Whamcloud - gitweb
tools/e2fsprogs.git
7 months agoLU-17117 build: update version to 1.47.0-wc5 v1.47.0-wc5
Li Dongyang [Wed, 27 Sep 2023 01:46:31 +0000 (11:46 +1000)]
LU-17117 build: update version to 1.47.0-wc5

b67310af4 LU-17117 journal: increase revoke block hash size
3ff538c1d LU-17117 misc: deduplicate log2/log10 functions
27d5daaad LU-15002 mke2fs: try to pack the GDT blocks together
03a6a831c LU-15002 mke2fs: do not set the BLOCK_UNINIT on groups has GDT
6e18f9e2c LU-15002 mke2fs: set free blocks accurately for groups has GDT
37440af03 LU-17117 debugfs: allow debugfs to disable MMP directly
d7e9c047d LU-15002 mke2fs: batch zeroing inode table
eb67ae2ec LU-15002 e2fsck: check all sparse_super backups
a59ac3441 LU-16750 tune2fs: add "-E iops" to set/clear IOPS groups
99deb0b3c LU-17028 build: add support for openEuler 20.03 LTS
4e82819ed LU-16169 e2fsck: improve parallel thread balance
3bf271a76 LU-14712 e2fsprogs: support EXT2_FLAG_BG_TRIMMED and EXT2_FLAGS_TRACK_TRIM

Change-Id: I8894011d9e63695c77b10c2460c826f5c1f28cab
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
7 months agoLU-17117 journal: increase revoke block hash size 86/52386/2
Andreas Dilger [Fri, 15 Sep 2023 08:08:09 +0000 (02:08 -0600)]
LU-17117 journal: increase revoke block hash size

Increase the size of the revoke block hash table to scale with the
size of the journal, so that we don't get long hash chains if there
are a large number of revoke blocks in the journal to replay.

The new hash size will default to 1/16 of the blocks in the journal.
This is about 1 byte per block in the hash table, but there are two
allocated.  The total amount of memory allocated for the revoke blocks
depends much more on how many are in the journal, and not on the size
of the hash table.  The system is regularly using this much memory for
the journal blocks, so the hash table size is not a big factor.

Consolidate duplicate code between recover_ext3_journal() and
ext2fs_open_ext3_journal() in debugfs.c.

Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Change-Id: Ibadf2a28c2f42fa92601f9da39a6ff73a43ebbe5
Reviewed-on: https://review.whamcloud.com/c/tools/e2fsprogs/+/52386
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>
Reviewed-by: Li Dongyang <dongyangli@ddn.com>
7 months agoLU-17117 misc: deduplicate log2/log10 functions 85/52385/3
Andreas Dilger [Fri, 15 Sep 2023 08:00:17 +0000 (02:00 -0600)]
LU-17117 misc: deduplicate log2/log10 functions

Remove duplicate log2() and log10() functions and replace them
with a single pair of functions ext2fs_log2() and ext2fs_log10().

Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Change-Id: Ifc86efe7e5f0243eb914c6d24319cc7dee3ebbe5
Reviewed-on: https://review.whamcloud.com/c/tools/e2fsprogs/+/52385
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Li Dongyang <dongyangli@ddn.com>
7 months agoLU-15002 mke2fs: try to pack the GDT blocks together 95/51295/10
Li Dongyang [Tue, 13 Jun 2023 10:12:22 +0000 (04:12 -0600)]
LU-15002 mke2fs: try to pack the GDT blocks together

Once a 4KiB block filesystem is 256TiB+ in size, the GDT grows
larger than the 128MiB size of the first block group, and would
overlap the backup superblock+GDT normally in the next group.
If this is the case, mke2fs will now automatically enable the
sparse_super2 and flex_bg features to allow the primary GDT to
spill into the second group, instead of enabling meta_bg, unless
meta_bg is explicitly requested.

Since sparse_super2 and flexbg already allow flexible placement
of the first and second backup superblock+GDT, no change is
needed to ext4 or e2fsck to use such a filesystem.

Using sparse_super2 and flexbg is preferable to meta_bg because it
packs the metadata close together at the start of the device. This
avoids millions of seeks at filesystem mount/open to read the
GDT blocks spread across a large filesystem by meta_bg.

Currently with sparse_super2 the backup superblock and GDT are put
in group #1 group and last group.  To allow the primary GDT to
spill into group #1, relocate the backups to the same group numbers
3^n, 5^n, 7^n that normal "sparse_super" backups are in.  The
first backup is close to the start of the device, while the second
backup is in a "sparse_super" group near the end of the device.

Remove the m_resize_inode_meta_bg test case, since mke2fs no
longer enables the meta_bg feature automatically.

Change-Id: I90a1d3b448fc17d4b11e8f52e41cf4ce87b89e08
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Reviewed-on: https://review.whamcloud.com/c/tools/e2fsprogs/+/51295
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
7 months agoLU-15002 mke2fs: do not set the BLOCK_UNINIT on groups has GDT 74/52274/3
Li Dongyang [Tue, 5 Sep 2023 13:36:24 +0000 (23:36 +1000)]
LU-15002 mke2fs: do not set the BLOCK_UNINIT on groups has GDT

This patch prepares the expansion of GDT blocks beyond a
single group, by make mke2fs to not set BLOCK_UNINIT on
groups with GDT blocks, block/inode bitmaps, or inode table
blocks allocated.

Otherwise, we still rely on kernel side to initialize the
block bitmap if the groups has BLOCK_UNINIT set, and the
kernel doesn't know a group could have GDT blocks allocated,
so it would make an bad block bitmap.

As a result, expect output of several tests needs to be changed,
especially if the test uses dumpe2fs to print the group summary.

Change-Id: I7847388d1f482d4e9cf6acb8fc6da3218d425e08
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Reviewed-on: https://review.whamcloud.com/c/tools/e2fsprogs/+/52274
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
7 months agoLU-15002 mke2fs: set free blocks accurately for groups has GDT 73/52273/3
Li Dongyang [Tue, 5 Sep 2023 13:12:35 +0000 (23:12 +1000)]
LU-15002 mke2fs: set free blocks accurately for groups has GDT

This patch is part of the preparation required to allow
GDT blocks expand beyond a single group,
it introduces 2 new interfaces:
- ext2fs_count_used_blocks(), to return the blocks used
in the bitmap range.
- ext2fs_reserve_super_and_bgd2() to return blocks used by
superblock/GDT blocks for every group, by looking up blocks used.

Change-Id: Ic26de963ed234af1099d00be6f7b6d0edc057c97
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Reviewed-on: https://review.whamcloud.com/c/tools/e2fsprogs/+/52273
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
7 months agoLU-17117 debugfs: allow debugfs to disable MMP directly 80/52380/4
Andreas Dilger [Fri, 15 Sep 2023 00:42:15 +0000 (18:42 -0600)]
LU-17117 debugfs: allow debugfs to disable MMP directly

Instead of needing to use "tune2fs -f -E clear_mmp" to clear the
MMP status, it should be possible to open and clear the MMP block
directly with debugfs in case of issues with tune2fs.

Add the "-m" option to debugfs and the "open" command to allow
opening the device and skipping the MMP status check.  Otherwise,
if sequence EXT4_MMP_SEQ_FSCK is set then debugfs will report an
error and the device cannot be not opened for write, even with -f:

    debugfs -f /dev/sdc
    MMP: e2fsck being run while trying to open /dev/sdc

Add the "clear_mmp" command to match the tune2fs extended command.
Add command aliases "mmp_clear", "mmp_dump", and "mmp_set_value"
so that they can be grouped together and found more easily.

Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Change-Id: I580ba4f30c366b0d0e75596c80cef2892c408fb0
Reviewed-on: https://review.whamcloud.com/c/tools/e2fsprogs/+/52380
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Li Dongyang <dongyangli@ddn.com>
7 months agoLU-15002 mke2fs: batch zeroing inode table 15/52215/7
Li Dongyang [Fri, 1 Sep 2023 03:59:57 +0000 (13:59 +1000)]
LU-15002 mke2fs: batch zeroing inode table

For flex_bg enabled fs, we could merge the
inode table blocks into a contiguous range,
this improves mke2fs time on large devices
when lazy_itable_init is disabled.

On a 977TB device, unpatched mke2fs was running
for 449m10s before getting terminated manually.
strace shows huge number of fallocate, given the
offset from fallocate it has done 41% of the inode
tables, the estimated time needed would be 1082m.

unpatched  patched
real 449m10.954s 4m20.531s
user 0m18.217s 0m16.147s
sys 0m20.311s 0m8.944s

Change-Id: I5257293ea35137cfc5321c15ed5dd8aa98c0612e
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Reviewed-on: https://review.whamcloud.com/c/tools/e2fsprogs/+/52215
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Emoly Liu <emoly@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
7 months agoLU-15002 e2fsck: check all sparse_super backups 19/52219/2
Andreas Dilger [Fri, 1 Sep 2023 07:14:49 +0000 (01:14 -0600)]
LU-15002 e2fsck: check all sparse_super backups

Teach e2fsck to look for backup super blocks in the "sparse_super"
groups, by checking group #1 first and then powers of 3^n, 5^n,
and 7^n, up to the limit of available block groups.

Export ext2fs_list_backups() function to efficiently iterate groups
for backup sb/GDT instead of checking every group.  Ensure that the
group counters do not try to overflow the 2^32-1 group limit, and
try to limit scanning to the size of the block device (if available).

Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Change-Id: I90a1d3b448fc17d4b11e8f52e41cf4ce873ebbe5
Reviewed-on: https://review.whamcloud.com/c/tools/e2fsprogs/+/52219
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
8 months agoLU-16750 tune2fs: add "-E iops" to set/clear IOPS groups 91/52091/4
Bobi Jam [Fri, 25 Aug 2023 04:00:52 +0000 (12:00 +0800)]
LU-16750 tune2fs: add "-E iops" to set/clear IOPS groups

Add an option to set/clear IOPS region of storage, for example:

-E iops=0-1024G,^4096-8192G

to set EXT4_BG_IOPS flag for block groups in 0 to 1024GiB and clear
the flag for those in 4096 to 8192GiB.

Change-Id: I52b773f9ad877a01b07f2e63d4d7ef4931499446
Signed-off-by: Bobi Jam <bobijam@whamcloud.com>
Reviewed-on: https://review.whamcloud.com/c/tools/e2fsprogs/+/52091
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
8 months agoLU-17028 build: add support for openEuler 20.03 LTS 56/51956/2
Xinliang Liu [Wed, 16 Aug 2023 07:53:02 +0000 (07:53 +0000)]
LU-17028 build: add support for openEuler 20.03 LTS

Add build support for openEuler 20.03 LTS.
And e2fsprogs rpm conflicts with openEuler's e2fsprogs-help man rpm,
should obsolete to replace it.

Signed-off-by: Xinliang Liu <xinliang.liu@linaro.org>
Change-Id: I12348e02cfec918167b1949205b86274a1651c27
Reviewed-on: https://review.whamcloud.com/c/tools/e2fsprogs/+/51956
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
8 months agoLU-16169 e2fsck: improve parallel thread balance 06/48806/8
Andreas Dilger [Sat, 8 Oct 2022 06:02:49 +0000 (00:02 -0600)]
LU-16169 e2fsck: improve parallel thread balance

Improve the balance of work allocated to each thread by distributing
a more equal number of inodes to each thread, rather than an equal
number of groups.

In some cases in real usage, the number of inodes allocated to
threads with equal numbers of groups can vary by 10x or more, which
leads to pass1 threads having a runtime time roughly propritional
to the number of inodes allocated to them.

Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Change-Id: Iea964cca33d19170e9b6d88aa725dc878cae6ce2
Reviewed-on: https://review.whamcloud.com/c/tools/e2fsprogs/+/48806
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Li Dongyang <dongyangli@ddn.com>
8 months agoLU-14712 e2fsprogs: support EXT2_FLAG_BG_TRIMMED and EXT2_FLAGS_TRACK_TRIM 24/51924/3
Li Dongyang [Fri, 11 Aug 2023 05:17:36 +0000 (15:17 +1000)]
LU-14712 e2fsprogs: support EXT2_FLAG_BG_TRIMMED and EXT2_FLAGS_TRACK_TRIM

This adds EXT2_FLAG_BG_TRIMMED, which is used on block group
descriptors during mke2fs after discard is done.
The EXT2_FLAG_BG_TRIMMED flag is cleared on the block group when
we free blocks.

Introduce EXT2_FLAGS_TRACK_TRIM, which is a new super block flag,
to indicate whether we should honour the EXT2_FLAG_BG_TRIMMED
set on each block group.
EXT2_FLAGS_TRACK_TRIM itself can be turned on/off via tune2fs.

Make dumpe2fs aware of the new flags.

Change-Id: I5af88fe4a32c8c9ab5604b84362a7140ae2cdde5
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Reviewed-on: https://review.whamcloud.com/c/tools/e2fsprogs/+/51924
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
8 months agoLU-16972 build: update version to 1.47.0-wc4 85/51885/2 1.47.0-wc4
Andreas Dilger [Mon, 7 Aug 2023 14:16:10 +0000 (08:16 -0600)]
LU-16972 build: update version to 1.47.0-wc4

LU-16750 mke2fs: add "-E iops" to set IOPS storage group
LU-16971 ext2fs: make sure we have at least EXT2_FIRST_INO + 1 inodes
ext2fs: don't retry discard/zeroout repeatedly
LU-16972 e2fsck: use rb-tree to track EA reference counts
LU-16972 e2fsck: fix merging ea_inode_refs

Change-Id: I23aa2ce53401a6cb71fdeffd1e7604028ceca4cf
Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-on: https://review.whamcloud.com/c/tools/e2fsprogs/+/51885

8 months agoLU-16750 mke2fs: add "-E iops" to set IOPS storage group 35/51735/5
Bobi Jam [Wed, 19 Jul 2023 10:07:05 +0000 (18:07 +0800)]
LU-16750 mke2fs: add "-E iops" to set IOPS storage group

With LVM it is possible to create an LV with SSD storage at the
beginning of the LV and HDD storage at the end of the LV, and use that
to separate ext4 metadata allocations (that need small random IOs)
from data allocations (that are better suited for large sequential
IOs) depending on the type of underlying storage.  Between 0.5-1.0% of
the filesystem capacity would need to be high-IOPS storage in order to
hold all of the internal metadata.

This would improve performance for inode and other metadata access,
such as ls, find, e2fsck, and in general improve file access latency,
modification, truncate, unlink, transaction commit, etc.

For mke2fs, using the sparse_super2 and packed_meta_blocks options
places all of the static metadata (group descriptors, block/inode
bitmaps, inode tables, journal) at the start of the device in the
(IOPS) flash region.

Add an option to mark which blocks are in the IOPS region of storage
at format time:

  -E iops=0-1024G,4096-8192G

so the ext4 mballoc code can then use the EXT4_BG_IOPS flag in the
group descriptors to decide which groups to allocate dynamic
filesystem metadata.

Change-Id: I13cc2820c71737848eab8a2d6e246748258a64df
Signed-off-by: Bobi Jam <bobijam@whamcloud.com>
Reviewed-on: https://review.whamcloud.com/c/tools/e2fsprogs/+/51735
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
8 months agoLU-16971 ext2fs: make sure we have at least EXT2_FIRST_INO + 1 inodes 22/51722/2
Li Dongyang [Thu, 20 Jul 2023 12:43:31 +0000 (22:43 +1000)]
LU-16971 ext2fs: make sure we have at least EXT2_FIRST_INO + 1 inodes

When creating a small fs with 100 1k blocks, mke2fs fails with:

Creating filesystem with 100 1k blocks and 8 inodes

Allocating group tables: done
Writing inode tables: done
ext2fs_mkdir: Could not allocate inode in ext2 filesystem while creating /lost+found

Increase s_inodes_per_group with a step of 8 to make
sure we have at least EXT2_FIRST_INO + 1 inodes.

Change-Id: I5ac30bdadf72e17780ebef93b2e81365a8a90641
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Reviewed-on: https://review.whamcloud.com/c/tools/e2fsprogs/+/51722
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
8 months agoext2fs: don't retry discard/zeroout repeatedly 94/51294/3
Andreas Dilger [Wed, 10 May 2023 04:26:39 +0000 (22:26 -0600)]
ext2fs: don't retry discard/zeroout repeatedly

Call safe_getenv(UNIX_IO_NOZEROOUT) once when the device is
opened and set CHANNEL_FLAG_NOZEROOUT if present instead of
getting uid/euid/getenv every time unix_zeroout() is called.

For unix_discard() and unix_zeroout() don't continue to call
them if the block device doesn't support these operations.

Change-Id: I3b3b79cf49a6f1cf9b254eb32103f174bacdfbd2
Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-on: https://review.whamcloud.com/c/tools/e2fsprogs/+/51294
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
8 months agoLU-16972 e2fsck: use rb-tree to track EA reference counts 29/51729/4
Li Dongyang [Wed, 26 Jul 2023 05:20:54 +0000 (15:20 +1000)]
LU-16972 e2fsck: use rb-tree to track EA reference counts

Using the sorted array to track the EA blocks and
its refs is not scalable.
When the file system has a huge number of EA blocks,
pass1 scanning could not be finished within a reasonable
time, as 95%+ of CPU time is spent in memmove() when
trying to enlarge the the sorted array.

On a file system with 20 million EA blocks on an NVMe device
pass1 time taken:
without patch:
time: 2014.78/1838.70/19.91
with patch:
time: 45.17/20.17/20.19

Change-Id: I6dc1ee3037dbf7a48deb610514af1f0e35a5a397
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Reviewed-on: https://review.whamcloud.com/c/tools/e2fsprogs/+/51729
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
8 months agoLU-16972 e2fsck: fix merging ea_inode_refs 63/51763/2
Li Dongyang [Wed, 26 Jul 2023 05:10:11 +0000 (15:10 +1000)]
LU-16972 e2fsck: fix merging ea_inode_refs

ea_inode_refs is used to track refs to ea inodes,
which has nothing to do with blocks.

Change-Id: Icd965a8e3a92edc51754c132b55044af9abbb9e4
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Reviewed-on: https://review.whamcloud.com/c/tools/e2fsprogs/+/51763
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
10 months agoLU-16879 build: update version to 1.47.0-wc3 v1.47.0-wc3
Li Dongyang [Thu, 22 Jun 2023 03:51:01 +0000 (13:51 +1000)]
LU-16879 build: update version to 1.47.0-wc3

2ad41ba20 LU-16879 mke2fs: disable metadata_csum_seed/orphan_file

Change-Id: I769bc9ce44e2c1d62fdeefb61183826b1347ff23
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
10 months agoLU-16879 mke2fs: disable metadata_csum_seed/orphan_file 40/51240/5
Andreas Dilger [Tue, 6 Jun 2023 17:52:40 +0000 (18:52 +0100)]
LU-16879 mke2fs: disable metadata_csum_seed/orphan_file

Disable the new metadata_csum_seed and orphan_file features in
/etc/mke2fs.conf by default, otherwise some older kernels cannot
mount ext4 filesystems formatted with these options.

Also disable metadata_csum by default, since Lustre MDTs do not
work with this option (it conflicts with dir_data), and it's
not considered totally stable before kernel 3.18.

Fixes: fe341d59e ("mke2fs.conf: enable the metadata_csum_seed and orphan_file features by default")
Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Change-Id: Ie1ad5bd485bdcdc65602a537790cd7c0ecf9a509
Reviewed-on: https://review.whamcloud.com/c/tools/e2fsprogs/+/51240
Tested-by: jenkins <devops@whamcloud.com>
Reviewed-by: Li Dongyang <dongyangli@ddn.com>
Reviewed-by: Artur Novik <anovik@ddn.com>
11 months agoLU-16835 build: update version to 1.47.0-wc2 v1.47.0-wc2
Li Dongyang [Thu, 25 May 2023 01:52:12 +0000 (11:52 +1000)]
LU-16835 build: update version to 1.47.0-wc2

c6af13873 LU-16835 tune2fs: fall back to old get/set fs label on error

Change-Id: I6c06e0dae4ce31e1b88920f01dca35dac0dc19a4
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
11 months agoLU-16835 tune2fs: fall back to old get/set fs label on error 73/51073/2
Li Dongyang [Sat, 20 May 2023 10:26:32 +0000 (20:26 +1000)]
LU-16835 tune2fs: fall back to old get/set fs label on error

If we fail to get/open the mount point for get/set
fs label ioctl, just fall back to old method and
silence the error messages.

Change-Id: I2ce5c02974540ff0fe883b992d0b510e68f6db26
Fixes: f85b4526f ("tune2fs: implement support for set/get label iocts")
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Reviewed-on: https://review.whamcloud.com/c/tools/e2fsprogs/+/51073
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
11 months agoLU-16531 build: update version to 1.47.0-wc1 08/49908/3 v1.47.0-wc1
Li Dongyang [Mon, 6 Feb 2023 10:18:37 +0000 (21:18 +1100)]
LU-16531 build: update version to 1.47.0-wc1

Rebase on new e2fsprogs-1.47.0 from upstream.

Change-Id: Ibc66dccb84839cf00a7792ffc4108e226e191b37
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Reviewed-on: https://review.whamcloud.com/c/tools/e2fsprogs/+/49908
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
12 months agotests: fix r_move_itable_realloc to run on Linux
Andreas Dilger [Mon, 9 Jan 2023 21:27:42 +0000 (14:27 -0700)]
tests: fix r_move_itable_realloc to run on Linux

The check for the various unsupported OSes incorrectly checked if
the string "FreeBSD" was true, which it always was.  Fix this.

Update the expect file as commit v1.46.4-17-g4ea80d031c7e did to
adjust the total number of blocks requested during resize.

Fixes: 5a3ea3905f ("tests: force test file systems to be built for Linux OS")
Change-Id: I4fbc74861bc68ad36467b2135e2abdb72898529b
Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
12 months agotests: fix u_direct_io to work with older losetup
Andreas Dilger [Sat, 7 Jan 2023 04:45:27 +0000 (21:45 -0700)]
tests: fix u_direct_io to work with older losetup

Older losetup does not have --sector-size, but this isn't really
needed for the test to work.  Instead specify the filesystem block
size directly to mke2fs, so that it works on all distros instead
of being skipped.

Change-Id: I5a0c82a9efdefd1b48f4d4288998c7725c9ae71e
Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
12 months agoLU-16067 misc: cleanup compiler warnings
Andreas Dilger [Tue, 2 Aug 2022 23:08:19 +0000 (17:08 -0600)]
LU-16067 misc: cleanup compiler warnings

Remove unused variables and various similar warnings.

Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Change-Id: I2a773d118960deb42f77c1ae8b038029e6f6cd18
Reviewed-by: Artem Blagodarenko <ablagodarenko@ddn.com>
Reviewed-by: Li Dongyang <dongyangli@ddn.com>
12 months agoLU-11446 e2fsck: check trusted.link when fixing nlink
Li Dongyang [Tue, 30 Mar 2021 11:22:40 +0000 (22:22 +1100)]
LU-11446 e2fsck: check trusted.link when fixing nlink

The inode link count could be higher than what is stored in the
local MDT inode because of remote file links from DNE MDTs.
If we find a mismatched link count, look up the "trusted.link"
xattr.  If it exists, do a sanity check on it, and use the
leh_reccount stored there if larger than the local link count.

If leh_overflow_time is set, then the "trusted.link" xattr may
not hold all of the links, so assume the maximum of available
link counts is valid until LFSCK clears leh_overflow_time.

If the inode is not linked into the namespace, link it into
lost+found before checking trusted.link for the DNE link count.

Update the f_trusted_link test case to include a locally
unreferenced inode that has a trusted.link xattr.  The
inode should be linked into lost+found because of the xattr,
even if it has no blocks/data, and the link cound should be
extracted from the trusted.link xattr.

Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Change-Id: I213d816a92043c348eb55374aaa98e98957ccf23
Reviewed-on: https://review.whamcloud.com/43169
Reviewed-by: Artem Blagodarenko <artem.blagodarenko@hpe.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
12 months agoLU-13797 e2fsck: quiet 'extent tree could be narrower'
Andreas Dilger [Sat, 12 Dec 2020 10:17:18 +0000 (03:17 -0700)]
LU-13797 e2fsck: quiet 'extent tree could be narrower'

The "extent tree (at level 1) could be narrower" message is
overly verbose and raises concerns by end users, even though
it is harmless.  On the flip side, this may save only a few
hundred blocks in the filesystem for a short period of time,
so there is relatively little benefit to be had.

Disable the extent optimization step in e2fsck by default by
adding the "no_optimize_extents" option to e2fsck.conf.

Change-Id: I0cee9407ef0653c3b946908513f3429816b89548
Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-on: https://review.whamcloud.com/40952
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Wang Shilong <wshilong@whamcloud.com>
12 months agoLU-12158 mke2fs: avoid too large stride and stripe_width
Wang Shilong [Tue, 21 May 2019 04:06:39 +0000 (12:06 +0800)]
LU-12158 mke2fs: avoid too large stride and stripe_width

According to benchmarks numbers, too large stripe and stripe_width
will make preassure with ext4 mballoc allocater and hurts performances
finally.
                    2MB Chunk size                 256K Chunk size
stripe_width,stride     Write(MB/s)    Read(MB/s)   Write(MB/s) Read(MB/s)
512,512                 10,810         10,124       10,492      6,923
1024,1024               10,793         10,064       10,431      6,921
2048,2048               8,047          10,080       6,629       7,381
4096,4096               7,350          10,089       6,505       7,282

Performance number comes from Shuichi Ihara.

This patch try to avoid use too large stride and stripe_width when mkfs.
If users really want large value they could do it by specify mkfs options
or run tune2fs later.

Change-Id: I768f1ecb39837338e08842b21b4fca8b98165d2a
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-on: https://review.whamcloud.com/34767
Tested-by: Jenkins
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: Andreas Dilger <adilger@whamcloud.com>
12 months agoLU-11609 mke2fs: allow 64bit feature without extents
Andreas Dilger [Sat, 3 Nov 2018 09:10:22 +0000 (03:10 -0600)]
LU-11609 mke2fs: allow 64bit feature without extents

The 64bit feature should be allowed without extents to for 32-bit
metadata_csum checksums to be stored in the group descriptor.
Change the extents check to check for more than 2^32 blocks instead
of the 64bit feature flag.  This also avoids warnings later if the
metadata_csum feature is enabled on a filesystem without 64bit.

Consolidate repeated 64bit feature checks into one check.

The fs_blocks_count value is always set when parse_fs_type() is
called, so no need to check if it is zero.

Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Change-Id: Ie73ff0ed50cfed5d6a9596260c6b6dc32d3ebbe5
Reviewed-on: https://review.whamcloud.com/33897
Tested-by: Jenkins
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Li Dongyang <dongyangli@ddn.com>
12 months agoLU-11215 libe2p: add large_xattr as an alias of ea_inode
Li Dongyang [Mon, 6 Aug 2018 01:33:44 +0000 (11:33 +1000)]
LU-11215 libe2p: add large_xattr as an alias of ea_inode

The upstream version of large xattr suppport doesn't
add "large_xattr" as a feature name, but it's still
used by our test scripts.

e2fsprogs commit:6a081f6d2a5cff0f5a077065aab39901d54bfb61

Change-Id: I48d7ac54b2b2ad252c404050bc4c3d9527865107
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Reviewed-on: https://review.whamcloud.com/32932
Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
12 months agoLU-9501 mke2fs: avoid inode number error with large FS
Alexey Lyashkov [Mon, 15 May 2017 14:45:38 +0000 (17:45 +0300)]
LU-9501 mke2fs: avoid inode number error with large FS

Limit inodes count instead of error return, to use all inodes in
the filesystem.

Change-Id: I76ee0a6c11b099b8c17d90b0778e3c0f0cf1f60a
Signed-off-by: Alexey Lyashkov <alexey.lyashkov@seagate.com>
Reviewed-on: https://review.whamcloud.com/27111
Tested-by: Jenkins
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
12 months agoLU-6722 jbd: double minimum journal size for RHEL7
Fan Yong [Fri, 26 Jun 2015 15:29:59 +0000 (23:29 +0800)]
LU-6722 jbd: double minimum journal size for RHEL7

In RHEL7 (kernel version >= 3.10.0), the maximum journal transaction
size is reduced to half of the RHEL6 case. That may cause Lustre to
complain that the declared transaction credits too large for very
small journal device. This patch increases the minimum journal size
from 4MB to 8MB for RHEL7 case, then counteract the above limitation
about the journal transaction size.

Signed-off-by: Fan Yong <fan.yong@intel.com>
Change-Id: Iec8a2c561416cb7b5acce342c8ebcb845c8d7a19
Reviewed-on: http://review.whamcloud.com/15401
Reviewed-by: Niu Yawei <yawei.niu@intel.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
12 months agoLU-1540 e2fsck: add missing symlink NUL terminator
Andreas Dilger [Sat, 14 Jul 2012 02:33:01 +0000 (20:33 -0600)]
LU-1540 e2fsck: add missing symlink NUL terminator

If a long symbolic link target is written into an external block
without a NUL terminator, its length is decided by the inode's size.
Make symlink check add a NUL termination in such cases if needed.

Such faulty symlinks were generated by osd-ldiskfs on the MDS until
Lustre 2.1.3 and Lustre 2.3.  The in-kernel code would handle such
unterminated symlinks correctly, since it used the inode size to
determine the symlink length, but e2fsck would assume the symlink
is broken if there wasn't a trailing NUL.

Signed-off-by: Andreas Dilger <andreas.dilger@intel.com>
Change-Id: I4419b30f1adb4a7d273796a936427aa351510213

12 months agodebugfs: decode and dump xattrs on inode stat
Andreas Dilger [Fri, 13 Apr 2012 18:55:45 +0000 (12:55 -0600)]
debugfs: decode and dump xattrs on inode stat

Decode and print Access Control Lists in debugfs stat/ea_list.

Print the Lustre "fid", "lma", and "link" xattr contents,
if present, with "debugfs stat" to simplify debugging.
Add "layout_version" and "range" values when printing a
trusted.fid xattr.

Increase the size of xattrs hex-dumped from 40 to 120.  But as
Lustre encrypted inodes may contain illegal characters in their
names, it is also needed to escape non-printable characters in
linkEA.

Change-Id: I3ee35ddb02582ab66d4e25ff76fae35935b54508
Reviewed-on: https://review.whamcloud.com/43384
Reviewed-by: Li Dongyang <dongyangli@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
12 months agotests: add basic tests for dirdata feature
Andreas Dilger [Fri, 18 May 2018 04:52:58 +0000 (14:52 +1000)]
tests: add basic tests for dirdata feature

Signed-off-by: Pravin Shelar <pravin@clusterfs.org>
Signed-off-by: Andreas Dilger <andreas.dilger@intel.com>
Signed-off-by: Artem Blagodarenko <artem.blagodarenko@gmail.com>
12 months agoe2fsck: add support for dirdata feature
Andreas Dilger [Thu, 17 May 2018 03:43:16 +0000 (13:43 +1000)]
e2fsck: add support for dirdata feature

Add support for the INCOMPAT_DIRDATA feature, which allows
storing extra data in the directory entry beyond the name.
This allows the Lustre File IDentifier to be accessed in
an efficient manner, and would be useful for expanding a
filesystem to allow more than 2^32 inodes in the future.

While adding the new file entry in directory block, link_proc()
calculates minimum record length of the existing directory entry
without considering the dirent data size and which leads to
corruption. Changed the code to use EXT2_DIR_REC_LEN() which will
return correct record length including dirent data size.

Lustre-bug-id: https://jira.hpdd.intel.com/browse/LU-4677
Signed-off-by: Pravin Shelar <pravin@clusterfs.com>
Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Artem Blagodarenko <artem.blagodarenko@gmail.com>
12 months agoext2fs: check if Lustre filesystem is mounted
Andreas Dilger [Fri, 13 Apr 2012 08:16:24 +0000 (02:16 -0600)]
ext2fs: check if Lustre filesystem is mounted

Add a check to ext2fs_check_mount_point() to loo in /proc/fs/lustre/*
to see if Lustre is mounted, since st_rdev of the mountpoint does not
match st_rdev of the block device itself, which confuses libext2fs.

  LU-6093 ext2fs: don't confuse mke2fs by Lustre zfs mount

  check_if_lustre_mounted() by efb735 will mistakenly report the
  device is mounted when there is a ZFS-based Lustre mount.

  This patch fix it by parsing the real device name from the pool
  name exported in Lustre proc file with "zpool" command.

  It also includes a typo fix in check_if_lustre_mounted() and a
  few other code cleanups.

Signed-off-by: Niu Yawei <yawei.niu@intel.com>
  Change-Id: I43dabbcd6fa721c569cd9cc0f4a6de93ab761392
  Reviewed-on: http://review.whamcloud.com/14223
Reviewed-by: John L. Hammond <john.hammond@intel.com>
  LU-12641 libext2fs: memory leak of check_if_lustre_mounted

  check_if_lustre_mounted() doesn't free the allocated memory
  if failure.

  Change-Id: I36b5c1e981ca4f1db8c9515be29dd98c074d14dc
Signed-off-by: Li Xi <lixi@ddn.com>
  Reviewed-on: https://review.whamcloud.com/35729
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: jenkins <devops@whamcloud.com>
Reviewed-by: Wang Shilong <wshilong@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Signed-off-by: Andreas Dilger <andreas.dilger@intel.com>
12 months agotests: verify large xattr inode support
Andreas Dilger [Fri, 13 Apr 2012 08:14:16 +0000 (02:14 -0600)]
tests: verify large xattr inode support

Verify that inodes with large EAs in a secondary inode are working:
* EA inode needs to have EA_INODE_FL set
* EA inode should reference parent inode number+generation

Signed-off-by: Kalpak Shah <kalpak@sun.com>
Signed-off-by: Andreas Dilger <andreas.dilger@intel.com>
12 months agolibext2fs: add a regression test for in-inode xattrs
Andreas Dilger [Fri, 13 Apr 2012 08:01:12 +0000 (02:01 -0600)]
libext2fs: add a regression test for in-inode xattrs

Add tst_read_ea test case to verify libext2fs xattr interfaces
work consistently with the kernel xattr handling.

The libattr-devel package 2.4.48 still has attr/xattr.h (el7, el8),
so keep checking for it.

Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
12 months agoe2fsck: clean up xattr checking code
Andreas Dilger [Fri, 13 Apr 2012 08:01:12 +0000 (02:01 -0600)]
e2fsck: clean up xattr checking code

Clean up xattr header/list processing for in-inode xattrs instead
of doing lots of explicit pointer math.

Signed-off-by: Andreas Dilger <andreas.dilger@intel.com>
12 months agoe2fsck: track errors/badness found for each inode
Andreas Dilger [Fri, 13 Apr 2012 07:13:58 +0000 (01:13 -0600)]
e2fsck: track errors/badness found for each inode

The present e2fsck code checks the inode, per field basis.  It
doesn't take into consideration to total sanity of the inode.
This may cause e2fsck turning a garbage inode into an apparently
sane inode ("It is a vessel of fertilizer, and none may abide
its strength").

Add heuristics into fix_problem() so that e2fsck tracks the degree
of badness of an inode, without needing to be added in many places
in code and explicitly set. An icount is used to keep track the
badness of every inode.  Take advantage of pctx->ino almost always
pointing at the bad inode and increment badness when fix_problem()
is called on an inode.  This also handles future problem addition.

That leaves only a handful of places in the code that need special
handling to either set a higher badness (e.g. inode blocks that are
directly referencing filesystem metadata or have wildly incorrect
timestamps), or should not contribute to inode badness at all.

Badness above a certain threshold value results in clearing the
inode.  The default badness threshold value is 12, it can be tuned
for e2fsck using "-E inode_badness_threshold=<value>" if needed.

When multiply-claimed blocks are found, this is often caused by a
corrupted inode or indirect block that causes a bad inode to overlap
with many good inodes.  This problem is made worse when running
on a large filesystem (16TB or more) because random 32-bit numbers
in the inode->i_blocks[] array are always "valid" block numbers
(with smaller filesystems the random block numbers would be detected
as an error).  Garbage triple/double/indirect blocks will also point
to random "valid" blocks that will themselves contain random 32-bit
block numbers and multiply duplicate blocks count exponentially.

Rather than clone all of those blocks, or possibly deleting/zeroing
all such inodes (as done with "-E shared=delete") it would be better
to find the "bad" inode(s) causing the most problems, and clear only
those inodes, rather than clearing all of inodes with shared blocks.
However, care should be taken to avoid spuriously clearing inodes
that only share blocks with a small number of peers, as it is
difficult to know for sure in this case which inode is the bad one.

An added difficulty in implementing this is that the full list of
inodes sharing a given block is only available in pass1d, at which
point it is already starting to clone the shared blocks.  Some checks
could be done in pass1b, by penalizing inodes wiht the most shared
*blocks*, but it is better to count the shared *inodes*.

If a bad inode is found in pass1b it can be cleared before adding it
to the shared cluster/inode dictionaries.  Otherwise, if found in
pass1d the code to remove inodes/cluster dict_delete_free() code is
non-functional.  For now, restart e2fsck pass1 if very bad inodes
are found.  This will not affect already-processed inodes, and should
reduce the number of duplicate blocks significantly.

Add a new f_ibadness_dup test case for a many-conflicting inode.
Update f_ind_inode_collision, which is a many-blocks conflicting
case that is much better handled by this new mechanism.

Signed-off-by: Girish Shilamkar <girish@clusterfs.com>
Change-Id: I9f5f48d979afebb9c953d9fb2777ebf570f30c15
Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
12 months agotests: add tests for expanding inode extra size
Andreas Dilger [Fri, 13 Apr 2012 00:05:03 +0000 (18:05 -0600)]
tests: add tests for expanding inode extra size

Signed-off-by: Kalpak Shah <kalpak@clusterfs.com>
Signed-off-by: Andreas Dilger <andreas.dilger@intel.com>
12 months agoLU-14768 e2fsck: merge expand_eisize_map properly
Wang Shilong [Tue, 25 Aug 2020 08:31:27 +0000 (16:31 +0800)]
LU-14768 e2fsck: merge expand_eisize_map properly

Min known min_extra_isize should be updated after
threads finish.

Original commit:
cb4f0f44 ("LU-8465 e2fsck: merge expand_eisize_map properly")

Add the missing commit back, in pass5 we won't pick up some of
the inodes marked in pass1 for eisize expanding.

This should be merged to
e47509eb ("e2fsck: add support for expanding the inode size")
in the next rebase.

Change-Id: Ie9703383821ecb11337630d74ffb0c7dcb2fcc1e
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Reviewed-on: https://review.whamcloud.com/44011
Reviewed-by: Wang Shilong <wshilong@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: jenkins <devops@whamcloud.com>
12 months agoe2fsck: add support for expanding the inode size
Andreas Dilger [Fri, 13 Apr 2012 00:03:37 +0000 (18:03 -0600)]
e2fsck: add support for expanding the inode size

This patch adds a "-E expand_extra_isize" feature which makes sure
that _every_ used inode has i_extra_isize >= s_min_extra_isize if
s_min_extra_isize is set. Else it makes sure that i_extra_isize
of every inode is equal to sizeof(ext2_inode_large) - 128.

This is useful for the case where nanosecond timestamps or 64-bit
inode version fields are required for all inodes in the filesystem.

There is also a fix for test f_itable_collision, the original
E2FSCK_TIME would overflow on a 32bit system when adding with
ctx->time_fudge in EXT4_XTIME_FUTURE, making us increase the
inode badness incorrectly.

  LU-10205 libext2fs: fix buffer overrun in ext2fs_expand_extra_isize

  In ext2fs_expand_extra_isize, we size buffer using 'size' but then
  do the memcpy with the rounded-up size, which can overflow the buffer.

  With MALLOC_CHECK_=2, I see:
  Error in `../e2fsck/e2fsck': free(): invalid pointer: <addr>

  Change-Id: I31be58de12d4d50646c7aa96959de0efc5c279c3
Signed-off-by: Jeff Mahoney <jeffm@suse.com>
  Reviewed-on: https://review.whamcloud.com/29975
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Change-Id: I306ff4f81d8bd6bdf0446c76d6772951043fead4
Signed-off-by: Kalpak Shah <kalpak@clusterfs.com>
Signed-off-by: Andreas Dilger <andreas.dilger@intel.com>
12 months agoe2fsck: allow deleting or zeroing shared blocks
Andreas Dilger [Thu, 12 Apr 2012 23:32:53 +0000 (17:32 -0600)]
e2fsck: allow deleting or zeroing shared blocks

E2fsck fixes files that are found to be sharing blocks by cloning
the shared blocks and giving each file a private copy in pass 1D.

Allowing all files claiming the shared blocks to have copies can
inadvertantly bypass access restrictions.  Deleting all the files,
zeroing the cloned blocks, or placing the files in the /lost+found
directory after cloning may be preferable in some secure environments.

The following patches implement config file and command line options
in e2fsck that allow pass 1D behavior to be tuned according to site
policy.  It adds two extended options and config file counterparts.
On the command line:

 -E clone=dup|zero

    Select the block cloning method.  "dup" is old behavior,
    and is the default.  "zero" is a new method that substitutes
    zero-filled blocks for the shared blocks in all the files
    that claim them.

 -E shared=preserve|lost+found|delete

    Select the disposition of files containing shared blocks.
    "preserve" is the old behavior which remains the default.
    "lost+found" causes files to be unlinked after cloning so
    they will be reconnected to /lost+found in pass 3.
    "delete" skips cloning entirely and simply deletes the files.

In the config file:
  [options]
      clone=dup|zero
      shared=preserve|lost+found|delete

Change-Id: I132a3122076b5c68c4f9ea8c1316cb48f8d94020
Signed-off-by: Jim Garlick <garlick@llnl.gov>
Signed-off-by: Andreas Dilger <andreas.dilger@intel.com>
12 months agoe2fsck: parse config file before command-line opts
Andreas Dilger [Thu, 12 Apr 2012 23:24:55 +0000 (17:24 -0600)]
e2fsck: parse config file before command-line opts

The patch changes the order that the config file and command line
are parsed so that command line has precedence.  It also parses
the -E option for every occurrence, otherwise the -E option is
not cumulative.

Signed-off-by: Jim Garlick <garlick@llnl.gov>
Signed-off-by: Andreas Dilger <andreas.dilger@intel.com>
12 months agofilefrag: Lustre changes to filefrag FIEMAP handling
Andreas Dilger [Thu, 12 Apr 2012 21:31:35 +0000 (15:31 -0600)]
filefrag: Lustre changes to filefrag FIEMAP handling

Add support for multiple-device filesystems by defining a new
fe_device field in the fiemap_extent structure.  This allows
printing the filesystem-relative or linux block device number
associated with each extent of a file.  If a single filesystem
extent is mirrored to multiple block devices, the fe_device
field can be used to disambiguate the multiple copies.

If the "-l" (device-logical) option is given to filefrag, then
all extents for a particular device of a file are returned
before returning extents for the next device.  This makes it
easier to see if extent allocation within a single device is
contiguous, instead of returning all of the blocks of a file
interleaved in file-logical-offset order.

Lustre uses the high 16bits of fiemap_extent:fe_device to record
the absolute stripe number being processed, so that next fiemap
call continues from that stripe, which allows files that have
multiple objects on the same device to work properly (FLR, PFL).
Older filefrag will print the stripe number with the OST index.

Change-Id: Ifb40cc159ddc61d2296c494f0e899ac11bf88b60
Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
12 months agoLU-13241 libext2fs: add quota to libext2fs
Li Dongyang [Mon, 23 Mar 2020 01:56:00 +0000 (12:56 +1100)]
LU-13241 libext2fs: add quota to libext2fs

add quota related interface to libext2fs and install the
relevant headers.

Change-Id: I17e6b5aa74e0f1bb1465168a1cf4e03184e003b0
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Reviewed-on: https://review.whamcloud.com/38027
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
12 months agoATM-1754 tests: skip f_large_dir on RHEL6
Andreas Dilger [Fri, 25 Sep 2020 02:52:03 +0000 (20:52 -0600)]
ATM-1754 tests: skip f_large_dir on RHEL6

Skip f_large_dir test on RHEL6 since it runs very slowly, and
has enough coverage on other branches.

Change-Id: I7e001f734d84fc7927bed1da861a33283577d490
Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-on: https://review.whamcloud.com/40042
Reviewed-by: Wang Shilong <wshilong@whamcloud.com>
12 months agoATM-1754 tests: skip f_large_dir(_csum) on aarch64
Andreas Dilger [Thu, 21 May 2020 20:57:49 +0000 (14:57 -0600)]
ATM-1754 tests: skip f_large_dir(_csum) on aarch64

Skip the f_large_dir(_csum) tests on the aarch64 builders because this is
extremely slow (90 minutes) and clogs up the build system when
there are many patches in the queue.

Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Change-Id: I8c3a588c34caae5332b4c643df940b0e855cab07
Reviewed-on: https://review.whamcloud.com/38693
Tested-by: jenkins <devops@whamcloud.com>
Reviewed-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Li Dongyang <dongyangli@ddn.com>
12 months agoLU-5307 build: add support for RHEL-7+ .spec file
Andreas Dilger [Fri, 18 Jul 2014 20:47:02 +0000 (14:47 -0600)]
LU-5307 build: add support for RHEL-7+ .spec file

Add the RHEL 7+ .spec file, which is also used by later RHEL,
derivatives (Alma, CentOs, Fedora, Rocky, Scientific, Euler).

Signed-off-by: Andreas Dilger <andreas.dilger@intel.com>
Change-Id: I00278b50c3669f56c5297a75b89775fec9476f43

12 months agoTT-177 build: add .spec file for SLES11+ packaging
Andreas Dilger [Fri, 13 Apr 2012 08:23:12 +0000 (02:23 -0600)]
TT-177 build: add .spec file for SLES11+ packaging

Include the upstream SLES11 .spec file to ensure the packages we
build match the upstream packages.  Any later patches that change
the packaging should patch the .spec file appropriately.

Add in the SLES-specific patches, excluding the replacement de.po
file, since the original SLES11 de.po file is only against 1.41.4,
and is missing a large number of changes to the translated messages
related to 64-bit format specifiers.

Always skip m_hugefile and t_replay_and_set tests, building rpms in
an ext3 environment will make the tests fail.

Need to add a line in the SUSE spec file for Provides: ldiskfsprogs.
This is present in the RHEL spec file and is needed to resolve
dependencies in lustre server rpms at rpm install time.

Ensure that the libcom_err and libext2fs2 packages installed match the
other packages built from e2fsprogs so that modifications to those
libraries (e.g. project quota) are available.

Changes in output from the lsb_release command in SLES 12
resulted in using the wrong .spec file in builds.
This patch corrects the problem for old and new versions.

Signed-off-by: Bob Glossman <bob.glossman@intel.com>
Signed-off-by: Andreas Dilger <andreas.dilger@intel.com>
Change-Id: I783d58bd78d7c4c66cc85ec5557ae1aaf64016ba

12 months agobuild: add RHEL6 .spec file for packaging
Andreas Dilger [Fri, 13 Apr 2012 08:19:19 +0000 (02:19 -0600)]
build: add RHEL6 .spec file for packaging

Include the upstream RHEL6 .spec file to ensure the packages we
build match the upstream packages.  Any later patches that change
the packaging should patch the .spec file appropriately.

Change-Id: Ia12d193ff1a9c5a40cee2585a9bc47d5f9f63f4d
Signed-off-by: Andreas Dilger <andreas.dilger@intel.com>
12 months agoLU-16391 build: allow '-' in the version
Andreas Dilger [Thu, 15 Dec 2022 01:34:07 +0000 (18:34 -0700)]
LU-16391 build: allow '-' in the version

Update configure.ac to separate Version from Release in the .spec.in
files so that RPM building doesn't choke if a '-' is in the version.
Add these changes into configure also, so they are immediately used.

Change-Id: I6cbf6efa25175f9cebeb349d28e1e8a565b4cb74
Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-on: https://review.whamcloud.com/c/tools/e2fsprogs/+/49421
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
12 months agobuild: update e2fsprogs.spec for distro builds
Andreas Dilger [Thu, 12 Apr 2012 21:39:04 +0000 (15:39 -0600)]
build: update e2fsprogs.spec for distro builds

Add the distro version to the RPM release number, so that it the
RPM names do not conflict.

Allow the RPM built from upstream to replace the split packages
provided by the distros.  At some point in the future it may be
desirable to also split the RPM built by this spec file, but this
bs complicated by the fact that SLES and RHEL have different splits.

Change-Id: Iab8f75a7f7bd9790899a2024d91079d55501ef65
Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
12 months agoLU-14432 configure.ac: fix AC_PREREQ
Li Dongyang [Sun, 14 Feb 2021 23:30:11 +0000 (10:30 +1100)]
LU-14432 configure.ac: fix AC_PREREQ

AS_HELP_STRING was introduced in autoconf 2.58
requiring autoconf 2.69 is not necessary and it breaks
several builds in our Jenkins.

Change required autoconf to 2.59 just to be safe.

Change-Id: I30af873879866858d32b715ac4ce78d095e10f7d
Fixes: 8c9e2921 ("configure.ac: convert all help strings to use AS_HELP_STRING")
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Reviewed-on: https://review.whamcloud.com/41665
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Wang Shilong <wshilong@whamcloud.com>
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
12 months agoe2fsck: Annotating fields in e2fsck_struct
Saranya Muruganandam [Tue, 17 Nov 2020 17:54:30 +0000 (09:54 -0800)]
e2fsck: Annotating fields in e2fsck_struct

Adding information on fields in e2fsck_struct
on how they are used when running parallel fsck.

E2fsprogs-commit: c2f04f7da367ffc4cd64be5b2285878ad23b3e7f

Change-Id: I5b97921f0191b1d2bf4bf6bbedefc7ae401e9e33
Signed-off-by: Saranya Muruganandam <saranyamohan@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoLU-13717 sec: support encrypted files handling in pfsck mode
Sebastien Buisson [Tue, 19 Oct 2021 11:02:12 +0000 (13:02 +0200)]
LU-13717 sec: support encrypted files handling in pfsck mode

e2fsck needs to be improved in order to support encrypted files
handling in parallel fsck mode. The e2fsck_merge_encrypted_info()
function is added to merge encrypted inodes info collected from
different threads in pass1, so that it can be used in pass2.

Signed-off-by: Sebastien Buisson <sbuisson@ddn.com>
Change-Id: Iaf495f399e38e3ca7914b92059607d4babc1ad44
Reviewed-on: https://review.whamcloud.com/45289
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Li Dongyang <dongyangli@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
12 months agoLU-14953 e2fsck: pfsck progress report
Li Dongyang [Mon, 30 Aug 2021 12:43:52 +0000 (22:43 +1000)]
LU-14953 e2fsck: pfsck progress report

Re-enable progress support for pfsck.
To make it work, in pass1 before calling e2fsck_update_progress,
figure out the the progress for other threads as well.

Change-Id: Ia039aa20f35990b550680169c8b3670b4346ce8b
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Reviewed-on: https://review.whamcloud.com/44784
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
12 months agoe2fsck: propagate number of threads
Saranya Muruganandam [Fri, 13 Nov 2020 23:33:05 +0000 (15:33 -0800)]
e2fsck: propagate number of threads

Sometimes, such as in orphan_inode case, e2fsck_pass1
is called after reading the block bitmaps. This results in
reading the block bitmap sequentially and multithreading
only gets kicked in later. Fix the thread count earlier
while setting up the file system.

E2fsprogs-commit: 3f56676f4daa7379dbd868a4114910f20d49527d

Change-Id: I51ab36ad8094c605a32e4b7929fadb326d77169e
Signed-off-by: Saranya Muruganandam <saranyamohan@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: misc cleanups for pfsck
Andreas Dilger [Mon, 28 Sep 2020 19:21:16 +0000 (13:21 -0600)]
e2fsck: misc cleanups for pfsck

Add -m option description to e2fsck.8 man page.

Rename e2fsck_struct fs_num_threads to pfs_num_threads to avoid
confusion with the ext2_filsys fs_num_threads field, and move
thread_info to be together with the other HAVE_PTHREAD fields.

Move ext2_filsys fs_num_threads to fit into the __u16 "pad" field
to avoid consuming one of the few remaining __u32 reserved fields.

Fix a few print format warnings.

E2fsprogs-commit: afad25acf8c6694c8cea08820fe5af84388a76f3

Change-Id: I5ab49a3ad41f1ebb374669c5d640afa1f3fc35fd
Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Wang Shilong <wshilong@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: fix memory leaks with pfsck enabled
Wang Shilong [Mon, 28 Sep 2020 02:27:12 +0000 (10:27 +0800)]
e2fsck: fix memory leaks with pfsck enabled

valgrind detected two memory leaks:

1) quota context is not released after merging.
2) @refcount_orig should be released

E2fsprogs-commit: d9aa090f713ecf3393c4e68d951a03ccba4a7073

Change-Id: I8b8379505524d25b5c18a0d67bb4dd8ede3494ef
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agotests: add pfsck test
Wang Shilong [Wed, 23 Sep 2020 15:58:23 +0000 (23:58 +0800)]
tests: add pfsck test

pfsck run on a clean fs should not return any errors.

Generate an image with possible features enabled,
especially EA shared blocks etc.

E2fsprogs-commit: 59ae5e5a627aaa24d663e8ce41955e1ded46633a

Change-Id: Ifa8d725993cf8bec08e847e2734bb44b0ddbfd04
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: reset @inodes_to_rebuild if restart
Wang Shilong [Thu, 24 Sep 2020 02:00:06 +0000 (10:00 +0800)]
e2fsck: reset @inodes_to_rebuild if restart

Verify multiple thread on a corrupted images hit following bug:

pass1.c:2902: e2fsck_pass1_thread_prepare:
Assertion `global_ctx->inodes_to_rebuild == NULL' failed.
Signal (6) SIGABRT si_code=SI_TKILL
./e2fsck/e2fsck[0x43829e]
/lib64/libpthread.so.0(+0x14b20)[0x7f3b45135b20]
/lib64/libc.so.6(gsignal+0x145)[0x7f3b44f2c625]
/lib64/libc.so.6(abort+0x12b)[0x7f3b44f158d9]
/lib64/libc.so.6(+0x257a9)[0x7f3b44f157a9]
/lib64/libc.so.6(+0x34a66)[0x7f3b44f24a66]
./e2fsck/e2fsck(e2fsck_pass1+0x1662)[0x423572]
./e2fsck/e2fsck(e2fsck_run+0x5a)[0x41611a]
./e2fsck/e2fsck(main+0x1608)[0x4121b8]
/lib64/libc.so.6(__libc_start_main+0xf3)[0x7f3b44f171a3]
./e2fsck/e2fsck(_start+0x2e)[0x413dde]

@inodes_to_rebuild could be not NULL after we restart pass1

E2fsprogs-commit: 3b7879e310d30d5f4c518c607aec830b63b2315b

Change-Id: I8ca6a37d41788333fe9494bd84816932f39fcb3c
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: update mmp block in one thread
Wang Shilong [Wed, 23 Sep 2020 14:36:41 +0000 (22:36 +0800)]
e2fsck: update mmp block in one thread

For multiple threads, different threads will try to
update mmp block at the same time, only allow one
thread to update it.

E2fsprogs-commit: 1806140014cbce395d0ca4a51d46fa21e2833241

Change-Id: Id78b949d338c384a64828b955ed7477aa816f8ee
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: make default smallest RA size to 1M
Wang Shilong [Tue, 15 Sep 2020 07:06:11 +0000 (15:06 +0800)]
e2fsck: make default smallest RA size to 1M

If we have a smaller inodes per group, default ra size could
be very small(etc 128KiB), this hurts performances.

Tune above 128K to 1M, i see pass1 time drop down from
677.12 seconds to 246 secons with 32 threads.

E2fsprogs-commit: 9226b3264b74044013f4933280b1989c33be42f3

Change-Id: Ia503b4f73dc8f3d5e6be705c7311bdacc5b874a9
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: avoid too much memory allocation for pfsck
Wang Shilong [Wed, 27 May 2020 07:14:23 +0000 (07:14 +0000)]
e2fsck: avoid too much memory allocation for pfsck

e2fsck init memory according to filesystem inodes/dir numbers
recorded in the superblock, this should be aware of filesystem
number of threads, otherwise, oom happen.

E2fsprogs-commit: ed2cec54da604efe1af2734182985d1730674138

Change-Id: Ia66cd782a330f3f0215c7119c25ef701edc8b143
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: cleanup e2fsck_pass1_thread_join()
Wang Shilong [Wed, 23 Sep 2020 10:05:34 +0000 (18:05 +0800)]
e2fsck: cleanup e2fsck_pass1_thread_join()

Use e2fsck_reset_context() to free memory to simpify
codes.

E2fsprogs-commit: 41f40f4215fd9a357a3d250a34745eb953982bda

Change-Id: I1fdf12eb520f738d90453bbc886dbe89bc9a2b17
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: wait fix thread finish before checking
Wang Shilong [Thu, 14 May 2020 14:16:50 +0000 (22:16 +0800)]
e2fsck: wait fix thread finish before checking

Before proceeding next inodes, waitting existed
fixing finished.

E2fsprogs-commit: ae4a78493b16d01d7dbe628ec9064216f960bd16

Change-Id: I159082381e66bd51b3558a65ef45e936aa741370
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Saranya Muruganandam <saranyamohan@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: set E2F_FLAG_ALLOC_OK after threads
Wang Shilong [Thu, 16 Apr 2020 07:08:31 +0000 (15:08 +0800)]
e2fsck: set E2F_FLAG_ALLOC_OK after threads

Only flag ALLOC OK after all threads finished without problem.

E2fsprogs-commit: 2c1e9ced674f0d8e634c3b6299ce9bdfbe4b8739

Change-Id: I764d08bab5299ec0aec9d7d5e9b21320c085dee9
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoLU-14768 e2fsck: merge casefolded dir lists after thread finish
Wang Shilong [Thu, 12 Mar 2020 11:40:27 +0000 (19:40 +0800)]
LU-14768 e2fsck: merge casefolded dir lists after thread finish

This is missed and should be done after threads finish.

Original commit:
18538b27 ("LU-8465 e2fsck: merge encrypted dir lists after thread finish")
Upstream is now using e2fsck_struct->casefolded_dirs instead of
encrypted_dirs.

Only for the correctness of pfsck, we won't use casefolded feature on
ldiskfs.

Change-Id: I6f3a28501886c2ed4bec26dc5c5e5a1914fece04
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Reviewed-on: https://review.whamcloud.com/44010
Reviewed-by: Wang Shilong <wshilong@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: jenkins <devops@whamcloud.com>
12 months agoe2fsck: simplify e2fsck context merging codes
Wang Shilong [Fri, 13 Mar 2020 10:46:24 +0000 (18:46 +0800)]
e2fsck: simplify e2fsck context merging codes

We tried to copy thread context to global context directly
and then copy back some saved variables before merging.

Since we have finished almost all necessary variables
in the e2fsck context, we could simplify codes, and
this could help us understand what is missing rather
than hide problems.

E2fsprogs-commit: 41c991af7ed7244da0fd799f63e4e6fd0d6c5b25

Change-Id: I450e3f5fb26c5c6fe8a075373bf538965f443d02
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: merge extent depth count after threads finish
Wang Shilong [Fri, 20 Mar 2020 01:55:18 +0000 (09:55 +0800)]
e2fsck: merge extent depth count after threads finish

tests covered by f_extent_htree.

E2fsprogs-commit: 291bc9b7fc5280166e58b0e586d5ec93738f31a7

Change-Id: I5b4f629e1031f719a2d032a9dd5edb1cfac90975
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: reset lost_and_found after threads finish
Wang Shilong [Fri, 20 Mar 2020 01:50:56 +0000 (09:50 +0800)]
e2fsck: reset lost_and_found after threads finish

This should not be kept, the reaons is similar to what
e2fsck_pass1 has done before.

E2fsprogs-commit: 10abbf8fd534d61c40fbf9699bf605ac6456b2e2

Change-Id: I0e431be7619720e31927ab1df7034aee1e3eb747
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: merge options after threads finish
Wang Shilong [Thu, 19 Mar 2020 14:05:30 +0000 (22:05 +0800)]
e2fsck: merge options after threads finish

It will be possible that threads might append E2F_OPT_YES,
so we need merge options to global, test f_yesall cover this.

E2fsprogs-commit: a68a5e890793eda2e8d7b4f7d27d23a7eeb474c5

Change-Id: I0e8bbc078fc538bd7d948ced1161506710401ab3
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Saranya Muruganandam <saranyamohan@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: fix readahead for pfsck of pass1
Wang Shilong [Wed, 11 Mar 2020 12:48:00 +0000 (20:48 +0800)]
e2fsck: fix readahead for pfsck of pass1

Several improvments for this patch:

1) move readahead_kb detection to preparing phase.
2) inode readahead should be aware of thread block group
boundary.
3) make readahead_kb aware of multiple threads.

E2fsprogs-commit: 02618b8f69c1ed453b092476cbd0d735206c939d

Change-Id: I2f9d7b8ae5039c0551bb3994f1c7814af257cee5
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoLU-14432 ext2fs: fix ext2fs_get_avg_group() warning
Andreas Dilger [Mon, 31 May 2021 21:08:13 +0000 (15:08 -0600)]
LU-14432 ext2fs: fix ext2fs_get_avg_group() warning

Move ext2fs_get_avg_group() into e2fsck/pass1.c instead of the
ext2fs.h header, since it is currently only used in that one
function.  It is too large to be a static inline function, and
it is currently causing many compile warnings because it is
not marked with "_INLINE_".

Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Change-Id: Ie948bc10f44dccf91558099a5217ea1fb5ce7057
Reviewed-on: https://review.whamcloud.com/43881
Tested-by: jenkins <devops@whamcloud.com>
Reviewed-by: Wang Shilong <wshilong@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
12 months agoe2fsck: adjust number of threads
Wang Shilong [Wed, 9 Sep 2020 07:07:10 +0000 (15:07 +0800)]
e2fsck: adjust number of threads

number of threads should not exceed flex bg numbers,
and output messages if we adjust threads number.

E2fsprogs-commit: 56070582ad951aa55b6eeecd7e8e2a19d53bbff6

Change-Id: I8bc87e6ca99205d7fc9e0c9f35d0ba9c3eed38ca
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: allow admin specify number of threads
Wang Shilong [Mon, 9 Mar 2020 07:11:10 +0000 (15:11 +0800)]
e2fsck: allow admin specify number of threads

-m option is introduced to specify number of threads for pfsck.

E2fsprogs-commit: e3bb22efe1c62f8c81686e1d8d065a4943e14bd3

Change-Id: I856d06e0c62f23bdd84e7b6bf53cdc4a46dd58da
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: kickoff mutex lock for block found map
Wang Shilong [Wed, 11 Mar 2020 02:12:18 +0000 (10:12 +0800)]
e2fsck: kickoff mutex lock for block found map

Now @block_found_map is no longer shared by multiple threads,
and @block_dup_map need be checked again after threads finish.

E2fsprogs-commit: 17df3e9f85594eca5515630753542ab5a5c92ee1

Change-Id: I4fb9ae84b70b7ee7d7ce2c0f7a8876ab89c31b5c
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Saranya Muruganandam <saranyamohan@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: merge EA blocks properly
Wang Shilong [Thu, 12 Mar 2020 08:05:51 +0000 (16:05 +0800)]
e2fsck: merge EA blocks properly

EA blocks might be shared, merge them carefully.

E2fsprogs-commit: b9c38a69ff1af2e0c2b9ddce96de965ce62b0de6

Change-Id: Ib01b7cf9c32f7e119bfd00972488b731b13f1ed9
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Saranya Muruganandam <saranyamohan@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: split and merge invalid bitmaps
Wang Shilong [Sat, 14 Dec 2019 05:12:36 +0000 (13:12 +0800)]
e2fsck: split and merge invalid bitmaps

Invalid bitmaps are splitted per thread, and we
should merge them after thread finish.

E2fsprogs-commit: e3e15f5c8f72f95d65d602929f59c1c78e72c5a0

Change-Id: I004f595c107adf390f3b16b90cb4b6aa3650f9de
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: move some fixes out of parallel pthreads
Wang Shilong [Sat, 7 Mar 2020 14:00:45 +0000 (22:00 +0800)]
e2fsck: move some fixes out of parallel pthreads

We could only use @found_map_block to find free blocks
after we have collectd all used blocks, so something like
handle_fs_bad_blocks(), ext2fs_create_resize_inode(),
e2fsck_pass1_dupblocks() really should be handled after
all threads has been finished.

E2fsprogs-commit: 6b1d459bf6018eea7ff0a0ae6b555276718dcda3

Change-Id: Ibdbaed3d218623daa0baec31837037d773799262
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Saranya Muruganandam <saranyamohan@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: serialize fix operations
Wang Shilong [Fri, 6 Mar 2020 08:46:47 +0000 (16:46 +0800)]
e2fsck: serialize fix operations

Allow different threads to fix at the same time could
be dangerous and error-prone now, and most of time
parallel scanning and checking is important.

So this patch adds a mutex to serialize
fix operations during pass1.

And the good benefit of this, we don't need block
allocations and free, superblock updates protection
any more, since only fix operations during pass1
could touch them.

E2fsprogs-commit: 5485120b668c952b6e83f24a1e1a74eaa7bf362e

Change-Id: Iedb64545240cc9325dc0c4ce737a970b27f4f3f6
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Saranya Muruganandam <saranyamohan@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: merge quota context after threads finish
Wang Shilong [Fri, 6 Mar 2020 15:08:07 +0000 (23:08 +0800)]
e2fsck: merge quota context after threads finish

Every threads calculate its own quota accounting,
merge them after threads finish.

E2fsprogs-commit: a0af18577fbb960eb20695afbb4af4e23b864909

Change-Id: If9f0e17e560715ed55183f85c443f21ebbce7d40
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: merge context flags properly
Wang Shilong [Mon, 24 Feb 2020 05:08:58 +0000 (13:08 +0800)]
e2fsck: merge context flags properly

e2fsck might restart after pass1, so we should keep
flags if possible, this patch try to fix f_illitable_flexbg failure

E2fsprogs-commit: c36fa1b28bd4171f53ab4f89a90f6b3b8bf80fd2

Change-Id: I6dda1520a2ff6255d244461af75d7609f47ba92b
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Saranya Muruganandam <saranyamohan@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: merge dirs_to_hash when threads finish
Wang Shilong [Mon, 24 Feb 2020 04:44:06 +0000 (12:44 +0800)]
e2fsck: merge dirs_to_hash when threads finish

@dirs_to_hash list need be merged after threads finish,
test covered by t_dangerous.

E2fsprogs-commit: e2db6c19943065909c11af3b03c2904c35e5e4f8

Change-Id: I99b44c40afa22ce8e4530c874b077493cf969e03
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: merge dx_dir_info after threads finish
Wang Shilong [Fri, 6 Dec 2019 15:37:53 +0000 (23:37 +0800)]
e2fsck: merge dx_dir_info after threads finish

Merge properly.

E2fsprogs-commit: 45519b8167a00979e86d5a5aa95d0bfb07f72520

Change-Id: Ifaed7a0022b716c42df908327c41aaaaa39aa365
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: merge fs flags when threads finish
Li Xi [Sun, 8 Sep 2019 08:15:35 +0000 (16:15 +0800)]
e2fsck: merge fs flags when threads finish

merge fs flags properly.

E2fsprogs-commit: d4c0519725d59c9380acc828d24bc070c5bc2c03

Change-Id: I80c773bebc0e321bf2fe3b436c3f0e601b1f1b79
Signed-off-by: Li Xi <lixi@ddn.com>
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: merge counts after threads finish
Wang Shilong [Tue, 22 Sep 2020 13:14:39 +0000 (21:14 +0800)]
e2fsck: merge counts after threads finish

Merge counts properly.

E2fsprogs-commit: 057060ccabf9b5b47afe435d90a1f7f68c03a231

Change-Id: Ieace889f0d1a20a1990424113b0e8fc81f945266
Signed-off-by: Li Xi <lixi@ddn.com>
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Saranya Muruganandam <saranyamohan@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: add debug codes for multiple threads
Li Xi [Thu, 5 Sep 2019 11:40:36 +0000 (19:40 +0800)]
e2fsck: add debug codes for multiple threads

These debug codes are added to run the multiple pass1 check
thread one by one in order. If all the codes are correct,
fsck of multiple threads should have exactly the same outcome
with single thread.

E2fsprogs-commit: a8b483a1e7987a6dddc32bc82db06f649d245630

Change-Id: I41b0493a0cd830b76434ff0f86a87c98d8d3ad22
Signed-off-by: Li Xi <lixi@ddn.com>
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: merge dblist after thread finishes
Li Xi [Thu, 5 Sep 2019 08:30:40 +0000 (16:30 +0800)]
e2fsck: merge dblist after thread finishes

Merge dblist properly.

E2fsprogs-commit: f3c2256490297201c63139ba0fea5c6b09bca3cf

Change-Id: Ifd907035ed9762c446c688ccc5af3f3fcaa52d24
Signed-off-by: Li Xi <lixi@ddn.com>
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: merge icounts after thread finishes
Li Xi [Wed, 4 Sep 2019 10:47:14 +0000 (18:47 +0800)]
e2fsck: merge icounts after thread finishes

Merge inode_count and inode_link_info properly after
threads finish.

E2fsprogs-commit: bb78742b5e2e95c3e7e9a76755144fdf8ec12285

Change-Id: If40930a87595010963fd59a54709a4d2b27d223b
Signed-off-by: Li Xi <lixi@ddn.com>
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: merge badblocks after thread finishes
Wang Shilong [Tue, 22 Sep 2020 10:00:39 +0000 (18:00 +0800)]
e2fsck: merge badblocks after thread finishes

Badblocks should be merged properly after threads finish.

E2fsprogs-commit: f39197f53ef5d90cd6952d7e75f2d2e3a02bf7ef

Change-Id: Ieb9a96a5e4438270ba6e939155e2666bb53177e4
Signed-off-by: Li Xi <lixi@ddn.com>
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: rbtree bitmap for dir
Wang Shilong [Mon, 21 Sep 2020 11:02:33 +0000 (19:02 +0800)]
e2fsck: rbtree bitmap for dir

Only rbtree support merge operation now, use it for bitmaps.

E2fsprogs-commit: 9bab3284e30752bccc36be3e07b9ca4dfd218041

Change-Id: I42aeb009ece78a4baf4369aff3355fc9df51089a
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: merge dir_info after thread finishes
Li Xi [Mon, 2 Sep 2019 08:14:38 +0000 (16:14 +0800)]
e2fsck: merge dir_info after thread finishes

dir_info need be merged after thread finish.

E2fsprogs-commit: fdbf5e226cf931a5713414001b95e4697997a464

Change-Id: I7e33c0035ec8c6bfa7c13d759e3edb468634eae3
Signed-off-by: Li Xi <lixi@ddn.com>
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: optimize the inserting of dir_info_db
Li Xi [Fri, 30 Aug 2019 09:56:10 +0000 (17:56 +0800)]
e2fsck: optimize the inserting of dir_info_db

Binary search is now used when inserting an dir info to the array.
Memmove is now used when moving array. Both of them improves
the performance of inserting.

This patch is also a prepartion for the merging of two dir db
arrays.

E2fsprogs-commit: 13d3c76475d050941d244c85cbb9256ffd980f88

Change-Id: I69041084dbd7e36eefa44744c3fb3737af8e906e
Signed-off-by: Li Xi <lixi@ddn.com>
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoe2fsck: do not change global variables
Li Xi [Fri, 30 Aug 2019 04:20:11 +0000 (12:20 +0800)]
e2fsck: do not change global variables

Global variables used in pass1 check are changed to local variables
in this patch. This will avoid conflict between threads.

E2fsprogs-commit: 5a1b22da5ca3e90728cee1dbe6bc1920b245dded

Change-Id: Ifc682874801af15e8e13764832ca569b7da4ac57
Signed-off-by: Li Xi <lixi@ddn.com>
Signed-off-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>