platform/kernel/linux-rpi.git
13 months agoigc: Work around HW bug causing missing timestamps
Vinicius Costa Gomes [Wed, 7 Jun 2023 21:32:32 +0000 (14:32 -0700)]
igc: Work around HW bug causing missing timestamps

There's an hardware issue that can cause missing timestamps. The bug
is that the interrupt is only cleared if the IGC_TXSTMPH_0 register is
read.

The bug can cause a race condition if a timestamp is captured at the
wrong time, and we will miss that timestamp. To reduce the time window
that the problem is able to happen, in case no timestamp was ready, we
read the "previous" value of the timestamp registers, and we compare
with the "current" one, if it didn't change we can be reasonably sure
that no timestamp was captured. If they are different, we use the new
value as the captured timestamp.

The HW bug is not easy to reproduce, got to reproduce it when smashing
the NIC with timestamping requests from multiple applications (e.g.
multiple ntpperf instances + ptp4l), after 10s of minutes.

This workaround has more impact when multiple timestamp registers are
used, and the IGC_TXSTMPH_0 register always need to be read, so the
interrupt is cleared.

Fixes: 2c344ae24501 ("igc: Add support for TX timestamping")
Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Tested-by: Naama Meir <naamax.meir@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
13 months agoigc: Retrieve TX timestamp during interrupt handling
Vinicius Costa Gomes [Wed, 7 Jun 2023 21:32:31 +0000 (14:32 -0700)]
igc: Retrieve TX timestamp during interrupt handling

When the interrupt is handled, the TXTT_0 bit in the TSYNCTXCTL
register should already be set and the timestamp value already loaded
in the appropriate register.

This simplifies the handling, and reduces the latency for retrieving
the TX timestamp, which increase the amount of TX timestamps that can
be handled in a given time period.

As the "work" function doesn't run in a workqueue anymore, rename it
to something more sensible, a event handler.

Using ntpperf[1] we can see the following performance improvements:

Before:

$ sudo ./ntpperf -i enp3s0 -m 10:22:22:22:22:21 -d 192.168.1.3 -s 172.18.0.0/16 -I -H -o -37
               |          responses            |     TX timestamp offset (ns)
rate   clients |  lost invalid   basic  xleave |    min    mean     max stddev
1000       100   0.00%   0.00%   0.00% 100.00%      -56      +9     +52     19
1500       150   0.00%   0.00%   0.00% 100.00%      -40     +30     +75     22
2250       225   0.00%   0.00%   0.00% 100.00%      -11     +29     +72     15
3375       337   0.00%   0.00%   0.00% 100.00%      -18     +40     +88     22
5062       506   0.00%   0.00%   0.00% 100.00%      -19     +23     +77     15
7593       759   0.00%   0.00%   0.00% 100.00%       +7     +47   +5168     43
11389     1138   0.00%   0.00%   0.00% 100.00%      -11     +41   +5240     39
17083     1708   0.00%   0.00%   0.00% 100.00%      +19     +60   +5288     50
25624     2562   0.00%   0.00%   0.00% 100.00%       +1     +56   +5368     58
38436     3843   0.00%   0.00%   0.00% 100.00%      -84     +12   +8847     66
57654     5765   0.00%   0.00% 100.00%   0.00%
86481     8648   0.00%   0.00% 100.00%   0.00%
129721   12972   0.00%   0.00% 100.00%   0.00%
194581   16384   0.00%   0.00% 100.00%   0.00%
291871   16384  27.35%   0.00%  72.65%   0.00%
437806   16384  50.05%   0.00%  49.95%   0.00%

After:

$ sudo ./ntpperf -i enp3s0 -m 10:22:22:22:22:21 -d 192.168.1.3 -s 172.18.0.0/16 -I -H -o -37
               |          responses            |     TX timestamp offset (ns)
rate   clients |  lost invalid   basic  xleave |    min    mean     max stddev
1000       100   0.00%   0.00%   0.00% 100.00%      -44      +0     +61     19
1500       150   0.00%   0.00%   0.00% 100.00%       -6     +39     +81     16
2250       225   0.00%   0.00%   0.00% 100.00%      -22     +25     +69     15
3375       337   0.00%   0.00%   0.00% 100.00%      -28     +15     +56     14
5062       506   0.00%   0.00%   0.00% 100.00%       +7     +78    +143     27
7593       759   0.00%   0.00%   0.00% 100.00%      -54     +24    +144     47
11389     1138   0.00%   0.00%   0.00% 100.00%      -90     -33     +28     21
17083     1708   0.00%   0.00%   0.00% 100.00%      -50      -2     +35     14
25624     2562   0.00%   0.00%   0.00% 100.00%      -62      +7     +66     23
38436     3843   0.00%   0.00%   0.00% 100.00%      -33     +30   +5395     36
57654     5765   0.00%   0.00% 100.00%   0.00%
86481     8648   0.00%   0.00% 100.00%   0.00%
129721   12972   0.00%   0.00% 100.00%   0.00%
194581   16384  19.50%   0.00%  80.50%   0.00%
291871   16384  35.81%   0.00%  64.19%   0.00%
437806   16384  55.40%   0.00%  44.60%   0.00%

[1] https://github.com/mlichvar/ntpperf

Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Reviewed-by: Kurt Kanzenbach <kurt@linutronix.de>
Tested-by: Naama Meir <naamax.meir@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
13 months agoigc: Check if hardware TX timestamping is enabled earlier
Vinicius Costa Gomes [Wed, 7 Jun 2023 21:32:30 +0000 (14:32 -0700)]
igc: Check if hardware TX timestamping is enabled earlier

Before requesting a packet transmission to be hardware timestamped,
check if the user has TX timestamping enabled. Fixes an issue that if
a packet was internally forwarded to the NIC, and it had the
SKBTX_HW_TSTAMP flag set, the driver would mark that timestamp as
skipped.

In reality, that timestamp was "not for us", as TX timestamp could
never be enabled in the NIC.

Checking if the TX timestamping is enabled earlier has a secondary
effect that when TX timestamping is disabled, there's no need to check
for timestamp timeouts.

We should only take care to free any pending timestamp when TX
timestamping is disabled, as that skb would never be released
otherwise.

Fixes: 2c344ae24501 ("igc: Add support for TX timestamping")
Suggested-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Tested-by: Naama Meir <naamax.meir@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
13 months agoigc: Fix race condition in PTP tx code
Vinicius Costa Gomes [Wed, 7 Jun 2023 21:32:29 +0000 (14:32 -0700)]
igc: Fix race condition in PTP tx code

Currently, the igc driver supports timestamping only one tx packet at a
time. During the transmission flow, the skb that requires hardware
timestamping is saved in adapter->ptp_tx_skb. Once hardware has the
timestamp, an interrupt is delivered, and adapter->ptp_tx_work is
scheduled. In igc_ptp_tx_work(), we read the timestamp register, update
adapter->ptp_tx_skb, and notify the network stack.

While the thread executing the transmission flow (the user process
running in kernel mode) and the thread executing ptp_tx_work don't
access adapter->ptp_tx_skb concurrently, there are two other places
where adapter->ptp_tx_skb is accessed: igc_ptp_tx_hang() and
igc_ptp_suspend().

igc_ptp_tx_hang() is executed by the adapter->watchdog_task worker
thread which runs periodically so it is possible we have two threads
accessing ptp_tx_skb at the same time. Consider the following scenario:
right after __IGC_PTP_TX_IN_PROGRESS is set in igc_xmit_frame_ring(),
igc_ptp_tx_hang() is executed. Since adapter->ptp_tx_start hasn't been
written yet, this is considered a timeout and adapter->ptp_tx_skb is
cleaned up.

This patch fixes the issue described above by adding the ptp_tx_lock to
protect access to ptp_tx_skb and ptp_tx_start fields from igc_adapter.
Since igc_xmit_frame_ring() called in atomic context by the networking
stack, ptp_tx_lock is defined as a spinlock, and the irq safe variants
of lock/unlock are used.

With the introduction of the ptp_tx_lock, the __IGC_PTP_TX_IN_PROGRESS
flag doesn't provide much of a use anymore so this patch gets rid of it.

Fixes: 2c344ae24501 ("igc: Add support for TX timestamping")
Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Reviewed-by: Kurt Kanzenbach <kurt@linutronix.de>
Tested-by: Naama Meir <naamax.meir@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
13 months agoMerge tag 'nf-23-06-21' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf
Paolo Abeni [Thu, 22 Jun 2023 12:39:06 +0000 (14:39 +0200)]
Merge tag 'nf-23-06-21' of git://git./linux/kernel/git/netfilter/nf

Pablo Neira Ayuso says:

====================
Netfilter/IPVS fixes for net

This is v3, including a crash fix for patch 01/14.

The following patchset contains Netfilter/IPVS fixes for net:

1) Fix UDP segmentation with IPVS tunneled traffic, from Terin Stock.

2) Fix chain binding transaction logic, add a bound flag to rule
   transactions. Remove incorrect logic in nft_data_hold() and
   nft_data_release().

3) Add a NFT_TRANS_PREPARE_ERROR deactivate state to deal with releasing
   the set/chain as a follow up to 1240eb93f061 ("netfilter: nf_tables:
   incorrect error path handling with NFT_MSG_NEWRULE")

4) Drop map element references from preparation phase instead of
   set destroy path, otherwise bogus EBUSY with transactions such as:

        flush chain ip x y
        delete chain ip x w

   where chain ip x y contains jump/goto from set elements.

5) Pipapo set type does not regard generation mask from the walk
   iteration.

6) Fix reference count underflow in set element reference to
   stateful object.

7) Several patches to tighten the nf_tables API:
   - disallow set element updates of bound anonymous set
   - disallow unbound anonymous set/chain at the end of transaction.
   - disallow updates of anonymous set.
   - disallow timeout configuration for anonymous sets.

8) Fix module reference leak in chain updates.

9) Fix nfnetlink_osf module autoload.

10) Fix deletion of basechain when NFTA_CHAIN_HOOK is specified as
    in iptables-nft.

This Netfilter batch is larger than usual at this stage, I am aware we
are fairly late in the -rc cycle, if you prefer to route them through
net-next, please let me know.

netfilter pull request 23-06-21

* tag 'nf-23-06-21' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf:
  netfilter: nf_tables: Fix for deleting base chains with payload
  netfilter: nfnetlink_osf: fix module autoload
  netfilter: nf_tables: drop module reference after updating chain
  netfilter: nf_tables: disallow timeout for anonymous sets
  netfilter: nf_tables: disallow updates of anonymous sets
  netfilter: nf_tables: reject unbound chain set before commit phase
  netfilter: nf_tables: reject unbound anonymous set before commit phase
  netfilter: nf_tables: disallow element updates of bound anonymous sets
  netfilter: nf_tables: fix underflow in object reference counter
  netfilter: nft_set_pipapo: .walk does not deal with generations
  netfilter: nf_tables: drop map element references from preparation phase
  netfilter: nf_tables: add NFT_TRANS_PREPARE_ERROR to deal with bound set/chain
  netfilter: nf_tables: fix chain binding transaction logic
  ipvs: align inner_mac_header for encapsulation
====================

Link: https://lore.kernel.org/r/20230621100731.68068-1-pablo@netfilter.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
13 months agorevert "net: align SO_RCVMARK required privileges with SO_MARK"
Maciej Żenczykowski [Sun, 18 Jun 2023 10:31:30 +0000 (03:31 -0700)]
revert "net: align SO_RCVMARK required privileges with SO_MARK"

This reverts commit 1f86123b9749 ("net: align SO_RCVMARK required
privileges with SO_MARK") because the reasoning in the commit message
is not really correct:
  SO_RCVMARK is used for 'reading' incoming skb mark (via cmsg), as such
  it is more equivalent to 'getsockopt(SO_MARK)' which has no priv check
  and retrieves the socket mark, rather than 'setsockopt(SO_MARK) which
  sets the socket mark and does require privs.

  Additionally incoming skb->mark may already be visible if
  sysctl_fwmark_reflect and/or sysctl_tcp_fwmark_accept are enabled.

  Furthermore, it is easier to block the getsockopt via bpf
  (either cgroup setsockopt hook, or via syscall filters)
  then to unblock it if it requires CAP_NET_RAW/ADMIN.

On Android the socket mark is (among other things) used to store
the network identifier a socket is bound to.  Setting it is privileged,
but retrieving it is not.  We'd like unprivileged userspace to be able
to read the network id of incoming packets (where mark is set via
iptables [to be moved to bpf])...

An alternative would be to add another sysctl to control whether
setting SO_RCVMARK is privilged or not.
(or even a MASK of which bits in the mark can be exposed)
But this seems like over-engineering...

Note: This is a non-trivial revert, due to later merged commit e42c7beee71d
("bpf: net: Consider has_current_bpf_ctx() when testing capable() in sk_setsockopt()")
which changed both 'ns_capable' into 'sockopt_ns_capable' calls.

Fixes: 1f86123b9749 ("net: align SO_RCVMARK required privileges with SO_MARK")
Cc: Larysa Zaremba <larysa.zaremba@intel.com>
Cc: Simon Horman <simon.horman@corigine.com>
Cc: Paolo Abeni <pabeni@redhat.com>
Cc: Eyal Birger <eyal.birger@gmail.com>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Patrick Rohr <prohr@google.com>
Signed-off-by: Maciej Żenczykowski <maze@google.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://lore.kernel.org/r/20230618103130.51628-1-maze@google.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
13 months agonet: wwan: iosm: Convert single instance struct member to flexible array
Kees Cook [Tue, 20 Jun 2023 19:42:38 +0000 (12:42 -0700)]
net: wwan: iosm: Convert single instance struct member to flexible array

struct mux_adth actually ends with multiple struct mux_adth_dg members.
This is seen both in the comments about the member:

/**
 * struct mux_adth - Structure of the Aggregated Datagram Table Header.
 ...
 * @dg: datagramm table with variable length
 */

and in the preparation for populating it:

                        adth_dg_size = offsetof(struct mux_adth, dg) +
                                        ul_adb->dg_count[i] * sizeof(*dg);
...
                        adth_dg_size -= offsetof(struct mux_adth, dg);
                        memcpy(&adth->dg, ul_adb->dg[i], adth_dg_size);

This was reported as a run-time false positive warning:

memcpy: detected field-spanning write (size 16) of single field "&adth->dg" at drivers/net/wwan/iosm/iosm_ipc_mux_codec.c:852 (size 8)

Adjust the struct mux_adth definition and associated sizeof() math; no binary
output differences are observed in the resulting object file.

Reported-by: Florian Klink <flokli@flokli.de>
Closes: https://lore.kernel.org/lkml/dbfa25f5-64c8-5574-4f5d-0151ba95d232@gmail.com/
Fixes: 1f52d7b62285 ("net: wwan: iosm: Enable M.2 7360 WWAN card support")
Cc: M Chetan Kumar <m.chetan.kumar@intel.com>
Cc: Bagas Sanjaya <bagasdotme@gmail.com>
Cc: Intel Corporation <linuxwwan@intel.com>
Cc: Loic Poulain <loic.poulain@linaro.org>
Cc: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Cc: Johannes Berg <johannes@sipsolutions.net>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Paolo Abeni <pabeni@redhat.com>
Cc: "Gustavo A. R. Silva" <gustavoars@kernel.org>
Cc: netdev@vger.kernel.org
Signed-off-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Link: https://lore.kernel.org/r/20230620194234.never.023-kees@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
13 months agosch_netem: acquire qdisc lock in netem_change()
Eric Dumazet [Tue, 20 Jun 2023 18:44:25 +0000 (18:44 +0000)]
sch_netem: acquire qdisc lock in netem_change()

syzbot managed to trigger a divide error [1] in netem.

It could happen if q->rate changes while netem_enqueue()
is running, since q->rate is read twice.

It turns out netem_change() always lacked proper synchronization.

[1]
divide error: 0000 [#1] SMP KASAN
CPU: 1 PID: 7867 Comm: syz-executor.1 Not tainted 6.1.30-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 05/25/2023
RIP: 0010:div64_u64 include/linux/math64.h:69 [inline]
RIP: 0010:packet_time_ns net/sched/sch_netem.c:357 [inline]
RIP: 0010:netem_enqueue+0x2067/0x36d0 net/sched/sch_netem.c:576
Code: 89 e2 48 69 da 00 ca 9a 3b 42 80 3c 28 00 4c 8b a4 24 88 00 00 00 74 0d 4c 89 e7 e8 c3 4f 3b fd 48 8b 4c 24 18 48 89 d8 31 d2 <49> f7 34 24 49 01 c7 4c 8b 64 24 48 4d 01 f7 4c 89 e3 48 c1 eb 03
RSP: 0018:ffffc9000dccea60 EFLAGS: 00010246
RAX: 000001a442624200 RBX: 000001a442624200 RCX: ffff888108a4f000
RDX: 0000000000000000 RSI: 000000000000070d RDI: 000000000000070d
RBP: ffffc9000dcceb90 R08: ffffffff849c5e26 R09: fffffbfff10e1297
R10: 0000000000000000 R11: dffffc0000000001 R12: ffff888108a4f358
R13: dffffc0000000000 R14: 0000001a8cd9a7ec R15: 0000000000000000
FS: 00007fa73fe18700(0000) GS:ffff8881f6b00000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007fa73fdf7718 CR3: 000000011d36e000 CR4: 0000000000350ee0
Call Trace:
<TASK>
[<ffffffff84714385>] __dev_xmit_skb net/core/dev.c:3931 [inline]
[<ffffffff84714385>] __dev_queue_xmit+0xcf5/0x3370 net/core/dev.c:4290
[<ffffffff84d22df2>] dev_queue_xmit include/linux/netdevice.h:3030 [inline]
[<ffffffff84d22df2>] neigh_hh_output include/net/neighbour.h:531 [inline]
[<ffffffff84d22df2>] neigh_output include/net/neighbour.h:545 [inline]
[<ffffffff84d22df2>] ip_finish_output2+0xb92/0x10d0 net/ipv4/ip_output.c:235
[<ffffffff84d21e63>] __ip_finish_output+0xc3/0x2b0
[<ffffffff84d10a81>] ip_finish_output+0x31/0x2a0 net/ipv4/ip_output.c:323
[<ffffffff84d10f14>] NF_HOOK_COND include/linux/netfilter.h:298 [inline]
[<ffffffff84d10f14>] ip_output+0x224/0x2a0 net/ipv4/ip_output.c:437
[<ffffffff84d123b5>] dst_output include/net/dst.h:444 [inline]
[<ffffffff84d123b5>] ip_local_out net/ipv4/ip_output.c:127 [inline]
[<ffffffff84d123b5>] __ip_queue_xmit+0x1425/0x2000 net/ipv4/ip_output.c:542
[<ffffffff84d12fdc>] ip_queue_xmit+0x4c/0x70 net/ipv4/ip_output.c:556

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Stephen Hemminger <stephen@networkplumber.org>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Jiri Pirko <jiri@resnulli.us>
Reviewed-by: Jamal Hadi Salim <jhs@mojatatu.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Link: https://lore.kernel.org/r/20230620184425.1179809-1-edumazet@google.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
13 months agoselftests: forwarding: Fix race condition in mirror installation
Danielle Ratson [Tue, 20 Jun 2023 12:45:15 +0000 (14:45 +0200)]
selftests: forwarding: Fix race condition in mirror installation

When mirroring to a gretap in hardware the device expects to be
programmed with the egress port and all the encapsulating headers. This
requires the driver to resolve the path the packet will take in the
software data path and program the device accordingly.

If the path cannot be resolved (in this case because of an unresolved
neighbor), then mirror installation fails until the path is resolved.
This results in a race that causes the test to sometimes fail.

Fix this by setting the neighbor's state to permanent in a couple of
tests, so that it is always valid.

Fixes: 35c31d5c323f ("selftests: forwarding: Test mirror-to-gretap w/ UL 802.1d")
Fixes: 239e754af854 ("selftests: forwarding: Test mirror-to-gretap w/ UL 802.1q")
Signed-off-by: Danielle Ratson <danieller@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Petr Machata <petrm@nvidia.com>
Link: https://lore.kernel.org/r/268816ac729cb6028c7a34d4dda6f4ec7af55333.1687264607.git.petrm@nvidia.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
13 months agowifi: mac80211: report all unusable beacon frames
Benjamin Berg [Wed, 21 Jun 2023 12:05:44 +0000 (14:05 +0200)]
wifi: mac80211: report all unusable beacon frames

Properly check for RX_DROP_UNUSABLE now that the new drop reason
infrastructure is used. Without this change, the comparison will always
be false as a more specific reason is given in the lower bits of result.

Fixes: baa951a1c177 ("mac80211: use the new drop reasons infrastructure")
Signed-off-by: Benjamin Berg <benjamin.berg@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Link: https://lore.kernel.org/r/20230621120543.412920-2-johannes@sipsolutions.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agoMerge branch 'mptcp-fixes-for-6-4'
Jakub Kicinski [Thu, 22 Jun 2023 05:44:59 +0000 (22:44 -0700)]
Merge branch 'mptcp-fixes-for-6-4'

Matthieu Baerts says:

====================
mptcp: fixes for 6.4

Patch 1 correctly handles disconnect() failures that can happen in some
specific cases: now the socket state is set as unconnected as expected.
That fixes an issue introduced in v6.2.

Patch 2 fixes a divide by zero bug in mptcp_recvmsg() with a fix similar
to a recent one from Eric Dumazet for TCP introducing sk_wait_pending
flag. It should address an issue present in MPTCP from almost the
beginning, from v5.9.

Patch 3 fixes a possible list corruption on passive MPJ even if the race
seems very unlikely, better be safe than sorry. The possible issue is
present from v5.17.

Patch 4 consolidates fallback and non fallback state machines to avoid
leaking some MPTCP sockets. The fix is likely needed for versions from
v5.11.

Patch 5 drops code that is no longer used after the introduction of
patch 4/6. This is not really a fix but this patch can probably land in
the -net tree as well not to leave unused code.

Patch 6 ensures listeners are unhashed before updating their sk status
to avoid possible deadlocks when diag info are going to be retrieved
with a lock. Even if it should not be visible with the way we are
currently getting diag info, the issue is present from v5.17.
====================

Link: https://lore.kernel.org/r/20230620-upstream-net-20230620-misc-fixes-for-v6-4-v1-0-f36aa5eae8b9@tessares.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agomptcp: ensure listener is unhashed before updating the sk status
Paolo Abeni [Tue, 20 Jun 2023 16:24:23 +0000 (18:24 +0200)]
mptcp: ensure listener is unhashed before updating the sk status

The MPTCP protocol access the listener subflow in a lockless
manner in a couple of places (poll, diag). That works only if
the msk itself leaves the listener status only after that the
subflow itself has been closed/disconnected. Otherwise we risk
deadlock in diag, as reported by Christoph.

Address the issue ensuring that the first subflow (the listener
one) is always disconnected before updating the msk socket status.

Reported-by: Christoph Paasch <cpaasch@apple.com>
Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/407
Fixes: b29fcfb54cd7 ("mptcp: full disconnect implementation")
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agomptcp: drop legacy code around RX EOF
Paolo Abeni [Tue, 20 Jun 2023 16:24:22 +0000 (18:24 +0200)]
mptcp: drop legacy code around RX EOF

Thanks to the previous patch -- "mptcp: consolidate fallback and non
fallback state machine" -- we can finally drop the "temporary hack"
used to detect rx eof.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agomptcp: consolidate fallback and non fallback state machine
Paolo Abeni [Tue, 20 Jun 2023 16:24:21 +0000 (18:24 +0200)]
mptcp: consolidate fallback and non fallback state machine

An orphaned msk releases the used resources via the worker,
when the latter first see the msk in CLOSED status.

If the msk status transitions to TCP_CLOSE in the release callback
invoked by the worker's final release_sock(), such instance of the
workqueue will not take any action.

Additionally the MPTCP code prevents scheduling the worker once the
socket reaches the CLOSE status: such msk resources will be leaked.

The only code path that can trigger the above scenario is the
__mptcp_check_send_data_fin() in fallback mode.

Address the issue removing the special handling of fallback socket
in __mptcp_check_send_data_fin(), consolidating the state machine
for fallback and non fallback socket.

Since non-fallback sockets do not send and do not receive data_fin,
the mptcp code can update the msk internal status to match the next
step in the SM every time data fin (ack) should be generated or
received.

As a consequence we can remove a bunch of checks for fallback from
the fastpath.

Fixes: 6e628cd3a8f7 ("mptcp: use mptcp release_cb for delayed tasks")
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agomptcp: fix possible list corruption on passive MPJ
Paolo Abeni [Tue, 20 Jun 2023 16:24:20 +0000 (18:24 +0200)]
mptcp: fix possible list corruption on passive MPJ

At passive MPJ time, if the msk socket lock is held by the user,
the new subflow is appended to the msk->join_list under the msk
data lock.

In mptcp_release_cb()/__mptcp_flush_join_list(), the subflows in
that list are moved from the join_list into the conn_list under the
msk socket lock.

Append and removal could race, possibly corrupting such list.
Address the issue splicing the join list into a temporary one while
still under the msk data lock.

Found by code inspection, the race itself should be almost impossible
to trigger in practice.

Fixes: 3e5014909b56 ("mptcp: cleanup MPJ subflow list handling")
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agomptcp: fix possible divide by zero in recvmsg()
Paolo Abeni [Tue, 20 Jun 2023 16:24:19 +0000 (18:24 +0200)]
mptcp: fix possible divide by zero in recvmsg()

Christoph reported a divide by zero bug in mptcp_recvmsg():

divide error: 0000 [#1] PREEMPT SMP
CPU: 1 PID: 19978 Comm: syz-executor.6 Not tainted 6.4.0-rc2-gffcc7899081b #20
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.11.0-2.el7 04/01/2014
RIP: 0010:__tcp_select_window+0x30e/0x420 net/ipv4/tcp_output.c:3018
Code: 11 ff 0f b7 cd c1 e9 0c b8 ff ff ff ff d3 e0 89 c1 f7 d1 01 cb 21 c3 eb 17 e8 2e 83 11 ff 31 db eb 0e e8 25 83 11 ff 89 d8 99 <f7> 7c 24 04 29 d3 65 48 8b 04 25 28 00 00 00 48 3b 44 24 10 75 60
RSP: 0018:ffffc90000a07a18 EFLAGS: 00010246
RAX: 000000000000ffd7 RBX: 000000000000ffd7 RCX: 0000000000040000
RDX: 0000000000000000 RSI: 000000000003ffff RDI: 0000000000040000
RBP: 000000000000ffd7 R08: ffffffff820cf297 R09: 0000000000000001
R10: 0000000000000000 R11: ffffffff8103d1a0 R12: 0000000000003f00
R13: 0000000000300000 R14: ffff888101cf3540 R15: 0000000000180000
FS:  00007f9af4c09640(0000) GS:ffff88813bd00000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000001b33824000 CR3: 000000012f241001 CR4: 0000000000170ee0
Call Trace:
 <TASK>
 __tcp_cleanup_rbuf+0x138/0x1d0 net/ipv4/tcp.c:1611
 mptcp_recvmsg+0xcb8/0xdd0 net/mptcp/protocol.c:2034
 inet_recvmsg+0x127/0x1f0 net/ipv4/af_inet.c:861
 ____sys_recvmsg+0x269/0x2b0 net/socket.c:1019
 ___sys_recvmsg+0xe6/0x260 net/socket.c:2764
 do_recvmmsg+0x1a5/0x470 net/socket.c:2858
 __do_sys_recvmmsg net/socket.c:2937 [inline]
 __se_sys_recvmmsg net/socket.c:2953 [inline]
 __x64_sys_recvmmsg+0xa6/0x130 net/socket.c:2953
 do_syscall_x64 arch/x86/entry/common.c:50 [inline]
 do_syscall_64+0x47/0xa0 arch/x86/entry/common.c:80
 entry_SYSCALL_64_after_hwframe+0x72/0xdc
RIP: 0033:0x7f9af58fc6a9
Code: 5c c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 4f 37 0d 00 f7 d8 64 89 01 48
RSP: 002b:00007f9af4c08cd8 EFLAGS: 00000246 ORIG_RAX: 000000000000012b
RAX: ffffffffffffffda RBX: 00000000006bc050 RCX: 00007f9af58fc6a9
RDX: 0000000000000001 RSI: 0000000020000140 RDI: 0000000000000004
RBP: 0000000000000000 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000f00 R11: 0000000000000246 R12: 00000000006bc05c
R13: fffffffffffffea8 R14: 00000000006bc050 R15: 000000000001fe40
 </TASK>

mptcp_recvmsg is allowed to release the msk socket lock when
blocking, and before re-acquiring it another thread could have
switched the sock to TCP_LISTEN status - with a prior
connect(AF_UNSPEC) - also clearing icsk_ack.rcv_mss.

Address the issue preventing the disconnect if some other process is
concurrently performing a blocking syscall on the same socket, alike
commit 4faeee0cf8a5 ("tcp: deny tcp_disconnect() when threads are waiting").

Fixes: a6b118febbab ("mptcp: add receive buffer auto-tuning")
Cc: stable@vger.kernel.org
Reported-by: Christoph Paasch <cpaasch@apple.com>
Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/404
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Tested-by: Christoph Paasch <cpaasch@apple.com>
Reviewed-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agomptcp: handle correctly disconnect() failures
Paolo Abeni [Tue, 20 Jun 2023 16:24:18 +0000 (18:24 +0200)]
mptcp: handle correctly disconnect() failures

Currently the mptcp code has assumes that disconnect() can fail only
at mptcp_sendmsg_fastopen() time - to avoid a deadlock scenario - and
don't even bother returning an error code.

Soon mptcp_disconnect() will handle more error conditions: let's track
them explicitly.

As a bonus, explicitly annotate TCP-level disconnect as not failing:
the mptcp code never blocks for event on the subflows.

Fixes: 7d803344fdc3 ("mptcp: fix deadlock in fastopen error path")
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Tested-by: Christoph Paasch <cpaasch@apple.com>
Reviewed-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agoMerge tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf
Jakub Kicinski [Wed, 21 Jun 2023 20:59:46 +0000 (13:59 -0700)]
Merge tag 'for-netdev' of https://git./linux/kernel/git/bpf/bpf

Daniel Borkmann says:

====================
pull-request: bpf 2023-06-21

We've added 7 non-merge commits during the last 14 day(s) which contain
a total of 7 files changed, 181 insertions(+), 15 deletions(-).

The main changes are:

1) Fix a verifier id tracking issue with scalars upon spill,
   from Maxim Mikityanskiy.

2) Fix NULL dereference if an exception is generated while a BPF
   subprogram is running, from Krister Johansen.

3) Fix a BTF verification failure when compiling kernel with LLVM_IAS=0,
   from Florent Revest.

4) Fix expected_attach_type enforcement for kprobe_multi link,
   from Jiri Olsa.

5) Fix a bpf_jit_dump issue for x86_64 to pick the correct JITed image,
   from Yonghong Song.

* tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf:
  bpf: Force kprobe multi expected_attach_type for kprobe_multi link
  bpf/btf: Accept function names that contain dots
  selftests/bpf: add a test for subprogram extables
  bpf: ensure main program has an extable
  bpf: Fix a bpf_jit_dump issue for x86_64 with sysctl bpf_jit_enable.
  selftests/bpf: Add test cases to assert proper ID tracking on spill
  bpf: Fix verifier id tracking of scalars on spill
====================

Link: https://lore.kernel.org/r/20230621101116.16122-1-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agobpf: Force kprobe multi expected_attach_type for kprobe_multi link
Jiri Olsa [Sun, 18 Jun 2023 13:14:14 +0000 (15:14 +0200)]
bpf: Force kprobe multi expected_attach_type for kprobe_multi link

We currently allow to create perf link for program with
expected_attach_type == BPF_TRACE_KPROBE_MULTI.

This will cause crash when we call helpers like get_attach_cookie or
get_func_ip in such program, because it will call the kprobe_multi's
version (current->bpf_ctx context setup) of those helpers while it
expects perf_link's current->bpf_ctx context setup.

Making sure that we use BPF_TRACE_KPROBE_MULTI expected_attach_type
only for programs attaching through kprobe_multi link.

Fixes: ca74823c6e16 ("bpf: Add cookie support to programs attached with kprobe multi link")
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20230618131414.75649-1-jolsa@kernel.org
13 months agobpf/btf: Accept function names that contain dots
Florent Revest [Thu, 15 Jun 2023 14:56:07 +0000 (16:56 +0200)]
bpf/btf: Accept function names that contain dots

When building a kernel with LLVM=1, LLVM_IAS=0 and CONFIG_KASAN=y, LLVM
leaves DWARF tags for the "asan.module_ctor" & co symbols. In turn,
pahole creates BTF_KIND_FUNC entries for these and this makes the BTF
metadata validation fail because they contain a dot.

In a dramatic turn of event, this BTF verification failure can cause
the netfilter_bpf initialization to fail, causing netfilter_core to
free the netfilter_helper hashmap and netfilter_ftp to trigger a
use-after-free. The risk of u-a-f in netfilter will be addressed
separately but the existence of "asan.module_ctor" debug info under some
build conditions sounds like a good enough reason to accept functions
that contain dots in BTF.

Although using only LLVM=1 is the recommended way to compile clang-based
kernels, users can certainly do LLVM=1, LLVM_IAS=0 as well and we still
try to support that combination according to Nick. To clarify:

  - > v5.10 kernel, LLVM=1 (LLVM_IAS=0 is not the default) is recommended,
    but user can still have LLVM=1, LLVM_IAS=0 to trigger the issue

  - <= 5.10 kernel, LLVM=1 (LLVM_IAS=0 is the default) is recommended in
    which case GNU as will be used

Fixes: 1dc92851849c ("bpf: kernel side support for BTF Var and DataSec")
Signed-off-by: Florent Revest <revest@chromium.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Cc: Yonghong Song <yhs@meta.com>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Link: https://lore.kernel.org/bpf/20230615145607.3469985-1-revest@chromium.org
13 months agoRevert "net: phy: dp83867: perform soft reset and retain established link"
Francesco Dolcini [Mon, 19 Jun 2023 15:44:35 +0000 (17:44 +0200)]
Revert "net: phy: dp83867: perform soft reset and retain established link"

This reverts commit da9ef50f545f86ffe6ff786174d26500c4db737a.

This fixes a regression in which the link would come up, but no
communication was possible.

The reverted commit was also removing a comment about
DP83867_PHYCR_FORCE_LINK_GOOD, this is not added back in this commits
since it seems that this is unrelated to the original code change.

Closes: https://lore.kernel.org/all/ZGuDJos8D7N0J6Z2@francesco-nb.int.toradex.com/
Fixes: da9ef50f545f ("net: phy: dp83867: perform soft reset and retain established link")
Signed-off-by: Francesco Dolcini <francesco.dolcini@toradex.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Praneeth Bajjuri <praneeth@ti.com>
Link: https://lore.kernel.org/r/20230619154435.355485-1-francesco@dolcini.it
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agonet: mdio: fix the wrong parameters
Jiawen Wu [Mon, 19 Jun 2023 09:49:48 +0000 (17:49 +0800)]
net: mdio: fix the wrong parameters

PHY address and device address are passed in the wrong order.

Cc: stable@vger.kernel.org
Fixes: 4e4aafcddbbf ("net: mdio: Add dedicated C45 API to MDIO bus drivers")
Signed-off-by: Jiawen Wu <jiawenwu@trustnetic.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://lore.kernel.org/r/20230619094948.84452-1-jiawenwu@trustnetic.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agonetfilter: nf_tables: Fix for deleting base chains with payload
Phil Sutter [Fri, 16 Jun 2023 15:56:11 +0000 (17:56 +0200)]
netfilter: nf_tables: Fix for deleting base chains with payload

When deleting a base chain, iptables-nft simply submits the whole chain
to the kernel, including the NFTA_CHAIN_HOOK attribute. The new code
added by fixed commit then turned this into a chain update, destroying
the hook but not the chain itself. Detect the situation by checking if
the chain type is either netdev or inet/ingress.

Fixes: 7d937b107108f ("netfilter: nf_tables: support for deleting devices in an existing netdev chain")
Signed-off-by: Phil Sutter <phil@nwl.cc>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
13 months agonetfilter: nfnetlink_osf: fix module autoload
Pablo Neira Ayuso [Thu, 15 Jun 2023 08:14:25 +0000 (10:14 +0200)]
netfilter: nfnetlink_osf: fix module autoload

Move the alias from xt_osf to nfnetlink_osf.

Fixes: f9324952088f ("netfilter: nfnetlink_osf: extract nfnetlink_subsystem code from xt_osf.c")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
13 months agonetfilter: nf_tables: drop module reference after updating chain
Pablo Neira Ayuso [Wed, 14 Jun 2023 21:20:18 +0000 (23:20 +0200)]
netfilter: nf_tables: drop module reference after updating chain

Otherwise the module reference counter is leaked.

Fixes b9703ed44ffb ("netfilter: nf_tables: support for adding new devices to an existing netdev chain")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
13 months agonetfilter: nf_tables: disallow timeout for anonymous sets
Pablo Neira Ayuso [Fri, 16 Jun 2023 13:22:18 +0000 (15:22 +0200)]
netfilter: nf_tables: disallow timeout for anonymous sets

Never used from userspace, disallow these parameters.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
13 months agonetfilter: nf_tables: disallow updates of anonymous sets
Pablo Neira Ayuso [Fri, 16 Jun 2023 13:22:01 +0000 (15:22 +0200)]
netfilter: nf_tables: disallow updates of anonymous sets

Disallow updates of set timeout and garbage collection parameters for
anonymous sets.

Fixes: 123b99619cca ("netfilter: nf_tables: honor set timeout and garbage collection updates")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
13 months agonetfilter: nf_tables: reject unbound chain set before commit phase
Pablo Neira Ayuso [Fri, 16 Jun 2023 13:21:39 +0000 (15:21 +0200)]
netfilter: nf_tables: reject unbound chain set before commit phase

Use binding list to track set transaction and to check for unbound
chains before entering the commit phase.

Bail out if chain binding remain unused before entering the commit
step.

Fixes: d0e2c7de92c7 ("netfilter: nf_tables: add NFT_CHAIN_BINDING")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
13 months agonetfilter: nf_tables: reject unbound anonymous set before commit phase
Pablo Neira Ayuso [Fri, 16 Jun 2023 13:21:33 +0000 (15:21 +0200)]
netfilter: nf_tables: reject unbound anonymous set before commit phase

Add a new list to track set transaction and to check for unbound
anonymous sets before entering the commit phase.

Bail out at the end of the transaction handling if an anonymous set
remains unbound.

Fixes: 96518518cc41 ("netfilter: add nftables")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
13 months agonetfilter: nf_tables: disallow element updates of bound anonymous sets
Pablo Neira Ayuso [Fri, 16 Jun 2023 13:20:16 +0000 (15:20 +0200)]
netfilter: nf_tables: disallow element updates of bound anonymous sets

Anonymous sets come with NFT_SET_CONSTANT from userspace. Although API
allows to create anonymous sets without NFT_SET_CONSTANT, it makes no
sense to allow to add and to delete elements for bound anonymous sets.

Fixes: 96518518cc41 ("netfilter: add nftables")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
13 months agonetfilter: nf_tables: fix underflow in object reference counter
Pablo Neira Ayuso [Fri, 16 Jun 2023 13:20:08 +0000 (15:20 +0200)]
netfilter: nf_tables: fix underflow in object reference counter

Since ("netfilter: nf_tables: drop map element references from
preparation phase"), integration with commit protocol is better,
therefore drop the workaround that b91d90368837 ("netfilter: nf_tables:
fix leaking object reference count") provides.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
13 months agonetfilter: nft_set_pipapo: .walk does not deal with generations
Pablo Neira Ayuso [Fri, 16 Jun 2023 13:20:04 +0000 (15:20 +0200)]
netfilter: nft_set_pipapo: .walk does not deal with generations

The .walk callback iterates over the current active set, but it might be
useful to iterate over the next generation set. Use the generation mask
to determine what set view (either current or next generation) is use
for the walk iteration.

Fixes: 3c4287f62044 ("nf_tables: Add set type for arbitrary concatenation of ranges")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
13 months agonetfilter: nf_tables: drop map element references from preparation phase
Pablo Neira Ayuso [Fri, 16 Jun 2023 12:51:49 +0000 (14:51 +0200)]
netfilter: nf_tables: drop map element references from preparation phase

set .destroy callback releases the references to other objects in maps.
This is very late and it results in spurious EBUSY errors. Drop refcount
from the preparation phase instead, update set backend not to drop
reference counter from set .destroy path.

Exceptions: NFT_TRANS_PREPARE_ERROR does not require to drop the
reference counter because the transaction abort path releases the map
references for each element since the set is unbound. The abort path
also deals with releasing reference counter for new elements added to
unbound sets.

Fixes: 591054469b3e ("netfilter: nf_tables: revisit chain/object refcounting from elements")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
13 months agonetfilter: nf_tables: add NFT_TRANS_PREPARE_ERROR to deal with bound set/chain
Pablo Neira Ayuso [Fri, 16 Jun 2023 12:45:26 +0000 (14:45 +0200)]
netfilter: nf_tables: add NFT_TRANS_PREPARE_ERROR to deal with bound set/chain

Add a new state to deal with rule expressions deactivation from the
newrule error path, otherwise the anonymous set remains in the list in
inactive state for the next generation. Mark the set/chain transaction
as unbound so the abort path releases this object, set it as inactive in
the next generation so it is not reachable anymore from this transaction
and reference counter is dropped.

Fixes: 1240eb93f061 ("netfilter: nf_tables: incorrect error path handling with NFT_MSG_NEWRULE")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
13 months agonetfilter: nf_tables: fix chain binding transaction logic
Pablo Neira Ayuso [Fri, 16 Jun 2023 12:45:22 +0000 (14:45 +0200)]
netfilter: nf_tables: fix chain binding transaction logic

Add bound flag to rule and chain transactions as in 6a0a8d10a366
("netfilter: nf_tables: use-after-free in failing rule with bound set")
to skip them in case that the chain is already bound from the abort
path.

This patch fixes an imbalance in the chain use refcnt that triggers a
WARN_ON on the table and chain destroy path.

This patch also disallows nested chain bindings, which is not
supported from userspace.

The logic to deal with chain binding in nft_data_hold() and
nft_data_release() is not correct. The NFT_TRANS_PREPARE state needs a
special handling in case a chain is bound but next expressions in the
same rule fail to initialize as described by 1240eb93f061 ("netfilter:
nf_tables: incorrect error path handling with NFT_MSG_NEWRULE").

The chain is left bound if rule construction fails, so the objects
stored in this chain (and the chain itself) are released by the
transaction records from the abort path, follow up patch ("netfilter:
nf_tables: add NFT_TRANS_PREPARE_ERROR to deal with bound set/chain")
completes this error handling.

When deleting an existing rule, chain bound flag is set off so the
rule expression .destroy path releases the objects.

Fixes: d0e2c7de92c7 ("netfilter: nf_tables: add NFT_CHAIN_BINDING")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
13 months agowifi: iwlwifi: pcie: Handle SO-F device for PCI id 0x7AF0
Mukesh Sisodiya [Mon, 19 Jun 2023 15:02:34 +0000 (17:02 +0200)]
wifi: iwlwifi: pcie: Handle SO-F device for PCI id 0x7AF0

Add support for AX1690i and AX1690s devices with
PCIE id 0x7AF0.

Cc: stable@vger.kernel.org # 6.1+
Signed-off-by: Mukesh Sisodiya <mukesh.sisodiya@intel.com>
Signed-off-by: Gregory Greenman <gregory.greenman@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Link: https://lore.kernel.org/r/20230619150233.461290-2-johannes@sipsolutions.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agobe2net: Extend xmit workaround to BE3 chip
Ross Lagerwall [Fri, 16 Jun 2023 16:45:49 +0000 (17:45 +0100)]
be2net: Extend xmit workaround to BE3 chip

We have seen a bug where the NIC incorrectly changes the length in the
IP header of a padded packet to include the padding bytes. The driver
already has a workaround for this so do the workaround for this NIC too.
This resolves the issue.

The NIC in question identifies itself as follows:

[    8.828494] be2net 0000:02:00.0: FW version is 10.7.110.31
[    8.834759] be2net 0000:02:00.0: Emulex OneConnect(be3): PF FLEX10 port 1

02:00.0 Ethernet controller: Emulex Corporation OneConnect 10Gb NIC (be3) (rev 01)

Fixes: ca34fe38f06d ("be2net: fix wrong usage of adapter->generation")
Signed-off-by: Ross Lagerwall <ross.lagerwall@citrix.com>
Link: https://lore.kernel.org/r/20230616164549.2863037-1-ross.lagerwall@citrix.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agoMerge tag 'ipsec-2023-06-20' of git://git.kernel.org/pub/scm/linux/kernel/git/klasser...
David S. Miller [Tue, 20 Jun 2023 12:33:50 +0000 (13:33 +0100)]
Merge tag 'ipsec-2023-06-20' of git://git./linux/kernel/git/klassert/ipsec

ipsec-2023-06-20

13 months agoMerge branch 'dsa-mt7530-fixes'
David S. Miller [Tue, 20 Jun 2023 08:40:26 +0000 (09:40 +0100)]
Merge branch 'dsa-mt7530-fixes'

Arınç ÜNAL says:

====================
net: dsa: mt7530: fix multiple CPU ports, BPDU and LLDP handling

This patch series fixes all non-theoretical issues regarding multiple CPU
ports and the handling of LLDP frames and BPDUs.

I am adding me as a maintainer, I've got some code improvements on the way.
I will keep an eye on this driver and the patches submitted for it in the
future.

Arınç

v6:
- Change a small portion of the comment in the diff on "net: dsa: mt7530:
  set all CPU ports in MT7531_CPU_PMAP" with Russell's suggestion.
- Change the patch log of "net: dsa: mt7530: fix trapping frames on
  non-MT7621 SoC MT7530 switch" with Vladimir's suggestion.
- Group the code for trapping frames into a common function and call that.
- Add Vladimir and Russell's reviewed-by tags to where they're given.

v5:
- Change the comment in the diff on the first patch with Russell's words.
- Change the patch log of the first patch to state that the patch is just
  preparatory work for change "net: dsa: introduce
  preferred_default_local_cpu_port and use on MT7530" and not a fix to an
  existing problem on the code base.
- Remove the "net: dsa: mt7530: fix trapping frames with multiple CPU ports
  on MT7530" patch. It fixes a theoretical issue, therefore it is net-next
  material.
- Remove unnecessary information from the patch logs. Remove the enum
  renaming change.
- Strengthen the point of the "net: dsa: introduce
  preferred_default_local_cpu_port and use on MT7530" patch.

v4: Make the patch logs and my comments in the code easier to understand.
v3: Fix the from header on the patches. Write a cover letter.
v2: Add patches to fix the handling of LLDP frames and BPDUs.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
13 months agoMAINTAINERS: add me as maintainer of MEDIATEK SWITCH DRIVER
Arınç ÜNAL [Sat, 17 Jun 2023 06:26:49 +0000 (09:26 +0300)]
MAINTAINERS: add me as maintainer of MEDIATEK SWITCH DRIVER

Add me as a maintainer of the MediaTek MT7530 DSA subdriver.

List maintainers in alphabetical order by first name.

Signed-off-by: Arınç ÜNAL <arinc.unal@arinc9.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Matthias Brugger <matthias.bgg@gmail.com>
Reviewed-by: Matthias Brugger <matthias.bgg@gmail.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 months agonet: dsa: introduce preferred_default_local_cpu_port and use on MT7530
Vladimir Oltean [Sat, 17 Jun 2023 06:26:48 +0000 (09:26 +0300)]
net: dsa: introduce preferred_default_local_cpu_port and use on MT7530

Since the introduction of the OF bindings, DSA has always had a policy that
in case multiple CPU ports are present in the device tree, the numerically
smallest one is always chosen.

The MT7530 switch family, except the switch on the MT7988 SoC, has 2 CPU
ports, 5 and 6, where port 6 is preferable on the MT7531BE switch because
it has higher bandwidth.

The MT7530 driver developers had 3 options:
- to modify DSA when the MT7531 switch support was introduced, such as to
  prefer the better port
- to declare both CPU ports in device trees as CPU ports, and live with the
  sub-optimal performance resulting from not preferring the better port
- to declare just port 6 in the device tree as a CPU port

Of course they chose the path of least resistance (3rd option), kicking the
can down the road. The hardware description in the device tree is supposed
to be stable - developers are not supposed to adopt the strategy of
piecemeal hardware description, where the device tree is updated in
lockstep with the features that the kernel currently supports.

Now, as a result of the fact that they did that, any attempts to modify the
device tree and describe both CPU ports as CPU ports would make DSA change
its default selection from port 6 to 5, effectively resulting in a
performance degradation visible to users with the MT7531BE switch as can be
seen below.

Without preferring port 6:

[ ID][Role] Interval           Transfer     Bitrate         Retr
[  5][TX-C]   0.00-20.00  sec   374 MBytes   157 Mbits/sec  734    sender
[  5][TX-C]   0.00-20.00  sec   373 MBytes   156 Mbits/sec    receiver
[  7][RX-C]   0.00-20.00  sec  1.81 GBytes   778 Mbits/sec    0    sender
[  7][RX-C]   0.00-20.00  sec  1.81 GBytes   777 Mbits/sec    receiver

With preferring port 6:

[ ID][Role] Interval           Transfer     Bitrate         Retr
[  5][TX-C]   0.00-20.00  sec  1.99 GBytes   856 Mbits/sec  273    sender
[  5][TX-C]   0.00-20.00  sec  1.99 GBytes   855 Mbits/sec    receiver
[  7][RX-C]   0.00-20.00  sec  1.72 GBytes   737 Mbits/sec   15    sender
[  7][RX-C]   0.00-20.00  sec  1.71 GBytes   736 Mbits/sec    receiver

Using one port for WAN and the other ports for LAN is a very popular use
case which is what this test emulates.

As such, this change proposes that we retroactively modify stable kernels
(which don't support the modification of the CPU port assignments, so as to
let user space fix the problem and restore the throughput) to keep the
mt7530 driver preferring port 6 even with device trees where the hardware
is more fully described.

Fixes: c288575f7810 ("net: dsa: mt7530: Add the support of MT7531 switch")
Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Arınç ÜNAL <arinc.unal@arinc9.com>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 months agonet: dsa: mt7530: fix handling of LLDP frames
Arınç ÜNAL [Sat, 17 Jun 2023 06:26:47 +0000 (09:26 +0300)]
net: dsa: mt7530: fix handling of LLDP frames

LLDP frames are link-local frames, therefore they must be trapped to the
CPU port. Currently, the MT753X switches treat LLDP frames as regular
multicast frames, therefore flooding them to user ports. To fix this, set
LLDP frames to be trapped to the CPU port(s).

Fixes: b8f126a8d543 ("net-next: dsa: add dsa support for Mediatek MT7530 switch")
Signed-off-by: Arınç ÜNAL <arinc.unal@arinc9.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 months agonet: dsa: mt7530: fix handling of BPDUs on MT7530 switch
Arınç ÜNAL [Sat, 17 Jun 2023 06:26:46 +0000 (09:26 +0300)]
net: dsa: mt7530: fix handling of BPDUs on MT7530 switch

BPDUs are link-local frames, therefore they must be trapped to the CPU
port. Currently, the MT7530 switch treats BPDUs as regular multicast
frames, therefore flooding them to user ports. To fix this, set BPDUs to be
trapped to the CPU port. Group this on mt7530_setup() and
mt7531_setup_common() into mt753x_trap_frames() and call that.

Fixes: b8f126a8d543 ("net-next: dsa: add dsa support for Mediatek MT7530 switch")
Signed-off-by: Arınç ÜNAL <arinc.unal@arinc9.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 months agonet: dsa: mt7530: fix trapping frames on non-MT7621 SoC MT7530 switch
Arınç ÜNAL [Sat, 17 Jun 2023 06:26:45 +0000 (09:26 +0300)]
net: dsa: mt7530: fix trapping frames on non-MT7621 SoC MT7530 switch

All MT7530 switch IP variants share the MT7530_MFC register, but the
current driver only writes it for the switch variant that is integrated in
the MT7621 SoC. Modify the code to include all MT7530 derivatives.

Fixes: b8f126a8d543 ("net-next: dsa: add dsa support for Mediatek MT7530 switch")
Suggested-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Arınç ÜNAL <arinc.unal@arinc9.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 months agonet: dsa: mt7530: set all CPU ports in MT7531_CPU_PMAP
Arınç ÜNAL [Sat, 17 Jun 2023 06:26:44 +0000 (09:26 +0300)]
net: dsa: mt7530: set all CPU ports in MT7531_CPU_PMAP

MT7531_CPU_PMAP represents the destination port mask for trapped-to-CPU
frames (further restricted by PCR_MATRIX).

Currently the driver sets the first CPU port as the single port in this bit
mask, which works fine regardless of whether the device tree defines port
5, 6 or 5+6 as CPU ports. This is because the logic coincides with DSA's
logic of picking the first CPU port as the CPU port that all user ports are
affine to, by default.

An upcoming change would like to influence DSA's selection of the default
CPU port to no longer be the first one, and in that case, this logic needs
adaptation.

Since there is no observed leakage or duplication of frames if all CPU
ports are defined in this bit mask, simply include them all.

Suggested-by: Russell King (Oracle) <linux@armlinux.org.uk>
Suggested-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Arınç ÜNAL <arinc.unal@arinc9.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 months agonet: dpaa2-mac: add 25gbase-r support
Josua Mayer [Fri, 16 Jun 2023 11:14:14 +0000 (14:14 +0300)]
net: dpaa2-mac: add 25gbase-r support

Layerscape MACs support 25Gbps network speed with dpmac "CAUI" mode.
Add the mappings between DPMAC_ETH_IF_* and HY_INTERFACE_MODE_*, as well
as the 25000 mac capability.

Tested on SolidRun LX2162a Clearfog, serdes 1 protocol 18.

Signed-off-by: Josua Mayer <josua@solid-run.com>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 months agoMerge tag 'ieee802154-for-net-2023-06-19' of git://git.kernel.org/pub/scm/linux/kerne...
David S. Miller [Tue, 20 Jun 2023 08:32:33 +0000 (09:32 +0100)]
Merge tag 'ieee802154-for-net-2023-06-19' of git://git./linux/kernel/git/wpan/wpan

Stefan Schmidt says:

====================
An update from ieee802154 for your *net* tree:

Two small fixes and MAINTAINERS update this time.

Azeem Shaikh ensured consistent use of strscpy through the tree and fixed
the usage in our trace.h.

Chen Aotian fixed a potential memory leak in the hwsim simulator for
ieee802154.

Miquel Raynal updated the MAINATINERS file with the new team git tree
locations and patchwork URLs.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
13 months agoipvs: align inner_mac_header for encapsulation
Terin Stock [Fri, 9 Jun 2023 20:58:42 +0000 (22:58 +0200)]
ipvs: align inner_mac_header for encapsulation

When using encapsulation the original packet's headers are copied to the
inner headers. This preserves the space for an inner mac header, which
is not used by the inner payloads for the encapsulation types supported
by IPVS. If a packet is using GUE or GRE encapsulation and needs to be
segmented, flow can be passed to __skb_udp_tunnel_segment() which
calculates a negative tunnel header length. A negative tunnel header
length causes pskb_may_pull() to fail, dropping the packet.

This can be observed by attaching probes to ip_vs_in_hook(),
__dev_queue_xmit(), and __skb_udp_tunnel_segment():

    perf probe --add '__dev_queue_xmit skb->inner_mac_header \
    skb->inner_network_header skb->mac_header skb->network_header'
    perf probe --add '__skb_udp_tunnel_segment:7 tnl_hlen'
    perf probe -m ip_vs --add 'ip_vs_in_hook skb->inner_mac_header \
    skb->inner_network_header skb->mac_header skb->network_header'

These probes the headers and tunnel header length for packets which
traverse the IPVS encapsulation path. A TCP packet can be forced into
the segmentation path by being smaller than a calculated clamped MSS,
but larger than the advertised MSS.

    probe:ip_vs_in_hook: inner_mac_header=0x0 inner_network_header=0x0 mac_header=0x44 network_header=0x52
    probe:ip_vs_in_hook: inner_mac_header=0x44 inner_network_header=0x52 mac_header=0x44 network_header=0x32
    probe:dev_queue_xmit: inner_mac_header=0x44 inner_network_header=0x52 mac_header=0x44 network_header=0x32
    probe:__skb_udp_tunnel_segment_L7: tnl_hlen=-2

When using veth-based encapsulation, the interfaces are set to be
mac-less, which does not preserve space for an inner mac header. This
prevents this issue from occurring.

In our real-world testing of sending a 32KB file we observed operation
time increasing from ~75ms for veth-based encapsulation to over 1.5s
using IPVS encapsulation due to retries from dropped packets.

This changeset modifies the packet on the encapsulation path in
ip_vs_tunnel_xmit() and ip_vs_tunnel_xmit_v6() to remove the inner mac
header offset. This fixes UDP segmentation for both encapsulation types,
and corrects the inner headers for any IPIP flows that may use it.

Fixes: 84c0d5e96f3a ("ipvs: allow tunneling with gue encapsulation")
Signed-off-by: Terin Stock <terin@cloudflare.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Acked-by: Simon Horman <horms@kernel.org>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
13 months agoMerge tag 'mlx5-fixes-2023-06-16' of git://git.kernel.org/pub/scm/linux/kernel/git...
David S. Miller [Mon, 19 Jun 2023 09:28:56 +0000 (10:28 +0100)]
Merge tag 'mlx5-fixes-2023-06-16' of git://git./linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
mlx5-fixes-2023-06-16

This series provides bug fixes to mlx5 driver.
Please pull and let me know if there is any problem.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
13 months agonet: qca_spi: Avoid high load if QCA7000 is not available
Stefan Wahren [Wed, 14 Jun 2023 21:06:56 +0000 (23:06 +0200)]
net: qca_spi: Avoid high load if QCA7000 is not available

In case the QCA7000 is not available via SPI (e.g. in reset),
the driver will cause a high load. The reason for this is
that the synchronization is never finished and schedule()
is never called. Since the synchronization is not timing
critical, it's safe to drop this from the scheduling condition.

Signed-off-by: Stefan Wahren <stefan.wahren@i2se.com>
Fixes: 291ab06ecf67 ("net: qualcomm: new Ethernet over SPI driver for QCA7000")
Signed-off-by: David S. Miller <davem@davemloft.net>
13 months agonet: phy: Manual remove LEDs to ensure correct ordering
Andrew Lunn [Sat, 17 Jun 2023 15:55:00 +0000 (17:55 +0200)]
net: phy: Manual remove LEDs to ensure correct ordering

If the core is left to remove the LEDs via devm_, it is performed too
late, after the PHY driver is removed from the PHY. This results in
dereferencing a NULL pointer when the LED core tries to turn the LED
off before destroying the LED.

Manually unregister the LEDs at a safe point in phy_remove.

Cc: stable@vger.kernel.org
Reported-by: Florian Fainelli <f.fainelli@gmail.com>
Suggested-by: Florian Fainelli <f.fainelli@gmail.com>
Fixes: 01e5b728e9e4 ("net: phy: Add a binding for PHY LEDs")
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 months agonfc: fdp: Add MODULE_FIRMWARE macros
Juerg Haefliger [Fri, 16 Jun 2023 12:22:18 +0000 (14:22 +0200)]
nfc: fdp: Add MODULE_FIRMWARE macros

The module loads firmware so add MODULE_FIRMWARE macros to provide that
information via modinfo.

Signed-off-by: Juerg Haefliger <juerg.haefliger@canonical.com>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 months agoieee802154/adf7242: Add MODULE_FIRMWARE macro
Juerg Haefliger [Fri, 16 Jun 2023 12:18:07 +0000 (14:18 +0200)]
ieee802154/adf7242: Add MODULE_FIRMWARE macro

The module loads firmware so add a MODULE_FIRMWARE macro to provide that
information via modinfo.

Signed-off-by: Juerg Haefliger <juerg.haefliger@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 months agosfc: use budget for TX completions
Íñigo Huguet [Thu, 15 Jun 2023 08:49:29 +0000 (10:49 +0200)]
sfc: use budget for TX completions

When running workloads heavy unbalanced towards TX (high TX, low RX
traffic), sfc driver can retain the CPU during too long times. Although
in many cases this is not enough to be visible, it can affect
performance and system responsiveness.

A way to reproduce it is to use a debug kernel and run some parallel
netperf TX tests. In some systems, this will lead to this message being
logged:
  kernel:watchdog: BUG: soft lockup - CPU#12 stuck for 22s!

The reason is that sfc driver doesn't account any NAPI budget for the TX
completion events work. With high-TX/low-RX traffic, this makes that the
CPU is held for long time for NAPI poll.

Documentations says "drivers can process completions for any number of Tx
packets but should only process up to budget number of Rx packets".
However, many drivers do limit the amount of TX completions that they
process in a single NAPI poll.

In the same way, this patch adds a limit for the TX work in sfc. With
the patch applied, the watchdog warning never appears.

Tested with netperf in different combinations: single process / parallel
processes, TCP / UDP and different sizes of UDP messages. Repeated the
tests before and after the patch, without any noticeable difference in
network or CPU performance.

Test hardware:
Intel(R) Xeon(R) CPU E5-1620 v4 @ 3.50GHz (4 cores, 2 threads/core)
Solarflare Communications XtremeScale X2522-25G Network Adapter

Fixes: 5227ecccea2d ("sfc: remove tx and MCDI handling from NAPI budget consideration")
Fixes: d19a53721863 ("sfc_ef100: TX path for EF100 NICs")
Reported-by: Fei Liu <feliu@redhat.com>
Signed-off-by: Íñigo Huguet <ihuguet@redhat.com>
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Link: https://lore.kernel.org/r/20230615084929.10506-1-ihuguet@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agoieee802154: Replace strlcpy with strscpy
Azeem Shaikh [Tue, 13 Jun 2023 00:33:25 +0000 (00:33 +0000)]
ieee802154: Replace strlcpy with strscpy

strlcpy() reads the entire source buffer first.
This read may exceed the destination size limit.
This is both inefficient and can lead to linear read
overflows if a source string is not NUL-terminated [1].
In an effort to remove strlcpy() completely [2], replace
strlcpy() here with strscpy().

Direct replacement is safe here since the return values
from the helper macros are ignored by the callers.

[1] https://www.kernel.org/doc/html/latest/process/deprecated.html#strlcpy
[2] https://github.com/KSPP/linux/issues/89

Signed-off-by: Azeem Shaikh <azeemshaikh38@gmail.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Link: https://lore.kernel.org/r/20230613003326.3538391-1-azeemshaikh38@gmail.com
Signed-off-by: Stefan Schmidt <stefan@datenfreihafen.org>
13 months agonet/mlx5e: Fix scheduling of IPsec ASO query while in atomic
Leon Romanovsky [Mon, 5 Jun 2023 08:09:52 +0000 (11:09 +0300)]
net/mlx5e: Fix scheduling of IPsec ASO query while in atomic

ASO query can be scheduled in atomic context as such it can't use usleep.
Use udelay as recommended in Documentation/timers/timers-howto.rst.

Fixes: 76e463f6508b ("net/mlx5e: Overcome slow response for first IPsec ASO WQE")
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
13 months agonet/mlx5e: Drop XFRM state lock when modifying flow steering
Leon Romanovsky [Mon, 5 Jun 2023 08:09:51 +0000 (11:09 +0300)]
net/mlx5e: Drop XFRM state lock when modifying flow steering

XFRM state which is changed to be XFRM_STATE_EXPIRED doesn't really
need to hold lock while modifying flow steering rules to drop traffic.

That state can be deleted only and as such mlx5e_ipsec_handle_tx_limit()
work will be canceled anyway and won't run in parallel.

Fixes: b2f7b01d36a9 ("net/mlx5e: Simulate missing IPsec TX limits hardware functionality")
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
13 months agonet/mlx5e: Fix ESN update kernel panic
Patrisious Haddad [Mon, 5 Jun 2023 08:09:50 +0000 (11:09 +0300)]
net/mlx5e: Fix ESN update kernel panic

Previously during mlx5e_ipsec_handle_event the driver tried to execute
an operation that could sleep, while holding a spinlock, which caused
the kernel panic mentioned below.

Move the function call that can sleep outside of the spinlock context.

 Call Trace:
 <TASK>
 dump_stack_lvl+0x49/0x6c
 __schedule_bug.cold+0x42/0x4e
 schedule_debug.constprop.0+0xe0/0x118
 __schedule+0x59/0x58a
 ? __mod_timer+0x2a1/0x3ef
 schedule+0x5e/0xd4
 schedule_timeout+0x99/0x164
 ? __pfx_process_timeout+0x10/0x10
 __wait_for_common+0x90/0x1da
 ? __pfx_schedule_timeout+0x10/0x10
 wait_func+0x34/0x142 [mlx5_core]
 mlx5_cmd_invoke+0x1f3/0x313 [mlx5_core]
 cmd_exec+0x1fe/0x325 [mlx5_core]
 mlx5_cmd_do+0x22/0x50 [mlx5_core]
 mlx5_cmd_exec+0x1c/0x40 [mlx5_core]
 mlx5_modify_ipsec_obj+0xb2/0x17f [mlx5_core]
 mlx5e_ipsec_update_esn_state+0x69/0xf0 [mlx5_core]
 ? wake_affine+0x62/0x1f8
 mlx5e_ipsec_handle_event+0xb1/0xc0 [mlx5_core]
 process_one_work+0x1e2/0x3e6
 ? __pfx_worker_thread+0x10/0x10
 worker_thread+0x54/0x3ad
 ? __pfx_worker_thread+0x10/0x10
 kthread+0xda/0x101
 ? __pfx_kthread+0x10/0x10
 ret_from_fork+0x29/0x37
 </TASK>
 BUG: workqueue leaked lock or atomic: kworker/u256:4/0x7fffffff/189754#012     last function: mlx5e_ipsec_handle_event [mlx5_core]
 CPU: 66 PID: 189754 Comm: kworker/u256:4 Kdump: loaded Tainted: G        W          6.2.0-2596.20230309201517_5.el8uek.rc1.x86_64 #2
 Hardware name: Oracle Corporation ORACLE SERVER X9-2/ASMMBX9-2, BIOS 61070300 08/17/2022
 Workqueue: mlx5e_ipsec: eth%d mlx5e_ipsec_handle_event [mlx5_core]
 Call Trace:
 <TASK>
 dump_stack_lvl+0x49/0x6c
 process_one_work.cold+0x2b/0x3c
 ? __pfx_worker_thread+0x10/0x10
 worker_thread+0x54/0x3ad
 ? __pfx_worker_thread+0x10/0x10
 kthread+0xda/0x101
 ? __pfx_kthread+0x10/0x10
 ret_from_fork+0x29/0x37
 </TASK>
 BUG: scheduling while atomic: kworker/u256:4/189754/0x00000000

Fixes: cee137a63431 ("net/mlx5e: Handle ESN update events")
Signed-off-by: Patrisious Haddad <phaddad@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
13 months agonet/mlx5e: Don't delay release of hardware objects
Leon Romanovsky [Mon, 5 Jun 2023 08:09:49 +0000 (11:09 +0300)]
net/mlx5e: Don't delay release of hardware objects

XFRM core provides two callbacks to release resources, one is .xdo_dev_policy_delete()
and another is .xdo_dev_policy_free(). This separation allows delayed release so
"ip xfrm policy free" commands won't starve. Unfortunately, mlx5 command interface
can't run in .xdo_dev_policy_free() callbacks as the latter runs in ATOMIC context.

 BUG: scheduling while atomic: swapper/7/0/0x00000100
 Modules linked in: act_mirred act_tunnel_key cls_flower sch_ingress vxlan mlx5_vdpa vringh vhost_iotlb vdpa rpcrdma rdma_ucm ib_iser libiscsi ib_umad scsi_transport_iscsi rdma_cm ib_ipoib iw_cm ib_cm mlx5_ib ib_uverbs ib_core xt_conntrack xt_MASQUERADE nf_conntrack_netlink nfnetlink xt_addrtype iptable_nat nf_nat br_netfilter rpcsec_gss_krb5 auth_rpcgss oid_registry overlay mlx5_core zram zsmalloc fuse
 CPU: 7 PID: 0 Comm: swapper/7 Not tainted 6.3.0+ #1
 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014
 Call Trace:
  <IRQ>
  dump_stack_lvl+0x33/0x50
  __schedule_bug+0x4e/0x60
  __schedule+0x5d5/0x780
  ? __mod_timer+0x286/0x3d0
  schedule+0x50/0x90
  schedule_timeout+0x7c/0xf0
  ? __bpf_trace_tick_stop+0x10/0x10
  __wait_for_common+0x88/0x190
  ? usleep_range_state+0x90/0x90
  cmd_exec+0x42e/0xb40 [mlx5_core]
  mlx5_cmd_do+0x1e/0x40 [mlx5_core]
  mlx5_cmd_exec+0x18/0x30 [mlx5_core]
  mlx5_cmd_delete_fte+0xa8/0xd0 [mlx5_core]
  del_hw_fte+0x60/0x120 [mlx5_core]
  mlx5_del_flow_rules+0xec/0x270 [mlx5_core]
  ? default_send_IPI_single_phys+0x26/0x30
  mlx5e_accel_ipsec_fs_del_pol+0x1a/0x60 [mlx5_core]
  mlx5e_xfrm_free_policy+0x15/0x20 [mlx5_core]
  xfrm_policy_destroy+0x5a/0xb0
  xfrm4_dst_destroy+0x7b/0x100
  dst_destroy+0x37/0x120
  rcu_core+0x2d6/0x540
  __do_softirq+0xcd/0x273
  irq_exit_rcu+0x82/0xb0
  sysvec_apic_timer_interrupt+0x72/0x90
  </IRQ>
  <TASK>
  asm_sysvec_apic_timer_interrupt+0x16/0x20
 RIP: 0010:default_idle+0x13/0x20
 Code: c0 08 00 00 00 4d 29 c8 4c 01 c7 4c 29 c2 e9 72 ff ff ff cc cc cc cc 8b 05 7a 4d ee 00 85 c0 7e 07 0f 00 2d 2f 98 2e 00 fb f4 <fa> c3 66 66 2e 0f 1f 84 00 00 00 00 00 65 48 8b 04 25 40 b4 02 00
 RSP: 0018:ffff888100843ee0 EFLAGS: 00000242
 RAX: 0000000000000001 RBX: ffff888100812b00 RCX: 4000000000000000
 RDX: 0000000000000001 RSI: 0000000000000083 RDI: 000000000002d2ec
 RBP: 0000000000000007 R08: 00000021daeded59 R09: 0000000000000001
 R10: 0000000000000000 R11: 000000000000000f R12: 0000000000000000
 R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000
  default_idle_call+0x30/0xb0
  do_idle+0x1c1/0x1d0
  cpu_startup_entry+0x19/0x20
  start_secondary+0xfe/0x120
  secondary_startup_64_no_verify+0xf3/0xfb
  </TASK>
 bad: scheduling from the idle thread!

Fixes: a5b8ca9471d3 ("net/mlx5e: Add XFRM policy offload logic")
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
13 months agonet/mlx5: Free IRQ rmap and notifier on kernel shutdown
Saeed Mahameed [Thu, 8 Jun 2023 19:00:54 +0000 (12:00 -0700)]
net/mlx5: Free IRQ rmap and notifier on kernel shutdown

The kernel IRQ system needs the irq affinity notifier to be clear
before attempting to free the irq, see WARN_ON log below.

On a normal driver unload we don't have this issue since we do the
complete cleanup of the irq resources.

To fix this, put the important resources cleanup in a helper function
and use it in both normal driver unload and shutdown flows.

[ 4497.498434] ------------[ cut here ]------------
[ 4497.498726] WARNING: CPU: 0 PID: 9 at kernel/irq/manage.c:2034 free_irq+0x295/0x340
[ 4497.499193] Modules linked in:
[ 4497.499386] CPU: 0 PID: 9 Comm: kworker/0:1 Tainted: G        W          6.4.0-rc4+ #10
[ 4497.499876] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.2-1.fc38 04/01/2014
[ 4497.500518] Workqueue: events do_poweroff
[ 4497.500849] RIP: 0010:free_irq+0x295/0x340
[ 4497.501132] Code: 85 c0 0f 84 1d ff ff ff 48 89 ef ff d0 0f 1f 00 e9 10 ff ff ff 0f 0b e9 72 ff ff ff 49 8d 7f 28 ff d0 0f 1f 00 e9 df fd ff ff <0f> 0b 48 c7 80 c0 008
[ 4497.502269] RSP: 0018:ffffc90000053da0 EFLAGS: 00010282
[ 4497.502589] RAX: ffff888100949600 RBX: ffff88810330b948 RCX: 0000000000000000
[ 4497.503035] RDX: ffff888100949600 RSI: ffff888100400490 RDI: 0000000000000023
[ 4497.503472] RBP: ffff88810330c7e0 R08: ffff8881004005d0 R09: ffffffff8273a260
[ 4497.503923] R10: 0000000000000000 R11: 0000000000000000 R12: ffff8881009ae000
[ 4497.504359] R13: ffff8881009ae148 R14: 0000000000000000 R15: ffff888100949600
[ 4497.504804] FS:  0000000000000000(0000) GS:ffff88813bc00000(0000) knlGS:0000000000000000
[ 4497.505302] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 4497.505671] CR2: 00007fce98806298 CR3: 000000000262e005 CR4: 0000000000370ef0
[ 4497.506104] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 4497.506540] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[ 4497.507002] Call Trace:
[ 4497.507158]  <TASK>
[ 4497.507299]  ? free_irq+0x295/0x340
[ 4497.507522]  ? __warn+0x7c/0x130
[ 4497.507740]  ? free_irq+0x295/0x340
[ 4497.507963]  ? report_bug+0x171/0x1a0
[ 4497.508197]  ? handle_bug+0x3c/0x70
[ 4497.508417]  ? exc_invalid_op+0x17/0x70
[ 4497.508662]  ? asm_exc_invalid_op+0x1a/0x20
[ 4497.508926]  ? free_irq+0x295/0x340
[ 4497.509146]  mlx5_irq_pool_free_irqs+0x48/0x90
[ 4497.509421]  mlx5_irq_table_free_irqs+0x38/0x50
[ 4497.509714]  mlx5_core_eq_free_irqs+0x27/0x40
[ 4497.509984]  shutdown+0x7b/0x100
[ 4497.510184]  pci_device_shutdown+0x30/0x60
[ 4497.510440]  device_shutdown+0x14d/0x240
[ 4497.510698]  kernel_power_off+0x30/0x70
[ 4497.510938]  process_one_work+0x1e6/0x3e0
[ 4497.511183]  worker_thread+0x49/0x3b0
[ 4497.511407]  ? __pfx_worker_thread+0x10/0x10
[ 4497.511679]  kthread+0xe0/0x110
[ 4497.511879]  ? __pfx_kthread+0x10/0x10
[ 4497.512114]  ret_from_fork+0x29/0x50
[ 4497.512342]  </TASK>

Fixes: 9c2d08010963 ("net/mlx5: Free irqs only on shutdown callback")
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
13 months agonet/mlx5: DR, Fix wrong action data allocation in decap action
Yevgeny Kliteynik [Sun, 4 Jun 2023 18:07:04 +0000 (21:07 +0300)]
net/mlx5: DR, Fix wrong action data allocation in decap action

When TUNNEL_L3_TO_L2 decap action was created, a pointer to a local
variable was passed as its HW action data, resulting in attempt to
free invalid address:

  BUG: KASAN: invalid-free in mlx5dr_action_destroy+0x318/0x410 [mlx5_core]

Fixes: 4781df92f4da ("net/mlx5: DR, Move STEv0 modify header logic")
Signed-off-by: Yevgeny Kliteynik <kliteyn@nvidia.com>
Reviewed-by: Alex Vesker <valex@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
13 months agonet/mlx5: DR, Support SW created encap actions for FW table
Yevgeny Kliteynik [Sun, 11 Jun 2023 22:05:48 +0000 (01:05 +0300)]
net/mlx5: DR, Support SW created encap actions for FW table

In some cases, steering might need to use SW-created action in
FW table, which results in wrong packet reformat being used:

  mlx5_core 0000:81:00.1: mlx5_cmd_check:756:(pid 1154):
      SET_FLOW_TABLE_ENTRY(0×936) op_mod(0×0) failed,
      status bad resource(0×5), syndrome (0xf2ff71)

This patch adds support for usage of SW-created packet reformat (encap)
actions in FW tables, and adds clear error flow for attempt to use
SW-created modify header on FW tables.

Fixes: 6a48faeeca10 ("net/mlx5: Add direct rule fs_cmd implementation")
Signed-off-by: Yevgeny Kliteynik <kliteyn@nvidia.com>
Reviewed-by: Erez Shitrit <erezsh@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
13 months agonet/mlx5e: TC, Cleanup ct resources for nic flow
Chris Mi [Thu, 6 Apr 2023 06:38:09 +0000 (09:38 +0300)]
net/mlx5e: TC, Cleanup ct resources for nic flow

The cited commit removes special handling of CT action. But it
removes too much. Pre ct/ct_nat tables and some other resources
are not destroyed due to the cited commit.

Fix it by adding it back.

Fixes: 08fe94ec5f77 ("net/mlx5e: TC, Remove special handling of CT action")
Signed-off-by: Chris Mi <cmi@nvidia.com>
Reviewed-by: Paul Blakey <paulb@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
13 months agonet/mlx5e: TC, Add null pointer check for hardware miss support
Chris Mi [Tue, 11 Apr 2023 08:17:35 +0000 (11:17 +0300)]
net/mlx5e: TC, Add null pointer check for hardware miss support

The cited commits add hardware miss support to tc action. But if
the rules can't be offloaded, the pointers are null and system
will panic when accessing them.

Fix it by checking null pointer.

Fixes: 08fe94ec5f77 ("net/mlx5e: TC, Remove special handling of CT action")
Fixes: 6702782845a5 ("net/mlx5e: TC, Set CT miss to the specific ct action instance")
Signed-off-by: Chris Mi <cmi@nvidia.com>
Reviewed-by: Paul Blakey <paulb@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
13 months agonet/mlx5: Fix driver load with single msix vector
Eli Cohen [Wed, 3 May 2023 12:10:05 +0000 (15:10 +0300)]
net/mlx5: Fix driver load with single msix vector

When a PCI device has just one msix vector available, we want to share
this vector between async and completion events. Current code fails to
do that assuming it will always have at least one dedicated vector for
completion events. Fix this by detecting when the pool contains just a
single vector.

Fixes: 3354822cde5a ("net/mlx5: Use dynamic msix vectors allocation")
Signed-off-by: Eli Cohen <elic@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
13 months agonet/mlx5e: xsk: Set napi_id to support busy polling on XSK RQ
Maxim Mikityanskiy [Wed, 14 Jun 2023 09:00:06 +0000 (12:00 +0300)]
net/mlx5e: xsk: Set napi_id to support busy polling on XSK RQ

The cited commit missed setting napi_id on XSK RQs, it only affected
regular RQs. Add the missing part to support socket busy polling on XSK
RQs.

Fixes: a2740f529da2 ("net/mlx5e: xsk: Set napi_id to support busy polling")
Signed-off-by: Maxim Mikityanskiy <maxtram95@gmail.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
13 months agonet/mlx5e: XDP, Allow growing tail for XDP multi buffer
Maxim Mikityanskiy [Wed, 14 Jun 2023 09:00:05 +0000 (12:00 +0300)]
net/mlx5e: XDP, Allow growing tail for XDP multi buffer

The cited commits missed passing frag_size to __xdp_rxq_info_reg, which
is required by bpf_xdp_adjust_tail to support growing the tail pointer
in fragmented packets. Pass the missing parameter when the current RQ
mode allows XDP multi buffer.

Fixes: ea5d49bdae8b ("net/mlx5e: Add XDP multi buffer support to the non-linear legacy RQ")
Fixes: 9cb9482ef10e ("net/mlx5e: Use fragments of the same size in non-linear legacy RQ with XDP")
Signed-off-by: Maxim Mikityanskiy <maxtram95@gmail.com>
Cc: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
13 months agoxfrm: Linearize the skb after offloading if needed.
Sebastian Andrzej Siewior [Wed, 14 Jun 2023 10:02:02 +0000 (12:02 +0200)]
xfrm: Linearize the skb after offloading if needed.

With offloading enabled, esp_xmit() gets invoked very late, from within
validate_xmit_xfrm() which is after validate_xmit_skb() validates and
linearizes the skb if the underlying device does not support fragments.

esp_output_tail() may add a fragment to the skb while adding the auth
tag/ IV. Devices without the proper support will then send skb->data
points to with the correct length so the packet will have garbage at the
end. A pcap sniffer will claim that the proper data has been sent since
it parses the skb properly.

It is not affected with INET_ESP_OFFLOAD disabled.

Linearize the skb after offloading if the sending hardware requires it.
It was tested on v4, v6 has been adopted.

Fixes: 7785bba299a8d ("esp: Add a software GRO codepath")
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
13 months agoMerge branch 'check-if-fips-mode-is-enabled-when-running-selftests'
Jakub Kicinski [Fri, 16 Jun 2023 05:24:03 +0000 (22:24 -0700)]
Merge branch 'check-if-fips-mode-is-enabled-when-running-selftests'

Magali Lemes says:

====================
Check if FIPS mode is enabled when running selftests

Some test cases from net/tls, net/fcnal-test and net/vrf-xfrm-tests
that rely on cryptographic functions to work and use non-compliant FIPS
algorithms fail in FIPS mode.

In order to allow these tests to pass in a wider set of kernels,
 - for net/tls, skip the test variants that use the ChaCha20-Poly1305
and SM4 algorithms, when FIPS mode is enabled;
 - for net/fcnal-test, skip the MD5 tests, when FIPS mode is enabled;
 - for net/vrf-xfrm-tests, replace the algorithms that are not
FIPS-compliant with compliant ones.

v1: https://lore.kernel.org/netdev/20230607174302.19542-1-magali.lemes@canonical.com/
v2: https://lore.kernel.org/netdev/20230609164324.497813-1-magali.lemes@canonical.com/
v3: https://lore.kernel.org/netdev/20230612125107.73795-1-magali.lemes@canonical.com/
====================

Link: https://lore.kernel.org/r/20230613123222.631897-1-magali.lemes@canonical.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agoselftests: net: fcnal-test: check if FIPS mode is enabled
Magali Lemes [Tue, 13 Jun 2023 12:32:22 +0000 (09:32 -0300)]
selftests: net: fcnal-test: check if FIPS mode is enabled

There are some MD5 tests which fail when the kernel is in FIPS mode,
since MD5 is not FIPS compliant. Add a check and only run those tests
if FIPS mode is not enabled.

Fixes: f0bee1ebb5594 ("fcnal-test: Add TCP MD5 tests")
Fixes: 5cad8bce26e01 ("fcnal-test: Add TCP MD5 tests for VRF")
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: Magali Lemes <magali.lemes@canonical.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agoselftests: net: vrf-xfrm-tests: change authentication and encryption algos
Magali Lemes [Tue, 13 Jun 2023 12:32:21 +0000 (09:32 -0300)]
selftests: net: vrf-xfrm-tests: change authentication and encryption algos

The vrf-xfrm-tests tests use the hmac(md5) and cbc(des3_ede)
algorithms for performing authentication and encryption, respectively.
This causes the tests to fail when fips=1 is set, since these algorithms
are not allowed in FIPS mode. Therefore, switch from hmac(md5) and
cbc(des3_ede) to hmac(sha1) and cbc(aes), which are FIPS compliant.

Fixes: 3f251d741150 ("selftests: Add tests for vrf and xfrms")
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: Magali Lemes <magali.lemes@canonical.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agoselftests: net: tls: check if FIPS mode is enabled
Magali Lemes [Tue, 13 Jun 2023 12:32:20 +0000 (09:32 -0300)]
selftests: net: tls: check if FIPS mode is enabled

TLS selftests use the ChaCha20-Poly1305 and SM4 algorithms, which are not
FIPS compliant. When fips=1, this set of tests fails. Add a check and only
run these tests if not in FIPS mode.

Fixes: 4f336e88a870 ("selftests/tls: add CHACHA20-POLY1305 to tls selftests")
Fixes: e506342a03c7 ("selftests/tls: add SM4 GCM/CCM to tls selftests")
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Magali Lemes <magali.lemes@canonical.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agoselftests/harness: allow tests to be skipped during setup
Magali Lemes [Tue, 13 Jun 2023 12:32:19 +0000 (09:32 -0300)]
selftests/harness: allow tests to be skipped during setup

Before executing each test from a fixture, FIXTURE_SETUP is run once.
When SKIP is used in FIXTURE_SETUP, the setup function returns early
but the test still proceeds to run, unless another SKIP macro is used
within the test definition, leading to some code repetition. Therefore,
allow tests to be skipped directly from the setup function.

Suggested-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Magali Lemes <magali.lemes@canonical.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agoMerge tag 'net-6.4-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Linus Torvalds [Fri, 16 Jun 2023 04:11:17 +0000 (21:11 -0700)]
Merge tag 'net-6.4-rc7' of git://git./linux/kernel/git/netdev/net

Pull networking fixes from Jakub Kicinski:
 "Including fixes from wireless, and netfilter.

  Selftests excluded - we have 58 patches and diff of +442/-199, which
  isn't really small but perhaps with the exception of the WiFi locking
  change it's old(ish) bugs.

  We have no known problems with v6.4.

  The selftest changes are rather large as MPTCP folks try to apply
  Greg's guidance that selftest from torvalds/linux should be able to
  run against stable kernels.

  Last thing I should call out is the DCCP/UDP-lite deprecation notices.
  We are fairly sure those are dead, but if we're wrong reverting them
  back in won't be fun.

  Current release - regressions:

   - wifi:
      - cfg80211: fix double lock bug in reg_wdev_chan_valid()
      - iwlwifi: mvm: spin_lock_bh() to fix lockdep regression

  Current release - new code bugs:

   - handshake: remove fput() that causes use-after-free

  Previous releases - regressions:

   - sched: cls_u32: fix reference counter leak leading to overflow

   - sched: cls_api: fix lockup on flushing explicitly created chain

  Previous releases - always broken:

   - nf_tables: integrate pipapo into commit protocol

   - nf_tables: incorrect error path handling with NFT_MSG_NEWRULE, fix
     dangling pointer on failure

   - ping6: fix send to link-local addresses with VRF

   - sched: act_pedit: parse L3 header for L4 offset, the skb may not
     have the offset saved

   - sched: act_ct: fix promotion of offloaded unreplied tuple

   - sched: refuse to destroy an ingress and clsact Qdiscs if there are
     lockless change operations in flight

   - wifi: mac80211: fix handful of bugs in multi-link operation

   - ipvlan: fix bound dev checking for IPv6 l3s mode

   - eth: enetc: correct the indexes of highest and 2nd highest TCs

   - eth: ice: fix XDP memory leak when NIC is brought up and down

  Misc:

   - add deprecation notices for UDP-lite and DCCP

   - selftests: mptcp: skip tests not supported by old kernels

   - sctp: handle invalid error codes without calling BUG()"

* tag 'net-6.4-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (91 commits)
  dccp: Print deprecation notice.
  udplite: Print deprecation notice.
  octeon_ep: Add missing check for ioremap
  selftests/ptp: Fix timestamp printf format for PTP_SYS_OFFSET
  net: ethernet: stmicro: stmmac: fix possible memory leak in __stmmac_open
  net: tipc: resize nlattr array to correct size
  sfc: fix XDP queues mode with legacy IRQ
  net: macsec: fix double free of percpu stats
  net: lapbether: only support ethernet devices
  MAINTAINERS: add reviewers for SMC Sockets
  s390/ism: Fix trying to free already-freed IRQ by repeated ism_dev_exit()
  net: dsa: felix: fix taprio guard band overflow at 10Mbps with jumbo frames
  net/sched: cls_api: Fix lockup on flushing explicitly created chain
  ice: Fix ice module unload
  net/handshake: remove fput() that causes use-after-free
  selftests: forwarding: hw_stats_l3: Set addrgenmode in a separate step
  net/sched: qdisc_destroy() old ingress and clsact Qdiscs before grafting
  net/sched: Refactor qdisc_graft() for ingress and clsact Qdiscs
  net/sched: act_ct: Fix promotion of offloaded unreplied tuple
  wifi: iwlwifi: mvm: spin_lock_bh() to fix lockdep regression
  ...

13 months agoMerge tag 'loongarch-fixes-6.4-1' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 16 Jun 2023 03:56:39 +0000 (20:56 -0700)]
Merge tag 'loongarch-fixes-6.4-1' of git://git./linux/kernel/git/chenhuacai/linux-loongson

Pull LoongArch fixes from Huacai Chen:
 "Some trivial bug fixes for v6.4-rc7"

* tag 'loongarch-fixes-6.4-1' of git://git.kernel.org/pub/scm/linux/kernel/git/chenhuacai/linux-loongson:
  LoongArch: Fix debugfs_create_dir() error checking
  LoongArch: Avoid uninitialized alignment_mask
  LoongArch: Fix perf event id calculation
  LoongArch: Fix the write_fcsr() macro
  LoongArch: Let pmd_present() return true when splitting pmd

13 months agoMerge tag 'for-6.4/dm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/device...
Linus Torvalds [Fri, 16 Jun 2023 03:19:21 +0000 (20:19 -0700)]
Merge tag 'for-6.4/dm-fixes' of git://git./linux/kernel/git/device-mapper/linux-dm

Pull device mapper fixes from Mike Snitzer:

 - Fix DM thinp discard performance regression introduced during this
   merge window where DM core was splitting large discards every 128K
   (max_sectors_kb) rather than every 64M (discard_max_bytes).

 - Extend DM core LOCKFS fix, made during 6.4 merge, to also fix race
   between do_mount and dm's do_suspend (in addition to the earlier
   fix's do_mount race with dm's do_resume).

 - Fix DM thin metadata operations to first check if the thin-pool is in
   "fail_io" mode; otherwise UAF can occur.

 - Fix DM thinp's call to __blkdev_issue_discard to use GFP_NOIO rather
   than GFP_NOWAIT (__blkdev_issue_discard cannot handle NULL return
   from bio_alloc).

* tag 'for-6.4/dm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
  dm: use op specific max_sectors when splitting abnormal io
  dm thin: fix issue_discard to pass GFP_NOIO to __blkdev_issue_discard
  dm thin metadata: check fail_io before using data_sm
  dm: don't lock fs when the map is NULL during suspend or resume

13 months agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma
Linus Torvalds [Fri, 16 Jun 2023 03:13:56 +0000 (20:13 -0700)]
Merge tag 'for-linus' of git://git./linux/kernel/git/rdma/rdma

Pull rdma fixes from Jason Gunthorpe:
 "This is an unusually large bunch of bug fixes for the later rc cycle,
  rxe and mlx5 both dumped a lot of things at once. rxe continues to fix
  itself, and mlx5 is fixing a bunch of "queue counters" related bugs.

  There is one highly notable bug fix regarding the qkey. This small
  security check was missed in the original 2005 implementation and it
  allows some significant issues.

  Summary:

   - Two rtrs bug fixes for error unwind bugs

   - Several rxe bug fixes:
      * Incorrect Rx packet validation
      * Using memory without a refcount
      * Syzkaller found use before initialization
      * Regression fix for missing locking with the tasklet conversion
        from this merge window

   - Have bnxt report the correct link properties to userspace, this was
     a regression in v6.3

   - Several mlx5 bug fixes:
      * Kernel crash triggerable by userspace for the RAW ethernet
        profile
      * Defend against steering refcounting issues created by userspace
      * Incorrect change of QP port affinity parameters in some LAG
        configurations

   - Fix mlx5 Q counters:
      * Do not over allocate Q counters to allow userspace to use the
        full port capacity
      * Kernel crash triggered by eswitch due to mis-use of Q counters
      * Incorrect mlx5_device for Q counters in some LAG configurations

   - Properly implement the IBA spec restricting privileged qkeys to
     root

   - Always an error when reading from a disassociated device's event
     queue

   - isert bug fixes:
      * Avoid a deadlock with the CM handler and CM ID destruction
      * Correct list corruption due to incorrect locking
      * Fix a use after free around connection tear down"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma:
  RDMA/rxe: Fix rxe_cq_post
  IB/isert: Fix incorrect release of isert connection
  IB/isert: Fix possible list corruption in CMA handler
  IB/isert: Fix dead lock in ib_isert
  RDMA/mlx5: Fix affinity assignment
  IB/uverbs: Fix to consider event queue closing also upon non-blocking mode
  RDMA/uverbs: Restrict usage of privileged QKEYs
  RDMA/cma: Always set static rate to 0 for RoCE
  RDMA/mlx5: Fix Q-counters query in LAG mode
  RDMA/mlx5: Remove vport Q-counters dependency on normal Q-counters
  RDMA/mlx5: Fix Q-counters per vport allocation
  RDMA/mlx5: Create an indirect flow table for steering anchor
  RDMA/mlx5: Initiate dropless RQ for RAW Ethernet functions
  RDMA/rxe: Fix the use-before-initialization error of resp_pkts
  RDMA/bnxt_re: Fix reporting active_{speed,width} attributes
  RDMA/rxe: Fix ref count error in check_rkey()
  RDMA/rxe: Fix packet length checks
  RDMA/rtrs: Fix rxe_dealloc_pd warning
  RDMA/rtrs: Fix the last iu->buf leak in err path

13 months agoMerge tag 'spi-fix-v6.4-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi
Linus Torvalds [Fri, 16 Jun 2023 03:03:15 +0000 (20:03 -0700)]
Merge tag 'spi-fix-v6.4-rc6' of git://git./linux/kernel/git/broonie/spi

Pull spi fixes from Mark Brown:
 "A few more driver specific fixes.

  The DesignWare fix is for an issue introduced by conversion to the
  chip select accessor functions and is pretty important but the other
  two are less severe"

* tag 'spi-fix-v6.4-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi:
  spi: dw: Replace incorrect spi_get_chipselect with set
  spi: fsl-dspi: avoid SCK glitches with continuous transfers
  spi: cadence-quadspi: Add missing check for dma_set_mask

13 months agoMerge tag 'regulator-fix-v6.4-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 16 Jun 2023 02:54:58 +0000 (19:54 -0700)]
Merge tag 'regulator-fix-v6.4-rc6' of git://git./linux/kernel/git/broonie/regulator

Pull regulator fix from Mark Brown:
 "The set of regulators described for the Qualcomm PM8550 just seems to
  have been completely wrong and would likely not have worked at all if
  anything tried to actually configure anything except for enabling and
  disabling at runtime"

* tag 'regulator-fix-v6.4-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator:
  regulator: qcom-rpmh: Fix regulators for PM8550

13 months agoMerge tag 'regmap-fix-v6.4-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 16 Jun 2023 02:50:57 +0000 (19:50 -0700)]
Merge tag 'regmap-fix-v6.4-rc6' of git://git./linux/kernel/git/broonie/regmap

Pull regmap fix from Mark Brown:
 "Another fix for the maple tree cache, Takashi noticed that unlike
  other caches the maple tree cache didn't check for read only registers
  before trying to sync which would result in spurious syncs for read
  only registers where we don't have a default.

  This was due to the check being open coded in the caches, we now check
  in the shared 'does this register need sync' function so that is fixed
  for this and future caches"

* tag 'regmap-fix-v6.4-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regmap:
  regmap: regcache: Don't sync read-only registers

13 months agoMerge tag 'media/v6.4-6' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab...
Linus Torvalds [Fri, 16 Jun 2023 02:13:45 +0000 (19:13 -0700)]
Merge tag 'media/v6.4-6' of git://git./linux/kernel/git/mchehab/linux-media

Pull media fixes from Mauro Carvalho Chehab:
 "A fix for dvb-core to avoid a race condition during DVB board
  registration"

* tag 'media/v6.4-6' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media:
  Revert "media: dvb-core: Fix use-after-free on race condition at dvb_frontend"

13 months agoMerge tag 'ext4_for_linus_stable' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 15 Jun 2023 22:40:58 +0000 (15:40 -0700)]
Merge tag 'ext4_for_linus_stable' of git://git./linux/kernel/git/tytso/ext4

Pull ext4 fixes from Ted Ts'o:
 "Fix two regressions in ext4, one report by syzkaller[1], and reported
  by multiple users (and tracked by regzbot[2])"

[1] https://syzkaller.appspot.com/bug?extid=4acc7d910e617b360859
[2] https://linux-regtracking.leemhuis.info/regzbot/regression/ZIauBR7YiV3rVAHL@glitch/

* tag 'ext4_for_linus_stable' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4:
  ext4: drop the call to ext4_error() from ext4_get_group_info()
  Revert "ext4: remove unnecessary check in ext4_bg_num_gdb_nometa"

13 months agoMerge tag '6.4-rc6-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Thu, 15 Jun 2023 22:24:33 +0000 (15:24 -0700)]
Merge tag '6.4-rc6-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6

Pull smb client fixes from Steve French:
 "Eight, mostly small, smb3 client fixes:

   - important fix for deferred close oops (race with unmount) found
     with xfstest generic/098 to some servers

   - important reconnect fix

   - fix problem with max_credits mount option

   - two multichannel (interface related) fixes

   - one trivial removal of confusing comment

   - two small debugging improvements (to better spot crediting
     problems)"

* tag '6.4-rc6-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6:
  cifs: add a warning when the in-flight count goes negative
  cifs: fix lease break oops in xfstest generic/098
  cifs: fix max_credits implementation
  cifs: fix sockaddr comparison in iface_cmp
  smb/client: print "Unknown" instead of bogus link speed value
  cifs: print all credit counters in DebugData
  cifs: fix status checks in cifs_tree_connect
  smb: remove obsolete comment

13 months agoMerge branch 'udplite-dccp-print-deprecation-notice'
Jakub Kicinski [Thu, 15 Jun 2023 22:09:00 +0000 (15:09 -0700)]
Merge branch 'udplite-dccp-print-deprecation-notice'

Kuniyuki Iwashima says:

====================
udplite/dccp: Print deprecation notice.

UDP-Lite is assumed to have no users for 7 years, and DCCP is
orphaned for 7 years too.

Let's add deprecation notice and see if anyone responds to it.
====================

Link: https://lore.kernel.org/r/20230614194705.90673-1-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agodccp: Print deprecation notice.
Kuniyuki Iwashima [Wed, 14 Jun 2023 19:47:05 +0000 (12:47 -0700)]
dccp: Print deprecation notice.

DCCP was marked as Orphan in the MAINTAINERS entry 2 years ago in commit
054c4610bd05 ("MAINTAINERS: dccp: move Gerrit Renker to CREDITS").  It says
we haven't heard from the maintainer for five years, so DCCP is not well
maintained for 7 years now.

Recently DCCP only receives updates for bugs, and major distros disable it
by default.

Removing DCCP would allow for better organisation of TCP fields to reduce
the number of cache lines hit in the fast path.

Let's add a deprecation notice when DCCP socket is created and schedule its
removal to 2025.

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agoudplite: Print deprecation notice.
Kuniyuki Iwashima [Wed, 14 Jun 2023 19:47:04 +0000 (12:47 -0700)]
udplite: Print deprecation notice.

Recently syzkaller reported a 7-year-old null-ptr-deref [0] that occurs
when a UDP-Lite socket tries to allocate a buffer under memory pressure.

Someone should have stumbled on the bug much earlier if UDP-Lite had been
used in a real app.  Also, we do not always need a large UDP-Lite workload
to hit the bug since UDP and UDP-Lite share the same memory accounting
limit.

Removing UDP-Lite would simplify UDP code removing a bunch of conditionals
in fast path.

Let's add a deprecation notice when UDP-Lite socket is created and schedule
its removal to 2025.

Link: https://lore.kernel.org/netdev/20230523163305.66466-1-kuniyu@amazon.com/
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agoocteon_ep: Add missing check for ioremap
Jiasheng Jiang [Thu, 15 Jun 2023 03:34:00 +0000 (11:34 +0800)]
octeon_ep: Add missing check for ioremap

Add check for ioremap() and return the error if it fails in order to
guarantee the success of ioremap().

Fixes: 862cd659a6fb ("octeon_ep: Add driver framework and device initialization")
Signed-off-by: Jiasheng Jiang <jiasheng@iscas.ac.cn>
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Link: https://lore.kernel.org/r/20230615033400.2971-1-jiasheng@iscas.ac.cn
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agoselftests/ptp: Fix timestamp printf format for PTP_SYS_OFFSET
Alex Maftei [Thu, 15 Jun 2023 08:34:04 +0000 (09:34 +0100)]
selftests/ptp: Fix timestamp printf format for PTP_SYS_OFFSET

Previously, timestamps were printed using "%lld.%u" which is incorrect
for nanosecond values lower than 100,000,000 as they're fractional
digits, therefore leading zeros are meaningful.

This patch changes the format strings to "%lld.%09u" in order to add
leading zeros to the nanosecond value.

Fixes: 568ebc5985f5 ("ptp: add the PTP_SYS_OFFSET ioctl to the testptp program")
Fixes: 4ec54f95736f ("ptp: Fix compiler warnings in the testptp utility")
Fixes: 6ab0e475f1f3 ("Documentation: fix misc. warnings")
Signed-off-by: Alex Maftei <alex.maftei@amd.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Link: https://lore.kernel.org/r/20230615083404.57112-1-alex.maftei@amd.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agonet: ethernet: stmicro: stmmac: fix possible memory leak in __stmmac_open
Christian Marangi [Wed, 14 Jun 2023 09:17:14 +0000 (11:17 +0200)]
net: ethernet: stmicro: stmmac: fix possible memory leak in __stmmac_open

Fix a possible memory leak in __stmmac_open when stmmac_init_phy fails.
It's also needed to free everything allocated by stmmac_setup_dma_desc
and not just the dma_conf struct.

Drop free_dma_desc_resources from __stmmac_open and correctly call
free_dma_desc_resources on each user of __stmmac_open on error.

Reported-by: Jose Abreu <Jose.Abreu@synopsys.com>
Fixes: ba39b344e924 ("net: ethernet: stmicro: stmmac: generate stmmac dma conf before open")
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Cc: stable@vger.kernel.org
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Reviewed-by: Jose Abreu <Jose.Abreu@synopsys.com>
Link: https://lore.kernel.org/r/20230614091714.15912-1-ansuelsmth@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agonet: tipc: resize nlattr array to correct size
Lin Ma [Wed, 14 Jun 2023 12:06:04 +0000 (20:06 +0800)]
net: tipc: resize nlattr array to correct size

According to nla_parse_nested_deprecated(), the tb[] is supposed to the
destination array with maxtype+1 elements. In current
tipc_nl_media_get() and __tipc_nl_media_set(), a larger array is used
which is unnecessary. This patch resize them to a proper size.

Fixes: 1e55417d8fc6 ("tipc: add media set to new netlink api")
Fixes: 46f15c6794fb ("tipc: add media get/dump to new netlink api")
Signed-off-by: Lin Ma <linma@zju.edu.cn>
Reviewed-by: Florian Westphal <fw@strlen.de>
Reviewed-by: Tung Nguyen <tung.q.nguyen@dektech.com.au>
Link: https://lore.kernel.org/r/20230614120604.1196377-1-linma@zju.edu.cn
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
13 months agodm: use op specific max_sectors when splitting abnormal io
Mike Snitzer [Thu, 15 Jun 2023 01:47:46 +0000 (21:47 -0400)]
dm: use op specific max_sectors when splitting abnormal io

Split abnormal IO in terms of the corresponding operation specific
max_sectors (max_discard_sectors, max_secure_erase_sectors or
max_write_zeroes_sectors).

This fixes a significant dm-thinp discard performance regression that
was introduced with commit e2dd8aca2d76 ("dm bio prison v1: improve
concurrent IO performance"). Relative to discard: max_discard_sectors
is used instead of max_sectors; which fixes excessive discard splitting
(e.g. max_sectors=128K vs max_discard_sectors=64M).

Tested by discarding an 1 Petabyte dm-thin device:
lvcreate -V 1125899906842624B -T test/pool -n thin
time blkdiscard /dev/test/thin

Before this fix (splitting discards every 128K): ~116m
 After this fix (splitting discards every 64M) : 0m33.460s

Reported-by: Zorro Lang <zlang@redhat.com>
Fixes: 06961c487a33 ("dm: split discards further if target sets max_discard_granularity")
Requires: 13f6facf3fae ("dm: allow targets to require splitting WRITE_ZEROES and SECURE_ERASE")
Fixes: e2dd8aca2d76 ("dm bio prison v1: improve concurrent IO performance")
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
13 months agodm thin: fix issue_discard to pass GFP_NOIO to __blkdev_issue_discard
Mike Snitzer [Wed, 14 Jun 2023 00:05:34 +0000 (20:05 -0400)]
dm thin: fix issue_discard to pass GFP_NOIO to __blkdev_issue_discard

issue_discard() passes GFP_NOWAIT to __blkdev_issue_discard() despite
its code assuming bio_alloc() always succeeds.

Commit 3dba53a958a75 ("dm thin: use __blkdev_issue_discard for async
discard support") clearly shows where things went bad:

Before commit 3dba53a958a75, dm-thin.c's open-coded
__blkdev_issue_discard_async() properly handled using GFP_NOWAIT.
Unfortunately __blkdev_issue_discard() doesn't and it was missed
during review.

Cc: stable@vger.kernel.org
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
13 months agodm thin metadata: check fail_io before using data_sm
Li Lingfeng [Tue, 6 Jun 2023 12:20:24 +0000 (20:20 +0800)]
dm thin metadata: check fail_io before using data_sm

Must check pmd->fail_io before using pmd->data_sm since
pmd->data_sm may be destroyed by other processes.

       P1(kworker)                             P2(message)
do_worker
 process_prepared
  process_prepared_discard_passdown_pt2
   dm_pool_dec_data_range
                                    pool_message
                                     commit
                                      dm_pool_commit_metadata
                                        ↓
                                       // commit failed
                                      metadata_operation_failed
                                       abort_transaction
                                        dm_pool_abort_metadata
                                         __open_or_format_metadata
                                           ↓
                                          dm_sm_disk_open
                                            ↓
                                           // open failed
                                           // pmd->data_sm is NULL
    dm_sm_dec_blocks
      ↓
     // try to access pmd->data_sm --> UAF

As shown above, if dm_pool_commit_metadata() and
dm_pool_abort_metadata() fail in pool_message process, kworker may
trigger UAF.

Fixes: be500ed721a6 ("dm space maps: improve performance with inc/dec on ranges of blocks")
Cc: stable@vger.kernel.org
Signed-off-by: Li Lingfeng <lilingfeng3@huawei.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
13 months agodm: don't lock fs when the map is NULL during suspend or resume
Li Lingfeng [Thu, 1 Jun 2023 06:14:23 +0000 (14:14 +0800)]
dm: don't lock fs when the map is NULL during suspend or resume

As described in commit 38d11da522aa ("dm: don't lock fs when the map is
NULL in process of resume"), a deadlock may be triggered between
do_resume() and do_mount().

This commit preserves the fix from commit 38d11da522aa but moves it to
where it also serves to fix a similar deadlock between do_suspend()
and do_mount().  It does so, if the active map is NULL, by clearing
DM_SUSPEND_LOCKFS_FLAG in dm_suspend() which is called by both
do_suspend() and do_resume().

Fixes: 38d11da522aa ("dm: don't lock fs when the map is NULL in process of resume")
Signed-off-by: Li Lingfeng <lilingfeng3@huawei.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
13 months agosfc: fix XDP queues mode with legacy IRQ
Íñigo Huguet [Tue, 13 Jun 2023 13:38:54 +0000 (15:38 +0200)]
sfc: fix XDP queues mode with legacy IRQ

In systems without MSI-X capabilities, xdp_txq_queues_mode is calculated
in efx_allocate_msix_channels, but when enabling MSI-X fails, it was not
changed to a proper default value. This was leading to the driver
thinking that it has dedicated XDP queues, when it didn't.

Fix it by setting xdp_txq_queues_mode to the correct value if the driver
fallbacks to MSI or legacy IRQ mode. The correct value is
EFX_XDP_TX_QUEUES_BORROWED because there are no XDP dedicated queues.

The issue can be easily visible if the kernel is started with pci=nomsi,
then a call trace is shown. It is not shown only with sfc's modparam
interrupt_mode=2. Call trace example:
 WARNING: CPU: 2 PID: 663 at drivers/net/ethernet/sfc/efx_channels.c:828 efx_set_xdp_channels+0x124/0x260 [sfc]
 [...skip...]
 Call Trace:
  <TASK>
  efx_set_channels+0x5c/0xc0 [sfc]
  efx_probe_nic+0x9b/0x15a [sfc]
  efx_probe_all+0x10/0x1a2 [sfc]
  efx_pci_probe_main+0x12/0x156 [sfc]
  efx_pci_probe_post_io+0x18/0x103 [sfc]
  efx_pci_probe.cold+0x154/0x257 [sfc]
  local_pci_probe+0x42/0x80

Fixes: 6215b608a8c4 ("sfc: last resort fallback for lack of xdp tx queues")
Reported-by: Yanghang Liu <yanghliu@redhat.com>
Signed-off-by: Íñigo Huguet <ihuguet@redhat.com>
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 months agonet: macsec: fix double free of percpu stats
Fedor Pchelkin [Tue, 13 Jun 2023 19:22:20 +0000 (22:22 +0300)]
net: macsec: fix double free of percpu stats

Inside macsec_add_dev() we free percpu macsec->secy.tx_sc.stats and
macsec->stats on some of the memory allocation failure paths. However, the
net_device is already registered to that moment: in macsec_newlink(), just
before calling macsec_add_dev(). This means that during unregister process
its priv_destructor - macsec_free_netdev() - will be called and will free
the stats again.

Remove freeing percpu stats inside macsec_add_dev() because
macsec_free_netdev() will correctly free the already allocated ones. The
pointers to unallocated stats stay NULL, and free_percpu() treats that
correctly.

Found by Linux Verification Center (linuxtesting.org) with Syzkaller.

Fixes: 0a28bfd4971f ("net/macsec: Add MACsec skb_metadata_dst Tx Data path support")
Fixes: c09440f7dcb3 ("macsec: introduce IEEE 802.1AE driver")
Signed-off-by: Fedor Pchelkin <pchelkin@ispras.ru>
Reviewed-by: Sabrina Dubroca <sd@queasysnail.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 months agonet: lapbether: only support ethernet devices
Eric Dumazet [Wed, 14 Jun 2023 16:18:02 +0000 (16:18 +0000)]
net: lapbether: only support ethernet devices

It probbaly makes no sense to support arbitrary network devices
for lapbether.

syzbot reported:

skbuff: skb_under_panic: text:ffff80008934c100 len:44 put:40 head:ffff0000d18dd200 data:ffff0000d18dd1ea tail:0x16 end:0x140 dev:bond1
kernel BUG at net/core/skbuff.c:200 !
Internal error: Oops - BUG: 00000000f2000800 [#1] PREEMPT SMP
Modules linked in:
CPU: 0 PID: 5643 Comm: dhcpcd Not tainted 6.4.0-rc5-syzkaller-g4641cff8e810 #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 05/25/2023
pstate: 60400005 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
pc : skb_panic net/core/skbuff.c:196 [inline]
pc : skb_under_panic+0x13c/0x140 net/core/skbuff.c:210
lr : skb_panic net/core/skbuff.c:196 [inline]
lr : skb_under_panic+0x13c/0x140 net/core/skbuff.c:210
sp : ffff8000973b7260
x29: ffff8000973b7270 x28: ffff8000973b7360 x27: dfff800000000000
x26: ffff0000d85d8150 x25: 0000000000000016 x24: ffff0000d18dd1ea
x23: ffff0000d18dd200 x22: 000000000000002c x21: 0000000000000140
x20: 0000000000000028 x19: ffff80008934c100 x18: ffff8000973b68a0
x17: 0000000000000000 x16: ffff80008a43bfbc x15: 0000000000000202
x14: 0000000000000000 x13: 0000000000000001 x12: 0000000000000001
x11: 0000000000000201 x10: 0000000000000000 x9 : f22f7eb937cced00
x8 : f22f7eb937cced00 x7 : 0000000000000001 x6 : 0000000000000001
x5 : ffff8000973b6b78 x4 : ffff80008df9ee80 x3 : ffff8000805974f4
x2 : 0000000000000001 x1 : 0000000100000201 x0 : 0000000000000086
Call trace:
skb_panic net/core/skbuff.c:196 [inline]
skb_under_panic+0x13c/0x140 net/core/skbuff.c:210
skb_push+0xf0/0x108 net/core/skbuff.c:2409
ip6gre_header+0xbc/0x738 net/ipv6/ip6_gre.c:1383
dev_hard_header include/linux/netdevice.h:3137 [inline]
lapbeth_data_transmit+0x1c4/0x298 drivers/net/wan/lapbether.c:257
lapb_data_transmit+0x8c/0xb0 net/lapb/lapb_iface.c:447
lapb_transmit_buffer+0x178/0x204 net/lapb/lapb_out.c:149
lapb_send_control+0x220/0x320 net/lapb/lapb_subr.c:251
lapb_establish_data_link+0x94/0xec
lapb_device_event+0x348/0x4e0
notifier_call_chain+0x1a4/0x510 kernel/notifier.c:93
raw_notifier_call_chain+0x3c/0x50 kernel/notifier.c:461
__dev_notify_flags+0x2bc/0x544
dev_change_flags+0xd0/0x15c net/core/dev.c:8643
devinet_ioctl+0x858/0x17e4 net/ipv4/devinet.c:1150
inet_ioctl+0x2ac/0x4d8 net/ipv4/af_inet.c:979
sock_do_ioctl+0x134/0x2dc net/socket.c:1201
sock_ioctl+0x4ec/0x858 net/socket.c:1318
vfs_ioctl fs/ioctl.c:51 [inline]
__do_sys_ioctl fs/ioctl.c:870 [inline]
__se_sys_ioctl fs/ioctl.c:856 [inline]
__arm64_sys_ioctl+0x14c/0x1c8 fs/ioctl.c:856
__invoke_syscall arch/arm64/kernel/syscall.c:38 [inline]
invoke_syscall+0x98/0x2c0 arch/arm64/kernel/syscall.c:52
el0_svc_common+0x138/0x244 arch/arm64/kernel/syscall.c:142
do_el0_svc+0x64/0x198 arch/arm64/kernel/syscall.c:191
el0_svc+0x4c/0x160 arch/arm64/kernel/entry-common.c:647
el0t_64_sync_handler+0x84/0xfc arch/arm64/kernel/entry-common.c:665
el0t_64_sync+0x190/0x194 arch/arm64/kernel/entry.S:591
Code: aa1803e6 aa1903e7 a90023f5 947730f5 (d4210000)

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Martin Schiller <ms@dev.tdt.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 months agoMAINTAINERS: add reviewers for SMC Sockets
Jan Karcher [Wed, 14 Jun 2023 06:54:56 +0000 (08:54 +0200)]
MAINTAINERS: add reviewers for SMC Sockets

adding three people from Alibaba as reviewers for SMC.
They are currently working on improving SMC on other architectures than
s390 and help with reviewing patches on top.

Thank you D. Wythe, Tony Lu and Wen Gu for your contributions and
collaboration and welcome on board as reviewers!

Reviewed-by: Wenjia Zhang <wenjia@linux.ibm.com>
Signed-off-by: Jan Karcher <jaka@linux.ibm.com>
Acked-by: Tony Lu <tonylu@linux.alibaba.com>
Acked-by: Wen Gu <guwen@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 months agos390/ism: Fix trying to free already-freed IRQ by repeated ism_dev_exit()
Julian Ruess [Tue, 13 Jun 2023 12:25:37 +0000 (14:25 +0200)]
s390/ism: Fix trying to free already-freed IRQ by repeated ism_dev_exit()

This patch prevents the system from crashing when unloading the ISM module.

How to reproduce: Attach an ISM device and execute 'rmmod ism'.

Error-Log:
- Trying to free already-free IRQ 0
- WARNING: CPU: 1 PID: 966 at kernel/irq/manage.c:1890 free_irq+0x140/0x540

After calling ism_dev_exit() for each ISM device in the exit routine,
pci_unregister_driver() will execute ism_remove() for each ISM device.
Because ism_remove() also calls ism_dev_exit(),
free_irq(pci_irq_vector(pdev, 0), ism) is called twice for each ISM
device. This results in a crash with the error
'Trying to free already-free IRQ'.

In the exit routine, it is enough to call pci_unregister_driver()
because it ensures that ism_dev_exit() is called once per
ISM device.

Cc: <stable@vger.kernel.org> # 6.3+
Fixes: 89e7d2ba61b7 ("net/ism: Add new API for client registration")
Reviewed-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Julian Ruess <julianr@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 months agoLoongArch: Fix debugfs_create_dir() error checking
Immad Mir [Thu, 15 Jun 2023 06:35:56 +0000 (14:35 +0800)]
LoongArch: Fix debugfs_create_dir() error checking

The debugfs_create_dir() returns ERR_PTR in case of an error and the
correct way of checking it is using the IS_ERR_OR_NULL inline function
rather than the simple null comparision. This patch fixes the issue.

Cc: stable@vger.kernel.org
Suggested-By: Ivan Orlov <ivan.orlov0322@gmail.com>
Signed-off-by: Immad Mir <mirimmad17@gmail.com>
Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>