[PATCH] [RESEND] arm: limit memblock base address for early_pte_alloc
Kim, Jong-Sung
neidhard.kim at lge.com
Thu Jun 28 01:43:17 EDT 2012
> From: Dave Martin [mailto:dave.martin at linaro.org]
> Sent: Thursday, June 28, 2012 1:02 AM
>
> For me, it appears that this block just contains the initial region passed
> in ATAG_MEM or on the command line, with some reservations for
> swapper_pg_dir, the kernel text/data, device tree and initramfs.
>
> So far as I can tell, the only memory guaranteed to be mapped here is the
> kernel image: there may be no guarantee that there is any unused space in
> this region which could be used to allocate extra page tables.
> The rest appears during the execution of map_lowmem().
>
> Cheers
> ---Dave
Thank you for your comment, Dave! It was not that sophisticated choice, but
I thought that normal embedded system trying to reduce the BOM would have a
big-enough first memblock memory region. However you're right. There can be
exceptional systems. Then, how do you think about following manner:
diff --git a/arch/arm/mm/mmu.c b/arch/arm/mm/mmu.c
index e5dad60..0bc5316 100644
--- a/arch/arm/mm/mmu.c
+++ b/arch/arm/mm/mmu.c
@@ -1094,6 +1094,16 @@ static void __init kmap_init(void)
static void __init map_lowmem(void)
{
struct memblock_region *reg;
+ phys_addr_t pmd_map_end = 0;
+
+ for_each_memblock(memory, reg) {
+ pmd_map_end = reg->base + reg->size;
+ if((reg->base | reg->size) & ~PMD_MASK)
+ break;
+ }
+ if(pmd_map_end > lowmem_limit)
+ pmd_map_end = lowmem_limit;
+ memblock_set_current_limit(pmd_map_end & PMD_MASK);
/* Map all the lowmem memory banks. */
for_each_memblock(memory, reg) {
@@ -1113,6 +1123,8 @@ static void __init map_lowmem(void)
create_mapping(&map);
}
+
+ memblock_set_current_limit(lowmem_limit);
}
/*
@@ -1123,8 +1135,6 @@ void __init paging_init(struct machine_desc *mdesc)
{
void *zero_page;
- memblock_set_current_limit(arm_lowmem_limit);
-
build_mem_type_table();
prepare_page_table();
map_lowmem();
This will not limit the PTE-allocation to near the end of first bank.
More information about the linux-arm-kernel
mailing list