[Info-vax] TCPIP$SMTP SYSTEM-F-EXQUOTA, SYSTEM-F-ACCVIO after upgrade to V8.3

[Steven Schweda]

   The system disk on my main system ("HP TCP/IP Services for
OpenVMS Alpha Version V5.4 - ECO 7 | on a COMPAQ Professional
Workstation XP1000 running OpenVMS V7.3-2") has started
dropping briefly into mount verification several times per
day, so this seemed like a good opportunity to do that VMS
V8.3 upgrade I've been postponing for a long time.

   So, I INITIALIZEd a new system disk (/STRUCTURE = 5),
booted off the V8.3 CD-ROM, did a BACKUP /IMAG /NOINIT /TRUN
/VERI from the old disk to the new one, and then did the OS
upgrade.  No obvious errors.

   The new BIND server goes berserk if it sees a name with an
"_" in it, but after replacing all those with "-", I got the
DNS working again.  However, the SMTP receiver fails every
time, with either %LIB-E-ACTIMAGE / -SYSTEM-F-EXQUOTA or
%SYSTEM-F-ACCVIO, depending.  Using the local name/address:

alp $ telnet /po = 25 alp-l
%TELNET-I-TRYING, Trying ... 10.0.0.9
%TELNET-I-SESSION, Session 01, host alp-l, port 25
-TELNET-I-ESCAPE, Escape character is ^]
%TELNET-S-REMCLOSED, Remote connection closed
-TELNET-I-SESSION, Session 01, host alp-l, port 25

ALP $
%%%%%%%%%%%  OPCOM  20-OCT-2009 23:01:44.03  %%%%%%%%%%%
Message from user INTERnet on ALP
INTERnet ACP SMTP Accept Request from Host: 10.0.0.9 Port:
49160

ALP $
%%%%%%%%%%%  OPCOM  20-OCT-2009 23:01:44.66  %%%%%%%%%%%
Message from user TCPIP$SMTP on ALP
-SYSTEM-F-EXQUOTA, process quota exceeded


   Using the outside name/address (via the DSL modem+router):

alp $ telnet /po = 25 alp
%TELNET-I-TRYING, Trying ... 209.98.249.184
%TELNET-I-SESSION, Session 01, host alp, port 25
-TELNET-I-ESCAPE, Escape character is ^]
%TELNET-S-REMCLOSED, Remote connection closed
-TELNET-I-SESSION, Session 01, host alp, port 25

ALP $
%%%%%%%%%%%  OPCOM  20-OCT-2009 23:02:20.81  %%%%%%%%%%%
Message from user INTERnet on ALP
INTERnet ACP SMTP Accept Request from Host: 209.98.249.184
Port: 17982

ALP $
%%%%%%%%%%%  OPCOM  20-OCT-2009 23:02:21.22  %%%%%%%%%%%
Message from user TCPIP$SMTP on ALP
%SYSTEM-F-ACCVIO, access violation, reason mask=00, virtual
address=0000000000000000, PC=0000000000031BCC, PS=0000001B


   These both work with the old system disk.

   The first time I tried all this ("What could go wrong?"), I
hit it with all the current patch kits (PCSI, UPDATE, SYS,
misc, ..., TCPIP ECO 4) before I tried to do any useful work.
Seeing that fail, I tried again with no patches (same
failures), and again after only the TCPIP ECO (same failures).
On the off chance that the UAF quotas for TCPIP$SMTP really
were too low, I tried increasing BIOlm: 18 -> 128, DIOlm: 18
-> 128, ASTlm: 100 -> 128, TQElm: 15 -> 128, Bytlm: 52200 ->
128000, with no effect.  I assume that it's trying to malloc()
2GB or something, but I couldn't prove it.

   Because the symptom changes, depending on which address is
used, I wonder about the DNS, but (after removing all the "_"
characters from the names), it all seems to work as I expect,
forward and back, on all the names and addresses.

   The TCPIP$SMTP_RECV_RUN.LOG files don't seem to add much:

$ Set NoOn
$ VERIFY = F$VERIFY(F$TRNLNM("SYLOGIN_VERIFY"))

%%%%%%%%%%%%                   20-OCT-2009 23:28:54.13  %%%%%%%%%%%%
%LIB-E-ACTIMAGE, error activating image
 ALP$DKA0:[SYS0.SYSCOMMON.][SYSLIB]UCX$IPC_SHR.EXE
-SYSTEM-F-EXQUOTA, process quota exceeded

%LIB-E-ACTIMAGE, error activating image !AS
  TCPIP$SMTP   job terminated at 20-OCT-2009 23:28:54.18
  Accounting information:
  Buffered I/O count:                 98      Peak working set
size:      15536
  Direct I/O count:                   66      Peak virtual
size:         208880
  Page faults:                     14351      Mounted
volumes:                0
  Charged CPU time:        0 00:00:00.23      Elapsed time:       0
00:00:00.62


$ Set NoOn
$ VERIFY = F$VERIFY(F$TRNLNM("SYLOGIN_VERIFY"))

%%%%%%%%%%%%                   20-OCT-2009 23:28:58.61  %%%%%%%%%%%%
%SYSTEM-F-ACCVIO, access violation, reason mask=00, virtual
 address=0000000000000000, PC=0000000000031BDC, PS=0000001B

%SYSTEM-F-ACCVIO, access violation, reason mask=00, virtual
 address=000000000000000C, PC=0000000000000012, PS=00000003
  TCPIP$SMTP   job terminated at 20-OCT-2009 23:28:58.66
  Accounting information:
  Buffered I/O count:                 93      Peak working set
size:      17936
  Direct I/O count:                   68      Peak virtual
size:         208240
  Page faults:                     14566      Mounted
volumes:                0
  Charged CPU time:        0 00:00:00.25      Elapsed time:       0
00:00:00.42


   I'm hoping to avoid a long series of slow experiments which
cripple my main system, so if anyone has any insight, I'm open
to suggestions.  (And back to V7.3-2.)

   I remember when stuff like this just worked.