linux/net/sched/sch_qfq.c

1540 lines
42 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0-only
/*
* net/sched/sch_qfq.c Quick Fair Queueing Plus Scheduler.
*
* Copyright (c) 2009 Fabio Checconi, Luigi Rizzo, and Paolo Valente.
* Copyright (c) 2012 Paolo Valente.
*/
#include <linux/module.h>
#include <linux/init.h>
#include <linux/bitops.h>
#include <linux/errno.h>
#include <linux/netdevice.h>
#include <linux/pkt_sched.h>
#include <net/sch_generic.h>
#include <net/pkt_sched.h>
#include <net/pkt_cls.h>
/* Quick Fair Queueing Plus
========================
Sources:
[1] Paolo Valente,
"Reducing the Execution Time of Fair-Queueing Schedulers."
http://algo.ing.unimo.it/people/paolo/agg-sched/agg-sched.pdf
Sources for QFQ:
[2] Fabio Checconi, Luigi Rizzo, and Paolo Valente: "QFQ: Efficient
Packet Scheduling with Tight Bandwidth Distribution Guarantees."
See also:
http://retis.sssup.it/~fabio/linux/qfq/
*/
/*
QFQ+ divides classes into aggregates of at most MAX_AGG_CLASSES
classes. Each aggregate is timestamped with a virtual start time S
and a virtual finish time F, and scheduled according to its
timestamps. S and F are computed as a function of a system virtual
time function V. The classes within each aggregate are instead
scheduled with DRR.
To speed up operations, QFQ+ divides also aggregates into a limited
number of groups. Which group a class belongs to depends on the
ratio between the maximum packet length for the class and the weight
of the class. Groups have their own S and F. In the end, QFQ+
schedules groups, then aggregates within groups, then classes within
aggregates. See [1] and [2] for a full description.
Virtual time computations.
S, F and V are all computed in fixed point arithmetic with
FRAC_BITS decimal bits.
QFQ_MAX_INDEX is the maximum index allowed for a group. We need
one bit per index.
QFQ_MAX_WSHIFT is the maximum power of two supported as a weight.
The layout of the bits is as below:
[ MTU_SHIFT ][ FRAC_BITS ]
[ MAX_INDEX ][ MIN_SLOT_SHIFT ]
^.__grp->index = 0
*.__grp->slot_shift
where MIN_SLOT_SHIFT is derived by difference from the others.
The max group index corresponds to Lmax/w_min, where
Lmax=1<<MTU_SHIFT, w_min = 1 .
From this, and knowing how many groups (MAX_INDEX) we want,
we can derive the shift corresponding to each group.
Because we often need to compute
F = S + len/w_i and V = V + len/wsum
instead of storing w_i store the value
inv_w = (1<<FRAC_BITS)/w_i
so we can do F = S + len * inv_w * wsum.
We use W_TOT in the formulas so we can easily move between
static and adaptive weight sum.
The per-scheduler-instance data contain all the data structures
for the scheduler: bitmaps and bucket lists.
*/
/*
* Maximum number of consecutive slots occupied by backlogged classes
* inside a group.
*/
#define QFQ_MAX_SLOTS 32
/*
* Shifts used for aggregate<->group mapping. We allow class weights that are
* in the range [1, 2^MAX_WSHIFT], and we try to map each aggregate i to the
* group with the smallest index that can support the L_i / r_i configured
* for the classes in the aggregate.
*
* grp->index is the index of the group; and grp->slot_shift
* is the shift for the corresponding (scaled) sigma_i.
*/
pkt_sched: enable QFQ to support TSO/GSO If the max packet size for some class (configured through tc) is violated by the actual size of the packets of that class, then QFQ would not schedule classes correctly, and the data structures implementing the bucket lists may get corrupted. This problem occurs with TSO/GSO even if the max packet size is set to the MTU, and is, e.g., the cause of the failure reported in [1]. Two patches have been proposed to solve this problem in [2], one of them is a preliminary version of this patch. This patch addresses the above issues by: 1) setting QFQ parameters to proper values for supporting TSO/GSO (in particular, setting the maximum possible packet size to 64KB), 2) automatically increasing the max packet size for a class, lmax, when a packet with a larger size than the current value of lmax arrives. The drawback of the first point is that the maximum weight for a class is now limited to 4096, which is equal to 1/16 of the maximum weight sum. Finally, this patch also forcibly caps the timestamps of a class if they are too high to be stored in the bucket list. This capping, taken from QFQ+ [3], handles the unfrequent case described in the comment to the function slot_insert. [1] http://marc.info/?l=linux-netdev&m=134968777902077&w=2 [2] http://marc.info/?l=linux-netdev&m=135096573507936&w=2 [3] http://marc.info/?l=linux-netdev&m=134902691421670&w=2 Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Tested-by: Cong Wang <amwang@redhat.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-11-05 20:29:24 +00:00
#define QFQ_MAX_INDEX 24
#define QFQ_MAX_WSHIFT 10
#define QFQ_MAX_WEIGHT (1<<QFQ_MAX_WSHIFT) /* see qfq_slot_insert */
#define QFQ_MAX_WSUM (64*QFQ_MAX_WEIGHT)
#define FRAC_BITS 30 /* fixed point arithmetic */
#define ONE_FP (1UL << FRAC_BITS)
pkt_sched: enable QFQ to support TSO/GSO If the max packet size for some class (configured through tc) is violated by the actual size of the packets of that class, then QFQ would not schedule classes correctly, and the data structures implementing the bucket lists may get corrupted. This problem occurs with TSO/GSO even if the max packet size is set to the MTU, and is, e.g., the cause of the failure reported in [1]. Two patches have been proposed to solve this problem in [2], one of them is a preliminary version of this patch. This patch addresses the above issues by: 1) setting QFQ parameters to proper values for supporting TSO/GSO (in particular, setting the maximum possible packet size to 64KB), 2) automatically increasing the max packet size for a class, lmax, when a packet with a larger size than the current value of lmax arrives. The drawback of the first point is that the maximum weight for a class is now limited to 4096, which is equal to 1/16 of the maximum weight sum. Finally, this patch also forcibly caps the timestamps of a class if they are too high to be stored in the bucket list. This capping, taken from QFQ+ [3], handles the unfrequent case described in the comment to the function slot_insert. [1] http://marc.info/?l=linux-netdev&m=134968777902077&w=2 [2] http://marc.info/?l=linux-netdev&m=135096573507936&w=2 [3] http://marc.info/?l=linux-netdev&m=134902691421670&w=2 Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Tested-by: Cong Wang <amwang@redhat.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-11-05 20:29:24 +00:00
#define QFQ_MTU_SHIFT 16 /* to support TSO/GSO */
#define QFQ_MIN_LMAX 512 /* see qfq_slot_insert */
#define QFQ_MAX_LMAX (1UL << QFQ_MTU_SHIFT)
#define QFQ_MAX_AGG_CLASSES 8 /* max num classes per aggregate allowed */
/*
* Possible group states. These values are used as indexes for the bitmaps
* array of struct qfq_queue.
*/
enum qfq_state { ER, IR, EB, IB, QFQ_MAX_STATE };
struct qfq_group;
struct qfq_aggregate;
struct qfq_class {
struct Qdisc_class_common common;
struct gnet_stats_basic_sync bstats;
struct gnet_stats_queue qstats;
struct net_rate_estimator __rcu *rate_est;
struct Qdisc *qdisc;
struct list_head alist; /* Link for active-classes list. */
struct qfq_aggregate *agg; /* Parent aggregate. */
int deficit; /* DRR deficit counter. */
};
struct qfq_aggregate {
struct hlist_node next; /* Link for the slot list. */
u64 S, F; /* flow timestamps (exact) */
/* group we belong to. In principle we would need the index,
* which is log_2(lmax/weight), but we never reference it
* directly, only the group.
*/
struct qfq_group *grp;
/* these are copied from the flowset. */
u32 class_weight; /* Weight of each class in this aggregate. */
/* Max pkt size for the classes in this aggregate, DRR quantum. */
int lmax;
u32 inv_w; /* ONE_FP/(sum of weights of classes in aggr.). */
u32 budgetmax; /* Max budget for this aggregate. */
u32 initial_budget, budget; /* Initial and current budget. */
int num_classes; /* Number of classes in this aggr. */
struct list_head active; /* DRR queue of active classes. */
struct hlist_node nonfull_next; /* See nonfull_aggs in qfq_sched. */
};
struct qfq_group {
u64 S, F; /* group timestamps (approx). */
unsigned int slot_shift; /* Slot shift. */
unsigned int index; /* Group index. */
unsigned int front; /* Index of the front slot. */
unsigned long full_slots; /* non-empty slots */
/* Array of RR lists of active aggregates. */
struct hlist_head slots[QFQ_MAX_SLOTS];
};
struct qfq_sched {
struct tcf_proto __rcu *filter_list;
struct tcf_block *block;
struct Qdisc_class_hash clhash;
u64 oldV, V; /* Precise virtual times. */
struct qfq_aggregate *in_serv_agg; /* Aggregate being served. */
u32 wsum; /* weight sum */
pkt_sched: sch_qfq: remove a source of high packet delay/jitter QFQ+ inherits from QFQ a design choice that may cause a high packet delay/jitter and a severe short-term unfairness. As QFQ, QFQ+ uses a special quantity, the system virtual time, to track the service provided by the ideal system it approximates. When a packet is dequeued, this quantity must be incremented by the size of the packet, divided by the sum of the weights of the aggregates waiting to be served. Tracking this sum correctly is a non-trivial task, because, to preserve tight service guarantees, the decrement of this sum must be delayed in a special way [1]: this sum can be decremented only after that its value would decrease also in the ideal system approximated by QFQ+. For efficiency, QFQ+ keeps track only of the 'instantaneous' weight sum, increased and decreased immediately as the weight of an aggregate changes, and as an aggregate is created or destroyed (which, in its turn, happens as a consequence of some class being created/destroyed/changed). However, to avoid the problems caused to service guarantees by these immediate decreases, QFQ+ increments the system virtual time using the maximum value allowed for the weight sum, 2^10, in place of the dynamic, instantaneous value. The instantaneous value of the weight sum is used only to check whether a request of weight increase or a class creation can be satisfied. Unfortunately, the problems caused by this choice are worse than the temporary degradation of the service guarantees that may occur, when a class is changed or destroyed, if the instantaneous value of the weight sum was used to update the system virtual time. In fact, the fraction of the link bandwidth guaranteed by QFQ+ to each aggregate is equal to the ratio between the weight of the aggregate and the sum of the weights of the competing aggregates. The packet delay guaranteed to the aggregate is instead inversely proportional to the guaranteed bandwidth. By using the maximum possible value, and not the actual value of the weight sum, QFQ+ provides each aggregate with the worst possible service guarantees, and not with service guarantees related to the actual set of competing aggregates. To see the consequences of this fact, consider the following simple example. Suppose that only the following aggregates are backlogged, i.e., that only the classes in the following aggregates have packets to transmit: one aggregate with weight 10, say A, and ten aggregates with weight 1, say B1, B2, ..., B10. In particular, suppose that these aggregates are always backlogged. Given the weight distribution, the smoothest and fairest service order would be: A B1 A B2 A B3 A B4 A B5 A B6 A B7 A B8 A B9 A B10 A B1 A B2 ... QFQ+ would provide exactly this optimal service if it used the actual value for the weight sum instead of the maximum possible value, i.e., 11 instead of 2^10. In contrast, since QFQ+ uses the latter value, it serves aggregates as follows (easy to prove and to reproduce experimentally): A B1 B2 B3 B4 B5 B6 B7 B8 B9 B10 A A A A A A A A A A B1 B2 ... B10 A A ... By replacing 10 with N in the above example, and by increasing N, one can increase at will the maximum packet delay and the jitter experienced by the classes in aggregate A. This patch addresses this issue by just using the above 'instantaneous' value of the weight sum, instead of the maximum possible value, when updating the system virtual time. After the instantaneous weight sum is decreased, QFQ+ may deviate from the ideal service for a time interval in the order of the time to serve one maximum-size packet for each backlogged class. The worst-case extent of the deviation exhibited by QFQ+ during this time interval [1] is basically the same as of the deviation described above (but, without this patch, QFQ+ suffers from such a deviation all the time). Finally, this patch modifies the comment to the function qfq_slot_insert, to make it coherent with the fact that the weight sum used by QFQ+ can now be lower than the maximum possible value. [1] P. Valente, "Extending WF2Q+ to support a dynamic traffic mix", Proceedings of AAA-IDEA'05, June 2005. Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-07-16 06:52:30 +00:00
u32 iwsum; /* inverse weight sum */
unsigned long bitmaps[QFQ_MAX_STATE]; /* Group bitmaps. */
struct qfq_group groups[QFQ_MAX_INDEX + 1]; /* The groups. */
u32 min_slot_shift; /* Index of the group-0 bit in the bitmaps. */
u32 max_agg_classes; /* Max number of classes per aggr. */
struct hlist_head nonfull_aggs; /* Aggs with room for more classes. */
};
/*
* Possible reasons why the timestamps of an aggregate are updated
* enqueue: the aggregate switches from idle to active and must scheduled
* for service
* requeue: the aggregate finishes its budget, so it stops being served and
* must be rescheduled for service
*/
enum update_reason {enqueue, requeue};
static struct qfq_class *qfq_find_class(struct Qdisc *sch, u32 classid)
{
struct qfq_sched *q = qdisc_priv(sch);
struct Qdisc_class_common *clc;
clc = qdisc_class_find(&q->clhash, classid);
if (clc == NULL)
return NULL;
return container_of(clc, struct qfq_class, common);
}
static const struct netlink_range_validation lmax_range = {
.min = QFQ_MIN_LMAX,
.max = QFQ_MAX_LMAX,
};
static const struct nla_policy qfq_policy[TCA_QFQ_MAX + 1] = {
[TCA_QFQ_WEIGHT] = NLA_POLICY_RANGE(NLA_U32, 1, QFQ_MAX_WEIGHT),
[TCA_QFQ_LMAX] = NLA_POLICY_FULL_RANGE(NLA_U32, &lmax_range),
};
/*
* Calculate a flow index, given its weight and maximum packet length.
* index = log_2(maxlen/weight) but we need to apply the scaling.
* This is used only once at flow creation.
*/
static int qfq_calc_index(u32 inv_w, unsigned int maxlen, u32 min_slot_shift)
{
u64 slot_size = (u64)maxlen * inv_w;
unsigned long size_map;
int index = 0;
size_map = slot_size >> min_slot_shift;
if (!size_map)
goto out;
index = __fls(size_map) + 1; /* basically a log_2 */
index -= !(slot_size - (1ULL << (index + min_slot_shift - 1)));
if (index < 0)
index = 0;
out:
pr_debug("qfq calc_index: W = %lu, L = %u, I = %d\n",
(unsigned long) ONE_FP/inv_w, maxlen, index);
return index;
}
static void qfq_deactivate_agg(struct qfq_sched *, struct qfq_aggregate *);
static void qfq_activate_agg(struct qfq_sched *, struct qfq_aggregate *,
enum update_reason);
static void qfq_init_agg(struct qfq_sched *q, struct qfq_aggregate *agg,
u32 lmax, u32 weight)
{
INIT_LIST_HEAD(&agg->active);
hlist_add_head(&agg->nonfull_next, &q->nonfull_aggs);
agg->lmax = lmax;
agg->class_weight = weight;
}
static struct qfq_aggregate *qfq_find_agg(struct qfq_sched *q,
u32 lmax, u32 weight)
{
struct qfq_aggregate *agg;
hlist: drop the node parameter from iterators I'm not sure why, but the hlist for each entry iterators were conceived list_for_each_entry(pos, head, member) The hlist ones were greedy and wanted an extra parameter: hlist_for_each_entry(tpos, pos, head, member) Why did they need an extra pos parameter? I'm not quite sure. Not only they don't really need it, it also prevents the iterator from looking exactly like the list iterator, which is unfortunate. Besides the semantic patch, there was some manual work required: - Fix up the actual hlist iterators in linux/list.h - Fix up the declaration of other iterators based on the hlist ones. - A very small amount of places were using the 'node' parameter, this was modified to use 'obj->member' instead. - Coccinelle didn't handle the hlist_for_each_entry_safe iterator properly, so those had to be fixed up manually. The semantic patch which is mostly the work of Peter Senna Tschudin is here: @@ iterator name hlist_for_each_entry, hlist_for_each_entry_continue, hlist_for_each_entry_from, hlist_for_each_entry_rcu, hlist_for_each_entry_rcu_bh, hlist_for_each_entry_continue_rcu_bh, for_each_busy_worker, ax25_uid_for_each, ax25_for_each, inet_bind_bucket_for_each, sctp_for_each_hentry, sk_for_each, sk_for_each_rcu, sk_for_each_from, sk_for_each_safe, sk_for_each_bound, hlist_for_each_entry_safe, hlist_for_each_entry_continue_rcu, nr_neigh_for_each, nr_neigh_for_each_safe, nr_node_for_each, nr_node_for_each_safe, for_each_gfn_indirect_valid_sp, for_each_gfn_sp, for_each_host; type T; expression a,c,d,e; identifier b; statement S; @@ -T b; <+... when != b ( hlist_for_each_entry(a, - b, c, d) S | hlist_for_each_entry_continue(a, - b, c) S | hlist_for_each_entry_from(a, - b, c) S | hlist_for_each_entry_rcu(a, - b, c, d) S | hlist_for_each_entry_rcu_bh(a, - b, c, d) S | hlist_for_each_entry_continue_rcu_bh(a, - b, c) S | for_each_busy_worker(a, c, - b, d) S | ax25_uid_for_each(a, - b, c) S | ax25_for_each(a, - b, c) S | inet_bind_bucket_for_each(a, - b, c) S | sctp_for_each_hentry(a, - b, c) S | sk_for_each(a, - b, c) S | sk_for_each_rcu(a, - b, c) S | sk_for_each_from -(a, b) +(a) S + sk_for_each_from(a) S | sk_for_each_safe(a, - b, c, d) S | sk_for_each_bound(a, - b, c) S | hlist_for_each_entry_safe(a, - b, c, d, e) S | hlist_for_each_entry_continue_rcu(a, - b, c) S | nr_neigh_for_each(a, - b, c) S | nr_neigh_for_each_safe(a, - b, c, d) S | nr_node_for_each(a, - b, c) S | nr_node_for_each_safe(a, - b, c, d) S | - for_each_gfn_sp(a, c, d, b) S + for_each_gfn_sp(a, c, d) S | - for_each_gfn_indirect_valid_sp(a, c, d, b) S + for_each_gfn_indirect_valid_sp(a, c, d) S | for_each_host(a, - b, c) S | for_each_host_safe(a, - b, c, d) S | for_each_mesh_entry(a, - b, c, d) S ) ...+> [akpm@linux-foundation.org: drop bogus change from net/ipv4/raw.c] [akpm@linux-foundation.org: drop bogus hunk from net/ipv6/raw.c] [akpm@linux-foundation.org: checkpatch fixes] [akpm@linux-foundation.org: fix warnings] [akpm@linux-foudnation.org: redo intrusive kvm changes] Tested-by: Peter Senna Tschudin <peter.senna@gmail.com> Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com> Cc: Wu Fengguang <fengguang.wu@intel.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Gleb Natapov <gleb@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-02-28 01:06:00 +00:00
hlist_for_each_entry(agg, &q->nonfull_aggs, nonfull_next)
if (agg->lmax == lmax && agg->class_weight == weight)
return agg;
return NULL;
}
/* Update aggregate as a function of the new number of classes. */
static void qfq_update_agg(struct qfq_sched *q, struct qfq_aggregate *agg,
int new_num_classes)
{
u32 new_agg_weight;
if (new_num_classes == q->max_agg_classes)
hlist_del_init(&agg->nonfull_next);
if (agg->num_classes > new_num_classes &&
new_num_classes == q->max_agg_classes - 1) /* agg no more full */
hlist_add_head(&agg->nonfull_next, &q->nonfull_aggs);
pkt_sched: sch_qfq: properly cap timestamps in charge_actual_service QFQ+ schedules the active aggregates in a group using a bucket list (one list per group). The bucket in which each aggregate is inserted depends on the aggregate's timestamps, and the number of buckets in a group is enough to accomodate the possible (range of) values of the timestamps of all the aggregates in the group. For this property to hold, timestamps must however be computed correctly. One necessary condition for computing timestamps correctly is that the number of bits dequeued for each aggregate, while the aggregate is in service, does not exceed the maximum budget budgetmax assigned to the aggregate. For each aggregate, budgetmax is proportional to the number of classes in the aggregate. If the number of classes of the aggregate is decreased through qfq_change_class(), then budgetmax is decreased automatically as well. Problems may occur if the aggregate is in service when budgetmax is decreased, because the current remaining budget of the aggregate and/or the service already received by the aggregate may happen to be larger than the new value of budgetmax. In this case, when the aggregate is eventually deselected and its timestamps are updated, the aggregate may happen to have received an amount of service larger than budgetmax. This may cause the aggregate to be assigned a higher virtual finish time than the maximum acceptable value for the last bucket in the bucket list of the group. This fix introduces a cap that addresses this issue. Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Reviewed-by: Fabio Checconi <fchecconi@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-03-05 08:04:57 +00:00
/* The next assignment may let
* agg->initial_budget > agg->budgetmax
* hold, we will take it into account in charge_actual_service().
*/
agg->budgetmax = new_num_classes * agg->lmax;
new_agg_weight = agg->class_weight * new_num_classes;
agg->inv_w = ONE_FP/new_agg_weight;
if (agg->grp == NULL) {
int i = qfq_calc_index(agg->inv_w, agg->budgetmax,
q->min_slot_shift);
agg->grp = &q->groups[i];
}
q->wsum +=
(int) agg->class_weight * (new_num_classes - agg->num_classes);
pkt_sched: sch_qfq: remove a source of high packet delay/jitter QFQ+ inherits from QFQ a design choice that may cause a high packet delay/jitter and a severe short-term unfairness. As QFQ, QFQ+ uses a special quantity, the system virtual time, to track the service provided by the ideal system it approximates. When a packet is dequeued, this quantity must be incremented by the size of the packet, divided by the sum of the weights of the aggregates waiting to be served. Tracking this sum correctly is a non-trivial task, because, to preserve tight service guarantees, the decrement of this sum must be delayed in a special way [1]: this sum can be decremented only after that its value would decrease also in the ideal system approximated by QFQ+. For efficiency, QFQ+ keeps track only of the 'instantaneous' weight sum, increased and decreased immediately as the weight of an aggregate changes, and as an aggregate is created or destroyed (which, in its turn, happens as a consequence of some class being created/destroyed/changed). However, to avoid the problems caused to service guarantees by these immediate decreases, QFQ+ increments the system virtual time using the maximum value allowed for the weight sum, 2^10, in place of the dynamic, instantaneous value. The instantaneous value of the weight sum is used only to check whether a request of weight increase or a class creation can be satisfied. Unfortunately, the problems caused by this choice are worse than the temporary degradation of the service guarantees that may occur, when a class is changed or destroyed, if the instantaneous value of the weight sum was used to update the system virtual time. In fact, the fraction of the link bandwidth guaranteed by QFQ+ to each aggregate is equal to the ratio between the weight of the aggregate and the sum of the weights of the competing aggregates. The packet delay guaranteed to the aggregate is instead inversely proportional to the guaranteed bandwidth. By using the maximum possible value, and not the actual value of the weight sum, QFQ+ provides each aggregate with the worst possible service guarantees, and not with service guarantees related to the actual set of competing aggregates. To see the consequences of this fact, consider the following simple example. Suppose that only the following aggregates are backlogged, i.e., that only the classes in the following aggregates have packets to transmit: one aggregate with weight 10, say A, and ten aggregates with weight 1, say B1, B2, ..., B10. In particular, suppose that these aggregates are always backlogged. Given the weight distribution, the smoothest and fairest service order would be: A B1 A B2 A B3 A B4 A B5 A B6 A B7 A B8 A B9 A B10 A B1 A B2 ... QFQ+ would provide exactly this optimal service if it used the actual value for the weight sum instead of the maximum possible value, i.e., 11 instead of 2^10. In contrast, since QFQ+ uses the latter value, it serves aggregates as follows (easy to prove and to reproduce experimentally): A B1 B2 B3 B4 B5 B6 B7 B8 B9 B10 A A A A A A A A A A B1 B2 ... B10 A A ... By replacing 10 with N in the above example, and by increasing N, one can increase at will the maximum packet delay and the jitter experienced by the classes in aggregate A. This patch addresses this issue by just using the above 'instantaneous' value of the weight sum, instead of the maximum possible value, when updating the system virtual time. After the instantaneous weight sum is decreased, QFQ+ may deviate from the ideal service for a time interval in the order of the time to serve one maximum-size packet for each backlogged class. The worst-case extent of the deviation exhibited by QFQ+ during this time interval [1] is basically the same as of the deviation described above (but, without this patch, QFQ+ suffers from such a deviation all the time). Finally, this patch modifies the comment to the function qfq_slot_insert, to make it coherent with the fact that the weight sum used by QFQ+ can now be lower than the maximum possible value. [1] P. Valente, "Extending WF2Q+ to support a dynamic traffic mix", Proceedings of AAA-IDEA'05, June 2005. Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-07-16 06:52:30 +00:00
q->iwsum = ONE_FP / q->wsum;
agg->num_classes = new_num_classes;
}
/* Add class to aggregate. */
static void qfq_add_to_agg(struct qfq_sched *q,
struct qfq_aggregate *agg,
struct qfq_class *cl)
{
cl->agg = agg;
qfq_update_agg(q, agg, agg->num_classes+1);
if (cl->qdisc->q.qlen > 0) { /* adding an active class */
list_add_tail(&cl->alist, &agg->active);
if (list_first_entry(&agg->active, struct qfq_class, alist) ==
cl && q->in_serv_agg != agg) /* agg was inactive */
qfq_activate_agg(q, agg, enqueue); /* schedule agg */
}
}
static struct qfq_aggregate *qfq_choose_next_agg(struct qfq_sched *);
static void qfq_destroy_agg(struct qfq_sched *q, struct qfq_aggregate *agg)
{
hlist_del_init(&agg->nonfull_next);
pkt_sched: sch_qfq: remove a source of high packet delay/jitter QFQ+ inherits from QFQ a design choice that may cause a high packet delay/jitter and a severe short-term unfairness. As QFQ, QFQ+ uses a special quantity, the system virtual time, to track the service provided by the ideal system it approximates. When a packet is dequeued, this quantity must be incremented by the size of the packet, divided by the sum of the weights of the aggregates waiting to be served. Tracking this sum correctly is a non-trivial task, because, to preserve tight service guarantees, the decrement of this sum must be delayed in a special way [1]: this sum can be decremented only after that its value would decrease also in the ideal system approximated by QFQ+. For efficiency, QFQ+ keeps track only of the 'instantaneous' weight sum, increased and decreased immediately as the weight of an aggregate changes, and as an aggregate is created or destroyed (which, in its turn, happens as a consequence of some class being created/destroyed/changed). However, to avoid the problems caused to service guarantees by these immediate decreases, QFQ+ increments the system virtual time using the maximum value allowed for the weight sum, 2^10, in place of the dynamic, instantaneous value. The instantaneous value of the weight sum is used only to check whether a request of weight increase or a class creation can be satisfied. Unfortunately, the problems caused by this choice are worse than the temporary degradation of the service guarantees that may occur, when a class is changed or destroyed, if the instantaneous value of the weight sum was used to update the system virtual time. In fact, the fraction of the link bandwidth guaranteed by QFQ+ to each aggregate is equal to the ratio between the weight of the aggregate and the sum of the weights of the competing aggregates. The packet delay guaranteed to the aggregate is instead inversely proportional to the guaranteed bandwidth. By using the maximum possible value, and not the actual value of the weight sum, QFQ+ provides each aggregate with the worst possible service guarantees, and not with service guarantees related to the actual set of competing aggregates. To see the consequences of this fact, consider the following simple example. Suppose that only the following aggregates are backlogged, i.e., that only the classes in the following aggregates have packets to transmit: one aggregate with weight 10, say A, and ten aggregates with weight 1, say B1, B2, ..., B10. In particular, suppose that these aggregates are always backlogged. Given the weight distribution, the smoothest and fairest service order would be: A B1 A B2 A B3 A B4 A B5 A B6 A B7 A B8 A B9 A B10 A B1 A B2 ... QFQ+ would provide exactly this optimal service if it used the actual value for the weight sum instead of the maximum possible value, i.e., 11 instead of 2^10. In contrast, since QFQ+ uses the latter value, it serves aggregates as follows (easy to prove and to reproduce experimentally): A B1 B2 B3 B4 B5 B6 B7 B8 B9 B10 A A A A A A A A A A B1 B2 ... B10 A A ... By replacing 10 with N in the above example, and by increasing N, one can increase at will the maximum packet delay and the jitter experienced by the classes in aggregate A. This patch addresses this issue by just using the above 'instantaneous' value of the weight sum, instead of the maximum possible value, when updating the system virtual time. After the instantaneous weight sum is decreased, QFQ+ may deviate from the ideal service for a time interval in the order of the time to serve one maximum-size packet for each backlogged class. The worst-case extent of the deviation exhibited by QFQ+ during this time interval [1] is basically the same as of the deviation described above (but, without this patch, QFQ+ suffers from such a deviation all the time). Finally, this patch modifies the comment to the function qfq_slot_insert, to make it coherent with the fact that the weight sum used by QFQ+ can now be lower than the maximum possible value. [1] P. Valente, "Extending WF2Q+ to support a dynamic traffic mix", Proceedings of AAA-IDEA'05, June 2005. Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-07-16 06:52:30 +00:00
q->wsum -= agg->class_weight;
if (q->wsum != 0)
q->iwsum = ONE_FP / q->wsum;
if (q->in_serv_agg == agg)
q->in_serv_agg = qfq_choose_next_agg(q);
kfree(agg);
}
/* Deschedule class from within its parent aggregate. */
static void qfq_deactivate_class(struct qfq_sched *q, struct qfq_class *cl)
{
struct qfq_aggregate *agg = cl->agg;
list_del(&cl->alist); /* remove from RR queue of the aggregate */
if (list_empty(&agg->active)) /* agg is now inactive */
qfq_deactivate_agg(q, agg);
}
/* Remove class from its parent aggregate. */
static void qfq_rm_from_agg(struct qfq_sched *q, struct qfq_class *cl)
pkt_sched: enable QFQ to support TSO/GSO If the max packet size for some class (configured through tc) is violated by the actual size of the packets of that class, then QFQ would not schedule classes correctly, and the data structures implementing the bucket lists may get corrupted. This problem occurs with TSO/GSO even if the max packet size is set to the MTU, and is, e.g., the cause of the failure reported in [1]. Two patches have been proposed to solve this problem in [2], one of them is a preliminary version of this patch. This patch addresses the above issues by: 1) setting QFQ parameters to proper values for supporting TSO/GSO (in particular, setting the maximum possible packet size to 64KB), 2) automatically increasing the max packet size for a class, lmax, when a packet with a larger size than the current value of lmax arrives. The drawback of the first point is that the maximum weight for a class is now limited to 4096, which is equal to 1/16 of the maximum weight sum. Finally, this patch also forcibly caps the timestamps of a class if they are too high to be stored in the bucket list. This capping, taken from QFQ+ [3], handles the unfrequent case described in the comment to the function slot_insert. [1] http://marc.info/?l=linux-netdev&m=134968777902077&w=2 [2] http://marc.info/?l=linux-netdev&m=135096573507936&w=2 [3] http://marc.info/?l=linux-netdev&m=134902691421670&w=2 Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Tested-by: Cong Wang <amwang@redhat.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-11-05 20:29:24 +00:00
{
struct qfq_aggregate *agg = cl->agg;
pkt_sched: enable QFQ to support TSO/GSO If the max packet size for some class (configured through tc) is violated by the actual size of the packets of that class, then QFQ would not schedule classes correctly, and the data structures implementing the bucket lists may get corrupted. This problem occurs with TSO/GSO even if the max packet size is set to the MTU, and is, e.g., the cause of the failure reported in [1]. Two patches have been proposed to solve this problem in [2], one of them is a preliminary version of this patch. This patch addresses the above issues by: 1) setting QFQ parameters to proper values for supporting TSO/GSO (in particular, setting the maximum possible packet size to 64KB), 2) automatically increasing the max packet size for a class, lmax, when a packet with a larger size than the current value of lmax arrives. The drawback of the first point is that the maximum weight for a class is now limited to 4096, which is equal to 1/16 of the maximum weight sum. Finally, this patch also forcibly caps the timestamps of a class if they are too high to be stored in the bucket list. This capping, taken from QFQ+ [3], handles the unfrequent case described in the comment to the function slot_insert. [1] http://marc.info/?l=linux-netdev&m=134968777902077&w=2 [2] http://marc.info/?l=linux-netdev&m=135096573507936&w=2 [3] http://marc.info/?l=linux-netdev&m=134902691421670&w=2 Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Tested-by: Cong Wang <amwang@redhat.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-11-05 20:29:24 +00:00
cl->agg = NULL;
if (agg->num_classes == 1) { /* agg being emptied, destroy it */
qfq_destroy_agg(q, agg);
return;
pkt_sched: enable QFQ to support TSO/GSO If the max packet size for some class (configured through tc) is violated by the actual size of the packets of that class, then QFQ would not schedule classes correctly, and the data structures implementing the bucket lists may get corrupted. This problem occurs with TSO/GSO even if the max packet size is set to the MTU, and is, e.g., the cause of the failure reported in [1]. Two patches have been proposed to solve this problem in [2], one of them is a preliminary version of this patch. This patch addresses the above issues by: 1) setting QFQ parameters to proper values for supporting TSO/GSO (in particular, setting the maximum possible packet size to 64KB), 2) automatically increasing the max packet size for a class, lmax, when a packet with a larger size than the current value of lmax arrives. The drawback of the first point is that the maximum weight for a class is now limited to 4096, which is equal to 1/16 of the maximum weight sum. Finally, this patch also forcibly caps the timestamps of a class if they are too high to be stored in the bucket list. This capping, taken from QFQ+ [3], handles the unfrequent case described in the comment to the function slot_insert. [1] http://marc.info/?l=linux-netdev&m=134968777902077&w=2 [2] http://marc.info/?l=linux-netdev&m=135096573507936&w=2 [3] http://marc.info/?l=linux-netdev&m=134902691421670&w=2 Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Tested-by: Cong Wang <amwang@redhat.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-11-05 20:29:24 +00:00
}
qfq_update_agg(q, agg, agg->num_classes-1);
}
pkt_sched: enable QFQ to support TSO/GSO If the max packet size for some class (configured through tc) is violated by the actual size of the packets of that class, then QFQ would not schedule classes correctly, and the data structures implementing the bucket lists may get corrupted. This problem occurs with TSO/GSO even if the max packet size is set to the MTU, and is, e.g., the cause of the failure reported in [1]. Two patches have been proposed to solve this problem in [2], one of them is a preliminary version of this patch. This patch addresses the above issues by: 1) setting QFQ parameters to proper values for supporting TSO/GSO (in particular, setting the maximum possible packet size to 64KB), 2) automatically increasing the max packet size for a class, lmax, when a packet with a larger size than the current value of lmax arrives. The drawback of the first point is that the maximum weight for a class is now limited to 4096, which is equal to 1/16 of the maximum weight sum. Finally, this patch also forcibly caps the timestamps of a class if they are too high to be stored in the bucket list. This capping, taken from QFQ+ [3], handles the unfrequent case described in the comment to the function slot_insert. [1] http://marc.info/?l=linux-netdev&m=134968777902077&w=2 [2] http://marc.info/?l=linux-netdev&m=135096573507936&w=2 [3] http://marc.info/?l=linux-netdev&m=134902691421670&w=2 Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Tested-by: Cong Wang <amwang@redhat.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-11-05 20:29:24 +00:00
/* Deschedule class and remove it from its parent aggregate. */
static void qfq_deact_rm_from_agg(struct qfq_sched *q, struct qfq_class *cl)
{
if (cl->qdisc->q.qlen > 0) /* class is active */
qfq_deactivate_class(q, cl);
pkt_sched: enable QFQ to support TSO/GSO If the max packet size for some class (configured through tc) is violated by the actual size of the packets of that class, then QFQ would not schedule classes correctly, and the data structures implementing the bucket lists may get corrupted. This problem occurs with TSO/GSO even if the max packet size is set to the MTU, and is, e.g., the cause of the failure reported in [1]. Two patches have been proposed to solve this problem in [2], one of them is a preliminary version of this patch. This patch addresses the above issues by: 1) setting QFQ parameters to proper values for supporting TSO/GSO (in particular, setting the maximum possible packet size to 64KB), 2) automatically increasing the max packet size for a class, lmax, when a packet with a larger size than the current value of lmax arrives. The drawback of the first point is that the maximum weight for a class is now limited to 4096, which is equal to 1/16 of the maximum weight sum. Finally, this patch also forcibly caps the timestamps of a class if they are too high to be stored in the bucket list. This capping, taken from QFQ+ [3], handles the unfrequent case described in the comment to the function slot_insert. [1] http://marc.info/?l=linux-netdev&m=134968777902077&w=2 [2] http://marc.info/?l=linux-netdev&m=135096573507936&w=2 [3] http://marc.info/?l=linux-netdev&m=134902691421670&w=2 Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Tested-by: Cong Wang <amwang@redhat.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-11-05 20:29:24 +00:00
qfq_rm_from_agg(q, cl);
pkt_sched: enable QFQ to support TSO/GSO If the max packet size for some class (configured through tc) is violated by the actual size of the packets of that class, then QFQ would not schedule classes correctly, and the data structures implementing the bucket lists may get corrupted. This problem occurs with TSO/GSO even if the max packet size is set to the MTU, and is, e.g., the cause of the failure reported in [1]. Two patches have been proposed to solve this problem in [2], one of them is a preliminary version of this patch. This patch addresses the above issues by: 1) setting QFQ parameters to proper values for supporting TSO/GSO (in particular, setting the maximum possible packet size to 64KB), 2) automatically increasing the max packet size for a class, lmax, when a packet with a larger size than the current value of lmax arrives. The drawback of the first point is that the maximum weight for a class is now limited to 4096, which is equal to 1/16 of the maximum weight sum. Finally, this patch also forcibly caps the timestamps of a class if they are too high to be stored in the bucket list. This capping, taken from QFQ+ [3], handles the unfrequent case described in the comment to the function slot_insert. [1] http://marc.info/?l=linux-netdev&m=134968777902077&w=2 [2] http://marc.info/?l=linux-netdev&m=135096573507936&w=2 [3] http://marc.info/?l=linux-netdev&m=134902691421670&w=2 Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Tested-by: Cong Wang <amwang@redhat.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-11-05 20:29:24 +00:00
}
/* Move class to a new aggregate, matching the new class weight and/or lmax */
static int qfq_change_agg(struct Qdisc *sch, struct qfq_class *cl, u32 weight,
u32 lmax)
{
struct qfq_sched *q = qdisc_priv(sch);
net/sched: sch_qfq: account for stab overhead in qfq_enqueue Lion says: ------- In the QFQ scheduler a similar issue to CVE-2023-31436 persists. Consider the following code in net/sched/sch_qfq.c: static int qfq_enqueue(struct sk_buff *skb, struct Qdisc *sch, struct sk_buff **to_free) { unsigned int len = qdisc_pkt_len(skb), gso_segs; // ... if (unlikely(cl->agg->lmax < len)) { pr_debug("qfq: increasing maxpkt from %u to %u for class %u", cl->agg->lmax, len, cl->common.classid); err = qfq_change_agg(sch, cl, cl->agg->class_weight, len); if (err) { cl->qstats.drops++; return qdisc_drop(skb, sch, to_free); } // ... } Similarly to CVE-2023-31436, "lmax" is increased without any bounds checks according to the packet length "len". Usually this would not impose a problem because packet sizes are naturally limited. This is however not the actual packet length, rather the "qdisc_pkt_len(skb)" which might apply size transformations according to "struct qdisc_size_table" as created by "qdisc_get_stab()" in net/sched/sch_api.c if the TCA_STAB option was set when modifying the qdisc. A user may choose virtually any size using such a table. As a result the same issue as in CVE-2023-31436 can occur, allowing heap out-of-bounds read / writes in the kmalloc-8192 cache. ------- We can create the issue with the following commands: tc qdisc add dev $DEV root handle 1: stab mtu 2048 tsize 512 mpu 0 \ overhead 999999999 linklayer ethernet qfq tc class add dev $DEV parent 1: classid 1:1 htb rate 6mbit burst 15k tc filter add dev $DEV parent 1: matchall classid 1:1 ping -I $DEV 1.1.1.2 This is caused by incorrectly assuming that qdisc_pkt_len() returns a length within the QFQ_MIN_LMAX < len < QFQ_MAX_LMAX. Fixes: 462dbc9101ac ("pkt_sched: QFQ Plus: fair-queueing service at DRR cost") Reported-by: Lion <nnamrec@gmail.com> Reviewed-by: Eric Dumazet <edumazet@google.com> Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com> Signed-off-by: Pedro Tammela <pctammela@mojatatu.com> Reviewed-by: Simon Horman <simon.horman@corigine.com> Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2023-07-11 21:01:02 +00:00
struct qfq_aggregate *new_agg;
net/sched: sch_qfq: account for stab overhead in qfq_enqueue Lion says: ------- In the QFQ scheduler a similar issue to CVE-2023-31436 persists. Consider the following code in net/sched/sch_qfq.c: static int qfq_enqueue(struct sk_buff *skb, struct Qdisc *sch, struct sk_buff **to_free) { unsigned int len = qdisc_pkt_len(skb), gso_segs; // ... if (unlikely(cl->agg->lmax < len)) { pr_debug("qfq: increasing maxpkt from %u to %u for class %u", cl->agg->lmax, len, cl->common.classid); err = qfq_change_agg(sch, cl, cl->agg->class_weight, len); if (err) { cl->qstats.drops++; return qdisc_drop(skb, sch, to_free); } // ... } Similarly to CVE-2023-31436, "lmax" is increased without any bounds checks according to the packet length "len". Usually this would not impose a problem because packet sizes are naturally limited. This is however not the actual packet length, rather the "qdisc_pkt_len(skb)" which might apply size transformations according to "struct qdisc_size_table" as created by "qdisc_get_stab()" in net/sched/sch_api.c if the TCA_STAB option was set when modifying the qdisc. A user may choose virtually any size using such a table. As a result the same issue as in CVE-2023-31436 can occur, allowing heap out-of-bounds read / writes in the kmalloc-8192 cache. ------- We can create the issue with the following commands: tc qdisc add dev $DEV root handle 1: stab mtu 2048 tsize 512 mpu 0 \ overhead 999999999 linklayer ethernet qfq tc class add dev $DEV parent 1: classid 1:1 htb rate 6mbit burst 15k tc filter add dev $DEV parent 1: matchall classid 1:1 ping -I $DEV 1.1.1.2 This is caused by incorrectly assuming that qdisc_pkt_len() returns a length within the QFQ_MIN_LMAX < len < QFQ_MAX_LMAX. Fixes: 462dbc9101ac ("pkt_sched: QFQ Plus: fair-queueing service at DRR cost") Reported-by: Lion <nnamrec@gmail.com> Reviewed-by: Eric Dumazet <edumazet@google.com> Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com> Signed-off-by: Pedro Tammela <pctammela@mojatatu.com> Reviewed-by: Simon Horman <simon.horman@corigine.com> Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2023-07-11 21:01:02 +00:00
/* 'lmax' can range from [QFQ_MIN_LMAX, pktlen + stab overhead] */
if (lmax > QFQ_MAX_LMAX)
return -EINVAL;
new_agg = qfq_find_agg(q, lmax, weight);
if (new_agg == NULL) { /* create new aggregate */
new_agg = kzalloc(sizeof(*new_agg), GFP_ATOMIC);
if (new_agg == NULL)
return -ENOBUFS;
qfq_init_agg(q, new_agg, lmax, weight);
}
qfq_deact_rm_from_agg(q, cl);
qfq_add_to_agg(q, new_agg, cl);
return 0;
}
pkt_sched: enable QFQ to support TSO/GSO If the max packet size for some class (configured through tc) is violated by the actual size of the packets of that class, then QFQ would not schedule classes correctly, and the data structures implementing the bucket lists may get corrupted. This problem occurs with TSO/GSO even if the max packet size is set to the MTU, and is, e.g., the cause of the failure reported in [1]. Two patches have been proposed to solve this problem in [2], one of them is a preliminary version of this patch. This patch addresses the above issues by: 1) setting QFQ parameters to proper values for supporting TSO/GSO (in particular, setting the maximum possible packet size to 64KB), 2) automatically increasing the max packet size for a class, lmax, when a packet with a larger size than the current value of lmax arrives. The drawback of the first point is that the maximum weight for a class is now limited to 4096, which is equal to 1/16 of the maximum weight sum. Finally, this patch also forcibly caps the timestamps of a class if they are too high to be stored in the bucket list. This capping, taken from QFQ+ [3], handles the unfrequent case described in the comment to the function slot_insert. [1] http://marc.info/?l=linux-netdev&m=134968777902077&w=2 [2] http://marc.info/?l=linux-netdev&m=135096573507936&w=2 [3] http://marc.info/?l=linux-netdev&m=134902691421670&w=2 Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Tested-by: Cong Wang <amwang@redhat.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-11-05 20:29:24 +00:00
static int qfq_change_class(struct Qdisc *sch, u32 classid, u32 parentid,
struct nlattr **tca, unsigned long *arg,
struct netlink_ext_ack *extack)
{
struct qfq_sched *q = qdisc_priv(sch);
struct qfq_class *cl = (struct qfq_class *)*arg;
bool existing = false;
struct nlattr *tb[TCA_QFQ_MAX + 1];
struct qfq_aggregate *new_agg = NULL;
u32 weight, lmax, inv_w;
pkt_sched: enable QFQ to support TSO/GSO If the max packet size for some class (configured through tc) is violated by the actual size of the packets of that class, then QFQ would not schedule classes correctly, and the data structures implementing the bucket lists may get corrupted. This problem occurs with TSO/GSO even if the max packet size is set to the MTU, and is, e.g., the cause of the failure reported in [1]. Two patches have been proposed to solve this problem in [2], one of them is a preliminary version of this patch. This patch addresses the above issues by: 1) setting QFQ parameters to proper values for supporting TSO/GSO (in particular, setting the maximum possible packet size to 64KB), 2) automatically increasing the max packet size for a class, lmax, when a packet with a larger size than the current value of lmax arrives. The drawback of the first point is that the maximum weight for a class is now limited to 4096, which is equal to 1/16 of the maximum weight sum. Finally, this patch also forcibly caps the timestamps of a class if they are too high to be stored in the bucket list. This capping, taken from QFQ+ [3], handles the unfrequent case described in the comment to the function slot_insert. [1] http://marc.info/?l=linux-netdev&m=134968777902077&w=2 [2] http://marc.info/?l=linux-netdev&m=135096573507936&w=2 [3] http://marc.info/?l=linux-netdev&m=134902691421670&w=2 Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Tested-by: Cong Wang <amwang@redhat.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-11-05 20:29:24 +00:00
int err;
int delta_w;
if (NL_REQ_ATTR_CHECK(extack, NULL, tca, TCA_OPTIONS)) {
NL_SET_ERR_MSG_MOD(extack, "missing options");
return -EINVAL;
}
netlink: make validation more configurable for future strictness We currently have two levels of strict validation: 1) liberal (default) - undefined (type >= max) & NLA_UNSPEC attributes accepted - attribute length >= expected accepted - garbage at end of message accepted 2) strict (opt-in) - NLA_UNSPEC attributes accepted - attribute length >= expected accepted Split out parsing strictness into four different options: * TRAILING - check that there's no trailing data after parsing attributes (in message or nested) * MAXTYPE - reject attrs > max known type * UNSPEC - reject attributes with NLA_UNSPEC policy entries * STRICT_ATTRS - strictly validate attribute size The default for future things should be *everything*. The current *_strict() is a combination of TRAILING and MAXTYPE, and is renamed to _deprecated_strict(). The current regular parsing has none of this, and is renamed to *_parse_deprecated(). Additionally it allows us to selectively set one of the new flags even on old policies. Notably, the UNSPEC flag could be useful in this case, since it can be arranged (by filling in the policy) to not be an incompatible userspace ABI change, but would then going forward prevent forgetting attribute entries. Similar can apply to the POLICY flag. We end up with the following renames: * nla_parse -> nla_parse_deprecated * nla_parse_strict -> nla_parse_deprecated_strict * nlmsg_parse -> nlmsg_parse_deprecated * nlmsg_parse_strict -> nlmsg_parse_deprecated_strict * nla_parse_nested -> nla_parse_nested_deprecated * nla_validate_nested -> nla_validate_nested_deprecated Using spatch, of course: @@ expression TB, MAX, HEAD, LEN, POL, EXT; @@ -nla_parse(TB, MAX, HEAD, LEN, POL, EXT) +nla_parse_deprecated(TB, MAX, HEAD, LEN, POL, EXT) @@ expression NLH, HDRLEN, TB, MAX, POL, EXT; @@ -nlmsg_parse(NLH, HDRLEN, TB, MAX, POL, EXT) +nlmsg_parse_deprecated(NLH, HDRLEN, TB, MAX, POL, EXT) @@ expression NLH, HDRLEN, TB, MAX, POL, EXT; @@ -nlmsg_parse_strict(NLH, HDRLEN, TB, MAX, POL, EXT) +nlmsg_parse_deprecated_strict(NLH, HDRLEN, TB, MAX, POL, EXT) @@ expression TB, MAX, NLA, POL, EXT; @@ -nla_parse_nested(TB, MAX, NLA, POL, EXT) +nla_parse_nested_deprecated(TB, MAX, NLA, POL, EXT) @@ expression START, MAX, POL, EXT; @@ -nla_validate_nested(START, MAX, POL, EXT) +nla_validate_nested_deprecated(START, MAX, POL, EXT) @@ expression NLH, HDRLEN, MAX, POL, EXT; @@ -nlmsg_validate(NLH, HDRLEN, MAX, POL, EXT) +nlmsg_validate_deprecated(NLH, HDRLEN, MAX, POL, EXT) For this patch, don't actually add the strict, non-renamed versions yet so that it breaks compile if I get it wrong. Also, while at it, make nla_validate and nla_parse go down to a common __nla_validate_parse() function to avoid code duplication. Ultimately, this allows us to have very strict validation for every new caller of nla_parse()/nlmsg_parse() etc as re-introduced in the next patch, while existing things will continue to work as is. In effect then, this adds fully strict validation for any new command. Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-04-26 12:07:28 +00:00
err = nla_parse_nested_deprecated(tb, TCA_QFQ_MAX, tca[TCA_OPTIONS],
qfq_policy, extack);
if (err < 0)
return err;
if (tb[TCA_QFQ_WEIGHT])
weight = nla_get_u32(tb[TCA_QFQ_WEIGHT]);
else
weight = 1;
if (tb[TCA_QFQ_LMAX]) {
lmax = nla_get_u32(tb[TCA_QFQ_LMAX]);
} else {
/* MTU size is user controlled */
pkt_sched: enable QFQ to support TSO/GSO If the max packet size for some class (configured through tc) is violated by the actual size of the packets of that class, then QFQ would not schedule classes correctly, and the data structures implementing the bucket lists may get corrupted. This problem occurs with TSO/GSO even if the max packet size is set to the MTU, and is, e.g., the cause of the failure reported in [1]. Two patches have been proposed to solve this problem in [2], one of them is a preliminary version of this patch. This patch addresses the above issues by: 1) setting QFQ parameters to proper values for supporting TSO/GSO (in particular, setting the maximum possible packet size to 64KB), 2) automatically increasing the max packet size for a class, lmax, when a packet with a larger size than the current value of lmax arrives. The drawback of the first point is that the maximum weight for a class is now limited to 4096, which is equal to 1/16 of the maximum weight sum. Finally, this patch also forcibly caps the timestamps of a class if they are too high to be stored in the bucket list. This capping, taken from QFQ+ [3], handles the unfrequent case described in the comment to the function slot_insert. [1] http://marc.info/?l=linux-netdev&m=134968777902077&w=2 [2] http://marc.info/?l=linux-netdev&m=135096573507936&w=2 [3] http://marc.info/?l=linux-netdev&m=134902691421670&w=2 Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Tested-by: Cong Wang <amwang@redhat.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-11-05 20:29:24 +00:00
lmax = psched_mtu(qdisc_dev(sch));
if (lmax < QFQ_MIN_LMAX || lmax > QFQ_MAX_LMAX) {
NL_SET_ERR_MSG_MOD(extack,
"MTU size out of bounds for qfq");
return -EINVAL;
}
}
inv_w = ONE_FP / weight;
weight = ONE_FP / inv_w;
if (cl != NULL &&
lmax == cl->agg->lmax &&
weight == cl->agg->class_weight)
return 0; /* nothing to change */
delta_w = weight - (cl ? cl->agg->class_weight : 0);
if (q->wsum + delta_w > QFQ_MAX_WSUM) {
NL_SET_ERR_MSG_FMT_MOD(extack,
"total weight out of range (%d + %u)\n",
delta_w, q->wsum);
return -EINVAL;
}
if (cl != NULL) { /* modify existing class */
if (tca[TCA_RATE]) {
err = gen_replace_estimator(&cl->bstats, NULL,
&cl->rate_est,
NULL,
net: sched: Remove Qdisc::running sequence counter The Qdisc::running sequence counter has two uses: 1. Reliably reading qdisc's tc statistics while the qdisc is running (a seqcount read/retry loop at gnet_stats_add_basic()). 2. As a flag, indicating whether the qdisc in question is running (without any retry loops). For the first usage, the Qdisc::running sequence counter write section, qdisc_run_begin() => qdisc_run_end(), covers a much wider area than what is actually needed: the raw qdisc's bstats update. A u64_stats sync point was thus introduced (in previous commits) inside the bstats structure itself. A local u64_stats write section is then started and stopped for the bstats updates. Use that u64_stats sync point mechanism for the bstats read/retry loop at gnet_stats_add_basic(). For the second qdisc->running usage, a __QDISC_STATE_RUNNING bit flag, accessed with atomic bitops, is sufficient. Using a bit flag instead of a sequence counter at qdisc_run_begin/end() and qdisc_is_running() leads to the SMP barriers implicitly added through raw_read_seqcount() and write_seqcount_begin/end() getting removed. All call sites have been surveyed though, and no required ordering was identified. Now that the qdisc->running sequence counter is no longer used, remove it. Note, using u64_stats implies no sequence counter protection for 64-bit architectures. This can lead to the qdisc tc statistics "packets" vs. "bytes" values getting out of sync on rare occasions. The individual values will still be valid. Signed-off-by: Ahmed S. Darwish <a.darwish@linutronix.de> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: David S. Miller <davem@davemloft.net>
2021-10-16 08:49:10 +00:00
true,
tca[TCA_RATE]);
if (err)
return err;
}
existing = true;
goto set_change_agg;
}
/* create and init new class */
cl = kzalloc(sizeof(struct qfq_class), GFP_KERNEL);
if (cl == NULL)
return -ENOBUFS;
gnet_stats_basic_sync_init(&cl->bstats);
cl->common.classid = classid;
cl->deficit = lmax;
cl->qdisc = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops,
classid, NULL);
if (cl->qdisc == NULL)
cl->qdisc = &noop_qdisc;
if (tca[TCA_RATE]) {
err = gen_new_estimator(&cl->bstats, NULL,
&cl->rate_est,
NULL,
net: sched: Remove Qdisc::running sequence counter The Qdisc::running sequence counter has two uses: 1. Reliably reading qdisc's tc statistics while the qdisc is running (a seqcount read/retry loop at gnet_stats_add_basic()). 2. As a flag, indicating whether the qdisc in question is running (without any retry loops). For the first usage, the Qdisc::running sequence counter write section, qdisc_run_begin() => qdisc_run_end(), covers a much wider area than what is actually needed: the raw qdisc's bstats update. A u64_stats sync point was thus introduced (in previous commits) inside the bstats structure itself. A local u64_stats write section is then started and stopped for the bstats updates. Use that u64_stats sync point mechanism for the bstats read/retry loop at gnet_stats_add_basic(). For the second qdisc->running usage, a __QDISC_STATE_RUNNING bit flag, accessed with atomic bitops, is sufficient. Using a bit flag instead of a sequence counter at qdisc_run_begin/end() and qdisc_is_running() leads to the SMP barriers implicitly added through raw_read_seqcount() and write_seqcount_begin/end() getting removed. All call sites have been surveyed though, and no required ordering was identified. Now that the qdisc->running sequence counter is no longer used, remove it. Note, using u64_stats implies no sequence counter protection for 64-bit architectures. This can lead to the qdisc tc statistics "packets" vs. "bytes" values getting out of sync on rare occasions. The individual values will still be valid. Signed-off-by: Ahmed S. Darwish <a.darwish@linutronix.de> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: David S. Miller <davem@davemloft.net>
2021-10-16 08:49:10 +00:00
true,
tca[TCA_RATE]);
if (err)
goto destroy_class;
}
if (cl->qdisc != &noop_qdisc)
qdisc_hash_add(cl->qdisc, true);
set_change_agg:
sch_tree_lock(sch);
new_agg = qfq_find_agg(q, lmax, weight);
if (new_agg == NULL) { /* create new aggregate */
sch_tree_unlock(sch);
new_agg = kzalloc(sizeof(*new_agg), GFP_KERNEL);
if (new_agg == NULL) {
err = -ENOBUFS;
gen_kill_estimator(&cl->rate_est);
goto destroy_class;
}
sch_tree_lock(sch);
qfq_init_agg(q, new_agg, lmax, weight);
}
if (existing)
qfq_deact_rm_from_agg(q, cl);
pkt_sched: sch_qfq: fix qfq_change_class() error path If qfq_change_class() is unable to allocate memory for qfq_aggregate, it frees the class that has been inserted in the class hash table, but does not unhash it. Defer the insertion after the problematic allocation. BUG: KASAN: use-after-free in hlist_add_head include/linux/list.h:884 [inline] BUG: KASAN: use-after-free in qdisc_class_hash_insert+0x200/0x210 net/sched/sch_api.c:731 Write of size 8 at addr ffff88814a534f10 by task syz-executor.4/31478 CPU: 0 PID: 31478 Comm: syz-executor.4 Not tainted 5.13.0-rc6-syzkaller #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 Call Trace: __dump_stack lib/dump_stack.c:79 [inline] dump_stack+0x141/0x1d7 lib/dump_stack.c:120 print_address_description.constprop.0.cold+0x5b/0x2f8 mm/kasan/report.c:233 __kasan_report mm/kasan/report.c:419 [inline] kasan_report.cold+0x7c/0xd8 mm/kasan/report.c:436 hlist_add_head include/linux/list.h:884 [inline] qdisc_class_hash_insert+0x200/0x210 net/sched/sch_api.c:731 qfq_change_class+0x96c/0x1990 net/sched/sch_qfq.c:489 tc_ctl_tclass+0x514/0xe50 net/sched/sch_api.c:2113 rtnetlink_rcv_msg+0x44e/0xad0 net/core/rtnetlink.c:5564 netlink_rcv_skb+0x153/0x420 net/netlink/af_netlink.c:2504 netlink_unicast_kernel net/netlink/af_netlink.c:1314 [inline] netlink_unicast+0x533/0x7d0 net/netlink/af_netlink.c:1340 netlink_sendmsg+0x856/0xd90 net/netlink/af_netlink.c:1929 sock_sendmsg_nosec net/socket.c:654 [inline] sock_sendmsg+0xcf/0x120 net/socket.c:674 ____sys_sendmsg+0x6e8/0x810 net/socket.c:2350 ___sys_sendmsg+0xf3/0x170 net/socket.c:2404 __sys_sendmsg+0xe5/0x1b0 net/socket.c:2433 do_syscall_64+0x3a/0xb0 arch/x86/entry/common.c:47 entry_SYSCALL_64_after_hwframe+0x44/0xae RIP: 0033:0x4665d9 Code: ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 bc ff ff ff f7 d8 64 89 01 48 RSP: 002b:00007fdc7b5f0188 EFLAGS: 00000246 ORIG_RAX: 000000000000002e RAX: ffffffffffffffda RBX: 000000000056bf80 RCX: 00000000004665d9 RDX: 0000000000000000 RSI: 00000000200001c0 RDI: 0000000000000003 RBP: 00007fdc7b5f01d0 R08: 0000000000000000 R09: 0000000000000000 R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000002 R13: 00007ffcf7310b3f R14: 00007fdc7b5f0300 R15: 0000000000022000 Allocated by task 31445: kasan_save_stack+0x1b/0x40 mm/kasan/common.c:38 kasan_set_track mm/kasan/common.c:46 [inline] set_alloc_info mm/kasan/common.c:428 [inline] ____kasan_kmalloc mm/kasan/common.c:507 [inline] ____kasan_kmalloc mm/kasan/common.c:466 [inline] __kasan_kmalloc+0x9b/0xd0 mm/kasan/common.c:516 kmalloc include/linux/slab.h:556 [inline] kzalloc include/linux/slab.h:686 [inline] qfq_change_class+0x705/0x1990 net/sched/sch_qfq.c:464 tc_ctl_tclass+0x514/0xe50 net/sched/sch_api.c:2113 rtnetlink_rcv_msg+0x44e/0xad0 net/core/rtnetlink.c:5564 netlink_rcv_skb+0x153/0x420 net/netlink/af_netlink.c:2504 netlink_unicast_kernel net/netlink/af_netlink.c:1314 [inline] netlink_unicast+0x533/0x7d0 net/netlink/af_netlink.c:1340 netlink_sendmsg+0x856/0xd90 net/netlink/af_netlink.c:1929 sock_sendmsg_nosec net/socket.c:654 [inline] sock_sendmsg+0xcf/0x120 net/socket.c:674 ____sys_sendmsg+0x6e8/0x810 net/socket.c:2350 ___sys_sendmsg+0xf3/0x170 net/socket.c:2404 __sys_sendmsg+0xe5/0x1b0 net/socket.c:2433 do_syscall_64+0x3a/0xb0 arch/x86/entry/common.c:47 entry_SYSCALL_64_after_hwframe+0x44/0xae Freed by task 31445: kasan_save_stack+0x1b/0x40 mm/kasan/common.c:38 kasan_set_track+0x1c/0x30 mm/kasan/common.c:46 kasan_set_free_info+0x20/0x30 mm/kasan/generic.c:357 ____kasan_slab_free mm/kasan/common.c:360 [inline] ____kasan_slab_free mm/kasan/common.c:325 [inline] __kasan_slab_free+0xfb/0x130 mm/kasan/common.c:368 kasan_slab_free include/linux/kasan.h:212 [inline] slab_free_hook mm/slub.c:1583 [inline] slab_free_freelist_hook+0xdf/0x240 mm/slub.c:1608 slab_free mm/slub.c:3168 [inline] kfree+0xe5/0x7f0 mm/slub.c:4212 qfq_change_class+0x10fb/0x1990 net/sched/sch_qfq.c:518 tc_ctl_tclass+0x514/0xe50 net/sched/sch_api.c:2113 rtnetlink_rcv_msg+0x44e/0xad0 net/core/rtnetlink.c:5564 netlink_rcv_skb+0x153/0x420 net/netlink/af_netlink.c:2504 netlink_unicast_kernel net/netlink/af_netlink.c:1314 [inline] netlink_unicast+0x533/0x7d0 net/netlink/af_netlink.c:1340 netlink_sendmsg+0x856/0xd90 net/netlink/af_netlink.c:1929 sock_sendmsg_nosec net/socket.c:654 [inline] sock_sendmsg+0xcf/0x120 net/socket.c:674 ____sys_sendmsg+0x6e8/0x810 net/socket.c:2350 ___sys_sendmsg+0xf3/0x170 net/socket.c:2404 __sys_sendmsg+0xe5/0x1b0 net/socket.c:2433 do_syscall_64+0x3a/0xb0 arch/x86/entry/common.c:47 entry_SYSCALL_64_after_hwframe+0x44/0xae The buggy address belongs to the object at ffff88814a534f00 which belongs to the cache kmalloc-128 of size 128 The buggy address is located 16 bytes inside of 128-byte region [ffff88814a534f00, ffff88814a534f80) The buggy address belongs to the page: page:ffffea0005294d00 refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x14a534 flags: 0x57ff00000000200(slab|node=1|zone=2|lastcpupid=0x7ff) raw: 057ff00000000200 ffffea00004fee00 0000000600000006 ffff8880110418c0 raw: 0000000000000000 0000000000100010 00000001ffffffff 0000000000000000 page dumped because: kasan: bad access detected page_owner tracks the page as allocated page last allocated via order 0, migratetype Unmovable, gfp_mask 0x12cc0(GFP_KERNEL|__GFP_NOWARN|__GFP_NORETRY), pid 29797, ts 604817765317, free_ts 604810151744 prep_new_page mm/page_alloc.c:2358 [inline] get_page_from_freelist+0x1033/0x2b60 mm/page_alloc.c:3994 __alloc_pages+0x1b2/0x500 mm/page_alloc.c:5200 alloc_pages+0x18c/0x2a0 mm/mempolicy.c:2272 alloc_slab_page mm/slub.c:1646 [inline] allocate_slab+0x2c5/0x4c0 mm/slub.c:1786 new_slab mm/slub.c:1849 [inline] new_slab_objects mm/slub.c:2595 [inline] ___slab_alloc+0x4a1/0x810 mm/slub.c:2758 __slab_alloc.constprop.0+0xa7/0xf0 mm/slub.c:2798 slab_alloc_node mm/slub.c:2880 [inline] slab_alloc mm/slub.c:2922 [inline] __kmalloc+0x315/0x330 mm/slub.c:4050 kmalloc include/linux/slab.h:561 [inline] kzalloc include/linux/slab.h:686 [inline] __register_sysctl_table+0x112/0x1090 fs/proc/proc_sysctl.c:1318 mpls_dev_sysctl_register+0x1b7/0x2d0 net/mpls/af_mpls.c:1421 mpls_add_dev net/mpls/af_mpls.c:1472 [inline] mpls_dev_notify+0x214/0x8b0 net/mpls/af_mpls.c:1588 notifier_call_chain+0xb5/0x200 kernel/notifier.c:83 call_netdevice_notifiers_info+0xb5/0x130 net/core/dev.c:2121 call_netdevice_notifiers_extack net/core/dev.c:2133 [inline] call_netdevice_notifiers net/core/dev.c:2147 [inline] register_netdevice+0x106b/0x1500 net/core/dev.c:10312 veth_newlink+0x585/0xac0 drivers/net/veth.c:1547 __rtnl_newlink+0x1062/0x1710 net/core/rtnetlink.c:3452 rtnl_newlink+0x64/0xa0 net/core/rtnetlink.c:3500 page last free stack trace: reset_page_owner include/linux/page_owner.h:24 [inline] free_pages_prepare mm/page_alloc.c:1298 [inline] free_pcp_prepare+0x223/0x300 mm/page_alloc.c:1342 free_unref_page_prepare mm/page_alloc.c:3250 [inline] free_unref_page+0x12/0x1d0 mm/page_alloc.c:3298 __vunmap+0x783/0xb60 mm/vmalloc.c:2566 free_work+0x58/0x70 mm/vmalloc.c:80 process_one_work+0x98d/0x1600 kernel/workqueue.c:2276 worker_thread+0x64c/0x1120 kernel/workqueue.c:2422 kthread+0x3b1/0x4a0 kernel/kthread.c:313 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:294 Memory state around the buggy address: ffff88814a534e00: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ffff88814a534e80: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc >ffff88814a534f00: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ^ ffff88814a534f80: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc ffff88814a535000: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 Fixes: 462dbc9101acd ("pkt_sched: QFQ Plus: fair-queueing service at DRR cost") Signed-off-by: Eric Dumazet <edumazet@google.com> Reported-by: syzbot <syzkaller@googlegroups.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2021-06-21 17:54:49 +00:00
else
qdisc_class_hash_insert(&q->clhash, &cl->common);
qfq_add_to_agg(q, new_agg, cl);
sch_tree_unlock(sch);
pkt_sched: sch_qfq: fix qfq_change_class() error path If qfq_change_class() is unable to allocate memory for qfq_aggregate, it frees the class that has been inserted in the class hash table, but does not unhash it. Defer the insertion after the problematic allocation. BUG: KASAN: use-after-free in hlist_add_head include/linux/list.h:884 [inline] BUG: KASAN: use-after-free in qdisc_class_hash_insert+0x200/0x210 net/sched/sch_api.c:731 Write of size 8 at addr ffff88814a534f10 by task syz-executor.4/31478 CPU: 0 PID: 31478 Comm: syz-executor.4 Not tainted 5.13.0-rc6-syzkaller #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 Call Trace: __dump_stack lib/dump_stack.c:79 [inline] dump_stack+0x141/0x1d7 lib/dump_stack.c:120 print_address_description.constprop.0.cold+0x5b/0x2f8 mm/kasan/report.c:233 __kasan_report mm/kasan/report.c:419 [inline] kasan_report.cold+0x7c/0xd8 mm/kasan/report.c:436 hlist_add_head include/linux/list.h:884 [inline] qdisc_class_hash_insert+0x200/0x210 net/sched/sch_api.c:731 qfq_change_class+0x96c/0x1990 net/sched/sch_qfq.c:489 tc_ctl_tclass+0x514/0xe50 net/sched/sch_api.c:2113 rtnetlink_rcv_msg+0x44e/0xad0 net/core/rtnetlink.c:5564 netlink_rcv_skb+0x153/0x420 net/netlink/af_netlink.c:2504 netlink_unicast_kernel net/netlink/af_netlink.c:1314 [inline] netlink_unicast+0x533/0x7d0 net/netlink/af_netlink.c:1340 netlink_sendmsg+0x856/0xd90 net/netlink/af_netlink.c:1929 sock_sendmsg_nosec net/socket.c:654 [inline] sock_sendmsg+0xcf/0x120 net/socket.c:674 ____sys_sendmsg+0x6e8/0x810 net/socket.c:2350 ___sys_sendmsg+0xf3/0x170 net/socket.c:2404 __sys_sendmsg+0xe5/0x1b0 net/socket.c:2433 do_syscall_64+0x3a/0xb0 arch/x86/entry/common.c:47 entry_SYSCALL_64_after_hwframe+0x44/0xae RIP: 0033:0x4665d9 Code: ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 bc ff ff ff f7 d8 64 89 01 48 RSP: 002b:00007fdc7b5f0188 EFLAGS: 00000246 ORIG_RAX: 000000000000002e RAX: ffffffffffffffda RBX: 000000000056bf80 RCX: 00000000004665d9 RDX: 0000000000000000 RSI: 00000000200001c0 RDI: 0000000000000003 RBP: 00007fdc7b5f01d0 R08: 0000000000000000 R09: 0000000000000000 R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000002 R13: 00007ffcf7310b3f R14: 00007fdc7b5f0300 R15: 0000000000022000 Allocated by task 31445: kasan_save_stack+0x1b/0x40 mm/kasan/common.c:38 kasan_set_track mm/kasan/common.c:46 [inline] set_alloc_info mm/kasan/common.c:428 [inline] ____kasan_kmalloc mm/kasan/common.c:507 [inline] ____kasan_kmalloc mm/kasan/common.c:466 [inline] __kasan_kmalloc+0x9b/0xd0 mm/kasan/common.c:516 kmalloc include/linux/slab.h:556 [inline] kzalloc include/linux/slab.h:686 [inline] qfq_change_class+0x705/0x1990 net/sched/sch_qfq.c:464 tc_ctl_tclass+0x514/0xe50 net/sched/sch_api.c:2113 rtnetlink_rcv_msg+0x44e/0xad0 net/core/rtnetlink.c:5564 netlink_rcv_skb+0x153/0x420 net/netlink/af_netlink.c:2504 netlink_unicast_kernel net/netlink/af_netlink.c:1314 [inline] netlink_unicast+0x533/0x7d0 net/netlink/af_netlink.c:1340 netlink_sendmsg+0x856/0xd90 net/netlink/af_netlink.c:1929 sock_sendmsg_nosec net/socket.c:654 [inline] sock_sendmsg+0xcf/0x120 net/socket.c:674 ____sys_sendmsg+0x6e8/0x810 net/socket.c:2350 ___sys_sendmsg+0xf3/0x170 net/socket.c:2404 __sys_sendmsg+0xe5/0x1b0 net/socket.c:2433 do_syscall_64+0x3a/0xb0 arch/x86/entry/common.c:47 entry_SYSCALL_64_after_hwframe+0x44/0xae Freed by task 31445: kasan_save_stack+0x1b/0x40 mm/kasan/common.c:38 kasan_set_track+0x1c/0x30 mm/kasan/common.c:46 kasan_set_free_info+0x20/0x30 mm/kasan/generic.c:357 ____kasan_slab_free mm/kasan/common.c:360 [inline] ____kasan_slab_free mm/kasan/common.c:325 [inline] __kasan_slab_free+0xfb/0x130 mm/kasan/common.c:368 kasan_slab_free include/linux/kasan.h:212 [inline] slab_free_hook mm/slub.c:1583 [inline] slab_free_freelist_hook+0xdf/0x240 mm/slub.c:1608 slab_free mm/slub.c:3168 [inline] kfree+0xe5/0x7f0 mm/slub.c:4212 qfq_change_class+0x10fb/0x1990 net/sched/sch_qfq.c:518 tc_ctl_tclass+0x514/0xe50 net/sched/sch_api.c:2113 rtnetlink_rcv_msg+0x44e/0xad0 net/core/rtnetlink.c:5564 netlink_rcv_skb+0x153/0x420 net/netlink/af_netlink.c:2504 netlink_unicast_kernel net/netlink/af_netlink.c:1314 [inline] netlink_unicast+0x533/0x7d0 net/netlink/af_netlink.c:1340 netlink_sendmsg+0x856/0xd90 net/netlink/af_netlink.c:1929 sock_sendmsg_nosec net/socket.c:654 [inline] sock_sendmsg+0xcf/0x120 net/socket.c:674 ____sys_sendmsg+0x6e8/0x810 net/socket.c:2350 ___sys_sendmsg+0xf3/0x170 net/socket.c:2404 __sys_sendmsg+0xe5/0x1b0 net/socket.c:2433 do_syscall_64+0x3a/0xb0 arch/x86/entry/common.c:47 entry_SYSCALL_64_after_hwframe+0x44/0xae The buggy address belongs to the object at ffff88814a534f00 which belongs to the cache kmalloc-128 of size 128 The buggy address is located 16 bytes inside of 128-byte region [ffff88814a534f00, ffff88814a534f80) The buggy address belongs to the page: page:ffffea0005294d00 refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x14a534 flags: 0x57ff00000000200(slab|node=1|zone=2|lastcpupid=0x7ff) raw: 057ff00000000200 ffffea00004fee00 0000000600000006 ffff8880110418c0 raw: 0000000000000000 0000000000100010 00000001ffffffff 0000000000000000 page dumped because: kasan: bad access detected page_owner tracks the page as allocated page last allocated via order 0, migratetype Unmovable, gfp_mask 0x12cc0(GFP_KERNEL|__GFP_NOWARN|__GFP_NORETRY), pid 29797, ts 604817765317, free_ts 604810151744 prep_new_page mm/page_alloc.c:2358 [inline] get_page_from_freelist+0x1033/0x2b60 mm/page_alloc.c:3994 __alloc_pages+0x1b2/0x500 mm/page_alloc.c:5200 alloc_pages+0x18c/0x2a0 mm/mempolicy.c:2272 alloc_slab_page mm/slub.c:1646 [inline] allocate_slab+0x2c5/0x4c0 mm/slub.c:1786 new_slab mm/slub.c:1849 [inline] new_slab_objects mm/slub.c:2595 [inline] ___slab_alloc+0x4a1/0x810 mm/slub.c:2758 __slab_alloc.constprop.0+0xa7/0xf0 mm/slub.c:2798 slab_alloc_node mm/slub.c:2880 [inline] slab_alloc mm/slub.c:2922 [inline] __kmalloc+0x315/0x330 mm/slub.c:4050 kmalloc include/linux/slab.h:561 [inline] kzalloc include/linux/slab.h:686 [inline] __register_sysctl_table+0x112/0x1090 fs/proc/proc_sysctl.c:1318 mpls_dev_sysctl_register+0x1b7/0x2d0 net/mpls/af_mpls.c:1421 mpls_add_dev net/mpls/af_mpls.c:1472 [inline] mpls_dev_notify+0x214/0x8b0 net/mpls/af_mpls.c:1588 notifier_call_chain+0xb5/0x200 kernel/notifier.c:83 call_netdevice_notifiers_info+0xb5/0x130 net/core/dev.c:2121 call_netdevice_notifiers_extack net/core/dev.c:2133 [inline] call_netdevice_notifiers net/core/dev.c:2147 [inline] register_netdevice+0x106b/0x1500 net/core/dev.c:10312 veth_newlink+0x585/0xac0 drivers/net/veth.c:1547 __rtnl_newlink+0x1062/0x1710 net/core/rtnetlink.c:3452 rtnl_newlink+0x64/0xa0 net/core/rtnetlink.c:3500 page last free stack trace: reset_page_owner include/linux/page_owner.h:24 [inline] free_pages_prepare mm/page_alloc.c:1298 [inline] free_pcp_prepare+0x223/0x300 mm/page_alloc.c:1342 free_unref_page_prepare mm/page_alloc.c:3250 [inline] free_unref_page+0x12/0x1d0 mm/page_alloc.c:3298 __vunmap+0x783/0xb60 mm/vmalloc.c:2566 free_work+0x58/0x70 mm/vmalloc.c:80 process_one_work+0x98d/0x1600 kernel/workqueue.c:2276 worker_thread+0x64c/0x1120 kernel/workqueue.c:2422 kthread+0x3b1/0x4a0 kernel/kthread.c:313 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:294 Memory state around the buggy address: ffff88814a534e00: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ffff88814a534e80: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc >ffff88814a534f00: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ^ ffff88814a534f80: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc ffff88814a535000: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 Fixes: 462dbc9101acd ("pkt_sched: QFQ Plus: fair-queueing service at DRR cost") Signed-off-by: Eric Dumazet <edumazet@google.com> Reported-by: syzbot <syzkaller@googlegroups.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2021-06-21 17:54:49 +00:00
qdisc_class_hash_grow(sch, &q->clhash);
*arg = (unsigned long)cl;
return 0;
destroy_class:
qdisc_put(cl->qdisc);
kfree(cl);
return err;
}
static void qfq_destroy_class(struct Qdisc *sch, struct qfq_class *cl)
{
struct qfq_sched *q = qdisc_priv(sch);
qfq_rm_from_agg(q, cl);
gen_kill_estimator(&cl->rate_est);
qdisc_put(cl->qdisc);
kfree(cl);
}
static int qfq_delete_class(struct Qdisc *sch, unsigned long arg,
struct netlink_ext_ack *extack)
{
struct qfq_sched *q = qdisc_priv(sch);
struct qfq_class *cl = (struct qfq_class *)arg;
if (qdisc_class_in_use(&cl->common)) {
NL_SET_ERR_MSG_MOD(extack, "QFQ class in use");
return -EBUSY;
}
sch_tree_lock(sch);
qdisc_purge_queue(cl->qdisc);
qdisc_class_hash_remove(&q->clhash, &cl->common);
sch_tree_unlock(sch);
qfq_destroy_class(sch, cl);
return 0;
}
static unsigned long qfq_search_class(struct Qdisc *sch, u32 classid)
{
return (unsigned long)qfq_find_class(sch, classid);
}
static struct tcf_block *qfq_tcf_block(struct Qdisc *sch, unsigned long cl,
struct netlink_ext_ack *extack)
{
struct qfq_sched *q = qdisc_priv(sch);
if (cl)
return NULL;
return q->block;
}
static unsigned long qfq_bind_tcf(struct Qdisc *sch, unsigned long parent,
u32 classid)
{
struct qfq_class *cl = qfq_find_class(sch, classid);
if (cl)
qdisc_class_get(&cl->common);
return (unsigned long)cl;
}
static void qfq_unbind_tcf(struct Qdisc *sch, unsigned long arg)
{
struct qfq_class *cl = (struct qfq_class *)arg;
qdisc_class_put(&cl->common);
}
static int qfq_graft_class(struct Qdisc *sch, unsigned long arg,
struct Qdisc *new, struct Qdisc **old,
struct netlink_ext_ack *extack)
{
struct qfq_class *cl = (struct qfq_class *)arg;
if (new == NULL) {
new = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops,
cl->common.classid, NULL);
if (new == NULL)
new = &noop_qdisc;
}
*old = qdisc_replace(sch, new, &cl->qdisc);
return 0;
}
static struct Qdisc *qfq_class_leaf(struct Qdisc *sch, unsigned long arg)
{
struct qfq_class *cl = (struct qfq_class *)arg;
return cl->qdisc;
}
static int qfq_dump_class(struct Qdisc *sch, unsigned long arg,
struct sk_buff *skb, struct tcmsg *tcm)
{
struct qfq_class *cl = (struct qfq_class *)arg;
struct nlattr *nest;
tcm->tcm_parent = TC_H_ROOT;
tcm->tcm_handle = cl->common.classid;
tcm->tcm_info = cl->qdisc->handle;
nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
if (nest == NULL)
goto nla_put_failure;
if (nla_put_u32(skb, TCA_QFQ_WEIGHT, cl->agg->class_weight) ||
nla_put_u32(skb, TCA_QFQ_LMAX, cl->agg->lmax))
goto nla_put_failure;
return nla_nest_end(skb, nest);
nla_put_failure:
nla_nest_cancel(skb, nest);
return -EMSGSIZE;
}
static int qfq_dump_class_stats(struct Qdisc *sch, unsigned long arg,
struct gnet_dump *d)
{
struct qfq_class *cl = (struct qfq_class *)arg;
struct tc_qfq_stats xstats;
memset(&xstats, 0, sizeof(xstats));
xstats.weight = cl->agg->class_weight;
xstats.lmax = cl->agg->lmax;
net: sched: Remove Qdisc::running sequence counter The Qdisc::running sequence counter has two uses: 1. Reliably reading qdisc's tc statistics while the qdisc is running (a seqcount read/retry loop at gnet_stats_add_basic()). 2. As a flag, indicating whether the qdisc in question is running (without any retry loops). For the first usage, the Qdisc::running sequence counter write section, qdisc_run_begin() => qdisc_run_end(), covers a much wider area than what is actually needed: the raw qdisc's bstats update. A u64_stats sync point was thus introduced (in previous commits) inside the bstats structure itself. A local u64_stats write section is then started and stopped for the bstats updates. Use that u64_stats sync point mechanism for the bstats read/retry loop at gnet_stats_add_basic(). For the second qdisc->running usage, a __QDISC_STATE_RUNNING bit flag, accessed with atomic bitops, is sufficient. Using a bit flag instead of a sequence counter at qdisc_run_begin/end() and qdisc_is_running() leads to the SMP barriers implicitly added through raw_read_seqcount() and write_seqcount_begin/end() getting removed. All call sites have been surveyed though, and no required ordering was identified. Now that the qdisc->running sequence counter is no longer used, remove it. Note, using u64_stats implies no sequence counter protection for 64-bit architectures. This can lead to the qdisc tc statistics "packets" vs. "bytes" values getting out of sync on rare occasions. The individual values will still be valid. Signed-off-by: Ahmed S. Darwish <a.darwish@linutronix.de> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: David S. Miller <davem@davemloft.net>
2021-10-16 08:49:10 +00:00
if (gnet_stats_copy_basic(d, NULL, &cl->bstats, true) < 0 ||
gnet_stats_copy_rate_est(d, &cl->rate_est) < 0 ||
qdisc_qstats_copy(d, cl->qdisc) < 0)
return -1;
return gnet_stats_copy_app(d, &xstats, sizeof(xstats));
}
static void qfq_walk(struct Qdisc *sch, struct qdisc_walker *arg)
{
struct qfq_sched *q = qdisc_priv(sch);
struct qfq_class *cl;
unsigned int i;
if (arg->stop)
return;
for (i = 0; i < q->clhash.hashsize; i++) {
hlist: drop the node parameter from iterators I'm not sure why, but the hlist for each entry iterators were conceived list_for_each_entry(pos, head, member) The hlist ones were greedy and wanted an extra parameter: hlist_for_each_entry(tpos, pos, head, member) Why did they need an extra pos parameter? I'm not quite sure. Not only they don't really need it, it also prevents the iterator from looking exactly like the list iterator, which is unfortunate. Besides the semantic patch, there was some manual work required: - Fix up the actual hlist iterators in linux/list.h - Fix up the declaration of other iterators based on the hlist ones. - A very small amount of places were using the 'node' parameter, this was modified to use 'obj->member' instead. - Coccinelle didn't handle the hlist_for_each_entry_safe iterator properly, so those had to be fixed up manually. The semantic patch which is mostly the work of Peter Senna Tschudin is here: @@ iterator name hlist_for_each_entry, hlist_for_each_entry_continue, hlist_for_each_entry_from, hlist_for_each_entry_rcu, hlist_for_each_entry_rcu_bh, hlist_for_each_entry_continue_rcu_bh, for_each_busy_worker, ax25_uid_for_each, ax25_for_each, inet_bind_bucket_for_each, sctp_for_each_hentry, sk_for_each, sk_for_each_rcu, sk_for_each_from, sk_for_each_safe, sk_for_each_bound, hlist_for_each_entry_safe, hlist_for_each_entry_continue_rcu, nr_neigh_for_each, nr_neigh_for_each_safe, nr_node_for_each, nr_node_for_each_safe, for_each_gfn_indirect_valid_sp, for_each_gfn_sp, for_each_host; type T; expression a,c,d,e; identifier b; statement S; @@ -T b; <+... when != b ( hlist_for_each_entry(a, - b, c, d) S | hlist_for_each_entry_continue(a, - b, c) S | hlist_for_each_entry_from(a, - b, c) S | hlist_for_each_entry_rcu(a, - b, c, d) S | hlist_for_each_entry_rcu_bh(a, - b, c, d) S | hlist_for_each_entry_continue_rcu_bh(a, - b, c) S | for_each_busy_worker(a, c, - b, d) S | ax25_uid_for_each(a, - b, c) S | ax25_for_each(a, - b, c) S | inet_bind_bucket_for_each(a, - b, c) S | sctp_for_each_hentry(a, - b, c) S | sk_for_each(a, - b, c) S | sk_for_each_rcu(a, - b, c) S | sk_for_each_from -(a, b) +(a) S + sk_for_each_from(a) S | sk_for_each_safe(a, - b, c, d) S | sk_for_each_bound(a, - b, c) S | hlist_for_each_entry_safe(a, - b, c, d, e) S | hlist_for_each_entry_continue_rcu(a, - b, c) S | nr_neigh_for_each(a, - b, c) S | nr_neigh_for_each_safe(a, - b, c, d) S | nr_node_for_each(a, - b, c) S | nr_node_for_each_safe(a, - b, c, d) S | - for_each_gfn_sp(a, c, d, b) S + for_each_gfn_sp(a, c, d) S | - for_each_gfn_indirect_valid_sp(a, c, d, b) S + for_each_gfn_indirect_valid_sp(a, c, d) S | for_each_host(a, - b, c) S | for_each_host_safe(a, - b, c, d) S | for_each_mesh_entry(a, - b, c, d) S ) ...+> [akpm@linux-foundation.org: drop bogus change from net/ipv4/raw.c] [akpm@linux-foundation.org: drop bogus hunk from net/ipv6/raw.c] [akpm@linux-foundation.org: checkpatch fixes] [akpm@linux-foundation.org: fix warnings] [akpm@linux-foudnation.org: redo intrusive kvm changes] Tested-by: Peter Senna Tschudin <peter.senna@gmail.com> Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com> Cc: Wu Fengguang <fengguang.wu@intel.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Gleb Natapov <gleb@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-02-28 01:06:00 +00:00
hlist_for_each_entry(cl, &q->clhash.hash[i], common.hnode) {
if (!tc_qdisc_stats_dump(sch, (unsigned long)cl, arg))
return;
}
}
}
static struct qfq_class *qfq_classify(struct sk_buff *skb, struct Qdisc *sch,
int *qerr)
{
struct qfq_sched *q = qdisc_priv(sch);
struct qfq_class *cl;
struct tcf_result res;
struct tcf_proto *fl;
int result;
if (TC_H_MAJ(skb->priority ^ sch->handle) == 0) {
pr_debug("qfq_classify: found %d\n", skb->priority);
cl = qfq_find_class(sch, skb->priority);
if (cl != NULL)
return cl;
}
*qerr = NET_XMIT_SUCCESS | __NET_XMIT_BYPASS;
fl = rcu_dereference_bh(q->filter_list);
result = tcf_classify(skb, NULL, fl, &res, false);
if (result >= 0) {
#ifdef CONFIG_NET_CLS_ACT
switch (result) {
case TC_ACT_QUEUED:
case TC_ACT_STOLEN:
case TC_ACT_TRAP:
*qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN;
fallthrough;
case TC_ACT_SHOT:
return NULL;
}
#endif
cl = (struct qfq_class *)res.class;
if (cl == NULL)
cl = qfq_find_class(sch, res.classid);
return cl;
}
return NULL;
}
/* Generic comparison function, handling wraparound. */
static inline int qfq_gt(u64 a, u64 b)
{
return (s64)(a - b) > 0;
}
/* Round a precise timestamp to its slotted value. */
static inline u64 qfq_round_down(u64 ts, unsigned int shift)
{
return ts & ~((1ULL << shift) - 1);
}
/* return the pointer to the group with lowest index in the bitmap */
static inline struct qfq_group *qfq_ffs(struct qfq_sched *q,
unsigned long bitmap)
{
int index = __ffs(bitmap);
return &q->groups[index];
}
/* Calculate a mask to mimic what would be ffs_from(). */
static inline unsigned long mask_from(unsigned long bitmap, int from)
{
return bitmap & ~((1UL << from) - 1);
}
/*
* The state computation relies on ER=0, IR=1, EB=2, IB=3
* First compute eligibility comparing grp->S, q->V,
* then check if someone is blocking us and possibly add EB
*/
static int qfq_calc_state(struct qfq_sched *q, const struct qfq_group *grp)
{
/* if S > V we are not eligible */
unsigned int state = qfq_gt(grp->S, q->V);
unsigned long mask = mask_from(q->bitmaps[ER], grp->index);
struct qfq_group *next;
if (mask) {
next = qfq_ffs(q, mask);
if (qfq_gt(grp->F, next->F))
state |= EB;
}
return state;
}
/*
* In principle
* q->bitmaps[dst] |= q->bitmaps[src] & mask;
* q->bitmaps[src] &= ~mask;
* but we should make sure that src != dst
*/
static inline void qfq_move_groups(struct qfq_sched *q, unsigned long mask,
int src, int dst)
{
q->bitmaps[dst] |= q->bitmaps[src] & mask;
q->bitmaps[src] &= ~mask;
}
static void qfq_unblock_groups(struct qfq_sched *q, int index, u64 old_F)
{
unsigned long mask = mask_from(q->bitmaps[ER], index + 1);
struct qfq_group *next;
if (mask) {
next = qfq_ffs(q, mask);
if (!qfq_gt(next->F, old_F))
return;
}
mask = (1UL << index) - 1;
qfq_move_groups(q, mask, EB, ER);
qfq_move_groups(q, mask, IB, IR);
}
/*
* perhaps
*
old_V ^= q->V;
old_V >>= q->min_slot_shift;
if (old_V) {
...
}
*
*/
static void qfq_make_eligible(struct qfq_sched *q)
{
unsigned long vslot = q->V >> q->min_slot_shift;
unsigned long old_vslot = q->oldV >> q->min_slot_shift;
if (vslot != old_vslot) {
unsigned long mask;
int last_flip_pos = fls(vslot ^ old_vslot);
if (last_flip_pos > 31) /* higher than the number of groups */
mask = ~0UL; /* make all groups eligible */
else
mask = (1UL << last_flip_pos) - 1;
qfq_move_groups(q, mask, IR, ER);
qfq_move_groups(q, mask, IB, EB);
}
}
/*
pkt_sched: sch_qfq: remove a source of high packet delay/jitter QFQ+ inherits from QFQ a design choice that may cause a high packet delay/jitter and a severe short-term unfairness. As QFQ, QFQ+ uses a special quantity, the system virtual time, to track the service provided by the ideal system it approximates. When a packet is dequeued, this quantity must be incremented by the size of the packet, divided by the sum of the weights of the aggregates waiting to be served. Tracking this sum correctly is a non-trivial task, because, to preserve tight service guarantees, the decrement of this sum must be delayed in a special way [1]: this sum can be decremented only after that its value would decrease also in the ideal system approximated by QFQ+. For efficiency, QFQ+ keeps track only of the 'instantaneous' weight sum, increased and decreased immediately as the weight of an aggregate changes, and as an aggregate is created or destroyed (which, in its turn, happens as a consequence of some class being created/destroyed/changed). However, to avoid the problems caused to service guarantees by these immediate decreases, QFQ+ increments the system virtual time using the maximum value allowed for the weight sum, 2^10, in place of the dynamic, instantaneous value. The instantaneous value of the weight sum is used only to check whether a request of weight increase or a class creation can be satisfied. Unfortunately, the problems caused by this choice are worse than the temporary degradation of the service guarantees that may occur, when a class is changed or destroyed, if the instantaneous value of the weight sum was used to update the system virtual time. In fact, the fraction of the link bandwidth guaranteed by QFQ+ to each aggregate is equal to the ratio between the weight of the aggregate and the sum of the weights of the competing aggregates. The packet delay guaranteed to the aggregate is instead inversely proportional to the guaranteed bandwidth. By using the maximum possible value, and not the actual value of the weight sum, QFQ+ provides each aggregate with the worst possible service guarantees, and not with service guarantees related to the actual set of competing aggregates. To see the consequences of this fact, consider the following simple example. Suppose that only the following aggregates are backlogged, i.e., that only the classes in the following aggregates have packets to transmit: one aggregate with weight 10, say A, and ten aggregates with weight 1, say B1, B2, ..., B10. In particular, suppose that these aggregates are always backlogged. Given the weight distribution, the smoothest and fairest service order would be: A B1 A B2 A B3 A B4 A B5 A B6 A B7 A B8 A B9 A B10 A B1 A B2 ... QFQ+ would provide exactly this optimal service if it used the actual value for the weight sum instead of the maximum possible value, i.e., 11 instead of 2^10. In contrast, since QFQ+ uses the latter value, it serves aggregates as follows (easy to prove and to reproduce experimentally): A B1 B2 B3 B4 B5 B6 B7 B8 B9 B10 A A A A A A A A A A B1 B2 ... B10 A A ... By replacing 10 with N in the above example, and by increasing N, one can increase at will the maximum packet delay and the jitter experienced by the classes in aggregate A. This patch addresses this issue by just using the above 'instantaneous' value of the weight sum, instead of the maximum possible value, when updating the system virtual time. After the instantaneous weight sum is decreased, QFQ+ may deviate from the ideal service for a time interval in the order of the time to serve one maximum-size packet for each backlogged class. The worst-case extent of the deviation exhibited by QFQ+ during this time interval [1] is basically the same as of the deviation described above (but, without this patch, QFQ+ suffers from such a deviation all the time). Finally, this patch modifies the comment to the function qfq_slot_insert, to make it coherent with the fact that the weight sum used by QFQ+ can now be lower than the maximum possible value. [1] P. Valente, "Extending WF2Q+ to support a dynamic traffic mix", Proceedings of AAA-IDEA'05, June 2005. Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-07-16 06:52:30 +00:00
* The index of the slot in which the input aggregate agg is to be
* inserted must not be higher than QFQ_MAX_SLOTS-2. There is a '-2'
* and not a '-1' because the start time of the group may be moved
* backward by one slot after the aggregate has been inserted, and
* this would cause non-empty slots to be right-shifted by one
* position.
*
* QFQ+ fully satisfies this bound to the slot index if the parameters
* of the classes are not changed dynamically, and if QFQ+ never
* happens to postpone the service of agg unjustly, i.e., it never
* happens that the aggregate becomes backlogged and eligible, or just
* eligible, while an aggregate with a higher approximated finish time
* is being served. In particular, in this case QFQ+ guarantees that
* the timestamps of agg are low enough that the slot index is never
* higher than 2. Unfortunately, QFQ+ cannot provide the same
* guarantee if it happens to unjustly postpone the service of agg, or
* if the parameters of some class are changed.
*
* As for the first event, i.e., an out-of-order service, the
* upper bound to the slot index guaranteed by QFQ+ grows to
* 2 +
* QFQ_MAX_AGG_CLASSES * ((1<<QFQ_MTU_SHIFT)/QFQ_MIN_LMAX) *
* (current_max_weight/current_wsum) <= 2 + 8 * 128 * 1.
pkt_sched: enable QFQ to support TSO/GSO If the max packet size for some class (configured through tc) is violated by the actual size of the packets of that class, then QFQ would not schedule classes correctly, and the data structures implementing the bucket lists may get corrupted. This problem occurs with TSO/GSO even if the max packet size is set to the MTU, and is, e.g., the cause of the failure reported in [1]. Two patches have been proposed to solve this problem in [2], one of them is a preliminary version of this patch. This patch addresses the above issues by: 1) setting QFQ parameters to proper values for supporting TSO/GSO (in particular, setting the maximum possible packet size to 64KB), 2) automatically increasing the max packet size for a class, lmax, when a packet with a larger size than the current value of lmax arrives. The drawback of the first point is that the maximum weight for a class is now limited to 4096, which is equal to 1/16 of the maximum weight sum. Finally, this patch also forcibly caps the timestamps of a class if they are too high to be stored in the bucket list. This capping, taken from QFQ+ [3], handles the unfrequent case described in the comment to the function slot_insert. [1] http://marc.info/?l=linux-netdev&m=134968777902077&w=2 [2] http://marc.info/?l=linux-netdev&m=135096573507936&w=2 [3] http://marc.info/?l=linux-netdev&m=134902691421670&w=2 Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Tested-by: Cong Wang <amwang@redhat.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-11-05 20:29:24 +00:00
*
pkt_sched: sch_qfq: remove a source of high packet delay/jitter QFQ+ inherits from QFQ a design choice that may cause a high packet delay/jitter and a severe short-term unfairness. As QFQ, QFQ+ uses a special quantity, the system virtual time, to track the service provided by the ideal system it approximates. When a packet is dequeued, this quantity must be incremented by the size of the packet, divided by the sum of the weights of the aggregates waiting to be served. Tracking this sum correctly is a non-trivial task, because, to preserve tight service guarantees, the decrement of this sum must be delayed in a special way [1]: this sum can be decremented only after that its value would decrease also in the ideal system approximated by QFQ+. For efficiency, QFQ+ keeps track only of the 'instantaneous' weight sum, increased and decreased immediately as the weight of an aggregate changes, and as an aggregate is created or destroyed (which, in its turn, happens as a consequence of some class being created/destroyed/changed). However, to avoid the problems caused to service guarantees by these immediate decreases, QFQ+ increments the system virtual time using the maximum value allowed for the weight sum, 2^10, in place of the dynamic, instantaneous value. The instantaneous value of the weight sum is used only to check whether a request of weight increase or a class creation can be satisfied. Unfortunately, the problems caused by this choice are worse than the temporary degradation of the service guarantees that may occur, when a class is changed or destroyed, if the instantaneous value of the weight sum was used to update the system virtual time. In fact, the fraction of the link bandwidth guaranteed by QFQ+ to each aggregate is equal to the ratio between the weight of the aggregate and the sum of the weights of the competing aggregates. The packet delay guaranteed to the aggregate is instead inversely proportional to the guaranteed bandwidth. By using the maximum possible value, and not the actual value of the weight sum, QFQ+ provides each aggregate with the worst possible service guarantees, and not with service guarantees related to the actual set of competing aggregates. To see the consequences of this fact, consider the following simple example. Suppose that only the following aggregates are backlogged, i.e., that only the classes in the following aggregates have packets to transmit: one aggregate with weight 10, say A, and ten aggregates with weight 1, say B1, B2, ..., B10. In particular, suppose that these aggregates are always backlogged. Given the weight distribution, the smoothest and fairest service order would be: A B1 A B2 A B3 A B4 A B5 A B6 A B7 A B8 A B9 A B10 A B1 A B2 ... QFQ+ would provide exactly this optimal service if it used the actual value for the weight sum instead of the maximum possible value, i.e., 11 instead of 2^10. In contrast, since QFQ+ uses the latter value, it serves aggregates as follows (easy to prove and to reproduce experimentally): A B1 B2 B3 B4 B5 B6 B7 B8 B9 B10 A A A A A A A A A A B1 B2 ... B10 A A ... By replacing 10 with N in the above example, and by increasing N, one can increase at will the maximum packet delay and the jitter experienced by the classes in aggregate A. This patch addresses this issue by just using the above 'instantaneous' value of the weight sum, instead of the maximum possible value, when updating the system virtual time. After the instantaneous weight sum is decreased, QFQ+ may deviate from the ideal service for a time interval in the order of the time to serve one maximum-size packet for each backlogged class. The worst-case extent of the deviation exhibited by QFQ+ during this time interval [1] is basically the same as of the deviation described above (but, without this patch, QFQ+ suffers from such a deviation all the time). Finally, this patch modifies the comment to the function qfq_slot_insert, to make it coherent with the fact that the weight sum used by QFQ+ can now be lower than the maximum possible value. [1] P. Valente, "Extending WF2Q+ to support a dynamic traffic mix", Proceedings of AAA-IDEA'05, June 2005. Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-07-16 06:52:30 +00:00
* The following function deals with this problem by backward-shifting
* the timestamps of agg, if needed, so as to guarantee that the slot
* index is never higher than QFQ_MAX_SLOTS-2. This backward-shift may
* cause the service of other aggregates to be postponed, yet the
* worst-case guarantees of these aggregates are not violated. In
* fact, in case of no out-of-order service, the timestamps of agg
* would have been even lower than they are after the backward shift,
* because QFQ+ would have guaranteed a maximum value equal to 2 for
* the slot index, and 2 < QFQ_MAX_SLOTS-2. Hence the aggregates whose
* service is postponed because of the backward-shift would have
* however waited for the service of agg before being served.
pkt_sched: enable QFQ to support TSO/GSO If the max packet size for some class (configured through tc) is violated by the actual size of the packets of that class, then QFQ would not schedule classes correctly, and the data structures implementing the bucket lists may get corrupted. This problem occurs with TSO/GSO even if the max packet size is set to the MTU, and is, e.g., the cause of the failure reported in [1]. Two patches have been proposed to solve this problem in [2], one of them is a preliminary version of this patch. This patch addresses the above issues by: 1) setting QFQ parameters to proper values for supporting TSO/GSO (in particular, setting the maximum possible packet size to 64KB), 2) automatically increasing the max packet size for a class, lmax, when a packet with a larger size than the current value of lmax arrives. The drawback of the first point is that the maximum weight for a class is now limited to 4096, which is equal to 1/16 of the maximum weight sum. Finally, this patch also forcibly caps the timestamps of a class if they are too high to be stored in the bucket list. This capping, taken from QFQ+ [3], handles the unfrequent case described in the comment to the function slot_insert. [1] http://marc.info/?l=linux-netdev&m=134968777902077&w=2 [2] http://marc.info/?l=linux-netdev&m=135096573507936&w=2 [3] http://marc.info/?l=linux-netdev&m=134902691421670&w=2 Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Tested-by: Cong Wang <amwang@redhat.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-11-05 20:29:24 +00:00
*
pkt_sched: sch_qfq: remove a source of high packet delay/jitter QFQ+ inherits from QFQ a design choice that may cause a high packet delay/jitter and a severe short-term unfairness. As QFQ, QFQ+ uses a special quantity, the system virtual time, to track the service provided by the ideal system it approximates. When a packet is dequeued, this quantity must be incremented by the size of the packet, divided by the sum of the weights of the aggregates waiting to be served. Tracking this sum correctly is a non-trivial task, because, to preserve tight service guarantees, the decrement of this sum must be delayed in a special way [1]: this sum can be decremented only after that its value would decrease also in the ideal system approximated by QFQ+. For efficiency, QFQ+ keeps track only of the 'instantaneous' weight sum, increased and decreased immediately as the weight of an aggregate changes, and as an aggregate is created or destroyed (which, in its turn, happens as a consequence of some class being created/destroyed/changed). However, to avoid the problems caused to service guarantees by these immediate decreases, QFQ+ increments the system virtual time using the maximum value allowed for the weight sum, 2^10, in place of the dynamic, instantaneous value. The instantaneous value of the weight sum is used only to check whether a request of weight increase or a class creation can be satisfied. Unfortunately, the problems caused by this choice are worse than the temporary degradation of the service guarantees that may occur, when a class is changed or destroyed, if the instantaneous value of the weight sum was used to update the system virtual time. In fact, the fraction of the link bandwidth guaranteed by QFQ+ to each aggregate is equal to the ratio between the weight of the aggregate and the sum of the weights of the competing aggregates. The packet delay guaranteed to the aggregate is instead inversely proportional to the guaranteed bandwidth. By using the maximum possible value, and not the actual value of the weight sum, QFQ+ provides each aggregate with the worst possible service guarantees, and not with service guarantees related to the actual set of competing aggregates. To see the consequences of this fact, consider the following simple example. Suppose that only the following aggregates are backlogged, i.e., that only the classes in the following aggregates have packets to transmit: one aggregate with weight 10, say A, and ten aggregates with weight 1, say B1, B2, ..., B10. In particular, suppose that these aggregates are always backlogged. Given the weight distribution, the smoothest and fairest service order would be: A B1 A B2 A B3 A B4 A B5 A B6 A B7 A B8 A B9 A B10 A B1 A B2 ... QFQ+ would provide exactly this optimal service if it used the actual value for the weight sum instead of the maximum possible value, i.e., 11 instead of 2^10. In contrast, since QFQ+ uses the latter value, it serves aggregates as follows (easy to prove and to reproduce experimentally): A B1 B2 B3 B4 B5 B6 B7 B8 B9 B10 A A A A A A A A A A B1 B2 ... B10 A A ... By replacing 10 with N in the above example, and by increasing N, one can increase at will the maximum packet delay and the jitter experienced by the classes in aggregate A. This patch addresses this issue by just using the above 'instantaneous' value of the weight sum, instead of the maximum possible value, when updating the system virtual time. After the instantaneous weight sum is decreased, QFQ+ may deviate from the ideal service for a time interval in the order of the time to serve one maximum-size packet for each backlogged class. The worst-case extent of the deviation exhibited by QFQ+ during this time interval [1] is basically the same as of the deviation described above (but, without this patch, QFQ+ suffers from such a deviation all the time). Finally, this patch modifies the comment to the function qfq_slot_insert, to make it coherent with the fact that the weight sum used by QFQ+ can now be lower than the maximum possible value. [1] P. Valente, "Extending WF2Q+ to support a dynamic traffic mix", Proceedings of AAA-IDEA'05, June 2005. Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-07-16 06:52:30 +00:00
* The other event that may cause the slot index to be higher than 2
* for agg is a recent change of the parameters of some class. If the
* weight of a class is increased or the lmax (max_pkt_size) of the
* class is decreased, then a new aggregate with smaller slot size
* than the original parent aggregate of the class may happen to be
* activated. The activation of this aggregate should be properly
* delayed to when the service of the class has finished in the ideal
* system tracked by QFQ+. If the activation of the aggregate is not
* delayed to this reference time instant, then this aggregate may be
* unjustly served before other aggregates waiting for service. This
* may cause the above bound to the slot index to be violated for some
* of these unlucky aggregates.
pkt_sched: enable QFQ to support TSO/GSO If the max packet size for some class (configured through tc) is violated by the actual size of the packets of that class, then QFQ would not schedule classes correctly, and the data structures implementing the bucket lists may get corrupted. This problem occurs with TSO/GSO even if the max packet size is set to the MTU, and is, e.g., the cause of the failure reported in [1]. Two patches have been proposed to solve this problem in [2], one of them is a preliminary version of this patch. This patch addresses the above issues by: 1) setting QFQ parameters to proper values for supporting TSO/GSO (in particular, setting the maximum possible packet size to 64KB), 2) automatically increasing the max packet size for a class, lmax, when a packet with a larger size than the current value of lmax arrives. The drawback of the first point is that the maximum weight for a class is now limited to 4096, which is equal to 1/16 of the maximum weight sum. Finally, this patch also forcibly caps the timestamps of a class if they are too high to be stored in the bucket list. This capping, taken from QFQ+ [3], handles the unfrequent case described in the comment to the function slot_insert. [1] http://marc.info/?l=linux-netdev&m=134968777902077&w=2 [2] http://marc.info/?l=linux-netdev&m=135096573507936&w=2 [3] http://marc.info/?l=linux-netdev&m=134902691421670&w=2 Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Tested-by: Cong Wang <amwang@redhat.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-11-05 20:29:24 +00:00
*
* Instead of delaying the activation of the new aggregate, which is
pkt_sched: sch_qfq: remove a source of high packet delay/jitter QFQ+ inherits from QFQ a design choice that may cause a high packet delay/jitter and a severe short-term unfairness. As QFQ, QFQ+ uses a special quantity, the system virtual time, to track the service provided by the ideal system it approximates. When a packet is dequeued, this quantity must be incremented by the size of the packet, divided by the sum of the weights of the aggregates waiting to be served. Tracking this sum correctly is a non-trivial task, because, to preserve tight service guarantees, the decrement of this sum must be delayed in a special way [1]: this sum can be decremented only after that its value would decrease also in the ideal system approximated by QFQ+. For efficiency, QFQ+ keeps track only of the 'instantaneous' weight sum, increased and decreased immediately as the weight of an aggregate changes, and as an aggregate is created or destroyed (which, in its turn, happens as a consequence of some class being created/destroyed/changed). However, to avoid the problems caused to service guarantees by these immediate decreases, QFQ+ increments the system virtual time using the maximum value allowed for the weight sum, 2^10, in place of the dynamic, instantaneous value. The instantaneous value of the weight sum is used only to check whether a request of weight increase or a class creation can be satisfied. Unfortunately, the problems caused by this choice are worse than the temporary degradation of the service guarantees that may occur, when a class is changed or destroyed, if the instantaneous value of the weight sum was used to update the system virtual time. In fact, the fraction of the link bandwidth guaranteed by QFQ+ to each aggregate is equal to the ratio between the weight of the aggregate and the sum of the weights of the competing aggregates. The packet delay guaranteed to the aggregate is instead inversely proportional to the guaranteed bandwidth. By using the maximum possible value, and not the actual value of the weight sum, QFQ+ provides each aggregate with the worst possible service guarantees, and not with service guarantees related to the actual set of competing aggregates. To see the consequences of this fact, consider the following simple example. Suppose that only the following aggregates are backlogged, i.e., that only the classes in the following aggregates have packets to transmit: one aggregate with weight 10, say A, and ten aggregates with weight 1, say B1, B2, ..., B10. In particular, suppose that these aggregates are always backlogged. Given the weight distribution, the smoothest and fairest service order would be: A B1 A B2 A B3 A B4 A B5 A B6 A B7 A B8 A B9 A B10 A B1 A B2 ... QFQ+ would provide exactly this optimal service if it used the actual value for the weight sum instead of the maximum possible value, i.e., 11 instead of 2^10. In contrast, since QFQ+ uses the latter value, it serves aggregates as follows (easy to prove and to reproduce experimentally): A B1 B2 B3 B4 B5 B6 B7 B8 B9 B10 A A A A A A A A A A B1 B2 ... B10 A A ... By replacing 10 with N in the above example, and by increasing N, one can increase at will the maximum packet delay and the jitter experienced by the classes in aggregate A. This patch addresses this issue by just using the above 'instantaneous' value of the weight sum, instead of the maximum possible value, when updating the system virtual time. After the instantaneous weight sum is decreased, QFQ+ may deviate from the ideal service for a time interval in the order of the time to serve one maximum-size packet for each backlogged class. The worst-case extent of the deviation exhibited by QFQ+ during this time interval [1] is basically the same as of the deviation described above (but, without this patch, QFQ+ suffers from such a deviation all the time). Finally, this patch modifies the comment to the function qfq_slot_insert, to make it coherent with the fact that the weight sum used by QFQ+ can now be lower than the maximum possible value. [1] P. Valente, "Extending WF2Q+ to support a dynamic traffic mix", Proceedings of AAA-IDEA'05, June 2005. Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-07-16 06:52:30 +00:00
* quite complex, the above-discussed capping of the slot index is
* used to handle also the consequences of a change of the parameters
* of a class.
*/
static void qfq_slot_insert(struct qfq_group *grp, struct qfq_aggregate *agg,
u64 roundedS)
{
u64 slot = (roundedS - grp->S) >> grp->slot_shift;
pkt_sched: enable QFQ to support TSO/GSO If the max packet size for some class (configured through tc) is violated by the actual size of the packets of that class, then QFQ would not schedule classes correctly, and the data structures implementing the bucket lists may get corrupted. This problem occurs with TSO/GSO even if the max packet size is set to the MTU, and is, e.g., the cause of the failure reported in [1]. Two patches have been proposed to solve this problem in [2], one of them is a preliminary version of this patch. This patch addresses the above issues by: 1) setting QFQ parameters to proper values for supporting TSO/GSO (in particular, setting the maximum possible packet size to 64KB), 2) automatically increasing the max packet size for a class, lmax, when a packet with a larger size than the current value of lmax arrives. The drawback of the first point is that the maximum weight for a class is now limited to 4096, which is equal to 1/16 of the maximum weight sum. Finally, this patch also forcibly caps the timestamps of a class if they are too high to be stored in the bucket list. This capping, taken from QFQ+ [3], handles the unfrequent case described in the comment to the function slot_insert. [1] http://marc.info/?l=linux-netdev&m=134968777902077&w=2 [2] http://marc.info/?l=linux-netdev&m=135096573507936&w=2 [3] http://marc.info/?l=linux-netdev&m=134902691421670&w=2 Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Tested-by: Cong Wang <amwang@redhat.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-11-05 20:29:24 +00:00
unsigned int i; /* slot index in the bucket list */
if (unlikely(slot > QFQ_MAX_SLOTS - 2)) {
u64 deltaS = roundedS - grp->S -
((u64)(QFQ_MAX_SLOTS - 2)<<grp->slot_shift);
agg->S -= deltaS;
agg->F -= deltaS;
pkt_sched: enable QFQ to support TSO/GSO If the max packet size for some class (configured through tc) is violated by the actual size of the packets of that class, then QFQ would not schedule classes correctly, and the data structures implementing the bucket lists may get corrupted. This problem occurs with TSO/GSO even if the max packet size is set to the MTU, and is, e.g., the cause of the failure reported in [1]. Two patches have been proposed to solve this problem in [2], one of them is a preliminary version of this patch. This patch addresses the above issues by: 1) setting QFQ parameters to proper values for supporting TSO/GSO (in particular, setting the maximum possible packet size to 64KB), 2) automatically increasing the max packet size for a class, lmax, when a packet with a larger size than the current value of lmax arrives. The drawback of the first point is that the maximum weight for a class is now limited to 4096, which is equal to 1/16 of the maximum weight sum. Finally, this patch also forcibly caps the timestamps of a class if they are too high to be stored in the bucket list. This capping, taken from QFQ+ [3], handles the unfrequent case described in the comment to the function slot_insert. [1] http://marc.info/?l=linux-netdev&m=134968777902077&w=2 [2] http://marc.info/?l=linux-netdev&m=135096573507936&w=2 [3] http://marc.info/?l=linux-netdev&m=134902691421670&w=2 Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Tested-by: Cong Wang <amwang@redhat.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-11-05 20:29:24 +00:00
slot = QFQ_MAX_SLOTS - 2;
}
i = (grp->front + slot) % QFQ_MAX_SLOTS;
hlist_add_head(&agg->next, &grp->slots[i]);
__set_bit(slot, &grp->full_slots);
}
/* Maybe introduce hlist_first_entry?? */
static struct qfq_aggregate *qfq_slot_head(struct qfq_group *grp)
{
return hlist_entry(grp->slots[grp->front].first,
struct qfq_aggregate, next);
}
/*
* remove the entry from the slot
*/
static void qfq_front_slot_remove(struct qfq_group *grp)
{
struct qfq_aggregate *agg = qfq_slot_head(grp);
BUG_ON(!agg);
hlist_del(&agg->next);
if (hlist_empty(&grp->slots[grp->front]))
__clear_bit(0, &grp->full_slots);
}
/*
* Returns the first aggregate in the first non-empty bucket of the
* group. As a side effect, adjusts the bucket list so the first
* non-empty bucket is at position 0 in full_slots.
*/
static struct qfq_aggregate *qfq_slot_scan(struct qfq_group *grp)
{
unsigned int i;
pr_debug("qfq slot_scan: grp %u full %#lx\n",
grp->index, grp->full_slots);
if (grp->full_slots == 0)
return NULL;
i = __ffs(grp->full_slots); /* zero based */
if (i > 0) {
grp->front = (grp->front + i) % QFQ_MAX_SLOTS;
grp->full_slots >>= i;
}
return qfq_slot_head(grp);
}
/*
* adjust the bucket list. When the start time of a group decreases,
* we move the index down (modulo QFQ_MAX_SLOTS) so we don't need to
* move the objects. The mask of occupied slots must be shifted
* because we use ffs() to find the first non-empty slot.
* This covers decreases in the group's start time, but what about
* increases of the start time ?
* Here too we should make sure that i is less than 32
*/
static void qfq_slot_rotate(struct qfq_group *grp, u64 roundedS)
{
unsigned int i = (grp->S - roundedS) >> grp->slot_shift;
grp->full_slots <<= i;
grp->front = (grp->front - i) % QFQ_MAX_SLOTS;
}
static void qfq_update_eligible(struct qfq_sched *q)
{
struct qfq_group *grp;
unsigned long ineligible;
ineligible = q->bitmaps[IR] | q->bitmaps[IB];
if (ineligible) {
if (!q->bitmaps[ER]) {
grp = qfq_ffs(q, ineligible);
if (qfq_gt(grp->S, q->V))
q->V = grp->S;
}
qfq_make_eligible(q);
}
}
/* Dequeue head packet of the head class in the DRR queue of the aggregate. */
net: sched: sch_qfq: Fix UAF in qfq_dequeue() When the plug qdisc is used as a class of the qfq qdisc it could trigger a UAF. This issue can be reproduced with following commands: tc qdisc add dev lo root handle 1: qfq tc class add dev lo parent 1: classid 1:1 qfq weight 1 maxpkt 512 tc qdisc add dev lo parent 1:1 handle 2: plug tc filter add dev lo parent 1: basic classid 1:1 ping -c1 127.0.0.1 and boom: [ 285.353793] BUG: KASAN: slab-use-after-free in qfq_dequeue+0xa7/0x7f0 [ 285.354910] Read of size 4 at addr ffff8880bad312a8 by task ping/144 [ 285.355903] [ 285.356165] CPU: 1 PID: 144 Comm: ping Not tainted 6.5.0-rc3+ #4 [ 285.357112] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.14.0-2 04/01/2014 [ 285.358376] Call Trace: [ 285.358773] <IRQ> [ 285.359109] dump_stack_lvl+0x44/0x60 [ 285.359708] print_address_description.constprop.0+0x2c/0x3c0 [ 285.360611] kasan_report+0x10c/0x120 [ 285.361195] ? qfq_dequeue+0xa7/0x7f0 [ 285.361780] qfq_dequeue+0xa7/0x7f0 [ 285.362342] __qdisc_run+0xf1/0x970 [ 285.362903] net_tx_action+0x28e/0x460 [ 285.363502] __do_softirq+0x11b/0x3de [ 285.364097] do_softirq.part.0+0x72/0x90 [ 285.364721] </IRQ> [ 285.365072] <TASK> [ 285.365422] __local_bh_enable_ip+0x77/0x90 [ 285.366079] __dev_queue_xmit+0x95f/0x1550 [ 285.366732] ? __pfx_csum_and_copy_from_iter+0x10/0x10 [ 285.367526] ? __pfx___dev_queue_xmit+0x10/0x10 [ 285.368259] ? __build_skb_around+0x129/0x190 [ 285.368960] ? ip_generic_getfrag+0x12c/0x170 [ 285.369653] ? __pfx_ip_generic_getfrag+0x10/0x10 [ 285.370390] ? csum_partial+0x8/0x20 [ 285.370961] ? raw_getfrag+0xe5/0x140 [ 285.371559] ip_finish_output2+0x539/0xa40 [ 285.372222] ? __pfx_ip_finish_output2+0x10/0x10 [ 285.372954] ip_output+0x113/0x1e0 [ 285.373512] ? __pfx_ip_output+0x10/0x10 [ 285.374130] ? icmp_out_count+0x49/0x60 [ 285.374739] ? __pfx_ip_finish_output+0x10/0x10 [ 285.375457] ip_push_pending_frames+0xf3/0x100 [ 285.376173] raw_sendmsg+0xef5/0x12d0 [ 285.376760] ? do_syscall_64+0x40/0x90 [ 285.377359] ? __static_call_text_end+0x136578/0x136578 [ 285.378173] ? do_syscall_64+0x40/0x90 [ 285.378772] ? kasan_enable_current+0x11/0x20 [ 285.379469] ? __pfx_raw_sendmsg+0x10/0x10 [ 285.380137] ? __sock_create+0x13e/0x270 [ 285.380673] ? __sys_socket+0xf3/0x180 [ 285.381174] ? __x64_sys_socket+0x3d/0x50 [ 285.381725] ? entry_SYSCALL_64_after_hwframe+0x6e/0xd8 [ 285.382425] ? __rcu_read_unlock+0x48/0x70 [ 285.382975] ? ip4_datagram_release_cb+0xd8/0x380 [ 285.383608] ? __pfx_ip4_datagram_release_cb+0x10/0x10 [ 285.384295] ? preempt_count_sub+0x14/0xc0 [ 285.384844] ? __list_del_entry_valid+0x76/0x140 [ 285.385467] ? _raw_spin_lock_bh+0x87/0xe0 [ 285.386014] ? __pfx__raw_spin_lock_bh+0x10/0x10 [ 285.386645] ? release_sock+0xa0/0xd0 [ 285.387148] ? preempt_count_sub+0x14/0xc0 [ 285.387712] ? freeze_secondary_cpus+0x348/0x3c0 [ 285.388341] ? aa_sk_perm+0x177/0x390 [ 285.388856] ? __pfx_aa_sk_perm+0x10/0x10 [ 285.389441] ? check_stack_object+0x22/0x70 [ 285.390032] ? inet_send_prepare+0x2f/0x120 [ 285.390603] ? __pfx_inet_sendmsg+0x10/0x10 [ 285.391172] sock_sendmsg+0xcc/0xe0 [ 285.391667] __sys_sendto+0x190/0x230 [ 285.392168] ? __pfx___sys_sendto+0x10/0x10 [ 285.392727] ? kvm_clock_get_cycles+0x14/0x30 [ 285.393328] ? set_normalized_timespec64+0x57/0x70 [ 285.393980] ? _raw_spin_unlock_irq+0x1b/0x40 [ 285.394578] ? __x64_sys_clock_gettime+0x11c/0x160 [ 285.395225] ? __pfx___x64_sys_clock_gettime+0x10/0x10 [ 285.395908] ? _copy_to_user+0x3e/0x60 [ 285.396432] ? exit_to_user_mode_prepare+0x1a/0x120 [ 285.397086] ? syscall_exit_to_user_mode+0x22/0x50 [ 285.397734] ? do_syscall_64+0x71/0x90 [ 285.398258] __x64_sys_sendto+0x74/0x90 [ 285.398786] do_syscall_64+0x64/0x90 [ 285.399273] ? exit_to_user_mode_prepare+0x1a/0x120 [ 285.399949] ? syscall_exit_to_user_mode+0x22/0x50 [ 285.400605] ? do_syscall_64+0x71/0x90 [ 285.401124] entry_SYSCALL_64_after_hwframe+0x6e/0xd8 [ 285.401807] RIP: 0033:0x495726 [ 285.402233] Code: ff ff ff f7 d8 64 89 02 48 c7 c0 ff ff ff ff eb b8 0f 1f 00 41 89 ca 64 8b 04 25 18 00 00 00 85 c0 75 11 b8 2c 00 00 00 0f 09 [ 285.404683] RSP: 002b:00007ffcc25fb618 EFLAGS: 00000246 ORIG_RAX: 000000000000002c [ 285.405677] RAX: ffffffffffffffda RBX: 0000000000000040 RCX: 0000000000495726 [ 285.406628] RDX: 0000000000000040 RSI: 0000000002518750 RDI: 0000000000000000 [ 285.407565] RBP: 00000000005205ef R08: 00000000005f8838 R09: 000000000000001c [ 285.408523] R10: 0000000000000000 R11: 0000000000000246 R12: 0000000002517634 [ 285.409460] R13: 00007ffcc25fb6f0 R14: 0000000000000003 R15: 0000000000000000 [ 285.410403] </TASK> [ 285.410704] [ 285.410929] Allocated by task 144: [ 285.411402] kasan_save_stack+0x1e/0x40 [ 285.411926] kasan_set_track+0x21/0x30 [ 285.412442] __kasan_slab_alloc+0x55/0x70 [ 285.412973] kmem_cache_alloc_node+0x187/0x3d0 [ 285.413567] __alloc_skb+0x1b4/0x230 [ 285.414060] __ip_append_data+0x17f7/0x1b60 [ 285.414633] ip_append_data+0x97/0xf0 [ 285.415144] raw_sendmsg+0x5a8/0x12d0 [ 285.415640] sock_sendmsg+0xcc/0xe0 [ 285.416117] __sys_sendto+0x190/0x230 [ 285.416626] __x64_sys_sendto+0x74/0x90 [ 285.417145] do_syscall_64+0x64/0x90 [ 285.417624] entry_SYSCALL_64_after_hwframe+0x6e/0xd8 [ 285.418306] [ 285.418531] Freed by task 144: [ 285.418960] kasan_save_stack+0x1e/0x40 [ 285.419469] kasan_set_track+0x21/0x30 [ 285.419988] kasan_save_free_info+0x27/0x40 [ 285.420556] ____kasan_slab_free+0x109/0x1a0 [ 285.421146] kmem_cache_free+0x1c2/0x450 [ 285.421680] __netif_receive_skb_core+0x2ce/0x1870 [ 285.422333] __netif_receive_skb_one_core+0x97/0x140 [ 285.423003] process_backlog+0x100/0x2f0 [ 285.423537] __napi_poll+0x5c/0x2d0 [ 285.424023] net_rx_action+0x2be/0x560 [ 285.424510] __do_softirq+0x11b/0x3de [ 285.425034] [ 285.425254] The buggy address belongs to the object at ffff8880bad31280 [ 285.425254] which belongs to the cache skbuff_head_cache of size 224 [ 285.426993] The buggy address is located 40 bytes inside of [ 285.426993] freed 224-byte region [ffff8880bad31280, ffff8880bad31360) [ 285.428572] [ 285.428798] The buggy address belongs to the physical page: [ 285.429540] page:00000000f4b77674 refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0xbad31 [ 285.430758] flags: 0x100000000000200(slab|node=0|zone=1) [ 285.431447] page_type: 0xffffffff() [ 285.431934] raw: 0100000000000200 ffff88810094a8c0 dead000000000122 0000000000000000 [ 285.432757] raw: 0000000000000000 00000000800c000c 00000001ffffffff 0000000000000000 [ 285.433562] page dumped because: kasan: bad access detected [ 285.434144] [ 285.434320] Memory state around the buggy address: [ 285.434828] ffff8880bad31180: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc [ 285.435580] ffff8880bad31200: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc [ 285.436264] >ffff8880bad31280: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb [ 285.436777] ^ [ 285.437106] ffff8880bad31300: fb fb fb fb fb fb fb fb fb fb fb fb fc fc fc fc [ 285.437616] ffff8880bad31380: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc [ 285.438126] ================================================================== [ 285.438662] Disabling lock debugging due to kernel taint Fix this by: 1. Changing sch_plug's .peek handler to qdisc_peek_dequeued(), a function compatible with non-work-conserving qdiscs 2. Checking the return value of qdisc_dequeue_peeked() in sch_qfq. Fixes: 462dbc9101ac ("pkt_sched: QFQ Plus: fair-queueing service at DRR cost") Reported-by: valis <sec@valis.email> Signed-off-by: valis <sec@valis.email> Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com> Link: https://lore.kernel.org/r/20230901162237.11525-1-jhs@mojatatu.com Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2023-09-01 16:22:37 +00:00
static struct sk_buff *agg_dequeue(struct qfq_aggregate *agg,
struct qfq_class *cl, unsigned int len)
{
net: sched: sch_qfq: Fix UAF in qfq_dequeue() When the plug qdisc is used as a class of the qfq qdisc it could trigger a UAF. This issue can be reproduced with following commands: tc qdisc add dev lo root handle 1: qfq tc class add dev lo parent 1: classid 1:1 qfq weight 1 maxpkt 512 tc qdisc add dev lo parent 1:1 handle 2: plug tc filter add dev lo parent 1: basic classid 1:1 ping -c1 127.0.0.1 and boom: [ 285.353793] BUG: KASAN: slab-use-after-free in qfq_dequeue+0xa7/0x7f0 [ 285.354910] Read of size 4 at addr ffff8880bad312a8 by task ping/144 [ 285.355903] [ 285.356165] CPU: 1 PID: 144 Comm: ping Not tainted 6.5.0-rc3+ #4 [ 285.357112] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.14.0-2 04/01/2014 [ 285.358376] Call Trace: [ 285.358773] <IRQ> [ 285.359109] dump_stack_lvl+0x44/0x60 [ 285.359708] print_address_description.constprop.0+0x2c/0x3c0 [ 285.360611] kasan_report+0x10c/0x120 [ 285.361195] ? qfq_dequeue+0xa7/0x7f0 [ 285.361780] qfq_dequeue+0xa7/0x7f0 [ 285.362342] __qdisc_run+0xf1/0x970 [ 285.362903] net_tx_action+0x28e/0x460 [ 285.363502] __do_softirq+0x11b/0x3de [ 285.364097] do_softirq.part.0+0x72/0x90 [ 285.364721] </IRQ> [ 285.365072] <TASK> [ 285.365422] __local_bh_enable_ip+0x77/0x90 [ 285.366079] __dev_queue_xmit+0x95f/0x1550 [ 285.366732] ? __pfx_csum_and_copy_from_iter+0x10/0x10 [ 285.367526] ? __pfx___dev_queue_xmit+0x10/0x10 [ 285.368259] ? __build_skb_around+0x129/0x190 [ 285.368960] ? ip_generic_getfrag+0x12c/0x170 [ 285.369653] ? __pfx_ip_generic_getfrag+0x10/0x10 [ 285.370390] ? csum_partial+0x8/0x20 [ 285.370961] ? raw_getfrag+0xe5/0x140 [ 285.371559] ip_finish_output2+0x539/0xa40 [ 285.372222] ? __pfx_ip_finish_output2+0x10/0x10 [ 285.372954] ip_output+0x113/0x1e0 [ 285.373512] ? __pfx_ip_output+0x10/0x10 [ 285.374130] ? icmp_out_count+0x49/0x60 [ 285.374739] ? __pfx_ip_finish_output+0x10/0x10 [ 285.375457] ip_push_pending_frames+0xf3/0x100 [ 285.376173] raw_sendmsg+0xef5/0x12d0 [ 285.376760] ? do_syscall_64+0x40/0x90 [ 285.377359] ? __static_call_text_end+0x136578/0x136578 [ 285.378173] ? do_syscall_64+0x40/0x90 [ 285.378772] ? kasan_enable_current+0x11/0x20 [ 285.379469] ? __pfx_raw_sendmsg+0x10/0x10 [ 285.380137] ? __sock_create+0x13e/0x270 [ 285.380673] ? __sys_socket+0xf3/0x180 [ 285.381174] ? __x64_sys_socket+0x3d/0x50 [ 285.381725] ? entry_SYSCALL_64_after_hwframe+0x6e/0xd8 [ 285.382425] ? __rcu_read_unlock+0x48/0x70 [ 285.382975] ? ip4_datagram_release_cb+0xd8/0x380 [ 285.383608] ? __pfx_ip4_datagram_release_cb+0x10/0x10 [ 285.384295] ? preempt_count_sub+0x14/0xc0 [ 285.384844] ? __list_del_entry_valid+0x76/0x140 [ 285.385467] ? _raw_spin_lock_bh+0x87/0xe0 [ 285.386014] ? __pfx__raw_spin_lock_bh+0x10/0x10 [ 285.386645] ? release_sock+0xa0/0xd0 [ 285.387148] ? preempt_count_sub+0x14/0xc0 [ 285.387712] ? freeze_secondary_cpus+0x348/0x3c0 [ 285.388341] ? aa_sk_perm+0x177/0x390 [ 285.388856] ? __pfx_aa_sk_perm+0x10/0x10 [ 285.389441] ? check_stack_object+0x22/0x70 [ 285.390032] ? inet_send_prepare+0x2f/0x120 [ 285.390603] ? __pfx_inet_sendmsg+0x10/0x10 [ 285.391172] sock_sendmsg+0xcc/0xe0 [ 285.391667] __sys_sendto+0x190/0x230 [ 285.392168] ? __pfx___sys_sendto+0x10/0x10 [ 285.392727] ? kvm_clock_get_cycles+0x14/0x30 [ 285.393328] ? set_normalized_timespec64+0x57/0x70 [ 285.393980] ? _raw_spin_unlock_irq+0x1b/0x40 [ 285.394578] ? __x64_sys_clock_gettime+0x11c/0x160 [ 285.395225] ? __pfx___x64_sys_clock_gettime+0x10/0x10 [ 285.395908] ? _copy_to_user+0x3e/0x60 [ 285.396432] ? exit_to_user_mode_prepare+0x1a/0x120 [ 285.397086] ? syscall_exit_to_user_mode+0x22/0x50 [ 285.397734] ? do_syscall_64+0x71/0x90 [ 285.398258] __x64_sys_sendto+0x74/0x90 [ 285.398786] do_syscall_64+0x64/0x90 [ 285.399273] ? exit_to_user_mode_prepare+0x1a/0x120 [ 285.399949] ? syscall_exit_to_user_mode+0x22/0x50 [ 285.400605] ? do_syscall_64+0x71/0x90 [ 285.401124] entry_SYSCALL_64_after_hwframe+0x6e/0xd8 [ 285.401807] RIP: 0033:0x495726 [ 285.402233] Code: ff ff ff f7 d8 64 89 02 48 c7 c0 ff ff ff ff eb b8 0f 1f 00 41 89 ca 64 8b 04 25 18 00 00 00 85 c0 75 11 b8 2c 00 00 00 0f 09 [ 285.404683] RSP: 002b:00007ffcc25fb618 EFLAGS: 00000246 ORIG_RAX: 000000000000002c [ 285.405677] RAX: ffffffffffffffda RBX: 0000000000000040 RCX: 0000000000495726 [ 285.406628] RDX: 0000000000000040 RSI: 0000000002518750 RDI: 0000000000000000 [ 285.407565] RBP: 00000000005205ef R08: 00000000005f8838 R09: 000000000000001c [ 285.408523] R10: 0000000000000000 R11: 0000000000000246 R12: 0000000002517634 [ 285.409460] R13: 00007ffcc25fb6f0 R14: 0000000000000003 R15: 0000000000000000 [ 285.410403] </TASK> [ 285.410704] [ 285.410929] Allocated by task 144: [ 285.411402] kasan_save_stack+0x1e/0x40 [ 285.411926] kasan_set_track+0x21/0x30 [ 285.412442] __kasan_slab_alloc+0x55/0x70 [ 285.412973] kmem_cache_alloc_node+0x187/0x3d0 [ 285.413567] __alloc_skb+0x1b4/0x230 [ 285.414060] __ip_append_data+0x17f7/0x1b60 [ 285.414633] ip_append_data+0x97/0xf0 [ 285.415144] raw_sendmsg+0x5a8/0x12d0 [ 285.415640] sock_sendmsg+0xcc/0xe0 [ 285.416117] __sys_sendto+0x190/0x230 [ 285.416626] __x64_sys_sendto+0x74/0x90 [ 285.417145] do_syscall_64+0x64/0x90 [ 285.417624] entry_SYSCALL_64_after_hwframe+0x6e/0xd8 [ 285.418306] [ 285.418531] Freed by task 144: [ 285.418960] kasan_save_stack+0x1e/0x40 [ 285.419469] kasan_set_track+0x21/0x30 [ 285.419988] kasan_save_free_info+0x27/0x40 [ 285.420556] ____kasan_slab_free+0x109/0x1a0 [ 285.421146] kmem_cache_free+0x1c2/0x450 [ 285.421680] __netif_receive_skb_core+0x2ce/0x1870 [ 285.422333] __netif_receive_skb_one_core+0x97/0x140 [ 285.423003] process_backlog+0x100/0x2f0 [ 285.423537] __napi_poll+0x5c/0x2d0 [ 285.424023] net_rx_action+0x2be/0x560 [ 285.424510] __do_softirq+0x11b/0x3de [ 285.425034] [ 285.425254] The buggy address belongs to the object at ffff8880bad31280 [ 285.425254] which belongs to the cache skbuff_head_cache of size 224 [ 285.426993] The buggy address is located 40 bytes inside of [ 285.426993] freed 224-byte region [ffff8880bad31280, ffff8880bad31360) [ 285.428572] [ 285.428798] The buggy address belongs to the physical page: [ 285.429540] page:00000000f4b77674 refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0xbad31 [ 285.430758] flags: 0x100000000000200(slab|node=0|zone=1) [ 285.431447] page_type: 0xffffffff() [ 285.431934] raw: 0100000000000200 ffff88810094a8c0 dead000000000122 0000000000000000 [ 285.432757] raw: 0000000000000000 00000000800c000c 00000001ffffffff 0000000000000000 [ 285.433562] page dumped because: kasan: bad access detected [ 285.434144] [ 285.434320] Memory state around the buggy address: [ 285.434828] ffff8880bad31180: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc [ 285.435580] ffff8880bad31200: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc [ 285.436264] >ffff8880bad31280: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb [ 285.436777] ^ [ 285.437106] ffff8880bad31300: fb fb fb fb fb fb fb fb fb fb fb fb fc fc fc fc [ 285.437616] ffff8880bad31380: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc [ 285.438126] ================================================================== [ 285.438662] Disabling lock debugging due to kernel taint Fix this by: 1. Changing sch_plug's .peek handler to qdisc_peek_dequeued(), a function compatible with non-work-conserving qdiscs 2. Checking the return value of qdisc_dequeue_peeked() in sch_qfq. Fixes: 462dbc9101ac ("pkt_sched: QFQ Plus: fair-queueing service at DRR cost") Reported-by: valis <sec@valis.email> Signed-off-by: valis <sec@valis.email> Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com> Link: https://lore.kernel.org/r/20230901162237.11525-1-jhs@mojatatu.com Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2023-09-01 16:22:37 +00:00
struct sk_buff *skb = qdisc_dequeue_peeked(cl->qdisc);
if (!skb)
return NULL;
cl->deficit -= (int) len;
if (cl->qdisc->q.qlen == 0) /* no more packets, remove from list */
list_del(&cl->alist);
else if (cl->deficit < qdisc_pkt_len(cl->qdisc->ops->peek(cl->qdisc))) {
cl->deficit += agg->lmax;
list_move_tail(&cl->alist, &agg->active);
}
net: sched: sch_qfq: Fix UAF in qfq_dequeue() When the plug qdisc is used as a class of the qfq qdisc it could trigger a UAF. This issue can be reproduced with following commands: tc qdisc add dev lo root handle 1: qfq tc class add dev lo parent 1: classid 1:1 qfq weight 1 maxpkt 512 tc qdisc add dev lo parent 1:1 handle 2: plug tc filter add dev lo parent 1: basic classid 1:1 ping -c1 127.0.0.1 and boom: [ 285.353793] BUG: KASAN: slab-use-after-free in qfq_dequeue+0xa7/0x7f0 [ 285.354910] Read of size 4 at addr ffff8880bad312a8 by task ping/144 [ 285.355903] [ 285.356165] CPU: 1 PID: 144 Comm: ping Not tainted 6.5.0-rc3+ #4 [ 285.357112] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.14.0-2 04/01/2014 [ 285.358376] Call Trace: [ 285.358773] <IRQ> [ 285.359109] dump_stack_lvl+0x44/0x60 [ 285.359708] print_address_description.constprop.0+0x2c/0x3c0 [ 285.360611] kasan_report+0x10c/0x120 [ 285.361195] ? qfq_dequeue+0xa7/0x7f0 [ 285.361780] qfq_dequeue+0xa7/0x7f0 [ 285.362342] __qdisc_run+0xf1/0x970 [ 285.362903] net_tx_action+0x28e/0x460 [ 285.363502] __do_softirq+0x11b/0x3de [ 285.364097] do_softirq.part.0+0x72/0x90 [ 285.364721] </IRQ> [ 285.365072] <TASK> [ 285.365422] __local_bh_enable_ip+0x77/0x90 [ 285.366079] __dev_queue_xmit+0x95f/0x1550 [ 285.366732] ? __pfx_csum_and_copy_from_iter+0x10/0x10 [ 285.367526] ? __pfx___dev_queue_xmit+0x10/0x10 [ 285.368259] ? __build_skb_around+0x129/0x190 [ 285.368960] ? ip_generic_getfrag+0x12c/0x170 [ 285.369653] ? __pfx_ip_generic_getfrag+0x10/0x10 [ 285.370390] ? csum_partial+0x8/0x20 [ 285.370961] ? raw_getfrag+0xe5/0x140 [ 285.371559] ip_finish_output2+0x539/0xa40 [ 285.372222] ? __pfx_ip_finish_output2+0x10/0x10 [ 285.372954] ip_output+0x113/0x1e0 [ 285.373512] ? __pfx_ip_output+0x10/0x10 [ 285.374130] ? icmp_out_count+0x49/0x60 [ 285.374739] ? __pfx_ip_finish_output+0x10/0x10 [ 285.375457] ip_push_pending_frames+0xf3/0x100 [ 285.376173] raw_sendmsg+0xef5/0x12d0 [ 285.376760] ? do_syscall_64+0x40/0x90 [ 285.377359] ? __static_call_text_end+0x136578/0x136578 [ 285.378173] ? do_syscall_64+0x40/0x90 [ 285.378772] ? kasan_enable_current+0x11/0x20 [ 285.379469] ? __pfx_raw_sendmsg+0x10/0x10 [ 285.380137] ? __sock_create+0x13e/0x270 [ 285.380673] ? __sys_socket+0xf3/0x180 [ 285.381174] ? __x64_sys_socket+0x3d/0x50 [ 285.381725] ? entry_SYSCALL_64_after_hwframe+0x6e/0xd8 [ 285.382425] ? __rcu_read_unlock+0x48/0x70 [ 285.382975] ? ip4_datagram_release_cb+0xd8/0x380 [ 285.383608] ? __pfx_ip4_datagram_release_cb+0x10/0x10 [ 285.384295] ? preempt_count_sub+0x14/0xc0 [ 285.384844] ? __list_del_entry_valid+0x76/0x140 [ 285.385467] ? _raw_spin_lock_bh+0x87/0xe0 [ 285.386014] ? __pfx__raw_spin_lock_bh+0x10/0x10 [ 285.386645] ? release_sock+0xa0/0xd0 [ 285.387148] ? preempt_count_sub+0x14/0xc0 [ 285.387712] ? freeze_secondary_cpus+0x348/0x3c0 [ 285.388341] ? aa_sk_perm+0x177/0x390 [ 285.388856] ? __pfx_aa_sk_perm+0x10/0x10 [ 285.389441] ? check_stack_object+0x22/0x70 [ 285.390032] ? inet_send_prepare+0x2f/0x120 [ 285.390603] ? __pfx_inet_sendmsg+0x10/0x10 [ 285.391172] sock_sendmsg+0xcc/0xe0 [ 285.391667] __sys_sendto+0x190/0x230 [ 285.392168] ? __pfx___sys_sendto+0x10/0x10 [ 285.392727] ? kvm_clock_get_cycles+0x14/0x30 [ 285.393328] ? set_normalized_timespec64+0x57/0x70 [ 285.393980] ? _raw_spin_unlock_irq+0x1b/0x40 [ 285.394578] ? __x64_sys_clock_gettime+0x11c/0x160 [ 285.395225] ? __pfx___x64_sys_clock_gettime+0x10/0x10 [ 285.395908] ? _copy_to_user+0x3e/0x60 [ 285.396432] ? exit_to_user_mode_prepare+0x1a/0x120 [ 285.397086] ? syscall_exit_to_user_mode+0x22/0x50 [ 285.397734] ? do_syscall_64+0x71/0x90 [ 285.398258] __x64_sys_sendto+0x74/0x90 [ 285.398786] do_syscall_64+0x64/0x90 [ 285.399273] ? exit_to_user_mode_prepare+0x1a/0x120 [ 285.399949] ? syscall_exit_to_user_mode+0x22/0x50 [ 285.400605] ? do_syscall_64+0x71/0x90 [ 285.401124] entry_SYSCALL_64_after_hwframe+0x6e/0xd8 [ 285.401807] RIP: 0033:0x495726 [ 285.402233] Code: ff ff ff f7 d8 64 89 02 48 c7 c0 ff ff ff ff eb b8 0f 1f 00 41 89 ca 64 8b 04 25 18 00 00 00 85 c0 75 11 b8 2c 00 00 00 0f 09 [ 285.404683] RSP: 002b:00007ffcc25fb618 EFLAGS: 00000246 ORIG_RAX: 000000000000002c [ 285.405677] RAX: ffffffffffffffda RBX: 0000000000000040 RCX: 0000000000495726 [ 285.406628] RDX: 0000000000000040 RSI: 0000000002518750 RDI: 0000000000000000 [ 285.407565] RBP: 00000000005205ef R08: 00000000005f8838 R09: 000000000000001c [ 285.408523] R10: 0000000000000000 R11: 0000000000000246 R12: 0000000002517634 [ 285.409460] R13: 00007ffcc25fb6f0 R14: 0000000000000003 R15: 0000000000000000 [ 285.410403] </TASK> [ 285.410704] [ 285.410929] Allocated by task 144: [ 285.411402] kasan_save_stack+0x1e/0x40 [ 285.411926] kasan_set_track+0x21/0x30 [ 285.412442] __kasan_slab_alloc+0x55/0x70 [ 285.412973] kmem_cache_alloc_node+0x187/0x3d0 [ 285.413567] __alloc_skb+0x1b4/0x230 [ 285.414060] __ip_append_data+0x17f7/0x1b60 [ 285.414633] ip_append_data+0x97/0xf0 [ 285.415144] raw_sendmsg+0x5a8/0x12d0 [ 285.415640] sock_sendmsg+0xcc/0xe0 [ 285.416117] __sys_sendto+0x190/0x230 [ 285.416626] __x64_sys_sendto+0x74/0x90 [ 285.417145] do_syscall_64+0x64/0x90 [ 285.417624] entry_SYSCALL_64_after_hwframe+0x6e/0xd8 [ 285.418306] [ 285.418531] Freed by task 144: [ 285.418960] kasan_save_stack+0x1e/0x40 [ 285.419469] kasan_set_track+0x21/0x30 [ 285.419988] kasan_save_free_info+0x27/0x40 [ 285.420556] ____kasan_slab_free+0x109/0x1a0 [ 285.421146] kmem_cache_free+0x1c2/0x450 [ 285.421680] __netif_receive_skb_core+0x2ce/0x1870 [ 285.422333] __netif_receive_skb_one_core+0x97/0x140 [ 285.423003] process_backlog+0x100/0x2f0 [ 285.423537] __napi_poll+0x5c/0x2d0 [ 285.424023] net_rx_action+0x2be/0x560 [ 285.424510] __do_softirq+0x11b/0x3de [ 285.425034] [ 285.425254] The buggy address belongs to the object at ffff8880bad31280 [ 285.425254] which belongs to the cache skbuff_head_cache of size 224 [ 285.426993] The buggy address is located 40 bytes inside of [ 285.426993] freed 224-byte region [ffff8880bad31280, ffff8880bad31360) [ 285.428572] [ 285.428798] The buggy address belongs to the physical page: [ 285.429540] page:00000000f4b77674 refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0xbad31 [ 285.430758] flags: 0x100000000000200(slab|node=0|zone=1) [ 285.431447] page_type: 0xffffffff() [ 285.431934] raw: 0100000000000200 ffff88810094a8c0 dead000000000122 0000000000000000 [ 285.432757] raw: 0000000000000000 00000000800c000c 00000001ffffffff 0000000000000000 [ 285.433562] page dumped because: kasan: bad access detected [ 285.434144] [ 285.434320] Memory state around the buggy address: [ 285.434828] ffff8880bad31180: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc [ 285.435580] ffff8880bad31200: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc [ 285.436264] >ffff8880bad31280: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb [ 285.436777] ^ [ 285.437106] ffff8880bad31300: fb fb fb fb fb fb fb fb fb fb fb fb fc fc fc fc [ 285.437616] ffff8880bad31380: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc [ 285.438126] ================================================================== [ 285.438662] Disabling lock debugging due to kernel taint Fix this by: 1. Changing sch_plug's .peek handler to qdisc_peek_dequeued(), a function compatible with non-work-conserving qdiscs 2. Checking the return value of qdisc_dequeue_peeked() in sch_qfq. Fixes: 462dbc9101ac ("pkt_sched: QFQ Plus: fair-queueing service at DRR cost") Reported-by: valis <sec@valis.email> Signed-off-by: valis <sec@valis.email> Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com> Link: https://lore.kernel.org/r/20230901162237.11525-1-jhs@mojatatu.com Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2023-09-01 16:22:37 +00:00
return skb;
}
static inline struct sk_buff *qfq_peek_skb(struct qfq_aggregate *agg,
struct qfq_class **cl,
unsigned int *len)
{
struct sk_buff *skb;
*cl = list_first_entry(&agg->active, struct qfq_class, alist);
skb = (*cl)->qdisc->ops->peek((*cl)->qdisc);
if (skb == NULL)
qdisc_warn_nonwc("qfq_dequeue", (*cl)->qdisc);
else
*len = qdisc_pkt_len(skb);
return skb;
}
/* Update F according to the actual service received by the aggregate. */
static inline void charge_actual_service(struct qfq_aggregate *agg)
{
pkt_sched: sch_qfq: properly cap timestamps in charge_actual_service QFQ+ schedules the active aggregates in a group using a bucket list (one list per group). The bucket in which each aggregate is inserted depends on the aggregate's timestamps, and the number of buckets in a group is enough to accomodate the possible (range of) values of the timestamps of all the aggregates in the group. For this property to hold, timestamps must however be computed correctly. One necessary condition for computing timestamps correctly is that the number of bits dequeued for each aggregate, while the aggregate is in service, does not exceed the maximum budget budgetmax assigned to the aggregate. For each aggregate, budgetmax is proportional to the number of classes in the aggregate. If the number of classes of the aggregate is decreased through qfq_change_class(), then budgetmax is decreased automatically as well. Problems may occur if the aggregate is in service when budgetmax is decreased, because the current remaining budget of the aggregate and/or the service already received by the aggregate may happen to be larger than the new value of budgetmax. In this case, when the aggregate is eventually deselected and its timestamps are updated, the aggregate may happen to have received an amount of service larger than budgetmax. This may cause the aggregate to be assigned a higher virtual finish time than the maximum acceptable value for the last bucket in the bucket list of the group. This fix introduces a cap that addresses this issue. Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Reviewed-by: Fabio Checconi <fchecconi@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-03-05 08:04:57 +00:00
/* Compute the service received by the aggregate, taking into
* account that, after decreasing the number of classes in
* agg, it may happen that
* agg->initial_budget - agg->budget > agg->bugdetmax
*/
u32 service_received = min(agg->budgetmax,
agg->initial_budget - agg->budget);
agg->F = agg->S + (u64)service_received * agg->inv_w;
}
/* Assign a reasonable start time for a new aggregate in group i.
* Admissible values for \hat(F) are multiples of \sigma_i
* no greater than V+\sigma_i . Larger values mean that
* we had a wraparound so we consider the timestamp to be stale.
*
* If F is not stale and F >= V then we set S = F.
* Otherwise we should assign S = V, but this may violate
* the ordering in EB (see [2]). So, if we have groups in ER,
* set S to the F_j of the first group j which would be blocking us.
* We are guaranteed not to move S backward because
* otherwise our group i would still be blocked.
*/
static void qfq_update_start(struct qfq_sched *q, struct qfq_aggregate *agg)
{
unsigned long mask;
u64 limit, roundedF;
int slot_shift = agg->grp->slot_shift;
roundedF = qfq_round_down(agg->F, slot_shift);
limit = qfq_round_down(q->V, slot_shift) + (1ULL << slot_shift);
if (!qfq_gt(agg->F, q->V) || qfq_gt(roundedF, limit)) {
/* timestamp was stale */
mask = mask_from(q->bitmaps[ER], agg->grp->index);
if (mask) {
struct qfq_group *next = qfq_ffs(q, mask);
if (qfq_gt(roundedF, next->F)) {
if (qfq_gt(limit, next->F))
agg->S = next->F;
else /* preserve timestamp correctness */
agg->S = limit;
return;
}
}
agg->S = q->V;
} else /* timestamp is not stale */
agg->S = agg->F;
}
/* Update the timestamps of agg before scheduling/rescheduling it for
* service. In particular, assign to agg->F its maximum possible
* value, i.e., the virtual finish time with which the aggregate
* should be labeled if it used all its budget once in service.
*/
static inline void
qfq_update_agg_ts(struct qfq_sched *q,
struct qfq_aggregate *agg, enum update_reason reason)
{
if (reason != requeue)
qfq_update_start(q, agg);
else /* just charge agg for the service received */
agg->S = agg->F;
agg->F = agg->S + (u64)agg->budgetmax * agg->inv_w;
}
pkt_sched: sch_qfq: serve activated aggregates immediately if the scheduler is empty If no aggregate is in service, then the function qfq_dequeue() does not dequeue any packet. For this reason, to guarantee QFQ+ to be work conserving, a just-activated aggregate must be set as in service immediately if it happens to be the only active aggregate. This is done by the function qfq_enqueue(). Unfortunately, the function qfq_add_to_agg(), used to add a class to an aggregate, does not perform this important additional operation. In particular, if: 1) qfq_add_to_agg() is invoked to complete the move of a class from a source aggregate, becoming, for this move, inactive, to a destination aggregate, becoming instead active, and 2) the destination aggregate becomes the only active aggregate, then this aggregate is not however set as in service. QFQ+ remains then in a non-work-conserving state until a new invocation of qfq_enqueue() recovers the situation. This fix solves the problem by moving the logic for setting an aggregate as in service directly into the function qfq_activate_agg(). Hence, from whatever point qfq_activate_aggregate() is invoked, QFQ+ remains work conserving. Since the more-complex logic of this new version of activate_aggregate() is not necessary, in qfq_dequeue(), to reschedule an aggregate that finishes its budget, then the aggregate is now rescheduled by invoking directly the functions needed. Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Reviewed-by: Fabio Checconi <fchecconi@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-03-05 08:04:59 +00:00
static void qfq_schedule_agg(struct qfq_sched *q, struct qfq_aggregate *agg);
static struct sk_buff *qfq_dequeue(struct Qdisc *sch)
{
struct qfq_sched *q = qdisc_priv(sch);
struct qfq_aggregate *in_serv_agg = q->in_serv_agg;
struct qfq_class *cl;
struct sk_buff *skb = NULL;
/* next-packet len, 0 means no more active classes in in-service agg */
unsigned int len = 0;
if (in_serv_agg == NULL)
return NULL;
if (!list_empty(&in_serv_agg->active))
skb = qfq_peek_skb(in_serv_agg, &cl, &len);
/*
* If there are no active classes in the in-service aggregate,
* or if the aggregate has not enough budget to serve its next
* class, then choose the next aggregate to serve.
*/
if (len == 0 || in_serv_agg->budget < len) {
charge_actual_service(in_serv_agg);
/* recharge the budget of the aggregate */
in_serv_agg->initial_budget = in_serv_agg->budget =
in_serv_agg->budgetmax;
pkt_sched: sch_qfq: serve activated aggregates immediately if the scheduler is empty If no aggregate is in service, then the function qfq_dequeue() does not dequeue any packet. For this reason, to guarantee QFQ+ to be work conserving, a just-activated aggregate must be set as in service immediately if it happens to be the only active aggregate. This is done by the function qfq_enqueue(). Unfortunately, the function qfq_add_to_agg(), used to add a class to an aggregate, does not perform this important additional operation. In particular, if: 1) qfq_add_to_agg() is invoked to complete the move of a class from a source aggregate, becoming, for this move, inactive, to a destination aggregate, becoming instead active, and 2) the destination aggregate becomes the only active aggregate, then this aggregate is not however set as in service. QFQ+ remains then in a non-work-conserving state until a new invocation of qfq_enqueue() recovers the situation. This fix solves the problem by moving the logic for setting an aggregate as in service directly into the function qfq_activate_agg(). Hence, from whatever point qfq_activate_aggregate() is invoked, QFQ+ remains work conserving. Since the more-complex logic of this new version of activate_aggregate() is not necessary, in qfq_dequeue(), to reschedule an aggregate that finishes its budget, then the aggregate is now rescheduled by invoking directly the functions needed. Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Reviewed-by: Fabio Checconi <fchecconi@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-03-05 08:04:59 +00:00
if (!list_empty(&in_serv_agg->active)) {
/*
* Still active: reschedule for
* service. Possible optimization: if no other
* aggregate is active, then there is no point
* in rescheduling this aggregate, and we can
* just keep it as the in-service one. This
* should be however a corner case, and to
* handle it, we would need to maintain an
* extra num_active_aggs field.
*/
pkt_sched: sch_qfq: serve activated aggregates immediately if the scheduler is empty If no aggregate is in service, then the function qfq_dequeue() does not dequeue any packet. For this reason, to guarantee QFQ+ to be work conserving, a just-activated aggregate must be set as in service immediately if it happens to be the only active aggregate. This is done by the function qfq_enqueue(). Unfortunately, the function qfq_add_to_agg(), used to add a class to an aggregate, does not perform this important additional operation. In particular, if: 1) qfq_add_to_agg() is invoked to complete the move of a class from a source aggregate, becoming, for this move, inactive, to a destination aggregate, becoming instead active, and 2) the destination aggregate becomes the only active aggregate, then this aggregate is not however set as in service. QFQ+ remains then in a non-work-conserving state until a new invocation of qfq_enqueue() recovers the situation. This fix solves the problem by moving the logic for setting an aggregate as in service directly into the function qfq_activate_agg(). Hence, from whatever point qfq_activate_aggregate() is invoked, QFQ+ remains work conserving. Since the more-complex logic of this new version of activate_aggregate() is not necessary, in qfq_dequeue(), to reschedule an aggregate that finishes its budget, then the aggregate is now rescheduled by invoking directly the functions needed. Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Reviewed-by: Fabio Checconi <fchecconi@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-03-05 08:04:59 +00:00
qfq_update_agg_ts(q, in_serv_agg, requeue);
qfq_schedule_agg(q, in_serv_agg);
} else if (sch->q.qlen == 0) { /* no aggregate to serve */
q->in_serv_agg = NULL;
return NULL;
}
/*
* If we get here, there are other aggregates queued:
* choose the new aggregate to serve.
*/
in_serv_agg = q->in_serv_agg = qfq_choose_next_agg(q);
skb = qfq_peek_skb(in_serv_agg, &cl, &len);
}
if (!skb)
return NULL;
sch->q.qlen--;
net: sched: sch_qfq: Fix UAF in qfq_dequeue() When the plug qdisc is used as a class of the qfq qdisc it could trigger a UAF. This issue can be reproduced with following commands: tc qdisc add dev lo root handle 1: qfq tc class add dev lo parent 1: classid 1:1 qfq weight 1 maxpkt 512 tc qdisc add dev lo parent 1:1 handle 2: plug tc filter add dev lo parent 1: basic classid 1:1 ping -c1 127.0.0.1 and boom: [ 285.353793] BUG: KASAN: slab-use-after-free in qfq_dequeue+0xa7/0x7f0 [ 285.354910] Read of size 4 at addr ffff8880bad312a8 by task ping/144 [ 285.355903] [ 285.356165] CPU: 1 PID: 144 Comm: ping Not tainted 6.5.0-rc3+ #4 [ 285.357112] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.14.0-2 04/01/2014 [ 285.358376] Call Trace: [ 285.358773] <IRQ> [ 285.359109] dump_stack_lvl+0x44/0x60 [ 285.359708] print_address_description.constprop.0+0x2c/0x3c0 [ 285.360611] kasan_report+0x10c/0x120 [ 285.361195] ? qfq_dequeue+0xa7/0x7f0 [ 285.361780] qfq_dequeue+0xa7/0x7f0 [ 285.362342] __qdisc_run+0xf1/0x970 [ 285.362903] net_tx_action+0x28e/0x460 [ 285.363502] __do_softirq+0x11b/0x3de [ 285.364097] do_softirq.part.0+0x72/0x90 [ 285.364721] </IRQ> [ 285.365072] <TASK> [ 285.365422] __local_bh_enable_ip+0x77/0x90 [ 285.366079] __dev_queue_xmit+0x95f/0x1550 [ 285.366732] ? __pfx_csum_and_copy_from_iter+0x10/0x10 [ 285.367526] ? __pfx___dev_queue_xmit+0x10/0x10 [ 285.368259] ? __build_skb_around+0x129/0x190 [ 285.368960] ? ip_generic_getfrag+0x12c/0x170 [ 285.369653] ? __pfx_ip_generic_getfrag+0x10/0x10 [ 285.370390] ? csum_partial+0x8/0x20 [ 285.370961] ? raw_getfrag+0xe5/0x140 [ 285.371559] ip_finish_output2+0x539/0xa40 [ 285.372222] ? __pfx_ip_finish_output2+0x10/0x10 [ 285.372954] ip_output+0x113/0x1e0 [ 285.373512] ? __pfx_ip_output+0x10/0x10 [ 285.374130] ? icmp_out_count+0x49/0x60 [ 285.374739] ? __pfx_ip_finish_output+0x10/0x10 [ 285.375457] ip_push_pending_frames+0xf3/0x100 [ 285.376173] raw_sendmsg+0xef5/0x12d0 [ 285.376760] ? do_syscall_64+0x40/0x90 [ 285.377359] ? __static_call_text_end+0x136578/0x136578 [ 285.378173] ? do_syscall_64+0x40/0x90 [ 285.378772] ? kasan_enable_current+0x11/0x20 [ 285.379469] ? __pfx_raw_sendmsg+0x10/0x10 [ 285.380137] ? __sock_create+0x13e/0x270 [ 285.380673] ? __sys_socket+0xf3/0x180 [ 285.381174] ? __x64_sys_socket+0x3d/0x50 [ 285.381725] ? entry_SYSCALL_64_after_hwframe+0x6e/0xd8 [ 285.382425] ? __rcu_read_unlock+0x48/0x70 [ 285.382975] ? ip4_datagram_release_cb+0xd8/0x380 [ 285.383608] ? __pfx_ip4_datagram_release_cb+0x10/0x10 [ 285.384295] ? preempt_count_sub+0x14/0xc0 [ 285.384844] ? __list_del_entry_valid+0x76/0x140 [ 285.385467] ? _raw_spin_lock_bh+0x87/0xe0 [ 285.386014] ? __pfx__raw_spin_lock_bh+0x10/0x10 [ 285.386645] ? release_sock+0xa0/0xd0 [ 285.387148] ? preempt_count_sub+0x14/0xc0 [ 285.387712] ? freeze_secondary_cpus+0x348/0x3c0 [ 285.388341] ? aa_sk_perm+0x177/0x390 [ 285.388856] ? __pfx_aa_sk_perm+0x10/0x10 [ 285.389441] ? check_stack_object+0x22/0x70 [ 285.390032] ? inet_send_prepare+0x2f/0x120 [ 285.390603] ? __pfx_inet_sendmsg+0x10/0x10 [ 285.391172] sock_sendmsg+0xcc/0xe0 [ 285.391667] __sys_sendto+0x190/0x230 [ 285.392168] ? __pfx___sys_sendto+0x10/0x10 [ 285.392727] ? kvm_clock_get_cycles+0x14/0x30 [ 285.393328] ? set_normalized_timespec64+0x57/0x70 [ 285.393980] ? _raw_spin_unlock_irq+0x1b/0x40 [ 285.394578] ? __x64_sys_clock_gettime+0x11c/0x160 [ 285.395225] ? __pfx___x64_sys_clock_gettime+0x10/0x10 [ 285.395908] ? _copy_to_user+0x3e/0x60 [ 285.396432] ? exit_to_user_mode_prepare+0x1a/0x120 [ 285.397086] ? syscall_exit_to_user_mode+0x22/0x50 [ 285.397734] ? do_syscall_64+0x71/0x90 [ 285.398258] __x64_sys_sendto+0x74/0x90 [ 285.398786] do_syscall_64+0x64/0x90 [ 285.399273] ? exit_to_user_mode_prepare+0x1a/0x120 [ 285.399949] ? syscall_exit_to_user_mode+0x22/0x50 [ 285.400605] ? do_syscall_64+0x71/0x90 [ 285.401124] entry_SYSCALL_64_after_hwframe+0x6e/0xd8 [ 285.401807] RIP: 0033:0x495726 [ 285.402233] Code: ff ff ff f7 d8 64 89 02 48 c7 c0 ff ff ff ff eb b8 0f 1f 00 41 89 ca 64 8b 04 25 18 00 00 00 85 c0 75 11 b8 2c 00 00 00 0f 09 [ 285.404683] RSP: 002b:00007ffcc25fb618 EFLAGS: 00000246 ORIG_RAX: 000000000000002c [ 285.405677] RAX: ffffffffffffffda RBX: 0000000000000040 RCX: 0000000000495726 [ 285.406628] RDX: 0000000000000040 RSI: 0000000002518750 RDI: 0000000000000000 [ 285.407565] RBP: 00000000005205ef R08: 00000000005f8838 R09: 000000000000001c [ 285.408523] R10: 0000000000000000 R11: 0000000000000246 R12: 0000000002517634 [ 285.409460] R13: 00007ffcc25fb6f0 R14: 0000000000000003 R15: 0000000000000000 [ 285.410403] </TASK> [ 285.410704] [ 285.410929] Allocated by task 144: [ 285.411402] kasan_save_stack+0x1e/0x40 [ 285.411926] kasan_set_track+0x21/0x30 [ 285.412442] __kasan_slab_alloc+0x55/0x70 [ 285.412973] kmem_cache_alloc_node+0x187/0x3d0 [ 285.413567] __alloc_skb+0x1b4/0x230 [ 285.414060] __ip_append_data+0x17f7/0x1b60 [ 285.414633] ip_append_data+0x97/0xf0 [ 285.415144] raw_sendmsg+0x5a8/0x12d0 [ 285.415640] sock_sendmsg+0xcc/0xe0 [ 285.416117] __sys_sendto+0x190/0x230 [ 285.416626] __x64_sys_sendto+0x74/0x90 [ 285.417145] do_syscall_64+0x64/0x90 [ 285.417624] entry_SYSCALL_64_after_hwframe+0x6e/0xd8 [ 285.418306] [ 285.418531] Freed by task 144: [ 285.418960] kasan_save_stack+0x1e/0x40 [ 285.419469] kasan_set_track+0x21/0x30 [ 285.419988] kasan_save_free_info+0x27/0x40 [ 285.420556] ____kasan_slab_free+0x109/0x1a0 [ 285.421146] kmem_cache_free+0x1c2/0x450 [ 285.421680] __netif_receive_skb_core+0x2ce/0x1870 [ 285.422333] __netif_receive_skb_one_core+0x97/0x140 [ 285.423003] process_backlog+0x100/0x2f0 [ 285.423537] __napi_poll+0x5c/0x2d0 [ 285.424023] net_rx_action+0x2be/0x560 [ 285.424510] __do_softirq+0x11b/0x3de [ 285.425034] [ 285.425254] The buggy address belongs to the object at ffff8880bad31280 [ 285.425254] which belongs to the cache skbuff_head_cache of size 224 [ 285.426993] The buggy address is located 40 bytes inside of [ 285.426993] freed 224-byte region [ffff8880bad31280, ffff8880bad31360) [ 285.428572] [ 285.428798] The buggy address belongs to the physical page: [ 285.429540] page:00000000f4b77674 refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0xbad31 [ 285.430758] flags: 0x100000000000200(slab|node=0|zone=1) [ 285.431447] page_type: 0xffffffff() [ 285.431934] raw: 0100000000000200 ffff88810094a8c0 dead000000000122 0000000000000000 [ 285.432757] raw: 0000000000000000 00000000800c000c 00000001ffffffff 0000000000000000 [ 285.433562] page dumped because: kasan: bad access detected [ 285.434144] [ 285.434320] Memory state around the buggy address: [ 285.434828] ffff8880bad31180: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc [ 285.435580] ffff8880bad31200: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc [ 285.436264] >ffff8880bad31280: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb [ 285.436777] ^ [ 285.437106] ffff8880bad31300: fb fb fb fb fb fb fb fb fb fb fb fb fc fc fc fc [ 285.437616] ffff8880bad31380: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc [ 285.438126] ================================================================== [ 285.438662] Disabling lock debugging due to kernel taint Fix this by: 1. Changing sch_plug's .peek handler to qdisc_peek_dequeued(), a function compatible with non-work-conserving qdiscs 2. Checking the return value of qdisc_dequeue_peeked() in sch_qfq. Fixes: 462dbc9101ac ("pkt_sched: QFQ Plus: fair-queueing service at DRR cost") Reported-by: valis <sec@valis.email> Signed-off-by: valis <sec@valis.email> Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com> Link: https://lore.kernel.org/r/20230901162237.11525-1-jhs@mojatatu.com Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2023-09-01 16:22:37 +00:00
skb = agg_dequeue(in_serv_agg, cl, len);
if (!skb) {
sch->q.qlen++;
return NULL;
}
qdisc_qstats_backlog_dec(sch, skb);
qdisc_bstats_update(sch, skb);
pkt_sched: sch_qfq: prevent budget from wrapping around after a dequeue Aggregate budgets are computed so as to guarantee that, after an aggregate has been selected for service, that aggregate has enough budget to serve at least one maximum-size packet for the classes it contains. For this reason, after a new aggregate has been selected for service, its next packet is immediately dequeued, without any further control. The maximum packet size for a class, lmax, can be changed through qfq_change_class(). In case the user sets lmax to a lower value than the the size of some of the still-to-arrive packets, QFQ+ will automatically push up lmax as it enqueues these packets. This automatic push up is likely to happen with TSO/GSO. In any case, if lmax is assigned a lower value than the size of some of the packets already enqueued for the class, then the following problem may occur: the size of the next packet to dequeue for the class may happen to be larger than lmax, after the aggregate to which the class belongs has been just selected for service. In this case, even the budget of the aggregate, which is an unsigned value, may be lower than the size of the next packet to dequeue. After dequeueing this packet and subtracting its size from the budget, the latter would wrap around. This fix prevents the budget from wrapping around after any packet dequeue. Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Reviewed-by: Fabio Checconi <fchecconi@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-03-05 08:05:00 +00:00
/* If lmax is lowered, through qfq_change_class, for a class
* owning pending packets with larger size than the new value
* of lmax, then the following condition may hold.
*/
if (unlikely(in_serv_agg->budget < len))
in_serv_agg->budget = 0;
else
in_serv_agg->budget -= len;
pkt_sched: sch_qfq: remove a source of high packet delay/jitter QFQ+ inherits from QFQ a design choice that may cause a high packet delay/jitter and a severe short-term unfairness. As QFQ, QFQ+ uses a special quantity, the system virtual time, to track the service provided by the ideal system it approximates. When a packet is dequeued, this quantity must be incremented by the size of the packet, divided by the sum of the weights of the aggregates waiting to be served. Tracking this sum correctly is a non-trivial task, because, to preserve tight service guarantees, the decrement of this sum must be delayed in a special way [1]: this sum can be decremented only after that its value would decrease also in the ideal system approximated by QFQ+. For efficiency, QFQ+ keeps track only of the 'instantaneous' weight sum, increased and decreased immediately as the weight of an aggregate changes, and as an aggregate is created or destroyed (which, in its turn, happens as a consequence of some class being created/destroyed/changed). However, to avoid the problems caused to service guarantees by these immediate decreases, QFQ+ increments the system virtual time using the maximum value allowed for the weight sum, 2^10, in place of the dynamic, instantaneous value. The instantaneous value of the weight sum is used only to check whether a request of weight increase or a class creation can be satisfied. Unfortunately, the problems caused by this choice are worse than the temporary degradation of the service guarantees that may occur, when a class is changed or destroyed, if the instantaneous value of the weight sum was used to update the system virtual time. In fact, the fraction of the link bandwidth guaranteed by QFQ+ to each aggregate is equal to the ratio between the weight of the aggregate and the sum of the weights of the competing aggregates. The packet delay guaranteed to the aggregate is instead inversely proportional to the guaranteed bandwidth. By using the maximum possible value, and not the actual value of the weight sum, QFQ+ provides each aggregate with the worst possible service guarantees, and not with service guarantees related to the actual set of competing aggregates. To see the consequences of this fact, consider the following simple example. Suppose that only the following aggregates are backlogged, i.e., that only the classes in the following aggregates have packets to transmit: one aggregate with weight 10, say A, and ten aggregates with weight 1, say B1, B2, ..., B10. In particular, suppose that these aggregates are always backlogged. Given the weight distribution, the smoothest and fairest service order would be: A B1 A B2 A B3 A B4 A B5 A B6 A B7 A B8 A B9 A B10 A B1 A B2 ... QFQ+ would provide exactly this optimal service if it used the actual value for the weight sum instead of the maximum possible value, i.e., 11 instead of 2^10. In contrast, since QFQ+ uses the latter value, it serves aggregates as follows (easy to prove and to reproduce experimentally): A B1 B2 B3 B4 B5 B6 B7 B8 B9 B10 A A A A A A A A A A B1 B2 ... B10 A A ... By replacing 10 with N in the above example, and by increasing N, one can increase at will the maximum packet delay and the jitter experienced by the classes in aggregate A. This patch addresses this issue by just using the above 'instantaneous' value of the weight sum, instead of the maximum possible value, when updating the system virtual time. After the instantaneous weight sum is decreased, QFQ+ may deviate from the ideal service for a time interval in the order of the time to serve one maximum-size packet for each backlogged class. The worst-case extent of the deviation exhibited by QFQ+ during this time interval [1] is basically the same as of the deviation described above (but, without this patch, QFQ+ suffers from such a deviation all the time). Finally, this patch modifies the comment to the function qfq_slot_insert, to make it coherent with the fact that the weight sum used by QFQ+ can now be lower than the maximum possible value. [1] P. Valente, "Extending WF2Q+ to support a dynamic traffic mix", Proceedings of AAA-IDEA'05, June 2005. Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-07-16 06:52:30 +00:00
q->V += (u64)len * q->iwsum;
pr_debug("qfq dequeue: len %u F %lld now %lld\n",
len, (unsigned long long) in_serv_agg->F,
(unsigned long long) q->V);
return skb;
}
static struct qfq_aggregate *qfq_choose_next_agg(struct qfq_sched *q)
{
struct qfq_group *grp;
struct qfq_aggregate *agg, *new_front_agg;
u64 old_F;
qfq_update_eligible(q);
q->oldV = q->V;
if (!q->bitmaps[ER])
return NULL;
grp = qfq_ffs(q, q->bitmaps[ER]);
old_F = grp->F;
agg = qfq_slot_head(grp);
/* agg starts to be served, remove it from schedule */
qfq_front_slot_remove(grp);
new_front_agg = qfq_slot_scan(grp);
if (new_front_agg == NULL) /* group is now inactive, remove from ER */
__clear_bit(grp->index, &q->bitmaps[ER]);
else {
u64 roundedS = qfq_round_down(new_front_agg->S,
grp->slot_shift);
unsigned int s;
if (grp->S == roundedS)
return agg;
grp->S = roundedS;
grp->F = roundedS + (2ULL << grp->slot_shift);
__clear_bit(grp->index, &q->bitmaps[ER]);
s = qfq_calc_state(q, grp);
__set_bit(grp->index, &q->bitmaps[s]);
}
qfq_unblock_groups(q, grp->index, old_F);
return agg;
}
static int qfq_enqueue(struct sk_buff *skb, struct Qdisc *sch,
struct sk_buff **to_free)
{
unsigned int len = qdisc_pkt_len(skb), gso_segs;
struct qfq_sched *q = qdisc_priv(sch);
struct qfq_class *cl;
struct qfq_aggregate *agg;
int err = 0;
bool first;
cl = qfq_classify(skb, sch, &err);
if (cl == NULL) {
if (err & __NET_XMIT_BYPASS)
qdisc_qstats_drop(sch);
__qdisc_drop(skb, to_free);
return err;
}
pr_debug("qfq_enqueue: cl = %x\n", cl->common.classid);
if (unlikely(cl->agg->lmax < len)) {
pkt_sched: enable QFQ to support TSO/GSO If the max packet size for some class (configured through tc) is violated by the actual size of the packets of that class, then QFQ would not schedule classes correctly, and the data structures implementing the bucket lists may get corrupted. This problem occurs with TSO/GSO even if the max packet size is set to the MTU, and is, e.g., the cause of the failure reported in [1]. Two patches have been proposed to solve this problem in [2], one of them is a preliminary version of this patch. This patch addresses the above issues by: 1) setting QFQ parameters to proper values for supporting TSO/GSO (in particular, setting the maximum possible packet size to 64KB), 2) automatically increasing the max packet size for a class, lmax, when a packet with a larger size than the current value of lmax arrives. The drawback of the first point is that the maximum weight for a class is now limited to 4096, which is equal to 1/16 of the maximum weight sum. Finally, this patch also forcibly caps the timestamps of a class if they are too high to be stored in the bucket list. This capping, taken from QFQ+ [3], handles the unfrequent case described in the comment to the function slot_insert. [1] http://marc.info/?l=linux-netdev&m=134968777902077&w=2 [2] http://marc.info/?l=linux-netdev&m=135096573507936&w=2 [3] http://marc.info/?l=linux-netdev&m=134902691421670&w=2 Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Tested-by: Cong Wang <amwang@redhat.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-11-05 20:29:24 +00:00
pr_debug("qfq: increasing maxpkt from %u to %u for class %u",
cl->agg->lmax, len, cl->common.classid);
err = qfq_change_agg(sch, cl, cl->agg->class_weight, len);
if (err) {
cl->qstats.drops++;
return qdisc_drop(skb, sch, to_free);
}
pkt_sched: enable QFQ to support TSO/GSO If the max packet size for some class (configured through tc) is violated by the actual size of the packets of that class, then QFQ would not schedule classes correctly, and the data structures implementing the bucket lists may get corrupted. This problem occurs with TSO/GSO even if the max packet size is set to the MTU, and is, e.g., the cause of the failure reported in [1]. Two patches have been proposed to solve this problem in [2], one of them is a preliminary version of this patch. This patch addresses the above issues by: 1) setting QFQ parameters to proper values for supporting TSO/GSO (in particular, setting the maximum possible packet size to 64KB), 2) automatically increasing the max packet size for a class, lmax, when a packet with a larger size than the current value of lmax arrives. The drawback of the first point is that the maximum weight for a class is now limited to 4096, which is equal to 1/16 of the maximum weight sum. Finally, this patch also forcibly caps the timestamps of a class if they are too high to be stored in the bucket list. This capping, taken from QFQ+ [3], handles the unfrequent case described in the comment to the function slot_insert. [1] http://marc.info/?l=linux-netdev&m=134968777902077&w=2 [2] http://marc.info/?l=linux-netdev&m=135096573507936&w=2 [3] http://marc.info/?l=linux-netdev&m=134902691421670&w=2 Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Tested-by: Cong Wang <amwang@redhat.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Acked-by: Stephen Hemminger <shemminger@vyatta.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-11-05 20:29:24 +00:00
}
gso_segs = skb_is_gso(skb) ? skb_shinfo(skb)->gso_segs : 1;
first = !cl->qdisc->q.qlen;
err = qdisc_enqueue(skb, cl->qdisc, to_free);
if (unlikely(err != NET_XMIT_SUCCESS)) {
pr_debug("qfq_enqueue: enqueue failed %d\n", err);
if (net_xmit_drop_count(err)) {
cl->qstats.drops++;
qdisc_qstats_drop(sch);
}
return err;
}
_bstats_update(&cl->bstats, len, gso_segs);
sch->qstats.backlog += len;
++sch->q.qlen;
agg = cl->agg;
/* if the queue was not empty, then done here */
if (!first) {
if (unlikely(skb == cl->qdisc->ops->peek(cl->qdisc)) &&
list_first_entry(&agg->active, struct qfq_class, alist)
== cl && cl->deficit < len)
list_move_tail(&cl->alist, &agg->active);
return err;
}
/* schedule class for service within the aggregate */
cl->deficit = agg->lmax;
list_add_tail(&cl->alist, &agg->active);
pkt_sched: sch_qfq: serve activated aggregates immediately if the scheduler is empty If no aggregate is in service, then the function qfq_dequeue() does not dequeue any packet. For this reason, to guarantee QFQ+ to be work conserving, a just-activated aggregate must be set as in service immediately if it happens to be the only active aggregate. This is done by the function qfq_enqueue(). Unfortunately, the function qfq_add_to_agg(), used to add a class to an aggregate, does not perform this important additional operation. In particular, if: 1) qfq_add_to_agg() is invoked to complete the move of a class from a source aggregate, becoming, for this move, inactive, to a destination aggregate, becoming instead active, and 2) the destination aggregate becomes the only active aggregate, then this aggregate is not however set as in service. QFQ+ remains then in a non-work-conserving state until a new invocation of qfq_enqueue() recovers the situation. This fix solves the problem by moving the logic for setting an aggregate as in service directly into the function qfq_activate_agg(). Hence, from whatever point qfq_activate_aggregate() is invoked, QFQ+ remains work conserving. Since the more-complex logic of this new version of activate_aggregate() is not necessary, in qfq_dequeue(), to reschedule an aggregate that finishes its budget, then the aggregate is now rescheduled by invoking directly the functions needed. Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Reviewed-by: Fabio Checconi <fchecconi@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-03-05 08:04:59 +00:00
if (list_first_entry(&agg->active, struct qfq_class, alist) != cl ||
q->in_serv_agg == agg)
return err; /* non-empty or in service, nothing else to do */
pkt_sched: sch_qfq: serve activated aggregates immediately if the scheduler is empty If no aggregate is in service, then the function qfq_dequeue() does not dequeue any packet. For this reason, to guarantee QFQ+ to be work conserving, a just-activated aggregate must be set as in service immediately if it happens to be the only active aggregate. This is done by the function qfq_enqueue(). Unfortunately, the function qfq_add_to_agg(), used to add a class to an aggregate, does not perform this important additional operation. In particular, if: 1) qfq_add_to_agg() is invoked to complete the move of a class from a source aggregate, becoming, for this move, inactive, to a destination aggregate, becoming instead active, and 2) the destination aggregate becomes the only active aggregate, then this aggregate is not however set as in service. QFQ+ remains then in a non-work-conserving state until a new invocation of qfq_enqueue() recovers the situation. This fix solves the problem by moving the logic for setting an aggregate as in service directly into the function qfq_activate_agg(). Hence, from whatever point qfq_activate_aggregate() is invoked, QFQ+ remains work conserving. Since the more-complex logic of this new version of activate_aggregate() is not necessary, in qfq_dequeue(), to reschedule an aggregate that finishes its budget, then the aggregate is now rescheduled by invoking directly the functions needed. Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Reviewed-by: Fabio Checconi <fchecconi@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-03-05 08:04:59 +00:00
qfq_activate_agg(q, agg, enqueue);
return err;
}
/*
* Schedule aggregate according to its timestamps.
*/
static void qfq_schedule_agg(struct qfq_sched *q, struct qfq_aggregate *agg)
{
struct qfq_group *grp = agg->grp;
u64 roundedS;
int s;
roundedS = qfq_round_down(agg->S, grp->slot_shift);
/*
* Insert agg in the correct bucket.
* If agg->S >= grp->S we don't need to adjust the
* bucket list and simply go to the insertion phase.
* Otherwise grp->S is decreasing, we must make room
* in the bucket list, and also recompute the group state.
* Finally, if there were no flows in this group and nobody
* was in ER make sure to adjust V.
*/
if (grp->full_slots) {
if (!qfq_gt(grp->S, agg->S))
goto skip_update;
/* create a slot for this agg->S */
qfq_slot_rotate(grp, roundedS);
/* group was surely ineligible, remove */
__clear_bit(grp->index, &q->bitmaps[IR]);
__clear_bit(grp->index, &q->bitmaps[IB]);
} else if (!q->bitmaps[ER] && qfq_gt(roundedS, q->V) &&
q->in_serv_agg == NULL)
q->V = roundedS;
grp->S = roundedS;
grp->F = roundedS + (2ULL << grp->slot_shift);
s = qfq_calc_state(q, grp);
__set_bit(grp->index, &q->bitmaps[s]);
pr_debug("qfq enqueue: new state %d %#lx S %lld F %lld V %lld\n",
s, q->bitmaps[s],
(unsigned long long) agg->S,
(unsigned long long) agg->F,
(unsigned long long) q->V);
skip_update:
qfq_slot_insert(grp, agg, roundedS);
}
/* Update agg ts and schedule agg for service */
static void qfq_activate_agg(struct qfq_sched *q, struct qfq_aggregate *agg,
enum update_reason reason)
{
pkt_sched: sch_qfq: serve activated aggregates immediately if the scheduler is empty If no aggregate is in service, then the function qfq_dequeue() does not dequeue any packet. For this reason, to guarantee QFQ+ to be work conserving, a just-activated aggregate must be set as in service immediately if it happens to be the only active aggregate. This is done by the function qfq_enqueue(). Unfortunately, the function qfq_add_to_agg(), used to add a class to an aggregate, does not perform this important additional operation. In particular, if: 1) qfq_add_to_agg() is invoked to complete the move of a class from a source aggregate, becoming, for this move, inactive, to a destination aggregate, becoming instead active, and 2) the destination aggregate becomes the only active aggregate, then this aggregate is not however set as in service. QFQ+ remains then in a non-work-conserving state until a new invocation of qfq_enqueue() recovers the situation. This fix solves the problem by moving the logic for setting an aggregate as in service directly into the function qfq_activate_agg(). Hence, from whatever point qfq_activate_aggregate() is invoked, QFQ+ remains work conserving. Since the more-complex logic of this new version of activate_aggregate() is not necessary, in qfq_dequeue(), to reschedule an aggregate that finishes its budget, then the aggregate is now rescheduled by invoking directly the functions needed. Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Reviewed-by: Fabio Checconi <fchecconi@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-03-05 08:04:59 +00:00
agg->initial_budget = agg->budget = agg->budgetmax; /* recharge budg. */
qfq_update_agg_ts(q, agg, reason);
pkt_sched: sch_qfq: serve activated aggregates immediately if the scheduler is empty If no aggregate is in service, then the function qfq_dequeue() does not dequeue any packet. For this reason, to guarantee QFQ+ to be work conserving, a just-activated aggregate must be set as in service immediately if it happens to be the only active aggregate. This is done by the function qfq_enqueue(). Unfortunately, the function qfq_add_to_agg(), used to add a class to an aggregate, does not perform this important additional operation. In particular, if: 1) qfq_add_to_agg() is invoked to complete the move of a class from a source aggregate, becoming, for this move, inactive, to a destination aggregate, becoming instead active, and 2) the destination aggregate becomes the only active aggregate, then this aggregate is not however set as in service. QFQ+ remains then in a non-work-conserving state until a new invocation of qfq_enqueue() recovers the situation. This fix solves the problem by moving the logic for setting an aggregate as in service directly into the function qfq_activate_agg(). Hence, from whatever point qfq_activate_aggregate() is invoked, QFQ+ remains work conserving. Since the more-complex logic of this new version of activate_aggregate() is not necessary, in qfq_dequeue(), to reschedule an aggregate that finishes its budget, then the aggregate is now rescheduled by invoking directly the functions needed. Signed-off-by: Paolo Valente <paolo.valente@unimore.it> Reviewed-by: Fabio Checconi <fchecconi@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-03-05 08:04:59 +00:00
if (q->in_serv_agg == NULL) { /* no aggr. in service or scheduled */
q->in_serv_agg = agg; /* start serving this aggregate */
/* update V: to be in service, agg must be eligible */
q->oldV = q->V = agg->S;
} else if (agg != q->in_serv_agg)
qfq_schedule_agg(q, agg);
}
static void qfq_slot_remove(struct qfq_sched *q, struct qfq_group *grp,
struct qfq_aggregate *agg)
{
unsigned int i, offset;
u64 roundedS;
roundedS = qfq_round_down(agg->S, grp->slot_shift);
offset = (roundedS - grp->S) >> grp->slot_shift;
i = (grp->front + offset) % QFQ_MAX_SLOTS;
hlist_del(&agg->next);
if (hlist_empty(&grp->slots[i]))
__clear_bit(offset, &grp->full_slots);
}
/*
* Called to forcibly deschedule an aggregate. If the aggregate is
* not in the front bucket, or if the latter has other aggregates in
* the front bucket, we can simply remove the aggregate with no other
* side effects.
* Otherwise we must propagate the event up.
*/
static void qfq_deactivate_agg(struct qfq_sched *q, struct qfq_aggregate *agg)
{
struct qfq_group *grp = agg->grp;
unsigned long mask;
u64 roundedS;
int s;
if (agg == q->in_serv_agg) {
charge_actual_service(agg);
q->in_serv_agg = qfq_choose_next_agg(q);
return;
}
agg->F = agg->S;
qfq_slot_remove(q, grp, agg);
if (!grp->full_slots) {
__clear_bit(grp->index, &q->bitmaps[IR]);
__clear_bit(grp->index, &q->bitmaps[EB]);
__clear_bit(grp->index, &q->bitmaps[IB]);
if (test_bit(grp->index, &q->bitmaps[ER]) &&
!(q->bitmaps[ER] & ~((1UL << grp->index) - 1))) {
mask = q->bitmaps[ER] & ((1UL << grp->index) - 1);
if (mask)
mask = ~((1UL << __fls(mask)) - 1);
else
mask = ~0UL;
qfq_move_groups(q, mask, EB, ER);
qfq_move_groups(q, mask, IB, IR);
}
__clear_bit(grp->index, &q->bitmaps[ER]);
} else if (hlist_empty(&grp->slots[grp->front])) {
agg = qfq_slot_scan(grp);
roundedS = qfq_round_down(agg->S, grp->slot_shift);
if (grp->S != roundedS) {
__clear_bit(grp->index, &q->bitmaps[ER]);
__clear_bit(grp->index, &q->bitmaps[IR]);
__clear_bit(grp->index, &q->bitmaps[EB]);
__clear_bit(grp->index, &q->bitmaps[IB]);
grp->S = roundedS;
grp->F = roundedS + (2ULL << grp->slot_shift);
s = qfq_calc_state(q, grp);
__set_bit(grp->index, &q->bitmaps[s]);
}
}
}
static void qfq_qlen_notify(struct Qdisc *sch, unsigned long arg)
{
struct qfq_sched *q = qdisc_priv(sch);
struct qfq_class *cl = (struct qfq_class *)arg;
qfq_deactivate_class(q, cl);
}
static int qfq_init_qdisc(struct Qdisc *sch, struct nlattr *opt,
struct netlink_ext_ack *extack)
{
struct qfq_sched *q = qdisc_priv(sch);
struct qfq_group *grp;
int i, j, err;
u32 max_cl_shift, maxbudg_shift, max_classes;
err = tcf_block_get(&q->block, &q->filter_list, sch, extack);
if (err)
return err;
err = qdisc_class_hash_init(&q->clhash);
if (err < 0)
return err;
sch_qfq: prevent shift-out-of-bounds in qfq_init_qdisc tx_queue_len can be set to ~0U, we need to be more careful about overflows. __fls(0) is undefined, as this report shows: UBSAN: shift-out-of-bounds in net/sched/sch_qfq.c:1430:24 shift exponent 51770272 is too large for 32-bit type 'int' CPU: 0 PID: 25574 Comm: syz-executor.0 Not tainted 5.16.0-rc7-syzkaller #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 Call Trace: <TASK> __dump_stack lib/dump_stack.c:88 [inline] dump_stack_lvl+0x201/0x2d8 lib/dump_stack.c:106 ubsan_epilogue lib/ubsan.c:151 [inline] __ubsan_handle_shift_out_of_bounds+0x494/0x530 lib/ubsan.c:330 qfq_init_qdisc+0x43f/0x450 net/sched/sch_qfq.c:1430 qdisc_create+0x895/0x1430 net/sched/sch_api.c:1253 tc_modify_qdisc+0x9d9/0x1e20 net/sched/sch_api.c:1660 rtnetlink_rcv_msg+0x934/0xe60 net/core/rtnetlink.c:5571 netlink_rcv_skb+0x200/0x470 net/netlink/af_netlink.c:2496 netlink_unicast_kernel net/netlink/af_netlink.c:1319 [inline] netlink_unicast+0x814/0x9f0 net/netlink/af_netlink.c:1345 netlink_sendmsg+0xaea/0xe60 net/netlink/af_netlink.c:1921 sock_sendmsg_nosec net/socket.c:704 [inline] sock_sendmsg net/socket.c:724 [inline] ____sys_sendmsg+0x5b9/0x910 net/socket.c:2409 ___sys_sendmsg net/socket.c:2463 [inline] __sys_sendmsg+0x280/0x370 net/socket.c:2492 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x44/0xd0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x44/0xae Fixes: 462dbc9101ac ("pkt_sched: QFQ Plus: fair-queueing service at DRR cost") Signed-off-by: Eric Dumazet <edumazet@google.com> Reported-by: syzbot <syzkaller@googlegroups.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2022-01-04 09:45:08 +00:00
max_classes = min_t(u64, (u64)qdisc_dev(sch)->tx_queue_len + 1,
QFQ_MAX_AGG_CLASSES);
/* max_cl_shift = floor(log_2(max_classes)) */
max_cl_shift = __fls(max_classes);
q->max_agg_classes = 1<<max_cl_shift;
/* maxbudg_shift = log2(max_len * max_classes_per_agg) */
maxbudg_shift = QFQ_MTU_SHIFT + max_cl_shift;
q->min_slot_shift = FRAC_BITS + maxbudg_shift - QFQ_MAX_INDEX;
for (i = 0; i <= QFQ_MAX_INDEX; i++) {
grp = &q->groups[i];
grp->index = i;
grp->slot_shift = q->min_slot_shift + i;
for (j = 0; j < QFQ_MAX_SLOTS; j++)
INIT_HLIST_HEAD(&grp->slots[j]);
}
INIT_HLIST_HEAD(&q->nonfull_aggs);
return 0;
}
static void qfq_reset_qdisc(struct Qdisc *sch)
{
struct qfq_sched *q = qdisc_priv(sch);
struct qfq_class *cl;
unsigned int i;
for (i = 0; i < q->clhash.hashsize; i++) {
hlist: drop the node parameter from iterators I'm not sure why, but the hlist for each entry iterators were conceived list_for_each_entry(pos, head, member) The hlist ones were greedy and wanted an extra parameter: hlist_for_each_entry(tpos, pos, head, member) Why did they need an extra pos parameter? I'm not quite sure. Not only they don't really need it, it also prevents the iterator from looking exactly like the list iterator, which is unfortunate. Besides the semantic patch, there was some manual work required: - Fix up the actual hlist iterators in linux/list.h - Fix up the declaration of other iterators based on the hlist ones. - A very small amount of places were using the 'node' parameter, this was modified to use 'obj->member' instead. - Coccinelle didn't handle the hlist_for_each_entry_safe iterator properly, so those had to be fixed up manually. The semantic patch which is mostly the work of Peter Senna Tschudin is here: @@ iterator name hlist_for_each_entry, hlist_for_each_entry_continue, hlist_for_each_entry_from, hlist_for_each_entry_rcu, hlist_for_each_entry_rcu_bh, hlist_for_each_entry_continue_rcu_bh, for_each_busy_worker, ax25_uid_for_each, ax25_for_each, inet_bind_bucket_for_each, sctp_for_each_hentry, sk_for_each, sk_for_each_rcu, sk_for_each_from, sk_for_each_safe, sk_for_each_bound, hlist_for_each_entry_safe, hlist_for_each_entry_continue_rcu, nr_neigh_for_each, nr_neigh_for_each_safe, nr_node_for_each, nr_node_for_each_safe, for_each_gfn_indirect_valid_sp, for_each_gfn_sp, for_each_host; type T; expression a,c,d,e; identifier b; statement S; @@ -T b; <+... when != b ( hlist_for_each_entry(a, - b, c, d) S | hlist_for_each_entry_continue(a, - b, c) S | hlist_for_each_entry_from(a, - b, c) S | hlist_for_each_entry_rcu(a, - b, c, d) S | hlist_for_each_entry_rcu_bh(a, - b, c, d) S | hlist_for_each_entry_continue_rcu_bh(a, - b, c) S | for_each_busy_worker(a, c, - b, d) S | ax25_uid_for_each(a, - b, c) S | ax25_for_each(a, - b, c) S | inet_bind_bucket_for_each(a, - b, c) S | sctp_for_each_hentry(a, - b, c) S | sk_for_each(a, - b, c) S | sk_for_each_rcu(a, - b, c) S | sk_for_each_from -(a, b) +(a) S + sk_for_each_from(a) S | sk_for_each_safe(a, - b, c, d) S | sk_for_each_bound(a, - b, c) S | hlist_for_each_entry_safe(a, - b, c, d, e) S | hlist_for_each_entry_continue_rcu(a, - b, c) S | nr_neigh_for_each(a, - b, c) S | nr_neigh_for_each_safe(a, - b, c, d) S | nr_node_for_each(a, - b, c) S | nr_node_for_each_safe(a, - b, c, d) S | - for_each_gfn_sp(a, c, d, b) S + for_each_gfn_sp(a, c, d) S | - for_each_gfn_indirect_valid_sp(a, c, d, b) S + for_each_gfn_indirect_valid_sp(a, c, d) S | for_each_host(a, - b, c) S | for_each_host_safe(a, - b, c, d) S | for_each_mesh_entry(a, - b, c, d) S ) ...+> [akpm@linux-foundation.org: drop bogus change from net/ipv4/raw.c] [akpm@linux-foundation.org: drop bogus hunk from net/ipv6/raw.c] [akpm@linux-foundation.org: checkpatch fixes] [akpm@linux-foundation.org: fix warnings] [akpm@linux-foudnation.org: redo intrusive kvm changes] Tested-by: Peter Senna Tschudin <peter.senna@gmail.com> Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com> Cc: Wu Fengguang <fengguang.wu@intel.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Gleb Natapov <gleb@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-02-28 01:06:00 +00:00
hlist_for_each_entry(cl, &q->clhash.hash[i], common.hnode) {
if (cl->qdisc->q.qlen > 0)
qfq_deactivate_class(q, cl);
qdisc_reset(cl->qdisc);
}
}
}
static void qfq_destroy_qdisc(struct Qdisc *sch)
{
struct qfq_sched *q = qdisc_priv(sch);
struct qfq_class *cl;
hlist: drop the node parameter from iterators I'm not sure why, but the hlist for each entry iterators were conceived list_for_each_entry(pos, head, member) The hlist ones were greedy and wanted an extra parameter: hlist_for_each_entry(tpos, pos, head, member) Why did they need an extra pos parameter? I'm not quite sure. Not only they don't really need it, it also prevents the iterator from looking exactly like the list iterator, which is unfortunate. Besides the semantic patch, there was some manual work required: - Fix up the actual hlist iterators in linux/list.h - Fix up the declaration of other iterators based on the hlist ones. - A very small amount of places were using the 'node' parameter, this was modified to use 'obj->member' instead. - Coccinelle didn't handle the hlist_for_each_entry_safe iterator properly, so those had to be fixed up manually. The semantic patch which is mostly the work of Peter Senna Tschudin is here: @@ iterator name hlist_for_each_entry, hlist_for_each_entry_continue, hlist_for_each_entry_from, hlist_for_each_entry_rcu, hlist_for_each_entry_rcu_bh, hlist_for_each_entry_continue_rcu_bh, for_each_busy_worker, ax25_uid_for_each, ax25_for_each, inet_bind_bucket_for_each, sctp_for_each_hentry, sk_for_each, sk_for_each_rcu, sk_for_each_from, sk_for_each_safe, sk_for_each_bound, hlist_for_each_entry_safe, hlist_for_each_entry_continue_rcu, nr_neigh_for_each, nr_neigh_for_each_safe, nr_node_for_each, nr_node_for_each_safe, for_each_gfn_indirect_valid_sp, for_each_gfn_sp, for_each_host; type T; expression a,c,d,e; identifier b; statement S; @@ -T b; <+... when != b ( hlist_for_each_entry(a, - b, c, d) S | hlist_for_each_entry_continue(a, - b, c) S | hlist_for_each_entry_from(a, - b, c) S | hlist_for_each_entry_rcu(a, - b, c, d) S | hlist_for_each_entry_rcu_bh(a, - b, c, d) S | hlist_for_each_entry_continue_rcu_bh(a, - b, c) S | for_each_busy_worker(a, c, - b, d) S | ax25_uid_for_each(a, - b, c) S | ax25_for_each(a, - b, c) S | inet_bind_bucket_for_each(a, - b, c) S | sctp_for_each_hentry(a, - b, c) S | sk_for_each(a, - b, c) S | sk_for_each_rcu(a, - b, c) S | sk_for_each_from -(a, b) +(a) S + sk_for_each_from(a) S | sk_for_each_safe(a, - b, c, d) S | sk_for_each_bound(a, - b, c) S | hlist_for_each_entry_safe(a, - b, c, d, e) S | hlist_for_each_entry_continue_rcu(a, - b, c) S | nr_neigh_for_each(a, - b, c) S | nr_neigh_for_each_safe(a, - b, c, d) S | nr_node_for_each(a, - b, c) S | nr_node_for_each_safe(a, - b, c, d) S | - for_each_gfn_sp(a, c, d, b) S + for_each_gfn_sp(a, c, d) S | - for_each_gfn_indirect_valid_sp(a, c, d, b) S + for_each_gfn_indirect_valid_sp(a, c, d) S | for_each_host(a, - b, c) S | for_each_host_safe(a, - b, c, d) S | for_each_mesh_entry(a, - b, c, d) S ) ...+> [akpm@linux-foundation.org: drop bogus change from net/ipv4/raw.c] [akpm@linux-foundation.org: drop bogus hunk from net/ipv6/raw.c] [akpm@linux-foundation.org: checkpatch fixes] [akpm@linux-foundation.org: fix warnings] [akpm@linux-foudnation.org: redo intrusive kvm changes] Tested-by: Peter Senna Tschudin <peter.senna@gmail.com> Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com> Cc: Wu Fengguang <fengguang.wu@intel.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Gleb Natapov <gleb@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-02-28 01:06:00 +00:00
struct hlist_node *next;
unsigned int i;
tcf_block_put(q->block);
for (i = 0; i < q->clhash.hashsize; i++) {
hlist: drop the node parameter from iterators I'm not sure why, but the hlist for each entry iterators were conceived list_for_each_entry(pos, head, member) The hlist ones were greedy and wanted an extra parameter: hlist_for_each_entry(tpos, pos, head, member) Why did they need an extra pos parameter? I'm not quite sure. Not only they don't really need it, it also prevents the iterator from looking exactly like the list iterator, which is unfortunate. Besides the semantic patch, there was some manual work required: - Fix up the actual hlist iterators in linux/list.h - Fix up the declaration of other iterators based on the hlist ones. - A very small amount of places were using the 'node' parameter, this was modified to use 'obj->member' instead. - Coccinelle didn't handle the hlist_for_each_entry_safe iterator properly, so those had to be fixed up manually. The semantic patch which is mostly the work of Peter Senna Tschudin is here: @@ iterator name hlist_for_each_entry, hlist_for_each_entry_continue, hlist_for_each_entry_from, hlist_for_each_entry_rcu, hlist_for_each_entry_rcu_bh, hlist_for_each_entry_continue_rcu_bh, for_each_busy_worker, ax25_uid_for_each, ax25_for_each, inet_bind_bucket_for_each, sctp_for_each_hentry, sk_for_each, sk_for_each_rcu, sk_for_each_from, sk_for_each_safe, sk_for_each_bound, hlist_for_each_entry_safe, hlist_for_each_entry_continue_rcu, nr_neigh_for_each, nr_neigh_for_each_safe, nr_node_for_each, nr_node_for_each_safe, for_each_gfn_indirect_valid_sp, for_each_gfn_sp, for_each_host; type T; expression a,c,d,e; identifier b; statement S; @@ -T b; <+... when != b ( hlist_for_each_entry(a, - b, c, d) S | hlist_for_each_entry_continue(a, - b, c) S | hlist_for_each_entry_from(a, - b, c) S | hlist_for_each_entry_rcu(a, - b, c, d) S | hlist_for_each_entry_rcu_bh(a, - b, c, d) S | hlist_for_each_entry_continue_rcu_bh(a, - b, c) S | for_each_busy_worker(a, c, - b, d) S | ax25_uid_for_each(a, - b, c) S | ax25_for_each(a, - b, c) S | inet_bind_bucket_for_each(a, - b, c) S | sctp_for_each_hentry(a, - b, c) S | sk_for_each(a, - b, c) S | sk_for_each_rcu(a, - b, c) S | sk_for_each_from -(a, b) +(a) S + sk_for_each_from(a) S | sk_for_each_safe(a, - b, c, d) S | sk_for_each_bound(a, - b, c) S | hlist_for_each_entry_safe(a, - b, c, d, e) S | hlist_for_each_entry_continue_rcu(a, - b, c) S | nr_neigh_for_each(a, - b, c) S | nr_neigh_for_each_safe(a, - b, c, d) S | nr_node_for_each(a, - b, c) S | nr_node_for_each_safe(a, - b, c, d) S | - for_each_gfn_sp(a, c, d, b) S + for_each_gfn_sp(a, c, d) S | - for_each_gfn_indirect_valid_sp(a, c, d, b) S + for_each_gfn_indirect_valid_sp(a, c, d) S | for_each_host(a, - b, c) S | for_each_host_safe(a, - b, c, d) S | for_each_mesh_entry(a, - b, c, d) S ) ...+> [akpm@linux-foundation.org: drop bogus change from net/ipv4/raw.c] [akpm@linux-foundation.org: drop bogus hunk from net/ipv6/raw.c] [akpm@linux-foundation.org: checkpatch fixes] [akpm@linux-foundation.org: fix warnings] [akpm@linux-foudnation.org: redo intrusive kvm changes] Tested-by: Peter Senna Tschudin <peter.senna@gmail.com> Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com> Cc: Wu Fengguang <fengguang.wu@intel.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Gleb Natapov <gleb@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-02-28 01:06:00 +00:00
hlist_for_each_entry_safe(cl, next, &q->clhash.hash[i],
common.hnode) {
qfq_destroy_class(sch, cl);
}
}
qdisc_class_hash_destroy(&q->clhash);
}
static const struct Qdisc_class_ops qfq_class_ops = {
.change = qfq_change_class,
.delete = qfq_delete_class,
.find = qfq_search_class,
.tcf_block = qfq_tcf_block,
.bind_tcf = qfq_bind_tcf,
.unbind_tcf = qfq_unbind_tcf,
.graft = qfq_graft_class,
.leaf = qfq_class_leaf,
.qlen_notify = qfq_qlen_notify,
.dump = qfq_dump_class,
.dump_stats = qfq_dump_class_stats,
.walk = qfq_walk,
};
static struct Qdisc_ops qfq_qdisc_ops __read_mostly = {
.cl_ops = &qfq_class_ops,
.id = "qfq",
.priv_size = sizeof(struct qfq_sched),
.enqueue = qfq_enqueue,
.dequeue = qfq_dequeue,
.peek = qdisc_peek_dequeued,
.init = qfq_init_qdisc,
.reset = qfq_reset_qdisc,
.destroy = qfq_destroy_qdisc,
.owner = THIS_MODULE,
};
net/sched: Add module aliases for cls_,sch_,act_ modules No functional change intended, aliases will be used in followup commits. Note for backporters: you may need to add aliases also for modules that are already removed in mainline kernel but still in your version. Patches were generated with the help of Coccinelle scripts like: cat >scripts/coccinelle/misc/tcf_alias.cocci <<EOD virtual patch virtual report @ haskernel @ @@ @ tcf_has_kind depends on report && haskernel @ identifier ops; constant K; @@ static struct tcf_proto_ops ops = { .kind = K, ... }; +char module_alias = K; EOD /usr/bin/spatch -D report --cocci-file scripts/coccinelle/misc/tcf_alias.cocci \ --dir . \ -I ./arch/x86/include -I ./arch/x86/include/generated -I ./include \ -I ./arch/x86/include/uapi -I ./arch/x86/include/generated/uapi \ -I ./include/uapi -I ./include/generated/uapi \ --include ./include/linux/compiler-version.h --include ./include/linux/kconfig.h \ --jobs 8 --chunksize 1 2>/dev/null | \ sed 's/char module_alias = "\([^"]*\)";/MODULE_ALIAS_NET_CLS("\1");/' And analogously for: static struct tc_action_ops ops = { .kind = K, static struct Qdisc_ops ops = { .id = K, (Someone familiar would be able to fit those into one .cocci file without sed post processing.) Signed-off-by: Michal Koutný <mkoutny@suse.com> Acked-by: Jamal Hadi Salim <jhs@mojatatu.com> Reviewed-by: Jiri Pirko <jiri@nvidia.com> Link: https://lore.kernel.org/r/20240201130943.19536-3-mkoutny@suse.com Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2024-02-01 13:09:41 +00:00
MODULE_ALIAS_NET_SCH("qfq");
static int __init qfq_init(void)
{
return register_qdisc(&qfq_qdisc_ops);
}
static void __exit qfq_exit(void)
{
unregister_qdisc(&qfq_qdisc_ops);
}
module_init(qfq_init);
module_exit(qfq_exit);
MODULE_LICENSE("GPL");
MODULE_DESCRIPTION("Quick Fair Queueing Plus qdisc");