Since the bulk queue used by XDP_REDIRECT now lives in struct net_device,
we can re-use the bulking for the non-map version of the bpf_redirect()
helper. This is a simple matter of having xdp_do_redirect_slow() queue the
frame on the bulk queue instead of sending it out with __bpf_tx_xdp().
Unfortunately we can't make the bpf_redirect() helper return an error if
the ifindex doesn't exit (as bpf_redirect_map() does), because we don't
have a reference to the network namespace of the ingress device at the time
the helper is called. So we have to leave it as-is and keep the device
lookup in xdp_do_redirect_slow().
Since this leaves less reason to have the non-map redirect code in a
separate function, so we get rid of the xdp_do_redirect_slow() function
entirely. This does lose us the tracepoint disambiguation, but fortunately
the xdp_redirect and xdp_redirect_map tracepoints use the same tracepoint
entry structures. This means both can contain a map index, so we can just
amend the tracepoint definitions so we always emit the xdp_redirect(_err)
tracepoints, but with the map ID only populated if a map is present. This
means we retire the xdp_redirect_map(_err) tracepoints entirely, but keep
the definitions around in case someone is still listening for them.
With this change, the performance of the xdp_redirect sample program goes
from 5Mpps to 8.4Mpps (a 68% increase).
Since the flush functions are no longer map-specific, rename the flush()
functions to drop _map from their names. One of the renamed functions is
the xdp_do_flush_map() callback used in all the xdp-enabled drivers. To
keep from having to update all drivers, use a #define to keep the old name
working, and only update the virtual drivers in this patch.
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/157918768505.1458396.17518057312953572912.stgit@toke.dk
Commit 96360004b8 ("xdp: Make devmap flush_list common for all map
instances"), changed devmap flushing to be a global operation instead of a
per-map operation. However, the queue structure used for bulking was still
allocated as part of the containing map.
This patch moves the devmap bulk queue into struct net_device. The
motivation for this is reusing it for the non-map variant of XDP_REDIRECT,
which will be changed in a subsequent commit. To avoid other fields of
struct net_device moving to different cache lines, we also move a couple of
other members around.
We defer the actual allocation of the bulk queue structure until the
NETDEV_REGISTER notification devmap.c. This makes it possible to check for
ndo_xdp_xmit support before allocating the structure, which is not possible
at the time struct net_device is allocated. However, we keep the freeing in
free_netdev() to avoid adding another RCU callback on NETDEV_UNREGISTER.
Because of this change, we lose the reference back to the map that
originated the redirect, so change the tracepoint to always return 0 as the
map ID and index. Otherwise no functional change is intended with this
patch.
After this patch, the relevant part of struct net_device looks like this,
according to pahole:
/* --- cacheline 14 boundary (896 bytes) --- */
struct netdev_queue * _tx __attribute__((__aligned__(64))); /* 896 8 */
unsigned int num_tx_queues; /* 904 4 */
unsigned int real_num_tx_queues; /* 908 4 */
struct Qdisc * qdisc; /* 912 8 */
unsigned int tx_queue_len; /* 920 4 */
spinlock_t tx_global_lock; /* 924 4 */
struct xdp_dev_bulk_queue * xdp_bulkq; /* 928 8 */
struct xps_dev_maps * xps_cpus_map; /* 936 8 */
struct xps_dev_maps * xps_rxqs_map; /* 944 8 */
struct mini_Qdisc * miniq_egress; /* 952 8 */
/* --- cacheline 15 boundary (960 bytes) --- */
struct hlist_head qdisc_hash[16]; /* 960 128 */
/* --- cacheline 17 boundary (1088 bytes) --- */
struct timer_list watchdog_timer; /* 1088 40 */
/* XXX last struct has 4 bytes of padding */
int watchdog_timeo; /* 1128 4 */
/* XXX 4 bytes hole, try to pack */
struct list_head todo_list; /* 1136 16 */
/* --- cacheline 18 boundary (1152 bytes) --- */
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/157918768397.1458396.12673224324627072349.stgit@toke.dk
When registering a umem area that is sufficiently large (>1G on an
x86), kmalloc cannot be used to allocate one of the internal data
structures, as the size requested gets too large. Use kvmalloc instead
that falls back on vmalloc if the allocation is too large for kmalloc.
Also add accounting for this structure as it is triggered by a user
space action (the XDP_UMEM_REG setsockopt) and it is by far the
largest structure of kernel allocated memory in xsk.
Reported-by: Ryan Goodfellow <rgoodfel@isi.edu>
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Link: https://lore.kernel.org/bpf/1578995365-7050-1-git-send-email-magnus.karlsson@intel.com
A negative value should be returned if map->map_type is invalid
although that is impossible now, but if we run into such situation
in future, then xdpbuff could be leaked.
Daniel Borkmann suggested:
-EBADRQC should be returned to stay consistent with generic XDP
for the tracepoint output and not to be confused with -EOPNOTSUPP
from other locations like dev_map_enqueue() when ndo_xdp_xmit is
missing and such.
Suggested-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Li RongQing <lirongqing@baidu.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/1578618277-18085-1-git-send-email-lirongqing@baidu.com
Add a helper to send out a tcp-ack. It will be used in the later
bpf_dctcp implementation that requires to send out an ack
when the CE state changed.
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Yonghong Song <yhs@fb.com>
Link: https://lore.kernel.org/bpf/20200109004551.3900448-1-kafai@fb.com
This patch makes "struct tcp_congestion_ops" to be the first user
of BPF STRUCT_OPS. It allows implementing a tcp_congestion_ops
in bpf.
The BPF implemented tcp_congestion_ops can be used like
regular kernel tcp-cc through sysctl and setsockopt. e.g.
[root@arch-fb-vm1 bpf]# sysctl -a | egrep congestion
net.ipv4.tcp_allowed_congestion_control = reno cubic bpf_cubic
net.ipv4.tcp_available_congestion_control = reno bic cubic bpf_cubic
net.ipv4.tcp_congestion_control = bpf_cubic
There has been attempt to move the TCP CC to the user space
(e.g. CCP in TCP). The common arguments are faster turn around,
get away from long-tail kernel versions in production...etc,
which are legit points.
BPF has been the continuous effort to join both kernel and
userspace upsides together (e.g. XDP to gain the performance
advantage without bypassing the kernel). The recent BPF
advancements (in particular BTF-aware verifier, BPF trampoline,
BPF CO-RE...) made implementing kernel struct ops (e.g. tcp cc)
possible in BPF. It allows a faster turnaround for testing algorithm
in the production while leveraging the existing (and continue growing)
BPF feature/framework instead of building one specifically for
userspace TCP CC.
This patch allows write access to a few fields in tcp-sock
(in bpf_tcp_ca_btf_struct_access()).
The optional "get_info" is unsupported now. It can be added
later. One possible way is to output the info with a btf-id
to describe the content.
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Acked-by: Yonghong Song <yhs@fb.com>
Link: https://lore.kernel.org/bpf/20200109003508.3856115-1-kafai@fb.com
The DSA drivers that implement .phylink_mac_link_state should normally
register an interrupt for the PCS, from which they should call
phylink_mac_change(). However not all switches implement this, and those
who don't should set this flag in dsa_switch in the .setup callback, so
that PHYLINK will poll for a few ms until the in-band AN link timer
expires and the PCS state settles.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is a cosmetic patch that makes the dp, tx_vid, queue_mapping and
pcp local variable definitions a bit closer in length, so they don't
look like an eyesore as much.
The 'ds' variable is not used otherwise, except for ds->dp.
Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There are 3 things that are wrong with the DSA deferred xmit mechanism:
1. Its introduction has made the DSA hotpath ever so slightly more
inefficient for everybody, since DSA_SKB_CB(skb)->deferred_xmit needs
to be initialized to false for every transmitted frame, in order to
figure out whether the driver requested deferral or not (a very rare
occasion, rare even for the only driver that does use this mechanism:
sja1105). That was necessary to avoid kfree_skb from freeing the skb.
2. Because L2 PTP is a link-local protocol like STP, it requires
management routes and deferred xmit with this switch. But as opposed
to STP, the deferred work mechanism needs to schedule the packet
rather quickly for the TX timstamp to be collected in time and sent
to user space. But there is no provision for controlling the
scheduling priority of this deferred xmit workqueue. Too bad this is
a rather specific requirement for a feature that nobody else uses
(more below).
3. Perhaps most importantly, it makes the DSA core adhere a bit too
much to the NXP company-wide policy "Innovate Where It Doesn't
Matter". The sja1105 is probably the only DSA switch that requires
some frames sent from the CPU to be routed to the slave port via an
out-of-band configuration (register write) rather than in-band (DSA
tag). And there are indeed very good reasons to not want to do that:
if that out-of-band register is at the other end of a slow bus such
as SPI, then you limit that Ethernet flow's throughput to effectively
the throughput of the SPI bus. So hardware vendors should definitely
not be encouraged to design this way. We do _not_ want more
widespread use of this mechanism.
Luckily we have a solution for each of the 3 issues:
For 1, we can just remove that variable in the skb->cb and counteract
the effect of kfree_skb with skb_get, much to the same effect. The
advantage, of course, being that anybody who doesn't use deferred xmit
doesn't need to do any extra operation in the hotpath.
For 2, we can create a kernel thread for each port's deferred xmit work.
If the user switch ports are named swp0, swp1, swp2, the kernel threads
will be named swp0_xmit, swp1_xmit, swp2_xmit (there appears to be a 15
character length limit on kernel thread names). With this, the user can
change the scheduling priority with chrt $(pidof swp2_xmit).
For 3, we can actually move the entire implementation to the sja1105
driver.
So this patch deletes the generic implementation from the DSA core and
adds a new one, more adequate to the requirements of PTP TX
timestamping, in sja1105_main.c.
Suggested-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Passing NULL to l2tp_pernet causes a crash via BUG_ON.
Dereferencing net in net_generic() also has the same effect.
This patch removes the redundant BUG_ON check on the same parameter.
Signed-off-by: Xu Wang <vulab@iscas.ac.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
Passing NULL to phonet_pernet causes a crash via BUG_ON.
Dereferencing net in net_generic() also has the same effect.
This patch removes the redundant BUG_ON check on the same parameter.
Signed-off-by: Xu Wang <vulab@iscas.ac.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
The argument is always ignored, so remove it.
Signed-off-by: Li RongQing <lirongqing@baidu.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fixes gcc '-Wunused-but-set-variable' warning:
net/ethtool/linkmodes.c: In function 'ethnl_set_linkmodes':
net/ethtool/linkmodes.c:326:32: warning:
variable 'lsettings' set but not used [-Wunused-but-set-variable]
struct ethtool_link_settings *lsettings;
^
It is never used, so remove it.
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Reviewed-by: Michal Kubecek <mkubecek@suse.cz>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
REXMIT_NEW is a macro for "FRTO-style
transmit of unsent/new packets", this patch
makes it more readable.
Signed-off-by: Mao Wenan <maowenan@huawei.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add support for userspace to specify a device index to limit the scope
of an entry via the TCP_MD5SIG_EXT setsockopt. The existing __tcpm_pad
is renamed to tcpm_ifindex and the new field is only checked if the new
TCP_MD5SIG_FLAG_IFINDEX is set in tcpm_flags. For now, the device index
must point to an L3 master device (e.g., VRF). The API and error
handling are setup to allow the constraint to be relaxed in the future
to any device index.
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add l3index to tcp_md5sig_key to represent the L3 domain of a key, and
add l3index to tcp_md5_do_add and tcp_md5_do_del to fill in the key.
With the key now based on an l3index, add the new parameter to the
lookup functions and consider the l3index when looking for a match.
The l3index comes from the skb when processing ingress packets leveraging
the helpers created for socket lookups, tcp_v4_sdif and inet_iif (and the
v6 variants). When the sdif index is set it means the packet ingressed a
device that is part of an L3 domain and inet_iif points to the VRF device.
For egress, the L3 domain is determined from the socket binding and
sk_bound_dev_if.
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The original ingress device index is saved to the cb space of the skb
and the cb is moved during tcp processing. Since tcp_v4_inbound_md5_hash
can be called before and after the cb move, pass dif and sdif to it so
the caller can save both prior to the cb move. Both are used by a later
patch.
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The original ingress device index is saved to the cb space of the skb
and the cb is moved during tcp processing. Since tcp_v6_inbound_md5_hash
can be called before and after the cb move, pass dif and sdif to it so
the caller can save both prior to the cb move. Both are used by a later
patch.
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Extract the typecast to (union tcp_md5_addr *) to a local variable
rather than the current long, inline declaration with function calls.
No functional change intended.
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When kernel is compiled without NUMA support, then page_pool NUMA
config setting (pool->p.nid) doesn't make any practical sense. The
compiler cannot see that it can remove the code paths.
This patch avoids reading pool->p.nid setting in case of !CONFIG_NUMA,
in allocation and numa check code, which helps compiler to see the
optimisation potential. It leaves update code intact to keep API the
same.
$ ./scripts/bloat-o-meter net/core/page_pool.o-numa-enabled \
net/core/page_pool.o-numa-disabled
add/remove: 0/0 grow/shrink: 0/3 up/down: 0/-113 (-113)
Function old new delta
page_pool_create 401 398 -3
__page_pool_alloc_pages_slow 439 426 -13
page_pool_refill_alloc_cache 425 328 -97
Total: Before=3611, After=3498, chg -3.13%
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The check in pool_page_reusable (page_to_nid(page) == pool->p.nid) is
not valid if page_pool was configured with pool->p.nid = NUMA_NO_NODE.
The goal of the NUMA changes in commit d5394610b1 ("page_pool: Don't
recycle non-reusable pages"), were to have RX-pages that belongs to the
same NUMA node as the CPU processing RX-packet during softirq/NAPI. As
illustrated by the performance measurements.
This patch moves the NAPI checks out of fast-path, and at the same time
solves the NUMA_NO_NODE issue.
First realize that alloc_pages_node() with pool->p.nid = NUMA_NO_NODE
will lookup current CPU nid (Numa ID) via numa_mem_id(), which is used
as the the preferred nid. It is only in rare situations, where
e.g. NUMA zone runs dry, that page gets doesn't get allocated from
preferred nid. The page_pool API allows drivers to control the nid
themselves via controlling pool->p.nid.
This patch moves the NAPI check to when alloc cache is refilled, via
dequeuing/consuming pages from the ptr_ring. Thus, we can allow placing
pages from remote NUMA into the ptr_ring, as the dequeue/consume step
will check the NUMA node. All current drivers using page_pool will
alloc/refill RX-ring from same CPU running softirq/NAPI process.
Drivers that control the nid explicitly, also use page_pool_update_nid
when changing nid runtime. To speed up transision to new nid the alloc
cache is now flushed on nid changes. This force pages to come from
ptr_ring, which does the appropate nid check.
For the NUMA_NO_NODE case, when a NIC IRQ is moved to another NUMA
node, we accept that transitioning the alloc cache doesn't happen
immediately. The preferred nid change runtime via consulting
numa_mem_id() based on the CPU processing RX-packets.
Notice, to avoid stressing the page buddy allocator and avoid doing too
much work under softirq with preempt disabled, the NUMA check at
ptr_ring dequeue will break the refill cycle, when detecting a NUMA
mismatch. This will cause a slower transition, but its done on purpose.
Fixes: d5394610b1 ("page_pool: Don't recycle non-reusable pages")
Reported-by: Li RongQing <lirongqing@baidu.com>
Reported-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Revert "net/sched: cls_u32: fix refcount leak in the error path of
u32_change()", and fix the u32 refcount leak in a more generic way that
preserves the semantic of rule dumping.
On tc filters that don't support lockless insertion/removal, there is no
need to guard against concurrent insertion when a removal is in progress.
Therefore, for most of them we can avoid a full walk() when deleting, and
just decrease the refcount, like it was done on older Linux kernels.
This fixes situations where walk() was wrongly detecting a non-empty
filter, like it happened with cls_u32 in the error path of change(), thus
leading to failures in the following tdc selftests:
6aa7: (filter, u32) Add/Replace u32 with source match and invalid indev
6658: (filter, u32) Add/Replace u32 with custom hash table and invalid handle
74c2: (filter, u32) Add/Replace u32 filter with invalid hash table id
On cls_flower, and on (future) lockless filters, this check is necessary:
move all the check_empty() logic in a callback so that each filter
can have its own implementation. For cls_flower, it's sufficient to check
if no IDRs have been allocated.
This reverts commit 275c44aa19.
Changes since v1:
- document the need for delete_empty() when TCF_PROTO_OPS_DOIT_UNLOCKED
is used, thanks to Vlad Buslov
- implement delete_empty() without doing fl_walk(), thanks to Vlad Buslov
- squash revert and new fix in a single patch, to be nice with bisect
tests that run tdc on u32 filter, thanks to Dave Miller
Fixes: 275c44aa19 ("net/sched: cls_u32: fix refcount leak in the error path of u32_change()")
Fixes: 6676d5e416 ("net: sched: set dedicated tcf_walker flag when tp is empty")
Suggested-by: Jamal Hadi Salim <jhs@mojatatu.com>
Suggested-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: Davide Caratti <dcaratti@redhat.com>
Reviewed-by: Vlad Buslov <vladbu@mellanox.com>
Tested-by: Jamal Hadi Salim <jhs@mojatatu.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
gma_flag was set at the time of GMA command request but it should
only be set after getting successful response. Movinng this flag
setting in GMA response handler.
This flag is used mainly for not repeating GMA command once
received MAC address.
Signed-off-by: Vijay Khemka <vijaykhemka@fb.com>
Reviewed-by: Samuel Mendoza-Jonas <sam@mendozajonas.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
sctp_outq_sack is the main function handles SACK, it is called very
frequently. As the commit "move trace_sctp_probe_path into sctp_outq_sack"
added below code to this function, sctp tracepoint is disabled most of time,
but the loop of transport list will be always called even though the
tracepoint is disabled, this is unnecessary.
+ /* SCTP path tracepoint for congestion control debugging. */
+ list_for_each_entry(transport, transport_list, transports) {
+ trace_sctp_probe_path(transport, asoc);
+ }
This patch is to add tracepoint enabled check at outside of the loop of
transport list, and avoid traversing the loop when trace is disabled,
it is a small optimization.
Signed-off-by: Kevin Kou <qdkevin.kou@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
>From commit 50895b9de1 ("tcp: highest_sack fix"), the logic about
setting tp->highest_sack to the head of the send queue was removed.
Of course the logic is error prone, but it is logical. Before we
remove the pointer to the highest sack skb and use the seq instead,
we need to set tp->highest_sack to NULL when there is no skb after
the last sack, and then replace NULL with the real skb when new skb
inserted into the rtx queue, because the NULL means the highest sack
seq is tp->snd_nxt. If tp->highest_sack is NULL and new data sent,
the next ACK with sack option will increase tp->reordering unexpectedly.
This patch sets tp->highest_sack to the tail of the rtx queue if
it's NULL and new data is sent. The patch keeps the rule that the
highest_sack can only be maintained by sack processing, except for
this only case.
Fixes: 50895b9de1 ("tcp: highest_sack fix")
Signed-off-by: Cambda Zhu <cambda@linux.alibaba.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pablo Neira Ayuso says:
====================
Netfilter updates for net-next
The following patchset contains Netfilter updates for net-next:
1) Remove #ifdef pollution around nf_ingress(), from Lukas Wunner.
2) Document ingress hook in netdevice, also from Lukas.
3) Remove htons() in tunnel metadata port netlink attributes,
from Xin Long.
4) Missing erspan netlink attribute validation also from Xin Long.
5) Missing erspan version in tunnel, from Xin Long.
6) Missing attribute nest in NFTA_TUNNEL_KEY_OPTS_{VXLAN,ERSPAN}
Patch from Xin Long.
7) Missing nla_nest_cancel() in tunnel netlink dump path,
from Xin Long.
8) Remove two exported conntrack symbols with no clients,
from Florian Westphal.
9) Add nft_meta_get_eval_time() helper to nft_meta, from Florian.
10) Add nft_meta_pkttype helper for loopback, also from Florian.
11) Add nft_meta_socket uid helper, from Florian Westphal.
12) Add nft_meta_cgroup helper, from Florian.
13) Add nft_meta_ifkind helper, from Florian.
14) Group all interface related meta selector, from Florian.
15) Add nft_prandom_u32() helper, from Florian.
16) Add nft_meta_rtclassid helper, from Florian.
17) Add support for matching on the slave device index,
from Florian.
This batch, among other things, contains updates for the netfilter
tunnel netlink interface: This extension is still incomplete and lacking
proper userspace support which is actually my fault, I did not find the
time to go back and finish this. This update is breaking tunnel UAPI in
some aspects to fix it but do it better sooner than never.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
It is possible to kill PTP on a DSA switch completely and absolutely,
until a reboot, with a simple command:
tcpdump -i eth2 -j adapter_unsynced
where eth2 is the switch's DSA master.
Why? Well, in short, the PTP API in place today is a bit rudimentary and
relies on applications to retrieve the TX timestamps by polling the
error queue and looking at the cmsg structure. But there is no timestamp
identification of any sorts (except whether it's HW or SW), you don't
know how many more timestamps are there to come, which one is this one,
from whom it is, etc. In other words, the SO_TIMESTAMPING API is
fundamentally limited in that you can get a single HW timestamp from the
stack.
And the "-j adapter_unsynced" flag of tcpdump enables hardware
timestamping.
So let's imagine what happens when the DSA master decides it wants to
deliver TX timestamps to the skb's socket too:
- The timestamp that the user space sees is taken by the DSA master.
Whereas the RX timestamp will eventually be overwritten by the DSA
switch. So the RX and TX timestamps will be in different time bases
(aka garbage).
- The user space applications have no way to deal with the second (real)
TX timestamp finally delivered by the DSA switch, or even to know to
wait for it.
Take ptp4l from the linuxptp project, for example. This is its behavior
after running tcpdump, before the patch:
ptp4l[172]: [6469.594] Unexpected data on socket err queue:
ptp4l[172]: [6469.693] rms 8 max 16 freq -21257 +/- 11 delay 748 +/- 0
ptp4l[172]: [6469.711] Unexpected data on socket err queue:
ptp4l[172]: 0020 00 00 00 1f 7b ff fe 63 02 48 00 03 aa 05 00 fd
ptp4l[172]: 0030 00 00 00 00 00 00 00 00 00 00
ptp4l[172]: [6469.721] Unexpected data on socket err queue:
ptp4l[172]: 0000 01 80 c2 00 00 0e 00 1f 7b 63 02 48 88 f7 10 02
ptp4l[172]: 0010 00 2c 00 00 02 00 00 00 00 00 00 00 00 00 00 00
ptp4l[172]: 0020 00 00 00 1f 7b ff fe 63 02 48 00 01 c6 b1 00 fd
ptp4l[172]: 0030 00 00 00 00 00 00 00 00 00 00
ptp4l[172]: [6469.838] Unexpected data on socket err queue:
ptp4l[172]: 0000 01 80 c2 00 00 0e 00 1f 7b 63 02 48 88 f7 10 02
ptp4l[172]: 0010 00 2c 00 00 02 00 00 00 00 00 00 00 00 00 00 00
ptp4l[172]: 0020 00 00 00 1f 7b ff fe 63 02 48 00 03 aa 06 00 fd
ptp4l[172]: 0030 00 00 00 00 00 00 00 00 00 00
ptp4l[172]: [6469.848] Unexpected data on socket err queue:
ptp4l[172]: 0000 01 80 c2 00 00 0e 00 1f 7b 63 02 48 88 f7 13 02
ptp4l[172]: 0010 00 36 00 00 02 00 00 00 00 00 00 00 00 00 00 00
ptp4l[172]: 0020 00 00 00 1f 7b ff fe 63 02 48 00 04 1a 45 05 7f
ptp4l[172]: 0030 00 00 5e 05 41 32 27 c2 1a 68 00 04 9f ff fe 05
ptp4l[172]: 0040 de 06 00 01
ptp4l[172]: [6469.855] Unexpected data on socket err queue:
ptp4l[172]: 0000 01 80 c2 00 00 0e 00 1f 7b 63 02 48 88 f7 10 02
ptp4l[172]: 0010 00 2c 00 00 02 00 00 00 00 00 00 00 00 00 00 00
ptp4l[172]: 0020 00 00 00 1f 7b ff fe 63 02 48 00 01 c6 b2 00 fd
ptp4l[172]: 0030 00 00 00 00 00 00 00 00 00 00
ptp4l[172]: [6469.974] Unexpected data on socket err queue:
ptp4l[172]: 0000 01 80 c2 00 00 0e 00 1f 7b 63 02 48 88 f7 10 02
ptp4l[172]: 0010 00 2c 00 00 02 00 00 00 00 00 00 00 00 00 00 00
ptp4l[172]: 0020 00 00 00 1f 7b ff fe 63 02 48 00 03 aa 07 00 fd
ptp4l[172]: 0030 00 00 00 00 00 00 00 00 00 00
The ptp4l program itself is heavily patched to show this (more details
here [0]). Otherwise, by default it just hangs.
On the other hand, with the DSA patch to disallow HW timestamping
applied:
tcpdump -i eth2 -j adapter_unsynced
tcpdump: SIOCSHWTSTAMP failed: Device or resource busy
So it is a fact of life that PTP timestamping on the DSA master is
incompatible with timestamping on the switch MAC, at least with the
current API. And if the switch supports PTP, taking the timestamps from
the switch MAC is highly preferable anyway, due to the fact that those
don't contain the queuing latencies of the switch. So just disallow PTP
on the DSA master if there is any PTP-capable switch attached.
[0]: https://sourceforge.net/p/linuxptp/mailman/message/36880648/
Fixes: 0336369d3a ("net: dsa: forward hardware timestamping ioctls to switch driver")
Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implement LINKSTATE_GET netlink request to get link state information.
At the moment, only link up flag as provided by ETHTOOL_GLINK ioctl command
is returned.
LINKSTATE_GET request can be used with NLM_F_DUMP (without device
identification) to request the information for all devices in current
network namespace providing the data.
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Send ETHTOOL_MSG_LINKMODES_NTF notification message whenever device link
settings or advertised modes are modified using ETHTOOL_MSG_LINKMODES_SET
netlink message or ETHTOOL_SLINKSETTINGS or ETHTOOL_SSET ioctl commands.
The notification message has the same format as reply to LINKMODES_GET
request. ETHTOOL_MSG_LINKMODES_SET netlink request only triggers the
notification if there is a change but the ioctl command handlers do not
check if there is an actual change and trigger the notification whenever
the commands are executed.
As all work is done by ethnl_default_notify() handler and callback
functions introduced to handle LINKMODES_GET requests, all that remains is
adding entries for ETHTOOL_MSG_LINKMODES_NTF into ethnl_notify_handlers and
ethnl_default_notify_ops lookup tables and calls to ethtool_notify() where
needed.
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implement LINKMODES_SET netlink request to set advertised linkmodes and
related attributes as ETHTOOL_SLINKSETTINGS and ETHTOOL_SSET commands do.
The request allows setting autonegotiation flag, speed, duplex and
advertised link modes.
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implement LINKMODES_GET netlink request to get link modes related
information provided by ETHTOOL_GLINKSETTINGS and ETHTOOL_GSET ioctl
commands.
This request provides supported, advertised and peer advertised link modes,
autonegotiation flag, speed and duplex.
LINKMODES_GET request can be used with NLM_F_DUMP (without device
identification) to request the information for all devices in current
network namespace providing the data.
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Send ETHTOOL_MSG_LINKINFO_NTF notification message whenever device link
settings are modified using ETHTOOL_MSG_LINKINFO_SET netlink message or
ETHTOOL_SLINKSETTINGS or ETHTOOL_SSET ioctl commands.
The notification message has the same format as reply to LINKINFO_GET
request. ETHTOOL_MSG_LINKINFO_SET netlink request only triggers the
notification if there is a change but the ioctl command handlers do not
check if there is an actual change and trigger the notification whenever
the commands are executed.
As all work is done by ethnl_default_notify() handler and callback
functions introduced to handle LINKINFO_GET requests, all that remains is
adding entries for ETHTOOL_MSG_LINKINFO_NTF into ethnl_notify_handlers and
ethnl_default_notify_ops lookup tables and calls to ethtool_notify() where
needed.
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The ethtool netlink notifications have the same format as related GET
replies so that if generic GET handling framework is used to process GET
requests, its callbacks and instance of struct get_request_ops can be
also used to compose corresponding notification message.
Provide function ethnl_std_notify() to be used as notification handler in
ethnl_notify_handlers table.
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implement LINKINFO_SET netlink request to set link settings queried by
LINKINFO_GET message.
Only physical port, phy MDIO address and MDI(-X) control can be set,
attempt to modify MDI(-X) status and transceiver is rejected.
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implement LINKINFO_GET netlink request to get basic link settings provided
by ETHTOOL_GLINKSETTINGS and ETHTOOL_GSET ioctl commands.
This request provides settings not directly related to autonegotiation and
link mode selection: physical port, phy MDIO address, MDI(-X) status,
MDI(-X) control and transceiver.
LINKINFO_GET request can be used with NLM_F_DUMP (without device
identification) to request the information for all devices in current
network namespace providing the data.
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Requests a contents of one or more string sets, i.e. indexed arrays of
strings; this information is provided by ETHTOOL_GSSET_INFO and
ETHTOOL_GSTRINGS commands of ioctl interface. Unlike ioctl interface, all
information can be retrieved with one request and mulitple string sets can
be requested at once.
There are three types of requests:
- no NLM_F_DUMP, no device: get "global" stringsets
- no NLM_F_DUMP, with device: get string sets related to the device
- NLM_F_DUMP, no device: get device related string sets for all devices
Client can request either all string sets of given type (global or device
related) or only specific sets. With ETHTOOL_A_STRSET_COUNTS flag set, only
set sizes (numbers of strings) are returned.
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Significant part of GET request processing is common for most request
types but unfortunately it cannot be easily separated from type specific
code as we need to alternate between common actions (parsing common request
header, allocating message and filling netlink/genetlink headers etc.) and
specific actions (querying the device, composing the reply). The processing
also happens in three different situations: "do" request, "dump" request
and notification, each doing things in slightly different way.
The request specific code is implemented in four or five callbacks defined
in an instance of struct get_request_ops:
parse_request() - parse incoming message
prepare_data() - retrieve data from driver or NIC
reply_size() - estimate reply message size
fill_reply() - compose reply message
cleanup_data() - (optional) clean up additional data
Other members of struct get_request_ops describe the data structure holding
information from client request and data used to compose the message. The
default handlers ethnl_default_doit(), ethnl_default_dumpit(),
ethnl_default_start() and ethnl_default_done() can be then used in genl_ops
handler. Notification handler will be introduced in a later patch.
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add infrastructure for ethtool netlink notifications. There is only one
multicast group "monitor" which is used to notify userspace about changes
and actions performed. Notification messages (types using suffix _NTF)
share the format with replies to GET requests.
Notifications are supposed to be broadcasted on every configuration change,
whether it is done using the netlink interface or ioctl one. Netlink SET
requests only trigger a notification if some data is actually changed.
To trigger an ethtool notification, both ethtool netlink and external code
use ethtool_notify() helper. This helper requires RTNL to be held and may
sleep. Handlers sending messages for specific notification message types
are registered in ethnl_notify_handlers array. As notifications can be
triggered from other code, ethnl_ok flag is used to prevent an attempt to
send notification before genetlink family is registered.
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The ethtool netlink code uses common framework for passing arbitrary
length bit sets to allow future extensions. A bitset can be a list (only
one bitmap) or can consist of value and mask pair (used e.g. when client
want to modify only some bits). A bitset can use one of two formats:
verbose (bit by bit) or compact.
Verbose format consists of bitset size (number of bits), list flag and
an array of bit nests, telling which bits are part of the list or which
bits are in the mask and which of them are to be set. In requests, bits
can be identified by index (position) or by name. In replies, kernel
provides both index and name. Verbose format is suitable for "one shot"
applications like standard ethtool command as it avoids the need to
either keep bit names (e.g. link modes) in sync with kernel or having to
add an extra roundtrip for string set request (e.g. for private flags).
Compact format uses one (list) or two (value/mask) arrays of 32-bit
words to store the bitmap(s). It is more suitable for long running
applications (ethtool in monitor mode or network management daemons)
which can retrieve the names once and then pass only compact bitmaps to
save space.
Userspace requests can use either format; ETHTOOL_FLAG_COMPACT_BITSETS
flag in request header tells kernel which format to use in reply.
Notifications always use compact format.
As some code uses arrays of unsigned long for internal representation and
some arrays of u32 (or even a single u32), two sets of parse/compose
helpers are introduced. To avoid code duplication, helpers for unsigned
long arrays are implemented as wrappers around helpers for u32 arrays.
There are two reasons for this choice: (1) u32 arrays are more frequent in
ethtool code and (2) unsigned long array can be always interpreted as an
u32 array on little endian 64-bit and all 32-bit architectures while we
would need special handling for odd number of u32 words in the opposite
direction.
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add common request/reply header definition and helpers to parse request
header and fill reply header. Provide ethnl_update_* helpers to update
structure members from request attributes (to be used for *_SET requests).
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Basic genetlink and init infrastructure for the netlink interface, register
genetlink family "ethtool". Add CONFIG_ETHTOOL_NETLINK Kconfig option to
make the build optional. Add initial overall interface description into
Documentation/networking/ethtool-netlink.rst, further patches will add more
detailed information.
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
There's no skb_pull performed when a mirred action is set at egress of a
mac device, with a target device/action that expects skb->data to point
at the network header.
As a result, either the target device is errornously given an skb with
data pointing to the mac (egress case), or the net stack receives the
skb with data pointing to the mac (ingress case).
E.g:
# tc qdisc add dev eth9 root handle 1: prio
# tc filter add dev eth9 parent 1: prio 9 protocol ip handle 9 basic \
action mirred egress redirect dev tun0
(tun0 is a tun device. result: tun0 errornously gets the eth header
instead of the iph)
Revise the push/pull logic of tcf_mirred_act() to not rely on the
skb_at_tc_ingress() vs tcf_mirred_act_wants_ingress() comparison, as it
does not cover all "pull" cases.
Instead, calculate whether the required action on the target device
requires the data to point at the network header, and compare this to
whether skb->data points to network header - and make the push/pull
adjustments as necessary.
Fixes: 1da177e4c3 ("Linux-2.6.12-rc2")
Signed-off-by: Shmulik Ladkani <sladkani@proofpoint.com>
Tested-by: Jamal Hadi Salim <jhs@mojatatu.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The function sctp_sf_eat_sack_6_2 now performs the Verification
Tag validation, Chunk length validation, Bogu check, and also
the detection of out-of-order SACK based on the RFC2960
Section 6.2 at the beginning, and finally performs the further
processing of SACK. The trace_sctp_probe now triggered before
the above necessary validation and check.
this patch is to do the trace_sctp_probe after the chunk sanity
tests, but keep doing trace if the SACK received is out of order,
for the out-of-order SACK is valuable to congestion control
debugging.
v1->v2:
- keep doing SCTP trace if the SACK is out of order as Marcelo's
suggestion.
v2->v3:
- regenerate the patch as v2 generated on top of v1, and add
'net-next' tag to the new one as Marcelo's comments.
Signed-off-by: Kevin Kou <qdkevin.kou@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For years we disabled Hystart ACK train detection at Google
because it was fooled by TCP pacing.
ACK train detection uses a simple heuristic, detecting if
we receive ACK past half the RTT, to exit slow start before
hitting the bottleneck and experience massive drops.
But pacing by design might delay packets up to RTT/2,
so we need to tweak the Hystart logic to be aware of this
extra delay.
Tested:
Added a 100 usec delay at receiver.
Before:
nstat -n;for f in {1..10}; do ./super_netperf 1 -H lpaa24 -l -4000000; done;nstat|egrep "Hystart"
9117
7057
9553
8300
7030
6849
9533
10126
6876
8473
TcpExtTCPHystartTrainDetect 10 0.0
TcpExtTCPHystartTrainCwnd 1230 0.0
After :
nstat -n;for f in {1..10}; do ./super_netperf 1 -H lpaa24 -l -4000000; done;nstat|egrep "Hystart"
9845
10103
10866
11096
11936
11487
11773
12188
11066
11894
TcpExtTCPHystartTrainDetect 10 0.0
TcpExtTCPHystartTrainCwnd 6462 0.0
Disabling Hystart ACK Train detection gives similar numbers
echo 2 >/sys/module/tcp_cubic/parameters/hystart_detect
nstat -n;for f in {1..10}; do ./super_netperf 1 -H lpaa24 -l -4000000; done;nstat|egrep "Hystart"
11173
10954
12455
10627
11578
11583
11222
10880
10665
11366
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
After switching ca->delay_min to usec resolution, we exit
slow start prematurely for very low RTT flows, setting
snd_ssthresh to 20.
The reason is that delay_min is fed with RTT of small packet
trains. Then as cwnd is increased, TCP sends bigger TSO packets.
LRO/GRO aggregation and/or interrupt mitigation strategies
on receiver tend to inflate RTT samples.
Fix this by adding to delay_min the expected delay of
two TSO packets, given current pacing rate.
Tested:
Sender uses pfifo_fast qdisc
Before :
$ nstat -n;for f in {1..10}; do ./super_netperf 1 -H lpaa24 -l -4000000; done;nstat|egrep "Hystart"
11348
11707
11562
11428
11773
11534
9878
11693
10597
10968
TcpExtTCPHystartTrainDetect 10 0.0
TcpExtTCPHystartTrainCwnd 200 0.0
After :
$ nstat -n;for f in {1..10}; do ./super_netperf 1 -H lpaa24 -l -4000000; done;nstat|egrep "Hystart"
14877
14517
15797
18466
17376
14833
17558
17933
16039
18059
TcpExtTCPHystartTrainDetect 10 0.0
TcpExtTCPHystartTrainCwnd 1670 0.0
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Current 1ms clock feeds ca->round_start, ca->delay_min,
ca->last_ack.
This is quite problematic for data-center flows, where delay_min
is way below 1 ms.
This means Hystart Train detection triggers every time jiffies value
is updated, since "((s32)(now - ca->round_start) > ca->delay_min >> 4)"
expression becomes true.
This kind of random behavior can be solved by reusing the existing
usec timestamp that TCP keeps in tp->tcp_mstamp
Note that a followup patch will tweak things a bit, because
during slow start, GRO aggregation on receivers naturally
increases the RTT as TSO packets gradually come to ~64KB size.
To recap, right after this patch CUBIC Hystart train detection
is more aggressive, since short RTT flows might exit slow start at
cwnd = 20, instead of being possibly unbounded.
Following patch will address this problem.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If we initialize ca->curr_rtt to ~0U, we do not need to test
for zero value in hystart_update()
We only read ca->curr_rtt if at least HYSTART_MIN_SAMPLES have
been processed, and thus ca->curr_rtt will have a sane value.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
