In preparation for adaptive IRQ coalescing, we first need to support
adjusting the settings at runtime. The existing code doesn't require any
locking because
- dma_start is the only function that modifies rx/tx_dma_cr. It is
always called with IRQs and NAPI disabled, so nothing else is touching
the hardware.
- The IRQs don't race with poll, since the latter is a softirq.
- The IRQs don't race with dma_stop since they both just clear the
control registers.
- dma_stop doesn't race with poll since the former is called with NAPI
disabled.
However, once we introduce another function that modifies rx/tx_dma_cr,
we need to have some locking to prevent races. Introduce two locks to
protect these variables and their registers.
The control register values are now generated where the coalescing
settings are set. Converting coalescing settings to control register
values may require sleeping because of clk_get_rate. However, the
read/modify/write of the control registers themselves can't sleep
because it needs to happen in IRQ context. By pre-calculating the
control register values, we avoid introducing an additional mutex.
Since axienet_dma_start writes the control settings when it runs, we
don't bother updating the CR registers when rx/tx_dma_started is false.
This prevents any issues from writing to the control registers in the
middle of a reset sequence.
Signed-off-by: Sean Anderson <sean.anderson@linux.dev>
Reviewed-by: Shannon Nelson <shannon.nelson@amd.com>
Link: https://patch.msgid.link/20250206201036.1516800-3-sean.anderson@linux.dev
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Alexander Lobakin says:
====================
xsk: the lost bits from Chapter III
Before introducing libeth_xdp, we need to add a couple more generic
helpers. Notably:
* 01: add generic loop unrolling hint helpers;
* 04: add helper to get both xdp_desc's DMA address and metadata
pointer in one go, saving several cycles and hotpath object
code size in drivers (especially when unrolling).
Bonus:
* 02, 03: convert two drivers which were using custom macros to
generic unrolled_count() (trivial, no object code changes).
====================
Link: https://patch.msgid.link/20250206182630.3914318-1-aleksander.lobakin@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currently, when your driver supports XSk Tx metadata and you want to
send an XSk frame, you need to do the following:
* call external xsk_buff_raw_get_dma();
* call inline xsk_buff_get_metadata(), which calls external
xsk_buff_raw_get_data() and then do some inline checks.
This effectively means that the following piece:
addr = pool->unaligned ? xp_unaligned_add_offset_to_addr(addr) : addr;
is done twice per frame, plus you have 2 external calls per frame, plus
this:
meta = pool->addrs + addr - pool->tx_metadata_len;
if (unlikely(!xsk_buff_valid_tx_metadata(meta)))
is always inlined, even if there's no meta or it's invalid.
Add xsk_buff_raw_get_ctx() (xp_raw_get_ctx() to be precise) to do that
in one go. It returns a small structure with 2 fields: DMA address,
filled unconditionally, and metadata pointer, non-NULL only if it's
present and valid. The address correction is performed only once and
you also have only 1 external call per XSk frame, which does all the
calculations and checks outside of your hotpath. You only need to
check `if (ctx.meta)` for the metadata presence.
To not copy any existing code, derive address correction and getting
virtual and DMA address into small helpers. bloat-o-meter reports no
object code changes for the existing functionality.
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Link: https://patch.msgid.link/20250206182630.3914318-5-aleksander.lobakin@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
There are cases when we need to explicitly unroll loops. For example,
cache operations, filling DMA descriptors on very high speeds etc.
Add compiler-specific attribute macros to give the compiler a hint
that we'd like to unroll a loop.
Example usage:
#define UNROLL_BATCH 8
unrolled_count(UNROLL_BATCH)
for (u32 i = 0; i < UNROLL_BATCH; i++)
op(priv, i);
Note that sometimes the compilers won't unroll loops if they think this
would have worse optimization and perf than without unrolling, and that
unroll attributes are available only starting GCC 8. For older compiler
versions, no hints/attributes will be applied.
For better unrolling/parallelization, don't have any variables that
interfere between iterations except for the iterator itself.
Co-developed-by: Jose E. Marchesi <jose.marchesi@oracle.com> # pragmas
Signed-off-by: Jose E. Marchesi <jose.marchesi@oracle.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Link: https://patch.msgid.link/20250206182630.3914318-2-aleksander.lobakin@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Jakub Kicinski says:
====================
eth: fbnic: support RSS contexts and ntuple filters
Add support for RSS contexts and ntuple filters in fbnic.
The device has only one context, intended for use by TCP zero-copy Rx.
First two patches add a check we seem to be missing in the core,
to avoid having to copy it to all drivers.
$ ./drivers/net/hw/rss_ctx.py
KTAP version 1
1..16
ok 1 rss_ctx.test_rss_key_indir
ok 2 rss_ctx.test_rss_queue_reconfigure
ok 3 rss_ctx.test_rss_resize
ok 4 rss_ctx.test_hitless_key_update
ok 5 rss_ctx.test_rss_context
# Failed to create context 2, trying to test what we got
ok 6 rss_ctx.test_rss_context4 # SKIP Tested only 1 contexts, wanted 4
# Increasing queue count 44 -> 66
# Failed to create context 2, trying to test what we got
ok 7 rss_ctx.test_rss_context32 # SKIP Tested only 1 contexts, wanted 32
# Added only 1 out of 3 contexts
ok 8 rss_ctx.test_rss_context_dump
# Driver does not support rss + queue offset
ok 9 rss_ctx.test_rss_context_queue_reconfigure
ok 10 rss_ctx.test_rss_context_overlap
ok 11 rss_ctx.test_rss_context_overlap2 # SKIP Test requires at least 2 contexts, but device only has 1
ok 12 rss_ctx.test_rss_context_out_of_order # SKIP Test requires at least 4 contexts, but device only has 1
# Failed to create context 2, trying to test what we got
ok 13 rss_ctx.test_rss_context4_create_with_cfg # SKIP Tested only 1 contexts, wanted 4
ok 14 rss_ctx.test_flow_add_context_missing
ok 15 rss_ctx.test_delete_rss_context_busy
ok 16 rss_ctx.test_rss_ntuple_addition # SKIP Ntuple filter with RSS and nonzero action not supported
# Totals: pass:10 fail:0 xfail:0 xpass:0 skip:6 error:0
====================
Link: https://patch.msgid.link/20250206235334.1425329-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
IPv6 addresses are huge so the device has 4 TCAMs used for narrowing
them down to a smaller key before the main match / action engine.
Add the tables in which we'll keep the IP addresses used by
ethtool n-tuple rules. Add the code for programming them
into the device, and code for allocating and freeing entries.
A bit of copy / paste here as we need to support IPv4 and
IPv6 in the same tables, and there is four of them.
But it makes the code easier to match up with the device.
Signed-off-by: Alexander Duyck <alexanderduyck@fb.com>
Link: https://patch.msgid.link/20250206235334.1425329-5-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Breno Leitao says:
====================
netconsole: Add support for CPU population
The current implementation of netconsole sends all log messages in
parallel, which can lead to an intermixed and interleaved output on the
receiving side. This makes it challenging to demultiplex the messages
and attribute them to their originating CPUs.
As a result, users and developers often struggle to effectively analyze
and debug the parallel log output received through netconsole.
Example of a message got from produciton hosts:
------------[ cut here ]------------
------------[ cut here ]------------
refcount_t: saturated; leaking memory.
WARNING: CPU: 2 PID: 1613668 at lib/refcount.c:22 refcount_warn_saturate+0x5e/0xe0
refcount_t: addition on 0; use-after-free.
WARNING: CPU: 26 PID: 4139916 at lib/refcount.c:25 refcount_warn_saturate+0x7d/0xe0
Modules linked in: bpf_preload(E) vhost_net(E) tun(E) vhost(E)
This series of patches introduces a new feature to the netconsole
subsystem that allows the automatic population of the CPU number in the
userdata field for each log message. This enhancement provides several
benefits:
* Improved demultiplexing of parallel log output: When multiple CPUs are
sending messages concurrently, the added CPU number in the userdata
makes it easier to differentiate and attribute the messages to their
originating CPUs.
* Better visibility into message sources: The CPU number information
gives users and developers more insight into which specific CPU a
particular log message came from, which can be valuable for debugging
and analysis.
The changes in this series are as follows Patches::
Patch "consolidate send buffers into netconsole_target struct"
=================================================
Move the static buffers to netconsole target, from static declaration
in send_msg_no_fragmentation() and send_msg_fragmented().
Patch "netconsole: Rename userdata to extradata"
=================================================
Create the a concept of extradata, which encompasses the concept of
userdata and the upcoming sysdatao
Sysdata is a new concept being added, which is basically fields that are
populated by the kernel. At this time only the CPU#, but, there is a
desire to add current task name, kernel release version, etc.
Patch "netconsole: Helper to count number of used entries"
===========================================================
Create a simple helper to count number of entries in extradata. I am
separating this in a function since it will need to count userdata and
sysdata. For instance, when the user adds an extra userdata, we need to
check if there is space, counting the previous data entries (from
userdata and cpu data)
Patch "Introduce configfs helpers for sysdata features"
======================================================
Create the concept of sysdata feature in the netconsole target, and
create the configfs helpers to enable the bit in nt->sysdata
Patch "Include sysdata in extradata entry count"
================================================
Add the concept of sysdata when counting for available space in the
buffer. This will protect users from creating new userdata/sysdata if
there is no more space
Patch "netconsole: add support for sysdata and CPU population"
===============================================================
This is the core patch. Basically add a new option to enable automatic
CPU number population in the netconsole userdata Provides a new "cpu_nr"
sysfs attribute to control this feature
Patch "netconsole: selftest: test CPU number auto-population"
=============================================================
Expands the existing netconsole selftest to verify the CPU number
auto-population functionality Ensures the received netconsole messages
contain the expected "cpu=<CPU>" entry in the message. Test different
permutation with userdata
Patch "netconsole: docs: Add documentation for CPU number auto-population"
=============================================================================
Updates the netconsole documentation to explain the new CPU number
auto-population feature Provides instructions on how to enable and use
the feature
I believe these changes will be a valuable addition to the netconsole
subsystem, enhancing its usefulness for kernel developers and users.
PS: This patchset is on top of the patch that created
netcons_fragmented_msg selftest:
https://lore.kernel.org/all/20250203-netcons_frag_msgs-v1-1-5bc6bedf2ac0@debian.org/
---
Changes in v5:
- Fixed a kernel doc syntax syntax (Simon)
- Link to v4: https://lore.kernel.org/r/20250204-netcon_cpu-v4-0-9480266ef556@debian.org
Changes in v4:
- Fixed Kernel doc for netconsole_target (Simon)
- Fixed a typo in disable_sysdata_feature (Simon)
- Improved sysdata_cpu_nr_show() to return !! in a bit-wise operation
- Link to v3: https://lore.kernel.org/r/20250124-netcon_cpu-v3-0-12a0d286ba1d@debian.org
Changes in v3:
- Moved the buffer into netconsole_target, avoiding static functions in
the send path (Jakub).
- Fix a documentation error (Randy Dunlap)
- Created a function that handle all the extradata, consolidating it in
a single place (Jakub)
- Split the patch even more, trying to simplify the review.
- Link to v2: https://lore.kernel.org/r/20250115-netcon_cpu-v2-0-95971b44dc56@debian.org
Changes in v2:
- Create the concept of extradata and sysdata. This will make the design
easier to understand, and the code easier to read.
* Basically extradata encompasses userdata and the new sysdata.
Userdata originates from user, and sysdata originates in kernel.
- Improved the test to send from a very specific CPU, which can be
checked to be correct on the other side, as suggested by Jakub.
- Fixed a bug where CPU # was populated at the wrong place
- Link to v1: https://lore.kernel.org/r/20241113-netcon_cpu-v1-0-d187bf7c0321@debian.org
====================
Signed-off-by: Breno Leitao <leitao@debian.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Update the netconsole documentation to explain the new feature that
allows automatic population of the CPU number.
The key changes include introducing a new section titled "CPU number
auto population in userdata", explaining how to enable the CPU number
auto-population feature by writing to the "populate_cpu_nr" file in the
netconsole configfs hierarchy.
This documentation update ensures users are aware of the new CPU number
auto-population functionality and how to leverage it for better
demultiplexing and visibility of parallel netconsole output.
Signed-off-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a new selftest to verify that the netconsole module correctly
handles CPU runtime data in sysdata. The test validates three scenarios:
1. Basic CPU sysdata functionality - verifies that cpu=X is appended to
messages
2. CPU sysdata with userdata - ensures CPU data works alongside userdata
3. Disabled CPU sysdata - confirms no CPU data is included when disabled
The test uses taskset to control which CPU sends messages and verifies
the reported CPU matches the one used. This helps ensure that netconsole
accurately tracks and reports the originating CPU of messages.
Signed-off-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add infrastructure to automatically append kernel-generated data (sysdata)
to netconsole messages. As the first use case, implement CPU number
population, which adds the CPU that sent the message.
This change introduces three distinct data types:
- extradata: The complete set of appended data (sysdata + userdata)
- userdata: User-provided key-value pairs from userspace
- sysdata: Kernel-populated data (e.g. cpu=XX)
The implementation adds a new configfs attribute 'cpu_nr' to control CPU
number population per target. When enabled, each message is tagged with
its originating CPU. The sysdata is dynamically updated at message time
and appended after any existing userdata.
The CPU number is formatted as "cpu=XX" and is added to the extradata
buffer, respecting the existing size limits.
Signed-off-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Modify count_extradata_entries() to include sysdata fields when
calculating the total number of extradata entries. This change ensures
that the sysdata feature, specifically the CPU number field, is
correctly counted against the MAX_EXTRADATA_ITEMS limit.
The modification adds a simple check for the CPU_NR flag in the
sysdata_fields, incrementing the entry count accordingly.
Signed-off-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch introduces a bitfield to store sysdata features in the
netconsole_target struct. It also adds configfs helpers to enable
or disable the CPU_NR feature, which populates the CPU number in
sysdata.
The patch provides the necessary infrastructure to set or unset the
CPU_NR feature, but does not modify the message itself.
Signed-off-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a helper function nr_extradata_entries() to count the number of used
extradata entries in a netconsole target. This refactors the duplicate
code for counting entries into a single function, which will be reused
by upcoming CPU sysdata changes.
The helper uses list_count_nodes() to count the number of children in
the userdata group configfs hierarchy.
Signed-off-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Rename "userdata" to "extradata" since this structure will hold both
user and system data in future patches. Keep "userdata" term only for
data that comes from userspace (configfs), while "extradata" encompasses
both userdata and future kerneldata.
These are the rules of the design
1. extradata_complete will hold userdata and sysdata (coming)
2. sysdata will come after userdata_length
3. extradata_complete[userdata_length] string will be replaced at every
message
5. userdata is replaced when configfs changes (update_userdata())
6. sysdata is replaced at every message
Example:
extradata_complete = "userkey=uservalue cpu=42"
userdata_length = 17
sysdata_length = 7 (space (" ") is part of sysdata)
Since sysdata is still not available, you will see the following in the
send functions:
extradata_len = nt->userdata_length;
The upcoming patches will, which will add support for sysdata, will
change it to:
extradata_len = nt->userdata_length + sysdata_len;
Signed-off-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move the static buffers from send_msg_no_fragmentation() and
send_msg_fragmented() into the netconsole_target structure. This
simplifies the code by:
- Eliminating redundant static buffers
- Centralizing buffer management in the target structure
- Reducing memory usage by 1KB (one buffer instead of two)
The buffer in netconsole_target is protected by target_list_lock,
maintaining the same synchronization semantics as the original code.
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jakub Kicinski says:
====================
net: improve core queue API handling while device is down
The core netdev_rx_queue_restart() doesn't currently take into account
that the device may be down. The current and proposed queue API
implementations deal with this by rejecting queue API calls while
the device is down. We can do better, in theory we can still allow
devmem binding when the device is down - we shouldn't stop and start
the queues just try to allocate the memory. The reason we allocate
the memory is that memory provider binding checks if any compatible
page pool has been created (page_pool_check_memory_provider()).
Alternatively we could reject installing MP while the device is down
but the MP assignment survives ifdown (so presumably MP doesn't cease
to exist while down), and in general we allow configuration while down.
Previously I thought we need this as a fix, but gve rejects page pool
calls while down, and so did Saeed in the patches he posted. So this
series just makes the core act more sensibly but practically should
be a noop for now.
v1: https://lore.kernel.org/20250205190131.564456-1-kuba@kernel.org
====================
Link: https://patch.msgid.link/20250206225638.1387810-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
We seem to be missing a netif_running() check from the devmem
installation path. Starting a queue on a stopped device makes
no sense. We still want to be able to allocate the memory, just
to test that the device is indeed setting up the page pools
in a memory provider compatible way.
This is not a bug fix, because existing drivers check if
the interface is down as part of the ops. But new drivers
shouldn't have to do this, as long as they can correctly
alloc/free while down.
Reviewed-by: Mina Almasry <almasrymina@google.com>
Link: https://patch.msgid.link/20250206225638.1387810-3-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Fabio Porcedda says:
====================
Add usb support for Telit Cinterion FN990B
Add usb support for Telit Cinterion FN990B.
Also fix Telit Cinterion FN990A name.
Connection with ModemManager was tested also AT ports.
====================
Link: https://patch.msgid.link/20250205171649.618162-1-fabio.porcedda@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Russell King says:
====================
net: stmmac: yet more EEE updates
Continuing on with the STMMAC EEE cleanups from last cycle, this series
further cleans up the EEE code, and fixes a problem with the existing
implementation - disabling EEE doesn't immediately disable LPI
signalling until the next packet is transmitted. It likely also fixes
a potential race condition when trying to disable LPI vs the software
timer.
====================
Link: https://patch.msgid.link/Z6NqGnM2yL7Ayo-T@shell.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add a new method to control LPI mode configuration. This is architected
to have three configuration states: LPI disabled, LPI forced (active),
or LPI under hardware timer control. This reflects the three modes
which the main body of the driver wishes to deal with.
We pass in whether transmit clock gating should be used, and the
hardware timer value in microseconds to be set when using hardware
timer control.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://patch.msgid.link/E1tffds-003ZIT-E8@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The bit definitions for the LPI control/status register are
identical across all MAC versions, with the exception that some
bits may not be implemented. Provide definitions for bits in this
register in common.h, convert to use them, and remove the core-
specific definitions.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://patch.msgid.link/E1tffdn-003ZIN-9p@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Phylink will not call the mac_disable_tx_lpi() and mac_enable_tx_lpi()
methods randomly - the first method to be called will be the enable
method, and then after, the disable method will be called once between
subsequent enable calls. Thus there is a guaranteed ordering.
Therefore, we know the previous state of priv->eee_enabled, and can
remove it from both methods.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://patch.msgid.link/E1tffdX-003ZI5-UV@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The tests for priv->dma_cap.eee in stmmac_mac_{en,dis}able_tx_lpi()
is useless as these methods will only be called when using phylink
managed EEE, and that will only be enabled if the LPI capabilities
in phylink_config have been populated during initialisation. This
only occurs when priv->dma_cap.eee was true.
As priv->dma_cap.eee remains constant during the lifetime of the driver
instance, there is no need to re-check it in these methods.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://patch.msgid.link/E1tffdN-003ZHt-Mq@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
