On Sat, Sep 15, 2012 at 9:52 PM, Nicholas A. Bellinger <nab@xxxxxxxxxxxxxxx> wrote: > On Fri, 2012-09-14 at 23:09 -0400, Daniel Lane wrote: >> On Fri, Sep 14, 2012 at 8:56 PM, Daniel Lane <dracodan@xxxxxxxxx> wrote: >> > On Fri, Sep 14, 2012 at 5:15 PM, Andy Grover <agrover@xxxxxxxxxx> wrote: >> >> On 09/14/2012 08:20 AM, Daniel Lane wrote: >> >>> I have another problem now though, not sure if I need to submit a new >> >>> E-mail to the list or not (for tracking), let me know if I do. >> >> >> >> Yeah I'd say you could start a new email thread next time, no biggie. >> >> >> >> (Also, please don't top post.) >> >> >> >>> The new problem: After rebuilding my initramfs and rebooting, I found >> >>> that roughly every 30 seconds the server would completely freeze for >> >>> about 30 seconds, but then go back to responding. This happens from >> >>> SSH and the local console, and no matter what I'm doing. Strangely a >> >>> continuous ping to the server never drops a packet or has any increase >> >>> response time. I tried using top to see if there was something causing >> >>> heavy CPU/Memory/IO but from what I could tell (top stops updating >> >>> during freezes). I checked /var/log/messages but there was nothing >> >>> pertaining to the problem. >> >>> >> >>> The one big thing that concerns me is the fact that I am NOT using a >> >>> "supported" HBA, my HBA is a QLogic QLA2362. I was curious whether >> >>> there was a reason why you guys say that only the 24xx+ cards are >> >>> compatible, were there issues with the older cards or were they just >> >>> not tested? >> >> >> >> I haven't been involved with the development but I believe support for >> >> pre-PCIe cards was explicitly REMOVED from the driver before it went >> >> into the mainline kernel. I'm surprised it even partially works. >> >> >> >> Regards -- Andy >> > >> > >> > That's disappointing... sure hope I win this auction for a QLE2462 so >> > I can give targetcli another shot. I would be interested in knowing >> > why it was removed since it was already in the driver and the card >> > worked perfectly with SCST. The abundance of PCI-X cards on ebay for >> > dirt cheap are what attracted me to the idea of using fiber channel >> > instead of iSCSI in the first place. With VMware labs being such a >> > popular use for this kind of thing I would think supporting old >> > hardware would be something to pay attention to... then again, I guess >> > the people doing home labs aren't paying customers (that wasn't meant >> > to sound derogatory). >> > > > So hch requested that we remove all legacy support for pre 24xx series > hardware for the initial upstream merge to avoid the mess of legacy > callbacks supporting !IS_FWI2_CAPABLE within the original out of tree > code. > > As there was not any interest at the time to keep these bits around, it > made the sense for mainline to not include this if no-one was actually > prepared to 'own it' for merge. > >> > Sorry about the top post, that's where the gmail web client tries to >> > put everything by default. >> > >> > Thanks for the info! >> > Dan >> >> Part of this doesn't sound right, here is what the linux-iscsi site >> says for support: >> "QLogic 2400 Series (single-port QLx246x), 4GBFC" >> >> That first x can be an E for PCI-Express or an A for PCI-X, so is it >> the 236x cards that support was removed for? Is the info in the wiki >> a mistake? >> > > The bulk of what got removed was supporting the really old > !IS_FWI2_CAPABLE cases. From your experience so far it sounds like the > 236x series target mode silicon might be close enough to semi-easily > forward-port legacy support to mainline code, preferably without all of > the extra !IS_FWI2_CAPABLE cruft required for <= 22xx silicon. > > I'm not sure where the line between ~IS_FWI2_CAPABLE is for 236x series, > but I would imagine the Qlogic/Linux folks would know. > > So the commit from early this year to drop can be found in the > historical lio-core-2.6.git tree here: > > qla2xxx: Drop legacy pre 24xx target mode support > http://git.kernel.org/?p=linux/kernel/git/nab/lio-core-2.6.git;a=commitdiff;h=db763cc35ca816d01cc25ee5f488f8b43bb74af5 > > Your more than welcome to try a forward-port (ask questions on the list) > of this code to suit your needs. > > Thanks! > > --nab > Thanks for the info, unfortunately I still have a few questions. The comment on the commit you sent has the erroneous line of "QLA 24xx (PCI-E) adapters for mainline target LLD code." Problem is that the QLA cards aren't PCI-E, they're PCI-X. I see that the wiki has been updated as of yesterday to reflect that, but I wanted to confirm this since PCI-X QLA2462 cards are roughly $30 on ebay where PCI-E QLE2462 cards are roughly $130. Another thing I just noticed that could be an even bigger issue is that the article says "QLogic 2400 Series (single-port QLE246x), 4GBFC", key word being single-port... Can you confirm that only single port 2400 series cards are supported? The price of 25xx cards is completely out of the question for a lab setup, as it is this has already gone from a cheap alternative to iSCSI to an impractical expense. If indeed only single port cards are supported, would it be possible to use two separate cards? Also, the wiki should be updated to say "QLogic 2400 Series (single-port QLE2460), 4GBFC". Thanks, Dan -- To unsubscribe from this list: send the line "unsubscribe target-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
