[HCD]: status report

Blog post by emitrax on Sat, 2008-08-23 08:51

It's been a bit since my last status update, so I guess it is time for another one.

First of all, I'd like to inform you that I received the first half HCD payment. Since it's a (fantastic) community based effort project, I thought you wanted to know where your donations ended up.

As of commit r27159 you should be able to read data from an UDF partition. The module has not yet been added back to the image, as I'd like to do some more tests, but as far as I can tell, the port of UDF to the new FS API is close to complete, and you can start testing by adding the module to the image and trying using DVD formatted with UDF, or iso image made with mkisofs. Feedbacks are welcome.

As for the other part of my HCD, in case you missed, bonnie++ was added in r26920 and it is available for the braves one, for testing purposes.

In r27052 I also fixed another BFS deadlock that would lock the file system when more then one thread was writing in the same directory. See this for more info.

Ok, going to back to UDF now. ;-)

Comments

Re: [HCD]: status report

Thanks for that update, Salvatore.
Not knowing too much about UDF, I looked at is entry in Wikipedia:
http://en.wikipedia.org/wiki/Universal_Disk_Format

There are supposed to be different revisions of UDF (1.02 ... 2.60) and different "builds" (plain, VAT and spared).

What are your plans for Haikus UDF for your HCD project and what would you consider necessary for the future (R1+)?

Thanks!
Humdinger

--=-=--=-=--=-=--=-=--=-=--=-=--=-=--=-=--=-=--=-=--=--
Deutsche Haiku News @ http://haiku-gazette.blogspot.com

Re: [HCD]: status report

Hi Humdinger,

as far as my HCD goes, I was supposed to port the existing module to the new FS API and that's it, as my mainly target was BFS (I'm more interested in helping making sure BFS is safe for R1). Although I did say in my proposal that I would have looked into write support if I had some time left, but I don't think that's an quick task anymore, so I probably won't do it now.

On the other hand though, I've already printed out UDF spec and I'll probably have a look at it, along with ECMA-167 once I printed them out. Can't tell you more right now. Sorry. :-)

Re: [HCD]: status report

What are your plans for stress testing BFS? I suppose it would be based on chapter 12 of Dominic's book. I hope we can get into coverity soon.
http://www.letterp.com/~dbg/practical-file-system-design.pdf

I liked this quote:
Fortunately in testing BFS our end user tester was amazingly devious and found endless clever ways to trash the file system. Surprisingly, most of the errors discovered were during operations that a seasoned Unix veteran would never imagine doing. For example, once I watched our lead tester start copying a large file hierarchy, begin archiving the hierarchy being created while removing it, and at the same time chopping up the archive file into many small files. This particular tester found myriad combinations of ways to run standard Unix tools, such as cp, mv, tar, and chop, that would not perform any useful work except for finding file system bugs. A good testing group that is clever and able to reliably describe what they did leading up to a crash is a big boon to the verification of a file system. BFS would not be nearly as robust as it is today were it not for this type of testing.