Messages in this thread | | | Date | Fri, 5 Nov 2004 05:03:09 +0100 | From | Andrea Arcangeli <> | Subject | Re: fix iounmap and a pageattr memleak (x86 and x86-64) |
| |
On Thu, Nov 04, 2004 at 06:23:11PM -0800, Dave Hansen wrote: > I'm not quite sure if this has any other weird effects, so I'll hold on > to it for a week or so and see if anything turns up.
this fixed the problem for me too.
However I'm not convinced this is correct, nothing in the kernel should ever free a bootmem piece of memory after the machine has booted.
If this helps, it also means we found an existing pte (not pmd) with page_count 0 during the first unmap event (bootmem allocated). The transition from mapped to unmapped works fine, but the transition from unmapped to mapped will thorw the pte away and we'll regenerate a 2M pmd where there was a pte instead. I wonder why there are 4k pages there in the first place.
Anyways I understand what's going on now thanks to your debugging, and I believe the only real fix is to use PageReserved to catch if we're working on a newly allocated page or not, I don't like to depend on the page_count being 0 for the bootmem pages like the previous code was doing. I believe my code would now fall apart even if you were using it with PSE disabled (nopentium or something). So I've to fix that bit at least and I will use PageReserved for that.
The page_count of bootmem pages really doesn't matter since they must never be freed. It really should remain 0 so we catch if anybody executes a put_page on it.
I'll fix it up... - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |