| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-cyborg - 2026-07-29 | |||
| 15:19:14 | opendevreview | Joan Gilabert proposed openstack/cyborg master: Refactor GPU driver to use MdevBusManager https://review.opendev.org/c/openstack/cyborg/+/997052 | |
| 15:19:14 | opendevreview | Joan Gilabert proposed openstack/cyborg master: Add generic mdev driver documentation https://review.opendev.org/c/openstack/cyborg/+/997053 | |
| 15:19:16 | opendevreview | Joan Gilabert proposed openstack/cyborg master: Add mdev support to pci-sim kernel module https://review.opendev.org/c/openstack/cyborg/+/997054 | |
| 15:19:16 | opendevreview | Joan Gilabert proposed openstack/cyborg master: Add mdev support to pci-sim DevStack plugin https://review.opendev.org/c/openstack/cyborg/+/997055 | |
| 15:19:18 | opendevreview | Joan Gilabert proposed openstack/cyborg master: [WIP] Configure mdev driver lifecycle tests https://review.opendev.org/c/openstack/cyborg/+/997364 | |
| 15:32:53 | jgilaber | chandankumar, IIUC your patch https://review.opendev.org/c/openstack/cyborg/+/999125 is building an abstraction for the persionality that my patch introduces, correct? | |
| 15:38:00 | chandankumar | jgilaber: yes correct | |
| 15:38:26 | chandankumar | let me fix the chain to make it clear | |
| 16:07:03 | sean-k-mooney | ya so i had planed ot work on this all next cycle, but because we had to pivot on the testign stragy for both the nvme and mdev driver we now hve 2.5 competing desings for htis although we are converging | |
| 16:07:49 | sean-k-mooney | jgilaber: you have start on the sysfs part and chandan has started on the abstraction layer which is good | |
| 16:08:23 | sean-k-mooney | the uart/serial was intened to be just one of many device the similator supproted | |
| 16:09:22 | sean-k-mooney | so seperatign that out into pci-sim/fake_pci_sriov_uart.c is inlien with that vission | |
| 16:11:20 | sean-k-mooney | althought he same feedback still aplies i woudl prefer to avoid a defautl personality | |
| 16:11:48 | sean-k-mooney | or default to a sential that you have to update before you can allocate the vfs | |
| 16:12:39 | sean-k-mooney | the uart was my strating pont to prove the concept not the end goal so we shoudl not make it a special default type | |
| 16:13:58 | jgilaber | I see, it took me a bit, just wanted to make sure I got it right, reading C is a bit harder than python :) | |
| 16:14:24 | sean-k-mooney | i have only skimed it so im not sure it has the right shape yet | |
| 16:14:37 | jgilaber | in my patch I moved the default to a 'unset', so that should carry into chandas patch once he rebases on top of mine | |
| 16:14:52 | sean-k-mooney | yep | |
| 16:15:42 | sean-k-mooney | i had orgianlly planend ot spend a month or so workign on this my self for what it worth after FF | |
| 16:16:28 | sean-k-mooney | i did the iniall high levle planign fo rhow tot evolve the framework but defered it to keep scope small for v1 | |
| 16:16:48 | sean-k-mooney | so we havenet really had time to talk about that or spec it out | |
| 16:17:07 | sean-k-mooney | if we had not change our testing plan | |
| 16:17:24 | sean-k-mooney | this was going to be somethign i teed up for the next ptg | |
| 16:18:16 | jgilaber | yes a PTG session would have been quite useful | |
| 16:18:44 | jgilaber | I realize there are a few things that I assumed worked in some way and was wrong | |
| 16:19:01 | sean-k-mooney | i think we can actully just do some one off desing session on this as open to everyone calls | |
| 16:19:43 | jgilaber | that might be helpful, yes | |
| 16:19:44 | sean-k-mooney | but what also going to impornat is making sure we docudment the desgin and rational in the contibutor docs | |
| 16:22:43 | sean-k-mooney | one thing im also thinking about | |
| 16:23:00 | sean-k-mooney | is the bash scritp testing i initally didn was jsut to have some testing | |
| 16:23:40 | sean-k-mooney | i want to eventually replace that with pyton based test in a new validation or other named test suite | |
| 16:24:17 | sean-k-mooney | pubally still under the pci-sim folder | |
| 16:25:23 | sean-k-mooney | but long term i want to move awasy form jsut bash script to something more maintianable | |
| 16:26:05 | sean-k-mooney | i havent really tought oo much abotu that but im remineid fo it when i start seeign lots of similar code in one of script files | |
| 16:26:47 | sean-k-mooney | so even if we keep it in bash woudl liek to move more toward devstack style bash with shared resuabel fucntions but ya i have nto really said that outload before | |
| 16:28:16 | sean-k-mooney | im glad you and chandan have start contibuting ot extending the pci-sim i just was not expecting ye too this cycle so i tought i woudl have time to do that cleanup before then | |
| 16:28:30 | jgilaber | yes, most of the code in those scripts is duplicated | |
| 16:29:06 | jgilaber | it's great to have it to test, but I found it a lot harder to work on | |
| 16:29:28 | jgilaber | AI has been really useful for this | |
| 16:29:56 | sean-k-mooney | yep, it will get eaiser once alot of the strucre is put in palce and more of the context on how it was intedned to work and be extened is wirten down | |
| 16:30:15 | jgilaber | although I'm less confident on its output since I have a harder time reviewing | |
| 16:30:17 | sean-k-mooney | but i agree ai has been makign building this type of test infra much more practical | |
| 16:30:28 | jgilaber | at least it seems functional | |
| 16:31:51 | sean-k-mooney | im undecieded if i still want to eventually look at upstreamign this to the linux kernel | |
| 16:32:05 | sean-k-mooney | im sure they woudl have lots of clear and valudable feedback | |
| 16:32:36 | jgilaber | it would probably quite a bit of work as well I guess | |
| 16:32:58 | jgilaber | depending on how we've diverged from the kernel style | |
| 16:33:07 | sean-k-mooney | ya its nto a discieon to make lightly or one we have to make in the near term | |
| 16:34:07 | sean-k-mooney | for me i think the cost of upstream is not warrented until we have something stable that we have provien to work for the usecase we intent to test | |
| 16:34:59 | sean-k-mooney | if in a cycle or two we find we do not need to extend it much and its stable we could start that conversation | |
| 16:36:13 | sean-k-mooney | its much easier to start a conversation by "hey we built this thing that we find useful and would like to share it to see if its useful to the kernel comunity too" | |
| 16:38:07 | jgilaber | I imagine it would be useful for many people | |
| 16:38:18 | jgilaber | since there does not seem to be many solutions out there | |
| 16:38:30 | jgilaber | or very limited ones like the mtty module | |
| 17:23:53 | opendevreview | chandan kumar proposed openstack/cyborg master: pci-sim: introduce personality ops-table dispatch for VF device types https://review.opendev.org/c/openstack/cyborg/+/999125 | |
| 17:35:54 | opendevreview | Joan Gilabert proposed openstack/cyborg master: Add mdev support to pci-sim kernel module https://review.opendev.org/c/openstack/cyborg/+/997054 | |
| 17:35:55 | opendevreview | Joan Gilabert proposed openstack/cyborg master: Add mdev support to pci-sim DevStack plugin https://review.opendev.org/c/openstack/cyborg/+/997055 | |
| 17:35:55 | opendevreview | Joan Gilabert proposed openstack/cyborg master: [WIP] Configure mdev driver lifecycle tests https://review.opendev.org/c/openstack/cyborg/+/997364 | |
| 18:56:21 | opendevreview | melanie witt proposed openstack/cyborg master: Enable port_device_profile ML2 extension driver for NIC drivers https://review.opendev.org/c/openstack/cyborg/+/998055 | |