real uid, effective uid
So, the real user id is who you really are (the one who owns the process), and the effective user id is what the operating system looks at to make a decision whether or not you are allowed to do something (most of the time, there are some exceptions).
贴一个试验代码, 子进程直接获取锁, 若获取不到则输出错误; 父进程睡3秒后退出.
- 如果该文件是自己创建的, 无法获取锁, 且错误为
Resource temporarily unavailable. - 如果该文件是进程创建的, 可以获取到锁.
- 在本例中, 我们让
aa.txt由进程创建, 让a.txt由用户创建. 则前者可以获取锁, 后者不能, 本文将分析原因.
#include
#include
#include
#include
#include
#include
#include
#include
extern int errno ;
int main() {
int fd = open("a.txt", O_CREAT| O_RDWR);
pid_t ret = fork();
if (ret == 0) {
// child lock ex.
printf("child try get ex lock.\n");
uid_t uid = getuid();
uid_t euid = geteuid();
printf("%d %d\n", uid, euid);
if (flock(fd, LOCK_EX | LOCK_NB) != 0) {
printf("lock ex failed.\n");
} else {
printf("child get ex lock.\n");
}
int err = errno;
printf("error no. %d\n", err);
printf("err msg: %s.\n", strerror( err ));
// fprintf(stdout, "Error opening file: %s\n", strerror( errno ));
while (1) {
printf("child sleep 10s.\n");
sleep(10);
}
return 0; // 直接退出, 此时父进程仍持有锁.
} else {
// parent lock shared.
sleep(3);
return 0;
// if (flock(fd, LOCK_SH) != 0) {
// printf("lock sh failed.\n");
// return 0;
// }
// printf("parent get shared lock.");
// fflush(stdout);
// while(1) {
// printf("parent sleep 10s\n");
// sleep(20);
// }
}
}
输出为:
child try get ex lock.
1000 1000
lock ex failed.
error no. 11
err msg: Resource temporarily unavailable.
child sleep 10s.
查看可知进程用户是自己:
[~/codes]$ id rasak
uid=1000(rasak) gid=1000(rasak) 组=1000(rasak),4(adm),24(cdrom),27(sudo),30(dip),46(plugdev),106(input),113(lpadmin),128(sambashare),999(docker)
权限问题?
我一度怀疑是权限问题, 导致无法获取锁, 于是查阅了不少关于权限的资料.
查看文件权限, 值得在意的是s, 和T. 查阅文献得知s是指setuid, T指sticky bit:
Real, Effective and Saved UserID in Linux 详细讲了三者的作用.
setuid
当执行该文件时, 执行者会拥有root权限. 如果让该文件能被所有用户执行, 就可以让所有用户以root身份去执行该文件的指令. 比如sudo:
root@host [~]# ls -l /usr/bin/sudo
-rwsr-xr-x 1 root root 136808 Jan 31 13:37 /usr/bin/sudo
它的执行内容需要root权限, 但它能被所有人执行.
setuid是否有权限:
If uid is the same as the real UID or the saved set-user-ID of the process, setuid() always succeeds and sets the effective UID. the real user ID and saved set-user-ID will remain unchanged.
Sticky BIT 权限: 总结一句话作用, 就是在文件上设置, 防止被文件夹写权限者误删.
---srwx--T 1 rasak rasak 0 2月 25 11:28 aa.txt
-rw-rw-r-- 1 rasak rasak 8 2月 24 05:22 a.txt
What does directory permission 'S' mean? (not lower case, but in upper case), 大写S代表setgid被设置, 小写s代表它还有组执行权限(也就是S+x). 应该t也同理.
'S' = The directory's setgid bit is set, but the execute bit isn't set.
's' = The directory's setgid bit is set, and the execute bit is set.
chmod用法总结了一些常用做法, 这里也记录以下:
所有人都获取读权限, 此处a代表所有人, r代表读权限:
chmod a+r file1.txt
效果等于如下, 其中u代表文件拥有者, g代表同群组, o代表其它群组的人:
chmod ugo+r file1.txt
将档案file1.txt和file2.txt设为该档案拥有者, 与其所属同一个群体者可写入, 带其它人不可写入:
chmod ug+w,o-w file1.txt file2.txt
进程问题
其实原因是父进程退出时, 没有发送信号给子进程让其终止, 导致后者成为了孤儿进程.
批量查找删除进程可用如下命令(另见xargs命令):
ps aux | grep -i process_name_to_kill | awk '{print $2}' | xargs sudo kill -9
另一种可参照Linux下批量杀掉筛选进程
ps -ef| grep override |grep -v grep |awk '{print $2}' | xargs kill -9
问题总结
与文件的执行权限并无关系, 之所以无法获取锁, 只是因为父进程退出后, 没有子进程变为孤儿进程, 且没有退出.
实验
子进程获取ex锁后, 父进程获取sh锁会成功, 并覆盖为sh锁.
此时启动另一个进程尝试获取sh锁会成功.
override_flock.c
#include
#include
#include
#include
#include
#include
#include
#include
extern int errno ;
int main() {
int fd = open("a.txt", O_CREAT| O_RDWR);
pid_t ret = fork();
if (ret == 0) {
// child lock ex.
printf("child try get ex lock.\n");
if (flock(fd, LOCK_EX | LOCK_NB) != 0) {
printf("lock ex failed.\n");
} else {
printf("child get ex lock.\n");
}
printf("child sleep 10s.\n");
sleep(15);
printf("child exit.\n");
} else {
// parent lock shared.
sleep(1);
if (flock(fd, LOCK_SH) != 0) {
printf("lock sh failed.\n");
return 0;
} else {
printf("parent get shared lock.\n");
}
printf("parent sleep 10s\n");
sleep(15);
}
}
override_flock_wait.c
#include
#include
#include
#include
#include
#include
#include
int main() {
printf("wait_process start.\n");
int fd = open("a.txt", O_RDONLY);
char *buf = (char *)malloc(sizeof(char) * 3);
flock(fd, LOCK_SH);
int ret = read(fd, buf, 5);
if (ret > 0) {
printf("read success\n");
printf(buf);
} else {
printf("read failed.\n");
}
}
实验结论
flock的锁视为持有人是open file description, 当fork后持有相同open file description的进程先后调用flock, 则都会成功, 且视为覆盖锁形式.
也就是说, 父进程获取ex锁后, 子进程获取sh锁, 则视为sh锁.