Blog

Moving from Santa Clara, CA to Hyattsville, MD

2025-11-12T05:00:00Z

Road Trip from Santa Clara, CA to Hyattsville, MD

Last modified : 12 November, 2025

In this post, I’m sharing data from my cross-country drive from Santa Clara, California to Hyattsville, Maryland. Along the way, I stopped in West Lafayette, Indiana to see family (a detour of 60 miles). I drove a 16 foot Penske truck and kept track of fuel stops, miles traveled, and costs throughout the journey.

The truck was relatively lightly packed with lots of room. I could’ve should’ve could’ve used a smaller 10 foot truck, except for the fact that I needed to transport my motorcycle, and 10 ft trucks don’t have a ramp. Maybe I should have bought my own ramp. The rental cost for the truck was $3308.46. I chose not to insure the truck separately because Penske was asking for almost $600 and my car insurance already covered rental vehicles. I also drove the truck extremely carefully and almost always 5-10 miles below the speed limit on cruise control.

🚗 Trip Summary

Starting point: Santa Clara, CA
Destination: Hyattsville, MD
Distance traveled: 2,924 miles
Fuel consumed: 242.83 gallons
Total cost: $738.89
Average fuel economy: ~12.0 mpg
Average cost per gallon: $3.04
Average cost per mile: $0.25

🗺️ Fueling and Mileage Log

#	Place	Miles to this place	Gallons to this place	Cost to this place	Cumulative miles	Cumulative gallons	Cumulative cost	MPG (segment)
1	Santa Clara, CA	–	–	–	0	0	$0.00	–
2	Fernley, NV	284	25.533	$82.70	284	25.533	$82.70	11.1
3	Elko, NV	259	22.079	$75.93	543	47.612	$158.63	11.7
4	West Valley, UT	224	18.651	$58.55	767	66.263	$217.18	12.0
5	Laramie, WY	396	30.504	$88.13	1,163	96.767	$305.31	13.0
6	Big Springs, NE	198	14.283	$40.84	1,361	111.050	$346.15	13.9
7	Lincoln, NE	294	25.588	$75.71	1,655	136.638	$421.86	11.5
8	Chillicothe, MO	213	17.741	$49.66	1,868	154.379	$471.52	12.0
9	Macon, MO	60	0.167	$0.49	1,928	154.546	$472.01	359.3
10	Hoopeston, IL	295	30.970	$95.36	2,223	185.516	$567.37	9.5
11	Centerville, IN	177	13.155	$39.45	2,400	198.671	$606.82	13.5
12	Morgantown, WV	318	26.139	$78.39	2,718	224.810	$685.21	12.2
13	Hyattsville, MD	206	18.019	$53.68	2,924	242.829	$738.89	11.4

📝 Note: The Macon, MO stop was a very small top-off, hence the unusually high calculated MPG for that leg.

⛽ Observations

The best fuel efficiency was between Big Springs, NE and Lincoln, NE at nearly 13.9 mpg.
The most expensive fuel stop was in Hoopeston, IL at $95.36.
The total cost of fuel for the entire trip came to $738.89.
The trip averaged around 12 mpg.

Hope this helps any future movers.

Networking on Linux.

2024-09-17T04:00:00Z

Networking on Linux.

Last modified : 17 September, 2024

This page talks about how the networking stack is set up on some Linux computers.

Device Enumeration.

The Linux kernel boots up and enumerates all devices attached to all the buses (e.g. PCI). The devices are recognized by some mechanism and appropriate device drivers are loaded in, so that the kernel can interface with the devices. The networking devices are enumerated in SysFS at /sys/class/net.

$ ls -al /sys/class/net
total 0
drwxr-xr-x.  2 root root 0 Sep 17 10:00 .
drwxr-xr-x. 84 root root 0 Sep  4 14:13 ..
lrwxrwxrwx.  1 root root 0 Sep  4 14:13 lo -> ../../devices/virtual/net/lo
lrwxrwxrwx.  1 root root 0 Sep  4 14:13 wlp166s0 -> ../../devices/pci0000:00/0000:00:1d.0/0000:a6:00.0/net/wlp166s0

The devices are named so that across different boots, the name stays the same.

To find out which device driver was loaded for each interface, you can use ethtool. e.g.

$ ethtool -i wlp166s0 
driver: iwlwifi
version: 6.10.6-200.fc40.x86_64
firmware-version: 89.202a2f7b.0 ty-a0-gf-a0-89.uc
expansion-rom-version: 
bus-info: 0000:a6:00.0
supports-statistics: yes
supports-test: no
supports-eeprom-access: no
supports-register-dump: no
supports-priv-flags: no

Also SysFS has a link to the driver for the device:

$ readlink /sys/class/net/wlp166s0/device/driver
../../../../bus/pci/drivers/iwlwifi

Bringing up the interface(s).

The kernel sends+receives Ethernet / 802.11 / (some equivalent) Layer 2 frames to+from the device driver. The device driver interfaces with the networking hardware and sends+receives those Layer 2 frames. To check statistics you can use the ip command. e.g.

$ ip -s link show wlp166s0 
2: wlp166s0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP mode DORMANT group default qlen 1000
    link/ether d2:5d:d0:e6:c8:5e brd ff:ff:ff:ff:ff:ff permaddr 8c:f8:c5:ed:92:fc
    RX:   bytes  packets errors dropped  missed   mcast           
    21506231662 18394460      0   14606       0       0 
    TX:   bytes  packets errors dropped carrier collsns           
     1289526154  3582885      0       0       0       0

For more detailed statistics for wireless hardware, you can use $ iw dev wlp166s0 station dump.

Routing table.

The kernel uses routing tables to determine which interface should handle a connection. Routing tables map IP address ranges to specific interfaces. We can view the routing table using:

$ ip route show
default via 192.168.68.1 dev wlp166s0 proto dhcp src 192.168.68.59 metric 600 
192.168.68.0/22 dev wlp166s0 proto kernel scope link src 192.168.68.59 metric 600

This routing table is populated by the kernel using (typically) DHCP. The DHCP client sends a request which is responded to by the DHCP server. $ nmcli device show can show the result of DHCP.

DNS

DHCP may also provide extra information for example DNS configuration.

$ resolvectl status
Global
         Protocols: LLMNR=resolve -mDNS -DNSOverTLS DNSSEC=no/unsupported
  resolv.conf mode: stub

Link 2 (wlp166s0)
    Current Scopes: DNS LLMNR/IPv4 LLMNR/IPv6
         Protocols: +DefaultRoute LLMNR=resolve -mDNS -DNSOverTLS DNSSEC=no/unsupported
Current DNS Server: 192.168.68.1
       DNS Servers: 206.225.75.226 206.225.75.225 192.168.68.1

VPN

When a VPN connection is established, it typically modifies the routing table on your machine so that your traffic is routed through the VPN’s virtual interface rather than directly through your physical network interfaces (like eth0 or wlan0).

There are different types of VPNs, but two of the most common protocols used are:

OpenVPN (uses tun or tap devices)
WireGuard (uses wg interfaces)

Using Linux bluetooth to spoof Tozo headphones to pair them

2024-01-11T05:00:00Z

Spoofing a TOZO headphone in order to pair the left with the right.

Last modified : 11 Jan, 2024

This seems to be a fairly common problem. I faced it too. Despite following instructions to “hard-reset”, the left and right headphones would just not pair with each other. I was able to pair each individual headphone with my laptop, which forced me to hear audio from only 1 side. I tried draining the battery completely in the hope that it would be akin to a “factory-reset”, but that didn’t help either.

I was able to make the 2 headphones (TOZO-T10-L and TOZO-T10-R) pair with each other using Linux bluetooth commands. My strategy was to

turn off TOZO-T10-L
Change my laptop bluetooth adapter to pretend to be TOZO-T10-L
Turn on TOZO-T10-R
Pair TOZO-T10-R with my laptop (which is pretending to be TOZO-T10-L)
Turn off TOZO-T10-R
Turn off bluetooth on my laptop.
Turn on both TOZO-T10-L and TOZO-T10-R

Turns out the above process did actually work for me and I was able to save my perfectly usable TOZO headphones from the trash.

I first gathered the information about the headphones using $ bt-device -l Added devices: TOZO-T10-L (00:00:00:03:07:8F) TOZO-T10-R (00:00:00:00:3B:7F)

I also saved my laptop’s real bluetooth configuration in a temporary file. The output of the command: $ btmgmt info Index list with 1 item hci0: Primary controller addr 01:23:45:67:89:00 version 12 manufacturer 2 class 0x7c010c supported settings: powered connectable fast-connectable discoverable bondable link-security ssp br/edr le advertising secure-conn debug-keys privacy configuration static-addr phy-configuration wide-band-speech cis-central cis-peripheral current settings: powered connectable discoverable bondable ssp br/edr le secure-conn cis-central cis-peripheral name something short name hci0: Configuration options supported options: public-address missing options:

Then turn off TOZO-T10-L and TOZO-T10-R. Make my laptop pretend to be TOZO-T10-L: $ sudo btmgmt public-addr 00:00:00:03:07:8F $ sudo btmgmt name TOZO-T10-L

Turn on TOZO-T10-R. Pair my laptop again with TOZO-T10-R using: $ sudo btmgmt unpair 00:00:00:00:3B:7F $ sudo btmgmt pair 00:00:00:00:3B:7F

Then turn off my laptop bluetooth (using the GUI). Turn off both headphones. Turn on both headphones. At this point my two headphones paired with each other and TOZO-T10-R was open to be paired with a device. I turned them both off and un-spoofed my laptop using. $ sudo btmgmt public-addr 01:23:45:67:89:00 $ sudo btmgmt name something

Turning them all again, I was able to pair everything perfectly and the audio came from left and right headphones correctly.

Benchmarks for SanDisk Extreme microSD cards.

2019-11-11T05:00:00Z

Benchmarks for SanDisk Extreme microSD cards.

Last modified : November 11, 2019

I recently bought 2 SanDisk Extreme microSD cards (256Gb). I also happened to have the same SD card brand for (32Gb). The device I had it in was also reporting errors writing to the SD card. I benchmarked before and after fixing it using the Linux Disk Utility. I ran the Linux filesystem benchmarks and thought it might be useful for people who are also considering purchasing it. Even though the SD cards are advertised as having a max read throughput of 160MB/s, I wasn’t able to get that performance. I tried with different filesystem formats (ext4, NTFS, FAT), still no dice

32Gb card before fixing:

32Gb card after fixing:

32Gb card after fixing with different filesystems:

Ext4

FAT

NTFS

256Gb card: SanDisk Extreme microSD UHS-I Card - 256GB. Product SKU SDSQXA1-25

Disassembling a Hard Drive

2019-06-12T04:00:00Z

Disassembling a hard drive

Last modified : 12 June, 2019

In the past year one of the hard drives in my desktop started to fail. It first manifested as SMART errors while booting up, and then one day a remarkably loud scratching noise. The irony is that this was the newer of the 2 hard disks in my RAID1 (mirroring) configuration. It was a Seagate Barracuda ST2000DM001. By my count, it lasted me ~3.5 years. My desktop is usually idle (no gaming, no constant write / read loads) so its a little disappointing given the other soldier in my RAID1 is going strong still. Oh well!

I decided to destructively disassemble the hdd, just for kicks. The following are the pictures I took and short blurbs. I’m glad I did it because its fascinating how such small things can store so much data. The process of disassembly itself was very easy too. All I needed was the correct screwdrivers (I think 3 of them) that any small screwdriver kit would have.

First to come off was the control board. This board plugs into the SATA port on the Motherboard. On the other side it had a 3-pin (I wonder why 3) connection to the motor which rotates the platter and a springy socket for reading the signals coming from the head. I’ve heard there are millions of lines of code running on the microcontroller on the control board. It exports the SATA API and schedules all the accesses on the physical media.

The next thing to unscrew was the top plate. There were a bunch of screws (and some hidden under stickers) but once all of them were unscrewed, the top plate came off very easily exposing the inner components.

The next thing I unscrewed were the platters. Again a bunch of screws. What I didn’t quite expect was that the platters were so shiny and thick. I bent them to destroy the data, and I couldn’t with my bare hands. I had to use pliers and good leverage. The surface was shiny like a mirror rather than the rusty brown color of iron oxide that was the color of the magnetic tape of floppy disks. I was also surprised that there were only 2 platters. Somehow I had expected more of them. I’m sure a lot of engineering effort has gone into optimizing the characteristics of these components. I did remember the white paper from Google on Disks for Data Centers (2016). Its a good read if you haven’t so far. I was also hoping I’d see some scratches owing to the loud scratching I had heard, but I wasn’t able to spot any with my eyesight.

The next component I took off was the head. It was mounted on a spring system. I probably damaged it while taking it off. The arms were made of solid metal, at the end of which was a triangular plastic protruding with a small sensor at the end. I tried getting good pictures of this from different angles. The base of the arms had a solenoidal loop of wire. I’m guessing the physics of this component are what determine the seek time of these disks.

The last thing to come off was the motor. I did note that the bearings were extremely smooth and efficient. I tried incorrectly to uncap the motor at first hitting it repeatedly with a heavy duty hammer before realizing I was going the wrong way. The rotor needs to be pushed out from the bottom of the hard drive towards the top. Even after such excessive abuse, the bearings were more or less smooth-ish. These bearings are probably the majority of the steady state power draw.

The case itself was pure solid metal. I don’t think any amount of hammering would have bent it. Here’s a picture of the tools that I used: