When a map free is called and in parallel a socket is closed we
have two paths that can potentially reset the socket prot ops, the
bpf close() path and the map free path. This creates a problem
with which prot ops should be used from the socket closed side.

If the map_free side completes first then we want to call the
original lowest level ops. However, if the tls path runs first
we want to call the sockmap ops. Additionally there was no locking
around prot updates in TLS code paths so the prot ops could
be changed multiple times once from TLS path and again from sockmap
side potentially leaving ops pointed at either TLS or sockmap
when psock and/or tls context have already been destroyed.

To fix this race first only update ops inside callback lock
so that TLS, sockmap and lowest level all agree on prot state.
Second and a ULP callback update() so that lower layers can
inform the upper layer when they are being removed allowing the
upper layer to reset prot ops.

This gets us close to allowing sockmap and tls to be stacked
in arbitrary order but will save that patch for *next trees.

v4:
 - make sure we don't free things for device;
 - remove the checks which swap the callbacks back
   only if TLS is at the top.

Reported-by: syzbot+06537213db7ba2745c4a@syzkaller.appspotmail.com
Fixes: 02c558b2 ("bpf: sockmap, support for msg_peek in sk_msg with redirect ingress")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Sockmap does not currently support adding sockets after TLS has been
enabled. There never was a real use case for this so it was never
added. But, we lost the test for ULP at some point so add it here
and fail the socket insert if TLS is enabled. Future work could
make sockmap support this use case but fixup the bug here.

Fixes: 604326b4 ("bpf, sockmap: convert to generic sk_msg interface")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

We need to have a synchronize_rcu before free'ing the sockmap because
any outstanding psock references will have a pointer to the map and
when they use this could trigger a use after free.

Fixes: 604326b4 ("bpf, sockmap: convert to generic sk_msg interface")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

__sock_map_delete() may be called from a tcp event such as unhash or
close from the following trace,

  tcp_bpf_close()
    tcp_bpf_remove()
      sk_psock_unlink()
        sock_map_delete_from_link()
          __sock_map_delete()

In this case the sock lock is held but this only protects against
duplicate removals on the TCP side. If the map is free'd then we have
this trace,

  sock_map_free
    xchg()                  <- replaces map entry
    sock_map_unref()
      sk_psock_put()
        sock_map_del_link()

The __sock_map_delete() call however uses a read, test, null over the
map entry which can result in both paths trying to free the map
entry.

To fix use xchg in TCP paths as well so we avoid having two references
to the same map entry.

Fixes: 604326b4 ("bpf, sockmap: convert to generic sk_msg interface")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

It is possible (via shutdown()) for TCP socks to go through TCP_CLOSE
state via tcp_disconnect() without actually calling tcp_close which
would then call the tls close callback. Because of this a user could
disconnect a socket then put it in a LISTEN state which would break
our assumptions about sockets always being ESTABLISHED state.

More directly because close() can call unhash() and unhash is
implemented by sockmap if a sockmap socket has TLS enabled we can
incorrectly destroy the psock from unhash() and then call its close
handler again. But because the psock (sockmap socket representation)
is already destroyed we call close handler in sk->prot. However,
in some cases (TLS BASE/BASE case) this will still point at the
sockmap close handler resulting in a circular call and crash reported
by syzbot.

To fix both above issues implement the unhash() routine for TLS.

v4:
 - add note about tls offload still needing the fix;
 - move sk_proto to the cold cache line;
 - split TX context free into "release" and "free",
   otherwise the GC work itself is in already freed
   memory;
 - more TX before RX for consistency;
 - reuse tls_ctx_free();
 - schedule the GC work after we're done with context
   to avoid UAF;
 - don't set the unhash in all modes, all modes "inherit"
   TLS_BASE's callbacks anyway;
 - disable the unhash hook for TLS_HW.

Fixes: 3c4d7559 ("tls: kernel TLS support")
Reported-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

The tls close() callback currently drops the sock lock to call
strp_done(). Split up the RX cleanup into stopping the strparser
and releasing most resources, syncing strparser and finally
freeing the context.

To avoid the need for a strp_done() call on the cleanup path
of device offload make sure we don't arm the strparser until
we are sure init will be successful.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

The tls close() callback currently drops the sock lock, makes a
cancel_delayed_work_sync() call, and then relocks the sock.

By restructuring the code we can avoid droping lock and then
reclaiming it. To simplify this we do the following,

 tls_sk_proto_close
 set_bit(CLOSING)
 set_bit(SCHEDULE)
 cancel_delay_work_sync() <- cancel workqueue
 lock_sock(sk)
 ...
 release_sock(sk)
 strp_done()

Setting the CLOSING bit prevents the SCHEDULE bit from being
cleared by any workqueue items e.g. if one happens to be
scheduled and run between when we set SCHEDULE bit and cancel
work. Then because SCHEDULE bit is set now no new work will
be scheduled.

Tested with net selftests and bpf selftests.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

The deprecated TOE offload doesn't actually do anything in
tls_sk_proto_close() - all TLS code is skipped and context
not freed. Remove the callback to make it easier to refactor
tls_sk_proto_close().
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

In tls_set_device_offload_rx() we prepare the software context
for RX fallback and proceed to add the connection to the device.
Unfortunately, software context prep includes arming strparser
so in case of a later error we have to release the socket lock
to call strp_done().

In preparation for not releasing the socket lock half way through
callbacks move arming strparser into a separate function.
Following patches will make use of that.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

If VAR in non-sanitized BTF was size less than 4, converting such VAR
into an INT with size=4 will cause BTF validation failure due to
violationg of STRUCT (into which DATASEC was converted) member size.
Fix by conservatively using size=1.
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

In case when BTF loading fails despite sanitization, but BPF object has
.BTF.ext loaded as well, we free and null obj->btf, but not
obj->btf_ext. This leads to an attempt to relocate .BTF.ext later on
during bpf_object__load(), which assumes obj->btf is present. This leads
to SIGSEGV on null pointer access. Fix bug by freeing and nulling
obj->btf_ext as well.
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Neal reported incorrect use of ns_capable() from bpf hook.

bpf_setsockopt(...TCP_CONGESTION...)
  -> tcp_set_congestion_control()
   -> ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)
    -> ns_capable_common()
     -> current_cred()
      -> rcu_dereference_protected(current->cred, 1)

Accessing 'current' in bpf context makes no sense, since packets
are processed from softirq context.

As Neal stated : The capability check in tcp_set_congestion_control()
was written assuming a system call context, and then was reused from
a BPF call site.

The fix is to add a new parameter to tcp_set_congestion_control(),
so that the ns_capable() call is only performed under the right
context.

Fixes: 91b5b21c ("bpf: Add support for changing congestion control")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Lawrence Brakmo <brakmo@fb.com>
Reported-by: Neal Cardwell <ncardwell@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Acked-by: Lawrence Brakmo <brakmo@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

In case of error, the function of_get_mac_address() returns ERR_PTR()
and never returns NULL. The NULL test in the return value check should
be replaced with IS_ERR().

Fixes: d51b6ce4 ("net: ethernet: add ag71xx driver")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Reviewed-by: Oleksij Rempel <o.rempel@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

Fix to return error code -ENOMEM from the dmam_alloc_coherent() error
handling case instead of 0, as done elsewhere in this function.

Fixes: d51b6ce4 ("net: ethernet: add ag71xx driver")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Reviewed-by: Oleksij Rempel <o.rempel@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

Signed-off-by: Rogan Dawes <rogan@dawes.za.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

Unlike legacy chips, 57500 chips don't need additional VNIC resources
for aRFS/ntuple.  Fix the code accordingly so that we don't reserve
and allocate additional VNICs on 57500 chips.  Without this patch,
the driver is failing to initialize when it tries to allocate extra
VNICs.

Fixes: ac33906c ("bnxt_en: Add support for aRFS on 57500 chips.")
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add the missing unlock before return from function sk_buff()
in the error handling case.

Fixes: f3097be2 ("net: dsa: sja1105: Add a state machine for RX timestamping")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Vivien Didelot <vivien.didelot@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Variables allocated by kvzalloc should not be freed by kfree.
Because they may be allocated by vmalloc.
So we replace kfree with kvfree here.
Signed-off-by: Chuhong Yuan <hslester96@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Alexei Starovoitov says:

====================
pull-request: bpf 2019-07-18

The following pull-request contains BPF updates for your *net* tree.

The main changes are:

1) verifier precision propagation fix, from Andrii.

2) BTF size fix for typedefs, from Andrii.

3) a bunch of big endian fixes, from Ilya.

4) wide load from bpf_sock_addr fixes, from Stanislav.

5) a bunch of misc fixes from a number of developers.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

test_xdp_noinline fails on s390 due to a handful of endianness issues.
Use ntohs for parsing eth_proto.
Replace bswaps with ntohs/htons.
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Acked-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

This test looks up a 32-bit map element and then loads it using a 64-bit
load. This does not work on s390, which is a big-endian machine.

Since the point of this test doesn't seem to be loading a smaller value
using a larger load, simply use a 32-bit load.
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Variable allocated by kvmalloc should not be freed by kfree.
Because it may be allocated by vmalloc.
So replace kfree with kvfree here.

Fixes: 9b1f2982 ("net/mlx5: Add support for FW fatal reporter dump")
Signed-off-by: Chuhong Yuan <hslester96@gmail.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add myself to maintainers since i provided the XDP and page_pool
implementation
Signed-off-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Acked-by: Jassi Brar <jaswinder.singh@linaro.org>
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

When a route needs to be appended to an existing multipath route,
fib6_add_rt2node() first appends it to the siblings list and increments
the number of sibling routes on each sibling.

Later, the function notifies the route via call_fib6_entry_notifiers().
In case the notification is vetoed, the route is not unlinked from the
siblings list, which can result in a use-after-free.

Fix this by unlinking the route from the siblings list before returning
an error.

Audited the rest of the call sites from which the FIB notification chain
is called and could not find more problems.

Fixes: 2233000c ("net/ipv6: Move call_fib6_entry_notifiers up for route adds")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reported-by: Alexander Petrovskiy <alexpe@mellanox.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge tag 'wireless-drivers-for-davem-2019-07-18' of git://git.kernel.org/pub/scm/linux/kernel/git/kvalo/wireless-drivers

Kalle Valo says:

====================
wireless-drivers fixes for 5.3

First set of fixes for 5.3.

iwlwifi

* add new cards for 9000 and 20000 series and qu c-step devices

ath10k

* workaround an uninitialised variable warning

rt2x00

* fix rx queue hand on USB
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Use vzalloc and vzalloc_node instead of using vmalloc and
vmalloc_node and then zeroing the allocated memory by
memset 0.
This simplifies the code.
Signed-off-by: Chuhong Yuan <hslester96@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Signed-off-by: Su Yanjun <suyj.fnst@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Currently we silently ignore filters if we cannot meet the filter
requirements. This will lead to the MAC dropping packets that are
expected to pass. A better solution would be to set the NIC to promisc
mode when the required filters cannot be met.

Also correct the number of MDF filters supported. It should be 17,
not 16.
Signed-off-by: Justin Chen <justinpopo6@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Paul reported that l2tp sessions were broken after the commit referenced
in the Fixes tag. Prior to this commit rt6_check returned NULL if the
rt6_info 'from' was NULL - ie., the dst_entry was disconnected from a FIB
entry. Restore that behavior.

Fixes: 93531c67 ("net/ipv6: separate handling of FIB entries from dst based routes")
Reported-by: Paul Donohue <linux-kernel@PaulSD.com>
Tested-by: Paul Donohue <linux-kernel@PaulSD.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The tipc_msg_validate() function leaves a boolean flag 'validated' in
the validated buffer's control block, to avoid performing this action
more than once. However, at reception of new packets, the position of
this field may already have been set by lower layer protocols, so
that the packet is erroneously perceived as already validated by TIPC.

We fix this by initializing the said field to 'false' before performing
the initial validation.
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Cong Wang says:

====================
ipv4: relax source validation check for loopback packets

This patchset fixes a corner case when loopback packets get dropped
by rp_filter when we route them from veth to lo. Patch 1 is the fix
and patch 2 provides a simplified test case for this scenario.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Add a test case to simulate the loopback packet case fixed
in the previous patch.

This test gets passed after the fix:

IPv4 rp_filter tests
    TEST: rp_filter passes local packets                                [ OK ]
    TEST: rp_filter passes loopback packets                             [ OK ]

Cc: David Ahern <dsahern@gmail.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

In a rare case where we redirect local packets from veth to lo,
these packets fail to pass the source validation when rp_filter
is turned on, as the tracing shows:

  <...>-311708 [040] ..s1 7951180.957825: fib_table_lookup: table 254 oif 0 iif 1 src 10.53.180.130 dst 10.53.180.130 tos 0 scope 0 flags 0
  <...>-311708 [040] ..s1 7951180.957826: fib_table_lookup_nh: nexthop dev eth0 oif 4 src 10.53.180.130

So, the fib table lookup returns eth0 as the nexthop even though
the packets are local and should be routed to loopback nonetheless,
but they can't pass the dev match check in fib_info_nh_uses_dev()
without this patch.

It should be safe to relax this check for this special case, as
normally packets coming out of loopback device still have skb_dst
so they won't even hit this slow path.

Cc: Julian Anastasov <ja@ssi.bg>
Cc: David Ahern <dsahern@gmail.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Ido Schimmel says:

====================
mlxsw: Two fixes

This patchset contains two fixes for mlxsw.

Patch #1 from Petr fixes an issue in which DSCP rewrite can occur even
if the egress port was switched to Trust L2 mode where priority mapping
is based on PCP.

Patch #2 fixes a problem where packets can be learned on a non-existing
FID if a tc filter with a redirect action is configured on a bridged
port. The problem and fix are explained in detail in the commit message.

Please consider both patches for 5.2.y
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

The switch periodically sends notifications about learned FDB entries.
Among other things, the notification includes the FID (Filtering
Identifier) and the port on which the MAC was learned.

In case the driver does not have the FID defined on the relevant port,
the following error will be periodically generated:

mlxsw_spectrum2 0000:06:00.0 swp32: Failed to find a matching {Port, VID} following FDB notification

This is not supposed to happen under normal conditions, but can happen
if an ingress tc filter with a redirect action is installed on a bridged
port. The redirect action will cause the packet's FID to be changed to
the dummy FID and a learning notification will be emitted with this FID
- which is not defined on the bridged port.

Fix this by having the driver ignore learning notifications generated
with the dummy FID and delete them from the device.

Another option is to chain an ignore action after the redirect action
which will cause the device to disable learning, but this means that we
need to consume another action whenever a redirect action is used. In
addition, the scenario described above is merely a corner case.

Fixes: cedbb8b2 ("mlxsw: spectrum_flower: Set dummy FID before forward action")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reported-by: Alex Kushnarov <alexanderk@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Tested-by: Alex Kushnarov <alexanderk@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Spectrum systems use DSCP rewrite map to update DSCP field in egressing
packets to correspond to priority that the packet has. Whether rewriting
will take place is determined at the point when the packet ingresses the
switch: if the port is in Trust L3 mode, packet priority is determined from
the DSCP map at the port, and DSCP rewrite will happen. If the port is in
Trust L2 mode, 802.1p is used for packet prioritization, and no DSCP
rewrite will happen.

The driver determines the port trust mode based on whether any DSCP
prioritization rules are in effect at given port. If there are any, trust
level is L3, otherwise it's L2. When the last DSCP rule is removed, the
port is switched to trust L2. Under that scenario, if DSCP of a packet
should be rewritten, it should be rewritten to 0.

However, when switching to Trust L2, the driver neglects to also update the
DSCP rewrite map. The last DSCP rule thus remains in effect, and packets
egressing through this port, if they have the right priority, will have
their DSCP set according to this rule.

Fix by first configuring the rewrite map, and only then switching to trust
L2 and bailing out.

Fixes: b2b1dab6 ("mlxsw: spectrum: Support ieee_setapp, ieee_delapp")
Signed-off-by: Petr Machata <petrm@mellanox.com>
Reported-by: Alex Veber <alexve@mellanox.com>
Tested-by: Alex Veber <alexve@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ag71xx uses devm_ioremap_nocache. This fixes usage of an implicit function

Fixes: d51b6ce4 ("net: ethernet: add ag71xx driver")
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

For qdisc's that support TC filters and set TCQ_F_CAN_BYPASS,
notably fq_codel, it makes no sense to let packets bypass the TC
filters we setup in any scenario, otherwise our packets steering
policy could not be enforced.

This can be reproduced easily with the following script:

 ip li add dev dummy0 type dummy
 ifconfig dummy0 up
 tc qd add dev dummy0 root fq_codel
 tc filter add dev dummy0 parent 8001: protocol arp basic action mirred egress redirect dev lo
 tc filter add dev dummy0 parent 8001: protocol ip basic action mirred egress redirect dev lo
 ping -I dummy0 192.168.112.1

Without this patch, packets are sent directly to dummy0 without
hitting any of the filters. With this patch, packets are redirected
to loopback as expected.

This fix is not perfect, it only unsets the flag but does not set it back
because we have to save the information somewhere in the qdisc if we
really want that. Note, both fq_codel and sfq clear this flag in their
->bind_tcf() but this is clearly not sufficient when we don't use any
class ID.

Fixes: 23624935 ("net_sched: TCQ_F_CAN_BYPASS generalization")
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Gerd Rausch says:

====================
net/rds: RDMA fixes

A number of net/rds fixes necessary to make "rds_rdma.ko"
pass some basic Oracle internal tests.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Otherwise, if an IB connection is torn down before "rds_ib_setup_qp"
is called, the value of "ic->i_fastreg_wrs" is still at zero
(as it wasn't initialized by "rds_ib_setup_qp").
Consequently "rds_ib_conn_path_shutdown" will spin forever,
waiting for it to go back to "RDS_IB_DEFAULT_FR_WR",
which of course will never happen as there are no
outstanding work requests.
Signed-off-by: Gerd Rausch <gerd.rausch@oracle.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>