shinys000114/linux
1
0
Fork 0
mirror of https://github.com/hardkernel/linux.git synced 2026-06-07 11:26:02 +09:00
Code Issues Packages Projects Releases Wiki Activity

video: rockchip: mpp: rkvenc2: Fix slice mode issue

Browse Source 
1. Use MPP_CMD_POLL_HW_IRQ to handle the hw slice irq.
   Old MPP_CMD_POLL_HW_FINISH will wait hardware finish and skip hw irq.
2. Clear poll request msgs when it is handled.

NOTE: When MPP_CMD_POLL_HW_IRQ return slice with zero length the
current task is completed. This way can reduce an extra
MPP_CMD_POLL_HW_FINISH ioctl.

Signed-off-by: Herman Chen <herman.chen@rock-chips.com>
Change-Id: I2c81d85c236d4290a497c1747ae2f3e5ac1ce55c
This commit is contained in:
Herman Chen
2022-03-18 17:04:33 +08:00
committed by Tao Huang
parent 298511efd0
commit 6cee981d5a
3 changed files with 39 additions and 19 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
drivers/video/rockchip/mpp/mpp_common.c
View File

@@ -1202,12 +1202,20 @@ static int mpp_process_request(struct mpp_session *session,
case MPP_CMD_POLL_HW_FINISH: {
msgs->flags |= req->flags;
msgs->poll_cnt++;
msgs->poll_req = NULL;
} break;
case MPP_CMD_POLL_HW_IRQ: {
if (msgs->poll_cnt || msgs->poll_req)
mpp_err("Do NOT poll hw irq when previous call not return\n");
msgs->flags |= req->flags;
msgs->poll_cnt++;
if (req->size && req->data) {
if (msgs->poll_req)
mpp_err("can not poll more than one request\n");
else
if (!msgs->poll_req)
msgs->poll_req = req;
} else {
msgs->poll_req = NULL;
}
} break;
case MPP_CMD_RESET_SESSION: {

1
drivers/video/rockchip/mpp/mpp_common.h
View File

@@ -115,6 +115,7 @@ enum MPP_DEV_COMMAND_TYPE {
MPP_CMD_POLL_BASE = 0x300,
MPP_CMD_POLL_HW_FINISH = MPP_CMD_POLL_BASE + 0,
MPP_CMD_POLL_HW_IRQ = MPP_CMD_POLL_BASE + 1,
MPP_CMD_POLL_BUTT,
MPP_CMD_CONTROL_BASE = 0x400,

43
drivers/video/rockchip/mpp/mpp_rkvenc2.c
View File

@@ -697,9 +697,9 @@ static int rkvenc2_is_split_task(struct rkvenc_task *task)
}
if (sli_split_en && slen_fifo_en && sli_flsh_en) {
mpp_dbg_slice("slice output with irq enabled\n");
if (!slc_done_en || slc_done_msk)
mpp_dbg_slice("slice output enabled but irq disabled!\n");
mpp_dbg_slice("task %d slice output enabled but irq disabled!\n",
task->mpp_task.task_id);
return 1;
}
@@ -984,7 +984,8 @@ static void rkvenc2_read_slice_len(struct mpp_dev *mpp, struct rkvenc_task *task
for (i = 0; i < sli_num; i++) {
u32 sli_len = mpp_read_relaxed(mpp, RKVENC2_REG_SLICE_LEN_BASE);
mpp_dbg_slice("wr %d:%d len %d\n", sli_num, i, sli_len);
mpp_dbg_slice("task %d wr len %d %d:%d\n",
task->mpp_task.task_id, sli_len, sli_num, i);
kfifo_in(&task->slice_len, &sli_len, 1);
}
}
@@ -993,7 +994,7 @@ static void rkvenc2_last_slice(struct rkvenc_task *task)
{
u32 sli_len = 0;
mpp_dbg_slice("wr last slice\n");
mpp_dbg_slice("task %d last slice found\n", task->mpp_task.task_id);
kfifo_in(&task->slice_len, &sli_len, 1);
}
@@ -1034,8 +1035,10 @@ static int rkvenc_irq(struct mpp_dev *mpp)
ret = IRQ_WAKE_THREAD;
if (task) {
rkvenc2_read_slice_len(mpp, task);
rkvenc2_last_slice(task);
if (task->task_split) {
rkvenc2_read_slice_len(mpp, task);
rkvenc2_last_slice(task);
}
wake_up(&mpp_task->wait);
}
}
@@ -1541,6 +1544,7 @@ static int rkvenc2_wait_result(struct mpp_session *session,
struct mpp_task *task;
struct mpp_dev *mpp;
u32 slice_len = 0;
u32 task_id;
int ret = 0;
mutex_lock(&session->pending_lock);
@@ -1555,6 +1559,9 @@ static int rkvenc2_wait_result(struct mpp_session *session,
mpp = mpp_get_task_used_device(task, session);
enc_task = to_rkvenc_task(task);
task_id = task->task_id;
req = cmpxchg(&msgs->poll_req, msgs->poll_req, NULL);
if (!enc_task->task_split || enc_task->task_split_done) {
task_done_ret:
@@ -1569,8 +1576,6 @@ task_done_ret:
return ret;
}
req = msgs->poll_req;
/* not slice return just wait all slice length */
if (!req) {
do {
@@ -1578,7 +1583,8 @@ task_done_ret:
kfifo_out(&enc_task->slice_len, &slice_len, 1),
msecs_to_jiffies(RKVENC2_WORK_TIMEOUT_DELAY));
if (ret > 0) {
mpp_dbg_slice("not ret rd len %d\n", slice_len);
mpp_dbg_slice("task %d skip slice len %d\n",
task_id, slice_len);
if (slice_len == 0)
goto task_done_ret;
@@ -1595,7 +1601,8 @@ task_done_ret:
return -EINVAL;
}
mpp_dbg_slice("count %d : %d\n", cfg.count_max, cfg.count_ret);
mpp_dbg_slice("task %d poll irq %d:%d\n", task->task_id,
cfg.count_max, cfg.count_ret);
cfg.count_ret = 0;
/* handle slice mode poll return */
@@ -1603,24 +1610,28 @@ task_done_ret:
kfifo_out(&enc_task->slice_len, &slice_len, 1),
msecs_to_jiffies(RKVENC2_WORK_TIMEOUT_DELAY));
if (ret > 0) {
mpp_dbg_slice("rd len %d\n", slice_len);
if (!slice_len)
enc_task->task_split_done = 1;
mpp_dbg_slice("task %d rd len %d\n", task_id, slice_len);
if (cfg.count_ret < cfg.count_max) {
struct rkvenc_poll_slice_cfg __user *ucfg =
(struct rkvenc_poll_slice_cfg __user *)(req->data);
u32 __user *dst = (u32 __user *)(ucfg + 1);
/* Do NOT return here when put_user error. Just continue */
if (put_user(slice_len, dst + cfg.count_ret))
return -EFAULT;
ret = -EFAULT;
cfg.count_ret++;
if (put_user(cfg.count_ret, &ucfg->count_ret))
return -EFAULT;
ret = -EFAULT;
}
return 0;
if (!slice_len) {
enc_task->task_split_done = 1;
goto task_done_ret;
}
return ret < 0 ? ret : 0;
}
rkvenc2_task_timeout_process(session, task);