Blackfin arch: move the init sections to the end of memory to help decrease memory fragementation

move the init sections to the end of memory, so that after they
are free, run time memory is all continugous - this should help decrease
memory fragementation. When doing this, we also pack some of the other
sections a little closer together, to make sure we don't waste memory.
To make this happen, we need to rename the .data.init_task section to
.init_task.data, so it doesn't get picked up by the linker script glob.

Signed-off-by: Robin Getz <robin.getz@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>

3 files changed, 33 insertions, 24 deletions

diff --git a/arch/blackfin/kernel/init_task.c b/arch/blackfin/kernel/init_task.c
index 673c860..c640154 100644
--- a/arch/blackfin/kernel/init_task.c
+++ b/arch/blackfin/kernel/init_task.c
@@ -57,5 +57,5 @@ EXPORT_SYMBOL(init_task);
  * "init_task" linker map entry.
  */
 union thread_union init_thread_union
-    __attribute__ ((__section__(".data.init_task"))) = {
+    __attribute__ ((__section__(".init_task.data"))) = {
 INIT_THREAD_INFO(init_task)};
diff --git a/arch/blackfin/kernel/setup.c b/arch/blackfin/kernel/setup.c
index 934234f..dfe06b0 100644
--- a/arch/blackfin/kernel/setup.c
+++ b/arch/blackfin/kernel/setup.c
@@ -236,7 +236,7 @@ void __init setup_arch(char **cmdline_p)
 	/* by now the stack is part of the init task */
 	memory_end = _ramend - DMA_UNCACHED_REGION;
 
-	_ramstart = (unsigned long)__bss_stop;
+	_ramstart = (unsigned long)_end;
 	memory_start = PAGE_ALIGN(_ramstart);
 
 #if defined(CONFIG_MTD_UCLINUX)
@@ -285,7 +285,7 @@ void __init setup_arch(char **cmdline_p)
 	}
 
 	/* Relocate MTD image to the top of memory after the uncached memory area */
-	dma_memcpy((char *)memory_end, __bss_stop, mtd_size);
+	dma_memcpy((char *)memory_end, _end, mtd_size);
 
 	memory_mtd_start = memory_end;
 	_ebss = memory_mtd_start;	/* define _ebss for compatible */
@@ -357,10 +357,10 @@ void __init setup_arch(char **cmdline_p)
 	printk(KERN_INFO "Memory map:\n"
 	       KERN_INFO "  text      = 0x%p-0x%p\n"
 	       KERN_INFO "  rodata    = 0x%p-0x%p\n"
+	       KERN_INFO "  bss       = 0x%p-0x%p\n"
 	       KERN_INFO "  data      = 0x%p-0x%p\n"
 	       KERN_INFO "    stack   = 0x%p-0x%p\n"
 	       KERN_INFO "  init      = 0x%p-0x%p\n"
-	       KERN_INFO "  bss       = 0x%p-0x%p\n"
 	       KERN_INFO "  available = 0x%p-0x%p\n"
 #ifdef CONFIG_MTD_UCLINUX
 	       KERN_INFO "  rootfs    = 0x%p-0x%p\n"
@@ -370,10 +370,10 @@ void __init setup_arch(char **cmdline_p)
 #endif
 	       , _stext, _etext,
 	       __start_rodata, __end_rodata,
+	       __bss_start, __bss_stop,
 	       _sdata, _edata,
 	       (void *)&init_thread_union, (void *)((int)(&init_thread_union) + 0x2000),
 	       __init_begin, __init_end,
-	       __bss_start, __bss_stop,
 	       (void *)_ramstart, (void *)memory_end
 #ifdef CONFIG_MTD_UCLINUX
 	       , (void *)memory_mtd_start, (void *)(memory_mtd_start + mtd_size)
diff --git a/arch/blackfin/kernel/vmlinux.lds.S b/arch/blackfin/kernel/vmlinux.lds.S
index 9b75bc8..5bdc0f7 100644
--- a/arch/blackfin/kernel/vmlinux.lds.S
+++ b/arch/blackfin/kernel/vmlinux.lds.S
@@ -41,6 +41,9 @@ _jiffies = _jiffies_64;
 SECTIONS
 {
 	. = CONFIG_BOOT_LOAD;
+	/* Neither the text, ro_data or bss section need to be aligned
+	 * So pack them back to back
+	 */
 	.text :
 	{
 		__text = .;
@@ -58,22 +61,25 @@ SECTIONS
 		*(__ex_table)
 		___stop___ex_table = .;
 
-		. = ALIGN(4);
 		__etext = .;
 	}
 
-	RO_DATA(PAGE_SIZE)
+	/* Just in case the first read only is a 32-bit access */
+	RO_DATA(4)
+
+	.bss :
+	{
+		. = ALIGN(4);
+		___bss_start = .;
+		*(.bss .bss.*)
+		*(COMMON)
+		___bss_stop = .;
+	}
 
 	.data :
 	{
-		/* make sure the init_task is aligned to the
-		 * kernel thread size so we can locate the kernel
-		 * stack properly and quickly.
-		 */
 		__sdata = .;
-		. = ALIGN(THREAD_SIZE);
-		*(.data.init_task)
-
+		/* This gets done first, so the glob doesn't suck it in */
 		. = ALIGN(32);
 		*(.data.cacheline_aligned)
 
@@ -81,10 +87,22 @@ SECTIONS
 		*(.data.*)
 		CONSTRUCTORS
 
+		/* make sure the init_task is aligned to the
+		 * kernel thread size so we can locate the kernel
+		 * stack properly and quickly.
+		 */
 		. = ALIGN(THREAD_SIZE);
+		*(.init_task.data)
+
 		__edata = .;
 	}
 
+	/* The init section should be last, so when we free it, it goes into
+	 * the general memory pool, and (hopefully) will decrease fragmentation
+	 * a tiny bit. The init section has a _requirement_ that it be
+	 * PAGE_SIZE aligned
+	 */
+	. = ALIGN(PAGE_SIZE);
 	___init_begin = .;
 
 	.init.text :
@@ -179,16 +197,7 @@ SECTIONS
 	. = ALIGN(PAGE_SIZE);
 	___init_end = .;
 
-	.bss :
-	{
-		. = ALIGN(4);
-		___bss_start = .;
-		*(.bss .bss.*)
-		*(COMMON)
-		. = ALIGN(4);
-		___bss_stop = .;
-		__end = .;
-	}
+	__end =.;
 
 	STABS_DEBUG