[PATCH v2 0/3] Optimize CONFIG_DEBUG_PAGEALLOC

STALE3872d

Revision v2 of 3 in this series.

7 messages, 3 authors, 2016-01-27 · open the first message on its own page

[PATCH v2 0/3] Optimize CONFIG_DEBUG_PAGEALLOC

From: Christian Borntraeger <hidden>
Date: 2016-01-27 08:50:04

As CONFIG_DEBUG_PAGEALLOC can be enabled/disabled via kernel
parameters we can optimize some cases by checking the enablement
state.

I have done s390 and x86 as examples.
s390 should be ok, I tested several combinations, x86 seems to
work as well, but I have some concerns about the 1GB pages.
The previous code did not bother at all with CONFIG_DEBUG_PAGEALLOC
and I do not fully understand why.

Power can probably do the same, Michael/Ben?
I am not sure about sparc. Sparc seems to allocate the TSB buffer
really early. David?


V1->V2:
- replace DEBUG_PAGEALLOC(disabled/enabled) with DEBUG_PAGEALLOC
  dump_stack for s390/x86
- add /* CONFIG_DEBUG_PAGEALLOC */ to else and endif


Christian Borntraeger (3):
  mm: provide debug_pagealloc_enabled() without CONFIG_DEBUG_PAGEALLOC
  x86: query dynamic DEBUG_PAGEALLOC setting
  s390: query dynamic DEBUG_PAGEALLOC setting

 arch/s390/kernel/dumpstack.c |  8 ++++----
 arch/s390/mm/vmem.c          | 10 ++++------
 arch/x86/kernel/dumpstack.c  |  5 ++---
 arch/x86/mm/init.c           |  7 ++++---
 arch/x86/mm/pageattr.c       | 14 ++++----------
 include/linux/mm.h           |  9 +++++++--
 6 files changed, 25 insertions(+), 28 deletions(-)

-- 
2.3.0

[PATCH v2 2/3] x86: query dynamic DEBUG_PAGEALLOC setting

From: Christian Borntraeger <hidden>
Date: 2016-01-27 08:50:11

We can use debug_pagealloc_enabled() to check if we can map
the identity mapping with 2MB pages. We can also add the state
into the dump_stack output.

The patch does not touch the code for the 1GB pages, which ignored
CONFIG_DEBUG_PAGEALLOC. Do we need to fence this as well?

Signed-off-by: Christian Borntraeger <redacted>
---
 arch/x86/kernel/dumpstack.c |  5 ++---
 arch/x86/mm/init.c          |  7 ++++---
 arch/x86/mm/pageattr.c      | 14 ++++----------
 3 files changed, 10 insertions(+), 16 deletions(-)

diff --git a/arch/x86/kernel/dumpstack.c b/arch/x86/kernel/dumpstack.c
index 9c30acf..32e5699 100644
--- a/arch/x86/kernel/dumpstack.c
+++ b/arch/x86/kernel/dumpstack.c

@@ -265,9 +265,8 @@ int __die(const char *str, struct pt_regs *regs, long err)
 #ifdef CONFIG_SMP
 	printk("SMP ");
 #endif
-#ifdef CONFIG_DEBUG_PAGEALLOC
-	printk("DEBUG_PAGEALLOC ");
-#endif
+	if (debug_pagealloc_enabled())
+		printk("DEBUG_PAGEALLOC ");
 #ifdef CONFIG_KASAN
 	printk("KASAN");
 #endif

diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c
index 493f541..39823fd 100644
--- a/arch/x86/mm/init.c
+++ b/arch/x86/mm/init.c

@@ -150,13 +150,14 @@ static int page_size_mask;
 
 static void __init probe_page_size_mask(void)
 {
-#if !defined(CONFIG_DEBUG_PAGEALLOC) && !defined(CONFIG_KMEMCHECK)
+#if !defined(CONFIG_KMEMCHECK)
 	/*
-	 * For CONFIG_DEBUG_PAGEALLOC, identity mapping will use small pages.
+	 * For CONFIG_KMEMCHECK or pagealloc debugging, identity mapping will
+	 * use small pages.
 	 * This will simplify cpa(), which otherwise needs to support splitting
 	 * large pages into small in interrupt context, etc.
 	 */
-	if (cpu_has_pse)
+	if (cpu_has_pse && !debug_pagealloc_enabled())
 		page_size_mask |= 1 << PG_LEVEL_2M;
 #endif

diff --git a/arch/x86/mm/pageattr.c b/arch/x86/mm/pageattr.c
index a3137a4..a49c8fd 100644
--- a/arch/x86/mm/pageattr.c
+++ b/arch/x86/mm/pageattr.c

@@ -103,12 +103,6 @@ static inline unsigned long highmap_end_pfn(void)
 
 #endif
 
-#ifdef CONFIG_DEBUG_PAGEALLOC
-# define debug_pagealloc 1
-#else
-# define debug_pagealloc 0
-#endif
-
 static inline int
 within(unsigned long addr, unsigned long start, unsigned long end)
 {

@@ -703,10 +697,10 @@ static int split_large_page(struct cpa_data *cpa, pte_t *kpte,
 {
 	struct page *base;
 
-	if (!debug_pagealloc)
+	if (!debug_pagealloc_enabled())
 		spin_unlock(&cpa_lock);
 	base = alloc_pages(GFP_KERNEL | __GFP_NOTRACK, 0);
-	if (!debug_pagealloc)
+	if (!debug_pagealloc_enabled())
 		spin_lock(&cpa_lock);
 	if (!base)
 		return -ENOMEM;

@@ -1326,10 +1320,10 @@ static int __change_page_attr_set_clr(struct cpa_data *cpa, int checkalias)
 		if (cpa->flags & (CPA_ARRAY | CPA_PAGES_ARRAY))
 			cpa->numpages = 1;
 
-		if (!debug_pagealloc)
+		if (!debug_pagealloc_enabled())
 			spin_lock(&cpa_lock);
 		ret = __change_page_attr(cpa, checkalias);
-		if (!debug_pagealloc)
+		if (!debug_pagealloc_enabled())
 			spin_unlock(&cpa_lock);
 		if (ret)
 			return ret;

-- 
2.3.0

[PATCH v2 3/3] s390: query dynamic DEBUG_PAGEALLOC setting

From: Christian Borntraeger <hidden>
Date: 2016-01-27 08:50:35

We can use debug_pagealloc_enabled() to check if we can map
the identity mapping with 1MB/2GB pages as well as to print
the current setting in dump_stack.

Signed-off-by: Christian Borntraeger <redacted>
---
 arch/s390/kernel/dumpstack.c |  8 ++++----
 arch/s390/mm/vmem.c          | 10 ++++------
 2 files changed, 8 insertions(+), 10 deletions(-)

diff --git a/arch/s390/kernel/dumpstack.c b/arch/s390/kernel/dumpstack.c
index dc8e204..3f352e9 100644
--- a/arch/s390/kernel/dumpstack.c
+++ b/arch/s390/kernel/dumpstack.c

@@ -11,6 +11,7 @@
 #include <linux/export.h>
 #include <linux/kdebug.h>
 #include <linux/ptrace.h>
+#include <linux/mm.h>
 #include <linux/module.h>
 #include <linux/sched.h>
 #include <asm/processor.h>

@@ -184,10 +185,9 @@ void die(struct pt_regs *regs, const char *str)
 #endif
 #ifdef CONFIG_SMP
 	printk("SMP ");
-#endif
-#ifdef CONFIG_DEBUG_PAGEALLOC
-	printk("DEBUG_PAGEALLOC");
-#endif
+#endif	
+if (debug_pagealloc_enabled())
+		printk("DEBUG_PAGEALLOC");
 	printk("\n");
 	notify_die(DIE_OOPS, str, regs, 0, regs->int_code & 0xffff, SIGSEGV);
 	print_modules();

diff --git a/arch/s390/mm/vmem.c b/arch/s390/mm/vmem.c
index ef7d6c8..d27fccba 100644
--- a/arch/s390/mm/vmem.c
+++ b/arch/s390/mm/vmem.c

@@ -94,16 +94,15 @@ static int vmem_add_mem(unsigned long start, unsigned long size, int ro)
 			pgd_populate(&init_mm, pg_dir, pu_dir);
 		}
 		pu_dir = pud_offset(pg_dir, address);
-#ifndef CONFIG_DEBUG_PAGEALLOC
 		if (MACHINE_HAS_EDAT2 && pud_none(*pu_dir) && address &&
-		    !(address & ~PUD_MASK) && (address + PUD_SIZE <= end)) {
+		    !(address & ~PUD_MASK) && (address + PUD_SIZE <= end) &&
+		     !debug_pagealloc_enabled()) {
 			pud_val(*pu_dir) = __pa(address) |
 				_REGION_ENTRY_TYPE_R3 | _REGION3_ENTRY_LARGE |
 				(ro ? _REGION_ENTRY_PROTECT : 0);
 			address += PUD_SIZE;
 			continue;
 		}
-#endif
 		if (pud_none(*pu_dir)) {
 			pm_dir = vmem_pmd_alloc();
 			if (!pm_dir)

@@ -111,9 +110,9 @@ static int vmem_add_mem(unsigned long start, unsigned long size, int ro)
 			pud_populate(&init_mm, pu_dir, pm_dir);
 		}
 		pm_dir = pmd_offset(pu_dir, address);
-#ifndef CONFIG_DEBUG_PAGEALLOC
 		if (MACHINE_HAS_EDAT1 && pmd_none(*pm_dir) && address &&
-		    !(address & ~PMD_MASK) && (address + PMD_SIZE <= end)) {
+		    !(address & ~PMD_MASK) && (address + PMD_SIZE <= end) &&
+		    !debug_pagealloc_enabled()) {
 			pmd_val(*pm_dir) = __pa(address) |
 				_SEGMENT_ENTRY | _SEGMENT_ENTRY_LARGE |
 				_SEGMENT_ENTRY_YOUNG |

@@ -121,7 +120,6 @@ static int vmem_add_mem(unsigned long start, unsigned long size, int ro)
 			address += PMD_SIZE;
 			continue;
 		}
-#endif
 		if (pmd_none(*pm_dir)) {
 			pt_dir = vmem_pte_alloc(address);
 			if (!pt_dir)

-- 
2.3.0

[PATCH v2 1/3] mm: provide debug_pagealloc_enabled() without CONFIG_DEBUG_PAGEALLOC

From: Christian Borntraeger <hidden>
Date: 2016-01-27 08:51:05

We can provide debug_pagealloc_enabled() also if CONFIG_DEBUG_PAGEALLOC
is not set. It will return false in that case.

Signed-off-by: Christian Borntraeger <redacted>
Acked-by: David Rientjes <rientjes@google.com>
---
 include/linux/mm.h | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/include/linux/mm.h b/include/linux/mm.h
index 7783073..56cab4e 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h

@@ -2147,13 +2147,18 @@ kernel_map_pages(struct page *page, int numpages, int enable)
 #ifdef CONFIG_HIBERNATION
 extern bool kernel_page_present(struct page *page);
 #endif /* CONFIG_HIBERNATION */
-#else
+#else  /* CONFIG_DEBUG_PAGEALLOC */
+static inline bool debug_pagealloc_enabled(void)
+{
+	return false;
+}
+
 static inline void
 kernel_map_pages(struct page *page, int numpages, int enable) {}
 #ifdef CONFIG_HIBERNATION
 static inline bool kernel_page_present(struct page *page) { return true; }
 #endif /* CONFIG_HIBERNATION */
-#endif
+#endif /* CONFIG_DEBUG_PAGEALLOC */
 
 #ifdef __HAVE_ARCH_GATE_AREA
 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);

-- 
2.3.0

Re: [PATCH v2 1/3] mm: provide debug_pagealloc_enabled() without CONFIG_DEBUG_PAGEALLOC

From: Thomas Gleixner <hidden>
Date: 2016-01-27 09:16:49

On Wed, 27 Jan 2016, Christian Borntraeger wrote:

We can provide debug_pagealloc_enabled() also if CONFIG_DEBUG_PAGEALLOC
is not set. It will return false in that case.

Signed-off-by: Christian Borntraeger <redacted>
Acked-by: David Rientjes <rientjes@google.com>

Acked-by: Thomas Gleixner <redacted>

Re: [PATCH v2 2/3] x86: query dynamic DEBUG_PAGEALLOC setting

From: Thomas Gleixner <hidden>
Date: 2016-01-27 09:17:09

On Wed, 27 Jan 2016, Christian Borntraeger wrote:

We can use debug_pagealloc_enabled() to check if we can map
the identity mapping with 2MB pages. We can also add the state
into the dump_stack output.

The patch does not touch the code for the 1GB pages, which ignored
CONFIG_DEBUG_PAGEALLOC. Do we need to fence this as well?

Signed-off-by: Christian Borntraeger <redacted>

Reviewed-by: Thomas Gleixner <redacted>

Re: [PATCH v2 3/3] s390: query dynamic DEBUG_PAGEALLOC setting

From: Heiko Carstens <hidden>
Date: 2016-01-27 09:55:56

On Wed, Jan 27, 2016 at 09:50:18AM +0100, Christian Borntraeger wrote:

quoted hunk

We can use debug_pagealloc_enabled() to check if we can map
the identity mapping with 1MB/2GB pages as well as to print
the current setting in dump_stack.

Signed-off-by: Christian Borntraeger <redacted>
---
 arch/s390/kernel/dumpstack.c |  8 ++++----
 arch/s390/mm/vmem.c          | 10 ++++------
 2 files changed, 8 insertions(+), 10 deletions(-)

diff --git a/arch/s390/kernel/dumpstack.c b/arch/s390/kernel/dumpstack.c
index dc8e204..3f352e9 100644
--- a/arch/s390/kernel/dumpstack.c
+++ b/arch/s390/kernel/dumpstack.c

@@ -11,6 +11,7 @@
 #include <linux/export.h>
 #include <linux/kdebug.h>
 #include <linux/ptrace.h>
+#include <linux/mm.h>
 #include <linux/module.h>
 #include <linux/sched.h>
 #include <asm/processor.h>

@@ -184,10 +185,9 @@ void die(struct pt_regs *regs, const char *str)
 #endif
 #ifdef CONFIG_SMP
 	printk("SMP ");
-#endif
-#ifdef CONFIG_DEBUG_PAGEALLOC
-	printk("DEBUG_PAGEALLOC");
-#endif
+#endif	
+if (debug_pagealloc_enabled())
+		printk("DEBUG_PAGEALLOC");
 	printk("\n");

Indentation is broken ("if").
Besides that

Reviewed-by: Heiko Carstens <redacted>

`h`	back out one level
`j`	next message in thread
`k`	previous message in thread
`l`	drill in
`Esc`	close help / fold thread tree
`?`	toggle this help