openkylin
/
linux
mirror of https://gitee.com/openkylin/linux.git
9
0
Fork 0
Code Issues Projects Releases Wiki Activity
linux/net/sched/sch_etf.c

485 lines
11 KiB
C
Raw Normal View History

net/sched: Introduce the ETF Qdisc The ETF (Earliest TxTime First) qdisc uses the information added earlier in this series (the socket option SO_TXTIME and the new role of sk_buff->tstamp) to schedule packets transmission based on absolute time. For some workloads, just bandwidth enforcement is not enough, and precise control of the transmission of packets is necessary. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf delta 100000 \ clockid CLOCK_TAI In this example, the Qdisc will provide SW best-effort for the control of the transmission time to the network adapter, the time stamp in the socket will be in reference to the clockid CLOCK_TAI and packets will leave the qdisc "delta" (100000) nanoseconds before its transmission time. The ETF qdisc will buffer packets sorted by their txtime. It will drop packets on enqueue() if their skbuff clockid does not match the clock reference of the Qdisc. Moreover, on dequeue(), a packet will be dropped if it expires while being enqueued. The qdisc also supports the SO_TXTIME deadline mode. For this mode, it will dequeue a packet as soon as possible and change the skb timestamp to 'now' during etf_dequeue(). Note that both the qdisc's and the SO_TXTIME ABIs allow for a clockid to be configured, but it's been decided that usage of CLOCK_TAI should be enforced until we decide to allow for other clockids to be used. The rationale here is that PTP times are usually in the TAI scale, thus no other clocks should be necessary. For now, the qdisc will return EINVAL if any clocks other than CLOCK_TAI are used. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:53 +08:00
// SPDX-License-Identifier: GPL-2.0
/* net/sched/sch_etf.c Earliest TxTime First queueing discipline.
*
* Authors: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com>
* Vinicius Costa Gomes <vinicius.gomes@intel.com>
*/
#include <linux/module.h>
#include <linux/types.h>
#include <linux/kernel.h>
#include <linux/string.h>
#include <linux/errno.h>
net/sched: Make etf report drops on error_queue Use the socket error queue for reporting dropped packets if the socket has enabled that feature through the SO_TXTIME API. Packets are dropped either on enqueue() if they aren't accepted by the qdisc or on dequeue() if the system misses their deadline. Those are reported as different errors so applications can react accordingly. Userspace can retrieve the errors through the socket error queue and the corresponding cmsg interfaces. A struct sock_extended_err* is used for returning the error data, and the packet's timestamp can be retrieved by adding both ee_data and ee_info fields as e.g.: ((__u64) serr->ee_data << 32) + serr->ee_info This feature is disabled by default and must be explicitly enabled by applications. Enabling it can bring some overhead for the Tx cycles of the application. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:43:00 +08:00
#include <linux/errqueue.h>
net/sched: Introduce the ETF Qdisc The ETF (Earliest TxTime First) qdisc uses the information added earlier in this series (the socket option SO_TXTIME and the new role of sk_buff->tstamp) to schedule packets transmission based on absolute time. For some workloads, just bandwidth enforcement is not enough, and precise control of the transmission of packets is necessary. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf delta 100000 \ clockid CLOCK_TAI In this example, the Qdisc will provide SW best-effort for the control of the transmission time to the network adapter, the time stamp in the socket will be in reference to the clockid CLOCK_TAI and packets will leave the qdisc "delta" (100000) nanoseconds before its transmission time. The ETF qdisc will buffer packets sorted by their txtime. It will drop packets on enqueue() if their skbuff clockid does not match the clock reference of the Qdisc. Moreover, on dequeue(), a packet will be dropped if it expires while being enqueued. The qdisc also supports the SO_TXTIME deadline mode. For this mode, it will dequeue a packet as soon as possible and change the skb timestamp to 'now' during etf_dequeue(). Note that both the qdisc's and the SO_TXTIME ABIs allow for a clockid to be configured, but it's been decided that usage of CLOCK_TAI should be enforced until we decide to allow for other clockids to be used. The rationale here is that PTP times are usually in the TAI scale, thus no other clocks should be necessary. For now, the qdisc will return EINVAL if any clocks other than CLOCK_TAI are used. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:53 +08:00
#include <linux/rbtree.h>
#include <linux/skbuff.h>
#include <linux/posix-timers.h>
#include <net/netlink.h>
#include <net/sch_generic.h>
#include <net/pkt_sched.h>
#include <net/sock.h>
#define DEADLINE_MODE_IS_ON(x) ((x)->flags & TC_ETF_DEADLINE_MODE_ON)
net/sched: Add HW offloading capability to ETF Add infra so etf qdisc supports HW offload of time-based transmission. For hw offload, the time sorted list is still used, so packets are dequeued always in order of txtime. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf offload delta 100000 \ clockid CLOCK_REALTIME In this example, the Qdisc will use HW offload for the control of the transmission time through the network adapter. The hrtimer used for packets scheduling inside the qdisc will use the clockid CLOCK_REALTIME as reference and packets leave the Qdisc "delta" (100000) nanoseconds before their transmission time. Because this will be using HW offload and since dynamic clocks are not supported by the hrtimer, the system clock and the PHC clock must be synchronized for this mode to behave as expected. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:54 +08:00
#define OFFLOAD_IS_ON(x) ((x)->flags & TC_ETF_OFFLOAD_ON)
net/sched: Introduce the ETF Qdisc The ETF (Earliest TxTime First) qdisc uses the information added earlier in this series (the socket option SO_TXTIME and the new role of sk_buff->tstamp) to schedule packets transmission based on absolute time. For some workloads, just bandwidth enforcement is not enough, and precise control of the transmission of packets is necessary. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf delta 100000 \ clockid CLOCK_TAI In this example, the Qdisc will provide SW best-effort for the control of the transmission time to the network adapter, the time stamp in the socket will be in reference to the clockid CLOCK_TAI and packets will leave the qdisc "delta" (100000) nanoseconds before its transmission time. The ETF qdisc will buffer packets sorted by their txtime. It will drop packets on enqueue() if their skbuff clockid does not match the clock reference of the Qdisc. Moreover, on dequeue(), a packet will be dropped if it expires while being enqueued. The qdisc also supports the SO_TXTIME deadline mode. For this mode, it will dequeue a packet as soon as possible and change the skb timestamp to 'now' during etf_dequeue(). Note that both the qdisc's and the SO_TXTIME ABIs allow for a clockid to be configured, but it's been decided that usage of CLOCK_TAI should be enforced until we decide to allow for other clockids to be used. The rationale here is that PTP times are usually in the TAI scale, thus no other clocks should be necessary. For now, the qdisc will return EINVAL if any clocks other than CLOCK_TAI are used. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:53 +08:00
struct etf_sched_data {
net/sched: Add HW offloading capability to ETF Add infra so etf qdisc supports HW offload of time-based transmission. For hw offload, the time sorted list is still used, so packets are dequeued always in order of txtime. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf offload delta 100000 \ clockid CLOCK_REALTIME In this example, the Qdisc will use HW offload for the control of the transmission time through the network adapter. The hrtimer used for packets scheduling inside the qdisc will use the clockid CLOCK_REALTIME as reference and packets leave the Qdisc "delta" (100000) nanoseconds before their transmission time. Because this will be using HW offload and since dynamic clocks are not supported by the hrtimer, the system clock and the PHC clock must be synchronized for this mode to behave as expected. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:54 +08:00
bool offload;
net/sched: Introduce the ETF Qdisc The ETF (Earliest TxTime First) qdisc uses the information added earlier in this series (the socket option SO_TXTIME and the new role of sk_buff->tstamp) to schedule packets transmission based on absolute time. For some workloads, just bandwidth enforcement is not enough, and precise control of the transmission of packets is necessary. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf delta 100000 \ clockid CLOCK_TAI In this example, the Qdisc will provide SW best-effort for the control of the transmission time to the network adapter, the time stamp in the socket will be in reference to the clockid CLOCK_TAI and packets will leave the qdisc "delta" (100000) nanoseconds before its transmission time. The ETF qdisc will buffer packets sorted by their txtime. It will drop packets on enqueue() if their skbuff clockid does not match the clock reference of the Qdisc. Moreover, on dequeue(), a packet will be dropped if it expires while being enqueued. The qdisc also supports the SO_TXTIME deadline mode. For this mode, it will dequeue a packet as soon as possible and change the skb timestamp to 'now' during etf_dequeue(). Note that both the qdisc's and the SO_TXTIME ABIs allow for a clockid to be configured, but it's been decided that usage of CLOCK_TAI should be enforced until we decide to allow for other clockids to be used. The rationale here is that PTP times are usually in the TAI scale, thus no other clocks should be necessary. For now, the qdisc will return EINVAL if any clocks other than CLOCK_TAI are used. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:53 +08:00
bool deadline_mode;
int clockid;
int queue;
s32 delta; /* in ns */
ktime_t last; /* The txtime of the last skb sent to the netdevice. */
struct rb_root head;
struct qdisc_watchdog watchdog;
ktime_t (*get_time)(void);
};
static const struct nla_policy etf_policy[TCA_ETF_MAX + 1] = {
[TCA_ETF_PARMS] = { .len = sizeof(struct tc_etf_qopt) },
};
static inline int validate_input_params(struct tc_etf_qopt *qopt,
struct netlink_ext_ack *extack)
{
/* Check if params comply to the following rules:
* * Clockid and delta must be valid.
*
* * Dynamic clockids are not supported.
*
* * Delta must be a positive integer.
net/sched: Add HW offloading capability to ETF Add infra so etf qdisc supports HW offload of time-based transmission. For hw offload, the time sorted list is still used, so packets are dequeued always in order of txtime. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf offload delta 100000 \ clockid CLOCK_REALTIME In this example, the Qdisc will use HW offload for the control of the transmission time through the network adapter. The hrtimer used for packets scheduling inside the qdisc will use the clockid CLOCK_REALTIME as reference and packets leave the Qdisc "delta" (100000) nanoseconds before their transmission time. Because this will be using HW offload and since dynamic clocks are not supported by the hrtimer, the system clock and the PHC clock must be synchronized for this mode to behave as expected. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:54 +08:00
*
* Also note that for the HW offload case, we must
* expect that system clocks have been synchronized to PHC.
net/sched: Introduce the ETF Qdisc The ETF (Earliest TxTime First) qdisc uses the information added earlier in this series (the socket option SO_TXTIME and the new role of sk_buff->tstamp) to schedule packets transmission based on absolute time. For some workloads, just bandwidth enforcement is not enough, and precise control of the transmission of packets is necessary. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf delta 100000 \ clockid CLOCK_TAI In this example, the Qdisc will provide SW best-effort for the control of the transmission time to the network adapter, the time stamp in the socket will be in reference to the clockid CLOCK_TAI and packets will leave the qdisc "delta" (100000) nanoseconds before its transmission time. The ETF qdisc will buffer packets sorted by their txtime. It will drop packets on enqueue() if their skbuff clockid does not match the clock reference of the Qdisc. Moreover, on dequeue(), a packet will be dropped if it expires while being enqueued. The qdisc also supports the SO_TXTIME deadline mode. For this mode, it will dequeue a packet as soon as possible and change the skb timestamp to 'now' during etf_dequeue(). Note that both the qdisc's and the SO_TXTIME ABIs allow for a clockid to be configured, but it's been decided that usage of CLOCK_TAI should be enforced until we decide to allow for other clockids to be used. The rationale here is that PTP times are usually in the TAI scale, thus no other clocks should be necessary. For now, the qdisc will return EINVAL if any clocks other than CLOCK_TAI are used. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:53 +08:00
*/
if (qopt->clockid < 0) {
NL_SET_ERR_MSG(extack, "Dynamic clockids are not supported");
return -ENOTSUPP;
}
if (qopt->clockid != CLOCK_TAI) {
NL_SET_ERR_MSG(extack, "Invalid clockid. CLOCK_TAI must be used");
return -EINVAL;
}
if (qopt->delta < 0) {
NL_SET_ERR_MSG(extack, "Delta must be positive");
return -EINVAL;
}
return 0;
}
static bool is_packet_valid(struct Qdisc *sch, struct sk_buff *nskb)
{
struct etf_sched_data *q = qdisc_priv(sch);
ktime_t txtime = nskb->tstamp;
struct sock *sk = nskb->sk;
ktime_t now;
if (!sk)
return false;
if (!sock_flag(sk, SOCK_TXTIME))
return false;
/* We don't perform crosstimestamping.
* Drop if packet's clockid differs from qdisc's.
*/
if (sk->sk_clockid != q->clockid)
return false;
if (sk->sk_txtime_deadline_mode != q->deadline_mode)
return false;
now = q->get_time();
if (ktime_before(txtime, now) || ktime_before(txtime, q->last))
return false;
return true;
}
static struct sk_buff *etf_peek_timesortedlist(struct Qdisc *sch)
{
struct etf_sched_data *q = qdisc_priv(sch);
struct rb_node *p;
p = rb_first(&q->head);
if (!p)
return NULL;
return rb_to_skb(p);
}
static void reset_watchdog(struct Qdisc *sch)
{
struct etf_sched_data *q = qdisc_priv(sch);
struct sk_buff *skb = etf_peek_timesortedlist(sch);
ktime_t next;
if (!skb)
return;
next = ktime_sub_ns(skb->tstamp, q->delta);
qdisc_watchdog_schedule_ns(&q->watchdog, ktime_to_ns(next));
}
net/sched: Make etf report drops on error_queue Use the socket error queue for reporting dropped packets if the socket has enabled that feature through the SO_TXTIME API. Packets are dropped either on enqueue() if they aren't accepted by the qdisc or on dequeue() if the system misses their deadline. Those are reported as different errors so applications can react accordingly. Userspace can retrieve the errors through the socket error queue and the corresponding cmsg interfaces. A struct sock_extended_err* is used for returning the error data, and the packet's timestamp can be retrieved by adding both ee_data and ee_info fields as e.g.: ((__u64) serr->ee_data << 32) + serr->ee_info This feature is disabled by default and must be explicitly enabled by applications. Enabling it can bring some overhead for the Tx cycles of the application. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:43:00 +08:00
static void report_sock_error(struct sk_buff *skb, u32 err, u8 code)
{
struct sock_exterr_skb *serr;
struct sk_buff *clone;
ktime_t txtime = skb->tstamp;
if (!skb->sk || !(skb->sk->sk_txtime_report_errors))
return;
clone = skb_clone(skb, GFP_ATOMIC);
if (!clone)
return;
serr = SKB_EXT_ERR(clone);
serr->ee.ee_errno = err;
serr->ee.ee_origin = SO_EE_ORIGIN_TXTIME;
serr->ee.ee_type = 0;
serr->ee.ee_code = code;
serr->ee.ee_pad = 0;
serr->ee.ee_data = (txtime >> 32); /* high part of tstamp */
serr->ee.ee_info = txtime; /* low part of tstamp */
if (sock_queue_err_skb(skb->sk, clone))
kfree_skb(clone);
}
net/sched: Introduce the ETF Qdisc The ETF (Earliest TxTime First) qdisc uses the information added earlier in this series (the socket option SO_TXTIME and the new role of sk_buff->tstamp) to schedule packets transmission based on absolute time. For some workloads, just bandwidth enforcement is not enough, and precise control of the transmission of packets is necessary. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf delta 100000 \ clockid CLOCK_TAI In this example, the Qdisc will provide SW best-effort for the control of the transmission time to the network adapter, the time stamp in the socket will be in reference to the clockid CLOCK_TAI and packets will leave the qdisc "delta" (100000) nanoseconds before its transmission time. The ETF qdisc will buffer packets sorted by their txtime. It will drop packets on enqueue() if their skbuff clockid does not match the clock reference of the Qdisc. Moreover, on dequeue(), a packet will be dropped if it expires while being enqueued. The qdisc also supports the SO_TXTIME deadline mode. For this mode, it will dequeue a packet as soon as possible and change the skb timestamp to 'now' during etf_dequeue(). Note that both the qdisc's and the SO_TXTIME ABIs allow for a clockid to be configured, but it's been decided that usage of CLOCK_TAI should be enforced until we decide to allow for other clockids to be used. The rationale here is that PTP times are usually in the TAI scale, thus no other clocks should be necessary. For now, the qdisc will return EINVAL if any clocks other than CLOCK_TAI are used. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:53 +08:00
static int etf_enqueue_timesortedlist(struct sk_buff *nskb, struct Qdisc *sch,
struct sk_buff **to_free)
{
struct etf_sched_data *q = qdisc_priv(sch);
struct rb_node **p = &q->head.rb_node, *parent = NULL;
ktime_t txtime = nskb->tstamp;
net/sched: Make etf report drops on error_queue Use the socket error queue for reporting dropped packets if the socket has enabled that feature through the SO_TXTIME API. Packets are dropped either on enqueue() if they aren't accepted by the qdisc or on dequeue() if the system misses their deadline. Those are reported as different errors so applications can react accordingly. Userspace can retrieve the errors through the socket error queue and the corresponding cmsg interfaces. A struct sock_extended_err* is used for returning the error data, and the packet's timestamp can be retrieved by adding both ee_data and ee_info fields as e.g.: ((__u64) serr->ee_data << 32) + serr->ee_info This feature is disabled by default and must be explicitly enabled by applications. Enabling it can bring some overhead for the Tx cycles of the application. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:43:00 +08:00
if (!is_packet_valid(sch, nskb)) {
report_sock_error(nskb, EINVAL,
SO_EE_CODE_TXTIME_INVALID_PARAM);
net/sched: Introduce the ETF Qdisc The ETF (Earliest TxTime First) qdisc uses the information added earlier in this series (the socket option SO_TXTIME and the new role of sk_buff->tstamp) to schedule packets transmission based on absolute time. For some workloads, just bandwidth enforcement is not enough, and precise control of the transmission of packets is necessary. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf delta 100000 \ clockid CLOCK_TAI In this example, the Qdisc will provide SW best-effort for the control of the transmission time to the network adapter, the time stamp in the socket will be in reference to the clockid CLOCK_TAI and packets will leave the qdisc "delta" (100000) nanoseconds before its transmission time. The ETF qdisc will buffer packets sorted by their txtime. It will drop packets on enqueue() if their skbuff clockid does not match the clock reference of the Qdisc. Moreover, on dequeue(), a packet will be dropped if it expires while being enqueued. The qdisc also supports the SO_TXTIME deadline mode. For this mode, it will dequeue a packet as soon as possible and change the skb timestamp to 'now' during etf_dequeue(). Note that both the qdisc's and the SO_TXTIME ABIs allow for a clockid to be configured, but it's been decided that usage of CLOCK_TAI should be enforced until we decide to allow for other clockids to be used. The rationale here is that PTP times are usually in the TAI scale, thus no other clocks should be necessary. For now, the qdisc will return EINVAL if any clocks other than CLOCK_TAI are used. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:53 +08:00
return qdisc_drop(nskb, sch, to_free);
net/sched: Make etf report drops on error_queue Use the socket error queue for reporting dropped packets if the socket has enabled that feature through the SO_TXTIME API. Packets are dropped either on enqueue() if they aren't accepted by the qdisc or on dequeue() if the system misses their deadline. Those are reported as different errors so applications can react accordingly. Userspace can retrieve the errors through the socket error queue and the corresponding cmsg interfaces. A struct sock_extended_err* is used for returning the error data, and the packet's timestamp can be retrieved by adding both ee_data and ee_info fields as e.g.: ((__u64) serr->ee_data << 32) + serr->ee_info This feature is disabled by default and must be explicitly enabled by applications. Enabling it can bring some overhead for the Tx cycles of the application. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:43:00 +08:00
}
net/sched: Introduce the ETF Qdisc The ETF (Earliest TxTime First) qdisc uses the information added earlier in this series (the socket option SO_TXTIME and the new role of sk_buff->tstamp) to schedule packets transmission based on absolute time. For some workloads, just bandwidth enforcement is not enough, and precise control of the transmission of packets is necessary. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf delta 100000 \ clockid CLOCK_TAI In this example, the Qdisc will provide SW best-effort for the control of the transmission time to the network adapter, the time stamp in the socket will be in reference to the clockid CLOCK_TAI and packets will leave the qdisc "delta" (100000) nanoseconds before its transmission time. The ETF qdisc will buffer packets sorted by their txtime. It will drop packets on enqueue() if their skbuff clockid does not match the clock reference of the Qdisc. Moreover, on dequeue(), a packet will be dropped if it expires while being enqueued. The qdisc also supports the SO_TXTIME deadline mode. For this mode, it will dequeue a packet as soon as possible and change the skb timestamp to 'now' during etf_dequeue(). Note that both the qdisc's and the SO_TXTIME ABIs allow for a clockid to be configured, but it's been decided that usage of CLOCK_TAI should be enforced until we decide to allow for other clockids to be used. The rationale here is that PTP times are usually in the TAI scale, thus no other clocks should be necessary. For now, the qdisc will return EINVAL if any clocks other than CLOCK_TAI are used. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:53 +08:00
while (*p) {
struct sk_buff *skb;
parent = *p;
skb = rb_to_skb(parent);
if (ktime_after(txtime, skb->tstamp))
p = &parent->rb_right;
else
p = &parent->rb_left;
}
rb_link_node(&nskb->rbnode, parent, p);
rb_insert_color(&nskb->rbnode, &q->head);
qdisc_qstats_backlog_inc(sch, nskb);
sch->q.qlen++;
/* Now we may need to re-arm the qdisc watchdog for the next packet. */
reset_watchdog(sch);
return NET_XMIT_SUCCESS;
}
static void timesortedlist_erase(struct Qdisc *sch, struct sk_buff *skb,
bool drop)
{
struct etf_sched_data *q = qdisc_priv(sch);
rb_erase(&skb->rbnode, &q->head);
/* The rbnode field in the skb re-uses these fields, now that
* we are done with the rbnode, reset them.
*/
skb->next = NULL;
skb->prev = NULL;
skb->dev = qdisc_dev(sch);
qdisc_qstats_backlog_dec(sch, skb);
if (drop) {
struct sk_buff *to_free = NULL;
net/sched: Make etf report drops on error_queue Use the socket error queue for reporting dropped packets if the socket has enabled that feature through the SO_TXTIME API. Packets are dropped either on enqueue() if they aren't accepted by the qdisc or on dequeue() if the system misses their deadline. Those are reported as different errors so applications can react accordingly. Userspace can retrieve the errors through the socket error queue and the corresponding cmsg interfaces. A struct sock_extended_err* is used for returning the error data, and the packet's timestamp can be retrieved by adding both ee_data and ee_info fields as e.g.: ((__u64) serr->ee_data << 32) + serr->ee_info This feature is disabled by default and must be explicitly enabled by applications. Enabling it can bring some overhead for the Tx cycles of the application. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:43:00 +08:00
report_sock_error(skb, ECANCELED, SO_EE_CODE_TXTIME_MISSED);
net/sched: Introduce the ETF Qdisc The ETF (Earliest TxTime First) qdisc uses the information added earlier in this series (the socket option SO_TXTIME and the new role of sk_buff->tstamp) to schedule packets transmission based on absolute time. For some workloads, just bandwidth enforcement is not enough, and precise control of the transmission of packets is necessary. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf delta 100000 \ clockid CLOCK_TAI In this example, the Qdisc will provide SW best-effort for the control of the transmission time to the network adapter, the time stamp in the socket will be in reference to the clockid CLOCK_TAI and packets will leave the qdisc "delta" (100000) nanoseconds before its transmission time. The ETF qdisc will buffer packets sorted by their txtime. It will drop packets on enqueue() if their skbuff clockid does not match the clock reference of the Qdisc. Moreover, on dequeue(), a packet will be dropped if it expires while being enqueued. The qdisc also supports the SO_TXTIME deadline mode. For this mode, it will dequeue a packet as soon as possible and change the skb timestamp to 'now' during etf_dequeue(). Note that both the qdisc's and the SO_TXTIME ABIs allow for a clockid to be configured, but it's been decided that usage of CLOCK_TAI should be enforced until we decide to allow for other clockids to be used. The rationale here is that PTP times are usually in the TAI scale, thus no other clocks should be necessary. For now, the qdisc will return EINVAL if any clocks other than CLOCK_TAI are used. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:53 +08:00
qdisc_drop(skb, sch, &to_free);
kfree_skb_list(to_free);
qdisc_qstats_overlimit(sch);
} else {
qdisc_bstats_update(sch, skb);
q->last = skb->tstamp;
}
sch->q.qlen--;
}
static struct sk_buff *etf_dequeue_timesortedlist(struct Qdisc *sch)
{
struct etf_sched_data *q = qdisc_priv(sch);
struct sk_buff *skb;
ktime_t now, next;
skb = etf_peek_timesortedlist(sch);
if (!skb)
return NULL;
now = q->get_time();
/* Drop if packet has expired while in queue. */
if (ktime_before(skb->tstamp, now)) {
timesortedlist_erase(sch, skb, true);
skb = NULL;
goto out;
}
/* When in deadline mode, dequeue as soon as possible and change the
* txtime from deadline to (now + delta).
*/
if (q->deadline_mode) {
timesortedlist_erase(sch, skb, false);
skb->tstamp = now;
goto out;
}
next = ktime_sub_ns(skb->tstamp, q->delta);
/* Dequeue only if now is within the [txtime - delta, txtime] range. */
if (ktime_after(now, next))
timesortedlist_erase(sch, skb, false);
else
skb = NULL;
out:
/* Now we may need to re-arm the qdisc watchdog for the next packet. */
reset_watchdog(sch);
return skb;
}
net/sched: Add HW offloading capability to ETF Add infra so etf qdisc supports HW offload of time-based transmission. For hw offload, the time sorted list is still used, so packets are dequeued always in order of txtime. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf offload delta 100000 \ clockid CLOCK_REALTIME In this example, the Qdisc will use HW offload for the control of the transmission time through the network adapter. The hrtimer used for packets scheduling inside the qdisc will use the clockid CLOCK_REALTIME as reference and packets leave the Qdisc "delta" (100000) nanoseconds before their transmission time. Because this will be using HW offload and since dynamic clocks are not supported by the hrtimer, the system clock and the PHC clock must be synchronized for this mode to behave as expected. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:54 +08:00
static void etf_disable_offload(struct net_device *dev,
struct etf_sched_data *q)
{
struct tc_etf_qopt_offload etf = { };
const struct net_device_ops *ops;
int err;
if (!q->offload)
return;
ops = dev->netdev_ops;
if (!ops->ndo_setup_tc)
return;
etf.queue = q->queue;
etf.enable = 0;
err = ops->ndo_setup_tc(dev, TC_SETUP_QDISC_ETF, &etf);
if (err < 0)
pr_warn("Couldn't disable ETF offload for queue %d\n",
etf.queue);
}
static int etf_enable_offload(struct net_device *dev, struct etf_sched_data *q,
struct netlink_ext_ack *extack)
{
const struct net_device_ops *ops = dev->netdev_ops;
struct tc_etf_qopt_offload etf = { };
int err;
if (q->offload)
return 0;
if (!ops->ndo_setup_tc) {
NL_SET_ERR_MSG(extack, "Specified device does not support ETF offload");
return -EOPNOTSUPP;
}
etf.queue = q->queue;
etf.enable = 1;
err = ops->ndo_setup_tc(dev, TC_SETUP_QDISC_ETF, &etf);
if (err < 0) {
NL_SET_ERR_MSG(extack, "Specified device failed to setup ETF hardware offload");
return err;
}
return 0;
}
net/sched: Introduce the ETF Qdisc The ETF (Earliest TxTime First) qdisc uses the information added earlier in this series (the socket option SO_TXTIME and the new role of sk_buff->tstamp) to schedule packets transmission based on absolute time. For some workloads, just bandwidth enforcement is not enough, and precise control of the transmission of packets is necessary. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf delta 100000 \ clockid CLOCK_TAI In this example, the Qdisc will provide SW best-effort for the control of the transmission time to the network adapter, the time stamp in the socket will be in reference to the clockid CLOCK_TAI and packets will leave the qdisc "delta" (100000) nanoseconds before its transmission time. The ETF qdisc will buffer packets sorted by their txtime. It will drop packets on enqueue() if their skbuff clockid does not match the clock reference of the Qdisc. Moreover, on dequeue(), a packet will be dropped if it expires while being enqueued. The qdisc also supports the SO_TXTIME deadline mode. For this mode, it will dequeue a packet as soon as possible and change the skb timestamp to 'now' during etf_dequeue(). Note that both the qdisc's and the SO_TXTIME ABIs allow for a clockid to be configured, but it's been decided that usage of CLOCK_TAI should be enforced until we decide to allow for other clockids to be used. The rationale here is that PTP times are usually in the TAI scale, thus no other clocks should be necessary. For now, the qdisc will return EINVAL if any clocks other than CLOCK_TAI are used. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:53 +08:00
static int etf_init(struct Qdisc *sch, struct nlattr *opt,
struct netlink_ext_ack *extack)
{
struct etf_sched_data *q = qdisc_priv(sch);
struct net_device *dev = qdisc_dev(sch);
struct nlattr *tb[TCA_ETF_MAX + 1];
struct tc_etf_qopt *qopt;
int err;
if (!opt) {
NL_SET_ERR_MSG(extack,
"Missing ETF qdisc options which are mandatory");
return -EINVAL;
}
err = nla_parse_nested(tb, TCA_ETF_MAX, opt, etf_policy, extack);
if (err < 0)
return err;
if (!tb[TCA_ETF_PARMS]) {
NL_SET_ERR_MSG(extack, "Missing mandatory ETF parameters");
return -EINVAL;
}
qopt = nla_data(tb[TCA_ETF_PARMS]);
net/sched: Add HW offloading capability to ETF Add infra so etf qdisc supports HW offload of time-based transmission. For hw offload, the time sorted list is still used, so packets are dequeued always in order of txtime. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf offload delta 100000 \ clockid CLOCK_REALTIME In this example, the Qdisc will use HW offload for the control of the transmission time through the network adapter. The hrtimer used for packets scheduling inside the qdisc will use the clockid CLOCK_REALTIME as reference and packets leave the Qdisc "delta" (100000) nanoseconds before their transmission time. Because this will be using HW offload and since dynamic clocks are not supported by the hrtimer, the system clock and the PHC clock must be synchronized for this mode to behave as expected. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:54 +08:00
pr_debug("delta %d clockid %d offload %s deadline %s\n",
net/sched: Introduce the ETF Qdisc The ETF (Earliest TxTime First) qdisc uses the information added earlier in this series (the socket option SO_TXTIME and the new role of sk_buff->tstamp) to schedule packets transmission based on absolute time. For some workloads, just bandwidth enforcement is not enough, and precise control of the transmission of packets is necessary. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf delta 100000 \ clockid CLOCK_TAI In this example, the Qdisc will provide SW best-effort for the control of the transmission time to the network adapter, the time stamp in the socket will be in reference to the clockid CLOCK_TAI and packets will leave the qdisc "delta" (100000) nanoseconds before its transmission time. The ETF qdisc will buffer packets sorted by their txtime. It will drop packets on enqueue() if their skbuff clockid does not match the clock reference of the Qdisc. Moreover, on dequeue(), a packet will be dropped if it expires while being enqueued. The qdisc also supports the SO_TXTIME deadline mode. For this mode, it will dequeue a packet as soon as possible and change the skb timestamp to 'now' during etf_dequeue(). Note that both the qdisc's and the SO_TXTIME ABIs allow for a clockid to be configured, but it's been decided that usage of CLOCK_TAI should be enforced until we decide to allow for other clockids to be used. The rationale here is that PTP times are usually in the TAI scale, thus no other clocks should be necessary. For now, the qdisc will return EINVAL if any clocks other than CLOCK_TAI are used. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:53 +08:00
qopt->delta, qopt->clockid,
net/sched: Add HW offloading capability to ETF Add infra so etf qdisc supports HW offload of time-based transmission. For hw offload, the time sorted list is still used, so packets are dequeued always in order of txtime. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf offload delta 100000 \ clockid CLOCK_REALTIME In this example, the Qdisc will use HW offload for the control of the transmission time through the network adapter. The hrtimer used for packets scheduling inside the qdisc will use the clockid CLOCK_REALTIME as reference and packets leave the Qdisc "delta" (100000) nanoseconds before their transmission time. Because this will be using HW offload and since dynamic clocks are not supported by the hrtimer, the system clock and the PHC clock must be synchronized for this mode to behave as expected. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:54 +08:00
OFFLOAD_IS_ON(qopt) ? "on" : "off",
net/sched: Introduce the ETF Qdisc The ETF (Earliest TxTime First) qdisc uses the information added earlier in this series (the socket option SO_TXTIME and the new role of sk_buff->tstamp) to schedule packets transmission based on absolute time. For some workloads, just bandwidth enforcement is not enough, and precise control of the transmission of packets is necessary. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf delta 100000 \ clockid CLOCK_TAI In this example, the Qdisc will provide SW best-effort for the control of the transmission time to the network adapter, the time stamp in the socket will be in reference to the clockid CLOCK_TAI and packets will leave the qdisc "delta" (100000) nanoseconds before its transmission time. The ETF qdisc will buffer packets sorted by their txtime. It will drop packets on enqueue() if their skbuff clockid does not match the clock reference of the Qdisc. Moreover, on dequeue(), a packet will be dropped if it expires while being enqueued. The qdisc also supports the SO_TXTIME deadline mode. For this mode, it will dequeue a packet as soon as possible and change the skb timestamp to 'now' during etf_dequeue(). Note that both the qdisc's and the SO_TXTIME ABIs allow for a clockid to be configured, but it's been decided that usage of CLOCK_TAI should be enforced until we decide to allow for other clockids to be used. The rationale here is that PTP times are usually in the TAI scale, thus no other clocks should be necessary. For now, the qdisc will return EINVAL if any clocks other than CLOCK_TAI are used. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:53 +08:00
DEADLINE_MODE_IS_ON(qopt) ? "on" : "off");
err = validate_input_params(qopt, extack);
if (err < 0)
return err;
q->queue = sch->dev_queue - netdev_get_tx_queue(dev, 0);
net/sched: Add HW offloading capability to ETF Add infra so etf qdisc supports HW offload of time-based transmission. For hw offload, the time sorted list is still used, so packets are dequeued always in order of txtime. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf offload delta 100000 \ clockid CLOCK_REALTIME In this example, the Qdisc will use HW offload for the control of the transmission time through the network adapter. The hrtimer used for packets scheduling inside the qdisc will use the clockid CLOCK_REALTIME as reference and packets leave the Qdisc "delta" (100000) nanoseconds before their transmission time. Because this will be using HW offload and since dynamic clocks are not supported by the hrtimer, the system clock and the PHC clock must be synchronized for this mode to behave as expected. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:54 +08:00
if (OFFLOAD_IS_ON(qopt)) {
err = etf_enable_offload(dev, q, extack);
if (err < 0)
return err;
}
net/sched: Introduce the ETF Qdisc The ETF (Earliest TxTime First) qdisc uses the information added earlier in this series (the socket option SO_TXTIME and the new role of sk_buff->tstamp) to schedule packets transmission based on absolute time. For some workloads, just bandwidth enforcement is not enough, and precise control of the transmission of packets is necessary. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf delta 100000 \ clockid CLOCK_TAI In this example, the Qdisc will provide SW best-effort for the control of the transmission time to the network adapter, the time stamp in the socket will be in reference to the clockid CLOCK_TAI and packets will leave the qdisc "delta" (100000) nanoseconds before its transmission time. The ETF qdisc will buffer packets sorted by their txtime. It will drop packets on enqueue() if their skbuff clockid does not match the clock reference of the Qdisc. Moreover, on dequeue(), a packet will be dropped if it expires while being enqueued. The qdisc also supports the SO_TXTIME deadline mode. For this mode, it will dequeue a packet as soon as possible and change the skb timestamp to 'now' during etf_dequeue(). Note that both the qdisc's and the SO_TXTIME ABIs allow for a clockid to be configured, but it's been decided that usage of CLOCK_TAI should be enforced until we decide to allow for other clockids to be used. The rationale here is that PTP times are usually in the TAI scale, thus no other clocks should be necessary. For now, the qdisc will return EINVAL if any clocks other than CLOCK_TAI are used. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:53 +08:00
/* Everything went OK, save the parameters used. */
q->delta = qopt->delta;
q->clockid = qopt->clockid;
net/sched: Add HW offloading capability to ETF Add infra so etf qdisc supports HW offload of time-based transmission. For hw offload, the time sorted list is still used, so packets are dequeued always in order of txtime. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf offload delta 100000 \ clockid CLOCK_REALTIME In this example, the Qdisc will use HW offload for the control of the transmission time through the network adapter. The hrtimer used for packets scheduling inside the qdisc will use the clockid CLOCK_REALTIME as reference and packets leave the Qdisc "delta" (100000) nanoseconds before their transmission time. Because this will be using HW offload and since dynamic clocks are not supported by the hrtimer, the system clock and the PHC clock must be synchronized for this mode to behave as expected. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:54 +08:00
q->offload = OFFLOAD_IS_ON(qopt);
net/sched: Introduce the ETF Qdisc The ETF (Earliest TxTime First) qdisc uses the information added earlier in this series (the socket option SO_TXTIME and the new role of sk_buff->tstamp) to schedule packets transmission based on absolute time. For some workloads, just bandwidth enforcement is not enough, and precise control of the transmission of packets is necessary. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf delta 100000 \ clockid CLOCK_TAI In this example, the Qdisc will provide SW best-effort for the control of the transmission time to the network adapter, the time stamp in the socket will be in reference to the clockid CLOCK_TAI and packets will leave the qdisc "delta" (100000) nanoseconds before its transmission time. The ETF qdisc will buffer packets sorted by their txtime. It will drop packets on enqueue() if their skbuff clockid does not match the clock reference of the Qdisc. Moreover, on dequeue(), a packet will be dropped if it expires while being enqueued. The qdisc also supports the SO_TXTIME deadline mode. For this mode, it will dequeue a packet as soon as possible and change the skb timestamp to 'now' during etf_dequeue(). Note that both the qdisc's and the SO_TXTIME ABIs allow for a clockid to be configured, but it's been decided that usage of CLOCK_TAI should be enforced until we decide to allow for other clockids to be used. The rationale here is that PTP times are usually in the TAI scale, thus no other clocks should be necessary. For now, the qdisc will return EINVAL if any clocks other than CLOCK_TAI are used. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:53 +08:00
q->deadline_mode = DEADLINE_MODE_IS_ON(qopt);
switch (q->clockid) {
case CLOCK_REALTIME:
q->get_time = ktime_get_real;
break;
case CLOCK_MONOTONIC:
q->get_time = ktime_get;
break;
case CLOCK_BOOTTIME:
q->get_time = ktime_get_boottime;
break;
case CLOCK_TAI:
q->get_time = ktime_get_clocktai;
break;
default:
NL_SET_ERR_MSG(extack, "Clockid is not supported");
return -ENOTSUPP;
}
qdisc_watchdog_init_clockid(&q->watchdog, sch, q->clockid);
return 0;
}
static void timesortedlist_clear(struct Qdisc *sch)
{
struct etf_sched_data *q = qdisc_priv(sch);
struct rb_node *p = rb_first(&q->head);
while (p) {
struct sk_buff *skb = rb_to_skb(p);
p = rb_next(p);
rb_erase(&skb->rbnode, &q->head);
rtnl_kfree_skbs(skb, skb);
sch->q.qlen--;
}
}
static void etf_reset(struct Qdisc *sch)
{
struct etf_sched_data *q = qdisc_priv(sch);
/* Only cancel watchdog if it's been initialized. */
if (q->watchdog.qdisc == sch)
qdisc_watchdog_cancel(&q->watchdog);
/* No matter which mode we are on, it's safe to clear both lists. */
timesortedlist_clear(sch);
__qdisc_reset_queue(&sch->q);
sch->qstats.backlog = 0;
sch->q.qlen = 0;
q->last = 0;
}
static void etf_destroy(struct Qdisc *sch)
{
struct etf_sched_data *q = qdisc_priv(sch);
net/sched: Add HW offloading capability to ETF Add infra so etf qdisc supports HW offload of time-based transmission. For hw offload, the time sorted list is still used, so packets are dequeued always in order of txtime. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf offload delta 100000 \ clockid CLOCK_REALTIME In this example, the Qdisc will use HW offload for the control of the transmission time through the network adapter. The hrtimer used for packets scheduling inside the qdisc will use the clockid CLOCK_REALTIME as reference and packets leave the Qdisc "delta" (100000) nanoseconds before their transmission time. Because this will be using HW offload and since dynamic clocks are not supported by the hrtimer, the system clock and the PHC clock must be synchronized for this mode to behave as expected. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:54 +08:00
struct net_device *dev = qdisc_dev(sch);
net/sched: Introduce the ETF Qdisc The ETF (Earliest TxTime First) qdisc uses the information added earlier in this series (the socket option SO_TXTIME and the new role of sk_buff->tstamp) to schedule packets transmission based on absolute time. For some workloads, just bandwidth enforcement is not enough, and precise control of the transmission of packets is necessary. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf delta 100000 \ clockid CLOCK_TAI In this example, the Qdisc will provide SW best-effort for the control of the transmission time to the network adapter, the time stamp in the socket will be in reference to the clockid CLOCK_TAI and packets will leave the qdisc "delta" (100000) nanoseconds before its transmission time. The ETF qdisc will buffer packets sorted by their txtime. It will drop packets on enqueue() if their skbuff clockid does not match the clock reference of the Qdisc. Moreover, on dequeue(), a packet will be dropped if it expires while being enqueued. The qdisc also supports the SO_TXTIME deadline mode. For this mode, it will dequeue a packet as soon as possible and change the skb timestamp to 'now' during etf_dequeue(). Note that both the qdisc's and the SO_TXTIME ABIs allow for a clockid to be configured, but it's been decided that usage of CLOCK_TAI should be enforced until we decide to allow for other clockids to be used. The rationale here is that PTP times are usually in the TAI scale, thus no other clocks should be necessary. For now, the qdisc will return EINVAL if any clocks other than CLOCK_TAI are used. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:53 +08:00
/* Only cancel watchdog if it's been initialized. */
if (q->watchdog.qdisc == sch)
qdisc_watchdog_cancel(&q->watchdog);
net/sched: Add HW offloading capability to ETF Add infra so etf qdisc supports HW offload of time-based transmission. For hw offload, the time sorted list is still used, so packets are dequeued always in order of txtime. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf offload delta 100000 \ clockid CLOCK_REALTIME In this example, the Qdisc will use HW offload for the control of the transmission time through the network adapter. The hrtimer used for packets scheduling inside the qdisc will use the clockid CLOCK_REALTIME as reference and packets leave the Qdisc "delta" (100000) nanoseconds before their transmission time. Because this will be using HW offload and since dynamic clocks are not supported by the hrtimer, the system clock and the PHC clock must be synchronized for this mode to behave as expected. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:54 +08:00
etf_disable_offload(dev, q);
net/sched: Introduce the ETF Qdisc The ETF (Earliest TxTime First) qdisc uses the information added earlier in this series (the socket option SO_TXTIME and the new role of sk_buff->tstamp) to schedule packets transmission based on absolute time. For some workloads, just bandwidth enforcement is not enough, and precise control of the transmission of packets is necessary. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf delta 100000 \ clockid CLOCK_TAI In this example, the Qdisc will provide SW best-effort for the control of the transmission time to the network adapter, the time stamp in the socket will be in reference to the clockid CLOCK_TAI and packets will leave the qdisc "delta" (100000) nanoseconds before its transmission time. The ETF qdisc will buffer packets sorted by their txtime. It will drop packets on enqueue() if their skbuff clockid does not match the clock reference of the Qdisc. Moreover, on dequeue(), a packet will be dropped if it expires while being enqueued. The qdisc also supports the SO_TXTIME deadline mode. For this mode, it will dequeue a packet as soon as possible and change the skb timestamp to 'now' during etf_dequeue(). Note that both the qdisc's and the SO_TXTIME ABIs allow for a clockid to be configured, but it's been decided that usage of CLOCK_TAI should be enforced until we decide to allow for other clockids to be used. The rationale here is that PTP times are usually in the TAI scale, thus no other clocks should be necessary. For now, the qdisc will return EINVAL if any clocks other than CLOCK_TAI are used. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:53 +08:00
}
static int etf_dump(struct Qdisc *sch, struct sk_buff *skb)
{
struct etf_sched_data *q = qdisc_priv(sch);
struct tc_etf_qopt opt = { };
struct nlattr *nest;
nest = nla_nest_start(skb, TCA_OPTIONS);
if (!nest)
goto nla_put_failure;
opt.delta = q->delta;
opt.clockid = q->clockid;
net/sched: Add HW offloading capability to ETF Add infra so etf qdisc supports HW offload of time-based transmission. For hw offload, the time sorted list is still used, so packets are dequeued always in order of txtime. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf offload delta 100000 \ clockid CLOCK_REALTIME In this example, the Qdisc will use HW offload for the control of the transmission time through the network adapter. The hrtimer used for packets scheduling inside the qdisc will use the clockid CLOCK_REALTIME as reference and packets leave the Qdisc "delta" (100000) nanoseconds before their transmission time. Because this will be using HW offload and since dynamic clocks are not supported by the hrtimer, the system clock and the PHC clock must be synchronized for this mode to behave as expected. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:54 +08:00
if (q->offload)
opt.flags |= TC_ETF_OFFLOAD_ON;
net/sched: Introduce the ETF Qdisc The ETF (Earliest TxTime First) qdisc uses the information added earlier in this series (the socket option SO_TXTIME and the new role of sk_buff->tstamp) to schedule packets transmission based on absolute time. For some workloads, just bandwidth enforcement is not enough, and precise control of the transmission of packets is necessary. Example: $ tc qdisc replace dev enp2s0 parent root handle 100 mqprio num_tc 3 \ map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 queues 1@0 1@1 2@2 hw 0 $ tc qdisc add dev enp2s0 parent 100:1 etf delta 100000 \ clockid CLOCK_TAI In this example, the Qdisc will provide SW best-effort for the control of the transmission time to the network adapter, the time stamp in the socket will be in reference to the clockid CLOCK_TAI and packets will leave the qdisc "delta" (100000) nanoseconds before its transmission time. The ETF qdisc will buffer packets sorted by their txtime. It will drop packets on enqueue() if their skbuff clockid does not match the clock reference of the Qdisc. Moreover, on dequeue(), a packet will be dropped if it expires while being enqueued. The qdisc also supports the SO_TXTIME deadline mode. For this mode, it will dequeue a packet as soon as possible and change the skb timestamp to 'now' during etf_dequeue(). Note that both the qdisc's and the SO_TXTIME ABIs allow for a clockid to be configured, but it's been decided that usage of CLOCK_TAI should be enforced until we decide to allow for other clockids to be used. The rationale here is that PTP times are usually in the TAI scale, thus no other clocks should be necessary. For now, the qdisc will return EINVAL if any clocks other than CLOCK_TAI are used. Signed-off-by: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2018-07-04 06:42:53 +08:00
if (q->deadline_mode)
opt.flags |= TC_ETF_DEADLINE_MODE_ON;
if (nla_put(skb, TCA_ETF_PARMS, sizeof(opt), &opt))
goto nla_put_failure;
return nla_nest_end(skb, nest);
nla_put_failure:
nla_nest_cancel(skb, nest);
return -1;
}
static struct Qdisc_ops etf_qdisc_ops __read_mostly = {
.id = "etf",
.priv_size = sizeof(struct etf_sched_data),
.enqueue = etf_enqueue_timesortedlist,
.dequeue = etf_dequeue_timesortedlist,
.peek = etf_peek_timesortedlist,
.init = etf_init,
.reset = etf_reset,
.destroy = etf_destroy,
.dump = etf_dump,
.owner = THIS_MODULE,
};
static int __init etf_module_init(void)
{
return register_qdisc(&etf_qdisc_ops);
}
static void __exit etf_module_exit(void)
{
unregister_qdisc(&etf_qdisc_ops);
}
module_init(etf_module_init)
module_exit(etf_module_exit)
MODULE_LICENSE("GPL");