The ERP kits provided in this Engineering
Advisory (EA) fix several AdvFS related problems that exist in HP
Tru64 UNIX V5.1A PK6, V5.1B-2, and V5.1B-3.
The HP Tru64 UNIX V5.1A PK6 ERP kit combines the fix for a newly
uncovered KMF in access_invalidate problem with a fix previously
announced in EA BU050414_EW01 (Panics During quotactl Calls).
0 stop_secondary_cpu
1 panic
2 event_timeout
3 printf
4 panic
5 trap
6 _XentMM
7 access_invalidate
8 bfs_dealloc
9 dmn_dealloc
10 bs_bfdmn_deactivate
11 fs_fset_get_info
12 msfs_real_syscall
13 msfs_syscall
14 syscall
15 _Xsyscall |
0 boot
1 panic
2 trap
3 _XentMM
4 access_invalidate
5 bfs_dealloc
6 dmn_dealloc
7 bs_bfdmn_deactivate
8 msfs_unmount
9 cfs_unmount
10 dounmount
11 cfs_do_unmount_accept
12 cms_handle_send_msg
13 cms_kgs_callback_thread
|
There are two types of panics which can occur
during quotactl calls, a panic with the panic string "lock_terminate:
lock held", and a kernel memory fault in dyn_hash_remove.
Partial stack traces for both types follow.
4 panic
5 lock_terminate
6 dqput
7 advfs_get_quota
8 advfs_quotactl
9 cfs_comm_quotactl
10 cfs_quotactl
11 quotactl
12 syscall
13 _Xsyscall
|
4 panic
5 trap
6 _XentMM
7 dyn_hash_remove
8 dqput
9 advfs_get_quota
10 advfs_quotactl
11 cfs_comm_quotactl
12 cfs_quotactl
13 quotactl
14 syscall
15 _Xsyscall |
HP Tru64 UNIX 5.1B-2 ERP Kit
The HP Tru64 UNIX V5.1B-2 ERP Kit contains the
fix for a newly uncovered KMF in access_invalidate problem. The
two typical stack traces for the KMF in access_invalidate are:
0 stop_secondary_cpu
1 panic
2 event_timeout
3 printf
4 panic
5 trap
6 _XentMM
7 access_invalidate
8 bfs_dealloc
9 dmn_dealloc
10 bs_bfdmn_deactivate
11 fs_fset_get_info
12 msfs_real_syscall
13 msfs_syscall
14 syscall
15 _Xsyscall |
0 boot
1 panic
2 trap
3 _XentMM
4 access_invalidate
5 bfs_dealloc
6 dmn_dealloc
7 bs_bfdmn_deactivate
8 msfs_unmount
9 cfs_unmount
10 dounmount
11 cfs_do_unmount_accept
12 cms_handle_send_msg
13 cms_kgs_callback_thread
|
HP Tru64 UNIX 5.1B-3 ERP Kit
The HP Tru64 UNIX V5.1B-3 ERP Kit combines the
fix for a newly uncovered KMF in access_invalidate problem with
fixes previously announced in EA c00640881 (Potential AdvFS
Hangs).
The two typical stack traces for the KMF in access_invalidate
are:
0 stop_secondary_cpu
1 panic
2 event_timeout
3 printf
4 panic
5 trap
6 _XentMM
7 access_invalidate
8 bfs_dealloc
9 dmn_dealloc
10 bs_bfdmn_deactivate
11 fs_fset_get_info
12 msfs_real_syscall
13 msfs_syscall
14 syscall
15 _Xsyscall |
0 boot
1 panic
2 trap
3 _XentMM
4 access_invalidate
5 bfs_dealloc
6 dmn_dealloc
7 bs_bfdmn_deactivate
8 msfs_unmount
9 cfs_unmount
10 dounmount
11 cfs_do_unmount_accept
12 cms_handle_send_msg
13 cms_kgs_callback_thread
|
This first type of hang could be seen by
customers with an AdvFS clone mounted and a file is being
extended using Direct I/O .
A forced crash during this type of hang produces the following
stack trace.
0 thread_block
1 sleep_prim
2 mpsleep
3 insert_actRange_onto_list
4 cow_get_locks
5 bs_cow_pg
6 bs_cow
7 bs_pinpg_clone
8 bs_pinpg_ftx
9 copy_and_del_frag
10 fs_write_add_stg
11 fs_write
12 msfs_write
13 vn_pwrite
14 prwuio
15 pwrite
16 syscall
17 _Xsyscall
This second type of hang could be seen by
customers with an AdvFS clone mounted and a file is being
truncated using Direct I/O .
A forced crash during this type of hang produces the following
stack trace.
0 thread_block
1 sleep_prim
2 mpsleep
3 insert_actRange_onto_list
4 fs_setattr
5 msfs_setattr
6 cfs_comm_setattr
7 cfscall_setattr
8 cfs_setattr
9 vn_open
10 copen
11 syscall
12 _Xsyscall
This third type of hang can be seen by customers
running vfast or defragment on an AdvFS domain, if the system
is in a cluster.
A forced crash during this type of hang may produce either of
the following stack traces.
0 thread_block
1 sleep_prim
2 mpsleep
3 insert_actRange_onto_list
4 msfs_flush_and_invalidate
5 cfs_comm_open
6 cfscall_open
7 cfs_open
8 vn_open
9 copen
10 syscall
11 _Xsyscall |
1 lock_write
2 cfs_condio_cc_excl_mode_enter
3 migrate_get_clu_locks
4 migrate_normal
5 mig_migrate
6 bs_migrate
7 ss_vd_migrate
8 ss_move_file
9 ss_work_thd_pool
|
This fourth type of hang hang can be seen by
customers with an AdvFS clone mounted and at least two
threads doing memory mapping to the same file.
A forced crash during this type of hang may produce either of
the following stack traces
0 thread_block
1 lock_read
2 cow_get_locks
3 bs_cow_pg
4 bs_cow
5 bs_pinpg_clone
6 bs_pinpg_get
7 msfs_getpage
8 u_vp_fault
9 u_map_fault
10 vm_fault
11 trap
12 _XentMM |
0 thread_block
1 lock_write
2 msfs_mmap
3 smmap
4 syscall
5 _Xsyscall
|