mirror of
https://github.com/torvalds/linux.git
synced 2024-11-29 23:51:37 +00:00
102b55ee92
Currently qdisc_run() checks the STATE_DEACTIVATED of lockless
qdisc before calling __qdisc_run(), which ultimately clear the
STATE_MISSED when all the skb is dequeued. If STATE_DEACTIVATED
is set before clearing STATE_MISSED, there may be rescheduling
of net_tx_action() at the end of qdisc_run_end(), see below:
CPU0(net_tx_atcion) CPU1(__dev_xmit_skb) CPU2(dev_deactivate)
. . .
. set STATE_MISSED .
. __netif_schedule() .
. . set STATE_DEACTIVATED
. . qdisc_reset()
. . .
.<--------------- . synchronize_net()
clear __QDISC_STATE_SCHED | . .
. | . .
. | . some_qdisc_is_busy()
. | . return *false*
. | . .
test STATE_DEACTIVATED | . .
__qdisc_run() *not* called | . .
. | . .
test STATE_MISS | . .
__netif_schedule()--------| . .
. . .
. . .
__qdisc_run() is not called by net_tx_atcion() in CPU0 because
CPU2 has set STATE_DEACTIVATED flag during dev_deactivate(), and
STATE_MISSED is only cleared in __qdisc_run(), __netif_schedule
is called at the end of qdisc_run_end(), causing tx action
rescheduling problem.
qdisc_run() called by net_tx_action() runs in the softirq context,
which should has the same semantic as the qdisc_run() called by
__dev_xmit_skb() protected by rcu_read_lock_bh(). And there is a
synchronize_net() between STATE_DEACTIVATED flag being set and
qdisc_reset()/some_qdisc_is_busy in dev_deactivate(), we can safely
bail out for the deactived lockless qdisc in net_tx_action(), and
qdisc_reset() will reset all skb not dequeued yet.
So add the rcu_read_lock() explicitly to protect the qdisc_run()
and do the STATE_DEACTIVATED checking in net_tx_action() before
calling qdisc_run_begin(). Another option is to do the checking in
the qdisc_run_end(), but it will add unnecessary overhead for
non-tx_action case, because __dev_queue_xmit() will not see qdisc
with STATE_DEACTIVATED after synchronize_net(), the qdisc with
STATE_DEACTIVATED can only be seen by net_tx_action() because of
__netif_schedule().
The STATE_DEACTIVATED checking in qdisc_run() is to avoid race
between net_tx_action() and qdisc_reset(), see:
commit d518d2ed86 ("net/sched: fix race between deactivation
and dequeue for NOLOCK qdisc"). As the bailout added above for
deactived lockless qdisc in net_tx_action() provides better
protection for the race without calling qdisc_run() at all, so
remove the STATE_DEACTIVATED checking in qdisc_run().
After qdisc_reset(), there is no skb in qdisc to be dequeued, so
clear the STATE_MISSED in dev_reset_queue() too.
Fixes: 6b3ba9146f ("net: sched: allow qdiscs to handle locking")
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
V8: Clearing STATE_MISSED before calling __netif_schedule() has
avoid the endless rescheduling problem, but there may still
be a unnecessary rescheduling, so adjust the commit log.
Signed-off-by: David S. Miller <davem@davemloft.net>
196 lines
5.0 KiB
C
196 lines
5.0 KiB
C
/* SPDX-License-Identifier: GPL-2.0 */
|
|
#ifndef __NET_PKT_SCHED_H
|
|
#define __NET_PKT_SCHED_H
|
|
|
|
#include <linux/jiffies.h>
|
|
#include <linux/ktime.h>
|
|
#include <linux/if_vlan.h>
|
|
#include <linux/netdevice.h>
|
|
#include <net/sch_generic.h>
|
|
#include <net/net_namespace.h>
|
|
#include <uapi/linux/pkt_sched.h>
|
|
|
|
#define DEFAULT_TX_QUEUE_LEN 1000
|
|
|
|
struct qdisc_walker {
|
|
int stop;
|
|
int skip;
|
|
int count;
|
|
int (*fn)(struct Qdisc *, unsigned long cl, struct qdisc_walker *);
|
|
};
|
|
|
|
static inline void *qdisc_priv(struct Qdisc *q)
|
|
{
|
|
return &q->privdata;
|
|
}
|
|
|
|
static inline struct Qdisc *qdisc_from_priv(void *priv)
|
|
{
|
|
return container_of(priv, struct Qdisc, privdata);
|
|
}
|
|
|
|
/*
|
|
Timer resolution MUST BE < 10% of min_schedulable_packet_size/bandwidth
|
|
|
|
Normal IP packet size ~ 512byte, hence:
|
|
|
|
0.5Kbyte/1Mbyte/sec = 0.5msec, so that we need 50usec timer for
|
|
10Mbit ethernet.
|
|
|
|
10msec resolution -> <50Kbit/sec.
|
|
|
|
The result: [34]86 is not good choice for QoS router :-(
|
|
|
|
The things are not so bad, because we may use artificial
|
|
clock evaluated by integration of network data flow
|
|
in the most critical places.
|
|
*/
|
|
|
|
typedef u64 psched_time_t;
|
|
typedef long psched_tdiff_t;
|
|
|
|
/* Avoid doing 64 bit divide */
|
|
#define PSCHED_SHIFT 6
|
|
#define PSCHED_TICKS2NS(x) ((s64)(x) << PSCHED_SHIFT)
|
|
#define PSCHED_NS2TICKS(x) ((x) >> PSCHED_SHIFT)
|
|
|
|
#define PSCHED_TICKS_PER_SEC PSCHED_NS2TICKS(NSEC_PER_SEC)
|
|
#define PSCHED_PASTPERFECT 0
|
|
|
|
static inline psched_time_t psched_get_time(void)
|
|
{
|
|
return PSCHED_NS2TICKS(ktime_get_ns());
|
|
}
|
|
|
|
static inline psched_tdiff_t
|
|
psched_tdiff_bounded(psched_time_t tv1, psched_time_t tv2, psched_time_t bound)
|
|
{
|
|
return min(tv1 - tv2, bound);
|
|
}
|
|
|
|
struct qdisc_watchdog {
|
|
u64 last_expires;
|
|
struct hrtimer timer;
|
|
struct Qdisc *qdisc;
|
|
};
|
|
|
|
void qdisc_watchdog_init_clockid(struct qdisc_watchdog *wd, struct Qdisc *qdisc,
|
|
clockid_t clockid);
|
|
void qdisc_watchdog_init(struct qdisc_watchdog *wd, struct Qdisc *qdisc);
|
|
|
|
void qdisc_watchdog_schedule_range_ns(struct qdisc_watchdog *wd, u64 expires,
|
|
u64 delta_ns);
|
|
|
|
static inline void qdisc_watchdog_schedule_ns(struct qdisc_watchdog *wd,
|
|
u64 expires)
|
|
{
|
|
return qdisc_watchdog_schedule_range_ns(wd, expires, 0ULL);
|
|
}
|
|
|
|
static inline void qdisc_watchdog_schedule(struct qdisc_watchdog *wd,
|
|
psched_time_t expires)
|
|
{
|
|
qdisc_watchdog_schedule_ns(wd, PSCHED_TICKS2NS(expires));
|
|
}
|
|
|
|
void qdisc_watchdog_cancel(struct qdisc_watchdog *wd);
|
|
|
|
extern struct Qdisc_ops pfifo_qdisc_ops;
|
|
extern struct Qdisc_ops bfifo_qdisc_ops;
|
|
extern struct Qdisc_ops pfifo_head_drop_qdisc_ops;
|
|
|
|
int fifo_set_limit(struct Qdisc *q, unsigned int limit);
|
|
struct Qdisc *fifo_create_dflt(struct Qdisc *sch, struct Qdisc_ops *ops,
|
|
unsigned int limit,
|
|
struct netlink_ext_ack *extack);
|
|
|
|
int register_qdisc(struct Qdisc_ops *qops);
|
|
int unregister_qdisc(struct Qdisc_ops *qops);
|
|
void qdisc_get_default(char *id, size_t len);
|
|
int qdisc_set_default(const char *id);
|
|
|
|
void qdisc_hash_add(struct Qdisc *q, bool invisible);
|
|
void qdisc_hash_del(struct Qdisc *q);
|
|
struct Qdisc *qdisc_lookup(struct net_device *dev, u32 handle);
|
|
struct Qdisc *qdisc_lookup_rcu(struct net_device *dev, u32 handle);
|
|
struct qdisc_rate_table *qdisc_get_rtab(struct tc_ratespec *r,
|
|
struct nlattr *tab,
|
|
struct netlink_ext_ack *extack);
|
|
void qdisc_put_rtab(struct qdisc_rate_table *tab);
|
|
void qdisc_put_stab(struct qdisc_size_table *tab);
|
|
void qdisc_warn_nonwc(const char *txt, struct Qdisc *qdisc);
|
|
bool sch_direct_xmit(struct sk_buff *skb, struct Qdisc *q,
|
|
struct net_device *dev, struct netdev_queue *txq,
|
|
spinlock_t *root_lock, bool validate);
|
|
|
|
void __qdisc_run(struct Qdisc *q);
|
|
|
|
static inline void qdisc_run(struct Qdisc *q)
|
|
{
|
|
if (qdisc_run_begin(q)) {
|
|
__qdisc_run(q);
|
|
qdisc_run_end(q);
|
|
}
|
|
}
|
|
|
|
/* Calculate maximal size of packet seen by hard_start_xmit
|
|
routine of this device.
|
|
*/
|
|
static inline unsigned int psched_mtu(const struct net_device *dev)
|
|
{
|
|
return dev->mtu + dev->hard_header_len;
|
|
}
|
|
|
|
static inline struct net *qdisc_net(struct Qdisc *q)
|
|
{
|
|
return dev_net(q->dev_queue->dev);
|
|
}
|
|
|
|
struct tc_cbs_qopt_offload {
|
|
u8 enable;
|
|
s32 queue;
|
|
s32 hicredit;
|
|
s32 locredit;
|
|
s32 idleslope;
|
|
s32 sendslope;
|
|
};
|
|
|
|
struct tc_etf_qopt_offload {
|
|
u8 enable;
|
|
s32 queue;
|
|
};
|
|
|
|
struct tc_taprio_sched_entry {
|
|
u8 command; /* TC_TAPRIO_CMD_* */
|
|
|
|
/* The gate_mask in the offloading side refers to traffic classes */
|
|
u32 gate_mask;
|
|
u32 interval;
|
|
};
|
|
|
|
struct tc_taprio_qopt_offload {
|
|
u8 enable;
|
|
ktime_t base_time;
|
|
u64 cycle_time;
|
|
u64 cycle_time_extension;
|
|
|
|
size_t num_entries;
|
|
struct tc_taprio_sched_entry entries[];
|
|
};
|
|
|
|
/* Reference counting */
|
|
struct tc_taprio_qopt_offload *taprio_offload_get(struct tc_taprio_qopt_offload
|
|
*offload);
|
|
void taprio_offload_free(struct tc_taprio_qopt_offload *offload);
|
|
|
|
/* Ensure skb_mstamp_ns, which might have been populated with the txtime, is
|
|
* not mistaken for a software timestamp, because this will otherwise prevent
|
|
* the dispatch of hardware timestamps to the socket.
|
|
*/
|
|
static inline void skb_txtime_consumed(struct sk_buff *skb)
|
|
{
|
|
skb->tstamp = ktime_set(0, 0);
|
|
}
|
|
|
|
#endif
|