Remove ARM meminfo

Apart from setting the limit of memblock, it's also useful to be able to get the limit to avoid recalculating it every time. Add the function to do so.

Acked-by Catalin Marinas <[email protected]>

Acked-by Santosh Shilimkar <[email protected]>

Acked-by Andrew Morton <[email protected]>

Signed-off-by Laura Abbott <[email protected]>

---
 include/linux/memblock.h |    2 ++
 mm/memblock.c            |    5 +++++
 2 files changed, 7 insertions(+), 0 deletions(-)

include/linux/memblock.h +2 -0

index 1ef6636..8a20a51 100644
@@ -252,6 +252,8 @@ static inline void memblock_dump_all(void)
 void memblock_set_current_limit(phys_addr_t limit);
 
 
+phys_addr_t memblock_get_current_limit(void);
+
 /*
  * pfn conversion functions
  *

mm/memblock.c +5 -0

index 39a31e7..7fe5354 100644
@@ -1407,6 +1407,11 @@ void __init_memblock memblock_set_current_limit(phys_addr_t limit)
 	memblock.current_limit = limit;
 }
 
+phys_addr_t __init_memblock memblock_get_current_limit(void)
+{
+	return memblock.current_limit;
+}
+
 static void __init_memblock memblock_dump(struct memblock_type *type, char *name)
 {
 	unsigned long long base, size;

Signature · 2 lines

The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
hosted by The Linux Foundation

memblock is now fully integrated into the kernel and is the prefered method for tracking memory. Rather than reinvent the wheel with meminfo, migrate to using memblock directly instead of meminfo as an intermediate.

#define NR_BANKS CONFIG_ARM_NR_BANKS

Quoted 7 lines · Laura Abbott

memblock is now fully integrated into the kernel and is the prefered method for tracking memory. Rather than reinvent the wheel with meminfo, migrate to using memblock directly instead of meminfo as an intermediate.

#define NR_BANKS CONFIG_ARM_NR_BANKS

arch/arm/include/asm/setup.h static const struct tagtable __tagtable_##fn __tag = { tag, fn } · line 38

Quoted 10 lines

arch/arm/include/asm/setup.h +0 -10

-struct membank {
-	phys_addr_t start;
-	phys_addr_t size;
-	unsigned int highmem;
-};
-
-struct meminfo {
-	int nr_banks;
-	struct membank bank[NR_BANKS];
-};

Doesn't this make NR_BANKS (and CONFIG_ARM_NR_BANKS) unused?

memblock is now fully integrated into the kernel and is the prefered method for tracking memory. Rather than reinvent the wheel with meminfo, migrate to using memblock directly instead of meminfo as an intermediate.

Acked-by Jason Cooper <[email protected]>

Acked-by Catalin Marinas <[email protected]>

Acked-by Santosh Shilimkar <[email protected]>

Acked-by Kukjin Kim <[email protected]>

Tested-by Marek Szyprowski <[email protected]>

Tested-by Leif Lindholm <[email protected]>

Signed-off-by Laura Abbott <[email protected]>

memblock is now fully integrated into the kernel and is the prefered method for tracking memory. Rather than reinvent the wheel with meminfo, migrate to using memblock directly instead of meminfo as an intermediate.

Acked-by Jason Cooper <[email protected]>

Acked-by Catalin Marinas <[email protected]>

Acked-by Santosh Shilimkar <[email protected]>

Acked-by Kukjin Kim <[email protected]>

Tested-by Marek Szyprowski <[email protected]>

Tested-by Leif Lindholm <[email protected]>

Signed-off-by Laura Abbott <[email protected]>

Quoted 5 lines · Laura Abbott

memblock is now fully integrated into the kernel and is the prefered method for tracking memory. Rather than reinvent the wheel with meminfo, migrate to using memblock directly instead of meminfo as an intermediate.

Quoted 7 lines

Acked-by Jason Cooper <[email protected]>

Acked-by Catalin Marinas <[email protected]>

Acked-by Santosh Shilimkar <[email protected]>

Acked-by Kukjin Kim <[email protected]>

Tested-by Marek Szyprowski <[email protected]>

Tested-by Leif Lindholm <[email protected]>

Signed-off-by Laura Abbott <[email protected]>

Laura,

This patch causes a bunch of platforms to no longer boot - imx6solo with 1GB of RAM boots, imx6q with 2GB of RAM doesn't. Versatile Express doesn't.

The early printk messages don't reveal anything too interesting:

Booting Linux on physical CPU 0x0
Linux version 3.14.0-rc6+ ([email protected]) (gcc version 4.6.4 (GCC) ) #630 SMP Wed Mar 12 01:13:36 GMT 2014
CPU: ARMv7 Processor [412fc09a] revision 10 (ARMv7), cr=10c53c7d
CPU: PIPT / VIPT nonaliasing data cache, VIPT aliasing instruction cache
Machine model: SolidRun Cubox-i Dual/Quad
cma: CMA: reserved 64 MiB at 8c000000
Memory policy: Data cache writealloc
<hang>

vs.

Booting Linux on physical CPU 0x0
Linux version 3.14.0-rc6+ ([email protected]) (gcc version 4.6.4 (GCC) ) #631 SMP Wed Mar 12 01:15:37 GMT 2014
CPU: ARMv7 Processor [412fc09a] revision 10 (ARMv7), cr=10c53c7d
CPU: PIPT / VIPT nonaliasing data cache, VIPT aliasing instruction cache
Machine model: SolidRun Cubox-i Dual/Quad
cma: CMA: reserved 64 MiB at 3b800000
Memory policy: Data cache writealloc
On node 0 totalpages: 524288
free_area_init_node: node 0, pgdat c09d0240, node_mem_map ea7d8000
Normal zone: 1520 pages used for memmap
Normal zone: 0 pages reserved
Normal zone: 194560 pages, LIFO batch:31
HighMem zone: 2576 pages used for memmap
HighMem zone: 329728 pages, LIFO batch:31
...

The only obvious difference is the address of that CMA reservation, CMA shouldn't make a difference here - but I suspect that other allocations which need to be in lowmem probably aren't.

Hi Russell,

Quoted 48 lines · Russell King - ARM Linux

Quoted 5 lines · Laura Abbott

memblock is now fully integrated into the kernel and is the prefered method for tracking memory. Rather than reinvent the wheel with meminfo, migrate to using memblock directly instead of meminfo as an intermediate.

Quoted 7 lines

Acked-by Jason Cooper <[email protected]>

Acked-by Catalin Marinas <[email protected]>

Acked-by Santosh Shilimkar <[email protected]>

Acked-by Kukjin Kim <[email protected]>

Tested-by Marek Szyprowski <[email protected]>

Tested-by Leif Lindholm <[email protected]>

Signed-off-by Laura Abbott <[email protected]>

Laura,

This patch causes a bunch of platforms to no longer boot - imx6solo with 1GB of RAM boots, imx6q with 2GB of RAM doesn't. Versatile Express doesn't.

The early printk messages don't reveal anything too interesting:

Booting Linux on physical CPU 0x0
Linux version 3.14.0-rc6+ (rmk at rmk-PC.arm.linux.org.uk) (gcc version 4.6.4 (GCC) ) #630 SMP Wed Mar 12 01:13:36 GMT 2014
CPU: ARMv7 Processor [412fc09a] revision 10 (ARMv7), cr=10c53c7d
CPU: PIPT / VIPT nonaliasing data cache, VIPT aliasing instruction cache
Machine model: SolidRun Cubox-i Dual/Quad
cma: CMA: reserved 64 MiB at 8c000000
Memory policy: Data cache writealloc
<hang>

vs.

Booting Linux on physical CPU 0x0
Linux version 3.14.0-rc6+ (rmk at rmk-PC.arm.linux.org.uk) (gcc version 4.6.4 (GCC) ) #631 SMP Wed Mar 12 01:15:37 GMT 2014
CPU: ARMv7 Processor [412fc09a] revision 10 (ARMv7), cr=10c53c7d
CPU: PIPT / VIPT nonaliasing data cache, VIPT aliasing instruction cache
Machine model: SolidRun Cubox-i Dual/Quad
cma: CMA: reserved 64 MiB at 3b800000
Memory policy: Data cache writealloc
On node 0 totalpages: 524288
free_area_init_node: node 0, pgdat c09d0240, node_mem_map ea7d8000
Normal zone: 1520 pages used for memmap
Normal zone: 0 pages reserved
Normal zone: 194560 pages, LIFO batch:31
HighMem zone: 2576 pages used for memmap
HighMem zone: 329728 pages, LIFO batch:31
...

Quoted 4 lines

The only obvious difference is the address of that CMA reservation, CMA shouldn't make a difference here - but I suspect that other allocations which need to be in lowmem probably aren't.

Could it be possible to enable memblock debug by adding "memblock=debug" in cmdline?

Try booting a machine with 2G of RAM with page offset set to 3GB and highmem enabled - it will fail as per the above.

Quoted 51 lines · Grygorii Strashko

Hi Russell,

Quoted 48 lines · Russell King - ARM Linux

Quoted 5 lines · Laura Abbott

memblock is now fully integrated into the kernel and is the prefered method for tracking memory. Rather than reinvent the wheel with meminfo, migrate to using memblock directly instead of meminfo as an intermediate.

Quoted 7 lines

Acked-by Jason Cooper <[email protected]>

Acked-by Catalin Marinas <[email protected]>

Acked-by Santosh Shilimkar <[email protected]>

Acked-by Kukjin Kim <[email protected]>

Tested-by Marek Szyprowski <[email protected]>

Tested-by Leif Lindholm <[email protected]>

Signed-off-by Laura Abbott <[email protected]>

Laura,

This patch causes a bunch of platforms to no longer boot - imx6solo with 1GB of RAM boots, imx6q with 2GB of RAM doesn't. Versatile Express doesn't.

The early printk messages don't reveal anything too interesting:

Booting Linux on physical CPU 0x0
Linux version 3.14.0-rc6+ ([email protected]) (gcc version 4.6.4 (GCC) ) #630 SMP Wed Mar 12 01:13:36 GMT 2014
CPU: ARMv7 Processor [412fc09a] revision 10 (ARMv7), cr=10c53c7d
CPU: PIPT / VIPT nonaliasing data cache, VIPT aliasing instruction cache
Machine model: SolidRun Cubox-i Dual/Quad
cma: CMA: reserved 64 MiB at 8c000000
Memory policy: Data cache writealloc
<hang>

vs.

Booting Linux on physical CPU 0x0
Linux version 3.14.0-rc6+ ([email protected]) (gcc version 4.6.4 (GCC) ) #631 SMP Wed Mar 12 01:15:37 GMT 2014
CPU: ARMv7 Processor [412fc09a] revision 10 (ARMv7), cr=10c53c7d
CPU: PIPT / VIPT nonaliasing data cache, VIPT aliasing instruction cache
Machine model: SolidRun Cubox-i Dual/Quad
cma: CMA: reserved 64 MiB at 3b800000
Memory policy: Data cache writealloc
On node 0 totalpages: 524288
free_area_init_node: node 0, pgdat c09d0240, node_mem_map ea7d8000
Normal zone: 1520 pages used for memmap
Normal zone: 0 pages reserved
Normal zone: 194560 pages, LIFO batch:31
HighMem zone: 2576 pages used for memmap
HighMem zone: 329728 pages, LIFO batch:31
...

Quoted 7 lines

Quoted 4 lines

The only obvious difference is the address of that CMA reservation, CMA shouldn't make a difference here - but I suspect that other allocations which need to be in lowmem probably aren't.

Could it be possible to enable memblock debug by adding "memblock=debug" in cmdline?

Here's with Laura's patch:

Uncompressing Linux... done, booting the kernel.
Booting Linux on physical CPU 0x0
Linux version 3.14.0-rc6+ ([email protected]) (gcc version 4.6.4 (GCC) ) #633 SMP Wed Mar 12 12:56:15 GMT 2014
CPU: ARMv7 Processor [412fc09a] revision 10 (ARMv7), cr=10c53c7d
CPU: PIPT / VIPT nonaliasing data cache, VIPT aliasing instruction cache
Machine model: SolidRun Cubox-i Dual/Quad
memblock_reserve: [0x00000010008240-0x0000001112c1f7] flags 0x0 arm_memblock_init+0x28/0x1a8
memblock_reserve: [0x00000020000040-0x000000201caa57] flags 0x0 arm_memblock_init+0x108/0x1a8
memblock_reserve: [0x00000010004000-0x00000010007fff] flags 0x0 arm_mm_memblock_reserve+0x1c/0x24
memblock_reserve: [0x00000018000000-0x0000001800b07f] flags 0x0 arm_dt_memblock_reserve+0x2c/0x70
memblock_reserve: [0x00000018000000-0x0000001800afff] flags 0x0 arm_dt_memblock_reserve+0x68/0x70
memblock_reserve: [0x00000020000040-0x000000201caa57] flags 0x0 arm_dt_memblock_reserve+0x68/0x70
memblock_reserve: [0x0000008c000000-0x0000008fffffff] flags 0x0 memblock_alloc_base_nid+0x40/0x54
cma: CMA: reserved 64 MiB at 8c000000
MEMBLOCK configuration:
memory size = 0x80000000 reserved size = 0x52fda50
memory.cnt = 0x1
memory[0x0] [0x00000010000000-0x0000008fffffff], 0x80000000 bytes flags: 0x06
reserved.cnt = 0x5
reserved[0x0] [0x00000010004000-0x00000010007fff], 0x4000 bytes flags: 0x0
reserved[0x1] [0x00000010008240-0x0000001112c1f7], 0x1123fb8 bytes flags: 0x0
reserved[0x2] [0x00000018000000-0x0000001800b07f], 0xb080 bytes flags: 0x0
reserved[0x3] [0x00000020000040-0x000000201caa57], 0x1caa18 bytes flags: 0x0
reserved[0x4] [0x0000008c000000-0x0000008fffffff], 0x4000000 bytes flags: 0x0
Memory policy: Data cache writealloc
memblock_reserve: [0x0000008bffffd8-0x0000008bffffff] flags 0x0 memblock_alloc_base_nid+0x40/0x54

Here's without:

Booting Linux on physical CPU 0x0
Linux version 3.14.0-rc6+ ([email protected]) (gcc version 4.6.4 (GCC) ) #635 SMP Wed Mar 12 13:22:15 GMT 2014
CPU: ARMv7 Processor [412fc09a] revision 10 (ARMv7), cr=10c53c7d
CPU: PIPT / VIPT nonaliasing data cache, VIPT aliasing instruction cache
Machine model: SolidRun Cubox-i Dual/Quad
memblock_reserve: [0x00000010008240-0x0000001112c277] flags 0x0 arm_memblock_init+0x54/0x1d4
memblock_reserve: [0x00000020000040-0x000000201caa57] flags 0x0 arm_memblock_init+0x134/0x1d4
memblock_reserve: [0x00000010004000-0x00000010007fff] flags 0x0 arm_mm_memblock_reserve+0x1c/0x24
memblock_reserve: [0x00000018000000-0x0000001800b07f] flags 0x0 arm_dt_memblock_reserve+0x2c/0x70
memblock_reserve: [0x00000018000000-0x0000001800afff] flags 0x0 arm_dt_memblock_reserve+0x68/0x70
memblock_reserve: [0x00000020000040-0x000000201caa57] flags 0x0 arm_dt_memblock_reserve+0x68/0x70
memblock_reserve: [0x0000003b800000-0x0000003f7fffff] flags 0x0 memblock_alloc_base_nid+0x40/0x54
cma: CMA: reserved 64 MiB at 3b800000
MEMBLOCK configuration:
memory size = 0x80000000 reserved size = 0x52fdad0
memory.cnt = 0x1
memory[0x0] [0x00000010000000-0x0000008fffffff], 0x80000000 bytes flags: 0x0
reserved.cnt = 0x5

  reserved[0x0]   [0x00000010004000-0x00000010007fff], 0x4000 bytes flags: 0x0
  reserved[0x1]   [0x00000010008240-0x0000001112c277], 0x1124038 bytes flags: 0x0
  reserved[0x2]   [0x00000018000000-0x0000001800b07f], 0xb080 bytes flags: 0x0
  reserved[0x3]   [0x00000020000040-0x000000201caa57], 0x1caa18 bytes flags: 0x0
  reserved[0x4]   [0x0000003b800000-0x0000003f7fffff], 0x4000000 bytes flags: 0x0

Memory policy: Data cache writealloc
memblock_reserve: [0x0000003b7fffd8-0x0000003b7fffff] flags 0x0 memblock_alloc_base_nid+0x40/0x54
memblock_reserve: [0x0000003b7fe000-0x0000003b7fefff] flags 0x0 memblock_alloc_base_nid+0x40/0x54
memblock_reserve: [0x0000003b7fd000-0x0000003b7fdfff] flags 0x0 memblock_alloc_base_nid+0x40/0x54
memblock_reserve: [0x0000003b7fc000-0x0000003b7fcfff] flags 0x0 memblock_alloc_base_nid+0x40/0x54
...

So it looks like allocations which must come from lowmem aren't being limited to lowmem.

Try booting a machine with 2G of RAM with page offset set to 3GB and highmem enabled - it will fail as per the above.

In fact, if we look at sanity_check_meminfo() post that patch, it's clearly wrong:

         for_each_memblock(memory, reg) {
                 phys_addr_t block_start = reg->base;
                 phys_addr_t block_end = reg->base + reg->size;
                 phys_addr_t size_limit = reg->size;

                 if (reg->base >= vmalloc_limit)
                         highmem = 1;
                 else
                         size_limit = vmalloc_limit - reg->base;
...
                 if (!highmem) {
                         if (block_end > arm_lowmem_limit)
                                 arm_lowmem_limit = block_end;

...

                 }

Quoted 130 lines · Russell King - ARM Linux

Quoted 51 lines · Grygorii Strashko

Hi Russell,

Quoted 48 lines · Russell King - ARM Linux

Quoted 5 lines · Laura Abbott

memblock is now fully integrated into the kernel and is the prefered method for tracking memory. Rather than reinvent the wheel with meminfo, migrate to using memblock directly instead of meminfo as an intermediate.

Quoted 7 lines

Acked-by Jason Cooper <[email protected]>

Acked-by Catalin Marinas <[email protected]>

Acked-by Santosh Shilimkar <[email protected]>

Acked-by Kukjin Kim <[email protected]>

Tested-by Marek Szyprowski <[email protected]>

Tested-by Leif Lindholm <[email protected]>

Signed-off-by Laura Abbott <[email protected]>

Laura,

This patch causes a bunch of platforms to no longer boot - imx6solo with 1GB of RAM boots, imx6q with 2GB of RAM doesn't. Versatile Express doesn't.

The early printk messages don't reveal anything too interesting:

Booting Linux on physical CPU 0x0
Linux version 3.14.0-rc6+ (rmk at rmk-PC.arm.linux.org.uk) (gcc version 4.6.4 (GCC) ) #630 SMP Wed Mar 12 01:13:36 GMT 2014
CPU: ARMv7 Processor [412fc09a] revision 10 (ARMv7), cr=10c53c7d
CPU: PIPT / VIPT nonaliasing data cache, VIPT aliasing instruction cache
Machine model: SolidRun Cubox-i Dual/Quad
cma: CMA: reserved 64 MiB at 8c000000
Memory policy: Data cache writealloc
<hang>

vs.

Booting Linux on physical CPU 0x0
Linux version 3.14.0-rc6+ (rmk at rmk-PC.arm.linux.org.uk) (gcc version 4.6.4 (GCC) ) #631 SMP Wed Mar 12 01:15:37 GMT 2014
CPU: ARMv7 Processor [412fc09a] revision 10 (ARMv7), cr=10c53c7d
CPU: PIPT / VIPT nonaliasing data cache, VIPT aliasing instruction cache
Machine model: SolidRun Cubox-i Dual/Quad
cma: CMA: reserved 64 MiB at 3b800000
Memory policy: Data cache writealloc
On node 0 totalpages: 524288
free_area_init_node: node 0, pgdat c09d0240, node_mem_map ea7d8000
Normal zone: 1520 pages used for memmap
Normal zone: 0 pages reserved
Normal zone: 194560 pages, LIFO batch:31
HighMem zone: 2576 pages used for memmap
HighMem zone: 329728 pages, LIFO batch:31
...

Quoted 7 lines

Quoted 4 lines

The only obvious difference is the address of that CMA reservation, CMA shouldn't make a difference here - but I suspect that other allocations which need to be in lowmem probably aren't.

Could it be possible to enable memblock debug by adding "memblock=debug" in cmdline?

Here's with Laura's patch:

Uncompressing Linux... done, booting the kernel.
Booting Linux on physical CPU 0x0
Linux version 3.14.0-rc6+ (rmk at rmk-PC.arm.linux.org.uk) (gcc version 4.6.4 (GCC) ) #633 SMP Wed Mar 12 12:56:15 GMT 2014
CPU: ARMv7 Processor [412fc09a] revision 10 (ARMv7), cr=10c53c7d
CPU: PIPT / VIPT nonaliasing data cache, VIPT aliasing instruction cache
Machine model: SolidRun Cubox-i Dual/Quad
memblock_reserve: [0x00000010008240-0x0000001112c1f7] flags 0x0 arm_memblock_init+0x28/0x1a8
memblock_reserve: [0x00000020000040-0x000000201caa57] flags 0x0 arm_memblock_init+0x108/0x1a8
memblock_reserve: [0x00000010004000-0x00000010007fff] flags 0x0 arm_mm_memblock_reserve+0x1c/0x24
memblock_reserve: [0x00000018000000-0x0000001800b07f] flags 0x0 arm_dt_memblock_reserve+0x2c/0x70
memblock_reserve: [0x00000018000000-0x0000001800afff] flags 0x0 arm_dt_memblock_reserve+0x68/0x70
memblock_reserve: [0x00000020000040-0x000000201caa57] flags 0x0 arm_dt_memblock_reserve+0x68/0x70
memblock_reserve: [0x0000008c000000-0x0000008fffffff] flags 0x0 memblock_alloc_base_nid+0x40/0x54
cma: CMA: reserved 64 MiB at 8c000000
MEMBLOCK configuration:

   memory size = 0x80000000 reserved size = 0x52fda50
   memory.cnt  = 0x1
   memory[0x0]   [0x00000010000000-0x0000008fffffff], 0x80000000 bytes flags: 0x06
   reserved.cnt  = 0x5
   reserved[0x0] [0x00000010004000-0x00000010007fff], 0x4000 bytes flags: 0x0
   reserved[0x1] [0x00000010008240-0x0000001112c1f7], 0x1123fb8 bytes flags: 0x0
   reserved[0x2] [0x00000018000000-0x0000001800b07f], 0xb080 bytes flags: 0x0
   reserved[0x3] [0x00000020000040-0x000000201caa57], 0x1caa18 bytes flags: 0x0
   reserved[0x4] [0x0000008c000000-0x0000008fffffff], 0x4000000 bytes flags: 0x0

Memory policy: Data cache writealloc
memblock_reserve: [0x0000008bffffd8-0x0000008bffffff] flags 0x0 memblock_alloc_base_nid+0x40/0x54

Here's without:

Booting Linux on physical CPU 0x0
Linux version 3.14.0-rc6+ (rmk at rmk-PC.arm.linux.org.uk) (gcc version 4.6.4 (GCC) ) #635 SMP Wed Mar 12 13:22:15 GMT 2014
CPU: ARMv7 Processor [412fc09a] revision 10 (ARMv7), cr=10c53c7d
CPU: PIPT / VIPT nonaliasing data cache, VIPT aliasing instruction cache
Machine model: SolidRun Cubox-i Dual/Quad
memblock_reserve: [0x00000010008240-0x0000001112c277] flags 0x0 arm_memblock_init+0x54/0x1d4
memblock_reserve: [0x00000020000040-0x000000201caa57] flags 0x0 arm_memblock_init+0x134/0x1d4
memblock_reserve: [0x00000010004000-0x00000010007fff] flags 0x0 arm_mm_memblock_reserve+0x1c/0x24
memblock_reserve: [0x00000018000000-0x0000001800b07f] flags 0x0 arm_dt_memblock_reserve+0x2c/0x70
memblock_reserve: [0x00000018000000-0x0000001800afff] flags 0x0 arm_dt_memblock_reserve+0x68/0x70
memblock_reserve: [0x00000020000040-0x000000201caa57] flags 0x0 arm_dt_memblock_reserve+0x68/0x70
memblock_reserve: [0x0000003b800000-0x0000003f7fffff] flags 0x0 memblock_alloc_base_nid+0x40/0x54
cma: CMA: reserved 64 MiB at 3b800000
MEMBLOCK configuration:

   memory size = 0x80000000 reserved size = 0x52fdad0
   memory.cnt  = 0x1
   memory[0x0]     [0x00000010000000-0x0000008fffffff], 0x80000000 bytes flags: 0x0
   reserved.cnt  = 0x5
   reserved[0x0]   [0x00000010004000-0x00000010007fff], 0x4000 bytes flags: 0x0
   reserved[0x1]   [0x00000010008240-0x0000001112c277], 0x1124038 bytes flags: 0x0
   reserved[0x2]   [0x00000018000000-0x0000001800b07f], 0xb080 bytes flags: 0x0
   reserved[0x3]   [0x00000020000040-0x000000201caa57], 0x1caa18 bytes flags: 0x0
   reserved[0x4]   [0x0000003b800000-0x0000003f7fffff], 0x4000000 bytes flags: 0x0

Memory policy: Data cache writealloc
memblock_reserve: [0x0000003b7fffd8-0x0000003b7fffff] flags 0x0 memblock_alloc_base_nid+0x40/0x54
memblock_reserve: [0x0000003b7fe000-0x0000003b7fefff] flags 0x0 memblock_alloc_base_nid+0x40/0x54
memblock_reserve: [0x0000003b7fd000-0x0000003b7fdfff] flags 0x0 memblock_alloc_base_nid+0x40/0x54
memblock_reserve: [0x0000003b7fc000-0x0000003b7fcfff] flags 0x0 memblock_alloc_base_nid+0x40/0x54
...

So it looks like allocations which must come from lowmem aren't being limited to lowmem.

Try booting a machine with 2G of RAM with page offset set to 3GB and highmem enabled - it will fail as per the above.

In fact, if we look at sanity_check_meminfo() post that patch, it's clearly wrong:

Quoted 13 lines

          for_each_memblock(memory, reg) {
                  phys_addr_t block_start = reg->base;
                  phys_addr_t block_end = reg->base + reg->size;
                  phys_addr_t size_limit = reg->size;

                  if (reg->base >= vmalloc_limit)
                          highmem = 1;
                  else
                          size_limit = vmalloc_limit - reg->base;
...
                  if (!highmem) {
                          if (block_end > arm_lowmem_limit)
                                  arm_lowmem_limit = block_end;

In v3, above was

		arm_lowmem_limit = reg->base + size_limit;

so, it has worked somehow, even arm_lowmem_limit can point on non existed address. It was changed because of my comment - sorry.

I think, it should be smth like:

if (!highmem) {
	if (block_end > arm_lowmem_limit)
		if (reg->size > size_limit)
			arm_lowmem_limit = vmalloc_limit;
		else
			arm_lowmem_limit = block_end;

I've created and attached the patch which allows me to boot on keystone.

Quoted 2 lines

...

                  }

regards,
-grygorii

From 3a210330f15c4cfc6728d83b28750c696d9eaefb Mon Sep 17 00:00:00 2001
From: Grygorii Strashko <x0174654@uglx0174654.(none)>
Date: Wed, 12 Mar 2014 18:04:02 +0200
Subject: [PATCH] arm: get rid of meminfo: fix lowmem_limit calculation

Signed-off-by Grygorii Strashko <x0174654@uglx0174654.(none)>

---
  arch/arm/mm/mmu.c |   10 +++++++---
  1 file changed, 7 insertions(+), 3 deletions(-)

arch/arm/mm/mmu.c +6 -2

index c3ae96c..f8b5175 100644
@@ -1096,8 +1096,12 @@ void __init sanity_check_meminfo(void)
  		}
 
  		if (!highmem) {
-			if (block_end > arm_lowmem_limit)
-				arm_lowmem_limit = block_end;
+			if (block_end > arm_lowmem_limit) {
+				if (reg->size > size_limit)
+					arm_lowmem_limit = vmalloc_limit;
+				else
+					arm_lowmem_limit = block_end;
+			}
 
 
  			/*

arch/arm/mm/mmu.c +31 -76

index b68c6b2..c3ae96c 100644
@@ -1061,74 +1061,44 @@ phys_addr_t arm_lowmem_limit __initdata = 0;
 void __init sanity_check_meminfo(void)
 {
 	phys_addr_t memblock_limit = 0;
-	int i, j, highmem = 0;
+	int highmem = 0;
 	phys_addr_t vmalloc_limit = __pa(vmalloc_min - 1) + 1;
+	struct memblock_region *reg;
 
-	for (i = 0, j = 0; i < meminfo.nr_banks; i++) {
-		struct membank *bank = &meminfo.bank[j];
-		phys_addr_t size_limit;
-
-		*bank = meminfo.bank[i];
-		size_limit = bank->size;
+	for_each_memblock(memory, reg) {
+		phys_addr_t block_start = reg->base;
+		phys_addr_t block_end = reg->base + reg->size;
+		phys_addr_t size_limit = reg->size;
 
-		if (bank->start >= vmalloc_limit)
+		if (reg->base >= vmalloc_limit)
 			highmem = 1;
 		else
-			size_limit = vmalloc_limit - bank->start;
+			size_limit = vmalloc_limit - reg->base;
 
-		bank->highmem = highmem;
 
-#ifdef CONFIG_HIGHMEM
-		/*
-		 * Split those memory banks which are partially overlapping
-		 * the vmalloc area greatly simplifying things later.
-		 */
-		if (!highmem && bank->size > size_limit) {
-			if (meminfo.nr_banks >= NR_BANKS) {
-				printk(KERN_CRIT "NR_BANKS too low, "
-						 "ignoring high memory\n");
-			} else {
-				memmove(bank + 1, bank,
-					(meminfo.nr_banks - i) * sizeof(*bank));
-				meminfo.nr_banks++;
-				i++;
-				bank[1].size -= size_limit;
-				bank[1].start = vmalloc_limit;
-				bank[1].highmem = highmem = 1;
-				j++;
+		if (!IS_ENABLED(CONFIG_HIGHMEM) || cache_is_vipt_aliasing()) {
+
+			if (highmem) {
+				pr_notice("Ignoring RAM at %pa-%pa (!CONFIG_HIGHMEM)\n",
+					&block_start, &block_end);
+				memblock_remove(reg->base, reg->size);
+				continue;
 			}
-			bank->size = size_limit;
-		}
-#else
-		/*
-		 * Highmem banks not allowed with !CONFIG_HIGHMEM.
-		 */
-		if (highmem) {
-			printk(KERN_NOTICE "Ignoring RAM at %.8llx-%.8llx "
-			       "(!CONFIG_HIGHMEM).\n",
-			       (unsigned long long)bank->start,
-			       (unsigned long long)bank->start + bank->size - 1);
-			continue;
-		}
 
-		/*
-		 * Check whether this memory bank would partially overlap
-		 * the vmalloc area.
-		 */
-		if (bank->size > size_limit) {
-			printk(KERN_NOTICE "Truncating RAM at %.8llx-%.8llx "
-			       "to -%.8llx (vmalloc region overlap).\n",
-			       (unsigned long long)bank->start,
-			       (unsigned long long)bank->start + bank->size - 1,
-			       (unsigned long long)bank->start + size_limit - 1);
-			bank->size = size_limit;
+			if (reg->size > size_limit) {
+				phys_addr_t overlap_size = reg->size - size_limit;
+
+				pr_notice("Truncating RAM at %pa-%pa to -%pa",
+				      &block_start, &block_end, &vmalloc_limit);
+				memblock_remove(vmalloc_limit, overlap_size);
+				block_end = vmalloc_limit;
+			}
 		}
-#endif
-		if (!bank->highmem) {
-			phys_addr_t bank_end = bank->start + bank->size;
 
-			if (bank_end > arm_lowmem_limit)
-				arm_lowmem_limit = bank_end;
+		if (!highmem) {
+			if (block_end > arm_lowmem_limit)
+				arm_lowmem_limit = block_end;
+
 
 			/*
 			 * Find the first non-section-aligned page, and point

@@ -1144,35 +1114,16 @@ void __init sanity_check_meminfo(void)
 			 * occurs before any free memory is mapped.
 			 */
 			if (!memblock_limit) {
-				if (!IS_ALIGNED(bank->start, SECTION_SIZE))
-					memblock_limit = bank->start;
-				else if (!IS_ALIGNED(bank_end, SECTION_SIZE))
-					memblock_limit = bank_end;
+				if (!IS_ALIGNED(block_start, SECTION_SIZE))
+					memblock_limit = block_start;
+				else if (!IS_ALIGNED(block_end, SECTION_SIZE))
+					memblock_limit = block_end;
 			}
-		}
-		j++;
-	}
-#ifdef CONFIG_HIGHMEM
-	if (highmem) {
-		const char *reason = NULL;
 
-		if (cache_is_vipt_aliasing()) {
-			/*
-			 * Interactions between kmap and other mappings
-			 * make highmem support with aliasing VIPT caches
-			 * rather difficult.
-			 */
-			reason = "with VIPT aliasing cache";
-		}
-		if (reason) {