NFS client bugfixes for Linux 4.20

Highlights include:
 
 Stable fixes:
 - Don't exit the NFSv4 state manager without clearing NFS4CLNT_MANAGER_RUNNING
 
 Bugfixes:
 - Fix an Oops when destroying the RPCSEC_GSS credential cache
 - Fix an Oops during delegation callbacks
 - Ensure that the NFSv4 state manager exits the loop on SIGKILL
 - Fix a bogus get/put in generic_key_to_expire()
 -----BEGIN PGP SIGNATURE-----
 
 iQIcBAABAgAGBQJb7Lh0AAoJEA4mA3inWBJc8uAQAIkrGChs3AFuEQ3G3H9RlxDX
 WFsPghRGmDwXf2sD+nWjl0r60v0v5fQaUhW/7EPe2kbVTF/rnjieXNeFOw33ZMFk
 MDq03nL1/I25DoNK/qg5GZ2NIltZ9oKKbwaN+0LxXKz69X5qIXYnDzYPHDR/PNTg
 Go7PvG8rU31Wd67E2pquwC6zZ6rCPf2BtQjZdzouLAEUWXAMHyJmszpFUxhLMJoz
 k6dZouphj8fkMse3cfKLnGDqbQ2bE6+Yb0B6Hi0p5nShYgZTaQNZ9KxrEJF7J05i
 cxH6IvLEawEMWXYzGEwr1LUDDrpwveuNTt/OroTgOcSsVpZx1DE0sOZkQ4pt/uTe
 c5NzZYKjEOb2DWxoGR2GEDkRasKVBkWvR5MegvyDgyAcXkAjN/6CgYXiniNYDxl6
 qk7sIqkJfug7fv+VW5YHwORKnvRIEDlFcwy5yZ0ij/Qa0dqUR3aczINGLwS6kcfn
 u7M42UR17FUo2zaI9pZhuijwntbtkXMIETWHGRctK7Mum6u37QSVySNCO2A4knBE
 jEy+oYPFCIUqH+ESpNp73otrVt1CTexScIJNsEi1naLmOhjQRW7YjUPEH1Xjg0Ss
 OGyqIjOf6ToF6ma39/XZI9miJe08k6x8b0aGUdG29Cko9UvjLH86ODEausSRAyFA
 OyZFFuHHAau5FGpNvZfj
 =AstN
 -----END PGP SIGNATURE-----

Merge tag 'nfs-for-4.20-3' of git://git.linux-nfs.org/projects/trondmy/linux-nfs

Pull NFS client bugfixes from Trond Myklebust:
 "Highlights include:

 Stable fixes:

   - Don't exit the NFSv4 state manager without clearing
     NFS4CLNT_MANAGER_RUNNING

  Bugfixes:

   - Fix an Oops when destroying the RPCSEC_GSS credential cache

   - Fix an Oops during delegation callbacks

   - Ensure that the NFSv4 state manager exits the loop on SIGKILL

   - Fix a bogus get/put in generic_key_to_expire()"

* tag 'nfs-for-4.20-3' of git://git.linux-nfs.org/projects/trondmy/linux-nfs:
  NFSv4: Fix an Oops during delegation callbacks
  SUNRPC: Fix a bogus get/put in generic_key_to_expire()
  SUNRPC: Fix a Oops when destroying the RPCSEC_GSS credential cache
  NFSv4: Ensure that the state manager exits the loop on SIGKILL
  NFSv4: Don't exit the state manager without clearing NFS4CLNT_MANAGER_RUNNING

fs/nfs/callback_proc.c

@@ -66,7 +66,7 @@ __be32 nfs4_callback_getattr(void *argp, void *resp,
 out_iput:
 	rcu_read_unlock();
 	trace_nfs4_cb_getattr(cps->clp, &args->fh, inode, -ntohl(res->status));
-	iput(inode);
+	nfs_iput_and_deactive(inode);
 out:
 	dprintk("%s: exit with status = %d\n", __func__, ntohl(res->status));
 	return res->status;
@@ -108,7 +108,7 @@ __be32 nfs4_callback_recall(void *argp, void *resp,
 	}
 	trace_nfs4_cb_recall(cps->clp, &args->fh, inode,
 			&args->stateid, -ntohl(res));
-	iput(inode);
+	nfs_iput_and_deactive(inode);
 out:
 	dprintk("%s: exit with status = %d\n", __func__, ntohl(res));
 	return res;

fs/nfs/delegation.c

@@ -850,16 +850,23 @@ nfs_delegation_find_inode_server(struct nfs_server *server,
 				 const struct nfs_fh *fhandle)
 {
 	struct nfs_delegation *delegation;
-	struct inode *res = NULL;
+	struct inode *freeme, *res = NULL;
 
 	list_for_each_entry_rcu(delegation, &server->delegations, super_list) {
 		spin_lock(&delegation->lock);
 		if (delegation->inode != NULL &&
 		    nfs_compare_fh(fhandle, &NFS_I(delegation->inode)->fh) == 0) {
-			res = igrab(delegation->inode);
+			freeme = igrab(delegation->inode);
+			if (freeme && nfs_sb_active(freeme->i_sb))
+				res = freeme;
 			spin_unlock(&delegation->lock);
 			if (res != NULL)
 				return res;
+			if (freeme) {
+				rcu_read_unlock();
+				iput(freeme);
+				rcu_read_lock();
+			}
 			return ERR_PTR(-EAGAIN);
 		}
 		spin_unlock(&delegation->lock);

fs/nfs/nfs4state.c

@@ -2601,11 +2601,12 @@ static void nfs4_state_manager(struct nfs_client *clp)
 		nfs4_clear_state_manager_bit(clp);
 		/* Did we race with an attempt to give us more work? */
 		if (clp->cl_state == 0)
-			break;
+			return;
 		if (test_and_set_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) != 0)
-			break;
-	} while (refcount_read(&clp->cl_count) > 1);
-	return;
+			return;
+	} while (refcount_read(&clp->cl_count) > 1 && !signalled());
+	goto out_drain;
+
 out_error:
 	if (strlen(section))
 		section_sep = ": ";
@@ -2613,6 +2614,7 @@ out_error:
 			" with error %d\n", section_sep, section,
 			clp->cl_hostname, -status);
 	ssleep(1);
+out_drain:
 	nfs4_end_drain_session(clp);
 	nfs4_clear_state_manager_bit(clp);
 }

net/sunrpc/auth_generic.c

@@ -281,13 +281,7 @@ static bool generic_key_to_expire(struct rpc_cred *cred)
 {
 	struct auth_cred *acred = &container_of(cred, struct generic_cred,
 						gc_base)->acred;
-	bool ret;
-
-	get_rpccred(cred);
-	ret = test_bit(RPC_CRED_KEY_EXPIRE_SOON, &acred->ac_flags);
-	put_rpccred(cred);
-
-	return ret;
+	return test_bit(RPC_CRED_KEY_EXPIRE_SOON, &acred->ac_flags);
 }
 
 static const struct rpc_credops generic_credops = {

net/sunrpc/auth_gss/auth_gss.c

@@ -1239,36 +1239,59 @@ gss_create(const struct rpc_auth_create_args *args, struct rpc_clnt *clnt)
 	return &gss_auth->rpc_auth;
 }
 
+static struct gss_cred *
+gss_dup_cred(struct gss_auth *gss_auth, struct gss_cred *gss_cred)
+{
+	struct gss_cred *new;
+
+	/* Make a copy of the cred so that we can reference count it */
+	new = kzalloc(sizeof(*gss_cred), GFP_NOIO);
+	if (new) {
+		struct auth_cred acred = {
+			.uid = gss_cred->gc_base.cr_uid,
+		};
+		struct gss_cl_ctx *ctx =
+			rcu_dereference_protected(gss_cred->gc_ctx, 1);
+
+		rpcauth_init_cred(&new->gc_base, &acred,
+				&gss_auth->rpc_auth,
+				&gss_nullops);
+		new->gc_base.cr_flags = 1UL << RPCAUTH_CRED_UPTODATE;
+		new->gc_service = gss_cred->gc_service;
+		new->gc_principal = gss_cred->gc_principal;
+		kref_get(&gss_auth->kref);
+		rcu_assign_pointer(new->gc_ctx, ctx);
+		gss_get_ctx(ctx);
+	}
+	return new;
+}
+
 /*
- * gss_destroying_context will cause the RPCSEC_GSS to send a NULL RPC call
+ * gss_send_destroy_context will cause the RPCSEC_GSS to send a NULL RPC call
  * to the server with the GSS control procedure field set to
  * RPC_GSS_PROC_DESTROY. This should normally cause the server to release
  * all RPCSEC_GSS state associated with that context.
  */
-static int
-gss_destroying_context(struct rpc_cred *cred)
+static void
+gss_send_destroy_context(struct rpc_cred *cred)
 {
 	struct gss_cred *gss_cred = container_of(cred, struct gss_cred, gc_base);
 	struct gss_auth *gss_auth = container_of(cred->cr_auth, struct gss_auth, rpc_auth);
 	struct gss_cl_ctx *ctx = rcu_dereference_protected(gss_cred->gc_ctx, 1);
+	struct gss_cred *new;
 	struct rpc_task *task;
 
-	if (test_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags) == 0)
-		return 0;
+	new = gss_dup_cred(gss_auth, gss_cred);
+	if (new) {
+		ctx->gc_proc = RPC_GSS_PROC_DESTROY;
 
-	ctx->gc_proc = RPC_GSS_PROC_DESTROY;
-	cred->cr_ops = &gss_nullops;
+		task = rpc_call_null(gss_auth->client, &new->gc_base,
+				RPC_TASK_ASYNC|RPC_TASK_SOFT);
+		if (!IS_ERR(task))
+			rpc_put_task(task);
 
-	/* Take a reference to ensure the cred will be destroyed either
-	 * by the RPC call or by the put_rpccred() below */
-	get_rpccred(cred);
-
-	task = rpc_call_null(gss_auth->client, cred, RPC_TASK_ASYNC|RPC_TASK_SOFT);
-	if (!IS_ERR(task))
-		rpc_put_task(task);
-
-	put_rpccred(cred);
-	return 1;
+		put_rpccred(&new->gc_base);
+	}
 }
 
 /* gss_destroy_cred (and gss_free_ctx) are used to clean up after failure
@@ -1330,8 +1353,8 @@ static void
 gss_destroy_cred(struct rpc_cred *cred)
 {
 
-	if (gss_destroying_context(cred))
-		return;
+	if (test_and_clear_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags) != 0)
+		gss_send_destroy_context(cred);
 	gss_destroy_nullcred(cred);
 }