diff mbox series

fs: Handle I_DONTCACHE in iput_final() instead of generic_drop_inode()

Message ID 20200831101313.168889-1-lihao2018.fnst@cn.fujitsu.com (mailing list archive)
State Superseded
Headers show
Series fs: Handle I_DONTCACHE in iput_final() instead of generic_drop_inode() | expand

Commit Message

Li, Hao Aug. 31, 2020, 10:13 a.m. UTC
If generic_drop_inode() returns true, it means iput_final() can evict
this inode regardless of whether it is dirty or not. If we check
I_DONTCACHE in generic_drop_inode(), any inode with this bit set will be
evicted unconditionally. This is not the desired behavior because
I_DONTCACHE only means the inode shouldn't be cached on the LRU list.
As for whether we need to evict this inode, this is what
generic_drop_inode() should do. This patch corrects the usage of
I_DONTCACHE.

This patch was proposed in [1].

[1]: https://lore.kernel.org/linux-fsdevel/20200831003407.GE12096@dread.disaster.area/

Signed-off-by: Hao Li <lihao2018.fnst@cn.fujitsu.com>
---
 fs/inode.c         | 3 ++-
 include/linux/fs.h | 3 +--
 2 files changed, 3 insertions(+), 3 deletions(-)

Comments

Ira Weiny Aug. 31, 2020, 5:12 p.m. UTC | #1
On Mon, Aug 31, 2020 at 06:13:13PM +0800, Hao Li wrote:
> If generic_drop_inode() returns true, it means iput_final() can evict
> this inode regardless of whether it is dirty or not. If we check
> I_DONTCACHE in generic_drop_inode(), any inode with this bit set will be
> evicted unconditionally. This is not the desired behavior because
> I_DONTCACHE only means the inode shouldn't be cached on the LRU list.
> As for whether we need to evict this inode, this is what
> generic_drop_inode() should do. This patch corrects the usage of
> I_DONTCACHE.
> 
> This patch was proposed in [1].
> 
> [1]: https://lore.kernel.org/linux-fsdevel/20200831003407.GE12096@dread.disaster.area/
> 
> Signed-off-by: Hao Li <lihao2018.fnst@cn.fujitsu.com>

Thanks!  I think this looks good, but shouldn't we add?  It seems like this is
a bug right?

Fixes: dae2f8ed7992 ("fs: Lift XFS_IDONTCACHE to the VFS layer")

Reviewed-by: Ira Weiny <ira.weiny@intel.com>

> ---
>  fs/inode.c         | 3 ++-
>  include/linux/fs.h | 3 +--
>  2 files changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/fs/inode.c b/fs/inode.c
> index 72c4c347afb7..4e45d5ea3d0f 100644
> --- a/fs/inode.c
> +++ b/fs/inode.c
> @@ -1625,7 +1625,8 @@ static void iput_final(struct inode *inode)
>  	else
>  		drop = generic_drop_inode(inode);
>  
> -	if (!drop && (sb->s_flags & SB_ACTIVE)) {
> +	if (!drop && !(inode->i_state & I_DONTCACHE) &&
> +			(sb->s_flags & SB_ACTIVE)) {
>  		inode_add_lru(inode);
>  		spin_unlock(&inode->i_lock);
>  		return;
> diff --git a/include/linux/fs.h b/include/linux/fs.h
> index e019ea2f1347..93caee80ce47 100644
> --- a/include/linux/fs.h
> +++ b/include/linux/fs.h
> @@ -2922,8 +2922,7 @@ extern int inode_needs_sync(struct inode *inode);
>  extern int generic_delete_inode(struct inode *inode);
>  static inline int generic_drop_inode(struct inode *inode)
>  {
> -	return !inode->i_nlink || inode_unhashed(inode) ||
> -		(inode->i_state & I_DONTCACHE);
> +	return !inode->i_nlink || inode_unhashed(inode);
>  }
>  extern void d_mark_dontcache(struct inode *inode);
>  
> -- 
> 2.28.0
> 
> 
>
Li, Hao Sept. 1, 2020, 3:48 a.m. UTC | #2
On 2020/9/1 1:12, Ira Weiny wrote:
> On Mon, Aug 31, 2020 at 06:13:13PM +0800, Hao Li wrote:
>> If generic_drop_inode() returns true, it means iput_final() can evict
>> this inode regardless of whether it is dirty or not. If we check
>> I_DONTCACHE in generic_drop_inode(), any inode with this bit set will be
>> evicted unconditionally. This is not the desired behavior because
>> I_DONTCACHE only means the inode shouldn't be cached on the LRU list.
>> As for whether we need to evict this inode, this is what
>> generic_drop_inode() should do. This patch corrects the usage of
>> I_DONTCACHE.
>>
>> This patch was proposed in [1].
>>
>> [1]: https://lore.kernel.org/linux-fsdevel/20200831003407.GE12096@dread.disaster.area/
>>
>> Signed-off-by: Hao Li <lihao2018.fnst@cn.fujitsu.com>
>
> Thanks!  I think this looks good, but shouldn't we add?  It seems like this is
> a bug right?
>
> Fixes: dae2f8ed7992 ("fs: Lift XFS_IDONTCACHE to the VFS layer")

Yeah, this is more meaningful.

I'm not sure if I need to submit a v2 patch, or this tag will be added
by the maintainer when applying this patch. I have no experience with
this before. Thanks!

Regards,
Hao Li

>
>
> Reviewed-by: Ira Weiny <ira.weiny@intel.com>
>
>> ---
>>  fs/inode.c         | 3 ++-
>>  include/linux/fs.h | 3 +--
>>  2 files changed, 3 insertions(+), 3 deletions(-)
>>
>> diff --git a/fs/inode.c b/fs/inode.c
>> index 72c4c347afb7..4e45d5ea3d0f 100644
>> --- a/fs/inode.c
>> +++ b/fs/inode.c
>> @@ -1625,7 +1625,8 @@ static void iput_final(struct inode *inode)
>>      else
>>          drop = generic_drop_inode(inode);
>>  
>> -    if (!drop && (sb->s_flags & SB_ACTIVE)) {
>> +    if (!drop && !(inode->i_state & I_DONTCACHE) &&
>> +            (sb->s_flags & SB_ACTIVE)) {
>>          inode_add_lru(inode);
>>          spin_unlock(&inode->i_lock);
>>          return;
>> diff --git a/include/linux/fs.h b/include/linux/fs.h
>> index e019ea2f1347..93caee80ce47 100644
>> --- a/include/linux/fs.h
>> +++ b/include/linux/fs.h
>> @@ -2922,8 +2922,7 @@ extern int inode_needs_sync(struct inode *inode);
>>  extern int generic_delete_inode(struct inode *inode);
>>  static inline int generic_drop_inode(struct inode *inode)
>>  {
>> -    return !inode->i_nlink || inode_unhashed(inode) ||
>> -        (inode->i_state & I_DONTCACHE);
>> +    return !inode->i_nlink || inode_unhashed(inode);
>>  }
>>  extern void d_mark_dontcache(struct inode *inode);
>>  
>> --
>> 2.28.0
>>
>>
>>
>
>
Dave Chinner Sept. 3, 2020, 9:58 p.m. UTC | #3
On Mon, Aug 31, 2020 at 06:13:13PM +0800, Hao Li wrote:
> If generic_drop_inode() returns true, it means iput_final() can evict
> this inode regardless of whether it is dirty or not. If we check
> I_DONTCACHE in generic_drop_inode(), any inode with this bit set will be
> evicted unconditionally. This is not the desired behavior because
> I_DONTCACHE only means the inode shouldn't be cached on the LRU list.
> As for whether we need to evict this inode, this is what
> generic_drop_inode() should do. This patch corrects the usage of
> I_DONTCACHE.
> 
> This patch was proposed in [1].
> 
> [1]: https://lore.kernel.org/linux-fsdevel/20200831003407.GE12096@dread.disaster.area/
> 
> Signed-off-by: Hao Li <lihao2018.fnst@cn.fujitsu.com>
> ---
>  fs/inode.c         | 3 ++-
>  include/linux/fs.h | 3 +--
>  2 files changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/fs/inode.c b/fs/inode.c
> index 72c4c347afb7..4e45d5ea3d0f 100644
> --- a/fs/inode.c
> +++ b/fs/inode.c
> @@ -1625,7 +1625,8 @@ static void iput_final(struct inode *inode)
>  	else
>  		drop = generic_drop_inode(inode);
>  
> -	if (!drop && (sb->s_flags & SB_ACTIVE)) {
> +	if (!drop && !(inode->i_state & I_DONTCACHE) &&
> +			(sb->s_flags & SB_ACTIVE)) {

FWIW, the format used in fs/inode.c is to align the logic
statements, not tab indent the additional lines in the statement.
i.e.

	if (!drop &&
	    !(inode->i_state & I_DONTCACHE) &&
	    (sb->s_flags & SB_ACTIVE)) {

Which gives a clear indication that there are all at the same
precedence and separate logic statements...

Otherwise the change looks good.

Probably best to resend with the fixes tag :)

Cheers,

Dave.
Li, Hao Sept. 4, 2020, 7:07 a.m. UTC | #4
On 2020/9/4 5:58, Dave Chinner wrote:
> On Mon, Aug 31, 2020 at 06:13:13PM +0800, Hao Li wrote:
>> If generic_drop_inode() returns true, it means iput_final() can evict
>> this inode regardless of whether it is dirty or not. If we check
>> I_DONTCACHE in generic_drop_inode(), any inode with this bit set will be
>> evicted unconditionally. This is not the desired behavior because
>> I_DONTCACHE only means the inode shouldn't be cached on the LRU list.
>> As for whether we need to evict this inode, this is what
>> generic_drop_inode() should do. This patch corrects the usage of
>> I_DONTCACHE.
>>
>> This patch was proposed in [1].
>>
>> [1]: https://lore.kernel.org/linux-fsdevel/20200831003407.GE12096@dread.disaster.area/
>>
>> Signed-off-by: Hao Li <lihao2018.fnst@cn.fujitsu.com>
>> ---
>>  fs/inode.c         | 3 ++-
>>  include/linux/fs.h | 3 +--
>>  2 files changed, 3 insertions(+), 3 deletions(-)
>>
>> diff --git a/fs/inode.c b/fs/inode.c
>> index 72c4c347afb7..4e45d5ea3d0f 100644
>> --- a/fs/inode.c
>> +++ b/fs/inode.c
>> @@ -1625,7 +1625,8 @@ static void iput_final(struct inode *inode)
>>      else
>>          drop = generic_drop_inode(inode);
>>  
>> -    if (!drop && (sb->s_flags & SB_ACTIVE)) {
>> +    if (!drop && !(inode->i_state & I_DONTCACHE) &&
>> +            (sb->s_flags & SB_ACTIVE)) {
>
> FWIW, the format used in fs/inode.c is to align the logic
> statements, not tab indent the additional lines in the statement.
> i.e.
>
>     if (!drop &&
>         !(inode->i_state & I_DONTCACHE) &&
>         (sb->s_flags & SB_ACTIVE)) {
>
> Which gives a clear indication that there are all at the same
> precedence and separate logic statements...
>
> Otherwise the change looks good.
>
> Probably best to resend with the fixes tag :)

Got it! Thanks.

>
>
> Cheers,
>
> Dave.
diff mbox series

Patch

diff --git a/fs/inode.c b/fs/inode.c
index 72c4c347afb7..4e45d5ea3d0f 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -1625,7 +1625,8 @@  static void iput_final(struct inode *inode)
 	else
 		drop = generic_drop_inode(inode);
 
-	if (!drop && (sb->s_flags & SB_ACTIVE)) {
+	if (!drop && !(inode->i_state & I_DONTCACHE) &&
+			(sb->s_flags & SB_ACTIVE)) {
 		inode_add_lru(inode);
 		spin_unlock(&inode->i_lock);
 		return;
diff --git a/include/linux/fs.h b/include/linux/fs.h
index e019ea2f1347..93caee80ce47 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -2922,8 +2922,7 @@  extern int inode_needs_sync(struct inode *inode);
 extern int generic_delete_inode(struct inode *inode);
 static inline int generic_drop_inode(struct inode *inode)
 {
-	return !inode->i_nlink || inode_unhashed(inode) ||
-		(inode->i_state & I_DONTCACHE);
+	return !inode->i_nlink || inode_unhashed(inode);
 }
 extern void d_mark_dontcache(struct inode *inode);