[PATCH v2 1/2] powerpc/mm: define TOP_ZONE as a constant

The zone that contains the top of memory will be either ZONE_NORMAL
or ZONE_HIGHMEM depending on the kernel config. There are two functions
that require this information and both of them use an #ifdef to set
a local variable (top_zone). This is a little silly so lets just make it
a constant.

Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Cc: linux-mm@kvack.org
---
 arch/powerpc/mm/mem.c | 17 +++++------------
 1 file changed, 5 insertions(+), 12 deletions(-)

diff --git a/arch/powerpc/mm/mem.c b/arch/powerpc/mm/mem.c
index ac79dbde1015..8f4c19789a38 100644
--- a/arch/powerpc/mm/mem.c
+++ b/arch/powerpc/mm/mem.c

@@ -68,12 +68,15 @@ pte_t *kmap_pte;
 EXPORT_SYMBOL(kmap_pte);
 pgprot_t kmap_prot;
 EXPORT_SYMBOL(kmap_prot);
+#define TOP_ZONE ZONE_HIGHMEM
 
 static inline pte_t *virt_to_kpte(unsigned long vaddr)
 {
 	return pte_offset_kernel(pmd_offset(pud_offset(pgd_offset_k(vaddr),
 			vaddr), vaddr), vaddr);
 }
+#else
+#define TOP_ZONE ZONE_NORMAL
 #endif
 
 int page_is_ram(unsigned long pfn)

@@ -267,14 +270,9 @@ void __init limit_zone_pfn(enum zone_type zone, unsigned long pfn_limit)
  */
 int dma_pfn_limit_to_zone(u64 pfn_limit)
 {
-	enum zone_type top_zone = ZONE_NORMAL;
 	int i;
 
-#ifdef CONFIG_HIGHMEM
-	top_zone = ZONE_HIGHMEM;
-#endif
-
-	for (i = top_zone; i >= 0; i--) {
+	for (i = TOP_ZONE; i >= 0; i--) {
 		if (max_zone_pfns[i] <= pfn_limit)
 			return i;
 	}

@@ -289,7 +287,6 @@ void __init paging_init(void)
 {
 	unsigned long long total_ram = memblock_phys_mem_size();
 	phys_addr_t top_of_ram = memblock_end_of_DRAM();
-	enum zone_type top_zone;
 
 #ifdef CONFIG_PPC32
 	unsigned long v = __fix_to_virt(__end_of_fixed_addresses - 1);

@@ -313,13 +310,9 @@ void __init paging_init(void)
 	       (long int)((top_of_ram - total_ram) >> 20));
 
 #ifdef CONFIG_HIGHMEM
-	top_zone = ZONE_HIGHMEM;
 	limit_zone_pfn(ZONE_NORMAL, lowmem_end_addr >> PAGE_SHIFT);
-#else
-	top_zone = ZONE_NORMAL;
 #endif
-
-	limit_zone_pfn(top_zone, top_of_ram >> PAGE_SHIFT);
+	limit_zone_pfn(TOP_ZONE, top_of_ram >> PAGE_SHIFT);
 	zone_limits_final = true;
 	free_area_init_nodes(max_zone_pfns);
 

-- 
2.5.5

The mm zone mechanism was traditionally used by arch specific code to
partition memory into allocation zones. However there are several zones
that are managed by the mm subsystem rather than the architecture. Most
architectures set the max PFN of these special zones to zero, however on
powerpc we set them to ~0ul. This, in conjunction with a bug in
free_area_init_nodes() results in all of system memory being placed in
ZONE_DEVICE when enabled. Device memory cannot be used for regular kernel
memory allocations so this will cause a kernel panic at boot.

Given the planned addition of more mm managed zones (ZONE_CMA) we should
aim to be consistent with every other architecture and set the max PFN for
these zones to zero.

Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Cc: linux-mm@kvack.org
---
 arch/powerpc/mm/mem.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/arch/powerpc/mm/mem.c b/arch/powerpc/mm/mem.c
index 8f4c19789a38..f0a058ebb6d7 100644
--- a/arch/powerpc/mm/mem.c
+++ b/arch/powerpc/mm/mem.c

@@ -239,8 +239,14 @@ static int __init mark_nonram_nosave(void)
 
 static bool zone_limits_final;
 
+/*
+ * The memory zones past TOP_ZONE are managed by generic mm code.
+ * These should be set to zero since that's what every other
+ * architecture does.
+ */
 static unsigned long max_zone_pfns[MAX_NR_ZONES] = {
-	[0 ... MAX_NR_ZONES - 1] = ~0UL
+	[0        ... TOP_ZONE     - 1] = ~0UL,
+	[TOP_ZONE ... MAX_NR_ZONES - 1] = 0
 };
 
 /*

-- 
2.5.5

On Thu, 2016-05-05 at 07:54:09 UTC, Oliver O'Halloran wrote:

The mm zone mechanism was traditionally used by arch specific code to
partition memory into allocation zones. However there are several zones
that are managed by the mm subsystem rather than the architecture. Most
architectures set the max PFN of these special zones to zero, however on
powerpc we set them to ~0ul. This, in conjunction with a bug in
free_area_init_nodes() results in all of system memory being placed in
ZONE_DEVICE when enabled. Device memory cannot be used for regular kernel
memory allocations so this will cause a kernel panic at boot.

This is breaking my freescale machine:

  Sorting __ex_table...
  Unable to handle kernel paging request for data at address 0xc000000101e28020
  Faulting instruction address: 0xc0000000009ab698
  cpu 0x0: Vector: 300 (Data Access) at [c000000000acbb30]
      pc: c0000000009ab698: .reserve_bootmem_region+0x64/0x8c
      lr: c0000000009883d0: .free_all_bootmem+0x70/0x200
      sp: c000000000acbdb0
     msr: 80021000
     dar: c000000101e28020
   dsisr: 800000
    current = 0xc000000000a07640
    paca    = 0xc00000003fff5000	 softe: 0	 irq_happened: 0x01
      pid   = 0, comm = swapper
  Linux version 4.6.0-rc3-00160-gc09920947f23 (michael@ka1) (gcc version 5.3.0 (GCC) ) #5 SMP Tue May 10 09:44:11 AEST 2016
  enter ? for help
  [link register   ] c0000000009883d0 .free_all_bootmem+0x70/0x200
  [c000000000acbdb0] c000000000988398 .free_all_bootmem+0x38/0x200 (unreliable)
  [c000000000acbe80] c00000000097b700 .mem_init+0x5c/0x7c
  [c000000000acbef0] c000000000971a0c .start_kernel+0x28c/0x4e4
  [c000000000acbf90] c000000000000544 start_here_common+0x20/0x5c
  0:mon> ? 

I can give you access some time if you need to debug it.

cheers

On Thu, 2016-05-05 at 07:54:08 UTC, Oliver O'Halloran wrote:

The zone that contains the top of memory will be either ZONE_NORMAL
or ZONE_HIGHMEM depending on the kernel config. There are two functions
that require this information and both of them use an #ifdef to set
a local variable (top_zone). This is a little silly so lets just make it
a constant.

Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Cc: linux-mm@kvack.org

Applied to powerpc next, thanks.

https://git.kernel.org/powerpc/c/d69777dbefd707974aed91918d

cheers

On Tue, 2016-05-10 at 09:55 +1000, Michael Ellerman wrote:

On Thu, 2016-05-05 at 07:54:09 UTC, Oliver O'Halloran wrote:

quoted

The mm zone mechanism was traditionally used by arch specific code to
partition memory into allocation zones. However there are several zones
that are managed by the mm subsystem rather than the architecture. Most
architectures set the max PFN of these special zones to zero, however on
powerpc we set them to ~0ul. This, in conjunction with a bug in
free_area_init_nodes() results in all of system memory being placed in
ZONE_DEVICE when enabled. Device memory cannot be used for regular kernel
memory allocations so this will cause a kernel panic at boot.

This is breaking my freescale machine:

  Sorting __ex_table...
  Unable to handle kernel paging request for data at address 0xc000000101e28020
  Faulting instruction address: 0xc0000000009ab698
  cpu 0x0: Vector: 300 (Data Access) at [c000000000acbb30]
      pc: c0000000009ab698: .reserve_bootmem_region+0x64/0x8c
      lr: c0000000009883d0: .free_all_bootmem+0x70/0x200
      sp: c000000000acbdb0
     msr: 80021000
     dar: c000000101e28020
   dsisr: 800000
    current = 0xc000000000a07640
    paca    = 0xc00000003fff5000	 softe: 0	 irq_happened: 0x01
      pid   = 0, comm = swapper
  Linux version 4.6.0-rc3-00160-gc09920947f23 (michael@ka1) (gcc version 5.3.0 (GCC) ) #5 SMP Tue May 10 09:44:11 AEST 2016
  enter ? for help
  [link register   ] c0000000009883d0 .free_all_bootmem+0x70/0x200
  [c000000000acbdb0] c000000000988398 .free_all_bootmem+0x38/0x200 (unreliable)
  [c000000000acbe80] c00000000097b700 .mem_init+0x5c/0x7c
  [c000000000acbef0] c000000000971a0c .start_kernel+0x28c/0x4e4
  [c000000000acbf90] c000000000000544 start_here_common+0x20/0x5c
  0:mon> ? 

I can give you access some time if you need to debug it.

Could you also please post the bits on the boot containing the zone
and node information. That would provide some information about what
is broken. Or you could just send the whole dmesg

Thanks,
Balbir Singh

The mm zone mechanism was traditionally used by arch specific code to
partition memory into allocation zones. However there are several zones
that are managed by the mm subsystem rather than the architecture. Most
architectures set the max PFN of these special zones to zero, however on
powerpc we set them to ~0ul. This, in conjunction with a bug in
free_area_init_nodes() results in all of system memory being placed in
ZONE_DEVICE when enabled. Device memory cannot be used for regular kernel
memory allocations so this will cause a kernel panic at boot. Given the
planned addition of more mm managed zones (ZONE_CMA) we should aim to be
consistent with every other architecture and set the max PFN for these
zones to zero.

Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Reviewed-by: Balbir Singh <bsingharora@gmail.com>
Cc: linux-mm@kvack.org
---
 arch/powerpc/mm/mem.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/arch/powerpc/mm/mem.c b/arch/powerpc/mm/mem.c
index 879e0bc6f82e..f35e6605c422 100644
--- a/arch/powerpc/mm/mem.c
+++ b/arch/powerpc/mm/mem.c

@@ -239,8 +239,14 @@ static int __init mark_nonram_nosave(void)
 
 static bool zone_limits_final;
 
+/*
+ * The memory zones past TOP_ZONE are managed by generic mm code.
+ * These should be set to zero since that's what every other
+ * architecture does.
+ */
 static unsigned long max_zone_pfns[MAX_NR_ZONES] = {
-	[0 ... MAX_NR_ZONES - 1] = ~0UL
+	[0            ... TOP_ZONE        ] = ~0UL,
+	[TOP_ZONE + 1 ... MAX_NR_ZONES - 1] = 0
 };
 
 /*

-- 
2.5.5

On Wed, 2016-11-05 at 09:22:18 UTC, Oliver O'Halloran wrote:

The mm zone mechanism was traditionally used by arch specific code to
partition memory into allocation zones. However there are several zones
that are managed by the mm subsystem rather than the architecture. Most
architectures set the max PFN of these special zones to zero, however on
powerpc we set them to ~0ul. This, in conjunction with a bug in
free_area_init_nodes() results in all of system memory being placed in
ZONE_DEVICE when enabled. Device memory cannot be used for regular kernel
memory allocations so this will cause a kernel panic at boot. Given the
planned addition of more mm managed zones (ZONE_CMA) we should aim to be
consistent with every other architecture and set the max PFN for these
zones to zero.

Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Reviewed-by: Balbir Singh <bsingharora@gmail.com>
Cc: linux-mm@kvack.org

Applied to powerpc next, thanks.

https://git.kernel.org/powerpc/c/3079abe555511031e2ba5d1e21

cheers