[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: [RFC] To port rv32 to Debian



Hi,
On Sat, May 20, 2023 at 12:16:44PM +0200, Aurelien Jarno wrote:
Hi,

On 2023-05-20 03:23, Jessica Clarke wrote:
On 20 May 2023, at 03:15, Bo YU <tsu.yubo@gmail.com> wrote:
>
> Hi,
>
> Maybe you've noticed a discussion about porting rv32 on this mailing
> list a few weeks ago. In the meantime, we,Gao Han<gaohan@iscas.ac.cn>
> and me, have almost finished rebootstrap's[0] support for riscv32,
> and already have its first users[1]. Given that, I hope to get more

That’s not riscv32, that’s riscv64ilp32, which is still cursed and
undesired in the context of Debian in my opinion. 32-bit architectures
are being killed off one-by-one, not being added.

I agree with that. I don't really see the point of riscv32 in the Debian
context. The 3/4GB limit is more and more an issue, even with less RAM
than that, as it prevent mmap on big files, or allocating big amount of
RAM to actually use it partially, a pattern seen more and more often.
From the security point of view, it also limits ASLR. In addition more
and more upstream packages also need porting as they do not care about
32-bit anymore and (admittedly wrongly) assuming the wrong type for the
variables. Therefore work to maintain such a port in a good condition is
going to be constantly increasing with time, with more and more
workarounds to be added in packages.

Choosing riscv64ilp32 looks even a more a wrong choice. Many attempts
have been done in Debian to support an ILP32 platform (arm64ilp32,
mipsn32, x32), and the porting work is even more important than a
standard 32-bit port. Only x32 have been able to somehow survive, and
still is just above 802% of the packages ported.

Okay, I think the above is enough to indicate that it is currently
inappropriate to port riscv32 to Debian. To put it simply, I thought there may be some users or vendors who will use the rootfs of riscv32.
Whether based on rv64ilp32 or pure 32 bit. If there is a less than
decent port here, I think it will probably be in < Tier 2.


Overall this also sends a bad signal for the riscv64 port, implying that
it is the first one of a long series to come, just like mips has done in
the past, or to a lesser extend arm.

Indeed. Porting rv64 is the highest priority and I am worried about the diversification of RISC-V will affect the overall judgment of rv64.

Anyway, I am very grateful to you and Jess for commenting out here at
least showing that this port is of no great benefit to Debian upstream.
Of course, for many other reasons, I(we) may maintain a basic Debian riscv32 rootfs locally, feel free to contact me if this helps you.:-)


Cheers
Aurelien

--
Aurelien Jarno                          GPG: 4096R/1DDD8C9B
aurelien@aurel32.net                     http://aurel32.net

--
Regards,
--
  Bo YU

Attachment: signature.asc
Description: PGP signature


Reply to: