| Message ID | 1371338108-26354-1-git-send-email-grant.likely@linaro.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Sun, Jun 16, 2013 at 12:15:08AM +0100, Grant Likely wrote: > I don't think this special case is needed anymore. From what I can tell, > setup_dma_zone() takes care of marking the region to be reserved for > DMA, and that happens before arm_memblock_init() which calls > arm_mm_memblock_reserve(). Therefore the mm setup should already know > about the reserved region and it won't need to be explicitly reserved > and then freed. That means the special case can be dropped from the > common code. No it doesn't. setup_dma_zone is all about setting the high watermark for GFP_DMA allocations. It has nothing to do with _temoprarily_ reserving the low 1MB, ensuring that no other user can use any of that precious SA1111 memory during the kernel boot.
On Sun, Jun 16, 2013 at 10:10 AM, Russell King - ARM Linux <linux@arm.linux.org.uk> wrote: > On Sun, Jun 16, 2013 at 12:15:08AM +0100, Grant Likely wrote: >> I don't think this special case is needed anymore. From what I can tell, >> setup_dma_zone() takes care of marking the region to be reserved for >> DMA, and that happens before arm_memblock_init() which calls >> arm_mm_memblock_reserve(). Therefore the mm setup should already know >> about the reserved region and it won't need to be explicitly reserved >> and then freed. That means the special case can be dropped from the >> common code. > > No it doesn't. setup_dma_zone is all about setting the high watermark > for GFP_DMA allocations. It has nothing to do with _temoprarily_ > reserving the low 1MB, ensuring that no other user can use any of that > precious SA1111 memory during the kernel boot. Ah, okay. So it's to protect against early allocations. All right, I've got an alternate patch that at least gets rid of one of the special case hooks. I'll post that one shortly. g.
diff --git a/arch/arm/mm/init.c b/arch/arm/mm/init.c index 9a5cdc0..73212fe 100644 --- a/arch/arm/mm/init.c +++ b/arch/arm/mm/init.c @@ -598,11 +598,6 @@ void __init mem_init(void) totalram_pages += free_all_bootmem(); -#ifdef CONFIG_SA1111 - /* now that our DMA memory is actually so designated, we can free it */ - free_reserved_area(__va(PHYS_PFN_OFFSET), swapper_pg_dir, 0, NULL); -#endif - free_highpages(); reserved_pages = free_pages = 0; diff --git a/arch/arm/mm/mmu.c b/arch/arm/mm/mmu.c index e0d8565..f281a24 100644 --- a/arch/arm/mm/mmu.c +++ b/arch/arm/mm/mmu.c @@ -1147,14 +1147,6 @@ void __init arm_mm_memblock_reserve(void) * and can only be in node 0. */ memblock_reserve(__pa(swapper_pg_dir), SWAPPER_PG_DIR_SIZE); - -#ifdef CONFIG_SA1111 - /* - * Because of the SA1111 DMA bug, we want to preserve our - * precious DMA-able memory... - */ - memblock_reserve(PHYS_OFFSET, __pa(swapper_pg_dir) - PHYS_OFFSET); -#endif } /*
I don't think this special case is needed anymore. From what I can tell, setup_dma_zone() takes care of marking the region to be reserved for DMA, and that happens before arm_memblock_init() which calls arm_mm_memblock_reserve(). Therefore the mm setup should already know about the reserved region and it won't need to be explicitly reserved and then freed. That means the special case can be dropped from the common code. Signed-off-by: Grant Likely <grant.likely@secretlab.ca> --- Hi Russell and Nicolas, I really don't know if this is correct. The MM code is not my strong area. Please take a look. Thanks, g. arch/arm/mm/init.c | 5 ----- arch/arm/mm/mmu.c | 8 -------- 2 files changed, 13 deletions(-)