This reverts commit 91948309.

There is only when ip_options_get() return zero need to free.
It already called kfree() when return error.

Fixes: 91948309 ("ipv4: fix memory leaks in ip_cmsg_send() callers")
Signed-off-by: Yajun Deng <yajun.deng@linux.dev>
Signed-off-by: David S. Miller <davem@davemloft.net>

Michael Chan says:

====================
bnxt_en: Bug fixes.

The first patch fixes an error recovery regression just introduced
about a week ago.  The other two patches fix issues related to
freeing rings in the bnxt_close() path under error conditions.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

We recently changed the completion ring page arrays to be dynamically
allocated to better support the expanded range of ring depths.  The
cleanup path for this was not quite complete.  It might cause the
shutdown path to crash if we need to abort before the completion ring
arrays have been allocated and initialized.

Fix it by initializing the ring_mem->pg_arr to NULL after freeing the
completion ring page array.  Add a check in bnxt_free_ring() to skip
referencing the rmem->pg_arr if it is NULL.

Fixes: 03c74487 ("bnxt_en: Don't use static arrays for completion ring pages")
Reviewed-by: Andy Gospodarek <gospo@broadcom.com>
Reviewed-by: Edwin Peer <edwin.peer@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The call to bnxt_free_mem(..., false) in the bnxt_half_open_nic() error
path will deallocate ring descriptor memory via bnxt_free_?x_rings(),
but because irq_re_init is false, the ring info itself is not freed.

To simplify error paths, deallocation functions have generally been
written to be safe when called on unallocated memory. It should always
be safe to call dev_close(), which calls bnxt_free_skbs() a second time,
even in this semi- allocated ring state.

Calling bnxt_free_skbs() a second time with the rings already freed will
cause NULL pointer dereference.  Fix it by checking the rings are valid
before proceeding in bnxt_free_tx_skbs() and
bnxt_free_one_rx_ring_skbs().

Fixes: 975bc99a ("bnxt_en: Refactor bnxt_free_rx_skbs().")
Signed-off-by: Edwin Peer <edwin.peer@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The recent patch has introduced a regression by not reading the reset
count in the ERROR_RECOVERY async event handler.  We may have just
gone through a reset and the reset count has just incremented.  If
we don't update the reset count in the ERROR_RECOVERY event handler,
the health check timer will see that the reset count has changed and
will initiate an unintended reset.

Restore the unconditional update of the reset count in
bnxt_async_event_process() if error recovery watchdog is enabled.
Also, update the reset count at the end of the reset sequence to
make it even more robust.

Fixes: 1b2b9183 ("bnxt_en: Fix possible unintended driver initiated error recovery")
Reviewed-by: Edwin Peer <edwin.peer@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

As noted in the "Deprecated Interfaces, Language Features, Attributes,
and Conventions" documentation [1], size calculations (especially
multiplication) should not be performed in memory allocator (or similar)
function arguments due to the risk of them overflowing. This could lead
to values wrapping around and a smaller allocation being made than the
caller was expecting. Using those allocations could lead to linear
overflows of heap memory and other misbehaviors.

So, use the struct_size() helper to do the arithmetic instead of the
argument "size + count * size" in the kzalloc() function.

[1] https://www.kernel.org/doc/html/v5.14/process/deprecated.html#open-coded-arithmetic-in-allocator-argumentsSigned-off-by: Len Baker <len.baker@gmx.com>
Reviewed-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

If altname deletion of the short alternative name fails, the error
message printed is: "Failed to add short alternative name".
This is obviously a typo, as we are testing altname deletion.

Fix this using a proper error message.

Fixes: f95e6c9c ("selftest: net: add alternative names test")
Signed-off-by: Andrea Claudi <aclaudi@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

When the mdio legacy mapping is used the mii_bus priv registered by DSA
refer to the dsa switch struct instead of the qca8k_priv struct and
causes a kernel panic. Create dedicated function when the internal
dedicated mdio driver is used to properly handle the 2 different
implementation.

Fixes: 759bafb8 ("net: dsa: qca8k: add support for internal phy and internal mdio")
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The Synopsys Ethernet IP uses the CSR clock as a base clock for MDC.
The divisor used is set in the MAC_MDIO_Address register field CR
(Clock Rate)

The divisor is there to change the CSR clock into a clock that falls
below the IEEE 802.3 specified max frequency of 2.5MHz.

If the CSR clock is 300MHz, the code falls back to using the reset
value in the MAC_MDIO_Address register, as described in the comment
above this code.

However, 300MHz is actually an allowed value and the proper divider
can be estimated quite easily (it's just 1Hz difference!)

A CSR frequency of 300MHz with the maximum clock rate value of 0x5
(STMMAC_CSR_250_300M, a divisor of 124) gives somewhere around
~2.42MHz which is below the IEEE 802.3 specified maximum.

For the ARTPEC-8 SoC, the CSR clock is this problematic 300MHz,
and unfortunately, the reset-value of the MAC_MDIO_Address CR field
is 0x0.

This leads to a clock rate of zero and a divisor of 42, and gives an
MDC frequency of ~7.14MHz.

Allow CSR clock of 300MHz by making the comparison inclusive.
Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The variable ret is being initialized with a value that is never read, it
is being updated later on. The assignment is redundant and can be removed.

Addresses-Coverity: ("Unused value")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Handle MFW (management FW) error response in order to avoid a crash
during recovery flows.

Changes from v1:
- Add "Fixes tag".

Fixes: tag 5e7ba042 ("qed: Fix reading stale configuration information")
Signed-off-by: Ariel Elior <aelior@marvell.com>
Signed-off-by: Shai Malin <smalin@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add pr_fmt macro to spell out the source of messages in prefix.

Before this patch:

  packet size is too long (1543 > 1518)

With this patch:

  af_packet: packet size is too long (1543 > 1518)
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>

A number of users have reported that they were not able to get the PHY
to successfully link up, especially after commit c36757eb ("net:
phy: consider AN_RESTART status when reading link status") where we
stopped reading just BMSR, but we also read BMCR to determine the link
status.

Andrius at NetBSD did a wonderful job at debugging the problem
and found out that the MDIO bus clock frequency would be incorrectly set
back to its default value which would prevent the MDIO bus controller
from reading PHY registers properly. Back when we only read BMSR, if we
read all 1s, we could falsely indicate a link status, though in general
there is a cable plugged in, so this went unnoticed. After a second read
of BMCR was added, a wrong read will lead to the inability to determine
a link UP condition which is when it started to be visibly broken, even
if it was long before that.

The fix consists in restoring the value of the MD_CSR register that was
set prior to the MAC reset.

Link: http://gnats.netbsd.org/cgi-bin/query-pr-single.pl?number=53494
Fixes: 90f750a8 ("r6040: consolidate MAC reset to its own function")
Reported-by: Andrius V <vezhlys@gmail.com>
Reported-by: Darek Strugacz <darek.strugacz@op.pl>
Tested-by: Darek Strugacz <darek.strugacz@op.pl>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

There are two cases where the current PF does not support RDMA
functionality.  The first is if the NVM loaded on the device is set
to not support RDMA (common_caps.rdma is false).  The second is if
the kernel bonding driver has included the current PF in an active
link aggregate.

When the driver has determined that this PF does not support RDMA, then
auxiliary devices should not be created on the auxiliary bus.  Without
a device on the auxiliary bus, even if the irdma driver is present, there
will be no RDMA activity attempted on this PF.

Currently, in the reset flow, an attempt to create auxiliary devices is
performed without regard to the ability of the PF.  There needs to be a
check in ice_aux_plug_dev (as the central point that creates auxiliary
devices) to see if the PF is in a state to support the functionality.

When disabling and re-enabling RDMA due to the inclusion/removal of the PF
in a link aggregate, we also need to set/clear the bit which controls
auxiliary device creation so that a reset recovery in a link aggregate
situation doesn't try to create auxiliary devices when it shouldn't.

Fixes: f9f5301e ("ice: Register auxiliary device to provide RDMA")
Reported-by: Yongxin Liu <yongxin.liu@windriver.com>
Signed-off-by: Dave Ertman <david.m.ertman@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Building alpha:allmodconfig results in the following error.

drivers/net/ethernet/amd/ni65.c: In function 'ni65_stop_start':
drivers/net/ethernet/amd/ni65.c:751:37: error:
	cast from pointer to integer of different size
		buffer[i] = (u32) isa_bus_to_virt(tmdp->u.buffer);

'buffer[]' is declared as unsigned long, so replace the typecast to u32
with a typecast to unsigned long to fix the problem.

Cc: Arnd Bergmann <arnd@kernel.org>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

Íñigo Huguet says:

====================
sfc: fallback for lack of xdp tx queues

If there are not enough hardware resources to allocate one tx queue per
CPU for XDP, XDP_TX and XDP_REDIRECT actions were unavailable, and using
them resulted each time with the packet being drop and this message in
the logs: XDP TX failed (-22)

These patches implement 2 fallback solutions for 2 different situations
that might happen:
1. There are not enough free resources for all the tx queues, but there
   are some free resources available
2. There are not enough free resources at all for tx queues.

Both solutions are based in sharing tx queues, using __netif_tx_lock for
synchronization. In the second case, as there are not XDP TX queues to
share, network stack queues are used instead, but since we're taking
__netif_tx_lock, concurrent access to the queues is correctly protected.

The solution for this second case might affect performance both of XDP
traffic and normal traffice due to lock contention if both are used
intensively. That's why I call it a "last resort" fallback: it's not a
desirable situation, but at least we have XDP TX working.

Some tests has shown good results and indicate that the non-fallback
case is not being damaged by this changes. They are also promising for
the fallback cases. This is the test:
1. From another machine, send high amount of packets with pktgen, script
   samples/pktgen/pktgen_sample04_many_flows.sh
2. In the tested machine, run samples/bpf/xdp_rxq_info with arguments
   "-a XDP_TX --swapmac" and see the results
3. In the tested machine, run also pktgen_sample04 to create high TX
   normal traffic, and see how xdp_rxq_info results vary

Note that this test doesn't check the worst situations for the fallback
solutions because XDP_TX will only be executed from the same CPUs that
are processed by sfc, and not from every CPU in the system, so the
performance drop due to the highest locking contention doesn't happen.
I'd like to test that, as well, but I don't have access right now to a
proper environment.

Test results:

Without doing TX:
Before changes: ~2,900,000 pps
After changes, 1 queues/core: ~2,900,000 pps
After changes, 2 queues/core: ~2,900,000 pps
After changes, 8 queues/core: ~2,900,000 pps
After changes, borrowing from network stack: ~2,900,000 pps

With multiflow TX at the same time:
Before changes: ~1,700,000 - 2,900,000 pps
After changes, 1 queues/core: ~1,700,000 - 2,900,000 pps
After changes, 2 queues/core: ~1,700,000 pps
After changes, 8 queues/core: ~1,700,000 pps
After changes, borrowing from network stack: 1,150,000 pps

Sporadic "XDP TX failed (-5)" warnings are shown when running xdp program
and pktgen simultaneously. This was expected because XDP doesn't have any
buffering system if the NIC is under very high pressure. Thousands of
these warnings are shown in the case of borrowing net stack queues. As I
said before, this was also expected.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Previous patch addressed the situation of having some free resources for
xdp tx but not enough for one tx queue per CPU. This patch address the
worst case of not having resources at all for xdp tx.

Instead of using queues dedicated to xdp, normal queues used by network
stack are shared for both cases, using __netif_tx_lock for
synchronization. Also queue stop/restart must be considered in the xdp
path to avoid freezing the queue.

This is not the ideal situation we might want to be, and a performance
penalty is expected both for normal and xdp traffic, but at least XDP
will work in all possible situations (with a warning in the logs),
improving a bit the pain of not knowing in what situations we can use it
and in what situations we cannot.
Signed-off-by: Íñigo Huguet <ihuguet@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

If there are not enough resources to allocate one TX queue per core for
XDP TX it was completely disabled.

This patch implements a fallback solution for sharing the available
queues using __netif_tx_lock for synchronization. In the normal case that
there is one TX queue per CPU, no locking is done, as it was before.

With this fallback solution, XDP TX will work in much more cases that
were failing, specially in machines with many CPUs. It's hard for XDP
users to know what features are supported across different NICs and
configurations, so they will benefit on having wider support.
Signed-off-by: Íñigo Huguet <ihuguet@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Use __maybe_unused for noirq_suspend()/noirq_resume() hooks to avoid
build warning with !CONFIG_PM_SLEEP:

>> drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c:796:12: error: 'stmmac_pltfr_noirq_resume' defined but not used [-Werror=unused-function]
     796 | static int stmmac_pltfr_noirq_resume(struct device *dev)
         |            ^~~~~~~~~~~~~~~~~~~~~~~~~
>> drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c:775:12: error: 'stmmac_pltfr_noirq_suspend' defined but not used [-Werror=unused-function]
     775 | static int stmmac_pltfr_noirq_suspend(struct device *dev)
         |            ^~~~~~~~~~~~~~~~~~~~~~~~~~
   cc1: all warnings being treated as errors

Fixes: 276aae37 ("net: stmmac: fix system hang caused by eee_ctrl_timer during suspend/resume")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Joakim Zhang <qiangqing.zhang@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The reference count leak issue may take place in an error handling
path. If both conditions of tunnel->version == L2TP_HDR_VER_3 and the
return value of l2tp_v3_ensure_opt_in_linear is nonzero, the function
would directly jump to label invalid, without decrementing the reference
count of the l2tp_session object session increased earlier by
l2tp_tunnel_get_session(). This may result in refcount leaks.

Fix this issue by decrease the reference count before jumping to the
label invalid.

Fixes: 4522a70d ("l2tp: fix reading optional fields of L2TPv3")
Signed-off-by: Xiyu Yang <xiyuyang19@fudan.edu.cn>
Signed-off-by: Xin Xiong <xiongx18@fudan.edu.cn>
Signed-off-by: Xin Tan <tanxin.ctf@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

syzbot reported another data-race in af_unix [1]

Lets change __skb_insert() to use WRITE_ONCE() when changing
skb head qlen.

Also, change unix_dgram_poll() to use lockless version
of unix_recvq_full()

It is verry possible we can switch all/most unix_recvq_full()
to the lockless version, this will be done in a future kernel version.

[1] HEAD commit: 8596e589

BUG: KCSAN: data-race in skb_queue_tail / unix_dgram_poll

write to 0xffff88814eeb24e0 of 4 bytes by task 25815 on cpu 0:
 __skb_insert include/linux/skbuff.h:1938 [inline]
 __skb_queue_before include/linux/skbuff.h:2043 [inline]
 __skb_queue_tail include/linux/skbuff.h:2076 [inline]
 skb_queue_tail+0x80/0xa0 net/core/skbuff.c:3264
 unix_dgram_sendmsg+0xff2/0x1600 net/unix/af_unix.c:1850
 sock_sendmsg_nosec net/socket.c:703 [inline]
 sock_sendmsg net/socket.c:723 [inline]
 ____sys_sendmsg+0x360/0x4d0 net/socket.c:2392
 ___sys_sendmsg net/socket.c:2446 [inline]
 __sys_sendmmsg+0x315/0x4b0 net/socket.c:2532
 __do_sys_sendmmsg net/socket.c:2561 [inline]
 __se_sys_sendmmsg net/socket.c:2558 [inline]
 __x64_sys_sendmmsg+0x53/0x60 net/socket.c:2558
 do_syscall_x64 arch/x86/entry/common.c:50 [inline]
 do_syscall_64+0x3d/0x90 arch/x86/entry/common.c:80
 entry_SYSCALL_64_after_hwframe+0x44/0xae

read to 0xffff88814eeb24e0 of 4 bytes by task 25834 on cpu 1:
 skb_queue_len include/linux/skbuff.h:1869 [inline]
 unix_recvq_full net/unix/af_unix.c:194 [inline]
 unix_dgram_poll+0x2bc/0x3e0 net/unix/af_unix.c:2777
 sock_poll+0x23e/0x260 net/socket.c:1288
 vfs_poll include/linux/poll.h:90 [inline]
 ep_item_poll fs/eventpoll.c:846 [inline]
 ep_send_events fs/eventpoll.c:1683 [inline]
 ep_poll fs/eventpoll.c:1798 [inline]
 do_epoll_wait+0x6ad/0xf00 fs/eventpoll.c:2226
 __do_sys_epoll_wait fs/eventpoll.c:2238 [inline]
 __se_sys_epoll_wait fs/eventpoll.c:2233 [inline]
 __x64_sys_epoll_wait+0xf6/0x120 fs/eventpoll.c:2233
 do_syscall_x64 arch/x86/entry/common.c:50 [inline]
 do_syscall_64+0x3d/0x90 arch/x86/entry/common.c:80
 entry_SYSCALL_64_after_hwframe+0x44/0xae

value changed: 0x0000001b -> 0x00000001

Reported by Kernel Concurrency Sanitizer on:
CPU: 1 PID: 25834 Comm: syz-executor.1 Tainted: G        W         5.14.0-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011

Fixes: 86b18aaa ("skbuff: fix a data race in skb_queue_len()")
Cc: Qian Cai <cai@lca.pw>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

plat_dev->dev->platform_data is released by platform_device_unregister(),
use of pclk and hclk is a use-after-free. Since device unregister won't
need a clk device we adjust the function call sequence to fix this issue.

[   31.261225] BUG: KASAN: use-after-free in macb_remove+0x77/0xc6 [macb_pci]
[   31.275563] Freed by task 306:
[   30.276782]  platform_device_release+0x25/0x80
Suggested-by: Nicolas Ferre <Nicolas.Ferre@microchip.com>
Signed-off-by: Tong Zhang <ztong0001@gmail.com>
Acked-by: Nicolas Ferre <nicolas.ferre@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

If a failover occurs before a login response is received, the login
response buffer maybe undefined. Check that there was no failover
before accessing the login response buffer.

Fixes: 032c5e82 ("Driver for IBM System i/p VNIC protocol")
Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

If the sendmsg() call in vhost_tx_batch() fails, both the 'batched_xdp'
and 'done_idx' indexes are left unchanged. If such failure happens
when batched_xdp == VHOST_NET_BATCH, the next call to
vhost_net_build_xdp() will access and write memory outside the xdp
buffers area.

Since sendmsg() can only error with EBADFD, this change addresses the
issue explicitly freeing the XDP buffers batch on error.

Fixes: 0a0be13b ("vhost_net: batch submitting XDP buffers to underlayer sockets")
Suggested-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

commit 5f585913 ("net: stmmac: delete the eee_ctrl_timer after
napi disabled"), this patch tries to fix system hang caused by eee_ctrl_timer,
unfortunately, it only can resolve it for system reboot stress test. System
hang also can be reproduced easily during system suspend/resume stess test
when mount NFS on i.MX8MP EVK board.

In stmmac driver, eee feature is combined to phylink framework. When do
system suspend, phylink_stop() would queue delayed work, it invokes
stmmac_mac_link_down(), where to deactivate eee_ctrl_timer synchronizly.
In above commit, try to fix issue by deactivating eee_ctrl_timer obviously,
but it is not enough. Looking into eee_ctrl_timer expire callback
stmmac_eee_ctrl_timer(), it could enable hareware eee mode again. What is
unexpected is that LPI interrupt (MAC_Interrupt_Enable.LPIEN bit) is always
asserted. This interrupt has chance to be issued when LPI state entry/exit
from the MAC, and at that time, clock could have been already disabled.
The result is that system hang when driver try to touch register from
interrupt handler.

The reason why above commit can fix system hang issue in stmmac_release()
is that, deactivate eee_ctrl_timer not just after napi disabled, further
after irq freed.

In conclusion, hardware would generate LPI interrupt when clock has been
disabled during suspend or resume, since hardware is in eee mode and LPI
interrupt enabled.

Interrupts from MAC, MTL and DMA level are enabled and never been disabled
when system suspend, so postpone clocks management from suspend stage to
noirq suspend stage should be more safe.

Fixes: 5f585913 ("net: stmmac: delete the eee_ctrl_timer after napi disabled")
Signed-off-by: Joakim Zhang <qiangqing.zhang@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

There is an off-by-one problem in ipa_table_init_add(), when
initializing filter tables.

In that function, the number of filter table entries is determined
based on the number of set bits in the filter map.  However that
count does *not* include the extra "slot" in the filter table that
holds the filter map itself.  Meanwhile, ipa_table_addr() *does*
include the filter map in the memory it returns, but because the
count it's provided doesn't include it, it includes one too few
table entries.

Fix this by including the extra slot for the filter map in the count
computed in ipa_table_init_add().

Note: ipa_filter_reset_table() does not have this problem; it resets
filter table entries one by one, but does not overwrite the filter
bitmap.

Fixes: 2b9feef2 ("soc: qcom: ipa: filter and routing tables")
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Currently changes to the advertising state via ethtool do not cause any
reselection of the configured interface mode after the SFP is already
inserted and initially configured.

While it is not typical to change the advertised link modes for an
interface using an SFP in certain use cases it is desirable. In the case
of a SFP port that is capable of handling both SFP and SFP+ modules it
will automatically select between 1G and 10G modes depending on the
supported mode of the SFP. However if the SFP module is capable of
working in multiple modes (e.g. a SFP+ DAC that can operate at 1G or
10G), one end of the cable may be attached to a SFP 1000base-x port thus
the SFP+ end must be manually configured to the 1000base-x mode in order
for the link to be established.

This change causes the ethtool setting of advertised mode changes to
reselect the interface mode so that the link can be established.
Additionally when a module is inserted the advertising mode is reset to
match the supported modes of the module.
Signed-off-by: Nathan Rossi <nathan.rossi@digi.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Geert noticed a warning on MIPS TX49xx, Atari and presuambly other
platforms when the driver is built-in but NETDEV_LEGACY_INIT is
disabled:

drivers/net/ethernet/8390/ne.c:909:20: warning: ‘ne_add_devices’ defined but not used [-Wunused-function]

Merge the two module init functions into a single one with an
IS_ENABLED() check to replace the incorrect #ifdef.

Fixes: 4228c394 ("make legacy ISA probe optional")
Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Geert Uytterhoeven <geert@linux-m68k.org>
Tested-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Saeed Mahameed says:

====================
mlx5 fixes 2021-09-07

This series introduces some fixes to mlx5 driver.
Please pull and let me know if there is any problem.

Included here, a patch which solves a build warning reported on
linux-kernel mailing list [1]:
Fix commit ("net/mlx5: Bridge, fix uninitialized variable usage")

I hope this series can make it to rc1.

[1] https://www.spinics.net/lists/netdev/msg765481.html
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

If a failover occurs before a login response is received, the login
response buffer maybe undefined. Check that there was no failover
before accessing the login response buffer.
Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The kernel test robot reports:

  [  843.509974][  T345] =============================
  [  843.524220][  T345] WARNING: suspicious RCU usage
  [  843.538791][  T345] 5.14.0-rc2-00606-g889b7da2 #1 Not tainted
  [  843.553617][  T345] -----------------------------
  [  843.567412][  T345] net/mctp/route.c:310 RCU-list traversed in non-reader section!!

- we're missing the rcu read lock acquire around the destruction path.

This change adds the acquire/release - the path is already atomic, and
we're using the _rcu list iterators.
Reported-by: kernel test robot <oliver.sang@intel.com>
Signed-off-by: Jeremy Kerr <jk@codeconstruct.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

Commit 2677d206 ("dccp: don't free ccid2_hc_tx_sock ...") fixed
a UAF but reintroduced CVE-2017-6074.

When the sock is cloned, two dccps_hc_tx_ccid will reference to the
same ccid. So one can free the ccid object twice from two socks after
cloning.

This issue was found by "Hadar Manor" as well and assigned with
CVE-2020-16119, which was fixed in Ubuntu's kernel. So here I port
the patch from Ubuntu to fix it.

The patch prevents cloned socks from referencing the same ccid.

Fixes: 2677d206 ("dccp: don't free ccid2_hc_tx_sock ...")
Signed-off-by: Zhenpeng Lin <zplin@psu.edu>
Signed-off-by: David S. Miller <davem@davemloft.net>

The D1 SoC contains EMAC hardware which is compatible with the A64 EMAC.
Add the new compatible string, with the A64 as a fallback.
Signed-off-by: Samuel Holland <samuel@sholland.org>
Acked-by: Maxime Ripard <maxime@cerno.tech>
Signed-off-by: David S. Miller <davem@davemloft.net>

When activating the PTP-RQ, redirect the RQT from drop-RQ to PTP-RQ.
Use mlx5e_channels_get_ptp_rqn to retrieve the rqn. This helper returns
a boolean (not status), hence caller should consider return value 0 as a
fail. Change the caller interpretation of the return value.

Fixes: 43ec0f41 ("net/mlx5e: Hide all implementation details of mlx5e_rx_res")
Signed-off-by: Aya Levin <ayal@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

Some profiles of the driver don't support a dedicated PTP-RQ, hence can't
support HW TS and CQE compression simultaneously. When HW TS is enabled
the COE compression is disabled, and should be restored when the HW TS
is turned off. Add rx_filter as an input to modifying CQE compression to
enforce this restriction.

Fixes: 256f79d1 ("net/mlx5e: Fix HW TS with CQE compression according to profile")
Signed-off-by: Aya Levin <ayal@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

Fixes the below flow of sleeping in atomic context by releasing
the RCU lock before calling to free_match_list.

build_match_list() <- disables preempt
-> free_match_list()
   -> tree_put_node()
      -> down_write_ref_node() <- take write lock

Fixes: 693c6883 ("net/mlx5: Add hash table for flow groups in flow table")
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Maor Gottlieb <maorg@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

Handle error flow on mlx5_core_alloc_pd() failure,
read_fw_strings_work must be canceled.

Fixes: c71ad41c ("net/mlx5: FW tracer, events handling")
Reported-by: Pavel Machek (CIP) <pavel@denx.de>
Suggested-by: Pavel Machek (CIP) <pavel@denx.de>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Reviewed-by: Aya Levin <ayal@nvidia.com>

In NICs that don't support LAG, the LAG control structure won't be
allocated. If it wasn't allocated it means LAG doesn't exists and can be
skipped.

Fixes: cac1eb2c ("net/mlx5: Lag, properly lock eswitch if needed")
Signed-off-by: Mark Bloch <mbloch@nvidia.com>
Reviewed-by: Maor Gottlieb <maorg@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

RDMA auxdev parameter registration was skipped for eswitch manager PCI PF.
Due to this when devlink parameter is read, it reads as false in below
code flow.

$ devlink dev reload pci/0000:06:00.0
  devlink_reload()
    mlx5_load_one()
      mlx5_attach_device()
        is_ib_enabled()
          devlink_param_driverinit_value_get()

Due to this, is_ib_enabled() returns false for the RDMA auxiliary
device. This results into a skipping RDMA auxiliary device creation on
reload.

There is no need to check for eswitch manager capability to support RDMA
auxiliary device. Hence, fix it by skipping eswitch manager capability.

Fixes: 87158ced ("net/mlx5: Support enable_rdma devlink dev param")
Signed-off-by: Parav Pandit <parav@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

In some conditions variable 'err' is not assigned with value in
mlx5_esw_bridge_port_obj_attr_set() and mlx5_esw_bridge_port_changeupper()
functions after recent changes to support LAG. Initialize the variable with
zero value in both cases.
Reported-by: Colin King <colin.king@canonical.com>
Reported-by: Tim Gardner <tim.gardner@canonical.com>
Reported-by: Naresh Kamboju <naresh.kamboju@linaro.org>
CC: linux-kernel@vger.kernel.org
Fixes: ff9b7521 ("net/mlx5: Bridge, support LAG")
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>