FreeBSD source tree https://cgit-dev.freebsd.org/src/atom?h=release%2F12.3.0 2021-03-15T03:00:38Z 2021-03-15T03:00:38Z Alexander Motin mav@FreeBSD.org 2021-01-28T20:53:49Z urn:sha1:3ef86cd7c3ccef27d42bb7cbdecd41f0fdd81f69 Move software iSCSI tunables/sysctls into kern.icl.soft subtree. Replace several hardcoded length constants there with variables. While there, stretch the limits to better match Linux' open-iscsi and our own initiator with new MAXPHYS of 1MB. Our CTL target is also optimized for up to 1MB I/Os, so there is also a match now. For Windows 10 and VMware 6.7 initiators at default settings it should make no change, since previous limits were sufficient there. Tests of QD1 1MB writes from FreeBSD over 10GigE link show throughput increase by 29% on idle connection and 132% with concurrent QD8 reads. MFC after: 3 days Sponsored by: iXsystems, Inc. (cherry picked from commit b75168ed24ca74f65929e5c57d4fed5f0ab08f2a) 2021-03-15T02:54:56Z Alexander Motin mav@FreeBSD.org 2021-03-03T20:21:26Z urn:sha1:6cd4542786f929ea57604662d7d5537f13261382 It closes tiny race when the flag could be set between being cleared and the space is checked, that would create us some more work. The flag setting is protected by both locks, so we can clear it in either place, but in between both locks are dropped. MFC after: 1 week (cherry picked from commit afc3e54eeee635a525c88e4678cc38e3219302c3) 2021-03-15T02:54:45Z Alexander Motin mav@FreeBSD.org 2021-03-03T16:58:04Z urn:sha1:2cd7a99cebafb932fc91e5a1c20cde1d4000f4c5 I think it allowed to avoid some TX thread wakeups while the socket buffer is full. But add there another options if ic_check_send_space is set, which means socket just reported that new space appeared, so it may have sense to pull more data from ic_to_send for better TX coalescing. MFC after: 1 week (cherry picked from commit aff9b9ee894e3e6b6d8c7e4182d6b973804df853) 2021-03-15T02:54:34Z Alexander Motin mav@FreeBSD.org 2021-03-02T23:39:44Z urn:sha1:cec95c50654097b8cb8e595abda7cae7a01d5884 Also remove stray STAILQ_REMOVE_AFTER(), not causing problems only because STAILQ_SWAP() fixed corrupted stqh_last. MFC after: 1 week (cherry picked from commit df3747c6607be12d48db825653e6adfc3041e97f) 2021-03-15T02:53:14Z Alexander Motin mav@FreeBSD.org 2021-03-02T21:07:22Z urn:sha1:cb89ac5a1daa7dd8d2306460ecbf5d8dac90b9a2 There are sessions without socket that are not disconnecting yet. MFC after: 3 weeks (cherry picked from commit 06e9c710998b83a3be21f7f264187fff5d590bc3) 2021-03-15T02:52:57Z Alexander Motin mav@FreeBSD.org 2021-03-02T04:31:34Z urn:sha1:3034c0dab755d29eb24d0187cfc00501b660a9c1 Before m_cat() each time traversed through all the coalesced chain. MFC after: 1 week (cherry picked from commit b85a67f54a40053e75658a17c620b89bafaba67f) 2021-03-15T02:45:25Z Alexander Motin mav@FreeBSD.org 2021-02-22T17:23:35Z urn:sha1:6469aab051dee71c88e27aa906c18efa09e77189 Instead of 2-4 socket reads per PDU this can do as low as one read per megabyte, dramatically reducing TCP overhead and lock contention. With this on iSCSI target I can write more than 4GB/s through a single connection. MFC after: 1 month (cherry picked from commit 6895f89fe54e0858aea70d2bd2a9651f45d7998e) 2021-03-04T14:40:37Z Alexander Motin mav@FreeBSD.org 2021-02-02T18:37:13Z urn:sha1:3e1da9a9f85bbfbb216829b5de528b86d16c888f DataSN for solicited Data-Out is per-R2T. Since we handle whole R2T in one go, we don't need to store it anywhere, especially in global per-command structure. This may allow us to handle multiple R2T per command at once, if we decide, or may be relax locking. Rename the second use of that field to io_referenced_task_tag. MFC after: 1 month (cherry picked from commit 3dd2a7a5ea2f1641c7525f692eed416fa02c28e6) 2021-02-23T00:33:20Z Alexander Motin mav@FreeBSD.org 2021-01-24T19:23:04Z urn:sha1:0004227346d0ceea713b456300794d55e787a5e6 RFC 7143 (11.2.1.8): An ITT value of 0xffffffff is reserved and MUST NOT be assigned for a task by the initiator. The only instance in which it may be seen on the wire is in a target-initiated NOP-In PDU (Section 11.19) and in the initiator response to that PDU, if necessary. MFC after: 1 month 2021-02-04T02:31:54Z Alexander Motin mav@FreeBSD.org 2021-01-21T02:33:14Z urn:sha1:bf96a6c762a79d974bd71322cd37e23984e0f3f3 For hardware offload solicited data may potentially be handled more efficiently than unsolicited due to direct data placement. Or there can be some unsolicited write buffering limitations. It may create situations where FirstBurstLength limit is really useful. Software driver though has no those factors, having to do memcopy() any way and having no so hard limit on the temporary storage. Same time more active use of unsolicited transfers allows to avoid some of Ready To Transfer (R2T) PDU round-trip times and processing. This change effectively doubles from 64KB to 128KB the maximum size of write command that can be transferred within one link RTT. Tests of (64KB, 128KB] QD1 writes mixed with simultaneous QD8 reads over the same connection, increasing RTT, shows almost double write speed and half latency, while we should be able to afford few megabytes of RAM for additional buffering on a target these days. Sponsored by: iXsystems, Inc. (cherry picked from commit ff751ee05c939eceab25c26ad60b1d56f989aec9)