Kumar Kartikeya Dwivedi
b198f8b67b
The existing BPF_PROG_STREAM_READ_BY_FD command only supports polling a program stream through repeated bpf() calls. It cannot block for new data or integrate with poll-based event loops. Add BPF_PROG_STREAM_OPEN to return a read-only, close-on-exec file descriptor for a selected program stream. Reads block by default and BPF_F_STREAM_NONBLOCK, the only accepted flag, provides non-blocking behavior. poll reports readable data and reports hangup once the program has been freed. Like pipes and sockets, the descriptor is not seekable and lseek fails with ESPIPE. A stream descriptor deliberately does not retain the program. Move each stream into a separately refcounted allocation so program teardown can mark it dead and wake descriptor users while outstanding descriptors drain buffered data safely. Readers sample the dead flag before looking for data, so EOF is reported only when the stream was already dead before it was found empty; data published right before teardown is never skipped. Only programs loaded through BPF_PROG_LOAD get streams. Classic BPF filters, JIT subprograms and shim programs never write to one, and kernel-side writers already resolve a subprogram to its main program, so those programs no longer carry stream state. Readiness needs its own counter. Stream capacity is charged before allocation and before an element is published to the stream log, so using that reservation as the read and poll condition can report readable data while no element exists: a blocking reader retries instead of sleeping and a lone non-blocking reader can see POLLIN followed by EAGAIN. Publish bytes with release ordering after adding elements to the lockless log, use acquire loads before consuming them or reporting readiness, limit each read to its readable snapshot and subtract only bytes actually copied. This keeps the aggregate count correct even when concurrent publishers update it out of publication order. With several readers on one stream, readiness remains advisory, as it is for pipes. The capacity counter is kept solely for enforcing the stream size limit. Wakeups are always deferred through irq_work. Stream writers run in whatever context the program runs in: NMI context for perf_event programs, sections with interrupts disabled inside bpf_spin_lock or rqspinlock critical sections since bpf_stream_vprintk() is KF_SPINLOCK_SAFE, and tracing programs attached anywhere in the kernel, including inside the wait queue and epoll code itself. Waking waiters directly from there can deadlock, and no cheap context check covers every case: on PREEMPT_RT, spinlock_t sections do not disable interrupts, so in_nmi() or irqs_disabled() cannot tell such a program apart from a benign one. Queue an irq_work item instead, as bpf_ringbuf does. Queue it only when a publication turns an empty stream readable. Readers block and pollers wait only after finding the stream empty, and the readable count never drops below zero because each read is bounded by its snapshot, so the first publication after such an observation is the one that makes the count positive, and it is the one that queues the wakeup. Publications into a stream that already holds data raise no interrupt, so a program that prints while nobody drains its stream pays for a single irq_work until the stream is emptied again. This matches bpf_ringbuf, which notifies only once the consumer has caught up. Blocking readers and level-triggered pollers re-check the readable count before waiting, so they cannot miss data, and edge-triggered epoll consumers drain until EAGAIN before waiting again, as epoll(7) requires. Synchronize pending work before releasing the final stream reference so the callback cannot outlive the stream, but only when the work was ever queued: irq_work_sync() waits for an RCU grace period on PREEMPT_RT and on architectures without an irq_work interrupt. Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://patch.msgid.link/20260925045536.1480933-3-memxor@gmail.com
libbpf
This is the official home of the libbpf library.
Please use this Github repository for building and packaging libbpf and when using it in your projects through Git submodule.
Libbpf authoritative source code is developed as part of bpf-next Linux source
tree under
tools/lib/bpf subdirectory and is periodically synced to Github. As such, all the
libbpf changes should be sent to BPF mailing list,
please don't open PRs here unless you are changing Github-specific parts of libbpf
(e.g., Github-specific Makefile).
Libbpf and general BPF usage questions
Libbpf documentation can be found here. It's an ongoing effort and has ways to go, but please take a look and consider contributing as well.
Please check out libbpf-bootstrap and the companion blog post for the examples of building BPF applications with libbpf. libbpf-tools are also a good source of the real-world libbpf-based tracing tools.
See also "BPF CO-RE reference guide" for the coverage of practical aspects of building BPF CO-RE applications and "BPF CO-RE" for general introduction into BPF portability issues and BPF CO-RE origins.
All general BPF questions, including kernel functionality, libbpf APIs and their application, should be sent to bpf@vger.kernel.org mailing list. You can subscribe to it here and search its archive here. Please search the archive before asking new questions. It very well might be that this was already addressed or answered before.
bpf@vger.kernel.org is monitored by many more people and they will happily try to help you with whatever issue you have. This repository's PRs and issues should be opened only for dealing with issues pertaining to specific way this libbpf mirror repo is set up and organized.
Building libbpf
libelf is an internal dependency of libbpf and thus it is required to link
against and must be installed on the system for applications to work.
pkg-config is used by default to find libelf, and the program called can be
overridden with PKG_CONFIG.
If using pkg-config at build time is not desired, it can be disabled by
setting NO_PKG_CONFIG=1 when calling make.
To build both static libbpf.a and shared libbpf.so:
$ cd src
$ make
To build only static libbpf.a library in directory build/ and install them together with libbpf headers in a staging directory root/:
$ cd src
$ mkdir build root
$ BUILD_STATIC_ONLY=y OBJDIR=build DESTDIR=root make install
To build both static libbpf.a and shared libbpf.so against a custom libelf dependency installed in /build/root/ and install them together with libbpf headers in a build directory /build/root/:
$ cd src
$ PKG_CONFIG_PATH=/build/root/lib64/pkgconfig DESTDIR=/build/root make install
BPF CO-RE (Compile Once – Run Everywhere)
Libbpf supports building BPF CO-RE-enabled applications, which, in contrast to BCC, do not require Clang/LLVM runtime being deployed to target servers and doesn't rely on kernel-devel headers being available.
It does rely on kernel to be built with BTF type information, though. Some major Linux distributions come with kernel BTF already built in:
- Fedora 31+
- RHEL 8.2+
- OpenSUSE Tumbleweed (in the next release, as of 2020-06-04)
- Arch Linux (from kernel 5.7.1.arch1-1)
- Manjaro (from kernel 5.4 if compiled after 2021-06-18)
- Ubuntu 20.10
- Debian 11 (amd64/arm64)
If your kernel doesn't come with BTF built-in, you'll need to build custom kernel. You'll need:
pahole1.16+ tool (part ofdwarvespackage), which performs DWARF to BTF conversion;- kernel built with
CONFIG_DEBUG_INFO_BTF=yoption; - you can check if your kernel has BTF built-in by looking for
/sys/kernel/btf/vmlinuxfile:
$ ls -la /sys/kernel/btf/vmlinux
-r--r--r--. 1 root root 3541561 Jun 2 18:16 /sys/kernel/btf/vmlinux
To develop and build BPF programs, you'll need Clang/LLVM 10+. The following distributions have Clang/LLVM 10+ packaged by default:
- Fedora 32+
- Ubuntu 20.04+
- Arch Linux
- Ubuntu 20.10 (LLVM 11)
- Debian 11 (LLVM 11)
- Alpine 3.13+
Otherwise, please make sure to update it on your system.
The following resources are useful to understand what BPF CO-RE is and how to use it:
- BPF CO-RE reference guide
- BPF Portability and CO-RE
- HOWTO: BCC to libbpf conversion
- libbpf-tools in BCC repo contain lots of real-world tools converted from BCC to BPF CO-RE. Consider converting some more to both contribute to the BPF community and gain some more experience with it.
Distributions
Distributions packaging libbpf from this mirror:
Benefits of packaging from the mirror over packaging from kernel sources:
- Consistent versioning across distributions.
- No ties to any specific kernel, transparent handling of older kernels. Libbpf is designed to be kernel-agnostic and work across multitude of kernel versions. It has built-in mechanisms to gracefully handle older kernels, that are missing some of the features, by working around or gracefully degrading functionality. Thus libbpf is not tied to a specific kernel version and can/should be packaged and versioned independently.
- Continuous integration testing via GitHub Actions.
- Static code analysis via LGTM and Coverity.
Package dependencies of libbpf, package names may vary across distros:
- zlib
- libelf
bpf-next to Github sync
All the gory details of syncing can be found in scripts/sync-kernel.sh
script. See SYNC.md for instruction.
Some header files in this repo (include/linux/*.h) are reduced versions of
their counterpart files at
bpf-next's
tools/include/linux/*.h to make compilation successful.
License
This work is dual-licensed under BSD 2-clause license and GNU LGPL v2.1 license. You can choose between one of them if you use this work.
SPDX-License-Identifier: BSD-2-Clause OR LGPL-2.1