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
zsun has joined #fedora-riscv
zsun has quit [Quit: Leaving.]
esv has quit [Quit: Leaving]
zsun has joined #fedora-riscv
davidlt has joined #fedora-riscv
zsun has quit [Quit: Leaving.]
<davidlt[m]> somlo: look like a kernel bug.
<davidlt[m]> somlo: This should be disabled on Fedora as I don't plane to support CONFIG_COMPAT (at least now).
<davidlt[m]> "This is relevant on all 32-bit architectures, and 64-bit architectures as part of compat syscall handling."
<davidlt[m]> Interesting that it's `def_bool (!64BIT && 64BIT_TIME) || COMPAT`
<davidlt[m]> Sorry, I was looking at patches, not actual master.
<davidlt[m]> config COMPAT_32BIT_TIME
<davidlt[m]> default !64BIT || COMPAT
<davidlt[m]> bool "Provide system calls for 32-bit time_t"
<davidlt[m]> Checking kernel-ark...
<davidlt[m]> redhat/configs/common/generic/CONFIG_COMPAT_32BIT_TIME
<davidlt[m]> CONFIG_COMPAT_32BIT_TIME=y
<davidlt[m]> somlo: set that to `n`
<cwt[m]> Starfive kernel doesn't enable this option too.
<davidlt[m]> cwt: there is only one company that is working on COMPAT and SoC that can support 32-bit and 64-bit land :) Alibaba T-HEAD.
<davidlt[m]> I have not seen any other company to have interest into it.
<davidlt[m]> somlo: epic job debugging that 6.2 kernel :)
<davidlt[m]> somlo: once you get the other problematic CONFIG_* option I will respin Fedora 37 image.
<davidlt[m]> NEWS: After updating U-Boot and OpenSBI we can now boot on QEMU setup.
<davidlt[m]> OpenSBI master was broken too. That got fixed this night, still to be tested.
<davidlt[m]> I still might want to play a bit with instructions, but I will try to put them in place over the weekend.
<davidlt[m]> This newer image is what we will use in builders (if nothing else pops up).
<davidlt[m]> There will be a small FU740 PCIe fix too.
<davidlt[m]> FYI, there are 4 Unmatched boards in stock at https://www.crowdsupply.com/sifive/hifive-unmatched#products
<davidlt[m]> djdelorie: any plans to update libffi on F37?
<davidlt[m]> There is 3.4.4 now.
<davidlt[m]> I am running 3.4.3 on F37 for riscv64.
<davidlt[m]> I am running 3.4.3 + backport.
troglodito has quit [Ping timeout: 252 seconds]
troglodito has joined #fedora-riscv
davidlt has quit [Ping timeout: 260 seconds]
davidlt_ has joined #fedora-riscv
guerby has quit [Read error: Connection reset by peer]
guerby has joined #fedora-riscv
zsun has joined #fedora-riscv
davidlt_ has quit [Ping timeout: 248 seconds]
davidlt_ has joined #fedora-riscv
esv has joined #fedora-riscv
davidlt_ has quit [Ping timeout: 246 seconds]
<somlo> davidlt[m]: good to hear CONFIG_COMPAT_32BIT_TIME shouldn't even have been enabled -- saves us the trouble of finding/fixing that particular kernel bug :)
<somlo> the idea is, once we have a full list that cause me problems on litex/rocket, you decide which ones need to be kept enabled, and I have to debug them; the rest I'm happy to ignore :D
davidlt_ has joined #fedora-riscv
<davidlt[m]> somlo: well there is definitely a bug in the kernel, but it shouldn't enabled in Fedora :)
<davidlt[m]> somlo: just to double check. You are working on find another one, correct?
<davidlt[m]> I will wait for that too before spinning a new kernel.
<somlo> yeah, I'm still testing my way through the list of suspects; so far it's CONFIG_PRINTK_INDEX and CONFIG_COMPAT_32BIT_TIME (independently of each other); but hold off making decisions based on my input until I've made it through the whole list
<davidlt[m]> Oh, I missed CONFIG_PRINTK_INDEX
<davidlt[m]> It's another redhat/configs/common/generic/CONFIG_PRINTK_INDEX Note, that nothing overrides it at all.
<davidlt[m]> Seems pretty stable thing.
<davidlt[m]> Config wise this one looks okay.
<somlo> right, so I'll have to figure out why it causes a crash for me -- which is perfectly fine
<davidlt[m]> bisection again :)
<somlo> this time the real kind, I hope :)
<davidlt[m]> I wonder if this is just one problem or multiple problems combined in a mix.
<davidlt[m]> somlo: how many more CONFIG_* options there are for you to test?
zsun has quit [Quit: Leaving.]
jednorozec has quit [Excess Flood]
jednorozec has joined #fedora-riscv
<somlo> I'll know later this afternoon (by the time it's morning again in your time zone) -- once I get back in front of my "real" computer
<somlo> but I could always just boldly re-enable all the remaining ones and see if they work -- and only go back to one-at-a-time if the experiment fails :)
<somlo> the equivalent of "speculative execution" :D :D
davidlt_ has quit [Ping timeout: 246 seconds]
davidlt_ has joined #fedora-riscv
<djdelorie> davidlt[m]: I hadn't planned on a libffi 3.4.4 backport to F37 but if you need it I can put it on the "please consider" list
<djdelorie> davidlt[m]: the canonical way to request a backport is a bugzilla bug
<djdelorie> but I'll mention it to Carlos too
<davidlt[m]> djdelorie: I can make bugzilla ticket, but just wanted to see if it's already planned or something.
<davidlt[m]> Sometimes things from Rawhide happen to show up in older releases after quite some time.
<davidlt[m]> Not strongly needed, but would be nice to have. The most important bit is that it's already in Rawhide.
<djdelorie> understood. libffi has had a rocky path to today, we've been careful about breaking ABI, but hopefully we're almost past that
<djdelorie> gimme a bit to hear back from Carlos, a bz might not be needed
<djdelorie> Carlos is testing a backport already
<davidlt[m]> All GCC 13 builds continue to cook, 70+ hours mark passed.
davidlt_ has quit [Ping timeout: 248 seconds]
esv_ has joined #fedora-riscv
esv has quit [Ping timeout: 268 seconds]
esv_ has quit [Ping timeout: 260 seconds]
esv_ has joined #fedora-riscv