[9fans] (no subject)

[9fans] (no subject)

[kokamoto]

Subkject: A68-N5000 cabini motherboard

I got it functional partially.

This is for energy-saving motherboard with APU (15W TDP)
on it without coolong fan.

I just moved the HDD used on a C2D machine
to this box, where it booted successfully with only
one problem, ether chip.

The ether chip on this board is:
1.0.0: net 02.00.00 10ec/8168 ....,
and the error message from the kernel is:
rtl8169: unknown mac 8168 4c000000.

No, the ether8169.c file does not have that line.
However ether82598.c file does.
Then, I checked out the ether8169 line from /sys/src/9/pc/pcf,
and tried to recompile the kernel, and got a huge sized one:
size 9pcf
  1309468t + 6491264d + 375712b = 8176444.

I think the second size may be too big!

I recompiled the 386 kernel many times and had
no problem like this.
The compilation after the last kernel recompilation are
go(1.3?) and amd64 kernel.

Anyone has an idea what'i going on?

Kenji

Re: [9fans] AMD A68-N5000 kabini mother

[kokamoto]

I forgot most important line.
This is 9front disk which moved from old machine.

> Subkject: A68-N5000 cabini motherboard
>
> I got it functional partially.
>
> This is for energy-saving motherboard with APU (15W TDP)
> on it without coolong fan.
>
> I just moved the HDD used on a C2D machine
> to this box, where it booted successfully with only
> one problem, ether chip.
>
> The ether chip on this board is:
> 1.0.0: net 02.00.00 10ec/8168 ....,
> and the error message from the kernel is:
> rtl8169: unknown mac 8168 4c000000.
>
> No, the ether8169.c file does not have that line.
> However ether82598.c file does.
> Then, I checked out the ether8169 line from /sys/src/9/pc/pcf,
> and tried to recompile the kernel, and got a huge sized one:
> size 9pcf
>   1309468t + 6491264d + 375712b = 8176444.
>
> I think the second size may be too big!
>
> I recompiled the 386 kernel many times and had
> no problem like this.
> The compilation after the last kernel recompilation are
> go(1.3?) and amd64 kernel.
>
> Anyone has an idea what'i going on?
>
> Kenji

Re: [9fans] AMD A68-N5000 kabini mother

[cinap_lenrek]

> No, the ether8169.c file does not have that line.

the 9front driver has the line:

ether8169.c:1108: 			print("rtl8169: unknown mac %.4ux %.8ux\n", p->did, macv);

but also,

	Macv40		= 0x4c000000,	/* RTL8168G */

so it should recognize the mac. you might be using an
older version?

> Anyone has an idea what'i going on?

no, i dont understand what your problem is.

--
cinap

Re: [9fans] AMD A68-N5000 kabini mother

[kokamoto]

> but also,
>
> 	Macv40		= 0x4c000000,	/* RTL8168G */
>
> so it should recognize the mac. you might be using an
> older version?

Ok, I'll back the HDD drive to the C2D machine, and
make it newest, then, recompile the kernel on that
older machione.   Then, return back the HDD drive to
new UEFI bios kabini machine.

Thanks a lot.

Kenji

PS. the older kernel's size:
1318345t + 2275244d + 375712b = 3969301
new UEFI kabini machine:
1309468t + 6491264d + 375712b = 8176444

Re: [9fans] AMD A68-N5000 kabini mother

[cinap_lenrek]

8mb is quite big. the current pc kernel on 9front is:

1319438t + 2154020d + 377288b = 3850746	9pcf

as your data size increased, you'r apparently including
more stuff in the #/ or bootfs.paq. candidates could be
/lib/firmware, but even with the iwl firmware blobs the
kernel gets just arround 6MB, not 8MB.

the question is, what did you add?

--
cinap

Re: [9fans] AMD A68-N5000 kabini mother

[kokamoto]

> more stuff in the #/ or bootfs.paq. candidates could be
> /lib/firmware, but even with the iwl firmware blobs the
> kernel gets just arround 6MB, not 8MB.
>
> the question is, what did you add?

I did not add anything.
The size of bootfs.paq is very big about 5.7MB (57949162 bytes).
I downloaded most recent sources from your site, and compiled
them on the C2D machine (not A68-N5000), and got about
same result.

In addition, I also compiled pc64 stuff, and got the size of
9pc64:   986827t + 6281440d + 39440b = 7662707.
This seems to have too big data size?

This happens only to compile kernels.
When I compile everything else, not such happens.

Above test shows this ios not the problem of the new
machine.

Kenji

Re: [9fans] AMD A68-N5000 kabini mother

[cinap_lenrek]

> This happens only to compile kernels.
> When I compile everything else, not such happens.
what does this mean?

anyway, you might have added stuff by accident or
theres some temporary files or some stuff that
unintentionally got included (you added /bin/tr there
remember for the fstype kfs detection). lets verify this.
please run:

cd /boot
du -n

paqfs /boot/bootfs.paq
cd /n/paq
du -n

theres the output of my bootfs as a reference:

cpu% cd /boot
cpu% du -n
29240	./boot
72285	./paqfs
245954	./factotum
1535020	./bootfs.paq
1882499	.

cpu% paqfs /boot/bootfs.paq
bootfs: Mon Jul 28 18:45:56 GMT 2014
fingerprint: ce48b5387fa103247e64c81531fd49f129526c7a
cpu% cd /n/paq
cpu% du -n
70210	./386/bin/9660srv
161885	./386/bin/awk
140892	./386/bin/auth/secstore
140892	./386/bin/auth
129072	./386/bin/aux/kbdfs
116502	./386/bin/aux/wpa
245574	./386/bin/aux
37172	./386/bin/bind
22049	./386/bin/cat
86647	./386/bin/cfs
72803	./386/bin/cryptsetup
28505	./386/bin/dd
69915	./386/bin/disk/fdisk
176625	./386/bin/disk/kfs
64404	./386/bin/disk/prep
310944	./386/bin/disk
93065	./386/bin/dossrv
35819	./386/bin/echo
287215	./386/bin/cwfs64x
50252	./386/bin/grep
109040	./386/bin/ip/ipconfig
109040	./386/bin/ip
50768	./386/bin/ls
38042	./386/bin/mkdir
99796	./386/bin/mntgen
45615	./386/bin/mount
41324	./386/bin/mv
166241	./386/bin/hjfs
85738	./386/bin/rc
38300	./386/bin/rm
58248	./386/bin/sed
1395	./386/bin/sleep
51597	./386/bin/srv
43474	./386/bin/test
21840	./386/bin/unmount
126197	./386/bin/nusb/usbd
146797	./386/bin/nusb/ether
151362	./386/bin/nusb/disk
89388	./386/bin/nusb/kb
513744	./386/bin/nusb
40815	./386/bin/xd
3049009	./386/bin
3049009	./386
579	./rc/lib/rcmain
1139	./rc/lib/local.rc
793	./rc/lib/net.rc
2511	./rc/lib
1528	./rc/bin/fstype
1138	./rc/bin/diskparts
1412	./rc/bin/nusbrc
3884	./rc/bin/bootrc
7962	./rc/bin
10473	./rc
0	./tmp
0	./lib/firmware
0	./lib
3059482	.

--
cinap

Re: [9fans] AMD A68-N5000 kabini mother

[kokamoto]

Hi, cinap.
It was solved(?).
I removed all the /lib/firmware stuff, and recompiled
then the result is as usual!

size:
1319729t + 2282308d + 377258d = 3979325

My /lib/firmware directory had many iwn-xxxx files.
All those files are neccessary?

Anyway, now I can use this AMD machine, and now I'm
writing this mail from that machine.

As I was requested before to describe this motherboard,
I'll do it here.

Model: BIOSTAR A68-N5000
CPU: AMD A4-5000   Quad Core APU  1500MHz, 1.044V
		with AMD HD8210 Graphic
Chipset: A500A-IBS
BIOS: AMI UEFIx64 2.3

This kernel say as follows:

cpu0: lapic clock at 100MHz
cpu0: 1497MHz Authentic AMD unknown (AX 0x700F01
	CX 0x36D8220B DX 0x178BFBFF)
cpu1: ... same as cpu0
cpu1: ...same as cpu0
cpu2: ...same as cpu0
cpu2 ...same as cpu0
cpu3: ...same as cpu0
cpu3: ...same as cpu0

It's interesting to use energy-saving 4 core machine with
small cubic case(17Wx20Lx20H) with 300W power supply!
Of course without CPU fan!

Kenji

Re: [9fans] AMD A68-N5000 kabini mother

[kokamoto]

Addition:

My small case can have only one 3.5" HDD.
This case allows us to use two 2.5" HDD, however,
I don't like.  The motherboard itself has two SATA
drive connectors.  You can choose another mini-IPX case.
This motherboard has two USB 3.0 and two 2.0.
I have one more USB 2.0 socket on the front(side).
keyboard/mouse are for PS/2.

Speed?   It's very good to compile kernels or cmd/gs.

Kenji

Re: [9fans] AMD A68-N5000 kabini mother

[erik quanstrom]

> cpu% paqfs /boot/bootfs.paq
> bootfs: Mon Jul 28 18:45:56 GMT 2014
> fingerprint: ce48b5387fa103247e64c81531fd49f129526c7a
> cpu% cd /n/paq
> cpu% du -n
> 70210	./386/bin/9660srv
> 161885	./386/bin/awk
[...]
[...]
> 0	./lib
> 3059482	.

cinap, how would you assess the tradeoffs in the 9front
boot process?  on the one hand, it is largely managed by
an rc script, but on the other, the rc script is fixed, and
not loaded dynamicly (such as via tftp) and a large number
of executables are hauled into the kernel.

- erik

Re: [9fans] AMD A68-N5000 kabini mother

[cinap_lenrek]

the bootfs.paq with all the binaries is 1.5mb. and it
has support for all the fileservers and network booting.

we just reuse the rc scripts and binaries that are already
there and not trying to make up special solutions just for
booting to save a bit of memory. and having self contained
rc environment in the kernel is great as it lets you debug
the hardware before you have a working driver. (or you can
instruct someone to run some rc commands to troubleshoot a
system without him having the capability of building a
new kernel (because theres no driver yet)).

instead of having half broken smarts in the bootloader
(like passing half the partitions from loader to kernel)
which needs to duplicate all the drivers, we can just do
that with the kernels drivers and rc scrips. the kernel
doesnt care how it got loaded. it gets to the same environment
every time regardless of who loaded it. i can boot terminal
that mounts the root filesystem over wpa encrypted wifi
network and there are no hacks there. it just runs the same
programs early that you would normally run from termrc.

if you want to make a specialized low memory overhead
kernel that will only work on your machines then you still
can by specifying different files and /boot/boot
in your kernel configuration.

--
cinap