[00:42:12] *** Joins: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) [00:54:43] *** Quits: ziyeyang_ (~ziyeyang@192.55.54.40) (Remote host closed the connection) [02:11:02] *** Quits: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) (Quit: My Mac Pro has gone to sleep. ZZZzzz…) [02:16:27] *** Joins: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) [04:56:42] *** Quits: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) (Quit: My Mac Pro has gone to sleep. ZZZzzz…) [05:35:55] *** Joins: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) [06:14:27] *** Quits: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) (Quit: My Mac Pro has gone to sleep. ZZZzzz…) [06:15:20] *** Joins: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) [07:07:27] *** Quits: tomzawadzki (~tzawadzk@134.134.139.77) (Ping timeout: 240 seconds) [08:20:05] QQ wrt blobstore... CPL stands for? [09:02:24] completion [09:24:32] LOL, thanks [10:19:41] drv: This is the only patch for vhost that is needed for release https://review.gerrithub.io/#/c/370016/ [10:19:46] and you already reviewed, so I just merged it [10:21:41] ok, sounds good [10:22:23] *** Joins: peluse_ (~peluse@ip72-201-144-44.ph.ph.cox.net) [10:22:40] I've got a series starting here: https://review.gerrithub.io/#/c/370525/ [10:22:46] that I think addresses the hotplug problem [10:22:56] it has a few implementation issues, but the idea is there [10:23:45] I think that's better than deferring put channel, then deferring unregister, then deferring bdev module unload stuff [10:24:25] it means all physical bdev modules do have to be written to handle the channel being deleted inside a completion callback [10:24:30] but so far that hasn't been hard [10:31:41] bwalker: I'm not sure if the aio patch is totally fixing the problem [10:32:26] if spdk_bdev_io_complete() can result in the I/O's channel being deleted, what happens if there are still more events, and the next I/O's completion callback tries to reference the channel? [10:33:16] or is it not possible for the end user to get the channel from the bdev_io? [11:46:20] we assert if you delete a channel and there are outstanding I/O [11:47:03] you basically can't delete a channel within the completion callback if there are other I/O outstanding [11:47:09] it has to be on the last one [11:48:51] that basically means that the aio poller was safe regardless [11:49:15] but I still think the patch to put the events on the stack is a good one [16:08:25] *** Joins: ziyeyang_ (ziyeyang@nat/intel/x-nxjnsfctpwdqasfs) [17:35:04] *** Quits: ziyeyang_ (ziyeyang@nat/intel/x-nxjnsfctpwdqasfs) (Remote host closed the connection) [17:35:25] *** Joins: ziyeyang_ (ziyeyang@nat/intel/x-japbpgozpahpfoun) [20:03:32] *** Quits: ziyeyang_ (ziyeyang@nat/intel/x-japbpgozpahpfoun) (Remote host closed the connection) [20:03:49] *** Joins: ziyeyang_ (~ziyeyang@134.134.139.73) [20:47:02] *** Quits: ziyeyang_ (~ziyeyang@134.134.139.73) (Remote host closed the connection) [20:47:19] *** Joins: ziyeyang_ (~ziyeyang@192.55.54.39) [23:07:01] *** Joins: ziyeyang__ (~ziyeyang@134.134.139.73) [23:07:01] *** Quits: ziyeyang_ (~ziyeyang@192.55.54.39) (Remote host closed the connection) [23:35:37] *** Joins: tomzawadzki (~tzawadzk@192.55.54.42)