Building an OS - 3 - The FAT file system

We continue writing our bootloader, and we learn to read a file from a FAT disk.

Tools:

Documentation:

[treytomes]

You actually move faster than I can type. I type pretty fast. I actually find this refreshing compared to other channels that I have to pump up to double-speed to keep up with me.

Also, thanks for going beyond the bootloader section of teaching OS development. Not many video series' make it that far.

[asharneyaz]

Hands down, this is one of the best explanations I have come across so far!

[bro-figc]

these videos are amazing, the best osdev content on youtube!
awesome job!!

[wkjagt]

What an amazing series. I discovered your channel this week and I'm learning a ton. A while ago I implemented a much simplified FAT file system (each entry in the FAT was one byte instead of 12 bits, and I didn't do subdirectories) on a homebrew 6502 computer storing files on serial eeproms, but doing FAT 12 is very cool, and I'm learning a ton from your videos. I'm going to try and run this on my 486 once I find the time to hook it up.

[bad-at-art-games]

This is incredible, thank you :) I just wanted to see some of how an OS is built in practise, as in university it was all quite abstract. I will not ever build one myself, unless for the fun of it. But my curiosity is being sparked perfectly by this

[radeksmola]

Thank you, very great article. I remember writing my own OS 25 years ago. You are master BTW ;)

[adrianstephens]

This has been so far a most enjoyable series. Thank you. I'll be interested to see how far you take it, and whether you delve into real-time. I'm a retired software engineer who has written bare metal OSs (no file systems!) for embedded instrumentation starting with a 6502 microprocessor in the 1980s. I also wrote an OS for an 80186-based WiFi product in about 1995 (well, it was 802.11 frequency hopping, before the WiFi term was coined) - this was an example in how *not* to write a real-time OS/application because I became unnecessarily mired in locks/semaphores protecting unnecessary critical regions. Eventually I pursuaded my bosses to let me re-engineer it, and the company was almost immediately sold off. Not sure what that says about life, the universe and everything.

[somethingwithtech]

Thanks for this amazing series!
Funny story I just searched for 1 hour why my kernel wouldnt print Hello world. I accidently mistyped dx instead of ds just infront of the jump instruction to the kernel. Therefore, he wasn't in sector 0x2 but in sector 0x0

[ashajjar]

I want to take a minute to appreciate your music choice 🤩

[thecodingpress]

This is definitely the best os development content i have seen so far, where did you find out how to do all this?

[SergeySergey-ks]

Thank you, your explanation is one of the best and it serves perfectly as a complementary to other materials

[thomascarlsen]

I think I have lost track. It could be that code part is going really fast and I am not totally sure why we do the things we do. I felt the same in episode 2. Maybe it would benefit to keep reminding us why we do the things we are doing to not loose track.
Otherwise, I Imagine that it takes a lot of time what you are doing and I gratefully thank you to put in the effort and creating such helpful videos. Cheers.

[NotAGrApe]

I hope the file system can lose weight…

[Scaik]

i dont get it, how do i get "root dir size"?

[UltimusTheSoldier]

I think there's a mistake. in 6:32 the table shows 1 sectors per cluster, but in 9:51 you wrote that we have 2 sectors per cluster.

[lens]

Ah, the wonderful days of bios interrupts.

At least it saves you from spending a week writing a disk driver 🤣

[neksodebe]

hey, i think i managed to fix your hard-coded equation for the sector number! i did have to shorten and remove a couple of my message strings at the end, but eventually i got it to fit, and WORK ON REAL HARDWARE!

just replace this:
```asm
mov ax, [kernel_cluster]
add ax, 31
```
with this:
```asm
mov ax, [kernel_cluster]
sub ax, 2 ; Clusters start at 2

; Multiply by sectors per cluster
movzx cx, byte [bdb_sectors_per_cluster]
mul cx ; AX = (cluster - 2) * sectors_per_cluster


; Add the data section start
add ax, 33
```

this actually calculates the sector number instead of making it a hard coded value. tested and working on a 16 GB pendrive. let me know if i'm wrong :)

[indrakumar]

Hi sir, i have carefully followed your instructions but when i run make command i get the following error :make: *** No rule to make target 'tools/fat/fat.c', needed by 'build/tools/fat'. Stop. Could you please guide me through it.

[WistrelChianti]

I love how the number of comments is dropping off the further in we get :D Just about hanging in there on the understanding of these.

[nizambhagaskara]

Hi, where can I learn more about Assembly?