Re: oprofile comments

[graydon h. <gr...@re...>]

On Mon, 2002-11-25 at 13:40, Ulrich Drepper wrote:

> I was just puzzled by Graydon writing he's looking into the last P4 bug.

well, I meant things which cause crashes or data loss (which I still
have 2 of on p4, though I think I know why). HT-active chips are safely
detected and cause oprofile to drop into RTC mode for the time being; I
tend to think of such well-behaved gaps in functionality as "missing
features", but if you like we can call it a bug. it's not one I planned
on addressing this week in any case. I can if you think it's real
important.

> Missing HT support is a bug and from my perspective it's awhole lot
> more important to fix than adding support for this horrible PPC crap.
> oprofile would be used on thousand times more HT P4s than PPC machines.
...
> > The 2.5 kernels (2.5.42 and newer) have oprofile support in them.
> 
> A completely outdated version without any P4 support.  I have this code
> in my 2.5 kernels.
...
> Again, the official oprofile in 2.5 does not have P4 support.  It needs
> forward-porting from the current 2.4 sources and I haven't seen any
> progress on this front for weeks. If your plans are to backport the 2.5
> approach to adding oprofile in the kernel (beacuse of the syscall
> situation) then you first have to look into getting the 2.5 support updated.

yeah, but these are work-scheduling issues. the reason we're doing
things in the order we are is that we expect a 2.4 kernel to go out the
door before a 2.5/2.6 series, and at the moment merging work (such as p4
and hammer) from public 2.4-supporting oprofile to RH private 2.4 kernel
is painful. we want to be rid of that task. 

once we've backported the 2.5 stuff, merging in the p4, hammer and ppc
will be very nearly identical to the task you're proposing of forward
porting those drivers to 2.5 series, since the underlying kernel
arrangement will be identical. when the time comes to ship a 2.5 series
kernel, the forward-port should be little more than copying the files
from one directory to another.

if you really think this schedule is wrong though, say if you have a
product deadline you think we're going to miss or something, I'm sure
you can convince john to adjust the schedule. 

-graydon