Align hfi property and memory file to pick interface changes
for encoder ring buffer support
Change-Id: Icde4cdffe5275d416c51c36e562bb832027b0fa1
Signed-off-by: Akshata Sahukar <quic_asahukar@quicinc.com>
1. Move DPB_TAG_LIST from ouput meta port to
input meta port;
2. Remove meta_max_reorder_num_frame;
Change-Id: I1783ee9cdb1092f90706dc485d247c51f36a9a65
Signed-off-by: Zhongbo Shi <quic_zhongbos@quicinc.com>
1. Move DPB_LIST subscription on input
port instead of output port.
2. Move decoder port PSC and properties
to platform data.
Change-Id: I5711e0ba915ebe4136f64d1440ae331b787f4d61
Signed-off-by: Zhongbo Shi <quic_zhongbos@quicinc.com>
Align hfi property and registers file with to pick synx fence
related interface changes
Change-Id: I19e5cce4d67f2f1eeacf917f9e688d2bb20f6585
Signed-off-by: Akshata Sahukar <quic_asahukar@quicinc.com>
In msm_vidc_open() create_singlethread_workqueue() is failing,
so msm_vidc_close() is called and i.e attempting to de-initialize
some uninitialized structures and resulting into stability issues.
So handle msm_vidc_open() failure with in same api, instead of
calling msm_vidc_close() and re-organize deinit sequence.
Change-Id: I843cf07eaf18f4ea764842fd1c87b78d115580d3
Signed-off-by: Govindaraj Rajagopal <quic_grajagop@quicinc.com>
In case of decode batch IDR frame can impact
the bitrate based clk scaling.
To tackle this use avg filled len in case
of decode batching.
Change-Id: I1b2701857118bfa1f1a4fd0c0498ee016c02e8d5
Signed-off-by: Ankush Mitra <quic_ankumitr@quicinc.com>
If there is any system error from firmware then driver will
remove instance from core->instances.list.
msm_vb2_detach(unmap)_dmabuf() returning failure because
get_inst_ref() returning NULL as the instance was removed from
in core->instances.list, but instance is still alive and buffers
can still be detached(unmapped). Skipping detach(unmap)
buffers resulting in below warning when the buffer's final
dma reference is removed by v4l2 framework (dma_buf_put(fd))
after msm_vb2_detach(unmap)_dmabuf() functions.
msm_dma_buf_freed: DMA buffer ffffff8884ebe600 destroyed with
outstanding iommu mappings
Call trace:
msm_dma_buf_freed+0x130/0x334 [msm_dma_iommu_mapping]
qcom_sg_release+0x30/0x50 [qcom_dma_heaps]
dma_buf_release+0x48/0xa8
__dentry_kill+0x178/0x268
dentry_kill+0xc4/0x110
dput+0xa4/0x110
__fput+0x170/0x260
____fput+0x18/0x28
task_work_run+0xc0/0xe8
Eventually when client closed the session, driver will try to
cleanup the missing buffers and hence driver calls unmap buffers
which is resulting in below list_del corruption issue as dma_buf
including sg_list was already removed for it
msm_dma_unmap_sg_attrs: (ffffff8916a92c00) was never mapped
WARNING: CPU: 2 PID: 2258 at drivers/iommu/msm_dma_iommu_mapping.c:377
msm_dma_unmap_sg_attrs+0x9c/0x1e4 [msm_dma_iommu_mapping]
Call trace:
msm_dma_unmap_sg_attrs+0x9c/0x1e4 [msm_dma_iommu_mapping]
qcom_sg_unmap_dma_buf+0xc0/0xf0 [qcom_dma_heaps]
dma_buf_unmap_attachment+0x48/0xa0
msm_vidc_dma_buf_unmap_attachment+0x2c/0x80 [msm_video]
msm_vidc_destroy_buffers+0x510/0xb74 [msm_video]
put_inst+0x158/0x3fc [msm_video]
msm_vidc_close+0xe4/0x16c [msm_video]
v4l2_release+0x7c/0x108
__fput+0xe4/0x260
____fput+0x18/0x28
task_work_run+0xc0/0xe8
list_del corruption. prev->next should be ffffff8891db3c90,
but was ffffff8916a92c10. (prev=ffffff8916a92c10)
As we know instance will always be alive in
msm_vb2_detach(unmap)_dmabuf() function, we can safely
remove get_inst_ref() to avoid skipping detach(unmap)
calls to avoid list_del corruption issues.
Change-Id: If292a39ab1cde17da3cbb634370b6ff6692e496a
Signed-off-by: Maheshwar Ajja <quic_majja@quicinc.com>
New hfi buffer flags to send memory region information to firmware.
Change-Id: I3a320622d0ed01a5e583dbeceb2648a57f618a66
Signed-off-by: Chinmay Sawarkar <quic_chinmays@quicinc.com>
use state specific checks inside state_handler instead of
relying global msm_vidc_allow_stop function.
Change-Id: I83d81aa0f840412a4c61f15673a6246f0f84be7f
Signed-off-by: Govindaraj Rajagopal <quic_grajagop@quicinc.com>
use state specific checks inside state_handler instead of
relying global msm_vidc_allow_start function.
Change-Id: I04182ee3b63fa9044acf4f40adc47ea97eaa7538
Signed-off-by: Govindaraj Rajagopal <quic_grajagop@quicinc.com>
use state specific checks inside state_handler instead of
relying global msm_vidc_allow_streamon function.
Change-Id: I67948831612ebb33cdd12a2e9f8561b6095e9433
Signed-off-by: Govindaraj Rajagopal <quic_grajagop@quicinc.com>
use state specific checks inside state_handler instead of
relying global msm_vidc_allow_reqbufs function.
Change-Id: Iaf7f23e4e2ef9b714e3d4c2900c2753d74150fd1
Signed-off-by: Govindaraj Rajagopal <quic_grajagop@quicinc.com>
use state specific checks inside state_handler instead of
relying global msm_vidc_allow_s_fmt function.
Change-Id: I92af08e4cd5d49f763c5a2d381af86eb10a4c061
Signed-off-by: Govindaraj Rajagopal <quic_grajagop@quicinc.com>
remove all conditional compilation in video driver.
With this change, all files willbe always compiled
and will be available in final kernel object file.
Change-Id: I9843c246e23bd1ee4fb8918e5cfa840e2defd432
Signed-off-by: Govindaraj Rajagopal <quic_grajagop@quicinc.com>
Set substate PM_SUSPEND even if
video core was already power collapsed.
Change-Id: I15740d8e44fe89c6b34c0412dca1064539cdac35
Signed-off-by: Ankush Mitra <quic_ankumitr@quicinc.com>
Move _clock_set_flag() function implementation from
resource.c to resource_ext.c file.
Change-Id: I4b977c7f840b94da578fcaac53b8fdc4bba86fca
Signed-off-by: Vedang Nagar <quic_vnagar@quicinc.com>
Introduce d_vpr_hs to print both VIDC_HIGH or
VIDC_STAT log.
Change-Id: Ie7bfa8df033b30689eca6ad6a4637978730eadf7
Signed-off-by: Ankush Mitra <quic_ankumitr@quicinc.com>
get allow_mask from substate allow table and decide to update/ignore
based on that mask.
Change-Id: I1cd465069717561aeadeeba7241128bd450d4933
Signed-off-by: Govindaraj Rajagopal <quic_grajagop@quicinc.com>
In downstream, driver vote for VIDEO_CC_MVS0_CLK_SRC which
needs a multiplication factor of 3 to be applied to
required freq while on upstream we vote for branch clk
i.e. VIDEO_CC_MVS0_CLK hence this multiplication factor is
not needed.
Change-Id: I43131558b3882da303cfaa9bdb5d76080a83fc42
Signed-off-by: Dikshita Agarwal <quic_dikshita@quicinc.com>
1. Revise to config encoder RAW port with buffer actual
aligned width and height in pixels to firmware;
2. Client also needs to config aligned input port
width and height to driver;
3. Driver must config encoder RAW port crops;
4. Client also needs to config encoder input port crops;
Change-Id: Idb85036df9a72e50239ddb46039ef1995b7f0689
Signed-off-by: Zhongbo Shi <quic_zhongbos@quicinc.com>
Reinitialize dmabuf pointer to NULL only after
dma_buf_put has been called on that buffer.
Change-Id: I05e8c1d0901438bc39a9bf9faeed11e228cd7675
Signed-off-by: Darshana Patil <quic_darshana@quicinc.com>
- introduce control to set colorspace. client will use this
to set private color info.
Change-Id: I0099ef1525f4562d3be3a6e518d046fe5b9ef894
Signed-off-by: Darshana Patil <quic_darshana@quicinc.com>
[1] In upstream side, udev is attampting to open video
nodes(video32 & video33) immediately after creation.
But by that time probe might not completed and i.e
resulting to NULL ptr dereference during open() call.
[2] component_match_add_release() must be called only
for video subdevices(contextbanks), but due to OPP table
more entries getting added into component_match table.
So master device bind is never getting invoked and probe
sequence is never getting completed.
Uploaded change to address above mentioned issuses.
Change-Id: I8d2e8fb5644da7076f5f99feda9365629e9130fd
Signed-off-by: Govindaraj Rajagopal <quic_grajagop@quicinc.com>
Allow streamoff from msm_vidc_vb2_queue_deinit after
errors to flush out the pending buffer in driver
in order to avoid v4l2 framework printing below warning.
videobuf2_common: driver bug: stop_streaming operation
is leaving buf in active state.
Call trace
__vb2_queue_cancel+0x214/0x288
vb2_core_queue_release+0x28/0x5c
vb2_queue_release+0x18/0x28
msm_vidc_vb2_queue_deinit+0x40/0x118 [msm_video]
msm_vidc_close+0x7c/0x144 [msm_video]
msm_v4l2_close+0x44/0x164 [msm_video]
There is a deadlock if driver acquires inst lock
in msm_vb2_stop_streaming, because this function
will be called from msm_vidc_close which already
acquired the inst lock. Hence, acquire inst lock
in msm_v4l2_stramoff instead of msm_vb2_stop_streaming.
Change-Id: Iec4c0e416a8a2705af28dbd5138f25d9f3016d12
Signed-off-by: Ankush Mitra <quic_ankumitr@quicinc.com>
Add support to allow 5 LTR count as per new
requirement.
Change-Id: Ic839ea61ae8dfe21bd8891f0f0dbaf26506d84ec
Signed-off-by: Akshata Sahukar <quic_asahukar@quicinc.com>
state handling changes were spread across multiple files.
So re-arrange all state related handling changes into
msm_vidc_state.h/.c file.
Change-Id: I3826daa678d1e2b5ce7e74380d465e70b1b824c6
Signed-off-by: Govindaraj Rajagopal <quic_grajagop@quicinc.com>
