The dev->can.state is set to CAN_STATE_ERROR_ACTIVE, after the device
has been started. On busy networks the CAN controller might receive
CAN frame between and go into an error state before the dev->can.state
is assigned.

Assign dev->can.state before starting the controller to close the race
window.

Fixes: d08e973a

 ("can: gs_usb: Added support for the GS_USB CAN devices")
Link: https://lore.kernel.org/all/20220920195216.232481-1-mkl@pengutronix.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

The following happened on an i.MX25 using flexcan with many packets on
the bus:

The rx-offload queue reached a length more than skb_queue_len_max. In
can_rx_offload_offload_one() the drop variable was set to true which
made the call to .mailbox_read() (here: flexcan_mailbox_read()) to
_always_ return ERR_PTR(-ENOBUFS) and drop the rx'ed CAN frame. So
can_rx_offload_offload_one() returned ERR_PTR(-ENOBUFS), too.

can_rx_offload_irq_offload_fifo() looks as follows:

| 	while (1) {
| 		skb = can_rx_offload_offload_one(offload, 0);
| 		if (IS_ERR(skb))
| 			continue;
| 		if (!skb)
| 			break;
| 		...
| 	}

The flexcan driver wrongly always returns ERR_PTR(-ENOBUFS) if drop is
requested, even if there is no CAN frame pending. As the i.MX25 is a
single core CPU, while the rx-offload processing is active, there is
no thread to process packets from the offload queue. So the queue
doesn't get any shorter and this results is a tight loop.

Instead of always returning ERR_PTR(-ENOBUFS) if drop is requested,
return NULL if no CAN frame is pending.

Changes since v1: https://lore.kernel.org/all/20220810144536.389237-1-u.kleine-koenig@pengutronix.de
- don't break in can_rx_offload_irq_offload_fifo() in case of an error,
  return NULL in flexcan_mailbox_read() in case of no pending CAN frame
  instead

Fixes: 4e9c9484

 ("can: rx-offload: Prepare for CAN FD support")
Link: https://lore.kernel.org/all/20220811094254.1864367-1-mkl@pengutronix.de
Cc: stable@vger.kernel.org # v5.5
Suggested-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Reviewed-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Tested-by: Thorsten Scherer <t.scherer@eckelmann.de>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

Since commit 744d23c7 ("net: phy: Warn about incorrect
mdio_bus_phy_resume() state"), a warning splat is printed during system
resume with Wake-on-LAN disabled:

	WARNING: CPU: 0 PID: 626 at drivers/net/phy/phy_device.c:323 mdio_bus_phy_resume+0xbc/0xe4

As the Renesas SuperH Ethernet driver already calls phy_{stop,start}()
in its suspend/resume callbacks, it is sufficient to just mark the MAC
responsible for managing the power state of the PHY.

Fixes: fba863b8

 ("net: phy: make PHY PM ops a no-op if MAC driver manages PHY PM")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Link: https://lore.kernel.org/r/c6e1331b9bef61225fa4c09db3ba3e2e7214ba2d.1663598886.git.geert+renesas@glider.be
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Since commit 744d23c7 ("net: phy: Warn about incorrect
mdio_bus_phy_resume() state"), a warning splat is printed during system
resume with Wake-on-LAN disabled:

        WARNING: CPU: 0 PID: 1197 at drivers/net/phy/phy_device.c:323 mdio_bus_phy_resume+0xbc/0xc8

As the Renesas Ethernet AVB driver already calls phy_{stop,start}() in
its suspend/resume callbacks, it is sufficient to just mark the MAC
responsible for managing the power state of the PHY.

Fixes: fba863b8

 ("net: phy: make PHY PM ops a no-op if MAC driver manages PHY PM")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Link: https://lore.kernel.org/r/8ec796f47620980fdd0403e21bd8b7200b4fa1d4.1663598796.git.geert+renesas@glider.be
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

As suggested by Vinod, adding myself as the reviewer
for the Qualcomm ETHQOS Ethernet driver.

Recently I have enabled this driver on a few Qualcomm
SoCs / boards and hence trying to keep a close eye on
it.

Signed-off-by: Bhupesh Sharma <bhupesh.sharma@linaro.org>
Acked-by: Vinod Koul <vkoul@kernel.org>
Link: https://lore.kernel.org/r/20220915112804.3950680-1-bhupesh.sharma@linaro.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Vladimir Oltean says:

====================
Fixes for tc-taprio software mode

While working on some new features for tc-taprio, I found some strange
behavior which looked like bugs. I was able to eventually trigger a NULL
pointer dereference. This patch set fixes 2 issues I saw. Detailed
explanation in patches.
====================

Link: https://lore.kernel.org/r/20220915100802.2308279-1-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

taprio can only operate as root qdisc, and to that end, there exists the
following check in taprio_init(), just as in mqprio:

	if (sch->parent != TC_H_ROOT)
		return -EOPNOTSUPP;

And indeed, when we try to attach taprio to an mqprio child, it fails as
expected:

$ tc qdisc add dev swp0 root handle 1: mqprio num_tc 8 \
	map 0 1 2 3 4 5 6 7 \
	queues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 hw 0
$ tc qdisc replace dev swp0 parent 1:2 taprio num_tc 8 \
	map 0 1 2 3 4 5 6 7 \
	queues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 \
	base-time 0 sched-entry S 0x7f 990000 sched-entry S 0x80 100000 \
	flags 0x0 clockid CLOCK_TAI
Error: sch_taprio: Can only be attached as root qdisc.

(extack message added by me)

But when we try to attach a taprio child to a taprio root qdisc,
surprisingly it doesn't fail:

$ tc qdisc replace dev swp0 root handle 1: taprio num_tc 8 \
	map 0 1 2 3 4 5 6 7 queues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 \
	base-time 0 sched-entry S 0x7f 990000 sched-entry S 0x80 100000 \
	flags 0x0 clockid CLOCK_TAI
$ tc qdisc replace dev swp0 parent 1:2 taprio num_tc 8 \
	map 0 1 2 3 4 5 6 7 \
	queues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 \
	base-time 0 sched-entry S 0x7f 990000 sched-entry S 0x80 100000 \
	flags 0x0 clockid CLOCK_TAI

This is because tc_modify_qdisc() behaves differently when mqprio is
root, vs when taprio is root.

In the mqprio case, it finds the parent qdisc through
p = qdisc_lookup(dev, TC_H_MAJ(clid)), and then the child qdisc through
q = qdisc_leaf(p, clid). This leaf qdisc q has handle 0, so it is
ignored according to the comment right below ("It may be default qdisc,
ignore it"). As a result, tc_modify_qdisc() goes through the
qdisc_create() code path, and this gives taprio_init() a chance to check
for sch_parent != TC_H_ROOT and error out.

Whereas in the taprio case, the returned q = qdisc_leaf(p, clid) is
different. It is not the default qdisc created for each netdev queue
(both taprio and mqprio call qdisc_create_dflt() and keep them in
a private q->qdiscs[], or priv->qdiscs[], respectively). Instead, taprio
makes qdisc_leaf() return the _root_ qdisc, aka itself.

When taprio does that, tc_modify_qdisc() goes through the qdisc_change()
code path, because the qdisc layer never finds out about the child qdisc
of the root. And through the ->change() ops, taprio has no reason to
check whether its parent is root or not, just through ->init(), which is
not called.

The problem is the taprio_leaf() implementation. Even though code wise,
it does the exact same thing as mqprio_leaf() which it is copied from,
it works with different input data. This is because mqprio does not
attach itself (the root) to each device TX queue, but one of the default
qdiscs from its private array.

In fact, since commit 13511704 ("net: taprio offload: enforce qdisc
to netdev queue mapping"), taprio does this too, but just for the full
offload case. So if we tried to attach a taprio child to a fully
offloaded taprio root qdisc, it would properly fail too; just not to a
software root taprio.

To fix the problem, stop looking at the Qdisc that's attached to the TX
queue, and instead, always return the default qdiscs that we've
allocated (and to which we privately enqueue and dequeue, in software
scheduling mode).

Since Qdisc_class_ops :: leaf  is only called from tc_modify_qdisc(),
the risk of unforeseen side effects introduced by this change is
minimal.

Fixes: 5a781ccb

 ("tc: Add support for configuring the taprio scheduler")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

In an incredibly strange API design decision, qdisc->destroy() gets
called even if qdisc->init() never succeeded, not exclusively since
commit 87b60cfa ("net_sched: fix error recovery at qdisc creation"),
but apparently also earlier (in the case of qdisc_create_dflt()).

The taprio qdisc does not fully acknowledge this when it attempts full
offload, because it starts off with q->flags = TAPRIO_FLAGS_INVALID in
taprio_init(), then it replaces q->flags with TCA_TAPRIO_ATTR_FLAGS
parsed from netlink (in taprio_change(), tail called from taprio_init()).

But in taprio_destroy(), we call taprio_disable_offload(), and this
determines what to do based on FULL_OFFLOAD_IS_ENABLED(q->flags).

But looking at the implementation of FULL_OFFLOAD_IS_ENABLED()
(a bitwise check of bit 1 in q->flags), it is invalid to call this macro
on q->flags when it contains TAPRIO_FLAGS_INVALID, because that is set
to U32_MAX, and therefore FULL_OFFLOAD_IS_ENABLED() will return true on
an invalid set of flags.

As a result, it is possible to crash the kernel if user space forces an
error between setting q->flags = TAPRIO_FLAGS_INVALID, and the calling
of taprio_enable_offload(). This is because drivers do not expect the
offload to be disabled when it was never enabled.

The error that we force here is to attach taprio as a non-root qdisc,
but instead as child of an mqprio root qdisc:

$ tc qdisc add dev swp0 root handle 1: \
	mqprio num_tc 8 map 0 1 2 3 4 5 6 7 \
	queues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 hw 0
$ tc qdisc replace dev swp0 parent 1:1 \
	taprio num_tc 8 map 0 1 2 3 4 5 6 7 \
	queues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 base-time 0 \
	sched-entry S 0x7f 990000 sched-entry S 0x80 100000 \
	flags 0x0 clockid CLOCK_TAI
Unable to handle kernel paging request at virtual address fffffffffffffff8
[fffffffffffffff8] pgd=0000000000000000, p4d=0000000000000000
Internal error: Oops: 96000004 [#1] PREEMPT SMP
Call trace:
 taprio_dump+0x27c/0x310
 vsc9959_port_setup_tc+0x1f4/0x460
 felix_port_setup_tc+0x24/0x3c
 dsa_slave_setup_tc+0x54/0x27c
 taprio_disable_offload.isra.0+0x58/0xe0
 taprio_destroy+0x80/0x104
 qdisc_create+0x240/0x470
 tc_modify_qdisc+0x1fc/0x6b0
 rtnetlink_rcv_msg+0x12c/0x390
 netlink_rcv_skb+0x5c/0x130
 rtnetlink_rcv+0x1c/0x2c

Fix this by keeping track of the operations we made, and undo the
offload only if we actually did it.

I've added "bool offloaded" inside a 4 byte hole between "int clockid"
and "atomic64_t picos_per_byte". Now the first cache line looks like
below:

$ pahole -C taprio_sched net/sched/sch_taprio.o
struct taprio_sched {
        struct Qdisc * *           qdiscs;               /*     0     8 */
        struct Qdisc *             root;                 /*     8     8 */
        u32                        flags;                /*    16     4 */
        enum tk_offsets            tk_offset;            /*    20     4 */
        int                        clockid;              /*    24     4 */
        bool                       offloaded;            /*    28     1 */

        /* XXX 3 bytes hole, try to pack */

        atomic64_t                 picos_per_byte;       /*    32     0 */

        /* XXX 8 bytes hole, try to pack */

        spinlock_t                 current_entry_lock;   /*    40     0 */

        /* XXX 8 bytes hole, try to pack */

        struct sched_entry *       current_entry;        /*    48     8 */
        struct sched_gate_list *   oper_sched;           /*    56     8 */
        /* --- cacheline 1 boundary (64 bytes) --- */

Fixes: 9c66d156

 ("taprio: Add support for hardware offloading")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

The global 'raw_v6_hashinfo' variable can be accessed even when IPv6 is
administratively disabled via the 'ipv6.disable=1' kernel command line
option, leading to a crash [1].

Fix by restoring the original behavior and always initializing the
variable, regardless of IPv6 support being administratively disabled or
not.

[1]
 BUG: unable to handle page fault for address: ffffffffffffffc8
 #PF: supervisor read access in kernel mode
 #PF: error_code(0x0000) - not-present page
 PGD 173e18067 P4D 173e18067 PUD 173e1a067 PMD 0
 Oops: 0000 [#1] PREEMPT SMP KASAN
 CPU: 3 PID: 271 Comm: ss Not tainted 6.0.0-rc4-custom-00136-g0727a9a5fbc1 #1396
 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.0-1.fc36 04/01/2014
 RIP: 0010:raw_diag_dump+0x310/0x7f0
 [...]
 Call Trace:
  <TASK>
  __inet_diag_dump+0x10f/0x2e0
  netlink_dump+0x575/0xfd0
  __netlink_dump_start+0x67b/0x940
  inet_diag_handler_cmd+0x273/0x2d0
  sock_diag_rcv_msg+0x317/0x440
  netlink_rcv_skb+0x15e/0x430
  sock_diag_rcv+0x2b/0x40
  netlink_unicast+0x53b/0x800
  netlink_sendmsg+0x945/0xe60
  ____sys_sendmsg+0x747/0x960
  ___sys_sendmsg+0x13a/0x1e0
  __sys_sendmsg+0x118/0x1e0
  do_syscall_64+0x34/0x80
  entry_SYSCALL_64_after_hwframe+0x63/0xcd

Fixes: 1da177e4 ("Linux-2.6.12-rc2")
Fixes: 0daf07e5

 ("raw: convert raw sockets to RCU")
Reported-by: Roberto Ricci <rroberto2r@gmail.com>
Tested-by: Roberto Ricci <rroberto2r@gmail.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20220916084821.229287-1-idosch@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

TSN features on the ENETC (taprio, cbs, gate, police) are configured
through a mix of command BD ring messages and port registers:
enetc_port_rd(), enetc_port_wr().

Port registers are a region of the ENETC memory map which are only
accessible from the PCIe Physical Function. They are not accessible from
the Virtual Functions.

Moreover, attempting to access these registers crashes the kernel:

$ echo 1 > /sys/bus/pci/devices/0000\:00\:00.0/sriov_numvfs
pci 0000:00:01.0: [1957:ef00] type 00 class 0x020001
fsl_enetc_vf 0000:00:01.0: Adding to iommu group 15
fsl_enetc_vf 0000:00:01.0: enabling device (0000 -> 0002)
fsl_enetc_vf 0000:00:01.0 eno0vf0: renamed from eth0
$ tc qdisc replace dev eno0vf0 root taprio num_tc 8 map 0 1 2 3 4 5 6 7 \
	queues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 base-time 0 \
	sched-entry S 0x7f 900000 sched-entry S 0x80 100000 flags 0x2
Unable to handle kernel paging request at virtual address ffff800009551a08
Internal error: Oops: 96000007 [#1] PREEMPT SMP
pc : enetc_setup_tc_taprio+0x170/0x47c
lr : enetc_setup_tc_taprio+0x16c/0x47c
Call trace:
 enetc_setup_tc_taprio+0x170/0x47c
 enetc_setup_tc+0x38/0x2dc
 taprio_change+0x43c/0x970
 taprio_init+0x188/0x1e0
 qdisc_create+0x114/0x470
 tc_modify_qdisc+0x1fc/0x6c0
 rtnetlink_rcv_msg+0x12c/0x390

Split enetc_setup_tc() into separate functions for the PF and for the
VF drivers. Also remove enetc_qos.o from being included into
enetc-vf.ko, since it serves absolutely no purpose there.

Fixes: 34c6adf1

 ("enetc: Configure the Time-Aware Scheduler via tc-taprio offload")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Link: https://lore.kernel.org/r/20220916133209.3351399-2-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

The VF netdev driver shouldn't respond to changes in the NETIF_F_HW_TC
flag; only PFs should. Moreover, TSN-specific code should go to
enetc_qos.c, which should not be included in the VF driver.

Fixes: 79e49982

 ("net: enetc: add hw tc hw offload features for PSPF capability")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Link: https://lore.kernel.org/r/20220916133209.3351399-1-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Jason A. Donenfeld says:

====================
wireguard patches for 6.0-rc6

1) The ratelimiter timing test doesn't help outside of development, yet
   it is currently preventing the module from being inserted on some
   kernels when it flakes at insertion time. So we disable it.

2) A fix for a build error on UML, caused by a recent change in a
   different tree.

3) A WARN_ON() is triggered by Kees' new fortified memcpy() patch, due
   to memcpy()ing over a sockaddr pointer with the size of a
   sockaddr_in[6]. The type safe fix is pretty simple. Given how classic
   of a thing sockaddr punning is, I suspect this may be the first in a
   few patches like this throughout the net tree, once Kees' fortify
   series is more widely deployed (current it's just in next).
====================

Link: https://lore.kernel.org/r/20220916143740.831881-1-Jason@zx2c4.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Doing a variable-sized memcpy is slower, and the compiler isn't smart
enough to turn this into a constant-size assignment.

Further, Kees' latest fortified memcpy will actually bark, because the
destination pointer is type sockaddr, not explicitly sockaddr_in or
sockaddr_in6, so it thinks there's an overflow:

    memcpy: detected field-spanning write (size 28) of single field
    "&endpoint.addr" at drivers/net/wireguard/netlink.c:446 (size 16)

Fix this by just assigning by using explicit casts for each checked
case.

Fixes: e7096c13

 ("net: WireGuard secure network tunnel")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reported-by:  <syzbot+a448cda4dba2dac50de5@syzkaller.appspotmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Since 1b620d53 ("kbuild: disable header exports for UML in a
straightforward way"), installing headers fails on UML, so just disable
installing them, since they're not needed anyway on the architecture.

Fixes: b438b3b8

 ("wireguard: selftests: support UML")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

A previous commit tried to make the ratelimiter timings test more
reliable but in the process made it less reliable on other
configurations. This is an impossible problem to solve without
increasingly ridiculous heuristics. And it's not even a problem that
actually needs to be solved in any comprehensive way, since this is only
ever used during development. So just cordon this off with a DEBUG_
ifdef, just like we do for the trie's randomized tests, so it can be
enabled while hacking on the code, and otherwise disabled in CI. In the
process we also revert 151c8e49.

Fixes: 151c8e49 ("wireguard: ratelimiter: use hrtimer in selftest")
Fixes: e7096c13

 ("net: WireGuard secure network tunnel")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Like in previous patch for sfc, prevent potential (but unlikely) NULL
pointer dereference.

Fixes: 12804793

 ("sfc: decouple TXQ type from label")
Reported-by: Tianhao Zhao <tizhao@redhat.com>
Signed-off-by: Íñigo Huguet <ihuguet@redhat.com>
Link: https://lore.kernel.org/r/20220915141958.16458-1-ihuguet@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

As in previous commit for sfc, fix TX channels offset when
efx_siena_separate_tx_channels is false (the default)

Fixes: 25bde571

 ("sfc/siena: fix wrong tx channel offset with efx_separate_tx_channels")
Reported-by: Tianhao Zhao <tizhao@redhat.com>
Signed-off-by: Íñigo Huguet <ihuguet@redhat.com>
Link: https://lore.kernel.org/r/20220915141653.15504-1-ihuguet@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

syzbot is still complaining uninit-value in tcp_recvmsg(), for
commit 1228b34c

 ("net: clear msg_get_inq in __sys_recvfrom() and
__copy_msghdr_from_user()") missed that __get_compat_msghdr() is called
instead of copy_msghdr_from_user() when MSG_CMSG_COMPAT is specified.

Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Fixes: 1228b34c

 ("net: clear msg_get_inq in __sys_recvfrom() and __copy_msghdr_from_user()")
Reviewed-by: Jens Axboe <axboe@kernel.dk>
Link: https://lore.kernel.org/r/d06d0f7f-696c-83b4-b2d5-70b5f2730a37@I-love.SAKURA.ne.jp
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Ido Schimmel says:

====================
ipmr: Always call ip{,6}_mr_forward() from RCU read-side critical section

Patch #1 fixes a bug in ipmr code.

Patch #2 adds corresponding test cases.
====================

Link: https://lore.kernel.org/r/20220914075339.4074096-1-idosch@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Add IPv4 and IPv6 test cases for unresolved multicast routes, testing
that queued packets are forwarded after installing a matching (S, G)
route.

The test cases can be used to reproduce the bugs fixed in "ipmr: Always
call ip{,6}_mr_forward() from RCU read-side critical section".

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

These functions expect to be called from RCU read-side critical section,
but this only happens when invoked from the data path via
ip{,6}_mr_input(). They can also be invoked from process context in
response to user space adding a multicast route which resolves a cache
entry with queued packets [1][2].

Fix by adding missing rcu_read_lock() / rcu_read_unlock() in these call
paths.

[1]
WARNING: suspicious RCU usage
6.0.0-rc3-custom-15969-g049d233c8bcc-dirty #1387 Not tainted
-----------------------------
net/ipv4/ipmr.c:84 suspicious rcu_dereference_check() usage!

other info that might help us debug this:

rcu_scheduler_active = 2, debug_locks = 1
1 lock held by smcrouted/246:
 #0: ffffffff862389b0 (rtnl_mutex){+.+.}-{3:3}, at: ip_mroute_setsockopt+0x11c/0x1420

stack backtrace:
CPU: 0 PID: 246 Comm: smcrouted Not tainted 6.0.0-rc3-custom-15969-g049d233c8bcc-dirty #1387
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.0-1.fc36 04/01/2014
Call Trace:
 <TASK>
 dump_stack_lvl+0x91/0xb9
 vif_dev_read+0xbf/0xd0
 ipmr_queue_xmit+0x135/0x1ab0
 ip_mr_forward+0xe7b/0x13d0
 ipmr_mfc_add+0x1a06/0x2ad0
 ip_mroute_setsockopt+0x5c1/0x1420
 do_ip_setsockopt+0x23d/0x37f0
 ip_setsockopt+0x56/0x80
 raw_setsockopt+0x219/0x290
 __sys_setsockopt+0x236/0x4d0
 __x64_sys_setsockopt+0xbe/0x160
 do_syscall_64+0x34/0x80
 entry_SYSCALL_64_after_hwframe+0x63/0xcd

[2]
WARNING: suspicious RCU usage
6.0.0-rc3-custom-15969-g049d233c8bcc-dirty #1387 Not tainted
-----------------------------
net/ipv6/ip6mr.c:69 suspicious rcu_dereference_check() usage!

other info that might help us debug this:

rcu_scheduler_active = 2, debug_locks = 1
1 lock held by smcrouted/246:
 #0: ffffffff862389b0 (rtnl_mutex){+.+.}-{3:3}, at: ip6_mroute_setsockopt+0x6b9/0x2630

stack backtrace:
CPU: 1 PID: 246 Comm: smcrouted Not tainted 6.0.0-rc3-custom-15969-g049d233c8bcc-dirty #1387
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.0-1.fc36 04/01/2014
Call Trace:
 <TASK>
 dump_stack_lvl+0x91/0xb9
 vif_dev_read+0xbf/0xd0
 ip6mr_forward2.isra.0+0xc9/0x1160
 ip6_mr_forward+0xef0/0x13f0
 ip6mr_mfc_add+0x1ff2/0x31f0
 ip6_mroute_setsockopt+0x1825/0x2630
 do_ipv6_setsockopt+0x462/0x4440
 ipv6_setsockopt+0x105/0x140
 rawv6_setsockopt+0xd8/0x690
 __sys_setsockopt+0x236/0x4d0
 __x64_sys_setsockopt+0xbe/0x160
 do_syscall_64+0x34/0x80
 entry_SYSCALL_64_after_hwframe+0x63/0xcd

Fixes: ebc31979

 ("ipmr: add rcu protection over (struct vif_device)->dev")
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

IPA can route packets between IPA-connected entities.  The AP and
modem are currently the only such entities supported, and no routing
is required to transfer packets between them.

The number of entries in each routing table is fixed, and defined at
initialization time.  Some of these entries are designated for use
by the modem, and the rest are available for the AP to use.  The AP
sends a QMI message to the modem which describes (among other
things) information about routing table memory available for the
modem to use.

Currently the QMI initialization packet gives wrong information in
its description of routing tables.  What *should* be supplied is the
maximum index that the modem can use for the routing table memory
located at a given location.  The current code instead supplies the
total *number* of routing table entries.  Furthermore, the modem is
granted the entire table, not just the subset it's supposed to use.

This patch fixes this.  First, the ipa_mem_bounds structure is
generalized so its "end" field can be interpreted either as a final
byte offset, or a final array index.  Second, the IPv4 and IPv6
(non-hashed and hashed) table information fields in the QMI
ipa_init_modem_driver_req structure are changed to be ipa_mem_bounds
rather than ipa_mem_array structures.  Third, we set the "end" value
for each routing table to be the last index, rather than setting the
"count" to be the number of indices.  Finally, instead of allowing
the modem to use all of a routing table's memory, it is limited to
just the portion meant to be used by the modem.  In all versions of
IPA currently supported, that is IPA_ROUTE_MODEM_COUNT (8) entries.

Update a few comments for clarity.

Fixes: 530f9216

 ("soc: qcom: ipa: AP/modem communications")
Signed-off-by: Alex Elder <elder@linaro.org>
Link: https://lore.kernel.org/r/20220913204602.1803004-1-elder@linaro.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

In of_mdiobus_register(), we should call of_node_put() for 'child'
escaped out of for_each_available_child_of_node().

Fixes: 66bdede4

 ("of_mdio: Fix broken PHY IRQ in case of probe deferral")
Co-developed-by: Miaoqian Lin <linmq006@gmail.com>
Signed-off-by: Miaoqian Lin <linmq006@gmail.com>
Signed-off-by: Liang He <windhl@126.com>
Link: https://lore.kernel.org/r/20220913125659.3331969-1-windhl@126.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Before we switched to ->read_skb(), ->read_sock() was passed with
desc.count=1, which technically indicates we only read one skb per
->sk_data_ready() call. However, for TCP, this is not true.

TCP at least has sk_rcvlowat which intentionally holds skb's in
receive queue until this watermark is reached. This means when
->sk_data_ready() is invoked there could be multiple skb's in the
queue, therefore we have to read multiple skbs in tcp_read_skb()
instead of one.

Fixes: 965b57b4

 ("net: Introduce a new proto_ops ->read_skb()")
Reported-by: Peilin Ye <peilin.ye@bytedance.com>
Cc: John Fastabend <john.fastabend@gmail.com>
Cc: Jakub Sitnicki <jakub@cloudflare.com>
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Link: https://lore.kernel.org/r/20220912173553.235838-1-xiyou.wangcong@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

Francesco Dolcini says:

====================
Revert fec PTP changes

Revert the last 2 FEC PTP changes from Csókás Bence, they are causing multiple
issues and we are at 6.0-rc5.
====================

Link: https://lore.kernel.org/r/20220912070143.98153-1-francesco.dolcini@toradex.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

This reverts commit b353b241, this is
creating multiple issues, just not ready to be merged yet.

Link: https://lore.kernel.org/all/CAHk-=wj1obPoTu1AHj9Bd_BGYjdjDyPP+vT5WMj8eheb3A9WHw@mail.gmail.com/
Link: https://lore.kernel.org/all/20220907143915.5w65kainpykfobte@pengutronix.de/
Fixes: b353b241

 ("net: fec: Use a spinlock to guard `fep->ptp_clk_on`")
Signed-off-by: Francesco Dolcini <francesco.dolcini@toradex.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Tested-by: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

This reverts commit f7995922, this is
creating multiple issues, just not ready to be merged yet.

Link: https://lore.kernel.org/all/20220905180542.GA3685102@roeck-us.net/
Link: https://lore.kernel.org/all/CAHk-=wj1obPoTu1AHj9Bd_BGYjdjDyPP+vT5WMj8eheb3A9WHw@mail.gmail.com/
Fixes: f7995922

 ("fec: Restart PPS after link state change")
Signed-off-by: Francesco Dolcini <francesco.dolcini@toradex.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Tested-by: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

Maximum frame length check is enabled in lan937x switch on POR, But it
is found to be disabled on driver during port setup operation. Due to
this, packets are not dropped when transmitted with greater than configured
value. For testing, setup made for lan1->lan2 transmission and configured
lan1 interface with a frame length (less than 1500 as mentioned in
documentation) and transmitted packets with greater than configured value.
Expected no packets at lan2 end, but packets observed at lan2.

Based on the documentation, packets should get discarded if the actual
packet length doesn't match the frame length configured. Frame length check
should be disabled only for cascaded ports due to tailtags.

This feature was disabled on ksz9477 series due to ptp issue, which is
not in lan937x series. But since lan937x took ksz9477 as base, frame
length check disabled here as well. Patch added to remove this portion
from port setup so that maximum frame length check will be active for
normal ports.

Fixes: 55ab6ffa

 ("net: dsa: microchip: add DSA support for microchip LAN937x")
Signed-off-by: Rakesh Sankaranarayanan <rakesh.sankaranarayanan@microchip.com>
Link: https://lore.kernel.org/r/20220912051228.1306074-1-rakesh.sankaranarayanan@microchip.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

Use GFP_ATOMIC when allocating pages out of the hotpath,
continue to use GFP_KERNEL when allocating pages during setup.

GFP_KERNEL will allow blocking which allows it to succeed
more often in a low memory enviornment but in the hotpath we do
not want to allow the allocation to block.

Fixes: 9b8dd5e5

 ("gve: DQO: Add RX path")
Signed-off-by: Shailend Chand <shailend@google.com>
Signed-off-by: Jeroen de Borst <jeroendb@google.com>
Link: https://lore.kernel.org/r/20220913000901.959546-1-jeroendb@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

The warning message of unsupported FW appears every time RX timestamps
are disabled on the interface. The patch fixes the flags to correct set
for the check.

Fixes: 66ed81dc

 ("bnxt_en: Enable packet timestamping for all RX packets")
Cc: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: Vadim Fedorenko <vfedorenko@novek.ru>
Reviewed-by: Andy Gospodarek <gospo@broadcom.com>
Reviewed-by: Michael Chan <michael.chan@broadcom.com>
Link: https://lore.kernel.org/r/20220915234932.25497-1-vfedorenko@novek.ru
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Kalle Valo says:

====================
wireless fixes for v6.0

Late stage fixes for v6.0. Temporarily mark iwlwifi's mei code broken
as it breaks suspend for iwd users and also don't spam nss trimming
messages. mt76 has fixes for aggregation sequence numbers and a
regression related to the VHT extended NSS BW feature.

* tag 'wireless-2022-09-19' of git://git.kernel.org/pub/scm/linux/kernel/git/wireless/wireless:
  wifi: mt76: fix 5 GHz connection regression on mt76x0/mt76x2
  wifi: mt76: fix reading current per-tid starting sequence number for aggregation
  wifi: iwlwifi: Mark IWLMEI as broken
  wifi: iwlwifi: don't spam logs with NSS>2 messages
====================

Link: https://lore.kernel.org/r/20220919105003.1EAE7C433B5@smtp.kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Simon Wunderlich says:

====================
Here is a batman-adv bugfix:

 - Fix hang up with small MTU hard-interface, by Shigeru Yoshida

* tag 'batadv-net-pullrequest-20220916' of git://git.open-mesh.org/linux-merge:
  batman-adv: Fix hang up with small MTU hard-interface
====================

Link: https://lore.kernel.org/r/20220916160931.1412407-1-sw@simonwunderlich.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Trying to get the channel from the tx_queue variable here is wrong
because we can only be here if tx_queue is NULL, so we shouldn't
dereference it. As the above comment in the code says, this is very
unlikely to happen, but it's wrong anyway so let's fix it.

I hit this issue because of a different bug that caused tx_queue to be
NULL. If that happens, this is the error message that we get here:
  BUG: unable to handle kernel NULL pointer dereference at 0000000000000020
  [...]
  RIP: 0010:efx_hard_start_xmit+0x153/0x170 [sfc]

Fixes: 12804793

 ("sfc: decouple TXQ type from label")
Reported-by: Tianhao Zhao <tizhao@redhat.com>
Signed-off-by: Íñigo Huguet <ihuguet@redhat.com>
Acked-by: Edward Cree <ecree.xilinx@gmail.com>
Link: https://lore.kernel.org/r/20220914111135.21038-1-ihuguet@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

In legacy interrupt mode the tx_channel_offset was hardcoded to 1, but
that's not correct if efx_sepparate_tx_channels is false. In that case,
the offset is 0 because the tx queues are in the single existing channel
at index 0, together with the rx queue.

Without this fix, as soon as you try to send any traffic, it tries to
get the tx queues from an uninitialized channel getting these errors:
  WARNING: CPU: 1 PID: 0 at drivers/net/ethernet/sfc/tx.c:540 efx_hard_start_xmit+0x12e/0x170 [sfc]
  [...]
  RIP: 0010:efx_hard_start_xmit+0x12e/0x170 [sfc]
  [...]
  Call Trace:
   <IRQ>
   dev_hard_start_xmit+0xd7/0x230
   sch_direct_xmit+0x9f/0x360
   __dev_queue_xmit+0x890/0xa40
  [...]
  BUG: unable to handle kernel NULL pointer dereference at 0000000000000020
  [...]
  RIP: 0010:efx_hard_start_xmit+0x153/0x170 [sfc]
  [...]
  Call Trace:
   <IRQ>
   dev_hard_start_xmit+0xd7/0x230
   sch_direct_xmit+0x9f/0x360
   __dev_queue_xmit+0x890/0xa40
  [...]

Fixes: c308dfd1

 ("sfc: fix wrong tx channel offset with efx_separate_tx_channels")
Reported-by: Tianhao Zhao <tizhao@redhat.com>
Signed-off-by: Íñigo Huguet <ihuguet@redhat.com>
Acked-by: Edward Cree <ecree.xilinx@gmail.com>
Link: https://lore.kernel.org/r/20220914103648.16902-1-ihuguet@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Luiz Augusto von Dentz says:

====================
bluetooth pull request for net:

 - Fix HCIGETDEVINFO regression

* tag 'for-net-2022-09-09' of git://git.kernel.org/pub/scm/linux/kernel/git/bluetooth/bluetooth:
  Bluetooth: Fix HCIGETDEVINFO regression
====================

Link: https://lore.kernel.org/r/20220909201642.3810565-1-luiz.dentz@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Disable page_pool/XDP support for MT7621 SoC in order fix a regression
introduce adding XDP for MT7986 SoC. There is no a real use case for XDP
on MT7621 since it is a low-end cpu. Moreover this patch reduces the
memory footprint.

Tested-by: Sergio Paracuellos <sergio.paracuellos@gmail.com>
Tested-by: Arınç ÜNAL <arinc.unal@arinc9.com>
Fixes: 23233e57

 ("net: ethernet: mtk_eth_soc: rely on page_pool for single page buffers")
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Link: https://lore.kernel.org/r/2bf31e27b888c43228b0d84dd2ef5033338269e2.1663074002.git.lorenzo@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Per GDMA spec, rmb is necessary after checking owner_bits, before
reading EQ or CQ entries.

Add rmb in these two places to comply with the specs.

Cc: stable@vger.kernel.org
Fixes: ca9c54d2

 ("net: mana: Add a driver for Microsoft Azure Network Adapter (MANA)")
Reported-by: Sinan Kaya <Sinan.Kaya@microsoft.com>
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: Dexuan Cui <decui@microsoft.com>
Link: https://lore.kernel.org/r/1662928805-15861-1-git-send-email-haiyangz@microsoft.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Updating active developers.

Signed-off-by: Jeroen de Borst <jeroendb@google.com>
Link: https://lore.kernel.org/r/20220913185319.1061909-1-jeroendb@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

The hwstats debugfs files are only writeable, but they are created with
read and write permissions, causing certain selftests to fail [1].

Fix by creating the files with write permission only.

[1]
 # ./test_offload.py
 Test destruction of generic XDP...
 Traceback (most recent call last):
   File "/home/idosch/code/linux/tools/testing/selftests/bpf/./test_offload.py", line 810, in <module>
     simdev = NetdevSimDev()
 [...]
 Exception: Command failed: cat /sys/kernel/debug/netdevsim/netdevsim0//ports/0/dev/hwstats/l3/disable_ifindex

 cat: /sys/kernel/debug/netdevsim/netdevsim0//ports/0/dev/hwstats/l3/disable_ifindex: Invalid argument

Fixes: 1a6d7ae7

 ("netdevsim: Introduce support for L3 offload xstats")
Reported-by: Jie2x Zhou <jie2x.zhou@intel.com>
Tested-by: Jie2x Zhou <jie2x.zhou@intel.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Link: https://lore.kernel.org/r/20220909153830.3732504-1-idosch@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

The MDIO gateway (GW) lock in BlueField-2 GIGE logic is
set after read.  This patch adds logic to make sure the
lock is always cleared at the end of each MDIO transaction.

Fixes: f92e1869

 ("Add Mellanox BlueField Gigabit Ethernet driver")
Reviewed-by: Asmaa Mnebhi <asmaa@nvidia.com>
Signed-off-by: David Thompson <davthompson@nvidia.com>
Link: https://lore.kernel.org/r/20220902164247.19862-1-davthompson@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>