| opendevreview | Ronelle Landy proposed openstack/cyborg master: Fix incorrect set syntax in placement client logging https://review.opendev.org/c/openstack/cyborg/+/994851 | 07:46 |
|---|---|---|
| rlandy | melwitt: hi https://review.opendev.org/c/openstack/cyborg/+/994851 has passing unit tests now. Could you review again when you get a chance? thank you | 08:23 |
| chandank` | sean-k-mooney: Hello | 11:50 |
| *** chandank` is now known as chandankumar | 11:50 | |
| chandankumar | I need to update the cleanup state diagram and few cleanups in the nvme spec. | 11:50 |
| chandankumar | Are you still reviewing it? or will I go ahead and make updates? | 11:51 |
| chandankumar | thank you! | 11:51 |
| chandankumar | sorry pinged on wrong channel earlier | 11:51 |
| sean-k-mooney | chandankumar: im lookign that the mdev sepc currntly | 13:08 |
| sean-k-mooney | i will get to yours after | 13:08 |
| sean-k-mooney | was at lunch | 13:08 |
| chandankumar | sean-k-mooney: thank you, I am working on few changes | 14:02 |
| opendevreview | chandan kumar proposed openstack/cyborg-specs master: Add generic NVMe driver spec with secure cleanup https://review.opendev.org/c/openstack/cyborg-specs/+/985349 | 14:36 |
| chandankumar | jgilaber: thank you for those valuable comments, I have addressed it, thank you! | 14:37 |
| chandankumar | sean-k-mooney: updated | 14:37 |
| jgilaber | thanks chandankumar I'll look again on Monday | 14:37 |
| chandankumar | sure | 14:37 |
| chandankumar | on pci-sim implementation side, I have 4k changes , It support nvme list on host as well as in guest | 14:38 |
| chandankumar | it also support nvme sanitize and zero write I think | 14:38 |
| chandankumar | I will push the changes on monday | 14:38 |
| sean-k-mooney | ok we will need to review that carfully | 14:39 |
| sean-k-mooney | but that is promising | 14:39 |
| chandankumar | all vide coded | 14:39 |
| chandankumar | but it works, I need to take a look at code and restructure it when I get time | 14:40 |
| sean-k-mooney | yes, i did many roudn of selfe review and reinemnet befor ei push the first public verion for review ot try and get the kernel driver ot follow the upstream kernel convetions and make it robost put also protable | 14:40 |
| sean-k-mooney | so there are two portablities target i have in mind currently | 14:41 |
| sean-k-mooney | 1 i want to make it so the pci-sim cna work on ubutnu 2026.04, debian 13 and eventualy centos 10 stream kernel | 14:41 |
| sean-k-mooney | the other portablity seam is it would be nice to make it work on x86_64 adn aarch64 in the future | 14:41 |
| chandankumar | this is what I have https://paste.openstack.org/raw/bBpYnMalpXkMgsaitSKi/ | 14:42 |
| sean-k-mooney | so we may need ot factor out some compaitbality code ectra to make that work | 14:42 |
| sean-k-mooney | ya so Phase 2 — stock nvme driver binding via reserved-memory BAR0: | 14:43 |
| sean-k-mooney | Uses boot-time reserved memory (memmap=SIZE$START) to back VF BAR0 | 14:43 |
| sean-k-mooney | addresses with real RAM. | 14:43 |
| sean-k-mooney | is somehtin i wanted to avoid | 14:43 |
| sean-k-mooney | i really dint want to bring that across form the nvmevirt approch if we coudl avoid it | 14:43 |
| chandankumar | but it is needed for showing nvme on host | 14:44 |
| sean-k-mooney | not nessiarly | 14:44 |
| chandankumar | I was not getting other idea so borrowed from nvmevirt | 14:44 |
| sean-k-mooney | i belive we can allcoate a much smaller ammoutn fo memroy and then back the nvme deevice with files on disk | 14:46 |
| sean-k-mooney | chandankumar: anyway push it for review when you think its ready for that | 14:47 |
| sean-k-mooney | and we can dicuss that in more detail | 14:47 |
| chandankumar | min memmap reservation: 112 KB of reserved physical RAM, one 16 KB NVMe BAR0 region per VF | 14:47 |
| chandankumar | ok pushing | 14:48 |
| sean-k-mooney | so the pci-sim today does not preacllotate resreved memory | 14:49 |
| sean-k-mooney | it allcoates a small amount at runtime for the pci config space and minimal bars | 14:49 |
| sean-k-mooney | i expect we shoudl be able ot do the same for the nvem emulation (basiclly the contole regeiseter and command/data queues) | 14:50 |
| sean-k-mooney | but keep the data storate just as a raw file on disk | 14:50 |
| sean-k-mooney | so we are not using ram for the nvme capscity jsut the hardware emulation | 14:50 |
| sean-k-mooney | chandankumar: https://paste.opendev.org/show/bjZuArbpAxmE35uoUUoV/ was my design chat on this topic from a week or to ago | 14:53 |
| sean-k-mooney | https://paste.opendev.org/show/bjZuArbpAxmE35uoUUoV/ | 14:53 |
| sean-k-mooney | was the draft planing prom i was creating to help design this work | 14:53 |
| opendevreview | chandan kumar proposed openstack/cyborg master: [WIP]pci-sim: add NVMe 1.4 controller emulation with host visibility https://review.opendev.org/c/openstack/cyborg/+/995079 | 14:54 |
| sean-k-mooney | you will noteice that still reference https://github.com/SeanMooney/cyborg-extra/blob/master/pci-sim/fake_pci_sriov.c | 14:54 |
| sean-k-mooney | since that was over a month ago | 14:54 |
| sean-k-mooney | May 17, 12:20 AM | 14:56 |
| chandankumar | ah thank you for sharing that , I will go over that | 14:57 |
| sean-k-mooney | the paste was the final promt i had seelte on as the requriemet for the next plannign session to design the overall vf_personality feature | 14:58 |
| sean-k-mooney | but i parked that work on until the intial pahse was merged and hook up into tempest testing as a phase 2 work | 14:58 |
| chandankumar | ok | 15:05 |
| chandankumar | see ya on monday, have a good weekend! | 15:05 |
| sean-k-mooney | o/ | 15:05 |
Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!