dm thin: wake worker when discard is prepared

[linux-2.6.git] / net / decnet / dn_nsp_out.c

/*
 * DECnet       An implementation of the DECnet protocol suite for the LINUX
 *              operating system.  DECnet is implemented using the  BSD Socket
 *              interface as the means of communication with the user level.
 *
 *              DECnet Network Services Protocol (Output)
 *
 * Author:      Eduardo Marcelo Serrat <emserrat@geocities.com>
 *
 * Changes:
 *
 *    Steve Whitehouse:  Split into dn_nsp_in.c and dn_nsp_out.c from
 *                       original dn_nsp.c.
 *    Steve Whitehouse:  Updated to work with my new routing architecture.
 *    Steve Whitehouse:  Added changes from Eduardo Serrat's patches.
 *    Steve Whitehouse:  Now conninits have the "return" bit set.
 *    Steve Whitehouse:  Fixes to check alloc'd skbs are non NULL!
 *                       Moved output state machine into one function
 *    Steve Whitehouse:  New output state machine
 *         Paul Koning:  Connect Confirm message fix.
 *      Eduardo Serrat:  Fix to stop dn_nsp_do_disc() sending malformed packets.
 *    Steve Whitehouse:  dn_nsp_output() and friends needed a spring clean
 *    Steve Whitehouse:  Moved dn_nsp_send() in here from route.h
 */

/******************************************************************************
    (c) 1995-1998 E.M. Serrat           emserrat@geocities.com

    This program is free software; you can redistribute it and/or modify
    it under the terms of the GNU General Public License as published by
    the Free Software Foundation; either version 2 of the License, or
    any later version.

    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.
*******************************************************************************/

#include <linux/errno.h>
#include <linux/types.h>
#include <linux/socket.h>
#include <linux/in.h>
#include <linux/kernel.h>
#include <linux/timer.h>
#include <linux/string.h>
#include <linux/sockios.h>
#include <linux/net.h>
#include <linux/netdevice.h>
#include <linux/inet.h>
#include <linux/route.h>
#include <linux/slab.h>
#include <net/sock.h>
#include <linux/fcntl.h>
#include <linux/mm.h>
#include <linux/termios.h>
#include <linux/interrupt.h>
#include <linux/proc_fs.h>
#include <linux/stat.h>
#include <linux/init.h>
#include <linux/poll.h>
#include <linux/if_packet.h>
#include <net/neighbour.h>
#include <net/dst.h>
#include <net/flow.h>
#include <net/dn.h>
#include <net/dn_nsp.h>
#include <net/dn_dev.h>
#include <net/dn_route.h>


static int nsp_backoff[NSP_MAXRXTSHIFT + 1] = { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 };

static void dn_nsp_send(struct sk_buff *skb)
{
        struct sock *sk = skb->sk;
        struct dn_scp *scp = DN_SK(sk);
        struct dst_entry *dst;
        struct flowidn fld;

        skb_reset_transport_header(skb);
        scp->stamp = jiffies;

        dst = sk_dst_check(sk, 0);
        if (dst) {
try_again:
                skb_dst_set(skb, dst);
                dst_output(skb);
                return;
        }

        memset(&fld, 0, sizeof(fld));
        fld.flowidn_oif = sk->sk_bound_dev_if;
        fld.saddr = dn_saddr2dn(&scp->addr);
        fld.daddr = dn_saddr2dn(&scp->peer);
        dn_sk_ports_copy(&fld, scp);
        fld.flowidn_proto = DNPROTO_NSP;
        if (dn_route_output_sock(&sk->sk_dst_cache, &fld, sk, 0) == 0) {
                dst = sk_dst_get(sk);
                sk->sk_route_caps = dst->dev->features;
                goto try_again;
        }

        sk->sk_err = EHOSTUNREACH;
        if (!sock_flag(sk, SOCK_DEAD))
                sk->sk_state_change(sk);
}


/*
 * If sk == NULL, then we assume that we are supposed to be making
 * a routing layer skb. If sk != NULL, then we are supposed to be
 * creating an skb for the NSP layer.
 *
 * The eventual aim is for each socket to have a cached header size
 * for its outgoing packets, and to set hdr from this when sk != NULL.
 */
struct sk_buff *dn_alloc_skb(struct sock *sk, int size, gfp_t pri)
{
        struct sk_buff *skb;
        int hdr = 64;

        if ((skb = alloc_skb(size + hdr, pri)) == NULL)
                return NULL;

        skb->protocol = htons(ETH_P_DNA_RT);
        skb->pkt_type = PACKET_OUTGOING;

        if (sk)
                skb_set_owner_w(skb, sk);

        skb_reserve(skb, hdr);

        return skb;
}

/*
 * Calculate persist timer based upon the smoothed round
 * trip time and the variance. Backoff according to the
 * nsp_backoff[] array.
 */
unsigned long dn_nsp_persist(struct sock *sk)
{
        struct dn_scp *scp = DN_SK(sk);

        unsigned long t = ((scp->nsp_srtt >> 2) + scp->nsp_rttvar) >> 1;

        t *= nsp_backoff[scp->nsp_rxtshift];

        if (t < HZ) t = HZ;
        if (t > (600*HZ)) t = (600*HZ);

        if (scp->nsp_rxtshift < NSP_MAXRXTSHIFT)
                scp->nsp_rxtshift++;

        /* printk(KERN_DEBUG "rxtshift %lu, t=%lu\n", scp->nsp_rxtshift, t); */

        return t;
}

/*
 * This is called each time we get an estimate for the rtt
 * on the link.
 */
static void dn_nsp_rtt(struct sock *sk, long rtt)
{
        struct dn_scp *scp = DN_SK(sk);
        long srtt = (long)scp->nsp_srtt;
        long rttvar = (long)scp->nsp_rttvar;
        long delta;

        /*
         * If the jiffies clock flips over in the middle of timestamp
         * gathering this value might turn out negative, so we make sure
         * that is it always positive here.
         */
        if (rtt < 0)
                rtt = -rtt;
        /*
         * Add new rtt to smoothed average
         */
        delta = ((rtt << 3) - srtt);
        srtt += (delta >> 3);
        if (srtt >= 1)
                scp->nsp_srtt = (unsigned long)srtt;
        else
                scp->nsp_srtt = 1;

        /*
         * Add new rtt varience to smoothed varience
         */
        delta >>= 1;
        rttvar += ((((delta>0)?(delta):(-delta)) - rttvar) >> 2);
        if (rttvar >= 1)
                scp->nsp_rttvar = (unsigned long)rttvar;
        else
                scp->nsp_rttvar = 1;

        /* printk(KERN_DEBUG "srtt=%lu rttvar=%lu\n", scp->nsp_srtt, scp->nsp_rttvar); */
}

/**
 * dn_nsp_clone_and_send - Send a data packet by cloning it
 * @skb: The packet to clone and transmit
 * @gfp: memory allocation flag
 *
 * Clone a queued data or other data packet and transmit it.
 *
 * Returns: The number of times the packet has been sent previously
 */
static inline unsigned int dn_nsp_clone_and_send(struct sk_buff *skb,
                                             gfp_t gfp)
{
        struct dn_skb_cb *cb = DN_SKB_CB(skb);
        struct sk_buff *skb2;
        int ret = 0;

        if ((skb2 = skb_clone(skb, gfp)) != NULL) {
                ret = cb->xmit_count;
                cb->xmit_count++;
                cb->stamp = jiffies;
                skb2->sk = skb->sk;
                dn_nsp_send(skb2);
        }

        return ret;
}

/**
 * dn_nsp_output - Try and send something from socket queues
 * @sk: The socket whose queues are to be investigated
 *
 * Try and send the packet on the end of the data and other data queues.
 * Other data gets priority over data, and if we retransmit a packet we
 * reduce the window by dividing it in two.
 *
 */
void dn_nsp_output(struct sock *sk)
{
        struct dn_scp *scp = DN_SK(sk);
        struct sk_buff *skb;
        unsigned int reduce_win = 0;

        /*
         * First we check for otherdata/linkservice messages
         */
        if ((skb = skb_peek(&scp->other_xmit_queue)) != NULL)
                reduce_win = dn_nsp_clone_and_send(skb, GFP_ATOMIC);

        /*
         * If we may not send any data, we don't.
         * If we are still trying to get some other data down the
         * channel, we don't try and send any data.
         */
        if (reduce_win || (scp->flowrem_sw != DN_SEND))
                goto recalc_window;

        if ((skb = skb_peek(&scp->data_xmit_queue)) != NULL)
                reduce_win = dn_nsp_clone_and_send(skb, GFP_ATOMIC);

        /*
         * If we've sent any frame more than once, we cut the
         * send window size in half. There is always a minimum
         * window size of one available.
         */
recalc_window:
        if (reduce_win) {
                scp->snd_window >>= 1;
                if (scp->snd_window < NSP_MIN_WINDOW)
                        scp->snd_window = NSP_MIN_WINDOW;
        }
}

int dn_nsp_xmit_timeout(struct sock *sk)
{
        struct dn_scp *scp = DN_SK(sk);

        dn_nsp_output(sk);

        if (!skb_queue_empty(&scp->data_xmit_queue) ||
            !skb_queue_empty(&scp->other_xmit_queue))
                scp->persist = dn_nsp_persist(sk);

        return 0;
}

static inline __le16 *dn_mk_common_header(struct dn_scp *scp, struct sk_buff *skb, unsigned char msgflag, int len)
{
        unsigned char *ptr = skb_push(skb, len);

        BUG_ON(len < 5);

        *ptr++ = msgflag;
        *((__le16 *)ptr) = scp->addrrem;
        ptr += 2;
        *((__le16 *)ptr) = scp->addrloc;
        ptr += 2;
        return (__le16 __force *)ptr;
}

static __le16 *dn_mk_ack_header(struct sock *sk, struct sk_buff *skb, unsigned char msgflag, int hlen, int other)
{
        struct dn_scp *scp = DN_SK(sk);
        unsigned short acknum = scp->numdat_rcv & 0x0FFF;
        unsigned short ackcrs = scp->numoth_rcv & 0x0FFF;
        __le16 *ptr;

        BUG_ON(hlen < 9);

        scp->ackxmt_dat = acknum;
        scp->ackxmt_oth = ackcrs;
        acknum |= 0x8000;
        ackcrs |= 0x8000;

        /* If this is an "other data/ack" message, swap acknum and ackcrs */
        if (other) {
                unsigned short tmp = acknum;
                acknum = ackcrs;
                ackcrs = tmp;
        }

        /* Set "cross subchannel" bit in ackcrs */
        ackcrs |= 0x2000;

        ptr = dn_mk_common_header(scp, skb, msgflag, hlen);

        *ptr++ = cpu_to_le16(acknum);
        *ptr++ = cpu_to_le16(ackcrs);

        return ptr;
}

static __le16 *dn_nsp_mk_data_header(struct sock *sk, struct sk_buff *skb, int oth)
{
        struct dn_scp *scp = DN_SK(sk);
        struct dn_skb_cb *cb = DN_SKB_CB(skb);
        __le16 *ptr = dn_mk_ack_header(sk, skb, cb->nsp_flags, 11, oth);

        if (unlikely(oth)) {
                cb->segnum = scp->numoth;
                seq_add(&scp->numoth, 1);
        } else {
                cb->segnum = scp->numdat;
                seq_add(&scp->numdat, 1);
        }
        *(ptr++) = cpu_to_le16(cb->segnum);

        return ptr;
}

void dn_nsp_queue_xmit(struct sock *sk, struct sk_buff *skb,
                        gfp_t gfp, int oth)
{
        struct dn_scp *scp = DN_SK(sk);
        struct dn_skb_cb *cb = DN_SKB_CB(skb);
        unsigned long t = ((scp->nsp_srtt >> 2) + scp->nsp_rttvar) >> 1;

        cb->xmit_count = 0;
        dn_nsp_mk_data_header(sk, skb, oth);

        /*
         * Slow start: If we have been idle for more than
         * one RTT, then reset window to min size.
         */
        if ((jiffies - scp->stamp) > t)
                scp->snd_window = NSP_MIN_WINDOW;

        if (oth)
                skb_queue_tail(&scp->other_xmit_queue, skb);
        else
                skb_queue_tail(&scp->data_xmit_queue, skb);

        if (scp->flowrem_sw != DN_SEND)
                return;

        dn_nsp_clone_and_send(skb, gfp);
}


int dn_nsp_check_xmit_queue(struct sock *sk, struct sk_buff *skb, struct sk_buff_head *q, unsigned short acknum)
{
        struct dn_skb_cb *cb = DN_SKB_CB(skb);
        struct dn_scp *scp = DN_SK(sk);
        struct sk_buff *skb2, *n, *ack = NULL;
        int wakeup = 0;
        int try_retrans = 0;
        unsigned long reftime = cb->stamp;
        unsigned long pkttime;
        unsigned short xmit_count;
        unsigned short segnum;

        skb_queue_walk_safe(q, skb2, n) {
                struct dn_skb_cb *cb2 = DN_SKB_CB(skb2);

                if (dn_before_or_equal(cb2->segnum, acknum))
                        ack = skb2;

                /* printk(KERN_DEBUG "ack: %s %04x %04x\n", ack ? "ACK" : "SKIP", (int)cb2->segnum, (int)acknum); */

                if (ack == NULL)
                        continue;

                /* printk(KERN_DEBUG "check_xmit_queue: %04x, %d\n", acknum, cb2->xmit_count); */

                /* Does _last_ packet acked have xmit_count > 1 */
                try_retrans = 0;
                /* Remember to wake up the sending process */
                wakeup = 1;
                /* Keep various statistics */
                pkttime = cb2->stamp;
                xmit_count = cb2->xmit_count;
                segnum = cb2->segnum;
                /* Remove and drop ack'ed packet */
                skb_unlink(ack, q);
                kfree_skb(ack);
                ack = NULL;

                /*
                 * We don't expect to see acknowledgements for packets we
                 * haven't sent yet.
                 */
                WARN_ON(xmit_count == 0);

                /*
                 * If the packet has only been sent once, we can use it
                 * to calculate the RTT and also open the window a little
                 * further.
                 */
                if (xmit_count == 1) {
                        if (dn_equal(segnum, acknum))
                                dn_nsp_rtt(sk, (long)(pkttime - reftime));

                        if (scp->snd_window < scp->max_window)
                                scp->snd_window++;
                }

                /*
                 * Packet has been sent more than once. If this is the last
                 * packet to be acknowledged then we want to send the next
                 * packet in the send queue again (assumes the remote host does
                 * go-back-N error control).
                 */
                if (xmit_count > 1)
                        try_retrans = 1;
        }

        if (try_retrans)
                dn_nsp_output(sk);

        return wakeup;
}

void dn_nsp_send_data_ack(struct sock *sk)
{
        struct sk_buff *skb = NULL;

        if ((skb = dn_alloc_skb(sk, 9, GFP_ATOMIC)) == NULL)
                return;

        skb_reserve(skb, 9);
        dn_mk_ack_header(sk, skb, 0x04, 9, 0);
        dn_nsp_send(skb);
}

void dn_nsp_send_oth_ack(struct sock *sk)
{
        struct sk_buff *skb = NULL;

        if ((skb = dn_alloc_skb(sk, 9, GFP_ATOMIC)) == NULL)
                return;

        skb_reserve(skb, 9);
        dn_mk_ack_header(sk, skb, 0x14, 9, 1);
        dn_nsp_send(skb);
}


void dn_send_conn_ack (struct sock *sk)
{
        struct dn_scp *scp = DN_SK(sk);
        struct sk_buff *skb = NULL;
        struct nsp_conn_ack_msg *msg;

        if ((skb = dn_alloc_skb(sk, 3, sk->sk_allocation)) == NULL)
                return;

        msg = (struct nsp_conn_ack_msg *)skb_put(skb, 3);
        msg->msgflg = 0x24;
        msg->dstaddr = scp->addrrem;

        dn_nsp_send(skb);
}

void dn_nsp_delayed_ack(struct sock *sk)
{
        struct dn_scp *scp = DN_SK(sk);

        if (scp->ackxmt_oth != scp->numoth_rcv)
                dn_nsp_send_oth_ack(sk);

        if (scp->ackxmt_dat != scp->numdat_rcv)
                dn_nsp_send_data_ack(sk);
}

static int dn_nsp_retrans_conn_conf(struct sock *sk)
{
        struct dn_scp *scp = DN_SK(sk);

        if (scp->state == DN_CC)
                dn_send_conn_conf(sk, GFP_ATOMIC);

        return 0;
}

void dn_send_conn_conf(struct sock *sk, gfp_t gfp)
{
        struct dn_scp *scp = DN_SK(sk);
        struct sk_buff *skb = NULL;
        struct nsp_conn_init_msg *msg;
        __u8 len = (__u8)le16_to_cpu(scp->conndata_out.opt_optl);

        if ((skb = dn_alloc_skb(sk, 50 + len, gfp)) == NULL)
                return;

        msg = (struct nsp_conn_init_msg *)skb_put(skb, sizeof(*msg));
        msg->msgflg = 0x28;
        msg->dstaddr = scp->addrrem;
        msg->srcaddr = scp->addrloc;
        msg->services = scp->services_loc;
        msg->info = scp->info_loc;
        msg->segsize = cpu_to_le16(scp->segsize_loc);

        *skb_put(skb,1) = len;

        if (len > 0)
                memcpy(skb_put(skb, len), scp->conndata_out.opt_data, len);


        dn_nsp_send(skb);

        scp->persist = dn_nsp_persist(sk);
        scp->persist_fxn = dn_nsp_retrans_conn_conf;
}


static __inline__ void dn_nsp_do_disc(struct sock *sk, unsigned char msgflg,
                        unsigned short reason, gfp_t gfp,
                        struct dst_entry *dst,
                        int ddl, unsigned char *dd, __le16 rem, __le16 loc)
{
        struct sk_buff *skb = NULL;
        int size = 7 + ddl + ((msgflg == NSP_DISCINIT) ? 1 : 0);
        unsigned char *msg;

        if ((dst == NULL) || (rem == 0)) {
                net_dbg_ratelimited("DECnet: dn_nsp_do_disc: BUG! Please report this to SteveW@ACM.org rem=%u dst=%p\n",
                                    le16_to_cpu(rem), dst);
                return;
        }

        if ((skb = dn_alloc_skb(sk, size, gfp)) == NULL)
                return;

        msg = skb_put(skb, size);
        *msg++ = msgflg;
        *(__le16 *)msg = rem;
        msg += 2;
        *(__le16 *)msg = loc;
        msg += 2;
        *(__le16 *)msg = cpu_to_le16(reason);
        msg += 2;
        if (msgflg == NSP_DISCINIT)
                *msg++ = ddl;

        if (ddl) {
                memcpy(msg, dd, ddl);
        }

        /*
         * This doesn't go via the dn_nsp_send() function since we need
         * to be able to send disc packets out which have no socket
         * associations.
         */
        skb_dst_set(skb, dst_clone(dst));
        dst_output(skb);
}


void dn_nsp_send_disc(struct sock *sk, unsigned char msgflg,
                        unsigned short reason, gfp_t gfp)
{
        struct dn_scp *scp = DN_SK(sk);
        int ddl = 0;

        if (msgflg == NSP_DISCINIT)
                ddl = le16_to_cpu(scp->discdata_out.opt_optl);

        if (reason == 0)
                reason = le16_to_cpu(scp->discdata_out.opt_status);

        dn_nsp_do_disc(sk, msgflg, reason, gfp, sk->sk_dst_cache, ddl,
                scp->discdata_out.opt_data, scp->addrrem, scp->addrloc);
}


void dn_nsp_return_disc(struct sk_buff *skb, unsigned char msgflg,
                        unsigned short reason)
{
        struct dn_skb_cb *cb = DN_SKB_CB(skb);
        int ddl = 0;
        gfp_t gfp = GFP_ATOMIC;

        dn_nsp_do_disc(NULL, msgflg, reason, gfp, skb_dst(skb), ddl,
                        NULL, cb->src_port, cb->dst_port);
}


void dn_nsp_send_link(struct sock *sk, unsigned char lsflags, char fcval)
{
        struct dn_scp *scp = DN_SK(sk);
        struct sk_buff *skb;
        unsigned char *ptr;
        gfp_t gfp = GFP_ATOMIC;

        if ((skb = dn_alloc_skb(sk, DN_MAX_NSP_DATA_HEADER + 2, gfp)) == NULL)
                return;

        skb_reserve(skb, DN_MAX_NSP_DATA_HEADER);
        ptr = skb_put(skb, 2);
        DN_SKB_CB(skb)->nsp_flags = 0x10;
        *ptr++ = lsflags;
        *ptr = fcval;

        dn_nsp_queue_xmit(sk, skb, gfp, 1);

        scp->persist = dn_nsp_persist(sk);
        scp->persist_fxn = dn_nsp_xmit_timeout;
}

static int dn_nsp_retrans_conninit(struct sock *sk)
{
        struct dn_scp *scp = DN_SK(sk);

        if (scp->state == DN_CI)
                dn_nsp_send_conninit(sk, NSP_RCI);

        return 0;
}

void dn_nsp_send_conninit(struct sock *sk, unsigned char msgflg)
{
        struct dn_scp *scp = DN_SK(sk);
        struct nsp_conn_init_msg *msg;
        unsigned char aux;
        unsigned char menuver;
        struct dn_skb_cb *cb;
        unsigned char type = 1;
        gfp_t allocation = (msgflg == NSP_CI) ? sk->sk_allocation : GFP_ATOMIC;
        struct sk_buff *skb = dn_alloc_skb(sk, 200, allocation);

        if (!skb)
                return;

        cb  = DN_SKB_CB(skb);
        msg = (struct nsp_conn_init_msg *)skb_put(skb,sizeof(*msg));

        msg->msgflg     = msgflg;
        msg->dstaddr    = 0x0000;               /* Remote Node will assign it*/

        msg->srcaddr    = scp->addrloc;
        msg->services   = scp->services_loc;    /* Requested flow control    */
        msg->info       = scp->info_loc;        /* Version Number            */
        msg->segsize    = cpu_to_le16(scp->segsize_loc);        /* Max segment size  */

        if (scp->peer.sdn_objnum)
                type = 0;

        skb_put(skb, dn_sockaddr2username(&scp->peer,
                                          skb_tail_pointer(skb), type));
        skb_put(skb, dn_sockaddr2username(&scp->addr,
                                          skb_tail_pointer(skb), 2));

        menuver = DN_MENUVER_ACC | DN_MENUVER_USR;
        if (scp->peer.sdn_flags & SDF_PROXY)
                menuver |= DN_MENUVER_PRX;
        if (scp->peer.sdn_flags & SDF_UICPROXY)
                menuver |= DN_MENUVER_UIC;

        *skb_put(skb, 1) = menuver;     /* Menu Version         */

        aux = scp->accessdata.acc_userl;
        *skb_put(skb, 1) = aux;
        if (aux > 0)
                memcpy(skb_put(skb, aux), scp->accessdata.acc_user, aux);

        aux = scp->accessdata.acc_passl;
        *skb_put(skb, 1) = aux;
        if (aux > 0)
                memcpy(skb_put(skb, aux), scp->accessdata.acc_pass, aux);

        aux = scp->accessdata.acc_accl;
        *skb_put(skb, 1) = aux;
        if (aux > 0)
                memcpy(skb_put(skb, aux), scp->accessdata.acc_acc, aux);

        aux = (__u8)le16_to_cpu(scp->conndata_out.opt_optl);
        *skb_put(skb, 1) = aux;
        if (aux > 0)
                memcpy(skb_put(skb, aux), scp->conndata_out.opt_data, aux);

        scp->persist = dn_nsp_persist(sk);
        scp->persist_fxn = dn_nsp_retrans_conninit;

        cb->rt_flags = DN_RT_F_RQR;

        dn_nsp_send(skb);
}