[00:01:00] *** Joins: sbasierx (sbasierx@nat/intel/x-imzncydprhpwxrqc) [00:28:44] *** Joins: ziyeyang_ (~ziyeyang@192.55.54.42) [00:51:51] *** Quits: ziyeyang_ (~ziyeyang@192.55.54.42) (Quit: Leaving) [01:13:35] *** Quits: Param (~Param@157.49.66.148) (Quit: Param) [01:42:49] *** Joins: Param (~Param@157.49.2.35) [03:02:08] *** Quits: Param (~Param@157.49.2.35) (Quit: Param) [04:16:36] *** Joins: Param (~Param@180.151.248.58) [05:23:38] Hi everyone, can I ask that someone runs autotest.sh with iSCSI tests on a physical server? Preferably one with FC26. After weekend I see lots of fails in these tests when running on Jenkins. [05:24:49] I have a SIGSEGV during *Problem Description:* [05:24:50] *Steps to Reproduce:* [05:24:50] *Add Logs where possible* [05:24:50] *Expected Results:* [05:24:50] *Actual Results:* [05:24:52] *Additional notes:* [05:25:06] oops, sorry, bad copy-paste:) [05:25:52] I have a SIGSEGV during ./test/iscsi_tgt/lvol/iscsi_lvol.sh [06:15:50] *** Parts: Param (~Param@180.151.248.58) () [06:16:55] *** Joins: Param (~Param@180.151.248.58) [06:26:44] *** Quits: Param (~Param@180.151.248.58) (Quit: Param) [07:09:30] *** Quits: sbasierx (sbasierx@nat/intel/x-imzncydprhpwxrqc) (Quit: Going offline, see ya! (www.adiirc.com)) [08:12:47] hi klateck - i'm trying it on my system now [08:24:29] *** Joins: Param (~Param@157.49.176.41) [08:37:46] klateck: can you try https://review.gerrithub.io/#/c/401479/ [09:23:25] jimharris, FYI I can get the crypto_null device to create OK so must be missing something small somewhere, progress though [09:24:40] does the aesni module show up on the application link command line? [09:25:10] for example: [09:25:13] yeah I just found it, that's what I was missing however when I tried that before it spewed out a zillion error messages [09:25:25] so I had it commented out [09:27:22] something is correctly connecting the dpdk lib and the intel crypto lib - supposed to be just one env var though which is set and works for me when I build/run the crypto perf app [09:27:29] isn't I mean, not is [09:28:12] env var? [09:28:30] there's definitely some Makefile changes needed in SPDK too [09:29:26] the env var that points to the Intel lib is correctly set and works when I build/run the perf test app. however I can't include the aesni lib in env.mk without getting a ton of link/undefined reference errors that look like functions from the Intel lib [09:30:04] maybe I need to add the Intel lib there too? I kinda figured that would all be behind the scenes [09:30:36] crypto perf app has its own Makefile - you should look at it to see what libraries it has to link [09:31:41] it'll need to link the Intel library too - I'm not sure what the env var does but it certainly isn't enough to link in an extra library [09:32:11] is it an LD_PRELOAD or some other env var? [09:34:38] AESNI_MULTI_BUFFER_LIB_PATH [09:34:47] yeah, I'm adding it to the list now [09:34:57] that env var is needed for DPDK to build/link it in [09:38:08] thinking for SPDK that we should have user pass in this AESNI path to configure, then we can set up this environment variable on behalf of the user [09:38:19] but you can do that later [09:38:47] I got it working now [09:39:09] yeah, I've got a few things hardcoded for now [09:39:44] I have pushed the code https://review.gerrithub.io/c/401314/ Commands Supported and Effects Log Page [09:40:46] cool [09:44:45] thanks Param - it looks like there were some failures found in the test pool [09:45:05] if you go to the gerrithub link that you posted, you will see a -1 vote from the test pool [09:45:22] in the comments section, there is a link that should take you to the results page [09:45:47] looks like a style issue found with our scripts/check_format.sh script [09:46:05] space missing between RESERVED and the closing */ [09:46:39] you can run this script on your system before pushing to gerrithub [09:49:05] yeah sure.. I will do that.. for some reason pool log page is not getting opened in my browser [09:51:55] you mean this link does not load? https://ci.spdk.io/spdk/builds/review/05294d9f1c103f0452783a27eba5572608c3c988.1519622567/ [09:52:45] No this under the resources we have Pool Log.. I am not able to open that one [09:53:45] sethhowe, drv: what is supposed to be in "POOL LOGS"? [09:54:16] Param: you can ignore the POOL LOGS for now - instead look at the per-test machine links further down the page [09:54:25] that is the log from autorun_post.py; it should normally be empty if nothing went wrong [09:54:34] for example, you can see that machine centos6 passed [09:54:37] but fedora-01 failed [09:54:42] click on fedora-01 [09:54:52] then on build.log [09:58:52] Yeah..Okay.. Thanks for the detailed explanation.. [09:59:50] And found that running scripts/check_format.sh did not do autoformatting.. [10:00:01] should i do the change manually in that case [10:00:32] once this patch is merged https://review.gerrithub.io/#/c/392240/ the pool logs will also contain information about which tests were completed on the machines. [10:01:15] drv: maybe we should rename POOL LOGS? i understand now what it's there for but to someone new to a test results page it's not really clear [10:01:41] send suggestions to sethhowe :) [10:01:56] it's really the logs from the pool aggregator; not sure what a concise name for that would be [10:02:45] Param: do you mean that when you run scripts/check_format.sh, it does not fix this style issue? [10:03:24] it won't for that comment-style check; that's enforced by a separate 'git grep', not astyle [10:03:34] yeah - i see that now [10:03:37] jimharris: Maybe we could call it supplemental logs? [10:03:44] Okay.. can I check that locally [10:03:44] so yes, you should fix it manually [10:03:50] sethhowe: +1 [10:09:31] jimharris: for https://review.gerrithub.io/#/c/401065/ - it looks like no new code is added; is the checking that the commit message mentions already in place before this point? [10:11:33] yes - see _spdk_blob_persist() [10:12:48] it's always been there - this patch could have been done independently of the rest of the patches in this series [10:15:37] ok, cool - I'm always happy to remove lines of code [10:24:21] i have pushed the code after changing manually [10:29:28] jimharris: I've reviewed everything up to https://review.gerrithub.io/#/c/401256/ if you wanted to re-upload the last couple of changes with fixes [10:31:28] oh yeah, with all the linking correctly it turns out that the EAL options don't need to be modified, it's either/or wrt the EAL options or calling the API later to init a virtual device... [10:31:38] peluse: perfect [10:31:43] yup yup [10:31:55] mszwed doesn't like my two spaces after a period [10:32:05] was anyone else taught that in school? [10:32:20] I think that's up there with "tabs vs spaces" and other holy wars :) [10:33:39] http://www.slate.com/articles/technology/technology/2011/01/space_invaders.html [10:33:46] (I was taught to put two spaces after a period too) [10:45:09] that how I learnt it. I think [10:46:01] drv: patches updated [11:28:05] *** Quits: Param (~Param@157.49.176.41) (Quit: Param) [11:28:59] *** Joins: Param (~Param@157.49.176.41) [11:37:27] should i add the assignee field in the review.gettithub.com [11:38:03] *should i add any one to the assignee in the review.gerrit.hub.io [11:38:15] for a commit that I had made [11:47:38] *** Quits: Param (~Param@157.49.176.41) (Quit: Param) [12:19:38] hi Param - no, you should not set the assignee field - the maintainers will see the patch in our view once it has passed the test pool [14:00:09] *** Quits: gila (~gila@5ED4D9C8.cm-7-5d.dynamic.ziggo.nl) (Read error: Connection reset by peer) [14:14:03] drv: i see locking in nvme_pcie_qpair_process_completions for the adminq - can you remind me why we don't have this in the nvme_rdma equivalent? [14:14:14] peluse: just FYI, we merged a patch that will require your passthru bdev to be rebased: https://github.com/spdk/spdk/commit/2939b715d05ad384f52f48d17678dd987a8d5883 [14:16:54] jimharris: the user-visible spdk_nvme_ctrlr_process_admin_completions() takes the ctrlr_lock, so that should handle locking when accessed via the API [14:17:07] but we may be missing a lock when calling it from within other NVMe library code [14:18:21] PCIe is probably different because it supports multi-process, since the adminq nvme_pcie_qpair_insert_pending_admin_request() stuff is supposed to be protected by ctrlr_lock too [14:18:56] ah - that makes sense [14:19:10] it could definitely use a clean-up, though [14:19:20] I really don't like this recursive mutex stuff - makes it hard to reason about locking [18:01:20] *** Joins: Shuhei (caf6fc61@gateway/web/freenode/ip.202.246.252.97) [18:39:28] *** Quits: sethhowe (sethhowe@nat/intel/x-gzrjnmobpivphihb) (Remote host closed the connection) [18:53:24] *** Joins: ziyeyang_ (ziyeyang@nat/intel/x-cbospnurwbpngwnf) [19:48:39] *** Quits: Shuhei (caf6fc61@gateway/web/freenode/ip.202.246.252.97) (Ping timeout: 260 seconds) [20:11:14] *** Joins: Param (~Param@157.49.190.18) [20:24:18] *** Quits: Param (~Param@157.49.190.18) (Quit: Param) [22:07:46] *** Joins: Shuhei (caf6fc61@gateway/web/freenode/ip.202.246.252.97) [22:21:06] *** Joins: Param (~Param@157.49.190.18) [22:59:37] *** Joins: sbasierx (sbasierx@nat/intel/x-ujvjxsmxaqdtzffg)