Re: LIVEDUMP

I hope some of you are using Editor and loading dumps in the context of
source code these days! It's not very often I run DA interactively
(equating TAPEIN). FYI, LIVEDUMP came about as a side-effect of analyzing
a FASTDUMP before it was unloaded (FASTDUMP is the mechanism where a
snapshot of memory is taken in a reserved memory area). DA/SDA was extended
to analyze this copy from memory. and then it was an easy step to say, what
happens if we analyze LIVE memory instead of a copy of it. The result was
LIVEDUMP.


"Andy Mountford" <andrew.mountford@xxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:3lsg235jjm5h4fkmj45s9pjifs5sq10jkh@xxxxxxxxxx
On 19 Apr 2007 05:33:32 -0700, "Roy.N.Latham@xxxxxxxxx"
<Roy.N.Latham@xxxxxxxxx> wrote:

On Apr 18, 5:46 pm, s...@xxxxxxxxxxxxx (Scott Lurndal) wrote:
MarkPe...@xxxxxxxxx writes:
I see that the A-Series (aka Large Systems and other names) MCP dump
analyzer now has a LIVEDUMP feature that peeks into the memory of a
running system. It's interesting that Ian Farbrother (?) and Roy
Latham had this working rather nicely on B1000 MCP some 22 years ago
(or a bit more).

medium systems had 0DM long before that (dump memory for mix-id 0, ie.
the MCP).

scott

0DM takes (took?) a snapshot of memory to disk and both B1000 and A-
Series had/have equivalent commands.

What Mark is referring to is just for the B1000 dump analyzer (IDA by
that time) to look out side its own memory space at the running
system. This meant that one could take a peek without the taking the
time hit to take a dump.

On A-Series you can of course use <mix> OT offset and if you know the
program well enough, a critical variable could be monitored. We used
that at S.W.I.F.T. one day when we knew from an earlier failure some
variable was going to get too big during the day. Luckily, it was on
the stack. We were able to monitor until a quiet time and then swover.
(Due to the nature of the problem, we were in no rush to change
systems. On the other hand, we didn't want to get the failure around a
time banks were busing closing/balancing books for their day.) Deepak
and I got a stack history monitor working - from a mix number, show
the current procedure call history at low cost.

Back on the B1000, several programs did the read outside of bounds
trick to get the time tick counter. RD and the system/monitor program
did too. All I did was put the pieces together one day in the English
Department in Goleta in the company of Ian and possibly Kelly.

By the way, we were not pusuing lofty goals in the English Department.
The bar had been opened by a professor of English at UCSB and it used
to have a poetry night in the back bar.

When the penny finally dropped with the faculty that the same person
was running both English Departments, Bob got an ultimatum. He did the
decent thing - resigned at the University. I think he saw more of his
students in the department he kept.

Roy

Roy,

The A-series has caught up somewhat as Mark was mentioning. If you
label-equate TAPEIN to LIVEDUMP, a privileged user can get to look at
the running system without taking a dump and it has proved invaluable.

The capability was introduced in 51.1 and you need ICs for MCP,
SDASUPPORT and DUMPANALYZER. At the moment, even a privileged user
cannot use the feature via a standard SDASUPPORT: you need either to
make SDASUPPORT privileged or use IGSDASUPPORT. (This limitation is
being addressed.)

I've used it to debug DA itself. Run a DA LIVEDUMP session during
which you cannot see yourself - SDASUPPORT takes the same snapshot at
initialization time as it does with an ordinary memory dump and this
happens before DA is up - and then run *another* LIVEDUMP session to
look at the first. This can get confusing...!

Andy


.

Relevant Pages

  • Re: Stop 0x50 occuring after Exch SP1 and e-mail bomb
    ... The results of my crash dump!analyze -v are at the end of this ... emptied and the queue folder was also empty. ... I only have one full dump to work with now. ... Invalid system memory was referenced. ...
    (microsoft.public.windows.server.sbs)
  • Re: Crashdump Win2k Server
    ... I'm afraid that mini dump is too limited to analyze and obtain enough ... I suggest that you config "Write debugging information" to Complete memory ...
    (microsoft.public.windows.server.general)
  • Re: LIVEDUMP
    ... medium systems had 0DM long before that (dump memory for mix-id 0, ... label-equate TAPEIN to LIVEDUMP, a privileged user can get to look at ... the running system without taking a dump and it has proved invaluable. ...
    (comp.sys.unisys)
  • Re: LIVEDUMP
    ... LIVEDUMP came about as a side-effect of analyzing ... a FASTDUMP before it was unloaded (FASTDUMP is the mechanism where a ... snapshot of memory is taken in a reserved memory area). ... happens if we analyze LIVE memory instead of a copy of it. ...
    (comp.sys.unisys)
  • Daily Report #4852
    ... Verify Guide Star Acquisition with Continuing FGSs ... Load and Dump Onboard Memory ... At the beginning of each test, the attitude control law ...
    (sci.astro.hubble)