That reminds me of a three week, three person group project I had back when I was in university. There were a couple of topics we could pick from, including writing a very simple operating system from scratch. I asked the professors if we could instead port MINIX3 to the Raspberry Pi (given that it already had an ARM port to the BeagleBoard) and they accepted.
What was supposed to be a reasonably straightforward project proved to be a huge challenge as we encountered numerous unforeseen technical difficulties. One particularly harrowing one was that the Raspberry Pi 3 booted into hypervisor mode instead of supervisor mode and QEMU's Raspberry Pi emulation accuracy was so abysmal back then to be borderline useless for osdev. I recall it took me an entire week of low-level hardware debugging just to figure that one out.
By the end, we pulled through and delivered a working port with UART, GPIO and framebuffer drivers that could run on the Raspberry Pi 2 and 3. We run our presentation with the port on real hardware, using a shell script that displayed bitmaps from the ramdisk and monitored GPIO pins to move slides forwards or backwards (I used a knife to short-circuit the pins as needed). It was by far the coolest presentation of all the groups just on originality alone and I think I still have the image of that SD card somewhere.
Maybe, but I already had a reputation of being the dark wizard back then. If anything, the other students in my group went along with this because they knew I could overcome any problem... regardless of the cost on my sanity.
> How did you handle the debugging the raspberry pi on real hardware?
Painfully through serial output. I didn't have access to a JTAG probe at the time (I'm not even sure the Raspberry Pi could be debugged that way) and documentation was exceedingly poor.
After that experience, I refuse to debug anything hardware-related without at the very least a GDB stub.
This is Broadcom we're talking about, where that's par for the course. Personally I'd choose a SoC from AllWinner or Rockchip or even Mediatek over them.
> I used a knife to short-circuit the pins as needed
Reminds me of shorting the two power-on pins on an ATX motherboard to start it without a switch installed. Obviously, your setup was far cooler. Nice work.
We did this sort of thing at SFU some 25-30 years ago; though we stopped short of running an OS and compiler on it, and it wasn't a group project.
For those interesting in trying this sort of experiment, but wish to have some guidance and accessible tooling, I highly recommend Turing Complete; you'll go from a few gates to a full computer. Components can be shared with the community; where you'll find things like a RiscV core and such. Anyhow, it's great fun. Do recommend. It's on Steam:
while building altogether cpu + compiler + os ,there's no platform under you. you're the platform.. every bug you hit is a law of your system.. most of us debug through layers someone else wrote. here, even the rules are homemade..op debugged his own rules
LOL in my 4th year Advanced Operating Systems Concepts course we wrote a toy x86 OS from scratch. We obviously didn't have to make our own hardware, but uhhhh I definitely added a bunch of printfs inside QEMU to dump out CPU states when we couldn't figure out the chain of events that led to hard faults.
On the other side... have also definitely used a pair of LEDs to try to debug an RTOS on a microcontroller with no JTAG access...
That reminds me of a three week, three person group project I had back when I was in university. There were a couple of topics we could pick from, including writing a very simple operating system from scratch. I asked the professors if we could instead port MINIX3 to the Raspberry Pi (given that it already had an ARM port to the BeagleBoard) and they accepted.
What was supposed to be a reasonably straightforward project proved to be a huge challenge as we encountered numerous unforeseen technical difficulties. One particularly harrowing one was that the Raspberry Pi 3 booted into hypervisor mode instead of supervisor mode and QEMU's Raspberry Pi emulation accuracy was so abysmal back then to be borderline useless for osdev. I recall it took me an entire week of low-level hardware debugging just to figure that one out.
By the end, we pulled through and delivered a working port with UART, GPIO and framebuffer drivers that could run on the Raspberry Pi 2 and 3. We run our presentation with the port on real hardware, using a shell script that displayed bitmaps from the ramdisk and monitored GPIO pins to move slides forwards or backwards (I used a knife to short-circuit the pins as needed). It was by far the coolest presentation of all the groups just on originality alone and I think I still have the image of that SD card somewhere.
Nice.
> I asked the professors if we could instead port MINIX3 to the Raspberry Pi
I think they were expecting you to fail.
> QEMU's Raspberry Pi emulation accuracy was so abysmal
When I did some hobby OS dev my strategy was to make it work on QEMU and then pray it would work on real hardware as well, which worked OK...
How did you handle the debugging the raspberry pi on real hardware?
> I think they were expecting you to fail.
Maybe, but I already had a reputation of being the dark wizard back then. If anything, the other students in my group went along with this because they knew I could overcome any problem... regardless of the cost on my sanity.
> How did you handle the debugging the raspberry pi on real hardware?
Painfully through serial output. I didn't have access to a JTAG probe at the time (I'm not even sure the Raspberry Pi could be debugged that way) and documentation was exceedingly poor.
After that experience, I refuse to debug anything hardware-related without at the very least a GDB stub.
and documentation was exceedingly poor.
This is Broadcom we're talking about, where that's par for the course. Personally I'd choose a SoC from AllWinner or Rockchip or even Mediatek over them.
> I used a knife to short-circuit the pins as needed
Reminds me of shorting the two power-on pins on an ATX motherboard to start it without a switch installed. Obviously, your setup was far cooler. Nice work.
We did this sort of thing at SFU some 25-30 years ago; though we stopped short of running an OS and compiler on it, and it wasn't a group project.
For those interesting in trying this sort of experiment, but wish to have some guidance and accessible tooling, I highly recommend Turing Complete; you'll go from a few gates to a full computer. Components can be shared with the community; where you'll find things like a RiscV core and such. Anyhow, it's great fun. Do recommend. It's on Steam:
https://store.steampowered.com/app/1444480/Turing_Complete/
That looks like the game-version of nand2tetris which I enjoyed doing a while back too (as another suggestion).
while building altogether cpu + compiler + os ,there's no platform under you. you're the platform.. every bug you hit is a law of your system.. most of us debug through layers someone else wrote. here, even the rules are homemade..op debugged his own rules
Previously:
https://news.ycombinator.com/item?id=24680109
Very impressive! Low end work is often tedious and time consuming, especially if you lack the essentials like a debugger.
Until you've used an oscilloscopes to debug your buggy kprintf, you haven't lived.
First thing to do is get a working UART.
First thing to do is get a GPIO pin or two to toggle. Plenty of debug patterns that one can blink out.
LOL in my 4th year Advanced Operating Systems Concepts course we wrote a toy x86 OS from scratch. We obviously didn't have to make our own hardware, but uhhhh I definitely added a bunch of printfs inside QEMU to dump out CPU states when we couldn't figure out the chain of events that led to hard faults.
On the other side... have also definitely used a pair of LEDs to try to debug an RTOS on a microcontroller with no JTAG access...
(2020)
Magic-1 and BMOW did similar, a while back. See: https://www.homebrewcpu.com/
The actual list of sites who have built their own CPUs: https://www.homebrewcpuring.org/
Now they gotta run to the semiconductor lab and get them to fabricate the CPU instead of configuring an FPGA to run it.