[Self-introduction] - Paulo Almeida

[Self-introduction] - Paulo Almeida

[Paulo Miguel Almeida]

Hi everyone,

My name is Paulo Almeida and as per the instructions listed on the KSPP
page, this is my self-introduction email :)

I will keep it short. 

- My background is in HPC and AI 
- I've been writing software for around 20 years now
- I've written my x86-64 hobbyist OS for fun and in my spare time I've
  writing a MOS 6502 emulator for the same reason.
- Contributing to KSPP is going to be a side project of mine that I plan
  to do outside of business hours... so expect a dedication of a few
  hours per week.

Q: What topics are you interested in?
A: kernel driver development, x86 & ARM hardware architecture, Math, Data
structures, Rust and virtualisation.

Q: What do you want to learn about?
A: I see the KSPP project/initiative as a way to get exposed to pieces
of code that I wouldn't normally come across which is always
appreciated :)

I am also aware of the calibre of developers I will be dealing with and
I'm sure that I will be learning really a lot from them :)

Q: What experience do you have with security, the kernel, programming, 
	or anything else you think is important.
A: 
I've contributed to the kernel a few times time in the past for both
adding features and janitorial tasks.

I took the Linux Kernel Internals (LF420) and the Linux Kernel Debugging
and Security (LF44) courses by the Linux Foundation.

As for other experiences, due to the fact that I wrote my hobbyist OS, I
do have a decent experience with the x86/x86-64 architecture. I also
spent quite sometime writing static analysis parsers.... so should those
experiences help anyone or any possible future plan for the KSPP, please
count on me.

Thanks!

Paulo Almeida

Re: [Self-introduction] - Paulo Almeida

[Kees Cook]

On Sun, Oct 09, 2022 at 07:32:38PM +1300, Paulo Miguel Almeida wrote:
> My name is Paulo Almeida and as per the instructions listed on the KSPP
> page, this is my self-introduction email :)

Hello! Welcome to the circus. :)

> I will keep it short. 
> 
> - My background is in HPC and AI 
> - I've been writing software for around 20 years now
> - I've written my x86-64 hobbyist OS for fun and in my spare time I've
>   writing a MOS 6502 emulator for the same reason.

Heh, nice. That made me wonder if there was a QEmu port, but it seems
it hasn't been touched in a decade?
https://github.com/AVEx-6502/qemu-6502

Is there a particular 6502 system you're working to emulate?

> - Contributing to KSPP is going to be a side project of mine that I plan
>   to do outside of business hours... so expect a dedication of a few
>   hours per week.

Excellent! We're always glad to have folks helping.

> Q: What topics are you interested in?
> A: kernel driver development, x86 & ARM hardware architecture, Math, Data
> structures, Rust and virtualisation.
> 
> Q: What do you want to learn about?
> A: I see the KSPP project/initiative as a way to get exposed to pieces
> of code that I wouldn't normally come across which is always
> appreciated :)

Yeah, you'll end up uncovering some really weird stuff, as you seem to
have already found[1]. ;)

> I am also aware of the calibre of developers I will be dealing with and
> I'm sure that I will be learning really a lot from them :)

One of the double-edged swords with KSPP is while you get to work with
all the kernel's subsystem maintainers, you also have to work with all
the kernel's subsystem maintainers. ;) You'll learn a lot, but you can
also end up discovering very different requirements as you send patches.

> Q: What experience do you have with security, the kernel, programming, 
> 	or anything else you think is important.
> A: 
> I've contributed to the kernel a few times time in the past for both
> adding features and janitorial tasks.

Excellent! Yeah, it looks like you've been helping clean up some staging
drivers[2].

I see you touched rtl8192u -- there's a lot of duplicate code spread
around in the rtl* drivers. It would be interesting to see those
consolidated some day. If you're interested, it may be worth asking
about it on the wireless list.

> I took the Linux Kernel Internals (LF420) and the Linux Kernel Debugging
> and Security (LF44) courses by the Linux Foundation.

Cool -- did anything stand out for you in those courses?

> As for other experiences, due to the fact that I wrote my hobbyist OS, I
> do have a decent experience with the x86/x86-64 architecture. I also
> spent quite sometime writing static analysis parsers.... so should those
> experiences help anyone or any possible future plan for the KSPP, please
> count on me.

Great! One area that needs some review and testing that is x86-specific
is the userspace CET support[3]. That spans a wide range of from chipset
all the way up through compiler, kernel, and glibc. Getting more people
to try that series out and post results ("it works for me" or "I
couldn't trigger the protection", etc) would be very welcome.

You've already found the "remove the 1-element arrays" work, and there
are plenty more like that on the issue tracker. Trying to really put an
end to strlcpy[4] is ongoing[5] too, as there has been a fairly
concerted effort to remove them lately:

Count of "git grep strlcpy | wc -l" over recent releases:

         v5.17:  1535
         v5.18:  1525
         v5.19:  1507
          v6.0:  1379
        master:   544
 next-20221010:   401

Thanks for introducing yourself!

-Kees

[1] https://lore.kernel.org/linux-hardening/Y0IsXXYnS4DnWkMW@mail.google.com/
[2] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/log/?qt=author&q=Paulo+Miguel+Almeida
[3] https://lore.kernel.org/lkml/Y0IsXXYnS4DnWkMW@mail.google.com/
[4] https://github.com/KSPP/linux/issues/89
[5] https://lore.kernel.org/lkml/?q=s%3A%22use+strscpy%22
    https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/log/?qt=grep&q=strlcpy

-- 
Kees Cook

Re: [Self-introduction] - Paulo Almeida

[Paulo Miguel Almeida]

On Mon, Oct 10, 2022 at 04:22:48PM -0700, Kees Cook wrote:
> On Sun, Oct 09, 2022 at 07:32:38PM +1300, Paulo Miguel Almeida wrote:
> > My name is Paulo Almeida and as per the instructions listed on the KSPP
> > page, this is my self-introduction email :)
> 
> Hello! Welcome to the circus. :)
> 
Thanks =)

> >   writing a MOS 6502 emulator for the same reason.
> 
> Heh, nice. That made me wonder if there was a QEmu port, but it seems
> it hasn't been touched in a decade?
> https://github.com/AVEx-6502/qemu-6502
> 
> Is there a particular 6502 system you're working to emulate?
> 
Yes, there is. I am trying to emulate Ricoh 2A03 (RP2A03). That's the
slightly modified 6502 CPU used in the NES system. It's a really
well-documented platform so I chose that one to dip my toes in writing
virtualisation-related software.

> > I took the Linux Kernel Internals (LF420) and the Linux Kernel Debugging
> > and Security (LF44) courses by the Linux Foundation.
> 
> Cool -- did anything stand out for you in those courses?
> 
I'm easily amused by debugging/instrumentation techniques and the
kernel has so many of them for various specific niches.

What really stood out at the time was learning how versatile ftrace can
be to the point that this "debugging/tracing" utility is one of the main
gears of kernel live patching. Initially it felt wrong but after
wrapping my head around it, that was a very creative solution.

> > As for other experiences, due to the fact that I wrote my hobbyist OS, I
> > do have a decent experience with the x86/x86-64 architecture. I also
> > spent quite sometime writing static analysis parsers.... so should those
> > experiences help anyone or any possible future plan for the KSPP, please
> > count on me.
> 
> Great! One area that needs some review and testing that is x86-specific
> is the userspace CET support[3]. That spans a wide range of from chipset
> all the way up through compiler, kernel, and glibc. Getting more people
> to try that series out and post results ("it works for me" or "I
> couldn't trigger the protection", etc) would be very welcome.
> 
That sounds like it's my cup of tea :) I just finished reading the Intel
CET paper, it's an interesting approach with lots of things to be
tested. Thanks for the recommendation, I will get involved with that. 

> You've already found the "remove the 1-element arrays" work, and there
> are plenty more like that on the issue tracker. Trying to really put an
> end to strlcpy[4] is ongoing[5] too, as there has been a fairly
> concerted effort to remove them lately:
> 
> Count of "git grep strlcpy | wc -l" over recent releases:
> 
>          v5.17:  1535
>          v5.18:  1525
>          v5.19:  1507
>           v6.0:  1379
>         master:   544
>  next-20221010:   401
> 
Fingers crossed! I saw many patches for the strlcpy replacement so you
seem to be well covered for that... I will continue with the one-element
array changes and try to get involved with the support for Intel CET =)

Thanks for all the suggestions Kees!

Paulo A.