login
Header Space

 
 

[patch 0/2] bootmem: Fix node-setup agnostic free_bootmem()

Score:
Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]
To: <linux-kernel@...>
Cc: Andi Kleen <ak@...>, Ingo Molnar <mingo@...>, Andrew Morton <akpm@...>, Yinghai Lu <yhlu.kernel@...>, Yasunori Goto <y-goto@...>, KAMEZAWA Hiroyuki <kamezawa.hiroyu@...>, Christoph Lameter <clameter@...>
Date: Saturday, April 12, 2008 - 6:33 pm

[the first send didn't make it to lkml, so here it is again]

Hi,

as I was doing some clean-ups in the bootmem allocator, I noticed the
patch

    5a982cbc7b3fe6cf72266f319286f29963c71b9e
    mm: fix boundary checking in free_bootmem_core

which seems to implement the opposite of what the subject promises.

It makes input arguments acceptable if they are `a bit wrong' and
silently aborts if they are completely off the whack.

Please have a look at the two patches I propose: one to revert the
original and the other one to reimplement what the whole thing was
really about: having free_bootmem() itself look up the node holding
the specified address range.

Note also that the reverted patch changed a helper function and
therefore also affected free_bootmem_node(), making the latter paper
over bugs as well.

	Hannes

Original version:

 mm/bootmem.c |   25 +++++++++++++++++++------
 1 files changed, 19 insertions(+), 6 deletions(-)

New version:

 bootmem.c |   10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)
--
Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]

Messages in current thread:
[patch 0/2] bootmem: Fix node-setup agnostic free_bootmem(), Johannes Weiner, (Sat Apr 12, 6:33 pm)
speck-geostationary