Another kernel RAID5 implementation

By Jonathan Corbet
October 18, 2011

There are many things that the kernel lacks, but RAID implementations is not on that list. Both the MD and DM subsystems currently have full RAID support, while the Btrfs filesystem has lower-level RAID support. RAID5/6 support for Btrfs has been posted a couple of times, but has not yet made it into the mainline. So, one might well be justified in wondering if yet another RAID5 implementation is needed in the kernel.

There will be one if Boaz Harrosh has his way; his RAID5 support patch has been posted to a few filesystem-related kernel development lists. Boaz's patch is aimed at adding RAID5 support to the "objects raid engine" code in the exofs filesystem, which provides a POSIX filesystem on top of object-storage devices. It also implements RAID5 for the pNFS object-storage backend.

According to Boaz, this work constitutes a nice, general-purpose RAID library that could be used in other settings; in particular, he says, Btrfs could make use of it. What would be even nicer, of course, is if some of the existing in-kernel RAID implementations could also move to this library - or if exofs could use one of those implementations. This version of RAID5 support may well be cleaner and more general than the others, but it may well take a stronger argument than that to get a new RAID subsystem merged at this point.

Index entries for this article
Kernel	Block layer/RAID
Kernel	RAID

Another kernel RAID5 implementation

Posted Oct 20, 2011 4:32 UTC (Thu) by dbenamy (guest, #39458) [Link] (7 responses)

There's a typo. It should be "There *are many things that the kernel lacks".

Another kernel RAID5 implementation

Posted Oct 20, 2011 14:03 UTC (Thu) by jake (editor, #205) [Link] (6 responses)

> There's a typo.

Indeed, thanks. In the future, though, please send typos to lwn@lwn.net, rather than post them here, thanks.

jake

Another kernel RAID5 implementation

Posted Oct 20, 2011 16:26 UTC (Thu) by mebrown (subscriber, #7960) [Link] (5 responses)

It seems to be a common case for people to report typos in comments and then for you to ask them to use the email address.

Perhaps you should add a "report typo" button somewhere in the comment posting process to divert these?

Another kernel RAID5 implementation

Posted Oct 20, 2011 17:40 UTC (Thu) by Slumberthud (subscriber, #45657) [Link]

Amen to that.

Another kernel RAID5 implementation

Posted Oct 20, 2011 19:58 UTC (Thu) by cladisch (✭ supporter ✭, #50193) [Link] (3 responses)

Or a notice like "Please do not post typos in the article as comments, send them to lwn@lwn.net instead." right above the comment box.

But there is reason to believe that not everybody would notice it … ;-)

Another kernel RAID5 implementation

Posted Oct 21, 2011 12:19 UTC (Fri) by sorpigal (guest, #36106) [Link] (2 responses)

Unnecessarily chatty web sites have long since trained me to scan right past blocks of text without reading it, especially if I think I know what I'm doing. Buttons, however, draw more attention.

Another kernel RAID5 implementation

Posted Oct 23, 2011 18:46 UTC (Sun) by smoogen (subscriber, #97) [Link]

I like: Reports of typos via comments versus proper form will result in a 2.00 sir-tax [misspelling on purpose.]

Another kernel RAID5 implementation

Posted Oct 25, 2011 15:16 UTC (Tue) by mebrown (subscriber, #7960) [Link]

Ooooh! I know!

<code>
if comment.text.contains("typo") or comment.text.contains("spell"):
typo_notification_widget.add_tag("<blink>")
typo_notification_widget.add_tag("<really_big>")
</code>

Another kernel RAID5 implementation

Posted Oct 20, 2011 22:25 UTC (Thu) by nix (subscriber, #2304) [Link]

It may be more general, but if you want code that's been tested to death, md's raid5 implementation must be that code. So it's probably unlikely that it'll be phased out at any time in the near future, since RAID code is by its very nature rife with complex stuff only executed on rare error paths and thus is a monster to test.