Command:
gst-launch-1.0 playbin uri=rtsp://192.168.1.88:556/vga
The memory of gst-launch-1.0 will increase more than 30M every hour, then the video stop, but audio soud ok。
Delete the libgstvpu.so , use the same commad. memory will not increase, but play the high resolution (1920x1080) not smooth.
Anybody has faced the same problem and resolved? Please help
I do more test.
Compare playing local video file and RTSP:
1) play 1920x1080 ((MPEG4\AVC_H_L3)local files, that seems not has memory leak. (log file : gplay-test-AVC_H_L3_Schindlers.log、gplay-test-MPEG4_Schindlers.log)
2) play 1920x1080 (AVC_H) RTSP, memory increase about 16M. (log file: gplay-test-AVC_H_RTSP.log)
Compare playing RTSP(960x540 AVC_H) with vpudec and avdec_h264 lib:
a) use vpudec , memory increase about 10M in 20 mins.( log file: gplay-test-AVC_H_RTSP_with_VPU.log)
b) use avdec_h264(delete /usr/lib/gstreamer-1.0/libgstvpu.so) , memory not increase. (log file: gplay-test-AVC_H_RTSP_with_avdec_h264.log)
CPU: IMX6DL
Kernel: 4.1.15
BSP: 2.0.1 GA
vpudec: imx-gst1.0-plugin-4.1.4
avdec_h264:gst-libav-1.8.1
b36401, do you have an update about this case?
Karina,
I escalated it to Experts Team.
Then internal ticket was created last monday.
Let's wait for the answer.
HI, Victor,
Did the Experts Team repeat this problem? Is there any good news?
R&D Team will provide an update next week due an urgent task for this week.
Sorry for the delay.
Thanks, waitting for your good news.
They did not reproduce the issue by VLC as the RTSP server.
Please provide us more details.
If the issue occurs only with your special RTSP sender, it is hard to identy the root cause.
Hi, Victor.
I found use this commad for VLC server will cause this issue.
vlc -vvv \\192.168.1.253\work\mynfs\AVC_H_L3_Schindlers.list.1993.HDTV.mkv --loop :sout=#transcode{vcodec=h264,vb=1024,acodec=mpga,ab=128,channels=2,samplerate=44100}:rtp{mux=ts,sdp=rtsp://:8554/vga} :sout-keep
I do not know the "mux=ts" meaning, but your team can try it.
Our RTSP sender is https://detail.tmall.com/item.htm?id=521274714573
When the RTSP_Format choose TS, gstreamer1.0 will cause memory leak, but ES will not.
But using the TS format, the memory will not increase when playing by gstream0.10(kernel is use the same one).
Also playing good with gstreamer1.0 by using libav_h264_dec.
I found mfw_gst_avc_fix_nalheader function /* size of SPS&PPS in NALs */in src/video/vpu_dec.full/src/mfw_gst_vpu_decoder.c(gst-fsl-plugins-4.0.3 in gstreamer0.10), but gstreamer1.0's vpudec didn't
Hi Liang JiaXing ,
Please try with attached libgstvpu.so, there is a known issue on memory alignment.
Can you replace the original vpu library on board with command "cp libgstvpu.so /usr/lib/gstreamer-1.0/ " to see if can still reproduce memory leak issue?
The HDMI encoder had been received, but the test condition is not ready yet.
Hi,Li,
thanks for replay. but this lib is not work for our board.
root@imx6dlsabresd:/home# gst-inspect-1.0 vpudec
(gst-plugin-scanner:14042): GStreamer-WARNING **: Failed to load plugin '/usr/lib/gstreamer-1.0/libgstvpu.so': libunwind.so.8: cannot open shared object file: No such file or directory
No such element or plugin 'vpudec'
my BSP is : BSP: 2.0.1 GA
bitbake config is :
===================================
MACHINE ??= 'imx6dlsabresd'
DISTRO ?= 'upt-mx6dl-eglfs'
PACKAGE_CLASSES ?= "package_rpm"
EXTRA_IMAGE_FEATURES ?= "debug-tweaks"
USER_CLASSES ?= "buildstats image-mklibs"
PATCHRESOLVE = "noop"
BB_DISKMON_DIRS = "\
STOPTASKS,${TMPDIR},1G,100K \
STOPTASKS,${DL_DIR},1G,100K \
STOPTASKS,${SSTATE_DIR},1G,100K \
STOPTASKS,/tmp,100M,100K \
ABORT,${TMPDIR},100M,1K \
ABORT,${DL_DIR},100M,1K \
ABORT,${SSTATE_DIR},100M,1K \
ABORT,/tmp,10M,1K"
PACKAGECONFIG_append_pn-qemu-native = " sdl"
PACKAGECONFIG_append_pn-nativesdk-qemu = " sdl"
CONF_VERSION = "1"
PACKAGECONFIG_append_pn-qtbase = " sql-sqlite"
IMAGE_INSTALL_append += " \
gstreamer1.0 \
gstreamer1.0-plugins-base \
gstreamer1.0-plugins-bad \
qtmultimedia \
qtmultimedia-plugins \
qtmultimedia-qmlplugins \
qtbase-plugins \
tslib \
tslib-calibrate \
tslib-conf \
tslib-tests \
"
DL_DIR ?= "${BSPDIR}/downloads/"
ACCEPT_FSL_EULA = "8"
HI,Li:
Has the testing environment with HDMI encoder been setup? this problem can repeat now?
When pc, imx6 board, hdmi encoder are connected to the same router, neither http://192.168.1.98/ nor the default IP address http://192.168.1.168/ can be reached to set the software parameters.
So, it seems we can't reproduce the issue on our side.
How about your network connection?
Some debug suggestion.
Hi,Qiang_FSL:
可以ping通192.168.1.98吗?是不是你们的路由器不在同一网段呢?
如果ping不通,可以复位背后的reset键,恢复出厂设置后默认ip是192.168.1.168.
寄过来的HDMI编码器是静态IP地址吗?固定192.168.1.98?
对的,固定192.168.1.98。
This week's update:
With the received HDMI encoder, there will be a 9 bytes package for each 30 frames, this is not a video frame, so it will be held in VPU.
Has this issue fixed?
Hi,Qiang:
I test this patch , it still has memory leak.,from 373m to 423m.
and gplay-1.0 exit.
Please check the log in
链接:https://pan.baidu.com/s/1KUgtP7Atw6UPMcEbjcMWMQ
提取码:9j1g
Can you zip the log file and attach here? Thanks!