net: phy: add helpers to handle sfp phy connect/disconnect
There are a few PHY drivers that can handle SFP modules through their
sfp_upstream_ops. Introduce Phylib helpers to keep track of connected
SFP PHYs in a netdevice's namespace, by adding the SFP PHY to the
upstream PHY's netdev's namespace.
By doing so, these SFP PHYs can be enumerated and exposed to users,
which will be able to use their capabilities.
net: sfp: pass the phy_device when disconnecting an sfp module's PHY
Pass the phy_device as a parameter to the sfp upstream .disconnect_phy
operation. This is preparatory work to help track phy devices across
a net_device's link.
net: phy: Introduce ethernet link topology representation
Link topologies containing multiple network PHYs attached to the same
net_device can be found when using a PHY as a media converter for use
with an SFP connector, on which an SFP transceiver containing a PHY can
be used.
With the current model, the transceiver's PHY can't be used for
operations such as cable testing, timestamping, macsec offload, etc.
The reason being that most of the logic for these configuration, coming
from either ethtool netlink or ioctls tend to use netdev->phydev, which
in multi-phy systems will reference the PHY closest to the MAC.
Introduce a numbering scheme allowing to enumerate PHY devices that
belong to any netdev, which can in turn allow userspace to take more
precise decisions with regard to each PHY's configuration.
The numbering is maintained per-netdev, in a phy_device_list.
The numbering works similarly to a netdevice's ifindex, with
identifiers that are only recycled once INT_MAX has been reached.
This prevents races that could occur between PHY listing and SFP
transceiver removal/insertion.
The identifiers are assigned at phy_attach time, as the numbering
depends on the netdevice the phy is attached to. The PHY index can be
re-used for PHYs that are persistent.
net: phy: marvell: implement cable test for 88E1111
The same implementation is also valid for 88E1145. VCT in 88E1111 is
similar to the 88E609x family. The main difference lies in register
organization and required workarounds.
It utilizes the same fields in registers but requires a simpler
implementation.
Jakub Kicinski [Wed, 3 Apr 2024 20:22:59 +0000 (13:22 -0700)]
netlink: add nlmsg_consume() and use it in devlink compat
devlink_compat_running_version() sticks out when running
netdevsim tests and watching dropped skbs. Add nlmsg_consume()
for cases were we want to free a netlink skb but it is expected,
rather than a drop. af_netlink code uses consume_skb() directly,
which is fine, but some may prefer the symmetry of nlmsg_new() /
nlmsg_consume().
Jakub Kicinski [Wed, 3 Apr 2024 20:21:39 +0000 (13:21 -0700)]
net: skbuff: generalize the skb->decrypted bit
The ->decrypted bit can be reused for other crypto protocols.
Remove the direct dependency on TLS, add helpers to clean up
the ifdefs leaking out everywhere.
Hangbin Liu [Thu, 4 Apr 2024 06:31:12 +0000 (14:31 +0800)]
ynl: rename array-nest to indexed-array
Some implementations, like bonding, has nest array with same attr type.
To support all kinds of entries under one nest array. As discussed[1],
let's rename array-nest to indexed-array, and assuming the value is
a nest by passing the type via sub-type.
The goal of this patch series is to introduce a common set of ethtool
statistics for hardware timestamping that a driver implementer can hook into.
The statistics counters added are based on what I believe are common
patterns/behaviors found across various hardware timestamping implementations
seen in the kernel tree today. The mlx5 family of devices is used
as the PoC for this patch series. Other vendors are more than welcome
to chime in on this series.
net/mlx5e: Introduce lost_cqe statistic counter for PTP Tx port timestamping CQ
Track the number of times a CQE was expected to not be delivered on PTP Tx
port timestamping CQ. A CQE is expected to not be delivered if a certain
amount of time passes since the corresponding CQE containing the DMA
timestamp information has arrived. Increment the late_cqe counter when such
a CQE does manage to be delivered to the CQ.
ethtool: add interface to read Tx hardware timestamping statistics
Multiple network devices that support hardware timestamping appear to have
common behavior with regards to timestamp handling. Implement common Tx
hardware timestamping statistics in a tx_stats struct_group. Common Rx
hardware timestamping statistics can subsequently be implemented in a
rx_stats struct_group for ethtool_ts_stats.
Jakub Kicinski [Sat, 6 Apr 2024 05:11:28 +0000 (22:11 -0700)]
Merge branch 'address-all-wunused-const-warnings'
Arnd Bergmann says:
====================
address all -Wunused-const warnings (the net part)
Compilers traditionally warn for unused 'static' variables, but not
if they are constant. The reason here is a custom for C++ programmers
to define named constants as 'static const' variables in header files
instead of using macros or enums.
In W=1 builds, we get warnings only static const variables in C
files, but not in headers, which is a good compromise, but this still
produces warning output in at least 30 files. These warnings are
almost all harmless, but also trivial to fix, and there is no
good reason to warn only about the non-const variables being unused.
I've gone through all the files that I found using randconfig and
allmodconfig builds and created patches to avoid these warnings,
with the goal of retaining a clean build once the option is enabled
by default.
Unfortunately, there is one fairly large patch ("drivers: remove
incorrect of_match_ptr/ACPI_PTR annotations") that touches
34 individual drivers that all need the same one-line change.
If necessary, I can split it up by driver or by subsystem,
but at least for reviewing I would keep it as one piece for
the moment.
====================
strncpy() is deprecated for use on NUL-terminated destination strings
[1] and as such we should prefer more robust and less ambiguous string
interfaces.
For 2 out of 3 of these changes we can simply swap in strscpy() as it
guarantess NUL-termination which is needed for the following trace
print.
trace_rpcgss_context() should use memcpy as its format specifier %.*s
allows for the length to be specifier (__entry->len). Due to this,
acceptor does not technically need to be NUL-terminated. Moreover,
swapping in strscpy() and keeping everything else the same could result
in truncation of the source string by one byte. To remedy this, we could
use `len + 1` but I am unsure of the size of the destination buffer so a
simple memcpy should suffice.
| TP_printk("win_size=%u expiry=%lu now=%lu timeout=%u acceptor=%.*s",
| __entry->window_size, __entry->expiry, __entry->now,
| __entry->timeout, __entry->len, __get_str(acceptor))
I suspect acceptor not to naturally be a NUL-terminated string due to
the presence of some stringify methods.
| .crstringify_acceptor = gss_stringify_acceptor,
net/mlx5e: Support FEC settings for 100G/lane modes
This consists of:
1. Expose the 100G/lane capability bit in the PCAM reg.
2. Expose the per link mode FEC capability masks in the PPLM reg.
3. Set the overrides according to ethtool parameters.
FEC for new modes is set if and only if the PCAM 100G/lane capability is
advertised and the capability mask for a given link mode reports that it
can accept the requested FEC mode.
net/mlx5e: Extract checking of FEC support for a link mode
The check of whether a given FEC mode is supported in a given link mode
is about to get more complicated, so extract it in a separate function
to avoid code duplication.
Refactor FDB dump code path for Microchip KSZ8xxx series. This series
mostly makes some cosmetic reworks and allows to forward errors detected
by the regmap.
Change logs are part of patch commit messages.
====================
Refactor ksz8_fdb_dump() to address potential issues:
- Limit the number of iterations to avoid endless loops.
- Handle error codes returned by ksz8_r_dyn_mac_table(), with
an exception for -ENXIO when no more dynamic entries are detected.
net/ipv4/ip_gre.c 17af420545a7 ("erspan: make sure erspan_base_hdr is present in skb->head") 5832c4a77d69 ("ip_tunnel: convert __be16 tunnel flags to bitmaps")
https://lore.kernel.org/all/20240402103253.3b54a1cf@canb.auug.org.au/
Adjacent changes:
net/ipv6/ip6_fib.c d21d40605bca ("ipv6: Fix infinite recursion in fib6_dump_done().") 5fc68320c1fb ("ipv6: remove RTNL protection from inet6_dump_fib()")
Merge tag 'net-6.9-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Pull networking fixes from Jakub Kicinski:
"Including fixes from netfilter, bluetooth and bpf.
Fairly usual collection of driver and core fixes. The large selftest
accompanying one of the fixes is also becoming a common occurrence.
Current release - regressions:
- ipv6: fix infinite recursion in fib6_dump_done()
- net/rds: fix possible null-deref in newly added error path
Current release - new code bugs:
- net: do not consume a full cacheline for system_page_pool
- bpf: fix bpf_arena-related file descriptor leaks in the verifier
- drv: ice: fix freeing uninitialized pointers, fixing misuse of the
newfangled __free() auto-cleanup
Previous releases - regressions:
- x86/bpf: fixes the BPF JIT with retbleed=stuff
- xen-netfront: add missing skb_mark_for_recycle, fix page pool
accounting leaks, revealed by recently added explicit warning
- tcp: fix bind() regression for v6-only wildcard and v4-mapped-v6
non-wildcard addresses
- Bluetooth:
- replace "hci_qca: Set BDA quirk bit if fwnode exists in DT" with
better workarounds to un-break some buggy Qualcomm devices
- set conn encrypted before conn establishes, fix re-connecting to
some headsets which use slightly unusual sequence of msgs
- mptcp:
- prevent BPF accessing lowat from a subflow socket
- don't account accept() of non-MPC client as fallback to TCP
- drv: mana: fix Rx DMA datasize and skb_over_panic
- drv: i40e: fix VF MAC filter removal
Previous releases - always broken:
- gro: various fixes related to UDP tunnels - netns crossing
problems, incorrect checksum conversions, and incorrect packet
transformations which may lead to panics
- bpf: support deferring bpf_link dealloc to after RCU grace period
- nf_tables:
- release batch on table validation from abort path
- release mutex after nft_gc_seq_end from abort path
- flush pending destroy work before exit_net release
- drv: r8169: skip DASH fw status checks when DASH is disabled"
* tag 'net-6.9-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (81 commits)
netfilter: validate user input for expected length
net/sched: act_skbmod: prevent kernel-infoleak
net: usb: ax88179_178a: avoid the interface always configured as random address
net: dsa: sja1105: Fix parameters order in sja1110_pcs_mdio_write_c45()
net: ravb: Always update error counters
net: ravb: Always process TX descriptor ring
netfilter: nf_tables: discard table flag update with pending basechain deletion
netfilter: nf_tables: Fix potential data-race in __nft_flowtable_type_get()
netfilter: nf_tables: reject new basechain after table flag update
netfilter: nf_tables: flush pending destroy work before exit_net release
netfilter: nf_tables: release mutex after nft_gc_seq_end from abort path
netfilter: nf_tables: release batch on table validation from abort path
Revert "tg3: Remove residual error handling in tg3_suspend"
tg3: Remove residual error handling in tg3_suspend
net: mana: Fix Rx DMA datasize and skb_over_panic
net/sched: fix lockdep splat in qdisc_tree_reduce_backlog()
net: phy: micrel: lan8814: Fix when enabling/disabling 1-step timestamping
net: stmmac: fix rx queue priority assignment
net: txgbe: fix i2c dev name cannot match clkdev
net: fec: Set mac_managed_pm during probe
...
Merge tag 'bcachefs-2024-04-03' of https://evilpiepirate.org/git/bcachefs
Pull bcachefs repair code from Kent Overstreet:
"A couple more small fixes, and new repair code.
We can now automatically recover from arbitrary corrupted interior
btree nodes by scanning, and we can reconstruct metadata as needed to
bring a filesystem back into a working, consistent, read-write state
and preserve access to whatevver wasn't corrupted.
Meaning - you can blow away all metadata except for extents and
dirents leaf nodes, and repair will reconstruct everything else and
give you your data, and under the correct paths. If inodes are missing
i_size will be slightly off and permissions/ownership/timestamps will
be gone, and we do still need the snapshots btree if snapshots were in
use - in the future we'll be able to guess the snapshot tree structure
in some situations.
IOW - aside from shaking out remaining bugs (fuzz testing is still
coming), repair code should be complete and if repair ever doesn't
work that's the highest priority bug that I want to know about
immediately.
This patchset was kindly tested by a user from India who accidentally
wiped one drive out of a three drive filesystem with no replication on
the family computer - it took a couple weeks but we got everything
important back"
* tag 'bcachefs-2024-04-03' of https://evilpiepirate.org/git/bcachefs:
bcachefs: reconstruct_inode()
bcachefs: Subvolume reconstruction
bcachefs: Check for extents that point to same space
bcachefs: Reconstruct missing snapshot nodes
bcachefs: Flag btrees with missing data
bcachefs: Topology repair now uses nodes found by scanning to fill holes
bcachefs: Repair pass for scanning for btree nodes
bcachefs: Don't skip fake btree roots in fsck
bcachefs: bch2_btree_root_alloc() -> bch2_btree_root_alloc_fake()
bcachefs: Etyzinger cleanups
bcachefs: bch2_shoot_down_journal_keys()
bcachefs: Clear recovery_passes_required as they complete without errors
bcachefs: ratelimit informational fsck errors
bcachefs: Check for bad needs_discard before doing discard
bcachefs: Improve bch2_btree_update_to_text()
mean_and_variance: Drop always failing tests
bcachefs: fix nocow lock deadlock
bcachefs: BCH_WATERMARK_interior_updates
bcachefs: Fix btree node reserve
Jakub Kicinski [Thu, 4 Apr 2024 18:37:39 +0000 (11:37 -0700)]
Merge tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf
Daniel Borkmann says:
====================
pull-request: bpf 2024-04-04
We've added 7 non-merge commits during the last 5 day(s) which contain
a total of 9 files changed, 75 insertions(+), 24 deletions(-).
The main changes are:
1) Fix x86 BPF JIT under retbleed=stuff which causes kernel panics due to
incorrect destination IP calculation and incorrect IP for relocations,
from Uros Bizjak and Joan Bruguera Micó.
2) Fix BPF arena file descriptor leaks in the verifier,
from Anton Protopopov.
3) Defer bpf_link deallocation to after RCU grace period as currently
running multi-{kprobes,uprobes} programs might still access cookie
information from the link, from Andrii Nakryiko.
4) Fix a BPF sockmap lock inversion deadlock in map_delete_elem reported
by syzkaller, from Jakub Sitnicki.
5) Fix resolve_btfids build with musl libc due to missing linux/types.h
include, from Natanael Copa.
* tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf:
bpf, sockmap: Prevent lock inversion deadlock in map delete elem
x86/bpf: Fix IP for relocating call depth accounting
x86/bpf: Fix IP after emitting call depth accounting
bpf: fix possible file descriptor leaks in verifier
tools/resolve_btfids: fix build with musl libc
bpf: support deferring bpf_link dealloc to after RCU grace period
bpf: put uprobe link's path and task in release callback
====================
Jakub Kicinski [Wed, 3 Apr 2024 02:34:21 +0000 (19:34 -0700)]
tools: ynl: copy netlink error to NlError
Typing e.nl_msg.error when processing exception is a bit tedious
and counter-intuitive. Set a local .error member to the positive
value of the netlink level error.
The buggy address belongs to the object at ffff88802cd73da0
which belongs to the cache kmalloc-8 of size 8
The buggy address is located 0 bytes inside of
allocated 1-byte region [ffff88802cd73da0, ffff88802cd73da1)
Memory state around the buggy address: ffff88802cd73c80: 07 fc fc fc 05 fc fc fc 05 fc fc fc fa fc fc fc ffff88802cd73d00: fa fc fc fc fa fc fc fc fa fc fc fc fa fc fc fc
>ffff88802cd73d80: fa fc fc fc 01 fc fc fc fa fc fc fc fa fc fc fc
^ ffff88802cd73e00: fa fc fc fc fa fc fc fc 05 fc fc fc 07 fc fc fc ffff88802cd73e80: 07 fc fc fc 07 fc fc fc 07 fc fc fc 07 fc fc fc
Jakub Kicinski [Thu, 4 Apr 2024 16:38:52 +0000 (09:38 -0700)]
Merge tag 'nf-24-04-04' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf
Pablo Neira Ayuso says:
====================
Netfilter fixes for net
The following patchset contains Netfilter fixes for net:
Patch #1 unlike early commit path stage which triggers a call to abort,
an explicit release of the batch is required on abort, otherwise
mutex is released and commit_list remains in place.
Patch #2 release mutex after nft_gc_seq_end() in commit path, otherwise
async GC worker could collect expired objects.
Patch #3 flush pending destroy work in module removal path, otherwise UaF
is possible.
Patch #4 and #6 restrict the table dormant flag with basechain updates
to fix state inconsistency in the hook registration.
Patch #5 adds missing RCU read side lock to flowtable type to avoid races
with module removal.
* tag 'nf-24-04-04' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf:
netfilter: nf_tables: discard table flag update with pending basechain deletion
netfilter: nf_tables: Fix potential data-race in __nft_flowtable_type_get()
netfilter: nf_tables: reject new basechain after table flag update
netfilter: nf_tables: flush pending destroy work before exit_net release
netfilter: nf_tables: release mutex after nft_gc_seq_end from abort path
netfilter: nf_tables: release batch on table validation from abort path
====================
net: usb: ax88179_178a: avoid the interface always configured as random address
After the commit d2689b6a86b9 ("net: usb: ax88179_178a: avoid two
consecutive device resets"), reset is not executed from bind operation and
mac address is not read from the device registers or the devicetree at that
moment. Since the check to configure if the assigned mac address is random
or not for the interface, happens after the bind operation from
usbnet_probe, the interface keeps configured as random address, although the
address is correctly read and set during open operation (the only reset
now).
In order to keep only one reset for the device and to avoid the interface
always configured as random address, after reset, configure correctly the
suitable field from the driver, if the mac address is read successfully from
the device registers or the devicetree. Take into account if a locally
administered address (random) was previously stored.
bnxt_en: Add warning message about disallowed speed change
Some chips may not allow changing default speed when dual rate
transceivers modules are used. Firmware on those chips will
indicate the same to the driver.
Add a warning message when speed change is not supported
because a dual rate transceiver is detected by the NIC.
This updated interface supports backing store APIs to
configure host FW trace buffers, updates transceivers ID
types, updates to TrueFlow features and other changes
for the basic L2 features.
- Change the last arg to xdp_prepare_buff to true from false.
- Ensure that when XDP_PASS is returned the xdp->data_meta area
is copied to the skb->data area. Account for the meta data
size on skb allocation and do a pull after to move it to the
"reserved" zone.
Change bnxt_rx_xdp() to take a pointer to xdp instead of stack
variable.
This is in prepartion for the XDP metadata patch change where
the BPF program can change the value of the xdp.meta_data.
Driver's Page Pool allocation code looks at the node local
to the PCIe device to determine where to allocate memory.
In scenarios where the core count per NUMA node is low (< default rings)
it makes sense to exhaust page pool allocations on
Node 0 first and then moving on to allocating page pools
for the remaining rings from Node 1.
With this patch, and the following configuration on the NIC
$ ethtool -L ens1f0np0 combined 16
(core count/node = 12, first 12 rings on node#0, last 4 rings node#1)
and traffic redirected to a ring on node#1 , we see a performance
improvement of ~20%
Enable XPS on default during NIC open. The choice of
Tx queue is based on the CPU executing the thread that
submits the Tx request. The pool of Tx queues will be
spread evenly across both device-attached NUMA nodes(local)
and remote NUMA nodes.
bnxt_en: Add delay to handle Downstream Port Containment (DPC) AER
In case of DPC, after issuing the hot reset, the
kernel waits for 100ms for the device to complete
the reset. However on some older chips, the firmware
may take up to 1 second to complete the reset, only
after which the driver can restart the card.
Introduce delay of 900ms to handle this scenario on
the older chipsets.
linke li [Wed, 3 Apr 2024 02:54:00 +0000 (10:54 +0800)]
net: ethernet: mtk_eth_soc: Reuse value using READ_ONCE instead of re-rereading it
In mtk_flow_entry_update_l2, the hwe->ib1 is read using READ_ONCE at the
beginning of the function, checked, and then re-read from hwe->ib1,
may void all guarantees of the checks. Reuse the value that was read by
READ_ONCE to ensure the consistency of the ib1 throughout the function.
Paul Barker [Tue, 2 Apr 2024 14:53:05 +0000 (15:53 +0100)]
net: ravb: Always update error counters
The error statistics should be updated each time the poll function is
called, even if the full RX work budget has been consumed. This prevents
the counts from becoming stuck when RX bandwidth usage is high.
This also ensures that error counters are not updated after we've
re-enabled interrupts as that could result in a race condition.
Paul Barker [Tue, 2 Apr 2024 14:53:04 +0000 (15:53 +0100)]
net: ravb: Always process TX descriptor ring
The TX queue should be serviced each time the poll function is called,
even if the full RX work budget has been consumed. This prevents
starvation of the TX queue when RX bandwidth usage is high.
netfilter: nf_tables: discard table flag update with pending basechain deletion
Hook unregistration is deferred to the commit phase, same occurs with
hook updates triggered by the table dormant flag. When both commands are
combined, this results in deleting a basechain while leaving its hook
still registered in the core.
Fixes: 179d9ba5559a ("netfilter: nf_tables: fix table flag updates") Signed-off-by: Pablo Neira Ayuso <[email protected]>
Ziyang Xuan [Wed, 3 Apr 2024 07:22:04 +0000 (15:22 +0800)]
netfilter: nf_tables: Fix potential data-race in __nft_flowtable_type_get()
nft_unregister_flowtable_type() within nf_flow_inet_module_exit() can
concurrent with __nft_flowtable_type_get() within nf_tables_newflowtable().
And thhere is not any protection when iterate over nf_tables_flowtables
list in __nft_flowtable_type_get(). Therefore, there is pertential
data-race of nf_tables_flowtables list entry.
Use list_for_each_entry_rcu() to iterate over nf_tables_flowtables list
in __nft_flowtable_type_get(), and use rcu_read_lock() in the caller
nft_flowtable_type_get() to protect the entire type query process.
Fixes: 3b49e2e94e6e ("netfilter: nf_tables: add flow table netlink frontend") Signed-off-by: Ziyang Xuan <[email protected]> Signed-off-by: Pablo Neira Ayuso <[email protected]>
netfilter: nf_tables: flush pending destroy work before exit_net release
Similar to 2c9f0293280e ("netfilter: nf_tables: flush pending destroy
work before netlink notifier") to address a race between exit_net and
the destroy workqueue.
The trace below shows an element to be released via destroy workqueue
while exit_net path (triggered via module removal) has already released
the set that is used in such transaction.
netfilter: nf_tables: release mutex after nft_gc_seq_end from abort path
The commit mutex should not be released during the critical section
between nft_gc_seq_begin() and nft_gc_seq_end(), otherwise, async GC
worker could collect expired objects and get the released commit lock
within the same GC sequence.
nf_tables_module_autoload() temporarily releases the mutex to load
module dependencies, then it goes back to replay the transaction again.
Move it at the end of the abort phase after nft_gc_seq_end() is called.
netfilter: nf_tables: release batch on table validation from abort path
Unlike early commit path stage which triggers a call to abort, an
explicit release of the batch is required on abort, otherwise mutex is
released and commit_list remains in place.
Add WARN_ON_ONCE to ensure commit_list is empty from the abort path
before releasing the mutex.
After this patch, commit_list is always assumed to be empty before
grabbing the mutex, therefore
03c1f1ef1584 ("netfilter: Cleanup nft_net->module_list from nf_tables_exit_net()")
only needs to release the pending modules for registration.
Cc: [email protected] Fixes: c0391b6ab810 ("netfilter: nf_tables: missing validation from the abort path") Signed-off-by: Pablo Neira Ayuso <[email protected]>
====================
mlxsw: Preparations for improving performance
Amit Cohen writes:
mlxsw driver will use NAPI for event processing in a next patch set.
Some additional improvements will be added later. This patch set
prepares the code for NAPI usage and refactor some relevant areas. See
more details in commit messages.
Patch Set overview:
Patches #1-#2 are preparations for patch #3
Patch #3 setups tasklets as part of queue initializtion
Patch #4 removes handling of unlikely scenario
Patch #5 removes unused counters
Patch #6 makes style change in mlxsw_pci_eq_tasklet()
Patch #7-#10 poll command interface instead of EQ0 usage
Patches #11-#12 make style change and break the function
mlxsw_pci_cq_tasklet()
Patches #13-#14 remove functions which can be replaced by a stored value
Patch #15 improves accessing to descriptor queue instance
====================
Amit Cohen [Tue, 2 Apr 2024 13:54:28 +0000 (15:54 +0200)]
mlxsw: pci: Store DQ pointer as part of CQ structure
Currently, for each completion, we check the number of descriptor queue
and take it via mlxsw_pci_{sdq,rdq}_get(). This is inefficient, the
DQ should be the same for all the completions in CQ, as each CQ handles
only one DQ - SDQ or RDQ. This mapping is handled as part of DQ
initialization via mlxsw_cmd_mbox_sw2hw_dq_cq_set().
Instead, as part of DQ initialization, set DQ pointer in the appropriate
CQ structure. When we handle completions, warn in case that the DQ number
that we expect is different from the number we get in the CQE. Call
WARN_ON_ONCE() only after checking the value, to avoid calling this method
for each completion.
Amit Cohen [Tue, 2 Apr 2024 13:54:27 +0000 (15:54 +0200)]
mlxsw: pci: Remove mlxsw_pci_cq_count()
Currently, for each interrupt we call mlxsw_pci_cq_count() to determine the
number of CQs. This call makes additional two function's calls. This can
be removed by storing this value as part of structure 'mlxsw_pci', as we
already do for number of SDQs. Remove the function and
__mlxsw_pci_queue_count() which is now not used and store the value
instead.
Amit Cohen [Tue, 2 Apr 2024 13:54:26 +0000 (15:54 +0200)]
mlxsw: pci: Remove mlxsw_pci_sdq_count()
The number of SDQs is stored as part of 'mlxsw_pci' structure. In some
cases, the driver uses this value and in some cases it calls
mlxsw_pci_sdq_count() to get the value. Align the code to use the
stored value. This simplifies the code and makes it clearer that the
value is always the same. Rename 'mlxsw_pci->num_sdq_cqs' to
'mlxsw_pci->num_sdqs' as now it is used not only in CQ context.
Amit Cohen [Tue, 2 Apr 2024 13:54:25 +0000 (15:54 +0200)]
mlxsw: pci: Break mlxsw_pci_cq_tasklet() into tasklets per queue type
Completion queues are used for completions of RDQ or SDQ. Each
completion queue is used for one DQ. The first CQs are used for SDQs and
the rest are used for RDQs.
Currently, for each CQE (completion queue element), we check 'sr' value
(send/receive) to know if it is completion of RDQ or SDQ. Actually, we
do not really have to check it, as according to the queue number we know
if it handles completions of Rx or Tx.
Break the tasklet into two - one for Rx (RDQ) and one for Tx (SDQ). Then,
setup the appropriate tasklet for each queue as part of queue
initialization. Use 'sr' value for unlikely case that we get completion
with type that we do not expect. Call WARN_ON_ONCE() only after checking
the value, to avoid calling this method for each completion.
A next patch set will use NAPI to handle events, then we will have a
separate poll method for Rx and Tx. This change is a preparation for
NAPI usage.
Amit Cohen [Tue, 2 Apr 2024 13:54:23 +0000 (15:54 +0200)]
mlxsw: pci: Remove unused wait queue
The previous patch changed the code to do not handle command interface
from event queue. With this change the wait queue is not used anymore.
Remove it and 'wait_done' variable.
Amit Cohen [Tue, 2 Apr 2024 13:54:22 +0000 (15:54 +0200)]
mlxsw: pci: Use only one event queue
The device supports two event queues. EQ0 is used for command interface
completion events. EQ1 is used for completion events of RDQ or SDQ.
Currently, for each EQE (event queue element), we check the queue number
and handle accordingly. More than that, for each interrupt we schedule
tasklets for both EQs. This is really ineffective, especially because of
the fact that EQ0 is used only as part of driver init/fini, when EMADs are
not available. There is no point to schedule the tasklet for it and check
each EQE.
A previous patch changed the code to poll command interface for each use of
it. It means that now there is no real reason to use EQ0, as we poll the
command interface.
Initialize only one event queue and use it as EQ1 (this is determined by
queue number). Then, for each interrupt we can schedule the tasklet only
for one queue and we do not have to check the queue number. This
simplifies the code and should improve performance. Note that polling
command interface is ok as we use it only as part of driver init/fini.
Amit Cohen [Tue, 2 Apr 2024 13:54:21 +0000 (15:54 +0200)]
mlxsw: pci: Rename MLXSW_PCI_EQS_COUNT
Currently we use MLXSW_PCI_EQS_COUNT event queues. A next patch will
change the driver to initialize only EQ1, as EQ0 is not required anymore
when we poll command interface.
Rename the macro to MLXSW_PCI_EQS_MAX as later we will not initialize
the maximum supported EQs, this value represents the maximum and a new
macro will be added to represent the actual used queues.
Amit Cohen [Tue, 2 Apr 2024 13:54:20 +0000 (15:54 +0200)]
mlxsw: pci: Poll command interface for each cmd_exec()
Command interface is used for configuring and querying FW when EMADs are
not available. During the time that the driver sets up the asynchronous
queues, it polls the command interface for getting completions. Then,
there is a short period when asynchronous queues work, but EMADs are not
available (marked in the code as nopoll = true). During this time, we
send commands via command interface, but we do not poll it, as we can get
an interrupt for the completion. Completions of command interface are
received from HW in EQ0 (event queue 0).
The usage of EQ0 instead of polling is done only 4 times during
initialization and one time during tear down, but it makes an overhead
during lifetime of the driver. For each interrupt, we have to check if
we get events in EQ0 or EQ1 and handle them. This is really ineffective,
especially because of the fact that EQ0 is used only as part of driver
init/fini.
Instead, we can poll command interface for each call of cmd_exec(). It
means that when we send a command via command interface (as EMADs are
not available), we will poll it, regardless of availability of the
asynchronous queues. This will allow us to configure later only EQ1 and
simplify the flow.
Remove 'nopoll' indication and change mlxsw_pci_cmd_exec() to poll till
answer/timeout regardless of queues' state. For now, completions are
handled also by EQ0, but it will be removed in next patch. Additional
cleanups will be added in next patches.
Amit Cohen [Tue, 2 Apr 2024 13:54:19 +0000 (15:54 +0200)]
mlxsw: pci: Make style changes in mlxsw_pci_eq_tasklet()
This function will be used later only for EQ1. As preparation, reorder
variables to reverse xmas tree and return earlier when it is possible, to
simplify the code.
Amit Cohen [Tue, 2 Apr 2024 13:54:18 +0000 (15:54 +0200)]
mlxsw: pci: Remove unused counters
The structure 'mlxsw_pci_queue' stores several counters which were consumed
via debugfs. Since commit 9a32562becd9 ("mlxsw: Remove debugfs interface"),
these counters are not used. Remove them. This makes the 'union u' and
'struct eq' redundant. Maintain 'struct cq' as it will be extended later.
Replace increasing 'q->u.eq.ev_other_count' with WARN_ON_ONCE(), as it is
used in an unreasonable case of receiving event in EQ which is not EQ0 or
EQ1. When the queues are initialized, we check number of event queues and
fail with the print "Unsupported number of queues" in case that the driver
tries to initialize more than two queues.
Amit Cohen [Tue, 2 Apr 2024 13:54:17 +0000 (15:54 +0200)]
mlxsw: pci: Arm CQ doorbell regardless of number of completions
Currently, as part of mlxsw_pci_cq_tasklet(), we check if any item
was handled, and only in such case we arm doorbell. This is unlikely case,
as we schedule tasklet only for CQs that we get an event for them, which
means that they contain completions to handle. Remove this check, which
is supposed to be true always, and even if it is false, it is not a mistake
to ring the doorbell. We can warn on such case, but it is not really worth
to add a check which will be run for each CQ handling when we do not expect
to reach it and it does not point to logic error that should be handled.
Amit Cohen [Tue, 2 Apr 2024 13:54:16 +0000 (15:54 +0200)]
mlxsw: pci: Do not setup tasklet from operation
Currently, the structure 'mlxsw_pci_queue_ops' holds a pointer to the
callback function of tasklet. This is used only for EQ and CQ. mlxsw
driver will use NAPI in a following patch set, so CQ will not use tasklet
anymore. As preparation, remove this pointer from the shared operation
structure and setup the tasklet as part of queue initialization.
For now, setup tasklet for EQ and CQ. Later, CQ code will be changed.
Jakub Kicinski [Thu, 4 Apr 2024 02:48:01 +0000 (19:48 -0700)]
Merge branch 'mlx5-misc-patches'
Tariq Toukan says:
====================
mlx5 misc patches
This patchset includes small features and misc code enhancements for the
mlx5 core and EN drivers.
Patches 1-4 by Gal improves the mlx5e ethtool stats implementation, for
example by using standard helpers ethtool_sprintf/puts.
Patch 5 by me adds a reset option for the FW command interface debugfs
stats entries. This allows explicit FW command interface stats reset
between different runs of a test case.
Patches 6 and 7 are simple cleanups.
Patch 8 by Gal adds driver support for 800Gbps link modes.
Patch 9 by Jianbo enhances the L4 steering abilities.
Patches 10-11 by Jianbo save redundant operations.
====================
Jianbo Liu [Tue, 2 Apr 2024 13:30:43 +0000 (16:30 +0300)]
net/mlx5: Don't call give_pages() if request 0 page
Firmware will return 0 on query BOOT/INIT PAGES for non-page supplier
functions (external host PF/VF/SF), so no page is needed to be
allocated for them.
Jianbo Liu [Tue, 2 Apr 2024 13:30:41 +0000 (16:30 +0300)]
net/mlx5: Support matching on l4_type for ttc_table
Replace matching on TCP and UDP protocols with new l4_type field which
is parsed by steering for ttc_table. It is enabled by the
outer_l4_type or inner_l4_type bits in nic_rx or port_sel flow table
capabilities and used only if pcc_ifa2 bit in HCA capabilities is set.
Starting from commit eb9b9fdcafe2 ("net/mlx5e: Introduce extended version for mlx5e_xmit_data")
sinfo is no longer passed as an argument to
mlx5e_xmit_xdp_frame(), the comment is inconsistent.
check_result must be zero when the packet is fragmented.
net/mlx5e: debugfs, Add reset option for command interface stats
Resetting stats just before some test/debug case allows us to eliminate
out the impact of previous commands. Useful in particular for the
average latency calculation.
The average_write() callback was unreachable, as "average" is a
read-only file. Extend, rename, and use it for a newly exposed
write-only "reset" file.
Gal Pressman [Tue, 2 Apr 2024 13:30:36 +0000 (16:30 +0300)]
net/mlx5e: Make stats group fill_stats callbacks consistent with the API
The fill_strings() callbacks were changed to accept a **data pointer,
and not rely on propagating the index value.
Make a similar change to fill_stats() callbacks to keep the API
consistent.
Gal Pressman [Tue, 2 Apr 2024 13:30:35 +0000 (16:30 +0300)]
net/mlx5e: Use ethtool_sprintf/puts() to fill stats strings
Use ethtool_sprintf/puts() helper functions which handle the common
pattern of printing a string into the ethtool strings interface and
incrementing the string pointer by ETH_GSTRING_LEN.
Change the fill_strings callback to accept a **data pointer, and remove
the index and return value.
Gal Pressman [Tue, 2 Apr 2024 13:30:34 +0000 (16:30 +0300)]
net/mlx5e: Use ethtool_sprintf/puts() to fill selftests strings
Use ethtool_sprintf/puts() helper functions which handle the common
pattern of printing a string into the ethtool strings interface and
incrementing the string pointer by ETH_GSTRING_LEN.
The int return value in mlx5e_self_test_fill_strings() is not removed as
it is still used to return the number of selftests.
Gal Pressman [Tue, 2 Apr 2024 13:30:33 +0000 (16:30 +0300)]
net/mlx5e: Use ethtool_sprintf/puts() to fill priv flags strings
Use ethtool_sprintf/puts() helper functions which handle the common
pattern of printing a string into the ethtool strings interface and
incrementing the string pointer by ETH_GSTRING_LEN.
projects / linux.git / log