memfault / interrupt Goto Github PK

"series of posts" is repeated twice in these second paragraph - the second "series of posts" should be removed as it is not a hyperlink:

In this series of posts series of posts, we’ll explore ways to save code space and ways not to do it. We will cover compiler options, coding style, logging, as well as desperate hacks when all you need is another 24 bytes.

https://interrupt.memfault.com/blog/best-firmware-size-tools

Hello,

First of all thank you for the article about the bootloader! 😄

I would like to point out that pointer assignment.

If we consider GNU LD, then to get proper access of a shared_data variable, we should use &:

struct shared_data *sd = (struct shared_data *)&_shared_data_stat;

instead of accessing external variable by value like it is in the article:

struct shared_data *sd = (struct shared_data *)_shared_data_stat;

I just stumbled across this post from 2021: https://interrupt.memfault.com/blog/pre-commit

Issue:
Upon trying to use this for a project that I am currently persuing - I found that the provided .pre-commit-config.yaml no longer works.

To re-create: (plain new python env)
Python: Python 3.10.14
pre-commit 3.6.2

What fixed it for me:
Bump isort version 5.10.1 to 5.12.0

Googling gave me your article ( https://interrupt.memfault.com/blog/emulating-raspberry-pi-in-qemu ) as the first result in asking about emulating a pi with qemu, so good for you, you're getting noticed by the algorithm! Thanks for the article, I learned (or re-learned) a lot by reading it. I especially appreciate the links at the end.

Since you asked people to open issues, here I am!

In the article you write about the host platform:

Since I’m on a Mac, and the purpose is to get a working Docker image, I set up an Ubuntu 20.04 Virtual Machine in Parallels. If you want to follow along, I recommend you to do the same and get onto a fresh Ubuntu 20.04 with QEMU...

Ok, you're running on a Mac, but what CPU architecture is in that Mac? Intel or ARM / Apple Silicon?

I think it ends up being relevant because I ran into the problem described here:

https://forums.raspberrypi.com/viewtopic.php?t=361643#p2169817

where a person replies with:

I have also seen this message, have no answer, but it works. Although very slow cause all is emulated, so not really interesting. KVM does anyhow not work on real ARM CPU with those RPi kernels, so it is more to show that it can be done, but not for normal use-cases. Use generic virt machine and default Debian distro kernel.

And I found that article because it did not work on an ARM64 CPU. In my case the description of the platform was so open-ended I said, let's try doing it on a Pi 4. I mean, emulating a Pi running Bookworm on a Pi running Bookworm should work? Apparently, not.

So, overall, I think the article would be better if it specified which kind of Mac you used, and if you expect it to work on platforms running ARM CPUs especially Apples or Pis.

And a minor nit, but sudo umount /mnt/image should happen somewhere once you're done with the mounted image, since that really forces all pending writes to hit the disk and it cleans up the namespace for next time.

And a super-nit: Maybe go more into why mount doesn't work inside Docker whereas mtools does? Or if reference states why, put a link to the reference in that section of the article.

Again, this is all small beer. Overall the article is very useful and very well written!

Regards,
Dave, N1AI

You might want to update item #20 in your reference links. The link you have is okay but probably obsolete. There is a new plugin that is based on the one you reference that is much more specific/advanced for Cortex devices.

https://marketplace.visualstudio.com/items?itemName=marus25.cortex-debug

I am one of the maintainers.

Hi,

Thanks for creating a great community around embedded knowledge. I've learnt a lot reading the blogs 👍

As a future topic idea, I wonder if we can explore initialisation patterns on embedded systems? A frequent cause of bugs in embedded projects that I've worked on is getting the initialisation order wrong which results in subtle and/or sometimes catastrophic bugs. Most initialisation glue logic I've seen have generally been gigantic messes that no one wants to touch.

What techniques/patterns has this community used in the past?

Thanks and look forward to a good discussion.

I've cloned repo and tried to build Renode example in /example/renode with make command and the following compiler warning occured:

renode-example.c:53:5: warning: no previous prototype for '_write' [-Wmissing-prototypes]
   53 | int _write(int file, char *ptr, int len)
      |     ^~~~~~

To fix the warning I've added function prototype

int _write(int file, char *ptr, int len);

to the top of the renode-example.c file.

I can add pull req to resolve this issue.

INFO:
Compiler version arm-none-eabi-gcc (15:9-2019-q4-0ubuntu1) 9.2.1 20191025 (release) [ARM/arm-9-branch revision 277599]

The section 1. Configuration Reset Behavior https://interrupt.memfault.com/blog/firmware-watchdog-best-practices#1-configuration-reset-behavior has a snip from the datasheet but it seems to be the wrong section.

This same image is used again (different URL though) later in the section 4. RESETREAS https://interrupt.memfault.com/blog/firmware-watchdog-best-practices#4-resetreas

Maybe it's just my monitor or my eyes, but I have a really hard time distinguishing links from plain bold text, e.g.:

I love the look of the blog overall, but IMO a slightly lighter blue would be helpful for link text.

Consider adding 'topic tags' to the posts. This way, when the number of posts increases it will be easier to read/find all the posts related to a particular topic.

P.S. I really enjoy reading the blog. Good job and keep going. :)

The "This guide" link appears to be missing the URL (via):

In the following file in line 507 (first paragraph of the variable section) the last sentence is not complete and thus makes no sense:

...It turns out, those variable came

See also the rendered version:
https://interrupt.memfault.com/blog/how-to-write-linker-scripts-for-firmware#variables

Hi, I find this blog by accident and learn a lot from here. When I try to compile the example demo of breakpoint and debugmon I get a error message at link stage.The error information is here:

Linking library
/usr/lib/gcc/arm-none-eabi/9.2.1/../../../arm-none-eabi/bin/ld: /usr/lib/gcc/arm-none-eabi/9.2.1/../../../arm-none-eabi/lib/thumb/v7e-m+fp/hard/libc.a(lib_a-abort.o): in function `abort':
/build/newlib-CVVEyx/newlib-3.3.0/build/arm-none-eabi/thumb/v7e-m+fp/hard/newlib/libc/stdlib/../../../../../../../../newlib/libc/stdlib/abort.c:59: undefined reference to `_exit'

My platform information:

kernel version: 5.8.0-28
gcc version: arm-none-eabi-gcc(15:9-2019-q4-0ubuntu1) 9.2.1 20191025(release)

After some search I found that is because these demo use some function from libc by include head file like stdio.h, and the lack symbol like _exit is system call which the function in libc invoke. But for these demo there are no need to use a operating system so there are no system call at all. I found the best way to fix it is add a --sepecs=nano.specs at the link flag and it work at my platform. I also make some change to Makefile to let them more useful.

I'm getting

Error response from daemon: pull access denied for memfault/interrupt, repository does not exist

and I can't see any image on dockerhub?

I have managed to build the Docker image locally with issue.

Thanks for writing up your how you got the Raspberry Pi Os to run in Docker

Is there any way to increase the size of the root filesystem? It's got 600Mb free right now, and I need more space for the task I'm working on.

You already seem to be increasing it to 8Gb, but it's not translating to more space for the mounted filesystem. Any suggestions would be welcome.