- 11 Mar, 2020 2 commits
-
-
Alexei Starovoitov authored
fmod_ret progs are emitted as: start = __bpf_prog_enter(); call fmod_ret *(u64 *)(rbp - 8) = rax __bpf_prog_exit(, start); test eax, eax jne do_fexit That 'test eax, eax' is working by accident. The compiler is free to use rax inside __bpf_prog_exit() or inside functions that __bpf_prog_exit() is calling. Which caused "test_progs -t modify_return" to sporadically fail depending on compiler version and kconfig. Fix it by using 'cmp [rbp - 8], 0' instead of 'test eax, eax'. Fixes: ae240823 ("bpf: Introduce BPF_MODIFY_RETURN") Reported-by:
Andrii Nakryiko <andriin@fb.com> Signed-off-by:
Alexei Starovoitov <ast@kernel.org> Signed-off-by:
Daniel Borkmann <daniel@iogearbox.net> Acked-by:
KP Singh <kpsingh@google.com> Link: https://lore.kernel.org/bpf/20200311003906.3643037-1-ast@kernel.org
-
Andrii Nakryiko authored
Add bpf_link_new_file() API for cases when we need to ensure anon_inode is successfully created before we proceed with expensive BPF program attachment procedure, which will require equally (if not more so) expensive and potentially failing compensation detachment procedure just because anon_inode creation failed. This API allows to simplify code by ensuring first that anon_inode is created and after BPF program is attached proceed with fd_install() that can't fail. After anon_inode file is created, link can't be just kfree()'d anymore, because its destruction will be performed by deferred file_operations->release call. For this, bpf_link API required specifying two separate operations: release() and dealloc(), former performing detachment only, while the latter frees memory used by bpf_link itself. dealloc() needs to be specified, because struct bpf_link is frequently embedded into link type-specific container struct (e.g., struct bpf_raw_tp_link), so bpf_link itself doesn't know how to properly free the memory. In case when anon_inode file was successfully created, but subsequent BPF attachment failed, bpf_link needs to be marked as "defunct", so that file's release() callback will perform only memory deallocation, but no detachment. Convert raw tracepoint and tracing attachment to new API and eliminate detachment from error handling path. Signed-off-by:
John Fastabend <john.fastabend@gmail.com> Link: https://lore.kernel.org/bpf/20200309231051.1270337-1-andriin@fb.com
-
- 09 Mar, 2020 16 commits
-
-
Song Liu authored
_bpftool_get_map_names => _bpftool_get_prog_names for prog-attach|detach. Fixes: 99f9863a ("bpftool: Match maps by name") Signed-off-by:
Song Liu <songliubraving@fb.com> Signed-off-by:
Quentin Monnet <quentin@isovalent.com> Link: https://lore.kernel.org/bpf/20200309173218.2739965-5-songliubraving@fb.com
-
Song Liu authored
Add bash completion for "bpftool prog profile" command. Signed-off-by:
-
Song Liu authored
Add documentation for the new bpftool prog profile command. Signed-off-by:
-
Song Liu authored
With fentry/fexit programs, it is possible to profile BPF program with hardware counters. Introduce bpftool "prog profile", which measures key metrics of a BPF program. bpftool prog profile command creates per-cpu perf events. Then it attaches fentry/fexit programs to the target BPF program. The fentry program saves perf event value to a map. The fexit program reads the perf event again, and calculates the difference, which is the instructions/cycles used by the target program. Example input and output: ./bpftool prog profile id 337 duration 3 cycles instructions llc_misses 4228 run_cnt 3403698 cycles (84.08%) 3525294 instructions # 1.04 insn per cycle (84.05%) 13 llc_misses # 3.69 LLC misses per million isns (83.50%) This command measures cycles and instructions for BPF program with id 337 for 3 seconds. The program has triggered 4228 times. The rest of the output is similar to perf-stat. In this example, the counters were only counting ~84% of the time because of time multiplexing of perf counters. Note that, this approach measures cycles and instructions in very small increments. So the fentry/fexit programs introduce noticeable errors to the measurement results. The fentry/fexit programs are generated with BPF skeletons. Therefore, we build bpftool twice. The first time _bpftool is built without skeletons. Then, _bpftool is used to generate the skeletons. The second time, bpftool is built with skeletons. Signed-off-by:Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/bpf/20200309173218.2739965-2-songliubraving@fb.com
-
Lorenz Bauer authored
Add Jakub and myself as maintainers for sockmap related code. Signed-off-by:
Lorenz Bauer <lmb@cloudflare.com> Signed-off-by:
Jakub Sitnicki <jakub@cloudflare.com> Acked-by:
-
Lorenz Bauer authored
Remove the guard that disables UDP tests now that sockmap has support for them. Signed-off-by:
-
Lorenz Bauer authored
Expand the TCP sockmap test suite to also check UDP sockets. Signed-off-by:
-
Lorenz Bauer authored
Most tests for TCP sockmap can be adapted to UDP sockmap if the listen call is skipped. Rename listen_loopback, etc. to socket_loopback and skip listen() for SOCK_DGRAM. Signed-off-by:
-
Lorenz Bauer authored
Allow adding hashed UDP sockets to sockmaps. Signed-off-by:
-
Lorenz Bauer authored
Add basic psock hooks for UDP sockets. This allows adding and removing sockets, as well as automatic removal on unhash and close. Signed-off-by:
-
Lorenz Bauer authored
We can take advantage of the fact that both callers of sock_map_init_proto are holding a RCU read lock, and have verified that psock is valid. Signed-off-by:
-
Lorenz Bauer authored
The init, close and unhash handlers from TCP sockmap are generic, and can be reused by UDP sockmap. Move the helpers into the sockmap code base and expose them. This requires tcp_bpf_get_proto and tcp_bpf_clone to be conditional on BPF_STREAM_PARSER. The moved functions are unmodified, except that sk_psock_unlink is renamed to sock_map_unlink to better match its behaviour. Signed-off-by:
-
Lorenz Bauer authored
tcp_bpf.c is only included in the build if CONFIG_NET_SOCK_MSG is selected. The declaration should therefore be guarded as such. Signed-off-by:
-
Lorenz Bauer authored
We need to ensure that sk->sk_prot uses certain callbacks, so that code that directly calls e.g. tcp_sendmsg in certain corner cases works. To avoid spurious asserts, we must to do this only if sk_psock_update_proto has not yet been called. The same invariants apply for tcp_bpf_check_v6_needs_rebuild, so move the call as well. Doing so allows us to merge tcp_bpf_init and tcp_bpf_reinit. Signed-off-by:
-
Lorenz Bauer authored
Only update psock->saved_* if psock->sk_proto has not been initialized yet. This allows us to get rid of tcp_bpf_reinit_sk_prot. Signed-off-by:
-
Lorenz Bauer authored
The sock map code checks that a socket does not have an active upper layer protocol before inserting it into the map. This requires casting via inet_csk, which isn't valid for UDP sockets. Guard checks for ULP by checking inet_sk(sk)->is_icsk first. Signed-off-by:
-
- 05 Mar, 2020 6 commits
-
-
KP Singh authored
test_run.o is not built when CONFIG_NET is not set and bpf_prog_test_run_tracing being referenced in bpf_trace.o causes the linker error: ld: kernel/trace/bpf_trace.o:(.rodata+0x38): undefined reference to `bpf_prog_test_run_tracing' Add a __weak function in bpf_trace.c to handle this. Fixes: da00d2f1 ("bpf: Add test ops for BPF_PROG_TYPE_TRACING") Signed-off-by:
Randy Dunlap <rdunlap@infradead.org> Acked-by:
-
KP Singh authored
While well intentioned, checking CAP_MAC_ADMIN for attaching BPF_MODIFY_RETURN tracing programs to "security_" functions is not necessary as tracing BPF programs already require CAP_SYS_ADMIN. Fixes: 6ba43b76 ("bpf: Attachment verification for BPF_MODIFY_RETURN") Signed-off-by:
-
Luke Nelson authored
Add a new entry for the 32-bit RISC-V JIT to MAINTAINERS and change mailing list to netdev and bpf following the guidelines from commit e42da4c6 ("docs/bpf: Update bpf development Q/A file"). Signed-off-by:
Xi Wang <xi.wang@gmail.com> Signed-off-by:
Luke Nelson <luke.r.nels@gmail.com> Signed-off-by:
Björn Töpel <bjorn.topel@gmail.com> Acked-by:
-
Luke Nelson authored
Update filter.txt and admin-guide to mention the BPF JIT for RV32G. Co-developed-by:
-
Luke Nelson authored
This is an eBPF JIT for RV32G, adapted from the JIT for RV64G and the 32-bit ARM JIT. There are two main changes required for this to work compared to the RV64 JIT. First, eBPF registers are 64-bit, while RV32G registers are 32-bit. BPF registers either map directly to 2 RISC-V registers, or reside in stack scratch space and are saved and restored when used. Second, many 64-bit ALU operations do not trivially map to 32-bit operations. Operations that move bits between high and low words, such as ADD, LSH, MUL, and others must emulate the 64-bit behavior in terms of 32-bit instructions. This patch also makes related changes to bpf_jit.h, such as adding RISC-V instructions required by the RV32 JIT. Supported features: The RV32 JIT supports the same features and instructions as the RV64 JIT, with the following exceptions: - ALU64 DIV/MOD: Requires loops to implement on 32-bit hardware. - BPF_XADD | BPF_DW: There's no 8-byte atomic instruction in RV32. These features are also unsupported on other BPF JITs for 32-bit architectures. Testing: - lib/test_bpf.c test_bpf: Summary: 378 PASSED, 0 FAILED, [349/366 JIT'ed] test_bpf: test_skb_segment: Summary: 2 PASSED, 0 FAILED The tests that are not JITed are all due to use of 64-bit div/mod or 64-bit xadd. - tools/testing/selftests/bpf/test_verifier.c Summary: 1415 PASSED, 122 SKIPPED, 43 FAILED Tested both with and without BPF JIT hardening. This is the same set of tests that pass using the BPF interpreter with the JIT disabled. Verification and synthesis: We developed the RV32 JIT using our automated verification tool, Serval. We have used Serval in the past to verify patches to the RV64 JIT. We also used Serval to superoptimize the resulting code through program synthesis. You can find the tool and a guide to the approach and results here: https://github.com/uw-unsat/serval-bpf/tree/rv32-jit-v5Co-developed-by:
-
Luke Nelson authored
This patch factors out code that can be used by both the RV64 and RV32 BPF JITs to a common bpf_jit.h and bpf_jit_core.c. Move struct definitions and macro-like functions to header. Rename rv_sb_insn/rv_uj_insn to rv_b_insn/rv_j_insn to match the RISC-V specification. Move reusable functions emit_body() and bpf_int_jit_compile() to bpf_jit_core.c with minor simplifications. Rename emit_insn() and build_{prologue,epilogue}() to be prefixed with "bpf_jit_" as they are no longer static. Rename bpf_jit_comp.c to bpf_jit_comp64.c to be more explicit. Co-developed-by:
-
- 04 Mar, 2020 16 commits
-
-
Alexei Starovoitov authored
KP Singh says: ==================== v3 -> v4: * Fix a memory leak noticed by Daniel. v2 -> v3: * bpf_trampoline_update_progs -> bpf_trampoline_get_progs + const qualification. * Typos in commit messages. * Added Andrii's Acks. v1 -> v2: * Adressed Andrii's feedback. * Fixed a bug that Alexei noticed about nop generation. * Rebase. This was brought up in the KRSI v4 discussion and found to be useful both for security and tracing programs. https://lore.kernel.org/bpf/20200225193108.GB22391@chromium.org/ The modify_return programs are allowed for security hooks (with an extra CAP_MAC_ADMIN check) and functions whitelisted for error injection (ALLOW_ERROR_INJECTION). The "security_" check is expected to be cleaned up with the KRSI patch series. Here is an example of how a fmod_ret program behaves: int func_to_be_attached(int a, int b) { <--- do_fentry do_fmod_ret: <update ret by calling fmod_ret> if (ret != 0) goto do_fexit; original_function: <side_effects_happen_here> } <--- do_fexit ALLOW_ERROR_INJECTION(func_to_be_attached, ERRNO) The fmod_ret program attached to this function can be defined as: SEC("fmod_ret/func_to_be_attached") int BPF_PROG(func_name, int a, int b, int ret) { // This will skip the original function logic. return -1; } ==================== Signed-off-by:
-
KP Singh authored
Test for two scenarios: * When the fmod_ret program returns 0, the original function should be called along with fentry and fexit programs. * When the fmod_ret program returns a non-zero value, the original function should not be called, no side effect should be observed and fentry and fexit programs should be called. The result from the kernel function call and whether a side-effect is observed is returned via the retval attr of the BPF_PROG_TEST_RUN (bpf) syscall. Signed-off-by: -
KP Singh authored
The current fexit and fentry tests rely on a different program to exercise the functions they attach to. Instead of doing this, implement the test operations for tracing which will also be used for BPF_MODIFY_RETURN in a subsequent patch. Also, clean up the fexit test to use the generated skeleton. Signed-off-by:
-
KP Singh authored
Signed-off-by:
-
KP Singh authored
- Allow BPF_MODIFY_RETURN attachment only to functions that are: * Whitelisted for error injection by checking within_error_injection_list. Similar discussions happened for the bpf_override_return helper. * security hooks, this is expected to be cleaned up with the LSM changes after the KRSI patches introduce the LSM_HOOK macro: https://lore.kernel.org/bpf/20200220175250.10795-1-kpsingh@chromium.org/ - The attachment is currently limited to functions that return an int. This can be extended later other types (e.g. PTR). Signed-off-by: -
KP Singh authored
When multiple programs are attached, each program receives the return value from the previous program on the stack and the last program provides the return value to the attached function. The fmod_ret bpf programs are run after the fentry programs and before the fexit programs. The original function is only called if all the fmod_ret programs return 0 to avoid any unintended side-effects. The success value, i.e. 0 is not currently configurable but can be made so where user-space can specify it at load time. For example: int func_to_be_attached(int a, int b) { <--- do_fentry do_fmod_ret: <update ret by calling fmod_ret> if (ret != 0) goto do_fexit; original_function: <side_effects_happen_here> } <--- do_fexit The fmod_ret program attached to this function can be defined as: SEC("fmod_ret/func_to_be_attached") int BPF_PROG(func_name, int a, int b, int ret) { // This will skip the original function logic. return 1; } The first fmod_ret program is passed 0 in its return argument. Signed-off-by: -
KP Singh authored
* Split the invoke_bpf program to prepare for special handling of fmod_ret programs introduced in a subsequent patch. * Move the definition of emit_cond_near_jump and emit_nops as they are needed for fmod_ret. * Refactor branch target alignment into its own generic helper function i.e. emit_align. Signed-off-by:
-
KP Singh authored
As we need to introduce a third type of attachment for trampolines, the flattened signature of arch_prepare_bpf_trampoline gets even more complicated. Refactor the prog and count argument to arch_prepare_bpf_trampoline to use bpf_tramp_progs to simplify the addition and accounting for new attachment types. Signed-off-by:
-
Andrii Nakryiko authored
Add detection of out-of-tree built vmlinux image for the purpose of VMLINUX_BTF detection. According to Documentation/kbuild/kbuild.rst, O takes precedence over KBUILD_OUTPUT. Also ensure ~/path/to/build/dir also works by relying on wildcard's resolution first, but then applying $(abspath) at the end to also handle O=../../whatever cases. Signed-off-by:
-
Kees Cook authored
When CONFIG_DEBUG_INFO is enabled, the two kallsyms linking steps spend time collecting and writing the dwarf sections to the temporary output files. kallsyms does not need this information, and leaving it off halves their linking time. This is especially noticeable without CONFIG_DEBUG_INFO_REDUCED. The BTF linking stage, however, does still need those details. Refactor the BTF and kallsyms generation stages slightly for more regularized temporary names. Skip debug during kallsyms links. Additionally move "info BTF" to the correct place since commit 8959e392 ("kbuild: Parameterize kallsyms generation and correct reporting"), which added "info LD ..." to vmlinux_link calls. For a full debug info build with BTF, my link time goes from 1m06s to 0m54s, saving about 12 seconds, or 18%. Signed-off-by:
Kees Cook <keescook@chromium.org> Signed-off-by:
-
Daniel Borkmann authored
Andrii Nakryiko says: ==================== Convert BPF-related UAPI constants, currently defined as #define macro, into anonymous enums. This has no difference in terms of usage of such constants in C code (they are still could be used in all the compile-time contexts that `#define`s can), but they are recorded as part of DWARF type info, and subsequently get recorded as part of kernel's BTF type info. This allows those constants to be emitted as part of vmlinux.h auto-generated header file and be used from BPF programs. Which is especially convenient for all kinds of BPF helper flags and makes CO-RE BPF programs nicer to write. libbpf's btf_dump logic currently assumes enum values are signed 32-bit values, but that doesn't match a typical case, so switch it to emit unsigned values. Once BTF encoding of BTF_KIND_ENUM is extended to capture signedness properly, this will be made more flexible. As an immediate validation of the approach, runqslower's copy of BPF_F_CURRENT_CPU #define is dropped in favor of its enum variant from vmlinux.h. v2->v3: - convert only constants usable from BPF programs (BPF helper flags, map create flags, etc) (Alexei); v1->v2: - fix up btf_dump test to use max 32-bit unsigned value instead of negative one. ==================== Signed-off-by: -
Andrii Nakryiko authored
With BPF_F_CURRENT_CPU being an enum, it is now captured in vmlinux.h and is readily usable by runqslower. So drop local copy/pasted definition in favor of the one coming from vmlinux.h. Signed-off-by:
-
Andrii Nakryiko authored
Currently, BTF_KIND_ENUM type doesn't record whether enum values should be interpreted as signed or unsigned. In Linux, most enums are unsigned, though, so interpreting them as unsigned matches real world better. Change btf_dump test case to test maximum 32-bit value, instead of negative value. Signed-off-by:
-
Andrii Nakryiko authored
Switch BPF UAPI constants, previously defined as #define macro, to anonymous enum values. This preserves constants values and behavior in expressions, but has added advantaged of being captured as part of DWARF and, subsequently, BTF type info. Which, in turn, greatly improves usefulness of generated vmlinux.h for BPF applications, as it will not require BPF users to copy/paste various flags and constants, which are frequently used with BPF helpers. Only those constants that are used/useful from BPF program side are converted. Signed-off-by:
-
Andrii Nakryiko authored
Internal functions, used by btf_dump__emit_type_decl(), assume field_name is never going to be NULL. Ensure it's always the case. Fixes: 9f81654e ("libbpf: Expose BTF-to-C type declaration emitting API") Signed-off-by:
-
Alexei Starovoitov authored
Willem de Bruijn says: ==================== See first patch for details. Patch split across three parts { kernel feature, uapi header, tools } following the custom for such __sk_buff changes. ==================== Acked-by:Petar Penkov <ppenkov@google.com> Signed-off-by:
-
