NFSv4.1: sp4_mach_cred cleanup (v2)
Subsystems (2)
cover NFSv4.1: sp4_mach_cred cleanup (v2)
Version 2 takes a new approach that Trond requested: we rely on the cl_machine_cred to have a reference held for the life of the nfs_client structure, so the sp4 stuff doesn't need to do any get/put_rpccred.
There's also two new patches:
- "fix SECINFO* use of put_rpccred" - recent changes (by me!) called put_rpccred of rpc_message.rpc_cred, but this value can change when nfs4_state_protect() is called. I searched through the rest of the client source and couldn't find another place where this happens.
- WARN_ON -> WARN_ON_ONCE - oops.
Weston Andros Adamson (4):
NFSv4.1: sp4_mach_cred: ask for WRITE and COMMIT NFSv4.1: fix SECINFO* use of put_rpccred NFSv4.1: sp4_mach_cred: no need to ref count creds NFSv4.1: sp4_mach_cred: WARN_ON -> WARN_ON_ONCE
fs/nfs/nfs4_fs.h | 10 +++++----- fs/nfs/nfs4proc.c | 22 ++++++++++++++-------- 2 files changed, 19 insertions(+), 13 deletions(-)
Signature · 1 line
1.7.12.4 (Apple Git-37)
PATCH 1/4 NFSv4.1: sp4_mach_cred: ask for WRITE and COMMIT
Request SP4_MACH_CRED WRITE and COMMIT support in spo_must_allow list -- they're already supported by the client.
--- fs/nfs/nfs4proc.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)
fs/nfs/nfs4proc.c +4 -2
@@ -6151,11 +6151,13 @@ static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = { }, .allow.u.words = { [0 ] =1 << (OP_CLOSE ) | - 1 << (OP_LOCKU), +1 << (OP_LOCKU ) | + 1 << (OP_COMMIT), [1 ] =1 << (OP_SECINFO -32 ) |1 << (OP_SECINFO_NO_NAME -32 ) |1 << (OP_TEST_STATEID -32 ) | -1 << (OP_FREE_STATEID -32 ) +1 << (OP_FREE_STATEID -32 ) | +1 << (OP_WRITE -32 ) } };
Signature · 1 line
1.7.12.4 (Apple Git-37)
PATCH 2/4 NFSv4.1: fix SECINFO* use of put_rpccred
In reply to Weston Andros AdamsonRecent SP4_MACH_CRED changes allows rpc_message.rpc_cred to change, so keep a separate pointer to the machine cred for put_rpccred.
--- fs/nfs/nfs4proc.c | 16 ++++++++++------ 1 file changed, 10 insertions(+), 6 deletions(-)
fs/nfs/nfs4proc.c +10 -6
@@ -6001,10 +6001,12 @@ static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct .rpc_resp = &res, };struct rpc_clnt *clnt =NFS_SERVER (dir)->client; +struct rpc_cred *cred =NULL ;if (use_integrity) { clnt =NFS_SERVER (dir)->nfs_client->cl_rpcclient; - msg.rpc_cred =nfs4_get_clid_cred (NFS_SERVER (dir)->nfs_client); + cred =nfs4_get_clid_cred (NFS_SERVER (dir)->nfs_client); + msg.rpc_cred = cred; }dprintk ("NFS call secinfo %s\n" , name->name);
@@ -6016,8 +6018,8 @@ static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct &res.seq_res, 0);dprintk ("NFS reply secinfo: %d\n" , status); -if (msg.rpc_cred) -put_rpccred (msg.rpc_cred); +if (cred) +put_rpccred (cred);return status; }
@@ -7498,11 +7500,13 @@ _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle, .rpc_resp = &res, };struct rpc_clnt *clnt = server->client; +struct rpc_cred *cred =NULL ;int status;if (use_integrity) { clnt = server->nfs_client->cl_rpcclient; - msg.rpc_cred =nfs4_get_clid_cred (server->nfs_client); + cred =nfs4_get_clid_cred (server->nfs_client); + msg.rpc_cred = cred; }dprintk ("--> %s\n" , __func__);
@@ -7510,8 +7514,8 @@ _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle, &res.seq_res, 0);dprintk ("<-- %s status=%d\n" , __func__, status); -if (msg.rpc_cred) -put_rpccred (msg.rpc_cred); +if (cred) +put_rpccred (cred);return status; }
Signature · 1 line
1.7.12.4 (Apple Git-37)
PATCH 3/4 NFSv4.1: sp4_mach_cred: no need to ref count creds
In reply to Weston Andros AdamsonThe cl_machine_cred doesn't need to be reference counted here - a reference is held is for the lifetime of the struct nfs_client. Also, no need to put_rpccred the rpc_message.rpc_cred.
--- fs/nfs/nfs4_fs.h | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
fs/nfs/nfs4_fs.h +3 -3
@@ -279,10 +279,10 @@ _nfs4_state_protect(struct nfs_client *clp, unsigned long sp4_mode,if (test_bit (sp4_mode, &clp->cl_sp4_flags)) {spin_lock (&clp->cl_lock);if (clp->cl_machine_cred !=NULL ) - newcred =get_rpccred (clp->cl_machine_cred); +/* don't call get_rpccred on the machine cred - +* a reference will be held for life of clp */ + newcred = clp->cl_machine_cred;spin_unlock (&clp->cl_lock); -if (msg->rpc_cred) -put_rpccred (msg->rpc_cred); msg->rpc_cred = newcred; flavor = clp->cl_rpcclient->cl_auth->au_flavor;
Signature · 1 line
1.7.12.4 (Apple Git-37)
Re: [PATCH 3/4] NFSv4.1: sp4_mach_cred: no need to ref count creds
Do we need to do a get_rpccred() after looking up the machine cred in nfs_alloc_client()? Where does our guaranteed reference come from?
Quoted 34 lines · Weston Andros Adamson
The
cl_machine_creddoesn't need to be reference counted here - a reference is held is for the lifetime of the structnfs_client. Also, no need toput_rpccredthe rpc_message.rpc_cred.Signed-off-by Weston Andros Adamson <[email protected]>--- fs/nfs/nfs4_fs.h | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)fs/nfs/nfs4_fs.h +3
-4@@ -279,10 +279,10 @@ _nfs4_state_protect(struct nfs_client *clp, unsigned long sp4_mode,if (test_bit (sp4_mode, &clp->cl_sp4_flags)) {spin_lock (&clp->cl_lock);if (clp->cl_machine_cred !=NULL ) - newcred =get_rpccred (clp->cl_machine_cred); +/* don't call get_rpccred on the machine cred - +* a reference will be held for life of clp */ + newcred = clp->cl_machine_cred;spin_unlock (&clp->cl_lock); -if (msg->rpc_cred) -put_rpccred (msg->rpc_cred); msg->rpc_cred = newcred; flavor = clp->cl_rpcclient->cl_auth->au_flavor; --1.7.12.4 (Apple Git-37)
--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
A message that is not in the archive (sent off-list or not yet received): CAFX2Jf=kbrS6byZBFOoZ64-rb9tCvUHDPZqkip3cy5K12FzaqQ@mail.gmail.com
Re: [PATCH 3/4] NFSv4.1: sp4_mach_cred: no need to ref count creds
T24gV2VkLCAyMDEzLTA5LTExIGF0IDA4OjU5IC0wNDAwLCBBbm5hIFNjaHVtYWtlciB3cm90ZToN Cj4gRG8gd2UgbmVlZCB0byBkbyBhIGdldF9ycGNjcmVkKCkgYWZ0ZXIgbG9va2luZyB1cCB0aGUg bWFjaGluZSBjcmVkIGluDQo+IG5mc19hbGxvY19jbGllbnQoKT8gIFdoZXJlIGRvZXMgb3VyIGd1 YXJhbnRlZWQgcmVmZXJlbmNlIGNvbWUgZnJvbT8NCg0KU2VlIG15IGNvbW1lbnQgdG8gRHJvcycg Zmlyc3QgcGF0Y2hzZXQgeWVzdGVyZGF5LiBJbiB0aGUgY2FzZSBvZg0KU1A0X01BQ0hfQ1JFRCwg d2UgY2Fubm90IGNoYW5nZSB0aGUgbWFjaGluZSBjcmVkZW50aWFsLCBzaW5jZSB0aGUgc3RhdGUN CnByb3RlY3Rpb24gaXMgY29tcGxldGVseSB0aWVkIHRvIHRoYXQgY3JlZGVudGlhbC4gSW4gZmFj dCBpZiB5b3UgbG9vayBhdA0KdGhlIGN1cnJlbnQgY29kZSwgd2Ugb25seSBhY3R1YWxseSBmcmVl IHRoZSBjbF9tYWNoaW5lX2NyZWQgaW4NCm5mc19mcmVlX2NsaWVudCgpLg0KDQpGb3IgdGhhdCBy ZWFzb24sIGl0IGlzIHNhZmUgdG8gYXNzdW1lIHRoYXQgaXQgaXMgYXZhaWxhYmxlIGZvciB0aGUN CmR1cmF0aW9uIG9mIHRoZSBSUEMgY2FsbC4gT2YgY291cnNlLCBycGNfdGFza19zZXRfcnBjX21l c3NhZ2UoKSB3aWxsDQphbHNvIHRha2UgYSByZWZlcmVuY2UsIGJ1dCB0aGF0IHRvbyBpcyByZWR1 bmRhbnQgaW4gdGhpcyBwYXJ0aWN1bGFyDQpjYXNlLg0KDQpDaGVlcnMNCiAgVHJvbmQNCg0KPiBB bm5hDQo+IA0KPiANCj4gDQo+IE9uIFR1ZSwgU2VwIDEwLCAyMDEzIGF0IDY6NDQgUE0sIFdlc3Rv biBBbmRyb3MgQWRhbXNvbg0KPiA8ZHJvc0BuZXRhcHAuY29tPiB3cm90ZToNCj4gICAgICAgICBU aGUgY2xfbWFjaGluZV9jcmVkIGRvZXNuJ3QgbmVlZCB0byBiZSByZWZlcmVuY2UgY291bnRlZCBo ZXJlDQo+ICAgICAgICAgLQ0KPiAgICAgICAgIGEgcmVmZXJlbmNlIGlzIGhlbGQgaXMgZm9yIHRo ZSBsaWZldGltZSBvZiB0aGUgc3RydWN0DQo+ICAgICAgICAgbmZzX2NsaWVudC4NCj4gICAgICAg ICBBbHNvLCBubyBuZWVkIHRvIHB1dF9ycGNjcmVkIHRoZSBycGNfbWVzc2FnZS5ycGNfY3JlZC4N Cj4gICAgICAgICANCj4gICAgICAgICBTaWduZWQtb2ZmLWJ5OiBXZXN0b24gQW5kcm9zIEFkYW1z b24gPGRyb3NAbmV0YXBwLmNvbT4NCj4gICAgICAgICAtLS0NCj4gICAgICAgICAgZnMvbmZzL25m czRfZnMuaCB8IDYgKysrLS0tDQo+ICAgICAgICAgIDEgZmlsZSBjaGFuZ2VkLCAzIGluc2VydGlv bnMoKyksIDMgZGVsZXRpb25zKC0pDQo+ICAgICAgICAgDQo+ICAgICAgICAgZGlmZiAtLWdpdCBh L2ZzL25mcy9uZnM0X2ZzLmggYi9mcy9uZnMvbmZzNF9mcy5oDQo+ICAgICAgICAgaW5kZXggZjUy MGExMS4uMDdhOGFhOSAxMDA2NDQNCj4gICAgICAgICAtLS0gYS9mcy9uZnMvbmZzNF9mcy5oDQo+ ICAgICAgICAgKysrIGIvZnMvbmZzL25mczRfZnMuaA0KPiAgICAgICAgIEBAIC0yNzksMTAgKzI3 OSwxMCBAQCBfbmZzNF9zdGF0ZV9wcm90ZWN0KHN0cnVjdCBuZnNfY2xpZW50DQo+ICAgICAgICAg KmNscCwgdW5zaWduZWQgbG9uZyBzcDRfbW9kZSwNCj4gICAgICAgICAgICAgICAgIGlmICh0ZXN0 X2JpdChzcDRfbW9kZSwgJmNscC0+Y2xfc3A0X2ZsYWdzKSkgew0KPiAgICAgICAgICAgICAgICAg ICAgICAgICBzcGluX2xvY2soJmNscC0+Y2xfbG9jayk7DQo+ICAgICAgICAgICAgICAgICAgICAg ICAgIGlmIChjbHAtPmNsX21hY2hpbmVfY3JlZCAhPSBOVUxMKQ0KPiAgICAgICAgIC0gICAgICAg ICAgICAgICAgICAgICAgIG5ld2NyZWQgPQ0KPiAgICAgICAgIGdldF9ycGNjcmVkKGNscC0+Y2xf bWFjaGluZV9jcmVkKTsNCj4gICAgICAgICArICAgICAgICAgICAgICAgICAgICAgICAvKiBkb24n dCBjYWxsIGdldF9ycGNjcmVkIG9uIHRoZQ0KPiAgICAgICAgIG1hY2hpbmUgY3JlZCAtDQo+ICAg ICAgICAgKyAgICAgICAgICAgICAgICAgICAgICAgICogYSByZWZlcmVuY2Ugd2lsbCBiZSBoZWxk IGZvciBsaWZlDQo+ICAgICAgICAgb2YgY2xwICovDQo+ICAgICAgICAgKyAgICAgICAgICAgICAg ICAgICAgICAgbmV3Y3JlZCA9IGNscC0+Y2xfbWFjaGluZV9jcmVkOw0KPiAgICAgICAgICAgICAg ICAgICAgICAgICBzcGluX3VubG9jaygmY2xwLT5jbF9sb2NrKTsNCj4gICAgICAgICAtICAgICAg ICAgICAgICAgaWYgKG1zZy0+cnBjX2NyZWQpDQo+ICAgICAgICAgLSAgICAgICAgICAgICAgICAg ICAgICAgcHV0X3JwY2NyZWQobXNnLT5ycGNfY3JlZCk7DQo+ICAgICAgICAgICAgICAgICAgICAg ICAgIG1zZy0+cnBjX2NyZWQgPSBuZXdjcmVkOw0KPiAgICAgICAgIA0KPiAgICAgICAgICAgICAg ICAgICAgICAgICBmbGF2b3IgPQ0KPiAgICAgICAgIGNscC0+Y2xfcnBjY2xpZW50LT5jbF9hdXRo LT5hdV9mbGF2b3I7DQo+ICAgICAgICAgLS0NCj4gICAgICAgICAxLjcuMTIuNCAoQXBwbGUgR2l0 LTM3KQ0KPiAgICAgICAgIA0KPiAgICAgICAgIC0tDQo+ICAgICAgICAgVG8gdW5zdWJzY3JpYmUg ZnJvbSB0aGlzIGxpc3Q6IHNlbmQgdGhlIGxpbmUgInVuc3Vic2NyaWJlDQo+ICAgICAgICAgbGlu dXgtbmZzIiBpbg0KPiAgICAgICAgIHRoZSBib2R5IG9mIGEgbWVzc2FnZSB0byBtYWpvcmRvbW9A dmdlci5rZXJuZWwub3JnDQo+ICAgICAgICAgTW9yZSBtYWpvcmRvbW8gaW5mbyBhdA0KPiAgICAg ICAgICBodHRwOi8vdmdlci5rZXJuZWwub3JnL21ham9yZG9tby1pbmZvLmh0bWwNCj4gDQo+IA0K DQotLSANClRyb25kIE15a2xlYnVzdA0KTGludXggTkZTIGNsaWVudCBtYWludGFpbmVyDQoNCk5l dEFwcA0KVHJvbmQuTXlrbGVidXN0QG5ldGFwcC5jb20NCnd3dy5uZXRhcHAuY29tDQo=
Re: [PATCH 3/4] NFSv4.1: sp4_mach_cred: no need to ref count creds
Okay, thanks! I think I missed that explanation yesterday.
Quoted 77 lines · Trond
Quoted 2 lines · Anna Schumaker
Do we need to do a
get_rpccred()after looking up the machine cred innfs_alloc_client()? Where does our guaranteed reference come from?See my comment to Dros' first patchset yesterday. In the case of
SP4_MACH_CRED, we cannot change the machine credential, since the state protection is completely tied to that credential. In fact if you look at the current code, we only actually free thecl_machine_credinnfs_free_client().For that reason, it is safe to assume that it is available for the duration of the RPC call. Of course,
rpc_task_set_rpc_message()will also take a reference, but that too is redundant in this particular case.Cheers
TrondQuoted 41 lines
Anna
On Tue, Sep 10, 2013 at 6:44 PM, Weston Andros Adamson <[email protected]> wrote:
The cl_machine_cred doesn't need to be reference counted here - a reference is held is for the lifetime of the struct nfs_client. Also, no need to put_rpccred the rpc_message.rpc_cred. Signed-off-by: Weston Andros Adamson <[email protected]> --- fs/nfs/nfs4_fs.h | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index f520a11..07a8aa9 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -279,10 +279,10 @@ _nfs4_state_protect(struct nfs_client *clp, unsigned long sp4_mode, if (test_bit(sp4_mode, &clp->cl_sp4_flags)) { spin_lock(&clp->cl_lock); if (clp->cl_machine_cred != NULL) - newcred = get_rpccred(clp->cl_machine_cred); + /* don't call get_rpccred on the machine cred - + * a reference will be held for life of clp */ + newcred = clp->cl_machine_cred; spin_unlock(&clp->cl_lock); - if (msg->rpc_cred) - put_rpccred(msg->rpc_cred); msg->rpc_cred = newcred; flavor = clp->cl_rpcclient->cl_auth->au_flavor; --Quoted 10 lines
1.7.12.4 (Apple Git-37) -- To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html--
Trond Myklebust
Linux NFS client maintainerNetApp
[email protected]
www.netapp.com
PATCH 4/4 NFSv4.1: sp4_mach_cred: WARN_ON -> WARN_ON_ONCE
In reply to Weston Andros AdamsonNo need to spam the logs
--- fs/nfs/nfs4_fs.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
fs/nfs/nfs4_fs.h +2 -2
@@ -286,8 +286,8 @@ _nfs4_state_protect(struct nfs_client *clp, unsigned long sp4_mode, msg->rpc_cred = newcred; flavor = clp->cl_rpcclient->cl_auth->au_flavor; -WARN_ON (flavor !=RPC_AUTH_GSS_KRB5I && - flavor !=RPC_AUTH_GSS_KRB5P ); +WARN_ON_ONCE (flavor !=RPC_AUTH_GSS_KRB5I && + flavor !=RPC_AUTH_GSS_KRB5P ); *clntp = clp->cl_rpcclient;return true;
Signature · 1 line
1.7.12.4 (Apple Git-37)