Skip to contentSkip to conversation

ext4: EXT4_I(sbi->s_buddy_cache)->i_state_flags is not initialized

PATCH In review Jiucheng Xu 2 participants 3 messages Jan 27 – Jan 28

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ext4: EXT4_I(sbi->s_buddy_cache)->i_state_flags is not initialized

The i_state_flags originates from an inode that was previously destroyed and then allocated to s_buddy_cache; it requires reinitialization.

The relevant log during umount is shown below:

EXT4-fs (mmcblk0p28): unmounting filesystem xxx-xxx EXT4-fs (mmcblk0p28): Inode 1 (39878178): inode tracked as orphan!

39878178: 1411f3c7 e0182705 78cc454d ac11f000  .....'..ME.x....
da10433b: 1a2e0146 792e03d0 9c2a04d1 0c788ad3  F......y..*...x.
a91573cf: 44270388 4f4202ea 721a12ea 340cbce0  ..'D..BO...r...4
89cb2f37: 0d13f000 4f270414 1a0b01f0 4f880fe0  ......'O.......O
810e3bc2: 3f0c02f0 482b0009 02e048d0 83f43f2a  ...?..+H.H..*?..
3f37c9f7: 02880aaf 00000000 00000000 00000000  ................
---
 fs/ext4/mballoc.c | 3 +++
 1 file changed, 3 insertions(+)
fs/ext4/mballoc.c +3 -0
index dbc82b65f810fed89da7fa7149d3a05de6f107d6..20b07b2bea31ea81ffbd0b4ace3a7b218c8f4dd5 100644
@@ -3521,6 +3521,9 @@ static int ext4_mb_init_backend(struct super_block *sb)
 	sbi->s_buddy_cache->i_ino = EXT4_BAD_INO;
 	EXT4_I(sbi->s_buddy_cache)->i_disksize = 0;
 	ext4_set_inode_mapping_order(sbi->s_buddy_cache);
+#if (BITS_PER_LONG < 64)
+	ext4_clear_state_flags(EXT4_I(sbi->s_buddy_cache));
+#endif
 
 	for (i = 0; i < ngroups; i++) {
 		cond_resched();

---
base-commit: 4f5e8e6f012349a107531b02eed5b5ace6181449
change-id: 20260126-origin-dev-9f84135b9555

Best regards,

Signature · 1 line
Jiucheng Xu <[email protected]>
reviewer
#1 lore raw
Quoted 10 lines · On Tue 27-01-26 17:34:10, Jiucheng Xu via B4 Relay wrote:

From: Jiucheng Xu <[email protected]>

The i_state_flags originates from an inode that was previously destroyed and then allocated to s_buddy_cache; it requires reinitialization.

The relevant log during umount is shown below:

EXT4-fs (mmcblk0p28): unmounting filesystem xxx-xxx EXT4-fs (mmcblk0p28): Inode 1 (39878178): inode tracked as orphan!

Quoted 8 lines
39878178: 1411f3c7 e0182705 78cc454d ac11f000  .....'..ME.x....
da10433b: 1a2e0146 792e03d0 9c2a04d1 0c788ad3  F......y..*...x.
a91573cf: 44270388 4f4202ea 721a12ea 340cbce0  ..'D..BO...r...4
89cb2f37: 0d13f000 4f270414 1a0b01f0 4f880fe0  ......'O.......O
810e3bc2: 3f0c02f0 482b0009 02e048d0 83f43f2a  ...?..+H.H..*?..
3f37c9f7: 02880aaf 00000000 00000000 00000000  ................

Thanks for the patch but this should be fixed since commit 4091c8206cfd ("ext4: clear i_state_flags when alloc inode"). Can you confirm you cannot reproduce the issue with the latest upstream kernel?

Honza

Quoted 29 lines
---
 fs/ext4/mballoc.c | 3 +++
 1 file changed, 3 insertions(+)
fs/ext4/mballoc.c +3 -1
index dbc82b65f810fed89da7fa7149d3a05de6f107d6..20b07b2bea31ea81ffbd0b4ace3a7b218c8f4dd5 100644
@@ -3521,6 +3521,9 @@ static int ext4_mb_init_backend(struct super_block *sb)
 	sbi->s_buddy_cache->i_ino = EXT4_BAD_INO;
 	EXT4_I(sbi->s_buddy_cache)->i_disksize = 0;
 	ext4_set_inode_mapping_order(sbi->s_buddy_cache);
+#if (BITS_PER_LONG < 64)
+	ext4_clear_state_flags(EXT4_I(sbi->s_buddy_cache));
+#endif
 
 	for (i = 0; i < ngroups; i++) {
 		cond_resched();
 
---

base-commit: 4f5e8e6f012349a107531b02eed5b5ace6181449
change-id: 20260126-origin-dev-9f84135b9555

Best regards,

Signature · 1 line
Jiucheng Xu <[email protected]>
Signature · 2 lines
Jan Kara <[email protected]>
SUSE Labs, CR
#2 lore raw
Quoted 22 lines · Jan Kara

[You don't often get email from [email protected]. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ]

[ EXTERNAL EMAIL ]

Quoted 17 lines · On Tue 27-01-26 17:34:10, Jiucheng Xu via B4 Relay wrote:

From: Jiucheng Xu <[email protected]>

The i_state_flags originates from an inode that was previously destroyed and then allocated to s_buddy_cache; it requires reinitialization.

The relevant log during umount is shown below:

EXT4-fs (mmcblk0p28): unmounting filesystem xxx-xxx EXT4-fs (mmcblk0p28): Inode 1 (39878178): inode tracked as orphan!

39878178: 1411f3c7 e0182705 78cc454d ac11f000  .....'..ME.x....
da10433b: 1a2e0146 792e03d0 9c2a04d1 0c788ad3  F......y..*...x.
a91573cf: 44270388 4f4202ea 721a12ea 340cbce0  ..'D..BO...r...4
89cb2f37: 0d13f000 4f270414 1a0b01f0 4f880fe0  ......'O.......O
810e3bc2: 3f0c02f0 482b0009 02e048d0 83f43f2a  ...?..+H.H..*?..
3f37c9f7: 02880aaf 00000000 00000000 00000000  ................
Quoted 8 lines
Quoted 1 line

Thanks for the patch but this should be fixed since commit 4091c8206cfd ("ext4: clear i_state_flags when alloc inode"). Can you confirm you cannot reproduce the issue with the latest upstream kernel?

Honza

Thanks a lot, Honza! My Android device does not support booting with the latest upstream kernel. It runs 6.12.58, which doesn't include commit 4091c8206cfd. I'm certain this patch will resolve my issue.. Thanks again!

Quoted 32 lines
Quoted 23 lines
---
  fs/ext4/mballoc.c | 3 +++
  1 file changed, 3 insertions(+)
fs/ext4/mballoc.c +3 -1
index dbc82b65f810fed89da7fa7149d3a05de6f107d6..20b07b2bea31ea81ffbd0b4ace3a7b218c8f4dd5 100644
@@ -3521,6 +3521,9 @@ static int ext4_mb_init_backend(struct super_block *sb)
       sbi->s_buddy_cache->i_ino = EXT4_BAD_INO;
       EXT4_I(sbi->s_buddy_cache)->i_disksize = 0;
       ext4_set_inode_mapping_order(sbi->s_buddy_cache);
+#if (BITS_PER_LONG < 64)
+     ext4_clear_state_flags(EXT4_I(sbi->s_buddy_cache));
+#endif
 
       for (i = 0; i < ngroups; i++) {
               cond_resched();
 
---

base-commit: 4f5e8e6f012349a107531b02eed5b5ace6181449
change-id: 20260126-origin-dev-9f84135b9555

Quoted 6 lines

Best regards,
--
Jiucheng Xu <[email protected]>

--
Jan Kara <[email protected]>
SUSE Labs, CR