user5265160's questions

I am running Ubuntu 20.04 and am using the Nvidia Riva Automatic Speech Recognition application to detect speech commands. I would like to direct/pipe them into another application to execute a control action. The problem I have is that the ASR application is deployed within a docker container and the control application is not.

Therefore. I would like to know how an application running outside the docker container may interface with an application running inside.

I have considered running everything inside the same container, however this is not practical because our product is mature, has lots of services interacting with each other, and also interfaces to linux kernel drivers.

Thanks in advance for any help.

Ubuntu 20.04.5 LTS

Docker version 24.0.2

I am trying to relocate the /var/lib/docker directory to my USB SSD flash drive to save memory on the internal storage, however, I am having difficulty doing so. I have attempted to modify the docker service to set a different data-root, but when I start the service, I get an error and think this is relates to the fact that the relocate destination is a USB flash drive.

I created a copy of the existing docker files by executing

sudo rsync -avxP /var/lib/docker/ /home/myuser/riva/docker/

where /home/myuser/riva is my mounted USB drive (myuser is my username), and then modified docker.service to set the data root to the new location

ExecStart=/usr/bin/dockerd --data-root /home/myuser/riva/docker -H fd:// --containerd=/run/containerd/containerd.sock

However, docker fails to start, producing the following error

time="2024-03-13T02:21:33.996540964-07:00" level=info msg="Starting up"
time="2024-03-13T02:21:34.000771936-07:00" level=info msg="detected 127.0.0.53 nameserver, assuming systemd-resolved, so using resolv.conf: /run/systemd/resolve/resolv.conf"
time="2024-03-13T02:21:34.100456272-07:00" level=error msg="failed to mount overlay: invalid argument" storage-driver=overlay2
time="2024-03-13T02:21:34.100577936-07:00" level=error msg="[graphdriver] prior storage driver overlay2 failed: driver not supported"
failed to start daemon: error initializing graphdriver: driver not supported

For test purposes, I tried the same procedure to relocate to another folder on internal storage and this worked well, so this proves the issue relates to the SSD Flash drive. When I execute ‘fdisk -l‘, the result reports ntfs, as expected.

/dev/sda1: LABEL="RIVA" UUID="18089E33089E103C" TYPE="ntfs" PARTLABEL="Basic data partition" PARTUUID="5fa27c32-117b-4223-9a76-6297b0fdc625"

Can someone help?

Thanks

I have symbols, defined in my makefile to represent the build version and date, however when I access them in my C++ program, the values are incorrect due to an offset being applied. My linker command is as follows

g++   -Xlinker --defsym -Xlinker __BUILD_DATE=$(date +'%Y%m%d') -Xlinker --defsym -Xlinker __BUILD_NUMBER=10000 -o MyProg main.o

and the symbols are defined, as shown below

readelf -s MyProg | grep BUILD
    58: 0000000000002710     0 NOTYPE  GLOBAL DEFAULT  ABS __BUILD_NUMBER
    65: 0000000001348c02     0 NOTYPE  GLOBAL DEFAULT  ABS __BUILD_DATE

However, when I run the folllowing code

#include <cstdio>

extern char   __BUILD_DATE;
extern char   __BUILD_NUMBER;

int main()
{
    printf("Build date  : %p\n", (void *)&__BUILD_DATE);
    printf("Build number: %p\n", (void *)&__BUILD_NUMBER);
    return 0;
}

The output is

Build date  : 0x55555689cc02
Build number: 0x555555556710

An offset of 0x555555554000 is being applied. I am running on ubuntu 18.4. The program runs fine on ubuntu 16.4. One of the replies in https://stackoverflow.com/questions/8368423/why-does-the-linker-modify-a-defsym-absolute-address states

The behavior of --defsym changed between gcc 5.4.0 (using Ubuntu 16.04.4) and 7.3.0 (Ubuntu 18.04). At 5.40, --defsym created a symbol representing an absolute, non-relocatable address. At 7.3.0, readelf -s shows the symbol as "ABS", but the symbol is in fact relocated when the program executes.

As a workaround I can add a known defsym value to determine the offest and use this to adjust the other symbols, however it would be good to know the proper way to de-reference symbols in such instances?

I'm trying to build a debian package for an ARM64 platform on my AMD64 Ubuntu OS. I execute the command

debuild -us -uc --host-arch arm64

but get the warning

dpkg-architecture: warning: specified GNU system type aarch64-linux-gnu does not match CC system type x86_64-linux-gnu, try setting a correct CC environment variable

and so would like to know how to set the CC environment variable, as the warning advises. For my target OS I would typically use

/opt/l4t-gcc-toolchain-64-bit-32.1/install/bin/aarch64-linux-gnu-gcc

Thanks.

I am creating a Debian package for a software application which gets executed from a service script and could like to know the best way to do this.

My current thoughts are for the package to copy the executable to /usr/bin, and the service to /etc/systemd/system/, and have a postinst script to execute

systemctl enable myapp.service

Is this the correct approach or is there a more 'standard' procedure for handling services within Debian package management?

I am working with an embedded linux device, which we typically connect to by static IP address during development. However, on-site we have a requirement to connect using DHCP and so, I would like to setup a primary DHCP connection, with a fallback static IP address on a different network. Is this possible?

I can use nmcli to configure a DHCP connection (IPV4.method=manual) and can successfully add an additional static IP address. Both will exist when simultaneously when the DHCP server is present, however when there is no DHCP server, the connection is disabled including the static IP address.

I have also tried setting up multiple connections, assigned to eth0, i.e. one static and one DHCP, and can manually enable them using

nmcli con up ConnectionName

but this does not meet the requirement because of the need to manually enable them in the event of the other failing. I can, of course run a script to check the connection status and enable the other if required but thought this would be a realistic expectation of a network manager.

Is there a way to tell network manager to attempt one connection and if this fails try another?

I am using Ubuntu 18.04.

Thanks.