The dpaa-eth driver is written for PowerPC and Arm SoCs which have 1-24
CPUs. It depends on CONFIG_NR_CPUS having a reasonably small value in
Kconfig. Otherwise, there are 2 functions which allocate on-stack arrays
of NR_CPUS elements, and these can quickly explode in size, leading to
warnings such as:
drivers/net/ethernet/freescale/dpaa/dpaa_eth.c:3280:12: warning:
stack frame size (16664) exceeds limit (2048) in 'dpaa_eth_probe' [-Wframe-larger-than]
The problem is twofold:
- Reducing the array size to the boot-time num_possible_cpus() (rather
than the compile-time NR_CPUS) creates a variable-length array,
which should be avoided in the Linux kernel.
- Using NR_CPUS as an array size makes the driver blow up in stack
consumption with generic, as opposed to hand-crafted, .config files.
A simple solution is to use dynamic allocation for num_possible_cpus()
elements (aka a small number determined at runtime).
Link: https://lore.kernel.org/all/202406261920.l5pzM1rj-lkp@intel.com/
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Breno Leitao <leitao@debian.org>
Acked-by: Madalin Bucur <madalin.bucur@oss.nxp.com>
Link: https://patch.msgid.link/20240713225336.1746343-2-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Steffen Klassert says:
====================
pull request (net-next): ipsec-next 2024-07-13
1) Support sending NAT keepalives in ESP in UDP states.
Userspace IKE daemon had to do this before, but the
kernel can better keep track of it.
From Eyal Birger.
2) Support IPsec crypto offload for IPv6 ESP and IPv4 UDP-encapsulated
ESP data paths. Currently, IPsec crypto offload is enabled for GRO
code path only. This patchset support UDP encapsulation for the non
GRO path. From Mike Yu.
* tag 'ipsec-next-2024-07-13' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec-next:
xfrm: Support crypto offload for outbound IPv4 UDP-encapsulated ESP packet
xfrm: Support crypto offload for inbound IPv4 UDP-encapsulated ESP packet
xfrm: Allow UDP encapsulation in crypto offload control path
xfrm: Support crypto offload for inbound IPv6 ESP packets not in GRO path
xfrm: support sending NAT keepalives in ESP in UDP states
====================
Link: https://patch.msgid.link/20240713102416.3272997-1-steffen.klassert@secunet.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Lorenzo Bianconi says:
====================
Introduce EN7581 ethernet support
Add airoha_eth driver in order to introduce ethernet support for
Airoha EN7581 SoC available on EN7581 development board.
EN7581 mac controller is mainly composed by Frame Engine (FE) and
QoS-DMA (QDMA) modules. FE is used for traffic offloading (just basic
functionalities are supported now) while QDMA is used for DMA operation
and QOS functionalities between mac layer and the dsa switch (hw QoS is
not available yet and it will be added in the future).
Currently only hw lan features are available, hw wan will be added with
subsequent patches.
====================
Link: https://patch.msgid.link/cover.1720818878.git.lorenzo@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add airoha_eth driver in order to introduce ethernet support for
Airoha EN7581 SoC available on EN7581 development board (en7581-evb).
EN7581 mac controller is mainly composed by the Frame Engine (PSE+PPE)
and QoS-DMA (QDMA) modules. FE is used for traffic offloading (just
basic functionalities are currently supported) while QDMA is used for
DMA operations and QOS functionalities between the mac layer and the
external modules conncted to the FE GDM ports (e.g MT7530 DSA switch
or external phys).
A general overview of airoha_eth architecture is reported below:
┌───────┐ ┌───────┐
│ QDMA2 │ │ QDMA1 │
└───┬───┘ └───┬───┘
│ │
┌───────▼─────────────────────────────────────────────▼────────┐
│ │
│ P5 P0 │
│ │
│ │
│ │ ┌──────┐
│ P3 ├────► GDM3 │
│ │ └──────┘
│ │
│ │
┌─────┐ │ │
│ PPE ◄────┤ P4 PSE │
└─────┘ │ │
│ │
│ │
│ │ ┌──────┐
│ P9 ├────► GDM4 │
│ │ └──────┘
│ │
│ │
│ │
│ P2 P1 │
└─────────┬───────────────────────────────────────────┬────────┘
│ │
┌───▼──┐ ┌──▼───┐
│ GDM2 │ │ GDM1 │
└──────┘ └──┬───┘
│
┌────▼─────┐
│ MT7530 │
└──────────┘
Currently only hw LAN features (QDMA1+GDM1) are available while hw WAN
(QDMA2+GDM{2,3,4}) ones will be added with subsequent patches introducing
traffic offloading support.
Tested-by: Benjamin Larsson <benjamin.larsson@genexis.eu>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Link: https://patch.msgid.link/274945d2391c195098ab180a46d0617b18b9e42c.1720818878.git.lorenzo@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Tony Nguyen says:
====================
ice: Switch API optimizations
Marcin Szycik says:
Optimize the process of creating a recipe in the switch block by removing
duplicate switch ID words and changing how result indexes are fitted into
recipes. In many cases this can decrease the number of recipes required to
add a certain set of rules, potentially allowing a more varied set of rules
to be created. Total rule count will also increase, since less words will
be left unused/wasted. There are only 64 rules available in total, so every
one counts.
After this modification, many fields and some structs became unused or were
simplified, resulting in overall simpler implementation.
* '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue:
ice: Add tracepoint for adding and removing switch rules
ice: Remove unused members from switch API
ice: Optimize switch recipe creation
ice: remove unused recipe bookkeeping data
ice: Simplify bitmap setting in adding recipe
ice: Remove reading all recipes before adding a new one
ice: Remove unused struct ice_prot_lkup_ext members
====================
Link: https://patch.msgid.link/20240711181312.2019606-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Tony Nguyen says:
====================
Intel Wired LAN Driver Updates 2024-07-11 (net/intel)
This series contains updates to most Intel network drivers.
Tony removes MODULE_AUTHOR from drivers containing the entry.
Simon Horman corrects a kdoc entry for i40e.
Pawel adds implementation for devlink param "local_forwarding" on ice.
Michal removes unneeded call, and code, for eswitch rebuild for ice.
Sasha removed a no longer used field from igc.
* '40GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue:
igc: Remove the internal 'eee_advert' field
ice: remove eswitch rebuild
ice: Add support for devlink local_forwarding param
i40e: correct i40e_addr_to_hkey() name in kdoc
net: intel: Remove MODULE_AUTHORs
====================
Link: https://patch.msgid.link/20240711201932.2019925-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This patch implements the TDR test procedure as described in
"Application Note DP83TD510E Cable Diagnostics Toolkit revC", section 3.2.
The procedure was tested with "draka 08 signalkabel 2x0.8mm". The reported
cable length was 5 meters more for each 20 meters of actual cable length.
For instance, a 20-meter cable showed as 25 meters, and a 40-meter cable
showed as 50 meters. Since other parts of the diagnostics provided by this
PHY (e.g., Active Link Cable Diagnostics) require accurate cable
characterization to provide proper results, this tuning can be implemented
in a separate patch/interface.
Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
changes v2:
- add comments
- change post silence time to 1000ms
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://patch.msgid.link/20240712152848.2479912-1-o.rempel@pengutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
ARFS depends on NTUPLE filters, but the inverse is not true.
Drivers which don't support ARFS commonly still support NTUPLE
filtering. mlx5 has a Kconfig option to disable ARFS (MLX5_EN_ARFS)
and does not advertise NTUPLE filters as a feature at all when ARFS
is compiled out. That's not correct, ntuple filters indeed still work
just fine (as long as MLX5_EN_RXNFC is enabled).
This is needed to make the RSS test not skip all RSS context
related testing.
Acked-by: Saeed Mahameed <saeedm@nvidia.com>
Link: https://patch.msgid.link/20240711223722.297676-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
It looks like we missed these two errors recently:
- SC2068: Double quote array expansions to avoid re-splitting elements.
- SC2145: Argument mixes string and array. Use * or separate argument.
Two simple fixes, it is not supposed to change the behaviour as the
variable names should not have any spaces in their names. Still, better
to fix them to easily spot new issues.
Fixes: f265d3119a ("selftests: mptcp: lib: use setup/cleanup_ns helpers")
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20240712-upstream-net-next-20240712-selftests-mptcp-fix-shellcheck-v1-1-1cb7180db40a@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Expose a new cap sf_eq_usage. The vhca_resource_manager can write this
cap, indicating the SF driver should use max_num_eqs_24b to determine
how many EQs to use.
Will be used in the next patch, to indicate to the SF driver from the PF
that the user has set the max io eqs via devlink. So the SF driver can
later query the proper max eq value from the new cap.
devlink port function set pci/0000:08:00.0/32768 max_io_eqs 32
Signed-off-by: Daniel Jurgens <danielj@nvidia.com>
Reviewed-by: William Tu <witu@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Link: https://patch.msgid.link/20240712003310.355106-2-saeed@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Change the data type of the variable freq in mvpp2_rx_time_coal_set()
and mvpp2_tx_time_coal_set() to u32 because port->priv->tclk also has
the data type u32.
Change the data type of the function parameter clk_hz in
mvpp2_usec_to_cycles() and mvpp2_cycles_to_usec() to u32 accordingly
and remove the following Coccinelle/coccicheck warning reported by
do_div.cocci:
WARNING: do_div() does a 64-by-32 division, please consider using div64_ul instead
Use min() to simplify the code and improve its readability.
Compile-tested only.
Signed-off-by: Thorsten Blum <thorsten.blum@toblux.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20240711154741.174745-1-thorsten.blum@toblux.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currently, during the module firmware flashing process, unicast
notifications are sent from the kernel using the same sequence number,
making it impossible for user space to track missed notifications.
Monotonically increase the message sequence number, so the order of
notifications could be tracked effectively.
Signed-off-by: Danielle Ratson <danieller@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Link: https://patch.msgid.link/20240711080934.2071869-1-danieller@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
tcp_ao/self-connect.c checked the following SNMP stats before/after
connect() to confirm that the test exercises the simultaneous connect()
path.
* TCPChallengeACK
* TCPSYNChallenge
But the stats should not be counted for self-connect in the first place,
and the assumption is no longer true.
Let's remove the check.
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Dmitry Safonov <dima@arista.com>
Link: https://patch.msgid.link/20240710171246.87533-3-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
RFC 9293 states that in the case of simultaneous connect(), the connection
gets established when SYN+ACK is received. [0]
TCP Peer A TCP Peer B
1. CLOSED CLOSED
2. SYN-SENT --> <SEQ=100><CTL=SYN> ...
3. SYN-RECEIVED <-- <SEQ=300><CTL=SYN> <-- SYN-SENT
4. ... <SEQ=100><CTL=SYN> --> SYN-RECEIVED
5. SYN-RECEIVED --> <SEQ=100><ACK=301><CTL=SYN,ACK> ...
6. ESTABLISHED <-- <SEQ=300><ACK=101><CTL=SYN,ACK> <-- SYN-RECEIVED
7. ... <SEQ=100><ACK=301><CTL=SYN,ACK> --> ESTABLISHED
However, since commit 0c24604b68 ("tcp: implement RFC 5961 4.2"), such a
SYN+ACK is dropped in tcp_validate_incoming() and responded with Challenge
ACK.
For example, the write() syscall in the following packetdrill script fails
with -EAGAIN, and wrong SNMP stats get incremented.
0 socket(..., SOCK_STREAM|SOCK_NONBLOCK, IPPROTO_TCP) = 3
+0 connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)
+0 > S 0:0(0) <mss 1460,sackOK,TS val 1000 ecr 0,nop,wscale 8>
+0 < S 0:0(0) win 1000 <mss 1000>
+0 > S. 0:0(0) ack 1 <mss 1460,sackOK,TS val 3308134035 ecr 0,nop,wscale 8>
+0 < S. 0:0(0) ack 1 win 1000
+0 write(3, ..., 100) = 100
+0 > P. 1:101(100) ack 1
--
# packetdrill cross-synack.pkt
cross-synack.pkt:13: runtime error in write call: Expected result 100 but got -1 with errno 11 (Resource temporarily unavailable)
# nstat
...
TcpExtTCPChallengeACK 1 0.0
TcpExtTCPSYNChallenge 1 0.0
The problem is that bpf_skops_established() is triggered by the Challenge
ACK instead of SYN+ACK. This causes the bpf prog to miss the chance to
check if the peer supports a TCP option that is expected to be exchanged
in SYN and SYN+ACK.
Let's accept a bare SYN+ACK for active-open TCP_SYN_RECV sockets to avoid
such a situation.
Note that tcp_ack_snd_check() in tcp_rcv_state_process() is skipped not to
send an unnecessary ACK, but this could be a bit risky for net.git, so this
targets for net-next.
Link: https://www.rfc-editor.org/rfc/rfc9293.html#section-3.5-7 [0]
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20240710171246.87533-2-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Mike Yu says:
====================
Currently, IPsec crypto offload is enabled for GRO code path. However, there
are other code paths where the XFRM stack is involved; for example, IPv6 ESP
packets handled by xfrm6_esp_rcv() in ESP layer, and IPv4 UDP-encapsulated
ESP packets handled by udp_rcv() in UDP layer.
This patchset extends the crypto offload support to cover these two cases.
This is useful for devices with traffic accounting (e.g., Android), where GRO
can lead to inaccurate accounting on the underlying network. For example, VPN
traffic might not be counted on the wifi network interface wlan0 if the packets
are handled in GRO code path before entering the network stack for accounting.
Below is the RX data path scenario the crypto offload can be applied to.
+-----------+ +-------+
| HW Driver |-->| wlan0 |--------+
+-----------+ +-------+ |
v
+---------------+ +------+
+------>| Network Stack |-->| Apps |
| +---------------+ +------+
| |
| v
+--------+ +------------+
| ipsec1 |<--| XFRM Stack |
+--------+ +------------+
====================
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Tony Nguyen says:
====================
idpf: XDP chapter I: convert Rx to libeth
Alexander Lobakin says:
XDP for idpf is currently 5 chapters:
* convert Rx to libeth (this);
* convert Tx and stats to libeth;
* generic XDP and XSk code changes, libeth_xdp;
* actual XDP for idpf via libeth_xdp;
* XSk for idpf (^).
Part I does the following:
* splits &idpf_queue into 4 (RQ, SQ, FQ, CQ) and puts them on a diet;
* ensures optimal cacheline placement, strictly asserts CL sizes;
* moves currently unused/dead singleq mode out of line;
* reuses libeth's Rx ptype definitions and helpers;
* uses libeth's Rx buffer management for both header and payload;
* eliminates memcpy()s and coherent DMA uses on hotpath, uses
napi_build_skb() instead of in-place short skb allocation.
Most idpf patches, except for the queue split, removes more lines
than adds.
Expect far better memory utilization and +5-8% on Rx depending on
the case (+17% on skb XDP_DROP :>).
* '200GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue:
idpf: use libeth Rx buffer management for payload buffer
idpf: convert header split mode to libeth + napi_build_skb()
libeth: support different types of buffers for Rx
idpf: remove legacy Page Pool Ethtool stats
idpf: reuse libeth's definitions of parsed ptype structures
idpf: compile singleq code only under default-n CONFIG_IDPF_SINGLEQ
idpf: merge singleq and splitq &net_device_ops
idpf: strictly assert cachelines of queue and queue vector structures
idpf: avoid bloating &idpf_q_vector with big %NR_CPUS
idpf: split &idpf_queue into 4 strictly-typed queue structures
idpf: stop using macros for accessing queue descriptors
libeth: add cacheline / struct layout assertion helpers
page_pool: use __cacheline_group_{begin, end}_aligned()
cache: add __cacheline_group_{begin, end}_aligned() (+ couple more)
====================
Link: https://patch.msgid.link/20240710203031.188081-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Daniel Borkmann says:
====================
pull-request: bpf-next 2024-07-12
We've added 23 non-merge commits during the last 3 day(s) which contain
a total of 18 files changed, 234 insertions(+), 243 deletions(-).
The main changes are:
1) Improve BPF verifier by utilizing overflow.h helpers to check
for overflows, from Shung-Hsi Yu.
2) Fix NULL pointer dereference in resolve_prog_type() for BPF_PROG_TYPE_EXT
when attr->attach_prog_fd was not specified, from Tengda Wu.
3) Fix arm64 BPF JIT when generating code for BPF trampolines with
BPF_TRAMP_F_CALL_ORIG which corrupted upper address bits,
from Puranjay Mohan.
4) Remove test_run callback from lwt_seg6local_prog_ops which never worked
in the first place and caused syzbot reports,
from Sebastian Andrzej Siewior.
5) Relax BPF verifier to accept non-zero offset on KF_TRUSTED_ARGS/
/KF_RCU-typed BPF kfuncs, from Matt Bobrowski.
6) Fix a long standing bug in libbpf with regards to handling of BPF
skeleton's forward and backward compatibility, from Andrii Nakryiko.
7) Annotate btf_{seq,snprintf}_show functions with __printf,
from Alan Maguire.
8) BPF selftest improvements to reuse common network helpers in sk_lookup
test and dropping the open-coded inetaddr_len() and make_socket() ones,
from Geliang Tang.
* tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next: (23 commits)
selftests/bpf: Test for null-pointer-deref bugfix in resolve_prog_type()
bpf: Fix null pointer dereference in resolve_prog_type() for BPF_PROG_TYPE_EXT
selftests/bpf: DENYLIST.aarch64: Skip fexit_sleep again
bpf: use check_sub_overflow() to check for subtraction overflows
bpf: use check_add_overflow() to check for addition overflows
bpf: fix overflow check in adjust_jmp_off()
bpf: Eliminate remaining "make W=1" warnings in kernel/bpf/btf.o
bpf: annotate BTF show functions with __printf
bpf, arm64: Fix trampoline for BPF_TRAMP_F_CALL_ORIG
selftests/bpf: Close obj in error path in xdp_adjust_tail
selftests/bpf: Null checks for links in bpf_tcp_ca
selftests/bpf: Use connect_fd_to_fd in sk_lookup
selftests/bpf: Use start_server_addr in sk_lookup
selftests/bpf: Use start_server_str in sk_lookup
selftests/bpf: Close fd in error path in drop_on_reuseport
selftests/bpf: Add ASSERT_OK_FD macro
selftests/bpf: Add backlog for network_helper_opts
selftests/bpf: fix compilation failure when CONFIG_NF_FLOW_TABLE=m
bpf: Remove tst_run from lwt_seg6local_prog_ops.
bpf: relax zero fixed offset constraint on KF_TRUSTED_ARGS/KF_RCU
...
====================
Link: https://patch.msgid.link/20240712212448.5378-1-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Cross-merge networking fixes after downstream PR.
Conflicts:
drivers/net/ethernet/broadcom/bnxt/bnxt.c
f7ce5eb2cb ("bnxt_en: Fix crash in bnxt_get_max_rss_ctx_ring()")
20c8ad72eb ("eth: bnxt: use the RSS context XArray instead of the local list")
Adjacent changes:
net/ethtool/ioctl.c
503757c809 ("net: ethtool: Fix RSS setting")
eac9122f0c ("net: ethtool: record custom RSS contexts in the XArray")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Jakub Kicinski says:
====================
eth: bnxt: use the new RSS API
Convert bnxt from using the set_rxfh API to separate create/modify/remove
callbacks.
Two small extensions to the core APIs are necessary:
- the ability to discard contexts if for some catastrophic reasons
device can no longer provide them;
- the ability to reserve space in the context for RSS table growth.
The driver is adjusted to store indirection tables on u32 to make
it easier to use core structs directly.
With that out of the way the conversion is fairly straightforward.
Since the opposition to discarding contexts was relatively mild
and its what bnxt does already, I'm sticking to that. We may very
well need to revisit that at a later time.
v1: https://lore.kernel.org/all/20240702234757.4188344-1-kuba@kernel.org/
====================
Link: https://patch.msgid.link/20240711220713.283778-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Instead of allocating a separate indir table in the vnic use
the one already present in the RSS context allocated by the core.
This saves some LoC and also we won't have to worry about syncing
the local version back to the core, once core learns how to dump
contexts.
Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Link: https://patch.msgid.link/20240711220713.283778-12-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
bnxt allocates tables of max size, and changes the used size
based on number of active rings. The unused entries get padded
out with zeros. bnxt_modify_rss() seems to always pad out
the table of the main / default RSS context, instead of
the table of the modified context.
I haven't observed any behavior change due to this patch,
so I don't think it's a fix. Not entirely sure what role
the padding plays, 0 is a valid queue ID.
Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Link: https://patch.msgid.link/20240711220713.283778-10-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Core can allocate space for per-context driver-private data,
use it for struct bnxt_rss_ctx. Inline bnxt_alloc_rss_ctx()
at this point, most of the init (as in the actions bnxt_del_one_rss_ctx()
will undo) is open coded in bnxt_create_rxfh_context(), anyway.
Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Link: https://patch.msgid.link/20240711220713.283778-8-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Use the new ethtool ops for RSS context management. The conversion
is pretty straightforward cut / paste of the right chunks of the
combined handler. Main change is that we let the core pick the IDs
(bitmap will be removed separately for ease of review), so we need
to tell the core when we lose a context.
Since the new API passes rxfh as const, change bnxt_modify_rss()
to also take const.
Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Link: https://patch.msgid.link/20240711220713.283778-5-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Contexts get deleted from FW when the device is down, but they
are kept in SW and re-added back on open. bnxt_set_rxfh_context()
apparently does not want to deal with complexity of dealing with
both the device down and device up cases. This is perhaps acceptable
for creating new contexts, but not being able to delete contexts
makes core-driven cleanups messy. Specifically with the new RSS
API core will try to delete contexts automatically after bringing
the device down.
Support the delete-while-down case. Skip the FW logic and delete
just the driver state.
Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Link: https://patch.msgid.link/20240711220713.283778-4-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Some drivers (bnxt but I think also mlx5 from ML discussions) change
the size of the indirection table depending on the number of Rx rings.
Decouple the max table size from the size of the currently used table,
so that we can reserve space in the context for table growth.
Static members in ethtool_ops are good enough for now, we can add
callbacks to read the max size more dynamically if someone needs
that.
Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Link: https://patch.msgid.link/20240711220713.283778-3-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
RSS contexts may get lost from a device, in various extreme circumstances.
Specifically if the firmware leaks resources and resets, or crashes and
either recovers in partially working state or the crash causes a
different FW version to run - creating the context again may fail.
Drivers should do their absolute best to prevent this from happening.
When it does, however, telling user that a context exists, when it can't
possibly be used any more is counter productive. Add a helper for
drivers to discard contexts. Print an error, in the future netlink
notification will also be sent.
More robust approaches were proposed, like keeping the contexts
but marking them as "dead" (but possibly resurrected by next reset).
That may be better but it's unclear at this stage whether the
effort is worth the benefits.
Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Link: https://patch.msgid.link/20240711220713.283778-2-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Pull more networking fixes from Jakub Kicinski:
"A quick follow up to yesterday's pull. We got a regressions report for
the bnxt patch as soon as it got to your tree. The ethtool fix is also
good to have, although it's an older regression.
Current release - regressions:
- eth: bnxt_en: fix crash in bnxt_get_max_rss_ctx_ring() on older HW
when user tries to decrease the ring count
Previous releases - regressions:
- ethtool: fix RSS setting, accept "no change" setting if the driver
doesn't support the new features
- eth: i40e: remove needless retries of NVM update, don't wait 20min
when we know the firmware update won't succeed"
* tag 'net-6.10-rc8-2' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net:
bnxt_en: Fix crash in bnxt_get_max_rss_ctx_ring()
octeontx2-af: fix issue with IPv4 match for RSS
octeontx2-af: fix issue with IPv6 ext match for RSS
octeontx2-af: fix detection of IP layer
octeontx2-af: fix a issue with cpt_lf_alloc mailbox
octeontx2-af: replace cpt slot with lf id on reg write
i40e: fix: remove needless retries of NVM update
net: ethtool: Fix RSS setting
This test verifies that resolve_prog_type() works as expected when
`attach_prog_fd` is not passed in.
`prog->aux->dst_prog` in resolve_prog_type() is assigned by
`attach_prog_fd`, and would be NULL if `attach_prog_fd` is not provided.
Loading EXT prog with bpf_dynptr_from_skb() kfunc call in this way will
lead to null-pointer-deref.
Verify that the null-pointer-deref bug in resolve_prog_type() is fixed.
Signed-off-by: Tengda Wu <wutengda@huaweicloud.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20240711145819.254178-3-wutengda@huaweicloud.com
When loading a EXT program without specifying `attr->attach_prog_fd`,
the `prog->aux->dst_prog` will be null. At this time, calling
resolve_prog_type() anywhere will result in a null pointer dereference.
Example stack trace:
[ 8.107863] Unable to handle kernel NULL pointer dereference at virtual address 0000000000000004
[ 8.108262] Mem abort info:
[ 8.108384] ESR = 0x0000000096000004
[ 8.108547] EC = 0x25: DABT (current EL), IL = 32 bits
[ 8.108722] SET = 0, FnV = 0
[ 8.108827] EA = 0, S1PTW = 0
[ 8.108939] FSC = 0x04: level 0 translation fault
[ 8.109102] Data abort info:
[ 8.109203] ISV = 0, ISS = 0x00000004, ISS2 = 0x00000000
[ 8.109399] CM = 0, WnR = 0, TnD = 0, TagAccess = 0
[ 8.109614] GCS = 0, Overlay = 0, DirtyBit = 0, Xs = 0
[ 8.109836] user pgtable: 4k pages, 48-bit VAs, pgdp=0000000101354000
[ 8.110011] [0000000000000004] pgd=0000000000000000, p4d=0000000000000000
[ 8.112624] Internal error: Oops: 0000000096000004 [#1] PREEMPT SMP
[ 8.112783] Modules linked in:
[ 8.113120] CPU: 0 PID: 99 Comm: may_access_dire Not tainted 6.10.0-rc3-next-20240613-dirty #1
[ 8.113230] Hardware name: linux,dummy-virt (DT)
[ 8.113390] pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[ 8.113429] pc : may_access_direct_pkt_data+0x24/0xa0
[ 8.113746] lr : add_subprog_and_kfunc+0x634/0x8e8
[ 8.113798] sp : ffff80008283b9f0
[ 8.113813] x29: ffff80008283b9f0 x28: ffff800082795048 x27: 0000000000000001
[ 8.113881] x26: ffff0000c0bb2600 x25: 0000000000000000 x24: 0000000000000000
[ 8.113897] x23: ffff0000c1134000 x22: 000000000001864f x21: ffff0000c1138000
[ 8.113912] x20: 0000000000000001 x19: ffff0000c12b8000 x18: ffffffffffffffff
[ 8.113929] x17: 0000000000000000 x16: 0000000000000000 x15: 0720072007200720
[ 8.113944] x14: 0720072007200720 x13: 0720072007200720 x12: 0720072007200720
[ 8.113958] x11: 0720072007200720 x10: 0000000000f9fca4 x9 : ffff80008021f4e4
[ 8.113991] x8 : 0101010101010101 x7 : 746f72705f6d656d x6 : 000000001e0e0f5f
[ 8.114006] x5 : 000000000001864f x4 : ffff0000c12b8000 x3 : 000000000000001c
[ 8.114020] x2 : 0000000000000002 x1 : 0000000000000000 x0 : 0000000000000000
[ 8.114126] Call trace:
[ 8.114159] may_access_direct_pkt_data+0x24/0xa0
[ 8.114202] bpf_check+0x3bc/0x28c0
[ 8.114214] bpf_prog_load+0x658/0xa58
[ 8.114227] __sys_bpf+0xc50/0x2250
[ 8.114240] __arm64_sys_bpf+0x28/0x40
[ 8.114254] invoke_syscall.constprop.0+0x54/0xf0
[ 8.114273] do_el0_svc+0x4c/0xd8
[ 8.114289] el0_svc+0x3c/0x140
[ 8.114305] el0t_64_sync_handler+0x134/0x150
[ 8.114331] el0t_64_sync+0x168/0x170
[ 8.114477] Code: 7100707f 54000081 f9401c00 f9403800 (b9400403)
[ 8.118672] ---[ end trace 0000000000000000 ]---
One way to fix it is by forcing `attach_prog_fd` non-empty when
bpf_prog_load(). But this will lead to `libbpf_probe_bpf_prog_type`
API broken which use verifier log to probe prog type and will log
nothing if we reject invalid EXT prog before bpf_check().
Another way is by adding null check in resolve_prog_type().
The issue was introduced by commit 4a9c7bbe2e ("bpf: Resolve to
prog->aux->dst_prog->type only for BPF_PROG_TYPE_EXT") which wanted
to correct type resolution for BPF_PROG_TYPE_TRACING programs. Before
that, the type resolution of BPF_PROG_TYPE_EXT prog actually follows
the logic below:
prog->aux->dst_prog ? prog->aux->dst_prog->type : prog->type;
It implies that when EXT program is not yet attached to `dst_prog`,
the prog type should be EXT itself. This code worked fine in the past.
So just keep using it.
Fix this by returning `prog->type` for BPF_PROG_TYPE_EXT if `dst_prog`
is not present in resolve_prog_type().
Fixes: 4a9c7bbe2e ("bpf: Resolve to prog->aux->dst_prog->type only for BPF_PROG_TYPE_EXT")
Signed-off-by: Tengda Wu <wutengda@huaweicloud.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/20240711145819.254178-2-wutengda@huaweicloud.com
Pull btrfs fixes from David Sterba:
"Fix a regression in extent map shrinker behaviour.
In the past weeks we got reports from users that there are huge
latency spikes or freezes. This was bisected to newly added shrinker
of extent maps (it was added to fix a build up of the structures in
memory).
I'm assuming that the freezes would happen to many users after release
so I'd like to get it merged now so it's in 6.10. Although the diff
size is not small the changes are relatively straightforward, the
reporters verified the fixes and we did testing on our side.
The fixes:
- adjust behaviour under memory pressure and check lock or scheduling
conditions, bail out if needed
- synchronize tracking of the scanning progress so inode ranges are
not skipped or work duplicated
- do a delayed iput when scanning a root so evicting an inode does
not slow things down in case of lots of dirty data, also fix
lockdep warning, a deadlock could happen when writing the dirty
data would need to start a transaction"
* tag 'for-6.10-rc7-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
btrfs: avoid races when tracking progress for extent map shrinking
btrfs: stop extent map shrinker if reschedule is needed
btrfs: use delayed iput during extent map shrinking
Pull ceph fixes from Ilya Dryomov:
"A fix for a possible use-after-free following "rbd unmap" or "umount"
marked for stable and two kernel-doc fixups"
* tag 'ceph-for-6.10-rc8' of https://github.com/ceph/ceph-client:
libceph: fix crush_choose_firstn() kernel-doc warnings
libceph: suppress crush_choose_indep() kernel-doc warnings
libceph: fix race between delayed_work() and ceph_monc_stop()
Pull pmdomain fix from Ulf Hansson:
- qcom: Skip retention level for rpmhpd's
* tag 'pmdomain-v6.10-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/linux-pm:
pmdomain: qcom: rpmhpd: Skip retention level for Power Domains
Pull ARM SoC fixes from Arnd Bergmann:
"Most of these changes are Qualcomm SoC specific and came in just after
I sent out the last set of fixes. This includes two regression fixes
for SoC drivers, a defconfig change to ensure the Lenovo X13s is
usable and 11 changes to DT files to fix regressions and minor
platform specific issues.
Tony and Chunyan step back from their respective maintainership roles
on the omap and unisoc platforms, and Christophe in turn takes over
maintaining some of the Freescale SoC drivers that he has been taking
care of in practice already.
Lastly, there are two trivial fixes for the davinci and sunxi
platforms"
* tag 'arm-fixes-6.10-3' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc:
MAINTAINERS: Update FREESCALE SOC DRIVERS and QUICC ENGINE LIBRARY
MAINTAINERS: Add more maintainers for omaps
ARM: davinci: Convert comma to semicolon
MAINTAINERS: Move myself from SPRD Maintainer to Reviewer
Revert "dt-bindings: cache: qcom,llcc: correct QDU1000 reg entries"
arm64: dts: qcom: qdu1000: Fix LLCC reg property
arm64: dts: qcom: sm6115: add iommu for sdhc_1
arm64: dts: qcom: x1e80100-crd: fix DAI used for headset recording
arm64: dts: qcom: x1e80100-crd: fix WCD audio codec TX port mapping
soc: qcom: pmic_glink: disable UCSI on sc8280xp
arm64: defconfig: enable Elan i2c-hid driver
arm64: dts: qcom: sc8280xp-crd: use external pull up for touch reset
arm64: dts: qcom: sc8280xp-x13s: fix touchscreen power on
arm64: dts: qcom: x1e80100: Fix PCIe 6a reg offsets and add MHI
arm64: dts: qcom: sa8775p: Correct IRQ number of EL2 non-secure physical timer
arm64: dts: allwinner: Fix PMIC interrupt number
arm64: dts: qcom: sc8280xp: Set status = "reserved" on PSHOLD
arm64: dts: qcom: x1e80100-*: Allocate some CMA buffers
arm64: dts: qcom: sc8180x: Fix LLCC reg property again
