Details
-
Bug
-
Resolution: Fixed
-
Minor
-
Lustre 2.5.0, Lustre 2.7.0, Lustre 2.8.0
-
3
-
9223372036854775807
Description
After Shadow's patch for LU-2655 (see http://review.whamcloud.com/5135), in ldiskfs_write_ldd() if SELinux is enabled then the security context of the temporary mount point is appended to the mount options. Unfortunately this is done by appending the context to ldd_mount_opts and so it gets written into CONFIGS/mountdata.
# lsb_release -d
Description: CentOS Linux release 7.1.1503 (Core)
# git describe
v2_7_57_0-60-g0addfa9
# getenforce
Enforcing
# insmod ./ldiskfs/ldiskfs.ko
# dd if=/dev/zero of=/tmp/lustre-mdt1 bs=1K count=200000
200000+0 records in
200000+0 records out
204800000 bytes (205 MB) copied, 1.40722 s, 146 MB/s
# mkfs.lustre --fsname=lustre --mgs --mdt --index=0 /tmp/lustre-mdt1
Permanent disk data:
Target: lustre:MDT0000
Index: 0
Lustre FS: lustre
Mount type: ldiskfs
Flags: 0x65
(MDT MGS first_time update )
Persistent mount opts: user_xattr,errors=remount-ro
Parameters:
checking for existing Lustre data: not found
formatting backing filesystem ldiskfs on /dev/loop0
target name lustre:MDT0000
4k blocks 0
options -I 512 -i 2048 -q -O dirdata,uninit_bg,^extents,dir_nlink,quota,huge_file,flex_bg -E lazy_journal_init -F
mkfs_cmd = mke2fs -j -b 4096 -L lustre:MDT0000 -I 512 -i 2048 -q -O dirdata,uninit_bg,^extents,dir_nlink,quota,huge_file,flex_bg -E lazy_journal_init -F /dev/loop0
Writing CONFIGS/mountdata
# mount /tmp/lustre-mdt1 /mnt/mds1 -t ldiskfs -o loop
# hexdump -C /mnt/mds1/CONFIGS/mountdata
00000000 01 00 d0 1d 00 00 00 00 00 00 00 00 00 00 00 00 |................|
00000010 01 00 00 00 65 00 00 00 00 00 00 00 01 00 00 00 |....e...........|
00000020 6c 75 73 74 72 65 00 00 00 00 00 00 00 00 00 00 |lustre..........|
00000030 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 |................|
*
00000060 6c 75 73 74 72 65 3a 4d 44 54 30 30 30 30 00 00 |lustre:MDT0000..|
00000070 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 |................|
*
00001000 75 73 65 72 5f 78 61 74 74 72 2c 65 72 72 6f 72 |user_xattr,error|
00001010 73 3d 72 65 6d 6f 75 6e 74 2d 72 6f 2c 63 6f 6e |s=remount-ro,con|
00001020 74 65 78 74 3d 75 6e 63 6f 6e 66 69 6e 65 64 5f |text=unconfined_|
00001030 75 3a 6f 62 6a 65 63 74 5f 72 3a 75 73 65 72 5f |u:object_r:user_|
00001040 74 6d 70 5f 74 3a 73 30 00 00 00 00 00 00 00 00 |tmp_t:s0........|
00001050 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 |................|
*
00003000
# umount /mnt/mds1
Then when the device is mounted as lustre mount.lustre tries to clear the LDD_F_UPDATE flag from CONFIGS/mountdata, so we call ldiskfs_write_ldd() again which appends the context to the new temporary directory to ldd_mount_opts and uses the result as the options to mount the target as ldiskfs. This fails since the context option can only be specified once.
# LOAD=1 ./lustre/tests/llmount.sh
Loading modules from /root/lustre-release/lustre
...
# strace -f -o mount.strace -s 1024 mount /tmp/lustre-mdt1 /mnt/mds1 -t lustre -o loop
mount.lustre: Unable to mount /dev/loop0: Invalid argument
mount.lustre FATAL: failed to write local files: Invalid argument
# awk '$2 ~ /^mount/' mount.strace
8672 mount("/dev/loop0", "/tmp/mntACvy85", "ldiskfs", 0, "user_xattr,errors=remount-ro,context=unconfined_u:object_r:user_tmp_t:s0,context=unconfined_u:object_r:user_tmp_t:s0") = -1 EINVAL (Invalid argument)
8672 mount("/dev/loop0", "/mnt/mds1", "lustre", MS_STRICTATIME, "osd=osd-ldiskfs,user_xattr,errors=remount-ro,context=unconfined_u:object_r:user_tmp_t:s0,mgs,update,svname=lustre-MDT0000,device=/dev/loop0") = 0
mount.lustre does succeed but the LDD_F_UPDATE flag is never cleared.
# mount ... /tmp/lustre-mdt1 on /mnt/mds1 type lustre (ro,seclabel) # umount /mnt/mds1 # mount /tmp/lustre-mdt1 /mnt/mds1 -t lustre -o loop mount.lustre: Unable to mount /dev/loop0: Invalid argument mount.lustre FATAL: failed to write local files: Invalid argument # mount ... /tmp/lustre-mdt1 on /mnt/mds1 type lustre (ro,seclabel)
