Reference to file size in nfsd_create_v3

Reference to file size in nfsd_create_v3

[Shehjar Tikoo]

Hi All

I am looking at the fs/nfsd/vfs.c:nfsd_create_v3 function.
In there, a comment says:
"furthermore, if the size is nonzero, we should ignore it
according to spec!"

Could someone please point out the section in RFC1813 where this
  particular point is specified?

Thanks
Shehjar

Re: Reference to file size in nfsd_create_v3

[J. Bruce Fields]

On Fri, Jun 26, 2009 at 02:40:12PM +0530, Shehjar Tikoo wrote:
> Hi All
>
> I am looking at the fs/nfsd/vfs.c:nfsd_create_v3 function.
> In there, a comment says:
> "furthermore, if the size is nonzero, we should ignore it
> according to spec!"
>
> Could someone please point out the section in RFC1813 where this
>  particular point is specified?

It's referring to the third paragraph of the DESCRIPTION section of the
OPEN operation (14.2.16) in rfc 3530.

--b.

Re: Reference to file size in nfsd_create_v3

[Shehjar Tikoo]

J. Bruce Fields wrote:
> On Fri, Jun 26, 2009 at 02:40:12PM +0530, Shehjar Tikoo wrote:
>> Hi All
>> 
>> I am looking at the fs/nfsd/vfs.c:nfsd_create_v3 function. In
>> there, a comment says: "furthermore, if the size is nonzero, we
>> should ignore it according to spec!"
>> 
>> Could someone please point out the section in RFC1813 where this 
>> particular point is specified?
> 
> It's referring to the third paragraph of the DESCRIPTION section of
> the OPEN operation (14.2.16) in rfc 3530.
> 
> --b.
Ok.

For NFSv3, is there a specified way to handle create or mkdir ops where
  the size is non-zero? The reason this came up is that I was testing
unfs3 with SpecSFS2k8 and an MKDIR op failed in unfs3 because SFS sent
the mkdir call with non-zero size in the attributes. I just wanted to
see how Linux nfsd handled it.


Thanks
-Shehjar

Re: Reference to file size in nfsd_create_v3

[Peter Staubach]

Shehjar Tikoo wrote:
> J. Bruce Fields wrote:
>> On Fri, Jun 26, 2009 at 02:40:12PM +0530, Shehjar Tikoo wrote:
>>> Hi All
>>>
>>> I am looking at the fs/nfsd/vfs.c:nfsd_create_v3 function. In
>>> there, a comment says: "furthermore, if the size is nonzero, we
>>> should ignore it according to spec!"
>>>
>>> Could someone please point out the section in RFC1813 where this
>>> particular point is specified?
>>
>> It's referring to the third paragraph of the DESCRIPTION section of
>> the OPEN operation (14.2.16) in rfc 3530.
>>
>> --b.
> Ok.
> 
> For NFSv3, is there a specified way to handle create or mkdir ops where
>  the size is non-zero? The reason this came up is that I was testing
> unfs3 with SpecSFS2k8 and an MKDIR op failed in unfs3 because SFS sent
> the mkdir call with non-zero size in the attributes. I just wanted to
> see how Linux nfsd handled it.
> 

Most servers in the market just ignore the size field for
MKDIR requests.  They also ignore the size field, unless it is
0, for CREATE requests.

Setting the size on a directory does not make sense and the
usual decision point for file creation is whether to truncate
the file to empty or not.

		ps

Re: Reference to file size in nfsd_create_v3

[J. Bruce Fields]

On Wed, Aug 26, 2009 at 09:38:46AM -0400, Peter Staubach wrote:
> Shehjar Tikoo wrote:
> > J. Bruce Fields wrote:
> >> On Fri, Jun 26, 2009 at 02:40:12PM +0530, Shehjar Tikoo wrote:
> >>> Hi All
> >>>
> >>> I am looking at the fs/nfsd/vfs.c:nfsd_create_v3 function. In
> >>> there, a comment says: "furthermore, if the size is nonzero, we
> >>> should ignore it according to spec!"
> >>>
> >>> Could someone please point out the section in RFC1813 where this
> >>> particular point is specified?
> >>
> >> It's referring to the third paragraph of the DESCRIPTION section of
> >> the OPEN operation (14.2.16) in rfc 3530.
> >>
> >> --b.
> > Ok.
> > 
> > For NFSv3, is there a specified way to handle create or mkdir ops where
> >  the size is non-zero? The reason this came up is that I was testing
> > unfs3 with SpecSFS2k8 and an MKDIR op failed in unfs3 because SFS sent
> > the mkdir call with non-zero size in the attributes. I just wanted to
> > see how Linux nfsd handled it.

But you've only seen the problem against unfs3, not against the kernel
nfsd?

> 
> Most servers in the market just ignore the size field for
> MKDIR requests.  They also ignore the size field, unless it is
> 0, for CREATE requests.

So it's probably a bug on both sides.  (SpecSFS shouldn't be sending a
non-zero size either.)

--b.

> 
> Setting the size on a directory does not make sense and the
> usual decision point for file creation is whether to truncate
> the file to empty or not.
> 
> 		ps
> --
> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Reference to file size in nfsd_create_v3

[Shehjar Tikoo]

J. Bruce Fields wrote:
> On Wed, Aug 26, 2009 at 09:38:46AM -0400, Peter Staubach wrote:
>> Shehjar Tikoo wrote:
>>> J. Bruce Fields wrote:
>>>> On Fri, Jun 26, 2009 at 02:40:12PM +0530, Shehjar Tikoo wrote:
>>>>> Hi All
>>>>>
>>>>> I am looking at the fs/nfsd/vfs.c:nfsd_create_v3 function. In
>>>>> there, a comment says: "furthermore, if the size is nonzero, we
>>>>> should ignore it according to spec!"
>>>>>
>>>>> Could someone please point out the section in RFC1813 where this
>>>>> particular point is specified?
>>>> It's referring to the third paragraph of the DESCRIPTION section of
>>>> the OPEN operation (14.2.16) in rfc 3530.
>>>>
>>>> --b.
>>> Ok.
>>>
>>> For NFSv3, is there a specified way to handle create or mkdir ops where
>>>  the size is non-zero? The reason this came up is that I was testing
>>> unfs3 with SpecSFS2k8 and an MKDIR op failed in unfs3 because SFS sent
>>> the mkdir call with non-zero size in the attributes. I just wanted to
>>> see how Linux nfsd handled it.
> 
> But you've only seen the problem against unfs3, not against the kernel
> nfsd?
> 

I didnt test SFS against the kernel nfsd so cant comment on it.

>> Most servers in the market just ignore the size field for
>> MKDIR requests.  They also ignore the size field, unless it is
>> 0, for CREATE requests.
> 
> So it's probably a bug on both sides.  (SpecSFS shouldn't be sending a
> non-zero size either.)
> 

Yes, I fixed the problem in unfs3 by making it ignore the size in
the MKDIR request. SpecSFS might need further investigation.

Thanks
-Shehjar

> --b.
> 
>> Setting the size on a directory does not make sense and the
>> usual decision point for file creation is whether to truncate
>> the file to empty or not.
>>
>> 		ps
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html