ON(1); + } + if ((s64)src_rq->curr_runnable_sum < 0) { + src_rq->curr_runnable_sum = 0; + WARN_ON(1); + } + + trace_walt_migration_update_sum(src_rq, p); + trace_walt_migration_update_sum(dest_rq, p); + + if (p->state == TASK_WAKING) + double_rq_unlock(src_rq, dest_rq); +} + +/* Keep track of max/min capacity possible across CPUs "currently" */ +static void __update_min_max_capacity(void) +{ + int i; + int max = 0, min = INT_MAX; + + for_each_online_cpu(i) { + if (cpu_rq(i)->capacity > max) + max = cpu_rq(i)->capacity; + if (cpu_rq(i)->capacity < min) + min = cpu_rq(i)->capacity; + } + + max_capacity = max; + min_capacity = min; +} + +static void update_min_max_capacity(void) +{ + unsigned long flags; + int i; + + local_irq_save(flags); + for_each_possible_cpu(i) + raw_spin_lock(&cpu_rq(i)->lock); + + __update_min_max_capacity(); + + for_each_possible_cpu(i) + raw_spin_unlock(&cpu_rq(i)->lock); + local_irq_restore(flags); +} + +/* + * Return 'capacity' of a cpu in reference to "least" efficient cpu, such that + * least efficient cpu gets capacity of 1024 + */ +static unsigned long capacity_scale_cpu_efficiency(int cpu) +{ + return (1024 * cpu_rq(cpu)->efficiency) / min_possible_efficiency; +} + +/* + * Return 'capacity' of a cpu in reference to cpu with lowest max_freq + * (min_max_freq), such that one with lowest max_freq gets capacity of 1024. + */ +static unsigned long capacity_scale_cpu_freq(int cpu) +{ + return (1024 * cpu_rq(cpu)->max_freq) / min_max_freq; +} + +/* + * Return load_scale_factor of a cpu in reference to "most" efficient cpu, so + * that "most" efficient cpu gets a load_scale_factor of 1 + */ +static unsigned long load_scale_cpu_efficiency(int cpu) +{ + return DIV_ROUND_UP(1024 * max_possible_efficiency, + cpu_rq(cpu)->efficiency); +} + +/* + * Return load_scale_factor of a cpu in reference to cpu with best max_freq + * (max_possible_freq), so that one with best max_freq gets a load_scale_factor + * of 1. + */ +static unsigned long load_scale_cpu_freq(int cpu) +{ + return DIV_ROUND_UP(1024 * max_possible_freq, cpu_rq(cpu)->max_freq); +} + +static int compute_capacity(int cpu) +{ + int capacity = 1024; + + capacity *= capacity_scale_cpu_efficiency(cpu); + capacity >>= 10; + + capacity *= capacity_scale_cpu_freq(cpu); + capacity >>= 10; + + return capacity; +} + +static int compute_load_scale_factor(int cpu) +{ + int load_scale = 1024; + + /* + * load_scale_factor accounts for the fact that task load + * is in reference to "best" performing cpu. Task's load will need to be + * scaled (up) by a factor to determine suitability to be placed on a + * (little) cpu. + */ + load_scale *= load_scale_cpu_efficiency(cpu); + load_scale >>= 10; + + load_scale *= load_scale_cpu_freq(cpu); + load_scale >>= 10; + + return load_scale; +} + +static int cpufreq_notifier_policy(struct notifier_block *nb, + unsigned long val, void *data) +{ + struct cpufreq_policy *policy = (struct cpufreq_policy *)data; + int i, update_max = 0; + u64 highest_mpc = 0, highest_mplsf = 0; + const struct cpumask *cpus = policy->related_cpus; + unsigned int orig_min_max_freq = min_max_freq; + unsigned int orig_max_possible_freq = max_possible_freq; + /* Initialized to policy->max in case policy->related_cpus is empty! */ + unsigned int orig_max_freq = policy->max; + + if (val != CPUFREQ_NOTIFY && val != CPUFREQ_REMOVE_POLICY && + val != CPUFREQ_CREATE_POLICY) + return 0; + + if (val == CPUFREQ_REMOVE_POLICY || val == CPUFREQ_CREATE_POLICY) { + update_min_max_capacity(); + return 0; + } + + for_each_cpu(i, policy->related_cpus) { + cpumask_copy(&cpu_rq(i)->freq_domain_cpumask, + policy->related_cpus); + orig_max_freq = cpu_rq(i)->max_freq; + cpu_rq(i)->min_freq = policy->min; + cpu_rq(i)->max_freq = policy->max; + cpu_rq(i)->cur_freq = policy->cur; + cpu_rq(i)->max_possible_freq = policy->cpuinfo.max_freq; + } + + max_possible_freq = max(max_possible_freq, policy->cpuinfo.max_freq); + if (min_max_freq == 1) + min_max_freq = UINT_MAX; + min_max_freq = min(min_max_freq, policy->cpuinfo.max_freq); + BUG_ON(!min_max_freq); + BUG_ON(!policy->max); + + /* Changes to policy other than max_freq don't require any updates */ + if (orig_max_freq == policy->max) + return 0; + + /* + * A changed min_max_freq or max_possible_freq (possible during bootup) + * needs to trigger re-computation of load_scale_factor and capacity for + * all possible cpus (even those offline). It also needs to trigger + * re-computation of nr_big_task count on all online cpus. + * + * A changed rq->max_freq otoh needs to trigger re-computation of + * load_scale_factor and capacity for just the cluster of cpus involved. + * Since small task definition depends on max_load_scale_factor, a + * changed load_scale_factor of one cluster could influence + * classification of tasks in another cluster. Hence a changed + * rq->max_freq will need to trigger re-computation of nr_big_task + * count on all online cpus. + * + * While it should be sufficient for nr_big_tasks to be + * re-computed for only online cpus, we have inadequate context + * information here (in policy notifier) with regard to hotplug-safety + * context in which notification is issued. As a result, we can't use + * get_online_cpus() here, as it can lead to deadlock. Until cpufreq is + * fixed up to issue notification always in hotplug-safe context, + * re-compute nr_big_task for all possible cpus. + */ + + if (orig_min_max_freq != min_max_freq || + orig_max_possible_freq != max_possible_freq) { + cpus = cpu_possible_mask; + update_max = 1; + } + + /* + * Changed load_scale_factor can trigger reclassification of tasks as + * big or small. Make this change "atomic" so that tasks are accounted + * properly due to changed load_scale_factor + */ + for_each_cpu(i, cpus) { + struct rq *rq = cpu_rq(i); + + rq->capacity = compute_capacity(i); + rq->load_scale_factor = compute_load_scale_factor(i); + + if (update_max) { + u64 mpc, mplsf; + + mpc = div_u64(((u64) rq->capacity) * + rq->max_possible_freq, rq->max_freq); + rq->max_possible_capacity = (int) mpc; + + mplsf = div_u64(((u64) rq->load_scale_factor) * + rq->max_possible_freq, rq->max_freq); + + if (mpc > highest_mpc) { + highest_mpc = mpc; + cpumask_clear(&mpc_mask); + cpumask_set_cpu(i, &mpc_mask); + } else if (mpc == highest_mpc) { + cpumask_set_cpu(i, &mpc_mask); + } + + if (mplsf > highest_mplsf) + highest_mplsf = mplsf; + } + } + + if (update_max) { + max_possible_capacity = highest_mpc; + max_load_scale_factor = highest_mplsf; + } + + __update_min_max_capacity(); + + return 0; +} + +static int cpufreq_notifier_trans(struct notifier_block *nb, + unsigned long val, void *data) +{ + struct cpufreq_freqs *freq = (struct cpufreq_freqs *)data; + unsigned int cpu = freq->cpu, new_freq = freq->new; + unsigned long flags; + int i; + + if (val != CPUFREQ_POSTCHANGE) + return 0; + + BUG_ON(!new_freq); + + if (cpu_rq(cpu)->cur_freq == new_freq) + return 0; + + for_each_cpu(i, &cpu_rq(cpu)->freq_domain_cpumask) { + struct rq *rq = cpu_rq(i); + + raw_spin_lock_irqsave(&rq->lock, flags); + walt_update_task_ravg(rq->curr, rq, TASK_UPDATE, + walt_ktime_clock(), 0); + rq->cur_freq = new_freq; + raw_spin_unlock_irqrestore(&rq->lock, flags); + } + + return 0; +} + +static struct notifier_block notifier_policy_block = { + .notifier_call = cpufreq_notifier_policy +}; + +static struct notifier_block notifier_trans_block = { + .notifier_call = cpufreq_notifier_trans +}; + +static int register_sched_callback(void) +{ + int ret; + + ret = cpufreq_register_notifier(¬ifier_policy_block, + CPUFREQ_POLICY_NOTIFIER); + + if (!ret) + ret = cpufreq_register_notifier(¬ifier_trans_block, + CPUFREQ_TRANSITION_NOTIFIER); + + return 0; +} + +/* + * cpufreq callbacks can be registered at core_initcall or later time. + * Any registration done prior to that is "forgotten" by cpufreq. See + * initialization of variable init_cpufreq_transition_notifier_list_called + * for further information. + */ +core_initcall(register_sched_callback); + +void walt_init_new_task_load(struct task_struct *p) +{ + int i; + u32 init_load_windows = + div64_u64((u64)sysctl_sched_walt_init_task_load_pct * + (u64)walt_ravg_window, 100); + u32 init_load_pct = current->init_load_pct; + + p->init_load_pct = 0; + memset(&p->ravg, 0, sizeof(struct ravg)); + + if (init_load_pct) { + init_load_windows = div64_u64((u64)init_load_pct * + (u64)walt_ravg_window, 100); + } + + p->ravg.demand = init_load_windows; + for (i = 0; i < RAVG_HIST_SIZE_MAX; ++i) + p->ravg.sum_history[i] = init_load_windows; +} diff --git a/kernel/sched/walt.h b/kernel/sched/walt.h new file mode 100644 index 000000000000..e181c87a928d --- /dev/null +++ b/kernel/sched/walt.h @@ -0,0 +1,62 @@ +/* + * Copyright (c) 2016, The Linux Foundation. All rights reserved. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 and + * only version 2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + */ + +#ifndef __WALT_H +#define __WALT_H + +#ifdef CONFIG_SCHED_WALT + +void walt_update_task_ravg(struct task_struct *p, struct rq *rq, int event, + u64 wallclock, u64 irqtime); +void walt_inc_cumulative_runnable_avg(struct rq *rq, struct task_struct *p); +void walt_dec_cumulative_runnable_avg(struct rq *rq, struct task_struct *p); +void walt_inc_cfs_cumulative_runnable_avg(struct cfs_rq *rq, + struct task_struct *p); +void walt_dec_cfs_cumulative_runnable_avg(struct cfs_rq *rq, + struct task_struct *p); +void walt_fixup_busy_time(struct task_struct *p, int new_cpu); +void walt_init_new_task_load(struct task_struct *p); +void walt_mark_task_starting(struct task_struct *p); +void walt_set_window_start(struct rq *rq); +void walt_migrate_sync_cpu(int cpu); +void walt_init_cpu_efficiency(void); +u64 walt_ktime_clock(void); +void walt_account_irqtime(int cpu, struct task_struct *curr, u64 delta, + u64 wallclock); + +u64 walt_irqload(int cpu); +int walt_cpu_high_irqload(int cpu); + +#else /* CONFIG_SCHED_WALT */ + +static inline void walt_update_task_ravg(struct task_struct *p, struct rq *rq, + int event, u64 wallclock, u64 irqtime) { } +static inline void walt_inc_cumulative_runnable_avg(struct rq *rq, struct task_struct *p) { } +static inline void walt_dec_cumulative_runnable_avg(struct rq *rq, struct task_struct *p) { } +static inline void walt_inc_cfs_cumulative_runnable_avg(struct cfs_rq *rq, + struct task_struct *p) { } +static inline void walt_dec_cfs_cumulative_runnable_avg(struct cfs_rq *rq, + struct task_struct *p) { } +static inline void walt_fixup_busy_time(struct task_struct *p, int new_cpu) { } +static inline void walt_init_new_task_load(struct task_struct *p) { } +static inline void walt_mark_task_starting(struct task_struct *p) { } +static inline void walt_set_window_start(struct rq *rq) { } +static inline void walt_migrate_sync_cpu(int cpu) { } +static inline void walt_init_cpu_efficiency(void) { } +static inline u64 walt_ktime_clock(void) { return 0; } + +#endif /* CONFIG_SCHED_WALT */ + +extern unsigned int walt_disabled; + +#endif diff --git a/kernel/sysctl.c b/kernel/sysctl.c index 46822df92c50..abb795e8a6f1 100644 --- a/kernel/sysctl.c +++ b/kernel/sysctl.c @@ -305,6 +305,64 @@ static struct ctl_table kern_table[] = { .extra2 = &max_sched_granularity_ns, }, { + .procname = "sched_is_big_little", + .data = &sysctl_sched_is_big_little, + .maxlen = sizeof(unsigned int), + .mode = 0644, + .proc_handler = proc_dointvec, + }, +#ifdef CONFIG_SCHED_WALT + { + .procname = "sched_use_walt_cpu_util", + .data = &sysctl_sched_use_walt_cpu_util, + .maxlen = sizeof(unsigned int), + .mode = 0644, + .proc_handler = proc_dointvec, + }, + { + .procname = "sched_use_walt_task_util", + .data = &sysctl_sched_use_walt_task_util, + .maxlen = sizeof(unsigned int), + .mode = 0644, + .proc_handler = proc_dointvec, + }, + { + .procname = "sched_walt_init_task_load_pct", + .data = &sysctl_sched_walt_init_task_load_pct, + .maxlen = sizeof(unsigned int), + .mode = 0644, + .proc_handler = proc_dointvec, + }, + { + .procname = "sched_walt_cpu_high_irqload", + .data = &sysctl_sched_walt_cpu_high_irqload, + .maxlen = sizeof(unsigned int), + .mode = 0644, + .proc_handler = proc_dointvec, + }, +#endif + { + .procname = "sched_sync_hint_enable", + .data = &sysctl_sched_sync_hint_enable, + .maxlen = sizeof(unsigned int), + .mode = 0644, + .proc_handler = proc_dointvec, + }, + { + .procname = "sched_initial_task_util", + .data = &sysctl_sched_initial_task_util, + .maxlen = sizeof(unsigned int), + .mode = 0644, + .proc_handler = proc_dointvec, + }, + { + .procname = "sched_cstate_aware", + .data = &sysctl_sched_cstate_aware, + .maxlen = sizeof(unsigned int), + .mode = 0644, + .proc_handler = proc_dointvec, + }, + { .procname = "sched_wakeup_granularity_ns", .data = &sysctl_sched_wakeup_granularity, .maxlen = sizeof(unsigned int), @@ -2097,6 +2155,21 @@ static int do_proc_dointvec_conv(bool *negp, unsigned long *lvalp, return 0; } +static int do_proc_douintvec_conv(bool *negp, unsigned long *lvalp, + int *valp, + int write, void *data) +{ + if (write) { + if (*negp) + return -EINVAL; + *valp = *lvalp; + } else { + unsigned int val = *valp; + *lvalp = (unsigned long)val; + } + return 0; +} + static const char proc_wspace_sep[] = { ' ', '\t', '\n' }; static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table, @@ -2224,8 +2297,27 @@ static int do_proc_dointvec(struct ctl_table *table, int write, int proc_dointvec(struct ctl_table *table, int write, void __user *buffer, size_t *lenp, loff_t *ppos) { - return do_proc_dointvec(table,write,buffer,lenp,ppos, - NULL,NULL); + return do_proc_dointvec(table, write, buffer, lenp, ppos, NULL, NULL); +} + +/** + * proc_douintvec - read a vector of unsigned integers + * @table: the sysctl table + * @write: %TRUE if this is a write to the sysctl file + * @buffer: the user buffer + * @lenp: the size of the user buffer + * @ppos: file position + * + * Reads/writes up to table->maxlen/sizeof(unsigned int) unsigned integer + * values from/to the user buffer, treated as an ASCII string. + * + * Returns 0 on success. + */ +int proc_douintvec(struct ctl_table *table, int write, + void __user *buffer, size_t *lenp, loff_t *ppos) +{ + return do_proc_dointvec(table, write, buffer, lenp, ppos, + do_proc_douintvec_conv, NULL); } /* @@ -2838,6 +2930,12 @@ int proc_dointvec(struct ctl_table *table, int write, return -ENOSYS; } +int proc_douintvec(struct ctl_table *table, int write, + void __user *buffer, size_t *lenp, loff_t *ppos) +{ + return -ENOSYS; +} + int proc_dointvec_minmax(struct ctl_table *table, int write, void __user *buffer, size_t *lenp, loff_t *ppos) { @@ -2883,6 +2981,7 @@ int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write, * exception granted :-) */ EXPORT_SYMBOL(proc_dointvec); +EXPORT_SYMBOL(proc_douintvec); EXPORT_SYMBOL(proc_dointvec_jiffies); EXPORT_SYMBOL(proc_dointvec_minmax); EXPORT_SYMBOL(proc_dointvec_userhz_jiffies); diff --git a/kernel/time/clocksource.c b/kernel/time/clocksource.c index 1347882d131e..b98810d2f3b4 100644 --- a/kernel/time/clocksource.c +++ b/kernel/time/clocksource.c @@ -323,13 +323,42 @@ static void clocksource_enqueue_watchdog(struct clocksource *cs) /* cs is a watchdog. */ if (cs->flags & CLOCK_SOURCE_IS_CONTINUOUS) cs->flags |= CLOCK_SOURCE_VALID_FOR_HRES; + } + spin_unlock_irqrestore(&watchdog_lock, flags); +} + +static void clocksource_select_watchdog(bool fallback) +{ + struct clocksource *cs, *old_wd; + unsigned long flags; + + spin_lock_irqsave(&watchdog_lock, flags); + /* save current watchdog */ + old_wd = watchdog; + if (fallback) + watchdog = NULL; + + list_for_each_entry(cs, &clocksource_list, list) { + /* cs is a clocksource to be watched. */ + if (cs->flags & CLOCK_SOURCE_MUST_VERIFY) + continue; + + /* Skip current if we were requested for a fallback. */ + if (fallback && cs == old_wd) + continue; + /* Pick the best watchdog. */ - if (!watchdog || cs->rating > watchdog->rating) { + if (!watchdog || cs->rating > watchdog->rating) watchdog = cs; - /* Reset watchdog cycles */ - clocksource_reset_watchdog(); - } } + /* If we failed to find a fallback restore the old one. */ + if (!watchdog) + watchdog = old_wd; + + /* If we changed the watchdog we need to reset cycles. */ + if (watchdog != old_wd) + clocksource_reset_watchdog(); + /* Check if the watchdog timer needs to be started. */ clocksource_start_watchdog(); spin_unlock_irqrestore(&watchdog_lock, flags); @@ -404,6 +433,7 @@ static void clocksource_enqueue_watchdog(struct clocksource *cs) cs->flags |= CLOCK_SOURCE_VALID_FOR_HRES; } +static void clocksource_select_watchdog(bool fallback) { } static inline void clocksource_dequeue_watchdog(struct clocksource *cs) { } static inline void clocksource_resume_watchdog(void) { } static inline int __clocksource_watchdog_kthread(void) { return 0; } @@ -736,6 +766,7 @@ int __clocksource_register_scale(struct clocksource *cs, u32 scale, u32 freq) clocksource_enqueue(cs); clocksource_enqueue_watchdog(cs); clocksource_select(); + clocksource_select_watchdog(false); mutex_unlock(&clocksource_mutex); return 0; } @@ -758,6 +789,7 @@ void clocksource_change_rating(struct clocksource *cs, int rating) mutex_lock(&clocksource_mutex); __clocksource_change_rating(cs, rating); clocksource_select(); + clocksource_select_watchdog(false); mutex_unlock(&clocksource_mutex); } EXPORT_SYMBOL(clocksource_change_rating); @@ -767,12 +799,12 @@ EXPORT_SYMBOL(clocksource_change_rating); */ static int clocksource_unbind(struct clocksource *cs) { - /* - * I really can't convince myself to support this on hardware - * designed by lobotomized monkeys. - */ - if (clocksource_is_watchdog(cs)) - return -EBUSY; + if (clocksource_is_watchdog(cs)) { + /* Select and try to install a replacement watchdog. */ + clocksource_select_watchdog(true); + if (clocksource_is_watchdog(cs)) + return -EBUSY; + } if (cs == curr_clocksource) { /* Select and try to install a replacement clock source */ diff --git a/kernel/time/hrtimer.c b/kernel/time/hrtimer.c index 58a321c34cfb..405536b22c0c 100644 --- a/kernel/time/hrtimer.c +++ b/kernel/time/hrtimer.c @@ -94,6 +94,9 @@ DEFINE_PER_CPU(struct hrtimer_cpu_base, hrtimer_bases) = }; static const int hrtimer_clock_to_base_table[MAX_CLOCKS] = { + /* Make sure we catch unsupported clockids */ + [0 ... MAX_CLOCKS - 1] = HRTIMER_MAX_CLOCK_BASES, + [CLOCK_REALTIME] = HRTIMER_BASE_REALTIME, [CLOCK_MONOTONIC] = HRTIMER_BASE_MONOTONIC, [CLOCK_BOOTTIME] = HRTIMER_BASE_BOOTTIME, @@ -102,7 +105,9 @@ static const int hrtimer_clock_to_base_table[MAX_CLOCKS] = { static inline int hrtimer_clockid_to_base(clockid_t clock_id) { - return hrtimer_clock_to_base_table[clock_id]; + int base = hrtimer_clock_to_base_table[clock_id]; + BUG_ON(base == HRTIMER_MAX_CLOCK_BASES); + return base; } /* diff --git a/kernel/time/ntp.c b/kernel/time/ntp.c index 149cc8086aea..ab861771e37f 100644 --- a/kernel/time/ntp.c +++ b/kernel/time/ntp.c @@ -674,8 +674,24 @@ int ntp_validate_timex(struct timex *txc) return -EINVAL; } - if ((txc->modes & ADJ_SETOFFSET) && (!capable(CAP_SYS_TIME))) - return -EPERM; + if (txc->modes & ADJ_SETOFFSET) { + /* In order to inject time, you gotta be super-user! */ + if (!capable(CAP_SYS_TIME)) + return -EPERM; + + if (txc->modes & ADJ_NANO) { + struct timespec ts; + + ts.tv_sec = txc->time.tv_sec; + ts.tv_nsec = txc->time.tv_usec; + if (!timespec_inject_offset_valid(&ts)) + return -EINVAL; + + } else { + if (!timeval_inject_offset_valid(&txc->time)) + return -EINVAL; + } + } /* * Check for potential multiplication overflows that can diff --git a/kernel/time/posix-cpu-timers.c b/kernel/time/posix-cpu-timers.c index f5e86d282d52..80016b329d94 100644 --- a/kernel/time/posix-cpu-timers.c +++ b/kernel/time/posix-cpu-timers.c @@ -808,6 +808,7 @@ static void posix_cpu_timer_get(struct k_itimer *timer, struct itimerspec *itp) timer->it.cpu.expires = 0; sample_to_timespec(timer->it_clock, timer->it.cpu.expires, &itp->it_value); + return; } else { cpu_timer_sample_group(timer->it_clock, p, &now); unlock_task_sighand(p, &flags); diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c index 99188ee5d9d0..445601c580d6 100644 --- a/kernel/time/timekeeping.c +++ b/kernel/time/timekeeping.c @@ -298,17 +298,34 @@ u32 (*arch_gettimeoffset)(void) = default_arch_gettimeoffset; static inline u32 arch_gettimeoffset(void) { return 0; } #endif +static inline s64 timekeeping_delta_to_ns(struct tk_read_base *tkr, + cycle_t delta) +{ + s64 nsec; + + nsec = delta * tkr->mult + tkr->xtime_nsec; + nsec >>= tkr->shift; + + /* If arch requires, add in get_arch_timeoffset() */ + return nsec + arch_gettimeoffset(); +} + static inline s64 timekeeping_get_ns(struct tk_read_base *tkr) { cycle_t delta; - s64 nsec; delta = timekeeping_get_delta(tkr); + return timekeeping_delta_to_ns(tkr, delta); +} - nsec = (delta * tkr->mult + tkr->xtime_nsec) >> tkr->shift; +static inline s64 timekeeping_cycles_to_ns(struct tk_read_base *tkr, + cycle_t cycles) +{ + cycle_t delta; - /* If arch requires, add in get_arch_timeoffset() */ - return nsec + arch_gettimeoffset(); + /* calculate the delta since the last update_wall_time */ + delta = clocksource_delta(cycles, tkr->cycle_last, tkr->mask); + return timekeeping_delta_to_ns(tkr, delta); } /** @@ -383,7 +400,13 @@ static __always_inline u64 __ktime_get_fast_ns(struct tk_fast *tkf) do { seq = raw_read_seqcount_latch(&tkf->seq); tkr = tkf->base + (seq & 0x01); - now = ktime_to_ns(tkr->base) + timekeeping_get_ns(tkr); + now = ktime_to_ns(tkr->base); + + now += timekeeping_delta_to_ns(tkr, + clocksource_delta( + tkr->read(tkr->clock), + tkr->cycle_last, + tkr->mask)); } while (read_seqcount_retry(&tkf->seq, seq)); return now; @@ -958,7 +981,7 @@ int timekeeping_inject_offset(struct timespec *ts) struct timespec64 ts64, tmp; int ret = 0; - if ((unsigned long)ts->tv_nsec >= NSEC_PER_SEC) + if (!timespec_inject_offset_valid(ts)) return -EINVAL; ts64 = timespec_to_timespec64(*ts); diff --git a/kernel/time/timekeeping_debug.c b/kernel/time/timekeeping_debug.c index f6bd65236712..107310a6f36f 100644 --- a/kernel/time/timekeeping_debug.c +++ b/kernel/time/timekeeping_debug.c @@ -23,7 +23,9 @@ #include "timekeeping_internal.h" -static unsigned int sleep_time_bin[32] = {0}; +#define NUM_BINS 32 + +static unsigned int sleep_time_bin[NUM_BINS] = {0}; static int tk_debug_show_sleep_time(struct seq_file *s, void *data) { @@ -69,6 +71,9 @@ late_initcall(tk_debug_sleep_time_init); void tk_debug_account_sleep_time(struct timespec64 *t) { - sleep_time_bin[fls(t->tv_sec)]++; + /* Cap bin index so we don't overflow the array */ + int bin = min(fls(t->tv_sec), NUM_BINS-1); + + sleep_time_bin[bin]++; } diff --git a/kernel/trace/Makefile b/kernel/trace/Makefile index ba04bf0c2653..a9bba37fab5a 100644 --- a/kernel/trace/Makefile +++ b/kernel/trace/Makefile @@ -1,4 +1,8 @@ +# We are fully aware of the dangers of __builtin_return_address() +FRAME_CFLAGS := $(call cc-disable-warning,frame-address) +KBUILD_CFLAGS += $(FRAME_CFLAGS) + # Do not instrument the tracer itself: ifdef CONFIG_FUNCTION_TRACER diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c index dd4d86ae8e21..bff7d4c69274 100644 --- a/kernel/trace/trace.c +++ b/kernel/trace/trace.c @@ -4816,19 +4816,20 @@ tracing_read_pipe(struct file *filp, char __user *ubuf, struct trace_iterator *iter = filp->private_data; ssize_t sret; - /* return any leftover data */ - sret = trace_seq_to_user(&iter->seq, ubuf, cnt); - if (sret != -EBUSY) - return sret; - - trace_seq_init(&iter->seq); - /* * Avoid more than one consumer on a single file descriptor * This is just a matter of traces coherency, the ring buffer itself * is protected. */ mutex_lock(&iter->mutex); + + /* return any leftover data */ + sret = trace_seq_to_user(&iter->seq, ubuf, cnt); + if (sret != -EBUSY) + goto out; + + trace_seq_init(&iter->seq); + if (iter->trace->read) { sret = iter->trace->read(iter, filp, ubuf, cnt, ppos); if (sret) @@ -5855,9 +5856,6 @@ tracing_buffers_splice_read(struct file *file, loff_t *ppos, return -EBUSY; #endif - if (splice_grow_spd(pipe, &spd)) - return -ENOMEM; - if (*ppos & (PAGE_SIZE - 1)) return -EINVAL; @@ -5867,6 +5865,9 @@ tracing_buffers_splice_read(struct file *file, loff_t *ppos, len &= PAGE_MASK; } + if (splice_grow_spd(pipe, &spd)) + return -ENOMEM; + again: trace_access_lock(iter->cpu_file); entries = ring_buffer_entries_cpu(iter->trace_buffer->buffer, iter->cpu_file); @@ -5924,19 +5925,21 @@ tracing_buffers_splice_read(struct file *file, loff_t *ppos, /* did we read anything? */ if (!spd.nr_pages) { if (ret) - return ret; + goto out; + ret = -EAGAIN; if ((file->f_flags & O_NONBLOCK) || (flags & SPLICE_F_NONBLOCK)) - return -EAGAIN; + goto out; ret = wait_on_pipe(iter, true); if (ret) - return ret; + goto out; goto again; } ret = splice_to_pipe(pipe, &spd); +out: splice_shrink_spd(&spd); return ret; |