If dsa_tag_8021q_setup() fails, for example due to the inability of the
device to install a VLAN, the tag_8021q context of the switch will leak.
Make sure it is freed on the error path.
Fixes: ee3f0b58ad79 ("net: dsa: tag_8021q: absorb dsa_8021q_setup into dsa_tag_8021q_{,un}register") Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Link: https://lore.kernel.org/r/20221209235242.480344-1-vladimir.oltean@nxp.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Whenever trying to load XDP prog on downed interface, function i40e_xdp
was passing vsi->rx_buf_len field to i40e_xdp_setup() which was equal 0.
i40e_open() calls i40e_vsi_configure_rx() which configures that field,
but that only happens when interface is up. When it is down, i40e_open()
is not being called, thus vsi->rx_buf_len is not set.
Solution for this is calculate buffer length in newly created
function - i40e_calculate_vsi_rx_buf_len() that return actual buffer
length. Buffer length is being calculated based on the same rules
applied previously in i40e_vsi_configure_rx() function.
Fixes: 34549e6d5bc3 ("i40e: Log error for oversized MTU on device") Fixes: 30067ecaebfe ("i40e: add XDP support for pass and drop actions") Signed-off-by: Bartosz Staszewski <bartoszx.staszewski@intel.com> Signed-off-by: Mateusz Palczewski <mateusz.palczewski@intel.com> Tested-by: Shwetha Nagaraju <Shwetha.nagaraju@intel.com> Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Reviewed-by: Saeed Mahameed <saeed@kernel.com> Link: https://lore.kernel.org/r/20221209185411.2519898-1-anthony.l.nguyen@intel.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
When the MAC is connected to a 10 Mb/s PHY and the PTP clock is derived
from the MAC reference clock (default), the clk_ptp_rate becomes too
small and the calculated sub second increment becomes 0 when computed by
the stmmac_config_sub_second_increment() function within
stmmac_init_tstamp_counter().
Therefore, the subsequent div_u64 in stmmac_init_tstamp_counter()
operation triggers a divide by 0 exception as shown below.
[ 95.062067] socfpga-dwmac ff700000.ethernet eth0: Register MEM_TYPE_PAGE_POOL RxQ-0
[ 95.076440] socfpga-dwmac ff700000.ethernet eth0: PHY [stmmac-0:08] driver [NCN26000] (irq=49)
[ 95.095964] dwmac1000: Master AXI performs any burst length
[ 95.101588] socfpga-dwmac ff700000.ethernet eth0: No Safety Features support found
[ 95.109428] Division by zero in kernel.
[ 95.113447] CPU: 0 PID: 239 Comm: ifconfig Not tainted 6.1.0-rc7-centurion3-1.0.3.0-01574-gb624218205b7-dirty #77
[ 95.123686] Hardware name: Altera SOCFPGA
[ 95.127695] unwind_backtrace from show_stack+0x10/0x14
[ 95.132938] show_stack from dump_stack_lvl+0x40/0x4c
[ 95.137992] dump_stack_lvl from Ldiv0+0x8/0x10
[ 95.142527] Ldiv0 from __aeabi_uidivmod+0x8/0x18
[ 95.147232] __aeabi_uidivmod from div_u64_rem+0x1c/0x40
[ 95.152552] div_u64_rem from stmmac_init_tstamp_counter+0xd0/0x164
[ 95.158826] stmmac_init_tstamp_counter from stmmac_hw_setup+0x430/0xf00
[ 95.165533] stmmac_hw_setup from __stmmac_open+0x214/0x2d4
[ 95.171117] __stmmac_open from stmmac_open+0x30/0x44
[ 95.176182] stmmac_open from __dev_open+0x11c/0x134
[ 95.181172] __dev_open from __dev_change_flags+0x168/0x17c
[ 95.186750] __dev_change_flags from dev_change_flags+0x14/0x50
[ 95.192662] dev_change_flags from devinet_ioctl+0x2b4/0x604
[ 95.198321] devinet_ioctl from inet_ioctl+0x1ec/0x214
[ 95.203462] inet_ioctl from sock_ioctl+0x14c/0x3c4
[ 95.208354] sock_ioctl from vfs_ioctl+0x20/0x38
[ 95.212984] vfs_ioctl from sys_ioctl+0x250/0x844
[ 95.217691] sys_ioctl from ret_fast_syscall+0x0/0x4c
[ 95.222743] Exception stack(0xd0ee1fa8 to 0xd0ee1ff0)
[ 95.227790] 1fa0: 00574c4fbe9aeca40000000300008914be9aeca4be9aec50
[ 95.235945] 1fc0: 00574c4fbe9aeca40059f07800000036be9aee8cbe9aef7a0000001500000000
[ 95.244096] 1fe0: 005a01f0be9aec38004d7484b6e67d74
In mcs_register_interrupts(), a call to request_irq() is not balanced by a
corresponding free_irq(), neither in the error handling path, nor in the
remove function.
It is not allowed to call kfree_skb() from hardware interrupt
context or with interrupts being disabled. So replace kfree_skb()
with dev_kfree_skb_irq() under spin_lock_irqsave().
Fixes: 11fa3ef31d3c ("Bluetooth: RFCOMM: Replace use of memcpy_from_msg with bt_skb_sendmmsg") Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
It is not allowed to call kfree_skb() from hardware interrupt
context or with interrupts being disabled. So replace kfree_skb()
with dev_kfree_skb_irq() under spin_lock_irqsave().
Fixes: e56c2d4d3833 ("Bluetooth: Add request cmd_complete and cmd_status functions") Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
It is not allowed to call kfree_skb() from hardware interrupt
context or with interrupts being disabled. So replace kfree_skb()
with dev_kfree_skb_irq() under spin_lock_irqsave().
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
It is not allowed to call kfree_skb() from hardware interrupt
context or with interrupts being disabled. So replace kfree_skb()
with dev_kfree_skb_irq() under spin_lock_irqsave().
Fixes: dae41d0b3727 ("Bluetooth: Fix/implement Three-wire reliable packet sending") Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
It is not allowed to call kfree_skb() from hardware interrupt
context or with interrupts being disabled. So replace kfree_skb()
with dev_kfree_skb_irq() under spin_lock_irqsave().
Fixes: 9a465d291061 ("[Bluetooth] Add UART driver for Texas Instruments' BRF63xx chips") Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
It is not allowed to call kfree_skb() from hardware interrupt
context or with interrupts being disabled. So replace kfree_skb()
with dev_kfree_skb_irq() under spin_lock_irqsave().
Fixes: 5aba0b2df148 ("Bluetooth: hciuart: Add support QCA chipset for UART") Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
It is not allowed to call kfree_skb() from hardware interrupt
context or with interrupts being disabled. So replace kfree_skb()
with dev_kfree_skb_irq() under spin_lock_irqsave().
Fixes: 37763e7882bf ("Bluetooth: btusb: Implement driver internal packet reassembly") Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
skb allocated by __hci_cmd_sync would not be used whether in checking
for supported iBT hardware variants or after, we should free it in all
error branches, this patch makes the case read version failed or default
error case free skb before return.
Fixes: 08409d4d4d71 ("Bluetooth: btintel: Fix the legacy bootloader returns tlv based version") Fixes: c292d4b696e8 ("Bluetooth: btintel: Refactoring setup routine for bootloader devices") Signed-off-by: Wang ShaoBo <bobo.shaobowang@huawei.com> Reviewed-by: Tedd Ho-Jeong An <tedd.an@intel.com> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
When validating the parameter length for MGMT_OP_ADD_EXT_ADV_PARAMS
command, use the correct op code in error status report:
was MGMT_OP_ADD_ADVERTISING, changed to MGMT_OP_ADD_EXT_ADV_PARAMS.
Fixes: 1a734d70b43ac ("Bluetooth: Break add adv into two mgmt commands") Signed-off-by: Inga Stotland <inga.stotland@intel.com> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
If hci_register_suspend_notifier() returns error, the hdev and rfkill
are leaked. We could disregard the error and print a warning message
instead to avoid leaks, as it just means we won't be handing suspend
requests.
Fixes: be6210b56614 ("Bluetooth: Handle PM_SUSPEND_PREPARE and PM_POST_SUSPEND") Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Recently, a customer reported that from their container whose
net namespace is different to the host's init_net, they can't set
the container's net.sctp.rto_max to any value smaller than
init_net.sctp.rto_min.
For instance,
Host:
sudo sysctl net.sctp.rto_min
net.sctp.rto_min = 1000
This is caused by the check made from this'commit d4b545c600c5
("sctp: add check rto_min and rto_max in sysctl")'
When validating the input value, it's always referring the boundary
value set for the init_net namespace.
Having container's rto_max smaller than host's init_net.sctp.rto_min
does make sense. Consider that the rto between two containers on the
same host is very likely smaller than it for two hosts.
So to fix this problem, as suggested by Marcelo, this patch makes the
extra pointers of rto_min, rto_max, pf_retrans, and ps_retrans point
to the corresponding variables from the newly created net namespace while
the new net namespace is being registered in sctp_sysctl_net_register.
Fixes: d4b545c600c5 ("sctp: add check rto_min and rto_max in sysctl") Reviewed-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com> Reviewed-by: Jakub Kicinski <kuba@kernel.org> Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com> Signed-off-by: Firo Yang <firo.yang@suse.com> Link: https://lore.kernel.org/r/20221209054854.23889-1-firo.yang@suse.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
TX/RX callback handlers (ntb_netdev_tx_handler(),
ntb_netdev_rx_handler()) can be called in interrupt
context via the DMA framework when the respective
DMA operations have completed. As such, any calls
by these routines to free skb's, should use the
interrupt context safe dev_kfree_skb_any() function.
Previously, these callback handlers would call the
interrupt unsafe version of dev_kfree_skb(). This has
not presented an issue on Intel IOAT DMA engines as
that driver utilizes tasklets rather than a hard
interrupt handler, like the AMD PTDMA DMA driver.
On AMD systems, a kernel WARNING message is
encountered, which is being issued from
skb_release_head_state() due to in_hardirq()
being true.
Besides the user visible WARNING from the kernel,
the other symptom of this bug was that TCP/IP performance
across the ntb_netdev interface was very poor, i.e.
approximately an order of magnitude below what was
expected. With the repair to use dev_kfree_skb_any(),
kernel WARNINGs from skb_release_head_state() ceased
and TCP/IP performance, as measured by iperf, was on
par with expected results, approximately 20 Gb/s on
AMD Milan based server. Note that this performance
is comparable with Intel based servers.
Fixes: 4d918621805ef ("ntb_netdev: correct skb leak") Fixes: 37b94e0273200 ("net: Add support for NTB virtual ethernet device") Signed-off-by: Eric Pilmore <epilmore@gigaio.com> Reviewed-by: Dave Jiang <dave.jiang@intel.com> Link: https://lore.kernel.org/r/20221209000659.8318-1-epilmore@gigaio.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
This patch fixes an issue where a read failure of a port statistic counter
will return unknown results. While it is highly unlikely the read will
ever fail, it is much cleaner to return a zero for the stat count.
Fixes: 3bbb5663f4fc ("net: dsa: add new DSA switch driver for the SMSC-LAN9303") Signed-off-by: Jerry Ray <jerry.ray@microchip.com> Reviewed-by: Vladimir Oltean <olteanv@gmail.com> Reviewed-by: Florian Fainelli <f.fainelli@gmail.com> Link: https://lore.kernel.org/r/20221209153502.7429-1-jerry.ray@microchip.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
TCAN4X5X_ERROR_STATUS is not a status register that needs clearing
during interrupt handling. Instead this is a masking register that masks
error interrupts. Writing TCAN4X5X_CLEAR_ALL_INT to this register
effectively masks everything.
Rename the register and mask all error interrupts only once by writing
to the register in tcan4x5x_init.
Fixes: b931004016bd ("can: tcan4x5x: Add tcan4x5x driver to the kernel") Signed-off-by: Markus Schneider-Pargmann <msp@baylibre.com> Link: https://lore.kernel.org/all/20221206115728.1056014-10-msp@baylibre.com Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de> Signed-off-by: Sasha Levin <sashal@kernel.org>
When we try to access the mcan message ram addresses during the probe,
hclk is gated by any other drivers or disabled, because of that probe
gets failed.
Move the mram init functionality to mcan chip config called by
m_can_start from mcan open function, by that time clocks are
enabled.
Suggested-by: Marc Kleine-Budde <mkl@pengutronix.de> Signed-off-by: Vivek Yadav <vivek.2311@samsung.com> Link: https://lore.kernel.org/all/20221207100632.96200-2-vivek.2311@samsung.com Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Stable-dep-of: 458f6f1b82c9 ("can: tcan4x5x: Fix use of register error status mask") Signed-off-by: Sasha Levin <sashal@kernel.org>
Register 0x824 TCAN4X5X_MCAN_INT_REG is a read-only register. Any writes
to this register do not have any effect.
Remove this write. The m_can driver aldready clears the interrupts in
m_can_isr() by writing to M_CAN_IR which is translated to register
0x1050 which is a writable version of this register.
Fixes: b931004016bd ("can: tcan4x5x: Add tcan4x5x driver to the kernel") Signed-off-by: Markus Schneider-Pargmann <msp@baylibre.com> Link: https://lore.kernel.org/all/20221206115728.1056014-9-msp@baylibre.com Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de> Signed-off-by: Sasha Levin <sashal@kernel.org>
There are cables that exist that can support speeds in excess of 10GbE.
The driver, however, restricts the EEPROM advertised nominal bitrate to
a specific range, which can prevent usage of cables that can support,
for example, up to 25GbE.
Rather than checking that an active or passive cable supports a specific
range, only check for a minimum supported speed.
Fixes: 7422f7a8a299 ("amd-xgbe: Add support for SFP+ modules") Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
SFP+ active and passive cables are copper cables with fixed SFP+ end
connectors. Due to a misinterpretation of this, SFP+ active cables could
end up not being recognized, causing the driver to fail to establish a
connection.
Introduce a new enum in SFP+ cable types, XGBE_SFP_CABLE_FIBER, that is
the default cable type, and handle active and passive cables when they are
specifically detected.
Fixes: 7422f7a8a299 ("amd-xgbe: Add support for SFP+ modules") Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
If register unix_stream_proto returns error, unix_dgram_proto needs
be unregistered.
Fixes: 8cee822891d7 ("af_unix: Add unix_stream_proto for sockmap") Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Reviewed-by: Simon Horman <simon.horman@corigine.com> Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
Change the behaviour of ip6_datagram_connect to consider the interface
set by the IPV6_UNICAST_IF socket option, similarly to udpv6_sendmsg.
This change is the IPv6 counterpart of the fix for IP_UNICAST_IF.
The tests introduced by that patch showed that the incorrect
behavior is present in IPv6 as well.
This patch fixes the broken test.
Reported-by: kernel test robot <oliver.sang@intel.com> Link: https://lore.kernel.org/r/202210062117.c7eef1a3-oliver.sang@intel.com Fixes: a25c52a3fa35 ("net-next: Fix IP_UNICAST_IF option behavior for connected sockets") Signed-off-by: Richard Gobert <richardbgobert@gmail.com> Reviewed-by: David Ahern <dsahern@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
It is not allowed to call kfree_skb() or consume_skb() from hardware
interrupt context or with hardware interrupts being disabled.
It should use dev_kfree_skb_irq() or dev_consume_skb_irq() instead.
The difference between them is free reason, dev_kfree_skb_irq() means
the SKB is dropped in error and dev_consume_skb_irq() means the SKB
is consumed in normal.
In these two cases, dev_kfree_skb() is called consume the xmited SKB,
so replace it with dev_consume_skb_irq().
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
It is not allowed to call kfree_skb() or consume_skb() from hardware
interrupt context or with hardware interrupts being disabled.
It should use dev_kfree_skb_irq() or dev_consume_skb_irq() instead.
The difference between them is free reason, dev_kfree_skb_irq() means
the SKB is dropped in error and dev_consume_skb_irq() means the SKB
is consumed in normal.
In scc_discard_buffers(), dev_kfree_skb() is called to discard the SKBs,
so replace it with dev_kfree_skb_irq().
In scc_net_tx(), dev_kfree_skb() is called to drop the SKB that exceed
queue length, so replace it with dev_kfree_skb_irq().
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
It is not allowed to call kfree_skb() or consume_skb() from hardware
interrupt context or with hardware interrupts being disabled.
In this case, the lock is used to protected 'bp', so we can move
dev_kfree_skb() after the spin_unlock_irqrestore().
Fixes: e5e99c202938 ("dnet: Dave DNET ethernet controller driver (updated)") Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
It is not allowed to call kfree_skb() or consume_skb() from hardware
interrupt context or with hardware interrupts being disabled.
It should use dev_kfree_skb_irq() or dev_consume_skb_irq() instead.
The difference between them is free reason, dev_kfree_skb_irq() means
the SKB is dropped in error and dev_consume_skb_irq() means the SKB
is consumed in normal.
In this case, dev_kfree_skb() is called in xemaclite_tx_timeout() to
drop the SKB, when tx timeout, so replace it with dev_kfree_skb_irq().
Fixes: aae0199b4648 ("net: add Xilinx emac lite device driver") Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
It is not allowed to call kfree_skb() or consume_skb() from hardware
interrupt context or with hardware interrupts being disabled.
It should use dev_kfree_skb_irq() or dev_consume_skb_irq() instead.
The difference between them is free reason, dev_kfree_skb_irq() means
the SKB is dropped in error and dev_consume_skb_irq() means the SKB
is consumed in normal.
In this case, dev_kfree_skb() is called in bmac_tx_timeout() to drop
the SKB, when tx timeout, so replace it with dev_kfree_skb_irq().
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
It is not allowed to call kfree_skb() or consume_skb() from hardware
interrupt context or with hardware interrupts being disabled.
It should use dev_kfree_skb_irq() or dev_consume_skb_irq() instead.
The difference between them is free reason, dev_kfree_skb_irq() means
the SKB is dropped in error and dev_consume_skb_irq() means the SKB
is consumed in normal.
In this case, dev_kfree_skb() is called in mace_tx_timeout() to drop
the SKB, when tx timeout, so replace it with dev_kfree_skb_irq().
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
There is a race condition in vxlan that when deleting a vxlan device
during receiving packets, there is a possibility that the sock is
released after getting vxlan_sock vs from sk_user_data. Then in
later vxlan_ecn_decapsulate(), vxlan_get_sk_family() we will got
NULL pointer dereference. e.g.
Fix this by waiting for all sk_user_data reader to finish before
releasing the sock.
Reported-by: Jianlin Shi <jishi@redhat.com> Suggested-by: Jakub Sitnicki <jakub@cloudflare.com> Fixes: b8e6e3c0aa28 ("udp-tunnel: Add a few more UDP tunnel APIs") Signed-off-by: Hangbin Liu <liuhangbin@gmail.com> Reviewed-by: Jiri Pirko <jiri@nvidia.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
The root cause is traced to the netdev and fst_card_info are not freed
when removes one fst in fst_remove_one(), which may trigger oom if
repeated insmod and rmmod module.
Fix it by adding free_netdev() and kfree() in fst_remove_one(), just as
the operations on the error handling path in fst_add_one().
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Signed-off-by: Li Zetao <lizetao1@huawei.com> Reviewed-by: Jiri Pirko <jiri@nvidia.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
It is not allowed to call kfree_skb() or consume_skb() from hardware
interrupt context or with hardware interrupts being disabled.
It should use dev_kfree_skb_irq() or dev_consume_skb_irq() instead.
The difference between them is free reason, dev_kfree_skb_irq() means
the SKB is dropped in error and dev_consume_skb_irq() means the SKB
is consumed in normal.
In this case, dev_kfree_skb() is called in free_tx_buffers() to drop
the SKBs in tx buffers, when the card is down, so replace it with
dev_kfree_skb_irq() here.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
If vp alloc failed in qlcnic_sriov_init(), all previously allocated vp
needs to be freed.
Fixes: 41d136013fb3 ("qlcnic: VF-PF communication channel implementation") Signed-off-by: Yuan Can <yuancan@huawei.com> Reviewed-by: Leon Romanovsky <leonro@nvidia.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
The bitmap_free() should be called to free priv->af_xdp_zc_qps
when create_singlethread_workqueue() fails, otherwise there will
be a memory leak, so we add the err path error_wq_init to fix it.
Fixes: 7e1b9610f54a ("net: stmmac: Enable RX via AF_XDP zero-copy") Signed-off-by: Gaosheng Cui <cuigaosheng1@huawei.com> Reviewed-by: Leon Romanovsky <leonro@nvidia.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
The skb allocated by stmmac_test_get_arp_skb() hasn't been released in
some error handling case, which will lead to a memory leak. Fix this up
by adding kfree_skb() to release skb.
Compile tested only.
Fixes: 9d940364745f ("net: stmmac: selftests: Implement the ARP Offload test") Signed-off-by: Zhang Changzhong <zhangchangzhong@huawei.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
vmci_transport_dgram_enqueue() does not check the return value
of memcpy_from_msg(). If memcpy_from_msg() fails, it is possible that
uninitialized memory contents are sent unintentionally instead of user's
message in the datagram to the destination. Return with an error if
memcpy_from_msg() fails.
Found by Linux Verification Center (linuxtesting.org) with SVACE.
Fixes: d107d5b48566 ("vmci_transport: switch ->enqeue_dgram, ->enqueue_stream and ->dequeue_stream to msghdr") Signed-off-by: Artem Chernyshev <artem.chernyshev@red-soft.ru> Reviewed-by: Stefano Garzarella <sgarzare@redhat.com> Reviewed-by: Vishnu Dasa <vdasa@vmware.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
In BPF all global functions, and BPF helpers return a 64-bit
value. For kfunc calls, this is not the case, and they can return
e.g. 32-bit values.
The return register R0 for kfuncs calls can therefore be marked as
subreg_def != DEF_NOT_SUBREG. In general, if a register is marked with
subreg_def != DEF_NOT_SUBREG, some archs (where bpf_jit_needs_zext()
returns true) require the verifier to insert explicit zero-extension
instructions.
For kfuncs calls, however, the caller should do sign/zero extension
for return values. In other words, the compiler is responsible to
insert proper instructions, not the verifier.
If the return value of 'foo()' is used in the BPF program, the proper
zero-extension will be done.
Currently, the verifier correctly marks, say, a 32-bit return value as
subreg_def != DEF_NOT_SUBREG, but will fail performing the actual
zero-extension, due to a verifier bug in
opt_subreg_zext_lo32_rnd_hi32(). load_reg is not properly set to R0,
and the following path will be taken:
if (WARN_ON(load_reg == -1)) {
verbose(env, "verifier bug. zext_dst is set, but no reg is defined\n");
return -EFAULT;
}
A longer discussion from v1 can be found in the link below.
Correct the verifier by avoiding doing explicit zero-extension of R0
for kfunc calls. Note that R0 will still be marked as a sub-register
for return values smaller than 64-bit.
When the blk_classic option is enabled, non-blktrace events must be
filtered out. Otherwise, events of other types are output in the blktrace
classic format, which is unexpected.
The problem can be triggered in the following ways:
Fix to return a negative error code instead of 0 when
brcmf_chip_set_active() fails. In addition, change the return
value for brcmf_pcie_exit_download_state() to keep consistent.
Fixes: 4c5cba42d1f9 ("brcmfmac: rename chip download functions") Signed-off-by: Wang Yufen <wangyufen@huawei.com> Reviewed-by: Arend van Spriel <arend.vanspriel@broadcom.com> Signed-off-by: Kalle Valo <kvalo@kernel.org> Link: https://lore.kernel.org/r/1669959342-27144-1-git-send-email-wangyufen@huawei.com Signed-off-by: Sasha Levin <sashal@kernel.org>
The gen 2 chips RTL8192EU and RTL8188FU periodically send the driver
reports about the TX rate, and the driver passes these reports to
sta_statistics. The reports from RTL8192EU may or may not include the
channel width. The reports from RTL8188FU do not include it.
Only access the c2h->ra_report.bw field if the report (skb) is big
enough.
The other problem fixed here is that the code was actually never
changing the channel width initially reported by
rtl8xxxu_bss_info_changed because the value of RATE_INFO_BW_20 is 0.
Fixes: 29f21f85efdb ("rtl8xxxu: Feed current txrate information for mac80211") Signed-off-by: Bitterblue Smith <rtl8821cerfe2@gmail.com> Reviewed-by: Ping-Ke Shih <pkshih@realtek.com> Signed-off-by: Kalle Valo <kvalo@kernel.org> Link: https://lore.kernel.org/r/5b41f1ae-72e7-6b7a-2459-b736399a1c40@gmail.com Signed-off-by: Sasha Levin <sashal@kernel.org>
The addition of 3WIRE support would affect MOSI direction even
when still in standard (4 wire) mode. This can lead to MOSI being
at an invalid logic level when a device driver sets an SPI
message with a NULL tx_buf.
spi.h states that if tx_buf is NULL then "zeros will be shifted
out ... " If MOSI is tristated then the data shifted out is subject
to pull resistors, keepers, or in the absence of those, noise.
This issue came to light when using spi-gpio connected to an
ADS7843 touchscreen controller. MOSI pulled high when clocking
MISO data in caused the SPI device to interpret this as a command
which would put the device in an unexpected and non-functional
state.
Fixes: a4d368b67b8e ("spi: spi-gpio: add SPI_3WIRE support") Fixes: a3f12beee6bf ("spi: gpio: Support 3WIRE high-impedance turn-around") Signed-off-by: Kris Bahnsen <kris@embeddedTS.com> Link: https://lore.kernel.org/r/20221207230853.6174-1-kris@embeddedTS.com Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
If clk_register() fails, @pll->rate_table may have allocated memory by
kmemdup(), so it needs to be freed, otherwise will cause memory leak
issue, this patch fixes it.
Fixes: 1268bd30087a ("clk: samsung: Add support to register rate_table for samsung plls") Signed-off-by: Xiu Jianfeng <xiujianfeng@huawei.com> Link: https://lore.kernel.org/r/20221123032015.63980-1-xiujianfeng@huawei.com Reviewed-by: Alim Akhtar <alim.akhtar@samsung.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
As the coda_iram_alloc may return NULL pointer,
it should be better to check the return value
in order to avoid NULL poineter dereference,
same as the others.
In configure_channels(), we should call of_node_put() when breaking
out of for_each_child_of_node() which will automatically increase
and decrease the refcount.
Fixes: b754abbeb59e ("[media] c8sectpfe: STiH407/10 Linux DVB demux support") Signed-off-by: Liang He <windhl@126.com> Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl> Signed-off-by: Sasha Levin <sashal@kernel.org>
Let's use pr_err() to output the error messages and let's extend a comment
to clarify why returning 0 (success) in one case make sense.
Fixes: 1957980b0e9b ("mmc: core: Read the SD function extension registers for power management") Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
[Ulf: Clarified the comment and the commit-msg] Link: https://lore.kernel.org/r/20221130134920.2109-1-thunder.leizhen@huawei.com Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
As the alloc_ordered_workqueue may return NULL pointer, it should be better
to add check for the return value. Moreover, the msb->io_queue should be
freed if error occurs later.
mmc_add_host() may return error, if we ignore its return value,
it will lead two issues:
1. The memory that allocated in mmc_alloc_host() is leaked.
2. In the remove() path, mmc_remove_host() will be called to
delete device, but it's not added yet, it will lead a kernel
crash because of null-ptr-deref in device_del().
So fix this by checking the return value and goto error path which
will call mmc_free_host().
mmc_add_host() may return error, if we ignore its return value,
it will lead two issues:
1. The memory that allocated in mmc_alloc_host() is leaked.
2. In the remove() path, mmc_remove_host() will be called to
delete device, but it's not added yet, it will lead a kernel
crash because of null-ptr-deref in device_del().
So fix this by checking the return value and goto error path which
will call mmc_free_host(), besides, other resources also need be
released.
mmc_add_host() may return error, if we ignore its return value,
it will lead two issues:
1. The memory that allocated in mmc_alloc_host() is leaked.
2. In the remove() path, mmc_remove_host() will be called to
delete device, but it's not added yet, it will lead a kernel
crash because of null-ptr-deref in device_del().
Fix this by checking the return value and goto error path which
will call mmc_free_host().
mmc_add_host() may return error, if we ignore its return value,
it will lead two issues:
1. The memory that allocated in mmc_alloc_host() is leaked.
2. In the remove() path, mmc_remove_host() will be called to
delete device, but it's not added yet, it will lead a kernel
crash because of null-ptr-deref in device_del().
Fix this by checking the return value and goto error path which
will call mmc_free_host().
Fixes: d9070c09fdfd ("MMC: meson: initial support for GX platforms") Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org> Link: https://lore.kernel.org/r/20221108123417.479045-1-yangyingliang@huawei.com Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
mmc_add_host() may return error, if we ignore its return value,
it will lead two issues:
1. The memory that allocated in mmc_alloc_host() is leaked.
2. In the remove() path, mmc_remove_host() will be called to
delete device, but it's not added yet, it will lead a kernel
crash because of null-ptr-deref in device_del().
Fix this by checking the return value and goto error path wihch
will call mmc_free_host().
Fixes: c96eead25096 ("[ARM] 5369/1: omap mmc: Add new omap hsmmc controller for 2430 and 34xx, v3") Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Link: https://lore.kernel.org/r/20221108121316.340354-1-yangyingliang@huawei.com Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
mmc_add_host() may return error, if we ignore its return value,
it will lead two issues:
1. The memory that allocated in mmc_alloc_host() is leaked.
2. In the remove() path, mmc_remove_host() will be called to
delete device, but it's not added yet, it will lead a kernel
crash because of null-ptr-deref in device_del().
So fix this by checking the return value and calling mmc_free_host()
in the error path.
mmc_add_host() may return error, if we ignore its return value, the memory
that allocated in mmc_alloc_host() will be leaked and it will lead a kernel
crash because of deleting not added device in the remove path.
So fix this by checking the return value and goto error path which will call
mmc_free_host(), besides, clk_disable_unprepare() also needs be called.
mmc_add_host() may return error, if we ignore its return value, the memory
that allocated in mmc_alloc_host() will be leaked and it will lead a kernel
crash because of deleting not added device in the remove path.
So fix this by checking the return value and goto error path which will call
mmc_free_host(), besides, the timer added before mmc_add_host() needs be del.
And this patch fixes another missing call mmc_free_host() if usb_control_msg()
fails.
mmc_add_host() may return error, if we ignore its return value, the memory
that allocated in mmc_alloc_host() will be leaked and it will lead a kernel
crash because of deleting not added device in the remove path.
So fix this by checking the return value and goto error path which will call
mmc_free_host(), besides, free_irq() also needs be called.
mmc_add_host() may return error, if we ignore its return value, the memory
that allocated in mmc_alloc_host() will be leaked and it will lead a kernel
crash because of deleting not added device in the remove path.
So fix this by checking the return value and calling mmc_free_host() in the
error path, besides, led_classdev_unregister() and pm_runtime_disable() also
need be called.
mmc_add_host() may return error, if we ignore its return value, the memory
that allocated in mmc_alloc_host() will be leaked and it will lead a kernel
crash because of deleting not added device in the remove path.
So fix this by checking the return value and calling mmc_free_host() in the
error path, beside, runtime PM also needs be disabled.
mmc_add_host() may return error, if we ignore its return value, the memory
that allocated in mmc_alloc_host() will be leaked and it will lead a kernel
crash because of deleting not added device in the remove path.
So fix this by checking the return value and goto error path which will call
mmc_free_host(), besides, ->exit() need be called to uninit the pdata.
mmc_add_host() may return error, if we ignore its return value, the memory
that allocated in mmc_alloc_host() will be leaked and it will lead a kernel
crash because of deleting not added device in the remove path.
So fix this by checking the return value and goto error path which will call
mmc_free_host().
Fixes: a65a4972f2f5 ("mmc: Add a MX2/MX3 specific SDHC driver") Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Link: https://lore.kernel.org/r/20221101063023.1664968-4-yangyingliang@huawei.com Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
mmc_add_host() may return error, if we ignore its return value, the memory
that allocated in mmc_alloc_host() will be leaked and it will lead a kernel
crash because of deleting not added device in the remove path.
So fix this by checking the return value and goto error path which will call
mmc_free_host().
mmc_add_host() may return error, if we ignore its return value, the memory
that allocated in mmc_alloc_host() will be leaked and it will lead a kernel
crash because of deleting not added device in the remove path.
So fix this by checking the return value and calling mmc_free_host() in the
error path.
There are fields 'last_hwmon_state' and 'last_thermal_state' in the
structure 'emc2305_cdev_data', which respectively store the cooling state
set by the 'hwmon' and 'thermal' subsystem, and the driver author hopes
that if the state set by 'hwmon' is lower than the value set by 'thermal',
the driver will just save it without actually setting the pwm. Currently,
the 'last_thermal_state' also be updated by 'hwmon', which will cause the
cooling state to never be set to a lower value. This patch fixes that.
The definitions of 'EMC2305_REG_PRODUCT_ID' and 'EMC2305_REG_DEVICE' are
both '0xfd', they actually return the same value, but the values returned
by emc2301/2/3/5 are different, so probe emc2301/2/3 will fail, This patch
fixes that.
Signed-off-by: Xingjiang Qiao <nanpuyue@gmail.com> Link: https://lore.kernel.org/r/20221206055331.170459-1-nanpuyue@gmail.com Fixes: 69ae061f605d4 ("hwmon: (emc2305) add support for EMC2301/2/3/5 RPM-based PWM Fan Speed Controller.") Signed-off-by: Guenter Roeck <linux@roeck-us.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
For BPF_PSEUDO_FUNC instruction, verifier will refill imm with
correct addresses of bpf_calls and then run last pass of JIT.
Since the emit_imm of RV64 is variable-length, which will emit
appropriate length instructions accorroding to the imm, it may
broke ctx->offset, and lead to unpredictable problem, such as
inaccurate jump. So let's fix it with fixed-length instructions.
Fixes: 359bfbf9bf1e ("bpf: Add bpf_for_each_map_elem() helper") Suggested-by: Björn Töpel <bjorn@rivosinc.com> Signed-off-by: Pu Lehui <pulehui@huawei.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Reviewed-by: Björn Töpel <bjorn@kernel.org> Acked-by: Björn Töpel <bjorn@kernel.org> Link: https://lore.kernel.org/bpf/20221206091410.1584784-1-pulehui@huaweicloud.com Signed-off-by: Sasha Levin <sashal@kernel.org>
If the state manager thread fails to start, then we should just mark the
client initialisation as failed so that other processes or threads don't
get stuck in nfs_wait_client_init_complete().
Reported-by: ChenXiaoSong <chenxiaosong2@huawei.com> Fixes: e44afd73e720 ("NFSv4: Fix a race in the net namespace mount notification") Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
8d6b622700d2 introduced nfs_io_size() to clamp the iosize to a multiple
of PAGE_SIZE. This had the unintended side effect of no longer allowing
iosizes less than a page, which could be useful in some situations.
UDP already has an exception that causes it to fall back on the
power-of-two style sizes instead. This patch adds an additional
exception for very small iosizes.
Reported-by: Jeff Layton <jlayton@kernel.org> Fixes: 8d6b622700d2 ("NFS: Allow setting rsize / wsize to a multiple of PAGE_SIZE") Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com> Reviewed-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The scratch_buf array is 16 bytes, but I was passing 32 to the
xdr_set_scratch_buffer() function. Fix this by using sizeof(), which is
what I probably should have been doing this whole time.
Fixes: a22895a7a24f ("NFS: Replace the READ_PLUS decoding code") Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com> Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
socket dynamically created is not released when getting an unintended
address family type in rpc_sockname(), direct to out_release for calling
sock_release().
Fixes: 99f53b6e051b ("SUNRPC: Add API to acquire source address") Signed-off-by: Wang ShaoBo <bobo.shaobowang@huawei.com> Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The thunderbay_add_functions() will free memory of thunderbay_funcs
when everything is ok, but thunderbay_funcs will not be freed when
thunderbay_add_functions() fails, then there will be a memory leak,
so we need to add kfree() when thunderbay_add_functions() fails to
fix it.
In addition, doing some cleaner works, moving kfree(funcs) from
thunderbay_add_functions() to thunderbay_build_functions().
When a PCM trigger-start fails at snd_pcm_do_start(), PCM core tries
to undo the action at snd_pcm_undo_start() by issuing the trigger STOP
manually. At that point, we forgot to set the stop_operating flag,
hence the sync-stop won't be issued at the next prepare or other
calls.
This patch adds the missing stop_operating flag at
snd_pcm_undo_start().
sock_map_free() calls release_sock(sk) without owning a reference
on the socket. This can cause use-after-free as syzbot found [1]
Jakub Sitnicki already took care of a similar issue
in sock_hash_free() in commit 5152c9504882 ("bpf, sockhash:
Synchronize delete from bucket list on map free")
The bpf_ct_set_nat_info() kfunc is defined in the nf_nat.ko module, and
takes as a parameter the nf_conn___init struct, which is allocated through
the bpf_xdp_ct_alloc() helper defined in the nf_conntrack.ko module.
However, because kernel modules can't deduplicate BTF types between each
other, and the nf_conn___init struct is not referenced anywhere in vmlinux
BTF, this leads to two distinct BTF IDs for the same type (one in each
module). This confuses the verifier, as described here:
As a workaround, add an explicit BTF_TYPE_EMIT for the type in
net/filter.c, so the type definition gets included in vmlinux BTF. This
way, both modules can refer to the same type ID (as they both build on top
of vmlinux BTF), and the verifier is no longer confused.
v2:
- Use BTF_TYPE_EMIT (which is a statement so it has to be inside a function
definition; use xdp_func_proto() for this, since this is mostly
xdp-related).
The JC42 compatible thermal sensor on Kingston KSM32ES8/16ME DIMMs
(using Micron E-Die) is an ST Microelectronics STTS2004 (manufacturer
0x104a, device 0x2201). It does not keep the previously programmed
minimum, maximum and critical temperatures after system suspend and
resume (which is a shutdown / startup cycle for the JC42 temperature
sensor). This results in an alarm on system resume because the hardware
default for these values is 0°C (so any environment temperature greater
than 0°C will trigger the alarm).
Example before system suspend:
jc42-i2c-0-1a
Adapter: SMBus PIIX4 adapter port 0 at 0b00
temp1: +34.8°C (low = +0.0°C)
(high = +85.0°C, hyst = +85.0°C)
(crit = +95.0°C, hyst = +95.0°C)
Example after system resume (without this change):
jc42-i2c-0-1a
Adapter: SMBus PIIX4 adapter port 0 at 0b00
temp1: +34.8°C (low = +0.0°C) ALARM (HIGH, CRIT)
(high = +0.0°C, hyst = +0.0°C)
(crit = +0.0°C, hyst = +0.0°C)
Apply the cached values from the JC42_REG_TEMP_UPPER,
JC42_REG_TEMP_LOWER, JC42_REG_TEMP_CRITICAL and JC42_REG_SMBUS (where
the SMBUS register is not related to this issue but a side-effect of
using regcache_sync() during system resume with the previously
cached/programmed values. This fixes the alarm due to the hardware
defaults of 0°C because the previously applied limits (set by userspace)
are re-applied on system resume.
Fixes: bfb20afe6900 ("hwmon: (jc42) Add support for STTS2004 and AT30TSE004") Reviewed-by: Guenter Roeck <linux@roeck-us.net> Signed-off-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com> Link: https://lore.kernel.org/r/20221023213157.11078-3-martin.blumenstingl@googlemail.com Signed-off-by: Guenter Roeck <linux@roeck-us.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
Switch the jc42 driver to use an I2C regmap to access the registers.
Also move over to regmap's built-in caching instead of adding a
custom caching implementation. This works for JC42_REG_TEMP_UPPER,
JC42_REG_TEMP_LOWER and JC42_REG_TEMP_CRITICAL as these values never
change except when explicitly written. The cache For JC42_REG_TEMP is
dropped (regmap can't cache it because it's volatile, meaning it can
change at any time) as well for simplicity and consistency with other
drivers.
After calling regulator_resolve_supply(), the 'rdev->supply' is set
by set_supply(), after this set, in the error path, the resources
need be released, so call regulator_put() to avoid the leaks.
Fixes: bbe1299acdc1 ("regulator: resolve supply after creating regulator") Fixes: 945db0f47e2d ("regulator: core: Resolve supply name earlier to prevent double-init") Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Link: https://lore.kernel.org/r/20221202025111.496402-1-yangyingliang@huawei.com Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
hsr_register_frame_out() compares new sequence_nr vs the old one
recorded in hsr_node::seq_out and if the new sequence_nr is higher then
it will be written to hsr_node::seq_out as the new value.
This operation isn't locked so it is possible that two frames with the
same sequence number arrive (via the two slave devices) and are fed to
hsr_register_frame_out() at the same time. Both will pass the check and
update the sequence counter later to the same value. As a result the
content of the same packet is fed into the stack twice.
This was noticed by running ping and observing DUP being reported from
time to time.
Instead of using the hsr_priv::seqnr_lock for the whole receive path (as
it is for sending in the master node) add an additional lock that is only
used for sequence number checks and updates.
Add a per-node lock that is used during sequence number reads and
updates.
Fixes: f8baa9c29c384 ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Sending frames via the hsr (master) device requires a sequence number
which is tracked in hsr_priv::sequence_nr and protected by
hsr_priv::seqnr_lock. Each time a new frame is sent, it will obtain a
new id and then send it via the slave devices.
Each time a packet is sent (via hsr_forward_do()) the sequence number is
checked via hsr_register_frame_out() to ensure that a frame is not
handled twice. This make sense for the receiving side to ensure that the
frame is not injected into the stack twice after it has been received
from both slave ports.
There is no locking to cover the sending path which means the following
scenario is possible:
Both skbs (or their struct hsr_frame_info) received an unique id.
However since skb2 was sent before skb1, the higher sequence number was
recorded in hsr_register_frame_out() and the late arriving skb1 was
dropped and never sent.
This scenario has been observed in a three node HSR setup, with node1 +
node2 having ping and iperf running in parallel. From time to time ping
reported a missing packet. Based on tracing that missing ping packet did
not leave the system.
It might be possible (didn't check) to drop the sequence number check on
the sending side. But if the higher sequence number leaves on wire
before the lower does and the destination receives them in that order
and it will drop the packet with the lower sequence number and never
inject into the stack.
Therefore it seems the only way is to lock the whole path from obtaining
the sequence number and sending via dev_queue_xmit() and assuming the
packets leave on wire in the same order (and don't get reordered by the
NIC).
Cover the whole path for the master interface from obtaining the ID
until after it has been forwarded via hsr_forward_skb() to ensure the
skbs are sent to the NIC in the order of the assigned sequence numbers.
Fixes: f8baa9c29c384 ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The hsr device is a software device. Its
net_device_ops::ndo_start_xmit() routine will process the packet and
then pass the resulting skb to dev_queue_xmit().
During processing, hsr acquires a lock with spin_lock_bh()
(hsr_add_node()) which needs to be promoted to the _irq() suffix in
order to avoid a potential deadlock.
Then there are the warnings in dev_queue_xmit() (due to
local_bh_disable() with disabled interrupts) left.
Instead trying to address those (there is qdisc and…) for netpoll sake,
just disable netpoll on hsr.
Disable netpoll on hsr and replace the _irqsave() locking with _bh().
Fixes: f8baa9c29c384 ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Due to the hashed-MAC optimisation one problem become visible:
hsr_handle_sup_frame() walks over the list of available nodes and merges
two node entries into one if based on the information in the supervision
both MAC addresses belong to one node. The list-walk happens on a RCU
protected list and delete operation happens under a lock.
If the supervision arrives on both slave interfaces at the same time
then this delete operation can occur simultaneously on two CPUs. The
result is the first-CPU deletes the from the list and the second CPUs
BUGs while attempting to dereference a poisoned list-entry. This happens
more likely with the optimisation because a new node for the mac_B entry
is created once a packet has been received and removed (merged) once the
supervision frame has been received.
Avoid removing/ cleaning up a hsr_node twice by adding a `removed' field
which is set to true after the removal and checked before the removal.
Fixes: 2da138a9a8154 ("net/hsr: Better frame dispatch") Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
projects / kernel.git / log