1/* 2 * Read-Copy Update mechanism for mutual exclusion (tree-based version) 3 * Internal non-public definitions. 4 * 5 * This program is free software; you can redistribute it and/or modify 6 * it under the terms of the GNU General Public License as published by 7 * the Free Software Foundation; either version 2 of the License, or 8 * (at your option) any later version. 9 * 10 * This program is distributed in the hope that it will be useful, 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 13 * GNU General Public License for more details. 14 * 15 * You should have received a copy of the GNU General Public License 16 * along with this program; if not, write to the Free Software 17 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. 18 * 19 * Copyright IBM Corporation, 2008 20 * 21 * Author: Ingo Molnar <mingo@elte.hu> 22 * Paul E. McKenney <paulmck@linux.vnet.ibm.com> 23 */ 24 25#include <linux/cache.h> 26#include <linux/spinlock.h> 27#include <linux/threads.h> 28#include <linux/cpumask.h> 29#include <linux/seqlock.h> 30 31/* 32 * Define shape of hierarchy based on NR_CPUS and CONFIG_RCU_FANOUT. 33 * In theory, it should be possible to add more levels straightforwardly. 34 * In practice, this has not been tested, so there is probably some 35 * bug somewhere. 36 */ 37#define MAX_RCU_LVLS 3 38#define RCU_FANOUT (CONFIG_RCU_FANOUT) 39#define RCU_FANOUT_SQ (RCU_FANOUT * RCU_FANOUT) 40#define RCU_FANOUT_CUBE (RCU_FANOUT_SQ * RCU_FANOUT) 41 42#if NR_CPUS <= RCU_FANOUT 43# define NUM_RCU_LVLS 1 44# define NUM_RCU_LVL_0 1 45# define NUM_RCU_LVL_1 (NR_CPUS) 46# define NUM_RCU_LVL_2 0 47# define NUM_RCU_LVL_3 0 48#elif NR_CPUS <= RCU_FANOUT_SQ 49# define NUM_RCU_LVLS 2 50# define NUM_RCU_LVL_0 1 51# define NUM_RCU_LVL_1 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT) 52# define NUM_RCU_LVL_2 (NR_CPUS) 53# define NUM_RCU_LVL_3 0 54#elif NR_CPUS <= RCU_FANOUT_CUBE 55# define NUM_RCU_LVLS 3 56# define NUM_RCU_LVL_0 1 57# define NUM_RCU_LVL_1 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_SQ) 58# define NUM_RCU_LVL_2 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT) 59# define NUM_RCU_LVL_3 NR_CPUS 60#else 61# error "CONFIG_RCU_FANOUT insufficient for NR_CPUS" 62#endif /* #if (NR_CPUS) <= RCU_FANOUT */ 63 64#define RCU_SUM (NUM_RCU_LVL_0 + NUM_RCU_LVL_1 + NUM_RCU_LVL_2 + NUM_RCU_LVL_3) 65#define NUM_RCU_NODES (RCU_SUM - NR_CPUS) 66 67/* 68 * Dynticks per-CPU state. 69 */ 70struct rcu_dynticks { 71 int dynticks_nesting; /* Track nesting level, sort of. */ 72 int dynticks; /* Even value for dynticks-idle, else odd. */ 73 int dynticks_nmi; /* Even value for either dynticks-idle or */ 74 /* not in nmi handler, else odd. So this */ 75 /* remains even for nmi from irq handler. */ 76}; 77 78/* 79 * Definition for node within the RCU grace-period-detection hierarchy. 80 */ 81struct rcu_node { 82 spinlock_t lock; /* Root rcu_node's lock protects some */ 83 /* rcu_state fields as well as following. */ 84 long gpnum; /* Current grace period for this node. */ 85 /* This will either be equal to or one */ 86 /* behind the root rcu_node's gpnum. */ 87 unsigned long qsmask; /* CPUs or groups that need to switch in */ 88 /* order for current grace period to proceed.*/ 89 /* In leaf rcu_node, each bit corresponds to */ 90 /* an rcu_data structure, otherwise, each */ 91 /* bit corresponds to a child rcu_node */ 92 /* structure. */ 93 unsigned long qsmaskinit; 94 /* Per-GP initialization for qsmask. */ 95 unsigned long grpmask; /* Mask to apply to parent qsmask. */ 96 /* Only one bit will be set in this mask. */ 97 int grplo; /* lowest-numbered CPU or group here. */ 98 int grphi; /* highest-numbered CPU or group here. */ 99 u8 grpnum; /* CPU/group number for next level up. */ 100 u8 level; /* root is at level 0. */ 101 struct rcu_node *parent; 102 struct list_head blocked_tasks[2]; 103 /* Tasks blocked in RCU read-side critsect. */ 104 /* Grace period number (->gpnum) x blocked */ 105 /* by tasks on the (x & 0x1) element of the */ 106 /* blocked_tasks[] array. */ 107} ____cacheline_internodealigned_in_smp; 108 109/* 110 * Do a full breadth-first scan of the rcu_node structures for the 111 * specified rcu_state structure. 112 */ 113#define rcu_for_each_node_breadth_first(rsp, rnp) \ 114 for ((rnp) = &(rsp)->node[0]; \ 115 (rnp) < &(rsp)->node[NUM_RCU_NODES]; (rnp)++) 116 117#define rcu_for_each_leaf_node(rsp, rnp) \ 118 for ((rnp) = (rsp)->level[NUM_RCU_LVLS - 1]; \ 119 (rnp) < &(rsp)->node[NUM_RCU_NODES]; (rnp)++) 120 121/* Index values for nxttail array in struct rcu_data. */ 122#define RCU_DONE_TAIL 0 /* Also RCU_WAIT head. */ 123#define RCU_WAIT_TAIL 1 /* Also RCU_NEXT_READY head. */ 124#define RCU_NEXT_READY_TAIL 2 /* Also RCU_NEXT head. */ 125#define RCU_NEXT_TAIL 3 126#define RCU_NEXT_SIZE 4 127 128/* Per-CPU data for read-copy update. */ 129struct rcu_data { 130 /* 1) quiescent-state and grace-period handling : */ 131 long completed; /* Track rsp->completed gp number */ 132 /* in order to detect GP end. */ 133 long gpnum; /* Highest gp number that this CPU */ 134 /* is aware of having started. */ 135 long passed_quiesc_completed; 136 /* Value of completed at time of qs. */ 137 bool passed_quiesc; /* User-mode/idle loop etc. */ 138 bool qs_pending; /* Core waits for quiesc state. */ 139 bool beenonline; /* CPU online at least once. */ 140 bool preemptable; /* Preemptable RCU? */ 141 struct rcu_node *mynode; /* This CPU's leaf of hierarchy */ 142 unsigned long grpmask; /* Mask to apply to leaf qsmask. */ 143 144 /* 2) batch handling */ 145 /* 146 * If nxtlist is not NULL, it is partitioned as follows. 147 * Any of the partitions might be empty, in which case the 148 * pointer to that partition will be equal to the pointer for 149 * the following partition. When the list is empty, all of 150 * the nxttail elements point to the ->nxtlist pointer itself, 151 * which in that case is NULL. 152 * 153 * [nxtlist, *nxttail[RCU_DONE_TAIL]): 154 * Entries that batch # <= ->completed 155 * The grace period for these entries has completed, and 156 * the other grace-period-completed entries may be moved 157 * here temporarily in rcu_process_callbacks(). 158 * [*nxttail[RCU_DONE_TAIL], *nxttail[RCU_WAIT_TAIL]): 159 * Entries that batch # <= ->completed - 1: waiting for current GP 160 * [*nxttail[RCU_WAIT_TAIL], *nxttail[RCU_NEXT_READY_TAIL]): 161 * Entries known to have arrived before current GP ended 162 * [*nxttail[RCU_NEXT_READY_TAIL], *nxttail[RCU_NEXT_TAIL]): 163 * Entries that might have arrived after current GP ended 164 * Note that the value of *nxttail[RCU_NEXT_TAIL] will 165 * always be NULL, as this is the end of the list. 166 */ 167 struct rcu_head *nxtlist; 168 struct rcu_head **nxttail[RCU_NEXT_SIZE]; 169 long qlen; /* # of queued callbacks */ 170 long qlen_last_fqs_check; 171 /* qlen at last check for QS forcing */ 172 unsigned long n_force_qs_snap; 173 /* did other CPU force QS recently? */ 174 long blimit; /* Upper limit on a processed batch */ 175 176#ifdef CONFIG_NO_HZ 177 /* 3) dynticks interface. */ 178 struct rcu_dynticks *dynticks; /* Shared per-CPU dynticks state. */ 179 int dynticks_snap; /* Per-GP tracking for dynticks. */ 180 int dynticks_nmi_snap; /* Per-GP tracking for dynticks_nmi. */ 181#endif /* #ifdef CONFIG_NO_HZ */ 182 183 /* 4) reasons this CPU needed to be kicked by force_quiescent_state */ 184#ifdef CONFIG_NO_HZ 185 unsigned long dynticks_fqs; /* Kicked due to dynticks idle. */ 186#endif /* #ifdef CONFIG_NO_HZ */ 187 unsigned long offline_fqs; /* Kicked due to being offline. */ 188 unsigned long resched_ipi; /* Sent a resched IPI. */ 189 190 /* 5) __rcu_pending() statistics. */ 191 long n_rcu_pending; /* rcu_pending() calls since boot. */ 192 long n_rp_qs_pending; 193 long n_rp_cb_ready; 194 long n_rp_cpu_needs_gp; 195 long n_rp_gp_completed; 196 long n_rp_gp_started; 197 long n_rp_need_fqs; 198 long n_rp_need_nothing; 199 200 int cpu; 201}; 202 203/* Values for signaled field in struct rcu_state. */ 204#define RCU_GP_IDLE 0 /* No grace period in progress. */ 205#define RCU_GP_INIT 1 /* Grace period being initialized. */ 206#define RCU_SAVE_DYNTICK 2 /* Need to scan dyntick state. */ 207#define RCU_FORCE_QS 3 /* Need to force quiescent state. */ 208#ifdef CONFIG_NO_HZ 209#define RCU_SIGNAL_INIT RCU_SAVE_DYNTICK 210#else /* #ifdef CONFIG_NO_HZ */ 211#define RCU_SIGNAL_INIT RCU_FORCE_QS 212#endif /* #else #ifdef CONFIG_NO_HZ */ 213 214#define RCU_JIFFIES_TILL_FORCE_QS 3 /* for rsp->jiffies_force_qs */ 215#ifdef CONFIG_RCU_CPU_STALL_DETECTOR 216#define RCU_SECONDS_TILL_STALL_CHECK (10 * HZ) /* for rsp->jiffies_stall */ 217#define RCU_SECONDS_TILL_STALL_RECHECK (30 * HZ) /* for rsp->jiffies_stall */ 218#define RCU_STALL_RAT_DELAY 2 /* Allow other CPUs time */ 219 /* to take at least one */ 220 /* scheduling clock irq */ 221 /* before ratting on them. */ 222 223#endif /* #ifdef CONFIG_RCU_CPU_STALL_DETECTOR */ 224 225/* 226 * RCU global state, including node hierarchy. This hierarchy is 227 * represented in "heap" form in a dense array. The root (first level) 228 * of the hierarchy is in ->node[0] (referenced by ->level[0]), the second 229 * level in ->node[1] through ->node[m] (->node[1] referenced by ->level[1]), 230 * and the third level in ->node[m+1] and following (->node[m+1] referenced 231 * by ->level[2]). The number of levels is determined by the number of 232 * CPUs and by CONFIG_RCU_FANOUT. Small systems will have a "hierarchy" 233 * consisting of a single rcu_node. 234 */ 235struct rcu_state { 236 struct rcu_node node[NUM_RCU_NODES]; /* Hierarchy. */ 237 struct rcu_node *level[NUM_RCU_LVLS]; /* Hierarchy levels. */ 238 u32 levelcnt[MAX_RCU_LVLS + 1]; /* # nodes in each level. */ 239 u8 levelspread[NUM_RCU_LVLS]; /* kids/node in each level. */ 240 struct rcu_data *rda[NR_CPUS]; /* array of rdp pointers. */ 241 242 /* The following fields are guarded by the root rcu_node's lock. */ 243 244 u8 signaled ____cacheline_internodealigned_in_smp; 245 /* Force QS state. */ 246 long gpnum; /* Current gp number. */ 247 long completed; /* # of last completed gp. */ 248 249 /* End of fields guarded by root rcu_node's lock. */ 250 251 spinlock_t onofflock; /* exclude on/offline and */ 252 /* starting new GP. Also */ 253 /* protects the following */ 254 /* orphan_cbs fields. */ 255 struct rcu_head *orphan_cbs_list; /* list of rcu_head structs */ 256 /* orphaned by all CPUs in */ 257 /* a given leaf rcu_node */ 258 /* going offline. */ 259 struct rcu_head **orphan_cbs_tail; /* And tail pointer. */ 260 long orphan_qlen; /* Number of orphaned cbs. */ 261 spinlock_t fqslock; /* Only one task forcing */ 262 /* quiescent states. */ 263 unsigned long jiffies_force_qs; /* Time at which to invoke */ 264 /* force_quiescent_state(). */ 265 unsigned long n_force_qs; /* Number of calls to */ 266 /* force_quiescent_state(). */ 267 unsigned long n_force_qs_lh; /* ~Number of calls leaving */ 268 /* due to lock unavailable. */ 269 unsigned long n_force_qs_ngp; /* Number of calls leaving */ 270 /* due to no GP active. */ 271#ifdef CONFIG_RCU_CPU_STALL_DETECTOR 272 unsigned long gp_start; /* Time at which GP started, */ 273 /* but in jiffies. */ 274 unsigned long jiffies_stall; /* Time at which to check */ 275 /* for CPU stalls. */ 276#endif /* #ifdef CONFIG_RCU_CPU_STALL_DETECTOR */ 277#ifdef CONFIG_NO_HZ 278 long dynticks_completed; /* Value of completed @ snap. */ 279#endif /* #ifdef CONFIG_NO_HZ */ 280}; 281 282#ifdef RCU_TREE_NONCORE 283 284/* 285 * RCU implementation internal declarations: 286 */ 287extern struct rcu_state rcu_sched_state; 288DECLARE_PER_CPU(struct rcu_data, rcu_sched_data); 289 290extern struct rcu_state rcu_bh_state; 291DECLARE_PER_CPU(struct rcu_data, rcu_bh_data); 292 293#ifdef CONFIG_TREE_PREEMPT_RCU 294extern struct rcu_state rcu_preempt_state; 295DECLARE_PER_CPU(struct rcu_data, rcu_preempt_data); 296#endif /* #ifdef CONFIG_TREE_PREEMPT_RCU */ 297 298#else /* #ifdef RCU_TREE_NONCORE */ 299 300/* Forward declarations for rcutree_plugin.h */ 301static inline void rcu_bootup_announce(void); 302long rcu_batches_completed(void); 303static void rcu_preempt_note_context_switch(int cpu); 304static int rcu_preempted_readers(struct rcu_node *rnp); 305#ifdef CONFIG_RCU_CPU_STALL_DETECTOR 306static void rcu_print_task_stall(struct rcu_node *rnp); 307#endif /* #ifdef CONFIG_RCU_CPU_STALL_DETECTOR */ 308static void rcu_preempt_check_blocked_tasks(struct rcu_node *rnp); 309#ifdef CONFIG_HOTPLUG_CPU 310static int rcu_preempt_offline_tasks(struct rcu_state *rsp, 311 struct rcu_node *rnp, 312 struct rcu_data *rdp); 313static void rcu_preempt_offline_cpu(int cpu); 314#endif /* #ifdef CONFIG_HOTPLUG_CPU */ 315static void rcu_preempt_check_callbacks(int cpu); 316static void rcu_preempt_process_callbacks(void); 317void call_rcu(struct rcu_head *head, void (*func)(struct rcu_head *rcu)); 318static int rcu_preempt_pending(int cpu); 319static int rcu_preempt_needs_cpu(int cpu); 320static void __cpuinit rcu_preempt_init_percpu_data(int cpu); 321static void rcu_preempt_send_cbs_to_orphanage(void); 322static void __init __rcu_init_preempt(void); 323 324#endif /* #else #ifdef RCU_TREE_NONCORE */ 325