In article <linux.kernel.AB4CB1CC6547D21197B00008C7F48FB402C10FE8@txexmta0.amd.com>,
<nathan.zook@amd.com> wrote:
>Concern: Some bios might hit random code which returns cf clear, ecx held
>constant, and edx copied to eax--the first few calls, after which they
>scribble on the HD (or worse).
>Solution: Aggressively test the registers (and data) for compliance. Test
>that buffer writes actually occur. Abort if anything is out of bounds.
And this is why the code has the paranoid #ifdef in it; Nathan is
looking at it from the viewpoint that The Computer Is Your Friend,
and I'm looking at it from the viewpoint that it's not worthwhile
to cater to the suicidal bios vendors.
I'm not willing to override him; I'm not sure what he thought, but
in any case I was running under the impression that the ifdeffed
code would satisfy both my desire for non-nonsense code and his
desire to work around the suicidal.
The Sunday patch does little more than change the existing e820 code
to return more data so that additional validations can be put in.
The detail that everyone who runs 2.3.x (x>18) isn't beating down
the doors to fix memory detection not working is, to me, a fairly
compelling reason to not make %es:%di checking mandatory. I'd
rather have the paranoid ifdef in the kernel side, but I'm willing
to defer that until later on.
____
david parsons \bi/ I test with paranoid on, but my production machines
\/ run with paranoid off.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Tue Feb 29 2000 - 21:00:21 EST