* [PATCH net-next v11 7/7] sch_cake: Conditionally split GSO segments
From: Toke Høiland-Jørgensen @ 2018-05-15 15:12 UTC (permalink / raw)
To: netdev; +Cc: cake
In-Reply-To: <152639705919.2525.3718182212727523960.stgit@alrua-kau>
At lower bandwidths, the transmission time of a single GSO segment can add
an unacceptable amount of latency due to HOL blocking. Furthermore, with a
software shaper, any tuning mechanism employed by the kernel to control the
maximum size of GSO segments is thrown off by the artificial limit on
bandwidth. For this reason, we split GSO segments into their individual
packets iff the shaper is active and configured to a bandwidth <= 1 Gbps.
Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
---
net/sched/sch_cake.c | 99 +++++++++++++++++++++++++++++++++++++-------------
1 file changed, 73 insertions(+), 26 deletions(-)
diff --git a/net/sched/sch_cake.c b/net/sched/sch_cake.c
index a4aad577bf8e..a0c2925b4c54 100644
--- a/net/sched/sch_cake.c
+++ b/net/sched/sch_cake.c
@@ -82,6 +82,7 @@
#define CAKE_QUEUES (1024)
#define CAKE_FLOW_MASK 63
#define CAKE_FLOW_NAT_FLAG 64
+#define CAKE_SPLIT_GSO_THRESHOLD (125000000) /* 1Gbps */
#define US2TIME(a) (a * (u64)NSEC_PER_USEC)
typedef u64 cobalt_time_t;
@@ -1477,36 +1478,73 @@ static s32 cake_enqueue(struct sk_buff *skb, struct Qdisc *sch,
if (unlikely(len > b->max_skblen))
b->max_skblen = len;
- cobalt_set_enqueue_time(skb, now);
- get_cobalt_cb(skb)->adjusted_len = cake_overhead(q, skb);
- flow_queue_add(flow, skb);
-
- if (q->ack_filter)
- ack = cake_ack_filter(q, flow);
+ if (skb_is_gso(skb) && q->rate_flags & CAKE_FLAG_SPLIT_GSO) {
+ struct sk_buff *segs, *nskb;
+ netdev_features_t features = netif_skb_features(skb);
+ unsigned int slen = 0;
+
+ segs = skb_gso_segment(skb, features & ~NETIF_F_GSO_MASK);
+ if (IS_ERR_OR_NULL(segs))
+ return qdisc_drop(skb, sch, to_free);
+
+ while (segs) {
+ nskb = segs->next;
+ segs->next = NULL;
+ qdisc_skb_cb(segs)->pkt_len = segs->len;
+ cobalt_set_enqueue_time(segs, now);
+ get_cobalt_cb(segs)->adjusted_len = cake_overhead(q,
+ segs);
+ flow_queue_add(flow, segs);
+
+ sch->q.qlen++;
+ slen += segs->len;
+ q->buffer_used += segs->truesize;
+ b->packets++;
+ segs = nskb;
+ }
- if (ack) {
- b->ack_drops++;
- sch->qstats.drops++;
- b->bytes += qdisc_pkt_len(ack);
- len -= qdisc_pkt_len(ack);
- q->buffer_used += skb->truesize - ack->truesize;
- if (q->rate_flags & CAKE_FLAG_INGRESS)
- cake_advance_shaper(q, b, ack, now, true);
+ /* stats */
+ b->bytes += slen;
+ b->backlogs[idx] += slen;
+ b->tin_backlog += slen;
+ sch->qstats.backlog += slen;
+ q->avg_window_bytes += slen;
- qdisc_tree_reduce_backlog(sch, 1, qdisc_pkt_len(ack));
- consume_skb(ack);
+ qdisc_tree_reduce_backlog(sch, 1, len);
+ consume_skb(skb);
} else {
- sch->q.qlen++;
- q->buffer_used += skb->truesize;
- }
+ /* not splitting */
+ cobalt_set_enqueue_time(skb, now);
+ get_cobalt_cb(skb)->adjusted_len = cake_overhead(q, skb);
+ flow_queue_add(flow, skb);
+
+ if (q->ack_filter)
+ ack = cake_ack_filter(q, flow);
+
+ if (ack) {
+ b->ack_drops++;
+ sch->qstats.drops++;
+ b->bytes += qdisc_pkt_len(ack);
+ len -= qdisc_pkt_len(ack);
+ q->buffer_used += skb->truesize - ack->truesize;
+ if (q->rate_flags & CAKE_FLAG_INGRESS)
+ cake_advance_shaper(q, b, ack, now, true);
+
+ qdisc_tree_reduce_backlog(sch, 1, qdisc_pkt_len(ack));
+ consume_skb(ack);
+ } else {
+ sch->q.qlen++;
+ q->buffer_used += skb->truesize;
+ }
- /* stats */
- b->packets++;
- b->bytes += len;
- b->backlogs[idx] += len;
- b->tin_backlog += len;
- sch->qstats.backlog += len;
- q->avg_window_bytes += len;
+ /* stats */
+ b->packets++;
+ b->bytes += len;
+ b->backlogs[idx] += len;
+ b->tin_backlog += len;
+ sch->qstats.backlog += len;
+ q->avg_window_bytes += len;
+ }
if (q->overflow_timeout)
cake_heapify_up(q, b->overflow_idx[idx]);
@@ -2314,6 +2352,11 @@ static int cake_change(struct Qdisc *sch, struct nlattr *opt,
if (tb[TCA_CAKE_MEMORY])
q->buffer_config_limit = nla_get_u32(tb[TCA_CAKE_MEMORY]);
+ if (q->rate_bps && q->rate_bps <= CAKE_SPLIT_GSO_THRESHOLD)
+ q->rate_flags |= CAKE_FLAG_SPLIT_GSO;
+ else
+ q->rate_flags &= ~CAKE_FLAG_SPLIT_GSO;
+
if (q->tins) {
sch_tree_lock(sch);
cake_reconfigure(sch);
@@ -2469,6 +2512,10 @@ static int cake_dump(struct Qdisc *sch, struct sk_buff *skb)
if (nla_put_u32(skb, TCA_CAKE_MPU, q->rate_mpu))
goto nla_put_failure;
+ if (nla_put_u32(skb, TCA_CAKE_SPLIT_GSO,
+ !!(q->rate_flags & CAKE_FLAG_SPLIT_GSO)))
+ goto nla_put_failure;
+
return nla_nest_end(skb, opts);
nla_put_failure:
^ permalink raw reply related
* [PATCH net-next v11 6/7] sch_cake: Add overhead compensation support to the rate shaper
From: Toke Høiland-Jørgensen @ 2018-05-15 15:12 UTC (permalink / raw)
To: netdev; +Cc: cake
In-Reply-To: <152639705919.2525.3718182212727523960.stgit@alrua-kau>
This commit adds configurable overhead compensation support to the rate
shaper. With this feature, userspace can configure the actual bottleneck
link overhead and encapsulation mode used, which will be used by the shaper
to calculate the precise duration of each packet on the wire.
This feature is needed because CAKE is often deployed one or two hops
upstream of the actual bottleneck (which can be, e.g., inside a DSL or
cable modem). In this case, the link layer characteristics and overhead
reported by the kernel does not match the actual bottleneck. Being able to
set the actual values in use makes it possible to configure the shaper rate
much closer to the actual bottleneck rate (our experience shows it is
possible to get with 0.1% of the actual physical bottleneck rate), thus
keeping latency low without sacrificing bandwidth.
The overhead compensation has three tunables: A fixed per-packet overhead
size (which, if set, will be accounted from the IP packet header), a
minimum packet size (MPU) and a framing mode supporting either ATM or PTM
framing. We include a set of common keywords in TC to help users configure
the right parameters. If no overhead value is set, the value reported by
the kernel is used.
Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
---
net/sched/sch_cake.c | 124 ++++++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 123 insertions(+), 1 deletion(-)
diff --git a/net/sched/sch_cake.c b/net/sched/sch_cake.c
index 4b13345f85dc..a4aad577bf8e 100644
--- a/net/sched/sch_cake.c
+++ b/net/sched/sch_cake.c
@@ -275,6 +275,7 @@ enum {
struct cobalt_skb_cb {
cobalt_time_t enqueue_time;
+ u32 adjusted_len;
};
static cobalt_time_t cobalt_get_time(void)
@@ -1129,6 +1130,88 @@ static cobalt_time_t cake_ewma(cobalt_time_t avg, cobalt_time_t sample,
return avg;
}
+static u32 cake_calc_overhead(struct cake_sched_data *q, u32 len, u32 off)
+{
+ if (q->rate_flags & CAKE_FLAG_OVERHEAD)
+ len -= off;
+
+ if (q->max_netlen < len)
+ q->max_netlen = len;
+ if (q->min_netlen > len)
+ q->min_netlen = len;
+
+ len += q->rate_overhead;
+
+ if (len < q->rate_mpu)
+ len = q->rate_mpu;
+
+ if (q->atm_mode == CAKE_ATM_ATM) {
+ len += 47;
+ len /= 48;
+ len *= 53;
+ } else if (q->atm_mode == CAKE_ATM_PTM) {
+ /* Add one byte per 64 bytes or part thereof.
+ * This is conservative and easier to calculate than the
+ * precise value.
+ */
+ len += (len + 63) / 64;
+ }
+
+ if (q->max_adjlen < len)
+ q->max_adjlen = len;
+ if (q->min_adjlen > len)
+ q->min_adjlen = len;
+
+ return len;
+}
+
+static u32 cake_overhead(struct cake_sched_data *q, const struct sk_buff *skb)
+{
+ const struct skb_shared_info *shinfo = skb_shinfo(skb);
+ unsigned int hdr_len, last_len = 0;
+ u32 off = skb_network_offset(skb);
+ u32 len = qdisc_pkt_len(skb);
+ u16 segs = 1;
+
+ q->avg_netoff = cake_ewma(q->avg_netoff, off << 16, 8);
+
+ if (!shinfo->gso_size)
+ return cake_calc_overhead(q, len, off);
+
+ /* borrowed from qdisc_pkt_len_init() */
+ hdr_len = skb_transport_header(skb) - skb_mac_header(skb);
+
+ /* + transport layer */
+ if (likely(shinfo->gso_type & (SKB_GSO_TCPV4 |
+ SKB_GSO_TCPV6))) {
+ const struct tcphdr *th;
+ struct tcphdr _tcphdr;
+
+ th = skb_header_pointer(skb, skb_transport_offset(skb),
+ sizeof(_tcphdr), &_tcphdr);
+ if (likely(th))
+ hdr_len += __tcp_hdrlen(th);
+ } else {
+ struct udphdr _udphdr;
+
+ if (skb_header_pointer(skb, skb_transport_offset(skb),
+ sizeof(_udphdr), &_udphdr))
+ hdr_len += sizeof(struct udphdr);
+ }
+
+ if (unlikely(shinfo->gso_type & SKB_GSO_DODGY))
+ segs = DIV_ROUND_UP(skb->len - hdr_len,
+ shinfo->gso_size);
+ else
+ segs = shinfo->gso_segs;
+
+ len = shinfo->gso_size + hdr_len;
+ last_len = skb->len - shinfo->gso_size * (segs - 1);
+
+ return (cake_calc_overhead(q, len, off) * (segs - 1) +
+ cake_calc_overhead(q, last_len, off));
+}
+
static void cake_heap_swap(struct cake_sched_data *q, u16 i, u16 j)
{
struct cake_heap_entry ii = q->overflow_heap[i];
@@ -1206,7 +1289,7 @@ static int cake_advance_shaper(struct cake_sched_data *q,
struct sk_buff *skb,
u64 now, bool drop)
{
- u32 len = qdisc_pkt_len(skb);
+ u32 len = get_cobalt_cb(skb)->adjusted_len;
/* charge packet bandwidth to this tin
* and to the global shaper.
@@ -1395,6 +1478,7 @@ static s32 cake_enqueue(struct sk_buff *skb, struct Qdisc *sch,
b->max_skblen = len;
cobalt_set_enqueue_time(skb, now);
+ get_cobalt_cb(skb)->adjusted_len = cake_overhead(q, skb);
flow_queue_add(flow, skb);
if (q->ack_filter)
@@ -2171,6 +2255,31 @@ static int cake_change(struct Qdisc *sch, struct nlattr *opt,
!!nla_get_u32(tb[TCA_CAKE_NAT]);
}
+ if (tb[TCA_CAKE_ATM])
+ q->atm_mode = nla_get_u32(tb[TCA_CAKE_ATM]);
+
+ if (tb[TCA_CAKE_OVERHEAD]) {
+ q->rate_overhead = nla_get_s32(tb[TCA_CAKE_OVERHEAD]);
+ q->rate_flags |= CAKE_FLAG_OVERHEAD;
+
+ q->max_netlen = 0;
+ q->max_adjlen = 0;
+ q->min_netlen = ~0;
+ q->min_adjlen = ~0;
+ }
+
+ if (tb[TCA_CAKE_RAW]) {
+ q->rate_flags &= ~CAKE_FLAG_OVERHEAD;
+
+ q->max_netlen = 0;
+ q->max_adjlen = 0;
+ q->min_netlen = ~0;
+ q->min_adjlen = ~0;
+ }
+
+ if (tb[TCA_CAKE_MPU])
+ q->rate_mpu = nla_get_u32(tb[TCA_CAKE_MPU]);
+
if (tb[TCA_CAKE_RTT]) {
q->interval = nla_get_u32(tb[TCA_CAKE_RTT]);
@@ -2347,6 +2456,19 @@ static int cake_dump(struct Qdisc *sch, struct sk_buff *skb)
!!(q->rate_flags & CAKE_FLAG_WASH)))
goto nla_put_failure;
+ if (nla_put_u32(skb, TCA_CAKE_OVERHEAD, q->rate_overhead))
+ goto nla_put_failure;
+
+ if (!(q->rate_flags & CAKE_FLAG_OVERHEAD))
+ if (nla_put_u32(skb, TCA_CAKE_RAW, 0))
+ goto nla_put_failure;
+
+ if (nla_put_u32(skb, TCA_CAKE_ATM, q->atm_mode))
+ goto nla_put_failure;
+
+ if (nla_put_u32(skb, TCA_CAKE_MPU, q->rate_mpu))
+ goto nla_put_failure;
+
return nla_nest_end(skb, opts);
nla_put_failure:
^ permalink raw reply related
* [PATCH net-next v11 2/7] sch_cake: Add ingress mode
From: Toke Høiland-Jørgensen @ 2018-05-15 15:12 UTC (permalink / raw)
To: netdev; +Cc: cake
In-Reply-To: <152639705919.2525.3718182212727523960.stgit@alrua-kau>
The ingress mode is meant to be enabled when CAKE runs downlink of the
actual bottleneck (such as on an IFB device). The mode changes the shaper
to also account dropped packets to the shaped rate, as these have already
traversed the bottleneck.
Enabling ingress mode will also tune the AQM to always keep at least two
packets queued *for each flow*. This is done by scaling the minimum queue
occupancy level that will disable the AQM by the number of active bulk
flows. The rationale for this is that retransmits are more expensive in
ingress mode, since dropped packets have to traverse the bottleneck again
when they are retransmitted; thus, being more lenient and keeping a minimum
number of packets queued will improve throughput in cases where the number
of active flows are so large that they saturate the bottleneck even at
their minimum window size.
This commit also adds a separate switch to enable ingress mode rate
autoscaling. If enabled, the autoscaling code will observe the actual
traffic rate and adjust the shaper rate to match it. This can help avoid
latency increases in the case where the actual bottleneck rate decreases
below the shaped rate. The scaling filters out spikes by an EWMA filter.
Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
---
net/sched/sch_cake.c | 78 +++++++++++++++++++++++++++++++++++++++++++++++---
1 file changed, 74 insertions(+), 4 deletions(-)
diff --git a/net/sched/sch_cake.c b/net/sched/sch_cake.c
index 9d7fc1848778..8d0823d6d8dd 100644
--- a/net/sched/sch_cake.c
+++ b/net/sched/sch_cake.c
@@ -442,7 +442,8 @@ static bool cobalt_queue_empty(struct cobalt_vars *vars,
static bool cobalt_should_drop(struct cobalt_vars *vars,
struct cobalt_params *p,
cobalt_time_t now,
- struct sk_buff *skb)
+ struct sk_buff *skb,
+ u32 bulk_flows)
{
bool next_due, over_target, drop = false;
cobalt_tdiff_t sojourn, schedule;
@@ -465,6 +466,7 @@ static bool cobalt_should_drop(struct cobalt_vars *vars,
sojourn = now - cobalt_get_enqueue_time(skb);
schedule = now - vars->drop_next;
over_target = sojourn > p->target &&
+ sojourn > p->mtu_time * bulk_flows * 2 &&
sojourn > p->mtu_time * 4;
next_due = vars->count && schedule >= 0;
@@ -914,6 +916,9 @@ static unsigned int cake_drop(struct Qdisc *sch, struct sk_buff **to_free)
b->tin_dropped++;
sch->qstats.drops++;
+ if (q->rate_flags & CAKE_FLAG_INGRESS)
+ cake_advance_shaper(q, b, skb, now, true);
+
__qdisc_drop(skb, to_free);
sch->q.qlen--;
@@ -989,8 +994,39 @@ static s32 cake_enqueue(struct sk_buff *skb, struct Qdisc *sch,
cake_heapify_up(q, b->overflow_idx[idx]);
/* incoming bandwidth capacity estimate */
- q->avg_window_bytes = 0;
- q->last_packet_time = now;
+ if (q->rate_flags & CAKE_FLAG_AUTORATE_INGRESS) {
+ u64 packet_interval = now - q->last_packet_time;
+
+ if (packet_interval > NSEC_PER_SEC)
+ packet_interval = NSEC_PER_SEC;
+
+ /* filter out short-term bursts, eg. wifi aggregation */
+ q->avg_packet_interval = cake_ewma(q->avg_packet_interval,
+ packet_interval,
+ packet_interval > q->avg_packet_interval ? 2 : 8);
+
+ q->last_packet_time = now;
+
+ if (packet_interval > q->avg_packet_interval) {
+ u64 window_interval = now - q->avg_window_begin;
+ u64 b = q->avg_window_bytes * (u64)NSEC_PER_SEC;
+
+ do_div(b, window_interval);
+ q->avg_peak_bandwidth =
+ cake_ewma(q->avg_peak_bandwidth, b,
+ b > q->avg_peak_bandwidth ? 2 : 8);
+ q->avg_window_bytes = 0;
+ q->avg_window_begin = now;
+
+ if (now - q->last_reconfig_time > (NSEC_PER_SEC / 4)) {
+ q->rate_bps = (q->avg_peak_bandwidth * 15) >> 4;
+ cake_reconfigure(sch);
+ }
+ }
+ } else {
+ q->avg_window_bytes = 0;
+ q->last_packet_time = now;
+ }
/* flowchain */
if (!flow->set || flow->set == CAKE_SET_DECAYING) {
@@ -1245,14 +1281,26 @@ static struct sk_buff *cake_dequeue(struct Qdisc *sch)
}
/* Last packet in queue may be marked, shouldn't be dropped */
- if (!cobalt_should_drop(&flow->cvars, &b->cparams, now, skb) ||
+ if (!cobalt_should_drop(&flow->cvars, &b->cparams, now, skb,
+ (b->bulk_flow_count *
+ !!(q->rate_flags &
+ CAKE_FLAG_INGRESS))) ||
!flow->head)
break;
+ /* drop this packet, get another one */
+ if (q->rate_flags & CAKE_FLAG_INGRESS) {
+ len = cake_advance_shaper(q, b, skb,
+ now, true);
+ flow->deficit -= len;
+ b->tin_deficit -= len;
+ }
b->tin_dropped++;
qdisc_tree_reduce_backlog(sch, 1, qdisc_pkt_len(skb));
qdisc_qstats_drop(sch);
kfree_skb(skb);
+ if (q->rate_flags & CAKE_FLAG_INGRESS)
+ goto retry;
}
b->tin_ecn_mark += !!flow->cvars.ecn_marked;
@@ -1431,6 +1479,20 @@ static int cake_change(struct Qdisc *sch, struct nlattr *opt,
q->target = 1;
}
+ if (tb[TCA_CAKE_AUTORATE]) {
+ if (!!nla_get_u32(tb[TCA_CAKE_AUTORATE]))
+ q->rate_flags |= CAKE_FLAG_AUTORATE_INGRESS;
+ else
+ q->rate_flags &= ~CAKE_FLAG_AUTORATE_INGRESS;
+ }
+
+ if (tb[TCA_CAKE_INGRESS]) {
+ if (!!nla_get_u32(tb[TCA_CAKE_INGRESS]))
+ q->rate_flags |= CAKE_FLAG_INGRESS;
+ else
+ q->rate_flags &= ~CAKE_FLAG_INGRESS;
+ }
+
if (tb[TCA_CAKE_MEMORY])
q->buffer_config_limit = nla_get_u32(tb[TCA_CAKE_MEMORY]);
@@ -1554,6 +1616,14 @@ static int cake_dump(struct Qdisc *sch, struct sk_buff *skb)
if (nla_put_u32(skb, TCA_CAKE_MEMORY, q->buffer_config_limit))
goto nla_put_failure;
+ if (nla_put_u32(skb, TCA_CAKE_AUTORATE,
+ !!(q->rate_flags & CAKE_FLAG_AUTORATE_INGRESS)))
+ goto nla_put_failure;
+
+ if (nla_put_u32(skb, TCA_CAKE_INGRESS,
+ !!(q->rate_flags & CAKE_FLAG_INGRESS)))
+ goto nla_put_failure;
+
return nla_nest_end(skb, opts);
nla_put_failure:
^ permalink raw reply related
* Re: [bpf-next V3 PATCH 4/4] xdp: change ndo_xdp_xmit API to support bulking
From: Alexei Starovoitov @ 2018-05-15 15:13 UTC (permalink / raw)
To: Jesper Dangaard Brouer
Cc: netdev, Daniel Borkmann, Christoph Hellwig, BjörnTöpel,
Magnus Karlsson, makita.toshiaki
In-Reply-To: <152638643041.9477.5642795713014271240.stgit@firesoul>
On Tue, May 15, 2018 at 02:13:50PM +0200, Jesper Dangaard Brouer wrote:
> This patch change the API for ndo_xdp_xmit to support bulking
> xdp_frames.
>
> When kernel is compiled with CONFIG_RETPOLINE, XDP sees a huge slowdown.
> Most of the slowdown is caused by DMA API indirect function calls, but
> also the net_device->ndo_xdp_xmit() call.
>
> Benchmarked patch with CONFIG_RETPOLINE, using xdp_redirect_map with
> single flow/core test (CPU E5-1650 v4 @ 3.60GHz), showed
> performance improved:
> for driver ixgbe: 6,042,682 pps -> 6,853,768 pps = +811,086 pps
> for driver i40e : 6,187,169 pps -> 6,724,519 pps = +537,350 pps
>
> With frames avail as a bulk inside the driver ndo_xdp_xmit call,
> further optimizations are possible, like bulk DMA-mapping for TX.
>
> Testing without CONFIG_RETPOLINE show the same performance for
> physical NIC drivers.
>
> The virtual NIC driver tun sees a huge performance boost, as it can
> avoid doing per frame producer locking, but instead amortize the
> locking cost over the bulk.
>
> V2: Fix compile errors reported by kbuild test robot <lkp@intel.com>
>
> Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
> ---
> drivers/net/ethernet/intel/i40e/i40e_txrx.c | 26 +++++++---
> drivers/net/ethernet/intel/i40e/i40e_txrx.h | 2 -
> drivers/net/ethernet/intel/ixgbe/ixgbe_main.c | 21 ++++++--
> drivers/net/tun.c | 37 +++++++++-----
> drivers/net/virtio_net.c | 66 +++++++++++++++++++------
> include/linux/netdevice.h | 14 +++--
> include/net/page_pool.h | 5 +-
> include/net/xdp.h | 1
> include/trace/events/xdp.h | 10 ++--
> kernel/bpf/devmap.c | 33 ++++++++-----
> net/core/filter.c | 4 +-
> net/core/xdp.c | 20 ++++++--
> samples/bpf/xdp_monitor_kern.c | 10 ++++
> samples/bpf/xdp_monitor_user.c | 35 +++++++++++--
> 14 files changed, 206 insertions(+), 78 deletions(-)
This patch has to be split into at least five:
- bpf and net core piece
- intel driver changes
- tun/virtio changes
- addition of tracepoints
- addition to samples
Putting changes from all over the areas into one patch makes it harder
to review, bisect, ack, test, merge conflicts.
Same issue with 3/4 as well. Please split it into two (core and samples).
^ permalink raw reply
* [PATCH v8] Add support for cake qdisc
From: Toke Høiland-Jørgensen @ 2018-05-15 15:13 UTC (permalink / raw)
To: netdev; +Cc: cake, Toke Høiland-Jørgensen, Dave Taht
sch_cake is intended to squeeze the most bandwidth and latency out of even
the slowest ISP links and routers, while presenting an API simple enough
that even an ISP can configure it.
Example of use on a cable ISP uplink:
tc qdisc add dev eth0 cake bandwidth 20Mbit nat docsis ack-filter
To shape a cable download link (ifb and tc-mirred setup elided)
tc qdisc add dev ifb0 cake bandwidth 200mbit nat docsis ingress wash besteffort
Cake is filled with:
* A hybrid Codel/Blue AQM algorithm, "Cobalt", tied to an FQ_Codel
derived Flow Queuing system, which autoconfigures based on the bandwidth.
* A novel "triple-isolate" mode (the default) which balances per-host
and per-flow FQ even through NAT.
* An deficit based shaper, that can also be used in an unlimited mode.
* 8 way set associative hashing to reduce flow collisions to a minimum.
* A reasonable interpretation of various diffserv latency/loss tradeoffs.
* Support for zeroing diffserv markings for entering and exiting traffic.
* Support for interacting well with Docsis 3.0 shaper framing.
* Support for DSL framing types and shapers.
* Support for ack filtering.
* Extensive statistics for measuring, loss, ecn markings, latency variation.
Various versions baking have been available as an out of tree build for
kernel versions going back to 3.10, as the embedded router world has been
running a few years behind mainline Linux. A stable version has been
generally available on lede-17.01 and later.
sch_cake replaces a combination of iptables, tc filter, htb and fq_codel
in the sqm-scripts, with sane defaults and vastly simpler configuration.
Cake's principal author is Jonathan Morton, with contributions from
Kevin Darbyshire-Bryant, Toke Høiland-Jørgensen, Sebastian Moeller,
Ryan Mounce, Guido Sarducci, Dean Scarff, Nils Andreas Svee, Dave Täht,
and Loganaden Velvindron.
Testing from Pete Heist, Georgios Amanakis, and the many other members of
the cake@lists.bufferbloat.net mailing list.
Signed-off-by: Dave Taht <dave.taht@gmail.com>
Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
---
Changelog:
v8:
- Change rates to 64bit values (apparently, 32 Gbps is not enough for
everyone).
v7:
- Move the target/interval presets to a table and check that only
one is passed.
v6:
- Identical to v5 because apparently I don't git so well... :/
v5:
- Print the SPLIT_GSO flag
- Switch to print_u64() for JSON output
- Fix a format string for mpu option output
v4:
- Switch stats parsing to use nested netlink attributes
- Tweaks to JSON stats output keys
v3:
- Remove accidentally included test flag
v2:
- Updated netlink config ABI
- Remove diffserv-llt mode
- Various tweaks and clean-ups of stats output
man/man8/tc-cake.8 | 632 ++++++++++++++++++++++++++++++++++++++
man/man8/tc.8 | 1 +
tc/Makefile | 1 +
tc/q_cake.c | 750 +++++++++++++++++++++++++++++++++++++++++++++
4 files changed, 1384 insertions(+)
create mode 100644 man/man8/tc-cake.8
create mode 100644 tc/q_cake.c
diff --git a/man/man8/tc-cake.8 b/man/man8/tc-cake.8
new file mode 100644
index 00000000..dff2e360
--- /dev/null
+++ b/man/man8/tc-cake.8
@@ -0,0 +1,632 @@
+.TH CAKE 8 "27 April 2018" "iproute2" "Linux"
+.SH NAME
+CAKE \- Common Applications Kept Enhanced (CAKE)
+.SH SYNOPSIS
+.B tc qdisc ... cake
+.br
+[
+.BR bandwidth
+RATE |
+.BR unlimited*
+|
+.BR autorate_ingress
+]
+.br
+[
+.BR rtt
+TIME |
+.BR datacentre
+|
+.BR lan
+|
+.BR metro
+|
+.BR regional
+|
+.BR internet*
+|
+.BR oceanic
+|
+.BR satellite
+|
+.BR interplanetary
+]
+.br
+[
+.BR besteffort
+|
+.BR diffserv8
+|
+.BR diffserv4
+|
+.BR diffserv3*
+]
+.br
+[
+.BR flowblind
+|
+.BR srchost
+|
+.BR dsthost
+|
+.BR hosts
+|
+.BR flows
+|
+.BR dual-srchost
+|
+.BR dual-dsthost
+|
+.BR triple-isolate*
+]
+.br
+[
+.BR nat
+|
+.BR nonat*
+]
+.br
+[
+.BR wash
+|
+.BR nowash*
+]
+.br
+[
+.BR ack-filter
+|
+.BR ack-filter-aggressive
+|
+.BR no-ack-filter*
+]
+.br
+[
+.BR memlimit
+LIMIT ]
+.br
+[
+.BR ptm
+|
+.BR atm
+|
+.BR noatm*
+]
+.br
+[
+.BR overhead
+N |
+.BR conservative
+|
+.BR raw*
+]
+.br
+[
+.BR mpu
+N ]
+.br
+[
+.BR ingress
+|
+.BR egress*
+]
+.br
+(* marks defaults)
+
+
+.SH DESCRIPTION
+CAKE (Common Applications Kept Enhanced) is a shaping-capable queue discipline
+which uses both AQM and FQ. It combines COBALT, which is an AQM algorithm
+combining Codel and BLUE, a shaper which operates in deficit mode, and a variant
+of DRR++ for flow isolation. 8-way set-associative hashing is used to virtually
+eliminate hash collisions. Priority queuing is available through a simplified
+diffserv implementation. Overhead compensation for various encapsulation
+schemes is tightly integrated.
+
+All settings are optional; the default settings are chosen to be sensible in
+most common deployments. Most people will only need to set the
+.B bandwidth
+parameter to get useful results, but reading the
+.B Overhead Compensation
+and
+.B Round Trip Time
+sections is strongly encouraged.
+
+.SH SHAPER PARAMETERS
+CAKE uses a deficit-mode shaper, which does not exhibit the initial burst
+typical of token-bucket shapers. It will automatically burst precisely as much
+as required to maintain the configured throughput. As such, it is very
+straightforward to configure.
+.PP
+.B unlimited
+(default)
+.br
+ No limit on the bandwidth.
+.PP
+.B bandwidth
+RATE
+.br
+ Set the shaper bandwidth. See
+.BR tc(8)
+or examples below for details of the RATE value.
+.PP
+.B autorate_ingress
+.br
+ Automatic capacity estimation based on traffic arriving at this qdisc.
+This is most likely to be useful with cellular links, which tend to change
+quality randomly. A
+.B bandwidth
+parameter can be used in conjunction to specify an initial estimate. The shaper
+will periodically be set to a bandwidth slightly below the estimated rate. This
+estimator cannot estimate the bandwidth of links downstream of itself.
+
+.SH OVERHEAD COMPENSATION PARAMETERS
+The size of each packet on the wire may differ from that seen by Linux. The
+following parameters allow CAKE to compensate for this difference by internally
+considering each packet to be bigger than Linux informs it. To assist users who
+are not expert network engineers, keywords have been provided to represent a
+number of common link technologies.
+
+.SS Manual Overhead Specification
+.B overhead
+BYTES
+.br
+ Adds BYTES to the size of each packet. BYTES may be negative; values
+between -64 and 256 (inclusive) are accepted.
+.PP
+.B mpu
+BYTES
+.br
+ Rounds each packet (including overhead) up to a minimum length
+BYTES. BYTES may not be negative; values between 0 and 256 (inclusive)
+are accepted.
+.PP
+.B atm
+.br
+ Compensates for ATM cell framing, which is normally found on ADSL links.
+This is performed after the
+.B overhead
+parameter above. ATM uses fixed 53-byte cells, each of which can carry 48 bytes
+payload.
+.PP
+.B ptm
+.br
+ Compensates for PTM encoding, which is normally found on VDSL2 links and
+uses a 64b/65b encoding scheme. It is even more efficient to simply
+derate the specified shaper bandwidth by a factor of 64/65 or 0.984. See
+ITU G.992.3 Annex N and IEEE 802.3 Section 61.3 for details.
+.PP
+.B noatm
+.br
+ Disables ATM and PTM compensation.
+
+.SS Failsafe Overhead Keywords
+These two keywords are provided for quick-and-dirty setup. Use them if you
+can't be bothered to read the rest of this section.
+.PP
+.B raw
+(default)
+.br
+ Turns off all overhead compensation in CAKE. The packet size reported
+by Linux will be used directly.
+.PP
+ Other overhead keywords may be added after "raw". The effect of this is
+to make the overhead compensation operate relative to the reported packet size,
+not the underlying IP packet size.
+.PP
+.B conservative
+.br
+ Compensates for more overhead than is likely to occur on any
+widely-deployed link technology.
+.br
+ Equivalent to
+.B overhead 48 atm.
+
+.SS ADSL Overhead Keywords
+Most ADSL modems have a way to check which framing scheme is in use. Often this
+is also specified in the settings document provided by the ISP. The keywords in
+this section are intended to correspond with these sources of information. All
+of them implicitly set the
+.B atm
+flag.
+.PP
+.B pppoa-vcmux
+.br
+ Equivalent to
+.B overhead 10 atm
+.PP
+.B pppoa-llc
+.br
+ Equivalent to
+.B overhead 14 atm
+.PP
+.B pppoe-vcmux
+.br
+ Equivalent to
+.B overhead 32 atm
+.PP
+.B pppoe-llcsnap
+.br
+ Equivalent to
+.B overhead 40 atm
+.PP
+.B bridged-vcmux
+.br
+ Equivalent to
+.B overhead 24 atm
+.PP
+.B bridged-llcsnap
+.br
+ Equivalent to
+.B overhead 32 atm
+.PP
+.B ipoa-vcmux
+.br
+ Equivalent to
+.B overhead 8 atm
+.PP
+.B ipoa-llcsnap
+.br
+ Equivalent to
+.B overhead 16 atm
+.PP
+See also the Ethernet Correction Factors section below.
+
+.SS VDSL2 Overhead Keywords
+ATM was dropped from VDSL2 in favour of PTM, which is a much more
+straightforward framing scheme. Some ISPs retained PPPoE for compatibility with
+their existing back-end systems.
+.PP
+.B pppoe-ptm
+.br
+ Equivalent to
+.B overhead 30 ptm
+
+.br
+ PPPoE: 2B PPP + 6B PPPoE +
+.br
+ ETHERNET: 6B dest MAC + 6B src MAC + 2B ethertype + 4B Frame Check Sequence +
+.br
+ PTM: 1B Start of Frame (S) + 1B End of Frame (Ck) + 2B TC-CRC (PTM-FCS)
+.br
+.PP
+.B bridged-ptm
+.br
+ Equivalent to
+.B overhead 22 ptm
+.br
+ ETHERNET: 6B dest MAC + 6B src MAC + 2B ethertype + 4B Frame Check Sequence +
+.br
+ PTM: 1B Start of Frame (S) + 1B End of Frame (Ck) + 2B TC-CRC (PTM-FCS)
+.br
+.PP
+See also the Ethernet Correction Factors section below.
+
+.SS DOCSIS Cable Overhead Keyword
+DOCSIS is the universal standard for providing Internet service over cable-TV
+infrastructure.
+
+In this case, the actual on-wire overhead is less important than the packet size
+the head-end equipment uses for shaping and metering. This is specified to be
+an Ethernet frame including the CRC (aka FCS).
+.PP
+.B docsis
+.br
+ Equivalent to
+.B overhead 18 mpu 64 noatm
+
+.SS Ethernet Overhead Keywords
+.PP
+.B ethernet
+.br
+ Accounts for Ethernet's preamble, inter-frame gap, and Frame Check
+Sequence. Use this keyword when the bottleneck being shaped for is an
+actual Ethernet cable.
+.br
+ Equivalent to
+.B overhead 38 mpu 84 noatm
+.PP
+.B ether-vlan
+.br
+ Adds 4 bytes to the overhead compensation, accounting for an IEEE 802.1Q
+VLAN header appended to the Ethernet frame header. NB: Some ISPs use one or
+even two of these within PPPoE; this keyword may be repeated as necessary to
+express this.
+
+.SH ROUND TRIP TIME PARAMETERS
+Active Queue Management (AQM) consists of embedding congestion signals in the
+packet flow, which receivers use to instruct senders to slow down when the queue
+is persistently occupied. CAKE uses ECN signalling when available, and packet
+drops otherwise, according to a combination of the Codel and BLUE AQM algorithms
+called COBALT.
+
+Very short latencies require a very rapid AQM response to adequately control
+latency. However, such a rapid response tends to impair throughput when the
+actual RTT is relatively long. CAKE allows specifying the RTT it assumes for
+tuning various parameters. Actual RTTs within an order of magnitude of this
+will generally work well for both throughput and latency management.
+
+At the 'lan' setting and below, the time constants are similar in magnitude to
+the jitter in the Linux kernel itself, so congestion might be signalled
+prematurely. The flows will then become sparse and total throughput reduced,
+leaving little or no back-pressure for the fairness logic to work against. Use
+the "metro" setting for local lans unless you have a custom kernel.
+.PP
+.B rtt
+TIME
+.br
+ Manually specify an RTT.
+.PP
+.B datacentre
+.br
+ For extremely high-performance 10GigE+ networks only. Equivalent to
+.B rtt 100us.
+.PP
+.B lan
+.br
+ For pure Ethernet (not Wi-Fi) networks, at home or in the office. Don't
+use this when shaping for an Internet access link. Equivalent to
+.B rtt 1ms.
+.PP
+.B metro
+.br
+ For traffic mostly within a single city. Equivalent to
+.B rtt 10ms.
+.PP
+.B regional
+.br
+ For traffic mostly within a European-sized country. Equivalent to
+.B rtt 30ms.
+.PP
+.B internet
+(default)
+.br
+ This is suitable for most Internet traffic. Equivalent to
+.B rtt 100ms.
+.PP
+.B oceanic
+.br
+ For Internet traffic with generally above-average latency, such as that
+suffered by Australasian residents. Equivalent to
+.B rtt 300ms.
+.PP
+.B satellite
+.br
+ For traffic via geostationary satellites. Equivalent to
+.B rtt 1000ms.
+.PP
+.B interplanetary
+.br
+ So named because Jupiter is about 1 light-hour from Earth. Use this to
+(almost) completely disable AQM actions. Equivalent to
+.B rtt 3600s.
+
+.SH FLOW ISOLATION PARAMETERS
+With flow isolation enabled, CAKE places packets from different flows into
+different queues, each of which carries its own AQM state. Packets from each
+queue are then delivered fairly, according to a DRR++ algorithm which minimises
+latency for "sparse" flows. CAKE uses a set-associative hashing algorithm to
+minimise flow collisions.
+
+These keywords specify whether fairness based on source address, destination
+address, individual flows, or any combination of those is desired.
+.PP
+.B flowblind
+.br
+ Disables flow isolation; all traffic passes through a single queue for
+each tin.
+.PP
+.B srchost
+.br
+ Flows are defined only by source address. Could be useful on the egress
+path of an ISP backhaul.
+.PP
+.B dsthost
+.br
+ Flows are defined only by destination address. Could be useful on the
+ingress path of an ISP backhaul.
+.PP
+.B hosts
+.br
+ Flows are defined by source-destination host pairs. This is host
+isolation, rather than flow isolation.
+.PP
+.B flows
+.br
+ Flows are defined by the entire 5-tuple of source address, destination
+address, transport protocol, source port and destination port. This is the type
+of flow isolation performed by SFQ and fq_codel.
+.PP
+.B dual-srchost
+.br
+ Flows are defined by the 5-tuple, and fairness is applied first over
+source addresses, then over individual flows. Good for use on egress traffic
+from a LAN to the internet, where it'll prevent any one LAN host from
+monopolising the uplink, regardless of the number of flows they use.
+.PP
+.B dual-dsthost
+.br
+ Flows are defined by the 5-tuple, and fairness is applied first over
+destination addresses, then over individual flows. Good for use on ingress
+traffic to a LAN from the internet, where it'll prevent any one LAN host from
+monopolising the downlink, regardless of the number of flows they use.
+.PP
+.B triple-isolate
+(default)
+.br
+ Flows are defined by the 5-tuple, and fairness is applied over source
+*and* destination addresses intelligently (ie. not merely by host-pairs), and
+also over individual flows. Use this if you're not certain whether to use
+dual-srchost or dual-dsthost; it'll do both jobs at once, preventing any one
+host on *either* side of the link from monopolising it with a large number of
+flows.
+.PP
+.B nat
+.br
+ Instructs Cake to perform a NAT lookup before applying flow-isolation
+rules, to determine the true addresses and port numbers of the packet, to
+improve fairness between hosts "inside" the NAT. This has no practical effect
+in "flowblind" or "flows" modes, or if NAT is performed on a different host.
+.PP
+.B nonat
+(default)
+.br
+ Cake will not perform a NAT lookup. Flow isolation will be performed
+using the addresses and port numbers directly visible to the interface Cake is
+attached to.
+
+.SH PRIORITY QUEUE PARAMETERS
+CAKE can divide traffic into "tins" based on the Diffserv field. Each tin has
+its own independent set of flow-isolation queues, and is serviced based on a WRR
+algorithm. To avoid perverse Diffserv marking incentives, tin weights have a
+"priority sharing" value when bandwidth used by that tin is below a threshold,
+and a lower "bandwidth sharing" value when above. Bandwidth is compared against
+the threshold using the same algorithm as the deficit-mode shaper.
+
+Detailed customisation of tin parameters is not provided. The following presets
+perform all necessary tuning, relative to the current shaper bandwidth and RTT
+settings.
+.PP
+.B besteffort
+.br
+ Disables priority queuing by placing all traffic in one tin.
+.PP
+.B precedence
+.br
+ Enables legacy interpretation of TOS "Precedence" field. Use of this
+preset on the modern Internet is firmly discouraged.
+.PP
+.B diffserv4
+.br
+ Provides a general-purpose Diffserv implementation with four tins:
+.br
+ Bulk (CS1), 6.25% threshold, generally low priority.
+.br
+ Best Effort (general), 100% threshold.
+.br
+ Video (AF4x, AF3x, CS3, AF2x, CS2, TOS4, TOS1), 50% threshold.
+.br
+ Voice (CS7, CS6, EF, VA, CS5, CS4), 25% threshold.
+.PP
+.B diffserv3
+(default)
+.br
+ Provides a simple, general-purpose Diffserv implementation with three tins:
+.br
+ Bulk (CS1), 6.25% threshold, generally low priority.
+.br
+ Best Effort (general), 100% threshold.
+.br
+ Voice (CS7, CS6, EF, VA, TOS4), 25% threshold, reduced Codel interval.
+
+.SH OTHER PARAMETERS
+.B memlimit
+LIMIT
+.br
+ Limit the memory consumed by Cake to LIMIT bytes. Note that this does
+not translate directly to queue size (so do not size this based on bandwidth
+delay product considerations, but rather on worst case acceptable memory
+consumption), as there is some overhead in the data structures containing the
+packets, especially for small packets.
+
+ By default, the limit is calculated based on the bandwidth and RTT
+settings.
+
+.PP
+.B wash
+
+.br
+ Traffic entering your diffserv domain is frequently mis-marked in
+transit from the perspective of your network, and traffic exiting yours may be
+mis-marked from the perspective of the transiting provider.
+
+Apply the wash option to clear all extra diffserv (but not ECN bits), after
+priority queuing has taken place.
+
+If you are shaping inbound, and cannot trust the diffserv markings (as is the
+case for Comcast Cable, among others), it is best to use a single queue
+"besteffort" mode with wash.
+
+.SH EXAMPLES
+# tc qdisc delete root dev eth0
+.br
+# tc qdisc add root dev eth0 cake bandwidth 100Mbit ethernet
+.br
+# tc -s qdisc show dev eth0
+.br
+qdisc cake 1: root refcnt 2 bandwidth 100Mbit diffserv3 triple-isolate rtt 100.0ms noatm overhead 38 mpu 84
+ Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
+ backlog 0b 0p requeues 0
+ memory used: 0b of 5000000b
+ capacity estimate: 100Mbit
+ min/max network layer size: 65535 / 0
+ min/max overhead-adjusted size: 65535 / 0
+ average network hdr offset: 0
+
+ Bulk Best Effort Voice
+ thresh 6250Kbit 100Mbit 25Mbit
+ target 5.0ms 5.0ms 5.0ms
+ interval 100.0ms 100.0ms 100.0ms
+ pk_delay 0us 0us 0us
+ av_delay 0us 0us 0us
+ sp_delay 0us 0us 0us
+ pkts 0 0 0
+ bytes 0 0 0
+ way_inds 0 0 0
+ way_miss 0 0 0
+ way_cols 0 0 0
+ drops 0 0 0
+ marks 0 0 0
+ ack_drop 0 0 0
+ sp_flows 0 0 0
+ bk_flows 0 0 0
+ un_flows 0 0 0
+ max_len 0 0 0
+ quantum 300 1514 762
+
+After some use:
+.br
+# tc -s qdisc show dev eth0
+
+qdisc cake 1: root refcnt 2 bandwidth 100Mbit diffserv3 triple-isolate rtt 100.0ms noatm overhead 38 mpu 84
+ Sent 44709231 bytes 31931 pkt (dropped 45, overlimits 93782 requeues 0)
+ backlog 33308b 22p requeues 0
+ memory used: 292352b of 5000000b
+ capacity estimate: 100Mbit
+ min/max network layer size: 28 / 1500
+ min/max overhead-adjusted size: 84 / 1538
+ average network hdr offset: 14
+
+ Bulk Best Effort Voice
+ thresh 6250Kbit 100Mbit 25Mbit
+ target 5.0ms 5.0ms 5.0ms
+ interval 100.0ms 100.0ms 100.0ms
+ pk_delay 8.7ms 6.9ms 5.0ms
+ av_delay 4.9ms 5.3ms 3.8ms
+ sp_delay 727us 1.4ms 511us
+ pkts 2590 21271 8137
+ bytes 3081804 30302659 11426206
+ way_inds 0 46 0
+ way_miss 3 17 4
+ way_cols 0 0 0
+ drops 20 15 10
+ marks 0 0 0
+ ack_drop 0 0 0
+ sp_flows 2 4 1
+ bk_flows 1 2 1
+ un_flows 0 0 0
+ max_len 1514 1514 1514
+ quantum 300 1514 762
+
+.SH SEE ALSO
+.BR tc (8),
+.BR tc-codel (8),
+.BR tc-fq_codel (8),
+.BR tc-red (8)
+
+.SH AUTHORS
+Cake's principal author is Jonathan Morton, with contributions from
+Tony Ambardar, Kevin Darbyshire-Bryant, Toke Høiland-Jørgensen,
+Sebastian Moeller, Ryan Mounce, Dean Scarff, Nils Andreas Svee, and Dave Täht.
+
+This manual page was written by Loganaden Velvindron. Please report corrections
+to the Linux Networking mailing list <netdev@vger.kernel.org>.
diff --git a/man/man8/tc.8 b/man/man8/tc.8
index 840880fb..716dfec5 100644
--- a/man/man8/tc.8
+++ b/man/man8/tc.8
@@ -795,6 +795,7 @@ was written by Alexey N. Kuznetsov and added in Linux 2.2.
.BR tc-basic (8),
.BR tc-bfifo (8),
.BR tc-bpf (8),
+.BR tc-cake (8),
.BR tc-cbq (8),
.BR tc-cgroup (8),
.BR tc-choke (8),
diff --git a/tc/Makefile b/tc/Makefile
index dfd00267..d9a43568 100644
--- a/tc/Makefile
+++ b/tc/Makefile
@@ -66,6 +66,7 @@ TCMODULES += q_codel.o
TCMODULES += q_fq_codel.o
TCMODULES += q_fq.o
TCMODULES += q_pie.o
+TCMODULES += q_cake.o
TCMODULES += q_hhf.o
TCMODULES += q_clsact.o
TCMODULES += e_bpf.o
diff --git a/tc/q_cake.c b/tc/q_cake.c
new file mode 100644
index 00000000..09c498a9
--- /dev/null
+++ b/tc/q_cake.c
@@ -0,0 +1,750 @@
+/* SPDX-License-Identifier: (GPL-2.0 OR BSD-3-Clause) */
+/*
+ * Common Applications Kept Enhanced -- CAKE
+ *
+ * Copyright (C) 2014-2018 Jonathan Morton <chromatix99@gmail.com>
+ * Copyright (C) 2017-2018 Toke Høiland-Jørgensen <toke@toke.dk>
+ */
+
+#include <stddef.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <syslog.h>
+#include <fcntl.h>
+#include <sys/socket.h>
+#include <netinet/in.h>
+#include <arpa/inet.h>
+#include <string.h>
+#include <inttypes.h>
+
+#include "utils.h"
+#include "tc_util.h"
+
+struct cake_preset {
+ char *name;
+ unsigned int target;
+ unsigned int interval;
+};
+
+static struct cake_preset presets[] = {
+ {"datacentre", 5, 100},
+ {"lan", 50, 1000},
+ {"metro", 500, 10000},
+ {"regional", 1500, 30000},
+ {"internet", 5000, 100000},
+ {"oceanic", 15000, 300000},
+ {"satellite", 50000, 1000000},
+ {"interplanetary", 50000000, 1000000000},
+};
+
+
+static struct cake_preset *find_preset(char *argv)
+{
+ for (int i = 0; i < ARRAY_SIZE(presets); i++)
+ if (!strcmp(argv, presets[i].name))
+ return &presets[i];
+ return NULL;
+}
+
+static void explain(void)
+{
+ fprintf(stderr,
+"Usage: ... cake [ bandwidth RATE | unlimited* | autorate_ingress ]\n"
+" [ rtt TIME | datacentre | lan | metro | regional |\n"
+" internet* | oceanic | satellite | interplanetary ]\n"
+" [ besteffort | diffserv8 | diffserv4 | diffserv3* ]\n"
+" [ flowblind | srchost | dsthost | hosts | flows |\n"
+" dual-srchost | dual-dsthost | triple-isolate* ]\n"
+" [ nat | nonat* ]\n"
+" [ wash | nowash* ]\n"
+" [ ack-filter | ack-filter-aggressive | no-ack-filter* ]\n"
+" [ memlimit LIMIT ]\n"
+" [ ptm | atm | noatm* ] [ overhead N | conservative | raw* ]\n"
+" [ mpu N ] [ ingress | egress* ]\n"
+" (* marks defaults)\n");
+}
+
+static int cake_parse_opt(struct qdisc_util *qu, int argc, char **argv,
+ struct nlmsghdr *n, const char *dev)
+{
+ int unlimited = 0;
+ __u64 bandwidth = 0;
+ unsigned interval = 0;
+ unsigned target = 0;
+ unsigned diffserv = 0;
+ unsigned memlimit = 0;
+ int overhead = 0;
+ bool overhead_set = false;
+ bool overhead_override = false;
+ int mpu = 0;
+ int flowmode = -1;
+ int nat = -1;
+ int atm = -1;
+ int autorate = -1;
+ int wash = -1;
+ int ingress = -1;
+ int ack_filter = -1;
+ struct rtattr *tail;
+ struct cake_preset *preset, *preset_set = NULL;
+
+ while (argc > 0) {
+ if (strcmp(*argv, "bandwidth") == 0) {
+ NEXT_ARG();
+ if (get_rate64(&bandwidth, *argv)) {
+ fprintf(stderr, "Illegal \"bandwidth\"\n");
+ return -1;
+ }
+ unlimited = 0;
+ autorate = 0;
+ } else if (strcmp(*argv, "unlimited") == 0) {
+ bandwidth = 0;
+ unlimited = 1;
+ autorate = 0;
+ } else if (strcmp(*argv, "autorate_ingress") == 0) {
+ autorate = 1;
+
+ } else if (strcmp(*argv, "rtt") == 0) {
+ NEXT_ARG();
+ if (get_time(&interval, *argv)) {
+ fprintf(stderr, "Illegal \"rtt\"\n");
+ return -1;
+ }
+ target = interval / 20;
+ if(!target)
+ target = 1;
+ } else if ((preset = find_preset(*argv))) {
+ if (preset_set)
+ duparg(*argv, preset_set->name);
+ preset_set = preset;
+ target = preset->target;
+ interval = preset->interval;
+
+ } else if (strcmp(*argv, "besteffort") == 0) {
+ diffserv = CAKE_DIFFSERV_BESTEFFORT;
+ } else if (strcmp(*argv, "precedence") == 0) {
+ diffserv = CAKE_DIFFSERV_PRECEDENCE;
+ } else if (strcmp(*argv, "diffserv8") == 0) {
+ diffserv = CAKE_DIFFSERV_DIFFSERV8;
+ } else if (strcmp(*argv, "diffserv4") == 0) {
+ diffserv = CAKE_DIFFSERV_DIFFSERV4;
+ } else if (strcmp(*argv, "diffserv") == 0) {
+ diffserv = CAKE_DIFFSERV_DIFFSERV4;
+ } else if (strcmp(*argv, "diffserv3") == 0) {
+ diffserv = CAKE_DIFFSERV_DIFFSERV3;
+
+ } else if (strcmp(*argv, "nowash") == 0) {
+ wash = 0;
+ } else if (strcmp(*argv, "wash") == 0) {
+ wash = 1;
+
+ } else if (strcmp(*argv, "flowblind") == 0) {
+ flowmode = CAKE_FLOW_NONE;
+ } else if (strcmp(*argv, "srchost") == 0) {
+ flowmode = CAKE_FLOW_SRC_IP;
+ } else if (strcmp(*argv, "dsthost") == 0) {
+ flowmode = CAKE_FLOW_DST_IP;
+ } else if (strcmp(*argv, "hosts") == 0) {
+ flowmode = CAKE_FLOW_HOSTS;
+ } else if (strcmp(*argv, "flows") == 0) {
+ flowmode = CAKE_FLOW_FLOWS;
+ } else if (strcmp(*argv, "dual-srchost") == 0) {
+ flowmode = CAKE_FLOW_DUAL_SRC;
+ } else if (strcmp(*argv, "dual-dsthost") == 0) {
+ flowmode = CAKE_FLOW_DUAL_DST;
+ } else if (strcmp(*argv, "triple-isolate") == 0) {
+ flowmode = CAKE_FLOW_TRIPLE;
+
+ } else if (strcmp(*argv, "nat") == 0) {
+ nat = 1;
+ } else if (strcmp(*argv, "nonat") == 0) {
+ nat = 0;
+
+ } else if (strcmp(*argv, "ptm") == 0) {
+ atm = CAKE_ATM_PTM;
+ } else if (strcmp(*argv, "atm") == 0) {
+ atm = CAKE_ATM_ATM;
+ } else if (strcmp(*argv, "noatm") == 0) {
+ atm = CAKE_ATM_NONE;
+
+ } else if (strcmp(*argv, "raw") == 0) {
+ atm = CAKE_ATM_NONE;
+ overhead = 0;
+ overhead_set = true;
+ overhead_override = true;
+ } else if (strcmp(*argv, "conservative") == 0) {
+ /*
+ * Deliberately over-estimate overhead:
+ * one whole ATM cell plus ATM framing.
+ * A safe choice if the actual overhead is unknown.
+ */
+ atm = CAKE_ATM_ATM;
+ overhead = 48;
+ overhead_set = true;
+
+ /* Various ADSL framing schemes, all over ATM cells */
+ } else if (strcmp(*argv, "ipoa-vcmux") == 0) {
+ atm = CAKE_ATM_ATM;
+ overhead += 8;
+ overhead_set = true;
+ } else if (strcmp(*argv, "ipoa-llcsnap") == 0) {
+ atm = CAKE_ATM_ATM;
+ overhead += 16;
+ overhead_set = true;
+ } else if (strcmp(*argv, "bridged-vcmux") == 0) {
+ atm = CAKE_ATM_ATM;
+ overhead += 24;
+ overhead_set = true;
+ } else if (strcmp(*argv, "bridged-llcsnap") == 0) {
+ atm = CAKE_ATM_ATM;
+ overhead += 32;
+ overhead_set = true;
+ } else if (strcmp(*argv, "pppoa-vcmux") == 0) {
+ atm = CAKE_ATM_ATM;
+ overhead += 10;
+ overhead_set = true;
+ } else if (strcmp(*argv, "pppoa-llc") == 0) {
+ atm = CAKE_ATM_ATM;
+ overhead += 14;
+ overhead_set = true;
+ } else if (strcmp(*argv, "pppoe-vcmux") == 0) {
+ atm = CAKE_ATM_ATM;
+ overhead += 32;
+ overhead_set = true;
+ } else if (strcmp(*argv, "pppoe-llcsnap") == 0) {
+ atm = CAKE_ATM_ATM;
+ overhead += 40;
+ overhead_set = true;
+
+ /* Typical VDSL2 framing schemes, both over PTM */
+ /* PTM has 64b/65b coding which absorbs some bandwidth */
+ } else if (strcmp(*argv, "pppoe-ptm") == 0) {
+ /* 2B PPP + 6B PPPoE + 6B dest MAC + 6B src MAC
+ * + 2B ethertype + 4B Frame Check Sequence
+ * + 1B Start of Frame (S) + 1B End of Frame (Ck)
+ * + 2B TC-CRC (PTM-FCS) = 30B
+ */
+ atm = CAKE_ATM_PTM;
+ overhead += 30;
+ overhead_set = true;
+ } else if (strcmp(*argv, "bridged-ptm") == 0) {
+ /* 6B dest MAC + 6B src MAC + 2B ethertype
+ * + 4B Frame Check Sequence
+ * + 1B Start of Frame (S) + 1B End of Frame (Ck)
+ * + 2B TC-CRC (PTM-FCS) = 22B
+ */
+ atm = CAKE_ATM_PTM;
+ overhead += 22;
+ overhead_set = true;
+
+ } else if (strcmp(*argv, "via-ethernet") == 0) {
+ /*
+ * We used to use this flag to manually compensate for
+ * Linux including the Ethernet header on Ethernet-type
+ * interfaces, but not on IP-type interfaces.
+ *
+ * It is no longer needed, because Cake now adjusts for
+ * that automatically, and is thus ignored.
+ *
+ * It would be deleted entirely, but it appears in the
+ * stats output when the automatic compensation is
+ * active.
+ */
+
+ } else if (strcmp(*argv, "ethernet") == 0) {
+ /* ethernet pre-amble & interframe gap & FCS
+ * you may need to add vlan tag */
+ overhead += 38;
+ overhead_set = true;
+ mpu = 84;
+
+ /* Additional Ethernet-related overhead used by some ISPs */
+ } else if (strcmp(*argv, "ether-vlan") == 0) {
+ /* 802.1q VLAN tag - may be repeated */
+ overhead += 4;
+ overhead_set = true;
+
+ /*
+ * DOCSIS cable shapers account for Ethernet frame with FCS,
+ * but not interframe gap or preamble.
+ */
+ } else if (strcmp(*argv, "docsis") == 0) {
+ atm = CAKE_ATM_NONE;
+ overhead += 18;
+ overhead_set = true;
+ mpu = 64;
+
+ } else if (strcmp(*argv, "overhead") == 0) {
+ char* p = NULL;
+ NEXT_ARG();
+ overhead = strtol(*argv, &p, 10);
+ if(!p || *p || !*argv || overhead < -64 || overhead > 256) {
+ fprintf(stderr, "Illegal \"overhead\", valid range is -64 to 256\\n");
+ return -1;
+ }
+ overhead_set = true;
+
+ } else if (strcmp(*argv, "mpu") == 0) {
+ char* p = NULL;
+ NEXT_ARG();
+ mpu = strtol(*argv, &p, 10);
+ if(!p || *p || !*argv || mpu < 0 || mpu > 256) {
+ fprintf(stderr, "Illegal \"mpu\", valid range is 0 to 256\\n");
+ return -1;
+ }
+
+ } else if (strcmp(*argv, "ingress") == 0) {
+ ingress = 1;
+ } else if (strcmp(*argv, "egress") == 0) {
+ ingress = 0;
+
+ } else if (strcmp(*argv, "no-ack-filter") == 0) {
+ ack_filter = CAKE_ACK_NONE;
+ } else if (strcmp(*argv, "ack-filter") == 0) {
+ ack_filter = CAKE_ACK_FILTER;
+ } else if (strcmp(*argv, "ack-filter-aggressive") == 0) {
+ ack_filter = CAKE_ACK_AGGRESSIVE;
+
+ } else if (strcmp(*argv, "memlimit") == 0) {
+ NEXT_ARG();
+ if(get_size(&memlimit, *argv)) {
+ fprintf(stderr, "Illegal value for \"memlimit\": \"%s\"\n", *argv);
+ return -1;
+ }
+
+ } else if (strcmp(*argv, "help") == 0) {
+ explain();
+ return -1;
+ } else {
+ fprintf(stderr, "What is \"%s\"?\n", *argv);
+ explain();
+ return -1;
+ }
+ argc--; argv++;
+ }
+
+ tail = NLMSG_TAIL(n);
+ addattr_l(n, 1024, TCA_OPTIONS, NULL, 0);
+ if (bandwidth || unlimited)
+ addattr_l(n, 1024, TCA_CAKE_BASE_RATE64, &bandwidth, sizeof(bandwidth));
+ if (diffserv)
+ addattr_l(n, 1024, TCA_CAKE_DIFFSERV_MODE, &diffserv, sizeof(diffserv));
+ if (atm != -1)
+ addattr_l(n, 1024, TCA_CAKE_ATM, &atm, sizeof(atm));
+ if (flowmode != -1)
+ addattr_l(n, 1024, TCA_CAKE_FLOW_MODE, &flowmode, sizeof(flowmode));
+ if (overhead_set)
+ addattr_l(n, 1024, TCA_CAKE_OVERHEAD, &overhead, sizeof(overhead));
+ if (overhead_override) {
+ unsigned zero = 0;
+ addattr_l(n, 1024, TCA_CAKE_RAW, &zero, sizeof(zero));
+ }
+ if (mpu > 0)
+ addattr_l(n, 1024, TCA_CAKE_MPU, &mpu, sizeof(mpu));
+ if (interval)
+ addattr_l(n, 1024, TCA_CAKE_RTT, &interval, sizeof(interval));
+ if (target)
+ addattr_l(n, 1024, TCA_CAKE_TARGET, &target, sizeof(target));
+ if (autorate != -1)
+ addattr_l(n, 1024, TCA_CAKE_AUTORATE, &autorate, sizeof(autorate));
+ if (memlimit)
+ addattr_l(n, 1024, TCA_CAKE_MEMORY, &memlimit, sizeof(memlimit));
+ if (nat != -1)
+ addattr_l(n, 1024, TCA_CAKE_NAT, &nat, sizeof(nat));
+ if (wash != -1)
+ addattr_l(n, 1024, TCA_CAKE_WASH, &wash, sizeof(wash));
+ if (ingress != -1)
+ addattr_l(n, 1024, TCA_CAKE_INGRESS, &ingress, sizeof(ingress));
+ if (ack_filter != -1)
+ addattr_l(n, 1024, TCA_CAKE_ACK_FILTER, &ack_filter, sizeof(ack_filter));
+
+ tail->rta_len = (void *) NLMSG_TAIL(n) - (void *) tail;
+ return 0;
+}
+
+
+static int cake_print_opt(struct qdisc_util *qu, FILE *f, struct rtattr *opt)
+{
+ struct rtattr *tb[TCA_CAKE_MAX + 1];
+ __u64 bandwidth = 0;
+ unsigned diffserv = 0;
+ unsigned flowmode = 0;
+ unsigned interval = 0;
+ unsigned memlimit = 0;
+ int overhead = 0;
+ int raw = 0;
+ int mpu = 0;
+ int atm = 0;
+ int nat = 0;
+ int autorate = 0;
+ int wash = 0;
+ int ingress = 0;
+ int ack_filter = 0;
+ int split_gso = 0;
+ SPRINT_BUF(b1);
+ SPRINT_BUF(b2);
+
+ if (opt == NULL)
+ return 0;
+
+ parse_rtattr_nested(tb, TCA_CAKE_MAX, opt);
+
+ if (tb[TCA_CAKE_BASE_RATE64] &&
+ RTA_PAYLOAD(tb[TCA_CAKE_BASE_RATE64]) >= sizeof(bandwidth)) {
+ bandwidth = rta_getattr_u64(tb[TCA_CAKE_BASE_RATE64]);
+ if(bandwidth) {
+ print_uint(PRINT_JSON, "bandwidth", NULL, bandwidth);
+ print_string(PRINT_FP, NULL, "bandwidth %s ", sprint_rate(bandwidth, b1));
+ } else
+ print_string(PRINT_ANY, "bandwidth", "bandwidth %s ", "unlimited");
+ }
+ if (tb[TCA_CAKE_AUTORATE] &&
+ RTA_PAYLOAD(tb[TCA_CAKE_AUTORATE]) >= sizeof(__u32)) {
+ autorate = rta_getattr_u32(tb[TCA_CAKE_AUTORATE]);
+ if(autorate == 1)
+ print_string(PRINT_ANY, "autorate", "autorate_%s ", "ingress");
+ else if(autorate)
+ print_string(PRINT_ANY, "autorate", "(?autorate?) ", "unknown");
+ }
+ if (tb[TCA_CAKE_DIFFSERV_MODE] &&
+ RTA_PAYLOAD(tb[TCA_CAKE_DIFFSERV_MODE]) >= sizeof(__u32)) {
+ diffserv = rta_getattr_u32(tb[TCA_CAKE_DIFFSERV_MODE]);
+ switch(diffserv) {
+ case CAKE_DIFFSERV_DIFFSERV3:
+ print_string(PRINT_ANY, "diffserv", "%s ", "diffserv3");
+ break;
+ case CAKE_DIFFSERV_DIFFSERV4:
+ print_string(PRINT_ANY, "diffserv", "%s ", "diffserv4");
+ break;
+ case CAKE_DIFFSERV_DIFFSERV8:
+ print_string(PRINT_ANY, "diffserv", "%s ", "diffserv8");
+ break;
+ case CAKE_DIFFSERV_BESTEFFORT:
+ print_string(PRINT_ANY, "diffserv", "%s ", "besteffort");
+ break;
+ case CAKE_DIFFSERV_PRECEDENCE:
+ print_string(PRINT_ANY, "diffserv", "%s ", "precedence");
+ break;
+ default:
+ print_string(PRINT_ANY, "diffserv", "(?diffserv?) ", "unknown");
+ break;
+ };
+ }
+ if (tb[TCA_CAKE_FLOW_MODE] &&
+ RTA_PAYLOAD(tb[TCA_CAKE_FLOW_MODE]) >= sizeof(__u32)) {
+ flowmode = rta_getattr_u32(tb[TCA_CAKE_FLOW_MODE]);
+ switch(flowmode) {
+ case CAKE_FLOW_NONE:
+ print_string(PRINT_ANY, "flowmode", "%s ", "flowblind");
+ break;
+ case CAKE_FLOW_SRC_IP:
+ print_string(PRINT_ANY, "flowmode", "%s ", "srchost");
+ break;
+ case CAKE_FLOW_DST_IP:
+ print_string(PRINT_ANY, "flowmode", "%s ", "dsthost");
+ break;
+ case CAKE_FLOW_HOSTS:
+ print_string(PRINT_ANY, "flowmode", "%s ", "hosts");
+ break;
+ case CAKE_FLOW_FLOWS:
+ print_string(PRINT_ANY, "flowmode", "%s ", "flows");
+ break;
+ case CAKE_FLOW_DUAL_SRC:
+ print_string(PRINT_ANY, "flowmode", "%s ", "dual-srchost");
+ break;
+ case CAKE_FLOW_DUAL_DST:
+ print_string(PRINT_ANY, "flowmode", "%s ", "dual-dsthost");
+ break;
+ case CAKE_FLOW_TRIPLE:
+ print_string(PRINT_ANY, "flowmode", "%s ", "triple-isolate");
+ break;
+ default:
+ print_string(PRINT_ANY, "flowmode", "(?flowmode?) ", "unknown");
+ break;
+ };
+
+ }
+
+ if (tb[TCA_CAKE_NAT] &&
+ RTA_PAYLOAD(tb[TCA_CAKE_NAT]) >= sizeof(__u32)) {
+ nat = rta_getattr_u32(tb[TCA_CAKE_NAT]);
+ }
+
+ if(nat)
+ print_string(PRINT_FP, NULL, "nat ", NULL);
+ print_bool(PRINT_JSON, "nat", NULL, nat);
+
+ if (tb[TCA_CAKE_WASH] &&
+ RTA_PAYLOAD(tb[TCA_CAKE_WASH]) >= sizeof(__u32)) {
+ wash = rta_getattr_u32(tb[TCA_CAKE_WASH]);
+ }
+ if (tb[TCA_CAKE_ATM] &&
+ RTA_PAYLOAD(tb[TCA_CAKE_ATM]) >= sizeof(__u32)) {
+ atm = rta_getattr_u32(tb[TCA_CAKE_ATM]);
+ }
+ if (tb[TCA_CAKE_OVERHEAD] &&
+ RTA_PAYLOAD(tb[TCA_CAKE_OVERHEAD]) >= sizeof(__s32)) {
+ overhead = *(__s32 *) RTA_DATA(tb[TCA_CAKE_OVERHEAD]);
+ }
+ if (tb[TCA_CAKE_MPU] &&
+ RTA_PAYLOAD(tb[TCA_CAKE_MPU]) >= sizeof(__u32)) {
+ mpu = rta_getattr_u32(tb[TCA_CAKE_MPU]);
+ }
+ if (tb[TCA_CAKE_INGRESS] &&
+ RTA_PAYLOAD(tb[TCA_CAKE_INGRESS]) >= sizeof(__u32)) {
+ ingress = rta_getattr_u32(tb[TCA_CAKE_INGRESS]);
+ }
+ if (tb[TCA_CAKE_ACK_FILTER] &&
+ RTA_PAYLOAD(tb[TCA_CAKE_ACK_FILTER]) >= sizeof(__u32)) {
+ ack_filter = rta_getattr_u32(tb[TCA_CAKE_ACK_FILTER]);
+ }
+ if (tb[TCA_CAKE_SPLIT_GSO] &&
+ RTA_PAYLOAD(tb[TCA_CAKE_SPLIT_GSO]) >= sizeof(__u32)) {
+ split_gso = rta_getattr_u32(tb[TCA_CAKE_SPLIT_GSO]);
+ }
+ if (tb[TCA_CAKE_RAW]) {
+ raw = 1;
+ }
+ if (tb[TCA_CAKE_RTT] &&
+ RTA_PAYLOAD(tb[TCA_CAKE_RTT]) >= sizeof(__u32)) {
+ interval = rta_getattr_u32(tb[TCA_CAKE_RTT]);
+ }
+
+ if (wash)
+ print_string(PRINT_FP, NULL, "wash ", NULL);
+ print_bool(PRINT_JSON, "wash", NULL, wash);
+
+ if (ingress)
+ print_string(PRINT_FP, NULL, "ingress ", NULL);
+ print_bool(PRINT_JSON, "ingress", NULL, ingress);
+
+ if (ack_filter == CAKE_ACK_AGGRESSIVE)
+ print_string(PRINT_ANY, "ack-filter", "ack-filter-%s ", "aggressive");
+ else if (ack_filter == CAKE_ACK_FILTER)
+ print_string(PRINT_ANY, "ack-filter", "ack-filter ", "enabled");
+ else
+ print_string(PRINT_JSON, "ack-filter", NULL, "disabled");
+
+ if (split_gso)
+ print_string(PRINT_FP, NULL, "split-gso ", NULL);
+ print_bool(PRINT_JSON, "split_gso", NULL, split_gso);
+
+ if (interval)
+ print_string(PRINT_FP, NULL, "rtt %s ", sprint_time(interval, b2));
+ print_uint(PRINT_JSON, "rtt", NULL, interval);
+
+ if (raw)
+ print_string(PRINT_FP, NULL, "raw ", NULL);
+ print_bool(PRINT_JSON, "raw", NULL, raw);
+
+ if (atm == CAKE_ATM_ATM)
+ print_string(PRINT_ANY, "atm", "%s ", "atm");
+ else if (atm == CAKE_ATM_PTM)
+ print_string(PRINT_ANY, "atm", "%s ", "ptm");
+ else if (!raw)
+ print_string(PRINT_ANY, "atm", "%s ", "noatm");
+
+ print_int(PRINT_ANY, "overhead", "overhead %d ", overhead);
+
+ if (mpu)
+ print_uint(PRINT_ANY, "mpu", "mpu %u ", mpu);
+
+ if (memlimit) {
+ print_uint(PRINT_JSON, "memlimit", NULL, memlimit);
+ print_string(PRINT_FP, NULL, "memlimit %s", sprint_size(memlimit, b1));
+ }
+
+ return 0;
+}
+
+static void cake_print_json_tin(struct rtattr **tstat)
+{
+#define PRINT_TSTAT_JSON(type, name, attr) if (tstat[TCA_CAKE_TIN_STATS_ ## attr]) \
+ print_u64(PRINT_JSON, name, NULL, \
+ rta_getattr_ ## type((struct rtattr *)tstat[TCA_CAKE_TIN_STATS_ ## attr]))
+
+ open_json_object(NULL);
+ PRINT_TSTAT_JSON(u64, "threshold_rate", THRESHOLD_RATE64);
+ PRINT_TSTAT_JSON(u32, "target_us", TARGET_US);
+ PRINT_TSTAT_JSON(u32, "interval_us", INTERVAL_US);
+ PRINT_TSTAT_JSON(u32, "peak_delay_us", PEAK_DELAY_US);
+ PRINT_TSTAT_JSON(u32, "avg_delay_us", AVG_DELAY_US);
+ PRINT_TSTAT_JSON(u32, "base_delay_us", BASE_DELAY_US);
+ PRINT_TSTAT_JSON(u32, "sent_packets", SENT_PACKETS);
+ PRINT_TSTAT_JSON(u64, "sent_bytes", SENT_BYTES64);
+ PRINT_TSTAT_JSON(u32, "way_indirect_hits", WAY_INDIRECT_HITS);
+ PRINT_TSTAT_JSON(u32, "way_misses", WAY_MISSES);
+ PRINT_TSTAT_JSON(u32, "way_collisions", WAY_COLLISIONS);
+ PRINT_TSTAT_JSON(u32, "drops", DROPPED_PACKETS);
+ PRINT_TSTAT_JSON(u32, "ecn_mark", ECN_MARKED_PACKETS);
+ PRINT_TSTAT_JSON(u32, "ack_drops", ACKS_DROPPED_PACKETS);
+ PRINT_TSTAT_JSON(u32, "sparse_flows", SPARSE_FLOWS);
+ PRINT_TSTAT_JSON(u32, "bulk_flows", BULK_FLOWS);
+ PRINT_TSTAT_JSON(u32, "unresponsive_flows", UNRESPONSIVE_FLOWS);
+ PRINT_TSTAT_JSON(u32, "max_pkt_len", MAX_SKBLEN);
+ PRINT_TSTAT_JSON(u32, "flow_quantum", FLOW_QUANTUM);
+ close_json_object();
+
+#undef PRINT_TSTAT_JSON
+}
+
+static int cake_print_xstats(struct qdisc_util *qu, FILE *f,
+ struct rtattr *xstats)
+{
+ SPRINT_BUF(b1);
+ struct rtattr *st[TCA_CAKE_STATS_MAX + 1];
+ int i;
+
+ if (xstats == NULL)
+ return 0;
+
+#define GET_STAT_U32(attr) rta_getattr_u32(st[TCA_CAKE_STATS_ ## attr])
+#define GET_STAT_U64(attr) rta_getattr_u64(st[TCA_CAKE_STATS_ ## attr])
+
+ parse_rtattr_nested(st, TCA_CAKE_STATS_MAX, xstats);
+
+ if (st[TCA_CAKE_STATS_MEMORY_USED] &&
+ st[TCA_CAKE_STATS_MEMORY_LIMIT]) {
+ print_string(PRINT_FP, NULL, " memory used: %s",
+ sprint_size(GET_STAT_U32(MEMORY_USED), b1));
+
+ print_string(PRINT_FP, NULL, " of %s\n",
+ sprint_size(GET_STAT_U32(MEMORY_LIMIT), b1));
+
+ print_uint(PRINT_JSON, "memory_used", NULL,
+ GET_STAT_U32(MEMORY_USED));
+ print_uint(PRINT_JSON, "memory_limit", NULL,
+ GET_STAT_U32(MEMORY_LIMIT));
+ }
+
+ if (st[TCA_CAKE_STATS_CAPACITY_ESTIMATE64]) {
+ print_string(PRINT_FP, NULL, " capacity estimate: %s\n",
+ sprint_rate(GET_STAT_U64(CAPACITY_ESTIMATE64), b1));
+ print_uint(PRINT_JSON, "capacity_estimate", NULL,
+ GET_STAT_U64(CAPACITY_ESTIMATE64));
+ }
+
+ if (st[TCA_CAKE_STATS_MIN_NETLEN] &&
+ st[TCA_CAKE_STATS_MAX_NETLEN]) {
+ print_uint(PRINT_ANY, "min_network_size",
+ " min/max network layer size: %12u",
+ GET_STAT_U32(MIN_NETLEN));
+ print_uint(PRINT_ANY, "max_network_size",
+ " /%8u\n", GET_STAT_U32(MAX_NETLEN));
+ }
+
+ if (st[TCA_CAKE_STATS_MIN_ADJLEN] &&
+ st[TCA_CAKE_STATS_MAX_ADJLEN]) {
+ print_uint(PRINT_ANY, "min_adj_size",
+ " min/max overhead-adjusted size: %8u",
+ GET_STAT_U32(MIN_ADJLEN));
+ print_uint(PRINT_ANY, "max_adj_size",
+ " /%8u\n", GET_STAT_U32(MAX_ADJLEN));
+ }
+
+ if (st[TCA_CAKE_STATS_AVG_NETOFF])
+ print_uint(PRINT_ANY, "avg_hdr_offset",
+ " average network hdr offset: %12u\n\n",
+ GET_STAT_U32(AVG_NETOFF));
+
+#undef GET_STAT_U32
+#undef GET_STAT_U64
+
+ if (st[TCA_CAKE_STATS_TIN_STATS]) {
+ struct rtattr *tins[TC_CAKE_MAX_TINS + 1];
+ struct rtattr *tstat[TC_CAKE_MAX_TINS][TCA_CAKE_TIN_STATS_MAX + 1];
+ int num_tins = 0;
+
+ parse_rtattr_nested(tins, TC_CAKE_MAX_TINS, st[TCA_CAKE_STATS_TIN_STATS]);
+
+ for (i = 1; i <= TC_CAKE_MAX_TINS && tins[i]; i++) {
+ parse_rtattr_nested(tstat[i-1], TCA_CAKE_TIN_STATS_MAX, tins[i]);
+ num_tins++;
+ }
+
+ if (!num_tins)
+ return 0;
+
+ if (is_json_context()) {
+ open_json_array(PRINT_JSON, "tins");
+ for (i = 0; i < num_tins; i++)
+ cake_print_json_tin(tstat[i]);
+ close_json_array(PRINT_JSON, NULL);
+
+ return 0;
+ }
+
+
+ switch(num_tins) {
+ case 3:
+ fprintf(f, " Bulk Best Effort Voice\n");
+ break;
+
+ case 4:
+ fprintf(f, " Bulk Best Effort Video Voice\n");
+ break;
+
+ default:
+ fprintf(f, " ");
+ for(i=0; i < num_tins; i++)
+ fprintf(f, " Tin %u", i);
+ fprintf(f, "\n");
+ };
+
+#define GET_TSTAT(i, attr) (tstat[i][TCA_CAKE_TIN_STATS_ ## attr])
+#define PRINT_TSTAT(name, attr, fmts, val) do { \
+ if (GET_TSTAT(0, attr)) { \
+ fprintf(f, name); \
+ for (i = 0; i < num_tins; i++) \
+ fprintf(f, " %12" fmts, val); \
+ fprintf(f, "\n"); \
+ } \
+ } while (0)
+
+#define SPRINT_TSTAT(pfunc, type, name, attr) PRINT_TSTAT( \
+ name, attr, "s", sprint_ ## pfunc( \
+ rta_getattr_ ## type(GET_TSTAT(i, attr)), b1))
+
+#define PRINT_TSTAT_U32(name, attr) PRINT_TSTAT( \
+ name, attr, "u", rta_getattr_u32(GET_TSTAT(i, attr)))
+
+#define PRINT_TSTAT_U64(name, attr) PRINT_TSTAT( \
+ name, attr, "llu", rta_getattr_u64(GET_TSTAT(i, attr)))
+
+ SPRINT_TSTAT(rate, u64, " thresh ", THRESHOLD_RATE64);
+ SPRINT_TSTAT(time, u32, " target ", TARGET_US);
+ SPRINT_TSTAT(time, u32, " interval", INTERVAL_US);
+ SPRINT_TSTAT(time, u32, " pk_delay", PEAK_DELAY_US);
+ SPRINT_TSTAT(time, u32, " av_delay", AVG_DELAY_US);
+ SPRINT_TSTAT(time, u32, " sp_delay", BASE_DELAY_US);
+
+ PRINT_TSTAT_U32(" pkts ", SENT_PACKETS);
+ PRINT_TSTAT_U64(" bytes ", SENT_BYTES64);
+
+ PRINT_TSTAT_U32(" way_inds", WAY_INDIRECT_HITS);
+ PRINT_TSTAT_U32(" way_miss", WAY_MISSES);
+ PRINT_TSTAT_U32(" way_cols", WAY_COLLISIONS);
+ PRINT_TSTAT_U32(" drops ", DROPPED_PACKETS);
+ PRINT_TSTAT_U32(" marks ", ECN_MARKED_PACKETS);
+ PRINT_TSTAT_U32(" ack_drop", ACKS_DROPPED_PACKETS);
+ PRINT_TSTAT_U32(" sp_flows", SPARSE_FLOWS);
+ PRINT_TSTAT_U32(" bk_flows", BULK_FLOWS);
+ PRINT_TSTAT_U32(" un_flows", UNRESPONSIVE_FLOWS);
+ PRINT_TSTAT_U32(" max_len ", MAX_SKBLEN);
+ PRINT_TSTAT_U32(" quantum ", FLOW_QUANTUM);
+
+#undef GET_STAT
+#undef PRINT_TSTAT
+#undef SPRINT_TSTAT
+#undef PRINT_TSTAT_U32
+#undef PRINT_TSTAT_U64
+ }
+ return 0;
+}
+
+struct qdisc_util cake_qdisc_util = {
+ .id = "cake",
+ .parse_qopt = cake_parse_opt,
+ .print_qopt = cake_print_opt,
+ .print_xstats = cake_print_xstats,
+};
--
2.17.0
^ permalink raw reply related
* Re: [PATCH v1 iproute2-next 2/3] rdma: print driver resource attributes
From: Jason Gunthorpe @ 2018-05-15 15:14 UTC (permalink / raw)
To: Steve Wise
Cc: 'Leon Romanovsky', dsahern, stephen, netdev, linux-rdma
In-Reply-To: <015b01d3ec5d$b1db4f90$1591eeb0$@opengridcomputing.com>
On Tue, May 15, 2018 at 10:02:08AM -0500, Steve Wise wrote:
> > On Tue, May 15, 2018 at 09:31:27AM -0500, Steve Wise wrote:
> > > > cap net admin is not high enough privledge to see unhashed kernel
> > > > pointers. CAP_RAW_IO? Or follow what printk does?
> > > >
> > >
> > > Do you mean CAP_NET_RAW? Here's the comments for it:
> >
> > Nope..
> >
> > > Func restricted_pointer() from lib/vsprintf.c uses CAP_SYSLOG. The
> > comment for CAP_SYSLOG:
> >
> > Yikes, yes, that is probably the required logic here, including the
> > kptr_restrict = 0 thing
> >
>
> Let's defer the ktpr_restrict issue for now; I want to finish the initial
> work this cycle, and adding that will likely take too much time. I'll use
> CAP_SYSLOG and add a FIXME comment. Ok?
Lets just drop wr_id instead...
Jason
^ permalink raw reply
* [PATCH 0/2] Use sbitmap instead of percpu_ida
From: Matthew Wilcox @ 2018-05-15 16:00 UTC (permalink / raw)
To: linux-kernel, linux-scsi, target-devel, linux1394-devel,
linux-usb, kvm, virtualization, netdev, Juergen Gross,
qla2xxx-upstream, Kent Overstreet, Jens Axboe
Cc: Matthew Wilcox
From: Matthew Wilcox <mawilcox@microsoft.com>
This is a pretty rough-and-ready conversion of the target drivers
from using percpu_ida to sbitmap. It compiles; I don't have a target
setup, so it's completely untested. I haven't tried to do anything
particularly clever here, so it's possible that, for example, the wait
queue in iscsi_target_util could be more clever, like the block layer
uses multiple wait queues to avoid pingpongs. Or maybe we could figure
out a way to not store the CPU that the ID was allocated on, or perhaps
the options I specified to sbitmap_queue_init() are suboptimal.
Patch 2 isn't interesting; it just deletes the implementation. Patch 1
will be where all the action is.
Matthew Wilcox (2):
Convert target drivers to use sbitmap
Remove percpu_ida
drivers/scsi/qla2xxx/qla_target.c | 16 +-
drivers/target/iscsi/iscsi_target_util.c | 34 +-
drivers/target/sbp/sbp_target.c | 8 +-
drivers/target/target_core_transport.c | 5 +-
drivers/target/tcm_fc/tfc_cmd.c | 11 +-
drivers/usb/gadget/function/f_tcm.c | 8 +-
drivers/vhost/scsi.c | 9 +-
drivers/xen/xen-scsiback.c | 8 +-
include/linux/percpu_ida.h | 83 -----
include/target/iscsi/iscsi_target_core.h | 1 +
include/target/target_core_base.h | 5 +-
lib/Makefile | 2 +-
lib/percpu_ida.c | 391 -----------------------
13 files changed, 74 insertions(+), 507 deletions(-)
delete mode 100644 include/linux/percpu_ida.h
delete mode 100644 lib/percpu_ida.c
--
2.17.0
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
^ permalink raw reply
* [PATCH 1/2] Convert target drivers to use sbitmap
From: Matthew Wilcox @ 2018-05-15 16:00 UTC (permalink / raw)
To: linux-kernel, linux-scsi, target-devel, linux1394-devel,
linux-usb, kvm, virtualization, netdev, Juergen Gross,
qla2xxx-upstream, Kent Overstreet, Jens Axboe
Cc: Matthew Wilcox
In-Reply-To: <20180515160043.27044-1-willy@infradead.org>
From: Matthew Wilcox <mawilcox@microsoft.com>
The sbitmap and the percpu_ida perform essentially the same task,
allocating tags for commands. Since the sbitmap is more used than
the percpu_ida, convert the percpu_ida users to the sbitmap API.
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
---
drivers/scsi/qla2xxx/qla_target.c | 16 ++++++-----
drivers/target/iscsi/iscsi_target_util.c | 34 +++++++++++++++++++++---
drivers/target/sbp/sbp_target.c | 8 +++---
drivers/target/target_core_transport.c | 5 ++--
drivers/target/tcm_fc/tfc_cmd.c | 11 ++++----
drivers/usb/gadget/function/f_tcm.c | 8 +++---
drivers/vhost/scsi.c | 9 ++++---
drivers/xen/xen-scsiback.c | 8 +++---
include/target/iscsi/iscsi_target_core.h | 1 +
include/target/target_core_base.h | 5 ++--
10 files changed, 73 insertions(+), 32 deletions(-)
diff --git a/drivers/scsi/qla2xxx/qla_target.c b/drivers/scsi/qla2xxx/qla_target.c
index 025dc2d3f3de..cdf671c2af61 100644
--- a/drivers/scsi/qla2xxx/qla_target.c
+++ b/drivers/scsi/qla2xxx/qla_target.c
@@ -3719,7 +3719,8 @@ void qlt_free_cmd(struct qla_tgt_cmd *cmd)
return;
}
cmd->jiffies_at_free = get_jiffies_64();
- percpu_ida_free(&sess->se_sess->sess_tag_pool, cmd->se_cmd.map_tag);
+ sbitmap_queue_clear(&sess->se_sess->sess_tag_pool, cmd->se_cmd.map_tag,
+ cmd->se_cmd.map_cpu);
}
EXPORT_SYMBOL(qlt_free_cmd);
@@ -4084,7 +4085,8 @@ static void __qlt_do_work(struct qla_tgt_cmd *cmd)
qlt_send_term_exchange(qpair, NULL, &cmd->atio, 1, 0);
qlt_decr_num_pend_cmds(vha);
- percpu_ida_free(&sess->se_sess->sess_tag_pool, cmd->se_cmd.map_tag);
+ sbitmap_queue_clear(&sess->se_sess->sess_tag_pool, cmd->se_cmd.map_tag,
+ cmd->se_cmd.map_cpu);
spin_unlock_irqrestore(qpair->qp_lock_ptr, flags);
spin_lock_irqsave(&ha->tgt.sess_lock, flags);
@@ -4215,9 +4217,9 @@ static struct qla_tgt_cmd *qlt_get_tag(scsi_qla_host_t *vha,
{
struct se_session *se_sess = sess->se_sess;
struct qla_tgt_cmd *cmd;
- int tag;
+ int tag, cpu;
- tag = percpu_ida_alloc(&se_sess->sess_tag_pool, TASK_RUNNING);
+ tag = sbitmap_queue_get(&se_sess->sess_tag_pool, &cpu);
if (tag < 0)
return NULL;
@@ -4230,6 +4232,7 @@ static struct qla_tgt_cmd *qlt_get_tag(scsi_qla_host_t *vha,
qlt_incr_num_pend_cmds(vha);
cmd->vha = vha;
cmd->se_cmd.map_tag = tag;
+ cmd->se_cmd.map_cpu = cpu;
cmd->sess = sess;
cmd->loop_id = sess->loop_id;
cmd->conf_compl_supported = sess->conf_compl_supported;
@@ -5212,7 +5215,7 @@ qlt_alloc_qfull_cmd(struct scsi_qla_host *vha,
struct fc_port *sess;
struct se_session *se_sess;
struct qla_tgt_cmd *cmd;
- int tag;
+ int tag, cpu;
unsigned long flags;
if (unlikely(tgt->tgt_stop)) {
@@ -5244,7 +5247,7 @@ qlt_alloc_qfull_cmd(struct scsi_qla_host *vha,
se_sess = sess->se_sess;
- tag = percpu_ida_alloc(&se_sess->sess_tag_pool, TASK_RUNNING);
+ tag = sbitmap_queue_get(&se_sess->sess_tag_pool, &cpu);
if (tag < 0)
return;
@@ -5275,6 +5278,7 @@ qlt_alloc_qfull_cmd(struct scsi_qla_host *vha,
cmd->reset_count = ha->base_qpair->chip_reset;
cmd->q_full = 1;
cmd->qpair = ha->base_qpair;
+ cmd->se_cmd.map_cpu = cpu;
if (qfull) {
cmd->q_full = 1;
diff --git a/drivers/target/iscsi/iscsi_target_util.c b/drivers/target/iscsi/iscsi_target_util.c
index 4435bf374d2d..28bcffae609f 100644
--- a/drivers/target/iscsi/iscsi_target_util.c
+++ b/drivers/target/iscsi/iscsi_target_util.c
@@ -17,7 +17,7 @@
******************************************************************************/
#include <linux/list.h>
-#include <linux/percpu_ida.h>
+#include <linux/sched/signal.h>
#include <net/ipv6.h> /* ipv6_addr_equal() */
#include <scsi/scsi_tcq.h>
#include <scsi/iscsi_proto.h>
@@ -147,6 +147,28 @@ void iscsit_free_r2ts_from_list(struct iscsi_cmd *cmd)
spin_unlock_bh(&cmd->r2t_lock);
}
+int iscsit_wait_for_tag(struct se_session *se_sess, int state, int *cpup)
+{
+ int tag = -1;
+ DEFINE_WAIT(wait);
+ struct sbq_wait_state *ws;
+
+ if (state == TASK_RUNNING)
+ return tag;
+
+ ws = &se_sess->sess_tag_pool.ws[0];
+ for (;;) {
+ prepare_to_wait_exclusive(&ws->wait, &wait, state);
+ if (signal_pending_state(state, current))
+ break;
+ schedule();
+ tag = sbitmap_queue_get(&se_sess->sess_tag_pool, cpup);
+ }
+
+ finish_wait(&ws->wait, &wait);
+ return tag;
+}
+
/*
* May be called from software interrupt (timer) context for allocating
* iSCSI NopINs.
@@ -155,9 +177,11 @@ struct iscsi_cmd *iscsit_allocate_cmd(struct iscsi_conn *conn, int state)
{
struct iscsi_cmd *cmd;
struct se_session *se_sess = conn->sess->se_sess;
- int size, tag;
+ int size, tag, cpu;
- tag = percpu_ida_alloc(&se_sess->sess_tag_pool, state);
+ tag = sbitmap_queue_get(&se_sess->sess_tag_pool, &cpu);
+ if (tag < 0)
+ tag = iscsit_wait_for_tag(se_sess, state, &cpu);
if (tag < 0)
return NULL;
@@ -166,6 +190,7 @@ struct iscsi_cmd *iscsit_allocate_cmd(struct iscsi_conn *conn, int state)
memset(cmd, 0, size);
cmd->se_cmd.map_tag = tag;
+ cmd->se_cmd.map_cpu = cpu;
cmd->conn = conn;
cmd->data_direction = DMA_NONE;
INIT_LIST_HEAD(&cmd->i_conn_node);
@@ -711,7 +736,8 @@ void iscsit_release_cmd(struct iscsi_cmd *cmd)
kfree(cmd->iov_data);
kfree(cmd->text_in_ptr);
- percpu_ida_free(&sess->se_sess->sess_tag_pool, se_cmd->map_tag);
+ sbitmap_queue_clear(&sess->se_sess->sess_tag_pool, se_cmd->map_tag,
+ se_cmd->map_cpu);
}
EXPORT_SYMBOL(iscsit_release_cmd);
diff --git a/drivers/target/sbp/sbp_target.c b/drivers/target/sbp/sbp_target.c
index fb1003921d85..c58f9f04c6be 100644
--- a/drivers/target/sbp/sbp_target.c
+++ b/drivers/target/sbp/sbp_target.c
@@ -926,15 +926,16 @@ static struct sbp_target_request *sbp_mgt_get_req(struct sbp_session *sess,
{
struct se_session *se_sess = sess->se_sess;
struct sbp_target_request *req;
- int tag;
+ int tag, cpu;
- tag = percpu_ida_alloc(&se_sess->sess_tag_pool, TASK_RUNNING);
+ tag = sbitmap_queue_get(&se_sess->sess_tag_pool, &cpu);
if (tag < 0)
return ERR_PTR(-ENOMEM);
req = &((struct sbp_target_request *)se_sess->sess_cmd_map)[tag];
memset(req, 0, sizeof(*req));
req->se_cmd.map_tag = tag;
+ req->se_cmd.map_cpu = cpu;
req->se_cmd.tag = next_orb;
return req;
@@ -1460,7 +1461,8 @@ static void sbp_free_request(struct sbp_target_request *req)
kfree(req->pg_tbl);
kfree(req->cmd_buf);
- percpu_ida_free(&se_sess->sess_tag_pool, se_cmd->map_tag);
+ sbitmap_queue_clear(&se_sess->sess_tag_pool, se_cmd->map_tag,
+ se_cmd->map_cpu);
}
static void sbp_mgt_agent_process(struct work_struct *work)
diff --git a/drivers/target/target_core_transport.c b/drivers/target/target_core_transport.c
index 4558f2e1fe1b..3103890ed109 100644
--- a/drivers/target/target_core_transport.c
+++ b/drivers/target/target_core_transport.c
@@ -260,7 +260,8 @@ int transport_alloc_session_tags(struct se_session *se_sess,
}
}
- rc = percpu_ida_init(&se_sess->sess_tag_pool, tag_num);
+ rc = sbitmap_queue_init_node(&se_sess->sess_tag_pool, tag_num, -1,
+ false, GFP_KERNEL, NUMA_NO_NODE);
if (rc < 0) {
pr_err("Unable to init se_sess->sess_tag_pool,"
" tag_num: %u\n", tag_num);
@@ -547,7 +548,7 @@ void transport_free_session(struct se_session *se_sess)
target_put_nacl(se_nacl);
}
if (se_sess->sess_cmd_map) {
- percpu_ida_destroy(&se_sess->sess_tag_pool);
+ sbitmap_queue_free(&se_sess->sess_tag_pool);
kvfree(se_sess->sess_cmd_map);
}
kmem_cache_free(se_sess_cache, se_sess);
diff --git a/drivers/target/tcm_fc/tfc_cmd.c b/drivers/target/tcm_fc/tfc_cmd.c
index ec372860106f..b3e3364b7147 100644
--- a/drivers/target/tcm_fc/tfc_cmd.c
+++ b/drivers/target/tcm_fc/tfc_cmd.c
@@ -28,7 +28,6 @@
#include <linux/configfs.h>
#include <linux/ctype.h>
#include <linux/hash.h>
-#include <linux/percpu_ida.h>
#include <asm/unaligned.h>
#include <scsi/scsi_tcq.h>
#include <scsi/libfc.h>
@@ -92,7 +91,8 @@ static void ft_free_cmd(struct ft_cmd *cmd)
if (fr_seq(fp))
fc_seq_release(fr_seq(fp));
fc_frame_free(fp);
- percpu_ida_free(&sess->se_sess->sess_tag_pool, cmd->se_cmd.map_tag);
+ sbitmap_queue_clear(&sess->se_sess->sess_tag_pool, cmd->se_cmd.map_tag,
+ cmd->se_cmd.map_cpu);
ft_sess_put(sess); /* undo get from lookup at recv */
}
@@ -448,9 +448,9 @@ static void ft_recv_cmd(struct ft_sess *sess, struct fc_frame *fp)
struct ft_cmd *cmd;
struct fc_lport *lport = sess->tport->lport;
struct se_session *se_sess = sess->se_sess;
- int tag;
+ int tag, cpu;
- tag = percpu_ida_alloc(&se_sess->sess_tag_pool, TASK_RUNNING);
+ tag = sbitmap_queue_get(&se_sess->sess_tag_pool, &cpu);
if (tag < 0)
goto busy;
@@ -458,10 +458,11 @@ static void ft_recv_cmd(struct ft_sess *sess, struct fc_frame *fp)
memset(cmd, 0, sizeof(struct ft_cmd));
cmd->se_cmd.map_tag = tag;
+ cmd->se_cmd.map_cpu = cpu;
cmd->sess = sess;
cmd->seq = fc_seq_assign(lport, fp);
if (!cmd->seq) {
- percpu_ida_free(&se_sess->sess_tag_pool, tag);
+ sbitmap_queue_clear(&se_sess->sess_tag_pool, tag, cpu);
goto busy;
}
cmd->req_frame = fp; /* hold frame during cmd */
diff --git a/drivers/usb/gadget/function/f_tcm.c b/drivers/usb/gadget/function/f_tcm.c
index d78dbb73bde8..b335f4f33bc3 100644
--- a/drivers/usb/gadget/function/f_tcm.c
+++ b/drivers/usb/gadget/function/f_tcm.c
@@ -1071,15 +1071,16 @@ static struct usbg_cmd *usbg_get_cmd(struct f_uas *fu,
{
struct se_session *se_sess = tv_nexus->tvn_se_sess;
struct usbg_cmd *cmd;
- int tag;
+ int tag, cpu;
- tag = percpu_ida_alloc(&se_sess->sess_tag_pool, TASK_RUNNING);
+ tag = sbitmap_queue_get(&se_sess->sess_tag_pool, &cpu);
if (tag < 0)
return ERR_PTR(-ENOMEM);
cmd = &((struct usbg_cmd *)se_sess->sess_cmd_map)[tag];
memset(cmd, 0, sizeof(*cmd));
cmd->se_cmd.map_tag = tag;
+ cmd->se_cmd.map_cpu = cpu;
cmd->se_cmd.tag = cmd->tag = scsi_tag;
cmd->fu = fu;
@@ -1288,7 +1289,8 @@ static void usbg_release_cmd(struct se_cmd *se_cmd)
struct se_session *se_sess = se_cmd->se_sess;
kfree(cmd->data_buf);
- percpu_ida_free(&se_sess->sess_tag_pool, se_cmd->map_tag);
+ sbitmap_queue_clear(&se_sess->sess_tag_pool, se_cmd->map_tag,
+ se_cmd->map_cpu);
}
static u32 usbg_sess_get_index(struct se_session *se_sess)
diff --git a/drivers/vhost/scsi.c b/drivers/vhost/scsi.c
index 7ad57094d736..1fadaa39f322 100644
--- a/drivers/vhost/scsi.c
+++ b/drivers/vhost/scsi.c
@@ -46,7 +46,6 @@
#include <linux/virtio_scsi.h>
#include <linux/llist.h>
#include <linux/bitmap.h>
-#include <linux/percpu_ida.h>
#include "vhost.h"
@@ -324,7 +323,8 @@ static void vhost_scsi_release_cmd(struct se_cmd *se_cmd)
}
vhost_scsi_put_inflight(tv_cmd->inflight);
- percpu_ida_free(&se_sess->sess_tag_pool, se_cmd->map_tag);
+ sbitmap_queue_clear(&se_sess->sess_tag_pool, se_cmd->map_tag,
+ se_cmd->map_cpu);
}
static u32 vhost_scsi_sess_get_index(struct se_session *se_sess)
@@ -567,7 +567,7 @@ vhost_scsi_get_tag(struct vhost_virtqueue *vq, struct vhost_scsi_tpg *tpg,
struct se_session *se_sess;
struct scatterlist *sg, *prot_sg;
struct page **pages;
- int tag;
+ int tag, cpu;
tv_nexus = tpg->tpg_nexus;
if (!tv_nexus) {
@@ -576,7 +576,7 @@ vhost_scsi_get_tag(struct vhost_virtqueue *vq, struct vhost_scsi_tpg *tpg,
}
se_sess = tv_nexus->tvn_se_sess;
- tag = percpu_ida_alloc(&se_sess->sess_tag_pool, TASK_RUNNING);
+ tag = sbitmap_queue_get(&se_sess->sess_tag_pool, &cpu);
if (tag < 0) {
pr_err("Unable to obtain tag for vhost_scsi_cmd\n");
return ERR_PTR(-ENOMEM);
@@ -591,6 +591,7 @@ vhost_scsi_get_tag(struct vhost_virtqueue *vq, struct vhost_scsi_tpg *tpg,
cmd->tvc_prot_sgl = prot_sg;
cmd->tvc_upages = pages;
cmd->tvc_se_cmd.map_tag = tag;
+ cmd->tvc_se_cmd.map_cpu = cpu;
cmd->tvc_tag = scsi_tag;
cmd->tvc_lun = lun;
cmd->tvc_task_attr = task_attr;
diff --git a/drivers/xen/xen-scsiback.c b/drivers/xen/xen-scsiback.c
index 7bc88fd43cfc..d2c71b8608f0 100644
--- a/drivers/xen/xen-scsiback.c
+++ b/drivers/xen/xen-scsiback.c
@@ -654,9 +654,9 @@ static struct vscsibk_pend *scsiback_get_pend_req(struct vscsiif_back_ring *ring
struct scsiback_nexus *nexus = tpg->tpg_nexus;
struct se_session *se_sess = nexus->tvn_se_sess;
struct vscsibk_pend *req;
- int tag, i;
+ int tag, cpu, i;
- tag = percpu_ida_alloc(&se_sess->sess_tag_pool, TASK_RUNNING);
+ tag = sbitmap_queue_get(&se_sess->sess_tag_pool, &cpu);
if (tag < 0) {
pr_err("Unable to obtain tag for vscsiif_request\n");
return ERR_PTR(-ENOMEM);
@@ -665,6 +665,7 @@ static struct vscsibk_pend *scsiback_get_pend_req(struct vscsiif_back_ring *ring
req = &((struct vscsibk_pend *)se_sess->sess_cmd_map)[tag];
memset(req, 0, sizeof(*req));
req->se_cmd.map_tag = tag;
+ req->se_cmd.map_cpu = cpu;
for (i = 0; i < VSCSI_MAX_GRANTS; i++)
req->grant_handles[i] = SCSIBACK_INVALID_HANDLE;
@@ -1379,7 +1380,8 @@ static void scsiback_release_cmd(struct se_cmd *se_cmd)
{
struct se_session *se_sess = se_cmd->se_sess;
- percpu_ida_free(&se_sess->sess_tag_pool, se_cmd->map_tag);
+ sbitmap_queue_clear(&se_sess->sess_tag_pool, se_cmd->map_tag,
+ se_cmd->map_cpu);
}
static u32 scsiback_sess_get_index(struct se_session *se_sess)
diff --git a/include/target/iscsi/iscsi_target_core.h b/include/target/iscsi/iscsi_target_core.h
index cf5f3fff1f1a..f2e6abea8490 100644
--- a/include/target/iscsi/iscsi_target_core.h
+++ b/include/target/iscsi/iscsi_target_core.h
@@ -4,6 +4,7 @@
#include <linux/dma-direction.h> /* enum dma_data_direction */
#include <linux/list.h> /* struct list_head */
+#include <linux/sched.h>
#include <linux/socket.h> /* struct sockaddr_storage */
#include <linux/types.h> /* u8 */
#include <scsi/iscsi_proto.h> /* itt_t */
diff --git a/include/target/target_core_base.h b/include/target/target_core_base.h
index 9f9f5902af38..cd417b17fee6 100644
--- a/include/target/target_core_base.h
+++ b/include/target/target_core_base.h
@@ -4,7 +4,7 @@
#include <linux/configfs.h> /* struct config_group */
#include <linux/dma-direction.h> /* enum dma_data_direction */
-#include <linux/percpu_ida.h> /* struct percpu_ida */
+#include <linux/sbitmap.h>
#include <linux/percpu-refcount.h>
#include <linux/semaphore.h> /* struct semaphore */
#include <linux/completion.h>
@@ -454,6 +454,7 @@ struct se_cmd {
int sam_task_attr;
/* Used for se_sess->sess_tag_pool */
unsigned int map_tag;
+ int map_cpu;
/* Transport protocol dependent state, see transport_state_table */
enum transport_state_table t_state;
/* See se_cmd_flags_table */
@@ -607,7 +608,7 @@ struct se_session {
struct list_head sess_wait_list;
spinlock_t sess_cmd_lock;
void *sess_cmd_map;
- struct percpu_ida sess_tag_pool;
+ struct sbitmap_queue sess_tag_pool;
};
struct se_device;
--
2.17.0
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
^ permalink raw reply related
* [PATCH 2/2] Remove percpu_ida
From: Matthew Wilcox @ 2018-05-15 16:00 UTC (permalink / raw)
To: linux-kernel, linux-scsi, target-devel, linux1394-devel,
linux-usb, kvm, virtualization, netdev, Juergen Gross,
qla2xxx-upstream, Kent Overstreet, Jens Axboe
Cc: Matthew Wilcox
In-Reply-To: <20180515160043.27044-1-willy@infradead.org>
From: Matthew Wilcox <mawilcox@microsoft.com>
With its one user gone, remove the library code.
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
---
include/linux/percpu_ida.h | 83 --------
lib/Makefile | 2 +-
lib/percpu_ida.c | 391 -------------------------------------
3 files changed, 1 insertion(+), 475 deletions(-)
delete mode 100644 include/linux/percpu_ida.h
delete mode 100644 lib/percpu_ida.c
diff --git a/include/linux/percpu_ida.h b/include/linux/percpu_ida.h
deleted file mode 100644
index 07d78e4653bc..000000000000
--- a/include/linux/percpu_ida.h
+++ /dev/null
@@ -1,83 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0 */
-#ifndef __PERCPU_IDA_H__
-#define __PERCPU_IDA_H__
-
-#include <linux/types.h>
-#include <linux/bitops.h>
-#include <linux/init.h>
-#include <linux/sched.h>
-#include <linux/spinlock_types.h>
-#include <linux/wait.h>
-#include <linux/cpumask.h>
-
-struct percpu_ida_cpu;
-
-struct percpu_ida {
- /*
- * number of tags available to be allocated, as passed to
- * percpu_ida_init()
- */
- unsigned nr_tags;
- unsigned percpu_max_size;
- unsigned percpu_batch_size;
-
- struct percpu_ida_cpu __percpu *tag_cpu;
-
- /*
- * Bitmap of cpus that (may) have tags on their percpu freelists:
- * steal_tags() uses this to decide when to steal tags, and which cpus
- * to try stealing from.
- *
- * It's ok for a freelist to be empty when its bit is set - steal_tags()
- * will just keep looking - but the bitmap _must_ be set whenever a
- * percpu freelist does have tags.
- */
- cpumask_t cpus_have_tags;
-
- struct {
- spinlock_t lock;
- /*
- * When we go to steal tags from another cpu (see steal_tags()),
- * we want to pick a cpu at random. Cycling through them every
- * time we steal is a bit easier and more or less equivalent:
- */
- unsigned cpu_last_stolen;
-
- /* For sleeping on allocation failure */
- wait_queue_head_t wait;
-
- /*
- * Global freelist - it's a stack where nr_free points to the
- * top
- */
- unsigned nr_free;
- unsigned *freelist;
- } ____cacheline_aligned_in_smp;
-};
-
-/*
- * Number of tags we move between the percpu freelist and the global freelist at
- * a time
- */
-#define IDA_DEFAULT_PCPU_BATCH_MOVE 32U
-/* Max size of percpu freelist, */
-#define IDA_DEFAULT_PCPU_SIZE ((IDA_DEFAULT_PCPU_BATCH_MOVE * 3) / 2)
-
-int percpu_ida_alloc(struct percpu_ida *pool, int state);
-void percpu_ida_free(struct percpu_ida *pool, unsigned tag);
-
-void percpu_ida_destroy(struct percpu_ida *pool);
-int __percpu_ida_init(struct percpu_ida *pool, unsigned long nr_tags,
- unsigned long max_size, unsigned long batch_size);
-static inline int percpu_ida_init(struct percpu_ida *pool, unsigned long nr_tags)
-{
- return __percpu_ida_init(pool, nr_tags, IDA_DEFAULT_PCPU_SIZE,
- IDA_DEFAULT_PCPU_BATCH_MOVE);
-}
-
-typedef int (*percpu_ida_cb)(unsigned, void *);
-int percpu_ida_for_each_free(struct percpu_ida *pool, percpu_ida_cb fn,
- void *data);
-
-unsigned percpu_ida_free_tags(struct percpu_ida *pool, int cpu);
-#endif /* __PERCPU_IDA_H__ */
diff --git a/lib/Makefile b/lib/Makefile
index ce20696d5a92..7626dece1d27 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -39,7 +39,7 @@ obj-y += bcd.o div64.o sort.o parser.o debug_locks.o random32.o \
bust_spinlocks.o kasprintf.o bitmap.o scatterlist.o \
gcd.o lcm.o list_sort.o uuid.o flex_array.o iov_iter.o clz_ctz.o \
bsearch.o find_bit.o llist.o memweight.o kfifo.o \
- percpu-refcount.o percpu_ida.o rhashtable.o reciprocal_div.o \
+ percpu-refcount.o rhashtable.o reciprocal_div.o \
once.o refcount.o usercopy.o errseq.o bucket_locks.o
obj-$(CONFIG_STRING_SELFTEST) += test_string.o
obj-y += string_helpers.o
diff --git a/lib/percpu_ida.c b/lib/percpu_ida.c
deleted file mode 100644
index 6016f1deb1f5..000000000000
--- a/lib/percpu_ida.c
+++ /dev/null
@@ -1,391 +0,0 @@
-/*
- * Percpu IDA library
- *
- * Copyright (C) 2013 Datera, Inc. Kent Overstreet
- *
- * This program is free software; you can redistribute it and/or
- * modify it under the terms of the GNU General Public License as
- * published by the Free Software Foundation; either version 2, or (at
- * your option) any later version.
- *
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- * General Public License for more details.
- */
-
-#include <linux/mm.h>
-#include <linux/bitmap.h>
-#include <linux/bitops.h>
-#include <linux/bug.h>
-#include <linux/err.h>
-#include <linux/export.h>
-#include <linux/init.h>
-#include <linux/kernel.h>
-#include <linux/percpu.h>
-#include <linux/sched/signal.h>
-#include <linux/string.h>
-#include <linux/spinlock.h>
-#include <linux/percpu_ida.h>
-
-struct percpu_ida_cpu {
- /*
- * Even though this is percpu, we need a lock for tag stealing by remote
- * CPUs:
- */
- spinlock_t lock;
-
- /* nr_free/freelist form a stack of free IDs */
- unsigned nr_free;
- unsigned freelist[];
-};
-
-static inline void move_tags(unsigned *dst, unsigned *dst_nr,
- unsigned *src, unsigned *src_nr,
- unsigned nr)
-{
- *src_nr -= nr;
- memcpy(dst + *dst_nr, src + *src_nr, sizeof(unsigned) * nr);
- *dst_nr += nr;
-}
-
-/*
- * Try to steal tags from a remote cpu's percpu freelist.
- *
- * We first check how many percpu freelists have tags
- *
- * Then we iterate through the cpus until we find some tags - we don't attempt
- * to find the "best" cpu to steal from, to keep cacheline bouncing to a
- * minimum.
- */
-static inline void steal_tags(struct percpu_ida *pool,
- struct percpu_ida_cpu *tags)
-{
- unsigned cpus_have_tags, cpu = pool->cpu_last_stolen;
- struct percpu_ida_cpu *remote;
-
- for (cpus_have_tags = cpumask_weight(&pool->cpus_have_tags);
- cpus_have_tags; cpus_have_tags--) {
- cpu = cpumask_next(cpu, &pool->cpus_have_tags);
-
- if (cpu >= nr_cpu_ids) {
- cpu = cpumask_first(&pool->cpus_have_tags);
- if (cpu >= nr_cpu_ids)
- BUG();
- }
-
- pool->cpu_last_stolen = cpu;
- remote = per_cpu_ptr(pool->tag_cpu, cpu);
-
- cpumask_clear_cpu(cpu, &pool->cpus_have_tags);
-
- if (remote == tags)
- continue;
-
- spin_lock(&remote->lock);
-
- if (remote->nr_free) {
- memcpy(tags->freelist,
- remote->freelist,
- sizeof(unsigned) * remote->nr_free);
-
- tags->nr_free = remote->nr_free;
- remote->nr_free = 0;
- }
-
- spin_unlock(&remote->lock);
-
- if (tags->nr_free)
- break;
- }
-}
-
-/*
- * Pop up to IDA_PCPU_BATCH_MOVE IDs off the global freelist, and push them onto
- * our percpu freelist:
- */
-static inline void alloc_global_tags(struct percpu_ida *pool,
- struct percpu_ida_cpu *tags)
-{
- move_tags(tags->freelist, &tags->nr_free,
- pool->freelist, &pool->nr_free,
- min(pool->nr_free, pool->percpu_batch_size));
-}
-
-static inline unsigned alloc_local_tag(struct percpu_ida_cpu *tags)
-{
- int tag = -ENOSPC;
-
- spin_lock(&tags->lock);
- if (tags->nr_free)
- tag = tags->freelist[--tags->nr_free];
- spin_unlock(&tags->lock);
-
- return tag;
-}
-
-/**
- * percpu_ida_alloc - allocate a tag
- * @pool: pool to allocate from
- * @state: task state for prepare_to_wait
- *
- * Returns a tag - an integer in the range [0..nr_tags) (passed to
- * tag_pool_init()), or otherwise -ENOSPC on allocation failure.
- *
- * Safe to be called from interrupt context (assuming it isn't passed
- * TASK_UNINTERRUPTIBLE | TASK_INTERRUPTIBLE, of course).
- *
- * @gfp indicates whether or not to wait until a free id is available (it's not
- * used for internal memory allocations); thus if passed __GFP_RECLAIM we may sleep
- * however long it takes until another thread frees an id (same semantics as a
- * mempool).
- *
- * Will not fail if passed TASK_UNINTERRUPTIBLE | TASK_INTERRUPTIBLE.
- */
-int percpu_ida_alloc(struct percpu_ida *pool, int state)
-{
- DEFINE_WAIT(wait);
- struct percpu_ida_cpu *tags;
- unsigned long flags;
- int tag;
-
- local_irq_save(flags);
- tags = this_cpu_ptr(pool->tag_cpu);
-
- /* Fastpath */
- tag = alloc_local_tag(tags);
- if (likely(tag >= 0)) {
- local_irq_restore(flags);
- return tag;
- }
-
- while (1) {
- spin_lock(&pool->lock);
-
- /*
- * prepare_to_wait() must come before steal_tags(), in case
- * percpu_ida_free() on another cpu flips a bit in
- * cpus_have_tags
- *
- * global lock held and irqs disabled, don't need percpu lock
- */
- if (state != TASK_RUNNING)
- prepare_to_wait(&pool->wait, &wait, state);
-
- if (!tags->nr_free)
- alloc_global_tags(pool, tags);
- if (!tags->nr_free)
- steal_tags(pool, tags);
-
- if (tags->nr_free) {
- tag = tags->freelist[--tags->nr_free];
- if (tags->nr_free)
- cpumask_set_cpu(smp_processor_id(),
- &pool->cpus_have_tags);
- }
-
- spin_unlock(&pool->lock);
- local_irq_restore(flags);
-
- if (tag >= 0 || state == TASK_RUNNING)
- break;
-
- if (signal_pending_state(state, current)) {
- tag = -ERESTARTSYS;
- break;
- }
-
- schedule();
-
- local_irq_save(flags);
- tags = this_cpu_ptr(pool->tag_cpu);
- }
- if (state != TASK_RUNNING)
- finish_wait(&pool->wait, &wait);
-
- return tag;
-}
-EXPORT_SYMBOL_GPL(percpu_ida_alloc);
-
-/**
- * percpu_ida_free - free a tag
- * @pool: pool @tag was allocated from
- * @tag: a tag previously allocated with percpu_ida_alloc()
- *
- * Safe to be called from interrupt context.
- */
-void percpu_ida_free(struct percpu_ida *pool, unsigned tag)
-{
- struct percpu_ida_cpu *tags;
- unsigned long flags;
- unsigned nr_free;
-
- BUG_ON(tag >= pool->nr_tags);
-
- local_irq_save(flags);
- tags = this_cpu_ptr(pool->tag_cpu);
-
- spin_lock(&tags->lock);
- tags->freelist[tags->nr_free++] = tag;
-
- nr_free = tags->nr_free;
- spin_unlock(&tags->lock);
-
- if (nr_free == 1) {
- cpumask_set_cpu(smp_processor_id(),
- &pool->cpus_have_tags);
- wake_up(&pool->wait);
- }
-
- if (nr_free == pool->percpu_max_size) {
- spin_lock(&pool->lock);
-
- /*
- * Global lock held and irqs disabled, don't need percpu
- * lock
- */
- if (tags->nr_free == pool->percpu_max_size) {
- move_tags(pool->freelist, &pool->nr_free,
- tags->freelist, &tags->nr_free,
- pool->percpu_batch_size);
-
- wake_up(&pool->wait);
- }
- spin_unlock(&pool->lock);
- }
-
- local_irq_restore(flags);
-}
-EXPORT_SYMBOL_GPL(percpu_ida_free);
-
-/**
- * percpu_ida_destroy - release a tag pool's resources
- * @pool: pool to free
- *
- * Frees the resources allocated by percpu_ida_init().
- */
-void percpu_ida_destroy(struct percpu_ida *pool)
-{
- free_percpu(pool->tag_cpu);
- free_pages((unsigned long) pool->freelist,
- get_order(pool->nr_tags * sizeof(unsigned)));
-}
-EXPORT_SYMBOL_GPL(percpu_ida_destroy);
-
-/**
- * percpu_ida_init - initialize a percpu tag pool
- * @pool: pool to initialize
- * @nr_tags: number of tags that will be available for allocation
- *
- * Initializes @pool so that it can be used to allocate tags - integers in the
- * range [0, nr_tags). Typically, they'll be used by driver code to refer to a
- * preallocated array of tag structures.
- *
- * Allocation is percpu, but sharding is limited by nr_tags - for best
- * performance, the workload should not span more cpus than nr_tags / 128.
- */
-int __percpu_ida_init(struct percpu_ida *pool, unsigned long nr_tags,
- unsigned long max_size, unsigned long batch_size)
-{
- unsigned i, cpu, order;
-
- memset(pool, 0, sizeof(*pool));
-
- init_waitqueue_head(&pool->wait);
- spin_lock_init(&pool->lock);
- pool->nr_tags = nr_tags;
- pool->percpu_max_size = max_size;
- pool->percpu_batch_size = batch_size;
-
- /* Guard against overflow */
- if (nr_tags > (unsigned) INT_MAX + 1) {
- pr_err("percpu_ida_init(): nr_tags too large\n");
- return -EINVAL;
- }
-
- order = get_order(nr_tags * sizeof(unsigned));
- pool->freelist = (void *) __get_free_pages(GFP_KERNEL, order);
- if (!pool->freelist)
- return -ENOMEM;
-
- for (i = 0; i < nr_tags; i++)
- pool->freelist[i] = i;
-
- pool->nr_free = nr_tags;
-
- pool->tag_cpu = __alloc_percpu(sizeof(struct percpu_ida_cpu) +
- pool->percpu_max_size * sizeof(unsigned),
- sizeof(unsigned));
- if (!pool->tag_cpu)
- goto err;
-
- for_each_possible_cpu(cpu)
- spin_lock_init(&per_cpu_ptr(pool->tag_cpu, cpu)->lock);
-
- return 0;
-err:
- percpu_ida_destroy(pool);
- return -ENOMEM;
-}
-EXPORT_SYMBOL_GPL(__percpu_ida_init);
-
-/**
- * percpu_ida_for_each_free - iterate free ids of a pool
- * @pool: pool to iterate
- * @fn: interate callback function
- * @data: parameter for @fn
- *
- * Note, this doesn't guarantee to iterate all free ids restrictly. Some free
- * ids might be missed, some might be iterated duplicated, and some might
- * be iterated and not free soon.
- */
-int percpu_ida_for_each_free(struct percpu_ida *pool, percpu_ida_cb fn,
- void *data)
-{
- unsigned long flags;
- struct percpu_ida_cpu *remote;
- unsigned cpu, i, err = 0;
-
- local_irq_save(flags);
- for_each_possible_cpu(cpu) {
- remote = per_cpu_ptr(pool->tag_cpu, cpu);
- spin_lock(&remote->lock);
- for (i = 0; i < remote->nr_free; i++) {
- err = fn(remote->freelist[i], data);
- if (err)
- break;
- }
- spin_unlock(&remote->lock);
- if (err)
- goto out;
- }
-
- spin_lock(&pool->lock);
- for (i = 0; i < pool->nr_free; i++) {
- err = fn(pool->freelist[i], data);
- if (err)
- break;
- }
- spin_unlock(&pool->lock);
-out:
- local_irq_restore(flags);
- return err;
-}
-EXPORT_SYMBOL_GPL(percpu_ida_for_each_free);
-
-/**
- * percpu_ida_free_tags - return free tags number of a specific cpu or global pool
- * @pool: pool related
- * @cpu: specific cpu or global pool if @cpu == nr_cpu_ids
- *
- * Note: this just returns a snapshot of free tags number.
- */
-unsigned percpu_ida_free_tags(struct percpu_ida *pool, int cpu)
-{
- struct percpu_ida_cpu *remote;
- if (cpu == nr_cpu_ids)
- return pool->nr_free;
- remote = per_cpu_ptr(pool->tag_cpu, cpu);
- return remote->nr_free;
-}
-EXPORT_SYMBOL_GPL(percpu_ida_free_tags);
--
2.17.0
^ permalink raw reply related
* Re: i.MX6S/DL and QCA8334 switch using DSA driver - CPU port not working
From: Andrew Lunn @ 2018-05-15 16:08 UTC (permalink / raw)
To: Michal Vokáč; +Cc: netdev, Vivien Didelot, Florian Fainelli
In-Reply-To: <46a92995-b976-130c-7658-4d9576d5bee9@gmail.com>
On Tue, May 15, 2018 at 04:25:49PM +0200, Michal Vokáč wrote:
> On 10.5.2018 16:29, Andrew Lunn wrote:
> >I would probably add code to dump all the qca8k registers. Compare the
> >values for your working setup and your non-working setup. Hopefully
> >they are not too different, and you can quickly get to the bits which
> >matter.
>
> Perfect! Thanks to your suggestion I did that again and much more carefully.
> After some tedious comparison I think I finally found the problem.
>
> The RGMII works only if I write 0x7e to the PORT0_STATUS (0x7c) register
> from setup. Then I found out that this setup is also described in
> Qualcomm QCA8334 Q&A document. When I do that, everything work as expected.
Great.
>
> Both PORT0 and PORT6 may be configured as xGMII, xMII and SerDes and their
> functions may be exchanged. In all cases the port status register should be
> set to 0x7X where X depends on the link speed setup.
>
> Translated into register bits this means:
> - clear BIT(12) - disable MAC flow control auto-negotiation (set by default)
> - clear BIT(7) - disable MAC Tx flow control in half-duplex (set by default)
> - set BIT(6) - use full-duplex
> - set BIT(5,4) - enable Rx/Tx flow control
> - set BIT(3) - enable Rx MAC - this one is tricky, the bit is described as
> R/O in datasheet but it does not work when not set.
> - set BIT(2) - enable Tx MAC
> - set BIT(1,0) - set speed to 1000Mb
The Marvell devices have something similar. What we do there is for
the cpu port is to always set the port up, full duplex, and the
fastest speed it supports. This covers 95% of boards. We have a few
boards where the SoC on the other end can only do 100Mbps, not
1G. Then we use a fixed-phy.
> I wonder whether there are some users of this driver and what may be their
> setup that they are not affected by that?
It could be the bootloader is setting up the CPU port? I don't really
like that.
> I would like to have confirmed that I understand it correctly and that
> the problem is really in the driver not handling fixed-link.
I would actually skip fixed-link, if you don't need it. Just hardwire
the CPU port, like the Marvell driver does:
https://elixir.bootlin.com/linux/latest/source/drivers/net/dsa/mv88e6xxx/chip.c#L1780
I would do it here:
https://elixir.bootlin.com/linux/latest/source/drivers/net/dsa/qca8k.c#L518
Andrew
^ permalink raw reply
* Re: [PATCH 1/2] Convert target drivers to use sbitmap
From: Jens Axboe @ 2018-05-15 16:11 UTC (permalink / raw)
To: Matthew Wilcox, linux-kernel, linux-scsi, target-devel,
linux1394-devel, linux-usb, kvm, virtualization, netdev,
Juergen Gross, qla2xxx-upstream, Kent Overstreet
Cc: Matthew Wilcox
In-Reply-To: <20180515160043.27044-2-willy@infradead.org>
On 5/15/18 10:00 AM, Matthew Wilcox wrote:
> From: Matthew Wilcox <mawilcox@microsoft.com>
>
> The sbitmap and the percpu_ida perform essentially the same task,
> allocating tags for commands. Since the sbitmap is more used than
> the percpu_ida, convert the percpu_ida users to the sbitmap API.
It should also be the same performance as percpu_ida in light use, and
performs much better at > 50% utilization of the tag space. I think
that's better justification than "more used than".
> diff --git a/drivers/target/iscsi/iscsi_target_util.c b/drivers/target/iscsi/iscsi_target_util.c
> index 4435bf374d2d..28bcffae609f 100644
> --- a/drivers/target/iscsi/iscsi_target_util.c
> +++ b/drivers/target/iscsi/iscsi_target_util.c
> @@ -17,7 +17,7 @@
> ******************************************************************************/
>
> #include <linux/list.h>
> -#include <linux/percpu_ida.h>
> +#include <linux/sched/signal.h>
> #include <net/ipv6.h> /* ipv6_addr_equal() */
> #include <scsi/scsi_tcq.h>
> #include <scsi/iscsi_proto.h>
> @@ -147,6 +147,28 @@ void iscsit_free_r2ts_from_list(struct iscsi_cmd *cmd)
> spin_unlock_bh(&cmd->r2t_lock);
> }
>
> +int iscsit_wait_for_tag(struct se_session *se_sess, int state, int *cpup)
> +{
> + int tag = -1;
> + DEFINE_WAIT(wait);
> + struct sbq_wait_state *ws;
> +
> + if (state == TASK_RUNNING)
> + return tag;
> +
> + ws = &se_sess->sess_tag_pool.ws[0];
> + for (;;) {
> + prepare_to_wait_exclusive(&ws->wait, &wait, state);
> + if (signal_pending_state(state, current))
> + break;
> + schedule();
> + tag = sbitmap_queue_get(&se_sess->sess_tag_pool, cpup);
> + }
> +
> + finish_wait(&ws->wait, &wait);
> + return tag;
> +}
Seems like that should be:
ws = &se_sess->sess_tag_pool.ws[0];
for (;;) {
prepare_to_wait_exclusive(&ws->wait, &wait, state);
if (signal_pending_state(state, current))
break;
tag = sbitmap_queue_get(&se_sess->sess_tag_pool, cpup);
if (tag != -1)
break;
schedule();
}
finish_wait(&ws->wait, &wait);
return tag;
> /*
> * May be called from software interrupt (timer) context for allocating
> * iSCSI NopINs.
> @@ -155,9 +177,11 @@ struct iscsi_cmd *iscsit_allocate_cmd(struct iscsi_conn *conn, int state)
> {
> struct iscsi_cmd *cmd;
> struct se_session *se_sess = conn->sess->se_sess;
> - int size, tag;
> + int size, tag, cpu;
>
> - tag = percpu_ida_alloc(&se_sess->sess_tag_pool, state);
> + tag = sbitmap_queue_get(&se_sess->sess_tag_pool, &cpu);
> + if (tag < 0)
> + tag = iscsit_wait_for_tag(se_sess, state, &cpu);
> if (tag < 0)
> return NULL;
Might make sense to just roll the whole thing into iscsi_get_tag(), that
would be cleaner.
sbitmap should provide a helper for that, but we can do that cleanup
later. That would encapsulate things like the per-cpu caching hint too,
for instance.
Rest looks fine to me.
--
Jens Axboe
^ permalink raw reply
* Re: i.MX6S/DL and QCA8334 switch using DSA driver - CPU port not working
From: Florian Fainelli @ 2018-05-15 16:17 UTC (permalink / raw)
To: Andrew Lunn, Michal Vokáč; +Cc: netdev, Vivien Didelot
In-Reply-To: <20180515160824.GJ27867@lunn.ch>
On 05/15/2018 09:08 AM, Andrew Lunn wrote:
> On Tue, May 15, 2018 at 04:25:49PM +0200, Michal Vokáč wrote:
>> On 10.5.2018 16:29, Andrew Lunn wrote:
>>> I would probably add code to dump all the qca8k registers. Compare the
>>> values for your working setup and your non-working setup. Hopefully
>>> they are not too different, and you can quickly get to the bits which
>>> matter.
>>
>> Perfect! Thanks to your suggestion I did that again and much more carefully.
>> After some tedious comparison I think I finally found the problem.
>>
>> The RGMII works only if I write 0x7e to the PORT0_STATUS (0x7c) register
>> from setup. Then I found out that this setup is also described in
>> Qualcomm QCA8334 Q&A document. When I do that, everything work as expected.
>
> Great.
>
>>
>> Both PORT0 and PORT6 may be configured as xGMII, xMII and SerDes and their
>> functions may be exchanged. In all cases the port status register should be
>> set to 0x7X where X depends on the link speed setup.
>>
>> Translated into register bits this means:
>> - clear BIT(12) - disable MAC flow control auto-negotiation (set by default)
>> - clear BIT(7) - disable MAC Tx flow control in half-duplex (set by default)
>> - set BIT(6) - use full-duplex
>> - set BIT(5,4) - enable Rx/Tx flow control
>> - set BIT(3) - enable Rx MAC - this one is tricky, the bit is described as
>> R/O in datasheet but it does not work when not set.
>> - set BIT(2) - enable Tx MAC
>> - set BIT(1,0) - set speed to 1000Mb
>
> The Marvell devices have something similar. What we do there is for
> the cpu port is to always set the port up, full duplex, and the
> fastest speed it supports. This covers 95% of boards. We have a few
> boards where the SoC on the other end can only do 100Mbps, not
> 1G. Then we use a fixed-phy.
>
>> I wonder whether there are some users of this driver and what may be their
>> setup that they are not affected by that?
>
> It could be the bootloader is setting up the CPU port? I don't really
> like that.
This is unfortunately not uncommon...
>
>> I would like to have confirmed that I understand it correctly and that
>> the problem is really in the driver not handling fixed-link.
>
> I would actually skip fixed-link, if you don't need it. Just hardwire
> the CPU port, like the Marvell driver does:
>
> https://elixir.bootlin.com/linux/latest/source/drivers/net/dsa/mv88e6xxx/chip.c#L1780
>
> I would do it here:
>
> https://elixir.bootlin.com/linux/latest/source/drivers/net/dsa/qca8k.c#L518
Agreed with Andrew here, though if you can implement fixed-link, this
should be more future proof.
As far as people using this driver, John submitted it, but we have not
see many fixes/enhancements, so I am not clear who is actually using it
these days... glad you are though!
--
Florian
^ permalink raw reply
* RE: [PATCH V2 8/8] dt-bindings: stm32: add compatible for syscon
From: Christophe ROULLIER @ 2018-05-15 16:19 UTC (permalink / raw)
To: Rob Herring
Cc: mark.rutland@arm.com, andrew@lunn.ch, Alexandre TORGUE,
devicetree@vger.kernel.org, netdev@vger.kernel.org,
mcoquelin.stm32@gmail.com, Peppe CAVALLARO,
linux-arm-kernel@lists.infradead.org
In-Reply-To: <20180507163537.GA15212@rob-hp-laptop>
Hi Rob,
I do not understand, so let me explain our status:
We have syscfg IP Harware in our SOC.
But we do not have SoC specific driver to manage syscfg, we are using a generic driver "syscon".
So can you tell me what you wish to describe this part in our SOC bindings ?
Thanks for your help.
Christophe.
-----Original Message-----
From: Rob Herring [mailto:robh@kernel.org]
Sent: lundi 7 mai 2018 18:36
To: Christophe ROULLIER <christophe.roullier@st.com>
Cc: mark.rutland@arm.com; mcoquelin.stm32@gmail.com; Alexandre TORGUE <alexandre.torgue@st.com>; Peppe CAVALLARO <peppe.cavallaro@st.com>; devicetree@vger.kernel.org; andrew@lunn.ch; linux-arm-kernel@lists.infradead.org; netdev@vger.kernel.org
Subject: Re: [PATCH V2 8/8] dt-bindings: stm32: add compatible for syscon
On Wed, May 02, 2018 at 04:18:43PM +0200, Christophe Roullier wrote:
> This patch describes syscon DT bindings.
>
> Signed-off-by: Christophe Roullier <christophe.roullier@st.com>
> ---
> Documentation/devicetree/bindings/arm/stm32.txt | 4 ++++
> 1 file changed, 4 insertions(+)
>
> diff --git a/Documentation/devicetree/bindings/arm/stm32.txt
> b/Documentation/devicetree/bindings/arm/stm32.txt
> index 6808ed9..06e3834 100644
> --- a/Documentation/devicetree/bindings/arm/stm32.txt
> +++ b/Documentation/devicetree/bindings/arm/stm32.txt
> @@ -8,3 +8,7 @@ using one of the following compatible strings:
> st,stm32f746
> st,stm32h743
> st,stm32mp157
> +
> +Required nodes:
> +- syscon: the soc bus node must have a system controller node
> +pointing to the
> + global control registers, with the compatible string "syscon";
You misunderstood my prior comment. 'syscon' alone is not valid. You need SoC specific compatible string for it and 'stm32' is not SoC specific. IOW, the compatible property for a syscon should imply every single register field in the block.
Rob
^ permalink raw reply
* Re: [PATCH 1/2] Convert target drivers to use sbitmap
From: Jens Axboe @ 2018-05-15 16:20 UTC (permalink / raw)
To: Matthew Wilcox, linux-kernel, linux-scsi, target-devel,
linux1394-devel, linux-usb, kvm, virtualization, netdev,
Juergen Gross, qla2xxx-upstream, Kent Overstreet
Cc: Matthew Wilcox
In-Reply-To: <3a56027b-47bc-dcb8-a465-3670031572f1@kernel.dk>
On 5/15/18 10:11 AM, Jens Axboe wrote:
> On 5/15/18 10:00 AM, Matthew Wilcox wrote:
>> From: Matthew Wilcox <mawilcox@microsoft.com>
>>
>> The sbitmap and the percpu_ida perform essentially the same task,
>> allocating tags for commands. Since the sbitmap is more used than
>> the percpu_ida, convert the percpu_ida users to the sbitmap API.
>
> It should also be the same performance as percpu_ida in light use, and
> performs much better at > 50% utilization of the tag space. I think
> that's better justification than "more used than".
Had to search long and hard for the perf numbers I did for percpu_ida
on higher utilization, but here it is:
https://lkml.org/lkml/2014/4/22/553
--
Jens Axboe
^ permalink raw reply
* KMSAN: uninit-value in __sctp_v6_cmp_addr
From: syzbot @ 2018-05-15 16:25 UTC (permalink / raw)
To: davem, linux-kernel, linux-sctp, netdev, nhorman, syzkaller-bugs,
vyasevich
Hello,
syzbot found the following crash on:
HEAD commit: 74ee2200b89f kmsan: bump .config.example to v4.17-rc3
git tree: https://github.com/google/kmsan.git/master
console output: https://syzkaller.appspot.com/x/log.txt?x=169efb5b800000
kernel config: https://syzkaller.appspot.com/x/.config?x=4ca1e57bafa8ab1f
dashboard link: https://syzkaller.appspot.com/bug?extid=85490c30c260afff22f2
compiler: clang version 7.0.0 (trunk 329391)
syzkaller repro:https://syzkaller.appspot.com/x/repro.syz?x=157e9237800000
C reproducer: https://syzkaller.appspot.com/x/repro.c?x=10fe5de7800000
IMPORTANT: if you fix the bug, please add the following tag to the commit:
Reported-by: syzbot+85490c30c260afff22f2@syzkaller.appspotmail.com
random: sshd: uninitialized urandom read (32 bytes read)
random: sshd: uninitialized urandom read (32 bytes read)
random: sshd: uninitialized urandom read (32 bytes read)
random: sshd: uninitialized urandom read (32 bytes read)
==================================================================
BUG: KMSAN: uninit-value in __sctp_v6_cmp_addr+0x49a/0x850
net/sctp/ipv6.c:580
CPU: 0 PID: 4453 Comm: syz-executor325 Not tainted 4.17.0-rc3+ #88
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS
Google 01/01/2011
Call Trace:
<IRQ>
__dump_stack lib/dump_stack.c:77 [inline]
dump_stack+0x185/0x1d0 lib/dump_stack.c:113
kmsan_report+0x142/0x240 mm/kmsan/kmsan.c:1067
__msan_warning_32+0x6c/0xb0 mm/kmsan/kmsan_instr.c:683
__sctp_v6_cmp_addr+0x49a/0x850 net/sctp/ipv6.c:580
sctp_inet6_cmp_addr+0x3dc/0x400 net/sctp/ipv6.c:898
sctp_bind_addr_match+0x18b/0x2f0 net/sctp/bind_addr.c:330
sctp_addrs_lookup_transport+0x904/0xa20 net/sctp/input.c:942
__sctp_lookup_association net/sctp/input.c:985 [inline]
__sctp_rcv_lookup net/sctp/input.c:1249 [inline]
sctp_rcv+0x15e6/0x4d30 net/sctp/input.c:170
ip_local_deliver_finish+0x874/0xec0 net/ipv4/ip_input.c:215
NF_HOOK include/linux/netfilter.h:288 [inline]
ip_local_deliver+0x43c/0x4e0 net/ipv4/ip_input.c:256
dst_input include/net/dst.h:450 [inline]
ip_rcv_finish+0xa36/0x1d00 net/ipv4/ip_input.c:396
NF_HOOK include/linux/netfilter.h:288 [inline]
ip_rcv+0x118f/0x16d0 net/ipv4/ip_input.c:492
__netif_receive_skb_core+0x47df/0x4a90 net/core/dev.c:4592
__netif_receive_skb net/core/dev.c:4657 [inline]
process_backlog+0x62d/0xe20 net/core/dev.c:5337
napi_poll net/core/dev.c:5735 [inline]
net_rx_action+0x7c1/0x1a70 net/core/dev.c:5801
__do_softirq+0x56d/0x93d kernel/softirq.c:285
do_softirq_own_stack+0x2a/0x40 arch/x86/entry/entry_64.S:1046
</IRQ>
do_softirq kernel/softirq.c:329 [inline]
__local_bh_enable_ip+0x114/0x140 kernel/softirq.c:182
local_bh_enable+0x36/0x40 include/linux/bottom_half.h:32
rcu_read_unlock_bh include/linux/rcupdate.h:728 [inline]
ip_finish_output2+0x135a/0x1470 net/ipv4/ip_output.c:231
ip_finish_output+0xcb2/0xff0 net/ipv4/ip_output.c:317
NF_HOOK_COND include/linux/netfilter.h:277 [inline]
ip_output+0x505/0x5d0 net/ipv4/ip_output.c:405
dst_output include/net/dst.h:444 [inline]
ip_local_out net/ipv4/ip_output.c:124 [inline]
ip_queue_xmit+0x1a1e/0x1d10 net/ipv4/ip_output.c:504
sctp_v4_xmit+0x188/0x210 net/sctp/protocol.c:983
sctp_packet_transmit+0x3eaa/0x4350 net/sctp/output.c:650
sctp_outq_flush+0x1a7a/0x6320 net/sctp/outqueue.c:1197
sctp_outq_uncork+0xd2/0xf0 net/sctp/outqueue.c:776
sctp_cmd_interpreter net/sctp/sm_sideeffect.c:1820 [inline]
sctp_side_effects net/sctp/sm_sideeffect.c:1220 [inline]
sctp_do_sm+0x8707/0x8d20 net/sctp/sm_sideeffect.c:1191
sctp_primitive_REQUESTHEARTBEAT+0x175/0x1a0 net/sctp/primitive.c:200
sctp_apply_peer_addr_params+0x207/0x1670 net/sctp/socket.c:2487
sctp_setsockopt_peer_addr_params net/sctp/socket.c:2683 [inline]
sctp_setsockopt+0x10e5f/0x11600 net/sctp/socket.c:4258
sock_common_setsockopt+0x136/0x170 net/core/sock.c:3039
__sys_setsockopt+0x4af/0x560 net/socket.c:1903
__do_sys_setsockopt net/socket.c:1914 [inline]
__se_sys_setsockopt net/socket.c:1911 [inline]
__x64_sys_setsockopt+0x15c/0x1c0 net/socket.c:1911
do_syscall_64+0x154/0x220 arch/x86/entry/common.c:287
entry_SYSCALL_64_after_hwframe+0x44/0xa9
RIP: 0033:0x43fef9
RSP: 002b:00007ffc00d9bfd8 EFLAGS: 00000207 ORIG_RAX: 0000000000000036
RAX: ffffffffffffffda RBX: 00000000004002c8 RCX: 000000000043fef9
RDX: 0000000000000009 RSI: 0000000000000084 RDI: 0000000000000003
RBP: 00000000006ca018 R08: 0000000000000098 R09: 000000000000001c
R10: 0000000020000180 R11: 0000000000000207 R12: 0000000000401820
R13: 00000000004018b0 R14: 0000000000000000 R15: 0000000000000000
Local variable description: ----dest@sctp_rcv
Variable was created at:
sctp_rcv+0x13d/0x4d30 net/sctp/input.c:97
ip_local_deliver_finish+0x874/0xec0 net/ipv4/ip_input.c:215
==================================================================
---
This bug is generated by a bot. It may contain errors.
See https://goo.gl/tpsmEJ for more information about syzbot.
syzbot engineers can be reached at syzkaller@googlegroups.com.
syzbot will keep track of this bug report. See:
https://goo.gl/tpsmEJ#bug-status-tracking for how to communicate with
syzbot.
syzbot can test patches for this bug, for details see:
https://goo.gl/tpsmEJ#testing-patches
^ permalink raw reply
* Re: [RESEND PATCH v2 1/1] net: phy: micrel: add 125MHz reference clock workaround
From: Andrew Lunn @ 2018-05-15 16:33 UTC (permalink / raw)
To: Marco Felsch
Cc: robh+dt, mark.rutland, f.fainelli, netdev, devicetree, kernel,
niebelm
In-Reply-To: <20180515081856.23322-2-m.felsch@pengutronix.de>
On Tue, May 15, 2018 at 10:18:56AM +0200, Marco Felsch wrote:
> From: Markus Niebel <Markus.Niebel@tqs.de>
>
> The micrel KSZ9031 phy has a optional clock pin (CLK125_NDO) which can be
> used as reference clock for the MAC unit. The clock signal must meet the
> RGMII requirements to ensure the correct data transmission between the
> MAC and the PHY. The KSZ9031 phy does not fulfill the duty cycle
> requirement if the phy is configured as slave. For a complete
> describtion look at the errata sheets: DS80000691D or DS80000692D.
>
> The errata sheet recommends to force the phy into master mode whenever
> there is a 1000Base-T link-up as work around. Only set the
> "micrel,force-master" property if you use the phy reference clock provided
> by CLK125_NDO pin as MAC reference clock in your application.
>
> Attenation, this workaround is only usable if the link partner can
> be configured to slave mode for 1000Base-T.
Thanks for adding all the documentation.
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Andrew
^ permalink raw reply
* Re: [PATCH v1 iproute2-next 2/3] rdma: print driver resource attributes
From: Doug Ledford @ 2018-05-15 16:35 UTC (permalink / raw)
To: Steve Wise, Leon Romanovsky; +Cc: dsahern, stephen, netdev, linux-rdma
In-Reply-To: <c57b8f18-858c-62df-50c6-453933a67913@opengridcomputing.com>
[-- Attachment #1: Type: text/plain, Size: 998 bytes --]
On Mon, 2018-05-14 at 09:51 -0500, Steve Wise wrote:
>
> On 5/13/2018 8:24 AM, Leon Romanovsky wrote:
> > On Mon, May 07, 2018 at 08:53:16AM -0700, Steve Wise wrote:
> > > This enhancement allows printing rdma device-specific state, if provided
> > > by the kernel. This is done in a generic manner, so rdma tool doesn't
> >
> > Double space between "." and "This".
> >
> > > need to know about the details of every type of rdma device.
> > >
> > > Driver attributes for a rdma resource are in the form of <key,
> > > [print_type], value> tuples, where the key is a string and the value can
> > > be any supported driver attribute. The print_type attribute, if present,
> >
> > ditto
>
> I'll fix these.
Fix it if you want, but don't do it because Leon told you to. A double
space after period is perfectly acceptable.
--
Doug Ledford <dledford@redhat.com>
GPG KeyID: B826A3330E572FDD
Key fingerprint = AE6B 1BDA 122B 23B4 265B 1274 B826 A333 0E57 2FDD
[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 833 bytes --]
^ permalink raw reply
* Re: [PATCH net] tcp: purge write queue in tcp_connect_init()
From: Neal Cardwell @ 2018-05-15 16:42 UTC (permalink / raw)
To: Eric Dumazet; +Cc: David Miller, Netdev, Eric Dumazet, Yuchung Cheng
In-Reply-To: <20180515041426.94062-1-edumazet@google.com>
On Tue, May 15, 2018 at 12:14 AM Eric Dumazet <edumazet@google.com> wrote:
> syzkaller found a reliable way to crash the host, hitting a BUG()
> in __tcp_retransmit_skb()
> Malicous MSG_FASTOPEN is the root cause. We need to purge write queue
> in tcp_connect_init() at the point we init snd_una/write_seq.
> This patch also replaces the BUG() by a less intrusive WARN_ON_ONCE()
> kernel BUG at net/ipv4/tcp_output.c:2837!
...
> Fixes: cf60af03ca4e ("net-tcp: Fast Open client - sendmsg(MSG_FASTOPEN)")
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> Cc: Yuchung Cheng <ycheng@google.com>
> Cc: Neal Cardwell <ncardwell@google.com>
> Reported-by: syzbot <syzkaller@googlegroups.com>
> ---
Acked-by: Neal Cardwell <ncardwell@google.com>
Thanks, Eric!
neal
^ permalink raw reply
* Re: [RFC bpf-next 07/11] bpf: Add helper to retrieve socket in BPF
From: Martin KaFai Lau @ 2018-05-15 16:48 UTC (permalink / raw)
To: Alexei Starovoitov, Joe Stringer; +Cc: daniel, netdev, ast, john fastabend
In-Reply-To: <20180515031657.4wza3jsatn2ggsj5@ast-mbp>
On Mon, May 14, 2018 at 08:16:59PM -0700, Alexei Starovoitov wrote:
> On Fri, May 11, 2018 at 05:54:33PM -0700, Joe Stringer wrote:
> > On 11 May 2018 at 14:41, Martin KaFai Lau <kafai@fb.com> wrote:
> > > On Fri, May 11, 2018 at 02:08:01PM -0700, Joe Stringer wrote:
> > >> On 10 May 2018 at 22:00, Martin KaFai Lau <kafai@fb.com> wrote:
> > >> > On Wed, May 09, 2018 at 02:07:05PM -0700, Joe Stringer wrote:
> > >> >> This patch adds a new BPF helper function, sk_lookup() which allows BPF
> > >> >> programs to find out if there is a socket listening on this host, and
> > >> >> returns a socket pointer which the BPF program can then access to
> > >> >> determine, for instance, whether to forward or drop traffic. sk_lookup()
> > >> >> takes a reference on the socket, so when a BPF program makes use of this
> > >> >> function, it must subsequently pass the returned pointer into the newly
> > >> >> added sk_release() to return the reference.
> > >> >>
> > >> >> By way of example, the following pseudocode would filter inbound
> > >> >> connections at XDP if there is no corresponding service listening for
> > >> >> the traffic:
> > >> >>
> > >> >> struct bpf_sock_tuple tuple;
> > >> >> struct bpf_sock_ops *sk;
> > >> >>
> > >> >> populate_tuple(ctx, &tuple); // Extract the 5tuple from the packet
> > >> >> sk = bpf_sk_lookup(ctx, &tuple, sizeof tuple, netns, 0);
> > >> >> if (!sk) {
> > >> >> // Couldn't find a socket listening for this traffic. Drop.
> > >> >> return TC_ACT_SHOT;
> > >> >> }
> > >> >> bpf_sk_release(sk, 0);
> > >> >> return TC_ACT_OK;
> > >> >>
> > >> >> Signed-off-by: Joe Stringer <joe@wand.net.nz>
> > >> >> ---
> > >>
> > >> ...
> > >>
> > >> >> @@ -4032,6 +4036,96 @@ static const struct bpf_func_proto bpf_skb_get_xfrm_state_proto = {
> > >> >> };
> > >> >> #endif
> > >> >>
> > >> >> +struct sock *
> > >> >> +sk_lookup(struct net *net, struct bpf_sock_tuple *tuple) {
> > >> > Would it be possible to have another version that
> > >> > returns a sk without taking its refcnt?
> > >> > It may have performance benefit.
> > >>
> > >> Not really. The sockets are not RCU-protected, and established sockets
> > >> may be torn down without notice. If we don't take a reference, there's
> > >> no guarantee that the socket will continue to exist for the duration
> > >> of running the BPF program.
> > >>
> > >> From what I follow, the comment below has a hidden implication which
> > >> is that sockets without SOCK_RCU_FREE, eg established sockets, may be
> > >> directly freed regardless of RCU.
> > > Right, SOCK_RCU_FREE sk is the one I am concern about.
> > > For example, TCP_LISTEN socket does not require taking a refcnt
> > > now. Doing a bpf_sk_lookup() may have a rather big
> > > impact on handling TCP syn flood. or the usual intention
> > > is to redirect instead of passing it up to the stack?
> >
> > I see, if you're only interested in listen sockets then probably this
> > series could be extended with a new flag, eg something like
> > BPF_F_SK_FIND_LISTENERS which restricts the set of possible sockets
> > found to only listen sockets, then the implementation would call into
> > __inet_lookup_listener() instead of inet_lookup(). The presence of
> > that flag in the relevant register during CALL instruction would show
> > that the verifier should not reference-track the result, then there'd
> > need to be a check on the release to ensure that this unreferenced
> > socket is never released. Just a thought, completely untested and I
> > could still be missing some detail..
> >
> > That said, I don't completely follow how you would expect to handle
> > the traffic for sockets that are already established - the helper
> > would no longer find those sockets, so you wouldn't know whether to
> > pass the traffic up the stack for established traffic or not.
>
> I think Martin has a valid concern here that if somebody starts using
> this helper on the rx traffic the bpf program (via these two new
> helpers) will be doing refcnt++ and refcnt-- even for listener
> sockets which will cause synflood to suffer.
> One can argue that this is bpf author mistake, but without fixes
> (and api changes) to the helper the programmer doesn't really have a way
> of avoiding this situation.
> Also udp sockets don't need refcnt at all.
> How about we split this single helper into three:
> - bpf_sk_lookup_tcp_established() that will return refcnt-ed socket
> and has to be bpf_sk_release()d by the program.
> - bpf_sk_lookup_tcp_listener() that doesn't refcnt, since progs
> run in rcu.
> - bpf_sk_lookup_udp() that also doesn't refcnt.
> The logic you want to put into this helper can be easily
> replicated with these three helpers and the whole thing will
> be much faster.
> Thoughts?
Just came to my mind.
or can we explore something like:
On the bpf_sk_lookup() side, use __inet[6]_lookup()
and __udp[46]_lib_lookup() instead. That should
only take refcnt if it has to.
On the bpf_sk_release() side, it skips refcnt--
if sk is SOCK_RCU_FREE.
^ permalink raw reply
* Re: [PATCH v1 iproute2-next 2/3] rdma: print driver resource attributes
From: Leon Romanovsky @ 2018-05-15 16:59 UTC (permalink / raw)
To: Doug Ledford; +Cc: Steve Wise, dsahern, stephen, netdev, linux-rdma
In-Reply-To: <13da78012c55e8f224c2226f3c0cfc53bd0522be.camel@redhat.com>
[-- Attachment #1: Type: text/plain, Size: 1431 bytes --]
On Tue, May 15, 2018 at 12:35:34PM -0400, Doug Ledford wrote:
> On Mon, 2018-05-14 at 09:51 -0500, Steve Wise wrote:
> >
> > On 5/13/2018 8:24 AM, Leon Romanovsky wrote:
> > > On Mon, May 07, 2018 at 08:53:16AM -0700, Steve Wise wrote:
> > > > This enhancement allows printing rdma device-specific state, if provided
> > > > by the kernel. This is done in a generic manner, so rdma tool doesn't
> > >
> > > Double space between "." and "This".
> > >
> > > > need to know about the details of every type of rdma device.
> > > >
> > > > Driver attributes for a rdma resource are in the form of <key,
> > > > [print_type], value> tuples, where the key is a string and the value can
> > > > be any supported driver attribute. The print_type attribute, if present,
> > >
> > > ditto
> >
> > I'll fix these.
>
> Fix it if you want, but don't do it because Leon told you to. A double
> space after period is perfectly acceptable.
It is very controversial thing [1],
"Most style guides indicate that single sentence spacing is proper for
final or published work today, and most publishers require manuscripts
to be submitted as they will appear in publication—single
sentence spaced."
[1] https://en.wikipedia.org/wiki/Sentence_spacing
>
> --
> Doug Ledford <dledford@redhat.com>
> GPG KeyID: B826A3330E572FDD
> Key fingerprint = AE6B 1BDA 122B 23B4 265B 1274 B826 A333 0E57 2FDD
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 801 bytes --]
^ permalink raw reply
* Re: [PATCH bpf-next v2 0/5] samples: bpf: fix build after move to full libbpf
From: Jakub Kicinski @ 2018-05-15 17:46 UTC (permalink / raw)
To: Jesper Dangaard Brouer
Cc: Alexei Starovoitov, daniel, oss-drivers, netdev,
Björn Töpel, Y Song
In-Reply-To: <20180515111643.1507fead@redhat.com>
On Tue, 15 May 2018 11:16:43 +0200, Jesper Dangaard Brouer wrote:
> > $ cd tools/lib/bpf/
> > $ make
> >
> > Auto-detecting system features:
> > ... libelf: [ on ]
> > ... bpf: [ on ]
> >
> > Warning: Kernel ABI header at 'tools/include/uapi/linux/bpf.h' differs from latest version at 'include/uapi/linux/bpf.h'
> > Warning: Kernel ABI header at 'tools/include/uapi/linux/if_link.h' differs from latest version at 'include/uapi/linux/if_link.h'
> > CC libbpf.o
> > CC bpf.o
> > CC nlattr.o
> > CC btf.o
> > LD libbpf-in.o
> > LINK libbpf.a
> > LINK libbpf.so
>
> SOLVED
>
> It seems that the "Auto-detecting system" needed a 'make clean'.
> My problem goes away when I did the following:
>
> $ cd tools/
> $ make clean
>
> $ cd lib/bpf/
> $ make
>
> Auto-detecting system features:
> ... libelf: [ on ]
> ... bpf: [ on ]
Interesting! I've seen the same thing (on Fedora, not sure distro is
relevant). I tracked it down to dependency files (*.d) in
tools/build/features/ containing a reference to a sigset.h header which
didn't exist :S I did git clean -xdf to fix my tree.
Thanks for testing and sorry about the breakage!
^ permalink raw reply
* RE: [PATCH v1 iproute2-next 2/3] rdma: print driver resource attributes
From: Steve Wise @ 2018-05-15 17:51 UTC (permalink / raw)
To: 'Leon Romanovsky', 'Doug Ledford'
Cc: dsahern, stephen, netdev, linux-rdma
In-Reply-To: <20180515165955.GD10381@mtr-leonro.mtl.com>
>
> On Tue, May 15, 2018 at 12:35:34PM -0400, Doug Ledford wrote:
> > On Mon, 2018-05-14 at 09:51 -0500, Steve Wise wrote:
> > >
> > > On 5/13/2018 8:24 AM, Leon Romanovsky wrote:
> > > > On Mon, May 07, 2018 at 08:53:16AM -0700, Steve Wise wrote:
> > > > > This enhancement allows printing rdma device-specific state, if
> provided
> > > > > by the kernel. This is done in a generic manner, so rdma tool doesn't
> > > >
> > > > Double space between "." and "This".
> > > >
> > > > > need to know about the details of every type of rdma device.
> > > > >
> > > > > Driver attributes for a rdma resource are in the form of <key,
> > > > > [print_type], value> tuples, where the key is a string and the value can
> > > > > be any supported driver attribute. The print_type attribute, if
> present,
> > > >
> > > > ditto
> > >
> > > I'll fix these.
> >
> > Fix it if you want, but don't do it because Leon told you to. A double
> > space after period is perfectly acceptable.
>
> It is very controversial thing [1],
>
> "Most style guides indicate that single sentence spacing is proper for
> final or published work today, and most publishers require manuscripts
> to be submitted as they will appear in publication—single
> sentence spaced."
>
> [1] https://en.wikipedia.org/wiki/Sentence_spacing
We're not writing a manuscript. 😉 Regardless, I made the changes and they are in v2 of the patch series, which I think is probably ready to merge.
Steve.
^ permalink raw reply
* SCM_RIGHTS and file descriptor limits
From: Andres Freund @ 2018-05-15 17:52 UTC (permalink / raw)
To: David S. Miller, mtk.manpages, Kees Cook, David Windsor,
Hans Liljestrand, Reshetova, Elena, Al Viro
Cc: linux-man, netdev, linux-kernel
Hi,
I'm not sure if this is a documentation omission, kernel code bug,
missing understanding on my part, or all of the above.
I'm looking at recvmsg()'s behaviour for AF_UNIX, where the sender has
sent an fd using SCM_RIGHTS, and the receiving process has already
exceeded RLIMIT_NOFILE.
By my reading of scm_detach_fds() (called from
unix_stream_read_generic() / unix_dgram_recvmsg(), via scm_recv()), the
error appears to just be thrown away, with no notification given to
userland at all:
void scm_detach_fds(struct msghdr *msg, struct scm_cookie *scm)
{
...
for (i=0, cmfptr=(__force int __user *)CMSG_DATA(cm); i<fdmax;
i++, cmfptr++)
{
struct socket *sock;
int new_fd;
err = security_file_receive(fp[i]);
if (err)
break;
err = get_unused_fd_flags(MSG_CMSG_CLOEXEC & msg->msg_flags
? O_CLOEXEC : 0);
if (err < 0)
break;
If I understand correctly no error is reported to the user (function
returns void, no put_user() calls for the unsuccessful fds/i), but all
the scm data is "consumed" via scm_destroy().
Silently consuming data without reporting an error doesn't strike me as
nice. Now admittedly scm is one of the more kludgey interfaces, and it's
not entirely obvious how one would do much better without significant
added complication - the correct behaviour seems to be to return EMFILE,
but that'd need to somehow rewind the received data, which seems
complicated... On the other hand it's not generally knowable how many
file descriptors the other side sent, so one really needs to duplicate
all scm information on the data level to make sure nothing gets lost.
In case this is something that won't be fixed, it seems appropriate to
add a section to the man pages (hence CCing linux man-pages list)
Greetings,
Andres Freund
^ permalink raw reply
* Re: [PATCH v1 iproute2-next 2/3] rdma: print driver resource attributes
From: Doug Ledford @ 2018-05-15 18:00 UTC (permalink / raw)
To: Leon Romanovsky; +Cc: Steve Wise, dsahern, stephen, netdev, linux-rdma
In-Reply-To: <20180515165955.GD10381@mtr-leonro.mtl.com>
[-- Attachment #1: Type: text/plain, Size: 2380 bytes --]
On Tue, 2018-05-15 at 19:59 +0300, Leon Romanovsky wrote:
> On Tue, May 15, 2018 at 12:35:34PM -0400, Doug Ledford wrote:
> > On Mon, 2018-05-14 at 09:51 -0500, Steve Wise wrote:
> > >
> > > On 5/13/2018 8:24 AM, Leon Romanovsky wrote:
> > > > On Mon, May 07, 2018 at 08:53:16AM -0700, Steve Wise wrote:
> > > > > This enhancement allows printing rdma device-specific state, if provided
> > > > > by the kernel. This is done in a generic manner, so rdma tool doesn't
> > > >
> > > > Double space between "." and "This".
> > > >
> > > > > need to know about the details of every type of rdma device.
> > > > >
> > > > > Driver attributes for a rdma resource are in the form of <key,
> > > > > [print_type], value> tuples, where the key is a string and the value can
> > > > > be any supported driver attribute. The print_type attribute, if present,
> > > >
> > > > ditto
> > >
> > > I'll fix these.
> >
> > Fix it if you want, but don't do it because Leon told you to. A double
> > space after period is perfectly acceptable.
>
> It is very controversial thing [1],
>
> "Most style guides indicate that single sentence spacing is proper for
> final or published work today, and most publishers require manuscripts
> to be submitted as they will appear in publication—single
> sentence spaced."
>
> [1] https://en.wikipedia.org/wiki/Sentence_spacing
Yes...and the justification is that proportional font systems resolve
the issue of delineating sentences without the need for a second space
(they actually don't because they are improperly implemented, but that's
another issue). But we don't work on proportional font systems, we work
with git, on command lines, with fixed spacing fonts, all the things
that indicate double spaces are actually preferred even by those people
that are proponents of single spacing. And on top of that, current
research suggests maybe the single spacers are wrong after all :-o
https://www.theatlantic.com/science/archive/2018/05/two-spaces-after-a-p
eriod/559304/
Anyway, my original point stands. It's controversial, as you said, so
Steve should do what Steve feels is best and not worry about anyone
else' opinion on the matter.
--
Doug Ledford <dledford@redhat.com>
GPG KeyID: B826A3330E572FDD
Key fingerprint = AE6B 1BDA 122B 23B4 265B 1274 B826 A333 0E57 2FDD
[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 833 bytes --]
^ permalink raw reply
* Re: [PATCH net-next 2/3] net: qualcomm: rmnet: Add support for ethtool private stats
From: Subash Abhinov Kasiviswanathan @ 2018-05-15 18:09 UTC (permalink / raw)
To: kbuild test robot; +Cc: kbuild-all, davem, netdev
In-Reply-To: <201805151651.Vgq47bPJ%fengguang.wu@intel.com>
On 2018-05-15 02:41, kbuild test robot wrote:
> Hi Subash,
>
> Thank you for the patch! Perhaps something to improve:
>
> [auto build test WARNING on net-next/master]
>
> url:
> https://github.com/0day-ci/linux/commits/Subash-Abhinov-Kasiviswanathan/net-qualcomm-rmnet-Updates-2018-05-14/20180515-115355
> reproduce:
> # apt-get install sparse
> make ARCH=x86_64 allmodconfig
> make C=1 CF=-D__CHECK_ENDIAN__
>
>
> sparse warnings: (new ones prefixed by >>)
>
>>> drivers/net/ethernet/qualcomm/rmnet/rmnet_vnd.c:199:26: sparse:
>>> symbol 'rmnet_ethtool_ops' was not declared. Should it be static?
>
> Please review and possibly fold the followup patch.
>
> ---
> 0-DAY kernel test infrastructure Open Source Technology
> Center
> https://lists.01.org/pipermail/kbuild-all Intel
> Corporation
Thanks Fengguang. I will fix both issues in v2
--
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project
^ permalink raw reply
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox