什么是进程

进程：是指计算机中已运行的程序。进程本身不是基本的运行单位，而是线程的容器。程序本身只是指令、数据及其组织形式的描述，进程才是程序（那些指令和数据）的真正运行实例。

进程是程序动态的形式。

进程生命周期Linux操作系统属于多任务操作系统，系统中的每个进程能够分时复用CPU时间片，通过有效的进程调度策略实现多任务并行执行。而进程在被CPU调度运行，等待CPU资源分配以及等待外部事件时会属于不同的状态。进程之间的状态关系：

运行：该进程此刻正在执行。
等待：进程能够运行，但没有得到许可，因为CPU分配给另一个进程。调度器可以在
下一次任务切换时选择该进程。
睡眠：进程正在睡眠无法运行，因为它在等待一个外部事件。调度器无法在下一次任
务切换时选择该进程。

Linux内核涉及进程和程序的所有算法都围绕一个名为task_struct的数据结构建立，该结构定义在include/linux/sched.h中。这是系统中主要的一个结构。在阐述调度器的实现之前，了解一下Linux管理进程的方式是很有必要的。task_struct包含很多成员，将进程与各个内核子系统联系，task_struct定义如下：

1 struct task_struct {　　　　进程状态
  2     volatile long state;    /* -1 unrunnable, 0 runnable, >0 stopped */
  3     void *stack;      内核栈
  4     atomic_t usage;    有几个进程在使用此结构
  5     unsigned int flags;    /* per process flags, defined below */
  6     unsigned int ptrace;    断点调试使用
  7 　　多处理使用
  8 #ifdef CONFIG_SMP
  9     struct llist_node wake_entry;
 10     int on_cpu;
 11     unsigned int wakee_flips;
 12     unsigned long wakee_flip_decay_ts;
 13     struct task_struct *last_wakee;
 14
 15     int wake_cpu;
 16 #endif
 17     int on_rq;
 18
 19     int prio, static_prio, normal_prio;
 20     unsigned int rt_priority;                优先级　　　　　　　　//进程调度
 21     const struct sched_class *sched_class;
 22     struct sched_entity se;
 23     struct sched_rt_entity rt;
 24 #ifdef CONFIG_CGROUP_SCHED
 25     struct task_group *sched_task_group;
 26 #endif
 27     struct sched_dl_entity dl;
 28
 29 #ifdef CONFIG_PREEMPT_NOTIFIERS
 30     /* list of struct preempt_notifier: */
 31     struct hlist_head preempt_notifiers;
 32 #endif
 33
 34 #ifdef CONFIG_BLK_DEV_IO_TRACE
 35     unsigned int btrace_seq;
 36 #endif
 37
 38     unsigned int policy;
 39     int nr_cpus_allowed;
 40     cpumask_t cpus_allowed;
 41
 42 #ifdef CONFIG_PREEMPT_RCU
 43     int rcu_read_lock_nesting;
 44     union rcu_special rcu_read_unlock_special;
 45     struct list_head rcu_node_entry;
 46     struct rcu_node *rcu_blocked_node;
 47 #endif /* #ifdef CONFIG_PREEMPT_RCU */
 48 #ifdef CONFIG_TASKS_RCU
 49     unsigned long rcu_tasks_nvcsw;
 50     bool rcu_tasks_holdout;
 51     struct list_head rcu_tasks_holdout_list;
 52     int rcu_tasks_idle_cpu;
 53 #endif /* #ifdef CONFIG_TASKS_RCU */
 54
 55 #ifdef CONFIG_SCHED_INFO
 56     struct sched_info sched_info;
 57 #endif
 58
 59     struct list_head tasks;
 60 #ifdef CONFIG_SMP
 61     struct plist_node pushable_tasks;
 62     struct rb_node pushable_dl_tasks;
 63 #endif
 64
 65     struct mm_struct *mm, *active_mm;
 66     /* per-thread vma caching */
 67     u32 vmacache_seqnum;
 68     struct vm_area_struct *vmacache[VMACACHE_SIZE];
 69 #if defined(SPLIT_RSS_COUNTING)
 70     struct task_rss_stat    rss_stat;
 71 #endif
 72 /* task state */
 73     int exit_state;
 74     int exit_code, exit_signal;
 75     int pdeath_signal;  /*  The signal sent when the parent dies  */
 76     unsigned long jobctl;    /* JOBCTL_*, siglock protected */
 77
 78     /* Used for emulating ABI behavior of previous Linux versions */
 79     unsigned int personality;
 80
 81     /* scheduler bits, serialized by scheduler locks */
 82     unsigned sched_reset_on_fork:1;
 83     unsigned sched_contributes_to_load:1;
 84     unsigned sched_migrated:1;
 85     unsigned :0; /* force alignment to the next boundary */
 86
 87     /* unserialized, strictly 'current' */
 88     unsigned in_execve:1; /* bit to tell LSMs we're in execve */
 89     unsigned in_iowait:1;
 90 #ifdef CONFIG_MEMCG
 91     unsigned memcg_may_oom:1;
 92 #endif
 93 #ifdef CONFIG_MEMCG_KMEM
 94     unsigned memcg_kmem_skip_account:1;
 95 #endif
 96 #ifdef CONFIG_COMPAT_BRK
 97     unsigned brk_randomized:1;
 98 #endif
 99
100     unsigned long atomic_flags; /* Flags needing atomic access. */
101
102     struct restart_block restart_block;
103
104     pid_t pid;     进程和父进程的pid
105     pid_t tgid;
106
107 #ifdef CONFIG_CC_STACKPROTECTOR
108     /* Canary value for the -fstack-protector gcc feature */
109     unsigned long stack_canary;
110 #endif
111     /*
112      * pointers to (original) parent process, youngest child, younger sibling,
113      * older sibling, respectively.  (p->father can be replaced with
114      * p->real_parent->pid)
115      */　　　　　　　　　　　父进程
116     struct task_struct __rcu *real_parent; /* real parent process */　　　　接收终止进程
117     struct task_struct __rcu *parent; /* recipient of SIGCHLD, wait4() reports */
118     /*
119      * children/sibling forms the list of my natural children
120      */　　　　　　子进程链表
121     struct list_head children;    /* list of my children */　　　　兄弟进程链表
122     struct list_head sibling;    /* linkage in my parent's children list */　　　　　　线程组组织
123     struct task_struct *group_leader;    /* threadgroup leader */
124
125     /*
126      * ptraced is the list of tasks this task is using ptrace on.

127      * This includes both natural children and PTRACE_ATTACH targets.

128      * p->ptrace_entry is p's link on the p->parent->ptraced list.

129      */　　　　系统调用调试使用
130     struct list_head ptraced;
131     struct list_head ptrace_entry;
132
133     /* PID/PID hash table linkage. */
134     struct pid_link pids[PIDTYPE_MAX];
135     struct list_head thread_group;
136     struct list_head thread_node;
137
138     struct completion *vfork_done;        /* for vfork() */
139     int __user *set_child_tid;        /* CLONE_CHILD_SETTID */
140     int __user *clear_child_tid;        /* CLONE_CHILD_CLEARTID */
141 　　　　utime 用户态执行时间   stime 内核态执行时间
142     cputime_t utime, stime, utimescaled, stimescaled;
143     cputime_t gtime;
144     struct prev_cputime prev_cputime;
145 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
146     seqlock_t vtime_seqlock;
147     unsigned long long vtime_snap;
148     enum {
149         VTIME_SLEEPING = 0,
150         VTIME_USER,
151         VTIME_SYS,
152     } vtime_snap_whence;
153 #endif
154     unsigned long nvcsw, nivcsw; /* context switch counts */
155     u64 start_time;        /* monotonic time in nsec */
156     u64 real_start_time;    /* boot based time in nsec */
157 /* mm fault and swap info: this can arguably be seen as either mm-specific or thread-specific */
158     unsigned long min_flt, maj_flt;
159
160     struct task_cputime cputime_expires;
161     struct list_head cpu_timers[3];
162
163 /* process credentials */
164     const struct cred __rcu *real_cred; /* objective and real subjective task
165                      * credentials (COW) */
166     const struct cred __rcu *cred;    /* effective (overridable) subjective task
167                      * credentials (COW) */
168     char comm[TASK_COMM_LEN]; /* executable name excluding path
169                      - access with [gs]et_task_comm (which lock
170                        it with task_lock())
171                      - initialized normally by setup_new_exec */
172 /* file system info */
173     struct nameidata *nameidata;
174 #ifdef CONFIG_SYSVIPC
175 /* ipc stuff */
176     struct sysv_sem sysvsem;
177     struct sysv_shm sysvshm;
178 #endif
179 #ifdef CONFIG_DETECT_HUNG_TASK
180 /* hung task detection */
181     unsigned long last_switch_count;
182 #endif
183 /* filesystem information */
184     struct fs_struct *fs;
185 /* open file information */
186     struct files_struct *files;
187 /* namespaces */
188     struct nsproxy *nsproxy;
189 /* signal handlers */
190     struct signal_struct *signal;
191     struct sighand_struct *sighand;
192
193     sigset_t blocked, real_blocked;
194     sigset_t saved_sigmask;    /* restored if set_restore_sigmask() was used */
195     struct sigpending pending;
196
197     unsigned long sas_ss_sp;
198     size_t sas_ss_size;
199
200     struct callback_head *task_works;
201
202     struct audit_context *audit_context;
203 #ifdef CONFIG_AUDITSYSCALL
204     kuid_t loginuid;
205     unsigned int sessionid;
206 #endif
207     struct seccomp seccomp;
208
209 /* Thread group tracking */
210        u32 parent_exec_id;
211        u32 self_exec_id;
212 /* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed,
213  * mempolicy */
214     spinlock_t alloc_lock;
215
216     /* Protection of the PI data structures: */
217     raw_spinlock_t pi_lock;
218
219     struct wake_q_node wake_q;
220
221 #ifdef CONFIG_RT_MUTEXES
222     /* PI waiters blocked on a rt_mutex held by this task */
223     struct rb_root pi_waiters;
224     struct rb_node *pi_waiters_leftmost;
225     /* Deadlock detection and priority inheritance handling */
226     struct rt_mutex_waiter *pi_blocked_on;
227 #endif
228
229 #ifdef CONFIG_DEBUG_MUTEXES
230     /* mutex deadlock detection */
231     struct mutex_waiter *blocked_on;
232 #endif
233 #ifdef CONFIG_TRACE_IRQFLAGS
234     unsigned int irq_events;
235     unsigned long hardirq_enable_ip;
236     unsigned long hardirq_disable_ip;
237     unsigned int hardirq_enable_event;
238     unsigned int hardirq_disable_event;
239     int hardirqs_enabled;
240     int hardirq_context;
241     unsigned long softirq_disable_ip;
242     unsigned long softirq_enable_ip;
243     unsigned int softirq_disable_event;
244     unsigned int softirq_enable_event;
245     int softirqs_enabled;
246     int softirq_context;
247 #endif
248 #ifdef CONFIG_LOCKDEP
249 # define MAX_LOCK_DEPTH 48UL
250     u64 curr_chain_key;
251     int lockdep_depth;
252     unsigned int lockdep_recursion;
253     struct held_lock held_locks[MAX_LOCK_DEPTH];
254     gfp_t lockdep_reclaim_gfp;
255 #endif
256
257 /* journalling filesystem info */
258     void *journal_info;
259
260 /* stacked block device info */
261     struct bio_list *bio_list;
262
263 #ifdef CONFIG_BLOCK
264 /* stack plugging */
265     struct blk_plug *plug;
266 #endif
267
268 /* VM state */
269     struct reclaim_state *reclaim_state;
270
271     struct backing_dev_info *backing_dev_info;
272
273     struct io_context *io_context;
274
275     unsigned long ptrace_message;
276     siginfo_t *last_siginfo; /* For ptrace use.  */
277     struct task_io_accounting ioac;
278 #if defined(CONFIG_TASK_XACCT)
279     u64 acct_rss_mem1;    /* accumulated rss usage */
280     u64 acct_vm_mem1;    /* accumulated virtual memory usage */
281     cputime_t acct_timexpd;    /* stime + utime since last update */
282 #endif
283 #ifdef CONFIG_CPUSETS
284     nodemask_t mems_allowed;    /* Protected by alloc_lock */
285     seqcount_t mems_allowed_seq;    /* Seqence no to catch updates */
286     int cpuset_mem_spread_rotor;
287     int cpuset_slab_spread_rotor;
288 #endif
289 #ifdef CONFIG_CGROUPS
290     /* Control Group info protected by css_set_lock */
291     struct css_set __rcu *cgroups;
292     /* cg_list protected by css_set_lock and tsk->alloc_lock */
293     struct list_head cg_list;
294 #endif
295 #ifdef CONFIG_FUTEX
296     struct robust_list_head __user *robust_list;
297 #ifdef CONFIG_COMPAT
298     struct compat_robust_list_head __user *compat_robust_list;
299 #endif
300     struct list_head pi_state_list;
301     struct futex_pi_state *pi_state_cache;
302 #endif
303 #ifdef CONFIG_PERF_EVENTS
304     struct perf_event_context *perf_event_ctxp[perf_nr_task_contexts];
305     struct mutex perf_event_mutex;
306     struct list_head perf_event_list;
307 #endif
308 #ifdef CONFIG_DEBUG_PREEMPT
309     unsigned long preempt_disable_ip;
310 #endif
311 #ifdef CONFIG_NUMA
312     struct mempolicy *mempolicy;    /* Protected by alloc_lock */
313     short il_next;
314     short pref_node_fork;
315 #endif
316 #ifdef CONFIG_NUMA_BALANCING
317     int numa_scan_seq;
318     unsigned int numa_scan_period;
319     unsigned int numa_scan_period_max;
320     int numa_preferred_nid;
321     unsigned long numa_migrate_retry;
322     u64 node_stamp;            /* migration stamp  */
323     u64 last_task_numa_placement;
324     u64 last_sum_exec_runtime;
325     struct callback_head numa_work;
326
327     struct list_head numa_entry;
328     struct numa_group *numa_group;
329
330     /*
331      * numa_faults is an array split into four regions:
332      * faults_memory, faults_cpu, faults_memory_buffer, faults_cpu_buffer
333      * in this precise order.

334      *
335      * faults_memory: Exponential decaying average of faults on a per-node
336      * basis. Scheduling placement decisions are made based on these
337      * counts. The values remain static for the duration of a PTE scan.

338      * faults_cpu: Track the nodes the process was running on when a NUMA
339      * hinting fault was incurred.

340      * faults_memory_buffer and faults_cpu_buffer: Record faults per node
341      * during the current scan window. When the scan completes, the counts
342      * in faults_memory and faults_cpu decay and these values are copied.

343      */
344     unsigned long *numa_faults;
345     unsigned long total_numa_faults;
346
347     /*
348      * numa_faults_locality tracks if faults recorded during the last
349      * scan window were remote/local or failed to migrate. The task scan
350      * period is adapted based on the locality of the faults with different
351      * weights depending on whether they were shared or private faults
352      */
353     unsigned long numa_faults_locality[3];
354
355     unsigned long numa_pages_migrated;
356 #endif /* CONFIG_NUMA_BALANCING */
357
358 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
359     struct tlbflush_unmap_batch tlb_ubc;
360 #endif
361
362     struct rcu_head rcu;
363
364     /*
365      * cache last used pipe for splice
366      */
367     struct pipe_inode_info *splice_pipe;
368
369     struct page_frag task_frag;
370
371 #ifdef    CONFIG_TASK_DELAY_ACCT
372     struct task_delay_info *delays;
373 #endif
374 #ifdef CONFIG_FAULT_INJECTION
375     int make_it_fail;
376 #endif
377     /*
378      * when (nr_dirtied >= nr_dirtied_pause), it's time to call
379      * balance_dirty_pages() for some dirty throttling pause
380      */
381     int nr_dirtied;
382     int nr_dirtied_pause;
383     unsigned long dirty_paused_when; /* start of a write-and-pause period */
384
385 #ifdef CONFIG_LATENCYTOP
386     int latency_record_count;
387     struct latency_record latency_record[LT_SAVECOUNT];
388 #endif
389     /*
390      * time slack values; these are used to round up poll() and
391      * select() etc timeout values. These are in nanoseconds.

392      */
393     unsigned long timer_slack_ns;
394     unsigned long default_timer_slack_ns;
395
396 #ifdef CONFIG_KASAN
397     unsigned int kasan_depth;
398 #endif
399 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
400     /* Index of current stored address in ret_stack */
401     int curr_ret_stack;
402     /* Stack of return addresses for return function tracing */
403     struct ftrace_ret_stack    *ret_stack;
404     /* time stamp for last schedule */
405     unsigned long long ftrace_timestamp;
406     /*
407      * Number of functions that haven't been traced
408      * because of depth overrun.

409      */
410     atomic_t trace_overrun;
411     /* Pause for the tracing */
412     atomic_t tracing_graph_pause;
413 #endif
414 #ifdef CONFIG_TRACING
415     /* state flags for use by tracers */
416     unsigned long trace;
417     /* bitmask and counter of trace recursion */
418     unsigned long trace_recursion;
419 #endif /* CONFIG_TRACING */
420 #ifdef CONFIG_MEMCG
421     struct mem_cgroup *memcg_in_oom;
422     gfp_t memcg_oom_gfp_mask;
423     int memcg_oom_order;
424
425     /* number of pages to reclaim on returning to userland */
426     unsigned int memcg_nr_pages_over_high;
427 #endif
428 #ifdef CONFIG_UPROBES
429     struct uprobe_task *utask;
430 #endif
431 #if defined(CONFIG_BCACHE) || defined(CONFIG_BCACHE_MODULE)
432     unsigned int    sequential_io;
433     unsigned int    sequential_io_avg;
434 #endif
435 #ifdef CONFIG_DEBUG_ATOMIC_SLEEP
436     unsigned long    task_state_change;
437 #endif
438     int pagefault_disabled;
439 /* CPU-specific state of this task */
440     struct thread_struct thread;
441 /*
442  * WARNING: on x86, 'thread_struct' contains a variable-sized
443  * structure.  It *MUST* be at the end of 'task_struct'.

444  *
445  * Do not put anything below here!

446  */
447 };

并非所有进程都具有相同的重要性。除了大多数我们所熟悉的进程优先级之外，进程还有不同的关键度类别，以满足不同需求先进行比较粗糙的划分，进程可以分为 实时进程和 非实时进程 （普通进程）。实时进程优先级（0-99）都比普通 进程的优先级（100-139）高。当系统中有实时进程 运行时，普通进程几乎无法分到赶时间片（只能分到5%的CPU时间）。

讨论fork和exec系列系统调用的实现。通常这些调用不是由应用程序直接发出的，而是通过一个中间层调用，即负责与内核通信的C标准库。从用户状态切换到核心态的方法，依不同的体系结构而各有不同。

1、进程复制

传统的UNIX中用于复制进程的系统调用是fork。但它并不是Linux为此实现的唯一调用，

实际上Linux实现了3个。

(1) fork是重量级调用，因为它建立了父进程的一个完整副本，然后作为子进程执行。为减少与该调用相关的工作量，Linux使用了写时复制（copy-on-write）技术。

(2) vfork类似于fork，但并不创建父进程数据的副本。相反，父子进程之间共享数据。这节省了大量CPU时间（如果一个进程操纵共享数据，则另一个会自动注意到）。

(3) clone产生线程，可以对父子进程之间的共享、复制进行精确控制。

五、进程系统调用【写时复制】

内核使用了写时复制（Copy-On-Write，COW）技术，以防止在fork执行时将父进程的所有数据复制到子进程。在调用fork时，内核通常对父进程的每个内存页，都为子进程创建一个相同的副本。

【执行系统调用】
fork、vfork和clone系统调用的入口点分别是sys_fork、sys_vfork和sys_clone函数。其定义依赖于
具体的体系结构，因为在用户空间和内核空间之间传递参数的方法因体系结构而异。

所有3个fork机制最终都调用kernel/fork.c中的do_fork（一个体系结构无关的函数），其代码流程

如图所示。

内核线程是直接由内核本身启动的进程。内核线程实际上是将内核函数委托给独立的
进程，与系统中其他进程”并行”执行（实际上，也并行于内核自身的执行）。内核线程经常
称之为（内核）守护进程。它们用于执行下列任务。
• 周期性地将修改的内存页与页来源块设备同步（例如，使用mmap的文件映射）。
• 如果内存页很少使用，则写入交换区。
• 管理延时动作（deferred action）。
• 实现文件系统的事务日志。

3、退出进程

进程必须用exit系统调用终止。这使得内核有机会将该进程使用的资源释放回系统。见kernel/exit.c——>do_exit。简而言之，该函数的实现就是将各个引用计数器减1，如果引用计数器归0而没有进程再使用对应的结构，那么将相应的内存区域返还给内存理模块。

exit () 系统调用一般由编译器在main函数结束位置添加

Original: https://www.cnblogs.com/thotf/p/16261566.html
Author: thotf
Title: 什么是进程