Re: [PATCH -next] slub: set PG_slab on all of slab pages

[Minchan Kim]

On Mon, Mar 05, 2012 at 05:42:56PM +0900, Namhyung Kim wrote:
> 2012-03-04 7:34 PM, Minchan Kim wrote:
> >Hi Namhyung,
> >
> 
> Hi Minchan,
> glad to see you here again :)

Thanks!

> 
> 
> >On Wed, Feb 29, 2012 at 05:54:34PM +0900, Namhyung Kim wrote:
> >>Unlike SLAB, SLUB doesn't set PG_slab on tail pages, so if a user would
> >>call free_pages() incorrectly on a object in a tail page, she will get
> >>i confused with the undefined result. Setting the flag would help her by
> >>emitting a warning on bad_page() in such a case.
> >>
> >>Reported-by: Sangseok Lee <sangseok.lee@xxxxxxx>
> >>Signed-off-by: Namhyung Kim <namhyung.kim@xxxxxxx>
> >
> >I read this thread and I feel the we don't reach right point.
> >I think it's not a compound page problem.
> >We can face above problem where we allocates big order page without __GFP_COMP
> >and free middle page of it.
> >
> >Fortunately, We can catch such a problem by put_page_testzero in __free_pages
> >if you enable CONFIG_DEBUG_VM.
> >
> >Did you tried that with CONFIG_DEBUG_VM?
> >
> 
> To be honest, I don't have a real test environment which brings this
> issue in the first place. On my simple test environment, enabling
> CONFIG_DEBUG_VM emits a bug when I tried to free middle of the slab
> pages. Thanks for pointing it out.
> 
> However I guess there was a chance to bypass that test anyhow since
> it did reach to __free_pages_ok(). If the page count was 0 already,
> free_pages() will prevent it from getting to the function even
> though CONFIG_DEBUG_VM was disabled. But I don't think it's a kernel
> bug - it seems entirely our fault :( I'll recheck and talk about it
> with my colleagues.

Let me ask a question.
Could you see bad page message by PG_slab with SLUB after you apply your patch?

--
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/