platform/kernel/linux-starfive.git
20 months agonet: microchip: sparx5: Adding initial tc flower support for VCAP API
Steen Hegelund [Thu, 20 Oct 2022 13:08:59 +0000 (15:08 +0200)]
net: microchip: sparx5: Adding initial tc flower support for VCAP API

This adds initial TC flower filter support to Sparx5 for the IS2 VCAP.

The support consists of the source and destination MAC addresses,
and the trap and pass actions.

This is how you can create a rule that test the functionality:

tc qdisc add dev eth0 clsact
tc filter add dev eth0 ingress chain 8000000 prio 10 handle 10 \
      protocol all flower skip_sw \
      dst_mac 0a:0b:0c:0d:0e:0f \
      src_mac 2:0:0:0:0:1 \
      action trap

The IS2 chains in Sparx5 are assigned like this:

- chain 8000000: IS2 Lookup 0
- chain 8100000: IS2 Lookup 1
- chain 8200000: IS2 Lookup 2
- chain 8300000: IS2 Lookup 3

Signed-off-by: Steen Hegelund <steen.hegelund@microchip.com>
Tested-by: Casper Andersson <casper.casan@gmail.com>
Reviewed-by: Casper Andersson <casper.casan@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: microchip: sparx5: Adding IS2 VCAP register interface
Steen Hegelund [Thu, 20 Oct 2022 13:08:58 +0000 (15:08 +0200)]
net: microchip: sparx5: Adding IS2 VCAP register interface

This adds the register interface needed to access the Sparx5 Ingress Stage
2 VCAP (IS2).

The Sparx5 Chip Register Model can be browsed at this location:
https://github.com/microchip-ung/sparx-5_reginfo

Signed-off-by: Steen Hegelund <steen.hegelund@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: microchip: sparx5: Adding IS2 VCAP model to VCAP API
Steen Hegelund [Thu, 20 Oct 2022 13:08:57 +0000 (15:08 +0200)]
net: microchip: sparx5: Adding IS2 VCAP model to VCAP API

This provides the Sparx5 Ingress Stage 2 (IS2) model and adds it to the
VCAP control instance that will be provided to the VCAP API.

The Sparx5 IS2 C code model is generated from the Sparx5 RTL design model.

Signed-off-by: Steen Hegelund <steen.hegelund@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: microchip: sparx5: Adding initial VCAP API support
Steen Hegelund [Thu, 20 Oct 2022 13:08:56 +0000 (15:08 +0200)]
net: microchip: sparx5: Adding initial VCAP API support

This provides the initial VCAP API framework and Sparx5 specific VCAP
implementation.

When the Sparx5 Switchdev driver is initialized it will also initialize its
VCAP module, and this hooks up the concrete Sparx5 VCAP model to the VCAP
API, so that the VCAP API knows what VCAP instances are available.

Signed-off-by: Steen Hegelund <steen.hegelund@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonfp: flower: tunnel neigh support bond offload
Yanguo Li [Thu, 20 Oct 2022 08:28:34 +0000 (09:28 +0100)]
nfp: flower: tunnel neigh support bond offload

Support hardware offload when tunnel neigh out port is bond.
These feature work with the nfp firmware. If the firmware
supports the NFP_FL_FEATS_TUNNEL_NEIGH_LAG feature, nfp driver
write the bond information to the firmware neighbor table or
do nothing for bond. when neighbor MAC changes, nfp driver
need to update the neighbor information too.

Signed-off-by: Yanguo Li <yanguo.li@corigine.com>
Reviewed-by: Louis Peens <louis.peens@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agoMerge branch 'inet6_destroy_sock-calls-remove'
David S. Miller [Mon, 24 Oct 2022 08:40:39 +0000 (09:40 +0100)]
Merge branch 'inet6_destroy_sock-calls-remove'

Kuniyuki Iwashima says:

====================
inet6: Remove inet6_destroy_sock() calls.

This is a follow-up series for commit d38afeec26ed ("tcp/udp: Call
inet6_destroy_sock() in IPv6 sk->sk_destruct().").

This series cleans up unnecessary inet6_destory_sock() calls in
sk->sk_prot->destroy() and call it from sk->sk_destruct() to make
sure we do not leak memory related to IPv6 specific-resources.

Changes:
  v2:
    * patch 1
      * Fix build failure for CONFIG_MPTCP_IPV6=y

  v1: https://lore.kernel.org/netdev/20221018190956.1308-1-kuniyu@amazon.com/
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agoinet6: Clean up failure path in do_ipv6_setsockopt().
Kuniyuki Iwashima [Wed, 19 Oct 2022 22:36:03 +0000 (15:36 -0700)]
inet6: Clean up failure path in do_ipv6_setsockopt().

We can reuse the unlock label above and need not repeat the same code.

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agoinet6: Remove inet6_destroy_sock().
Kuniyuki Iwashima [Wed, 19 Oct 2022 22:36:02 +0000 (15:36 -0700)]
inet6: Remove inet6_destroy_sock().

The last user of inet6_destroy_sock() is its wrapper inet6_cleanup_sock().
Let's rename inet6_destroy_sock() to inet6_cleanup_sock().

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agosctp: Call inet6_destroy_sock() via sk->sk_destruct().
Kuniyuki Iwashima [Wed, 19 Oct 2022 22:36:01 +0000 (15:36 -0700)]
sctp: Call inet6_destroy_sock() via sk->sk_destruct().

After commit d38afeec26ed ("tcp/udp: Call inet6_destroy_sock()
in IPv6 sk->sk_destruct()."), we call inet6_destroy_sock() in
sk->sk_destruct() by setting inet6_sock_destruct() to it to make
sure we do not leak inet6-specific resources.

SCTP sets its own sk->sk_destruct() in the sctp_init_sock(), and
SCTPv6 socket reuses it as the init function.

To call inet6_sock_destruct() from SCTPv6 sk->sk_destruct(), we
set sctp_v6_destruct_sock() in a new init function.

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agodccp: Call inet6_destroy_sock() via sk->sk_destruct().
Kuniyuki Iwashima [Wed, 19 Oct 2022 22:36:00 +0000 (15:36 -0700)]
dccp: Call inet6_destroy_sock() via sk->sk_destruct().

After commit d38afeec26ed ("tcp/udp: Call inet6_destroy_sock()
in IPv6 sk->sk_destruct()."), we call inet6_destroy_sock() in
sk->sk_destruct() by setting inet6_sock_destruct() to it to make
sure we do not leak inet6-specific resources.

DCCP sets its own sk->sk_destruct() in the dccp_init_sock(), and
DCCPv6 socket shares it by calling the same init function via
dccp_v6_init_sock().

To call inet6_sock_destruct() from DCCPv6 sk->sk_destruct(), we
export it and set dccp_v6_sk_destruct() in the init function.

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agoinet6: Remove inet6_destroy_sock() in sk->sk_prot->destroy().
Kuniyuki Iwashima [Wed, 19 Oct 2022 22:35:59 +0000 (15:35 -0700)]
inet6: Remove inet6_destroy_sock() in sk->sk_prot->destroy().

After commit d38afeec26ed ("tcp/udp: Call inet6_destroy_sock()
in IPv6 sk->sk_destruct()."), we call inet6_destroy_sock() in
sk->sk_destruct() by setting inet6_sock_destruct() to it to make
sure we do not leak inet6-specific resources.

Now we can remove unnecessary inet6_destroy_sock() calls in
sk->sk_prot->destroy().

DCCP and SCTP have their own sk->sk_destruct() function, so we
change them separately in the following patches.

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agoMerge branch 'dpaa2-eth-AF_XDP-zc'
David S. Miller [Mon, 24 Oct 2022 08:22:12 +0000 (09:22 +0100)]
Merge branch 'dpaa2-eth-AF_XDP-zc'

Ioana Ciornei says:

====================
net: dpaa2-eth: AF_XDP zero-copy support

This patch set adds support for AF_XDP zero-copy in the dpaa2-eth
driver. The support is available on the LX2160A SoC and its variants and
only on interfaces (DPNIs) with a maximum of 8 queues (HW limitations
are the root cause).

We are first implementing the .get_channels() callback since this a
dependency for further work.

Patches 2-3 are working on making the necessary changes for multiple
buffer pools on a single interface. By default, without an AF_XDP socket
attached, only a single buffer pool will be used and shared between all
the queues. The changes in the functions are made in this patch, but the
actual allocation and setup of a new BP is done in patch#10.

Patches 4-5 are improving the information exposed in debugfs. We are
exposing a new file to show which buffer pool is used by what channels
and how many buffers it currently has.

The 6th patch updates the dpni_set_pools() firmware API so that we are
capable of setting up a different buffer per queue in later patches.

In the 7th patch the generic dev_open/close APIs are used instead of the
dpaa2-eth internal ones.

Patches 8-9 are rearranging the existing code in dpaa2-eth.c in order to
create new functions which will be used in the XSK implementation in
dpaa2-xsk.c

Finally, the last 3 patches are adding the actual support for both the
Rx and Tx path of AF_XDP zero-copy and some associated tracepoints.
Details on the implementation can be found in the actual patch.

Changes in v2:
 - 3/12:  Export dpaa2_eth_allocate_dpbp/dpaa2_eth_free_dpbp in this
   patch to avoid a build warning. The functions will be used in next
   patches.
 - 6/12:  Use __le16 instead of u16 for the dpbp_id field.
 - 12/12: Use xdp_buff->data_hard_start when tracing the BP seeding.

Changes in v3:
 - 3/12: fix leaking of bp on error path
====================

Acked-by: Björn Töpel <bjorn@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: dpaa2-eth: add trace points on XSK events
Robert-Ionut Alexa [Tue, 18 Oct 2022 14:19:01 +0000 (17:19 +0300)]
net: dpaa2-eth: add trace points on XSK events

Define the dpaa2_tx_xsk_fd and dpaa2_rx_xsk_fd trace events for the XSK
zero-copy Rx and Tx path.  Also, define the dpaa2_eth_buf as an event
class so that both dpaa2_eth_buf_seed and dpaa2_xsk_buf_seed traces can
derive from the same class.

Signed-off-by: Robert-Ionut Alexa <robert-ionut.alexa@nxp.com>
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: dpaa2-eth: AF_XDP TX zero copy support
Robert-Ionut Alexa [Tue, 18 Oct 2022 14:19:00 +0000 (17:19 +0300)]
net: dpaa2-eth: AF_XDP TX zero copy support

Add support in dpaa2-eth for packet processing on the Tx path using
AF_XDP zero copy mode.

The newly added dpaa2_xsk_tx() function will handle enqueuing AF_XDP Tx
packets into the appropriate queue and update any necessary statistics.

On a more detailed note, the dpaa2_xsk_tx_build_fd() function handles
creating a Scatter-Gather frame descriptor with only one data buffer.
This is needed because otherwise we would need to impose a headroom in
the Tx buffer to store our software annotation structures.
This tactic is already used on the normal data path of the dpaa2-eth
driver, thus we are reusing the dpaa2_eth_sgt_get/dpaa2_eth_sgt_recycle
functions in order to allocate and recycle the Scatter-Gather table
buffers.

In case we have reached the maximum number of Tx XSK packets to be sent
in a NAPI cycle, we'll exit the dpaa2_eth_poll() and hope to be
rescheduled again.

On the XSK Tx confirmation path, we are just unmapping the SGT buffer
and recycle it for further use.

Signed-off-by: Robert-Ionut Alexa <robert-ionut.alexa@nxp.com>
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: dpaa2-eth: AF_XDP RX zero copy support
Robert-Ionut Alexa [Tue, 18 Oct 2022 14:18:59 +0000 (17:18 +0300)]
net: dpaa2-eth: AF_XDP RX zero copy support

This patch adds the support for receiving packets via the AF_XDP
zero-copy mechanism in the dpaa2-eth driver. The support is available
only on the LX2160A SoC and variants because we are relying on the HW
capability to associate a buffer pool to a specific queue (QDBIN), only
available on newer WRIOP versions.

On the control path, the dpaa2_xsk_enable_pool() function is responsible
to allocate a buffer pool (BP), setup this new BP to be used only on the
requested queue and change the consume function to point to the XSK ZC
one.
We are forced to call dev_close() in order to change the queue to buffer
pool association (dpaa2_xsk_set_bp_per_qdbin) . This also works in our
favor since at dev_close() the buffer pools will be drained and at the
later dev_open() call they will be again seeded, this time with buffers
allocated from the XSK pool if needed.

On the data path, a new software annotation type is defined to be used
only for the XSK scenarios. This will enable us to pass keep necessary
information about a packet buffer between the moment in which it was
seeded and when it's received by the driver. In the XSK case, we are
keeping the associated xdp_buff.
Depending on the action returned by the BPF program, we will do the
following:
 - XDP_PASS: copy the contents of the packet into a brand new skb,
   recycle the initial buffer.
 - XDP_TX: just enqueue the same frame descriptor back into the Tx path,
   the buffer will get automatically released into the initial BP.
 - XDP_REDIRECT: call xdp_do_redirect() and exit.

Signed-off-by: Robert-Ionut Alexa <robert-ionut.alexa@nxp.com>
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: dpaa2-eth: create and export the dpaa2_eth_receive_skb() function
Robert-Ionut Alexa [Tue, 18 Oct 2022 14:18:58 +0000 (17:18 +0300)]
net: dpaa2-eth: create and export the dpaa2_eth_receive_skb() function

Carve out code from the dpaa2_eth_rx() function in order to create and
export the dpaa2_eth_receive_skb() function. Do this in order to reuse
this code also from the XSK path which will be introduced in a later
patch.

Signed-off-by: Robert-Ionut Alexa <robert-ionut.alexa@nxp.com>
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: dpaa2-eth: create and export the dpaa2_eth_alloc_skb function
Robert-Ionut Alexa [Tue, 18 Oct 2022 14:18:57 +0000 (17:18 +0300)]
net: dpaa2-eth: create and export the dpaa2_eth_alloc_skb function

The dpaa2_eth_alloc_skb() function is added by moving code from the
dpaa2_eth_copybreak() previously defined function. What the new API does
is to allocate a new skb, copy the frame data from the passed FD to the
new skb and then return the skb.
Export this new function since we'll need the this functionality also
from the XSK code path.

Signed-off-by: Robert-Ionut Alexa <robert-ionut.alexa@nxp.com>
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: dpaa2-eth: use dev_close/open instead of the internal functions
Ioana Ciornei [Tue, 18 Oct 2022 14:18:56 +0000 (17:18 +0300)]
net: dpaa2-eth: use dev_close/open instead of the internal functions

Instead of calling the internal functions which implement .ndo_stop and
.ndo_open, we can simply call dev_close and dev_open, so that we keep
the code cleaner.

Also, in the next patches we'll use the same APIs from other files
without needing to export the internal functions.

Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: dpaa2-eth: update the dpni_set_pools() API to support per QDBIN pools
Robert-Ionut Alexa [Tue, 18 Oct 2022 14:18:55 +0000 (17:18 +0300)]
net: dpaa2-eth: update the dpni_set_pools() API to support per QDBIN pools

Update the dpni_set_pool() firmware API so that in the next patches we
can configure per Rx queue (per QDBIN) buffer pools.
This is a hard requirement of the AF_XDP, thus we need the newer API
version.

Signed-off-by: Robert-Ionut Alexa <robert-ionut.alexa@nxp.com>
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: dpaa2-eth: export buffer pool info into a new debugfs file
Ioana Ciornei [Tue, 18 Oct 2022 14:18:54 +0000 (17:18 +0300)]
net: dpaa2-eth: export buffer pool info into a new debugfs file

Export the allocated buffer pools, the number of buffers that they have
currently and which channels are using which BP.

The output looks like below:

Buffer pool info for eth2:
IDX        BPID      Buf count      CH#0      CH#1      CH#2      CH#3
BP#0         1           5124         x         x         x         x

Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: dpaa2-eth: export the CH#<index> in the 'ch_stats' debug file
Ioana Ciornei [Tue, 18 Oct 2022 14:18:53 +0000 (17:18 +0300)]
net: dpaa2-eth: export the CH#<index> in the 'ch_stats' debug file

Just give out an index for each channel that we export into the debug
file in the form of CH#<index>. This is purely to help corelate each
channel information from one debugfs file to another one.

Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: dpaa2-eth: add support for multiple buffer pools per DPNI
Robert-Ionut Alexa [Tue, 18 Oct 2022 14:18:52 +0000 (17:18 +0300)]
net: dpaa2-eth: add support for multiple buffer pools per DPNI

This patch allows the configuration of multiple buffer pools associated
with a single DPNI object, each distinct DPBP object not necessarily
shared among all queues.
The user can interogate both the number of buffer pools and the buffer
count in each buffer pool by using the .get_ethtool_stats() callback.

Signed-off-by: Robert-Ionut Alexa <robert-ionut.alexa@nxp.com>
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: dpaa2-eth: rearrange variable in dpaa2_eth_get_ethtool_stats
Ioana Ciornei [Tue, 18 Oct 2022 14:18:51 +0000 (17:18 +0300)]
net: dpaa2-eth: rearrange variable in dpaa2_eth_get_ethtool_stats

Rearrange the variables in the dpaa2_eth_get_ethtool_stats() function so
that we adhere to the reverse Christmas tree rule.
Also, in the next patch we are adding more variables and I didn't know
where to place them with the current ordering.

Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: dpaa2-eth: add support to query the number of queues through ethtool
Robert-Ionut Alexa [Tue, 18 Oct 2022 14:18:50 +0000 (17:18 +0300)]
net: dpaa2-eth: add support to query the number of queues through ethtool

The .get_channels() ethtool_ops callback is implemented and exports the
number of queues: Rx, Tx, Tx conf and Rx err.
The last two ones, Tx confirmation and Rx err, are counted as 'others'.

The .set_channels() callback is not implemented since the DPAA2
software/firmware architecture does not allow the dynamic
reconfiguration of the number of queues.

Signed-off-by: Robert-Ionut Alexa <robert-ionut.alexa@nxp.com>
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agodt-bindings: net: adin1110: Document reset
Alexandru Tachici [Wed, 19 Oct 2022 17:13:14 +0000 (20:13 +0300)]
dt-bindings: net: adin1110: Document reset

Document GPIO for HW reset.

Signed-off-by: Alexandru Tachici <alexandru.tachici@analog.com>
Acked-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: ethernet: adi: adin1110: add reset GPIO
Alexandru Tachici [Wed, 19 Oct 2022 17:13:13 +0000 (20:13 +0300)]
net: ethernet: adi: adin1110: add reset GPIO

Add an optional GPIO to be used for a hardware reset of the IC.

Signed-off-by: Alexandru Tachici <alexandru.tachici@analog.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: ipa: Make QMI message rules const
Jeff Johnson [Tue, 18 Oct 2022 21:17:18 +0000 (14:17 -0700)]
net: ipa: Make QMI message rules const

Commit ff6d365898d4 ("soc: qcom: qmi: use const for struct
qmi_elem_info") allows QMI message encoding/decoding rules to be
const, so do that for IPA.

Signed-off-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Reviewed-by: Alex Elder <elder@linaro.org>
Reviewed-by: Sibi Sankar <quic_sibis@quicinc.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: bcmgenet: add RX_CLS_LOC_ANY support
Doug Berger [Wed, 19 Oct 2022 21:51:23 +0000 (14:51 -0700)]
net: bcmgenet: add RX_CLS_LOC_ANY support

If a matching flow spec exists its current location is as good
as ANY. If not add the new flow spec at the first available
location.

Signed-off-by: Doug Berger <opendmb@gmail.com>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Link: https://lore.kernel.org/r/20221019215123.316997-1-opendmb@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
20 months agosctp: remove unnecessary NULL checks in sctp_enqueue_event()
Alexey Kodanev [Wed, 19 Oct 2022 18:07:35 +0000 (21:07 +0300)]
sctp: remove unnecessary NULL checks in sctp_enqueue_event()

After commit 178ca044aa60 ("sctp: Make sctp_enqueue_event tak an
skb list."), skb_list cannot be NULL.

Detected using the static analysis tool - Svace.
Signed-off-by: Alexey Kodanev <aleksei.kodanev@bell-sw.com>
Reviewed-by: Xin Long <lucien.xin@gmail.com>
Link: https://lore.kernel.org/r/20221019180735.161388-3-aleksei.kodanev@bell-sw.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
20 months agosctp: remove unnecessary NULL check in sctp_ulpq_tail_event()
Alexey Kodanev [Wed, 19 Oct 2022 18:07:34 +0000 (21:07 +0300)]
sctp: remove unnecessary NULL check in sctp_ulpq_tail_event()

After commit 013b96ec6461 ("sctp: Pass sk_buff_head explicitly to
sctp_ulpq_tail_event().") there is one more unneeded check of
skb_list for NULL.

Detected using the static analysis tool - Svace.
Signed-off-by: Alexey Kodanev <aleksei.kodanev@bell-sw.com>
Reviewed-by: Xin Long <lucien.xin@gmail.com>
Link: https://lore.kernel.org/r/20221019180735.161388-2-aleksei.kodanev@bell-sw.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
20 months agosctp: remove unnecessary NULL check in sctp_association_init()
Alexey Kodanev [Wed, 19 Oct 2022 18:07:33 +0000 (21:07 +0300)]
sctp: remove unnecessary NULL check in sctp_association_init()

'&asoc->ulpq' passed to sctp_ulpq_init() as the first argument,
then sctp_qlpq_init() initializes it and eventually returns the
address of the struct member back. Therefore, in this case, the
return pointer cannot be NULL.

Moreover, it seems sctp_ulpq_init() has always been used only in
sctp_association_init(), so there's really no need to return ulpq
anymore.

Detected using the static analysis tool - Svace.
Signed-off-by: Alexey Kodanev <aleksei.kodanev@bell-sw.com>
Reviewed-by: Xin Long <lucien.xin@gmail.com>
Link: https://lore.kernel.org/r/20221019180735.161388-1-aleksei.kodanev@bell-sw.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
20 months agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Jakub Kicinski [Fri, 21 Oct 2022 00:49:10 +0000 (17:49 -0700)]
Merge git://git./linux/kernel/git/netdev/net

No conflicts.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
20 months agoMerge tag 'net-6.1-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Linus Torvalds [Fri, 21 Oct 2022 00:24:59 +0000 (17:24 -0700)]
Merge tag 'net-6.1-rc2' of git://git./linux/kernel/git/netdev/net

Pull networking fixes from Paolo Abeni:
 "Including fixes from netfilter.

  Current release - regressions:

   - revert "net: fix cpu_max_bits_warn() usage in
     netif_attrmask_next{,_and}"

   - revert "net: sched: fq_codel: remove redundant resource cleanup in
     fq_codel_init()"

   - dsa: uninitialized variable in dsa_slave_netdevice_event()

   - eth: sunhme: uninitialized variable in happy_meal_init()

  Current release - new code bugs:

   - eth: octeontx2: fix resource not freed after malloc

  Previous releases - regressions:

   - sched: fix return value of qdisc ingress handling on success

   - sched: fix race condition in qdisc_graft()

   - udp: update reuse->has_conns under reuseport_lock.

   - tls: strp: make sure the TCP skbs do not have overlapping data

   - hsr: avoid possible NULL deref in skb_clone()

   - tipc: fix an information leak in tipc_topsrv_kern_subscr

   - phylink: add mac_managed_pm in phylink_config structure

   - eth: i40e: fix DMA mappings leak

   - eth: hyperv: fix a RX-path warning

   - eth: mtk: fix memory leaks

  Previous releases - always broken:

   - sched: cake: fix null pointer access issue when cake_init() fails"

* tag 'net-6.1-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (43 commits)
  net: phy: dp83822: disable MDI crossover status change interrupt
  net: sched: fix race condition in qdisc_graft()
  net: hns: fix possible memory leak in hnae_ae_register()
  wwan_hwsim: fix possible memory leak in wwan_hwsim_dev_new()
  sfc: include vport_id in filter spec hash and equal()
  genetlink: fix kdoc warnings
  selftests: add selftest for chaining of tc ingress handling to egress
  net: Fix return value of qdisc ingress handling on success
  net: sched: sfb: fix null pointer access issue when sfb_init() fails
  Revert "net: sched: fq_codel: remove redundant resource cleanup in fq_codel_init()"
  net: sched: cake: fix null pointer access issue when cake_init() fails
  ethernet: marvell: octeontx2 Fix resource not freed after malloc
  netfilter: nf_tables: relax NFTA_SET_ELEM_KEY_END set flags requirements
  netfilter: rpfilter/fib: Set ->flowic_uid correctly for user namespaces.
  ionic: catch NULL pointer issue on reconfig
  net: hsr: avoid possible NULL deref in skb_clone()
  bnxt_en: fix memory leak in bnxt_nvm_test()
  ip6mr: fix UAF issue in ip6mr_sk_done() when addrconf_init_net() failed
  udp: Update reuse->has_conns under reuseport_lock.
  net: ethernet: mediatek: ppe: Remove the unused function mtk_foe_entry_usable()
  ...

20 months agoMerge tag 'ata-6.1-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/dlemoal...
Linus Torvalds [Fri, 21 Oct 2022 00:07:54 +0000 (17:07 -0700)]
Merge tag 'ata-6.1-rc2' of git://git./linux/kernel/git/dlemoal/libata

Pull ata fixes from Damien Le Moal:
 "Several minor fixes:

   - Fix the module alias for the ahci_imx driver to get autoloading to
     work (Alexander)

   - Fix a potential array-index-out-of-bounds problem with the
     enclosure managment support in the ahci driver (Kai-Heng)

   - Several patches to fix compilation warnings thrown by clang in the
     ahci_st, sata_rcar, ahci_brcm, ahci_xgene, ahci_imx and ahci_qoriq
     drivers (me)"

* tag 'ata-6.1-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/dlemoal/libata:
  ata: ahci_qoriq: Fix compilation warning
  ata: ahci_imx: Fix compilation warning
  ata: ahci_xgene: Fix compilation warning
  ata: ahci_brcm: Fix compilation warning
  ata: sata_rcar: Fix compilation warning
  ata: ahci_st: Fix compilation warning
  ata: ahci: Match EM_MAX_SLOTS with SATA_PMP_MAX_PORTS
  ata: ahci-imx: Fix MODULE_ALIAS

20 months agoMerge tag 'for-6.1/dm-changes-v2' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 21 Oct 2022 00:00:54 +0000 (17:00 -0700)]
Merge tag 'for-6.1/dm-changes-v2' of git://git./linux/kernel/git/device-mapper/linux-dm

Pull device mapper updates from Mike Snitzer:

 - Fix dm-bufio to use test_bit_acquire to properly test_bit on arches
   with weaker memory ordering.

 - DM core replace DMWARN with DMERR or DMCRIT for fatal errors.

 - Enable WQ_HIGHPRI on DM verity target's verify_wq.

 - Add documentation for DM verity's try_verify_in_tasklet option.

 - Various typo and redundant word fixes in code and/or comments.

* tag 'for-6.1/dm-changes-v2' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
  dm clone: Fix typo in block_device format specifier
  dm: remove unnecessary assignment statement in alloc_dev()
  dm verity: Add documentation for try_verify_in_tasklet option
  dm cache: delete the redundant word 'each' in comment
  dm raid: fix typo in analyse_superblocks code comment
  dm verity: enable WQ_HIGHPRI on verify_wq
  dm raid: delete the redundant word 'that' in comment
  dm: change from DMWARN to DMERR or DMCRIT for fatal errors
  dm bufio: use the acquire memory barrier when testing for B_READING

20 months agonet: ipa: Proactively round up to kmalloc bucket size
Kees Cook [Tue, 18 Oct 2022 09:28:27 +0000 (02:28 -0700)]
net: ipa: Proactively round up to kmalloc bucket size

Instead of discovering the kmalloc bucket size _after_ allocation, round
up proactively so the allocation is explicitly made for the full size,
allowing the compiler to correctly reason about the resulting size of
the buffer through the existing __alloc_size() hint.

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: netdev@vger.kernel.org
Reviewed-by: Alex Elder <elder@linaro.org>
Link: https://lore.kernel.org/lkml/4d75a9fd-1b94-7208-9de8-5a0102223e68@ieee.org
Signed-off-by: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20221018092724.give.735-kees@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
20 months agonet: phy: dp83822: disable MDI crossover status change interrupt
Felix Riemann [Tue, 18 Oct 2022 10:47:54 +0000 (12:47 +0200)]
net: phy: dp83822: disable MDI crossover status change interrupt

If the cable is disconnected the PHY seems to toggle between MDI and
MDI-X modes. With the MDI crossover status interrupt active this causes
roughly 10 interrupts per second.

As the crossover status isn't checked by the driver, the interrupt can
be disabled to reduce the interrupt load.

Fixes: 87461f7a58ab ("net: phy: DP83822 initial driver submission")
Signed-off-by: Felix Riemann <felix.riemann@sma.de>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20221018104755.30025-1-svc.sw.rte.linux@sma.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
20 months agonet: sched: fix race condition in qdisc_graft()
Eric Dumazet [Tue, 18 Oct 2022 20:32:58 +0000 (20:32 +0000)]
net: sched: fix race condition in qdisc_graft()

We had one syzbot report [1] in syzbot queue for a while.
I was waiting for more occurrences and/or a repro but
Dmitry Vyukov spotted the issue right away.

<quoting Dmitry>
qdisc_graft() drops reference to qdisc in notify_and_destroy
while it's still assigned to dev->qdisc
</quoting>

Indeed, RCU rules are clear when replacing a data structure.
The visible pointer (dev->qdisc in this case) must be updated
to the new object _before_ RCU grace period is started
(qdisc_put(old) in this case).

[1]
BUG: KASAN: use-after-free in __tcf_qdisc_find.part.0+0xa3a/0xac0 net/sched/cls_api.c:1066
Read of size 4 at addr ffff88802065e038 by task syz-executor.4/21027

CPU: 0 PID: 21027 Comm: syz-executor.4 Not tainted 6.0.0-rc3-syzkaller-00363-g7726d4c3e60b #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 08/26/2022
Call Trace:
<TASK>
__dump_stack lib/dump_stack.c:88 [inline]
dump_stack_lvl+0xcd/0x134 lib/dump_stack.c:106
print_address_description mm/kasan/report.c:317 [inline]
print_report.cold+0x2ba/0x719 mm/kasan/report.c:433
kasan_report+0xb1/0x1e0 mm/kasan/report.c:495
__tcf_qdisc_find.part.0+0xa3a/0xac0 net/sched/cls_api.c:1066
__tcf_qdisc_find net/sched/cls_api.c:1051 [inline]
tc_new_tfilter+0x34f/0x2200 net/sched/cls_api.c:2018
rtnetlink_rcv_msg+0x955/0xca0 net/core/rtnetlink.c:6081
netlink_rcv_skb+0x153/0x420 net/netlink/af_netlink.c:2501
netlink_unicast_kernel net/netlink/af_netlink.c:1319 [inline]
netlink_unicast+0x543/0x7f0 net/netlink/af_netlink.c:1345
netlink_sendmsg+0x917/0xe10 net/netlink/af_netlink.c:1921
sock_sendmsg_nosec net/socket.c:714 [inline]
sock_sendmsg+0xcf/0x120 net/socket.c:734
____sys_sendmsg+0x6eb/0x810 net/socket.c:2482
___sys_sendmsg+0x110/0x1b0 net/socket.c:2536
__sys_sendmsg+0xf3/0x1c0 net/socket.c:2565
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x63/0xcd
RIP: 0033:0x7f5efaa89279
Code: ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 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 c7 c1 b8 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007f5efbc31168 EFLAGS: 00000246 ORIG_RAX: 000000000000002e
RAX: ffffffffffffffda RBX: 00007f5efab9bf80 RCX: 00007f5efaa89279
RDX: 0000000000000000 RSI: 0000000020000140 RDI: 0000000000000005
RBP: 00007f5efaae32e9 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
R13: 00007f5efb0cfb1f R14: 00007f5efbc31300 R15: 0000000000022000
</TASK>

Allocated by task 21027:
kasan_save_stack+0x1e/0x40 mm/kasan/common.c:38
kasan_set_track mm/kasan/common.c:45 [inline]
set_alloc_info mm/kasan/common.c:437 [inline]
____kasan_kmalloc mm/kasan/common.c:516 [inline]
____kasan_kmalloc mm/kasan/common.c:475 [inline]
__kasan_kmalloc+0xa9/0xd0 mm/kasan/common.c:525
kmalloc_node include/linux/slab.h:623 [inline]
kzalloc_node include/linux/slab.h:744 [inline]
qdisc_alloc+0xb0/0xc50 net/sched/sch_generic.c:938
qdisc_create_dflt+0x71/0x4a0 net/sched/sch_generic.c:997
attach_one_default_qdisc net/sched/sch_generic.c:1152 [inline]
netdev_for_each_tx_queue include/linux/netdevice.h:2437 [inline]
attach_default_qdiscs net/sched/sch_generic.c:1170 [inline]
dev_activate+0x760/0xcd0 net/sched/sch_generic.c:1229
__dev_open+0x393/0x4d0 net/core/dev.c:1441
__dev_change_flags+0x583/0x750 net/core/dev.c:8556
rtnl_configure_link+0xee/0x240 net/core/rtnetlink.c:3189
rtnl_newlink_create net/core/rtnetlink.c:3371 [inline]
__rtnl_newlink+0x10b8/0x17e0 net/core/rtnetlink.c:3580
rtnl_newlink+0x64/0xa0 net/core/rtnetlink.c:3593
rtnetlink_rcv_msg+0x43a/0xca0 net/core/rtnetlink.c:6090
netlink_rcv_skb+0x153/0x420 net/netlink/af_netlink.c:2501
netlink_unicast_kernel net/netlink/af_netlink.c:1319 [inline]
netlink_unicast+0x543/0x7f0 net/netlink/af_netlink.c:1345
netlink_sendmsg+0x917/0xe10 net/netlink/af_netlink.c:1921
sock_sendmsg_nosec net/socket.c:714 [inline]
sock_sendmsg+0xcf/0x120 net/socket.c:734
____sys_sendmsg+0x6eb/0x810 net/socket.c:2482
___sys_sendmsg+0x110/0x1b0 net/socket.c:2536
__sys_sendmsg+0xf3/0x1c0 net/socket.c:2565
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x63/0xcd

Freed by task 21020:
kasan_save_stack+0x1e/0x40 mm/kasan/common.c:38
kasan_set_track+0x21/0x30 mm/kasan/common.c:45
kasan_set_free_info+0x20/0x30 mm/kasan/generic.c:370
____kasan_slab_free mm/kasan/common.c:367 [inline]
____kasan_slab_free+0x166/0x1c0 mm/kasan/common.c:329
kasan_slab_free include/linux/kasan.h:200 [inline]
slab_free_hook mm/slub.c:1754 [inline]
slab_free_freelist_hook+0x8b/0x1c0 mm/slub.c:1780
slab_free mm/slub.c:3534 [inline]
kfree+0xe2/0x580 mm/slub.c:4562
rcu_do_batch kernel/rcu/tree.c:2245 [inline]
rcu_core+0x7b5/0x1890 kernel/rcu/tree.c:2505
__do_softirq+0x1d3/0x9c6 kernel/softirq.c:571

Last potentially related work creation:
kasan_save_stack+0x1e/0x40 mm/kasan/common.c:38
__kasan_record_aux_stack+0xbe/0xd0 mm/kasan/generic.c:348
call_rcu+0x99/0x790 kernel/rcu/tree.c:2793
qdisc_put+0xcd/0xe0 net/sched/sch_generic.c:1083
notify_and_destroy net/sched/sch_api.c:1012 [inline]
qdisc_graft+0xeb1/0x1270 net/sched/sch_api.c:1084
tc_modify_qdisc+0xbb7/0x1a00 net/sched/sch_api.c:1671
rtnetlink_rcv_msg+0x43a/0xca0 net/core/rtnetlink.c:6090
netlink_rcv_skb+0x153/0x420 net/netlink/af_netlink.c:2501
netlink_unicast_kernel net/netlink/af_netlink.c:1319 [inline]
netlink_unicast+0x543/0x7f0 net/netlink/af_netlink.c:1345
netlink_sendmsg+0x917/0xe10 net/netlink/af_netlink.c:1921
sock_sendmsg_nosec net/socket.c:714 [inline]
sock_sendmsg+0xcf/0x120 net/socket.c:734
____sys_sendmsg+0x6eb/0x810 net/socket.c:2482
___sys_sendmsg+0x110/0x1b0 net/socket.c:2536
__sys_sendmsg+0xf3/0x1c0 net/socket.c:2565
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x63/0xcd

Second to last potentially related work creation:
kasan_save_stack+0x1e/0x40 mm/kasan/common.c:38
__kasan_record_aux_stack+0xbe/0xd0 mm/kasan/generic.c:348
kvfree_call_rcu+0x74/0x940 kernel/rcu/tree.c:3322
neigh_destroy+0x431/0x630 net/core/neighbour.c:912
neigh_release include/net/neighbour.h:454 [inline]
neigh_cleanup_and_release+0x1f8/0x330 net/core/neighbour.c:103
neigh_del net/core/neighbour.c:225 [inline]
neigh_remove_one+0x37d/0x460 net/core/neighbour.c:246
neigh_forced_gc net/core/neighbour.c:276 [inline]
neigh_alloc net/core/neighbour.c:447 [inline]
___neigh_create+0x18b5/0x29a0 net/core/neighbour.c:642
ip6_finish_output2+0xfb8/0x1520 net/ipv6/ip6_output.c:125
__ip6_finish_output net/ipv6/ip6_output.c:195 [inline]
ip6_finish_output+0x690/0x1160 net/ipv6/ip6_output.c:206
NF_HOOK_COND include/linux/netfilter.h:296 [inline]
ip6_output+0x1ed/0x540 net/ipv6/ip6_output.c:227
dst_output include/net/dst.h:451 [inline]
NF_HOOK include/linux/netfilter.h:307 [inline]
NF_HOOK include/linux/netfilter.h:301 [inline]
mld_sendpack+0xa09/0xe70 net/ipv6/mcast.c:1820
mld_send_cr net/ipv6/mcast.c:2121 [inline]
mld_ifc_work+0x71c/0xdc0 net/ipv6/mcast.c:2653
process_one_work+0x991/0x1610 kernel/workqueue.c:2289
worker_thread+0x665/0x1080 kernel/workqueue.c:2436
kthread+0x2e4/0x3a0 kernel/kthread.c:376
ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:306

The buggy address belongs to the object at ffff88802065e000
which belongs to the cache kmalloc-1k of size 1024
The buggy address is located 56 bytes inside of
1024-byte region [ffff88802065e000ffff88802065e400)

The buggy address belongs to the physical page:
page:ffffea0000819600 refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x20658
head:ffffea0000819600 order:3 compound_mapcount:0 compound_pincount:0
flags: 0xfff00000010200(slab|head|node=0|zone=1|lastcpupid=0x7ff)
raw: 00fff00000010200 0000000000000000 dead000000000001 ffff888011841dc0
raw: 0000000000000000 0000000000100010 00000001ffffffff 0000000000000000
page dumped because: kasan: bad access detected
page_owner tracks the page as allocated
page last allocated via order 3, migratetype Unmovable, gfp_mask 0xd20c0(__GFP_IO|__GFP_FS|__GFP_NOWARN|__GFP_NORETRY|__GFP_COMP|__GFP_NOMEMALLOC), pid 3523, tgid 3523 (sshd), ts 41495190986, free_ts 41417713212
prep_new_page mm/page_alloc.c:2532 [inline]
get_page_from_freelist+0x109b/0x2ce0 mm/page_alloc.c:4283
__alloc_pages+0x1c7/0x510 mm/page_alloc.c:5515
alloc_pages+0x1a6/0x270 mm/mempolicy.c:2270
alloc_slab_page mm/slub.c:1824 [inline]
allocate_slab+0x27e/0x3d0 mm/slub.c:1969
new_slab mm/slub.c:2029 [inline]
___slab_alloc+0x7f1/0xe10 mm/slub.c:3031
__slab_alloc.constprop.0+0x4d/0xa0 mm/slub.c:3118
slab_alloc_node mm/slub.c:3209 [inline]
__kmalloc_node_track_caller+0x2f2/0x380 mm/slub.c:4955
kmalloc_reserve net/core/skbuff.c:358 [inline]
__alloc_skb+0xd9/0x2f0 net/core/skbuff.c:430
alloc_skb_fclone include/linux/skbuff.h:1307 [inline]
tcp_stream_alloc_skb+0x38/0x580 net/ipv4/tcp.c:861
tcp_sendmsg_locked+0xc36/0x2f80 net/ipv4/tcp.c:1325
tcp_sendmsg+0x2b/0x40 net/ipv4/tcp.c:1483
inet_sendmsg+0x99/0xe0 net/ipv4/af_inet.c:819
sock_sendmsg_nosec net/socket.c:714 [inline]
sock_sendmsg+0xcf/0x120 net/socket.c:734
sock_write_iter+0x291/0x3d0 net/socket.c:1108
call_write_iter include/linux/fs.h:2187 [inline]
new_sync_write fs/read_write.c:491 [inline]
vfs_write+0x9e9/0xdd0 fs/read_write.c:578
ksys_write+0x1e8/0x250 fs/read_write.c:631
page last free stack trace:
reset_page_owner include/linux/page_owner.h:24 [inline]
free_pages_prepare mm/page_alloc.c:1449 [inline]
free_pcp_prepare+0x5e4/0xd20 mm/page_alloc.c:1499
free_unref_page_prepare mm/page_alloc.c:3380 [inline]
free_unref_page+0x19/0x4d0 mm/page_alloc.c:3476
__unfreeze_partials+0x17c/0x1a0 mm/slub.c:2548
qlink_free mm/kasan/quarantine.c:168 [inline]
qlist_free_all+0x6a/0x170 mm/kasan/quarantine.c:187
kasan_quarantine_reduce+0x180/0x200 mm/kasan/quarantine.c:294
__kasan_slab_alloc+0xa2/0xc0 mm/kasan/common.c:447
kasan_slab_alloc include/linux/kasan.h:224 [inline]
slab_post_alloc_hook mm/slab.h:727 [inline]
slab_alloc_node mm/slub.c:3243 [inline]
slab_alloc mm/slub.c:3251 [inline]
__kmem_cache_alloc_lru mm/slub.c:3258 [inline]
kmem_cache_alloc+0x267/0x3b0 mm/slub.c:3268
kmem_cache_zalloc include/linux/slab.h:723 [inline]
alloc_buffer_head+0x20/0x140 fs/buffer.c:2974
alloc_page_buffers+0x280/0x790 fs/buffer.c:829
create_empty_buffers+0x2c/0xee0 fs/buffer.c:1558
ext4_block_write_begin+0x1004/0x1530 fs/ext4/inode.c:1074
ext4_da_write_begin+0x422/0xae0 fs/ext4/inode.c:2996
generic_perform_write+0x246/0x560 mm/filemap.c:3738
ext4_buffered_write_iter+0x15b/0x460 fs/ext4/file.c:270
ext4_file_write_iter+0x44a/0x1660 fs/ext4/file.c:679
call_write_iter include/linux/fs.h:2187 [inline]
new_sync_write fs/read_write.c:491 [inline]
vfs_write+0x9e9/0xdd0 fs/read_write.c:578

Fixes: af356afa010f ("net_sched: reintroduce dev->qdisc for use by sch_api")
Reported-by: syzbot <syzkaller@googlegroups.com>
Diagnosed-by: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20221018203258.2793282-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
20 months agonet: hns: fix possible memory leak in hnae_ae_register()
Yang Yingliang [Tue, 18 Oct 2022 12:24:51 +0000 (20:24 +0800)]
net: hns: fix possible memory leak in hnae_ae_register()

Inject fault while probing module, if device_register() fails,
but the refcount of kobject is not decreased to 0, the name
allocated in dev_set_name() is leaked. Fix this by calling
put_device(), so that name can be freed in callback function
kobject_cleanup().

unreferenced object 0xffff00c01aba2100 (size 128):
  comm "systemd-udevd", pid 1259, jiffies 4294903284 (age 294.152s)
  hex dump (first 32 bytes):
    68 6e 61 65 30 00 00 00 18 21 ba 1a c0 00 ff ff  hnae0....!......
    00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
  backtrace:
    [<0000000034783f26>] slab_post_alloc_hook+0xa0/0x3e0
    [<00000000748188f2>] __kmem_cache_alloc_node+0x164/0x2b0
    [<00000000ab0743e8>] __kmalloc_node_track_caller+0x6c/0x390
    [<000000006c0ffb13>] kvasprintf+0x8c/0x118
    [<00000000fa27bfe1>] kvasprintf_const+0x60/0xc8
    [<0000000083e10ed7>] kobject_set_name_vargs+0x3c/0xc0
    [<000000000b87affc>] dev_set_name+0x7c/0xa0
    [<000000003fd8fe26>] hnae_ae_register+0xcc/0x190 [hnae]
    [<00000000fe97edc9>] hns_dsaf_ae_init+0x9c/0x108 [hns_dsaf]
    [<00000000c36ff1eb>] hns_dsaf_probe+0x548/0x748 [hns_dsaf]

Fixes: 6fe6611ff275 ("net: add Hisilicon Network Subsystem hnae framework support")
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Link: https://lore.kernel.org/r/20221018122451.1749171-1-yangyingliang@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
20 months agowwan_hwsim: fix possible memory leak in wwan_hwsim_dev_new()
Yang Yingliang [Tue, 18 Oct 2022 13:16:07 +0000 (21:16 +0800)]
wwan_hwsim: fix possible memory leak in wwan_hwsim_dev_new()

Inject fault while probing module, if device_register() fails,
but the refcount of kobject is not decreased to 0, the name
allocated in dev_set_name() is leaked. Fix this by calling
put_device(), so that name can be freed in callback function
kobject_cleanup().

unreferenced object 0xffff88810152ad20 (size 8):
  comm "modprobe", pid 252, jiffies 4294849206 (age 22.713s)
  hex dump (first 8 bytes):
    68 77 73 69 6d 30 00 ff                          hwsim0..
  backtrace:
    [<000000009c3504ed>] __kmalloc_node_track_caller+0x44/0x1b0
    [<00000000c0228a5e>] kvasprintf+0xb5/0x140
    [<00000000cff8c21f>] kvasprintf_const+0x55/0x180
    [<0000000055a1e073>] kobject_set_name_vargs+0x56/0x150
    [<000000000a80b139>] dev_set_name+0xab/0xe0

Fixes: f36a111a74e7 ("wwan_hwsim: WWAN device simulator")
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Reviewed-by: Loic Poulain <loic.poulain@linaro.org>
Acked-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Link: https://lore.kernel.org/r/20221018131607.1901641-1-yangyingliang@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
20 months agosfc: include vport_id in filter spec hash and equal()
Pieter Jansen van Vuuren [Tue, 18 Oct 2022 09:28:41 +0000 (10:28 +0100)]
sfc: include vport_id in filter spec hash and equal()

Filters on different vports are qualified by different implicit MACs and/or
VLANs, so shouldn't be considered equal even if their other match fields
are identical.

Fixes: 7c460d9be610 ("sfc: Extend and abstract efx_filter_spec to cover Huntington/EF10")
Co-developed-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Pieter Jansen van Vuuren <pieter.jansen-van-vuuren@amd.com>
Reviewed-by: Martin Habets <habetsm.xilinx@gmail.com>
Link: https://lore.kernel.org/r/20221018092841.32206-1-pieter.jansen-van-vuuren@amd.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
20 months agoopenvswitch: Use kmalloc_size_roundup() to match ksize() usage
Kees Cook [Tue, 18 Oct 2022 09:06:33 +0000 (02:06 -0700)]
openvswitch: Use kmalloc_size_roundup() to match ksize() usage

Round up allocations with kmalloc_size_roundup() so that openvswitch's
use of ksize() is always accurate and no special handling of the memory
is needed by KASAN, UBSAN_BOUNDS, nor FORTIFY_SOURCE.

Cc: Pravin B Shelar <pshelar@ovn.org>
Cc: dev@openvswitch.org
Signed-off-by: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20221018090628.never.537-kees@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
20 months agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf
Jakub Kicinski [Wed, 19 Oct 2022 22:45:54 +0000 (15:45 -0700)]
Merge git://git./linux/kernel/git/netfilter/nf

Pablo Neira Ayuso says:

====================
Netfilter fixes for net

1) Missing flowi uid field in nft_fib expression, from Guillaume Nault.
   This is broken since the creation of the fib expression.

2) Relax sanity check to fix bogus EINVAL error when deleting elements
   belonging set intervals. Broken since 6.0-rc.

* git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf:
  netfilter: nf_tables: relax NFTA_SET_ELEM_KEY_END set flags requirements
  netfilter: rpfilter/fib: Set ->flowic_uid correctly for user namespaces.
====================

Link: https://lore.kernel.org/r/20221019065225.1006344-1-pablo@netfilter.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
20 months agogenetlink: fix kdoc warnings
Jakub Kicinski [Tue, 18 Oct 2022 23:13:10 +0000 (16:13 -0700)]
genetlink: fix kdoc warnings

Address a bunch of kdoc warnings:

include/net/genetlink.h:81: warning: Function parameter or member 'module' not described in 'genl_family'
include/net/genetlink.h:243: warning: expecting prototype for struct genl_info. Prototype was for struct genl_dumpit_info instead
include/net/genetlink.h:419: warning: Function parameter or member 'net' not described in 'genlmsg_unicast'
include/net/genetlink.h:438: warning: expecting prototype for gennlmsg_data(). Prototype was for genlmsg_data() instead
include/net/genetlink.h:244: warning: Function parameter or member 'op' not described in 'genl_dumpit_info'

Link: https://lore.kernel.org/r/20221018231310.1040482-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
20 months agoMerge branch 'netlink-formatted-extacks'
Jakub Kicinski [Wed, 19 Oct 2022 20:00:09 +0000 (13:00 -0700)]
Merge branch 'netlink-formatted-extacks'

Edward Cree says:

====================
netlink: formatted extacks

Currently, netlink extacks can only carry fixed string messages, which
 is limiting when reporting failures in complex systems.  This series
 adds the ability to return printf-formatted messages, and uses it in
 the sfc driver's TC offload code.
Formatted extack messages are limited in length to a fixed buffer size,
 currently 80 characters.  If the message exceeds this, the full message
 will be logged (ratelimited) to the console and a truncated version
 returned over netlink.
There is no change to the netlink uAPI; only internal kernel changes
 are needed.
====================

Link: https://lore.kernel.org/r/cover.1666102698.git.ecree.xilinx@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
20 months agosfc: remove 'log-tc-errors' ethtool private flag
Edward Cree [Tue, 18 Oct 2022 14:37:29 +0000 (15:37 +0100)]
sfc: remove 'log-tc-errors' ethtool private flag

It no longer does anything now that we're using formatted extacks instead.
So we can remove the driver's whole get/set priv_flags implementation.

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
20 months agosfc: use formatted extacks instead of efx_tc_err()
Edward Cree [Tue, 18 Oct 2022 14:37:28 +0000 (15:37 +0100)]
sfc: use formatted extacks instead of efx_tc_err()

Since we can now get a formatted message back to the user with
 NL_SET_ERR_MSG_FMT_MOD(), there's no need for our special logging.

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
20 months agonetlink: add support for formatted extack messages
Edward Cree [Tue, 18 Oct 2022 14:37:27 +0000 (15:37 +0100)]
netlink: add support for formatted extack messages

Include an 80-byte buffer in struct netlink_ext_ack that can be used
 for scnprintf()ed messages.  This does mean that the resulting string
 can't be enumerated, translated etc. in the way NL_SET_ERR_MSG() was
 designed to allow.

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
20 months agonet: ethernet: adi: adin1110: Fix SPI transfers
Alexandru Tachici [Mon, 17 Oct 2022 16:37:03 +0000 (19:37 +0300)]
net: ethernet: adi: adin1110: Fix SPI transfers

No need to use more than one SPI transfer for reads.
Use only one from now as ADIN1110/2111 does not tolerate
CS changes during reads.

The BCM2711/2708 SPI controllers worked fine, but the NXP
IMX8MM could not keep CS lowered during SPI bursts.

This change aims to make the ADIN1110/2111 driver compatible
with both SPI controllers, without any loss of bandwidth/other
capabilities.

Fixes: bc93e19d088b ("net: ethernet: adi: Add ADIN1110 support")
Signed-off-by: Alexandru Tachici <alexandru.tachici@analog.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agoMerge branch 'qdisc-ingress-success'
David S. Miller [Wed, 19 Oct 2022 13:04:36 +0000 (14:04 +0100)]
Merge branch 'qdisc-ingress-success'

Paul Blakey says:

====================
net: Fix return value of qdisc ingress handling on success

Fix patch + self-test with the currently broken scenario.

v4->v3:
  Removed new line in self test and rebase (Paolo).

v2->v3:
  Added DROP return to TC_ACT_SHOT case (Cong).

v1->v2:
  Changed blamed commit
  Added self-test
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agoselftests: add selftest for chaining of tc ingress handling to egress
Paul Blakey [Tue, 18 Oct 2022 07:34:39 +0000 (10:34 +0300)]
selftests: add selftest for chaining of tc ingress handling to egress

This test runs a simple ingress tc setup between two veth pairs,
then adds a egress->ingress rule to test the chaining of tc ingress
pipeline to tc egress piepline.

Signed-off-by: Paul Blakey <paulb@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: Fix return value of qdisc ingress handling on success
Paul Blakey [Tue, 18 Oct 2022 07:34:38 +0000 (10:34 +0300)]
net: Fix return value of qdisc ingress handling on success

Currently qdisc ingress handling (sch_handle_ingress()) doesn't
set a return value and it is left to the old return value of
the caller (__netif_receive_skb_core()) which is RX drop, so if
the packet is consumed, caller will stop and return this value
as if the packet was dropped.

This causes a problem in the kernel tcp stack when having a
egress tc rule forwarding to a ingress tc rule.
The tcp stack sending packets on the device having the egress rule
will see the packets as not successfully transmitted (although they
actually were), will not advance it's internal state of sent data,
and packets returning on such tcp stream will be dropped by the tcp
stack with reason ack-of-unsent-data. See reproduction in [0] below.

Fix that by setting the return value to RX success if
the packet was handled successfully.

[0] Reproduction steps:
 $ ip link add veth1 type veth peer name peer1
 $ ip link add veth2 type veth peer name peer2
 $ ifconfig peer1 5.5.5.6/24 up
 $ ip netns add ns0
 $ ip link set dev peer2 netns ns0
 $ ip netns exec ns0 ifconfig peer2 5.5.5.5/24 up
 $ ifconfig veth2 0 up
 $ ifconfig veth1 0 up

 #ingress forwarding veth1 <-> veth2
 $ tc qdisc add dev veth2 ingress
 $ tc qdisc add dev veth1 ingress
 $ tc filter add dev veth2 ingress prio 1 proto all flower \
   action mirred egress redirect dev veth1
 $ tc filter add dev veth1 ingress prio 1 proto all flower \
   action mirred egress redirect dev veth2

 #steal packet from peer1 egress to veth2 ingress, bypassing the veth pipe
 $ tc qdisc add dev peer1 clsact
 $ tc filter add dev peer1 egress prio 20 proto ip flower \
   action mirred ingress redirect dev veth1

 #run iperf and see connection not running
 $ iperf3 -s&
 $ ip netns exec ns0 iperf3 -c 5.5.5.6 -i 1

 #delete egress rule, and run again, now should work
 $ tc filter del dev peer1 egress
 $ ip netns exec ns0 iperf3 -c 5.5.5.6 -i 1

Fixes: f697c3e8b35c ("[NET]: Avoid unnecessary cloning for ingress filtering")
Signed-off-by: Paul Blakey <paulb@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agoMerge branch 'net-bridge-mc-cleanups'
David S. Miller [Wed, 19 Oct 2022 13:01:08 +0000 (14:01 +0100)]
Merge branch 'net-bridge-mc-cleanups'

Ido Schimmel says:

====================
bridge: A few multicast cleanups

Clean up a few issues spotted while working on the bridge multicast code
and running its selftests.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agobridge: mcast: Simplify MDB entry creation
Ido Schimmel [Tue, 18 Oct 2022 06:40:01 +0000 (09:40 +0300)]
bridge: mcast: Simplify MDB entry creation

Before creating a new MDB entry, br_multicast_new_group() will call
br_mdb_ip_get() to see if one exists and return it if so.

Therefore, simply call br_multicast_new_group() and omit the call to
br_mdb_ip_get().

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Acked-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agobridge: mcast: Use spin_lock() instead of spin_lock_bh()
Ido Schimmel [Tue, 18 Oct 2022 06:40:00 +0000 (09:40 +0300)]
bridge: mcast: Use spin_lock() instead of spin_lock_bh()

IGMPv3 / MLDv2 Membership Reports are only processed from the data path
with softIRQ disabled, so there is no need to call spin_lock_bh(). Use
spin_lock() instead.

This is consistent with how other IGMP / MLD packets are processed.

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Acked-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agoselftests: bridge_igmp: Remove unnecessary address deletion
Ido Schimmel [Tue, 18 Oct 2022 06:39:59 +0000 (09:39 +0300)]
selftests: bridge_igmp: Remove unnecessary address deletion

The test group address is added and removed in v2reportleave_test().
There is no need to delete it again during cleanup as it results in the
following error message:

 # bash -x ./bridge_igmp.sh
 [...]
 + cleanup
 + pre_cleanup
 [...]
 + ip address del dev swp4 239.10.10.10/32
 RTNETLINK answers: Cannot assign requested address
 + h2_destroy

Solve by removing the unnecessary address deletion.

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Acked-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agoselftests: bridge_vlan_mcast: Delete qdiscs during cleanup
Ido Schimmel [Tue, 18 Oct 2022 06:39:58 +0000 (09:39 +0300)]
selftests: bridge_vlan_mcast: Delete qdiscs during cleanup

The qdiscs are added during setup, but not deleted during cleanup,
resulting in the following error messages:

 # ./bridge_vlan_mcast.sh
 [...]
 # ./bridge_vlan_mcast.sh
 Error: Exclusivity flag on, cannot modify.
 Error: Exclusivity flag on, cannot modify.

Solve by deleting the qdiscs during cleanup.

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Acked-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agoMerge branch 'qdisc-null-deref'
David S. Miller [Wed, 19 Oct 2022 12:47:09 +0000 (13:47 +0100)]
Merge branch 'qdisc-null-deref'

Zhengchao Shao says:

====================
net: fix null pointer access issue in qdisc

These three patches fix the same type of problem. Set the default qdisc,
and then construct an init failure scenario when the dev qdisc is
configured on mqprio to trigger the reset process. NULL pointer access
may occur during the reset process.

---
v2: for fq_codel, revert the patch
---
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: sched: sfb: fix null pointer access issue when sfb_init() fails
Zhengchao Shao [Tue, 18 Oct 2022 06:32:01 +0000 (14:32 +0800)]
net: sched: sfb: fix null pointer access issue when sfb_init() fails

When the default qdisc is sfb, if the qdisc of dev_queue fails to be
inited during mqprio_init(), sfb_reset() is invoked to clear resources.
In this case, the q->qdisc is NULL, and it will cause gpf issue.

The process is as follows:
qdisc_create_dflt()
sfb_init()
tcf_block_get()          --->failed, q->qdisc is NULL
...
qdisc_put()
...
sfb_reset()
qdisc_reset(q->qdisc)    --->q->qdisc is NULL
ops = qdisc->ops

The following is the Call Trace information:
general protection fault, probably for non-canonical address
0xdffffc0000000003: 0000 [#1] PREEMPT SMP KASAN
KASAN: null-ptr-deref in range [0x0000000000000018-0x000000000000001f]
RIP: 0010:qdisc_reset+0x2b/0x6f0
Call Trace:
<TASK>
sfb_reset+0x37/0xd0
qdisc_reset+0xed/0x6f0
qdisc_destroy+0x82/0x4c0
qdisc_put+0x9e/0xb0
qdisc_create_dflt+0x2c3/0x4a0
mqprio_init+0xa71/0x1760
qdisc_create+0x3eb/0x1000
tc_modify_qdisc+0x408/0x1720
rtnetlink_rcv_msg+0x38e/0xac0
netlink_rcv_skb+0x12d/0x3a0
netlink_unicast+0x4a2/0x740
netlink_sendmsg+0x826/0xcc0
sock_sendmsg+0xc5/0x100
____sys_sendmsg+0x583/0x690
___sys_sendmsg+0xe8/0x160
__sys_sendmsg+0xbf/0x160
do_syscall_64+0x35/0x80
entry_SYSCALL_64_after_hwframe+0x46/0xb0
RIP: 0033:0x7f2164122d04
</TASK>

Fixes: e13e02a3c68d ("net_sched: SFB flow scheduler")
Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agoRevert "net: sched: fq_codel: remove redundant resource cleanup in fq_codel_init()"
Zhengchao Shao [Tue, 18 Oct 2022 06:32:00 +0000 (14:32 +0800)]
Revert "net: sched: fq_codel: remove redundant resource cleanup in fq_codel_init()"

This reverts commit 494f5063b86cd6e972cb41a27e083c9a3664319d.

When the default qdisc is fq_codel, if the qdisc of dev_queue fails to be
inited during mqprio_init(), fq_codel_reset() is invoked to clear
resources. In this case, the flow is NULL, and it will cause gpf issue.

The process is as follows:
qdisc_create_dflt()
fq_codel_init()
...
q->flows_cnt = 1024;
...
q->flows = kvcalloc(...)      --->failed, q->flows is NULL
...
qdisc_put()
...
fq_codel_reset()
...
flow = q->flows + i   --->q->flows is NULL

The following is the Call Trace information:
general protection fault, probably for non-canonical address
0xdffffc0000000001: 0000 [#1] PREEMPT SMP KASAN
KASAN: null-ptr-deref in range [0x0000000000000008-0x000000000000000f]
RIP: 0010:fq_codel_reset+0x14d/0x350
Call Trace:
<TASK>
qdisc_reset+0xed/0x6f0
qdisc_destroy+0x82/0x4c0
qdisc_put+0x9e/0xb0
qdisc_create_dflt+0x2c3/0x4a0
mqprio_init+0xa71/0x1760
qdisc_create+0x3eb/0x1000
tc_modify_qdisc+0x408/0x1720
rtnetlink_rcv_msg+0x38e/0xac0
netlink_rcv_skb+0x12d/0x3a0
netlink_unicast+0x4a2/0x740
netlink_sendmsg+0x826/0xcc0
sock_sendmsg+0xc5/0x100
____sys_sendmsg+0x583/0x690
___sys_sendmsg+0xe8/0x160
__sys_sendmsg+0xbf/0x160
do_syscall_64+0x35/0x80
entry_SYSCALL_64_after_hwframe+0x46/0xb0
RIP: 0033:0x7fd272b22d04
</TASK>

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: sched: cake: fix null pointer access issue when cake_init() fails
Zhengchao Shao [Tue, 18 Oct 2022 06:31:59 +0000 (14:31 +0800)]
net: sched: cake: fix null pointer access issue when cake_init() fails

When the default qdisc is cake, if the qdisc of dev_queue fails to be
inited during mqprio_init(), cake_reset() is invoked to clear
resources. In this case, the tins is NULL, and it will cause gpf issue.

The process is as follows:
qdisc_create_dflt()
cake_init()
q->tins = kvcalloc(...)        --->failed, q->tins is NULL
...
qdisc_put()
...
cake_reset()
...
cake_dequeue_one()
b = &q->tins[...]   --->q->tins is NULL

The following is the Call Trace information:
general protection fault, probably for non-canonical address
0xdffffc0000000000: 0000 [#1] PREEMPT SMP KASAN
KASAN: null-ptr-deref in range [0x0000000000000000-0x0000000000000007]
RIP: 0010:cake_dequeue_one+0xc9/0x3c0
Call Trace:
<TASK>
cake_reset+0xb1/0x140
qdisc_reset+0xed/0x6f0
qdisc_destroy+0x82/0x4c0
qdisc_put+0x9e/0xb0
qdisc_create_dflt+0x2c3/0x4a0
mqprio_init+0xa71/0x1760
qdisc_create+0x3eb/0x1000
tc_modify_qdisc+0x408/0x1720
rtnetlink_rcv_msg+0x38e/0xac0
netlink_rcv_skb+0x12d/0x3a0
netlink_unicast+0x4a2/0x740
netlink_sendmsg+0x826/0xcc0
sock_sendmsg+0xc5/0x100
____sys_sendmsg+0x583/0x690
___sys_sendmsg+0xe8/0x160
__sys_sendmsg+0xbf/0x160
do_syscall_64+0x35/0x80
entry_SYSCALL_64_after_hwframe+0x46/0xb0
RIP: 0033:0x7f89e5122d04
</TASK>

Fixes: 046f6fd5daef ("sched: Add Common Applications Kept Enhanced (cake) qdisc")
Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Acked-by: Toke Høiland-Jørgensen <toke@toke.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agoethernet: marvell: octeontx2 Fix resource not freed after malloc
Manank Patel [Tue, 18 Oct 2022 05:33:18 +0000 (11:03 +0530)]
ethernet: marvell: octeontx2 Fix resource not freed after malloc

fix rxsc and txsc not getting freed before going out of scope

Fixes: c54ffc73601c ("octeontx2-pf: mcs: Introduce MACSEC hardware offloading")
Signed-off-by: Manank Patel <pmanank200502@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agoMerge branch 'dpaa-phylink'
David S. Miller [Wed, 19 Oct 2022 12:25:09 +0000 (13:25 +0100)]
Merge branch 'dpaa-phylink'

Sean Anderson says:

====================
net: dpaa: Convert to phylink

This series converts the DPAA driver to phylink.

I have tried to maintain backwards compatibility with existing device
trees whereever possible. However, one area where I was unable to
achieve this was with QSGMII. Please refer to patch 2 for details.

All mac drivers have now been converted. I would greatly appreciate if
anyone has T-series or P-series boards they can test/debug this series
on. I only have an LS1046ARDB. Everything but QSGMII should work without
breakage; QSGMII needs patches 7 and 8. For this reason, the last 4
patches in this series should be applied together (and should not go
through separate trees).

Changes in v7:
- provide phylink_validate_mask_caps() helper
- Fix oops if memac_pcs_create returned -EPROBE_DEFER
- Fix using pcs-names instead of pcs-handle-names
- Fix not checking for -ENODATA when looking for sgmii pcs
- Fix 81-character line
- Simplify memac_validate with phylink_validate_mask_caps

Changes in v6:
- Remove unnecessary $ref from renesas,rzn1-a5psw
- Remove unnecessary type from pcs-handle-names
- Add maxItems to pcs-handle
- Fix 81-character line
- Fix uninitialized variable in dtsec_mac_config

Changes in v5:
- Add Lynx PCS binding

Changes in v4:
- Use pcs-handle-names instead of pcs-names, as discussed
- Don't fail if phy support was not compiled in
- Split off rate adaptation series
- Split off DPAA "preparation" series
- Split off Lynx 10G support
- t208x: Mark MAC1 and MAC2 as 10G
- Add XFI PCS for t208x MAC1/MAC2

Changes in v3:
- Expand pcs-handle to an array
- Add vendor prefix 'fsl,' to rgmii and mii properties.
- Set maxItems for pcs-names
- Remove phy-* properties from example because dt-schema complains and I
  can't be bothered to figure out how to make it work.
- Add pcs-handle as a preferred version of pcsphy-handle
- Deprecate pcsphy-handle
- Remove mii/rmii properties
- Put the PCS mdiodev only after we are done with it (since the PCS
  does not perform a get itself).
- Remove _return label from memac_initialization in favor of returning
  directly
- Fix grabbing the default PCS not checking for -ENODATA from
  of_property_match_string
- Set DTSEC_ECNTRL_R100M in dtsec_link_up instead of dtsec_mac_config
- Remove rmii/mii properties
- Replace 1000Base... with 1000BASE... to match IEEE capitalization
- Add compatibles for QSGMII PCSs
- Split arm and powerpcs dts updates

Changes in v2:
- Better document how we select which PCS to use in the default case
- Move PCS_LYNX dependency to fman Kconfig
- Remove unused variable slow_10g_if
- Restrict valid link modes based on the phy interface. This is easier
  to set up, and mostly captures what I intended to do the first time.
  We now have a custom validate which restricts half-duplex for some SoCs
  for RGMII, but generally just uses the default phylink validate.
- Configure the SerDes in enable/disable
- Properly implement all ethtool ops and ioctls. These were mostly
  stubbed out just enough to compile last time.
- Convert 10GEC and dTSEC as well
- Fix capitalization of mEMAC in commit messages
- Add nodes for QSGMII PCSs
- Add nodes for QSGMII PCSs
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agoarm64: dts: layerscape: Add nodes for QSGMII PCSs
Sean Anderson [Mon, 17 Oct 2022 20:22:41 +0000 (16:22 -0400)]
arm64: dts: layerscape: Add nodes for QSGMII PCSs

Now that we actually read registers from QSGMII PCSs, it's important
that we have the correct address (instead of hoping that we're the MAC
with all the QSGMII PCSs on its bus). This adds nodes for the QSGMII
PCSs.  The exact mapping of QSGMII to MACs depends on the SoC.

Since the first QSGMII PCSs share an address with the SGMII and XFI
PCSs, we only add new nodes for PCSs 2-4. This avoids address conflicts
on the bus.

Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agopowerpc: dts: qoriq: Add nodes for QSGMII PCSs
Sean Anderson [Mon, 17 Oct 2022 20:22:40 +0000 (16:22 -0400)]
powerpc: dts: qoriq: Add nodes for QSGMII PCSs

Now that we actually read registers from QSGMII PCSs, it's important
that we have the correct address (instead of hoping that we're the MAC
with all the QSGMII PCSs on its bus). This adds nodes for the QSGMII
PCSs. They have the same addresses on all SoCs (e.g. if QSGMIIA is
present it's used for MACs 1 through 4).

Since the first QSGMII PCSs share an address with the SGMII and XFI
PCSs, we only add new nodes for PCSs 2-4. This avoids address conflicts
on the bus.

Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agopowerpc: dts: t208x: Mark MAC1 and MAC2 as 10G
Sean Anderson [Mon, 17 Oct 2022 20:22:39 +0000 (16:22 -0400)]
powerpc: dts: t208x: Mark MAC1 and MAC2 as 10G

On the T208X SoCs, MAC1 and MAC2 support XGMII. Add some new MAC dtsi
fragments, and mark the QMAN ports as 10G.

Fixes: da414bb923d9 ("powerpc/mpc85xx: Add FSL QorIQ DPAA FMan support to the SoC device tree(s)")
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: dpaa: Convert to phylink
Sean Anderson [Mon, 17 Oct 2022 20:22:38 +0000 (16:22 -0400)]
net: dpaa: Convert to phylink

This converts DPAA to phylink. All macs are converted. This should work
with no device tree modifications (including those made in this series),
except for QSGMII (as noted previously).

The mEMAC configuration is one of the tricker areas. I have tried to
capture all the restrictions across the various models. Most of the time,
we assume that if the serdes supports a mode or the phy-interface-mode
specifies it, then we support it. The only place we can't do this is
(RG)MII, since there's no serdes. In that case, we rely on a (new)
devicetree property. There are also several cases where half-duplex is
broken. Unfortunately, only a single compatible is used for the MAC, so we
have to use the board compatible instead.

The 10GEC conversion is very straightforward, since it only supports XAUI.
There is generally nothing to configure.

The dTSEC conversion is broadly similar to mEMAC, but is simpler because we
don't support configuring the SerDes (though this can be easily added) and
we don't have multiple PCSs. From what I can tell, there's nothing
different in the driver or documentation between SGMII and 1000BASE-X
except for the advertising. Similarly, I couldn't find anything about
2500BASE-X. In both cases, I treat them like SGMII. These modes aren't used
by any in-tree boards. Similarly, despite being mentioned in the driver, I
couldn't find any documented SoCs which supported QSGMII.  I have left it
unimplemented for now.

Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: fman: memac: Use lynx pcs driver
Sean Anderson [Mon, 17 Oct 2022 20:22:37 +0000 (16:22 -0400)]
net: fman: memac: Use lynx pcs driver

Although not stated in the datasheet, as far as I can tell PCS for mEMACs
is a "Lynx." By reusing the existing driver, we can remove the PCS
management code from the memac driver. This requires calling some PCS
functions manually which phylink would usually do for us, but we will let
it do that soon.

One problem is that we don't actually have a PCS for QSGMII. We pretend
that each mEMAC's MDIO bus has four QSGMII PCSs, but this is not the case.
Only the "base" mEMAC's MDIO bus has the four QSGMII PCSs. This is not an
issue yet, because we never get the PCS state. However, it will be once the
conversion to phylink is complete, since the links will appear to never
come up. To get around this, we allow specifying multiple PCSs in pcsphy.
This breaks backwards compatibility with old device trees, but only for
QSGMII. IMO this is the only reasonable way to figure out what the actual
QSGMII PCS is.

Additionally, we now also support a separate XFI PCS. This can allow the
SerDes driver to set different addresses for the SGMII and XFI PCSs so they
can be accessed at the same time.

Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: fman: memac: Add serdes support
Sean Anderson [Mon, 17 Oct 2022 20:22:36 +0000 (16:22 -0400)]
net: fman: memac: Add serdes support

This adds support for using a serdes which has to be configured. This is
primarly in preparation for phylink conversion, which will then change the
serdes mode dynamically.

Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonet: phylink: provide phylink_validate_mask_caps() helper
Russell King (Oracle) [Mon, 17 Oct 2022 20:22:35 +0000 (16:22 -0400)]
net: phylink: provide phylink_validate_mask_caps() helper

Provide a helper that restricts the link modes according to the
phylink capabilities.

Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
[rebased on net-next/master and added documentation]
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agodt-bindings: net: fman: Add additional interface properties
Sean Anderson [Mon, 17 Oct 2022 20:22:34 +0000 (16:22 -0400)]
dt-bindings: net: fman: Add additional interface properties

At the moment, mEMACs are configured almost completely based on the
phy-connection-type. That is, if the phy interface is RGMII, it assumed
that RGMII is supported. For some interfaces, it is assumed that the
RCW/bootloader has set up the SerDes properly. This is generally OK, but
restricts runtime reconfiguration. The actual link state is never
reported.

To address these shortcomings, the driver will need additional
information. First, it needs to know how to access the PCS/PMAs (in
order to configure them and get the link status). The SGMII PCS/PMA is
the only currently-described PCS/PMA. Add the XFI and QSGMII PCS/PMAs as
well. The XFI (and 10GBASE-KR) PCS/PMA is a c45 "phy" which sits on the
same MDIO bus as SGMII PCS/PMA. By default they will have conflicting
addresses, but they are also not enabled at the same time by default.
Therefore, we can let the XFI PCS/PMA be the default when
phy-connection-type is xgmii. This will allow for
backwards-compatibility.

QSGMII, however, cannot work with the current binding. This is because
the QSGMII PCS/PMAs are only present on one MAC's MDIO bus. At the
moment this is worked around by having every MAC write to the PCS/PMA
addresses (without checking if they are present). This only works if
each MAC has the same configuration, and only if we don't need to know
the status. Because the QSGMII PCS/PMA will typically be located on a
different MDIO bus than the MAC's SGMII PCS/PMA, there is no fallback
for the QSGMII PCS/PMA.

Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agodt-bindings: net: Add Lynx PCS binding
Sean Anderson [Mon, 17 Oct 2022 20:22:33 +0000 (16:22 -0400)]
dt-bindings: net: Add Lynx PCS binding

This binding is fairly bare-bones for now, since the Lynx driver doesn't
parse any properties (or match based on the compatible). We just need it
in order to prevent the PCS nodes from having phy devices attached to
them. This is not really a problem, but it is a bit inefficient.

This binding is really for three separate PCSs (SGMII, QSGMII, and XFI).
However, the driver treats all of them the same. This works because the
SGMII and XFI devices typically use the same address, and the SerDes
driver (or RCW) muxes between them. The QSGMII PCSs have the same
register layout as the SGMII PCSs. To do things properly, we'd probably
do something like

ethernet-pcs@0 {
#pcs-cells = <1>;
compatible = "fsl,lynx-pcs";
reg = <0>, <1>, <2>, <3>;
};

but that would add complexity, and we can describe the hardware just
fine using separate PCSs for now.

Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agodt-bindings: net: Expand pcs-handle to an array
Sean Anderson [Mon, 17 Oct 2022 20:22:32 +0000 (16:22 -0400)]
dt-bindings: net: Expand pcs-handle to an array

This allows multiple phandles to be specified for pcs-handle, such as
when multiple PCSs are present for a single MAC. To differentiate
between them, also add a pcs-handle-names property.

Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agoMerge branch 'net-marvell-yaml'
David S. Miller [Wed, 19 Oct 2022 08:49:38 +0000 (09:49 +0100)]
Merge branch 'net-marvell-yaml'

Michał Grzelak says:

====================
net: further improvements to marvell,pp2.yaml

This patchset addresses problems with reg ranges and
additional $refs. It also limits phy-mode and aligns examples.

Best regards,
Michał

---
Changelog:
v4->v5
- drop '+' from all patternProperties
- restrict range of patternProperties to [0-2] in top level
- drop the $ref in patternProperties:'^...':properties:reg
- add patternProperties:'^...':properties:reg:maximum:2
- drop $ref in patternProperties:'^...':properties:phys
- add patternProperties:'^...':properties:phys:maxItems:1
- limit phy-mode to the subset found in dts files
- reflect the order of subnodes' properties in subnodes' required:
- restrict range of pattern to [0-2] in marvell,armada-7k-pp22 case
- restrict range of pattern to [0-1] in marvell,armada-375-pp2 case
- align to 4 spaces all examples:
- add specified maximum to allOf:if:then-else:properties:reg

v3->v4
- change commit message of first patch
- move allOf:$ref to patternProperties:'^...':$ref
- deprecate port-id in favour of reg
- move reg to front of properties list in patternProperties
- reflect the order of properties in required list in
  patternProperties
- add unevaluatedProperties: false to patternProperties
- change unevaluated- to additionalProperties at top level
- add property phys: to ports subnode
- extend example binding with additional information about phys and sfp
- hook phys property to phy-consumer.yaml schema

v2->v3
- move 'reg:description' to 'allOf:if:then'
- change '#size-cells: true' and '#address-cells: true'
  to '#size-cells: const: 0' and '#address-cells: const: 1'
- replace all occurences of pattern "^eth\{hex_num}*"
  with "^(ethernet-)?port@[0-9]+$"
- add description in 'patternProperties:^...'
- add 'patternProperties:^...:interrupt-names:minItems: 1'
- add 'patternProperties:^...:reg:description'
- update 'patternProperties:^...:port-id:description'
- add 'patternProperties:^...:required: - reg'
- update '*:description:' to uppercase
- add 'allOf:then:required:marvell,system-controller'
- skip quotation marks from 'allOf:$ref'
- add 'else' schema to match 'allOf:if:then'
- restrict 'clocks' in 'allOf:if:then'
- restrict 'clock-names' in 'allOf:if:then'
- add #address-cells=<1>; #size-cells=<0>; in 'examples:'
- change every "ethX" to "ethernet-port@X" in 'examples:'
- add "reg" and comment in all ports in 'examples:'
- change /ethernet/eth0/phy-mode in examples://Armada-375
  to "rgmii-id"
- replace each cpm_ with cp0_ in 'examples:'
- replace each _syscon0 with _clk0 in 'examples:'
- remove each eth0X label in 'examples:'
- update armada-375.dtsi and armada-cp11x.dtsi to match
  marvell,pp2.yaml

v1->v2
- move 'properties' to the front of the file
- remove blank line after 'properties'
- move 'compatible' to the front of 'properties'
- move 'clocks', 'clock-names' and 'reg' definitions to 'properties'
- substitute all occurences of 'marvell,armada-7k-pp2' with
  'marvell,armada-7k-pp22'
- add properties:#size-cells and properties:#address-cells
- specify list in 'interrupt-names'
- remove blank lines after 'patternProperties'
- remove '^interrupt' and '^#.*-cells$' patterns
- remove blank line after 'allOf'
- remove first 'if-then-else' block from 'allOf'
- negate the condition in allOf:if schema
- delete 'interrupt-controller' from section 'examples'
- delete '#interrupt-cells' from section 'examples'
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agoARM: dts: armada-375: Update network description to match schema
Marcin Wojtas [Fri, 14 Oct 2022 21:32:54 +0000 (23:32 +0200)]
ARM: dts: armada-375: Update network description to match schema

Update the PP2 ethernet ports subnodes' names to match
schema enforced by the marvell,pp2.yaml contents.

Add new required properties ('reg') which contains information
about the port ID, keeping 'port-id' ones for backward
compatibility.

Signed-off-by: Marcin Wojtas <mw@semihalf.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agoarm64: dts: marvell: Update network description to match schema
Marcin Wojtas [Fri, 14 Oct 2022 21:32:53 +0000 (23:32 +0200)]
arm64: dts: marvell: Update network description to match schema

Update the PP2 ethernet ports subnodes' names to match
schema enforced by the marvell,pp2.yaml contents.

Add new required properties ('reg') which contains information
about the port ID, keeping 'port-id' ones for backward
compatibility.

Signed-off-by: Marcin Wojtas <mw@semihalf.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agodt-bindings: net: marvell,pp2: convert to json-schema
Michał Grzelak [Fri, 14 Oct 2022 21:32:52 +0000 (23:32 +0200)]
dt-bindings: net: marvell,pp2: convert to json-schema

Convert the marvell,pp2 bindings from text to proper schema.

Move 'marvell,system-controller' and 'dma-coherent' properties from
port up to the controller node, to match what is actually done in DT.

Rename all subnodes to match "^(ethernet-)?port@[0-2]$" and deprecate
port-id in favour of 'reg'.

Signed-off-by: Michał Grzelak <mig@semihalf.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
20 months agonetfilter: nf_tables: relax NFTA_SET_ELEM_KEY_END set flags requirements
Pablo Neira Ayuso [Mon, 17 Oct 2022 12:12:58 +0000 (14:12 +0200)]
netfilter: nf_tables: relax NFTA_SET_ELEM_KEY_END set flags requirements

Otherwise EINVAL is bogusly reported to userspace when deleting a set
element. NFTA_SET_ELEM_KEY_END does not need to be set in case of:

- insertion: if not present, start key is used as end key.
- deletion: only start key needs to be specified, end key is ignored.

Hence, relax the sanity check.

Fixes: 88cccd908d51 ("netfilter: nf_tables: NFTA_SET_ELEM_KEY_END requires concat and interval flags")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
20 months agonetfilter: rpfilter/fib: Set ->flowic_uid correctly for user namespaces.
Guillaume Nault [Thu, 13 Oct 2022 14:37:47 +0000 (16:37 +0200)]
netfilter: rpfilter/fib: Set ->flowic_uid correctly for user namespaces.

Currently netfilter's rpfilter and fib modules implicitely initialise
->flowic_uid with 0. This is normally the root UID. However, this isn't
the case in user namespaces, where user ID 0 is mapped to a different
kernel UID. By initialising ->flowic_uid with sock_net_uid(), we get
the root UID of the user namespace, thus keeping the same behaviour
whether or not we're running in a user namepspace.

Note, this is similar to commit 8bcfd0925ef1 ("ipv4: add missing
initialization for flowi4_uid"), which fixed the rp_filter sysctl.

Fixes: 622ec2c9d524 ("net: core: add UID to flows, rules, and routes")
Signed-off-by: Guillaume Nault <gnault@redhat.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
20 months agoenic: define constants for legacy interrupts offset
Govindarajulu Varadarajan [Tue, 18 Oct 2022 00:58:04 +0000 (17:58 -0700)]
enic: define constants for legacy interrupts offset

Use macro instead of function calls. These values are constant and will
not change.

Signed-off-by: Govindarajulu Varadarajan <govind.varadar@gmail.com>
Link: https://lore.kernel.org/r/20221018005804.188643-1-govind.varadar@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
20 months agonet: fec: remove the unused functions
Shenwei Wang [Mon, 17 Oct 2022 16:12:36 +0000 (11:12 -0500)]
net: fec: remove the unused functions

Removed those unused functions since we simplified the driver
by using the page pool to manage RX buffers.

Signed-off-by: Shenwei Wang <shenwei.wang@nxp.com>
Link: https://lore.kernel.org/r/20221017161236.1563975-1-shenwei.wang@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
20 months agonet: remove smc911x driver
Arnd Bergmann [Mon, 17 Oct 2022 12:18:26 +0000 (14:18 +0200)]
net: remove smc911x driver

This driver was used on Arm and SH machines until 2009, when the
last platforms moved to the smsc911x driver for the same hardware.

Time to retire this version.

Link: https://lore.kernel.org/netdev/1232010482-3744-1-git-send-email-steve.glendinning@smsc.com/
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Link: https://lore.kernel.org/r/20221017121900.3520108-1-arnd@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
20 months agoionic: catch NULL pointer issue on reconfig
Brett Creeley [Mon, 17 Oct 2022 23:31:23 +0000 (16:31 -0700)]
ionic: catch NULL pointer issue on reconfig

It's possible that the driver will dereference a qcq that doesn't exist
when calling ionic_reconfigure_queues(), which causes a page fault BUG.

If a reduction in the number of queues is followed by a different
reconfig such as changing the ring size, the driver can hit a NULL
pointer when trying to clean up non-existent queues.

Fix this by checking to make sure both the qcqs array and qcq entry
exists bofore trying to use and free the entry.

Fixes: 101b40a0171f ("ionic: change queue count with no reset")
Signed-off-by: Brett Creeley <brett@pensando.io>
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Link: https://lore.kernel.org/r/20221017233123.15869-1-snelson@pensando.io
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
20 months agonet: hsr: avoid possible NULL deref in skb_clone()
Eric Dumazet [Mon, 17 Oct 2022 16:59:28 +0000 (16:59 +0000)]
net: hsr: avoid possible NULL deref in skb_clone()

syzbot got a crash [1] in skb_clone(), caused by a bug
in hsr_get_untagged_frame().

When/if create_stripped_skb_hsr() returns NULL, we must
not attempt to call skb_clone().

While we are at it, replace a WARN_ONCE() by netdev_warn_once().

[1]
general protection fault, probably for non-canonical address 0xdffffc000000000f: 0000 [#1] PREEMPT SMP KASAN
KASAN: null-ptr-deref in range [0x0000000000000078-0x000000000000007f]
CPU: 1 PID: 754 Comm: syz-executor.0 Not tainted 6.0.0-syzkaller-02734-g0326074ff465 #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 09/22/2022
RIP: 0010:skb_clone+0x108/0x3c0 net/core/skbuff.c:1641
Code: 93 02 00 00 49 83 7c 24 28 00 0f 85 e9 00 00 00 e8 5d 4a 29 fa 4c 8d 75 7e 48 b8 00 00 00 00 00 fc ff df 4c 89 f2 48 c1 ea 03 <0f> b6 04 02 4c 89 f2 83 e2 07 38 d0 7f 08 84 c0 0f 85 9e 01 00 00
RSP: 0018:ffffc90003ccf4e0 EFLAGS: 00010207

RAX: dffffc0000000000 RBX: ffffc90003ccf5f8 RCX: ffffc9000c24b000
RDX: 000000000000000f RSI: ffffffff8751cb13 RDI: 0000000000000000
RBP: 0000000000000000 R08: 00000000000000f0 R09: 0000000000000140
R10: fffffbfff181d972 R11: 0000000000000000 R12: ffff888161fc3640
R13: 0000000000000a20 R14: 000000000000007e R15: ffffffff8dc5f620
FS: 00007feb621e4700(0000) GS:ffff8880b9b00000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007feb621e3ff8 CR3: 00000001643a9000 CR4: 00000000003506e0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Call Trace:
<TASK>
hsr_get_untagged_frame+0x4e/0x610 net/hsr/hsr_forward.c:164
hsr_forward_do net/hsr/hsr_forward.c:461 [inline]
hsr_forward_skb+0xcca/0x1d50 net/hsr/hsr_forward.c:623
hsr_handle_frame+0x588/0x7c0 net/hsr/hsr_slave.c:69
__netif_receive_skb_core+0x9fe/0x38f0 net/core/dev.c:5379
__netif_receive_skb_one_core+0xae/0x180 net/core/dev.c:5483
__netif_receive_skb+0x1f/0x1c0 net/core/dev.c:5599
netif_receive_skb_internal net/core/dev.c:5685 [inline]
netif_receive_skb+0x12f/0x8d0 net/core/dev.c:5744
tun_rx_batched+0x4ab/0x7a0 drivers/net/tun.c:1544
tun_get_user+0x2686/0x3a00 drivers/net/tun.c:1995
tun_chr_write_iter+0xdb/0x200 drivers/net/tun.c:2025
call_write_iter include/linux/fs.h:2187 [inline]
new_sync_write fs/read_write.c:491 [inline]
vfs_write+0x9e9/0xdd0 fs/read_write.c:584
ksys_write+0x127/0x250 fs/read_write.c:637
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x63/0xcd

Fixes: f266a683a480 ("net/hsr: Better frame dispatch")
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20221017165928.2150130-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
20 months agobnxt_en: fix memory leak in bnxt_nvm_test()
Vikas Gupta [Mon, 17 Oct 2022 15:32:22 +0000 (11:32 -0400)]
bnxt_en: fix memory leak in bnxt_nvm_test()

Free the kzalloc'ed buffer before returning in the success path.

Fixes: 5b6ff128fdf6 ("bnxt_en: implement callbacks for devlink selftests")
Signed-off-by: Vikas Gupta <vikas.gupta@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Link: https://lore.kernel.org/r/1666020742-25834-1-git-send-email-michael.chan@broadcom.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
20 months agoMerge tag 'for-netdev' of git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next
Jakub Kicinski [Wed, 19 Oct 2022 01:56:41 +0000 (18:56 -0700)]
Merge tag 'for-netdev' of git://git./linux/kernel/git/bpf/bpf-next

Daniel Borkmann says:

====================
pull-request: bpf-next 2022-10-18

We've added 33 non-merge commits during the last 14 day(s) which contain
a total of 31 files changed, 874 insertions(+), 538 deletions(-).

The main changes are:

1) Add RCU grace period chaining to BPF to wait for the completion
   of access from both sleepable and non-sleepable BPF programs,
   from Hou Tao & Paul E. McKenney.

2) Improve helper UAPI by explicitly defining BPF_FUNC_xxx integer
   values. In the wild we have seen OS vendors doing buggy backports
   where helper call numbers mismatched. This is an attempt to make
   backports more foolproof, from Andrii Nakryiko.

3) Add libbpf *_opts API-variants for bpf_*_get_fd_by_id() functions,
   from Roberto Sassu.

4) Fix libbpf's BTF dumper for structs with padding-only fields,
   from Eduard Zingerman.

5) Fix various libbpf bugs which have been found from fuzzing with
   malformed BPF object files, from Shung-Hsi Yu.

6) Clean up an unneeded check on existence of SSE2 in BPF x86-64 JIT,
   from Jie Meng.

7) Fix various ASAN bugs in both libbpf and selftests when running
   the BPF selftest suite on arm64, from Xu Kuohai.

8) Fix missing bpf_iter_vma_offset__destroy() call in BPF iter selftest
   and use in-skeleton link pointer to remove an explicit bpf_link__destroy(),
   from Jiri Olsa.

9) Fix BPF CI breakage by pointing to iptables-legacy instead of relying
   on symlinked iptables which got upgraded to iptables-nft,
   from Martin KaFai Lau.

10) Minor BPF selftest improvements all over the place, from various others.

* tag 'for-netdev' of git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next: (33 commits)
  bpf/docs: Update README for most recent vmtest.sh
  bpf: Use rcu_trace_implies_rcu_gp() for program array freeing
  bpf: Use rcu_trace_implies_rcu_gp() in local storage map
  bpf: Use rcu_trace_implies_rcu_gp() in bpf memory allocator
  rcu-tasks: Provide rcu_trace_implies_rcu_gp()
  selftests/bpf: Use sys_pidfd_open() helper when possible
  libbpf: Fix null-pointer dereference in find_prog_by_sec_insn()
  libbpf: Deal with section with no data gracefully
  libbpf: Use elf_getshdrnum() instead of e_shnum
  selftest/bpf: Fix error usage of ASSERT_OK in xdp_adjust_tail.c
  selftests/bpf: Fix error failure of case test_xdp_adjust_tail_grow
  selftest/bpf: Fix memory leak in kprobe_multi_test
  selftests/bpf: Fix memory leak caused by not destroying skeleton
  libbpf: Fix memory leak in parse_usdt_arg()
  libbpf: Fix use-after-free in btf_dump_name_dups
  selftests/bpf: S/iptables/iptables-legacy/ in the bpf_nf and xdp_synproxy test
  selftests/bpf: Alphabetize DENYLISTs
  selftests/bpf: Add tests for _opts variants of bpf_*_get_fd_by_id()
  libbpf: Introduce bpf_link_get_fd_by_id_opts()
  libbpf: Introduce bpf_btf_get_fd_by_id_opts()
  ...
====================

Link: https://lore.kernel.org/r/20221018210631.11211-1-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
20 months agodm clone: Fix typo in block_device format specifier
Nikos Tsironis [Thu, 29 Sep 2022 14:11:48 +0000 (17:11 +0300)]
dm clone: Fix typo in block_device format specifier

Use %pg for printing the block device name, instead of %pd.

Fixes: 385411ffba0c ("dm: stop using bdevname")
Cc: stable@vger.kernel.org # v5.18+
Signed-off-by: Nikos Tsironis <ntsironis@arrikto.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
20 months agodm: remove unnecessary assignment statement in alloc_dev()
Genjian Zhang [Thu, 29 Sep 2022 08:20:36 +0000 (16:20 +0800)]
dm: remove unnecessary assignment statement in alloc_dev()

Fixes: 74fe6ba923949 ("dm: convert to blk_alloc_disk/blk_cleanup_disk")
Signed-off-by: Genjian Zhang <zhanggenjian@kylinos.cn>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
20 months agodm verity: Add documentation for try_verify_in_tasklet option
Milan Broz [Tue, 27 Sep 2022 18:42:26 +0000 (20:42 +0200)]
dm verity: Add documentation for try_verify_in_tasklet option

Add documentation that was missing from commit 5721d4e5a9cd ("dm
verity: Add optional "try_verify_in_tasklet" feature").

Signed-off-by: Milan Broz <gmazyland@gmail.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
20 months agodm cache: delete the redundant word 'each' in comment
Shaomin Deng [Sun, 4 Sep 2022 16:04:27 +0000 (12:04 -0400)]
dm cache: delete the redundant word 'each' in comment

Signed-off-by: Shaomin Deng <dengshaomin@cdjrlc.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
20 months agodm raid: fix typo in analyse_superblocks code comment
Jiangshan Yi [Mon, 5 Sep 2022 02:45:52 +0000 (10:45 +0800)]
dm raid: fix typo in analyse_superblocks code comment

Reported-by: k2ci <kernel-bot@kylinos.cn>
Signed-off-by: Jiangshan Yi <yijiangshan@kylinos.cn>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
20 months agodm verity: enable WQ_HIGHPRI on verify_wq
Nathan Huckleberry [Tue, 30 Aug 2022 18:44:44 +0000 (18:44 +0000)]
dm verity: enable WQ_HIGHPRI on verify_wq

WQ_HIGHPRI increases throughput and decreases disk latency when using
dm-verity. This is important in Android for camera startup speed.

The following tests were run by doing 60 seconds of random reads using
a dm-verity device backed by two ramdisks.

Without WQ_HIGHPRI
lat (usec): min=13, max=3947, avg=69.53, stdev=50.55
READ: bw=51.1MiB/s (53.6MB/s), 51.1MiB/s-51.1MiB/s (53.6MB/s-53.6MB/s)

With WQ_HIGHPRI:
lat (usec): min=13, max=7854, avg=31.15, stdev=30.42
READ: bw=116MiB/s (121MB/s), 116MiB/s-116MiB/s (121MB/s-121MB/s)

Further testing was done by measuring how long it takes to open a
camera on an Android device.

Without WQ_HIGHPRI
Total verity work queue wait times (ms):
880.960, 789.517, 898.852

With WQ_HIGHPRI:
Total verity work queue wait times (ms):
528.824, 439.191, 433.300

The average time to open the camera is reduced by 350ms (or 40-50%).

Signed-off-by: Nathan Huckleberry <nhuck@google.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
20 months agodm raid: delete the redundant word 'that' in comment
Jilin Yuan [Tue, 30 Aug 2022 15:33:45 +0000 (23:33 +0800)]
dm raid: delete the redundant word 'that' in comment

Signed-off-by: Jilin Yuan <yuanjilin@cdjrlc.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
20 months agodm: change from DMWARN to DMERR or DMCRIT for fatal errors
Mikulas Patocka [Wed, 24 Aug 2022 11:25:57 +0000 (07:25 -0400)]
dm: change from DMWARN to DMERR or DMCRIT for fatal errors

Change DMWARN to DMERR in cases when there is an unrecoverable error.
Change DMWARN to DMCRIT when handling of a case is unimplemented.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
20 months agobpf/docs: Update README for most recent vmtest.sh
Daniel Müller [Mon, 17 Oct 2022 23:24:58 +0000 (23:24 +0000)]
bpf/docs: Update README for most recent vmtest.sh

Since commit 40b09653b197 ("selftests/bpf: Adjust vmtest.sh to use local
kernel configuration") the vmtest.sh script no longer downloads a kernel
configuration but uses the local, in-repository one.
This change updates the README, which still mentions the old behavior.

Signed-off-by: Daniel Müller <deso@posteo.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20221017232458.1272762-1-deso@posteo.net
20 months agoMerge tag 'for-6.1-rc1-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave...
Linus Torvalds [Tue, 18 Oct 2022 18:25:50 +0000 (11:25 -0700)]
Merge tag 'for-6.1-rc1-tag' of git://git./linux/kernel/git/kdave/linux

Pull btrfs fixes from David Sterba:

 - fiemap fixes:
      - add missing path cache update
      - fix processing of delayed data and tree refs during backref
        walking, this could lead to reporting incorrect extent sharing

 - fix extent range locking under heavy contention to avoid deadlocks

 - make it possible to test send v3 in debugging mode

 - update links in MAINTAINERS

* tag 'for-6.1-rc1-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  MAINTAINERS: update btrfs website links and files
  btrfs: ignore fiemap path cache if we have multiple leaves for a data extent
  btrfs: fix processing of delayed tree block refs during backref walking
  btrfs: fix processing of delayed data refs during backref walking
  btrfs: delete stale comments after merge conflict resolution
  btrfs: unlock locked extent area if we have contention
  btrfs: send: update command for protocol version check
  btrfs: send: allow protocol version 3 with CONFIG_BTRFS_DEBUG
  btrfs: add missing path cache update during fiemap

20 months agoMerge tag 'erofs-for-6.1-rc2-fixes' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Tue, 18 Oct 2022 18:18:26 +0000 (11:18 -0700)]
Merge tag 'erofs-for-6.1-rc2-fixes' of git://git./linux/kernel/git/xiang/erofs

Pull erofs fixes from Gao Xiang:

 - Fix invalid unmapped accesses when initializing compressed inodes

 - Fix up very rare hung on page lock after enabling compressed data
   deduplication

 - Fix up inplace decompression success rate

 - Take s_inode_list_lock to protect sb->s_inodes for fscache shared
   domain

* tag 'erofs-for-6.1-rc2-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs:
  erofs: protect s_inodes with s_inode_list_lock for fscache
  erofs: fix up inplace decompression success rate
  erofs: shouldn't churn the mapping page for duplicated copies
  erofs: fix illegal unmapped accesses in z_erofs_fill_inode_lazy()

20 months agoMerge branch 'Remove unnecessary RCU grace period chaining'
Alexei Starovoitov [Tue, 18 Oct 2022 17:27:02 +0000 (10:27 -0700)]
Merge branch 'Remove unnecessary RCU grace period chaining'

Hou Tao says:

====================
Now bpf uses RCU grace period chaining to wait for the completion of
access from both sleepable and non-sleepable bpf program: calling
call_rcu_tasks_trace() firstly to wait for a RCU-tasks-trace grace
period, then in its callback calls call_rcu() or kfree_rcu() to wait for
a normal RCU grace period.

According to the implementation of RCU Tasks Trace, it inovkes
->postscan_func() to wait for one RCU-tasks-trace grace period and
rcu_tasks_trace_postscan() inovkes synchronize_rcu() to wait for one
normal RCU grace period in turn, so one RCU-tasks-trace grace period
will imply one normal RCU grace period. To codify the implication,
introduces rcu_trace_implies_rcu_gp() in patch #1. And using it in patch
Other two uses of call_rcu_tasks_trace() are unchanged: for
__bpf_prog_put_rcu() there is no gp chain and for
__bpf_tramp_image_put_rcu_tasks() it chains RCU tasks trace GP and RCU
tasks GP.

An alternative way to remove these unnecessary RCU grace period
chainings is using the RCU polling API to check whether or not a normal
RCU grace period has passed (e.g. get_state_synchronize_rcu()). But it
needs an unsigned long space for each free element or each call, and
it is not affordable for local storage element, so as for now always
rcu_trace_implies_rcu_gp().

Comments are always welcome.

Change Log:

v2:
 * codify the implication of RCU Tasks Trace grace period instead of
   assuming for it

v1: https://lore.kernel.org/bpf/20221011071128.3470622-1-houtao@huaweicloud.com

Hou Tao (3):
  bpf: Use rcu_trace_implies_rcu_gp() in bpf memory allocator
  bpf: Use rcu_trace_implies_rcu_gp() in local storage map
  bpf: Use rcu_trace_implies_rcu_gp() for program array freeing
====================

Reviewed-by: Paul E. McKenney <paulmck@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
20 months agobpf: Use rcu_trace_implies_rcu_gp() for program array freeing
Hou Tao [Fri, 14 Oct 2022 11:39:46 +0000 (19:39 +0800)]
bpf: Use rcu_trace_implies_rcu_gp() for program array freeing

To support both sleepable and normal uprobe bpf program, the freeing of
trace program array chains a RCU-tasks-trace grace period and a normal
RCU grace period one after the other.

With the introduction of rcu_trace_implies_rcu_gp(),
__bpf_prog_array_free_sleepable_cb() can check whether or not a normal
RCU grace period has also passed after a RCU-tasks-trace grace period
has passed. If it is true, it is safe to invoke kfree() directly.

Signed-off-by: Hou Tao <houtao1@huawei.com>
Link: https://lore.kernel.org/r/20221014113946.965131-5-houtao@huaweicloud.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
20 months agobpf: Use rcu_trace_implies_rcu_gp() in local storage map
Hou Tao [Fri, 14 Oct 2022 11:39:45 +0000 (19:39 +0800)]
bpf: Use rcu_trace_implies_rcu_gp() in local storage map

Local storage map is accessible for both sleepable and non-sleepable bpf
program, and its memory is freed by using both call_rcu_tasks_trace() and
kfree_rcu() to wait for both RCU-tasks-trace grace period and RCU grace
period to pass.

With the introduction of rcu_trace_implies_rcu_gp(), both
bpf_selem_free_rcu() and bpf_local_storage_free_rcu() can check whether
or not a normal RCU grace period has also passed after a RCU-tasks-trace
grace period has passed. If it is true, it is safe to call kfree()
directly.

Signed-off-by: Hou Tao <houtao1@huawei.com>
Link: https://lore.kernel.org/r/20221014113946.965131-4-houtao@huaweicloud.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>