Sashiko found possible double close of btf object fd [1],
which happens when strdup in load_module_btfs fails at which
point the obj->btf_module_cnt is already incremented.
The error path close btf fd and so does later cleanup code in
bpf_object_post_load_cleanup function.
Also libbpf_ensure_mem failure leaves btf object not assigned
and it's leaked.
Replacing the err_out label with break to make the error path
less confusing as suggested by Alan.
Incrementing obj->btf_module_cnt only if there's no failure
and releasing btf object in error path.
Fixes: 91abb4a6d79d ("libbpf: Support attachment of BPF tracing programs to kernel modules")
[1] https://sashiko.dev/#/patchset/20260324081846.2334094-1-jolsa%40kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Link: https://lore.kernel.org/r/20260416100034.1610852-1-jolsa@kernel.org
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The bridge-stp usermode helper is currently restricted to the initial
network namespace, preventing userspace STP daemons (e.g. mstpd) from
operating on bridges in other network namespaces. Since commit
ff62198553e4 ("bridge: Only call /sbin/bridge-stp for the initial
network namespace"), bridges in non-init namespaces silently fall
back to kernel STP with no way to use userspace STP.
Add a new bridge attribute IFLA_BR_STP_MODE that allows explicit
per-bridge control over STP mode selection:
BR_STP_MODE_AUTO (default) - Existing behavior: invoke the
/sbin/bridge-stp helper in init_net only; fall back to kernel STP
if it fails or in non-init namespaces.
BR_STP_MODE_USER - Directly enable userspace STP (BR_USER_STP)
without invoking the helper. Works in any network namespace.
Userspace is responsible for ensuring an STP daemon manages the
bridge.
BR_STP_MODE_KERNEL - Directly enable kernel STP (BR_KERNEL_STP)
without invoking the helper.
The mode can only be changed while STP is disabled, or set to the
same value (-EBUSY otherwise). IFLA_BR_STP_MODE is processed before
IFLA_BR_STP_STATE in br_changelink(), so both can be set atomically
in a single netlink message. The mode can also be changed in the
same message that disables STP.
The stp_mode struct field is u8 since all possible values fit, while
NLA_U32 is used for the netlink attribute since it occupies the same
space in the netlink message as NLA_U8.
A new stp_helper_active boolean tracks whether the /sbin/bridge-stp
helper was invoked during br_stp_start(), so that br_stp_stop() only
calls the helper for stop when it was called for start. This avoids
calling the helper asymmetrically when stp_mode changes between
start and stop.
Suggested-by: Ido Schimmel <idosch@nvidia.com>
Assisted-by: Claude:claude-opus-4-6
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Acked-by: Nikolay Aleksandrov <nikolay@nvidia.com>
Signed-off-by: Andy Roulin <aroulin@nvidia.com>
Link: https://patch.msgid.link/20260405205224.3163000-2-aroulin@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add a single device mode for netkit instead of netkit pairs. The primary
target for the paired devices is to connect network namespaces, of course,
and support has been implemented in projects like Cilium [0]. For the rxq
leasing the plan is to support two main scenarios related to single device
mode:
* For the use-case of io_uring zero-copy, the control plane can either
set up a netkit pair where the peer device can perform rxq leasing which
is then tied to the lifetime of the peer device, or the control plane
can use a regular netkit pair to connect the hostns to a Pod/container
and dynamically add/remove rxq leasing through a single device without
having to interrupt the device pair. In the case of io_uring, the memory
pool is used as skb non-linear pages, and thus the skb will go its way
through the regular stack into netkit. Things like the netkit policy when
no BPF is attached or skb scrubbing etc apply as-is in case the paired
devices are used, or if the backend memory is tied to the single device
and traffic goes through a paired device.
* For the use-case of AF_XDP, the control plane needs to use netkit in the
single device mode. The single device mode currently enforces only a
pass policy when no BPF is attached, and does not yet support BPF link
attachments for AF_XDP. skbs sent to that device get dropped at the
moment. Given AF_XDP operates at a lower layer of the stack tying this
to the netkit pair did not make sense. In future, the plan is to allow
BPF at the XDP layer which can: i) process traffic coming from the AF_XDP
application (e.g. QEMU with AF_XDP backend) to filter egress traffic or
to push selected egress traffic up to the single netkit device to the
local stack (e.g. DHCP requests), and ii) vice-versa skbs sent to the
single netkit into the AF_XDP application (e.g. DHCP replies). Also,
the control-plane can dynamically manage rxq leasing for the single
netkit device without having to interrupt (e.g. down/up cycle) the main
netkit pair for the Pod which has traffic going in and out.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Co-developed-by: David Wei <dw@davidwei.uk>
Signed-off-by: David Wei <dw@davidwei.uk>
Reviewed-by: Jordan Rife <jordan@jrife.io>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Link: https://docs.cilium.io/en/stable/operations/performance/tuning/#netkit-device-mode [0]
Link: https://patch.msgid.link/20260402231031.447597-11-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add a ynl netdev family operation called queue-create that creates a
new queue on a netdevice:
name: queue-create
attribute-set: queue
flags: [admin-perm]
do:
request:
attributes:
- ifindex
- type
- lease
reply: &queue-create-op
attributes:
- id
This is a generic operation such that it can be extended for various
use cases in future. Right now it is mandatory to specify ifindex,
the queue type which is enforced to rx and a lease. The newly created
queue id is returned to the caller.
A queue from a virtual device can have a lease which refers to another
queue from a physical device. This is useful for memory providers
and AF_XDP operations which take an ifindex and queue id to allow
applications to bind against virtual devices in containers. The lease
couples both queues together and allows to proxy the operations from
a virtual device in a container to the physical device.
In future, the nested lease attribute can be lifted and made optional
for other use-cases such as dynamic queue creation for physical
netdevs. The lack of lease and the specification of the physical
device as an ifindex will imply that we need a real queue to be
allocated. Similarly, the queue type enforcement to rx can then be
lifted as well to support tx.
An early implementation had only driver-specific integration [0], but
in order for other virtual devices to reuse, it makes sense to have
this as a generic API in core net.
For leasing queues, the virtual netdev must have real_num_rx_queues
less than num_rx_queues at the time of calling queue-create. The
queue-type must be rx as only rx queues are supported for leasing
for now. We also enforce that the queue-create ifindex must point
to a virtual device, and that the nested lease attribute's ifindex
must point to a physical device. The nested lease attribute set
contains a netns-id attribute which is optional and can specify a
netns-id relative to the caller's netns. It requires cap_net_admin
and if the netns-id attribute is not specified, the lease ifindex
will be retrieved from the current netns. Also, it is modeled as
an s32 type similarly as done elsewhere in the stack.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Co-developed-by: David Wei <dw@davidwei.uk>
Signed-off-by: David Wei <dw@davidwei.uk>
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Link: https://bpfconf.ebpf.io/bpfconf2025/bpfconf2025_material/lsfmmbpf_2025_netkit_borkmann.pdf [0]
Link: https://patch.msgid.link/20260402231031.447597-2-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Allow bpf object feat_cache assignment in BPF selftests
to simulate missing features via inclusion of libbpf_internal.h
and use of bpf_object_set_feat_cache() and bpf_object__sanitize_btf() to
test BTF sanitization for cases where missing features are simulated.
Signed-off-by: Alan Maguire <alan.maguire@oracle.com>
Link: https://lore.kernel.org/r/20260408165735.843763-2-alan.maguire@oracle.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
CO-RE accessor strings are colon-separated indices that describe a path
from a root BTF type to a target field, e.g. "0:1:2" walks through
nested struct members. bpf_core_parse_spec() parses each component with
sscanf("%d"), so negative values like -1 are silently accepted. The
subsequent bounds checks (access_idx >= btf_vlen(t)) only guard the
upper bound and always pass for negative values because C integer
promotion converts the __u16 btf_vlen result to int, making the
comparison (int)(-1) >= (int)(N) false for any positive N.
When -1 reaches btf_member_bit_offset() it gets cast to u32 0xffffffff,
producing an out-of-bounds read far past the members array. A crafted
BPF program with a negative CO-RE accessor on any struct that exists in
vmlinux BTF (e.g. task_struct) crashes the kernel deterministically
during BPF_PROG_LOAD on any system with CONFIG_DEBUG_INFO_BTF=y
(default on major distributions). The bug is reachable with CAP_BPF:
BUG: unable to handle page fault for address: ffffed11818b6626
#PF: supervisor read access in kernel mode
#PF: error_code(0x0000) - not-present page
Oops: Oops: 0000 [#1] SMP KASAN NOPTI
CPU: 0 UID: 0 PID: 85 Comm: poc Not tainted 7.0.0-rc6 #18 PREEMPT(full)
RIP: 0010:bpf_core_parse_spec (tools/lib/bpf/relo_core.c:354)
RAX: 00000000ffffffff
Call Trace:
<TASK>
bpf_core_calc_relo_insn (tools/lib/bpf/relo_core.c:1321)
bpf_core_apply (kernel/bpf/btf.c:9507)
check_core_relo (kernel/bpf/verifier.c:19475)
bpf_check (kernel/bpf/verifier.c:26031)
bpf_prog_load (kernel/bpf/syscall.c:3089)
__sys_bpf (kernel/bpf/syscall.c:6228)
</TASK>
CO-RE accessor indices are inherently non-negative (struct member index,
array element index, or enumerator index), so reject them immediately
after parsing.
Fixes: ddc7c3042614 ("libbpf: implement BPF CO-RE offset relocation algorithm")
Reported-by: Xiang Mei <xmei5@asu.edu>
Signed-off-by: Weiming Shi <bestswngs@gmail.com>
Reviewed-by: Emil Tsalapatis <emil@etsalapatis.com>
Acked-by: Paul Chaignon <paul.chaignon@gmail.com>
Link: https://lore.kernel.org/r/20260404161221.961828-2-bestswngs@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
bpf_program__attach_kprobe_opts() documents single-kprobe attach
through func_name, with an optional offset. For the PMU-based path,
func_name = NULL with an absolute address in offset already works as
well, but that is not described in the API.
This commit clarifies this existing non-legacy behavior. For PMU-based
attach, callers can use func_name = NULL with an absolute address in
offset as the raw-address form. For legacy tracefs/debugfs kprobes,
reject this form explicitly.
Signed-off-by: Hoyeon Lee <hoyeon.lee@suse.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Link: https://lore.kernel.org/bpf/20260401143116.185049-3-hoyeon.lee@suse.com
Align bpf_program__clone() with bpf_object_load_prog() by gating
BTF func/line info on FEAT_BTF_FUNC kernel support, and resolve
caller-provided prog_btf_fd before checking obj->btf so that callers
with their own BTF can use clone() even when the object has no BTF
loaded.
While at it, treat func_info and line_info fields as atomic groups
to prevent mismatches between pointer and count from different sources.
Move bpf_program__clone() to libbpf 1.8.
Fixes: 970bd2dced35 ("libbpf: Introduce bpf_program__clone()")
Signed-off-by: Mykyta Yatsenko <yatsenko@meta.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20260401151640.356419-1-mykyta.yatsenko5@gmail.com
Clarify bpf_ringbuf_discard() documentation for BPF_RB_NO_WAKEUP.
Discarded ring buffer records are still left in the ring buffer and are
only skipped when user space consumes them. This can matter when
BPF_RB_NO_WAKEUP is used: a later submit relying on adaptive wakeup
might not wake the consumer, because the discarded record still needs to
be consumed first.
Scenario:
epoll_wait(rb_fd); // blocks
rec = bpf_ringbuf_reserve(&rb, ...);
bpf_ringbuf_discard(rec, BPF_RB_NO_WAKEUP);
rec = bpf_ringbuf_reserve(&rb, ...);
bpf_ringbuf_submit(rec, 0); // valid record, but no wakeup
Document this in bpf_ringbuf_discard() to make the interaction between
discarded records, user-space consumption, and adaptive wakeups explicit.
Reported-by: Shmulik Ladkani <shmulik.ladkani@gmail.com>
Signed-off-by: Eyal Birger <eyal.birger@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20260331130612.3762433-1-eyal.birger@gmail.com
----
v2: adapt wording per feedback from Andrii.
Add a FEAT_BTF_LAYOUT feature check which checks if the
kernel supports BTF layout information. Also sanitize
BTF if it contains layout data but the kernel does not
support it. The sanitization requires rewriting raw
BTF data to update the header and eliminate the layout
section (since it lies between the types and strings),
so refactor sanitization to do the raw BTF retrieval
and creation of updated BTF, returning that new BTF
on success.
Signed-off-by: Alan Maguire <alan.maguire@oracle.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20260326145444.2076244-7-alan.maguire@oracle.com
Support encoding of BTF layout data via btf__new_empty_opts().
Current supported opts are base_btf and add_layout.
Layout information is maintained in btf.c in the layouts[] array;
when BTF is created with the add_layout option it represents the
current view of supported BTF kinds.
Signed-off-by: Alan Maguire <alan.maguire@oracle.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20260326145444.2076244-5-alan.maguire@oracle.com
Support reading in layout fixing endian issues on reading;
also support writing layout section to raw BTF object.
There is not yet an API to populate the layout with meaningful
information.
As part of this, we need to consider multiple valid BTF header
sizes; the original or the layout-extended headers.
So to support this, the "struct btf" representation is modified
to contain a "struct btf_header" and we copy the valid
portion from the raw data to it; this means we can always safely
check fields like btf->hdr.layout_len .
Note if parsed-in BTF has extra header information beyond
sizeof(struct btf_header) - if so we make that BTF ineligible
for modification by setting btf->has_hdr_extra .
Ensure that we handle endianness issues for BTF layout section,
though currently only field that needs this (flags) is unused.
Signed-off-by: Alan Maguire <alan.maguire@oracle.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20260326145444.2076244-3-alan.maguire@oracle.com
Add bpf_program__clone() API that loads a single BPF program from a
prepared BPF object into the kernel, returning a file descriptor owned
by the caller.
After bpf_object__prepare(), callers can use bpf_program__clone() to
load individual programs with custom bpf_prog_load_opts, instead of
loading all programs at once via bpf_object__load(). Non-zero fields in
opts override the defaults derived from the program and object
internals; passing NULL opts populates everything automatically.
Signed-off-by: Mykyta Yatsenko <yatsenko@meta.com>
Reviewed-by: Emil Tsalapatis <emil@etsalapatis.com>
Link: https://lore.kernel.org/r/20260317-veristat_prepare-v4-1-74193d4cc9d9@meta.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Update .mailmap based on libbpf's list of contributors and on the latest
.mailmap version in the upstream repository.
Signed-off-by: Ihor Solodrai <ihor.solodrai@linux.dev>
Detect exact function names (no wildcards) in
bpf_program__attach_kprobe_multi_opts() and bypass kallsyms parsing,
passing the symbol directly to the kernel via syms[] array. This
benefits all callers, not just kprobe.session.
When the pattern contains no '*' or '?' characters, set syms to point
directly at the pattern string and cnt to 1, skipping the expensive
/proc/kallsyms or available_filter_functions parsing (~150ms per
function).
Error code normalization: the fast path returns ESRCH from kernel's
ftrace_lookup_symbols(), while the slow path returns ENOENT from
userspace kallsyms parsing. Convert ESRCH to ENOENT in the
bpf_link_create error path to maintain API consistency - both paths
now return identical error codes for "symbol not found".
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@crowdstrike.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20260302200837.317907-2-andrey.grodzovsky@crowdstrike.com
btf__add_btf() currently rejects split BTF sources with -ENOTSUP.
This prevents merging types from multiple kernel module BTFs that
are all split against the same vmlinux base.
Extend btf__add_btf() to handle split BTF sources by:
- Replacing the blanket -ENOTSUP with a validation that src and dst
share the same base BTF pointer when both are split, returning
-EOPNOTSUPP on mismatch.
- Computing src_start_id from the source's base to distinguish base
type ID references (which must remain unchanged) from split type
IDs (which must be remapped to new positions in the destination).
- Using src_btf->nr_types instead of btf__type_cnt()-1 for the type
count, which is correct for both split and non-split sources.
- Skipping base string offsets (< start_str_off) during the string
rewrite loop, mirroring the type ID skip pattern. Since src and
dst share the same base BTF, base string offsets are already valid
and need no remapping.
For non-split sources the behavior is identical: src_start_id is 1,
the type_id < 1 guard is never true (VOID is already skipped), and
the remapping formula reduces to the original. start_str_off is 0
so no string offsets are skipped.
Assisted-by: Claude:claude-opus-4-6
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Reviewed-by: Alan Maguire <alan.maguire@oracle.com>
Link: https://lore.kernel.org/bpf/c00216ed48cf7897078d9645679059d5ebf42738.1772657690.git.josef@toxicpanda.com
Adding support to detect nop,nop5 instructions combo for usdt probe
by checking on probe's following nop5 instruction.
When the nop,nop5 combo is detected together with uprobe syscall,
we can place the probe on top of nop5 and get it optimized.
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Link: https://lore.kernel.org/r/20260224103915.1369690-3-jolsa@kernel.org
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Netlink requires that the recv buffer used during dumps is at least
min(PAGE_SIZE, 8k) (see the man page). Otherwise the messages will
get truncated. Make sure bpftool follows this requirement, avoid
missing information on systems with large pages.
Acked-by: Quentin Monnet <qmo@kernel.org>
Fixes: 7084566a236f ("tools/bpftool: Remove libbpf_internal.h usage in bpftool")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Link: https://lore.kernel.org/r/20260217194150.734701-1-kuba@kernel.org
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
github.head_ref is empty for push and schedule events, so all those
runs shared the same concurrency group and cancelled each other. Fall
back to github.run_id to give each non-PR run a unique group.
Assisted-by: Claude:claude-opus-4-6
Signed-off-by: Ihor Solodrai <ihor.solodrai@linux.dev>
The libbpf CI always builds the kernel from source at CHECKPOINT-COMMIT;
there is no prebuilt-kernel matrix entry. Remove the `kernel` input and
the conditional build-vs-download logic, simplifying the workflow.
The run-vmtest action defaults KERNEL to "LATEST" internally when the
env var is unset, so DENYLIST-LATEST is still picked up. The vmlinuz
path is auto-discovered via `make -s image_name` when not passed.
Assisted-by: Claude:claude-opus-4-6
Signed-off-by: Ihor Solodrai <ihor.solodrai@linux.dev>
Merge DENYLIST-LATEST into DENYLIST and remove the per-kernel denylist
files. With LATEST being the only kernel mode, there's no need for
separate files. Also remove the s390x denylist (libbpf CI only tests
x86_64) and drop stale entries fixed upstream.
Add map_kptr, test_profiler (kprobes not available in VM kernel), and
sockmap udp multi channels (flaky) based on CI run results.
Assisted-by: Claude:claude-opus-4-6
Signed-off-by: Ihor Solodrai <ihor.solodrai@linux.dev>
All 4 patches either fail to apply (context mismatch) or are already
applied upstream. They produce noise in CI logs.
Assisted-by: Claude:claude-opus-4-6
Signed-off-by: Ihor Solodrai <ihor.solodrai@linux.dev>
Commit 728ff167910e ("libbpf: Add gating for arena globals relocation feature")
adds a feature gate check that loads a map and BPF program to
test the running kernel supports large direct offsets for LDIMM64
instructions. This check is currently used to calculate arena symbol
offsets during bpf_object__collect_relos, itself called by
bpf_object_open.
However, the program calling bpf_object_open may not have the permissions to
load maps and programs. This is the case with the BPF selftests, where
bpftool is invoked at compilation time during skeleton generation. This
causes errors as the feature gate unexpectedly fails with -EPERM.
Avoid this by moving all the use of the FEAT_LDIMM64_FULL_RANGE_OFF feature gate
to BPF object preparation time instead.
Fixes: 728ff167910e ("libbpf: Add gating for arena globals relocation feature")
Signed-off-by: Emil Tsalapatis <emil@etsalapatis.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20260217204345.548648-3-emil@etsalapatis.com
Commit 728ff167910e uses a PROG_TYPE_TRACEPOINT BPF test program to
check whether the running kernel supports large LDIMM64 offsets. The
feature gate incorrectly assumes that the program will fail at
verification time with one of two messages, depending on whether the
feature is supported by the running kernel. However,
PROG_TYPE_TRACEPOINT programs may fail to load before verification even
starts, e.g., if the shell does not have the appropriate capabilities.
Use a BPF_PROG_TYPE_SOCKET_FILTER program for the feature gate instead.
Also fix two minor issues. First, ensure the log buffer for the test is
initialized: Failing program load before verification led to libbpf dumping
uninitialized data to stdout. Also, ensure that close() is only called
for program_fd in the probe if the program load actually succeeded. The
call was currently failing silently with -EBADF most of the time.
Fixes: 728ff167910e ("libbpf: Add gating for arena globals relocation feature")
Reported-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Emil Tsalapatis <emil@etsalapatis.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20260217204345.548648-2-emil@etsalapatis.com
This reverts commit 77b9c4a438fc66e2ab004c411056b3fb71a54f2c, reversing
changes made to 4515ec4ad58a37e70a9e1256c0b993958c9b7497:
931420a2fc36 ("selftests/net: Add netkit container tests")
ab771c938d9a ("selftests/net: Make NetDrvContEnv support queue leasing")
6be87fbb2776 ("selftests/net: Add env for container based tests")
61d99ce3dfc2 ("selftests/net: Add bpf skb forwarding program")
920da3634194 ("netkit: Add xsk support for af_xdp applications")
eef51113f8af ("netkit: Add netkit notifier to check for unregistering devices")
b5ef109d22d4 ("netkit: Implement rtnl_link_ops->alloc and ndo_queue_create")
b5c3fa4a0b16 ("netkit: Add single device mode for netkit")
0073d2fd679d ("xsk: Proxy pool management for leased queues")
1ecea95dd3b5 ("xsk: Extend xsk_rcv_check validation")
804bf334d08a ("net: Proxy netdev_queue_get_dma_dev for leased queues")
0caa9a8ddec3 ("net: Proxy net_mp_{open,close}_rxq for leased queues")
ff8889ff9107 ("net, ethtool: Disallow leased real rxqs to be resized")
9e2103f36110 ("net: Add lease info to queue-get response")
31127deddef4 ("net: Implement netdev_nl_queue_create_doit")
a5546e18f77c ("net: Add queue-create operation")
The series will conflict with io_uring work, and the code needs more
polish.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add a single device mode for netkit instead of netkit pairs. The primary
target for the paired devices is to connect network namespaces, of course,
and support has been implemented in projects like Cilium [0]. For the rxq
leasing the plan is to support two main scenarios related to single device
mode:
* For the use-case of io_uring zero-copy, the control plane can either
set up a netkit pair where the peer device can perform rxq leasing which
is then tied to the lifetime of the peer device, or the control plane
can use a regular netkit pair to connect the hostns to a Pod/container
and dynamically add/remove rxq leasing through a single device without
having to interrupt the device pair. In the case of io_uring, the memory
pool is used as skb non-linear pages, and thus the skb will go its way
through the regular stack into netkit. Things like the netkit policy when
no BPF is attached or skb scrubbing etc apply as-is in case the paired
devices are used, or if the backend memory is tied to the single device
and traffic goes through a paired device.
* For the use-case of AF_XDP, the control plane needs to use netkit in the
single device mode. The single device mode currently enforces only a
pass policy when no BPF is attached, and does not yet support BPF link
attachments for AF_XDP. skbs sent to that device get dropped at the
moment. Given AF_XDP operates at a lower layer of the stack tying this
to the netkit pair did not make sense. In future, the plan is to allow
BPF at the XDP layer which can: i) process traffic coming from the AF_XDP
application (e.g. QEMU with AF_XDP backend) to filter egress traffic or
to push selected egress traffic up to the single netkit device to the
local stack (e.g. DHCP requests), and ii) vice-versa skbs sent to the
single netkit into the AF_XDP application (e.g. DHCP replies). Also,
the control-plane can dynamically manage rxq leasing for the single
netkit device without having to interrupt (e.g. down/up cycle) the main
netkit pair for the Pod which has traffic going in and out.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Co-developed-by: David Wei <dw@davidwei.uk>
Signed-off-by: David Wei <dw@davidwei.uk>
Reviewed-by: Jordan Rife <jordan@jrife.io>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Link: https://docs.cilium.io/en/stable/operations/performance/tuning/#netkit-device-mode [0]
Link: https://patch.msgid.link/20260115082603.219152-10-daniel@iogearbox.net
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Add a ynl netdev family operation called queue-create that creates a
new queue on a netdevice:
name: queue-create
attribute-set: queue
flags: [admin-perm]
do:
request:
attributes:
- ifindex
- type
- lease
reply: &queue-create-op
attributes:
- id
This is a generic operation such that it can be extended for various
use cases in future. Right now it is mandatory to specify ifindex,
the queue type which is enforced to rx and a lease. The newly created
queue id is returned to the caller.
A queue from a virtual device can have a lease which refers to another
queue from a physical device. This is useful for memory providers
and AF_XDP operations which take an ifindex and queue id to allow
applications to bind against virtual devices in containers. The lease
couples both queues together and allows to proxy the operations from
a virtual device in a container to the physical device.
In future, the nested lease attribute can be lifted and made optional
for other use-cases such as dynamic queue creation for physical
netdevs. The lack of lease and the specification of the physical
device as an ifindex will imply that we need a real queue to be
allocated. Similarly, the queue type enforcement to rx can then be
lifted as well to support tx.
An early implementation had only driver-specific integration [0], but
in order for other virtual devices to reuse, it makes sense to have
this as a generic API in core net.
For leasing queues, the virtual netdev must have real_num_rx_queue
less than num_rx_queues at the time of calling queue-create. The
queue-type must be rx as only rx queues are supported for leasing
for now. We also enforce that the queue-create ifindex must point
to a virtual device, and that the nested lease attribute's ifindex
must point to a physical device. The nested lease attribute set
contains a netns-id attribute which is currently only intended for
dumping as part of the queue-get operation. Also, it is modeled as
an s32 type similarly as done elsewhere in the stack.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Co-developed-by: David Wei <dw@davidwei.uk>
Signed-off-by: David Wei <dw@davidwei.uk>
Link: https://bpfconf.ebpf.io/bpfconf2025/bpfconf2025_material/lsfmmbpf_2025_netkit_borkmann.pdf [0]
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Link: https://patch.msgid.link/20260115082603.219152-2-daniel@iogearbox.net
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
This commit adds shared shaper state across the cake instances beneath a
cake_mq qdisc. It works by periodically tracking the number of active
instances, and scaling the configured rate by the number of active
queues.
The scan is lockless and simply reads the qlen and the last_active state
variable of each of the instances configured beneath the parent cake_mq
instance. Locking is not required since the values are only updated by
the owning instance, and eventual consistency is sufficient for the
purpose of estimating the number of active queues.
The interval for scanning the number of active queues is set to 200 us.
We found this to be a good tradeoff between overhead and response time.
For a detailed analysis of this aspect see the Netdevconf talk:
https://netdevconf.info/0x19/docs/netdev-0x19-paper16-talk-paper.pdf
Reviewed-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Jonas Köppeler <j.koeppeler@tu-berlin.de>
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Link: https://patch.msgid.link/20260109-mq-cake-sub-qdisc-v8-5-8d613fece5d8@redhat.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Update .mailmap based on libbpf's list of contributors and on the latest
.mailmap version in the upstream repository.
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Jiri Olsa
Andy Roulin
Daniel Borkmann
Hangbin Liu
Menglong Dong
Alan Maguire
Andrii Nakryiko
Weiming Shi
Hoyeon Lee
Mykyta Yatsenko
Eyal Birger
Ihor Solodrai
Andrey Grodzovsky
Josef Bacik
Jakub Kicinski
Emil Tsalapatis
Paolo Abeni
Jonas Köppeler
Amery Hung