asmorkalov
commented
1 year ago Please attach reproducer code with images and full log with error.
Open JackIRose opened 1 year ago
asmorkalov
commented
1 year ago Please attach reproducer code with images and full log with error.
Kumataro
commented
1 year ago Sorry this is only advice to write bug issue, and it is not to fix problem directly.
I updated to the latest OpenCV version and the issue is still there, And maybe I guess this is not OpenCV bug. If OpenCV detects any trouble/wrong conditio, it calls cv::Error() to tell for you. However this log has no information about it. So it is hard to comment/help .
( e.g. I suspect that the memory allocated to the GPU is being exhausted. )
JackIRose
commented
1 year ago
.
The txt file below is the complete log report. 1026_gdb_bt_full.txt
Kumataro
commented
1 year ago I'm sorry but It seems like you want to say that copyTo() is the cause, but there isn't enough evidence to confirm that.
There are no information about it.
#8 0x0000007f9da691fc in cv::error(int, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, char const*, char const*, int) ()
from /mnt/system/mvsystem/lib/libKeylabAVM.so
No symbol table info available.
#9 0x0000007f9d9fbadc in cv::ocl::OpenCLAllocator::upload(cv::UMatData*, void const*, int, unsigned long const*, unsigned long const*, unsigned long const*, unsigned long const*) const ()
from /mnt/system/mvsystem/lib/libKeylabAVM.so
No symbol table info available.
#10 0x0000007f9d912fe8 in cv::Mat::copyTo(cv::_OutputArray const&) const () from /mnt/system/mvsystem/lib/libKeylabAVM.so
No symbol table info available.So We don't/cannot know where exception is happened at cv::ocl::OpenCLAllocator: :upload(). Only you can identify at which line the Exception occurs. Please could you detemine it ? Perhaps there is a reason.
Code: https://github.com/opencv/opencv/blob/4.7.0/modules/core/src/ocl.cpp#L6228-L6230
Method:
I think it will be difficult to help you further unless new information is shared to us.
zirid
commented
12 months ago A quick look at the code, it seems like nothing ensures thatImageSrcRoi was successfully created. check if that image was created before copying. I would also check that ROI rect at the 3rd line.
JackIRose
commented
12 months ago I'm sorry but It seems like you want to say that
copyTo()is the cause, but there isn't enough evidence to confirm that.There are no information about it.
#8 0x0000007f9da691fc in cv::error(int, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, char const*, char const*, int) () from /mnt/system/mvsystem/lib/libKeylabAVM.so No symbol table info available. #9 0x0000007f9d9fbadc in cv::ocl::OpenCLAllocator::upload(cv::UMatData*, void const*, int, unsigned long const*, unsigned long const*, unsigned long const*, unsigned long const*) const () from /mnt/system/mvsystem/lib/libKeylabAVM.so No symbol table info available. #10 0x0000007f9d912fe8 in cv::Mat::copyTo(cv::_OutputArray const&) const () from /mnt/system/mvsystem/lib/libKeylabAVM.so No symbol table info available.So We don't/cannot know where exception is happened at
cv::ocl::OpenCLAllocator: :upload(). Only you can identify at which line the Exception occurs. Please could you detemine it ? Perhaps there is a reason.Code: https://github.com/opencv/opencv/blob/4.7.0/modules/core/src/ocl.cpp#L6228-L6230
Method:
- building opencv with debug option
- adding verbose log
I think it will be difficult to help you further unless new information is shared to us.
After review, we located the specific openCV error location, which is in the CV::ocl::OpenCLALLocator::upload line, but we still cannot decipher what problem caused the crash and the role of this line of function. The specific details log and pictures are as follows.
Log:
what(): OpenCV(4.7.0) /home/hisense/桌面/SoftTokit/opencv-4.7.03/modules/core/src/ocl.cpp:6330: error: (-220:Unknown error code -220) OpenCL error CL_INVALID_COMMAND_QUEUE (-36) during call: clEnqueueWriteBuffer(q, handle=0x7f3c01e010, CL_TRUE, offset=0, sz=2188800, data=0x7f3c9386c0, 0, 0, 0) in function 'upload'
![Uploading Screenshot from 2023-11-08 09-25-21.png…]()
Kumataro
commented
12 months ago Thank you for your investigation! It's quite a deep topic, so my help may not be enough. sorry.
Umm... (if my recognize is wrong, sorry) it seems that CL_INVALID_COMMAND_QUEUE means q is invalid.
https://man.opencl.org/clEnqueueWriteBuffer.html
CL_INVALID_COMMAND_QUEUE if command_queue is not a valid command-queue.
q comes from here.
I believe that this source code shows usually returns valid same queue.
queue_ is initilized once and it is reused many times.
(Sorry, it's a low possibility)if possible, please could you check whether there is a return with L3417 ? A case where the queue becomes invalid should not be expected here...
JackIRose
commented
12 months ago (Sorry, it's a low possibility)if possible, please could you check whether there is a return with L3417 ? A case where the queue becomes invalid should not be expected here...
Thank very much for your answer. After the program crashes, the system also triggers an error log, prompting page allocation failure. Will this affect the validity of q? The log is following:
25186.354250] avm360Main: page allocation failure: order:7, mode:0x40dc0(GFP_KERNEL|__GFP_COMP|__GFP_ZERO), nodemask=(null),cpuset=/,mems_allowed=0 [25186.354291] CPU: 1 PID: 1775 Comm: avm360Main Not tainted 5.10.110 #83 [25186.354295] Hardware name: Rockchip RK3588 EVB7 LP4 V10 Board (DT) [25186.354300] Call tr2023-11-06 12:36:48 ##G3_DEBUG[g3module.c,147]Send AT: AT+CSQ
If you encounter such a problem, do you have any good solutions? This has really bothered us for a long time. We would be grateful if you could give us some valuable advice.
System Information
OpenCV version: 4.7.0 Operating System / Platform: Ubuntu 20.04 Compiler & compiler version: GCC 11.3
Detailed description
We are working on a 360 panoramic splicing project. In the code, we use the CV::Mat::Copyto function under multi-threading to copy the pictures obtained by the camera from the CPU to the GPU, and convert them from mat to Umat format. Then it runs normally on the bus, but the program crashes after about a week. This has happened three times. Through the logs, it is found that it is a problem with the copyto::upload function code in Opencv. We cannot reproduce this problem in the laboratory environment because it can always run normally in the laboratory. Therefore, we suspect that it is a problem with the opencv::copyto function.
The detailed log is as follows
0x00000071b287968 in cv::error (cv: Exception const&) () from /mnt/system/mvsystem/lib/libKeylabAVM.so 0x0000007fb28b8b9c in cv::error (int, Cxx11: :basic string<char, std:: char traits, std: :allocator > const&, char const, char const, int) () from /mnt/system/mvsystem/lib/libKeylabAVM.so 0x0000007fb284b47c in cv::oc1::OpenCLAllocator: :upload (cv::UMatData, void const, int, unsigned long unsigned long const, unsigned long const, () from /mnt/system/mvsystem/lib/libKeylabAVM.so 579 580 581 282 583 584 585 ：86 587 0x0000007£b2762988 in cv: Mat: :copyTo (cv: :_OutputArray consts) from /mnt/system/mvsystem/lib/libKeylabAVM.so 0x00000071b26791a8 in nameKeylabAVM: KeylabAVMOnline:: Inference RectifyBird
Steps to reproduce
Issue submission checklist