Skip to contentSkip to conversation

ext4: Fix remount with 'abort' option isn't effective

PATCH In review Ye Bin 5 participants 12 messages Dec 21, 2021 – Mar 3, 2022 ext4 file system

lore
Share & tools
lore

Ye Bin

author
#0 lore raw

PATCH 1/1 ext4: Fix remount with 'abort' option isn't effective

We test remount with 'abort' option as follows:
[root@localhost home]# mount /dev/sda test [root@localhost home]# mount | grep test /dev/sda on /home/test type ext4 (rw,relatime) [root@localhost home]# mount -o remount,abort test [root@localhost home]# mount | grep test /dev/sda on /home/test type ext4 (rw,relatime)

Obviously, remount 'abort' option isn't effective.
After 6e47a3cc68fc commit we process abort option with 'ctx_set_mount_flags':
static inline void ctx_set_mount_flags(struct ext4_fs_context *ctx, int flag)

{
	ctx->mask_s_mount_flags |= flag;
	ctx->vals_s_mount_flags |= flag;
}

But we test 'abort' option with 'ext4_test_mount_flag':
static inline int ext4_test_mount_flag(struct super_block *sb, int bit)

{
        return test_bit(bit, &EXT4_SB(sb)->s_mount_flags);
}

To solve this issue, pass (1 << EXT4_MF_FS_ABORTED) to 'ctx_set_mount_flags'.

Fixes 6e47a3cc68fc("ext4: get rid of super block and sbi from handle_mount_ops()")
---
 fs/ext4/super.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
fs/ext4/super.c +1 -1
index b72d989b77fb..071b7b3c5678 100644
@@ -2236,7 +2236,7 @@ static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
 			 param->key);
 		return 0;
 	case Opt_abort:
-		ctx_set_mount_flags(ctx, EXT4_MF_FS_ABORTED);
+		ctx_set_mount_flags(ctx, 1 << EXT4_MF_FS_ABORTED);
 		return 0;
 	case Opt_i_version:
 		ctx_set_flags(ctx, SB_I_VERSION);
Signature · 1 line
2.31.1
#1 lore raw

Hi,

nice catch. This is a bug indeed. However I am currently in a process of changing the ctx_set/clear/test_ helpers because currently it generates functions that are unused.

While I am at it I can just create a custom ctx_set_mount_flags() helper that would behave as expected so that we won't have to specify "1 < EXT4_MF_FS_ABORTED" which is not really obvious and hence error prone.

My plan is to send my patch set including this one tomorrow, will that be fine with you ?

-Lukas

Quoted 46 lines · Ye Bin

We test remount with 'abort' option as follows:
[root@localhost home]# mount /dev/sda test [root@localhost home]# mount | grep test /dev/sda on /home/test type ext4 (rw,relatime) [root@localhost home]# mount -o remount,abort test [root@localhost home]# mount | grep test /dev/sda on /home/test type ext4 (rw,relatime)

Obviously, remount 'abort' option isn't effective.
After 6e47a3cc68fc commit we process abort option with 'ctx_set_mount_flags':
static inline void ctx_set_mount_flags(struct ext4_fs_context *ctx, int flag)

{
	ctx->mask_s_mount_flags |= flag;
	ctx->vals_s_mount_flags |= flag;
}

But we test 'abort' option with 'ext4_test_mount_flag':
static inline int ext4_test_mount_flag(struct super_block *sb, int bit)

{
        return test_bit(bit, &EXT4_SB(sb)->s_mount_flags);
}

To solve this issue, pass (1 << EXT4_MF_FS_ABORTED) to 'ctx_set_mount_flags'.

Fixes 6e47a3cc68fc("ext4: get rid of super block and sbi from handle_mount_ops()")
---
 fs/ext4/super.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
fs/ext4/super.c +1 -1
index b72d989b77fb..071b7b3c5678 100644
@@ -2236,7 +2236,7 @@ static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
 			 param->key);
 		return 0;
 	case Opt_abort:
-		ctx_set_mount_flags(ctx, EXT4_MF_FS_ABORTED);
+		ctx_set_mount_flags(ctx, 1 << EXT4_MF_FS_ABORTED);
 		return 0;
 	case Opt_i_version:
 		ctx_set_flags(ctx, SB_I_VERSION);
Signature · 1 line
2.31.1
author
#2 lore raw
Quoted 10 lines · Lukas Czerner

Hi,

nice catch. This is a bug indeed. However I am currently in a process of changing the ctx_set/clear/test_ helpers because currently it generates functions that are unused.

While I am at it I can just create a custom ctx_set_mount_flags() helper that would behave as expected so that we won't have to specify "1 < EXT4_MF_FS_ABORTED" which is not really obvious and hence error prone.

Actually, I fixed the first version as follows:

fs/ext4/super.c +8 -3
index b72d989b77fb..199920ffc7d3 100644
@@ -2049,8 +2049,8 @@ struct ext4_fs_context {
 	unsigned int	mask_s_mount_opt;
 	unsigned int	vals_s_mount_opt2;
 	unsigned int	mask_s_mount_opt2;
-	unsigned int	vals_s_mount_flags;
-	unsigned int	mask_s_mount_flags;
+	unsigned long	vals_s_mount_flags;
+	unsigned long	mask_s_mount_flags;
 	unsigned int	opt_flags;	/* MOPT flags */
 	unsigned int	spec;
 	u32		s_max_batch_time;
@@ -2166,7 +2166,12 @@ static inline bool ctx_test_##name(struct ext4_fs_context *ctx, int flag)\
 EXT4_SET_CTX(flags);
 EXT4_SET_CTX(mount_opt);
 EXT4_SET_CTX(mount_opt2);
-EXT4_SET_CTX(mount_flags);
+
+static inline void ctx_set_mount_flags(struct ext4_fs_context *ctx, int bit)
+{
+	set_bit(bit, &ctx->mask_s_mount_flags);
+	set_bit(bit, &ctx->vals_s_mount_flags);
+}
 
 

I think 'mask_s_mount_flags' is useless now.

Quoted 55 lines

My plan is to send my patch set including this one tomorrow, will that be fine with you ?

-Lukas

Quoted 46 lines · Ye Bin

We test remount with 'abort' option as follows:
[root@localhost home]# mount /dev/sda test [root@localhost home]# mount | grep test /dev/sda on /home/test type ext4 (rw,relatime) [root@localhost home]# mount -o remount,abort test [root@localhost home]# mount | grep test /dev/sda on /home/test type ext4 (rw,relatime)

Obviously, remount 'abort' option isn't effective.
After 6e47a3cc68fc commit we process abort option with 'ctx_set_mount_flags':
static inline void ctx_set_mount_flags(struct ext4_fs_context *ctx, int flag)

{
	ctx->mask_s_mount_flags |= flag;
	ctx->vals_s_mount_flags |= flag;
}

But we test 'abort' option with 'ext4_test_mount_flag':
static inline int ext4_test_mount_flag(struct super_block *sb, int bit)

{
         return test_bit(bit, &EXT4_SB(sb)->s_mount_flags);
}

To solve this issue, pass (1 << EXT4_MF_FS_ABORTED) to 'ctx_set_mount_flags'.

Fixes 6e47a3cc68fc("ext4: get rid of super block and sbi from handle_mount_ops()")
---
  fs/ext4/super.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
fs/ext4/super.c +1 -1
index b72d989b77fb..071b7b3c5678 100644
@@ -2236,7 +2236,7 @@ static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
  			 param->key);
  		return 0;
  	case Opt_abort:
-		ctx_set_mount_flags(ctx, EXT4_MF_FS_ABORTED);
+		ctx_set_mount_flags(ctx, 1 << EXT4_MF_FS_ABORTED);
  		return 0;
  	case Opt_i_version:
  		ctx_set_flags(ctx, SB_I_VERSION);
Signature · 1 line
2.31.1

.

#3 lore raw
Quoted 9 lines · yebin
Quoted 6 lines · Lukas Czerner

Hi,

nice catch. This is a bug indeed. However I am currently in a process of changing the ctx_set/clear/test_ helpers because currently it generates functions that are unused.

Quoted 5 lines
Quoted 4 lines

While I am at it I can just create a custom ctx_set_mount_flags() helper that would behave as expected so that we won't have to specify "1 < EXT4_MF_FS_ABORTED" which is not really obvious and hence error prone.

Actually, I fixed the first version as follows:

Allright, this looks better.

Quoted 17 lines
fs/ext4/super.c +2 -2
index b72d989b77fb..199920ffc7d3 100644
@@ -2049,8 +2049,8 @@ struct ext4_fs_context {
 	unsigned int	mask_s_mount_opt;
 	unsigned int	vals_s_mount_opt2;
 	unsigned int	mask_s_mount_opt2;
-	unsigned int	vals_s_mount_flags;
-	unsigned int	mask_s_mount_flags;
+	unsigned long	vals_s_mount_flags;
+	unsigned long	mask_s_mount_flags;
 	unsigned int	opt_flags;	/* MOPT flags */
 	unsigned int	spec;
 	u32		s_max_batch_time;

fs/ext4/super.c static inline bool ctx_test_##name(struct ext4_fs_context *ctx, int flag)\ · line 2170

Quoted 6 lines
fs/ext4/super.c +2 -1
@@ -2166,7 +2166,12 @@ static inline bool ctx_test_##name(struct ext4_fs_context *ctx, int flag)\
 EXT4_SET_CTX(flags);
 EXT4_SET_CTX(mount_opt);
 EXT4_SET_CTX(mount_opt2);
-EXT4_SET_CTX(mount_flags);
+
+static inline void ctx_set_mount_flags(struct ext4_fs_context *ctx, int bit)

Maybe ctx_set_mount_flag since you can't really set more than one this way?

Quoted 7 lines
fs/ext4/super.c +4 -0
+{
+	set_bit(bit, &ctx->mask_s_mount_flags);
+	set_bit(bit, &ctx->vals_s_mount_flags);
+}
 
 

I think 'mask_s_mount_flags' is useless now.

So how would we know what flags have changed ? Sure, there is currently no need to clear the flag but that can come in future and once it does we'll only need to create a clear helper, the rest will be ready. I'd rather keep it.

-Lukas

Quoted 56 lines
Quoted 55 lines

My plan is to send my patch set including this one tomorrow, will that be fine with you ?

-Lukas

Quoted 46 lines · Ye Bin

We test remount with 'abort' option as follows:
[root@localhost home]# mount /dev/sda test [root@localhost home]# mount | grep test /dev/sda on /home/test type ext4 (rw,relatime) [root@localhost home]# mount -o remount,abort test [root@localhost home]# mount | grep test /dev/sda on /home/test type ext4 (rw,relatime)

Obviously, remount 'abort' option isn't effective.
After 6e47a3cc68fc commit we process abort option with 'ctx_set_mount_flags':
static inline void ctx_set_mount_flags(struct ext4_fs_context *ctx, int flag)

{
	ctx->mask_s_mount_flags |= flag;
	ctx->vals_s_mount_flags |= flag;
}

But we test 'abort' option with 'ext4_test_mount_flag':
static inline int ext4_test_mount_flag(struct super_block *sb, int bit)

{
         return test_bit(bit, &EXT4_SB(sb)->s_mount_flags);
}

To solve this issue, pass (1 << EXT4_MF_FS_ABORTED) to 'ctx_set_mount_flags'.

Fixes 6e47a3cc68fc("ext4: get rid of super block and sbi from handle_mount_ops()")
---
  fs/ext4/super.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
fs/ext4/super.c +1 -1
index b72d989b77fb..071b7b3c5678 100644
@@ -2236,7 +2236,7 @@ static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
  			 param->key);
  		return 0;
  	case Opt_abort:
-		ctx_set_mount_flags(ctx, EXT4_MF_FS_ABORTED);
+		ctx_set_mount_flags(ctx, 1 << EXT4_MF_FS_ABORTED);
  		return 0;
  	case Opt_i_version:
  		ctx_set_flags(ctx, SB_I_VERSION);
Signature · 1 line
2.31.1

.

author
#4 lore raw
Quoted 32 lines · Lukas Czerner
Quoted 8 lines · yebin
Quoted 6 lines · Lukas Czerner

Hi,

nice catch. This is a bug indeed. However I am currently in a process of changing the ctx_set/clear/test_ helpers because currently it generates functions that are unused.

Quoted 5 lines
Quoted 4 lines

While I am at it I can just create a custom ctx_set_mount_flags() helper that would behave as expected so that we won't have to specify "1 < EXT4_MF_FS_ABORTED" which is not really obvious and hence error prone.

Actually, I fixed the first version as follows:

Allright, this looks better.

Quoted 16 lines
fs/ext4/super.c +2 -2
index b72d989b77fb..199920ffc7d3 100644
@@ -2049,8 +2049,8 @@ struct ext4_fs_context {
  	unsigned int	mask_s_mount_opt;
  	unsigned int	vals_s_mount_opt2;
  	unsigned int	mask_s_mount_opt2;
-	unsigned int	vals_s_mount_flags;
-	unsigned int	mask_s_mount_flags;
+	unsigned long	vals_s_mount_flags;
+	unsigned long	mask_s_mount_flags;
  	unsigned int	opt_flags;	/* MOPT flags */
  	unsigned int	spec;
  	u32		s_max_batch_time;
@@ -2166,7 +2166,12 @@ static inline bool ctx_test_##name(struct ext4_fs_context *ctx, int flag)\
Quoted 8 lines

fs/ext4/super.c static inline bool ctx_test_##name(struct ext4_fs_context *ctx, int flag)\ · line 2170

Quoted 6 lines
fs/ext4/super.c +2 -1
  EXT4_SET_CTX(flags);
  EXT4_SET_CTX(mount_opt);
  EXT4_SET_CTX(mount_opt2);
-EXT4_SET_CTX(mount_flags);
+
+static inline void ctx_set_mount_flags(struct ext4_fs_context *ctx, int bit)

Maybe ctx_set_mount_flag since you can't really set more than one this way?

So I think it's a little inappropriate to use the current repair scheme.

Quoted 6 lines
Quoted 6 lines
fs/ext4/super.c +4 -0
+{
+	set_bit(bit, &ctx->mask_s_mount_flags);
+	set_bit(bit, &ctx->vals_s_mount_flags);
+}
 
 
Quoted 7 lines
Quoted 1 line

I think 'mask_s_mount_flags' is useless now.

So how would we know what flags have changed ? Sure, there is currently no need to clear the flag but that can come in future and once it does we'll only need to create a clear helper, the rest will be ready. I'd rather keep it.

-Lukas

From this point of view, I agree with you.

Quoted 56 lines
Quoted 46 lines
Quoted 46 lines

My plan is to send my patch set including this one tomorrow, will that be fine with you ?

-Lukas

Quoted 40 lines · Ye Bin

We test remount with 'abort' option as follows:
[root@localhost home]# mount /dev/sda test [root@localhost home]# mount | grep test /dev/sda on /home/test type ext4 (rw,relatime) [root@localhost home]# mount -o remount,abort test [root@localhost home]# mount | grep test /dev/sda on /home/test type ext4 (rw,relatime)

Obviously, remount 'abort' option isn't effective.
After 6e47a3cc68fc commit we process abort option with 'ctx_set_mount_flags':
static inline void ctx_set_mount_flags(struct ext4_fs_context *ctx, int flag)

{
	ctx->mask_s_mount_flags |= flag;
	ctx->vals_s_mount_flags |= flag;
}

But we test 'abort' option with 'ext4_test_mount_flag':
static inline int ext4_test_mount_flag(struct super_block *sb, int bit)

{
          return test_bit(bit, &EXT4_SB(sb)->s_mount_flags);
}

To solve this issue, pass (1 << EXT4_MF_FS_ABORTED) to 'ctx_set_mount_flags'.

Fixes 6e47a3cc68fc("ext4: get rid of super block and sbi from handle_mount_ops()")
---
   fs/ext4/super.c | 2 +-
   1 file changed, 1 insertion(+), 1 deletion(-)
fs/ext4/super.c +1 -1
index b72d989b77fb..071b7b3c5678 100644
@@ -2236,7 +2236,7 @@ static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
   			 param->key);
   		return 0;
   	case Opt_abort:
-		ctx_set_mount_flags(ctx, EXT4_MF_FS_ABORTED);
+		ctx_set_mount_flags(ctx, 1 << EXT4_MF_FS_ABORTED);
Quoted 8 lines
Quoted 8 lines
Quoted 6 lines
fs/ext4/super.c +0 -0
   		return 0;
   	case Opt_i_version:
   		ctx_set_flags(ctx, SB_I_VERSION);
Signature · 1 line
2.31.1

.

.

maintainer
#5 lore raw
Quoted 11 lines · yebin
Quoted 9 lines · Lukas Czerner
Quoted 8 lines · yebin
Quoted 6 lines · Lukas Czerner

Hi,

nice catch. This is a bug indeed. However I am currently in a process of changing the ctx_set/clear/test_ helpers because currently it generates functions that are unused.

Quoted 6 lines
Quoted 6 lines
Quoted 5 lines
Quoted 4 lines

While I am at it I can just create a custom ctx_set_mount_flags() helper that would behave as expected so that we won't have to specify "1 < EXT4_MF_FS_ABORTED" which is not really obvious and hence error prone.

Actually, I fixed the first version as follows:

Allright, this looks better.

Was there an update to this patch? I can't seem to find it in my inbox or in patchwork....

Thanks,

- Ted

#6 lore raw

PATCH 1/1 ext4: fix remount with 'abort' option

After commit 6e47a3cc68fc ("ext4: get rid of super block and sbi from handle_mount_ops()") the 'abort' options stopped working. This is because we're using ctx_set_mount_flags() helper that's expecting an argument with the appropriate bit set, but instead got EXT4_MF_FS_ABORTED which is a bit position. ext4_set_mount_flag() is using set_bit() while ctx_set_mount_flags() was using bitwise OR.

Create a separate helper ctx_set_mount_flag() to handle setting the mount_flags correctly.

While we're at it clean up the EXT4_SET_CTX macros so that we're only creating helpers that we actually use to avoid warnings.

Fixes 6e47a3cc68fc ("ext4: get rid of super block and sbi from handle_mount_ops()")
Cc Ye Bin <[email protected]>
---
 fs/ext4/super.c | 29 +++++++++++++++++++++--------
 1 file changed, 21 insertions(+), 8 deletions(-)
fs/ext4/super.c +21 -8
index eee0d9ebfa6c..6f74cd51df2e 100644
@@ -2045,8 +2045,8 @@ struct ext4_fs_context {
 	unsigned int	mask_s_mount_opt;
 	unsigned int	vals_s_mount_opt2;
 	unsigned int	mask_s_mount_opt2;
-	unsigned int	vals_s_mount_flags;
-	unsigned int	mask_s_mount_flags;
+	unsigned long	vals_s_mount_flags;
+	unsigned long	mask_s_mount_flags;
 	unsigned int	opt_flags;	/* MOPT flags */
 	unsigned int	spec;
 	u32		s_max_batch_time;
@@ -2149,23 +2149,36 @@ static inline void ctx_set_##name(struct ext4_fs_context *ctx,		\
 {									\
 	ctx->mask_s_##name |= flag;					\
 	ctx->vals_s_##name |= flag;					\
-}									\
+}
+
+#define EXT4_CLEAR_CTX(name)						\
 static inline void ctx_clear_##name(struct ext4_fs_context *ctx,	\
 				    unsigned long flag)			\
 {									\
 	ctx->mask_s_##name |= flag;					\
 	ctx->vals_s_##name &= ~flag;					\
-}									\
+}
+
+#define EXT4_TEST_CTX(name)						\
 static inline unsigned long						\
 ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag)	\
 {									\
 	return (ctx->vals_s_##name & flag);				\
-}									\
+}
 
-EXT4_SET_CTX(flags);
+EXT4_SET_CTX(flags); /* set only */
 EXT4_SET_CTX(mount_opt);
+EXT4_CLEAR_CTX(mount_opt);
+EXT4_TEST_CTX(mount_opt);
 EXT4_SET_CTX(mount_opt2);
-EXT4_SET_CTX(mount_flags);
+EXT4_CLEAR_CTX(mount_opt2);
+EXT4_TEST_CTX(mount_opt2);
+
+static inline void ctx_set_mount_flag(struct ext4_fs_context *ctx, int bit)
+{
+	set_bit(bit, &ctx->mask_s_mount_flags);
+	set_bit(bit, &ctx->vals_s_mount_flags);
+}
 
 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
 {
@@ -2235,7 +2248,7 @@ static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
 			 param->key);
 		return 0;
 	case Opt_abort:
-		ctx_set_mount_flags(ctx, EXT4_MF_FS_ABORTED);
+		ctx_set_mount_flag(ctx, EXT4_MF_FS_ABORTED);
 		return 0;
 	case Opt_i_version:
 		ext4_msg(NULL, KERN_WARNING, deprecated_msg, param->key, "5.20");
Signature · 1 line
2.31.1
Tested-by on v1 1/1
#7 lore raw

Re: [PATCH] ext4: fix remount with 'abort' option

Quoted 10 lines · Lukas Czerner

After commit 6e47a3cc68fc ("ext4: get rid of super block and sbi from handle_mount_ops()") the 'abort' options stopped working. This is because we're using ctx_set_mount_flags() helper that's expecting an argument with the appropriate bit set, but instead got EXT4_MF_FS_ABORTED which is a bit position. ext4_set_mount_flag() is using set_bit() while ctx_set_mount_flags() was using bitwise OR.

Create a separate helper ctx_set_mount_flag() to handle setting the mount_flags correctly.

Quoted 6 lines

While we're at it clean up the EXT4_SET_CTX macros so that we're only creating helpers that we actually use to avoid warnings.

Fixes 6e47a3cc68fc ("ext4: get rid of super block and sbi from handle_mount_ops()")
Cc Ye Bin <[email protected]>

We also observed this abort issue as a regression of an fanotify LTP test (fanotify22). This patch fixes that test case for me.

Quoted 70 lines
---
 fs/ext4/super.c | 29 +++++++++++++++++++++--------
 1 file changed, 21 insertions(+), 8 deletions(-)
fs/ext4/super.c +21 -8
index eee0d9ebfa6c..6f74cd51df2e 100644
@@ -2045,8 +2045,8 @@ struct ext4_fs_context {
 	unsigned int	mask_s_mount_opt;
 	unsigned int	vals_s_mount_opt2;
 	unsigned int	mask_s_mount_opt2;
-	unsigned int	vals_s_mount_flags;
-	unsigned int	mask_s_mount_flags;
+	unsigned long	vals_s_mount_flags;
+	unsigned long	mask_s_mount_flags;
 	unsigned int	opt_flags;	/* MOPT flags */
 	unsigned int	spec;
 	u32		s_max_batch_time;
@@ -2149,23 +2149,36 @@ static inline void ctx_set_##name(struct ext4_fs_context *ctx,		\
 {									\
 	ctx->mask_s_##name |= flag;					\
 	ctx->vals_s_##name |= flag;					\
-}									\
+}
+
+#define EXT4_CLEAR_CTX(name)						\
 static inline void ctx_clear_##name(struct ext4_fs_context *ctx,	\
 				    unsigned long flag)			\
 {									\
 	ctx->mask_s_##name |= flag;					\
 	ctx->vals_s_##name &= ~flag;					\
-}									\
+}
+
+#define EXT4_TEST_CTX(name)						\
 static inline unsigned long						\
 ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag)	\
 {									\
 	return (ctx->vals_s_##name & flag);				\
-}									\
+}
 
-EXT4_SET_CTX(flags);
+EXT4_SET_CTX(flags); /* set only */
 EXT4_SET_CTX(mount_opt);
+EXT4_CLEAR_CTX(mount_opt);
+EXT4_TEST_CTX(mount_opt);
 EXT4_SET_CTX(mount_opt2);
-EXT4_SET_CTX(mount_flags);
+EXT4_CLEAR_CTX(mount_opt2);
+EXT4_TEST_CTX(mount_opt2);
+
+static inline void ctx_set_mount_flag(struct ext4_fs_context *ctx, int bit)
+{
+	set_bit(bit, &ctx->mask_s_mount_flags);
+	set_bit(bit, &ctx->vals_s_mount_flags);
+}
 
 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
 {
@@ -2235,7 +2248,7 @@ static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
 			 param->key);
 		return 0;
 	case Opt_abort:
-		ctx_set_mount_flags(ctx, EXT4_MF_FS_ABORTED);
+		ctx_set_mount_flag(ctx, EXT4_MF_FS_ABORTED);
 		return 0;
 	case Opt_i_version:
 		ext4_msg(NULL, KERN_WARNING, deprecated_msg, param->key, "5.20");
Signature · 1 line
Gabriel Krisman Bertazi
Reviewed-by on v1 1/1
#8 lore raw

Re: [PATCH] ext4: fix remount with 'abort' option

In reply to Lukas Czerner
Quoted 10 lines · Lukas Czerner

After commit 6e47a3cc68fc ("ext4: get rid of super block and sbi from handle_mount_ops()") the 'abort' options stopped working. This is because we're using ctx_set_mount_flags() helper that's expecting an argument with the appropriate bit set, but instead got EXT4_MF_FS_ABORTED which is a bit position. ext4_set_mount_flag() is using set_bit() while ctx_set_mount_flags() was using bitwise OR.

Create a separate helper ctx_set_mount_flag() to handle setting the mount_flags correctly.

Quoted 6 lines

While we're at it clean up the EXT4_SET_CTX macros so that we're only creating helpers that we actually use to avoid warnings.

Fixes 6e47a3cc68fc ("ext4: get rid of super block and sbi from handle_mount_ops()")
Cc Ye Bin <[email protected]>

ok thinking out loud here - reducing this to the functional change that fixes the bug, (apologies for the surely-mangled whitespace, I'm being lazy), it looks something like:

fs/ext4/super.c +10 -4
index eee0d9ebfa6c..a3047b033053 100644
@@ -2045,8 +2045,8 @@ struct ext4_fs_context {
        unsigned int    mask_s_mount_opt;
        unsigned int    vals_s_mount_opt2;
        unsigned int    mask_s_mount_opt2;
-       unsigned int    vals_s_mount_flags;
-       unsigned int    mask_s_mount_flags;
+       unsigned long   vals_s_mount_flags;
+       unsigned long   mask_s_mount_flags;
        unsigned int    opt_flags;      /* MOPT flags */
        unsigned int    spec;
        u32             s_max_batch_time;
@@ -2165,7 +2165,13 @@ ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag) \
 EXT4_SET_CTX(flags);
 EXT4_SET_CTX(mount_opt);
 EXT4_SET_CTX(mount_opt2);
-EXT4_SET_CTX(mount_flags);
+
+/* Setting the mount_flags field is special, it takes a bit number */
+static inline void ctx_set_mount_flag(struct ext4_fs_context *ctx, int bit)
+{
+       set_bit(bit, &ctx->mask_s_mount_flags);
+       set_bit(bit, &ctx->vals_s_mount_flags);
+}
 
 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
 {
@@ -2235,7 +2241,7 @@ static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
                         param->key);
                return 0;
        case Opt_abort:
-               ctx_set_mount_flags(ctx, EXT4_MF_FS_ABORTED);
+               ctx_set_mount_flag(ctx, EXT4_MF_FS_ABORTED);
                return 0;
        case Opt_i_version:
                ext4_msg(NULL, KERN_WARNING, deprecated_msg, param->key, "5.20");

and that makes sense to me. I wonder if there's any further cleanup that could make it slightly more clear or foolproof re: which flag matches which ctx_set_* generated function, so the wrong thing doesn't get sent to the wrong routine, but that's a different issue, so for the patch as you sent it,

Quoted 70 lines
---
 fs/ext4/super.c | 29 +++++++++++++++++++++--------
 1 file changed, 21 insertions(+), 8 deletions(-)
fs/ext4/super.c +21 -8
index eee0d9ebfa6c..6f74cd51df2e 100644
@@ -2045,8 +2045,8 @@ struct ext4_fs_context {
 	unsigned int	mask_s_mount_opt;
 	unsigned int	vals_s_mount_opt2;
 	unsigned int	mask_s_mount_opt2;
-	unsigned int	vals_s_mount_flags;
-	unsigned int	mask_s_mount_flags;
+	unsigned long	vals_s_mount_flags;
+	unsigned long	mask_s_mount_flags;
 	unsigned int	opt_flags;	/* MOPT flags */
 	unsigned int	spec;
 	u32		s_max_batch_time;
@@ -2149,23 +2149,36 @@ static inline void ctx_set_##name(struct ext4_fs_context *ctx,		\
 {									\
 	ctx->mask_s_##name |= flag;					\
 	ctx->vals_s_##name |= flag;					\
-}									\
+}
+
+#define EXT4_CLEAR_CTX(name)						\
 static inline void ctx_clear_##name(struct ext4_fs_context *ctx,	\
 				    unsigned long flag)			\
 {									\
 	ctx->mask_s_##name |= flag;					\
 	ctx->vals_s_##name &= ~flag;					\
-}									\
+}
+
+#define EXT4_TEST_CTX(name)						\
 static inline unsigned long						\
 ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag)	\
 {									\
 	return (ctx->vals_s_##name & flag);				\
-}									\
+}
 
-EXT4_SET_CTX(flags);
+EXT4_SET_CTX(flags); /* set only */
 EXT4_SET_CTX(mount_opt);
+EXT4_CLEAR_CTX(mount_opt);
+EXT4_TEST_CTX(mount_opt);
 EXT4_SET_CTX(mount_opt2);
-EXT4_SET_CTX(mount_flags);
+EXT4_CLEAR_CTX(mount_opt2);
+EXT4_TEST_CTX(mount_opt2);
+
+static inline void ctx_set_mount_flag(struct ext4_fs_context *ctx, int bit)
+{
+	set_bit(bit, &ctx->mask_s_mount_flags);
+	set_bit(bit, &ctx->vals_s_mount_flags);
+}
 
 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
 {
@@ -2235,7 +2248,7 @@ static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
 			 param->key);
 		return 0;
 	case Opt_abort:
-		ctx_set_mount_flags(ctx, EXT4_MF_FS_ABORTED);
+		ctx_set_mount_flag(ctx, EXT4_MF_FS_ABORTED);
 		return 0;
 	case Opt_i_version:
 		ext4_msg(NULL, KERN_WARNING, deprecated_msg, param->key, "5.20");
#9 lore raw

Re: [PATCH] ext4: fix remount with 'abort' option

In reply to Lukas Czerner

Hi Ted,

this problem is still generating warnings. Can you please take this in when you have time?

Thanks!
-Lukas

Quoted 89 lines · Lukas Czerner

After commit 6e47a3cc68fc ("ext4: get rid of super block and sbi from handle_mount_ops()") the 'abort' options stopped working. This is because we're using ctx_set_mount_flags() helper that's expecting an argument with the appropriate bit set, but instead got EXT4_MF_FS_ABORTED which is a bit position. ext4_set_mount_flag() is using set_bit() while ctx_set_mount_flags() was using bitwise OR.

Create a separate helper ctx_set_mount_flag() to handle setting the mount_flags correctly.

While we're at it clean up the EXT4_SET_CTX macros so that we're only creating helpers that we actually use to avoid warnings.

Fixes 6e47a3cc68fc ("ext4: get rid of super block and sbi from handle_mount_ops()")
Cc Ye Bin <[email protected]>
---
 fs/ext4/super.c | 29 +++++++++++++++++++++--------
 1 file changed, 21 insertions(+), 8 deletions(-)
fs/ext4/super.c +21 -8
index eee0d9ebfa6c..6f74cd51df2e 100644
@@ -2045,8 +2045,8 @@ struct ext4_fs_context {
 	unsigned int	mask_s_mount_opt;
 	unsigned int	vals_s_mount_opt2;
 	unsigned int	mask_s_mount_opt2;
-	unsigned int	vals_s_mount_flags;
-	unsigned int	mask_s_mount_flags;
+	unsigned long	vals_s_mount_flags;
+	unsigned long	mask_s_mount_flags;
 	unsigned int	opt_flags;	/* MOPT flags */
 	unsigned int	spec;
 	u32		s_max_batch_time;
@@ -2149,23 +2149,36 @@ static inline void ctx_set_##name(struct ext4_fs_context *ctx,		\
 {									\
 	ctx->mask_s_##name |= flag;					\
 	ctx->vals_s_##name |= flag;					\
-}									\
+}
+
+#define EXT4_CLEAR_CTX(name)						\
 static inline void ctx_clear_##name(struct ext4_fs_context *ctx,	\
 				    unsigned long flag)			\
 {									\
 	ctx->mask_s_##name |= flag;					\
 	ctx->vals_s_##name &= ~flag;					\
-}									\
+}
+
+#define EXT4_TEST_CTX(name)						\
 static inline unsigned long						\
 ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag)	\
 {									\
 	return (ctx->vals_s_##name & flag);				\
-}									\
+}
 
-EXT4_SET_CTX(flags);
+EXT4_SET_CTX(flags); /* set only */
 EXT4_SET_CTX(mount_opt);
+EXT4_CLEAR_CTX(mount_opt);
+EXT4_TEST_CTX(mount_opt);
 EXT4_SET_CTX(mount_opt2);
-EXT4_SET_CTX(mount_flags);
+EXT4_CLEAR_CTX(mount_opt2);
+EXT4_TEST_CTX(mount_opt2);
+
+static inline void ctx_set_mount_flag(struct ext4_fs_context *ctx, int bit)
+{
+	set_bit(bit, &ctx->mask_s_mount_flags);
+	set_bit(bit, &ctx->vals_s_mount_flags);
+}
 
 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
 {
@@ -2235,7 +2248,7 @@ static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
 			 param->key);
 		return 0;
 	case Opt_abort:
-		ctx_set_mount_flags(ctx, EXT4_MF_FS_ABORTED);
+		ctx_set_mount_flag(ctx, EXT4_MF_FS_ABORTED);
 		return 0;
 	case Opt_i_version:
 		ext4_msg(NULL, KERN_WARNING, deprecated_msg, param->key, "5.20");
Signature · 1 line
2.31.1
maintainer
#10 lore raw

Re: [PATCH] ext4: fix remount with 'abort' option

Quoted 4 lines · Lukas Czerner

Hi Ted,

this problem is still generating warnings. Can you please take this in when you have time?

Oops, sorry. I processed a bunch of patches last week, and had even pushed them them out to ext4.git on git.kernel.org, but I had forgotten to run "b4 ty". (I was waiting for the regression tests to finish, and then forgot to send out the e-mail ack's.) I'll send them out now.

- Ted

maintainer Applied
#11 lore raw

Re: [PATCH] ext4: fix remount with 'abort' option

In reply to Lukas Czerner
Quoted 8 lines · Lukas Czerner

After commit 6e47a3cc68fc ("ext4: get rid of super block and sbi from handle_mount_ops()") the 'abort' options stopped working. This is because we're using ctx_set_mount_flags() helper that's expecting an argument with the appropriate bit set, but instead got EXT4_MF_FS_ABORTED which is a bit position. ext4_set_mount_flag() is using set_bit() while ctx_set_mount_flags() was using bitwise OR.

[...]

Applied, thanks!

[1/1] ext4: fix remount with 'abort' option

      commit: e3952fcce1aad934f1322843b564ff86256444b2

Best regards,

Signature · 1 line
Theodore Ts'o <[email protected]>