PROBLEM:  (74651, 75705) (PATCH ID: OSF445-013)
********

This patch fixes a problem where a root user was not allowed to check
file access on behalf of a user without completely becoming the user.
The functionality is needed by the ASU (Advanced Server for UNIX) product.

PROBLEM:  (MGO90721A) (PATCH ID: OSF445-029)
********


This patch fixes a problem with simple_lock: time limit exceeded panic
seen on SMP systems with the namei.nch_lockinfo lock.  A typical entry
in the /var/adm/messages file would be:

    simple_lock: time limit exceeded

        pc of caller:         0xfffffc00004445d8
        lock address:         0xfffffc0000eb50e8
        lock info addr:       0xfffffc0000772be0
        lock class name:      namei.nch_lockinfo
        current lock state:   0x800000be00443b45
                              (cpu=0,pc=0xfffffc0000443b44,busy)

PROBLEM: (TKTR30001) (PATCH ID: OSF445-037)
********


This patch fixes a race condition in the UBC code
where a lookup is done on a page being invalidated (freed).
The lookup continued to access the data structure after it was
freed, and depending on who uses the data structure next,
this causes the lookup to think that the invalidation did not
complete and wait.

PROBLEM:  (58325) (PATCH ID: OSF445-052)
********

This patch fixes UFS metadata update performance by adding a mount option,
 "delayed", for UFS to convert synchronous metadata writes into
 delayed metadata writes. The File System is often used as a cache for
transitory data such as in web servers. Applications such as Squid web
proxy server will get a significant performance boost by enabling this
option. Restoring from tape is another application that can be benefited
from this.
PROBLEM: (79937) (PATCH ID: OSF445-076)
********
This patch fixes a hang or simple_lock_state_violation panic in biodone.
This patch places mount throttle fields under the new mount_throttle_lock
(locking requirements described in mount.h).  These fields previously were
under the mount lock.  The problem is that it was possible for the mount
lock to be held in a context which could be interrupted by a higher
priority event, such as biodone, which could also require this lock.  The
solution is to protect these fields under a different lock and spl. 

The key pieces of the stack trace for this problem are:
 biodone()              -- wants mount lock
    ...
    ...                 -- interrupt service routines
    ...
 _XentInt()             -- interrupt
    ...
 vfs_unbusy()           -- some routine which takes the mount lock
                           (vfs_unbusy is not unique here)
    ...