💾 Archived View for ecs.d2evs.net › feed.xml captured on 2021-11-30 at 19:37:34.

-=-=-=-=-=-=-

<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
	<channel>
		<title>ecs's blog</title>
		<description>I do stuff, most of which does not involve writing blog posts.</description>
		<link>gemini://ecs.d2evs.net</link>
		<atom:link href="gemini://ecs.d2evs.net/feed.xml" rel="self" type="application/rss+xml" />
<item>
<title>Apple sucks and I'm tired of it</title>
<description># Apple sucks and I'm tired of it

All Apple products that I've ever used are incredibly shoddy. Let's break down why.

## General issues

Apple, in general, views its products as, well, products. It attempts to control what its users can and cannot do with their computers, waging constant warfare on the concept of general-purpose computing.

## Software

For obvious reasons, I haven't purposefully come within a ten-foot radius of Apple software since I've become tech-literate, so I'm positive that there are more issues with Apple software. Patches are welcome!

macOS's POSIX implementation is extraordinarily buggy. All the good parts are stolen from BSD and Mach. Contrary to popular belief, the UI is terrible - it doesn't give users the tools to become more knowledgeable. Lack of access to the source code makes users reliant on Apple to provide support. Proprietary APIs abound, locking developers into Apple's platform. macOS is increasingly similar to iOS, I would be surprised if it lasted another five years without becoming a thinly rebranded iOS knockoff.

There's so much wrong with iOS that I can't even begin to fathom how to express it all. The UI is the literal worst interface I could possibly imagine. The amount of lock-in and lockdown should be a criminal offence punishable by dismemberment of the executives responsible. The advertising and tracking APIs are violations of the trust that users put in the software vendor to look out for them. The propaganda about Apple being committed to privacy is a disgrace to any vendor even remotely privacy-conscious. The walled garden of apps is a crude mockery of any semblance of sane package management.

I don't have any first-hand experience with watchOS or tvOS, but from what I've seen they exhibit all the problems that other Apple software has - putting Apple's interests before the user's and having dubious technical merit.

## Firmware

OpenFirmware isn't actually open, and USB booting on the only PowerBook I've used caused the system to hang indefinitely. EFI is a pain to deal with on MacBooks. I haven't yet had the "joy" of dealing with ARM MacBooks, but I bet the boot story won't be pretty.

iOS bootup is a nightmare hellscape, the only good thing is that at least there's an exploit which allows arbitrary code execution early in the boot sequence. The fact that this requires an exploit is exactly the issue.

As of the time of publication, there are no known watchOS/tvOS bootloader exploits, therefore these devices are worth less than an equivalently sized pile of toxic gas.

## Hardware

I use Apple hardware (currently a MacBook 3,1, previously a MacBook 5,1, both running Alpine Linux) as my daily driver, so I have the most first-hand experience with the hardware side of things (lucky me!).

Update 2020-01-06: I have replaced my MacBooks with an Optiplex 9010 and a ThinkPad T420, both corebooted. Best decision I've made in quite a while.

The MacBook 3,1 runs extraordinarily loudly, the MacBook 5,1 runs extraordinarily hot. Regardless of what it tries, Apple seems to be incapable of coping with the fact that electronics produce heat. The (proprietary, yay) chargers break (3,1's charger is extraordinarily frayed, had to buy a replacement for the PowerBook G4 charger). All Apple products are extraordinarily overpriced (or extraordinarily slow, depending on how you look at it). They somehow manage to make the laptop look (subjectively) uglier than a Thinkpad in exchange for terrible thermals. The 5,1 recently straight up died. Locks up completely somewhere inside the initramfs. I'd try to debug that, but it's practically impossible to take a MacBook apart.

Update 2020-01-06 (though the events in question happened a while ago): The issue with the 5,1 was a dead HDD. It worked mostly fine with an SSD, though the GPU is ridiculously terrible - it's actually /faster/ with LIBGL_ALWAYS_SOFTWARE set globally, and without that it randomly locks up.

Remember the iPhone 4 "you're holding it wrong" fiasco? 'Nuff said.

In conclusion, all Apple executives deserve to be drawn, quartered, resurrected, have lemon juice poured in their eyes, be launched into space with no protective equipment, brought back, cut up into small chunks, powdered, and used as animal feed.</description>
<pubDate>24 Oct 2020 00:00:00 +0000</pubDate>
<link>https://ecs.d2evs.net/posts/2020-10-24-Apple.gmi</link>
<guid isPermaLink="true">gemini://ecs.d2evs.net/posts/2020-10-24-Apple.gmi</guid>
</item>
<item>
<title>Self-hosting my life</title>
<description># Self-hosting my life

Over the last few months, I've grown to despise the way most people now interact with computers. A typical person:


	Uses Gmail for their email. Some folks may still some other megacorp email, but rarely will folks even use their own domain name.
	Uses Google to search the web. Because we all know that Google can be trusted:


=&gt; https://amp.dev
=&gt; https://en.wikipedia.org/wiki/Dragonfly_(search_engine)
=&gt; https://support.google.com/youtube/thread/19190975
=&gt; https://en.wikipedia.org/wiki/Censorship_by_Google


	Uses Discord, Skype, Zoom, and other proprietary chat programs.
	Uses Spotify (or a competitor) for music.
	Uses Netflix to watch movies and TV shows.
	Uses Facebook, Twitter, or whatever the proprietary social media du jour is.
	Uses Windows or macOS. Some people don't even have non-mobile devices.


You get the idea.

Each service we use that's operated by someone other than ourselves is another point of failure in our lives. Each of these corporations is handling staggeringly large amounts of personal data. When (not if) these companies go belly-up, they inevitably will take peoples' online lives with them. I think it's clear to see that these are horrible companies, to be avoided at all cost. But what's the alternative?

Enter FOSS[0]. Over the past 30ish years, millions of man-hours have been put into creating an ecosystem of open-source[1] tools. It's feasible to run a daily-driver computer with 100% free software. But why stop there? There are open-source self-hostable solutions for practically all of these.

A quick overview, only including what I was able to set up within 20
days:


	Postfix/Dovecot for email, as well as alps webmail
	The Lounge for some the non-techies, and a normal IRC bouncer for myself
	I'm actually quite happy with a combination of mpd and mpv, though I'm considering setting up FunkWhale
	A Pleroma instance
	Finally, this blog


=&gt; https://git.sr.ht/~emersion/alps alps webmail

Surprisingly enough, the last one took the most time, since I opted to use wersh instead of a more "normal" framework. I'm rather proud of where I ended up, check out the code if you're curious!

=&gt; https://git.d2evs.net/~ecs/wersh wersh
=&gt; https://git.d2evs.net/~ecs/ecs.d2evs.net This site's source code

While setting these up took some time, it's a fixed time-sink, with a low maintenance cost. If you'd like to migrate off of any of these services but don't have the technical experience, send me an email at ecs@d2evs.net!

[0]: Free and Open-Source Software. Loosely, this is where programmers make the secret sauce that powers their programs tick publicly available. Even better, they take contributions!
[1]: I use the terms "FOSS", "free software", and "open-source" interchangeably. Some people don't, and that's a topic for another day.</description>
<pubDate>22 May 2020 00:00:00 +0000</pubDate>
<link>https://ecs.d2evs.net/posts/2020-05-22-Selfhosting.gmi</link>
<guid isPermaLink="true">gemini://ecs.d2evs.net/posts/2020-05-22-Selfhosting.gmi</guid>
</item>
<item>
<title>Installing OpenBSD on a PowerBook G4</title>
<description># Installing OpenBSD on a PowerBook G4

TL;DR:


	Download the OpenBSD install ISO
	If you have a CD and your CD drive works, do that
	If that doesn't work, flash the ISO onto a USB, and try booting to it
	If that doesn't work, flash the ISO onto the hard drive
	Install OpenBSD normally. Make sure to put / on the first partition.


### My ISO flashing procedure

	Remove the battery
	Unscrew all 18 screws
	Take off the memory bay
	Remove the RAM stick closest to the battery
	Unplug the ribbon cable beneath that RAM stick
	Unscrew both Torx T8 screws inside the memory bay
	Carefully remove the keyboard/plamrest/touchpad
	Unplug the ribbon cable going over the hard drive
	Unplug the ribbon cable coming from the hard drive
	Remove the ribbon cable from the hard drive
	The hard drive is a 2.5" IDE drive. Connect it to your computer and flash the ISO:

dd if=openbsd.iso of=/dev/sdb



	Follow these steps in reverse order to put everything back together.
	Boot to OpenFirmware, and run:

boot hd:,\ofwboot /6.6/macppc/bsd.rd

I recently found an old computer lying around in my house -- an old 17" PowerBook G4 (A1085, or 5,5). It had stopped working years ago, but with a new charger, I got it to boot to OS X. However, I wanted to run an open-source OS on it.

32-bit PowerPC is old enough that practically no-one still supports it. I ended up choosing OpenBSD, because I managed to boot it before I managed to boot NetBSD or FreeBSD, and I didn't want to have to deal with Gentoo's compilation hell on a computer that old.

I downloaded the OpenBSD install ISO, and got to work. The first challenge was booting to it. The CD drive wasn't working, so I couldn't boot from there, and for whatever reason, OpenFirmware would hang whenever I tried to boot from a USB.

So I booted to MacOS (which could read the USB) and used it to copy the contents of the ISO onto the HFS partition.

It worked... almost. I realized my mistake as soon as I had repartitioned the hard drive. The sets (gzipped tarballs that are untarred over the new root) were on the now-destroyed HFS partition. In addition, I had completely messed up `/dev` earlier (by running `sh /dev/MAKEDEV all` and running out of space on the ramdisk), so I wasn't able to copy over sets from my USB. I ended up accidentally killing my shell without a `/dev/console`, leading to a wedged system.

However, I still had the OpenBSD kernel running, so I was hoping that maybe, just maybe, if I hot-swapped the hard drive...

So I pulled out a stick of RAM, unplugged the internal keyboard, and it must have touched some circuit because there was a flash and magic smoke was leaking out of the logic board. I quickly unplugged it and removed the battery (thank God for easily removable batteries!).

Note to future self: DO NOT try to take apart a computer while it is running!

When I turned it on, it no longer chimed on boot, but was otherwise unharmed.

However, I still had a bit of a problem. I needed to boot to OpenBSD, but I had no way of booting a CD or USB, and the hard drive was wiped.

Thankfully, a quick `dd if=openbsd.iso of=/dev/sdb` after plugging the hard drive into my computer worked, and the installer booted just fine from there:

boot hd:,\ofwboot /6.6/macppc/bsd.rd


This time, I made sure that I could access the sets before I partitioned the hard drive. After putting the ISO in the first partition of an MBR, I was able to mount `/dev/sd0i`. Success! Now I just had to finish the installer, and reboot.

Not so fast. The install didn't work, and I wasn't able to debug because I wasn't able to mount the OpenBSD partition on Linux. So I re-installed the installer (by this point, I was quite proficient at taking the PowerBook apart), reinstalled OpenBSD, and got a shell after the install process, in order to see what was going on.

It went fine, but my custom partitioning had put the kernel in the second partition (with the first partition being swap), which apparently confused the bootloader. Once I fixed that, it worked!</description>
<pubDate>13 Jan 2020 00:00:00 +0000</pubDate>
<link>https://ecs.d2evs.net/posts/2020-01-13-PowerBook-G4-OpenBSD.gmi</link>
<guid isPermaLink="true">gemini://ecs.d2evs.net/posts/2020-01-13-PowerBook-G4-OpenBSD.gmi</guid>
</item>
<item>
<title>Adventures installing OpenWRT</title>
<description># Adventures installing OpenWRT

I recently got my MikroTik RB2011U, and I wanted to install OpenWRT on it.

=&gt; https://openwrt.org/toh/mikrotik/rb2011uias MikroTik RB2011U

The MikroTik RouterBOARD has a "Netinstall" mode, in which it connects to `eth1` and receives a kernel to boot from using TFTP (the Trivial File Transfer Protocol). I ended up with this small script, based on something from the OpenWRT documentation:

!/bin/sh

Replace `enp2s0f1` with your ethernet interface

Serve TFTP from $TFTP_DIR/$TFTP_FILE

TFTP_DIR=~/tmp/openwrt

TFTP_FILE=stable.elf

sudo ifconfig enp2s0f1 192.168.1.10 up

sudo dnsmasq -i enp2s0f1 --dhcp-range=192.168.1.100,192.168.1.200 \

--dhcp-boot=$TFTP_FILE --enable-tftp --tftp-root=$TFTP_DIR \

--enable-tftp -d -u $USER -p0 -K --log-dhcp --bootp-dynamic


Once the router attempted to connect to the DHCP server with the hostname `OpenWrt`, I killed the TFTP server.

This worked - the router was now running OpenWRT from RAM.

Then, I disconnected my laptop from `eth1` and plugged it into `eth2`. The OpenWRT initramfs is configured to act as a DHCP server on `eth2` through `eth10`.

At this point, I `ssh`ed into the router:

$ ssh root@OpenWrt


(In the future, commands starting with `


gemini - kennedy.gemi.dev




 are on the host computer, and commands starting with `#` are on the router.)

I now had a root shell on the router! So far so good.

I then `scp`ed the OpenWRT sysupgrade[1] onto the router:

$ scp ~/tmp/openwrt/stable.bin root@OpenWrt:/tmp/sysupgrade.bin


and proceeded to `sysupgrade` the router:

sysupgrade /tmp/sysupgrade.bin

Image metadata not found

Cannot save config while running from ramdisk.

Commencing upgrade. Closing all shell sessions.


At this point, the router closed the SSH connection and entered a boot loop.

Huh.

That's interesting.

I tried repeating the previous steps, replacing stable (18.06.4) OpenWRT with the snapshot. (The RB2011U v2 is listed as only supported by OpenWRT snapshot.)

No luck.

In vain, I tried pinging the router as it boot looped. It never responded.

This meant that it was crashing and burning *before* networking was initialized. (This also meant that OpenWRT failsafe mode wouldn't help.)

So, I tried to read the kernel logs from flash memory:

mount /dev/mtdblock6 /mnt

mount: mounting /dev/mtdblock6 on /mnt/ failed: Invalid argument


Maybe I need to specify the filesystem type? JFFS2 seems likely:

mount -t jffs2 /dev/mtdblock6 /mnt

(hangs for a really long time)

mount: mounting /dev/mtdblock6 on /mnt/ failed: Invalid argument


That last case was interesting, so I pulled up another SSH connection and:

dmesg | tail -1

[ 703.468505] jffs2: jffs2_scan_eraseblock(): Magic bitmask 0x1985 not

found at 0x0244003c: 0xa94c instead

[ 855.211369] jffs2: Cowardly refusing to erase blocks on filesystem

with no valid JFFS2 nodes

[ 855.220100] jffs2: empty_blocks 0, bad_blocks 3, c->nr_blocks 992


It appears that there's no JFFS2 filesystem on /dev/mtdblock6. That makes sense, given:

cat /proc/mtd | grep mtd6

mtd0: 0000c000 00001000 "routerboot"

mtd1: 00001000 00001000 "hard_config"

mtd2: 00001000 00001000 "bios"

mtd3: 00001000 00001000 "soft_config"

mtd4: 00040000 00020000 "booter"

mtd5: 003c0000 00020000 "kernel"

mtd6: 07c00000 00020000 "ubi"


It would appear that `mtd6` has UBIFS on it, but:

mount -t ubifs /dev/mtdblock6 /mnt

mount: mounting /dev/mtdblock6 on /mnt/ failed: Invalid argument


At this point, I decided to dump `/dev/mtd6` to my laptop, and do some forensics:

dd if=/dev/mtd6 | ssh ecs@192.168.1.10 dd of=/tmp/mtd6

blah blah blah fingerprints blah blah blah

ecs@192.168.1.10's password:

253952+0 records in

253952+0 records out

253944+10 records in

253952+0 records out

130023424 bytes (130 MB, 124 MiB) copied, 313.463 s, 415 kB/s


Wait! UBI is `ubi`, not `ubifs`:

mount -t ubi /dev/mtdblock6 /mnt

mount: mounting /dev/mtdblock6 on /mnt/ failed: No such device


What?!?!

ls -l /dev/mtdblock6

brw------- 1 root root 31, 6 Jan 1 1970 /dev/mtdblock6


By this point, I've installed `ubi_reader` on my laptop and I go back there to do some diagnostics:

$ ubireader_list_files mtd6

UBI Fatal: Less than 2 layout blocks found.

$ less mtd6

UBI#^A^@^@^@...^E^@^@^B...

...

<FF><FF><FF><FF><FF>...


Interesting. It appeared that OpenWRT did not successfully flash, but it failed consistently. (By this time, I had re-flashed OpenWRT quite a few times.)

At this point, it was 5:00 in the morning, so I decided to go to sleep.

Update: I returned that router and bought another one. Same problem. I decided to give up.

[1]: OpenWRT has a `sysupgrade` command which takes a firmware image and installs it. The install procedure for MikroTik RouterBOARDs is to run OpenWRT from RAM using TFTS, then use `sysupgrade` to flash a firmware upgrade.</description>
<pubDate>18 Oct 2019 00:00:00 +0000</pubDate>
<link>https://ecs.d2evs.net/posts/2019-10-18-Installing-OpenWRT.gmi</link>
<guid isPermaLink="true">gemini://ecs.d2evs.net/posts/2019-10-18-Installing-OpenWRT.gmi</guid>
</item>
	</channel>
</rss>

!/bin/sh

Replace `enp2s0f1` with your ethernet interface

Serve TFTP from $TFTP_DIR/$TFTP_FILE

sysupgrade /tmp/sysupgrade.bin

mount /dev/mtdblock6 /mnt

mount -t jffs2 /dev/mtdblock6 /mnt

(hangs for a *really* long time)

dmesg | tail -1

cat /proc/mtd | grep mtd6

mount -t ubifs /dev/mtdblock6 /mnt

dd if=/dev/mtd6 | ssh ecs@192.168.1.10 dd of=/tmp/mtd6

blah blah blah fingerprints blah blah blah

mount -t ubi /dev/mtdblock6 /mnt

ls -l /dev/mtdblock6

UBI#^A^@^@^@...^E^@^@^B...

...

&lt;FF&gt;&lt;FF&gt;&lt;FF&gt;&lt;FF&gt;&lt;FF&gt;...

(hangs for a really long time)

<FF><FF><FF><FF><FF>...