On 3/3/22 2:04 AM, Jeff Layton wrote:
On Tue, 2022-03-01 at 21:17 +0800, xiubli@xxxxxxxxxx wrote:
From: Xiubo Li <xiubli@xxxxxxxxxx>
Reviewed-by: Jeff Layton <jlayton@xxxxxxxxxx>
Signed-off-by: Xiubo Li <xiubli@xxxxxxxxxx>
---
fs/ceph/dir.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
index 0cf6afe283e9..bf69678d6434 100644
--- a/fs/ceph/dir.c
+++ b/fs/ceph/dir.c
@@ -478,8 +478,10 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx)
2 : (fpos_off(rde->offset) + 1);
err = note_last_dentry(dfi, rde->name, rde->name_len,
next_offset);
- if (err)
+ if (err) {
+ ceph_mdsc_put_request(dfi->last_readdir);
return err;
+ }
} else if (req->r_reply_info.dir_end) {
dfi->next_offset = 2;
/* keep last name */
@@ -521,6 +523,7 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx)
ceph_present_ino(inode->i_sb, le64_to_cpu(rde->inode.in->ino)),
le32_to_cpu(rde->inode.in->mode) >> 12)) {
dout("filldir stopping us...\n");
+ ceph_mdsc_put_request(dfi->last_readdir);
return 0;
}
ctx->pos++;
This patch is reliably causing a KASAN warning about a UAF with xfstest
generic/006:
[ 1170.050701] ==================================================================
[ 1170.054738] BUG: KASAN: use-after-free in ceph_readdir+0x274/0x1cc0 [ceph]
[ 1170.056908] Read of size 4 at addr ffff888116978db8 by task find/2367
[ 1170.058814]
[ 1170.059310] CPU: 1 PID: 2367 Comm: find Tainted: G OE 5.17.0-rc6+ #163
[ 1170.061601] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.15.0-1.fc35 04/01/2014
[ 1170.064099] Call Trace:
[ 1170.064925] <TASK>
[ 1170.065608] dump_stack_lvl+0x59/0x73
[ 1170.066743] print_address_description.constprop.0+0x1f/0x150
[ 1170.068476] ? ceph_readdir+0x274/0x1cc0 [ceph]
[ 1170.069983] kasan_report.cold+0x7f/0x11b
[ 1170.071227] ? ceph_readdir+0x274/0x1cc0 [ceph]
[ 1170.072752] ceph_readdir+0x274/0x1cc0 [ceph]
[ 1170.074212] ? lock_release+0x410/0x410
[ 1170.075393] ? do_raw_spin_unlock+0x86/0xf0
[ 1170.076706] ? mutex_lock_io_nested+0xbc0/0xbc0
[ 1170.087733] ? ceph_d_revalidate+0x7b0/0x7b0 [ceph]
[ 1170.098993] ? down_write_killable+0xc7/0x130
[ 1170.109955] ? __down_interruptible+0x1d0/0x1d0
[ 1170.121021] iterate_dir+0x107/0x2e0
[ 1170.127058] __x64_sys_getdents64+0xe2/0x1b0
[ 1170.131463] ? filldir+0x270/0x270
[ 1170.136034] ? __ia32_sys_getdents+0x1a0/0x1a0
[ 1170.140761] ? lockdep_hardirqs_on_prepare+0x129/0x220
[ 1170.145606] ? syscall_enter_from_user_mode+0x21/0x70
[ 1170.150368] do_syscall_64+0x3b/0x90
[ 1170.154915] entry_SYSCALL_64_after_hwframe+0x44/0xae
[ 1170.159623] RIP: 0033:0x7f7a4bd0ffd7
[ 1170.164023] Code: 19 fb ff 4c 89 e0 5b 5d 41 5c c3 0f 1f 84 00 00 00 00 00 f3 0f 1e fa b8 ff ff ff 7f 48 39 c2 48 0f 47 d0 b8 d9 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 01 c3 48 8b 15 21 0e 12 00 f7 d8 64 89 02 48
[ 1170.174817] RSP: 002b:00007ffd260e00f8 EFLAGS: 00000293 ORIG_RAX: 00000000000000d9
[ 1170.180113] RAX: ffffffffffffffda RBX: 000055b5d33be380 RCX: 00007f7a4bd0ffd7
[ 1170.185346] RDX: 0000000000010000 RSI: 000055b5d33be380 RDI: 0000000000000004
[ 1170.190507] RBP: 000055b5d33be354 R08: 0000000000000003 R09: 0000000000000001
[ 1170.195623] R10: 0000000000000fff R11: 0000000000000293 R12: fffffffffffffe98
[ 1170.200699] R13: 0000000000000000 R14: 000055b5d33be350 R15: 00000000000010fe
[ 1170.205731] </TASK>
[ 1170.210052]
[ 1170.214314] Allocated by task 2367:
[ 1170.218734] kasan_save_stack+0x1e/0x40
[ 1170.223152] __kasan_slab_alloc+0x90/0xc0
[ 1170.227598] kmem_cache_alloc+0x1bc/0x470
[ 1170.232004] ceph_mdsc_create_request+0x2f/0x270 [ceph]
[ 1170.236476] ceph_readdir+0xd8d/0x1cc0 [ceph]
[ 1170.240894] iterate_dir+0x107/0x2e0
[ 1170.245245] __x64_sys_getdents64+0xe2/0x1b0
[ 1170.249498] do_syscall_64+0x3b/0x90
[ 1170.253664] entry_SYSCALL_64_after_hwframe+0x44/0xae
[ 1170.257883]
[ 1170.261666] Freed by task 2367:
[ 1170.265636] kasan_save_stack+0x1e/0x40
[ 1170.269662] kasan_set_track+0x21/0x30
[ 1170.273521] kasan_set_free_info+0x20/0x30
[ 1170.277346] ____kasan_slab_free+0x12f/0x160
[ 1170.281127] slab_free_freelist_hook+0xd6/0x1b0
[ 1170.285122] kmem_cache_free+0x12e/0x590
[ 1170.288917] ceph_readdir+0x15e2/0x1cc0 [ceph]
[ 1170.292839] iterate_dir+0x107/0x2e0
[ 1170.296432] __x64_sys_getdents64+0xe2/0x1b0
[ 1170.300321] do_syscall_64+0x3b/0x90
[ 1170.304117] entry_SYSCALL_64_after_hwframe+0x44/0xae
[ 1170.308067]
[ 1170.311570] The buggy address belongs to the object at ffff888116978a90
[ 1170.311570] which belongs to the cache ceph_mds_request of size 1224
[ 1170.319951] The buggy address is located 808 bytes inside of
[ 1170.319951] 1224-byte region [ffff888116978a90, ffff888116978f58)
[ 1170.327861] The buggy address belongs to the page:
[ 1170.331994] page:000000002aea1f14 refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x116978
[ 1170.336790] head:000000002aea1f14 order:3 compound_mapcount:0 compound_pincount:0
[ 1170.341392] flags: 0x17ffffc0010200(slab|head|node=0|zone=2|lastcpupid=0x1fffff)
[ 1170.346032] raw: 0017ffffc0010200 0000000000000000 dead000000000122 ffff88810c356000
[ 1170.350771] raw: 0000000000000000 0000000080180018 00000001ffffffff 0000000000000000
[ 1170.355530] page dumped because: kasan: bad access detected
[ 1170.359971]
[ 1170.363912] Memory state around the buggy address:
[ 1170.368132] ffff888116978c80: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[ 1170.372861] ffff888116978d00: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[ 1170.377299] >ffff888116978d80: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[ 1170.381594] ^
[ 1170.385746] ffff888116978e00: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[ 1170.390163] ffff888116978e80: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[ 1170.394629] ==================================================================
[ 1170.399041] Disabling lock debugging due to kernel taint
[ 1170.403354] ------------[ cut here ]------------
[ 1170.413479] refcount_t: underflow; use-after-free.
[ 1170.424453] WARNING: CPU: 7 PID: 2367 at lib/refcount.c:28 refcount_warn_saturate+0xc5/0x110
[ 1170.436622] Modules linked in: ceph(OE) libceph(OE) rpcsec_gss_krb5(E) auth_rpcgss(E) nfsv4(E) dns_resolver(E) nfs(E) lockd(E) grace(E) nft_fib_inet(E) nft_fib_ipv4(E) nft_fib_ipv6(E) nft_fib(E) nft_reject_inet(E) nf_reject_ipv4(E) nf_reject_ipv6(E) nft_reject(E) nft_ct(E) nft_chain_nat(E) nf_nat(E) nf_conntrack(E) nf_defrag_ipv6(E) nf_defrag_ipv4(E) bridge(E) ip_set(E) stp(E) llc(E) rfkill(E) nf_tables(E) nfnetlink(E) cachefiles(E) fscache(E) netfs(E) sunrpc(E) iTCO_wdt(E) intel_pmc_bxt(E) iTCO_vendor_support(E) lpc_ich(E) intel_rapl_msr(E) virtio_balloon(E) joydev(E) i2c_i801(E) i2c_smbus(E) intel_rapl_common(E) fuse(E) zram(E) ip_tables(E) xfs(E) crct10dif_pclmul(E) crc32_pclmul(E) crc32c_intel(E) ghash_clmulni_intel(E) serio_raw(E) virtio_gpu(E) virtio_dma_buf(E) drm_shmem_helper(E) virtio_blk(E) drm_kms_helper(E) virtio_console(E) virtio_net(E) cec(E) net_failover(E) failover(E) drm(E) qemu_fw_cfg(E) [last unloaded: libceph]
[ 1170.506605] CPU: 2 PID: 2367 Comm: find Tainted: G B OE 5.17.0-rc6+ #163
[ 1170.521537] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.15.0-1.fc35 04/01/2014
[ 1170.535837] RIP: 0010:refcount_warn_saturate+0xc5/0x110
[ 1170.549516] Code: 88 fd 66 02 01 e8 28 a4 93 00 0f 0b eb 99 80 3d 75 fd 66 02 00 75 90 48 c7 c7 20 9b aa 88 c6 05 65 fd 66 02 01 e8 08 a4 93 00 <0f> 0b e9 76 ff ff ff 80 3d 50 fd 66 02 00 0f 85 69 ff ff ff 48 c7
[ 1170.577563] RSP: 0018:ffff8881507b7cd0 EFLAGS: 00010282
[ 1170.591094] RAX: 0000000000000000 RBX: 0000000000000003 RCX: 0000000000000000
[ 1170.607700] RDX: 0000000000000001 RSI: ffffffff88aadf60 RDI: ffffed102a0f6f90
[ 1170.624135] RBP: ffff888116978ab8 R08: ffffffff87187ec4 R09: ffff8884187bdb47
[ 1170.640644] R10: ffffed10830f7b68 R11: 0000000000000001 R12: ffff8881507b7eb0
[ 1170.655872] R13: 0ff7c9ea00000003 R14: 0000030000009ee1 R15: ffff8881507b7eb8
[ 1170.673101] FS: 00007f7a4bb9d800(0000) GS:ffff888418780000(0000) knlGS:0000000000000000
[ 1170.689317] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 1170.705066] CR2: 00007f2a487a70d8 CR3: 000000011f6be000 CR4: 00000000003506e0
[ 1170.721472] Call Trace:
[ 1170.735726] <TASK>
[ 1170.750251] ? __ia32_sys_getdents+0x1a0/0x1a0
[ 1170.766809] ceph_readdir+0x1428/0x1cc0 [ceph]
[ 1170.782120] ? mutex_lock_io_nested+0xbc0/0xbc0
[ 1170.797099] ? ceph_d_revalidate+0x7b0/0x7b0 [ceph]
[ 1170.812069] ? down_write_killable+0xc7/0x130
[ 1170.825664] ? __down_interruptible+0x1d0/0x1d0
[ 1170.832980] iterate_dir+0x107/0x2e0
[ 1170.845448] __x64_sys_getdents64+0xe2/0x1b0
[ 1170.859218] ? filldir+0x270/0x270
[ 1170.871956] ? __ia32_sys_getdents+0x1a0/0x1a0
[ 1170.885328] ? lockdep_hardirqs_on_prepare+0x129/0x220
[ 1170.898675] ? syscall_enter_from_user_mode+0x21/0x70
[ 1170.911432] do_syscall_64+0x3b/0x90
[ 1170.924215] entry_SYSCALL_64_after_hwframe+0x44/0xae
[ 1170.937487] RIP: 0033:0x7f7a4bd0ffd7
[ 1170.951209] Code: 19 fb ff 4c 89 e0 5b 5d 41 5c c3 0f 1f 84 00 00 00 00 00 f3 0f 1e fa b8 ff ff ff 7f 48 39 c2 48 0f 47 d0 b8 d9 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 01 c3 48 8b 15 21 0e 12 00 f7 d8 64 89 02 48
[ 1170.978466] RSP: 002b:00007ffd260e00f8 EFLAGS: 00000293 ORIG_RAX: 00000000000000d9
[ 1170.985171] RAX: ffffffffffffffda RBX: 000055b5d33be380 RCX: 00007f7a4bd0ffd7
[ 1170.991427] RDX: 0000000000010000 RSI: 000055b5d33be380 RDI: 0000000000000004
[ 1171.006641] RBP: 000055b5d33be354 R08: 0000000000000003 R09: 0000000000000001
[ 1171.022724] R10: 0000000000000fff R11: 0000000000000293 R12: fffffffffffffe98
[ 1171.037966] R13: 0000000000000000 R14: 000055b5d33be350 R15: 00000000000010fe
[ 1171.047204] </TASK>
[ 1171.053887] irq event stamp: 10358
[ 1171.060503] hardirqs last enabled at (10357): [<ffffffff88317cb1>] syscall_enter_from_user_mode+0x21/0x70
[ 1171.067647] hardirqs last disabled at (10358): [<ffffffff88331f80>] _raw_spin_lock_irqsave+0x60/0x70
[ 1171.076273] softirqs last enabled at (8366): [<ffffffff875c292c>] touch_atime+0x36c/0x380
[ 1171.084243] softirqs last disabled at (8362): [<ffffffff87472e50>] wb_wakeup_delayed+0x40/0xa0
[ 1171.090668] ---[ end trace 0000000000000000 ]---
I think the original bugs are real, but I suspect that this fix is
running afoul of some of the last_readdir tracking in this code.
We should reset the 'dfi->last_readdir' to NULL after release the
request. The 'dfi' is a private data in the 'struct file', so for the
next readdir it will try to release it again.
diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
index 0cf6afe283e9..c8720944036f 100644
--- a/fs/ceph/dir.c
+++ b/fs/ceph/dir.c
@@ -478,8 +478,11 @@ static int ceph_readdir(struct file *file, struct
dir_context *ctx)
2 : (fpos_off(rde->offset) + 1);
err = note_last_dentry(dfi, rde->name,
rde->name_len,
next_offset);
- if (err)
+ if (err) {
+ ceph_mdsc_put_request(dfi->last_readdir);
+ dfi->last_readdir = NULL;
return err;
+ }
} else if (req->r_reply_info.dir_end) {
dfi->next_offset = 2;
/* keep last name */
@@ -521,6 +524,8 @@ static int ceph_readdir(struct file *file, struct
dir_context *ctx)
ceph_present_ino(inode->i_sb,
le64_to_cpu(rde->inode.in->ino)),
le32_to_cpu(rde->inode.in->mode) >> 12)) {
dout("filldir stopping us...\n");
+ ceph_mdsc_put_request(dfi->last_readdir);
+ dfi->last_readdir = NULL;
return 0;
}
ctx->pos++;
Ilya, can you drop patch this from master for now?
Thanks,
