The skb->mark field is a union with reserved_tailroom which is used
in the TCP code paths from stream memory allocation. Allowing SK_SKB
programs to set this field creates a conflict with future code
optimizations, such as "gifting" the skb to the egress path instead
of creating a new skb and doing a memcpy.
Because we do not have a released version of SK_SKB yet lets just
remove it for now. A more appropriate scratch pad to use at the
socket layer is dev_scratch, but lets add that in future kernels
when needed.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
SK_SKB BPF programs are run from the socket/tcp context but early in
the stack before much of the TCP metadata is needed in tcp_skb_cb. So
we can use some unused fields to place BPF metadata needed for SK_SKB
programs when implementing the redirect function.
This allows us to drop the preempt disable logic. It does however
require an API change so sk_redirect_map() has been updated to
additionally provide ctx_ptr to skb. Note, we do however continue to
disable/enable preemption around actual BPF program running to account
for map updates.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Only TCP sockets have been tested and at the moment the state change
callback only handles TCP sockets. This adds a check to ensure that
sockets actually being added are TCP sockets.
For net-next we can consider UDP support.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now sctp processes icmp redirect packet in sctp_icmp_redirect where
it calls sctp_transport_dst_check in which tp->dst can be released.
The problem is before calling sctp_transport_dst_check, it doesn't
check sock_owned_by_user, which means tp->dst could be freed while
a process is accessing it with owning the socket.
An use-after-free issue could be triggered by this.
This patch is to fix it by checking sock_owned_by_user before calling
sctp_transport_dst_check in sctp_icmp_redirect, so that it would not
release tp->dst if users still hold sock lock.
Besides, the same issue fixed in commit 45caeaa5ac ("dccp/tcp: fix
routing redirect race") on sctp also needs this check.
Fixes: 55be7a9c60 ("ipv4: Add redirect support to all protocol icmp error handlers")
Reported-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now when peeling off an association to the sock in another netns, all
transports in this assoc are not to be rehashed and keep use the old
key in hashtable.
As a transport uses sk->net as the hash key to insert into hashtable,
it would miss removing these transports from hashtable due to the new
netns when closing the sock and all transports are being freeed, then
later an use-after-free issue could be caused when looking up an asoc
and dereferencing those transports.
This is a very old issue since very beginning, ChunYu found it with
syzkaller fuzz testing with this series:
socket$inet6_sctp()
bind$inet6()
sendto$inet6()
unshare(0x40000000)
getsockopt$inet_sctp6_SCTP_GET_ASSOC_ID_LIST()
getsockopt$inet_sctp6_SCTP_SOCKOPT_PEELOFF()
This patch is to block this call when peeling one assoc off from one
netns to another one, so that the netns of all transport would not
go out-sync with the key in hashtable.
Note that this patch didn't fix it by rehashing transports, as it's
difficult to handle the situation when the tuple is already in use
in the new netns. Besides, no one would like to peel off one assoc
to another netns, considering ipaddrs, ifaces, etc. are usually
different.
Reported-by: ChunYu Wang <chunwang@redhat.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Daniel Borkmann says:
====================
bpf: Fix for BPF devmap percpu allocation splat
The set fixes a splat in devmap percpu allocation when we alloc
the flush bitmap. Patch 1 is a prerequisite for the fix in patch 2,
patch 1 is rather small, so if this could be routed via -net, for
example, with Tejun's Ack that would be good. Patch 3 gets rid of
remaining PCPU_MIN_UNIT_SIZE checks, which are percpu allocator
internals and should not be used.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
PCPU_MIN_UNIT_SIZE is an implementation detail of the percpu
allocator. Given we support __GFP_NOWARN now, lets just let
the allocation request fail naturally instead. The two call
sites from BPF mistakenly assumed __GFP_NOWARN would work, so
no changes needed to their actual __alloc_percpu_gfp() calls
which use the flag already.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It was reported that syzkaller was able to trigger a splat on
devmap percpu allocation due to illegal/unsupported allocation
request size passed to __alloc_percpu():
[ 70.094249] illegal size (32776) or align (8) for percpu allocation
[ 70.094256] ------------[ cut here ]------------
[ 70.094259] WARNING: CPU: 3 PID: 3451 at mm/percpu.c:1365 pcpu_alloc+0x96/0x630
[...]
[ 70.094325] Call Trace:
[ 70.094328] __alloc_percpu_gfp+0x12/0x20
[ 70.094330] dev_map_alloc+0x134/0x1e0
[ 70.094331] SyS_bpf+0x9bc/0x1610
[ 70.094333] ? selinux_task_setrlimit+0x5a/0x60
[ 70.094334] ? security_task_setrlimit+0x43/0x60
[ 70.094336] entry_SYSCALL_64_fastpath+0x1a/0xa5
This was due to too large max_entries for the map such that we
surpassed the upper limit of PCPU_MIN_UNIT_SIZE. It's fine to
fail naturally here, so switch to __alloc_percpu_gfp() and pass
__GFP_NOWARN instead.
Fixes: 11393cc9b9 ("xdp: Add batching support to redirect map")
Reported-by: Mark Rutland <mark.rutland@arm.com>
Reported-by: Shankara Pailoor <sp3485@columbia.edu>
Reported-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: John Fastabend <john.fastabend@gmail.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add an option for pcpu_alloc() to support __GFP_NOWARN flag.
Currently, we always throw a warning when size or alignment
is unsupported (and also dump stack on failed allocation
requests). The warning itself is harmless since we return
NULL anyway for any failed request, which callers are
required to handle anyway. However, it becomes harmful when
panic_on_warn is set.
The rationale for the WARN() in pcpu_alloc() is that it can
be tracked when larger than supported allocation requests are
made such that allocations limits can be tweaked if warranted.
This makes sense for in-kernel users, however, there are users
of pcpu allocator where allocation size is derived from user
space requests, e.g. when creating BPF maps. In these cases,
the requests should fail gracefully without throwing a splat.
The current work-around was to check allocation size against
the upper limit of PCPU_MIN_UNIT_SIZE from call-sites for
bailing out prior to a call to pcpu_alloc() in order to
avoid throwing the WARN(). This is bad in multiple ways since
PCPU_MIN_UNIT_SIZE is an implementation detail, and having
the checks on call-sites only complicates the code for no
good reason. Thus, lets fix it generically by supporting the
__GFP_NOWARN flag that users can then use with calling the
__alloc_percpu_gfp() helper instead.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: Tejun Heo <tj@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Netanel Belgazal says:
====================
ENA ethernet driver bug fixes
Some fixes for ENA ethernet driver
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
ethtool ena_get_channels() expose the max number of queues as the max
number of queues ENA supports (128 queues) and not the actual number
of created queues.
Signed-off-by: Netanel Belgazal <netanel@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Decrease log level of checksum errors as these messages can be
triggered remotely by bad packets.
Signed-off-by: Netanel Belgazal <netanel@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit f1174f77b5 ("bpf/verifier: rework value tracking")
removed the crafty selection of which pointer types are
allowed to be modified. This is OK for most pointer types
since adjust_ptr_min_max_vals() will catch operations on
immutable pointers. One exception is PTR_TO_CTX which is
now allowed to be offseted freely.
The intent of aforementioned commit was to allow context
access via modified registers. The offset passed to
->is_valid_access() verifier callback has been adjusted
by the value of the variable offset.
What is missing, however, is taking the variable offset
into account when the context register is used. Or in terms
of the code adding the offset to the value passed to the
->convert_ctx_access() callback. This leads to the following
eBPF user code:
r1 += 68
r0 = *(u32 *)(r1 + 8)
exit
being translated to this in kernel space:
0: (07) r1 += 68
1: (61) r0 = *(u32 *)(r1 +180)
2: (95) exit
Offset 8 is corresponding to 180 in the kernel, but offset
76 is valid too. Verifier will "accept" access to offset
68+8=76 but then "convert" access to offset 8 as 180.
Effective access to offset 248 is beyond the kernel context.
(This is a __sk_buff example on a debug-heavy kernel -
packet mark is 8 -> 180, 76 would be data.)
Dereferencing the modified context pointer is not as easy
as dereferencing other types, because we have to translate
the access to reading a field in kernel structures which is
usually at a different offset and often of a different size.
To allow modifying the pointer we would have to make sure
that given eBPF instruction will always access the same
field or the fields accessed are "compatible" in terms of
offset and size...
Disallow dereferencing modified context pointers and add
to selftests the test case described here.
Fixes: f1174f77b5 ("bpf/verifier: rework value tracking")
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It seems that it's possible to toggle NETLINK_F_EXT_ACK
through setsockopt() while another thread/CPU is building
a message inside netlink_ack(), which could then trigger
the WARN_ON()s I added since if it goes from being turned
off to being turned on between allocating and filling the
message, the skb could end up being too small.
Avoid this whole situation by storing the value of this
flag in a separate variable and using that throughout the
function instead.
Fixes: 2d4bc93368 ("netlink: extended ACK reporting")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch correctly sets the number of additional header descriptors
that will be sent in an indirect SCRQ entry.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When an EMAD is transmitted, a timeout work item is scheduled with a
delay of 200ms, so that another EMAD will be retried until a maximum of
five retries.
In certain situations, it's possible for the function waiting on the
EMAD to be associated with a work item that is queued on the same
workqueue (`mlxsw_core`) as the timeout work item. This results in
flushing a work item on the same workqueue.
According to commit e159489baa ("workqueue: relax lockdep annotation
on flush_work()") the above may lead to a deadlock in case the workqueue
has only one worker active or if the system in under memory pressure and
the rescue worker is in use. The latter explains the very rare and
random nature of the lockdep splats we have been seeing:
[ 52.730240] ============================================
[ 52.736179] WARNING: possible recursive locking detected
[ 52.742119] 4.14.0-rc3jiri+ #4 Not tainted
[ 52.746697] --------------------------------------------
[ 52.752635] kworker/1:3/599 is trying to acquire lock:
[ 52.758378] (mlxsw_core_driver_name){+.+.}, at: [<ffffffff811c4fa4>] flush_work+0x3a4/0x5e0
[ 52.767837]
but task is already holding lock:
[ 52.774360] (mlxsw_core_driver_name){+.+.}, at: [<ffffffff811c65c4>] process_one_work+0x7d4/0x12f0
[ 52.784495]
other info that might help us debug this:
[ 52.791794] Possible unsafe locking scenario:
[ 52.798413] CPU0
[ 52.801144] ----
[ 52.803875] lock(mlxsw_core_driver_name);
[ 52.808556] lock(mlxsw_core_driver_name);
[ 52.813236]
*** DEADLOCK ***
[ 52.819857] May be due to missing lock nesting notation
[ 52.827450] 3 locks held by kworker/1:3/599:
[ 52.832221] #0: (mlxsw_core_driver_name){+.+.}, at: [<ffffffff811c65c4>] process_one_work+0x7d4/0x12f0
[ 52.842846] #1: ((&(&bridge->fdb_notify.dw)->work)){+.+.}, at: [<ffffffff811c65c4>] process_one_work+0x7d4/0x12f0
[ 52.854537] #2: (rtnl_mutex){+.+.}, at: [<ffffffff822ad8e7>] rtnl_lock+0x17/0x20
[ 52.863021]
stack backtrace:
[ 52.867890] CPU: 1 PID: 599 Comm: kworker/1:3 Not tainted 4.14.0-rc3jiri+ #4
[ 52.875773] Hardware name: Mellanox Technologies Ltd. "MSN2100-CB2F"/"SA001017", BIOS 5.6.5 06/07/2016
[ 52.886267] Workqueue: mlxsw_core mlxsw_sp_fdb_notify_work [mlxsw_spectrum]
[ 52.894060] Call Trace:
[ 52.909122] __lock_acquire+0xf6f/0x2a10
[ 53.025412] lock_acquire+0x158/0x440
[ 53.047557] flush_work+0x3c4/0x5e0
[ 53.087571] __cancel_work_timer+0x3ca/0x5e0
[ 53.177051] cancel_delayed_work_sync+0x13/0x20
[ 53.182142] mlxsw_reg_trans_bulk_wait+0x12d/0x7a0 [mlxsw_core]
[ 53.194571] mlxsw_core_reg_access+0x586/0x990 [mlxsw_core]
[ 53.225365] mlxsw_reg_query+0x10/0x20 [mlxsw_core]
[ 53.230882] mlxsw_sp_fdb_notify_work+0x2a3/0x9d0 [mlxsw_spectrum]
[ 53.237801] process_one_work+0x8f1/0x12f0
[ 53.321804] worker_thread+0x1fd/0x10c0
[ 53.435158] kthread+0x28e/0x370
[ 53.448703] ret_from_fork+0x2a/0x40
[ 53.453017] mlxsw_spectrum 0000:01:00.0: EMAD retries (2/5) (tid=bf4549b100000774)
[ 53.453119] mlxsw_spectrum 0000:01:00.0: EMAD retries (5/5) (tid=bf4549b100000770)
[ 53.453132] mlxsw_spectrum 0000:01:00.0: EMAD reg access failed (tid=bf4549b100000770,reg_id=200b(sfn),type=query,status=0(operation performed))
[ 53.453143] mlxsw_spectrum 0000:01:00.0: Failed to get FDB notifications
Fix this by creating another workqueue for EMAD timeouts, thereby
preventing the situation of a work item trying to flush a work item
queued on the same workqueue.
Fixes: caf7297e7a ("mlxsw: core: Introduce support for asynchronous EMAD register access")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reported-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Johannes Berg says:
====================
Just a single fix, for a WoWLAN-related part of CVE-2017-13080.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
When changing dev tx_queue_len via netlink or net-sysfs,
a NETDEV_CHANGE_TX_QUEUE_LEN event notification will be
called.
But dev_ioctl missed this event notification, which could
cause no userspace notification would be sent.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit 7091d8c '(net/sched: cls_flower: Add offload support using egress
Hardware device') made sure (when fl_hw_replace_filter is called) to put
the egress_dev mark on persisent structure instance. Hence, following calls
into the HW driver for stats and deletion will note it and act accordingly.
With commit de4784ca03 this property is lost and hence when called,
the HW driver failes to operate (stats, delete) on the offloaded flow.
Fix it by setting the egress_dev flag whenever the ingress device is
different from the hw device since this is exactly the condition under
which we're calling into the HW driver through the egress port net-device.
Fixes: de4784ca03 ('net: sched: get rid of struct tc_to_netdev')
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roi Dayan <roid@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
register_netdevice() could fail early when we have an invalid
dev name, in which case ->ndo_uninit() is not called. For tun
device, this is a problem because a timer etc. are already
initialized and it expects ->ndo_uninit() to clean them up.
We could move these initializations into a ->ndo_init() so
that register_netdevice() knows better, however this is still
complicated due to the logic in tun_detach().
Therefore, I choose to just call dev_get_valid_name() before
register_netdevice(), which is quicker and much easier to audit.
And for this specific case, it is already enough.
Fixes: 96442e4242 ("tuntap: choose the txq based on rxq")
Reported-by: Dmitry Alexeev <avekceeb@gmail.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
IFLA_IFALIAS is defined as NLA_STRING. It means that the minimal length of
the attribute is 1 ("\0"). However, to remove an alias, the attribute
length must be 0 (see dev_set_alias()).
Let's define the type to NLA_BINARY to allow 0-length string, so that the
alias can be removed.
Example:
$ ip l s dummy0 alias foo
$ ip l l dev dummy0
5: dummy0: <BROADCAST,NOARP> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
link/ether ae:20:30:4f:a7:f3 brd ff:ff:ff:ff:ff:ff
alias foo
Before the patch:
$ ip l s dummy0 alias ""
RTNETLINK answers: Numerical result out of range
After the patch:
$ ip l s dummy0 alias ""
$ ip l l dev dummy0
5: dummy0: <BROADCAST,NOARP> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
link/ether ae:20:30:4f:a7:f3 brd ff:ff:ff:ff:ff:ff
CC: Oliver Hartkopp <oliver@hartkopp.net>
CC: Stephen Hemminger <stephen@networkplumber.org>
Fixes: 96ca4a2cc1 ("net: remove ifalias on empty given alias")
Reported-by: Julien FLoret <julien.floret@6wind.com>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Xin Long says:
====================
rtnetlink: a bunch of fixes for userspace notifications in changing dev properties
Whenever any property of a link, address, route, etc. changes by whatever way,
kernel should notify the programs that listen for such events in userspace.
The patchet "rtnetlink: Cleanup user notifications for netdev events" tried to
fix a redundant notifications issue, but it also introduced a side effect.
After that, user notifications could only be sent when changing dev properties
via netlink api. As it removed some events process in rtnetlink_event where
the notifications was sent to users.
It resulted in no notification generated when dev properties are changed via
other ways, like ioctl, sysfs, etc. It may cause some user programs doesn't
work as expected because of the missing notifications.
This patchset will fix it by bringing some of these netdev events back and
also fix the old redundant notifications issue with a proper way.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
NETDEV_CHANGE_TX_QUEUE_LEN event process in rtnetlink_event would
send a notification for userspace and tx_queue_len's setting in
do_setlink would trigger NETDEV_CHANGE_TX_QUEUE_LEN.
So it shouldn't set DO_SETLINK_NOTIFY status for this change to
send a notification any more.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The check 'status & DO_SETLINK_NOTIFY' in do_setlink doesn't really
work after status & DO_SETLINK_MODIFIED, as:
DO_SETLINK_MODIFIED 0x1
DO_SETLINK_NOTIFY 0x3
Considering that notifications are suppposed to be sent only when
status have the flag DO_SETLINK_NOTIFY, the right check would be:
(status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY
This would avoid lots of duplicated notifications when setting some
properties of a link.
Fixes: ba9989069f ("rtnl/do_setlink(): notify when a netdev is modified")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: David Ahern <dsahern@gmail.com>
Acked-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
John Fastabend
Xin Long
David S. Miller
Daniel Borkmann
Netanel Belgazal
Jakub Kicinski
Johannes Berg
Thomas Falcon
Ido Schimmel
Or Gerlitz
Cong Wang
Nicolas Dichtel