Hi, Shane
I’ve been slowly reconcile our driver with the existing driver. Basically implementing all the needed s_ctrl methods. I also experimented with different pix_clk_hz. Right now when I run my pipeline with nvcamerasrc I got very similar out put with existing driver at the beginning, with extensive logging of tegra vi, csi and channe and nvcamera-daemon. However my pipeline always times out after showing a few LSC:patch height less than 10, this is shown before capture thread step 5 and the capturing timeout after step 6. There also is an Error: waitCsiFrameStart timeout guid 0 later. Attached is my nvcamperasrc verbose logging as well as dmesg logs. Do you know how can I investigate further the root cause of this failure? When I use v4l2src pipeline I can see there are csi requests for capture each frame and finish capture but in nvcamerasrc pipeline there is no csi requests at all so I’m stuck.
thanks a lot
The device has its own clk and it is the standard 24mhz. I’m not very sure about pix_clk_hz and mcl_multiplier, are they refer to MIPI’s pixel_clk and the multiplier wrt to mclk_khz?
Specifies the sensor pixel clock for calculating the exposure, frame rate, and so forth.
how is this value used to calculate the exposure and frame rate?
thanks
ShaneCCC:
@xiaoyongtijee
mode0 { // IMX219_MODE_3280X2464
mclk_khz = "12000";
num_lanes = "2";
tegra_sinterface = "serial_a";
discontinuous_clk = "yes";
cil_settletime = "0";
active_w = "3280";
active_h = "2464";
pixel_t = "bayer_rggb";
readout_orientation = "90";
line_length = "3448";
inherent_gain = "1";
mclk_multiplier = "25";
pix_clk_hz = "170000000";
min_gain_val = "1.0";
max_gain_val = "16.0";
min_hdr_ratio = "1";
max_hdr_ratio = "64";
min_framerate = "1.462526";
max_framerate = "21";
min_exp_time = "13";
max_exp_time = "683709";
}; };
nvcamera-daemon-new.log (78.1 KB)
Please use the v4l2-ctl to capture the raw data and verify it is validate first. And did you try the discontinuous_clk for yes and no.
Hi, Shane
Really appreciate for the quick response. the v4l2-ctl capture works fine now. it may have error for the first frame or two and sometime has a timeout after it finishes but in between is ok. I’ve tried both discontinous_clk yes and no and it does not make a difference for the nvcamerasrc case
thanks
If the v4l2-ctl still have problem you can’t say the v4l2-ctl capture works fine.
v4l2-ctl -d /dev/video0 --set-ctrl bypass_mode=0 --stream-mmap --stream-count=500
This command runs mostly ok, the outputs are lines of
dmesg log is about 500 of below (with my verbose output).
[ +0.000042] video4linux video0: tegra_channel_error_status : error status
last frame has below error though
vi vi: nvhost_syncpt_wait_timeout_ext: thresh 531, timeout 20
overall stat after it all finishes is
[ +0.000050] vi2_channel_read
Thanks a lot
ShaneCCC:
If the v4l2-ctl still have problem you can’t say the v4l2-ctl capture works fine.
v4l2-ctl -d /dev/video0 --set-ctrl bypass_mode=0 --stream-mmap --stream-count=500
one additional thing I noticed for the nvcamera-daemon verbose output is there is a
Are there any additional kernal logging that I can enable in /sys/kernel/debug to understand this issue more?
thanks
This command runs mostly ok, the outputs are lines of
dmesg log is about 500 of below (with my verbose output).
[ +0.000042] video4linux video0: tegra_channel_error_status : error status
last frame has below error though
vi vi: nvhost_syncpt_wait_timeout_ext: thresh 531, timeout 20
overall stat after it all finishes is
[ +0.000050] vi2_channel_read
Thanks a lot
ShaneCCC:
If the v4l2-ctl still have problem you can’t say the v4l2-ctl capture works fine.
v4l2-ctl -d /dev/video0 --set-ctrl bypass_mode=0 --stream-mmap --stream-count=500
@xiaoyongtijee
Due to the v4l2-ctl and nvcamerasrc use different VI/CSI driver may have different behavior for the error handling.
@ShaneCCC
I’ll give that a try. Does nvcamera-daemon has some setting to dump reading of registers as in tegra v4l2 vi/csi drivers?
thanks a lot
@xiaoyongtijee
@ShaneCCC
It finally works now!!! Really appreciate all your supports!
Fixed CRC as well as word count in mipi, it turns out our mipi rx misses two words.
I accidentally found we have embedded_metadata so modified the DT accordingly.
Now I’ll try to port it to 28.1
thanks again!