@aha

Yes, I noticed that too.

WBInfo() probably does something like (used_blocks * 100 / total_blocks) using 32-bit math only which overflows on the multiplication when the number of used blocks gets too large.

Great!

In the Screenshot: 13 % used, 16% free?

"usbscsi.device" doesn't support NSCMD_DEVICEQUERY so I need to implement a fallback for this case.

Edit: Turns out that it actually does but for some reason the NSDeviceQueryResult structure needs to be cleared first.

Here's a screenshot of a 500GB NTFS formatted USB harddrive being mounted under VirtualBox using filesysbox.library/ntfs3g-handler:

Thanks for the update - I'm looking forwards to trying this.

Cheers,
Nigel.

Quote:

Turns out that the gcc optimiser didn't like some of my code so I've rewritten it to be less "clever" (also slightly less efficient) and now it works correctly. Since I was sure it wasn't an endian issue I figured it had to be something like this so I spent last evening looking over the code for anything that looked like something that might break with optimisation.

Still some minor things to implement and also I need to test ntfs3g-handler with some device driver that isn't "diskimage.device" which is what I've been using up til now.

Asynchronous doslist operations are now implemented and working. After getting this out of the way I found and fixed a bug in ntfs3g-handler that would cause a crash when formatting the same disk/partition twice in a row.

I also found and fixed a bug in the splay tree implementation so that the block cache is working better now.

Something that still is not working though is the BlockCacheFlush() operation. For some reason after sorting the list of dirty cache nodes the list is empty and I've yet to figure out why. The sorting algorithm used is a merge sort and is exactly the same as I use on AmigaOS 4 and it does not have this bug there.

+1

Thank you!

Thanks for the information on progress!

I implemented the block cache system but it seems to have some bugs so I have disabled it for now as I'm fixing another issue that turned up during testing.

This other issue being that FbxSetupVolume() deadlocks when filesystem is uninhibited after formatting. I'm working on fixing this issue by making the AddDosEntry() and RemDosEntry() calls asynchronous similar to how they already are in the AmigaOS 4 version of filesysbox.

Also I replaced GetSysTime() by some code using ReadEClock() because as GetSysTime() isn't guaranteed to be always increasing it's not really suitable to use for the update timeouts (the AmigaOS 4 version of filesysbox uses GetUpTime() which doesn't exist on AROS BTW).

Thanks for the progress report! I hope you'll find time to port exFAT after the bounty conditions have been satisfied. It would be very useful to read SDXC cards.