On 09/22/2010 02:08 PM, FUJITA Tomonori wrote: > On Wed, 22 Sep 2010 21:05:01 +0900 > FUJITA Tomonori <fujita.tomonori at lab.ntt.co.jp> wrote: > >> On Tue, 21 Sep 2010 14:33:31 +0200 >> Alexander Nezhinsky <alexandern at Voltaire.COM> wrote: >> >> This patch breaks the dependence of iscsi login/text code on the >> static structures within struct iscsi_conn. > Can you start with the simpler approach like this? > > http://lists.wpkg.org/pipermail/stgt/2010-July/003970.html It's been a long time since i promised a patch with the login stuff. This proved to be more difficult than i envisioned. The login code is quite messy, and i failed in changing it to receive only char * buffers and sizes. So the simple approach did not work. The compromise was to pass pdu structures instead. Note, that the patch that i sent does not break anything. > As I said before, duplicating the code is fine by now. If the new code > really fixes the iser bugs, better to merge the code quickly. As long > as there is no risk to break the tcp iscsi code, the patchset can be > merged. iser code (that you've got) fixes the bug, but all the objections and reservations, that you have raised previously (about the big patch etc.) are still valid. So i am a bit surprised that your highest priority was the merge. > And please add bidi support to your new code. No feature regression > compared with the current code, please. I can work on it, sure. Again, it has not been my perception that this was the only blocker. I believe that i have a rough idea now, about how to proceed in a series of relatively small controlled patches and get the things right, with a cleaner design and minimal code duplication. If you prefer to merge new iser "as is", as quickly as possible, I'm all in for it, of course. I can fix bidi, as I said. But as a minimum, i'd prefer avoiding login/text code duplication. I'm less bothered by other duplicated stuff, like connections, iscsi logic etc. Two reasons. First, it will shrink the patch. Second, now after working out the login code, i see that its iser-duplicated version has some holes (not that i'm happy with the original one, but this is another story). Thus i think the patch i have sent is worth taking. If you are afraid of breaking iscsi/tcp by applying the patch directly to "master", perhaps another approach will work. You can start a new branch, from the "master", apply the last patch there, and i'll start sending patches destined for the new branch. I'll resend the new iser code, w/out duplicated login, then add a patch for bidi support and a few other small fixes i've made recently to iser. Anyway, the nearest milestone will be having a working iser + bidi, with common login. When this is attained and tested, you can merge the new branch. And then we can go on with the "deduplication" attempts. Alexander -- To unsubscribe from this list: send the line "unsubscribe stgt" in the body of a message to majordomo at vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html |