Skip to contentSkip to conversation

Dwrite with non-aligned offset and size

f2fs-dev He YunLei 3 participants 7 messages Jun 1, 2015 – Jul 17, 2015

Subsystems (2)
lore
Share & tools
lore

He YunLei

author
#0 lore raw

Hi Jaegeuk,

We run ltp testcase with f2fs and obtain a TFAIL in diotest4, the result in detail is as fallow:

dio04

<<<test_start>>>
tag=dio04 stime=1432278894
cmdline="diotest4"
contacts=""
analysis=exit
<<<test_output>>>

diotest4    1  TPASS  :  Negative Offset
diotest4    2  TPASS  :  removed
diotest4    3  TFAIL  :  diotest4.c:129: write allows odd count.returns 1: Success
diotest4    4  TFAIL  :  diotest4.c:183: Odd count of read and write
diotest4    5  TPASS  :  Read beyond the file size

......

the result of ext4 with same environment:

dio04

<<<test_start>>>
tag=dio04 stime=1432259643
cmdline="diotest4"
contacts=""
analysis=exit
<<<test_output>>>

diotest4    1  TPASS  :  Negative Offset
diotest4    2  TPASS  :  removed
diotest4    3  TPASS  :  Odd count of read and write
diotest4    4  TPASS  :  Read beyond the file size

......

Does f2fs allow dwrite with non-aligned offset and size? I check the code and found dwrite with non-aligned offset and size will turn into buffered write. Whether it will have some impact on user layer applications?

I wrote a patch, not well tested, how do you think of it?

fs/f2fs/data.c +3 -2
index 9bedfa8..ba5d94c 100644
@@ -2010,8 +2010,9 @@ static ssize_t f2fs_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
        if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode))
                return 0;
 
-       if (check_direct_IO(inode, iter, offset))
-               return 0;
+       err = check_direct_IO(inode, iter, offset)
+       if (err)
+               return -EINVAL;
 
        trace_f2fs_direct_IO_enter(inode, offset, count, iov_iter_rw(iter));
 

I wish you and other developers in this list could help me in a correct way.

Thanks,
He

maintainer
#1 lore raw
Quoted 34 lines · He YunLei

Hi Jaegeuk,

We run ltp testcase with f2fs and obtain a TFAIL in diotest4, the result in detail is as fallow:

dio04

<<<test_start>>>
tag=dio04 stime=1432278894
cmdline="diotest4"
contacts=""
analysis=exit
<<<test_output>>>

diotest4    1  TPASS  :  Negative Offset
diotest4    2  TPASS  :  removed
diotest4    3  TFAIL  :  diotest4.c:129: write allows odd count.returns 1: Success
diotest4    4  TFAIL  :  diotest4.c:183: Odd count of read and write
diotest4    5  TPASS  :  Read beyond the file size

......

the result of ext4 with same environment:

dio04

<<<test_start>>>
tag=dio04 stime=1432259643
cmdline="diotest4"
contacts=""
analysis=exit
<<<test_output>>>

diotest4    1  TPASS  :  Negative Offset
diotest4    2  TPASS  :  removed
diotest4    3  TPASS  :  Odd count of read and write
diotest4    4  TPASS  :  Read beyond the file size
Quoted 5 lines

......

Does f2fs allow dwrite with non-aligned offset and size? I check the code and found dwrite with non-aligned offset and size will turn into buffered write. Whether it will have some impact on user layer applications?

It's not a big deal to return -EINVAL.
When I take a look at other filesystem behaviors, it seems there is no restriction.

Quoted 2 lines

I wrote a patch, not well tested, how do you think of it?

Returning the error number would be good to me.
Could you write and sumbit a complete one?

Thanks,

Quoted 21 lines
fs/f2fs/data.c +3 -2
index 9bedfa8..ba5d94c 100644
@@ -2010,8 +2010,9 @@ static ssize_t f2fs_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
        if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode))
                return 0;
 
-       if (check_direct_IO(inode, iter, offset))
-               return 0;
+       err = check_direct_IO(inode, iter, offset)
+       if (err)
+               return -EINVAL;
 
        trace_f2fs_direct_IO_enter(inode, offset, count, iov_iter_rw(iter));
 

I wish you and other developers in this list could help me in a correct way.

Thanks,
He

author
#2 lore raw
Quoted 37 lines · Jaegeuk Kim
Quoted 36 lines · He YunLei

Hi Jaegeuk,

We run ltp testcase with f2fs and obtain a TFAIL in diotest4, the result in detail is as fallow:

dio04

<<<test_start>>>
tag=dio04 stime=1432278894
cmdline="diotest4"
contacts=""
analysis=exit
<<<test_output>>>

diotest4    1  TPASS  :  Negative Offset
diotest4    2  TPASS  :  removed
diotest4    3  TFAIL  :  diotest4.c:129: write allows odd count.returns 1: Success
diotest4    4  TFAIL  :  diotest4.c:183: Odd count of read and write
diotest4    5  TPASS  :  Read beyond the file size

......

the result of ext4 with same environment:

dio04

<<<test_start>>>
tag=dio04 stime=1432259643
cmdline="diotest4"
contacts=""
analysis=exit
<<<test_output>>>

diotest4    1  TPASS  :  Negative Offset
diotest4    2  TPASS  :  removed
diotest4    3  TPASS  :  Odd count of read and write
diotest4    4  TPASS  :  Read beyond the file size

......

Quoted 7 lines
Quoted 3 lines

Does f2fs allow dwrite with non-aligned offset and size? I check the code and found dwrite with non-aligned offset and size will turn into buffered write. Whether it will have some impact on user layer applications?

It's not a big deal to return -EINVAL.
When I take a look at other filesystem behaviors, it seems there is no restriction.

Ext4 do a check in the function do_blockdev_direct_IO:

         if (align & blocksize_mask) {
             if (bdev)
                 blkbits = blksize_bits(bdev_logical_block_size(bdev));
             blocksize_mask = (1 << blkbits) - 1;
             if (align & blocksize_mask)
                 goto out;
         }

It will return -EINVAL if the alignment is not satisfied.

In f2fs, it do the check by check_direct_IO() before blockdev_direct_IO(). The difference between the two methods is whether turn dwrite with non-aligned offset and size into buffered write. I am not very clear which one is better!

Thanks,
He

Quoted 32 lines
Quoted 2 lines

I wrote a patch, not well tested, how do you think of it?

Returning the error number would be good to me.
Could you write and sumbit a complete one?

Thanks,

Quoted 19 lines
fs/f2fs/data.c +3 -2
index 9bedfa8..ba5d94c 100644
@@ -2010,8 +2010,9 @@ static ssize_t f2fs_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
         if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode))
                 return 0;
 
-       if (check_direct_IO(inode, iter, offset))
-               return 0;
+       err = check_direct_IO(inode, iter, offset)
+       if (err)
+               return -EINVAL;
 
         trace_f2fs_direct_IO_enter(inode, offset, count, iov_iter_rw(iter));
 

I wish you and other developers in this list could help me in a correct way.

Quoted 2 lines

Thanks,
He

.

#3 lore raw

Hi Yunlei,

Sorry for the long delay.

Quoted 63 lines

-----Original Message-----
From: He YunLei [mailto:[email protected]]
Sent: Tuesday, June 02, 2015 12:21 PM
To: Jaegeuk Kim
Cc: [email protected]; [email protected]
Subject: Re: [f2fs-dev] Dwrite with non-aligned offset and size

Quoted 37 lines · Jaegeuk Kim
Quoted 36 lines · He YunLei

Hi Jaegeuk,

We run ltp testcase with f2fs and obtain a TFAIL in diotest4, the result in detail is as fallow:

dio04

<<<test_start>>>
tag=dio04 stime=1432278894
cmdline="diotest4"
contacts=""
analysis=exit
<<<test_output>>>

diotest4    1  TPASS  :  Negative Offset
diotest4    2  TPASS  :  removed
diotest4    3  TFAIL  :  diotest4.c:129: write allows odd count.returns 1: Success
diotest4    4  TFAIL  :  diotest4.c:183: Odd count of read and write
diotest4    5  TPASS  :  Read beyond the file size

......

the result of ext4 with same environment:

dio04

<<<test_start>>>
tag=dio04 stime=1432259643
cmdline="diotest4"
contacts=""
analysis=exit
<<<test_output>>>

diotest4    1  TPASS  :  Negative Offset
diotest4    2  TPASS  :  removed
diotest4    3  TPASS  :  Odd count of read and write
diotest4    4  TPASS  :  Read beyond the file size

......

Quoted 7 lines
Quoted 3 lines

Does f2fs allow dwrite with non-aligned offset and size? I check the code and found dwrite with non-aligned offset and size will turn into buffered write. Whether it will have some impact on user layer applications?

It's not a big deal to return -EINVAL.
When I take a look at other filesystem behaviors, it seems there is no restriction.

Ext4 do a check in the function do_blockdev_direct_IO:

          if (align & blocksize_mask) {
              if (bdev)
                  blkbits = blksize_bits(bdev_logical_block_size(bdev));
              blocksize_mask = (1 << blkbits) - 1;
              if (align & blocksize_mask)
                  goto out;
          }
Quoted 1 line

It will return -EINVAL if the alignment is not satisfied.

I think we can get hint from the error case description in write(2) manual:

"EINVAL fd is attached to an object which is unsuitable for writing; or the file was opened with the O_DIRECT flag, and either the address specified in buf, the value specified in count, or the current file offset is not suitably aligned."

So if the alignment is not satisfied, we should return '-EINVAL' instead of letting user fall back to buffered write.

Do you have time to make and send us a patch for fixing this issue?

Thanks,

Quoted 47 lines

In f2fs, it do the check by check_direct_IO() before blockdev_direct_IO(). The difference between the two methods is whether turn dwrite with non-aligned offset and size into buffered write. I am not very clear which one is better!

Thanks,
He

Quoted 23 lines
Quoted 2 lines

I wrote a patch, not well tested, how do you think of it?

Returning the error number would be good to me.
Could you write and sumbit a complete one?

Thanks,

Quoted 15 lines
fs/f2fs/data.c +3 -2
index 9bedfa8..ba5d94c 100644
@@ -2010,8 +2010,9 @@ static ssize_t f2fs_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
         if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode))
                 return 0;
 
-       if (check_direct_IO(inode, iter, offset))
-               return 0;
+       err = check_direct_IO(inode, iter, offset)
+       if (err)
+               return -EINVAL;
 
Quoted 9 lines
Quoted 6 lines
fs/f2fs/data.c +0 -0
         trace_f2fs_direct_IO_enter(inode, offset, count, iov_iter_rw(iter));
 

I wish you and other developers in this list could help me in a correct way.

Thanks,
He

.

------------------------------------------------------------------------------
_______________________________________________
Linux-f2fs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

author
#4 lore raw
Quoted 75 lines · Chao Yu

Hi Yunlei,

Sorry for the long delay.

Quoted 63 lines

-----Original Message-----
From: He YunLei [mailto:[email protected]]
Sent: Tuesday, June 02, 2015 12:21 PM
To: Jaegeuk Kim
Cc: [email protected]; [email protected]
Subject: Re: [f2fs-dev] Dwrite with non-aligned offset and size

Quoted 37 lines · Jaegeuk Kim
Quoted 36 lines · He YunLei

Hi Jaegeuk,

We run ltp testcase with f2fs and obtain a TFAIL in diotest4, the result in detail is as fallow:

dio04

<<<test_start>>>
tag=dio04 stime=1432278894
cmdline="diotest4"
contacts=""
analysis=exit
<<<test_output>>>

diotest4    1  TPASS  :  Negative Offset
diotest4    2  TPASS  :  removed
diotest4    3  TFAIL  :  diotest4.c:129: write allows odd count.returns 1: Success
diotest4    4  TFAIL  :  diotest4.c:183: Odd count of read and write
diotest4    5  TPASS  :  Read beyond the file size

......

the result of ext4 with same environment:

dio04

<<<test_start>>>
tag=dio04 stime=1432259643
cmdline="diotest4"
contacts=""
analysis=exit
<<<test_output>>>

diotest4    1  TPASS  :  Negative Offset
diotest4    2  TPASS  :  removed
diotest4    3  TPASS  :  Odd count of read and write
diotest4    4  TPASS  :  Read beyond the file size

......

Quoted 7 lines
Quoted 3 lines

Does f2fs allow dwrite with non-aligned offset and size? I check the code and found dwrite with non-aligned offset and size will turn into buffered write. Whether it will have some impact on user layer applications?

It's not a big deal to return -EINVAL.
When I take a look at other filesystem behaviors, it seems there is no restriction.

Ext4 do a check in the function do_blockdev_direct_IO:

           if (align & blocksize_mask) {
               if (bdev)
                   blkbits = blksize_bits(bdev_logical_block_size(bdev));
               blocksize_mask = (1 << blkbits) - 1;
               if (align & blocksize_mask)
                   goto out;
           }
Quoted 1 line

It will return -EINVAL if the alignment is not satisfied.

I think we can get hint from the error case description in write(2) manual:

"EINVAL fd is attached to an object which is unsuitable for writing; or the file was opened with the O_DIRECT flag, and either the address specified in buf, the value specified in count, or the current file offset is not suitably aligned."

Quoted 7 lines

So if the alignment is not satisfied, we should return '-EINVAL' instead of letting user fall back to buffered write.

Do you have time to make and send us a patch for fixing this issue?

Thanks,

In man page of open(2)

The O_DIRECT flag may impose alignment restrictions on the length and address of user-space buffers and the file offset of I/Os. In Linux alignment restrictions vary by filesystem and kernel version and might be absent entirely. However there is currently no filesystem-independent interface for an application to discover these restrictions for a given file or filesystem.

So now I don't know which one is better, falling back to buffered write or not?

Quoted 42 lines
Quoted 42 lines

In f2fs, it do the check by check_direct_IO() before blockdev_direct_IO(). The difference between the two methods is whether turn dwrite with non-aligned offset and size into buffered write. I am not very clear which one is better!

Thanks,
He

Quoted 32 lines
Quoted 2 lines

I wrote a patch, not well tested, how do you think of it?

Returning the error number would be good to me.
Could you write and sumbit a complete one?

Thanks,

Quoted 19 lines
fs/f2fs/data.c +3 -2
index 9bedfa8..ba5d94c 100644
@@ -2010,8 +2010,9 @@ static ssize_t f2fs_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
          if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode))
                  return 0;
 
-       if (check_direct_IO(inode, iter, offset))
-               return 0;
+       err = check_direct_IO(inode, iter, offset)
+       if (err)
+               return -EINVAL;
 
          trace_f2fs_direct_IO_enter(inode, offset, count, iov_iter_rw(iter));
 

I wish you and other developers in this list could help me in a correct way.

Quoted 2 lines

Thanks,
He

.

Quoted 9 lines
Quoted 5 lines

------------------------------------------------------------------------------
_______________________________________________
Linux-f2fs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

.

------------------------------------------------------------------------------
Don't Limit Your Business. Reach for the Cloud.
GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today.
https://www.gigenetcloud.com/

#5 lore raw
Quoted 90 lines

-----Original Message-----
From: He YunLei [mailto:[email protected]]
Sent: Thursday, July 16, 2015 10:58 AM
To: Chao Yu
Cc: 'Jaegeuk Kim'; [email protected]; [email protected]
Subject: Re: [f2fs-dev] Dwrite with non-aligned offset and size

Quoted 75 lines · Chao Yu

Hi Yunlei,

Sorry for the long delay.

Quoted 63 lines

-----Original Message-----
From: He YunLei [mailto:[email protected]]
Sent: Tuesday, June 02, 2015 12:21 PM
To: Jaegeuk Kim
Cc: [email protected]; [email protected]
Subject: Re: [f2fs-dev] Dwrite with non-aligned offset and size

Quoted 37 lines · Jaegeuk Kim
Quoted 36 lines · He YunLei

Hi Jaegeuk,

We run ltp testcase with f2fs and obtain a TFAIL in diotest4, the result in detail is as fallow:

dio04

<<<test_start>>>
tag=dio04 stime=1432278894
cmdline="diotest4"
contacts=""
analysis=exit
<<<test_output>>>

diotest4    1  TPASS  :  Negative Offset
diotest4    2  TPASS  :  removed
diotest4    3  TFAIL  :  diotest4.c:129: write allows odd count.returns 1: Success
diotest4    4  TFAIL  :  diotest4.c:183: Odd count of read and write
diotest4    5  TPASS  :  Read beyond the file size

......

the result of ext4 with same environment:

dio04

<<<test_start>>>
tag=dio04 stime=1432259643
cmdline="diotest4"
contacts=""
analysis=exit
<<<test_output>>>

diotest4    1  TPASS  :  Negative Offset
diotest4    2  TPASS  :  removed
diotest4    3  TPASS  :  Odd count of read and write
diotest4    4  TPASS  :  Read beyond the file size

......

Quoted 7 lines
Quoted 3 lines

Does f2fs allow dwrite with non-aligned offset and size? I check the code and found dwrite with non-aligned offset and size will turn into buffered write. Whether it will have some impact on user layer applications?

It's not a big deal to return -EINVAL.
When I take a look at other filesystem behaviors, it seems there is no restriction.

Ext4 do a check in the function do_blockdev_direct_IO:

           if (align & blocksize_mask) {
               if (bdev)
                   blkbits = blksize_bits(bdev_logical_block_size(bdev));
               blocksize_mask = (1 << blkbits) - 1;
               if (align & blocksize_mask)
                   goto out;
           }
Quoted 1 line

It will return -EINVAL if the alignment is not satisfied.

I think we can get hint from the error case description in write(2) manual:

"EINVAL fd is attached to an object which is unsuitable for writing; or the file was opened with the O_DIRECT flag, and either the address specified in buf, the value specified in count, or the current file offset is not suitably aligned."

Quoted 7 lines

So if the alignment is not satisfied, we should return '-EINVAL' instead of letting user fall back to buffered write.

Do you have time to make and send us a patch for fixing this issue?

Thanks,

Quoted 8 lines

In man page of open(2)

The O_DIRECT flag may impose alignment restrictions on the length and address of user-space buffers and the file offset of I/Os. In Linux alignment restrictions vary by filesystem and kernel version and might be absent entirely. However there is currently no filesystem-independent interface for an application to discover these restrictions for a given file or filesystem.

Right, there is no such interface here, and I don't think we need this, since developers already have the manual for guiding them to use the interfaces of filesystem correctly. Manuals should describe the restrictions of interfaces which is exported by VFS, if some filesystems did not follow theses restricting rules, we should add related description in manual.

Quoted 2 lines

So now I don't know which one is better, falling back to buffered write or not?

IMO, DIO is used for special purpose, if the target if not be achieved because of wrong parameters, why not feedback to our user as write(2) manual descripted?

Any special reason to fallback to buffered write in f2fs?

Thanks,

Quoted 52 lines
Quoted 14 lines
Quoted 14 lines

In f2fs, it do the check by check_direct_IO() before blockdev_direct_IO(). The difference between the two methods is whether turn dwrite with non-aligned offset and size into buffered write. I am not very clear which one is better!

Thanks,
He

Quoted 6 lines
Quoted 2 lines

I wrote a patch, not well tested, how do you think of it?

Returning the error number would be good to me.
Could you write and sumbit a complete one?

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

Thanks,

fs/f2fs/data.c static ssize_t f2fs_direct_IO(struct kiocb *iocb, struct iov_iter

Quoted 6 lines
fs/f2fs/data.c +0 -0
index 9bedfa8..ba5d94c 100644
@@ -2010,8 +2010,9 @@ static ssize_t f2fs_direct_IO(struct kiocb *iocb, struct iov_iter

*iter,

Quoted 29 lines
Quoted 20 lines
Quoted 9 lines
Quoted 9 lines
fs/f2fs/data.c +3 -2
          if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode))
                  return 0;
 
-       if (check_direct_IO(inode, iter, offset))
-               return 0;
+       err = check_direct_IO(inode, iter, offset)
+       if (err)
+               return -EINVAL;
 
Quoted 9 lines
Quoted 6 lines
fs/f2fs/data.c +0 -0
          trace_f2fs_direct_IO_enter(inode, offset, count, iov_iter_rw(iter));
 

I wish you and other developers in this list could help me in a correct way.

Thanks,
He

.

Quoted 5 lines

------------------------------------------------------------------------------
_______________________________________________
Linux-f2fs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

.

maintainer
#6 lore raw

Hi,

Quoted 106 lines · Chao Yu
Quoted 90 lines

-----Original Message-----
From: He YunLei [mailto:[email protected]]
Sent: Thursday, July 16, 2015 10:58 AM
To: Chao Yu
Cc: 'Jaegeuk Kim'; [email protected]; [email protected]
Subject: Re: [f2fs-dev] Dwrite with non-aligned offset and size

Quoted 75 lines · Chao Yu

Hi Yunlei,

Sorry for the long delay.

Quoted 63 lines

-----Original Message-----
From: He YunLei [mailto:[email protected]]
Sent: Tuesday, June 02, 2015 12:21 PM
To: Jaegeuk Kim
Cc: [email protected]; [email protected]
Subject: Re: [f2fs-dev] Dwrite with non-aligned offset and size

Quoted 37 lines · Jaegeuk Kim
Quoted 36 lines · He YunLei

Hi Jaegeuk,

We run ltp testcase with f2fs and obtain a TFAIL in diotest4, the result in detail is as fallow:

dio04

<<<test_start>>>
tag=dio04 stime=1432278894
cmdline="diotest4"
contacts=""
analysis=exit
<<<test_output>>>

diotest4    1  TPASS  :  Negative Offset
diotest4    2  TPASS  :  removed
diotest4    3  TFAIL  :  diotest4.c:129: write allows odd count.returns 1: Success
diotest4    4  TFAIL  :  diotest4.c:183: Odd count of read and write
diotest4    5  TPASS  :  Read beyond the file size

......

the result of ext4 with same environment:

dio04

<<<test_start>>>
tag=dio04 stime=1432259643
cmdline="diotest4"
contacts=""
analysis=exit
<<<test_output>>>

diotest4    1  TPASS  :  Negative Offset
diotest4    2  TPASS  :  removed
diotest4    3  TPASS  :  Odd count of read and write
diotest4    4  TPASS  :  Read beyond the file size

......

Quoted 7 lines
Quoted 3 lines

Does f2fs allow dwrite with non-aligned offset and size? I check the code and found dwrite with non-aligned offset and size will turn into buffered write. Whether it will have some impact on user layer applications?

It's not a big deal to return -EINVAL.
When I take a look at other filesystem behaviors, it seems there is no restriction.

Ext4 do a check in the function do_blockdev_direct_IO:

           if (align & blocksize_mask) {
               if (bdev)
                   blkbits = blksize_bits(bdev_logical_block_size(bdev));
               blocksize_mask = (1 << blkbits) - 1;
               if (align & blocksize_mask)
                   goto out;
           }
Quoted 1 line

It will return -EINVAL if the alignment is not satisfied.

I think we can get hint from the error case description in write(2) manual:

"EINVAL fd is attached to an object which is unsuitable for writing; or the file was opened with the O_DIRECT flag, and either the address specified in buf, the value specified in count, or the current file offset is not suitably aligned."

Quoted 7 lines

So if the alignment is not satisfied, we should return '-EINVAL' instead of letting user fall back to buffered write.

Do you have time to make and send us a patch for fixing this issue?

Thanks,

Quoted 8 lines

In man page of open(2)

The O_DIRECT flag may impose alignment restrictions on the length and address of user-space buffers and the file offset of I/Os. In Linux alignment restrictions vary by filesystem and kernel version and might be absent entirely. However there is currently no filesystem-independent interface for an application to discover these restrictions for a given file or filesystem.

Right, there is no such interface here, and I don't think we need this, since developers already have the manual for guiding them to use the interfaces of filesystem correctly. Manuals should describe the restrictions of interfaces which is exported by VFS, if some filesystems did not follow theses restricting rules, we should add related description in manual.

Quoted 1 line
Quoted 6 lines
Quoted 1 line

So now I don't know which one is better, falling back to buffered write or not?

IMO, DIO is used for special purpose, if the target if not be achieved because of wrong parameters, why not feedback to our user as write(2) manual descripted?

Any special reason to fallback to buffered write in f2fs?

I think it'd be better to return -EINVAL, when unaligned offset was given, since what users try to request dios means they want to avoid the filesystem layer and get the maximum storage performance.
So, we need to alarm them to guide correct settings.

The only case that we need to fall back to buffered write would be some specific conditions like inline_data and encryption.

Thanks,

Quoted 55 lines

Thanks,

Quoted 52 lines
Quoted 14 lines
Quoted 14 lines

In f2fs, it do the check by check_direct_IO() before blockdev_direct_IO(). The difference between the two methods is whether turn dwrite with non-aligned offset and size into buffered write. I am not very clear which one is better!

Thanks,
He

Quoted 6 lines
Quoted 2 lines

I wrote a patch, not well tested, how do you think of it?

Returning the error number would be good to me.
Could you write and sumbit a complete one?

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

Thanks,

fs/f2fs/data.c static ssize_t f2fs_direct_IO(struct kiocb *iocb, struct iov_iter

Quoted 6 lines
fs/f2fs/data.c +0 -0
index 9bedfa8..ba5d94c 100644
@@ -2010,8 +2010,9 @@ static ssize_t f2fs_direct_IO(struct kiocb *iocb, struct iov_iter

*iter,

Quoted 29 lines
Quoted 20 lines
Quoted 9 lines
Quoted 9 lines
fs/f2fs/data.c +3 -2
          if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode))
                  return 0;
 
-       if (check_direct_IO(inode, iter, offset))
-               return 0;
+       err = check_direct_IO(inode, iter, offset)
+       if (err)
+               return -EINVAL;
 
Quoted 9 lines
Quoted 6 lines
fs/f2fs/data.c +0 -0
          trace_f2fs_direct_IO_enter(inode, offset, count, iov_iter_rw(iter));
 

I wish you and other developers in this list could help me in a correct way.

Thanks,
He

.

Quoted 5 lines

------------------------------------------------------------------------------
_______________________________________________
Linux-f2fs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

.