2008-08-26 22:05:38

by Harvey Harrison

[permalink] [raw]
Subject: [PATCH] afs: fsclient.c sparse endian annotations of operation_ID

fs/afs/fsclient.c:319:11: warning: restricted __be32 degrades to integer
fs/afs/fsclient.c:465:21: warning: incorrect type in assignment (different base types)
fs/afs/fsclient.c:465:21: expected restricted __be32 [usertype] operation_ID
fs/afs/fsclient.c:465:21: got int
fs/afs/fsclient.c:510:21: warning: incorrect type in assignment (different base types)
fs/afs/fsclient.c:510:21: expected restricted __be32 [usertype] operation_ID
fs/afs/fsclient.c:510:21: got int
fs/afs/fsclient.c:1304:10: warning: restricted __be32 degrades to integer
fs/afs/fsclient.c:1366:21: warning: incorrect type in assignment (different base types)
fs/afs/fsclient.c:1366:21: expected restricted __be32 [usertype] operation_ID
fs/afs/fsclient.c:1366:21: got int
fs/afs/fsclient.c:1417:21: warning: incorrect type in assignment (different base types)
fs/afs/fsclient.c:1417:21: expected restricted __be32 [usertype] operation_ID
fs/afs/fsclient.c:1417:21: got int
fs/afs/fsclient.c:1463:21: warning: incorrect type in assignment (different base types)
fs/afs/fsclient.c:1463:21: expected restricted __be32 [usertype] operation_ID
fs/afs/fsclient.c:1463:21: got int

Signed-off-by: Harvey Harrison <[email protected]>
---
fs/afs/fsclient.c | 14 +++++++-------
1 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/fs/afs/fsclient.c b/fs/afs/fsclient.c
index 023b95b..2928067 100644
--- a/fs/afs/fsclient.c
+++ b/fs/afs/fsclient.c
@@ -316,7 +316,7 @@ static int afs_deliver_fs_fetch_data(struct afs_call *call,
case 0:
call->offset = 0;
call->unmarshall++;
- if (call->operation_ID != FSFETCHDATA64) {
+ if (call->operation_ID != __constant_htonl(FSFETCHDATA64)) {
call->unmarshall++;
goto no_msw;
}
@@ -462,7 +462,7 @@ static int afs_fs_fetch_data64(struct afs_server *server,
call->reply3 = buffer;
call->service_id = FS_SERVICE;
call->port = htons(AFS_FS_PORT);
- call->operation_ID = FSFETCHDATA64;
+ call->operation_ID = __constant_htonl(FSFETCHDATA64);

/* marshall the parameters */
bp = call->request;
@@ -507,7 +507,7 @@ int afs_fs_fetch_data(struct afs_server *server,
call->reply3 = buffer;
call->service_id = FS_SERVICE;
call->port = htons(AFS_FS_PORT);
- call->operation_ID = FSFETCHDATA;
+ call->operation_ID = __constant_htonl(FSFETCHDATA);

/* marshall the parameters */
bp = call->request;
@@ -1301,7 +1301,7 @@ static int afs_deliver_fs_store_status(struct afs_call *call,

/* unmarshall the reply once we've received all of it */
store_version = NULL;
- if (call->operation_ID == FSSTOREDATA)
+ if (call->operation_ID == __constant_htonl(FSSTOREDATA))
store_version = &call->store_version;

bp = call->buffer;
@@ -1363,7 +1363,7 @@ static int afs_fs_setattr_size64(struct afs_server *server, struct key *key,
call->service_id = FS_SERVICE;
call->port = htons(AFS_FS_PORT);
call->store_version = vnode->status.data_version + 1;
- call->operation_ID = FSSTOREDATA;
+ call->operation_ID = __constant_htonl(FSSTOREDATA);

/* marshall the parameters */
bp = call->request;
@@ -1414,7 +1414,7 @@ static int afs_fs_setattr_size(struct afs_server *server, struct key *key,
call->service_id = FS_SERVICE;
call->port = htons(AFS_FS_PORT);
call->store_version = vnode->status.data_version + 1;
- call->operation_ID = FSSTOREDATA;
+ call->operation_ID = __constant_htonl(FSSTOREDATA);

/* marshall the parameters */
bp = call->request;
@@ -1460,7 +1460,7 @@ int afs_fs_setattr(struct afs_server *server, struct key *key,
call->reply = vnode;
call->service_id = FS_SERVICE;
call->port = htons(AFS_FS_PORT);
- call->operation_ID = FSSTORESTATUS;
+ call->operation_ID = __constant_htonl(FSSTORESTATUS);

/* marshall the parameters */
bp = call->request;
--
1.6.0.340.g84854



2008-08-27 11:12:56

by David Howells

[permalink] [raw]
Subject: Re: [PATCH] afs: fsclient.c sparse endian annotations of operation_ID

Harvey Harrison <[email protected]> wrote:

> + if (call->operation_ID != __constant_htonl(FSFETCHDATA64)) {

Doesn't htonl() resolve to this for a constant argument? Following through
the definitions, it certainly looks like it ought to:

<linux/byteorder/generic.h>
#undef htonl
#define ___htonl(x) __cpu_to_be32(x)

<linux/byteorder/little_endian.h>
#define __cpu_to_be32(x) ((__force __be32)__swab32((x)))

<linux/byteorder/swab.h>
# define __swab32(x) \
(__builtin_constant_p((__u32)(x)) ? \
___constant_swab32((x)) : \
__fswab32((x)))

at least for GCC with optimisation enabled.

Of course, linux/byteorder.h and linux/swab.h seem to do much the same. Any
idea why we have both sets? There seems to be unnecessary redundancy.

David

2008-08-27 11:17:19

by David Miller

[permalink] [raw]
Subject: Re: [PATCH] afs: fsclient.c sparse endian annotations of operation_ID

From: David Howells <[email protected]>
Date: Wed, 27 Aug 2008 12:12:24 +0100

> Harvey Harrison <[email protected]> wrote:
>
> > + if (call->operation_ID != __constant_htonl(FSFETCHDATA64)) {
>
> Doesn't htonl() resolve to this for a constant argument? Following through
> the definitions, it certainly looks like it ought to:

Yes it does.

The __constant_*() interfaces should only be uses for things that
must be evaluated at compile time (static data initializations,
switch statement case values etc.).

2008-08-28 02:22:48

by Al Viro

[permalink] [raw]
Subject: Re: [PATCH] afs: fsclient.c sparse endian annotations of operation_ID

On Wed, Aug 27, 2008 at 04:17:05AM -0700, David Miller wrote:
> From: David Howells <[email protected]>
> Date: Wed, 27 Aug 2008 12:12:24 +0100
>
> > Harvey Harrison <[email protected]> wrote:
> >
> > > + if (call->operation_ID != __constant_htonl(FSFETCHDATA64)) {
> >
> > Doesn't htonl() resolve to this for a constant argument? Following through
> > the definitions, it certainly looks like it ought to:
>
> Yes it does.
>
> The __constant_*() interfaces should only be uses for things that
> must be evaluated at compile time (static data initializations,
> switch statement case values etc.).

It's not just "must be evaluated at compile time"; it's "must be C constant
expressions", which is much stricter. And no, in the case quoted above
__constant_ form is not needed. Just use htonl()...