I plan to run potentially dangerous applications (browser, etc.) on a separate X-server, but since each server has its own clipboard, it is impossible to copy the link / text from one window to another.
Most articles suggest doing this using scripts using xclip and other similar utilities.
But how to CORRECTLY create a common clipboard so as not to accidentally create a new vulnerability?
NOTE: Updating answer to reflect change of strategy - instead of running Xephyr on a server/container, I am running it on host/main-use environment. The reason for that is that running Xephyr on the server/container is like putting a lock on an inside door instead of the front door - bad actors would go around the inside door and access the clipboard through the X pipe remote socket directly.
Facing the same problem of clipboard vulnerability, I approached it by running a Xephyr on the host (my personal workspace), and forwarding X from the server-on-container(s) to the local Xephyr.
The Xephyr is running on display ':2' while my personal workspace windows and browser are running on the default display ':0'. These two displays do not share a clipboard - each has it's own clipboard. That is the only way to prevent snooping of my personal workspace clipboard on display ':0'. Then I have hot keys set up (e.g., function keys), one to transfer the clipboard content from ':0' to ':2', and another for ':2' to ':0', so allowing complete control.
In shell script the code could look like
xsel --display :0 --clipboard -o | xsel --display :2 --clipboard -i
although I am using javascript as shown at the bottom of this post.
Shell script to start up the X forwarding could look something like this
Xephyr <args> :2
DISPLAY=:2 ssh -X -R 44713:localhost:4713 user@container <<EOF
DISPLAY=:10 PULSE_SERVER=tcp:localhost:44713 openbox --startup firefox
EOF
although I am using a javascript program to do it.
Here is the javascript code for copying between ':0' and ":2" which is mapped to by hot keys. You can see it pops up a transient message box to confirm it worked.
#!/usr/bin/env node
`strict`;
const child_process = require('child_process');
// TODO: write unviewable error messasges to system log
function notifySend(title, msg){
title = title.replace(/"/g, '\\"');
msg = msg.replace(/"/g, '\\"');
//msg = msg.replace(/'/g, '\\'')
try {
child_process.execSync(`notify-send "${title}" "${msg}"`);
} catch(e) {
console.log(`system call "notify-send" failed with ${e.message}`);
}
}
async function clipXfer(fromDispNum,toDispNum){
var clipValue;
let cmd1 = `xsel --display :${fromDispNum} --clipboard --output`;
let cmd2 = `xsel --display :${toDispNum} --clipboard --input`;
try {
clipValue = child_process.execSync(cmd1, { encoding: 'utf-8' });
} catch(e) {
throw Error(`Display ${fromDispNum} clipboard is empty`);
}
await new Promise((resolve, reject)=>{
// eslint-disable-next-line no-unused-vars
var proc = child_process.exec(cmd2, (error, stdout, stderr) => {
//if (stdout) console.log(stdout);
//if (stderr) console.log(stderr);
if (error) {
reject(Error(`${error.message}`));
}
resolve();
});
if (!proc.stdin.write(clipValue))
reject(Error('clipToCont(), pipe write failed'));
proc.stdin.end();
});
}
async function main()
{
let argOff=2;
if (process.argv.length-argOff<2)
throw Error('clip-xfer requires two arguments: fromDispNum, toDispNum');
let fromDispNum = process.argv[argOff];
let toDispNum = process.argv[argOff+1];
argOff+=2;
let f = (outcome)=>{
notifySend("clipXfer", `${fromDispNum} => ${toDispNum} ${outcome}`);
};
await clipXfer(fromDispNum,toDispNum)
.then(f('SUCCESS'))
.catch((e)=>{f(`FAILURE, ${e.message}`); throw e;});
}
//console.log(process.argv);
//console.log(process.env);
main()
.then(()=>{process.exitCode=0;})
.catch((e)=>{
console.log(e);
process.exitCode=1;
});
I use xclipsync. It works in 98%. In 2% I'm doing this:
On dislpay :0 run copyq. On display :1 run parcellite. To sync clipboard i use
copyq add $(parcellite)
or to backward
copyq read | parcellite -c
It works 100%! :-)
