Saeed Mahameed says:
====================
mlx5 fixes 2023-07-26
This series provides bug fixes to mlx5 driver.
* tag 'mlx5-fixes-2023-07-26' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux:
net/mlx5: Unregister devlink params in case interface is down
net/mlx5: DR, Fix peer domain namespace setting
net/mlx5: fs_chains: Fix ft prio if ignore_flow_level is not supported
net/mlx5e: kTLS, Fix protection domain in use syndrome when devlink reload
net/mlx5: Bridge, set debugfs access right to root-only
net/mlx5e: xsk: Fix crash on regular rq reactivation
net/mlx5e: xsk: Fix invalid buffer access for legacy rq
net/mlx5e: Move representor neigh cleanup to profile cleanup_tx
net/mlx5e: Fix crash moving to switchdev mode when ntuple offload is set
net/mlx5e: Don't hold encap tbl lock if there is no encap action
net/mlx5: Honor user input for migratable port fn attr
net/mlx5e: fix return value check in mlx5e_ipsec_remove_trailer()
net/mlx5: fix potential memory leak in mlx5e_init_rep_rx
net/mlx5: DR, fix memory leak in mlx5dr_cmd_create_reformat_ctx
net/mlx5e: fix double free in macsec_fs_tx_create_crypto_table_groups
====================
Link: https://lore.kernel.org/r/20230726213206.47022-1-saeed@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The offending patch is based on the assumption that for PFs,
mlx5_get_dev_index() is the same as vhca_id. However, this assumption
is wrong in case of DPU (ECPF).
Fix it by using vhca_id directly, and switch the array of peers to
xarray.
Fixes: 6d5b7321d8 ("net/mlx5: DR, handle more than one peer domain")
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Yevgeny Kliteynik <kliteyn@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
The cited commit sets ft prio to fs_base_prio. But if
ignore_flow_level it not supported, ft prio must be set based on
tc filter prio. Otherwise, all the ft prio are the same on the same
chain. It is invalid if ignore_flow_level is not supported.
Fix it by setting ft prio based on tc filter prio and setting
fs_base_prio to 0 for fdb.
Fixes: 8e80e56480 ("net/mlx5: fs_chains: Refactor to detach chains from tc usage")
Signed-off-by: Chris Mi <cmi@nvidia.com>
Reviewed-by: Paul Blakey <paulb@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
There are DEK objects cached in DEK pool after kTLS is used, and they
are freed only in mlx5e_ktls_cleanup().
mlx5e_destroy_mdev_resources() is called in mlx5e_suspend() to
free mdev resources, including protection domain (PD). However, PD is
still referenced by the cached DEK objects in this case, because
profile->cleanup() (and therefore mlx5e_ktls_cleanup()) is called
after mlx5e_suspend() during devlink reload. So the following FW
syndrome is generated:
mlx5_cmd_out_err:803:(pid 12948): DEALLOC_PD(0x801) op_mod(0x0) failed,
status bad resource state(0x9), syndrome (0xef0c8a), err(-22)
To avoid this syndrome, move DEK pool destruction to
mlx5e_ktls_cleanup_tx(), which is called by profile->cleanup_tx(). And
move pool creation to mlx5e_ktls_init_tx() for symmetry.
Fixes: f741db1a51 ("net/mlx5e: kTLS, Improve connection rate by using fast update encryption key")
Signed-off-by: Jianbo Liu <jianbol@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
When the regular rq is reactivated after the XSK socket is closed
it could be reading stale cqes which eventually corrupts the rq.
This leads to no more traffic being received on the regular rq and a
crash on the next close or deactivation of the rq.
Kal Cuttler Conely reported this issue as a crash on the release
path when the xdpsock sample program is stopped (killed) and restarted
in sequence while traffic is running.
This patch flushes all cqes when during the rq flush. The cqe flushing
is done in the reset state of the rq. mlx5e_rq_to_ready code is moved
into the flush function to allow for this.
Fixes: 082a9edf12 ("net/mlx5e: xsk: Flush RQ on XSK activation to save memory")
Reported-by: Kal Cutter Conley <kal.conley@dectris.com>
Closes: https://lore.kernel.org/xdp-newbies/CAHApi-nUAs4TeFWUDV915CZJo07XVg2Vp63-no7UDfj6wur9nQ@mail.gmail.com
Signed-off-by: Dragos Tatulea <dtatulea@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
The cited commit holds encap tbl lock unconditionally when setting
up dests. But it may cause the following deadlock:
PID: 1063722 TASK: ffffa062ca5d0000 CPU: 13 COMMAND: "handler8"
#0 [ffffb14de05b7368] __schedule at ffffffffa1d5aa91
#1 [ffffb14de05b7410] schedule at ffffffffa1d5afdb
#2 [ffffb14de05b7430] schedule_preempt_disabled at ffffffffa1d5b528
#3 [ffffb14de05b7440] __mutex_lock at ffffffffa1d5d6cb
#4 [ffffb14de05b74e8] mutex_lock_nested at ffffffffa1d5ddeb
#5 [ffffb14de05b74f8] mlx5e_tc_tun_encap_dests_set at ffffffffc12f2096 [mlx5_core]
#6 [ffffb14de05b7568] post_process_attr at ffffffffc12d9fc5 [mlx5_core]
#7 [ffffb14de05b75a0] mlx5e_tc_add_fdb_flow at ffffffffc12de877 [mlx5_core]
#8 [ffffb14de05b75f0] __mlx5e_add_fdb_flow at ffffffffc12e0eef [mlx5_core]
#9 [ffffb14de05b7660] mlx5e_tc_add_flow at ffffffffc12e12f7 [mlx5_core]
#10 [ffffb14de05b76b8] mlx5e_configure_flower at ffffffffc12e1686 [mlx5_core]
#11 [ffffb14de05b7720] mlx5e_rep_indr_offload at ffffffffc12e3817 [mlx5_core]
#12 [ffffb14de05b7730] mlx5e_rep_indr_setup_tc_cb at ffffffffc12e388a [mlx5_core]
#13 [ffffb14de05b7740] tc_setup_cb_add at ffffffffa1ab2ba8
#14 [ffffb14de05b77a0] fl_hw_replace_filter at ffffffffc0bdec2f [cls_flower]
#15 [ffffb14de05b7868] fl_change at ffffffffc0be6caa [cls_flower]
#16 [ffffb14de05b7908] tc_new_tfilter at ffffffffa1ab71f0
[1031218.028143] wait_for_completion+0x24/0x30
[1031218.028589] mlx5e_update_route_decap_flows+0x9a/0x1e0 [mlx5_core]
[1031218.029256] mlx5e_tc_fib_event_work+0x1ad/0x300 [mlx5_core]
[1031218.029885] process_one_work+0x24e/0x510
Actually no need to hold encap tbl lock if there is no encap action.
Fix it by checking if encap action exists or not before holding
encap tbl lock.
Fixes: 37c3b9fa7c ("net/mlx5e: Prevent encap offload when neigh update is running")
Signed-off-by: Chris Mi <cmi@nvidia.com>
Reviewed-by: Vlad Buslov <vladbu@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Currently, whenever a user is setting migratable port fn attr, the
driver is always turn migratable capability on.
Fix it by honor the user input
Fixes: e5b9642a33 ("net/mlx5: E-Switch, Implement devlink port function cmds to control migratable")
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
mlx5e_ipsec_remove_trailer() should return an error code if function
pskb_trim() returns an unexpected value.
Fixes: 2ac9cfe782 ("net/mlx5e: IPSec, Add Innova IPSec offload TX data path")
Signed-off-by: Yuanjun Gong <ruc_gongyuanjun@163.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
The memory pointed to by the priv->rx_res pointer is not freed in the error
path of mlx5e_init_rep_rx, which can lead to a memory leak. Fix by freeing
the memory in the error path, thereby making the error path identical to
mlx5e_cleanup_rep_rx().
Fixes: af8bbf7300 ("net/mlx5e: Convert mlx5e_flow_steering member of mlx5e_priv to pointer")
Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
when mlx5_cmd_exec failed in mlx5dr_cmd_create_reformat_ctx, the memory
pointed by 'in' is not released, which will cause memory leak. Move memory
release after mlx5_cmd_exec.
Fixes: 1d9186476e ("net/mlx5: DR, Add direct rule command utilities")
Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
In function macsec_fs_tx_create_crypto_table_groups(), when the ft->g
memory is successfully allocated but the 'in' memory fails to be
allocated, the memory pointed to by ft->g is released once. And in function
macsec_fs_tx_create(), macsec_fs_tx_destroy() is called to release the
memory pointed to by ft->g again. This will cause double free problem.
Fixes: e467b283ff ("net/mlx5e: Add MACsec TX steering rules")
Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
The qca8k switch doesn't support using 0 as VID and require a default
VID to be always set. MDB add/del function doesn't currently handle
this and are currently setting the default VID.
Fix this by correctly handling this corner case and internally use the
default VID for VID 0 case.
Fixes: ba8f870dfa ("net: dsa: qca8k: add support for mdb_add/del")
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Cc: stable@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
On deleting an MDB entry for a port, fdb_search_and_del is used.
An FDB entry can't be modified so it needs to be deleted and readded
again with the new portmap (and the port deleted as requested)
We use the SEARCH operator to search the entry to edit by vid and mac
address and then we check the aging if we actually found an entry.
Currently the code suffer from a bug where the searched fdb entry is
never read again with the found values (if found) resulting in the code
always returning -EINVAL as aging was always 0.
Fix this by correctly read the fdb entry after it was searched.
Fixes: ba8f870dfa ("net: dsa: qca8k: add support for mdb_add/del")
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Cc: stable@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
On inserting a mdb entry, fdb_search_and_insert is used to add a port to
the qca8k target entry in the FDB db.
A FDB entry can't be modified so it needs to be removed and insert again
with the new values.
To detect if an entry already exist, the SEARCH operation is used and we
check the aging of the entry. If the entry is not 0, the entry exist and
we proceed to delete it.
Current code have 2 main problem:
- The condition to check if the FDB entry exist is wrong and should be
the opposite.
- When a FDB entry doesn't exist, aging was never actually set to the
STATIC value resulting in allocating an invalid entry.
Fix both problem by adding aging support to the function, calling the
function with STATIC as aging by default and finally by correct the
condition to check if the entry actually exist.
Fixes: ba8f870dfa ("net: dsa: qca8k: add support for mdb_add/del")
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Cc: stable@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
The qca8xxx switch supports 2 way to write reg values, a slow way using
mdio and a fast way by sending specially crafted mgmt packet to
read/write reg.
The fast way can support up to 32 bytes of data as eth packet are used
to send/receive.
This correctly works for almost the entire regmap of the switch but with
the use of some kernel selftests for dsa drivers it was found a funny
and interesting hw defect/limitation.
For some specific reg, bulk write won't work and will result in writing
only part of the requested regs resulting in half data written. This was
especially hard to track and discover due to the total strangeness of
the problem and also by the specific regs where this occurs.
This occurs in the specific regs of the ATU table, where multiple entry
needs to be written to compose the entire entry.
It was discovered that with a bulk write of 12 bytes on
QCA8K_REG_ATU_DATA0 only QCA8K_REG_ATU_DATA0 and QCA8K_REG_ATU_DATA2
were written, but QCA8K_REG_ATU_DATA1 was always zero.
Tcpdump was used to make sure the specially crafted packet was correct
and this was confirmed.
The problem was hard to track as the lack of QCA8K_REG_ATU_DATA1
resulted in an entry somehow possible as the first bytes of the mac
address are set in QCA8K_REG_ATU_DATA0 and the entry type is set in
QCA8K_REG_ATU_DATA2.
Funlly enough writing QCA8K_REG_ATU_DATA1 results in the same problem
with QCA8K_REG_ATU_DATA2 empty and QCA8K_REG_ATU_DATA1 and
QCA8K_REG_ATU_FUNC correctly written.
A speculation on the problem might be that there are some kind of
indirection internally when accessing these regs and they can't be
accessed all together, due to the fact that it's really a table mapped
somewhere in the switch SRAM.
Even more funny is the fact that every other reg was tested with all
kind of combination and they are not affected by this problem. Read
operation was also tested and always worked so it's not affected by this
problem.
The problem is not present if we limit writing a single reg at times.
To handle this hardware defect, enable use_single_write so that bulk
api can correctly split the write in multiple different operation
effectively reverting to a non-bulk write.
Cc: Mark Brown <broonie@kernel.org>
Fixes: c766e077d9 ("net: dsa: qca8k: convert to regmap read/write API")
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Cc: stable@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
Last year, the code that manages GSI channel transactions switched
from using spinlock-protected linked lists to using indexes into the
ring buffer used for a channel. Recently, Google reported seeing
transaction reference count underflows occasionally during shutdown.
Doug Anderson found a way to reproduce the issue reliably, and
bisected the issue to the commit that eliminated the linked lists
and the lock. The root cause was ultimately determined to be
related to unused transactions being committed as part of the modem
shutdown cleanup activity. Unused transactions are not normally
expected (except in error cases).
The modem uses some ranges of IPA-resident memory, and whenever it
shuts down we zero those ranges. In ipa_filter_reset_table() a
transaction is allocated to zero modem filter table entries. If
hashing is not supported, hashed table memory should not be zeroed.
But currently nothing prevents that, and the result is an unused
transaction. Something similar occurs when we zero routing table
entries for the modem.
By preventing any attempt to clear hashed tables when hashing is not
supported, the reference count underflow is avoided in this case.
Note that there likely remains an issue with properly freeing unused
transactions (if they occur due to errors). This patch addresses
only the underflows that Google originally reported.
Cc: <stable@vger.kernel.org> # 6.1.x
Fixes: d338ae28d8 ("net: ipa: kill all other transaction lists")
Tested-by: Douglas Anderson <dianders@chromium.org>
Signed-off-by: Alex Elder <elder@linaro.org>
Link: https://lore.kernel.org/r/20230724224055.1688854-1-elder@linaro.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The previous commit 954d1fa1ac ("macvlan: Add netlink attribute for
broadcast cutoff") added one additional attribute named
IFLA_MACVLAN_BC_CUTOFF to allow broadcast cutfoff.
However, it forgot to describe the nla_policy at macvlan_policy
(drivers/net/macvlan.c). Hence, this suppose NLA_S32 (4 bytes) integer
can be faked as empty (0 bytes) by a malicious user, which could leads
to OOB in heap just like CVE-2023-3773.
To fix it, this commit just completes the nla_policy description for
IFLA_MACVLAN_BC_CUTOFF. This enforces the length check and avoids the
potential OOB read.
Fixes: 954d1fa1ac ("macvlan: Add netlink attribute for broadcast cutoff")
Signed-off-by: Lin Ma <linma@zju.edu.cn>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Link: https://lore.kernel.org/r/20230723080205.3715164-1-linma@zju.edu.cn
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
commit a3a57bf07d ("net: stmmac: work
around sporadic tx issue on link-up") worked around a problem with TX
sometimes not working after a link-up by avoiding a redundant write to
MAC_CTRL_REG (aka GMAC_CONFIG), since the IP appeared to have problems
with handling multiple writes to that register in some cases.
That commit however only added the work around to dwmac_lib.c (apart
from the common code in stmmac_main.c), but my systems with version
4.21a of the IP exhibit the same problem, so add the work around to
dwmac4_lib.c too.
Fixes: a3a57bf07d ("net: stmmac: work around sporadic tx issue on link-up")
Signed-off-by: Vincent Whitchurch <vincent.whitchurch@axis.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Link: https://lore.kernel.org/r/20230721-stmmac-tx-workaround-v1-1-9411cbd5ee07@axis.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
As of today, hash extraction support is enabled for all the silicons.
Because of which we are facing initialization issues when the silicon
does not support hash extraction. During creation of the hardware
parsing table for IPv6 address, we need to consider if hash extraction
is enabled then extract only 32 bit, otherwise 128 bit needs to be
extracted. This patch fixes the issue and configures the hardware parser
based on the availability of the feature.
Fixes: a95ab93550 ("octeontx2-af: Use hashed field in MCAM key")
Signed-off-by: Suman Ghosh <sumang@marvell.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Link: https://lore.kernel.org/r/20230721061222.2632521-1-sumang@marvell.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
When adding a point to point downlink to team device, we neglected to reset
the team's flags, which were still using flags like BROADCAST and
MULTICAST. Consequently, this would initiate ARP/DAD for P2P downlink
interfaces, such as when adding a GRE device to team device. Fix this by
remove multicast/broadcast flags and add p2p and noarp flags.
After removing the none ethernet interface and adding an ethernet interface
to team, we need to reset team interface flags. Unlike bonding interface,
team do not need restore IFF_MASTER, IFF_SLAVE flags.
Reported-by: Liang Li <liali@redhat.com>
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2221438
Fixes: 1d76efe157 ("team: add support for non-ethernet devices")
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
When adding a point to point downlink to the bond, we neglected to reset
the bond's flags, which were still using flags like BROADCAST and
MULTICAST. Consequently, this would initiate ARP/DAD for P2P downlink
interfaces, such as when adding a GRE device to the bonding.
To address this issue, let's reset the bond's flags for P2P interfaces.
Before fix:
7: gre0@NONE: <POINTOPOINT,NOARP,SLAVE,UP,LOWER_UP> mtu 1500 qdisc noqueue master bond0 state UNKNOWN group default qlen 1000
link/gre6 2006:70:10::1 peer 2006:70:10::2 permaddr 167f:18:f188::
8: bond0: <BROADCAST,MULTICAST,MASTER,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
link/gre6 2006:70:10::1 brd 2006:70:10::2
inet6 fe80::200:ff:fe00:0/64 scope link
valid_lft forever preferred_lft forever
After fix:
7: gre0@NONE: <POINTOPOINT,NOARP,SLAVE,UP,LOWER_UP> mtu 1500 qdisc noqueue master bond2 state UNKNOWN group default qlen 1000
link/gre6 2006:70:10::1 peer 2006:70:10::2 permaddr c29e:557a:e9d9::
8: bond0: <POINTOPOINT,NOARP,MASTER,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
link/gre6 2006:70:10::1 peer 2006:70:10::2
inet6 fe80::1/64 scope link
valid_lft forever preferred_lft forever
Reported-by: Liang Li <liali@redhat.com>
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2221438
Fixes: 872254dd6b ("net/bonding: Enable bonding to enslave non ARPHRD_ETHER")
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Tony Nguyen says:
====================
Intel Wired LAN Driver Updates 2023-07-21 (i40e, iavf)
This series contains updates to i40e and iavf drivers.
Wang Ming corrects an error check on i40e.
Jake unlocks crit_lock on allocation failure to prevent deadlock and
stops re-enabling of interrupts when it's not intended for iavf.
* '40GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
iavf: check for removal state before IAVF_FLAG_PF_COMMS_FAILED
iavf: fix potential deadlock on allocation failure
i40e: Fix an NULL vs IS_ERR() bug for debugfs_create_dir()
====================
Link: https://lore.kernel.org/r/20230721155812.1292752-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Marc Kleine-Budde says:
====================
pull-request: can 2023-07-24
The first patch is by me and adds a missing set of CAN state to
CAN_STATE_STOPPED on close in the gs_usb driver.
The last patch is by Eric Dumazet and fixes a lockdep issue in the CAN
raw protocol.
* tag 'linux-can-fixes-for-6.5-20230724' of git://git.kernel.org/pub/scm/linux/kernel/git/mkl/linux-can:
can: raw: fix lockdep issue in raw_release()
can: gs_usb: gs_can_close(): add missing set of CAN state to CAN_STATE_STOPPED
====================
Link: https://lore.kernel.org/r/20230724150141.766047-1-mkl@pengutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Fix ethtool FDIR logic to not use memory after its release.
In the ice_ethtool_fdir.c file there are 2 spots where code can
refer to pointers which may be missing.
In the ice_cfg_fdir_xtrct_seq() function seg may be freed but
even then may be still used by memcpy(&tun_seg[1], seg, sizeof(*seg)).
In the ice_add_fdir_ethtool() function struct ice_fdir_fltr *input
may first fail to be added via ice_fdir_update_list_entry() but then
may be deleted by ice_fdir_update_list_entry.
Terminate in both cases when the returned value of the previous
operation is other than 0, free memory and don't use it anymore.
Reported-by: Michal Schmidt <mschmidt@redhat.com>
Link: https://bugzilla.redhat.com/show_bug.cgi?id=2208423
Fixes: cac2a27cd9 ("ice: Support IPv4 Flow Director filters")
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Signed-off-by: Jedrzej Jagielski <jedrzej.jagielski@intel.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Link: https://lore.kernel.org/r/20230721155854.1292805-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
According to the implementation of XDP of FEC driver, the XDP path
shares the transmit queues with the kernel network stack, so it is
possible to lead to a tx timeout event when XDP uses the tx queue
pretty much exclusively. And this event will cause the reset of the
FEC hardware.
To avoid timeout in this case, we use the txq_trans_cond_update()
interface to update txq->trans_start to jiffies so that watchdog
won't generate a transmit timeout warning.
Fixes: 6d6b39f180 ("net: fec: add initial XDP support")
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Link: https://lore.kernel.org/r/20230721083559.2857312-1-wei.fang@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In VXLAN-GPE, there may not be an Ethernet header following the VXLAN
header. But in GRO, the vxlan driver calls eth_gro_receive
unconditionally, which means the following header is incorrectly parsed
as Ethernet.
Introduce GPE specific GRO handling.
For better performance, do not check for GPE during GRO but rather
install a different set of functions at setup time.
Fixes: e1e5314de0 ("vxlan: implement GPE")
Reported-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Jiri Benc <jbenc@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The vxlan_parse_gpe_hdr function extracts the next protocol value from
the GPE header and marks GPE bits as parsed.
In order to be used in the next patch, split the function into protocol
extraction and bit marking. The bit marking is meaningful only in
vxlan_rcv; move it directly there.
Rename the function to vxlan_parse_gpe_proto to reflect what it now
does. Remove unused arguments skb and vxflags. Move the function earlier
in the file to allow it to be called from more places in the next patch.
Signed-off-by: Jiri Benc <jbenc@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
in atl1c_tso_csum, it should check the return value of pskb_trim(),
and return an error code if an unexpected value is returned
by pskb_trim().
Signed-off-by: Yuanjun Gong <ruc_gongyuanjun@163.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
VXLAN-GPE does not add an extra inner Ethernet header. Take that into
account when calculating header length.
This causes problems in skb_tunnel_check_pmtu, where incorrect PMTU is
cached.
In the collect_md mode (which is the only mode that VXLAN-GPE
supports), there's no magic auto-setting of the tunnel interface MTU.
It can't be, since the destination and thus the underlying interface
may be different for each packet.
So, the administrator is responsible for setting the correct tunnel
interface MTU. Apparently, the administrators are capable enough to
calculate that the maximum MTU for VXLAN-GPE is (their_lower_MTU - 36).
They set the tunnel interface MTU to 1464. If you run a TCP stream over
such interface, it's then segmented according to the MTU 1464, i.e.
producing 1514 bytes frames. Which is okay, this still fits the lower
MTU.
However, skb_tunnel_check_pmtu (called from vxlan_xmit_one) uses 50 as
the header size and thus incorrectly calculates the frame size to be
1528. This leads to ICMP too big message being generated (locally),
PMTU of 1450 to be cached and the TCP stream to be resegmented.
The fix is to use the correct actual header size, especially for
skb_tunnel_check_pmtu calculation.
Fixes: e1e5314de0 ("vxlan: implement GPE")
Signed-off-by: Jiri Benc <jbenc@redhat.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In dwrr mode, the default bandwidth weight of disabled tc is set to 0.
If the bandwidth weight is 0, the mode will change to sp.
Therefore, disabled tc default bandwidth weight need changed to 1,
and 0 is returned when query the bandwidth weight of disabled tc.
In addition, driver need stop configure bandwidth weight if tc is disabled.
Fixes: 848440544b ("net: hns3: Add support of TX Scheduler & Shaper to HNS3 driver")
Signed-off-by: Jie Wang <wangjie125@huawei.com>
Signed-off-by: Jijie Shao <shaojijie@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, the weight saved by the driver is used as the query result,
which may be different from the actual weight in the register.
Therefore, the register value read from the firmware is used
as the query result
Fixes: 0e32038dc8 ("net: hns3: refactor dump tc of debugfs")
Signed-off-by: Jijie Shao <shaojijie@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When the tm module is configured with traffic, traffic
may be abnormal. This patch fixes this problem.
Before the tm module is configured, traffic processing
should be stopped. After the tm module is configured,
traffic processing is enabled.
Signed-off-by: Hao Lan <lanhao@huawei.com>
Signed-off-by: Jijie Shao <shaojijie@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Current only the first 32 bits of the capability flag bit are considered.
When the matching capability flag bit is greater than 31 bits,
it will get an error bit.This patch use bitmap to solve this issue.
It can handle each capability bit whitout bit width limit.
Fixes: da77aef9cc ("net: hns3: create common cmdq resource allocate/free/query APIs")
Signed-off-by: Hao Lan <lanhao@huawei.com>
Signed-off-by: Jijie Shao <shaojijie@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Clear MV_V2_PORT_CTRL_PWRDOWN bit to set power up for 88x3310 PHY,
it sometimes does not take effect immediately. And a read of this
register causes the bit not to clear. This will cause mv3310_reset()
to time out, which will fail the config initialization. So add a delay
before the next access.
Fixes: c9cc1c815d ("net: phy: marvell10g: place in powersave mode at probe")
Signed-off-by: Jiawen Wu <jiawenwu@trustnetic.com>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
In iavf_adminq_task(), if the function can't acquire the
adapter->crit_lock, it checks if the driver is removing. If so, it simply
exits without re-enabling the interrupt. This is done to ensure that the
task stops processing as soon as possible once the driver is being removed.
However, if the IAVF_FLAG_PF_COMMS_FAILED is set, the function checks this
before attempting to acquire the lock. In this case, the function exits
early and re-enables the interrupt. This will happen even if the driver is
already removing.
Avoid this, by moving the check to after the adapter->crit_lock is
acquired. This way, if the driver is removing, we will not re-enable the
interrupt.
Fixes: fc2e6b3b13 ("iavf: Rework mutexes for better synchronisation")
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Rafal Romanowski <rafal.romanowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
