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

Re: where, besides raspian can I find a full armhf installer that works on an rpi4b?



On Tuesday, January 25, 2022 2:41:22 PM EST gene heskett wrote:
> On Tuesday, January 25, 2022 1:26:39 PM EST Andrew M.A. Cater wrote:
> > On Tue, Jan 25, 2022 at 06:17:16PM +0100, LinAdmin wrote:
> > > When on 16 Mar 2021 I gave the solution in
> > > https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=981586
> > > 
> > > I was told from the omnipotent gurus of Debian that nobody needs 32
> > > Bit and the bug was closed.
> > > 
> > > After that, everybody requiring 32 Bit went with Ubuntu 20.4 LTS
> > > which is not as stubborn as Debian.
> > 
> > No - not quite. That's not what the bug history says. You will find
> > that Ubuntu isn't supported on all Pi models, for example, but only
> > on
> > later models that are ARMv7. In general, people are opting for 64 bit
> > for Pi 3 and Pi 4 - see also Fedora.
> > 
> > If you actually talk to the people doing the work, you often find the
> > reasons why - they're not all "omnipotent gurus" and they're
> > generally
> > approachable.
> > 
> > As it stands today, we can't support the Raspberry Pi (at any
> > version)
> > with an official Debian installer because of the method of booting
> > and
> > the need for non-free firmware. There's a good port for UEFI - and
> > that's what Fedora is also using, for example - but the need for
> > non-free firmware persists.
> > 
> > > LinAdmin
> > > 
> > > On 24.01.22 02:23, gene heskett wrote:
> > > > I thought I had some cornered earlier today, but when written to
> > > > u-sd
> > > > and booted, were arm64.
> > > > 
> > > > For low latency reasons when a realtime kernel is installed it
> > > > must
> > > > be
> > > > for armhf.
> > > > 
> > > > Cheers, Gene Heskett.
> > 
> > ?? Why must it be armhf - a Pi 4 is massively faster/more capable
> > than
> > a Pi 1 ?? Could you explain what difference 32 bitness makes?
> 
> Simple Andy, and has been explained many times. The bigger stack frame
> of arm64 negatively impacts the IRQ latency, making the response to an
> IRQ take several microseconds longer. Running a fully preempt-rt
> kernel on an i5 can get that time down to 4 u-secs unless you're
> running something that needs the nvidia proprietary video driver,
> which can tie things up with the irq's locked out for 3+ milliseconds.
> That's a full showstopper.
> 
> On armhf, on a 2gig rpi4b, a kernel I built,
> 
> 4.19.71-rt24-v7l+ #1 SMP PREEMPT RT Thu Feb 6 07:09:18 EST 2020 armv7l
> GNU/Linux,
> 
> almost 2 years ago makes about 12 microseconds which is fast enough to
> run over half a ton of 75 yo sheldon lathe I rebuilt for cnc control.
> The Linuxcnc latency-test won't even run on an arm64 because the
> realtime irq capability isn't there, and when measured by other means
> can be 100 microseconds worse. That's valuable time lost that will
> manifest itself in machine crashes that break expensive tooling. Even
> the half your little fingernail sized carbide chips that tip our
> cutting tools cost $20 or more a copy if we buy the better stuff.
> 
> The guys that inhabit the linux-rt list are putting a lot of what
> they've learned back into the regular linux kernel, with the future
> target being capable of doing this reatime work, but its not all done
> yet. And its a heck of a lot faster now than it was at 2.2 because of
> this, despite the line count being multiplied by 4 or more since then.
> 
> It is Linus's stated target that this happens.
> 
> But there is a lot of work yet to be done too, lots of legacy code that
> needs re-written to take advantage of what has been learned.
> 
> We are the folks who if we want music while we work, will find a radio
> and turn it on.  Its a different environment for sure but in the long
> view it will come together.
> 
> We are the unreasonable people G. B. Shaw wrote about when he said all
> progress is made by unreasonable people, reasonable folks adapt to the
> status quo and get on with it even when things are not optimal.
> 
> > All the very best, as ever,
> 
> To you too Andy, take care and stay well, all of you reading this to
> see what trash talk Gene is spewing today.
> 
> > Andy Cater
> 
> Cheers, Gene Heskett.

FWIW, I unpacked a fresh 64G u-sd card, wrote the 20211030 raspian 
bullseye armhf.img install to it. Then mounted it and edited 
/etc/dhcpdcp.conf to fill in the fallback data at the bottom of that file 
and wrote my hosts file over what was there, and edited hostname to name 
it.

Placed the card in my pi and powered it up, skipping the wifi discovery. 
It is now in the midst of installing about 130 updates since that release 
was spun.  So the fix is now published. When that gets done, I'll mount 
my work drive and see if I can install the newly built realtime kernel.

Cheers, Gene Heskett.
-- 
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author, 1940)
If we desire respect for the law, we must first make the law respectable.
 - Louis D. Brandeis
Genes Web page <http://geneslinuxbox.net:6309/gene>




Reply to: