btrfs-progs: mounted filesystems checks
cover btrfs-progs: mounted filesystems checks
The following two patches add code that prevents btrfsck to run on mounted filesystems. The first patch makes check_mounted() work with multidevice filesystems. The second patch finally adds the check to btrfsck.
Cheers,
Andi
PATCH 1/2 btrfs-progs: multidevice support for check_mounted
Check_mount() should also work with multi device filesystems. This patch adds checks that allow to detect if a file is a device file used by a mounted single or multi device btrfs or if it is a regular file used by a loopback device that is part of a mounted single or multi device btrfs.
The single device checks also work for non-btrfs filesystems. This might be helpful to prevent users from running btrfs programs (e.g. mkfs.btrfs) accidentally on a filesystem used somewhere else.
---
kerncompat.h +4 -0
@@ -42,7 +42,11 @@#define GFP_NOFS 0#define __read_mostly#define ARRAY_SIZE (x) (sizeof(x) / sizeof((x)[0])) + +#ifndef ULONG_MAX #define ULONG_MAX (~0UL) +#endif +#define BUG () abort()#ifdef __CHECKER__#define __force __attribute__((force))
utils.c +206 -33
267 lines not on this page · raw message
utils.h +1 -1
@@ -36,7 +36,7 @@ int btrfs_scan_for_fsid(struct btrfs_fs_devices *fs_devices, u64 total_devs, int run_ioctls);void btrfs_register_one_device (char *fname);int btrfs_scan_one_dir (char *dirname,int run_ioctl); -int check_mounted (char *devicename); +int check_mounted (const char *devicename);int btrfs_device_already_in_root (struct btrfs_root *root,int fd,int super_offset);char *pretty_sizes (u64 size);
PATCH 2/2 btrfs-progs: prevent btrfsck to run on mounted filesystems
In reply to Andi DrebesAs recently discussed on the list, btrfsck should only be run on unmounted filesystems. This patch adds a short check for the mount status at the beginning of btrfsck. If the FS is mounted, the program aborts showing an error message.
---
btrfsck.c +10 -0
@@ -28,6 +28,7 @@#include "transaction.h" #include "list.h" #include "version.h" +#include "utils.h" static u64 bytes_used =0 ;static u64 total_csum_bytes =0 ;
@@ -2821,6 +2822,15 @@ int main(int ac, char **av)radix_tree_init ();cache_tree_init (&root_cache); + +if ((ret =check_mounted (av[1 ])) <0 ) { +fprintf (stderr,"Could not check mount status: %s\n" ,strerror (ret)); +return ret; + }else if (ret) { +fprintf (stderr,"%s is currently mounted. Aborting.\n" , av[1 ]); +return -EBUSY ; + } + root =open_ctree (av[1 ],0 ,0 );if (root ==NULL )
Re: [PATCH 1/2] btrfs-progs: multidevice support for check_mounted
In reply to Andi DrebesQuoted 5 lines · Andi Drebes
utils.c +5
-0+ +/* Checks if an mntentry represents a pseudo FS */ +int is_pseudo_fs (const struct mntent * mnt) +{ +struct stat st_buf;
utils.c error: · line 723
Quoted 10 lines
utils.c +10
-0+ +if (stat (mnt->mnt_fsname, &st_buf) <0 ) { +if (errno ==ENOENT ) +return 1 ; +else +return -errno; + } + +return 0 ; +}
This is bad idea. The mnt_fsname field could be an arbitrary string include valid paths.
# grep sysfs /proc/mounts
/sys /sys sysfs rw,relatime 0 0Karel
Signature · 1 line
Karel Zak <[email protected]>
Re: [PATCH 1/2] btrfs-progs: multidevice support for check_mounted
Hi!
Thanks for the reply!
Quoted 19 lines
Quoted 5 lines
(quoted diff) +5
-0+ +/* Checks if an mntentry represents a pseudo FS */ +int is_pseudo_fs(const struct mntent* mnt) +{ + struct stat st_buf;Review
Quoted 10 lines
(quoted diff) +10
-0+ + if(stat(mnt->mnt_fsname, &st_buf) < 0) { + if(errno == ENOENT) + return 1; + else + return -errno; + } + + return 0; +}This is bad idea. The
mnt_fsnamefield could be an arbitrary string include valid paths.
Quoted 2 lines
# grep sysfs /proc/mounts
/sys /sys sysfs rw,relatime 0 0Yes, indeed. Although this doesn't really cause problems (because if the string is a valid path, we would just do one check too much), the name of the function doesn't really correspond to what it does. In the new patch below, is_pseudo_fs() is replaced by is_existing_blk_or_reg_file(). We ignore entries associated with an invalid path or paths that don't point to a regular or block file. However, if a path used in a pseudo-filesystem entry points to the file that is being checked, check_mounted() returns 1. In my eyes, this is extremely unlikely.
Here's the new patch:
---
kerncompat.h +4 -0
@@ -42,7 +42,11 @@#define GFP_NOFS 0#define __read_mostly#define ARRAY_SIZE (x) (sizeof(x) / sizeof((x)[0])) + +#ifndef ULONG_MAX #define ULONG_MAX (~0UL) +#endif +#define BUG () abort()#ifdef __CHECKER__#define __force __attribute__((force))
utils.c +205 -32
266 lines not on this page · raw message
utils.h +1 -1
@@ -36,7 +36,7 @@ int btrfs_scan_for_fsid(struct btrfs_fs_devices *fs_devices, u64 total_devs, int run_ioctls);void btrfs_register_one_device (char *fname);int btrfs_scan_one_dir (char *dirname,int run_ioctl); -int check_mounted (char *devicename); +int check_mounted (const char *devicename);int btrfs_device_already_in_root (struct btrfs_root *root,int fd,int super_offset);char *pretty_sizes (u64 size);
Re: [PATCH 1/2] btrfs-progs: multidevice support for check_mounted
Quoted 6 lines · Andi Drebes
In the new patch below,
is_pseudo_fs()is replaced byis_existing_blk_or_reg_file(). We ignore entries associated with an invalid path or paths that don't point to a regular or block file. However, if a path used in a pseudo-filesystem entry points to the file that is being checked,check_mounted()returns 1. In my eyes, this is extremely unlikely.
...
Quoted 4 lines
utils.c +4
-0+/* Checks if a file exists and is a block or regular file*/ +int is_existing_blk_or_reg_file (const char * filename) +{ +struct stat st_buf;
utils.c error: · line 723
Quoted 10 lines
utils.c +10
-0+ +if (stat (filename, &st_buf) <0 ) { +if (errno ==ENOENT ) +return 0 ; +else +return -errno; + } + +return (S_ISBLK (st_buf.st_mode) ||S_ISREG (st_buf.st_mode)); +}
Yes, this looks better.
Karel
Signature · 1 line
Karel Zak <[email protected]>