ovl: ignore lower entries when checking purity of non-directory entries
diff mbox

Message ID 145424627339.27058.6895511767458306300.stgit@zurg
State New
Headers show

Commit Message

Konstantin Khlebnikov Jan. 31, 2016, 1:17 p.m. UTC
After rename file dentry still holds reference to lower dentry from
previous location. This doesn't matter for data access because data
cames from upper dentry. But this stale lower dentry taints dentry
at new location and turns it into non-pure upper. Such file leaves
visible whiteout entry after remove in directory which shouldn't
have whiteouts at all.

Overlayfs already tracks pureness of file location in oe->opaque.
This patch just uses that for detecting actual path type.

Comment from Vivek Goyal's patch:

Here are the details of the problem. Do following.

$ mkdir upper lower work merged upper/dir/
$ touch lower/test
$ sudo mount -t overlay overlay -olowerdir=lower,upperdir=upper,workdir=work
merged
$ mv merged/test merged/dir/
$ rm merged/dir/test
$ ls -l merged/dir/
/usr/bin/ls: cannot access merged/dir/test: No such file or directory
total 0
c????????? ? ? ? ?            ? test

Basic problem seems to be that once a file has been unlinked, a
whiteout has been left behind which was not needed and hence it becomes
visible.

whiteout is visible because parent dir is of not type MERGE, hence
od->is_real is set during ovl_dir_open(). And that means ovl_iterate()
passes on iterate handling directly to underlying fs. Underlying fs does
not know/filter whiteouts so it becomes visible to user.

Why did we leave a whiteout to begin with when we should not have.
ovl_do_remove() checks for OVL_TYPE_PURE_UPPER() and does not leave
whiteout if file is pure upper. In this case file is not found to be
pure upper hence whiteout is left.

So why file was not PURE_UPPER in this case? I think because dentry is
still carrying some leftover state which was valid before rename. For example,
od->numlower was set to 1 as it was a lower file. After rename, this state
is not valid anymore as there is no such file in lower.

Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com>
Reported-by: Viktor Stanchev <me@viktorstanchev.com>
Diagnosed-by: Vivek Goyal <vgoyal@redhat.com>
Link: https://bugzilla.kernel.org/show_bug.cgi?id=109611
---
 fs/overlayfs/dir.c   |    7 +++++++
 fs/overlayfs/super.c |   12 +++++++-----
 2 files changed, 14 insertions(+), 5 deletions(-)


--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Comments

Vivek Goyal Feb. 1, 2016, 2:14 p.m. UTC | #1
On Sun, Jan 31, 2016 at 04:17:53PM +0300, Konstantin Khlebnikov wrote:
> After rename file dentry still holds reference to lower dentry from
> previous location. This doesn't matter for data access because data
> cames from upper dentry. But this stale lower dentry taints dentry
> at new location and turns it into non-pure upper. Such file leaves
> visible whiteout entry after remove in directory which shouldn't
> have whiteouts at all.
> 
> Overlayfs already tracks pureness of file location in oe->opaque.
> This patch just uses that for detecting actual path type.
> 
> Comment from Vivek Goyal's patch:
> 
> Here are the details of the problem. Do following.
> 
> $ mkdir upper lower work merged upper/dir/
> $ touch lower/test
> $ sudo mount -t overlay overlay -olowerdir=lower,upperdir=upper,workdir=work
> merged
> $ mv merged/test merged/dir/
> $ rm merged/dir/test
> $ ls -l merged/dir/
> /usr/bin/ls: cannot access merged/dir/test: No such file or directory
> total 0
> c????????? ? ? ? ?            ? test
> 
> Basic problem seems to be that once a file has been unlinked, a
> whiteout has been left behind which was not needed and hence it becomes
> visible.
> 
> whiteout is visible because parent dir is of not type MERGE, hence
> od->is_real is set during ovl_dir_open(). And that means ovl_iterate()
> passes on iterate handling directly to underlying fs. Underlying fs does
> not know/filter whiteouts so it becomes visible to user.
> 
> Why did we leave a whiteout to begin with when we should not have.
> ovl_do_remove() checks for OVL_TYPE_PURE_UPPER() and does not leave
> whiteout if file is pure upper. In this case file is not found to be
> pure upper hence whiteout is left.
> 
> So why file was not PURE_UPPER in this case? I think because dentry is
> still carrying some leftover state which was valid before rename. For example,
> od->numlower was set to 1 as it was a lower file. After rename, this state
> is not valid anymore as there is no such file in lower.
> 
> Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com>
> Reported-by: Viktor Stanchev <me@viktorstanchev.com>
> Diagnosed-by: Vivek Goyal <vgoyal@redhat.com>
> Link: https://bugzilla.kernel.org/show_bug.cgi?id=109611

Hi Konstantin,

Thanks for the patch. This patch works for me also does not break
unionmount-testsuite.

This change also sounds reasonable to me. (given we can't free lower). So
over rename we will be keeping references to dentries which we should
not have to. I guess there does not seem to be a better option.

Acked-by: Vivek Goyal <vgoyal@redhat.com>

Thanks
Vivek



> ---
>  fs/overlayfs/dir.c   |    7 +++++++
>  fs/overlayfs/super.c |   12 +++++++-----
>  2 files changed, 14 insertions(+), 5 deletions(-)
> 
> diff --git a/fs/overlayfs/dir.c b/fs/overlayfs/dir.c
> index ed95272d57a6..edf83f325bca 100644
> --- a/fs/overlayfs/dir.c
> +++ b/fs/overlayfs/dir.c
> @@ -903,6 +903,13 @@ static int ovl_rename2(struct inode *olddir, struct dentry *old,
>  	if (!overwrite && new_is_dir && !old_opaque && new_opaque)
>  		ovl_remove_opaque(newdentry);
>  
> +	/*
> +	 * Old dentry now lives in different location. Dentries in
> +	 * lowerstack are stale. We cannot drop them here because
> +	 * access to them is lockless. This could be only pure upper
> +	 * or opaque directory - numlower is zero. Or upper non-dir
> +	 * entry - its pureness is tracked by flag opaque.
> +	 */
>  	if (old_opaque != new_opaque) {
>  		ovl_dentry_set_opaque(old, new_opaque);
>  		if (!overwrite)
> diff --git a/fs/overlayfs/super.c b/fs/overlayfs/super.c
> index 8d826bd56b26..ba28b007005e 100644
> --- a/fs/overlayfs/super.c
> +++ b/fs/overlayfs/super.c
> @@ -76,12 +76,14 @@ enum ovl_path_type ovl_path_type(struct dentry *dentry)
>  	if (oe->__upperdentry) {
>  		type = __OVL_PATH_UPPER;
>  
> -		if (oe->numlower) {
> -			if (S_ISDIR(dentry->d_inode->i_mode))
> -				type |= __OVL_PATH_MERGE;
> -		} else if (!oe->opaque) {
> +		/*
> +		 * Non-dir dentry can hold lower dentry from previous
> +		 * location. Its purity depends only on opaque flag.
> +		 */
> +		if (oe->numlower && S_ISDIR(dentry->d_inode->i_mode))
> +			type |= __OVL_PATH_MERGE;
> +		else if (!oe->opaque)
>  			type |= __OVL_PATH_PURE;
> -		}
>  	} else {
>  		if (oe->numlower > 1)
>  			type |= __OVL_PATH_MERGE;
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Konstantin Khlebnikov Feb. 1, 2016, 3:43 p.m. UTC | #2
On Mon, Feb 1, 2016 at 5:14 PM, Vivek Goyal <vgoyal@redhat.com> wrote:
> On Sun, Jan 31, 2016 at 04:17:53PM +0300, Konstantin Khlebnikov wrote:
>> After rename file dentry still holds reference to lower dentry from
>> previous location. This doesn't matter for data access because data
>> cames from upper dentry. But this stale lower dentry taints dentry
>> at new location and turns it into non-pure upper. Such file leaves
>> visible whiteout entry after remove in directory which shouldn't
>> have whiteouts at all.
>>
>> Overlayfs already tracks pureness of file location in oe->opaque.
>> This patch just uses that for detecting actual path type.
>>
>> Comment from Vivek Goyal's patch:
>>
>> Here are the details of the problem. Do following.
>>
>> $ mkdir upper lower work merged upper/dir/
>> $ touch lower/test
>> $ sudo mount -t overlay overlay -olowerdir=lower,upperdir=upper,workdir=work
>> merged
>> $ mv merged/test merged/dir/
>> $ rm merged/dir/test
>> $ ls -l merged/dir/
>> /usr/bin/ls: cannot access merged/dir/test: No such file or directory
>> total 0
>> c????????? ? ? ? ?            ? test
>>
>> Basic problem seems to be that once a file has been unlinked, a
>> whiteout has been left behind which was not needed and hence it becomes
>> visible.
>>
>> whiteout is visible because parent dir is of not type MERGE, hence
>> od->is_real is set during ovl_dir_open(). And that means ovl_iterate()
>> passes on iterate handling directly to underlying fs. Underlying fs does
>> not know/filter whiteouts so it becomes visible to user.
>>
>> Why did we leave a whiteout to begin with when we should not have.
>> ovl_do_remove() checks for OVL_TYPE_PURE_UPPER() and does not leave
>> whiteout if file is pure upper. In this case file is not found to be
>> pure upper hence whiteout is left.
>>
>> So why file was not PURE_UPPER in this case? I think because dentry is
>> still carrying some leftover state which was valid before rename. For example,
>> od->numlower was set to 1 as it was a lower file. After rename, this state
>> is not valid anymore as there is no such file in lower.
>>
>> Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com>
>> Reported-by: Viktor Stanchev <me@viktorstanchev.com>
>> Diagnosed-by: Vivek Goyal <vgoyal@redhat.com>
>> Link: https://bugzilla.kernel.org/show_bug.cgi?id=109611
>
> Hi Konstantin,
>
> Thanks for the patch. This patch works for me also does not break
> unionmount-testsuite.

Interesting. I haven't heard about this testsuite. Never read Documentation. =)

I've send couple of patches for xfstests: this case and fixed bugs in setattr.
This seems much better place for them.

>
> This change also sounds reasonable to me. (given we can't free lower). So
> over rename we will be keeping references to dentries which we should
> not have to. I guess there does not seem to be a better option.
>
> Acked-by: Vivek Goyal <vgoyal@redhat.com>
>
> Thanks
> Vivek
>
>
>
>> ---
>>  fs/overlayfs/dir.c   |    7 +++++++
>>  fs/overlayfs/super.c |   12 +++++++-----
>>  2 files changed, 14 insertions(+), 5 deletions(-)
>>
>> diff --git a/fs/overlayfs/dir.c b/fs/overlayfs/dir.c
>> index ed95272d57a6..edf83f325bca 100644
>> --- a/fs/overlayfs/dir.c
>> +++ b/fs/overlayfs/dir.c
>> @@ -903,6 +903,13 @@ static int ovl_rename2(struct inode *olddir, struct dentry *old,
>>       if (!overwrite && new_is_dir && !old_opaque && new_opaque)
>>               ovl_remove_opaque(newdentry);
>>
>> +     /*
>> +      * Old dentry now lives in different location. Dentries in
>> +      * lowerstack are stale. We cannot drop them here because
>> +      * access to them is lockless. This could be only pure upper
>> +      * or opaque directory - numlower is zero. Or upper non-dir
>> +      * entry - its pureness is tracked by flag opaque.
>> +      */
>>       if (old_opaque != new_opaque) {
>>               ovl_dentry_set_opaque(old, new_opaque);
>>               if (!overwrite)
>> diff --git a/fs/overlayfs/super.c b/fs/overlayfs/super.c
>> index 8d826bd56b26..ba28b007005e 100644
>> --- a/fs/overlayfs/super.c
>> +++ b/fs/overlayfs/super.c
>> @@ -76,12 +76,14 @@ enum ovl_path_type ovl_path_type(struct dentry *dentry)
>>       if (oe->__upperdentry) {
>>               type = __OVL_PATH_UPPER;
>>
>> -             if (oe->numlower) {
>> -                     if (S_ISDIR(dentry->d_inode->i_mode))
>> -                             type |= __OVL_PATH_MERGE;
>> -             } else if (!oe->opaque) {
>> +             /*
>> +              * Non-dir dentry can hold lower dentry from previous
>> +              * location. Its purity depends only on opaque flag.
>> +              */
>> +             if (oe->numlower && S_ISDIR(dentry->d_inode->i_mode))
>> +                     type |= __OVL_PATH_MERGE;
>> +             else if (!oe->opaque)
>>                       type |= __OVL_PATH_PURE;
>> -             }
>>       } else {
>>               if (oe->numlower > 1)
>>                       type |= __OVL_PATH_MERGE;
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Vivek Goyal Feb. 3, 2016, 12:24 p.m. UTC | #3
On Mon, Feb 01, 2016 at 06:43:19PM +0300, Konstantin Khlebnikov wrote:

[..]
> > Hi Konstantin,
> >
> > Thanks for the patch. This patch works for me also does not break
> > unionmount-testsuite.
> 
> Interesting. I haven't heard about this testsuite. Never read Documentation. =)
> 
> I've send couple of patches for xfstests: this case and fixed bugs in setattr.
> This seems much better place for them.

I guess running xfstests on overlayfs makes sense too, in an attempt to
figure out how overyalfs behavior is different from non layerd file
systems like xfs, ext4 etc and how many of those differences can be
fixed. 

Thanks
Vivek
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Miklos Szeredi March 2, 2016, 2:15 p.m. UTC | #4
On Mon, Feb 1, 2016 at 3:14 PM, Vivek Goyal <vgoyal@redhat.com> wrote:
> On Sun, Jan 31, 2016 at 04:17:53PM +0300, Konstantin Khlebnikov wrote:
>> After rename file dentry still holds reference to lower dentry from
>> previous location. This doesn't matter for data access because data
>> cames from upper dentry. But this stale lower dentry taints dentry
>> at new location and turns it into non-pure upper. Such file leaves
>> visible whiteout entry after remove in directory which shouldn't
>> have whiteouts at all.
>>
>> Overlayfs already tracks pureness of file location in oe->opaque.
>> This patch just uses that for detecting actual path type.
>>
>> Comment from Vivek Goyal's patch:
>>
>> Here are the details of the problem. Do following.
>>
>> $ mkdir upper lower work merged upper/dir/
>> $ touch lower/test
>> $ sudo mount -t overlay overlay -olowerdir=lower,upperdir=upper,workdir=work
>> merged
>> $ mv merged/test merged/dir/
>> $ rm merged/dir/test
>> $ ls -l merged/dir/
>> /usr/bin/ls: cannot access merged/dir/test: No such file or directory
>> total 0
>> c????????? ? ? ? ?            ? test
>>
>> Basic problem seems to be that once a file has been unlinked, a
>> whiteout has been left behind which was not needed and hence it becomes
>> visible.
>>
>> whiteout is visible because parent dir is of not type MERGE, hence
>> od->is_real is set during ovl_dir_open(). And that means ovl_iterate()
>> passes on iterate handling directly to underlying fs. Underlying fs does
>> not know/filter whiteouts so it becomes visible to user.
>>
>> Why did we leave a whiteout to begin with when we should not have.
>> ovl_do_remove() checks for OVL_TYPE_PURE_UPPER() and does not leave
>> whiteout if file is pure upper. In this case file is not found to be
>> pure upper hence whiteout is left.
>>
>> So why file was not PURE_UPPER in this case? I think because dentry is
>> still carrying some leftover state which was valid before rename. For example,
>> od->numlower was set to 1 as it was a lower file. After rename, this state
>> is not valid anymore as there is no such file in lower.
>>
>> Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com>
>> Reported-by: Viktor Stanchev <me@viktorstanchev.com>
>> Diagnosed-by: Vivek Goyal <vgoyal@redhat.com>
>> Link: https://bugzilla.kernel.org/show_bug.cgi?id=109611
>
> Hi Konstantin,
>
> Thanks for the patch. This patch works for me also does not break
> unionmount-testsuite.

Thanks for the analysis and patch.  Added to the queue.

Thanks,
Miklos
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch
diff mbox

diff --git a/fs/overlayfs/dir.c b/fs/overlayfs/dir.c
index ed95272d57a6..edf83f325bca 100644
--- a/fs/overlayfs/dir.c
+++ b/fs/overlayfs/dir.c
@@ -903,6 +903,13 @@  static int ovl_rename2(struct inode *olddir, struct dentry *old,
 	if (!overwrite && new_is_dir && !old_opaque && new_opaque)
 		ovl_remove_opaque(newdentry);
 
+	/*
+	 * Old dentry now lives in different location. Dentries in
+	 * lowerstack are stale. We cannot drop them here because
+	 * access to them is lockless. This could be only pure upper
+	 * or opaque directory - numlower is zero. Or upper non-dir
+	 * entry - its pureness is tracked by flag opaque.
+	 */
 	if (old_opaque != new_opaque) {
 		ovl_dentry_set_opaque(old, new_opaque);
 		if (!overwrite)
diff --git a/fs/overlayfs/super.c b/fs/overlayfs/super.c
index 8d826bd56b26..ba28b007005e 100644
--- a/fs/overlayfs/super.c
+++ b/fs/overlayfs/super.c
@@ -76,12 +76,14 @@  enum ovl_path_type ovl_path_type(struct dentry *dentry)
 	if (oe->__upperdentry) {
 		type = __OVL_PATH_UPPER;
 
-		if (oe->numlower) {
-			if (S_ISDIR(dentry->d_inode->i_mode))
-				type |= __OVL_PATH_MERGE;
-		} else if (!oe->opaque) {
+		/*
+		 * Non-dir dentry can hold lower dentry from previous
+		 * location. Its purity depends only on opaque flag.
+		 */
+		if (oe->numlower && S_ISDIR(dentry->d_inode->i_mode))
+			type |= __OVL_PATH_MERGE;
+		else if (!oe->opaque)
 			type |= __OVL_PATH_PURE;
-		}
 	} else {
 		if (oe->numlower > 1)
 			type |= __OVL_PATH_MERGE;