unrelenting.technology

Another weird LLVM mystery solved!

So, I was porting LDC to FreeBSD/aarch64, wondering why global constructors (you know, the before-main() code you can make in C using an __attribute__ thingy) aren’t running… but only when the executable is linked with LLD. Turns out:

• .init_array is the only supported way to do constructors on AArch64
• and everything in general is moving towards .init_array — but the default in LLVM is still to emit .ctors
• clang has code to enable .init_array when appropriate, ldc did not
• and the reason it all worked fine with bfd and gold is that these linkers SILENTLY convert .ctors to .init_array. For PERFORMANCE REASONS.

It’s almost 2019, so using a SATA SSD as the boot drive for your main development OS is not cool anymore… and I was running out of space on this 128gb one, so I bought an NVMe drive to replace it. Yay.

Because I don’t have anything with two or more M.2 slots and I was too lazy to find/make a bootable FreeBSD USB drive, moving the system involved inserting the new drive into another machine (server) and using ZFS replication to copy the data. (And forgetting to set bootfs on the pool, of course.)

But the fun part was that my 10G network card stopped working. Moving the card into the middle slot (from the bottom one) fixed it. Reported a FreeBSD bug.

The weirdest discovery of the day though was that MSI mainboards persist the “above 4G” PCIe setting across CMOS clears. What in the actual heck. This is the setting that breaks display output on most GPUs (funnily enough, mine did display non-EFI things such as the network card’s boot prompt and the glitchy way FreeBSD displays the console when booting on UEFI with the EFI framebuffer disabled). It’s a setting you very much need to clear.

yay, Firefox will use SHM_ANON on FreeBSD now

An SPI flash chip module arrived in the mail today. I ordered that for replicating this basically. To (network) boot the Orange Pi PC without microSD.

Naturally, instead of booting Linux to flash or (?) trying flash from U-Boot itself, time was spent on making flashrom work on FreeBSD :)

Scaleway’s ARM64 VPS has been successfully depenguinated! :) Now you can run FreeBSD on four ThunderX cores, 2GB RAM and 50GB SSD for 3€/month. Awesome!

Also, in the process, I finally discovered the cause of GPT partitions sometimes disappearing on reboot. It was the size of the partition table. It’s 128 by default, but sometimes it’s not — e.g. on the FreeBSD installer memstick image, it’s 2. Creating a third partition with gpart “succeeded”, but the partition disappeared on reboot.

For 0.5 dollars per hour (or currently 0.1/hr if you reserve for 24h?) on packet.net you can get access to a dedicated dual-socket Cavium ThunderX server with 128GB RAM and a 250GB SSD. I took it for a few hours and now lang/crystal, lang/mono and some other ports work on aarch64.

Ironically, these two builds have involved long single-threaded compile processes. In the mono case, parallelism had to be disabled for C# compilation to work around a concurrency bug.

At least building things like WebKitGTK+ (to test a one line patch) and Krita felt awesome :D

BREAKING FreeBSD NEWS: AMDGPU DC works!

The 4.15 drm update apparently did not have amdgpu working yet… but I tried it anyway, since that’s just an assertion failure.

Well… it does work. Even with DC (compat.linuxkpi.dc=”1”)! Weston didn’t work at first, but commenting out its “Disable all the planes” section that was causing a kernel warning solved the problem.

So quite possibly we have Vega support?

Booted into FreeBSD, launched Epiphany (GNOME Web), noticed slowness. Opened htop: 800-1200% load in kernel! Ran dtrace to find hot kernel stacks: it’s all ZFS write threads, trying to compress. WTF? I mean, my fault for choosing gzip (I have a tiny SSD for this system, want max compression), but I expect zero I/O right now?! Found a DTrace script to show file I/O.

Turns out Epiphany is constantly writing to and reading from ~/.config/epiphany/gsb-threats.db. Yeah, it was downloading the whole Google Safe Browsing database.

I guess that’s the most privacy-friendly way to do Safe Browsing, but wow, that’s a very surprising behavior. What if I had, like, a dial-up connection or something? :D

It is well known that Enlightenment/EFL is pretty awful. But I had to build it on my FreeBSD box since I’m testing a LuaJIT update and EFL does include LuaJIT support. So while I’m at it, I decided to add Wayland support to the port.

So at least it works, but its behavior is indeed kinda funny. Running their terminal with GL acceleration (ELM_DISPLAY=wl ELM_ACCEL=opengl terminology) actually does work. Running it without acceleration (ELM_DISPLAY=wl ELM_ACCEL=no terminology) results in the surface not showing up and… this message:

DRM_IOCTL_I915_GEM_APERTURE failed: Invalid argument
Assuming 131072kB available aperture size.
May lead to reduced performance or incorrect rendering.
get chip id failed: -1 [2]
param: 4, val: 0

It’s no “SPANK! SPANK! SPANK!” but how exactly did it decide to use an Intel GPU related system call? On my machine that has a Radeon card?!?! When I asked for software rendering??

haha, Phoronix wrote about the stuff I posted on wayland-devel@. So yeah, I’m working on Rust bindings for libweston that would eventually allow me to write the best Wayland compositor ever :)

And fractional HiDPI scaling was pretty easy to add. Wayland apps look awesome. However, X11 apps are blurry now, and bypassing the scaling for Xwayland is not as easy… So I made some changes to my Ports fork to enable Wayland support in more apps. Turns out a lot of complex applications run fine — LibreOffice (!), Inkscape, MyPaint, RawTherapee, Darktable.

The Firefox Wayland support though… is not usable yet :( It looks awesome but EGL isn’t working and, even worse, the screen doesn’t refresh when it needs to — so you’re typing and letters don’t appear until you scroll or some time passes. Hopefully this will be fixed soon.