Re: tmpfs inode leakage when opening file with O_TMP_FILE
From: Andrew Morton
Date: Thu Feb 14 2019 - 18:44:07 EST
On Mon, 11 Feb 2019 15:18:11 +0100 Matej Kupljen <matej.kupljen@xxxxxxxxx> wrote:
> it seems that when opening file on file system that is mounted on
> tmpfs with the O_TMPFILE flag and using linkat call after that, it
> uses 2 inodes instead of 1.
> This is simple test case:
> #include <sys/types.h>
> #include <sys/stat.h>
> #include <fcntl.h>
> #include <unistd.h>
> #include <string.h>
> #include <stdio.h>
> #include <stdlib.h>
> #include <linux/limits.h>
> #include <errno.h>
> #define TEST_STRING "Testing\n"
> #define TMP_PATH "/tmp/ping/"
> #define TMP_FILE "file.txt"
> int main(int argc, char* argv)
> char path[PATH_MAX];
> int fd;
> int rc;
> fd = open(TMP_PATH, __O_TMPFILE | O_RDWR,
> S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP |
> S_IROTH | S_IWOTH);
> rc = write(fd, TEST_STRING, strlen(TEST_STRING));
> snprintf(path, PATH_MAX, "/proc/self/fd/%d", fd);
> linkat(AT_FDCWD, path, AT_FDCWD, TMP_PATH TMP_FILE, AT_SYMLINK_FOLLOW);
> return 0;
> I have checked indoes with "df -i" tool. The first inode is used when
> the call to open is executed and the second one when the call to
> linkat is executed.
> It is not decreased when close is executed.
> I have also tested this on an ext4 mounted fs and there only one inode is used.
> I tested this on:
> $ cat /etc/lsb-release
> DISTRIB_DESCRIPTION="Ubuntu 18.04.1 LTS"
> $ uname -a
> Linux Orion 4.15.0-43-generic #46-Ubuntu SMP Thu Dec 6 14:45:28 UTC
> 2018 x86_64 x86_64 x86_64 GNU/Linux
> If you need any more information, please let me know.
> And please CC me when replying, I am not subscribed to the list.
> Thanks and BR,