Commit 488b5ec8 authored by Linus Torvalds's avatar Linus Torvalds

Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ericvh/v9fs

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ericvh/v9fs:
  9p: fix p9_printfcall export
  9p: transport API reorganization
  9p: add remove function to trans_virtio
  9p: Convert semaphore to spinlock for p9_idpool
  9p: fix mmap to be read-only
  9p: add support for sticky bit
  9p: Fix soft lockup in virtio transport
  9p: fix bug in attach-per-user
  9p: block-based virtio client
  9p: create transport rpc cut-thru
  9p: fix bug in p9_clone_stat
parents a80a438b 72767443
......@@ -175,7 +175,7 @@ struct p9_fid *v9fs_fid_lookup(struct dentry *dentry)
if (!wnames)
return ERR_PTR(-ENOMEM);
for (d = dentry, i = n; i >= 0; i--, d = d->d_parent)
for (d = dentry, i = (n-1); i >= 0; i--, d = d->d_parent)
wnames[i] = (char *) d->d_name.name;
clone = 1;
......@@ -183,7 +183,7 @@ struct p9_fid *v9fs_fid_lookup(struct dentry *dentry)
while (i < n) {
l = min(n - i, P9_MAXWELEM);
fid = p9_client_walk(fid, l, &wnames[i], clone);
if (!fid) {
if (IS_ERR(fid)) {
kfree(wnames);
return fid;
}
......
......@@ -3,7 +3,7 @@
*
* This file contains functions assisting in mapping VFS to 9P2000
*
* Copyright (C) 2004 by Eric Van Hensbergen <ericvh@gmail.com>
* Copyright (C) 2004-2008 by Eric Van Hensbergen <ericvh@gmail.com>
* Copyright (C) 2002 by Ron Minnich <rminnich@lanl.gov>
*
* This program is free software; you can redistribute it and/or modify
......@@ -31,7 +31,6 @@
#include <linux/idr.h>
#include <net/9p/9p.h>
#include <net/9p/transport.h>
#include <net/9p/conn.h>
#include <net/9p/client.h>
#include "v9fs.h"
#include "v9fs_vfs.h"
......@@ -43,11 +42,11 @@
enum {
/* Options that take integer arguments */
Opt_debug, Opt_msize, Opt_dfltuid, Opt_dfltgid, Opt_afid,
Opt_debug, Opt_dfltuid, Opt_dfltgid, Opt_afid,
/* String options */
Opt_uname, Opt_remotename, Opt_trans,
/* Options that take no arguments */
Opt_legacy, Opt_nodevmap,
Opt_nodevmap,
/* Cache options */
Opt_cache_loose,
/* Access options */
......@@ -58,14 +57,11 @@ enum {
static match_table_t tokens = {
{Opt_debug, "debug=%x"},
{Opt_msize, "msize=%u"},
{Opt_dfltuid, "dfltuid=%u"},
{Opt_dfltgid, "dfltgid=%u"},
{Opt_afid, "afid=%u"},
{Opt_uname, "uname=%s"},
{Opt_remotename, "aname=%s"},
{Opt_trans, "trans=%s"},
{Opt_legacy, "noextend"},
{Opt_nodevmap, "nodevmap"},
{Opt_cache_loose, "cache=loose"},
{Opt_cache_loose, "loose"},
......@@ -85,16 +81,14 @@ static void v9fs_parse_options(struct v9fs_session_info *v9ses)
char *options;
substring_t args[MAX_OPT_ARGS];
char *p;
int option;
int ret;
int option = 0;
char *s, *e;
int ret;
/* setup defaults */
v9ses->maxdata = 8192;
v9ses->afid = ~0;
v9ses->debug = 0;
v9ses->cache = 0;
v9ses->trans = v9fs_default_trans();
if (!v9ses->options)
return;
......@@ -106,7 +100,8 @@ static void v9fs_parse_options(struct v9fs_session_info *v9ses)
continue;
token = match_token(p, tokens, args);
if (token < Opt_uname) {
if ((ret = match_int(&args[0], &option)) < 0) {
ret = match_int(&args[0], &option);
if (ret < 0) {
P9_DPRINTK(P9_DEBUG_ERROR,
"integer field, but no integer?\n");
continue;
......@@ -119,9 +114,7 @@ static void v9fs_parse_options(struct v9fs_session_info *v9ses)
p9_debug_level = option;
#endif
break;
case Opt_msize:
v9ses->maxdata = option;
break;
case Opt_dfltuid:
v9ses->dfltuid = option;
break;
......@@ -131,18 +124,12 @@ static void v9fs_parse_options(struct v9fs_session_info *v9ses)
case Opt_afid:
v9ses->afid = option;
break;
case Opt_trans:
v9ses->trans = v9fs_match_trans(&args[0]);
break;
case Opt_uname:
match_strcpy(v9ses->uname, &args[0]);
break;
case Opt_remotename:
match_strcpy(v9ses->aname, &args[0]);
break;
case Opt_legacy:
v9ses->flags &= ~V9FS_EXTENDED;
break;
case Opt_nodevmap:
v9ses->nodev = 1;
break;
......@@ -185,7 +172,6 @@ struct p9_fid *v9fs_session_init(struct v9fs_session_info *v9ses,
const char *dev_name, char *data)
{
int retval = -EINVAL;
struct p9_trans *trans = NULL;
struct p9_fid *fid;
v9ses->uname = __getname();
......@@ -207,24 +193,7 @@ struct p9_fid *v9fs_session_init(struct v9fs_session_info *v9ses,
v9ses->options = kstrdup(data, GFP_KERNEL);
v9fs_parse_options(v9ses);
if (v9ses->trans == NULL) {
retval = -EPROTONOSUPPORT;
P9_DPRINTK(P9_DEBUG_ERROR,
"No transport defined or default transport\n");
goto error;
}
trans = v9ses->trans->create(dev_name, v9ses->options);
if (IS_ERR(trans)) {
retval = PTR_ERR(trans);
trans = NULL;
goto error;
}
if ((v9ses->maxdata+P9_IOHDRSZ) > v9ses->trans->maxsize)
v9ses->maxdata = v9ses->trans->maxsize-P9_IOHDRSZ;
v9ses->clnt = p9_client_create(trans, v9ses->maxdata+P9_IOHDRSZ,
v9fs_extended(v9ses));
v9ses->clnt = p9_client_create(dev_name, v9ses->options);
if (IS_ERR(v9ses->clnt)) {
retval = PTR_ERR(v9ses->clnt);
......@@ -236,6 +205,8 @@ struct p9_fid *v9fs_session_init(struct v9fs_session_info *v9ses,
if (!v9ses->clnt->dotu)
v9ses->flags &= ~V9FS_EXTENDED;
v9ses->maxdata = v9ses->clnt->msize;
/* for legacy mode, fall back to V9FS_ACCESS_ANY */
if (!v9fs_extended(v9ses) &&
((v9ses->flags&V9FS_ACCESS_MASK) == V9FS_ACCESS_USER)) {
......
/*
* V9FS definitions.
*
* Copyright (C) 2004 by Eric Van Hensbergen <ericvh@gmail.com>
* Copyright (C) 2004-2008 by Eric Van Hensbergen <ericvh@gmail.com>
* Copyright (C) 2002 by Ron Minnich <rminnich@lanl.gov>
*
* This program is free software; you can redistribute it and/or modify
......@@ -28,7 +28,6 @@
struct v9fs_session_info {
/* options */
unsigned int maxdata;
unsigned char flags; /* session flags */
unsigned char nodev; /* set to 1 if no disable device mapping */
unsigned short debug; /* debug level */
......@@ -38,10 +37,10 @@ struct v9fs_session_info {
char *options; /* copy of mount options */
char *uname; /* user name to mount as */
char *aname; /* name of remote hierarchy being mounted */
unsigned int maxdata; /* max data for client interface */
unsigned int dfltuid; /* default uid/muid for legacy support */
unsigned int dfltgid; /* default gid for legacy support */
u32 uid; /* if ACCESS_SINGLE, the uid that has access */
struct p9_trans_module *trans; /* 9p transport */
struct p9_client *clnt; /* 9p client */
struct dentry *debugfs_dir;
};
......
......@@ -184,7 +184,7 @@ static const struct file_operations v9fs_cached_file_operations = {
.open = v9fs_file_open,
.release = v9fs_dir_release,
.lock = v9fs_file_lock,
.mmap = generic_file_mmap,
.mmap = generic_file_readonly_mmap,
};
const struct file_operations v9fs_file_operations = {
......@@ -194,5 +194,5 @@ const struct file_operations v9fs_file_operations = {
.open = v9fs_file_open,
.release = v9fs_dir_release,
.lock = v9fs_file_lock,
.mmap = generic_file_mmap,
.mmap = generic_file_readonly_mmap,
};
......@@ -77,6 +77,8 @@ static int unixmode2p9mode(struct v9fs_session_info *v9ses, int mode)
res |= P9_DMSETUID;
if ((mode & S_ISGID) == S_ISGID)
res |= P9_DMSETGID;
if ((mode & S_ISVTX) == S_ISVTX)
res |= P9_DMSETVTX;
if ((mode & P9_DMLINK))
res |= P9_DMLINK;
}
......@@ -119,6 +121,9 @@ static int p9mode2unixmode(struct v9fs_session_info *v9ses, int mode)
if ((mode & P9_DMSETGID) == P9_DMSETGID)
res |= S_ISGID;
if ((mode & P9_DMSETVTX) == P9_DMSETVTX)
res |= S_ISVTX;
}
return res;
......
......@@ -124,6 +124,7 @@ enum {
P9_DMSOCKET = 0x00100000,
P9_DMSETUID = 0x00080000,
P9_DMSETGID = 0x00040000,
P9_DMSETVTX = 0x00010000,
};
/* qid.types */
......
......@@ -3,6 +3,7 @@
*
* 9P Client Definitions
*
* Copyright (C) 2008 by Eric Van Hensbergen <ericvh@gmail.com>
* Copyright (C) 2007 by Latchesar Ionkov <lucho@ionkov.net>
*
* This program is free software; you can redistribute it and/or modify
......@@ -29,6 +30,7 @@ struct p9_client {
spinlock_t lock; /* protect client structure */
int msize;
unsigned char dotu;
struct p9_trans_module *trans_mod;
struct p9_trans *trans;
struct p9_conn *conn;
......@@ -52,8 +54,7 @@ struct p9_fid {
struct list_head dlist; /* list of all fids attached to a dentry */
};
struct p9_client *p9_client_create(struct p9_trans *trans, int msize,
int dotu);
struct p9_client *p9_client_create(const char *dev_name, char *options);
void p9_client_destroy(struct p9_client *clnt);
void p9_client_disconnect(struct p9_client *clnt);
struct p9_fid *p9_client_attach(struct p9_client *clnt, struct p9_fid *afid,
......
/*
* include/net/9p/conn.h
*
* Connection Definitions
*
* Copyright (C) 2005 by Latchesar Ionkov <lucho@ionkov.net>
* Copyright (C) 2004 by Eric Van Hensbergen <ericvh@gmail.com>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2
* as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to:
* Free Software Foundation
* 51 Franklin Street, Fifth Floor
* Boston, MA 02111-1301 USA
*
*/
#ifndef NET_9P_CONN_H
#define NET_9P_CONN_H
#undef P9_NONBLOCK
struct p9_conn;
struct p9_req;
/**
* p9_mux_req_callback - callback function that is called when the
* response of a request is received. The callback is called from
* a workqueue and shouldn't block.
*
* @req - request
* @a - the pointer that was specified when the request was send to be
* passed to the callback
*/
typedef void (*p9_conn_req_callback)(struct p9_req *req, void *a);
struct p9_conn *p9_conn_create(struct p9_trans *trans, int msize,
unsigned char *dotu);
void p9_conn_destroy(struct p9_conn *);
int p9_conn_rpc(struct p9_conn *m, struct p9_fcall *tc, struct p9_fcall **rc);
#ifdef P9_NONBLOCK
int p9_conn_rpcnb(struct p9_conn *m, struct p9_fcall *tc,
p9_conn_req_callback cb, void *a);
#endif /* P9_NONBLOCK */
void p9_conn_cancel(struct p9_conn *m, int err);
#endif /* NET_9P_CONN_H */
......@@ -4,7 +4,7 @@
* Transport Definition
*
* Copyright (C) 2005 by Latchesar Ionkov <lucho@ionkov.net>
* Copyright (C) 2004 by Eric Van Hensbergen <ericvh@gmail.com>
* Copyright (C) 2004-2008 by Eric Van Hensbergen <ericvh@gmail.com>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2
......@@ -34,11 +34,12 @@ enum p9_trans_status {
struct p9_trans {
enum p9_trans_status status;
int msize;
unsigned char extended;
void *priv;
int (*write) (struct p9_trans *, void *, int);
int (*read) (struct p9_trans *, void *, int);
void (*close) (struct p9_trans *);
unsigned int (*poll)(struct p9_trans *, struct poll_table_struct *);
int (*rpc) (struct p9_trans *t, struct p9_fcall *tc,
struct p9_fcall **rc);
};
struct p9_trans_module {
......@@ -46,7 +47,7 @@ struct p9_trans_module {
char *name; /* name of transport */
int maxsize; /* max message size of transport */
int def; /* this transport should be default */
struct p9_trans * (*create)(const char *devname, char *options);
struct p9_trans * (*create)(const char *, char *, int, unsigned char);
};
void v9fs_register_trans(struct p9_trans_module *m);
......
......@@ -4,7 +4,6 @@ obj-$(CONFIG_NET_9P_VIRTIO) += 9pnet_virtio.o
9pnet-objs := \
mod.o \
mux.o \
client.o \
conv.o \
error.o \
......
......@@ -3,6 +3,7 @@
*
* 9P Client
*
* Copyright (C) 2008 by Eric Van Hensbergen <ericvh@gmail.com>
* Copyright (C) 2007 by Latchesar Ionkov <lucho@ionkov.net>
*
* This program is free software; you can redistribute it and/or modify
......@@ -25,6 +26,7 @@
#include <linux/module.h>
#include <linux/errno.h>
#include <linux/fs.h>
#include <linux/poll.h>
#include <linux/idr.h>
#include <linux/mutex.h>
#include <linux/sched.h>
......@@ -32,15 +34,97 @@
#include <net/9p/9p.h>
#include <linux/parser.h>
#include <net/9p/transport.h>
#include <net/9p/conn.h>
#include <net/9p/client.h>
static struct p9_fid *p9_fid_create(struct p9_client *clnt);
static void p9_fid_destroy(struct p9_fid *fid);
static struct p9_stat *p9_clone_stat(struct p9_stat *st, int dotu);
struct p9_client *p9_client_create(struct p9_trans *trans, int msize,
int dotu)
/*
* Client Option Parsing (code inspired by NFS code)
* - a little lazy - parse all client options
*/
enum {
Opt_msize,
Opt_trans,
Opt_legacy,
Opt_err,
};
static match_table_t tokens = {
{Opt_msize, "msize=%u"},
{Opt_legacy, "noextend"},
{Opt_trans, "trans=%s"},
{Opt_err, NULL},
};
/**
* v9fs_parse_options - parse mount options into session structure
* @options: options string passed from mount
* @v9ses: existing v9fs session information
*
*/
static void parse_opts(char *options, struct p9_client *clnt)
{
char *p;
substring_t args[MAX_OPT_ARGS];
int option;
int ret;
clnt->trans_mod = v9fs_default_trans();
clnt->dotu = 1;
clnt->msize = 8192;
if (!options)
return;
while ((p = strsep(&options, ",")) != NULL) {
int token;
if (!*p)
continue;
token = match_token(p, tokens, args);
if (token < Opt_trans) {
ret = match_int(&args[0], &option);
if (ret < 0) {
P9_DPRINTK(P9_DEBUG_ERROR,
"integer field, but no integer?\n");
continue;
}
}
switch (token) {
case Opt_msize:
clnt->msize = option;
break;
case Opt_trans:
clnt->trans_mod = v9fs_match_trans(&args[0]);
break;
case Opt_legacy:
clnt->dotu = 0;
break;
default:
continue;
}
}
}
/**
* p9_client_rpc - sends 9P request and waits until a response is available.
* The function can be interrupted.
* @c: client data
* @tc: request to be sent
* @rc: pointer where a pointer to the response is stored
*/
int
p9_client_rpc(struct p9_client *c, struct p9_fcall *tc,
struct p9_fcall **rc)
{
return c->trans->rpc(c->trans, tc, rc);
}
struct p9_client *p9_client_create(const char *dev_name, char *options)
{
int err, n;
struct p9_client *clnt;
......@@ -54,12 +138,7 @@ struct p9_client *p9_client_create(struct p9_trans *trans, int msize,
if (!clnt)
return ERR_PTR(-ENOMEM);
P9_DPRINTK(P9_DEBUG_9P, "clnt %p trans %p msize %d dotu %d\n",
clnt, trans, msize, dotu);
spin_lock_init(&clnt->lock);
clnt->trans = trans;
clnt->msize = msize;
clnt->dotu = dotu;
INIT_LIST_HEAD(&clnt->fidlist);
clnt->fidpool = p9_idpool_create();
if (!clnt->fidpool) {
......@@ -68,13 +147,29 @@ struct p9_client *p9_client_create(struct p9_trans *trans, int msize,
goto error;
}
clnt->conn = p9_conn_create(clnt->trans, clnt->msize, &clnt->dotu);
if (IS_ERR(clnt->conn)) {
err = PTR_ERR(clnt->conn);
clnt->conn = NULL;
parse_opts(options, clnt);
if (clnt->trans_mod == NULL) {
err = -EPROTONOSUPPORT;
P9_DPRINTK(P9_DEBUG_ERROR,
"No transport defined or default transport\n");
goto error;
}
P9_DPRINTK(P9_DEBUG_9P, "clnt %p trans %p msize %d dotu %d\n",
clnt, clnt->trans_mod, clnt->msize, clnt->dotu);
clnt->trans = clnt->trans_mod->create(dev_name, options, clnt->msize,
clnt->dotu);
if (IS_ERR(clnt->trans)) {
err = PTR_ERR(clnt->trans);
clnt->trans = NULL;
goto error;
}
if ((clnt->msize+P9_IOHDRSZ) > clnt->trans_mod->maxsize)
clnt->msize = clnt->trans_mod->maxsize-P9_IOHDRSZ;
tc = p9_create_tversion(clnt->msize, clnt->dotu?"9P2000.u":"9P2000");
if (IS_ERR(tc)) {
err = PTR_ERR(tc);
......@@ -82,7 +177,7 @@ struct p9_client *p9_client_create(struct p9_trans *trans, int msize,
goto error;
}
err = p9_conn_rpc(clnt->conn, tc, &rc);
err = p9_client_rpc(clnt, tc, &rc);
if (err)
goto error;
......@@ -117,10 +212,6 @@ void p9_client_destroy(struct p9_client *clnt)
struct p9_fid *fid, *fidptr;
P9_DPRINTK(P9_DEBUG_9P, "clnt %p\n", clnt);
if (clnt->conn) {
p9_conn_destroy(clnt->conn);
clnt->conn = NULL;
}
if (clnt->trans) {
clnt->trans->close(clnt->trans);
......@@ -142,7 +233,6 @@ void p9_client_disconnect(struct p9_client *clnt)
{
P9_DPRINTK(P9_DEBUG_9P, "clnt %p\n", clnt);
clnt->trans->status = Disconnected;
p9_conn_cancel(clnt->conn, -EIO);
}
EXPORT_SYMBOL(p9_client_disconnect);
......@@ -174,7 +264,7 @@ struct p9_fid *p9_client_attach(struct p9_client *clnt, struct p9_fid *afid,
goto error;
}
err = p9_conn_rpc(clnt->conn, tc, &rc);
err = p9_client_rpc(clnt, tc, &rc);
if (err)
goto error;
......@@ -219,7 +309,7 @@ struct p9_fid *p9_client_auth(struct p9_client *clnt, char *uname,
goto error;
}
err = p9_conn_rpc(clnt->conn, tc, &rc);
err = p9_client_rpc(clnt, tc, &rc);
if (err)
goto error;
......@@ -270,7 +360,7 @@ struct p9_fid *p9_client_walk(struct p9_fid *oldfid, int nwname, char **wnames,
goto error;
}
err = p9_conn_rpc(clnt->conn, tc, &rc);
err = p9_client_rpc(clnt, tc, &rc);
if (err) {
if (rc && rc->id == P9_RWALK)
goto clunk_fid;
......@@ -305,7 +395,7 @@ clunk_fid:
goto error;
}
p9_conn_rpc(clnt->conn, tc, &rc);
p9_client_rpc(clnt, tc, &rc);
error:
kfree(tc);
......@@ -339,7 +429,7 @@ int p9_client_open(struct p9_fid *fid, int mode)
goto done;
}
err = p9_conn_rpc(clnt->conn, tc, &rc);
err = p9_client_rpc(clnt, tc, &rc);
if (err)
goto done;
......@@ -378,7 +468,7 @@ int p9_client_fcreate(struct p9_fid *fid, char *name, u32 perm, int mode,
goto done;
}
err = p9_conn_rpc(clnt->conn, tc, &rc);
err = p9_client_rpc(clnt, tc, &rc);
if (err)
goto done;
......@@ -411,7 +501,7 @@ int p9_client_clunk(struct p9_fid *fid)
goto done;
}
err = p9_conn_rpc(clnt->conn, tc, &rc);
err = p9_client_rpc(clnt, tc, &rc);
if (err)
goto done;
......@@ -443,7 +533,7 @@ int p9_client_remove(struct p9_fid *fid)
goto done;
}
err = p9_conn_rpc(clnt->conn, tc, &rc);
err = p9_client_rpc(clnt, tc, &rc);
if (err)
goto done;
......@@ -485,7 +575,7 @@ int p9_client_read(struct p9_fid *fid, char *data, u64 offset, u32 count)
goto error;
}
err = p9_conn_rpc(clnt->conn, tc, &rc);
err = p9_client_rpc(clnt, tc, &rc);
if (err)
goto error;
......@@ -542,7 +632,7 @@ int p9_client_write(struct p9_fid *fid, char *data, u64 offset, u32 count)
goto error;
}
err = p9_conn_rpc(clnt->conn, tc, &rc);
err = p9_client_rpc(clnt, tc, &rc);
if (err)
goto error;
......@@ -596,7 +686,7 @@ p9_client_uread(struct p9_fid *fid, char __user *data, u64 offset, u32 count)
goto error;
}
err = p9_conn_rpc(clnt->conn, tc, &rc);
err = p9_client_rpc(clnt, tc, &rc);
if (err)
goto error;
......@@ -660,7 +750,7 @@ p9_client_uwrite(struct p9_fid *fid, const char __user *data, u64 offset,
goto error;
}
err = p9_conn_rpc(clnt->conn, tc, &rc);
err = p9_client_rpc(clnt, tc, &rc);
if (err)
goto error;
......@@ -731,7 +821,7 @@ struct p9_stat *p9_client_stat(struct p9_fid *fid)
goto error;
}
err = p9_conn_rpc(clnt->conn, tc, &rc);
err = p9_client_rpc(clnt, tc, &rc);
if (err)
goto error;
......@@ -773,7 +863,7 @@ int p9_client_wstat(struct p9_fid *fid, struct p9_wstat *wst)
goto done;
}
err = p9_conn_rpc(clnt->conn, tc, &rc);
err = p9_client_rpc(clnt, tc, &rc);
done:
kfree(tc);
......@@ -830,7 +920,7 @@ struct p9_stat *p9_client_dirread(struct p9_fid *fid, u64 offset)
goto error;
}
err = p9_conn_rpc(clnt->conn, tc, &rc);
err = p9_client_rpc(clnt, tc, &rc);
if (err)
goto error;
......@@ -901,16 +991,21 @@ static struct p9_stat *p9_clone_stat(struct p9_stat *st, int dotu)
memmove(ret, st, sizeof(struct p9_stat));
p = ((char *) ret) + sizeof(struct p9_stat);
memmove(p, st->name.str, st->name.len);
ret->name.str = p;
p += st->name.len;
memmove(p, st->uid.str, st->uid.len);
ret->uid.str = p;
p += st->uid.len;
memmove(p, st->gid.str, st->gid.len);
ret->gid.str = p;
p += st->gid.len;
memmove(p, st->muid.str, st->muid.len);
ret->muid.str = p;
p += st->muid.len;
if (dotu) {
memmove(p, st->extension.str, st->extension.len);
ret->extension.str = p;
p += st->extension.len;
}
......
......@@ -347,12 +347,12 @@ p9_printfcall(char *buf, int buflen, struct p9_fcall *fc, int extended)
return ret;
}
#else
int
p9_printfcall(char *buf, int buflen, struct p9_fcall *fc, int extended)
{
return 0;
}
EXPORT_SYMBOL(p9_printfcall);
#endif /* CONFIG_NET_9P_DEBUG */
EXPORT_SYMBOL(p9_printfcall);
......@@ -106,15 +106,10 @@ EXPORT_SYMBOL(v9fs_default_trans);
*/
static int __init init_p9(void)
{
int ret;
int ret = 0;
p9_error_init();
printk(KERN_INFO "Installing 9P2000 support\n");
ret = p9_mux_global_init();
if (ret) {
printk(KERN_WARNING "9p: starting mux failed\n");
return ret;
}
return ret;
}
......@@ -126,7 +121,7 @@ static int __init init_p9(void)
static void __exit exit_p9(void)
{
p9_mux_global_exit();
printk(KERN_INFO "Unloading 9P2000 support\n");
}
module_init(init_p9)
......
/*
* net/9p/mux.c
*
* Protocol Multiplexer
*
* Copyright (C) 2004 by Eric Van Hensbergen <ericvh@gmail.com>
* Copyright (C) 2004-2005 by Latchesar Ionkov <lucho@ionkov.net>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2
* as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to:
* Free Software Foundation
* 51 Franklin Street, Fifth Floor
* Boston, MA 02111-1301 USA
*
*/
#include <linux/module.h>
#include <linux/errno.h>
#include <linux/fs.h>
#include <linux/poll.h>
#include <linux/kthread.h>
#include <linux/idr.h>
#include <linux/mutex.h>
#include <net/9p/9p.h>
#include <linux/parser.h>
#include <net/9p/transport.h>
#include <net/9p/conn.h>
#define ERREQFLUSH 1
#define SCHED_TIMEOUT 10
#define MAXPOLLWADDR 2
enum {
Rworksched = 1, /* read work scheduled or running */
Rpending = 2, /* can read */
Wworksched = 4, /* write work scheduled or running */
Wpending = 8, /* can write */
};
enum {
None,
Flushing,
Flushed,
};
struct p9_mux_poll_task;
struct p9_req {
spinlock_t lock; /* protect request structure */
int tag;
struct p9_fcall *tcall;
struct p9_fcall *rcall;
int err;
p9_conn_req_callback cb;
void *cba;
int flush;
struct list_head req_list;
};
struct p9_conn {
spinlock_t lock; /* protect lock structure */
struct list_head mux_list;
struct p9_mux_poll_task *poll_task;
int msize;
unsigned char *extended;
struct p9_trans *trans;
struct p9_idpool *tagpool;
int err;
wait_queue_head_t equeue;
struct list_head req_list;
struct list_head unsent_req_list;
struct p9_fcall *rcall;
int rpos;
char *rbuf;
int wpos;
int wsize;
char *wbuf;
wait_queue_t poll_wait[MAXPOLLWADDR];
wait_queue_head_t *poll_waddr[MAXPOLLWADDR];
poll_table pt;
struct work_struct rq;
struct work_struct wq;
unsigned long wsched;
};
struct p9_mux_poll_task {
struct task_struct *task;
struct list_head mux_list;
int muxnum;
};
struct p9_mux_rpc {
struct p9_conn *m;
int err;
struct p9_fcall *tcall;
struct p9_fcall *rcall;
wait_queue_head_t wqueue;
};
static int p9_poll_proc(void *);
static void p9_read_work(struct work_struct *work);
static void p9_write_work(struct work_struct *work);
static void p9_pollwait(struct file *filp, wait_queue_head_t *wait_address,
poll_table * p);
static u16 p9_mux_get_tag(struct p9_conn *);
static void p9_mux_put_tag(struct p9_conn *, u16);
static DEFINE_MUTEX(p9_mux_task_lock);
static struct workqueue_struct *p9_mux_wq;
static int p9_mux_num;
static int p9_mux_poll_task_num;
static struct p9_mux_poll_task p9_mux_poll_tasks[100];
int p9_mux_global_init(void)
{
int i;
for (i = 0; i < ARRAY_SIZE(p9_mux_poll_tasks); i++)
p9_mux_poll_tasks[i].task = NULL;
p9_mux_wq = create_workqueue("v9fs");
if (!p9_mux_wq) {
printk(KERN_WARNING "v9fs: mux: creating workqueue failed\n");
return -ENOMEM;
}
return 0;
}
void p9_mux_global_exit(void)
{
destroy_workqueue(p9_mux_wq);
}
/**
* p9_mux_calc_poll_procs - calculates the number of polling procs
* based on the number of mounted v9fs filesystems.
*
* The current implementation returns sqrt of the number of mounts.
*/
static int p9_mux_calc_poll_procs(int muxnum)
{
int n;
if (p9_mux_poll_task_num)
n = muxnum / p9_mux_poll_task_num +
(muxnum % p9_mux_poll_task_num ? 1 : 0);
else
n = 1;
if (n > ARRAY_SIZE(p9_mux_poll_tasks))
n = ARRAY_SIZE(p9_mux_poll_tasks);
return n;
}
static int p9_mux_poll_start(struct p9_conn *m)
{
int i, n;
struct p9_mux_poll_task *vpt, *vptlast;
struct task_struct *pproc;
P9_DPRINTK(P9_DEBUG_MUX, "mux %p muxnum %d procnum %d\n", m, p9_mux_num,
p9_mux_poll_task_num);
mutex_lock(&p9_mux_task_lock);
n = p9_mux_calc_poll_procs(p9_mux_num + 1);
if (n > p9_mux_poll_task_num) {
for (i = 0; i < ARRAY_SIZE(p9_mux_poll_tasks); i++) {
if (p9_mux_poll_tasks[i].task == NULL) {
vpt = &p9_mux_poll_tasks[i];
P9_DPRINTK(P9_DEBUG_MUX, "create proc %p\n",
vpt);
pproc = kthread_create(p9_poll_proc, vpt,
"v9fs-poll");
if (!IS_ERR(pproc)) {
vpt->task = pproc;
INIT_LIST_HEAD(&vpt->mux_list);
vpt->muxnum = 0;
p9_mux_poll_task_num++;
wake_up_process(vpt->task);
}
break;
}
}
if (i >= ARRAY_SIZE(p9_mux_poll_tasks))
P9_DPRINTK(P9_DEBUG_ERROR,
"warning: no free poll slots\n");
}
n = (p9_mux_num + 1) / p9_mux_poll_task_num +
((p9_mux_num + 1) % p9_mux_poll_task_num ? 1 : 0);
vptlast = NULL;
for (i = 0; i < ARRAY_SIZE(p9_mux_poll_tasks); i++) {
vpt = &p9_mux_poll_tasks[i];
if (vpt->task != NULL) {
vptlast = vpt;
if (vpt->muxnum < n) {
P9_DPRINTK(P9_DEBUG_MUX, "put in proc %d\n", i);
list_add(&m->mux_list, &vpt->mux_list);
vpt->muxnum++;
m->poll_task = vpt;
memset(&m->poll_waddr, 0,
sizeof(m->poll_waddr));
init_poll_funcptr(&m->pt, p9_pollwait);
break;
}
}
}
if (i >= ARRAY_SIZE(p9_mux_poll_tasks)) {
if (vptlast == NULL) {
mutex_unlock(&p9_mux_task_lock);
return -ENOMEM;
}
P9_DPRINTK(P9_DEBUG_MUX, "put in proc %d\n", i);
list_add(&m->mux_list, &vptlast->mux_list);
vptlast->muxnum++;
m->poll_task = vptlast;
memset(&m->poll_waddr, 0, sizeof(m->poll_waddr));
init_poll_funcptr(&m->pt, p9_pollwait);
}
p9_mux_num++;
mutex_unlock(&p9_mux_task_lock);
return 0;
}
static void p9_mux_poll_stop(struct p9_conn *m)
{
int i;
struct p9_mux_poll_task *vpt;
mutex_lock(&p9_mux_task_lock);
vpt = m->poll_task;
list_del(&m->mux_list);
for (i = 0; i < ARRAY_SIZE(m->poll_waddr); i++) {
if (m->poll_waddr[i] != NULL) {
remove_wait_queue(m->poll_waddr[i], &m->poll_wait[i]);
m->poll_waddr[i] = NULL;
}
}
vpt->muxnum--;
if (!vpt->muxnum) {
P9_DPRINTK(P9_DEBUG_MUX, "destroy proc %p\n", vpt);
kthread_stop(vpt->task);
vpt->task = NULL;
p9_mux_poll_task_num--;
}
p9_mux_num--;
mutex_unlock(&p9_mux_task_lock);
}
/**
* p9_conn_create - allocate and initialize the per-session mux data
* Creates the polling task if this is the first session.
*
* @trans - transport structure
* @msize - maximum message size
* @extended - pointer to the extended flag
*/
struct p9_conn *p9_conn_create(struct p9_trans *trans, int msize,
unsigned char *extended)
{
int i, n;
struct p9_conn *m, *mtmp;
P9_DPRINTK(P9_DEBUG_MUX, "transport %p msize %d\n", trans, msize);
m = kmalloc(sizeof(struct p9_conn), GFP_KERNEL);
if (!m)
return ERR_PTR(-ENOMEM);
spin_lock_init(&m->lock);
INIT_LIST_HEAD(&m->mux_list);
m->msize = msize;
m->extended = extended;
m->trans = trans;
m->tagpool = p9_idpool_create();
if (IS_ERR(m->tagpool)) {
mtmp = ERR_PTR(-ENOMEM);
kfree(m);
return mtmp;
}
m->err = 0;
init_waitqueue_head(&m->equeue);
INIT_LIST_HEAD(&m->req_list);
INIT_LIST_HEAD(&m->unsent_req_list);
m->rcall = NULL;
m->rpos = 0;
m->rbuf = NULL;
m->wpos = m->wsize = 0;
m->wbuf = NULL;
INIT_WORK(&m->rq, p9_read_work);
INIT_WORK(&m->wq, p9_write_work);
m->wsched = 0;
memset(&m->poll_waddr, 0, sizeof(m->poll_waddr));
m->poll_task = NULL;
n = p9_mux_poll_start(m);
if (n) {
kfree(m);
return ERR_PTR(n);
}
n = trans->poll(trans, &m->pt);
if (n & POLLIN) {
P9_DPRINTK(P9_DEBUG_MUX, "mux %p can read\n", m);
set_bit(Rpending, &m->wsched);
}
if (n & POLLOUT) {
P9_DPRINTK(P9_DEBUG_MUX, "mux %p can write\n", m);
set_bit(Wpending, &m->wsched);
}
for (i = 0; i < ARRAY_SIZE(m->poll_waddr); i++) {
if (IS_ERR(m->poll_waddr[i])) {
p9_mux_poll_stop(m);
mtmp = (void *)m->poll_waddr; /* the error code */
kfree(m);
m = mtmp;
break;
}
}
return m;
}
EXPORT_SYMBOL(p9_conn_create);
/**
* p9_mux_destroy - cancels all pending requests and frees mux resources
*/
void p9_conn_destroy(struct p9_conn *m)
{
P9_DPRINTK(P9_DEBUG_MUX, "mux %p prev %p next %p\n", m,
m->mux_list.prev, m->mux_list.next);
p9_conn_cancel(m, -ECONNRESET);
if (!list_empty(&m->req_list)) {
/* wait until all processes waiting on this session exit */
P9_DPRINTK(P9_DEBUG_MUX,
"mux %p waiting for empty request queue\n", m);
wait_event_timeout(m->equeue, (list_empty(&m->req_list)), 5000);
P9_DPRINTK(P9_DEBUG_MUX, "mux %p request queue empty: %d\n", m,
list_empty(&m->req_list));
}
p9_mux_poll_stop(m);
m->trans = NULL;
p9_idpool_destroy(m->tagpool);
kfree(m);
}
EXPORT_SYMBOL(p9_conn_destroy);
/**
* p9_pollwait - called by files poll operation to add v9fs-poll task
* to files wait queue
*/
static void
p9_pollwait(struct file *filp, wait_queue_head_t *wait_address,
poll_table * p)
{
int i;
struct p9_conn *m;
m = container_of(p, struct p9_conn, pt);
for (i = 0; i < ARRAY_SIZE(m->poll_waddr); i++)
if (m->poll_waddr[i] == NULL)
break;
if (i >= ARRAY_SIZE(m->poll_waddr)) {
P9_DPRINTK(P9_DEBUG_ERROR, "not enough wait_address slots\n");
return;
}
m->poll_waddr[i] = wait_address;
if (!wait_address) {
P9_DPRINTK(P9_DEBUG_ERROR, "no wait_address\n");
m->poll_waddr[i] = ERR_PTR(-EIO);
return;
}
init_waitqueue_entry(&m->poll_wait[i], m->poll_task->task);
add_wait_queue(wait_address, &m->poll_wait[i]);
}
/**
* p9_poll_mux - polls a mux and schedules read or write works if necessary
*/
static void p9_poll_mux(struct p9_conn *m)
{
int n;
if (m->err < 0)
return;
n = m->trans->poll(m->trans, NULL);
if (n < 0 || n & (POLLERR | POLLHUP | POLLNVAL)) {
P9_DPRINTK(P9_DEBUG_MUX, "error mux %p err %d\n", m, n);
if (n >= 0)
n = -ECONNRESET;
p9_conn_cancel(m, n);
}
if (n & POLLIN) {
set_bit(Rpending, &m->wsched);
P9_DPRINTK(P9_DEBUG_MUX, "mux %p can read\n", m);
if (!test_and_set_bit(Rworksched, &m->wsched)) {
P9_DPRINTK(P9_DEBUG_MUX, "schedule read work %p\n", m);
queue_work(p9_mux_wq, &m->rq);
}
}
if (n & POLLOUT) {
set_bit(Wpending, &m->wsched);
P9_DPRINTK(P9_DEBUG_MUX, "mux %p can write\n", m);
if ((m->wsize || !list_empty(&m->unsent_req_list))
&& !test_and_set_bit(Wworksched, &m->wsched)) {
P9_DPRINTK(P9_DEBUG_MUX, "schedule write work %p\n", m);
queue_work(p9_mux_wq, &m->wq);
}
}
}
/**
* p9_poll_proc - polls all v9fs transports for new events and queues
* the appropriate work to the work queue
*/
static int p9_poll_proc(void *a)
{
struct p9_conn *m, *mtmp;
struct p9_mux_poll_task *vpt;
vpt = a;
P9_DPRINTK(P9_DEBUG_MUX, "start %p %p\n", current, vpt);
while (!kthread_should_stop()) {
set_current_state(TASK_INTERRUPTIBLE);
list_for_each_entry_safe(m, mtmp, &vpt->mux_list, mux_list) {
p9_poll_mux(m);
}
P9_DPRINTK(P9_DEBUG_MUX, "sleeping...\n");
schedule_timeout(SCHED_TIMEOUT * HZ);
}
__set_current_state(TASK_RUNNING);
P9_DPRINTK(P9_DEBUG_MUX, "finish\n");
return 0;
}
/**
* p9_write_work - called when a transport can send some data
*/
static void p9_write_work(struct work_struct *work)
{
int n, err;
struct p9_conn *m;
struct p9_req *req;
m = container_of(work, struct p9_conn, wq);
if (m->err < 0) {
clear_bit(Wworksched, &m->wsched);
return;
}
if (!m->wsize) {
if (list_empty(&m->unsent_req_list)) {
clear_bit(Wworksched, &m->wsched);
return;
}
spin_lock(&m->lock);
again:
req = list_entry(m->unsent_req_list.next, struct p9_req,
req_list);
list_move_tail(&req->req_list, &m->req_list);
if (req->err == ERREQFLUSH)
goto again;
m->wbuf = req->tcall->sdata;
m->wsize = req->tcall->size;
m->wpos = 0;
spin_unlock(&m->lock);
}
P9_DPRINTK(P9_DEBUG_MUX, "mux %p pos %d size %d\n", m, m->wpos,
m->wsize);
clear_bit(Wpending, &m->wsched);
err = m->trans->write(m->trans, m->wbuf + m->wpos, m->wsize - m->wpos);
P9_DPRINTK(P9_DEBUG_MUX, "mux %p sent %d bytes\n", m, err);
if (err == -EAGAIN) {
clear_bit(Wworksched, &m->wsched);
return;
}
if (err < 0)
goto error;
else if (err == 0) {
err = -EREMOTEIO;
goto error;
}
m->wpos += err;
if (m->wpos == m->wsize)
m->wpos = m->wsize = 0;
if (m->wsize == 0 && !list_empty(&m->unsent_req_list)) {
if (test_and_clear_bit(Wpending, &m->wsched))
n = POLLOUT;
else
n = m->trans->poll(m->trans, NULL);
if (n & POLLOUT) {
P9_DPRINTK(P9_DEBUG_MUX, "schedule write work %p\n", m);
queue_work(p9_mux_wq, &m->wq);
} else
clear_bit(Wworksched, &m->wsched);
} else
clear_bit(Wworksched, &m->wsched);
return;
error:
p9_conn_cancel(m, err);
clear_bit(Wworksched, &m->wsched);
}
static void process_request(struct p9_conn *m, struct p9_req *req)
{
int ecode;
struct p9_str *ename;
if (!req->err && req->rcall->id == P9_RERROR) {
ecode = req->rcall->params.rerror.errno;
ename = &req->rcall->params.rerror.error;
P9_DPRINTK(P9_DEBUG_MUX, "Rerror %.*s\n", ename->len,
ename->str);
if (*m->extended)
req->err = -ecode;
if (!req->err) {
req->err = p9_errstr2errno(ename->str, ename->len);
if (!req->err) { /* string match failed */
PRINT_FCALL_ERROR("unknown error", req->rcall);
}
if (!req->err)
req->err = -ESERVERFAULT;
}
} else if (req->tcall && req->rcall->id != req->tcall->id + 1) {
P9_DPRINTK(P9_DEBUG_ERROR,
"fcall mismatch: expected %d, got %d\n",
req->tcall->id + 1, req->rcall->id);
if (!req->err)
req->err = -EIO;
}
}
/**
* p9_read_work - called when there is some data to be read from a transport
*/
static void p9_read_work(struct work_struct *work)
{
int n, err;
struct p9_conn *m;
struct p9_req *req, *rptr, *rreq;
struct p9_fcall *rcall;
char *rbuf;
m = container_of(work, struct p9_conn, rq);
if (m->err < 0)
return;
rcall = NULL;
P9_DPRINTK(P9_DEBUG_MUX, "start mux %p pos %d\n", m, m->rpos);
if (!m->rcall) {
m->rcall =
kmalloc(sizeof(struct p9_fcall) + m->msize, GFP_KERNEL);
if (!m->rcall) {
err = -ENOMEM;
goto error;
}
m->rbuf = (char *)m->rcall + sizeof(struct p9_fcall);
m->rpos = 0;
}
clear_bit(Rpending, &m->wsched);
err = m->trans->read(m->trans, m->rbuf + m->rpos, m->msize - m->rpos);
P9_DPRINTK(P9_DEBUG_MUX, "mux %p got %d bytes\n", m, err);
if (err == -EAGAIN) {
clear_bit(Rworksched, &m->wsched);
return;
}
if (err <= 0)
goto error;
m->rpos += err;
while (m->rpos > 4) {
n = le32_to_cpu(*(__le32 *) m->rbuf);
if (n >= m->msize) {
P9_DPRINTK(P9_DEBUG_ERROR,
"requested packet size too big: %d\n", n);
err = -EIO;
goto error;
}
if (m->rpos < n)
break;
err =
p9_deserialize_fcall(m->rbuf, n, m->rcall, *m->extended);
if (err < 0) {
goto error;
}
#ifdef CONFIG_NET_9P_DEBUG
if ((p9_debug_level&P9_DEBUG_FCALL) == P9_DEBUG_FCALL) {
char buf[150];
p9_printfcall(buf, sizeof(buf), m->rcall,
*m->extended);
printk(KERN_NOTICE ">>> %p %s\n", m, buf);
}
#endif
rcall = m->rcall;
rbuf = m->rbuf;
if (m->rpos > n) {
m->rcall = kmalloc(sizeof(struct p9_fcall) + m->msize,
GFP_KERNEL);
if (!m->rcall) {
err = -ENOMEM;
goto error;
}
m->rbuf = (char *)m->rcall + sizeof(struct p9_fcall);
memmove(m->rbuf, rbuf + n, m->rpos - n);
m->rpos -= n;
} else {
m->rcall = NULL;
m->rbuf = NULL;
m->rpos = 0;
}
P9_DPRINTK(P9_DEBUG_MUX, "mux %p fcall id %d tag %d\n", m,
rcall->id, rcall->tag);
req = NULL;
spin_lock(&m->lock);
list_for_each_entry_safe(rreq, rptr, &m->req_list, req_list) {
if (rreq->tag == rcall->tag) {
req = rreq;
if (req->flush != Flushing)
list_del(&req->req_list);
break;
}
}
spin_unlock(&m->lock);
if (req) {
req->rcall = rcall;
process_request(m, req);
if (req->flush != Flushing) {
if (req->cb)
(*req->cb) (req, req->cba);
else
kfree(req->rcall);
wake_up(&m->equeue);
}
} else {
if (err >= 0 && rcall->id != P9_RFLUSH)
P9_DPRINTK(P9_DEBUG_ERROR,
"unexpected response mux %p id %d tag %d\n",
m, rcall->id, rcall->tag);
kfree(rcall);
}
}
if (!list_empty(&m->req_list)) {
if (test_and_clear_bit(Rpending, &m->wsched))
n = POLLIN;
else
n = m->trans->poll(m->trans, NULL);
if (n & POLLIN) {
P9_DPRINTK(P9_DEBUG_MUX, "schedule read work %p\n", m);
queue_work(p9_mux_wq, &m->rq);
} else
clear_bit(Rworksched, &m->wsched);
} else
clear_bit(Rworksched, &m->wsched);
return;
error:
p9_conn_cancel(m, err);
clear_bit(Rworksched, &m->wsched);
}
/**
* p9_send_request - send 9P request
* The function can sleep until the request is scheduled for sending.
* The function can be interrupted. Return from the function is not
* a guarantee that the request is sent successfully. Can return errors
* that can be retrieved by PTR_ERR macros.
*
* @m: mux data
* @tc: request to be sent
* @cb: callback function to call when response is received
* @cba: parameter to pass to the callback function
*/
static struct p9_req *p9_send_request(struct p9_conn *m,
struct p9_fcall *tc,
p9_conn_req_callback cb, void *cba)
{
int n;
struct p9_req *req;
P9_DPRINTK(P9_DEBUG_MUX, "mux %p task %p tcall %p id %d\n", m, current,
tc, tc->id);
if (m->err < 0)
return ERR_PTR(m->err);
req = kmalloc(sizeof(struct p9_req), GFP_KERNEL);
if (!req)
return ERR_PTR(-ENOMEM);
if (tc->id == P9_TVERSION)
n = P9_NOTAG;
else
n = p9_mux_get_tag(m);
if (n < 0)
return ERR_PTR(-ENOMEM);
p9_set_tag(tc, n);
#ifdef CONFIG_NET_9P_DEBUG
if ((p9_debug_level&P9_DEBUG_FCALL) == P9_DEBUG_FCALL) {
char buf[150];
p9_printfcall(buf, sizeof(buf), tc, *m->extended);
printk(KERN_NOTICE "<<< %p %s\n", m, buf);
}
#endif
spin_lock_init(&req->lock);
req->tag = n;
req->tcall = tc;
req->rcall = NULL;
req->err = 0;
req->cb = cb;
req->cba = cba;
req->flush = None;
spin_lock(&m->lock);
list_add_tail(&req->req_list, &m->unsent_req_list);
spin_unlock(&m->lock);
if (test_and_clear_bit(Wpending, &m->wsched))
n = POLLOUT;
else
n = m->trans->poll(m->trans, NULL);
if (n & POLLOUT && !test_and_set_bit(Wworksched, &m->wsched))
queue_work(p9_mux_wq, &m->wq);
return req;
}
static void p9_mux_free_request(struct p9_conn *m, struct p9_req *req)
{
p9_mux_put_tag(m, req->tag);
kfree(req);
}
static void p9_mux_flush_cb(struct p9_req *freq, void *a)
{
p9_conn_req_callback cb;
int tag;
struct p9_conn *m;
struct p9_req *req, *rreq, *rptr;
m = a;
P9_DPRINTK(P9_DEBUG_MUX, "mux %p tc %p rc %p err %d oldtag %d\n", m,
freq->tcall, freq->rcall, freq->err,
freq->tcall->params.tflush.oldtag);
spin_lock(&m->lock);
cb = NULL;
tag = freq->tcall->params.tflush.oldtag;
req = NULL;
list_for_each_entry_safe(rreq, rptr, &m->req_list, req_list) {
if (rreq->tag == tag) {
req = rreq;
list_del(&req->req_list);
break;
}
}
spin_unlock(&m->lock);
if (req) {
spin_lock(&req->lock);
req->flush = Flushed;
spin_unlock(&req->lock);
if (req->cb)
(*req->cb) (req, req->cba);
else
kfree(req->rcall);
wake_up(&m->equeue);
}
kfree(freq->tcall);
kfree(freq->rcall);
p9_mux_free_request(m, freq);
}
static int
p9_mux_flush_request(struct p9_conn *m, struct p9_req *req)
{
struct p9_fcall *fc;
struct p9_req *rreq, *rptr;
P9_DPRINTK(P9_DEBUG_MUX, "mux %p req %p tag %d\n", m, req, req->tag);
/* if a response was received for a request, do nothing */
spin_lock(&req->lock);
if (req->rcall || req->err) {
spin_unlock(&req->lock);
P9_DPRINTK(P9_DEBUG_MUX,
"mux %p req %p response already received\n", m, req);
return 0;
}
req->flush = Flushing;
spin_unlock(&req->lock);
spin_lock(&m->lock);
/* if the request is not sent yet, just remove it from the list */
list_for_each_entry_safe(rreq, rptr, &m->unsent_req_list, req_list) {
if (rreq->tag == req->tag) {
P9_DPRINTK(P9_DEBUG_MUX,
"mux %p req %p request is not sent yet\n", m, req);
list_del(&rreq->req_list);
req->flush = Flushed;
spin_unlock(&m->lock);
if (req->cb)
(*req->cb) (req, req->cba);
return 0;
}
}
spin_unlock(&m->lock);
clear_thread_flag(TIF_SIGPENDING);
fc = p9_create_tflush(req->tag);
p9_send_request(m, fc, p9_mux_flush_cb, m);
return 1;
}
static void
p9_conn_rpc_cb(struct p9_req *req, void *a)
{
struct p9_mux_rpc *r;
P9_DPRINTK(P9_DEBUG_MUX, "req %p r %p\n", req, a);
r = a;
r->rcall = req->rcall;
r->err = req->err;
if (req->flush != None && !req->err)
r->err = -ERESTARTSYS;
wake_up(&r->wqueue);
}
/**
* p9_mux_rpc - sends 9P request and waits until a response is available.
* The function can be interrupted.
* @m: mux data
* @tc: request to be sent
* @rc: pointer where a pointer to the response is stored
*/
int
p9_conn_rpc(struct p9_conn *m, struct p9_fcall *tc,
struct p9_fcall **rc)
{
int err, sigpending;
unsigned long flags;
struct p9_req *req;
struct p9_mux_rpc r;
r.err = 0;
r.tcall = tc;
r.rcall = NULL;
r.m = m;
init_waitqueue_head(&r.wqueue);
if (rc)
*rc = NULL;
sigpending = 0;
if (signal_pending(current)) {
sigpending = 1;
clear_thread_flag(TIF_SIGPENDING);
}
req = p9_send_request(m, tc, p9_conn_rpc_cb, &r);
if (IS_ERR(req)) {
err = PTR_ERR(req);
P9_DPRINTK(P9_DEBUG_MUX, "error %d\n", err);
return err;
}
err = wait_event_interruptible(r.wqueue, r.rcall != NULL || r.err < 0);
if (r.err < 0)
err = r.err;
if (err == -ERESTARTSYS && m->trans->status == Connected
&& m->err == 0) {
if (p9_mux_flush_request(m, req)) {
/* wait until we get response of the flush message */
do {
clear_thread_flag(TIF_SIGPENDING);
err = wait_event_interruptible(r.wqueue,
r.rcall || r.err);
} while (!r.rcall && !r.err && err == -ERESTARTSYS &&
m->trans->status == Connected && !m->err);
err = -ERESTARTSYS;
}
sigpending = 1;
}
if (sigpending) {
spin_lock_irqsave(&current->sighand->siglock, flags);
recalc_sigpending();
spin_unlock_irqrestore(&current->sighand->siglock, flags);
}
if (rc)
*rc = r.rcall;
else
kfree(r.rcall);
p9_mux_free_request(m, req);
if (err > 0)
err = -EIO;
return err;
}
EXPORT_SYMBOL(p9_conn_rpc);
#ifdef P9_NONBLOCK
/**
* p9_conn_rpcnb - sends 9P request without waiting for response.
* @m: mux data
* @tc: request to be sent
* @cb: callback function to be called when response arrives
* @cba: value to pass to the callback function
*/
int p9_conn_rpcnb(struct p9_conn *m, struct p9_fcall *tc,
p9_conn_req_callback cb, void *a)
{
int err;
struct p9_req *req;
req = p9_send_request(m, tc, cb, a);
if (IS_ERR(req)) {
err = PTR_ERR(req);
P9_DPRINTK(P9_DEBUG_MUX, "error %d\n", err);
return PTR_ERR(req);
}
P9_DPRINTK(P9_DEBUG_MUX, "mux %p tc %p tag %d\n", m, tc, req->tag);
return 0;
}
EXPORT_SYMBOL(p9_conn_rpcnb);
#endif /* P9_NONBLOCK */
/**
* p9_conn_cancel - cancel all pending requests with error
* @m: mux data
* @err: error code
*/
void p9_conn_cancel(struct p9_conn *m, int err)
{
struct p9_req *req, *rtmp;
LIST_HEAD(cancel_list);
P9_DPRINTK(P9_DEBUG_ERROR, "mux %p err %d\n", m, err);
m->err = err;
spin_lock(&m->lock);
list_for_each_entry_safe(req, rtmp, &m->req_list, req_list) {
list_move(&req->req_list, &cancel_list);
}
list_for_each_entry_safe(req, rtmp, &m->unsent_req_list, req_list) {
list_move(&req->req_list, &cancel_list);
}
spin_unlock(&m->lock);
list_for_each_entry_safe(req, rtmp, &cancel_list, req_list) {
list_del(&req->req_list);
if (!req->err)
req->err = err;
if (req->cb)
(*req->cb) (req, req->cba);
else
kfree(req->rcall);
}
wake_up(&m->equeue);
}
EXPORT_SYMBOL(p9_conn_cancel);
static u16 p9_mux_get_tag(struct p9_conn *m)
{
int tag;
tag = p9_idpool_get(m->tagpool);
if (tag < 0)
return P9_NOTAG;
else
return (u16) tag;
}
static void p9_mux_put_tag(struct p9_conn *m, u16 tag)
{
if (tag != P9_NOTAG && p9_idpool_check(tag, m->tagpool))
p9_idpool_put(tag, m->tagpool);
}
......@@ -5,7 +5,7 @@
*
* Copyright (C) 2006 by Russ Cox <rsc@swtch.com>
* Copyright (C) 2004-2005 by Latchesar Ionkov <lucho@ionkov.net>
* Copyright (C) 2004-2007 by Eric Van Hensbergen <ericvh@gmail.com>
* Copyright (C) 2004-2008 by Eric Van Hensbergen <ericvh@gmail.com>
* Copyright (C) 1997-2002 by Ron Minnich <rminnich@sarnoff.com>
*
* This program is free software; you can redistribute it and/or modify
......@@ -29,6 +29,7 @@
#include <linux/module.h>
#include <linux/net.h>
#include <linux/ipv6.h>
#include <linux/kthread.h>
#include <linux/errno.h>
#include <linux/kernel.h>
#include <linux/un.h>
......@@ -42,7 +43,9 @@
#define P9_PORT 564
#define MAX_SOCK_BUF (64*1024)
#define ERREQFLUSH 1
#define SCHED_TIMEOUT 10
#define MAXPOLLWADDR 2
struct p9_fd_opts {
int rfd;
......@@ -53,6 +56,7 @@ struct p9_fd_opts {
struct p9_trans_fd {
struct file *rd;
struct file *wr;
struct p9_conn *conn;
};
/*
......@@ -72,6 +76,1028 @@ static match_table_t tokens = {
{Opt_err, NULL},
};
enum {
Rworksched = 1, /* read work scheduled or running */
Rpending = 2, /* can read */
Wworksched = 4, /* write work scheduled or running */
Wpending = 8, /* can write */
};
enum {
None,
Flushing,
Flushed,
};
struct p9_req;
typedef void (*p9_conn_req_callback)(struct p9_req *req, void *a);
struct p9_req {
spinlock_t lock; /* protect request structure */
int tag;
struct p9_fcall *tcall;
struct p9_fcall *rcall;
int err;
p9_conn_req_callback cb;
void *cba;
int flush;
struct list_head req_list;
};
struct p9_mux_poll_task;
struct p9_conn {
spinlock_t lock; /* protect lock structure */
struct list_head mux_list;
struct p9_mux_poll_task *poll_task;
int msize;
unsigned char extended;
struct p9_trans *trans;
struct p9_idpool *tagpool;
int err;
wait_queue_head_t equeue;
struct list_head req_list;
struct list_head unsent_req_list;
struct p9_fcall *rcall;
int rpos;
char *rbuf;
int wpos;
int wsize;
char *wbuf;
wait_queue_t poll_wait[MAXPOLLWADDR];
wait_queue_head_t *poll_waddr[MAXPOLLWADDR];
poll_table pt;
struct work_struct rq;
struct work_struct wq;
unsigned long wsched;
};
struct p9_mux_poll_task {
struct task_struct *task;
struct list_head mux_list;
int muxnum;
};
struct p9_mux_rpc {
struct p9_conn *m;
int err;
struct p9_fcall *tcall;
struct p9_fcall *rcall;
wait_queue_head_t wqueue;
};
static int p9_poll_proc(void *);
static void p9_read_work(struct work_struct *work);
static void p9_write_work(struct work_struct *work);
static void p9_pollwait(struct file *filp, wait_queue_head_t *wait_address,
poll_table *p);
static int p9_fd_write(struct p9_trans *trans, void *v, int len);
static int p9_fd_read(struct p9_trans *trans, void *v, int len);
static DEFINE_MUTEX(p9_mux_task_lock);
static struct workqueue_struct *p9_mux_wq;
static int p9_mux_num;
static int p9_mux_poll_task_num;
static struct p9_mux_poll_task p9_mux_poll_tasks[100];
static void p9_conn_destroy(struct p9_conn *);
static unsigned int p9_fd_poll(struct p9_trans *trans,
struct poll_table_struct *pt);
#ifdef P9_NONBLOCK
static int p9_conn_rpcnb(struct p9_conn *m, struct p9_fcall *tc,
p9_conn_req_callback cb, void *a);
#endif /* P9_NONBLOCK */
static void p9_conn_cancel(struct p9_conn *m, int err);
static int p9_mux_global_init(void)
{
int i;
for (i = 0; i < ARRAY_SIZE(p9_mux_poll_tasks); i++)
p9_mux_poll_tasks[i].task = NULL;
p9_mux_wq = create_workqueue("v9fs");
if (!p9_mux_wq) {
printk(KERN_WARNING "v9fs: mux: creating workqueue failed\n");
return -ENOMEM;
}
return 0;
}
static u16 p9_mux_get_tag(struct p9_conn *m)
{
int tag;
tag = p9_idpool_get(m->tagpool);
if (tag < 0)
return P9_NOTAG;
else
return (u16) tag;
}
static void p9_mux_put_tag(struct p9_conn *m, u16 tag)
{
if (tag != P9_NOTAG && p9_idpool_check(tag, m->tagpool))
p9_idpool_put(tag, m->tagpool);
}
/**
* p9_mux_calc_poll_procs - calculates the number of polling procs
* based on the number of mounted v9fs filesystems.
*
* The current implementation returns sqrt of the number of mounts.
*/
static int p9_mux_calc_poll_procs(int muxnum)
{
int n;
if (p9_mux_poll_task_num)
n = muxnum / p9_mux_poll_task_num +
(muxnum % p9_mux_poll_task_num ? 1 : 0);
else
n = 1;
if (n > ARRAY_SIZE(p9_mux_poll_tasks))
n = ARRAY_SIZE(p9_mux_poll_tasks);
return n;
}
static int p9_mux_poll_start(struct p9_conn *m)
{
int i, n;
struct p9_mux_poll_task *vpt, *vptlast;
struct task_struct *pproc;
P9_DPRINTK(P9_DEBUG_MUX, "mux %p muxnum %d procnum %d\n", m, p9_mux_num,
p9_mux_poll_task_num);
mutex_lock(&p9_mux_task_lock);
n = p9_mux_calc_poll_procs(p9_mux_num + 1);
if (n > p9_mux_poll_task_num) {
for (i = 0; i < ARRAY_SIZE(p9_mux_poll_tasks); i++) {
if (p9_mux_poll_tasks[i].task == NULL) {
vpt = &p9_mux_poll_tasks[i];
P9_DPRINTK(P9_DEBUG_MUX, "create proc %p\n",
vpt);
pproc = kthread_create(p9_poll_proc, vpt,
"v9fs-poll");
if (!IS_ERR(pproc)) {
vpt->task = pproc;
INIT_LIST_HEAD(&vpt->mux_list);
vpt->muxnum = 0;
p9_mux_poll_task_num++;
wake_up_process(vpt->task);
}
break;
}
}
if (i >= ARRAY_SIZE(p9_mux_poll_tasks))
P9_DPRINTK(P9_DEBUG_ERROR,
"warning: no free poll slots\n");
}
n = (p9_mux_num + 1) / p9_mux_poll_task_num +
((p9_mux_num + 1) % p9_mux_poll_task_num ? 1 : 0);
vptlast = NULL;
for (i = 0; i < ARRAY_SIZE(p9_mux_poll_tasks); i++) {
vpt = &p9_mux_poll_tasks[i];
if (vpt->task != NULL) {
vptlast = vpt;
if (vpt->muxnum < n) {
P9_DPRINTK(P9_DEBUG_MUX, "put in proc %d\n", i);
list_add(&m->mux_list, &vpt->mux_list);
vpt->muxnum++;
m->poll_task = vpt;
memset(&m->poll_waddr, 0,
sizeof(m->poll_waddr));
init_poll_funcptr(&m->pt, p9_pollwait);
break;
}
}
}
if (i >= ARRAY_SIZE(p9_mux_poll_tasks)) {
if (vptlast == NULL) {
mutex_unlock(&p9_mux_task_lock);
return -ENOMEM;
}
P9_DPRINTK(P9_DEBUG_MUX, "put in proc %d\n", i);
list_add(&m->mux_list, &vptlast->mux_list);
vptlast->muxnum++;
m->poll_task = vptlast;
memset(&m->poll_waddr, 0, sizeof(m->poll_waddr));
init_poll_funcptr(&m->pt, p9_pollwait);
}
p9_mux_num++;
mutex_unlock(&p9_mux_task_lock);
return 0;
}
static void p9_mux_poll_stop(struct p9_conn *m)
{
int i;
struct p9_mux_poll_task *vpt;
mutex_lock(&p9_mux_task_lock);
vpt = m->poll_task;
list_del(&m->mux_list);
for (i = 0; i < ARRAY_SIZE(m->poll_waddr); i++) {
if (m->poll_waddr[i] != NULL) {
remove_wait_queue(m->poll_waddr[i], &m->poll_wait[i]);
m->poll_waddr[i] = NULL;
}
}
vpt->muxnum--;
if (!vpt->muxnum) {
P9_DPRINTK(P9_DEBUG_MUX, "destroy proc %p\n", vpt);
kthread_stop(vpt->task);
vpt->task = NULL;
p9_mux_poll_task_num--;
}
p9_mux_num--;
mutex_unlock(&p9_mux_task_lock);
}
/**
* p9_conn_create - allocate and initialize the per-session mux data
* Creates the polling task if this is the first session.
*
* @trans - transport structure
* @msize - maximum message size
* @extended - extended flag
*/
static struct p9_conn *p9_conn_create(struct p9_trans *trans)
{
int i, n;
struct p9_conn *m, *mtmp;
P9_DPRINTK(P9_DEBUG_MUX, "transport %p msize %d\n", trans,
trans->msize);
m = kmalloc(sizeof(struct p9_conn), GFP_KERNEL);
if (!m)
return ERR_PTR(-ENOMEM);
spin_lock_init(&m->lock);
INIT_LIST_HEAD(&m->mux_list);
m->msize = trans->msize;
m->extended = trans->extended;
m->trans = trans;
m->tagpool = p9_idpool_create();
if (IS_ERR(m->tagpool)) {
mtmp = ERR_PTR(-ENOMEM);
kfree(m);
return mtmp;
}
m->err = 0;
init_waitqueue_head(&m->equeue);
INIT_LIST_HEAD(&m->req_list);
INIT_LIST_HEAD(&m->unsent_req_list);
m->rcall = NULL;
m->rpos = 0;
m->rbuf = NULL;
m->wpos = m->wsize = 0;
m->wbuf = NULL;
INIT_WORK(&m->rq, p9_read_work);
INIT_WORK(&m->wq, p9_write_work);
m->wsched = 0;
memset(&m->poll_waddr, 0, sizeof(m->poll_waddr));
m->poll_task = NULL;
n = p9_mux_poll_start(m);
if (n) {
kfree(m);
return ERR_PTR(n);
}
n = p9_fd_poll(trans, &m->pt);
if (n & POLLIN) {
P9_DPRINTK(P9_DEBUG_MUX, "mux %p can read\n", m);
set_bit(Rpending, &m->wsched);
}
if (n & POLLOUT) {
P9_DPRINTK(P9_DEBUG_MUX, "mux %p can write\n", m);
set_bit(Wpending, &m->wsched);
}
for (i = 0; i < ARRAY_SIZE(m->poll_waddr); i++) {
if (IS_ERR(m->poll_waddr[i])) {
p9_mux_poll_stop(m);
mtmp = (void *)m->poll_waddr; /* the error code */
kfree(m);
m = mtmp;
break;
}
}
return m;
}
/**
* p9_mux_destroy - cancels all pending requests and frees mux resources
*/
static void p9_conn_destroy(struct p9_conn *m)
{
P9_DPRINTK(P9_DEBUG_MUX, "mux %p prev %p next %p\n", m,
m->mux_list.prev, m->mux_list.next);
p9_conn_cancel(m, -ECONNRESET);
if (!list_empty(&m->req_list)) {
/* wait until all processes waiting on this session exit */
P9_DPRINTK(P9_DEBUG_MUX,
"mux %p waiting for empty request queue\n", m);
wait_event_timeout(m->equeue, (list_empty(&m->req_list)), 5000);
P9_DPRINTK(P9_DEBUG_MUX, "mux %p request queue empty: %d\n", m,
list_empty(&m->req_list));
}
p9_mux_poll_stop(m);
m->trans = NULL;
p9_idpool_destroy(m->tagpool);
kfree(m);
}
/**
* p9_pollwait - called by files poll operation to add v9fs-poll task
* to files wait queue
*/
static void
p9_pollwait(struct file *filp, wait_queue_head_t *wait_address, poll_table *p)
{
int i;
struct p9_conn *m;
m = container_of(p, struct p9_conn, pt);
for (i = 0; i < ARRAY_SIZE(m->poll_waddr); i++)
if (m->poll_waddr[i] == NULL)
break;
if (i >= ARRAY_SIZE(m->poll_waddr)) {
P9_DPRINTK(P9_DEBUG_ERROR, "not enough wait_address slots\n");
return;
}
m->poll_waddr[i] = wait_address;
if (!wait_address) {
P9_DPRINTK(P9_DEBUG_ERROR, "no wait_address\n");
m->poll_waddr[i] = ERR_PTR(-EIO);
return;
}
init_waitqueue_entry(&m->poll_wait[i], m->poll_task->task);
add_wait_queue(wait_address, &m->poll_wait[i]);
}
/**
* p9_poll_mux - polls a mux and schedules read or write works if necessary
*/
static void p9_poll_mux(struct p9_conn *m)
{
int n;
if (m->err < 0)
return;
n = p9_fd_poll(m->trans, NULL);
if (n < 0 || n & (POLLERR | POLLHUP | POLLNVAL)) {
P9_DPRINTK(P9_DEBUG_MUX, "error mux %p err %d\n", m, n);
if (n >= 0)
n = -ECONNRESET;
p9_conn_cancel(m, n);
}
if (n & POLLIN) {
set_bit(Rpending, &m->wsched);
P9_DPRINTK(P9_DEBUG_MUX, "mux %p can read\n", m);
if (!test_and_set_bit(Rworksched, &m->wsched)) {
P9_DPRINTK(P9_DEBUG_MUX, "schedule read work %p\n", m);
queue_work(p9_mux_wq, &m->rq);
}
}
if (n & POLLOUT) {
set_bit(Wpending, &m->wsched);
P9_DPRINTK(P9_DEBUG_MUX, "mux %p can write\n", m);
if ((m->wsize || !list_empty(&m->unsent_req_list))
&& !test_and_set_bit(Wworksched, &m->wsched)) {
P9_DPRINTK(P9_DEBUG_MUX, "schedule write work %p\n", m);
queue_work(p9_mux_wq, &m->wq);
}
}
}
/**
* p9_poll_proc - polls all v9fs transports for new events and queues
* the appropriate work to the work queue
*/
static int p9_poll_proc(void *a)
{
struct p9_conn *m, *mtmp;
struct p9_mux_poll_task *vpt;
vpt = a;
P9_DPRINTK(P9_DEBUG_MUX, "start %p %p\n", current, vpt);
while (!kthread_should_stop()) {
set_current_state(TASK_INTERRUPTIBLE);
list_for_each_entry_safe(m, mtmp, &vpt->mux_list, mux_list) {
p9_poll_mux(m);
}
P9_DPRINTK(P9_DEBUG_MUX, "sleeping...\n");
schedule_timeout(SCHED_TIMEOUT * HZ);
}
__set_current_state(TASK_RUNNING);
P9_DPRINTK(P9_DEBUG_MUX, "finish\n");
return 0;
}
/**
* p9_write_work - called when a transport can send some data
*/
static void p9_write_work(struct work_struct *work)
{
int n, err;
struct p9_conn *m;
struct p9_req *req;
m = container_of(work, struct p9_conn, wq);
if (m->err < 0) {
clear_bit(Wworksched, &m->wsched);
return;
}
if (!m->wsize) {
if (list_empty(&m->unsent_req_list)) {
clear_bit(Wworksched, &m->wsched);
return;
}
spin_lock(&m->lock);
again:
req = list_entry(m->unsent_req_list.next, struct p9_req,
req_list);
list_move_tail(&req->req_list, &m->req_list);
if (req->err == ERREQFLUSH)
goto again;
m->wbuf = req->tcall->sdata;
m->wsize = req->tcall->size;
m->wpos = 0;
spin_unlock(&m->lock);
}
P9_DPRINTK(P9_DEBUG_MUX, "mux %p pos %d size %d\n", m, m->wpos,
m->wsize);
clear_bit(Wpending, &m->wsched);
err = p9_fd_write(m->trans, m->wbuf + m->wpos, m->wsize - m->wpos);
P9_DPRINTK(P9_DEBUG_MUX, "mux %p sent %d bytes\n", m, err);
if (err == -EAGAIN) {
clear_bit(Wworksched, &m->wsched);
return;
}
if (err < 0)
goto error;
else if (err == 0) {
err = -EREMOTEIO;
goto error;
}
m->wpos += err;
if (m->wpos == m->wsize)
m->wpos = m->wsize = 0;
if (m->wsize == 0 && !list_empty(&m->unsent_req_list)) {
if (test_and_clear_bit(Wpending, &m->wsched))
n = POLLOUT;
else
n = p9_fd_poll(m->trans, NULL);
if (n & POLLOUT) {
P9_DPRINTK(P9_DEBUG_MUX, "schedule write work %p\n", m);
queue_work(p9_mux_wq, &m->wq);
} else
clear_bit(Wworksched, &m->wsched);
} else
clear_bit(Wworksched, &m->wsched);
return;
error:
p9_conn_cancel(m, err);
clear_bit(Wworksched, &m->wsched);
}
static void process_request(struct p9_conn *m, struct p9_req *req)
{
int ecode;
struct p9_str *ename;
if (!req->err && req->rcall->id == P9_RERROR) {
ecode = req->rcall->params.rerror.errno;
ename = &req->rcall->params.rerror.error;
P9_DPRINTK(P9_DEBUG_MUX, "Rerror %.*s\n", ename->len,
ename->str);
if (m->extended)
req->err = -ecode;
if (!req->err) {
req->err = p9_errstr2errno(ename->str, ename->len);
/* string match failed */
if (!req->err) {
PRINT_FCALL_ERROR("unknown error", req->rcall);
req->err = -ESERVERFAULT;
}
}
} else if (req->tcall && req->rcall->id != req->tcall->id + 1) {
P9_DPRINTK(P9_DEBUG_ERROR,
"fcall mismatch: expected %d, got %d\n",
req->tcall->id + 1, req->rcall->id);
if (!req->err)
req->err = -EIO;
}
}
/**
* p9_read_work - called when there is some data to be read from a transport
*/
static void p9_read_work(struct work_struct *work)
{
int n, err;
struct p9_conn *m;
struct p9_req *req, *rptr, *rreq;
struct p9_fcall *rcall;
char *rbuf;
m = container_of(work, struct p9_conn, rq);
if (m->err < 0)
return;
rcall = NULL;
P9_DPRINTK(P9_DEBUG_MUX, "start mux %p pos %d\n", m, m->rpos);
if (!m->rcall) {
m->rcall =
kmalloc(sizeof(struct p9_fcall) + m->msize, GFP_KERNEL);
if (!m->rcall) {
err = -ENOMEM;
goto error;
}
m->rbuf = (char *)m->rcall + sizeof(struct p9_fcall);
m->rpos = 0;
}
clear_bit(Rpending, &m->wsched);
err = p9_fd_read(m->trans, m->rbuf + m->rpos, m->msize - m->rpos);
P9_DPRINTK(P9_DEBUG_MUX, "mux %p got %d bytes\n", m, err);
if (err == -EAGAIN) {
clear_bit(Rworksched, &m->wsched);
return;
}
if (err <= 0)
goto error;
m->rpos += err;
while (m->rpos > 4) {
n = le32_to_cpu(*(__le32 *) m->rbuf);
if (n >= m->msize) {
P9_DPRINTK(P9_DEBUG_ERROR,
"requested packet size too big: %d\n", n);
err = -EIO;
goto error;
}
if (m->rpos < n)
break;
err =
p9_deserialize_fcall(m->rbuf, n, m->rcall, m->extended);
if (err < 0)
goto error;
#ifdef CONFIG_NET_9P_DEBUG
if ((p9_debug_level&P9_DEBUG_FCALL) == P9_DEBUG_FCALL) {
char buf[150];
p9_printfcall(buf, sizeof(buf), m->rcall,
m->extended);
printk(KERN_NOTICE ">>> %p %s\n", m, buf);
}
#endif
rcall = m->rcall;
rbuf = m->rbuf;
if (m->rpos > n) {
m->rcall = kmalloc(sizeof(struct p9_fcall) + m->msize,
GFP_KERNEL);
if (!m->rcall) {
err = -ENOMEM;
goto error;
}
m->rbuf = (char *)m->rcall + sizeof(struct p9_fcall);
memmove(m->rbuf, rbuf + n, m->rpos - n);
m->rpos -= n;
} else {
m->rcall = NULL;
m->rbuf = NULL;
m->rpos = 0;
}
P9_DPRINTK(P9_DEBUG_MUX, "mux %p fcall id %d tag %d\n", m,
rcall->id, rcall->tag);
req = NULL;
spin_lock(&m->lock);
list_for_each_entry_safe(rreq, rptr, &m->req_list, req_list) {
if (rreq->tag == rcall->tag) {
req = rreq;
if (req->flush != Flushing)
list_del(&req->req_list);
break;
}
}
spin_unlock(&m->lock);
if (req) {
req->rcall = rcall;
process_request(m, req);
if (req->flush != Flushing) {
if (req->cb)
(*req->cb) (req, req->cba);
else
kfree(req->rcall);
wake_up(&m->equeue);
}
} else {
if (err >= 0 && rcall->id != P9_RFLUSH)
P9_DPRINTK(P9_DEBUG_ERROR,
"unexpected response mux %p id %d tag %d\n",
m, rcall->id, rcall->tag);
kfree(rcall);
}
}
if (!list_empty(&m->req_list)) {
if (test_and_clear_bit(Rpending, &m->wsched))
n = POLLIN;
else
n = p9_fd_poll(m->trans, NULL);
if (n & POLLIN) {
P9_DPRINTK(P9_DEBUG_MUX, "schedule read work %p\n", m);
queue_work(p9_mux_wq, &m->rq);
} else
clear_bit(Rworksched, &m->wsched);
} else
clear_bit(Rworksched, &m->wsched);
return;
error:
p9_conn_cancel(m, err);
clear_bit(Rworksched, &m->wsched);
}
/**
* p9_send_request - send 9P request
* The function can sleep until the request is scheduled for sending.
* The function can be interrupted. Return from the function is not
* a guarantee that the request is sent successfully. Can return errors
* that can be retrieved by PTR_ERR macros.
*
* @m: mux data
* @tc: request to be sent
* @cb: callback function to call when response is received
* @cba: parameter to pass to the callback function
*/
static struct p9_req *p9_send_request(struct p9_conn *m,
struct p9_fcall *tc,
p9_conn_req_callback cb, void *cba)
{
int n;
struct p9_req *req;
P9_DPRINTK(P9_DEBUG_MUX, "mux %p task %p tcall %p id %d\n", m, current,
tc, tc->id);
if (m->err < 0)
return ERR_PTR(m->err);
req = kmalloc(sizeof(struct p9_req), GFP_KERNEL);
if (!req)
return ERR_PTR(-ENOMEM);
if (tc->id == P9_TVERSION)
n = P9_NOTAG;
else
n = p9_mux_get_tag(m);
if (n < 0)
return ERR_PTR(-ENOMEM);
p9_set_tag(tc, n);
#ifdef CONFIG_NET_9P_DEBUG
if ((p9_debug_level&P9_DEBUG_FCALL) == P9_DEBUG_FCALL) {
char buf[150];
p9_printfcall(buf, sizeof(buf), tc, m->extended);
printk(KERN_NOTICE "<<< %p %s\n", m, buf);
}
#endif
spin_lock_init(&req->lock);
req->tag = n;
req->tcall = tc;
req->rcall = NULL;
req->err = 0;
req->cb = cb;
req->cba = cba;
req->flush = None;
spin_lock(&m->lock);
list_add_tail(&req->req_list, &m->unsent_req_list);
spin_unlock(&m->lock);
if (test_and_clear_bit(Wpending, &m->wsched))
n = POLLOUT;
else
n = p9_fd_poll(m->trans, NULL);
if (n & POLLOUT && !test_and_set_bit(Wworksched, &m->wsched))
queue_work(p9_mux_wq, &m->wq);
return req;
}
static void p9_mux_free_request(struct p9_conn *m, struct p9_req *req)
{
p9_mux_put_tag(m, req->tag);
kfree(req);
}
static void p9_mux_flush_cb(struct p9_req *freq, void *a)
{
p9_conn_req_callback cb;
int tag;
struct p9_conn *m;
struct p9_req *req, *rreq, *rptr;
m = a;
P9_DPRINTK(P9_DEBUG_MUX, "mux %p tc %p rc %p err %d oldtag %d\n", m,
freq->tcall, freq->rcall, freq->err,
freq->tcall->params.tflush.oldtag);
spin_lock(&m->lock);
cb = NULL;
tag = freq->tcall->params.tflush.oldtag;
req = NULL;
list_for_each_entry_safe(rreq, rptr, &m->req_list, req_list) {
if (rreq->tag == tag) {
req = rreq;
list_del(&req->req_list);
break;
}
}
spin_unlock(&m->lock);
if (req) {
spin_lock(&req->lock);
req->flush = Flushed;
spin_unlock(&req->lock);
if (req->cb)
(*req->cb) (req, req->cba);
else
kfree(req->rcall);
wake_up(&m->equeue);
}
kfree(freq->tcall);
kfree(freq->rcall);
p9_mux_free_request(m, freq);
}
static int
p9_mux_flush_request(struct p9_conn *m, struct p9_req *req)
{
struct p9_fcall *fc;
struct p9_req *rreq, *rptr;
P9_DPRINTK(P9_DEBUG_MUX, "mux %p req %p tag %d\n", m, req, req->tag);
/* if a response was received for a request, do nothing */
spin_lock(&req->lock);
if (req->rcall || req->err) {
spin_unlock(&req->lock);
P9_DPRINTK(P9_DEBUG_MUX,
"mux %p req %p response already received\n", m, req);
return 0;
}
req->flush = Flushing;
spin_unlock(&req->lock);
spin_lock(&m->lock);
/* if the request is not sent yet, just remove it from the list */
list_for_each_entry_safe(rreq, rptr, &m->unsent_req_list, req_list) {
if (rreq->tag == req->tag) {
P9_DPRINTK(P9_DEBUG_MUX,
"mux %p req %p request is not sent yet\n", m, req);
list_del(&rreq->req_list);
req->flush = Flushed;
spin_unlock(&m->lock);
if (req->cb)
(*req->cb) (req, req->cba);
return 0;
}
}
spin_unlock(&m->lock);
clear_thread_flag(TIF_SIGPENDING);
fc = p9_create_tflush(req->tag);
p9_send_request(m, fc, p9_mux_flush_cb, m);
return 1;
}
static void
p9_conn_rpc_cb(struct p9_req *req, void *a)
{
struct p9_mux_rpc *r;
P9_DPRINTK(P9_DEBUG_MUX, "req %p r %p\n", req, a);
r = a;
r->rcall = req->rcall;
r->err = req->err;
if (req->flush != None && !req->err)
r->err = -ERESTARTSYS;
wake_up(&r->wqueue);
}
/**
* p9_fd_rpc- sends 9P request and waits until a response is available.
* The function can be interrupted.
* @m: mux data
* @tc: request to be sent
* @rc: pointer where a pointer to the response is stored
*/
int
p9_fd_rpc(struct p9_trans *t, struct p9_fcall *tc, struct p9_fcall **rc)
{
struct p9_trans_fd *p = t->priv;
struct p9_conn *m = p->conn;
int err, sigpending;
unsigned long flags;
struct p9_req *req;
struct p9_mux_rpc r;
r.err = 0;
r.tcall = tc;
r.rcall = NULL;
r.m = m;
init_waitqueue_head(&r.wqueue);
if (rc)
*rc = NULL;
sigpending = 0;
if (signal_pending(current)) {
sigpending = 1;
clear_thread_flag(TIF_SIGPENDING);
}
req = p9_send_request(m, tc, p9_conn_rpc_cb, &r);
if (IS_ERR(req)) {
err = PTR_ERR(req);
P9_DPRINTK(P9_DEBUG_MUX, "error %d\n", err);
return err;
}
err = wait_event_interruptible(r.wqueue, r.rcall != NULL || r.err < 0);
if (r.err < 0)
err = r.err;
if (err == -ERESTARTSYS && m->trans->status == Connected
&& m->err == 0) {
if (p9_mux_flush_request(m, req)) {
/* wait until we get response of the flush message */
do {
clear_thread_flag(TIF_SIGPENDING);
err = wait_event_interruptible(r.wqueue,
r.rcall || r.err);
} while (!r.rcall && !r.err && err == -ERESTARTSYS &&
m->trans->status == Connected && !m->err);
err = -ERESTARTSYS;
}
sigpending = 1;
}
if (sigpending) {
spin_lock_irqsave(&current->sighand->siglock, flags);
recalc_sigpending();
spin_unlock_irqrestore(&current->sighand->siglock, flags);
}
if (rc)
*rc = r.rcall;
else
kfree(r.rcall);
p9_mux_free_request(m, req);
if (err > 0)
err = -EIO;
return err;
}
#ifdef P9_NONBLOCK
/**
* p9_conn_rpcnb - sends 9P request without waiting for response.
* @m: mux data
* @tc: request to be sent
* @cb: callback function to be called when response arrives
* @cba: value to pass to the callback function
*/
int p9_conn_rpcnb(struct p9_conn *m, struct p9_fcall *tc,
p9_conn_req_callback cb, void *a)
{
int err;
struct p9_req *req;
req = p9_send_request(m, tc, cb, a);
if (IS_ERR(req)) {
err = PTR_ERR(req);
P9_DPRINTK(P9_DEBUG_MUX, "error %d\n", err);
return PTR_ERR(req);
}
P9_DPRINTK(P9_DEBUG_MUX, "mux %p tc %p tag %d\n", m, tc, req->tag);
return 0;
}
#endif /* P9_NONBLOCK */
/**
* p9_conn_cancel - cancel all pending requests with error
* @m: mux data
* @err: error code
*/
void p9_conn_cancel(struct p9_conn *m, int err)
{
struct p9_req *req, *rtmp;
LIST_HEAD(cancel_list);
P9_DPRINTK(P9_DEBUG_ERROR, "mux %p err %d\n", m, err);
m->err = err;
spin_lock(&m->lock);
list_for_each_entry_safe(req, rtmp, &m->req_list, req_list) {
list_move(&req->req_list, &cancel_list);
}
list_for_each_entry_safe(req, rtmp, &m->unsent_req_list, req_list) {
list_move(&req->req_list, &cancel_list);
}
spin_unlock(&m->lock);
list_for_each_entry_safe(req, rtmp, &cancel_list, req_list) {
list_del(&req->req_list);
if (!req->err)
req->err = err;
if (req->cb)
(*req->cb) (req, req->cba);
else
kfree(req->rcall);
}
wake_up(&m->equeue);
}
/**
* v9fs_parse_options - parse mount options into session structure
* @options: options string passed from mount
......@@ -268,7 +1294,7 @@ end:
}
/**
* p9_sock_close - shutdown socket
* p9_fd_close - shutdown socket
* @trans: private socket structure
*
*/
......@@ -284,6 +1310,8 @@ static void p9_fd_close(struct p9_trans *trans)
if (!ts)
return;
p9_conn_destroy(ts->conn);
trans->status = Disconnected;
if (ts->rd)
fput(ts->rd);
......@@ -292,13 +1320,15 @@ static void p9_fd_close(struct p9_trans *trans)
kfree(ts);
}
static struct p9_trans *p9_trans_create_tcp(const char *addr, char *args)
static struct p9_trans *
p9_trans_create_tcp(const char *addr, char *args, int msize, unsigned char dotu)
{
int err;
struct p9_trans *trans;
struct socket *csocket;
struct sockaddr_in sin_server;
struct p9_fd_opts opts;
struct p9_trans_fd *p;
parse_opts(args, &opts);
......@@ -306,11 +1336,10 @@ static struct p9_trans *p9_trans_create_tcp(const char *addr, char *args)
trans = kmalloc(sizeof(struct p9_trans), GFP_KERNEL);
if (!trans)
return ERR_PTR(-ENOMEM);
trans->write = p9_fd_write;
trans->read = p9_fd_read;
trans->msize = msize;
trans->extended = dotu;
trans->rpc = p9_fd_rpc;
trans->close = p9_fd_close;
trans->poll = p9_fd_poll;
sin_server.sin_family = AF_INET;
sin_server.sin_addr.s_addr = in_aton(addr);
......@@ -337,6 +1366,14 @@ static struct p9_trans *p9_trans_create_tcp(const char *addr, char *args)
if (err < 0)
goto error;
p = (struct p9_trans_fd *) trans->priv;
p->conn = p9_conn_create(trans);
if (IS_ERR(p->conn)) {
err = PTR_ERR(p->conn);
p->conn = NULL;
goto error;
}
return trans;
error:
......@@ -347,22 +1384,23 @@ error:
return ERR_PTR(err);
}
static struct p9_trans *p9_trans_create_unix(const char *addr, char *args)
static struct p9_trans *
p9_trans_create_unix(const char *addr, char *args, int msize,
unsigned char dotu)
{
int err;
struct socket *csocket;
struct sockaddr_un sun_server;
struct p9_trans *trans;
struct p9_trans_fd *p;
csocket = NULL;
trans = kmalloc(sizeof(struct p9_trans), GFP_KERNEL);
if (!trans)
return ERR_PTR(-ENOMEM);
trans->write = p9_fd_write;
trans->read = p9_fd_read;
trans->rpc = p9_fd_rpc;
trans->close = p9_fd_close;
trans->poll = p9_fd_poll;
if (strlen(addr) > UNIX_PATH_MAX) {
P9_EPRINTK(KERN_ERR, "p9_trans_unix: address too long: %s\n",
......@@ -387,6 +1425,16 @@ static struct p9_trans *p9_trans_create_unix(const char *addr, char *args)
if (err < 0)
goto error;
trans->msize = msize;
trans->extended = dotu;
p = (struct p9_trans_fd *) trans->priv;
p->conn = p9_conn_create(trans);
if (IS_ERR(p->conn)) {
err = PTR_ERR(p->conn);
p->conn = NULL;
goto error;
}
return trans;
error:
......@@ -397,11 +1445,14 @@ error:
return ERR_PTR(err);
}
static struct p9_trans *p9_trans_create_fd(const char *name, char *args)
static struct p9_trans *
p9_trans_create_fd(const char *name, char *args, int msize,
unsigned char extended)
{
int err;
struct p9_trans *trans;
struct p9_fd_opts opts;
struct p9_trans_fd *p;
parse_opts(args, &opts);
......@@ -414,15 +1465,23 @@ static struct p9_trans *p9_trans_create_fd(const char *name, char *args)
if (!trans)
return ERR_PTR(-ENOMEM);
trans->write = p9_fd_write;
trans->read = p9_fd_read;
trans->rpc = p9_fd_rpc;
trans->close = p9_fd_close;
trans->poll = p9_fd_poll;
err = p9_fd_open(trans, opts.rfd, opts.wfd);
if (err < 0)
goto error;
trans->msize = msize;
trans->extended = extended;
p = (struct p9_trans_fd *) trans->priv;
p->conn = p9_conn_create(trans);
if (IS_ERR(p->conn)) {
err = PTR_ERR(p->conn);
p->conn = NULL;
goto error;
}
return trans;
error:
......@@ -453,6 +1512,12 @@ static struct p9_trans_module p9_fd_trans = {
static int __init p9_trans_fd_init(void)
{
int ret = p9_mux_global_init();
if (ret) {
printk(KERN_WARNING "9p: starting mux failed\n");
return ret;
}
v9fs_register_trans(&p9_tcp_trans);
v9fs_register_trans(&p9_unix_trans);
v9fs_register_trans(&p9_fd_trans);
......@@ -460,13 +1525,7 @@ static int __init p9_trans_fd_init(void)
return 1;
}
static void __exit p9_trans_fd_exit(void) {
printk(KERN_ERR "Removal of 9p transports not implemented\n");
BUG();
}
module_init(p9_trans_fd_init);
module_exit(p9_trans_fd_exit);
MODULE_AUTHOR("Latchesar Ionkov <lucho@ionkov.net>");
MODULE_AUTHOR("Eric Van Hensbergen <ericvh@gmail.com>");
......
/*
* The Guest 9p transport driver
*
* This is a trivial pipe-based transport driver based on the lguest console
* code: we use lguest's DMA mechanism to send bytes out, and register a
* DMA buffer to receive bytes in. It is assumed to be present and available
* from the very beginning of boot.
*
* This may be have been done by just instaniating another HVC console,
* but HVC's blocksize of 16 bytes is annoying and painful to performance.
*
* A more efficient transport could be built based on the virtio block driver
* but it requires some changes in the 9p transport model (which are in
* progress)
* This is a block based transport driver based on the lguest block driver
* code.
*
*/
/*
......@@ -55,11 +46,25 @@
#include <linux/virtio.h>
#include <linux/virtio_9p.h>
#define VIRTQUEUE_NUM 128
/* a single mutex to manage channel initialization and attachment */
static DECLARE_MUTEX(virtio_9p_lock);
/* global which tracks highest initialized channel */
static int chan_index;
#define P9_INIT_MAXTAG 16
#define REQ_STATUS_IDLE 0
#define REQ_STATUS_SENT 1
#define REQ_STATUS_RCVD 2
#define REQ_STATUS_FLSH 3
struct p9_req_t {
int status;
wait_queue_head_t *wq;
};
/* We keep all per-channel information in a structure.
* This structure is allocated within the devices dev->mem space.
* A pointer to the structure will get put in the transport private.
......@@ -68,146 +73,198 @@ static struct virtio_chan {
bool initialized; /* channel is initialized */
bool inuse; /* channel is in use */
struct virtqueue *in_vq, *out_vq;
spinlock_t lock;
struct virtio_device *vdev;
struct virtqueue *vq;
/* This is our input buffer, and how much data is left in it. */
unsigned int in_len;
char *in, *inbuf;
struct p9_idpool *tagpool;
struct p9_req_t *reqs;
int max_tag;
wait_queue_head_t wq; /* waitq for buffer */
/* Scatterlist: can be too big for stack. */
struct scatterlist sg[VIRTQUEUE_NUM];
} channels[MAX_9P_CHAN];
/* Lookup requests by tag */
static struct p9_req_t *p9_lookup_tag(struct virtio_chan *c, u16 tag)
{
/* This looks up the original request by tag so we know which
* buffer to read the data into */
tag++;
while (tag >= c->max_tag) {
int old_max = c->max_tag;
int count;
if (c->max_tag)
c->max_tag *= 2;
else
c->max_tag = P9_INIT_MAXTAG;
c->reqs = krealloc(c->reqs, sizeof(struct p9_req_t)*c->max_tag,
GFP_ATOMIC);
if (!c->reqs) {
printk(KERN_ERR "Couldn't grow tag array\n");
BUG();
}
for (count = old_max; count < c->max_tag; count++) {
c->reqs[count].status = REQ_STATUS_IDLE;
c->reqs[count].wq = kmalloc(sizeof(wait_queue_t),
GFP_ATOMIC);
if (!c->reqs[count].wq) {
printk(KERN_ERR "Couldn't grow tag array\n");
BUG();
}
init_waitqueue_head(c->reqs[count].wq);
}
}
return &c->reqs[tag];
}
/* How many bytes left in this page. */
static unsigned int rest_of_page(void *data)
{
return PAGE_SIZE - ((unsigned long)data % PAGE_SIZE);
}
static int p9_virtio_write(struct p9_trans *trans, void *buf, int count)
static void p9_virtio_close(struct p9_trans *trans)
{
struct virtio_chan *chan = (struct virtio_chan *) trans->priv;
struct virtqueue *out_vq = chan->out_vq;
struct scatterlist sg[1];
unsigned int len;
struct virtio_chan *chan = trans->priv;
int count;
unsigned int flags;
P9_DPRINTK(P9_DEBUG_TRANS, "9p debug: virtio write (%d)\n", count);
spin_lock_irqsave(&chan->lock, flags);
p9_idpool_destroy(chan->tagpool);
for (count = 0; count < chan->max_tag; count++)
kfree(chan->reqs[count].wq);
kfree(chan->reqs);
chan->max_tag = 0;
spin_unlock_irqrestore(&chan->lock, flags);
/* keep it simple - make sure we don't overflow a page */
if (rest_of_page(buf) < count)
count = rest_of_page(buf);
down(&virtio_9p_lock);
chan->inuse = false;
up(&virtio_9p_lock);
sg_init_one(sg, buf, count);
kfree(trans);
}
/* add_buf wants a token to identify this buffer: we hand it any
* non-NULL pointer, since there's only ever one buffer. */
if (out_vq->vq_ops->add_buf(out_vq, sg, 1, 0, (void *)1) == 0) {
/* Tell Host to go! */
out_vq->vq_ops->kick(out_vq);
/* Chill out until it's done with the buffer. */
while (!out_vq->vq_ops->get_buf(out_vq, &len))
cpu_relax();
static void req_done(struct virtqueue *vq)
{
struct virtio_chan *chan = vq->vdev->priv;
struct p9_fcall *rc;
unsigned int len;
unsigned long flags;
struct p9_req_t *req;
spin_lock_irqsave(&chan->lock, flags);
while ((rc = chan->vq->vq_ops->get_buf(chan->vq, &len)) != NULL) {
req = p9_lookup_tag(chan, rc->tag);
req->status = REQ_STATUS_RCVD;
wake_up(req->wq);
}
P9_DPRINTK(P9_DEBUG_TRANS, "9p debug: virtio wrote (%d)\n", count);
/* We're expected to return the amount of data we wrote: all of it. */
return count;
/* In case queue is stopped waiting for more buffers. */
spin_unlock_irqrestore(&chan->lock, flags);
}
/* Create a scatter-gather list representing our input buffer and put it in the
* queue. */
static void add_inbuf(struct virtio_chan *chan)
static int
pack_sg_list(struct scatterlist *sg, int start, int limit, char *data,
int count)
{
struct scatterlist sg[1];
sg_init_one(sg, chan->inbuf, PAGE_SIZE);
/* We should always be able to add one buffer to an empty queue. */
if (chan->in_vq->vq_ops->add_buf(chan->in_vq, sg, 0, 1, chan->inbuf))
int s;
int index = start;
while (count) {
s = rest_of_page(data);
if (s > count)
s = count;
sg_set_buf(&sg[index++], data, s);
count -= s;
data += s;
if (index > limit)
BUG();
chan->in_vq->vq_ops->kick(chan->in_vq);
}
return index-start;
}
static int p9_virtio_read(struct p9_trans *trans, void *buf, int count)
static int
p9_virtio_rpc(struct p9_trans *t, struct p9_fcall *tc, struct p9_fcall **rc)
{
struct virtio_chan *chan = (struct virtio_chan *) trans->priv;
struct virtqueue *in_vq = chan->in_vq;
P9_DPRINTK(P9_DEBUG_TRANS, "9p debug: virtio read (%d)\n", count);
int in, out;
int n, err, size;
struct virtio_chan *chan = t->priv;
char *rdata;
struct p9_req_t *req;
unsigned long flags;
if (*rc == NULL) {
*rc = kmalloc(sizeof(struct p9_fcall) + t->msize, GFP_KERNEL);
if (!*rc)
return -ENOMEM;
}
/* If we don't have an input queue yet, we can't get input. */
BUG_ON(!in_vq);
rdata = (char *)*rc+sizeof(struct p9_fcall);
/* No buffer? Try to get one. */
if (!chan->in_len) {
chan->in = in_vq->vq_ops->get_buf(in_vq, &chan->in_len);
if (!chan->in)
return 0;
n = P9_NOTAG;
if (tc->id != P9_TVERSION) {
n = p9_idpool_get(chan->tagpool);
if (n < 0)
return -ENOMEM;
}
/* You want more than we have to give? Well, try wanting less! */
if (chan->in_len < count)
count = chan->in_len;
spin_lock_irqsave(&chan->lock, flags);
req = p9_lookup_tag(chan, n);
spin_unlock_irqrestore(&chan->lock, flags);
/* Copy across to their buffer and increment offset. */
memcpy(buf, chan->in, count);
chan->in += count;
chan->in_len -= count;
p9_set_tag(tc, n);
/* Finished? Re-register buffer so Host will use it again. */
if (chan->in_len == 0)
add_inbuf(chan);
P9_DPRINTK(P9_DEBUG_TRANS, "9p debug: virtio rpc tag %d\n", n);
P9_DPRINTK(P9_DEBUG_TRANS, "9p debug: virtio finished read (%d)\n",
count);
out = pack_sg_list(chan->sg, 0, VIRTQUEUE_NUM, tc->sdata, tc->size);
in = pack_sg_list(chan->sg, out, VIRTQUEUE_NUM-out, rdata, t->msize);
return count;
}
req->status = REQ_STATUS_SENT;
/* The poll function is used by 9p transports to determine if there
* is there is activity available on a particular channel. In our case
* we use it to wait for a callback from the input routines.
*/
static unsigned int
p9_virtio_poll(struct p9_trans *trans, struct poll_table_struct *pt)
{
struct virtio_chan *chan = (struct virtio_chan *)trans->priv;
struct virtqueue *in_vq = chan->in_vq;
int ret = POLLOUT; /* we can always handle more output */
if (chan->vq->vq_ops->add_buf(chan->vq, chan->sg, out, in, tc)) {
P9_DPRINTK(P9_DEBUG_TRANS,
"9p debug: virtio rpc add_buf returned failure");
return -EIO;
}
poll_wait(NULL, &chan->wq, pt);
chan->vq->vq_ops->kick(chan->vq);
/* No buffer? Try to get one. */
if (!chan->in_len)
chan->in = in_vq->vq_ops->get_buf(in_vq, &chan->in_len);
wait_event(*req->wq, req->status == REQ_STATUS_RCVD);
if (chan->in_len)
ret |= POLLIN;
size = le32_to_cpu(*(__le32 *) rdata);
return ret;
}
err = p9_deserialize_fcall(rdata, size, *rc, t->extended);
if (err < 0) {
P9_DPRINTK(P9_DEBUG_TRANS,
"9p debug: virtio rpc deserialize returned %d\n", err);
return err;
}
static void p9_virtio_close(struct p9_trans *trans)
{
struct virtio_chan *chan = trans->priv;
#ifdef CONFIG_NET_9P_DEBUG
if ((p9_debug_level&P9_DEBUG_FCALL) == P9_DEBUG_FCALL) {
char buf[150];
down(&virtio_9p_lock);
chan->inuse = false;
up(&virtio_9p_lock);
p9_printfcall(buf, sizeof(buf), *rc, t->extended);
printk(KERN_NOTICE ">>> %p %s\n", t, buf);
}
#endif
kfree(trans);
}
if (n != P9_NOTAG && p9_idpool_check(n, chan->tagpool))
p9_idpool_put(n, chan->tagpool);
static void p9_virtio_intr(struct virtqueue *q)
{
struct virtio_chan *chan = q->vdev->priv;
req->status = REQ_STATUS_IDLE;
P9_DPRINTK(P9_DEBUG_TRANS, "9p poll_wakeup: %p\n", &chan->wq);
wake_up_interruptible(&chan->wq);
return 0;
}
static int p9_virtio_probe(struct virtio_device *dev)
static int p9_virtio_probe(struct virtio_device *vdev)
{
int err;
struct virtio_chan *chan;
......@@ -221,44 +278,29 @@ static int p9_virtio_probe(struct virtio_device *dev)
if (chan_index > MAX_9P_CHAN) {
printk(KERN_ERR "9p: virtio: Maximum channels exceeded\n");
BUG();
}
chan->vdev = dev;
/* This is the scratch page we use to receive console input */
chan->inbuf = kmalloc(PAGE_SIZE, GFP_KERNEL);
if (!chan->inbuf) {
err = -ENOMEM;
goto fail;
}
/* Find the input queue. */
dev->priv = chan;
chan->in_vq = dev->config->find_vq(dev, 0, p9_virtio_intr);
if (IS_ERR(chan->in_vq)) {
err = PTR_ERR(chan->in_vq);
goto free;
}
chan->vdev = vdev;
chan->out_vq = dev->config->find_vq(dev, 1, NULL);
if (IS_ERR(chan->out_vq)) {
err = PTR_ERR(chan->out_vq);
goto free_in_vq;
/* We expect one virtqueue, for requests. */
chan->vq = vdev->config->find_vq(vdev, 0, req_done);
if (IS_ERR(chan->vq)) {
err = PTR_ERR(chan->vq);
goto out_free_vq;
}
chan->vq->vdev->priv = chan;
spin_lock_init(&chan->lock);
init_waitqueue_head(&chan->wq);
sg_init_table(chan->sg, VIRTQUEUE_NUM);
/* Register the input buffer the first time. */
add_inbuf(chan);
chan->inuse = false;
chan->initialized = true;
return 0;
free_in_vq:
dev->config->del_vq(chan->in_vq);
free:
kfree(chan->inbuf);
out_free_vq:
vdev->config->del_vq(chan->vq);
fail:
down(&virtio_9p_lock);
chan_index--;
......@@ -271,11 +313,13 @@ fail:
* alternate channels by matching devname versus a virtio_config entry.
* We use a simple reference count mechanism to ensure that only a single
* mount has a channel open at a time. */
static struct p9_trans *p9_virtio_create(const char *devname, char *args)
static struct p9_trans *
p9_virtio_create(const char *devname, char *args, int msize,
unsigned char extended)
{
struct p9_trans *trans;
int index = 0;
struct virtio_chan *chan = channels;
int index = 0;
down(&virtio_9p_lock);
while (index < MAX_9P_CHAN) {
......@@ -290,25 +334,45 @@ static struct p9_trans *p9_virtio_create(const char *devname, char *args)
up(&virtio_9p_lock);
if (index >= MAX_9P_CHAN) {
printk(KERN_ERR "9p: virtio: couldn't find a free channel\n");
return NULL;
printk(KERN_ERR "9p: no channels available\n");
return ERR_PTR(-ENODEV);
}
chan->tagpool = p9_idpool_create();
if (IS_ERR(chan->tagpool)) {
printk(KERN_ERR "9p: couldn't allocate tagpool\n");
return ERR_PTR(-ENOMEM);
}
p9_idpool_get(chan->tagpool); /* reserve tag 0 */
chan->max_tag = 0;
chan->reqs = NULL;
trans = kmalloc(sizeof(struct p9_trans), GFP_KERNEL);
if (!trans) {
printk(KERN_ERR "9p: couldn't allocate transport\n");
return ERR_PTR(-ENOMEM);
}
trans->write = p9_virtio_write;
trans->read = p9_virtio_read;
trans->extended = extended;
trans->msize = msize;
trans->close = p9_virtio_close;
trans->poll = p9_virtio_poll;
trans->rpc = p9_virtio_rpc;
trans->priv = chan;
return trans;
}
static void p9_virtio_remove(struct virtio_device *vdev)
{
struct virtio_chan *chan = vdev->priv;
BUG_ON(chan->inuse);
if (chan->initialized) {
vdev->config->del_vq(chan->vq);
chan->initialized = false;
}
}
#define VIRTIO_ID_9P 9
static struct virtio_device_id id_table[] = {
......@@ -322,12 +386,13 @@ static struct virtio_driver p9_virtio_drv = {
.driver.owner = THIS_MODULE,
.id_table = id_table,
.probe = p9_virtio_probe,
.remove = p9_virtio_remove,
};
static struct p9_trans_module p9_virtio_trans = {
.name = "virtio",
.create = p9_virtio_create,
.maxsize = PAGE_SIZE,
.maxsize = PAGE_SIZE*16,
.def = 0,
};
......@@ -343,7 +408,13 @@ static int __init p9_virtio_init(void)
return register_virtio_driver(&p9_virtio_drv);
}
static void __exit p9_virtio_cleanup(void)
{
unregister_virtio_driver(&p9_virtio_drv);
}
module_init(p9_virtio_init);
module_exit(p9_virtio_cleanup);
MODULE_DEVICE_TABLE(virtio, id_table);
MODULE_AUTHOR("Eric Van Hensbergen <ericvh@gmail.com>");
......
......@@ -33,7 +33,7 @@
#include <net/9p/9p.h>
struct p9_idpool {
struct semaphore lock;
spinlock_t lock;
struct idr pool;
};
......@@ -45,7 +45,7 @@ struct p9_idpool *p9_idpool_create(void)
if (!p)
return ERR_PTR(-ENOMEM);
init_MUTEX(&p->lock);
spin_lock_init(&p->lock);
idr_init(&p->pool);
return p;
......@@ -71,19 +71,17 @@ int p9_idpool_get(struct p9_idpool *p)
{
int i = 0;
int error;
unsigned int flags;
retry:
if (idr_pre_get(&p->pool, GFP_KERNEL) == 0)
return 0;
if (down_interruptible(&p->lock) == -EINTR) {
P9_EPRINTK(KERN_WARNING, "Interrupted while locking\n");
return -1;
}
spin_lock_irqsave(&p->lock, flags);
/* no need to store exactly p, we just need something non-null */
error = idr_get_new(&p->pool, p, &i);
up(&p->lock);
spin_unlock_irqrestore(&p->lock, flags);
if (error == -EAGAIN)
goto retry;
......@@ -104,12 +102,10 @@ EXPORT_SYMBOL(p9_idpool_get);
void p9_idpool_put(int id, struct p9_idpool *p)
{
if (down_interruptible(&p->lock) == -EINTR) {
P9_EPRINTK(KERN_WARNING, "Interrupted while locking\n");
return;
}
unsigned int flags;
spin_lock_irqsave(&p->lock, flags);
idr_remove(&p->pool, id);
up(&p->lock);
spin_unlock_irqrestore(&p->lock, flags);
}
EXPORT_SYMBOL(p9_idpool_put);
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment