Good to hear that you found the cause! Our own testing has mostly
focused on 64-bit systems, so that could be the reason why we haven't
noticed this particular bug. If there isn't a ticket for the problem
yet, please write one.
OpenSAF is moving towards a continuous delivery model which means that
we can make new OpenSAF deliveries with less effort than before. The
next planned delivery is early October, but if you feel that this
bug-fix is urgent then we can deliver a new version of OpenSAF as soon
as the bug-fix is in and all tests have passed.
Regarding Zoran's comment: yes you are right that revents is an output
variable that shouldn't have to be initialized before the call to
poll(). I think the actual problem here is that the nfds argument to
poll() is variable and if we don't have a CLM fd then
fds[FD_CLM].revents is uninitialized (not touched by poll()). I usually
prefer to keep nfds fixed and set fds[FD_CLM].fd to -1 in case we don't
yet have a CLM fd, in which case fds[FD_CLM].revents will always be
written by poll(). With the current approach of having a variable nfds
the following condition needs to change:
if (fds[FD_CLM].revents & POLLIN)
Instead if should be:
if (nfds == 5 && (fds[FD_CLM].revents & POLLIN))
regards,
Anders Widell
On 08/07/2017 12:41 PM, Zoran Milinkovic wrote:
Hi James,
Can you be more specific with "OpenSAF hasn't been started successfully" ?
When clmDispatch returns ERR_BAD_HANDLE, IMMND logs messages with warning and
notice log level.
In general in all services, when *Dispatch functions fail with ERR_BAD_HANDLE
error code, it's because the handle might be corrupted, finalized, or the
operation cannot be done (for example, a callback with a long DN parameter to a
client where long DN is not supported, etc...)
Setting revents parameter to 0 before the poll call does not make sense to me.
Revents is the output parameter, and I don't see how your proposal can be a
solution to your problem.
Thanks,
Zoran
-----Original Message-----
From: Carroll, James R [mailto:[email protected]]
Sent: den 5 augusti 2017 17:45
To: [email protected]
Subject: Re: [users] opensaf 5.2, not starting on 32-bit linux platform
Hi,
I never had any feedback on this, which is a little disheartening. In any
case, we continued to troubleshoot the issue, and it appears there is an
uninitialized variable in the source code that is causing the error.
The file is immnd_main.c, and the line number is 355.
We added the line: fds[FD_CLM].revents = 0;
This resolved our issue, and we were able to bring up opensaf 5.2. For the
record, we were never able to bring up opensaf 5.2, without this fix, on our
32-bit systems. We did also test this out on our 64-bit systems, and found
that we were about 50/50 when staring OpenSAF - sometimes is started, and
sometimes it failed, without this fix. Once this fix was applied, both our
32-bit and 64-bit systems started consistently.
Is there an open ticket for this issue? Also, with the just announced release
of OpenSaf 5.2.1, will there be an update to capture this?
How are high-priority fixes applied to OpenSAF?
Thanks.
Jim
From: Carroll, James R (US)
Sent: Friday, July 21, 2017 5:07 PM
To: [email protected]
Subject: opensaf 5.2, not starting on 32-bit linux platform
Hi,
We have recently built and upgraded to opensaf 5.2. We were previously running
using opensaf 5.0.
We compiled opensaf as 32-bit applications. We are running on a 32-bit Redhat
OS (v6.5).
We have been unable to start opensaf successfully, it appears to be a timing
issue between IMM and CLM.
The code that is failing appears to be a new feature added to 5.2, reference
the ticket number below.
The file is immnd_main.c, line number 435, failing on saClmDispatch() call. We
are getting SA_AIS_ERR_BAD_HANDLE_ERROR
Ticket Information:
1640 - imm agent changes for integrating IMM with CLMS
Has anyone built opensaf 5.2 as a 32-bit application, and been able to run it
successfully?
Thanks.
Jim
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most engaging tech
sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Opensaf-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/opensaf-users
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Opensaf-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/opensaf-users
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Opensaf-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/opensaf-users
