memory conflict between codec engine and omx in DM8168

Hi Livio,

Are you using EZSDK 5.05.02.00?

BR

Pavel

Hi Pavel,

yes, I'm using EZSDK 5.05.02.00

Thanks 

LL

It seems that the error is in the syslink module.

ti-ezsdk_dm816x-evm_5_05_02_00/component-sources/syslink_2_20_02_20/packages/ti/syslink/family/hlos/knl/ti81xx/Platform.c :

/* add new shared region entries to table */
    for (i = 0; i < handle->slaveCfg.numSRs; i++) {
        GT_2trace(curTrace, GT_1CLASS, "handle->slaveSRCfg[%d].entryBase %p\n",
            i, handle->slaveSRCfg[i].entryBase);

        /* get entry from table */
        status = SharedRegion_getEntry(handle->slaveSRCfg[i].id, &entry);

        if (status < 0) {
            status = Platform_E_FAIL;
            GT_setFailureReason(curTrace, GT_4CLASS, "Platform_loadCallback",
                status, "SharedRegion_getEntry failed");
            goto leave;
        }

        /* if entry already started, validate properties on new entry */
        if (entry.isValid) {
            if (entry.len != handle->slaveSRCfg[i].entryLen) {
                status = -1;
                GT_setFailureReason(curTrace, GT_4CLASS,
                    "Platform_loadCallback", status,
                    "length of slave region does not match host entry");
                goto leave;
            }
            if (entry.ownerProcId != handle->slaveSRCfg[i].ownerProcId) {
                status = -1;
                GT_setFailureReason(curTrace, GT_4CLASS,
                    "Platform_loadCallback", status,
                    "owner of slave region does not match host entry");
                goto leave;
            }

.............

leave:
    /* if error, release resources in reverse order of acquisition */
    if (status < 0) {
        /* TODO: release map entries acquired in for loop */
        /* TODO: if Ipc_start failed, slave already running, what now? */
        if (handle->bslaveAdditionalReg != NULL) {
            Memory_free(NULL, handle->bslaveAdditionalReg,
                handle->slaveCfg.numSRs * sizeof(UInt8));
            handle->bslaveAdditionalReg = NULL;
        }
        if (handle->slaveSRCfg != NULL) {
            Memory_free(NULL, handle->slaveSRCfg,
                handle->slaveCfg.numSRs * sizeof(Platform_SlaveSRConfig));
            handle->slaveSRCfg = NULL;
        }
    }
    GT_1trace(curTrace, GT_LEAVE, "Platform_loadCallback: [0x%08x]", status);

    return (status);
}

The validation of the properties of the new entry fails. I will check with the Syslink/BIOS team for any clues on why the validation fails.

Meanwhile you can refer to the syslink documentation in ti-ezsdk_dm816x-evm_5_05_02_00/component-sources/syslink_2_20_02_20/docs for information on how to change the memory map. 

Here is some documentation from the EZSDK which you might want to take a look for some information for a particular supported configuration and additional information about how to make changes to the memory map:

http://processors.wiki.ti.com/index.php/EZSDK_Memory_Map

Regards,

Pavel

Hi Pavel,

thank you for the reply. I'll wait for any updates from TI about the issue. 

In the meanwhile you propose to change the memory map. I red some documentation even if I've never tried to do it since I use the default map.

By the way is not completely clear for me how to modify the memory map, in particular which areas have to be modified in order to solve my original problem.

thanks

LL 

Lima:

   Tracing through the code, the error says that the SharedRegion owner ID for some SharedRegion entry, read from the slave memory configuration after loading the DSP (the universal_copy load?), does not match the owner ID already set by a previous call to SharedRegion_setEntry() from a previous Platform_loadCallback() call (from the capture_encode example?).   These threads give some clues as to which files to look at for the mismatch:

http://e2e.ti.com/support/embedded/multimedia_software_codecs/f/356/t/226819.aspx

http://e2e.ti.com/support/dsp/davinci_digital_media_processors/f/717/t/149505.aspx

http://e2e.ti.com/support/embedded/multimedia_software_codecs/f/356/p/226819/801152.aspx

   From component-sources/codec_engine_3_22_01_06/examples/ti/sdo/ce/examples/servers/all_codecs/ti_platforms_evmTI816X.cfg:

ti_platforms_evmTI816X.cfg:entry.ownerProcId = MultiProc.getIdMeta("HOST");
ti_platforms_evmTI816X.cfg:entry2.ownerProcId = MultiProc.getIdMeta("HOST");

  Looking at the MulitProc config table in ipc Settings.xs,

    'TMS320TI816X'      : ["DSP", "VIDEO-M3", "VPSS-M3", "HOST"],

   that appears to set ownerProcId to '3'.

   The next question is where this OMX capture example set's it's Shared Regions, and owner Ids.  

   The link at http://processors.wiki.ti.com/index.php/EZSDK_Memory_Map#DSP_for_OMX gives some hints, though digging through the code, I see multiple places where an ownerId could be set.  

   But, looking at board-support/media-controller-utils_3_00_00_05/src/ldrmemcfg/src/ldr_memcfg.c,

   #ifdef LDR_DUMPMEMSEGINFO
    ldr_dumpMemSegInfo (pMemSeg);
    #endif

   it appears that if you #define LDR_DUMPMEMSEGINFO, and rebuild the mm_host_util, it will dump for you the memory segments that get configured into the bin file which is fed to the firmware loader.

   You could also directly put some trace in the Platform_loadCallback() and print the differing values of the SharedRegion owner procId, and the Region ID.

   See if that helps at all, and we can keep digging on the next post.