8

I have to explicitly test the case what my application does if it tries to access some corrupted part of an NTFS file system. I looked around for similar questions but those were about corrupting specific files (i.e. destroy the syntax of an xml file, etc).

I need to create a scenario where chkdsk would report errors in the file system in a specific location.

Does anybody have any idea how I could provoke such a situation to test my application.

Just in case you worry: I wouldn't do that on my production system, I have a VM set up for things like that.

klaustopher
  • 6,702
  • 1
  • 22
  • 25
  • 1
    See [this answer](http://serverfault.com/a/498931/1774) to a similar question at Server Fault. The Q&A relate to linux, but it seems like there should be (or ought to be) a similar capability available to Windows via a device driver laying over the physical disk's driver. – RBerteig Apr 13 '13 at 01:06

3 Answers3

2

If the concern is about what your app would do when encountering corrupt files, then why bother corrupting the filesystem that contains the files? Why not simply corrupt the files themselves?

The operating system does a lot to ensure that filesystems are robust, I'm not sure that intentionally scribbling garbage on the file system will do anything other than fuzz test the filesystem implementation.

Larry Osterman
  • 16,086
  • 32
  • 60
  • 1
    My tool helps migrating files from one OS installation to another, and there is absolutely no way to implement checks for the integrity of every file type. I am pretty sure that my checks for I/O errors are correct and that the program exits with a certain error code. But it is required that I explicitly test that case. – klaustopher Nov 29 '11 at 08:29
2

Since there is no better solution here, I did it the hard way: I used Sysinternals NTFSInfo to find the location of the Master File Table, started a Linux Live CD and edited some bits in both locations of the Master File Table (MFT). Took some time to figure out the right places, but I can produce a corrupted file system no to test the error routines in my application.

Thanks to those contributing ideas, but I will award the answer to myself. If anybody has the same problem in the future, feel free to contact me.

klaustopher
  • 6,702
  • 1
  • 22
  • 25
  • 2
    Your question is excellent, but this answer is not. For someone (like myself!) who is trying to artificially corrupt a file entry in the MFT this answer has little detail to go on. Could you flesh out some details beyond `Took some time to figure out the right places`? – AlainD Oct 05 '17 at 17:59
  • @AlainD Sorry, all of this was 7 years ago. I basically tried poking in different places and setting a bunch of bytes to zeros until chkdsk broke. – klaustopher Mar 28 '18 at 12:14
0

I would try to write some random bytes at some random locations of the disk (or disk partition).

Basile Starynkevitch
  • 223,805
  • 18
  • 296
  • 547
  • That would corrupt the disk somewhere ... I need to specifically have errors in a given folder like C:\MyApp\Data – klaustopher Nov 28 '11 at 14:03
  • 1
    Perhaps you could use Linux NTFS driver, patch it to write garbage when accessing the specific `MyApp/Data`, and use it from Linux? (I agree it take more than a day of work) – Basile Starynkevitch Nov 28 '11 at 14:18
  • Yeah, that was my first intention, but I hope there is an easier way. Cause I think that the chance of destroying the whole system and being unable to boot after all is pretty high... So if there is a better solution, I am thankful for any advice – klaustopher Nov 28 '11 at 14:24