[Ocfs2-devel] [PATCH 1/1] Ocfs2: Teach truncating and punching-hole codes to handle fastsymlink.

[Ocfs2-devel] [PATCH 1/1] Ocfs2: Teach truncating and punching-hole codes to handle fastsymlink.

[Tristan Ye]

Fast symlink can be treated the same way as inline file for truncating
and hole punching, since the mechanism is quite simliar per se.

Signed-off-by: Tristan Ye <tristan.ye@oracle.com>
---
 fs/ocfs2/alloc.c |   18 ++++++++++++++----
 fs/ocfs2/file.c  |    7 +++++--
 2 files changed, 19 insertions(+), 6 deletions(-)

diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c
index 9f8bd91..f13ce3e 100644
--- a/fs/ocfs2/alloc.c
+++ b/fs/ocfs2/alloc.c
@@ -50,6 +50,7 @@
 #include "uptodate.h"
 #include "xattr.h"
 #include "refcounttree.h"
+#include "symlink.h"
 
 #include "buffer_head_io.h"
 
@@ -7668,6 +7669,10 @@ bail:
 
 /*
  * 'start' is inclusive, 'end' is not.
+ *
+ * The name ocfs2_truncate_inline() may be misleading a little bit, since
+ * it handles truncating for fast symlink as well, cause they actually will
+ * be doing one thing per se.
  */
 int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh,
 			  unsigned int start, unsigned int end, int trunc)
@@ -7684,12 +7689,14 @@ int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh,
 
 	BUG_ON(start >= end);
 
-	if (!(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) ||
+	if ((!(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) ||
 	    !(le16_to_cpu(di->i_dyn_features) & OCFS2_INLINE_DATA_FL) ||
-	    !ocfs2_supports_inline_data(osb)) {
+	    !ocfs2_supports_inline_data(osb)) &&
+	    (!ocfs2_inode_is_fast_symlink(inode))) {
 		ocfs2_error(inode->i_sb,
 			    "Inline data flags for inode %llu don't agree! "
-			    "Disk: 0x%x, Memory: 0x%x, Superblock: 0x%x\n",
+			    "Disk: 0x%x, Memory: 0x%x, Superblock: 0x%x, "
+			    "and it's not a fast symlink as well.\n",
 			    (unsigned long long)OCFS2_I(inode)->ip_blkno,
 			    le16_to_cpu(di->i_dyn_features),
 			    OCFS2_I(inode)->ip_dyn_features,
@@ -7713,7 +7720,10 @@ int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh,
 	}
 
 	numbytes = end - start;
-	memset(idata->id_data + start, 0, numbytes);
+	if (ocfs2_inode_is_fast_symlink(inode))
+		memset(di->id2.i_symlink + start, 0, numbytes);
+	else
+		memset(idata->id_data + start, 0, numbytes);
 
 	/*
 	 * No need to worry about the data page here - it's been
diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c
index 17947dc..290a3b0 100644
--- a/fs/ocfs2/file.c
+++ b/fs/ocfs2/file.c
@@ -60,6 +60,7 @@
 #include "acl.h"
 #include "quota.h"
 #include "refcounttree.h"
+#include "symlink.h"
 
 #include "buffer_head_io.h"
 
@@ -498,7 +499,8 @@ static int ocfs2_truncate_file(struct inode *inode,
 	unmap_mapping_range(inode->i_mapping, new_i_size + PAGE_SIZE - 1, 0, 1);
 	truncate_inode_pages(inode->i_mapping, new_i_size);
 
-	if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
+	if ((OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) ||
+	    (ocfs2_inode_is_fast_symlink(inode))) {
 		status = ocfs2_truncate_inline(inode, di_bh, new_i_size,
 					       i_size_read(inode), 1);
 		if (status)
@@ -1450,7 +1452,8 @@ static int ocfs2_remove_inode_range(struct inode *inode,
 	if (byte_len == 0)
 		return 0;
 
-	if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
+	if ((OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) ||
+	    (ocfs2_inode_is_fast_symlink(inode))) {
 		ret = ocfs2_truncate_inline(inode, di_bh, byte_start,
 					    byte_start + byte_len, 0);
 		if (ret) {
-- 
1.5.5

[Ocfs2-devel] [PATCH 1/1] Ocfs2: Teach truncating and punching-hole codes to handle fastsymlink.

[Joel Becker]

On Tue, Mar 23, 2010 at 04:04:44PM +0800, Tristan Ye wrote:
> Fast symlink can be treated the same way as inline file for truncating
> and hole punching, since the mechanism is quite simliar per se.
> 
> Signed-off-by: Tristan Ye <tristan.ye@oracle.com>

	The patch looks good, with perhaps a bit over-zealous
parentheses ;-)  Is this fixing a bug someone has hit?

Joel


-- 

"In a crisis, don't hide behind anything or anybody. They're going
 to find you anyway."
	- Paul "Bear" Bryant

Joel Becker
Principal Software Developer
Oracle
E-mail: joel.becker at oracle.com
Phone: (650) 506-8127

[Ocfs2-devel] [PATCH 1/1] Ocfs2: Teach truncating and punching-hole codes to handle fastsymlink.

[tristan]

Joel Becker wrote:
> On Tue, Mar 23, 2010 at 04:04:44PM +0800, Tristan Ye wrote:
>> Fast symlink can be treated the same way as inline file for truncating
>> and hole punching, since the mechanism is quite simliar per se.
>>
>> Signed-off-by: Tristan Ye <tristan.ye@oracle.com>
>
> 	The patch looks good, with perhaps a bit over-zealous
> parentheses ;-)  Is this fixing a bug someone has hit?

Not exactly,

Sunil reported such a bug from userspace in libocfs2, I suddenly 
realised that we could also do the same in fs.

Tristan.

>
> Joel
>
>

[Ocfs2-devel] [PATCH 1/1] Ocfs2: Teach truncating and punching-hole codes to handle fastsymlink.

[Joel Becker]

On Wed, Mar 24, 2010 at 09:10:13AM +0800, tristan wrote:
> Joel Becker wrote:
> > On Tue, Mar 23, 2010 at 04:04:44PM +0800, Tristan Ye wrote:
> >> Fast symlink can be treated the same way as inline file for truncating
> >> and hole punching, since the mechanism is quite simliar per se.
> >>
> >> Signed-off-by: Tristan Ye <tristan.ye@oracle.com>
> >
> > 	The patch looks good, with perhaps a bit over-zealous
> > parentheses ;-)  Is this fixing a bug someone has hit?
> 
> Not exactly,
> 
> Sunil reported such a bug from userspace in libocfs2, I suddenly 
> realised that we could also do the same in fs.

	The difference is that we never call the truncate code for fast
symlinks or inline data in the kernel.  We do in libocfs2.

Joel

-- 

"If you took all of the grains of sand in the world, and lined
 them up end to end in a row, you'd be working for the government!"
	- Mr. Interesting

Joel Becker
Principal Software Developer
Oracle
E-mail: joel.becker at oracle.com
Phone: (650) 506-8127

[Ocfs2-devel] [PATCH 1/1] Ocfs2: Teach truncating and punching-hole codes to handle fastsymlink.

[tristan]

Joel Becker wrote:
> On Wed, Mar 24, 2010 at 09:10:13AM +0800, tristan wrote:
>> Joel Becker wrote:
>>> On Tue, Mar 23, 2010 at 04:04:44PM +0800, Tristan Ye wrote:
>>>> Fast symlink can be treated the same way as inline file for truncating
>>>> and hole punching, since the mechanism is quite simliar per se.
>>>>
>>>> Signed-off-by: Tristan Ye <tristan.ye@oracle.com>
>>> 	The patch looks good, with perhaps a bit over-zealous
>>> parentheses ;-)  Is this fixing a bug someone has hit?
>> Not exactly,
>>
>> Sunil reported such a bug from userspace in libocfs2, I suddenly 
>> realised that we could also do the same in fs.
>
> 	The difference is that we never call the truncate code for fast
> symlinks or inline data in the kernel.  We do in libocfs2.

Really?

Truncating for inline data is common I guess, for symlink, we may lack 
of method to truncate it from userspace via ftruncate(2).

But it's ok to be there, right?

Tristan.

>
> Joel
>

[Ocfs2-devel] [PATCH 1/1] Ocfs2: Teach truncating and punching-hole codes to handle fastsymlink.

[Joel Becker]

On Wed, Mar 24, 2010 at 10:21:57AM +0800, tristan wrote:
> Joel Becker wrote:
> > 	The difference is that we never call the truncate code for fast
> > symlinks or inline data in the kernel.  We do in libocfs2.
> 
> Really?
> 
> Truncating for inline data is common I guess, for symlink, we may lack 
> of method to truncate it from userspace via ftruncate(2).
> 
> But it's ok to be there, right?

	It's not OK to be there.  The truncate system call will error if
it is not a regular file.  Same with OCFS2_IOC_UNRESVP.  Thus no symlink
can get there.  ocfs2_truncate_for_delete() checks i_clusters before
calling the real truncate code.  Thus it avoids fast symlinks and inline
data.

Joel

-- 

"Too much walking shoes worn thin.
 Too much trippin' and my soul's worn thin.
 Time to catch a ride it leaves today
 Her name is what it means.
 Too much walking shoes worn thin."

Joel Becker
Principal Software Developer
Oracle
E-mail: joel.becker at oracle.com
Phone: (650) 506-8127

[Ocfs2-devel] [PATCH 1/1] Ocfs2: Teach truncating and punching-hole codes to handle fastsymlink.

[tristan]

Joel Becker wrote:
> On Wed, Mar 24, 2010 at 10:21:57AM +0800, tristan wrote:
>> Joel Becker wrote:
>>> 	The difference is that we never call the truncate code for fast
>>> symlinks or inline data in the kernel.  We do in libocfs2.
>> Really?
>>
>> Truncating for inline data is common I guess, for symlink, we may lack 
>> of method to truncate it from userspace via ftruncate(2).
>>
>> But it's ok to be there, right?
>
> 	It's not OK to be there.  The truncate system call will error if
> it is not a regular file.  Same with OCFS2_IOC_UNRESVP.  Thus no symlink
> can get there.  ocfs2_truncate_for_delete() checks i_clusters before
> calling the real truncate code.  Thus it avoids fast symlinks and inline
> data.

Joel,

I agreed with your point on the fact that truncating for fast symlink 
may be meaningless, since sys_truncate() will block all none-regular 
files. while it does make sense for inline data, see 
ocfs2_truncate_inline(), which was used to deal with inline file 
separately. that's the way we're treating inline file when doing 
truncating, how did you say we avoid this? that makes me confused.


Tristan.

>
> Joel
>

[Ocfs2-devel] [PATCH 1/1] Ocfs2: Teach truncating and punching-hole codes to handle fastsymlink.

[Joel Becker]

On Wed, Mar 24, 2010 at 10:51:24AM +0800, tristan wrote:
> > 	It's not OK to be there.  The truncate system call will error if
> > it is not a regular file.  Same with OCFS2_IOC_UNRESVP.  Thus no symlink
> > can get there.  ocfs2_truncate_for_delete() checks i_clusters before
> > calling the real truncate code.  Thus it avoids fast symlinks and inline
> > data.
> 
> Joel,
> 
> I agreed with your point on the fact that truncating for fast symlink 
> may be meaningless, since sys_truncate() will block all none-regular 
> files. while it does make sense for inline data, see 
> ocfs2_truncate_inline(), which was used to deal with inline file 
> separately. that's the way we're treating inline file when doing 
> truncating, how did you say we avoid this? that makes me confused.

	For inline data, we shortcut into ocfs2_truncate_inline() and
never go into the meaty truncate code.  That's all I meant.

Joel

-- 

"Reality is merely an illusion, albeit a very persistent one."
        - Albert Einstien

Joel Becker
Principal Software Developer
Oracle
E-mail: joel.becker at oracle.com
Phone: (650) 506-8127

[Ocfs2-devel] [PATCH 1/1] Ocfs2: Teach truncating and punching-hole codes to handle fastsymlink.

[Sunil Mushran]

tristan wrote:
>>> Not exactly,
>>>
>>> Sunil reported such a bug from userspace in libocfs2, I suddenly 
>>> realised that we could also do the same in fs.
>>>       
>> 	

I did not see the patch for libocfs2? Is it in the pipeline or did
I miss it?

[Ocfs2-devel] [PATCH 1/1] Ocfs2: Teach truncating and punching-hole codes to handle fastsymlink.

[tristan]

Sunil Mushran wrote:
> tristan wrote:
>>>> Not exactly,
>>>>
>>>> Sunil reported such a bug from userspace in libocfs2, I suddenly 
>>>> realised that we could also do the same in fs.
>>>>       
>>>     
>
> I did not see the patch for libocfs2? Is it in the pipeline or did
> I miss it?

Oh,

I did send it along with 'setting dtime' patch, let me forward the patch 
to you.

Tristan.

[Ocfs2-devel] [PATCH 1/1] Ocfs2: Teach truncating and punching-hole codes to handle fastsymlink.

[tristan]

Joel Becker wrote:
> On Tue, Mar 23, 2010 at 04:04:44PM +0800, Tristan Ye wrote:
>> Fast symlink can be treated the same way as inline file for truncating
>> and hole punching, since the mechanism is quite simliar per se.
>>
>> Signed-off-by: Tristan Ye <tristan.ye@oracle.com>
>
> 	The patch looks good, with perhaps a bit over-zealous
> parentheses ;-)  Is this fixing a bug someone has hit?

Joel,

You're right, there are some parentheses abuse somewhere, I'll simplify 
them.

>
> Joel
>
>

[Ocfs2-devel] [PATCH 1/1] Ocfs2: Teach truncating and punching-hole codes to handle fastsymlink.

[Tristan Ye]

Fast symlink can be treated the same way as inline file for truncating
and hole punching, since the mechanism is quite simliar per se.

Signed-off-by: Tristan Ye <tristan.ye@oracle.com>
---
 fs/ocfs2/alloc.c |   18 ++++++++++++++----
 fs/ocfs2/file.c  |    7 +++++--
 2 files changed, 19 insertions(+), 6 deletions(-)

diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c
index 9f8bd91..a6d6d8e 100644
--- a/fs/ocfs2/alloc.c
+++ b/fs/ocfs2/alloc.c
@@ -50,6 +50,7 @@
 #include "uptodate.h"
 #include "xattr.h"
 #include "refcounttree.h"
+#include "symlink.h"
 
 #include "buffer_head_io.h"
 
@@ -7668,6 +7669,10 @@ bail:
 
 /*
  * 'start' is inclusive, 'end' is not.
+ *
+ * The name ocfs2_truncate_inline() may be misleading a little bit, since
+ * it handles truncating for fast symlink as well, cause they actually will
+ * be doing one thing per se.
  */
 int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh,
 			  unsigned int start, unsigned int end, int trunc)
@@ -7684,12 +7689,14 @@ int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh,
 
 	BUG_ON(start >= end);
 
-	if (!(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) ||
+	if ((!(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) ||
 	    !(le16_to_cpu(di->i_dyn_features) & OCFS2_INLINE_DATA_FL) ||
-	    !ocfs2_supports_inline_data(osb)) {
+	    !ocfs2_supports_inline_data(osb)) &&
+	    !ocfs2_inode_is_fast_symlink(inode)) {
 		ocfs2_error(inode->i_sb,
 			    "Inline data flags for inode %llu don't agree! "
-			    "Disk: 0x%x, Memory: 0x%x, Superblock: 0x%x\n",
+			    "Disk: 0x%x, Memory: 0x%x, Superblock: 0x%x, "
+			    "and it's not a fast symlink as well.\n",
 			    (unsigned long long)OCFS2_I(inode)->ip_blkno,
 			    le16_to_cpu(di->i_dyn_features),
 			    OCFS2_I(inode)->ip_dyn_features,
@@ -7713,7 +7720,10 @@ int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh,
 	}
 
 	numbytes = end - start;
-	memset(idata->id_data + start, 0, numbytes);
+	if (ocfs2_inode_is_fast_symlink(inode))
+		memset(di->id2.i_symlink + start, 0, numbytes);
+	else
+		memset(idata->id_data + start, 0, numbytes);
 
 	/*
 	 * No need to worry about the data page here - it's been
diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c
index 17947dc..9f6051b 100644
--- a/fs/ocfs2/file.c
+++ b/fs/ocfs2/file.c
@@ -60,6 +60,7 @@
 #include "acl.h"
 #include "quota.h"
 #include "refcounttree.h"
+#include "symlink.h"
 
 #include "buffer_head_io.h"
 
@@ -498,7 +499,8 @@ static int ocfs2_truncate_file(struct inode *inode,
 	unmap_mapping_range(inode->i_mapping, new_i_size + PAGE_SIZE - 1, 0, 1);
 	truncate_inode_pages(inode->i_mapping, new_i_size);
 
-	if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
+	if ((OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) ||
+	    ocfs2_inode_is_fast_symlink(inode)) {
 		status = ocfs2_truncate_inline(inode, di_bh, new_i_size,
 					       i_size_read(inode), 1);
 		if (status)
@@ -1450,7 +1452,8 @@ static int ocfs2_remove_inode_range(struct inode *inode,
 	if (byte_len == 0)
 		return 0;
 
-	if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
+	if ((OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) ||
+	    ocfs2_inode_is_fast_symlink(inode)) {
 		ret = ocfs2_truncate_inline(inode, di_bh, byte_start,
 					    byte_start + byte_len, 0);
 		if (ret) {
-- 
1.5.5