[PATCH 1/3] zsmalloc: s/firstpage/page in new copy map funcs

[PATCH 1/3] zsmalloc: s/firstpage/page in new copy map funcs

[Seth Jennings]

firstpage already has precedent and meaning the first page
of a zspage.  In the case of the copy mapping functions,
it is the first of a pair of pages needing to be mapped.

This patch just renames the firstpage argument to "page" to
avoid confusion.

Signed-off-by: Seth Jennings <sjenning@linux.vnet.ibm.com>
---
 drivers/staging/zsmalloc/zsmalloc-main.c |   12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/drivers/staging/zsmalloc/zsmalloc-main.c b/drivers/staging/zsmalloc/zsmalloc-main.c
index 8b0bcb6..3c83c65 100644
--- a/drivers/staging/zsmalloc/zsmalloc-main.c
+++ b/drivers/staging/zsmalloc/zsmalloc-main.c
@@ -470,15 +470,15 @@ static struct page *find_get_zspage(struct size_class *class)
 	return page;
 }
 
-static void zs_copy_map_object(char *buf, struct page *firstpage,
+static void zs_copy_map_object(char *buf, struct page *page,
 				int off, int size)
 {
 	struct page *pages[2];
 	int sizes[2];
 	void *addr;
 
-	pages[0] = firstpage;
-	pages[1] = get_next_page(firstpage);
+	pages[0] = page;
+	pages[1] = get_next_page(page);
 	BUG_ON(!pages[1]);
 
 	sizes[0] = PAGE_SIZE - off;
@@ -493,15 +493,15 @@ static void zs_copy_map_object(char *buf, struct page *firstpage,
 	kunmap_atomic(addr);
 }
 
-static void zs_copy_unmap_object(char *buf, struct page *firstpage,
+static void zs_copy_unmap_object(char *buf, struct page *page,
 				int off, int size)
 {
 	struct page *pages[2];
 	int sizes[2];
 	void *addr;
 
-	pages[0] = firstpage;
-	pages[1] = get_next_page(firstpage);
+	pages[0] = page;
+	pages[1] = get_next_page(page);
 	BUG_ON(!pages[1]);
 
 	sizes[0] = PAGE_SIZE - off;
-- 
1.7.9.5

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 2/3] zsmalloc: prevent mappping in interrupt context

[Seth Jennings]

Because we use per-cpu mapping areas shared among the
pools/users, we can't allow mapping in interrupt context
because it can corrupt another users mappings.

Signed-off-by: Seth Jennings <sjenning@linux.vnet.ibm.com>
---
 drivers/staging/zsmalloc/zsmalloc-main.c |    8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/drivers/staging/zsmalloc/zsmalloc-main.c b/drivers/staging/zsmalloc/zsmalloc-main.c
index 3c83c65..b86133f 100644
--- a/drivers/staging/zsmalloc/zsmalloc-main.c
+++ b/drivers/staging/zsmalloc/zsmalloc-main.c
@@ -75,6 +75,7 @@
 #include <linux/cpumask.h>
 #include <linux/cpu.h>
 #include <linux/vmalloc.h>
+#include <linux/hardirq.h>
 
 #include "zsmalloc.h"
 #include "zsmalloc_int.h"
@@ -761,6 +762,13 @@ void *zs_map_object(struct zs_pool *pool, unsigned long handle,
 
 	BUG_ON(!handle);
 
+	/*
+	 * Because we use per-cpu mapping areas shared among the
+	 * pools/users, we can't allow mapping in interrupt context
+	 * because it can corrupt another users mappings.
+	 */
+	BUG_ON(in_interrupt());
+
 	obj_handle_to_location(handle, &page, &obj_idx);
 	get_zspage_mapping(get_first_page(page), &class_idx, &fg);
 	class = &pool->size_class[class_idx];
-- 
1.7.9.5

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 2/3] zsmalloc: prevent mappping in interrupt context

[Minchan Kim]

On Wed, Jul 18, 2012 at 11:55:55AM -0500, Seth Jennings wrote:
> Because we use per-cpu mapping areas shared among the
> pools/users, we can't allow mapping in interrupt context
> because it can corrupt another users mappings.
> 
> Signed-off-by: Seth Jennings <sjenning@linux.vnet.ibm.com>
Acked-by: Minchan Kim <minchan@kernel.org>

-- 
Kind regards,
Minchan Kim

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 3/3] zsmalloc: add page table mapping method

[Seth Jennings]

This patchset provides page mapping via the page table.
On some archs, most notably ARM, this method has been
demonstrated to be faster than copying.

The logic controlling the method selection (copy vs page table)
is controlled by the definition of USE_PGTABLE_MAPPING which
is/can be defined for any arch that performs better with page
table mapping.

Signed-off-by: Seth Jennings <sjenning@linux.vnet.ibm.com>
---
 drivers/staging/zsmalloc/zsmalloc-main.c |  182 ++++++++++++++++++++++--------
 drivers/staging/zsmalloc/zsmalloc_int.h  |    6 -
 2 files changed, 134 insertions(+), 54 deletions(-)

diff --git a/drivers/staging/zsmalloc/zsmalloc-main.c b/drivers/staging/zsmalloc/zsmalloc-main.c
index b86133f..defe350 100644
--- a/drivers/staging/zsmalloc/zsmalloc-main.c
+++ b/drivers/staging/zsmalloc/zsmalloc-main.c
@@ -89,6 +89,30 @@
 #define CLASS_IDX_MASK	((1 << CLASS_IDX_BITS) - 1)
 #define FULLNESS_MASK	((1 << FULLNESS_BITS) - 1)
 
+/*
+ * By default, zsmalloc uses a copy-based object mapping method to access
+ * allocations that span two pages. However, if a particular architecture
+ * 1) Implements local_flush_tlb_kernel_range() and 2) Performs VM mapping
+ * faster than copying, then it should be added here so that
+ * USE_PGTABLE_MAPPING is defined. This causes zsmalloc to use page table
+ * mapping rather than copying
+ * for object mapping.
+*/
+#if defined(CONFIG_ARM)
+#define USE_PGTABLE_MAPPING
+#endif
+
+struct mapping_area {
+#ifdef USE_PGTABLE_MAPPING
+	struct vm_struct *vm; /* vm area for mapping object that span pages */
+#else
+	char *vm_buf; /* copy buffer for objects that span pages */
+#endif
+	char *vm_addr; /* address of kmap_atomic()'ed pages */
+	enum zs_mapmode vm_mm; /* mapping mode */
+};
+
+
 /* per-cpu VM mapping areas for zspage accesses that cross page boundaries */
 static DEFINE_PER_CPU(struct mapping_area, zs_map_area);
 
@@ -471,16 +495,83 @@ static struct page *find_get_zspage(struct size_class *class)
 	return page;
 }
 
-static void zs_copy_map_object(char *buf, struct page *page,
-				int off, int size)
+#ifdef USE_PGTABLE_MAPPING
+static inline int __zs_cpu_up(struct mapping_area *area)
+{
+	/*
+	 * Make sure we don't leak memory if a cpu UP notification
+	 * and zs_init() race and both call zs_cpu_up() on the same cpu
+	 */
+	if (area->vm)
+		return 0;
+	area->vm = alloc_vm_area(PAGE_SIZE * 2, NULL);
+	if (!area->vm)
+		return -ENOMEM;
+	return 0;
+}
+
+static inline void __zs_cpu_down(struct mapping_area *area)
+{
+	if (area->vm)
+		free_vm_area(area->vm);
+	area->vm = NULL;
+}
+
+static inline void *__zs_map_object(struct mapping_area *area,
+				struct page *pages[2], int off, int size)
+{
+	BUG_ON(map_vm_area(area->vm, PAGE_KERNEL, &pages));
+	area->vm_addr = area->vm->addr;
+	return area->vm_addr + off;
+}
+
+static inline void __zs_unmap_object(struct mapping_area *area,
+				struct page *pages[2], int off, int size)
+{
+	unsigned long addr = (unsigned long)area->vm_addr;
+	unsigned long end = addr + (PAGE_SIZE * 2);
+
+	flush_cache_vunmap(addr, end);
+	unmap_kernel_range_noflush(addr, PAGE_SIZE * 2);
+	local_flush_tlb_kernel_range(addr, end);
+}
+
+#else /* USE_PGTABLE_MAPPING */
+
+static inline int __zs_cpu_up(struct mapping_area *area)
+{
+	/*
+	 * Make sure we don't leak memory if a cpu UP notification
+	 * and zs_init() race and both call zs_cpu_up() on the same cpu
+	 */
+	if (area->vm_buf)
+		return 0;
+	area->vm_buf = (char *)__get_free_page(GFP_KERNEL);
+	if (!area->vm_buf)
+		return -ENOMEM;
+	return 0;
+}
+
+static inline void __zs_cpu_down(struct mapping_area *area)
+{
+	if (area->vm_buf)
+		free_page((unsigned long)area->vm_buf);
+	area->vm_buf = NULL;
+}
+
+static void *__zs_map_object(struct mapping_area *area,
+			struct page *pages[2], int off, int size)
 {
-	struct page *pages[2];
 	int sizes[2];
 	void *addr;
+	char *buf = area->vm_buf;
 
-	pages[0] = page;
-	pages[1] = get_next_page(page);
-	BUG_ON(!pages[1]);
+	/* disable page faults to match kmap_atomic() return conditions */
+	pagefault_disable();
+
+	/* no read fastpath */
+	if (area->vm_mm == ZS_MM_WO)
+		goto out;
 
 	sizes[0] = PAGE_SIZE - off;
 	sizes[1] = size - sizes[0];
@@ -492,18 +583,20 @@ static void zs_copy_map_object(char *buf, struct page *page,
 	addr = kmap_atomic(pages[1]);
 	memcpy(buf + sizes[0], addr, sizes[1]);
 	kunmap_atomic(addr);
+out:
+	return area->vm_buf;
 }
 
-static void zs_copy_unmap_object(char *buf, struct page *page,
-				int off, int size)
+static void __zs_unmap_object(struct mapping_area *area,
+			struct page *pages[2], int off, int size)
 {
-	struct page *pages[2];
 	int sizes[2];
 	void *addr;
+	char *buf = area->vm_buf;
 
-	pages[0] = page;
-	pages[1] = get_next_page(page);
-	BUG_ON(!pages[1]);
+	/* no write fastpath */
+	if (area->vm_mm == ZS_MM_RO)
+		goto out;
 
 	sizes[0] = PAGE_SIZE - off;
 	sizes[1] = size - sizes[0];
@@ -515,34 +608,31 @@ static void zs_copy_unmap_object(char *buf, struct page *page,
 	addr = kmap_atomic(pages[1]);
 	memcpy(addr, buf + sizes[0], sizes[1]);
 	kunmap_atomic(addr);
+
+out:
+	/* enable page faults to match kunmap_atomic() return conditions */
+	pagefault_enable();
 }
 
+#endif /* USE_PGTABLE_MAPPING */
+
 static int zs_cpu_notifier(struct notifier_block *nb, unsigned long action,
 				void *pcpu)
 {
-	int cpu = (long)pcpu;
+	int ret, cpu = (long)pcpu;
 	struct mapping_area *area;
 
 	switch (action) {
 	case CPU_UP_PREPARE:
 		area = &per_cpu(zs_map_area, cpu);
-		/*
-		 * Make sure we don't leak memory if a cpu UP notification
-		 * and zs_init() race and both call zs_cpu_up() on the same cpu
-		 */
-		if (area->vm_buf)
-			return 0;
-		area->vm_buf = (char *)__get_free_page(GFP_KERNEL);
-		if (!area->vm_buf)
-			return -ENOMEM;
-		return 0;
+		ret = __zs_cpu_up(area);
+		if (ret)
+			return notifier_from_errno(ret);
 		break;
 	case CPU_DEAD:
 	case CPU_UP_CANCELED:
 		area = &per_cpu(zs_map_area, cpu);
-		if (area->vm_buf)
-			free_page((unsigned long)area->vm_buf);
-		area->vm_buf = NULL;
+		__zs_cpu_down(area);
 		break;
 	}
 
@@ -759,6 +849,7 @@ void *zs_map_object(struct zs_pool *pool, unsigned long handle,
 	enum fullness_group fg;
 	struct size_class *class;
 	struct mapping_area *area;
+	struct page *pages[2];
 
 	BUG_ON(!handle);
 
@@ -775,19 +866,19 @@ void *zs_map_object(struct zs_pool *pool, unsigned long handle,
 	off = obj_idx_to_offset(page, obj_idx, class->size);
 
 	area = &get_cpu_var(zs_map_area);
+	area->vm_mm = mm;
 	if (off + class->size <= PAGE_SIZE) {
 		/* this object is contained entirely within a page */
 		area->vm_addr = kmap_atomic(page);
 		return area->vm_addr + off;
 	}
 
-	/* disable page faults to match kmap_atomic() return conditions */
-	pagefault_disable();
+	/* this object spans two pages */
+	pages[0] = page;
+	pages[1] = get_next_page(page);
+	BUG_ON(!pages[1]);
 
-	if (mm != ZS_MM_WO)
-		zs_copy_map_object(area->vm_buf, page, off, class->size);
-	area->vm_addr = NULL;
-	return area->vm_buf;
+	return __zs_map_object(area, pages, off, class->size);
 }
 EXPORT_SYMBOL_GPL(zs_map_object);
 
@@ -801,17 +892,6 @@ void zs_unmap_object(struct zs_pool *pool, unsigned long handle)
 	struct size_class *class;
 	struct mapping_area *area;
 
-	area = &__get_cpu_var(zs_map_area);
-	/* single-page object fastpath */
-	if (area->vm_addr) {
-		kunmap_atomic(area->vm_addr);
-		goto out;
-	}
-
-	/* no write fastpath */
-	if (area->vm_mm == ZS_MM_RO)
-		goto pfenable;
-
 	BUG_ON(!handle);
 
 	obj_handle_to_location(handle, &page, &obj_idx);
@@ -819,12 +899,18 @@ void zs_unmap_object(struct zs_pool *pool, unsigned long handle)
 	class = &pool->size_class[class_idx];
 	off = obj_idx_to_offset(page, obj_idx, class->size);
 
-	zs_copy_unmap_object(area->vm_buf, page, off, class->size);
+	area = &__get_cpu_var(zs_map_area);
+	if (off + class->size <= PAGE_SIZE)
+		kunmap_atomic(area->vm_addr);
+	else {
+		struct page *pages[2];
+
+		pages[0] = page;
+		pages[1] = get_next_page(page);
+		BUG_ON(!pages[1]);
 
-pfenable:
-	/* enable page faults to match kunmap_atomic() return conditions */
-	pagefault_enable();
-out:
+		__zs_unmap_object(area, pages, off, class->size);
+	}
 	put_cpu_var(zs_map_area);
 }
 EXPORT_SYMBOL_GPL(zs_unmap_object);
diff --git a/drivers/staging/zsmalloc/zsmalloc_int.h b/drivers/staging/zsmalloc/zsmalloc_int.h
index 52805176..8c0b344 100644
--- a/drivers/staging/zsmalloc/zsmalloc_int.h
+++ b/drivers/staging/zsmalloc/zsmalloc_int.h
@@ -109,12 +109,6 @@ enum fullness_group {
  */
 static const int fullness_threshold_frac = 4;
 
-struct mapping_area {
-	char *vm_buf; /* copy buffer for objects that span pages */
-	char *vm_addr; /* address of kmap_atomic()'ed pages */
-	enum zs_mapmode vm_mm; /* mapping mode */
-};
-
 struct size_class {
 	/*
 	 * Size of objects stored in this class. Must be multiple
-- 
1.7.9.5

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 3/3] zsmalloc: add page table mapping method

[Minchan Kim]

On Wed, Jul 18, 2012 at 11:55:56AM -0500, Seth Jennings wrote:
> This patchset provides page mapping via the page table.
> On some archs, most notably ARM, this method has been
> demonstrated to be faster than copying.
> 
> The logic controlling the method selection (copy vs page table)
> is controlled by the definition of USE_PGTABLE_MAPPING which
> is/can be defined for any arch that performs better with page
> table mapping.
> 
> Signed-off-by: Seth Jennings <sjenning@linux.vnet.ibm.com>
> ---
>  drivers/staging/zsmalloc/zsmalloc-main.c |  182 ++++++++++++++++++++++--------
>  drivers/staging/zsmalloc/zsmalloc_int.h  |    6 -
>  2 files changed, 134 insertions(+), 54 deletions(-)
> 
> diff --git a/drivers/staging/zsmalloc/zsmalloc-main.c b/drivers/staging/zsmalloc/zsmalloc-main.c
> index b86133f..defe350 100644
> --- a/drivers/staging/zsmalloc/zsmalloc-main.c
> +++ b/drivers/staging/zsmalloc/zsmalloc-main.c
> @@ -89,6 +89,30 @@
>  #define CLASS_IDX_MASK	((1 << CLASS_IDX_BITS) - 1)
>  #define FULLNESS_MASK	((1 << FULLNESS_BITS) - 1)
>  
> +/*
> + * By default, zsmalloc uses a copy-based object mapping method to access
> + * allocations that span two pages. However, if a particular architecture
> + * 1) Implements local_flush_tlb_kernel_range() and 2) Performs VM mapping
> + * faster than copying, then it should be added here so that

How about adding your benchmark url?

> + * USE_PGTABLE_MAPPING is defined. This causes zsmalloc to use page table
> + * mapping rather than copying
> + * for object mapping.

unnecessary new line.

> +*/
> +#if defined(CONFIG_ARM)
> +#define USE_PGTABLE_MAPPING
> +#endif

I had no better idea and I would like to add zsmalloc into mainline.
So no objection.
Nitin?

> +
> +struct mapping_area {
> +#ifdef USE_PGTABLE_MAPPING
> +	struct vm_struct *vm; /* vm area for mapping object that span pages */
> +#else
> +	char *vm_buf; /* copy buffer for objects that span pages */
> +#endif
> +	char *vm_addr; /* address of kmap_atomic()'ed pages */
> +	enum zs_mapmode vm_mm; /* mapping mode */
> +};
> +
> +
>  /* per-cpu VM mapping areas for zspage accesses that cross page boundaries */
>  static DEFINE_PER_CPU(struct mapping_area, zs_map_area);
>  
> @@ -471,16 +495,83 @@ static struct page *find_get_zspage(struct size_class *class)
>  	return page;
>  }
>  
> -static void zs_copy_map_object(char *buf, struct page *page,
> -				int off, int size)
> +#ifdef USE_PGTABLE_MAPPING
> +static inline int __zs_cpu_up(struct mapping_area *area)
> +{
> +	/*
> +	 * Make sure we don't leak memory if a cpu UP notification
> +	 * and zs_init() race and both call zs_cpu_up() on the same cpu
> +	 */
> +	if (area->vm)
> +		return 0;
> +	area->vm = alloc_vm_area(PAGE_SIZE * 2, NULL);
> +	if (!area->vm)
> +		return -ENOMEM;
> +	return 0;
> +}
> +
> +static inline void __zs_cpu_down(struct mapping_area *area)
> +{
> +	if (area->vm)
> +		free_vm_area(area->vm);
> +	area->vm = NULL;
> +}
> +
> +static inline void *__zs_map_object(struct mapping_area *area,
> +				struct page *pages[2], int off, int size)
> +{
> +	BUG_ON(map_vm_area(area->vm, PAGE_KERNEL, &pages));
> +	area->vm_addr = area->vm->addr;
> +	return area->vm_addr + off;
> +}
> +
> +static inline void __zs_unmap_object(struct mapping_area *area,
> +				struct page *pages[2], int off, int size)
> +{
> +	unsigned long addr = (unsigned long)area->vm_addr;
> +	unsigned long end = addr + (PAGE_SIZE * 2);
> +
> +	flush_cache_vunmap(addr, end);
> +	unmap_kernel_range_noflush(addr, PAGE_SIZE * 2);
> +	local_flush_tlb_kernel_range(addr, end);
> +}
> +
> +#else /* USE_PGTABLE_MAPPING */
> +
> +static inline int __zs_cpu_up(struct mapping_area *area)
> +{
> +	/*
> +	 * Make sure we don't leak memory if a cpu UP notification
> +	 * and zs_init() race and both call zs_cpu_up() on the same cpu
> +	 */
> +	if (area->vm_buf)
> +		return 0;
> +	area->vm_buf = (char *)__get_free_page(GFP_KERNEL);
> +	if (!area->vm_buf)
> +		return -ENOMEM;
> +	return 0;
> +}
> +
> +static inline void __zs_cpu_down(struct mapping_area *area)
> +{
> +	if (area->vm_buf)
> +		free_page((unsigned long)area->vm_buf);
> +	area->vm_buf = NULL;
> +}
> +
> +static void *__zs_map_object(struct mapping_area *area,
> +			struct page *pages[2], int off, int size)
>  {
> -	struct page *pages[2];
>  	int sizes[2];
>  	void *addr;
> +	char *buf = area->vm_buf;
>  
> -	pages[0] = page;
> -	pages[1] = get_next_page(page);
> -	BUG_ON(!pages[1]);
> +	/* disable page faults to match kmap_atomic() return conditions */
> +	pagefault_disable();
> +
> +	/* no read fastpath */
> +	if (area->vm_mm == ZS_MM_WO)
> +		goto out;
>  
>  	sizes[0] = PAGE_SIZE - off;
>  	sizes[1] = size - sizes[0];
> @@ -492,18 +583,20 @@ static void zs_copy_map_object(char *buf, struct page *page,
>  	addr = kmap_atomic(pages[1]);
>  	memcpy(buf + sizes[0], addr, sizes[1]);
>  	kunmap_atomic(addr);
> +out:
> +	return area->vm_buf;
>  }
>  
> -static void zs_copy_unmap_object(char *buf, struct page *page,
> -				int off, int size)
> +static void __zs_unmap_object(struct mapping_area *area,
> +			struct page *pages[2], int off, int size)
>  {
> -	struct page *pages[2];
>  	int sizes[2];
>  	void *addr;
> +	char *buf = area->vm_buf;
>  
> -	pages[0] = page;
> -	pages[1] = get_next_page(page);
> -	BUG_ON(!pages[1]);
> +	/* no write fastpath */
> +	if (area->vm_mm == ZS_MM_RO)
> +		goto out;
>  
>  	sizes[0] = PAGE_SIZE - off;
>  	sizes[1] = size - sizes[0];
> @@ -515,34 +608,31 @@ static void zs_copy_unmap_object(char *buf, struct page *page,
>  	addr = kmap_atomic(pages[1]);
>  	memcpy(addr, buf + sizes[0], sizes[1]);
>  	kunmap_atomic(addr);
> +
> +out:
> +	/* enable page faults to match kunmap_atomic() return conditions */
> +	pagefault_enable();
>  }
>  
> +#endif /* USE_PGTABLE_MAPPING */
> +
>  static int zs_cpu_notifier(struct notifier_block *nb, unsigned long action,
>  				void *pcpu)
>  {
> -	int cpu = (long)pcpu;
> +	int ret, cpu = (long)pcpu;
>  	struct mapping_area *area;
>  
>  	switch (action) {
>  	case CPU_UP_PREPARE:
>  		area = &per_cpu(zs_map_area, cpu);
> -		/*
> -		 * Make sure we don't leak memory if a cpu UP notification
> -		 * and zs_init() race and both call zs_cpu_up() on the same cpu
> -		 */
> -		if (area->vm_buf)
> -			return 0;
> -		area->vm_buf = (char *)__get_free_page(GFP_KERNEL);
> -		if (!area->vm_buf)
> -			return -ENOMEM;
> -		return 0;
> +		ret = __zs_cpu_up(area);
> +		if (ret)
> +			return notifier_from_errno(ret);
>  		break;
>  	case CPU_DEAD:
>  	case CPU_UP_CANCELED:
>  		area = &per_cpu(zs_map_area, cpu);
> -		if (area->vm_buf)
> -			free_page((unsigned long)area->vm_buf);
> -		area->vm_buf = NULL;
> +		__zs_cpu_down(area);
>  		break;
>  	}
>  
> @@ -759,6 +849,7 @@ void *zs_map_object(struct zs_pool *pool, unsigned long handle,
>  	enum fullness_group fg;
>  	struct size_class *class;
>  	struct mapping_area *area;
> +	struct page *pages[2];
>  
>  	BUG_ON(!handle);
>  
> @@ -775,19 +866,19 @@ void *zs_map_object(struct zs_pool *pool, unsigned long handle,
>  	off = obj_idx_to_offset(page, obj_idx, class->size);
>  
>  	area = &get_cpu_var(zs_map_area);
> +	area->vm_mm = mm;
>  	if (off + class->size <= PAGE_SIZE) {
>  		/* this object is contained entirely within a page */
>  		area->vm_addr = kmap_atomic(page);
>  		return area->vm_addr + off;
>  	}
>  
> -	/* disable page faults to match kmap_atomic() return conditions */
> -	pagefault_disable();
> +	/* this object spans two pages */
> +	pages[0] = page;
> +	pages[1] = get_next_page(page);
> +	BUG_ON(!pages[1]);
>  
> -	if (mm != ZS_MM_WO)
> -		zs_copy_map_object(area->vm_buf, page, off, class->size);
> -	area->vm_addr = NULL;
> -	return area->vm_buf;
> +	return __zs_map_object(area, pages, off, class->size);
>  }
>  EXPORT_SYMBOL_GPL(zs_map_object);
>  
> @@ -801,17 +892,6 @@ void zs_unmap_object(struct zs_pool *pool, unsigned long handle)
>  	struct size_class *class;
>  	struct mapping_area *area;
>  
> -	area = &__get_cpu_var(zs_map_area);
> -	/* single-page object fastpath */
> -	if (area->vm_addr) {
> -		kunmap_atomic(area->vm_addr);
> -		goto out;
> -	}
> -
> -	/* no write fastpath */
> -	if (area->vm_mm == ZS_MM_RO)
> -		goto pfenable;
> -
>  	BUG_ON(!handle);
>  
>  	obj_handle_to_location(handle, &page, &obj_idx);
> @@ -819,12 +899,18 @@ void zs_unmap_object(struct zs_pool *pool, unsigned long handle)
>  	class = &pool->size_class[class_idx];
>  	off = obj_idx_to_offset(page, obj_idx, class->size);
>  
> -	zs_copy_unmap_object(area->vm_buf, page, off, class->size);
> +	area = &__get_cpu_var(zs_map_area);
> +	if (off + class->size <= PAGE_SIZE)
> +		kunmap_atomic(area->vm_addr);
> +	else {
> +		struct page *pages[2];
> +
> +		pages[0] = page;
> +		pages[1] = get_next_page(page);
> +		BUG_ON(!pages[1]);
>  
> -pfenable:
> -	/* enable page faults to match kunmap_atomic() return conditions */
> -	pagefault_enable();
> -out:
> +		__zs_unmap_object(area, pages, off, class->size);
> +	}
>  	put_cpu_var(zs_map_area);
>  }
>  EXPORT_SYMBOL_GPL(zs_unmap_object);
> diff --git a/drivers/staging/zsmalloc/zsmalloc_int.h b/drivers/staging/zsmalloc/zsmalloc_int.h
> index 52805176..8c0b344 100644
> --- a/drivers/staging/zsmalloc/zsmalloc_int.h
> +++ b/drivers/staging/zsmalloc/zsmalloc_int.h
> @@ -109,12 +109,6 @@ enum fullness_group {
>   */
>  static const int fullness_threshold_frac = 4;
>  
> -struct mapping_area {
> -	char *vm_buf; /* copy buffer for objects that span pages */
> -	char *vm_addr; /* address of kmap_atomic()'ed pages */
> -	enum zs_mapmode vm_mm; /* mapping mode */
> -};
> -
>  struct size_class {
>  	/*
>  	 * Size of objects stored in this class. Must be multiple
> -- 
> 1.7.9.5
> 
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to majordomo@kvack.org.  For more info on Linux MM,
> see: http://www.linux-mm.org/ .
> Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

-- 
Kind regards,
Minchan Kim

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 3/3] zsmalloc: add page table mapping method

[Seth Jennings]

On 07/22/2012 07:26 PM, Minchan Kim wrote:
> On Wed, Jul 18, 2012 at 11:55:56AM -0500, Seth Jennings wrote:
>> This patchset provides page mapping via the page table.
>> On some archs, most notably ARM, this method has been
>> demonstrated to be faster than copying.
>>
>> The logic controlling the method selection (copy vs page table)
>> is controlled by the definition of USE_PGTABLE_MAPPING which
>> is/can be defined for any arch that performs better with page
>> table mapping.
>>
>> Signed-off-by: Seth Jennings <sjenning@linux.vnet.ibm.com>
>> ---
>>  drivers/staging/zsmalloc/zsmalloc-main.c |  182 ++++++++++++++++++++++--------
>>  drivers/staging/zsmalloc/zsmalloc_int.h  |    6 -
>>  2 files changed, 134 insertions(+), 54 deletions(-)
>>
>> diff --git a/drivers/staging/zsmalloc/zsmalloc-main.c b/drivers/staging/zsmalloc/zsmalloc-main.c
>> index b86133f..defe350 100644
>> --- a/drivers/staging/zsmalloc/zsmalloc-main.c
>> +++ b/drivers/staging/zsmalloc/zsmalloc-main.c
>> @@ -89,6 +89,30 @@
>>  #define CLASS_IDX_MASK	((1 << CLASS_IDX_BITS) - 1)
>>  #define FULLNESS_MASK	((1 << FULLNESS_BITS) - 1)
>>  
>> +/*
>> + * By default, zsmalloc uses a copy-based object mapping method to access
>> + * allocations that span two pages. However, if a particular architecture
>> + * 1) Implements local_flush_tlb_kernel_range() and 2) Performs VM mapping
>> + * faster than copying, then it should be added here so that
> 
> How about adding your benchmark url?
> 
>> + * USE_PGTABLE_MAPPING is defined. This causes zsmalloc to use page table
>> + * mapping rather than copying
>> + * for object mapping.
> 
> unnecessary new line.

Since these aren't functional issues with the code, if I
_promise_ to send a follow-up patch to address these, can I
get your Ack?

--
Seth

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 3/3] zsmalloc: add page table mapping method

[Minchan Kim]

Hi Seth,

On Sun, Jul 22, 2012 at 07:33:40PM -0500, Seth Jennings wrote:
> On 07/22/2012 07:26 PM, Minchan Kim wrote:
> > On Wed, Jul 18, 2012 at 11:55:56AM -0500, Seth Jennings wrote:
> >> This patchset provides page mapping via the page table.
> >> On some archs, most notably ARM, this method has been
> >> demonstrated to be faster than copying.
> >>
> >> The logic controlling the method selection (copy vs page table)
> >> is controlled by the definition of USE_PGTABLE_MAPPING which
> >> is/can be defined for any arch that performs better with page
> >> table mapping.
> >>
> >> Signed-off-by: Seth Jennings <sjenning@linux.vnet.ibm.com>
> >> ---
> >>  drivers/staging/zsmalloc/zsmalloc-main.c |  182 ++++++++++++++++++++++--------
> >>  drivers/staging/zsmalloc/zsmalloc_int.h  |    6 -
> >>  2 files changed, 134 insertions(+), 54 deletions(-)
> >>
> >> diff --git a/drivers/staging/zsmalloc/zsmalloc-main.c b/drivers/staging/zsmalloc/zsmalloc-main.c
> >> index b86133f..defe350 100644
> >> --- a/drivers/staging/zsmalloc/zsmalloc-main.c
> >> +++ b/drivers/staging/zsmalloc/zsmalloc-main.c
> >> @@ -89,6 +89,30 @@
> >>  #define CLASS_IDX_MASK	((1 << CLASS_IDX_BITS) - 1)
> >>  #define FULLNESS_MASK	((1 << FULLNESS_BITS) - 1)
> >>  
> >> +/*
> >> + * By default, zsmalloc uses a copy-based object mapping method to access
> >> + * allocations that span two pages. However, if a particular architecture
> >> + * 1) Implements local_flush_tlb_kernel_range() and 2) Performs VM mapping
> >> + * faster than copying, then it should be added here so that
> > 
> > How about adding your benchmark url?
> > 
> >> + * USE_PGTABLE_MAPPING is defined. This causes zsmalloc to use page table
> >> + * mapping rather than copying
> >> + * for object mapping.
> > 
> > unnecessary new line.
> 
> Since these aren't functional issues with the code, if I
> _promise_ to send a follow-up patch to address these, can I
> get your Ack?

Sure!

Thanks for your effort!

-- 
Kind regards,
Minchan Kim

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 3/3] zsmalloc: add page table mapping method

[Nitin Gupta]

On 07/22/2012 08:26 PM, Minchan Kim wrote:
> On Wed, Jul 18, 2012 at 11:55:56AM -0500, Seth Jennings wrote:
>> This patchset provides page mapping via the page table.
>> On some archs, most notably ARM, this method has been
>> demonstrated to be faster than copying.
>>
>> The logic controlling the method selection (copy vs page table)
>> is controlled by the definition of USE_PGTABLE_MAPPING which
>> is/can be defined for any arch that performs better with page
>> table mapping.
>>
>> Signed-off-by: Seth Jennings <sjenning@linux.vnet.ibm.com>
>> ---
>>  drivers/staging/zsmalloc/zsmalloc-main.c |  182 ++++++++++++++++++++++--------
>>  drivers/staging/zsmalloc/zsmalloc_int.h  |    6 -
>>  2 files changed, 134 insertions(+), 54 deletions(-)
>>
>> diff --git a/drivers/staging/zsmalloc/zsmalloc-main.c b/drivers/staging/zsmalloc/zsmalloc-main.c
>> index b86133f..defe350 100644
>> --- a/drivers/staging/zsmalloc/zsmalloc-main.c
>> +++ b/drivers/staging/zsmalloc/zsmalloc-main.c
>> @@ -89,6 +89,30 @@
>>  #define CLASS_IDX_MASK	((1 << CLASS_IDX_BITS) - 1)
>>  #define FULLNESS_MASK	((1 << FULLNESS_BITS) - 1)
>>  
>> +/*
>> + * By default, zsmalloc uses a copy-based object mapping method to access
>> + * allocations that span two pages. However, if a particular architecture
>> + * 1) Implements local_flush_tlb_kernel_range() and 2) Performs VM mapping
>> + * faster than copying, then it should be added here so that
> 
> How about adding your benchmark url?
> 
>> + * USE_PGTABLE_MAPPING is defined. This causes zsmalloc to use page table
>> + * mapping rather than copying
>> + * for object mapping.
> 
> unnecessary new line.
> 
>> +*/
>> +#if defined(CONFIG_ARM)
>> +#define USE_PGTABLE_MAPPING
>> +#endif
> 
> I had no better idea and I would like to add zsmalloc into mainline.
> So no objection.
> Nitin?
> 

Same here. I just cannot think of anything better for now.

Thanks for your efforts.
Nitin


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 1/3] zsmalloc: s/firstpage/page in new copy map funcs

[Minchan Kim]

On Wed, Jul 18, 2012 at 11:55:54AM -0500, Seth Jennings wrote:
> firstpage already has precedent and meaning the first page
> of a zspage.  In the case of the copy mapping functions,
> it is the first of a pair of pages needing to be mapped.
> 
> This patch just renames the firstpage argument to "page" to
> avoid confusion.
> 
> Signed-off-by: Seth Jennings <sjenning@linux.vnet.ibm.com>
Acked-by: Minchan Kim <minchan@kernel.org>

-- 
Kind regards,
Minchan Kim

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 1/3] zsmalloc: s/firstpage/page in new copy map funcs

[Seth Jennings]

Greg,

I know it's the first Monday after a kernel release and
things are crazy for you.  I was hoping to get this zsmalloc
stuff in before the merge window hit so I wouldn't have to
bother you :-/  But, alas, it didn't happen that way.

Minchan acked these yesterday.  When you get a chance, could
you pull these 3 patches?  I'm wanting to send out a
promotion patch for zsmalloc and zcache based on these.

Thanks Greg!

--
Seth

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 1/3] zsmalloc: s/firstpage/page in new copy map funcs

[Greg Kroah-Hartman]

On Mon, Jul 23, 2012 at 05:10:39PM -0500, Seth Jennings wrote:
> Greg,
> 
> I know it's the first Monday after a kernel release and
> things are crazy for you.  I was hoping to get this zsmalloc
> stuff in before the merge window hit so I wouldn't have to
> bother you :-/  But, alas, it didn't happen that way.

Nope, sorry, it missed them.  It needed to be at least a week previous
to when the final kernel comes out to get into the next one.

> Minchan acked these yesterday.  When you get a chance, could
> you pull these 3 patches?  I'm wanting to send out a
> promotion patch for zsmalloc and zcache based on these.

Sorry, it will have to wait until after 3.6-rc1 is out before I will add
them to my tree for 3.7, that's the merge rules, that you well know :)

greg k-h

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 1/3] zsmalloc: s/firstpage/page in new copy map funcs

[Minchan Kim]

Hi Greg,

On Mon, Jul 23, 2012 at 03:27:49PM -0700, Greg Kroah-Hartman wrote:
> On Mon, Jul 23, 2012 at 05:10:39PM -0500, Seth Jennings wrote:
> > Greg,
> > 
> > I know it's the first Monday after a kernel release and
> > things are crazy for you.  I was hoping to get this zsmalloc
> > stuff in before the merge window hit so I wouldn't have to
> > bother you :-/  But, alas, it didn't happen that way.
> 
> Nope, sorry, it missed them.  It needed to be at least a week previous
> to when the final kernel comes out to get into the next one.
> 
> > Minchan acked these yesterday.  When you get a chance, could
> > you pull these 3 patches?  I'm wanting to send out a
> > promotion patch for zsmalloc and zcache based on these.
> 
> Sorry, it will have to wait until after 3.6-rc1 is out before I will add
> them to my tree for 3.7, that's the merge rules, that you well know :)

I think it is good time that zram/zsmalloc is out of staging because of
removing arch dependency and many clean up with some bug fix.
I hope it's out of staging in this chance.
If you have a concern about that, please let me know it.

Thanks!

-- 
Kind regards,
Minchan Kim

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 1/3] zsmalloc: s/firstpage/page in new copy map funcs

[Greg Kroah-Hartman]

On Tue, Jul 24, 2012 at 11:24:49AM +0900, Minchan Kim wrote:
> Hi Greg,
> 
> On Mon, Jul 23, 2012 at 03:27:49PM -0700, Greg Kroah-Hartman wrote:
> > On Mon, Jul 23, 2012 at 05:10:39PM -0500, Seth Jennings wrote:
> > > Greg,
> > > 
> > > I know it's the first Monday after a kernel release and
> > > things are crazy for you.  I was hoping to get this zsmalloc
> > > stuff in before the merge window hit so I wouldn't have to
> > > bother you :-/  But, alas, it didn't happen that way.
> > 
> > Nope, sorry, it missed them.  It needed to be at least a week previous
> > to when the final kernel comes out to get into the next one.
> > 
> > > Minchan acked these yesterday.  When you get a chance, could
> > > you pull these 3 patches?  I'm wanting to send out a
> > > promotion patch for zsmalloc and zcache based on these.
> > 
> > Sorry, it will have to wait until after 3.6-rc1 is out before I will add
> > them to my tree for 3.7, that's the merge rules, that you well know :)
> 
> I think it is good time that zram/zsmalloc is out of staging because of
> removing arch dependency and many clean up with some bug fix.
> I hope it's out of staging in this chance.
> If you have a concern about that, please let me know it.

That would have to wait until 3.7, moving it out is not for 3.6, sorry.

greg k-h

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 1/3] zsmalloc: s/firstpage/page in new copy map funcs

[Minchan Kim]

Hi Greg,

When do you merge this series?
Thanks.

On Wed, Jul 18, 2012 at 11:55:54AM -0500, Seth Jennings wrote:
> firstpage already has precedent and meaning the first page
> of a zspage.  In the case of the copy mapping functions,
> it is the first of a pair of pages needing to be mapped.
> 
> This patch just renames the firstpage argument to "page" to
> avoid confusion.
> 
> Signed-off-by: Seth Jennings <sjenning@linux.vnet.ibm.com>
> ---
>  drivers/staging/zsmalloc/zsmalloc-main.c |   12 ++++++------
>  1 file changed, 6 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/staging/zsmalloc/zsmalloc-main.c b/drivers/staging/zsmalloc/zsmalloc-main.c
> index 8b0bcb6..3c83c65 100644
> --- a/drivers/staging/zsmalloc/zsmalloc-main.c
> +++ b/drivers/staging/zsmalloc/zsmalloc-main.c
> @@ -470,15 +470,15 @@ static struct page *find_get_zspage(struct size_class *class)
>  	return page;
>  }
>  
> -static void zs_copy_map_object(char *buf, struct page *firstpage,
> +static void zs_copy_map_object(char *buf, struct page *page,
>  				int off, int size)
>  {
>  	struct page *pages[2];
>  	int sizes[2];
>  	void *addr;
>  
> -	pages[0] = firstpage;
> -	pages[1] = get_next_page(firstpage);
> +	pages[0] = page;
> +	pages[1] = get_next_page(page);
>  	BUG_ON(!pages[1]);
>  
>  	sizes[0] = PAGE_SIZE - off;
> @@ -493,15 +493,15 @@ static void zs_copy_map_object(char *buf, struct page *firstpage,
>  	kunmap_atomic(addr);
>  }
>  
> -static void zs_copy_unmap_object(char *buf, struct page *firstpage,
> +static void zs_copy_unmap_object(char *buf, struct page *page,
>  				int off, int size)
>  {
>  	struct page *pages[2];
>  	int sizes[2];
>  	void *addr;
>  
> -	pages[0] = firstpage;
> -	pages[1] = get_next_page(firstpage);
> +	pages[0] = page;
> +	pages[1] = get_next_page(page);
>  	BUG_ON(!pages[1]);
>  
>  	sizes[0] = PAGE_SIZE - off;
> -- 
> 1.7.9.5
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

-- 
Kind regards,
Minchan Kim

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>