This change enables congestion control to update cwnd based on
not only packet cumulatively acked but also packets delivered
out-of-order. This makes congestion control robust against packet
reordering because it may raise cwnd as long as packets are being
delivered once reordering has been detected (i.e., it only cares
the amount of packets delivered, not the ordering among them).
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A small refactoring that gets number of packets cumulatively acked
from tcp_clean_rtx_queue() directly.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch changes the accounting of how many packets are
newly acked or sacked when the sender receives an ACK.
The current approach basically computes
newly_acked_sacked = (prior_packets - prior_sacked) -
(tp->packets_out - tp->sacked_out)
where prior_packets and prior_sacked out are snapshot
at the beginning of the ACK processing.
The new approach tracks the delivery information via a new
TCP state variable "delivered" which monotically increases
as new packets are delivered in order or out-of-order.
The reason for this change is that the current approach is
brittle that produces negative or inaccurate estimate.
1) For non-SACK connections, an ACK that advances the SND.UNA
could reset the DUPACK counters (tp->sacked_out) in
tcp_process_loss() or tcp_fastretrans_alert(). This inflates
the inflight suddenly and causes under-estimate or even
negative estimate. Here is a real example:
before after (processing ACK)
packets_out 75 73
sacked_out 23 0
ca state Loss Open
The old approach computes (75-23) - (73 - 0) = -21 delivered
while the new approach computes 1 delivered since it
considers the 2nd-24th packets are delivered OOO.
2) MSS change would re-count packets_out and sacked_out so
the estimate is in-accurate and can even become negative.
E.g., the inflight is doubled when MSS is halved.
3) Spurious retransmission signaled by DSACK is not accounted
The new approach is simpler and more robust. For SACK connections,
tp->delivered increments as packets are being acked or sacked in
SACK and ACK processing.
For non-sack connections, it's done in tcp_remove_reno_sacks() and
tcp_add_reno_sack(). When an ACK advances the SND.UNA, tp->delivered
is incremented by the number of packets ACKed (less the current
number of DUPACKs received plus one packet hole). Upon receiving
a DUPACK, tp->delivered is incremented assuming one out-of-order
packet is delivered.
Upon receiving a DSACK, tp->delivered is incremtened assuming one
retransmission is delivered in tcp_sacktag_write_queue().
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently the cwnd is reduced and increased in various different
places. The reduction happens in various places in the recovery
state processing (tcp_fastretrans_alert) while the increase
happens afterward.
A better sequence is to identify lost packets and update
the congestion control state (icsk_ca_state) first. Then base
on the new state, up/down the cwnd in one central place. It's
more clear to reason cwnd changes.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The retransmission and F-RTO transmission currently happen inside
recovery state processing (tcp_fastretrans_alert) but before
congestion control. This refactoring moves the logic after both
s.t. we can determine how much to send (cwnd) before deciding what to
send.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When we acknowledge a FIN, it is not enough to ack the sequence number
and queue the skb into receive queue. We also have to call tcp_fin()
to properly update socket state and send proper poll() notifications.
It seems we also had the problem if we received a SYN packet with the
FIN flag set, but it does not seem an urgent issue, as no known
implementation can do that.
Fixes: 61d2bcae99 ("tcp: fastopen: accept data/FIN present in SYNACK message")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If we remove the SYN flag from the skbs that tcp_fastopen_add_skb()
places in socket receive queue, then we can remove the test that
tcp_recvmsg() has to perform in fast path.
All we have to do is to adjust SEQ in the slow path.
For the moment, we place an unlikely() and output a message
if we find an skb having SYN flag set.
Goal would be to get rid of the test completely.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
RFC 7413 (TCP Fast Open) 4.2.2 states that the SYNACK message
MAY include data and/or FIN
This patch adds support for the client side :
If we receive a SYNACK with payload or FIN, queue the skb instead
of ignoring it.
Since we already support the same for SYN, we refactor the existing
code and reuse it. Note we need to clone the skb, so this operation
might fail under memory pressure.
Sara Dickinson pointed out FreeBSD server Fast Open implementation
was planned to generate such SYNACK in the future.
The server side might be implemented on linux later.
Reported-by: Sara Dickinson <sara@sinodun.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull networking fixes from David Miller:
"This looks like a lot but it's a mixture of regression fixes as well
as fixes for longer standing issues.
1) Fix on-channel cancellation in mac80211, from Johannes Berg.
2) Handle CHECKSUM_COMPLETE properly in xt_TCPMSS netfilter xtables
module, from Eric Dumazet.
3) Avoid infinite loop in UDP SO_REUSEPORT logic, also from Eric
Dumazet.
4) Avoid a NULL deref if we try to set SO_REUSEPORT after a socket is
bound, from Craig Gallek.
5) GRO key comparisons don't take lightweight tunnels into account,
from Jesse Gross.
6) Fix struct pid leak via SCM credentials in AF_UNIX, from Eric
Dumazet.
7) We need to set the rtnl_link_ops of ipv6 SIT tunnels before we
register them, otherwise the NEWLINK netlink message is missing
the proper attributes. From Thadeu Lima de Souza Cascardo.
8) Several Spectrum chip bug fixes for mlxsw switch driver, from Ido
Schimmel
9) Handle fragments properly in ipv4 easly socket demux, from Eric
Dumazet.
10) Don't ignore the ifindex key specifier on ipv6 output route
lookups, from Paolo Abeni"
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (128 commits)
tcp: avoid cwnd undo after receiving ECN
irda: fix a potential use-after-free in ircomm_param_request
net: tg3: avoid uninitialized variable warning
net: nb8800: avoid uninitialized variable warning
net: vxge: avoid unused function warnings
net: bgmac: clarify CONFIG_BCMA dependency
net: hp100: remove unnecessary #ifdefs
net: davinci_cpdma: use dma_addr_t for DMA address
ipv6/udp: use sticky pktinfo egress ifindex on connect()
ipv6: enforce flowi6_oif usage in ip6_dst_lookup_tail()
netlink: not trim skb for mmaped socket when dump
vxlan: fix a out of bounds access in __vxlan_find_mac
net: dsa: mv88e6xxx: fix port VLAN maps
fib_trie: Fix shift by 32 in fib_table_lookup
net: moxart: use correct accessors for DMA memory
ipv4: ipconfig: avoid unused ic_proto_used symbol
bnxt_en: Fix crash in bnxt_free_tx_skbs() during tx timeout.
bnxt_en: Exclude rx_drop_pkts hw counter from the stack's rx_dropped counter.
bnxt_en: Ring free response from close path should use completion ring
net_sched: drr: check for NULL pointer in drr_dequeue
...
RFC 4015 section 3.4 says the TCP sender MUST refrain from
reversing the congestion control state when the ACK signals
congestion through the ECN-Echo flag. Currently we may not
always do that when prior_ssthresh is reset upon receiving
ACKs with ECE marks. This patch fixes that.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The fib_table_lookup function had a shift by 32 that triggered a UBSAN
warning. This was due to the fact that I had placed the shift first and
then followed it with the check for the suffix length to ignore the
undefined behavior. If we reorder this so that we verify the suffix is
less than 32 before shifting the value we can avoid the issue.
Reported-by: Toralf Förster <toralf.foerster@gmx.de>
Signed-off-by: Alexander Duyck <aduyck@mirantis.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When CONFIG_PROC_FS, CONFIG_IP_PNP_BOOTP, CONFIG_IP_PNP_DHCP and
CONFIG_IP_PNP_RARP are all disabled, we get a warning about the
ic_proto_used variable being unused:
net/ipv4/ipconfig.c:146:12: error: 'ic_proto_used' defined but not used [-Werror=unused-variable]
This avoids the warning, by making the definition conditional on
whether a dynamic IP configuration protocol is configured. If not,
we know that the value is always zero, so we can optimize away the
variable and all code that depends on it.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
We should not assume a valid protocol header is present,
as this is not the case for IPv4 fragments.
Lets avoid extra cache line misses and potential bugs
if we actually find a socket and incorrectly uses its dst.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
With some combinations of user provided flags in netlink command,
it is possible to call tcp_get_info() with a buffer that is not 8-bytes
aligned.
It does matter on some arches, so we need to use put_unaligned() to
store the u64 fields.
Current iproute2 package does not trigger this particular issue.
Fixes: 0df48c26d8 ("tcp: add tcpi_bytes_acked to tcp_info")
Fixes: 977cb0ecf8 ("tcp: add pacing_rate information into tcp_info")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This commit fixes a corner case in tcp_mark_head_lost() which was
causing the WARN_ON(len > skb->len) in tcp_fragment() to fire.
tcp_mark_head_lost() was assuming that if a packet has
tcp_skb_pcount(skb) of N, then it's safe to fragment off a prefix of
M*mss bytes, for any M < N. But with the tricky way TCP pcounts are
maintained, this is not always true.
For example, suppose the sender sends 4 1-byte packets and have the
last 3 packet sacked. It will merge the last 3 packets in the write
queue into an skb with pcount = 3 and len = 3 bytes. If another
recovery happens after a sack reneging event, tcp_mark_head_lost()
may attempt to split the skb assuming it has more than 2*MSS bytes.
This sounds very counterintuitive, but as the commit description for
the related commit c0638c247f ("tcp: don't fragment SACKed skbs in
tcp_mark_head_lost()") notes, this is because tcp_shifted_skb()
coalesces adjacent regions of SACKed skbs, and when doing this it
preserves the sum of their packet counts in order to reflect the
real-world dynamics on the wire. The c0638c247f commit tried to
avoid problems by not fragmenting SACKed skbs, since SACKed skbs are
where the non-proportionality between pcount and skb->len/mss is known
to be possible. However, that commit did not handle the case where
during a reneging event one of these weird SACKed skbs becomes an
un-SACKed skb, which tcp_mark_head_lost() can then try to fragment.
The fix is to simply mark the entire skb lost when this happens.
This makes the recovery slightly more aggressive in such corner
cases before we detect reordering. But once we detect reordering
this code path is by-passed because FACK is disabled.
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch replaces uses of the long obsolete hash interface with
ahash.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: David S. Miller <davem@davemloft.net>
The ESP algorithms using CBC mode require echainiv. Hence INET*_ESP have
to select CRYPTO_ECHAINIV in order to work properly. This solves the
issues caused by a misconfiguration as described in [1].
The original approach, patching crypto/Kconfig was turned down by
Herbert Xu [2].
[1] https://lists.strongswan.org/pipermail/users/2015-December/009074.html
[2] http://marc.info/?l=linux-crypto-vger&m=145224655809562&w=2
Signed-off-by: Thomas Egerer <hakke_007@gmx.de>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
There are many locations that do
if (memory_was_allocated_by_vmalloc)
vfree(ptr);
else
kfree(ptr);
but kvfree() can handle both kmalloc()ed memory and vmalloc()ed memory
using is_vmalloc_addr(). Unless callers have special reasons, we can
replace this branch with kvfree(). Please check and reply if you found
problems.
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Jan Kara <jack@suse.com>
Acked-by: Russell King <rmk+kernel@arm.linux.org.uk>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Acked-by: "Rafael J. Wysocki" <rjw@rjwysocki.net>
Acked-by: David Rientjes <rientjes@google.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Oleg Drokin <oleg.drokin@intel.com>
Cc: Boris Petkov <bp@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Neal reported crashes with this stack trace :
RIP: 0010:[<ffffffff8c57231b>] tcp_v4_send_ack+0x41/0x20f
...
CR2: 0000000000000018 CR3: 000000044005c000 CR4: 00000000001427e0
...
[<ffffffff8c57258e>] tcp_v4_reqsk_send_ack+0xa5/0xb4
[<ffffffff8c1a7caa>] tcp_check_req+0x2ea/0x3e0
[<ffffffff8c19e420>] tcp_rcv_state_process+0x850/0x2500
[<ffffffff8c1a6d21>] tcp_v4_do_rcv+0x141/0x330
[<ffffffff8c56cdb2>] sk_backlog_rcv+0x21/0x30
[<ffffffff8c098bbd>] tcp_recvmsg+0x75d/0xf90
[<ffffffff8c0a8700>] inet_recvmsg+0x80/0xa0
[<ffffffff8c17623e>] sock_aio_read+0xee/0x110
[<ffffffff8c066fcf>] do_sync_read+0x6f/0xa0
[<ffffffff8c0673a1>] SyS_read+0x1e1/0x290
[<ffffffff8c5ca262>] system_call_fastpath+0x16/0x1b
The problem here is the skb we provide to tcp_v4_send_ack() had to
be parked in the backlog of a new TCP fastopen child because this child
was owned by the user at the time an out of window packet arrived.
Before queuing a packet, TCP has to set skb->dev to NULL as the device
could disappear before packet is removed from the queue.
Fix this issue by using the net pointer provided by the socket (being a
timewait or a request socket).
IPv6 is immune to the bug : tcp_v6_send_response() already gets the net
pointer from the socket if provided.
Fixes: 168a8f5805 ("tcp: TCP Fast Open Server - main code path")
Reported-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Jerry Chu <hkchu@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Lorenzo reported that we could not properly find v4mapped sockets
in inet_diag_find_one_icsk(). This patch fixes the issue.
Reported-by: Lorenzo Colitti <lorenzo@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Lorenzo Colitti <lorenzo@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
tcp_memcontrol.c only contains legacy memory.tcp.kmem.* file definitions
and mem_cgroup->tcp_mem init/destroy stuff. This doesn't belong to
network subsys. Let's move it to memcontrol.c. This also allows us to
reuse generic code for handling legacy memcg files.
Signed-off-by: Vladimir Davydov <vdavydov@virtuozzo.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: "David S. Miller" <davem@davemloft.net>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Let the user know that CONFIG_MEMCG_KMEM does not apply to the cgroup2
interface. This also makes legacy-only code sections stand out better.
[arnd@arndb.de: mm: memcontrol: only manage socket pressure for CONFIG_INET]
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Acked-by: Vladimir Davydov <vdavydov@virtuozzo.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This series adds accounting of the historical "kmem" memory consumers to
the cgroup2 memory controller.
These consumers include the dentry cache, the inode cache, kernel stack
pages, and a few others that are pointed out in patch 7/8. The
footprint of these consumers is directly tied to userspace activity in
common workloads, and so they have to be part of the minimally viable
configuration in order to present a complete feature to our users.
The cgroup2 interface of the memory controller is far from complete, but
this series, along with the socket memory accounting series, provides
the final semantic changes for the existing memory knobs in the cgroup2
interface, which is scheduled for initial release in the next merge
window.
This patch (of 8):
Remove unused css argument frmo memcg_init_kmem()
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Tejun Heo <tj@kernel.org>
Acked-by: Vladimir Davydov <vdavydov@virtuozzo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Using a combination of connected and un-connected sockets, Dmitry
was able to trigger soft lockups with his fuzzer.
The problem is that sockets in the SO_REUSEPORT array might have
different scores.
Right after sk2=socket(), setsockopt(sk2,...,SO_REUSEPORT, on) and
bind(sk2, ...), but _before_ the connect(sk2) is done, sk2 is added into
the soreuseport array, with a score which is smaller than the score of
first socket sk1 found in hash table (I am speaking of the regular UDP
hash table), if sk1 had the connect() done, giving a +8 to its score.
hash bucket [X] -> sk1 -> sk2 -> NULL
sk1 score = 14 (because it did a connect())
sk2 score = 6
SO_REUSEPORT fast selection is an optimization. If it turns out the
score of the selected socket does not match score of first socket, just
fallback to old SO_REUSEPORT logic instead of trying to be too smart.
Normal SO_REUSEPORT users do not mix different kind of sockets, as this
mechanism is used for load balance traffic.
Fixes: e32ea7e747 ("soreuseport: fast reuseport UDP socket selection")
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Craig Gallek <kraigatgoog@gmail.com>
Acked-by: Craig Gallek <kraig@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull networking fixes from David Miller:
"A quick set of bug fixes after there initial networking merge:
1) Netlink multicast group storage allocator only was tested with
nr_groups equal to 1, make it work for other values too. From
Matti Vaittinen.
2) Check build_skb() return value in macb and hip04_eth drivers, from
Weidong Wang.
3) Don't leak x25_asy on x25_asy_open() failure.
4) More DMA map/unmap fixes in 3c59x from Neil Horman.
5) Don't clobber IP skb control block during GSO segmentation, from
Konstantin Khlebnikov.
6) ECN helpers for ipv6 don't fixup the checksum, from Eric Dumazet.
7) Fix SKB segment utilization estimation in xen-netback, from David
Vrabel.
8) Fix lockdep splat in bridge addrlist handling, from Nikolay
Aleksandrov"
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (26 commits)
bgmac: Fix reversed test of build_skb() return value.
bridge: fix lockdep addr_list_lock false positive splat
net: smsc: Add support h8300
xen-netback: free queues after freeing the net device
xen-netback: delete NAPI instance when queue fails to initialize
xen-netback: use skb to determine number of required guest Rx requests
net: sctp: Move sequence start handling into sctp_transport_get_idx()
ipv6: update skb->csum when CE mark is propagated
net: phy: turn carrier off on phy attach
net: macb: clear interrupts when disabling them
sctp: support to lookup with ep+paddr in transport rhashtable
net: hns: fixes no syscon error when init mdio
dts: hisi: fixes no syscon fault when init mdio
net: preserve IP control block during GSO segmentation
fsl/fman: Delete one function call "put_device" in dtsec_config()
hip04_eth: fix missing error handle for build_skb failed
3c59x: fix another page map/single unmap imbalance
3c59x: balance page maps and unmaps
x25_asy: Free x25_asy on x25_asy_open() failure.
mlxsw: fix SWITCHDEV_OBJ_ID_PORT_MDB
...
Skb_gso_segment() uses skb control block during segmentation.
This patch adds 32-bytes room for previous control block which
will be copied into all resulting segments.
This patch fixes kernel crash during fragmenting forwarded packets.
Fragmentation requires valid IP CB in skb for clearing ip options.
Also patch removes custom save/restore in ovs code, now it's redundant.
Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com>
Link: http://lkml.kernel.org/r/CALYGNiP-0MZ-FExV2HutTvE9U-QQtkKSoE--KN=JQE5STYsjAA@mail.gmail.com
Signed-off-by: David S. Miller <davem@davemloft.net>
According to <linux/jump_label.h> the direct use of struct static_key is
deprecated. Update the socket and slab accounting code accordingly.
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: David S. Miller <davem@davemloft.net>
Reported-by: Jason Baron <jbaron@akamai.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The unified hierarchy memory controller is going to use this jump label
as well to control the networking callbacks. Move it to the memory
controller code and give it a more generic name.
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Vladimir Davydov <vdavydov@virtuozzo.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
There won't be any separate counters for socket memory consumed by
protocols other than TCP in the future. Remove the indirection and link
sockets directly to their owning memory cgroup.
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Vladimir Davydov <vdavydov@virtuozzo.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
There won't be a tcp control soft limit, so integrating the memcg code
into the global skmem limiting scheme complicates things unnecessarily.
Replace this with simple and clear charge and uncharge calls--hidden
behind a jump label--to account skb memory.
Note that this is not purely aesthetic: as a result of shoehorning the
per-memcg code into the same memory accounting functions that handle the
global level, the old code would compare the per-memcg consumption
against the smaller of the per-memcg limit and the global limit. This
allowed the total consumption of multiple sockets to exceed the global
limit, as long as the individual sockets stayed within bounds. After
this change, the code will always compare the per-memcg consumption to
the per-memcg limit, and the global consumption to the global limit, and
thus close this loophole.
Without a soft limit, the per-memcg memory pressure state in sockets is
generally questionable. However, we did it until now, so we continue to
enter it when the hard limit is hit, and packets are dropped, to let
other sockets in the cgroup know that they shouldn't grow their transmit
windows, either. However, keep it simple in the new callback model and
leave memory pressure lazily when the next packet is accepted (as
opposed to doing it synchroneously when packets are processed). When
packets are dropped, network performance will already be in the toilet,
so that should be a reasonable trade-off.
As described above, consumption is now checked on the per-memcg level
and the global level separately. Likewise, memory pressure states are
maintained on both the per-memcg level and the global level, and a
socket is considered under pressure when either level asserts as much.
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Vladimir Davydov <vdavydov@virtuozzo.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
tcp_memcontrol replicates the global sysctl_mem limit array per cgroup,
but it only ever sets these entries to the value of the memory_allocated
page_counter limit. Use the latter directly.
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Vladimir Davydov <vdavydov@virtuozzo.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The number of allocated sockets is used for calculations in the soft
limit phase, where packets are accepted but the socket is under memory
pressure.
Since there is no soft limit phase in tcp_memcontrol, and memory
pressure is only entered when packets are already dropped, this is
actually dead code. Remove it.
As this is the last user of parent_cg_proto(), remove that too.
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: David S. Miller <davem@davemloft.net>
Reviewed-by: Vladimir Davydov <vdavydov@virtuozzo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Move the jump-label from sock_update_memcg() and sock_release_memcg() to
the callsite, and so eliminate those function calls when socket
accounting is not enabled.
This also eliminates the need for dummy functions because the calls will
be optimized away if the Kconfig options are not enabled.
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: David S. Miller <davem@davemloft.net>
Reviewed-by: Vladimir Davydov <vdavydov@virtuozzo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
There are two bits defined for cg_proto->flags - MEMCG_SOCK_ACTIVATED
and MEMCG_SOCK_ACTIVE - both are set in tcp_update_limit, but the former
is never cleared while the latter can be cleared by unsetting the limit.
This allows to disable tcp socket accounting for new sockets after it
was enabled by writing -1 to memory.kmem.tcp.limit_in_bytes while still
guaranteeing that memcg_socket_limit_enabled static key will be
decremented on memcg destruction.
This functionality looks dubious, because it is not clear what a use
case would be. By enabling tcp accounting a user accepts the price. If
they then find the performance degradation unacceptable, they can always
restart their workload with tcp accounting disabled. It does not seem
there is any need to flip it while the workload is running.
Besides, it contradicts to how kmem accounting API works: writing
whatever to memory.kmem.limit_in_bytes enables kmem accounting for the
cgroup in question, after which it cannot be disabled. Therefore one
might expect that writing -1 to memory.kmem.tcp.limit_in_bytes just
enables socket accounting w/o limiting it, which might be useful by
itself, but it isn't true.
Since this API peculiarity is not documented anywhere, I propose to drop
it. This will allow to simplify the code by dropping cg_proto->flags.
Signed-off-by: Vladimir Davydov <vdavydov@virtuozzo.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Conflicts:
drivers/net/bonding/bond_main.c
drivers/net/ethernet/mellanox/mlxsw/spectrum.h
drivers/net/ethernet/mellanox/mlxsw/spectrum_switchdev.c
The bond_main.c and mellanox switch conflicts were cases of
overlapping changes.
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit acf8dd0a9d ("udp: only allow UFO for packets from SOCK_DGRAM
sockets") disallows UFO for packets sent from raw sockets. We need to do
the same also for SOCK_DGRAM sockets with SO_NO_CHECK options, even if
for a bit different reason: while such socket would override the
CHECKSUM_PARTIAL set by ip_ufo_append_data(), gso_size is still set and
bad offloading flags warning is triggered in __skb_gso_segment().
In the IPv6 case, SO_NO_CHECK option is ignored but we need to disallow
UFO for packets sent by sockets with UDP_NO_CHECK6_TX option.
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Tested-by: Shannon Nelson <shannon.nelson@intel.com>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
For tcp_yeah, use an ssthresh floor of 2, the same floor used by Reno
and CUBIC, per RFC 5681 (equation 4).
tcp_yeah_ssthresh() was sometimes returning a 0 or negative ssthresh
value if the intended reduction is as big or bigger than the current
cwnd. Congestion control modules should never return a zero or
negative ssthresh. A zero ssthresh generally results in a zero cwnd,
causing the connection to stall. A negative ssthresh value will be
interpreted as a u32 and will set a target cwnd for PRR near 4
billion.
Oleksandr Natalenko reported that a system using tcp_yeah with ECN
could see a warning about a prior_cwnd of 0 in
tcp_cwnd_reduction(). Testing verified that this was due to
tcp_yeah_ssthresh() misbehaving in this way.
Reported-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is the final part required to namespaceify the tcp
keep alive mechanism.
Signed-off-by: Nikolay Borisov <kernel@kyup.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is required to have full tcp keepalive mechanism namespace
support.
Signed-off-by: Nikolay Borisov <kernel@kyup.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Different net namespaces might have different requirements as to
the keepalive time of tcp sockets. This might be required in cases
where different firewall rules are in place which require tcp
timeout sockets to be increased/decreased independently of the host.
Signed-off-by: Nikolay Borisov <kernel@kyup.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
udp tunnel offloads tend to aggregate datagrams based on inner
headers. gro engine gets notified by tunnel implementations about
possible offloads. The match is solely based on the port number.
Imagine a tunnel bound to port 53, the offloading will look into all
DNS packets and tries to aggregate them based on the inner data found
within. This could lead to data corruption and malformed DNS packets.
While this patch minimizes the problem and helps an administrator to find
the issue by querying ip tunnel/fou, a better way would be to match on
the specific destination ip address so if a user space socket is bound
to the same address it will conflict.
Cc: Tom Herbert <tom@herbertland.com>
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add lock release/acquire annotations to ping_seq_start() and
ping_seq_stop() to satisfy sparse.
Signed-off-by: Lance Richardson <lrichard@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eliminate endianness mismatch warnings (reported by sparse) in this file by
using appropriate nla_put_*()/nla_get_*() calls.
Signed-off-by: Lance Richardson <lrichard@redhat.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, they are:
1) Release nf_tables objects on netns destructions via
nft_release_afinfo().
2) Destroy basechain and rules on netdevice removal in the new netdev
family.
3) Get rid of defensive check against removal of inactive objects in
nf_tables.
4) Pass down netns pointer to our existing nfnetlink callbacks, as well
as commit() and abort() nfnetlink callbacks.
5) Allow to invert limit expression in nf_tables, so we can throttle
overlimit traffic.
6) Add packet duplication for the netdev family.
7) Add forward expression for the netdev family.
8) Define pr_fmt() in conntrack helpers.
9) Don't leave nfqueue configuration on inconsistent state in case of
errors, from Ken-ichirou MATSUZAWA, follow up patches are also from
him.
10) Skip queue option handling after unbind.
11) Return error on unknown both in nfqueue and nflog command.
12) Autoload ctnetlink when NFQA_CFG_F_CONNTRACK is set.
13) Add new NFTA_SET_USERDATA attribute to store user data in sets,
from Carlos Falgueras.
14) Add support for 64 bit byteordering changes nf_tables, from Florian
Westphal.
15) Add conntrack byte/packet counter matching support to nf_tables,
also from Florian.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Patch 3759824da8 ("tcp: PRR uses CRB mode by default and SS mode
conditionally") introduced a bug that cwnd may become 0 when both
inflight and sndcnt are 0 (cwnd = inflight + sndcnt). This may lead
to a div-by-zero if the connection starts another cwnd reduction
phase by setting tp->prior_cwnd to the current cwnd (0) in
tcp_init_cwnd_reduction().
To prevent this we skip PRR operation when nothing is acked or
sacked. Then cwnd must be positive in all cases as long as ssthresh
is positive:
1) The proportional reduction mode
inflight > ssthresh > 0
2) The reduction bound mode
a) inflight == ssthresh > 0
b) inflight < ssthresh
sndcnt > 0 since newly_acked_sacked > 0 and inflight < ssthresh
Therefore in all cases inflight and sndcnt can not both be 0.
We check invalid tp->prior_cwnd to avoid potential div0 bugs.
In reality this bug is triggered only with a sequence of less common
events. For example, the connection is terminating an ECN-triggered
cwnd reduction with an inflight 0, then it receives reordered/old
ACKs or DSACKs from prior transmission (which acks nothing). Or the
connection is in fast recovery stage that marks everything lost,
but fails to retransmit due to local issues, then receives data
packets from other end which acks nothing.
Fixes: 3759824da8 ("tcp: PRR uses CRB mode by default and SS mode conditionally")
Reported-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This socket-lookup path did not pass along the skb in question
in my original BPF-based socket selection patch. The skb in the
udpN_lib_lookup2 path can be used for BPF-based socket selection just
like it is in the 'traditional' udpN_lib_lookup path.
udpN_lib_lookup2 kicks in when there are greater than 10 sockets in
the same hlist slot. Coincidentally, I chose 10 sockets per
reuseport group in my functional test, so the lookup2 path was not
excersised. This adds an additional set of tests with 20 sockets.
Fixes: 538950a1b7 ("soreuseport: setsockopt SO_ATTACH_REUSEPORT_[CE]BPF")
Fixes: 3ca8e40299 ("soreuseport: BPF selection functional test")
Suggested-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Craig Gallek <kraig@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The only user was removed in commit
029f7f3b87 ("netfilter: ipv6: nf_defrag: avoid/free clone operations").
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commands run in a vrf context are not failing as expected on a route lookup:
root@kenny:~# ip ro ls table vrf-red
unreachable default
root@kenny:~# ping -I vrf-red -c1 -w1 10.100.1.254
ping: Warning: source address might be selected on device other than vrf-red.
PING 10.100.1.254 (10.100.1.254) from 0.0.0.0 vrf-red: 56(84) bytes of data.
--- 10.100.1.254 ping statistics ---
2 packets transmitted, 0 received, 100% packet loss, time 999ms
Since the vrf table does not have a route for 10.100.1.254 the ping
should have failed. The saddr lookup causes a full VRF table lookup.
Propogating a lookup failure to the user allows the command to fail as
expected:
root@kenny:~# ping -I vrf-red -c1 -w1 10.100.1.254
connect: No route to host
Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Expose socket options for setting a classic or extended BPF program
for use when selecting sockets in an SO_REUSEPORT group. These options
can be used on the first socket to belong to a group before bind or
on any socket in the group after bind.
This change includes refactoring of the existing sk_filter code to
allow reuse of the existing BPF filter validation checks.
Signed-off-by: Craig Gallek <kraig@google.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Include a struct sock_reuseport instance when a UDP socket binds to
a specific address for the first time with the reuseport flag set.
When selecting a socket for an incoming UDP packet, use the information
available in sock_reuseport if present.
This required adding an additional field to the UDP source address
equality function to differentiate between exact and wildcard matches.
The original use case allowed wildcard matches when checking for
existing port uses during bind. The new use case of adding a socket
to a reuseport group requires exact address matching.
Performance test (using a machine with 2 CPU sockets and a total of
48 cores): Create reuseport groups of varying size. Use one socket
from this group per user thread (pinning each thread to a different
core) calling recvmmsg in a tight loop. Record number of messages
received per second while saturating a 10G link.
10 sockets: 18% increase (~2.8M -> 3.3M pkts/s)
20 sockets: 14% increase (~2.9M -> 3.3M pkts/s)
40 sockets: 13% increase (~3.0M -> 3.4M pkts/s)
This work is based off a similar implementation written by
Ying Cai <ycai@google.com> for implementing policy-based reuseport
selection.
Signed-off-by: Craig Gallek <kraig@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Backport of this upstream commit into stable kernels :
89c22d8c3b ("net: Fix skb csum races when peeking")
exposed a bug in udp stack vs MSG_PEEK support, when user provides
a buffer smaller than skb payload.
In this case,
skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr),
msg->msg_iov);
returns -EFAULT.
This bug does not happen in upstream kernels since Al Viro did a great
job to replace this into :
skb_copy_and_csum_datagram_msg(skb, sizeof(struct udphdr), msg);
This variant is safe vs short buffers.
For the time being, instead reverting Herbert Xu patch and add back
skb->ip_summed invalid changes, simply store the result of
udp_lib_checksum_complete() so that we avoid computing the checksum a
second time, and avoid the problematic
skb_copy_and_csum_datagram_iovec() call.
This patch can be applied on recent kernels as it avoids a double
checksumming, then backported to stable kernels as a bug fix.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
We have to release the existing objects on netns removal otherwise we
leak them. Chains are unregistered in first place to make sure no
packets are walking on our rules and sets anymore.
The object release happens by when we unregister the family via
nft_release_afinfo() which is called from nft_unregister_afinfo() from
the corresponding __net_exit path in every family.
Reported-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
By moving stats update into iptunnel_xmit(), we can simplify
iptunnel_xmit() usage. With this change there is no need to
call another function (iptunnel_xmit_stats()) to update stats
in tunnel xmit code path.
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Hannes points out that when we generate tcp reset for timewait sockets we
pretend we found no socket and pass NULL sk to tcp_vX_send_reset().
Make it cope with inet tw sockets and then provide tw sk.
This makes RSTs appear on correct interface when SO_BINDTODEVICE is used.
Packetdrill test case:
// want default route to be used, we rely on BINDTODEVICE
`ip route del 192.0.2.0/24 via 192.168.0.2 dev tun0`
0.000 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
// test case still works due to BINDTODEVICE
0.001 setsockopt(3, SOL_SOCKET, SO_BINDTODEVICE, "tun0", 4) = 0
0.100...0.200 connect(3, ..., ...) = 0
0.100 > S 0:0(0) <mss 1460,sackOK,nop,nop>
0.200 < S. 0:0(0) ack 1 win 32792 <mss 1460,sackOK,nop,nop>
0.200 > . 1:1(0) ack 1
0.210 close(3) = 0
0.210 > F. 1:1(0) ack 1 win 29200
0.300 < . 1:1(0) ack 2 win 46
// more data while in FIN_WAIT2, expect RST
1.300 < P. 1:1001(1000) ack 1 win 46
// fails without this change -- default route is used
1.301 > R 1:1(0) win 0
Reported-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: Florian Westphal <fw@strlen.de>
Acked-by: Eric Dumazet <edumazet@google.com>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
tcp_md5_do_lookup requires a full socket, so once we extend
_send_reset() to also accept timewait socket we would have to change
if (!sk && hash_location)
to something like
if ((!sk || !sk_fullsock(sk)) && hash_location) {
...
} else {
(sk && sk_fullsock(sk)) tcp_md5_do_lookup()
}
Switch the two branches: check if we have a socket first, then
fall back to a listener lookup if we saw a md5 option (hash_location).
Signed-off-by: Florian Westphal <fw@strlen.de>
Acked-by: Eric Dumazet <edumazet@google.com>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Steffen Klassert says:
====================
pull request (net): ipsec 2015-12-22
Just one patch to fix dst_entries_init with multiple namespaces.
From Dan Streetman.
Please pull or let me know if there are problems.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
When closing a listen socket, tcp_abort currently calls
tcp_done without clearing the request queue. If the socket has a
child socket that is established but not yet accepted, the child
socket is then left without a parent, causing a leak.
Fix this by setting the socket state to TCP_CLOSE and calling
inet_csk_listen_stop with the socket lock held, like tcp_close
does.
Tested using net_test. With this patch, calling SOCK_DESTROY on a
listen socket that has an established but not yet accepted child
socket results in the parent and the child being closed, such
that they no longer appear in sock_diag dumps.
Reported-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Lorenzo Colitti <lorenzo@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
IP-TTL case is already handled in ip_tunnel_ioctl() API.
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Adding support for SYN_RECV request sockets to tcp_abort()
is quite easy after our tcp listener rewrite.
Note that we also need to better handle listeners, or we might
leak not yet accepted children, because of a missing
inet_csk_listen_stop() call.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Lorenzo Colitti <lorenzo@google.com>
Tested-by: Lorenzo Colitti <lorenzo@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pablo Neira Ayuso says:
====================
Netfilter updates for net-next
The following patchset contains the first batch of Netfilter updates for
the upcoming 4.5 kernel. This batch contains userspace netfilter header
compilation fixes, support for packet mangling in nf_tables, the new
tracing infrastructure for nf_tables and cgroup2 support for iptables.
More specifically, they are:
1) Two patches to include dependencies in our netfilter userspace
headers to resolve compilation problems, from Mikko Rapeli.
2) Four comestic cleanup patches for the ebtables codebase, from Ian Morris.
3) Remove duplicate include in the netfilter reject infrastructure,
from Stephen Hemminger.
4) Two patches to simplify the netfilter defragmentation code for IPv6,
patch from Florian Westphal.
5) Fix root ownership of /proc/net netfilter for unpriviledged net
namespaces, from Philip Whineray.
6) Get rid of unused fields in struct nft_pktinfo, from Florian Westphal.
7) Add mangling support to our nf_tables payload expression, from
Patrick McHardy.
8) Introduce a new netlink-based tracing infrastructure for nf_tables,
from Florian Westphal.
9) Change setter functions in nfnetlink_log to be void, from
Rami Rosen.
10) Add netns support to the cttimeout infrastructure.
11) Add cgroup2 support to iptables, from Tejun Heo.
12) Introduce nfnl_dereference_protected() in nfnetlink, from Florian.
13) Add support for mangling pkttype in the nf_tables meta expression,
also from Florian.
BTW, I need that you pull net into net-next, I have another batch that
requires changes that I don't yet see in net.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Allow accepted sockets to derive their sk_bound_dev_if setting from the
l3mdev domain in which the packets originated. A sysctl setting is added
to control the behavior which is similar to sk_mark and
sysctl_tcp_fwmark_accept.
This effectively allow a process to have a "VRF-global" listen socket,
with child sockets bound to the VRF device in which the packet originated.
A similar behavior can be achieved using sk_mark, but a solution using marks
is incomplete as it does not handle duplicate addresses in different L3
domains/VRFs. Allowing sockets to inherit the sk_bound_dev_if from l3mdev
domain provides a complete solution.
Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Conflicts:
drivers/net/geneve.c
Here we had an overlapping change, where in 'net' the extraneous stats
bump was being removed whilst in 'net-next' the final argument to
udp_tunnel6_xmit_skb() was being changed.
Signed-off-by: David S. Miller <davem@davemloft.net>
Yuchung tracked a regression caused by commit 57be5bdad7 ("ip: convert
tcp_sendmsg() to iov_iter primitives") for TCP Fast Open.
Some Fast Open users do not actually add any data in the SYN packet.
Fixes: 57be5bdad7 ("ip: convert tcp_sendmsg() to iov_iter primitives")
Reported-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Acked-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
fou->udp_offloads is managed by RCU. As it is actually included inside
the fou sockets, we cannot let the memory go out of scope before a grace
period. We either can synchronize_rcu or switch over to kfree_rcu to
manage the sockets. kfree_rcu seems appropriate as it is used by vxlan
and geneve.
Fixes: 23461551c0 ("fou: Support for foo-over-udp RX path")
Cc: Tom Herbert <tom@herbertland.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This implements SOCK_DESTROY for TCP sockets. It causes all
blocking calls on the socket to fail fast with ECONNABORTED and
causes a protocol close of the socket. It informs the other end
of the connection by sending a RST, i.e., initiating a TCP ABORT
as per RFC 793. ECONNABORTED was chosen for consistency with
FreeBSD.
Signed-off-by: Lorenzo Colitti <lorenzo@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This passes the SOCK_DESTROY operation to the underlying protocol
diag handler, or returns -EOPNOTSUPP if that handler does not
define a destroy operation.
Most of this patch is just renaming functions. This is not
strictly necessary, but it would be fairly counterintuitive to
have the code to destroy inet sockets be in a function whose name
starts with inet_diag_get.
Signed-off-by: Lorenzo Colitti <lorenzo@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, inet_diag_dump_one_icsk finds a socket and then dumps
its information to userspace. Split it into a part that finds the
socket and a part that dumps the information.
Signed-off-by: Lorenzo Colitti <lorenzo@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In tcp_send_sendpage and tcp_sendmsg we check the route capabilities to
determine if checksum offload can be performed. This check currently
does not take the IP protocol into account for devices that advertise
only one of NETIF_F_IPV6_CSUM or NETIF_F_IP_CSUM. This patch adds a
function to check capabilities for checksum offload with a socket
called sk_check_csum_caps. This function checks for specific IPv4 or
IPv6 offload support based on the family of the socket.
Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
These netif flags are unnecessary convolutions. It is more
straightforward to just use NETIF_F_HW_CSUM, NETIF_F_IP_CSUM,
and NETIF_F_IPV6_CSUM directly.
This patch also:
- Cleans up can_checksum_protocol
- Simplifies netdev_intersect_features
Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The name NETIF_F_ALL_CSUM is a misnomer. This does not correspond to the
set of features for offloading all checksums. This is a mask of the
checksum offload related features bits. It is incorrect to set both
NETIF_F_HW_CSUM and NETIF_F_IP_CSUM or NETIF_F_IPV6 at the same time for
features of a device.
This patch:
- Changes instances of NETIF_F_ALL_CSUM to NETIF_F_CSUM_MASK (where
NETIF_F_ALL_CSUM is being used as a mask).
- Changes bonding, sfc/efx, ipvlan, macvlan, vlan, and team drivers to
use NEITF_F_HW_CSUM in features list instead of NETIF_F_ALL_CSUM.
Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
David Wilder reported crashes caused by dst reuse.
<quote David>
I am seeing a crash on a distro V4.2.3 kernel caused by a double
release of a dst_entry. In ipv4_dst_destroy() the call to
list_empty() finds a poisoned next pointer, indicating the dst_entry
has already been removed from the list and freed. The crash occurs
18 to 24 hours into a run of a network stress exerciser.
</quote>
Thanks to his detailed report and analysis, we were able to understand
the core issue.
IP early demux can associate a dst to skb, after a lookup in TCP/UDP
sockets.
When socket cache is not properly set, we want to store into
sk->sk_dst_cache the dst for future IP early demux lookups,
by acquiring a stable refcount on the dst.
Problem is this acquisition is simply using an atomic_inc(),
which works well, unless the dst was queued for destruction from
dst_release() noticing dst refcount went to zero, if DST_NOCACHE
was set on dst.
We need to make sure current refcount is not zero before incrementing
it, or risk double free as David reported.
This patch, being a stable candidate, adds two new helpers, and use
them only from IP early demux problematic paths.
It might be possible to merge in net-next skb_dst_force() and
skb_dst_force_safe(), but I prefer having the smallest patch for stable
kernels : Maybe some skb_dst_force() callers do not expect skb->dst
can suddenly be cleared.
Can probably be backported back to linux-3.6 kernels
Reported-by: David J. Wilder <dwilder@us.ibm.com>
Tested-by: David J. Wilder <dwilder@us.ibm.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
郭永刚 reported that one could simply crash the kernel as root by
using a simple program:
int socket_fd;
struct sockaddr_in addr;
addr.sin_port = 0;
addr.sin_addr.s_addr = INADDR_ANY;
addr.sin_family = 10;
socket_fd = socket(10,3,0x40000000);
connect(socket_fd , &addr,16);
AF_INET, AF_INET6 sockets actually only support 8-bit protocol
identifiers. inet_sock's skc_protocol field thus is sized accordingly,
thus larger protocol identifiers simply cut off the higher bits and
store a zero in the protocol fields.
This could lead to e.g. NULL function pointer because as a result of
the cut off inet_num is zero and we call down to inet_autobind, which
is NULL for raw sockets.
kernel: Call Trace:
kernel: [<ffffffff816db90e>] ? inet_autobind+0x2e/0x70
kernel: [<ffffffff816db9a4>] inet_dgram_connect+0x54/0x80
kernel: [<ffffffff81645069>] SYSC_connect+0xd9/0x110
kernel: [<ffffffff810ac51b>] ? ptrace_notify+0x5b/0x80
kernel: [<ffffffff810236d8>] ? syscall_trace_enter_phase2+0x108/0x200
kernel: [<ffffffff81645e0e>] SyS_connect+0xe/0x10
kernel: [<ffffffff81779515>] tracesys_phase2+0x84/0x89
I found no particular commit which introduced this problem.
CVE: CVE-2015-8543
Cc: Cong Wang <cwang@twopensource.com>
Reported-by: 郭永刚 <guoyonggang@360.cn>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Resolve conflict between commit 264640fc2c ("ipv6: distinguish frag
queues by device for multicast and link-local packets") from the net
tree and commit 029f7f3b87 ("netfilter: ipv6: nf_defrag: avoid/free
clone operations") from the nf-next tree.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Conflicts:
net/ipv6/netfilter/nf_conntrack_reasm.c
Pablo Neira Ayuso says:
====================
netfilter fixes for net
The following patchset contains Netfilter fixes for you net tree,
specifically for nf_tables and nfnetlink_queue, they are:
1) Avoid a compilation warning in nfnetlink_queue that was introduced
in the previous merge window with the simplification of the conntrack
integration, from Arnd Bergmann.
2) nfnetlink_queue is leaking the pernet subsystem registration from
a failure path, patch from Nikolay Borisov.
3) Pass down netns pointer to batch callback in nfnetlink, this is the
largest patch and it is not a bugfix but it is a dependency to
resolve a splat in the correct way.
4) Fix a splat due to incorrect socket memory accounting with nfnetlink
skbuff clones.
5) Add missing conntrack dependencies to NFT_DUP_IPV4 and NFT_DUP_IPV6.
6) Traverse the nftables commit list in reverse order from the commit
path, otherwise we crash when the user applies an incremental update
via 'nft -f' that deletes an object that was just introduced in this
batch, from Xin Long.
Regarding the compilation warning fix, many people have sent us (and
keep sending us) patches to address this, that's why I'm including this
batch even if this is not critical.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
The VRF driver cycles netdevs when an interface is enslaved or released:
the down event is used to flush neighbor and route tables and the up
event (if the interface was already up) effectively moves local and
connected routes to the proper table.
As of 4f823defdd the local route is left hanging around after a link
down, so when a netdev is moved from one VRF to another (or released
from a VRF altogether) local routes are left in the wrong table.
Fix by handling the NETDEV_CHANGEUPPER event. When the upper dev is
an L3mdev then call fib_disable_ip to flush all routes, local ones
to.
Fixes: 4f823defdd ("ipv4: fix to not remove local route on link down")
Cc: Julian Anastasov <ja@ssi.bg>
Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
CONFIG_NF_CONNTRACK=m
CONFIG_NF_DUP_IPV4=y
results in:
net/built-in.o: In function `nf_dup_ipv4':
>> (.text+0xd434f): undefined reference to `nf_conntrack_untracked'
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Conflicts:
drivers/net/ethernet/renesas/ravb_main.c
kernel/bpf/syscall.c
net/ipv4/ipmr.c
All three conflicts were cases of overlapping changes.
Signed-off-by: David S. Miller <davem@davemloft.net>
When a multicast group is joined on a socket, a struct ip_mc_socklist
is appended to the sockets mc_list containing information about the
joined group.
If the interface is hot unplugged, this entry becomes stale. Prior to
commit 52ad353a53 ("igmp: fix the problem when mc leave group") it
was possible to remove the stale entry by performing a
IP_DROP_MEMBERSHIP, passing either the old ifindex or ip address on
the interface. However, this fix enforces that the interface must
still exist. Thus with time, the number of stale entries grows, until
sysctl_igmp_max_memberships is reached and then it is not possible to
join and more groups.
The previous patch fixes an issue where a IP_DROP_MEMBERSHIP is
performed without specifying the interface, either by ifindex or ip
address. However here we do supply one of these. So loosen the
restriction on device existence to only apply when the interface has
not been specified. This then restores the ability to clean up the
stale entries.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Fixes: 52ad353a53 "(igmp: fix the problem when mc leave group")
Signed-off-by: David S. Miller <davem@davemloft.net>
If tcp_send_ack() can not allocate skb, we properly handle this
and setup a timer to try later.
Use __GFP_NOWARN to avoid polluting syslog in the case host is
under memory pressure, so that pertinent messages are not lost under
a flood of useless information.
sk_gfp_atomic() can use its gfp_mask argument (all callers currently
were using GFP_ATOMIC before this patch)
We rename sk_gfp_atomic() to sk_gfp_mask() to clearly express this
function now takes into account its second argument (gfp_mask)
Note that when tcp_transmit_skb() is called with clone_it set to false,
we do not attempt memory allocations, so can pass a 0 gfp_mask, which
most compilers can emit faster than a non zero or constant value.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch is a cleanup to make following patch easier to
review.
Goal is to move SOCK_ASYNC_NOSPACE and SOCK_ASYNC_WAITDATA
from (struct socket)->flags to a (struct socket_wq)->flags
to benefit from RCU protection in sock_wake_async()
To ease backports, we rename both constants.
Two new helpers, sk_set_bit(int nr, struct sock *sk)
and sk_clear_bit(int net, struct sock *sk) are added so that
following patch can change their implementation.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Dmitry provided a syzkaller (http://github.com/google/syzkaller)
generated program that triggers the WARNING at
net/ipv4/tcp.c:1729 in tcp_recvmsg() :
WARN_ON(tp->copied_seq != tp->rcv_nxt &&
!(flags & (MSG_PEEK | MSG_TRUNC)));
His program is specifically attempting a Cross SYN TCP exchange,
that we support (for the pleasure of hackers ?), but it looks we
lack proper tcp->copied_seq initialization.
Thanks again Dmitry for your report and testings.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Tested-by: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds support to add and remove MFC entries. It uses the
same attributes like the already present dump support in order to be
consistent. There's one new entry - RTA_PREFSRC, it's used to denote an
MFC_PROXY entry (see MRT_ADD_MFC vs MRT_ADD_MFC_PROXY).
The already existing infrastructure is used to create and delete the
entries, the netlink message gets converted internally to a struct mfcctl
which is used with ipmr_mfc_add/delete.
The other used attributes are:
RTA_IIF - used for mfcc_parent (when adding it's required to be valid)
RTA_SRC - used for mfcc_origin
RTA_DST - used for mfcc_mcastgrp
RTA_TABLE - the MRT table id
RTA_MULTIPATH - the "oifs" ttl array
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We can have both errors and we'll return the second one, fix it to
return an error at a time as it's normal. I've overlooked this in my
previous set.
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move the inline pimsm_enabled() to pim.h and rename it to
ipmr_pimsm_enabled to show it's for the ipv4 ipmr code since pim.h is
used by IPv6 too.
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move the definitions of VIF_EXISTS() and struct mr_table to mroute.h
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
MFC_NOTIFY was introduced in kernel 2.1.68 but afaik it hasn't been used
and I couldn't find any users currently so just remove it. Only
MFC_STATIC is left, so move it into an enum, add a description and use
BIT().
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It looks like many files are including mroute.h unnecessarily, so remove
the include. Most importantly remove it from ipv6.
CC: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
CC: Steffen Klassert <steffen.klassert@secunet.com>
CC: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since (at least) commit b17a7c179d ("[NET]: Do sysfs registration as
part of register_netdevice."), netdev_run_todo() deals only with
unregistration, so we don't need to do the rtnl_unlock/lock cycle to
finish registration when failing pimreg or dvmrp device creation. In
fact that opens a race condition where someone can delete the device
while rtnl is unlocked because it's fully registered. The problem gets
worse when netlink support is introduced as there are more points of entry
that can cause it and it also makes reusing that code correctly impossible.
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Reviewed-by: Cong Wang <cwang@twopensource.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit 09605cc12c ("net ipv4: use preferred log methods") replaced
a few calls of pr_cont() after a console print without a trailing
newline by pr_info(), causing lines to be split during IP
autoconfiguration, like:
.
,
OK
IP-Config: Got DHCP answer from 192.168.97.254,
my address is 192.168.97.44
Convert these back to using pr_cont(), so it prints again:
., OK
IP-Config: Got DHCP answer from 192.168.97.254, my address is 192.168.97.44
Absorb the printing of "my address ..." into the previous call to
pr_info(), as there's no reason to use a continuation there.
Convert one more pr_info() to print nameservers while we're at it.
Fixes: 09605cc12c ("net ipv4: use preferred log methods")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
Factor out common vif init code used in both tunnel and pimreg
initialization and create ipmr_init_vif_indev() function.
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Take rtnl in the beginning unconditionally as most options already need
it (one exception - MRT_DONE, see the comment inside), make the
lock/unlock places central and move out the switch() local variables.
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It's not necessary to check for null as SLAB_PANIC is used and we'll
panic if the alloc fails, so just drop it.
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Trivial replace of ifdef with IS_BUILTIN().
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use a switch to determine if optname is correct and set val accordingly.
This produces a much more straight-forward and readable code.
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Trivial code and comment style fixes, also removed some extra newlines,
spaces and tabs.
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add the helper pimsm_enabled() which replaces the old CONFIG_IP_PIMSM
define and is used to check if any version of PIM-SM has been enabled.
Use a single if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
for the pim-sm shared code. This is okay w.r.t IGMPMSG_WHOLEPKT because
only a VIFF_REGISTER device can send such packet, and it can't be
created if pimsm_enabled() is false.
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Before mroute_reg_vif_num was defined only if any of the CONFIG_PIMSM_
options were set, but that's not really necessary as the size of the
struct is the same in both cases (checked with pahole, both cases size
is 3256 bytes) and we can remove some unnecessary ifdefs to simplify the
code.
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move the table id check in ipmr_new_table and make it return error
pointer. We need this change for the upcoming netlink table manipulation
support in order to avoid code duplication and a race condition.
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When destroying an mrt table the static mfc entries and the static
devices are kept, which leads to devices that can never be destroyed
(because of refcnt taken) and leaked memory, for example:
unreferenced object 0xffff880034c144c0 (size 192):
comm "mfc-broken", pid 4777, jiffies 4320349055 (age 46001.964s)
hex dump (first 32 bytes):
98 53 f0 34 00 88 ff ff 98 53 f0 34 00 88 ff ff .S.4.....S.4....
ef 0a 0a 14 01 02 03 04 00 00 00 00 01 00 00 00 ................
backtrace:
[<ffffffff815c1b9e>] kmemleak_alloc+0x4e/0xb0
[<ffffffff811ea6e0>] kmem_cache_alloc+0x190/0x300
[<ffffffff815931cb>] ip_mroute_setsockopt+0x5cb/0x910
[<ffffffff8153d575>] do_ip_setsockopt.isra.11+0x105/0xff0
[<ffffffff8153e490>] ip_setsockopt+0x30/0xa0
[<ffffffff81564e13>] raw_setsockopt+0x33/0x90
[<ffffffff814d1e14>] sock_common_setsockopt+0x14/0x20
[<ffffffff814d0b51>] SyS_setsockopt+0x71/0xc0
[<ffffffff815cdbf6>] entry_SYSCALL_64_fastpath+0x16/0x7a
[<ffffffffffffffff>] 0xffffffffffffffff
Make sure that everything is cleaned on netns destruction.
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Reviewed-by: Cong Wang <cwang@twopensource.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
tcp_send_rcvq() is used for re-injecting data into tcp receive queue.
Problems :
- No check against size is performed, allowed user to fool kernel in
attempting very large memory allocations, eventually triggering
OOM when memory is fragmented.
- In case of fault during the copy we do not return correct errno.
Lets use alloc_skb_with_frags() to cook optimal skbs.
Fixes: 292e8d8c85 ("tcp: Move rcvq sending to tcp_input.c")
Fixes: c0e88ff0f2 ("tcp: Repair socket queues")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Pavel Emelyanov <xemul@parallels.com>
Acked-by: Pavel Emelyanov <xemul@parallels.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix incrementing TCPFastOpenActiveFailed snmp stats multiple times
when the handshake experiences multiple SYN timeouts.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some middle-boxes black-hole the data after the Fast Open handshake
(https://www.ietf.org/proceedings/94/slides/slides-94-tcpm-13.pdf).
The exact reason is unknown. The work-around is to disable Fast Open
temporarily after multiple recurring timeouts with few or no data
delivered in the established state.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Christoph Paasch <cpaasch@apple.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When a passive TCP is created, we eventually call tcp_md5_do_add()
with sk pointing to the child. It is not owner by the user yet (we
will add this socket into listener accept queue a bit later anyway)
But we do own the spinlock, so amend the lockdep annotation to avoid
following splat :
[ 8451.090932] net/ipv4/tcp_ipv4.c:923 suspicious rcu_dereference_protected() usage!
[ 8451.090932]
[ 8451.090932] other info that might help us debug this:
[ 8451.090932]
[ 8451.090934]
[ 8451.090934] rcu_scheduler_active = 1, debug_locks = 1
[ 8451.090936] 3 locks held by socket_sockopt_/214795:
[ 8451.090936] #0: (rcu_read_lock){.+.+..}, at: [<ffffffff855c6ac1>] __netif_receive_skb_core+0x151/0xe90
[ 8451.090947] #1: (rcu_read_lock){.+.+..}, at: [<ffffffff85618143>] ip_local_deliver_finish+0x43/0x2b0
[ 8451.090952] #2: (slock-AF_INET){+.-...}, at: [<ffffffff855acda5>] sk_clone_lock+0x1c5/0x500
[ 8451.090958]
[ 8451.090958] stack backtrace:
[ 8451.090960] CPU: 7 PID: 214795 Comm: socket_sockopt_
[ 8451.091215] Call Trace:
[ 8451.091216] <IRQ> [<ffffffff856fb29c>] dump_stack+0x55/0x76
[ 8451.091229] [<ffffffff85123b5b>] lockdep_rcu_suspicious+0xeb/0x110
[ 8451.091235] [<ffffffff8564544f>] tcp_md5_do_add+0x1bf/0x1e0
[ 8451.091239] [<ffffffff85645751>] tcp_v4_syn_recv_sock+0x1f1/0x4c0
[ 8451.091242] [<ffffffff85642b27>] ? tcp_v4_md5_hash_skb+0x167/0x190
[ 8451.091246] [<ffffffff85647c78>] tcp_check_req+0x3c8/0x500
[ 8451.091249] [<ffffffff856451ae>] ? tcp_v4_inbound_md5_hash+0x11e/0x190
[ 8451.091253] [<ffffffff85647170>] tcp_v4_rcv+0x3c0/0x9f0
[ 8451.091256] [<ffffffff85618143>] ? ip_local_deliver_finish+0x43/0x2b0
[ 8451.091260] [<ffffffff856181b6>] ip_local_deliver_finish+0xb6/0x2b0
[ 8451.091263] [<ffffffff85618143>] ? ip_local_deliver_finish+0x43/0x2b0
[ 8451.091267] [<ffffffff85618d38>] ip_local_deliver+0x48/0x80
[ 8451.091270] [<ffffffff85618510>] ip_rcv_finish+0x160/0x700
[ 8451.091273] [<ffffffff8561900e>] ip_rcv+0x29e/0x3d0
[ 8451.091277] [<ffffffff855c74b7>] __netif_receive_skb_core+0xb47/0xe90
Fixes: a8afca0329 ("tcp: md5: protects md5sig_info with RCU")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Replace printk calls with preferred unconditional log method calls to keep
kernel messages clean.
Added newline to "too small MTU" message.
Signed-off-by: Bastian Stender <bst@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Sending ICMP packets with raw sockets ends up in the SNMP counters
logging the type as the first byte of the IPv4 header rather than
the ICMP header. This is fixed by adding the IP Header Length to
the casting into a icmphdr struct.
Signed-off-by: Ben Cartwright-Cox <ben@benjojo.co.uk>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some functions access TCP sockets without holding a lock and
might output non consistent data, depending on compiler and or
architecture.
tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ...
Introduce sk_state_load() and sk_state_store() to fix the issues,
and more clearly document where this lack of locking is happening.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The uninitialized tuple structure caused incorrect hash calculation
and the lookup failed.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=106441
Signed-off-by: Anthony Lineham <anthony.lineham@alliedtelesis.co.nz>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
I mistakenly took wrong request sock pointer when calling tcp_move_syn()
@req_unhash is either a copy of @req, or a NULL value for
FastOpen connexions (as we do not expect to unhash the temporary
request sock from ehash table)
Fixes: 805c4bc057 ("tcp: fix req->saved_syn race")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Ying Cai <ycai@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Before converting a 'socket pointer' into inet socket,
use sk_fullsock() to detect timewait or request sockets.
Fixes: ca6fb06518 ("tcp: attach SYNACK messages to request sockets instead of listener")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Tested-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
For the reasons explained in commit ce1050089c ("tcp/dccp: fix
ireq->pktopts race"), we need to make sure we do not access
req->saved_syn unless we own the request sock.
This fixes races for listeners using TCP_SAVE_SYN option.
Fixes: e994b2f0fb ("tcp: do not lock listener to process SYN packets")
Fixes: 079096f103 ("tcp/dccp: install syn_recv requests into ehash table")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Ying Cai <ycai@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A bug report (https://bugzilla.kernel.org/show_bug.cgi?id=107071) noted
that the follwoing ip command is failing with v4.3:
$ ip route add 10.248.5.0/24 dev bond0.250 table vlan_250 src 10.248.5.154
RTNETLINK answers: Invalid argument
021dd3b8a1 changed the lookup of the given preferred source address to
use the table id passed in, but this assumes the local entries are in the
given table which is not necessarily true for non-VRF use cases. When
validating the preferred source fallback to the local table on failure.
Fixes: 021dd3b8a1 ("net: Add routes to the table associated with the device")
Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Sasha reported the following lockdep warning:
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(sk_lock-AF_INET);
lock(rtnl_mutex);
lock(sk_lock-AF_INET);
lock(rtnl_mutex);
This is due to that for IP_MSFILTER and MCAST_MSFILTER, we take
rtnl lock before the socket lock in setsockopt() path, but take
the socket lock before rtnl lock in getsockopt() path. All the
rest optnames are setsockopt()-only.
Fix this by aligning the getsockopt() path with the setsockopt()
path, so that all mcast socket path would be locked in the same
order.
Note, IPv6 part is different where rtnl lock is not held.
Fixes: 54ff9ef36b ("ipv4, ipv6: kill ip_mc_{join, leave}_group and ipv6_sock_mc_{join, drop}")
Reported-by: Sasha Levin <sasha.levin@oracle.com>
Cc: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Reviewed-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Minor overlapping changes in net/ipv4/ipmr.c, in 'net' we were
fixing the "BH-ness" of the counter bumps whilst in 'net-next'
the functions were modified to take an explicit 'net' parameter.
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove the dst_entries_init/destroy calls for xfrm4 and xfrm6 dst_ops
templates; their dst_entries counters will never be used. Move the
xfrm dst_ops initialization from the common xfrm/xfrm_policy.c to
xfrm4/xfrm4_policy.c and xfrm6/xfrm6_policy.c, and call dst_entries_init
and dst_entries_destroy for each net namespace.
The ipv4 and ipv6 xfrms each create dst_ops template, and perform
dst_entries_init on the templates. The template values are copied to each
net namespace's xfrm.xfrm*_dst_ops. The problem there is the dst_ops
pcpuc_entries field is a percpu counter and cannot be used correctly by
simply copying it to another object.
The result of this is a very subtle bug; changes to the dst entries
counter from one net namespace may sometimes get applied to a different
net namespace dst entries counter. This is because of how the percpu
counter works; it has a main count field as well as a pointer to the
percpu variables. Each net namespace maintains its own main count
variable, but all point to one set of percpu variables. When any net
namespace happens to change one of the percpu variables to outside its
small batch range, its count is moved to the net namespace's main count
variable. So with multiple net namespaces operating concurrently, the
dst_ops entries counter can stray from the actual value that it should
be; if counts are consistently moved from one net namespace to another
(which my testing showed is likely), then one net namespace winds up
with a negative dst_ops count while another winds up with a continually
increasing count, eventually reaching its gc_thresh limit, which causes
all new traffic on the net namespace to fail with -ENOBUFS.
Signed-off-by: Dan Streetman <dan.streetman@canonical.com>
Signed-off-by: Dan Streetman <ddstreet@ieee.org>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
This patch fixes following problems :
1) percpu_counter_init() can return an error, therefore
init_frag_mem_limit() must propagate this error so that
inet_frags_init_net() can do the same up to its callers.
2) If ip[46]_frags_ns_ctl_register() fail, we must unwind
properly and free the percpu_counter.
Without this fix, we leave freed object in percpu_counters
global list (if CONFIG_HOTPLUG_CPU) leading to crashes.
This bug was detected by KASAN and syzkaller tool
(http://github.com/google/syzkaller)
Fixes: 6d7b857d54 ("net: use lib/percpu_counter API for fragmentation mem accounting")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Hannes Frederic Sowa <hannes@stressinduktion.org>
Cc: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
skb_set_owner_w() is called from various places that assume
skb->sk always point to a full blown socket (as it changes
sk->sk_wmem_alloc)
We'd like to attach skb to request sockets, and in the future
to timewait sockets as well. For these kind of pseudo sockets,
we need to take a traditional refcount and use sock_edemux()
as the destructor.
It is now time to un-inline skb_set_owner_w(), being too big.
Fixes: ca6fb06518 ("tcp: attach SYNACK messages to request sockets instead of listener")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Bisected-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch changes how the multipath hash is computed for locally
generated flows: now the hash comprises l4 information.
This allows better utilization of the available paths when the existing
flows have the same source IP and the same destination IP: with l3 hash,
even when multiple connections are in place simultaneously, a single path
will be used, while with l4 hash we can use all the available paths.
v2 changes:
- use get_hash_from_flowi4() instead of implementing just another l4 hash
function
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When nexthop is part of multipath route we should clear the
LINKDOWN flag when link goes UP or when first address is added.
This is needed because we always set LINKDOWN flag when DEAD flag
was set but now on UP the nexthop is not dead anymore. Examples when
LINKDOWN bit can be forgotten when no NETDEV_CHANGE is delivered:
- link goes down (LINKDOWN is set), then link goes UP and device
shows carrier OK but LINKDOWN remains set
- last address is deleted (LINKDOWN is set), then address is
added and device shows carrier OK but LINKDOWN remains set
Steps to reproduce:
modprobe dummy
ifconfig dummy0 192.168.168.1 up
here add a multipath route where one nexthop is for dummy0:
ip route add 1.2.3.4 nexthop dummy0 nexthop SOME_OTHER_DEVICE
ifconfig dummy0 down
ifconfig dummy0 up
now ip route shows nexthop that is not dead. Now set the sysctl var:
echo 1 > /proc/sys/net/ipv4/conf/dummy0/ignore_routes_with_linkdown
now ip route will show a dead nexthop because the forgotten
RTNH_F_LINKDOWN is propagated as RTNH_F_DEAD.
Fixes: 8a3d03166f ("net: track link-status of ipv4 nexthops")
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: David S. Miller <davem@davemloft.net>
When fib_netdev_event calls fib_disable_ip on NETDEV_DOWN event
we should not delete the local routes if the local address
is still present. The confusion comes from the fact that both
fib_netdev_event and fib_inetaddr_event use the NETDEV_DOWN
constant. Fix it by returning back the variable 'force'.
Steps to reproduce:
modprobe dummy
ifconfig dummy0 192.168.168.1 up
ifconfig dummy0 down
ip route list table local | grep dummy | grep host
local 192.168.168.1 dev dummy0 proto kernel scope host src 192.168.168.1
Fixes: 8a3d03166f ("net: track link-status of ipv4 nexthops")
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: David S. Miller <davem@davemloft.net>
CHECKSUM_PARTIAL skbs should never arrive in ip_fragment. If we get one
of those warn about them once and handle them gracefully by recalculating
the checksum.
Cc: Eric Dumazet <edumazet@google.com>
Cc: Vlad Yasevich <vyasevich@gmail.com>
Cc: Benjamin Coddington <bcodding@redhat.com>
Cc: Tom Herbert <tom@herbertland.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
We cannot reliable calculate packet size on MSG_MORE corked sockets
and thus cannot decide if they are going to be fragmented later on,
so better not use CHECKSUM_PARTIAL in the first place.
Cc: Eric Dumazet <edumazet@google.com>
Cc: Vlad Yasevich <vyasevich@gmail.com>
Cc: Benjamin Coddington <bcodding@redhat.com>
Cc: Tom Herbert <tom@herbertland.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Steffen Klassert says:
====================
pull request (net-next): ipsec-next 2015-10-30
1) The flow cache is limited by the flow cache limit which
depends on the number of cpus and the xfrm garbage collector
threshold which is independent of the number of cpus. This
leads to the fact that on systems with more than 16 cpus
we hit the xfrm garbage collector limit and refuse new
allocations, so new flows are dropped. On systems with 16
or less cpus, we hit the flowcache limit. In this case, we
shrink the flow cache instead of refusing new flows.
We increase the xfrm garbage collector threshold to INT_MAX
to get the same behaviour, independent of the number of cpus.
2) Fix some unaligned accesses on sparc systems.
From Sowmini Varadhan.
3) Fix some header checks in _decode_session4. We may call
pskb_may_pull with a negative value converted to unsigened
int from pskb_may_pull. This can lead to incorrect policy
lookups. We fix this by a check of the data pointer position
before we call pskb_may_pull.
4) Reload skb header pointers after calling pskb_may_pull
in _decode_session4 as this may change the pointers into
the packet.
5) Add a missing statistic counter on inner mode errors.
Please pull or let me know if there are problems.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
We were computing the child index in cases where the key value we were
looking for was actually less than the base key of the tnode. As a result
we were getting incorrect index values that would cause us to skip over
some children.
To fix this I have added a test that will force us to use child index 0 if
the key we are looking for is less than the key of the current tnode.
Fixes: 8be33e955c ("fib_trie: Fib walk rcu should take a tnode and key instead of a trie and a leaf")
Reported-by: Brian Rak <brak@gameservers.com>
Signed-off-by: Alexander Duyck <aduyck@mirantis.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
gre_gso_segment() chokes if SIT frames were aggregated by GRO engine.
Fixes: 61c1db7fae ("ipv6: sit: add GSO/TSO support")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Conflicts:
net/ipv6/xfrm6_output.c
net/openvswitch/flow_netlink.c
net/openvswitch/vport-gre.c
net/openvswitch/vport-vxlan.c
net/openvswitch/vport.c
net/openvswitch/vport.h
The openvswitch conflicts were overlapping changes. One was
the egress tunnel info fix in 'net' and the other was the
vport ->send() op simplification in 'net-next'.
The xfrm6_output.c conflicts was also a simplification
overlapping a bug fix.
Signed-off-by: David S. Miller <davem@davemloft.net>
Multiple cpus can process duplicates of incoming ACK messages
matching a SYN_RECV request socket. This is a rare event under
normal operations, but definitely can happen.
Only one must win the race, otherwise corruption would occur.
To fix this without adding new atomic ops, we use logic in
inet_ehash_nolisten() to detect the request was present in the same
ehash bucket where we try to insert the new child.
If request socket was not found, we have to undo the child creation.
This actually removes a spin_lock()/spin_unlock() pair in
reqsk_queue_unlink() for the fast path.
Fixes: e994b2f0fb ("tcp: do not lock listener to process SYN packets")
Fixes: 079096f103 ("tcp/dccp: install syn_recv requests into ehash table")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently adding a new ipv4 address always cause the creation of the
related network route, with default metric. When a host has multiple
interfaces on the same network, multiple routes with the same metric
are created.
If the userspace wants to set specific metric on each routes, i.e.
giving better metric to ethernet links in respect to Wi-Fi ones,
the network routes must be deleted and recreated, which is error-prone.
This patch implements the support for IFA_F_NOPREFIXROUTE for ipv4
address. When an address is added with such flag set, no associated
network route is created, no network route is deleted when
said IP is gone and it's up to the user space manage such route.
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If alpha is strictly reduced by alpha >> dctcp_shift_g and if alpha is less
than 1 << dctcp_shift_g, then alpha may never reach zero. For example,
given shift_g=4 and alpha=15, alpha >> dctcp_shift_g yields 0 and alpha
remains 15. The effect isn't noticeable in this case below cwnd=137, but
could gradually drive uncongested flows with leftover alpha down to
cwnd=137. A larger dctcp_shift_g would have a greater effect.
This change causes alpha=15 to drop to 0 instead of being decrementing by 1
as it would when alpha=16. However, it requires one less conditional to
implement since it doesn't have to guard against subtracting 1 from 0U. A
decay of 15 is not unreasonable since an equal or greater amount occurs at
alpha >= 240.
Signed-off-by: Andrew G. Shewmaker <agshew@gmail.com>
Acked-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
A call to pskb_may_pull may change the pointers into the packet,
so reload the pointers after the call.
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
We skip the header informations if the data pointer points
already behind the header in question for some protocols.
This is because we call pskb_may_pull with a negative value
converted to unsigened int from pskb_may_pull in this case.
Skipping the header informations can lead to incorrect policy
lookups, so fix it by a check of the data pointer position
before we call pskb_may_pull.
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
While transitioning to netdev based vport we broke OVS
feature which allows user to retrieve tunnel packet egress
information for lwtunnel devices. Following patch fixes it
by introducing ndo operation to get the tunnel egress info.
Same ndo operation can be used for lwtunnel devices and compat
ovs-tnl-vport devices. So after adding such device operation
we can remove similar operation from ovs-vport.
Fixes: 614732eaa1 ("openvswitch: Use regular VXLAN net_device device").
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Steffen Klassert says:
====================
pull request (net): ipsec 2015-10-22
1) Fix IPsec pre-encap fragmentation for GSO packets.
From Herbert Xu.
2) Fix some header checks in _decode_session6.
We skip the header informations if the data pointer points
already behind the header in question for some protocols.
This is because we call pskb_may_pull with a negative value
converted to unsigened int from pskb_may_pull in this case.
Skipping the header informations can lead to incorrect policy
lookups. From Mathias Krause.
3) Allow to change the replay threshold and expiry timer of a
state without having to set other attributes like replay
counter and byte lifetime. Changing these other attributes
may break the SA. From Michael Rossberg.
4) Fix pmtu discovery for local generated packets.
We may fail dispatch to the inner address family.
As a reault, the local error handler is not called
and the mtu value is not reported back to userspace.
Please pull or let me know if there are problems.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Pablo Neira Ayuso says:
====================
Netfilter fixes for net
The following patchset contains four Netfilter fixes for net, they are:
1) Fix Kconfig dependencies of new nf_dup_ipv4 and nf_dup_ipv6.
2) Remove bogus test nh_scope in IPv4 rpfilter match that is breaking
--accept-local, from Xin Long.
3) Wait for RCU grace period after dropping the pending packets in the
nfqueue, from Florian Westphal.
4) Fix sleeping allocation while holding spin_lock_bh, from Nikolay Borisov.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
if_nlmsg_size() overestimates the minimum allocation size of netlink
dump request (when called from rtnl_calcit()) or the size of the
message (when called from rtnl_getlink()). This is because
ext_filter_mask is not supported by rtnl_link_get_af_size() and
rtnl_link_get_size().
The over-estimation is significant when at least one netdev has many
VLANs configured (8 bytes for each configured VLAN).
This patch-set "rightsizes" the protocol specific attribute size
calculation by propagating ext_filter_mask to rtnl_link_get_af_size()
and adding this a argument to get_link_af_size op in rtnl_af_ops.
Bridge module already used filtering aware sizing for notifications.
br_get_link_af_size_filtered() is consistent with the modified
get_link_af_size op so it replaces br_get_link_af_size() in br_af_ops.
br_get_link_af_size() becomes unused and thus removed.
Signed-off-by: Ronen Arad <ronen.arad@intel.com>
Acked-by: Sridhar Samudrala <sridhar.samudrala@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch implements the second half of RACK that uses the the most
recent transmit time among all delivered packets to detect losses.
tcp_rack_mark_lost() is called upon receiving a dubious ACK.
It then checks if an not-yet-sacked packet was sent at least
"reo_wnd" prior to the sent time of the most recently delivered.
If so the packet is deemed lost.
The "reo_wnd" reordering window starts with 1msec for fast loss
detection and changes to min-RTT/4 when reordering is observed.
We found 1msec accommodates well on tiny degree of reordering
(<3 pkts) on faster links. We use min-RTT instead of SRTT because
reordering is more of a path property but SRTT can be inflated by
self-inflicated congestion. The factor of 4 is borrowed from the
delayed early retransmit and seems to work reasonably well.
Since RACK is still experimental, it is now used as a supplemental
loss detection on top of existing algorithms. It is only effective
after the fast recovery starts or after the timeout occurs. The
fast recovery is still triggered by FACK and/or dupack threshold
instead of RACK.
We introduce a new sysctl net.ipv4.tcp_recovery for future
experiments of loss recoveries. For now RACK can be disabled by
setting it to 0.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch is the first half of the RACK loss recovery.
RACK loss recovery uses the notion of time instead
of packet sequence (FACK) or counts (dupthresh). It's inspired by the
previous FACK heuristic in tcp_mark_lost_retrans(): when a limited
transmit (new data packet) is sacked, then current retransmitted
sequence below the newly sacked sequence must been lost,
since at least one round trip time has elapsed.
But it has several limitations:
1) can't detect tail drops since it depends on limited transmit
2) is disabled upon reordering (assumes no reordering)
3) only enabled in fast recovery ut not timeout recovery
RACK (Recently ACK) addresses these limitations with the notion
of time instead: a packet P1 is lost if a later packet P2 is s/acked,
as at least one round trip has passed.
Since RACK cares about the time sequence instead of the data sequence
of packets, it can detect tail drops when later retransmission is
s/acked while FACK or dupthresh can't. For reordering RACK uses a
dynamically adjusted reordering window ("reo_wnd") to reduce false
positives on ever (small) degree of reordering.
This patch implements tcp_advanced_rack() which tracks the
most recent transmission time among the packets that have been
delivered (ACKed or SACKed) in tp->rack.mstamp. This timestamp
is the key to determine which packet has been lost.
Consider an example that the sender sends six packets:
T1: P1 (lost)
T2: P2
T3: P3
T4: P4
T100: sack of P2. rack.mstamp = T2
T101: retransmit P1
T102: sack of P2,P3,P4. rack.mstamp = T4
T205: ACK of P4 since the hole is repaired. rack.mstamp = T101
We need to be careful about spurious retransmission because it may
falsely advance tp->rack.mstamp by an RTT or an RTO, causing RACK
to falsely mark all packets lost, just like a spurious timeout.
We identify spurious retransmission by the ACK's TS echo value.
If TS option is not applicable but the retransmission is acknowledged
less than min-RTT ago, it is likely to be spurious. We refrain from
using the transmission time of these spurious retransmissions.
The second half is implemented in the next patch that marks packet
lost using RACK timestamp.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
a helper to prepare the main RACK patch
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove the existing lost retransmit detection because RACK subsumes
it completely. This also stops the overloading the ack_seq field of
the skb control block.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Kathleen Nichols' algorithm for tracking the minimum RTT of a
data stream over some measurement window. It uses constant space
and constant time per update. Yet it almost always delivers
the same minimum as an implementation that has to keep all
the data in the window. The measurement window is tunable via
sysctl.net.ipv4.tcp_min_rtt_wlen with a default value of 5 minutes.
The algorithm keeps track of the best, 2nd best & 3rd best min
values, maintaining an invariant that the measurement time of
the n'th best >= n-1'th best. It also makes sure that the three
values are widely separated in the time window since that bounds
the worse case error when that data is monotonically increasing
over the window.
Upon getting a new min, we can forget everything earlier because
it has no value - the new min is less than everything else in the
window by definition and it's the most recent. So we restart fresh
on every new min and overwrites the 2nd & 3rd choices. The same
property holds for the 2nd & 3rd best.
Therefore we have to maintain two invariants to maximize the
information in the samples, one on values (1st.v <= 2nd.v <=
3rd.v) and the other on times (now-win <=1st.t <= 2nd.t <= 3rd.t <=
now). These invariants determine the structure of the code
The RTT input to the windowed filter is the minimum RTT measured
from ACK or SACK, or as the last resort from TCP timestamps.
The accessor tcp_min_rtt() returns the minimum RTT seen in the
window. ~0U indicates it is not available. The minimum is 1usec
even if the true RTT is below that.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently ca_seq_rtt_us does not use Kern's check. Fix that by
checking if any packet acked is a retransmit, for both RTT used
for RTT estimation and congestion control.
Fixes: 5b08e47ca ("tcp: prefer packet timing to TS-ECR for RTT")
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Conflicts:
drivers/net/usb/asix_common.c
net/ipv4/inet_connection_sock.c
net/switchdev/switchdev.c
In the inet_connection_sock.c case the request socket hashing scheme
is completely different in net-next.
The other two conflicts were overlapping changes.
Signed-off-by: David S. Miller <davem@davemloft.net>
Yuchung Cheng