[PATCH 12/11] unpaged: fix sound Bad page states

[PATCH 12/11] unpaged: fix sound Bad page states

[Hugh Dickins]

Earlier I unifdefed PageCompound, so that snd_pcm_mmap_control_nopage
and others can give out a 0-order component of a higher-order page,
which won't be mistakenly freed when zap_pte_range unmaps it.  But
many Bad page states reported a PG_reserved was freed after all: I had
missed that we need to say __GFP_COMP to get compound page behaviour.

Some of these higher-order pages are allocated by snd_malloc_pages, some
by snd_malloc_dev_pages; or if SBUS, by sbus_alloc_consistent - but that
has no gfp arg, so add __GFP_COMP into its sparc32/64 implementations.

I'm still rather puzzled that DRM seems not to need a similar change.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
---

 arch/sparc/kernel/ioport.c |    2 +-
 arch/sparc64/kernel/sbus.c |    2 +-
 sound/core/memalloc.c      |    2 ++
 3 files changed, 4 insertions(+), 2 deletions(-)

--- 2.6.15-rc1-mm2/arch/sparc/kernel/ioport.c	2005-06-17 20:48:29.000000000 +0100
+++ linux/arch/sparc/kernel/ioport.c	2005-11-19 19:11:04.000000000 +0000
@@ -252,7 +252,7 @@ void *sbus_alloc_consistent(struct sbus_
 	}
 
 	order = get_order(len_total);
-	if ((va = __get_free_pages(GFP_KERNEL, order)) == 0)
+	if ((va = __get_free_pages(GFP_KERNEL|__GFP_COMP, order)) == 0)
 		goto err_nopages;
 
 	if ((res = kmalloc(sizeof(struct resource), GFP_KERNEL)) == NULL)
--- 2.6.15-rc1-mm2/arch/sparc64/kernel/sbus.c	2005-11-12 09:00:36.000000000 +0000
+++ linux/arch/sparc64/kernel/sbus.c	2005-11-19 19:12:53.000000000 +0000
@@ -327,7 +327,7 @@ void *sbus_alloc_consistent(struct sbus_
 	order = get_order(size);
 	if (order >= 10)
 		return NULL;
-	first_page = __get_free_pages(GFP_KERNEL, order);
+	first_page = __get_free_pages(GFP_KERNEL|__GFP_COMP, order);
 	if (first_page == 0UL)
 		return NULL;
 	memset((char *)first_page, 0, PAGE_SIZE << order);
--- 2.6.15-rc1-mm2/sound/core/memalloc.c	2005-11-12 09:01:28.000000000 +0000
+++ linux/sound/core/memalloc.c	2005-11-19 19:03:32.000000000 +0000
@@ -197,6 +197,7 @@ void *snd_malloc_pages(size_t size, gfp_
 
 	snd_assert(size > 0, return NULL);
 	snd_assert(gfp_flags != 0, return NULL);
+	gfp_flags |= __GFP_COMP;	/* compound page lets parts be mapped */
 	pg = get_order(size);
 	if ((res = (void *) __get_free_pages(gfp_flags, pg)) != NULL) {
 		mark_pages(virt_to_page(res), pg);
@@ -241,6 +242,7 @@ static void *snd_malloc_dev_pages(struct
 	snd_assert(dma != NULL, return NULL);
 	pg = get_order(size);
 	gfp_flags = GFP_KERNEL
+		| __GFP_COMP	/* compound page lets parts be mapped */
 		| __GFP_NORETRY /* don't trigger OOM-killer */
 		| __GFP_NOWARN; /* no stack trace print - this call is non-critical */
 	res = dma_alloc_coherent(dev, PAGE_SIZE << pg, dma, gfp_flags);

Re: [PATCH 12/11] unpaged: fix sound Bad page states

[Andrew Morton]

Hugh Dickins <hugh@veritas.com> wrote:
>
>   arch/sparc/kernel/ioport.c |    2 +-
>   arch/sparc64/kernel/sbus.c |    2 +-
>   sound/core/memalloc.c      |    2 ++
>   3 files changed, 4 insertions(+), 2 deletions(-)

So with this, do you think everything is in place for a mainline merge?

Re: [PATCH 12/11] unpaged: fix sound Bad page states

[William Lee Irwin III]

On Mon, Nov 21, 2005 at 06:12:38PM +0000, Hugh Dickins wrote:
> Earlier I unifdefed PageCompound, so that snd_pcm_mmap_control_nopage
> and others can give out a 0-order component of a higher-order page,
> which won't be mistakenly freed when zap_pte_range unmaps it.  But
> many Bad page states reported a PG_reserved was freed after all: I had
> missed that we need to say __GFP_COMP to get compound page behaviour.
> Some of these higher-order pages are allocated by snd_malloc_pages, some
> by snd_malloc_dev_pages; or if SBUS, by sbus_alloc_consistent - but that
> has no gfp arg, so add __GFP_COMP into its sparc32/64 implementations.
> I'm still rather puzzled that DRM seems not to need a similar change.
> Signed-off-by: Hugh Dickins <hugh@veritas.com>

Acked-by: William Irwin <wli@holomorphy.com>


-- wli

Re: [PATCH 12/11] unpaged: fix sound Bad page states

[Hugh Dickins]

On Mon, 21 Nov 2005, Andrew Morton wrote:
> Hugh Dickins <hugh@veritas.com> wrote:
> >
> >   arch/sparc/kernel/ioport.c |    2 +-
> >   arch/sparc64/kernel/sbus.c |    2 +-
> >   sound/core/memalloc.c      |    2 ++
> >   3 files changed, 4 insertions(+), 2 deletions(-)
> 
> So with this, do you think everything is in place for a mainline merge?

Yes, I believe so: thanks.

Hugh