Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1481148 > unrolled thread

[PATCH 0/2] do not squash mapping flags and gfp_mask together (was: Re: [PATCH -v2] mm: Don't use radix tree writeback tags for pages in)

Started byMichal Hocko <mhocko@kernel.org>
First post2016-09-12 13:20 +0200
Last post2016-09-13 23:40 +0200
Articles 8 — 2 participants

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  [PATCH 0/2] do not squash mapping flags and gfp_mask together (was: Re: [PATCH -v2] mm: Don't use radix tree writeback tags for pages in) Michal Hocko <mhocko@kernel.org> - 2016-09-12 13:20 +0200
    [PATCH 2/2] mm: split gfp_mask and mapping flags into separate fields Michal Hocko <mhocko@kernel.org> - 2016-09-12 13:20 +0200
      Re: [PATCH 2/2] mm: split gfp_mask and mapping flags into separate  fields Michal Hocko <mhocko@kernel.org> - 2016-09-12 13:50 +0200
    [PATCH 1/2] fs: use mapping_set_error instead of opencoded set_bit Michal Hocko <mhocko@kernel.org> - 2016-09-12 13:20 +0200
      Re: [PATCH 1/2] fs: use mapping_set_error instead of opencoded  set_bit Andrew Morton <akpm@linux-foundation.org> - 2016-09-13 00:20 +0200
        Re: [PATCH 1/2] fs: use mapping_set_error instead of opencoded  set_bit Andrew Morton <akpm@linux-foundation.org> - 2016-09-13 00:20 +0200
          Re: [PATCH 1/2] fs: use mapping_set_error instead of opencoded  set_bit Michal Hocko <mhocko@kernel.org> - 2016-09-13 09:00 +0200
            Re: [PATCH 1/2] fs: use mapping_set_error instead of opencoded  set_bit Andrew Morton <akpm@linux-foundation.org> - 2016-09-13 23:40 +0200

#1481148 — [PATCH 0/2] do not squash mapping flags and gfp_mask together (was: Re: [PATCH -v2] mm: Don't use radix tree writeback tags for pages in)

FromMichal Hocko <mhocko@kernel.org>
Date2016-09-12 13:20 +0200
Subject[PATCH 0/2] do not squash mapping flags and gfp_mask together (was: Re: [PATCH -v2] mm: Don't use radix tree writeback tags for pages in)
Message-ID<sgxrk-2vU-9@gated-at.bofh.it>
On Thu 01-09-16 11:13:47, Michal Hocko wrote:
> On Wed 31-08-16 14:30:31, Andrew Morton wrote:
> > On Wed, 31 Aug 2016 10:14:59 +0100 Mel Gorman <mgorman@techsingularity.net> wrote:
[...]
> > > I didn't see anything wrong with the patch but it's worth highlighting
> > > that this hunk means we are now out of GFP bits.
> > 
> > Well ugh.  What are we to do about that?
> 
> Can we simply give these AS_ flags their own word in mapping rather than
> squash them together with gfp flags and impose the restriction on the
> number of gfp flags. There was some demand for new gfp flags already and
> mapping flags were in the way.

OK, it seems this got unnoticed. What do you think about the following
two patches? I have only compile tested them and git grep suggests
nobody else should be relying on storing gfp_mask into flags directly.
So either I my grep-foo fools me or this should be safe. The two patches
will come as a reply to this email.

[toc] | [next] | [standalone]

#1481149 — [PATCH 2/2] mm: split gfp_mask and mapping flags into separate fields

FromMichal Hocko <mhocko@kernel.org>
Date2016-09-12 13:20 +0200
Subject[PATCH 2/2] mm: split gfp_mask and mapping flags into separate fields
Message-ID<sgxrk-2vU-7@gated-at.bofh.it>
In reply to#1481148
From: Michal Hocko <mhocko@suse.com>

mapping->flags currently encodes two different things into a single
flag. It contains sticky gfp_mask for page cache allocations and AS_
codes used to report errors/enospace and other states which are mapping
specific. Condensing the two semantically unrelated things saves few
bytes but it also complicates other things. For one thing the gfp flags
space is reduced and in fact we are already running out of available
bits. It can be assumed that more gfp flags will be necessary later on.

To not introduce the address_space grow (at least on x86_64) we can
stick it right after private_lock because we have a hole there.

struct address_space {
        struct inode *             host;                 /*     0     8 */
        struct radix_tree_root     page_tree;            /*     8    16 */
        spinlock_t                 tree_lock;            /*    24     4 */
        atomic_t                   i_mmap_writable;      /*    28     4 */
        struct rb_root             i_mmap;               /*    32     8 */
        struct rw_semaphore        i_mmap_rwsem;         /*    40    40 */
        /* --- cacheline 1 boundary (64 bytes) was 16 bytes ago --- */
        long unsigned int          nrpages;              /*    80     8 */
        long unsigned int          nrexceptional;        /*    88     8 */
        long unsigned int          writeback_index;      /*    96     8 */
        const struct address_space_operations  * a_ops;  /*   104     8 */
        long unsigned int          flags;                /*   112     8 */
        spinlock_t                 private_lock;         /*   120     4 */

        /* XXX 4 bytes hole, try to pack */

        /* --- cacheline 2 boundary (128 bytes) --- */
        struct list_head           private_list;         /*   128    16 */
        void *                     private_data;         /*   144     8 */

        /* size: 152, cachelines: 3, members: 14 */
        /* sum members: 148, holes: 1, sum holes: 4 */
        /* last cacheline: 24 bytes */
};

Signed-off-by: Michal Hocko <mhocko@suse.com>
---
 include/linux/fs.h      |  3 ++-
 include/linux/pagemap.h | 20 +++++++++-----------
 2 files changed, 11 insertions(+), 12 deletions(-)

diff --git a/include/linux/fs.h b/include/linux/fs.h
index cd8a5e1d5580..41d7213946af 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -443,7 +443,8 @@ struct address_space {
 	unsigned long		nrexceptional;
 	pgoff_t			writeback_index;/* writeback starts here */
 	const struct address_space_operations *a_ops;	/* methods */
-	unsigned long		flags;		/* error bits/gfp mask */
+	unsigned long		flags;		/* error bits */
+	gfp_t			gfp_mask;	/* implicit gfp mask for allocations */
 	spinlock_t		private_lock;	/* for use by the address_space */
 	struct list_head	private_list;	/* ditto */
 	void			*private_data;	/* ditto */
diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h
index 76f151ab4f62..0385a954465c 100644
--- a/include/linux/pagemap.h
+++ b/include/linux/pagemap.h
@@ -16,17 +16,16 @@
 #include <linux/hugetlb_inline.h>
 
 /*
- * Bits in mapping->flags.  The lower __GFP_BITS_SHIFT bits are the page
- * allocation mode flags.
+ * Bits in mapping->flags.
  */
 enum mapping_flags {
-	AS_EIO		= __GFP_BITS_SHIFT + 0,	/* IO error on async write */
-	AS_ENOSPC	= __GFP_BITS_SHIFT + 1,	/* ENOSPC on async write */
-	AS_MM_ALL_LOCKS	= __GFP_BITS_SHIFT + 2,	/* under mm_take_all_locks() */
-	AS_UNEVICTABLE	= __GFP_BITS_SHIFT + 3,	/* e.g., ramdisk, SHM_LOCK */
-	AS_EXITING	= __GFP_BITS_SHIFT + 4, /* final truncate in progress */
+	AS_EIO		= 0,	/* IO error on async write */
+	AS_ENOSPC	= 1,	/* ENOSPC on async write */
+	AS_MM_ALL_LOCKS	= 2,	/* under mm_take_all_locks() */
+	AS_UNEVICTABLE	= 3,	/* e.g., ramdisk, SHM_LOCK */
+	AS_EXITING	= 4, 	/* final truncate in progress */
 	/* writeback related tags are not used */
-	AS_NO_WRITEBACK_TAGS = __GFP_BITS_SHIFT + 5,
+	AS_NO_WRITEBACK_TAGS = 5,
 
 	AS_LAST_FLAG,
 };
@@ -80,7 +79,7 @@ static inline int mapping_use_writeback_tags(struct address_space *mapping)
 
 static inline gfp_t mapping_gfp_mask(struct address_space * mapping)
 {
-	return (__force gfp_t)mapping->flags & __GFP_BITS_MASK;
+	return mapping->gfp_mask;
 }
 
 /* Restricts the given gfp_mask to what the mapping allows. */
@@ -96,8 +95,7 @@ static inline gfp_t mapping_gfp_constraint(struct address_space *mapping,
  */
 static inline void mapping_set_gfp_mask(struct address_space *m, gfp_t mask)
 {
-	m->flags = (m->flags & ~(__force unsigned long)__GFP_BITS_MASK) |
-				(__force unsigned long)mask;
+	m->gfp_mask = mask;
 }
 
 void release_pages(struct page **pages, int nr, bool cold);
-- 
2.9.3

[toc] | [prev] | [next] | [standalone]

#1481165 — Re: [PATCH 2/2] mm: split gfp_mask and mapping flags into separate fields

FromMichal Hocko <mhocko@kernel.org>
Date2016-09-12 13:50 +0200
SubjectRe: [PATCH 2/2] mm: split gfp_mask and mapping flags into separate fields
Message-ID<sgxUm-2FS-1@gated-at.bofh.it>
In reply to#1481149
Errr, the gfp_mask move behind private_lock didn't make it into the
commit. Here is the updated patch. Btw. with this patch we can drop
mm-check-that-we-havent-used-more-than-32-bits-in-address_spaceflags.patch
---
From a8200e0de375886bbb41bbae4df7fa65ec619d05 Mon Sep 17 00:00:00 2001
From: Michal Hocko <mhocko@suse.com>
Date: Mon, 12 Sep 2016 12:50:09 +0200
Subject: [PATCH] mm: split gfp_mask and mapping flags into separate fields

mapping->flags currently encodes two different things into a single
flag. It contains sticky gfp_mask for page cache allocations and AS_
codes used to report errors/enospace and other states which are mapping
specific. Condensing the two semantically unrelated things saves few
bytes but it also complicates other things. For one thing the gfp flags
space is reduced and in fact we are already running out of available
bits. It can be assumed that more gfp flags will be necessary later on.

To not introduce the address_space grow (at least on x86_64) we can
stick it right after private_lock because we have a hole there.

struct address_space {
        struct inode *             host;                 /*     0     8 */
        struct radix_tree_root     page_tree;            /*     8    16 */
        spinlock_t                 tree_lock;            /*    24     4 */
        atomic_t                   i_mmap_writable;      /*    28     4 */
        struct rb_root             i_mmap;               /*    32     8 */
        struct rw_semaphore        i_mmap_rwsem;         /*    40    40 */
        /* --- cacheline 1 boundary (64 bytes) was 16 bytes ago --- */
        long unsigned int          nrpages;              /*    80     8 */
        long unsigned int          nrexceptional;        /*    88     8 */
        long unsigned int          writeback_index;      /*    96     8 */
        const struct address_space_operations  * a_ops;  /*   104     8 */
        long unsigned int          flags;                /*   112     8 */
        spinlock_t                 private_lock;         /*   120     4 */

        /* XXX 4 bytes hole, try to pack */

        /* --- cacheline 2 boundary (128 bytes) --- */
        struct list_head           private_list;         /*   128    16 */
        void *                     private_data;         /*   144     8 */

        /* size: 152, cachelines: 3, members: 14 */
        /* sum members: 148, holes: 1, sum holes: 4 */
        /* last cacheline: 24 bytes */
};

Signed-off-by: Michal Hocko <mhocko@suse.com>
---
 include/linux/fs.h      |  3 ++-
 include/linux/pagemap.h | 20 +++++++++-----------
 2 files changed, 11 insertions(+), 12 deletions(-)

diff --git a/include/linux/fs.h b/include/linux/fs.h
index cd8a5e1d5580..34b9d88cb788 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -443,8 +443,9 @@ struct address_space {
 	unsigned long		nrexceptional;
 	pgoff_t			writeback_index;/* writeback starts here */
 	const struct address_space_operations *a_ops;	/* methods */
-	unsigned long		flags;		/* error bits/gfp mask */
+	unsigned long		flags;		/* error bits */
 	spinlock_t		private_lock;	/* for use by the address_space */
+	gfp_t			gfp_mask;	/* implicit gfp mask for allocations */
 	struct list_head	private_list;	/* ditto */
 	void			*private_data;	/* ditto */
 } __attribute__((aligned(sizeof(long))));
diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h
index 76f151ab4f62..0385a954465c 100644
--- a/include/linux/pagemap.h
+++ b/include/linux/pagemap.h
@@ -16,17 +16,16 @@
 #include <linux/hugetlb_inline.h>
 
 /*
- * Bits in mapping->flags.  The lower __GFP_BITS_SHIFT bits are the page
- * allocation mode flags.
+ * Bits in mapping->flags.
  */
 enum mapping_flags {
-	AS_EIO		= __GFP_BITS_SHIFT + 0,	/* IO error on async write */
-	AS_ENOSPC	= __GFP_BITS_SHIFT + 1,	/* ENOSPC on async write */
-	AS_MM_ALL_LOCKS	= __GFP_BITS_SHIFT + 2,	/* under mm_take_all_locks() */
-	AS_UNEVICTABLE	= __GFP_BITS_SHIFT + 3,	/* e.g., ramdisk, SHM_LOCK */
-	AS_EXITING	= __GFP_BITS_SHIFT + 4, /* final truncate in progress */
+	AS_EIO		= 0,	/* IO error on async write */
+	AS_ENOSPC	= 1,	/* ENOSPC on async write */
+	AS_MM_ALL_LOCKS	= 2,	/* under mm_take_all_locks() */
+	AS_UNEVICTABLE	= 3,	/* e.g., ramdisk, SHM_LOCK */
+	AS_EXITING	= 4, 	/* final truncate in progress */
 	/* writeback related tags are not used */
-	AS_NO_WRITEBACK_TAGS = __GFP_BITS_SHIFT + 5,
+	AS_NO_WRITEBACK_TAGS = 5,
 
 	AS_LAST_FLAG,
 };
@@ -80,7 +79,7 @@ static inline int mapping_use_writeback_tags(struct address_space *mapping)
 
 static inline gfp_t mapping_gfp_mask(struct address_space * mapping)
 {
-	return (__force gfp_t)mapping->flags & __GFP_BITS_MASK;
+	return mapping->gfp_mask;
 }
 
 /* Restricts the given gfp_mask to what the mapping allows. */
@@ -96,8 +95,7 @@ static inline gfp_t mapping_gfp_constraint(struct address_space *mapping,
  */
 static inline void mapping_set_gfp_mask(struct address_space *m, gfp_t mask)
 {
-	m->flags = (m->flags & ~(__force unsigned long)__GFP_BITS_MASK) |
-				(__force unsigned long)mask;
+	m->gfp_mask = mask;
 }
 
 void release_pages(struct page **pages, int nr, bool cold);
-- 
2.9.3

-- 
Michal Hocko
SUSE Labs

[toc] | [prev] | [next] | [standalone]

#1481150 — [PATCH 1/2] fs: use mapping_set_error instead of opencoded set_bit

FromMichal Hocko <mhocko@kernel.org>
Date2016-09-12 13:20 +0200
Subject[PATCH 1/2] fs: use mapping_set_error instead of opencoded set_bit
Message-ID<sgxrk-2vU-19@gated-at.bofh.it>
In reply to#1481148
From: Michal Hocko <mhocko@suse.com>

mapping_set_error helper sets the correct AS_ flag for the mapping so
there is no reason to open code it. Use the helper directly.

Signed-off-by: Michal Hocko <mhocko@suse.com>
---
 drivers/staging/lustre/lustre/llite/vvp_page.c | 5 +----
 fs/afs/write.c                                 | 5 ++---
 fs/buffer.c                                    | 4 ++--
 fs/exofs/inode.c                               | 2 +-
 fs/ext4/page-io.c                              | 2 +-
 fs/f2fs/data.c                                 | 2 +-
 fs/jbd2/commit.c                               | 3 +--
 7 files changed, 9 insertions(+), 14 deletions(-)

diff --git a/drivers/staging/lustre/lustre/llite/vvp_page.c b/drivers/staging/lustre/lustre/llite/vvp_page.c
index 6cd2af7a958f..96194b6f118e 100644
--- a/drivers/staging/lustre/lustre/llite/vvp_page.c
+++ b/drivers/staging/lustre/lustre/llite/vvp_page.c
@@ -247,10 +247,7 @@ static void vvp_vmpage_error(struct inode *inode, struct page *vmpage, int ioret
 		obj->vob_discard_page_warned = 0;
 	} else {
 		SetPageError(vmpage);
-		if (ioret == -ENOSPC)
-			set_bit(AS_ENOSPC, &inode->i_mapping->flags);
-		else
-			set_bit(AS_EIO, &inode->i_mapping->flags);
+		mapping_set_error(inode->i_mapping, ioret);
 
 		if ((ioret == -ESHUTDOWN || ioret == -EINTR) &&
 		     obj->vob_discard_page_warned == 0) {
diff --git a/fs/afs/write.c b/fs/afs/write.c
index 14d506efd1aa..20ed04ab833c 100644
--- a/fs/afs/write.c
+++ b/fs/afs/write.c
@@ -398,8 +398,7 @@ static int afs_write_back_from_locked_page(struct afs_writeback *wb,
 		switch (ret) {
 		case -EDQUOT:
 		case -ENOSPC:
-			set_bit(AS_ENOSPC,
-				&wb->vnode->vfs_inode.i_mapping->flags);
+			mapping_set_error(wb->vnode->vfs_inode.i_mapping, -ENOSPC);
 			break;
 		case -EROFS:
 		case -EIO:
@@ -409,7 +408,7 @@ static int afs_write_back_from_locked_page(struct afs_writeback *wb,
 		case -ENOMEDIUM:
 		case -ENXIO:
 			afs_kill_pages(wb->vnode, true, first, last);
-			set_bit(AS_EIO, &wb->vnode->vfs_inode.i_mapping->flags);
+			mapping_set_error(wb->vnode->vfs_inode.i_mapping, -ENXIO);
 			break;
 		case -EACCES:
 		case -EPERM:
diff --git a/fs/buffer.c b/fs/buffer.c
index 754813a6962b..467e1ac3fac6 100644
--- a/fs/buffer.c
+++ b/fs/buffer.c
@@ -350,7 +350,7 @@ void end_buffer_async_write(struct buffer_head *bh, int uptodate)
 		set_buffer_uptodate(bh);
 	} else {
 		buffer_io_error(bh, ", lost async page write");
-		set_bit(AS_EIO, &page->mapping->flags);
+		mapping_set_error(page->mapping, -EIO);
 		set_buffer_write_io_error(bh);
 		clear_buffer_uptodate(bh);
 		SetPageError(page);
@@ -3180,7 +3180,7 @@ drop_buffers(struct page *page, struct buffer_head **buffers_to_free)
 	bh = head;
 	do {
 		if (buffer_write_io_error(bh) && page->mapping)
-			set_bit(AS_EIO, &page->mapping->flags);
+			mapping_set_error(page->mapping, -EIO);
 		if (buffer_busy(bh))
 			goto failed;
 		bh = bh->b_this_page;
diff --git a/fs/exofs/inode.c b/fs/exofs/inode.c
index 9dc4c6dbf3c9..a405db82e060 100644
--- a/fs/exofs/inode.c
+++ b/fs/exofs/inode.c
@@ -778,7 +778,7 @@ static int writepage_strip(struct page *page,
 fail:
 	EXOFS_DBGMSG("Error: writepage_strip(0x%lx, 0x%lx)=>%d\n",
 		     inode->i_ino, page->index, ret);
-	set_bit(AS_EIO, &page->mapping->flags);
+	mapping_set_error(page->mapping, -EIO);
 	unlock_page(page);
 	return ret;
 }
diff --git a/fs/ext4/page-io.c b/fs/ext4/page-io.c
index 2a01df9cc1c3..8073d63e37a7 100644
--- a/fs/ext4/page-io.c
+++ b/fs/ext4/page-io.c
@@ -89,7 +89,7 @@ static void ext4_finish_bio(struct bio *bio)
 
 		if (bio->bi_error) {
 			SetPageError(page);
-			set_bit(AS_EIO, &page->mapping->flags);
+			mapping_set_error(page->mapping, -EIO);
 		}
 		bh = head = page_buffers(page);
 		/*
diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index c80dda4bdff8..b728d284778e 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -67,7 +67,7 @@ static void f2fs_write_end_io(struct bio *bio)
 		fscrypt_pullback_bio_page(&page, true);
 
 		if (unlikely(bio->bi_error)) {
-			set_bit(AS_EIO, &page->mapping->flags);
+			mapping_set_error(page->mapping, -EIO);
 			f2fs_stop_checkpoint(sbi, true);
 		}
 		end_page_writeback(page);
diff --git a/fs/jbd2/commit.c b/fs/jbd2/commit.c
index 70078096117d..f3d5746f2446 100644
--- a/fs/jbd2/commit.c
+++ b/fs/jbd2/commit.c
@@ -269,8 +269,7 @@ static int journal_finish_inode_data_buffers(journal_t *journal,
 			 * filemap_fdatawait_range(), set it again so
 			 * that user process can get -EIO from fsync().
 			 */
-			set_bit(AS_EIO,
-				&jinode->i_vfs_inode->i_mapping->flags);
+			mapping_set_error(jinode->i_vfs_inode->i_mapping, -EIO);
 
 			if (!ret)
 				ret = err;
-- 
2.9.3

[toc] | [prev] | [next] | [standalone]

#1482032 — Re: [PATCH 1/2] fs: use mapping_set_error instead of opencoded set_bit

FromAndrew Morton <akpm@linux-foundation.org>
Date2016-09-13 00:20 +0200
SubjectRe: [PATCH 1/2] fs: use mapping_set_error instead of opencoded set_bit
Message-ID<sgHK2-1cH-31@gated-at.bofh.it>
In reply to#1481150
On Mon, 12 Sep 2016 13:16:07 +0200 Michal Hocko <mhocko@kernel.org> wrote:

> From: Michal Hocko <mhocko@suse.com>
> 
> mapping_set_error helper sets the correct AS_ flag for the mapping so
> there is no reason to open code it. Use the helper directly.
> 
> ...
>
> --- a/drivers/staging/lustre/lustre/llite/vvp_page.c
> +++ b/drivers/staging/lustre/lustre/llite/vvp_page.c
> @@ -247,10 +247,7 @@ static void vvp_vmpage_error(struct inode *inode, struct page *vmpage, int ioret
>  		obj->vob_discard_page_warned = 0;
>  	} else {
>  		SetPageError(vmpage);
> -		if (ioret == -ENOSPC)
> -			set_bit(AS_ENOSPC, &inode->i_mapping->flags);
> -		else
> -			set_bit(AS_EIO, &inode->i_mapping->flags);
> +		mapping_set_error(inode->i_mapping, ioret);
>  
>  		if ((ioret == -ESHUTDOWN || ioret == -EINTR) &&
>  		     obj->vob_discard_page_warned == 0) {
> diff --git a/fs/afs/write.c b/fs/afs/write.c
> index 14d506efd1aa..20ed04ab833c 100644
> --- a/fs/afs/write.c
> +++ b/fs/afs/write.c
> @@ -398,8 +398,7 @@ static int afs_write_back_from_locked_page(struct afs_writeback *wb,
>  		switch (ret) {
>  		case -EDQUOT:
>  		case -ENOSPC:
> -			set_bit(AS_ENOSPC,
> -				&wb->vnode->vfs_inode.i_mapping->flags);
> +			mapping_set_error(wb->vnode->vfs_inode.i_mapping, -ENOSPC);
>  			break;
>  		case -EROFS:
>  		case -EIO:
> @@ -409,7 +408,7 @@ static int afs_write_back_from_locked_page(struct afs_writeback *wb,
>  		case -ENOMEDIUM:
>  		case -ENXIO:
>  			afs_kill_pages(wb->vnode, true, first, last);
> -			set_bit(AS_EIO, &wb->vnode->vfs_inode.i_mapping->flags);
> +			mapping_set_error(wb->vnode->vfs_inode.i_mapping, -ENXIO);

This one is a functional change: mapping_set_error() will rewrite
-ENXIO into -EIO.  Doesn't seem at all important though.

> ...

[toc] | [prev] | [next] | [standalone]

#1482037 — Re: [PATCH 1/2] fs: use mapping_set_error instead of opencoded set_bit

FromAndrew Morton <akpm@linux-foundation.org>
Date2016-09-13 00:20 +0200
SubjectRe: [PATCH 1/2] fs: use mapping_set_error instead of opencoded set_bit
Message-ID<sgHK2-1cH-39@gated-at.bofh.it>
In reply to#1482032
On Mon, 12 Sep 2016 15:11:46 -0700 Andrew Morton <akpm@linux-foundation.org> wrote:

> > @@ -409,7 +408,7 @@ static int afs_write_back_from_locked_page(struct afs_writeback *wb,
> >  		case -ENOMEDIUM:
> >  		case -ENXIO:
> >  			afs_kill_pages(wb->vnode, true, first, last);
> > -			set_bit(AS_EIO, &wb->vnode->vfs_inode.i_mapping->flags);
> > +			mapping_set_error(wb->vnode->vfs_inode.i_mapping, -ENXIO);
> 
> This one is a functional change: mapping_set_error() will rewrite
> -ENXIO into -EIO.  Doesn't seem at all important though.

hm, OK, it's not a functional change - the code was already doing
s/ENXIO/EIO/.

Let's make it look more truthful?

--- a/fs/afs/write.c~fs-use-mapping_set_error-instead-of-opencoded-set_bit-fix
+++ a/fs/afs/write.c
@@ -408,7 +408,7 @@ no_more:
 		case -ENOMEDIUM:
 		case -ENXIO:
 			afs_kill_pages(wb->vnode, true, first, last);
-			mapping_set_error(wb->vnode->vfs_inode.i_mapping, -ENXIO);
+			mapping_set_error(wb->vnode->vfs_inode.i_mapping, -EIO);
 			break;
 		case -EACCES:
 		case -EPERM:
_

[toc] | [prev] | [next] | [standalone]

#1482180 — Re: [PATCH 1/2] fs: use mapping_set_error instead of opencoded set_bit

FromMichal Hocko <mhocko@kernel.org>
Date2016-09-13 09:00 +0200
SubjectRe: [PATCH 1/2] fs: use mapping_set_error instead of opencoded set_bit
Message-ID<sgPRf-6Fc-11@gated-at.bofh.it>
In reply to#1482037
On Mon 12-09-16 15:18:23, Andrew Morton wrote:
> On Mon, 12 Sep 2016 15:11:46 -0700 Andrew Morton <akpm@linux-foundation.org> wrote:
> 
> > > @@ -409,7 +408,7 @@ static int afs_write_back_from_locked_page(struct afs_writeback *wb,
> > >  		case -ENOMEDIUM:
> > >  		case -ENXIO:
> > >  			afs_kill_pages(wb->vnode, true, first, last);
> > > -			set_bit(AS_EIO, &wb->vnode->vfs_inode.i_mapping->flags);
> > > +			mapping_set_error(wb->vnode->vfs_inode.i_mapping, -ENXIO);
> > 
> > This one is a functional change: mapping_set_error() will rewrite
> > -ENXIO into -EIO.  Doesn't seem at all important though.
> 
> hm, OK, it's not a functional change - the code was already doing
> s/ENXIO/EIO/.

Yes the rewrite is silent but I've decided to keep the current errno
because I have no idea whether this can change in future. It doesn't
sound probable but it also sounds safer to do an overwrite at a single
place rather than all over the place /me thinks.

> Let's make it look more truthful?
> 
> --- a/fs/afs/write.c~fs-use-mapping_set_error-instead-of-opencoded-set_bit-fix
> +++ a/fs/afs/write.c
> @@ -408,7 +408,7 @@ no_more:
>  		case -ENOMEDIUM:
>  		case -ENXIO:
>  			afs_kill_pages(wb->vnode, true, first, last);
> -			mapping_set_error(wb->vnode->vfs_inode.i_mapping, -ENXIO);
> +			mapping_set_error(wb->vnode->vfs_inode.i_mapping, -EIO);
>  			break;
>  		case -EACCES:
>  		case -EPERM:
> _
> 
> 

-- 
Michal Hocko
SUSE Labs

[toc] | [prev] | [next] | [standalone]

#1482784 — Re: [PATCH 1/2] fs: use mapping_set_error instead of opencoded set_bit

FromAndrew Morton <akpm@linux-foundation.org>
Date2016-09-13 23:40 +0200
SubjectRe: [PATCH 1/2] fs: use mapping_set_error instead of opencoded set_bit
Message-ID<sh3AR-7tK-3@gated-at.bofh.it>
In reply to#1482180
On Tue, 13 Sep 2016 08:53:01 +0200 Michal Hocko <mhocko@kernel.org> wrote:

> On Mon 12-09-16 15:18:23, Andrew Morton wrote:
> > On Mon, 12 Sep 2016 15:11:46 -0700 Andrew Morton <akpm@linux-foundation.org> wrote:
> > 
> > > > @@ -409,7 +408,7 @@ static int afs_write_back_from_locked_page(struct afs_writeback *wb,
> > > >  		case -ENOMEDIUM:
> > > >  		case -ENXIO:
> > > >  			afs_kill_pages(wb->vnode, true, first, last);
> > > > -			set_bit(AS_EIO, &wb->vnode->vfs_inode.i_mapping->flags);
> > > > +			mapping_set_error(wb->vnode->vfs_inode.i_mapping, -ENXIO);
> > > 
> > > This one is a functional change: mapping_set_error() will rewrite
> > > -ENXIO into -EIO.  Doesn't seem at all important though.
> > 
> > hm, OK, it's not a functional change - the code was already doing
> > s/ENXIO/EIO/.
> 
> Yes the rewrite is silent but I've decided to keep the current errno
> because I have no idea whether this can change in future. It doesn't
> sound probable but it also sounds safer to do an overwrite at a single
> place rather than all over the place /me thinks.

Well, this is the only place in the kernel where we attempt to set
anything other than EIO.  I do think it's better to be honest about
what's happening, right here at the callsite.

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web