#2889 imm: avoid BAD_HANDLE result on requests after timeout on CCB apply

[Vu Minh Nguyen]

If CCB apply (IMMND_EVT_A2ND_CCB_APPLY) gets TIMEOUT (e.g: PBE is hung during that time), here is what will be happening at IMMND and IMMA:

IMMND purges that synchronous request by doing following main jobs:
1) Attemp to abort CCB even the CCB state is in critical.
2) Remove client node from client dabase.

At IMMA, inside saImmOmCcbApply() call, imma_proc_recover_ccb_result() is invoked to get CCB result by sending IMMND_EVT_A2ND_RECOVER_CCB_OUTCOME to the IMMND. After the CCB is commited, imma_proc_recover_ccb_result() returns SA_AIS_OK to saImmOmCcbApply() and then SA_AIS_OK to user.

Later on, when user performs any IMM call on that IMM OM handle such as saImmOmFinalize() will result in BAD_HANDLE as the client node has already removed in #2 above.

See syslog and imma trace from the attachment.

[Vu Minh Nguyen]

https://sourceforge.net/p/opensaf/mailman/message/36361509/

[Vu Minh Nguyen]

If user does recovery the handle in this case, then no functionality is impacted. But if there are choices about getting SA_AIS_OK or getting BAD_HANDLE in subsequence call,
user probably prefer to the former one if there is no consequence/issue at all.

So, change this ticket to enhancement instead.

[Vu Minh Nguyen]

commit ede5191f9caf41836a65acaffd648e7ac0b00590 (HEAD -> develop, origin/develop, ticket-2889)
Author: Vu Minh Nguyen vu.m.nguyen@dektech.com.au
Date: Fri Jul 27 09:37:28 2018 +0700

imm: avoid BAD_HANDLE result on requests after timeout on CCB apply [#2889]

When CCB apply (IMMND_EVT_A2ND_CCB_APPLY) gets TIMEOUT, IMMND will be
informed the timeout via msg type IMMND_EVT_A2ND_CL_TIMEOUT, then
doing following things:
1) Attemp to abort CCB even the CCB state is in critical.
2) Remove client node from client dabase.

Once the CCB has been comitted, saImmOmCcbApply() returns SA_AIS_OK to user.
As the client is removed at #2 above, any IMM operations on the IMM OM handle
such as saImmOmCcbObjectDelete() or saImmOmFinalize() will result in BAD_HANDLE.

With this enhancement, the client having critical CCB will not be purged
when timeout happens at IMM sync requests.