All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH 1/2] f2fs: fix to avoid potential negative .f_bfree
@ 2019-04-26  9:57 ` Chao Yu
  0 siblings, 0 replies; 9+ messages in thread
From: Chao Yu @ 2019-04-26  9:57 UTC (permalink / raw)
  To: jaegeuk; +Cc: linux-f2fs-devel, linux-kernel, chao, drosen, Chao Yu

When calculating .f_bfree value in f2fs_statfs(), sbi->unusable_block_count
can be increased after the judgment condition, result in overflow of
.f_bfree in later calculation. This patch fixes to use a temporary signed
variable to save the calculation result of .f_bfree.

	if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
 		buf->f_bfree = 0;
 	else
		buf->f_bfree -= sbi->unusable_block_count;

Signed-off-by: Chao Yu <yuchao0@huawei.com>
---
 fs/f2fs/super.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
index 2376bb01b5c4..fcc9793dbc2c 100644
--- a/fs/f2fs/super.c
+++ b/fs/f2fs/super.c
@@ -1216,6 +1216,7 @@ static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
 	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
 	block_t total_count, user_block_count, start_count;
 	u64 avail_node_count;
+	long long bfree;
 
 	total_count = le64_to_cpu(sbi->raw_super->block_count);
 	user_block_count = sbi->user_block_count;
@@ -1226,10 +1227,12 @@ static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
 	buf->f_blocks = total_count - start_count;
 	buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
 						sbi->current_reserved_blocks;
-	if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
+
+	bfree = buf->f_bfree - sbi->unusable_block_count;
+	if (unlikely(bfree < 0))
 		buf->f_bfree = 0;
 	else
-		buf->f_bfree -= sbi->unusable_block_count;
+		buf->f_bfree = bfree;
 
 	if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
 		buf->f_bavail = buf->f_bfree -
-- 
2.18.0.rc1


^ permalink raw reply related	[flat|nested] 9+ messages in thread
* [PATCH 1/2] f2fs: fix to avoid potential negative .f_bfree
@ 2019-04-26  9:57 ` Chao Yu
  0 siblings, 0 replies; 9+ messages in thread
From: Chao Yu @ 2019-04-26  9:57 UTC (permalink / raw)
  To: jaegeuk; +Cc: linux-f2fs-devel, linux-kernel, chao, drosen, Chao Yu

When calculating .f_bfree value in f2fs_statfs(), sbi->unusable_block_count
can be increased after the judgment condition, result in overflow of
.f_bfree in later calculation. This patch fixes to use a temporary signed
variable to save the calculation result of .f_bfree.

	if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
 		buf->f_bfree = 0;
 	else
		buf->f_bfree -= sbi->unusable_block_count;

Signed-off-by: Chao Yu <yuchao0@huawei.com>
---
 fs/f2fs/super.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
index 2376bb01b5c4..fcc9793dbc2c 100644
--- a/fs/f2fs/super.c
+++ b/fs/f2fs/super.c
@@ -1216,6 +1216,7 @@ static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
 	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
 	block_t total_count, user_block_count, start_count;
 	u64 avail_node_count;
+	long long bfree;
 
 	total_count = le64_to_cpu(sbi->raw_super->block_count);
 	user_block_count = sbi->user_block_count;
@@ -1226,10 +1227,12 @@ static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
 	buf->f_blocks = total_count - start_count;
 	buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
 						sbi->current_reserved_blocks;
-	if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
+
+	bfree = buf->f_bfree - sbi->unusable_block_count;
+	if (unlikely(bfree < 0))
 		buf->f_bfree = 0;
 	else
-		buf->f_bfree -= sbi->unusable_block_count;
+		buf->f_bfree = bfree;
 
 	if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
 		buf->f_bavail = buf->f_bfree -
-- 
2.18.0.rc1

^ permalink raw reply related	[flat|nested] 9+ messages in thread
* [PATCH 2/2] f2fs: fix to handle error in f2fs_disable_checkpoint()
  2019-04-26  9:57 ` Chao Yu
@ 2019-04-26  9:57   ` Chao Yu
  -1 siblings, 0 replies; 9+ messages in thread
From: Chao Yu @ 2019-04-26  9:57 UTC (permalink / raw)
  To: jaegeuk; +Cc: linux-f2fs-devel, linux-kernel, chao, drosen, Chao Yu

In f2fs_disable_checkpoint(), it needs to detect and propagate error
number returned from f2fs_write_checkpoint().

Signed-off-by: Chao Yu <yuchao0@huawei.com>
---
 fs/f2fs/super.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
index fcc9793dbc2c..ad0faa19cb69 100644
--- a/fs/f2fs/super.c
+++ b/fs/f2fs/super.c
@@ -1507,9 +1507,12 @@ static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
 	mutex_lock(&sbi->gc_mutex);
 	cpc.reason = CP_PAUSE;
 	set_sbi_flag(sbi, SBI_CP_DISABLED);
-	f2fs_write_checkpoint(sbi, &cpc);
+	err = f2fs_write_checkpoint(sbi, &cpc);
+	if (err)
+		goto out_unlock;
 
 	sbi->unusable_block_count = 0;
+out_unlock:
 	mutex_unlock(&sbi->gc_mutex);
 restore_flag:
 	sbi->sb->s_flags = s_flags;	/* Restore MS_RDONLY status */
-- 
2.18.0.rc1


^ permalink raw reply related	[flat|nested] 9+ messages in thread
* [PATCH 2/2] f2fs: fix to handle error in f2fs_disable_checkpoint()
@ 2019-04-26  9:57   ` Chao Yu
  0 siblings, 0 replies; 9+ messages in thread
From: Chao Yu @ 2019-04-26  9:57 UTC (permalink / raw)
  To: jaegeuk; +Cc: linux-f2fs-devel, linux-kernel, chao, drosen, Chao Yu

In f2fs_disable_checkpoint(), it needs to detect and propagate error
number returned from f2fs_write_checkpoint().

Signed-off-by: Chao Yu <yuchao0@huawei.com>
---
 fs/f2fs/super.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
index fcc9793dbc2c..ad0faa19cb69 100644
--- a/fs/f2fs/super.c
+++ b/fs/f2fs/super.c
@@ -1507,9 +1507,12 @@ static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
 	mutex_lock(&sbi->gc_mutex);
 	cpc.reason = CP_PAUSE;
 	set_sbi_flag(sbi, SBI_CP_DISABLED);
-	f2fs_write_checkpoint(sbi, &cpc);
+	err = f2fs_write_checkpoint(sbi, &cpc);
+	if (err)
+		goto out_unlock;
 
 	sbi->unusable_block_count = 0;
+out_unlock:
 	mutex_unlock(&sbi->gc_mutex);
 restore_flag:
 	sbi->sb->s_flags = s_flags;	/* Restore MS_RDONLY status */
-- 
2.18.0.rc1

^ permalink raw reply related	[flat|nested] 9+ messages in thread
* Re: [PATCH 1/2] f2fs: fix to avoid potential negative .f_bfree
  2019-04-26  9:57 ` Chao Yu
  (?)
  (?)
@ 2019-04-28 13:47 ` Jaegeuk Kim
  2019-04-29 14:13   ` Chao Yu
  -1 siblings, 1 reply; 9+ messages in thread
From: Jaegeuk Kim @ 2019-04-28 13:47 UTC (permalink / raw)
  To: Chao Yu; +Cc: linux-f2fs-devel, linux-kernel, chao, drosen

On 04/26, Chao Yu wrote:
> When calculating .f_bfree value in f2fs_statfs(), sbi->unusable_block_count
> can be increased after the judgment condition, result in overflow of
> .f_bfree in later calculation. This patch fixes to use a temporary signed
> variable to save the calculation result of .f_bfree.
> 
> 	if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
>  		buf->f_bfree = 0;
>  	else
> 		buf->f_bfree -= sbi->unusable_block_count;

Do we just need stat_lock for this?

> 
> Signed-off-by: Chao Yu <yuchao0@huawei.com>
> ---
>  fs/f2fs/super.c | 7 +++++--
>  1 file changed, 5 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
> index 2376bb01b5c4..fcc9793dbc2c 100644
> --- a/fs/f2fs/super.c
> +++ b/fs/f2fs/super.c
> @@ -1216,6 +1216,7 @@ static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
>  	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
>  	block_t total_count, user_block_count, start_count;
>  	u64 avail_node_count;
> +	long long bfree;
>  
>  	total_count = le64_to_cpu(sbi->raw_super->block_count);
>  	user_block_count = sbi->user_block_count;
> @@ -1226,10 +1227,12 @@ static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
>  	buf->f_blocks = total_count - start_count;
>  	buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
>  						sbi->current_reserved_blocks;
> -	if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
> +
> +	bfree = buf->f_bfree - sbi->unusable_block_count;
> +	if (unlikely(bfree < 0))
>  		buf->f_bfree = 0;
>  	else
> -		buf->f_bfree -= sbi->unusable_block_count;
> +		buf->f_bfree = bfree;
>  
>  	if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
>  		buf->f_bavail = buf->f_bfree -
> -- 
> 2.18.0.rc1

^ permalink raw reply	[flat|nested] 9+ messages in thread
* Re: [PATCH 1/2] f2fs: fix to avoid potential negative .f_bfree
  2019-04-28 13:47 ` [PATCH 1/2] f2fs: fix to avoid potential negative .f_bfree Jaegeuk Kim
@ 2019-04-29 14:13   ` Chao Yu
  2019-04-30  2:54     ` Jaegeuk Kim
  0 siblings, 1 reply; 9+ messages in thread
From: Chao Yu @ 2019-04-29 14:13 UTC (permalink / raw)
  To: Jaegeuk Kim, Chao Yu; +Cc: linux-f2fs-devel, linux-kernel, drosen

On 2019-4-28 21:47, Jaegeuk Kim wrote:
> On 04/26, Chao Yu wrote:
>> When calculating .f_bfree value in f2fs_statfs(), sbi->unusable_block_count
>> can be increased after the judgment condition, result in overflow of
>> .f_bfree in later calculation. This patch fixes to use a temporary signed
>> variable to save the calculation result of .f_bfree.
>>
>> 	if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
>>  		buf->f_bfree = 0;
>>  	else
>> 		buf->f_bfree -= sbi->unusable_block_count;
> 
> Do we just need stat_lock for this?

Like we access other stat value in statfs(), we just need the instantaneous
value of .unusable_block_count, so we don't need additional stat_lock, right?

Thanks,

> 
>>
>> Signed-off-by: Chao Yu <yuchao0@huawei.com>
>> ---
>>  fs/f2fs/super.c | 7 +++++--
>>  1 file changed, 5 insertions(+), 2 deletions(-)
>>
>> diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
>> index 2376bb01b5c4..fcc9793dbc2c 100644
>> --- a/fs/f2fs/super.c
>> +++ b/fs/f2fs/super.c
>> @@ -1216,6 +1216,7 @@ static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
>>  	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
>>  	block_t total_count, user_block_count, start_count;
>>  	u64 avail_node_count;
>> +	long long bfree;
>>  
>>  	total_count = le64_to_cpu(sbi->raw_super->block_count);
>>  	user_block_count = sbi->user_block_count;
>> @@ -1226,10 +1227,12 @@ static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
>>  	buf->f_blocks = total_count - start_count;
>>  	buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
>>  						sbi->current_reserved_blocks;
>> -	if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
>> +
>> +	bfree = buf->f_bfree - sbi->unusable_block_count;
>> +	if (unlikely(bfree < 0))
>>  		buf->f_bfree = 0;
>>  	else
>> -		buf->f_bfree -= sbi->unusable_block_count;
>> +		buf->f_bfree = bfree;
>>  
>>  	if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
>>  		buf->f_bavail = buf->f_bfree -
>> -- 
>> 2.18.0.rc1

^ permalink raw reply	[flat|nested] 9+ messages in thread
* Re: [PATCH 1/2] f2fs: fix to avoid potential negative .f_bfree
  2019-04-29 14:13   ` Chao Yu
@ 2019-04-30  2:54     ` Jaegeuk Kim
  2019-04-30  3:14         ` Chao Yu
  0 siblings, 1 reply; 9+ messages in thread
From: Jaegeuk Kim @ 2019-04-30  2:54 UTC (permalink / raw)
  To: Chao Yu; +Cc: Chao Yu, linux-f2fs-devel, linux-kernel, drosen

On 04/29, Chao Yu wrote:
> On 2019-4-28 21:47, Jaegeuk Kim wrote:
> > On 04/26, Chao Yu wrote:
> >> When calculating .f_bfree value in f2fs_statfs(), sbi->unusable_block_count
> >> can be increased after the judgment condition, result in overflow of
> >> .f_bfree in later calculation. This patch fixes to use a temporary signed
> >> variable to save the calculation result of .f_bfree.
> >>
> >> 	if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
> >>  		buf->f_bfree = 0;
> >>  	else
> >> 		buf->f_bfree -= sbi->unusable_block_count;
> > 
> > Do we just need stat_lock for this?
> 
> Like we access other stat value in statfs(), we just need the instantaneous
> value of .unusable_block_count, so we don't need additional stat_lock, right?

What I've concerend is whether or not this fixes all the inconsistent values.
The original intention was providing stats in best effort, so we wouldn't use
any lock.

> 
> Thanks,
> 
> > 
> >>
> >> Signed-off-by: Chao Yu <yuchao0@huawei.com>
> >> ---
> >>  fs/f2fs/super.c | 7 +++++--
> >>  1 file changed, 5 insertions(+), 2 deletions(-)
> >>
> >> diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
> >> index 2376bb01b5c4..fcc9793dbc2c 100644
> >> --- a/fs/f2fs/super.c
> >> +++ b/fs/f2fs/super.c
> >> @@ -1216,6 +1216,7 @@ static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
> >>  	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
> >>  	block_t total_count, user_block_count, start_count;
> >>  	u64 avail_node_count;
> >> +	long long bfree;
> >>  
> >>  	total_count = le64_to_cpu(sbi->raw_super->block_count);
> >>  	user_block_count = sbi->user_block_count;
> >> @@ -1226,10 +1227,12 @@ static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
> >>  	buf->f_blocks = total_count - start_count;
> >>  	buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
> >>  						sbi->current_reserved_blocks;
> >> -	if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
> >> +
> >> +	bfree = buf->f_bfree - sbi->unusable_block_count;
> >> +	if (unlikely(bfree < 0))
> >>  		buf->f_bfree = 0;
> >>  	else
> >> -		buf->f_bfree -= sbi->unusable_block_count;
> >> +		buf->f_bfree = bfree;
> >>  
> >>  	if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
> >>  		buf->f_bavail = buf->f_bfree -
> >> -- 
> >> 2.18.0.rc1

^ permalink raw reply	[flat|nested] 9+ messages in thread
* Re: [PATCH 1/2] f2fs: fix to avoid potential negative .f_bfree
  2019-04-30  2:54     ` Jaegeuk Kim
@ 2019-04-30  3:14         ` Chao Yu
  0 siblings, 0 replies; 9+ messages in thread
From: Chao Yu @ 2019-04-30  3:14 UTC (permalink / raw)
  To: Jaegeuk Kim, Chao Yu; +Cc: linux-f2fs-devel, linux-kernel, drosen

On 2019/4/30 10:54, Jaegeuk Kim wrote:
> On 04/29, Chao Yu wrote:
>> On 2019-4-28 21:47, Jaegeuk Kim wrote:
>>> On 04/26, Chao Yu wrote:
>>>> When calculating .f_bfree value in f2fs_statfs(), sbi->unusable_block_count
>>>> can be increased after the judgment condition, result in overflow of
>>>> .f_bfree in later calculation. This patch fixes to use a temporary signed
>>>> variable to save the calculation result of .f_bfree.
>>>>
>>>> 	if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
>>>>  		buf->f_bfree = 0;
>>>>  	else
>>>> 		buf->f_bfree -= sbi->unusable_block_count;
>>>
>>> Do we just need stat_lock for this?
>>
>> Like we access other stat value in statfs(), we just need the instantaneous
>> value of .unusable_block_count, so we don't need additional stat_lock, right?
> 
> What I've concerend is whether or not this fixes all the inconsistent values.
> The original intention was providing stats in best effort, so we wouldn't use
> any lock.

Hmm.. I've made a patch to protect .unusable_block_count update/access as below,
how about merging this two patch, in addition, in this patch, let's add
stat_lock around accessing .f_bfree/.unusable_block_count.

From b927209d10fc222243037d05ccc899f48569e773 Mon Sep 17 00:00:00 2001
From: Chao Yu <yuchao0@huawei.com>
Date: Fri, 26 Apr 2019 15:51:22 +0800
Subject: [PATCH] f2fs: fix to cover sbi->unusable_block_count with stat_lock

Signed-off-by: Chao Yu <yuchao0@huawei.com>
---
 fs/f2fs/checkpoint.c | 4 ++++
 fs/f2fs/segment.c    | 5 ++++-
 fs/f2fs/super.c      | 2 ++
 3 files changed, 10 insertions(+), 1 deletion(-)

diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c
index f6ba9a743a2d..526a70ea7433 100644
--- a/fs/f2fs/checkpoint.c
+++ b/fs/f2fs/checkpoint.c
@@ -1536,7 +1536,11 @@ static int do_checkpoint(struct f2fs_sb_info *sbi, struct
cp_control *cpc)
 	clear_sbi_flag(sbi, SBI_IS_DIRTY);
 	clear_sbi_flag(sbi, SBI_NEED_CP);
 	clear_sbi_flag(sbi, SBI_QUOTA_SKIP_FLUSH);
+
+	spin_lock(&sbi->stat_lock);
 	sbi->unusable_block_count = 0;
+	spin_unlock(&sbi->stat_lock);
+
 	__set_cp_next_pack(sbi);

 	/*
diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
index 5c3f07540db1..f10c394cf467 100644
--- a/fs/f2fs/segment.c
+++ b/fs/f2fs/segment.c
@@ -2203,8 +2203,11 @@ static void update_sit_entry(struct f2fs_sb_info *sbi,
block_t blkaddr, int del)
 			 * before, we must track that to know how much space we
 			 * really have.
 			 */
-			if (f2fs_test_bit(offset, se->ckpt_valid_map))
+			if (f2fs_test_bit(offset, se->ckpt_valid_map)) {
+				spin_lock(&sbi->stat_lock);
 				sbi->unusable_block_count++;
+				spin_unlock(&sbi->stat_lock);
+			}
 		}

 		if (f2fs_test_and_clear_bit(offset, se->discard_map))
diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
index 1a05a636bd2a..dd6e7b351b58 100644
--- a/fs/f2fs/super.c
+++ b/fs/f2fs/super.c
@@ -1546,7 +1546,9 @@ static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
 	set_sbi_flag(sbi, SBI_CP_DISABLED);
 	f2fs_write_checkpoint(sbi, &cpc);

+	spin_lock(&sbi->stat_lock);
 	sbi->unusable_block_count = 0;
+	spin_unlock(&sbi->stat_lock);
 	mutex_unlock(&sbi->gc_mutex);
 restore_flag:
 	sbi->sb->s_flags = s_flags;	/* Restore MS_RDONLY status */
-- 
2.18.0.rc1



> 
>>
>> Thanks,
>>
>>>
>>>>
>>>> Signed-off-by: Chao Yu <yuchao0@huawei.com>
>>>> ---
>>>>  fs/f2fs/super.c | 7 +++++--
>>>>  1 file changed, 5 insertions(+), 2 deletions(-)
>>>>
>>>> diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
>>>> index 2376bb01b5c4..fcc9793dbc2c 100644
>>>> --- a/fs/f2fs/super.c
>>>> +++ b/fs/f2fs/super.c
>>>> @@ -1216,6 +1216,7 @@ static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
>>>>  	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
>>>>  	block_t total_count, user_block_count, start_count;
>>>>  	u64 avail_node_count;
>>>> +	long long bfree;
>>>>  
>>>>  	total_count = le64_to_cpu(sbi->raw_super->block_count);
>>>>  	user_block_count = sbi->user_block_count;
>>>> @@ -1226,10 +1227,12 @@ static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
>>>>  	buf->f_blocks = total_count - start_count;
>>>>  	buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
>>>>  						sbi->current_reserved_blocks;
>>>> -	if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
>>>> +
>>>> +	bfree = buf->f_bfree - sbi->unusable_block_count;
>>>> +	if (unlikely(bfree < 0))
>>>>  		buf->f_bfree = 0;
>>>>  	else
>>>> -		buf->f_bfree -= sbi->unusable_block_count;
>>>> +		buf->f_bfree = bfree;
>>>>  
>>>>  	if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
>>>>  		buf->f_bavail = buf->f_bfree -
>>>> -- 
>>>> 2.18.0.rc1
> .
> 

^ permalink raw reply related	[flat|nested] 9+ messages in thread
* Re: [PATCH 1/2] f2fs: fix to avoid potential negative .f_bfree
@ 2019-04-30  3:14         ` Chao Yu
  0 siblings, 0 replies; 9+ messages in thread
From: Chao Yu @ 2019-04-30  3:14 UTC (permalink / raw)
  To: Jaegeuk Kim, Chao Yu; +Cc: linux-f2fs-devel, linux-kernel, drosen

On 2019/4/30 10:54, Jaegeuk Kim wrote:
> On 04/29, Chao Yu wrote:
>> On 2019-4-28 21:47, Jaegeuk Kim wrote:
>>> On 04/26, Chao Yu wrote:
>>>> When calculating .f_bfree value in f2fs_statfs(), sbi->unusable_block_count
>>>> can be increased after the judgment condition, result in overflow of
>>>> .f_bfree in later calculation. This patch fixes to use a temporary signed
>>>> variable to save the calculation result of .f_bfree.
>>>>
>>>> 	if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
>>>>  		buf->f_bfree = 0;
>>>>  	else
>>>> 		buf->f_bfree -= sbi->unusable_block_count;
>>>
>>> Do we just need stat_lock for this?
>>
>> Like we access other stat value in statfs(), we just need the instantaneous
>> value of .unusable_block_count, so we don't need additional stat_lock, right?
> 
> What I've concerend is whether or not this fixes all the inconsistent values.
> The original intention was providing stats in best effort, so we wouldn't use
> any lock.

Hmm.. I've made a patch to protect .unusable_block_count update/access as below,
how about merging this two patch, in addition, in this patch, let's add
stat_lock around accessing .f_bfree/.unusable_block_count.

>From b927209d10fc222243037d05ccc899f48569e773 Mon Sep 17 00:00:00 2001
From: Chao Yu <yuchao0@huawei.com>
Date: Fri, 26 Apr 2019 15:51:22 +0800
Subject: [PATCH] f2fs: fix to cover sbi->unusable_block_count with stat_lock

Signed-off-by: Chao Yu <yuchao0@huawei.com>
---
 fs/f2fs/checkpoint.c | 4 ++++
 fs/f2fs/segment.c    | 5 ++++-
 fs/f2fs/super.c      | 2 ++
 3 files changed, 10 insertions(+), 1 deletion(-)

diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c
index f6ba9a743a2d..526a70ea7433 100644
--- a/fs/f2fs/checkpoint.c
+++ b/fs/f2fs/checkpoint.c
@@ -1536,7 +1536,11 @@ static int do_checkpoint(struct f2fs_sb_info *sbi, struct
cp_control *cpc)
 	clear_sbi_flag(sbi, SBI_IS_DIRTY);
 	clear_sbi_flag(sbi, SBI_NEED_CP);
 	clear_sbi_flag(sbi, SBI_QUOTA_SKIP_FLUSH);
+
+	spin_lock(&sbi->stat_lock);
 	sbi->unusable_block_count = 0;
+	spin_unlock(&sbi->stat_lock);
+
 	__set_cp_next_pack(sbi);

 	/*
diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
index 5c3f07540db1..f10c394cf467 100644
--- a/fs/f2fs/segment.c
+++ b/fs/f2fs/segment.c
@@ -2203,8 +2203,11 @@ static void update_sit_entry(struct f2fs_sb_info *sbi,
block_t blkaddr, int del)
 			 * before, we must track that to know how much space we
 			 * really have.
 			 */
-			if (f2fs_test_bit(offset, se->ckpt_valid_map))
+			if (f2fs_test_bit(offset, se->ckpt_valid_map)) {
+				spin_lock(&sbi->stat_lock);
 				sbi->unusable_block_count++;
+				spin_unlock(&sbi->stat_lock);
+			}
 		}

 		if (f2fs_test_and_clear_bit(offset, se->discard_map))
diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
index 1a05a636bd2a..dd6e7b351b58 100644
--- a/fs/f2fs/super.c
+++ b/fs/f2fs/super.c
@@ -1546,7 +1546,9 @@ static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
 	set_sbi_flag(sbi, SBI_CP_DISABLED);
 	f2fs_write_checkpoint(sbi, &cpc);

+	spin_lock(&sbi->stat_lock);
 	sbi->unusable_block_count = 0;
+	spin_unlock(&sbi->stat_lock);
 	mutex_unlock(&sbi->gc_mutex);
 restore_flag:
 	sbi->sb->s_flags = s_flags;	/* Restore MS_RDONLY status */
-- 
2.18.0.rc1



> 
>>
>> Thanks,
>>
>>>
>>>>
>>>> Signed-off-by: Chao Yu <yuchao0@huawei.com>
>>>> ---
>>>>  fs/f2fs/super.c | 7 +++++--
>>>>  1 file changed, 5 insertions(+), 2 deletions(-)
>>>>
>>>> diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
>>>> index 2376bb01b5c4..fcc9793dbc2c 100644
>>>> --- a/fs/f2fs/super.c
>>>> +++ b/fs/f2fs/super.c
>>>> @@ -1216,6 +1216,7 @@ static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
>>>>  	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
>>>>  	block_t total_count, user_block_count, start_count;
>>>>  	u64 avail_node_count;
>>>> +	long long bfree;
>>>>  
>>>>  	total_count = le64_to_cpu(sbi->raw_super->block_count);
>>>>  	user_block_count = sbi->user_block_count;
>>>> @@ -1226,10 +1227,12 @@ static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
>>>>  	buf->f_blocks = total_count - start_count;
>>>>  	buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
>>>>  						sbi->current_reserved_blocks;
>>>> -	if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
>>>> +
>>>> +	bfree = buf->f_bfree - sbi->unusable_block_count;
>>>> +	if (unlikely(bfree < 0))
>>>>  		buf->f_bfree = 0;
>>>>  	else
>>>> -		buf->f_bfree -= sbi->unusable_block_count;
>>>> +		buf->f_bfree = bfree;
>>>>  
>>>>  	if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
>>>>  		buf->f_bavail = buf->f_bfree -
>>>> -- 
>>>> 2.18.0.rc1
> .
> 

^ permalink raw reply related	[flat|nested] 9+ messages in thread
end of thread, other threads:[~2019-04-30  3:15 UTC | newest]

Thread overview: 9+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-04-26  9:57 [PATCH 1/2] f2fs: fix to avoid potential negative .f_bfree Chao Yu
2019-04-26  9:57 ` Chao Yu
2019-04-26  9:57 ` [PATCH 2/2] f2fs: fix to handle error in f2fs_disable_checkpoint() Chao Yu
2019-04-26  9:57   ` Chao Yu
2019-04-28 13:47 ` [PATCH 1/2] f2fs: fix to avoid potential negative .f_bfree Jaegeuk Kim
2019-04-29 14:13   ` Chao Yu
2019-04-30  2:54     ` Jaegeuk Kim
2019-04-30  3:14       ` Chao Yu
2019-04-30  3:14         ` Chao Yu

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.