net: phy: nxp-c45-tja11xx: prepare the ground for TJA1120
Between TJA1120 and TJA1103 the hardware was improved, but some register
addresses were changed and some bit fields were moved from one register
to another.
Introduce the nxp_c45_reg_field structure and its associated functions to
abstract the differences between the PHYs.
Remove the defined bits and register addresses that are not common
between TJA1103 and TJA1120 and replace them with reg_fields and
register addresses from phydev->drv->driver_data.
net: phy: nxp-c45-tja11xx: use phylib master/slave implementation
Remove the custom implementation of master/save setup and read status
and use genphy_c45_config_aneg and genphy_c45_read_status since phylib
has support for master/slave setup and master/slave status.
====================
virtio_net: add per queue interrupt coalescing support
Currently, coalescing parameters are grouped for all transmit and receive
virtqueues. This patch series add support to set or get the parameters for
a specified virtqueue.
When the traffic between virtqueues is unbalanced, for example, one virtqueue
is busy and another virtqueue is idle, then it will be very useful to
control coalescing parameters at the virtqueue granularity.
Example command:
$ ethtool -Q eth5 queue_mask 0x1 --coalesce tx-packets 10
Would set max_packets=10 to VQ 1.
$ ethtool -Q eth5 queue_mask 0x1 --coalesce rx-packets 10
Would set max_packets=10 to VQ 0.
$ ethtool -Q eth5 queue_mask 0x1 --show-coalesce
Queue: 0
Adaptive RX: off TX: off
stats-block-usecs: 0
sample-interval: 0
pkt-rate-low: 0
pkt-rate-high: 0
Gavin Li [Mon, 31 Jul 2023 07:06:55 +0000 (10:06 +0300)]
virtio_net: support per queue interrupt coalesce command
Add interrupt_coalesce config in send_queue and receive_queue to cache user
config.
Send per virtqueue interrupt moderation config to underlying device in
order to have more efficient interrupt moderation and cpu utilization of
guest VM.
Additionally, address all the VQs when updating the global configuration,
as now the individual VQs configuration can diverge from the global
configuration.
net/macmace: Replace zero-length array with DECLARE_FLEX_ARRAY() helper
Since zero-length arrays are deprecated, we are replacing
them with C99 flexible-array members. As a result, instead
of declaring a zero-length array, use the new
DECLARE_FLEX_ARRAY() helper macro.
This fixes warnings such as:
./drivers/net/ethernet/apple/macmace.c:80:4-8: WARNING use flexible-array member instead (https://www.kernel.org/doc/html/latest/process/deprecated.html#zero-length-and-one-element-arrays)
net: dsa: qca8k: limit user ports access to the first CPU port on setup
In preparation for multi-CPU support, set CPU port LOOKUP MEMBER outside
the port loop and setup the LOOKUP MEMBER mask for user ports only to
the first CPU port.
This is to handle flooding condition where every CPU port is set as
target and prevent packet duplication for unknown frames from user ports.
Secondary CPU port LOOKUP MEMBER mask will be setup later when
port_change_master will be implemented.
net: dsa: qca8k: make learning configurable and keep off if standalone
Address learning should initially be turned off by the driver for port
operation in standalone mode, then the DSA core handles changes to it
via ds->ops->port_bridge_flags().
Currently this is not the case for qca8k where learning is enabled
unconditionally in qca8k_setup for every user port.
Handle ports configured in standalone mode by making the learning
configurable and not enabling it by default.
Implement .port_pre_bridge_flags and .port_bridge_flags dsa ops to
enable learning for bridge that request it and tweak
.port_stp_state_set to correctly disable learning when port is
configured in standalone mode.
====================
net/sched: improve class lifetime handling
Valis says[0]:
============
Three classifiers (cls_fw, cls_u32 and cls_route) always copy
tcf_result struct into the new instance of the filter on update.
This causes a problem when updating a filter bound to a class,
as tcf_unbind_filter() is always called on the old instance in the
success path, decreasing filter_cnt of the still referenced class
and allowing it to be deleted, leading to a use-after-free.
============
Turns out these could have been spotted easily with proper warnings.
Improve the current class lifetime with wrappers that check for
overflow/underflow.
While at it add an extack for when a class in use is deleted.
Pedro Tammela [Fri, 28 Jul 2023 15:35:33 +0000 (12:35 -0300)]
net/sched: wrap open coded Qdics class filter counter
The 'filter_cnt' counter is used to control a Qdisc class lifetime.
Each filter referecing this class by its id will eventually
increment/decrement this counter in their respective
'add/update/delete' routines.
As these operations are always serialized under rtnl lock, we don't
need an atomic type like 'refcount_t'.
It also means that we lose the overflow/underflow checks already
present in refcount_t, which are valuable to hunt down bugs
where the unsigned counter wraps around as it aids automated tools
like syzkaller to scream in such situations.
Wrap the open coded increment/decrement into helper functions and
add overflow checks to the operations.
====================
mptcp: cleanup and improvements in the selftests
This small series of 4 patches adds some improvements in MPTCP
selftests:
- Patch 1 reworks the detailed report of mptcp_join.sh selftest to
better display what went well or wrong per test.
- Patch 2 adds colours (if supported, forced and/or not disabled) in
mptcp_join.sh selftest output to help spotting issues.
- Patch 3 modifies an MPTCP selftest tool to interact with the
path-manager via Netlink to always look for errors if any. This makes
sure odd behaviours can be seen in the logs and errors can be caught
later if needed.
- Patch 4 removes stdout and stderr redirections to /dev/null when using
pm_nl_ctl if no errors are expected in order to log odd behaviours.
====================
All pm_nl_ctl commands were muted. If there was an unexpected error with
one of them, this was simply not visible in the logs, making the
analysis very hard. It could also hide misuse of commands by mistake.
Now the output is only muted when we do expect to have an error, e.g.
when giving invalid arguments on purpose.
selftests: mptcp: pm_nl_ctl: always look for errors
If a Netlink command for the MPTCP path-managers is not valid, it is
important to check if there are errors. If yes, they need to be reported
instead of being ignored and exiting without errors.
Now if no replies are expected, an ACK from the kernelspace is asked by
the userspace in order to always expect a reply. We can use the same
buffer that is currently always >1024 bytes. Then we can check if there
is an error (err->error), print it if any and report the error.
After this modification, it is required to mute expected errors in
mptcp_join.sh and pm_netlink.sh selftests:
- when trying to add a bad endpoint, e.g. duplicated
- when trying to set the two limits above the hard limit
This patch modifies how the detailed results are printed, mainly to
improve what is displayed in case of issue:
- Now the test name (title) is printed earlier, when starting the test
if it is not intentionally skipped: by doing that, errors linked to
a test will be printed after having written the test name and then
avoid confusions.
- Due to the previous item, it is required to add a new line after
having printed the test name because in case of error with a command,
it is better not to have the output in the middle of the screen.
- Each check is printed on a dedicated line with aligned status (ok,
skip, fail): it is easier to spot which one has failed, simpler to
manage in the code not having to deal with alignment case by case and
helpers can be used to uniform what is done. These helpers can also be
useful later to do more actions depending on the results or change in
one place what is printed.
- Info messages have been reduced and aligned as well. And info messages
about the creation of the default test files of 1 KB are no longer
printed.
Example:
001 no JOIN
syn [ ok ]
synack [ ok ]
ack [ ok ]
Or with a skip and a failure:
001 no JOIN
syn [ ok ]
synack [fail] got 42 JOIN[s] synack expected 0
Server ns stats
(...)
Client ns stats
(...)
ack [skip]
Or with info:
104 Infinite map
Test file (size 128 KB) for client
Test file (size 128 KB) for server
file received by server has inverted byte at 169
5 corrupted pkts
syn [ ok ]
synack [ ok ]
While at it, verify_listener_events() now also print more info in case
of failure and in pm_nl_check_endpoint(), the test is marked as failed
instead of skipped if no ID has been given (internal selftest issue).
The syzbot reproducer triggered a case that the qdisc refcnt is not
zero during dev_shutdown().
tcx_uninstall() will then WARN_ON_ONCE(tcx_entry(entry)->miniq_active)
because the miniq is still active and the entry should not be freed.
The latter assumed that qdisc destruction happens before tcx teardown.
This fix is to avoid tcx_uninstall() doing tcx_entry_free() when the
miniq is still alive and let the clsact_destroy() do the free later, so
that we do not assume any specific ordering for either of them.
If still active, tcx_uninstall() does clear the entry when flushing out
the prog/link. clsact_destroy() will then notice the "!tcx_entry_is_active()"
and then does the tcx_entry_free() eventually.
commit f9aab6f2ce57 ("net/smc: immediate freeing in smc_lgr_cleanup_early()")
left behind smc_lgr_schedule_free_work_fast() declaration.
And since commit 349d43127dac ("net/smc: fix kernel panic caused by race of smc_sock")
smc_ib_modify_qp_reset() is not used anymore.
commit 8a59f9d1e3d4 ("sock: Introduce sk->sk_prot->psock_update_sk_prot()")
left behind tcp_bpf_get_proto() declaration. And tcp_v4_tw_remember_stamp()
function is remove in ccb7c410ddc0 ("timewait_sock: Create and use getpeer op.").
Since commit 686989700cab ("tcp: simplify tcp_mark_skb_lost")
tcp_skb_mark_lost_uncond_verify() declaration is not used anymore.
net: Use sockaddr_storage for getsockopt(SO_PEERNAME).
Commit df8fc4e934c1 ("kbuild: Enable -fstrict-flex-arrays=3") started
applying strict rules to standard string functions.
It does not work well with conventional socket code around each protocol-
specific sockaddr_XXX struct, which is cast from sockaddr_storage and has
a bigger size than fortified functions expect. See these commits:
commit 06d4c8a80836 ("af_unix: Fix fortify_panic() in unix_bind_bsd().")
commit ecb4534b6a1c ("af_unix: Terminate sun_path when bind()ing pathname socket.")
commit a0ade8404c3b ("af_packet: Fix warning of fortified memcpy() in packet_getname().")
We must cast the protocol-specific address back to sockaddr_storage
to call such functions.
However, in the case of getsockaddr(SO_PEERNAME), the rationale is a bit
unclear as the buffer is defined by char[128] which is the same size as
sockaddr_storage.
Yang Li [Thu, 27 Jul 2023 00:57:41 +0000 (08:57 +0800)]
team: Remove NULL check before dev_{put, hold}
The call netdev_{put, hold} of dev_{put, hold} will check NULL,
so there is no need to check before using dev_{put, hold},
remove it to silence the warning:
./drivers/net/team/team.c:2325:3-10: WARNING: NULL check before dev_{put, hold} functions is not needed.
Referring to platform_get_irq()'s definition, the return value has
already been checked, error message also been printed via
dev_err_probe() if ret < 0. Calling dev_err_probe() one more time
outside platform_get_irq() is obviously redundant.
Removing dev_err_probe() outside platform_get_irq() to clean up
above problem.
Hayes Wang [Wed, 26 Jul 2023 03:08:07 +0000 (11:08 +0800)]
r8152: adjust generic_ocp_write function
Reduce the control transfer if all bytes of first or the last DWORD are
written.
The original method is to split the control transfer into three parts
(the first DWORD, middle continuous data, and the last DWORD). However,
they could be combined if whole bytes of the first DWORD or last DWORD
are written. That is, the first DWORD or the last DWORD could be combined
with the middle continuous data, if the byte_en is 0xff.
====================
In-kernel support for the TLS Alert protocol
IMO the kernel doesn't need user space (ie, tlshd) to handle the TLS
Alert protocol. Instead, a set of small helper functions can be used
to handle sending and receiving TLS Alerts for in-kernel TLS
consumers.
====================
Merged on top of a tag in case it's needed in the NFS tree.
Chuck Lever [Thu, 27 Jul 2023 17:37:37 +0000 (13:37 -0400)]
SUNRPC: Use new helpers to handle TLS Alerts
Use the helpers to parse the level and description fields in
incoming alerts. "Warning" alerts are discarded, and "fatal"
alerts mean the session is no longer valid.
Chuck Lever [Thu, 27 Jul 2023 17:35:50 +0000 (13:35 -0400)]
net/tls: Add TLS Alert definitions
I'm about to add support for kernel handshake API consumers to send
TLS Alerts, so introduce the needed protocol definitions in the new
header tls_prot.h.
This presages support for Closure alerts. Also, support for alerts
is a pre-requite for handling session re-keying, where one peer will
signal the need for a re-key by sending a TLS Alert.
Chuck Lever [Thu, 27 Jul 2023 17:35:23 +0000 (13:35 -0400)]
net/tls: Move TLS protocol elements to a separate header
Kernel TLS consumers will need definitions of various parts of the
TLS protocol, but often do not need the function declarations and
other infrastructure provided in <net/tls.h>.
Break out existing standardized protocol elements into a separate
header, and make room for a few more elements in subsequent patches.
Jakub Kicinski [Thu, 27 Jul 2023 19:07:26 +0000 (12:07 -0700)]
eth: bnxt: fix warning for define in struct_group
Fix C=1 warning with sparse 0.6.4:
drivers/net/ethernet/broadcom/bnxt/bnxt.c: note: in included file:
drivers/net/ethernet/broadcom/bnxt/bnxt_dcb.h:30:1: warning: directive in macro's argument list
Jakub Kicinski [Thu, 27 Jul 2023 19:07:25 +0000 (12:07 -0700)]
eth: bnxt: fix one of the W=1 warnings about fortified memcpy()
Fix a W=1 warning with gcc 13.1:
In function ‘fortify_memcpy_chk’,
inlined from ‘bnxt_hwrm_queue_cos2bw_cfg’ at drivers/net/ethernet/broadcom/bnxt/bnxt_dcb.c:133:3:
include/linux/fortify-string.h:592:25: warning: call to ‘__read_overflow2_field’ declared with attribute warning: detected read beyond size of field (2nd parameter); maybe use struct_group()? [-Wattribute-warning]
592 | __read_overflow2_field(q_size_field, size);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The field group is already defined and starts at queue_id:
Jakub Kicinski [Fri, 28 Jul 2023 20:41:59 +0000 (13:41 -0700)]
Merge tag 'mlx5-updates-2023-07-24' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux
Saeed Mahameed says:
====================
mlx5-updates-2023-07-24
1) Generalize devcom implementation to be independent of number of ports
or device's GUID.
2) Save memory on command interface statistics.
3) General code cleanups
* tag 'mlx5-updates-2023-07-24' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux:
net/mlx5: Give esw_offloads_load/unload_rep() "mlx5_" prefix
net/mlx5: Make mlx5_eswitch_load/unload_vport() static
net/mlx5: Make mlx5_esw_offloads_rep_load/unload() static
net/mlx5: Remove pointless devlink_rate checks
net/mlx5: Don't check vport->enabled in port ops
net/mlx5e: Make flow classification filters static
net/mlx5e: Remove duplicate code for user flow
net/mlx5: Allocate command stats with xarray
net/mlx5: split mlx5_cmd_init() to probe and reload routines
net/mlx5: Remove redundant cmdif revision check
net/mlx5: Re-organize mlx5_cmd struct
net/mlx5e: E-Switch, Allow devcom initialization on more vports
net/mlx5e: E-Switch, Register devcom device with switch id key
net/mlx5: Devcom, Infrastructure changes
net/mlx5: Use shared code for checking lag is supported
====================
====================
mlxsw: Avoid non-tracker helpers when holding and putting netdevices
Using the tracking helpers, netdev_hold() and netdev_put(), makes it easier
to debug netdevice refcount imbalances when CONFIG_NET_DEV_REFCNT_TRACKER
is enabled. For example, the following traceback shows the callpath to the
point of an outstanding hold that was never put:
unregister_netdevice: waiting for swp3 to become free. Usage count = 6
ref_tracker: eth%d@ffff888123c9a580 has 1/5 users at
mlxsw_sp_switchdev_event+0x6bd/0xcc0 [mlxsw_spectrum]
notifier_call_chain+0xbf/0x3b0
atomic_notifier_call_chain+0x78/0x200
br_switchdev_fdb_notify+0x25f/0x2c0 [bridge]
fdb_notify+0x16a/0x1a0 [bridge]
[...]
In this patchset, get rid of all non-ref-tracking helpers in mlxsw.
- Patch #1 drops two functions that are not used anymore, but contain
dev_hold() / dev_put() calls.
- Patch #2 avoids taking a reference in one function which is called
under RTNL.
- The remaining patches convert individual hold/put sites one by one
from trackerless to tracker-enabled.
Petr Machata [Thu, 27 Jul 2023 15:59:20 +0000 (17:59 +0200)]
mlxsw: spectrum_nve: Do not take reference when looking up netdevice
mlxsw_sp_nve_fid_disable() is always called under RTNL. It is therefore
safe to call __dev_get_by_index() to get the netdevice pointer without
bumping the reference count, because we can be sure the netdevice is not
going away. That then obviates the need to put the netdevice later in the
function.
Petr Machata [Thu, 27 Jul 2023 15:59:19 +0000 (17:59 +0200)]
mlxsw: spectrum: Drop unused functions mlxsw_sp_port_lower_dev_hold/_put()
As of commit 151b89f6025a ("mlxsw: spectrum_router: Reuse work neighbor
initialization in work scheduler"), the functions
mlxsw_sp_port_lower_dev_hold() and mlxsw_sp_port_dev_put() have no users.
Drop them.
Patrick Rohr [Wed, 26 Jul 2023 23:07:01 +0000 (16:07 -0700)]
net: change accept_ra_min_rtr_lft to affect all RA lifetimes
accept_ra_min_rtr_lft only considered the lifetime of the default route
and discarded entire RAs accordingly.
This change renames accept_ra_min_rtr_lft to accept_ra_min_lft, and
applies the value to individual RA sections; in particular, router
lifetime, PIO preferred lifetime, and RIO lifetime. If any of those
lifetimes are lower than the configured value, the specific RA section
is ignored.
In order for the sysctl to be useful to Android, it should really apply
to all lifetimes in the RA, since that is what determines the minimum
frequency at which RAs must be processed by the kernel. Android uses
hardware offloads to drop RAs for a fraction of the minimum of all
lifetimes present in the RA (some networks have very frequent RAs (5s)
with high lifetimes (2h)). Despite this, we have encountered networks
that set the router lifetime to 30s which results in very frequent CPU
wakeups. Instead of disabling IPv6 (and dropping IPv6 ethertype in the
WiFi firmware) entirely on such networks, it seems better to ignore the
misconfigured routers while still processing RAs from other IPv6 routers
on the same network (i.e. to support IoT applications).
The previous implementation dropped the entire RA based on router
lifetime. This turned out to be hard to expand to the other lifetimes
present in the RA in a consistent manner; dropping the entire RA based
on RIO/PIO lifetimes would essentially require parsing the whole thing
twice.
Jakub Kicinski [Wed, 26 Jul 2023 18:55:29 +0000 (11:55 -0700)]
net: store netdevs in an xarray
Iterating over the netdev hash table for netlink dumps is hard.
Dumps are done in "chunks" so we need to save the position
after each chunk, so we know where to restart from. Because
netdevs are stored in a hash table we remember which bucket
we were in and how many devices we dumped.
Since we don't hold any locks across the "chunks" - devices may
come and go while we're dumping. If that happens we may miss
a device (if device is deleted from the bucket we were in).
We indicate to user space that this may have happened by setting
NLM_F_DUMP_INTR. User space is supposed to dump again (I think)
if it sees that. Somehow I doubt most user space gets this right..
To illustrate let's look at an example:
System state:
start: # [A, B, C]
del: B # [A, C]
with the hash table we may dump [A, B], missing C completely even
tho it existed both before and after the "del B".
Add an xarray and use it to allocate ifindexes. This way we
can iterate ifindexes in order, without the worry that we'll
skip one. We may still generate a dump of a state which "never
existed", for example for a set of values and sequence of ops:
System state:
start: # [A, B]
add: C # [A, C, B]
del: B # [A, C]
we may generate a dump of [A], if C got an index between A and B.
System has never been in such state. But I'm 90% sure that's perfectly
fine, important part is that we can't _miss_ devices which exist before
and after. User space which wants to mirror kernel's state subscribes
to notifications and does periodic dumps so it will know that C exists
from the notification about its creation or from the next dump
(next dump is _guaranteed_ to include C, if it doesn't get removed).
To avoid any perf regressions keep the hash table for now. Most
net namespaces have very few devices and microbenchmarking 1M lookups
on Skylake I get the following results (not counting loopback
to number of devs):
No surprises since the hash table has 256 entries.
The microbenchmark scans indexes in order, if the pattern is more
random xa starts to win at 512 devices already. But that's a lot
of devices, in practice.
Simon mentioned in another thread that it makes kdoc happy
and Jakub confirms that commit e27cb89a22ad ("scripts: kernel-doc: support
private / public marking for enums") actually added the needed
support.
David S. Miller [Fri, 28 Jul 2023 10:03:57 +0000 (11:03 +0100)]
Merge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/nex
t-queue
Tony Nguyen says:
====================
ice: Implement support for SRIOV + LAG
Dave Ertman says:
Implement support for SRIOV VF's on interfaces that are in an
aggregate interface.
The first interface added into the aggregate will be flagged as
the primary interface, and this primary interface will be
responsible for managing the VF's resources. VF's created on the
primary are the only VFs that will be supported on the aggregate.
Only Active-Backup mode will be supported and only aggregates whose
primary interface is in switchdev mode will be supported.
The ice-lag DDP must be loaded to support this feature.
Additional restrictions on what interfaces can be added to the aggregate
and still support SRIOV VFs are:
- interfaces have to all be on the same physical NIC
- all interfaces have to have the same QoS settings
- interfaces have to have the FW LLDP agent disabled
- only the primary interface is to be put into switchdev mode
- no more than two interfaces in the aggregate
---
v2:
- Move NULL check for q_ctx in ice_lag_qbuf_recfg() earlier (patch 6)
David S. Miller [Fri, 28 Jul 2023 09:59:40 +0000 (10:59 +0100)]
Merge branch 'selftest-ptp'
Alex Maftei says:
====================
selftests/ptp: Add support for new timestamp IOCTLs
PTP_SYS_OFFSET_EXTENDED was added in November 2018 in 361800876f80 (" ptp: add PTP_SYS_OFFSET_EXTENDED ioctl")
and PTP_SYS_OFFSET_PRECISE was added in February 2016 in 719f1aa4a671 ("ptp: Add PTP_SYS_OFFSET_PRECISE for driver crosstimestamping")
The PTP selftest code is lacking support for these two IOCTLS.
This short series of patches adds support for them.
Changes in v2:
- Fixed rebase issues (v1 somehow ended up with patch 1 being from the
first manual split of my changes and patch 2 being from rebase 2 out
of 3)
- Rebased on top of net-next
====================
Alex Maftei [Tue, 25 Jul 2023 21:53:33 +0000 (22:53 +0100)]
selftests/ptp: Add -x option for testing PTP_SYS_OFFSET_EXTENDED
The -x option (where 'x' stands for eXtended) takes an argument which
represents the number of samples to request from the PTP device.
The help message will display the maximum number of samples allowed.
Providing an invalid argument will also display the maximum number of
samples allowed.
this is a pull request of 21 patches for net-next/master.
The 1st patch is by Gerhard Uttenthaler, which adds Gerhard as the
maintainer ems_pci driver.
Peter Seiderer's patch removes a unused function from the peak_usb
driver.
The next 4 patches are by John Watts and add support for the sun4i_can
driver on the Allwinner D1.
Rob Herring's patch corrects the DT includes in various CAN drivers.
Followed by 14 patches from me concerning the gs_usb driver. The first
11 are various cleanups consisting of coding style improvements, error
path printout cleanups, and removal of unneeded
usb_kill_anchored_urbs(). The last 3 convert the driver to use NAPI to
avoid out-of-order reception of CAN frames.
====================
David S. Miller [Fri, 28 Jul 2023 08:54:18 +0000 (09:54 +0100)]
Merge branch 'sfc-siena-next'
Martin Habets says:
====================
sfc: Remove Siena bits from sfc.ko
Last year we split off Siena into it's own driver under directory siena.
This patch series removes the now unused Falcon and Siena code from sfc.ko.
No functional changes are intended.
====================
Martin Habets [Thu, 27 Jul 2023 10:41:21 +0000 (11:41 +0100)]
sfc: Remove struct efx_special_buffer
The attributes index and entries are no longer needed, so use
struct efx_buffer instead.
next_buffer_table was also Siena specific.
Removed some checkpatch warnings.
Martin Habets [Thu, 27 Jul 2023 10:41:05 +0000 (11:41 +0100)]
sfc: Remove PTP code for Siena
rx_tx_inline is now always true.
The special event list is no longer needed, event handling is always
inline.
Event MCDI_EVENT_CODE_PTP_RX is no longer needed.
Martin Habets [Thu, 27 Jul 2023 10:40:43 +0000 (11:40 +0100)]
sfc: Remove falcon references
Siena was using functions from the Falcon architecture.
These start with the efx_farch prefix.
Now that both of these are in separate modules the references
are no longer used in the sfc.ko module.
1. ETHTOOL_GRXFH supports custom RSS contexts, but ETHTOOL_SRXFH does
not. I moved the copy logic out of ETHTOOL_GRXFH and into a helper so
that both ETHTOOL_{G,S}RXFH now call it, which fixes ETHTOOL_SRXFH. This
is patch 1/2.
2. mlx5 defaulted to RSS context 0 for both ETHTOOL_{G,S}RXFH paths. I
have modified the driver to support custom contexts for both paths. It
is now possible to get and set the flow hash key for custom RSS contexts
with mlx5. This is patch 2/2.
See commit messages for more details.
Thanks.
v2:
- Rebased on net-next
- Adjusted arguments of mlx5e_rx_res_rss_get_hash_fields and
mlx5e_rx_res_rss_set_hash_fields to move rss_idx next to the rss
argument
- Changed return value of both mlx5e_rx_res_rss_get_hash_fields and
mlx5e_rx_res_rss_set_hash_fields to -ENOENT when the rss entry is
NULL
- Changed order of local variables in mlx5e_get_rss_hash_opt and
mlx5e_set_rss_hash_opt
====================
Joe Damato [Tue, 25 Jul 2023 20:56:55 +0000 (20:56 +0000)]
net/mlx5: Fix flowhash key set/get for custom RSS
mlx5 flow hash field retrieval and set only worked on the default
RSS context as of commit f01cc58c18d6 ("net/mlx5e: Support multiple RSS
contexts"), not custom RSS contexts.
For example, before this patch attempting to retrieve the flow hash fields
for RSS context 1 fails:
$ sudo ethtool -u eth1 rx-flow-hash tcp4 context 1
For RSS context 1:
TCP over IPV4 flows use these fields for computing Hash flow key:
IP SA
IP DA
L4 bytes 0 & 1 [TCP/UDP src port]
L4 bytes 2 & 3 [TCP/UDP dst port]
Joe Damato [Tue, 25 Jul 2023 20:56:54 +0000 (20:56 +0000)]
net: ethtool: Unify ETHTOOL_{G,S}RXFH rxnfc copy
ETHTOOL_GRXFH correctly copies in the full struct ethtool_rxnfc when
FLOW_RSS is set; ETHTOOL_SRXFH needs a similar code path to handle the
FLOW_RSS case so that ethtool can set the flow hash for custom RSS
contexts (if supported by the driver).
The copy code from ETHTOOL_GRXFH has been pulled out in to a helper so
that it can be called in both ETHTOOL_{G,S}RXFH code paths.
projects / linux.git / log