Tobias Nygren

Tweets about stuff related to @NetBSD, @pkgsrc and magical adventures of using devices in ways the manufacturer did not intend.

[Tobias Nygren @tnn2]

@astr0baby I wonder if it would "just work" to desolder the 64 meg GDDR3 chip and plop a 128 one in. It looks like some versions of the Xbox 360 may have shipped with theoretically compatible Samsung memory chips.

[Tobias Nygren @tnn2]

@herdware @AnnatarTheMaia The biggest problem is GCC got desupported by upstream before version 5 and the rest of the world moved on to require C11, C++17, cmake, rust and so on. It can be fixed for a subset of packages, but is a losing battle. (That said, of course you can check out a historical branch.)

[Tobias Nygren @tnn2]

@MakeAugusta Must admit I have never seen a Centronics drywall socket. I would guess some sort of proprietary PBX connector.

[Tobias Nygren @tnn2]

@jmcwhatever Will try to revert my local hacks ASAP. (Wow, I spent many hours hunting for that bug!). Another minor issue is the system panics in arm_efirt_settime(). I know that it is an Ampere-specific EFI mapping problem and that Linux has a workaround.

[Tobias Nygren @tnn2]

Pointed MSI-X vectors to kmem and am servicing PCI interrupts from kernel threads in a busy loop. 🙈 But GPT is detected on NVMe and no more reset! Now to figure out why plcom_acpi doesn't attach. If enough other issues are sorted out maybe the GICv3 problem will go away?

[Tobias Nygren @tnn2]

@jmmv There is a userland interface. Basically MAKEDEV spi0, open(2) it and issue the ioctl(2) defined in <dev/spi/spi_io.h>

[Tobias Nygren @tnn2]

@bexcran I got this one, can confirm it fits on the aadk at least

[Tobias Nygren @tnn2]

@jonmasters Hardware for developers should be interpreted as hardware for operating system developers. If devs have time to complain on forums they have time to hack on drivers. It doesn't become a turnkey solution until someone puts in the hours and upstreams code.

[Tobias Nygren @tnn2]

@herdware @PalleLyckegaard Mine is somewhat broken (one of the PCIe bridges is bad I think) which panics the Solaris installer. But it runs OpenBSD pretty well.

[Tobias Nygren @tnn2]

@hennichodernich Cool! What part numbers are U2 and U3 if you don't mind sharing?

[Tobias Nygren @tnn2]

@bexcran This makes updating the firmware much less scary. But I hope the user (with vendor's permission) can post the correct pinout table as well so we don't have to decipher it from the picture.

[Tobias Nygren @tnn2]

@NordicPGDay @redpilllinpro UPDATE social_media_templates SET year = year + 1;

[Tobias Nygren @tnn2]

radeondrmkms can map the framebuffer and do wscons. But has the same issue as the LX2K where the ROM BAR is not enabled by EDK2 and all bios extraction methods thus fail. Dumped the bios on an amd64 box and compiled it into the kernel. md kludge for mapping the bios needed.

[Tobias Nygren @tnn2]

Reached multi-user mode with serial, NVMe onboard Ethernet and almost-working wsdisplay. No USB yet due to buggy ACPI tables.

[Tobias Nygren @tnn2]

@JoeSpeeds Meshify 2 XL. Already ordered but logistics lost it somewhere. Got a message they found it today so hopefully it will show up.

[Tobias Nygren @tnn2]

Parts for new workstation build arrived. Still need to pick up a case and PSU after work.

[Tobias Nygren @tnn2]

@WhatAintInside @jonmasters I think it comes down to if the platform has active maintainers. m68k still has a very active community (those Amiga weirdos). OTOH nobody is enthusiastic about Itanium.

[Tobias Nygren @tnn2]

@jmcwhatever Good advice, thanks. Unfortunately I don't think we have an smmu driver yet. But this gives me some ideas where to start looking.

[Tobias Nygren @tnn2]

Managed work around the _CRS issue for now. But taking MSI-X interrupts causes a hard reset and MSI interrupts just seem to not get delivered. MADT and/or GICv3 ITS bug?

[Tobias Nygren @tnn2]

@bexcran @olofj Yep, they included it in the box (shipping label created Oct 23rd)

[Tobias Nygren @tnn2]

@olofj It seems the fan can at least be regulated with the BFAN_1 adapter since it varies speed during POST. But the firmware does have some annoying PCIe resource assignment issues I need to work around first.