/*
* This file is part of the Distributed Network Block Device 3
*
* Copyright(c) 2019 Frederic Robra <frederic@robra.org>
* Parts copyright 2011-2012 Johann Latocha <johann@latocha.de>
*
* This file may be licensed under the terms of of the
* GNU General Public License Version 2 (the ``GPL'').
*
* Software distributed under the License is distributed
* on an ``AS IS'' basis, WITHOUT WARRANTY OF ANY KIND, either
* express or implied. See the GPL for the specific language
* governing rights and limitations.
*
* You should have received a copy of the GPL along with this
* program. If not, go to http://www.gnu.org/licenses/gpl.html
* or write to the Free Software Foundation, Inc.,
* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
*
*/
#ifndef DNBD_H_
#define DNBD_H_
#include <linux/blk-mq.h>
#include <linux/types.h>
#include <linux/refcount.h>
#include <linux/blkdev.h>
#define KERNEL_MODULE
#include "types.h"
#include "serialize.h"
/**
* the number of parallel connections
*/
#define NUMBER_CONNECTIONS 4
#define RTT_FACTOR(rtt) (((rtt) * 3) / 2)
/**
* limit to which the other connected servers are only allowed to be that worser
* then the best rtt
*/
#define RTT_THRESOULD_LIMIT(best_rtt) ((best_rtt) * 10)
/**
* turn on/off debug information (1/0)
*/
#define DEBUG 1
/**
* turn on/off file and line information (1/0)
*/
#define DEBUG_FILE 1
//TODO test with ip tables regel. jedes x paket an port 58007 zerstoeren
// iptables -A OUTPUT -p tcp -m statistic --mode random --probability 0.1 -j REJECT --reject-with tcp-reset
//TODO setup mit 2 vms eine langsam eine schnell tc (traffic control)
// -> geschwindigkkeit sollte in Richtung schnellen VM gehen mit 2 Queues
extern struct workqueue_struct *dnbd3_wq;
/**
* struct dnbd3_server - defining a server for dnbd3
* @host: host of this server
* @rtts: last four round trip time measurements in microseconds
* @protocol_version: dnbd3 protocol version of this server
* @failures: how many times the server was unreachable
*/
struct dnbd3_server {
dnbd3_host_t host;
uint64_t rtts[4];
uint64_t avg_rtt;
uint16_t protocol_version;
uint16_t failures; // TODO failures runterzaehlen wenn wieder verbindung moeglich, seltener messen bei hohem failure count
/*
* TODO would this help?
* uint8_t failures[4]
* uint8_t avg_failures;
*
* in timer, calculate each x iterations average over failure array
* then move elements in array to left
* for (i = 3; i > 0; i--) {
* server->failures[i] = server->failures[i - 1]
* }
* server->failures[0] = 0;
* failures always go to failures[0] / *failures
*
* - is this to much effort?
* - what are the benefits?
* - could increase the rtt e.g. (rtt = rtt * failures)
*/
};
/**
* struct dnbd3_sock - defining a socket for dnbd3
* @sock_nr: nr of this socket
* @device: the dnbd3_device this socket belongs to
* @server: the server this socket is connected to, 'NULL' if not connected
* @tx_lock: mutex to lock when sending
* @sock: the socket, 'NULL' if not connected
* @panic: 'true' if it is not possible to send or receive
* @cookie: is incremented for every send, used to find the mq request in the
* receiver
* @keepalive_worker: worker to send a keepalive package
* @receive_worker: worker to handle the incoming packages
* @pending: the pending request which is going to be send
*/
struct dnbd3_sock {
uint8_t sock_nr;
struct dnbd3_device *device;
struct dnbd3_server *server;
struct mutex tx_lock;
struct socket *sock;
bool panic;
uint32_t cookie;
struct work_struct keepalive_worker;
struct work_struct receive_worker;
struct request *pending;
};
/**
* struct dnbd3_device - defining a dnbd3 device
* @minor: minor number of this device
* @tag_set: the blk mq tag set
* @list: list of previous and next dnbd3_device
* @disk: the gendisk
* @kobj: the kobject
* @device_lock: mutex to lock when device changes
* @socks: array of dnbd3_sock to connect to
* @imgname: the connected image name
* @initial_server: the server which was configured with ioctl, will not be
* overriden
* @alt_servers: array of alternative servers
* @new_servers_num: number of new alternative servers that are waiting to be
* copied to above array
* @new_servers: pending new alternative servers
* @update_available: 'true' if the rid has changes
* @use_server_provided_alts: 'true' if the alt_servers array is upated by the
* alternatives provided by the server
* @rid: the revision ID? TODO
* @reported_size: the size of the image
* @panic_worker: worker to handle panics and to connect if all connections are
* down
* @discovery_worker: worker to update the alt_servers, make rtt meassurement
* and reconnect to better servers
* @discovery_count: counter for the discovery worker
* @timer: timer to start the appropriate workers
* @timer_count: counter for the timer
*/
struct dnbd3_device {
int minor;
struct blk_mq_tag_set tag_set;
struct list_head list;
struct gendisk *disk;
struct kobject kobj;
struct mutex device_lock;
struct dnbd3_sock socks[NUMBER_CONNECTIONS];
char *imgname;
struct dnbd3_server initial_server;
struct dnbd3_server alt_servers[NUMBER_SERVERS];//TODO just add server to list do not remove
int new_servers_num;
dnbd3_server_entry_t new_servers[NUMBER_SERVERS];
bool update_available;
bool use_server_provided_alts;
uint16_t rid;
uint64_t reported_size;
struct work_struct panic_worker;
struct work_struct discovery_worker;
uint8_t discovery_count;
struct timer_list timer;
uint8_t timer_count;
};
/**
* struct dnbd3_cmd - defines a mq cmd for dnbd3
* @dnbd3: the dnbd3 device this command belongs to
* @lock: mutex to lock the cmd
* @cookie: the cookie of the command, this is set by the socket when sending
* @status: the status of the command
* @requed: 'true' if the command is requed
*/
struct dnbd3_cmd {
//TODO do we want the socket here (index)?
struct dnbd3_device *dnbd3;
struct mutex lock;
uint32_t cookie;
blk_status_t status;
bool requed;
};
#define dnbd3_avg_rtt(server) \
(( (server)->rtts[0] + (server)->rtts[1] \
+ (server)->rtts[2] + (server)->rtts[3] ) / 4 )
#define dnbd3_set_rtt_unreachable(server) \
(server)->rtts[0] = (server)->rtts[1] = (server)->rtts[2] \
= (server)->rtts[3] = (server)->avg_rtt = \
RTT_UNREACHABLE;
#define dnbd3_set_rtt_unknown(server) \
(server)->rtts[0] = (server)->rtts[1] = (server)->rtts[2] \
= (server)->rtts[3] = (server)->avg_rtt = \
RTT_UNREACHABLE / 2;
/**
* macros for logging
* levels:
* - debug
* - info
* - warn
* - error
* methods:
* <level>(fmt,...)
* print fmt
* <level>_dev(dev, fmt,...)
* print fmt, adds device information to the log
* <level>_sock(sock, fmt,...)
* print fmt, adds sock and device information to log
* <level>_server(dev, server, fmt,...)
* print fmt, adds device and appends server information to the log
* DEBUG - 1/0 switch all debug messages on
* DEBUG_FILE - 1/0 switch on file and line number information
*/
#if IS_ENABLED(DEBUG_FILE)
#define __print(level, fmt,...) \
printk(level "%s:%d " fmt "\n", __FILE__, __LINE__, ## __VA_ARGS__)
#else
#define __print(level, fmt,...) \
printk(level fmt "\n", ## __VA_ARGS__)
#endif
#define __print_sock(level, sock, fmt, ...) \
do { \
if ((sock)->server->host.type == HOST_IP4) { \
__print( \
level, \
"dnbd%d/%d %pI4:%d: " fmt, \
(sock)->device->minor, \
(sock)->sock_nr, \
(sock)->server->host.addr, \
(sock)->server->host.port, \
## __VA_ARGS__); \
} else { \
__print( \
level, \
"dnbd%d/%d %pI6:%d: " fmt, \
(sock)->device->minor, \
(sock)->sock_nr, \
(sock)->server->host.addr, \
(sock)->server->host.port, \
## __VA_ARGS__); \
} \
} while (0)
#define __print_server(level, dev, server, fmt, ...) \
do { \
if ((server)->host.type == HOST_IP4) { \
__print( \
level, \
"dnbd%d: " fmt " %pI4:%d", \
(dev)->minor, \
## __VA_ARGS__, \
(server)->host.addr, \
(server)->host.port); \
} else { \
__print( \
level, \
"dnbd%d: " fmt " %pI6:%d", \
(dev)->minor, \
## __VA_ARGS__, \
(server)->host.addr, \
(server)->host.port); \
} \
} while (0)
#if IS_ENABLED(DEBUG)
#define debug(fmt, ...) \
__print(KERN_DEBUG, "dnbd: " fmt, ## __VA_ARGS__)
#define debug_dev(dev, fmt, ...) \
__print(KERN_DEBUG, "dnbd%d: " fmt, (dev)->minor, ## __VA_ARGS__)
#define debug_sock(sock, fmt, ...) \
__print_sock(KERN_DEBUG, sock, fmt, ## __VA_ARGS__)
#define debug_server(dev, server, fmt, ...) \
__print_server(KERN_DEBUG, dev, server, fmt, ## __VA_ARGS__)
#else
#define debug(fmt, ...)
#define debug_dev(dev, fmt, ...)
#define debug_sock(sock, fmt, ...)
#define debug_server(dev, server, fmt, ...)
#endif
#define info(fmt, ...) \
__print(KERN_INFO, "dnbd: " fmt, ## __VA_ARGS__)
#define info_dev(dev, fmt, ...) \
__print(KERN_INFO, "dnbd%d: " fmt, (dev)->minor, ## __VA_ARGS__)
#define info_sock(sock, fmt, ...) \
__print_sock(KERN_INFO, sock, fmt, ## __VA_ARGS__)
#define info_server(dev, server, fmt, ...) \
__print_server(KERN_INFO, dev, server, fmt, ## __VA_ARGS__)
#define warn(fmt, ...) \
__print(KERN_WARNING, "dnbd: " fmt, ## __VA_ARGS__)
#define warn_dev(dev, fmt, ...) \
__print(KERN_WARNING, "dnbd%d: " fmt, (dev)->minor, ## __VA_ARGS__)
#define warn_sock(sock, fmt, ...) \
__print_sock(KERN_WARNING, sock, fmt, ## __VA_ARGS__)
#define warn_server(dev, server, fmt, ...) \
__print_server(KERN_WARNING, dev, server, fmt, ## __VA_ARGS__)
#define error(fmt, ...) \
__print(KERN_ERR, "dnbd: " fmt, ## __VA_ARGS__)
#define error_dev(dev, fmt, ...) \
__print(KERN_ERR, "dnbd%d: " fmt, (dev)->minor, ## __VA_ARGS__)
#define error_sock(sock, fmt, ...) \
__print_sock(KERN_ERR, sock, fmt, ## __VA_ARGS__)
#define error_server(dev, server, fmt, ...) \
__print_server(KERN_ERR, dev, server, fmt, ## __VA_ARGS__)
#endif /* DNBD_H_ */