In some situations 255 bytes offset is not enough to match or manipulate
the desired packet field. Increase the offset limit to 65535 or U16_MAX.
In addition, the nla policy maximum value is not set anymore as it is
limited to s16. Instead, the maximum value is checked during the payload
expression initialization function.
Tested with the nft command line tool.
table ip filter {
chain output {
@nh,2040,8 set 0xff
@nh,524280,8 set 0xff
@nh,524280,8 0xff
@nh,2040,8 0xff
}
}
Signed-off-by: Fernando Fernandez Mancera <fmancera@suse.de>
Signed-off-by: Florian Westphal <fw@strlen.de>
Sebastian points out that avx2 depends on avx, see check_cpufeature_deps()
in arch/x86/kernel/cpu/cpuid-deps.c:
avx2 feature bit will be cleared when avx isn't available.
No functional change intended.
Reported-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Now that nft_setelem_flush is not called with rcu read lock held or
disabled softinterrupts anymore this can now use GFP_KERNEL too.
This is the last atomic allocation of transaction elements, so remove
all gfp_t arguments and the wrapper function.
This makes attempts to delete large sets much more reliable, before
this was prone to transient memory allocation failures.
Signed-off-by: Florian Westphal <fw@strlen.de>
Quoting Sven Auhagen:
we do see on occasions that we get the following error message, more so on
x86 systems than on arm64:
Error: Could not process rule: Cannot allocate memory delete table inet filter
It is not a consistent error and does not happen all the time.
We are on Kernel 6.6.80, seems to me like we have something along the lines
of the nf_tables: allow clone callbacks to sleep problem using GFP_ATOMIC.
As hinted at by Sven, this is because of GFP_ATOMIC allocations during
set flush.
When set is flushed, all elements are deactivated. This triggers a set
walk and each element gets added to the transaction list.
The rbtree and rhashtable sets don't allow the iter callback to sleep:
rbtree walk acquires read side of an rwlock with bh disabled, rhashtable
walk happens with rcu read lock held.
Rbtree is simple enough to resolve:
When the walk context is ITER_READ, no change is needed (the iter
callback must not deactivate elements; we're not in a transaction).
When the iter type is ITER_UPDATE, the rwlock isn't needed because the
caller holds the transaction mutex, this prevents any and all changes to
the ruleset, including add/remove of set elements.
Rhashtable is slightly more complex.
When the iter type is ITER_READ, no change is needed, like rbtree.
For ITER_UPDATE, we hold transaction mutex which prevents elements from
getting free'd, even outside of rcu read lock section.
So build a temporary list of all elements while doing the rcu iteration
and then call the iterator in a second pass.
The disadvantage is the need to iterate twice, but this cost comes with
the benefit to allow the iter callback to use GFP_KERNEL allocations in
a followup patch.
The new list based logic makes it necessary to catch recursive calls to
the same set earlier.
Such walk -> iter -> walk recursion for the same set can happen during
ruleset validation in case userspace gave us a bogus (cyclic) ruleset
where verdict map m jumps to chain that sooner or later also calls
"vmap @m".
Before the new ->in_update_walk test, the ruleset is rejected because the
infinite recursion causes ctx->level to exceed the allowed maximum.
But with the new logic added here, elements would get skipped:
nft_rhash_walk_update would see elements that are on the walk_list of
an older stack frame.
As all recursive calls into same map results in -EMLINK, we can avoid this
problem by using the new in_update_walk flag and reject immediately.
Next patch converts the problematic GFP_ATOMIC allocations.
Reported-by: Sven Auhagen <Sven.Auhagen@belden.com>
Closes: https://lore.kernel.org/netfilter-devel/BY1PR18MB5874110CAFF1ED098D0BC4E7E07BA@BY1PR18MB5874.namprd18.prod.outlook.com/
Signed-off-by: Florian Westphal <fw@strlen.de>
Open coded calculation can be avoided and replaced by the
equivalent csum_replace4() in nft_csum_replace().
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Florian Westphal <fw@strlen.de>
Remove array_size() calls and replace vmalloc() with vmalloc_array() to
simplify the code. vmalloc_array() is also optimized better, uses fewer
instructions, and handles overflow more concisely[1].
[1]: https://lore.kernel.org/lkml/abc66ec5-85a4-47e1-9759-2f60ab111971@vivo.com/
Signed-off-by: Qianfeng Rong <rongqianfeng@vivo.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Saeed Mahameed says:
====================
E-Switch vport sharing & delegation
An mlx5 E-Switch FDB table can manage vports belonging to other sibling
physical functions, such as ECPF (ARM embedded cores) and Host PF (x86).
This enables a single source of truth for SDN software to manage network
pipelines from one host. While such functionality already exists in mlx5,
it is currently limited by static vport allocation,
meaning the number of vports shared between multi-host functions
must be known pre-boot.
This patchset enables delegated/external vports to be discovered
dynamically when switchdev mode is enabled, leveraging new firmware
capabilities for dynamic vport creation.
Adjacent functions that delegate their SR-IOV VFs to sibling PFs, can be
dynamically discovered on the sibling PF's switchdev mode enabling,
after sriov was enabled on the originating PF, allowing for more
flexible and scalable management in multi-host and ECPF-to-host
scenarios.
The patchset consists of the following changes:
- Refactoring of ACL root namespace handling: The storage of vport ACL root
namespaces is converted from a linear array to an xarray, allowing dynamic
creation of ACLs per individual vport.
- Improvements for vhca_id to vport mapping.
- Dynamic querying and creation of delegated functions/vports.
====================
Link: https://patch.msgid.link/20250829223722.900629-1-saeed@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The mlx5 driver caches many capabilities to be used by mlx5 layers.
In SW and HW steering we can use the cached vhca_id instead of invoking
FW commands.
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250829223722.900629-8-saeed@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Adjacent vfs get their devlink port information from firmware,
use the information (pfnum, function id) from FW when populating the
devlink port attributes.
Before:
$ devlink port show
pci/0000:00:03.0/180225: type eth netdev eth0 flavour pcivf controller 0 pfnum 0 vfnum 49152 external false splittable false
function:
hw_addr 00:00:00:00:00:00
After:
$ devlink port show
pci/0000:00:03.0/180225: type eth netdev enp0s3npf0vf2 flavour pcivf controller 0 pfnum 0 vfnum 2 external false splittable false
function:
hw_addr 00:00:00:00:00:00
Signed-off-by: Adithya Jayachandran <ajayachandra@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250829223722.900629-7-saeed@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Register representors for adjacent vports dynamically when they are
discovered. Dynamically added representors state will now be set to
'REGISTERED' when the representor type was already registered,
otherwise they won't be loaded.
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250829223722.900629-6-saeed@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Use the new vport acl root namespace add/remove API to create the
missing acl root name spaces per each new adjacent function vport.
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250829223722.900629-5-saeed@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Adding driver support to query adjacent functions vports, AKA
delegated vports.
Adjacent functions can delegate their sriov vfs to other sibling PF in
the system, to be managed by the eswitch capable sibling PF.
E.g, ECPF to Host PF, multi host PF between each other, etc.
Only supported in switchdev mode.
Signed-off-by: Adithya Jayachandran <ajayachandra@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250829223722.900629-4-saeed@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Move the loop that creates the vports ACLs root name spaces to eswitch,
since it is the eswitch responsibility to decide when and how many
vports ACLs root namespaces to create, in the next patches we will use
the fs_core vport ACL root namespace APIs to create/remove root ns
ACLs dynamically for dynamically created vports.
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250829223722.900629-3-saeed@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Before this patch it was a linear array and could only support a certain
number of vports, in the next patches, vport numbers are not bound to a
well known limit, thus convert acl root name space storage to xarray.
In addition create fs_core public API to add/remove vport acl namespaces
as it is the eswitch responsibility to create the vports and their
root name spaces for acls, in the next patch we will move
mlx5_fs_ingress_acls_{init,cleanup} to eswitch and will use
the individual mlx5_fs_vport_{egress,ingresS}_acl_ns_{add,remove}
APIs for dynamically create vports.
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250829223722.900629-2-saeed@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Wei Fang says:
====================
Add NETC Timer PTP driver and add PTP support for i.MX95
This series adds NETC Timer PTP clock driver, which supports precise
periodic pulse, time capture on external pulse and PTP synchronization.
It also adds PTP support to the enetc v4 driver for i.MX95 and optimizes
the PTP-related code in the enetc driver.
====================
Link: https://patch.msgid.link/20250829050615.1247468-1-wei.fang@nxp.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
For ENETC v4, the hardware has the capability to support Tx checksum
offload. so the enetc driver does not need to update the UDP checksum
of PTP sync packets if Tx checksum offload is enabled.
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Link: https://patch.msgid.link/20250829050615.1247468-15-wei.fang@nxp.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Regarding PTP, ENETC v4 has some changes compared to ENETC v1 (LS1028A),
mainly as follows.
1. ENETC v4 uses a different PTP driver, so the way to get phc_index is
different from LS1028A. Therefore, enetc_get_ts_info() has been modified
appropriately to be compatible with ENETC v1 and v4.
2. The PMa_SINGLE_STEP register has changed in ENETC v4, not only the
register offset, but also some register fields. Therefore, two helper
functions are added, enetc_set_one_step_ts() for ENETC v1 and
enetc4_set_one_step_ts() for ENETC v4.
3. Since the generic helper functions from ptp_clock are used to get
the PHC index of the PTP clock, so FSL_ENETC_CORE depends on Kconfig
symbol "PTP_1588_CLOCK_OPTIONAL". But FSL_ENETC_CORE can only be
selected, so add the dependency to FSL_ENETC, FSL_ENETC_VF and
NXP_ENETC4. Perhaps the best approach would be to change FSL_ENETC_CORE
to a visible menu entry. Then make FSL_ENETC, FSL_ENETC_VF, and
NXP_ENETC4 depend on it, but this is not the goal of this patch, so this
may be changed in the future.
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Link: https://patch.msgid.link/20250829050615.1247468-14-wei.fang@nxp.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Move sync packet content modification before dma_map_single() to follow
correct DMA usage process, even though the previous sequence worked due
to hardware DMA-coherence support (LS1028A). But for the upcoming i.MX95,
its ENETC (v4) does not support "dma-coherent", so this step is very
necessary. Otherwise, the originTimestamp and correction fields of the
sent packets will still be the values before the modification.
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Link: https://patch.msgid.link/20250829050615.1247468-13-wei.fang@nxp.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The ENETC_F_RX_TSTAMP flag of priv->active_offloads can only be set when
CONFIG_FSL_ENETC_PTP_CLOCK is enabled. Similarly, rx_ring->ext_en can
only be set when CONFIG_FSL_ENETC_PTP_CLOCK is enabled as well. So it is
safe to remove unnecessary CONFIG_FSL_ENETC_PTP_CLOCK check.
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Reviewed-by: Vadim Fedorenko <vadim.fedorenko@linux.dev>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Link: https://patch.msgid.link/20250829050615.1247468-12-wei.fang@nxp.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Move PTP Sync packet processing from enetc_map_tx_buffs() to a new helper
function enetc_update_ptp_sync_msg() to simplify the original function.
Prepare for upcoming ENETC v4 one-step support. There is no functional
change. It is worth mentioning that ENETC_TXBD_TSTAMP is added to replace
0x3fffffff.
Prepare for upcoming ENETC v4 one-step support.
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Link: https://patch.msgid.link/20250829050615.1247468-11-wei.fang@nxp.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Currently, the Tx PTP packets are parsed twice in the enetc driver, once
in enetc_xmit() and once in enetc_map_tx_buffs(). The latter is duplicate
and is unnecessary, since the parsed information can be saved to skb->cb
so that enetc_map_tx_buffs() can get the previously parsed data from
skb->cb. Therefore, add struct enetc_skb_cb as the format of the data
in the skb->cb buffer to save the parsed information of PTP packet. Use
saved information in enetc_map_tx_buffs() to avoid parsing data again.
In addition, rename variables offset1 and offset2 in enetc_map_tx_buffs()
to corr_off and tstamp_off for better readability.
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Link: https://patch.msgid.link/20250829050615.1247468-10-wei.fang@nxp.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The NETC Timer is capable of recording the timestamp on receipt of an
external pulse on a GPIO pin. It supports two such external triggers.
The recorded value is saved in a 16 entry FIFO accessed by
TMR_ETTSa_H/L. An interrupt can be generated when the trigger occurs,
when the FIFO reaches a threshold or overflows.
Signed-off-by: F.S. Peng <fushi.peng@nxp.com>
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Link: https://patch.msgid.link/20250829050615.1247468-8-wei.fang@nxp.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
NETC Timer has three pulse channels, all of which support periodic pulse
output. Bind the channel to a ALARM register and then sets a future time
into the ALARM register. When the current time is greater than the ALARM
value, the FIPER register will be triggered to count down, and when the
count reaches 0, the pulse will be triggered. The PPS signal is also
implemented in this way.
i.MX95 only has ALARM1 can be used as an indication to the FIPER start
down counting, but i.MX943 has ALARM1 and ALARM2 can be used. Therefore,
only one channel can work for i.MX95, two channels for i.MX943 as most.
In addition, change the PPS channel to be dynamically selected from fixed
number (0) because add PTP_CLK_REQ_PEROUT support.
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Link: https://patch.msgid.link/20250829050615.1247468-7-wei.fang@nxp.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The NETC Timer is capable of generating a PPS interrupt to the host. To
support this feature, a 64-bit alarm time (which is a integral second
of PHC in the future) is set to TMR_ALARM, and the period is set to
TMR_FIPER. The alarm time is compared to the current time on each update,
then the alarm trigger is used as an indication to the TMR_FIPER starts
down counting. After the period has passed, the PPS event is generated.
According to the NETC block guide, the Timer has three FIPERs, any of
which can be used to generate the PPS events, but in the current
implementation, we only need one of them to implement the PPS feature,
so FIPER 0 is used as the default PPS generator. Also, the Timer has
2 ALARMs, currently, ALARM 0 is used as the default time comparator.
However, if the time is adjusted or the integer of period is changed when
PPS is enabled, the PPS event will not be generated at an integral second
of PHC. The suggested steps from IP team if time drift happens:
1. Disable FIPER before adjusting the hardware time
2. Rearm ALARM after the time adjustment to make the next PPS event be
generated at an integral second of PHC.
3. Re-enable FIPER.
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Link: https://patch.msgid.link/20250829050615.1247468-6-wei.fang@nxp.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
NETC V4 Timer provides current time with nanosecond resolution, precise
periodic pulse, pulse on timeout (alarm), and time capture on external
pulse support. And it supports time synchronization as required for
IEEE 1588 and IEEE 802.1AS-2020.
Inside NETC, ENETC can capture the timestamp of the sent/received packet
through the PHC provided by the Timer and record it on the Tx/Rx BD. And
through the relevant PHC interfaces provided by the driver, the enetc V4
driver can support PTP time synchronization.
In addition, NETC V4 Timer is similar to the QorIQ 1588 timer, but it is
not exactly the same. The current ptp-qoriq driver is not compatible with
NETC V4 Timer, most of the code cannot be reused, see below reasons.
1. The architecture of ptp-qoriq driver makes the register offset fixed,
however, the offsets of all the high registers and low registers of V4
are swapped, and V4 also adds some new registers. so extending ptp-qoriq
to make it compatible with V4 Timer is tantamount to completely rewriting
ptp-qoriq driver.
2. The usage of some functions is somewhat different from QorIQ timer,
such as the setting of TCLK_PERIOD and TMR_ADD, the logic of configuring
PPS, etc., so making the driver compatible with V4 Timer will undoubtedly
increase the complexity of the code and reduce readability.
3. QorIQ is an expired brand. It is difficult for us to verify whether
it works stably on the QorIQ platforms if we refactor the driver, and
this will make maintenance difficult, so refactoring the driver obviously
does not bring any benefits.
Therefore, add this new driver for NETC V4 Timer. Note that the missing
features like PEROUT, PPS and EXTTS will be added in subsequent patches.
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Link: https://patch.msgid.link/20250829050615.1247468-5-wei.fang@nxp.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Some Ethernet controllers do not have an integrated PTP timer function.
Instead, the PTP timer is a separated device and provides PTP hardware
clock to the Ethernet controller to use. Therefore, the Ethernet
controller driver needs to obtain the PTP clock's phc_index in its
ethtool_ops::get_ts_info(). Currently, most drivers implement this in
the following ways.
1. The PTP device driver adds a custom API and exports it to the Ethernet
controller driver.
2. The PTP device driver adds private data to its device structure. So
the private data structure needs to be exposed to the Ethernet controller
driver.
When registering the ptp clock, ptp_clock_register() always saves the
ptp_clock pointer to the private data of ptp_clock::dev. Therefore, as
long as ptp_clock::dev is obtained, the phc_index can be obtained. So
the following generic APIs can be added to the ptp driver to obtain the
phc_index.
1. ptp_clock_index_by_dev(): Obtain the phc_index by the device pointer
of the PTP device.
2.ptp_clock_index_by_of_node(): Obtain the phc_index by the of_node
pointer of the PTP device.
Also, we can add another API like ptp_clock_index_by_fwnode() to get the
phc_index by fwnode of PTP device. However, this API is not used in this
patch set, so it is better to add it when needed.
Suggested-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Link: https://patch.msgid.link/20250829050615.1247468-4-wei.fang@nxp.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
For some Ethernet controllers, the PTP timer function is not integrated.
Instead, the PTP timer is a separate device and provides PTP Hardware
Clock (PHC) to the Ethernet controller to use, such as NXP FMan MAC,
ENETC, etc. Therefore, a property is needed to indicate this hardware
relationship between the Ethernet controller and the PTP timer.
Since this use case is also very common, it is better to add a generic
property to ethernet-controller.yaml. According to the existing binding
docs, there are two good candidates, one is the "ptp-timer" defined in
fsl,fman-dtsec.yaml, and the other is the "ptimer-handle" defined in
fsl,fman.yaml. From the perspective of the name, the former is more
straightforward, so move the "ptp-timer" from fsl,fman-dtsec.yaml to
ethernet-controller.yaml.
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Reviewed-by: Rob Herring (Arm) <robh@kernel.org>
Link: https://patch.msgid.link/20250829050615.1247468-3-wei.fang@nxp.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
NXP NETC (Ethernet Controller) is a multi-function PCIe Root Complex
Integrated Endpoint (RCiEP), the Timer is one of its functions which
provides current time with nanosecond resolution, precise periodic
pulse, pulse on timeout (alarm), and time capture on external pulse
support. And also supports time synchronization as required for IEEE
1588 and IEEE 802.1AS-2020. So add device tree binding doc for the PTP
clock based on NETC Timer.
NETC Timer has three reference clock sources, but the clock mux is inside
the IP. Therefore, the driver will parse the clock name to select the
desired clock source. If the clocks property is not present, NETC Timer
will use the system clock of NETC IP as its reference clock. Because the
Timer is a PCIe function of NETC IP, the system clock of NETC is always
available to the Timer.
Signed-off-by: Wei Fang <wei.fang@nxp.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://patch.msgid.link/20250829050615.1247468-2-wei.fang@nxp.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Eric Dumazet says:
====================
inet: ping: misc changes
First and third patches improve security a bit.
Second patch (ping_hash removal) is a cleanup.
Fourth patch uses EXPORT_IPV6_MOD[_GPL].
====================
Link: https://patch.msgid.link/20250829153054.474201-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Provide isolation between netns for ping idents.
Randomize initial ping_port_rover value at netns creation.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20250829153054.474201-4-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
There is no point in keeping ping_hash().
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Reviewed-by: Yue Haibing <yuehaibing@huawei.com>
Link: https://patch.msgid.link/20250829153054.474201-3-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
We need to check socket netns before considering them in ping_get_port().
Otherwise, one malicious netns could 'consume' all ports.
Add corresponding check in ping_lookup().
Fixes: c319b4d76b ("net: ipv4: add IPPROTO_ICMP socket kind")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Reviewed-by: Yue Haibing <yuehaibing@huawei.com>
Link: https://patch.msgid.link/20250829153054.474201-2-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Commit 3c7826d0b1 ("net: stmmac: Separate C22 and C45 transactions
for xgmac") missed a change that happened in commit e2d0acd40c
("net: stmmac: using pm_runtime_resume_and_get instead of
pm_runtime_get_sync").
Update the two clause 45 functions that didn't get switched to
pm_runtime_resume_and_get().
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://patch.msgid.link/E1urv09-00000000gJ1-3SxO@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Fix several spelling and grammatical mistakes in output messages from
the net selftests to improve readability.
Only the message strings for the test output have been modified. No
changes to the functional logic of the tests have been made.
Signed-off-by: Praveen Balakrishnan <praveen.balakrishnan@magd.ox.ac.uk>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250828211100.51019-1-praveen.balakrishnan@magd.ox.ac.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Networking drivers implementing PTP clocks and kernel socket code
handling hardware timestamps use the 64-bit signed ktime_t type counting
nanoseconds. When a PTP clock reaches the maximum value in year 2262,
the timestamps returned to applications will overflow into year 1667.
The same thing happens when injecting a large offset with
clock_adjtime(ADJ_SETOFFSET).
The commit 7a8e61f847 ("timekeeping: Force upper bound for setting
CLOCK_REALTIME") limited the maximum accepted value setting the system
clock to 30 years before the maximum representable value (i.e. year
2232) to avoid the overflow, assuming the system will not run for more
than 30 years.
Enforce the same limit for PTP clocks. Don't allow negative values and
values closer than 30 years to the maximum value. Drivers may implement
an even lower limit if the hardware registers cannot represent the whole
interval between years 1970 and 2262 in the required resolution.
Signed-off-by: Miroslav Lichvar <mlichvar@redhat.com>
Cc: Richard Cochran <richardcochran@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: John Stultz <jstultz@google.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Vadim Fedorenko <vadim.fedorenko@linux.dev>
Link: https://patch.msgid.link/20250828103300.1387025-1-mlichvar@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The Qualcomm Technologies, Inc. Packet Process Engine (PPE) is only
present on Qualcomm IPQ SoCs. Hence add a dependency on ARCH_QCOM, to
prevent asking the user about this driver when configuring a kernel
without Qualcomm platform support,
Fixes: 353a0f1d5b ("net: ethernet: qualcomm: Add PPE driver for IPQ9574 SoC")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/eb7bd6e6ce27eb6d602a63184d9daa80127e32bd.1756466786.git.geert+renesas@glider.be
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
TCP tracks the number of orphaned (SOCK_DEAD but not yet destructed)
sockets in tcp_orphan_count.
In some code that was shared with DCCP, tcp_orphan_count is referenced
via sk->sk_prot->orphan_count.
Let's reference tcp_orphan_count directly.
inet_csk_prepare_for_destroy_sock() is moved to inet_connection_sock.c
due to header dependency.
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Reviewed-by: Jason Xing <kerneljasonxing@gmail.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20250829215641.711664-1-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Eric Dumazet says:
====================
net: add rcu safety to dst->dev
Followup of commit 88fe14253e ("net: dst: add four helpers
to annotate data-races around dst->dev").
Use lockdep enabled helpers to convert our unsafe dst->dev
uses one at a time.
More to come...
====================
Link: https://patch.msgid.link/20250828195823.3958522-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Change icmpv4_xrlim_allow(), ip_defrag() to prevent possible UAF.
Change ipmr_prepare_xmit(), ipmr_queue_fwd_xmit(), ip_mr_output(),
ipv4_neigh_lookup() to use lockdep enabled dst_dev_rcu().
Fixes: 4a6ce2b6f2 ("net: introduce a new function dst_dev_put()")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20250828195823.3958522-9-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Use RCU to avoid a pair of atomic operations and a potential
UAF on dst_dev()->flags.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20250828195823.3958522-8-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Replace three dst_dev() with a lockdep enabled helper.
Fixes: 4a6ce2b6f2 ("net: introduce a new function dst_dev_put()")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20250828195823.3958522-7-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Use RCU to protect accesses to dst->dev from sk_setup_caps()
and sk_dst_gso_max_size().
Also use dst_dev_rcu() in ip6_dst_mtu_maybe_forward(),
and ip_dst_mtu_maybe_forward().
ip4_dst_hoplimit() can use dst_dev_net_rcu().
Fixes: 4a6ce2b6f2 ("net: introduce a new function dst_dev_put()")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20250828195823.3958522-6-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Use RCU in ip6_output() in order to use dst_dev_rcu() to prevent
possible UAF.
We can remove rcu_read_lock()/rcu_read_unlock() pairs
from ip6_finish_output2().
Fixes: 4a6ce2b6f2 ("net: introduce a new function dst_dev_put()")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20250828195823.3958522-5-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Use RCU in ip6_xmit() in order to use dst_dev_rcu() to prevent
possible UAF.
Fixes: 4a6ce2b6f2 ("net: introduce a new function dst_dev_put()")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20250828195823.3958522-4-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Refactor icmpv6_xrlim_allow() and ip6_dst_hoplimit()
so that we acquire rcu_read_lock() a bit longer
to be able to use dst_dev_rcu() instead of dst_dev().
__ip6_rt_update_pmtu() and rt6_do_redirect can directly
use dst_dev_rcu() in sections already holding rcu_read_lock().
Small changes to use dst_dev_net_rcu() in
ip6_default_advmss(), ipv6_sock_ac_join(),
ip6_mc_find_dev() and ndisc_send_skb().
Fixes: 4a6ce2b6f2 ("net: introduce a new function dst_dev_put()")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20250828195823.3958522-3-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Followup of commit 88fe14253e ("net: dst: add four helpers
to annotate data-races around dst->dev").
We want to gradually add explicit RCU protection to dst->dev,
including lockdep support.
Add an union to alias dst->dev_rcu and dst->dev.
Add dst_dev_net_rcu() helper.
Fixes: 4a6ce2b6f2 ("net: introduce a new function dst_dev_put()")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20250828195823.3958522-2-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Eric Dumazet says:
====================
inet_diag: make dumps faster with simple filters
inet_diag_bc_sk() pulls five cache lines per socket,
while most filters only need the two first ones.
We can change it to only pull needed cache lines,
to make things like "ss -temoi src :21456" much faster.
First patches (1-3) are annotating data-races as a first step.
====================
Link: https://patch.msgid.link/20250828102738.2065992-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
