Trond Myklebust wrote:
> On Mon, 2008-02-04 at 19:04 -0500, Chuck Lever wrote:
>
>> Refresh the help text for Kconfig items related to the NFS client. Remove
>> obsolete URLs, and make the language consistent among the options.
>>
>>
>> config NFS_DIRECTIO
>> bool "Allow direct I/O on NFS files"
>> depends on NFS_FS
>> + default N
>>
>
> Why? Isn't O_DIRECT pretty much a standard feature that is enabled in
> most distros at this time?
> IMO we should rather be aiming to phase out NFS_DIRECTIO.
>
Agreed.
>
>> help
>> - This option enables applications to perform uncached I/O on files
>> - in NFS file systems using the O_DIRECT open() flag. When O_DIRECT
>> - is set for a file, its data is not cached in the system's page
>> - cache. Data is moved to and from user-level application buffers
>> - directly. Unlike local disk-based file systems, NFS O_DIRECT has
>> - no alignment restrictions.
>> + This option enables support for uncached I/O on files accessed
>> + via an NFS mount point. Applications request direct I/O by
>> + setting the O_DIRECT flag when opening a file.
>> +
>> + When direct I/O is requested, the NFS client moves data to and
>> + from user-level application buffers directly to NFS servers.
>> + Direct I/O on NFS files does not suffer from any I/O or buffer
>> + alignment restrictions, as does direct I/O on files that reside
>> + in local file systems.
>>
It might be an idea to mention that direct IO does *not* bypass the
server's page cache, an unobvious semantic which escapes many people.
>> +
>> + For details, see the open(2) man page.
>>
FYI, I recently posted an update to the O_DIRECT documentation in open(2).
http://marc.info/?l=linux-man&m=120107005026531&w=2
--
Greg Banks, R&D Software Engineer, SGI Australian Software Group.
The cake is *not* a lie.
I don't speak for SGI.