2015-11-28 21:51:41

by Darrick J. Wong

[permalink] [raw]
Subject: [PATCH] e2fsck: zap extent-format inode with no extent header

The kernel requires all inodes with the extent flag set to have a
valid extent tree header in i_block. The ext2fs_extent_open2 prefers
to initialize the header if i_block is zeroed, but e2fsck never writes
the new header to disk. Since the kernel won't create inodes with the
flag and no header anyway, zap such files.

Reported-by: Bo Branten <[email protected]>
Signed-off-by: Darrick J. Wong <[email protected]>
---
e2fsck/pass1.c | 13 +++++++++++++
e2fsck/problem.c | 5 +++++
e2fsck/problem.h | 3 +++
tests/f_extents/expect.1 | 3 +--
tests/f_zeroed_ext_header/expect.1 | 22 ++++++++++++++++++++++
tests/f_zeroed_ext_header/expect.2 | 7 +++++++
tests/f_zeroed_ext_header/image.gz | Bin
tests/f_zeroed_ext_header/name | 1 +
8 files changed, 52 insertions(+), 2 deletions(-)
create mode 100644 tests/f_zeroed_ext_header/expect.1
create mode 100644 tests/f_zeroed_ext_header/expect.2
create mode 100644 tests/f_zeroed_ext_header/image.gz
create mode 100644 tests/f_zeroed_ext_header/name

diff --git a/e2fsck/pass1.c b/e2fsck/pass1.c
index 0a8e23e..bc986c9 100644
--- a/e2fsck/pass1.c
+++ b/e2fsck/pass1.c
@@ -2820,7 +2820,20 @@ static void check_blocks_extents(e2fsck_t ctx, struct problem_context *pctx,
ext2_ino_t ino = pctx->ino;
errcode_t retval;
blk64_t eof_lblk;
+ struct ext3_extent_header *eh;

+ /* Check for a proper extent header... */
+ eh = (struct ext3_extent_header *) &inode->i_block[0];
+ retval = ext2fs_extent_header_verify(eh, sizeof(inode->i_block));
+ if (retval) {
+ if (fix_problem(ctx, PR_1_MISSING_EXTENT_HEADER, pctx))
+ e2fsck_clear_inode(ctx, ino, inode, 0,
+ "check_blocks_extents");
+ pctx->errcode = 0;
+ return;
+ }
+
+ /* ...since this function doesn't fail if i_block is zeroed. */
pctx->errcode = ext2fs_extent_open2(fs, ino, inode, &ehandle);
if (pctx->errcode) {
if (fix_problem(ctx, PR_1_READ_EXTENT, pctx))
diff --git a/e2fsck/problem.c b/e2fsck/problem.c
index 6a35e93..971b0ad 100644
--- a/e2fsck/problem.c
+++ b/e2fsck/problem.c
@@ -1119,6 +1119,11 @@ static struct e2fsck_problem problem_table[] = {
N_("@i %i on bigalloc @f cannot be @b mapped. "),
PROMPT_FIX, 0 },

+ /* Inode has corrupt extent header */
+ { PR_1_MISSING_EXTENT_HEADER,
+ N_("@i %i has corrupt @x header. "),
+ PROMPT_CLEAR_INODE, 0 },
+
/* Pass 1b errors */

/* Pass 1B: Rescan for duplicate/bad blocks */
diff --git a/e2fsck/problem.h b/e2fsck/problem.h
index 943024b..bc1a5d5 100644
--- a/e2fsck/problem.h
+++ b/e2fsck/problem.h
@@ -654,6 +654,9 @@ struct problem_context {
/* bigalloc fs cannot have blockmap files */
#define PR_1_NO_BIGALLOC_BLOCKMAP_FILES 0x010080

+/* missing extent header */
+#define PR_1_MISSING_EXTENT_HEADER 0x010081
+
/*
* Pass 1b errors
*/
diff --git a/tests/f_extents/expect.1 b/tests/f_extents/expect.1
index e9d9e5b..da65f94 100644
--- a/tests/f_extents/expect.1
+++ b/tests/f_extents/expect.1
@@ -27,8 +27,7 @@ Inode 17 extent tree (at level 1) could be shorter. Fix? yes

Inode 17, i_blocks is 32, should be 0. Fix? yes

-Error while reading over extent tree in inode 18: Corrupt extent header
-Clear inode? yes
+Inode 18 has corrupt extent header. Clear inode? yes

Inode 18, i_blocks is 2, should be 0. Fix? yes

diff --git a/tests/f_zeroed_ext_header/expect.1 b/tests/f_zeroed_ext_header/expect.1
new file mode 100644
index 0000000..2613e9f
--- /dev/null
+++ b/tests/f_zeroed_ext_header/expect.1
@@ -0,0 +1,22 @@
+Pass 1: Checking inodes, blocks, and sizes
+Inode 12 has corrupt extent header. Clear inode? yes
+
+Pass 2: Checking directory structure
+Entry 'testa' in / (2) has deleted/unused inode 12. Clear? yes
+
+Pass 3: Checking directory connectivity
+Pass 4: Checking reference counts
+Pass 5: Checking group summary information
+Inode bitmap differences: -12
+Fix? yes
+
+Free inodes count wrong for group #0 (115, counted=116).
+Fix? yes
+
+Free inodes count wrong (115, counted=116).
+Fix? yes
+
+
+test_filesys: ***** FILE SYSTEM WAS MODIFIED *****
+test_filesys: 12/128 files (0.0% non-contiguous), 15/256 blocks
+Exit status is 1
diff --git a/tests/f_zeroed_ext_header/expect.2 b/tests/f_zeroed_ext_header/expect.2
new file mode 100644
index 0000000..177288f
--- /dev/null
+++ b/tests/f_zeroed_ext_header/expect.2
@@ -0,0 +1,7 @@
+Pass 1: Checking inodes, blocks, and sizes
+Pass 2: Checking directory structure
+Pass 3: Checking directory connectivity
+Pass 4: Checking reference counts
+Pass 5: Checking group summary information
+test_filesys: 12/128 files (0.0% non-contiguous), 15/256 blocks
+Exit status is 0
diff --git a/tests/f_zeroed_ext_header/image.gz b/tests/f_zeroed_ext_header/image.gz
new file mode 100644
index 0000000000000000000000000000000000000000..67a43341680f578cddb53f54ff7702887dbff384
GIT binary patch
literal 1447
zcmb2|=3o$(j|yXAetXw2TR2dL;X`?N>kKz(Z^1b^>YCHe^o6qAm0?+$7xp+r+H2hf
zM}xmSuWtD6Z<xQJd3N{W*=a)UN{gy%rti++U}aC1zI>nm-}_HRKy-ZmaoZi!TNvhq
z%-!SRwBqR5?GkT4{oT@2*X(yw&+5mHGIQR&_vWtmtc&;|^Sb%G*QvKTl{c<RTu(l8
z<i>8B|G!S}ojuij{=4(_6}z?;$vm%m_;mN@?bq`1_e9)yWfLD&IlumX)Xw}LHpet4
zCNVJhOxUu=-u~+T|M}wI|1dBx90+-7`}=SmkoDky`N<l2AY<9xgnRSX*UKIWw5u;D
z&=vc-=>O6@!K3qL21sUwe_X-#-yz*x?Dy*SC-tAMTb7qC$<zJu<mUO~^RFI0w%<gD
zZ>zQNVyUkCW-EF({tw@Mber7qz1@bd?%w}dE}|TNum0ZUBFk@j<?f&NYplLM`JdFr
zhcR~QwLoi;z%jl5`}aqATspS?dgiSRk=HTePNsJ=wr<w$kG?&-zpHfp*8;1Pv);aO
zELJ~#-F*LC?N_fi?yFe*wcxjBv2k{a^-OPWplQO%WhwvuC$F3FpBcytUiMB1NHzSl
p2a*Z@_<<w{$Sg0N4q=UwqaiRF0z)wb9=QDGRO%~w&%hwS004O^>~R19

literal 0
HcmV?d00001

diff --git a/tests/f_zeroed_ext_header/name b/tests/f_zeroed_ext_header/name
new file mode 100644
index 0000000..31394cf
--- /dev/null
+++ b/tests/f_zeroed_ext_header/name
@@ -0,0 +1 @@
+zap inode with zeroed extent header


2015-11-29 17:40:02

by Bo Branten

[permalink] [raw]
Subject: Re: [PATCH] e2fsck: zap extent-format inode with no extent header


Thank you for this patch, I tryed it on my computer and it does fix the
error our bug caused:

bo@ubuntu:~/src/e2fsprogs-1.42.13$ sudo mount -t ext4 /dev/sdb5 /mnt
bo@ubuntu:~/src/e2fsprogs-1.42.13$ ls -l /mnt
ls: cannot access /mnt/testa: Input/output error
total 20
drwx------ 2 root root 16384 Nov 28 18:09 lost+found
-????????? ? ? ? ? ? testa
-rw-r--r-- 1 root root 1 Nov 28 18:15 testb
bo@ubuntu:~/src/e2fsprogs-1.42.13$ sudo umount /mnt
bo@ubuntu:~/src/e2fsprogs-1.42.13$ sudo ./e2fsck/e2fsck -fy /dev/sdb5
e2fsck 1.42.13 (17-May-2015)
Pass 1: Checking inodes, blocks, and sizes
Inode 12 has corrupt extent header. Clear inode? yes

Pass 2: Checking directory structure
Entry 'testa' in / (2) has deleted/unused inode 12. Clear? yes

Pass 3: Checking directory connectivity
Pass 4: Checking reference counts
Pass 5: Checking group summary information
Inode bitmap differences: -12
Fix? yes

Free inodes count wrong for group #0 (115, counted=116).
Fix? yes

Free inodes count wrong (115, counted=116).
Fix? yes


/dev/sdb5: ***** FILE SYSTEM WAS MODIFIED *****
/dev/sdb5: 12/128 files (0.0% non-contiguous), 15/256 blocks
bo@ubuntu:~/src/e2fsprogs-1.42.13$ sudo mount -t ext4 /dev/sdb5 /mnt
bo@ubuntu:~/src/e2fsprogs-1.42.13$ ls -l /mnt
total 20
drwx------ 2 root root 16384 Nov 28 18:09 lost+found
-rw-r--r-- 1 root root 1 Nov 28 18:15 testb
bo@ubuntu:~/src/e2fsprogs-1.42.13$


2015-11-30 19:22:20

by Theodore Ts'o

[permalink] [raw]
Subject: Re: [PATCH] e2fsck: zap extent-format inode with no extent header

On Sat, Nov 28, 2015 at 01:51:32PM -0800, Darrick J. Wong wrote:
> The kernel requires all inodes with the extent flag set to have a
> valid extent tree header in i_block. The ext2fs_extent_open2 prefers
> to initialize the header if i_block is zeroed, but e2fsck never writes
> the new header to disk. Since the kernel won't create inodes with the
> flag and no header anyway, zap such files.
>
> Reported-by: Bo Branten <[email protected]>
> Signed-off-by: Darrick J. Wong <[email protected]>

Thanks, applied.

- Ted