Re: [PATCH v2 10/13] arc: use FLATMEM with freeing of unused memory map instead of DISCONTIGMEM

2020-11-23 Thread Vineet Gupta

On 11/16/20 10:57 PM, Mike Rapoport wrote:

On Tue, Nov 17, 2020 at 06:40:16AM +, Vineet Gupta wrote:

Hi Mike,

On 11/1/20 9:04 AM, Mike Rapoport wrote:

From: Mike Rapoport 

Currently ARC uses DISCONTIGMEM to cope with sparse physical memory address
space on systems with 2 memory banks. While DISCONTIGMEM avoids wasting
memory on unpopulated memory map, it adds both memory and CPU overhead
relatively to FLATMEM. Moreover, DISCONTINGMEM is generally considered
deprecated.

The obvious replacement for DISCONTIGMEM would be SPARSEMEM, but it is also
less efficient than FLATMEM in pfn_to_page() and page_to_pfn() conversions.
Besides it requires tuning of SECTION_SIZE which is not trivial for
possible ARC memory configuration.

Since the memory map for both banks is always allocated from the "lowmem"
bank, it is possible to use FLATMEM for two-bank configuration and simply
free the unused hole in the memory map. All is required for that is to
provide ARC-specific pfn_valid() that will take into account actual
physical memory configuration and define HAVE_ARCH_PFN_VALID.

The resulting kernel image configured with defconfig + HIGHMEM=y is
smaller:

$ size a/vmlinux b/vmlinux
 textdata bss dec hex filename
4673503 1245456  279756 6198715  5e95bb a/vmlinux
4658706 1246864  279756 6185326  5e616e b/vmlinux

$ ./scripts/bloat-o-meter a/vmlinux b/vmlinux
add/remove: 28/30 grow/shrink: 42/399 up/down: 10986/-29025 (-18039)
...
Total: Before=4709315, After=4691276, chg -0.38%

Booting nSIM with haps_ns.dts results in the following memory usage
reports:

a:
Memory: 1559104K/1572864K available (3531K kernel code, 595K rwdata, 752K 
rodata, 136K init, 275K bss, 13760K reserved, 0K cma-reserved, 1048576K highmem)

b:
Memory: 1559112K/1572864K available (3519K kernel code, 594K rwdata, 752K 
rodata, 136K init, 280K bss, 13752K reserved, 0K cma-reserved, 1048576K highmem)

Signed-off-by: Mike Rapoport 


To avoid any surprises later, I tested that highmem was actually working 
on real hardware (HSDK-4xD dev platform) with modified hsdk.dts to 
enable 2 GB of memory.


   reg = <0x0 0x8000 0x0 0x4000/* 1 GB low mem */
- 0x1 0x 0x0 0x4000>;  /* 1 GB highmem PAE */
+ 0x0 0x 0x0 0x4000>;  /* 1 GB highmem low phy mem*/
};

A simple malloc+memset program can allocate upto 1.98 GB of memory.

# cat /proc/meminfo | grep Mem
MemTotal:2077984 kB
MemFree: 2047512 kB
MemAvailable:2005712 kB

# /oom 1000 &
# malloc 1000 MB
# Done memset, sleeping for 20 secs

# cat /proc/meminfo | grep Mem
MemTotal:2077984 kB
MemFree: 1163888 kB
MemAvailable:1122088 kB

# /oom 980 &
# malloc 980 MB
# Done memset, sleeping for 20 secs

# cat /proc/meminfo | grep Mem
MemTotal:2077984 kB
MemFree:  239096 kB
MemAvailable: 197296 kB

# Done free.
Done free.

So this is all hunky-dory. Thanks for working on this Mike and improving 
things.


Acked-by: Vineet Gupta 

-Vineet



Sorry this fell through the cracks. Do you have a branch I can checkout
and do a quick test.


It's in mmotm and in my tree:
https://git.kernel.org/pub/scm/linux/kernel/git/rppt/linux.git 
memory-models/rm-discontig/v0


Thx,
-Vineet


---
   arch/arc/Kconfig|  3 ++-
   arch/arc/include/asm/page.h | 20 +---
   arch/arc/mm/init.c  | 29 ++---
   3 files changed, 41 insertions(+), 11 deletions(-)

diff --git a/arch/arc/Kconfig b/arch/arc/Kconfig
index 0a89cc9def65..c874f8ab0341 100644
--- a/arch/arc/Kconfig
+++ b/arch/arc/Kconfig
@@ -67,6 +67,7 @@ config GENERIC_CSUM
   
   config ARCH_DISCONTIGMEM_ENABLE

def_bool n
+   depends on BROKEN
   
   config ARCH_FLATMEM_ENABLE

def_bool y
@@ -506,7 +507,7 @@ config LINUX_RAM_BASE
   
   config HIGHMEM

bool "High Memory Support"
-   select ARCH_DISCONTIGMEM_ENABLE
+   select HAVE_ARCH_PFN_VALID
help
  With ARC 2G:2G address split, only upper 2G is directly addressable by
  kernel. Enable this to potentially allow access to rest of 2G and PAE
diff --git a/arch/arc/include/asm/page.h b/arch/arc/include/asm/page.h
index b0dfed0f12be..23e41e890eda 100644
--- a/arch/arc/include/asm/page.h
+++ b/arch/arc/include/asm/page.h
@@ -82,11 +82,25 @@ typedef pte_t * pgtable_t;
*/
   #define virt_to_pfn(kaddr)   (__pa(kaddr) >> PAGE_SHIFT)
   
-#define ARCH_PFN_OFFSET		virt_to_pfn(CONFIG_LINUX_RAM_BASE)

+/*
+ * When HIGHMEM is enabled we have holes in the memory map so we need
+ * pfn_valid() that takes into account the actual extents of the physical
+ * memory
+ */
+#ifdef CONFIG_HIGHMEM
+
+extern unsigned long arch_pfn_offset;
+#define ARCH_PFN_OFFSETarch_pfn_offset
+
+extern int pfn_valid(unsigned long pfn);
+#define pfn_valid  pfn_valid
   
-#ifdef CONFIG_FLATMEM

+#else /* CONFIG_HIGHMEM */
+
+#define ARCH_PFN_OFFSETvirt_to_pfn(CONFIG_LINUX

[GIT PULL] ARC fixes for 5.10-rc6

2020-11-23 Thread Vineet Gupta
Hi Linus,

A couple more stack unwinder related fixes. Please pull !

Thx,
-Vineet
--->
The following changes since commit 3b57533b460c8dc22a432684b7e8d22571f34d2e:

   ARC: [plat-hsdk] Remap CCMs super early in asm boot trampoline 
(2020-11-02 11:45:09 -0800)

are available in the Git repository at:

   git://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc.git/ 
tags/arc-5.10-rc6

for you to fetch changes up to f737561c709667013d832316dd3198a7fe3d1260:

   ARC: stack unwinding: reorganize how initial register state setup 
(2020-11-17 20:12:01 -0800)


ARC fixes for 5.10-rc6

  - More stack unwinding updates

  - Miscll minor fixes


Flavio Suligoi (1):
   ARC: mm: fix spelling mistakes

Gustavo Pimentel (1):
   ARC: bitops: Remove unecessary operation and value

Vineet Gupta (2):
   ARC: stack unwinding: don't assume non-current task is sleeping
   ARC: stack unwinding: reorganize how initial register state setup

  arch/arc/include/asm/bitops.h |  4 +---
  arch/arc/kernel/stacktrace.c  | 56 
---
  arch/arc/mm/tlb.c | 24 +--
  3 files changed, 44 insertions(+), 40 deletions(-)
___
linux-snps-arc mailing list
linux-snps-arc@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-snps-arc