Re: lowmem_reserve (replaces protection)

[Andrea Arcangeli]

On Tue, Oct 26, 2004 at 09:00:11PM -0400, Rik van Riel wrote:
> However, doesn't the protection also serve a purpose for
> NUMA fallback ?  How is that handled by your code ?

It doesn't.

The only reason any NUMA folk were interested in this stuff was to nuke
the incremental min weak band-aid that Linus introduced in 2.4 while
rejecting my real lowmem_reserve algorithm (that has been reinvented in
2.6.5 first and against 2.6.7 with the name of protection but still
buggy and disabled by default, so not very useful.., especially after
even the incremental-min was gone from 2.6.8+).

NUMA folks only want to nuke the incremental-min hack since that
escalated to huge waste of ram on the hundred nodes.

the major point of the rework in the protection algorithm happened
between 2.6.5 and 2.6.8 was to drop any NUMA effect from the algorithm
that guarantees DMA allocations on x86-64 and NORMAL allocations on x86
PAE. So being it unrelated to NUMA is a feature as far as I can tell.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/