public inbox for [email protected]
 help / color / mirror / Atom feed
From: Stefan Roesch <[email protected]>
To: Clay Harris <[email protected]>
Cc: <[email protected]>, <[email protected]>
Subject: Re: [PATCH v2 0/3] io_uring: add getdents64 support
Date: Tue, 30 Nov 2021 22:01:30 -0800	[thread overview]
Message-ID: <[email protected]> (raw)
In-Reply-To: <[email protected]>



On 11/24/21 8:42 PM, Clay Harris wrote:
> 
> I seem to recall making a few comments the last time a getdents64
> for io_uring was proposed; in particular I wanted to bring up one
> here.  This applies only to altering the internal interface, which
> io_uring would use, although wiring up a new syscall would be a nice
> addition.
> 
> The current interface has 2 issues:
> 
> 1)
> getdents64 requires at least two calls to read a directory.
> One or more to get the dents and a final call to see the EOF.
> With small directories, this means literally 50% of the calls
> are wasted.
> 
> 2)
> The fpos cannot be changed atomically with a read, so it is not
> possible to to safely perform concurrent reads on the same fd.
> 
> But, the kernel knows (most, if not all of the time) that it is at
> EOF at the time it returns the last buffer.  So, it would be very
> useful to get an EOF indicator back with the final buffer.  This
> could just a flag, or for instance make an fpos parameter which is
> both input and output, returning the (post read) fpos or zero at
> EOF.
> 
> Futhermore, for input, one could supply:
> 	0:	Start from directory beginning
> 	-1:	Read from current position
> 	other:	(output from previous call) Read from here
> 

While I can understand the wish to optimize the getdents call, this
has its own set of challenges:

- The getdents API is following the logic of other read API's. None
  of these API's has the logic you described above. This would be
  inconsistent.
- The eof needs to be stored in another field. The dirent structure
  does not have space in the field, so a new data structure needs to be defined.
- However the goal is to provide a familiar interface to the user.
- If the user wants to reduce the number of calls he can still provide
  a bigger user buffer.

> On Wed, Nov 24 2021 at 15:16:57 -0800, Stefan Roesch quoth thus:
> 
>> This series adds support for getdents64 in liburing. The intent is to
>> provide a more complete I/O interface for io_uring.
>>
>> Patch 1: fs: add parameter use_fpos to iterate_dir()
>>   This adds a new parameter to the function iterate_dir() so the
>>   caller can specify if the position is the file position or the
>>   position stored in the buffer context.
>>
>> Patch 2: fs: split off vfs_getdents function from getdents64 system call
>>   This splits of the iterate_dir part of the syscall in its own
>>   dedicated function. This allows to call the function directly from
>>   liburing.
>>
>> Patch 3: io_uring: add support for getdents64
>>   Adds the functions to io_uring to support getdents64.
>>
>> There is also a patch series for the changes to liburing. This includes
>> a new test. The patch series is called "liburing: add getdents support."
>>
>> The following tests have been performed:
>> - new liburing getdents test program has been run
>> - xfstests have been run
>> - both tests have been repeated with the kernel memory leak checker
>>   and no leaks have been reported.
>>
>> Signed-off-by: Stefan Roesch <[email protected]>
>> ---
>> V2: Updated the iterate_dir calls in fs/ksmbd, fs/ecryptfs and arch/alpha with
>>     the additional parameter.
>>
>> Stefan Roesch (3):
>>   fs: add parameter use_fpos to iterate_dir function
>>   fs: split off vfs_getdents function of getdents64 syscall
>>   io_uring: add support for getdents64
>>
>>  arch/alpha/kernel/osf_sys.c   |  2 +-
>>  fs/ecryptfs/file.c            |  2 +-
>>  fs/exportfs/expfs.c           |  2 +-
>>  fs/internal.h                 |  8 +++++
>>  fs/io_uring.c                 | 52 ++++++++++++++++++++++++++++
>>  fs/ksmbd/smb2pdu.c            |  2 +-
>>  fs/ksmbd/vfs.c                |  4 +--
>>  fs/nfsd/nfs4recover.c         |  2 +-
>>  fs/nfsd/vfs.c                 |  2 +-
>>  fs/overlayfs/readdir.c        |  6 ++--
>>  fs/readdir.c                  | 64 ++++++++++++++++++++++++++---------
>>  include/linux/fs.h            |  2 +-
>>  include/uapi/linux/io_uring.h |  1 +
>>  13 files changed, 121 insertions(+), 28 deletions(-)
>>
>>
>> base-commit: f0afafc21027c39544a2c1d889b0cff75b346932
>> -- 
>> 2.30.2

  reply	other threads:[~2021-12-01  6:01 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-11-24 23:16 [PATCH v2 0/3] io_uring: add getdents64 support Stefan Roesch
2021-11-24 23:16 ` [PATCH v2 1/3] fs: add parameter use_fpos to iterate_dir function Stefan Roesch
2021-11-24 23:16 ` [PATCH v2 2/3] fs: split off vfs_getdents function of getdents64 syscall Stefan Roesch
2021-11-24 23:17 ` [PATCH v2 3/3] io_uring: add support for getdents64 Stefan Roesch
2021-11-25  4:42 ` [PATCH v2 0/3] io_uring: add getdents64 support Clay Harris
2021-12-01  6:01   ` Stefan Roesch [this message]
2021-12-01  7:11     ` Clay Harris
2021-12-15 21:09 ` Jens Axboe
2021-12-31 23:27   ` Al Viro

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox