[06:44:45] drv, one more thing... you mentioned arb should be primary and the rest secondary there in multi_proc but arb will exit when its done so its only primary for those that happen to fire up while its still running which is a race but most if not all of the time ends up being the apps started before that final while loop, then the first perf in that loop gets primary which is why a delay is needed after that before starting the first identify which will be [06:44:45] secondary. I'll propose the simplest sync for this section of nvme.sh sometime today... it looked so simple :) [06:48:29] doesn't explain the errors you guys were seeing running vtophys earlier though... ugh. Fixes mine though [08:35:24] heads up, pausing the build pool momentarily to perform a couple of tests related to soft roCE. [08:59:41] drv: darek has a long patch of vhost-scsi patches prepping for hot remove and some vhost-scsi/blk generalizations [09:00:12] they almost all look OK to me - just a small piece of feedback on the first patch in the series - could you take a look at them today to see if there's anything else you'd like him to address? [09:00:22] ok, I'll take a look [09:00:32] cool - thanks [09:13:19] jimharris: did you want to mark this as -1 since the series will have to be re-run anyway? https://review.gerrithub.io/#/c/366718/ [09:16:44] same for the patch right after that [09:39:11] i'll leave it up to darek - that code gets fixed when it gets moved in the later patch [09:54:11] peluse: your mock patch still has a dpdk submodule change in it [09:54:22] if you were using stgit I could help you with how to fix it [09:54:45] but since you aren't you'll be stuck with bwalker or drv to help :) [09:55:29] jimharris, heh. I'll just try and delete it, rebase and see what happens [10:05:51] peluse: in your mock_ut patch, MOCK_PASS means passthrough (not pass/fail)? [10:06:57] yes, pass through/ I can change it to PASS_THRU to be clearer [10:07:07] didn't think about pass/fail :) [10:10:22] i posted a question on your mock_ut patch [10:12:10] yeah, just saw that. I couldn't get one macro to work right for both kinds of args. tried so many things I cant tell you how many reasons why... I can try what you sugegsted though [10:12:20] I'm going to be out for a few hrs starting here in just a few though [10:14:21] note also that I didn't make it so you can just return NULL. You can return anything you want for a * stub [11:04:35] peluse: looks like you actually deleted the submodule in your latest patch rather than updating it [11:50:16] *** Joins: jmoyer (jmoyer@nat/redhat/x-wjahwctulstoznvg) [13:53:38] base<=>virtual bdev relationships: https://review.gerrithub.io/#/c/367610/ [13:54:07] bdev open/close semantics: https://review.gerrithub.io/#/c/367611/ [14:31:32] *** Quits: jmoyer (jmoyer@nat/redhat/x-wjahwctulstoznvg) (Quit: .) [15:13:57] drv, yeah I tried deleting and rebasing to get the latest. WTF. I feel like crap so am about done for the day, too much sun or fun or somethin... [15:21:50] this submodule thing is a PITA. I can't figure out how to do anything with it. I delete it and rebase doesn't pick it up. I manually update .gitmodules and then do git submodule update or init or sync and none of them do anything... [15:49:35] jimharris, where/how is the correct dpdk hash kept? [15:49:38] never mind [15:49:52] I think I have it updated correctly now [16:01:45] what is today, Fri? [16:09:53] if you round up [16:31:14] well hell it's Sat then! See ya on Wed next week :) [16:34:21] hey back to nvme.sh sync between pri and sec w/arb, identify and perf. I think just a sleep in the required spots is best. The idea of dropping marker files like the stub sentinel isn't really solid since those apps aren't designed to stick around until killed by the test script. [16:34:58] will submit the patch tomorrow, its working good now [16:35:52] yeah, let's do that for now [16:37:22] rock on