Re: Debian on mac68k (LC III)
On 11/3/16 4:50 PM, Finn Thain wrote:
>
> On Thu, 3 Nov 2016, Mac User wrote:
>
>> On 11/2/16 10:29 PM, Finn Thain wrote:
>>>
>>> On Wed, 2 Nov 2016, Mac User wrote:
>>>
>>>> On 11/2/16 8:20 PM, Finn Thain wrote:
>>>>
>>>> My mistake earlier -- I cut and pasted the hang from "version
>>>> 4.1.35-mac_scsi+" instead of capturing the result of
>>>> "vmlinux-4.1.35-mac_scsi-egret+" on a serial console.
>>>>
>>>
>>> Confusingly, I sent two kernel binaries with the same patch level. But
>>> you can tell them apart from the banner (also /proc/version):
>>>
>>> Linux version 4.1.35-mac_scsi-egret+ (fthain@nippy) (gcc version 4.6.3
>>> (GCC) ) #2 Tue Oct 25 13:37:53 AEDT 2016
>>>
>>> Linux version 4.1.35-mac_scsi-egret+ (fthain@nippy) (gcc version 4.6.3
>>> (GCC) ) #1 Wed Nov 2 13:09:32 AEDT 2016
>>>
>>> If I understand correctly, you've only observed two possible outcomes:
>>> either a hang shortly after "ABCFGHIJK" or else "kernel BUG at
>>> block/deadline-iosched.c".
>>>
>>> Does those outcomes apply to both of these kernel binaries, given the
>>> same Penguin settings etc?
>>>
>>
>> Almost.
>>
>> With 4.1.35-mac_scsi+, "kernel BUG at block/deadline-iosched.c:332!"
>> happens consistently when booting with extensions off and setting the
>> serial ports in Penguin. When booting with extensions on and not setting
>> the serial ports, there is a random hang (about one third of the time)
>> at "ABCFGHIJK".
>>
>
> It's most likely a spurious ethernet or serial interrupt that causes the
> intermittent hang. It's not actually a Linux bug, but an inadequacy of the
> Penguin bootloader. It's a known issue, so I don't need you to test
> Penguin configuration variations.
>
> The only Penguin settings that I'd have you use are the reliable ones,
> i.e. extentions off and serial ports pre-configured. Please keep the
> settings the same for every test, including your kernel parameters,
> root=/dev/sda6 console=ttyS0 earlyprintk
>
>> With 4.1.35-mac_scsi-egret+, even booting with extensions off, the
>> system hangs consistently at "ABCFGHIJK" regardless of Penguin settings.
>
> Can you confirm that this holds for both of the 4.1.35-mac_scsi-egret+
> builds: Oct 25 and Nov 2? The difference between them is a small .config
> change. I'd be interested to know whether they generate the same console
> output before they hang, or just "ABCFGHIJK".
>
> You can tell these binaries apart on the MacOS side by calculating the MD5
> digests using the md5fl utility from Paul C. Pratt.
> http://www.gryphel.com/c/minivmac/extras/index.html
Thanks for the md5fl link; I had been using FTP from NCSA Telnet 2.6
to transfer files, but one of the kernels didn't transfer properly,
so now I'll know to just use AppleTalk (and verify checksums).
>
> https://www.telegraphics.com.au/~fthain/stan/vmlinux-4.1.35-mac_scsi-egret+
> https://www.telegraphics.com.au/~fthain/stan/20161102/vmlinux-4.1.35-mac_scsi-egret+
>
> MD5 digests:
> 1e3917d2d1964efb8dff379cc13a79f8 vmlinux-4.1.35-mac_scsi-egret+
> cd56acd9e749ed9b4a07b44fa6747679 20161102/vmlinux-4.1.35-mac_scsi-egret+
>
> Thanks for your help.
>
Unfortunately, there was no change:
1) 10/25/2016 - "vmlinux-4.1.35-mac_scsi-egret+"
ABCFGHIJK
Linux version 4.1.35-mac_scsi-egret+ (fthain@nippy) (gcc version 4.6.3
(GCC) ) #2 Tue Oct 25 13:37:53 AEDT 2016
Saving 402 bytes of bootinfo
bootconsole [debug0] enabled
Detected Macintosh model: 27
VIA1 at 50f00000 is a 6522 or clone
VIA2 at 50f26000 is an RBV
2) 11/02/2016 - "vmlinux-4.1.35-mac_scsi-egret+"
ABCFGHIJK
Linux version 4.1.35-mac_scsi-egret+ (fthain@nippy) (gcc version 4.6.3
(GCC) ) #1 Wed Nov 2 13:09:32 AEDT 2016
Saving 402 bytes of bootinfo
bootconsole [debug0] enabled
Detected Macintosh model: 27
VIA1 at 50f00000 is a 6522 or clone
VIA2 at 50f26000 is an RBV
Let me know of anything else to try...
-Stan
Reply to: