Tested with your patch. It works here.

It still has a bug, as you know CD-ROM drives must be one after the other, right now you can't mount drives that break the order, however right now you can still unmount drives in the middle; in other words, you can always unmount images.
Other bug I missed in the first patch is, when you unmount cdrom drives that were mounted from a directory (not an iso file) the drive letter is not freed. I already fixed that, I'll attach a patch.

* Uploaded a small fix in this patch

Okay, this is an step closer to a full unmount.
I've added a couple things to fix the bug that let unmounting cdrom drives between two other cdroms (breaking the chain of cd drives) this should no longer be a problem.

CD Images Unmount has become CD Unmount (full support) and it's quickly morphing to * Unmount 😀

Here's the fixed patch.

* Sorry the patch had a bug it and was corrupt, I've fixed it now (I think)
** I have added a PhysFS friendly version of the patch 😀
For this friendly version, apply the PhysFS first (you can still use this same patch without the PhysFS installed (if you don't want to use PhysFS, just skip the 'drive_physfs.cpp' file patch, ignoring 3 of 3 hunks and also ignore the 2 of 8 hunks FAILED message on file 'drives.h'.)

I cannot mount cd-rom drive with limited support.

does that make sense to mount a directory with -ioctl?

Ah, that's my mistake. 😀
attached a different screenshot...

Hmm.. true, and this is really weird as I don't recall touching anything that might have broken that... I'll have to check 😜

I now know where the problem is... I suppose it's a small design fault in MSCDEX_AddDrive() it returns an errorcode, the errorcode it's 0 if no error was found, and another number if something went wrong... or does it? well, almost.
0 = MSCDEX_SUCCESS
1 = MSCDEX_ERROR_MULTIPLE_CDROMS
2 = MSCDEX_ERROR_NOT_SUPPORTED
3 = MSCDEX_ERROR_PATH
4 = MSCDEX_TOO_MANY_DRIVES
5 = MSCDEX_LIMITED_SUPPORT
6 = MSCDEX_UNKNOWN_ERROR
Every single one of those non-zero codes is a "fatal error" making the mount to fail (and every one of them should destroy the half-created cdrom object) however, MSCDEX_LIMITED_SUPPORT isn't a fatal error, actually it's a success code (even when the "cdrom" has limited access, it succeeded to mount it.) This breaks the "0 means succeded" supposition I had. I think those error codes need to have a close look and better more consistent values; perhaps fatal errors could be negative integers, while success errors could be positive numbers and the 0... perhaps implementing a global common errorcode scheme with macros as SUCCEEDED() and FAILED() would be a good thing.

* Okay, I decided to post a temporary patch, this would make it easier for you 😀

Thanks, krounz. It's now possible to mount the cdrom drive.