Skip to contentSkip to conversation

bcachefs: Mark bch_inode_info as SLAB_ACCOUNT

PATCH Applied Youling Tang 2 participants 5 messages Jul 3, 2024 – Jul 12, 2024

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

PATCH 1/1 bcachefs: Mark bch_inode_info as SLAB_ACCOUNT

From: Youling Tang <[email protected]>

After commit 230e9fc28604 ("slab: add SLAB_ACCOUNT flag"), we need to mark the inode cache as SLAB_ACCOUNT, similar to commit 5d097056c9a0 ("kmemcg:
account for certain kmem allocations to memcg")

---
 fs/bcachefs/fs.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
fs/bcachefs/fs.c +2 -1
index be7eda94b459..092f95426dce 100644
@@ -2126,7 +2126,8 @@ int __init bch2_vfs_init(void)
 {
 	int ret = -ENOMEM;
 
-	bch2_inode_cache = KMEM_CACHE(bch_inode_info, SLAB_RECLAIM_ACCOUNT);
+	bch2_inode_cache = KMEM_CACHE(bch_inode_info, SLAB_RECLAIM_ACCOUNT |
+				      SLAB_ACCOUNT);
 	if (!bch2_inode_cache)
 		goto err;
 
Signature · 1 line
2.34.1
#1 lore raw
Quoted 7 lines · Youling Tang

From: Youling Tang <[email protected]>

After commit 230e9fc28604 ("slab: add SLAB_ACCOUNT flag"), we need to mark the inode cache as SLAB_ACCOUNT, similar to commit 5d097056c9a0 ("kmemcg:
account for certain kmem allocations to memcg")

Thanks, applied

#2 lore raw
In reply to Youling Tang
Quoted 7 lines · Youling Tang

From: Youling Tang <[email protected]>

After commit 230e9fc28604 ("slab: add SLAB_ACCOUNT flag"), we need to mark the inode cache as SLAB_ACCOUNT, similar to commit 5d097056c9a0 ("kmemcg:
account for certain kmem allocations to memcg")

Turns out this was never tested with memcg enabled (!).

I'm reverting it, please feel free to send me a fixed version.

#3 lore raw
Quoted 11 lines · Kent Overstreet
Quoted 7 lines · Youling Tang

From: Youling Tang <[email protected]>

After commit 230e9fc28604 ("slab: add SLAB_ACCOUNT flag"), we need to mark the inode cache as SLAB_ACCOUNT, similar to commit 5d097056c9a0 ("kmemcg:
account for certain kmem allocations to memcg")

Turns out this was never tested with memcg enabled (!).

I'm reverting it, please feel free to send me a fixed version.

Sorry, my oversight.

The following null pointer dereference is triggered after MEMCG configuration is enabled.

```
BUG: kernel NULL pointer dereference, address: 0000000000000008
#PF: supervisor read access in kernel mode
#PF: error_code(0x0000) - not-present page
PGD 0 P4D 0
Oops: Oops: 0000 [#1] SMP
CPU: 5 PID: 1702 Comm: umount Not tainted 6.10.0-rc7-ktest-00003-g557bd05b0d4c-dirty #12
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.15.0-1 04/01/2014
RIP: 0010:list_lru_add+0x83/0x100
Code: 5f 5d c3 48 8b 45 d0 48 85 c0 74 13 41 80 7c 24 1c 00 48 63 b0 68 06 00 00 74 04 85 f6 79 5e 4d 03 2c 24 49 83 c5 08 4c 89 ea <49> 8b 45 08 49 89 5d 08 48 89 13 48 89 43 08 48 89 18 49 8b 45 10
RSP: 0018:ffff8881178efd10 EFLAGS: 00010246
RAX: 0000000000000000 RBX: ffff88810ec140f0 RCX: 0000000000000000
RDX: 0000000000000000 RSI: 0000000000000017 RDI: ffff8881178efcc8
RBP: ffff8881178efd48 R08: ffff8881009de780 R09: ffffffff822e0de0
R10: 0000000000000000 R11: 0000000000000000 R12: ffff888102075c80
R13: 0000000000000000 R14: ffff88810443e6c0 R15: 0000000000000000
FS:  00007f9ed1840800(0000) GS:ffff888179940000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000000000008 CR3: 00000001062b9005 CR4: 0000000000370eb0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400

Call Trace:
 <TASK>
 ? show_regs+0x69/0x70
 ? __die+0x29/0x70
 ? page_fault_oops+0x14f/0x3c0
 ? do_user_addr_fault+0x2d0/0x5b0
 ? default_wake_function+0x1e/0x30
 ? exc_page_fault+0x6d/0x130
 ? asm_exc_page_fault+0x2b/0x30
 ? list_lru_add+0x83/0x100
 list_lru_add_obj+0x4b/0x60
 iput+0x1fe/0x220
 dentry_unlink_inode+0xbd/0x120
 __dentry_kill+0x78/0x180
 dput+0xc7/0x170
 shrink_dcache_for_umount+0xe8/0x120
 generic_shutdown_super+0x23/0x150
 bch2_kill_sb+0x1b/0x30
 deactivate_locked_super+0x34/0xb0
 deactivate_super+0x44/0x50
 cleanup_mnt+0x105/0x160
 __cleanup_mnt+0x16/0x20
 task_work_run+0x63/0x90
 syscall_exit_to_user_mode+0x10d/0x110
 do_syscall_64+0x57/0x100
 entry_SYSCALL_64_after_hwframe+0x4b/0x53
RIP: 0033:0x7f9ed1a7a6e7
Code: 0c 00 f7 d8 64 89 02 b8 ff ff ff ff c3 66 0f 1f 44 00 00 31 f6 e9 09 00 00 00 66 0f 1f 84 00 00 00 00 00 b8 a6 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 01 c3 48 8b 15 09 97 0c 00 f7 d8 64 89 02 b8
RSP: 002b:00007ffef8a29128 EFLAGS: 00000246 ORIG_RAX: 00000000000000a6
RAX: 0000000000000000 RBX: 000055f4671acad8 RCX: 00007f9ed1a7a6e7
RDX: 0000000000000000 RSI: 0000000000000000 RDI: 000055f4671b1240
RBP: 0000000000000000 R08: 0000000000000001 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 00007f9ed1bc6244
R13: 000055f4671b1240 R14: 000055f4671acde0 R15: 000055f4671ac9d0
 </TASK>
```

I'm going to analyze it.

Thanks,
Youling.

#4 lore raw

Hi, Kent

Quoted 65 lines · Youling Tang
Quoted 11 lines · Kent Overstreet
Quoted 7 lines · Youling Tang

From: Youling Tang <[email protected]>

After commit 230e9fc28604 ("slab: add SLAB_ACCOUNT flag"), we need to mark the inode cache as SLAB_ACCOUNT, similar to commit 5d097056c9a0 ("kmemcg:
account for certain kmem allocations to memcg")

Turns out this was never tested with memcg enabled (!).

I'm reverting it, please feel free to send me a fixed version.

Sorry, my oversight.

The following null pointer dereference is triggered after MEMCG configuration is enabled.

```
BUG: kernel NULL pointer dereference, address: 0000000000000008
#PF: supervisor read access in kernel mode
#PF: error_code(0x0000) - not-present page
PGD 0 P4D 0
Oops: Oops: 0000 [#1] SMP
CPU: 5 PID: 1702 Comm: umount Not tainted 6.10.0-rc7-ktest-00003-g557bd05b0d4c-dirty #12
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.15.0-1 04/01/2014
RIP: 0010:list_lru_add+0x83/0x100
Code: 5f 5d c3 48 8b 45 d0 48 85 c0 74 13 41 80 7c 24 1c 00 48 63 b0 68 06 00 00 74 04 85 f6 79 5e 4d 03 2c 24 49 83 c5 08 4c 89 ea <49> 8b 45 08 49 89 5d 08 48 89 13 48 89 43 08 48 89 18 49 8b 45 10
RSP: 0018:ffff8881178efd10 EFLAGS: 00010246
RAX: 0000000000000000 RBX: ffff88810ec140f0 RCX: 0000000000000000
RDX: 0000000000000000 RSI: 0000000000000017 RDI: ffff8881178efcc8
RBP: ffff8881178efd48 R08: ffff8881009de780 R09: ffffffff822e0de0
R10: 0000000000000000 R11: 0000000000000000 R12: ffff888102075c80
R13: 0000000000000000 R14: ffff88810443e6c0 R15: 0000000000000000
FS:  00007f9ed1840800(0000) GS:ffff888179940000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000000000008 CR3: 00000001062b9005 CR4: 0000000000370eb0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400

Call Trace:
 <TASK>
 ? show_regs+0x69/0x70
 ? __die+0x29/0x70
 ? page_fault_oops+0x14f/0x3c0
 ? do_user_addr_fault+0x2d0/0x5b0
 ? default_wake_function+0x1e/0x30
 ? exc_page_fault+0x6d/0x130
 ? asm_exc_page_fault+0x2b/0x30
 ? list_lru_add+0x83/0x100
 list_lru_add_obj+0x4b/0x60
 iput+0x1fe/0x220
 dentry_unlink_inode+0xbd/0x120
 __dentry_kill+0x78/0x180
 dput+0xc7/0x170
 shrink_dcache_for_umount+0xe8/0x120
 generic_shutdown_super+0x23/0x150
 bch2_kill_sb+0x1b/0x30
 deactivate_locked_super+0x34/0xb0
 deactivate_super+0x44/0x50
 cleanup_mnt+0x105/0x160
 __cleanup_mnt+0x16/0x20
 task_work_run+0x63/0x90
 syscall_exit_to_user_mode+0x10d/0x110
 do_syscall_64+0x57/0x100
 entry_SYSCALL_64_after_hwframe+0x4b/0x53
RIP: 0033:0x7f9ed1a7a6e7
Code: 0c 00 f7 d8 64 89 02 b8 ff ff ff ff c3 66 0f 1f 44 00 00 31 f6 e9 09 00 00 00 66 0f 1f 84 00 00 00 00 00 b8 a6 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 01 c3 48 8b 15 09 97 0c 00 f7 d8 64 89 02 b8
Quoted 8 lines
RSP: 002b:00007ffef8a29128 EFLAGS: 00000246 ORIG_RAX: 00000000000000a6
RAX: 0000000000000000 RBX: 000055f4671acad8 RCX: 00007f9ed1a7a6e7
RDX: 0000000000000000 RSI: 0000000000000000 RDI: 000055f4671b1240
RBP: 0000000000000000 R08: 0000000000000001 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 00007f9ed1bc6244
R13: 000055f4671b1240 R14: 000055f4671acde0 R15: 000055f4671ac9d0
 </TASK>
```

The direct cause of the BUG is that the return value of list_lru_from_memcg_idx() is NULL, and the execution of l->list will cause NULL pointer dereference.

The return value of list_lru_from_memcg_idx() needs to be determined, similar to commit 5abc1e37afa0 ("mm: list_lru: allocate list_lru_one only when needed").

Modified as follows:

mm/list_lru.c +0 -0
index 3fd64736bc45..ee7424c3879d 100644
@@ -94,6 +94,9 @@ bool list_lru_add(struct list_lru *lru, struct list_head *item, int nid,

        spin_lock(&nlru->lock);
        if (list_empty(item)) {
                l = list_lru_from_memcg_idx(lru, nid, memcg_kmem_id(memcg)); +               if (!l)
+                       goto out;
+
                list_add_tail(item, &l->list);
                /* Set shrinker bit if the first element was added */
                if (!l->nr_items++)
@@ -102,6 +105,7 @@ bool list_lru_add(struct list_lru *lru, struct list_head *item, int nid, spin_unlock(&nlru->lock);
                return true;
        }
+out:
        spin_unlock(&nlru->lock);
        return false;
 }

After ktest test tests/bcachefs/xfstests.ktest can continue to test (enable MEMCG and MEMCG_KMEM).

Thanks,
Youling.