[00:28:19] *** Joins: dlw (~Thunderbi@119.103.179.51) [00:48:42] *** Quits: dlw (~Thunderbi@119.103.179.51) (Ping timeout: 246 seconds) [00:52:39] *** Joins: dlw (~Thunderbi@171.82.179.15) [00:59:39] *** Joins: tkulasek (~tkulasek@192.55.54.45) [01:47:27] *** Quits: lhodev (~lhodev@66-90-218-190.dyn.grandenetworks.net) (Ping timeout: 240 seconds) [01:48:32] *** Joins: lhodev (~lhodev@66-90-218-190.dyn.grandenetworks.net) [01:56:42] *** Joins: lhodev_ (~lhodev@66-90-218-190.dyn.grandenetworks.net) [01:57:58] *** Quits: lhodev (~lhodev@66-90-218-190.dyn.grandenetworks.net) (Ping timeout: 265 seconds) [02:36:21] *** Quits: dlw (~Thunderbi@171.82.179.15) (Ping timeout: 264 seconds) [05:36:26] *** Joins: dlw (~Thunderbi@123.118.169.166) [06:28:06] *** Quits: tkulasek (~tkulasek@192.55.54.45) (Quit: Leaving) [06:46:47] *** Quits: dlw (~Thunderbi@123.118.169.166) (Ping timeout: 276 seconds) [06:53:30] *** Parts: lhodev_ (~lhodev@66-90-218-190.dyn.grandenetworks.net) ("Textual IRC Client: www.textualapp.com") [06:53:39] *** Joins: lhodev_ (~lhodev@66-90-218-190.dyn.grandenetworks.net) [06:53:50] *** lhodev_ is now known as lhodev [07:03:50] *** Joins: dlw (~Thunderbi@123.118.169.166) [07:12:24] *** Joins: tomzawadzki (tomzawadzk@nat/intel/x-kydheyklqkaibjqy) [07:19:12] *** Quits: dlw (~Thunderbi@123.118.169.166) (Ping timeout: 268 seconds) [07:51:05] *** Quits: tomzawadzki (tomzawadzk@nat/intel/x-kydheyklqkaibjqy) (Ping timeout: 240 seconds) [08:21:26] *** Joins: gila (~gila@5ED4D9C8.cm-7-5d.dynamic.ziggo.nl) [08:24:27] *** Joins: tkulasek (~tkulasek@192.55.54.45) [09:03:31] *** Joins: travis-ci (~travis-ci@ec2-54-198-17-152.compute-1.amazonaws.com) [09:03:32] (spdk/master) bdev: defer freeing bdev_io mempool (Dariusz Stojaczyk) [09:03:32] Diff URL: https://github.com/spdk/spdk/compare/2e1dbc458758...6c0078300b52 [09:03:32] *** Parts: travis-ci (~travis-ci@ec2-54-198-17-152.compute-1.amazonaws.com) () [09:07:37] *** Joins: tomzawadzki (tomzawadzk@nat/intel/x-fxfczvbmhbvreyzg) [09:09:13] jimharris: I pushed a new revision of https://review.gerrithub.io/#/c/408193/ with the comment removed - does everything else look good to you in that patch? [09:10:26] lgtm [09:10:57] good - I'd like to get these in and then retest with Seth's patch applied [09:11:07] hopefully that fixes the intermittent failures that we were seeing yesterday [09:15:58] *** Quits: tkulasek (~tkulasek@192.55.54.45) (Ping timeout: 264 seconds) [09:24:07] jimharris: replied on (io_channel: always set dev->unregister_thread under a mutex) https://review.gerrithub.io/c/408194/ [09:25:29] that explanation works for me [09:26:56] thanks for fixing these issues darsto - we realized the patch had problems last night and I was going to spend a chunk of my day fixing it, but now I'm free to do other things! [09:28:04] :) [09:28:08] we should try to run our tests under Thread Sanitizer at some point (I suspect that will flag a lot of issues, though) [09:28:17] I did at one point - it goes boom [09:28:38] but when I analyzed the issues, I didn't think any of the ones I happened to look at were "real" [09:28:56] we're pretty tricky with our threading stuff [09:29:14] but we should take a look again [09:33:25] *** Joins: travis-ci (~travis-ci@ec2-54-224-75-159.compute-1.amazonaws.com) [09:33:26] (spdk/master) io_channel: always set dev->unregister_thread under a mutex (Dariusz Stojaczyk) [09:33:26] Diff URL: https://github.com/spdk/spdk/compare/6c0078300b52...3ed915618fc5 [09:33:26] *** Parts: travis-ci (~travis-ci@ec2-54-224-75-159.compute-1.amazonaws.com) () [09:34:50] *** Joins: travis-ci (~travis-ci@ec2-54-198-17-152.compute-1.amazonaws.com) [09:34:51] (spdk/master) bdev/iSCSI: Make the connection connect in async mode. (Ziye Yang) [09:34:51] Diff URL: https://github.com/spdk/spdk/compare/3ed915618fc5...ec5cf8a22118 [09:34:51] *** Parts: travis-ci (~travis-ci@ec2-54-198-17-152.compute-1.amazonaws.com) () [09:35:42] *** Joins: travis-ci (~travis-ci@ec2-54-198-17-152.compute-1.amazonaws.com) [09:35:43] (spdk/master) virtio: rename ops->get_queue_num callback (Dariusz Stojaczyk) [09:35:43] Diff URL: https://github.com/spdk/spdk/compare/ec5cf8a22118...0a1a4ce11628 [09:35:43] *** Parts: travis-ci (~travis-ci@ec2-54-198-17-152.compute-1.amazonaws.com) () [09:36:50] *** Joins: travis-ci (~travis-ci@ec2-54-198-17-152.compute-1.amazonaws.com) [09:36:51] (spdk/master) io_channel: add asserts in spdk_io_device_register (Daniel Verkamp) [09:36:51] Diff URL: https://github.com/spdk/spdk/compare/0a1a4ce11628...37039565d0ad [09:36:51] *** Parts: travis-ci (~travis-ci@ec2-54-198-17-152.compute-1.amazonaws.com) () [10:12:32] bwalker: +2 on https://review.gerrithub.io/#/c/407356/ but I added some comments [10:13:38] Daniel and I had the same discussion about embedding the qos structure [10:13:57] I think I'd like to embed it at some point just to reduce chances to leak [10:14:09] I'd have to add a flag to it for enable/disable then [10:14:34] maybe I could trigger off of rate limit, but as other types of restrictions are added a flag may be better overall [10:14:51] yeah, I'd prefer to just have a separate qos_enabled flag [10:15:01] (or the pointer to the QoS struct if we keep it separate) [11:14:15] need one more +2 on this, and then we can merge two virtio test patches: https://review.gerrithub.io/#/c/398915/ [11:30:44] *** Quits: sethhowe (~sethhowe@192.55.54.42) (Remote host closed the connection) [11:32:26] *** Joins: sethhowe (~sethhowe@192.55.54.42) [12:03:50] about those iscsi initiator async connect/disconect patches form Ziye [12:04:36] when you plan to merge all of them so RPC configuration patches can be reworked? [12:05:51] *** Quits: tomzawadzki (tomzawadzk@nat/intel/x-fxfczvbmhbvreyzg) (Ping timeout: 240 seconds) [12:08:04] *** Quits: sethhowe (~sethhowe@192.55.54.42) (Remote host closed the connection) [12:17:31] can you link them? I'll review [12:32:47] the async connect patch has been merged already [12:33:06] and the async disconnect patch has outstanding review comments [12:38:59] *** Joins: travis-ci (~travis-ci@ec2-54-205-141-74.compute-1.amazonaws.com) [12:39:00] (spdk/master) lvol: Make lvol rename RPC arguments required (Ben Walker) [12:39:00] Diff URL: https://github.com/spdk/spdk/compare/37039565d0ad...880e3b153edf [12:39:00] *** Parts: travis-ci (~travis-ci@ec2-54-205-141-74.compute-1.amazonaws.com) () [12:43:59] *** Joins: travis-ci (~travis-ci@ec2-54-198-17-152.compute-1.amazonaws.com) [12:44:00] (spdk/master) nvmf: make spdk_nvmf_subsystem_remove_ns asynchronous (Seth Howell) [12:44:01] Diff URL: https://github.com/spdk/spdk/compare/880e3b153edf...4ca87a01b4f0 [12:44:01] *** Parts: travis-ci (~travis-ci@ec2-54-198-17-152.compute-1.amazonaws.com) () [13:05:29] *** Joins: sethhowe (~sethhowe@192.55.54.42) [13:14:50] *** Quits: sethhowe (~sethhowe@192.55.54.42) (Remote host closed the connection) [13:16:23] drv: Hi Daniel. Do you have a few minutes to discuss some code related to extended LBA support and end-to-end protection? I'm attempting to reconcile a very old patch (from someone within our company that was never upstream'd) against code I see top of tree. It's an area about which I currently lack familiarity and am trying to gain some perspective before a meeting in a couple of hours; i.e. haven't digested all the relevant [13:16:23] sections in the NVMe spec. [13:16:55] I can try to answer - changpe1 would also be a good resource, since he's done a lot of the enabling work for extended LBA/end-to-end protection in SPDK [13:17:27] drv: Thanks. When I did a "git blame", your name appeared alongside the specific code ;-) [13:17:39] I blame him for everything too [13:17:51] *** Joins: sethhowe (sethhowe@nat/intel/x-hzlxmneaddnfjbnr) [13:20:32] Lol. At issue, once upon a time in nvme_ns_identify_update(), there was an expression that was tested before testing another to determine whether SPDK_NVME_NS_EXTENDED_LBA_SUPPORTED was OR'd into ns->flags. That later was changed such that it did *not* evaluate the earlier expression which tested for a non-zero metadata size for a given format and that non-zero pit value. [13:23:12] In both _nvme_ns_cmd_split_request() and _nvme_ns_cmd_rw(), however, they test for SPDK_NVME_NS_DPS_PI_SUPPORTED before adjusting sector_size. The old patch I'm examining eliminates that check and they put in a comment of the form: Allow extended LBA event without Protection Information. (Ignore PI completely). [13:24:20] I'm wondering if this was a bug or a different interpretation of the spec. [13:26:58] the only place we should be considering PI_SUPPORTED is where we do the special case adjustment of the extended LBA size when PI is enabled and PRACT is set [13:27:10] (meaning the device should generate the checksums, and we should not send them) [13:28:02] extended LBA should be usable with or without PI enabled [13:29:19] *** Quits: sethhowe (sethhowe@nat/intel/x-hzlxmneaddnfjbnr) (Remote host closed the connection) [13:31:42] *** Joins: sethhowe (~sethhowe@192.55.54.42) [13:33:25] drv: Hmmm. I need to read some more in the spec, but at least this gives me some more context. It's unfortunate that the patch I'm examining is quite old (and therefore based on much earlier version of the SPDK NVMe driver) so it's not a simple diff. What they had done might have made sense at the time of the version of the driver to which they were applying it, but not now; i.e. what's upstream may fully address the intention [13:33:25] of the patch. I'll just work through it. Thanks. [13:33:33] to be specific, the weird exception where we need to check if PI is enabled is when the metadata size is exactly 8 bytes (the size of the PI info), and the I/O has the PRACT bit set, in which case we should not add the metadata buffer to each block at all [13:33:46] this is section 8.3.1 ("The PRACT Bit") in the NVMe 1.3 spec [13:34:15] (it's possible we've still got it wrong somehow, so definitely double check and let us know if you find anything to the contrary) [13:37:11] Will do. I was also looking at an older version of the NVMe spec (1.2 Gold). Dunno yet how much, if any, the sections relevant to this topic may have been amended or otherwise. Will take a look. [13:37:34] I think it should be the same across all versions [13:37:39] it has some nice pretty pictures too :) [13:44:34] jimharris can you look at this https://review.gerrithub.io/#/c/407355/ and the one after it? [13:44:40] the qos rpcs depend on those two now [13:45:21] i'm on it [13:46:48] lhodev: this patch you're trying to help get rebased - are you able to share that patch and do you know what spdk rev it is based off of? [13:48:15] bwalker: done [13:48:32] jimharris: Actually, I'm attempting to evaluate whether it's even a candidate. I don't know the spdk rev against which it's based, but am only guestimating the time range. The patch was crafted in the Fall of 2016. Prehistoric ;-) [13:51:12] i was thinking more that if we could see the patch, we may be able to decipher what it was trying to do - best case, we say "yeah, we already fixes this but in a different way" [13:51:25] fixes => fixed [13:52:11] *** Joins: travis-ci (~travis-ci@ec2-54-205-141-74.compute-1.amazonaws.com) [13:52:12] (spdk/master) bdev: Fix data type mismatch for io_per_sec (Ben Walker) [13:52:12] Diff URL: https://github.com/spdk/spdk/compare/4ca87a01b4f0...87211463a267 [13:52:12] *** Parts: travis-ci (~travis-ci@ec2-54-205-141-74.compute-1.amazonaws.com) () [13:52:25] *** Joins: travis-ci (~travis-ci@ec2-54-224-75-159.compute-1.amazonaws.com) [13:52:26] (spdk/master) scripts/rpc.py: use dashes in --long-args (Daniel Verkamp) [13:52:26] Diff URL: https://github.com/spdk/spdk/compare/87211463a267...3ac5d089dcfc [13:52:26] *** Parts: travis-ci (~travis-ci@ec2-54-224-75-159.compute-1.amazonaws.com) () [14:24:25] jimharris: that's awfully nice of you to offer. I'm heading into a meeting where I hope I can learn what I can do about these patches. Back to you a bit later. [15:11:46] hey, annoying thing I get with Hexchat internally here at Intel, wondering if anyone else get this and has addressed it. If I am offline for a while, like moving my laptop from home to work, when I fire Hexchat back up again I get all the history I missed but its buried behind every private IRC conversation I've ever had. All that history shows up last so I have to scrollback like 20 pages to see IRC history from like the last hour or [15:11:46] something. Anyone? [15:14:30] i'm using LimeChat - when I fire it up, i'll get that same history, but I get it per-channel, meaning I don't have to filter #spdk stuff out from other channels or private conversations [15:15:16] I get it too peluse - haven't tried to figure out a way to get rid of it [15:15:21] it's annoying, but not annoying enough [15:15:28] yet at least [15:16:15] it's an interaction between ZNC sending the last X lines and the irc client not interpreting them as private messages [15:16:57] bwalker, OK at least I'm not nuts. thanks [15:17:09] or at least your not nuts for that reason [15:17:14] ha! you're [15:17:28] before the apostrophe police show up [15:17:30] * jimharris looks at drv [15:17:38] LOL [15:17:56] sometimes you feel like a nut, and sometimes you don't [15:18:00] I should have said "at least Im not nut's" [16:58:52] bwalker, you think maybe if you delete my ZNC account and recreate that I might at least get rid of the last several months of crap? Or send me an email (not private IRC please!) on how to do it and I will :) [17:16:41] jimharris, I know you gave the crypto patch a once over before, the latest is a pretty big rework of when/how init is done. Probably not worth looking at right now, let me get QAT basically working first in case I need to change things around again - hard to tell with this stuff until implementing it as the example test/perf code is all very specific to, well, either test or perf :) [17:17:05] I'll let you know when I think it's "stabilized" but it does indeed work as posted now [17:17:18] (for AESNI-MB) [17:17:46] ok - i have plenty of other patches to look at :-) let me know once you're ready for me to take a look again [17:18:35] will do - indeed I know you do which is why I mentioned it [18:26:42] *** Joins: dlw (~Thunderbi@114.255.44.143) [19:05:18] *** Quits: dlw (~Thunderbi@114.255.44.143) (Remote host closed the connection) [19:23:42] *** Joins: dlw (~Thunderbi@114.255.44.143) [21:11:52] *** Joins: travis-ci (~travis-ci@ec2-54-205-141-74.compute-1.amazonaws.com) [21:11:53] (spdk/master) vhost: Handle null pointer when dumping SCSI JSON config (Ben Walker) [21:11:53] Diff URL: https://github.com/spdk/spdk/compare/3ac5d089dcfc...c1f7f02cfefb [21:11:53] *** Parts: travis-ci (~travis-ci@ec2-54-205-141-74.compute-1.amazonaws.com) () [22:16:03] peluse: do you have 'Auto Clear Query Buffer' option set? [22:16:31] it's on znc website in user settings [22:17:27] it should make znc send each private message just once [23:38:24] *** Joins: tomzawadzki (tomzawadzk@nat/intel/x-hwqejthksggadckl)