Hi All.

I am running a script that copies files to a USB disk.

What I would like to do is run another script if the USB disk get disconnected before the copying is done.

TIA.

Ant.

And what have you done in an attempt to solve this problem?

Dear Grail.

Thanks for your helpful reply. I did not realise that posting what I had already tried was essential.

Anyway, what I do now is at the end of the script I check to see if the USB device is still present and if not, run the error script.

The problem with this is that the device may still be present, but there was a write error or some other issue with the cp command.

What I was hoping for was a more sophisticated solution so that when/if the cp command fails, it jumps straight to the error.

Unless you can lead me to a better process, I guess I will stick with this.

Ant.

It would help if you actually post the script. Use code tags (Click 'Post Reply' then click on '#') for that.

One problem you have is that if the device gets unplugged, you may have corrupted the filesystem on it...

Now if your trap handler does nothing but request the user to put it back in... things MIGHT be able to continue. Normally, this is not handled by the shell though - it is handled by the USB controller so that any copy to/from isn't aborted, and the system buffers remain consistent (and the storage device doesn't get corrupted). If, howerever, the USB actually resets, I think things can get nasty - the device is now corrupted, and might not be mountable.

I believe the usual way to trap shell errors within the shell (and not abort the script) is by using an "eval" on a function. If the function aborts, the eval will catch it and return the error status.

Here is the script.
What I would like to do is somehow trap is the cp command did not complete, then run the error.sh script.
Thanks for looking at it.

Umm... you can't check for a file on a removed device without having yet another failure...

All you can do is evaluate the exit status of the cp.

If the cp command fails for some reason, it will exit with a non-zero exit code. The actual exit code will be stored in the $? variable which you can then test for a non-zero value:Or you can simply use the cp command as a condition in the if statement:Also, as jpollard pointed out, checking whether a device is mounted or not by looking for a specific file, is perhaps not the best idea: If the device is not mounted, you'll generate another error. The method seems to have the advantage of being able to identify a specific device as being mounted (if the file is there), but that assumes that the file hasn't been deleted, and that a file by the same name was not created on a different device.

If you want to see whether a device is mounted at a specific mount point, I'd recommend parsing the output from the mount command. If the path to the mount point is listed, a device of some kind must obviously be mounted there.

It's more difficult to verify conclusively that a specific device is mounted at that mount point. You could check for a filesystem UUID, but I believe that information can only be read with a filesystem-specific utility. For instance, tune2fs is able to extract the UUID from an ext2/3/4 filesystem.

Checking for a particular partition table GUID could be an alternative, but AFAIK those are only found on GPT-partitioned disks.

1 members found this post helpful.

Thanks Ser Olmy for your detailed reply.
This solution
works perfectly and has really helped me out.

Ant.