[00:59:46] *** Joins: tkulasek_ (tkulasek@nat/intel/x-qnytcfsoyymxrlxy) [01:58:50] *** Joins: tkulasek (~tkulasek@134.134.139.73) [02:01:35] *** Quits: tkulasek_ (tkulasek@nat/intel/x-qnytcfsoyymxrlxy) (Ping timeout: 256 seconds) [02:04:39] *** Joins: alekseymmm (050811aa@gateway/web/freenode/ip.5.8.17.170) [02:23:14] *** Quits: alekseymmm (050811aa@gateway/web/freenode/ip.5.8.17.170) (Ping timeout: 260 seconds) [02:42:27] *** Joins: alekseymmm (050811aa@gateway/web/freenode/ip.5.8.17.170) [03:43:57] *** Quits: dlw (~Thunderbi@114.255.44.143) (Ping timeout: 240 seconds) [05:45:51] *** Joins: lyan (~lyan@2605:a000:160e:2124:4a4d:7eff:fef2:eea3) [05:53:46] *** Joins: tkulasek_ (tkulasek@nat/intel/x-qbyowibbbjejgqji) [05:57:41] *** Quits: tkulasek (~tkulasek@134.134.139.73) (Ping timeout: 256 seconds) [07:39:16] *** Joins: tomzawadzki (~tomzawadz@134.134.139.76) [07:47:34] *** Quits: tomzawadzki (~tomzawadz@134.134.139.76) (Ping timeout: 260 seconds) [09:03:33] *** Quits: alekseymmm (050811aa@gateway/web/freenode/ip.5.8.17.170) (Quit: Page closed) [09:20:55] *** Quits: lyan (~lyan@2605:a000:160e:2124:4a4d:7eff:fef2:eea3) (Remote host closed the connection) [09:25:24] jimharris: I see one spot that stores an LBA as uint32_t in blobstore - _spdk_blob_serialize_extent() - which definitely looks wrong [09:25:31] but I'm not sure if there are more that I'm missing in my grep [09:27:47] looking [09:28:22] in your test, did you create the lvol and then immediately export it over nvme-of? [09:28:39] or did you shut nvmf-tgt down and then restart it? [09:29:42] *** Joins: lyan (~lyan@2605:a000:160e:2124:4a4d:7eff:fef2:eea3) [09:29:55] I immediately exported it [09:30:06] that's probably why it worked [09:30:18] I tried to shut down the target and re-start it, but that was where I was seeing all sorts of problems (CRC errors, or the lvol just didn't show up at all after restart) [09:30:36] I'm going to try that again now after changing that lba to uint64_t and see if that fixes it [09:32:47] hmm, it still doesn't seem to work (construct lvol store, construct lvol bdev, see lvol bdev in get_bdevs; kill app, restart app; get_lvol_stores shows the lvs, but no lvol bdev shows up) [09:33:01] blobcli -D? [09:35:03] working on that now [09:35:37] let me re-create it from scratch to be sure it's clean [09:37:41] https://gist.github.com/dverkamp-intel/e33d577569afeef7fa25aacfbc1945d7 [09:38:03] I created the lvol with size=3811716 MB [09:38:37] which should be 952929 clusters, so that looks right [09:39:55] but still, when I turn around and start up spdk_tgt and construct the NVMe bdev, it just exposes Nvme0n1, no lvol bdev [09:42:27] (btw, that test and dump was with my patch to change the serialize_extent lba to 64 bits - going to try it on master now) [09:44:41] it seems to get the same blobdump output on master, which makes sense, since the LBA itself is at the beginning of the disk (so it doesn't get truncated) [10:05:08] *** Quits: lyan (~lyan@2605:a000:160e:2124:4a4d:7eff:fef2:eea3) (Remote host closed the connection) [10:23:28] *** Joins: lyan (~lyan@2605:a000:160e:2124:4a4d:7eff:fef2:eea3) [10:44:53] *** Joins: travis-ci (~travis-ci@ec2-54-162-144-157.compute-1.amazonaws.com) [10:44:54] (spdk/master) bdev: add spdk_bdev_queue_io_wait() (Jim Harris) [10:44:54] Diff URL: https://github.com/spdk/spdk/compare/714776c02874...32d7c91cbc48 [10:44:54] *** Parts: travis-ci (~travis-ci@ec2-54-162-144-157.compute-1.amazonaws.com) () [10:46:24] *** Joins: travis-ci (~travis-ci@ec2-54-147-45-122.compute-1.amazonaws.com) [10:46:25] (spdk/master) bdevperf: Enable the debug -L option (Ziye Yang) [10:46:25] Diff URL: https://github.com/spdk/spdk/compare/32d7c91cbc48...1c3c50a6cb44 [10:46:25] *** Parts: travis-ci (~travis-ci@ec2-54-147-45-122.compute-1.amazonaws.com) () [10:53:23] *** Joins: travis-ci (~travis-ci@ec2-54-147-45-122.compute-1.amazonaws.com) [10:53:24] (spdk/master) bdev/nvme: Refactor code to create a controller (Ben Walker) [10:53:24] Diff URL: https://github.com/spdk/spdk/compare/1c3c50a6cb44...c5b861fcec5a [10:53:24] *** Parts: travis-ci (~travis-ci@ec2-54-147-45-122.compute-1.amazonaws.com) () [11:07:47] *** Quits: darsto (~darsto@89-68-124-254.dynamic.chello.pl) (Quit: /quit) [11:08:05] *** Joins: darsto (~darsto@89-68-124-254.dynamic.chello.pl) [12:24:16] sethhowe_: looking at your last bdev_io patch, I think ::link should also be moved into the internal data structure - you can do that in a separate patch though [12:27:37] *** Joins: travis-ci (~travis-ci@ec2-54-147-45-122.compute-1.amazonaws.com) [12:27:38] (spdk/master) bdev: finish encapsulating members of spdk_bdev_io (Seth Howell) [12:27:38] Diff URL: https://github.com/spdk/spdk/compare/c5b861fcec5a...b55efd2ac769 [12:27:38] *** Parts: travis-ci (~travis-ci@ec2-54-147-45-122.compute-1.amazonaws.com) () [12:36:52] *** Joins: travis-ci (~travis-ci@ec2-54-162-144-157.compute-1.amazonaws.com) [12:36:53] (spdk/master) test: fix annoying error from hash pdflatex (John Meneghini) [12:36:53] Diff URL: https://github.com/spdk/spdk/compare/b55efd2ac769...65c48bfb9574 [12:36:53] *** Parts: travis-ci (~travis-ci@ec2-54-162-144-157.compute-1.amazonaws.com) () [12:38:20] jimharris: right you are, thanks. [12:38:53] *** Quits: tkulasek_ (tkulasek@nat/intel/x-qbyowibbbjejgqji) (Ping timeout: 256 seconds) [12:39:28] drv: https://review.gerrithub.io/#/c/spdk/spdk/+/416231/ [12:39:53] yeah, i was concerned about those other fields too [12:39:54] i think i might have figured out the misreported namespace size issue [12:39:58] but we have lba_count as uint32_t all over the place [12:40:02] so there may be more fixes necessary [12:40:17] we do - but those are all for a single I/O [12:40:21] or at least most of them are [12:40:35] but for sure we need to scrub it [12:40:37] hmm, true, probably OK (although maybe not for large unmap?) [12:40:45] but yeah, I can respin this patch with all 3 of those as uint64_t [12:41:30] i'll bet for that issue that was submitted, his SSD ends up with a number of clusters that's divisible evenly by 8 [12:42:53] instead of ((segment >> (i % 8)) & 1U), you could do: [12:43:46] if ((segment & (1 << (i % 8)))... well crap, that really is not better [12:44:37] yeah, it isn't great either way, probably a wash in terms of efficiency too [12:45:22] we could have a separate mask that rotates through along with i, but that seems even more convoluted [12:45:41] *** Joins: alekseymmm (bcf3adf1@gateway/web/freenode/ip.188.243.173.241) [12:48:01] *** Quits: alekseymmm (bcf3adf1@gateway/web/freenode/ip.188.243.173.241) (Client Quit) [12:50:14] this still only explains the issue if the bug submitter is shutting down and restarting, right? [12:50:27] if it is created and exported over NVMe-oF in one run, it should all be fine [12:51:13] (aside from any other similar bugs where we truncate things down to uint32_t elsewhere) [12:56:13] correct [13:20:04] my 4 TB device supports both 512 and 4096 sector sizes - I'm going to try to reformat it as 512 and verify that I can reproduce the bug [13:20:53] *** Joins: travis-ci (~travis-ci@ec2-54-162-144-157.compute-1.amazonaws.com) [13:20:54] (spdk/master) build: add -lcrypto (OpenSSL) to SYS_LIBS (Daniel Verkamp) [13:20:54] Diff URL: https://github.com/spdk/spdk/compare/65c48bfb9574...fdf5c1d7d267 [13:20:54] *** Parts: travis-ci (~travis-ci@ec2-54-162-144-157.compute-1.amazonaws.com) () [13:24:14] yeah, blobstore is definitely truncating the size somewhere - I get a 1.7 TB bdev now [13:24:33] (even in the same run - no serialization involved) [13:45:01] even with your fixes? [13:47:32] I ran it on master, but I can try it with my fixes [13:47:42] I don't think the serialize one should matter, though, since i'm not reloading it from disk [13:47:58] there's a lot of uint32_t to audit [13:47:59] hmmm [14:08:33] *** Quits: lyan (~lyan@2605:a000:160e:2124:4a4d:7eff:fef2:eea3) (Remote host closed the connection) [14:33:08] jimharris, bwalker: I simplified the first blobstore mask fix a little more: https://review.gerrithub.io/#/c/spdk/spdk/+/416230/ [14:33:18] and then a couple of follow-up cleanups, but we don't need those for correctness [15:10:10] *** Joins: travis-ci (~travis-ci@ec2-54-147-45-122.compute-1.amazonaws.com) [15:10:11] (spdk/master) nvmf: Reimplement abort handling (Ben Walker) [15:10:11] Diff URL: https://github.com/spdk/spdk/compare/fdf5c1d7d267...7d87273f60f8 [15:10:11] *** Parts: travis-ci (~travis-ci@ec2-54-147-45-122.compute-1.amazonaws.com) () [15:16:40] sethhowe_: I think you can put the 'error' union in bdev_io inside the internal struct too [15:17:00] since status is already in internal, there's no way to know which branch of error is valid [15:17:11] and modules should be using the helper functions to set those anyway [15:42:50] bwalker: the FreeBSD hugemem limit patch could use your +2: https://review.gerrithub.io/#/c/spdk/spdk/+/416214/ [15:46:38] *** Joins: travis-ci (~travis-ci@ec2-54-147-45-122.compute-1.amazonaws.com) [15:46:39] (spdk/master) bdev/nvme: Add a HostNQN parameter to NVMe bdev creation (Ben Walker) [15:46:39] Diff URL: https://github.com/spdk/spdk/compare/7d87273f60f8...d733654b5c8e [15:46:39] *** Parts: travis-ci (~travis-ci@ec2-54-147-45-122.compute-1.amazonaws.com) () [15:47:06] done [15:51:34] *** Joins: travis-ci (~travis-ci@ec2-54-159-117-127.compute-1.amazonaws.com) [15:51:35] (spdk/master) autotest: reduce HUGEMEM allocation to 2GB on FreeBSD (Jim Harris) [15:51:35] Diff URL: https://github.com/spdk/spdk/compare/d733654b5c8e...696d9eaed023 [15:51:35] *** Parts: travis-ci (~travis-ci@ec2-54-159-117-127.compute-1.amazonaws.com) () [16:37:44] drv: Good point. I left it in because it was used in scsi_nvme.c, but upon closer look, that file is really an internal only file independent of the modules. [16:38:00] yeah, that should be part of the generic bdev layer, not a module [16:40:20] it's just split out into a separate file for readability [18:34:27] *** Joins: dlw (~Thunderbi@114.255.44.143) [20:19:06] *** Joins: johnmeneghini (~johnmeneg@pool-100-0-53-181.bstnma.fios.verizon.net) [20:19:37] Are we having the community meeting tonight, or is it going to be a bug crawl? [20:20:29] tonight will be a normal community meeting [20:20:58] OK. I've added a bunch of stuff to the Trello agenda and I think Madu will be attending too. [20:21:07] Thanks [22:20:54] darsto, drv: looks like there is a memory region we can pick that ASAN won't interfere with for mmap [22:20:55] https://groups.google.com/forum/#!msg/address-sanitizer/w7wFo4Mnj9w/iYCBhrtiDwAJ [22:22:50] *** Joins: travis-ci (~travis-ci@ec2-107-22-7-197.compute-1.amazonaws.com) [22:22:51] (spdk/master) examples: make debug flag option consistent (Ziye Yang) [22:22:51] Diff URL: https://github.com/spdk/spdk/compare/696d9eaed023...8f26b74e020a [22:22:51] *** Parts: travis-ci (~travis-ci@ec2-107-22-7-197.compute-1.amazonaws.com) () [22:30:51] jimharris: this seems to work on my system [22:55:55] *** Quits: johnmeneghini (~johnmeneg@pool-100-0-53-181.bstnma.fios.verizon.net) (Quit: Leaving.)