Re: [PATCH 24/44] docs: filesystems: convert inotify.txt to ReST

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Mon 17-02-20 17:12:10, Mauro Carvalho Chehab wrote:
> - Add a SPDX header;
> - Add a document title;
> - Adjust document title;
> - Fix list markups;
> - Some whitespace fixes and new line breaks;
> - Add it to filesystems/index.rst.
> 
> Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@xxxxxxxxxx>

Thanks. You can add:

Acked-by: Jan Kara <jack@xxxxxxx>

or tell me if you want me to pick up this patch.

								Honza

> ---
>  Documentation/filesystems/index.rst           |  1 +
>  .../filesystems/{inotify.txt => inotify.rst}  | 33 ++++++++++++-------
>  2 files changed, 23 insertions(+), 11 deletions(-)
>  rename Documentation/filesystems/{inotify.txt => inotify.rst} (83%)
> 
> diff --git a/Documentation/filesystems/index.rst b/Documentation/filesystems/index.rst
> index 3fbe2fa0b5c5..5a737722652c 100644
> --- a/Documentation/filesystems/index.rst
> +++ b/Documentation/filesystems/index.rst
> @@ -70,6 +70,7 @@ Documentation for filesystem implementations.
>     hfs
>     hfsplus
>     hpfs
> +   inotify
>     fuse
>     overlayfs
>     virtiofs
> diff --git a/Documentation/filesystems/inotify.txt b/Documentation/filesystems/inotify.rst
> similarity index 83%
> rename from Documentation/filesystems/inotify.txt
> rename to Documentation/filesystems/inotify.rst
> index 51f61db787fb..7f7ef8af0e1e 100644
> --- a/Documentation/filesystems/inotify.txt
> +++ b/Documentation/filesystems/inotify.rst
> @@ -1,27 +1,36 @@
> -				   inotify
> -	    a powerful yet simple file change notification system
> +.. SPDX-License-Identifier: GPL-2.0
> +
> +===============================================================
> +Inotify - A Powerful yet Simple File Change Notification System
> +===============================================================
>  
>  
>  
>  Document started 15 Mar 2005 by Robert Love <rml@xxxxxxxxxx>
> +
>  Document updated 4 Jan 2015 by Zhang Zhen <zhenzhang.zhang@xxxxxxxxxx>
> -	--Deleted obsoleted interface, just refer to manpages for user interface.
> +
> +	- Deleted obsoleted interface, just refer to manpages for user interface.
>  
>  (i) Rationale
>  
> -Q: What is the design decision behind not tying the watch to the open fd of
> +Q:
> +   What is the design decision behind not tying the watch to the open fd of
>     the watched object?
>  
> -A: Watches are associated with an open inotify device, not an open file.
> +A:
> +   Watches are associated with an open inotify device, not an open file.
>     This solves the primary problem with dnotify: keeping the file open pins
>     the file and thus, worse, pins the mount.  Dnotify is therefore infeasible
>     for use on a desktop system with removable media as the media cannot be
>     unmounted.  Watching a file should not require that it be open.
>  
> -Q: What is the design decision behind using an-fd-per-instance as opposed to
> +Q:
> +   What is the design decision behind using an-fd-per-instance as opposed to
>     an fd-per-watch?
>  
> -A: An fd-per-watch quickly consumes more file descriptors than are allowed,
> +A:
> +   An fd-per-watch quickly consumes more file descriptors than are allowed,
>     more fd's than are feasible to manage, and more fd's than are optimally
>     select()-able.  Yes, root can bump the per-process fd limit and yes, users
>     can use epoll, but requiring both is a silly and extraneous requirement.
> @@ -29,8 +38,8 @@ A: An fd-per-watch quickly consumes more file descriptors than are allowed,
>     spaces is thus sensible.  The current design is what user-space developers
>     want: Users initialize inotify, once, and add n watches, requiring but one
>     fd and no twiddling with fd limits.  Initializing an inotify instance two
> -   thousand times is silly.  If we can implement user-space's preferences 
> -   cleanly--and we can, the idr layer makes stuff like this trivial--then we 
> +   thousand times is silly.  If we can implement user-space's preferences
> +   cleanly--and we can, the idr layer makes stuff like this trivial--then we
>     should.
>  
>     There are other good arguments.  With a single fd, there is a single
> @@ -65,9 +74,11 @@ A: An fd-per-watch quickly consumes more file descriptors than are allowed,
>     need not be a one-fd-per-process mapping; it is one-fd-per-queue and a
>     process can easily want more than one queue.
>  
> -Q: Why the system call approach?
> +Q:
> +   Why the system call approach?
>  
> -A: The poor user-space interface is the second biggest problem with dnotify.
> +A:
> +   The poor user-space interface is the second biggest problem with dnotify.
>     Signals are a terrible, terrible interface for file notification.  Or for
>     anything, for that matter.  The ideal solution, from all perspectives, is a
>     file descriptor-based one that allows basic file I/O and poll/select.
> -- 
> 2.24.1
> 
-- 
Jan Kara <jack@xxxxxxxx>
SUSE Labs, CR



[Index of Archives]     [Linux Ext4 Filesystem]     [Union Filesystem]     [Filesystem Testing]     [Ceph Users]     [Ecryptfs]     [AutoFS]     [Kernel Newbies]     [Share Photos]     [Security]     [Netfilter]     [Bugtraq]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux Cachefs]     [Reiser Filesystem]     [Linux RAID]     [Samba]     [Device Mapper]     [CEPH Development]

  Powered by Linux