Re: [patch 2/2] bootmem: Node-setup agnostic free_bootmem()

[Johannes Weiner <hannes@saeurebad.de>]

Hi,

"Yinghai Lu" <yhlu.kernel@gmail.com> writes:

> On Tue, Apr 15, 2008 at 12:51 PM, Johannes Weiner <hannes@saeurebad.de> wrote:
>> Hi,
>>
>>
>>  "Yinghai Lu" <yhlu.kernel@gmail.com> writes:
>>
>>  > On Tue, Apr 15, 2008 at 4:53 AM, Johannes Weiner <hannes@saeurebad.de> wrote:
>>  >>
>>  >> Andi Kleen <andi@firstfloor.org> writes:
>>  >>
>>  >>  > Andrew Morton wrote:
>>  >>  >> On Sun, 13 Apr 2008 18:56:57 +0200 Andi Kleen <andi@firstfloor.org> wrote:
>>  >>  >>
>>  >>  >>> Johannes Weiner <hannes@saeurebad.de> writes:
>>  >>  >>>
>>  >>  >>>> Make free_bootmem() look up the node holding the specified address
>>  >>  >>>> range which lets it work transparently on single-node and multi-node
>>  >>  >>>> configurations.
>>  >>  >>> Acked-by: Andi Kleen <andi@firstfloor.org>
>>  >>  >>>
>>  >>  >>> This is far better than the original change it replaces and which
>>  >>  >>> I also objected to in review.
>>  >>  >>>
>>  >>  >>
>>  >>  >> So...  do we think these two patches are sufficiently safe and important for
>>  >>  >> 2.6.25?
>>  >>  >
>>  >>  > It's only strictly needed for .26 I think for some (also slightly
>>  >>  > dubious) changes queued in git-x86.
>>  >>
>>  >>  Does anything yet rely on this new free_bootmem() behaviour?  If not,
>>  >>  the safest thing would be to just revert the original patch in mainline
>>  >>  and drop the second patch completely.
>>  >
>>  > 1. free_bootmem(ramdisk_image, ramdisk_size) in setup_arch of x86_64
>>  > need that
>>  > 2. another patch in x86.git need that.
>>
>>  Ok, to avoid confusion: we are talking about free_bootmem() iterating
>>  over nodes and looking up an area WITHIN a node or free_bootmem()
>>  freeing an area ACROSS nodes?
>>
>>  The first is what my patch does _only_.
>
> Yes, your patch for free_bootmem only can free blocks in the same
> node.

Yep.

> but the free_bootmem(ramdisk_image,...) in setup_arch could cross
> node... , and some other via reserve_early...
>
> for example two nodes, every node have 2G, and in case use
> memmap=NN$SS to execlude some memory on node1. the ramdisk could sit
> cross the boundary.

Now it gets clear.  Alright, then my patches should be dropped and I'll
whip something up for the 2.6.26 merge window.

	Hannes