dgilmore changed the topic of #fedora-riscv to: Fedora on RISC-V https://fedoraproject.org/wiki/Architectures/RISC-V || Logs: https://libera.irclog.whitequark.org/fedora-riscv || Alt Arch discussions are welcome in #fedora-alt-arches
* BryceII is in a daze,..
masami has quit [Remote host closed the connection]
masami has joined #fedora-riscv
masami has quit [Remote host closed the connection]
masami has joined #fedora-riscv
jimwilson_ has joined #fedora-riscv
jimwilson has quit [*.net *.split]
jbowen has joined #fedora-riscv
jbowen has quit [Client Quit]
jbowen has joined #fedora-riscv
jbowen has quit [Client Quit]
jbowen has joined #fedora-riscv
jbowen has quit [Client Quit]
jbowen has joined #fedora-riscv
ChanServ has quit [shutting down]
ChanServ has joined #fedora-riscv
davidlt has joined #fedora-riscv
bkircher has joined #fedora-riscv
jcajka has joined #fedora-riscv
<rwmjones> morning
<rwmjones> davidlt: so we don't have rust yet? I was trying to compile rust-hyperfine yesterday but got lots of missing deps
<davidlt> rwmjones, we do
<davidlt> rwmjones, example of missing pacakge?
<rwmjones> one sec
<rwmjones> $ sudo dnf builddep rust-hyperfine-1.11.0-3.fc34.src.rpm
<rwmjones> Problem 1: nothing provides requested (crate(colored/default) >= 2.0.0 with crate(colored/default) < 3.0.0~)
<rwmjones> Problem 2: nothing provides requested (crate(indicatif/default) >= 0.15.0 with crate(indicatif/default) < 0.16.0~)
<rwmjones> Problem 3: nothing provides requested (crate(rust_decimal/default) >= 1.8.0 with crate(rust_decimal/default) < 2.0.0~)
<rwmjones> actually looking again it seems like a versioning problem
<rwmjones> also yesterday I had a lot more errors than that, not sure what changed
<davidlt> rwmjones, these seems to be from modules
<davidlt> we don't have modules
<davidlt> we do have: rust-colored-1.9.3-1.fc33
<davidlt> but "default" suggest this is default some kind of module ?
<davidlt> yeah
<davidlt> I see rust-colored has module builds in the Koji
<rwmjones> the whole spec for this package is weird, it has not BuildRequires at all
<rwmjones> no
<davidlt> Rust works slightly different
<davidlt> rust-packaging (the only BR) is responsible for generating true BRs
<davidlt> ha
<davidlt> interesting
<davidlt> ah, this is from Rawhide
<davidlt> We still didn't start Rawhide build
<rwmjones> oh should I try the f33 build? let me try it
<davidlt> F33 Rust tooling seems to generate slightly different Provides: and the version is slightly older, 1.9.3
<davidlt> Sorry, I haven't started F35 yet
<rwmjones> only one error in f33:
<rwmjones> Problem: nothing provides requested (crate(indicatif/default) >= 0.15.0 with crate(indicatif/default) < 0.16.0)
* davidlt checking...
<rwmjones> oh I see the package in koji is too old
<rwmjones> so probably I can build that first and get it to work
<davidlt> interesting
<davidlt> did you try F33 RPM?
<davidlt> My script says everything is up to date
<davidlt> I checked Koji, and required version (0.15 or higher) never made the final F33
<davidlt> is completely untagged
<rwmjones> the rust-indicatif package in fedora/risc-v is too old, but I'm trying to build the newer version from primary koji now
<davidlt> was part of f33-rebuild, but only somehow got into f34
<rwmjones> weird, maybe there was a problem with it and someone untagged it?
<davidlt> rwmjones, almost all Rust package are noarch so just importing is enough :)
<davidlt> rwmjones, most likely
<rwmjones> so it is ..
<davidlt> rwmjones, but it could be that other packages were built against it in a short amount of thime and those landed
<davidlt> I have seen this before
<davidlt> there the requires package to build is not part of the release
<davidlt> where
* davidlt brain is not working yet
<rwmjones> those noarch packages install fine
<rwmjones> let's see if I can build hyperfine now ..
<rwmjones> it's building ..
davidlt_ has joined #fedora-riscv
davidlt has quit [Ping timeout: 268 seconds]
<rwmjones> hyperfine compiled ok
cmuellner has quit [Ping timeout: 268 seconds]
cmuellner has joined #fedora-riscv
cmuellner has quit [Ping timeout: 252 seconds]
cmuellner has joined #fedora-riscv
* rwmjones is building the other hifive unmatched now ...
<rwmjones> my plan is to temporarily use the ubuntu bootloader and redirect it to boot Fedora from the NVMe drive
cmuellner has quit [Ping timeout: 252 seconds]
<somlo> davidlt: another archaeology question: if one had a f31-era rawhide fedora-rv64gc VM, with riscv-specific URLs in /etc/yum.repos.d pointed at "http://fedora-riscv.tranquility.se/..."; (and kojipkgs.fedoraproject.org/repos/rawhide/latest/riscv64 in /etc/mock/fedora-rawhide-riscv64) -- how would one re-point those URLs to something age-appropriate for that version?
<rwmjones> there may be other links under kojipkgs.fedoraproject.org/repos but unlikely that we kept anything from f31
<rwmjones> that was probably before we switched the build system over from the .se site
<somlo> rwmjones: thanks!
<rwmjones> fyi jive got rebooted because I had to plug in a machine that was obscured by the plug
cmuellner has joined #fedora-riscv
<BryceII> yeah
<BryceII> you basically provide a disk as a internet capable drive
<BryceII> oops
jbowen has quit [Quit: leaving]
<rwmjones> BryceII: ?
<BryceII> rwmjones: oh collegue of mine in another irc window,.. we're discussing diskless pxe booting and raid1 ramdisk scenarios for teh riscv
<rwmjones> hehe
<rwmjones> I'm using nbdkit-tmpdisk-plugin as a remote tmpfs
<BryceII> (that said I'm having more fun with rpm crapping out on rpms with cap_foo (apability issue though I dunno what exactly
<BryceII> error: unpacking of archive failed on file /usr/bin/newgidmap;60d088f2: cpio: cap_set_file failed - No such file or directory
* BryceII scratchs his head,.. seems a common enough problem just not nailed since it's rpm's own cpio version that spits it out
<BryceII> [root@risc-v fedora-33-riscv64]# getfattr -n security.capability /etc/fstab
<BryceII> /etc/fstab: security.capability: Operation not supported
* BryceII scratchs his head,.. if this is selinux I'n going to fetch the 3-in-1 oil and a book of matches...
<rwmjones> my two boards have mac addresses that differ by +2
<BryceII> I'll be a lot happier when we can write a FSBL/BBL into the SPI-NOR I have a habit of breaking SD cards or loosing them under heavy furniture hiding a hoarde of dust bunnies
ahs3 has joined #fedora-riscv
ahs3|afk has joined #fedora-riscv
ahs3 has quit [Quit: Ex-Chat]
ahs3|afk is now known as ahs3
<rwmjones> so intel are buying sifive ..?
<BryceII> that was the rumour
<BryceII> for a few billion in small change
* BryceII wonders if anyone had stock options,...
<sorear> did something happen since last week
<sorear> so, no
somlo has quit [Remote host closed the connection]
somlo has joined #fedora-riscv
<BryceII> when did that originally get submitted? a year ago?
<BryceII> ugh wc sorry
bkircher has quit [Ping timeout: 244 seconds]
davidlt_ has quit [Ping timeout: 265 seconds]
<rwmjones> $ nbdkit --filter=partition --filter=xz file Fedora-Developer-Rawhide-20210421.n.0-sda.raw.xz partition=2 --run 'qemu-img convert "$uri" -O qcow2 -c Fedora-Developer-Rawhide-20210421.n.0-root.qcow2 -p'
bkircher has joined #fedora-riscv
bkircher has quit [Ping timeout: 258 seconds]
bkeys45 has joined #fedora-riscv
<bkeys45> No updates about Fedora on the Unmatched I take it?
<BryceII> define update? (kernel, rpms, fc33 -> fc34 -> fc35, nvidia drivers for video...)
* BryceII was going to weigh in on EFI/uboot/ SPI-NOR but there are better folk working on it
<bkeys45> How about booting into Fedora in the first place?
<BryceII> sure,..
<BryceII> (unmatched)
<BryceII> [root@risc-v linux]# df -h
<BryceII> Filesystem Size Used Avail Use% Mounted on
<BryceII> /dev/root 916G 19G 851G 3% /
<bkeys45> How did you manage to get Fedora on the unmatched?
<BryceII> where the NVME is a 1TB 970
<bkeys45> Do you have a bootable image? Or did you run the ubuntu kernel with a fedora userland?
<BryceII> chicken blood and voodoo!
<BryceII> pfft nah, it's simpler than that,.. the unmached came with a SD card with the sifive SD boot
<bkeys45> Mine just had the freedom SDK, something like that
<BryceII> yeap,..
<BryceII> ok well
<BryceII> the easy way is boot the unmatched's SD card,.. then partition up the NVME as you see fit,.. I expect odd things sI I want with this layout,.. sec lemme get my notes
<BryceII> parted --script /dev/nvme0n1 \
<BryceII> mkpart primary 1MiB 2MiB \
<BryceII> mkpart primary fat16 6MiB 518MiB \
<BryceII> mkpart primary 2MiB 6MiB \
<BryceII> mklabel gpt \
<BryceII> mkpart primary ext4 518MiB 100% \
<BryceII> print
<BryceII> you can then grab yourself the fedora 33 rootfs and copy it over PREFERABLY from another network machine ,... the SD card speed is *****d
<BryceII> mkfs.msdos /dev/nvme0n1p3
<BryceII> mkfs.ext4 /dev/nvme0n1p4
<BryceII> mount /dev/nvme0n1p4 /mnt
<BryceII> if you have the fedora rootfs mount (kpartx -av foo.img) then you can cd to that mounted area and use something like
<BryceII> ssh -l root 192.168.1.100 'tar -C /mnt --one-file-system --selinux --acls --xattrs -cf - . ' | dd status=progress bs=$((512 * 16)) | tar -C /mnt -xf-
<BryceII> (the dd is just to let you know there is movement happeneing)
<bkeys45> Is this nicely written up somewhere?
<BryceII> after that you just edit the fstab and the SD cards /boot/extlinux/extlinux.conf file to point roo to the nmve partition,.. oh and fix the UUID's (blkid will show them) and lastly copy over the sifive's /lib/modules/5.* directory to the nvme or you'll loose fun stuff like bluetoothand wireless)
bkeys has joined #fedora-riscv
<BryceII> it will be,.. I was writing it up for my college since he's more toolchain orientated
<BryceII> <-- more a system integration specialist
<bkeys> BryceII where did you get the fedora riscv image from?
<bkeys> The qemu image they distribute in the wiki?
<BryceII> the only fly in the ointment I've hit so far is that the sifive kernel doesn;t seem to have Selinux/capabilities enabled which causes,.. interesting times
<bkeys> Oh no
<BryceII> yeah the one off the wicki,.. you'll probably be fine with just the rootfs,.. the wic I couldn;t get that to work on the unmatched but the SD card hates me anyway
bkeys45 has quit [Ping timeout: 246 seconds]
<bkeys> Okay, well I gotta download the images and all that so I will probably start on this in an hour or two
<bkeys> I would really like to have Fedora on the unmatched
<rwmjones> bkeys: hey
<BryceII> I have a badly written email if you want it as guidence
<rwmjones> I have it working now actually
<bkeys> BryceII That would be good
<bkeys> rwmjones Do you have a bootable image?
<BryceII> I was trying to figure out getting pxe/iscsi going to make it all diskless
<rwmjones> nope, I'm using the ubuntu uboot + kernel to boot the fedora userspace :-/
<BryceII> ok who wahts a crappy email?
<rwmjones> I wouldn't mind seeing that
<bkeys> rwmjones I have attempted something like that but failed; do you have a script or something written up?
<BryceII> You must log in with services to message this user ... huh?
<rwmjones> (rjones@redhat.com)
<BryceII> me old stomping ground!
<rwmjones> bkeys: so what I did basically was to unpack the Fedora second partition onto a partition on the NVMe drive
<rwmjones> then edit /boot/extlinux/extlinux.conf on the ubuntu filesystem to point to that, ie. setting root=LABEL=<label-of-my-fedora-root>
<rwmjones> I also had to resize2fs the fedora partition
<rwmjones> and edit /etc/fstab in *Fedora* root to remove reference to /boot
<rwmjones> my handy nbdkit command for extracting the second partition was:
<BryceII> blkid is what you ned there to find the UUID's in effect
<rwmjones> $ nbdkit --filter=partition --filter=xz file Fedora-Developer-Rawhide-20210421.n.0-sda.raw.xz partition=2 --run 'qemu-img convert "$uri" -O qcow2 -c Fedora-Developer-Rawhide-20210421.n.0-root.qcow2 -p'
<rwmjones> it's a hack though, I'm sure BryceII's method is better, especially since the ubuntu kernel doesn't have selinux
<BryceII> mine is more or less the same actually though I wrote up a more agile extlinux.conf
<bkeys> I am just spinning my wheels downloading images right now
<bkeys> rwmjones How far are you from having a bootable image? Or is that not what you are working on?
<rwmjones> I'm not working on that at the moment
<bkeys> It should hopefully be a moot point once they get UEFI on the unmatched
<rwmjones> davidlt may be working on this, he's got a pile of unmatched machines
<rwmjones> interesting that -march=native doesn't work
<BryceII> I put in for an unmatched way back at the end of febuary,.. I was starting to think it was a scam the numbver of times it kept getting put back
<bkeys> Is this the most up to date image out there?
<rwmjones> bkeys: no the latest is:
<bkeys> I got one about 2 weeks after I ordered it, but that was at the beginning of this month
<BryceII> I'm certain there must have been later
<rwmjones> yeah I think davidlt "curates" images and puts then there, but the fedora.riscv.rocks link is the latest
<rwmjones> Fedora-Developer-Rawhide-20210421.n.0-sda.raw.xz I'm using works fine
<BryceII> 800 xzcat Fedora-Developer-Rawhide-20210421.n.0-sda.raw.xz > /tmp/risc-v.img
<BryceII> thats the one I used
<BryceII> definitly need a kernel srpm ,.. sifive would have us all building ostrees and yocto manglements
<bkeys> What we need is UEFI
* BryceII bites his lip,.. <conflicted>
<rwmjones> yeah uefi would solve a lot of this
<rwmjones> I think it exists at least in theory
<BryceII> I've read the pdf's for uefi on risc by the WD fokl
<rwmjones> thanks for that email, I guess davidlt will know what patches we're missing
<BryceII> the sifive guys have about 25? sec
<BryceII> lynx --dump https://github.com/sifive/meta-sifive/tree/2021.05/recipes-kernel/linux/files | grep "\.patch" | cut -f 2- -d '.' | xargs -I{} curl -L -O {}
<BryceII> not tidy but works,..
<rwmjones> so these boards are reasonably fast now I'm using NVMe for everything, not going to break any speed records though
<rwmjones> like the early aarch64 boards
<BryceII> definitly not I was a lttle dissapointed at the nvme speed too but anything is better than the SD card.
<BryceII> I also find that the serial console hets lossy when anything marginbally intesive is happeneing on the boartd
<rwmjones> I wonder if there's some performance that can be gained with nvme, it's surely not running at pcie full speed
<bkeys> That it is not
<BryceII> it's probably running at x1 instad of x4?
<rwmjones> those cases are ugly but very cheap
<BryceII> I should be getting at least somewhere in the region of 600MB/s but it's forking out about 89MB/s
<BryceII> I got a really cheap itx (35 quid?) which came with the PSU.. annoyingly, no reset switch
<BryceII> 141.79 MB/sec.. for a 970 thats poor
bkeys has quit [Ping timeout: 246 seconds]
bkeys has joined #fedora-riscv
jcajka has quit [Quit: Leaving]
<BryceII> LnkCap: Port #0, Speed 8GT/s, Width x4, ASPM L1, Exit Latency L1 <64us
bkeys has quit [Quit: Client closed]
ahs3 is now known as ahs3|afk
ahs3|afk is now known as ahs3