[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Questions about the Real Time Clock

On Friday, 22 Feb, Donald R. Spoon wrote:
> out.  But WHY is it still trying to load something with modprobe?  The 
> complaints happen right after the hwclock is accessed on boot-up, so I 
> suspect it is connected to whatever the hwclock script is doing.  I have 

Exactly; see below for explanations.

> managed to dream up the following questions for the "gurus":
> 1.  Are the "hardware clock" and the "real time clock" the same or are 
> they different devices in the BIOS?
They are the same physical device.  None of them is "in the BIOS,"
whatever this may mean.

> 2.  If they are the same, then does removing RTC support from the kernel 
> keep the system from accessing the hardware clock?
Sometimes.  There are two methods of accessing the hardware clock: via
/dev/rtc and by directly messing with I/O ports.  The first one is
reliable, the second one is not: some Alpha systems exhibit incorrect
behaviour when hardware is accessed directly from user mode code.  This is
why hwclock tries /dev/rtc first.

> 3.  If different, is RTC support needed in anyway to run the NTPDATE or 
> NTP programs?
No.  (I know this answer is redundant, because they are not different.)

> 4.  If the RTC is not needed and I have support turned off in the 
> kernel, can I just remove the /dev/rtc entry and make these complaints 
> go away without hurting anyting else on the system?
There are more gentle ways of doing this:
(a) there may be flags to hwclock that prevent it from trying /dev/rtc; or
(b) you can add `alias char-major-10-135 off' to a file in /etc/modutils
(say, /etc/modutils/aliases) and run update-modules.
Removing /dev/rtc should also work (at least for a while).

> that function.  The "current" OS is Debian "Woody", and the kernel is 
> 2.2.20-generic + 256 Megs of RAM + Matrox Millenium Video card w/ 8 megs.

2.2.x kernels were known for doing silly things in their RTC driver;
don't know if it still holds for the recent ones such as 2.2.20.  For
2.2.x, a "light-weight" version of the driver exists as a patch.  The
light-weight driver only does what is needed for hwclock, disabling
advanced functionality such as programmable interrupts for user programs.
The RTC driver in 2.4.x is already light-weight.


Reply to: