FreeBSD source tree https://cgit-dev.freebsd.org/src/atom?h=releng%2F12.2 2019-08-27T04:01:09Z 2019-08-27T04:01:09Z Alexander Motin mav@FreeBSD.org 2019-08-27T04:01:09Z urn:sha1:90fda462d4849aeb19e3ae82082adaca7f26feb0 With support for multiple namespaces and multiple ports in NVMe there is now a need for reliable unique namespace identification alike to SCSI. Sponsored by: iXsystems, Inc. 2019-08-13T19:17:44Z Alexander Motin mav@FreeBSD.org 2019-08-13T19:17:44Z urn:sha1:7dae83c6425ed24fa5ef38857ce3a1ec801c535d 2019-02-12T00:53:14Z Alexander Motin mav@FreeBSD.org 2019-02-12T00:53:14Z urn:sha1:798c2b4b44872fe82a887b9b7b502dab16ad5cfc This fixes BIO_ORDERED semantics while also improving performance by: - sleeping also before BIO_ORDERED bio, as defined, not only after; - not queueing BIO_ORDERED bio to taskqueue if no other bios running; - waking up sleeping taskqueue explicitly rather then rely on polling. On Samsung SSD 970 PRO this shows sync write latency, measured with `diskinfo -wS`, reduction from ~2ms to ~1.1ms by not sleeping without reason till next HZ tick. On the same device ZFS pool with 8 ZVOLs synchronously writing 4KB blocks shows ~950 IOPS instead of ~750 IOPS before. I suspect ZFS does not need BIO_ORDERED on BIO_FLUSH at all, but that will be next question. 2019-01-25T19:56:06Z Alexander Motin mav@FreeBSD.org 2019-01-25T19:56:06Z urn:sha1:05729e22d0a8fc27428fb5029993142704b3d6e3 Previous code typically crashed in case of NVMe device unplug or even clean detach while some I/Os are still in flight. To fix this the new code calls disk_gone() and waits for confirmation of all references gone before calling disk_destroy(), freeing other resources and allowing controller detach. While there, fix disk lists locking and reimplement unit numbers assignment. 2017-11-27T14:52:40Z Pedro F. Giffuni pfg@FreeBSD.org 2017-11-27T14:52:40Z urn:sha1:718cf2ccb9956613756ab15d7a0e28f2c8e91cab Mainly focus on files that use BSD 2-Clause license, however the tool I was using misidentified many licenses so this was mostly a manual - error prone - task. The Software Package Data Exchange (SPDX) group provides a specification to make it easier for automated tools to detect and summarize well known opensource licenses. We are gradually adopting the specification, noting that the tags are considered only advisory and do not, in any way, superceed or replace the license texts. 2017-08-04T03:40:01Z Warner Losh imp@FreeBSD.org 2017-08-04T03:40:01Z urn:sha1:8a5d94f94dc26ec417c6aa149c34cc5794ab7362 Introduce hw.nvme.use_nvd tunable. This tunable allows both nvd and nda to be installed in the kernel, while allowing only one of them to create devices. This is an all-or-nothing setting, and you can't change it after boot-time. However, it will allow easier A/B testing. Differential Revision: https://reviews.freebsd.org/D11825 2017-01-12T08:53:10Z Alexander Motin mav@FreeBSD.org 2017-01-12T08:53:10Z urn:sha1:17160457b41ff26b4384d8f82ed176a2844eb272 While doing it, introduce respective constants in geom_disk.h. MFC after: 1 week 2016-07-19T19:19:03Z Scott Long scottl@FreeBSD.org 2016-07-19T19:19:03Z urn:sha1:3c7910115096f6ca3a03ddbbe8ccdc73389c7524 2016-07-19T19:09:23Z Scott Long scottl@FreeBSD.org 2016-07-19T19:09:23Z urn:sha1:49e20d2420925a967e78702fe743655badce9e3e logic. Switch to a 5us delay since most NVME devices can easily do 200,000 iops. Submitted by: imp MFC after: 3 days Sponsored by: Netflix, Inc. 2016-07-19T03:13:51Z Scott Long scottl@FreeBSD.org 2016-07-19T03:13:51Z urn:sha1:a498975ef70e00fa25af3cc0a6ad5560f9bcff85 MFC after: 3 days Sponsored by: Netflix, Inc.