Re: [PATCH] reinstate mac rtc

[Finn Thain <fthain@telegraphics.com.au>]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 3607 bytes --]



On Sun, 19 Oct 2008, Kolbjørn Barmen wrote:

> On Sun, 19 Oct 2008, Finn Thain wrote:
> 
> > On Sat, 18 Oct 2008, Kolbjørn Barmen wrote:
> > 
> > > On Sat, 18 Oct 2008, Riccardo wrote:
> > > 
> > > > I don't know "how", but with 2.2 kernels I always had a correct 
> > > > date at boot without any tricks. It is true though that with high 
> > > > CPU load we had clockskew and that we didn't save back the date 
> > > > and hourtime to the clock, thus any clock setting needed to be 
> > > > done from the mac side. A compromoise, but better than the current 
> > > > situation.
> > > 
> > > The clockproblem is something I stumble upon every now and then on 
> > > various machinees - I really wish there was a kernel parameter where 
> > > one could set a date string, then the bootloader could pass it on.
> > 
> > It isn't a problem if you disable the time-stamp triggered fsck and 
> > set the clock from the network (rdate or ntp).
> 
> This is the answer I keep getting, but it isnt really helping.

Works for me. Especially when I cannot source a clock battery. One must 
fabricate the 4.5 pack for the LC 630 and similar models (New Old Stock 
batteries don't last long enough to be worth the cost even if you can find 
them).

In fact, many of the several dozen machines I must test have no clock 
battery at all since even the 3.6 V 1/2 AA lithium cells cost me 
AUD$10-$15 each. Replacing the battery in a powerbook is difficult because 
it always means disassembly and, in some models, it means replacing a PCB 
mount cell. (But they need to be removed, I guess, since the old NiCd ones 
in all of my PB 150s were beginning to leak. I don't generally replace 
them, since that's inviting PMU flakiness when the go flat again).

BTW, I've seen old Li 1/2 AA cells leak too (but not the Li button 
cells... so far).

> * SSL/TLS and certificate validation. For example my gumtix uses 802.1X 
>   to get online, when it is running in "1970-01-01", the server and 
>   client certificates are ofcourse not valid and the authentication 
>   _will_ fail, and it wont go online, and therefor no ntp. As .1x is 
>   becoming more common even on the wire, this problem will grow, at 
>   least for me - and .1X is not the only "get online"-method that relies 
>   on certificate validation.
> 
> * Logs, erronous datestamps all over the place, and files made in 1970 
>   or whatever (my acer laptop always starts in 1988-01-01)

Right. The RTC is important. Perhaps you can configure eth0 using kernel 
parameters (like with NFS root) and then use eth0 to set the clock from 
userspace before filesystems are mounted r/w.

> There are so many kernel parameters for the strangest things, all I ask 
> for is one with a timestamp - then I could tell my gumstick to always 
> boot on a given time where the .1x certificates would be valid instead 
> of 1970-01-01. And on my laptop, I could just change a value in grub 
> before booting. On the mac, penguin could just take system time from 
> macos and pass it on, maybe even emile could do it.

I'd much prefer to see that effort spent on reverse engineering the VIA1 
accesses in question. This problem only applies to the Quadra 900/950 
AFAIK. None of the other several dozen models would see any benefit from 
changes to the mac bootloaders.

Finn

> 
> -- kolla
> --
> To unsubscribe from this list: send the line "unsubscribe linux-m68k" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>