summaryrefslogtreecommitdiffstats
path: root/domx
Commit message (Collapse)AuthorAgeFilesLines
* domx: connection robustness improvementsLuden2016-05-011-16/+29
| | | | | | | | | | | | | | | | | CMA-enabled kernel for tuna devices will unload Ducati firmware when it's not in use to free extra memory for applications. On the first access to /dev/rpmsg-omx1 it will reload the firmware. However, in the process of this reloading device drivers will be reset, therefore causing errors when either opening the device or performing the first operation on it. Moreover, the previously opened handle will be invalidated in this case. Therefore, the connection loop is modified to retry both open() and first ioctl()calls instead of just open() and to reopen the device on retries. Also, because the reload can happen relatively fast when there are not too many pages to migrate, we use smaller waiting period but increased number of retries to improve camera start latency. Change-Id: I4b919f457c0e3bc43f99381c405e364b3326f583
* domx: fix up DCC data loading when USES_LEGACY_DOMX_DCC is setZiyan2016-02-061-1/+8
| | | | | | | The 'dcc_loaded' local static variable was never reset properly. Change-Id: Id4ceddea6153d897456e0f845f3818be5dfbae36 (cherry picked from commit 881a349f9721837f2556b6ed8214fecbf7a19f27)
* domx: add support for tunaZiyan2016-01-034-0/+73
| | | | | Change-Id: Ibdf65581caca78b2daa41ac9898383f4c3d38097 (cherry picked from commit e57f2b6f67a2b87da5b9c2f2b82e6cc28eef819b)
* domx: Fix GetParameter crashing with OMX_IndexParamConsumerUsageBitsZiyan2015-10-241-0/+7
| | | | | | | | | | Thanks to Kyle Repinski <repinski23@gmail.com> for pinpointing the issue, I just ifdef'd it and modified the return value to OMX_ErrorNotImplemented. Since OMX_IndexParamConsumerUsageBits is fairly new, we're assuming none of the current omap4 ducatis implement it. Change-Id: Ib1e26c8be36d90ef1624699a5d4107ae25794951
* domx: pass ANDROID_API_CFLAGSZiyan2015-10-211-0/+1
| | | | Change-Id: I4b2f8b72601aae7c4baedd91ddfb6d6437de73f7
* domx: Add structures and indices for Motorola devicesMichael Gernoth2015-02-212-1/+51
| | | | | | | These additions are only enabled when BOARD_USE_MOTOROLA_DOMX_ENHANCEMENTS is defined. Change-Id: I47d559d4908b7f5da2b1b3370a736a13376ef464
* fix hardware/ti/omap4xxx referencesHashcode2015-02-031-3/+2
|
* domx: revert mistaken change of DCC_PATHHashcode2015-02-021-1/+1
|
* domx: link liblog when building libOMX_CoreHashcode2015-02-021-1/+2
|
* domx: add TI_CAMERAHAL_USES_LEGACY_DOMX_DCC board flag to enable legacy DCCHashcode2015-02-023-4/+19
|
* domx: remove duplicate OMX_TI_COLOR_FormatYUV420PackedSemiPlanar definitionHashcode2015-02-021-2/+0
|
* domx: remove duplicate include for VP8Hashcode2015-02-021-2/+1
|
* set correct DCC file locationHashcode2015-02-021-1/+1
| | | | Change-Id: If1a49cddc1ea241008b9e8faeae26ace8415b188
* revert: DOMX : Removes DCC loading from OMX Camera ProxyHashcode2015-02-022-1/+335
| | | | | | | This reverts the DCC handling by camera HAL and loads the dcc files using the legacy method (works better with ICS-based ducati binaries) Change-Id: I2b8d4615beee224068ea39a746ff8b4796699d20
* domx: remove duplicate OMX headers and refer to originals where neededHashcode2015-02-0215-7132/+22
|
* domx: add board config for LOW_SECURE_HEAPMichael Gernoth2015-02-022-1/+11
| | | | | | | Change-Id: I1a888975945ec88d011e3bfc4e8af7d8ac36fc59 Conflicts: domx/domx/plugins/memplugin_ion.c
* domx: create subdir for domx to merge into omap4-nextHashcode2015-01-30101-90/+31810
|
* Use ion_ti instead of ionDaniel Levin2012-11-193-3/+3
| | | | | | | | | | | | Since Google approach is to have own ion library in parallel with TI ion library, latter has been renamed to ion_ti in JB MR1 and moved into hardware/ti/omap4xxx together with kernel headers. This patch fixes inclusion of ion library from <ion_ti/ion.h> and linking to libion_ti appropriately. Change-Id: Ibea0d45b013a49310d75c77561724add4ee01999 Signed-off-by: Daniel Levin <dendy@ti.com>
* DOMX: Changed return code for non-successful memory allocation cases.Oleksii Gulchenko2012-11-082-0/+3
| | | | | | | | | | Functions "MemPlugin_ION_Alloc", "PROXY_AllocateBuffer" and "PROXY_UseBuffer" always returns successfull result, in unsuccessfull memory allocation cases. This behaviour caused crashes in code, which used these functions. Change-Id: I014905d7972d7a74d2de5051d34fdf52d90bca83 Signed-off-by: Oleksii Gulchenko <oleksii.gulchenko@ti.com>
* [PROXY] Remove PROXY_AllocateBufferIonCarveout methodVidhoon Viswanathan2012-10-191-26/+0
| | | | | | | | | | | | | | | | With the introduction of generic MEMUTIL APIs, the old wrapper method PROXY_AllocateBufferIonCarveout used to make ION specific allocations in DOMX PROXY is no longer used. This patch removes the method from code. This patch should be integrated after the integration of following patches: http://review.omapzoom.org/#/c/27669/ http://review.omapzoom.org/#/c/27671/ http://review.omapzoom.org/#/c/27776/ Change-Id: I10e0f251aaf3a15b903cfdf3af38695cda9c4193 Signed-off-by: Vidhoon Viswanathan <vidhoon@ti.com>
* [MEMPLUGIN ION] Patch to enable secure buffer allocationVidhoon Viswanathan2012-10-192-6/+17
| | | | | | | | | | | This patch allows allocation of secure buffer in memplugin ION. Based on alloc flags, retry with tiler 1d is done only for default buffer alloc flags. For any special alloc flags, no retry done. This patch has been verified on OMAP5 JB DB 139 for android sanity. Change-Id: Ia95b8a9eba5b42c1cb69342fe2a5222e8a718a7e Signed-off-by: Vidhoon Viswanathan <vidhoon@ti.com>
* [FIX] Pass PluginExtendedInfo properly to MemPluginsVidhoon Viswanathan2012-10-192-10/+11
| | | | | | | | | | | | This patch fixes the following: *initialization for pPluginExtendedInfo was missed previously. *pMemPluginHandle->pPluginExtendedInfo was not properly pointed and passed for allocation. These two issues are fixed in this patch and tested on OMAP5 DB 139. Change-Id: I5150c4c71eccc7b24d1e4fad01bb0cf1312b0170 Signed-off-by: Vidhoon Viswanathan <vidhoon@ti.com>
* [DOMX OMXRPC] Fix memory leakAlberto Aguirre2012-10-121-4/+20
| | | | | | | | | | Fix memory leak when receiving OMX_ErrorHardware from the remote core. In such cases, the listener will not free the allocated packet; instead omx_rpc should check for this condition and free the packet. Change-Id: I3694a8fef4fee93495070671fe574abd61b3b948 Signed-off-by: Alberto Aguirre <a-aguirre@ti.com>
* [PROXY] Integrate MemPlugin API within DOMX PROXY (PART1)Vidhoon Viswanathan2012-10-043-159/+103
| | | | | | | | | | | | | | | | | | | | | | | | | | | This patch integrates the generic memory manager APIs from MemPlugin introduced by http://review.omapzoom.org/#/c/27669/ into the DOMX proxy COMMON and components. This involves the following changes: *New element pMemPluginHandle has been introduced in PROXY_COMPONENT_PRIVATE to contain memory manager info and to link to corresponding APIs. *Remove ION APIs that are directly used in DOMX PROXY code. PROXY common is completely cleaned out in this patch. In component proxies, camera proxy is not completely clean yet since it requires more effort and testing. That will be done in PART 2. *Make changes to compile MemPlugin files that have been integrated and used within PROXY code. Tested on DB 285 with android and OMTB sanity. Change-Id: I31e9ecefcf5be77da237a7b1b6234b7f34cf99f7 Signed-off-by: Vidhoon Viswanathan <vidhoon@ti.com>
* [PLUGINS] MemPlugins: Plugin based integration of Memory managersVidhoon Viswanathan2012-10-045-0/+960
| | | | | | | | | | | | | | | | This patch introduces the MemPlugin generic API interface for buffer actions. All types and definitions are contained in memplugin.h & memplugin.c ION memory manager is "plugged in" through internal methods defined in memplugin_ion.h & memplugin_ion.c. The generic APIs are linked to the specific memory manager methods internally based on the mapping table defined in memplugin_table.c. Change-Id: Ibff94732c40749f9c8cf34aa18bc0c43ddd298de Signed-off-by: Vidhoon Viswanathan <vidhoon@ti.com>
* PROXY_BUFFER_ACCESSOR: interfaces to all types of accessorsVidhoon Viswanathan2012-09-142-77/+84
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This patch introduces a new struct PROXY_BUFFER_ACCESSOR. This contains all possible accessors for a buffer including: *buffer handle pBufferHandle *context free FD bufferFd *Registered handle pRegBufferHandle *Mapped buffer address pBufferMappedAddress Three elements of this struct are included in the PROXY_BUFFER_INFO struct replacing the following elements that were not well organized: OMX_PTR pYBuffer; OMX_PTR pMetaDataBuffer; int mmap_fd; int mmap_fd_metadata_buff; OMX_PTR pRegisteredAufBux0; OMX_PTR pRegisteredAufBux1; OMX_PTR pRegisteredAufBux2; This patch also removes the ION dependency from these elements. In addition, all changes to introduce this new struct and element in PROXY_BUFFER_INFO are included in this patch. Tested on DB 285 with android and OMTB sanity. Change-Id: I6f4ea6239bf6b8d5395fece04008f6b8298f0a95 Signed-off-by: Vidhoon Viswanathan <vidhoon@ti.com>
* [FIX] Add missing Unregistration in PROXY_ComponentDeInit methodVidhoon Viswanathan2012-09-141-2/+29
| | | | | | | | | | | | | | | | Buffer registration and unregistration was added to facilitate buffer tracking from RPMSG. Registered buffer handle should be unregistered before they are freed. Buffers are freed either in FreeBuffer or ComponentDeInit. Presently Unregister calls are included in FreeBuffer method. This patch adds them to ComponentDeInit which was missed previously. Now registered buffer handles are unregistered at all places before the buffer is freed. Change-Id: I58919b8ad6f955bf2632d18b9480b6b24dac9472 Signed-off-by: Vidhoon Viswanathan <vidhoon@ti.com>
* [PROXY] Removing ION dependency from PROXY_COMPONENT_PRIVATEVidhoon Viswanathan2012-09-142-29/+30
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | The following members were nested under USE_ION preprocessor flag in PROXY_COMPONENT_PRIVATE structure: *bUseIon *bMapIonBuffers *ion_fd Out of them, bMapIonBuffers and ion_fd are generic and denote: *bMapIonBuffers - if buffers needs to be mapped in userspace *ion_fd - userspace mem manager[ION] client for creating/destroying buffers Hence they have been generalized in this patch into: *bMapIonBuffers -> bMapBuffers *ion_fd -> nMemmgrClientDesc Also these elements are removed from the USE_ION # flag. Changes to element 'bUseIon' has greater impacts in the domx code base and hence will be done in a separate patch. Note: This patch has nothing to do with COMMON DOMX, GLUE and UTILS. This is a seperate clean up. This patch has been tested in REL DB 285 with android and OMTB sanity. Signed-off-by: Vidhoon Viswanathan <vidhoon@ti.com> Change-Id: I4ca29ceb0e0f5f5c2ae306f802cab2a888a18f03
* Raw dump utility: fix for continuos NV12.Devaraj Rangasamy2012-09-121-1/+2
| | | | | | | | | | Raw dump utility was broken due to changing NV12 to be continuous in memory, this patch changes the algorithm for getting Y and UV parts according the new implementation. Change-Id: I639471d0159bab5d027d762719e0d97afd1586d6 Signed-off-by: Illia Ragozin <x0160199@ti.com> Signed-off-by: Devaraj Rangasamy <dev@ti.com>
* Include ION headers from <ion/ion.h>Daniel Levin2012-08-282-2/+2
| | | | | | | | | | Reconciling with JB ION headers inclusion style. The ION headers should be included from <ion/ion.h> to avoid ambiguity. Depends on: http://review.omapzoom.org/#/c/23883/ Change-Id: Ie0c93f5b03b67660825ca3ed77af6ca1ceb7fc3e Signed-off-by: Daniel Levin <dendy@ti.com>
* [OMX] Propogate GetParam IndexUseBufferDescriptor to DucatiVidhoon Viswanathan2012-08-231-7/+5
| | | | | | | | | | | | | | | | | | | The ducati patches: https://gerrit.ext.ti.com/gerrit/omap/#change,8128 https://gerrit.ext.ti.com/gerrit/omap/#change,8261 implement Get/Setparam for OMX_TI_IndexUseBufferDescriptor at Ducati. The implementation of OMX_TI_IndexUseBufferDescriptor mandates ducati side OMX components to be aware of attributes such as buffer type to proceed with buffer processing appropriately. This patch propogates the GetParameter call that was so far handled internally within DOMX to ducati OMX component utilizing the GetParameter implementation at components. Now DOMX is aware of component defaults to the structure related to this index. Change-Id: Icad3abb5236c927edda5ad668ac7ef89db92dfef Signed-off-by: Vidhoon Viswanathan <vidhoon@ti.com>
* DOMX: Allow 1D gralloc buffersEmilian Peev2012-08-211-13/+31
| | | | | | | | | | | | | | | | | | | | For CPCAM usecases, gralloc buffers could be 1D also. This patch makes corresponding changes in DOMX to allow this. - In RPC_RegisterBuffer, looks for num_handles returned from rpmsg ioctl call and returns the corresponding valid handles only. Rest are NULL. - While passing gralloc buffers to ducati via UseBuffer call, fetch only number of components as indicated in pBuffer->numFds. Set other place holders to NULL. This patch has a kernel dependency: http://review.omapzoom.org/#/c/26668/ Change-Id: Ief80d45e8cbe4719aca4eaf09200b2034078e688 Signed-off-by: Emilian Peev <epeev@mm-sol.com> Signed-off-by: Vidhoon Viswanathan <vidhoon@ti.com>
* DOMX: Remove stride based classification of 1D or 2D buffersYordan Lilov2012-08-141-258/+17
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | stride value (4096) was used to determine 2D buffers in - PROXY_UseBuffer - PROXY_AllocateBuffer - PROXY_FreeBuffer In case of PROXY_UseBuffer - GrallocPointers, BufferDescriptorVirtual2D are always 2D and EncoderMetadataPointers are always 1D. So the stride check was not required as none of the handled buffer types coexist in 1D and 2D forms. In case of PROXY_AllocateBuffer - Tiler 2D buffers are not implemented and a request for such a buffer was detected using nStride == LINUX_PAGE_SIZE condition. But tiler 2d buffers are presently not used and hence, this check is removed. In case of PROXY_FreeBuffer - BufferDescriptorVirtual2D (always 2D) is handled within the stride check. So the stride check was not required as it does not coexist in 1D and 2D forms. This patch will enable usecases that require 2D buffers of multiple stride values. Previously, the filling of pAuxBuf1 was hampered by the stride check. Change-Id: I31a26442dc8401f18edf5cc23ed0e7697336c72c Signed-off-by: Yordan Lilov <ylilov@mm-sol.com> Signed-off-by: Vidhoon Viswanathan <vidhoon@ti.com>
* Include buffer type specfic (un)registration within RPC (Un)Register APIVidhoon Viswanathan2012-08-071-56/+148
| | | | | | | | | | | | | | | | | | | For (un)registration of 2D ION, Gralloc buffers, multiple calls from PROXY APIs were being made to RPC_(Un)RegisterBuffer for each component of the buffer. To avoid this, all the buffer type specific logic has been moved to (Un)Register APIs. To facilitate this, the prototype of these APIs have been changed accordingly. Minor change: Previously OMX_IOCPVRREGISTER was used to enable registration implementation for gralloc buffers (to not affect REL DBs). Now it can be replaced by ENABLE_GRALLOC_BUFFERS which is the conventional macro within DOMX for compile time enablement of gralloc buffers. Change-Id: I90ac8e0eb350a53e4f9aff5ecaeb66b5598b6509 Signed-off-by: Vidhoon Viswanathan <vidhoon@ti.com>
* DOMX: Fix to release the shared ion bufferPrasanna Kumar M.R2012-07-201-3/+7
| | | | | | | | | | | In secure playback case, we dont map the ion buffers. We allocate the buffer and perform ion share to share it across to other components. In this case, we also need to close the fd inadditional to the existing ion free call inorder to ensure that buffer gets released as otherwise the reference count is still valid in ion driver and is not freed. Change-Id: I621077e132af55ada2a5ab0c7f4a01ed7afc7d02 Signed-off-by: Prasanna Kumar M.R. <prasanna.kumarmr@ti.com>
* DOMX: multi instance adaptation of OMX buffer monitoring traceFrederic Turgis2012-07-131-1/+3
| | | | | | | | | | | | | | * simple alignment with Ducati instrumentation + OMX component index is added to component name to monitor several instances in parallel, i.e. DECODE->DECODE0, H264E->H264E1 * 2 minor issues fixed: + cast timestamp to long long to avoid negative values after 2^32 us elapsed (1 hour testing) + strncpy does not put \0 at end of string in all cases Change-Id: Ieb07d4900b5ac43295dc95328c4aef954be4f2e5 Signed-off-by: Frederic Turgis <f-turgis@ti.com>
* DOMX: Changes for secure playback ION handle mappingPrasanna Kumar M.R2012-07-021-3/+10
| | | | | | | | | | In secure playback case, we dont map the ion buffers. When we pass the Ion handle, the Rpmsg fails in buffer registration. Hence we need to share the ion buffer and then pass it to Rpmsg/Ducati. Change-Id: I2bea9323aa557d340c2a2c2d653fd5f0480160a4 Signed-off-by: Prasanna Kumar M.R. <prasanna.kumarmr@ti.com> Ported-by: Bryan Buckley <bryan.buckley@ti.com>
* DOMX: framework to dynamically enable/disable perf trace + OMX buffer events ↵Frederic Turgis2012-06-225-3/+407
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | trace Added framework to control activation of perf trace: - set DEBUG_DOMX_KPI_STATUS environment variable - or set debug.domx.kpi_status android property. Env variable takes precedence kpi_status is a bitfield to control enabling of various perf traces. Currently only OMX buffers events are available with bitfield value of 1. It will trace EmptyThisBuffer, FillThisBuffer, EmptyBufferDone, FillBufferDone. It is aligned with Ducati equivalent trace. Enabling the 2 allows computing MPU<->Ducati latency Following parameters are included for each message: - message type: ETB, FTB, EBD, FBD - OMX component name: DECODE, H264E, CAMERA, ... - number of such requests for this component - time in us, from clock_gettime(CLOCK_MONOTONIC, ...) - Remote buffer header to match Ducati trace - FBD nTimeStamp metadata to identify an FBD buffer among framework SW layers visual post-processing is described at http://opbuwiki.dal.design.ti.com/index.php/Benelli-Ducati_performance_instrumentation It is the same for MPU or Ducati trace Example: - setprop debug.domx.trace_level 3 (perf trace is trace level >= 3) - setprop debug.domx.kpi_status 1 - logcat DOMX:D *:S Change-Id: I3b8a996fb861ba84a9f96c98285a30209880423a Signed-off-by: Frederic Turgis <f-turgis@ti.com>
* DOMX: Update trace level at component init and introduce PROFILING levelFrederic Turgis2012-06-222-0/+3
| | | | | | | | | | | | | | | | | | * Profiling trace level introduced after WARN and before INFO levels * Cleaning of trace level usage through enum * "Trace level update" functionality called at every proxy common init * Trace level can be updated through Android property debug.domx.trace_level. Environment variable still has precedence over Android property To modify trace level, do on terminal either: - export TIMM_OSAL_TRACE_DEBUG_LEVEL=<trace level> - setprop debug.domx.trace_level <trace level> (Android only) For example setprop debug.domx.trace_level 3 to set PROFILING level The new value is taken into account only at init of a new component. Change-Id: I04de66a6160c11e9339a8f5a63ca0676755a17e0 Signed-off-by: Frederic Turgis <f-turgis@ti.com>
* [DOMX]: Fix error handling in RPC_sendPacket_syncVidhoon Viswanathan2012-06-201-3/+7
| | | | | | | | | | | | | | | | | | Ducati faulty state is determined by the following inference condition: if(status<0 && errno==ENXIO) { //ducati in faulty state } After write call, errno was checked in wrong manner. Also, write call returns number of bytes written. This patch checks errno in case of write call failure and detects faulty state of ducati. And it uses status == nPacketSize as a tighter condition to validate write call success. Change-Id: I2062e128c4754b0faef2c33ca9465129105ade0c Signed-off-by: Vidhoon Viswanathan <vidhoon@ti.com>
* DOMX: Use Tiler-1D buffer when ION 1D allocation failsDevaraj Rangasamy2012-06-181-1/+12
| | | | | | | | | | | By default Camera proxy and DOMX components uses ION 1D heap for buffer allocation. For 512MB system configuration, ION 1D heap maynot available. So use tiler 1D buffers as fallback mechanism. Applicable for both 1GB and 512MB configuration. Change-Id: Iff9b264f13d03f99975c4ff9dce9d474edeac34c Signed-off-by: Devaraj Rangasamy <dev@ti.com>
* proxy common: make register/unregister conditionalTyler Luu2012-05-241-17/+29
| | | | | | | | | | | | since some kernels do not support OMX_IOCPVRREGISTER, we are making the compilation on whether or not it is defined. If OMX_IOCPVRREGISTER is defined in FS but return ENOTTY, this means that kernel support for this new ioctl has not been merged yet. Change-Id: I691a0ba53a53c443efa5341dae93f429f60705d0 Signed-off-by: Tyler Luu <tluu@ti.com>
* proxy: add support for sgx register ioctlTyler Luu2012-05-241-24/+38
| | | | | | | | | | | | | | rpmsg_omx requires clients to register gralloc (sgx) buffers using seperate ioctl (OMX_IOCPVRREGISTER). Handles received from regsiter need to be released using OMX_IOCIONUNREGISTER when done. the exploit of passing ion handles is no longer supported directly to rpmsg is no longer supported. rpmsg mappings will fail if handles are not first registered. Change-Id: I8d2bad7a9a6a7c34a96bef494d4327e988c78e8b Signed-off-by: Tyler Luu <tluu@ti.com>
* proxy common: choose buffer 1st in AllocateBufferTyler Luu2012-05-241-19/+21
| | | | | | | | | | | Moving choose buffer logic to the beginning of AllocateBuffer so we can use it to index tBufList array when saving ion handle information Fixes an issue missed by "domx: fix video record" Change-Id: Icf5ddcda8b2ffdc67f5ac337b4e7b683b90c053f Signed-off-by: Tyler Luu <tluu@ti.com>
* domx: pass registered buffer address to FreeBufferTyler Luu2012-05-241-0/+3
| | | | | | | | | Pass registered buffer address to RPC_FreeBuffer in PROXY_FreeBuffer. rpmsg expects a registered address to be passed to it when mapping is required. Change-Id: I50abeaf5a7cfc31298832294648a0b8991cea3ad Signed-off-by: Tyler Luu <tluu@ti.com>
* domx: fix video recordTyler Luu2012-05-242-44/+53
| | | | | | | | | | | | | With cpcam kernel changes for rpmsg driver, rpmsg omx is requiring all buffers be registered and unregistered starting here: http://review.omapzoom.org/#/c/20916 Since encoders are using this little trick of dynamically reassigning handles for each input buffer, we will need to register/unregister the handles every ETB. Change-Id: Iaefcd8d3ba56605edfc398814bdec9a6524fa196 Signed-off-by: Tyler Luu <tluu@ti.com>
* domx: add register/unregister for buffersTyler Luu2012-05-212-33/+239
| | | | | | | | | need to register/unregister ion and gralloc handles with rpmsg before sending. this includes shared buffers used in get/set config and parameters. Change-Id: I23e4c09349c9c4f87c06b65e270b320dba9f0372 Signed-off-by: Tyler Luu <tluu@ti.com>
* domx: fixes for reprocessingTyler Luu2012-05-211-0/+9
| | | | | | | | | | | | | | 1. For OMX_TI_IndexUseNativeBuffers, properly set gralloc usage for false case. If client set this index to true, a subsequent set to false was not being handled causing domx to misintrepret handles. 2. For OMX_TI_IndexUseBufferDescriptor, RPC SetParameters needs to be called so Ducati also knows about the descriptor. This index was previously only being handled internally. Change-Id: I928dda6d5ae9b76682c3b60dce4375a1accf09e4 Signed-off-by: Tyler Luu <tluu@ti.com>
* DOMX: Pass the correct buf address in PROXY_FreeBufferPavel Nedev2012-04-031-2/+26
| | | | | | | | This fixes the address passed from PROXY_FreeBuffer to RPC_FreeBuffer to be the same that is passed to RPC_UseBuffer from PROXY_UseBuffer. Change-Id: I8581ec7a6abfafcb8ef11dd360a6114ba4d704d2 Signed-off-by: Pavel Nedev <pnedev@mm-sol.com>
* [DOMX] Fix Buffer Map type during free bufferPradeep Venkatasubbarao2012-04-031-1/+4
| | | | | | | | | | | | | | | | | The buffer sent from RPC calls (UseBuffer, AllocateBuffer, FreeBuffer) will get mapped to approriate space depending on the map type that is set in the packet. Even though a pBuffer was sent in the packet, the map type (RPC_OMX_MAP_INFO_TYPE) was set to RPC_OMX_MAP_INFO_NONE. This resulted in conversion not being done in rpmsg. This patch sets the RPC_OMX_MAP_INFO_TYPE correctly in the packet to RPC_OMX_MAP_INFO_ONE_BUF. With this rpmsg will convert the pBuffer address to ducati side (tiler) address. Change-Id: I722b618687e178817e2e8ffdb9670e7a5f54d5e9 Signed-off-by: Pradeep Venkatasubbarao <pradeepv@ti.com>
generated by cgit v1.1 at 2025-03-08 08:23:15 +0000