Skip to contentSkip to conversation

"btrfs rescue super-recover" memory corruption

PATCH In review Eric Sandeen 3 participants 6 messages Sep 17, 2014 – Sep 22, 2014 Btrfs file system

lore
Share & tools
lore
#0 lore raw

This:

# truncate --size=8g
# dd if=/dev/zero of=file conv=notrunc  bs=4 seek=16384 count=1
# valgrind ./btrfs rescue super-recover file -v

yields:

==4604== Memcheck, a memory error detector
==4604== Copyright (C) 2002-2012, and GNU GPL'd, by Julian Seward et al. ==4604== Using Valgrind-3.8.1 and LibVEX; rerun with -h for copyright info ==4604== Command: ./btrfs rescue super-recover file -v ==4604==
All Devices:
Device: id = 1, name = file

Before Recovering:
[All good supers]:
device name = file
superblock bytenr = 67108864

[All bad supers]:
device name = file
superblock bytenr = 65536

Make sure this is a btrfs disk otherwise the tool will destroy other fs, Are you sure? [y/N]: y Recovered bad superblocks successful
==4604== Invalid read of size 8

==4604==    at 0x426B55: btrfs_recover_superblocks (list.h:204)
==4604==    by 0x421C79: cmd_super_recover (cmds-rescue.c:148)
==4604==    by 0x40420A: handle_command_group (btrfs.c:145)
==4604==    by 0x421B54: cmd_rescue (cmds-rescue.c:162)
==4604==    by 0x404199: main (btrfs.c:247)
==4604==  Address 0x4c250b0 is 48 bytes inside a block of size 96 free'd
==4604==    at 0x4A063F0: free (vg_replace_malloc.c:446)
==4604==    by 0x43C77E: btrfs_close_devices (volumes.c:196)
==4604==    by 0x42F5D1: close_ctree (disk-io.c:1404)
==4604==    by 0x426A85: btrfs_recover_superblocks (super-recover.c:340)
==4604==    by 0x421C79: cmd_super_recover (cmds-rescue.c:148)
==4604==    by 0x40420A: handle_command_group (btrfs.c:145)
==4604==    by 0x421B54: cmd_rescue (cmds-rescue.c:162)
==4604==    by 0x404199: main (btrfs.c:247)
==4604== 
==4604== Invalid free() / delete / delete[] / realloc()
==4604==    at 0x4A063F0: free (vg_replace_malloc.c:446)
==4604==    by 0x426B9E: btrfs_recover_superblocks (super-recover.c:85)
==4604==    by 0x421C79: cmd_super_recover (cmds-rescue.c:148)
==4604==    by 0x40420A: handle_command_group (btrfs.c:145)
==4604==    by 0x421B54: cmd_rescue (cmds-rescue.c:162)
==4604==    by 0x404199: main (btrfs.c:247)
==4604==  Address 0x4c25080 is 0 bytes inside a block of size 96 free'd
==4604==    at 0x4A063F0: free (vg_replace_malloc.c:446)
==4604==    by 0x43C77E: btrfs_close_devices (volumes.c:196)
==4604==    by 0x42F5D1: close_ctree (disk-io.c:1404)
==4604==    by 0x426A85: btrfs_recover_superblocks (super-recover.c:340)
==4604==    by 0x421C79: cmd_super_recover (cmds-rescue.c:148)
==4604==    by 0x40420A: handle_command_group (btrfs.c:145)
==4604==    by 0x421B54: cmd_rescue (cmds-rescue.c:162)
==4604==    by 0x404199: main (btrfs.c:247)

==4604==
==4604==
==4604== HEAP SUMMARY:
==4604== in use at exit: 0 bytes in 0 blocks
==4604== total heap usage: 72 allocs, 73 frees, 140,384 bytes allocated
==4604==
==4604== All heap blocks were freed -- no leaks are possible ==4604==
==4604== For counts of detected and suppressed errors, rerun with: -v ==4604== ERROR SUMMARY: 2 errors from 2 contexts (suppressed: 6 from 6)

i.e. I think we are double freeing memory:

        close_ctree(root); // <-- here

no_recover:
recover_err_str(ret); free_recover_superblock(&recover); // <-- and here

I can't really work out what all this is all doing, but maybe the fix is obvious to Wang Shilong (who wrote the original code)?

Thanks,
-Eric

#1 lore raw
Quoted 3 lines · Eric Sandeen

This:

# truncate --size=8g

oops, s/b:

# truncate --size=8g file
Quoted 2 lines
# dd if=/dev/zero of=file conv=notrunc  bs=4 seek=16384 count=1
# valgrind ./btrfs rescue super-recover file -v
#2 lore raw
In reply to Eric Sandeen

Hi Eric,

Quoted 73 lines

This:

# truncate --size=8g
# dd if=/dev/zero of=file conv=notrunc  bs=4 seek=16384 count=1
# valgrind ./btrfs rescue super-recover file -v

yields:

==4604== Memcheck, a memory error detector
==4604== Copyright (C) 2002-2012, and GNU GPL'd, by Julian Seward et al. ==4604== Using Valgrind-3.8.1 and LibVEX; rerun with -h for copyright info ==4604== Command: ./btrfs rescue super-recover file -v ==4604==
All Devices:
Device: id = 1, name = file

Before Recovering:
[All good supers]:
device name = file
superblock bytenr = 67108864

[All bad supers]:
device name = file
superblock bytenr = 65536

Make sure this is a btrfs disk otherwise the tool will destroy other fs, Are you sure? [y/N]: y Recovered bad superblocks successful
==4604== Invalid read of size 8

==4604==    at 0x426B55: btrfs_recover_superblocks (list.h:204)
==4604==    by 0x421C79: cmd_super_recover (cmds-rescue.c:148)
==4604==    by 0x40420A: handle_command_group (btrfs.c:145)
==4604==    by 0x421B54: cmd_rescue (cmds-rescue.c:162)
==4604==    by 0x404199: main (btrfs.c:247)
==4604==  Address 0x4c250b0 is 48 bytes inside a block of size 96 free'd
==4604==    at 0x4A063F0: free (vg_replace_malloc.c:446)
==4604==    by 0x43C77E: btrfs_close_devices (volumes.c:196)
==4604==    by 0x42F5D1: close_ctree (disk-io.c:1404)
==4604==    by 0x426A85: btrfs_recover_superblocks (super-recover.c:340)
==4604==    by 0x421C79: cmd_super_recover (cmds-rescue.c:148)
==4604==    by 0x40420A: handle_command_group (btrfs.c:145)
==4604==    by 0x421B54: cmd_rescue (cmds-rescue.c:162)
==4604==    by 0x404199: main (btrfs.c:247)
==4604== 
==4604== Invalid free() / delete / delete[] / realloc()
==4604==    at 0x4A063F0: free (vg_replace_malloc.c:446)
==4604==    by 0x426B9E: btrfs_recover_superblocks (super-recover.c:85)
==4604==    by 0x421C79: cmd_super_recover (cmds-rescue.c:148)
==4604==    by 0x40420A: handle_command_group (btrfs.c:145)
==4604==    by 0x421B54: cmd_rescue (cmds-rescue.c:162)
==4604==    by 0x404199: main (btrfs.c:247)
==4604==  Address 0x4c25080 is 0 bytes inside a block of size 96 free'd
==4604==    at 0x4A063F0: free (vg_replace_malloc.c:446)
==4604==    by 0x43C77E: btrfs_close_devices (volumes.c:196)
==4604==    by 0x42F5D1: close_ctree (disk-io.c:1404)
==4604==    by 0x426A85: btrfs_recover_superblocks (super-recover.c:340)
==4604==    by 0x421C79: cmd_super_recover (cmds-rescue.c:148)
==4604==    by 0x40420A: handle_command_group (btrfs.c:145)
==4604==    by 0x421B54: cmd_rescue (cmds-rescue.c:162)
==4604==    by 0x404199: main (btrfs.c:247)

==4604==
==4604==
==4604== HEAP SUMMARY:
==4604== in use at exit: 0 bytes in 0 blocks
==4604== total heap usage: 72 allocs, 73 frees, 140,384 bytes allocated
==4604==
==4604== All heap blocks were freed -- no leaks are possible ==4604==
==4604== For counts of detected and suppressed errors, rerun with: -v ==4604== ERROR SUMMARY: 2 errors from 2 contexts (suppressed: 6 from 6)

i.e. I think we are double freeing memory:

Quoted 7 lines
       close_ctree(root); // <-- here

no_recover:
recover_err_str(ret); free_recover_superblock(&recover); // <-- and here

I can't really work out what all this is all doing, but maybe the fix is obvious to Wang Shilong (who wrote the original code)?

Though i no longer spend much time on btrfs, i will take a look at this. ^_^

Thanks Eric

Quoted 3 lines

Thanks,
-Eric

Best Regards,
Wang Shilong

#3 lore raw

PATCH 1/1 Btrfs-progs: super-recover: fix double free fs_devices memory

In reply to Eric Sandeen

super-recover collects btrfs devices infomation using existed functions scan_one_devices().

Problem is fs_devices is freed twice in close_ctree() and free_recover_superblock() for super correction path.

Fix this problem by checking whether fs_devices memory have been freed before we free it.

Cc Eric Sandeen <[email protected]>
Cc Chris Murphy <[email protected]>
---
 super-recover.c | 13 +++----------
 1 file changed, 3 insertions(+), 10 deletions(-)
super-recover.c +3 -10
index 767de4b..419b86a 100644
@@ -69,21 +69,11 @@ void init_recover_superblock(struct btrfs_recover_superblock *recover)
 static
 void free_recover_superblock(struct btrfs_recover_superblock *recover)
 {
-	struct btrfs_device *device;
 	struct super_block_record *record;
 
 	if (!recover->fs_devices)
 		return;
 
-	while (!list_empty(&recover->fs_devices->devices)) {
-		device = list_entry(recover->fs_devices->devices.next,
-				struct btrfs_device, dev_list);
-		list_del_init(&device->dev_list);
-		free(device->name);
-		free(device);
-	}
-	free(recover->fs_devices);
-
 	while (!list_empty(&recover->good_supers)) {
 		record = list_entry(recover->good_supers.next,
 				struct super_block_record, list);
@@ -341,6 +331,9 @@ int btrfs_recover_superblocks(const char *dname,
 no_recover:
 	recover_err_str(ret);
 	free_recover_superblock(&recover);
+	/* check if we have freed fs_deivces in close_ctree() */
+	if (!root)
+		btrfs_close_devices(recover.fs_devices);
 	return ret;
 }
 
Signature · 1 line
1.9.3
#4 lore raw

Re: [PATCH] Btrfs-progs: super-recover: fix double free fs_devices memory

Quoted 43 lines · Wang Shilong

super-recover collects btrfs devices infomation using existed functions scan_one_devices().

Problem is fs_devices is freed twice in close_ctree() and free_recover_superblock() for super correction path.

Fix this problem by checking whether fs_devices memory have been freed before we free it.

Cc Eric Sandeen <[email protected]>
Cc Chris Murphy <[email protected]>
---
 super-recover.c | 13 +++----------
 1 file changed, 3 insertions(+), 10 deletions(-)
super-recover.c +0 -10
index 767de4b..419b86a 100644
@@ -69,21 +69,11 @@ void init_recover_superblock(struct btrfs_recover_superblock *recover)
 static
 void free_recover_superblock(struct btrfs_recover_superblock *recover)
 {
-	struct btrfs_device *device;
 	struct super_block_record *record;
 
 	if (!recover->fs_devices)
 		return;
 
-	while (!list_empty(&recover->fs_devices->devices)) {
-		device = list_entry(recover->fs_devices->devices.next,
-				struct btrfs_device, dev_list);
-		list_del_init(&device->dev_list);
-		free(device->name);
-		free(device);
-	}
-	free(recover->fs_devices);
-
 	while (!list_empty(&recover->good_supers)) {
 		record = list_entry(recover->good_supers.next,
 				struct super_block_record, list);

super-recover.c int btrfs_recover_superblocks(const char *dname, · line 339

Quoted 9 lines
super-recover.c +3 -0
@@ -341,6 +331,9 @@ int btrfs_recover_superblocks(const char *dname,
 no_recover:
 	recover_err_str(ret);
 	free_recover_superblock(&recover);
+	/* check if we have freed fs_deivces in close_ctree() */
+	if (!root)
+		btrfs_close_devices(recover.fs_devices);
 	return ret;
 }
 

nice catch! "+20, recorded". ^_^

author Acked-by on v1 1/1
#5 lore raw

Re: [PATCH] Btrfs-progs: super-recover: fix double free fs_devices memory

In reply to Wang Shilong
Quoted 6 lines · Wang Shilong

super-recover collects btrfs devices infomation using existed functions scan_one_devices().

Problem is fs_devices is freed twice in close_ctree() and free_recover_superblock() for super correction path.

Quoted 6 lines

Fix this problem by checking whether fs_devices memory have been freed before we free it.

Cc Eric Sandeen <[email protected]>
Cc Chris Murphy <[email protected]>

That does seem to fix the testcase. Thanks!

Quoted 41 lines
---
 super-recover.c | 13 +++----------
 1 file changed, 3 insertions(+), 10 deletions(-)
super-recover.c +3 -10
index 767de4b..419b86a 100644
@@ -69,21 +69,11 @@ void init_recover_superblock(struct btrfs_recover_superblock *recover)
 static
 void free_recover_superblock(struct btrfs_recover_superblock *recover)
 {
-	struct btrfs_device *device;
 	struct super_block_record *record;
 
 	if (!recover->fs_devices)
 		return;
 
-	while (!list_empty(&recover->fs_devices->devices)) {
-		device = list_entry(recover->fs_devices->devices.next,
-				struct btrfs_device, dev_list);
-		list_del_init(&device->dev_list);
-		free(device->name);
-		free(device);
-	}
-	free(recover->fs_devices);
-
 	while (!list_empty(&recover->good_supers)) {
 		record = list_entry(recover->good_supers.next,
 				struct super_block_record, list);
@@ -341,6 +331,9 @@ int btrfs_recover_superblocks(const char *dname,
 no_recover:
 	recover_err_str(ret);
 	free_recover_superblock(&recover);
+	/* check if we have freed fs_deivces in close_ctree() */
+	if (!root)
+		btrfs_close_devices(recover.fs_devices);
 	return ret;
 }