[01:06:02] *** Joins: pwodkowx (pwodkowx@nat/intel/x-enturdmasxmyjxcw) [01:42:59] *** Quits: pwodkowx (pwodkowx@nat/intel/x-enturdmasxmyjxcw) (Quit: Leaving) [01:43:39] *** Joins: pwodkowx (pwodkowx@nat/intel/x-ipeoxdlkaybkbfdi) [01:47:56] *** Quits: pwodkowx (pwodkowx@nat/intel/x-ipeoxdlkaybkbfdi) (Client Quit) [02:22:00] *** Joins: pwodkowx (pwodkowx@nat/intel/x-msdsczfqltivhdjy) [02:27:15] *** Joins: pwodkowx_ (pwodkowx@nat/intel/x-lajqervznqniyzbv) [02:27:20] *** Quits: pwodkowx (pwodkowx@nat/intel/x-msdsczfqltivhdjy) (Quit: Leaving) [02:27:43] *** Parts: pwodkowx_ (pwodkowx@nat/intel/x-lajqervznqniyzbv) () [02:28:02] *** Joins: pwodkowx_ (pwodkowx@nat/intel/x-lajqervznqniyzbv) [02:28:15] *** Quits: pwodkowx_ (pwodkowx@nat/intel/x-lajqervznqniyzbv) (Client Quit) [02:28:40] *** Joins: pwodkowx (pwodkowx@nat/intel/x-xzterofghtqhstdo) [06:03:44] *** Quits: pwodkowx (pwodkowx@nat/intel/x-xzterofghtqhstdo) (Ping timeout: 240 seconds) [06:09:16] *** Joins: pwodkowx (pwodkowx@nat/intel/x-htgskzvaeknpucik) [07:15:35] *** Quits: pwodkowx (pwodkowx@nat/intel/x-htgskzvaeknpucik) (Ping timeout: 255 seconds) [07:18:58] *** Joins: lhodev (~Adium@66-90-218-190.dyn.grandenetworks.net) [07:43:52] bwalker, question for you on bstore MD thread thing on Trello at https://trello.com/c/KH8p4TQp [07:48:04] also, I can't seem to move cards in the blobstore board that you made, need to change permissions or something? I'll move the couple of blobstore items over from the Things To Do board once I'm able [07:50:44] *** Joins: pwodkowx (pwodkowx@nat/intel/x-komooyakdtydjqpt) [08:22:45] *** Quits: lhodev (~Adium@66-90-218-190.dyn.grandenetworks.net) (Quit: Leaving.) [09:04:45] peluse: I think Jim or Daniel added you to the trello board, but I tried to add everyone else [09:05:32] bwalker, OK maybe one of them can check permissions. Let me know after you get a chance to read the question [09:05:45] I responded already [09:06:02] it looks like you added 2 cards to the blobstore board - is it still not working? [09:09:09] bwalker, OK, will look. So I can add cards but I can't move like the MD thread card into the doing column for some reason [09:09:22] oh [09:09:29] I'm such a trello noob [09:09:38] I'll see if I can't figure it out [09:10:31] bwalker, OK, thanks. Yeah saw your reply, glad you like Scenario A :) [09:11:37] bwalker, I'll also put together a quick list of funcs, scope and what structs they touch that need to be protected to make sure I've got the full picture in that area correct... thx [09:17:39] *** Joins: lhodev (~Adium@inet-hqmc02-o.oracle.com) [11:04:41] *** Quits: pwodkowx (pwodkowx@nat/intel/x-komooyakdtydjqpt) (Ping timeout: 240 seconds) [11:17:49] *** Joins: pwodkowx (~pwodkowx@192.55.54.36) [11:56:42] bwalker: removing g_nvmf_tgt: Could we just embed discovery_log_page in spdk_nvmf_tgt instead of allocating it separately? [11:57:48] sure - but I don't think I changed that [11:57:50] did I? [11:58:18] no - it's still a separate allocation - just thinking that would be a nice simplification (in a different patch) [11:58:23] oh, yeah [11:58:48] drv wants to rework how discovery is done so that the user adds subsystems to the discovery service manually [11:58:50] it looks like previously we would leak that discovery log page? [11:58:55] so I was staying out of that [11:58:57] ok [11:59:05] yeah it would have been leaked [11:59:06] now it isn't [12:00:21] i just pushed three of your patches - looks like this patch we're talking about needs to be rebased [12:00:26] ok, will do [12:00:33] (and I have like 6 more on top of it ready to go) [12:00:58] I'm writing the patch that actually fixes the scaling problem now - I've got all the infrastructure in place finally [12:34:01] ugh - I just fumblefingered and pushed the first patch in your env series [12:34:20] your == bwalker's [13:16:16] I am working on cleaning up the bdev APIs to all work in blocks instead of bytes [13:16:30] I'm thinking about introducing new APIs with _blocks to keep compat - any opinions on whether that's worth it? [13:16:48] e.g. adding spdk_bdev_read_blocks() vs. changing the existing spdk_bdev_read() [14:09:36] went ahead with the "add new APIs" approach for now, still open to comments: https://review.gerrithub.io/#/c/376255/ [14:45:25] *** Quits: patrickmacarthur (~smuxi@2606:4100:3880:1240:85a6:a4ef:6ebd:4ce4) (Ping timeout: 246 seconds) [14:48:22] *** Joins: patrickmacarthur (~smuxi@2606:4100:3880:1240:9dd:9387:a602:df88) [16:41:58] Dan, I am looking at the patch. Does it support block sizes other than 4K? Is the block size configurable? [16:50:49] jkkariu: yes, it doesn't change what block sizes are supported, just which units the offset and length use [16:51:08] the current bdev API functions (spdk_bdev_read(), etc.) all take offset and length in bytes [16:51:28] they already have to be block size multiples; the new APIs just simplify usage when you already have blocks, and it allows us to remove a couple of divisions in the I/O path [16:51:58] (thanks for checking out the patch) [16:58:10] Dan, thanks for the info. Do we need test cases for block sizes other than 512? [16:59:48] yes, currently we don't do any automated testing of NVMe devices with non-512 block sizes [17:00:00] although we do have malloc bdevs with 4096 block size in some tests, I believe [18:39:12] *** Quits: lhodev (~Adium@inet-hqmc02-o.oracle.com) (Quit: Leaving.) [23:53:13] *** Joins: pwodkowx_ (pwodkowx@nat/intel/x-gehkussnkdrrkvwr)