diff mbox

nfs: RPC_MAX_AUTH_SIZE is in bytes

Message ID 20170920164213.GF14329@fieldses.org (mailing list archive)
State New, archived
Headers show

Commit Message

J. Bruce Fields Sept. 20, 2017, 4:42 p.m. UTC
From: "J. Bruce Fields" <bfields@redhat.com>

The units of RPC_MAX_AUTH_SIZE is bytes, not 4-byte words.  This causes
the client to request a larger-than-necessary session replay slot size.

Signed-off-by: J. Bruce Fields <bfields@redhat.com>
---
 fs/nfs/nfs4xdr.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

Comments

J. Bruce Fields Sept. 20, 2017, 4:49 p.m. UTC | #1
On Wed, Sep 20, 2017 at 12:42:13PM -0400, bfields wrote:
> From: "J. Bruce Fields" <bfields@redhat.com>
> 
> The units of RPC_MAX_AUTH_SIZE is bytes, not 4-byte words.  This causes
> the client to request a larger-than-necessary session replay slot size.

By the way, the client's still asking for 3428 bytes after that, which
seems high.  It's mostly the fault of NFS4_MAXLABELLEN, which is 2048.
I haven't spotted an improvement there yet.

Noticed because knfsd was hitting session drc cache limits much to early
and failing CREATE_SESSION (hence mount).  The main trouble is on the
server side, so I'm relaxing the limits there (and revisiting Trond's
dynamic slot renegotiation patches).  But I thought I should check for
any easy improvements here too.

--b.

> 
> Signed-off-by: J. Bruce Fields <bfields@redhat.com>
> ---
>  fs/nfs/nfs4xdr.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c
> index 37c8af003275..14ed9791ec9c 100644
> --- a/fs/nfs/nfs4xdr.c
> +++ b/fs/nfs/nfs4xdr.c
> @@ -1842,8 +1842,8 @@ static void encode_create_session(struct xdr_stream *xdr,
>  	 * Assumes OPEN is the biggest non-idempotent compound.
>  	 * 2 is the verifier.
>  	 */
> -	max_resp_sz_cached = (NFS4_dec_open_sz + RPC_REPHDRSIZE +
> -			      RPC_MAX_AUTH_SIZE + 2) * XDR_UNIT;
> +	max_resp_sz_cached = (NFS4_dec_open_sz + RPC_REPHDRSIZE + 2)
> +				* XDR_UNIT + RPC_MAX_AUTH_SIZE;
>  
>  	encode_op_hdr(xdr, OP_CREATE_SESSION, decode_create_session_maxsz, hdr);
>  	p = reserve_space(xdr, 16 + 2*28 + 20 + clnt->cl_nodelen + 12);
> -- 
> 2.13.5
> 
--
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
Chuck Lever Sept. 20, 2017, 4:56 p.m. UTC | #2
> On Sep 20, 2017, at 12:49 PM, J. Bruce Fields <bfields@fieldses.org> wrote:
> 
> On Wed, Sep 20, 2017 at 12:42:13PM -0400, bfields wrote:
>> From: "J. Bruce Fields" <bfields@redhat.com>
>> 
>> The units of RPC_MAX_AUTH_SIZE is bytes, not 4-byte words.  This causes
>> the client to request a larger-than-necessary session replay slot size.
> 
> By the way, the client's still asking for 3428 bytes after that, which
> seems high.  It's mostly the fault of NFS4_MAXLABELLEN, which is 2048.

The maximum size of NFSv4 LOOKUP replies is also quite large, thanks
to label support (IIRC). It would be nicer for NFS/RDMA, at least, if
the maximum reply sizes of common operations were small (say, less
than 1024 bytes).


> I haven't spotted an improvement there yet.

I haven't either.


> Noticed because knfsd was hitting session drc cache limits much to early
> and failing CREATE_SESSION (hence mount).  The main trouble is on the
> server side, so I'm relaxing the limits there (and revisiting Trond's
> dynamic slot renegotiation patches).  But I thought I should check for
> any easy improvements here too.
> 
> --b.
> 
>> 
>> Signed-off-by: J. Bruce Fields <bfields@redhat.com>
>> ---
>> fs/nfs/nfs4xdr.c | 4 ++--
>> 1 file changed, 2 insertions(+), 2 deletions(-)
>> 
>> diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c
>> index 37c8af003275..14ed9791ec9c 100644
>> --- a/fs/nfs/nfs4xdr.c
>> +++ b/fs/nfs/nfs4xdr.c
>> @@ -1842,8 +1842,8 @@ static void encode_create_session(struct xdr_stream *xdr,
>> 	 * Assumes OPEN is the biggest non-idempotent compound.
>> 	 * 2 is the verifier.
>> 	 */
>> -	max_resp_sz_cached = (NFS4_dec_open_sz + RPC_REPHDRSIZE +
>> -			      RPC_MAX_AUTH_SIZE + 2) * XDR_UNIT;
>> +	max_resp_sz_cached = (NFS4_dec_open_sz + RPC_REPHDRSIZE + 2)
>> +				* XDR_UNIT + RPC_MAX_AUTH_SIZE;
>> 
>> 	encode_op_hdr(xdr, OP_CREATE_SESSION, decode_create_session_maxsz, hdr);
>> 	p = reserve_space(xdr, 16 + 2*28 + 20 + clnt->cl_nodelen + 12);
>> -- 
>> 2.13.5
>> 
> --
> 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

--
Chuck Lever



--
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
J. Bruce Fields Sept. 20, 2017, 6:15 p.m. UTC | #3
On Wed, Sep 20, 2017 at 12:56:07PM -0400, Chuck Lever wrote:
> 
> > On Sep 20, 2017, at 12:49 PM, J. Bruce Fields <bfields@fieldses.org> wrote:
> > 
> > On Wed, Sep 20, 2017 at 12:42:13PM -0400, bfields wrote:
> >> From: "J. Bruce Fields" <bfields@redhat.com>
> >> 
> >> The units of RPC_MAX_AUTH_SIZE is bytes, not 4-byte words.  This causes
> >> the client to request a larger-than-necessary session replay slot size.
> > 
> > By the way, the client's still asking for 3428 bytes after that, which
> > seems high.  It's mostly the fault of NFS4_MAXLABELLEN, which is 2048.
> 
> The maximum size of NFSv4 LOOKUP replies is also quite large, thanks
> to label support (IIRC). It would be nicer for NFS/RDMA, at least, if
> the maximum reply sizes of common operations were small (say, less
> than 1024 bytes).
> 
> 
> > I haven't spotted an improvement there yet.
> 
> I haven't either.

Session (and I assume RDMA) limits have to be negotiated pretty early,
before we know server capabilities.  And even if we renegotiated
later--I think security label support is per-filesystem, while the
connection parameters are server-wide.

So, I dunno, all I can think of is putting off fetching the security
label till it's needed, but maybe that's not practical--I don't know
where it's used.

Well, whatever, we can live with it.

--b.

> 
> 
> > Noticed because knfsd was hitting session drc cache limits much to early
> > and failing CREATE_SESSION (hence mount).  The main trouble is on the
> > server side, so I'm relaxing the limits there (and revisiting Trond's
> > dynamic slot renegotiation patches).  But I thought I should check for
> > any easy improvements here too.
> > 
> > --b.
> > 
> >> 
> >> Signed-off-by: J. Bruce Fields <bfields@redhat.com>
> >> ---
> >> fs/nfs/nfs4xdr.c | 4 ++--
> >> 1 file changed, 2 insertions(+), 2 deletions(-)
> >> 
> >> diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c
> >> index 37c8af003275..14ed9791ec9c 100644
> >> --- a/fs/nfs/nfs4xdr.c
> >> +++ b/fs/nfs/nfs4xdr.c
> >> @@ -1842,8 +1842,8 @@ static void encode_create_session(struct xdr_stream *xdr,
> >> 	 * Assumes OPEN is the biggest non-idempotent compound.
> >> 	 * 2 is the verifier.
> >> 	 */
> >> -	max_resp_sz_cached = (NFS4_dec_open_sz + RPC_REPHDRSIZE +
> >> -			      RPC_MAX_AUTH_SIZE + 2) * XDR_UNIT;
> >> +	max_resp_sz_cached = (NFS4_dec_open_sz + RPC_REPHDRSIZE + 2)
> >> +				* XDR_UNIT + RPC_MAX_AUTH_SIZE;
> >> 
> >> 	encode_op_hdr(xdr, OP_CREATE_SESSION, decode_create_session_maxsz, hdr);
> >> 	p = reserve_space(xdr, 16 + 2*28 + 20 + clnt->cl_nodelen + 12);
> >> -- 
> >> 2.13.5
> >> 
> > --
> > 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
> 
> --
> Chuck Lever
> 
> 
--
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
diff mbox

Patch

diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c
index 37c8af003275..14ed9791ec9c 100644
--- a/fs/nfs/nfs4xdr.c
+++ b/fs/nfs/nfs4xdr.c
@@ -1842,8 +1842,8 @@  static void encode_create_session(struct xdr_stream *xdr,
 	 * Assumes OPEN is the biggest non-idempotent compound.
 	 * 2 is the verifier.
 	 */
-	max_resp_sz_cached = (NFS4_dec_open_sz + RPC_REPHDRSIZE +
-			      RPC_MAX_AUTH_SIZE + 2) * XDR_UNIT;
+	max_resp_sz_cached = (NFS4_dec_open_sz + RPC_REPHDRSIZE + 2)
+				* XDR_UNIT + RPC_MAX_AUTH_SIZE;
 
 	encode_op_hdr(xdr, OP_CREATE_SESSION, decode_create_session_maxsz, hdr);
 	p = reserve_space(xdr, 16 + 2*28 + 20 + clnt->cl_nodelen + 12);