[PATCH 0/2] RFC: iomap-based swap file activation

[PATCH 0/2] RFC: iomap-based swap file activation

[Aleksei Besogonov]

This patch set adds iomap-based swap file initialization, this allows
it to use files with holes on XFS for swap.

Aleksei Besogonov (2):
  fs: Add iomap_swap_activate
  xfs: add support for iomap-based swapfile activation

 fs/iomap.c            | 87 +++++++++++++++++++++++++++++++++++++++++++++++++++
 fs/xfs/xfs_aops.c     |  7 +++++
 include/linux/iomap.h |  5 +++
 mm/swapfile.c         |  1 +
 4 files changed, 100 insertions(+)

-- 
2.14.1

[PATCH 1/2] fs: Add iomap_swap_activate

[Aleksei Besogonov]

This commit adds an alternative to generic_swapfile_activate function
based on iomap family. The major driver for this is swap file support
on XFS, currently it doesn't work on files with holes due to deficiencies
in the old bmap() interface that is used by generic_swapfile_activate.
---
 fs/iomap.c            | 87 +++++++++++++++++++++++++++++++++++++++++++++++++++
 include/linux/iomap.h |  5 +++
 mm/swapfile.c         |  1 +
 3 files changed, 93 insertions(+)

diff --git a/fs/iomap.c b/fs/iomap.c
index afd163586aa0..9f0980637dbc 100644
--- a/fs/iomap.c
+++ b/fs/iomap.c
@@ -1089,3 +1089,90 @@ iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
 	return ret;
 }
 EXPORT_SYMBOL_GPL(iomap_dio_rw);
+
+struct iomap_swap_walker_context
+{
+    struct swap_info_struct *sis;
+    unsigned int page_no;
+    unsigned int nr_extents;
+    sector_t *span;
+};
+
+static loff_t
+iomap_swap_map(struct inode *inode, loff_t pos, loff_t length,
+		void *data, struct iomap *iomap)
+{
+	struct iomap_swap_walker_context *ctx = data;
+	unsigned int blocks_per_page = PAGE_SIZE >> inode->i_blkbits;
+	u64 num_pages;
+	u64 aligned_address;
+	sector_t disk_pos;
+	loff_t eff_length;
+	loff_t ret;
+
+	/*
+	 * Swap extents need to be PAGE_SIZE aligned on the disk
+	 */
+	aligned_address = ALIGN(iomap->addr, PAGE_SIZE);
+	eff_length = length - (aligned_address - iomap->addr);
+	BUG_ON(eff_length > length);
+	disk_pos = aligned_address >> inode->i_blkbits;
+
+	if (eff_length < 0)
+		return length; /* Continue probing */
+
+	num_pages = eff_length / PAGE_SIZE;
+
+	/* Can't add blocks less than 1 page */
+	if (num_pages == 0)
+		return length; /* Continue probing */
+
+	ret = add_swap_extent(ctx->sis, ctx->page_no, num_pages, disk_pos);
+	if (ret < 0)
+		return ret;
+
+	ctx->nr_extents += ret;
+	ctx->page_no += num_pages;
+	*(ctx->span) += num_pages * blocks_per_page;
+
+	return length; /* continue probing */
+}
+
+int
+iomap_swap_activate(struct swap_info_struct *sis, struct file *file,
+		sector_t *span, const struct iomap_ops *ops)
+{
+	struct iomap_swap_walker_context ctx = {
+		.sis = sis,
+		.page_no = 0,
+		.nr_extents = 0,
+		.span = span
+	};
+
+	struct inode * inode = file_inode(file);
+	loff_t len = i_size_read(inode);
+	loff_t pos = 0;
+	loff_t ret = -EINVAL;
+
+	while(len > 0) {
+		ret = iomap_apply(inode, pos, len, IOMAP_REPORT, ops,
+				  &ctx, iomap_swap_map);
+		if (ret <= 0)
+			break;
+
+		pos += ret;
+		len -= ret;
+	}
+
+	if (ret < 0) {
+		pr_err("Failed to activate the swap file\n");
+		return -EINVAL;
+	}
+
+	sis->max = ctx.page_no == 0 ? 1 : ctx.page_no;
+	sis->pages = ctx.page_no;
+	sis->highest_bit = ctx.page_no;
+
+	return ctx.nr_extents;
+}
+EXPORT_SYMBOL_GPL(iomap_swap_activate);
diff --git a/include/linux/iomap.h b/include/linux/iomap.h
index 19a07de28212..a85da0db6c2e 100644
--- a/include/linux/iomap.h
+++ b/include/linux/iomap.h
@@ -10,6 +10,8 @@ struct iov_iter;
 struct kiocb;
 struct vm_area_struct;
 struct vm_fault;
+struct file;
+struct swap_info_struct;
 
 /*
  * Types of block ranges for iomap mappings:
@@ -106,4 +108,7 @@ typedef int (iomap_dio_end_io_t)(struct kiocb *iocb, ssize_t ret,
 ssize_t iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
 		const struct iomap_ops *ops, iomap_dio_end_io_t end_io);
 
+int iomap_swap_activate(struct swap_info_struct *sis, struct file *file,
+			sector_t *span, const struct iomap_ops *ops);
+
 #endif /* LINUX_IOMAP_H */
diff --git a/mm/swapfile.c b/mm/swapfile.c
index c7a33717d079..9e6d3c1a882e 100644
--- a/mm/swapfile.c
+++ b/mm/swapfile.c
@@ -2381,6 +2381,7 @@ add_swap_extent(struct swap_info_struct *sis, unsigned long start_page,
 	list_add_tail(&new_se->list, &sis->first_swap_extent.list);
 	return 1;
 }
+EXPORT_SYMBOL_GPL(add_swap_extent);
 
 /*
  * A `swap extent' is a simple thing which maps a contiguous range of pages
-- 
2.14.1

[PATCH 2/2] xfs: add support for iomap-based swapfile activation

[Aleksei Besogonov]

Use iomap-based swapfile activation instead of the default
generic_swapfile_activate bmap-based infrastructure.
---
 fs/xfs/xfs_aops.c | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c
index 31f1f10eecd1..878ff3872883 100644
--- a/fs/xfs/xfs_aops.c
+++ b/fs/xfs/xfs_aops.c
@@ -1491,6 +1491,12 @@ xfs_vm_set_page_dirty(
 	return newly_dirty;
 }
 
+static int xfs_swap_activate(struct swap_info_struct *sis, struct file *file,
+			     sector_t *span)
+{
+	return iomap_swap_activate(sis, file, span, &xfs_iomap_ops);
+}
+
 const struct address_space_operations xfs_address_space_operations = {
 	.readpage		= xfs_vm_readpage,
 	.readpages		= xfs_vm_readpages,
@@ -1504,4 +1510,5 @@ const struct address_space_operations xfs_address_space_operations = {
 	.migratepage		= buffer_migrate_page,
 	.is_partially_uptodate  = block_is_partially_uptodate,
 	.error_remove_page	= generic_error_remove_page,
+	.swap_activate          = xfs_swap_activate,
 };
-- 
2.14.1

Re: [PATCH 0/2] RFC: iomap-based swap file activation

[Darrick J. Wong]

On Fri, Apr 20, 2018 at 10:08:13PM +0000, Aleksei Besogonov wrote:
> This patch set adds iomap-based swap file initialization, this allows
> it to use files with holes on XFS for swap.

Um... there's already a patch[1] implementing this out for review on the
mailing list.  Could you please take a look at that instead?

--D

[1] https://marc.info/?l=linux-xfs&m=152401984810062&w=2

> 
> Aleksei Besogonov (2):
>   fs: Add iomap_swap_activate
>   xfs: add support for iomap-based swapfile activation
> 
>  fs/iomap.c            | 87 +++++++++++++++++++++++++++++++++++++++++++++++++++
>  fs/xfs/xfs_aops.c     |  7 +++++
>  include/linux/iomap.h |  5 +++
>  mm/swapfile.c         |  1 +
>  4 files changed, 100 insertions(+)
> 
> -- 
> 2.14.1
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 0/2] RFC: iomap-based swap file activation

[Besogonov, Aleksei]

Sorry for top-posting.

I'm sorry I missed your patch. It's definitely better than mine in the number
of validations it does.

My only comment is, shouldn't we allow blocks that are not page-aligned?
The current generic_swapfile_activate will simply discard the leading unaligned
sectors. Your patch will immediately error out in this case.

I tried to replicate the same behavior by doing ALIGN on the iomap addr.

On 4/20/18, 15:26, "Darrick J. Wong" <darrick.wong@oracle.com> wrote:

    On Fri, Apr 20, 2018 at 10:08:13PM +0000, Aleksei Besogonov wrote:
    > This patch set adds iomap-based swap file initialization, this allows
    > it to use files with holes on XFS for swap.
    
    Um... there's already a patch[1] implementing this out for review on the
    mailing list.  Could you please take a look at that instead?
    
    --D
    
    [1] https://marc.info/?l=linux-xfs&m=152401984810062&w=2
    
    > 
    > Aleksei Besogonov (2):
    >   fs: Add iomap_swap_activate
    >   xfs: add support for iomap-based swapfile activation
    > 
    >  fs/iomap.c            | 87 +++++++++++++++++++++++++++++++++++++++++++++++++++
    >  fs/xfs/xfs_aops.c     |  7 +++++
    >  include/linux/iomap.h |  5 +++
    >  mm/swapfile.c         |  1 +
    >  4 files changed, 100 insertions(+)
    > 
    > -- 
    > 2.14.1
    > 
    > --
    > To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
    > the body of a message to majordomo@vger.kernel.org
    > More majordomo info at  http://vger.kernel.org/majordomo-info.html
    

Re: [PATCH 1/2] fs: Add iomap_swap_activate

[kbuild test robot]

[-- Attachment #1: Type: text/plain, Size: 2483 bytes --]

Hi Aleksei,

Thank you for the patch! Yet something to improve:

[auto build test ERROR on linus/master]
[also build test ERROR on v4.17-rc2 next-20180424]
[cannot apply to dgc-xfs/for-next]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]

url:    https://github.com/0day-ci/linux/commits/Aleksei-Besogonov/fs-Add-iomap_swap_activate/20180423-061348
config: x86_64-randconfig-s4-04260252 (attached as .config)
compiler: gcc-7 (Debian 7.3.0-16) 7.3.0
reproduce:
        # save the attached .config to linux build tree
        make ARCH=x86_64 

All errors (new ones prefixed by >>):

   fs/iomap.c: In function 'iomap_swap_map':
>> fs/iomap.c:1130:8: error: implicit declaration of function 'add_swap_extent'; did you mean '__swap_count'? [-Werror=implicit-function-declaration]
     ret = add_swap_extent(ctx->sis, ctx->page_no, num_pages, disk_pos);
           ^~~~~~~~~~~~~~~
           __swap_count
   cc1: some warnings being treated as errors

vim +1130 fs/iomap.c

  1100	
  1101	static loff_t
  1102	iomap_swap_map(struct inode *inode, loff_t pos, loff_t length,
  1103			void *data, struct iomap *iomap)
  1104	{
  1105		struct iomap_swap_walker_context *ctx = data;
  1106		unsigned int blocks_per_page = PAGE_SIZE >> inode->i_blkbits;
  1107		u64 num_pages;
  1108		u64 aligned_address;
  1109		sector_t disk_pos;
  1110		loff_t eff_length;
  1111		loff_t ret;
  1112	
  1113		/*
  1114		 * Swap extents need to be PAGE_SIZE aligned on the disk
  1115		 */
  1116		aligned_address = ALIGN(iomap->addr, PAGE_SIZE);
  1117		eff_length = length - (aligned_address - iomap->addr);
  1118		BUG_ON(eff_length > length);
  1119		disk_pos = aligned_address >> inode->i_blkbits;
  1120	
  1121		if (eff_length < 0)
  1122			return length; /* Continue probing */
  1123	
  1124		num_pages = eff_length / PAGE_SIZE;
  1125	
  1126		/* Can't add blocks less than 1 page */
  1127		if (num_pages == 0)
  1128			return length; /* Continue probing */
  1129	
> 1130		ret = add_swap_extent(ctx->sis, ctx->page_no, num_pages, disk_pos);
  1131		if (ret < 0)
  1132			return ret;
  1133	
  1134		ctx->nr_extents += ret;
  1135		ctx->page_no += num_pages;
  1136		*(ctx->span) += num_pages * blocks_per_page;
  1137	
  1138		return length; /* continue probing */
  1139	}
  1140	

---
0-DAY kernel test infrastructure                Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all                   Intel Corporation

[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 34283 bytes --]

Re: [PATCH 1/2] fs: Add iomap_swap_activate

[kbuild test robot]

[-- Attachment #1: Type: text/plain, Size: 2455 bytes --]

Hi Aleksei,

Thank you for the patch! Yet something to improve:

[auto build test ERROR on linus/master]
[also build test ERROR on v4.17-rc2 next-20180424]
[cannot apply to dgc-xfs/for-next]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]

url:    https://github.com/0day-ci/linux/commits/Aleksei-Besogonov/fs-Add-iomap_swap_activate/20180423-061348
config: x86_64-randconfig-s2-04260250 (attached as .config)
compiler: gcc-6 (Debian 6.4.0-9) 6.4.0 20171026
reproduce:
        # save the attached .config to linux build tree
        make ARCH=x86_64 

All errors (new ones prefixed by >>):

   fs/iomap.c: In function 'iomap_swap_map':
>> fs/iomap.c:1130:8: error: implicit declaration of function 'add_swap_extent' [-Werror=implicit-function-declaration]
     ret = add_swap_extent(ctx->sis, ctx->page_no, num_pages, disk_pos);
           ^~~~~~~~~~~~~~~
   cc1: some warnings being treated as errors

vim +/add_swap_extent +1130 fs/iomap.c

  1100	
  1101	static loff_t
  1102	iomap_swap_map(struct inode *inode, loff_t pos, loff_t length,
  1103			void *data, struct iomap *iomap)
  1104	{
  1105		struct iomap_swap_walker_context *ctx = data;
  1106		unsigned int blocks_per_page = PAGE_SIZE >> inode->i_blkbits;
  1107		u64 num_pages;
  1108		u64 aligned_address;
  1109		sector_t disk_pos;
  1110		loff_t eff_length;
  1111		loff_t ret;
  1112	
  1113		/*
  1114		 * Swap extents need to be PAGE_SIZE aligned on the disk
  1115		 */
  1116		aligned_address = ALIGN(iomap->addr, PAGE_SIZE);
  1117		eff_length = length - (aligned_address - iomap->addr);
  1118		BUG_ON(eff_length > length);
  1119		disk_pos = aligned_address >> inode->i_blkbits;
  1120	
  1121		if (eff_length < 0)
  1122			return length; /* Continue probing */
  1123	
  1124		num_pages = eff_length / PAGE_SIZE;
  1125	
  1126		/* Can't add blocks less than 1 page */
  1127		if (num_pages == 0)
  1128			return length; /* Continue probing */
  1129	
> 1130		ret = add_swap_extent(ctx->sis, ctx->page_no, num_pages, disk_pos);
  1131		if (ret < 0)
  1132			return ret;
  1133	
  1134		ctx->nr_extents += ret;
  1135		ctx->page_no += num_pages;
  1136		*(ctx->span) += num_pages * blocks_per_page;
  1137	
  1138		return length; /* continue probing */
  1139	}
  1140	

---
0-DAY kernel test infrastructure                Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all                   Intel Corporation

[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 32387 bytes --]

Re: [PATCH 0/2] RFC: iomap-based swap file activation

[Darrick J. Wong]

On Fri, Apr 20, 2018 at 10:08:13PM +0000, Aleksei Besogonov wrote:
> This patch set adds iomap-based swap file initialization, this allows
> it to use files with holes on XFS for swap.

Just out of curiosity (well after the fact, sorry...), what's the
motivation for having sparse swap files?

--D

> Aleksei Besogonov (2):
>   fs: Add iomap_swap_activate
>   xfs: add support for iomap-based swapfile activation
> 
>  fs/iomap.c            | 87 +++++++++++++++++++++++++++++++++++++++++++++++++++
>  fs/xfs/xfs_aops.c     |  7 +++++
>  include/linux/iomap.h |  5 +++
>  mm/swapfile.c         |  1 +
>  4 files changed, 100 insertions(+)
> 
> -- 
> 2.14.1
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 1/2] fs: Add iomap_swap_activate

[kbuild test robot]

[-- Attachment #1: Type: text/plain, Size: 2973 bytes --]

Hi Aleksei,

Thank you for the patch! Yet something to improve:

[auto build test ERROR on linus/master]
[also build test ERROR on v4.17-rc3 next-20180501]
[cannot apply to dgc-xfs/for-next]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]

url:    https://github.com/0day-ci/linux/commits/Aleksei-Besogonov/fs-Add-iomap_swap_activate/20180502-034017
config: i386-randconfig-a1-201817 (attached as .config)
compiler: gcc-4.9 (Debian 4.9.4-2) 4.9.4
reproduce:
        # save the attached .config to linux build tree
        make ARCH=i386 

All errors (new ones prefixed by >>):

   fs/iomap.c: In function 'iomap_swap_map':
>> fs/iomap.c:1142:2: error: implicit declaration of function 'add_swap_extent' [-Werror=implicit-function-declaration]
     ret = add_swap_extent(ctx->sis, ctx->page_no, num_pages, disk_pos);
     ^
   cc1: some warnings being treated as errors

vim +/add_swap_extent +1142 fs/iomap.c

  1101	
  1102	static loff_t iomap_swap_map(struct inode *inode, loff_t pos, loff_t length,
  1103				     void *data, struct iomap *iomap)
  1104	{
  1105		struct iomap_swap_walker_context *ctx = data;
  1106		unsigned int blocks_per_page = PAGE_SIZE >> inode->i_blkbits;
  1107		u64 num_pages;
  1108		u64 aligned_address;
  1109		sector_t disk_pos;
  1110		loff_t eff_length;
  1111		loff_t ret;
  1112	
  1113		/* Validate the extent */
  1114	
  1115		/* Only one bdev per swap file. */
  1116		if (iomap->bdev != ctx->sis->bdev)
  1117			goto err;
  1118		/* Only real or unwritten extents. */
  1119		if (iomap->type != IOMAP_MAPPED && iomap->type != IOMAP_UNWRITTEN)
  1120			goto err;
  1121		/* No uncommitted metadata or shared blocks or inline data. */
  1122		if (iomap->flags & (IOMAP_F_DIRTY | IOMAP_F_SHARED | IOMAP_F_DATA_INLINE))
  1123			goto err;
  1124	
  1125		/*
  1126		 * Swap extents need to be PAGE_SIZE aligned on the disk
  1127		 */
  1128		aligned_address = ALIGN(iomap->addr, PAGE_SIZE);
  1129		eff_length = length - (aligned_address - iomap->addr);
  1130		BUG_ON(eff_length > length);
  1131		disk_pos = aligned_address >> inode->i_blkbits;
  1132	
  1133		if (eff_length < 0)
  1134			return length; /* Continue probing */
  1135	
  1136		num_pages = eff_length / PAGE_SIZE;
  1137	
  1138		/* Can't add blocks less than 1 page */
  1139		if (num_pages == 0)
  1140			return length; /* Continue probing */
  1141	
> 1142		ret = add_swap_extent(ctx->sis, ctx->page_no, num_pages, disk_pos);
  1143		if (ret < 0)
  1144			return ret;
  1145	
  1146		ctx->nr_extents += ret;
  1147		ctx->page_no += num_pages;
  1148		(*ctx->span) += num_pages * blocks_per_page;
  1149	
  1150		return length; /* continue probing */
  1151	err:
  1152		pr_err("swapon: swapfile has holes\n");
  1153		return -EINVAL;
  1154	}
  1155	

---
0-DAY kernel test infrastructure                Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all                   Intel Corporation

[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 32248 bytes --]

Re: [PATCH 1/2] fs: Add iomap_swap_activate

[kbuild test robot]

[-- Attachment #1: Type: text/plain, Size: 3034 bytes --]

Hi Aleksei,

Thank you for the patch! Yet something to improve:

[auto build test ERROR on linus/master]
[also build test ERROR on v4.17-rc3 next-20180501]
[cannot apply to dgc-xfs/for-next]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]

url:    https://github.com/0day-ci/linux/commits/Aleksei-Besogonov/fs-Add-iomap_swap_activate/20180502-034017
config: x86_64-randconfig-x009-201817 (attached as .config)
compiler: gcc-7 (Debian 7.3.0-16) 7.3.0
reproduce:
        # save the attached .config to linux build tree
        make ARCH=x86_64 

All errors (new ones prefixed by >>):

   fs/iomap.c: In function 'iomap_swap_map':
>> fs/iomap.c:1142:8: error: implicit declaration of function 'add_swap_extent'; did you mean '__swap_count'? [-Werror=implicit-function-declaration]
     ret = add_swap_extent(ctx->sis, ctx->page_no, num_pages, disk_pos);
           ^~~~~~~~~~~~~~~
           __swap_count
   cc1: some warnings being treated as errors

vim +1142 fs/iomap.c

  1101	
  1102	static loff_t iomap_swap_map(struct inode *inode, loff_t pos, loff_t length,
  1103				     void *data, struct iomap *iomap)
  1104	{
  1105		struct iomap_swap_walker_context *ctx = data;
  1106		unsigned int blocks_per_page = PAGE_SIZE >> inode->i_blkbits;
  1107		u64 num_pages;
  1108		u64 aligned_address;
  1109		sector_t disk_pos;
  1110		loff_t eff_length;
  1111		loff_t ret;
  1112	
  1113		/* Validate the extent */
  1114	
  1115		/* Only one bdev per swap file. */
  1116		if (iomap->bdev != ctx->sis->bdev)
  1117			goto err;
  1118		/* Only real or unwritten extents. */
  1119		if (iomap->type != IOMAP_MAPPED && iomap->type != IOMAP_UNWRITTEN)
  1120			goto err;
  1121		/* No uncommitted metadata or shared blocks or inline data. */
  1122		if (iomap->flags & (IOMAP_F_DIRTY | IOMAP_F_SHARED | IOMAP_F_DATA_INLINE))
  1123			goto err;
  1124	
  1125		/*
  1126		 * Swap extents need to be PAGE_SIZE aligned on the disk
  1127		 */
  1128		aligned_address = ALIGN(iomap->addr, PAGE_SIZE);
  1129		eff_length = length - (aligned_address - iomap->addr);
  1130		BUG_ON(eff_length > length);
  1131		disk_pos = aligned_address >> inode->i_blkbits;
  1132	
  1133		if (eff_length < 0)
  1134			return length; /* Continue probing */
  1135	
  1136		num_pages = eff_length / PAGE_SIZE;
  1137	
  1138		/* Can't add blocks less than 1 page */
  1139		if (num_pages == 0)
  1140			return length; /* Continue probing */
  1141	
> 1142		ret = add_swap_extent(ctx->sis, ctx->page_no, num_pages, disk_pos);
  1143		if (ret < 0)
  1144			return ret;
  1145	
  1146		ctx->nr_extents += ret;
  1147		ctx->page_no += num_pages;
  1148		(*ctx->span) += num_pages * blocks_per_page;
  1149	
  1150		return length; /* continue probing */
  1151	err:
  1152		pr_err("swapon: swapfile has holes\n");
  1153		return -EINVAL;
  1154	}
  1155	

---
0-DAY kernel test infrastructure                Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all                   Intel Corporation

[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 27280 bytes --]

[PATCH 1/2] fs: Add iomap_swap_activate

[Aleksei Besogonov]

This commit adds an alternative to generic_swapfile_activate function
based on iomap family. The major driver for this is swap file support
on XFS, currently it doesn't work on files with holes due to deficiencies
in the old bmap() interface that is used by generic_swapfile_activate.
---
 fs/iomap.c            | 106 ++++++++++++++++++++++++++++++++++++++++++++++++++
 include/linux/iomap.h |   5 +++
 mm/swapfile.c         |   1 +
 3 files changed, 112 insertions(+)

diff --git a/fs/iomap.c b/fs/iomap.c
index afd163586aa0..8ac6dcbfbeb6 100644
--- a/fs/iomap.c
+++ b/fs/iomap.c
@@ -94,6 +94,7 @@ iomap_apply(struct inode *inode, loff_t pos, loff_t length, unsigned flags,
 
 	return written ? written : ret;
 }
+EXPORT_SYMBOL_GPL(iomap_apply);
 
 static void
 iomap_write_failed(struct inode *inode, loff_t pos, unsigned len)
@@ -1089,3 +1090,108 @@ iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
 	return ret;
 }
 EXPORT_SYMBOL_GPL(iomap_dio_rw);
+
+struct iomap_swap_walker_context
+{
+    struct swap_info_struct *sis;
+    unsigned int page_no;
+    unsigned int nr_extents;
+    sector_t *span;
+};
+
+static loff_t iomap_swap_map(struct inode *inode, loff_t pos, loff_t length,
+			     void *data, struct iomap *iomap)
+{
+	struct iomap_swap_walker_context *ctx = data;
+	unsigned int blocks_per_page = PAGE_SIZE >> inode->i_blkbits;
+	u64 num_pages;
+	u64 aligned_address;
+	sector_t disk_pos;
+	loff_t eff_length;
+	loff_t ret;
+
+	/* Validate the extent */
+
+	/* Only one bdev per swap file. */
+	if (iomap->bdev != ctx->sis->bdev)
+		goto err;
+	/* Only real or unwritten extents. */
+	if (iomap->type != IOMAP_MAPPED && iomap->type != IOMAP_UNWRITTEN)
+		goto err;
+	/* No uncommitted metadata or shared blocks or inline data. */
+	if (iomap->flags & (IOMAP_F_DIRTY | IOMAP_F_SHARED | IOMAP_F_DATA_INLINE))
+		goto err;
+
+	/*
+	 * Swap extents need to be PAGE_SIZE aligned on the disk
+	 */
+	aligned_address = ALIGN(iomap->addr, PAGE_SIZE);
+	eff_length = length - (aligned_address - iomap->addr);
+	BUG_ON(eff_length > length);
+	disk_pos = aligned_address >> inode->i_blkbits;
+
+	if (eff_length < 0)
+		return length; /* Continue probing */
+
+	num_pages = eff_length / PAGE_SIZE;
+
+	/* Can't add blocks less than 1 page */
+	if (num_pages == 0)
+		return length; /* Continue probing */
+
+	ret = add_swap_extent(ctx->sis, ctx->page_no, num_pages, disk_pos);
+	if (ret < 0)
+		return ret;
+
+	ctx->nr_extents += ret;
+	ctx->page_no += num_pages;
+	(*ctx->span) += num_pages * blocks_per_page;
+
+	return length; /* continue probing */
+err:
+	pr_err("swapon: swapfile has holes\n");
+	return -EINVAL;
+}
+
+int iomap_swap_activate(struct swap_info_struct *sis, struct file *file,
+			sector_t *span, const struct iomap_ops *ops)
+{
+	struct iomap_swap_walker_context ctx = {
+		.sis = sis,
+		.page_no = 0,
+		.nr_extents = 0,
+		.span = span
+	};
+
+	struct inode * inode = file_inode(file);
+	loff_t len = i_size_read(inode);
+	loff_t pos = 0;
+	loff_t ret = -EINVAL;
+
+
+	ret = filemap_write_and_wait(inode->i_mapping);
+	if (ret)
+		return ret;
+
+	while(len > 0) {
+		ret = iomap_apply(inode, pos, len, IOMAP_REPORT, ops,
+				  &ctx, iomap_swap_map);
+		if (ret <= 0)
+			break;
+
+		pos += ret;
+		len -= ret;
+	}
+
+	if (ret < 0) {
+		pr_err("Failed to activate the swap file\n");
+		return -EINVAL;
+	}
+
+	sis->max = ctx.page_no == 0 ? 1 : ctx.page_no;
+	sis->pages = ctx.page_no;
+	sis->highest_bit = ctx.page_no;
+
+	return ctx.nr_extents;
+}
+EXPORT_SYMBOL_GPL(iomap_swap_activate);
diff --git a/include/linux/iomap.h b/include/linux/iomap.h
index 19a07de28212..a85da0db6c2e 100644
--- a/include/linux/iomap.h
+++ b/include/linux/iomap.h
@@ -10,6 +10,8 @@ struct iov_iter;
 struct kiocb;
 struct vm_area_struct;
 struct vm_fault;
+struct file;
+struct swap_info_struct;
 
 /*
  * Types of block ranges for iomap mappings:
@@ -106,4 +108,7 @@ typedef int (iomap_dio_end_io_t)(struct kiocb *iocb, ssize_t ret,
 ssize_t iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
 		const struct iomap_ops *ops, iomap_dio_end_io_t end_io);
 
+int iomap_swap_activate(struct swap_info_struct *sis, struct file *file,
+			sector_t *span, const struct iomap_ops *ops);
+
 #endif /* LINUX_IOMAP_H */
diff --git a/mm/swapfile.c b/mm/swapfile.c
index c7a33717d079..9e6d3c1a882e 100644
--- a/mm/swapfile.c
+++ b/mm/swapfile.c
@@ -2381,6 +2381,7 @@ add_swap_extent(struct swap_info_struct *sis, unsigned long start_page,
 	list_add_tail(&new_se->list, &sis->first_swap_extent.list);
 	return 1;
 }
+EXPORT_SYMBOL_GPL(add_swap_extent);
 
 /*
  * A `swap extent' is a simple thing which maps a contiguous range of pages
-- 
2.14.1