Hi Alex,
Reviewed and tested the patch.
Ack.
/Neel.
On 2016/10/07 11:10 PM, Alex Jones wrote:
> osaf/services/saf/smfsv/smfd/SmfStepState.cc | 13 ++++
> osaf/services/saf/smfsv/smfd/SmfUpgradeProcedure.cc | 68
> +++++++++++++++++---
> osaf/services/saf/smfsv/smfd/SmfUpgradeProcedure.hh | 6 +
> 3 files changed, 76 insertions(+), 11 deletions(-)
>
>
> Sep 27 00:34:14 q50-s1 osafsmfd[6667]: NO SA_AMF_ADMIN_SI_SWAP [rc=1]
> successfully initiated
> Sep 27 00:34:15 q50-s1 osafimmnd[6571]: NO ERR_BAD_OPERATION: Mismatch on
> administrative owner '' != 'SMFSERVICE'
> Sep 27 00:34:17 q50-s1 osafsmfd[6667]: NO Fail to invoke admin operation,
> rc=SA_AIS_ERR_BAD_OPERATION (20). dn=[safSi=SC-2N,safApp=OpenSAF], opId=[7]
> Sep 27 00:34:17 q50-s1 osafsmfd[6667]: NO Admin op SA_AMF_ADMIN_SI_SWAP fail
> [rc = 20]
> Sep 27 00:34:17 q50-s1 osafsmfd[6667]: NO CAMP: Procedure
> safSmfProc=RollingUpgrade returned FAILED
> Sep 27 00:36:14 q50-s1 osafsmfd[6667]: NO Campaign thread does not disappear
> within 120 seconds after SA_AMF_ADMIN_SI_SWAP, the operation was assumed
> failed.
> Sep 27 00:36:14 q50-s1 kernel: [14934029.531187] osafsmfd[32024]: segfault at
> 4 ip 00000000004425b6 sp 00007f67f7ffe1c0 error 4 in osafsmfd[400000+9a000]
> Sep 27 00:36:14 q50-s1 osafamfnd[6649]: NO
> 'safComp=SMF,safSu=SC-1,safSg=2N,safApp=OpenSAF' faulted due to 'avaDown' :
> Recovery is 'nodeFailfast'
> Sep 27 00:36:14 q50-s1 osafamfnd[6649]: ER
> safComp=SMF,safSu=SC-1,safSg=2N,safApp=OpenSAF Faulted due to:avaDown
> Recovery is:nodeFailfast
>
> There are a few problems here. One is that the SmfSwapThread is pointing to a
> deleted procedure when the original active controller is reassigned active.
> The
> second problem is that a new SmfSwapThread is created when the original active
> controller is reassigned active, so now there are two running. The first
> thread
> tries to use its proc pointer (which has been deleted when the original active
> goes to quiesced) and causes the segfault.
>
> The proposed solution is a little different from that proposed in the ticket
> description. This solution proposes to use the existence of the SmfSwapThread
> as
> a test. When the original active controller is reassigned active because the
> si-swap failed, it will still remove the RestartIndicator as it does now. But,
> if the SmfSwapThread is still running, it will not create a new one, but
> update
> it with the recreated procedure pointer, and let it handle the si-swap
> timeout.
> Then it will report the error. I believe this solution is backwards compatible
> because no IMM changes are made like the ones proposed in the ticket.
>
> diff --git a/osaf/services/saf/smfsv/smfd/SmfStepState.cc
> b/osaf/services/saf/smfsv/smfd/SmfStepState.cc
> --- a/osaf/services/saf/smfsv/smfd/SmfStepState.cc
> +++ b/osaf/services/saf/smfsv/smfd/SmfStepState.cc
> @@ -424,6 +424,19 @@ SmfStepStateUndone::execute(SmfUpgradeSt
> {
> TRACE_ENTER();
>
> + if (i_step->calculateStepType() != SA_AIS_OK) {
> + LOG_ER("SmfStepStateUndone: Failed to calculate step type");
> + changeState(i_step, SmfStepStateFailed::instance());
> + TRACE_LEAVE();
> + return SMF_STEP_FAILED;
> + }
> +
> + if (i_step->getSwitchOver() == true) {
> + TRACE("Switch over is needed in this step");
> + TRACE_LEAVE();
> + return SMF_STEP_SWITCHOVER;
> + }
> +
> i_step->setRetryCount(0); /* Reset the retry counter */
> changeState(i_step, SmfStepStateExecuting::instance());
>
> diff --git a/osaf/services/saf/smfsv/smfd/SmfUpgradeProcedure.cc
> b/osaf/services/saf/smfsv/smfd/SmfUpgradeProcedure.cc
> --- a/osaf/services/saf/smfsv/smfd/SmfUpgradeProcedure.cc
> +++ b/osaf/services/saf/smfsv/smfd/SmfUpgradeProcedure.cc
> @@ -482,12 +482,17 @@ SmfUpgradeProcedure::switchOver()
> osafassert(0);
> }
>
> - TRACE("SmfUpgradeProcedure::switchOver: Create the restart indicator");
> -
> SmfCampaignThread::instance()->campaign()->getUpgradeCampaign()->createSmfRestartIndicator();
> -
> - SmfSwapThread *swapThread = new SmfSwapThread(this);
> - TRACE("SmfUpgradeProcedure::switchOver, Starting SI_SWAP thread");
> - swapThread->start();
> + if (!SmfSwapThread::running()) {
> + TRACE("SmfUpgradeProcedure::switchOver: Create the restart
> indicator");
> +
> SmfCampaignThread::instance()->campaign()->getUpgradeCampaign()->createSmfRestartIndicator();
> +
> + SmfSwapThread *swapThread = new SmfSwapThread(this);
> + TRACE("SmfUpgradeProcedure::switchOver, Starting SI_SWAP
> thread");
> + swapThread->start();
> + } else {
> + TRACE("SmfUpgradeProcedure::switchOver, SI_SWAP thread already
> running");
> + SmfSwapThread::setProc(this);
> + }
>
> TRACE_LEAVE();
> }
> @@ -4156,6 +4161,31 @@ SmfUpgradeProcedure::resetProcCounter()
> /* Static methods */
> /*====================================================================*/
>
> +SmfSwapThread *SmfSwapThread::me(0);
> +std::mutex SmfSwapThread::m_mutex;
> +
> +/**
> + * SmfSmfSwapThread::running
> + * Is the thread currently running?
> + */
> +bool
> +SmfSwapThread::running(void)
> +{
> + std::lock_guard<std::mutex> guard(m_mutex);
> + return me ? true : false;
> +}
> +
> +/**
> + * SmfSmfSwapThread::setProc
> + * Set the procedure pointer to the newly created procedure
> + */
> +void
> +SmfSwapThread::setProc(SmfUpgradeProcedure *newProc)
> +{
> + std::lock_guard<std::mutex> guard(m_mutex);
> + me->m_proc = newProc;
> +}
> +
> /**
> * SmfSmfSwapThread::main
> * static main for the thread
> @@ -4181,6 +4211,8 @@ SmfSwapThread::SmfSwapThread(SmfUpgradeP
> m_proc(i_proc)
> {
> sem_init(&m_semaphore, 0, 0);
> + std::lock_guard<std::mutex> guard(m_mutex);
> + me = this;
> }
>
> /**
> @@ -4188,6 +4220,8 @@ SmfSwapThread::SmfSwapThread(SmfUpgradeP
> */
> SmfSwapThread::~SmfSwapThread()
> {
> + std::lock_guard<std::mutex> guard(m_mutex);
> + me = 0;
> }
>
> /**
> @@ -4309,13 +4343,25 @@ SmfSwapThread::main(void)
>
> exit_error:
> if (SmfCampaignThread::instance() != NULL) {
> - SmfProcStateExecFailed::instance()->changeState(m_proc,
> SmfProcStateExecFailed::instance());
> - }
> -
> - if (SmfCampaignThread::instance() != NULL) {
> + std::lock_guard<std::mutex> guard(m_mutex);
> +
> + SmfProcStateExecuting::instance()->changeState(m_proc,
> SmfProcStateStepUndone::instance());
> +
> + // find the failed upgrade step and set it to Undone
> + std::vector<SmfUpgradeStep *>&
> upgradeSteps(m_proc->getProcSteps());
> + for (std::vector<SmfUpgradeStep *>::iterator
> it(upgradeSteps.begin()); it != upgradeSteps.end(); ++it) {
> + if ((*it)->getSwitchOver()) {
> +
> (*it)->changeState(SmfStepStateUndone::instance());
> + break;
> + }
> + }
> +
> + std::string error("si-swap of middleware failed");
> + SmfCampaignThread::instance()->campaign()->setError(error);
> +
> CAMPAIGN_EVT *evt = new CAMPAIGN_EVT();
> evt->type = CAMPAIGN_EVT_PROCEDURE_RC;
> - evt->event.procResult.rc = SMF_PROC_FAILED;
> + evt->event.procResult.rc = SMF_PROC_STEPUNDONE;
> evt->event.procResult.procedure = m_proc;
> SmfCampaignThread::instance()->send(evt);
> }
> diff --git a/osaf/services/saf/smfsv/smfd/SmfUpgradeProcedure.hh
> b/osaf/services/saf/smfsv/smfd/SmfUpgradeProcedure.hh
> --- a/osaf/services/saf/smfsv/smfd/SmfUpgradeProcedure.hh
> +++ b/osaf/services/saf/smfsv/smfd/SmfUpgradeProcedure.hh
> @@ -29,6 +29,7 @@
> #include <vector>
> #include <list>
> #include <map>
> +#include <mutex>
>
> #include <saSmf.h>
> #include <saImmOi.h>
> @@ -791,7 +792,12 @@ class SmfSwapThread {
> ~SmfSwapThread();
> int start(void);
>
> + static bool running(void);
> + static void setProc(SmfUpgradeProcedure *);
> +
> private:
> + static SmfSwapThread *me;
> + static std::mutex m_mutex;
>
> void main(void);
> int init(void);
>
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
Opensaf-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/opensaf-devel