[00:55:46] *** Joins: tkulasek_ (tkulasek@nat/intel/x-mxnhlpvlfonlvmiz) [03:38:05] *** Quits: dlw (~Thunderbi@114.255.44.143) (Ping timeout: 256 seconds) [03:43:06] *** Joins: johnmeneghini (~johnmeneg@pool-100-0-53-181.bstnma.fios.verizon.net) [04:58:49] *** Joins: lyan (~lyan@2605:a000:160e:2124:4a4d:7eff:fef2:eea3) [07:51:23] *** Quits: gila (~gila@5ED74129.cm-7-8b.dynamic.ziggo.nl) (Ping timeout: 256 seconds) [07:51:54] *** Joins: gila (~gila@static.214.50.9.5.clients.your-server.de) [08:10:26] *** Quits: gila (~gila@static.214.50.9.5.clients.your-server.de) (Ping timeout: 260 seconds) [08:13:15] *** Joins: gila (~gila@static.214.50.9.5.clients.your-server.de) [08:21:18] *** Joins: travis-ci (~travis-ci@ec2-54-211-251-47.compute-1.amazonaws.com) [08:21:19] (spdk/master) bdev: add delete_malloc_bdev call (Tomasz Zawadzki) [08:21:19] Diff URL: https://github.com/spdk/spdk/compare/635a1aa8a962...59dee78bd6bb [08:21:19] *** Parts: travis-ci (~travis-ci@ec2-54-211-251-47.compute-1.amazonaws.com) () [08:35:48] *** Quits: gila (~gila@static.214.50.9.5.clients.your-server.de) (Ping timeout: 268 seconds) [08:38:10] *** Joins: gila (~gila@5ED74129.cm-7-8b.dynamic.ziggo.nl) [08:59:08] bwalker: can you check out the discussion on your zcopy patch? [09:16:30] *** Quits: lyan (~lyan@2605:a000:160e:2124:4a4d:7eff:fef2:eea3) (Remote host closed the connection) [09:18:57] I've been tracking it [09:19:17] I was working on updating the nvmf target to use the changes because I think that may be the best way to communicate how to use it [09:24:24] we have a couple patches in the test queue [09:24:31] just a few here and there [09:25:13] those test machines just get hammered all day long every day [09:25:30] surprised we haven't worn out some of the SSDs yet [10:27:59] *** Joins: travis-ci (~travis-ci@ec2-54-211-251-47.compute-1.amazonaws.com) [10:28:00] (spdk/master) bdev: add delete_iscsi_bdev call (Tomasz Zawadzki) [10:28:00] Diff URL: https://github.com/spdk/spdk/compare/9539edca0231...6be654aad72f [10:28:00] *** Parts: travis-ci (~travis-ci@ec2-54-211-251-47.compute-1.amazonaws.com) () [10:28:09] *** Joins: travis-ci (~travis-ci@ec2-54-81-206-218.compute-1.amazonaws.com) [10:28:10] (spdk/master) event: Add tsc stats to the reactor (Vishal Verma) [10:28:10] Diff URL: https://github.com/spdk/spdk/compare/6be654aad72f...5b822d4c925a [10:28:10] *** Parts: travis-ci (~travis-ci@ec2-54-81-206-218.compute-1.amazonaws.com) () [10:29:44] *** Joins: travis-ci (~travis-ci@ec2-54-81-206-218.compute-1.amazonaws.com) [10:29:46] (spdk/master) nvme: add nvme error injection support (Changpeng Liu) [10:29:46] Diff URL: https://github.com/spdk/spdk/compare/5b822d4c925a...c391fed3e1c3 [10:29:46] *** Parts: travis-ci (~travis-ci@ec2-54-81-206-218.compute-1.amazonaws.com) () [10:31:23] *** Joins: travis-ci (~travis-ci@ec2-54-81-206-218.compute-1.amazonaws.com) [10:31:24] (spdk/v18.04.x) blob: fix load of non-multiple-of-8 masks (Daniel Verkamp) [10:31:24] Diff URL: https://github.com/spdk/spdk/compare/727a80b32839...5bf3a5f3835b [10:31:24] *** Parts: travis-ci (~travis-ci@ec2-54-81-206-218.compute-1.amazonaws.com) () [11:07:17] *** Joins: travis-ci (~travis-ci@ec2-54-81-206-218.compute-1.amazonaws.com) [11:07:18] (spdk/master) pkgdep: support pycodestyle in check_format.sh (John Meneghini) [11:07:18] Diff URL: https://github.com/spdk/spdk/compare/c391fed3e1c3...b7322649dbbe [11:07:18] *** Parts: travis-ci (~travis-ci@ec2-54-81-206-218.compute-1.amazonaws.com) () [11:19:17] *** Quits: gila (~gila@5ED74129.cm-7-8b.dynamic.ziggo.nl) (Quit: Textual IRC Client: www.textualapp.com) [11:27:52] *** Joins: tkulasek (~tkulasek@134.134.139.74) [11:29:32] *** Quits: tkulasek_ (tkulasek@nat/intel/x-mxnhlpvlfonlvmiz) (Ping timeout: 276 seconds) [12:14:18] *** Quits: tkulasek (~tkulasek@134.134.139.74) (Ping timeout: 260 seconds) [14:05:16] bwalker: https://review.gerrithub.io/#/c/spdk/spdk/+/416562/ [14:11:26] done [14:13:38] *** Joins: travis-ci (~travis-ci@ec2-54-211-251-47.compute-1.amazonaws.com) [14:13:39] (spdk/master) test/iscsi: allow choosing net framework for iSCSI tests (Tomasz Zawadzki) [14:13:39] Diff URL: https://github.com/spdk/spdk/compare/79cffd37cadd...19dd8ea57676 [14:13:39] *** Parts: travis-ci (~travis-ci@ec2-54-211-251-47.compute-1.amazonaws.com) () [14:14:33] *** Joins: travis-ci (~travis-ci@ec2-184-73-113-211.compute-1.amazonaws.com) [14:14:34] (spdk/master) spdkcli: add Ceph rbd bdev create option (Karol Latecki) [14:14:34] Diff URL: https://github.com/spdk/spdk/compare/19dd8ea57676...356b68a7c1be [14:14:34] *** Parts: travis-ci (~travis-ci@ec2-184-73-113-211.compute-1.amazonaws.com) () [14:15:34] *** Joins: travis-ci (~travis-ci@ec2-184-73-113-211.compute-1.amazonaws.com) [14:15:35] (spdk/master) rpc/bdev: construct RPC return a string instead array (Tomasz Zawadzki) [14:15:35] Diff URL: https://github.com/spdk/spdk/compare/356b68a7c1be...160850f0a605 [14:15:35] *** Parts: travis-ci (~travis-ci@ec2-184-73-113-211.compute-1.amazonaws.com) () [14:15:59] gracias [14:16:19] the test pool backlog was down to only 6 or 7 patches and I just couldn't have that [14:18:38] *** Joins: travis-ci (~travis-ci@ec2-54-167-158-100.compute-1.amazonaws.com) [14:18:39] (spdk/v18.04.x) blob: always use uint64_t to represent page_idx (Jim Harris) [14:18:39] Diff URL: https://github.com/spdk/spdk/compare/5bf3a5f3835b...f49fa23d9ad0 [14:18:39] *** Parts: travis-ci (~travis-ci@ec2-54-167-158-100.compute-1.amazonaws.com) () [14:37:00] I switched bdevperf read path to the new zcopy locally [14:37:08] and it works fine, if the bdev module supports zcopy [14:37:14] my zcopy emulation is broken though [14:37:44] not shocked [14:37:54] *** Joins: travis-ci (~travis-ci@ec2-54-167-158-100.compute-1.amazonaws.com) [14:37:55] (spdk/master) ocssd: vector commands (Jakub Radtke) [14:37:55] Diff URL: https://github.com/spdk/spdk/compare/160850f0a605...c182da27b58a [14:37:55] *** Parts: travis-ci (~travis-ci@ec2-54-167-158-100.compute-1.amazonaws.com) () [14:56:40] looking at that patch now [14:56:51] i really like how this is coming along [14:57:17] I found the first bug, on to the second [14:57:25] bt [14:57:33] i'll wait to review this patch i think [14:57:36] whoops - obviously I'm in gdb hah [14:57:47] lol [14:59:30] should we even allow bdev modules that don't support read and write? [15:00:02] i guess there's a use case for a read-only device [15:00:06] we were thinking for read-only [15:00:14] but now I'm not so sure - we could enforce that in a different way [15:00:32] through descriptors - giving modules a hook on open [15:00:56] which is better because then a bdev can be temporarily read-only [15:02:01] zero copy should be possible with zcopy_start(populate=true) and zcopy_end(commit=false) if the backend only supports read [15:02:06] but i'm fine with how you have it for now [15:02:49] the reason I required write is because we can't know if that zcopy_start buffer is the "real" data or not [15:02:57] i.e. if you memcpy into it [15:03:01] whether that counts [15:03:37] for malloc bdev it's the real deal - so requiring just read would let you bypass that and perform data modifications [15:05:21] i meant io_supported [15:05:31] totally agree - the desc has to be write-enabled for zcopy [15:05:44] *** Joins: Abbott_ (0cda5282@gateway/web/freenode/ip.12.218.82.130) [15:07:10] I am using perf for nvmef drive with kernel based nvmef target and get "nvmet: ctrl 1 keep-alive timer (10 seconds) expired!". Anything I missed configuring? [15:09:47] And followed with "nvmet: ctrl 1 fatal error occurred!" [15:09:52] I believe I fixed that on master last week [15:10:01] I know I did in the fio plugins [15:10:31] Thank you. I will take the new master and try it. Will the fix work for perf? [15:11:17] I just looked - I think I only fixed it in the bdev and nvme fio_plugins [15:11:35] the problem is that the performance tools don't continually poll the admin queue [15:11:46] but polling the admin queue is what triggers the keep alive processing [15:12:51] ppelplin posted a patch to do the two-pass examine - https://review.gerrithub.io/#/c/spdk/spdk/+/413913/ [15:13:04] @bwalker: Ok. Will try with fio_plugins of the new master. [15:13:27] jimharris: I saw that - my first question is do we want to do more examine callbacks, or pass an enum to examine? [15:13:43] @bwalker: Have a good weekend. [15:14:05] i like the separate examine callbacks, because it makes the bdev implementation so darn simple - the config-based examines are always synchronous [15:15:11] that's true [15:15:22] i'm not sure how an enum would work actually - we want the config-based modules a chance to claim them first [15:15:36] it would call examine multiple times, with a different enum for each phase [15:15:45] but synchronous vs. asynchronous doesn't work out [15:15:58] so if the calling semantics is actually different, then different calls is better [15:19:01] what about examine_io instead of examine_disk? [15:19:08] maybe examine_disk is good enough [15:29:44] ok, emulated bdev zcopy read path is working [15:29:53] with bdevperf [15:30:07] I'll replace scsi next, then we can actually eliminate the null pointer hack on read [16:34:51] drv: https://review.gerrithub.io/#/c/spdk/spdk/+/416575/2/lib/nvme/nvme_fabric.c [16:51:24] jimharris: I posted a response [18:43:44] *** Quits: Abbott_ (0cda5282@gateway/web/freenode/ip.12.218.82.130) (Ping timeout: 260 seconds) [21:38:07] *** Quits: johnmeneghini (~johnmeneg@pool-100-0-53-181.bstnma.fios.verizon.net) (Quit: Leaving.) [21:39:53] *** Joins: johnmeneghini (~johnmeneg@pool-100-0-53-181.bstnma.fios.verizon.net) [21:49:02] *** Quits: johnmeneghini (~johnmeneg@pool-100-0-53-181.bstnma.fios.verizon.net) (Quit: Leaving.)