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>
LPIATE enables the hardware timer for entering LPI mode. To sure that
the correct mode is used, clear LPIATE when using manual/software-timed
mode to prevent the hardware using the timer.
stmmac_main.c avoids this being a problem at the moment by calling
stmmac_set_eee_lpi_timer(..., 0) before switching to software mode.
We no longer need to call stmmac_set_eee_lpi_timer(..., 0) when
disabling EEE as stmmac_reset_eee_mode() will now clear all LPI
settings.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://patch.msgid.link/E1tffdD-003ZHh-Ew@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When EEE is disabled, we call stmmac_set_eee_lpi_timer(..., 0).
For dwmac4, this will result in LPIATE being cleared, but LPIEN and
LPITXA being set, causing LPI mode to be signalled (if it wasn't
before).
For others MACs, stmmac_set_eee_lpi_timer() does nothing, which means
that LPI mode will continue to be signalled despite the expectation
for it to be disabled.
In both cases, LPI mode will be terminated when the transmitter has
a packet to send, and LPIEN will be cleared by hardware.
Call stmmac_reset_eee_mode() to ensure that LPI mode is disabled when
EEE mode is requested to be disabled.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://patch.msgid.link/E1tffd8-003ZHb-AX@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Biju Das says:
====================
Add of_get_available_child_by_name()
There are lot of net drivers using of_get_child_by_name() followed by
of_device_is_available() to find the available child node by name for a
given parent. Provide a helper for these users to simplify the code.
v1->v2:
* Make it as a series as per [1] to cover the dependency.
* Added Rb tag from Rob for patch#1 and this patch can be merged through
net as it is the main user.
* Updated all the patches with patch suffix net-next
* Dropped _free() usage.
[1]
https://lore.kernel.org/all/CAL_JsqLo4uSGYMcLXN=0iSUMHdW8RaGCY+o8ThQHq3_eUTV9wQ@mail.gmail.com/
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Simplify a5psw_probe() by using of_get_available_child_by_name().
While at it, move of_node_put(mdio) inside the if block to avoid code
duplication.
Signed-off-by: Biju Das <biju.das.jz@bp.renesas.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
There are lot of drivers using of_get_child_by_name() followed by
of_device_is_available() to find the available child node by name for a
given parent. Provide a helper for these users to simplify the code.
Suggested-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Biju Das <biju.das.jz@bp.renesas.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Tony Nguyen says:
====================
ice: managing MSI-X in driver
Michal Swiatkowski says:
It is another try to allow user to manage amount of MSI-X used for each
feature in ice. First was via devlink resources API, it wasn't accepted
in upstream. Also static MSI-X allocation using devlink resources isn't
really user friendly.
This try is using more dynamic way. "Dynamic" across whole kernel when
platform supports it and "dynamic" across the driver when not.
To achieve that reuse global devlink parameter pf_msix_max and
pf_msix_min. It fits how ice hardware counts MSI-X. In case of ice amount
of MSI-X reported on PCI is a whole MSI-X for the card (with MSI-X for
VFs also). Having pf_msix_max allow user to statically set how many
MSI-X he wants on PF and how many should be reserved for VFs.
pf_msix_min is used to set minimum number of MSI-X with which ice driver
should probe correctly.
Meaning of this field in case of dynamic vs static allocation:
- on system with dynamic MSI-X allocation support
* alloc pf_msix_min as static, rest will be allocated dynamically
- on system without dynamic MSI-X allocation support
* try alloc pf_msix_max as static, minimum acceptable result is
pf_msix_min
As Jesse and Piotr suggested pf_msix_max and pf_msix_min can (an
probably should) be stored in NVM. This patchset isn't implementing
that.
Dynamic (kernel or driver) way means that splitting MSI-X across the
RDMA and eth in case there is a MSI-X shortage isn't correct. Can work
when dynamic is only on driver site, but can't when dynamic is on kernel
site.
Let's remove this code and move to MSI-X allocation feature by feature.
If there is no more MSI-X for a feature, a feature is working with less
MSI-X or it is turned off.
There is a regression here. With MSI-X splitting user can run RDMA and
eth even on system with not enough MSI-X. Now only eth will work. RDMA
can be turned on by changing number of PF queues (lowering) and reprobe
RDMA driver.
Example:
72 CPU number, eth, RDMA and flow director (1 MSI-X), 1 MSI-X for OICR
on PF, and 1 more for RDMA. Card is using 1 + 72 + 1 + 72 + 1 = 147.
We set pf_msix_min = 2, pf_msix_max = 128
OICR: 1
eth: 72
flow director: 1
RDMA: 128 - 74 = 54
We can change number of queues on pf to 36 and do devlink reinit
OICR: 1
eth: 36
RDMA: 73
flow director: 1
We can also (implemented in "ice: enable_rdma devlink param") turned
RDMA off.
OICR: 1
eth: 72
RDMA: 0 (turned off)
flow director: 1
After this changes we have a static base vector for SRIOV (SIOV probably
in the feature). Last patch from this series is simplifying managing VF
MSI-X code based on static vector.
Now changing queues using ethtool is also changing MSI-X. If there is
enough MSI-X it is always one to one. When there is not enough there
will be more queues than MSI-X.
* 'for-next' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue:
ice: init flow director before RDMA
ice: simplify VF MSI-X managing
ice: enable_rdma devlink param
ice: treat dyn_allowed only as suggestion
ice, irdma: move interrupts code to irdma
ice: get rid of num_lan_msix field
ice: remove splitting MSI-X between features
ice: devlink PF MSI-X max and min parameter
ice: count combined queues using Rx/Tx count
====================
Link: https://patch.msgid.link/20250205185512.895887-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
John Daley says:
====================
enic: Use Page Pool API for receiving packets
Use the Page Pool API for RX. The Page Pool API improves bandwidth and
CPU overhead by recycling pages instead of allocating new buffers in the
driver. Also, page pool fragment allocation for smaller MTUs is used
allow multiple packets to share pages.
RX code was moved to its own file and some refactoring was done
beforehand to make the page pool changes more trasparent and to simplify
the resulting code.
====================
Link: https://patch.msgid.link/20250205235416.25410-1-johndale@cisco.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
With the move to using the Page Pool API for RX, rx copybreak was not
showing any improvement in host CPU overhead, latency or bandwidth so
the driver no longer makes use of the rx_copybreak setting. This patch
removes the ethtool tuneable hooks to set and get the rx copybreak since
they and now no-ops. Rx copybreak was the only tunable supported, so
remove the set and get tunable callbacks all together.
Co-developed-by: Nelson Escobar <neescoba@cisco.com>
Signed-off-by: Nelson Escobar <neescoba@cisco.com>
Co-developed-by: Satish Kharat <satishkh@cisco.com>
Signed-off-by: Satish Kharat <satishkh@cisco.com>
Signed-off-by: John Daley <johndale@cisco.com>
Link: https://patch.msgid.link/20250205235416.25410-5-johndale@cisco.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The Page Pool API improves bandwidth and CPU overhead by recycling pages
instead of allocating new buffers in the driver. Make use of page pool
fragment allocation for smaller MTUs so that multiple packets can share
a page. For MTUs larger than PAGE_SIZE, adjust the 'order' page
parameter so that contiguous pages can be used to receive the larger
packets.
The RQ descriptor field 'os_buf' is repurposed to hold page pointers
allocated from page_pool instead of SKBs. When packets arrive, SKBs are
allocated and the page pointers are attached instead of preallocating SKBs.
'alloc_fail' netdev statistic is incremented when page_pool_dev_alloc()
fails.
Co-developed-by: Nelson Escobar <neescoba@cisco.com>
Signed-off-by: Nelson Escobar <neescoba@cisco.com>
Co-developed-by: Satish Kharat <satishkh@cisco.com>
Signed-off-by: Satish Kharat <satishkh@cisco.com>
Signed-off-by: John Daley <johndale@cisco.com>
Link: https://patch.msgid.link/20250205235416.25410-4-johndale@cisco.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
There are at least two cases where napi_id may not present and the
napi_id should be elided:
1. Queues could be created, but napi_enable may not have been called
yet. In this case, there may be a NAPI but it may not have an ID and
output of a napi_id should be elided.
2. TX-only NAPIs currently do not have NAPI IDs. If a TX queue happens
to be linked with a TX-only NAPI, elide the NAPI ID from the netlink
output as a NAPI ID of 0 is not useful for users.
Signed-off-by: Joe Damato <jdamato@fastly.com>
Reviewed-by: Sridhar Samudrala <sridhar.samudrala@intel.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20250205193751.297211-1-jdamato@fastly.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
David Wei says:
====================
io_uring zero copy rx
This patchset contains net/ patches needed by a new io_uring request
implementing zero copy rx into userspace pages, eliminating a kernel
to user copy.
We configure a page pool that a driver uses to fill a hw rx queue to
hand out user pages instead of kernel pages. Any data that ends up
hitting this hw rx queue will thus be dma'd into userspace memory
directly, without needing to be bounced through kernel memory. 'Reading'
data out of a socket instead becomes a _notification_ mechanism, where
the kernel tells userspace where the data is. The overall approach is
similar to the devmem TCP proposal.
This relies on hw header/data split, flow steering and RSS to ensure
packet headers remain in kernel memory and only desired flows hit a hw
rx queue configured for zero copy. Configuring this is outside of the
scope of this patchset.
We share netdev core infra with devmem TCP. The main difference is that
io_uring is used for the uAPI and the lifetime of all objects are bound
to an io_uring instance. Data is 'read' using a new io_uring request
type. When done, data is returned via a new shared refill queue. A zero
copy page pool refills a hw rx queue from this refill queue directly. Of
course, the lifetime of these data buffers are managed by io_uring
rather than the networking stack, with different refcounting rules.
This patchset is the first step adding basic zero copy support. We will
extend this iteratively with new features e.g. dynamically allocated
zero copy areas, THP support, dmabuf support, improved copy fallback,
general optimisations and more.
In terms of netdev support, we're first targeting Broadcom bnxt. Patches
aren't included since Taehee Yoo has already sent a more comprehensive
patchset adding support in [1]. Google gve should already support this,
and Mellanox mlx5 support is WIP pending driver changes.
===========
Performance
===========
Note: Comparison with epoll + TCP_ZEROCOPY_RECEIVE isn't done yet.
Test setup:
* AMD EPYC 9454
* Broadcom BCM957508 200G
* Kernel v6.11 base [2]
* liburing fork [3]
* kperf fork [4]
* 4K MTU
* Single TCP flow
With application thread + net rx softirq pinned to _different_ cores:
+-------------------------------+
| epoll | io_uring |
|-----------|-------------------|
| 82.2 Gbps | 116.2 Gbps (+41%) |
+-------------------------------+
Pinned to _same_ core:
+-------------------------------+
| epoll | io_uring |
|-----------|-------------------|
| 62.6 Gbps | 80.9 Gbps (+29%) |
+-------------------------------+
=====
Links
=====
Broadcom bnxt support:
[1]: https://lore.kernel.org/20241003160620.1521626-8-ap420073@gmail.com
Linux kernel branch including io_uring bits:
[2]: https://github.com/isilence/linux.git zcrx/v13
liburing for testing:
[3]: https://github.com/isilence/liburing.git zcrx/next
kperf for testing:
[4]: https://git.kernel.dk/kperf.git
====================
Link: https://patch.msgid.link/20250204215622.695511-1-dw@davidwei.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add helpers for memory providers to interact with page pools.
net_mp_niov_{set,clear}_page_pool() serve to [dis]associate a net_iov
with a page pool. If used, the memory provider is responsible to match
"set" calls with "clear" once a net_iov is not going to be used by a page
pool anymore, changing a page pool, etc.
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: David Wei <dw@davidwei.uk>
Link: https://patch.msgid.link/20250204215622.695511-10-dw@davidwei.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currently net_iov stores a pointer to struct dmabuf_genpool_chunk_owner,
which serves as a useful abstraction to share data and provide a
context. However, it's too devmem specific, and we want to reuse it for
other memory providers, and for that we need to decouple net_iov from
devmem. Make net_iov to point to a new base structure called
net_iov_area, which dmabuf_genpool_chunk_owner extends.
Reviewed-by: Mina Almasry <almasrymina@google.com>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: David Wei <dw@davidwei.uk>
Link: https://patch.msgid.link/20250204215622.695511-4-dw@davidwei.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The Makefile.deps lists uAPI headers to make the build work when
system headers are older than in-tree headers. The problem doesn't
occur for new headers, because system headers are not there at all.
But out-of-tree YNL clone on GH also uses this header to identify
header dependencies, and one day the system headers will exist,
and will get out of date. So let's add the headers we missed.
I don't think this is a fix, but FWIW the commits which added
the missing headers are:
commit 04e65df94b ("netlink: spec: add shaper YAML spec")
commit 49922401c2 ("ethtool: separate definitions that are gonna be generated")
Reviewed-by: Donald Hunter <donald.hunter@gmail.com>
Link: https://patch.msgid.link/20250205173352.446704-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Pull networking fixes from Paolo Abeni:
"Interestingly the recent kmemleak improvements allowed our CI to catch
a couple of percpu leaks addressed here.
We (mostly Jakub, to be accurate) are working to increase review
coverage over the net code-base tweaking the MAINTAINER entries.
Current release - regressions:
- core: harmonize tstats and dstats
- ipv6: fix dst refleaks in rpl, seg6 and ioam6 lwtunnels
- eth: tun: revert fix group permission check
- eth: stmmac: revert "specify hardware capability value when FIFO
size isn't specified"
Previous releases - regressions:
- udp: gso: do not drop small packets when PMTU reduces
- rxrpc: fix race in call state changing vs recvmsg()
- eth: ice: fix Rx data path for heavy 9k MTU traffic
- eth: vmxnet3: fix tx queue race condition with XDP
Previous releases - always broken:
- sched: pfifo_tail_enqueue: drop new packet when sch->limit == 0
- ethtool: ntuple: fix rss + ring_cookie check
- rxrpc: fix the rxrpc_connection attend queue handling
Misc:
- recognize Kuniyuki Iwashima as a maintainer"
* tag 'net-6.14-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (34 commits)
Revert "net: stmmac: Specify hardware capability value when FIFO size isn't specified"
MAINTAINERS: add a sample ethtool section entry
MAINTAINERS: add entry for ethtool
rxrpc: Fix race in call state changing vs recvmsg()
rxrpc: Fix call state set to not include the SERVER_SECURING state
net: sched: Fix truncation of offloaded action statistics
tun: revert fix group permission check
selftests/tc-testing: Add a test case for qdisc_tree_reduce_backlog()
netem: Update sch->q.qlen before qdisc_tree_reduce_backlog()
selftests/tc-testing: Add a test case for pfifo_head_drop qdisc when limit==0
pfifo_tail_enqueue: Drop new packet when sch->limit == 0
selftests: mptcp: connect: -f: no reconnect
net: rose: lock the socket in rose_bind()
net: atlantic: fix warning during hot unplug
rxrpc: Fix the rxrpc_connection attend queue handling
net: harmonize tstats and dstats
selftests: drv-net: rss_ctx: don't fail reconfigure test if queue offset not supported
selftests: drv-net: rss_ctx: add missing cleanup in queue reconfigure
ethtool: ntuple: fix rss + ring_cookie check
ethtool: rss: fix hiding unsupported fields in dumps
...
This reverts commit 8865d22656, which caused breakage for platforms
which are not using xgmac2 or gmac4. Only these two cores have the
capability of providing the FIFO sizes from hardware capability fields
(which are provided in priv->dma_cap.[tr]x_fifo_size.)
All other cores can not, which results in these two fields containing
zero. We also have platforms that do not provide a value in
priv->plat->[tr]x_fifo_size, resulting in these also being zero.
This causes the new tests introduced by the reverted commit to fail,
and produce e.g.:
stmmaceth f0804000.eth: Can't specify Rx FIFO size
An example of such a platform which fails is QEMU's npcm750-evb.
This uses dwmac1000 which, as noted above, does not have the capability
to provide the FIFO sizes from hardware.
Therefore, revert the commit to maintain compatibility with the way
the driver used to work.
Reported-by: Guenter Roeck <linux@roeck-us.net>
Link: https://lore.kernel.org/r/4e98f967-f636-46fb-9eca-d383b9495b86@roeck-us.net
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Tested-by: Steven Price <steven.price@arm.com>
Fixes: 8865d22656 ("net: stmmac: Specify hardware capability value when FIFO size isn't specified")
Link: https://patch.msgid.link/E1tfeyR-003YGJ-Gb@rmk-PC.armlinux.org.uk
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
