2
0
mirror of https://github.com/edk2-porting/linux-next.git synced 2024-12-21 03:33:59 +08:00
linux-next/net/sched/sch_atm.c
Eric Dumazet 520ac30f45 net_sched: drop packets after root qdisc lock is released
Qdisc performance suffers when packets are dropped at enqueue()
time because drops (kfree_skb()) are done while qdisc lock is held,
delaying a dequeue() draining the queue.

Nominal throughput can be reduced by 50 % when this happens,
at a time we would like the dequeue() to proceed as fast as possible.

Even FQ is vulnerable to this problem, while one of FQ goals was
to provide some flow isolation.

This patch adds a 'struct sk_buff **to_free' parameter to all
qdisc->enqueue(), and in qdisc_drop() helper.

I measured a performance increase of up to 12 %, but this patch
is a prereq so that future batches in enqueue() can fly.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-06-25 12:19:35 -04:00

682 lines
19 KiB
C

/* net/sched/sch_atm.c - ATM VC selection "queueing discipline" */
/* Written 1998-2000 by Werner Almesberger, EPFL ICA */
#include <linux/module.h>
#include <linux/slab.h>
#include <linux/init.h>
#include <linux/interrupt.h>
#include <linux/string.h>
#include <linux/errno.h>
#include <linux/skbuff.h>
#include <linux/atmdev.h>
#include <linux/atmclip.h>
#include <linux/rtnetlink.h>
#include <linux/file.h> /* for fput */
#include <net/netlink.h>
#include <net/pkt_sched.h>
/*
* The ATM queuing discipline provides a framework for invoking classifiers
* (aka "filters"), which in turn select classes of this queuing discipline.
* Each class maps the flow(s) it is handling to a given VC. Multiple classes
* may share the same VC.
*
* When creating a class, VCs are specified by passing the number of the open
* socket descriptor by which the calling process references the VC. The kernel
* keeps the VC open at least until all classes using it are removed.
*
* In this file, most functions are named atm_tc_* to avoid confusion with all
* the atm_* in net/atm. This naming convention differs from what's used in the
* rest of net/sched.
*
* Known bugs:
* - sometimes messes up the IP stack
* - any manipulations besides the few operations described in the README, are
* untested and likely to crash the system
* - should lock the flow while there is data in the queue (?)
*/
#define VCC2FLOW(vcc) ((struct atm_flow_data *) ((vcc)->user_back))
struct atm_flow_data {
struct Qdisc *q; /* FIFO, TBF, etc. */
struct tcf_proto __rcu *filter_list;
struct atm_vcc *vcc; /* VCC; NULL if VCC is closed */
void (*old_pop)(struct atm_vcc *vcc,
struct sk_buff *skb); /* chaining */
struct atm_qdisc_data *parent; /* parent qdisc */
struct socket *sock; /* for closing */
u32 classid; /* x:y type ID */
int ref; /* reference count */
struct gnet_stats_basic_packed bstats;
struct gnet_stats_queue qstats;
struct list_head list;
struct atm_flow_data *excess; /* flow for excess traffic;
NULL to set CLP instead */
int hdr_len;
unsigned char hdr[0]; /* header data; MUST BE LAST */
};
struct atm_qdisc_data {
struct atm_flow_data link; /* unclassified skbs go here */
struct list_head flows; /* NB: "link" is also on this
list */
struct tasklet_struct task; /* dequeue tasklet */
};
/* ------------------------- Class/flow operations ------------------------- */
static inline struct atm_flow_data *lookup_flow(struct Qdisc *sch, u32 classid)
{
struct atm_qdisc_data *p = qdisc_priv(sch);
struct atm_flow_data *flow;
list_for_each_entry(flow, &p->flows, list) {
if (flow->classid == classid)
return flow;
}
return NULL;
}
static int atm_tc_graft(struct Qdisc *sch, unsigned long arg,
struct Qdisc *new, struct Qdisc **old)
{
struct atm_qdisc_data *p = qdisc_priv(sch);
struct atm_flow_data *flow = (struct atm_flow_data *)arg;
pr_debug("atm_tc_graft(sch %p,[qdisc %p],flow %p,new %p,old %p)\n",
sch, p, flow, new, old);
if (list_empty(&flow->list))
return -EINVAL;
if (!new)
new = &noop_qdisc;
*old = flow->q;
flow->q = new;
if (*old)
qdisc_reset(*old);
return 0;
}
static struct Qdisc *atm_tc_leaf(struct Qdisc *sch, unsigned long cl)
{
struct atm_flow_data *flow = (struct atm_flow_data *)cl;
pr_debug("atm_tc_leaf(sch %p,flow %p)\n", sch, flow);
return flow ? flow->q : NULL;
}
static unsigned long atm_tc_get(struct Qdisc *sch, u32 classid)
{
struct atm_qdisc_data *p __maybe_unused = qdisc_priv(sch);
struct atm_flow_data *flow;
pr_debug("atm_tc_get(sch %p,[qdisc %p],classid %x)\n", sch, p, classid);
flow = lookup_flow(sch, classid);
if (flow)
flow->ref++;
pr_debug("atm_tc_get: flow %p\n", flow);
return (unsigned long)flow;
}
static unsigned long atm_tc_bind_filter(struct Qdisc *sch,
unsigned long parent, u32 classid)
{
return atm_tc_get(sch, classid);
}
/*
* atm_tc_put handles all destructions, including the ones that are explicitly
* requested (atm_tc_destroy, etc.). The assumption here is that we never drop
* anything that still seems to be in use.
*/
static void atm_tc_put(struct Qdisc *sch, unsigned long cl)
{
struct atm_qdisc_data *p = qdisc_priv(sch);
struct atm_flow_data *flow = (struct atm_flow_data *)cl;
pr_debug("atm_tc_put(sch %p,[qdisc %p],flow %p)\n", sch, p, flow);
if (--flow->ref)
return;
pr_debug("atm_tc_put: destroying\n");
list_del_init(&flow->list);
pr_debug("atm_tc_put: qdisc %p\n", flow->q);
qdisc_destroy(flow->q);
tcf_destroy_chain(&flow->filter_list);
if (flow->sock) {
pr_debug("atm_tc_put: f_count %ld\n",
file_count(flow->sock->file));
flow->vcc->pop = flow->old_pop;
sockfd_put(flow->sock);
}
if (flow->excess)
atm_tc_put(sch, (unsigned long)flow->excess);
if (flow != &p->link)
kfree(flow);
/*
* If flow == &p->link, the qdisc no longer works at this point and
* needs to be removed. (By the caller of atm_tc_put.)
*/
}
static void sch_atm_pop(struct atm_vcc *vcc, struct sk_buff *skb)
{
struct atm_qdisc_data *p = VCC2FLOW(vcc)->parent;
pr_debug("sch_atm_pop(vcc %p,skb %p,[qdisc %p])\n", vcc, skb, p);
VCC2FLOW(vcc)->old_pop(vcc, skb);
tasklet_schedule(&p->task);
}
static const u8 llc_oui_ip[] = {
0xaa, /* DSAP: non-ISO */
0xaa, /* SSAP: non-ISO */
0x03, /* Ctrl: Unnumbered Information Command PDU */
0x00, /* OUI: EtherType */
0x00, 0x00,
0x08, 0x00
}; /* Ethertype IP (0800) */
static const struct nla_policy atm_policy[TCA_ATM_MAX + 1] = {
[TCA_ATM_FD] = { .type = NLA_U32 },
[TCA_ATM_EXCESS] = { .type = NLA_U32 },
};
static int atm_tc_change(struct Qdisc *sch, u32 classid, u32 parent,
struct nlattr **tca, unsigned long *arg)
{
struct atm_qdisc_data *p = qdisc_priv(sch);
struct atm_flow_data *flow = (struct atm_flow_data *)*arg;
struct atm_flow_data *excess = NULL;
struct nlattr *opt = tca[TCA_OPTIONS];
struct nlattr *tb[TCA_ATM_MAX + 1];
struct socket *sock;
int fd, error, hdr_len;
void *hdr;
pr_debug("atm_tc_change(sch %p,[qdisc %p],classid %x,parent %x,"
"flow %p,opt %p)\n", sch, p, classid, parent, flow, opt);
/*
* The concept of parents doesn't apply for this qdisc.
*/
if (parent && parent != TC_H_ROOT && parent != sch->handle)
return -EINVAL;
/*
* ATM classes cannot be changed. In order to change properties of the
* ATM connection, that socket needs to be modified directly (via the
* native ATM API. In order to send a flow to a different VC, the old
* class needs to be removed and a new one added. (This may be changed
* later.)
*/
if (flow)
return -EBUSY;
if (opt == NULL)
return -EINVAL;
error = nla_parse_nested(tb, TCA_ATM_MAX, opt, atm_policy);
if (error < 0)
return error;
if (!tb[TCA_ATM_FD])
return -EINVAL;
fd = nla_get_u32(tb[TCA_ATM_FD]);
pr_debug("atm_tc_change: fd %d\n", fd);
if (tb[TCA_ATM_HDR]) {
hdr_len = nla_len(tb[TCA_ATM_HDR]);
hdr = nla_data(tb[TCA_ATM_HDR]);
} else {
hdr_len = RFC1483LLC_LEN;
hdr = NULL; /* default LLC/SNAP for IP */
}
if (!tb[TCA_ATM_EXCESS])
excess = NULL;
else {
excess = (struct atm_flow_data *)
atm_tc_get(sch, nla_get_u32(tb[TCA_ATM_EXCESS]));
if (!excess)
return -ENOENT;
}
pr_debug("atm_tc_change: type %d, payload %d, hdr_len %d\n",
opt->nla_type, nla_len(opt), hdr_len);
sock = sockfd_lookup(fd, &error);
if (!sock)
return error; /* f_count++ */
pr_debug("atm_tc_change: f_count %ld\n", file_count(sock->file));
if (sock->ops->family != PF_ATMSVC && sock->ops->family != PF_ATMPVC) {
error = -EPROTOTYPE;
goto err_out;
}
/* @@@ should check if the socket is really operational or we'll crash
on vcc->send */
if (classid) {
if (TC_H_MAJ(classid ^ sch->handle)) {
pr_debug("atm_tc_change: classid mismatch\n");
error = -EINVAL;
goto err_out;
}
} else {
int i;
unsigned long cl;
for (i = 1; i < 0x8000; i++) {
classid = TC_H_MAKE(sch->handle, 0x8000 | i);
cl = atm_tc_get(sch, classid);
if (!cl)
break;
atm_tc_put(sch, cl);
}
}
pr_debug("atm_tc_change: new id %x\n", classid);
flow = kzalloc(sizeof(struct atm_flow_data) + hdr_len, GFP_KERNEL);
pr_debug("atm_tc_change: flow %p\n", flow);
if (!flow) {
error = -ENOBUFS;
goto err_out;
}
RCU_INIT_POINTER(flow->filter_list, NULL);
flow->q = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops, classid);
if (!flow->q)
flow->q = &noop_qdisc;
pr_debug("atm_tc_change: qdisc %p\n", flow->q);
flow->sock = sock;
flow->vcc = ATM_SD(sock); /* speedup */
flow->vcc->user_back = flow;
pr_debug("atm_tc_change: vcc %p\n", flow->vcc);
flow->old_pop = flow->vcc->pop;
flow->parent = p;
flow->vcc->pop = sch_atm_pop;
flow->classid = classid;
flow->ref = 1;
flow->excess = excess;
list_add(&flow->list, &p->link.list);
flow->hdr_len = hdr_len;
if (hdr)
memcpy(flow->hdr, hdr, hdr_len);
else
memcpy(flow->hdr, llc_oui_ip, sizeof(llc_oui_ip));
*arg = (unsigned long)flow;
return 0;
err_out:
if (excess)
atm_tc_put(sch, (unsigned long)excess);
sockfd_put(sock);
return error;
}
static int atm_tc_delete(struct Qdisc *sch, unsigned long arg)
{
struct atm_qdisc_data *p = qdisc_priv(sch);
struct atm_flow_data *flow = (struct atm_flow_data *)arg;
pr_debug("atm_tc_delete(sch %p,[qdisc %p],flow %p)\n", sch, p, flow);
if (list_empty(&flow->list))
return -EINVAL;
if (rcu_access_pointer(flow->filter_list) || flow == &p->link)
return -EBUSY;
/*
* Reference count must be 2: one for "keepalive" (set at class
* creation), and one for the reference held when calling delete.
*/
if (flow->ref < 2) {
pr_err("atm_tc_delete: flow->ref == %d\n", flow->ref);
return -EINVAL;
}
if (flow->ref > 2)
return -EBUSY; /* catch references via excess, etc. */
atm_tc_put(sch, arg);
return 0;
}
static void atm_tc_walk(struct Qdisc *sch, struct qdisc_walker *walker)
{
struct atm_qdisc_data *p = qdisc_priv(sch);
struct atm_flow_data *flow;
pr_debug("atm_tc_walk(sch %p,[qdisc %p],walker %p)\n", sch, p, walker);
if (walker->stop)
return;
list_for_each_entry(flow, &p->flows, list) {
if (walker->count >= walker->skip &&
walker->fn(sch, (unsigned long)flow, walker) < 0) {
walker->stop = 1;
break;
}
walker->count++;
}
}
static struct tcf_proto __rcu **atm_tc_find_tcf(struct Qdisc *sch,
unsigned long cl)
{
struct atm_qdisc_data *p = qdisc_priv(sch);
struct atm_flow_data *flow = (struct atm_flow_data *)cl;
pr_debug("atm_tc_find_tcf(sch %p,[qdisc %p],flow %p)\n", sch, p, flow);
return flow ? &flow->filter_list : &p->link.filter_list;
}
/* --------------------------- Qdisc operations ---------------------------- */
static int atm_tc_enqueue(struct sk_buff *skb, struct Qdisc *sch,
struct sk_buff **to_free)
{
struct atm_qdisc_data *p = qdisc_priv(sch);
struct atm_flow_data *flow;
struct tcf_result res;
int result;
int ret = NET_XMIT_SUCCESS | __NET_XMIT_BYPASS;
pr_debug("atm_tc_enqueue(skb %p,sch %p,[qdisc %p])\n", skb, sch, p);
result = TC_ACT_OK; /* be nice to gcc */
flow = NULL;
if (TC_H_MAJ(skb->priority) != sch->handle ||
!(flow = (struct atm_flow_data *)atm_tc_get(sch, skb->priority))) {
struct tcf_proto *fl;
list_for_each_entry(flow, &p->flows, list) {
fl = rcu_dereference_bh(flow->filter_list);
if (fl) {
result = tc_classify(skb, fl, &res, true);
if (result < 0)
continue;
flow = (struct atm_flow_data *)res.class;
if (!flow)
flow = lookup_flow(sch, res.classid);
goto done;
}
}
flow = NULL;
done:
;
}
if (!flow) {
flow = &p->link;
} else {
if (flow->vcc)
ATM_SKB(skb)->atm_options = flow->vcc->atm_options;
/*@@@ looks good ... but it's not supposed to work :-) */
#ifdef CONFIG_NET_CLS_ACT
switch (result) {
case TC_ACT_QUEUED:
case TC_ACT_STOLEN:
__qdisc_drop(skb, to_free);
return NET_XMIT_SUCCESS | __NET_XMIT_STOLEN;
case TC_ACT_SHOT:
__qdisc_drop(skb, to_free);
goto drop;
case TC_ACT_RECLASSIFY:
if (flow->excess)
flow = flow->excess;
else
ATM_SKB(skb)->atm_options |= ATM_ATMOPT_CLP;
break;
}
#endif
}
ret = qdisc_enqueue(skb, flow->q, to_free);
if (ret != NET_XMIT_SUCCESS) {
drop: __maybe_unused
if (net_xmit_drop_count(ret)) {
qdisc_qstats_drop(sch);
if (flow)
flow->qstats.drops++;
}
return ret;
}
/*
* Okay, this may seem weird. We pretend we've dropped the packet if
* it goes via ATM. The reason for this is that the outer qdisc
* expects to be able to q->dequeue the packet later on if we return
* success at this place. Also, sch->q.qdisc needs to reflect whether
* there is a packet egligible for dequeuing or not. Note that the
* statistics of the outer qdisc are necessarily wrong because of all
* this. There's currently no correct solution for this.
*/
if (flow == &p->link) {
sch->q.qlen++;
return NET_XMIT_SUCCESS;
}
tasklet_schedule(&p->task);
return NET_XMIT_SUCCESS | __NET_XMIT_BYPASS;
}
/*
* Dequeue packets and send them over ATM. Note that we quite deliberately
* avoid checking net_device's flow control here, simply because sch_atm
* uses its own channels, which have nothing to do with any CLIP/LANE/or
* non-ATM interfaces.
*/
static void sch_atm_dequeue(unsigned long data)
{
struct Qdisc *sch = (struct Qdisc *)data;
struct atm_qdisc_data *p = qdisc_priv(sch);
struct atm_flow_data *flow;
struct sk_buff *skb;
pr_debug("sch_atm_dequeue(sch %p,[qdisc %p])\n", sch, p);
list_for_each_entry(flow, &p->flows, list) {
if (flow == &p->link)
continue;
/*
* If traffic is properly shaped, this won't generate nasty
* little bursts. Otherwise, it may ... (but that's okay)
*/
while ((skb = flow->q->ops->peek(flow->q))) {
if (!atm_may_send(flow->vcc, skb->truesize))
break;
skb = qdisc_dequeue_peeked(flow->q);
if (unlikely(!skb))
break;
qdisc_bstats_update(sch, skb);
bstats_update(&flow->bstats, skb);
pr_debug("atm_tc_dequeue: sending on class %p\n", flow);
/* remove any LL header somebody else has attached */
skb_pull(skb, skb_network_offset(skb));
if (skb_headroom(skb) < flow->hdr_len) {
struct sk_buff *new;
new = skb_realloc_headroom(skb, flow->hdr_len);
dev_kfree_skb(skb);
if (!new)
continue;
skb = new;
}
pr_debug("sch_atm_dequeue: ip %p, data %p\n",
skb_network_header(skb), skb->data);
ATM_SKB(skb)->vcc = flow->vcc;
memcpy(skb_push(skb, flow->hdr_len), flow->hdr,
flow->hdr_len);
atomic_add(skb->truesize,
&sk_atm(flow->vcc)->sk_wmem_alloc);
/* atm.atm_options are already set by atm_tc_enqueue */
flow->vcc->send(flow->vcc, skb);
}
}
}
static struct sk_buff *atm_tc_dequeue(struct Qdisc *sch)
{
struct atm_qdisc_data *p = qdisc_priv(sch);
struct sk_buff *skb;
pr_debug("atm_tc_dequeue(sch %p,[qdisc %p])\n", sch, p);
tasklet_schedule(&p->task);
skb = qdisc_dequeue_peeked(p->link.q);
if (skb)
sch->q.qlen--;
return skb;
}
static struct sk_buff *atm_tc_peek(struct Qdisc *sch)
{
struct atm_qdisc_data *p = qdisc_priv(sch);
pr_debug("atm_tc_peek(sch %p,[qdisc %p])\n", sch, p);
return p->link.q->ops->peek(p->link.q);
}
static int atm_tc_init(struct Qdisc *sch, struct nlattr *opt)
{
struct atm_qdisc_data *p = qdisc_priv(sch);
pr_debug("atm_tc_init(sch %p,[qdisc %p],opt %p)\n", sch, p, opt);
INIT_LIST_HEAD(&p->flows);
INIT_LIST_HEAD(&p->link.list);
list_add(&p->link.list, &p->flows);
p->link.q = qdisc_create_dflt(sch->dev_queue,
&pfifo_qdisc_ops, sch->handle);
if (!p->link.q)
p->link.q = &noop_qdisc;
pr_debug("atm_tc_init: link (%p) qdisc %p\n", &p->link, p->link.q);
RCU_INIT_POINTER(p->link.filter_list, NULL);
p->link.vcc = NULL;
p->link.sock = NULL;
p->link.classid = sch->handle;
p->link.ref = 1;
tasklet_init(&p->task, sch_atm_dequeue, (unsigned long)sch);
return 0;
}
static void atm_tc_reset(struct Qdisc *sch)
{
struct atm_qdisc_data *p = qdisc_priv(sch);
struct atm_flow_data *flow;
pr_debug("atm_tc_reset(sch %p,[qdisc %p])\n", sch, p);
list_for_each_entry(flow, &p->flows, list)
qdisc_reset(flow->q);
sch->q.qlen = 0;
}
static void atm_tc_destroy(struct Qdisc *sch)
{
struct atm_qdisc_data *p = qdisc_priv(sch);
struct atm_flow_data *flow, *tmp;
pr_debug("atm_tc_destroy(sch %p,[qdisc %p])\n", sch, p);
list_for_each_entry(flow, &p->flows, list)
tcf_destroy_chain(&flow->filter_list);
list_for_each_entry_safe(flow, tmp, &p->flows, list) {
if (flow->ref > 1)
pr_err("atm_destroy: %p->ref = %d\n", flow, flow->ref);
atm_tc_put(sch, (unsigned long)flow);
}
tasklet_kill(&p->task);
}
static int atm_tc_dump_class(struct Qdisc *sch, unsigned long cl,
struct sk_buff *skb, struct tcmsg *tcm)
{
struct atm_qdisc_data *p = qdisc_priv(sch);
struct atm_flow_data *flow = (struct atm_flow_data *)cl;
struct nlattr *nest;
pr_debug("atm_tc_dump_class(sch %p,[qdisc %p],flow %p,skb %p,tcm %p)\n",
sch, p, flow, skb, tcm);
if (list_empty(&flow->list))
return -EINVAL;
tcm->tcm_handle = flow->classid;
tcm->tcm_info = flow->q->handle;
nest = nla_nest_start(skb, TCA_OPTIONS);
if (nest == NULL)
goto nla_put_failure;
if (nla_put(skb, TCA_ATM_HDR, flow->hdr_len, flow->hdr))
goto nla_put_failure;
if (flow->vcc) {
struct sockaddr_atmpvc pvc;
int state;
memset(&pvc, 0, sizeof(pvc));
pvc.sap_family = AF_ATMPVC;
pvc.sap_addr.itf = flow->vcc->dev ? flow->vcc->dev->number : -1;
pvc.sap_addr.vpi = flow->vcc->vpi;
pvc.sap_addr.vci = flow->vcc->vci;
if (nla_put(skb, TCA_ATM_ADDR, sizeof(pvc), &pvc))
goto nla_put_failure;
state = ATM_VF2VS(flow->vcc->flags);
if (nla_put_u32(skb, TCA_ATM_STATE, state))
goto nla_put_failure;
}
if (flow->excess) {
if (nla_put_u32(skb, TCA_ATM_EXCESS, flow->classid))
goto nla_put_failure;
} else {
if (nla_put_u32(skb, TCA_ATM_EXCESS, 0))
goto nla_put_failure;
}
return nla_nest_end(skb, nest);
nla_put_failure:
nla_nest_cancel(skb, nest);
return -1;
}
static int
atm_tc_dump_class_stats(struct Qdisc *sch, unsigned long arg,
struct gnet_dump *d)
{
struct atm_flow_data *flow = (struct atm_flow_data *)arg;
if (gnet_stats_copy_basic(qdisc_root_sleeping_running(sch),
d, NULL, &flow->bstats) < 0 ||
gnet_stats_copy_queue(d, NULL, &flow->qstats, flow->q->q.qlen) < 0)
return -1;
return 0;
}
static int atm_tc_dump(struct Qdisc *sch, struct sk_buff *skb)
{
return 0;
}
static const struct Qdisc_class_ops atm_class_ops = {
.graft = atm_tc_graft,
.leaf = atm_tc_leaf,
.get = atm_tc_get,
.put = atm_tc_put,
.change = atm_tc_change,
.delete = atm_tc_delete,
.walk = atm_tc_walk,
.tcf_chain = atm_tc_find_tcf,
.bind_tcf = atm_tc_bind_filter,
.unbind_tcf = atm_tc_put,
.dump = atm_tc_dump_class,
.dump_stats = atm_tc_dump_class_stats,
};
static struct Qdisc_ops atm_qdisc_ops __read_mostly = {
.cl_ops = &atm_class_ops,
.id = "atm",
.priv_size = sizeof(struct atm_qdisc_data),
.enqueue = atm_tc_enqueue,
.dequeue = atm_tc_dequeue,
.peek = atm_tc_peek,
.init = atm_tc_init,
.reset = atm_tc_reset,
.destroy = atm_tc_destroy,
.dump = atm_tc_dump,
.owner = THIS_MODULE,
};
static int __init atm_init(void)
{
return register_qdisc(&atm_qdisc_ops);
}
static void __exit atm_exit(void)
{
unregister_qdisc(&atm_qdisc_ops);
}
module_init(atm_init)
module_exit(atm_exit)
MODULE_LICENSE("GPL");