Skip to contentSkip to conversation

AoE inclusion into 2.4.x

DISCUSSION Sam Hopkins 2 participants 3 messages Apr 21, 2004

lore
Share & tools
lore
#0 lore raw

Greetings,

I would like to offer our driver for inclusion into the 2.4 kernel. Marcelo suggested (strongly) that I make this request officially to the list.

The driver code is available from:

	http://www.coraid.com/support/aoe-1.6.tar

I'm now going to cheat and copy our prior correspondence. Please CC me on any comments / questions as I'm not a lkml subscriber.

Cheers,

Sam Hopkins
[email protected]

-----

Quoted 1 line

From cyclades.com!marcelo.tosatti Wed Apr 21 12:18:18 EDT 2004

Date: Wed, 21 Apr 2004 13:16:35 -0300
From: Marcelo Tosatti <[email protected]>
To: Sam Hopkins <[email protected]>
Subject: Re: AoE driver inclusion into standard kernel release
In-Reply-To: <[email protected]>

mail [email protected]
Subject: AoE inclusion into 2.4.x
Greetings,

I would like to offer our driver for inclusion into the 2.4 kernel. Marcelo suggested (strongly) that I make this request officially to the list.

The driver code is available from:

	http://www.coraid.com/support/aoe-1.6.tar

I'm now going to cheat and copy our prior correspondence. Please CC me on any comments / questions as I'm not a lkml subscriber.

Cheers,

Sam Hopkins
[email protected]

-----

Quoted 1 line

From cyclades.com!marcelo.tosatti Wed Apr 21 12:18:18 EDT 2004

Date: Wed, 21 Apr 2004 13:16:35 -0300
From: Marcelo Tosatti <[email protected]>
To: Sam Hopkins <[email protected]>
Subject: Re: AoE driver inclusion into standard kernel release
In-Reply-To: <[email protected]>

Hi Sam,

Sorry for taking so long to answer.

Currently 2.4.x is in maintenance-only mode, but I dont see a problem including your driver. I strongly suggest you to post this announcement to the linux-kernel mailing list to get wider review, please.

Thanks!

Quoted 42 lines · Sam Hopkins

Hello,

I would like to discuss getting my AoE protocol driver into the standard kernel release. AoE (ATA over Ethernet) is an open protocol for transmitting ATA commands over ethernet frames. AoE has an assigned IANA Ethernet type of 0x88a2 and an assigned linux kernel major number of 152. The AoE driver translates block device requests into AoE requests to EtherDrive blades, the only device to currently support the AoE protocol. An EtherDrive blade is a small computer that answers AoE requests by performing them upon the attached ATA drive. Each blade slides into a rackmount shelf where it gets access to power and ethernet. Each blade has its own ethernet address. For more information about EtherDrive blades, please visit www.coraid.com.

The current addressing scheme for accessing EtherDrive blades has changed since the initial registration of the major number. Previously we had a mechanism by which users associated blades with a minor number and then accessed the blade by accessing /dev/etherd/<minor number>. This was scrapped in favor of a simpler method. The current mechanism relies on the notion that each blade slides into one of 18 slots in an addressable rack. The blade in shelf 0, slot 1 is addressed as /dev/etherd/00:01. For char devices, the following is now employed:

	0 = /dev/etherd/ctl
	1 = /dev/etherd/stat
	2 = /dev/etherd/err

We may add /dev/etherd/raw back in the future, but currently it is no longer necessary. Documentation at lanana.org should probably be updated to reflect these changes.

We've been using the AoE driver with EtherDrive blades and software raid for a while and I feel the driver is ready for this next step. Please let me know if there is anything you need for me to do in order to move this along.

Cheers,

Sam Hopkins
[email protected]

#1 lore raw
Quoted 6 lines · Sam Hopkins

Greetings,

I would like to offer our driver for inclusion into the 2.4 kernel. Marcelo suggested (strongly) that I make this request officially to the list.

Quoted 6 lines

The driver code is available from:

	http://www.coraid.com/support/aoe-1.6.tar

I'm now going to cheat and copy our prior correspondence. Please CC me on any comments / questions as I'm not a lkml subscriber.

Overall it seems reasonably clean, though the following combination seems a bit silly:

if(access_ok(VERIFY_WRITE, arg, sizeof(long))) if(!copy_to_user((long *) arg, &d->disk.ssize, sizeof (long)))

#2 lore raw
In reply to Sam Hopkins
Quoted 5 lines · Sam Hopkins

Greetings,

I would like to offer our driver for inclusion into the 2.4 kernel. Marcelo suggested (strongly) that I make this request officially to the list.

Oh, why submit to 2.4 before 2.6? Usually the other way around is preferred.

Jeff