Same boot SSD: 8GByte PowerMac G5 Dual boots fine; 12GByte and 16GByte PowerMac G5 Quad cores usually hang...

Sat Sep 13 13:04:38 UTC 2014

I have kept at the testing and got an example failure without using startxfce4. Just shutdown -p now then power-on produced the failure on the PCI-X 8 GByte two single-core processor PowerMac G5. (So much for that G5 not having the problem. This does make the span include the radeon card context as well --all using modern 10.1-PRERELEASE builds.)

I'll periodically check more and report any more cases of failures that I come up with. May be the only consistent thing will end up being that powerpc/GENERIC never has the problem on any of the G5's.

===
Mark Millard
markmi at dsl-only.net

On Sep 13, 2014, at 4:26 AM, Mark Millard <markmi at dsl-only.net> wrote:

I should have also mentioned that I've never had any PowerMac G5 fail to boot from powerpc/GENERIC boot SSDs, only from powerpc64/GENERIC64 ones. For example:

FreeBSD FBSDG4S0 10.1-PRERELEASE FreeBSD 10.1-PRERELEASE #0 r271215: Sat Sep  6 23:56:15 PDT 2014     root at FBSDG4S0:/usr/obj/usr/src/sys/GENERIC  powerpc

always has worked. (This, of course, largely ignores most of the 8GByte+ of RAM.) This is true no matter what I've done after booting before trying to reboot (or shutdown -p now then power on), such as use startxfce4 (which is the way I normally start Xorg). (See below for why I mention startxfce4 as the example here.)

Now narrowing the context down to the problematical Quad Core PowerMac G5's where I frequently see the boot problem with powerpc64/GENERIC64...

But on thinking of this I realized something else: After such a powerpc/GENERIC boot of a Quad Core G5 PowerMac the first boot after switching to a powerpc64/GENERIC64 boot SSD also always worked. For example the fairly modern:

FreeBSD FBSDG5S1 10.1-PRERELEASE FreeBSD 10.1-PRERELEASE #0 r271243: Mon Sep  8 06:28:03 UTC 2014     root at releng1.nyi.freebsd.org:/usr/obj/powerpc.powerpc64/usr/src/sys/GENERIC64  powerpc

does in my experiments. And in experimenting I've discovered that I can then reboot using that powerpc64/GENERIC64 SSD over and over just fine --as long as I do not use startxfce4/Xorg at any point. (I've not tried other Xorg uses in any of my activities.)

But if I use startxfce4 then after that I have blank-screen/fans-spin-up problems for many/most later reboots (or shutdown -p now then power on) --until I go back and boot from the powerpc/GENERIC SSD.

This description holds for both the 12GByte Quad Core PowerMac G5 and the 16 GByte one. (Both have NVIDIA GeForce 7800 GT's, unfortunately. I currently do not have access to a Radeon for PCI-Express PowerMacs. Nor to some other NVIDIA card for them. No comparison/contrast material to try.)

(I'll keep monitoring for a violation of this pattern in my future use. But I've not noticed an violation of the pattern yet.)

This suggests some sort of recording that carries over context across reboots and even across power-off/power-on --a recording that is messed up after startxfce4 and cleaned up by powerpc/GENERIC use. Possibly problems configuring the GeForce 7800 GT cards if they can carry such a context across off/on.

It also tends to mean that just having the large amount of RAM is not sufficient context to initiate the boot problem.

The only other PowerMac G5 I've access to has Radeon video hardware, PCI-X/AGP instead of PCI-Express, and has 8 GBytes of RAM (its limit), single-core processors (2 of them). (I do not have access to a NVIDIA video board for this PowerMac G5.) This PowerMac G5 does not show the problem when I experiment with it, at least not with modern builds. I can use startxfce4 and I do not have later boot problems.

[I have updated my ports and have picked up the updated Xorg, drivers, and the change to lang/gcc being 4.8.3. The recent experiments have that new context involved.]

===
Mark Millard
markmi at dsl-only.net

On Sep 13, 2014, at 2:12 AM, Mark Millard <markmi at dsl-only.net> wrote:

Context: a boot SSD with

FreeBSD FBSDG5S1 10.1-PRERELEASE FreeBSD 10.1-PRERELEASE #0 r271243: Mon Sep  8 06:28:03 UTC 2014     root at releng1.nyi.freebsd.org:/usr/obj/powerpc.powerpc64/usr/src/sys/GENERIC64  powerpc

moved between PowerMacs. One PowerMac kind has no boot problems. The other kind frequently ends up hung with a blank screen just before the Copyright would normally show up: the start of the FreeBSD boot messages never show up. (The fans eventually speed up.)

The working PowerMac is the Dual Processor (single core each) PCI-X based one:

FreeBSD 10.1-PRERELEASE #0 r271243: Mon Sep  8 06:28:03 UTC 2014
    root at releng1.nyi.freebsd.org:/usr/obj/powerpc.powerpc64/usr/src/sys/GENERIC64 powerpc
gcc version 4.2.1 20070831 patched [FreeBSD]
VT: running with driver "ofwfb".
cpu0: IBM PowerPC 970 revision 2.2, 2000.23 MHz
cpu0: Features dc000000<PPC32,PPC64,ALTIVEC,FPU,MMU>
cpu0: HID0 511081<NAP,DPM,NHR,TBEN,ENATTN>
real memory  = 8569122816 (8172 MB)
avail memory = 8148094976 (7770 MB)
FreeBSD/SMP: Multiprocessor System Detected: 2 CPUs
cpu0: dev=ff887e10 (BSP)
cpu1: dev=ff889150

[   277.833] (--) RADEON(0): Chipset: "ATI Radeon 9800PRO NH (AGP)" (ChipID = 0x4e48)

An example of the frequently failing-boot kind of PowerMac Quad core G5 PCI-Express context is:

FreeBSD 10.1-PRERELEASE #0 r271278: Mon Sep  8 12:40:56 PDT 2014
    root at FBSDG5S0:/usr/obj/usr/src/sys/GENERIC64 powerpc
gcc version 4.2.1 20070831 patched [FreeBSD]
VT: running with driver "ofwfb".
cpu0: IBM PowerPC 970MP revision 1.1, 2500.34 MHz
cpu0: Features dc000000<PPC32,PPC64,ALTIVEC,FPU,MMU>
cpu0: HID0 1511081<DEEPNAP,NAP,DPM,NHR,TBEN,ENATTN>
real memory  = 17152716800 (16358 MB)
avail memory = 16374759424 (15616 MB)
FreeBSD/SMP: Multiprocessor System Detected: 4 CPUs
cpu0: dev=ff89d680 (BSP)
cpu1: dev=ff89eb70
cpu2: dev=ff89f248
cpu3: dev=ff89f920

(The other example I have access to has 12 GBytes of RAM instead. Both examples behave the same.)

[  3310.669] (--) NV(0): Chipset: "GeForce 7800 GT"

Some of the differences are (working context on left of "vs.", failing on right):

PowerPC 970 rev 2.2 vs. PowerPC 970MP rev 1.1
(I.e., two single-core processors vs. two dual-core processors)
8 GBytes RAM vs. 12 or 16 GBytes RAM (and slower vs. faster RAM)
ATI Radeon 9800PRO NH (AGP) vs. GeForce 7800 GT
clock rate: 2GHz vs. 2.5 GHz
PCI-X vs. PCI-Express
gem ethernet vs. bge ethernet
Whatever the OpenFirmware and such details are for the two kinds of PowerMacs.

Merely having more than 2 GBytes RAM or 4 GBytes RAM is not enough context for there to be a boot-hang problem.

===
Mark Millard
markmi at dsl-only.net