verpeaux
ADSM.ORG Member
Hi,
We are using HSM on AIX. Since the last release, il is possible to migrate JFS2 extended attributes V2 filesystems. Since we converted one file system to EA V2, we experience deadlocks in the kernel:
bos.mp64 5.3.7.2
tivoli.tsm.client.hsm.jfs2 5.5.0.1
kdb session:
(0)> dla
Deadlock from tid 60001. This tid waits for the first line lock,
owned by Owner-Id that waits for the next line lock, and so on ...
LOCK NAME | ADDRESS | OWNER-ID | LOCK STATUS | WAITING FUNCTION
? ? ? | 0xF100010030506EE0 | Tid 18202F | 0x20000000 | slock
? ? ? | 0xF100010030506EE0 | Tid 18202F | 0x20000000 | slock
(0)> ttid 18202F
SLOT NAME STATE TID PRI RQ CPUID CL WCHAN
pvthread+018200 386 dsmc SLEEP 18202F 03C 3 0 F100010030506EE0 slist_table+000510
NAME................ dsmc
WTYPE............... WLOCK
(0)> sw 386
Switch to thread: <pvthread+018200>
(0)> f
pvthread+018200 STACK:
[002F7388]slock+0006A4 (00000000002F6FB4, 80000000000090B2 [??])
[00009558].simple_lock+000058 ()
[002666E0]vPut+00006C (??)
[00266884]iUnbind+0000C4 (??, ??)
[0026AED0]iRecycle+0000B8 (??, ??)
[00266F5C]iAlloc+00034C (??, ??)
[0026B440]iGet+0001D4 (??, ??, ??, ??)
[002A796C]eaLookup+0000DC (??, ??, ??, ??, ??)
[002B4364]dmGetEA+000064 (??, ??, ??, ??, ??)
[002B49C0]dmGetAttr+00008C (??, ??)
[002B6578]kvAlloc+000108 (??, ??)
[0026D130]vGet@AF79_20+0000BC (??, ??, ??, ??)
[0026B878]iGet+00060C (??, ??, ??, ??)
[002C49EC]j2_lookup+0001F4 (??, ??, ??, ??, ??, ??)
[04794870]hook_lookup+00009C (F1000100484893F8, F000000030046BB8,
F000000030046BC0, 0000000000000082, 0000000000000000, F10001003063597C)
[0040004C]vnop_lookup+000068 (??, ??, ??, ??, ??, ??)
[003CE19C]lookuppn+000508 (??, ??, ??, ??, ??, ??)
[003CE778]lookupname_cur+0000EC (??, ??, ??, ??, ??, ??, ??)
[003CE7D0]lookupname+00001C (??, ??, ??, ??, ??, ??)
[00440104]statx+000190 (??, ??, ??, ??)
[00003810].svc_instr+000110 ()
[D03516D0]lstat64+000014 (??, ??)
[1013DA40]1013DA40 ()
[1013CD58]1013CD58 ()
[10068954]10068954 ()
[1006C3B4]1006C3B4 ()
[1008D5B4]1008D5B4 ()
[103292BC]103292BC ()
[10329A74]10329A74 ()
[10329A74]10329A74 ()
[10327CF4]10327CF4 ()
[1032E3C4]1032E3C4 ()
[1032D63C]1032D63C ()
[103199EC]103199EC ()
[103194F4]103194F4 ()
[10319E7C]10319E7C ()
[10106A00]10106A00 ()
[10106940]10106940 ()
[D027A770]_pthread_body+000118 (??)
(0)> (0)> ttid 60001
SLOT NAME STATE TID PRI RQ CPUID CL WCHAN
pvthread+006000 96 cron SLEEP 060001 03C 2 0 F1000100303ED160 slist_table+000340
NAME................ cron(0)
WTYPE............... WLOCKREAD
(0)> sw 96
Switch to thread: <pvthread+006000>
(0)> f
pvthread+006000 STACK:
[002F2F9C]complex_lock_sleep_ppc+000168 (00000000D0353064, 000000000000F0B2,
0000000000000000, 000000002FF1E290 [??])
[002F35D4]lock_read_ppc+00028C (??)
[002A1738]j2_access+000040 (??, ??, ??, ??)
[00401C58]vnop_access+000054 (??, ??, ??, ??)
[004F621C]accessx+000228 (??, ??, ??)
[00003810].svc_instr+000110 ()
[D04A2384]tempnam+0000E0 (??, ??)
[10003110]ex+00023C (??)
[10004ED4]main+0005B4 (??, ??)
[10000198]__start+000098 ()
Deadlocks occur in various places, but the common point is that one the locked processes is in dmGetEA.
Pierre Verpeaux
We are using HSM on AIX. Since the last release, il is possible to migrate JFS2 extended attributes V2 filesystems. Since we converted one file system to EA V2, we experience deadlocks in the kernel:
bos.mp64 5.3.7.2
tivoli.tsm.client.hsm.jfs2 5.5.0.1
kdb session:
(0)> dla
Deadlock from tid 60001. This tid waits for the first line lock,
owned by Owner-Id that waits for the next line lock, and so on ...
LOCK NAME | ADDRESS | OWNER-ID | LOCK STATUS | WAITING FUNCTION
? ? ? | 0xF100010030506EE0 | Tid 18202F | 0x20000000 | slock
? ? ? | 0xF100010030506EE0 | Tid 18202F | 0x20000000 | slock
(0)> ttid 18202F
SLOT NAME STATE TID PRI RQ CPUID CL WCHAN
pvthread+018200 386 dsmc SLEEP 18202F 03C 3 0 F100010030506EE0 slist_table+000510
NAME................ dsmc
WTYPE............... WLOCK
(0)> sw 386
Switch to thread: <pvthread+018200>
(0)> f
pvthread+018200 STACK:
[002F7388]slock+0006A4 (00000000002F6FB4, 80000000000090B2 [??])
[00009558].simple_lock+000058 ()
[002666E0]vPut+00006C (??)
[00266884]iUnbind+0000C4 (??, ??)
[0026AED0]iRecycle+0000B8 (??, ??)
[00266F5C]iAlloc+00034C (??, ??)
[0026B440]iGet+0001D4 (??, ??, ??, ??)
[002A796C]eaLookup+0000DC (??, ??, ??, ??, ??)
[002B4364]dmGetEA+000064 (??, ??, ??, ??, ??)
[002B49C0]dmGetAttr+00008C (??, ??)
[002B6578]kvAlloc+000108 (??, ??)
[0026D130]vGet@AF79_20+0000BC (??, ??, ??, ??)
[0026B878]iGet+00060C (??, ??, ??, ??)
[002C49EC]j2_lookup+0001F4 (??, ??, ??, ??, ??, ??)
[04794870]hook_lookup+00009C (F1000100484893F8, F000000030046BB8,
F000000030046BC0, 0000000000000082, 0000000000000000, F10001003063597C)
[0040004C]vnop_lookup+000068 (??, ??, ??, ??, ??, ??)
[003CE19C]lookuppn+000508 (??, ??, ??, ??, ??, ??)
[003CE778]lookupname_cur+0000EC (??, ??, ??, ??, ??, ??, ??)
[003CE7D0]lookupname+00001C (??, ??, ??, ??, ??, ??)
[00440104]statx+000190 (??, ??, ??, ??)
[00003810].svc_instr+000110 ()
[D03516D0]lstat64+000014 (??, ??)
[1013DA40]1013DA40 ()
[1013CD58]1013CD58 ()
[10068954]10068954 ()
[1006C3B4]1006C3B4 ()
[1008D5B4]1008D5B4 ()
[103292BC]103292BC ()
[10329A74]10329A74 ()
[10329A74]10329A74 ()
[10327CF4]10327CF4 ()
[1032E3C4]1032E3C4 ()
[1032D63C]1032D63C ()
[103199EC]103199EC ()
[103194F4]103194F4 ()
[10319E7C]10319E7C ()
[10106A00]10106A00 ()
[10106940]10106940 ()
[D027A770]_pthread_body+000118 (??)
(0)> (0)> ttid 60001
SLOT NAME STATE TID PRI RQ CPUID CL WCHAN
pvthread+006000 96 cron SLEEP 060001 03C 2 0 F1000100303ED160 slist_table+000340
NAME................ cron(0)
WTYPE............... WLOCKREAD
(0)> sw 96
Switch to thread: <pvthread+006000>
(0)> f
pvthread+006000 STACK:
[002F2F9C]complex_lock_sleep_ppc+000168 (00000000D0353064, 000000000000F0B2,
0000000000000000, 000000002FF1E290 [??])
[002F35D4]lock_read_ppc+00028C (??)
[002A1738]j2_access+000040 (??, ??, ??, ??)
[00401C58]vnop_access+000054 (??, ??, ??, ??)
[004F621C]accessx+000228 (??, ??, ??)
[00003810].svc_instr+000110 ()
[D04A2384]tempnam+0000E0 (??, ??)
[10003110]ex+00023C (??)
[10004ED4]main+0005B4 (??, ??)
[10000198]__start+000098 ()
Deadlocks occur in various places, but the common point is that one the locked processes is in dmGetEA.
Pierre Verpeaux