aboutsummaryrefslogtreecommitdiffstats
path: root/kernel
diff options
context:
space:
mode:
authorDavid S. Miller <davem@davemloft.net>2015-03-03 21:16:48 -0500
committerDavid S. Miller <davem@davemloft.net>2015-03-03 21:16:48 -0500
commit71a83a6db6138b9d41d8a0b6b91cb59f6dc4742c (patch)
treef74b6e4e48257ec6ce40b95645ecb8533b9cc1f8 /kernel
parentb97526f3ff95f92b107f0fb52cbb8627e395429b (diff)
parenta6c5170d1edea97c538c81e377e56c7b5c5b7e63 (diff)
Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Conflicts: drivers/net/ethernet/rocker/rocker.c The rocker commit was two overlapping changes, one to rename the ->vport member to ->pport, and another making the bitmask expression use '1ULL' instead of plain '1'. Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'kernel')
-rw-r--r--kernel/debug/debug_core.c19
-rw-r--r--kernel/debug/kdb/kdb_io.c46
-rw-r--r--kernel/debug/kdb/kdb_main.c16
-rw-r--r--kernel/debug/kdb/kdb_private.h4
-rw-r--r--kernel/gcov/Makefile36
-rw-r--r--kernel/livepatch/core.c10
-rw-r--r--kernel/locking/rtmutex.c4
-rw-r--r--kernel/printk/printk.c2
-rw-r--r--kernel/rcu/tree_plugin.h1
-rw-r--r--kernel/sched/auto_group.c6
-rw-r--r--kernel/sched/completion.c19
-rw-r--r--kernel/sched/core.c113
-rw-r--r--kernel/sched/deadline.c33
-rw-r--r--kernel/sched/sched.h76
-rw-r--r--kernel/sys.c15
-rw-r--r--kernel/time/ntp.c10
16 files changed, 247 insertions, 163 deletions
diff --git a/kernel/debug/debug_core.c b/kernel/debug/debug_core.c
index 07ce18ca71e0..0874e2edd275 100644
--- a/kernel/debug/debug_core.c
+++ b/kernel/debug/debug_core.c
@@ -604,7 +604,7 @@ return_normal:
604 online_cpus) 604 online_cpus)
605 cpu_relax(); 605 cpu_relax();
606 if (!time_left) 606 if (!time_left)
607 pr_crit("KGDB: Timed out waiting for secondary CPUs.\n"); 607 pr_crit("Timed out waiting for secondary CPUs.\n");
608 608
609 /* 609 /*
610 * At this point the primary processor is completely 610 * At this point the primary processor is completely
@@ -696,6 +696,14 @@ kgdb_handle_exception(int evector, int signo, int ecode, struct pt_regs *regs)
696 696
697 if (arch_kgdb_ops.enable_nmi) 697 if (arch_kgdb_ops.enable_nmi)
698 arch_kgdb_ops.enable_nmi(0); 698 arch_kgdb_ops.enable_nmi(0);
699 /*
700 * Avoid entering the debugger if we were triggered due to an oops
701 * but panic_timeout indicates the system should automatically
702 * reboot on panic. We don't want to get stuck waiting for input
703 * on such systems, especially if its "just" an oops.
704 */
705 if (signo != SIGTRAP && panic_timeout)
706 return 1;
699 707
700 memset(ks, 0, sizeof(struct kgdb_state)); 708 memset(ks, 0, sizeof(struct kgdb_state));
701 ks->cpu = raw_smp_processor_id(); 709 ks->cpu = raw_smp_processor_id();
@@ -828,6 +836,15 @@ static int kgdb_panic_event(struct notifier_block *self,
828 unsigned long val, 836 unsigned long val,
829 void *data) 837 void *data)
830{ 838{
839 /*
840 * Avoid entering the debugger if we were triggered due to a panic
841 * We don't want to get stuck waiting for input from user in such case.
842 * panic_timeout indicates the system should automatically
843 * reboot on panic.
844 */
845 if (panic_timeout)
846 return NOTIFY_DONE;
847
831 if (dbg_kdb_mode) 848 if (dbg_kdb_mode)
832 kdb_printf("PANIC: %s\n", (char *)data); 849 kdb_printf("PANIC: %s\n", (char *)data);
833 kgdb_breakpoint(); 850 kgdb_breakpoint();
diff --git a/kernel/debug/kdb/kdb_io.c b/kernel/debug/kdb/kdb_io.c
index 7c70812caea5..fc1ef736253c 100644
--- a/kernel/debug/kdb/kdb_io.c
+++ b/kernel/debug/kdb/kdb_io.c
@@ -439,7 +439,7 @@ poll_again:
439 * substituted for %d, %x or %o in the prompt. 439 * substituted for %d, %x or %o in the prompt.
440 */ 440 */
441 441
442char *kdb_getstr(char *buffer, size_t bufsize, char *prompt) 442char *kdb_getstr(char *buffer, size_t bufsize, const char *prompt)
443{ 443{
444 if (prompt && kdb_prompt_str != prompt) 444 if (prompt && kdb_prompt_str != prompt)
445 strncpy(kdb_prompt_str, prompt, CMD_BUFLEN); 445 strncpy(kdb_prompt_str, prompt, CMD_BUFLEN);
@@ -548,7 +548,7 @@ static int kdb_search_string(char *searched, char *searchfor)
548 return 0; 548 return 0;
549} 549}
550 550
551int vkdb_printf(const char *fmt, va_list ap) 551int vkdb_printf(enum kdb_msgsrc src, const char *fmt, va_list ap)
552{ 552{
553 int diag; 553 int diag;
554 int linecount; 554 int linecount;
@@ -680,6 +680,12 @@ int vkdb_printf(const char *fmt, va_list ap)
680 size_avail = sizeof(kdb_buffer) - len; 680 size_avail = sizeof(kdb_buffer) - len;
681 goto kdb_print_out; 681 goto kdb_print_out;
682 } 682 }
683 if (kdb_grepping_flag >= KDB_GREPPING_FLAG_SEARCH)
684 /*
685 * This was a interactive search (using '/' at more
686 * prompt) and it has completed. Clear the flag.
687 */
688 kdb_grepping_flag = 0;
683 /* 689 /*
684 * at this point the string is a full line and 690 * at this point the string is a full line and
685 * should be printed, up to the null. 691 * should be printed, up to the null.
@@ -691,19 +697,20 @@ kdb_printit:
691 * Write to all consoles. 697 * Write to all consoles.
692 */ 698 */
693 retlen = strlen(kdb_buffer); 699 retlen = strlen(kdb_buffer);
700 cp = (char *) printk_skip_level(kdb_buffer);
694 if (!dbg_kdb_mode && kgdb_connected) { 701 if (!dbg_kdb_mode && kgdb_connected) {
695 gdbstub_msg_write(kdb_buffer, retlen); 702 gdbstub_msg_write(cp, retlen - (cp - kdb_buffer));
696 } else { 703 } else {
697 if (dbg_io_ops && !dbg_io_ops->is_console) { 704 if (dbg_io_ops && !dbg_io_ops->is_console) {
698 len = retlen; 705 len = retlen - (cp - kdb_buffer);
699 cp = kdb_buffer; 706 cp2 = cp;
700 while (len--) { 707 while (len--) {
701 dbg_io_ops->write_char(*cp); 708 dbg_io_ops->write_char(*cp2);
702 cp++; 709 cp2++;
703 } 710 }
704 } 711 }
705 while (c) { 712 while (c) {
706 c->write(c, kdb_buffer, retlen); 713 c->write(c, cp, retlen - (cp - kdb_buffer));
707 touch_nmi_watchdog(); 714 touch_nmi_watchdog();
708 c = c->next; 715 c = c->next;
709 } 716 }
@@ -711,7 +718,10 @@ kdb_printit:
711 if (logging) { 718 if (logging) {
712 saved_loglevel = console_loglevel; 719 saved_loglevel = console_loglevel;
713 console_loglevel = CONSOLE_LOGLEVEL_SILENT; 720 console_loglevel = CONSOLE_LOGLEVEL_SILENT;
714 printk(KERN_INFO "%s", kdb_buffer); 721 if (printk_get_level(kdb_buffer) || src == KDB_MSGSRC_PRINTK)
722 printk("%s", kdb_buffer);
723 else
724 pr_info("%s", kdb_buffer);
715 } 725 }
716 726
717 if (KDB_STATE(PAGER)) { 727 if (KDB_STATE(PAGER)) {
@@ -794,11 +804,23 @@ kdb_printit:
794 kdb_nextline = linecount - 1; 804 kdb_nextline = linecount - 1;
795 kdb_printf("\r"); 805 kdb_printf("\r");
796 suspend_grep = 1; /* for this recursion */ 806 suspend_grep = 1; /* for this recursion */
807 } else if (buf1[0] == '/' && !kdb_grepping_flag) {
808 kdb_printf("\r");
809 kdb_getstr(kdb_grep_string, KDB_GREP_STRLEN,
810 kdbgetenv("SEARCHPROMPT") ?: "search> ");
811 *strchrnul(kdb_grep_string, '\n') = '\0';
812 kdb_grepping_flag += KDB_GREPPING_FLAG_SEARCH;
813 suspend_grep = 1; /* for this recursion */
797 } else if (buf1[0] && buf1[0] != '\n') { 814 } else if (buf1[0] && buf1[0] != '\n') {
798 /* user hit something other than enter */ 815 /* user hit something other than enter */
799 suspend_grep = 1; /* for this recursion */ 816 suspend_grep = 1; /* for this recursion */
800 kdb_printf("\nOnly 'q' or 'Q' are processed at more " 817 if (buf1[0] != '/')
801 "prompt, input ignored\n"); 818 kdb_printf(
819 "\nOnly 'q', 'Q' or '/' are processed at "
820 "more prompt, input ignored\n");
821 else
822 kdb_printf("\n'/' cannot be used during | "
823 "grep filtering, input ignored\n");
802 } else if (kdb_grepping_flag) { 824 } else if (kdb_grepping_flag) {
803 /* user hit enter */ 825 /* user hit enter */
804 suspend_grep = 1; /* for this recursion */ 826 suspend_grep = 1; /* for this recursion */
@@ -844,7 +866,7 @@ int kdb_printf(const char *fmt, ...)
844 int r; 866 int r;
845 867
846 va_start(ap, fmt); 868 va_start(ap, fmt);
847 r = vkdb_printf(fmt, ap); 869 r = vkdb_printf(KDB_MSGSRC_INTERNAL, fmt, ap);
848 va_end(ap); 870 va_end(ap);
849 871
850 return r; 872 return r;
diff --git a/kernel/debug/kdb/kdb_main.c b/kernel/debug/kdb/kdb_main.c
index 7b40c5f07dce..4121345498e0 100644
--- a/kernel/debug/kdb/kdb_main.c
+++ b/kernel/debug/kdb/kdb_main.c
@@ -50,8 +50,7 @@
50static int kdb_cmd_enabled = CONFIG_KDB_DEFAULT_ENABLE; 50static int kdb_cmd_enabled = CONFIG_KDB_DEFAULT_ENABLE;
51module_param_named(cmd_enable, kdb_cmd_enabled, int, 0600); 51module_param_named(cmd_enable, kdb_cmd_enabled, int, 0600);
52 52
53#define GREP_LEN 256 53char kdb_grep_string[KDB_GREP_STRLEN];
54char kdb_grep_string[GREP_LEN];
55int kdb_grepping_flag; 54int kdb_grepping_flag;
56EXPORT_SYMBOL(kdb_grepping_flag); 55EXPORT_SYMBOL(kdb_grepping_flag);
57int kdb_grep_leading; 56int kdb_grep_leading;
@@ -870,7 +869,7 @@ static void parse_grep(const char *str)
870 len = strlen(cp); 869 len = strlen(cp);
871 if (!len) 870 if (!len)
872 return; 871 return;
873 if (len >= GREP_LEN) { 872 if (len >= KDB_GREP_STRLEN) {
874 kdb_printf("search string too long\n"); 873 kdb_printf("search string too long\n");
875 return; 874 return;
876 } 875 }
@@ -915,13 +914,12 @@ int kdb_parse(const char *cmdstr)
915 char *cp; 914 char *cp;
916 char *cpp, quoted; 915 char *cpp, quoted;
917 kdbtab_t *tp; 916 kdbtab_t *tp;
918 int i, escaped, ignore_errors = 0, check_grep; 917 int i, escaped, ignore_errors = 0, check_grep = 0;
919 918
920 /* 919 /*
921 * First tokenize the command string. 920 * First tokenize the command string.
922 */ 921 */
923 cp = (char *)cmdstr; 922 cp = (char *)cmdstr;
924 kdb_grepping_flag = check_grep = 0;
925 923
926 if (KDB_FLAG(CMD_INTERRUPT)) { 924 if (KDB_FLAG(CMD_INTERRUPT)) {
927 /* Previous command was interrupted, newline must not 925 /* Previous command was interrupted, newline must not
@@ -1247,7 +1245,6 @@ static int kdb_local(kdb_reason_t reason, int error, struct pt_regs *regs,
1247 kdb_printf("due to NonMaskable Interrupt @ " 1245 kdb_printf("due to NonMaskable Interrupt @ "
1248 kdb_machreg_fmt "\n", 1246 kdb_machreg_fmt "\n",
1249 instruction_pointer(regs)); 1247 instruction_pointer(regs));
1250 kdb_dumpregs(regs);
1251 break; 1248 break;
1252 case KDB_REASON_SSTEP: 1249 case KDB_REASON_SSTEP:
1253 case KDB_REASON_BREAK: 1250 case KDB_REASON_BREAK:
@@ -1281,6 +1278,9 @@ static int kdb_local(kdb_reason_t reason, int error, struct pt_regs *regs,
1281 */ 1278 */
1282 kdb_nextline = 1; 1279 kdb_nextline = 1;
1283 KDB_STATE_CLEAR(SUPPRESS); 1280 KDB_STATE_CLEAR(SUPPRESS);
1281 kdb_grepping_flag = 0;
1282 /* ensure the old search does not leak into '/' commands */
1283 kdb_grep_string[0] = '\0';
1284 1284
1285 cmdbuf = cmd_cur; 1285 cmdbuf = cmd_cur;
1286 *cmdbuf = '\0'; 1286 *cmdbuf = '\0';
@@ -2256,7 +2256,7 @@ static int kdb_cpu(int argc, const char **argv)
2256 /* 2256 /*
2257 * Validate cpunum 2257 * Validate cpunum
2258 */ 2258 */
2259 if ((cpunum > NR_CPUS) || !kgdb_info[cpunum].enter_kgdb) 2259 if ((cpunum >= CONFIG_NR_CPUS) || !kgdb_info[cpunum].enter_kgdb)
2260 return KDB_BADCPUNUM; 2260 return KDB_BADCPUNUM;
2261 2261
2262 dbg_switch_cpu = cpunum; 2262 dbg_switch_cpu = cpunum;
@@ -2583,7 +2583,7 @@ static int kdb_summary(int argc, const char **argv)
2583#define K(x) ((x) << (PAGE_SHIFT - 10)) 2583#define K(x) ((x) << (PAGE_SHIFT - 10))
2584 kdb_printf("\nMemTotal: %8lu kB\nMemFree: %8lu kB\n" 2584 kdb_printf("\nMemTotal: %8lu kB\nMemFree: %8lu kB\n"
2585 "Buffers: %8lu kB\n", 2585 "Buffers: %8lu kB\n",
2586 val.totalram, val.freeram, val.bufferram); 2586 K(val.totalram), K(val.freeram), K(val.bufferram));
2587 return 0; 2587 return 0;
2588} 2588}
2589 2589
diff --git a/kernel/debug/kdb/kdb_private.h b/kernel/debug/kdb/kdb_private.h
index eaacd1693954..75014d7f4568 100644
--- a/kernel/debug/kdb/kdb_private.h
+++ b/kernel/debug/kdb/kdb_private.h
@@ -196,7 +196,9 @@ extern int kdb_main_loop(kdb_reason_t, kdb_reason_t,
196 196
197/* Miscellaneous functions and data areas */ 197/* Miscellaneous functions and data areas */
198extern int kdb_grepping_flag; 198extern int kdb_grepping_flag;
199#define KDB_GREPPING_FLAG_SEARCH 0x8000
199extern char kdb_grep_string[]; 200extern char kdb_grep_string[];
201#define KDB_GREP_STRLEN 256
200extern int kdb_grep_leading; 202extern int kdb_grep_leading;
201extern int kdb_grep_trailing; 203extern int kdb_grep_trailing;
202extern char *kdb_cmds[]; 204extern char *kdb_cmds[];
@@ -209,7 +211,7 @@ extern void kdb_ps1(const struct task_struct *p);
209extern void kdb_print_nameval(const char *name, unsigned long val); 211extern void kdb_print_nameval(const char *name, unsigned long val);
210extern void kdb_send_sig_info(struct task_struct *p, struct siginfo *info); 212extern void kdb_send_sig_info(struct task_struct *p, struct siginfo *info);
211extern void kdb_meminfo_proc_show(void); 213extern void kdb_meminfo_proc_show(void);
212extern char *kdb_getstr(char *, size_t, char *); 214extern char *kdb_getstr(char *, size_t, const char *);
213extern void kdb_gdb_state_pass(char *buf); 215extern void kdb_gdb_state_pass(char *buf);
214 216
215/* Defines for kdb_symbol_print */ 217/* Defines for kdb_symbol_print */
diff --git a/kernel/gcov/Makefile b/kernel/gcov/Makefile
index 52aa7e8de927..752d6486b67e 100644
--- a/kernel/gcov/Makefile
+++ b/kernel/gcov/Makefile
@@ -1,33 +1,7 @@
1ccflags-y := -DSRCTREE='"$(srctree)"' -DOBJTREE='"$(objtree)"' 1ccflags-y := -DSRCTREE='"$(srctree)"' -DOBJTREE='"$(objtree)"'
2 2
3# if-lt 3obj-y := base.o fs.o
4# Usage VAR := $(call if-lt, $(a), $(b)) 4obj-$(CONFIG_GCOV_FORMAT_3_4) += gcc_3_4.o
5# Returns 1 if (a < b) 5obj-$(CONFIG_GCOV_FORMAT_4_7) += gcc_4_7.o
6if-lt = $(shell [ $(1) -lt $(2) ] && echo 1) 6obj-$(CONFIG_GCOV_FORMAT_AUTODETECT) += $(call cc-ifversion, -lt, 0407, \
7 7 gcc_3_4.o, gcc_4_7.o)
8ifeq ($(CONFIG_GCOV_FORMAT_3_4),y)
9 cc-ver := 0304
10else ifeq ($(CONFIG_GCOV_FORMAT_4_7),y)
11 cc-ver := 0407
12else
13# Use cc-version if available, otherwise set 0
14#
15# scripts/Kbuild.include, which contains cc-version function, is not included
16# during make clean "make -f scripts/Makefile.clean obj=kernel/gcov"
17# Meaning cc-ver is empty causing if-lt test to fail with
18# "/bin/sh: line 0: [: -lt: unary operator expected" error mesage.
19# This has no affect on the clean phase, but the error message could be
20# confusing/annoying. So this dummy workaround sets cc-ver to zero if cc-version
21# is not available. We can probably move if-lt to Kbuild.include, so it's also
22# not defined during clean or to include Kbuild.include in
23# scripts/Makefile.clean. But the following workaround seems least invasive.
24 cc-ver := $(if $(call cc-version),$(call cc-version),0)
25endif
26
27obj-$(CONFIG_GCOV_KERNEL) := base.o fs.o
28
29ifeq ($(call if-lt, $(cc-ver), 0407),1)
30 obj-$(CONFIG_GCOV_KERNEL) += gcc_3_4.o
31else
32 obj-$(CONFIG_GCOV_KERNEL) += gcc_4_7.o
33endif
diff --git a/kernel/livepatch/core.c b/kernel/livepatch/core.c
index ff7f47d026ac..782172f073c5 100644
--- a/kernel/livepatch/core.c
+++ b/kernel/livepatch/core.c
@@ -314,12 +314,12 @@ static void notrace klp_ftrace_handler(unsigned long ip,
314 rcu_read_lock(); 314 rcu_read_lock();
315 func = list_first_or_null_rcu(&ops->func_stack, struct klp_func, 315 func = list_first_or_null_rcu(&ops->func_stack, struct klp_func,
316 stack_node); 316 stack_node);
317 rcu_read_unlock();
318
319 if (WARN_ON_ONCE(!func)) 317 if (WARN_ON_ONCE(!func))
320 return; 318 goto unlock;
321 319
322 klp_arch_set_pc(regs, (unsigned long)func->new_func); 320 klp_arch_set_pc(regs, (unsigned long)func->new_func);
321unlock:
322 rcu_read_unlock();
323} 323}
324 324
325static int klp_disable_func(struct klp_func *func) 325static int klp_disable_func(struct klp_func *func)
@@ -731,7 +731,7 @@ static int klp_init_func(struct klp_object *obj, struct klp_func *func)
731 func->state = KLP_DISABLED; 731 func->state = KLP_DISABLED;
732 732
733 return kobject_init_and_add(&func->kobj, &klp_ktype_func, 733 return kobject_init_and_add(&func->kobj, &klp_ktype_func,
734 obj->kobj, func->old_name); 734 obj->kobj, "%s", func->old_name);
735} 735}
736 736
737/* parts of the initialization that is done only when the object is loaded */ 737/* parts of the initialization that is done only when the object is loaded */
@@ -807,7 +807,7 @@ static int klp_init_patch(struct klp_patch *patch)
807 patch->state = KLP_DISABLED; 807 patch->state = KLP_DISABLED;
808 808
809 ret = kobject_init_and_add(&patch->kobj, &klp_ktype_patch, 809 ret = kobject_init_and_add(&patch->kobj, &klp_ktype_patch,
810 klp_root_kobj, patch->mod->name); 810 klp_root_kobj, "%s", patch->mod->name);
811 if (ret) 811 if (ret)
812 goto unlock; 812 goto unlock;
813 813
diff --git a/kernel/locking/rtmutex.c b/kernel/locking/rtmutex.c
index 3059bc2f022d..6357265a31ad 100644
--- a/kernel/locking/rtmutex.c
+++ b/kernel/locking/rtmutex.c
@@ -1193,7 +1193,9 @@ rt_mutex_slowlock(struct rt_mutex *lock, int state,
1193 ret = __rt_mutex_slowlock(lock, state, timeout, &waiter); 1193 ret = __rt_mutex_slowlock(lock, state, timeout, &waiter);
1194 1194
1195 if (unlikely(ret)) { 1195 if (unlikely(ret)) {
1196 remove_waiter(lock, &waiter); 1196 __set_current_state(TASK_RUNNING);
1197 if (rt_mutex_has_waiters(lock))
1198 remove_waiter(lock, &waiter);
1197 rt_mutex_handle_deadlock(ret, chwalk, &waiter); 1199 rt_mutex_handle_deadlock(ret, chwalk, &waiter);
1198 } 1200 }
1199 1201
diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index c06df7de0963..01cfd69c54c6 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -1811,7 +1811,7 @@ int vprintk_default(const char *fmt, va_list args)
1811 1811
1812#ifdef CONFIG_KGDB_KDB 1812#ifdef CONFIG_KGDB_KDB
1813 if (unlikely(kdb_trap_printk)) { 1813 if (unlikely(kdb_trap_printk)) {
1814 r = vkdb_printf(fmt, args); 1814 r = vkdb_printf(KDB_MSGSRC_PRINTK, fmt, args);
1815 return r; 1815 return r;
1816 } 1816 }
1817#endif 1817#endif
diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h
index 0d7bbe3095ad..0a571e9a0f1d 100644
--- a/kernel/rcu/tree_plugin.h
+++ b/kernel/rcu/tree_plugin.h
@@ -326,6 +326,7 @@ void rcu_read_unlock_special(struct task_struct *t)
326 special = t->rcu_read_unlock_special; 326 special = t->rcu_read_unlock_special;
327 if (special.b.need_qs) { 327 if (special.b.need_qs) {
328 rcu_preempt_qs(); 328 rcu_preempt_qs();
329 t->rcu_read_unlock_special.b.need_qs = false;
329 if (!t->rcu_read_unlock_special.s) { 330 if (!t->rcu_read_unlock_special.s) {
330 local_irq_restore(flags); 331 local_irq_restore(flags);
331 return; 332 return;
diff --git a/kernel/sched/auto_group.c b/kernel/sched/auto_group.c
index 8a2e230fb86a..eae160dd669d 100644
--- a/kernel/sched/auto_group.c
+++ b/kernel/sched/auto_group.c
@@ -87,8 +87,7 @@ static inline struct autogroup *autogroup_create(void)
87 * so we don't have to move tasks around upon policy change, 87 * so we don't have to move tasks around upon policy change,
88 * or flail around trying to allocate bandwidth on the fly. 88 * or flail around trying to allocate bandwidth on the fly.
89 * A bandwidth exception in __sched_setscheduler() allows 89 * A bandwidth exception in __sched_setscheduler() allows
90 * the policy change to proceed. Thereafter, task_group() 90 * the policy change to proceed.
91 * returns &root_task_group, so zero bandwidth is required.
92 */ 91 */
93 free_rt_sched_group(tg); 92 free_rt_sched_group(tg);
94 tg->rt_se = root_task_group.rt_se; 93 tg->rt_se = root_task_group.rt_se;
@@ -115,9 +114,6 @@ bool task_wants_autogroup(struct task_struct *p, struct task_group *tg)
115 if (tg != &root_task_group) 114 if (tg != &root_task_group)
116 return false; 115 return false;
117 116
118 if (p->sched_class != &fair_sched_class)
119 return false;
120
121 /* 117 /*
122 * We can only assume the task group can't go away on us if 118 * We can only assume the task group can't go away on us if
123 * autogroup_move_group() can see us on ->thread_group list. 119 * autogroup_move_group() can see us on ->thread_group list.
diff --git a/kernel/sched/completion.c b/kernel/sched/completion.c
index 7052d3fd4e7b..8d0f35debf35 100644
--- a/kernel/sched/completion.c
+++ b/kernel/sched/completion.c
@@ -274,7 +274,7 @@ bool try_wait_for_completion(struct completion *x)
274 * first without taking the lock so we can 274 * first without taking the lock so we can
275 * return early in the blocking case. 275 * return early in the blocking case.
276 */ 276 */
277 if (!ACCESS_ONCE(x->done)) 277 if (!READ_ONCE(x->done))
278 return 0; 278 return 0;
279 279
280 spin_lock_irqsave(&x->wait.lock, flags); 280 spin_lock_irqsave(&x->wait.lock, flags);
@@ -297,6 +297,21 @@ EXPORT_SYMBOL(try_wait_for_completion);
297 */ 297 */
298bool completion_done(struct completion *x) 298bool completion_done(struct completion *x)
299{ 299{
300 return !!ACCESS_ONCE(x->done); 300 if (!READ_ONCE(x->done))
301 return false;
302
303 /*
304 * If ->done, we need to wait for complete() to release ->wait.lock
305 * otherwise we can end up freeing the completion before complete()
306 * is done referencing it.
307 *
308 * The RMB pairs with complete()'s RELEASE of ->wait.lock and orders
309 * the loads of ->done and ->wait.lock such that we cannot observe
310 * the lock before complete() acquires it while observing the ->done
311 * after it's acquired the lock.
312 */
313 smp_rmb();
314 spin_unlock_wait(&x->wait.lock);
315 return true;
301} 316}
302EXPORT_SYMBOL(completion_done); 317EXPORT_SYMBOL(completion_done);
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 13049aac05a6..f0f831e8a345 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -307,66 +307,6 @@ __read_mostly int scheduler_running;
307int sysctl_sched_rt_runtime = 950000; 307int sysctl_sched_rt_runtime = 950000;
308 308
309/* 309/*
310 * __task_rq_lock - lock the rq @p resides on.
311 */
312static inline struct rq *__task_rq_lock(struct task_struct *p)
313 __acquires(rq->lock)
314{
315 struct rq *rq;
316
317 lockdep_assert_held(&p->pi_lock);
318
319 for (;;) {
320 rq = task_rq(p);
321 raw_spin_lock(&rq->lock);
322 if (likely(rq == task_rq(p) && !task_on_rq_migrating(p)))
323 return rq;
324 raw_spin_unlock(&rq->lock);
325
326 while (unlikely(task_on_rq_migrating(p)))
327 cpu_relax();
328 }
329}
330
331/*
332 * task_rq_lock - lock p->pi_lock and lock the rq @p resides on.
333 */
334static struct rq *task_rq_lock(struct task_struct *p, unsigned long *flags)
335 __acquires(p->pi_lock)
336 __acquires(rq->lock)
337{
338 struct rq *rq;
339
340 for (;;) {
341 raw_spin_lock_irqsave(&p->pi_lock, *flags);
342 rq = task_rq(p);
343 raw_spin_lock(&rq->lock);
344 if (likely(rq == task_rq(p) && !task_on_rq_migrating(p)))
345 return rq;
346 raw_spin_unlock(&rq->lock);
347 raw_spin_unlock_irqrestore(&p->pi_lock, *flags);
348
349 while (unlikely(task_on_rq_migrating(p)))
350 cpu_relax();
351 }
352}
353
354static void __task_rq_unlock(struct rq *rq)
355 __releases(rq->lock)
356{
357 raw_spin_unlock(&rq->lock);
358}
359
360static inline void
361task_rq_unlock(struct rq *rq, struct task_struct *p, unsigned long *flags)
362 __releases(rq->lock)
363 __releases(p->pi_lock)
364{
365 raw_spin_unlock(&rq->lock);
366 raw_spin_unlock_irqrestore(&p->pi_lock, *flags);
367}
368
369/*
370 * this_rq_lock - lock this runqueue and disable interrupts. 310 * this_rq_lock - lock this runqueue and disable interrupts.
371 */ 311 */
372static struct rq *this_rq_lock(void) 312static struct rq *this_rq_lock(void)
@@ -2899,7 +2839,7 @@ void __sched schedule_preempt_disabled(void)
2899 preempt_disable(); 2839 preempt_disable();
2900} 2840}
2901 2841
2902static void preempt_schedule_common(void) 2842static void __sched notrace preempt_schedule_common(void)
2903{ 2843{
2904 do { 2844 do {
2905 __preempt_count_add(PREEMPT_ACTIVE); 2845 __preempt_count_add(PREEMPT_ACTIVE);
@@ -4418,36 +4358,29 @@ EXPORT_SYMBOL_GPL(yield_to);
4418 * This task is about to go to sleep on IO. Increment rq->nr_iowait so 4358 * This task is about to go to sleep on IO. Increment rq->nr_iowait so
4419 * that process accounting knows that this is a task in IO wait state. 4359 * that process accounting knows that this is a task in IO wait state.
4420 */ 4360 */
4421void __sched io_schedule(void)
4422{
4423 struct rq *rq = raw_rq();
4424
4425 delayacct_blkio_start();
4426 atomic_inc(&rq->nr_iowait);
4427 blk_flush_plug(current);
4428 current->in_iowait = 1;
4429 schedule();
4430 current->in_iowait = 0;
4431 atomic_dec(&rq->nr_iowait);
4432 delayacct_blkio_end();
4433}
4434EXPORT_SYMBOL(io_schedule);
4435
4436long __sched io_schedule_timeout(long timeout) 4361long __sched io_schedule_timeout(long timeout)
4437{ 4362{
4438 struct rq *rq = raw_rq(); 4363 int old_iowait = current->in_iowait;
4364 struct rq *rq;
4439 long ret; 4365 long ret;
4440 4366
4367 current->in_iowait = 1;
4368 if (old_iowait)
4369 blk_schedule_flush_plug(current);
4370 else
4371 blk_flush_plug(current);
4372
4441 delayacct_blkio_start(); 4373 delayacct_blkio_start();
4374 rq = raw_rq();
4442 atomic_inc(&rq->nr_iowait); 4375 atomic_inc(&rq->nr_iowait);
4443 blk_flush_plug(current);
4444 current->in_iowait = 1;
4445 ret = schedule_timeout(timeout); 4376 ret = schedule_timeout(timeout);
4446 current->in_iowait = 0; 4377 current->in_iowait = old_iowait;
4447 atomic_dec(&rq->nr_iowait); 4378 atomic_dec(&rq->nr_iowait);
4448 delayacct_blkio_end(); 4379 delayacct_blkio_end();
4380
4449 return ret; 4381 return ret;
4450} 4382}
4383EXPORT_SYMBOL(io_schedule_timeout);
4451 4384
4452/** 4385/**
4453 * sys_sched_get_priority_max - return maximum RT priority. 4386 * sys_sched_get_priority_max - return maximum RT priority.
@@ -7642,6 +7575,12 @@ static inline int tg_has_rt_tasks(struct task_group *tg)
7642{ 7575{
7643 struct task_struct *g, *p; 7576 struct task_struct *g, *p;
7644 7577
7578 /*
7579 * Autogroups do not have RT tasks; see autogroup_create().
7580 */
7581 if (task_group_is_autogroup(tg))
7582 return 0;
7583
7645 for_each_process_thread(g, p) { 7584 for_each_process_thread(g, p) {
7646 if (rt_task(p) && task_group(p) == tg) 7585 if (rt_task(p) && task_group(p) == tg)
7647 return 1; 7586 return 1;
@@ -7734,6 +7673,17 @@ static int tg_set_rt_bandwidth(struct task_group *tg,
7734{ 7673{
7735 int i, err = 0; 7674 int i, err = 0;
7736 7675
7676 /*
7677 * Disallowing the root group RT runtime is BAD, it would disallow the
7678 * kernel creating (and or operating) RT threads.
7679 */
7680 if (tg == &root_task_group && rt_runtime == 0)
7681 return -EINVAL;
7682
7683 /* No period doesn't make any sense. */
7684 if (rt_period == 0)
7685 return -EINVAL;
7686
7737 mutex_lock(&rt_constraints_mutex); 7687 mutex_lock(&rt_constraints_mutex);
7738 read_lock(&tasklist_lock); 7688 read_lock(&tasklist_lock);
7739 err = __rt_schedulable(tg, rt_period, rt_runtime); 7689 err = __rt_schedulable(tg, rt_period, rt_runtime);
@@ -7790,9 +7740,6 @@ static int sched_group_set_rt_period(struct task_group *tg, long rt_period_us)
7790 rt_period = (u64)rt_period_us * NSEC_PER_USEC; 7740 rt_period = (u64)rt_period_us * NSEC_PER_USEC;
7791 rt_runtime = tg->rt_bandwidth.rt_runtime; 7741 rt_runtime = tg->rt_bandwidth.rt_runtime;
7792 7742
7793 if (rt_period == 0)
7794 return -EINVAL;
7795
7796 return tg_set_rt_bandwidth(tg, rt_period, rt_runtime); 7743 return tg_set_rt_bandwidth(tg, rt_period, rt_runtime);
7797} 7744}
7798 7745
diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
index a027799ae130..3fa8fa6d9403 100644
--- a/kernel/sched/deadline.c
+++ b/kernel/sched/deadline.c
@@ -511,16 +511,10 @@ static enum hrtimer_restart dl_task_timer(struct hrtimer *timer)
511 struct sched_dl_entity, 511 struct sched_dl_entity,
512 dl_timer); 512 dl_timer);
513 struct task_struct *p = dl_task_of(dl_se); 513 struct task_struct *p = dl_task_of(dl_se);
514 unsigned long flags;
514 struct rq *rq; 515 struct rq *rq;
515again:
516 rq = task_rq(p);
517 raw_spin_lock(&rq->lock);
518 516
519 if (rq != task_rq(p)) { 517 rq = task_rq_lock(current, &flags);
520 /* Task was moved, retrying. */
521 raw_spin_unlock(&rq->lock);
522 goto again;
523 }
524 518
525 /* 519 /*
526 * We need to take care of several possible races here: 520 * We need to take care of several possible races here:
@@ -541,6 +535,26 @@ again:
541 535
542 sched_clock_tick(); 536 sched_clock_tick();
543 update_rq_clock(rq); 537 update_rq_clock(rq);
538
539 /*
540 * If the throttle happened during sched-out; like:
541 *
542 * schedule()
543 * deactivate_task()
544 * dequeue_task_dl()
545 * update_curr_dl()
546 * start_dl_timer()
547 * __dequeue_task_dl()
548 * prev->on_rq = 0;
549 *
550 * We can be both throttled and !queued. Replenish the counter
551 * but do not enqueue -- wait for our wakeup to do that.
552 */
553 if (!task_on_rq_queued(p)) {
554 replenish_dl_entity(dl_se, dl_se);
555 goto unlock;
556 }
557
544 enqueue_task_dl(rq, p, ENQUEUE_REPLENISH); 558 enqueue_task_dl(rq, p, ENQUEUE_REPLENISH);
545 if (dl_task(rq->curr)) 559 if (dl_task(rq->curr))
546 check_preempt_curr_dl(rq, p, 0); 560 check_preempt_curr_dl(rq, p, 0);
@@ -555,7 +569,7 @@ again:
555 push_dl_task(rq); 569 push_dl_task(rq);
556#endif 570#endif
557unlock: 571unlock:
558 raw_spin_unlock(&rq->lock); 572 task_rq_unlock(rq, current, &flags);
559 573
560 return HRTIMER_NORESTART; 574 return HRTIMER_NORESTART;
561} 575}
@@ -898,6 +912,7 @@ static void yield_task_dl(struct rq *rq)
898 rq->curr->dl.dl_yielded = 1; 912 rq->curr->dl.dl_yielded = 1;
899 p->dl.runtime = 0; 913 p->dl.runtime = 0;
900 } 914 }
915 update_rq_clock(rq);
901 update_curr_dl(rq); 916 update_curr_dl(rq);
902} 917}
903 918
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 0870db23d79c..dc0f435a2779 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -1380,6 +1380,82 @@ static inline void sched_avg_update(struct rq *rq) { }
1380 1380
1381extern void start_bandwidth_timer(struct hrtimer *period_timer, ktime_t period); 1381extern void start_bandwidth_timer(struct hrtimer *period_timer, ktime_t period);
1382 1382
1383/*
1384 * __task_rq_lock - lock the rq @p resides on.
1385 */
1386static inline struct rq *__task_rq_lock(struct task_struct *p)
1387 __acquires(rq->lock)
1388{
1389 struct rq *rq;
1390
1391 lockdep_assert_held(&p->pi_lock);
1392
1393 for (;;) {
1394 rq = task_rq(p);
1395 raw_spin_lock(&rq->lock);
1396 if (likely(rq == task_rq(p) && !task_on_rq_migrating(p)))
1397 return rq;
1398 raw_spin_unlock(&rq->lock);
1399
1400 while (unlikely(task_on_rq_migrating(p)))
1401 cpu_relax();
1402 }
1403}
1404
1405/*
1406 * task_rq_lock - lock p->pi_lock and lock the rq @p resides on.
1407 */
1408static inline struct rq *task_rq_lock(struct task_struct *p, unsigned long *flags)
1409 __acquires(p->pi_lock)
1410 __acquires(rq->lock)
1411{
1412 struct rq *rq;
1413
1414 for (;;) {
1415 raw_spin_lock_irqsave(&p->pi_lock, *flags);
1416 rq = task_rq(p);
1417 raw_spin_lock(&rq->lock);
1418 /*
1419 * move_queued_task() task_rq_lock()
1420 *
1421 * ACQUIRE (rq->lock)
1422 * [S] ->on_rq = MIGRATING [L] rq = task_rq()
1423 * WMB (__set_task_cpu()) ACQUIRE (rq->lock);
1424 * [S] ->cpu = new_cpu [L] task_rq()
1425 * [L] ->on_rq
1426 * RELEASE (rq->lock)
1427 *
1428 * If we observe the old cpu in task_rq_lock, the acquire of
1429 * the old rq->lock will fully serialize against the stores.
1430 *
1431 * If we observe the new cpu in task_rq_lock, the acquire will
1432 * pair with the WMB to ensure we must then also see migrating.
1433 */
1434 if (likely(rq == task_rq(p) && !task_on_rq_migrating(p)))
1435 return rq;
1436 raw_spin_unlock(&rq->lock);
1437 raw_spin_unlock_irqrestore(&p->pi_lock, *flags);
1438
1439 while (unlikely(task_on_rq_migrating(p)))
1440 cpu_relax();
1441 }
1442}
1443
1444static inline void __task_rq_unlock(struct rq *rq)
1445 __releases(rq->lock)
1446{
1447 raw_spin_unlock(&rq->lock);
1448}
1449
1450static inline void
1451task_rq_unlock(struct rq *rq, struct task_struct *p, unsigned long *flags)
1452 __releases(rq->lock)
1453 __releases(p->pi_lock)
1454{
1455 raw_spin_unlock(&rq->lock);
1456 raw_spin_unlock_irqrestore(&p->pi_lock, *flags);
1457}
1458
1383#ifdef CONFIG_SMP 1459#ifdef CONFIG_SMP
1384#ifdef CONFIG_PREEMPT 1460#ifdef CONFIG_PREEMPT
1385 1461
diff --git a/kernel/sys.c b/kernel/sys.c
index ea9c88109894..a03d9cd23ed7 100644
--- a/kernel/sys.c
+++ b/kernel/sys.c
@@ -97,6 +97,12 @@
97#ifndef MPX_DISABLE_MANAGEMENT 97#ifndef MPX_DISABLE_MANAGEMENT
98# define MPX_DISABLE_MANAGEMENT(a) (-EINVAL) 98# define MPX_DISABLE_MANAGEMENT(a) (-EINVAL)
99#endif 99#endif
100#ifndef GET_FP_MODE
101# define GET_FP_MODE(a) (-EINVAL)
102#endif
103#ifndef SET_FP_MODE
104# define SET_FP_MODE(a,b) (-EINVAL)
105#endif
100 106
101/* 107/*
102 * this is where the system-wide overflow UID and GID are defined, for 108 * this is where the system-wide overflow UID and GID are defined, for
@@ -1102,6 +1108,7 @@ DECLARE_RWSEM(uts_sem);
1102/* 1108/*
1103 * Work around broken programs that cannot handle "Linux 3.0". 1109 * Work around broken programs that cannot handle "Linux 3.0".
1104 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40 1110 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1111 * And we map 4.x to 2.6.60+x, so 4.0 would be 2.6.60.
1105 */ 1112 */
1106static int override_release(char __user *release, size_t len) 1113static int override_release(char __user *release, size_t len)
1107{ 1114{
@@ -1121,7 +1128,7 @@ static int override_release(char __user *release, size_t len)
1121 break; 1128 break;
1122 rest++; 1129 rest++;
1123 } 1130 }
1124 v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 40; 1131 v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 60;
1125 copy = clamp_t(size_t, len, 1, sizeof(buf)); 1132 copy = clamp_t(size_t, len, 1, sizeof(buf));
1126 copy = scnprintf(buf, copy, "2.6.%u%s", v, rest); 1133 copy = scnprintf(buf, copy, "2.6.%u%s", v, rest);
1127 ret = copy_to_user(release, buf, copy + 1); 1134 ret = copy_to_user(release, buf, copy + 1);
@@ -2219,6 +2226,12 @@ SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2219 return -EINVAL; 2226 return -EINVAL;
2220 error = MPX_DISABLE_MANAGEMENT(me); 2227 error = MPX_DISABLE_MANAGEMENT(me);
2221 break; 2228 break;
2229 case PR_SET_FP_MODE:
2230 error = SET_FP_MODE(me, arg2);
2231 break;
2232 case PR_GET_FP_MODE:
2233 error = GET_FP_MODE(me);
2234 break;
2222 default: 2235 default:
2223 error = -EINVAL; 2236 error = -EINVAL;
2224 break; 2237 break;
diff --git a/kernel/time/ntp.c b/kernel/time/ntp.c
index 4b585e0fdd22..0f60b08a4f07 100644
--- a/kernel/time/ntp.c
+++ b/kernel/time/ntp.c
@@ -633,10 +633,14 @@ int ntp_validate_timex(struct timex *txc)
633 if ((txc->modes & ADJ_SETOFFSET) && (!capable(CAP_SYS_TIME))) 633 if ((txc->modes & ADJ_SETOFFSET) && (!capable(CAP_SYS_TIME)))
634 return -EPERM; 634 return -EPERM;
635 635
636 if (txc->modes & ADJ_FREQUENCY) { 636 /*
637 if (LONG_MIN / PPM_SCALE > txc->freq) 637 * Check for potential multiplication overflows that can
638 * only happen on 64-bit systems:
639 */
640 if ((txc->modes & ADJ_FREQUENCY) && (BITS_PER_LONG == 64)) {
641 if (LLONG_MIN / PPM_SCALE > txc->freq)
638 return -EINVAL; 642 return -EINVAL;
639 if (LONG_MAX / PPM_SCALE < txc->freq) 643 if (LLONG_MAX / PPM_SCALE < txc->freq)
640 return -EINVAL; 644 return -EINVAL;
641 } 645 }
642 646