Skip to contentSkip to conversation

use lmb with x86

PATCH In review Yinghai Lu 6 participants 73 messages Mar 29, 2010 – Mar 30, 2010

Subsystems (13)
lore
Share & tools
lore

Yinghai Lu

author
#0 lore raw

cover use lmb with x86

the new lmb could be used to early_res in x86.

Suggested by: David, Ben, and Thomas

First three patches should go into 2.6.34

-v6: change sequence as requested by Thomas
-v7: seperate them to more patches
-v8: add boundary checking to make sure not free partial page. -v9: use lmb_debug to control print out of reserve_lmb.
add e820 clean up, and e820 become __initdata

Quoted 1 line

size vmlinux.*

text data bss dec hex filename
20195694 4149812 12627536 36973042 23429f2 vmlinux.before_lmb_patchset
20198187 4163892 12614736 36976815 23438af vmlinux.after_lmb_patchset

Before:

[   12.124431] Freeing unused kernel memory: 2740k freed
After:
[   11.514822] Freeing unused kernel memory: 2764k freed

So We move about 24k to .init

Please check

Thanks

Yinghai

author
#1 lore raw

PATCH 1/31 x86: Make smp_locks end with page alignment

Try to fix

------------[ cut here ]------------
WARNING: at arch/x86/mm/init.c:342 free_init_pages+0x4c/0xfa()
free_init_pages: range [0x40daf000, 0x40db5c24] is not aligned
Modules linked in:
Pid: 0, comm: swapper Not tainted 2.6.34-rc2-tip-03946-g4f16b23-dirty #50
Call Trace:
 [<40232e9f>] warn_slowpath_common+0x65/0x7c
 [<4021c9f0>] ? free_init_pages+0x4c/0xfa
 [<40881434>] ? _etext+0x0/0x24
 [<40232eea>] warn_slowpath_fmt+0x24/0x27
 [<4021c9f0>] free_init_pages+0x4c/0xfa
 [<40881434>] ? _etext+0x0/0x24
 [<40d3f4bd>] alternative_instructions+0xf6/0x100
 [<40d3fe4f>] check_bugs+0xbd/0xbf
 [<40d398a7>] start_kernel+0x2d5/0x2e4
 [<40d390ce>] i386_start_kernel+0xce/0xd5
---[ end trace 4eaa2a86a8e2da22 ]---

Comments in vmlinux.lds.S already said:

|        /*
|         * smp_locks might be freed after init
|         * start/end must be page aligned
|         */
---
 arch/x86/kernel/vmlinux.lds.S |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
arch/x86/kernel/vmlinux.lds.S +1 -1
index 44879df..2cc2497 100644
@@ -291,8 +291,8 @@ SECTIONS
 	.smp_locks : AT(ADDR(.smp_locks) - LOAD_OFFSET) {
 		__smp_locks = .;
 		*(.smp_locks)
-		__smp_locks_end = .;
 		. = ALIGN(PAGE_SIZE);
+		__smp_locks_end = .;
 	}
 
 #ifdef CONFIG_X86_64
Signature · 1 line
1.6.4.2
#2 lore raw

[tip:x86/urgent] x86: Make smp_locks end with page alignment

Commit-ID:  596b711ed6b5235f8545680ef38ace00f9898c32
Gitweb:     http://git.kernel.org/tip/596b711ed6b5235f8545680ef38ace00f9898c32
Author:     Yinghai Lu <[email protected]>
AuthorDate: Sun, 28 Mar 2010 19:42:54 -0700
Committer:  Ingo Molnar <[email protected]>
CommitDate: Mon, 29 Mar 2010 18:42:30 +0200

x86: Make smp_locks end with page alignment

Fix:

 ------------[ cut here ]------------
 WARNING: at arch/x86/mm/init.c:342 free_init_pages+0x4c/0xfa()
 free_init_pages: range [0x40daf000, 0x40db5c24] is not aligned
 Modules linked in:
 Pid: 0, comm: swapper Not tainted
 2.6.34-rc2-tip-03946-g4f16b23-dirty #50 Call Trace:
  [<40232e9f>] warn_slowpath_common+0x65/0x7c
  [<4021c9f0>] ? free_init_pages+0x4c/0xfa
  [<40881434>] ? _etext+0x0/0x24
  [<40232eea>] warn_slowpath_fmt+0x24/0x27
  [<4021c9f0>] free_init_pages+0x4c/0xfa
  [<40881434>] ? _etext+0x0/0x24
  [<40d3f4bd>] alternative_instructions+0xf6/0x100
  [<40d3fe4f>] check_bugs+0xbd/0xbf
  [<40d398a7>] start_kernel+0x2d5/0x2e4
  [<40d390ce>] i386_start_kernel+0xce/0xd5
 ---[ end trace 4eaa2a86a8e2da22 ]---

Comments in vmlinux.lds.S already said:

 |        /*
 |         * smp_locks might be freed after init
 |         * start/end must be page aligned
 |         */
Cc David Miller <[email protected]>
Cc Benjamin Herrenschmidt <[email protected]>
Cc Linus Torvalds <[email protected]>

LKML-Reference: <[email protected]>

---
 arch/x86/kernel/vmlinux.lds.S |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
arch/x86/kernel/vmlinux.lds.S +1 -1
index 44879df..2cc2497 100644
@@ -291,8 +291,8 @@ SECTIONS
 	.smp_locks : AT(ADDR(.smp_locks) - LOAD_OFFSET) {
 		__smp_locks = .;
 		*(.smp_locks)
-		__smp_locks_end = .;
 		. = ALIGN(PAGE_SIZE);
+		__smp_locks_end = .;
 	}
 
 #ifdef CONFIG_X86_64
author
#3 lore raw

PATCH 2/31 x86: Make sure free_init_pages() free pages in boundary

In reply to Yinghai Lu

When CONFIG_NO_BOOTMEM, it could use memory more effient, or more compact.

Example is:
Allocated new RAMDISK: 00ec2000 - 0248ce57
Move RAMDISK from 000000002ea04000 - 000000002ffcee56 to 00ec2000 - 0248ce56

The new RAMDISK's end is not page aligned.
Last page could use shared with other user.

When free_init_pages are called for initrd or .init, the page could be freed could have chance to corrupt other data.

code segment in free_init_pages()

|        for (; addr < end; addr += PAGE_SIZE) {
|                ClearPageReserved(virt_to_page(addr));
|                init_page_count(virt_to_page(addr));
|                memset((void *)(addr & ~(PAGE_SIZE-1)),
|                        POISON_FREE_INITMEM, PAGE_SIZE);
|                free_page(addr);
|                totalram_pages++;
|        }

last half page could be used as one whole free page.

Try to make the boundaries to be page aligned.

-v2: make the original initramdisk to be aligned, according to Johannes.
otherwise we have chance to lose one page.
we still need to keep initrd_end not aligned, otherwise it could confuse decompresser.
-v3: change to WARN_ON instead according to Johannes.
-v4: use PAGE_ALIGN according to Johannes.
We may fix that MARCO name later to PAGE_ALIGN_UP, and PAGE_ALIGN_DOWN Add comments about assuming ramdisk start is aligned
in relocate_initrd(), change to re get ramdisk_image instead of save it
to make diff smaller.
Add WARN for wrong range according Johannes.
-v6: remove one WARN
We need align begin in free_init_pages()
not copy more than ramdisk_size according to Johannes

---
 arch/x86/kernel/head32.c |    3 ++-
 arch/x86/kernel/head64.c |    3 ++-
 arch/x86/kernel/setup.c  |   10 ++++++----
 arch/x86/mm/init.c       |   32 ++++++++++++++++++++++++++------
 4 files changed, 36 insertions(+), 12 deletions(-)
arch/x86/kernel/head32.c +2 -1
index adedeef..9a97504 100644
@@ -44,9 +44,10 @@ void __init i386_start_kernel(void)
 #ifdef CONFIG_BLK_DEV_INITRD
 	/* Reserve INITRD */
 	if (boot_params.hdr.type_of_loader && boot_params.hdr.ramdisk_image) {
+		/* Assume only end is not page aligned */
 		u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 		u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
-		u64 ramdisk_end   = ramdisk_image + ramdisk_size;
+		u64 ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
 		reserve_early(ramdisk_image, ramdisk_end, "RAMDISK");
 	}
 #endif
arch/x86/kernel/head64.c +2 -1
index b5a9896..7147143 100644
@@ -103,9 +103,10 @@ void __init x86_64_start_reservations(char *real_mode_data)
 #ifdef CONFIG_BLK_DEV_INITRD
 	/* Reserve INITRD */
 	if (boot_params.hdr.type_of_loader && boot_params.hdr.ramdisk_image) {
+		/* Assume only end is not page aligned */
 		unsigned long ramdisk_image = boot_params.hdr.ramdisk_image;
 		unsigned long ramdisk_size  = boot_params.hdr.ramdisk_size;
-		unsigned long ramdisk_end   = ramdisk_image + ramdisk_size;
+		unsigned long ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
 		reserve_early(ramdisk_image, ramdisk_end, "RAMDISK");
 	}
 #endif
arch/x86/kernel/setup.c +6 -4
index 5d7ba1a..d76e185 100644
@@ -314,16 +314,17 @@ static void __init reserve_brk(void)
 #define MAX_MAP_CHUNK	(NR_FIX_BTMAPS << PAGE_SHIFT)
 static void __init relocate_initrd(void)
 {
-
+	/* Assume only end is not page aligned */
 	u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 	u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
+	u64 area_size     = PAGE_ALIGN(ramdisk_size);
 	u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
 	u64 ramdisk_here;
 	unsigned long slop, clen, mapaddr;
 	char *p, *q;
 
 	/* We need to move the initrd down into lowmem */
-	ramdisk_here = find_e820_area(0, end_of_lowmem, ramdisk_size,
+	ramdisk_here = find_e820_area(0, end_of_lowmem, area_size,
 					 PAGE_SIZE);
 
 	if (ramdisk_here == -1ULL)
@@ -332,7 +333,7 @@ static void __init relocate_initrd(void)
 
 	/* Note: this includes all the lowmem currently occupied by
 	   the initrd, we rely on that fact to keep the data intact. */
-	reserve_early(ramdisk_here, ramdisk_here + ramdisk_size,
+	reserve_early(ramdisk_here, ramdisk_here + area_size,
 			 "NEW RAMDISK");
 	initrd_start = ramdisk_here + PAGE_OFFSET;
 	initrd_end   = initrd_start + ramdisk_size;
@@ -376,9 +377,10 @@ static void __init relocate_initrd(void)
 
 static void __init reserve_initrd(void)
 {
+	/* Assume only end is not page aligned */
 	u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 	u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
-	u64 ramdisk_end   = ramdisk_image + ramdisk_size;
+	u64 ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
 	u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
 
 	if (!boot_params.hdr.type_of_loader ||
arch/x86/mm/init.c +26 -6
index e71c5cb..452ee5b 100644
@@ -331,11 +331,23 @@ int devmem_is_allowed(unsigned long pagenr)
 
 void free_init_pages(char *what, unsigned long begin, unsigned long end)
 {
-	unsigned long addr = begin;
+	unsigned long addr;
+	unsigned long begin_aligned, end_aligned;
 
-	if (addr >= end)
+	/* Make sure boundaries are page aligned */
+	begin_aligned = PAGE_ALIGN(begin);
+	end_aligned   = end & PAGE_MASK;
+
+	if (WARN_ON(begin_aligned != begin || end_aligned != end)) {
+		begin = begin_aligned;
+		end   = end_aligned;
+	}
+
+	if (begin >= end)
 		return;
 
+	addr = begin;
+
 	/*
 	 * If debugging page accesses then do not free this memory but
 	 * mark them not present - any buggy init-section access will
@@ -343,7 +355,7 @@ void free_init_pages(char *what, unsigned long begin, unsigned long end)
 	 */
 #ifdef CONFIG_DEBUG_PAGEALLOC
 	printk(KERN_INFO "debug: unmapping init memory %08lx..%08lx\n",
-		begin, PAGE_ALIGN(end));
+		begin, end);
 	set_memory_np(begin, (end - begin) >> PAGE_SHIFT);
 #else
 	/*
@@ -358,8 +370,7 @@ void free_init_pages(char *what, unsigned long begin, unsigned long end)
 	for (; addr < end; addr += PAGE_SIZE) {
 		ClearPageReserved(virt_to_page(addr));
 		init_page_count(virt_to_page(addr));
-		memset((void *)(addr & ~(PAGE_SIZE-1)),
-			POISON_FREE_INITMEM, PAGE_SIZE);
+		memset((void *)addr, POISON_FREE_INITMEM, PAGE_SIZE);
 		free_page(addr);
 		totalram_pages++;
 	}
@@ -376,6 +387,15 @@ void free_initmem(void)
 #ifdef CONFIG_BLK_DEV_INITRD
 void free_initrd_mem(unsigned long start, unsigned long end)
 {
-	free_init_pages("initrd memory", start, end);
+	/*
+	 * end could be not aligned, and We can not align that,
+	 * decompresser could be confused by aligned initrd_end
+	 * We already reserve the end partial page before in
+	 *   - i386_start_kernel()
+	 *   - x86_64_start_kernel()
+	 *   - relocate_initrd()
+	 * So here We can do PAGE_ALIGN() safely to get partial page to be freed
+	 */
+	free_init_pages("initrd memory", start, PAGE_ALIGN(end));
 }
 #endif
Signature · 1 line
1.6.4.2
#4 lore raw

Re: [PATCH 02/31] x86: Make sure free_init_pages() free pages in boundary

Note, i applied the fix below, without it the 32-bit defconfig build would fail with:

arch/x86/kernel/head32.c: In function 'i386_start_kernel':
arch/x86/kernel/head32.c:50: error: implicit declaration of function 'PAGE_ALIGN'

Ingo

arch/x86/kernel/head32.c +1 -0
Index: linux/arch/x86/kernel/head32.c
===================================================================
@@ -7,6 +7,7 @@
 
 #include <linux/init.h>
 #include <linux/start_kernel.h>
+#include <linux/mm.h>
 
 #include <asm/setup.h>
 #include <asm/sections.h>
author
#5 lore raw

Re: [PATCH 02/31] x86: Make sure free_init_pages() free pages in boundary

Quoted 7 lines · Ingo Molnar

Note, i applied the fix below, without it the 32-bit defconfig build would fail with:

arch/x86/kernel/head32.c: In function 'i386_start_kernel':
arch/x86/kernel/head32.c:50: error: implicit declaration of function 'PAGE_ALIGN'

thanks.

after lmb work, will check all those
round_up
round_down
roundup
PAGE_ALIGN
ALIGN
...

We promised to Andrew, We need to clean them up.

Yinghai

#6 lore raw

[tip:x86/urgent] x86: Make sure free_init_pages() frees pages on page boundary

In reply to Yinghai Lu
Commit-ID:  c967da6a0ba837f762042e931d4afcf72045547c
Gitweb:     http://git.kernel.org/tip/c967da6a0ba837f762042e931d4afcf72045547c
Author:     Yinghai Lu <[email protected]>
AuthorDate: Sun, 28 Mar 2010 19:42:55 -0700
Committer:  Ingo Molnar <[email protected]>
CommitDate: Mon, 29 Mar 2010 18:55:33 +0200

x86: Make sure free_init_pages() frees pages on page boundary

When CONFIG_NO_BOOTMEM=y, it could use memory more effiently, or in a more compact fashion.

Example:

Allocated new RAMDISK: 00ec2000 - 0248ce57
Move RAMDISK from 000000002ea04000 - 000000002ffcee56 to 00ec2000 - 0248ce56

The new RAMDISK's end is not page aligned.
Last page could be shared with other users.

When free_init_pages are called for initrd or .init, the page could be freed and we could corrupt other data.

code segment in free_init_pages():

 |        for (; addr < end; addr += PAGE_SIZE) {
 |                ClearPageReserved(virt_to_page(addr));
 |                init_page_count(virt_to_page(addr));
 |                memset((void *)(addr & ~(PAGE_SIZE-1)),
 |                        POISON_FREE_INITMEM, PAGE_SIZE);
 |                free_page(addr);
 |                totalram_pages++;
 |        }

last half page could be used as one whole free page.

So page align the boundaries.

-v2: make the original initramdisk to be aligned, according to
Johannes, otherwise we have the chance to lose one page.
we still need to keep initrd_end not aligned, otherwise it could confuse decompressor.
-v3: change to WARN_ON instead, suggested by Johannes.
-v4: use PAGE_ALIGN, suggested by Johannes.
We may fix that macro name later to PAGE_ALIGN_UP, and PAGE_ALIGN_DOWN Add comments about assuming ramdisk start is aligned
in relocate_initrd(), change to re get ramdisk_image instead of save it to make diff smaller. Add warning for wrong range, suggested by Johannes.
-v6: remove one WARN()
We need to align beginning in free_init_pages()
do not copy more than ramdisk_size, noticed by Johannes

LKML-Reference: <[email protected]>

---
 arch/x86/kernel/head32.c |    4 +++-
 arch/x86/kernel/head64.c |    3 ++-
 arch/x86/kernel/setup.c  |   10 ++++++----
 arch/x86/mm/init.c       |   32 ++++++++++++++++++++++++++------
 4 files changed, 37 insertions(+), 12 deletions(-)
arch/x86/kernel/head32.c +3 -1
index adedeef..b2e2460 100644
@@ -7,6 +7,7 @@
 
 #include <linux/init.h>
 #include <linux/start_kernel.h>
+#include <linux/mm.h>
 
 #include <asm/setup.h>
 #include <asm/sections.h>
@@ -44,9 +45,10 @@ void __init i386_start_kernel(void)
 #ifdef CONFIG_BLK_DEV_INITRD
 	/* Reserve INITRD */
 	if (boot_params.hdr.type_of_loader && boot_params.hdr.ramdisk_image) {
+		/* Assume only end is not page aligned */
 		u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 		u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
-		u64 ramdisk_end   = ramdisk_image + ramdisk_size;
+		u64 ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
 		reserve_early(ramdisk_image, ramdisk_end, "RAMDISK");
 	}
 #endif
arch/x86/kernel/head64.c +2 -1
index b5a9896..7147143 100644
@@ -103,9 +103,10 @@ void __init x86_64_start_reservations(char *real_mode_data)
 #ifdef CONFIG_BLK_DEV_INITRD
 	/* Reserve INITRD */
 	if (boot_params.hdr.type_of_loader && boot_params.hdr.ramdisk_image) {
+		/* Assume only end is not page aligned */
 		unsigned long ramdisk_image = boot_params.hdr.ramdisk_image;
 		unsigned long ramdisk_size  = boot_params.hdr.ramdisk_size;
-		unsigned long ramdisk_end   = ramdisk_image + ramdisk_size;
+		unsigned long ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
 		reserve_early(ramdisk_image, ramdisk_end, "RAMDISK");
 	}
 #endif
arch/x86/kernel/setup.c +6 -4
index 5d7ba1a..d76e185 100644
@@ -314,16 +314,17 @@ static void __init reserve_brk(void)
 #define MAX_MAP_CHUNK	(NR_FIX_BTMAPS << PAGE_SHIFT)
 static void __init relocate_initrd(void)
 {
-
+	/* Assume only end is not page aligned */
 	u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 	u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
+	u64 area_size     = PAGE_ALIGN(ramdisk_size);
 	u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
 	u64 ramdisk_here;
 	unsigned long slop, clen, mapaddr;
 	char *p, *q;
 
 	/* We need to move the initrd down into lowmem */
-	ramdisk_here = find_e820_area(0, end_of_lowmem, ramdisk_size,
+	ramdisk_here = find_e820_area(0, end_of_lowmem, area_size,
 					 PAGE_SIZE);
 
 	if (ramdisk_here == -1ULL)
@@ -332,7 +333,7 @@ static void __init relocate_initrd(void)
 
 	/* Note: this includes all the lowmem currently occupied by
 	   the initrd, we rely on that fact to keep the data intact. */
-	reserve_early(ramdisk_here, ramdisk_here + ramdisk_size,
+	reserve_early(ramdisk_here, ramdisk_here + area_size,
 			 "NEW RAMDISK");
 	initrd_start = ramdisk_here + PAGE_OFFSET;
 	initrd_end   = initrd_start + ramdisk_size;
@@ -376,9 +377,10 @@ static void __init relocate_initrd(void)
 
 static void __init reserve_initrd(void)
 {
+	/* Assume only end is not page aligned */
 	u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 	u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
-	u64 ramdisk_end   = ramdisk_image + ramdisk_size;
+	u64 ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
 	u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
 
 	if (!boot_params.hdr.type_of_loader ||
arch/x86/mm/init.c +26 -6
index e71c5cb..452ee5b 100644
@@ -331,11 +331,23 @@ int devmem_is_allowed(unsigned long pagenr)
 
 void free_init_pages(char *what, unsigned long begin, unsigned long end)
 {
-	unsigned long addr = begin;
+	unsigned long addr;
+	unsigned long begin_aligned, end_aligned;
 
-	if (addr >= end)
+	/* Make sure boundaries are page aligned */
+	begin_aligned = PAGE_ALIGN(begin);
+	end_aligned   = end & PAGE_MASK;
+
+	if (WARN_ON(begin_aligned != begin || end_aligned != end)) {
+		begin = begin_aligned;
+		end   = end_aligned;
+	}
+
+	if (begin >= end)
 		return;
 
+	addr = begin;
+
 	/*
 	 * If debugging page accesses then do not free this memory but
 	 * mark them not present - any buggy init-section access will
@@ -343,7 +355,7 @@ void free_init_pages(char *what, unsigned long begin, unsigned long end)
 	 */
 #ifdef CONFIG_DEBUG_PAGEALLOC
 	printk(KERN_INFO "debug: unmapping init memory %08lx..%08lx\n",
-		begin, PAGE_ALIGN(end));
+		begin, end);
 	set_memory_np(begin, (end - begin) >> PAGE_SHIFT);
 #else
 	/*
@@ -358,8 +370,7 @@ void free_init_pages(char *what, unsigned long begin, unsigned long end)
 	for (; addr < end; addr += PAGE_SIZE) {
 		ClearPageReserved(virt_to_page(addr));
 		init_page_count(virt_to_page(addr));
-		memset((void *)(addr & ~(PAGE_SIZE-1)),
-			POISON_FREE_INITMEM, PAGE_SIZE);
+		memset((void *)addr, POISON_FREE_INITMEM, PAGE_SIZE);
 		free_page(addr);
 		totalram_pages++;
 	}
@@ -376,6 +387,15 @@ void free_initmem(void)
 #ifdef CONFIG_BLK_DEV_INITRD
 void free_initrd_mem(unsigned long start, unsigned long end)
 {
-	free_init_pages("initrd memory", start, end);
+	/*
+	 * end could be not aligned, and We can not align that,
+	 * decompresser could be confused by aligned initrd_end
+	 * We already reserve the end partial page before in
+	 *   - i386_start_kernel()
+	 *   - x86_64_start_kernel()
+	 *   - relocate_initrd()
+	 * So here We can do PAGE_ALIGN() safely to get partial page to be freed
+	 */
+	free_init_pages("initrd memory", start, PAGE_ALIGN(end));
 }
 #endif
author
#7 lore raw

PATCH 3/31 x86: Do not free zero sized per cpu areas

In reply to Yinghai Lu

From: Ian Campbell <[email protected]>

This avoids an infinite loop in free_early_partial().

Add a warning to free_early_partial to catch future problems.

-v5: put back start > end back into WARN_ONCE() -v6: use one line for warning according to linus -v7: more test by
-v8: remove the function name according to Johannes
WARN_ONCE will print that function name.

---
 kernel/early_res.c |    6 ++++++
 1 files changed, 6 insertions(+), 0 deletions(-)
kernel/early_res.c +6 -0
index 3cb2c66..31aa933 100644
@@ -333,6 +333,12 @@ void __init free_early_partial(u64 start, u64 end)
 	struct early_res *r;
 	int i;
 
+	if (start == end)
+		return;
+
+	if (WARN_ONCE(start > end, "  wrong range [%#llx, %#llx]\n", start, end))
+		return;
+
 try_next:
 	i = find_overlapped_early(start, end);
 	if (i >= max_early_res)
Signature · 1 line
1.6.4.2
#8 lore raw

[tip:x86/urgent] x86: Do not free zero sized per cpu areas

Commit-ID:  eed63519e3e74d515d2007ecd895338d0ba2a85c
Gitweb:     http://git.kernel.org/tip/eed63519e3e74d515d2007ecd895338d0ba2a85c
Author:     Ian Campbell <[email protected]>
AuthorDate: Sun, 28 Mar 2010 19:42:56 -0700
Committer:  Ingo Molnar <[email protected]>
CommitDate: Mon, 29 Mar 2010 18:55:40 +0200

x86: Do not free zero sized per cpu areas

This avoids an infinite loop in free_early_partial().

Add a warning to free_early_partial() to catch future problems.

-v5: put back start > end back into WARN_ONCE() -v6: use one line for warning, suggested by Linus -v7: more tests
-v8: remove the function name as suggested by Johannes
WARN_ONCE() will print out that function name.

LKML-Reference: <[email protected]>

---
 kernel/early_res.c |    6 ++++++
 1 files changed, 6 insertions(+), 0 deletions(-)
kernel/early_res.c +6 -0
index 3cb2c66..31aa933 100644
@@ -333,6 +333,12 @@ void __init free_early_partial(u64 start, u64 end)
 	struct early_res *r;
 	int i;
 
+	if (start == end)
+		return;
+
+	if (WARN_ONCE(start > end, "  wrong range [%#llx, %#llx]\n", start, end))
+		return;
+
 try_next:
 	i = find_overlapped_early(start, end);
 	if (i >= max_early_res)
author
#9 lore raw

PATCH 4/31 lmb: Move lmb.c to mm/

In reply to Yinghai Lu

that is memory related, so move to mm/ according to Ingo

---
 lib/Makefile      |    2 --
 mm/Makefile       |    2 ++
 {lib => mm}/lmb.c |    0
 3 files changed, 2 insertions(+), 2 deletions(-)
 rename {lib => mm}/lmb.c (100%)
lib/Makefile +0 -2
index 2e152ae..a463a4d 100644
@@ -85,8 +85,6 @@ obj-$(CONFIG_FAULT_INJECTION) += fault-inject.o
 
 lib-$(CONFIG_GENERIC_BUG) += bug.o
 
-obj-$(CONFIG_HAVE_LMB) += lmb.o
-
 obj-$(CONFIG_HAVE_ARCH_TRACEHOOK) += syscall.o
 
 obj-$(CONFIG_DYNAMIC_DEBUG) += dynamic_debug.o
mm/Makefile +2 -0
index 7a68d2a..df22fd1 100644
@@ -15,6 +15,8 @@ obj-y			:= bootmem.o filemap.o mempool.o oom_kill.o fadvise.o \
 			   $(mmu-y)
 obj-y += init-mm.o
 
+obj-$(CONFIG_HAVE_LMB) += lmb.o
+
 obj-$(CONFIG_BOUNCE)	+= bounce.o
 obj-$(CONFIG_SWAP)	+= page_io.o swap_state.o swapfile.o thrash.o
 obj-$(CONFIG_HAS_DMA)	+= dmapool.o
mm/lmb.c renamed from lib/lmb.c +0 -0
similarity index 100%
rename from lib/lmb.c
rename to mm/lmb.c
Signature · 1 line
1.6.4.2
author
#10 lore raw

PATCH 5/31 lmb: Seperate region array from lmb_region struct

In reply to Yinghai Lu

lmb_init() will connect them back.

Add nr_regions in struct lmb_region to track region array size.

---
 include/linux/lmb.h |    3 ++-
 mm/lmb.c            |    9 ++++++++-
 2 files changed, 10 insertions(+), 2 deletions(-)
include/linux/lmb.h +2 -1
index f3d1433..e14ea8d 100644
@@ -26,7 +26,8 @@ struct lmb_property {
 struct lmb_region {
 	unsigned long cnt;
 	u64 size;
-	struct lmb_property region[MAX_LMB_REGIONS+1];
+	struct lmb_property *region;
+	unsigned long nr_regions;
 };
 
 struct lmb {
mm/lmb.c +8 -1
index b1fc526..65b62dc 100644
@@ -18,6 +18,8 @@
 #define LMB_ALLOC_ANYWHERE	0
 
 struct lmb lmb;
+static struct lmb_property lmb_memory_region[MAX_LMB_REGIONS + 1];
+static struct lmb_property lmb_reserved_region[MAX_LMB_REGIONS + 1];
 
 static int lmb_debug;
 
@@ -106,6 +108,11 @@ static void lmb_coalesce_regions(struct lmb_region *rgn,
 
 void __init lmb_init(void)
 {
+	lmb.memory.region   = lmb_memory_region;
+	lmb.reserved.region = lmb_reserved_region;
+	lmb.memory.nr_regions   = ARRAY_SIZE(lmb_memory_region);
+	lmb.reserved.nr_regions = ARRAY_SIZE(lmb_reserved_region);
+
 	/* Create a dummy zero size LMB which will get coalesced away later.
 	 * This simplifies the lmb_add() code below...
 	 */
@@ -169,7 +176,7 @@ static long lmb_add_region(struct lmb_region *rgn, u64 base, u64 size)
 
 	if (coalesced)
 		return coalesced;
-	if (rgn->cnt >= MAX_LMB_REGIONS)
+	if (rgn->cnt > rgn->nr_regions)
 		return -1;
 
 	/* Couldn't coalesce the LMB, so add it to the sorted table. */
Signature · 1 line
1.6.4.2
author
#11 lore raw

PATCH 6/31 lmb: Add find_lmb_area()

In reply to Yinghai Lu

It will try find area according with size/align in specified range (start, end).

Need use it find correct buffer for new lmb.reserved.region.

also make it more easy for x86 to use lmb.
x86 early_res is using find/reserve pattern instead of alloc.

find_lmb_area() will hohor goal

When we need temperary buff for range array etc for range work, if We are using lmb_alloc(), We will need to add some post fix code for buffer that is used by range array, because it is in the lmb.reserved already.

---
 include/linux/lmb.h |    4 ++
 mm/lmb.c            |   81 +++++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 85 insertions(+), 0 deletions(-)
include/linux/lmb.h +4 -0
index e14ea8d..05234bd 100644
@@ -83,6 +83,10 @@ lmb_end_pfn(struct lmb_region *type, unsigned long region_nr)
 	       lmb_size_pages(type, region_nr);
 }
 
+u64 __find_lmb_area(u64 ei_start, u64 ei_last, u64 start, u64 end,
+			 u64 size, u64 align);
+u64 find_lmb_area(u64 start, u64 end, u64 size, u64 align);
+
 #include <asm/lmb.h>
 
 #endif /* __KERNEL__ */
mm/lmb.c +81 -0

93 lines not on this page · raw message

Signature · 1 line
1.6.4.2
author
#12 lore raw

PATCH 7/31 lmb: Add reserve_lmb/free_lmb

In reply to Yinghai Lu

They will check if the region array is big enough.

__check_and_double_region_array will try to double the region if that array spare slots if not big enough.
find_lmb_area() is used to find good postion for new region array. Old array will be copied to new array.

Arch code should provide to get_max_mapped, so the new array have accessiable address

---
 include/linux/lmb.h |    4 ++
 mm/lmb.c            |   89 +++++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 93 insertions(+), 0 deletions(-)
include/linux/lmb.h +4 -0
index 05234bd..95ae3f4 100644
@@ -83,9 +83,13 @@ lmb_end_pfn(struct lmb_region *type, unsigned long region_nr)
 	       lmb_size_pages(type, region_nr);
 }
 
+void reserve_lmb(u64 start, u64 end, char *name);
+void free_lmb(u64 start, u64 end);
+void add_lmb_memory(u64 start, u64 end);
 u64 __find_lmb_area(u64 ei_start, u64 ei_last, u64 start, u64 end,
 			 u64 size, u64 align);
 u64 find_lmb_area(u64 start, u64 end, u64 size, u64 align);
+u64 get_max_mapped(void);
 
 #include <asm/lmb.h>
 
mm/lmb.c +89 -0

95 lines not on this page · raw message

Signature · 1 line
1.6.4.2
#13 lore raw

Re: [PATCH 07/31] lmb: Add reserve_lmb/free_lmb

Quoted 9 lines · Yinghai Lu

They will check if the region array is big enough.

__check_and_double_region_array will try to double the region if that array spare slots if not big enough.
find_lmb_area() is used to find good postion for new region array. Old array will be copied to new array.

Arch code should provide to get_max_mapped, so the new array have accessiable address

..

Quoted 6 lines
mm/lmb.c +0 -0
index d5d5dc4..9798458 100644
@@ -551,6 +551,95 @@ int lmb_find(struct lmb_property *res)
 	return -1;

mm/lmb.c int lmb_find(struct lmb_property *res) · line 561

Quoted 10 lines
mm/lmb.c +8 -0
 }
 
+u64 __weak __init get_max_mapped(void)
+{
+	u64 end = max_low_pfn;
+
+	end <<= PAGE_SHIFT;
+
+	return end;
+}

^ This is (sort of) what lmb.rmo_size represents. So maybe instead of adding this function, we could just say that the arch code needs to set rmo_size up with an appropriate value, and then use that below. Though maybe that's conflating things.

...

Quoted 21 lines
mm/lmb.c +21 -0
+
+void __init add_lmb_memory(u64 start, u64 end)
+{
+	__check_and_double_region_array(&lmb.memory, &lmb_memory_region[0], start, end);
+	lmb_add(start, end - start);
+}
+
+void __init reserve_lmb(u64 start, u64 end, char *name)
+{
+	if (start == end)
+		return;
+
+	if (WARN_ONCE(start > end, "reserve_lmb: wrong range [%#llx, %#llx]\n", start, end))
+		return;
+
+	__check_and_double_region_array(&lmb.reserved, &lmb_reserved_region[0], start, end);
+	lmb_reserve(start, end - start);
+}
+
+void __init free_lmb(u64 start, u64 end)
+{

mm/lmb.c int lmb_find(struct lmb_property *res) · line 641

Quoted 10 lines
mm/lmb.c +10 -0
+	if (start == end)
+		return;
+
+	if (WARN_ONCE(start > end, "free_lmb: wrong range [%#llx, %#llx]\n", start, end))
+		return;
+
+	/* keep punching hole, could run out of slots too */
+	__check_and_double_region_array(&lmb.reserved, &lmb_reserved_region[0], start, end);
+	lmb_free(start, end - start);
+}

Doesn't this mean that if I call lmb_alloc() or lmb_free() too many times then I'll potentially run out of space? So doesn't that essentially break the existing API?

It seems to me that rather than adding these "special" routines that check for enough space on the way in, instead you should be checking in lmb_add_region() - which is where AFAICS all allocs/frees/reserves eventually end up if they need to insert a new region.

cheers

author
#14 lore raw

Re: [PATCH 07/31] lmb: Add reserve_lmb/free_lmb

Quoted 19 lines · Michael Ellerman
Quoted 9 lines · Yinghai Lu

They will check if the region array is big enough.

__check_and_double_region_array will try to double the region if that array spare slots if not big enough.
find_lmb_area() is used to find good postion for new region array. Old array will be copied to new array.

Arch code should provide to get_max_mapped, so the new array have accessiable address

..

Quoted 8 lines
mm/lmb.c +0 -0
index d5d5dc4..9798458 100644
@@ -551,6 +551,95 @@ int lmb_find(struct lmb_property *res)
 	return -1;
 }
 
Quoted 13 lines

mm/lmb.c int lmb_find(struct lmb_property *res) · line 561

Quoted 8 lines
mm/lmb.c +8 -0
+u64 __weak __init get_max_mapped(void)
+{
+	u64 end = max_low_pfn;
+
+	end <<= PAGE_SHIFT;
+
+	return end;
+}

^ This is (sort of) what lmb.rmo_size represents. So maybe instead of adding this function, we could just say that the arch code needs to set rmo_size up with an appropriate value, and then use that below. Though maybe that's conflating things.

ok

will have another patch following this patchset. to use rmo_size replace get_max_mapped()

long __init_lmb lmb_add(u64 base, u64 size)

{
        struct lmb_region *_rgn = &lmb.memory;

        /* On pSeries LPAR systems, the first LMB is our RMO region. */
        if (base == 0)
                lmb.rmo_size = size;

        return lmb_add_region(_rgn, base, size);

}

looks scary.
maybe later powerpc could used lmb_find and set_lmb_rmo_size in their arch code.

Quoted 29 lines

...

Quoted 27 lines
mm/lmb.c +27 -0
+
+void __init add_lmb_memory(u64 start, u64 end)
+{
+	__check_and_double_region_array(&lmb.memory, &lmb_memory_region[0], start, end);
+	lmb_add(start, end - start);
+}
+
+void __init reserve_lmb(u64 start, u64 end, char *name)
+{
+	if (start == end)
+		return;
+
+	if (WARN_ONCE(start > end, "reserve_lmb: wrong range [%#llx, %#llx]\n", start, end))
+		return;
+
+	__check_and_double_region_array(&lmb.reserved, &lmb_reserved_region[0], start, end);
+	lmb_reserve(start, end - start);
+}
+
+void __init free_lmb(u64 start, u64 end)
+{
+	if (start == end)
+		return;
+
+	if (WARN_ONCE(start > end, "free_lmb: wrong range [%#llx, %#llx]\n", start, end))
+		return;
+
Quoted 8 lines

mm/lmb.c int lmb_find(struct lmb_property *res) · line 641

Quoted 4 lines
mm/lmb.c +4 -0
+	/* keep punching hole, could run out of slots too */
+	__check_and_double_region_array(&lmb.reserved, &lmb_reserved_region[0], start, end);
+	lmb_free(start, end - start);
+}

Doesn't this mean that if I call lmb_alloc() or lmb_free() too many times then I'll potentially run out of space? So doesn't that essentially break the existing API?

No, I didn't touch existing API, arches other than x86 should have little change about lmb.memory.region
lmb.reserved.region
become pointer from array.

Quoted 5 lines

It seems to me that rather than adding these "special" routines that check for enough space on the way in, instead you should be checking in lmb_add_region() - which is where AFAICS all allocs/frees/reserves eventually end up if they need to insert a new region.

later i prefer to replace lmb_alloc with find_lmb_area + reserve_lmb.

Thanks

Yinghai

#15 lore raw

Re: [PATCH 07/31] lmb: Add reserve_lmb/free_lmb

Quoted 34 lines · Yinghai Lu
Quoted 19 lines · Michael Ellerman
Quoted 9 lines · Yinghai Lu

They will check if the region array is big enough.

__check_and_double_region_array will try to double the region if that array spare slots if not big enough.
find_lmb_area() is used to find good postion for new region array. Old array will be copied to new array.

Arch code should provide to get_max_mapped, so the new array have accessiable address

..

Quoted 8 lines
mm/lmb.c +0 -0
index d5d5dc4..9798458 100644
@@ -551,6 +551,95 @@ int lmb_find(struct lmb_property *res)
 	return -1;
 }
 
Quoted 13 lines

mm/lmb.c int lmb_find(struct lmb_property *res) · line 561

Quoted 8 lines
mm/lmb.c +8 -0
+u64 __weak __init get_max_mapped(void)
+{
+	u64 end = max_low_pfn;
+
+	end <<= PAGE_SHIFT;
+
+	return end;
+}

^ This is (sort of) what lmb.rmo_size represents. So maybe instead of adding this function, we could just say that the arch code needs to set rmo_size up with an appropriate value, and then use that below. Though maybe that's conflating things.

Quoted 3 lines

ok

will have another patch following this patchset. to use rmo_size replace get_max_mapped()

No don't, Benh's idea was better. Leave rmo_size for now, we can clean that up later.

We just need a lmb.alloc_limit and a lmb_set_alloc_limit() which arch code calls when it knows what the alloc limit is (and can call multiple times during boot). Or maybe it should be called "default_alloc_limit", but that's getting a bit long winded.

Quoted 11 lines

long __init_lmb lmb_add(u64 base, u64 size)

{
        struct lmb_region *_rgn = &lmb.memory;

        /* On pSeries LPAR systems, the first LMB is our RMO region. */
        if (base == 0)
                lmb.rmo_size = size;

        return lmb_add_region(_rgn, base, size);
Quoted 4 lines

}

looks scary.
maybe later powerpc could used lmb_find and set_lmb_rmo_size in their arch code.

It's not really scary, and it gives you a hint where the code came from originally :)

We can remove that later though, with some powerpc code to detect the first memory region before we put it into lmb.

Quoted 37 lines
Quoted 28 lines

...

Quoted 27 lines
mm/lmb.c +27 -0
+
+void __init add_lmb_memory(u64 start, u64 end)
+{
+	__check_and_double_region_array(&lmb.memory, &lmb_memory_region[0], start, end);
+	lmb_add(start, end - start);
+}
+
+void __init reserve_lmb(u64 start, u64 end, char *name)
+{
+	if (start == end)
+		return;
+
+	if (WARN_ONCE(start > end, "reserve_lmb: wrong range [%#llx, %#llx]\n", start, end))
+		return;
+
+	__check_and_double_region_array(&lmb.reserved, &lmb_reserved_region[0], start, end);
+	lmb_reserve(start, end - start);
+}
+
+void __init free_lmb(u64 start, u64 end)
+{
+	if (start == end)
+		return;
+
+	if (WARN_ONCE(start > end, "free_lmb: wrong range [%#llx, %#llx]\n", start, end))
+		return;
+
Quoted 8 lines

mm/lmb.c int lmb_find(struct lmb_property *res) · line 641

Quoted 4 lines
mm/lmb.c +4 -0
+	/* keep punching hole, could run out of slots too */
+	__check_and_double_region_array(&lmb.reserved, &lmb_reserved_region[0], start, end);
+	lmb_free(start, end - start);
+}

Doesn't this mean that if I call lmb_alloc() or lmb_free() too many times then I'll potentially run out of space? So doesn't that essentially break the existing API?

Quoted 4 lines

No, I didn't touch existing API, arches other than x86 should have little change about lmb.memory.region
lmb.reserved.region
become pointer from array.

But that's my point. You shouldn't need to touch the existing API, and you shouldn't need to add a new parallel API. You should just be able to add the logic for doubling the array in the lmb core, and then everyone gets dynamically expandable lmb. I don't see any reason why we want to have two APIs.

Quoted 6 lines
Quoted 4 lines

It seems to me that rather than adding these "special" routines that check for enough space on the way in, instead you should be checking in lmb_add_region() - which is where AFAICS all allocs/frees/reserves eventually end up if they need to insert a new region.

later i prefer to replace lmb_alloc with find_lmb_area + reserve_lmb.

Why? The existing code has been working for years and is well tested?

cheers

author
#16 lore raw

Re: [PATCH 07/31] lmb: Add reserve_lmb/free_lmb

Quoted 39 lines · Michael Ellerman
Quoted 34 lines · Yinghai Lu
Quoted 19 lines · Michael Ellerman
Quoted 9 lines · Yinghai Lu

They will check if the region array is big enough.

__check_and_double_region_array will try to double the region if that array spare slots if not big enough.
find_lmb_area() is used to find good postion for new region array. Old array will be copied to new array.

Arch code should provide to get_max_mapped, so the new array have accessiable address

..

Quoted 8 lines
mm/lmb.c +0 -0
index d5d5dc4..9798458 100644
@@ -551,6 +551,95 @@ int lmb_find(struct lmb_property *res)
 	return -1;
 }
 
Quoted 13 lines

mm/lmb.c int lmb_find(struct lmb_property *res) · line 561

Quoted 8 lines
mm/lmb.c +8 -0
+u64 __weak __init get_max_mapped(void)
+{
+	u64 end = max_low_pfn;
+
+	end <<= PAGE_SHIFT;
+
+	return end;
+}

^ This is (sort of) what lmb.rmo_size represents. So maybe instead of adding this function, we could just say that the arch code needs to set rmo_size up with an appropriate value, and then use that below. Though maybe that's conflating things.

Quoted 3 lines

ok

will have another patch following this patchset. to use rmo_size replace get_max_mapped()

Quoted 7 lines

No don't, Benh's idea was better. Leave rmo_size for now, we can clean that up later.

We just need a lmb.alloc_limit and a lmb_set_alloc_limit() which arch code calls when it knows what the alloc limit is (and can call multiple times during boot). Or maybe it should be called "default_alloc_limit", but that's getting a bit long winded.

ok, I will get_max_mapped() for now, an will change to new field...later

Quoted 14 lines
Quoted 14 lines

long __init_lmb lmb_add(u64 base, u64 size)

{
        struct lmb_region *_rgn = &lmb.memory;

        /* On pSeries LPAR systems, the first LMB is our RMO region. */
        if (base == 0)
                lmb.rmo_size = size;

        return lmb_add_region(_rgn, base, size);

}

Quoted 8 lines
Quoted 2 lines

looks scary.
maybe later powerpc could used lmb_find and set_lmb_rmo_size in their arch code.

It's not really scary, and it gives you a hint where the code came from originally :)

We can remove that later though, with some powerpc code to detect the first memory region before we put it into lmb.

good.

Quoted 43 lines
Quoted 38 lines
Quoted 29 lines

...

Quoted 27 lines
mm/lmb.c +27 -0
+
+void __init add_lmb_memory(u64 start, u64 end)
+{
+	__check_and_double_region_array(&lmb.memory, &lmb_memory_region[0], start, end);
+	lmb_add(start, end - start);
+}
+
+void __init reserve_lmb(u64 start, u64 end, char *name)
+{
+	if (start == end)
+		return;
+
+	if (WARN_ONCE(start > end, "reserve_lmb: wrong range [%#llx, %#llx]\n", start, end))
+		return;
+
+	__check_and_double_region_array(&lmb.reserved, &lmb_reserved_region[0], start, end);
+	lmb_reserve(start, end - start);
+}
+
+void __init free_lmb(u64 start, u64 end)
+{
+	if (start == end)
+		return;
+
+	if (WARN_ONCE(start > end, "free_lmb: wrong range [%#llx, %#llx]\n", start, end))
+		return;
+
Quoted 8 lines

mm/lmb.c int lmb_find(struct lmb_property *res) · line 641

Quoted 4 lines
mm/lmb.c +4 -0
+	/* keep punching hole, could run out of slots too */
+	__check_and_double_region_array(&lmb.reserved, &lmb_reserved_region[0], start, end);
+	lmb_free(start, end - start);
+}

Doesn't this mean that if I call lmb_alloc() or lmb_free() too many times then I'll potentially run out of space? So doesn't that essentially break the existing API?

Quoted 4 lines

No, I didn't touch existing API, arches other than x86 should have little change about lmb.memory.region
lmb.reserved.region
become pointer from array.

Quoted 5 lines

But that's my point. You shouldn't need to touch the existing API, and you shouldn't need to add a new parallel API. You should just be able to add the logic for doubling the array in the lmb core, and then everyone gets dynamically expandable lmb. I don't see any reason why we want to have two APIs.

that will have too much change for all x86 caller.

that set of API is __init, and will be freed later.

and these two are just wrapper for old one to make x86 transition more easily.

Quoted 9 lines
Quoted 7 lines
Quoted 5 lines

It seems to me that rather than adding these "special" routines that check for enough space on the way in, instead you should be checking in lmb_add_region() - which is where AFAICS all allocs/frees/reserves eventually end up if they need to insert a new region.

later i prefer to replace lmb_alloc with find_lmb_area + reserve_lmb.

Why? The existing code has been working for years and is well tested?

We need to find_lmb_area to find good position for new reserved region array.

static void __init __reserve_lmb(u64 start, u64 end, char *name)

{
        __check_and_double_region_array(&lmb.reserved, &lmb_reserved_region[0], start, end);
        lmb_reserve(start, end - start);
}

with the execlude area in __check_and_double_region_lmb() will not put the new array overlap with area that we are going to reserve.

and find_lmb_area + reserve_lmb should be identical to lmb_alloc...

Thanks

Yinghai Lu

#17 lore raw

Re: [PATCH 07/31] lmb: Add reserve_lmb/free_lmb

Quoted 6 lines · Yinghai Lu
Quoted 4 lines

We just need a lmb.alloc_limit and a lmb_set_alloc_limit() which arch code calls when it knows what the alloc limit is (and can call multiple times during boot). Or maybe it should be called "default_alloc_limit", but that's getting a bit long winded.

ok, I will get_max_mapped() for now, an will change to new field...later

No. Do it now. get_max_mapped() sucks as an identifier.

Cheers,
Ben.

author
#18 lore raw

Re: [PATCH 07/31] lmb: Add reserve_lmb/free_lmb

Quoted 10 lines · Benjamin Herrenschmidt
Quoted 6 lines · Yinghai Lu
Quoted 4 lines

We just need a lmb.alloc_limit and a lmb_set_alloc_limit() which arch code calls when it knows what the alloc limit is (and can call multiple times during boot). Or maybe it should be called "default_alloc_limit", but that's getting a bit long winded.

ok, I will get_max_mapped() for now, an will change to new field...later

No. Do it now. get_max_mapped() sucks as an identifier.

ok, can i reuse rmo_size, or introduce one new member in struct lmb.

default_limit?

Thanks

Yinghai

#19 lore raw

Re: [PATCH 07/31] lmb: Add reserve_lmb/free_lmb

Quoted 7 lines · Yinghai Lu
Quoted 6 lines · Benjamin Herrenschmidt
Quoted 5 lines · Yinghai Lu
Quoted 4 lines

We just need a lmb.alloc_limit and a lmb_set_alloc_limit() which arch code calls when it knows what the alloc limit is (and can call multiple times during boot). Or maybe it should be called "default_alloc_limit", but that's getting a bit long winded.

Quoted 8 lines
Quoted 4 lines
Quoted 1 line

ok, I will get_max_mapped() for now, an will change to new field...later

No. Do it now. get_max_mapped() sucks as an identifier.

ok, can i reuse rmo_size, or introduce one new member in struct lmb.

default_limit?

alloc_limit or default_alloc_limit

cheers

author
#20 lore raw

Re: [PATCH 07/31] lmb: Add reserve_lmb/free_lmb

Quoted 10 lines · Michael Ellerman
Quoted 9 lines · Yinghai Lu
Quoted 8 lines · Benjamin Herrenschmidt
Quoted 6 lines · Yinghai Lu
Quoted 4 lines

We just need a lmb.alloc_limit and a lmb_set_alloc_limit() which arch code calls when it knows what the alloc limit is (and can call multiple times during boot). Or maybe it should be called "default_alloc_limit", but that's getting a bit long winded.

ok, I will get_max_mapped() for now, an will change to new field...later

Quoted 9 lines
Quoted 6 lines
Quoted 2 lines

No. Do it now. get_max_mapped() sucks as an identifier.

ok, can i reuse rmo_size, or introduce one new member in struct lmb.

default_limit?

alloc_limit or default_alloc_limit

looks that is not accurate.

if someone want to find some area, but not going to access that range, then we should let them alloc it.

how about access_limit?

Thanks

Yinghai Lu

#21 lore raw

Re: [PATCH 07/31] lmb: Add reserve_lmb/free_lmb

Quoted 5 lines · Yinghai Lu

if someone want to find some area, but not going to access that range, then we should let them alloc it.

how about access_limit?

No, no you don't get it. default_alloc_limit is fine. It's the -default- limit. It doesn't apply to lmb_alloc_base() that takes an explicit limit...

Something along those lines.

Cheers,
Ben.

author
#22 lore raw

Re: [PATCH 07/31] lmb: Add reserve_lmb/free_lmb

Quoted 11 lines · Benjamin Herrenschmidt
Quoted 5 lines · Yinghai Lu

if someone want to find some area, but not going to access that range, then we should let them alloc it.

how about access_limit?

No, no you don't get it. default_alloc_limit is fine. It's the -default- limit. It doesn't apply to lmb_alloc_base() that takes an explicit limit...

fine, will use default_limit.

Yinghai

#23 lore raw

Re: [PATCH 07/31] lmb: Add reserve_lmb/free_lmb

In reply to Yinghai Lu
Quoted 4 lines · Yinghai Lu

ok, can i reuse rmo_size, or introduce one new member in struct lmb.

default_limit?

lmb.default_limit sounds good to me.

Cheers,
Ben.

#24 lore raw

Re: [PATCH 07/31] lmb: Add reserve_lmb/free_lmb

In reply to Michael Ellerman
Quoted 6 lines · Michael Ellerman

But that's my point. You shouldn't need to touch the existing API, and you shouldn't need to add a new parallel API. You should just be able to add the logic for doubling the array in the lmb core, and then everyone gets dynamically expandable lmb. I don't see any reason why we want to have two APIs.

Ack.

Quoted 8 lines
Quoted 6 lines
Quoted 4 lines

It seems to me that rather than adding these "special" routines that check for enough space on the way in, instead you should be checking in lmb_add_region() - which is where AFAICS all allocs/frees/reserves eventually end up if they need to insert a new region.

later i prefer to replace lmb_alloc with find_lmb_area + reserve_lmb.

Why? The existing code has been working for years and is well tested?

I still don't totally understand why he needs a find_lmb_area() anyways.

It might be justified ... or not. I just want it to be better documented.

Cheers,
Ben.

author
#25 lore raw

Re: [PATCH 07/31] lmb: Add reserve_lmb/free_lmb

Quoted 9 lines · Benjamin Herrenschmidt
Quoted 6 lines · Michael Ellerman

But that's my point. You shouldn't need to touch the existing API, and you shouldn't need to add a new parallel API. You should just be able to add the logic for doubling the array in the lmb core, and then everyone gets dynamically expandable lmb. I don't see any reason why we want to have two APIs.

Ack.

ok, we can merge them later.

Quoted 8 lines
Quoted 8 lines
Quoted 7 lines
Quoted 5 lines

It seems to me that rather than adding these "special" routines that check for enough space on the way in, instead you should be checking in lmb_add_region() - which is where AFAICS all allocs/frees/reserves eventually end up if they need to insert a new region.

later i prefer to replace lmb_alloc with find_lmb_area + reserve_lmb.

Quoted 7 lines
Quoted 1 line

Why? The existing code has been working for years and is well tested?

I still don't totally understand why he needs a find_lmb_area() anyways.

It might be justified ... or not. I just want it to be better documented.

current changelog for that

------------------

Subject: [PATCH 6/31] lmb: Add lmb_find_area()

It will try find area according with size/align in specified range (start, end).

Need use it find correct buffer for new lmb.reserved.region.

also make it more easy for x86 to use lmb.
x86 early_res is using find/reserve pattern instead of alloc.

lmb_find_area() will honor goal

When we need temporary buff for range array etc for range work, if We are using lmb_alloc(), We will need to add some post fix code for buffer that is used by range array, because it is in the lmb.reserved already.

----------------

in short: It could make us to avoid use the range that we are going to reserve,
when we try to get new position new lmb.reserved.region.

Thanks

Yinghai

#26 lore raw

Re: [PATCH 07/31] lmb: Add reserve_lmb/free_lmb

Quoted 4 lines · Yinghai Lu

in short: It could make us to avoid use the range that we are going to reserve,
when we try to get new position new lmb.reserved.region.

I'm not too sure I follow you. For the resizing, I would just basically call a low level variant of alloc (__lmb_alloc ?) that explicitely doesn't honor the total-2 "reserved" entries in the array.

Ie. It should all be one single find/allocation function.

In fact, you want to split lmb_find from lmb_reserve, then just make lmb_alloc use the above, I don't want 2 implementations of the same thing (maybe call it __lmb_find to expose the fact that it's a low level function to avoid for normal use).

Cheers,
Ben.

author
#27 lore raw

Re: [PATCH 07/31] lmb: Add reserve_lmb/free_lmb

Quoted 9 lines · Benjamin Herrenschmidt
Quoted 4 lines · Yinghai Lu

in short: It could make us to avoid use the range that we are going to reserve,
when we try to get new position new lmb.reserved.region.

I'm not too sure I follow you. For the resizing, I would just basically call a low level variant of alloc (__lmb_alloc ?) that explicitely doesn't honor the total-2 "reserved" entries in the array.

1. you want to reserve rangeA
2. before that will check if region array is big enough,
3. if region is not big enough, will call lmb_alloc to get new range. lmb_alloc could return rangB that is overlapped with rangeA
4. current lmb_alloc only honor limit, and doesn't honor low limit.

another usage is: for temporary buffer, like range array for subtraction. we don't need to do free later.

Quoted 7 lines

Ie. It should all be one single find/allocation function.

In fact, you want to split lmb_find from lmb_reserve, then just make lmb_alloc use the above, I don't want 2 implementations of the same thing (maybe call it __lmb_find to expose the fact that it's a low level function to avoid for normal use).

that is some difference between them, and lmb_alloc doesn't honor low limit.

we can provide
lmb_find_area
lmb_reserve_area
lmb_free_area

and use lmb_find_area + lmb_reserve_area to get one lmb_alloc()

x86 sometime is using find_lmb_area to find big area, and use those area and later reserve actually used area.

you could use lmb_alloc, and later lmb_free not used. that is equal to lmb_find + lmb_reserve + lmb_free ...

Thanks

Yinghai Lu

#28 lore raw

Re: [PATCH 07/31] lmb: Add reserve_lmb/free_lmb

Quoted 7 lines · Yinghai Lu
Quoted 6 lines · Benjamin Herrenschmidt
Quoted 4 lines · Yinghai Lu

in short: It could make us to avoid use the range that we are going to reserve,
when we try to get new position new lmb.reserved.region.

Quoted 8 lines
Quoted 3 lines

I'm not too sure I follow you. For the resizing, I would just basically call a low level variant of alloc (__lmb_alloc ?) that explicitely doesn't honor the total-2 "reserved" entries in the array.

1. you want to reserve rangeA
2. before that will check if region array is big enough,
3. if region is not big enough, will call lmb_alloc to get new range. lmb_alloc could return rangB that is overlapped with rangeA

So instead you do it the other way.

1. you want to reserve rangeA
2. you reserve rangeA
3. if reserving rangeA consumed a slot in the array then you check if you have at least two free slots. If not you realloc. You don't need any special tricks because you have space to lmb_alloc() a new area and move everything over.

cheers

author
#29 lore raw

Re: [PATCH 07/31] lmb: Add reserve_lmb/free_lmb

Quoted 17 lines · Michael Ellerman
Quoted 7 lines · Yinghai Lu
Quoted 6 lines · Benjamin Herrenschmidt
Quoted 4 lines · Yinghai Lu

in short: It could make us to avoid use the range that we are going to reserve,
when we try to get new position new lmb.reserved.region.

Quoted 8 lines
Quoted 3 lines

I'm not too sure I follow you. For the resizing, I would just basically call a low level variant of alloc (__lmb_alloc ?) that explicitely doesn't honor the total-2 "reserved" entries in the array.

1. you want to reserve rangeA
2. before that will check if region array is big enough,
3. if region is not big enough, will call lmb_alloc to get new range. lmb_alloc could return rangB that is overlapped with rangeA

Quoted 8 lines

So instead you do it the other way.

1. you want to reserve rangeA
2. you reserve rangeA
3. if reserving rangeA consumed a slot in the array then you check if you have at least two free slots. If not you realloc. You don't need any special tricks because you have space to lmb_alloc() a new area and move everything over.

so that is check it later. should work.

one less find_lmb_area user.

Thanks

Yinghai

#30 lore raw

Re: [PATCH 07/31] lmb: Add reserve_lmb/free_lmb

In reply to Yinghai Lu
Quoted 6 lines · Yinghai Lu

1. you want to reserve rangeA
2. before that will check if region array is big enough,
3. if region is not big enough, will call lmb_alloc to get new range. lmb_alloc could return rangB that is overlapped with rangeA
4. current lmb_alloc only honor limit, and doesn't honor low limit.

I see. This is a direct consequence of you wanting to use find/reserve instead of alloc tho :-)

This is also easily fixed. Instead of doing the resize "on demand" like I originally proposed, do it at the end of reserve/alloc. If the number of free entries is down to 1 or 0, then alloc a new chunk.

Of course, all of that requires that reservations done from FW to take memory out because it must not be accessed need to be all done before the first grow of the array, and so the static array must be sized accordingly. We may want to catch these things too. We don't want to warn on multiple overlapping lmb_reserve() tho on powerpc...

Quoted 2 lines

another usage is: for temporary buffer, like range array for subtraction. we don't need to do free later.

Sorry, doesn't parse.

Quoted 10 lines
Quoted 5 lines

Ie. It should all be one single find/allocation function.

In fact, you want to split lmb_find from lmb_reserve, then just make lmb_alloc use the above, I don't want 2 implementations of the same thing (maybe call it __lmb_find to expose the fact that it's a low

level

Quoted 1 line

function to avoid for normal use).

that is some difference between them, and lmb_alloc doesn't honor low limit.

If you want a low and a high limit, then add the low limit to lmb_alloc_base(), it's easy to fix all callers, there aren't many, and make not one but two defaults for lmb_alloc(), one for the low limit and one for the high limit. Problem solved.

Quoted 6 lines

we can provide
lmb_find_area
lmb_reserve_area
lmb_free_area

and use lmb_find_area + lmb_reserve_area to get one lmb_alloc()

I still don't understand why you insist on using find + reserve instead of alloc in x86 land. Can you give me a proper explanation as to why that is needed since it seems to be causing problems, and so far and I don't see what it solves.

Quoted 2 lines

x86 sometime is using find_lmb_area to find big area, and use those area and later reserve actually used area.

That's very wrong. If you use something, alloc/reserve it. You can always free it later.

Quoted 2 lines

you could use lmb_alloc, and later lmb_free not used. that is equal to lmb_find + lmb_reserve + lmb_free ...

Sure, then just do alloc + later free.

Cheers,
Ben.

author
#31 lore raw

Re: [PATCH 07/31] lmb: Add reserve_lmb/free_lmb

Quoted 8 lines · Benjamin Herrenschmidt
Quoted 6 lines · Yinghai Lu

1. you want to reserve rangeA
2. before that will check if region array is big enough,
3. if region is not big enough, will call lmb_alloc to get new range. lmb_alloc could return rangB that is overlapped with rangeA
4. current lmb_alloc only honor limit, and doesn't honor low limit.

Quoted 6 lines

I see. This is a direct consequence of you wanting to use find/reserve instead of alloc tho :-)

This is also easily fixed. Instead of doing the resize "on demand" like I originally proposed, do it at the end of reserve/alloc. If the number of free entries is down to 1 or 0, then alloc a new chunk.

already done. last night, Michael point that out.

Quoted 11 lines

Of course, all of that requires that reservations done from FW to take memory out because it must not be accessed need to be all done before the first grow of the array, and so the static array must be sized accordingly. We may want to catch these things too. We don't want to warn on multiple overlapping lmb_reserve() tho on powerpc...

Quoted 2 lines

another usage is: for temporary buffer, like range array for subtraction. we don't need to do free later.

Sorry, doesn't parse.

never mind.

Quoted 9 lines
Quoted 9 lines
Quoted 6 lines

Ie. It should all be one single find/allocation function.

In fact, you want to split lmb_find from lmb_reserve, then just make lmb_alloc use the above, I don't want 2 implementations of the same thing (maybe call it __lmb_find to expose the fact that it's a low

level

Quoted 1 line

function to avoid for normal use).

Quoted 7 lines
Quoted 2 lines

that is some difference between them, and lmb_alloc doesn't honor low limit.

If you want a low and a high limit, then add the low limit to lmb_alloc_base(), it's easy to fix all callers, there aren't many, and make not one but two defaults for lmb_alloc(), one for the low limit and one for the high limit. Problem solved.

make 32 bit x86 can work from high to low now.

Quoted 16 lines
Quoted 7 lines

we can provide
lmb_find_area
lmb_reserve_area
lmb_free_area

and use lmb_find_area + lmb_reserve_area to get one lmb_alloc()

I still don't understand why you insist on using find + reserve instead of alloc in x86 land. Can you give me a proper explanation as to why that is needed since it seems to be causing problems, and so far and I don't see what it solves.

Quoted 2 lines

x86 sometime is using find_lmb_area to find big area, and use those area and later reserve actually used area.

Quoted 7 lines

That's very wrong. If you use something, alloc/reserve it. You can always free it later.

Quoted 2 lines

you could use lmb_alloc, and later lmb_free not used. that is equal to lmb_find + lmb_reserve + lmb_free ...

Sure, then just do alloc + later free.

ok, I will replace that later after it get stable. but at first will expose the find_lmb_area.

will send out -v11, please check that version.

Thanks

Yinghai

#32 lore raw

Re: [PATCH 07/31] lmb: Add reserve_lmb/free_lmb

In reply to Michael Ellerman
Quoted 4 lines · Michael Ellerman

^ This is (sort of) what lmb.rmo_size represents. So maybe instead of adding this function, we could just say that the arch code needs to set rmo_size up with an appropriate value, and then use that below. Though maybe that's conflating things.

Well... not quite.

The RMO (which really is the RMA, historical misnaming) is the region of memory we can access very early during boot (in real mode on ppc64 but I plan to use it to represent the boot-time fixed mapping on ppc32 at some stage). It's not strictly equivalent to max lowmem. However, on ppc64, it happens to be the size of the first added LMB entry/

In any case, the LMBs should really not care. They allocate where you tell them to.

IE. That stuff is arch specific enough that I suspect we should just move it out, while the concept of max_lowmem is common enough (at least for 32-bit archs) that I'm happy to have some provisions for it inside the LMB core.

Maybe what we need is an arch call to set the allocation "limit". It could be set in stages during boot. To the initial mapped memory (bolted TLB) on ppc32 very early, and then pushed up to max_low_pfn as soon as the full MMU setup is done for example. IE. All we need is an lmb_set_alloc_limit() called by the arch in the right spots, that defines the default allocation limit for lmb_alloc() though of course lmb_alloc_base() can be used by callers to enforce explicit limits.

Cheers,
Ben.

author
#33 lore raw

PATCH 8/31 lmb: Add find_lmb_area_size()

In reply to Yinghai Lu

The same as find_lmb_area(), but size is returned according free range.

Will be used to find free ranges for early_memtest and memory corruption check

---
 include/linux/lmb.h |    1 +
 mm/lmb.c            |   80 +++++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 81 insertions(+), 0 deletions(-)
include/linux/lmb.h +1 -0
index 95ae3f4..7301072 100644
@@ -89,6 +89,7 @@ void add_lmb_memory(u64 start, u64 end);
 u64 __find_lmb_area(u64 ei_start, u64 ei_last, u64 start, u64 end,
 			 u64 size, u64 align);
 u64 find_lmb_area(u64 start, u64 end, u64 size, u64 align);
+u64 find_lmb_area_size(u64 start, u64 *sizep, u64 align);
 u64 get_max_mapped(void);
 
 #include <asm/lmb.h>
mm/lmb.c +80 -0

95 lines not on this page · raw message

Signature · 1 line
1.6.4.2
author
#34 lore raw

PATCH 9/31 bootmem, x86: Add weak version of reserve_bootmem_generic

In reply to Yinghai Lu

It will be used lmb_to_bootmem converting

It is an wrapper for reserve_bootmem, and x86 64bit is using special one.

Also clean up that version for x86_64. We don't need to take care of numa path for that, bootmem can handle it how

---
 arch/x86/mm/init_32.c |    6 ------
 arch/x86/mm/init_64.c |   20 ++------------------
 mm/bootmem.c          |    6 ++++++
 3 files changed, 8 insertions(+), 24 deletions(-)
arch/x86/mm/init_32.c +0 -6
index 5cb3f0f..804bbe9 100644
@@ -1069,9 +1069,3 @@ void mark_rodata_ro(void)
 #endif
 }
 #endif
-
-int __init reserve_bootmem_generic(unsigned long phys, unsigned long len,
-				   int flags)
-{
-	return reserve_bootmem(phys, len, flags);
-}
arch/x86/mm/init_64.c +2 -18
index e9b040e..5ba6b0e 100644
@@ -798,13 +798,10 @@ void mark_rodata_ro(void)
 
 #endif
 
+#ifndef CONFIG_NO_BOOTMEM
 int __init reserve_bootmem_generic(unsigned long phys, unsigned long len,
 				   int flags)
 {
-#ifdef CONFIG_NUMA
-	int nid, next_nid;
-	int ret;
-#endif
 	unsigned long pfn = phys >> PAGE_SHIFT;
 
 	if (pfn >= max_pfn) {
@@ -820,21 +817,7 @@ int __init reserve_bootmem_generic(unsigned long phys, unsigned long len,
 		return -EFAULT;
 	}
 
-	/* Should check here against the e820 map to avoid double free */
-#ifdef CONFIG_NUMA
-	nid = phys_to_nid(phys);
-	next_nid = phys_to_nid(phys + len - 1);
-	if (nid == next_nid)
-		ret = reserve_bootmem_node(NODE_DATA(nid), phys, len, flags);
-	else
-		ret = reserve_bootmem(phys, len, flags);
-
-	if (ret != 0)
-		return ret;
-
-#else
 	reserve_bootmem(phys, len, flags);
-#endif
 
 	if (phys+len <= MAX_DMA_PFN*PAGE_SIZE) {
 		dma_reserve += len / PAGE_SIZE;
@@ -843,6 +826,7 @@ int __init reserve_bootmem_generic(unsigned long phys, unsigned long len,
 
 	return 0;
 }
+#endif
 
 int kern_addr_valid(unsigned long addr)
 {
mm/bootmem.c +6 -0
index 9b13446..fadbc3b 100644
@@ -512,6 +512,12 @@ int __init reserve_bootmem(unsigned long addr, unsigned long size,
 }
 
 #ifndef CONFIG_NO_BOOTMEM
+int __weak __init reserve_bootmem_generic(unsigned long phys, unsigned long len,
+				   int flags)
+{
+	return reserve_bootmem(phys, len, flags);
+}
+
 static unsigned long __init align_idx(struct bootmem_data *bdata,
 				      unsigned long idx, unsigned long step)
 {
Signature · 1 line
1.6.4.2
author
#35 lore raw

PATCH 10/31 lmb: Add lmb_to_bootmem()

In reply to Yinghai Lu

lmb_to_bootmem() will reserve lmb.reserved.region in bootmem

We can use it to with all arches that support lmb.

---
 include/linux/lmb.h |    2 ++
 mm/lmb.c            |   32 ++++++++++++++++++++++++++++++++
 2 files changed, 34 insertions(+), 0 deletions(-)
include/linux/lmb.h +2 -0
index 7301072..f5071e1 100644
@@ -92,6 +92,8 @@ u64 find_lmb_area(u64 start, u64 end, u64 size, u64 align);
 u64 find_lmb_area_size(u64 start, u64 *sizep, u64 align);
 u64 get_max_mapped(void);
 
+void lmb_to_bootmem(u64 start, u64 end);
+
 #include <asm/lmb.h>
 
 #endif /* __KERNEL__ */
mm/lmb.c +32 -0
index a91f48d..71a45b4 100644
@@ -640,6 +640,38 @@ void __init free_lmb(u64 start, u64 end)
 	lmb_free(start, end - start);
 }
 
+#ifndef CONFIG_NO_BOOTMEM
+void __init lmb_to_bootmem(u64 start, u64 end)
+{
+	int i, count;
+	u64 final_start, final_end;
+
+	/* Take out region array itself */
+	if (lmb.reserved.region != lmb_reserved_region)
+		lmb_free(__pa(lmb.reserved.region), sizeof(struct lmb_property) * lmb.reserved.nr_regions);
+
+	count  = lmb.reserved.cnt;
+	pr_info("(%d early reservations) ==> bootmem [%010llx - %010llx]\n", count, start, end);
+	for (i = 0; i < count; i++) {
+		struct lmb_property *r = &lmb.reserved.region[i];
+		pr_info("  #%d [%010llx - %010llx] ", i, r->base, r->base + r->size);
+		final_start = max(start, r->base);
+		final_end = min(end, r->base + r->size);
+		if (final_start >= final_end) {
+			pr_cont("\n");
+			continue;
+		}
+		pr_cont(" ==> [%010llx - %010llx]\n", final_start, final_end);
+		reserve_bootmem_generic(final_start, final_end - final_start, BOOTMEM_DEFAULT);
+	}
+	/* Clear them to avoid misusing ? */
+	memset(&lmb.reserved.region[0], 0, sizeof(struct lmb_property) * lmb.reserved.nr_regions);
+	lmb.reserved.region = NULL;
+	lmb.reserved.nr_regions = 0;
+	lmb.reserved.cnt = 0;
+}
+#endif
+
 static int __init find_overlapped_early(u64 start, u64 end)
 {
 	int i;
Signature · 1 line
1.6.4.2
author
#36 lore raw

PATCH 11/31 lmb: Add get_free_all_memory_range()

In reply to Yinghai Lu

get_free_all_memory_range is for CONFIG_NO_BOOTMEM, and will be called by free_all_memory_core_early().

It will use early_node_map aka active ranges subtract lmb.reserved to get all free range.

-v2: Update with Jan Beulich's patch "fix allocation done in get_free_all_memory_range()", that one is for early_res.

---
 include/linux/lmb.h |    2 +
 mm/lmb.c            |   86 ++++++++++++++++++++++++++++++++++++++++++++++++++-
 2 files changed, 87 insertions(+), 1 deletions(-)
include/linux/lmb.h +2 -0
index f5071e1..9e2dcf5 100644
@@ -93,6 +93,8 @@ u64 find_lmb_area_size(u64 start, u64 *sizep, u64 align);
 u64 get_max_mapped(void);
 
 void lmb_to_bootmem(u64 start, u64 end);
+struct range;
+int get_free_all_memory_range(struct range **rangep, int nodeid);
 
 #include <asm/lmb.h>
 
mm/lmb.c +85 -1

92 lines not on this page · raw message

Signature · 1 line
1.6.4.2
author
#37 lore raw

PATCH 12/31 lmb: Add lmb_register_active_regions() and lmb_hole_size()

In reply to Yinghai Lu

lmb_register_active_regions() will be used to fill early_node_map, the result will be lmb.memory.region AND numa data

lmb_hole_size will be used to find hole size on lmb.memory.region with specified range.

---
 include/linux/lmb.h |    4 +++
 mm/lmb.c            |   68 +++++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 72 insertions(+), 0 deletions(-)
include/linux/lmb.h +4 -0
index 9e2dcf5..6568e9d 100644
@@ -96,6 +96,10 @@ void lmb_to_bootmem(u64 start, u64 end);
 struct range;
 int get_free_all_memory_range(struct range **rangep, int nodeid);
 
+void lmb_register_active_regions(int nid, unsigned long start_pfn,
+					 unsigned long last_pfn);
+u64 lmb_hole_size(u64 start, u64 end);
+
 #include <asm/lmb.h>
 
 #endif /* __KERNEL__ */
mm/lmb.c +68 -0

71 lines not on this page · raw message

Signature · 1 line
1.6.4.2
author
#38 lore raw

PATCH 13/31 lmb: Prepare to include linux/lmb.h in core file

In reply to Yinghai Lu

Need to add protection in linux/lmb.h, to prepare to include it in mm/page_alloc.c and mm/bootmem.c etc.

---
 include/linux/lmb.h |    3 +++
 1 files changed, 3 insertions(+), 0 deletions(-)
include/linux/lmb.h +3 -0
index 6568e9d..143f97d 100644
@@ -2,6 +2,7 @@
 #define _LINUX_LMB_H
 #ifdef __KERNEL__
 
+#ifdef CONFIG_HAVE_LMB
 /*
  * Logical memory blocks.
  *
@@ -102,6 +103,8 @@ u64 lmb_hole_size(u64 start, u64 end);
 
 #include <asm/lmb.h>
 
+#endif /* CONFIG_HAVE_LMB */
+
 #endif /* __KERNEL__ */
 
 #endif /* _LINUX_LMB_H */
Signature · 1 line
1.6.4.2
author
#39 lore raw

PATCH 14/31 lmb: Add find_memory_core_early()

In reply to Yinghai Lu

According to node range in early_node_map[] with __find_lmb_area to find free range.

Will be used by find_lmb_area_node()

find_lmb_area_node will be used to find right buffer for NODE_DATA

---
 include/linux/mm.h |    2 ++
 mm/page_alloc.c    |   29 +++++++++++++++++++++++++++++
 2 files changed, 31 insertions(+), 0 deletions(-)
include/linux/mm.h +2 -0
index c8442b6..8070bd8 100644
@@ -1159,6 +1159,8 @@ extern void free_bootmem_with_active_regions(int nid,
 						unsigned long max_low_pfn);
 int add_from_early_node_map(struct range *range, int az,
 				   int nr_range, int nid);
+u64 __init find_memory_core_early(int nid, u64 size, u64 align,
+					u64 goal, u64 limit);
 void *__alloc_memory_core_early(int nodeid, u64 size, u64 align,
 				 u64 goal, u64 limit);
 typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
mm/page_alloc.c +29 -0
index d03c946..faae23a 100644
@@ -21,6 +21,7 @@
 #include <linux/pagemap.h>
 #include <linux/jiffies.h>
 #include <linux/bootmem.h>
+#include <linux/lmb.h>
 #include <linux/compiler.h>
 #include <linux/kernel.h>
 #include <linux/kmemcheck.h>
@@ -3393,6 +3394,34 @@ void __init free_bootmem_with_active_regions(int nid,
 	}
 }
 
+#ifdef CONFIG_HAVE_LMB
+u64 __init find_memory_core_early(int nid, u64 size, u64 align,
+					u64 goal, u64 limit)
+{
+	int i;
+
+	/* Need to go over early_node_map to find out good range for node */
+	for_each_active_range_index_in_nid(i, nid) {
+		u64 addr;
+		u64 ei_start, ei_last;
+
+		ei_last = early_node_map[i].end_pfn;
+		ei_last <<= PAGE_SHIFT;
+		ei_start = early_node_map[i].start_pfn;
+		ei_start <<= PAGE_SHIFT;
+		addr = __find_lmb_area(ei_start, ei_last,
+					 goal, limit, size, align);
+
+		if (addr == -1ULL)
+			continue;
+
+		return addr;
+	}
+
+	return -1ULL;
+}
+#endif
+
 int __init add_from_early_node_map(struct range *range, int az,
 				   int nr_range, int nid)
 {
Signature · 1 line
1.6.4.2
author
#40 lore raw

PATCH 15/31 lmb: Add find_lmb_area_node()

In reply to Yinghai Lu

It can be used to find NODE_DATA for numa.

Need to make sure early_node_map[] is filled before it is called

---
 include/linux/lmb.h |    1 +
 mm/lmb.c            |   15 +++++++++++++++
 2 files changed, 16 insertions(+), 0 deletions(-)
include/linux/lmb.h +1 -0
index 143f97d..019520a 100644
@@ -91,6 +91,7 @@ u64 __find_lmb_area(u64 ei_start, u64 ei_last, u64 start, u64 end,
 			 u64 size, u64 align);
 u64 find_lmb_area(u64 start, u64 end, u64 size, u64 align);
 u64 find_lmb_area_size(u64 start, u64 *sizep, u64 align);
+u64 find_lmb_area_node(int nid, u64 start, u64 end, u64 size, u64 align);
 u64 get_max_mapped(void);
 
 void lmb_to_bootmem(u64 start, u64 end);
mm/lmb.c +15 -0
index 7e3c2c1..addfcb1 100644
@@ -913,6 +913,21 @@ u64 __init find_lmb_area_size(u64 start, u64 *sizep, u64 align)
 }
 
 /*
+ * Need to call this function after lmb_register_active_regions,
+ * so early_node_map[] is filled already.
+ */
+u64 __init find_lmb_area_node(int nid, u64 start, u64 end, u64 size, u64 align)
+{
+	u64 addr;
+	addr = find_memory_core_early(nid, size, align, start, end);
+	if (addr != -1ULL)
+		return addr;
+
+	/* Fallback, should already have start end within node range */
+	return find_lmb_area(start, end, size, align);
+}
+
+/*
  * Finds an active region in the address range from start_pfn to last_pfn and
  * returns its range in ei_startpfn and ei_endpfn for the lmb entry.
  */
Signature · 1 line
1.6.4.2
author
#41 lore raw

PATCH 16/31 lmb: Add lmb_free_memory_size()

In reply to Yinghai Lu

It will return free memory size in specified range.

We can not use memory_size - reserved_size here, because some reserved area may not be in the scope of lmb.memory.region.

Use lmb.memory.region subtracting lmb.reserved.region to get free range array. then count size of all free ranges.

---
 include/linux/lmb.h |    1 +
 mm/lmb.c            |   51 +++++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 52 insertions(+), 0 deletions(-)
include/linux/lmb.h +1 -0
index 019520a..51a8653 100644
@@ -101,6 +101,7 @@ int get_free_all_memory_range(struct range **rangep, int nodeid);
 void lmb_register_active_regions(int nid, unsigned long start_pfn,
 					 unsigned long last_pfn);
 u64 lmb_hole_size(u64 start, u64 end);
+u64 lmb_free_memory_size(u64 addr, u64 limit);
 
 #include <asm/lmb.h>
 
mm/lmb.c +51 -0

57 lines not on this page · raw message

Signature · 1 line
1.6.4.2
author
#42 lore raw

PATCH 17/31 lmb: Add lmb_memory_size()

In reply to Yinghai Lu

It will return memory size in specified range according to lmb.memory.region

Try to share some code with lmb_free_memory_size() by passing get_free to __lmb_memory_size().

---
 include/linux/lmb.h |    1 +
 mm/lmb.c            |   18 +++++++++++++++++-
 2 files changed, 18 insertions(+), 1 deletions(-)
include/linux/lmb.h +1 -0
index 51a8653..285f287 100644
@@ -102,6 +102,7 @@ void lmb_register_active_regions(int nid, unsigned long start_pfn,
 					 unsigned long last_pfn);
 u64 lmb_hole_size(u64 start, u64 end);
 u64 lmb_free_memory_size(u64 addr, u64 limit);
+u64 lmb_memory_size(u64 addr, u64 limit);
 
 #include <asm/lmb.h>
 
mm/lmb.c +17 -1
index 233c40d..f49d6c8 100644
@@ -756,7 +756,7 @@ void __init lmb_to_bootmem(u64 start, u64 end)
 }
 #endif
 
-u64 __init lmb_free_memory_size(u64 addr, u64 limit)
+static u64 __init __lmb_memory_size(u64 addr, u64 limit, bool get_free)
 {
 	int i, count;
 	struct range *range;
@@ -786,6 +786,10 @@ u64 __init lmb_free_memory_size(u64 addr, u64 limit)
 	}
 	subtract_range(range, count, 0, addr);
 	subtract_range(range, count, limit, -1ULL);
+
+	/* Subtract lmb.reserved.region in range ? */
+	if (!get_free)
+		goto sort_and_count_them;
 	for (i = 0; i < lmb.reserved.cnt; i++) {
 		struct lmb_property *r = &lmb.reserved.region[i];
 
@@ -798,6 +802,8 @@ u64 __init lmb_free_memory_size(u64 addr, u64 limit)
 
 		subtract_range(range, count, final_start, final_end);
 	}
+
+sort_and_count_them:
 	nr_range = clean_sort_range(range, count);
 
 	free_size = 0;
@@ -807,6 +813,16 @@ u64 __init lmb_free_memory_size(u64 addr, u64 limit)
 	return free_size << PAGE_SHIFT;
 }
 
+u64 __init lmb_free_memory_size(u64 addr, u64 limit)
+{
+	return __lmb_memory_size(addr, limit, true);
+}
+
+u64 __init lmb_memory_size(u64 addr, u64 limit)
+{
+	return __lmb_memory_size(addr, limit, false);
+}
+
 static int __init find_overlapped_early(u64 start, u64 end)
 {
 	int i;
Signature · 1 line
1.6.4.2
author
#43 lore raw

PATCH 18/31 lmb: Add reserve_lmb_overlap_ok()

In reply to Yinghai Lu

Some areas from firmware could be reserved several times from different callers.

If these area are overlapped, We may have overlapped entries in lmb.reserved.

Try to free the area at first, before rerserve them again.

---
 include/linux/lmb.h |    1 +
 mm/lmb.c            |   27 +++++++++++++++++++++++++--
 2 files changed, 26 insertions(+), 2 deletions(-)
include/linux/lmb.h +1 -0
index 285f287..1e11fe0 100644
@@ -85,6 +85,7 @@ lmb_end_pfn(struct lmb_region *type, unsigned long region_nr)
 }
 
 void reserve_lmb(u64 start, u64 end, char *name);
+void reserve_lmb_overlap_ok(u64 start, u64 end, char *name);
 void free_lmb(u64 start, u64 end);
 void add_lmb_memory(u64 start, u64 end);
 u64 __find_lmb_area(u64 ei_start, u64 ei_last, u64 start, u64 end,
mm/lmb.c +25 -2
index f49d6c8..a709522 100644
@@ -615,6 +615,12 @@ void __init add_lmb_memory(u64 start, u64 end)
 	lmb_add(start, end - start);
 }
 
+static void __init __reserve_lmb(u64 start, u64 end, char *name)
+{
+	__check_and_double_region_array(&lmb.reserved, &lmb_reserved_region[0], start, end);
+	lmb_reserve(start, end - start);
+}
+
 void __init reserve_lmb(u64 start, u64 end, char *name)
 {
 	if (start == end)
@@ -623,8 +629,25 @@ void __init reserve_lmb(u64 start, u64 end, char *name)
 	if (WARN_ONCE(start > end, "reserve_lmb: wrong range [%#llx, %#llx]\n", start, end))
 		return;
 
-	__check_and_double_region_array(&lmb.reserved, &lmb_reserved_region[0], start, end);
-	lmb_reserve(start, end - start);
+	__reserve_lmb(start, end, name);
+}
+
+/*
+ * Could be used to avoid having overlap entries in lmb.reserved.region.
+ *  Don't need to use it with area that is from find_lmb_area()
+ *  Only use it for the area that fw hidden area.
+ */
+void __init reserve_lmb_overlap_ok(u64 start, u64 end, char *name)
+{
+	if (start == end)
+		return;
+
+	if (WARN_ONCE(start > end, "reserve_lmb_overlap_ok: wrong range [%#llx, %#llx]\n", start, end))
+		return;
+
+	/* Free that region at first */
+	lmb_free(start, end - start);
+	__reserve_lmb(start, end, name);
 }
 
 void __init free_lmb(u64 start, u64 end)
Signature · 1 line
1.6.4.2
author
#44 lore raw

PATCH 19/31 lmb: Use lmb_debug to control debug message print out

In reply to Yinghai Lu

Also let reserve_lmb/free_lmb could print out name if lmb=debug is specified

---
 mm/lmb.c |   29 +++++++++++++++++++++--------
 1 files changed, 21 insertions(+), 8 deletions(-)
mm/lmb.c +21 -8
index a709522..91a2b33 100644
@@ -600,8 +600,9 @@ static void __init __check_and_double_region_array(struct lmb_region *type,
 	memset(&old[0], 0, sizeof(struct lmb_property) * rgnsz);
 	type->region = new;
 	type->nr_regions = rgnsz * 2;
-	printk(KERN_DEBUG "lmb.reserved.region array is doubled to %ld at [%llx - %llx]\n",
-		type->nr_regions, mem, mem + size - 1);
+	if (lmb_debug)
+		pr_info("lmb.reserved.region array is doubled to %ld at [%010llx - %010llx]\n",
+				type->nr_regions, mem, mem + size - 1);
 
 	/* Reserve new array and free old one */
 	lmb_reserve(mem, sizeof(struct lmb_property) * rgnsz * 2);
@@ -629,6 +630,8 @@ void __init reserve_lmb(u64 start, u64 end, char *name)
 	if (WARN_ONCE(start > end, "reserve_lmb: wrong range [%#llx, %#llx]\n", start, end))
 		return;
 
+	if (lmb_debug)
+		pr_info("    reserve_lmb: [%010llx, %010llx] %16s\n", start, end, name);
 	__reserve_lmb(start, end, name);
 }
 
@@ -645,6 +648,8 @@ void __init reserve_lmb_overlap_ok(u64 start, u64 end, char *name)
 	if (WARN_ONCE(start > end, "reserve_lmb_overlap_ok: wrong range [%#llx, %#llx]\n", start, end))
 		return;
 
+	if (lmb_debug)
+		pr_info("    reserve_lmb_overlap_ok: [%010llx, %010llx] %16s\n", start, end, name);
 	/* Free that region at first */
 	lmb_free(start, end - start);
 	__reserve_lmb(start, end, name);
@@ -658,6 +663,8 @@ void __init free_lmb(u64 start, u64 end)
 	if (WARN_ONCE(start > end, "free_lmb: wrong range [%#llx, %#llx]\n", start, end))
 		return;
 
+	if (lmb_debug)
+		pr_info("       free_lmb: [%010llx, %010llx]\n", start, end);
 	/* keep punching hole, could run out of slots too */
 	__check_and_double_region_array(&lmb.reserved, &lmb_reserved_region[0], start, end);
 	lmb_free(start, end - start);
@@ -698,11 +705,13 @@ static void __init subtract_lmb_reserved(struct range *range, int az)
 
 	count  = lmb.reserved.cnt;
 
-	pr_info("Subtract (%d early reservations)\n", count);
+	if (lmb_debug)
+		pr_info("Subtract (%d early reservations)\n", count);
 
 	for (i = 0; i < count; i++) {
 		struct lmb_property *r = &lmb.reserved.region[i];
-		pr_info("  #%d [%010llx - %010llx]\n", i, r->base, r->base + r->size);
+		if (lmb_debug)
+			pr_info("  #%03d [%010llx - %010llx]\n", i, r->base, r->base + r->size);
 		final_start = PFN_DOWN(r->base);
 		final_end = PFN_UP(r->base + r->size);
 		if (final_start >= final_end)
@@ -758,17 +767,21 @@ void __init lmb_to_bootmem(u64 start, u64 end)
 		lmb_free(__pa(lmb.reserved.region), sizeof(struct lmb_property) * lmb.reserved.nr_regions);
 
 	count  = lmb.reserved.cnt;
-	pr_info("(%d early reservations) ==> bootmem [%010llx - %010llx]\n", count, start, end);
+	if (lmb_debug)
+		pr_info("(%d early reservations) ==> bootmem [%010llx - %010llx]\n", count, start, end);
 	for (i = 0; i < count; i++) {
 		struct lmb_property *r = &lmb.reserved.region[i];
-		pr_info("  #%d [%010llx - %010llx] ", i, r->base, r->base + r->size);
+		if (lmb_debug)
+			pr_info("  #%03d [%010llx - %010llx] ", i, r->base, r->base + r->size);
 		final_start = max(start, r->base);
 		final_end = min(end, r->base + r->size);
 		if (final_start >= final_end) {
-			pr_cont("\n");
+			if (lmb_debug)
+				pr_cont("\n");
 			continue;
 		}
-		pr_cont(" ==> [%010llx - %010llx]\n", final_start, final_end);
+		if (lmb_debug)
+			pr_cont(" ==> [%010llx - %010llx]\n", final_start, final_end);
 		reserve_bootmem_generic(final_start, final_end - final_start, BOOTMEM_DEFAULT);
 	}
 	/* Clear them to avoid misusing ? */
Signature · 1 line
1.6.4.2
author
#45 lore raw

PATCH 20/31 lmb: Add __NOT_KEEP_LMB to put lmb code to .init

In reply to Yinghai Lu

So those lmb bits could released after kernel is booted up.

Arch code could define __NOT_KEEP_LMB in asm/lmb.h, __init_lmb will become __init

x86 code will use that.

---
 include/linux/lmb.h |    8 +++++++
 mm/lmb.c            |   54 ++++++++++++++++++++++++++++++--------------------
 2 files changed, 40 insertions(+), 22 deletions(-)
include/linux/lmb.h +8 -0
index 1e11fe0..7ae52c8 100644
@@ -107,6 +107,14 @@ u64 lmb_memory_size(u64 addr, u64 limit);
 
 #include <asm/lmb.h>
 
+#ifdef __NOT_KEEP_LMB
+#define __init_lmb __init
+#define __initdata_lmb __initdata
+#else
+#define __init_lmb
+#define __initdata_lmb
+#endif
+
 #endif /* CONFIG_HAVE_LMB */
 
 #endif /* __KERNEL__ */
mm/lmb.c +32 -22

210 lines not on this page · raw message

Signature · 1 line
1.6.4.2
#46 lore raw

Re: [PATCH 20/31] lmb: Add __NOT_KEEP_LMB to put lmb code to .init

Quoted 3 lines · Yinghai Lu

So those lmb bits could released after kernel is booted up.

Arch code could define __NOT_KEEP_LMB in asm/lmb.h, __init_lmb will become __init

ARCH_KEEP_LMB or ARCH_DISCARD_LMB, or something like that, would be more in keeping with existing flags like this.

cheers

author
#47 lore raw

Re: [PATCH 20/31] lmb: Add __NOT_KEEP_LMB to put lmb code to .init

Quoted 8 lines · Michael Ellerman
Quoted 3 lines · Yinghai Lu

So those lmb bits could released after kernel is booted up.

Arch code could define __NOT_KEEP_LMB in asm/lmb.h, __init_lmb will become __init

ARCH_KEEP_LMB or ARCH_DISCARD_LMB, or something like that, would be more in keeping with existing flags like this.

ok, will use ARCH_DISCARD_LMB.

BTW, it seems only PowerPC need to keep lmb after init stage, right ?

Thanks

Yinghai

#48 lore raw

Re: [PATCH 20/31] lmb: Add __NOT_KEEP_LMB to put lmb code to .init

From: Yinghai Lu <[email protected]>
Date: Mon, 29 Mar 2010 09:20:28 -0700

Quoted 1 line

BTW, it seems only PowerPC need to keep lmb after init stage, right ?

For now. Sparc64 will need it at some point in the future.

author
#49 lore raw

Re: [PATCH 20/31] lmb: Add __NOT_KEEP_LMB to put lmb code to .init

Quoted 6 lines · David Miller

From: Yinghai Lu <[email protected]>
Date: Mon, 29 Mar 2010 09:20:28 -0700

Quoted 1 line

BTW, it seems only PowerPC need to keep lmb after init stage, right ?

For now. Sparc64 will need it at some point in the future.

for memory hotplug support?

Yinghai

#50 lore raw

Re: [PATCH 20/31] lmb: Add __NOT_KEEP_LMB to put lmb code to .init

From: Yinghai Lu <[email protected]>
Date: Mon, 29 Mar 2010 11:39:52 -0700

Quoted 9 lines · David Miller
Quoted 6 lines

From: Yinghai Lu <[email protected]>
Date: Mon, 29 Mar 2010 09:20:28 -0700

Quoted 1 line

BTW, it seems only PowerPC need to keep lmb after init stage, right ?

For now. Sparc64 will need it at some point in the future.

for memory hotplug support?

Yes.

#51 lore raw

Re: [PATCH 20/31] lmb: Add __NOT_KEEP_LMB to put lmb code to .init

Quoted 7 lines · David Miller

From: Yinghai Lu <[email protected]>
Date: Mon, 29 Mar 2010 11:39:52 -0700

Quoted 4 lines · David Miller
Quoted 3 lines

From: Yinghai Lu <[email protected]>
Date: Mon, 29 Mar 2010 09:20:28 -0700

Quoted 7 lines
Quoted 5 lines
Quoted 3 lines
Quoted 1 line

BTW, it seems only PowerPC need to keep lmb after init stage, right ?

For now. Sparc64 will need it at some point in the future.

for memory hotplug support?

Yes.

We also can use it to implement page_is_ram() which might help get /dev/mem cachability setting right by default one day :-)

However, we only need to keep the memory list, not the reserve list.

Cheers,
Ben.

author
#52 lore raw

PATCH 21/31 x86: Add sanitize_e820_map()

In reply to Yinghai Lu

So We don't need to take e820.map with it.

Also change e820_saved to initdata to get some bytes memory back.

---
 arch/x86/include/asm/e820.h |    5 ++---
 arch/x86/kernel/e820.c      |   26 ++++++++++++++++++--------
 arch/x86/kernel/efi.c       |    2 +-
 arch/x86/kernel/setup.c     |   10 +++++-----
 arch/x86/xen/setup.c        |    4 +---
 5 files changed, 27 insertions(+), 20 deletions(-)
arch/x86/include/asm/e820.h +2 -3
index ec8a52d..0457c49 100644
@@ -75,15 +75,14 @@ struct e820map {
 #ifdef __KERNEL__
 /* see comment in arch/x86/kernel/e820.c */
 extern struct e820map e820;
-extern struct e820map e820_saved;
 
 extern unsigned long pci_mem_start;
 extern int e820_any_mapped(u64 start, u64 end, unsigned type);
 extern int e820_all_mapped(u64 start, u64 end, unsigned type);
 extern void e820_add_region(u64 start, u64 size, int type);
 extern void e820_print_map(char *who);
-extern int
-sanitize_e820_map(struct e820entry *biosmap, int max_nr_map, u32 *pnr_map);
+int sanitize_e820_map(void);
+void save_e820_map(void);
 extern u64 e820_update_range(u64 start, u64 size, unsigned old_type,
 			       unsigned new_type);
 extern u64 e820_remove_range(u64 start, u64 size, unsigned old_type,
arch/x86/kernel/e820.c +18 -8
index 740b440..0eb9830 100644
@@ -35,7 +35,7 @@
  * next kernel with full memory.
  */
 struct e820map e820;
-struct e820map e820_saved;
+static struct e820map __initdata e820_saved;
 
 /* For PCI or other memory-mapped resources */
 unsigned long pci_mem_start = 0xaeedbabe;
@@ -224,7 +224,7 @@ void __init e820_print_map(char *who)
  *	   ______________________4_
  */
 
-int __init sanitize_e820_map(struct e820entry *biosmap, int max_nr_map,
+static int __init __sanitize_e820_map(struct e820entry *biosmap, int max_nr_map,
 			     u32 *pnr_map)
 {
 	struct change_member {
@@ -383,6 +383,11 @@ int __init sanitize_e820_map(struct e820entry *biosmap, int max_nr_map,
 	return 0;
 }
 
+int __init sanitize_e820_map(void)
+{
+	return __sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
+}
+
 static int __init __append_e820_map(struct e820entry *biosmap, int nr_map)
 {
 	while (nr_map) {
@@ -555,7 +560,7 @@ void __init update_e820(void)
 	u32 nr_map;
 
 	nr_map = e820.nr_map;
-	if (sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &nr_map))
+	if (__sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &nr_map))
 		return;
 	e820.nr_map = nr_map;
 	printk(KERN_INFO "modified physical RAM map:\n");
@@ -566,7 +571,7 @@ static void __init update_e820_saved(void)
 	u32 nr_map;
 
 	nr_map = e820_saved.nr_map;
-	if (sanitize_e820_map(e820_saved.map, ARRAY_SIZE(e820_saved.map), &nr_map))
+	if (__sanitize_e820_map(e820_saved.map, ARRAY_SIZE(e820_saved.map), &nr_map))
 		return;
 	e820_saved.nr_map = nr_map;
 }
@@ -661,7 +666,7 @@ void __init parse_e820_ext(struct setup_data *sdata, unsigned long pa_data)
 		sdata = early_ioremap(pa_data, map_len);
 	extmap = (struct e820entry *)(sdata->data);
 	__append_e820_map(extmap, entries);
-	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
+	sanitize_e820_map();
 	if (map_len > PAGE_SIZE)
 		early_iounmap(sdata, map_len);
 	printk(KERN_INFO "extended physical RAM map:\n");
@@ -1028,7 +1033,7 @@ void __init finish_e820_parsing(void)
 	if (userdef) {
 		u32 nr = e820.nr_map;
 
-		if (sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &nr) < 0)
+		if (__sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &nr) < 0)
 			early_panic("Invalid user supplied memory map");
 		e820.nr_map = nr;
 
@@ -1158,7 +1163,7 @@ char *__init default_machine_specific_memory_setup(void)
 	 * the next section from 1mb->appropriate_mem_k
 	 */
 	new_nr = boot_params.e820_entries;
-	sanitize_e820_map(boot_params.e820_map,
+	__sanitize_e820_map(boot_params.e820_map,
 			ARRAY_SIZE(boot_params.e820_map),
 			&new_nr);
 	boot_params.e820_entries = new_nr;
@@ -1185,12 +1190,17 @@ char *__init default_machine_specific_memory_setup(void)
 	return who;
 }
 
+void __init save_e820_map(void)
+{
+	memcpy(&e820_saved, &e820, sizeof(struct e820map));
+}
+
 void __init setup_memory_map(void)
 {
 	char *who;
 
 	who = x86_init.resources.memory_setup();
-	memcpy(&e820_saved, &e820, sizeof(struct e820map));
+	save_e820_map();
 	printk(KERN_INFO "BIOS-provided physical RAM map:\n");
 	e820_print_map(who);
 }
arch/x86/kernel/efi.c +1 -1
index c2fa9b8..299f03f 100644
@@ -272,7 +272,7 @@ static void __init do_add_efi_memmap(void)
 		}
 		e820_add_region(start, size, e820_type);
 	}
-	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
+	sanitize_e820_map();
 }
 
 void __init efi_reserve_early(void)
arch/x86/kernel/setup.c +5 -5
index d76e185..7ca6878 100644
@@ -463,8 +463,8 @@ static void __init e820_reserve_setup_data(void)
 	if (!found)
 		return;
 
-	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
-	memcpy(&e820_saved, &e820, sizeof(struct e820map));
+	sanitize_e820_map();
+	save_e820_map();
 	printk(KERN_INFO "extended physical RAM map:\n");
 	e820_print_map("reserve setup_data");
 }
@@ -616,7 +616,7 @@ static int __init dmi_low_memory_corruption(const struct dmi_system_id *d)
 		d->ident);
 
 	e820_update_range(0, 0x10000, E820_RAM, E820_RESERVED);
-	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
+	sanitize_e820_map();
 
 	return 0;
 }
@@ -685,7 +685,7 @@ static void __init trim_bios_range(void)
 	 * take them out.
 	 */
 	e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
-	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
+	sanitize_e820_map();
 }
 
 /*
@@ -856,7 +856,7 @@ void __init setup_arch(char **cmdline_p)
 	if (ppro_with_ram_bug()) {
 		e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
 				  E820_RESERVED);
-		sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
+		sanitize_e820_map();
 		printk(KERN_INFO "fixed physical RAM map:\n");
 		e820_print_map("bad_ppro");
 	}
arch/x86/xen/setup.c +1 -3
index ad0047f..3f2c411 100644
@@ -43,8 +43,6 @@ char * __init xen_memory_setup(void)
 
 	max_pfn = min(MAX_DOMAIN_PAGES, max_pfn);
 
-	e820.nr_map = 0;
-
 	e820_add_region(0, PFN_PHYS((u64)max_pfn), E820_RAM);
 
 	/*
@@ -65,7 +63,7 @@ char * __init xen_memory_setup(void)
 		      __pa(xen_start_info->pt_base),
 			"XEN START INFO");
 
-	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
+	sanitize_e820_map();
 
 	return "Xen";
 }
Signature · 1 line
1.6.4.2
author
#53 lore raw

PATCH 22/31 x86: Use lmb to replace early_res

In reply to Yinghai Lu

1. replace find_e820_area with find_lmb_area
2. replace reserve_early with reserve_lmb
3. replace free_early with free_lmb.
4. NO_BOOTMEM will switch to use lmb too.
5. use _e820, _early wrap in the patch, in following patch, will replace them all
6. because free_lmb support partial free, we can remove some special care
7. Need to make sure that find_lmb_area() is called after fill_lmb_memory() so adjust some calling later in setup.c::setup_arch()
-- corruption_check and mptable_update

-v2: Move reserve_brk() early
Before fill_lmb_area, to avoid overlap between brk and find_lmb_area() that could happen We have more then 128 RAM entry in E820 tables, and fill_lmb_memory() could use find_lmb_area() to find a new place for lmb.memory.region array.
and We don't need to use extend_brk() after fill_lmb_area()
-v3: Move find_smp_config early
To make sure find_lmb_area not find wrong place, if BIOS doesn't put mptable in right place.
-v4: Treat RESERVED_KERN as RAM in lmb.memory. and they are already in
lmb.reserved already..
use __NOT_KEEP_LMB to make sure lmb related code could be freed later.

So move reserve_brk() early before fill_lmb_area().

---
 arch/x86/Kconfig               |    9 +--
 arch/x86/include/asm/e820.h    |   15 +++--
 arch/x86/include/asm/lmb.h     |   10 +++
 arch/x86/kernel/check.c        |   14 ++--
 arch/x86/kernel/e820.c         |  139 +++++++++++-----------------------------
 arch/x86/kernel/head.c         |    3 +-
 arch/x86/kernel/head32.c       |    6 +-
 arch/x86/kernel/head64.c       |    3 +
 arch/x86/kernel/mpparse.c      |    5 +-
 arch/x86/kernel/setup.c        |   29 ++++++---
 arch/x86/kernel/setup_percpu.c |    6 --
 arch/x86/mm/numa_64.c          |    5 +-
 kernel/Makefile                |    1 -
 mm/bootmem.c                   |    1 +
 mm/page_alloc.c                |   35 ++--------
 mm/sparse-vmemmap.c            |   11 ---
 16 files changed, 111 insertions(+), 181 deletions(-)
 create mode 100644 arch/x86/include/asm/lmb.h
arch/x86/Kconfig +3 -6
index 6a80bce..3117de5 100644
@@ -27,6 +27,7 @@ config X86
 	select HAVE_PERF_EVENTS if (!M386 && !M486)
 	select HAVE_IOREMAP_PROT
 	select HAVE_KPROBES
+	select HAVE_LMB
 	select ARCH_WANT_OPTIONAL_GPIOLIB
 	select ARCH_WANT_FRAME_POINTERS
 	select HAVE_DMA_ATTRS
@@ -192,9 +193,6 @@ config ARCH_SUPPORTS_OPTIMIZED_INLINING
 config ARCH_SUPPORTS_DEBUG_PAGEALLOC
 	def_bool y
 
-config HAVE_EARLY_RES
-	def_bool y
-
 config HAVE_INTEL_TXT
 	def_bool y
 	depends on EXPERIMENTAL && DMAR && ACPI
@@ -589,14 +587,13 @@ config NO_BOOTMEM
 	default y
 	bool "Disable Bootmem code"
 	---help---
-	  Use early_res directly instead of bootmem before slab is ready.
+	  Use lmb directly instead of bootmem before slab is ready.
 		- allocator (buddy) [generic]
 		- early allocator (bootmem) [generic]
-		- very early allocator (reserve_early*()) [x86]
+		- very early allocator (lmb) [some generic]
 		- very very early allocator (early brk model) [x86]
 	  So reduce one layer between early allocator to final allocator
 
-
 config MEMTEST
 	bool "Memtest"
 	---help---
arch/x86/include/asm/e820.h +9 -6
index 0457c49..396c849 100644
@@ -116,24 +116,27 @@ extern unsigned long end_user_pfn;
 extern u64 find_e820_area(u64 start, u64 end, u64 size, u64 align);
 extern u64 find_e820_area_size(u64 start, u64 *sizep, u64 align);
 extern u64 early_reserve_e820(u64 startt, u64 sizet, u64 align);
-#include <linux/early_res.h>
 
 extern unsigned long e820_end_of_ram_pfn(void);
 extern unsigned long e820_end_of_low_ram_pfn(void);
-extern int e820_find_active_region(const struct e820entry *ei,
-				  unsigned long start_pfn,
-				  unsigned long last_pfn,
-				  unsigned long *ei_startpfn,
-				  unsigned long *ei_endpfn);
 extern void e820_register_active_regions(int nid, unsigned long start_pfn,
 					 unsigned long end_pfn);
 extern u64 e820_hole_size(u64 start, u64 end);
+
+extern u64 early_reserve_e820(u64 startt, u64 sizet, u64 align);
+
+void init_lmb_memory(void);
+void fill_lmb_memory(void);
+
 extern void finish_e820_parsing(void);
 extern void e820_reserve_resources(void);
 extern void e820_reserve_resources_late(void);
 extern void setup_memory_map(void);
 extern char *default_machine_specific_memory_setup(void);
 
+void reserve_early(u64 start, u64 end, char *name);
+void free_early(u64 start, u64 end);
+
 /*
  * Returns true iff the specified range [s,e) is completely contained inside
  * the ISA region.
arch/x86/include/asm/lmb.h new file +10 -0
new file mode 100644
index 0000000..2fd1db4
@@ -0,0 +1,10 @@
+#ifndef _X86_LMB_H
+#define _X86_LMB_H
+
+#define LMB_DBG(fmt...) printk(fmt)
+
+#define LMB_REAL_LIMIT 0
+
+#define __NOT_KEEP_LMB
+
+#endif
arch/x86/kernel/check.c +8 -6
index fc999e6..90236af 100644
@@ -2,7 +2,8 @@
 #include <linux/sched.h>
 #include <linux/kthread.h>
 #include <linux/workqueue.h>
-#include <asm/e820.h>
+#include <linux/lmb.h>
+
 #include <asm/proto.h>
 
 /*
@@ -18,10 +19,12 @@ static int __read_mostly memory_corruption_check = -1;
 static unsigned __read_mostly corruption_check_size = 64*1024;
 static unsigned __read_mostly corruption_check_period = 60; /* seconds */
 
-static struct e820entry scan_areas[MAX_SCAN_AREAS];
+static struct scan_area {
+	u64 addr;
+	u64 size;
+} scan_areas[MAX_SCAN_AREAS];
 static int num_scan_areas;
 
-
 static __init int set_corruption_check(char *arg)
 {
 	char *end;
@@ -81,7 +84,7 @@ void __init setup_bios_corruption_check(void)
 
 	while (addr < corruption_check_size && num_scan_areas < MAX_SCAN_AREAS) {
 		u64 size;
-		addr = find_e820_area_size(addr, &size, PAGE_SIZE);
+		addr = find_lmb_area_size(addr, &size, PAGE_SIZE);
 
 		if (!(addr + 1))
 			break;
@@ -92,7 +95,7 @@ void __init setup_bios_corruption_check(void)
 		if ((addr + size) > corruption_check_size)
 			size = corruption_check_size - addr;
 
-		e820_update_range(addr, size, E820_RAM, E820_RESERVED);
+		reserve_lmb(addr, addr + size, "SCAN RAM");
 		scan_areas[num_scan_areas].addr = addr;
 		scan_areas[num_scan_areas].size = size;
 		num_scan_areas++;
@@ -105,7 +108,6 @@ void __init setup_bios_corruption_check(void)
 
 	printk(KERN_INFO "Scanning %d areas for low memory corruption\n",
 	       num_scan_areas);
-	update_e820();
 }
 
 
arch/x86/kernel/e820.c +38 -101

194 lines not on this page · raw message

arch/x86/kernel/head.c +2 -1
index 3e66bd3..0341b09 100644
@@ -1,5 +1,6 @@
 #include <linux/kernel.h>
 #include <linux/init.h>
+#include <linux/lmb.h>
 
 #include <asm/setup.h>
 #include <asm/bios_ebda.h>
@@ -51,5 +52,5 @@ void __init reserve_ebda_region(void)
 		lowmem = 0x9f000;
 
 	/* reserve all memory between lowmem and the 1MB mark */
-	reserve_early_overlap_ok(lowmem, 0x100000, "BIOS reserved");
+	reserve_lmb_overlap_ok(lowmem, 0x100000, "BIOS reserved");
 }
arch/x86/kernel/head32.c +4 -2
index 9a97504..cbab479 100644
@@ -7,6 +7,7 @@
 
 #include <linux/init.h>
 #include <linux/start_kernel.h>
+#include <linux/lmb.h>
 
 #include <asm/setup.h>
 #include <asm/sections.h>
@@ -29,14 +30,15 @@ static void __init i386_default_early_setup(void)
 
 void __init i386_start_kernel(void)
 {
+	init_lmb_memory();
+
 #ifdef CONFIG_X86_TRAMPOLINE
 	/*
 	 * But first pinch a few for the stack/trampoline stuff
 	 * FIXME: Don't need the extra page at 4K, but need to fix
 	 * trampoline before removing it. (see the GDT stuff)
 	 */
-	reserve_early_overlap_ok(PAGE_SIZE, PAGE_SIZE + PAGE_SIZE,
-					 "EX TRAMPOLINE");
+	reserve_lmb(PAGE_SIZE, PAGE_SIZE + PAGE_SIZE, "EX TRAMPOLINE");
 #endif
 
 	reserve_early(__pa_symbol(&_text), __pa_symbol(&__bss_stop), "TEXT DATA BSS");
arch/x86/kernel/head64.c +3 -0
index 7147143..89dd2de 100644
@@ -12,6 +12,7 @@
 #include <linux/percpu.h>
 #include <linux/start_kernel.h>
 #include <linux/io.h>
+#include <linux/lmb.h>
 
 #include <asm/processor.h>
 #include <asm/proto.h>
@@ -96,6 +97,8 @@ void __init x86_64_start_kernel(char * real_mode_data)
 
 void __init x86_64_start_reservations(char *real_mode_data)
 {
+	init_lmb_memory();
+
 	copy_bootdata(__va(real_mode_data));
 
 	reserve_early(__pa_symbol(&_text), __pa_symbol(&__bss_stop), "TEXT DATA BSS");
arch/x86/kernel/mpparse.c +3 -2
index a2c1edd..5779c27 100644
@@ -11,6 +11,7 @@
 #include <linux/init.h>
 #include <linux/delay.h>
 #include <linux/bootmem.h>
+#include <linux/lmb.h>
 #include <linux/kernel_stat.h>
 #include <linux/mc146818rtc.h>
 #include <linux/bitops.h>
@@ -664,7 +665,7 @@ static void __init smp_reserve_memory(struct mpf_intel *mpf)
 {
 	unsigned long size = get_mpc_size(mpf->physptr);
 
-	reserve_early(mpf->physptr, mpf->physptr+size, "MP-table mpc");
+	reserve_lmb_overlap_ok(mpf->physptr, mpf->physptr+size, "MP-table mpc");
 }
 
 static int __init smp_scan_config(unsigned long base, unsigned long length)
@@ -693,7 +694,7 @@ static int __init smp_scan_config(unsigned long base, unsigned long length)
 			       mpf, (u64)virt_to_phys(mpf));
 
 			mem = virt_to_phys(mpf);
-			reserve_early(mem, mem + sizeof(*mpf), "MP-table mpf");
+			reserve_lmb_overlap_ok(mem, mem + sizeof(*mpf), "MP-table mpf");
 			if (mpf->physptr)
 				smp_reserve_memory(mpf);
 
arch/x86/kernel/setup.c +19 -10
index 7ca6878..a5c029f 100644
@@ -31,6 +31,7 @@
 #include <linux/apm_bios.h>
 #include <linux/initrd.h>
 #include <linux/bootmem.h>
+#include <linux/lmb.h>
 #include <linux/seq_file.h>
 #include <linux/console.h>
 #include <linux/mca.h>
@@ -870,8 +871,6 @@ void __init setup_arch(char **cmdline_p)
 	 */
 	max_pfn = e820_end_of_ram_pfn();
 
-	/* preallocate 4k for mptable mpc */
-	early_reserve_e820_mpc_new();
 	/* update e820 for memory not covered by WB MTRRs */
 	mtrr_bp_init();
 	if (mtrr_trim_uncached_memory(max_pfn))
@@ -896,6 +895,23 @@ void __init setup_arch(char **cmdline_p)
 	max_pfn_mapped = KERNEL_IMAGE_SIZE >> PAGE_SHIFT;
 #endif
 
+	/*
+	 * Find and reserve possible boot-time SMP configuration:
+	 */
+	find_smp_config();
+
+	/*
+	 * Need to conclude brk, before fill_lmb_memory()
+	 *  it could use find_lmb_area, could overlap with
+	 *  brk area.
+	 */
+	reserve_brk();
+
+	fill_lmb_memory();
+
+	/* preallocate 4k for mptable mpc */
+	early_reserve_e820_mpc_new();
+
 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
 	setup_bios_corruption_check();
 #endif
@@ -903,13 +919,6 @@ void __init setup_arch(char **cmdline_p)
 	printk(KERN_DEBUG "initial memory mapped : 0 - %08lx\n",
 			max_pfn_mapped<<PAGE_SHIFT);
 
-	reserve_brk();
-
-	/*
-	 * Find and reserve possible boot-time SMP configuration:
-	 */
-	find_smp_config();
-
 	reserve_trampoline_memory();
 
 #ifdef CONFIG_ACPI_SLEEP
@@ -972,7 +981,7 @@ void __init setup_arch(char **cmdline_p)
 
 	initmem_init(0, max_pfn, acpi, k8);
 #ifndef CONFIG_NO_BOOTMEM
-	early_res_to_bootmem(0, max_low_pfn<<PAGE_SHIFT);
+	lmb_to_bootmem(0, max_low_pfn<<PAGE_SHIFT);
 #endif
 
 	dma32_reserve_bootmem();
arch/x86/kernel/setup_percpu.c +0 -6
index ef6370b..35abcb8 100644
@@ -137,13 +137,7 @@ static void * __init pcpu_fc_alloc(unsigned int cpu, size_t size, size_t align)
 
 static void __init pcpu_fc_free(void *ptr, size_t size)
 {
-#ifdef CONFIG_NO_BOOTMEM
-	u64 start = __pa(ptr);
-	u64 end = start + size;
-	free_early_partial(start, end);
-#else
 	free_bootmem(__pa(ptr), size);
-#endif
 }
 
 static int __init pcpu_cpu_distance(unsigned int from, unsigned int to)
arch/x86/mm/numa_64.c +3 -2
index 8948f47..fc9a403 100644
@@ -7,6 +7,7 @@
 #include <linux/string.h>
 #include <linux/init.h>
 #include <linux/bootmem.h>
+#include <linux/lmb.h>
 #include <linux/mmzone.h>
 #include <linux/ctype.h>
 #include <linux/module.h>
@@ -174,7 +175,7 @@ static void * __init early_node_mem(int nodeid, unsigned long start,
 	if (start < (MAX_DMA32_PFN<<PAGE_SHIFT) &&
 	    end > (MAX_DMA32_PFN<<PAGE_SHIFT))
 		start = MAX_DMA32_PFN<<PAGE_SHIFT;
-	mem = find_e820_area(start, end, size, align);
+	mem = find_lmb_area_node(nodeid, start, end, size, align);
 	if (mem != -1L)
 		return __va(mem);
 
@@ -184,7 +185,7 @@ static void * __init early_node_mem(int nodeid, unsigned long start,
 		start = MAX_DMA32_PFN<<PAGE_SHIFT;
 	else
 		start = MAX_DMA_PFN<<PAGE_SHIFT;
-	mem = find_e820_area(start, end, size, align);
+	mem = find_lmb_area_node(nodeid, start, end, size, align);
 	if (mem != -1L)
 		return __va(mem);
 
kernel/Makefile +0 -1
index d5c3006..754bf79 100644
@@ -11,7 +11,6 @@ obj-y     = sched.o fork.o exec_domain.o panic.o printk.o \
 	    hrtimer.o rwsem.o nsproxy.o srcu.o semaphore.o \
 	    notifier.o ksysfs.o pm_qos_params.o sched_clock.o cred.o \
 	    async.o range.o
-obj-$(CONFIG_HAVE_EARLY_RES) += early_res.o
 obj-y += groups.o
 
 ifdef CONFIG_FUNCTION_TRACER
mm/bootmem.c +1 -0
index fadbc3b..81f9350 100644
@@ -14,6 +14,7 @@
 #include <linux/module.h>
 #include <linux/kmemleak.h>
 #include <linux/range.h>
+#include <linux/lmb.h>
 
 #include <asm/bug.h>
 #include <asm/io.h>
mm/page_alloc.c +8 -27
index faae23a..42bc09e 100644
@@ -3441,38 +3441,19 @@ int __init add_from_early_node_map(struct range *range, int az,
 void * __init __alloc_memory_core_early(int nid, u64 size, u64 align,
 					u64 goal, u64 limit)
 {
-	int i;
 	void *ptr;
 
-	/* need to go over early_node_map to find out good range for node */
-	for_each_active_range_index_in_nid(i, nid) {
-		u64 addr;
-		u64 ei_start, ei_last;
-
-		ei_last = early_node_map[i].end_pfn;
-		ei_last <<= PAGE_SHIFT;
-		ei_start = early_node_map[i].start_pfn;
-		ei_start <<= PAGE_SHIFT;
-		addr = find_early_area(ei_start, ei_last,
-					 goal, limit, size, align);
-
-		if (addr == -1ULL)
-			continue;
+	u64 addr;
 
-#if 0
-		printk(KERN_DEBUG "alloc (nid=%d %llx - %llx) (%llx - %llx) %llx %llx => %llx\n",
-				nid,
-				ei_start, ei_last, goal, limit, size,
-				align, addr);
-#endif
+	addr = find_memory_core_early(nid, size, align, goal, limit);
 
-		ptr = phys_to_virt(addr);
-		memset(ptr, 0, size);
-		reserve_early_without_check(addr, addr + size, "BOOTMEM");
-		return ptr;
-	}
+	if (addr == -1ULL)
+		return NULL;
 
-	return NULL;
+	ptr = phys_to_virt(addr);
+	memset(ptr, 0, size);
+	reserve_lmb(addr, addr + size, "BOOTMEM");
+	return ptr;
 }
 #endif
 
mm/sparse-vmemmap.c +0 -11
index 392b9bb..cb917d5 100644
@@ -219,18 +219,7 @@ void __init sparse_mem_maps_populate_node(struct page **map_map,
 
 	if (vmemmap_buf_start) {
 		/* need to free left buf */
-#ifdef CONFIG_NO_BOOTMEM
-		free_early(__pa(vmemmap_buf_start), __pa(vmemmap_buf_end));
-		if (vmemmap_buf_start < vmemmap_buf) {
-			char name[15];
-
-			snprintf(name, sizeof(name), "MEMMAP %d", nodeid);
-			reserve_early_without_check(__pa(vmemmap_buf_start),
-						    __pa(vmemmap_buf), name);
-		}
-#else
 		free_bootmem(__pa(vmemmap_buf), vmemmap_buf_end - vmemmap_buf);
-#endif
 		vmemmap_buf = NULL;
 		vmemmap_buf_end = NULL;
 	}
Signature · 1 line
1.6.4.2
author
#54 lore raw

PATCH 23/31 x86: Replace e820_/_early string with lmb_

In reply to Yinghai Lu

1.include linux/lmb.h directly. so later could reduce e820.h reference. 2 this patch is done by sed scripts mainly

---
 arch/x86/include/asm/efi.h      |    2 +-
 arch/x86/kernel/acpi/sleep.c    |    5 +++--
 arch/x86/kernel/apic/numaq_32.c |    3 ++-
 arch/x86/kernel/efi.c           |    5 +++--
 arch/x86/kernel/head32.c        |    4 ++--
 arch/x86/kernel/head64.c        |    4 ++--
 arch/x86/kernel/setup.c         |   25 ++++++++++++-------------
 arch/x86/kernel/trampoline.c    |    6 +++---
 arch/x86/mm/init.c              |    5 +++--
 arch/x86/mm/init_32.c           |   10 ++++++----
 arch/x86/mm/init_64.c           |    9 +++++----
 arch/x86/mm/k8topology_64.c     |    4 +++-
 arch/x86/mm/memtest.c           |    7 +++----
 arch/x86/mm/numa_32.c           |   17 +++++++++--------
 arch/x86/mm/numa_64.c           |   32 ++++++++++++++++----------------
 arch/x86/mm/srat_32.c           |    3 ++-
 arch/x86/mm/srat_64.c           |    9 +++++----
 arch/x86/xen/mmu.c              |    5 +++--
 arch/x86/xen/setup.c            |    3 ++-
 mm/bootmem.c                    |    4 ++--
 20 files changed, 87 insertions(+), 75 deletions(-)
arch/x86/include/asm/efi.h +1 -1
index 8406ed7..f756536 100644
@@ -90,7 +90,7 @@ extern void __iomem *efi_ioremap(unsigned long addr, unsigned long size,
 #endif /* CONFIG_X86_32 */
 
 extern int add_efi_memmap;
-extern void efi_reserve_early(void);
+extern void efi_reserve_lmb(void);
 extern void efi_call_phys_prelog(void);
 extern void efi_call_phys_epilog(void);
 
arch/x86/kernel/acpi/sleep.c +3 -2
index f996103..0cabfaa 100644
@@ -7,6 +7,7 @@
 
 #include <linux/acpi.h>
 #include <linux/bootmem.h>
+#include <linux/lmb.h>
 #include <linux/dmi.h>
 #include <linux/cpumask.h>
 #include <asm/segment.h>
@@ -133,7 +134,7 @@ void __init acpi_reserve_wakeup_memory(void)
 		return;
 	}
 
-	mem = find_e820_area(0, 1<<20, WAKEUP_SIZE, PAGE_SIZE);
+	mem = find_lmb_area(0, 1<<20, WAKEUP_SIZE, PAGE_SIZE);
 
 	if (mem == -1L) {
 		printk(KERN_ERR "ACPI: Cannot allocate lowmem, S3 disabled.\n");
@@ -141,7 +142,7 @@ void __init acpi_reserve_wakeup_memory(void)
 	}
 	acpi_realmode = (unsigned long) phys_to_virt(mem);
 	acpi_wakeup_address = mem;
-	reserve_early(mem, mem + WAKEUP_SIZE, "ACPI WAKEUP");
+	reserve_lmb(mem, mem + WAKEUP_SIZE, "ACPI WAKEUP");
 }
 
 
arch/x86/kernel/apic/numaq_32.c +2 -1
index 3e28401..c71e494 100644
@@ -26,6 +26,7 @@
 #include <linux/nodemask.h>
 #include <linux/topology.h>
 #include <linux/bootmem.h>
+#include <linux/lmb.h>
 #include <linux/threads.h>
 #include <linux/cpumask.h>
 #include <linux/kernel.h>
@@ -88,7 +89,7 @@ static inline void numaq_register_node(int node, struct sys_cfg_data *scd)
 	node_end_pfn[node] =
 		 MB_TO_PAGES(eq->hi_shrd_mem_start + eq->hi_shrd_mem_size);
 
-	e820_register_active_regions(node, node_start_pfn[node],
+	lmb_register_active_regions(node, node_start_pfn[node],
 						node_end_pfn[node]);
 
 	memory_present(node, node_start_pfn[node], node_end_pfn[node]);
arch/x86/kernel/efi.c +3 -2
index 299f03f..35038a7 100644
@@ -30,6 +30,7 @@
 #include <linux/init.h>
 #include <linux/efi.h>
 #include <linux/bootmem.h>
+#include <linux/lmb.h>
 #include <linux/spinlock.h>
 #include <linux/uaccess.h>
 #include <linux/time.h>
@@ -275,7 +276,7 @@ static void __init do_add_efi_memmap(void)
 	sanitize_e820_map();
 }
 
-void __init efi_reserve_early(void)
+void __init efi_reserve_lmb(void)
 {
 	unsigned long pmap;
 
@@ -290,7 +291,7 @@ void __init efi_reserve_early(void)
 		boot_params.efi_info.efi_memdesc_size;
 	memmap.desc_version = boot_params.efi_info.efi_memdesc_version;
 	memmap.desc_size = boot_params.efi_info.efi_memdesc_size;
-	reserve_early(pmap, pmap + memmap.nr_map * memmap.desc_size,
+	reserve_lmb(pmap, pmap + memmap.nr_map * memmap.desc_size,
 		      "EFI memmap");
 }
 
arch/x86/kernel/head32.c +2 -2
index cbab479..9d09d26 100644
@@ -41,7 +41,7 @@ void __init i386_start_kernel(void)
 	reserve_lmb(PAGE_SIZE, PAGE_SIZE + PAGE_SIZE, "EX TRAMPOLINE");
 #endif
 
-	reserve_early(__pa_symbol(&_text), __pa_symbol(&__bss_stop), "TEXT DATA BSS");
+	reserve_lmb(__pa_symbol(&_text), __pa_symbol(&__bss_stop), "TEXT DATA BSS");
 
 #ifdef CONFIG_BLK_DEV_INITRD
 	/* Reserve INITRD */
@@ -50,7 +50,7 @@ void __init i386_start_kernel(void)
 		u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 		u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
 		u64 ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
-		reserve_early(ramdisk_image, ramdisk_end, "RAMDISK");
+		reserve_lmb(ramdisk_image, ramdisk_end, "RAMDISK");
 	}
 #endif
 
arch/x86/kernel/head64.c +2 -2
index 89dd2de..d4442a8 100644
@@ -101,7 +101,7 @@ void __init x86_64_start_reservations(char *real_mode_data)
 
 	copy_bootdata(__va(real_mode_data));
 
-	reserve_early(__pa_symbol(&_text), __pa_symbol(&__bss_stop), "TEXT DATA BSS");
+	reserve_lmb(__pa_symbol(&_text), __pa_symbol(&__bss_stop), "TEXT DATA BSS");
 
 #ifdef CONFIG_BLK_DEV_INITRD
 	/* Reserve INITRD */
@@ -110,7 +110,7 @@ void __init x86_64_start_reservations(char *real_mode_data)
 		unsigned long ramdisk_image = boot_params.hdr.ramdisk_image;
 		unsigned long ramdisk_size  = boot_params.hdr.ramdisk_size;
 		unsigned long ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
-		reserve_early(ramdisk_image, ramdisk_end, "RAMDISK");
+		reserve_lmb(ramdisk_image, ramdisk_end, "RAMDISK");
 	}
 #endif
 
arch/x86/kernel/setup.c +12 -13
index a5c029f..3d43f12 100644
@@ -303,7 +303,7 @@ static inline void init_gbpages(void)
 static void __init reserve_brk(void)
 {
 	if (_brk_end > _brk_start)
-		reserve_early(__pa(_brk_start), __pa(_brk_end), "BRK");
+		reserve_lmb(__pa(_brk_start), __pa(_brk_end), "BRK");
 
 	/* Mark brk area as locked down and no longer taking any
 	   new allocations */
@@ -325,7 +325,7 @@ static void __init relocate_initrd(void)
 	char *p, *q;
 
 	/* We need to move the initrd down into lowmem */
-	ramdisk_here = find_e820_area(0, end_of_lowmem, area_size,
+	ramdisk_here = find_lmb_area(0, end_of_lowmem, area_size,
 					 PAGE_SIZE);
 
 	if (ramdisk_here == -1ULL)
@@ -334,8 +334,7 @@ static void __init relocate_initrd(void)
 
 	/* Note: this includes all the lowmem currently occupied by
 	   the initrd, we rely on that fact to keep the data intact. */
-	reserve_early(ramdisk_here, ramdisk_here + area_size,
-			 "NEW RAMDISK");
+	reserve_lmb(ramdisk_here, ramdisk_here + area_size, "NEW RAMDISK");
 	initrd_start = ramdisk_here + PAGE_OFFSET;
 	initrd_end   = initrd_start + ramdisk_size;
 	printk(KERN_INFO "Allocated new RAMDISK: %08llx - %08llx\n",
@@ -391,7 +390,7 @@ static void __init reserve_initrd(void)
 	initrd_start = 0;
 
 	if (ramdisk_size >= (end_of_lowmem>>1)) {
-		free_early(ramdisk_image, ramdisk_end);
+		free_lmb(ramdisk_image, ramdisk_end);
 		printk(KERN_ERR "initrd too large to handle, "
 		       "disabling initrd\n");
 		return;
@@ -414,7 +413,7 @@ static void __init reserve_initrd(void)
 
 	relocate_initrd();
 
-	free_early(ramdisk_image, ramdisk_end);
+	free_lmb(ramdisk_image, ramdisk_end);
 }
 #else
 static void __init reserve_initrd(void)
@@ -470,7 +469,7 @@ static void __init e820_reserve_setup_data(void)
 	e820_print_map("reserve setup_data");
 }
 
-static void __init reserve_early_setup_data(void)
+static void __init reserve_lmb_setup_data(void)
 {
 	struct setup_data *data;
 	u64 pa_data;
@@ -482,7 +481,7 @@ static void __init reserve_early_setup_data(void)
 	while (pa_data) {
 		data = early_memremap(pa_data, sizeof(*data));
 		sprintf(buf, "setup data %x", data->type);
-		reserve_early(pa_data, pa_data+sizeof(*data)+data->len, buf);
+		reserve_lmb(pa_data, pa_data+sizeof(*data)+data->len, buf);
 		pa_data = data->next;
 		early_iounmap(data, sizeof(*data));
 	}
@@ -520,7 +519,7 @@ static void __init reserve_crashkernel(void)
 	if (crash_base <= 0) {
 		const unsigned long long alignment = 16<<20;	/* 16M */
 
-		crash_base = find_e820_area(alignment, ULONG_MAX, crash_size,
+		crash_base = find_lmb_area(alignment, ULONG_MAX, crash_size,
 				 alignment);
 		if (crash_base == -1ULL) {
 			pr_info("crashkernel reservation failed - No suitable area found.\n");
@@ -529,14 +528,14 @@ static void __init reserve_crashkernel(void)
 	} else {
 		unsigned long long start;
 
-		start = find_e820_area(crash_base, ULONG_MAX, crash_size,
+		start = find_lmb_area(crash_base, ULONG_MAX, crash_size,
 				 1<<20);
 		if (start != crash_base) {
 			pr_info("crashkernel reservation failed - memory is in use.\n");
 			return;
 		}
 	}
-	reserve_early(crash_base, crash_base + crash_size, "CRASH KERNEL");
+	reserve_lmb(crash_base, crash_base + crash_size, "CRASH KERNEL");
 
 	printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
 			"for crashkernel (System RAM: %ldMB)\n",
@@ -756,7 +755,7 @@ void __init setup_arch(char **cmdline_p)
 #endif
 	 4)) {
 		efi_enabled = 1;
-		efi_reserve_early();
+		efi_reserve_lmb();
 	}
 #endif
 
@@ -816,7 +815,7 @@ void __init setup_arch(char **cmdline_p)
 	vmi_activate();
 
 	/* after early param, so could get panic from serial */
-	reserve_early_setup_data();
+	reserve_lmb_setup_data();
 
 	if (acpi_mps_check()) {
 #ifdef CONFIG_X86_LOCAL_APIC
arch/x86/kernel/trampoline.c +3 -3
index c652ef6..4a634f6 100644
@@ -1,7 +1,7 @@
 #include <linux/io.h>
+#include <linux/lmb.h>
 
 #include <asm/trampoline.h>
-#include <asm/e820.h>
 
 #if defined(CONFIG_X86_64) && defined(CONFIG_ACPI_SLEEP)
 #define __trampinit
@@ -19,12 +19,12 @@ void __init reserve_trampoline_memory(void)
 	unsigned long mem;
 
 	/* Has to be in very low memory so we can execute real-mode AP code. */
-	mem = find_e820_area(0, 1<<20, TRAMPOLINE_SIZE, PAGE_SIZE);
+	mem = find_lmb_area(0, 1<<20, TRAMPOLINE_SIZE, PAGE_SIZE);
 	if (mem == -1L)
 		panic("Cannot allocate trampoline\n");
 
 	trampoline_base = __va(mem);
-	reserve_early(mem, mem + TRAMPOLINE_SIZE, "TRAMPOLINE");
+	reserve_lmb(mem, mem + TRAMPOLINE_SIZE, "TRAMPOLINE");
 }
 
 /*
arch/x86/mm/init.c +3 -2
index 452ee5b..837aaa2 100644
@@ -1,6 +1,7 @@
 #include <linux/initrd.h>
 #include <linux/ioport.h>
 #include <linux/swap.h>
+#include <linux/lmb.h>
 
 #include <asm/cacheflush.h>
 #include <asm/e820.h>
@@ -74,7 +75,7 @@ static void __init find_early_table_space(unsigned long end, int use_pse,
 #else
 	start = 0x8000;
 #endif
-	e820_table_start = find_e820_area(start, max_pfn_mapped<<PAGE_SHIFT,
+	e820_table_start = find_lmb_area(start, max_pfn_mapped<<PAGE_SHIFT,
 					tables, PAGE_SIZE);
 	if (e820_table_start == -1UL)
 		panic("Cannot find space for the kernel page tables");
@@ -298,7 +299,7 @@ unsigned long __init_refok init_memory_mapping(unsigned long start,
 	__flush_tlb_all();
 
 	if (!after_bootmem && e820_table_end > e820_table_start)
-		reserve_early(e820_table_start << PAGE_SHIFT,
+		reserve_lmb(e820_table_start << PAGE_SHIFT,
 				 e820_table_end << PAGE_SHIFT, "PGTABLE");
 
 	if (!after_bootmem)
arch/x86/mm/init_32.c +6 -4
index 804bbe9..26b9ceb 100644
@@ -25,6 +25,7 @@
 #include <linux/pfn.h>
 #include <linux/poison.h>
 #include <linux/bootmem.h>
+#include <linux/lmb.h>
 #include <linux/slab.h>
 #include <linux/proc_fs.h>
 #include <linux/memory_hotplug.h>
@@ -712,14 +713,14 @@ void __init initmem_init(unsigned long start_pfn, unsigned long end_pfn,
 	highstart_pfn = highend_pfn = max_pfn;
 	if (max_pfn > max_low_pfn)
 		highstart_pfn = max_low_pfn;
-	e820_register_active_regions(0, 0, highend_pfn);
+	lmb_register_active_regions(0, 0, highend_pfn);
 	sparse_memory_present_with_active_regions(0);
 	printk(KERN_NOTICE "%ldMB HIGHMEM available.\n",
 		pages_to_mb(highend_pfn - highstart_pfn));
 	num_physpages = highend_pfn;
 	high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1;
 #else
-	e820_register_active_regions(0, 0, max_low_pfn);
+	lmb_register_active_regions(0, 0, max_low_pfn);
 	sparse_memory_present_with_active_regions(0);
 	num_physpages = max_low_pfn;
 	high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1;
@@ -781,11 +782,11 @@ void __init setup_bootmem_allocator(void)
 	 * Initialize the boot-time allocator (with low memory only):
 	 */
 	bootmap_size = bootmem_bootmap_pages(max_low_pfn)<<PAGE_SHIFT;
-	bootmap = find_e820_area(0, max_pfn_mapped<<PAGE_SHIFT, bootmap_size,
+	bootmap = find_lmb_area(0, max_pfn_mapped<<PAGE_SHIFT, bootmap_size,
 				 PAGE_SIZE);
 	if (bootmap == -1L)
 		panic("Cannot find bootmem map of size %ld\n", bootmap_size);
-	reserve_early(bootmap, bootmap + bootmap_size, "BOOTMAP");
+	reserve_lmb(bootmap, bootmap + bootmap_size, "BOOTMAP");
 #endif
 
 	printk(KERN_INFO "  mapped low ram: 0 - %08lx\n",
@@ -1069,3 +1070,4 @@ void mark_rodata_ro(void)
 #endif
 }
 #endif
+
arch/x86/mm/init_64.c +5 -4
index 5ba6b0e..b86492e 100644
@@ -21,6 +21,7 @@
 #include <linux/initrd.h>
 #include <linux/pagemap.h>
 #include <linux/bootmem.h>
+#include <linux/lmb.h>
 #include <linux/proc_fs.h>
 #include <linux/pci.h>
 #include <linux/pfn.h>
@@ -576,18 +577,18 @@ void __init initmem_init(unsigned long start_pfn, unsigned long end_pfn,
 	unsigned long bootmap_size, bootmap;
 
 	bootmap_size = bootmem_bootmap_pages(end_pfn)<<PAGE_SHIFT;
-	bootmap = find_e820_area(0, end_pfn<<PAGE_SHIFT, bootmap_size,
+	bootmap = find_lmb_area(0, end_pfn<<PAGE_SHIFT, bootmap_size,
 				 PAGE_SIZE);
 	if (bootmap == -1L)
 		panic("Cannot find bootmem map of size %ld\n", bootmap_size);
-	reserve_early(bootmap, bootmap + bootmap_size, "BOOTMAP");
+	reserve_lmb(bootmap, bootmap + bootmap_size, "BOOTMAP");
 	/* don't touch min_low_pfn */
 	bootmap_size = init_bootmem_node(NODE_DATA(0), bootmap >> PAGE_SHIFT,
 					 0, end_pfn);
-	e820_register_active_regions(0, start_pfn, end_pfn);
+	lmb_register_active_regions(0, start_pfn, end_pfn);
 	free_bootmem_with_active_regions(0, end_pfn);
 #else
-	e820_register_active_regions(0, start_pfn, end_pfn);
+	lmb_register_active_regions(0, start_pfn, end_pfn);
 #endif
 }
 #endif
arch/x86/mm/k8topology_64.c +3 -1
index 970ed57..d7d031b 100644
@@ -11,6 +11,8 @@
 #include <linux/string.h>
 #include <linux/module.h>
 #include <linux/nodemask.h>
+#include <linux/lmb.h>
+
 #include <asm/io.h>
 #include <linux/pci_ids.h>
 #include <linux/acpi.h>
@@ -222,7 +224,7 @@ int __init k8_scan_nodes(void)
 	for_each_node_mask(i, node_possible_map) {
 		int j;
 
-		e820_register_active_regions(i,
+		lmb_register_active_regions(i,
 				nodes[i].start >> PAGE_SHIFT,
 				nodes[i].end >> PAGE_SHIFT);
 		for (j = apicid_base; j < cores + apicid_base; j++)
arch/x86/mm/memtest.c +3 -4
index 18d244f..f889776 100644
@@ -6,8 +6,7 @@
 #include <linux/smp.h>
 #include <linux/init.h>
 #include <linux/pfn.h>
-
-#include <asm/e820.h>
+#include <linux/lmb.h>
 
 static u64 patterns[] __initdata = {
 	0,
@@ -35,7 +34,7 @@ static void __init reserve_bad_mem(u64 pattern, u64 start_bad, u64 end_bad)
 	       (unsigned long long) pattern,
 	       (unsigned long long) start_bad,
 	       (unsigned long long) end_bad);
-	reserve_early(start_bad, end_bad, "BAD RAM");
+	reserve_lmb(start_bad, end_bad, "BAD RAM");
 }
 
 static void __init memtest(u64 pattern, u64 start_phys, u64 size)
@@ -74,7 +73,7 @@ static void __init do_one_pass(u64 pattern, u64 start, u64 end)
 	u64 size = 0;
 
 	while (start < end) {
-		start = find_e820_area_size(start, &size, 1);
+		start = find_lmb_area_size(start, &size, 1);
 
 		/* done ? */
 		if (start >= end)
arch/x86/mm/numa_32.c +9 -8
index 809baaa..3434a93 100644
@@ -24,6 +24,7 @@
 
 #include <linux/mm.h>
 #include <linux/bootmem.h>
+#include <linux/lmb.h>
 #include <linux/mmzone.h>
 #include <linux/highmem.h>
 #include <linux/initrd.h>
@@ -120,7 +121,7 @@ int __init get_memcfg_numa_flat(void)
 
 	node_start_pfn[0] = 0;
 	node_end_pfn[0] = max_pfn;
-	e820_register_active_regions(0, 0, max_pfn);
+	lmb_register_active_regions(0, 0, max_pfn);
 	memory_present(0, 0, max_pfn);
 	node_remap_size[0] = node_memmap_size_bytes(0, 0, max_pfn);
 
@@ -161,14 +162,14 @@ static void __init allocate_pgdat(int nid)
 		NODE_DATA(nid) = (pg_data_t *)node_remap_start_vaddr[nid];
 	else {
 		unsigned long pgdat_phys;
-		pgdat_phys = find_e820_area(min_low_pfn<<PAGE_SHIFT,
+		pgdat_phys = find_lmb_area(min_low_pfn<<PAGE_SHIFT,
 				 max_pfn_mapped<<PAGE_SHIFT,
 				 sizeof(pg_data_t),
 				 PAGE_SIZE);
 		NODE_DATA(nid) = (pg_data_t *)(pfn_to_kaddr(pgdat_phys>>PAGE_SHIFT));
 		memset(buf, 0, sizeof(buf));
 		sprintf(buf, "NODE_DATA %d",  nid);
-		reserve_early(pgdat_phys, pgdat_phys + sizeof(pg_data_t), buf);
+		reserve_lmb(pgdat_phys, pgdat_phys + sizeof(pg_data_t), buf);
 	}
 	printk(KERN_DEBUG "allocate_pgdat: node %d NODE_DATA %08lx\n",
 		nid, (unsigned long)NODE_DATA(nid));
@@ -291,7 +292,7 @@ static __init unsigned long calculate_numa_remap_pages(void)
 						 PTRS_PER_PTE);
 		node_kva_target <<= PAGE_SHIFT;
 		do {
-			node_kva_final = find_e820_area(node_kva_target,
+			node_kva_final = find_lmb_area(node_kva_target,
 					((u64)node_end_pfn[nid])<<PAGE_SHIFT,
 						((u64)size)<<PAGE_SHIFT,
 						LARGE_PAGE_BYTES);
@@ -318,9 +319,9 @@ static __init unsigned long calculate_numa_remap_pages(void)
 		 *  but we could have some hole in high memory, and it will only
 		 *  check page_is_ram(pfn) && !page_is_reserved_early(pfn) to decide
 		 *  to use it as free.
-		 *  So reserve_early here, hope we don't run out of that array
+		 *  So reserve_lmb here, hope we don't run out of that array
 		 */
-		reserve_early(node_kva_final,
+		reserve_lmb(node_kva_final,
 			      node_kva_final+(((u64)size)<<PAGE_SHIFT),
 			      "KVA RAM");
 
@@ -367,7 +368,7 @@ void __init initmem_init(unsigned long start_pfn, unsigned long end_pfn,
 
 	kva_target_pfn = round_down(max_low_pfn - kva_pages, PTRS_PER_PTE);
 	do {
-		kva_start_pfn = find_e820_area(kva_target_pfn<<PAGE_SHIFT,
+		kva_start_pfn = find_lmb_area(kva_target_pfn<<PAGE_SHIFT,
 					max_low_pfn<<PAGE_SHIFT,
 					kva_pages<<PAGE_SHIFT,
 					PTRS_PER_PTE<<PAGE_SHIFT) >> PAGE_SHIFT;
@@ -382,7 +383,7 @@ void __init initmem_init(unsigned long start_pfn, unsigned long end_pfn,
 	printk(KERN_INFO "max_pfn = %lx\n", max_pfn);
 
 	/* avoid clash with initrd */
-	reserve_early(kva_start_pfn<<PAGE_SHIFT,
+	reserve_lmb(kva_start_pfn<<PAGE_SHIFT,
 		      (kva_start_pfn + kva_pages)<<PAGE_SHIFT,
 		     "KVA PG");
 #ifdef CONFIG_HIGHMEM
arch/x86/mm/numa_64.c +16 -16
index fc9a403..80b568d 100644
@@ -90,7 +90,7 @@ static int __init allocate_cachealigned_memnodemap(void)
 
 	addr = 0x8000;
 	nodemap_size = roundup(sizeof(s16) * memnodemapsize, L1_CACHE_BYTES);
-	nodemap_addr = find_e820_area(addr, max_pfn<<PAGE_SHIFT,
+	nodemap_addr = find_lmb_area(addr, max_pfn<<PAGE_SHIFT,
 				      nodemap_size, L1_CACHE_BYTES);
 	if (nodemap_addr == -1UL) {
 		printk(KERN_ERR
@@ -99,7 +99,7 @@ static int __init allocate_cachealigned_memnodemap(void)
 		return -1;
 	}
 	memnodemap = phys_to_virt(nodemap_addr);
-	reserve_early(nodemap_addr, nodemap_addr + nodemap_size, "MEMNODEMAP");
+	reserve_lmb(nodemap_addr, nodemap_addr + nodemap_size, "MEMNODEMAP");
 
 	printk(KERN_DEBUG "NUMA: Allocated memnodemap from %lx - %lx\n",
 	       nodemap_addr, nodemap_addr + nodemap_size);
@@ -230,7 +230,7 @@ setup_node_bootmem(int nodeid, unsigned long start, unsigned long end)
 	if (node_data[nodeid] == NULL)
 		return;
 	nodedata_phys = __pa(node_data[nodeid]);
-	reserve_early(nodedata_phys, nodedata_phys + pgdat_size, "NODE_DATA");
+	reserve_lmb(nodedata_phys, nodedata_phys + pgdat_size, "NODE_DATA");
 	printk(KERN_INFO "  NODE_DATA [%016lx - %016lx]\n", nodedata_phys,
 		nodedata_phys + pgdat_size - 1);
 	nid = phys_to_nid(nodedata_phys);
@@ -249,7 +249,7 @@ setup_node_bootmem(int nodeid, unsigned long start, unsigned long end)
 	 * Find a place for the bootmem map
 	 * nodedata_phys could be on other nodes by alloc_bootmem,
 	 * so need to sure bootmap_start not to be small, otherwise
-	 * early_node_mem will get that with find_e820_area instead
+	 * early_node_mem will get that with find_lmb_area instead
 	 * of alloc_bootmem, that could clash with reserved range
 	 */
 	bootmap_pages = bootmem_bootmap_pages(last_pfn - start_pfn);
@@ -261,12 +261,12 @@ setup_node_bootmem(int nodeid, unsigned long start, unsigned long end)
 	bootmap = early_node_mem(nodeid, bootmap_start, end,
 				 bootmap_pages<<PAGE_SHIFT, PAGE_SIZE);
 	if (bootmap == NULL)  {
-		free_early(nodedata_phys, nodedata_phys + pgdat_size);
+		free_lmb(nodedata_phys, nodedata_phys + pgdat_size);
 		node_data[nodeid] = NULL;
 		return;
 	}
 	bootmap_start = __pa(bootmap);
-	reserve_early(bootmap_start, bootmap_start+(bootmap_pages<<PAGE_SHIFT),
+	reserve_lmb(bootmap_start, bootmap_start+(bootmap_pages<<PAGE_SHIFT),
 			"BOOTMAP");
 
 	bootmap_size = init_bootmem_node(NODE_DATA(nodeid),
@@ -420,7 +420,7 @@ static int __init split_nodes_interleave(u64 addr, u64 max_addr,
 		nr_nodes = MAX_NUMNODES;
 	}
 
-	size = (max_addr - addr - e820_hole_size(addr, max_addr)) / nr_nodes;
+	size = (max_addr - addr - lmb_hole_size(addr, max_addr)) / nr_nodes;
 	/*
 	 * Calculate the number of big nodes that can be allocated as a result
 	 * of consolidating the remainder.
@@ -456,7 +456,7 @@ static int __init split_nodes_interleave(u64 addr, u64 max_addr,
 			 * non-reserved memory is less than the per-node size.
 			 */
 			while (end - physnodes[i].start -
-				e820_hole_size(physnodes[i].start, end) < size) {
+				lmb_hole_size(physnodes[i].start, end) < size) {
 				end += FAKE_NODE_MIN_SIZE;
 				if (end > physnodes[i].end) {
 					end = physnodes[i].end;
@@ -470,7 +470,7 @@ static int __init split_nodes_interleave(u64 addr, u64 max_addr,
 			 * this one must extend to the boundary.
 			 */
 			if (end < dma32_end && dma32_end - end -
-			    e820_hole_size(end, dma32_end) < FAKE_NODE_MIN_SIZE)
+			    lmb_hole_size(end, dma32_end) < FAKE_NODE_MIN_SIZE)
 				end = dma32_end;
 
 			/*
@@ -479,7 +479,7 @@ static int __init split_nodes_interleave(u64 addr, u64 max_addr,
 			 * physical node.
 			 */
 			if (physnodes[i].end - end -
-			    e820_hole_size(end, physnodes[i].end) < size)
+			    lmb_hole_size(end, physnodes[i].end) < size)
 				end = physnodes[i].end;
 
 			/*
@@ -507,7 +507,7 @@ static u64 __init find_end_of_node(u64 start, u64 max_addr, u64 size)
 {
 	u64 end = start + size;
 
-	while (end - start - e820_hole_size(start, end) < size) {
+	while (end - start - lmb_hole_size(start, end) < size) {
 		end += FAKE_NODE_MIN_SIZE;
 		if (end > max_addr) {
 			end = max_addr;
@@ -536,7 +536,7 @@ static int __init split_nodes_size_interleave(u64 addr, u64 max_addr, u64 size)
 	 * creates a uniform distribution of node sizes across the entire
 	 * machine (but not necessarily over physical nodes).
 	 */
-	min_size = (max_addr - addr - e820_hole_size(addr, max_addr)) /
+	min_size = (max_addr - addr - lmb_hole_size(addr, max_addr)) /
 						MAX_NUMNODES;
 	min_size = max(min_size, FAKE_NODE_MIN_SIZE);
 	if ((min_size & FAKE_NODE_MIN_HASH_MASK) < min_size)
@@ -569,7 +569,7 @@ static int __init split_nodes_size_interleave(u64 addr, u64 max_addr, u64 size)
 			 * this one must extend to the boundary.
 			 */
 			if (end < dma32_end && dma32_end - end -
-			    e820_hole_size(end, dma32_end) < FAKE_NODE_MIN_SIZE)
+			    lmb_hole_size(end, dma32_end) < FAKE_NODE_MIN_SIZE)
 				end = dma32_end;
 
 			/*
@@ -578,7 +578,7 @@ static int __init split_nodes_size_interleave(u64 addr, u64 max_addr, u64 size)
 			 * physical node.
 			 */
 			if (physnodes[i].end - end -
-			    e820_hole_size(end, physnodes[i].end) < size)
+			    lmb_hole_size(end, physnodes[i].end) < size)
 				end = physnodes[i].end;
 
 			/*
@@ -642,7 +642,7 @@ static int __init numa_emulation(unsigned long start_pfn,
 	 */
 	remove_all_active_ranges();
 	for_each_node_mask(i, node_possible_map) {
-		e820_register_active_regions(i, nodes[i].start >> PAGE_SHIFT,
+		lmb_register_active_regions(i, nodes[i].start >> PAGE_SHIFT,
 						nodes[i].end >> PAGE_SHIFT);
 		setup_node_bootmem(i, nodes[i].start, nodes[i].end);
 	}
@@ -695,7 +695,7 @@ void __init initmem_init(unsigned long start_pfn, unsigned long last_pfn,
 	node_set(0, node_possible_map);
 	for (i = 0; i < nr_cpu_ids; i++)
 		numa_set_node(i, 0);
-	e820_register_active_regions(0, start_pfn, last_pfn);
+	lmb_register_active_regions(0, start_pfn, last_pfn);
 	setup_node_bootmem(0, start_pfn << PAGE_SHIFT, last_pfn << PAGE_SHIFT);
 }
 
arch/x86/mm/srat_32.c +2 -1
index 9324f13..68dd606 100644
@@ -25,6 +25,7 @@
  */
 #include <linux/mm.h>
 #include <linux/bootmem.h>
+#include <linux/lmb.h>
 #include <linux/mmzone.h>
 #include <linux/acpi.h>
 #include <linux/nodemask.h>
@@ -264,7 +265,7 @@ int __init get_memcfg_from_srat(void)
 		if (node_read_chunk(chunk->nid, chunk))
 			continue;
 
-		e820_register_active_regions(chunk->nid, chunk->start_pfn,
+		lmb_register_active_regions(chunk->nid, chunk->start_pfn,
 					     min(chunk->end_pfn, max_pfn));
 	}
 	/* for out of order entries in SRAT */
arch/x86/mm/srat_64.c +5 -4
index 28c6876..9eca2bc 100644
@@ -16,6 +16,7 @@
 #include <linux/module.h>
 #include <linux/topology.h>
 #include <linux/bootmem.h>
+#include <linux/lmb.h>
 #include <linux/mm.h>
 #include <asm/proto.h>
 #include <asm/numa.h>
@@ -98,7 +99,7 @@ void __init acpi_numa_slit_init(struct acpi_table_slit *slit)
 	unsigned long phys;
 
 	length = slit->header.length;
-	phys = find_e820_area(0, max_pfn_mapped<<PAGE_SHIFT, length,
+	phys = find_lmb_area(0, max_pfn_mapped<<PAGE_SHIFT, length,
 		 PAGE_SIZE);
 
 	if (phys == -1L)
@@ -106,7 +107,7 @@ void __init acpi_numa_slit_init(struct acpi_table_slit *slit)
 
 	acpi_slit = __va(phys);
 	memcpy(acpi_slit, slit, length);
-	reserve_early(phys, phys + length, "ACPI SLIT");
+	reserve_lmb(phys, phys + length, "ACPI SLIT");
 }
 
 /* Callback for Proximity Domain -> x2APIC mapping */
@@ -324,7 +325,7 @@ static int __init nodes_cover_memory(const struct bootnode *nodes)
 			pxmram = 0;
 	}
 
-	e820ram = max_pfn - (e820_hole_size(0, max_pfn<<PAGE_SHIFT)>>PAGE_SHIFT);
+	e820ram = max_pfn - (lmb_hole_size(0, max_pfn<<PAGE_SHIFT)>>PAGE_SHIFT);
 	/* We seem to lose 3 pages somewhere. Allow 1M of slack. */
 	if ((long)(e820ram - pxmram) >= (1<<(20 - PAGE_SHIFT))) {
 		printk(KERN_ERR
@@ -373,7 +374,7 @@ int __init acpi_scan_nodes(unsigned long start, unsigned long end)
 	}
 
 	for_each_node_mask(i, nodes_parsed)
-		e820_register_active_regions(i, nodes[i].start >> PAGE_SHIFT,
+		lmb_register_active_regions(i, nodes[i].start >> PAGE_SHIFT,
 						nodes[i].end >> PAGE_SHIFT);
 	/* for out of order entries in SRAT */
 	sort_node_map();
arch/x86/xen/mmu.c +3 -2
index f9eb7de..e61658e 100644
@@ -43,6 +43,7 @@
 #include <linux/debugfs.h>
 #include <linux/bug.h>
 #include <linux/module.h>
+#include <linux/lmb.h>
 
 #include <asm/pgtable.h>
 #include <asm/tlbflush.h>
@@ -1734,7 +1735,7 @@ __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd,
 	__xen_write_cr3(true, __pa(pgd));
 	xen_mc_issue(PARAVIRT_LAZY_CPU);
 
-	reserve_early(__pa(xen_start_info->pt_base),
+	reserve_lmb(__pa(xen_start_info->pt_base),
 		      __pa(xen_start_info->pt_base +
 			   xen_start_info->nr_pt_frames * PAGE_SIZE),
 		      "XEN PAGETABLES");
@@ -1772,7 +1773,7 @@ __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd,
 
 	pin_pagetable_pfn(MMUEXT_PIN_L3_TABLE, PFN_DOWN(__pa(swapper_pg_dir)));
 
-	reserve_early(__pa(xen_start_info->pt_base),
+	reserve_lmb(__pa(xen_start_info->pt_base),
 		      __pa(xen_start_info->pt_base +
 			   xen_start_info->nr_pt_frames * PAGE_SIZE),
 		      "XEN PAGETABLES");
arch/x86/xen/setup.c +2 -1
index 3f2c411..e8b3ca1 100644
@@ -8,6 +8,7 @@
 #include <linux/sched.h>
 #include <linux/mm.h>
 #include <linux/pm.h>
+#include <linux/lmb.h>
 
 #include <asm/elf.h>
 #include <asm/vdso.h>
@@ -59,7 +60,7 @@ char * __init xen_memory_setup(void)
 	 *  - xen_start_info
 	 * See comment above "struct start_info" in <xen/interface/xen.h>
 	 */
-	reserve_early(__pa(xen_start_info->mfn_list),
+	reserve_lmb(__pa(xen_start_info->mfn_list),
 		      __pa(xen_start_info->pt_base),
 			"XEN START INFO");
 
mm/bootmem.c +2 -2
index 81f9350..7e10d41 100644
@@ -421,7 +421,7 @@ void __init free_bootmem_node(pg_data_t *pgdat, unsigned long physaddr,
 			      unsigned long size)
 {
 #ifdef CONFIG_NO_BOOTMEM
-	free_early(physaddr, physaddr + size);
+	free_lmb(physaddr, physaddr + size);
 #else
 	unsigned long start, end;
 
@@ -446,7 +446,7 @@ void __init free_bootmem_node(pg_data_t *pgdat, unsigned long physaddr,
 void __init free_bootmem(unsigned long addr, unsigned long size)
 {
 #ifdef CONFIG_NO_BOOTMEM
-	free_early(addr, addr + size);
+	free_lmb(addr, addr + size);
 #else
 	unsigned long start, end;
 
Signature · 1 line
1.6.4.2
author
#55 lore raw

PATCH 24/31 x86: Remove not used early_res code

In reply to Yinghai Lu

and some functions in e820.c that are not used anymore

---
 arch/x86/include/asm/e820.h |   14 -
 arch/x86/kernel/e820.c      |   41 ---
 include/linux/early_res.h   |   23 --
 kernel/early_res.c          |  584 -------------------------------------------
 4 files changed, 0 insertions(+), 662 deletions(-)
 delete mode 100644 include/linux/early_res.h
 delete mode 100644 kernel/early_res.c
arch/x86/include/asm/e820.h +0 -14
index 396c849..de6cd06 100644
@@ -111,32 +111,18 @@ static inline void early_memtest(unsigned long start, unsigned long end)
 }
 #endif
 
-extern unsigned long end_user_pfn;
-
-extern u64 find_e820_area(u64 start, u64 end, u64 size, u64 align);
-extern u64 find_e820_area_size(u64 start, u64 *sizep, u64 align);
-extern u64 early_reserve_e820(u64 startt, u64 sizet, u64 align);
-
 extern unsigned long e820_end_of_ram_pfn(void);
 extern unsigned long e820_end_of_low_ram_pfn(void);
-extern void e820_register_active_regions(int nid, unsigned long start_pfn,
-					 unsigned long end_pfn);
-extern u64 e820_hole_size(u64 start, u64 end);
-
 extern u64 early_reserve_e820(u64 startt, u64 sizet, u64 align);
 
 void init_lmb_memory(void);
 void fill_lmb_memory(void);
-
 extern void finish_e820_parsing(void);
 extern void e820_reserve_resources(void);
 extern void e820_reserve_resources_late(void);
 extern void setup_memory_map(void);
 extern char *default_machine_specific_memory_setup(void);
 
-void reserve_early(u64 start, u64 end, char *name);
-void free_early(u64 start, u64 end);
-
 /*
  * Returns true iff the specified range [s,e) is completely contained inside
  * the ISA region.
arch/x86/kernel/e820.c +0 -41

59 lines not on this page · raw message

include/linux/early_res.h deleted +0 -23
deleted file mode 100644
index 29c09f5..0000000
@@ -1,23 +0,0 @@
-#ifndef _LINUX_EARLY_RES_H
-#define _LINUX_EARLY_RES_H
-#ifdef __KERNEL__
-
-extern void reserve_early(u64 start, u64 end, char *name);
-extern void reserve_early_overlap_ok(u64 start, u64 end, char *name);
-extern void free_early(u64 start, u64 end);
-void free_early_partial(u64 start, u64 end);
-extern void early_res_to_bootmem(u64 start, u64 end);
-
-void reserve_early_without_check(u64 start, u64 end, char *name);
-u64 find_early_area(u64 ei_start, u64 ei_last, u64 start, u64 end,
-			 u64 size, u64 align);
-u64 find_early_area_size(u64 ei_start, u64 ei_last, u64 start,
-			 u64 *sizep, u64 align);
-u64 find_fw_memmap_area(u64 start, u64 end, u64 size, u64 align);
-u64 get_max_mapped(void);
-#include <linux/range.h>
-int get_free_all_memory_range(struct range **rangep, int nodeid);
-
-#endif /* __KERNEL__ */
-
-#endif /* _LINUX_EARLY_RES_H */
kernel/early_res.c deleted +0 -584

584 lines not on this page · raw message

Signature · 1 line
1.6.4.2
author
#56 lore raw

PATCH 25/31 x86, lmb: Use lmb_memory_size()/lmb_free_memory_size() to get correct dma_reserve

In reply to Yinghai Lu

lmb_memory_size() will return memory size in lmb.memory.region. lmb_free_memory_size() will return free memory size in lmb.memory.region.

So We can get exact reseved size.

Set the size right after initmem_init(), because later bootmem API will get area from 16M. (except some fallback).

Later after we remove the bootmem, We could call that just before paging_init().

---
 arch/x86/include/asm/e820.h |    2 ++
 arch/x86/kernel/e820.c      |   17 +++++++++++++++++
 arch/x86/kernel/setup.c     |    1 +
 arch/x86/mm/init_64.c       |    7 -------
 4 files changed, 20 insertions(+), 7 deletions(-)
arch/x86/include/asm/e820.h +2 -0
index de6cd06..334281f 100644
@@ -117,6 +117,8 @@ extern u64 early_reserve_e820(u64 startt, u64 sizet, u64 align);
 
 void init_lmb_memory(void);
 void fill_lmb_memory(void);
+void find_lmb_dma_reserve(void);
+
 extern void finish_e820_parsing(void);
 extern void e820_reserve_resources(void);
 extern void e820_reserve_resources_late(void);
arch/x86/kernel/e820.c +17 -0
index 47eb188..9c4b9da 100644
@@ -1100,3 +1100,20 @@ void __init fill_lmb_memory(void)
 	lmb_analyze();
 	lmb_dump_all();
 }
+
+void __init find_lmb_dma_reserve(void)
+{
+#ifdef CONFIG_X86_64
+	u64 free_size_pfn;
+	u64 mem_size_pfn;
+	/*
+	 * need to find out used area below MAX_DMA_PFN
+	 * need to use lmb to get free size in [0, MAX_DMA_PFN]
+	 * at first, and assume boot_mem will not take below MAX_DMA_PFN
+	 */
+	mem_size_pfn = lmb_memory_size(0, MAX_DMA_PFN << PAGE_SHIFT) >> PAGE_SHIFT;
+	free_size_pfn = lmb_free_memory_size(0, MAX_DMA_PFN << PAGE_SHIFT) >> PAGE_SHIFT;
+	set_dma_reserve(mem_size_pfn - free_size_pfn);
+#endif
+}
+
arch/x86/kernel/setup.c +1 -0
index 3d43f12..b00ccc4 100644
@@ -979,6 +979,7 @@ void __init setup_arch(char **cmdline_p)
 #endif
 
 	initmem_init(0, max_pfn, acpi, k8);
+	find_lmb_dma_reserve();
 #ifndef CONFIG_NO_BOOTMEM
 	lmb_to_bootmem(0, max_low_pfn<<PAGE_SHIFT);
 #endif
arch/x86/mm/init_64.c +0 -7
index b86492e..f1af41c 100644
@@ -52,8 +52,6 @@
 #include <asm/init.h>
 #include <linux/bootmem.h>
 
-static unsigned long dma_reserve __initdata;
-
 static int __init parse_direct_gbpages_off(char *arg)
 {
 	direct_gbpages = 0;
@@ -820,11 +818,6 @@ int __init reserve_bootmem_generic(unsigned long phys, unsigned long len,
 
 	reserve_bootmem(phys, len, flags);
 
-	if (phys+len <= MAX_DMA_PFN*PAGE_SIZE) {
-		dma_reserve += len / PAGE_SIZE;
-		set_dma_reserve(dma_reserve);
-	}
-
 	return 0;
 }
 #endif
Signature · 1 line
1.6.4.2
author
#57 lore raw

PATCH 26/31 x86: Align e820 ram range to page

In reply to Yinghai Lu

to workaround wrong BIOS memory map.

---
 arch/x86/kernel/e820.c |   44 ++++++++++++++++++++++++++++++++++++++++++++
 1 files changed, 44 insertions(+), 0 deletions(-)
arch/x86/kernel/e820.c +44 -0
index 9c4b9da..6fdd9e9 100644
@@ -903,6 +903,47 @@ static int __init parse_memmap_opt(char *p)
 }
 early_param("memmap", parse_memmap_opt);
 
+static void __init e820_align_ram_page(void)
+{
+	int i;
+	bool changed = false;;
+
+	for (i = 0; i < e820.nr_map; i++) {
+		struct e820entry *entry = &e820.map[i];
+		u64 start, end;
+		u64 start_aligned, end_aligned;
+
+		if (entry->type != E820_RAM)
+			continue;
+
+		start = entry->addr;
+		end = start + entry->size;
+
+		start_aligned = round_up(start, PAGE_SIZE);
+		end_aligned = round_down(end, PAGE_SIZE);
+
+		if (end_aligned <= start_aligned) {
+			e820_update_range(start, end - start, E820_RAM, E820_RESERVED);
+			changed = true;
+			continue;
+		}
+		if (start < start_aligned) {
+			e820_update_range(start, start_aligned - start, E820_RAM, E820_RESERVED);
+			changed = true;
+		}
+		if (end_aligned < end) {
+			e820_update_range(end_aligned, end - end_aligned, E820_RAM, E820_RESERVED);
+			changed = true;
+		}
+	}
+
+	if (changed) {
+		sanitize_e820_map();
+		printk(KERN_INFO "aligned physical RAM map:\n");
+		e820_print_map("aligned");
+	}
+}
+
 void __init finish_e820_parsing(void)
 {
 	if (userdef) {
@@ -915,6 +956,9 @@ void __init finish_e820_parsing(void)
 		printk(KERN_INFO "user-defined physical RAM map:\n");
 		e820_print_map("user");
 	}
+
+	/* In case, We have RAM entres that are not PAGE aligned */
+	e820_align_ram_page();
 }
 
 static inline const char *e820_type_to_string(int e820_type)
Signature · 1 line
1.6.4.2
author
#58 lore raw

PATCH 27/31 x86: Use wake_system_ram_range instead of e820_any_mapped in agp path

In reply to Yinghai Lu

Move apterture_valid back to .c

and early path still use e820_any_mapped()

so later we can make e820_any_mapped() to _init

---
 arch/x86/include/asm/gart.h   |   22 ----------------------
 arch/x86/kernel/aperture_64.c |   22 ++++++++++++++++++++++
 drivers/char/agp/amd64-agp.c  |   39 ++++++++++++++++++++++++++++++++++++++-
 3 files changed, 60 insertions(+), 23 deletions(-)
arch/x86/include/asm/gart.h +0 -22
index 4ac5b0f..2b63a91 100644
@@ -74,26 +74,4 @@ static inline void enable_gart_translation(struct pci_dev *dev, u64 addr)
         pci_write_config_dword(dev, AMD64_GARTAPERTURECTL, ctl);
 }
 
-static inline int aperture_valid(u64 aper_base, u32 aper_size, u32 min_size)
-{
-	if (!aper_base)
-		return 0;
-
-	if (aper_base + aper_size > 0x100000000ULL) {
-		printk(KERN_INFO "Aperture beyond 4GB. Ignoring.\n");
-		return 0;
-	}
-	if (e820_any_mapped(aper_base, aper_base + aper_size, E820_RAM)) {
-		printk(KERN_INFO "Aperture pointing to e820 RAM. Ignoring.\n");
-		return 0;
-	}
-	if (aper_size < min_size) {
-		printk(KERN_INFO "Aperture too small (%d MB) than (%d MB)\n",
-				 aper_size>>20, min_size>>20);
-		return 0;
-	}
-
-	return 1;
-}
-
 #endif /* _ASM_X86_GART_H */
arch/x86/kernel/aperture_64.c +22 -0
index 3704997..f6e6270 100644
@@ -145,6 +145,28 @@ static u32 __init find_cap(int bus, int slot, int func, int cap)
 	return 0;
 }
 
+static int __init aperture_valid(u64 aper_base, u32 aper_size, u32 min_size)
+{
+	if (!aper_base)
+		return 0;
+
+	if (aper_base + aper_size > 0x100000000ULL) {
+		printk(KERN_INFO "Aperture beyond 4GB. Ignoring.\n");
+		return 0;
+	}
+	if (e820_any_mapped(aper_base, aper_base + aper_size, E820_RAM)) {
+		printk(KERN_INFO "Aperture pointing to e820 RAM. Ignoring.\n");
+		return 0;
+	}
+	if (aper_size < min_size) {
+		printk(KERN_INFO "Aperture too small (%d MB) than (%d MB)\n",
+				 aper_size>>20, min_size>>20);
+		return 0;
+	}
+
+	return 1;
+}
+
 /* Read a standard AGPv3 bridge header */
 static u32 __init read_agp(int bus, int slot, int func, int cap, u32 *order)
 {
drivers/char/agp/amd64-agp.c +38 -1
index fd50ead..85cabd0 100644
@@ -14,7 +14,6 @@
 #include <linux/agp_backend.h>
 #include <linux/mmzone.h>
 #include <asm/page.h>		/* PAGE_SIZE */
-#include <asm/e820.h>
 #include <asm/k8.h>
 #include <asm/gart.h>
 #include "agp.h"
@@ -231,6 +230,44 @@ static const struct agp_bridge_driver amd_8151_driver = {
 	.agp_type_to_mask_type  = agp_generic_type_to_mask_type,
 };
 
+static int __devinit
+__is_ram(unsigned long pfn, unsigned long nr_pages, void *arg)
+{
+	return 1;
+}
+
+static int __devinit any_ram_in_range(u64 base, u64 size)
+{
+	unsigned long pfn, nr_pages;
+
+	pfn = base >> PAGE_SHIFT;
+	nr_pages = size >> PAGE_SHIFT;
+
+	return walk_system_ram_range(pfn, nr_pages, NULL, __is_ram) == 1;
+}
+
+static int __devinit aperture_valid(u64 aper_base, u32 aper_size, u32 min_size)
+{
+	if (!aper_base)
+		return 0;
+
+	if (aper_base + aper_size > 0x100000000ULL) {
+		printk(KERN_INFO "Aperture beyond 4GB. Ignoring.\n");
+		return 0;
+	}
+	if (any_ram_in_range(aper_base, aper_size)) {
+		printk(KERN_INFO "Aperture pointing to E820 RAM. Ignoring.\n");
+		return 0;
+	}
+	if (aper_size < min_size) {
+		printk(KERN_INFO "Aperture too small (%d MB) than (%d MB)\n",
+				 aper_size>>20, min_size>>20);
+		return 0;
+	}
+
+	return 1;
+}
+
 /* Some basic sanity checks for the aperture. */
 static int __devinit agp_aperture_valid(u64 aper, u32 size)
 {
Signature · 1 line
1.6.4.2
author
#59 lore raw

PATCH 28/31 x86: Add get_centaur_ram_top()

In reply to Yinghai Lu

So we can avoid to access e820.map[] directly.

later we could move e820 to static and _initdata

---
 arch/x86/include/asm/e820.h   |    9 ++++++
 arch/x86/kernel/cpu/centaur.c |   53 +-------------------------------------
 arch/x86/kernel/e820.c        |   56 +++++++++++++++++++++++++++++++++++++++++
 arch/x86/kernel/setup.c       |    2 +
 4 files changed, 69 insertions(+), 51 deletions(-)
arch/x86/include/asm/e820.h +9 -0
index 334281f..cd7de51 100644
@@ -76,6 +76,15 @@ struct e820map {
 /* see comment in arch/x86/kernel/e820.c */
 extern struct e820map e820;
 
+#if defined(CONFIG_X86_OOSTORE) && defined(CONFIG_CPU_SUP_CENTAUR)
+extern int centaur_ram_top;
+void get_centaur_ram_top(void);
+#else
+static inline void get_centaur_ram_top(void)
+{
+}
+#endif
+
 extern unsigned long pci_mem_start;
 extern int e820_any_mapped(u64 start, u64 end, unsigned type);
 extern int e820_all_mapped(u64 start, u64 end, unsigned type);
arch/x86/kernel/cpu/centaur.c +2 -51

65 lines not on this page · raw message

arch/x86/kernel/e820.c +56 -0

62 lines not on this page · raw message

arch/x86/kernel/setup.c +2 -0
index b00ccc4..0e52435 100644
@@ -875,6 +875,8 @@ void __init setup_arch(char **cmdline_p)
 	if (mtrr_trim_uncached_memory(max_pfn))
 		max_pfn = e820_end_of_ram_pfn();
 
+	get_centaur_ram_top();
+
 #ifdef CONFIG_X86_32
 	/* max_low_pfn get updated here */
 	find_low_pfn_range();
Signature · 1 line
1.6.4.2
author
#60 lore raw

PATCH 29/31 x86: Make e820_any_mapped to __init

In reply to Yinghai Lu

We don't need to expose e820_any_mapped anymore

---
 arch/x86/kernel/e820.c |    6 +++---
 1 files changed, 3 insertions(+), 3 deletions(-)
arch/x86/kernel/e820.c +3 -3
index 6b17893..2182bb0 100644
@@ -47,9 +47,10 @@ EXPORT_SYMBOL(pci_mem_start);
 /*
  * This function checks if any part of the range <start,end> is mapped
  * with type.
+ * phys_pud_init() is using it and is _meminit, but we have !after_bootmem
+ * so could use refok here
  */
-int
-e820_any_mapped(u64 start, u64 end, unsigned type)
+int __init_refok e820_any_mapped(u64 start, u64 end, unsigned type)
 {
 	int i;
 
@@ -64,7 +65,6 @@ e820_any_mapped(u64 start, u64 end, unsigned type)
 	}
 	return 0;
 }
-EXPORT_SYMBOL_GPL(e820_any_mapped);
 
 /*
  * This function checks if the entire range <start,end> is mapped with type.
Signature · 1 line
1.6.4.2
author
#61 lore raw

PATCH 30/31 x86: Use walk_system_ream_range()instead of e820.map directly

In reply to Yinghai Lu

So we can make e820 to be __initdata

---
 arch/x86/kernel/tboot.c |   22 +++++++++-------------
 1 files changed, 9 insertions(+), 13 deletions(-)
arch/x86/kernel/tboot.c +9 -13
index cc2c604..cf27d64 100644
@@ -170,34 +170,30 @@ static void tboot_create_trampoline(void)
 
 #ifdef CONFIG_ACPI_SLEEP
 
-static void add_mac_region(phys_addr_t start, unsigned long size)
+static int
+add_mac_region(unsigned long start_pfn, unsigned long nr_pages, void  *arg)
 {
+	u64 start = start_pfn;
+	u64 size = nr_pages;
 	struct tboot_mac_region *mr;
-	phys_addr_t end = start + size;
 
 	if (tboot->num_mac_regions >= MAX_TB_MAC_REGIONS)
 		panic("tboot: Too many MAC regions\n");
 
 	if (start && size) {
 		mr = &tboot->mac_regions[tboot->num_mac_regions++];
-		mr->start = round_down(start, PAGE_SIZE);
-		mr->size  = round_up(end, PAGE_SIZE) - mr->start;
+		mr->start = start << PAGE_SHIFT;
+		mr->size  = (u32) (size << PAGE_SHIFT);
 	}
+
+	return 0;
 }
 
 static int tboot_setup_sleep(void)
 {
-	int i;
-
 	tboot->num_mac_regions = 0;
 
-	for (i = 0; i < e820.nr_map; i++) {
-		if ((e820.map[i].type != E820_RAM)
-		 && (e820.map[i].type != E820_RESERVED_KERN))
-			continue;
-
-		add_mac_region(e820.map[i].addr, e820.map[i].size);
-	}
+	walk_system_ram_range(0, max_pfn, NULL, add_mac_region);
 
 	tboot->acpi_sinfo.kernel_s3_resume_vector = acpi_wakeup_address;
 
Signature · 1 line
1.6.4.2
author
#62 lore raw

PATCH 31/31 x86: make e820 to be __initdata

In reply to Yinghai Lu

Finally no users after init boot stage. We can free it

---
 arch/x86/include/asm/e820.h |    2 --
 arch/x86/kernel/e820.c      |    2 +-
 2 files changed, 1 insertions(+), 3 deletions(-)
arch/x86/include/asm/e820.h +0 -2
index cd7de51..f2ab72e 100644
@@ -73,8 +73,6 @@ struct e820map {
 #define BIOS_END		0x00100000
 
 #ifdef __KERNEL__
-/* see comment in arch/x86/kernel/e820.c */
-extern struct e820map e820;
 
 #if defined(CONFIG_X86_OOSTORE) && defined(CONFIG_CPU_SUP_CENTAUR)
 extern int centaur_ram_top;
arch/x86/kernel/e820.c +1 -1
index 2182bb0..5a0d688 100644
@@ -35,7 +35,7 @@
  * user can e.g. boot the original kernel with mem=1G while still booting the
  * next kernel with full memory.
  */
-struct e820map e820;
+static struct e820map __initdata e820;
 static struct e820map __initdata e820_saved;
 
 /* For PCI or other memory-mapped resources */
Signature · 1 line
1.6.4.2
#63 lore raw
In reply to Yinghai Lu

Review

Quoted 11 lines · Yinghai Lu

the new lmb could be used to early_res in x86.

Suggested by: David, Ben, and Thomas

First three patches should go into 2.6.34

(quoted diff) +0 -4
-v6: change sequence as requested by Thomas
-v7: seperate them to more patches
-v8: add boundary checking to make sure not free partial page.
-v9: use lmb_debug to control print out of reserve_lmb.
     add e820 clean up, and e820 become __initdata

Bike shedding perhaps, but can you maintain the naming convention, ie. lmb_xxx() rather than xxx_lmb(). Neither is necessarily better, but all the existing functions use the lmb_xxx() style.

cheers

author
#64 lore raw
Quoted 7 lines · Michael Ellerman
Quoted 6 lines · Yinghai Lu

the new lmb could be used to early_res in x86.

Suggested by: David, Ben, and Thomas

First three patches should go into 2.6.34

Quoted 10 lines

Review

Quoted 5 lines
(quoted diff) +0 -4
-v6: change sequence as requested by Thomas
-v7: seperate them to more patches
-v8: add boundary checking to make sure not free partial page.
-v9: use lmb_debug to control print out of reserve_lmb.
     add e820 clean up, and e820 become __initdata

Bike shedding perhaps, but can you maintain the naming convention, ie. lmb_xxx() rather than xxx_lmb(). Neither is necessarily better, but all the existing functions use the lmb_xxx() style.

so you want

find_lmb_area ==> lmb_find_area
reserve_lmb ==> lmb_reserve
free_lmb ==> lmb_free

first one is ok,

but next two we already have lmb_reserved and lmb_free without checking and increasing the size of region array.

should i use
lmb_reserve_with_check?

thanks

yinghai

author
#65 lore raw
Quoted 25 lines · Yinghai Lu
Quoted 7 lines · Michael Ellerman
Quoted 6 lines · Yinghai Lu

the new lmb could be used to early_res in x86.

Suggested by: David, Ben, and Thomas

First three patches should go into 2.6.34

Quoted 10 lines

Review

Quoted 5 lines
(quoted diff) +0 -4
-v6: change sequence as requested by Thomas
-v7: seperate them to more patches
-v8: add boundary checking to make sure not free partial page.
-v9: use lmb_debug to control print out of reserve_lmb.
     add e820 clean up, and e820 become __initdata

Bike shedding perhaps, but can you maintain the naming convention, ie. lmb_xxx() rather than xxx_lmb(). Neither is necessarily better, but all the existing functions use the lmb_xxx() style.

so you want

find_lmb_area ==> lmb_find_area
reserve_lmb ==> lmb_reserve
free_lmb ==> lmb_free

Quoted 7 lines

first one is ok,

but next two we already have lmb_reserved and lmb_free without checking and increasing the size of region array.

should i use
lmb_reserve_with_check?

I change

find_lmb_area ==> lmb_find_area
reserve_lmb ==> lmb_reserve_area
free_lmb ==> lmb_free_area

does that look ok to you?

Thanks

Yinghai

#66 lore raw
In reply to Yinghai Lu
Quoted 21 lines · Yinghai Lu
Quoted 7 lines · Michael Ellerman
Quoted 6 lines · Yinghai Lu

the new lmb could be used to early_res in x86.

Suggested by: David, Ben, and Thomas

First three patches should go into 2.6.34

Quoted 10 lines

Review

Quoted 5 lines
(quoted diff) +0 -4
-v6: change sequence as requested by Thomas
-v7: seperate them to more patches
-v8: add boundary checking to make sure not free partial page.
-v9: use lmb_debug to control print out of reserve_lmb.
     add e820 clean up, and e820 become __initdata

Bike shedding perhaps, but can you maintain the naming convention, ie. lmb_xxx() rather than xxx_lmb(). Neither is necessarily better, but all the existing functions use the lmb_xxx() style.

so you want

Quoted 7 lines

find_lmb_area ==> lmb_find_area
reserve_lmb ==> lmb_reserve
free_lmb ==> lmb_free

first one is ok,

but next two we already have lmb_reserved and lmb_free without checking and increasing the size of region array.

That was the point of my other mail. We now have two lmb APIs, one which checks if the array will overflow and one which doesn't. That seems like a bad idea. Having one called lmb_free() and one called free_lmb() is definitely a bad idea, because it's completely non obvious which one caters for overflow.

cheers

author
#67 lore raw
Quoted 28 lines · Michael Ellerman
Quoted 27 lines · Yinghai Lu
Quoted 7 lines · Michael Ellerman
Quoted 6 lines · Yinghai Lu

the new lmb could be used to early_res in x86.

Suggested by: David, Ben, and Thomas

First three patches should go into 2.6.34

Quoted 10 lines

Review

Quoted 5 lines
(quoted diff) +0 -4
-v6: change sequence as requested by Thomas
-v7: seperate them to more patches
-v8: add boundary checking to make sure not free partial page.
-v9: use lmb_debug to control print out of reserve_lmb.
     add e820 clean up, and e820 become __initdata

Bike shedding perhaps, but can you maintain the naming convention, ie. lmb_xxx() rather than xxx_lmb(). Neither is necessarily better, but all the existing functions use the lmb_xxx() style.

so you want

find_lmb_area ==> lmb_find_area
reserve_lmb ==> lmb_reserve
free_lmb ==> lmb_free

first one is ok,

Quoted 7 lines
Quoted 1 line

but next two we already have lmb_reserved and lmb_free without checking and increasing the size of region array.

That was the point of my other mail. We now have two lmb APIs, one which checks if the array will overflow and one which doesn't. That seems like a bad idea. Having one called lmb_free() and one called free_lmb() is definitely a bad idea, because it's completely non obvious which one caters for overflow.

I want to keep the affects to other lmb users to minium at first.

and we can merge those functions later.

or you insist on merging them in this patchset?

Yinghai

#68 lore raw
Quoted 31 lines · Yinghai Lu
Quoted 30 lines · Michael Ellerman
Quoted 28 lines · Yinghai Lu
Quoted 7 lines · Michael Ellerman
Quoted 6 lines · Yinghai Lu

the new lmb could be used to early_res in x86.

Suggested by: David, Ben, and Thomas

First three patches should go into 2.6.34

Quoted 10 lines

Review

Quoted 5 lines
(quoted diff) +0 -4
-v6: change sequence as requested by Thomas
-v7: seperate them to more patches
-v8: add boundary checking to make sure not free partial page.
-v9: use lmb_debug to control print out of reserve_lmb.
     add e820 clean up, and e820 become __initdata

Bike shedding perhaps, but can you maintain the naming convention, ie. lmb_xxx() rather than xxx_lmb(). Neither is necessarily better, but all the existing functions use the lmb_xxx() style.

so you want

find_lmb_area ==> lmb_find_area
reserve_lmb ==> lmb_reserve
free_lmb ==> lmb_free

first one is ok,

but next two we already have lmb_reserved and lmb_free without checking and increasing the size of region array.

Quoted 7 lines
Quoted 5 lines

That was the point of my other mail. We now have two lmb APIs, one which checks if the array will overflow and one which doesn't. That seems like a bad idea. Having one called lmb_free() and one called free_lmb() is definitely a bad idea, because it's completely non obvious which one caters for overflow.

I want to keep the affects to other lmb users to minium at first.

That's a good plan, but I don't think this is the nicest way to do it.

Quoted 3 lines

and we can merge those functions later.

or you insist on merging them in this patchset?

No I don't insist.

I _suggest_ that if we want to avoid affecting existing lmb users, then the checking logic should go into the existing API, but be #ifdef'ed for now - eg. CONFIG_DYNAMIC_LMB or something. That way you avoid affecting existing users (more or less), but you don't add a new API that you then have to remove later.

Having said that I don't think it really does affect existing users that much. We still have the statically defined region arrays, and they're still the same size, so sparc and powerpc should never need to resize, except on machines where we currently run out of space in the array anyway.

cheers

author
#69 lore raw
Quoted 5 lines · Michael Ellerman
Quoted 2 lines

I want to keep the affects to other lmb users to minium at first.

That's a good plan, but I don't think this is the nicest way to do it.

Too much to be test. so better to keep less impact to current lmb users.

Quoted 12 lines
Quoted 3 lines

and we can merge those functions later.

or you insist on merging them in this patchset?

No I don't insist.

I _suggest_ that if we want to avoid affecting existing lmb users, then the checking logic should go into the existing API, but be #ifdef'ed for now - eg. CONFIG_DYNAMIC_LMB or something. That way you avoid affecting existing users (more or less), but you don't add a new API that you then have to remove later.

Quoted 5 lines

Having said that I don't think it really does affect existing users that much. We still have the statically defined region arrays, and they're still the same size, so sparc and powerpc should never need to resize, except on machines where we currently run out of space in the array anyway.

dynamic region array for lmb.memory and lmb.reserved, actually is only need when plan to use lmb to replace bootmem code.

Thanks

Yinghai

#70 lore raw
In reply to Michael Ellerman
Quoted 11 lines · Michael Ellerman

I _suggest_ that if we want to avoid affecting existing lmb users, then the checking logic should go into the existing API, but be #ifdef'ed for now - eg. CONFIG_DYNAMIC_LMB or something. That way you avoid affecting existing users (more or less), but you don't add a new API that you then have to remove later.

Having said that I don't think it really does affect existing users that much. We still have the statically defined region arrays, and they're still the same size, so sparc and powerpc should never need to resize, except on machines where we currently run out of space in the array anyway.

We'll want the dynamic sizing if we switch bootmem to lmb (though last I looked, that still needs sparsemem to be fixed as well as it still uses bootmem). There's also some interest on ARM side I've heard.

I still have my evil plan to turn it into lists anyways, but we'll see how that goes later. In any case, the doubling ability does the job for now and yes, it should definitely be part of the core.

LMB should be able to use it's own storage to do the doubling. IE. Always guarantee that you have at least 1 or 2 free entries in the table, if you're going past that threshold, then use the remaining entry to allocate a new table. Easy :-)

Cheers,
Ben.

#71 lore raw
In reply to Yinghai Lu
Quoted 6 lines · Yinghai Lu
Quoted 1 line

That was the point of my other mail. We now have two lmb APIs, one

which

Quoted 1 line

checks if the array will overflow and one which doesn't. That seems

like

Quoted 1 line

a bad idea. Having one called lmb_free() and one called free_lmb()

is

Quoted 8 lines
Quoted 2 lines

definitely a bad idea, because it's completely non obvious which one caters for overflow.

I want to keep the affects to other lmb users to minium at first.

and we can merge those functions later.

or you insist on merging them in this patchset?

As a separate patch sure, but you should really separate the patch series that changes LMB from the one that moves x86 to it imho. It would make things much clearer.

It would also allow you to spend some time properly -documenting- why you need to change LMB the way you do, since it's non obvious for those not familiar with x86 needs. I'm not objecting to the changes, I'm just asking for much better documentation as to why they are needed and what function they provide.

Cheers,
Ben.

author
#72 lore raw
Quoted 12 lines · Benjamin Herrenschmidt
Quoted 11 lines · Yinghai Lu
Quoted 1 line

That was the point of my other mail. We now have two lmb APIs, one

which

Quoted 1 line

checks if the array will overflow and one which doesn't. That seems

like

Quoted 1 line

a bad idea. Having one called lmb_free() and one called free_lmb()

is

Quoted 2 lines

definitely a bad idea, because it's completely non obvious which one caters for overflow.

I want to keep the affects to other lmb users to minium at first.

Quoted 7 lines
Quoted 3 lines

and we can merge those functions later.

or you insist on merging them in this patchset?

As a separate patch sure, but you should really separate the patch series that changes LMB from the one that moves x86 to it imho. It would make things much clearer.

Those patches should go through tip/x86 ?

Please check the patches only have "lmb:" in the subject.

Quoted 6 lines

It would also allow you to spend some time properly -documenting- why you need to change LMB the way you do, since it's non obvious for those not familiar with x86 needs. I'm not objecting to the changes, I'm just asking for much better documentation as to why they are needed and what function they provide.

Thanks, will try to write more changelog for next reversion.

Yinghai