[PATCH v3] media: iris: Refine internal buffer reconfiguration logic for resolution change
Improve the condition used to determine when input internal buffers need to be reconfigured during streamon on the capture port. Previously, the check relied on the INPUT_PAUSE sub-state, which was also being set during seek operations. This led to input buffers being queued multiple times to the firmware, causing session errors due to duplicate buffer submissions.
This change introduces a more accurate check using the FIRST_IPSC and DRC sub-states to ensure that input buffer reconfiguration is triggered only during resolution change scenarios, such as streamoff/on on the capture port. This avoids duplicate buffer queuing during seek operations.
Fixes: c1f8b2cc72ec ("media: iris: handle streamoff/on from client in dynamic resolution change") Cc: stable@vger.kernel.org Reported-by: Val Packett val@packett.cool Closes: https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/4700 Signed-off-by: Dikshita Agarwal dikshita.agarwal@oss.qualcomm.com --- Changes in v3: - Fixed the compilation issue - Added stable@vger.kernel.org in Cc - Link to v2: https://lore.kernel.org/r/20251104-iris-seek-fix-v2-1-c9dace39b43d@oss.qualc...
Changes in v2: - Removed spurious space and addressed other comments (Nicolas) - Remove the unnecessary initializations (Self) - Link to v1: https://lore.kernel.org/r/20251103-iris-seek-fix-v1-1-6db5f5e17722@oss.qualc... --- drivers/media/platform/qcom/iris/iris_common.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)
diff --git a/drivers/media/platform/qcom/iris/iris_common.c b/drivers/media/platform/qcom/iris/iris_common.c index 9fc663bdaf3fc989fe1273b4d4280a87f68de85d..7f1c7fe144f707accc2e3da65ce37cd6d9dfeaff 100644 --- a/drivers/media/platform/qcom/iris/iris_common.c +++ b/drivers/media/platform/qcom/iris/iris_common.c @@ -91,12 +91,14 @@ int iris_process_streamon_input(struct iris_inst *inst) int iris_process_streamon_output(struct iris_inst *inst) { const struct iris_hfi_command_ops *hfi_ops = inst->core->hfi_ops; - bool drain_active = false, drc_active = false; enum iris_inst_sub_state clear_sub_state = 0; + bool drain_active, drc_active, first_ipsc; int ret = 0;
iris_scale_power(inst);
+ first_ipsc = inst->sub_state & IRIS_INST_SUB_FIRST_IPSC; + drain_active = inst->sub_state & IRIS_INST_SUB_DRAIN && inst->sub_state & IRIS_INST_SUB_DRAIN_LAST;
@@ -108,7 +110,8 @@ int iris_process_streamon_output(struct iris_inst *inst) else if (drain_active) clear_sub_state = IRIS_INST_SUB_DRAIN | IRIS_INST_SUB_DRAIN_LAST;
- if (inst->domain == DECODER && inst->sub_state & IRIS_INST_SUB_INPUT_PAUSE) { + /* Input internal buffer reconfiguration required in case of resolution change */ + if (first_ipsc || drc_active) { ret = iris_alloc_and_queue_input_int_bufs(inst); if (ret) return ret;
--- base-commit: 163917839c0eea3bdfe3620f27f617a55fd76302 change-id: 20251103-iris-seek-fix-7a25af22fa52
Best regards,
On 11/5/2025 11:17 AM, Dikshita Agarwal wrote:
Improve the condition used to determine when input internal buffers need to be reconfigured during streamon on the capture port. Previously, the check relied on the INPUT_PAUSE sub-state, which was also being set during seek operations. This led to input buffers being queued multiple times to the firmware, causing session errors due to duplicate buffer submissions.
This change introduces a more accurate check using the FIRST_IPSC and DRC sub-states to ensure that input buffer reconfiguration is triggered only during resolution change scenarios, such as streamoff/on on the capture port. This avoids duplicate buffer queuing during seek operations.
Fixes: c1f8b2cc72ec ("media: iris: handle streamoff/on from client in dynamic resolution change") Cc:stable@vger.kernel.org Reported-by: Val Packettval@packett.cool Closes:https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/4700 Signed-off-by: Dikshita Agarwaldikshita.agarwal@oss.qualcomm.com
Reviewed-by: Vikash Garodiavikash.garodia@oss.qualcomm.com
On Wed, Nov 05, 2025 at 11:17:37AM +0530, Dikshita Agarwal wrote:
Improve the condition used to determine when input internal buffers need to be reconfigured during streamon on the capture port. Previously, the check relied on the INPUT_PAUSE sub-state, which was also being set during seek operations. This led to input buffers being queued multiple times to the firmware, causing session errors due to duplicate buffer submissions.
This change introduces a more accurate check using the FIRST_IPSC and DRC sub-states to ensure that input buffer reconfiguration is triggered only during resolution change scenarios, such as streamoff/on on the capture port. This avoids duplicate buffer queuing during seek operations.
Fixes: c1f8b2cc72ec ("media: iris: handle streamoff/on from client in dynamic resolution change") Cc: stable@vger.kernel.org Reported-by: Val Packett val@packett.cool Closes: https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/4700 Signed-off-by: Dikshita Agarwal dikshita.agarwal@oss.qualcomm.com
Changes in v3:
- Fixed the compilation issue
- Added stable@vger.kernel.org in Cc
- Link to v2: https://lore.kernel.org/r/20251104-iris-seek-fix-v2-1-c9dace39b43d@oss.qualc...
Changes in v2:
- Removed spurious space and addressed other comments (Nicolas)
- Remove the unnecessary initializations (Self)
- Link to v1: https://lore.kernel.org/r/20251103-iris-seek-fix-v1-1-6db5f5e17722@oss.qualc...
drivers/media/platform/qcom/iris/iris_common.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)
diff --git a/drivers/media/platform/qcom/iris/iris_common.c b/drivers/media/platform/qcom/iris/iris_common.c index 9fc663bdaf3fc989fe1273b4d4280a87f68de85d..7f1c7fe144f707accc2e3da65ce37cd6d9dfeaff 100644 --- a/drivers/media/platform/qcom/iris/iris_common.c +++ b/drivers/media/platform/qcom/iris/iris_common.c @@ -91,12 +91,14 @@ int iris_process_streamon_input(struct iris_inst *inst) int iris_process_streamon_output(struct iris_inst *inst) { const struct iris_hfi_command_ops *hfi_ops = inst->core->hfi_ops;
- bool drain_active = false, drc_active = false; enum iris_inst_sub_state clear_sub_state = 0;
- bool drain_active, drc_active, first_ipsc; int ret = 0;
iris_scale_power(inst);
- first_ipsc = inst->sub_state & IRIS_INST_SUB_FIRST_IPSC;
- drain_active = inst->sub_state & IRIS_INST_SUB_DRAIN && inst->sub_state & IRIS_INST_SUB_DRAIN_LAST;
@@ -108,7 +110,8 @@ int iris_process_streamon_output(struct iris_inst *inst) else if (drain_active) clear_sub_state = IRIS_INST_SUB_DRAIN | IRIS_INST_SUB_DRAIN_LAST;
- if (inst->domain == DECODER && inst->sub_state & IRIS_INST_SUB_INPUT_PAUSE) {
- /* Input internal buffer reconfiguration required in case of resolution change */
- if (first_ipsc || drc_active) { ret = iris_alloc_and_queue_input_int_bufs(inst); if (ret) return ret;
I will repeat my (unanswered) question from v2:
After this line comes manual writing of STAGE and PIPE. Could you please point out where is the driver updating the resolution in the firmware? And if it does, why do we need to write STAGE and PIPE again?
On 11/10/2025 2:46 AM, Dmitry Baryshkov wrote:
On Wed, Nov 05, 2025 at 11:17:37AM +0530, Dikshita Agarwal wrote:
Improve the condition used to determine when input internal buffers need to be reconfigured during streamon on the capture port. Previously, the check relied on the INPUT_PAUSE sub-state, which was also being set during seek operations. This led to input buffers being queued multiple times to the firmware, causing session errors due to duplicate buffer submissions.
This change introduces a more accurate check using the FIRST_IPSC and DRC sub-states to ensure that input buffer reconfiguration is triggered only during resolution change scenarios, such as streamoff/on on the capture port. This avoids duplicate buffer queuing during seek operations.
Fixes: c1f8b2cc72ec ("media: iris: handle streamoff/on from client in dynamic resolution change") Cc: stable@vger.kernel.org Reported-by: Val Packett val@packett.cool Closes: https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/4700 Signed-off-by: Dikshita Agarwal dikshita.agarwal@oss.qualcomm.com
Changes in v3:
- Fixed the compilation issue
- Added stable@vger.kernel.org in Cc
- Link to v2: https://lore.kernel.org/r/20251104-iris-seek-fix-v2-1-c9dace39b43d@oss.qualc...
Changes in v2:
- Removed spurious space and addressed other comments (Nicolas)
- Remove the unnecessary initializations (Self)
- Link to v1: https://lore.kernel.org/r/20251103-iris-seek-fix-v1-1-6db5f5e17722@oss.qualc...
drivers/media/platform/qcom/iris/iris_common.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)
diff --git a/drivers/media/platform/qcom/iris/iris_common.c b/drivers/media/platform/qcom/iris/iris_common.c index 9fc663bdaf3fc989fe1273b4d4280a87f68de85d..7f1c7fe144f707accc2e3da65ce37cd6d9dfeaff 100644 --- a/drivers/media/platform/qcom/iris/iris_common.c +++ b/drivers/media/platform/qcom/iris/iris_common.c @@ -91,12 +91,14 @@ int iris_process_streamon_input(struct iris_inst *inst) int iris_process_streamon_output(struct iris_inst *inst) { const struct iris_hfi_command_ops *hfi_ops = inst->core->hfi_ops;
- bool drain_active = false, drc_active = false; enum iris_inst_sub_state clear_sub_state = 0;
- bool drain_active, drc_active, first_ipsc; int ret = 0;
iris_scale_power(inst);
- first_ipsc = inst->sub_state & IRIS_INST_SUB_FIRST_IPSC;
- drain_active = inst->sub_state & IRIS_INST_SUB_DRAIN && inst->sub_state & IRIS_INST_SUB_DRAIN_LAST;
@@ -108,7 +110,8 @@ int iris_process_streamon_output(struct iris_inst *inst) else if (drain_active) clear_sub_state = IRIS_INST_SUB_DRAIN | IRIS_INST_SUB_DRAIN_LAST;
- if (inst->domain == DECODER && inst->sub_state & IRIS_INST_SUB_INPUT_PAUSE) {
- /* Input internal buffer reconfiguration required in case of resolution change */
- if (first_ipsc || drc_active) { ret = iris_alloc_and_queue_input_int_bufs(inst); if (ret) return ret;
I will repeat my (unanswered) question from v2:
After this line comes manual writing of STAGE and PIPE. Could you please point out where is the driver updating the resolution in the firmware? And if it does, why do we need to write STAGE and PIPE again?
Sorry for late reply,
During streamon on the output port, the driver sets the resolution in the firmware. However, during Dynamic Resolution Change (DRC), the resolution update originates from the firmware and is communicated to the driver. As a result, the driver does not proactively update the resolution in the firmware during DRC.
STAGE parameter depends on the resolution, the driver must update the firmware with the new STAGE value after a resolution change to ensure proper operation.
On the other hand, the PIPE value is independent of resolution. It is typically updated to 1 for interlaced content, which is identified during the sequence change. Currently, the Iris driver does not support interlaced content, so updating the PIPE value during DRC handling is redundant. However, this update is harmless and will be necessary once interlace support is added in the future.
Thanks, Dikshita
On Mon, Nov 10, 2025 at 02:31:20PM +0530, Dikshita Agarwal wrote:
On 11/10/2025 2:46 AM, Dmitry Baryshkov wrote:
On Wed, Nov 05, 2025 at 11:17:37AM +0530, Dikshita Agarwal wrote:
Improve the condition used to determine when input internal buffers need to be reconfigured during streamon on the capture port. Previously, the check relied on the INPUT_PAUSE sub-state, which was also being set during seek operations. This led to input buffers being queued multiple times to the firmware, causing session errors due to duplicate buffer submissions.
This change introduces a more accurate check using the FIRST_IPSC and DRC sub-states to ensure that input buffer reconfiguration is triggered only during resolution change scenarios, such as streamoff/on on the capture port. This avoids duplicate buffer queuing during seek operations.
Fixes: c1f8b2cc72ec ("media: iris: handle streamoff/on from client in dynamic resolution change") Cc: stable@vger.kernel.org Reported-by: Val Packett val@packett.cool Closes: https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/4700 Signed-off-by: Dikshita Agarwal dikshita.agarwal@oss.qualcomm.com
Changes in v3:
- Fixed the compilation issue
- Added stable@vger.kernel.org in Cc
- Link to v2: https://lore.kernel.org/r/20251104-iris-seek-fix-v2-1-c9dace39b43d@oss.qualc...
Changes in v2:
- Removed spurious space and addressed other comments (Nicolas)
- Remove the unnecessary initializations (Self)
- Link to v1: https://lore.kernel.org/r/20251103-iris-seek-fix-v1-1-6db5f5e17722@oss.qualc...
drivers/media/platform/qcom/iris/iris_common.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)
diff --git a/drivers/media/platform/qcom/iris/iris_common.c b/drivers/media/platform/qcom/iris/iris_common.c index 9fc663bdaf3fc989fe1273b4d4280a87f68de85d..7f1c7fe144f707accc2e3da65ce37cd6d9dfeaff 100644 --- a/drivers/media/platform/qcom/iris/iris_common.c +++ b/drivers/media/platform/qcom/iris/iris_common.c @@ -91,12 +91,14 @@ int iris_process_streamon_input(struct iris_inst *inst) int iris_process_streamon_output(struct iris_inst *inst) { const struct iris_hfi_command_ops *hfi_ops = inst->core->hfi_ops;
- bool drain_active = false, drc_active = false; enum iris_inst_sub_state clear_sub_state = 0;
- bool drain_active, drc_active, first_ipsc; int ret = 0;
iris_scale_power(inst);
- first_ipsc = inst->sub_state & IRIS_INST_SUB_FIRST_IPSC;
- drain_active = inst->sub_state & IRIS_INST_SUB_DRAIN && inst->sub_state & IRIS_INST_SUB_DRAIN_LAST;
@@ -108,7 +110,8 @@ int iris_process_streamon_output(struct iris_inst *inst) else if (drain_active) clear_sub_state = IRIS_INST_SUB_DRAIN | IRIS_INST_SUB_DRAIN_LAST;
- if (inst->domain == DECODER && inst->sub_state & IRIS_INST_SUB_INPUT_PAUSE) {
- /* Input internal buffer reconfiguration required in case of resolution change */
- if (first_ipsc || drc_active) { ret = iris_alloc_and_queue_input_int_bufs(inst); if (ret) return ret;
I will repeat my (unanswered) question from v2:
After this line comes manual writing of STAGE and PIPE. Could you please point out where is the driver updating the resolution in the firmware? And if it does, why do we need to write STAGE and PIPE again?
Sorry for late reply,
During streamon on the output port, the driver sets the resolution in the firmware. However, during Dynamic Resolution Change (DRC), the resolution update originates from the firmware and is communicated to the driver. As a result, the driver does not proactively update the resolution in the firmware during DRC.
STAGE parameter depends on the resolution, the driver must update the firmware with the new STAGE value after a resolution change to ensure proper operation.
On the other hand, the PIPE value is independent of resolution. It is typically updated to 1 for interlaced content, which is identified during the sequence change. Currently, the Iris driver does not support interlaced content, so updating the PIPE value during DRC handling is redundant. However, this update is harmless and will be necessary once interlace support is added in the future.
Ack, thanks for the explanation.
On Wed, Nov 05, 2025 at 11:17:37AM +0530, Dikshita Agarwal wrote:
Improve the condition used to determine when input internal buffers need to be reconfigured during streamon on the capture port. Previously, the check relied on the INPUT_PAUSE sub-state, which was also being set during seek operations. This led to input buffers being queued multiple times to the firmware, causing session errors due to duplicate buffer submissions.
This change introduces a more accurate check using the FIRST_IPSC and DRC sub-states to ensure that input buffer reconfiguration is triggered only during resolution change scenarios, such as streamoff/on on the capture port. This avoids duplicate buffer queuing during seek operations.
Fixes: c1f8b2cc72ec ("media: iris: handle streamoff/on from client in dynamic resolution change") Cc: stable@vger.kernel.org Reported-by: Val Packett val@packett.cool Closes: https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/4700 Signed-off-by: Dikshita Agarwal dikshita.agarwal@oss.qualcomm.com
Changes in v3:
- Fixed the compilation issue
- Added stable@vger.kernel.org in Cc
- Link to v2: https://lore.kernel.org/r/20251104-iris-seek-fix-v2-1-c9dace39b43d@oss.qualc...
Changes in v2:
- Removed spurious space and addressed other comments (Nicolas)
- Remove the unnecessary initializations (Self)
- Link to v1: https://lore.kernel.org/r/20251103-iris-seek-fix-v1-1-6db5f5e17722@oss.qualc...
drivers/media/platform/qcom/iris/iris_common.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)
diff --git a/drivers/media/platform/qcom/iris/iris_common.c b/drivers/media/platform/qcom/iris/iris_common.c index 9fc663bdaf3fc989fe1273b4d4280a87f68de85d..7f1c7fe144f707accc2e3da65ce37cd6d9dfeaff 100644 --- a/drivers/media/platform/qcom/iris/iris_common.c +++ b/drivers/media/platform/qcom/iris/iris_common.c @@ -91,12 +91,14 @@ int iris_process_streamon_input(struct iris_inst *inst) int iris_process_streamon_output(struct iris_inst *inst) { const struct iris_hfi_command_ops *hfi_ops = inst->core->hfi_ops;
- bool drain_active = false, drc_active = false; enum iris_inst_sub_state clear_sub_state = 0;
- bool drain_active, drc_active, first_ipsc; int ret = 0;
iris_scale_power(inst);
- first_ipsc = inst->sub_state & IRIS_INST_SUB_FIRST_IPSC;
- drain_active = inst->sub_state & IRIS_INST_SUB_DRAIN && inst->sub_state & IRIS_INST_SUB_DRAIN_LAST;
@@ -108,7 +110,8 @@ int iris_process_streamon_output(struct iris_inst *inst) else if (drain_active) clear_sub_state = IRIS_INST_SUB_DRAIN | IRIS_INST_SUB_DRAIN_LAST;
- if (inst->domain == DECODER && inst->sub_state & IRIS_INST_SUB_INPUT_PAUSE) {
- /* Input internal buffer reconfiguration required in case of resolution change */
- if (first_ipsc || drc_active) {
Another question: can this now result in PIPE being sent for the ENCODER instance?
ret = iris_alloc_and_queue_input_int_bufs(inst); if (ret) return ret;
base-commit: 163917839c0eea3bdfe3620f27f617a55fd76302 change-id: 20251103-iris-seek-fix-7a25af22fa52
Best regards,
Dikshita Agarwal dikshita.agarwal@oss.qualcomm.com
On 11/10/2025 7:05 PM, Dmitry Baryshkov wrote:
On Wed, Nov 05, 2025 at 11:17:37AM +0530, Dikshita Agarwal wrote:
Improve the condition used to determine when input internal buffers need to be reconfigured during streamon on the capture port. Previously, the check relied on the INPUT_PAUSE sub-state, which was also being set during seek operations. This led to input buffers being queued multiple times to the firmware, causing session errors due to duplicate buffer submissions.
This change introduces a more accurate check using the FIRST_IPSC and DRC sub-states to ensure that input buffer reconfiguration is triggered only during resolution change scenarios, such as streamoff/on on the capture port. This avoids duplicate buffer queuing during seek operations.
Fixes: c1f8b2cc72ec ("media: iris: handle streamoff/on from client in dynamic resolution change") Cc: stable@vger.kernel.org Reported-by: Val Packett val@packett.cool Closes: https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/4700 Signed-off-by: Dikshita Agarwal dikshita.agarwal@oss.qualcomm.com
Changes in v3:
- Fixed the compilation issue
- Added stable@vger.kernel.org in Cc
- Link to v2: https://lore.kernel.org/r/20251104-iris-seek-fix-v2-1-c9dace39b43d@oss.qualc...
Changes in v2:
- Removed spurious space and addressed other comments (Nicolas)
- Remove the unnecessary initializations (Self)
- Link to v1: https://lore.kernel.org/r/20251103-iris-seek-fix-v1-1-6db5f5e17722@oss.qualc...
drivers/media/platform/qcom/iris/iris_common.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)
diff --git a/drivers/media/platform/qcom/iris/iris_common.c b/drivers/media/platform/qcom/iris/iris_common.c index 9fc663bdaf3fc989fe1273b4d4280a87f68de85d..7f1c7fe144f707accc2e3da65ce37cd6d9dfeaff 100644 --- a/drivers/media/platform/qcom/iris/iris_common.c +++ b/drivers/media/platform/qcom/iris/iris_common.c @@ -91,12 +91,14 @@ int iris_process_streamon_input(struct iris_inst *inst) int iris_process_streamon_output(struct iris_inst *inst) { const struct iris_hfi_command_ops *hfi_ops = inst->core->hfi_ops;
- bool drain_active = false, drc_active = false; enum iris_inst_sub_state clear_sub_state = 0;
- bool drain_active, drc_active, first_ipsc; int ret = 0;
iris_scale_power(inst);
- first_ipsc = inst->sub_state & IRIS_INST_SUB_FIRST_IPSC;
- drain_active = inst->sub_state & IRIS_INST_SUB_DRAIN && inst->sub_state & IRIS_INST_SUB_DRAIN_LAST;
@@ -108,7 +110,8 @@ int iris_process_streamon_output(struct iris_inst *inst) else if (drain_active) clear_sub_state = IRIS_INST_SUB_DRAIN | IRIS_INST_SUB_DRAIN_LAST;
- if (inst->domain == DECODER && inst->sub_state & IRIS_INST_SUB_INPUT_PAUSE) {
- /* Input internal buffer reconfiguration required in case of resolution change */
- if (first_ipsc || drc_active) {
Another question: can this now result in PIPE being sent for the ENCODER instance?
This state check will never be true for Encoder.
Thanks, Dikshita
ret = iris_alloc_and_queue_input_int_bufs(inst); if (ret) return ret;
base-commit: 163917839c0eea3bdfe3620f27f617a55fd76302 change-id: 20251103-iris-seek-fix-7a25af22fa52
Best regards,
Dikshita Agarwal dikshita.agarwal@oss.qualcomm.com
linux-stable-mirror@lists.linaro.org
-
Dikshita Agarwal -
Dmitry Baryshkov -
Vikash Garodia