Menu
▾
▴
[Aceos-developers] SF.net SVN: aceos:[308] src
|
From: <sam...@us...> - 2008-08-03 11:48:22
|
Revision: 308
http://aceos.svn.sourceforge.net/aceos/?rev=308&view=rev
Author: samueldotj
Date: 2008-08-03 11:48:30 +0000 (Sun, 03 Aug 2008)
Log Message:
-----------
Fixed few bugs in pmem.c.
Added code to print the system memory map.
Now the kernel boots without triple fault in VMWare and also in real machine.
Modified Paths:
--------------
src/include/kernel/mm/pmem.h
src/kernel/i386/mm/pmem_init.c
src/kernel/kernel.ld
src/kernel/mm/virtual_page.c
src/kernel/mm/vm.c
Modified: src/include/kernel/mm/pmem.h
===================================================================
--- src/include/kernel/mm/pmem.h 2008-08-01 11:00:10 UTC (rev 307)
+++ src/include/kernel/mm/pmem.h 2008-08-03 11:48:30 UTC (rev 308)
@@ -19,14 +19,30 @@
#define MAX_MEMORY_AREAS 32
#define MAX_PHYSICAL_REGIONS 16
+/*http://www.ctyme.com/intr/rb-1741.htm
+Values for System Memory Map address type:
+01h memory, available to OS
+02h reserved, not available (e.g. system ROM, memory-mapped device)
+03h ACPI Reclaim Memory (usable by OS after reading ACPI tables)
+04h ACPI NVS Memory (OS is required to save this memory between NVS
+sessions)
+other not defined yet -- treat as Reserved*/
+#define PMEM_TYPE_AVAILABLE 0x1
+#define PMEM_TYPE_RESERVED 0x2
+#define PMEM_TYPE_ACPI_RECLAIM 0x3
+#define PMEM_TYPE_ACPI_NVS 0x4
+
+
typedef struct physical_memory_region
{
UINT32 start_physical_address; //starting physical address
UINT32 end_physical_address; //ending physical address
-
+
VIRTUAL_PAGE_PTR virtual_page_array; //virutal page array for this region
- UINT32 virtual_page_count;
+ UINT32 virtual_page_count; //total pages in this region
+ UINT32 type; //usable, reserved, etc
+
}PHYSICAL_MEMORY_REGION, * PHYSICAL_MEMORY_REGION_PTR;
typedef struct memory_area
Modified: src/kernel/i386/mm/pmem_init.c
===================================================================
--- src/kernel/i386/mm/pmem_init.c 2008-08-01 11:00:10 UTC (rev 307)
+++ src/kernel/i386/mm/pmem_init.c 2008-08-03 11:48:30 UTC (rev 308)
@@ -16,7 +16,7 @@
#include <kernel/mm/pmem.h>
#include <kernel/i386/pmem.h>
-extern UINT32 ebss;
+extern UINT32 ebss, kernel_code_start;
#define BOOT_ADDRESS(addr) (((UINT32)addr - KERNEL_VIRTUAL_ADDRESS_TEXT_START)+KERNEL_PHYSICAL_ADDRESS_LOAD )
@@ -50,43 +50,47 @@
static UINT32 InitMemoryArea(MEMORY_AREA_PTR ma_pa, MEMORY_MAP_PTR memory_map_array, int memory_map_count)
{
UINT32 total_size = 0; //total bytes occupied by virtual page array
+ int i;
+ UINT32 kernel_start = PAGE_ALIGN_UP ( BOOT_ADDRESS(&kernel_code_start) );
UINT32 kernel_end = PAGE_ALIGN_UP ( BOOT_ADDRESS(&ebss) );
- int i;
- for(i=0; i<memory_map_count; i++)
+
+ ma_pa->physical_memory_regions_count = 0;
+ for(i=0; i<memory_map_count && i<MAX_PHYSICAL_REGIONS; i++)
{
- //use only available RAM
- if( memory_map_array[i].type == 1 )
+ PHYSICAL_MEMORY_REGION_PTR pmr_pa;
+ //get physical address of the physical memory region
+ pmr_pa = &ma_pa->physical_memory_regions[i];
+
+ pmr_pa->start_physical_address = memory_map_array[i].base_addr_low;
+ pmr_pa->end_physical_address = pmr_pa->start_physical_address + memory_map_array[i].length_low;
+ pmr_pa->virtual_page_array = NULL;
+ pmr_pa->virtual_page_count = 0;
+ pmr_pa->type = memory_map_array[i].type;
+
+ ma_pa->physical_memory_regions_count++;
+ //create vm_page_array only for available RAM
+ if( pmr_pa->type == PMEM_TYPE_AVAILABLE )
{
- PHYSICAL_MEMORY_REGION_PTR pmr_pa;
int total_virtual_pages, virtual_page_array_size;
-
- //if memory start is less than 1MB then skip it because we cant use it
- if ( memory_map_array[i].base_addr_low < (1024 * 1024) )
- continue;
+ UINT32 region_size;
+
//we cant use the kernel code and data area
- if ( memory_map_array[i].base_addr_low < kernel_end )
- {
- memory_map_array[i].length_low -= kernel_end - memory_map_array[i].base_addr_low;
- memory_map_array[i].base_addr_low = kernel_end;
- }
+ if ( ( pmr_pa->start_physical_address <= kernel_start && pmr_pa->end_physical_address > kernel_start ) ||
+ ( pmr_pa->start_physical_address >= kernel_start && pmr_pa->start_physical_address < kernel_end ) )
+ pmr_pa->start_physical_address = kernel_end;
+ region_size = pmr_pa->end_physical_address - pmr_pa->start_physical_address;
+
//calculate virtual page array size
- total_virtual_pages = memory_map_array[i].length_low / PAGE_SIZE;
+ total_virtual_pages = region_size / PAGE_SIZE;
virtual_page_array_size = PAGE_ALIGN_UP ( total_virtual_pages * sizeof(VIRTUAL_PAGE) );
//adjust total_virtual_pages
- total_virtual_pages = (memory_map_array[i].length_low - virtual_page_array_size) / PAGE_SIZE;
+ total_virtual_pages = (region_size - virtual_page_array_size) / PAGE_SIZE;
+ pmr_pa->virtual_page_array = (VIRTUAL_PAGE_PTR)pmr_pa->start_physical_address; /*this physical address will be converted into virtual address by InitKernelPagediretory()*/
+ pmr_pa->virtual_page_count = total_virtual_pages;
+
total_size += virtual_page_array_size;
-
- //get physical address of the physical memory region
- pmr_pa = &ma_pa->physical_memory_regions[ma_pa->physical_memory_regions_count];
-
- pmr_pa->start_physical_address = memory_map_array[i].base_addr_low + virtual_page_array_size;
- pmr_pa->end_physical_address = memory_map_array[i].base_addr_low + memory_map_array[i].length_low;
- pmr_pa->virtual_page_array = (VIRTUAL_PAGE_PTR)memory_map_array[i].base_addr_low;
- pmr_pa->virtual_page_count = total_virtual_pages;
-
- ma_pa->physical_memory_regions_count++;
}
}
return total_size;
@@ -100,26 +104,30 @@
int i;
MEMORY_AREA_PTR ma_pa;
ma_pa = (MEMORY_AREA_PTR)BOOT_ADDRESS ( &memory_areas[0] );
+
for(i=ma_pa->physical_memory_regions_count-1; i >= 0 ; i--)
{
PHYSICAL_MEMORY_REGION_PTR pmr_pa;
void * pa;
pmr_pa = &ma_pa->physical_memory_regions[i];
- //skip the region if there is not enough space
- if ( (pmr_pa->end_physical_address - pmr_pa->start_physical_address) <= PAGE_SIZE )
- continue;
- //get the first page
- pa = (void *)pmr_pa->start_physical_address;
-
- //adjust the region
- pmr_pa->start_physical_address += PAGE_SIZE;
- pmr_pa->virtual_page_count--;
-
- memset(pa, 0, PAGE_SIZE);
-
- return pa;
+ //if the region has enough free space allocate and return
+ if ( pmr_pa->type == PMEM_TYPE_AVAILABLE && (pmr_pa->end_physical_address - pmr_pa->start_physical_address) > PAGE_SIZE )
+ {
+ //get the last page
+ pa = (void *)(pmr_pa->end_physical_address-PAGE_SIZE);
+
+ //adjust the region
+ pmr_pa->end_physical_address = (UINT32)pa;
+ pmr_pa->virtual_page_count--;
+
+ memset(pa, 0, PAGE_SIZE);
+
+ return pa;
+ }
}
+ /*panic if we dont find a free physical page*/
+ asm("cli;hlt");
return NULL;
}
@@ -175,16 +183,18 @@
PHYSICAL_MEMORY_REGION_PTR pmr_pa;
pmr_pa = &ma_pa->physical_memory_regions[i];
- physical_address = (UINT32)pmr_pa->virtual_page_array;
- end_address = physical_address + (pmr_pa->virtual_page_count * sizeof(VIRTUAL_PAGE));
- pmr_pa->virtual_page_array = (VIRTUAL_PAGE_PTR)va;
- do
+ if ( pmr_pa->type == PMEM_TYPE_AVAILABLE )
{
- EnterKernelPageTableEntry( va, physical_address);
- physical_address += PAGE_SIZE;
- va += PAGE_SIZE;
- }while( physical_address <= ( end_address ) );
-
+ physical_address = (UINT32)pmr_pa->virtual_page_array;
+ end_address = physical_address + (pmr_pa->virtual_page_count * sizeof(VIRTUAL_PAGE));
+ pmr_pa->virtual_page_array = (VIRTUAL_PAGE_PTR)va;
+ do
+ {
+ EnterKernelPageTableEntry( va, physical_address);
+ physical_address += PAGE_SIZE;
+ va += PAGE_SIZE;
+ }while( physical_address <= ( end_address ) );
+ }
}
/*self mapping*/
k_page_dir[PT_SELF_MAP_INDEX] = ((UINT32)k_page_dir) | KERNEL_PTE_FLAG;
@@ -217,7 +227,8 @@
page_table = (PAGE_TABLE_ENTRY_PTR) pa;
/*enter pde*/
- k_page_dir[pd_index].all = PA_TO_PFN(pa) | KERNEL_PTE_FLAG;
+ k_page_dir[pd_index].all = KERNEL_PTE_FLAG;
+ k_page_dir[pd_index]._.page_table_pfn = PA_TO_PFN(pa);
}
else
{
@@ -228,7 +239,8 @@
//enter pte in the page table.
if ( !page_table[ pt_index ]._.present )
{
- page_table[pt_index].all = PA_TO_PFN(pa) | KERNEL_PTE_FLAG;
+ page_table[pt_index].all = KERNEL_PTE_FLAG;
+ page_table[pt_index]._.page_pfn = PA_TO_PFN(pa);
}
}
/*! 1) This phase will removes the unnessary page table entries that is created before enabling paging.
@@ -237,6 +249,8 @@
void InitPhysicalMemoryManagerPhaseII()
{
int i;
+ UINT32 kernel_start = PAGE_ALIGN_UP ( BOOT_ADDRESS(&kernel_code_start) );
+ UINT32 kernel_end = PAGE_ALIGN_UP ( BOOT_ADDRESS(&ebss) );
/*initialize the kernel physical map*/
InitSpinLock( &kernel_physical_map.lock );
@@ -251,11 +265,27 @@
for(i=0; i<memory_area_count; i++ )
{
int j;
+ kprintf("System map: START END PAGES TYPE\n");
for(j=0; j<memory_areas[i].physical_memory_regions_count; j++ )
{
PHYSICAL_MEMORY_REGION_PTR pmr = &memory_areas[i].physical_memory_regions[j];
- InitVirtualPageArray(pmr->virtual_page_array, pmr->virtual_page_count, pmr->start_physical_address);
- vm_data.total_memory_pages += pmr->virtual_page_count;
+ kprintf(" %9p %9p %9d %10s\n", pmr->start_physical_address, pmr->end_physical_address, pmr->virtual_page_count,
+ pmr->type == PMEM_TYPE_AVAILABLE ? "Available" : pmr->type == PMEM_TYPE_ACPI_RECLAIM ? "ACPI Reclaim" : pmr->type == PMEM_TYPE_ACPI_NVS ? "ACPI NVS" : "Reserved" );
+
+ if ( pmr->start_physical_address < (1024*1024) )
+ continue;
+ //we cant use the kernel code and data area
+ if ( ( pmr->start_physical_address <= kernel_start && pmr->end_physical_address > kernel_start ) ||
+ ( pmr->start_physical_address >= kernel_start && pmr->start_physical_address < kernel_end ) )
+ {
+ continue;
+ }
+
+ if ( pmr->type == PMEM_TYPE_AVAILABLE )
+ {
+ InitVirtualPageArray(pmr->virtual_page_array, pmr->virtual_page_count, pmr->start_physical_address);
+ vm_data.total_memory_pages += pmr->virtual_page_count;
+ }
}
}
Modified: src/kernel/kernel.ld
===================================================================
--- src/kernel/kernel.ld 2008-08-01 11:00:10 UTC (rev 307)
+++ src/kernel/kernel.ld 2008-08-03 11:48:30 UTC (rev 308)
@@ -15,6 +15,7 @@
}
.text :
{
+ kernel_code_start = .;
*(.text)
}
.rodata :
Modified: src/kernel/mm/virtual_page.c
===================================================================
--- src/kernel/mm/virtual_page.c 2008-08-01 11:00:10 UTC (rev 307)
+++ src/kernel/mm/virtual_page.c 2008-08-03 11:48:30 UTC (rev 308)
@@ -261,7 +261,7 @@
PHYSICAL_MEMORY_REGION_PTR pmr;
for(j=0;j<memory_areas[i].physical_memory_regions_count;j++)
{
- pmr = memory_areas[i].physical_memory_regions;
+ pmr = &memory_areas[i].physical_memory_regions[j];
if ( physical_address >= pmr->start_physical_address && physical_address < pmr->end_physical_address )
{
UINT32 index;
Modified: src/kernel/mm/vm.c
===================================================================
--- src/kernel/mm/vm.c 2008-08-01 11:00:10 UTC (rev 307)
+++ src/kernel/mm/vm.c 2008-08-03 11:48:30 UTC (rev 308)
@@ -48,7 +48,7 @@
/*complete physical memory initialization*/
InitPhysicalMemoryManagerPhaseII();
- kprintf("System memory: %dMB (PAGE_SIZE %d)\n", (vm_data.total_memory_pages * PAGE_SIZE) / (1024*1024), PAGE_SIZE );
+ kprintf("Total memory: %d KB (PAGE_SIZE %d)\n", (vm_data.total_memory_pages * PAGE_SIZE) / (1024), PAGE_SIZE );
/*Initialize the kernel memory allocator*/
kmem_start_va = kernel_free_virtual_address;
This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.
|
