Merge /home/trondmy/scm/kernel/git/torvalds/linux-2.6
diff --git a/fs/exec.c b/fs/exec.c
index a04a575..d2208f7 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -126,8 +126,7 @@
struct nameidata nd;
int error;
- nd.intent.open.flags = FMODE_READ;
- error = __user_walk(library, LOOKUP_FOLLOW|LOOKUP_OPEN, &nd);
+ error = __user_path_lookup_open(library, LOOKUP_FOLLOW, &nd, FMODE_READ);
if (error)
goto out;
@@ -139,7 +138,7 @@
if (error)
goto exit;
- file = dentry_open(nd.dentry, nd.mnt, O_RDONLY);
+ file = nameidata_to_filp(&nd, O_RDONLY);
error = PTR_ERR(file);
if (IS_ERR(file))
goto out;
@@ -167,6 +166,7 @@
out:
return error;
exit:
+ release_open_intent(&nd);
path_release(&nd);
goto out;
}
@@ -490,8 +490,7 @@
int err;
struct file *file;
- nd.intent.open.flags = FMODE_READ;
- err = path_lookup(name, LOOKUP_FOLLOW|LOOKUP_OPEN, &nd);
+ err = path_lookup_open(name, LOOKUP_FOLLOW, &nd, FMODE_READ);
file = ERR_PTR(err);
if (!err) {
@@ -504,7 +503,7 @@
err = -EACCES;
file = ERR_PTR(err);
if (!err) {
- file = dentry_open(nd.dentry, nd.mnt, O_RDONLY);
+ file = nameidata_to_filp(&nd, O_RDONLY);
if (!IS_ERR(file)) {
err = deny_write_access(file);
if (err) {
@@ -516,6 +515,7 @@
return file;
}
}
+ release_open_intent(&nd);
path_release(&nd);
}
goto out;
diff --git a/fs/lockd/host.c b/fs/lockd/host.c
index 82c77df..c4c8601 100644
--- a/fs/lockd/host.c
+++ b/fs/lockd/host.c
@@ -173,11 +173,10 @@
/* If we've already created an RPC client, check whether
* RPC rebind is required
- * Note: why keep rebinding if we're on a tcp connection?
*/
if ((clnt = host->h_rpcclnt) != NULL) {
xprt = clnt->cl_xprt;
- if (!xprt->stream && time_after_eq(jiffies, host->h_nextrebind)) {
+ if (time_after_eq(jiffies, host->h_nextrebind)) {
clnt->cl_port = 0;
host->h_nextrebind = jiffies + NLM_HOST_REBIND;
dprintk("lockd: next rebind in %ld jiffies\n",
@@ -189,7 +188,6 @@
goto forgetit;
xprt_set_timeout(&xprt->timeout, 5, nlmsvc_timeout);
- xprt->nocong = 1; /* No congestion control for NLM */
xprt->resvport = 1; /* NLM requires a reserved port */
/* Existing NLM servers accept AUTH_UNIX only */
diff --git a/fs/locks.c b/fs/locks.c
index f7daa5f..a1e8b22 100644
--- a/fs/locks.c
+++ b/fs/locks.c
@@ -316,21 +316,22 @@
/* POSIX-1996 leaves the case l->l_len < 0 undefined;
POSIX-2001 defines it. */
start += l->l_start;
- end = start + l->l_len - 1;
- if (l->l_len < 0) {
- end = start - 1;
- start += l->l_len;
- }
-
if (start < 0)
return -EINVAL;
- if (l->l_len > 0 && end < 0)
- return -EOVERFLOW;
-
+ fl->fl_end = OFFSET_MAX;
+ if (l->l_len > 0) {
+ end = start + l->l_len - 1;
+ fl->fl_end = end;
+ } else if (l->l_len < 0) {
+ end = start - 1;
+ fl->fl_end = end;
+ start += l->l_len;
+ if (start < 0)
+ return -EINVAL;
+ }
fl->fl_start = start; /* we record the absolute position */
- fl->fl_end = end;
- if (l->l_len == 0)
- fl->fl_end = OFFSET_MAX;
+ if (fl->fl_end < fl->fl_start)
+ return -EOVERFLOW;
fl->fl_owner = current->files;
fl->fl_pid = current->tgid;
@@ -362,14 +363,21 @@
return -EINVAL;
}
- if (((start += l->l_start) < 0) || (l->l_len < 0))
+ start += l->l_start;
+ if (start < 0)
return -EINVAL;
- fl->fl_end = start + l->l_len - 1;
- if (l->l_len > 0 && fl->fl_end < 0)
- return -EOVERFLOW;
+ fl->fl_end = OFFSET_MAX;
+ if (l->l_len > 0) {
+ fl->fl_end = start + l->l_len - 1;
+ } else if (l->l_len < 0) {
+ fl->fl_end = start - 1;
+ start += l->l_len;
+ if (start < 0)
+ return -EINVAL;
+ }
fl->fl_start = start; /* we record the absolute position */
- if (l->l_len == 0)
- fl->fl_end = OFFSET_MAX;
+ if (fl->fl_end < fl->fl_start)
+ return -EOVERFLOW;
fl->fl_owner = current->files;
fl->fl_pid = current->tgid;
@@ -829,12 +837,16 @@
/* Detect adjacent or overlapping regions (if same lock type)
*/
if (request->fl_type == fl->fl_type) {
+ /* In all comparisons of start vs end, use
+ * "start - 1" rather than "end + 1". If end
+ * is OFFSET_MAX, end + 1 will become negative.
+ */
if (fl->fl_end < request->fl_start - 1)
goto next_lock;
/* If the next lock in the list has entirely bigger
* addresses than the new one, insert the lock here.
*/
- if (fl->fl_start > request->fl_end + 1)
+ if (fl->fl_start - 1 > request->fl_end)
break;
/* If we come here, the new and old lock are of the
diff --git a/fs/namei.c b/fs/namei.c
index aa62dbd..aaaa810 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -28,6 +28,7 @@
#include <linux/syscalls.h>
#include <linux/mount.h>
#include <linux/audit.h>
+#include <linux/file.h>
#include <asm/namei.h>
#include <asm/uaccess.h>
@@ -317,6 +318,18 @@
mntput_no_expire(nd->mnt);
}
+/**
+ * release_open_intent - free up open intent resources
+ * @nd: pointer to nameidata
+ */
+void release_open_intent(struct nameidata *nd)
+{
+ if (nd->intent.open.file->f_dentry == NULL)
+ put_filp(nd->intent.open.file);
+ else
+ fput(nd->intent.open.file);
+}
+
/*
* Internal lookup() using the new generic dcache.
* SMP-safe
@@ -750,6 +763,7 @@
struct qstr this;
unsigned int c;
+ nd->flags |= LOOKUP_CONTINUE;
err = exec_permission_lite(inode, nd);
if (err == -EAGAIN) {
err = permission(inode, MAY_EXEC, nd);
@@ -802,7 +816,6 @@
if (err < 0)
break;
}
- nd->flags |= LOOKUP_CONTINUE;
/* This does the actual lookups.. */
err = do_lookup(nd, &this, &next);
if (err)
@@ -1052,6 +1065,70 @@
return retval;
}
+static int __path_lookup_intent_open(const char *name, unsigned int lookup_flags,
+ struct nameidata *nd, int open_flags, int create_mode)
+{
+ struct file *filp = get_empty_filp();
+ int err;
+
+ if (filp == NULL)
+ return -ENFILE;
+ nd->intent.open.file = filp;
+ nd->intent.open.flags = open_flags;
+ nd->intent.open.create_mode = create_mode;
+ err = path_lookup(name, lookup_flags|LOOKUP_OPEN, nd);
+ if (IS_ERR(nd->intent.open.file)) {
+ if (err == 0) {
+ err = PTR_ERR(nd->intent.open.file);
+ path_release(nd);
+ }
+ } else if (err != 0)
+ release_open_intent(nd);
+ return err;
+}
+
+/**
+ * path_lookup_open - lookup a file path with open intent
+ * @name: pointer to file name
+ * @lookup_flags: lookup intent flags
+ * @nd: pointer to nameidata
+ * @open_flags: open intent flags
+ */
+int path_lookup_open(const char *name, unsigned int lookup_flags,
+ struct nameidata *nd, int open_flags)
+{
+ return __path_lookup_intent_open(name, lookup_flags, nd,
+ open_flags, 0);
+}
+
+/**
+ * path_lookup_create - lookup a file path with open + create intent
+ * @name: pointer to file name
+ * @lookup_flags: lookup intent flags
+ * @nd: pointer to nameidata
+ * @open_flags: open intent flags
+ * @create_mode: create intent flags
+ */
+int path_lookup_create(const char *name, unsigned int lookup_flags,
+ struct nameidata *nd, int open_flags, int create_mode)
+{
+ return __path_lookup_intent_open(name, lookup_flags|LOOKUP_CREATE, nd,
+ open_flags, create_mode);
+}
+
+int __user_path_lookup_open(const char __user *name, unsigned int lookup_flags,
+ struct nameidata *nd, int open_flags)
+{
+ char *tmp = getname(name);
+ int err = PTR_ERR(tmp);
+
+ if (!IS_ERR(tmp)) {
+ err = __path_lookup_intent_open(tmp, lookup_flags, nd, open_flags, 0);
+ putname(tmp);
+ }
+ return err;
+}
+
/*
* Restricted form of lookup. Doesn't follow links, single-component only,
* needs parent already locked. Doesn't follow mounts.
@@ -1416,27 +1493,27 @@
*/
int open_namei(const char * pathname, int flag, int mode, struct nameidata *nd)
{
- int acc_mode, error = 0;
+ int acc_mode, error;
struct path path;
struct dentry *dir;
int count = 0;
acc_mode = ACC_MODE(flag);
+ /* O_TRUNC implies we need access checks for write permissions */
+ if (flag & O_TRUNC)
+ acc_mode |= MAY_WRITE;
+
/* Allow the LSM permission hook to distinguish append
access from general write access. */
if (flag & O_APPEND)
acc_mode |= MAY_APPEND;
- /* Fill in the open() intent data */
- nd->intent.open.flags = flag;
- nd->intent.open.create_mode = mode;
-
/*
* The simplest case - just a plain lookup.
*/
if (!(flag & O_CREAT)) {
- error = path_lookup(pathname, lookup_flags(flag)|LOOKUP_OPEN, nd);
+ error = path_lookup_open(pathname, lookup_flags(flag), nd, flag);
if (error)
return error;
goto ok;
@@ -1445,7 +1522,7 @@
/*
* Create - we need to know the parent.
*/
- error = path_lookup(pathname, LOOKUP_PARENT|LOOKUP_OPEN|LOOKUP_CREATE, nd);
+ error = path_lookup_create(pathname, LOOKUP_PARENT, nd, flag, mode);
if (error)
return error;
@@ -1520,6 +1597,8 @@
exit_dput:
dput_path(&path, nd);
exit:
+ if (!IS_ERR(nd->intent.open.file))
+ release_open_intent(nd);
path_release(nd);
return error;
diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c
index 4a36839..44135af 100644
--- a/fs/nfs/delegation.c
+++ b/fs/nfs/delegation.c
@@ -142,7 +142,7 @@
/*
* Basic procedure for returning a delegation to the server
*/
-int nfs_inode_return_delegation(struct inode *inode)
+int __nfs_inode_return_delegation(struct inode *inode)
{
struct nfs4_client *clp = NFS_SERVER(inode)->nfs4_state;
struct nfs_inode *nfsi = NFS_I(inode);
diff --git a/fs/nfs/delegation.h b/fs/nfs/delegation.h
index 3f6c45a..8017846 100644
--- a/fs/nfs/delegation.h
+++ b/fs/nfs/delegation.h
@@ -25,7 +25,7 @@
int nfs_inode_set_delegation(struct inode *inode, struct rpc_cred *cred, struct nfs_openres *res);
void nfs_inode_reclaim_delegation(struct inode *inode, struct rpc_cred *cred, struct nfs_openres *res);
-int nfs_inode_return_delegation(struct inode *inode);
+int __nfs_inode_return_delegation(struct inode *inode);
int nfs_async_inode_return_delegation(struct inode *inode, const nfs4_stateid *stateid);
struct inode *nfs_delegation_find_inode(struct nfs4_client *clp, const struct nfs_fh *fhandle);
@@ -47,11 +47,25 @@
return 1;
return 0;
}
+
+static inline int nfs_inode_return_delegation(struct inode *inode)
+{
+ int err = 0;
+
+ if (NFS_I(inode)->delegation != NULL)
+ err = __nfs_inode_return_delegation(inode);
+ return err;
+}
#else
static inline int nfs_have_delegation(struct inode *inode, int flags)
{
return 0;
}
+
+static inline int nfs_inode_return_delegation(struct inode *inode)
+{
+ return 0;
+}
#endif
#endif
diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
index 2df639f..eb50c19 100644
--- a/fs/nfs/dir.c
+++ b/fs/nfs/dir.c
@@ -565,8 +565,6 @@
}
}
unlock_kernel();
- if (desc->error < 0)
- return desc->error;
if (res < 0)
return res;
return 0;
@@ -803,6 +801,7 @@
*/
static void nfs_dentry_iput(struct dentry *dentry, struct inode *inode)
{
+ nfs_inode_return_delegation(inode);
if (dentry->d_flags & DCACHE_NFSFS_RENAMED) {
lock_kernel();
inode->i_nlink--;
@@ -916,7 +915,6 @@
static struct dentry *nfs_atomic_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
{
struct dentry *res = NULL;
- struct inode *inode = NULL;
int error;
/* Check that we are indeed trying to open this file */
@@ -930,8 +928,10 @@
dentry->d_op = NFS_PROTO(dir)->dentry_ops;
/* Let vfs_create() deal with O_EXCL */
- if (nd->intent.open.flags & O_EXCL)
- goto no_entry;
+ if (nd->intent.open.flags & O_EXCL) {
+ d_add(dentry, NULL);
+ goto out;
+ }
/* Open the file on the server */
lock_kernel();
@@ -945,32 +945,30 @@
if (nd->intent.open.flags & O_CREAT) {
nfs_begin_data_update(dir);
- inode = nfs4_atomic_open(dir, dentry, nd);
+ res = nfs4_atomic_open(dir, dentry, nd);
nfs_end_data_update(dir);
} else
- inode = nfs4_atomic_open(dir, dentry, nd);
+ res = nfs4_atomic_open(dir, dentry, nd);
unlock_kernel();
- if (IS_ERR(inode)) {
- error = PTR_ERR(inode);
+ if (IS_ERR(res)) {
+ error = PTR_ERR(res);
switch (error) {
/* Make a negative dentry */
case -ENOENT:
- inode = NULL;
- break;
+ res = NULL;
+ goto out;
/* This turned out not to be a regular file */
+ case -EISDIR:
+ case -ENOTDIR:
+ goto no_open;
case -ELOOP:
if (!(nd->intent.open.flags & O_NOFOLLOW))
goto no_open;
- /* case -EISDIR: */
/* case -EINVAL: */
default:
- res = ERR_PTR(error);
goto out;
}
- }
-no_entry:
- res = d_add_unique(dentry, inode);
- if (res != NULL)
+ } else if (res != NULL)
dentry = res;
nfs_renew_times(dentry);
nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
@@ -1014,7 +1012,7 @@
*/
lock_kernel();
verifier = nfs_save_change_attribute(dir);
- ret = nfs4_open_revalidate(dir, dentry, openflags);
+ ret = nfs4_open_revalidate(dir, dentry, openflags, nd);
if (!ret)
nfs_set_verifier(dentry, verifier);
unlock_kernel();
@@ -1137,7 +1135,7 @@
lock_kernel();
nfs_begin_data_update(dir);
- error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags);
+ error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags, nd);
nfs_end_data_update(dir);
if (error != 0)
goto out_err;
@@ -1332,6 +1330,7 @@
nfs_begin_data_update(dir);
if (inode != NULL) {
+ nfs_inode_return_delegation(inode);
nfs_begin_data_update(inode);
error = NFS_PROTO(dir)->remove(dir, &dentry->d_name);
/* The VFS may want to delete this inode */
@@ -1512,9 +1511,11 @@
*/
if (!new_inode)
goto go_ahead;
- if (S_ISDIR(new_inode->i_mode))
- goto out;
- else if (atomic_read(&new_dentry->d_count) > 2) {
+ if (S_ISDIR(new_inode->i_mode)) {
+ error = -EISDIR;
+ if (!S_ISDIR(old_inode->i_mode))
+ goto out;
+ } else if (atomic_read(&new_dentry->d_count) > 2) {
int err;
/* copy the target dentry's name */
dentry = d_alloc(new_dentry->d_parent,
@@ -1539,7 +1540,8 @@
#endif
goto out;
}
- }
+ } else
+ new_inode->i_nlink--;
go_ahead:
/*
@@ -1549,6 +1551,7 @@
nfs_wb_all(old_inode);
shrink_dcache_parent(old_dentry);
}
+ nfs_inode_return_delegation(old_inode);
if (new_inode)
d_delete(new_dentry);
diff --git a/fs/nfs/file.c b/fs/nfs/file.c
index 6bdcfa9..572d859 100644
--- a/fs/nfs/file.c
+++ b/fs/nfs/file.c
@@ -376,22 +376,31 @@
static int do_getlk(struct file *filp, int cmd, struct file_lock *fl)
{
+ struct file_lock *cfl;
struct inode *inode = filp->f_mapping->host;
int status = 0;
lock_kernel();
- /* Use local locking if mounted with "-onolock" */
- if (!(NFS_SERVER(inode)->flags & NFS_MOUNT_NONLM))
- status = NFS_PROTO(inode)->lock(filp, cmd, fl);
- else {
- struct file_lock *cfl = posix_test_lock(filp, fl);
-
- fl->fl_type = F_UNLCK;
- if (cfl != NULL)
- memcpy(fl, cfl, sizeof(*fl));
+ /* Try local locking first */
+ cfl = posix_test_lock(filp, fl);
+ if (cfl != NULL) {
+ locks_copy_lock(fl, cfl);
+ goto out;
}
+
+ if (nfs_have_delegation(inode, FMODE_READ))
+ goto out_noconflict;
+
+ if (NFS_SERVER(inode)->flags & NFS_MOUNT_NONLM)
+ goto out_noconflict;
+
+ status = NFS_PROTO(inode)->lock(filp, cmd, fl);
+out:
unlock_kernel();
return status;
+out_noconflict:
+ fl->fl_type = F_UNLCK;
+ goto out;
}
static int do_vfs_lock(struct file *file, struct file_lock *fl)
diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c
index d4eadee..65d5ab4 100644
--- a/fs/nfs/inode.c
+++ b/fs/nfs/inode.c
@@ -358,6 +358,35 @@
return no_root_error;
}
+static void nfs_init_timeout_values(struct rpc_timeout *to, int proto, unsigned int timeo, unsigned int retrans)
+{
+ to->to_initval = timeo * HZ / 10;
+ to->to_retries = retrans;
+ if (!to->to_retries)
+ to->to_retries = 2;
+
+ switch (proto) {
+ case IPPROTO_TCP:
+ if (!to->to_initval)
+ to->to_initval = 60 * HZ;
+ if (to->to_initval > NFS_MAX_TCP_TIMEOUT)
+ to->to_initval = NFS_MAX_TCP_TIMEOUT;
+ to->to_increment = to->to_initval;
+ to->to_maxval = to->to_initval + (to->to_increment * to->to_retries);
+ to->to_exponential = 0;
+ break;
+ case IPPROTO_UDP:
+ default:
+ if (!to->to_initval)
+ to->to_initval = 11 * HZ / 10;
+ if (to->to_initval > NFS_MAX_UDP_TIMEOUT)
+ to->to_initval = NFS_MAX_UDP_TIMEOUT;
+ to->to_maxval = NFS_MAX_UDP_TIMEOUT;
+ to->to_exponential = 1;
+ break;
+ }
+}
+
/*
* Create an RPC client handle.
*/
@@ -367,22 +396,12 @@
struct rpc_timeout timeparms;
struct rpc_xprt *xprt = NULL;
struct rpc_clnt *clnt = NULL;
- int tcp = (data->flags & NFS_MOUNT_TCP);
+ int proto = (data->flags & NFS_MOUNT_TCP) ? IPPROTO_TCP : IPPROTO_UDP;
- /* Initialize timeout values */
- timeparms.to_initval = data->timeo * HZ / 10;
- timeparms.to_retries = data->retrans;
- timeparms.to_maxval = tcp ? RPC_MAX_TCP_TIMEOUT : RPC_MAX_UDP_TIMEOUT;
- timeparms.to_exponential = 1;
-
- if (!timeparms.to_initval)
- timeparms.to_initval = (tcp ? 600 : 11) * HZ / 10;
- if (!timeparms.to_retries)
- timeparms.to_retries = 5;
+ nfs_init_timeout_values(&timeparms, proto, data->timeo, data->retrans);
/* create transport and client */
- xprt = xprt_create_proto(tcp ? IPPROTO_TCP : IPPROTO_UDP,
- &server->addr, &timeparms);
+ xprt = xprt_create_proto(proto, &server->addr, &timeparms);
if (IS_ERR(xprt)) {
dprintk("%s: cannot create RPC transport. Error = %ld\n",
__FUNCTION__, PTR_ERR(xprt));
@@ -576,7 +595,6 @@
{ NFS_MOUNT_SOFT, ",soft", ",hard" },
{ NFS_MOUNT_INTR, ",intr", "" },
{ NFS_MOUNT_POSIX, ",posix", "" },
- { NFS_MOUNT_TCP, ",tcp", ",udp" },
{ NFS_MOUNT_NOCTO, ",nocto", "" },
{ NFS_MOUNT_NOAC, ",noac", "" },
{ NFS_MOUNT_NONLM, ",nolock", ",lock" },
@@ -585,6 +603,8 @@
};
struct proc_nfs_info *nfs_infop;
struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
+ char buf[12];
+ char *proto;
seq_printf(m, ",v%d", nfss->rpc_ops->version);
seq_printf(m, ",rsize=%d", nfss->rsize);
@@ -603,6 +623,18 @@
else
seq_puts(m, nfs_infop->nostr);
}
+ switch (nfss->client->cl_xprt->prot) {
+ case IPPROTO_TCP:
+ proto = "tcp";
+ break;
+ case IPPROTO_UDP:
+ proto = "udp";
+ break;
+ default:
+ snprintf(buf, sizeof(buf), "%u", nfss->client->cl_xprt->prot);
+ proto = buf;
+ }
+ seq_printf(m, ",proto=%s", proto);
seq_puts(m, ",addr=");
seq_escape(m, nfss->hostname, " \t\n\\");
return 0;
@@ -821,6 +853,11 @@
filemap_fdatawait(inode->i_mapping);
nfs_wb_all(inode);
}
+ /*
+ * Return any delegations if we're going to change ACLs
+ */
+ if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
+ nfs_inode_return_delegation(inode);
error = NFS_PROTO(inode)->setattr(dentry, &fattr, attr);
if (error == 0)
nfs_refresh_inode(inode, &fattr);
@@ -1639,8 +1676,7 @@
struct nfs_inode *nfsi = NFS_I(inode);
/* If we are holding a delegation, return it! */
- if (nfsi->delegation != NULL)
- nfs_inode_return_delegation(inode);
+ nfs_inode_return_delegation(inode);
/* First call standard NFS clear_inode() code */
nfs_clear_inode(inode);
/* Now clear out any remaining state */
@@ -1669,7 +1705,7 @@
struct rpc_clnt *clnt = NULL;
struct rpc_timeout timeparms;
rpc_authflavor_t authflavour;
- int proto, err = -EIO;
+ int err = -EIO;
sb->s_blocksize_bits = 0;
sb->s_blocksize = 0;
@@ -1687,30 +1723,8 @@
server->acdirmax = data->acdirmax*HZ;
server->rpc_ops = &nfs_v4_clientops;
- /* Initialize timeout values */
- timeparms.to_initval = data->timeo * HZ / 10;
- timeparms.to_retries = data->retrans;
- timeparms.to_exponential = 1;
- if (!timeparms.to_retries)
- timeparms.to_retries = 5;
-
- proto = data->proto;
- /* Which IP protocol do we use? */
- switch (proto) {
- case IPPROTO_TCP:
- timeparms.to_maxval = RPC_MAX_TCP_TIMEOUT;
- if (!timeparms.to_initval)
- timeparms.to_initval = 600 * HZ / 10;
- break;
- case IPPROTO_UDP:
- timeparms.to_maxval = RPC_MAX_UDP_TIMEOUT;
- if (!timeparms.to_initval)
- timeparms.to_initval = 11 * HZ / 10;
- break;
- default:
- return -EINVAL;
- }
+ nfs_init_timeout_values(&timeparms, data->proto, data->timeo, data->retrans);
clp = nfs4_get_client(&server->addr.sin_addr);
if (!clp) {
@@ -1735,7 +1749,7 @@
down_write(&clp->cl_sem);
if (IS_ERR(clp->cl_rpcclient)) {
- xprt = xprt_create_proto(proto, &server->addr, &timeparms);
+ xprt = xprt_create_proto(data->proto, &server->addr, &timeparms);
if (IS_ERR(xprt)) {
up_write(&clp->cl_sem);
err = PTR_ERR(xprt);
diff --git a/fs/nfs/nfs3proc.c b/fs/nfs/nfs3proc.c
index edc9551..e4a1cd4 100644
--- a/fs/nfs/nfs3proc.c
+++ b/fs/nfs/nfs3proc.c
@@ -299,7 +299,7 @@
*/
static int
nfs3_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
- int flags)
+ int flags, struct nameidata *nd)
{
struct nfs_fh fhandle;
struct nfs_fattr fattr;
@@ -735,7 +735,7 @@
static void
nfs3_read_done(struct rpc_task *task)
{
- struct nfs_write_data *data = (struct nfs_write_data *) task->tk_calldata;
+ struct nfs_read_data *data = (struct nfs_read_data *) task->tk_calldata;
if (nfs3_async_handle_jukebox(task))
return;
diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h
index ec1a22d..78a53f5 100644
--- a/fs/nfs/nfs4_fs.h
+++ b/fs/nfs/nfs4_fs.h
@@ -93,25 +93,50 @@
};
/*
+ * struct rpc_sequence ensures that RPC calls are sent in the exact
+ * order that they appear on the list.
+ */
+struct rpc_sequence {
+ struct rpc_wait_queue wait; /* RPC call delay queue */
+ spinlock_t lock; /* Protects the list */
+ struct list_head list; /* Defines sequence of RPC calls */
+};
+
+#define NFS_SEQID_CONFIRMED 1
+struct nfs_seqid_counter {
+ struct rpc_sequence *sequence;
+ int flags;
+ u32 counter;
+};
+
+struct nfs_seqid {
+ struct nfs_seqid_counter *sequence;
+ struct list_head list;
+};
+
+static inline void nfs_confirm_seqid(struct nfs_seqid_counter *seqid, int status)
+{
+ if (seqid_mutating_err(-status))
+ seqid->flags |= NFS_SEQID_CONFIRMED;
+}
+
+/*
* NFS4 state_owners and lock_owners are simply labels for ordered
* sequences of RPC calls. Their sole purpose is to provide once-only
* semantics by allowing the server to identify replayed requests.
- *
- * The ->so_sema is held during all state_owner seqid-mutating operations:
- * OPEN, OPEN_DOWNGRADE, and CLOSE. Its purpose is to properly serialize
- * so_seqid.
*/
struct nfs4_state_owner {
+ spinlock_t so_lock;
struct list_head so_list; /* per-clientid list of state_owners */
struct nfs4_client *so_client;
u32 so_id; /* 32-bit identifier, unique */
- struct semaphore so_sema;
- u32 so_seqid; /* protected by so_sema */
atomic_t so_count;
struct rpc_cred *so_cred; /* Associated cred */
struct list_head so_states;
struct list_head so_delegations;
+ struct nfs_seqid_counter so_seqid;
+ struct rpc_sequence so_sequence;
};
/*
@@ -132,7 +157,7 @@
fl_owner_t ls_owner; /* POSIX lock owner */
#define NFS_LOCK_INITIALIZED 1
int ls_flags;
- u32 ls_seqid;
+ struct nfs_seqid_counter ls_seqid;
u32 ls_id;
nfs4_stateid ls_stateid;
atomic_t ls_count;
@@ -153,7 +178,6 @@
struct inode *inode; /* Pointer to the inode */
unsigned long flags; /* Do we hold any locks? */
- struct semaphore lock_sema; /* Serializes file locking operations */
spinlock_t state_lock; /* Protects the lock_states list */
nfs4_stateid stateid;
@@ -191,8 +215,8 @@
extern int nfs4_proc_async_renew(struct nfs4_client *);
extern int nfs4_proc_renew(struct nfs4_client *);
extern int nfs4_do_close(struct inode *inode, struct nfs4_state *state, mode_t mode);
-extern struct inode *nfs4_atomic_open(struct inode *, struct dentry *, struct nameidata *);
-extern int nfs4_open_revalidate(struct inode *, struct dentry *, int);
+extern struct dentry *nfs4_atomic_open(struct inode *, struct dentry *, struct nameidata *);
+extern int nfs4_open_revalidate(struct inode *, struct dentry *, int, struct nameidata *);
extern struct nfs4_state_recovery_ops nfs4_reboot_recovery_ops;
extern struct nfs4_state_recovery_ops nfs4_network_partition_recovery_ops;
@@ -224,12 +248,17 @@
extern void nfs4_put_open_state(struct nfs4_state *);
extern void nfs4_close_state(struct nfs4_state *, mode_t);
extern struct nfs4_state *nfs4_find_state(struct inode *, struct rpc_cred *, mode_t mode);
-extern void nfs4_increment_seqid(int status, struct nfs4_state_owner *sp);
extern void nfs4_schedule_state_recovery(struct nfs4_client *);
+extern void nfs4_put_lock_state(struct nfs4_lock_state *lsp);
extern int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl);
-extern void nfs4_increment_lock_seqid(int status, struct nfs4_lock_state *ls);
extern void nfs4_copy_stateid(nfs4_stateid *, struct nfs4_state *, fl_owner_t);
+extern struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter);
+extern int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task);
+extern void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid);
+extern void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid);
+extern void nfs_free_seqid(struct nfs_seqid *seqid);
+
extern const nfs4_stateid zero_stateid;
/* nfs4xdr.c */
diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index 9701ca8..9c1da34 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -47,6 +47,7 @@
#include <linux/nfs_page.h>
#include <linux/smp_lock.h>
#include <linux/namei.h>
+#include <linux/mount.h>
#include "nfs4_fs.h"
#include "delegation.h"
@@ -56,10 +57,11 @@
#define NFS4_POLL_RETRY_MIN (1*HZ)
#define NFS4_POLL_RETRY_MAX (15*HZ)
+static int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nfs_fh *fh, struct nfs4_state_owner *sp, nfs4_stateid *stateid, struct nfs_seqid *seqid);
static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
-static int nfs4_async_handle_error(struct rpc_task *, struct nfs_server *);
+static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *);
static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry);
-static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception);
+static int nfs4_handle_exception(const struct nfs_server *server, int errorcode, struct nfs4_exception *exception);
extern u32 *nfs4_decode_dirent(u32 *p, struct nfs_entry *entry, int plus);
extern struct rpc_procinfo nfs4_procedures[];
@@ -189,12 +191,28 @@
nfsi->change_attr = cinfo->after;
}
+/* Helper for asynchronous RPC calls */
+static int nfs4_call_async(struct rpc_clnt *clnt, rpc_action tk_begin,
+ rpc_action tk_exit, void *calldata)
+{
+ struct rpc_task *task;
+
+ if (!(task = rpc_new_task(clnt, tk_exit, RPC_TASK_ASYNC)))
+ return -ENOMEM;
+
+ task->tk_calldata = calldata;
+ task->tk_action = tk_begin;
+ rpc_execute(task);
+ return 0;
+}
+
static void update_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, int open_flags)
{
struct inode *inode = state->inode;
open_flags &= (FMODE_READ|FMODE_WRITE);
/* Protect against nfs4_find_state() */
+ spin_lock(&state->owner->so_lock);
spin_lock(&inode->i_lock);
state->state |= open_flags;
/* NB! List reordering - see the reclaim code for why. */
@@ -204,12 +222,12 @@
state->nreaders++;
memcpy(&state->stateid, stateid, sizeof(state->stateid));
spin_unlock(&inode->i_lock);
+ spin_unlock(&state->owner->so_lock);
}
/*
* OPEN_RECLAIM:
* reclaim state on the server after a reboot.
- * Assumes caller is holding the sp->so_sem
*/
static int _nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
{
@@ -218,7 +236,6 @@
struct nfs_delegation *delegation = NFS_I(inode)->delegation;
struct nfs_openargs o_arg = {
.fh = NFS_FH(inode),
- .seqid = sp->so_seqid,
.id = sp->so_id,
.open_flags = state->state,
.clientid = server->nfs4_state->cl_clientid,
@@ -245,8 +262,13 @@
}
o_arg.u.delegation_type = delegation->type;
}
+ o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid);
+ if (o_arg.seqid == NULL)
+ return -ENOMEM;
status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
- nfs4_increment_seqid(status, sp);
+ /* Confirm the sequence as being established */
+ nfs_confirm_seqid(&sp->so_seqid, status);
+ nfs_increment_open_seqid(status, o_arg.seqid);
if (status == 0) {
memcpy(&state->stateid, &o_res.stateid, sizeof(state->stateid));
if (o_res.delegation_type != 0) {
@@ -256,6 +278,7 @@
nfs_async_inode_return_delegation(inode, &o_res.stateid);
}
}
+ nfs_free_seqid(o_arg.seqid);
clear_bit(NFS_DELEGATED_STATE, &state->flags);
/* Ensure we update the inode attributes */
NFS_CACHEINV(inode);
@@ -302,23 +325,35 @@
};
int status = 0;
- down(&sp->so_sema);
if (!test_bit(NFS_DELEGATED_STATE, &state->flags))
goto out;
if (state->state == 0)
goto out;
- arg.seqid = sp->so_seqid;
+ arg.seqid = nfs_alloc_seqid(&sp->so_seqid);
+ status = -ENOMEM;
+ if (arg.seqid == NULL)
+ goto out;
arg.open_flags = state->state;
memcpy(arg.u.delegation.data, state->stateid.data, sizeof(arg.u.delegation.data));
status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
- nfs4_increment_seqid(status, sp);
+ nfs_increment_open_seqid(status, arg.seqid);
+ if (status != 0)
+ goto out_free;
+ if(res.rflags & NFS4_OPEN_RESULT_CONFIRM) {
+ status = _nfs4_proc_open_confirm(server->client, NFS_FH(inode),
+ sp, &res.stateid, arg.seqid);
+ if (status != 0)
+ goto out_free;
+ }
+ nfs_confirm_seqid(&sp->so_seqid, 0);
if (status >= 0) {
memcpy(state->stateid.data, res.stateid.data,
sizeof(state->stateid.data));
clear_bit(NFS_DELEGATED_STATE, &state->flags);
}
+out_free:
+ nfs_free_seqid(arg.seqid);
out:
- up(&sp->so_sema);
dput(parent);
return status;
}
@@ -345,11 +380,11 @@
return err;
}
-static inline int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nfs_fh *fh, struct nfs4_state_owner *sp, nfs4_stateid *stateid)
+static int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nfs_fh *fh, struct nfs4_state_owner *sp, nfs4_stateid *stateid, struct nfs_seqid *seqid)
{
struct nfs_open_confirmargs arg = {
.fh = fh,
- .seqid = sp->so_seqid,
+ .seqid = seqid,
.stateid = *stateid,
};
struct nfs_open_confirmres res;
@@ -362,7 +397,9 @@
int status;
status = rpc_call_sync(clnt, &msg, RPC_TASK_NOINTR);
- nfs4_increment_seqid(status, sp);
+ /* Confirm the sequence as being established */
+ nfs_confirm_seqid(&sp->so_seqid, status);
+ nfs_increment_open_seqid(status, seqid);
if (status >= 0)
memcpy(stateid, &res.stateid, sizeof(*stateid));
return status;
@@ -380,21 +417,37 @@
int status;
/* Update sequence id. The caller must serialize! */
- o_arg->seqid = sp->so_seqid;
o_arg->id = sp->so_id;
o_arg->clientid = sp->so_client->cl_clientid;
status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
- nfs4_increment_seqid(status, sp);
+ if (status == 0) {
+ /* OPEN on anything except a regular file is disallowed in NFSv4 */
+ switch (o_res->f_attr->mode & S_IFMT) {
+ case S_IFREG:
+ break;
+ case S_IFLNK:
+ status = -ELOOP;
+ break;
+ case S_IFDIR:
+ status = -EISDIR;
+ break;
+ default:
+ status = -ENOTDIR;
+ }
+ }
+
+ nfs_increment_open_seqid(status, o_arg->seqid);
if (status != 0)
goto out;
update_changeattr(dir, &o_res->cinfo);
if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
status = _nfs4_proc_open_confirm(server->client, &o_res->fh,
- sp, &o_res->stateid);
+ sp, &o_res->stateid, o_arg->seqid);
if (status != 0)
goto out;
}
+ nfs_confirm_seqid(&sp->so_seqid, 0);
if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
status = server->rpc_ops->getattr(server, &o_res->fh, o_res->f_attr);
out:
@@ -465,6 +518,10 @@
set_bit(NFS_DELEGATED_STATE, &state->flags);
goto out;
}
+ o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid);
+ status = -ENOMEM;
+ if (o_arg.seqid == NULL)
+ goto out;
status = _nfs4_proc_open(dir, sp, &o_arg, &o_res);
if (status != 0)
goto out_nodeleg;
@@ -490,6 +547,7 @@
nfs_inode_reclaim_delegation(inode, sp->so_cred, &o_res);
}
out_nodeleg:
+ nfs_free_seqid(o_arg.seqid);
clear_bit(NFS_DELEGATED_STATE, &state->flags);
out:
dput(parent);
@@ -564,7 +622,6 @@
dprintk("%s: nfs4_get_state_owner failed!\n", __FUNCTION__);
goto out_err;
}
- down(&sp->so_sema);
state = nfs4_get_open_state(inode, sp);
if (state == NULL)
goto out_err;
@@ -589,7 +646,6 @@
set_bit(NFS_DELEGATED_STATE, &state->flags);
update_open_stateid(state, &delegation->stateid, open_flags);
out_ok:
- up(&sp->so_sema);
nfs4_put_state_owner(sp);
up_read(&nfsi->rwsem);
up_read(&clp->cl_sem);
@@ -600,11 +656,12 @@
if (sp != NULL) {
if (state != NULL)
nfs4_put_open_state(state);
- up(&sp->so_sema);
nfs4_put_state_owner(sp);
}
up_read(&nfsi->rwsem);
up_read(&clp->cl_sem);
+ if (err != -EACCES)
+ nfs_inode_return_delegation(inode);
return err;
}
@@ -665,8 +722,10 @@
} else
o_arg.u.attrs = sattr;
/* Serialization for the sequence id */
- down(&sp->so_sema);
+ o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid);
+ if (o_arg.seqid == NULL)
+ return -ENOMEM;
status = _nfs4_proc_open(dir, sp, &o_arg, &o_res);
if (status != 0)
goto out_err;
@@ -681,7 +740,7 @@
update_open_stateid(state, &o_res.stateid, flags);
if (o_res.delegation_type != 0)
nfs_inode_set_delegation(inode, cred, &o_res);
- up(&sp->so_sema);
+ nfs_free_seqid(o_arg.seqid);
nfs4_put_state_owner(sp);
up_read(&clp->cl_sem);
*res = state;
@@ -690,7 +749,7 @@
if (sp != NULL) {
if (state != NULL)
nfs4_put_open_state(state);
- up(&sp->so_sema);
+ nfs_free_seqid(o_arg.seqid);
nfs4_put_state_owner(sp);
}
/* Note: clp->cl_sem must be released before nfs4_put_open_state()! */
@@ -718,7 +777,7 @@
* It is actually a sign of a bug on the client or on the server.
*
* If we receive a BAD_SEQID error in the particular case of
- * doing an OPEN, we assume that nfs4_increment_seqid() will
+ * doing an OPEN, we assume that nfs_increment_open_seqid() will
* have unhashed the old state_owner for us, and that we can
* therefore safely retry using a new one. We should still warn
* the user though...
@@ -728,6 +787,16 @@
exception.retry = 1;
continue;
}
+ /*
+ * BAD_STATEID on OPEN means that the server cancelled our
+ * state before it received the OPEN_CONFIRM.
+ * Recover by retrying the request as per the discussion
+ * on Page 181 of RFC3530.
+ */
+ if (status == -NFS4ERR_BAD_STATEID) {
+ exception.retry = 1;
+ continue;
+ }
res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
status, &exception));
} while (exception.retry);
@@ -789,17 +858,27 @@
struct nfs_closeres res;
};
+static void nfs4_free_closedata(struct nfs4_closedata *calldata)
+{
+ struct nfs4_state *state = calldata->state;
+ struct nfs4_state_owner *sp = state->owner;
+
+ nfs4_put_open_state(calldata->state);
+ nfs_free_seqid(calldata->arg.seqid);
+ nfs4_put_state_owner(sp);
+ kfree(calldata);
+}
+
static void nfs4_close_done(struct rpc_task *task)
{
struct nfs4_closedata *calldata = (struct nfs4_closedata *)task->tk_calldata;
struct nfs4_state *state = calldata->state;
- struct nfs4_state_owner *sp = state->owner;
struct nfs_server *server = NFS_SERVER(calldata->inode);
/* hmm. we are done with the inode, and in the process of freeing
* the state_owner. we keep this around to process errors
*/
- nfs4_increment_seqid(task->tk_status, sp);
+ nfs_increment_open_seqid(task->tk_status, calldata->arg.seqid);
switch (task->tk_status) {
case 0:
memcpy(&state->stateid, &calldata->res.stateid,
@@ -817,24 +896,46 @@
}
}
state->state = calldata->arg.open_flags;
- nfs4_put_open_state(state);
- up(&sp->so_sema);
- nfs4_put_state_owner(sp);
- up_read(&server->nfs4_state->cl_sem);
- kfree(calldata);
+ nfs4_free_closedata(calldata);
}
-static inline int nfs4_close_call(struct rpc_clnt *clnt, struct nfs4_closedata *calldata)
+static void nfs4_close_begin(struct rpc_task *task)
{
+ struct nfs4_closedata *calldata = (struct nfs4_closedata *)task->tk_calldata;
+ struct nfs4_state *state = calldata->state;
struct rpc_message msg = {
.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
.rpc_argp = &calldata->arg,
.rpc_resp = &calldata->res,
- .rpc_cred = calldata->state->owner->so_cred,
+ .rpc_cred = state->owner->so_cred,
};
- if (calldata->arg.open_flags != 0)
+ int mode = 0;
+ int status;
+
+ status = nfs_wait_on_sequence(calldata->arg.seqid, task);
+ if (status != 0)
+ return;
+ /* Don't reorder reads */
+ smp_rmb();
+ /* Recalculate the new open mode in case someone reopened the file
+ * while we were waiting in line to be scheduled.
+ */
+ if (state->nreaders != 0)
+ mode |= FMODE_READ;
+ if (state->nwriters != 0)
+ mode |= FMODE_WRITE;
+ if (test_bit(NFS_DELEGATED_STATE, &state->flags))
+ state->state = mode;
+ if (mode == state->state) {
+ nfs4_free_closedata(calldata);
+ task->tk_exit = NULL;
+ rpc_exit(task, 0);
+ return;
+ }
+ if (mode != 0)
msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
- return rpc_call_async(clnt, &msg, 0, nfs4_close_done, calldata);
+ calldata->arg.open_flags = mode;
+ rpc_call_setup(task, &msg, 0);
}
/*
@@ -851,39 +952,52 @@
int nfs4_do_close(struct inode *inode, struct nfs4_state *state, mode_t mode)
{
struct nfs4_closedata *calldata;
- int status;
+ int status = -ENOMEM;
- /* Tell caller we're done */
- if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
- state->state = mode;
- return 0;
- }
- calldata = (struct nfs4_closedata *)kmalloc(sizeof(*calldata), GFP_KERNEL);
+ calldata = kmalloc(sizeof(*calldata), GFP_KERNEL);
if (calldata == NULL)
- return -ENOMEM;
+ goto out;
calldata->inode = inode;
calldata->state = state;
calldata->arg.fh = NFS_FH(inode);
+ calldata->arg.stateid = &state->stateid;
/* Serialization for the sequence id */
- calldata->arg.seqid = state->owner->so_seqid;
- calldata->arg.open_flags = mode;
- memcpy(&calldata->arg.stateid, &state->stateid,
- sizeof(calldata->arg.stateid));
- status = nfs4_close_call(NFS_SERVER(inode)->client, calldata);
- /*
- * Return -EINPROGRESS on success in order to indicate to the
- * caller that an asynchronous RPC call has been launched, and
- * that it will release the semaphores on completion.
- */
- return (status == 0) ? -EINPROGRESS : status;
+ calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid);
+ if (calldata->arg.seqid == NULL)
+ goto out_free_calldata;
+
+ status = nfs4_call_async(NFS_SERVER(inode)->client, nfs4_close_begin,
+ nfs4_close_done, calldata);
+ if (status == 0)
+ goto out;
+
+ nfs_free_seqid(calldata->arg.seqid);
+out_free_calldata:
+ kfree(calldata);
+out:
+ return status;
}
-struct inode *
+static void nfs4_intent_set_file(struct nameidata *nd, struct dentry *dentry, struct nfs4_state *state)
+{
+ struct file *filp;
+
+ filp = lookup_instantiate_filp(nd, dentry, NULL);
+ if (!IS_ERR(filp)) {
+ struct nfs_open_context *ctx;
+ ctx = (struct nfs_open_context *)filp->private_data;
+ ctx->state = state;
+ } else
+ nfs4_close_state(state, nd->intent.open.flags);
+}
+
+struct dentry *
nfs4_atomic_open(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
{
struct iattr attr;
struct rpc_cred *cred;
struct nfs4_state *state;
+ struct dentry *res;
if (nd->flags & LOOKUP_CREATE) {
attr.ia_mode = nd->intent.open.create_mode;
@@ -897,16 +1011,23 @@
cred = rpcauth_lookupcred(NFS_SERVER(dir)->client->cl_auth, 0);
if (IS_ERR(cred))
- return (struct inode *)cred;
+ return (struct dentry *)cred;
state = nfs4_do_open(dir, dentry, nd->intent.open.flags, &attr, cred);
put_rpccred(cred);
- if (IS_ERR(state))
- return (struct inode *)state;
- return state->inode;
+ if (IS_ERR(state)) {
+ if (PTR_ERR(state) == -ENOENT)
+ d_add(dentry, NULL);
+ return (struct dentry *)state;
+ }
+ res = d_add_unique(dentry, state->inode);
+ if (res != NULL)
+ dentry = res;
+ nfs4_intent_set_file(nd, dentry, state);
+ return res;
}
int
-nfs4_open_revalidate(struct inode *dir, struct dentry *dentry, int openflags)
+nfs4_open_revalidate(struct inode *dir, struct dentry *dentry, int openflags, struct nameidata *nd)
{
struct rpc_cred *cred;
struct nfs4_state *state;
@@ -919,18 +1040,30 @@
if (IS_ERR(state))
state = nfs4_do_open(dir, dentry, openflags, NULL, cred);
put_rpccred(cred);
- if (state == ERR_PTR(-ENOENT) && dentry->d_inode == 0)
- return 1;
- if (IS_ERR(state))
- return 0;
+ if (IS_ERR(state)) {
+ switch (PTR_ERR(state)) {
+ case -EPERM:
+ case -EACCES:
+ case -EDQUOT:
+ case -ENOSPC:
+ case -EROFS:
+ lookup_instantiate_filp(nd, (struct dentry *)state, NULL);
+ return 1;
+ case -ENOENT:
+ if (dentry->d_inode == NULL)
+ return 1;
+ }
+ goto out_drop;
+ }
inode = state->inode;
+ iput(inode);
if (inode == dentry->d_inode) {
- iput(inode);
+ nfs4_intent_set_file(nd, dentry, state);
return 1;
}
- d_drop(dentry);
nfs4_close_state(state, openflags);
- iput(inode);
+out_drop:
+ d_drop(dentry);
return 0;
}
@@ -1431,7 +1564,7 @@
static int
nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
- int flags)
+ int flags, struct nameidata *nd)
{
struct nfs4_state *state;
struct rpc_cred *cred;
@@ -1453,13 +1586,13 @@
struct nfs_fattr fattr;
status = nfs4_do_setattr(NFS_SERVER(dir), &fattr,
NFS_FH(state->inode), sattr, state);
- if (status == 0) {
+ if (status == 0)
nfs_setattr_update_inode(state->inode, sattr);
- goto out;
- }
- } else if (flags != 0)
- goto out;
- nfs4_close_state(state, flags);
+ }
+ if (status == 0 && nd != NULL && (nd->flags & LOOKUP_OPEN))
+ nfs4_intent_set_file(nd, dentry, state);
+ else
+ nfs4_close_state(state, flags);
out:
return status;
}
@@ -2106,65 +2239,6 @@
return 0;
}
-/*
- * We will need to arrange for the VFS layer to provide an atomic open.
- * Until then, this open method is prone to inefficiency and race conditions
- * due to the lookup, potential create, and open VFS calls from sys_open()
- * placed on the wire.
- */
-static int
-nfs4_proc_file_open(struct inode *inode, struct file *filp)
-{
- struct dentry *dentry = filp->f_dentry;
- struct nfs_open_context *ctx;
- struct nfs4_state *state = NULL;
- struct rpc_cred *cred;
- int status = -ENOMEM;
-
- dprintk("nfs4_proc_file_open: starting on (%.*s/%.*s)\n",
- (int)dentry->d_parent->d_name.len,
- dentry->d_parent->d_name.name,
- (int)dentry->d_name.len, dentry->d_name.name);
-
-
- /* Find our open stateid */
- cred = rpcauth_lookupcred(NFS_SERVER(inode)->client->cl_auth, 0);
- if (IS_ERR(cred))
- return PTR_ERR(cred);
- ctx = alloc_nfs_open_context(dentry, cred);
- put_rpccred(cred);
- if (unlikely(ctx == NULL))
- return -ENOMEM;
- status = -EIO; /* ERACE actually */
- state = nfs4_find_state(inode, cred, filp->f_mode);
- if (unlikely(state == NULL))
- goto no_state;
- ctx->state = state;
- nfs4_close_state(state, filp->f_mode);
- ctx->mode = filp->f_mode;
- nfs_file_set_open_context(filp, ctx);
- put_nfs_open_context(ctx);
- if (filp->f_mode & FMODE_WRITE)
- nfs_begin_data_update(inode);
- return 0;
-no_state:
- printk(KERN_WARNING "NFS: v4 raced in function %s\n", __FUNCTION__);
- put_nfs_open_context(ctx);
- return status;
-}
-
-/*
- * Release our state
- */
-static int
-nfs4_proc_file_release(struct inode *inode, struct file *filp)
-{
- if (filp->f_mode & FMODE_WRITE)
- nfs_end_data_update(inode);
- nfs_file_clear_open_context(filp);
- return 0;
-}
-
static inline int nfs4_server_supports_acls(struct nfs_server *server)
{
return (server->caps & NFS_CAP_ACLS)
@@ -2285,7 +2359,7 @@
return -ENOMEM;
args.acl_pages[0] = localpage;
args.acl_pgbase = 0;
- args.acl_len = PAGE_SIZE;
+ resp_len = args.acl_len = PAGE_SIZE;
} else {
resp_buf = buf;
buf_to_pages(buf, buflen, args.acl_pages, &args.acl_pgbase);
@@ -2345,6 +2419,7 @@
if (!nfs4_server_supports_acls(server))
return -EOPNOTSUPP;
+ nfs_inode_return_delegation(inode);
buf_to_pages(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
ret = rpc_call_sync(NFS_SERVER(inode)->client, &msg, 0);
if (ret == 0)
@@ -2353,7 +2428,7 @@
}
static int
-nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server)
+nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server)
{
struct nfs4_client *clp = server->nfs4_state;
@@ -2431,7 +2506,7 @@
/* This is the error handling routine for processes that are allowed
* to sleep.
*/
-int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
+int nfs4_handle_exception(const struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
{
struct nfs4_client *clp = server->nfs4_state;
int ret = errorcode;
@@ -2632,7 +2707,6 @@
down_read(&clp->cl_sem);
nlo.clientid = clp->cl_clientid;
- down(&state->lock_sema);
status = nfs4_set_lock_state(state, request);
if (status != 0)
goto out;
@@ -2659,7 +2733,6 @@
status = 0;
}
out:
- up(&state->lock_sema);
up_read(&clp->cl_sem);
return status;
}
@@ -2696,67 +2769,125 @@
return res;
}
-static int _nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
+struct nfs4_unlockdata {
+ struct nfs_lockargs arg;
+ struct nfs_locku_opargs luargs;
+ struct nfs_lockres res;
+ struct nfs4_lock_state *lsp;
+ struct nfs_open_context *ctx;
+ atomic_t refcount;
+ struct completion completion;
+};
+
+static void nfs4_locku_release_calldata(struct nfs4_unlockdata *calldata)
{
- struct inode *inode = state->inode;
- struct nfs_server *server = NFS_SERVER(inode);
- struct nfs4_client *clp = server->nfs4_state;
- struct nfs_lockargs arg = {
- .fh = NFS_FH(inode),
- .type = nfs4_lck_type(cmd, request),
- .offset = request->fl_start,
- .length = nfs4_lck_length(request),
- };
- struct nfs_lockres res = {
- .server = server,
- };
+ if (atomic_dec_and_test(&calldata->refcount)) {
+ nfs_free_seqid(calldata->luargs.seqid);
+ nfs4_put_lock_state(calldata->lsp);
+ put_nfs_open_context(calldata->ctx);
+ kfree(calldata);
+ }
+}
+
+static void nfs4_locku_complete(struct nfs4_unlockdata *calldata)
+{
+ complete(&calldata->completion);
+ nfs4_locku_release_calldata(calldata);
+}
+
+static void nfs4_locku_done(struct rpc_task *task)
+{
+ struct nfs4_unlockdata *calldata = (struct nfs4_unlockdata *)task->tk_calldata;
+
+ nfs_increment_lock_seqid(task->tk_status, calldata->luargs.seqid);
+ switch (task->tk_status) {
+ case 0:
+ memcpy(calldata->lsp->ls_stateid.data,
+ calldata->res.u.stateid.data,
+ sizeof(calldata->lsp->ls_stateid.data));
+ break;
+ case -NFS4ERR_STALE_STATEID:
+ case -NFS4ERR_EXPIRED:
+ nfs4_schedule_state_recovery(calldata->res.server->nfs4_state);
+ break;
+ default:
+ if (nfs4_async_handle_error(task, calldata->res.server) == -EAGAIN) {
+ rpc_restart_call(task);
+ return;
+ }
+ }
+ nfs4_locku_complete(calldata);
+}
+
+static void nfs4_locku_begin(struct rpc_task *task)
+{
+ struct nfs4_unlockdata *calldata = (struct nfs4_unlockdata *)task->tk_calldata;
struct rpc_message msg = {
.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
- .rpc_argp = &arg,
- .rpc_resp = &res,
- .rpc_cred = state->owner->so_cred,
+ .rpc_argp = &calldata->arg,
+ .rpc_resp = &calldata->res,
+ .rpc_cred = calldata->lsp->ls_state->owner->so_cred,
};
- struct nfs4_lock_state *lsp;
- struct nfs_locku_opargs luargs;
int status;
-
- down_read(&clp->cl_sem);
- down(&state->lock_sema);
- status = nfs4_set_lock_state(state, request);
- if (status != 0)
- goto out;
- lsp = request->fl_u.nfs4_fl.owner;
- /* We might have lost the locks! */
- if ((lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0)
- goto out;
- luargs.seqid = lsp->ls_seqid;
- memcpy(&luargs.stateid, &lsp->ls_stateid, sizeof(luargs.stateid));
- arg.u.locku = &luargs;
- status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
- nfs4_increment_lock_seqid(status, lsp);
- if (status == 0)
- memcpy(&lsp->ls_stateid, &res.u.stateid,
- sizeof(lsp->ls_stateid));
-out:
- up(&state->lock_sema);
- if (status == 0)
- do_vfs_lock(request->fl_file, request);
- up_read(&clp->cl_sem);
- return status;
+ status = nfs_wait_on_sequence(calldata->luargs.seqid, task);
+ if (status != 0)
+ return;
+ if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
+ nfs4_locku_complete(calldata);
+ task->tk_exit = NULL;
+ rpc_exit(task, 0);
+ return;
+ }
+ rpc_call_setup(task, &msg, 0);
}
static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
{
- struct nfs4_exception exception = { };
- int err;
+ struct nfs4_unlockdata *calldata;
+ struct inode *inode = state->inode;
+ struct nfs_server *server = NFS_SERVER(inode);
+ struct nfs4_lock_state *lsp;
+ int status;
- do {
- err = nfs4_handle_exception(NFS_SERVER(state->inode),
- _nfs4_proc_unlck(state, cmd, request),
- &exception);
- } while (exception.retry);
- return err;
+ status = nfs4_set_lock_state(state, request);
+ if (status != 0)
+ return status;
+ lsp = request->fl_u.nfs4_fl.owner;
+ /* We might have lost the locks! */
+ if ((lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0)
+ return 0;
+ calldata = kmalloc(sizeof(*calldata), GFP_KERNEL);
+ if (calldata == NULL)
+ return -ENOMEM;
+ calldata->luargs.seqid = nfs_alloc_seqid(&lsp->ls_seqid);
+ if (calldata->luargs.seqid == NULL) {
+ kfree(calldata);
+ return -ENOMEM;
+ }
+ calldata->luargs.stateid = &lsp->ls_stateid;
+ calldata->arg.fh = NFS_FH(inode);
+ calldata->arg.type = nfs4_lck_type(cmd, request);
+ calldata->arg.offset = request->fl_start;
+ calldata->arg.length = nfs4_lck_length(request);
+ calldata->arg.u.locku = &calldata->luargs;
+ calldata->res.server = server;
+ calldata->lsp = lsp;
+ atomic_inc(&lsp->ls_count);
+
+ /* Ensure we don't close file until we're done freeing locks! */
+ calldata->ctx = get_nfs_open_context((struct nfs_open_context*)request->fl_file->private_data);
+
+ atomic_set(&calldata->refcount, 2);
+ init_completion(&calldata->completion);
+
+ status = nfs4_call_async(NFS_SERVER(inode)->client, nfs4_locku_begin,
+ nfs4_locku_done, calldata);
+ if (status == 0)
+ wait_for_completion_interruptible(&calldata->completion);
+ do_vfs_lock(request->fl_file, request);
+ nfs4_locku_release_calldata(calldata);
+ return status;
}
static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *request, int reclaim)
@@ -2764,11 +2895,23 @@
struct inode *inode = state->inode;
struct nfs_server *server = NFS_SERVER(inode);
struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
+ struct nfs_lock_opargs largs = {
+ .lock_stateid = &lsp->ls_stateid,
+ .open_stateid = &state->stateid,
+ .lock_owner = {
+ .clientid = server->nfs4_state->cl_clientid,
+ .id = lsp->ls_id,
+ },
+ .reclaim = reclaim,
+ };
struct nfs_lockargs arg = {
.fh = NFS_FH(inode),
.type = nfs4_lck_type(cmd, request),
.offset = request->fl_start,
.length = nfs4_lck_length(request),
+ .u = {
+ .lock = &largs,
+ },
};
struct nfs_lockres res = {
.server = server,
@@ -2779,53 +2922,39 @@
.rpc_resp = &res,
.rpc_cred = state->owner->so_cred,
};
- struct nfs_lock_opargs largs = {
- .reclaim = reclaim,
- .new_lock_owner = 0,
- };
- int status;
+ int status = -ENOMEM;
- if (!(lsp->ls_flags & NFS_LOCK_INITIALIZED)) {
+ largs.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid);
+ if (largs.lock_seqid == NULL)
+ return -ENOMEM;
+ if (!(lsp->ls_seqid.flags & NFS_SEQID_CONFIRMED)) {
struct nfs4_state_owner *owner = state->owner;
- struct nfs_open_to_lock otl = {
- .lock_owner = {
- .clientid = server->nfs4_state->cl_clientid,
- },
- };
- otl.lock_seqid = lsp->ls_seqid;
- otl.lock_owner.id = lsp->ls_id;
- memcpy(&otl.open_stateid, &state->stateid, sizeof(otl.open_stateid));
- largs.u.open_lock = &otl;
+ largs.open_seqid = nfs_alloc_seqid(&owner->so_seqid);
+ if (largs.open_seqid == NULL)
+ goto out;
largs.new_lock_owner = 1;
- arg.u.lock = &largs;
- down(&owner->so_sema);
- otl.open_seqid = owner->so_seqid;
status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
- /* increment open_owner seqid on success, and
- * seqid mutating errors */
- nfs4_increment_seqid(status, owner);
- up(&owner->so_sema);
- if (status == 0) {
- lsp->ls_flags |= NFS_LOCK_INITIALIZED;
- lsp->ls_seqid++;
+ /* increment open seqid on success, and seqid mutating errors */
+ if (largs.new_lock_owner != 0) {
+ nfs_increment_open_seqid(status, largs.open_seqid);
+ if (status == 0)
+ nfs_confirm_seqid(&lsp->ls_seqid, 0);
}
- } else {
- struct nfs_exist_lock el = {
- .seqid = lsp->ls_seqid,
- };
- memcpy(&el.stateid, &lsp->ls_stateid, sizeof(el.stateid));
- largs.u.exist_lock = ⪙
- arg.u.lock = &largs;
+ nfs_free_seqid(largs.open_seqid);
+ } else
status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
- /* increment seqid on success, and * seqid mutating errors*/
- nfs4_increment_lock_seqid(status, lsp);
- }
+ /* increment lock seqid on success, and seqid mutating errors*/
+ nfs_increment_lock_seqid(status, largs.lock_seqid);
/* save the returned stateid. */
- if (status == 0)
- memcpy(&lsp->ls_stateid, &res.u.stateid, sizeof(nfs4_stateid));
- else if (status == -NFS4ERR_DENIED)
+ if (status == 0) {
+ memcpy(lsp->ls_stateid.data, res.u.stateid.data,
+ sizeof(lsp->ls_stateid.data));
+ lsp->ls_flags |= NFS_LOCK_INITIALIZED;
+ } else if (status == -NFS4ERR_DENIED)
status = -EAGAIN;
+out:
+ nfs_free_seqid(largs.lock_seqid);
return status;
}
@@ -2865,11 +2994,9 @@
int status;
down_read(&clp->cl_sem);
- down(&state->lock_sema);
status = nfs4_set_lock_state(state, request);
if (status == 0)
status = _nfs4_do_setlk(state, cmd, request, 0);
- up(&state->lock_sema);
if (status == 0) {
/* Note: we always want to sleep here! */
request->fl_flags |= FL_SLEEP;
@@ -3024,8 +3151,8 @@
.read_setup = nfs4_proc_read_setup,
.write_setup = nfs4_proc_write_setup,
.commit_setup = nfs4_proc_commit_setup,
- .file_open = nfs4_proc_file_open,
- .file_release = nfs4_proc_file_release,
+ .file_open = nfs_open,
+ .file_release = nfs_release,
.lock = nfs4_proc_lock,
.clear_acl_cache = nfs4_zap_acl_attr,
};
diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c
index afe587d..2d5a6a2 100644
--- a/fs/nfs/nfs4state.c
+++ b/fs/nfs/nfs4state.c
@@ -264,13 +264,16 @@
{
struct nfs4_state_owner *sp;
- sp = kmalloc(sizeof(*sp),GFP_KERNEL);
+ sp = kzalloc(sizeof(*sp),GFP_KERNEL);
if (!sp)
return NULL;
- init_MUTEX(&sp->so_sema);
- sp->so_seqid = 0; /* arbitrary */
+ spin_lock_init(&sp->so_lock);
INIT_LIST_HEAD(&sp->so_states);
INIT_LIST_HEAD(&sp->so_delegations);
+ rpc_init_wait_queue(&sp->so_sequence.wait, "Seqid_waitqueue");
+ sp->so_seqid.sequence = &sp->so_sequence;
+ spin_lock_init(&sp->so_sequence.lock);
+ INIT_LIST_HEAD(&sp->so_sequence.list);
atomic_set(&sp->so_count, 1);
return sp;
}
@@ -359,7 +362,6 @@
memset(state->stateid.data, 0, sizeof(state->stateid.data));
atomic_set(&state->count, 1);
INIT_LIST_HEAD(&state->lock_states);
- init_MUTEX(&state->lock_sema);
spin_lock_init(&state->state_lock);
return state;
}
@@ -437,21 +439,23 @@
if (state)
goto out;
new = nfs4_alloc_open_state();
+ spin_lock(&owner->so_lock);
spin_lock(&inode->i_lock);
state = __nfs4_find_state_byowner(inode, owner);
if (state == NULL && new != NULL) {
state = new;
- /* Caller *must* be holding owner->so_sem */
- /* Note: The reclaim code dictates that we add stateless
- * and read-only stateids to the end of the list */
- list_add_tail(&state->open_states, &owner->so_states);
state->owner = owner;
atomic_inc(&owner->so_count);
list_add(&state->inode_states, &nfsi->open_states);
state->inode = igrab(inode);
spin_unlock(&inode->i_lock);
+ /* Note: The reclaim code dictates that we add stateless
+ * and read-only stateids to the end of the list */
+ list_add_tail(&state->open_states, &owner->so_states);
+ spin_unlock(&owner->so_lock);
} else {
spin_unlock(&inode->i_lock);
+ spin_unlock(&owner->so_lock);
if (new)
nfs4_free_open_state(new);
}
@@ -461,19 +465,21 @@
/*
* Beware! Caller must be holding exactly one
- * reference to clp->cl_sem and owner->so_sema!
+ * reference to clp->cl_sem!
*/
void nfs4_put_open_state(struct nfs4_state *state)
{
struct inode *inode = state->inode;
struct nfs4_state_owner *owner = state->owner;
- if (!atomic_dec_and_lock(&state->count, &inode->i_lock))
+ if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
return;
+ spin_lock(&inode->i_lock);
if (!list_empty(&state->inode_states))
list_del(&state->inode_states);
- spin_unlock(&inode->i_lock);
list_del(&state->open_states);
+ spin_unlock(&inode->i_lock);
+ spin_unlock(&owner->so_lock);
iput(inode);
BUG_ON (state->state != 0);
nfs4_free_open_state(state);
@@ -481,20 +487,17 @@
}
/*
- * Beware! Caller must be holding no references to clp->cl_sem!
- * of owner->so_sema!
+ * Close the current file.
*/
void nfs4_close_state(struct nfs4_state *state, mode_t mode)
{
struct inode *inode = state->inode;
struct nfs4_state_owner *owner = state->owner;
- struct nfs4_client *clp = owner->so_client;
int newstate;
atomic_inc(&owner->so_count);
- down_read(&clp->cl_sem);
- down(&owner->so_sema);
/* Protect against nfs4_find_state() */
+ spin_lock(&owner->so_lock);
spin_lock(&inode->i_lock);
if (mode & FMODE_READ)
state->nreaders--;
@@ -507,6 +510,7 @@
list_move_tail(&state->open_states, &owner->so_states);
}
spin_unlock(&inode->i_lock);
+ spin_unlock(&owner->so_lock);
newstate = 0;
if (state->state != 0) {
if (state->nreaders)
@@ -515,14 +519,16 @@
newstate |= FMODE_WRITE;
if (state->state == newstate)
goto out;
- if (nfs4_do_close(inode, state, newstate) == -EINPROGRESS)
+ if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
+ state->state = newstate;
+ goto out;
+ }
+ if (nfs4_do_close(inode, state, newstate) == 0)
return;
}
out:
nfs4_put_open_state(state);
- up(&owner->so_sema);
nfs4_put_state_owner(owner);
- up_read(&clp->cl_sem);
}
/*
@@ -546,19 +552,16 @@
* Return a compatible lock_state. If no initialized lock_state structure
* exists, return an uninitialized one.
*
- * The caller must be holding state->lock_sema
*/
static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
{
struct nfs4_lock_state *lsp;
struct nfs4_client *clp = state->owner->so_client;
- lsp = kmalloc(sizeof(*lsp), GFP_KERNEL);
+ lsp = kzalloc(sizeof(*lsp), GFP_KERNEL);
if (lsp == NULL)
return NULL;
- lsp->ls_flags = 0;
- lsp->ls_seqid = 0; /* arbitrary */
- memset(lsp->ls_stateid.data, 0, sizeof(lsp->ls_stateid.data));
+ lsp->ls_seqid.sequence = &state->owner->so_sequence;
atomic_set(&lsp->ls_count, 1);
lsp->ls_owner = fl_owner;
spin_lock(&clp->cl_lock);
@@ -572,7 +575,7 @@
* Return a compatible lock_state. If no initialized lock_state structure
* exists, return an uninitialized one.
*
- * The caller must be holding state->lock_sema and clp->cl_sem
+ * The caller must be holding clp->cl_sem
*/
static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner)
{
@@ -605,7 +608,7 @@
* Release reference to lock_state, and free it if we see that
* it is no longer in use
*/
-static void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
+void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
{
struct nfs4_state *state;
@@ -673,29 +676,94 @@
nfs4_put_lock_state(lsp);
}
-/*
-* Called with state->lock_sema and clp->cl_sem held.
-*/
-void nfs4_increment_lock_seqid(int status, struct nfs4_lock_state *lsp)
+struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter)
{
- if (status == NFS_OK || seqid_mutating_err(-status))
- lsp->ls_seqid++;
+ struct nfs_seqid *new;
+
+ new = kmalloc(sizeof(*new), GFP_KERNEL);
+ if (new != NULL) {
+ new->sequence = counter;
+ INIT_LIST_HEAD(&new->list);
+ }
+ return new;
+}
+
+void nfs_free_seqid(struct nfs_seqid *seqid)
+{
+ struct rpc_sequence *sequence = seqid->sequence->sequence;
+
+ if (!list_empty(&seqid->list)) {
+ spin_lock(&sequence->lock);
+ list_del(&seqid->list);
+ spin_unlock(&sequence->lock);
+ }
+ rpc_wake_up_next(&sequence->wait);
+ kfree(seqid);
}
/*
-* Called with sp->so_sema and clp->cl_sem held.
-*
-* Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
-* failed with a seqid incrementing error -
-* see comments nfs_fs.h:seqid_mutating_error()
-*/
-void nfs4_increment_seqid(int status, struct nfs4_state_owner *sp)
+ * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
+ * failed with a seqid incrementing error -
+ * see comments nfs_fs.h:seqid_mutating_error()
+ */
+static inline void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
{
- if (status == NFS_OK || seqid_mutating_err(-status))
- sp->so_seqid++;
- /* If the server returns BAD_SEQID, unhash state_owner here */
- if (status == -NFS4ERR_BAD_SEQID)
+ switch (status) {
+ case 0:
+ break;
+ case -NFS4ERR_BAD_SEQID:
+ case -NFS4ERR_STALE_CLIENTID:
+ case -NFS4ERR_STALE_STATEID:
+ case -NFS4ERR_BAD_STATEID:
+ case -NFS4ERR_BADXDR:
+ case -NFS4ERR_RESOURCE:
+ case -NFS4ERR_NOFILEHANDLE:
+ /* Non-seqid mutating errors */
+ return;
+ };
+ /*
+ * Note: no locking needed as we are guaranteed to be first
+ * on the sequence list
+ */
+ seqid->sequence->counter++;
+}
+
+void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid)
+{
+ if (status == -NFS4ERR_BAD_SEQID) {
+ struct nfs4_state_owner *sp = container_of(seqid->sequence,
+ struct nfs4_state_owner, so_seqid);
nfs4_drop_state_owner(sp);
+ }
+ return nfs_increment_seqid(status, seqid);
+}
+
+/*
+ * Increment the seqid if the LOCK/LOCKU succeeded, or
+ * failed with a seqid incrementing error -
+ * see comments nfs_fs.h:seqid_mutating_error()
+ */
+void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid)
+{
+ return nfs_increment_seqid(status, seqid);
+}
+
+int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
+{
+ struct rpc_sequence *sequence = seqid->sequence->sequence;
+ int status = 0;
+
+ if (sequence->list.next == &seqid->list)
+ goto out;
+ spin_lock(&sequence->lock);
+ if (!list_empty(&sequence->list)) {
+ rpc_sleep_on(&sequence->wait, task, NULL, NULL);
+ status = -EAGAIN;
+ } else
+ list_add(&seqid->list, &sequence->list);
+ spin_unlock(&sequence->lock);
+out:
+ return status;
}
static int reclaimer(void *);
@@ -791,8 +859,6 @@
if (state->state == 0)
continue;
status = ops->recover_open(sp, state);
- list_for_each_entry(lock, &state->lock_states, ls_locks)
- lock->ls_flags &= ~NFS_LOCK_INITIALIZED;
if (status >= 0) {
status = nfs4_reclaim_locks(ops, state);
if (status < 0)
@@ -831,6 +897,28 @@
return status;
}
+static void nfs4_state_mark_reclaim(struct nfs4_client *clp)
+{
+ struct nfs4_state_owner *sp;
+ struct nfs4_state *state;
+ struct nfs4_lock_state *lock;
+
+ /* Reset all sequence ids to zero */
+ list_for_each_entry(sp, &clp->cl_state_owners, so_list) {
+ sp->so_seqid.counter = 0;
+ sp->so_seqid.flags = 0;
+ spin_lock(&sp->so_lock);
+ list_for_each_entry(state, &sp->so_states, open_states) {
+ list_for_each_entry(lock, &state->lock_states, ls_locks) {
+ lock->ls_seqid.counter = 0;
+ lock->ls_seqid.flags = 0;
+ lock->ls_flags &= ~NFS_LOCK_INITIALIZED;
+ }
+ }
+ spin_unlock(&sp->so_lock);
+ }
+}
+
static int reclaimer(void *ptr)
{
struct reclaimer_args *args = (struct reclaimer_args *)ptr;
@@ -864,6 +952,7 @@
default:
ops = &nfs4_network_partition_recovery_ops;
};
+ nfs4_state_mark_reclaim(clp);
status = __nfs4_init_client(clp);
if (status)
goto out_error;
diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c
index 6c564ef..cd76264 100644
--- a/fs/nfs/nfs4xdr.c
+++ b/fs/nfs/nfs4xdr.c
@@ -602,10 +602,10 @@
{
uint32_t *p;
- RESERVE_SPACE(8+sizeof(arg->stateid.data));
+ RESERVE_SPACE(8+sizeof(arg->stateid->data));
WRITE32(OP_CLOSE);
- WRITE32(arg->seqid);
- WRITEMEM(arg->stateid.data, sizeof(arg->stateid.data));
+ WRITE32(arg->seqid->sequence->counter);
+ WRITEMEM(arg->stateid->data, sizeof(arg->stateid->data));
return 0;
}
@@ -729,22 +729,18 @@
WRITE64(arg->length);
WRITE32(opargs->new_lock_owner);
if (opargs->new_lock_owner){
- struct nfs_open_to_lock *ol = opargs->u.open_lock;
-
RESERVE_SPACE(40);
- WRITE32(ol->open_seqid);
- WRITEMEM(&ol->open_stateid, sizeof(ol->open_stateid));
- WRITE32(ol->lock_seqid);
- WRITE64(ol->lock_owner.clientid);
+ WRITE32(opargs->open_seqid->sequence->counter);
+ WRITEMEM(opargs->open_stateid->data, sizeof(opargs->open_stateid->data));
+ WRITE32(opargs->lock_seqid->sequence->counter);
+ WRITE64(opargs->lock_owner.clientid);
WRITE32(4);
- WRITE32(ol->lock_owner.id);
+ WRITE32(opargs->lock_owner.id);
}
else {
- struct nfs_exist_lock *el = opargs->u.exist_lock;
-
RESERVE_SPACE(20);
- WRITEMEM(&el->stateid, sizeof(el->stateid));
- WRITE32(el->seqid);
+ WRITEMEM(opargs->lock_stateid->data, sizeof(opargs->lock_stateid->data));
+ WRITE32(opargs->lock_seqid->sequence->counter);
}
return 0;
@@ -775,8 +771,8 @@
RESERVE_SPACE(44);
WRITE32(OP_LOCKU);
WRITE32(arg->type);
- WRITE32(opargs->seqid);
- WRITEMEM(&opargs->stateid, sizeof(opargs->stateid));
+ WRITE32(opargs->seqid->sequence->counter);
+ WRITEMEM(opargs->stateid->data, sizeof(opargs->stateid->data));
WRITE64(arg->offset);
WRITE64(arg->length);
@@ -826,7 +822,7 @@
*/
RESERVE_SPACE(8);
WRITE32(OP_OPEN);
- WRITE32(arg->seqid);
+ WRITE32(arg->seqid->sequence->counter);
encode_share_access(xdr, arg->open_flags);
RESERVE_SPACE(16);
WRITE64(arg->clientid);
@@ -941,7 +937,7 @@
RESERVE_SPACE(8+sizeof(arg->stateid.data));
WRITE32(OP_OPEN_CONFIRM);
WRITEMEM(arg->stateid.data, sizeof(arg->stateid.data));
- WRITE32(arg->seqid);
+ WRITE32(arg->seqid->sequence->counter);
return 0;
}
@@ -950,10 +946,10 @@
{
uint32_t *p;
- RESERVE_SPACE(8+sizeof(arg->stateid.data));
+ RESERVE_SPACE(8+sizeof(arg->stateid->data));
WRITE32(OP_OPEN_DOWNGRADE);
- WRITEMEM(arg->stateid.data, sizeof(arg->stateid.data));
- WRITE32(arg->seqid);
+ WRITEMEM(arg->stateid->data, sizeof(arg->stateid->data));
+ WRITE32(arg->seqid->sequence->counter);
encode_share_access(xdr, arg->open_flags);
return 0;
}
@@ -1437,6 +1433,9 @@
};
int status;
+ status = nfs_wait_on_sequence(args->seqid, req->rq_task);
+ if (status != 0)
+ goto out;
xdr_init_encode(&xdr, &req->rq_snd_buf, p);
encode_compound_hdr(&xdr, &hdr);
status = encode_putfh(&xdr, args->fh);
@@ -1464,6 +1463,9 @@
};
int status;
+ status = nfs_wait_on_sequence(args->seqid, req->rq_task);
+ if (status != 0)
+ goto out;
xdr_init_encode(&xdr, &req->rq_snd_buf, p);
encode_compound_hdr(&xdr, &hdr);
status = encode_putfh(&xdr, args->fh);
@@ -1485,6 +1487,9 @@
};
int status;
+ status = nfs_wait_on_sequence(args->seqid, req->rq_task);
+ if (status != 0)
+ goto out;
xdr_init_encode(&xdr, &req->rq_snd_buf, p);
encode_compound_hdr(&xdr, &hdr);
status = encode_putfh(&xdr, args->fh);
@@ -1525,8 +1530,15 @@
struct compound_hdr hdr = {
.nops = 2,
};
+ struct nfs_lock_opargs *opargs = args->u.lock;
int status;
+ status = nfs_wait_on_sequence(opargs->lock_seqid, req->rq_task);
+ if (status != 0)
+ goto out;
+ /* Do we need to do an open_to_lock_owner? */
+ if (opargs->lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)
+ opargs->new_lock_owner = 0;
xdr_init_encode(&xdr, &req->rq_snd_buf, p);
encode_compound_hdr(&xdr, &hdr);
status = encode_putfh(&xdr, args->fh);
@@ -2890,8 +2902,8 @@
status = decode_op_hdr(xdr, OP_LOCK);
if (status == 0) {
- READ_BUF(sizeof(nfs4_stateid));
- COPYMEM(&res->u.stateid, sizeof(res->u.stateid));
+ READ_BUF(sizeof(res->u.stateid.data));
+ COPYMEM(res->u.stateid.data, sizeof(res->u.stateid.data));
} else if (status == -NFS4ERR_DENIED)
return decode_lock_denied(xdr, &res->u.denied);
return status;
@@ -2913,8 +2925,8 @@
status = decode_op_hdr(xdr, OP_LOCKU);
if (status == 0) {
- READ_BUF(sizeof(nfs4_stateid));
- COPYMEM(&res->u.stateid, sizeof(res->u.stateid));
+ READ_BUF(sizeof(res->u.stateid.data));
+ COPYMEM(res->u.stateid.data, sizeof(res->u.stateid.data));
}
return status;
}
@@ -3243,7 +3255,8 @@
if (attrlen <= *acl_len)
xdr_read_pages(xdr, attrlen);
*acl_len = attrlen;
- }
+ } else
+ status = -EOPNOTSUPP;
out:
return status;
diff --git a/fs/nfs/proc.c b/fs/nfs/proc.c
index be23c3f..8fef865 100644
--- a/fs/nfs/proc.c
+++ b/fs/nfs/proc.c
@@ -216,7 +216,7 @@
static int
nfs_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
- int flags)
+ int flags, struct nameidata *nd)
{
struct nfs_fh fhandle;
struct nfs_fattr fattr;
diff --git a/fs/open.c b/fs/open.c
index f0d90cf..8d06ec9 100644
--- a/fs/open.c
+++ b/fs/open.c
@@ -739,7 +739,8 @@
}
static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
- int flags, struct file *f)
+ int flags, struct file *f,
+ int (*open)(struct inode *, struct file *))
{
struct inode *inode;
int error;
@@ -761,11 +762,14 @@
f->f_op = fops_get(inode->i_fop);
file_move(f, &inode->i_sb->s_files);
- if (f->f_op && f->f_op->open) {
- error = f->f_op->open(inode,f);
+ if (!open && f->f_op)
+ open = f->f_op->open;
+ if (open) {
+ error = open(inode, f);
if (error)
goto cleanup_all;
}
+
f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
@@ -814,28 +818,75 @@
{
int namei_flags, error;
struct nameidata nd;
- struct file *f;
namei_flags = flags;
if ((namei_flags+1) & O_ACCMODE)
namei_flags++;
- if (namei_flags & O_TRUNC)
- namei_flags |= 2;
-
- error = -ENFILE;
- f = get_empty_filp();
- if (f == NULL)
- return ERR_PTR(error);
error = open_namei(filename, namei_flags, mode, &nd);
if (!error)
- return __dentry_open(nd.dentry, nd.mnt, flags, f);
+ return nameidata_to_filp(&nd, flags);
- put_filp(f);
return ERR_PTR(error);
}
EXPORT_SYMBOL(filp_open);
+/**
+ * lookup_instantiate_filp - instantiates the open intent filp
+ * @nd: pointer to nameidata
+ * @dentry: pointer to dentry
+ * @open: open callback
+ *
+ * Helper for filesystems that want to use lookup open intents and pass back
+ * a fully instantiated struct file to the caller.
+ * This function is meant to be called from within a filesystem's
+ * lookup method.
+ * Note that in case of error, nd->intent.open.file is destroyed, but the
+ * path information remains valid.
+ * If the open callback is set to NULL, then the standard f_op->open()
+ * filesystem callback is substituted.
+ */
+struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
+ int (*open)(struct inode *, struct file *))
+{
+ if (IS_ERR(nd->intent.open.file))
+ goto out;
+ if (IS_ERR(dentry))
+ goto out_err;
+ nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->mnt),
+ nd->intent.open.flags - 1,
+ nd->intent.open.file,
+ open);
+out:
+ return nd->intent.open.file;
+out_err:
+ release_open_intent(nd);
+ nd->intent.open.file = (struct file *)dentry;
+ goto out;
+}
+EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
+
+/**
+ * nameidata_to_filp - convert a nameidata to an open filp.
+ * @nd: pointer to nameidata
+ * @flags: open flags
+ *
+ * Note that this function destroys the original nameidata
+ */
+struct file *nameidata_to_filp(struct nameidata *nd, int flags)
+{
+ struct file *filp;
+
+ /* Pick up the filp from the open intent */
+ filp = nd->intent.open.file;
+ /* Has the filesystem initialised the file for us? */
+ if (filp->f_dentry == NULL)
+ filp = __dentry_open(nd->dentry, nd->mnt, flags, filp, NULL);
+ else
+ path_release(nd);
+ return filp;
+}
+
struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags)
{
int error;
@@ -846,7 +897,7 @@
if (f == NULL)
return ERR_PTR(error);
- return __dentry_open(dentry, mnt, flags, f);
+ return __dentry_open(dentry, mnt, flags, f, NULL);
}
EXPORT_SYMBOL(dentry_open);
diff --git a/include/linux/namei.h b/include/linux/namei.h
index 7db67b0..1c975d0 100644
--- a/include/linux/namei.h
+++ b/include/linux/namei.h
@@ -8,6 +8,7 @@
struct open_intent {
int flags;
int create_mode;
+ struct file *file;
};
enum { MAX_NESTED_LINKS = 5 };
@@ -65,6 +66,13 @@
extern void path_release(struct nameidata *);
extern void path_release_on_umount(struct nameidata *);
+extern int __user_path_lookup_open(const char __user *, unsigned lookup_flags, struct nameidata *nd, int open_flags);
+extern int path_lookup_open(const char *, unsigned lookup_flags, struct nameidata *, int open_flags);
+extern struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
+ int (*open)(struct inode *, struct file *));
+extern struct file *nameidata_to_filp(struct nameidata *nd, int flags);
+extern void release_open_intent(struct nameidata *);
+
extern struct dentry * lookup_one_len(const char *, struct dentry *, int);
extern struct dentry * lookup_hash(struct qstr *, struct dentry *);
diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h
index 9a6047f..7bac278 100644
--- a/include/linux/nfs_fs.h
+++ b/include/linux/nfs_fs.h
@@ -41,6 +41,10 @@
#define NFS_MAX_FILE_IO_BUFFER_SIZE 32768
#define NFS_DEF_FILE_IO_BUFFER_SIZE 4096
+/* Default timeout values */
+#define NFS_MAX_UDP_TIMEOUT (60*HZ)
+#define NFS_MAX_TCP_TIMEOUT (600*HZ)
+
/*
* superblock magic number for NFS
*/
diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h
index a2bf691..60086da 100644
--- a/include/linux/nfs_xdr.h
+++ b/include/linux/nfs_xdr.h
@@ -96,12 +96,13 @@
u64 after;
};
+struct nfs_seqid;
/*
* Arguments to the open call.
*/
struct nfs_openargs {
const struct nfs_fh * fh;
- __u32 seqid;
+ struct nfs_seqid * seqid;
int open_flags;
__u64 clientid;
__u32 id;
@@ -136,7 +137,7 @@
struct nfs_open_confirmargs {
const struct nfs_fh * fh;
nfs4_stateid stateid;
- __u32 seqid;
+ struct nfs_seqid * seqid;
};
struct nfs_open_confirmres {
@@ -148,8 +149,8 @@
*/
struct nfs_closeargs {
struct nfs_fh * fh;
- nfs4_stateid stateid;
- __u32 seqid;
+ nfs4_stateid * stateid;
+ struct nfs_seqid * seqid;
int open_flags;
};
@@ -164,30 +165,19 @@
u32 id;
};
-struct nfs_open_to_lock {
- __u32 open_seqid;
- nfs4_stateid open_stateid;
- __u32 lock_seqid;
- struct nfs_lowner lock_owner;
-};
-
-struct nfs_exist_lock {
- nfs4_stateid stateid;
- __u32 seqid;
-};
-
struct nfs_lock_opargs {
+ struct nfs_seqid * lock_seqid;
+ nfs4_stateid * lock_stateid;
+ struct nfs_seqid * open_seqid;
+ nfs4_stateid * open_stateid;
+ struct nfs_lowner lock_owner;
__u32 reclaim;
__u32 new_lock_owner;
- union {
- struct nfs_open_to_lock *open_lock;
- struct nfs_exist_lock *exist_lock;
- } u;
};
struct nfs_locku_opargs {
- __u32 seqid;
- nfs4_stateid stateid;
+ struct nfs_seqid * seqid;
+ nfs4_stateid * stateid;
};
struct nfs_lockargs {
@@ -722,7 +712,7 @@
int (*write) (struct nfs_write_data *);
int (*commit) (struct nfs_write_data *);
int (*create) (struct inode *, struct dentry *,
- struct iattr *, int);
+ struct iattr *, int, struct nameidata *);
int (*remove) (struct inode *, struct qstr *);
int (*unlink_setup) (struct rpc_message *,
struct dentry *, struct qstr *);
diff --git a/include/linux/sunrpc/auth.h b/include/linux/sunrpc/auth.h
index 04ebc24..b68c11a 100644
--- a/include/linux/sunrpc/auth.h
+++ b/include/linux/sunrpc/auth.h
@@ -66,7 +66,12 @@
struct rpc_auth {
unsigned int au_cslack; /* call cred size estimate */
- unsigned int au_rslack; /* reply verf size guess */
+ /* guess at number of u32's auth adds before
+ * reply data; normally the verifier size: */
+ unsigned int au_rslack;
+ /* for gss, used to calculate au_rslack: */
+ unsigned int au_verfsize;
+
unsigned int au_flags; /* various flags */
struct rpc_authops * au_ops; /* operations */
rpc_authflavor_t au_flavor; /* pseudoflavor (note may
diff --git a/include/linux/sunrpc/debug.h b/include/linux/sunrpc/debug.h
index eadb31e..1a42d90 100644
--- a/include/linux/sunrpc/debug.h
+++ b/include/linux/sunrpc/debug.h
@@ -32,6 +32,7 @@
#define RPCDBG_AUTH 0x0010
#define RPCDBG_PMAP 0x0020
#define RPCDBG_SCHED 0x0040
+#define RPCDBG_TRANS 0x0080
#define RPCDBG_SVCSOCK 0x0100
#define RPCDBG_SVCDSP 0x0200
#define RPCDBG_MISC 0x0400
@@ -94,6 +95,8 @@
CTL_NLMDEBUG,
CTL_SLOTTABLE_UDP,
CTL_SLOTTABLE_TCP,
+ CTL_MIN_RESVPORT,
+ CTL_MAX_RESVPORT,
};
#endif /* _LINUX_SUNRPC_DEBUG_H_ */
diff --git a/include/linux/sunrpc/gss_api.h b/include/linux/sunrpc/gss_api.h
index 689262f..9b8bcf1 100644
--- a/include/linux/sunrpc/gss_api.h
+++ b/include/linux/sunrpc/gss_api.h
@@ -40,14 +40,21 @@
struct gss_ctx **ctx_id);
u32 gss_get_mic(
struct gss_ctx *ctx_id,
- u32 qop,
struct xdr_buf *message,
struct xdr_netobj *mic_token);
u32 gss_verify_mic(
struct gss_ctx *ctx_id,
struct xdr_buf *message,
- struct xdr_netobj *mic_token,
- u32 *qstate);
+ struct xdr_netobj *mic_token);
+u32 gss_wrap(
+ struct gss_ctx *ctx_id,
+ int offset,
+ struct xdr_buf *outbuf,
+ struct page **inpages);
+u32 gss_unwrap(
+ struct gss_ctx *ctx_id,
+ int offset,
+ struct xdr_buf *inbuf);
u32 gss_delete_sec_context(
struct gss_ctx **ctx_id);
@@ -56,7 +63,6 @@
struct pf_desc {
u32 pseudoflavor;
- u32 qop;
u32 service;
char *name;
char *auth_domain_name;
@@ -85,14 +91,21 @@
struct gss_ctx *ctx_id);
u32 (*gss_get_mic)(
struct gss_ctx *ctx_id,
- u32 qop,
struct xdr_buf *message,
struct xdr_netobj *mic_token);
u32 (*gss_verify_mic)(
struct gss_ctx *ctx_id,
struct xdr_buf *message,
- struct xdr_netobj *mic_token,
- u32 *qstate);
+ struct xdr_netobj *mic_token);
+ u32 (*gss_wrap)(
+ struct gss_ctx *ctx_id,
+ int offset,
+ struct xdr_buf *outbuf,
+ struct page **inpages);
+ u32 (*gss_unwrap)(
+ struct gss_ctx *ctx_id,
+ int offset,
+ struct xdr_buf *buf);
void (*gss_delete_sec_context)(
void *internal_ctx_id);
};
diff --git a/include/linux/sunrpc/gss_err.h b/include/linux/sunrpc/gss_err.h
index 92608a2..a680786 100644
--- a/include/linux/sunrpc/gss_err.h
+++ b/include/linux/sunrpc/gss_err.h
@@ -66,16 +66,6 @@
/*
- * Define the default Quality of Protection for per-message services. Note
- * that an implementation that offers multiple levels of QOP may either reserve
- * a value (for example zero, as assumed here) to mean "default protection", or
- * alternatively may simply equate GSS_C_QOP_DEFAULT to a specific explicit
- * QOP value. However a value of 0 should always be interpreted by a GSSAPI
- * implementation as a request for the default protection level.
- */
-#define GSS_C_QOP_DEFAULT 0
-
-/*
* Expiration time of 2^32-1 seconds means infinite lifetime for a
* credential or security context
*/
diff --git a/include/linux/sunrpc/gss_krb5.h b/include/linux/sunrpc/gss_krb5.h
index ffe31d2..2c3601d 100644
--- a/include/linux/sunrpc/gss_krb5.h
+++ b/include/linux/sunrpc/gss_krb5.h
@@ -116,18 +116,22 @@
s32
make_checksum(s32 cksumtype, char *header, int hdrlen, struct xdr_buf *body,
- struct xdr_netobj *cksum);
+ int body_offset, struct xdr_netobj *cksum);
+
+u32 gss_get_mic_kerberos(struct gss_ctx *, struct xdr_buf *,
+ struct xdr_netobj *);
+
+u32 gss_verify_mic_kerberos(struct gss_ctx *, struct xdr_buf *,
+ struct xdr_netobj *);
u32
-krb5_make_token(struct krb5_ctx *context_handle, int qop_req,
- struct xdr_buf *input_message_buffer,
- struct xdr_netobj *output_message_buffer, int toktype);
+gss_wrap_kerberos(struct gss_ctx *ctx_id, int offset,
+ struct xdr_buf *outbuf, struct page **pages);
u32
-krb5_read_token(struct krb5_ctx *context_handle,
- struct xdr_netobj *input_token_buffer,
- struct xdr_buf *message_buffer,
- int *qop_state, int toktype);
+gss_unwrap_kerberos(struct gss_ctx *ctx_id, int offset,
+ struct xdr_buf *buf);
+
u32
krb5_encrypt(struct crypto_tfm * key,
@@ -137,6 +141,13 @@
krb5_decrypt(struct crypto_tfm * key,
void *iv, void *in, void *out, int length);
+int
+gss_encrypt_xdr_buf(struct crypto_tfm *tfm, struct xdr_buf *outbuf, int offset,
+ struct page **pages);
+
+int
+gss_decrypt_xdr_buf(struct crypto_tfm *tfm, struct xdr_buf *inbuf, int offset);
+
s32
krb5_make_seq_num(struct crypto_tfm * key,
int direction,
diff --git a/include/linux/sunrpc/gss_spkm3.h b/include/linux/sunrpc/gss_spkm3.h
index b5c9968..0beb2cf 100644
--- a/include/linux/sunrpc/gss_spkm3.h
+++ b/include/linux/sunrpc/gss_spkm3.h
@@ -41,9 +41,9 @@
#define SPKM_WRAP_TOK 5
#define SPKM_DEL_TOK 6
-u32 spkm3_make_token(struct spkm3_ctx *ctx, int qop_req, struct xdr_buf * text, struct xdr_netobj * token, int toktype);
+u32 spkm3_make_token(struct spkm3_ctx *ctx, struct xdr_buf * text, struct xdr_netobj * token, int toktype);
-u32 spkm3_read_token(struct spkm3_ctx *ctx, struct xdr_netobj *read_token, struct xdr_buf *message_buffer, int *qop_state, int toktype);
+u32 spkm3_read_token(struct spkm3_ctx *ctx, struct xdr_netobj *read_token, struct xdr_buf *message_buffer, int toktype);
#define CKSUMTYPE_RSA_MD5 0x0007
diff --git a/include/linux/sunrpc/msg_prot.h b/include/linux/sunrpc/msg_prot.h
index 15f1153..f43f237 100644
--- a/include/linux/sunrpc/msg_prot.h
+++ b/include/linux/sunrpc/msg_prot.h
@@ -76,5 +76,30 @@
#define RPC_MAXNETNAMELEN 256
+/*
+ * From RFC 1831:
+ *
+ * "A record is composed of one or more record fragments. A record
+ * fragment is a four-byte header followed by 0 to (2**31) - 1 bytes of
+ * fragment data. The bytes encode an unsigned binary number; as with
+ * XDR integers, the byte order is from highest to lowest. The number
+ * encodes two values -- a boolean which indicates whether the fragment
+ * is the last fragment of the record (bit value 1 implies the fragment
+ * is the last fragment) and a 31-bit unsigned binary value which is the
+ * length in bytes of the fragment's data. The boolean value is the
+ * highest-order bit of the header; the length is the 31 low-order bits.
+ * (Note that this record specification is NOT in XDR standard form!)"
+ *
+ * The Linux RPC client always sends its requests in a single record
+ * fragment, limiting the maximum payload size for stream transports to
+ * 2GB.
+ */
+
+typedef u32 rpc_fraghdr;
+
+#define RPC_LAST_STREAM_FRAGMENT (1U << 31)
+#define RPC_FRAGMENT_SIZE_MASK (~RPC_LAST_STREAM_FRAGMENT)
+#define RPC_MAX_FRAGMENT_SIZE ((1U << 31) - 1)
+
#endif /* __KERNEL__ */
#endif /* _LINUX_SUNRPC_MSGPROT_H_ */
diff --git a/include/linux/sunrpc/xdr.h b/include/linux/sunrpc/xdr.h
index 23448d0..5da9687 100644
--- a/include/linux/sunrpc/xdr.h
+++ b/include/linux/sunrpc/xdr.h
@@ -161,14 +161,10 @@
typedef size_t (*skb_read_actor_t)(skb_reader_t *desc, void *to, size_t len);
+extern int csum_partial_copy_to_xdr(struct xdr_buf *, struct sk_buff *);
extern ssize_t xdr_partial_copy_from_skb(struct xdr_buf *, unsigned int,
skb_reader_t *, skb_read_actor_t);
-struct socket;
-struct sockaddr;
-extern int xdr_sendpages(struct socket *, struct sockaddr *, int,
- struct xdr_buf *, unsigned int, int);
-
extern int xdr_encode_word(struct xdr_buf *, int, u32);
extern int xdr_decode_word(struct xdr_buf *, int, u32 *);
diff --git a/include/linux/sunrpc/xprt.h b/include/linux/sunrpc/xprt.h
index e618c16..3b8b6e8 100644
--- a/include/linux/sunrpc/xprt.h
+++ b/include/linux/sunrpc/xprt.h
@@ -1,5 +1,5 @@
/*
- * linux/include/linux/sunrpc/clnt_xprt.h
+ * linux/include/linux/sunrpc/xprt.h
*
* Declarations for the RPC transport interface.
*
@@ -15,20 +15,6 @@
#include <linux/sunrpc/sched.h>
#include <linux/sunrpc/xdr.h>
-/*
- * The transport code maintains an estimate on the maximum number of out-
- * standing RPC requests, using a smoothed version of the congestion
- * avoidance implemented in 44BSD. This is basically the Van Jacobson
- * congestion algorithm: If a retransmit occurs, the congestion window is
- * halved; otherwise, it is incremented by 1/cwnd when
- *
- * - a reply is received and
- * - a full number of requests are outstanding and
- * - the congestion window hasn't been updated recently.
- *
- * Upper procedures may check whether a request would block waiting for
- * a free RPC slot by using the RPC_CONGESTED() macro.
- */
extern unsigned int xprt_udp_slot_table_entries;
extern unsigned int xprt_tcp_slot_table_entries;
@@ -36,36 +22,25 @@
#define RPC_DEF_SLOT_TABLE (16U)
#define RPC_MAX_SLOT_TABLE (128U)
-#define RPC_CWNDSHIFT (8U)
-#define RPC_CWNDSCALE (1U << RPC_CWNDSHIFT)
-#define RPC_INITCWND RPC_CWNDSCALE
-#define RPC_MAXCWND(xprt) ((xprt)->max_reqs << RPC_CWNDSHIFT)
-#define RPCXPRT_CONGESTED(xprt) ((xprt)->cong >= (xprt)->cwnd)
-
-/* Default timeout values */
-#define RPC_MAX_UDP_TIMEOUT (60*HZ)
-#define RPC_MAX_TCP_TIMEOUT (600*HZ)
-
/*
- * Wait duration for an RPC TCP connection to be established. Solaris
- * NFS over TCP uses 60 seconds, for example, which is in line with how
- * long a server takes to reboot.
- */
-#define RPC_CONNECT_TIMEOUT (60*HZ)
-
-/*
- * Delay an arbitrary number of seconds before attempting to reconnect
- * after an error.
- */
-#define RPC_REESTABLISH_TIMEOUT (15*HZ)
-
-/* RPC call and reply header size as number of 32bit words (verifier
+ * RPC call and reply header size as number of 32bit words (verifier
* size computed separately)
*/
#define RPC_CALLHDRSIZE 6
#define RPC_REPHDRSIZE 4
/*
+ * Parameters for choosing a free port
+ */
+extern unsigned int xprt_min_resvport;
+extern unsigned int xprt_max_resvport;
+
+#define RPC_MIN_RESVPORT (1U)
+#define RPC_MAX_RESVPORT (65535U)
+#define RPC_DEF_MIN_RESVPORT (650U)
+#define RPC_DEF_MAX_RESVPORT (1023U)
+
+/*
* This describes a timeout strategy
*/
struct rpc_timeout {
@@ -76,6 +51,9 @@
unsigned char to_exponential;
};
+struct rpc_task;
+struct rpc_xprt;
+
/*
* This describes a complete RPC request
*/
@@ -95,7 +73,10 @@
int rq_cong; /* has incremented xprt->cong */
int rq_received; /* receive completed */
u32 rq_seqno; /* gss seq no. used on req. */
-
+ int rq_enc_pages_num;
+ struct page **rq_enc_pages; /* scratch pages for use by
+ gss privacy code */
+ void (*rq_release_snd_buf)(struct rpc_rqst *); /* release rq_enc_pages */
struct list_head rq_list;
struct xdr_buf rq_private_buf; /* The receive buffer
@@ -121,12 +102,21 @@
#define rq_svec rq_snd_buf.head
#define rq_slen rq_snd_buf.len
-#define XPRT_LAST_FRAG (1 << 0)
-#define XPRT_COPY_RECM (1 << 1)
-#define XPRT_COPY_XID (1 << 2)
-#define XPRT_COPY_DATA (1 << 3)
+struct rpc_xprt_ops {
+ void (*set_buffer_size)(struct rpc_xprt *xprt, size_t sndsize, size_t rcvsize);
+ int (*reserve_xprt)(struct rpc_task *task);
+ void (*release_xprt)(struct rpc_xprt *xprt, struct rpc_task *task);
+ void (*connect)(struct rpc_task *task);
+ int (*send_request)(struct rpc_task *task);
+ void (*set_retrans_timeout)(struct rpc_task *task);
+ void (*timer)(struct rpc_task *task);
+ void (*release_request)(struct rpc_task *task);
+ void (*close)(struct rpc_xprt *xprt);
+ void (*destroy)(struct rpc_xprt *xprt);
+};
struct rpc_xprt {
+ struct rpc_xprt_ops * ops; /* transport methods */
struct socket * sock; /* BSD socket layer */
struct sock * inet; /* INET layer */
@@ -137,11 +127,13 @@
unsigned long cong; /* current congestion */
unsigned long cwnd; /* congestion window */
- unsigned int rcvsize, /* socket receive buffer size */
- sndsize; /* socket send buffer size */
+ size_t rcvsize, /* transport rcv buffer size */
+ sndsize; /* transport send buffer size */
size_t max_payload; /* largest RPC payload size,
in bytes */
+ unsigned int tsh_size; /* size of transport specific
+ header */
struct rpc_wait_queue sending; /* requests waiting to send */
struct rpc_wait_queue resend; /* requests waiting to resend */
@@ -150,11 +142,9 @@
struct list_head free; /* free slots */
struct rpc_rqst * slot; /* slot table storage */
unsigned int max_reqs; /* total slots */
- unsigned long sockstate; /* Socket state */
+ unsigned long state; /* transport state */
unsigned char shutdown : 1, /* being shut down */
- nocong : 1, /* no congestion control */
- resvport : 1, /* use a reserved port */
- stream : 1; /* TCP */
+ resvport : 1; /* use a reserved port */
/*
* XID
@@ -171,22 +161,27 @@
unsigned long tcp_copied, /* copied to request */
tcp_flags;
/*
- * Connection of sockets
+ * Connection of transports
*/
- struct work_struct sock_connect;
+ unsigned long connect_timeout,
+ bind_timeout,
+ reestablish_timeout;
+ struct work_struct connect_worker;
unsigned short port;
+
/*
- * Disconnection of idle sockets
+ * Disconnection of idle transports
*/
struct work_struct task_cleanup;
struct timer_list timer;
- unsigned long last_used;
+ unsigned long last_used,
+ idle_timeout;
/*
* Send stuff
*/
- spinlock_t sock_lock; /* lock socket info */
- spinlock_t xprt_lock; /* lock xprt info */
+ spinlock_t transport_lock; /* lock transport info */
+ spinlock_t reserve_lock; /* lock slot table */
struct rpc_task * snd_task; /* Task blocked in send */
struct list_head recv;
@@ -195,37 +190,111 @@
void (*old_data_ready)(struct sock *, int);
void (*old_state_change)(struct sock *);
void (*old_write_space)(struct sock *);
-
- wait_queue_head_t cong_wait;
};
+#define XPRT_LAST_FRAG (1 << 0)
+#define XPRT_COPY_RECM (1 << 1)
+#define XPRT_COPY_XID (1 << 2)
+#define XPRT_COPY_DATA (1 << 3)
+
#ifdef __KERNEL__
-struct rpc_xprt * xprt_create_proto(int proto, struct sockaddr_in *addr,
- struct rpc_timeout *toparms);
-int xprt_destroy(struct rpc_xprt *);
-void xprt_set_timeout(struct rpc_timeout *, unsigned int,
- unsigned long);
+/*
+ * Transport operations used by ULPs
+ */
+struct rpc_xprt * xprt_create_proto(int proto, struct sockaddr_in *addr, struct rpc_timeout *to);
+void xprt_set_timeout(struct rpc_timeout *to, unsigned int retr, unsigned long incr);
-void xprt_reserve(struct rpc_task *);
-int xprt_prepare_transmit(struct rpc_task *);
-void xprt_transmit(struct rpc_task *);
-void xprt_receive(struct rpc_task *);
+/*
+ * Generic internal transport functions
+ */
+void xprt_connect(struct rpc_task *task);
+void xprt_reserve(struct rpc_task *task);
+int xprt_reserve_xprt(struct rpc_task *task);
+int xprt_reserve_xprt_cong(struct rpc_task *task);
+int xprt_prepare_transmit(struct rpc_task *task);
+void xprt_transmit(struct rpc_task *task);
+void xprt_abort_transmit(struct rpc_task *task);
int xprt_adjust_timeout(struct rpc_rqst *req);
-void xprt_release(struct rpc_task *);
-void xprt_connect(struct rpc_task *);
-void xprt_sock_setbufsize(struct rpc_xprt *);
+void xprt_release_xprt(struct rpc_xprt *xprt, struct rpc_task *task);
+void xprt_release_xprt_cong(struct rpc_xprt *xprt, struct rpc_task *task);
+void xprt_release(struct rpc_task *task);
+int xprt_destroy(struct rpc_xprt *xprt);
-#define XPRT_LOCKED 0
-#define XPRT_CONNECT 1
-#define XPRT_CONNECTING 2
+static inline u32 *xprt_skip_transport_header(struct rpc_xprt *xprt, u32 *p)
+{
+ return p + xprt->tsh_size;
+}
-#define xprt_connected(xp) (test_bit(XPRT_CONNECT, &(xp)->sockstate))
-#define xprt_set_connected(xp) (set_bit(XPRT_CONNECT, &(xp)->sockstate))
-#define xprt_test_and_set_connected(xp) (test_and_set_bit(XPRT_CONNECT, &(xp)->sockstate))
-#define xprt_test_and_clear_connected(xp) \
- (test_and_clear_bit(XPRT_CONNECT, &(xp)->sockstate))
-#define xprt_clear_connected(xp) (clear_bit(XPRT_CONNECT, &(xp)->sockstate))
+/*
+ * Transport switch helper functions
+ */
+void xprt_set_retrans_timeout_def(struct rpc_task *task);
+void xprt_set_retrans_timeout_rtt(struct rpc_task *task);
+void xprt_wake_pending_tasks(struct rpc_xprt *xprt, int status);
+void xprt_wait_for_buffer_space(struct rpc_task *task);
+void xprt_write_space(struct rpc_xprt *xprt);
+void xprt_update_rtt(struct rpc_task *task);
+void xprt_adjust_cwnd(struct rpc_task *task, int result);
+struct rpc_rqst * xprt_lookup_rqst(struct rpc_xprt *xprt, u32 xid);
+void xprt_complete_rqst(struct rpc_task *task, int copied);
+void xprt_release_rqst_cong(struct rpc_task *task);
+void xprt_disconnect(struct rpc_xprt *xprt);
+
+/*
+ * Socket transport setup operations
+ */
+int xs_setup_udp(struct rpc_xprt *xprt, struct rpc_timeout *to);
+int xs_setup_tcp(struct rpc_xprt *xprt, struct rpc_timeout *to);
+
+/*
+ * Reserved bit positions in xprt->state
+ */
+#define XPRT_LOCKED (0)
+#define XPRT_CONNECTED (1)
+#define XPRT_CONNECTING (2)
+
+static inline void xprt_set_connected(struct rpc_xprt *xprt)
+{
+ set_bit(XPRT_CONNECTED, &xprt->state);
+}
+
+static inline void xprt_clear_connected(struct rpc_xprt *xprt)
+{
+ clear_bit(XPRT_CONNECTED, &xprt->state);
+}
+
+static inline int xprt_connected(struct rpc_xprt *xprt)
+{
+ return test_bit(XPRT_CONNECTED, &xprt->state);
+}
+
+static inline int xprt_test_and_set_connected(struct rpc_xprt *xprt)
+{
+ return test_and_set_bit(XPRT_CONNECTED, &xprt->state);
+}
+
+static inline int xprt_test_and_clear_connected(struct rpc_xprt *xprt)
+{
+ return test_and_clear_bit(XPRT_CONNECTED, &xprt->state);
+}
+
+static inline void xprt_clear_connecting(struct rpc_xprt *xprt)
+{
+ smp_mb__before_clear_bit();
+ clear_bit(XPRT_CONNECTING, &xprt->state);
+ smp_mb__after_clear_bit();
+}
+
+static inline int xprt_connecting(struct rpc_xprt *xprt)
+{
+ return test_bit(XPRT_CONNECTING, &xprt->state);
+}
+
+static inline int xprt_test_and_set_connecting(struct rpc_xprt *xprt)
+{
+ return test_and_set_bit(XPRT_CONNECTING, &xprt->state);
+}
#endif /* __KERNEL__*/
diff --git a/net/sunrpc/Makefile b/net/sunrpc/Makefile
index 46a2ce0..cdcab9c 100644
--- a/net/sunrpc/Makefile
+++ b/net/sunrpc/Makefile
@@ -6,7 +6,7 @@
obj-$(CONFIG_SUNRPC) += sunrpc.o
obj-$(CONFIG_SUNRPC_GSS) += auth_gss/
-sunrpc-y := clnt.o xprt.o sched.o \
+sunrpc-y := clnt.o xprt.o socklib.o xprtsock.o sched.o \
auth.o auth_null.o auth_unix.o \
svc.o svcsock.o svcauth.o svcauth_unix.o \
pmap_clnt.o timer.o xdr.o \
diff --git a/net/sunrpc/auth.c b/net/sunrpc/auth.c
index 505e2d4..a415d99 100644
--- a/net/sunrpc/auth.c
+++ b/net/sunrpc/auth.c
@@ -11,7 +11,6 @@
#include <linux/module.h>
#include <linux/slab.h>
#include <linux/errno.h>
-#include <linux/socket.h>
#include <linux/sunrpc/clnt.h>
#include <linux/spinlock.h>
diff --git a/net/sunrpc/auth_gss/Makefile b/net/sunrpc/auth_gss/Makefile
index fe1b874..f3431a7 100644
--- a/net/sunrpc/auth_gss/Makefile
+++ b/net/sunrpc/auth_gss/Makefile
@@ -10,7 +10,7 @@
obj-$(CONFIG_RPCSEC_GSS_KRB5) += rpcsec_gss_krb5.o
rpcsec_gss_krb5-objs := gss_krb5_mech.o gss_krb5_seal.o gss_krb5_unseal.o \
- gss_krb5_seqnum.o
+ gss_krb5_seqnum.o gss_krb5_wrap.o
obj-$(CONFIG_RPCSEC_GSS_SPKM3) += rpcsec_gss_spkm3.o
diff --git a/net/sunrpc/auth_gss/auth_gss.c b/net/sunrpc/auth_gss/auth_gss.c
index 2f7b867..f44f46f 100644
--- a/net/sunrpc/auth_gss/auth_gss.c
+++ b/net/sunrpc/auth_gss/auth_gss.c
@@ -42,9 +42,8 @@
#include <linux/init.h>
#include <linux/types.h>
#include <linux/slab.h>
-#include <linux/socket.h>
-#include <linux/in.h>
#include <linux/sched.h>
+#include <linux/pagemap.h>
#include <linux/sunrpc/clnt.h>
#include <linux/sunrpc/auth.h>
#include <linux/sunrpc/auth_gss.h>
@@ -846,10 +845,8 @@
/* We compute the checksum for the verifier over the xdr-encoded bytes
* starting with the xid and ending at the end of the credential: */
- iov.iov_base = req->rq_snd_buf.head[0].iov_base;
- if (task->tk_client->cl_xprt->stream)
- /* See clnt.c:call_header() */
- iov.iov_base += 4;
+ iov.iov_base = xprt_skip_transport_header(task->tk_xprt,
+ req->rq_snd_buf.head[0].iov_base);
iov.iov_len = (u8 *)p - (u8 *)iov.iov_base;
xdr_buf_from_iov(&iov, &verf_buf);
@@ -857,9 +854,7 @@
*p++ = htonl(RPC_AUTH_GSS);
mic.data = (u8 *)(p + 1);
- maj_stat = gss_get_mic(ctx->gc_gss_ctx,
- GSS_C_QOP_DEFAULT,
- &verf_buf, &mic);
+ maj_stat = gss_get_mic(ctx->gc_gss_ctx, &verf_buf, &mic);
if (maj_stat == GSS_S_CONTEXT_EXPIRED) {
cred->cr_flags &= ~RPCAUTH_CRED_UPTODATE;
} else if (maj_stat != 0) {
@@ -890,10 +885,8 @@
gss_validate(struct rpc_task *task, u32 *p)
{
struct rpc_cred *cred = task->tk_msg.rpc_cred;
- struct gss_cred *gss_cred = container_of(cred, struct gss_cred,
- gc_base);
struct gss_cl_ctx *ctx = gss_cred_get_ctx(cred);
- u32 seq, qop_state;
+ u32 seq;
struct kvec iov;
struct xdr_buf verf_buf;
struct xdr_netobj mic;
@@ -914,23 +907,14 @@
mic.data = (u8 *)p;
mic.len = len;
- maj_stat = gss_verify_mic(ctx->gc_gss_ctx, &verf_buf, &mic, &qop_state);
+ maj_stat = gss_verify_mic(ctx->gc_gss_ctx, &verf_buf, &mic);
if (maj_stat == GSS_S_CONTEXT_EXPIRED)
cred->cr_flags &= ~RPCAUTH_CRED_UPTODATE;
if (maj_stat)
goto out_bad;
- switch (gss_cred->gc_service) {
- case RPC_GSS_SVC_NONE:
- /* verifier data, flavor, length: */
- task->tk_auth->au_rslack = XDR_QUADLEN(len) + 2;
- break;
- case RPC_GSS_SVC_INTEGRITY:
- /* verifier data, flavor, length, length, sequence number: */
- task->tk_auth->au_rslack = XDR_QUADLEN(len) + 4;
- break;
- case RPC_GSS_SVC_PRIVACY:
- goto out_bad;
- }
+ /* We leave it to unwrap to calculate au_rslack. For now we just
+ * calculate the length of the verifier: */
+ task->tk_auth->au_verfsize = XDR_QUADLEN(len) + 2;
gss_put_ctx(ctx);
dprintk("RPC: %4u GSS gss_validate: gss_verify_mic succeeded.\n",
task->tk_pid);
@@ -975,8 +959,7 @@
p = iov->iov_base + iov->iov_len;
mic.data = (u8 *)(p + 1);
- maj_stat = gss_get_mic(ctx->gc_gss_ctx,
- GSS_C_QOP_DEFAULT, &integ_buf, &mic);
+ maj_stat = gss_get_mic(ctx->gc_gss_ctx, &integ_buf, &mic);
status = -EIO; /* XXX? */
if (maj_stat == GSS_S_CONTEXT_EXPIRED)
cred->cr_flags &= ~RPCAUTH_CRED_UPTODATE;
@@ -990,6 +973,113 @@
return 0;
}
+static void
+priv_release_snd_buf(struct rpc_rqst *rqstp)
+{
+ int i;
+
+ for (i=0; i < rqstp->rq_enc_pages_num; i++)
+ __free_page(rqstp->rq_enc_pages[i]);
+ kfree(rqstp->rq_enc_pages);
+}
+
+static int
+alloc_enc_pages(struct rpc_rqst *rqstp)
+{
+ struct xdr_buf *snd_buf = &rqstp->rq_snd_buf;
+ int first, last, i;
+
+ if (snd_buf->page_len == 0) {
+ rqstp->rq_enc_pages_num = 0;
+ return 0;
+ }
+
+ first = snd_buf->page_base >> PAGE_CACHE_SHIFT;
+ last = (snd_buf->page_base + snd_buf->page_len - 1) >> PAGE_CACHE_SHIFT;
+ rqstp->rq_enc_pages_num = last - first + 1 + 1;
+ rqstp->rq_enc_pages
+ = kmalloc(rqstp->rq_enc_pages_num * sizeof(struct page *),
+ GFP_NOFS);
+ if (!rqstp->rq_enc_pages)
+ goto out;
+ for (i=0; i < rqstp->rq_enc_pages_num; i++) {
+ rqstp->rq_enc_pages[i] = alloc_page(GFP_NOFS);
+ if (rqstp->rq_enc_pages[i] == NULL)
+ goto out_free;
+ }
+ rqstp->rq_release_snd_buf = priv_release_snd_buf;
+ return 0;
+out_free:
+ for (i--; i >= 0; i--) {
+ __free_page(rqstp->rq_enc_pages[i]);
+ }
+out:
+ return -EAGAIN;
+}
+
+static inline int
+gss_wrap_req_priv(struct rpc_cred *cred, struct gss_cl_ctx *ctx,
+ kxdrproc_t encode, struct rpc_rqst *rqstp, u32 *p, void *obj)
+{
+ struct xdr_buf *snd_buf = &rqstp->rq_snd_buf;
+ u32 offset;
+ u32 maj_stat;
+ int status;
+ u32 *opaque_len;
+ struct page **inpages;
+ int first;
+ int pad;
+ struct kvec *iov;
+ char *tmp;
+
+ opaque_len = p++;
+ offset = (u8 *)p - (u8 *)snd_buf->head[0].iov_base;
+ *p++ = htonl(rqstp->rq_seqno);
+
+ status = encode(rqstp, p, obj);
+ if (status)
+ return status;
+
+ status = alloc_enc_pages(rqstp);
+ if (status)
+ return status;
+ first = snd_buf->page_base >> PAGE_CACHE_SHIFT;
+ inpages = snd_buf->pages + first;
+ snd_buf->pages = rqstp->rq_enc_pages;
+ snd_buf->page_base -= first << PAGE_CACHE_SHIFT;
+ /* Give the tail its own page, in case we need extra space in the
+ * head when wrapping: */
+ if (snd_buf->page_len || snd_buf->tail[0].iov_len) {
+ tmp = page_address(rqstp->rq_enc_pages[rqstp->rq_enc_pages_num - 1]);
+ memcpy(tmp, snd_buf->tail[0].iov_base, snd_buf->tail[0].iov_len);
+ snd_buf->tail[0].iov_base = tmp;
+ }
+ maj_stat = gss_wrap(ctx->gc_gss_ctx, offset, snd_buf, inpages);
+ /* RPC_SLACK_SPACE should prevent this ever happening: */
+ BUG_ON(snd_buf->len > snd_buf->buflen);
+ status = -EIO;
+ /* We're assuming that when GSS_S_CONTEXT_EXPIRED, the encryption was
+ * done anyway, so it's safe to put the request on the wire: */
+ if (maj_stat == GSS_S_CONTEXT_EXPIRED)
+ cred->cr_flags &= ~RPCAUTH_CRED_UPTODATE;
+ else if (maj_stat)
+ return status;
+
+ *opaque_len = htonl(snd_buf->len - offset);
+ /* guess whether we're in the head or the tail: */
+ if (snd_buf->page_len || snd_buf->tail[0].iov_len)
+ iov = snd_buf->tail;
+ else
+ iov = snd_buf->head;
+ p = iov->iov_base + iov->iov_len;
+ pad = 3 - ((snd_buf->len - offset - 1) & 3);
+ memset(p, 0, pad);
+ iov->iov_len += pad;
+ snd_buf->len += pad;
+
+ return 0;
+}
+
static int
gss_wrap_req(struct rpc_task *task,
kxdrproc_t encode, void *rqstp, u32 *p, void *obj)
@@ -1017,6 +1107,8 @@
rqstp, p, obj);
break;
case RPC_GSS_SVC_PRIVACY:
+ status = gss_wrap_req_priv(cred, ctx, encode,
+ rqstp, p, obj);
break;
}
out:
@@ -1054,8 +1146,7 @@
if (xdr_buf_read_netobj(rcv_buf, &mic, mic_offset))
return status;
- maj_stat = gss_verify_mic(ctx->gc_gss_ctx, &integ_buf,
- &mic, NULL);
+ maj_stat = gss_verify_mic(ctx->gc_gss_ctx, &integ_buf, &mic);
if (maj_stat == GSS_S_CONTEXT_EXPIRED)
cred->cr_flags &= ~RPCAUTH_CRED_UPTODATE;
if (maj_stat != GSS_S_COMPLETE)
@@ -1063,6 +1154,35 @@
return 0;
}
+static inline int
+gss_unwrap_resp_priv(struct rpc_cred *cred, struct gss_cl_ctx *ctx,
+ struct rpc_rqst *rqstp, u32 **p)
+{
+ struct xdr_buf *rcv_buf = &rqstp->rq_rcv_buf;
+ u32 offset;
+ u32 opaque_len;
+ u32 maj_stat;
+ int status = -EIO;
+
+ opaque_len = ntohl(*(*p)++);
+ offset = (u8 *)(*p) - (u8 *)rcv_buf->head[0].iov_base;
+ if (offset + opaque_len > rcv_buf->len)
+ return status;
+ /* remove padding: */
+ rcv_buf->len = offset + opaque_len;
+
+ maj_stat = gss_unwrap(ctx->gc_gss_ctx, offset, rcv_buf);
+ if (maj_stat == GSS_S_CONTEXT_EXPIRED)
+ cred->cr_flags &= ~RPCAUTH_CRED_UPTODATE;
+ if (maj_stat != GSS_S_COMPLETE)
+ return status;
+ if (ntohl(*(*p)++) != rqstp->rq_seqno)
+ return status;
+
+ return 0;
+}
+
+
static int
gss_unwrap_resp(struct rpc_task *task,
kxdrproc_t decode, void *rqstp, u32 *p, void *obj)
@@ -1071,6 +1191,9 @@
struct gss_cred *gss_cred = container_of(cred, struct gss_cred,
gc_base);
struct gss_cl_ctx *ctx = gss_cred_get_ctx(cred);
+ u32 *savedp = p;
+ struct kvec *head = ((struct rpc_rqst *)rqstp)->rq_rcv_buf.head;
+ int savedlen = head->iov_len;
int status = -EIO;
if (ctx->gc_proc != RPC_GSS_PROC_DATA)
@@ -1084,8 +1207,14 @@
goto out;
break;
case RPC_GSS_SVC_PRIVACY:
+ status = gss_unwrap_resp_priv(cred, ctx, rqstp, &p);
+ if (status)
+ goto out;
break;
}
+ /* take into account extra slack for integrity and privacy cases: */
+ task->tk_auth->au_rslack = task->tk_auth->au_verfsize + (p - savedp)
+ + (savedlen - head->iov_len);
out_decode:
status = decode(rqstp, p, obj);
out:
diff --git a/net/sunrpc/auth_gss/gss_krb5_crypto.c b/net/sunrpc/auth_gss/gss_krb5_crypto.c
index ee6ae74..3f3d543 100644
--- a/net/sunrpc/auth_gss/gss_krb5_crypto.c
+++ b/net/sunrpc/auth_gss/gss_krb5_crypto.c
@@ -139,17 +139,91 @@
sg->length = len;
}
+static int
+process_xdr_buf(struct xdr_buf *buf, int offset, int len,
+ int (*actor)(struct scatterlist *, void *), void *data)
+{
+ int i, page_len, thislen, page_offset, ret = 0;
+ struct scatterlist sg[1];
+
+ if (offset >= buf->head[0].iov_len) {
+ offset -= buf->head[0].iov_len;
+ } else {
+ thislen = buf->head[0].iov_len - offset;
+ if (thislen > len)
+ thislen = len;
+ buf_to_sg(sg, buf->head[0].iov_base + offset, thislen);
+ ret = actor(sg, data);
+ if (ret)
+ goto out;
+ offset = 0;
+ len -= thislen;
+ }
+ if (len == 0)
+ goto out;
+
+ if (offset >= buf->page_len) {
+ offset -= buf->page_len;
+ } else {
+ page_len = buf->page_len - offset;
+ if (page_len > len)
+ page_len = len;
+ len -= page_len;
+ page_offset = (offset + buf->page_base) & (PAGE_CACHE_SIZE - 1);
+ i = (offset + buf->page_base) >> PAGE_CACHE_SHIFT;
+ thislen = PAGE_CACHE_SIZE - page_offset;
+ do {
+ if (thislen > page_len)
+ thislen = page_len;
+ sg->page = buf->pages[i];
+ sg->offset = page_offset;
+ sg->length = thislen;
+ ret = actor(sg, data);
+ if (ret)
+ goto out;
+ page_len -= thislen;
+ i++;
+ page_offset = 0;
+ thislen = PAGE_CACHE_SIZE;
+ } while (page_len != 0);
+ offset = 0;
+ }
+ if (len == 0)
+ goto out;
+
+ if (offset < buf->tail[0].iov_len) {
+ thislen = buf->tail[0].iov_len - offset;
+ if (thislen > len)
+ thislen = len;
+ buf_to_sg(sg, buf->tail[0].iov_base + offset, thislen);
+ ret = actor(sg, data);
+ len -= thislen;
+ }
+ if (len != 0)
+ ret = -EINVAL;
+out:
+ return ret;
+}
+
+static int
+checksummer(struct scatterlist *sg, void *data)
+{
+ struct crypto_tfm *tfm = (struct crypto_tfm *)data;
+
+ crypto_digest_update(tfm, sg, 1);
+
+ return 0;
+}
+
/* checksum the plaintext data and hdrlen bytes of the token header */
s32
make_checksum(s32 cksumtype, char *header, int hdrlen, struct xdr_buf *body,
- struct xdr_netobj *cksum)
+ int body_offset, struct xdr_netobj *cksum)
{
char *cksumname;
struct crypto_tfm *tfm = NULL; /* XXX add to ctx? */
struct scatterlist sg[1];
u32 code = GSS_S_FAILURE;
- int len, thislen, offset;
- int i;
switch (cksumtype) {
case CKSUMTYPE_RSA_MD5:
@@ -169,33 +243,8 @@
crypto_digest_init(tfm);
buf_to_sg(sg, header, hdrlen);
crypto_digest_update(tfm, sg, 1);
- if (body->head[0].iov_len) {
- buf_to_sg(sg, body->head[0].iov_base, body->head[0].iov_len);
- crypto_digest_update(tfm, sg, 1);
- }
-
- len = body->page_len;
- if (len != 0) {
- offset = body->page_base & (PAGE_CACHE_SIZE - 1);
- i = body->page_base >> PAGE_CACHE_SHIFT;
- thislen = PAGE_CACHE_SIZE - offset;
- do {
- if (thislen > len)
- thislen = len;
- sg->page = body->pages[i];
- sg->offset = offset;
- sg->length = thislen;
- crypto_digest_update(tfm, sg, 1);
- len -= thislen;
- i++;
- offset = 0;
- thislen = PAGE_CACHE_SIZE;
- } while(len != 0);
- }
- if (body->tail[0].iov_len) {
- buf_to_sg(sg, body->tail[0].iov_base, body->tail[0].iov_len);
- crypto_digest_update(tfm, sg, 1);
- }
+ process_xdr_buf(body, body_offset, body->len - body_offset,
+ checksummer, tfm);
crypto_digest_final(tfm, cksum->data);
code = 0;
out:
@@ -204,3 +253,154 @@
}
EXPORT_SYMBOL(make_checksum);
+
+struct encryptor_desc {
+ u8 iv[8]; /* XXX hard-coded blocksize */
+ struct crypto_tfm *tfm;
+ int pos;
+ struct xdr_buf *outbuf;
+ struct page **pages;
+ struct scatterlist infrags[4];
+ struct scatterlist outfrags[4];
+ int fragno;
+ int fraglen;
+};
+
+static int
+encryptor(struct scatterlist *sg, void *data)
+{
+ struct encryptor_desc *desc = data;
+ struct xdr_buf *outbuf = desc->outbuf;
+ struct page *in_page;
+ int thislen = desc->fraglen + sg->length;
+ int fraglen, ret;
+ int page_pos;
+
+ /* Worst case is 4 fragments: head, end of page 1, start
+ * of page 2, tail. Anything more is a bug. */
+ BUG_ON(desc->fragno > 3);
+ desc->infrags[desc->fragno] = *sg;
+ desc->outfrags[desc->fragno] = *sg;
+
+ page_pos = desc->pos - outbuf->head[0].iov_len;
+ if (page_pos >= 0 && page_pos < outbuf->page_len) {
+ /* pages are not in place: */
+ int i = (page_pos + outbuf->page_base) >> PAGE_CACHE_SHIFT;
+ in_page = desc->pages[i];
+ } else {
+ in_page = sg->page;
+ }
+ desc->infrags[desc->fragno].page = in_page;
+ desc->fragno++;
+ desc->fraglen += sg->length;
+ desc->pos += sg->length;
+
+ fraglen = thislen & 7; /* XXX hardcoded blocksize */
+ thislen -= fraglen;
+
+ if (thislen == 0)
+ return 0;
+
+ ret = crypto_cipher_encrypt_iv(desc->tfm, desc->outfrags, desc->infrags,
+ thislen, desc->iv);
+ if (ret)
+ return ret;
+ if (fraglen) {
+ desc->outfrags[0].page = sg->page;
+ desc->outfrags[0].offset = sg->offset + sg->length - fraglen;
+ desc->outfrags[0].length = fraglen;
+ desc->infrags[0] = desc->outfrags[0];
+ desc->infrags[0].page = in_page;
+ desc->fragno = 1;
+ desc->fraglen = fraglen;
+ } else {
+ desc->fragno = 0;
+ desc->fraglen = 0;
+ }
+ return 0;
+}
+
+int
+gss_encrypt_xdr_buf(struct crypto_tfm *tfm, struct xdr_buf *buf, int offset,
+ struct page **pages)
+{
+ int ret;
+ struct encryptor_desc desc;
+
+ BUG_ON((buf->len - offset) % crypto_tfm_alg_blocksize(tfm) != 0);
+
+ memset(desc.iv, 0, sizeof(desc.iv));
+ desc.tfm = tfm;
+ desc.pos = offset;
+ desc.outbuf = buf;
+ desc.pages = pages;
+ desc.fragno = 0;
+ desc.fraglen = 0;
+
+ ret = process_xdr_buf(buf, offset, buf->len - offset, encryptor, &desc);
+ return ret;
+}
+
+EXPORT_SYMBOL(gss_encrypt_xdr_buf);
+
+struct decryptor_desc {
+ u8 iv[8]; /* XXX hard-coded blocksize */
+ struct crypto_tfm *tfm;
+ struct scatterlist frags[4];
+ int fragno;
+ int fraglen;
+};
+
+static int
+decryptor(struct scatterlist *sg, void *data)
+{
+ struct decryptor_desc *desc = data;
+ int thislen = desc->fraglen + sg->length;
+ int fraglen, ret;
+
+ /* Worst case is 4 fragments: head, end of page 1, start
+ * of page 2, tail. Anything more is a bug. */
+ BUG_ON(desc->fragno > 3);
+ desc->frags[desc->fragno] = *sg;
+ desc->fragno++;
+ desc->fraglen += sg->length;
+
+ fraglen = thislen & 7; /* XXX hardcoded blocksize */
+ thislen -= fraglen;
+
+ if (thislen == 0)
+ return 0;
+
+ ret = crypto_cipher_decrypt_iv(desc->tfm, desc->frags, desc->frags,
+ thislen, desc->iv);
+ if (ret)
+ return ret;
+ if (fraglen) {
+ desc->frags[0].page = sg->page;
+ desc->frags[0].offset = sg->offset + sg->length - fraglen;
+ desc->frags[0].length = fraglen;
+ desc->fragno = 1;
+ desc->fraglen = fraglen;
+ } else {
+ desc->fragno = 0;
+ desc->fraglen = 0;
+ }
+ return 0;
+}
+
+int
+gss_decrypt_xdr_buf(struct crypto_tfm *tfm, struct xdr_buf *buf, int offset)
+{
+ struct decryptor_desc desc;
+
+ /* XXXJBF: */
+ BUG_ON((buf->len - offset) % crypto_tfm_alg_blocksize(tfm) != 0);
+
+ memset(desc.iv, 0, sizeof(desc.iv));
+ desc.tfm = tfm;
+ desc.fragno = 0;
+ desc.fraglen = 0;
+ return process_xdr_buf(buf, offset, buf->len - offset, decryptor, &desc);
+}
+
+EXPORT_SYMBOL(gss_decrypt_xdr_buf);
diff --git a/net/sunrpc/auth_gss/gss_krb5_mech.c b/net/sunrpc/auth_gss/gss_krb5_mech.c
index 606a8a8..5f1f806 100644
--- a/net/sunrpc/auth_gss/gss_krb5_mech.c
+++ b/net/sunrpc/auth_gss/gss_krb5_mech.c
@@ -39,7 +39,6 @@
#include <linux/types.h>
#include <linux/slab.h>
#include <linux/sunrpc/auth.h>
-#include <linux/in.h>
#include <linux/sunrpc/gss_krb5.h>
#include <linux/sunrpc/xdr.h>
#include <linux/crypto.h>
@@ -191,43 +190,12 @@
kfree(kctx);
}
-static u32
-gss_verify_mic_kerberos(struct gss_ctx *ctx,
- struct xdr_buf *message,
- struct xdr_netobj *mic_token,
- u32 *qstate) {
- u32 maj_stat = 0;
- int qop_state;
- struct krb5_ctx *kctx = ctx->internal_ctx_id;
-
- maj_stat = krb5_read_token(kctx, mic_token, message, &qop_state,
- KG_TOK_MIC_MSG);
- if (!maj_stat && qop_state)
- *qstate = qop_state;
-
- dprintk("RPC: gss_verify_mic_kerberos returning %d\n", maj_stat);
- return maj_stat;
-}
-
-static u32
-gss_get_mic_kerberos(struct gss_ctx *ctx,
- u32 qop,
- struct xdr_buf *message,
- struct xdr_netobj *mic_token) {
- u32 err = 0;
- struct krb5_ctx *kctx = ctx->internal_ctx_id;
-
- err = krb5_make_token(kctx, qop, message, mic_token, KG_TOK_MIC_MSG);
-
- dprintk("RPC: gss_get_mic_kerberos returning %d\n",err);
-
- return err;
-}
-
static struct gss_api_ops gss_kerberos_ops = {
.gss_import_sec_context = gss_import_sec_context_kerberos,
.gss_get_mic = gss_get_mic_kerberos,
.gss_verify_mic = gss_verify_mic_kerberos,
+ .gss_wrap = gss_wrap_kerberos,
+ .gss_unwrap = gss_unwrap_kerberos,
.gss_delete_sec_context = gss_delete_sec_context_kerberos,
};
@@ -242,6 +210,11 @@
.service = RPC_GSS_SVC_INTEGRITY,
.name = "krb5i",
},
+ [2] = {
+ .pseudoflavor = RPC_AUTH_GSS_KRB5P,
+ .service = RPC_GSS_SVC_PRIVACY,
+ .name = "krb5p",
+ },
};
static struct gss_api_mech gss_kerberos_mech = {
diff --git a/net/sunrpc/auth_gss/gss_krb5_seal.c b/net/sunrpc/auth_gss/gss_krb5_seal.c
index afeeb87..13f8ae9 100644
--- a/net/sunrpc/auth_gss/gss_krb5_seal.c
+++ b/net/sunrpc/auth_gss/gss_krb5_seal.c
@@ -70,22 +70,13 @@
# define RPCDBG_FACILITY RPCDBG_AUTH
#endif
-static inline int
-gss_krb5_padding(int blocksize, int length) {
- /* Most of the code is block-size independent but in practice we
- * use only 8: */
- BUG_ON(blocksize != 8);
- return 8 - (length & 7);
-}
-
u32
-krb5_make_token(struct krb5_ctx *ctx, int qop_req,
- struct xdr_buf *text, struct xdr_netobj *token,
- int toktype)
+gss_get_mic_kerberos(struct gss_ctx *gss_ctx, struct xdr_buf *text,
+ struct xdr_netobj *token)
{
+ struct krb5_ctx *ctx = gss_ctx->internal_ctx_id;
s32 checksum_type;
struct xdr_netobj md5cksum = {.len = 0, .data = NULL};
- int blocksize = 0, tmsglen;
unsigned char *ptr, *krb5_hdr, *msg_start;
s32 now;
@@ -93,9 +84,6 @@
now = get_seconds();
- if (qop_req != 0)
- goto out_err;
-
switch (ctx->signalg) {
case SGN_ALG_DES_MAC_MD5:
checksum_type = CKSUMTYPE_RSA_MD5;
@@ -111,21 +99,13 @@
goto out_err;
}
- if (toktype == KG_TOK_WRAP_MSG) {
- blocksize = crypto_tfm_alg_blocksize(ctx->enc);
- tmsglen = blocksize + text->len
- + gss_krb5_padding(blocksize, blocksize + text->len);
- } else {
- tmsglen = 0;
- }
-
- token->len = g_token_size(&ctx->mech_used, 22 + tmsglen);
+ token->len = g_token_size(&ctx->mech_used, 22);
ptr = token->data;
- g_make_token_header(&ctx->mech_used, 22 + tmsglen, &ptr);
+ g_make_token_header(&ctx->mech_used, 22, &ptr);
- *ptr++ = (unsigned char) ((toktype>>8)&0xff);
- *ptr++ = (unsigned char) (toktype&0xff);
+ *ptr++ = (unsigned char) ((KG_TOK_MIC_MSG>>8)&0xff);
+ *ptr++ = (unsigned char) (KG_TOK_MIC_MSG&0xff);
/* ptr now at byte 2 of header described in rfc 1964, section 1.2.1: */
krb5_hdr = ptr - 2;
@@ -133,17 +113,9 @@
*(u16 *)(krb5_hdr + 2) = htons(ctx->signalg);
memset(krb5_hdr + 4, 0xff, 4);
- if (toktype == KG_TOK_WRAP_MSG)
- *(u16 *)(krb5_hdr + 4) = htons(ctx->sealalg);
- if (toktype == KG_TOK_WRAP_MSG) {
- /* XXX removing support for now */
- goto out_err;
- } else { /* Sign only. */
- if (make_checksum(checksum_type, krb5_hdr, 8, text,
- &md5cksum))
+ if (make_checksum(checksum_type, krb5_hdr, 8, text, 0, &md5cksum))
goto out_err;
- }
switch (ctx->signalg) {
case SGN_ALG_DES_MAC_MD5:
diff --git a/net/sunrpc/auth_gss/gss_krb5_unseal.c b/net/sunrpc/auth_gss/gss_krb5_unseal.c
index 8767fc5..2030475 100644
--- a/net/sunrpc/auth_gss/gss_krb5_unseal.c
+++ b/net/sunrpc/auth_gss/gss_krb5_unseal.c
@@ -68,21 +68,14 @@
#endif
-/* message_buffer is an input if toktype is MIC and an output if it is WRAP:
- * If toktype is MIC: read_token is a mic token, and message_buffer is the
- * data that the mic was supposedly taken over.
- * If toktype is WRAP: read_token is a wrap token, and message_buffer is used
- * to return the decrypted data.
- */
+/* read_token is a mic token, and message_buffer is the data that the mic was
+ * supposedly taken over. */
-/* XXX will need to change prototype and/or just split into a separate function
- * when we add privacy (because read_token will be in pages too). */
u32
-krb5_read_token(struct krb5_ctx *ctx,
- struct xdr_netobj *read_token,
- struct xdr_buf *message_buffer,
- int *qop_state, int toktype)
+gss_verify_mic_kerberos(struct gss_ctx *gss_ctx,
+ struct xdr_buf *message_buffer, struct xdr_netobj *read_token)
{
+ struct krb5_ctx *ctx = gss_ctx->internal_ctx_id;
int signalg;
int sealalg;
s32 checksum_type;
@@ -100,16 +93,12 @@
read_token->len))
goto out;
- if ((*ptr++ != ((toktype>>8)&0xff)) || (*ptr++ != (toktype&0xff)))
+ if ((*ptr++ != ((KG_TOK_MIC_MSG>>8)&0xff)) ||
+ (*ptr++ != ( KG_TOK_MIC_MSG &0xff)) )
goto out;
/* XXX sanity-check bodysize?? */
- if (toktype == KG_TOK_WRAP_MSG) {
- /* XXX gone */
- goto out;
- }
-
/* get the sign and seal algorithms */
signalg = ptr[0] + (ptr[1] << 8);
@@ -120,14 +109,7 @@
if ((ptr[4] != 0xff) || (ptr[5] != 0xff))
goto out;
- if (((toktype != KG_TOK_WRAP_MSG) && (sealalg != 0xffff)) ||
- ((toktype == KG_TOK_WRAP_MSG) && (sealalg == 0xffff)))
- goto out;
-
- /* in the current spec, there is only one valid seal algorithm per
- key type, so a simple comparison is ok */
-
- if ((toktype == KG_TOK_WRAP_MSG) && !(sealalg == ctx->sealalg))
+ if (sealalg != 0xffff)
goto out;
/* there are several mappings of seal algorithms to sign algorithms,
@@ -154,7 +136,7 @@
switch (signalg) {
case SGN_ALG_DES_MAC_MD5:
ret = make_checksum(checksum_type, ptr - 2, 8,
- message_buffer, &md5cksum);
+ message_buffer, 0, &md5cksum);
if (ret)
goto out;
@@ -175,9 +157,6 @@
/* it got through unscathed. Make sure the context is unexpired */
- if (qop_state)
- *qop_state = GSS_C_QOP_DEFAULT;
-
now = get_seconds();
ret = GSS_S_CONTEXT_EXPIRED;
diff --git a/net/sunrpc/auth_gss/gss_krb5_wrap.c b/net/sunrpc/auth_gss/gss_krb5_wrap.c
new file mode 100644
index 0000000..af777cf
--- /dev/null
+++ b/net/sunrpc/auth_gss/gss_krb5_wrap.c
@@ -0,0 +1,363 @@
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/jiffies.h>
+#include <linux/sunrpc/gss_krb5.h>
+#include <linux/random.h>
+#include <linux/pagemap.h>
+#include <asm/scatterlist.h>
+#include <linux/crypto.h>
+
+#ifdef RPC_DEBUG
+# define RPCDBG_FACILITY RPCDBG_AUTH
+#endif
+
+static inline int
+gss_krb5_padding(int blocksize, int length)
+{
+ /* Most of the code is block-size independent but currently we
+ * use only 8: */
+ BUG_ON(blocksize != 8);
+ return 8 - (length & 7);
+}
+
+static inline void
+gss_krb5_add_padding(struct xdr_buf *buf, int offset, int blocksize)
+{
+ int padding = gss_krb5_padding(blocksize, buf->len - offset);
+ char *p;
+ struct kvec *iov;
+
+ if (buf->page_len || buf->tail[0].iov_len)
+ iov = &buf->tail[0];
+ else
+ iov = &buf->head[0];
+ p = iov->iov_base + iov->iov_len;
+ iov->iov_len += padding;
+ buf->len += padding;
+ memset(p, padding, padding);
+}
+
+static inline int
+gss_krb5_remove_padding(struct xdr_buf *buf, int blocksize)
+{
+ u8 *ptr;
+ u8 pad;
+ int len = buf->len;
+
+ if (len <= buf->head[0].iov_len) {
+ pad = *(u8 *)(buf->head[0].iov_base + len - 1);
+ if (pad > buf->head[0].iov_len)
+ return -EINVAL;
+ buf->head[0].iov_len -= pad;
+ goto out;
+ } else
+ len -= buf->head[0].iov_len;
+ if (len <= buf->page_len) {
+ int last = (buf->page_base + len - 1)
+ >>PAGE_CACHE_SHIFT;
+ int offset = (buf->page_base + len - 1)
+ & (PAGE_CACHE_SIZE - 1);
+ ptr = kmap_atomic(buf->pages[last], KM_SKB_SUNRPC_DATA);
+ pad = *(ptr + offset);
+ kunmap_atomic(ptr, KM_SKB_SUNRPC_DATA);
+ goto out;
+ } else
+ len -= buf->page_len;
+ BUG_ON(len > buf->tail[0].iov_len);
+ pad = *(u8 *)(buf->tail[0].iov_base + len - 1);
+out:
+ /* XXX: NOTE: we do not adjust the page lengths--they represent
+ * a range of data in the real filesystem page cache, and we need
+ * to know that range so the xdr code can properly place read data.
+ * However adjusting the head length, as we do above, is harmless.
+ * In the case of a request that fits into a single page, the server
+ * also uses length and head length together to determine the original
+ * start of the request to copy the request for deferal; so it's
+ * easier on the server if we adjust head and tail length in tandem.
+ * It's not really a problem that we don't fool with the page and
+ * tail lengths, though--at worst badly formed xdr might lead the
+ * server to attempt to parse the padding.
+ * XXX: Document all these weird requirements for gss mechanism
+ * wrap/unwrap functions. */
+ if (pad > blocksize)
+ return -EINVAL;
+ if (buf->len > pad)
+ buf->len -= pad;
+ else
+ return -EINVAL;
+ return 0;
+}
+
+static inline void
+make_confounder(char *p, int blocksize)
+{
+ static u64 i = 0;
+ u64 *q = (u64 *)p;
+
+ /* rfc1964 claims this should be "random". But all that's really
+ * necessary is that it be unique. And not even that is necessary in
+ * our case since our "gssapi" implementation exists only to support
+ * rpcsec_gss, so we know that the only buffers we will ever encrypt
+ * already begin with a unique sequence number. Just to hedge my bets
+ * I'll make a half-hearted attempt at something unique, but ensuring
+ * uniqueness would mean worrying about atomicity and rollover, and I
+ * don't care enough. */
+
+ BUG_ON(blocksize != 8);
+ *q = i++;
+}
+
+/* Assumptions: the head and tail of inbuf are ours to play with.
+ * The pages, however, may be real pages in the page cache and we replace
+ * them with scratch pages from **pages before writing to them. */
+/* XXX: obviously the above should be documentation of wrap interface,
+ * and shouldn't be in this kerberos-specific file. */
+
+/* XXX factor out common code with seal/unseal. */
+
+u32
+gss_wrap_kerberos(struct gss_ctx *ctx, int offset,
+ struct xdr_buf *buf, struct page **pages)
+{
+ struct krb5_ctx *kctx = ctx->internal_ctx_id;
+ s32 checksum_type;
+ struct xdr_netobj md5cksum = {.len = 0, .data = NULL};
+ int blocksize = 0, plainlen;
+ unsigned char *ptr, *krb5_hdr, *msg_start;
+ s32 now;
+ int headlen;
+ struct page **tmp_pages;
+
+ dprintk("RPC: gss_wrap_kerberos\n");
+
+ now = get_seconds();
+
+ switch (kctx->signalg) {
+ case SGN_ALG_DES_MAC_MD5:
+ checksum_type = CKSUMTYPE_RSA_MD5;
+ break;
+ default:
+ dprintk("RPC: gss_krb5_seal: kctx->signalg %d not"
+ " supported\n", kctx->signalg);
+ goto out_err;
+ }
+ if (kctx->sealalg != SEAL_ALG_NONE && kctx->sealalg != SEAL_ALG_DES) {
+ dprintk("RPC: gss_krb5_seal: kctx->sealalg %d not supported\n",
+ kctx->sealalg);
+ goto out_err;
+ }
+
+ blocksize = crypto_tfm_alg_blocksize(kctx->enc);
+ gss_krb5_add_padding(buf, offset, blocksize);
+ BUG_ON((buf->len - offset) % blocksize);
+ plainlen = blocksize + buf->len - offset;
+
+ headlen = g_token_size(&kctx->mech_used, 22 + plainlen) -
+ (buf->len - offset);
+
+ ptr = buf->head[0].iov_base + offset;
+ /* shift data to make room for header. */
+ /* XXX Would be cleverer to encrypt while copying. */
+ /* XXX bounds checking, slack, etc. */
+ memmove(ptr + headlen, ptr, buf->head[0].iov_len - offset);
+ buf->head[0].iov_len += headlen;
+ buf->len += headlen;
+ BUG_ON((buf->len - offset - headlen) % blocksize);
+
+ g_make_token_header(&kctx->mech_used, 22 + plainlen, &ptr);
+
+
+ *ptr++ = (unsigned char) ((KG_TOK_WRAP_MSG>>8)&0xff);
+ *ptr++ = (unsigned char) (KG_TOK_WRAP_MSG&0xff);
+
+ /* ptr now at byte 2 of header described in rfc 1964, section 1.2.1: */
+ krb5_hdr = ptr - 2;
+ msg_start = krb5_hdr + 24;
+ /* XXXJBF: */ BUG_ON(buf->head[0].iov_base + offset + headlen != msg_start + blocksize);
+
+ *(u16 *)(krb5_hdr + 2) = htons(kctx->signalg);
+ memset(krb5_hdr + 4, 0xff, 4);
+ *(u16 *)(krb5_hdr + 4) = htons(kctx->sealalg);
+
+ make_confounder(msg_start, blocksize);
+
+ /* XXXJBF: UGH!: */
+ tmp_pages = buf->pages;
+ buf->pages = pages;
+ if (make_checksum(checksum_type, krb5_hdr, 8, buf,
+ offset + headlen - blocksize, &md5cksum))
+ goto out_err;
+ buf->pages = tmp_pages;
+
+ switch (kctx->signalg) {
+ case SGN_ALG_DES_MAC_MD5:
+ if (krb5_encrypt(kctx->seq, NULL, md5cksum.data,
+ md5cksum.data, md5cksum.len))
+ goto out_err;
+ memcpy(krb5_hdr + 16,
+ md5cksum.data + md5cksum.len - KRB5_CKSUM_LENGTH,
+ KRB5_CKSUM_LENGTH);
+
+ dprintk("RPC: make_seal_token: cksum data: \n");
+ print_hexl((u32 *) (krb5_hdr + 16), KRB5_CKSUM_LENGTH, 0);
+ break;
+ default:
+ BUG();
+ }
+
+ kfree(md5cksum.data);
+
+ /* XXX would probably be more efficient to compute checksum
+ * and encrypt at the same time: */
+ if ((krb5_make_seq_num(kctx->seq, kctx->initiate ? 0 : 0xff,
+ kctx->seq_send, krb5_hdr + 16, krb5_hdr + 8)))
+ goto out_err;
+
+ if (gss_encrypt_xdr_buf(kctx->enc, buf, offset + headlen - blocksize,
+ pages))
+ goto out_err;
+
+ kctx->seq_send++;
+
+ return ((kctx->endtime < now) ? GSS_S_CONTEXT_EXPIRED : GSS_S_COMPLETE);
+out_err:
+ if (md5cksum.data) kfree(md5cksum.data);
+ return GSS_S_FAILURE;
+}
+
+u32
+gss_unwrap_kerberos(struct gss_ctx *ctx, int offset, struct xdr_buf *buf)
+{
+ struct krb5_ctx *kctx = ctx->internal_ctx_id;
+ int signalg;
+ int sealalg;
+ s32 checksum_type;
+ struct xdr_netobj md5cksum = {.len = 0, .data = NULL};
+ s32 now;
+ int direction;
+ s32 seqnum;
+ unsigned char *ptr;
+ int bodysize;
+ u32 ret = GSS_S_DEFECTIVE_TOKEN;
+ void *data_start, *orig_start;
+ int data_len;
+ int blocksize;
+
+ dprintk("RPC: gss_unwrap_kerberos\n");
+
+ ptr = (u8 *)buf->head[0].iov_base + offset;
+ if (g_verify_token_header(&kctx->mech_used, &bodysize, &ptr,
+ buf->len - offset))
+ goto out;
+
+ if ((*ptr++ != ((KG_TOK_WRAP_MSG>>8)&0xff)) ||
+ (*ptr++ != (KG_TOK_WRAP_MSG &0xff)) )
+ goto out;
+
+ /* XXX sanity-check bodysize?? */
+
+ /* get the sign and seal algorithms */
+
+ signalg = ptr[0] + (ptr[1] << 8);
+ sealalg = ptr[2] + (ptr[3] << 8);
+
+ /* Sanity checks */
+
+ if ((ptr[4] != 0xff) || (ptr[5] != 0xff))
+ goto out;
+
+ if (sealalg == 0xffff)
+ goto out;
+
+ /* in the current spec, there is only one valid seal algorithm per
+ key type, so a simple comparison is ok */
+
+ if (sealalg != kctx->sealalg)
+ goto out;
+
+ /* there are several mappings of seal algorithms to sign algorithms,
+ but few enough that we can try them all. */
+
+ if ((kctx->sealalg == SEAL_ALG_NONE && signalg > 1) ||
+ (kctx->sealalg == SEAL_ALG_1 && signalg != SGN_ALG_3) ||
+ (kctx->sealalg == SEAL_ALG_DES3KD &&
+ signalg != SGN_ALG_HMAC_SHA1_DES3_KD))
+ goto out;
+
+ if (gss_decrypt_xdr_buf(kctx->enc, buf,
+ ptr + 22 - (unsigned char *)buf->head[0].iov_base))
+ goto out;
+
+ /* compute the checksum of the message */
+
+ /* initialize the the cksum */
+ switch (signalg) {
+ case SGN_ALG_DES_MAC_MD5:
+ checksum_type = CKSUMTYPE_RSA_MD5;
+ break;
+ default:
+ ret = GSS_S_DEFECTIVE_TOKEN;
+ goto out;
+ }
+
+ switch (signalg) {
+ case SGN_ALG_DES_MAC_MD5:
+ ret = make_checksum(checksum_type, ptr - 2, 8, buf,
+ ptr + 22 - (unsigned char *)buf->head[0].iov_base, &md5cksum);
+ if (ret)
+ goto out;
+
+ ret = krb5_encrypt(kctx->seq, NULL, md5cksum.data,
+ md5cksum.data, md5cksum.len);
+ if (ret)
+ goto out;
+
+ if (memcmp(md5cksum.data + 8, ptr + 14, 8)) {
+ ret = GSS_S_BAD_SIG;
+ goto out;
+ }
+ break;
+ default:
+ ret = GSS_S_DEFECTIVE_TOKEN;
+ goto out;
+ }
+
+ /* it got through unscathed. Make sure the context is unexpired */
+
+ now = get_seconds();
+
+ ret = GSS_S_CONTEXT_EXPIRED;
+ if (now > kctx->endtime)
+ goto out;
+
+ /* do sequencing checks */
+
+ ret = GSS_S_BAD_SIG;
+ if ((ret = krb5_get_seq_num(kctx->seq, ptr + 14, ptr + 6, &direction,
+ &seqnum)))
+ goto out;
+
+ if ((kctx->initiate && direction != 0xff) ||
+ (!kctx->initiate && direction != 0))
+ goto out;
+
+ /* Copy the data back to the right position. XXX: Would probably be
+ * better to copy and encrypt at the same time. */
+
+ blocksize = crypto_tfm_alg_blocksize(kctx->enc);
+ data_start = ptr + 22 + blocksize;
+ orig_start = buf->head[0].iov_base + offset;
+ data_len = (buf->head[0].iov_base + buf->head[0].iov_len) - data_start;
+ memmove(orig_start, data_start, data_len);
+ buf->head[0].iov_len -= (data_start - orig_start);
+ buf->len -= (data_start - orig_start);
+
+ ret = GSS_S_DEFECTIVE_TOKEN;
+ if (gss_krb5_remove_padding(buf, blocksize))
+ goto out;
+
+ ret = GSS_S_COMPLETE;
+out:
+ if (md5cksum.data) kfree(md5cksum.data);
+ return ret;
+}
diff --git a/net/sunrpc/auth_gss/gss_mech_switch.c b/net/sunrpc/auth_gss/gss_mech_switch.c
index 9dfb683..b048bf67 100644
--- a/net/sunrpc/auth_gss/gss_mech_switch.c
+++ b/net/sunrpc/auth_gss/gss_mech_switch.c
@@ -35,7 +35,6 @@
#include <linux/types.h>
#include <linux/slab.h>
-#include <linux/socket.h>
#include <linux/module.h>
#include <linux/sunrpc/msg_prot.h>
#include <linux/sunrpc/gss_asn1.h>
@@ -251,13 +250,11 @@
u32
gss_get_mic(struct gss_ctx *context_handle,
- u32 qop,
struct xdr_buf *message,
struct xdr_netobj *mic_token)
{
return context_handle->mech_type->gm_ops
->gss_get_mic(context_handle,
- qop,
message,
mic_token);
}
@@ -267,16 +264,34 @@
u32
gss_verify_mic(struct gss_ctx *context_handle,
struct xdr_buf *message,
- struct xdr_netobj *mic_token,
- u32 *qstate)
+ struct xdr_netobj *mic_token)
{
return context_handle->mech_type->gm_ops
->gss_verify_mic(context_handle,
message,
- mic_token,
- qstate);
+ mic_token);
}
+u32
+gss_wrap(struct gss_ctx *ctx_id,
+ int offset,
+ struct xdr_buf *buf,
+ struct page **inpages)
+{
+ return ctx_id->mech_type->gm_ops
+ ->gss_wrap(ctx_id, offset, buf, inpages);
+}
+
+u32
+gss_unwrap(struct gss_ctx *ctx_id,
+ int offset,
+ struct xdr_buf *buf)
+{
+ return ctx_id->mech_type->gm_ops
+ ->gss_unwrap(ctx_id, offset, buf);
+}
+
+
/* gss_delete_sec_context: free all resources associated with context_handle.
* Note this differs from the RFC 2744-specified prototype in that we don't
* bother returning an output token, since it would never be used anyway. */
diff --git a/net/sunrpc/auth_gss/gss_spkm3_mech.c b/net/sunrpc/auth_gss/gss_spkm3_mech.c
index 6c97d61..39b3edc 100644
--- a/net/sunrpc/auth_gss/gss_spkm3_mech.c
+++ b/net/sunrpc/auth_gss/gss_spkm3_mech.c
@@ -224,18 +224,13 @@
static u32
gss_verify_mic_spkm3(struct gss_ctx *ctx,
struct xdr_buf *signbuf,
- struct xdr_netobj *checksum,
- u32 *qstate) {
+ struct xdr_netobj *checksum)
+{
u32 maj_stat = 0;
- int qop_state = 0;
struct spkm3_ctx *sctx = ctx->internal_ctx_id;
dprintk("RPC: gss_verify_mic_spkm3 calling spkm3_read_token\n");
- maj_stat = spkm3_read_token(sctx, checksum, signbuf, &qop_state,
- SPKM_MIC_TOK);
-
- if (!maj_stat && qop_state)
- *qstate = qop_state;
+ maj_stat = spkm3_read_token(sctx, checksum, signbuf, SPKM_MIC_TOK);
dprintk("RPC: gss_verify_mic_spkm3 returning %d\n", maj_stat);
return maj_stat;
@@ -243,15 +238,15 @@
static u32
gss_get_mic_spkm3(struct gss_ctx *ctx,
- u32 qop,
struct xdr_buf *message_buffer,
- struct xdr_netobj *message_token) {
+ struct xdr_netobj *message_token)
+{
u32 err = 0;
struct spkm3_ctx *sctx = ctx->internal_ctx_id;
dprintk("RPC: gss_get_mic_spkm3\n");
- err = spkm3_make_token(sctx, qop, message_buffer,
+ err = spkm3_make_token(sctx, message_buffer,
message_token, SPKM_MIC_TOK);
return err;
}
@@ -264,8 +259,8 @@
};
static struct pf_desc gss_spkm3_pfs[] = {
- {RPC_AUTH_GSS_SPKM, 0, RPC_GSS_SVC_NONE, "spkm3"},
- {RPC_AUTH_GSS_SPKMI, 0, RPC_GSS_SVC_INTEGRITY, "spkm3i"},
+ {RPC_AUTH_GSS_SPKM, RPC_GSS_SVC_NONE, "spkm3"},
+ {RPC_AUTH_GSS_SPKMI, RPC_GSS_SVC_INTEGRITY, "spkm3i"},
};
static struct gss_api_mech gss_spkm3_mech = {
diff --git a/net/sunrpc/auth_gss/gss_spkm3_seal.c b/net/sunrpc/auth_gss/gss_spkm3_seal.c
index 2533986..148201e 100644
--- a/net/sunrpc/auth_gss/gss_spkm3_seal.c
+++ b/net/sunrpc/auth_gss/gss_spkm3_seal.c
@@ -51,7 +51,7 @@
*/
u32
-spkm3_make_token(struct spkm3_ctx *ctx, int qop_req,
+spkm3_make_token(struct spkm3_ctx *ctx,
struct xdr_buf * text, struct xdr_netobj * token,
int toktype)
{
@@ -68,8 +68,6 @@
dprintk("RPC: spkm3_make_token\n");
now = jiffies;
- if (qop_req != 0)
- goto out_err;
if (ctx->ctx_id.len != 16) {
dprintk("RPC: spkm3_make_token BAD ctx_id.len %d\n",
diff --git a/net/sunrpc/auth_gss/gss_spkm3_unseal.c b/net/sunrpc/auth_gss/gss_spkm3_unseal.c
index 65ce81b..c3c0d95 100644
--- a/net/sunrpc/auth_gss/gss_spkm3_unseal.c
+++ b/net/sunrpc/auth_gss/gss_spkm3_unseal.c
@@ -52,7 +52,7 @@
spkm3_read_token(struct spkm3_ctx *ctx,
struct xdr_netobj *read_token, /* checksum */
struct xdr_buf *message_buffer, /* signbuf */
- int *qop_state, int toktype)
+ int toktype)
{
s32 code;
struct xdr_netobj wire_cksum = {.len =0, .data = NULL};
diff --git a/net/sunrpc/auth_gss/svcauth_gss.c b/net/sunrpc/auth_gss/svcauth_gss.c
index e330819..e4ada15 100644
--- a/net/sunrpc/auth_gss/svcauth_gss.c
+++ b/net/sunrpc/auth_gss/svcauth_gss.c
@@ -566,8 +566,7 @@
if (rqstp->rq_deferred) /* skip verification of revisited request */
return SVC_OK;
- if (gss_verify_mic(ctx_id, &rpchdr, &checksum, NULL)
- != GSS_S_COMPLETE) {
+ if (gss_verify_mic(ctx_id, &rpchdr, &checksum) != GSS_S_COMPLETE) {
*authp = rpcsec_gsserr_credproblem;
return SVC_DENIED;
}
@@ -604,7 +603,7 @@
xdr_buf_from_iov(&iov, &verf_data);
p = rqstp->rq_res.head->iov_base + rqstp->rq_res.head->iov_len;
mic.data = (u8 *)(p + 1);
- maj_stat = gss_get_mic(ctx_id, 0, &verf_data, &mic);
+ maj_stat = gss_get_mic(ctx_id, &verf_data, &mic);
if (maj_stat != GSS_S_COMPLETE)
return -1;
*p++ = htonl(mic.len);
@@ -710,7 +709,7 @@
goto out;
if (read_bytes_from_xdr_buf(buf, integ_len + 4, mic.data, mic.len))
goto out;
- maj_stat = gss_verify_mic(ctx, &integ_buf, &mic, NULL);
+ maj_stat = gss_verify_mic(ctx, &integ_buf, &mic);
if (maj_stat != GSS_S_COMPLETE)
goto out;
if (ntohl(svc_getu32(&buf->head[0])) != seq)
@@ -1012,7 +1011,7 @@
resv = &resbuf->tail[0];
}
mic.data = (u8 *)resv->iov_base + resv->iov_len + 4;
- if (gss_get_mic(gsd->rsci->mechctx, 0, &integ_buf, &mic))
+ if (gss_get_mic(gsd->rsci->mechctx, &integ_buf, &mic))
goto out_err;
svc_putu32(resv, htonl(mic.len));
memset(mic.data + mic.len, 0,
diff --git a/net/sunrpc/auth_null.c b/net/sunrpc/auth_null.c
index 9b72d3a..f56767a 100644
--- a/net/sunrpc/auth_null.c
+++ b/net/sunrpc/auth_null.c
@@ -7,9 +7,7 @@
*/
#include <linux/types.h>
-#include <linux/socket.h>
#include <linux/module.h>
-#include <linux/in.h>
#include <linux/utsname.h>
#include <linux/sunrpc/clnt.h>
#include <linux/sched.h>
diff --git a/net/sunrpc/auth_unix.c b/net/sunrpc/auth_unix.c
index 4ff297a..890fb5e 100644
--- a/net/sunrpc/auth_unix.c
+++ b/net/sunrpc/auth_unix.c
@@ -9,8 +9,6 @@
#include <linux/types.h>
#include <linux/sched.h>
#include <linux/module.h>
-#include <linux/socket.h>
-#include <linux/in.h>
#include <linux/sunrpc/clnt.h>
#include <linux/sunrpc/auth.h>
diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index f17e615..702ede3 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -1,5 +1,5 @@
/*
- * linux/net/sunrpc/rpcclnt.c
+ * linux/net/sunrpc/clnt.c
*
* This file contains the high-level RPC interface.
* It is modeled as a finite state machine to support both synchronous
@@ -27,7 +27,6 @@
#include <linux/types.h>
#include <linux/mm.h>
#include <linux/slab.h>
-#include <linux/in.h>
#include <linux/utsname.h>
#include <linux/sunrpc/clnt.h>
@@ -53,6 +52,7 @@
static void call_encode(struct rpc_task *task);
static void call_decode(struct rpc_task *task);
static void call_bind(struct rpc_task *task);
+static void call_bind_status(struct rpc_task *task);
static void call_transmit(struct rpc_task *task);
static void call_status(struct rpc_task *task);
static void call_refresh(struct rpc_task *task);
@@ -517,15 +517,8 @@
rpc_setbufsize(struct rpc_clnt *clnt, unsigned int sndsize, unsigned int rcvsize)
{
struct rpc_xprt *xprt = clnt->cl_xprt;
-
- xprt->sndsize = 0;
- if (sndsize)
- xprt->sndsize = sndsize + RPC_SLACK_SPACE;
- xprt->rcvsize = 0;
- if (rcvsize)
- xprt->rcvsize = rcvsize + RPC_SLACK_SPACE;
- if (xprt_connected(xprt))
- xprt_sock_setbufsize(xprt);
+ if (xprt->ops->set_buffer_size)
+ xprt->ops->set_buffer_size(xprt, sndsize, rcvsize);
}
/*
@@ -685,13 +678,11 @@
static void
call_encode(struct rpc_task *task)
{
- struct rpc_clnt *clnt = task->tk_client;
struct rpc_rqst *req = task->tk_rqstp;
struct xdr_buf *sndbuf = &req->rq_snd_buf;
struct xdr_buf *rcvbuf = &req->rq_rcv_buf;
unsigned int bufsiz;
kxdrproc_t encode;
- int status;
u32 *p;
dprintk("RPC: %4d call_encode (status %d)\n",
@@ -719,11 +710,15 @@
rpc_exit(task, -EIO);
return;
}
- if (encode && (status = rpcauth_wrap_req(task, encode, req, p,
- task->tk_msg.rpc_argp)) < 0) {
- printk(KERN_WARNING "%s: can't encode arguments: %d\n",
- clnt->cl_protname, -status);
- rpc_exit(task, status);
+ if (encode == NULL)
+ return;
+
+ task->tk_status = rpcauth_wrap_req(task, encode, req, p,
+ task->tk_msg.rpc_argp);
+ if (task->tk_status == -ENOMEM) {
+ /* XXX: Is this sane? */
+ rpc_delay(task, 3*HZ);
+ task->tk_status = -EAGAIN;
}
}
@@ -734,43 +729,95 @@
call_bind(struct rpc_task *task)
{
struct rpc_clnt *clnt = task->tk_client;
- struct rpc_xprt *xprt = clnt->cl_xprt;
- dprintk("RPC: %4d call_bind xprt %p %s connected\n", task->tk_pid,
- xprt, (xprt_connected(xprt) ? "is" : "is not"));
+ dprintk("RPC: %4d call_bind (status %d)\n",
+ task->tk_pid, task->tk_status);
- task->tk_action = (xprt_connected(xprt)) ? call_transmit : call_connect;
-
+ task->tk_action = call_connect;
if (!clnt->cl_port) {
- task->tk_action = call_connect;
- task->tk_timeout = RPC_CONNECT_TIMEOUT;
+ task->tk_action = call_bind_status;
+ task->tk_timeout = task->tk_xprt->bind_timeout;
rpc_getport(task, clnt);
}
}
/*
- * 4a. Connect to the RPC server (TCP case)
+ * 4a. Sort out bind result
+ */
+static void
+call_bind_status(struct rpc_task *task)
+{
+ int status = -EACCES;
+
+ if (task->tk_status >= 0) {
+ dprintk("RPC: %4d call_bind_status (status %d)\n",
+ task->tk_pid, task->tk_status);
+ task->tk_status = 0;
+ task->tk_action = call_connect;
+ return;
+ }
+
+ switch (task->tk_status) {
+ case -EACCES:
+ dprintk("RPC: %4d remote rpcbind: RPC program/version unavailable\n",
+ task->tk_pid);
+ rpc_delay(task, 3*HZ);
+ goto retry_bind;
+ case -ETIMEDOUT:
+ dprintk("RPC: %4d rpcbind request timed out\n",
+ task->tk_pid);
+ if (RPC_IS_SOFT(task)) {
+ status = -EIO;
+ break;
+ }
+ goto retry_bind;
+ case -EPFNOSUPPORT:
+ dprintk("RPC: %4d remote rpcbind service unavailable\n",
+ task->tk_pid);
+ break;
+ case -EPROTONOSUPPORT:
+ dprintk("RPC: %4d remote rpcbind version 2 unavailable\n",
+ task->tk_pid);
+ break;
+ default:
+ dprintk("RPC: %4d unrecognized rpcbind error (%d)\n",
+ task->tk_pid, -task->tk_status);
+ status = -EIO;
+ break;
+ }
+
+ rpc_exit(task, status);
+ return;
+
+retry_bind:
+ task->tk_status = 0;
+ task->tk_action = call_bind;
+ return;
+}
+
+/*
+ * 4b. Connect to the RPC server
*/
static void
call_connect(struct rpc_task *task)
{
- struct rpc_clnt *clnt = task->tk_client;
+ struct rpc_xprt *xprt = task->tk_xprt;
- dprintk("RPC: %4d call_connect status %d\n",
- task->tk_pid, task->tk_status);
+ dprintk("RPC: %4d call_connect xprt %p %s connected\n",
+ task->tk_pid, xprt,
+ (xprt_connected(xprt) ? "is" : "is not"));
- if (xprt_connected(clnt->cl_xprt)) {
- task->tk_action = call_transmit;
- return;
+ task->tk_action = call_transmit;
+ if (!xprt_connected(xprt)) {
+ task->tk_action = call_connect_status;
+ if (task->tk_status < 0)
+ return;
+ xprt_connect(task);
}
- task->tk_action = call_connect_status;
- if (task->tk_status < 0)
- return;
- xprt_connect(task);
}
/*
- * 4b. Sort out connect result
+ * 4c. Sort out connect result
*/
static void
call_connect_status(struct rpc_task *task)
@@ -778,6 +825,9 @@
struct rpc_clnt *clnt = task->tk_client;
int status = task->tk_status;
+ dprintk("RPC: %5u call_connect_status (status %d)\n",
+ task->tk_pid, task->tk_status);
+
task->tk_status = 0;
if (status >= 0) {
clnt->cl_stats->netreconn++;
@@ -785,17 +835,19 @@
return;
}
- /* Something failed: we may have to rebind */
+ /* Something failed: remote service port may have changed */
if (clnt->cl_autobind)
clnt->cl_port = 0;
+
switch (status) {
case -ENOTCONN:
case -ETIMEDOUT:
case -EAGAIN:
- task->tk_action = (clnt->cl_port == 0) ? call_bind : call_connect;
+ task->tk_action = call_bind;
break;
default:
rpc_exit(task, -EIO);
+ break;
}
}
@@ -815,10 +867,12 @@
if (task->tk_status != 0)
return;
/* Encode here so that rpcsec_gss can use correct sequence number. */
- if (!task->tk_rqstp->rq_bytes_sent)
+ if (task->tk_rqstp->rq_bytes_sent == 0) {
call_encode(task);
- if (task->tk_status < 0)
- return;
+ /* Did the encode result in an error condition? */
+ if (task->tk_status != 0)
+ goto out_nosend;
+ }
xprt_transmit(task);
if (task->tk_status < 0)
return;
@@ -826,6 +880,10 @@
task->tk_action = NULL;
rpc_wake_up_task(task);
}
+ return;
+out_nosend:
+ /* release socket write lock before attempting to handle error */
+ xprt_abort_transmit(task);
}
/*
@@ -1020,13 +1078,12 @@
call_header(struct rpc_task *task)
{
struct rpc_clnt *clnt = task->tk_client;
- struct rpc_xprt *xprt = clnt->cl_xprt;
struct rpc_rqst *req = task->tk_rqstp;
u32 *p = req->rq_svec[0].iov_base;
/* FIXME: check buffer size? */
- if (xprt->stream)
- *p++ = 0; /* fill in later */
+
+ p = xprt_skip_transport_header(task->tk_xprt, p);
*p++ = req->rq_xid; /* XID */
*p++ = htonl(RPC_CALL); /* CALL */
*p++ = htonl(RPC_VERSION); /* RPC version */
diff --git a/net/sunrpc/pmap_clnt.c b/net/sunrpc/pmap_clnt.c
index 4e81f27..a398575 100644
--- a/net/sunrpc/pmap_clnt.c
+++ b/net/sunrpc/pmap_clnt.c
@@ -26,7 +26,7 @@
#define PMAP_GETPORT 3
static struct rpc_procinfo pmap_procedures[];
-static struct rpc_clnt * pmap_create(char *, struct sockaddr_in *, int);
+static struct rpc_clnt * pmap_create(char *, struct sockaddr_in *, int, int);
static void pmap_getport_done(struct rpc_task *);
static struct rpc_program pmap_program;
static DEFINE_SPINLOCK(pmap_lock);
@@ -65,7 +65,7 @@
map->pm_binding = 1;
spin_unlock(&pmap_lock);
- pmap_clnt = pmap_create(clnt->cl_server, sap, map->pm_prot);
+ pmap_clnt = pmap_create(clnt->cl_server, sap, map->pm_prot, 0);
if (IS_ERR(pmap_clnt)) {
task->tk_status = PTR_ERR(pmap_clnt);
goto bailout;
@@ -112,7 +112,7 @@
NIPQUAD(sin->sin_addr.s_addr), prog, vers, prot);
sprintf(hostname, "%u.%u.%u.%u", NIPQUAD(sin->sin_addr.s_addr));
- pmap_clnt = pmap_create(hostname, sin, prot);
+ pmap_clnt = pmap_create(hostname, sin, prot, 0);
if (IS_ERR(pmap_clnt))
return PTR_ERR(pmap_clnt);
@@ -171,7 +171,7 @@
sin.sin_family = AF_INET;
sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
- pmap_clnt = pmap_create("localhost", &sin, IPPROTO_UDP);
+ pmap_clnt = pmap_create("localhost", &sin, IPPROTO_UDP, 1);
if (IS_ERR(pmap_clnt)) {
error = PTR_ERR(pmap_clnt);
dprintk("RPC: couldn't create pmap client. Error = %d\n", error);
@@ -198,7 +198,7 @@
}
static struct rpc_clnt *
-pmap_create(char *hostname, struct sockaddr_in *srvaddr, int proto)
+pmap_create(char *hostname, struct sockaddr_in *srvaddr, int proto, int privileged)
{
struct rpc_xprt *xprt;
struct rpc_clnt *clnt;
@@ -208,6 +208,8 @@
if (IS_ERR(xprt))
return (struct rpc_clnt *)xprt;
xprt->addr.sin_port = htons(RPC_PMAP_PORT);
+ if (!privileged)
+ xprt->resvport = 0;
/* printk("pmap: create clnt\n"); */
clnt = rpc_new_client(xprt, hostname,
diff --git a/net/sunrpc/socklib.c b/net/sunrpc/socklib.c
new file mode 100644
index 0000000..8f97e90
--- /dev/null
+++ b/net/sunrpc/socklib.c
@@ -0,0 +1,175 @@
+/*
+ * linux/net/sunrpc/socklib.c
+ *
+ * Common socket helper routines for RPC client and server
+ *
+ * Copyright (C) 1995, 1996 Olaf Kirch <okir@monad.swb.de>
+ */
+
+#include <linux/types.h>
+#include <linux/pagemap.h>
+#include <linux/udp.h>
+#include <linux/sunrpc/xdr.h>
+
+
+/**
+ * skb_read_bits - copy some data bits from skb to internal buffer
+ * @desc: sk_buff copy helper
+ * @to: copy destination
+ * @len: number of bytes to copy
+ *
+ * Possibly called several times to iterate over an sk_buff and copy
+ * data out of it.
+ */
+static size_t skb_read_bits(skb_reader_t *desc, void *to, size_t len)
+{
+ if (len > desc->count)
+ len = desc->count;
+ if (skb_copy_bits(desc->skb, desc->offset, to, len))
+ return 0;
+ desc->count -= len;
+ desc->offset += len;
+ return len;
+}
+
+/**
+ * skb_read_and_csum_bits - copy and checksum from skb to buffer
+ * @desc: sk_buff copy helper
+ * @to: copy destination
+ * @len: number of bytes to copy
+ *
+ * Same as skb_read_bits, but calculate a checksum at the same time.
+ */
+static size_t skb_read_and_csum_bits(skb_reader_t *desc, void *to, size_t len)
+{
+ unsigned int csum2, pos;
+
+ if (len > desc->count)
+ len = desc->count;
+ pos = desc->offset;
+ csum2 = skb_copy_and_csum_bits(desc->skb, pos, to, len, 0);
+ desc->csum = csum_block_add(desc->csum, csum2, pos);
+ desc->count -= len;
+ desc->offset += len;
+ return len;
+}
+
+/**
+ * xdr_partial_copy_from_skb - copy data out of an skb
+ * @xdr: target XDR buffer
+ * @base: starting offset
+ * @desc: sk_buff copy helper
+ * @copy_actor: virtual method for copying data
+ *
+ */
+ssize_t xdr_partial_copy_from_skb(struct xdr_buf *xdr, unsigned int base, skb_reader_t *desc, skb_read_actor_t copy_actor)
+{
+ struct page **ppage = xdr->pages;
+ unsigned int len, pglen = xdr->page_len;
+ ssize_t copied = 0;
+ int ret;
+
+ len = xdr->head[0].iov_len;
+ if (base < len) {
+ len -= base;
+ ret = copy_actor(desc, (char *)xdr->head[0].iov_base + base, len);
+ copied += ret;
+ if (ret != len || !desc->count)
+ goto out;
+ base = 0;
+ } else
+ base -= len;
+
+ if (unlikely(pglen == 0))
+ goto copy_tail;
+ if (unlikely(base >= pglen)) {
+ base -= pglen;
+ goto copy_tail;
+ }
+ if (base || xdr->page_base) {
+ pglen -= base;
+ base += xdr->page_base;
+ ppage += base >> PAGE_CACHE_SHIFT;
+ base &= ~PAGE_CACHE_MASK;
+ }
+ do {
+ char *kaddr;
+
+ /* ACL likes to be lazy in allocating pages - ACLs
+ * are small by default but can get huge. */
+ if (unlikely(*ppage == NULL)) {
+ *ppage = alloc_page(GFP_ATOMIC);
+ if (unlikely(*ppage == NULL)) {
+ if (copied == 0)
+ copied = -ENOMEM;
+ goto out;
+ }
+ }
+
+ len = PAGE_CACHE_SIZE;
+ kaddr = kmap_atomic(*ppage, KM_SKB_SUNRPC_DATA);
+ if (base) {
+ len -= base;
+ if (pglen < len)
+ len = pglen;
+ ret = copy_actor(desc, kaddr + base, len);
+ base = 0;
+ } else {
+ if (pglen < len)
+ len = pglen;
+ ret = copy_actor(desc, kaddr, len);
+ }
+ flush_dcache_page(*ppage);
+ kunmap_atomic(kaddr, KM_SKB_SUNRPC_DATA);
+ copied += ret;
+ if (ret != len || !desc->count)
+ goto out;
+ ppage++;
+ } while ((pglen -= len) != 0);
+copy_tail:
+ len = xdr->tail[0].iov_len;
+ if (base < len)
+ copied += copy_actor(desc, (char *)xdr->tail[0].iov_base + base, len - base);
+out:
+ return copied;
+}
+
+/**
+ * csum_partial_copy_to_xdr - checksum and copy data
+ * @xdr: target XDR buffer
+ * @skb: source skb
+ *
+ * We have set things up such that we perform the checksum of the UDP
+ * packet in parallel with the copies into the RPC client iovec. -DaveM
+ */
+int csum_partial_copy_to_xdr(struct xdr_buf *xdr, struct sk_buff *skb)
+{
+ skb_reader_t desc;
+
+ desc.skb = skb;
+ desc.offset = sizeof(struct udphdr);
+ desc.count = skb->len - desc.offset;
+
+ if (skb->ip_summed == CHECKSUM_UNNECESSARY)
+ goto no_checksum;
+
+ desc.csum = csum_partial(skb->data, desc.offset, skb->csum);
+ if (xdr_partial_copy_from_skb(xdr, 0, &desc, skb_read_and_csum_bits) < 0)
+ return -1;
+ if (desc.offset != skb->len) {
+ unsigned int csum2;
+ csum2 = skb_checksum(skb, desc.offset, skb->len - desc.offset, 0);
+ desc.csum = csum_block_add(desc.csum, csum2, desc.offset);
+ }
+ if (desc.count)
+ return -1;
+ if ((unsigned short)csum_fold(desc.csum))
+ return -1;
+ return 0;
+no_checksum:
+ if (xdr_partial_copy_from_skb(xdr, 0, &desc, skb_read_bits) < 0)
+ return -1;
+ if (desc.count)
+ return -1;
+ return 0;
+}
diff --git a/net/sunrpc/sunrpc_syms.c b/net/sunrpc/sunrpc_syms.c
index ed48ff0..2387e7b 100644
--- a/net/sunrpc/sunrpc_syms.c
+++ b/net/sunrpc/sunrpc_syms.c
@@ -10,7 +10,6 @@
#include <linux/module.h>
#include <linux/types.h>
-#include <linux/socket.h>
#include <linux/sched.h>
#include <linux/uio.h>
#include <linux/unistd.h>
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 691dea4..f16e7cd 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -548,9 +548,6 @@
/*
* Receive a datagram from a UDP socket.
*/
-extern int
-csum_partial_copy_to_xdr(struct xdr_buf *xdr, struct sk_buff *skb);
-
static int
svc_udp_recvfrom(struct svc_rqst *rqstp)
{
diff --git a/net/sunrpc/sysctl.c b/net/sunrpc/sysctl.c
index 1b9616a..d0c9f46 100644
--- a/net/sunrpc/sysctl.c
+++ b/net/sunrpc/sysctl.c
@@ -119,8 +119,18 @@
return 0;
}
+unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE;
+unsigned int xprt_tcp_slot_table_entries = RPC_DEF_SLOT_TABLE;
+unsigned int xprt_min_resvport = RPC_DEF_MIN_RESVPORT;
+EXPORT_SYMBOL(xprt_min_resvport);
+unsigned int xprt_max_resvport = RPC_DEF_MAX_RESVPORT;
+EXPORT_SYMBOL(xprt_max_resvport);
+
+
static unsigned int min_slot_table_size = RPC_MIN_SLOT_TABLE;
static unsigned int max_slot_table_size = RPC_MAX_SLOT_TABLE;
+static unsigned int xprt_min_resvport_limit = RPC_MIN_RESVPORT;
+static unsigned int xprt_max_resvport_limit = RPC_MAX_RESVPORT;
static ctl_table debug_table[] = {
{
@@ -177,6 +187,28 @@
.extra1 = &min_slot_table_size,
.extra2 = &max_slot_table_size
},
+ {
+ .ctl_name = CTL_MIN_RESVPORT,
+ .procname = "min_resvport",
+ .data = &xprt_min_resvport,
+ .maxlen = sizeof(unsigned int),
+ .mode = 0644,
+ .proc_handler = &proc_dointvec_minmax,
+ .strategy = &sysctl_intvec,
+ .extra1 = &xprt_min_resvport_limit,
+ .extra2 = &xprt_max_resvport_limit
+ },
+ {
+ .ctl_name = CTL_MAX_RESVPORT,
+ .procname = "max_resvport",
+ .data = &xprt_max_resvport,
+ .maxlen = sizeof(unsigned int),
+ .mode = 0644,
+ .proc_handler = &proc_dointvec_minmax,
+ .strategy = &sysctl_intvec,
+ .extra1 = &xprt_min_resvport_limit,
+ .extra2 = &xprt_max_resvport_limit
+ },
{ .ctl_name = 0 }
};
diff --git a/net/sunrpc/xdr.c b/net/sunrpc/xdr.c
index fde16f4..32df433 100644
--- a/net/sunrpc/xdr.c
+++ b/net/sunrpc/xdr.c
@@ -6,15 +6,12 @@
* Copyright (C) 1995, 1996 Olaf Kirch <okir@monad.swb.de>
*/
+#include <linux/module.h>
#include <linux/types.h>
-#include <linux/socket.h>
#include <linux/string.h>
#include <linux/kernel.h>
#include <linux/pagemap.h>
#include <linux/errno.h>
-#include <linux/in.h>
-#include <linux/net.h>
-#include <net/sock.h>
#include <linux/sunrpc/xdr.h>
#include <linux/sunrpc/msg_prot.h>
@@ -176,178 +173,6 @@
xdr->buflen += len;
}
-ssize_t
-xdr_partial_copy_from_skb(struct xdr_buf *xdr, unsigned int base,
- skb_reader_t *desc,
- skb_read_actor_t copy_actor)
-{
- struct page **ppage = xdr->pages;
- unsigned int len, pglen = xdr->page_len;
- ssize_t copied = 0;
- int ret;
-
- len = xdr->head[0].iov_len;
- if (base < len) {
- len -= base;
- ret = copy_actor(desc, (char *)xdr->head[0].iov_base + base, len);
- copied += ret;
- if (ret != len || !desc->count)
- goto out;
- base = 0;
- } else
- base -= len;
-
- if (pglen == 0)
- goto copy_tail;
- if (base >= pglen) {
- base -= pglen;
- goto copy_tail;
- }
- if (base || xdr->page_base) {
- pglen -= base;
- base += xdr->page_base;
- ppage += base >> PAGE_CACHE_SHIFT;
- base &= ~PAGE_CACHE_MASK;
- }
- do {
- char *kaddr;
-
- /* ACL likes to be lazy in allocating pages - ACLs
- * are small by default but can get huge. */
- if (unlikely(*ppage == NULL)) {
- *ppage = alloc_page(GFP_ATOMIC);
- if (unlikely(*ppage == NULL)) {
- if (copied == 0)
- copied = -ENOMEM;
- goto out;
- }
- }
-
- len = PAGE_CACHE_SIZE;
- kaddr = kmap_atomic(*ppage, KM_SKB_SUNRPC_DATA);
- if (base) {
- len -= base;
- if (pglen < len)
- len = pglen;
- ret = copy_actor(desc, kaddr + base, len);
- base = 0;
- } else {
- if (pglen < len)
- len = pglen;
- ret = copy_actor(desc, kaddr, len);
- }
- flush_dcache_page(*ppage);
- kunmap_atomic(kaddr, KM_SKB_SUNRPC_DATA);
- copied += ret;
- if (ret != len || !desc->count)
- goto out;
- ppage++;
- } while ((pglen -= len) != 0);
-copy_tail:
- len = xdr->tail[0].iov_len;
- if (base < len)
- copied += copy_actor(desc, (char *)xdr->tail[0].iov_base + base, len - base);
-out:
- return copied;
-}
-
-
-int
-xdr_sendpages(struct socket *sock, struct sockaddr *addr, int addrlen,
- struct xdr_buf *xdr, unsigned int base, int msgflags)
-{
- struct page **ppage = xdr->pages;
- unsigned int len, pglen = xdr->page_len;
- int err, ret = 0;
- ssize_t (*sendpage)(struct socket *, struct page *, int, size_t, int);
-
- len = xdr->head[0].iov_len;
- if (base < len || (addr != NULL && base == 0)) {
- struct kvec iov = {
- .iov_base = xdr->head[0].iov_base + base,
- .iov_len = len - base,
- };
- struct msghdr msg = {
- .msg_name = addr,
- .msg_namelen = addrlen,
- .msg_flags = msgflags,
- };
- if (xdr->len > len)
- msg.msg_flags |= MSG_MORE;
-
- if (iov.iov_len != 0)
- err = kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len);
- else
- err = kernel_sendmsg(sock, &msg, NULL, 0, 0);
- if (ret == 0)
- ret = err;
- else if (err > 0)
- ret += err;
- if (err != iov.iov_len)
- goto out;
- base = 0;
- } else
- base -= len;
-
- if (pglen == 0)
- goto copy_tail;
- if (base >= pglen) {
- base -= pglen;
- goto copy_tail;
- }
- if (base || xdr->page_base) {
- pglen -= base;
- base += xdr->page_base;
- ppage += base >> PAGE_CACHE_SHIFT;
- base &= ~PAGE_CACHE_MASK;
- }
-
- sendpage = sock->ops->sendpage ? : sock_no_sendpage;
- do {
- int flags = msgflags;
-
- len = PAGE_CACHE_SIZE;
- if (base)
- len -= base;
- if (pglen < len)
- len = pglen;
-
- if (pglen != len || xdr->tail[0].iov_len != 0)
- flags |= MSG_MORE;
-
- /* Hmm... We might be dealing with highmem pages */
- if (PageHighMem(*ppage))
- sendpage = sock_no_sendpage;
- err = sendpage(sock, *ppage, base, len, flags);
- if (ret == 0)
- ret = err;
- else if (err > 0)
- ret += err;
- if (err != len)
- goto out;
- base = 0;
- ppage++;
- } while ((pglen -= len) != 0);
-copy_tail:
- len = xdr->tail[0].iov_len;
- if (base < len) {
- struct kvec iov = {
- .iov_base = xdr->tail[0].iov_base + base,
- .iov_len = len - base,
- };
- struct msghdr msg = {
- .msg_flags = msgflags,
- };
- err = kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len);
- if (ret == 0)
- ret = err;
- else if (err > 0)
- ret += err;
- }
-out:
- return ret;
-}
-
/*
* Helper routines for doing 'memmove' like operations on a struct xdr_buf
diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c
index 3c654e0..6dda386 100644
--- a/net/sunrpc/xprt.c
+++ b/net/sunrpc/xprt.c
@@ -10,12 +10,12 @@
* one is available. Otherwise, it sleeps on the backlog queue
* (xprt_reserve).
* - Next, the caller puts together the RPC message, stuffs it into
- * the request struct, and calls xprt_call().
- * - xprt_call transmits the message and installs the caller on the
- * socket's wait list. At the same time, it installs a timer that
+ * the request struct, and calls xprt_transmit().
+ * - xprt_transmit sends the message and installs the caller on the
+ * transport's wait list. At the same time, it installs a timer that
* is run after the packet's timeout has expired.
* - When a packet arrives, the data_ready handler walks the list of
- * pending requests for that socket. If a matching XID is found, the
+ * pending requests for that transport. If a matching XID is found, the
* caller is woken up, and the timer removed.
* - When no reply arrives within the timeout interval, the timer is
* fired by the kernel and runs xprt_timer(). It either adjusts the
@@ -33,36 +33,17 @@
*
* Copyright (C) 1995-1997, Olaf Kirch <okir@monad.swb.de>
*
- * TCP callback races fixes (C) 1998 Red Hat Software <alan@redhat.com>
- * TCP send fixes (C) 1998 Red Hat Software <alan@redhat.com>
- * TCP NFS related read + write fixes
- * (C) 1999 Dave Airlie, University of Limerick, Ireland <airlied@linux.ie>
- *
- * Rewrite of larges part of the code in order to stabilize TCP stuff.
- * Fix behaviour when socket buffer is full.
- * (C) 1999 Trond Myklebust <trond.myklebust@fys.uio.no>
+ * Transport switch API copyright (C) 2005, Chuck Lever <cel@netapp.com>
*/
+#include <linux/module.h>
+
#include <linux/types.h>
-#include <linux/slab.h>
-#include <linux/capability.h>
-#include <linux/sched.h>
-#include <linux/errno.h>
-#include <linux/socket.h>
-#include <linux/in.h>
-#include <linux/net.h>
-#include <linux/mm.h>
-#include <linux/udp.h>
-#include <linux/tcp.h>
-#include <linux/sunrpc/clnt.h>
-#include <linux/file.h>
+#include <linux/interrupt.h>
#include <linux/workqueue.h>
#include <linux/random.h>
-#include <net/sock.h>
-#include <net/checksum.h>
-#include <net/udp.h>
-#include <net/tcp.h>
+#include <linux/sunrpc/clnt.h>
/*
* Local variables
@@ -73,93 +54,62 @@
# define RPCDBG_FACILITY RPCDBG_XPRT
#endif
-#define XPRT_MAX_BACKOFF (8)
-#define XPRT_IDLE_TIMEOUT (5*60*HZ)
-#define XPRT_MAX_RESVPORT (800)
-
/*
* Local functions
*/
static void xprt_request_init(struct rpc_task *, struct rpc_xprt *);
static inline void do_xprt_reserve(struct rpc_task *);
-static void xprt_disconnect(struct rpc_xprt *);
static void xprt_connect_status(struct rpc_task *task);
-static struct rpc_xprt * xprt_setup(int proto, struct sockaddr_in *ap,
- struct rpc_timeout *to);
-static struct socket *xprt_create_socket(struct rpc_xprt *, int, int);
-static void xprt_bind_socket(struct rpc_xprt *, struct socket *);
static int __xprt_get_cong(struct rpc_xprt *, struct rpc_task *);
-static int xprt_clear_backlog(struct rpc_xprt *xprt);
-
-#ifdef RPC_DEBUG_DATA
/*
- * Print the buffer contents (first 128 bytes only--just enough for
- * diropres return).
+ * The transport code maintains an estimate on the maximum number of out-
+ * standing RPC requests, using a smoothed version of the congestion
+ * avoidance implemented in 44BSD. This is basically the Van Jacobson
+ * congestion algorithm: If a retransmit occurs, the congestion window is
+ * halved; otherwise, it is incremented by 1/cwnd when
+ *
+ * - a reply is received and
+ * - a full number of requests are outstanding and
+ * - the congestion window hasn't been updated recently.
*/
-static void
-xprt_pktdump(char *msg, u32 *packet, unsigned int count)
-{
- u8 *buf = (u8 *) packet;
- int j;
+#define RPC_CWNDSHIFT (8U)
+#define RPC_CWNDSCALE (1U << RPC_CWNDSHIFT)
+#define RPC_INITCWND RPC_CWNDSCALE
+#define RPC_MAXCWND(xprt) ((xprt)->max_reqs << RPC_CWNDSHIFT)
- dprintk("RPC: %s\n", msg);
- for (j = 0; j < count && j < 128; j += 4) {
- if (!(j & 31)) {
- if (j)
- dprintk("\n");
- dprintk("0x%04x ", j);
- }
- dprintk("%02x%02x%02x%02x ",
- buf[j], buf[j+1], buf[j+2], buf[j+3]);
- }
- dprintk("\n");
-}
-#else
-static inline void
-xprt_pktdump(char *msg, u32 *packet, unsigned int count)
-{
- /* NOP */
-}
-#endif
+#define RPCXPRT_CONGESTED(xprt) ((xprt)->cong >= (xprt)->cwnd)
-/*
- * Look up RPC transport given an INET socket
+/**
+ * xprt_reserve_xprt - serialize write access to transports
+ * @task: task that is requesting access to the transport
+ *
+ * This prevents mixing the payload of separate requests, and prevents
+ * transport connects from colliding with writes. No congestion control
+ * is provided.
*/
-static inline struct rpc_xprt *
-xprt_from_sock(struct sock *sk)
+int xprt_reserve_xprt(struct rpc_task *task)
{
- return (struct rpc_xprt *) sk->sk_user_data;
-}
-
-/*
- * Serialize write access to sockets, in order to prevent different
- * requests from interfering with each other.
- * Also prevents TCP socket connects from colliding with writes.
- */
-static int
-__xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task)
-{
+ struct rpc_xprt *xprt = task->tk_xprt;
struct rpc_rqst *req = task->tk_rqstp;
- if (test_and_set_bit(XPRT_LOCKED, &xprt->sockstate)) {
+ if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) {
if (task == xprt->snd_task)
return 1;
+ if (task == NULL)
+ return 0;
goto out_sleep;
}
- if (xprt->nocong || __xprt_get_cong(xprt, task)) {
- xprt->snd_task = task;
- if (req) {
- req->rq_bytes_sent = 0;
- req->rq_ntrans++;
- }
- return 1;
+ xprt->snd_task = task;
+ if (req) {
+ req->rq_bytes_sent = 0;
+ req->rq_ntrans++;
}
- smp_mb__before_clear_bit();
- clear_bit(XPRT_LOCKED, &xprt->sockstate);
- smp_mb__after_clear_bit();
+ return 1;
+
out_sleep:
- dprintk("RPC: %4d failed to lock socket %p\n", task->tk_pid, xprt);
+ dprintk("RPC: %4d failed to lock transport %p\n",
+ task->tk_pid, xprt);
task->tk_timeout = 0;
task->tk_status = -EAGAIN;
if (req && req->rq_ntrans)
@@ -169,26 +119,92 @@
return 0;
}
-static inline int
-xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task)
+/*
+ * xprt_reserve_xprt_cong - serialize write access to transports
+ * @task: task that is requesting access to the transport
+ *
+ * Same as xprt_reserve_xprt, but Van Jacobson congestion control is
+ * integrated into the decision of whether a request is allowed to be
+ * woken up and given access to the transport.
+ */
+int xprt_reserve_xprt_cong(struct rpc_task *task)
+{
+ struct rpc_xprt *xprt = task->tk_xprt;
+ struct rpc_rqst *req = task->tk_rqstp;
+
+ if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) {
+ if (task == xprt->snd_task)
+ return 1;
+ goto out_sleep;
+ }
+ if (__xprt_get_cong(xprt, task)) {
+ xprt->snd_task = task;
+ if (req) {
+ req->rq_bytes_sent = 0;
+ req->rq_ntrans++;
+ }
+ return 1;
+ }
+ smp_mb__before_clear_bit();
+ clear_bit(XPRT_LOCKED, &xprt->state);
+ smp_mb__after_clear_bit();
+out_sleep:
+ dprintk("RPC: %4d failed to lock transport %p\n", task->tk_pid, xprt);
+ task->tk_timeout = 0;
+ task->tk_status = -EAGAIN;
+ if (req && req->rq_ntrans)
+ rpc_sleep_on(&xprt->resend, task, NULL, NULL);
+ else
+ rpc_sleep_on(&xprt->sending, task, NULL, NULL);
+ return 0;
+}
+
+static inline int xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task)
{
int retval;
- spin_lock_bh(&xprt->sock_lock);
- retval = __xprt_lock_write(xprt, task);
- spin_unlock_bh(&xprt->sock_lock);
+ spin_lock_bh(&xprt->transport_lock);
+ retval = xprt->ops->reserve_xprt(task);
+ spin_unlock_bh(&xprt->transport_lock);
return retval;
}
+static void __xprt_lock_write_next(struct rpc_xprt *xprt)
+{
+ struct rpc_task *task;
+ struct rpc_rqst *req;
-static void
-__xprt_lock_write_next(struct rpc_xprt *xprt)
+ if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
+ return;
+
+ task = rpc_wake_up_next(&xprt->resend);
+ if (!task) {
+ task = rpc_wake_up_next(&xprt->sending);
+ if (!task)
+ goto out_unlock;
+ }
+
+ req = task->tk_rqstp;
+ xprt->snd_task = task;
+ if (req) {
+ req->rq_bytes_sent = 0;
+ req->rq_ntrans++;
+ }
+ return;
+
+out_unlock:
+ smp_mb__before_clear_bit();
+ clear_bit(XPRT_LOCKED, &xprt->state);
+ smp_mb__after_clear_bit();
+}
+
+static void __xprt_lock_write_next_cong(struct rpc_xprt *xprt)
{
struct rpc_task *task;
- if (test_and_set_bit(XPRT_LOCKED, &xprt->sockstate))
+ if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
return;
- if (!xprt->nocong && RPCXPRT_CONGESTED(xprt))
+ if (RPCXPRT_CONGESTED(xprt))
goto out_unlock;
task = rpc_wake_up_next(&xprt->resend);
if (!task) {
@@ -196,7 +212,7 @@
if (!task)
goto out_unlock;
}
- if (xprt->nocong || __xprt_get_cong(xprt, task)) {
+ if (__xprt_get_cong(xprt, task)) {
struct rpc_rqst *req = task->tk_rqstp;
xprt->snd_task = task;
if (req) {
@@ -207,87 +223,52 @@
}
out_unlock:
smp_mb__before_clear_bit();
- clear_bit(XPRT_LOCKED, &xprt->sockstate);
+ clear_bit(XPRT_LOCKED, &xprt->state);
smp_mb__after_clear_bit();
}
-/*
- * Releases the socket for use by other requests.
+/**
+ * xprt_release_xprt - allow other requests to use a transport
+ * @xprt: transport with other tasks potentially waiting
+ * @task: task that is releasing access to the transport
+ *
+ * Note that "task" can be NULL. No congestion control is provided.
*/
-static void
-__xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task)
+void xprt_release_xprt(struct rpc_xprt *xprt, struct rpc_task *task)
{
if (xprt->snd_task == task) {
xprt->snd_task = NULL;
smp_mb__before_clear_bit();
- clear_bit(XPRT_LOCKED, &xprt->sockstate);
+ clear_bit(XPRT_LOCKED, &xprt->state);
smp_mb__after_clear_bit();
__xprt_lock_write_next(xprt);
}
}
-static inline void
-xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task)
+/**
+ * xprt_release_xprt_cong - allow other requests to use a transport
+ * @xprt: transport with other tasks potentially waiting
+ * @task: task that is releasing access to the transport
+ *
+ * Note that "task" can be NULL. Another task is awoken to use the
+ * transport if the transport's congestion window allows it.
+ */
+void xprt_release_xprt_cong(struct rpc_xprt *xprt, struct rpc_task *task)
{
- spin_lock_bh(&xprt->sock_lock);
- __xprt_release_write(xprt, task);
- spin_unlock_bh(&xprt->sock_lock);
+ if (xprt->snd_task == task) {
+ xprt->snd_task = NULL;
+ smp_mb__before_clear_bit();
+ clear_bit(XPRT_LOCKED, &xprt->state);
+ smp_mb__after_clear_bit();
+ __xprt_lock_write_next_cong(xprt);
+ }
}
-/*
- * Write data to socket.
- */
-static inline int
-xprt_sendmsg(struct rpc_xprt *xprt, struct rpc_rqst *req)
+static inline void xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task)
{
- struct socket *sock = xprt->sock;
- struct xdr_buf *xdr = &req->rq_snd_buf;
- struct sockaddr *addr = NULL;
- int addrlen = 0;
- unsigned int skip;
- int result;
-
- if (!sock)
- return -ENOTCONN;
-
- xprt_pktdump("packet data:",
- req->rq_svec->iov_base,
- req->rq_svec->iov_len);
-
- /* For UDP, we need to provide an address */
- if (!xprt->stream) {
- addr = (struct sockaddr *) &xprt->addr;
- addrlen = sizeof(xprt->addr);
- }
- /* Dont repeat bytes */
- skip = req->rq_bytes_sent;
-
- clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags);
- result = xdr_sendpages(sock, addr, addrlen, xdr, skip, MSG_DONTWAIT);
-
- dprintk("RPC: xprt_sendmsg(%d) = %d\n", xdr->len - skip, result);
-
- if (result >= 0)
- return result;
-
- switch (result) {
- case -ECONNREFUSED:
- /* When the server has died, an ICMP port unreachable message
- * prompts ECONNREFUSED.
- */
- case -EAGAIN:
- break;
- case -ECONNRESET:
- case -ENOTCONN:
- case -EPIPE:
- /* connection broken */
- if (xprt->stream)
- result = -ENOTCONN;
- break;
- default:
- printk(KERN_NOTICE "RPC: sendmsg returned error %d\n", -result);
- }
- return result;
+ spin_lock_bh(&xprt->transport_lock);
+ xprt->ops->release_xprt(xprt, task);
+ spin_unlock_bh(&xprt->transport_lock);
}
/*
@@ -321,26 +302,40 @@
return;
req->rq_cong = 0;
xprt->cong -= RPC_CWNDSCALE;
- __xprt_lock_write_next(xprt);
+ __xprt_lock_write_next_cong(xprt);
}
-/*
- * Adjust RPC congestion window
+/**
+ * xprt_release_rqst_cong - housekeeping when request is complete
+ * @task: RPC request that recently completed
+ *
+ * Useful for transports that require congestion control.
+ */
+void xprt_release_rqst_cong(struct rpc_task *task)
+{
+ __xprt_put_cong(task->tk_xprt, task->tk_rqstp);
+}
+
+/**
+ * xprt_adjust_cwnd - adjust transport congestion window
+ * @task: recently completed RPC request used to adjust window
+ * @result: result code of completed RPC request
+ *
* We use a time-smoothed congestion estimator to avoid heavy oscillation.
*/
-static void
-xprt_adjust_cwnd(struct rpc_xprt *xprt, int result)
+void xprt_adjust_cwnd(struct rpc_task *task, int result)
{
- unsigned long cwnd;
+ struct rpc_rqst *req = task->tk_rqstp;
+ struct rpc_xprt *xprt = task->tk_xprt;
+ unsigned long cwnd = xprt->cwnd;
- cwnd = xprt->cwnd;
if (result >= 0 && cwnd <= xprt->cong) {
/* The (cwnd >> 1) term makes sure
* the result gets rounded properly. */
cwnd += (RPC_CWNDSCALE * RPC_CWNDSCALE + (cwnd >> 1)) / cwnd;
if (cwnd > RPC_MAXCWND(xprt))
cwnd = RPC_MAXCWND(xprt);
- __xprt_lock_write_next(xprt);
+ __xprt_lock_write_next_cong(xprt);
} else if (result == -ETIMEDOUT) {
cwnd >>= 1;
if (cwnd < RPC_CWNDSCALE)
@@ -349,11 +344,89 @@
dprintk("RPC: cong %ld, cwnd was %ld, now %ld\n",
xprt->cong, xprt->cwnd, cwnd);
xprt->cwnd = cwnd;
+ __xprt_put_cong(xprt, req);
+}
+
+/**
+ * xprt_wake_pending_tasks - wake all tasks on a transport's pending queue
+ * @xprt: transport with waiting tasks
+ * @status: result code to plant in each task before waking it
+ *
+ */
+void xprt_wake_pending_tasks(struct rpc_xprt *xprt, int status)
+{
+ if (status < 0)
+ rpc_wake_up_status(&xprt->pending, status);
+ else
+ rpc_wake_up(&xprt->pending);
+}
+
+/**
+ * xprt_wait_for_buffer_space - wait for transport output buffer to clear
+ * @task: task to be put to sleep
+ *
+ */
+void xprt_wait_for_buffer_space(struct rpc_task *task)
+{
+ struct rpc_rqst *req = task->tk_rqstp;
+ struct rpc_xprt *xprt = req->rq_xprt;
+
+ task->tk_timeout = req->rq_timeout;
+ rpc_sleep_on(&xprt->pending, task, NULL, NULL);
+}
+
+/**
+ * xprt_write_space - wake the task waiting for transport output buffer space
+ * @xprt: transport with waiting tasks
+ *
+ * Can be called in a soft IRQ context, so xprt_write_space never sleeps.
+ */
+void xprt_write_space(struct rpc_xprt *xprt)
+{
+ if (unlikely(xprt->shutdown))
+ return;
+
+ spin_lock_bh(&xprt->transport_lock);
+ if (xprt->snd_task) {
+ dprintk("RPC: write space: waking waiting task on xprt %p\n",
+ xprt);
+ rpc_wake_up_task(xprt->snd_task);
+ }
+ spin_unlock_bh(&xprt->transport_lock);
+}
+
+/**
+ * xprt_set_retrans_timeout_def - set a request's retransmit timeout
+ * @task: task whose timeout is to be set
+ *
+ * Set a request's retransmit timeout based on the transport's
+ * default timeout parameters. Used by transports that don't adjust
+ * the retransmit timeout based on round-trip time estimation.
+ */
+void xprt_set_retrans_timeout_def(struct rpc_task *task)
+{
+ task->tk_timeout = task->tk_rqstp->rq_timeout;
}
/*
- * Reset the major timeout value
+ * xprt_set_retrans_timeout_rtt - set a request's retransmit timeout
+ * @task: task whose timeout is to be set
+ *
+ * Set a request's retransmit timeout using the RTT estimator.
*/
+void xprt_set_retrans_timeout_rtt(struct rpc_task *task)
+{
+ int timer = task->tk_msg.rpc_proc->p_timer;
+ struct rpc_rtt *rtt = task->tk_client->cl_rtt;
+ struct rpc_rqst *req = task->tk_rqstp;
+ unsigned long max_timeout = req->rq_xprt->timeout.to_maxval;
+
+ task->tk_timeout = rpc_calc_rto(rtt, timer);
+ task->tk_timeout <<= rpc_ntimeo(rtt, timer) + req->rq_retries;
+ if (task->tk_timeout > max_timeout || task->tk_timeout == 0)
+ task->tk_timeout = max_timeout;
+}
+
static void xprt_reset_majortimeo(struct rpc_rqst *req)
{
struct rpc_timeout *to = &req->rq_xprt->timeout;
@@ -368,8 +441,10 @@
req->rq_majortimeo += jiffies;
}
-/*
- * Adjust timeout values etc for next retransmit
+/**
+ * xprt_adjust_timeout - adjust timeout values for next retransmit
+ * @req: RPC request containing parameters to use for the adjustment
+ *
*/
int xprt_adjust_timeout(struct rpc_rqst *req)
{
@@ -391,9 +466,9 @@
req->rq_retries = 0;
xprt_reset_majortimeo(req);
/* Reset the RTT counters == "slow start" */
- spin_lock_bh(&xprt->sock_lock);
+ spin_lock_bh(&xprt->transport_lock);
rpc_init_rtt(req->rq_task->tk_client->cl_rtt, to->to_initval);
- spin_unlock_bh(&xprt->sock_lock);
+ spin_unlock_bh(&xprt->transport_lock);
pprintk("RPC: %lu timeout\n", jiffies);
status = -ETIMEDOUT;
}
@@ -405,133 +480,52 @@
return status;
}
-/*
- * Close down a transport socket
- */
-static void
-xprt_close(struct rpc_xprt *xprt)
-{
- struct socket *sock = xprt->sock;
- struct sock *sk = xprt->inet;
-
- if (!sk)
- return;
-
- write_lock_bh(&sk->sk_callback_lock);
- xprt->inet = NULL;
- xprt->sock = NULL;
-
- sk->sk_user_data = NULL;
- sk->sk_data_ready = xprt->old_data_ready;
- sk->sk_state_change = xprt->old_state_change;
- sk->sk_write_space = xprt->old_write_space;
- write_unlock_bh(&sk->sk_callback_lock);
-
- sk->sk_no_check = 0;
-
- sock_release(sock);
-}
-
-static void
-xprt_socket_autoclose(void *args)
+static void xprt_autoclose(void *args)
{
struct rpc_xprt *xprt = (struct rpc_xprt *)args;
xprt_disconnect(xprt);
- xprt_close(xprt);
+ xprt->ops->close(xprt);
xprt_release_write(xprt, NULL);
}
-/*
- * Mark a transport as disconnected
+/**
+ * xprt_disconnect - mark a transport as disconnected
+ * @xprt: transport to flag for disconnect
+ *
*/
-static void
-xprt_disconnect(struct rpc_xprt *xprt)
+void xprt_disconnect(struct rpc_xprt *xprt)
{
dprintk("RPC: disconnected transport %p\n", xprt);
- spin_lock_bh(&xprt->sock_lock);
+ spin_lock_bh(&xprt->transport_lock);
xprt_clear_connected(xprt);
- rpc_wake_up_status(&xprt->pending, -ENOTCONN);
- spin_unlock_bh(&xprt->sock_lock);
+ xprt_wake_pending_tasks(xprt, -ENOTCONN);
+ spin_unlock_bh(&xprt->transport_lock);
}
-/*
- * Used to allow disconnection when we've been idle
- */
static void
xprt_init_autodisconnect(unsigned long data)
{
struct rpc_xprt *xprt = (struct rpc_xprt *)data;
- spin_lock(&xprt->sock_lock);
+ spin_lock(&xprt->transport_lock);
if (!list_empty(&xprt->recv) || xprt->shutdown)
goto out_abort;
- if (test_and_set_bit(XPRT_LOCKED, &xprt->sockstate))
+ if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
goto out_abort;
- spin_unlock(&xprt->sock_lock);
- /* Let keventd close the socket */
- if (test_bit(XPRT_CONNECTING, &xprt->sockstate) != 0)
+ spin_unlock(&xprt->transport_lock);
+ if (xprt_connecting(xprt))
xprt_release_write(xprt, NULL);
else
schedule_work(&xprt->task_cleanup);
return;
out_abort:
- spin_unlock(&xprt->sock_lock);
+ spin_unlock(&xprt->transport_lock);
}
-static void xprt_socket_connect(void *args)
-{
- struct rpc_xprt *xprt = (struct rpc_xprt *)args;
- struct socket *sock = xprt->sock;
- int status = -EIO;
-
- if (xprt->shutdown || xprt->addr.sin_port == 0)
- goto out;
-
- /*
- * Start by resetting any existing state
- */
- xprt_close(xprt);
- sock = xprt_create_socket(xprt, xprt->prot, xprt->resvport);
- if (sock == NULL) {
- /* couldn't create socket or bind to reserved port;
- * this is likely a permanent error, so cause an abort */
- goto out;
- }
- xprt_bind_socket(xprt, sock);
- xprt_sock_setbufsize(xprt);
-
- status = 0;
- if (!xprt->stream)
- goto out;
-
- /*
- * Tell the socket layer to start connecting...
- */
- status = sock->ops->connect(sock, (struct sockaddr *) &xprt->addr,
- sizeof(xprt->addr), O_NONBLOCK);
- dprintk("RPC: %p connect status %d connected %d sock state %d\n",
- xprt, -status, xprt_connected(xprt), sock->sk->sk_state);
- if (status < 0) {
- switch (status) {
- case -EINPROGRESS:
- case -EALREADY:
- goto out_clear;
- }
- }
-out:
- if (status < 0)
- rpc_wake_up_status(&xprt->pending, status);
- else
- rpc_wake_up(&xprt->pending);
-out_clear:
- smp_mb__before_clear_bit();
- clear_bit(XPRT_CONNECTING, &xprt->sockstate);
- smp_mb__after_clear_bit();
-}
-
-/*
- * Attempt to connect a TCP socket.
+/**
+ * xprt_connect - schedule a transport connect operation
+ * @task: RPC task that is requesting the connect
*
*/
void xprt_connect(struct rpc_task *task)
@@ -552,37 +546,19 @@
if (!xprt_lock_write(xprt, task))
return;
if (xprt_connected(xprt))
- goto out_write;
+ xprt_release_write(xprt, task);
+ else {
+ if (task->tk_rqstp)
+ task->tk_rqstp->rq_bytes_sent = 0;
- if (task->tk_rqstp)
- task->tk_rqstp->rq_bytes_sent = 0;
-
- task->tk_timeout = RPC_CONNECT_TIMEOUT;
- rpc_sleep_on(&xprt->pending, task, xprt_connect_status, NULL);
- if (!test_and_set_bit(XPRT_CONNECTING, &xprt->sockstate)) {
- /* Note: if we are here due to a dropped connection
- * we delay reconnecting by RPC_REESTABLISH_TIMEOUT/HZ
- * seconds
- */
- if (xprt->sock != NULL)
- schedule_delayed_work(&xprt->sock_connect,
- RPC_REESTABLISH_TIMEOUT);
- else {
- schedule_work(&xprt->sock_connect);
- if (!RPC_IS_ASYNC(task))
- flush_scheduled_work();
- }
+ task->tk_timeout = xprt->connect_timeout;
+ rpc_sleep_on(&xprt->pending, task, xprt_connect_status, NULL);
+ xprt->ops->connect(task);
}
return;
- out_write:
- xprt_release_write(xprt, task);
}
-/*
- * We arrive here when awoken from waiting on connection establishment.
- */
-static void
-xprt_connect_status(struct rpc_task *task)
+static void xprt_connect_status(struct rpc_task *task)
{
struct rpc_xprt *xprt = task->tk_xprt;
@@ -592,31 +568,42 @@
return;
}
- /* if soft mounted, just cause this RPC to fail */
- if (RPC_IS_SOFT(task))
- task->tk_status = -EIO;
-
switch (task->tk_status) {
case -ECONNREFUSED:
case -ECONNRESET:
+ dprintk("RPC: %4d xprt_connect_status: server %s refused connection\n",
+ task->tk_pid, task->tk_client->cl_server);
+ break;
case -ENOTCONN:
- return;
+ dprintk("RPC: %4d xprt_connect_status: connection broken\n",
+ task->tk_pid);
+ break;
case -ETIMEDOUT:
- dprintk("RPC: %4d xprt_connect_status: timed out\n",
+ dprintk("RPC: %4d xprt_connect_status: connect attempt timed out\n",
task->tk_pid);
break;
default:
- printk(KERN_ERR "RPC: error %d connecting to server %s\n",
- -task->tk_status, task->tk_client->cl_server);
+ dprintk("RPC: %4d xprt_connect_status: error %d connecting to server %s\n",
+ task->tk_pid, -task->tk_status, task->tk_client->cl_server);
+ xprt_release_write(xprt, task);
+ task->tk_status = -EIO;
+ return;
}
- xprt_release_write(xprt, task);
+
+ /* if soft mounted, just cause this RPC to fail */
+ if (RPC_IS_SOFT(task)) {
+ xprt_release_write(xprt, task);
+ task->tk_status = -EIO;
+ }
}
-/*
- * Look up the RPC request corresponding to a reply, and then lock it.
+/**
+ * xprt_lookup_rqst - find an RPC request corresponding to an XID
+ * @xprt: transport on which the original request was transmitted
+ * @xid: RPC XID of incoming reply
+ *
*/
-static inline struct rpc_rqst *
-xprt_lookup_rqst(struct rpc_xprt *xprt, u32 xid)
+struct rpc_rqst *xprt_lookup_rqst(struct rpc_xprt *xprt, u32 xid)
{
struct list_head *pos;
struct rpc_rqst *req = NULL;
@@ -631,556 +618,68 @@
return req;
}
-/*
- * Complete reply received.
- * The TCP code relies on us to remove the request from xprt->pending.
+/**
+ * xprt_update_rtt - update an RPC client's RTT state after receiving a reply
+ * @task: RPC request that recently completed
+ *
*/
-static void
-xprt_complete_rqst(struct rpc_xprt *xprt, struct rpc_rqst *req, int copied)
+void xprt_update_rtt(struct rpc_task *task)
{
- struct rpc_task *task = req->rq_task;
- struct rpc_clnt *clnt = task->tk_client;
+ struct rpc_rqst *req = task->tk_rqstp;
+ struct rpc_rtt *rtt = task->tk_client->cl_rtt;
+ unsigned timer = task->tk_msg.rpc_proc->p_timer;
- /* Adjust congestion window */
- if (!xprt->nocong) {
- unsigned timer = task->tk_msg.rpc_proc->p_timer;
- xprt_adjust_cwnd(xprt, copied);
- __xprt_put_cong(xprt, req);
- if (timer) {
- if (req->rq_ntrans == 1)
- rpc_update_rtt(clnt->cl_rtt, timer,
- (long)jiffies - req->rq_xtime);
- rpc_set_timeo(clnt->cl_rtt, timer, req->rq_ntrans - 1);
- }
+ if (timer) {
+ if (req->rq_ntrans == 1)
+ rpc_update_rtt(rtt, timer,
+ (long)jiffies - req->rq_xtime);
+ rpc_set_timeo(rtt, timer, req->rq_ntrans - 1);
}
+}
-#ifdef RPC_PROFILE
- /* Profile only reads for now */
- if (copied > 1024) {
- static unsigned long nextstat;
- static unsigned long pkt_rtt, pkt_len, pkt_cnt;
+/**
+ * xprt_complete_rqst - called when reply processing is complete
+ * @task: RPC request that recently completed
+ * @copied: actual number of bytes received from the transport
+ *
+ * Caller holds transport lock.
+ */
+void xprt_complete_rqst(struct rpc_task *task, int copied)
+{
+ struct rpc_rqst *req = task->tk_rqstp;
- pkt_cnt++;
- pkt_len += req->rq_slen + copied;
- pkt_rtt += jiffies - req->rq_xtime;
- if (time_before(nextstat, jiffies)) {
- printk("RPC: %lu %ld cwnd\n", jiffies, xprt->cwnd);
- printk("RPC: %ld %ld %ld %ld stat\n",
- jiffies, pkt_cnt, pkt_len, pkt_rtt);
- pkt_rtt = pkt_len = pkt_cnt = 0;
- nextstat = jiffies + 5 * HZ;
- }
- }
-#endif
+ dprintk("RPC: %5u xid %08x complete (%d bytes received)\n",
+ task->tk_pid, ntohl(req->rq_xid), copied);
- dprintk("RPC: %4d has input (%d bytes)\n", task->tk_pid, copied);
list_del_init(&req->rq_list);
req->rq_received = req->rq_private_buf.len = copied;
-
- /* ... and wake up the process. */
rpc_wake_up_task(task);
- return;
}
-static size_t
-skb_read_bits(skb_reader_t *desc, void *to, size_t len)
+static void xprt_timer(struct rpc_task *task)
{
- if (len > desc->count)
- len = desc->count;
- if (skb_copy_bits(desc->skb, desc->offset, to, len))
- return 0;
- desc->count -= len;
- desc->offset += len;
- return len;
-}
-
-static size_t
-skb_read_and_csum_bits(skb_reader_t *desc, void *to, size_t len)
-{
- unsigned int csum2, pos;
-
- if (len > desc->count)
- len = desc->count;
- pos = desc->offset;
- csum2 = skb_copy_and_csum_bits(desc->skb, pos, to, len, 0);
- desc->csum = csum_block_add(desc->csum, csum2, pos);
- desc->count -= len;
- desc->offset += len;
- return len;
-}
-
-/*
- * We have set things up such that we perform the checksum of the UDP
- * packet in parallel with the copies into the RPC client iovec. -DaveM
- */
-int
-csum_partial_copy_to_xdr(struct xdr_buf *xdr, struct sk_buff *skb)
-{
- skb_reader_t desc;
-
- desc.skb = skb;
- desc.offset = sizeof(struct udphdr);
- desc.count = skb->len - desc.offset;
-
- if (skb->ip_summed == CHECKSUM_UNNECESSARY)
- goto no_checksum;
-
- desc.csum = csum_partial(skb->data, desc.offset, skb->csum);
- if (xdr_partial_copy_from_skb(xdr, 0, &desc, skb_read_and_csum_bits) < 0)
- return -1;
- if (desc.offset != skb->len) {
- unsigned int csum2;
- csum2 = skb_checksum(skb, desc.offset, skb->len - desc.offset, 0);
- desc.csum = csum_block_add(desc.csum, csum2, desc.offset);
- }
- if (desc.count)
- return -1;
- if ((unsigned short)csum_fold(desc.csum))
- return -1;
- return 0;
-no_checksum:
- if (xdr_partial_copy_from_skb(xdr, 0, &desc, skb_read_bits) < 0)
- return -1;
- if (desc.count)
- return -1;
- return 0;
-}
-
-/*
- * Input handler for RPC replies. Called from a bottom half and hence
- * atomic.
- */
-static void
-udp_data_ready(struct sock *sk, int len)
-{
- struct rpc_task *task;
- struct rpc_xprt *xprt;
- struct rpc_rqst *rovr;
- struct sk_buff *skb;
- int err, repsize, copied;
- u32 _xid, *xp;
-
- read_lock(&sk->sk_callback_lock);
- dprintk("RPC: udp_data_ready...\n");
- if (!(xprt = xprt_from_sock(sk))) {
- printk("RPC: udp_data_ready request not found!\n");
- goto out;
- }
-
- dprintk("RPC: udp_data_ready client %p\n", xprt);
-
- if ((skb = skb_recv_datagram(sk, 0, 1, &err)) == NULL)
- goto out;
-
- if (xprt->shutdown)
- goto dropit;
-
- repsize = skb->len - sizeof(struct udphdr);
- if (repsize < 4) {
- printk("RPC: impossible RPC reply size %d!\n", repsize);
- goto dropit;
- }
-
- /* Copy the XID from the skb... */
- xp = skb_header_pointer(skb, sizeof(struct udphdr),
- sizeof(_xid), &_xid);
- if (xp == NULL)
- goto dropit;
-
- /* Look up and lock the request corresponding to the given XID */
- spin_lock(&xprt->sock_lock);
- rovr = xprt_lookup_rqst(xprt, *xp);
- if (!rovr)
- goto out_unlock;
- task = rovr->rq_task;
-
- dprintk("RPC: %4d received reply\n", task->tk_pid);
-
- if ((copied = rovr->rq_private_buf.buflen) > repsize)
- copied = repsize;
-
- /* Suck it into the iovec, verify checksum if not done by hw. */
- if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb))
- goto out_unlock;
-
- /* Something worked... */
- dst_confirm(skb->dst);
-
- xprt_complete_rqst(xprt, rovr, copied);
-
- out_unlock:
- spin_unlock(&xprt->sock_lock);
- dropit:
- skb_free_datagram(sk, skb);
- out:
- read_unlock(&sk->sk_callback_lock);
-}
-
-/*
- * Copy from an skb into memory and shrink the skb.
- */
-static inline size_t
-tcp_copy_data(skb_reader_t *desc, void *p, size_t len)
-{
- if (len > desc->count)
- len = desc->count;
- if (skb_copy_bits(desc->skb, desc->offset, p, len)) {
- dprintk("RPC: failed to copy %zu bytes from skb. %zu bytes remain\n",
- len, desc->count);
- return 0;
- }
- desc->offset += len;
- desc->count -= len;
- dprintk("RPC: copied %zu bytes from skb. %zu bytes remain\n",
- len, desc->count);
- return len;
-}
-
-/*
- * TCP read fragment marker
- */
-static inline void
-tcp_read_fraghdr(struct rpc_xprt *xprt, skb_reader_t *desc)
-{
- size_t len, used;
- char *p;
-
- p = ((char *) &xprt->tcp_recm) + xprt->tcp_offset;
- len = sizeof(xprt->tcp_recm) - xprt->tcp_offset;
- used = tcp_copy_data(desc, p, len);
- xprt->tcp_offset += used;
- if (used != len)
- return;
- xprt->tcp_reclen = ntohl(xprt->tcp_recm);
- if (xprt->tcp_reclen & 0x80000000)
- xprt->tcp_flags |= XPRT_LAST_FRAG;
- else
- xprt->tcp_flags &= ~XPRT_LAST_FRAG;
- xprt->tcp_reclen &= 0x7fffffff;
- xprt->tcp_flags &= ~XPRT_COPY_RECM;
- xprt->tcp_offset = 0;
- /* Sanity check of the record length */
- if (xprt->tcp_reclen < 4) {
- printk(KERN_ERR "RPC: Invalid TCP record fragment length\n");
- xprt_disconnect(xprt);
- }
- dprintk("RPC: reading TCP record fragment of length %d\n",
- xprt->tcp_reclen);
-}
-
-static void
-tcp_check_recm(struct rpc_xprt *xprt)
-{
- dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u, tcp_flags = %lx\n",
- xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen, xprt->tcp_flags);
- if (xprt->tcp_offset == xprt->tcp_reclen) {
- xprt->tcp_flags |= XPRT_COPY_RECM;
- xprt->tcp_offset = 0;
- if (xprt->tcp_flags & XPRT_LAST_FRAG) {
- xprt->tcp_flags &= ~XPRT_COPY_DATA;
- xprt->tcp_flags |= XPRT_COPY_XID;
- xprt->tcp_copied = 0;
- }
- }
-}
-
-/*
- * TCP read xid
- */
-static inline void
-tcp_read_xid(struct rpc_xprt *xprt, skb_reader_t *desc)
-{
- size_t len, used;
- char *p;
-
- len = sizeof(xprt->tcp_xid) - xprt->tcp_offset;
- dprintk("RPC: reading XID (%Zu bytes)\n", len);
- p = ((char *) &xprt->tcp_xid) + xprt->tcp_offset;
- used = tcp_copy_data(desc, p, len);
- xprt->tcp_offset += used;
- if (used != len)
- return;
- xprt->tcp_flags &= ~XPRT_COPY_XID;
- xprt->tcp_flags |= XPRT_COPY_DATA;
- xprt->tcp_copied = 4;
- dprintk("RPC: reading reply for XID %08x\n",
- ntohl(xprt->tcp_xid));
- tcp_check_recm(xprt);
-}
-
-/*
- * TCP read and complete request
- */
-static inline void
-tcp_read_request(struct rpc_xprt *xprt, skb_reader_t *desc)
-{
- struct rpc_rqst *req;
- struct xdr_buf *rcvbuf;
- size_t len;
- ssize_t r;
-
- /* Find and lock the request corresponding to this xid */
- spin_lock(&xprt->sock_lock);
- req = xprt_lookup_rqst(xprt, xprt->tcp_xid);
- if (!req) {
- xprt->tcp_flags &= ~XPRT_COPY_DATA;
- dprintk("RPC: XID %08x request not found!\n",
- ntohl(xprt->tcp_xid));
- spin_unlock(&xprt->sock_lock);
- return;
- }
-
- rcvbuf = &req->rq_private_buf;
- len = desc->count;
- if (len > xprt->tcp_reclen - xprt->tcp_offset) {
- skb_reader_t my_desc;
-
- len = xprt->tcp_reclen - xprt->tcp_offset;
- memcpy(&my_desc, desc, sizeof(my_desc));
- my_desc.count = len;
- r = xdr_partial_copy_from_skb(rcvbuf, xprt->tcp_copied,
- &my_desc, tcp_copy_data);
- desc->count -= r;
- desc->offset += r;
- } else
- r = xdr_partial_copy_from_skb(rcvbuf, xprt->tcp_copied,
- desc, tcp_copy_data);
-
- if (r > 0) {
- xprt->tcp_copied += r;
- xprt->tcp_offset += r;
- }
- if (r != len) {
- /* Error when copying to the receive buffer,
- * usually because we weren't able to allocate
- * additional buffer pages. All we can do now
- * is turn off XPRT_COPY_DATA, so the request
- * will not receive any additional updates,
- * and time out.
- * Any remaining data from this record will
- * be discarded.
- */
- xprt->tcp_flags &= ~XPRT_COPY_DATA;
- dprintk("RPC: XID %08x truncated request\n",
- ntohl(xprt->tcp_xid));
- dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u\n",
- xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen);
- goto out;
- }
-
- dprintk("RPC: XID %08x read %Zd bytes\n",
- ntohl(xprt->tcp_xid), r);
- dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u\n",
- xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen);
-
- if (xprt->tcp_copied == req->rq_private_buf.buflen)
- xprt->tcp_flags &= ~XPRT_COPY_DATA;
- else if (xprt->tcp_offset == xprt->tcp_reclen) {
- if (xprt->tcp_flags & XPRT_LAST_FRAG)
- xprt->tcp_flags &= ~XPRT_COPY_DATA;
- }
-
-out:
- if (!(xprt->tcp_flags & XPRT_COPY_DATA)) {
- dprintk("RPC: %4d received reply complete\n",
- req->rq_task->tk_pid);
- xprt_complete_rqst(xprt, req, xprt->tcp_copied);
- }
- spin_unlock(&xprt->sock_lock);
- tcp_check_recm(xprt);
-}
-
-/*
- * TCP discard extra bytes from a short read
- */
-static inline void
-tcp_read_discard(struct rpc_xprt *xprt, skb_reader_t *desc)
-{
- size_t len;
-
- len = xprt->tcp_reclen - xprt->tcp_offset;
- if (len > desc->count)
- len = desc->count;
- desc->count -= len;
- desc->offset += len;
- xprt->tcp_offset += len;
- dprintk("RPC: discarded %Zu bytes\n", len);
- tcp_check_recm(xprt);
-}
-
-/*
- * TCP record receive routine
- * We first have to grab the record marker, then the XID, then the data.
- */
-static int
-tcp_data_recv(read_descriptor_t *rd_desc, struct sk_buff *skb,
- unsigned int offset, size_t len)
-{
- struct rpc_xprt *xprt = rd_desc->arg.data;
- skb_reader_t desc = {
- .skb = skb,
- .offset = offset,
- .count = len,
- .csum = 0
- };
-
- dprintk("RPC: tcp_data_recv\n");
- do {
- /* Read in a new fragment marker if necessary */
- /* Can we ever really expect to get completely empty fragments? */
- if (xprt->tcp_flags & XPRT_COPY_RECM) {
- tcp_read_fraghdr(xprt, &desc);
- continue;
- }
- /* Read in the xid if necessary */
- if (xprt->tcp_flags & XPRT_COPY_XID) {
- tcp_read_xid(xprt, &desc);
- continue;
- }
- /* Read in the request data */
- if (xprt->tcp_flags & XPRT_COPY_DATA) {
- tcp_read_request(xprt, &desc);
- continue;
- }
- /* Skip over any trailing bytes on short reads */
- tcp_read_discard(xprt, &desc);
- } while (desc.count);
- dprintk("RPC: tcp_data_recv done\n");
- return len - desc.count;
-}
-
-static void tcp_data_ready(struct sock *sk, int bytes)
-{
- struct rpc_xprt *xprt;
- read_descriptor_t rd_desc;
-
- read_lock(&sk->sk_callback_lock);
- dprintk("RPC: tcp_data_ready...\n");
- if (!(xprt = xprt_from_sock(sk))) {
- printk("RPC: tcp_data_ready socket info not found!\n");
- goto out;
- }
- if (xprt->shutdown)
- goto out;
-
- /* We use rd_desc to pass struct xprt to tcp_data_recv */
- rd_desc.arg.data = xprt;
- rd_desc.count = 65536;
- tcp_read_sock(sk, &rd_desc, tcp_data_recv);
-out:
- read_unlock(&sk->sk_callback_lock);
-}
-
-static void
-tcp_state_change(struct sock *sk)
-{
- struct rpc_xprt *xprt;
-
- read_lock(&sk->sk_callback_lock);
- if (!(xprt = xprt_from_sock(sk)))
- goto out;
- dprintk("RPC: tcp_state_change client %p...\n", xprt);
- dprintk("RPC: state %x conn %d dead %d zapped %d\n",
- sk->sk_state, xprt_connected(xprt),
- sock_flag(sk, SOCK_DEAD),
- sock_flag(sk, SOCK_ZAPPED));
-
- switch (sk->sk_state) {
- case TCP_ESTABLISHED:
- spin_lock_bh(&xprt->sock_lock);
- if (!xprt_test_and_set_connected(xprt)) {
- /* Reset TCP record info */
- xprt->tcp_offset = 0;
- xprt->tcp_reclen = 0;
- xprt->tcp_copied = 0;
- xprt->tcp_flags = XPRT_COPY_RECM | XPRT_COPY_XID;
- rpc_wake_up(&xprt->pending);
- }
- spin_unlock_bh(&xprt->sock_lock);
- break;
- case TCP_SYN_SENT:
- case TCP_SYN_RECV:
- break;
- default:
- xprt_disconnect(xprt);
- break;
- }
- out:
- read_unlock(&sk->sk_callback_lock);
-}
-
-/*
- * Called when more output buffer space is available for this socket.
- * We try not to wake our writers until they can make "significant"
- * progress, otherwise we'll waste resources thrashing sock_sendmsg
- * with a bunch of small requests.
- */
-static void
-xprt_write_space(struct sock *sk)
-{
- struct rpc_xprt *xprt;
- struct socket *sock;
-
- read_lock(&sk->sk_callback_lock);
- if (!(xprt = xprt_from_sock(sk)) || !(sock = sk->sk_socket))
- goto out;
- if (xprt->shutdown)
- goto out;
-
- /* Wait until we have enough socket memory */
- if (xprt->stream) {
- /* from net/core/stream.c:sk_stream_write_space */
- if (sk_stream_wspace(sk) < sk_stream_min_wspace(sk))
- goto out;
- } else {
- /* from net/core/sock.c:sock_def_write_space */
- if (!sock_writeable(sk))
- goto out;
- }
-
- if (!test_and_clear_bit(SOCK_NOSPACE, &sock->flags))
- goto out;
-
- spin_lock_bh(&xprt->sock_lock);
- if (xprt->snd_task)
- rpc_wake_up_task(xprt->snd_task);
- spin_unlock_bh(&xprt->sock_lock);
-out:
- read_unlock(&sk->sk_callback_lock);
-}
-
-/*
- * RPC receive timeout handler.
- */
-static void
-xprt_timer(struct rpc_task *task)
-{
- struct rpc_rqst *req = task->tk_rqstp;
+ struct rpc_rqst *req = task->tk_rqstp;
struct rpc_xprt *xprt = req->rq_xprt;
- spin_lock(&xprt->sock_lock);
- if (req->rq_received)
- goto out;
+ dprintk("RPC: %4d xprt_timer\n", task->tk_pid);
- xprt_adjust_cwnd(req->rq_xprt, -ETIMEDOUT);
- __xprt_put_cong(xprt, req);
-
- dprintk("RPC: %4d xprt_timer (%s request)\n",
- task->tk_pid, req ? "pending" : "backlogged");
-
- task->tk_status = -ETIMEDOUT;
-out:
+ spin_lock(&xprt->transport_lock);
+ if (!req->rq_received) {
+ if (xprt->ops->timer)
+ xprt->ops->timer(task);
+ task->tk_status = -ETIMEDOUT;
+ }
task->tk_timeout = 0;
rpc_wake_up_task(task);
- spin_unlock(&xprt->sock_lock);
+ spin_unlock(&xprt->transport_lock);
}
-/*
- * Place the actual RPC call.
- * We have to copy the iovec because sendmsg fiddles with its contents.
+/**
+ * xprt_prepare_transmit - reserve the transport before sending a request
+ * @task: RPC task about to send a request
+ *
*/
-int
-xprt_prepare_transmit(struct rpc_task *task)
+int xprt_prepare_transmit(struct rpc_task *task)
{
struct rpc_rqst *req = task->tk_rqstp;
struct rpc_xprt *xprt = req->rq_xprt;
@@ -1191,12 +690,12 @@
if (xprt->shutdown)
return -EIO;
- spin_lock_bh(&xprt->sock_lock);
+ spin_lock_bh(&xprt->transport_lock);
if (req->rq_received && !req->rq_bytes_sent) {
err = req->rq_received;
goto out_unlock;
}
- if (!__xprt_lock_write(xprt, task)) {
+ if (!xprt->ops->reserve_xprt(task)) {
err = -EAGAIN;
goto out_unlock;
}
@@ -1206,39 +705,42 @@
goto out_unlock;
}
out_unlock:
- spin_unlock_bh(&xprt->sock_lock);
+ spin_unlock_bh(&xprt->transport_lock);
return err;
}
void
-xprt_transmit(struct rpc_task *task)
+xprt_abort_transmit(struct rpc_task *task)
{
- struct rpc_clnt *clnt = task->tk_client;
+ struct rpc_xprt *xprt = task->tk_xprt;
+
+ xprt_release_write(xprt, task);
+}
+
+/**
+ * xprt_transmit - send an RPC request on a transport
+ * @task: controlling RPC task
+ *
+ * We have to copy the iovec because sendmsg fiddles with its contents.
+ */
+void xprt_transmit(struct rpc_task *task)
+{
struct rpc_rqst *req = task->tk_rqstp;
struct rpc_xprt *xprt = req->rq_xprt;
- int status, retry = 0;
-
+ int status;
dprintk("RPC: %4d xprt_transmit(%u)\n", task->tk_pid, req->rq_slen);
- /* set up everything as needed. */
- /* Write the record marker */
- if (xprt->stream) {
- u32 *marker = req->rq_svec[0].iov_base;
-
- *marker = htonl(0x80000000|(req->rq_slen-sizeof(*marker)));
- }
-
smp_rmb();
if (!req->rq_received) {
if (list_empty(&req->rq_list)) {
- spin_lock_bh(&xprt->sock_lock);
+ spin_lock_bh(&xprt->transport_lock);
/* Update the softirq receive buffer */
memcpy(&req->rq_private_buf, &req->rq_rcv_buf,
sizeof(req->rq_private_buf));
/* Add request to the receive list */
list_add_tail(&req->rq_list, &xprt->recv);
- spin_unlock_bh(&xprt->sock_lock);
+ spin_unlock_bh(&xprt->transport_lock);
xprt_reset_majortimeo(req);
/* Turn off autodisconnect */
del_singleshot_timer_sync(&xprt->timer);
@@ -1246,40 +748,19 @@
} else if (!req->rq_bytes_sent)
return;
- /* Continue transmitting the packet/record. We must be careful
- * to cope with writespace callbacks arriving _after_ we have
- * called xprt_sendmsg().
- */
- while (1) {
- req->rq_xtime = jiffies;
- status = xprt_sendmsg(xprt, req);
-
- if (status < 0)
- break;
-
- if (xprt->stream) {
- req->rq_bytes_sent += status;
-
- /* If we've sent the entire packet, immediately
- * reset the count of bytes sent. */
- if (req->rq_bytes_sent >= req->rq_slen) {
- req->rq_bytes_sent = 0;
- goto out_receive;
- }
- } else {
- if (status >= req->rq_slen)
- goto out_receive;
- status = -EAGAIN;
- break;
- }
-
- dprintk("RPC: %4d xmit incomplete (%d left of %d)\n",
- task->tk_pid, req->rq_slen - req->rq_bytes_sent,
- req->rq_slen);
-
- status = -EAGAIN;
- if (retry++ > 50)
- break;
+ status = xprt->ops->send_request(task);
+ if (status == 0) {
+ dprintk("RPC: %4d xmit complete\n", task->tk_pid);
+ spin_lock_bh(&xprt->transport_lock);
+ xprt->ops->set_retrans_timeout(task);
+ /* Don't race with disconnect */
+ if (!xprt_connected(xprt))
+ task->tk_status = -ENOTCONN;
+ else if (!req->rq_received)
+ rpc_sleep_on(&xprt->pending, task, NULL, xprt_timer);
+ xprt->ops->release_xprt(xprt, task);
+ spin_unlock_bh(&xprt->transport_lock);
+ return;
}
/* Note: at this point, task->tk_sleeping has not yet been set,
@@ -1289,60 +770,19 @@
task->tk_status = status;
switch (status) {
- case -EAGAIN:
- if (test_bit(SOCK_ASYNC_NOSPACE, &xprt->sock->flags)) {
- /* Protect against races with xprt_write_space */
- spin_lock_bh(&xprt->sock_lock);
- /* Don't race with disconnect */
- if (!xprt_connected(xprt))
- task->tk_status = -ENOTCONN;
- else if (test_bit(SOCK_NOSPACE, &xprt->sock->flags)) {
- task->tk_timeout = req->rq_timeout;
- rpc_sleep_on(&xprt->pending, task, NULL, NULL);
- }
- spin_unlock_bh(&xprt->sock_lock);
- return;
- }
- /* Keep holding the socket if it is blocked */
- rpc_delay(task, HZ>>4);
- return;
case -ECONNREFUSED:
- task->tk_timeout = RPC_REESTABLISH_TIMEOUT;
rpc_sleep_on(&xprt->sending, task, NULL, NULL);
+ case -EAGAIN:
case -ENOTCONN:
return;
default:
- if (xprt->stream)
- xprt_disconnect(xprt);
+ break;
}
xprt_release_write(xprt, task);
return;
- out_receive:
- dprintk("RPC: %4d xmit complete\n", task->tk_pid);
- /* Set the task's receive timeout value */
- spin_lock_bh(&xprt->sock_lock);
- if (!xprt->nocong) {
- int timer = task->tk_msg.rpc_proc->p_timer;
- task->tk_timeout = rpc_calc_rto(clnt->cl_rtt, timer);
- task->tk_timeout <<= rpc_ntimeo(clnt->cl_rtt, timer) + req->rq_retries;
- if (task->tk_timeout > xprt->timeout.to_maxval || task->tk_timeout == 0)
- task->tk_timeout = xprt->timeout.to_maxval;
- } else
- task->tk_timeout = req->rq_timeout;
- /* Don't race with disconnect */
- if (!xprt_connected(xprt))
- task->tk_status = -ENOTCONN;
- else if (!req->rq_received)
- rpc_sleep_on(&xprt->pending, task, NULL, xprt_timer);
- __xprt_release_write(xprt, task);
- spin_unlock_bh(&xprt->sock_lock);
}
-/*
- * Reserve an RPC call slot.
- */
-static inline void
-do_xprt_reserve(struct rpc_task *task)
+static inline void do_xprt_reserve(struct rpc_task *task)
{
struct rpc_xprt *xprt = task->tk_xprt;
@@ -1362,22 +802,25 @@
rpc_sleep_on(&xprt->backlog, task, NULL, NULL);
}
-void
-xprt_reserve(struct rpc_task *task)
+/**
+ * xprt_reserve - allocate an RPC request slot
+ * @task: RPC task requesting a slot allocation
+ *
+ * If no more slots are available, place the task on the transport's
+ * backlog queue.
+ */
+void xprt_reserve(struct rpc_task *task)
{
struct rpc_xprt *xprt = task->tk_xprt;
task->tk_status = -EIO;
if (!xprt->shutdown) {
- spin_lock(&xprt->xprt_lock);
+ spin_lock(&xprt->reserve_lock);
do_xprt_reserve(task);
- spin_unlock(&xprt->xprt_lock);
+ spin_unlock(&xprt->reserve_lock);
}
}
-/*
- * Allocate a 'unique' XID
- */
static inline u32 xprt_alloc_xid(struct rpc_xprt *xprt)
{
return xprt->xid++;
@@ -1388,11 +831,7 @@
get_random_bytes(&xprt->xid, sizeof(xprt->xid));
}
-/*
- * Initialize RPC request
- */
-static void
-xprt_request_init(struct rpc_task *task, struct rpc_xprt *xprt)
+static void xprt_request_init(struct rpc_task *task, struct rpc_xprt *xprt)
{
struct rpc_rqst *req = task->tk_rqstp;
@@ -1400,128 +839,104 @@
req->rq_task = task;
req->rq_xprt = xprt;
req->rq_xid = xprt_alloc_xid(xprt);
+ req->rq_release_snd_buf = NULL;
dprintk("RPC: %4d reserved req %p xid %08x\n", task->tk_pid,
req, ntohl(req->rq_xid));
}
-/*
- * Release an RPC call slot
+/**
+ * xprt_release - release an RPC request slot
+ * @task: task which is finished with the slot
+ *
*/
-void
-xprt_release(struct rpc_task *task)
+void xprt_release(struct rpc_task *task)
{
struct rpc_xprt *xprt = task->tk_xprt;
struct rpc_rqst *req;
if (!(req = task->tk_rqstp))
return;
- spin_lock_bh(&xprt->sock_lock);
- __xprt_release_write(xprt, task);
- __xprt_put_cong(xprt, req);
+ spin_lock_bh(&xprt->transport_lock);
+ xprt->ops->release_xprt(xprt, task);
+ if (xprt->ops->release_request)
+ xprt->ops->release_request(task);
if (!list_empty(&req->rq_list))
list_del(&req->rq_list);
xprt->last_used = jiffies;
if (list_empty(&xprt->recv) && !xprt->shutdown)
- mod_timer(&xprt->timer, xprt->last_used + XPRT_IDLE_TIMEOUT);
- spin_unlock_bh(&xprt->sock_lock);
+ mod_timer(&xprt->timer,
+ xprt->last_used + xprt->idle_timeout);
+ spin_unlock_bh(&xprt->transport_lock);
task->tk_rqstp = NULL;
+ if (req->rq_release_snd_buf)
+ req->rq_release_snd_buf(req);
memset(req, 0, sizeof(*req)); /* mark unused */
dprintk("RPC: %4d release request %p\n", task->tk_pid, req);
- spin_lock(&xprt->xprt_lock);
+ spin_lock(&xprt->reserve_lock);
list_add(&req->rq_list, &xprt->free);
- xprt_clear_backlog(xprt);
- spin_unlock(&xprt->xprt_lock);
+ rpc_wake_up_next(&xprt->backlog);
+ spin_unlock(&xprt->reserve_lock);
}
-/*
- * Set default timeout parameters
+/**
+ * xprt_set_timeout - set constant RPC timeout
+ * @to: RPC timeout parameters to set up
+ * @retr: number of retries
+ * @incr: amount of increase after each retry
+ *
*/
-static void
-xprt_default_timeout(struct rpc_timeout *to, int proto)
-{
- if (proto == IPPROTO_UDP)
- xprt_set_timeout(to, 5, 5 * HZ);
- else
- xprt_set_timeout(to, 5, 60 * HZ);
-}
-
-/*
- * Set constant timeout
- */
-void
-xprt_set_timeout(struct rpc_timeout *to, unsigned int retr, unsigned long incr)
+void xprt_set_timeout(struct rpc_timeout *to, unsigned int retr, unsigned long incr)
{
to->to_initval =
to->to_increment = incr;
- to->to_maxval = incr * retr;
+ to->to_maxval = to->to_initval + (incr * retr);
to->to_retries = retr;
to->to_exponential = 0;
}
-unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE;
-unsigned int xprt_tcp_slot_table_entries = RPC_DEF_SLOT_TABLE;
-
-/*
- * Initialize an RPC client
- */
-static struct rpc_xprt *
-xprt_setup(int proto, struct sockaddr_in *ap, struct rpc_timeout *to)
+static struct rpc_xprt *xprt_setup(int proto, struct sockaddr_in *ap, struct rpc_timeout *to)
{
+ int result;
struct rpc_xprt *xprt;
- unsigned int entries;
- size_t slot_table_size;
struct rpc_rqst *req;
- dprintk("RPC: setting up %s transport...\n",
- proto == IPPROTO_UDP? "UDP" : "TCP");
-
- entries = (proto == IPPROTO_TCP)?
- xprt_tcp_slot_table_entries : xprt_udp_slot_table_entries;
-
if ((xprt = kmalloc(sizeof(struct rpc_xprt), GFP_KERNEL)) == NULL)
return ERR_PTR(-ENOMEM);
memset(xprt, 0, sizeof(*xprt)); /* Nnnngh! */
- xprt->max_reqs = entries;
- slot_table_size = entries * sizeof(xprt->slot[0]);
- xprt->slot = kmalloc(slot_table_size, GFP_KERNEL);
- if (xprt->slot == NULL) {
- kfree(xprt);
- return ERR_PTR(-ENOMEM);
- }
- memset(xprt->slot, 0, slot_table_size);
xprt->addr = *ap;
- xprt->prot = proto;
- xprt->stream = (proto == IPPROTO_TCP)? 1 : 0;
- if (xprt->stream) {
- xprt->cwnd = RPC_MAXCWND(xprt);
- xprt->nocong = 1;
- xprt->max_payload = (1U << 31) - 1;
- } else {
- xprt->cwnd = RPC_INITCWND;
- xprt->max_payload = (1U << 16) - (MAX_HEADER << 3);
+
+ switch (proto) {
+ case IPPROTO_UDP:
+ result = xs_setup_udp(xprt, to);
+ break;
+ case IPPROTO_TCP:
+ result = xs_setup_tcp(xprt, to);
+ break;
+ default:
+ printk(KERN_ERR "RPC: unrecognized transport protocol: %d\n",
+ proto);
+ result = -EIO;
+ break;
}
- spin_lock_init(&xprt->sock_lock);
- spin_lock_init(&xprt->xprt_lock);
- init_waitqueue_head(&xprt->cong_wait);
+ if (result) {
+ kfree(xprt);
+ return ERR_PTR(result);
+ }
+
+ spin_lock_init(&xprt->transport_lock);
+ spin_lock_init(&xprt->reserve_lock);
INIT_LIST_HEAD(&xprt->free);
INIT_LIST_HEAD(&xprt->recv);
- INIT_WORK(&xprt->sock_connect, xprt_socket_connect, xprt);
- INIT_WORK(&xprt->task_cleanup, xprt_socket_autoclose, xprt);
+ INIT_WORK(&xprt->task_cleanup, xprt_autoclose, xprt);
init_timer(&xprt->timer);
xprt->timer.function = xprt_init_autodisconnect;
xprt->timer.data = (unsigned long) xprt;
xprt->last_used = jiffies;
- xprt->port = XPRT_MAX_RESVPORT;
-
- /* Set timeout parameters */
- if (to) {
- xprt->timeout = *to;
- } else
- xprt_default_timeout(&xprt->timeout, xprt->prot);
+ xprt->cwnd = RPC_INITCWND;
rpc_init_wait_queue(&xprt->pending, "xprt_pending");
rpc_init_wait_queue(&xprt->sending, "xprt_sending");
@@ -1529,139 +944,25 @@
rpc_init_priority_wait_queue(&xprt->backlog, "xprt_backlog");
/* initialize free list */
- for (req = &xprt->slot[entries-1]; req >= &xprt->slot[0]; req--)
+ for (req = &xprt->slot[xprt->max_reqs-1]; req >= &xprt->slot[0]; req--)
list_add(&req->rq_list, &xprt->free);
xprt_init_xid(xprt);
- /* Check whether we want to use a reserved port */
- xprt->resvport = capable(CAP_NET_BIND_SERVICE) ? 1 : 0;
-
dprintk("RPC: created transport %p with %u slots\n", xprt,
xprt->max_reqs);
return xprt;
}
-/*
- * Bind to a reserved port
+/**
+ * xprt_create_proto - create an RPC client transport
+ * @proto: requested transport protocol
+ * @sap: remote peer's address
+ * @to: timeout parameters for new transport
+ *
*/
-static inline int xprt_bindresvport(struct rpc_xprt *xprt, struct socket *sock)
-{
- struct sockaddr_in myaddr = {
- .sin_family = AF_INET,
- };
- int err, port;
-
- /* Were we already bound to a given port? Try to reuse it */
- port = xprt->port;
- do {
- myaddr.sin_port = htons(port);
- err = sock->ops->bind(sock, (struct sockaddr *) &myaddr,
- sizeof(myaddr));
- if (err == 0) {
- xprt->port = port;
- return 0;
- }
- if (--port == 0)
- port = XPRT_MAX_RESVPORT;
- } while (err == -EADDRINUSE && port != xprt->port);
-
- printk("RPC: Can't bind to reserved port (%d).\n", -err);
- return err;
-}
-
-static void
-xprt_bind_socket(struct rpc_xprt *xprt, struct socket *sock)
-{
- struct sock *sk = sock->sk;
-
- if (xprt->inet)
- return;
-
- write_lock_bh(&sk->sk_callback_lock);
- sk->sk_user_data = xprt;
- xprt->old_data_ready = sk->sk_data_ready;
- xprt->old_state_change = sk->sk_state_change;
- xprt->old_write_space = sk->sk_write_space;
- if (xprt->prot == IPPROTO_UDP) {
- sk->sk_data_ready = udp_data_ready;
- sk->sk_no_check = UDP_CSUM_NORCV;
- xprt_set_connected(xprt);
- } else {
- tcp_sk(sk)->nonagle = 1; /* disable Nagle's algorithm */
- sk->sk_data_ready = tcp_data_ready;
- sk->sk_state_change = tcp_state_change;
- xprt_clear_connected(xprt);
- }
- sk->sk_write_space = xprt_write_space;
-
- /* Reset to new socket */
- xprt->sock = sock;
- xprt->inet = sk;
- write_unlock_bh(&sk->sk_callback_lock);
-
- return;
-}
-
-/*
- * Set socket buffer length
- */
-void
-xprt_sock_setbufsize(struct rpc_xprt *xprt)
-{
- struct sock *sk = xprt->inet;
-
- if (xprt->stream)
- return;
- if (xprt->rcvsize) {
- sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
- sk->sk_rcvbuf = xprt->rcvsize * xprt->max_reqs * 2;
- }
- if (xprt->sndsize) {
- sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
- sk->sk_sndbuf = xprt->sndsize * xprt->max_reqs * 2;
- sk->sk_write_space(sk);
- }
-}
-
-/*
- * Datastream sockets are created here, but xprt_connect will create
- * and connect stream sockets.
- */
-static struct socket * xprt_create_socket(struct rpc_xprt *xprt, int proto, int resvport)
-{
- struct socket *sock;
- int type, err;
-
- dprintk("RPC: xprt_create_socket(%s %d)\n",
- (proto == IPPROTO_UDP)? "udp" : "tcp", proto);
-
- type = (proto == IPPROTO_UDP)? SOCK_DGRAM : SOCK_STREAM;
-
- if ((err = sock_create_kern(PF_INET, type, proto, &sock)) < 0) {
- printk("RPC: can't create socket (%d).\n", -err);
- return NULL;
- }
-
- /* If the caller has the capability, bind to a reserved port */
- if (resvport && xprt_bindresvport(xprt, sock) < 0) {
- printk("RPC: can't bind to reserved port.\n");
- goto failed;
- }
-
- return sock;
-
-failed:
- sock_release(sock);
- return NULL;
-}
-
-/*
- * Create an RPC client transport given the protocol and peer address.
- */
-struct rpc_xprt *
-xprt_create_proto(int proto, struct sockaddr_in *sap, struct rpc_timeout *to)
+struct rpc_xprt *xprt_create_proto(int proto, struct sockaddr_in *sap, struct rpc_timeout *to)
{
struct rpc_xprt *xprt;
@@ -1673,46 +974,26 @@
return xprt;
}
-/*
- * Prepare for transport shutdown.
- */
-static void
-xprt_shutdown(struct rpc_xprt *xprt)
+static void xprt_shutdown(struct rpc_xprt *xprt)
{
xprt->shutdown = 1;
rpc_wake_up(&xprt->sending);
rpc_wake_up(&xprt->resend);
- rpc_wake_up(&xprt->pending);
+ xprt_wake_pending_tasks(xprt, -EIO);
rpc_wake_up(&xprt->backlog);
- wake_up(&xprt->cong_wait);
del_timer_sync(&xprt->timer);
-
- /* synchronously wait for connect worker to finish */
- cancel_delayed_work(&xprt->sock_connect);
- flush_scheduled_work();
}
-/*
- * Clear the xprt backlog queue
+/**
+ * xprt_destroy - destroy an RPC transport, killing off all requests.
+ * @xprt: transport to destroy
+ *
*/
-static int
-xprt_clear_backlog(struct rpc_xprt *xprt) {
- rpc_wake_up_next(&xprt->backlog);
- wake_up(&xprt->cong_wait);
- return 1;
-}
-
-/*
- * Destroy an RPC transport, killing off all requests.
- */
-int
-xprt_destroy(struct rpc_xprt *xprt)
+int xprt_destroy(struct rpc_xprt *xprt)
{
dprintk("RPC: destroying transport %p\n", xprt);
xprt_shutdown(xprt);
- xprt_disconnect(xprt);
- xprt_close(xprt);
- kfree(xprt->slot);
+ xprt->ops->destroy(xprt);
kfree(xprt);
return 0;
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
new file mode 100644
index 0000000..2e15292
--- /dev/null
+++ b/net/sunrpc/xprtsock.c
@@ -0,0 +1,1252 @@
+/*
+ * linux/net/sunrpc/xprtsock.c
+ *
+ * Client-side transport implementation for sockets.
+ *
+ * TCP callback races fixes (C) 1998 Red Hat Software <alan@redhat.com>
+ * TCP send fixes (C) 1998 Red Hat Software <alan@redhat.com>
+ * TCP NFS related read + write fixes
+ * (C) 1999 Dave Airlie, University of Limerick, Ireland <airlied@linux.ie>
+ *
+ * Rewrite of larges part of the code in order to stabilize TCP stuff.
+ * Fix behaviour when socket buffer is full.
+ * (C) 1999 Trond Myklebust <trond.myklebust@fys.uio.no>
+ *
+ * IP socket transport implementation, (C) 2005 Chuck Lever <cel@netapp.com>
+ */
+
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/capability.h>
+#include <linux/sched.h>
+#include <linux/pagemap.h>
+#include <linux/errno.h>
+#include <linux/socket.h>
+#include <linux/in.h>
+#include <linux/net.h>
+#include <linux/mm.h>
+#include <linux/udp.h>
+#include <linux/tcp.h>
+#include <linux/sunrpc/clnt.h>
+#include <linux/file.h>
+
+#include <net/sock.h>
+#include <net/checksum.h>
+#include <net/udp.h>
+#include <net/tcp.h>
+
+/*
+ * How many times to try sending a request on a socket before waiting
+ * for the socket buffer to clear.
+ */
+#define XS_SENDMSG_RETRY (10U)
+
+/*
+ * Time out for an RPC UDP socket connect. UDP socket connects are
+ * synchronous, but we set a timeout anyway in case of resource
+ * exhaustion on the local host.
+ */
+#define XS_UDP_CONN_TO (5U * HZ)
+
+/*
+ * Wait duration for an RPC TCP connection to be established. Solaris
+ * NFS over TCP uses 60 seconds, for example, which is in line with how
+ * long a server takes to reboot.
+ */
+#define XS_TCP_CONN_TO (60U * HZ)
+
+/*
+ * Wait duration for a reply from the RPC portmapper.
+ */
+#define XS_BIND_TO (60U * HZ)
+
+/*
+ * Delay if a UDP socket connect error occurs. This is most likely some
+ * kind of resource problem on the local host.
+ */
+#define XS_UDP_REEST_TO (2U * HZ)
+
+/*
+ * The reestablish timeout allows clients to delay for a bit before attempting
+ * to reconnect to a server that just dropped our connection.
+ *
+ * We implement an exponential backoff when trying to reestablish a TCP
+ * transport connection with the server. Some servers like to drop a TCP
+ * connection when they are overworked, so we start with a short timeout and
+ * increase over time if the server is down or not responding.
+ */
+#define XS_TCP_INIT_REEST_TO (3U * HZ)
+#define XS_TCP_MAX_REEST_TO (5U * 60 * HZ)
+
+/*
+ * TCP idle timeout; client drops the transport socket if it is idle
+ * for this long. Note that we also timeout UDP sockets to prevent
+ * holding port numbers when there is no RPC traffic.
+ */
+#define XS_IDLE_DISC_TO (5U * 60 * HZ)
+
+#ifdef RPC_DEBUG
+# undef RPC_DEBUG_DATA
+# define RPCDBG_FACILITY RPCDBG_TRANS
+#endif
+
+#ifdef RPC_DEBUG_DATA
+static void xs_pktdump(char *msg, u32 *packet, unsigned int count)
+{
+ u8 *buf = (u8 *) packet;
+ int j;
+
+ dprintk("RPC: %s\n", msg);
+ for (j = 0; j < count && j < 128; j += 4) {
+ if (!(j & 31)) {
+ if (j)
+ dprintk("\n");
+ dprintk("0x%04x ", j);
+ }
+ dprintk("%02x%02x%02x%02x ",
+ buf[j], buf[j+1], buf[j+2], buf[j+3]);
+ }
+ dprintk("\n");
+}
+#else
+static inline void xs_pktdump(char *msg, u32 *packet, unsigned int count)
+{
+ /* NOP */
+}
+#endif
+
+#define XS_SENDMSG_FLAGS (MSG_DONTWAIT | MSG_NOSIGNAL)
+
+static inline int xs_send_head(struct socket *sock, struct sockaddr *addr, int addrlen, struct xdr_buf *xdr, unsigned int base, unsigned int len)
+{
+ struct kvec iov = {
+ .iov_base = xdr->head[0].iov_base + base,
+ .iov_len = len - base,
+ };
+ struct msghdr msg = {
+ .msg_name = addr,
+ .msg_namelen = addrlen,
+ .msg_flags = XS_SENDMSG_FLAGS,
+ };
+
+ if (xdr->len > len)
+ msg.msg_flags |= MSG_MORE;
+
+ if (likely(iov.iov_len))
+ return kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len);
+ return kernel_sendmsg(sock, &msg, NULL, 0, 0);
+}
+
+static int xs_send_tail(struct socket *sock, struct xdr_buf *xdr, unsigned int base, unsigned int len)
+{
+ struct kvec iov = {
+ .iov_base = xdr->tail[0].iov_base + base,
+ .iov_len = len - base,
+ };
+ struct msghdr msg = {
+ .msg_flags = XS_SENDMSG_FLAGS,
+ };
+
+ return kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len);
+}
+
+/**
+ * xs_sendpages - write pages directly to a socket
+ * @sock: socket to send on
+ * @addr: UDP only -- address of destination
+ * @addrlen: UDP only -- length of destination address
+ * @xdr: buffer containing this request
+ * @base: starting position in the buffer
+ *
+ */
+static inline int xs_sendpages(struct socket *sock, struct sockaddr *addr, int addrlen, struct xdr_buf *xdr, unsigned int base)
+{
+ struct page **ppage = xdr->pages;
+ unsigned int len, pglen = xdr->page_len;
+ int err, ret = 0;
+ ssize_t (*sendpage)(struct socket *, struct page *, int, size_t, int);
+
+ if (unlikely(!sock))
+ return -ENOTCONN;
+
+ clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags);
+
+ len = xdr->head[0].iov_len;
+ if (base < len || (addr != NULL && base == 0)) {
+ err = xs_send_head(sock, addr, addrlen, xdr, base, len);
+ if (ret == 0)
+ ret = err;
+ else if (err > 0)
+ ret += err;
+ if (err != (len - base))
+ goto out;
+ base = 0;
+ } else
+ base -= len;
+
+ if (unlikely(pglen == 0))
+ goto copy_tail;
+ if (unlikely(base >= pglen)) {
+ base -= pglen;
+ goto copy_tail;
+ }
+ if (base || xdr->page_base) {
+ pglen -= base;
+ base += xdr->page_base;
+ ppage += base >> PAGE_CACHE_SHIFT;
+ base &= ~PAGE_CACHE_MASK;
+ }
+
+ sendpage = sock->ops->sendpage ? : sock_no_sendpage;
+ do {
+ int flags = XS_SENDMSG_FLAGS;
+
+ len = PAGE_CACHE_SIZE;
+ if (base)
+ len -= base;
+ if (pglen < len)
+ len = pglen;
+
+ if (pglen != len || xdr->tail[0].iov_len != 0)
+ flags |= MSG_MORE;
+
+ /* Hmm... We might be dealing with highmem pages */
+ if (PageHighMem(*ppage))
+ sendpage = sock_no_sendpage;
+ err = sendpage(sock, *ppage, base, len, flags);
+ if (ret == 0)
+ ret = err;
+ else if (err > 0)
+ ret += err;
+ if (err != len)
+ goto out;
+ base = 0;
+ ppage++;
+ } while ((pglen -= len) != 0);
+copy_tail:
+ len = xdr->tail[0].iov_len;
+ if (base < len) {
+ err = xs_send_tail(sock, xdr, base, len);
+ if (ret == 0)
+ ret = err;
+ else if (err > 0)
+ ret += err;
+ }
+out:
+ return ret;
+}
+
+/**
+ * xs_nospace - place task on wait queue if transmit was incomplete
+ * @task: task to put to sleep
+ *
+ */
+static void xs_nospace(struct rpc_task *task)
+{
+ struct rpc_rqst *req = task->tk_rqstp;
+ struct rpc_xprt *xprt = req->rq_xprt;
+
+ dprintk("RPC: %4d xmit incomplete (%u left of %u)\n",
+ task->tk_pid, req->rq_slen - req->rq_bytes_sent,
+ req->rq_slen);
+
+ if (test_bit(SOCK_ASYNC_NOSPACE, &xprt->sock->flags)) {
+ /* Protect against races with write_space */
+ spin_lock_bh(&xprt->transport_lock);
+
+ /* Don't race with disconnect */
+ if (!xprt_connected(xprt))
+ task->tk_status = -ENOTCONN;
+ else if (test_bit(SOCK_NOSPACE, &xprt->sock->flags))
+ xprt_wait_for_buffer_space(task);
+
+ spin_unlock_bh(&xprt->transport_lock);
+ } else
+ /* Keep holding the socket if it is blocked */
+ rpc_delay(task, HZ>>4);
+}
+
+/**
+ * xs_udp_send_request - write an RPC request to a UDP socket
+ * @task: address of RPC task that manages the state of an RPC request
+ *
+ * Return values:
+ * 0: The request has been sent
+ * EAGAIN: The socket was blocked, please call again later to
+ * complete the request
+ * ENOTCONN: Caller needs to invoke connect logic then call again
+ * other: Some other error occured, the request was not sent
+ */
+static int xs_udp_send_request(struct rpc_task *task)
+{
+ struct rpc_rqst *req = task->tk_rqstp;
+ struct rpc_xprt *xprt = req->rq_xprt;
+ struct xdr_buf *xdr = &req->rq_snd_buf;
+ int status;
+
+ xs_pktdump("packet data:",
+ req->rq_svec->iov_base,
+ req->rq_svec->iov_len);
+
+ req->rq_xtime = jiffies;
+ status = xs_sendpages(xprt->sock, (struct sockaddr *) &xprt->addr,
+ sizeof(xprt->addr), xdr, req->rq_bytes_sent);
+
+ dprintk("RPC: xs_udp_send_request(%u) = %d\n",
+ xdr->len - req->rq_bytes_sent, status);
+
+ if (likely(status >= (int) req->rq_slen))
+ return 0;
+
+ /* Still some bytes left; set up for a retry later. */
+ if (status > 0)
+ status = -EAGAIN;
+
+ switch (status) {
+ case -ENETUNREACH:
+ case -EPIPE:
+ case -ECONNREFUSED:
+ /* When the server has died, an ICMP port unreachable message
+ * prompts ECONNREFUSED. */
+ break;
+ case -EAGAIN:
+ xs_nospace(task);
+ break;
+ default:
+ dprintk("RPC: sendmsg returned unrecognized error %d\n",
+ -status);
+ break;
+ }
+
+ return status;
+}
+
+static inline void xs_encode_tcp_record_marker(struct xdr_buf *buf)
+{
+ u32 reclen = buf->len - sizeof(rpc_fraghdr);
+ rpc_fraghdr *base = buf->head[0].iov_base;
+ *base = htonl(RPC_LAST_STREAM_FRAGMENT | reclen);
+}
+
+/**
+ * xs_tcp_send_request - write an RPC request to a TCP socket
+ * @task: address of RPC task that manages the state of an RPC request
+ *
+ * Return values:
+ * 0: The request has been sent
+ * EAGAIN: The socket was blocked, please call again later to
+ * complete the request
+ * ENOTCONN: Caller needs to invoke connect logic then call again
+ * other: Some other error occured, the request was not sent
+ *
+ * XXX: In the case of soft timeouts, should we eventually give up
+ * if sendmsg is not able to make progress?
+ */
+static int xs_tcp_send_request(struct rpc_task *task)
+{
+ struct rpc_rqst *req = task->tk_rqstp;
+ struct rpc_xprt *xprt = req->rq_xprt;
+ struct xdr_buf *xdr = &req->rq_snd_buf;
+ int status, retry = 0;
+
+ xs_encode_tcp_record_marker(&req->rq_snd_buf);
+
+ xs_pktdump("packet data:",
+ req->rq_svec->iov_base,
+ req->rq_svec->iov_len);
+
+ /* Continue transmitting the packet/record. We must be careful
+ * to cope with writespace callbacks arriving _after_ we have
+ * called sendmsg(). */
+ while (1) {
+ req->rq_xtime = jiffies;
+ status = xs_sendpages(xprt->sock, NULL, 0, xdr,
+ req->rq_bytes_sent);
+
+ dprintk("RPC: xs_tcp_send_request(%u) = %d\n",
+ xdr->len - req->rq_bytes_sent, status);
+
+ if (unlikely(status < 0))
+ break;
+
+ /* If we've sent the entire packet, immediately
+ * reset the count of bytes sent. */
+ req->rq_bytes_sent += status;
+ if (likely(req->rq_bytes_sent >= req->rq_slen)) {
+ req->rq_bytes_sent = 0;
+ return 0;
+ }
+
+ status = -EAGAIN;
+ if (retry++ > XS_SENDMSG_RETRY)
+ break;
+ }
+
+ switch (status) {
+ case -EAGAIN:
+ xs_nospace(task);
+ break;
+ case -ECONNREFUSED:
+ case -ECONNRESET:
+ case -ENOTCONN:
+ case -EPIPE:
+ status = -ENOTCONN;
+ break;
+ default:
+ dprintk("RPC: sendmsg returned unrecognized error %d\n",
+ -status);
+ xprt_disconnect(xprt);
+ break;
+ }
+
+ return status;
+}
+
+/**
+ * xs_close - close a socket
+ * @xprt: transport
+ *
+ * This is used when all requests are complete; ie, no DRC state remains
+ * on the server we want to save.
+ */
+static void xs_close(struct rpc_xprt *xprt)
+{
+ struct socket *sock = xprt->sock;
+ struct sock *sk = xprt->inet;
+
+ if (!sk)
+ return;
+
+ dprintk("RPC: xs_close xprt %p\n", xprt);
+
+ write_lock_bh(&sk->sk_callback_lock);
+ xprt->inet = NULL;
+ xprt->sock = NULL;
+
+ sk->sk_user_data = NULL;
+ sk->sk_data_ready = xprt->old_data_ready;
+ sk->sk_state_change = xprt->old_state_change;
+ sk->sk_write_space = xprt->old_write_space;
+ write_unlock_bh(&sk->sk_callback_lock);
+
+ sk->sk_no_check = 0;
+
+ sock_release(sock);
+}
+
+/**
+ * xs_destroy - prepare to shutdown a transport
+ * @xprt: doomed transport
+ *
+ */
+static void xs_destroy(struct rpc_xprt *xprt)
+{
+ dprintk("RPC: xs_destroy xprt %p\n", xprt);
+
+ cancel_delayed_work(&xprt->connect_worker);
+ flush_scheduled_work();
+
+ xprt_disconnect(xprt);
+ xs_close(xprt);
+ kfree(xprt->slot);
+}
+
+static inline struct rpc_xprt *xprt_from_sock(struct sock *sk)
+{
+ return (struct rpc_xprt *) sk->sk_user_data;
+}
+
+/**
+ * xs_udp_data_ready - "data ready" callback for UDP sockets
+ * @sk: socket with data to read
+ * @len: how much data to read
+ *
+ */
+static void xs_udp_data_ready(struct sock *sk, int len)
+{
+ struct rpc_task *task;
+ struct rpc_xprt *xprt;
+ struct rpc_rqst *rovr;
+ struct sk_buff *skb;
+ int err, repsize, copied;
+ u32 _xid, *xp;
+
+ read_lock(&sk->sk_callback_lock);
+ dprintk("RPC: xs_udp_data_ready...\n");
+ if (!(xprt = xprt_from_sock(sk)))
+ goto out;
+
+ if ((skb = skb_recv_datagram(sk, 0, 1, &err)) == NULL)
+ goto out;
+
+ if (xprt->shutdown)
+ goto dropit;
+
+ repsize = skb->len - sizeof(struct udphdr);
+ if (repsize < 4) {
+ dprintk("RPC: impossible RPC reply size %d!\n", repsize);
+ goto dropit;
+ }
+
+ /* Copy the XID from the skb... */
+ xp = skb_header_pointer(skb, sizeof(struct udphdr),
+ sizeof(_xid), &_xid);
+ if (xp == NULL)
+ goto dropit;
+
+ /* Look up and lock the request corresponding to the given XID */
+ spin_lock(&xprt->transport_lock);
+ rovr = xprt_lookup_rqst(xprt, *xp);
+ if (!rovr)
+ goto out_unlock;
+ task = rovr->rq_task;
+
+ if ((copied = rovr->rq_private_buf.buflen) > repsize)
+ copied = repsize;
+
+ /* Suck it into the iovec, verify checksum if not done by hw. */
+ if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb))
+ goto out_unlock;
+
+ /* Something worked... */
+ dst_confirm(skb->dst);
+
+ xprt_adjust_cwnd(task, copied);
+ xprt_update_rtt(task);
+ xprt_complete_rqst(task, copied);
+
+ out_unlock:
+ spin_unlock(&xprt->transport_lock);
+ dropit:
+ skb_free_datagram(sk, skb);
+ out:
+ read_unlock(&sk->sk_callback_lock);
+}
+
+static inline size_t xs_tcp_copy_data(skb_reader_t *desc, void *p, size_t len)
+{
+ if (len > desc->count)
+ len = desc->count;
+ if (skb_copy_bits(desc->skb, desc->offset, p, len)) {
+ dprintk("RPC: failed to copy %zu bytes from skb. %zu bytes remain\n",
+ len, desc->count);
+ return 0;
+ }
+ desc->offset += len;
+ desc->count -= len;
+ dprintk("RPC: copied %zu bytes from skb. %zu bytes remain\n",
+ len, desc->count);
+ return len;
+}
+
+static inline void xs_tcp_read_fraghdr(struct rpc_xprt *xprt, skb_reader_t *desc)
+{
+ size_t len, used;
+ char *p;
+
+ p = ((char *) &xprt->tcp_recm) + xprt->tcp_offset;
+ len = sizeof(xprt->tcp_recm) - xprt->tcp_offset;
+ used = xs_tcp_copy_data(desc, p, len);
+ xprt->tcp_offset += used;
+ if (used != len)
+ return;
+
+ xprt->tcp_reclen = ntohl(xprt->tcp_recm);
+ if (xprt->tcp_reclen & RPC_LAST_STREAM_FRAGMENT)
+ xprt->tcp_flags |= XPRT_LAST_FRAG;
+ else
+ xprt->tcp_flags &= ~XPRT_LAST_FRAG;
+ xprt->tcp_reclen &= RPC_FRAGMENT_SIZE_MASK;
+
+ xprt->tcp_flags &= ~XPRT_COPY_RECM;
+ xprt->tcp_offset = 0;
+
+ /* Sanity check of the record length */
+ if (unlikely(xprt->tcp_reclen < 4)) {
+ dprintk("RPC: invalid TCP record fragment length\n");
+ xprt_disconnect(xprt);
+ return;
+ }
+ dprintk("RPC: reading TCP record fragment of length %d\n",
+ xprt->tcp_reclen);
+}
+
+static void xs_tcp_check_recm(struct rpc_xprt *xprt)
+{
+ dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u, tcp_flags = %lx\n",
+ xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen, xprt->tcp_flags);
+ if (xprt->tcp_offset == xprt->tcp_reclen) {
+ xprt->tcp_flags |= XPRT_COPY_RECM;
+ xprt->tcp_offset = 0;
+ if (xprt->tcp_flags & XPRT_LAST_FRAG) {
+ xprt->tcp_flags &= ~XPRT_COPY_DATA;
+ xprt->tcp_flags |= XPRT_COPY_XID;
+ xprt->tcp_copied = 0;
+ }
+ }
+}
+
+static inline void xs_tcp_read_xid(struct rpc_xprt *xprt, skb_reader_t *desc)
+{
+ size_t len, used;
+ char *p;
+
+ len = sizeof(xprt->tcp_xid) - xprt->tcp_offset;
+ dprintk("RPC: reading XID (%Zu bytes)\n", len);
+ p = ((char *) &xprt->tcp_xid) + xprt->tcp_offset;
+ used = xs_tcp_copy_data(desc, p, len);
+ xprt->tcp_offset += used;
+ if (used != len)
+ return;
+ xprt->tcp_flags &= ~XPRT_COPY_XID;
+ xprt->tcp_flags |= XPRT_COPY_DATA;
+ xprt->tcp_copied = 4;
+ dprintk("RPC: reading reply for XID %08x\n",
+ ntohl(xprt->tcp_xid));
+ xs_tcp_check_recm(xprt);
+}
+
+static inline void xs_tcp_read_request(struct rpc_xprt *xprt, skb_reader_t *desc)
+{
+ struct rpc_rqst *req;
+ struct xdr_buf *rcvbuf;
+ size_t len;
+ ssize_t r;
+
+ /* Find and lock the request corresponding to this xid */
+ spin_lock(&xprt->transport_lock);
+ req = xprt_lookup_rqst(xprt, xprt->tcp_xid);
+ if (!req) {
+ xprt->tcp_flags &= ~XPRT_COPY_DATA;
+ dprintk("RPC: XID %08x request not found!\n",
+ ntohl(xprt->tcp_xid));
+ spin_unlock(&xprt->transport_lock);
+ return;
+ }
+
+ rcvbuf = &req->rq_private_buf;
+ len = desc->count;
+ if (len > xprt->tcp_reclen - xprt->tcp_offset) {
+ skb_reader_t my_desc;
+
+ len = xprt->tcp_reclen - xprt->tcp_offset;
+ memcpy(&my_desc, desc, sizeof(my_desc));
+ my_desc.count = len;
+ r = xdr_partial_copy_from_skb(rcvbuf, xprt->tcp_copied,
+ &my_desc, xs_tcp_copy_data);
+ desc->count -= r;
+ desc->offset += r;
+ } else
+ r = xdr_partial_copy_from_skb(rcvbuf, xprt->tcp_copied,
+ desc, xs_tcp_copy_data);
+
+ if (r > 0) {
+ xprt->tcp_copied += r;
+ xprt->tcp_offset += r;
+ }
+ if (r != len) {
+ /* Error when copying to the receive buffer,
+ * usually because we weren't able to allocate
+ * additional buffer pages. All we can do now
+ * is turn off XPRT_COPY_DATA, so the request
+ * will not receive any additional updates,
+ * and time out.
+ * Any remaining data from this record will
+ * be discarded.
+ */
+ xprt->tcp_flags &= ~XPRT_COPY_DATA;
+ dprintk("RPC: XID %08x truncated request\n",
+ ntohl(xprt->tcp_xid));
+ dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u\n",
+ xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen);
+ goto out;
+ }
+
+ dprintk("RPC: XID %08x read %Zd bytes\n",
+ ntohl(xprt->tcp_xid), r);
+ dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u\n",
+ xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen);
+
+ if (xprt->tcp_copied == req->rq_private_buf.buflen)
+ xprt->tcp_flags &= ~XPRT_COPY_DATA;
+ else if (xprt->tcp_offset == xprt->tcp_reclen) {
+ if (xprt->tcp_flags & XPRT_LAST_FRAG)
+ xprt->tcp_flags &= ~XPRT_COPY_DATA;
+ }
+
+out:
+ if (!(xprt->tcp_flags & XPRT_COPY_DATA))
+ xprt_complete_rqst(req->rq_task, xprt->tcp_copied);
+ spin_unlock(&xprt->transport_lock);
+ xs_tcp_check_recm(xprt);
+}
+
+static inline void xs_tcp_read_discard(struct rpc_xprt *xprt, skb_reader_t *desc)
+{
+ size_t len;
+
+ len = xprt->tcp_reclen - xprt->tcp_offset;
+ if (len > desc->count)
+ len = desc->count;
+ desc->count -= len;
+ desc->offset += len;
+ xprt->tcp_offset += len;
+ dprintk("RPC: discarded %Zu bytes\n", len);
+ xs_tcp_check_recm(xprt);
+}
+
+static int xs_tcp_data_recv(read_descriptor_t *rd_desc, struct sk_buff *skb, unsigned int offset, size_t len)
+{
+ struct rpc_xprt *xprt = rd_desc->arg.data;
+ skb_reader_t desc = {
+ .skb = skb,
+ .offset = offset,
+ .count = len,
+ .csum = 0
+ };
+
+ dprintk("RPC: xs_tcp_data_recv started\n");
+ do {
+ /* Read in a new fragment marker if necessary */
+ /* Can we ever really expect to get completely empty fragments? */
+ if (xprt->tcp_flags & XPRT_COPY_RECM) {
+ xs_tcp_read_fraghdr(xprt, &desc);
+ continue;
+ }
+ /* Read in the xid if necessary */
+ if (xprt->tcp_flags & XPRT_COPY_XID) {
+ xs_tcp_read_xid(xprt, &desc);
+ continue;
+ }
+ /* Read in the request data */
+ if (xprt->tcp_flags & XPRT_COPY_DATA) {
+ xs_tcp_read_request(xprt, &desc);
+ continue;
+ }
+ /* Skip over any trailing bytes on short reads */
+ xs_tcp_read_discard(xprt, &desc);
+ } while (desc.count);
+ dprintk("RPC: xs_tcp_data_recv done\n");
+ return len - desc.count;
+}
+
+/**
+ * xs_tcp_data_ready - "data ready" callback for TCP sockets
+ * @sk: socket with data to read
+ * @bytes: how much data to read
+ *
+ */
+static void xs_tcp_data_ready(struct sock *sk, int bytes)
+{
+ struct rpc_xprt *xprt;
+ read_descriptor_t rd_desc;
+
+ read_lock(&sk->sk_callback_lock);
+ dprintk("RPC: xs_tcp_data_ready...\n");
+ if (!(xprt = xprt_from_sock(sk)))
+ goto out;
+ if (xprt->shutdown)
+ goto out;
+
+ /* We use rd_desc to pass struct xprt to xs_tcp_data_recv */
+ rd_desc.arg.data = xprt;
+ rd_desc.count = 65536;
+ tcp_read_sock(sk, &rd_desc, xs_tcp_data_recv);
+out:
+ read_unlock(&sk->sk_callback_lock);
+}
+
+/**
+ * xs_tcp_state_change - callback to handle TCP socket state changes
+ * @sk: socket whose state has changed
+ *
+ */
+static void xs_tcp_state_change(struct sock *sk)
+{
+ struct rpc_xprt *xprt;
+
+ read_lock(&sk->sk_callback_lock);
+ if (!(xprt = xprt_from_sock(sk)))
+ goto out;
+ dprintk("RPC: xs_tcp_state_change client %p...\n", xprt);
+ dprintk("RPC: state %x conn %d dead %d zapped %d\n",
+ sk->sk_state, xprt_connected(xprt),
+ sock_flag(sk, SOCK_DEAD),
+ sock_flag(sk, SOCK_ZAPPED));
+
+ switch (sk->sk_state) {
+ case TCP_ESTABLISHED:
+ spin_lock_bh(&xprt->transport_lock);
+ if (!xprt_test_and_set_connected(xprt)) {
+ /* Reset TCP record info */
+ xprt->tcp_offset = 0;
+ xprt->tcp_reclen = 0;
+ xprt->tcp_copied = 0;
+ xprt->tcp_flags = XPRT_COPY_RECM | XPRT_COPY_XID;
+ xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
+ xprt_wake_pending_tasks(xprt, 0);
+ }
+ spin_unlock_bh(&xprt->transport_lock);
+ break;
+ case TCP_SYN_SENT:
+ case TCP_SYN_RECV:
+ break;
+ default:
+ xprt_disconnect(xprt);
+ break;
+ }
+ out:
+ read_unlock(&sk->sk_callback_lock);
+}
+
+/**
+ * xs_udp_write_space - callback invoked when socket buffer space
+ * becomes available
+ * @sk: socket whose state has changed
+ *
+ * Called when more output buffer space is available for this socket.
+ * We try not to wake our writers until they can make "significant"
+ * progress, otherwise we'll waste resources thrashing kernel_sendmsg
+ * with a bunch of small requests.
+ */
+static void xs_udp_write_space(struct sock *sk)
+{
+ read_lock(&sk->sk_callback_lock);
+
+ /* from net/core/sock.c:sock_def_write_space */
+ if (sock_writeable(sk)) {
+ struct socket *sock;
+ struct rpc_xprt *xprt;
+
+ if (unlikely(!(sock = sk->sk_socket)))
+ goto out;
+ if (unlikely(!(xprt = xprt_from_sock(sk))))
+ goto out;
+ if (unlikely(!test_and_clear_bit(SOCK_NOSPACE, &sock->flags)))
+ goto out;
+
+ xprt_write_space(xprt);
+ }
+
+ out:
+ read_unlock(&sk->sk_callback_lock);
+}
+
+/**
+ * xs_tcp_write_space - callback invoked when socket buffer space
+ * becomes available
+ * @sk: socket whose state has changed
+ *
+ * Called when more output buffer space is available for this socket.
+ * We try not to wake our writers until they can make "significant"
+ * progress, otherwise we'll waste resources thrashing kernel_sendmsg
+ * with a bunch of small requests.
+ */
+static void xs_tcp_write_space(struct sock *sk)
+{
+ read_lock(&sk->sk_callback_lock);
+
+ /* from net/core/stream.c:sk_stream_write_space */
+ if (sk_stream_wspace(sk) >= sk_stream_min_wspace(sk)) {
+ struct socket *sock;
+ struct rpc_xprt *xprt;
+
+ if (unlikely(!(sock = sk->sk_socket)))
+ goto out;
+ if (unlikely(!(xprt = xprt_from_sock(sk))))
+ goto out;
+ if (unlikely(!test_and_clear_bit(SOCK_NOSPACE, &sock->flags)))
+ goto out;
+
+ xprt_write_space(xprt);
+ }
+
+ out:
+ read_unlock(&sk->sk_callback_lock);
+}
+
+static void xs_udp_do_set_buffer_size(struct rpc_xprt *xprt)
+{
+ struct sock *sk = xprt->inet;
+
+ if (xprt->rcvsize) {
+ sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
+ sk->sk_rcvbuf = xprt->rcvsize * xprt->max_reqs * 2;
+ }
+ if (xprt->sndsize) {
+ sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
+ sk->sk_sndbuf = xprt->sndsize * xprt->max_reqs * 2;
+ sk->sk_write_space(sk);
+ }
+}
+
+/**
+ * xs_udp_set_buffer_size - set send and receive limits
+ * @xprt: generic transport
+ * @sndsize: requested size of send buffer, in bytes
+ * @rcvsize: requested size of receive buffer, in bytes
+ *
+ * Set socket send and receive buffer size limits.
+ */
+static void xs_udp_set_buffer_size(struct rpc_xprt *xprt, size_t sndsize, size_t rcvsize)
+{
+ xprt->sndsize = 0;
+ if (sndsize)
+ xprt->sndsize = sndsize + 1024;
+ xprt->rcvsize = 0;
+ if (rcvsize)
+ xprt->rcvsize = rcvsize + 1024;
+
+ xs_udp_do_set_buffer_size(xprt);
+}
+
+/**
+ * xs_udp_timer - called when a retransmit timeout occurs on a UDP transport
+ * @task: task that timed out
+ *
+ * Adjust the congestion window after a retransmit timeout has occurred.
+ */
+static void xs_udp_timer(struct rpc_task *task)
+{
+ xprt_adjust_cwnd(task, -ETIMEDOUT);
+}
+
+static int xs_bindresvport(struct rpc_xprt *xprt, struct socket *sock)
+{
+ struct sockaddr_in myaddr = {
+ .sin_family = AF_INET,
+ };
+ int err;
+ unsigned short port = xprt->port;
+
+ do {
+ myaddr.sin_port = htons(port);
+ err = sock->ops->bind(sock, (struct sockaddr *) &myaddr,
+ sizeof(myaddr));
+ if (err == 0) {
+ xprt->port = port;
+ dprintk("RPC: xs_bindresvport bound to port %u\n",
+ port);
+ return 0;
+ }
+ if (port <= xprt_min_resvport)
+ port = xprt_max_resvport;
+ else
+ port--;
+ } while (err == -EADDRINUSE && port != xprt->port);
+
+ dprintk("RPC: can't bind to reserved port (%d).\n", -err);
+ return err;
+}
+
+/**
+ * xs_udp_connect_worker - set up a UDP socket
+ * @args: RPC transport to connect
+ *
+ * Invoked by a work queue tasklet.
+ */
+static void xs_udp_connect_worker(void *args)
+{
+ struct rpc_xprt *xprt = (struct rpc_xprt *) args;
+ struct socket *sock = xprt->sock;
+ int err, status = -EIO;
+
+ if (xprt->shutdown || xprt->addr.sin_port == 0)
+ goto out;
+
+ dprintk("RPC: xs_udp_connect_worker for xprt %p\n", xprt);
+
+ /* Start by resetting any existing state */
+ xs_close(xprt);
+
+ if ((err = sock_create_kern(PF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock)) < 0) {
+ dprintk("RPC: can't create UDP transport socket (%d).\n", -err);
+ goto out;
+ }
+
+ if (xprt->resvport && xs_bindresvport(xprt, sock) < 0) {
+ sock_release(sock);
+ goto out;
+ }
+
+ if (!xprt->inet) {
+ struct sock *sk = sock->sk;
+
+ write_lock_bh(&sk->sk_callback_lock);
+
+ sk->sk_user_data = xprt;
+ xprt->old_data_ready = sk->sk_data_ready;
+ xprt->old_state_change = sk->sk_state_change;
+ xprt->old_write_space = sk->sk_write_space;
+ sk->sk_data_ready = xs_udp_data_ready;
+ sk->sk_write_space = xs_udp_write_space;
+ sk->sk_no_check = UDP_CSUM_NORCV;
+
+ xprt_set_connected(xprt);
+
+ /* Reset to new socket */
+ xprt->sock = sock;
+ xprt->inet = sk;
+
+ write_unlock_bh(&sk->sk_callback_lock);
+ }
+ xs_udp_do_set_buffer_size(xprt);
+ status = 0;
+out:
+ xprt_wake_pending_tasks(xprt, status);
+ xprt_clear_connecting(xprt);
+}
+
+/*
+ * We need to preserve the port number so the reply cache on the server can
+ * find our cached RPC replies when we get around to reconnecting.
+ */
+static void xs_tcp_reuse_connection(struct rpc_xprt *xprt)
+{
+ int result;
+ struct socket *sock = xprt->sock;
+ struct sockaddr any;
+
+ dprintk("RPC: disconnecting xprt %p to reuse port\n", xprt);
+
+ /*
+ * Disconnect the transport socket by doing a connect operation
+ * with AF_UNSPEC. This should return immediately...
+ */
+ memset(&any, 0, sizeof(any));
+ any.sa_family = AF_UNSPEC;
+ result = sock->ops->connect(sock, &any, sizeof(any), 0);
+ if (result)
+ dprintk("RPC: AF_UNSPEC connect return code %d\n",
+ result);
+}
+
+/**
+ * xs_tcp_connect_worker - connect a TCP socket to a remote endpoint
+ * @args: RPC transport to connect
+ *
+ * Invoked by a work queue tasklet.
+ */
+static void xs_tcp_connect_worker(void *args)
+{
+ struct rpc_xprt *xprt = (struct rpc_xprt *)args;
+ struct socket *sock = xprt->sock;
+ int err, status = -EIO;
+
+ if (xprt->shutdown || xprt->addr.sin_port == 0)
+ goto out;
+
+ dprintk("RPC: xs_tcp_connect_worker for xprt %p\n", xprt);
+
+ if (!xprt->sock) {
+ /* start from scratch */
+ if ((err = sock_create_kern(PF_INET, SOCK_STREAM, IPPROTO_TCP, &sock)) < 0) {
+ dprintk("RPC: can't create TCP transport socket (%d).\n", -err);
+ goto out;
+ }
+
+ if (xprt->resvport && xs_bindresvport(xprt, sock) < 0) {
+ sock_release(sock);
+ goto out;
+ }
+ } else
+ /* "close" the socket, preserving the local port */
+ xs_tcp_reuse_connection(xprt);
+
+ if (!xprt->inet) {
+ struct sock *sk = sock->sk;
+
+ write_lock_bh(&sk->sk_callback_lock);
+
+ sk->sk_user_data = xprt;
+ xprt->old_data_ready = sk->sk_data_ready;
+ xprt->old_state_change = sk->sk_state_change;
+ xprt->old_write_space = sk->sk_write_space;
+ sk->sk_data_ready = xs_tcp_data_ready;
+ sk->sk_state_change = xs_tcp_state_change;
+ sk->sk_write_space = xs_tcp_write_space;
+
+ /* socket options */
+ sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
+ sock_reset_flag(sk, SOCK_LINGER);
+ tcp_sk(sk)->linger2 = 0;
+ tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF;
+
+ xprt_clear_connected(xprt);
+
+ /* Reset to new socket */
+ xprt->sock = sock;
+ xprt->inet = sk;
+
+ write_unlock_bh(&sk->sk_callback_lock);
+ }
+
+ /* Tell the socket layer to start connecting... */
+ status = sock->ops->connect(sock, (struct sockaddr *) &xprt->addr,
+ sizeof(xprt->addr), O_NONBLOCK);
+ dprintk("RPC: %p connect status %d connected %d sock state %d\n",
+ xprt, -status, xprt_connected(xprt), sock->sk->sk_state);
+ if (status < 0) {
+ switch (status) {
+ case -EINPROGRESS:
+ case -EALREADY:
+ goto out_clear;
+ case -ECONNREFUSED:
+ case -ECONNRESET:
+ /* retry with existing socket, after a delay */
+ break;
+ default:
+ /* get rid of existing socket, and retry */
+ xs_close(xprt);
+ break;
+ }
+ }
+out:
+ xprt_wake_pending_tasks(xprt, status);
+out_clear:
+ xprt_clear_connecting(xprt);
+}
+
+/**
+ * xs_connect - connect a socket to a remote endpoint
+ * @task: address of RPC task that manages state of connect request
+ *
+ * TCP: If the remote end dropped the connection, delay reconnecting.
+ *
+ * UDP socket connects are synchronous, but we use a work queue anyway
+ * to guarantee that even unprivileged user processes can set up a
+ * socket on a privileged port.
+ *
+ * If a UDP socket connect fails, the delay behavior here prevents
+ * retry floods (hard mounts).
+ */
+static void xs_connect(struct rpc_task *task)
+{
+ struct rpc_xprt *xprt = task->tk_xprt;
+
+ if (xprt_test_and_set_connecting(xprt))
+ return;
+
+ if (xprt->sock != NULL) {
+ dprintk("RPC: xs_connect delayed xprt %p for %lu seconds\n",
+ xprt, xprt->reestablish_timeout / HZ);
+ schedule_delayed_work(&xprt->connect_worker,
+ xprt->reestablish_timeout);
+ xprt->reestablish_timeout <<= 1;
+ if (xprt->reestablish_timeout > XS_TCP_MAX_REEST_TO)
+ xprt->reestablish_timeout = XS_TCP_MAX_REEST_TO;
+ } else {
+ dprintk("RPC: xs_connect scheduled xprt %p\n", xprt);
+ schedule_work(&xprt->connect_worker);
+
+ /* flush_scheduled_work can sleep... */
+ if (!RPC_IS_ASYNC(task))
+ flush_scheduled_work();
+ }
+}
+
+static struct rpc_xprt_ops xs_udp_ops = {
+ .set_buffer_size = xs_udp_set_buffer_size,
+ .reserve_xprt = xprt_reserve_xprt_cong,
+ .release_xprt = xprt_release_xprt_cong,
+ .connect = xs_connect,
+ .send_request = xs_udp_send_request,
+ .set_retrans_timeout = xprt_set_retrans_timeout_rtt,
+ .timer = xs_udp_timer,
+ .release_request = xprt_release_rqst_cong,
+ .close = xs_close,
+ .destroy = xs_destroy,
+};
+
+static struct rpc_xprt_ops xs_tcp_ops = {
+ .reserve_xprt = xprt_reserve_xprt,
+ .release_xprt = xprt_release_xprt,
+ .connect = xs_connect,
+ .send_request = xs_tcp_send_request,
+ .set_retrans_timeout = xprt_set_retrans_timeout_def,
+ .close = xs_close,
+ .destroy = xs_destroy,
+};
+
+/**
+ * xs_setup_udp - Set up transport to use a UDP socket
+ * @xprt: transport to set up
+ * @to: timeout parameters
+ *
+ */
+int xs_setup_udp(struct rpc_xprt *xprt, struct rpc_timeout *to)
+{
+ size_t slot_table_size;
+
+ dprintk("RPC: setting up udp-ipv4 transport...\n");
+
+ xprt->max_reqs = xprt_udp_slot_table_entries;
+ slot_table_size = xprt->max_reqs * sizeof(xprt->slot[0]);
+ xprt->slot = kmalloc(slot_table_size, GFP_KERNEL);
+ if (xprt->slot == NULL)
+ return -ENOMEM;
+ memset(xprt->slot, 0, slot_table_size);
+
+ xprt->prot = IPPROTO_UDP;
+ xprt->port = xprt_max_resvport;
+ xprt->tsh_size = 0;
+ xprt->resvport = capable(CAP_NET_BIND_SERVICE) ? 1 : 0;
+ /* XXX: header size can vary due to auth type, IPv6, etc. */
+ xprt->max_payload = (1U << 16) - (MAX_HEADER << 3);
+
+ INIT_WORK(&xprt->connect_worker, xs_udp_connect_worker, xprt);
+ xprt->bind_timeout = XS_BIND_TO;
+ xprt->connect_timeout = XS_UDP_CONN_TO;
+ xprt->reestablish_timeout = XS_UDP_REEST_TO;
+ xprt->idle_timeout = XS_IDLE_DISC_TO;
+
+ xprt->ops = &xs_udp_ops;
+
+ if (to)
+ xprt->timeout = *to;
+ else
+ xprt_set_timeout(&xprt->timeout, 5, 5 * HZ);
+
+ return 0;
+}
+
+/**
+ * xs_setup_tcp - Set up transport to use a TCP socket
+ * @xprt: transport to set up
+ * @to: timeout parameters
+ *
+ */
+int xs_setup_tcp(struct rpc_xprt *xprt, struct rpc_timeout *to)
+{
+ size_t slot_table_size;
+
+ dprintk("RPC: setting up tcp-ipv4 transport...\n");
+
+ xprt->max_reqs = xprt_tcp_slot_table_entries;
+ slot_table_size = xprt->max_reqs * sizeof(xprt->slot[0]);
+ xprt->slot = kmalloc(slot_table_size, GFP_KERNEL);
+ if (xprt->slot == NULL)
+ return -ENOMEM;
+ memset(xprt->slot, 0, slot_table_size);
+
+ xprt->prot = IPPROTO_TCP;
+ xprt->port = xprt_max_resvport;
+ xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32);
+ xprt->resvport = capable(CAP_NET_BIND_SERVICE) ? 1 : 0;
+ xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
+
+ INIT_WORK(&xprt->connect_worker, xs_tcp_connect_worker, xprt);
+ xprt->bind_timeout = XS_BIND_TO;
+ xprt->connect_timeout = XS_TCP_CONN_TO;
+ xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
+ xprt->idle_timeout = XS_IDLE_DISC_TO;
+
+ xprt->ops = &xs_tcp_ops;
+
+ if (to)
+ xprt->timeout = *to;
+ else
+ xprt_set_timeout(&xprt->timeout, 2, 60 * HZ);
+
+ return 0;
+}
