NAME

kcmp - compare two processes to determine if they share a kernel resource

LIBRARY

Standard C library (libc, -lc)

SYNOPSIS

#include <linux/kcmp.h> /* Definition of KCMP_* constants */
#include <sys/syscall.h> /* Definition of SYS_* constants */
#include <unistd.h>
int syscall(SYS_kcmp, pid_t pid1, pid_t pid2, int type,
 unsigned long idx1, unsigned long idx2);

Note: glibc provides no wrapper for kcmp(), necessitating the use of syscall(2).

DESCRIPTION

The kcmp() system call can be used to check whether the two processes identified by pid1 and pid2 share a kernel resource such as virtual memory, file descriptors, and so on.

Permission to employ kcmp() is governed by ptrace access mode PTRACE_MODE_READ_REALCREDS checks against both pid1 and pid2; see ptrace(2).

The type argument specifies which resource is to be compared in the two processes. It has one of the following values:

KCMP_FILE

Check whether a file descriptor idx1 in the process pid1 refers to the same open file description (see open(2)) as file descriptor idx2 in the process pid2. The existence of two file descriptors that refer to the same open file description can occur as a result of dup(2) (and similar) fork(2), or passing file descriptors via a domain socket (see unix(7)).

KCMP_FILES

Check whether the processes share the same set of open file descriptors. The arguments idx1 and idx2 are ignored. See the discussion of the CLONE_FILES flag in clone(2).

KCMP_FS

Check whether the processes share the same filesystem information (i.e., file mode creation mask, working directory, and filesystem root). The arguments idx1 and idx2 are ignored. See the discussion of the CLONE_FS flag in clone(2).

KCMP_IO

Check whether the processes share I/O context. The arguments idx1 and idx2 are ignored. See the discussion of the CLONE_IO flag in clone(2).

KCMP_SIGHAND

Check whether the processes share the same table of signal dispositions. The arguments idx1 and idx2 are ignored. See the discussion of the CLONE_SIGHAND flag in clone(2).

KCMP_SYSVSEM

Check whether the processes share the same list of System V semaphore undo operations. The arguments idx1 and idx2 are ignored. See the discussion of the CLONE_SYSVSEM flag in clone(2).

KCMP_VM

Check whether the processes share the same address space. The arguments idx1 and idx2 are ignored. See the discussion of the CLONE_VM flag in clone(2).

KCMP_EPOLL_TFD (since Linux 4.13)

Check whether the file descriptor idx1 of the process pid1 is present in the epoll(7) instance described by idx2 of the process pid2. The argument idx2 is a pointer to a structure where the target file is described. This structure has the form:

struct kcmp_epoll_slot {
    __u32 efd;
    __u32 tfd;
    __u64 toff;
};

Within this structure, efd is an epoll file descriptor returned from epoll_create(2), tfd is a target file descriptor number, and toff is a target file offset counted from zero. Several different targets may be registered with the same file descriptor number and setting a specific offset helps to investigate each of them.

Note the kcmp() is not protected against false positives which may occur if the processes are currently running. One should stop the processes by sending SIGSTOP (see signal(7)) prior to inspection with this system call to obtain meaningful results.

RETURN VALUE

The return value of a successful call to kcmp() is simply the result of arithmetic comparison of kernel pointers (when the kernel compares resources, it uses their memory addresses).

The easiest way to explain is to consider an example. Suppose that v1 and v2 are the addresses of appropriate resources, then the return value is one of the following:

0

v1 is equal to v2; in other words, the two processes share the resource.

1

v1 is less than v2.

2

v1 is greater than v2.

3

v1 is not equal to v2, but ordering information is unavailable.

On error, -1 is returned, and errno is set to indicate the error.

kcmp() was designed to return values suitable for sorting. This is particularly handy if one needs to compare a large number of file descriptors.

EXAMPLES

The program below uses kcmp() to test whether pairs of file descriptors refer to the same open file description. The program tests different cases for the file descriptor pairs, as described in the program output. An example run of the program is as follows:

$ ./a.out
Parent PID is 1144
Parent opened file on FD 3
PID of child of fork() is 1145
	Compare duplicate FDs from different processes:
		kcmp(1145, 1144, KCMP_FILE, 3, 3) ==> same
Child opened file on FD 4
	Compare FDs from distinct open()s in same process:
		kcmp(1145, 1145, KCMP_FILE, 3, 4) ==> different
Child duplicated FD 3 to create FD 5
	Compare duplicated FDs in same process:
		kcmp(1145, 1145, KCMP_FILE, 3, 5) ==> same

Program source

#define _GNU_SOURCE
#include <err.h>
#include <fcntl.h>
#include <linux/kcmp.h>
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include <sys/syscall.h>
#include <sys/wait.h>
#include <unistd.h>
static int
kcmp(pid_t pid1, pid_t pid2, int type,
     unsigned long idx1, unsigned long idx2)
{
    return syscall(SYS_kcmp, pid1, pid2, type, idx1, idx2);
}
static void
test_kcmp(char *msg, pid_t pid1, pid_t pid2, int fd_a, int fd_b)
{
    printf("\t%s\n", msg);
    printf("\t\tkcmp(%jd, %jd, KCMP_FILE, %d, %d) ==> %s\n",
           (intmax_t) pid1, (intmax_t) pid2, fd_a, fd_b,
           (kcmp(pid1, pid2, KCMP_FILE, fd_a, fd_b) == 0) ?
                        "same" : "different");
}
int
main(void)
{
    int                fd1, fd2, fd3;
    static const char  pathname[] = "/tmp/kcmp.test";
    fd1 = open(pathname, O_CREAT | O_RDWR, 0600);
    if (fd1 == -1)
        err(EXIT_FAILURE, "open");
    printf("Parent PID is %jd\n", (intmax_t) getpid());
    printf("Parent opened file on FD %d\n\n", fd1);
    switch (fork()) {
    case -1:
        err(EXIT_FAILURE, "fork");
    case 0:
        printf("PID of child of fork() is %jd\n", (intmax_t) getpid());
        test_kcmp("Compare duplicate FDs from different processes:",
                  getpid(), getppid(), fd1, fd1);
        fd2 = open(pathname, O_CREAT | O_RDWR, 0600);
        if (fd2 == -1)
            err(EXIT_FAILURE, "open");
        printf("Child opened file on FD %d\n", fd2);
        test_kcmp("Compare FDs from distinct open()s in same process:",
                  getpid(), getpid(), fd1, fd2);
        fd3 = dup(fd1);
        if (fd3 == -1)
            err(EXIT_FAILURE, "dup");
        printf("Child duplicated FD %d to create FD %d\n", fd1, fd3);
        test_kcmp("Compare duplicated FDs in same process:",
                  getpid(), getpid(), fd1, fd3);
        break;
    default:
        wait(NULL);
    }
    exit(EXIT_SUCCESS);
}

ERRORS

EBADF

type is KCMP_FILE and fd1 or fd2 is not an open file descriptor.

EFAULT

The epoll slot addressed by idx2 is outside of the user's address space.

EINVAL

type is invalid.

ENOENT

The target file is not present in epoll(7) instance.

EPERM

Insufficient permission to inspect process resources. The CAP_SYS_PTRACE capability is required to inspect processes that you do not own. Other ptrace limitations may also apply, such as CONFIG_SECURITY_YAMA, which, when /proc/sys/kernel/yama/ptrace_scope is 2, limits kcmp() to child processes; see ptrace(2).

ESRCH

Process pid1 or pid2 does not exist.

STANDARDS

Linux.

HISTORY

Linux 3.5.

Before Linux 5.12, this system call is available only if the kernel is configured with CONFIG_CHECKPOINT_RESTORE, since the original purpose of the system call was for the checkpoint/restore in user space (CRIU) feature. (The alternative to this system call would have been to expose suitable process information via the proc(5) filesystem; this was deemed to be unsuitable for security reasons.) Since Linux 5.12, this system call is also available if the kernel is configured with CONFIG_KCMP.

NOTES

See clone(2) for some background information on the shared resources referred to on this page.

SEE ALSO

clone(2), unshare(2)