I’m back to working on my WG3526-similar device, as it’s too cold to work on the cars, and the last HP 4100 was (re)built before lunch. It’s nice to have a functional panel on it.

I’ve found someone else who received a device which was based on the ZBT WE1326, and has a slightly different design than mine internally. Of course, his works with the stock LEDE DTS for the WE1326. Naturally!

I really like the default design of the hnyman WNDR3×00 builds which have dyndns, luci-ssl, et al- so I started working on one of my own now that, indeed, the 5Ghz is stable with a standard WE1326, even if the 2.4Ghz radios are still pretty useless (which is bad for my needs- I’m still primarily 2.4Ghz- one of the perils of not upgrading every couple years for the sake of upgrading).

Anyhow, onto the “Well duh” moments (some I’ve discovered though trial and error, yet others, I was aware of due to others’ trial and error):

  • Don’t install openssh server by default- it has wisely disabled root by default, whereas dropbear will let you login, which is useful for a clean flash (which has no password). dropbear also uses fewer resources. Don’t be wasteful on embedded targets!
  • Don’t forget to not only update your packages, but to install them. luci-ssl isn’t all that useful when your docroot is empty.
  • Always sysupgrade from /tmp – I’ve never not done so, but I can’t imagine it being great to try to do so from a bit-banged MMC.
  • Always be aware of your umask. Anything except umask 022 is going to cause pain!

Since my same-but-not device is a different revision than that of the only other person I know who owns one- I really have to question how we’re going to get this supported in LEDE since we don’t have a real definitive platform for this device, or revision history.

It looks like this device I am trying to get mainline support into LEDE was done in a greymarket way. It has an older kernel, but is built against a version of OpenWRT that is 2 years old (15.05). It looks like they may have used DreamBox Binary Modules. I don’t want to lock myself down, especially to a Dirty COW (if not patched) kernel, but this may be a way to get a semi-recent LEDE working, and stable.

I thought I should write a little cautionary note for anyone considering the not-as-well know Teleflora.com.

Two years ago, I used them to deliver roses on Mothers’ Day, and they arrived wilted and half dead. When I contacted them about my dissatisfaction, they did not offer a replacement or refund- they offered me 10% off my subsequent order. I told them that this was not sufficient, and they offered no further solution, despite asking to have them replaced.

Needless to say, there was no second order. Stick to FTD, 1-800-FLOWERS, or your trusted local florist.

root@deux:~# /usr/bin/wget http://ftp3.usa.openbsd.org/pub/OpenBSD/6.1/amd64/cd60.iso


Even if you’re used to it, brctl is deprecated. Here’s how to setup a bridge using iproute2, take note that before you can add the bridge to the device, the device has to be up:

ip link set dev $DEVICE up
ip link add $BRIDGE type bridge
ip link set $DEVICE master $BRIDGE
ip link set dev $BRIDGE up

Chances are that at some time, you’ll need to repair a corrupted filesystem, or worse, fix something you accidentally broke. Sometimes, your operating system will do this, especially if you are a fan of Linux. Such as the case with the Core2, today:

[    0.179209] Unpacking initramfs...
[    1.176420] ------------[ cut here ]------------
[    1.176514] WARNING: at arch/x86/mm/init.c:504 free_init_pages+0x53/0x11f()
[...]
[    1.177589] Call Trace:
[    1.177672]  [<ffffffff81734b42>] dump_stack+0x19/0x1f
[    1.177759]  [<ffffffff8104d6b6>] warn_slowpath_common+0x62/0x7b
[    1.177846]  [<ffffffff8104d6e4>] warn_slowpath_null+0x15/0x17
[    1.177933]  [<ffffffff81037181>] free_init_pages+05x3/0x11f
[    1.178025]  [<ffffffff81d32c4b>] free_initrd_mem+0x30/0x35
[    1.178112]  [<ffffffff81d1db2b>] populate_rootfs+0xeb/0x10c
[    1.178199]  [<ffffffff81d1da40>] ? unpack_to_rootfs+0x289/0x289
[    1.178286]  [<ffffffff8100209b>] do_one_initcall+0x9b/0x135
[    1.178373]  [<ffffffff81d1b8eb>] kernel_init_freeable+0x154/0x1ea
[    1.178461]  [<ffffffff81d1b981>] ? kernel_init_freeable+0x1ea/0x1ea
[    1.178550]  [<ffffffff81725900>] ? rest_init+0x74/0x74
[    1.178636]  [<ffffffff81725909>] kernel_init+0x9/0xd6
[    1.178723]  [<ffffffff8173d73c>] ret_from_fork+0x7c/0xb0
[    1.178810]  [<ffffffff81725900>] ? rest_init+0x74/0x74
[    1.178900] ---[ end trace ded000897d630586 ]---
[    1.179052] Freeing initrd memory: 11812k freed

Long story short, the kernel became unhappy with the latest update, and being a headless machine with no remote hands, I had to figure how to fix it.

Now, although I don’t have remote hands, I have a fairly dated (and limited) sysrescue image which I can boot the system from, and manipulate my installation from there.

Despite architectural and distribution differences, a 3.x kernel with most options available can get you back online with minimal effort.

Althought teaching you how to handle LVM/etc is not something for a quick little diary entry, I will offer this bit of advice: The following mounting heirarchy tends to work OK for Debian, CentOS, and almost any other Linux distribution I’ve thrown at it:

# fsck.{filesystem} /dev/{device}{partition} #root partition
# mount -t {filesystem} /dev/{device}{partition} /mnt
# mount -o bind /dev /mnt/dev
# mount -t sysfs sys /mnt/sys
# mount -t devpts pts /mnt/dev/pts
# mount -t proc proc /mnt/proc
# cd /mnt
# {mount/fsck any subpartitions here}
# chroot /mnt /bin/{dash|bash|ash}

Happy repairs!

If you browse back a few pages, you might notice that along with the mass of printers, there are some PCs, LCDs, and associated components:

I’m hoping to continue refurbishing and selling the handful of laser printers to recoop my expenses (25’ U-Haul 1-way rentals are NOT cheap), but my wish is to rebuild these computers for those who cannot afford them.

Coming from a government auction, they’re all functional- but missing hard drives (and needing new CMOS batteries, but I have plenty CR2032s). There are 5 complete units that are worth salvaging, ranging from P4HT and Core2 based eras. 15”-17” LCDs, keyboards, and mice are all included, of course.

They’ve all got Windows license stickers on the machines themselves, so I can reinstall an OS- should I be able to find drives. I don’t have drives, but I’m hoping to get a bulk order of ~160GB SATA1 disks cheap enough.