[9front] build iso from arm64

[9front] build iso from arm64

[william]

Here are the exact commands I ran followed by the error at the end. I must be missing
a step or not understanding something. 

	cd /
	for(i in /sys/src/cmd/7[acl]){cd $i && mk install}

build remaining arm64 binaries

	cd /sys/src
	objtype=arm64 
	mk install

Building an ISO

Put your root file system into /n/src9

	bind /root /n/src9

Put your hg repository there

	bind -ac /dist/plan9front /n/src9

Build the iso

	cd /sys/lib/dist
	mk /tmp/9front.$objtype.iso

mk: don't know how to make '/tmp/9front.arm64.iso' in directory /sys/lib/dist

Re: [9front] build iso from arm64

[ori]

Quoth william@thinktankworkspaces.com:
> Here are the exact commands I ran followed by the error at the end. I must be missing
> a step or not understanding something. 
> 
> 	cd /
> 	for(i in /sys/src/cmd/7[acl]){cd $i && mk install}
> 
> build remaining arm64 binaries
> 
> 	cd /sys/src
> 	objtype=arm64 
> 	mk install
> 
> Building an ISO
> 
> Put your root file system into /n/src9
> 
> 	bind /root /n/src9
> 
> Put your hg repository there
> 
> 	bind -ac /dist/plan9front /n/src9
> 
> Build the iso
> 
> 	cd /sys/lib/dist
> 	mk /tmp/9front.$objtype.iso
> 
> mk: don't know how to make '/tmp/9front.arm64.iso' in directory /sys/lib/dist
> 
> 

I don't think we have any arm images that boot from an iso,
so we have no rules for building it.

pick a target that exists -- say, amd64 or 386.

read the mkfile to see what we can build.

Re: [9front] build iso from arm64

[william]

Thanks. I miss the obvious. 

I have another question. Looks like I need to add
an entry form arm64. But where is the kernel for arm64?

It can't be 9pc64 that is AMD, bcm is for pi i.e broadcom. I know 9pc64 is wrong but ran it anyways.

%.arm64.iso:
	@{
	objtype=arm64
	kernel=/n/src9/$objtype/9pc64
	mk $target.$pid.pc.iso
	mv $target.$pid.pc.iso $target
	}

/sys/lib/dist
cpu% mk /tmp/9front.$objtype.iso
@{
objtype=arm64
kernel=/n/src9/$objtype/9pc64
mk /tmp/9front.arm64.iso.60008.pc.iso
mv /tmp/9front.arm64.iso.60008.pc.iso /tmp/9front.arm64.iso
}
mk: don't know how to make '/n/src9/arm64/9pc64' in directory /sys/lib/dist
mk: @{ objtype=arm64 kernel=/n/src9/$objtype/9pc64 ...  : exit status=rc 60010: rc 60012: mk 60013: error

I went to /sys/src9/386 and kernel boot and a few other files exist for 386. I suspect I have
more leg work to do 

Is it /sys/src/9/kw or is it /sys/src/9/omap

I guess this is the next decision its an ARM Cortex-A53 

I guess I need set CONF and try running make in either of those directories. Then copy the kernel over.  



Quoth ori@eigenstate.org:
> Quoth william@thinktankworkspaces.com:
> > Here are the exact commands I ran followed by the error at the end. I must be missing
> > a step or not understanding something. 
> > 
> > 	cd /
> > 	for(i in /sys/src/cmd/7[acl]){cd $i && mk install}
> > 
> > build remaining arm64 binaries
> > 
> > 	cd /sys/src
> > 	objtype=arm64 
> > 	mk install
> > 
> > Building an ISO
> > 
> > Put your root file system into /n/src9
> > 
> > 	bind /root /n/src9
> > 
> > Put your hg repository there
> > 
> > 	bind -ac /dist/plan9front /n/src9
> > 
> > Build the iso
> > 
> > 	cd /sys/lib/dist
> > 	mk /tmp/9front.$objtype.iso
> > 
> > mk: don't know how to make '/tmp/9front.arm64.iso' in directory /sys/lib/dist
> > 
> > 
> 
> I don't think we have any arm images that boot from an iso,
> so we have no rules for building it.
> 
> pick a target that exists -- say, amd64 or 386.
> 
> read the mkfile to see what we can build.
> 
> 
> 
> 

Re: [9front] build iso from arm64

[ori]

Quoth william@thinktankworkspaces.com:
> Thanks. I miss the obvious. 
> 
> I have another question. Looks like I need to add
> an entry form arm64. But where is the kernel for arm64?
> 
> It can't be 9pc64 that is AMD, bcm is for pi i.e broadcom. I know 9pc64 is wrong but ran it anyways.

arm isn't a system; it's a processor type. bcm64 is the only
system that runs on arm64 that's currently supported; if you
want to run it on another arm64 processor, you need to write
code -- mostly drivers.

Re: [9front] build iso from arm64

[ori]

Quoth william@thinktankworkspaces.com:
> I have another question. Looks like I need to add
> an entry form arm64. But where is the kernel for arm64?

To put it another way: every arm board is different,
wtih different busses, drivers, peripherals, etc.

We don't have one kernel for all of them. Each
image targets one specific arm system.

Arm is a pain that way.

Re: [9front] build iso from arm64

[mkf9]

May 7, 2022 08:34:11 ori@eigenstate.org:

> Quoth william@thinktankworkspaces.com:
>> I have another question. Looks like I need to add
>> an entry form arm64. But where is the kernel for arm64?
> 
> To put it another way: every arm board is different,
> wtih different busses, drivers, peripherals, etc.
> 
> We don't have one kernel for all of them. Each
> image targets one specific arm system.
> 
> Arm is a pain that way.
Is it a problem with arm itself,
or we don't have code to have
a unified arm kernel (bcm+omap+etc) yet?

Re: [9front] build iso from arm64

[Kurt H Maier]

On Sat, May 07, 2022 at 04:36:29AM +0000, mkf9 wrote:
> Is it a problem with arm itself,
> or we don't have code to have
> a unified arm kernel (bcm+omap+etc) yet?

It's a problem with ARM itself.  Everyone pays ARM for processor designs
and then bolts random shit to it.  Since the random shit is not subject
to any standardization effort, every device is a special snowflake.

khm

Re: [9front] build iso from arm64

[william]

Quoth Kurt H Maier <khm@sciops.net>:
> On Sat, May 07, 2022 at 04:36:29AM +0000, mkf9 wrote:
> > Is it a problem with arm itself,
> > or we don't have code to have
> > a unified arm kernel (bcm+omap+etc) yet?
> 
> It's a problem with ARM itself.  Everyone pays ARM for processor designs
> and then bolts random shit to it.  Since the random shit is not subject
> to any standardization effort, every device is a special snowflake.
> 
> khm
> 

Good to know. I'm glad I didn't waste too much time and getting further lost. 

Basically I need to pour through the different arm repo's and puzzle what I can for
NXP/Freescale i.MX8MQ with 4x ARM Cortex-A53 cores (1.5 GHz), 1x Cortex-M4F core.

not familiar with paqfs. So that was interesting but I could comment all that out. 

This is an MNT reform laptop. After putting it together it seems pretty stable and I'm interest
in porting plan9 to this thing. I expect I might need to dig into the ARM® Cortex®-A53 MPCore Processor
Revision: r0p3 and try to figure this thing out. 

But before I do this has anyone made any headway in this area? I'm enjoying the laptop. The keyboard
is a bit funky the mapping is different but all in all its enjoyable laptop and a bit sturdy. 

Recommendations greatly welcomed.

Regards, 
-Will

Re: [9front] build iso from arm64

[Stanley Lieber]

[-- Attachment #1: Type: text/plain, Size: 1686 bytes --]

a few of us have these things. cinap has it booting to userspace over serial connection, but nobody has graphics working yet. i don’t think anything has been committed yet.

http://mnt.stanleylieber.com/reform/

sl

> On May 7, 2022, at 1:30 AM, william@thinktankworkspaces.com wrote:
> 
> Quoth Kurt H Maier <khm@sciops.net>:
>>> On Sat, May 07, 2022 at 04:36:29AM +0000, mkf9 wrote:
>>> Is it a problem with arm itself,
>>> or we don't have code to have
>>> a unified arm kernel (bcm+omap+etc) yet?
>> 
>> It's a problem with ARM itself.  Everyone pays ARM for processor designs
>> and then bolts random shit to it.  Since the random shit is not subject
>> to any standardization effort, every device is a special snowflake.
>> 
>> khm
>> 
> 
> Good to know. I'm glad I didn't waste too much time and getting further lost. 
> 
> Basically I need to pour through the different arm repo's and puzzle what I can for
> NXP/Freescale i.MX8MQ with 4x ARM Cortex-A53 cores (1.5 GHz), 1x Cortex-M4F core.
> 
> not familiar with paqfs. So that was interesting but I could comment all that out. 
> 
> This is an MNT reform laptop. After putting it together it seems pretty stable and I'm interest
> in porting plan9 to this thing. I expect I might need to dig into the ARM® Cortex®-A53 MPCore Processor
> Revision: r0p3 and try to figure this thing out. 
> 
> But before I do this has anyone made any headway in this area? I'm enjoying the laptop. The keyboard
> is a bit funky the mapping is different but all in all its enjoyable laptop and a bit sturdy. 
> 
> Recommendations greatly welcomed.
> 
> Regards, 
> -Will
> 
> 
> 
> 
> 
> 
> 

[-- Attachment #2: Type: text/html, Size: 3109 bytes --]

Re: [9front] build iso from arm64

[Thaddeus Woskowiak]

I dont think it's fair to blame Arm as no one bothered to build a
standard Arm platform in the same vein as the PC. Before the IBM PC XT
and later AT standard existed the home computer industry was in a
similar situation with random intel, wd, zilog, motorola systems
kicking about with incompatibilities between them. IBM was very
influential and relevant at the time so it stands to reason that the
PC was in a unique position to take off. Who is going to IBM the Arm
platform? Microsoft? Amazon? Google? Facebook? They're too busy
building ant farms.

The PC platform is an agreement between manufacturers to use a
standard boot firmware (bios/uefi), standard hardware interfaces (e.g.
pci, ata, vesa, etc, ac97/hda), and booting procedures (mbr/gpt).
Hardware is wired predictably so video, keyboard, com ports, timers,
etc are always on the same addresses, interrupt and dma ports meaning
drivers don't have to be tweaked or re-written between vendors.
Nowadays we have such complex (batshit even) hardware that we need to
manage it via a subsystem which we know as acpi that adds a whole
layer of abstraction to talk to fans and firmwares.

The closest we have had is the Microsoft Windows RT Surface which
bolted wintel PC baggage to Arm such as uEFI and ACPI. Basically an
Arm PC. Raspberry pi is an honorable mention by fitting their firmware
with pc like booting from usb, pxe and so on but pi firmware isn't
portable as the pi hardware is batshit wherein the gpu handles the
bootstrapping, not the Arm cpus....

It would be nice to have a platform controller, firmware and api using
9p so we can mount and walk the platform controller on boot then find
and configure what we need. Though, for anything to succeed, it would
need the weight of an IBM behind it.

Re: [9front] build iso from arm64

[Kurt H Maier]

On Sat, May 07, 2022 at 11:14:37AM -0400, Thaddeus Woskowiak wrote:
> Who is going to IBM the Arm platform? Microsoft? Amazon? Google? 
> Facebook?

Google could easily drive the entire ARM SoC market themselves, without
question, just by making requirements in Android. In fact the ARM/Google
relationship would nicely parallel the Intel/Microsoft reign of terror.

> The PC platform is an agreement between manufacturers to use a
> standard boot firmware (bios/uefi), standard hardware interfaces (e.g.
> pci, ata, vesa, etc, ac97/hda), and booting procedures (mbr/gpt).

No, it's not an agreement between manufacturers.  Every one of the
standards you mentioned came from Intel except for ATA.  If you want an
Intel chip in your computer, you play ball.  Fortunately it's easy to
play ball, because once you sign the NDAs Intel will happily do half the
work for you.

> The closest we have had is the Microsoft Windows RT Surface

The closest we've had with a chance at being a competitor is the Google
Chromebook specifications, but it's severely limiting in what hardware
it allows, which leads back to a free-for-all once you step outside the
spec.  Another could-have-been was the OLPC XO-4 ARM machines, which
shipped with a beautiful OpenFirmware build that supported all the
hardware natively, no ACPI required.

> Though, for anything to succeed, it would need the weight of an IBM 
> behind it.

No, I think ARM could step up and offer meaningful reference designs,
and that's about all it would take.  But the world is committed to a
bespoke u-boot and device-tree per SKU, and most of the development
industry have Stockholmed themselves into believing this is for the
best.

khm