[PATCHv3 0/2] creating non-section aligned lowmem mappings

[PATCHv3 0/2] creating non-section aligned lowmem mappings

[Min-Hua Chen]

Hi,

I found problems when I reserved non-section aligned memory blocks
by the device tree. The problem is that the memblock_set_current_limit()
point to the first section aligned memblock in the
first place (sanity_check_meminfo), but the memblock may be split
into non-section aligned memblocks because the memblock_reserve() calls
in arm_memblock_init().


*** BLURB HERE ***

Min-Hua Chen (2):
  arm: fix non-section-aligned low memory mapping
  arm: use max_lowmem_limit in find_limit()

 arch/arm/mm/init.c |    2 +-
 arch/arm/mm/mmu.c  |   48 ++++++++++++++----------------------------------
 2 files changed, 15 insertions(+), 35 deletions(-)

-- 
1.7.10.4

[PATCHv3 1/2] arm: fix non-section-aligned low memory mapping

[Min-Hua Chen]

In current design, the memblock.current_limit is set to
a section-aligned value in sanity_check_meminfo().

However, the section-aligned memblock may become non-section-aligned
after arm_memblock_init(). For example, the first section-aligned
memblock is 0x00000000-0x01000000 and sanity_check_meminfo sets
current_limit to 0x01000000. After arm_memblock_init, two memory blocks
[0x00c00000 - 0x00d00000] and [0x00ff0000 - 0x01000000] are reserved
by memblock_reserve() and make the original memory block
[0x00000000-0x01000000] becomes:

[0x00000000-0x00c00000]
[0x00d00000-0x00ff0000]

When creating the low memory mapping for [0x00d00000-0x00ff0000],
since the memory block is non-section-aligned, it will need to create
a second level page table. But the current_limit is set to 0x01000000,
and it's possible to allocate a unmapped memory block.

call flow:

setup_arch
 + sanity_check_meminfo
 + arm_memblock_init
 + paging_init
    + map_lowmem
    + bootmem_init

Move the memblock_set_current_limit logic to map_lowmem(), we point
the memblock current_limit to the first section-aligned memblock block.
Since map_lowmem() is called after arm_memblock_init(), there is no way to
change memblock layout. So we can say that the first section-aligned limit
is valid during map_lowmem(). Hence fix the problem described above.

Signed-off-by: Min-Hua Chen <orca.chen@gmail.com>
---
 arch/arm/mm/mmu.c |   48 ++++++++++++++----------------------------------
 1 file changed, 14 insertions(+), 34 deletions(-)

diff --git a/arch/arm/mm/mmu.c b/arch/arm/mm/mmu.c
index 4e6ef89..73e64ab 100644
--- a/arch/arm/mm/mmu.c
+++ b/arch/arm/mm/mmu.c
@@ -1068,7 +1068,6 @@ phys_addr_t arm_lowmem_limit __initdata = 0;
 
 void __init sanity_check_meminfo(void)
 {
-	phys_addr_t memblock_limit = 0;
 	int highmem = 0;
 	phys_addr_t vmalloc_limit = __pa(vmalloc_min - 1) + 1;
 	struct memblock_region *reg;
@@ -1110,43 +1109,10 @@ void __init sanity_check_meminfo(void)
 				else
 					arm_lowmem_limit = block_end;
 			}
-
-			/*
-			 * Find the first non-section-aligned page, and point
-			 * memblock_limit at it. This relies on rounding the
-			 * limit down to be section-aligned, which happens at
-			 * the end of this function.
-			 *
-			 * With this algorithm, the start or end of almost any
-			 * bank can be non-section-aligned. The only exception
-			 * is that the start of the bank 0 must be section-
-			 * aligned, since otherwise memory would need to be
-			 * allocated when mapping the start of bank 0, which
-			 * occurs before any free memory is mapped.
-			 */
-			if (!memblock_limit) {
-				if (!IS_ALIGNED(block_start, SECTION_SIZE))
-					memblock_limit = block_start;
-				else if (!IS_ALIGNED(block_end, SECTION_SIZE))
-					memblock_limit = arm_lowmem_limit;
-			}
-
 		}
 	}
 
 	high_memory = __va(arm_lowmem_limit - 1) + 1;
-
-	/*
-	 * Round the memblock limit down to a section size.  This
-	 * helps to ensure that we will allocate memory from the
-	 * last full section, which should be mapped.
-	 */
-	if (memblock_limit)
-		memblock_limit = round_down(memblock_limit, SECTION_SIZE);
-	if (!memblock_limit)
-		memblock_limit = arm_lowmem_limit;
-
-	memblock_set_current_limit(memblock_limit);
 }
 
 static inline void prepare_page_table(void)
@@ -1331,6 +1297,7 @@ static void __init map_lowmem(void)
 	struct memblock_region *reg;
 	phys_addr_t kernel_x_start = round_down(__pa(_stext), SECTION_SIZE);
 	phys_addr_t kernel_x_end = round_up(__pa(__init_end), SECTION_SIZE);
+	phys_addr_t section_block_limit = 0;
 
 	/* Map all the lowmem memory banks. */
 	for_each_memblock(memory, reg) {
@@ -1384,6 +1351,19 @@ static void __init map_lowmem(void)
 				create_mapping(&map);
 			}
 		}
+
+		/*
+		 * The first memblock MUST be section-size-aligned. Otherwise
+		 * there is no valid low memory mapping to create 2nd level
+		 * page tables.
+		 * After the first mapping is created, other 2nd level
+		 * page tables can be created from the memory allocated
+		 * from the first memblock.
+		 */
+		if (!section_memblock_limit) {
+			section_memblock_limit = end;
+			memblock_set_current_limit(section_memblock_limit);
+		}
 	}
 }
 
-- 
1.7.10.4

[PATCHv3 1/2] arm: fix non-section-aligned low memory mapping

[Russell King - ARM Linux]

On Thu, Jun 11, 2015 at 02:59:32AM +0800, Min-Hua Chen wrote:
> In current design, the memblock.current_limit is set to
> a section-aligned value in sanity_check_meminfo().
> 
> However, the section-aligned memblock may become non-section-aligned
> after arm_memblock_init(). For example, the first section-aligned
> memblock is 0x00000000-0x01000000 and sanity_check_meminfo sets
> current_limit to 0x01000000. After arm_memblock_init, two memory blocks
> [0x00c00000 - 0x00d00000] and [0x00ff0000 - 0x01000000] are reserved
> by memblock_reserve() and make the original memory block
> [0x00000000-0x01000000] becomes:

There isn't a problem with memblock_reserve().  That just marks the
memory as reserved, it doesn't steal the memory from the lowmem
mappings - in fact, it is still expected that reserved memory
claimed in this way will be mapped.

Somehow, I don't think this is what you're doing though, because you
go on to describe a problem which can only happen if you steal memory
after arm_memblock_init() has returned.

Don't do this.  There is a specific point in the boot sequence where you
are permitted to steal memory, which is done inside arm_memblock_init().
Stealing outside of that is not permitted.

arm_memblock_steal() is written to BUG_ON() if you attempt to do this
outside of the permissible code paths.

-- 
FTTC broadband for 0.8mile line: currently at 10.5Mbps down 400kbps up
according to speedtest.net.

[PATCHv3 1/2] arm: fix non-section-aligned low memory mapping

[Min-Hua Chen]

On Wed, Jun 10, 2015 at 11:40:59PM +0100, Russell King - ARM Linux wrote:
> On Thu, Jun 11, 2015 at 02:59:32AM +0800, Min-Hua Chen wrote:
> > In current design, the memblock.current_limit is set to
> > a section-aligned value in sanity_check_meminfo().
> > 
> > However, the section-aligned memblock may become non-section-aligned
> > after arm_memblock_init(). For example, the first section-aligned
> > memblock is 0x00000000-0x01000000 and sanity_check_meminfo sets
> > current_limit to 0x01000000. After arm_memblock_init, two memory blocks
> > [0x00c00000 - 0x00d00000] and [0x00ff0000 - 0x01000000] are reserved
> > by memblock_reserve() and make the original memory block
> > [0x00000000-0x01000000] becomes:
> 
> There isn't a problem with memblock_reserve().  That just marks the
> memory as reserved, it doesn't steal the memory from the lowmem
> mappings - in fact, it is still expected that reserved memory
> claimed in this way will be mapped.
> 
> Somehow, I don't think this is what you're doing though, because you
> go on to describe a problem which can only happen if you steal memory
> after arm_memblock_init() has returned.

Yes, your are right. The probelm is not caused by memblock_reserve().
It's caused by the memory reserving code in early_init_fdt_scan_reserved_mem(),
which is in arm_memblock_init().

The memory reserving code in of_of_reserved_mem.c allows the reserved
memory blocks to have a "no-map" property. When a reserved-memory
is marked "no-map", the mapping will be removed by memblock_remove() like
arm_memblock_steal() does.

> Don't do this.  There is a specific point in the boot sequence where you
> are permitted to steal memory, which is done inside arm_memblock_init().
> Stealing outside of that is not permitted.
> 
> arm_memblock_steal() is written to BUG_ON() if you attempt to do this
> outside of the permissible code paths.
> 
> -- 
> FTTC broadband for 0.8mile line: currently at 10.5Mbps down 400kbps up
> according to speedtest.net.

[PATCHv3 2/2] arm: use max_lowmem_limit in find_limit()

[Min-Hua Chen]

In commit: 1c2f87c22566cd057bc8cde10c37ae9da1a1bb76, the max_low is
set by memblock_get_current_limit(). However memblock.current_limit
can be changed by memblock_set_current_limit() any point before
find_limits().

It's better to use arm_lowmem_limit to be max_lowmem in two ways:
First, arm_lowmem_limit cannot be changed by a public API. Second, the
high_memory is set by arm_lowmem_limit and is a natural limit of
low memory area in bootmem_init().

Signed-off-by: Min-Hua Chen <orca.chen@gmail.com>
---
 arch/arm/mm/init.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/arm/mm/init.c b/arch/arm/mm/init.c
index be92fa0..b4f9513 100644
--- a/arch/arm/mm/init.c
+++ b/arch/arm/mm/init.c
@@ -89,7 +89,7 @@ __tagtable(ATAG_INITRD2, parse_tag_initrd2);
 static void __init find_limits(unsigned long *min, unsigned long *max_low,
 			       unsigned long *max_high)
 {
-	*max_low = PFN_DOWN(memblock_get_current_limit());
+	*max_low = PFN_DOWN(arm_lowmem_limit);
 	*min = PFN_UP(memblock_start_of_DRAM());
 	*max_high = PFN_DOWN(memblock_end_of_DRAM());
 }
-- 
1.7.10.4

[PATCHv3 2/2] arm: use max_lowmem_limit in find_limit()

[Russell King - ARM Linux]

On Thu, Jun 11, 2015 at 02:59:33AM +0800, Min-Hua Chen wrote:
> In commit: 1c2f87c22566cd057bc8cde10c37ae9da1a1bb76, the max_low is
> set by memblock_get_current_limit(). However memblock.current_limit
> can be changed by memblock_set_current_limit() any point before
> find_limits().
> 
> It's better to use arm_lowmem_limit to be max_lowmem in two ways:
> First, arm_lowmem_limit cannot be changed by a public API. Second, the
> high_memory is set by arm_lowmem_limit and is a natural limit of
> low memory area in bootmem_init().
> 
> Signed-off-by: Min-Hua Chen <orca.chen@gmail.com>

NAK, this is a known build breaker.

-- 
FTTC broadband for 0.8mile line: currently at 10.5Mbps down 400kbps up
according to speedtest.net.