
PROBLEM: (90119) (PATCH ID: TCR520-034)
********
The patch fixes a panic in the distributed lock manager when another
node in the cluster is halted.


PROBLEM: (90540) (PATCH ID: TCR520-074)
********
This patch fixes a problem in the distributed lock manager deadlock
detecton code.


PROBLEM: (92337) (PATCH ID: TCR520-123)
********
This patch fixes a problem in the dlm lock teardown path.
Processes using dlm take upto 10 minutes to exit. Sometimes this 
results in an "ics_unable_to_make_progress: input thread stalled" panic.


PROBLEM: (91358, 91668, 91725, 91990) (PATCH ID: TCR520-122)
********
The first situation that this patch fixes is a crash while using CNX.  
The stack trace will show a race condition between cnx setting dlm_suspended to 
0 via the reconfig pending callback and the last step of the current rebuild 
which resets dlm_suspended to 1.  The machine will print out a panic string 
reading, "panic (cpu 0): dlm_suspended file
 ../../../../src/kernel/dlm/dlm_msdlm_g.c line 3534."
The second situation is one in which the machine would crash with a held lock.
The stack trace will show a thread accessing the dlm_membership table while
the rebuild thread holds the lock to this table.  The panic string for this 
will read, "kcpanic (cpu 0): mbr->node != DLM_INVALID_NODE_ID && mbr->state == 1
file ../../../../src/kernel/dlm/dlm_rebuild.c line 988."

The third situation that this patch fixes is one in which the initial manager of
a resource would panic if a holder of a sublock rebooted.  The stack trace will
show that the thread was attempting to free memory that had just been freed. 
This panic's string will read, "dlm_state == dlm_rs_cleanup && lkbp->lcktype == 
DLM_MASTER && (!(lkbp->qstate & DLM_NOTFQ)) && (!(lkbp->qstate & DLM_TIMEOUTQ)) 
file ../../../../src/kernel/dlm/dlm_rebuild.c line 2158."


PROBLEM: (HPAQ10TKX, 91496, HPAQ10S0M, 91617) (PATCH ID: TCR520-115)
********
This patch corrects two problems in the DLM subsystem.
    - A cluster member panic with the panic string "ns->rootlks > 0 file
      ../../../../src/kernel/dlm/dlm_lock.c line 4359"
    - DLM was not always returning the resource block info for the sublock
      even if the sub lock was held.

PROBLEM: (CH_G07364, 96573) (PATCH ID: TCR520-291)
********
This patch fixes a Distributed Lock Manager panic when calling the 
dlm_get_lkinfo() routine passing an lkid of a lock block that has 
already been declared dead by the deadlock detection thread. The 
panic would occur if the resource the lock originally was associated
with had already been deleted.  Deletion of a resource block can occur
when all the corresponding lock requests have either been unlocked or
declared dead.  The following is a sample trace:
   PANIC: "VALID_RSB(rsbp) file ../../../../src/kernel/dlm/dlm_lock.c line 3404"


   crash> tf
      0 boot                 src/kernel/arch/alpha/machdep.c : 2672
      1 panic                src/kernel/bsd/subr_prf.c : 1401
      2 dlm_panic            src/kernel/dlm/dlm_cfg.c : 684
      3 get_lkinfo           src/kernel/dlm/dlm_lock.c : 3404
      4 dlm                  src/kernel/dlm/dlm_lock.c : 2205
      5 syscall              src/kernel/arch/alpha/syscall_trap.c : 725
      6 _Xsyscall            src/kernel/arch/alpha/locore.s : 1814



