linux下poll和epoll内核源代码剖析

晨曦之光 发布于 2012/03/09 14:50
阅读 1K+
收藏 1

解读下一代网络:算力网络正从理想照进现实!>>>

作者:董昊 博客链接http://donghao.org/uii/


poll和epoll的使用应该不用再多说了。当fd很多时,使用epoll比poll效率更高。

我们通过内核源码分析来看看到底是为什么。

poll剖析
poll系统调用:
int poll(struct pollfd *fds, nfds_t nfds, int timeout);
内核2.6.9对应的实现代码为:
[fs/select.c -->sys_poll]
456 asmlinkage long sys_poll(struct pollfd __user * ufds, unsigned int nfds, long timeout)
457 {
458 struct poll_wqueues table;
459 int fdcount, err;
460 unsigned int i;
461 struct poll_list *head;
462 struct poll_list *walk;
463
464 /* Do a sanity check on nfds ... */ /* 用户给的nfds数不可以超过一个struct file结构支持
的最大fd数(默认是256)*/
465 if (nfds > current->files->max_fdset && nfds > OPEN_MAX)
466 return -EINVAL;
467
468 if (timeout) {
469 /* Careful about overflow in the intermediate values */
470 if ((unsigned long) timeout < MAX_SCHEDULE_TIMEOUT / HZ)
471 timeout = (unsigned long)(timeout*HZ+999)/1000+1;
472 else /* Negative or overflow */
473 timeout = MAX_SCHEDULE_TIMEOUT;
474 }
475
476 poll_initwait(&table);
其中poll_initwait较为关键,从字面上看,应该是初始化变量table,注意此处table在整个执行poll的过
程中是很关键的变量。
而struct poll_table其实就只包含了一个函数指针:
[fs/poll.h]
16 /*
17 * structures and helpers for f_op->poll implementations
18 */
19 typedef void (*poll_queue_proc)(struct file *, wait_queue_head_t *, struct
poll_table_struct *);
20
21 typedef struct poll_table_struct {
22 poll_queue_proc qproc;
23 } poll_table;
现在我们来看看poll_initwait到底在做些什么
[fs/select.c]
57 void __pollwait(struct file *filp, wait_queue_head_t *wait_address, poll_table *p);
58
59 void poll_initwait(struct poll_wqueues *pwq)
60 {
61 &(pwq->pt)->qproc = __pollwait; /*此行已经被我“翻译”了,方便观看*/
62 pwq->error = 0;
63 pwq->table = NULL;
64 }
很明显,poll_initwait的主要动作就是把table变量的成员poll_table对应的回调函数置为__pollwait。这
个__pollwait不仅是poll系统调用需要,select系统调用也一样是用这个__pollwait,说白了,这是个操
作系统的异步操作的“御用”回调函数。当然了,epoll没有用这个,它另外新增了一个回调函数,以达到其
高效运转的目的,这是后话,暂且不表。
我们先不讨论__pollwait的具体实现,还是继续看sys_poll:
[fs/select.c -->sys_poll]
478 head = NULL;
479 walk = NULL;
480 i = nfds;
481 err = -ENOMEM;
482 while(i!=0) {
483 struct poll_list *pp;
484 pp = kmalloc(sizeof(struct poll_list)+
485 sizeof(struct pollfd)*
486 (i>POLLFD_PER_PAGE?POLLFD_PER_PAGE:i),
487 GFP_KERNEL);
488 if(pp==NULL)
489 goto out_fds;
490 pp->next=NULL;
491 pp->len = (i>POLLFD_PER_PAGE?POLLFD_PER_PAGE:i);
492 if (head == NULL)
493 head = pp;
494 else
495 walk->next = pp;
496
497 walk = pp;
498 if (copy_from_user(pp->entries, ufds + nfds-i,
499 sizeof(struct pollfd)*pp->len)) {
500 err = -EFAULT;
501 goto out_fds;
502 }
503 i -= pp->len;
504 }
505 fdcount = do_poll(nfds, head, &table, timeout);
这一大堆代码就是建立一个链表,每个链表的节点是一个page大小(通常是4k),这链表节点由一个指向
struct poll_list的指针掌控,而众多的struct pollfd就通过struct_list的entries成员访问。上面的循环就
是把用户态的struct pollfd拷进这些entries里。通常用户程序的poll调用就监控几个fd,所以上面这个链
表通常也就只需要一个节点,即操作系统的一页。但是,当用户传入的fd很多时,由于poll系统调用每次都
要把所有struct pollfd拷进内核,所以参数传递和页分配此时就成了poll系统调用的性能瓶颈。
最后一句do_poll,我们跟进去:
[fs/select.c-->sys_poll()-->do_poll()]
395 static void do_pollfd(unsigned int num, struct pollfd * fdpage,
396 poll_table ** pwait, int *count)
397 {
398 int i;
399
400 for (i = 0; i < num; i++) {
401 int fd;
402 unsigned int mask;
403 struct pollfd *fdp;
404
405 mask = 0;
406 fdp = fdpage+i;
407 fd = fdp->fd;
408 if (fd >= 0) {
409 struct file * file = fget(fd);
410 mask = POLLNVAL;
411 if (file != NULL) {
412 mask = DEFAULT_POLLMASK;
413 if (file->f_op && file->f_op->poll)
414 mask = file->f_op->poll(file, *pwait);
415 mask &= fdp->events | POLLERR | POLLHUP;
416 fput(file);
417 }
418 if (mask) {
419 *pwait = NULL;
420 (*count)++;
421 }
422 }
423 fdp->revents = mask;
424 }
425 }
426
427 static int do_poll(unsigned int nfds, struct poll_list *list,
428 struct poll_wqueues *wait, long timeout)
429 {
430 int count = 0;
431 poll_table* pt = &wait->pt;
432
433 if (!timeout)
434 pt = NULL;
435
436 for (;;) {
437 struct poll_list *walk;
438 set_current_state(TASK_INTERRUPTIBLE);
439 walk = list;
440 while(walk != NULL) {
441 do_pollfd( walk->len, walk->entries, &pt, &count);
442 walk = walk->next;
443 }
444 pt = NULL;
445 if (count || !timeout || signal_pending(current))
446 break;
447 count = wait->error;
448 if (count)
449 break;
450 timeout = schedule_timeout(timeout); /* 让current挂起,别的进程跑,timeout到了
以后再回来运行current*/
451 }
452 __set_current_state(TASK_RUNNING);
453 return count;
454 }
注意438行的set_current_state和445行的signal_pending,它们两句保障了当用户程序在调用poll后
挂起时,发信号可以让程序迅速推出poll调用,而通常的系统调用是不会被信号打断的。
纵览do_poll函数,主要是在循环内等待,直到count大于0才跳出循环,而count主要是靠do_pollfd函数
处理。
注意标红的440-443行,当用户传入的fd很多时(比如1000个),对do_pollfd就会调用很多次,poll效
率瓶颈的另一原因就在这里。
do_pollfd就是针对每个传进来的fd,调用它们各自对应的poll函数,简化一下调用过程,如下:
struct file* file = fget(fd);
file->f_op->poll(file, &(table->pt));
如果fd对应的是某个socket,do_pollfd调用的就是网络设备驱动实现的poll;如果fd对应的是某个ext3文
件系统上的一个打开文件,那do_pollfd调用的就是ext3文件系统驱动实现的poll。一句话,这个file-
>f_op->poll是设备驱动程序实现的,那设备驱动程序的poll实现通常又是什么样子呢?其实,设备驱动
程序的标准实现是:调用poll_wait,即以设备自己的等待队列为参数(通常设备都有自己的等待队列,不
然一个不支持异步操作的设备会让人很郁闷)调用struct poll_table的回调函数。
作为驱动程序的代表,我们看看socket在使用tcp时的代码:
[net/ipv4/tcp.c-->tcp_poll]
329 unsigned int tcp_poll(struct file *file, struct socket *sock, poll_table *wait)
330 {
331 unsigned int mask;
332 struct sock *sk = sock->sk;
333 struct tcp_opt *tp = tcp_sk(sk);
334
335 poll_wait(file, sk->sk_sleep, wait);
代码就看这些,剩下的无非就是判断状态、返回状态值,tcp_poll的核心实现就是poll_wait,而
poll_wait就是调用struct poll_table对应的回调函数,那poll系统调用对应的回调函数就是
__poll_wait,所以这里几乎就可以把tcp_poll理解为一个语句:
__poll_wait(file, sk->sk_sleep, wait);
由此也可以看出,每个socket自己都带有一个等待队列sk_sleep,所以上面我们所说的“设备的等待队列”
其实不止一个。
这时候我们再看看__poll_wait的实现:
[fs/select.c-->__poll_wait()]
89 void __pollwait(struct file *filp, wait_queue_head_t *wait_address, poll_table *_p)
90 {
91 struct poll_wqueues *p = container_of(_p, struct poll_wqueues, pt);
92 struct poll_table_page *table = p->table;
93
94 if (!table || POLL_TABLE_FULL(table)) {
95 struct poll_table_page *new_table;
96
97 new_table = (struct poll_table_page *) __get_free_page(GFP_KERNEL);
98 if (!new_table) {
99 p->error = -ENOMEM;
100 __set_current_state(TASK_RUNNING);
101 return;
102 }
103 new_table->entry = new_table->entries;
104 new_table->next = table;
105 p->table = new_table;
106 table = new_table;
107 }
108
109 /* Add a new entry */
110 {
111 struct poll_table_entry * entry = table->entry;
112 table->entry = entry+1;
113 get_file(filp);
114 entry->filp = filp;
115 entry->wait_address = wait_address;
116 init_waitqueue_entry(&entry->wait, current);
117 add_wait_queue(wait_address,&entry->wait);
118 }
119 }

 


__poll_wait的作用就是创建了上图所示的数据结构(一次__poll_wait即一次设备poll调用只创建一个
poll_table_entry),并通过struct poll_table_entry的wait成员,把current挂在了设备的等待队列
上,此处的等待队列是wait_address,对应tcp_poll里的sk->sk_sleep。
现在我们可以回顾一下poll系统调用的原理了:先注册回调函数__poll_wait,再初始化table变量(类型
为struct poll_wqueues),接着拷贝用户传入的struct pollfd(其实主要是fd),然后轮流调用所有fd对
应的poll(把current挂到各个fd对应的设备等待队列上)。在设备收到一条消息(网络设备)或填写完文
件数据(磁盘设备)后,会唤醒设备等待队列上的进程,这时current便被唤醒了。current醒来后离开
sys_poll的操作相对简单,这里就不逐行分析了。

epoll原理简介


通过上面的分析,poll运行效率的两个瓶颈已经找出,现在的问题是怎么改进。首先,每次poll都要把
1000个fd 拷入内核,太不科学了,内核干嘛不自己保存已经拷入的fd呢?答对了,epoll就是自己保存拷
入的fd,它的API就已经说明了这一点——不是 epoll_wait的时候才传入fd,而是通过epoll_ctl把所有fd
传入内核再一起"wait",这就省掉了不必要的重复拷贝。其次,在 epoll_wait时,也不是把current轮流
的加入fd对应的设备等待队列,而是在设备等待队列醒来时调用一个回调函数(当然,这就需要“唤醒回
调”机制),把产生事件的fd归入一个链表,然后返回这个链表上的fd。
epoll剖析
epoll是个module,所以先看看module的入口eventpoll_init
[fs/eventpoll.c-->evetpoll_init()]
1582 static int __init eventpoll_init(void)
1583 {
1584 int error;
1585
1586 init_MUTEX(&epsem);
1587
1588 /* Initialize the structure used to perform safe poll wait head wake ups */
1589 ep_poll_safewake_init(&psw);
1590
1591 /* Allocates slab cache used to allocate "struct epitem" items */
1592 epi_cache = kmem_cache_create("eventpoll_epi", sizeof(struct epitem),
1593 0, SLAB_HWCACHE_ALIGN|EPI_SLAB_DEBUG|SLAB_PANIC,
1594 NULL, NULL);
1595
1596 /* Allocates slab cache used to allocate "struct eppoll_entry" */
1597 pwq_cache = kmem_cache_create("eventpoll_pwq",
1598 sizeof(struct eppoll_entry), 0,
1599 EPI_SLAB_DEBUG|SLAB_PANIC, NULL, NULL);
1600
1601 /*
1602 * Register the virtual file system that will be the source of inodes
1603 * for the eventpoll files
1604 */
1605 error = register_filesystem(&eventpoll_fs_type);
1606 if (error)
1607 goto epanic;
1608
1609 /* Mount the above commented virtual file system */
1610 eventpoll_mnt = kern_mount(&eventpoll_fs_type);
1611 error = PTR_ERR(eventpoll_mnt);
1612 if (IS_ERR(eventpoll_mnt))
1613 goto epanic;
1614
1615 DNPRINTK(3, (KERN_INFO "[%p] eventpoll: successfully initialized.\n",
1616 current));
1617 return 0;
1618
1619 epanic:
1620 panic("eventpoll_init() failed\n");
1621 }
很有趣,这个module在初始化时注册了一个新的文件系统,叫"eventpollfs"(在eventpoll_fs_type结
构里),然后挂载此文件系统。另外创建两个内核cache(在内核编程中,如果需要频繁分配小块内存,
应该创建kmem_cahe来做“内存池”),分别用于存放struct epitem和eppoll_entry。如果以后要开发新
的文件系统,可以参考这段代码。
现在想想epoll_create为什么会返回一个新的fd?因为它就是在这个叫做"eventpollfs"的文件系统里创建
了一个新文件!如下:
[fs/eventpoll.c-->sys_epoll_create()]
476 asmlinkage long sys_epoll_create(int size)
477 {
478 int error, fd;
479 struct inode *inode;
480 struct file *file;
481
482 DNPRINTK(3, (KERN_INFO "[%p] eventpoll: sys_epoll_create(%d)\n",
483 current, size));
484
485 /* Sanity check on the size parameter */
486 error = -EINVAL;
487 if (size <= 0)
488 goto eexit_1;
489
490 /*
491 * Creates all the items needed to setup an eventpoll file. That is,
492 * a file structure, and inode and a free file descriptor.
493 */
494 error = ep_getfd(&fd, &inode, &file);
495 if (error)
496 goto eexit_1;
497
498 /* Setup the file internal data structure ( "struct eventpoll" ) */
499 error = ep_file_init(file);
500 if (error)
501 goto eexit_2;
函数很简单,其中ep_getfd看上去是“get”,其实在第一次调用epoll_create时,它是要创建新inode、
新的file、新的fd。而ep_file_init则要创建一个struct eventpoll结构,并把它放入file-
>private_data,注意,这个private_data后面还要用到的。
看到这里,也许有人要问了,为什么epoll的开发者不做一个内核的超级大map把用户要创建的epoll句柄
存起来,在epoll_create时返回一个指针?那似乎很直观呀。但是,仔细看看,linux的系统调用有多少是
返回指针的?你会发现几乎没有!(特此强调,malloc不是系统调用,malloc调用的brk才是)因为linux
做为unix的最杰出的继承人,它遵循了unix的一个巨大优点——一切皆文件,输入输出是文件、socket也
是文件,一切皆文件意味着使用这个操作系统的程序可以非常简单,因为一切都是文件操作而已!(unix
还不是完全做到,plan 9才算)。而且使用文件系统有个好处:epoll_create返回的是一个fd,而不是该
死的指针,指针如果指错了,你简直没办法判断,而fd则可以通过current->files->fd_array[]找到其真
伪。
epoll_create好了,该epoll_ctl了,我们略去判断性的代码:
[fs/eventpoll.c-->sys_epoll_ctl()]
524 asmlinkage long
525 sys_epoll_ctl(int epfd, int op, int fd, struct epoll_event __user *event)
526 {
527 int error;
528 struct file *file, *tfile;
529 struct eventpoll *ep;
530 struct epitem *epi;
531 struct epoll_event epds;
....
575 epi = ep_find(ep, tfile, fd);
576
577 error = -EINVAL;
578 switch (op) {
579 case EPOLL_CTL_ADD:
580 if (!epi) {
581 epds.events |= POLLERR | POLLHUP;
582
583 error = ep_insert(ep, &epds, tfile, fd);
584 } else
585 error = -EEXIST;
586 break;
587 case EPOLL_CTL_DEL:
588 if (epi)
589 error = ep_remove(ep, epi);
590 else
591 error = -ENOENT;
592 break;
593 case EPOLL_CTL_MOD:
594 if (epi) {
595 epds.events |= POLLERR | POLLHUP;
596 error = ep_modify(ep, epi, &epds);
597 } else
598 error = -ENOENT;
599 break;
600 }
原来就是在一个大的结构(现在先不管是什么大结构)里先ep_find,如果找到了struct epitem而用户操
作是ADD,那么返回-EEXIST;如果是DEL,则ep_remove。如果找不到struct epitem而用户操作是
ADD,就ep_insert创建并插入一个。很直白。那这个“大结构”是什么呢?看ep_find的调用方式,ep参数
应该是指向这个“大结构”的指针,再看ep = file->private_data,我们才明白,原来这个“大结构”就是那
个在epoll_create时创建的struct eventpoll,具体再看看ep_find的实现,发现原来是struct eventpoll
的rbr成员(struct rb_root),原来这是一个红黑树的根!而红黑树上挂的都是struct epitem。
现在清楚了,一个新创建的epoll文件带有一个struct eventpoll结构,这个结构上再挂一个红黑树,而这
个红黑树就是每次epoll_ctl时fd存放的地方!
现在数据结构都已经清楚了,我们来看最核心的:
[fs/eventpoll.c-->sys_epoll_wait()]
627 asmlinkage long sys_epoll_wait(int epfd, struct epoll_event __user *events,
628 int maxevents, int timeout)
629 {
630 int error;
631 struct file *file;
632 struct eventpoll *ep;
633
634 DNPRINTK(3, (KERN_INFO "[%p] eventpoll: sys_epoll_wait(%d, %p, %d, %d)\n",
635 current, epfd, events, maxevents, timeout));
636
637 /* The maximum number of event must be greater than zero */
638 if (maxevents <= 0)
639 return -EINVAL;
640
641 /* Verify that the area passed by the user is writeable */
642 if ((error = verify_area(VERIFY_WRITE, events, maxevents * sizeof(struct
epoll_event))))
643 goto eexit_1;
644
645 /* Get the "struct file *" for the eventpoll file */
646 error = -EBADF;
647 file = fget(epfd);
648 if (!file)
649 goto eexit_1;
650
651 /*
652 * We have to check that the file structure underneath the fd
653 * the user passed to us _is_ an eventpoll file.
654 */
655 error = -EINVAL;
656 if (!IS_FILE_EPOLL(file))
657 goto eexit_2;
658
659 /*
660 * At this point it is safe to assume that the "private_data" contains
661 * our own data structure.
662 */
663 ep = file->private_data;
664
665 /* Time to fish for events ... */
666 error = ep_poll(ep, events, maxevents, timeout);
667
668 eexit_2:
669 fput(file);
670 eexit_1:
671 DNPRINTK(3, (KERN_INFO "[%p] eventpoll: sys_epoll_wait(%d, %p, %d, %d) =
%d\n",
672 current, epfd, events, maxevents, timeout, error));
673
674 return error;
675 }
故伎重演,从file->private_data中拿到struct eventpoll,再调用ep_poll
[fs/eventpoll.c-->sys_epoll_wait()->ep_poll()]
1468 static int ep_poll(struct eventpoll *ep, struct epoll_event __user *events,
1469 int maxevents, long timeout)
1470 {
1471 int res, eavail;
1472 unsigned long flags;
1473 long jtimeout;
1474 wait_queue_t wait;
1475
1476 /*
1477 * Calculate the timeout by checking for the "infinite" value ( -1 )
1478 * and the overflow condition. The passed timeout is in milliseconds,
1479 * that why (t * HZ) / 1000.
1480 */
1481 jtimeout = timeout == -1 || timeout > (MAX_SCHEDULE_TIMEOUT - 1000) / HZ ?
1482 MAX_SCHEDULE_TIMEOUT: (timeout * HZ + 999) / 1000;
1483
1484 retry:
1485 write_lock_irqsave(&ep->lock, flags);
1486
1487 res = 0;
1488 if (list_empty(&ep->rdllist)) {
1489 /*
1490 * We don't have any available event to return to the caller.
1491 * We need to sleep here, and we will be wake up by
1492 * ep_poll_callback() when events will become available.
1493 */
1494 init_waitqueue_entry(&wait, current);
1495 add_wait_queue(&ep->wq, &wait);
1496
1497 for (;;) {
1498 /*
1499 * We don't want to sleep if the ep_poll_callback() sends us
1500 * a wakeup in between. That's why we set the task state
1501 * to TASK_INTERRUPTIBLE before doing the checks.
1502 */
1503 set_current_state(TASK_INTERRUPTIBLE);
1504 if (!list_empty(&ep->rdllist) || !jtimeout)
1505 break;
1506 if (signal_pending(current)) {
1507 res = -EINTR;
1508 break;
1509 }
1510
1511 write_unlock_irqrestore(&ep->lock, flags);
1512 jtimeout = schedule_timeout(jtimeout);
1513 write_lock_irqsave(&ep->lock, flags);
1514 }
1515 remove_wait_queue(&ep->wq, &wait);
1516
1517 set_current_state(TASK_RUNNING);
1518 }
....
又是一个大循环,不过这个大循环比poll的那个好,因为仔细一看——它居然除了睡觉和判断ep->rdllist
是否为空以外,啥也没做!
什么也没做当然效率高了,但到底是谁来让ep->rdllist不为空呢?
答案是ep_insert时设下的回调函数:
[fs/eventpoll.c-->sys_epoll_ctl()-->ep_insert()]
923 static int ep_insert(struct eventpoll *ep, struct epoll_event *event,
924 struct file *tfile, int fd)
925 {
926 int error, revents, pwake = 0;
927 unsigned long flags;
928 struct epitem *epi;
929 struct ep_pqueue epq;
930
931 error = -ENOMEM;
932 if (!(epi = EPI_MEM_ALLOC()))
933 goto eexit_1;
934
935 /* Item initialization follow here ... */
936 EP_RB_INITNODE(&epi->rbn);
937 INIT_LIST_HEAD(&epi->rdllink);
938 INIT_LIST_HEAD(&epi->fllink);
939 INIT_LIST_HEAD(&epi->txlink);
940 INIT_LIST_HEAD(&epi->pwqlist);
941 epi->ep = ep;
942 EP_SET_FFD(&epi->ffd, tfile, fd);
943 epi->event = *event;
944 atomic_set(&epi->usecnt, 1);
945 epi->nwait = 0;
946
947 /* Initialize the poll table using the queue callback */
948 epq.epi = epi;
949 init_poll_funcptr(&epq.pt, ep_ptable_queue_proc);
950
951 /*
952 * Attach the item to the poll hooks and get current event bits.
953 * We can safely use the file* here because its usage count has
954 * been increased by the caller of this function.
955 */
956 revents = tfile->f_op->poll(tfile, &epq.pt);
我们注意949行,其实就是
&(epq.pt)->qproc = ep_ptable_queue_proc;
紧接着 tfile->f_op->poll(tfile, &epq.pt)其实就是调用被监控文件(epoll里叫“target file”)的poll方
法,而这个poll其实就是调用poll_wait(还记得poll_wait吗?每个支持poll的设备驱动程序都要调用
的),最后就是调用ep_ptable_queue_proc。这是比较难解的一个调用关系,因为不是语言级的直接调
用。
ep_insert还把struct epitem放到struct file里的f_ep_links连表里,以方便查找,struct epitem里的
fllink就是担负这个使命的。
[fs/eventpoll.c-->ep_ptable_queue_proc()]
883 static void ep_ptable_queue_proc(struct file *file, wait_queue_head_t *whead,
884 poll_table *pt)
885 {
886 struct epitem *epi = EP_ITEM_FROM_EPQUEUE(pt);
887 struct eppoll_entry *pwq;
888
889 if (epi->nwait >= 0 && (pwq = PWQ_MEM_ALLOC())) {
890 init_waitqueue_func_entry(&pwq->wait, ep_poll_callback);
891 pwq->whead = whead;
892 pwq->base = epi;
893 add_wait_queue(whead, &pwq->wait);
894 list_add_tail(&pwq->llink, &epi->pwqlist);
895 epi->nwait++;
896 } else {
897 /* We have to signal that an error occurred */
898 epi->nwait = -1;
899 }
900 }
上面的代码就是ep_insert中要做的最重要的事:创建struct eppoll_entry,设置其唤醒回调函数为
ep_poll_callback,然后加入设备等待队列(注意这里的whead就是上一章所说的每个设备驱动都要带的
等待队列)。只有这样,当设备就绪,唤醒等待队列上的等待着时,ep_poll_callback就会被调用。每次
调用poll系统调用,操作系统都要把current(当前进程)挂到fd对应的所有设备的等待队列上,可以想
象,fd多到上千的时候,这样“挂”法很费事;而每次调用epoll_wait则没有这么罗嗦,epoll只在epoll_ctl
时把current挂一遍(这第一遍是免不了的)并给每个fd一个命令“好了就调回调函数”,如果设备有事件
了,通过回调函数,会把fd放入rdllist,而每次调用epoll_wait就只是收集rdllist里的fd就可以了
——epoll巧妙的利用回调函数,实现了更高效的事件驱动模型。
现在我们猜也能猜出来ep_poll_callback会干什么了——肯定是把红黑树上的收到event的epitem(代表
每个fd)插入ep->rdllist中,这样,当epoll_wait返回时,rdllist里就都是就绪的fd了!
[fs/eventpoll.c-->ep_poll_callback()]
1206 static int ep_poll_callback(wait_queue_t *wait, unsigned mode, int sync, void *key)
1207 {
1208 int pwake = 0;
1209 unsigned long flags;
1210 struct epitem *epi = EP_ITEM_FROM_WAIT(wait);
1211 struct eventpoll *ep = epi->ep;
1212
1213 DNPRINTK(3, (KERN_INFO "[%p] eventpoll: poll_callback(%p) epi=%p
ep=%p\n",
1214 current, epi->file, epi, ep));
1215
1216 write_lock_irqsave(&ep->lock, flags);
1217
1218 /*
1219 * If the event mask does not contain any poll(2) event, we consider the
1220 * descriptor to be disabled. This condition is likely the effect of the
1221 * EPOLLONESHOT bit that disables the descriptor when an event is received,
1222 * until the next EPOLL_CTL_MOD will be issued.
1223 */
1224 if (!(epi->event.events & ~EP_PRIVATE_BITS))
1225 goto is_disabled;
1226
1227 /* If this file is already in the ready list we exit soon */
1228 if (EP_IS_LINKED(&epi->rdllink))
1229 goto is_linked;
1230
1231 list_add_tail(&epi->rdllink, &ep->rdllist);
1232
1233 is_linked:
1234 /*
1235 * Wake up ( if active ) both the eventpoll wait list and the ->poll()
1236 * wait list.
1237 */
1238 if (waitqueue_active(&ep->wq))
1239 wake_up(&ep->wq);
1240 if (waitqueue_active(&ep->poll_wait))
1241 pwake++;
1242
1243 is_disabled:
1244 write_unlock_irqrestore(&ep->lock, flags);
1245
1246 /* We have to call this outside the lock */
1247 if (pwake)
1248 ep_poll_safewake(&psw, &ep->poll_wait);
1249
1250 return 1;
1251 }
真正重要的只有1231行的只一句,就是把struct epitem放到struct eventpoll的rdllist中去。现在我们
可以画出epoll的核心数据结构图了:


epoll独有的EPOLLET


EPOLLET是epoll系统调用独有的flag,ET就是Edge Trigger(边缘触发)的意思,具体含义和应用大家
可google之。有了EPOLLET,重复的事件就不会总是出来打扰程序的判断,故而常被使用。那EPOLLET
的原理是什么呢?
上篇我们讲到epoll把fd都挂上一个回调函数,当fd对应的设备有消息时,就把fd放入rdllist链表,这样
epoll_wait只要检查这个rdllist链表就可以知道哪些fd有事件了。我们看看ep_poll的最后几行代码:
[fs/eventpoll.c->ep_poll()]
1524
1525 /*
1526 * Try to transfer events to user space. In case we get 0 events and
1527 * there's still timeout left over, we go trying again in search of
1528 * more luck.
1529 */
1530 if (!res && eavail &&
1531 !(res = ep_events_transfer(ep, events, maxevents)) && jtimeout)
1532 goto retry;
1533
1534 return res;
1535 }
把rdllist里的fd拷到用户空间,这个任务是ep_events_transfer做的:
[fs/eventpoll.c->ep_events_transfer()]
1439 static int ep_events_transfer(struct eventpoll *ep,
1440 struct epoll_event __user *events, int maxevents)
1441 {
1442 int eventcnt = 0;
1443 struct list_head txlist;
1444
1445 INIT_LIST_HEAD(&txlist);
1446
1447 /*
1448 * We need to lock this because we could be hit by
1449 * eventpoll_release_file() and epoll_ctl(EPOLL_CTL_DEL).
1450 */
1451 down_read(&ep->sem);
1452
1453 /* Collect/extract ready items */
1454 if (ep_collect_ready_items(ep, &txlist, maxevents) > 0) {
1455 /* Build result set in userspace */
1456 eventcnt = ep_send_events(ep, &txlist, events);
1457
1458 /* Reinject ready items into the ready list */
1459 ep_reinject_items(ep, &txlist);
1460 }
1461
1462 up_read(&ep->sem);
1463
1464 return eventcnt;
1465 }
代码很少,其中ep_collect_ready_items把rdllist里的fd挪到txlist里(挪完后rdllist就空了),接着
ep_send_events把txlist里的fd拷给用户空间,然后ep_reinject_items把一部分fd从txlist里“返还”给
rdllist以便下次还能从rdllist里发现它。
其中ep_send_events的实现:
[fs/eventpoll.c->ep_send_events()]
1337 static int ep_send_events(struct eventpoll *ep, struct list_head *txlist,
1338 struct epoll_event __user *events)
1339 {
1340 int eventcnt = 0;
1341 unsigned int revents;
1342 struct list_head *lnk;
1343 struct epitem *epi;
1344
1345 /*
1346 * We can loop without lock because this is a task private list.
1347 * The test done during the collection loop will guarantee us that
1348 * another task will not try to collect this file. Also, items
1349 * cannot vanish during the loop because we are holding "sem".
1350 */
1351 list_for_each(lnk, txlist) {
1352 epi = list_entry(lnk, struct epitem, txlink);
1353
1354 /*
1355 * Get the ready file event set. We can safely use the file
1356 * because we are holding the "sem" in read and this will
1357 * guarantee that both the file and the item will not vanish.
1358 */
1359 revents = epi->ffd.file->f_op->poll(epi->ffd.file, NULL);
1360
1361 /*
1362 * Set the return event set for the current file descriptor.
1363 * Note that only the task task was successfully able to link
1364 * the item to its "txlist" will write this field.
1365 */
1366 epi->revents = revents & epi->event.events;
1367
1368 if (epi->revents) {
1369 if (__put_user(epi->revents,
1370 &events[eventcnt].events) ||
1371 __put_user(epi->event.data,
1372 &events[eventcnt].data))
1373 return -EFAULT;
1374 if (epi->event.events & EPOLLONESHOT)
1375 epi->event.events &= EP_PRIVATE_BITS;
1376 eventcnt++;
1377 }
1378 }
1379 return eventcnt;
1380 }
这个拷贝实现其实没什么可看的,但是请注意1359行,这个poll很狡猾,它把第二个参数置为NULL来调
用。我们先看一下设备驱动通常是怎么实现poll的:
static unsigned int scull_p_poll(struct file *filp, poll_table *wait)
{
struct scull_pipe *dev = filp->private_data;
unsigned int mask = 0;
/*
* The buffer is circular; it is considered full
* if "wp" is right behind "rp" and empty if the
* two are equal.
*/
down(&dev->sem);
poll_wait(filp, &dev->inq, wait);
poll_wait(filp, &dev->outq, wait);
if (dev->rp != dev->wp)
mask |= POLLIN | POLLRDNORM; /* readable */
if (spacefree(dev))
mask |= POLLOUT | POLLWRNORM; /* writable */
up(&dev->sem);
return mask;
}
上面这段代码摘自《linux设备驱动程序(第三版)》,绝对经典,设备先要把current(当前进程)挂在
inq和outq两个队列上(这个“挂”操作是wait回调函数指针做的),然后等设备来唤醒,唤醒后就能通过
mask拿到事件掩码了(注意那个mask参数,它就是负责拿事件掩码的)。那如果wait为NULL,
poll_wait会做些什么呢?
[include/linux/poll.h->poll_wait]
25 static inline void poll_wait(struct file * filp, wait_queue_head_t * wait_address,
poll_table *p)
26 {
27 if (p && wait_address)
28 p->qproc(filp, wait_address, p);
29 }
喏,看见了,如果poll_table为空,什么也不做。我们倒回ep_send_events,那句标红的poll,实际上
就是“我不想休眠,我只想拿到事件掩码”的意思。然后再把拿到的事件掩码拷给用户空间。
ep_send_events完成后,就轮到ep_reinject_items了:
[fs/eventpoll.c->ep_reinject_items]
1389 static void ep_reinject_items(struct eventpoll *ep, struct list_head *txlist)
1390 {
1391 int ricnt = 0, pwake = 0;
1392 unsigned long flags;
1393 struct epitem *epi;
1394
1395 write_lock_irqsave(&ep->lock, flags);
1396
1397 while (!list_empty(txlist)) {
1398 epi = list_entry(txlist->next, struct epitem, txlink);
1399
1400 /* Unlink the current item from the transfer list */
1401 EP_LIST_DEL(&epi->txlink);
1402
1403 /*
1404 * If the item is no more linked to the interest set, we don't
1405 * have to push it inside the ready list because the following
1406 * ep_release_epitem() is going to drop it. Also, if the current
1407 * item is set to have an Edge Triggered behaviour, we don't have
1408 * to push it back either.
1409 */
1410 if (EP_RB_LINKED(&epi->rbn) && !(epi->event.events & EPOLLET) &&
1411 (epi->revents & epi->event.events) && !EP_IS_LINKED(&epi->rdllink)) {
1412 list_add_tail(&epi->rdllink, &ep->rdllist);
1413 ricnt++;
1414 }
1415 }
1416
1417 if (ricnt) {
1418 /*
1419 * Wake up ( if active ) both the eventpoll wait list and the ->poll()
1420 * wait list.
1421 */
1422 if (waitqueue_active(&ep->wq))
1423 wake_up(&ep->wq);
1424 if (waitqueue_active(&ep->poll_wait))
1425 pwake++;
1426 }
1427
1428 write_unlock_irqrestore(&ep->lock, flags);
1429
1430 /* We have to call this outside the lock */
1431 if (pwake)
1432 ep_poll_safewake(&psw, &ep->poll_wait);
1433 }
ep_reinject_items把txlist里的一部分fd又放回rdllist,那么,是把哪一部分fd放回去呢?看上面1410行
的那个判断——是哪些“没有标上EPOLLET”(标红代码)且“事件被关注”(标蓝代码)的fd被重新放回了
rdllist。那么下次epoll_wait当然会又把rdllist里的fd拿来拷给用户了。
举个例子。假设一个socket,只是connect,还没有收发数据,那么它的poll事件掩码总是有POLLOUT的
(参见上面的驱动示例),每次调用epoll_wait总是返回POLLOUT事件(比较烦),因为它的fd就总是被
放回rdllist;假如此时有人往这个socket里写了一大堆数据,造成socket塞住(不可写了),那么1411行
里标蓝色的判断就不成立了(没有POLLOUT了),fd不会放回rdllist,epoll_wait将不会再返回用户
POLLOUT事件。现在我们给这个socket加上EPOLLET,然后connect,没有收发数据,此时,1410行标
红的判断又不成立了,所以epoll_wait只会返回一次POLLOUT通知给用户(因为此fd不会再回到rdllist
了),接下来的epoll_wait都不会有任何事件通知了。


原文链接:http://blog.csdn.net/21aspnet/article/details/2627662
加载中
返回顶部
顶部