11:35 <Guest51> Where can I find the latest Fedora image for the SiFive Unmatched?

11:42 <davidlt[m]> Guest51: https://dl.fedoraproject.org/pub/alt/risc-v/disk_images/Fedora-Developer-Rawhide-20220705.n.0.SiFive.Unmatched/

11:59 <defolos> davidlt: does that one work for the visionfive as well?

12:03 <davidlt[m]> No

12:03 <defolos> ah, bummer

12:04 <davidlt[m]> I am not sure that JH7100 even got upstream support

12:04 <davidlt[m]> I think Ubuntu has the patches in, but not sure about the rest

12:06 <davidlt[m]> 13 hours ago this was posted: "[PATCH 0/3] Enable initial support for StarFive VisionFive V1 SBC"

12:06 <davidlt[m]> But I wonder what "boots" means

12:07 <davidlt[m]> Conor D. (Microchip) even commented: "I know that most jn7100 stuff is not really wanted upstream, but I'd

12:07 <davidlt[m]> say that the minimal vision5 dts is an exception to that, so with the

12:07 <davidlt[m]> one comment on patch 1 resolved:"

12:08 <davidlt[m]> I think the upstream suppose to boot (at least on BeagleV) but doesn't do anything useful.

12:08 <rwmjones> why would it not be wanted upstream?

12:08 <davidlt[m]> Anyways this was replaced by JH7110 aka VisionFive V2

12:09 <davidlt[m]> Because things are not connected in cache coherent way

12:09 <rwmjones> interesting - does that make it impossible to run normal oses?

12:09 <davidlt[m]> They connected peripherals to a wrong port in their SoC design

12:09 <rwmjones> oops

12:09 <davidlt[m]> It runs, but they have a bunch of l2 cache management bits in various drivers IIRC

12:10 <rwmjones> I have the first samsung big.LITTLE chip that infamously has cache coherency problems

12:10 <davidlt[m]> There was a patch or a thread how to deal with DMA stuff IIRC

12:10 <davidlt[m]> They also have a HW bug in L2 too

12:10 <rwmjones> so are they going to fix it or just yolo it?

12:10 <rwmjones> btw I backed the visionfive 2 yesterday

12:11 <davidlt[m]> So the fix is JH7110 aka Vision V2

12:11 <davidlt[m]> JH7100 was a test chip, let's be honest :)

12:11 <rwmjones> ah ok

12:11 <davidlt[m]> It never hard a proper full wafer production too

12:11 <rwmjones> that makes more sense then

12:12 <davidlt[m]> This is also the reason why there is so little of it around :)

12:13 <davidlt[m]> VisionFive V2 unboxing: https://www.youtube.com/watch?v=bezXFWWmsuQ

12:18 <rwmjones> it's interesting that it has 2 ethernet ports

12:18 <davidlt[m]> Ah, I see Bruce Hoult already got the information bout it.

12:18 <davidlt[m]> So JH7110 contains Zba and Zbb (B) extensions

12:18 <rwmjones> can't think of another SBC with 2 ethernet, although I have some ancient Arm board that was designed for routing that has 3

12:19 <davidlt[m]> Plenty of SoCs these days support 2 Ethernets

12:19 <rwmjones> sure

12:19 <rwmjones> I guess the sockets cost a few cents

12:19 <rwmjones> just a bit unusual

12:19 <davidlt[m]> PINE64 will sells boards with 1 and 2 Ethernets (it will be a few dollars cheaper)

12:20 <davidlt[m]> SiFive and Intel should hurry up and release their board :)

13:02 <Guest51>  @davidit Thanks for the link!

13:22 <neil> davidlt[m]: following up from yesterday. Definitely possible with Rocky. For altarches i'm less concerned about breaking the rules since they wouldn't exist otherwise. Aiming for 10 sounds like a great plan

13:23 <davidlt[m]> neil: could you share your email?

13:23 <neil> neil@shrug.pw or neil@rockylinux.org (personal vs rocky)

13:24 <neil> my fedora account is the former