[02:20:37] *** Quits: Shuhei (caf6fc61@gateway/web/freenode/ip.202.246.252.97) (Quit: Page closed) [07:19:42] *** Quits: lhodev (~Adium@66-90-218-190.dyn.grandenetworks.net) (Quit: Leaving.) [07:20:26] *** Joins: lhodev (~Adium@inet-hqmc01-o.oracle.com) [07:22:04] *** Parts: lhodev (~Adium@inet-hqmc01-o.oracle.com) () [09:36:19] jimharris: I still get some valgrind "Conditional jump or move depends on uninitialised value(s)" errors in the unit tests after your iscsi patch - do you see those as well? [09:38:13] i'm not seeing them [09:38:28] trying non-debug build now [09:39:10] no - i'm not seeing any issues like that [09:41:30] I'll send a patch - it's the same thing except with ig instead of pg [09:42:38] https://review.gerrithub.io/391000 [09:43:03] spdk_iscsi_pg_map_add_ig_map() does ig->ref++, and nobody else initialized ig [09:48:44] do we have valgrind on any of the systems in the test pool running the unit tests? [09:50:30] sethhowe_ is turning it on now, I believe [09:52:06] yeah - i actually see this failure now on fedora-03 [09:52:11] on one of my unrelated patches [09:52:18] also, somebody should write up a changelog entry for https://github.com/spdk/spdk/commit/0709cc48c90b2a83edabb529e91c25df894344e1 [09:52:22] guess it got missed in the review [09:52:59] you're right - that's my fault [09:53:18] (we should also avoid churning the RPC method names without a good reason) [09:55:24] jimharris: fedora-03 is now running valgrind. [10:12:01] *** Joins: lhodev (~Adium@inet-hqmc01-o.oracle.com) [11:25:00] hmm, it looks like there's a double free in the NBD stop path - both the polling thread and the RPC call end up calling spdk_nbd_stop() [11:27:13] sorry, not actually a double free, but the free() in spdk_nbd_stop() races with the nbd->dev_fd access in nbd_start_kernel() after NBD_DO_IT ioctl returns [11:50:33] jimharris: I was able to reproduce the error from issue #221 on github. in kernel version 13.13, the symlink we are looking for is up one directory from where we expect it to be in 14.04. [11:51:06] excellent! [11:51:11] kernel version 13.13? [11:51:37] sorry 3.13 [11:52:44] so maybe we can check for the link in the current place, and if its empty, then check one directory up - could you see if you can create a patch to setup.sh for that? [11:53:08] yeah, I will do that now. [11:53:15] and post that patch to github #221 so the original submitter can try it out [11:53:20] thanks seth [12:15:37] Hey guys. I created a Trello account, but it's unclear to me how I may request to join the SPDK team thereon. [12:16:11] hi lhodev - what is your Trello handle? [12:16:29] @lancehartmann [12:17:15] ok - i've added you to the SPDK team [12:17:19] Thank you! Rec'd notification. [12:29:09] jimharris: https://review.gerrithub.io/#/c/391019/ for the setup.sh fix [12:38:06] proposed NBD fix is here: https://review.gerrithub.io/391021 [12:50:40] sethhow_: patch looks good, but in the commit message you should just say "older kernels" instead of calling out 3.13 specifically [13:17:51] *** Quits: ChanServ (ChanServ@services.) (shutting down) [13:23:58] *** Joins: ChanServ (ChanServ@services.) [13:23:58] *** wolfe.freenode.net sets mode: +o ChanServ [13:27:31] bwalker: on the blobstore API changes - what if we just always allocate an I/O channel internally for an blobstore-level API calls? [13:27:37] an => any [13:28:14] I think that avoids a lot of confusion around which blobstore API calls take an I/O channel parameter and which ones don't [13:49:34] it would also eliminate the need for spdk_bs_[alloc|free]_io_channel [15:16:40] jimharris: with the new API you can perform blobstore-wide operations from any thread, so we'd have to create the channel on demand each time [15:16:52] for an operation like creating a blob, do you think that is too heavy-weight? [15:17:06] are we doing that often enough that creating a channel every time would be a problem? [15:17:18] normally it's just bumping a ref count [15:17:21] of course, it won't create it if you have blobs open on the current thread because it will have ref counts [15:17:25] right [15:17:41] but if you eliminate alloc_io_channel entirely, you won't always have a ref count [15:17:46] so we might need to go optimize the io_channel code but that's super well contained [15:18:12] that was my only hesitation in just saying we always handle the channels behind the scenes [15:18:27] I agree it makes the API look a lot nicer [15:18:34] *** bwalker_ is now known as bwalker [15:18:43] *** ChanServ sets mode: +o bwalker [15:23:58] *** Joins: Shuhei (72b10586@gateway/web/freenode/ip.114.177.5.134) [15:28:22] Hi Jim, Daniel, if possible I would like to hear your thought quickly. Preventing an LUN from being shared among multiple iSCSI targets should be done by another patch? Thanks. [15:34:34] hi Shuhei, I think that makes sense, but I'm not the iSCSI expert, so I'd liek to hear jimharris's thoughts [15:45:31] HI Daniel, thank you. [15:52:30] hi Shuhei - today, without the "lun name" patch that is out for review - a bdev cannot be attached to more than one LUN in one iSCSI target node [15:53:13] the "lun name" patch would remove this restriction - but we can add it back by having the SCSI layer "claim" the bdev after opening it [15:53:55] we need to think about this would work in a target that supports both iSCSI and NVMe-oF though [16:01:31] Hi Jim, thank you and I agreed. [16:07:47] *** Quits: lhodev (~Adium@inet-hqmc01-o.oracle.com) (Quit: Leaving.) [16:08:35] *** Quits: Shuhei (72b10586@gateway/web/freenode/ip.114.177.5.134) (Ping timeout: 260 seconds) [22:04:47] *** Quits: vermavis (~vermavis@192.55.54.41) (*.net *.split) [22:04:47] *** Quits: darsto (~dstojacx@192.55.54.41) (*.net *.split) [22:04:47] *** Quits: sethhowe_ (sethhowe@nat/intel/x-jvyugvawtxkoywle) (*.net *.split) [22:04:47] *** Quits: guerby (~guerby@april/board/guerby) (*.net *.split) [22:04:49] *** Quits: klateck (~klateck@192.55.54.41) (*.net *.split) [22:04:49] *** Quits: pawelkax (~pawelkax@192.55.54.41) (*.net *.split) [22:04:49] *** Quits: gangcao (~gangcao@192.55.54.41) (*.net *.split) [22:04:49] *** Quits: mszwed (~mszwed@192.55.54.41) (*.net *.split) [22:04:50] *** Quits: jstern (~jstern@192.55.54.41) (*.net *.split) [22:04:50] *** Quits: cunyinch (~cunyinch@192.55.54.41) (*.net *.split) [22:04:50] *** Quits: pwodkowx (~pwodkowx@192.55.54.41) (*.net *.split) [22:04:50] *** Quits: qdai2 (~qdai2@192.55.54.41) (*.net *.split) [22:04:50] *** Quits: kjakimia (~kjakimia@192.55.54.41) (*.net *.split) [22:04:51] *** Quits: sage__ (~quassel@64.111.99.127) (*.net *.split) [22:04:52] *** Quits: jkkariu (~jkkariu@192.55.54.41) (*.net *.split) [22:04:52] *** Quits: jimharris (~jimharris@192.55.54.41) (*.net *.split) [22:04:52] *** Quits: pniedzwx (~pniedzwx@192.55.54.41) (*.net *.split) [22:04:52] *** Quits: changpe1 (~changpe1@192.55.54.41) (*.net *.split) [22:04:52] *** Quits: nKumar (uid239884@gateway/web/irccloud.com/x-szmizwohfgjvstgr) (*.net *.split) [22:04:52] *** Quits: peluse (~peluse@192.55.54.41) (*.net *.split) [22:04:53] *** Quits: lgalkax (~lgalkax@192.55.54.41) (*.net *.split) [22:04:55] *** Quits: drv (daniel@oak.drv.nu) (*.net *.split) [22:04:55] *** Quits: darsto_ (~dstojacx@89-68-135-211.dynamic.chello.pl) (*.net *.split) [22:04:57] *** Quits: ChanServ (ChanServ@services.) (*.net *.split) [22:11:44] *** Joins: drv (daniel@oak.drv.nu) [22:11:44] *** Joins: guerby (~guerby@april/board/guerby) [22:11:44] *** Joins: sethhowe_ (sethhowe@nat/intel/x-jvyugvawtxkoywle) [22:11:44] *** Joins: darsto (~dstojacx@192.55.54.41) [22:11:44] *** Joins: vermavis (~vermavis@192.55.54.41) [22:11:44] *** Joins: jstern (~jstern@192.55.54.41) [22:11:44] *** Joins: jkkariu (~jkkariu@192.55.54.41) [22:11:44] *** Joins: jimharris (~jimharris@192.55.54.41) [22:11:44] *** Joins: cunyinch (~cunyinch@192.55.54.41) [22:11:44] *** Joins: pniedzwx (~pniedzwx@192.55.54.41) [22:11:44] *** Joins: klateck (~klateck@192.55.54.41) [22:11:44] *** Joins: changpe1 (~changpe1@192.55.54.41) [22:11:44] *** Joins: pwodkowx (~pwodkowx@192.55.54.41) [22:11:44] *** Joins: qdai2 (~qdai2@192.55.54.41) [22:11:44] *** Joins: pawelkax (~pawelkax@192.55.54.41) [22:11:44] *** Joins: kjakimia (~kjakimia@192.55.54.41) [22:11:44] *** Joins: peluse (~peluse@192.55.54.41) [22:11:44] *** Joins: lgalkax (~lgalkax@192.55.54.41) [22:11:44] *** Joins: gangcao (~gangcao@192.55.54.41) [22:11:44] *** Joins: mszwed (~mszwed@192.55.54.41) [22:11:44] *** Joins: darsto_ (~dstojacx@89-68-135-211.dynamic.chello.pl) [22:11:44] *** Joins: nKumar (uid239884@gateway/web/irccloud.com/x-szmizwohfgjvstgr) [22:11:44] *** Joins: sage__ (~quassel@64.111.99.127) [22:11:44] *** wolfe.freenode.net sets mode: +oo drv jimharris [22:11:52] *** Joins: ChanServ (ChanServ@services.) [22:11:52] *** wolfe.freenode.net sets mode: +o ChanServ