diff mbox series

[3/4] sch_api: Allow reducing queue backlog by a negative value

Message ID 20190107194733.31138-4-toke@toke.dk
State Changes Requested
Delegated to: David Miller
Headers show
Series sched: Fix qdisc interactions exposed by using sch_cake as a leaf qdisc | expand

Commit Message

Toke Høiland-Jørgensen Jan. 7, 2019, 7:47 p.m. UTC 
From: Toke Høiland-Jørgensen <toke@redhat.com>

With GSO splitting in sch_cake, we can decrease as well as increase the
qlen. To make it possible to signal this up the qdisc tree, change
qdisc_tree_reduce_backlog() to accept signed integer values as the number
of packets and bytes to reduce the backlog by.

Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
---
 include/net/sch_generic.h | 3 +--
 net/sched/sch_api.c       | 3 +--
 2 files changed, 2 insertions(+), 4 deletions(-)

Comments

Eric Dumazet Jan. 7, 2019, 8:54 p.m. UTC | #1 
On 01/07/2019 11:47 AM, Toke Høiland-Jørgensen wrote:
> From: Toke Høiland-Jørgensen <toke@redhat.com>
> 
> With GSO splitting in sch_cake, we can decrease as well as increase the
> qlen. To make it possible to signal this up the qdisc tree, change
> qdisc_tree_reduce_backlog() to accept signed integer values as the number
> of packets and bytes to reduce the backlog by.


Not sure why this patch is needed, given we use u32 integers
for q.qlen & qstats.backlog

0xFFFFFFFF for n is really -1 once folded...

sch->q.qlen -= n;
sch->qstats.backlog -= len;
Toke Høiland-Jørgensen Jan. 8, 2019, 9:55 a.m. UTC | #2 
Eric Dumazet <eric.dumazet@gmail.com> writes:

> On 01/07/2019 11:47 AM, Toke Høiland-Jørgensen wrote:
>> From: Toke Høiland-Jørgensen <toke@redhat.com>
>> 
>> With GSO splitting in sch_cake, we can decrease as well as increase the
>> qlen. To make it possible to signal this up the qdisc tree, change
>> qdisc_tree_reduce_backlog() to accept signed integer values as the number
>> of packets and bytes to reduce the backlog by.
>
>
> Not sure why this patch is needed, given we use u32 integers
> for q.qlen & qstats.backlog
>
> 0xFFFFFFFF for n is really -1 once folded...
>
> sch->q.qlen -= n;
> sch->qstats.backlog -= len;

Sure, it works without the patch, but changing to int properly
communicates the intent (that negative values are allowed) rather than
rely on overflow behaviour...

-Toke
diff mbox series

Patch

diff --git a/include/net/sch_generic.h b/include/net/sch_generic.h
index 9481f2c142e2..7a4957599874 100644
--- a/include/net/sch_generic.h
+++ b/include/net/sch_generic.h
@@ -580,8 +580,7 @@  struct Qdisc *dev_graft_qdisc(struct netdev_queue *dev_queue,
 void qdisc_reset(struct Qdisc *qdisc);
 void qdisc_put(struct Qdisc *qdisc);
 void qdisc_put_unlocked(struct Qdisc *qdisc);
-void qdisc_tree_reduce_backlog(struct Qdisc *qdisc, unsigned int n,
-			       unsigned int len);
+void qdisc_tree_reduce_backlog(struct Qdisc *qdisc, int n, int len);
 #ifdef CONFIG_NET_SCHED
 int qdisc_offload_dump_helper(struct Qdisc *q, enum tc_setup_type type,
 			      void *type_data);
diff --git a/net/sched/sch_api.c b/net/sched/sch_api.c
index 7e4d1ccf4c87..03e26e8d0ec9 100644
--- a/net/sched/sch_api.c
+++ b/net/sched/sch_api.c
@@ -758,8 +758,7 @@  static u32 qdisc_alloc_handle(struct net_device *dev)
 	return 0;
 }
 
-void qdisc_tree_reduce_backlog(struct Qdisc *sch, unsigned int n,
-			       unsigned int len)
+void qdisc_tree_reduce_backlog(struct Qdisc *sch, int n, int len)
 {
 	bool qdisc_is_offloaded = sch->flags & TCQ_F_OFFLOADED;
 	const struct Qdisc_class_ops *cops;