I'm using Ubuntu 11.10 on amd64, with ntfs-3g 2011.4.12AR.4.

At first I thought there was a problem with TrueCrypt as the system froze when I tried to create a filesystem inside a container. It happened when filling the device with zeros, at about 1% or 2% (the file had 50GB in this case, but didn't seem relevant).


Then I tried with dm-crypt/LUKS, and after setting up a file-backed disk, and the same occurred with mkntfs.


When using mke2fs it worked and mounted fine in both previous cases. I believe the problem is related to /dev/mapper/* devices as it worked fine with /dev/loopX only. And I think it has to do with mkntfs because "dd if=/dev/zero of=/dev/mapper/truecrypt1" works as well. Finally, mkntfs succeeds if using a "quick format" (-f). To summarize:


After the last command, the filesystem can be mounted and used normally. So it isn't so serious, but I wanted to report it as it crashes the system with default settings.

BTW is there a reason to keep zeroing on by default (AFAIK modern disks manage bad sectors internally and it is useless when using a file instead of a device)?

Hi,


Strange. mkntfs is a user-level application, it should not be able to cause the system to freeze, the more as it is just writing a zeroed buffer (and apparently the buffer is correctly allocated).
Did you use some unusual option ?

I think this was done to get the list of bad clusters and mark them as unallocatable... but with current disk sizes, you have to be patient.

Regards

Jean-Pierre

Thank you for your reply.



Right, the problem must be in the kernel (or module) code. But what surprises me is that it's triggered only by mkntfs and not dd.

I forgot to mention that I ran all commands as root. There were no additional command arguments other than shown.

As the system freezes, I'm not sure how this could be investigated further.



What I was trying to say is, if modern disks detect and correct bad sectors (which in turn could be assigned to a filesystem cluster) internally, there is no need to check them in software. For instance, mke2fs and others don't do it (and they run very fast by default, because of that).