<geist>
Hmm guess the question is is all the DMA on the sifive designs cache coherent, such that flushing the cache isn't required much
<geist>
in general since the designs i've worked with at least (u54, u74) don't have any ability to mark pages cached or uncached, i've just assumed the whole thing is cache coherent
<geist>
otherwise having to do a translatino on every page as you flush it through the L2/L3 is really really expensive
motherfsck has quit [Ping timeout: 252 seconds]
BootLayer has quit [Quit: Leaving]
crossdev has joined #riscv
motherfsck has joined #riscv
<conchuod>
geist: Yeah it is coherent, which is why I don't get them having that stuff turned on in a jh7110 config
<rneese>
i think that alot of it came from the sifive unmtched and they used it by default
<rneese>
but we will be turning it off in ours
<rneese>
but we had planned to support as many boards as we can in the long run
<rneese>
but the starfive v2 is the default now
<rneese>
and the v1 for those still using it as a dev
<rneese>
and the beaglev until they canupgrade
junaid_ has joined #riscv
<drmpeg>
No, unmatched uses a mainline kernel. None of that stuff is in mainline.
<conchuod>
rneese: it comes from the jh7100
<conchuod>
rneese: Are you doing custom kernels for every board, or the same kernel everywhere?
<rneese>
well we where planning on 1 kernel for all the diff boards since its all riscv
<rneese>
sor starfive and sifive boards
<rneese>
the d1/nezha/mangopi will have thier own
<rneese>
at the min we are working to fix a fb bug
<conchuod>
Why do the d1 boards need a different kernel?
<rneese>
some how multi fb got activated and not its spews some random stuff on screen
<rneese>
that is the fully updated . but needs a few patched
<rneese>
the sounds wm9860.c needs the gpl added to the end the Licensing
<rneese>
the jh71110-viaionfive-v2.dtsi needs a patch for the mene to fix the 4 gig to read 8 gig
<conchuod>
Why does it need a patch? Won't that break the 4 GiB boards?
<rneese>
no
<conchuod>
Shouldn't the bootloader be setting that?
<rneese>
the bl needs some major work
<rneese>
and as we are not using the dtb files from the 69 img
<rneese>
your welcome to the config
<conchuod>
Nah, dw
<rneese>
if there is a furthe fix and the bl neds to be fixed
<rneese>
my board will be here tuesday/wednesday
<rneese>
and I plan to pound out things hard and figure otu all the issues
<rneese>
and report what we find
<rneese>
as we compile the kernel for our builds
<geist>
ah yeah: re the L2 stuff the reason I'm thiking about it is i'm worknig on another OS and will need to worry about these coherency issues
<geist>
was hoping the whole 'cycle through L2/L3 by physical tag' wasnt going to be a thing here
<rneese>
ah conchuod we do plan to support the vf2/unmatched/nezha/mangopi/star64
<rneese>
but I am waiting for a star64 to be released to test with it also
<conchuod>
I'm just a curious bystander, support what your users want :)
<rneese>
my mangopi is on backorder
<geist>
oh hey my vf2 just shipped from ameridroid
<rneese>
cool
<rneese>
yeah they are starting to ship
<rneese>
i have a 120 gig nvme ready and a 64 gig emmc and8 gig emmc
<rneese>
and to find out if a single header ver works or if it has tobe dual header version
<rneese>
they dont have docs on the emmc
pedja has quit [Quit: Leaving]
<rneese>
I have single header emmc
larsen has joined #riscv
larsen has quit [Read error: Connection reset by peer]
<rneese>
who flashed thier board and it stopped working
<rneese>
thereis a bl recover
<rneese>
in the qst
larsen has joined #riscv
bryanz has joined #riscv
<larsen>
I had just flashed my VisionFive 2 board yesterday, first to Debian's image-55 to update the firmware then to the current image-69. Would I have to reflash the OS each time to update or would Debian allow upgrading through apt?
<mps>
larsen: yes, it should allow upgrade
crossdev has quit [Remote host closed the connection]
junaid_ has quit [Remote host closed the connection]
rurtty has quit [Quit: Leaving]
<jrtc27>
conchuod: their uboot has a working command that patches the device tree to have the right memory range defined
<jrtc27>
unfortunately it doesn't get called during the default boot sequence...
<conchuod>
I wouldn't know, I threw out their u-boot env ;)