Confirm? New local root exploits

[Philip Molter]

Dan Hollis wrote:
> On Fri, 7 Jan 2005, Philip Molter wrote:
> 
>>Note that if you upgrade, they also made a major change to way the 
>>kernel handles large memory sizes (the so-called 4g/4g patch was 
>>dropped).  It probably doesn't affect a university shell server, but it 
>>sure screwed me over.
> 
> 
> Specifics?

Specifically, I have a machine with 4GB of memory that requires a lot of 
memory listed as Buffers:  (upwards of 2.5GB).  This machine is very 
heavy on I/O and highly tuned for its design.  Not having the 4G/4G 
patch, Buffers: is limited to about 600MB, and the system goes to I/O hell.

I did recompile the kernel with a 2G/2G split rather than a 1G/3G and 
that helps (3G/1G is out of the question for me), but the previous 4G/4G 
setup was much better for my load, even if it did use about 5-10% more 
CPU (the box was, again, tuned for that extra CPU usage).

It still kind of irks me that such a major kernel change was made in a 
x.y.z_a.b security release, but one can't complain too much about 
software they don't pay for.  Most people aren't affected by the change, 
I suppose.  It would've been nice if the patch was left in, so people 
could recompile and enable it.

Philip