 |
≫ |
|
|
|
パッチ名: PHNE_33498
パッチ摘要: s700_800 11.11 ONC/NFS 一般リリース/性能パッチ
作成日: 05/11/07
公開日: 05/12/22
ハードウェアプラットフォームおよびOSリリース:
s700: 11.11
s800: 11.11
現象:
PHNE_33498:
1.(SR:8606167807 CR:JAGad37090)
ypinit(1m)のマンページで、廃止されたvhe関数が参照されています。
2.(SR:8606406594 CR:JAGaf66500)
ソフトNFSマウントを使ってマウントされたファイルシステムにアクセスしよ
うとすると、"RPC: Unable to receive"というエラーメッセージが表示されま
す。
3.(SR:8606413268 CR:JAGaf73129)
CacheFSのアンマウントに時間がかかります。
4.(SR:8606413731 CR:JAGaf73591)
rfs3call()からエラーが戻されても、ファイルハンドルキャッシュがパージさ
れません。
問題点の説明:
PHNE_33498:
1.(SR:8606167807 CR:JAGad37090)
vhe関数を廃止したときに、vhe関数の参照が一部削除されていませんでした。
解決方法:
vhe関数の参照をすべてypinit(1m)のマンページから削除しました。
2.(SR:8606406594 CR:JAGaf66500)
長さゼロ以外のデータを含む非RPC UDPメッセージがSTREAMS読み取りキューに
挿入されることがありました。そのため、有効なRPCデータを受信する前に、
許容受信再試行回数がこれらのメッセージによって使い果たされていました。
その結果、"ls"のようなコマンドが、ソフトNFSマウントを使ってUDPトランス
ポート上にマウントされたファイルシステムにアクセスできませんでした。
解決方法:
無効なRPC UDPメッセージを受信したら、再試行カウンタに影響を与えないた
めに、それらのメッセージを無視するようにコードを修正しました。
3.(SR:8606413268 CR:JAGaf73129)
CacheFSは直列検索を使って、マウントポイントのvノードによって割り当てら
れたバッファを解放していました。そのため、バッファキャッシュのサイズが
大きい(たとえば、単一の"ll"コマンドを使って大規模なディレクトリをマウ
ントした場合)と、アンマウントに約30秒〜5分以上かかることがありました。
解決方法:
マウントポイントをアンマウントする場合は、新たなメカニズムを使って、
そのマウントポインに関連付けられたバッファを解放するようにコードを修正
しました。この修正を有効にするには、PHKL_33367もインストールする必要が
あります。
4.(SR:8606413731 CR:JAGaf73591)
rfs3call()からエラーが戻されると、VFSはNFSに対してファイルハンドル
キャッシュをパージするよう要求します。ところが、キャッシュがパージされ
なかったため、ハングが起きていました。
解決方法:
rfs3call()からエラーが戻されたらファイルハンドルキャッシュをパージする
ようにコードを修正しました。
-----------------------------------------------------------------------------
Patch Name: PHNE_33498
Patch Description: s700_800 11.11 ONC/NFS General Release/Performance Patch
Creation Date: 05/11/07
Post Date: 05/12/22
Hardware Platforms - OS Releases:
s700: 11.11
s800: 11.11
Products: N/A
Filesets:
NFS.KEY-CORE,fr=B.11.11,fa=HP-UX_B.11.11_32/64,v=HP
NFS.NFS-64ALIB,fr=B.11.11,fa=HP-UX_B.11.11_32/64,v=HP
NFS.NFS-64SLIB,fr=B.11.11,fa=HP-UX_B.11.11_32/64,v=HP
NFS.NFS-CLIENT,fr=B.11.11,fa=HP-UX_B.11.11_32/64,v=HP
NFS.NFS-CORE,fr=B.11.11,fa=HP-UX_B.11.11_32/64,v=HP
NFS.NFS-ENG-A-MAN,fr=B.11.11,fa=HP-UX_B.11.11_32/64,v=HP
NFS.NFS-PRG,fr=B.11.11,fa=HP-UX_B.11.11_32/64,v=HP
NFS.NFS-SERVER,fr=B.11.11,fa=HP-UX_B.11.11_32/64,v=HP
NFS.NFS-SHLIBS,fr=B.11.11,fa=HP-UX_B.11.11_32/64,v=HP
NFS.NIS-CLIENT,fr=B.11.11,fa=HP-UX_B.11.11_32/64,v=HP
NFS.NIS-CORE,fr=B.11.11,fa=HP-UX_B.11.11_32/64,v=HP
NFS.NIS-SERVER,fr=B.11.11,fa=HP-UX_B.11.11_32/64,v=HP
NFS.NISPLUS-CORE,fr=B.11.11,fa=HP-UX_B.11.11_32/64,v=HP
OS-Core.CORE-ENG-A-MAN,fr=B.11.11,fa=HP-UX_B.11.11_32/64,v=HP
NFS.NFS-KRN,fr=B.11.11,fa=HP-UX_B.11.11_32,v=HP
NFS.NFS-KRN,fr=B.11.11,fa=HP-UX_B.11.11_64,v=HP
Automatic Reboot?: Yes
Status: General Release
Critical:
No (superseded patches were critical)
PHNE_33315: HANG ABORT MEMORY_LEAK
PHNE_32811: CORRUPTION ABORT
PHNE_32477: ABORT HANG PANIC MEMORY_LEAK
PHNE_31097: HANG ABORT PANIC
PHNE_30661: ABORT HANG PANIC
PHNE_30380: ABORT
PHNE_30378: HANG ABORT
PHNE_29883: ABORT PANIC
PHNE_29303: PANIC HANG
PHNE_28983: ABORT PANIC MEMORY_LEAK HANG
PHNE_28137: ABORT PANIC CORRUPTION MEMORY_LEAK HANG
PHNE_28103: HANG
PHNE_27218: PANIC MEMORY_LEAK CORRUPTION HANG
PHNE_26388: PANIC
PHNE_25627: ABORT PANIC MEMORY_LEAK HANG
PHNE_25625: ABORT
PHNE_24910: ABORT PANIC CORRUPTION HANG
PHNE_24035: ABORT CORRUPTION HANG
PHNE_23502: ABORT PANIC HANG
PHNE_22878: PANIC HANG
Category Tags:
defect_repair enhancement general_release critical panic
halts_system corruption memory_leak manual_dependencies
Path Name: /hp-ux_patches/s700_800/11.X/PHNE_33498
Symptoms:
PHNE_33498:
( SR:8606167807 CR:JAGad37090 )
The ypinit(1m) man page contains references to obsolete vhe
functions.
( SR:8606406594 CR:JAGaf66500 )
Attempts to access a file system mounted with a soft NFS
mount fail with the error message "RPC: Unable to receive".
( SR:8606413268 CR:JAGaf73129 )
CacheFS unmounts are slow.
( SR:8606413731 CR:JAGaf73591 )
The file handle cache is not purged when an error is
returned from rfs3call().
PHNE_33315:
( SR:8606385895 CR:JAGaf46049 )
NFS write performance slows when writing files larger than
2 GB.
( SR:8606394445 CR:JAGaf54468 )
When an AutoFS file system is unmounted, and the IP Address
of the server that was AutoFS mounted is changed, AutoFS
uses the old IP address if the file system is remounted.
( SR:8606399902 CR:JAGaf59858 )
RPC applications consume a lot of CPU resources.
( SR:8606400085 CR:JAGaf60041 )
The man page describing the rpc_broadcast_exp function has
one of the parameters listed out of order.
( SR:8606401738 CR:JAGaf61675 )
The rpc.lockd and rpc.statd processes hang if logging is
enabled and disabled many times.
( SR:8606402519 CR:JAGaf62454 )
The rpc.lockd and rpc.statd processes dump core when trying
to log an I/O error with logging disabled.
( SR:8606405252 CR:JAGaf65172 )
The rpc.rquotad process increases in size and may dump core.
PHNE_32811:
( SR:8606345889 CR:JAGaf06735 )
The server's rpc.lockd grants reclaim requests to clients
after the grace period expires.
( SR:8606377207 CR:JAGaf37485 )
rpc.lockd terminates when it receives a SIGALRM signal. If
rpc.lockd logging is enabled, the message "signal handler:
signal = 14" appears in the /var/adm/rpc.lockd.log file.
( SR:8606389345 CR:JAGaf49493 )
The mount_nfs(1M) man page states that the default value of
the 'retrans' option is 5. However, if an NFS client sends a
request with no reply from the server, only 4
retransmissions are attempted before giving up in the case
of a soft mount. The documentation of the default retrans
value is also inconsistent.
( SR:8606393839 CR:JAGaf53894 )
The rename() system call returns a "Not Owner" error when
renaming a directory in a cachefs-mounted file system to a
name which is an existing directory.
( SR:8606394223 CR:JAGaf54255 )
The "ypwhich -m" command returns the error message:
"ypwhich: can't get the list of maps: RPC: Unable to
receive; buffer not large enough".
PHNE_32477:
( SR:8606386532 CR:JAGaf46686 )
An application using clnt_tli_create() dumps core when
called with large send/receive buffers for connection-
oriented transports.
( SR:8606387765 CR:JAGaf47918 )
If a write() system call to an NFS file is interrupted by a
signal (e.g. by sending SIGINT to the process) before the
data is written to the server, the call succeeds. This
occurs only on files configured to do synchronous I/O
(opened with the O_DSYNC flag set).
( SR:8606387807 CR:JAGaf47960 )
An NFS server panics with the panic string "kalloc: out of
kernel virtual space". The panic stack trace is:
panic+0x6c
kalloc_4K+0x270
kalloc_from_superpage+0x488
kalloc+0x14
alloc_mem+0x44
get_kmem+0x8c
kmem_arena_xlarge_alloc+0x74
kmalloc+0x1e4
nfs_kmem_alloc+0x18
rfs3_read+0x310
rfsexp_dispatch+0x5f0
svc_getreq+0x1ec
svc_run+0x6e8
nfsexp_svc+0x1f8
nfs_stub_svc+0xa4
coerce_scall_args+0xe0
syscall+0x204
syscallinit+0x55c
( SR:8606389142 CR:JAGaf49290 )
rpc.lockd dumps core with SIGBUS.
( SR:8606389255 CR:JAGaf49403 )
A multi-threaded process hangs and cannot be killed. The
system needs to be rebooted to clear the situation.
( SR:8606348790 CR:JAGaf09611 )
Difficulties in correctly configuring automounts with direct
maps using the automount(1m) man page.
( SR:8606381800 CR:JAGaf42012 )
There is a fix for an AutoFS 1.2 defect which is not needed
if the Enhanced AutoFS product is installed.
( SR:8606393701 CR:JAGaf53756 )
AutoFS mounts and unmounts fail with the "clnt_dg_create:
out of memory" message in the syslog.log file.
PHNE_31929:
( SR:8606373075 CR:JAGaf33480 )
The mlock call is succeeding after truncation of a memory
mapped file over NFS.
( SR:8606375121 CR:JAGaf35426 )
The on command fails with the error message "Unable to
mount" when executed from an AutoFS mount point.
( SR:8606379022 CR:JAGaf39271 )
Provide the tunable support to adjust the delay time
variable in response to an NFS3ERR_JUKEBOX error from an NFS
server.
PHNE_31097:
( SR:8606345966 CR:JAGaf06811 )
The message "/net -hosts -nosuid,soft" appears in the
/etc/rc.log file and on stderr.
( SR:8606347603 CR:JAGaf08425 )
Enhancement: RPC applications using UDP transport can now
identify the end-point pair.
( SR:8606354067 CR:JAGaf14861 )
Previously running UDP nfsd's are unregistered after nfsd's
are restarted.
( SR:8606359725 CR:JAGaf20421 )
The automountd process dumps core.
( SR:8606360695 CR:JAGaf21402 )
An NFS TCP client hangs even though its TCP connection to
the server appears to be good, or panics reporting "kalloc:
out of kernel virtual space" as the panic string.
( SR:8606364296 CR:JAGaf24937 )
The function netdir_getbyname() returns the message
"successful completion" when it should return the error
"service name not found".
( SR:8606364383 CR:JAGaf25022 )
If an NFS server is not available, processes accessing NFS
filesystems hang and cannot be killed.
( SR:8606365866 CR:JAGaf26496 )
An NFS client crashes in the read path when an NFS server
returns more data than requested.
( SR:8606369177 CR:JAGaf29721 )
rpcbind(1M) core dumps with SIGBUS or SIGSEGV.
( SR:8606366266 CR:JAGaf26870 )
The same lock is granted to two blocking clients, when one
of the locks is obtained by a loopback mount (LOFS).
( SR:8606379191 CR:JAGaf39440 )
Support for largefiles with rpc.rquotad.
( SR:8606379708 CR:JAGaf39957 )
When a "tcp_discon" is issued via ndd against the local TCP
end-point for NFS/TCP client requests to a specific NFS
server, the NFS/TCP client will attempt to re-connect to the
NFS server using the same local port.
PHNE_30661:
( SR:8606354850 CR:JAGaf15606 )
NIS slave server configuration fails with the error message:
"ypxfr: can't dbm close temp map".
( SR:8606355618 CR:JAGaf16332 )
A client's rpc.lockd dumps core after a server failover.
( SR:8606357066 CR:JAGaf17766 )
The exportfs command fails with the error message "exportfs
error: /etc/xtab in use. Please try again." when /etc/xtab
is not in use.
( SR:8606357361 CR:JAGaf18061 )
The CacheFS mount command dumps core.
( SR:8606358606 CR:JAGaf19305 )
A connectionless NFS server hangs with the following
messages in syslog.log:
xdr_bytes: bad size FAILED
NOTICE: nfs_server: bad sendreply; proc=6
( SR:8606363276 CR:JAGaf23938 )
There is no documentation for the nfs_portmon variable.
( SR:8606329363 CR:JAGae90487 )
The system panics with the panic string "Data page fault".
The panic stack trace is:
bwrite+0xd4
rw3vp+0x8d8
nfs3_write+0x184
nfs3_rdwr+0x50
vno_rw+0x80
write+0x108
syscall+0x750
syscallinit+0x554
PHNE_30380:
( SR:8606382150 CR:JAGaf42361 )
An RPC application using TCP dumps core with SIGBUS when
started with inetd.
PHNE_30378:
( SR:8606319551 CR:JAGae82039 )
The rpc_soc man page is missing a description about using
sockets in an application.
( SR:8606339927 CR:JAGaf00848 )
The automount daemon hangs during startup and the
/etc/mnttab file contains inconsistent entries.
( SR:8606344107 CR:JAGaf04959 )
A locking application hangs and the rpc.lockd daemon aborts.
If logging is enabled for rpc.lockd, the message "ABORT in
add_reclock" is logged in the /var/adm/rpc.lockd.log file.
( SR:8606345310 CR:JAGaf06160 )
When there are multiple subdomains configured, the
rpc.nispasswdd daemon may fail to start and the nisstat
command may return with the following error message:
"nisstat: Error talking to host <NIS+ master>: Error in RPC
subsystem".
( SR:8606346094 CR:JAGaf06921 )
The setoncenv command accepts NFS_TCP as a valid
configurable environment variable.
( SR:8606346132 CR:JAGaf06959 )
Incorrect NFS version information is being passed to the
kernel profiling code.
( SR:8606347226 CR:JAGaf08050 )
Commands operating on an NFS file system mounted as a soft
mount over UDP transport protocol fail with the error
message: "RPC: Unable to receive".
( SR:8606347264 CR:JAGaf08087 )
The error message "No such file or directory" results from
the /usr/bin/pwd command with AutoFS direct maps.
( SR:8606351701 CR:JAGaf12506 )
The mount_nfs man page states that quota checking on NFS
file systems is enforced by default.
( SR:8606355273 CR:JAGaf16027 )
Netselect and Netdir output some error messages incorrectly.
( SR:8606356288 CR:JAGaf16994 )
A hang occurs during an NFS write operation.
PHNE_30086:
( SR:8606324736 CR:JAGae87042 )
An NFS client operating on filesystems mounted over TCP from
an NFS server is unable to contact the server and fails with
repeated messages such as "NFS server <xxx> not responding
still trying" or "NFS getattr failed for server <xxx> : RPC:
(unknown error code)". The client's TCP connection state
shows that the connection is in the CLOSE_WAIT state. The
connection persists in this state until you reboot the
client or reset the connection with ndd.
( SR:8606336931 CR:JAGae97978 )
An NFS-TCP client running processes over an NFS-TCP hard
mount fails to reconnect with the NFS server after a network
outage has been restored. The client fails with repeated
messages such as "NFS server <xxx> not responding still
trying" or "NFS getattr failed for server <xxx> : RPC:
(unknown error code)". The client's TCP connection state
shows that an established connection to the server exists.
This failure persists for about six minutes until the
server times out and closes the existing TCP connection.
PHNE_29883:
( SR:8606312940 CR:JAGae75747 )
An NFS PV3 client panics while reading the path name of a
long symbolic link.
( SR:8606320583 CR:JAGae83065 )
The system panics while AutoFS is unmounting the file
systems that are being managed by AutoFS.
( SR:8606330100 CR:JAGae91223 )
The error message "rpcbind cannot accept connection: bad
call sequence number (current state 4)" appears in the
syslog.log file.
( SR:8606333242 CR:JAGae94337 )
There is a problem with nested AutoFS maps.
( SR:8606334605 CR:JAGae95685 )
A core dump occurs when an application uses 64-bit
xdr_char() and xdr_u_char() routines.
( SR:8606336578 CR:JAGae97630 )
It is not clear how to choose the number of nfsd's to be run
when configuring an HP-UX NFS server.
PHNE_29783:
( SR:8606347801 CR:JAGaf08623 )
Problem with rpc.ypupdated.
PHNE_29303:
( SR:8606286517 CR:JAGae50456 )
AutoFS mount points do not get mounted under heavy load.
This product update is a member of a set needed to enable
the optional HP-UX Enhanced AutoFS feature. Upon
installation, the HP-UX Enhanced AutoFS bundle (Enhanced
AutoFS) will install the full set of product updates
(including this one) to enable the HP-UX Enhanced AutoFS
feature. If the HP-UX Enhanced AutoFS product is not
installed, this product update will have no impact on your
system.
( SR:8606302457 CR:JAGae65816 )
Applications that issue lock requests from an HP-UX client
to an EMC Server sometimes hang.
( SR:8606304146 CR:JAGae67490 )
In an NIS+ trusted environment, if a non-root user executes
a password change on an NIS+ client, logins to other hosts
will fail because the pwchanger field of the trusted table
is not updated.
( SR:8606306059 CR:JAGae69107 )
A non-root NIS+ user will hang when issuing a rexec command
to another NIS+ client.
( SR:8606306319 CR:JAGae69367 )
The NIS+ client authentication fails when the /var
filesystem is full.
( SR:8606312085 CR:JAGae74904 )
When an NFS client tries to create a symbolic link to a file
or directory with a large path name, the NFS server panics
with a data page fault.
( SR:8606313428 CR:JAGae76226 )
Unexpected RPC error messages such as "Program not
registered", "Miscellaneous tli error", and "Program
unavailable" result from the showmount and other commands.
( SR:8606322456 CR:JAGae84924 )
The NFS mount command mounts with NFS V2 instead of failing
when NFS V3 is requested and is not available on the server.
( SR:8606335982 CR:JAGae97035 )
This product update is a member of a set needed to enable
the optional HP-UX Enhanced AutoFS feature. Upon
installation, the HP-UX Enhanced AutoFS bundle (Enhanced
AutoFS) will install the full set of product updates
(including this one) to enable the HP-UX Enhanced AutoFS
feature. If the HP-UX Enhanced AutoFS product is not
installed, this product update will have no impact on your
system.
( SR:8606332784 CR:JAGae93899 )
The system panics with the panic string "Data page fault".
The panic stack trace is:
rtablehash+0x18
rfind+0x18
make_rnode+0x24
makenfs3node+0x34
nfs3readdirplus+0x280
do_nfs3readdir+0x68
nfs3_readdir2+0xc0
getdents+0x1b0
syscall+0x28c
syscallinit+0x54c
PHNE_29211:
( SR:8606317529 CR:JAGae80090 )
The performance level of connection-oriented RPC transports
is being reduced by network delays.
( SR:8606320560 CR:JAGae83042 )
The rpc.mountd code returns an incorrect error message.
( SR:8606317535 CR:JAGae80096 )
The performance level of connection-oriented RPC servers is
being reduced by network delays.
( SR:8606337943 CR:JAGae98946 )
NIS+ commands and applications fail with the error message
"Error in RPC subsystem".
PHNE_28983:
( SR:8606220736 CR:JAGad89872 )
When a new NIS+ user changes the password to log in for the
first time, the login fails with the error message "Password
information update failed".
( SR:8606278781 CR:JAGae42837 )
AutoFS mounted directories do not get unmounted, even if
not accessed during the timeout period. These mounts are
unmounted only after a reboot.
( SR:8606294191 CR:JAGae57931 )
Directories that are mounted with AutoFS are not getting
unmounted by AutoFS.
( SR:8606294387 CR:JAGae58084 )
Under heavy load, AutoFS-mounted file systems never get
unmounted.
( SR:8606294761 CR:JAGae58456 )
A large /etc/exports file causes the system to boot very
slowly.
( SR:8606295723 CR:JAGae59388 )
Unable to mount an NFS file system from a PC.
( SR:8606296110 CR:JAGae59732 )
The defaults listed in the /etc/nsswitch.hp_defaults file do
not match the nsswitch.conf man page.
( SR:8606296698 CR:JAGae60267 )
An NFS Server hangs and the error message "NFS Server not
responding" is seen when accessing NFS loopback mounts.
( SR:8606296789 CR:JAGae60355 )
A 64-bit system panics when executing an "ls" command on an
NFS PV3 mounted file system.
( SR:8606297163 CR:JAGae60682 )
When open() is called with O_EXCL via NFS, the date,
timestamp, and privileges of the file are not updated
properly.
( SR:8606297688 CR:JAGae61190 )
AutoFS directories are not getting remounted after a manual
unmount.
( SR:8606297816 CR:JAGae61318 )
The data segment of the rpc.lockd process grows larger with
time, and response time increases. The process may
ultimately crash because of a lack of memory.
( SR:8606300458 CR:JAGae63927 )
Poor performance while remounting an AutoFS hierarchical
mount.
( SR:8606301831 CR:JAGae65194 )
The /etc/services file is cleared when the machine is booted
with /usr as a dead mount point.
( SR:8606302828 CR:JAGae66186 )
Stale AutoFS directories remain when a mount attempt to a
non-existent file system fails.
( SR:8606303150 CR:JAGae66506 )
An NFS Server hangs and the error message "NFS Server not
responding" is seen when accessing NFS loopback mounts.
( SR:8606308183 CR:JAGae71218 )
The nfsd.1m and setoncenv.1m man pages contain references to
NFS_TCP which is not relevant in this HP-UX release.
( SR:8606308342 CR:JAGae71377 )
A non-root user is not able to unmount an NFS mounted
directory.
( SR:8606309647 CR:JAGae72522 )
NIS+ man pages do not mention that the HP-UX 11.23 release
will be the last release on which NIS+ is supported.
PHNE_28930:
( SR:8606306693 CR:JAGae69727 )
Some of the XDR routines do not function properly.
PHNE_28568:
( SR:8606308270 CR:JAGae71305 )
The "nfsstat -m" command does not display NFS PV3
information.
PHNE_28137:
( SR:8606284507 CR:JAGae48453 )
NFS client calls close() and write() timeout on hard NFS
mounts.
( SR:8606279504 CR:JAGae43557 )
The result of executing a binary linked with a shared
library from a CacheFS mount point differs from the result
of executing the same binary from a native file system.
( SR:8606274703 CR:JAGae38780 )
NFS pathconf(_PC_CHOWN_RESTRICTED) and fpathconf
(_PC_CHOWN_RESTRICTED) return 0 when chown is not restricted
instead of -1, the POSIX compliant result.
( SR:8606272309 CR:JAGae36449 )
Using loopback NFS mounts, the server may hang due to a
deadlock condition.
( SR:8606284964 CR:JAGae48906 )
Data corruption while using XDR function xdrmblk_getlong()
with a 64-bit kernel.
( SR:8606275551 CR:JAGae39627 )
If an NFS server listed in the /etc/mnttab file is down or
there are network problems, the "nfsstat -m" call hangs.
( SR:8606284885 CR:JAGae48827 )
The number of file locks (nflocks) grows over a long period
of time, eventually causing NFS to lock up. The only
workaround is to restart NFS.
( SR:8606274519 CR:JAGae38596 )
After unmounting all the AutoFS mount points through the
script "nfs.client stop", the system panics if it tries to
access an unmounted directory.
( SR:8606256016 CR:JAGae20331 )
Threads accessing an NFS directory hang for a long time with
"bindresvport: loop-twice port assigned ..." messages other
than port 512 logged in syslog.log.
( SR:8606271824 CR:JAGae36003 )
The system reports "vmunix: bindresvport: loop_twice port
assigned 512" messages.
( SR:8606269865 CR:JAGae34105 )
When Intergraph's DiskShare NFS server returns zero bytes
for the preferred and/or maximum transfer size of the
READ/WRITE operation, the HP-UX NFS client panics.
( SR:8606267227 CR:JAGae31470 )
The system panics after a read/write NFS exported file
system is unmounted and then re-mounted as read-only without
re-exporting the file system.
( SR:8606276756 CR:JAGae40831 )
An NFS server sometimes panics if pathconf() is called with
a file in an NFS mounted directory.
( SR:8606267508 CR:JAGae31750 )
A process is able to write into a read-only automounted file
system.
( SR:8606267320 CR:JAGae31562 )
Poor NFS PV3 performance on a 64-bit client system.
( SR:8606287136 CR:JAGae51075 )
The contents of a directory mounted with AutoFS become
unavailable, and the directory cannot be used, mounted, or
unmounted.
( SR:8606278982 CR:JAGae43038 )
An AutoFS mounted file system using the "-hosts" map can
take a long time to mount when the environment contains dead
NFS servers.
( SR:8606288334 CR:JAGae52265 )
The NFS mount command does not update the /etc/mnttab file
with protocol information when the UDP protocol is used.
( SR:8606282498 CR:JAGae46451 )
Memory/file-descriptor leak in RPC multi-threaded server
programs, if the client tries to connect using a version
number that is unsupported by the server. The memory leak
may result in the server dumping core and terminating. This
can also cause the server to quickly run out of file
descriptors.
( SR:8606288821 CR:JAGae52752 )
Calls to netdir_getbyname() from a multithreaded process may
hang or terminate with SIGSEGV or SIGBUS.
( SR:8606282662 CR:JAGae46615 )
The message "NFS server xxxx not responding still trying" is
seen on a client while attempting to establish an NFS TCP
connection, and the connection is in the LISTEN state.
( SR:8606291642 CR:JAGae55409 )
Memory based XDR stream implementations do not function
properly.
PHNE_28103:
( SR:8606276016 CR:JAGae40093 )
The rpc.yppasswdd process dies.
( SR:8606274502 CR:JAGae38579 )
RPC does not function properly.
( SR:8606288394 CR:JAGae52325 )
Incorrect permissions for the log file.
( SR:8606281883 CR:JAGae45826 )
The netdir_getbyname() routine using NIS can hang under
certain circumstances.
( SR:8606276334 CR:JAGae40412 )
Problem with ypxfrd.
( SR:8606286563 CR:JAGae50502 )
Incorrect configuration for AutoFS.
( SR:8606274555 CR:JAGae38632 )
Showmount dumps the core with SIGSEGV.
PHNE_27218:
( SR:8606249576 CR:JAGae15966 )
A client's attempt to acquire a lock may result in a delay
or a hang.
( SR:8606249559 CR:JAGae15949 )
The '-p' option for rpc.mountd described in the man page
does not function as described.
( SR:8606180770 CR:JAGad49991 )
Blocked lock requests on a client system hang.
( SR:8606233931 CR:JAGae03154 )
After the AutoFS client unmount timer expires, if the
unmount of the parent file system fails and the remount of
hierarchical mount points fails, then zero entries are found
while accessing the hierarchical mount points under the
parent file system. Once this condition occurs the failed
hierarchical mounts will not retry mounting.
( SR:8606234034 CR:JAGae03255 )
The nfsconfcol script is not removing the target data files
before writing to them.
( SR:8606256845 CR:JAGae21158 )
The /var/adm/automount.log file is shipped incorrectly.
( SR:8606256846 CR:JAGae21159 )
The /var/spool/rexd directory is shipped incorrectly.
( SR:8606259831 CR:JAGae24149 )
A client system hangs when trying to lock a file.
( SR:8606255006 CR:JAGae19335 )
Exported directories cannot be mounted.
( SR:8606277106 CR:JAGae41178 )
Under heavy load, file systems mounted with AutoFS never get
unmounted, even after the system load has decreased.
( SR:8606257688 CR:JAGae21991 )
Client hangs when trying to lock a remote file.
( SR:8606260873 CR:JAGae25195 )
Large files become corrupt when writing across NFS mounts.
( SR:8606242021 CR:JAGae09276 )
1. Many "dup_remove_pending: WARNING: duplicate request
cache size (210) > XID_CACHE_SIZE (200), continuing"
messages in syslog.log and automount.log.
2. Many "Rmdir: err=247, mnt='<mount-point-dir>'" messages
in automount.log.
( SR:8606226778 CR:JAGad95840 )
If an AutoFS mounted filesystem using a simple direct map is
manually unmounted, a panic may result if that filesystem is
referenced simultaneously (e.g. by two or more ls commands).
The panic string is:
"Data page fault".
The panic stack trace is:
panic+0x14
report_trap_or_int_and_panic+0x80
trap+0xdb8
nokgdb+0x8
nfs3_root+0x68
hpauto_trigger_mount+0x20c
auto_access+0x30
chdirec_seg+0x5c
chdirec+0x14
chdir+0x34
syscall+0x29c
$syscallrtn+0x0
( SR:8606248269 CR:JAGae14669 )
If a client's /var/statmon/sm directory contents are
destroyed, future lock requests hang.
( SR:8606245329 CR:JAGae11799 )
A system hangs or panics when issuing a "ln" command.
( SR:8606264317 CR:JAGae28647 )
This product update is a member of a set needed to enable
the optional HP-UX Device IDs feature. Upon installation,
the HP-UX Device IDs bundle (DevIDs11i) will install the
full set of product updates (including this one) necessary
to enable the Device IDs feature.
If the HP-UX Device IDs product (DeviceIDs11i) is not
installed, this product update to HP-UX will have no
impact on your system.
( SR:8606265228 CR:JAGae29556 )
Spinlock deadlock panic due to a lock being held too long.
( SR:8606259713 CR:JAGae24031 )
AutoFS problem.
( SR:8606234968 CR:JAGae04152 )
When automountd gets a stale NFS file handle error for a
direct mount point, it removes the corresponding AutoFS
direct map entry for that mount point from the file
/etc/mnttab. Automountd needs to be stopped and restarted to
get back the direct AutoFS mount point.
( SR:8606235172 CR:JAGae04352 )
The NFS mount operation returns an incorrect error message
when all ports are in use. The error message is:
nfs mount: bindresvport: t_errno = 8
nfs mount: get_fh: Couldn't bind to a reserved port.
return = -1
( SR:8606245876 CR:JAGae12335 )
System panics when the mount of an exported file system that
contains an extraneous or missing "/" fails. The panic
string is: rm_autonode: No entry for <address>
( SR:8606257219 CR:JAGae21525 )
Accessing a TCP AutoFS mount point causes the system to
hang.
( SR:8606244127 CR:JAGae10618 )
If rpc.lockd and rpc.statd are restarted and the
/var/statmon/sm directory is destroyed on the server, the
client's rpc.lockd takes 35 minutes to recover.
( SR:8606264087 CR:JAGae28412 )
When a 64-bit RPC application uses UNIX authentication
flavor, the RPC server does not properly decode the client
credentials.
( SR:8606271227 CR:JAGae35462 )
The "ls" command will hang when users try to access a dead
server after some systems with mounted directories are shut
down.
( SR:8606147221 CR:JAGad16564 )
Problem with privileged groups.
( SR:8606248870 CR:JAGae15267 )
The "cd" command to an AutoFS direct offset mount directory
hangs, and automountd cannot be killed by pressing CTRL-C to
stop the "cd" command.
( SR:8606232532 CR:JAGae01767 )
AutoFS mounted directories are empty.
( SR:8606254087 CR:JAGae18419 )
The rpcbind process increases in size and stops responding
to client requests. This requires rpcbind needs to be killed
and restarted.
PHNE_26388:
( SR:8606248624 CR:JAGae15022 )
AutoFS causes a system panic with the panic string
"rm_autonode: No entry for <<address>>".
PHNE_25627:
( SR:8606230660 CR:JAGad99710 )
A server stub generated by rpcgen logs error messages to
/dev/console.
( SR:8606188628 CR:JAGad57842 )
The man page for nsswitch.conf does not reflect ldap as an
additional database source for nss backends.
( SR:8606228306 CR:JAGad97363 )
A file has an unusual timestamp when created with open(2)
using O_EXCL flag.
( SR:8606224657 CR:JAGad93745 )
File lock recovery fails when the server reboots if the
server has multiple network interfaces and the client has
mounted the file system with an interface name different
than the server's hostname.
( SR:8606218377 CR:JAGad87526 )
A multi-threaded server leaks file descriptors under heavy
load if clients are running on the same box.
( SR:8606186978 CR:JAGad56188 )
Partially cached large binaries are removed at unmount time
by CacheFS.
( SR:8606225370 CR:JAGad94457 )
The system panics due to a data page fault on the NFS server
when processing a TCP client mount request.
( SR:8606230685 CR:JAGad99735 )
Excessive warning messages being logged:
WARNING: filegrp_destroy_metadata:
fileno 61 slot 7-5 fgp 41f2c1d0 not allocated.
( SR:8606219527 CR:JAGad88667 )
On a client system, the old automounter is trying to access
a non-existent entry in the NIS map and that is generating
an excessive number of packets over the network which causes
slow response time.
( SR:8606226735 CR:JAGad95797 )
The inetd is unable to resolve services.
( SR:8606237277 CR:JAGae06322 )
NFS PV2 client does not properly show the quota exceeded
error.
( SR:8606221217 CR:JAGad90351 )
The "ll" command hangs in the case of an automounted file
system if the server is down.
( SR:8606221909 CR:JAGad91027 )
System panic occurs while autofs is unmounting the idle
autofs mount points. The panic string is "spinlocks held
with M_WAITOK".
( SR:8606221102 CR:JAGad90236 )
Reactivating a user account that has been disabled on a
trusted NIS+ clients fails.
( SR:8606172908 CR:JAGad42168 )
When an HANFS package is manually failed over back-and-forth
between two different cluster server nodes within a period
of time shorter than five minutes, then on any NFS client
which have the HANFS package filesystem mounted over TCP,
NFS access to that filesystem can appear hung for a
relatively long period of time - usually more than four
minutes, with no apparent response from the server. Also
the NFS client issues the error message "NFS server XXXXXXX
not responding still trying", about every 30 seconds.
( SR:8606226170 CR:JAGad95241 )
A "umount" of an NFS mounted file system using CacheFS
causes an RPC error.
( SR:8606229377 CR:JAGad98430 )
AutoFS mounted file systems never get unmounted.
( SR:8606204999 CR:JAGad74175 )
The ttsyncd process loops, occupying much of the CPU usage
and causing the system to slow down.
( SR:8606225120 CR:JAGad94209 )
rpc.rusersd core dumps if more than 180 users have logged
in.
PHNE_25625:
( SR:8606245747 CR:JAGae12206 )
rpc.lockd core dumps with a SIGBUS error during heavy NFS
file locking.
PHNE_24910:
( SR:8606184366 CR:JAGad53575 )
Instruction page fault where the instruction address being
accessed is 0.
( SR:8606201620 CR:JAGad70794 )
On SUN's NIS clients, it is possible to specify a list of
available NIS servers for a given domain. HP's NIS clients
do not provide this functionality.
( SR:8606147778 CR:JAGad17120 )
RPC client call rejection from an RPC server because of
unsupported RPC protocol version can result in an error
incorrectly indicating a non-response from the RPC server.
For example, an NFS operation results in "NFS server
<server name> not responding still trying" error.
( SR:8606216548 CR:JAGad85717 )
A password expires unexpectedly in an NIS+ environment.
( SR:8606205779 CR:JAGad74954 )
The automountd (AUTOFS) hangs/cores while attempting to
automount exported FS and reports 100% CPU utilization.
( SR:8606205013 CR:JAGad74189 )
The mount option 'grpid' does not return the correct 'gid'
value on NFS PV2 and PV3.
( SR:8606212242 CR:JAGad81430 )
When a customer specifies a user name with leading blank
characters as an argument to nispasswd and passwd(1),
nispasswd modifies the username to include those leading
blanks in the NIS+ passwd table and trusted table.
( SR:8606187776 CR:JAGad56985 )
An NFS/TCP client panics while under heavy NFS load. Problem
has been seen most often while client performed 100 NFS/TCP
mounts in succession.
( SR:8606220062 CR:JAGad89203 )
The ovcmsgrd process aborts and produces a core file when
the Network Node Manager managed processes are started with
"ovstart opc".
( SR:8606219558 CR:JAGad88698 )
During the login process, NIS+/trusted clients force NIS+
users to modify their password when their password is
expired but fail to un-expire the password when their
password is changed. Users will always be forced to modify
their password as soon as it has expired.
( SR:8606213087 CR:JAGad82274 )
NFS PV3 allows the file to be opened with mandatory lock
(02760) permission, while NFS only supports advisory lock.
( SR:8606199444 CR:JAGad68631 )
The yppasswd command updates fields other than the user
password field in an NIS map.
( SR:8606198866 CR:JAGad68055 )
Can not cache large files (> 2GB) with cachefs mount.
( SR:8606169485 CR:JAGad38760 )
NFS/TCP client does not handle out-of-order streams messages
correctly while attempting to connect to the server and can
result in a client panic.
PHNE_24035:
( SR:8606196567 CR:JAGad65770 )
AutoFS cannot unmount a cachefs type file system.
( SR:8606208446 CR:JAGad77633 )
NFS file lock recovery fails because rpc.statd matches the
wrong hostname.
( SR:8606207461 CR:JAGad76637 )
rpcbind crashes with signal 10, Bus error (SIGBUS).
( SR:8606164133 CR:JAGad33433 )
The ypxfr command fails with the error message:
"ypxfr: bind_to_server clntudp_create error: RPC unknown
host".
( SR:8606184904 CR:JAGad54106 )
rpc.lockd can cause a user process to hang, if a loopback
mount is specified with an alias hostname.
( SR:8606207826 CR:JAGad77003 )
The nisplus.server script does not kill the rpc.nispasswdd
process when performing the "/sbin/init.d/nisplus.server
stop" command.
( SR:8606181688 CR:JAGad50904 )
On a system with active lofs mounts managed by AutoFS, the
"/sbin/shutdown" command may hang when the file systems are
unmounted.
( SR:8606160777 CR:JAGad30096 )
AutoFS can not unmount hfs or cdrom file system types.
( SR:8606168123 CR:JAGad37405 )
An NFS/TCP client operation receives "NFS server not
responding still trying" messages while attempting to access
the server, even though the server system is up. In this
case, NFS/TCP on the server shows many TCP connections for
local address port 2049 in a CLOSE_WAIT state when
performing a "netstat -an" command.
( SR:8606203242 CR:JAGad72416 )
A process sequentially reading a large file across an NFS
mount runs extremely slow if, at the same time, there is
another process sequentially writing a large file to an
NFS mounted directory, and one or more biods are running on
the system.
( SR:8606152534 CR:JAGad21864 )
AutoFS can mishandle unmounts when a server is down, which
may appear as bdf command anomalies.
( SR:8606202370 CR:JAGad71544 )
NFS locks remain when killing NFS loopback mounted client
applications.
( SR:8606204523 CR:JAGad73705 )
There is no support for cachefs.
( SR:8606176017 CR:JAGad45257 )
AutoFS directories are empty under heavy NFS load.
( SR:8606193515 CR:JAGad62727 )
Duplicate entries for direct AutoFS mounts show up in the
/etc/mnttab file.
( SR:8606186548 CR:JAGad55752 )
An RPC server running in TCP multithreaded mode develops a
file descriptor leak when clients are on the same machine as
the server. Eventually the server fails to process the
clients' requests when the process limit is attained.
( SR:8606146332 CR:JAGad15675 )
The I/O throughput on a locked file is low compared to the
throughput on an unlocked file.
PHNE_23502:
( SR:8606183171 CR:JAGad52387 )
UDP packets sent by remote users can cause RPC applications
to abort.
( SR:8606160596 CR:JAGad29916 )
The /var/adm/automount.log file fills up the /var file
system.
( SR:8606178618 CR:JAGad47845 )
The rename() system call always fails when trying to rename
a directory in a cachefs-mounted file system, if the target
directory name exists. 'errno' gets set to EPERM ("Not
Owner").
( SR:8606128739 CR:JAGac86286 )
File corruption over NFS when NFS file server is TOC'ed.
( SR:8606140609 CR:JAGad09968 )
Core dump occurs when user belongs to more than 16 groups.
( SR:8606159451 CR:JAGad28779 )
A multithreaded process can hang when performing NFS file
operations and concurrently writing core, forking, using a
debugger, or undergoing job control. A similar situation
arises in any process where a user can issue a CTRL-Z, send
a SIGSTOP, or use a debugger. These hung processes also
carry the potential of hanging the system.
( SR:8606180764 CR:JAGad49985 )
rpc.lockd causes user process to hang.
( SR:8606180765 CR:JAGad49986 )
rpc.lockd causes user process to hang.
( SR:8606175862 CR:JAGad45102 )
Mount points of automounter directories are not visible when
executing the "ls" command.
( SR:8606175781 CR:JAGad45021 )
The "du" output is eight times higher over NFS with an HP
server / Sun Client configuration when compared to an HP
server / HP Client configuration.
( SR:8606172747 CR:JAGad42007 )
Changing an NIS+ user's password with nispasswd() or
passwd() while the system is in trusted mode without
specifying the dot "." notation (non fully qualified domain)
results in an incorrect sub-domain being created.
( SR:8606165859 CR:JAGad35150 )
Data page fault panic while copying a file using NFS V3
between an HP-UX client and an NT InterDrive server where
the target filesystem is full.
( SR:8606199460 CR:JAGaa35208 )
The libnsl library uses a large amount of BSS and DATA
memory.
( SR:8606174017 CR:JAGad43270 )
rpc.lockd drops the lock request from the same process for a
different file.
( SR:8606185238 CR:JAGad54440 )
Partial record unlocking in NFS does not work properly.
( SR:8606173363 CR:JAGad42623 )
The "man nfsstat" command does not display all of the
statistical data fields.
( SR:8606178998 CR:JAGad48223 )
Unnecessary NFS read before write is causing severe
decreases in write performance.
( SR:8606176384 CR:JAGad45622 )
The boot process of a Linux client fails due to rpc.mountd
not allowing AUTH_NULL authentication.
( SR:1653311530 CR:JAGab46804 )
Unable to use fsirand() on JFS file system.
( SR:8606169796 CR:JAGad39062 )
Data page fault panic due to bad machine name length in NFS
request.
( SR:8606171662 CR:JAGad40926 )
exportfs sometimes does not work correctly when invalid host
names are present in a host list in the /etc/exports file.
( SR:8606179200 CR:JAGad48424 )
rpc.lockd core dumps when the user process is interrupted
with CTRL-C.
( SR:8606169011 CR:JAGad38290 )
The automount process hangs, making automounter mount points
inaccessible.
( SR:8606179951 CR:JAGad49173 )
getservbyname() using NIS returns NULL when trying to
resolve service aliases.
( SR:8606167053 CR:JAGad36339 )
An NFS/TCP client operation receives "NFS server not
responding still trying" messages while attempting to access
the server, even though the server system is up. The server
displays "vmunix: WARNING: tcpd_thread_create: thread_create
failed: 11" messages in /var/adm/syslog/syslog.log.
PHNE_23248:
( SR:8606188447 CR:JAGad57655 )
autofs does not appear to remove inactive mounts from
/etc/mnttab.
PHNE_22878:
( SR:8606127276 CR:JAGac68822 )
A failing mount request may return a wrong error code.
( SR:8606168603 CR:JAGad37881 )
Loading binary via CacheFS is slower than via NFS.
( SR:8606166823 CR:JAGad36110 )
NFS client panics when the server is down.
( SR:8606168604 CR:JAGad37882 )
Periodically the CacheFS umount command hangs.
( SR:8606135856 CR:JAGad04988 )
Panic occurs when trying to umount a ufs filesystem which is
still in the process of mounting.
( SR:8606160808 CR:JAGad30127 )
When mounting an NFS directory with the quota option, the
quota does not get enforced.
( SR:8606168606 CR:JAGad37884 )
No visible increase in performance while reloading an
application binary.
( SR:8606166314 CR:JAGad35601 )
The 'kill' and 'stop' NFS processes on a client do not work.
( SR:8606145476 CR:JAGad14812 )
Remote locking of a file occasionally leads to a system
hang.
( SR:8606155843 CR:JAGad25155 )
NIS fails to return all groups for id/group command when
concurrent access is allowed to maps.
( SR:8606146885 CR:JAGad16228 )
The "rpcinfo -u" command fails when the nfsd unregisters
a running nfsd process on startup.
( SR:8606159736 CR:JAGad29062 )
Unexpected delay in unmounting of Autofs mounts.
( SR:8606158230 CR:JAGad27560 )
Occasionally umount may cause panic within
nfs3_close/nfs_close.
( SR:8606145517 CR:JAGad14853 )
If NFS is stopped and restarted, the nfsd processes do not
get re-registered with rpcbind, and rpcinfo commands fail.
( SR:8606169239 CR:JAGad38515 )
Cachefsstat is not returning the correct statistics.
( SR:8606157592 CR:JAGad26923 )
Newkey does not recognize the root password of the remote
machine and returns an error, "ERROR, passwd differs from
login passwd".
Defect Description:
PHNE_33498:
( SR:8606167807 CR:JAGad37090 )
Some references to vhe functions were not removed when the
functions were obsoleted.
Resolution:
Removed all references to the vhe functions from the
ypinit(1m) man page.
( SR:8606406594 CR:JAGaf66500 )
Non-RPC UDP messages containing non-zero-length data can be
inserted into the STREAMS read queue. These messages exhaust
the number of allowed receive retries before any valid RPC
data is received. This causes commands such as "ls" which
attempt to access a file system with a soft NFS mount over
the UDP transport to fail.
Resolution:
The invalid RPC UDP messages are received and ignored
without affecting the retry counter.
( SR:8606413268 CR:JAGaf73129 )
CacheFS uses a linear search to release buffers allocated by
vnodes for a mount point. If the buffer cache is large,
which can occur with a single "ll" command on a large
directory, the unmount can take anywhere from 30 seconds to
over five minutes to complete.
Resolution:
A new mechanism is used to release the buffers associated
with a mount point when the mount point is unmounted. This
fix requires PHKL_33367 to be installed.
( SR:8606413731 CR:JAGaf73591 )
VFS requires NFS to purge the file handle cache if an error
is returned from rfs3call(). If the cache is not purged, a
hang occurs.
Resolution:
Purge the file handle cache if an error is returned from
rfs3call().
PHNE_33315:
( SR:8606385895 CR:JAGaf46049 )
When writing over an NFS mount point, there are unnecessary
read-before-write calls done after 2 GB of data has been
written to the file. This occurs because the calculation of
the file offset using the block number and block size
results in an overflow condition. This causes read
avoidance logic to fail and affects performance.
Resolution:
Modified the file offset calculation to avoid an overflow
condition.
( SR:8606394445 CR:JAGaf54468 )
RPC maintains a server address cache that is used by AutoFS.
The server addresses remain in cache until they are flushed
out, and there is no timeout mechanism for the cache. If one
of the IP addresses is modified, AutoFS could use the old IP
address.
Resolution:
A timeout mechanism has been implemented for the RPC server
address cache so that AutoFS uses the correct IP address.
The cache expires after 180 seconds.
( SR:8606399902 CR:JAGaf59858 )
Network congestion results in sending EAGAIN or TFLOW
(flow control) messages to the server, which is in
non-blocking I/O mode. The server continues to send data to
the client, resulting in more EAGAIN and TFLOW messages.
This causes high CPU utilization.
Resolution:
The server now waits for the transport to be ready to accept
data before trying to re-send data to the client.
( SR:8606400085 CR:JAGaf60041 )
The rpc_clnt_calls(3N) man page lists the parameter
"eachresult" for the rpc_broadcast_exp function as the tenth
parameter. It should be listed as the eighth parameter.
Resolution:
Corrected the order of the parameters for the
rpc_broadcast_exp function in the rpc_clnt_calls(3N) man
page.
( SR:8606401738 CR:JAGaf61675 )
The rpc.lockd and rpc.statd processes do not close the file
descriptor when logging is disabled with SIGUSR2. If all of
file descriptors are used this causes the process to hang.
Resolution:
The logging routines now close the file descriptor when
logging is disabled.
( SR:8606402519 CR:JAGaf62454 )
The rpc.lockd and rpc.statd processes attempt to log I/O
errors without checking whether logging is enabled or not.
If logging is not enabled then an attempt is made to write
to a null file descriptor. If libc patch PHCO_24942 (or a
superseding patch) is installed this causes a core dump,
otherwise an EBADF error is returned. This problem occurs
when the processes are started without using the "-l"
option.
Resolution:
The logging procedures now check whether or not logging is
enabled before attempting to log I/O errors.
( SR:8606405252 CR:JAGaf65172 )
Memory allocated for arguments in rpc.rquotad is not freed.
This results in a memory leak which can lead to a core dump.
Resolution:
The memory allocated for arguments in rpc.rquotad is now
freed before returning from the function.
PHNE_32811:
( SR:8606345889 CR:JAGaf06735 )
After a server's rpc.lockd is restarted, the clients attempt
to reclaim their locks. The server waits for a grace period
for all client rpc.lockd reclaim requests. When the grace
period expires, granting reclaim requests can cause data
corruption. The server's rpc.lockd does not check whether
the lock request is a new request or a reclaim request after
the grace period expires.
Resolution:
The server's rpc.lockd was modified to deny reclaim requests
after the grace period expires.
( SR:8606377207 CR:JAGaf37485 )
If rpc.lockd is performing a re-transmission and it receives
a SIGALRM signal, the rpc.lockd process exits.
Resolution:
Ignore the SIGALRM signal while performing the
re-transmission.
( SR:8606389345 CR:JAGaf49493 )
With a soft NFS mount, the number of retransmissions that
actually occur is one less than the value specified in the
retrans option (or the default) for the NFS mount command.
Also, there exists an inconsistency in the documentation of
the default retrans value between the mount_nfs man page
and the NFS Services Administrator's Guide.
Resolution:
Corrected the inconsistency in the retrans value displayed
by nfsstat when used with -c and -m. Corrected the
inconsistency in the documentation of the default retrans
value by updating the NFS Services Administrator's Guide
to correct the default retrans value to 5.
( SR:8606393839 CR:JAGaf53894 )
The rename() system call behaves incorrectly when renaming
a directory in a CacheFS-mounted file system using NFSv2,
if the target name is an existing directory. The incorrect
behavior is due to the reference count for a directory
getting incremented to more than 1 (due to CacheFS and NFS).
This causes "unlink-open-file" logic to be executed, which
is only applicable to regular files and not directories.
As a result a "Not Owner" error returned. If the target name
is an empty directory, the rename() call should succeed. If
the target name is a non-empty directory, a "File exists"
error should be returned.
Resolution:
An additional check for the file type is added before the
"unlink-open-file" logic is executed.
( SR:8606394223 CR:JAGaf54255 )
The "ypwhich -m" command fails when the client is bound to
a Solaris NIS server. This is because the receiving buffer
is not large enough to handle the message returned by the
server if there is a large number of NIS domains.
Resolution:
Increased the size of the client's receiving buffer. Also
increased the size of the message returned by the server.
PHNE_32477:
( SR:8606386532 CR:JAGaf46686 )
When clnt_tli_create() is called with large send/receive
buffers, certain error checking is not done which results in
the application dumping core.
Resolution:
Additional error checking is done when calling
clnt_tli_create() with large send/receive buffers.
( SR:8606387765 CR:JAGaf47918 )
If a write() system call is interrupted by a signal before
it writes any data to the NFS server (this could happen if
the server is down), it should return -1 and have errno set
to EINTR. Instead, the return value of write() is
inconsistent. It can return success and have errno set to
0, or it can return -1 and have errno set to EINTR. This is
because NFS does a "read before write" if the write length
is smaller than the block size - it reads one block,
modifies it, and writes it back. If "read before write" is
used, the signal interrupts the read path and the behavior
is as expected (returns -1). If no "read before write" is
used (the file is being created, the buffer has already been
read because of a previous write, or the write is one block
long) the signal interrupts the write path and the behavior
is faulty and returns success. This occurs because the value
returned by bwrite() is ignored for synchronous writes.
Resolution:
The value returned by bwrite() for synchronous writes is now
checked and returned to the application.
( SR:8606387807 CR:JAGaf47960 )
If an NFS client sends a read request with an invalid read
size, the server crashes while allocating memory.
Resolution:
Reject read requests greater than the maximum allowable
transfer size for PV2. Allocate memory for the maximum
allowable transfer size for read requests greater than the
maximum allowable transfer size for PV3.
( SR:8606389142 CR:JAGaf49290 )
A TCP transport handle is passed to a UDP routine because a
flag is not set correctly in some rpc.lockd procedures. This
causes rpc.lockd to dump core.
Resolution:
The rpc.lockd procedures now check whether the transport
handle is TCP or UDP and set the flag correctly.
( SR:8606389255 CR:JAGaf49403 )
A deadlock situation occurs between three NFS threads in a
single process. Thread A sleeps non-interruptibly in the
following code path:
_swtch+0xc4
_sleep_one+0x1a8
ksleep_one+0x2f4
hp_cv_wait+0x60
nfs_purge_rddir_cache+0x6c
nfs3create+0x258
nfs3_create+0xc4
vns_create+0xa4
vn_create+0xd4
vn_open+0x19c
copen+0xa8
open+0x38
syscall+0x768
syscallinit+0x55c
Thread A waits for a resource owned by thread B (in this
case it waits for thread B to finish accessing directory
data that is being updating) but thread B sleeps
interruptibly in the following code path:
_swtch+0xc4
thread_process_suspend+0x20c
issig+0x338
_sleep_one+0x2e0
ksleep_one+0x2f4
hp_cv_wait_sig+0x64
nfs3_readdir2+0x1f4
getdents+0xc8
syscall+0x768
syscallinit+0x55c
Thread C calls vfork() which sends SIGSTOP to all the
threads in the process, suspending thread B. Thread C is in
the following code path:
_sleep+0x218
thread_suspend_wait+0x2e8
for_specific_threads+0xc0
process_wide_suspend_wait+0x28
newproc+0x1f8
fork1+0x220
vfork+0x14
syscall+0x768
syscallinit+0x55c
Thread A is still waiting for the resource to be
freed by thread B, and thread C waits for thread A to
process SIGSTOP, which causes the hang.
Resolution:
When an NFS thread does an interruptible sleep, block
SIGSTOP so that the thread does not become suspended.
( SR:8606348790 CR:JAGaf09611 )
The master map syntax for specifying a direct map is not
described in the automount(1m) man page.
Resolution:
Added a description on how to use direct maps.
( SR:8606381800 CR:JAGaf42012 )
The Enhanced AutoFS product implementation has obsoleted the
need for the JAGad50904 fix. This fix prevented a possible
hang during system shutdown.
Resolution:
Added logic to the nfs.client script so that the fix for
JAGad50904 is only exercised if the Enhanced AutoFS product
is not installed.
( SR:8606393701 CR:JAGaf53756 )
There is a small memory leak in libnsl which occurs every
time an automountd thread is created.
Resolution:
Modified the libnsl code to eliminate the memory leak.
PHNE_31929:
( SR:8606373075 CR:JAGaf33480 )
Memory mappings are not removed while truncating memory
mapped files over NFS.
Resolution:
Modified the NFS truncate path to completely remove the
memory mappings.
( SR:8606375121 CR:JAGaf35426 )
For an AutoFS loopback mount (LOFS), the filesystem name is
not passed correctly from the on command. This causes a
mount failure and thus the on command also fails.
Resolution:
Modified the on command to mount the correct filesystem for
loopback mounts (LOFS).
( SR:8606379022 CR:JAGaf39271 )
If an NFSv3 client attempts to access a remote NFS server
and the requested file is not currently available, the
server returns an NFS3ERR_JUKEBOX error to the client. Upon
receiving this error, the client re-issues the request 10
seconds later. This delay time affects NFS performance.
Resolution:
Made the delay time variable global so that it can be
adjusted by adb. The default delay time is still 10 seconds.
To modify the delay time, set the kernel variable
nfs3_jukebox_delay by patching the vmunix file and
rebooting, or by patching the kernel image in memory. To
patch the vmunix file (survives a reboot), use the
following command:
echo 'nfs3_jukebox_delay?W 0d300' |
adb -w /stand/vmunix /dev/kmem
The above example sets the delay time to 3 seconds.
The unit of the delay time variable nfs3_jukebox_delay is
the number of clock ticks (number of seconds * 100).
To patch the in-memory kernel (does not survive a reboot),
use the command (for a delay time of 3 seconds):
echo 'nfs3_jukebox_delay/W 0d300' |
adb -w /stand/vmunix /dev/kmem
A kernel regeneration will require these steps to be done
again.
PHNE_31097:
( SR:8606345966 CR:JAGaf06811 )
If AUTO_MASTER is set to "" in the /etc/rc.config.d/nfsconf
file, and AutoFS is started with the
"/sbin/init.d/nfs.client start" command, the
/etc/auto_master file is not created. The text that should
be in the /etc/auto_master file is echoed to a null variable
and appears in the /etc/rc.log file and also on stderr.
Resolution:
The nfs.client script was modified to create the
/etc/auto_master map with the default entry if the
AUTO_MASTER variable in /etc/rc.config.d/nfsconf is set to
"".
( SR:8606347603 CR:JAGaf08425 )
Currently libnsl does not have any provision to identify
the end-point pair for UDP based RPC applications.
Resolution:
Modified libnsl to add functionality to identify the
end-point pair for UDP based RPC applications.
( SR:8606354067 CR:JAGaf14861 )
If nfsd's need to be restarted, then previously running
nfsd's must be stopped first. This is not documented in the
nfsd man page.
Resolution:
The nfsd man page has been updated to mention that nfsd's
must be stopped before new nfsd's are started.
( SR:8606359725 CR:JAGaf20421 )
The automountd process may dump core during an RPC unmount
call. This occurs if automountd is started with the "-v"
option and if an error occurs in the RPC umount call. This
is because a client handle is passed to a function after it
has been destroyed.
Resolution:
Relocated the function calls which destroy the client
handle.
( SR:8606360695 CR:JAGaf21402 )
An NFS TCP client continuously adds data that is received
from the NFS server to an RPC message fragment while trying
to assemble an RPC reply message from the server when the
fragment size specified in the record-marking header for a
fragment sent from the server is wrong.
Resolution:
The NFS TCP client is modified to disconnect if the size of
an RPC message fragment received goes beyond a specified
limit.
( SR:8606364296 CR:JAGaf24937 )
The error number is not set prior to returning from the
netdir_getbyname() function.
Resolution:
Modified the netdir_getbyname() function to set the error
number before returning.
( SR:8606364383 CR:JAGaf25022 )
Threads that are waiting for an NFS rnode lock cannot be
interrupted. As a result they will sleep until the NFS
server becomes available or until the lock gets released,
which causes applications to hang.
Resolution:
Created a kernel variable that allows toggling between the
existing behavior of non-interruptible threads and allowing
the threads that are waiting for the rnode lock to be
interrupted. This fix is disabled by default. To enable the
fix, set the kernel parameter nfs_new_rnode_lock_code to 1
by patching the vmunix file and rebooting, or by patching
the kernel image in memory. To patch the vmunix file
(survives a reboot), use this command:
echo 'nfs_new_rnode_lock_code?W 0d1' |
adb -w -k /stand/vmunix /dev/kmem
To patch the in-memory kernel (does not survive a reboot),
use this command:
echo 'nfs_new_rnode_lock_code/W 0d1' |
adb -w -k /stand/vmunix /dev/kmem
A kernel regeneration will require these steps to be done
again.
( SR:8606365866 CR:JAGaf26496 )
An NFS client does not check whether the byte count returned
by the server exceeds the requested number of bytes. If
additional data is returned by the server, the client
crashes while attempting to copy the extra data.
Resolution:
Modified the NFS client to have read requests fail with the
error EIO if the server returns more data than requested.
( SR:8606369177 CR:JAGaf29721 )
Under heavy load, memory in an internal linked list is
overwritten causing rpcbind(1M) to core dump.
Resolution:
Modified the buffer allocation in rpcbind to avoid memory
corruption.
( SR:8606366266 CR:JAGaf26870 )
The server's rpc.lockd grants the same lock to two blocking
clients. One client is the server with a loopback mount
using a server name other than the name returned by the
hostname command.
Resolution:
The server name for locking requests is changed to the name
returned by the gethostname function for loopback mounts if
using a name other than that returned by the hostname
command.
( SR:8606379191 CR:JAGaf39440 )
There is a restriction that limits establishing quotas to
user ID's less than 67,000,000. The rpc.rquotad does not
currently support largefiles. Largefiles must be supported
to establish quotas for user ID's greater than or equal to
67,000,000.
Resolution:
Modified rpc.rquotad to support largefiles. This change, in
conjunction with the fix for JAGaf21431, eliminates the
restriction of only being able to establish quotas for user
ID's less than 67,000,000.
( SR:8606379708 CR:JAGaf39957 )
Kernel RPC performs re-connects to an NFS/TCP server on
behalf of an NFS/TCP client using the same local TCP
end-point which was originally connected. It does not create
a new local TCP end-point bound to a different port.
Resolution:
The kernel RPC/TCP client is modified to create a new local
TCP end-point for connection to a remote TCP end-point
whenever a disconnect is received for an existing local-
remote TCP connection tuple.
PHNE_30661:
( SR:8606354850 CR:JAGaf15606 )
The ypxfr failure is caused by improper use of the
dbmclose() API.
Resolution:
Modified the ypxfr code was to call dbmclose() as specified
in its prototype definition.
( SR:8606355618 CR:JAGaf16332 )
The client's rpc.lockd sends a re-claim lock request for
pending unlock, cancel, and test locks. It should only send
re-claim lock requests for granted locks. If the server is
performing rpc.lockd crash recovery, the client's rpc.lockd
dumps core after sending a re-claim lock request for a
pending lock.
Resolution:
Modified rpc.lockd so that the client only sends re-claim
requests for granted locks.
( SR:8606357066 CR:JAGaf17766 )
The error message "exportfs error: /etc/xtab in use. Please
try again." is not accurate. The global file lock table
maintained in the kernel is full.
Resolution:
The exportfs command is modified to print the error message
"Cannot open or lock /etc/xtab. Please try again." Note that
if you are using HP Serviceguard NFS Toolkit, it is
recommended to install version A.11.11.05 since this fix may
affect package startup and shutdown.
( SR:8606357361 CR:JAGaf18061 )
The CacheFS mount command dumps core if the name of the
directory being mounted contains more than 126 characters.
Resolution:
Modified the CacheFS mount code to return the error "cacheid
too long" if the total number of characters in the name of
the directory being mounted exceeds the limit.
( SR:8606358606 CR:JAGaf19305 )
Memory allocated for a message block is not being freed when
an XDR failure occurs while replying to an NFS/UDP client.
Resolution:
Memory allocated for a message block in the NFS server is
now freed after an XDR failure while replying to an NFS/UDP
client.
( SR:8606363276 CR:JAGaf23938 )
The nfs_portmon variable determines whether or not an NFS
server only processes requests from privileged ports. This
functionality is not documented.
Resolution:
Documented the tunable parameter nfs_portmon in the nfsd man
page.
( SR:8606329363 CR:JAGae90487 )
There is a race condition between the lock and write threads
which causes the panic. An uncached temporary buffer is used
for buffered I/O.
Resolution:
Temporary buffers are released after I/O completes. This
prevents the temporary buffer from taking the buffered I/O
path.
PHNE_30380:
( SR:8606382150 CR:JAGaf42361 )
When an RPC application is started by inetd, a connection is
established. The transport interface state for the
connection is not checked, which causes libnsl to write to
an incorrect memory location. This can lead to the
application terminating unexpectedly and dumping core. This
occurs only with applications that use TCP and call
svctcp_create() or svc_tli_create(). Applications that call
svc_tli_create() with a null netconfig parameter (the second
parameter) are not affected.
Resolution:
Modified libnsl to check for the transport interface state
to avoid writing to an incorrect memory location.
PHNE_30378:
( SR:8606319551 CR:JAGae82039 )
The rpc_soc man page does not mention the RPC socket library
(-lrpcsoc).
Resolution:
The rpc_soc man page was modified to include information
about the RPC socket library (-lrpcsoc).
( SR:8606339927 CR:JAGaf00848 )
If the automount process is killed with SIGTERM, there are
entries left in the /etc/mnttab file which cause automount
to hang when it is re-started. This is caused by a
synchronization problem between the signal handler and the
mount module.
Resolution:
Fixed the synchronization problem between the signal handler
and the mount module.
( SR:8606344107 CR:JAGaf04959 )
The problem occurs when packets are delivered out of order
and some of the packets are dropped due to the network
environment being stressed. The response to the unlock
request arrives later than the response to the blocked lock
request. The client's rpc.lockd discards the reply from the
server for the blocked lock request, since the lock in
question is still held by another process on the client
machine, and it removes the blocked lock request from the
wait queue. If the reply to the cancel request arrives when
there is no entry for this lock in the wait queue, the
client's rpc.lockd daemon tries to add this lock to the
granted lock queue whose pre_fe field is null, which causes
the core dump.
Resolution:
When the client's rpc.lockd receives the granted lock for
the blocked lock request, it removes the lock from the
wait queue only if the lock is successfully added to the
granted lock queue.
( SR:8606345310 CR:JAGaf06160 )
Insufficient buffer size allocated to store NIS+ subdomain
information.
Resolution:
The buffer size has been increased to the maximum buffer
size supported by xdr_string().
( SR:8606346094 CR:JAGaf06921 )
The NFS_TCP configurable environment variable is valid only
for the 11.00 release of HP-UX. The setoncenv command was
not modified to remove it from subsequent HP-UX releases.
Resolution:
Modified the setoncenv command so that NFS_TCP is not a
configurable variable.
( SR:8606346132 CR:JAGaf06959 )
The NFS version information being passed to the kernel
profiling code is not being interpreted correctly on 64-bit
systems.
Resolution:
The code passing NFS version information to the kernel
profiling code has been modified with the correct variable
type.
( SR:8606347226 CR:JAGaf08050 )
Non-RPC UDP messages containing zero-length data can be
inserted onto the STREAM read queue and exhaust the number
of allowed receive retries before any valid RPC data is
received.
Resolution:
Modified the code to discard invalid zero-length data
packets without affecting the received retry counter while
receiving these invalid packets.
( SR:8606347264 CR:JAGaf08087 )
AutoFS does not handle direct maps properly when looking up
the current and parent directories.
Resolution:
Modified AutoFS to work properly when looking up the current
and parent directories with direct maps.
( SR:8606351701 CR:JAGaf12506 )
NFS enforces quotas only when the mount option "quota" is
used.
Resolution:
Modified the mount_nfs man page to remove the statement that
quota checking is done by default.
( SR:8606355273 CR:JAGaf16027 )
Netselect's nc_perror() outputs the error message for
NC_BADLINE incorrectly. Netdir's netdir_perror() outputs an
additional newline for all its errors.
Resolution:
Modified the affected error messages so the output is
correct.
( SR:8606356288 CR:JAGaf16994 )
UDP only supports up to 32K for the transfer size. Some
servers permit a 64K transfer size, which causes the client
to hang during a write operation to an NFS mounted
directory.
Resolution:
Set the default transfer size for NFS clients to 32K.
PHNE_30086:
( SR:8606324736 CR:JAGae87042 )
The RPC client, loaded such that it is operating under
memory pressure, may fail to allocate a STREAMS message
block to request the TCP transport layer to complete an
orderly connection close. This failure puts the RPC client
into a state in which it is unable to reset or finish
closing the connection and causes the connection to remain
in the CLOSE_WAIT state.
Resolution:
RPC client TCP connection management is modified to ensure
that a STREAMS message block to request TCP transport to
initiate or complete an orderly TCP connection close will
always be available.
( SR:8606336931 CR:JAGae97978 )
A race condition occurs between the RPC client TCP
connection timeout and the TCP transport layer connection
timeout. During a TCP connection attempt the RPC client may
miss the indication from the TCP transport layer when a
connection has been established.
Resolution:
The TCP connection timeout used by the RPC client is
modified to match the default TCP connection timeout,
instead of using the RPC call wait time.
PHNE_29883:
( SR:8606312940 CR:JAGae75747 )
When the length of a symbolic link name exceeds 1024
characters, the string that is passed from the function
reading the symbolic link to the strlen function contains a
-1. This causes the NFS client to crash.
Resolution:
Dependency on PHKL_30398.
The length of the symbolic link name is checked to see if it
exceeds 1024 characters. If it does, an error is returned.
( SR:8606320583 CR:JAGae83065 )
While AutoFS is unmounting a file system that is being
managed by AutoFS, it is not able to recognize whether the
file system was manually unmounted or not. This is because
of improper synchronization in AutoFS.
Resolution:
AutoFS has been modified to provide the proper
synchronization when the file systems are unmounted.
( SR:8606330100 CR:JAGae91223 )
An error message is returned while the rpcbind server is
processing a disconnection message for a connection that has
not yet been accepted. This error is not handled correctly.
Resolution:
The error returned while accepting a disconnected connection
is now handled properly. The syslog.log message is no longer
logged for that error.
( SR:8606333242 CR:JAGae94337 )
There is a problem with nested AutoFS maps.
Resolution:
The behavior of nested AutoFS maps has been modified.
( SR:8606334605 CR:JAGae95685 )
The 64-bit xdr_char() and xdr_u_char() routines do not check
for encoding/decoding the XDR stream failures, and do not
return an error. This may cause an application to core dump.
Resolution:
FALSE is now returned upon failure of encoding/decoding in
the xdr_char() and xdr_u_char() routines.
( SR:8606336578 CR:JAGae97630 )
Inconsistent recommendations, coming from multiple
references about the number of nfsd's configured to be run
on an HP-UX NFS server, have caused confusion.
Resolution:
The default number of nfsd's configured to be run on an
HP-UX NFS server has been modified in
/usr/newconfig/etc/rc.config.d/nfsconf. The nfsd man page
has also been modified to give a clearer description about
choosing the number of nfsd's to be run on an HP-UX NFS
server.
PHNE_29783:
( SR:8606347801 CR:JAGaf08623 )
The rpc.ypupdated code does not function properly.
Resolution:
The rpc.ypupdated code has been modified to fix the problem.
PHNE_29303:
( SR:8606286517 CR:JAGae50456 )
This product update contains minor enhancements required to
enable the HP-UX Enhanced AutoFS product.
Under heavy load, automountd may treat a duplicate mount
request as a new mount request causing it to retry an NFS
mount even though the file system was mounted earlier. This
leads to a case where AutoFS mount points cannot be mounted.
Resolution:
Developed a new and improved duplicate request cache in
libnsl (which the automountd in the HP-UX Enhanced AutoFS
product uses) to eliminate existing problems.
( SR:8606302457 CR:JAGae65816 )
The client's rpc.lockd routine sends an initial request,
then waits for a reply from the server. If no reply is
received the client will begin retransmitting the request.
The retransmissions are controlled by a timer, sending a
request every other timer cycle. However, the first
retransmission occurs at the beginning of the first timer
cycle. The server treats this first retransmission as a new
lock request because of the short time interval between the
initial request and the first retransmission, which causes
the client to hang.
Resolution:
Modified the rpc.lockd code to send the first retransmission
request at the beginning of the second timer cycle.
( SR:8606304146 CR:JAGae67490 )
The pwchanger field of the trusted table is not cleared when
the password is changed. This is due to the restriction
imposed on the size of the XDR memory stream buffer. This
problem only occurs when the admin group has more than 312
members.
Resolution:
Increased the size of the XDR memory stream buffer.
( SR:8606306059 CR:JAGae69107 )
When an NIS+ user's keys are changed in the NIS+ tables, the
keys are not updated in the NIS+ clients' keyserv daemon. As
a result, there is a mismatch between the NIS+ credential
tables, causing the NIS+ user to not be authenticated, and
the NIS+ client to hang.
Resolution:
The NIS+ user is authenticated as "nobody" on the NIS+
client, and an authentication failure is logged in the
syslog.log file. This is documented in the newkey(1m) man
page.
( SR:8606306319 CR:JAGae69367 )
When the /var filesystem becomes full, the
/var/nis/NIS_COLD_START file is overwritten with a zero-byte
file. This causes the NIS+ client authentication to fail.
Resolution:
The /var/nis/NIS_COLD_START file is no longer overwritten
when there is no space left in the /var filesystem.
( SR:8606312085 CR:JAGae74904 )
When a symbolic link is attempted from a Linux NFS client to
a file or directory that has a full path name length of more
than 1024 characters, the NFS server receives a -1 in the
name field. When passed to strlen this results in a system
crash.
Resolution:
Added a test to see if the length of the symbolic link's
name exceeds 1024 characters. If it does, an error is
returned.
( SR:8606313428 CR:JAGae76226 )
If the /etc/hosts file has entries with IP addresses which
are not in the proper dot notation format, some ONC commands
give incorrect results.
Resolution:
Recognize incorrect IP addresses and return an appropriate
error condition value.
( SR:8606322456 CR:JAGae84924 )
If the vers=<version number> option of the NFS mount command
is specified, then the mount command is expected to mount
the requested version. If this version is not supported by
the NFS server, then the mount command should fail. If the
user does not provide a version number with the NFS mount
command, a mount is attempted with the highest available
version. If the highest version is not available it will
attempt to perform a mount with a lower version. Currently,
the NFS mount command uses this approach, even when a
version number is specified.
Resolution:
Avoid negotiations with the NFS server for lower mount
versions when the mount command specifies the
vers=<version number> option, and return a failure if the
specified version is not available.
( SR:8606335982 CR:JAGae97035 )
This product update contains minor enhancements required to
enable the HP-UX Enhanced AutoFS product.
Resolution:
AutoFS has been modified to operate in enhanced mode when
the HP-UX Enhanced AutoFS product is installed. In addition,
the nfs.client script has been modified to work with HP-UX
Enhanced AutoFS. The AutoFS man pages have been modified to
include information for using HP-UX Enhanced AutoFS.
( SR:8606332784 CR:JAGae93899 )
The panic occurs because the nfs3readdirplus() routine does
not check for unaligned buffers. This causes an incorrect
type value to be passed to the makenfs3node() routine, which
crashes the system.
Resolution:
Modified the nfs3readdirplus() and xdr_getdirpluslist()
routines to align the XDR buffer.
PHNE_29211:
( SR:8606317529 CR:JAGae80090 )
Connection-oriented RPC packets are not being decoded
effectively during network delays.
Resolution:
The non-blocking I/O feature is being provided to improve
the performance of connection-oriented RPC transports during
network delays.
( SR:8606320560 CR:JAGae83042 )
The rpc.mountd code returns an incorrect error message.
Resolution:
The rpc.mountd code has been modified to fix the error
message.
( SR:8606317535 CR:JAGae80096 )
Connection-oriented RPC packets are not being decoded
effectively during network delays.
Resolution:
The non-blocking I/O feature is being provided to improve
the performance of connection-oriented RPC servers during
network delays.
( SR:8606337943 CR:JAGae98946 )
If "Error in RPC subsystem" messages are displayed or logged
for NIS+ related commands or applications that use NIS+
API's, a possible cause is that the default maximum RPC
record size for NIS+ (9000 bytes) is too small. Examples of
NIS+ commands that could be affected are nistbladm,
nisgrpadm, and nisaddent. NIS+ API's that could be affected
include nis_add() and nis_modify().
Resolution:
Added the rpc.nisd -z option to set the maximum record size
needed by the application. Some experimentation may be
required to determine the optimal value for a particular
application. The NIS+ server needs to be restarted after
changing this value for it to take effect. Please see the
rpc.nisd man page for details on using the -z option.
PHNE_28983:
( SR:8606220736 CR:JAGad89872 )
After an NIS+ user is deleted, its user ID is reused for a
different user. The shadow field for each user is set to age
the password such that the password must be changed at the
first login, which failed. This occurs due to internal
caching within the rpc.nispasswdd daemon. The cache contains
credential information for the deleted user.
Resolution:
Added functionality to determine if the cache is obsolete,
and if so, remove the obsolete cache entry.
( SR:8606278781 CR:JAGae42837 )
The automount daemon is getting saturated with mount
requests, and is not able to handle duplicate requests
properly. The daemon is considering each duplicate request
as a new request, and this new request creates a mount point
without a proper state which allows it to be unmounted.
Resolution:
Modified the automount daemon to properly handle duplicate
mount requests.
( SR:8606294191 CR:JAGae57931 )
When a mount request for a directory is issued, AutoFS
increments a reference count. The reference count must be
decremented when a mount request fails, but AutoFS is not
doing so. When a mount request for the same directory
succeeds, AutoFS is unable to unmount those directories
after the unmount timer expires. So directories mounted with
AutoFS will remain mounted and will never be unmounted by
AutoFS.
Resolution:
Fixed AutoFS to maintain a correct reference count, even
when a mount request fails, so that directories will always
be unmounted properly.
( SR:8606294387 CR:JAGae58084 )
Under heavy load, automountd is not removing stale entries
when the unmount request list runs out of memory.
Automountd is not able to process unmount requests and
AutoFS mounted file systems never get unmounted.
Resolution:
Modified automountd to correctly remove stale entries from
the unmount request list.
( SR:8606294761 CR:JAGae58456 )
Re-exporting a large number of entries in the /etc/exports
file takes almost twice as much time as the original export.
Resolution:
Modified the exportfs code so that the re-exporting time is
now approximately the same as the original export.
( SR:8606295723 CR:JAGae59388 )
Due to a Makefile problem, pcnfsd is only allowing NFS
access for PC users with a default shell ending with "sh".
Resolution:
Modified the Makefile to verify NFS access from a PC by
checking the /etc/shells file.
( SR:8606296110 CR:JAGae59732 )
The nsswitch.conf man page implies that the values in the
/etc/nsswitch.hp_defaults file are the default values.
However, this file is only for 10.20 compatibility and does
not contain the default values.
Resolution:
Updated the nsswitch.conf man page with the correct default
values.
( SR:8606296698 CR:JAGae60267 )
The NFS Server hangs due to a deadlock condition between the
NFS Client and the buffer cache with NFS loopback mounts.
This occurs while flushing buffers containing removed NFS
files. The NFS Server attempts to hold a semaphore lock that
is already being held, resulting in the deadlock.
Resolution:
Buffers containing removed NFS files are flushed whenever a
file is removed or its last reference is closed.
( SR:8606296789 CR:JAGae60355 )
The panic from the ls command occurs because
nfs3readdirplus() does not check for unaligned buffers on
64-bit systems. This causes an incorrect type value to be
passed to makenfs3node(), which crashes the system.
Resolution:
Modified nfs3readdirplus() to handle alignment on 64-bit
systems. Also modified makenfs3node() to verify that the
type value is in the proper range.
( SR:8606297163 CR:JAGae60682 )
While trying to open a file in an NFSv3 mounted file system
from a Solaris NFS server with O_EXCL and the setgid bit of
the parent directory set, the last access time and last
modified time attributes are not updated properly and
changing the group ownership fails. This leaves a file with
no access permissions on the server.
Resolution:
Update the last access time and the last modified time
attributes with one SETATTR call, and update the group ID
with a separate SETATTR call.
( SR:8606297688 CR:JAGae61190 )
All file systems being managed by AutoFS are not getting
remounted after a manual unmount. A manual unmount of those
file systems sometimes gives unpredictable results.
Therefore, manual unmounting is not supported on the file
systems that are being managed by AutoFS.
Resolution:
Manual unmounting of file systems which are being managed by
AutoFS is not supported. This is documented in the man page
of automountd.
( SR:8606297816 CR:JAGae61318 )
A client whose portmapper is not responding is trying to
acquire a lock on a remote file. The server is not able to
communicate the status to the client because it is unable to
get the client's port number for lockd, and is queuing all
of the lock retransmissions from the client in the reply
queue. The queue grows bigger with time, causing a
performance hit and delays in response time, ultimately
causing a server lockd crash.
Resolution:
Verify that a new reply queue add entry request results in a
duplicate entry in the reply queue, and that the request is
dropped if it is a duplicate.
( SR:8606300458 CR:JAGae63927 )
Poor performance is caused by a 20-second delay while AutoFS
is remounting the hierarchy. This delay occurs because a
mount request is sent from the unmount module.
Resolution:
Improved AutoFS' performance by eliminating the need for
waiting 20 seconds for AutoFS to remount the hierarchy and
removing redundant code.
( SR:8606301831 CR:JAGae65194 )
No check is made to see if the /usr/bin/sed command
succeeded in the /sbin/init.d/nfs.core startup script. If
the command fails, the /etc/services file gets overwritten
with a zero-byte file.
Resolution:
Added a check to see if the /usr/bin/sed command succeeded
in the /sbin/init.d/nfs.core script.
( SR:8606302828 CR:JAGae66186 )
AutoFS is not removing stale directories after a mount
request to a non-existent file system fails because AutoFS
is not identifying those directories as stale.
Resolution:
AutoFS has been modified to identify the stale AutoFS
directories, and remove them.
( SR:8606303150 CR:JAGae66506 )
The NFS Server hangs due to a deadlock condition between the
NFS Client and the buffer cache with NFS loopback mounts.
This occurs while invalidating buffers containing removed
NFS files. The NFS Server attempts to hold a semaphore lock
that is already held, resulting in the deadlock.
Resolution:
Verify that buffers containing removed NFS files are
invalidated in the close() routine.
( SR:8606308183 CR:JAGae71218 )
The NFS_TCP configuration environment variable was
introduced in the 11.00 release of HP-UX to support TCP/IP
NFS connections. This is not needed in the 11.11 release
since TCP/IP is enabled by default.
Resolution:
Removed the references to NFS_TCP in the nfsd(1M) and
setoncenv(1M) man pages.
( SR:8606308342 CR:JAGae71377 )
A non-root user cannot unmount an NFS mounted directory
even when the suid bit is set for the umount executable.
Resolution:
NFS umount code is modified to correctly check for
executable permission.
( SR:8606309647 CR:JAGae72522 )
NIS+ man pages do not contain NIS+ obsolescence information.
Resolution:
Added the obsolescence note for NIS+ in the WARNINGS section
of all the NIS+ man pages.
PHNE_28930:
( SR:8606306693 CR:JAGae69727 )
Some of the XDR routines are doing faulty length checks in
some cases.
Resolution:
Code changes have been made to fix the defect.
PHNE_28568:
( SR:8606308270 CR:JAGae71305 )
If device ID's are enabled, and the file systems are
configured such that the index for NFS PV3 file systems is a
value other than 3, the information for NFS PV3 file systems
will not be displayed when the "nfsstat -m" command is
executed.
Resolution:
Added a function call to get the file system index instead
of setting it to a constant value.
PHNE_28137:
( SR:8606284507 CR:JAGae48453 )
An error ETIMEDOUT is passed to user space even when Kernel
RPC succeeds in getting a message.
Resolution:
The error is reset to 0 if Kernel RPC succeeds in getting a
message.
( SR:8606279504 CR:JAGae43557 )
CacheFS is caching only a part of the required data, but it
marks its internal data structures to reflect that it has
cached all the required data. Hence, CacheFS was caching
partially and resulting in incorrect execution.
Resolution:
CacheFS read page code is modified so that CacheFS marks its
internal data structures to track the actual amount of data
cached.
( SR:8606274703 CR:JAGae38780 )
The functions pathconf(_PC_CHOWN_RESTRICTED) and fpathconf
(_PC_CHOWN_RESTRICTED) of NFS are returning the wrong value
if chown is not restricted.
Resolution:
Created a kernel variable that allows toggling between the
existing behavior and the POSIX conformance. This fix is
disabled by default. To enable the fix, set the NFS kernel
global variable pathconf_chown_restricted_posix to 1 by
patching the vmunix file and rebooting, or by patching the
kernel image in memory. To patch the vmunix file (survives a
reboot), use this command:
echo 'pathconf_chown_restricted_posix?W 0d1' |
adb -w -k /stand/vmunix /dev/kmem
To patch the in-memory kernel (does not survive a reboot),
use this command:
echo 'pathconf_chown_restricted_posix/W 0d1' |
adb -w -k /stand/vmunix /dev/kmem
A kernel regeneration will require these steps to be done
again.
( SR:8606272309 CR:JAGae36449 )
This deadlock lock happens only in loopback mounts. This
happens while adding an entry into directory name lookup
cache (DNLC) for a file. The NFS server attempts to hold a
vnode lock that is already held. This leads to a deadlock
situation.
Resolution:
Avoid DNLC caching of removed NFS files in NFS loopback
mounts.
( SR:8606284964 CR:JAGae48906 )
When copying 32-bit XDR long variables to 64-bit kernel
variables, the upper bits are left the way they were before
the copy.
Resolution:
Added code to make sure the upper 32 bits of 64-bit variable
data is now properly set.
( SR:8606275551 CR:JAGae39627 )
When "nfsstat -m" is executed from the client, a statfs call
is made across the wire to each server listed in the
/etc/mnttab file. The statfs call waits for a response from
the server, so if no response is received, the client hangs.
Resolution:
With the HP-UX Device IDs product (DeviceIDs11i) installed,
the device id information is available in the /etc/mnttab
file on the client and the statfs call is not made across
the wire.
( SR:8606284885 CR:JAGae48827 )
A certain sequence of shared (read) locking and partial
unlocking on overlapping regions by multiple processes over
NFS causes the server's lockd to skip unlocking a portion of
the range in the kernel, thereby losing track of these
locks. These locks accumulate over a period of time and
consequently exhaust all the available lock structures in
the kernel, at which time the kernel can no longer give out
any locks.
Resolution:
The code in lockd dealing with unlocking regions in the
kernel for overlapping shared (read) locks has been
rewritten. The server lockd now correctly identifies non-
overlapping regions and issues unlock requests whenever
necessary.
( SR:8606274519 CR:JAGae38596 )
When the script "nfs.client stop" is run, it unmounts all
the AutoFS mount points. While unmounting the AutoFS mount
points, AutoFS does not verify whether or not a directory
entry under such an AutoFS mount point is busy and therefore
it frees the internal data structures of the corresponding
AutoFS mount point. The system panics when some process
tries to access a directory entry under such an AutoFS mount
point because AutoFS' internal data structures no longer
exist.
Resolution:
AutoFS verifies whether or not the directory entries under
an AutoFS mount point are busy. If busy, the mount point is
not unmounted.
( SR:8606256016 CR:JAGae20331 )
An error condition returned while binding to a reserved port
is not handled in the code.
Resolution:
The error condition returned while binding to a reserved
port is now handled properly.
( SR:8606271824 CR:JAGae36003 )
The system reports "vmunix: bindresvport: loop_twice port
assigned 512" many times in syslog.log, even though there
are reserved ports available.
Resolution:
Provide a kernel variable that by default disables these
messages. To re-enable these messages, the kernel parameter
'loop_twice_msg' must be set to one, by patching the vmunix
file and rebooting, or by patching the kernel image in
memory. To patch the vmunix file (survives a reboot), use
this command:
echo 'loop_twice_msg?W 0d1' |
adb -w /stand/vmunix /dev/kmem
To patch the in-memory kernel (does not survive a reboot),
use this command:
echo 'loop_twice_msg/W 0d1' |
adb -w /stand/vmunix /dev/kmem
A kernel regeneration will require these steps to be done
again.
( SR:8606269865 CR:JAGae34105 )
Intergraph's DiskShare NFS server returns zero bytes for the
preferred and/or maximum transfer size of READ/WRITE, which
makes the HP-UX NFS client panic. The correct behavior
should be to 1) return an I/O error message when the
preferred and/or maximum size has a value of zero, and 2)
allow normal operation when the maximum transfer size has a
non-zero value.
Resolution:
When an NFS server returns zero or MAX_LARGE_FILES for a
READ/WRITE operation's transfer size, the HP-UX client now
responds correctly instead of panicking.
( SR:8606267227 CR:JAGae31470 )
The NFS server does not see any changes made to an exported
file system. Because of this, making an exported read/write
mount point read-only through re-mounting will result in a
system crash in the NFS server stack.
Resolution:
Modified the NFS server to check whether an exported file
system permits alterations before each attempt to modify it,
and to reject all modification requests for a read-only file
system.
( SR:8606276756 CR:JAGae40831 )
When an NFS server receives a pathconf RPC request from a
client, an uninitialized variable may be passed to the
locking routine. This uninitialized variable causes a panic.
Resolution:
Initialized the variable that caused the panic.
( SR:8606267508 CR:JAGae31750 )
AutoFS is allowing a process to write into a read-only file
system. This is because AutoFS is not verifying whether the
underlying file system is read-only or not.
Resolution:
AutoFS is now verifying whether the underlying file system
is read-only or not. If the underlying file system is read-
only then AutoFS returns an error.
( SR:8606267320 CR:JAGae31562 )
There is an NFS performance problem due to bad
compare_cred() logic in nfs3_vnops.c. The current
calculation sizeof(*cr1) - sizeof(cr1->cr_ref) -
sizeof(cr1->cr_pag) yields 100 bytes. However, cr1->cr_pag
is an 8-byte structure in a 64 bit OS. Since it is 8-byte
aligned, there are 4 bytes between cr1->cr_ref and
cr1->cr_pag that were unaccounted for.
Resolution:
Have compare_cred() use 96 bytes for comparison instead of
100 bytes.
( SR:8606287136 CR:JAGae51075 )
Threads in the AutoFS code are not synchronized properly
while one thread is trying to remove a directory and another
thread is sending a mount request for the same directory.
This causes AutoFS to create a fresh directory with null
flags. Therefore AutoFS is unable send a mount request for
the newly created directory.
Resolution:
Measures have been taken to provide a proper synchronization
between the threads in AutoFS that are accessing the same
directory.
( SR:8606278982 CR:JAGae43038 )
The AutoFS daemon automountd serializes mount and unmount
requests. When a mount request for a dead NFS server is
received it takes 75 seconds for automountd to fail the
mount request. During this time no other mount or unmount
requests can be processed, even if the requests are for a
live NFS server.
Resolution:
The number of mount requests sent from the kernel to
automountd has been reduced which will decrease the chance
of automountd getting a request for a dead NFS server. Also,
no mount requests will be sent to automountd without first
checking to see if the NFS server is up.
( SR:8606288334 CR:JAGae52265 )
Since the NFS Client kernel does not provide protocol
information to update the /etc/mnttab file, proto=udp is not
included in the file.
Resolution:
Modified the NFS Client to provide protocol information to
update the /etc/mnttab file when UDP is used.
( SR:8606282498 CR:JAGae46451 )
File descriptors are not closed when the server detects an
RPC_PROGVERSMISMATCH error.
Resolution:
Modified libnsl to close open file descriptors and free
transport handles when an RPC_PROGVERSMISMATCH error is
returned to the client.
( SR:8606288821 CR:JAGae52752 )
Mutex variables are not initialized which causes the calling
process to hang or terminate.
Resolution:
Functions for initializing the mutex variables were added.
( SR:8606282662 CR:JAGae46615 )
While binding a TCP connection, an uninitialized variable
causes the connection to incorrectly be in the LISTEN state.
Resolution:
Initialized the variable to fix the state of the connection.
( SR:8606291642 CR:JAGae55409 )
XDR stream manipulation is not handled correctly in some
cases.
Resolution:
Modified the XDR code to correct the problem.
PHNE_28103:
( SR:8606276016 CR:JAGae40093 )
The rpc.yppasswdd process dumps the core.
Resolution:
rpc.yppasswdd has been modified to correct the problem.
( SR:8606274502 CR:JAGae38579 )
RPC does not function properly.
Resolution:
RPC has been modified to correct the problem.
( SR:8606288394 CR:JAGae52325 )
Incorrect permissions for the log file.
Resolution:
Automount has been modified to set the proper permissions
for the default log file.
( SR:8606281883 CR:JAGae45826 )
When netdir_getbyname() calls ypmatch a hang can result
under certain circumstances.
Resolution:
The logic in ypmatch has been modified.
( SR:8606276334 CR:JAGae40412 )
ypxfrd picks up files with the wrong pathname.
Resolution:
Code changes were made to check for the wrong pathname. Some
restrictions regarding invalid domain and map names are now
enforced. The domain and map names of "." and ".." and those
containing "/" are now rejected.
( SR:8606286563 CR:JAGae50502 )
Incorrect configuration for AutoFS.
Resolution:
The AutoFS code has been modified to address the problem.
( SR:8606274555 CR:JAGae38632 )
Showmount dumps the core with SIGSEGV.
Resolution:
Code has been added to address the problem.
PHNE_27218:
( SR:8606249576 CR:JAGae15966 )
To obtain the port information of client's rpc.lockd, a
server needs to communicate with the client's portmapper.
A file descriptor is allocated on the server and it polls to
check whether replies arrive from the client's portmapper.
For this polling process, the file descriptor should be
registered with the polling list. However, file descriptors
with values that are multiples of 32 do not get registered,
and thus the server does not get the port information.
Resolution:
Fixed the routine that registers the file descriptors.
( SR:8606249559 CR:JAGae15949 )
The '-p' option for rpc.mountd described in the man page is
obsolete.
Resolution:
Modify the man page to indicate that the '-p' option for
rpc.mountd is obsolete.
( SR:8606180770 CR:JAGad49991 )
The client currently does not retransmit the blocked lock
requests. As a result, if the NLM_GRANTED_MSG sent by the
server is lost and it doesn't retry the request, the blocked
lock requests on the client will wait forever.
Resolution:
The client lockd is modified to retransmit the blocked lock
requests. If the NLM_GRANTED_MSG is lost and the server does
not retry then the retransmission of the blocked lock
requests will prevent the hangs on the client.
( SR:8606233931 CR:JAGae03154 )
The parent file system has inconsistent hierarchical mount
points when the automount daemon cannot remount them at the
time the unmount of the parent file system fails. The
unmount of the parent file system can fail if a process is
accessing that file system at the same time the automount
daemon is trying to unmount it. The automount daemon could
not remount the hierarchical mount points due to
unavailability of the remote mount daemon (i.e. the remote
mount daemon is not running).
Resolution:
The automount daemon has been modified so that it informs
the AutoFS client whether or not it could remount the
hierarchical mount points when the unmount of the parent
file system fails. AutoFS has been modified to act based on
the return value from the automount daemon. AutoFS sends a
remount request for the autonode if the automount daemon
could not remount the hierarchical mount points under the
parent file system.
( SR:8606234034 CR:JAGae03255 )
The nfsconfcol script is not removing the target and
temporary data files before writing data to them. If the
script aborts, the files may be wrong.
Resolution:
Remove the target and temporary data files before writing to
them.
( SR:8606256845 CR:JAGae21158 )
The /var/adm/automount.log file is shipped incorrectly.
Resolution:
The /var/adm/automount.log file is now shipped properly.
( SR:8606256846 CR:JAGae21159 )
The /var/spool/rexd directory is shipped incorrectly.
Resolution:
The /var/spool/rexd directory is now shipped properly.
( SR:8606259831 CR:JAGae24149 )
When rpc.lockd starts after a crash, it sends a crash
notification to rpc.statd. If rpc.statd is not yet
registered with rpcbind, rpc.lockd does not re-try the crash
notification, and crash recovery does not take place. This
causes apparent hangs to clients trying to lock files.
Resolution:
Modified the rpc.lockd code to retry the crash notification
to rpc.statd 20 times, due to the delay in rpc.statd
registration.
( SR:8606255006 CR:JAGae19335 )
While performing multiple concurrent exportfs operations,
the /etc/xtab file's modification time remains the same.
This leads to incorrect mount and showmount output, and at
least one of the exported directories cannot all be mounted.
Resolution:
Changed the logic so that rpc.mountd considers both the
modification time and the file size of the /etc/xtab file to
update the output of the mount and showmount commands.
( SR:8606277106 CR:JAGae41178 )
There is a race condition in the AutoFS kernel code where
two threads executing the same code can cause one of the
threads to wait forever. The thread that waits forever
causes the AutoFS mounted file system to be busy,
preventing it from ever being unmounted.
Resolution:
The AutoFS kernel code was modified to prevent this race
condition. The race condition is avoided by changing the
conditions for which a thread will block when executing
this particular AutoFS kernel code.
( SR:8606257688 CR:JAGae21991 )
If a kill signal is delivered to a thread sending an unlock
request on a single CPU machine, the thread enters a tight
loop in order to clean up the locking information on the
server. This causes the thread to hold the CPU and the
machine hangs.
Resolution:
When the interrupt error (EINTR) returns to the kernel lock
manager due to signals, the kernel lock manager makes the
next RPC call uninterruptible, which causes threads to sleep
during the RPC calls and releases the CPU.
( SR:8606260873 CR:JAGae25195 )
The number of bytes to be written to a file is stored as a
signed long value. On 32-bit clients, this value overflows
when a file is greater than 2.5 GB, and the file becomes
corrupted.
Resolution:
Modified the code so that variable overflow does not occur.
( SR:8606242021 CR:JAGae09276 )
1. If there are multiple mount requests for a non-existent
server, then the automount daemon tries to move these
pending requests to cache. When this operation causes the
cache size to exceed a certain threshold value, the
automount daemon prints "dup_remove_pending: WARNING:
duplicate request cache size (210) > XID_CACHE_SIZE (200),
continuing" to the syslog.log file
2. When a mount request fails, the automount daemon first
removes the child node, and then tries to remove the parent
node. If the parent node is mounted or if there are other
files in the parent directory, the automount daemon logs
"Rmdir: err=247, mnt='<mount point-directory>'" to the
/var/adm/automount.log file.
Resolution:
1. The trace level must now be set to 10 (instead of greater
than 0) to log the message "dup_remove_pending: WARNING:
duplicate request cache size (210) > XID_CACHE_SIZE (200),
continuing" to /var/adm/automount.log. It will be logged to
/var/adm/syslog/syslog.log only if the verbose option is
enabled.
2. The trace level must now be set to 10 (instead of greater
than 0) to log the message "Rmdir: err=247,
mnt='<mount point-directory>'" to /var/adm/automount.log.
( SR:8606226778 CR:JAGad95840 )
After a manual unmount, two accesses take place
simultaneously. One access is forced to initiate a mount,
and before that mount completes, the other access
dereferences a pointer that hasn't been set yet. This
causes a data page fault which results in the panic.
Resolution:
Added a condition to check for a mount-in-progress which
makes the second access follow an existing code path to wait
for the first mount to complete.
( SR:8606248269 CR:JAGae14669 )
If the /var/statmon/sm contents are destroyed on the client,
the client's rpc.statd will not know which servers to notify
that the client has crashed and recovered. Since the server
never receives notification of the client system reboot, it
will continue holding onto locks for this client that are
now "stale". The server will also send any replies to new
lock requests to the old lockd port number on the client. In
most cases, the new instance of lockd on the client will not
be listening on the same port as the previous instance so it
will not receive these lock replies. The result is that
future lock requests from this client will hang, causing the
applications issuing these lock requests to hang.
Resolution:
The clear_locks(1M) command is provided to allow a root user
to forcibly notify the NFS server's rpc.lockd to discard all
locks and cached port information for a specified client.
This tool should only be used in cases where a client system
experienced a state change (i.e. re-installed or had
rpc.lockd or rpc.statd killed and re-started) and was not
able to successfully notify the server system using the
normal lockd/statd recovery mechanism.
( SR:8606245329 CR:JAGae11799 )
When a server does not support a PV3 link system call from a
client, the server returns an error message which hangs or
panics the client system.
Resolution:
When the server returns the error message for the PV3 link
system call, the client now releases a semaphore to avoid
the deadlock/hang.
( SR:8606264317 CR:JAGae28647 )
This product update contains minor enhancements required to
enable the HP-UX Device IDs feature.
When AutoFS unmounts a filesystem, it obtains the device id
from the filesystem server. If that server is slow, or not
responding, there may be a delay of 15 seconds or more. On
systems with many AutoFS managed filesystems and many slow
or non-responding servers, the cumulative delay may be quite
significant.
Resolution:
HP-UX is implementing Device IDs in the /etc/mnttab file.
AutoFS will now make use of the Device IDs and not query
each server when performing unmounts.
( SR:8606265228 CR:JAGae29556 )
System panics if all the entries in duplicate cache are
marked DUP_INPROGRESS, when a new duplicate cache request
arrives after reaching cache maximum limit.
Resolution:
Duplicate cache requests will return an error if cache
reaches maximum limit and all entries are marked
DUP_INPROGRESS.
( SR:8606259713 CR:JAGae24031 )
AutoFS problem.
Resolution:
Fixed the AutoFS problem.
( SR:8606234968 CR:JAGae04152 )
The unmount module of automountd is unable to find the
difference between the mount points with ESTALE errors and
the mount points that are to be unmounted. As a result,
apart from unmounting the mount points that are to be
unmounted, mount points with ESTALE errors are also
unmounted. Because of this, corresponding direct AutoFS
entries for those mount points are removed from the file
/etc/mnttab.
Resolution:
In automountd, the unmount module, which finds the mount
points that are to be unmounted, is modified so that it is
able to find the difference between the mount points with
ESTALE errors and the mount points that are to be unmounted.
( SR:8606235172 CR:JAGae04352 )
During an NFS mount operation, the client performs a bind
operation. When all ports are in use, the error message
returned is "system error" (t_errno = 8) instead of "address
busy" (t_errno = 23).
Resolution:
Return the error message "address busy" instead of "system
error" when all ports are busy.
( SR:8606245876 CR:JAGae12335 )
When an automount daemon's mount request fails, it counts
the number of slashes in the exported file system (e.g.
/dir/) |
