[PATCH] Reserve low memory
From: Or Sagi
Date: Sun Oct 21 2007 - 12:54:32 EST
Adds a CONFIG_RESERVE_LOWMEM option which reserves all memory below
CONFIG_PHYSICAL_START address.
Cleans up some of the static early-boot memory allocations (nodemap, trampoline)
Useful in particular for running virtual machines which require a 1:1 mapping
of the physical memory. This lets a virtual machine control a dma-capable device
directly without hypervisor involvement.
Signed-off-by: Or Sagi <ors@xxxxxxxxx>
---
arch/x86/kernel/e820_64.c | 11 +++++++++++
arch/x86/kernel/head_64.S | 8 ++++----
arch/x86/kernel/setup_64.c | 3 +++
arch/x86/kernel/suspend_64.c | 2 +-
arch/x86/mm/init_64.c | 11 ++++++++++-
arch/x86/mm/numa_64.c | 12 ++++++++++++
arch/x86/mm/pageattr_64.c | 4 ++--
arch/x86_64/Kconfig | 10 +++++++++-
include/asm-x86/page_64.h | 7 ++++++-
include/asm-x86/pgtable_64.h | 2 +-
init/initramfs.c | 6 +++++-
11 files changed, 64 insertions(+), 12 deletions(-)
diff --git a/arch/x86/kernel/e820_64.c b/arch/x86/kernel/e820_64.c
index 5761686..29b927a 100644
--- a/arch/x86/kernel/e820_64.c
+++ b/arch/x86/kernel/e820_64.c
@@ -49,6 +49,14 @@ static unsigned long __initdata end_user_pfn = MAXMEM>>PAGE_SHIFT;
extern struct resource code_resource, data_resource;
+#ifdef CONFIG_RESERVE_LOWMEM
+static struct resource lowmem_res = {
+ .start = 1024*1024,
+ .end = CONFIG_PHYSICAL_START - 1,
+ .name = "reserved low memory"
+};
+#endif
+
/* Check for some hardcoded bad areas that early boot is not allowed to touch */
static inline int bad_addr(unsigned long *addrp, unsigned long size)
{
@@ -229,6 +237,9 @@ void __init e820_reserve_resources(void)
if (crashk_res.start != crashk_res.end)
request_resource(res, &crashk_res);
#endif
+#ifdef CONFIG_RESERVE_LOWMEM
+ request_resource(res, &lowmem_res);
+#endif
}
}
}
diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S
index b6167fe..9fd2be0 100644
--- a/arch/x86/kernel/head_64.S
+++ b/arch/x86/kernel/head_64.S
@@ -342,12 +342,12 @@ NEXT_PAGE(level2_ident_pgt)
PMDS(0x0000000000000000, __PAGE_KERNEL_LARGE_EXEC, PTRS_PER_PMD)
NEXT_PAGE(level2_kernel_pgt)
- /* 40MB kernel mapping. The kernel code cannot be bigger than that.
- When you change this change KERNEL_TEXT_SIZE in page.h too. */
+ /* kernel mapping. The kernel code cannot be bigger than that.
+ When you change this change KERNEL_TEXT_TOP in page.h too. */
/* (2^48-(2*1024*1024*1024)-((2^39)*511)-((2^30)*510)) = 0 */
- PMDS(0x0000000000000000, __PAGE_KERNEL_LARGE_EXEC|_PAGE_GLOBAL, KERNEL_TEXT_SIZE/PMD_SIZE)
+ PMDS(0x0000000000000000, __PAGE_KERNEL_LARGE_EXEC|_PAGE_GLOBAL, KERNEL_TEXT_TOP/PMD_SIZE)
/* Module mapping starts here */
- .fill (PTRS_PER_PMD - (KERNEL_TEXT_SIZE/PMD_SIZE)),8,0
+ .fill (PTRS_PER_PMD - (KERNEL_TEXT_TOP/PMD_SIZE)),8,0
NEXT_PAGE(level2_spare_pgt)
.fill 512,8,0
diff --git a/arch/x86/kernel/setup_64.c b/arch/x86/kernel/setup_64.c
index 31322d4..7db301e 100644
--- a/arch/x86/kernel/setup_64.c
+++ b/arch/x86/kernel/setup_64.c
@@ -363,6 +363,9 @@ void __init setup_arch(char **cmdline_p)
/* Reserve SMP trampoline */
reserve_bootmem_generic(SMP_TRAMPOLINE_BASE, 2*PAGE_SIZE);
#endif
+#ifdef CONFIG_RESERVE_LOWMEM
+ reserve_bootmem_generic(0x100000, CONFIG_PHYSICAL_START - 0x100000 + 1);
+#endif
#ifdef CONFIG_ACPI_SLEEP
/*
diff --git a/arch/x86/kernel/suspend_64.c b/arch/x86/kernel/suspend_64.c
index bc9f59c..1f17db7 100644
--- a/arch/x86/kernel/suspend_64.c
+++ b/arch/x86/kernel/suspend_64.c
@@ -207,7 +207,7 @@ static int res_kernel_text_pud_init(pud_t *pud, unsigned long start)
if (!pmd)
return -ENOMEM;
set_pud(pud + pud_index(start), __pud(__pa(pmd) | _KERNPG_TABLE));
- for (paddr = 0; paddr < KERNEL_TEXT_SIZE; pmd++, paddr += PMD_SIZE) {
+ for (paddr = 0; paddr < KERNEL_TEXT_TOP; pmd++, paddr += PMD_SIZE) {
unsigned long pe;
pe = __PAGE_KERNEL_LARGE_EXEC | _PAGE_GLOBAL | paddr;
diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c
index 1e3862e..4f320d2 100644
--- a/arch/x86/mm/init_64.c
+++ b/arch/x86/mm/init_64.c
@@ -329,7 +329,11 @@ static void __init find_early_table_space(unsigned long end)
/* RED-PEN putting page tables only on node 0 could
cause a hotspot and fill up ZONE_DMA. The page tables
need roughly 0.5KB per GB. */
- start = 0x8000;
+#ifdef CONFIG_RESERVE_LOWMEM
+ start = CONFIG_PHYSICAL_START;
+#else
+ start = 0x8000;
+#endif
table_start = find_e820_area(start, end, tables);
if (table_start == -1UL)
panic("Cannot find space for the kernel page tables");
@@ -526,6 +530,11 @@ void __init mem_init(void)
reservedpages = 0;
+#ifdef CONFIG_RESERVE_LOWMEM
+ /* see also mm/numa.c:allocate_cachealigned_memnodemap. */
+ reserve_bootmem_generic(0, 0xa0000);
+#endif
+
/* this will put all low memory onto the freelists */
#ifdef CONFIG_NUMA
totalram_pages = numa_free_all_bootmem();
diff --git a/arch/x86/mm/numa_64.c b/arch/x86/mm/numa_64.c
index 3d6926b..70940b6 100644
--- a/arch/x86/mm/numa_64.c
+++ b/arch/x86/mm/numa_64.c
@@ -82,7 +82,11 @@ static int __init allocate_cachealigned_memnodemap(void)
return 0;
pad = L1_CACHE_BYTES - 1;
+#ifdef CONFIG_RESERVE_LOWMEM
+ pad_addr = CONFIG_PHYSICAL_START + 0x8000;
+#else
pad_addr = 0x8000;
+#endif
nodemap_size = pad + memnodemapsize;
nodemap_addr = find_e820_area(pad_addr, end_pfn<<PAGE_SHIFT,
nodemap_size);
@@ -189,7 +193,15 @@ void __init setup_node_bootmem(int nodeid, unsigned long start, unsigned long en
start_pfn = start >> PAGE_SHIFT;
end_pfn = end >> PAGE_SHIFT;
+#ifdef CONFIG_RESERVE_LOWMEM
+ {
+ unsigned long s = start;
+ if (s < CONFIG_PHYSICAL_START) s += CONFIG_PHYSICAL_START;
+ node_data[nodeid] = early_node_mem(nodeid, s, end, pgdat_size);
+ }
+#else
node_data[nodeid] = early_node_mem(nodeid, start, end, pgdat_size);
+#endif
if (node_data[nodeid] == NULL)
return;
nodedata_phys = __pa(node_data[nodeid]);
diff --git a/arch/x86/mm/pageattr_64.c b/arch/x86/mm/pageattr_64.c
index c7b7dfe..158a448 100644
--- a/arch/x86/mm/pageattr_64.c
+++ b/arch/x86/mm/pageattr_64.c
@@ -188,7 +188,7 @@ int change_page_attr_addr(unsigned long address, int numpages, pgprot_t prot)
int i;
if (address >= __START_KERNEL_map
- && address < __START_KERNEL_map + KERNEL_TEXT_SIZE) {
+ && address < __START_KERNEL_map + KERNEL_TEXT_TOP) {
address = (unsigned long)__va(__pa(address));
kernel_map = 1;
}
@@ -204,7 +204,7 @@ int change_page_attr_addr(unsigned long address, int numpages, pgprot_t prot)
}
/* Handle kernel mapping too which aliases part of the
* lowmem */
- if (__pa(address) < KERNEL_TEXT_SIZE) {
+ if (__pa(address) < KERNEL_TEXT_TOP) {
unsigned long addr2;
pgprot_t prot2;
addr2 = __START_KERNEL_map + __pa(address);
diff --git a/arch/x86_64/Kconfig b/arch/x86_64/Kconfig
index aab25f3..391f4e0 100644
--- a/arch/x86_64/Kconfig
+++ b/arch/x86_64/Kconfig
@@ -576,6 +576,14 @@ config KEXEC
support. As of this writing the exact hardware interface is
strongly in flux, so no good recommendation can be made.
+config RESERVE_LOWMEM
+ bool "Reserve memory below CONFIG_PHYSICAL_START (EXPERIMENTAL)"
+ depends on EXPERIMENTAL && !RELOCATABLE && !CRASH_DUMP
+ help
+ Does not make use of memory below CONFIG_PHYSICAL_START.
+ This is work in progress, and low memory may still be used
+ or run over during bootstrap and/or shutdown.
+
config CRASH_DUMP
bool "kernel crash dumps (EXPERIMENTAL)"
depends on EXPERIMENTAL
@@ -607,7 +615,7 @@ config RELOCATABLE
(CONFIG_PHYSICAL_START) is ignored.
config PHYSICAL_START
- hex "Physical address where the kernel is loaded" if (EMBEDDED || CRASH_DUMP)
+ hex "Physical address where the kernel is loaded" if (EMBEDDED || CRASH_DUMP || RESERVE_LOWMEM)
default "0x200000"
help
This gives the physical address where the kernel is loaded. It
diff --git a/include/asm-x86/page_64.h b/include/asm-x86/page_64.h
index c3b52bc..e25221e 100644
--- a/include/asm-x86/page_64.h
+++ b/include/asm-x86/page_64.h
@@ -103,7 +103,12 @@ extern unsigned long phys_base;
#define __VIRTUAL_MASK_SHIFT 48
#define __VIRTUAL_MASK ((_AC(1,UL) << __VIRTUAL_MASK_SHIFT) - 1)
-#define KERNEL_TEXT_SIZE (40*1024*1024)
+/*
+ * This used to be called KERNEL_TEXT_SIZE, which it never was.
+ * Note that going beond 1GB requires code modifications
+ * (arch/x86_64/kernel/head.S)
+ */
+#define KERNEL_TEXT_TOP (576*1024*1024)
#define KERNEL_TEXT_START _AC(0xffffffff80000000, UL)
#define PAGE_OFFSET __PAGE_OFFSET
diff --git a/include/asm-x86/pgtable_64.h b/include/asm-x86/pgtable_64.h
index 9b0ff47..26cc42e 100644
--- a/include/asm-x86/pgtable_64.h
+++ b/include/asm-x86/pgtable_64.h
@@ -138,7 +138,7 @@ static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm, unsigned long
#define VMALLOC_START _AC(0xffffc20000000000, UL)
#define VMALLOC_END _AC(0xffffe1ffffffffff, UL)
#define VMEMMAP_START _AC(0xffffe20000000000, UL)
-#define MODULES_VADDR _AC(0xffffffff88000000, UL)
+#define MODULES_VADDR _AC(0xffffffffa8000000, UL)
#define MODULES_END _AC(0xfffffffffff00000, UL)
#define MODULES_LEN (MODULES_END - MODULES_VADDR)
diff --git a/init/initramfs.c b/init/initramfs.c
index 1db02a0..c0a36df 100644
--- a/init/initramfs.c
+++ b/init/initramfs.c
@@ -513,10 +513,14 @@ static void __init free_initrd(void)
unsigned long crashk_start = (unsigned long)__va(crashk_res.start);
unsigned long crashk_end = (unsigned long)__va(crashk_res.end);
#endif
+#ifdef CONFIG_RESERVE_LOWMEM
+ unsigned long crashk_start = 1024*1024;
+ unsigned long crashk_end = CONFIG_PHYSICAL_START - 1;
+#endif
if (do_retain_initrd)
goto skip;
-#ifdef CONFIG_KEXEC
+#if defined(CONFIG_KEXEC) || defined(CONFIG_RESERVE_LOWMEM)
/*
* If the initrd region is overlapped with crashkernel reserved region,
* free only memory that is not part of crashkernel region.
--
1.5.0
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/