idevice commands fail on Linux when /var/lib/lockdown pre-exists about libimobiledevice

[nikias]

The thing is that you need latest usbmuxd that will handle the pair records in /var/lib/lockdown

[greg-dennis]

I do have the latest usbmuxd and it does properly handle the pair records in /var/lib/lockdown. The issue is simply that it was unable to create the /var/lib/lockdown directory to begin with. It attempted to do so here:
https://github.com/libimobiledevice/usbmuxd/blob/master/src/main.c#L542

/var/lib isn't generally world writable, or it at least isn't on a stock Ubuntu install, and so that directory cannot be created at that point.

My suggestion is for sudo make install of either libimobiledevice or usbmuxd to create it and either make it writable by anyone or owned by the usbmux user.

Also, when usbmuxd fails to create the directory, it fails silently, and the idevice commands silently echo no output to the user.

[nikias]

Hm well actually usbmuxd creates the directory before dropping the privileges:

http://cgit.sukimashita.com/usbmuxd.git/tree/src/main.c#n537

And before finally dropping the privileges it changes ownership and permissions:

http://cgit.sukimashita.com/usbmuxd.git/tree/src/main.c#n565

In a normal setup usbmuxd will be executed by udev as root so this should work...

[greg-dennis]

You are right. My bad. Seems I had an error in my build that was causing HAVE_LIBIMOBILEDEVICE to not be set, causing the directory to never be made. Sorry to bother!

[dborca]

I had the same problem as @progressnerd.
It was very frustrating to see everything failing silently.
And now usbmuxd has a semantic dependency to libimobiledevice?

[greg-dennis]

I think there should be a more explicit error message to the user when attempting to use the latest libimobiledevice without updating usbmuxd, or if usbmuxd is built before libimobiledevice is and HAVE_LIBIMOBILEDEVICE is false. I don't know why HAVE_LIBIMOBILEDEVICE is even an compilation option.

[dborca]

"I don't know why HAVE_LIBIMOBILEDEVICE is even an compilation option."

Exactly.

Besides, I find /var/lib a rather curious location for userpref. It holds state information pertaining to an application or the system, not an user. http://www.pathname.com/fhs/pub/fhs-2.3.html#VARLIBVARIABLESTATEINFORMATION

[FunkyM]

Thanks for you input, but you guys are debating about a feature that is WIP and needs heavy refactoring of all related components. The directory and permissions issue will be resolved and we are very aware of it, too.
The next stable release will work properly and have the needed refactoring in place.

[greg-dennis]

Upon further investigation and reports from other users, I've concluded this should be considered a genuine bug. If the userpref directory already exists, usbmuxd should either explode with an intelligble error or chmod and chown it to the permissions needed by libimobiledevice. Many Linux distros are coming with usbmuxd installed by default, and these default installs are creating the lockdown directory with permissions that make usbmuxd and limd fail. Currently, usbmuxd only calls chmod and chown when it creates the directory itself.

[FunkyM]

You must be aware of the fact that limd will no longer manage pair records nor access the pair record directory, only usbmuxd will. Thus this problem will go away anyways.

[greg-dennis]

Will it necessarily? This directory sometimes pre-exists with root permissions, and since usbmuxd won't change the permissions in that case, I think usbmuxd will be unable to access it after it drops privileges.

[nikias]

@progressnerd I changed usbmuxd a while back to create the directory and set proper permissions. Nevertheless @FunkyM is right and the pair record management will move into usbmuxd. I already added the necessary functionality to usbmuxd and adapted the libusbmuxd interface (usbmuxd_read_buid, usbmuxd_read_pair_record, usbmuxd_save_pair_record, and usbmuxd_delete_pair_record) but libimobiledevice integration is still missing.
Closing.

[greg-dennis]

Thank you! I see it was fixed on 12/10/13 by commit libimobiledevice/usbmuxd@ed73edd, only 6 days after I reopened this bug. Now that's service! :)