-
Bug
-
Resolution: Unresolved
-
Medium
-
None
-
None
-
None
-
3
-
9223372036854775807
When we test with some customer's kernel. The crash has been hit:
[ 215.684912] RIP: 0010:memcpy_orig+0x115/0x140 [ 215.689371] Code: 0f 1f 44 00 00 83 fa 04 72 1b 8b 0e 44 8b 44 16 fc 89 0f 44 89 44 17 fc c3 cc cc cc cc 0f 1f 84 00 00 00 00 00 83 ea 01 72 19 <0f> b6 0e 74 12 4c 0f b6 46 01 4c 0f b6 0c 16 44 88 47 01 44 88 0c [ 215.708244] RSP: 0018:ffffbb9080d63b48 EFLAGS: 00010202 [ 215.713569] RAX: ffffa0501e52e998 RBX: 0000000000000002 RCX: ffffffffc0824385 [ 215.720802] RDX: 0000000000000001 RSI: 000000000000001c RDI: ffffa0501e52e998 [ 215.728039] RBP: ffffbb9080d63bb8 R08: 0000000000000007 R09: 6c63615f7869736f [ 215.735272] R10: 6c63615f7869736f R11: ffffffffc071e030 R12: ffffbb9080d63ce8 [ 215.742508] R13: 0000000000000000 R14: 0000000000000018 R15: ffffa04f55091608 [ 215.749742] FS: 00007b209ddad740(0000) GS:ffffa0522fd00000(0000) knlGS:0000000000000000 [ 215.757931] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 215.763775] CR2: 000000000000001c CR3: 000000021eaaa001 CR4: 00000000003706f0 [ 215.771015] Call Trace: [ 215.773578] <TASK> [ 215.775779] mdc_xattr_common+0x4c2/0x5d0 [mdc] [ 215.780434] mdc_setxattr+0xb3/0xd0 [mdc] [ 215.784555] lmv_setxattr+0x2c3/0x310 [lmv] [ 215.788853] ll_set_acl+0x301/0x3e0 [lustre] [ 215.793265] vfs_set_acl+0x2a6/0x380 [ 215.796954] do_set_acl+0x7d/0xd0 [ 215.800463] path_setxattr+0x1ec/0x310 [ 215.804327] __x64_sys_setxattr+0x2f/0x40 [ 215.808529] do_syscall_64+0x61/0x140 [ 215.812310] ? clear_bhb_loop+0x40/0x90 [ 215.816250] entry_SYSCALL_64_after_hwframe+0x76/0x7e [ 215.821405] RIP: 0033:0x7b209deb0fba [ 215.825084] Code: 48 8b 0d 71 0e 0c 00 f7 d8 64 89 01 4
The RSI is source address map to:
....
if (input_size) {
tmp = req_capsule_client_get(&req->rq_pill, &RMF_EADATA);
memcpy(tmp, input, input_size);
}
...........
But it is 0x1c. Look into the binary:
crash> posix_acl_to_xattr
posix_acl_to_xattr = $1 =
{int (user_namespace *, const posix_acl *, void *, size_t)} 0xffffffffa24eb450 <posix_acl_to_xattr>
---- the kernel api is old style
0xffffffffc09febfc <ll_set_acl+204>: and $0xf000,%eax
0xffffffffc09fec01 <ll_set_acl+209>: mov $0xffffffffc0ada5ef,%r15
0xffffffffc09fec08 <ll_set_acl+216>: cmp $0x4000,%eax
0xffffffffc09fec0d <ll_set_acl+221>: je 0xffffffffc09fec33 <ll_set_acl+259>
0xffffffffc09fec0f <ll_set_acl+223>: mov $0xfffffff3,%r15d
0xffffffffc09fec15 <ll_set_acl+229>: test %r13,%r13
0xffffffffc09fec18 <ll_set_acl+232>: jne 0xffffffffc09feea1 <ll_set_acl+881>
0xffffffffc09fec1e <ll_set_acl+238>: mov %r13,-0x58(%rbp)
0xffffffffc09fec22 <ll_set_acl+242>: movq $0xffffffffc0ada5ef,-0x50(%rbp)
0xffffffffc09fec2a <ll_set_acl+250>: jmp 0xffffffffc09fec6d <ll_set_acl+317>
0xffffffffc09fec2c <ll_set_acl+252>: mov $0xffffffffc0ad58bb,%r15
0xffffffffc09fec33 <ll_set_acl+259>: test %r13,%r13
0xffffffffc09fec36 <ll_set_acl+262>: je 0xffffffffc09fec65 <ll_set_acl+309>
0xffffffffc09fec38 <ll_set_acl+264>: lea -0x40(%rbp),%rdx
0xffffffffc09fec3c <ll_set_acl+268>: mov $0xc40,%ecx
0xffffffffc09fec41 <ll_set_acl+273>: mov $0xffffffffa3e5fc20,%rdi
0xffffffffc09fec48 <ll_set_acl+280>: mov %r13,%rsi
0xffffffffc09fec4b <ll_set_acl+283>: call 0xffffffffa24eb450 <posix_acl_to_xattr>
0xffffffffc09fec50 <ll_set_acl+288>: test %eax,%eax
----- calling it without kmalloc
So I think we should add a prefix to remind people that we redefine it, and also to keep the call chain clear.