Operating System - HP-UX
1748168 Members
4055 Online
108758 Solutions
New Discussion юеВ

Re: Could Memory corruption happen in HPUX10.0

 
Dias George
Occasional Advisor

Could Memory corruption happen in HPUX10.0

I have a process which takes data and puts it into a buffer , which is then picked up by another process. The issue here is that the data is put into the buffer fine but when it is picked back , there is some corruption happening ... i will show you an example ...we put in DESMIADT08T .. but when we retrieve FESMIADT08T is what we get . Also the second process core dumps . Could there be a issue memory or the disk . Please advise
10 REPLIES 10
TTr
Honored Contributor

Re: Could Memory corruption happen in HPUX10.0

It is more likely a programming issue with the memory addressing used to exchange the data. Does this happen consistently if you re-run the two processes? What is the core dump reason of the second process' core when you run "file core"?
Hardware memory corruptions usually are captured at the hardware level as single bit errors or page deallocation tables.

What memory mechanism do you use to exchange the data?
Dias George
Occasional Advisor

Re: Could Memory corruption happen in HPUX10.0

Yes it happens consistently, it never had an issue for 15 years but it started recently and nothing new has come up or changed. This process runs on 6 servers but only one server core dumps . File core says segmentation fault , I tried to use GDB but could not get the stack trace.
Steven E. Protter
Exalted Contributor

Re: Could Memory corruption happen in HPUX10.0

Shalom,

Use cstm or mstm or xstm if part of the OS to physically test memory.

If it checks out without problem, then begin to suspect your software is a contributing factor.

You could have some pointer or other programming problems creating the core dump.

To answer your question, yes, memory corruption is possible, but if you can not independently create the problem without your software, then suspect your software.

SEP
Steven E Protter
Owner of ISN Corporation
http://isnamerica.com
http://hpuxconsulting.com
Sponsor: http://hpux.ws
Twitter: http://twitter.com/hpuxlinux
Founder http://newdatacloud.com
TTr
Honored Contributor

Re: Could Memory corruption happen in HPUX10.0

> Yes it happens consistently

Do you always get the "DES..." changed to "FES..."? The reason I asked was that there is no guarantee that you will get the same exact memory hardware address each time you run the two processes. If that's the case most likely it is not hardware related.
What does "file core" show as the reason for the core dump?
> it never had an issue for 15 years but it started recently and nothing new has come up or changed...
When something like this happens without any system changes it is usually data related, a data type changed or a threshold was exceeded or something within the data itself.
Dias George
Occasional Advisor

Re: Could Memory corruption happen in HPUX10.0

core: core file from 'RTNUP' - received SIGSEGV

Not like that actually sometimes the first character is replaced by F ,otherwise more than one character but always the starting character is replaced by F.
Dias George
Occasional Advisor

Re: Could Memory corruption happen in HPUX10.0

One more thing , this error is not produced on the same data always . Always the different data , the pattern here is the last data in the buffer gets affected whatever it is.
Dennis Handly
Acclaimed Contributor

Re: Could Memory corruption happen in HPUX10.0

Is this really 10.20? This OS version isn't supported.

>buffer, which is then picked up by another process.

Please be more specific. Is this shared memory or a mapped file?
What are you using for synchronization?

>Also the second process core dumps.

Is this related to the corrupted string or something else?

>tried to use GDB but could not get the stack trace.

What did you get instead? What version of gdb do you have?
Dias George
Occasional Advisor

Re: Could Memory corruption happen in HPUX10.0

Please be more specific. Is this shared memory or a mapped file?
What are you using for synchronization?

> it is shared memory

Is this related to the corrupted string or something else?

> I think yes .. the second process just logs the data it retrieves using pointers to generate a report , now the data is picked up from shared memory put in by the first process

What did you get instead? What version of gdb do you have?

> got nothing actually , we used the debug executable and tried to use gdb and adb to debug, but we got ????? in place of function names , in adb we got something like stack could not be unwind . Everything is outdated , like you said the OS too , gdb version is HP gdb 1.1 and i cannot update it too ....so only have to use what I have


Dennis Handly
Acclaimed Contributor

Re: Could Memory corruption happen in HPUX10.0

>gdb version is HP gdb 1.1 and i cannot update it too

The oldest WDB archive is 5.0 for 11.00.
You might be able to copy the corefile, executable and all of the shlibs and dld.sl to a 11.11 box and debug?