Hi Anders Widell,
Please find my response .
On 1/5/2016 3:26 PM, Anders Widell wrote:
> See comments inline.
>
> regards,
> Anders Widell
>
> On 01/04/2016 11:00 AM, A V Mahesh wrote:
>> Hi Anders Widell,
>>
>> Please find my answers.
>>
>> On 1/4/2016 2:20 PM, Anders Widell wrote:
>>> Ok good, that will solve the backwards compatibility issue! Though I
>>> still don't understand what the problem would be if we let the
>>> function time out at the time specified by the user, even when the
>>> user wants us to time out after less than fourteen seconds. If we go
>>> for a new API version, wouldn't it be best if we let the new version
>>> of the API treat any non-negative timeout parameter value as valid?
>> [AVM] Currently all OpenSaf (IMMA/CPA/LGA/NTFA/ect .... ) API are
>> tested using default ~10 second as its MAX Transport MDS timeout
>> value (Implicitly by CPA or explicitly passed by user/application) ,
>> and we haven't tested Opensaf with less value than
>> this on a normal 4 node cluster , if we have a requirement of even
>> when the user wants us to time out after less than default timeout
>> vault based
>> on his cluster size or performance of system, we can
>> make this as exportable/configurable variable at Transport level.
> [AndersW] The synchronous SAF API functions that don't take an
> explicit timeout parameter is a different story - for them I think it
> is fine to just test with the default timeout of 10 seconds although
> the user has the possibility to change the timeout using an
> environment variable (at least for the IMM API).
>
> For the API functions that take an explicit timeout parameter, we
> can't expect the user to always set this parameter to ten seconds. I
> understand that you cannot test with all possible values of the
> timeout parameter, since the timeout parameter is a 64-bit integer.
> You ought to have one test case where the SAF API call does time out,
> and anther test case where it doesn't time out. That, plus a few tests
> with extreme timeout values (0, 1, SA_TIME_MAX, and -1) ought to be
> enough.
>
> If we go for bumping the CKPT API version then we don't have to worry
> about backwards compatibility, and then I think the new version of the
> API should support the full range of the timeout parameter, i.e. [0,
> SA_TIME_MAX]. That is better than imposing some arbitrary limits (why
> 14 seconds?). It shouldn't be difficult to implement support for the
> full range of timeout values, since the underlying POSIX functions
> that we use do support them. And for the testing: we anyhow do need to
> test both the case when the timeout happens and when it doesn't.
[AVM] If we should support the full range of the timeout parameter,
i.e. [0, SA_TIME_MAX] , even we don't have an issue of application
backwards compatibility ( bumping the CKPT API version is NOT required ).
OK then I will do following change in new ticket for the
SAF API functions in all services that take an explicit timeout parameter:
1) API should support the full range of the timeout
parameter, i.e. [0, SA_TIME_MAX].
2) API will unconditionally return
SA_AIS_ERR_TIMEOUT if application specify > 0x98968000000000 as timeout
and it will be documented as well ( because the current implementation
of MDS header
adopting the size of the timeout parameter from 32
bits to 64 bits is complex , so I will explore the possibility of this
in an enhancement ) .
>>
>>>
>>> Another thing: it looks like the timeout parameter is used both by
>>> the agent (to time out waiting for the reply from the server) as
>>> well as by the server. Or, at least it is passed as a parameter in
>>> the Open request message to the CKPT node director. However, when I
>>> look at the CKPT node director code, as far as I can tell it looks
>>> like the timeout parameter isn't really used for anything. If I am
>>> correct, we end up in the cpnd_evt_proc_ckpt_open() function. It has
>>> a local variable called "timeout", and it reads the timeout
>>> parameter from the message it has received from the agent. But it is
>>> not used! Could you please take a look at this?
>> [AVM] This timeout is used only for Transport (MDS) sync_send
>> message timeout between CPA<--->CPND (
>> mds_info.info.svc_send.info.sndrsp.i_time_to_wait )
>> and CPND only has interaction with /dev/shmR/W
>> operation to create/read/write to Physical ckpt replicas ( the real
>> checkpoint data) , which varies for system to system configuration.
>>
>> CPND--->CPD--->CPND only control data, so the the
>> default timeout value is used between CPND & CPD .
> [AndersW] Ok, so then the part for handling the timeout variable in
> cpnd_evt_proc_ckpt_open() is just dead code that ought to be removed?
[AVM] yes this can be removed.
-AVM
>>
>> -AVM
>>
>>>
>>> thanks,
>>> Anders Widell
>>>
>>> On 01/04/2016 07:19 AM, A V Mahesh wrote:
>>>> Hi Anders Widell
>>>>
>>>> Ok I will bump the minor version of the CKPT API, or to use ELF
>>>> symbol version as you suggested.
>>>>
>>>> -AVM
>>>>
>>>> On 12/29/2015 3:49 PM, Anders Widell wrote:
>>>>> Hi!
>>>>>
>>>>> See my comments marked [AndersW] inline.
>>>>>
>>>>> regards,
>>>>> Anders Widell
>>>>>
>>>>> On 12/28/2015 05:35 AM, A V Mahesh wrote:
>>>>>> Hi Anders Widell,
>>>>>>
>>>>>> Please find my answers.
>>>>>>
>>>>>> On 12/22/2015 10:41 PM, Anders Widell wrote:
>>>>>>> See comments inline.
>>>>>>>
>>>>>>> regards,
>>>>>>> Anders Widell
>>>>>>>
>>>>>>> On 12/21/2015 11:04 AM, A V Mahesh wrote:
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> - The timeout here is the duration for which the API should
>>>>>>>> wait before returning to the user.
>>>>>>>> - The patch does not changes the minimum value of 14 seconds.
>>>>>>>> IE. Before this patch also the default value was 14 seconds.
>>>>>>>> This value was set based on implementation specific fine
>>>>>>>> tuning, particularly in large size cluster.
>>>>>>> I think it does change the minimum timeout. I am here referring
>>>>>>> to the functions that take an SaTimeT timeout parameter. You can
>>>>>>> try this example program, that used to work before ticket
>>>>>>> [#1583] was pushed:
>>>>>>>
>>>>>>> const SaNameT checkpointName = {strlen("safCkpt=hello"),
>>>>>>> "safCkpt=hello"};
>>>>>>> do {
>>>>>>> err = saCkptCheckpointOpen(
>>>>>>> ckptHandle,
>>>>>>> &checkpointName,
>>>>>>> &checkpointCreationAttributes,
>>>>>>> SA_CKPT_CHECKPOINT_READ |
>>>>>>> SA_CKPT_CHECKPOINT_WRITE |
>>>>>>> SA_CKPT_CHECKPOINT_CREATE,
>>>>>>> 10 * SA_TIME_ONE_SECOND,
>>>>>>> &checkpointHandle);
>>>>>>>
>>>>>> [AVM]
>>>>>>
>>>>>> Even though previously ( before this patch)
>>>>>> Application was passing 10 * SA_TIME_ONE_SECOND ( )
>>>>>> because of #1583 reported real issue of `SaTimeT
>>>>>> timeout parameter is not being converted in to TEN_MILLI_SEC()
>>>>>> while passing to mds_msg_sync_send()`
>>>>>> internally CPA was assuming Application is
>>>>>> passed 10 SECOND * 10000000 ( 10 * 1000000000 * 10000000)
>>>>>> ,which is being treated as larger than the default
>>>>>> timeout time of 14 SECOND, so previously the
>>>>>> Application only has an impression that it will get time out in
>>>>>> 10 SECOND , but in piratical it was considered larger
>>>>>> than 14 SECOND default time.
>>>>> [AndersW] Yes this is a bug and I am not arguing against fixing
>>>>> it. I don't think a fix for this would be backwards incompatible,
>>>>> since an application shouldn't depend on a bug in OpenSAF (we
>>>>> don't need to be bug-backwards-compatible, except perhaps in some
>>>>> unusual situations). But this is not the problem that I am talking
>>>>> about. The problem I am talking about is that the range of allowed
>>>>> timeout values has become smaller as part of the fix. Such a
>>>>> change is typically not backwards compatible, since an application
>>>>> could be using a timeout value which lies within the previously
>>>>> allowed range of timeout values, but which is outside the new
>>>>> (smaller) range of allowed values. The example program I provided
>>>>> demonstrates this problem. It works before [#1583] was pushed, but
>>>>> not afterwards. The program uses the checkpoint API in a perfectly
>>>>> valid and reasonable way. Thus, it provides a clear evidence that
>>>>> the change is not backwards compatible. If you don't believe me,
>>>>> please try the program for yourself.
>>>>>>
>>>>>> =================================================================================================================
>>>>>>
>>>>>>
>>>>>> Before #1583 Fix :
>>>>>> saCkptCheckpointOpen( ckptHandle,
>>>>>> &checkpointName, ....., 10 * SA_TIME_ONE_SECOND,
>>>>>> &checkpointHandle);
>>>>>> timeout = 10 * SA_TIME_ONE_SECOND;
>>>>>> proc_rc =
>>>>>> cpa_mds_msg_sync_send(cb->cpa_mds_hdl, &(cb->cpnd_mds_dest),
>>>>>> &evt, &out_evt, timeout);
>>>>>>
>>>>>> After #1583 Fix :
>>>>>> timeout = 10 * SA_TIME_ONE_SECOND;
>>>>>> time_out =
>>>>>> m_CPSV_CONVERT_SATIME_TEN_MILLI_SEC(timeout);
>>>>>> proc_rc =
>>>>>> cpa_mds_msg_sync_send(cb->cpa_mds_hdl, &(cb->cpnd_mds_dest),
>>>>>> &evt, &out_evt, time_out);
>>>>>>
>>>>>> =================================================================================================================
>>>>>>
>>>>>>
>>>>>>
>>>>>> In other words , before this patch in reality the
>>>>>> saCkptCheckpointOpen() was not being returned TIMEOUT even after
>>>>>> 10 SECOND ( but Application expecting it).
>>>>>> Application was only in assumption it was passing
>>>>>> 10 SECOND , so no change in behavioral difference in Application
>>>>>> before and after this patch.
>>>>> [AndersW] As I said above, this is not the part of the change that
>>>>> I have problems with. It is the following passage:
>>>>>
>>>>> Before (note: NCS_SAF_MIN_ACCEPT_TIME has the value 10):
>>>>>
>>>>> if (time_out < NCS_SAF_MIN_ACCEPT_TIME) {
>>>>> rc = SA_AIS_ERR_TIMEOUT;
>>>>> TRACE_4("Cpa CkptOpen failed Api failed with return
>>>>> value:%d,ckptHandle:%llx,timeout:%llu", rc, ckptHandle, timeout);
>>>>> goto mds_send_fail;
>>>>> }
>>>>>
>>>>>
>>>>> After (note: CPSV_WAIT_TIME has the value 1400):
>>>>>
>>>>> if (time_out < CPSV_WAIT_TIME) {
>>>>> rc = SA_AIS_ERR_TIMEOUT;
>>>>> TRACE_4("Cpa CkptOpen failed Api failed with return
>>>>> value:%d,ckptHandle:%llx,timeout:%llu",
>>>>> rc, ckptHandle, timeout);
>>>>> goto mds_send_fail;
>>>>> }
>>>>>
>>>>> So yes, there is a change in behaviour from the application's
>>>>> point of view. If the timeout (after conversion to a unit of 10
>>>>> milliseconds) falls within the range 10 to 1400, the application
>>>>> will now always get the SA_AIS_ERR_TIMEOUT. Previously, the
>>>>> function may have succeeded.
>>>>>>
>>>>>>>
>>>>>>>> - Even though it is not a practical case of passing 49 days
>>>>>>>> (greater than or equal to 0x98968000000000) as a timeout value,
>>>>>>>> yes what you say can be done across all services.
>>>>>>> Shouldn't we then document 0x98968000000000 as the upper limit
>>>>>>> for the timeout parameter, and validate that the parameter given
>>>>>>> by the user is less than this upper limit. As it is coded right
>>>>>>> now, the function will unconditionally return SA_AIS_ERR_TIMEOUT
>>>>>>> if you specify 0x98968000000000 as timeout (you can try this by
>>>>>>> replacing 10 * SA_TIME_ONE_SECOND with 0x98968000000000 in the
>>>>>>> example program above).
>>>>>>
>>>>>> [AVM] Ok ,as if this issue exist with all service , I will
>>>>>> raise a new ticket and will be fixed/document for all services
>>>>>> point of view .
>>>>> [AndersW] Instead of documenting this as a limitation, please fix
>>>>> MDS by changing the size of the timeout parameter from 32 bits to
>>>>> 64 bits. Note that there are perfectly reasonable timeout values
>>>>> that are very large, e.g. someone might be using SA_TIME_MAX to
>>>>> get a (for all practical purposes) infinite timeout.
>>>>>>
>>>>>> -AVM
>>>>>>>>
>>>>>>>> Thanks,
>>>>>>>> AV.
>>>>>>>>
>>>>>>>>
>>>>>>>> On 12/18/2015 2:43 PM, Anders Widell wrote:
>>>>>>>>> Hi!
>>>>>>>>>
>>>>>>>>> I have some questions regarding this change:
>>>>>>>>>
>>>>>>>>> 1) This change is not backwards compatible. Have you
>>>>>>>>> considered to either bump the minor version of the CKPT API,
>>>>>>>>> or to use ELF symbol versioning, to avoid breaking old
>>>>>>>>> applications?
>>>>>>>>>
>>>>>>>>> 2) Why was the minimum possible time-out increased from 0.1
>>>>>>>>> seconds to 14 seconds? 14 seconds is like an eternity in a
>>>>>>>>> real-time system.
>>>>>>>>>
>>>>>>>>> 3) If the user-specified timeout is less than 14 seconds, the
>>>>>>>>> functions will always fail without even trying. Wouldn't it be
>>>>>>>>> better to at least try to perform the operation within the
>>>>>>>>> time limit specified by the user? The user may have a loop
>>>>>>>>> where it keeps retrying the API function call when the
>>>>>>>>> SA_AIS_ERR_TIMEOUT error code is returned. I would expect such
>>>>>>>>> a loop to eventually terminate, especially if the timeout
>>>>>>>>> parameter has a reasonable value (and there are many
>>>>>>>>> reasonable timeout values which are less than 14 seconds).
>>>>>>>>>
>>>>>>>>> 3) Can't the conversion from nanoseconds represented as a
>>>>>>>>> signed 64-bit integer to a unit of 10 milliseconds stored in
>>>>>>>>> an unsigned 32-bit integer ('uint32_t time_out' in the code
>>>>>>>>> below) overflow? I.e. if the timeout parameter has a value
>>>>>>>>> greater than or equal to 0x98968000000000?
>>>>>>>>>
>>>>>>>>> / Anders Widell
>>>>>>>>>
>>>>>>>>> On 11/06/2015 06:33 AM, mahesh.va...@oracle.com wrote:
>>>>>>>>>> osaf/libs/agents/saf/cpa/cpa_api.c | 111
>>>>>>>>>> ++++++++++++++++++++++++++++++------
>>>>>>>>>> 1 files changed, 91 insertions(+), 20 deletions(-)
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> If we test upper and lower limits of SaTimeT timeout
>>>>>>>>>> arguments to CPSV API ,
>>>>>>>>>> some of CPA APIs are doesn't returns or hungs.
>>>>>>>>>>
>>>>>>>>>> Current cpa code doesn't validating or passing SaTimeT
>>>>>>>>>> timeout arguments to CPSV API ,
>>>>>>>>>> which are below default API timeout limit of CPSV_WAIT_TIME
>>>>>>>>>> (1400 X 10ms) and
>>>>>>>>>> above SaTimeT, type long long int Maximum value
>>>>>>>>>> 9223372036854775807 (2^63-1)
>>>>>>>>>>
>>>>>>>>>> For example :
>>>>>>>>>>
>>>>>>>>>> i) const SaTimeT timeout = 1215752192;
>>>>>>>>>> ii) const SaTimeT timeout = 2 * SA_TIME_ONE_MINUTE;
>>>>>>>>>> iii) const SaTimeT timeout = 9223372036854775809;
>>>>>>>>>>
>>>>>>>>>> saCkptCheckpointOpen(...., timeout, ...);
>>>>>>>>>>
>>>>>>>>>> Because of that in some of variables are over flowing and ,
>>>>>>>>>> API calls are not working properly
>>>>>>>>>>
>>>>>>>>>> Fix :
>>>>>>>>>>
>>>>>>>>>> 1 - Validated if (time_out < CPSV_WAIT_TIME) return
>>>>>>>>>> SA_AIS_ERR_INVALID_PARAM
>>>>>>>>>> 2 - Validates if (expirationTime < 0 ) return
>>>>>>>>>> SA_AIS_ERR_INVALID_PARAM
>>>>>>>>>> 3 - Pass proper timeout values to internal functions of Opensaf
>>>>>>>>>>
>>>>>>>>>> diff --git a/osaf/libs/agents/saf/cpa/cpa_api.c
>>>>>>>>>> b/osaf/libs/agents/saf/cpa/cpa_api.c
>>>>>>>>>> --- a/osaf/libs/agents/saf/cpa/cpa_api.c
>>>>>>>>>> +++ b/osaf/libs/agents/saf/cpa/cpa_api.c
>>>>>>>>>> @@ -27,7 +27,6 @@
>>>>>>>>>> ******************************************************************************/
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> #include "cpa.h"
>>>>>>>>>> -#define NCS_SAF_MIN_ACCEPT_TIME 10
>>>>>>>>>> /****************************************************************************
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Name : SaCkptInitialize
>>>>>>>>>> @@ -889,6 +888,14 @@ SaAisErrorT saCkptCheckpointOpen(SaCkptH
>>>>>>>>>> return SA_AIS_ERR_INVALID_PARAM;
>>>>>>>>>> }
>>>>>>>>>> + if (timeout < 0 ) {
>>>>>>>>>> + TRACE_4("Cpa CkptOpen: failed invalid timeout return
>>>>>>>>>> value:%d,ckptHandle:%llx",
>>>>>>>>>> + SA_AIS_ERR_INVALID_PARAM, ckptHandle);
>>>>>>>>>> + TRACE_LEAVE2("API return code = %u", rc);
>>>>>>>>>> + return SA_AIS_ERR_INVALID_PARAM;
>>>>>>>>>> + }
>>>>>>>>>> +
>>>>>>>>>> +
>>>>>>>>>> /* Draft Validations */
>>>>>>>>>> rc =
>>>>>>>>>> cpa_open_attr_validate(checkpointCreationAttributes,
>>>>>>>>>> checkpointOpenFlags, checkpointName);
>>>>>>>>>> if (rc != SA_AIS_OK) {
>>>>>>>>>> @@ -975,13 +982,13 @@ SaAisErrorT saCkptCheckpointOpen(SaCkptH
>>>>>>>>>> }
>>>>>>>>>> evt.info.cpnd.info.openReq.ckpt_flags =
>>>>>>>>>> checkpointOpenFlags;
>>>>>>>>>> -
>>>>>>>>>> /* convert the timeout to 10 ms value and add it to the
>>>>>>>>>> sync send timeout */
>>>>>>>>>> time_out = m_CPSV_CONVERT_SATIME_TEN_MILLI_SEC(timeout);
>>>>>>>>>> - if (time_out < NCS_SAF_MIN_ACCEPT_TIME) {
>>>>>>>>>> + if (time_out < CPSV_WAIT_TIME) {
>>>>>>>>>> rc = SA_AIS_ERR_TIMEOUT;
>>>>>>>>>> - TRACE_4("Cpa CkptOpen failed Api failed with return
>>>>>>>>>> value:%d,ckptHandle:%llx,timeout:%llu", rc, ckptHandle,
>>>>>>>>>> timeout);
>>>>>>>>>> + TRACE_4("Cpa CkptOpen failed Api failed with return
>>>>>>>>>> value:%d,ckptHandle:%llx,timeout:%llu",
>>>>>>>>>> + rc, ckptHandle, timeout);
>>>>>>>>>> goto mds_send_fail;
>>>>>>>>>> }
>>>>>>>>>> @@ -998,9 +1005,8 @@ SaAisErrorT saCkptCheckpointOpen(SaCkptH
>>>>>>>>>> TRACE_4("cpa CkptOpen:CPND_DOWN Api failed with
>>>>>>>>>> return value:%d,ckptHandle:%llx", rc, ckptHandle);
>>>>>>>>>> goto mds_send_fail;
>>>>>>>>>> }
>>>>>>>>>> -
>>>>>>>>>> /* Send the evt to CPND */
>>>>>>>>>> - proc_rc = cpa_mds_msg_sync_send(cb->cpa_mds_hdl,
>>>>>>>>>> &(cb->cpnd_mds_dest), &evt, &out_evt, timeout);
>>>>>>>>>> + proc_rc = cpa_mds_msg_sync_send(cb->cpa_mds_hdl,
>>>>>>>>>> &(cb->cpnd_mds_dest), &evt, &out_evt, time_out);
>>>>>>>>>> /* Generate rc from proc_rc */
>>>>>>>>>> switch (proc_rc) {
>>>>>>>>>> @@ -1709,6 +1715,13 @@ SaAisErrorT saCkptCheckpointRetentionDur
>>>>>>>>>> goto done;
>>>>>>>>>> }
>>>>>>>>>> + if (retentionDuration < 0 ) {
>>>>>>>>>> + TRACE_4("Cpa RetentionDurationSet: failed
>>>>>>>>>> invalid retentionDuration return value:%d,ckptHandle:%llx",
>>>>>>>>>> + SA_AIS_ERR_INVALID_PARAM, checkpointHandle);
>>>>>>>>>> + rc = SA_AIS_ERR_INVALID_PARAM;
>>>>>>>>>> + goto done;
>>>>>>>>>> + }
>>>>>>>>>> +
>>>>>>>>>> /* get the CB Lock */
>>>>>>>>>> if (m_NCS_LOCK(&cb->cb_lock, NCS_LOCK_WRITE) !=
>>>>>>>>>> NCSCC_RC_SUCCESS) {
>>>>>>>>>> rc = SA_AIS_ERR_LIBRARY;
>>>>>>>>>> @@ -2183,6 +2196,7 @@ SaAisErrorT saCkptSectionCreate(SaCkptCh
>>>>>>>>>> bool gen_sec_flag = false;
>>>>>>>>>> SaCkptSectionIdT app_ptr;
>>>>>>>>>> CPA_CLIENT_NODE *cl_node = NULL;
>>>>>>>>>> + uint32_t time_out;
>>>>>>>>>> TRACE_ENTER2("SaCkptCheckpointHandleT passed is
>>>>>>>>>> %llx",checkpointHandle);
>>>>>>>>>> /* Validate the Input Parameters */
>>>>>>>>>> @@ -2312,8 +2326,21 @@ SaAisErrorT saCkptSectionCreate(SaCkptCh
>>>>>>>>>> evt.info.cpnd.info.sec_creatReq.init_size =
>>>>>>>>>> initialDataSize;
>>>>>>>>>> m_NCS_UNLOCK(&cb->cb_lock, NCS_LOCK_WRITE);
>>>>>>>>>> +
>>>>>>>>>> + time_out = CPA_WAIT_TIME(initialDataSize);
>>>>>>>>>> +
>>>>>>>>>> + if (time_out < 0 ) {
>>>>>>>>>> + rc = SA_AIS_ERR_TIMEOUT;
>>>>>>>>>> + TRACE_4("cpa SectCreate Api failed because of larg
>>>>>>>>>> initialDataSize return :%d,ckptHandle:%llx",
>>>>>>>>>> + rc, checkpointHandle);
>>>>>>>>>> + goto fail1;
>>>>>>>>>> + }
>>>>>>>>>> + else if (time_out < CPSV_WAIT_TIME) {
>>>>>>>>>> + time_out = CPSV_WAIT_TIME;
>>>>>>>>>> + }
>>>>>>>>>> +
>>>>>>>>>> proc_rc = cpa_mds_msg_sync_send(cb->cpa_mds_hdl,
>>>>>>>>>> &(gc_node->active_mds_dest),
>>>>>>>>>> - &evt, &out_evt,
>>>>>>>>>> CPA_WAIT_TIME(initialDataSize));
>>>>>>>>>> + &evt, &out_evt, time_out);
>>>>>>>>>> /* Generate rc from proc_rc */
>>>>>>>>>> switch (proc_rc) {
>>>>>>>>>> @@ -2684,6 +2711,14 @@ SaAisErrorT saCkptSectionExpirationTimeS
>>>>>>>>>> return rc;
>>>>>>>>>> }
>>>>>>>>>> + if (expirationTime < 0 ) {
>>>>>>>>>> + TRACE_4("Cpa ExpirationTimeSet: failed
>>>>>>>>>> invalid expirationTime return value:%d,ckptHandle:%llx",
>>>>>>>>>> + SA_AIS_ERR_INVALID_PARAM, checkpointHandle);
>>>>>>>>>> + rc = SA_AIS_ERR_INVALID_PARAM;
>>>>>>>>>> + TRACE_LEAVE2("API return code = %u", rc);
>>>>>>>>>> + return rc;
>>>>>>>>>> + }
>>>>>>>>>> +
>>>>>>>>>> /* retrieve CPA CB */
>>>>>>>>>> m_CPA_RETRIEVE_CB(cb);
>>>>>>>>>> if (!cb) {
>>>>>>>>>> @@ -2847,6 +2882,14 @@ SaAisErrorT saCkptSectionIterationInitia
>>>>>>>>>> return rc;
>>>>>>>>>> }
>>>>>>>>>> + if (expirationTime < 0 ) {
>>>>>>>>>> + TRACE_4("Cpa SectionIterationInitialize: failed
>>>>>>>>>> invalid expirationTime return value:%d,ckptHandle:%llx",
>>>>>>>>>> + SA_AIS_ERR_INVALID_PARAM, checkpointHandle);
>>>>>>>>>> + rc = SA_AIS_ERR_INVALID_PARAM;
>>>>>>>>>> + TRACE_LEAVE2("API return code = %u", rc);
>>>>>>>>>> + return rc;
>>>>>>>>>> + }
>>>>>>>>>> +
>>>>>>>>>> /* retrieve CPA CB */
>>>>>>>>>> m_CPA_RETRIEVE_CB(cb);
>>>>>>>>>> if (!cb) {
>>>>>>>>>> @@ -3360,6 +3403,7 @@ SaAisErrorT saCkptCheckpointWrite(SaCkpt
>>>>>>>>>> CPA_CLIENT_NODE *cl_node = NULL;
>>>>>>>>>> SaSizeT all_ioVector_size = 0;
>>>>>>>>>> uint32_t err_flag = 0;
>>>>>>>>>> + uint32_t time_out;
>>>>>>>>>>
>>>>>>>>>> TRACE_ENTER2("SaCkptCheckpointHandleT passed is
>>>>>>>>>> %llx",checkpointHandle);
>>>>>>>>>> @@ -3479,18 +3523,27 @@ SaAisErrorT
>>>>>>>>>> saCkptCheckpointWrite(SaCkpt
>>>>>>>>>> /* Unlock cpa_lock before calling mds api */
>>>>>>>>>> m_NCS_UNLOCK(&cb->cb_lock, NCS_LOCK_WRITE);
>>>>>>>>>> -#if 1
>>>>>>>>>> - for (iter = 0; iter < numberOfElements; iter++)
>>>>>>>>>> +
>>>>>>>>>> + for (iter = 0; iter < numberOfElements; iter++) {
>>>>>>>>>> all_ioVector_size += ioVector[iter].dataSize;
>>>>>>>>>> -
>>>>>>>>>> - proc_rc = cpa_mds_msg_sync_send(cb->cpa_mds_hdl,
>>>>>>>>>> &(gc_node->active_mds_dest),
>>>>>>>>>> - &evt, &out_evt,
>>>>>>>>>> CPA_WAIT_TIME(all_ioVector_size));
>>>>>>>>>> -#else
>>>>>>>>>> - proc_rc = cpa_mds_msg_sync_send(cb->cpa_mds_hdl,
>>>>>>>>>> &(gc_node->active_mds_dest), &evt, &out_evt, CPSV_WAIT_TIME);
>>>>>>>>>> -#endif
>>>>>>>>>> - /* Generate rc from proc_rc */
>>>>>>>>>> - switch (proc_rc)
>>>>>>>>>> - {
>>>>>>>>>> + time_out = CPA_WAIT_TIME(all_ioVector_size);
>>>>>>>>>> +
>>>>>>>>>> + if (time_out < 0 ) {
>>>>>>>>>> + rc = SA_AIS_ERR_TIMEOUT;
>>>>>>>>>> + TRACE_4("cpa ckpointWrite Api failed because of
>>>>>>>>>> larg ioVector size return :%d,ckptHandle:%llx",
>>>>>>>>>> + rc, checkpointHandle);
>>>>>>>>>> + goto done;
>>>>>>>>>> + }
>>>>>>>>>> + else if (time_out < CPSV_WAIT_TIME) {
>>>>>>>>>> + time_out = CPSV_WAIT_TIME;
>>>>>>>>>> + }
>>>>>>>>>> + proc_rc = cpa_mds_msg_sync_send(cb->cpa_mds_hdl,
>>>>>>>>>> &(gc_node->active_mds_dest),
>>>>>>>>>> + &evt, &out_evt, time_out);
>>>>>>>>>> +
>>>>>>>>>> + }
>>>>>>>>>> + /* Generate rc from proc_rc */
>>>>>>>>>> + switch (proc_rc)
>>>>>>>>>> + {
>>>>>>>>>> case NCSCC_RC_SUCCESS:
>>>>>>>>>> break;
>>>>>>>>>> case NCSCC_RC_REQ_TIMOUT:
>>>>>>>>>> @@ -3577,6 +3630,7 @@ SaAisErrorT saCkptSectionOverwrite(SaCkp
>>>>>>>>>> CPA_GLOBAL_CKPT_NODE *gc_node = NULL;
>>>>>>>>>> bool add_flag = false;
>>>>>>>>>> CPA_CLIENT_NODE *cl_node = NULL;
>>>>>>>>>> + uint32_t time_out;
>>>>>>>>>> TRACE_ENTER2("SaCkptCheckpointHandleT passed is
>>>>>>>>>> %llx",checkpointHandle);
>>>>>>>>>> memset(&ckpt_data, '\0', sizeof(CPSV_CKPT_DATA));
>>>>>>>>>> @@ -3684,8 +3738,19 @@ SaAisErrorT saCkptSectionOverwrite(SaCkp
>>>>>>>>>> /* Unlock cpa_lock before calling mds api */
>>>>>>>>>> m_NCS_UNLOCK(&cb->cb_lock, NCS_LOCK_WRITE);
>>>>>>>>>> + time_out = CPA_WAIT_TIME(dataSize);
>>>>>>>>>> +
>>>>>>>>>> + if (time_out < 0 ) {
>>>>>>>>>> + rc = SA_AIS_ERR_TIMEOUT;
>>>>>>>>>> + TRACE_4("cpa SectionOverwrite Api failed because of
>>>>>>>>>> larg dataSize return :%d,ckptHandle:%llx", rc,
>>>>>>>>>> checkpointHandle);
>>>>>>>>>> + goto done;
>>>>>>>>>> + }
>>>>>>>>>> + else if (time_out < CPSV_WAIT_TIME) {
>>>>>>>>>> + time_out = CPSV_WAIT_TIME;
>>>>>>>>>> + }
>>>>>>>>>> +
>>>>>>>>>> proc_rc = cpa_mds_msg_sync_send(cb->cpa_mds_hdl,
>>>>>>>>>> &(gc_node->active_mds_dest),
>>>>>>>>>> - &evt, &out_evt, CPA_WAIT_TIME(dataSize));
>>>>>>>>>> + &evt, &out_evt, time_out);
>>>>>>>>>> /* Generate rc from proc_rc */
>>>>>>>>>> switch (proc_rc) {
>>>>>>>>>> @@ -4053,6 +4118,12 @@ SaAisErrorT saCkptCheckpointSynchronize(
>>>>>>>>>> /* For Ckpt Sync CPA will not perform any operation
>>>>>>>>>> other than passing the request to CPND */
>>>>>>>>>> + if (timeout < 0 ) {
>>>>>>>>>> + TRACE_4("Cpa CheckpointSynchronize: failed with
>>>>>>>>>> invalied timeout return value:%d,ckptHandle:%llx",
>>>>>>>>>> SA_AIS_ERR_INVALID_PARAM, checkpointHandle);
>>>>>>>>>> + rc =SA_AIS_ERR_INVALID_PARAM;
>>>>>>>>>> + goto done;
>>>>>>>>>> + }
>>>>>>>>>> +
>>>>>>>>>> /* retrieve CPA CB */
>>>>>>>>>> m_CPA_RETRIEVE_CB(cb);
>>>>>>>>>> if (!cb) {
>>>>>>>>>> @@ -4139,7 +4210,7 @@ SaAisErrorT saCkptCheckpointSynchronize(
>>>>>>>>>> /* Convert the time from saTimeT to millisecs */
>>>>>>>>>> timeout = m_CPSV_CONVERT_SATIME_TEN_MILLI_SEC(timeout);
>>>>>>>>>> - if (timeout < NCS_SAF_MIN_ACCEPT_TIME) {
>>>>>>>>>> + if (timeout < CPSV_WAIT_TIME) {
>>>>>>>>>> rc = SA_AIS_ERR_TIMEOUT;
>>>>>>>>>> TRACE_4("cpa CkptSynchronize Api failed with return
>>>>>>>>>> value:%d,ckptHandle:%llx ", rc, checkpointHandle);
>>>>>>>>>> goto fail1;
>>>>>>>>>>
>>>>>>>>>> ------------------------------------------------------------------------------
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> _______________________________________________
>>>>>>>>>> Opensaf-devel mailing list
>>>>>>>>>> Opensaf-devel@lists.sourceforge.net
>>>>>>>>>> https://lists.sourceforge.net/lists/listinfo/opensaf-devel
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>>
>
------------------------------------------------------------------------------
_______________________________________________
Opensaf-devel mailing list
Opensaf-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/opensaf-devel
