The driver supports many hw statistics. This patch supports
dump statistics through ethtool_ops and ndo.get_stats64().
The type of hw statistics register is u32,
To prevent the statistics register from overflowing,
the driver dump the statistics every 30 seconds.
in a scheduled task.
Signed-off-by: Jijie Shao <shaojijie@huawei.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Similar to PPE support for Mediatek devices, introduce PPE debugfs
in order to dump binded and unbinded flows.
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Enable hw redirection for traffic received on GDM2 port to GDM{3,4}.
This is required to apply Qdisc offloading (HTB or ETS) for traffic to
and from GDM{3,4} port.
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Introduce netfilter flowtable integration in order to allow airoha_eth
driver to offload 5-tuple flower rules learned by the PPE module if the
user accelerates them using a nft configuration similar to the one reported
below:
table inet filter {
flowtable ft {
hook ingress priority filter
devices = { lan1, lan2, lan3, lan4, eth1 }
flags offload;
}
chain forward {
type filter hook forward priority filter; policy accept;
meta l4proto { tcp, udp } flow add @ft
}
}
Tested-by: Sayantan Nandy <sayantan.nandy@airoha.com>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Packet Processor Engine (PPE) module available on EN7581 SoC populates
the PPE table with 5-tuples flower rules learned from traffic forwarded
between the GDM ports connected to the Packet Switch Engine (PSE) module.
The airoha_eth driver can enable hw acceleration of learned 5-tuples
rules if the user configure them in netfilter flowtable (netfilter
flowtable support will be added with subsequent patches).
airoha_eth driver configures and collects data from the PPE module via a
Network Processor Unit (NPU) RISC-V module available on the EN7581 SoC.
Introduce basic support for Airoha NPU module.
Tested-by: Sayantan Nandy <sayantan.nandy@airoha.com>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Introduce the airoha,npu property for the NPU node available on
EN7581 SoC. The airoha Network Processor Unit (NPU) is used to
offload network traffic forwarded between Packet Switch Engine
(PSE) ports.
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
This patch adds the NPU document binding for EN7581 SoC.
The Airoha Network Processor Unit (NPU) provides a configuration interface
to implement wired and wireless hardware flow offloading programming Packet
Processor Engine (PPE) flow table.
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Rename airoha_set_gdm_port() in airoha_set_vip_for_gdm_port().
Get rid of airoha_set_gdm_ports routine.
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Move REG_GDM_FWD_CFG() register initialization in airoha_dev_init
routine. Moreover, always send traffic PPE module in order to be
processed by hw accelerator.
This is a preliminary patch to enable netfilter flowtable hw offloading
on EN7581 SoC.
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
In the current codebase airoha_eth driver supports just a single
net_device connected to the Packet Switch Engine (PSE) lan port (GDM1).
As shown in commit 23020f0493 ("net: airoha: Introduce ethernet
support for EN7581 SoC"), PSE can switch packets between four GDM ports.
Enable the capability to create a net_device for each GDM port of the
PSE module. Moreover, since the QDMA blocks can be shared between
net_devices, do not stop TX/RX DMA in airoha_dev_stop() if there are
active net_devices for this QDMA block.
This is a preliminary patch to enable flowtable hw offloading for EN7581
SoC.
Co-developed-by: Christian Marangi <ansuelsmth@gmail.com>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Packet Processor Engine (PPE) module used for hw acceleration on EN7581
mac block, in order to properly parse packets, requires DSA untagged
packets on TX side and read DSA tag from DMA descriptor on RX side.
For this reason, enable RX Special Tag (SPTAG) for EN7581 SoC.
This is a preliminary patch to enable netfilter flowtable hw offloading
on EN7581 SoC.
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Packet Processor Engine (PPE) module reads DSA tags from the DMA descriptor
and requires untagged DSA packets to properly parse them. Move DSA tag
in the DMA descriptor on TX side and read DSA tag from DMA descriptor
on RX side. In order to avoid skb reallocation, store tag in skb_dst on
RX side.
This is a preliminary patch to enable netfilter flowtable hw offloading
on EN7581 SoC.
Tested-by: Sayantan Nandy <sayantan.nandy@airoha.com>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Move common airoha_eth register definitions in airoha_regs.h in order
to reuse them for Packet Processor Engine (PPE) codebase.
PPE module is used to enable support for flowtable hw offloading in
airoha_eth driver.
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
This is a preliminary patch to introduce flowtable hw offloading
support for airoha_eth driver.
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Move common airoha_eth definitions in airoha_eth.h in order to reuse
them for Packet Processor Engine (PPE) codebase.
PPE module is used to enable support for flowtable hw offloading in
airoha_eth driver.
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The airoha_eth driver has no codebase shared with mtk_eth_soc one.
Moreover, the upcoming features (flowtable hw offloading, PCS, ..) will
not reuse any code from MediaTek driver. Move the Airoha driver in a
dedicated folder.
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
It could be hard to understand why the netlink command fails. For example,
if dev->netns_immutable is set, the error is "Invalid argument".
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Since commit 05c1280a2b ("netdev_features: convert NETIF_F_NETNS_LOCAL to
dev->netns_local"), there is no way to see if the netns_immutable property
s set on a device. Let's add a netlink attribute to advertise it.
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Peter Seiderer says:
====================
Some pktgen fixes/improvments (part II)
While taking a look at '[PATCH net] pktgen: Avoid out-of-range in
get_imix_entries' ([1]) and '[PATCH net v2] pktgen: Avoid out-of-bounds
access in get_imix_entries' ([2], [3]) and doing some tests and code review
I detected that the /proc/net/pktgen/... parsing logic does not honour the
user given buffer bounds (resulting in out-of-bounds access).
This can be observed e.g. by the following simple test (sometimes the
old/'longer' previous value is re-read from the buffer):
$ echo add_device lo@0 > /proc/net/pktgen/kpktgend_0
$ echo "min_pkt_size 12345" > /proc/net/pktgen/lo\@0 && grep min_pkt_size /proc/net/pktgen/lo\@0
Params: count 1000 min_pkt_size: 12345 max_pkt_size: 0
Result: OK: min_pkt_size=12345
$ echo -n "min_pkt_size 123" > /proc/net/pktgen/lo\@0 && grep min_pkt_size /proc/net/pktgen/lo\@0
Params: count 1000 min_pkt_size: 12345 max_pkt_size: 0
Result: OK: min_pkt_size=12345
$ echo "min_pkt_size 123" > /proc/net/pktgen/lo\@0 && grep min_pkt_size /proc/net/pktgen/lo\@0
Params: count 1000 min_pkt_size: 123 max_pkt_size: 0
Result: OK: min_pkt_size=123
So fix the out-of-bounds access (and some minor findings) and add a simple
proc_net_pktgen selftest...
Patch set splited into part I (now already applied to net-next)
- net: pktgen: replace ENOTSUPP with EOPNOTSUPP
- net: pktgen: enable 'param=value' parsing
- net: pktgen: fix hex32_arg parsing for short reads
- net: pktgen: fix 'rate 0' error handling (return -EINVAL)
- net: pktgen: fix 'ratep 0' error handling (return -EINVAL)
- net: pktgen: fix ctrl interface command parsing
- net: pktgen: fix access outside of user given buffer in pktgen_thread_write()
nd part II (this one):
- net: pktgen: use defines for the various dec/hex number parsing digits lengths
- net: pktgen: fix mix of int/long
- net: pktgen: remove extra tmp variable (re-use len instead)
- net: pktgen: remove some superfluous variable initializing
- net: pktgen: fix mpls maximum labels list parsing
- net: pktgen: fix access outside of user given buffer in pktgen_if_write()
- net: pktgen: fix mpls reset parsing
- net: pktgen: remove all superfluous index assignements
- selftest: net: add proc_net_pktgen
[1] https://lore.kernel.org/netdev/20241006221221.3744995-1-artem.chernyshev@red-soft.ru/
[2] https://lore.kernel.org/netdev/20250109083039.14004-1-pchelkin@ispras.ru/
[3] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=76201b5979768500bca362871db66d77cb4c225e
====================
Link: https://patch.msgid.link/20250227135604.40024-1-ps.report@gmx.net
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Add some test for /proc/net/pktgen/... interface.
- enable 'CONFIG_NET_PKTGEN=m' in tools/testing/selftests/net/config
Signed-off-by: Peter Seiderer <ps.report@gmx.net>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Remove all superfluous index ('i += len') assignements (value not used
afterwards).
Signed-off-by: Peter Seiderer <ps.report@gmx.net>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Honour the user given buffer size for the hex32_arg(), num_arg(),
strn_len(), get_imix_entries() and get_labels() calls (otherwise they will
access memory outside of the user given buffer).
Signed-off-by: Peter Seiderer <ps.report@gmx.net>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Fix mpls maximum labels list parsing up to MAX_MPLS_LABELS entries (instead
of up to MAX_MPLS_LABELS - 1).
Addresses the following:
$ echo "mpls 00000f00,00000f01,00000f02,00000f03,00000f04,00000f05,00000f06,00000f07,00000f08,00000f09,00000f0a,00000f0b,00000f0c,00000f0d,00000f0e,00000f0f" > /proc/net/pktgen/lo\@0
-bash: echo: write error: Argument list too long
Signed-off-by: Peter Seiderer <ps.report@gmx.net>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Remove some superfluous variable initializing before hex32_arg call (as the
same init is done here already).
Signed-off-by: Peter Seiderer <ps.report@gmx.net>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Remove extra tmp variable in pktgen_if_write (re-use len instead).
Signed-off-by: Peter Seiderer <ps.report@gmx.net>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Fix mix of int/long (and multiple conversion from/to) by using consequently
size_t for i and max and ssize_t for len and adjust function signatures
of hex32_arg(), count_trail_chars(), num_arg() and strn_len() accordingly.
Signed-off-by: Peter Seiderer <ps.report@gmx.net>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Geliang Tang says:
====================
add sock_kmemdup helper
While developing MPTCP BPF path manager [1], I found it's useful to
add a new sock_kmemdup() helper.
My use case is this:
In mptcp_userspace_pm_append_new_local_addr() function (see patch 3
in this patchset), it uses sock_kmalloc() to allocate an address
entry "e", then immediately duplicate the input "entry" to it:
'''
e = sock_kmalloc(sk, sizeof(*e), GFP_ATOMIC);
if (!e) {
ret = -ENOMEM;
goto append_err;
}
*e = *entry;
'''
When I implemented MPTCP BPF path manager, I needed to implement a
code similar to this in BPF.
The kfunc sock_kmalloc() can be easily invoked in BPF to allocate
an entry "e", but the code "*e = *entry;" that assigns "entry" to
"e" is not easy to implemented.
I had to implement such a "copy entry" helper in BPF:
'''
static void mptcp_pm_copy_addr(struct mptcp_addr_info *dst,
struct mptcp_addr_info *src)
{
dst->id = src->id;
dst->family = src->family;
dst->port = src->port;
if (src->family == AF_INET) {
dst->addr.s_addr = src->addr.s_addr;
} else if (src->family == AF_INET6) {
dst->addr6.s6_addr32[0] = src->addr6.s6_addr32[0];
dst->addr6.s6_addr32[1] = src->addr6.s6_addr32[1];
dst->addr6.s6_addr32[2] = src->addr6.s6_addr32[2];
dst->addr6.s6_addr32[3] = src->addr6.s6_addr32[3];
}
}
static void mptcp_pm_copy_entry(struct mptcp_pm_addr_entry *dst,
struct mptcp_pm_addr_entry *src)
{
mptcp_pm_copy_addr(&dst->addr, &src->addr);
dst->flags = src->flags;
dst->ifindex = src->ifindex;
}
'''
And add "write permission" for BPF to each field of mptcp_pm_addr_entry:
'''
@@ static int bpf_mptcp_pm_btf_struct_access(struct bpf_verifier_log *log,
case offsetof(struct mptcp_pm_addr_entry, addr.port):
end = offsetofend(struct mptcp_pm_addr_entry, addr.port);
break;
#if IS_ENABLED(CONFIG_MPTCP_IPV6)
case offsetof(struct mptcp_pm_addr_entry, addr.addr6.s6_addr32[0]):
end = offsetofend(struct mptcp_pm_addr_entry, addr.addr6.s6_addr32[0]);
break;
case offsetof(struct mptcp_pm_addr_entry, addr.addr6.s6_addr32[1]):
end = offsetofend(struct mptcp_pm_addr_entry, addr.addr6.s6_addr32[1]);
break;
case offsetof(struct mptcp_pm_addr_entry, addr.addr6.s6_addr32[2]):
end = offsetofend(struct mptcp_pm_addr_entry, addr.addr6.s6_addr32[2]);
break;
case offsetof(struct mptcp_pm_addr_entry, addr.addr6.s6_addr32[3]):
end = offsetofend(struct mptcp_pm_addr_entry, addr.addr6.s6_addr32[3]);
break;
#else
case offsetof(struct mptcp_pm_addr_entry, addr.addr.s_addr):
end = offsetofend(struct mptcp_pm_addr_entry, addr.addr.s_addr);
break;
#endif
'''
But if there's a sock_kmemdup() helper, it will become much simpler,
only need to call kfunc sock_kmemdup() instead in BPF.
So this patchset adds this new helper and uses it in several places.
[1]
https://lore.kernel.org/mptcp/cover.1738924875.git.tanggeliang@kylinos.cn/
====================
Link: https://patch.msgid.link/cover.1740735165.git.tanggeliang@kylinos.cn
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Instead of using sock_kmalloc() to allocate an address
entry "e" and then immediately duplicate the input "entry"
to it, the newly added sock_kmemdup() helper can be used in
mptcp_userspace_pm_append_new_local_addr() to simplify the code.
More importantly, the code "*e = *entry;" that assigns "entry"
to "e" is not easy to implemented in BPF if we use the same code
to implement an append_new_local_addr() helper of a BFP path
manager. This patch avoids this type of memory assignment
operation.
Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
Acked-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/3e5a307aed213038a87e44ff93b5793229b16279.1740735165.git.tanggeliang@kylinos.cn
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
fib_valid_key_len() is called in the beginning of fib_table_insert()
or fib_table_delete() to check if the prefix length is valid.
fib_table_insert() and fib_table_delete() are called from 3 paths
- ip_rt_ioctl()
- inet_rtm_newroute() / inet_rtm_delroute()
- fib_magic()
In the first ioctl() path, rtentry_to_fib_config() checks the prefix
length with bad_mask(). Also, fib_magic() always passes the correct
prefix: 32 or ifa->ifa_prefixlen, which is already validated.
Let's move fib_valid_key_len() to the rtnetlink path, rtm_to_fib_config().
While at it, 2 direct returns in rtm_to_fib_config() are changed to
goto to match other places in the same function
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20250228042328.96624-12-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
We will convert RTM_NEWROUTE and RTM_DELROUTE to per-netns RTNL.
Then, we need to have per-netns hash tables for struct fib_info.
Let's allocate the hash tables per netns.
fib_info_hash, fib_info_hash_bits, and fib_info_cnt are now moved
to struct netns_ipv4 and accessed with net->ipv4.fib_XXX.
Also, the netns checks are removed from fib_find_info_nh() and
fib_find_info().
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20250228042328.96624-9-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When the number of struct fib_info exceeds the hash table size in
fib_create_info(), we try to allocate a new hash table with the
doubled size.
The allocation is done in fib_create_info(), and if successful, each
struct fib_info is moved to the new hash table by fib_info_hash_move().
Let's integrate the allocation and fib_info_hash_move() as
fib_info_hash_grow() to make the following change cleaner.
While at it, fib_info_hash_grow() is placed near other hash-table-specific
functions.
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20250228042328.96624-8-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
