bcc uses libbpf repo as a submodule. It brings in libbpf source
code and builds everything together to produce shared libraries.
With latest libbpf, I got the following errors:
/bin/ld: libbcc_bpf.so.0.10.0: version node not found for symbol xsk_umem__create@LIBBPF_0.0.2
/bin/ld: failed to set dynamic section sizes: Bad value
collect2: error: ld returned 1 exit status
make[2]: *** [src/cc/libbcc_bpf.so.0.10.0] Error 1
In xsk.c, we have
asm(".symver xsk_umem__create_v0_0_2, xsk_umem__create@LIBBPF_0.0.2");
asm(".symver xsk_umem__create_v0_0_4, xsk_umem__create@@LIBBPF_0.0.4");
The linker thinks the built is for LIBBPF but cannot find proper version
LIBBPF_0.0.2/4, so emit errors.
I also confirmed that using libbpf.a to produce a shared library also
has issues:
-bash-4.4$ cat t.c
extern void *xsk_umem__create;
void * test() { return xsk_umem__create; }
-bash-4.4$ gcc -c -fPIC t.c
-bash-4.4$ gcc -shared t.o libbpf.a -o t.so
/bin/ld: t.so: version node not found for symbol xsk_umem__create@LIBBPF_0.0.2
/bin/ld: failed to set dynamic section sizes: Bad value
collect2: error: ld returned 1 exit status
-bash-4.4$
Symbol versioning does happens in commonly used libraries, e.g., elfutils
and glibc. For static libraries, for a versioned symbol, the old definitions
will be ignored, and the symbol will be an alias to the latest definition.
For example, glibc sched_setaffinity is versioned.
-bash-4.4$ readelf -s /usr/lib64/libc.so.6 | grep sched_setaffinity
756: 000000000013d3d0 13 FUNC GLOBAL DEFAULT 13 sched_setaffinity@GLIBC_2.3.3
757: 00000000000e2e70 455 FUNC GLOBAL DEFAULT 13 sched_setaffinity@@GLIBC_2.3.4
1800: 0000000000000000 0 FILE LOCAL DEFAULT ABS sched_setaffinity.c
4228: 00000000000e2e70 455 FUNC LOCAL DEFAULT 13 __sched_setaffinity_new
4648: 000000000013d3d0 13 FUNC LOCAL DEFAULT 13 __sched_setaffinity_old
7338: 000000000013d3d0 13 FUNC GLOBAL DEFAULT 13 sched_setaffinity@GLIBC_2
7380: 00000000000e2e70 455 FUNC GLOBAL DEFAULT 13 sched_setaffinity@@GLIBC_
-bash-4.4$
For static library, the definition of sched_setaffinity aliases to the new definition.
-bash-4.4$ readelf -s /usr/lib64/libc.a | grep sched_setaffinity
File: /usr/lib64/libc.a(sched_setaffinity.o)
8: 0000000000000000 455 FUNC GLOBAL DEFAULT 1 __sched_setaffinity_new
12: 0000000000000000 455 FUNC WEAK DEFAULT 1 sched_setaffinity
For both elfutils and glibc, additional macros are used to control different handling
of symbol versioning w.r.t static and shared libraries.
For elfutils, the macro is SYMBOL_VERSIONING
(https://sourceware.org/git/?p=elfutils.git;a=blob;f=lib/eu-config.h).
For glibc, the macro is SHARED
(https://sourceware.org/git/?p=glibc.git;a=blob;f=include/shlib-compat.h;hb=refs/heads/master)
This patch used SHARED as the macro name. After this patch, the libbpf.a has
-bash-4.4$ readelf -s libbpf.a | grep xsk_umem__create
372: 0000000000017145 1190 FUNC GLOBAL DEFAULT 1 xsk_umem__create_v0_0_4
405: 0000000000017145 1190 FUNC GLOBAL DEFAULT 1 xsk_umem__create
499: 00000000000175eb 103 FUNC GLOBAL DEFAULT 1 xsk_umem__create_v0_0_2
-bash-4.4$
No versioned symbols for xsk_umem__create.
The libbpf.a can be used to build a shared library succesfully.
-bash-4.4$ cat t.c
extern void *xsk_umem__create;
void * test() { return xsk_umem__create; }
-bash-4.4$ gcc -c -fPIC t.c
-bash-4.4$ gcc -shared t.o libbpf.a -o t.so
-bash-4.4$
Fixes: 10d30e3017 ("libbpf: add flags to umem config")
Cc: Kevin Laatz <kevin.laatz@intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Andrii Nakryiko <andriin@fb.com>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
On my local machine, I have the following compilation errors:
=====
In file included from prog_tests/core_reloc.c:3:0:
./progs/core_reloc_types.h:517:46: error: expected ‘=’, ‘,’, ‘;’, ‘asm’ or ‘__attribute__’ before ‘fancy_char_ptr_t’
typedef const char * const volatile restrict fancy_char_ptr_t;
^
./progs/core_reloc_types.h:527:2: error: unknown type name ‘fancy_char_ptr_t’
fancy_char_ptr_t d;
^
=====
I am using gcc 4.8.5. Later compilers may change their behavior not emitting the
error. Nevertheless, let us fix the issue. "restrict" can be tested
without typedef.
Fixes: 9654e2ae90 ("selftests/bpf: add CO-RE relocs modifiers/typedef tests")
Cc: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
With CONFIG_MODULES=n, the following compiler warning occurs:
/data/users/yhs/work/net-next/kernel/trace/bpf_trace.c:605:13: warning:
‘do_bpf_send_signal’ defined but not used [-Wunused-function]
static void do_bpf_send_signal(struct irq_work *entry)
The __init function send_signal_irq_work_init(), which calls
do_bpf_send_signal(), is defined under CONFIG_MODULES. Hence,
when CONFIG_MODULES=n, nobody calls static function do_bpf_send_signal(),
hence the warning.
The init function send_signal_irq_work_init() should work without
CONFIG_MODULES. Moving it out of CONFIG_MODULES
code section fixed the compiler warning, and also make bpf_send_signal()
helper work without CONFIG_MODULES.
Fixes: 8b401f9ed2 ("bpf: implement bpf_send_signal() helper")
Reported-By: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Yonghong Song <yhs@fb.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Commit 8b401f9ed2 ("bpf: implement bpf_send_signal() helper")
introduced bpf_send_signal() helper. If the context is nmi,
the sending signal work needs to be deferred to irq_work.
If the signal is invalid, the error will appear in irq_work
and it won't be propagated to user.
This patch did an early check in the helper itself to notify
user invalid signal, as suggested by Daniel.
Suggested-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
The bpf uapi header include/uapi/linux/bpf.h is sync'ed
to tools/include/uapi/linux/bpf.h.
Acked-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This patch tries to solve the following specific use case.
Currently, bpf program can already collect stack traces
through kernel function get_perf_callchain()
when certain events happens (e.g., cache miss counter or
cpu clock counter overflows). But such stack traces are
not enough for jitted programs, e.g., hhvm (jited php).
To get real stack trace, jit engine internal data structures
need to be traversed in order to get the real user functions.
bpf program itself may not be the best place to traverse
the jit engine as the traversing logic could be complex and
it is not a stable interface either.
Instead, hhvm implements a signal handler,
e.g. for SIGALARM, and a set of program locations which
it can dump stack traces. When it receives a signal, it will
dump the stack in next such program location.
Such a mechanism can be implemented in the following way:
. a perf ring buffer is created between bpf program
and tracing app.
. once a particular event happens, bpf program writes
to the ring buffer and the tracing app gets notified.
. the tracing app sends a signal SIGALARM to the hhvm.
But this method could have large delays and causing profiling
results skewed.
This patch implements bpf_send_signal() helper to send
a signal to hhvm in real time, resulting in intended stack traces.
Acked-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
For a host which has a lower rlimit for max locked memory (e.g., 64KB),
the following error occurs in one of our production systems:
# /usr/sbin/bpftool prog load /paragon/pods/52877437/home/mark.o \
/sys/fs/bpf/paragon_mark_21 type cgroup/skb \
map idx 0 pinned /sys/fs/bpf/paragon_map_21
libbpf: Error in bpf_object__probe_name():Operation not permitted(1).
Couldn't load basic 'r0 = 0' BPF program.
Error: failed to open object file
The reason is due to low locked memory during bpf_object__probe_name()
which probes whether program name is supported in kernel or not
during __bpf_object__open_xattr().
bpftool program load already tries to relax mlock rlimit before
bpf_object__load(). Let us move set_max_rlimit() before
__bpf_object__open_xattr(), which fixed the issue here.
Fixes: 47eff61777 ("bpf, libbpf: introduce bpf_object__probe_caps to test BPF capabilities")
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Test test_libbpf.sh failed on my development server with failure
-bash-4.4$ sudo ./test_libbpf.sh
[0] libbpf: Error in bpf_object__probe_name():Operation not permitted(1).
Couldn't load basic 'r0 = 0' BPF program.
test_libbpf: failed at file test_l4lb.o
selftests: test_libbpf [FAILED]
-bash-4.4$
The reason is because my machine has 64KB locked memory by default which
is not enough for this program to get locked memory.
Similar to other bpf selftests, let us increase RLIMIT_MEMLOCK
to infinity, which fixed the issue.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
I hit the following compilation error with gcc 4.8.5.
prog_tests/flow_dissector.c: In function ‘test_flow_dissector’:
prog_tests/flow_dissector.c:155:2: error: ‘for’ loop initial declarations are only allowed in C99 mode
for (int i = 0; i < ARRAY_SIZE(tests); i++) {
^
prog_tests/flow_dissector.c:155:2: note: use option -std=c99 or -std=gnu99 to compile your code
Let us fix the issue by avoiding this particular c99 feature.
Fixes: a5cb33464e ("selftests/bpf: make flow dissector tests more extensible")
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The issue is reported at https://github.com/libbpf/libbpf/issues/28.
Basically, per C standard, for
void *memcpy(void *dest, const void *src, size_t n)
if "dest" or "src" is NULL, regardless of whether "n" is 0 or not,
the result of memcpy is undefined. clang ubsan reported three such
instances in bpf.c with the following pattern:
memcpy(dest, 0, 0).
Although in practice, no known compiler will cause issues when
copy size is 0. Let us still fix the issue to silence ubsan
warnings.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
For multiple dimensional arrays like below,
int a[2][3]
both llvm and pahole generated one BTF_KIND_ARRAY type like
. element_type: int
. index_type: unsigned int
. number of elements: 6
Such a collapsed BTF_KIND_ARRAY type will cause the divergence
in BTF vs. the user code. In the compile-once-run-everywhere
project, the header file is generated from BTF and used for bpf
program, and the definition in the header file will be different
from what user expects.
But the kernel actually supports chained multi-dimensional array
types properly. The above "int a[2][3]" can be represented as
Type #n:
. element_type: int
. index_type: unsigned int
. number of elements: 3
Type #(n+1):
. element_type: type #n
. index_type: unsigned int
. number of elements: 2
The following llvm commit
https://reviews.llvm.org/rL357215
also enables llvm to generated proper chained multi-dimensional arrays.
The test_btf already has a raw test ("struct test #1") for chained
multi-dimensional arrays. This patch added amended bpffs test for
chained multi-dimensional arrays.
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
The test_progs subtests, test_spin_lock() and test_map_lock(),
requires BTF present to run successfully.
Currently, when BTF failed to load, test_progs will segfault,
$ ./test_progs
...
12: (bf) r1 = r8
13: (85) call bpf_spin_lock#93
map 'hash_map' has to have BTF in order to use bpf_spin_lock
libbpf: -- END LOG --
libbpf: failed to load program 'map_lock_demo'
libbpf: failed to load object './test_map_lock.o'
test_map_lock:bpf_prog_load errno 13
Segmentation fault
The segfault is caused by uninitialized variable "obj", which
is used in bpf_object__close(obj), when bpf prog failed to load.
Initializing variable "obj" to NULL in two occasions fixed the problem.
$ ./test_progs
...
Summary: 219 PASSED, 2 FAILED
Fixes: b4d4556c32 ("selftests/bpf: add bpf_spin_lock verifier tests")
Fixes: ba72a7b4ba ("selftests/bpf: test for BPF_F_LOCK")
Reported-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Yonghong Song <yhs@fb.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Compiling xdpsock_user.c with 4.8.5, I hit the following
compilation warning:
HOSTCC samples/bpf/xdpsock_user.o
/data/users/yhs/work/net-next/samples/bpf/xdpsock_user.c: In function ‘main’:
/data/users/yhs/work/net-next/samples/bpf/xdpsock_user.c:449:6: warning: ‘idx_cq’ may be used unini
tialized in this function [-Wmaybe-uninitialized]
u32 idx_cq, idx_fq;
^
/data/users/yhs/work/net-next/samples/bpf/xdpsock_user.c:606:7: warning: ‘idx_rx’ may be used unini
tialized in this function [-Wmaybe-uninitialized]
u32 idx_rx, idx_tx = 0;
^
/data/users/yhs/work/net-next/samples/bpf/xdpsock_user.c:506:6: warning: ‘idx_rx’ may be used unini
tialized in this function [-Wmaybe-uninitialized]
u32 idx_rx, idx_fq = 0;
As an example, the code pattern looks like:
u32 idx_cq;
...
ret = xsk_ring_prod__reserve(&xsk->umem->fq, rcvd, &idx_fq);
if (ret) {
...
}
... idx_fq ...
The compiler warns since it does not know whether &idx_fq is assigned
or not inside the library function xsk_ring_prod__reserve().
Let us assign an initial value 0 to such auto variables to silence
compiler warning.
Fixes: 248c7f9c0e ("samples/bpf: convert xdpsock to use libbpf for AF_XDP access")
Signed-off-by: Yonghong Song <yhs@fb.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This is to avoid permission denied error. A lot of systems
may have a much lower number, e.g., 64KB, for RLIMIT_MEMLOCK,
which may not be sufficient for the test to run successfully.
Fixes: e0b27b3f97 ("bpf: Add test_sock_fields for skb->sk and bpf_tcp_sock")
Signed-off-by: Yonghong Song <yhs@fb.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Commit d83525ca62 ("bpf: introduce bpf_spin_lock")
introduced bpf_spin_lock and the field spin_lock_off
in kernel internal structure bpf_map has the following
meaning:
>=0 valid offset, <0 error
For every map created, the kernel will ensure
spin_lock_off has correct value.
Currently, bpf_map->spin_lock_off is not copied
from the inner map to the map_in_map inner_map_meta
during a map_in_map type map creation, so
inner_map_meta->spin_lock_off = 0.
This will give verifier wrong information that
inner_map has bpf_spin_lock and the bpf_spin_lock
is defined at offset 0. An access to offset 0
of a value pointer will trigger the following error:
bpf_spin_lock cannot be accessed directly by load/store
This patch fixed the issue by copy inner map's spin_lock_off
value to inner_map_meta->spin_lock_off.
Fixes: d83525ca62 ("bpf: introduce bpf_spin_lock")
Signed-off-by: Yonghong Song <yhs@fb.com>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The kernel verifier has three levels of logs:
0: no logs
1: logs mostly useful
> 1: verbose
Current libbpf API functions bpf_load_program_xattr() and
bpf_load_program() cannot specify log_level.
The bcc, however, provides an interface for user to
specify log_level 2 for verbose output.
This patch added log_level into structure
bpf_load_program_attr, so users, including bcc, can use
bpf_load_program_xattr() to change log_level. The
supported log_level is 0, 1, and 2.
The bpf selftest test_sock.c is modified to enable log_level = 2.
If the "verbose" in test_sock.c is changed to true,
the test will output logs like below:
$ ./test_sock
func#0 @0
0: R1=ctx(id=0,off=0,imm=0) R10=fp0,call_-1
0: (bf) r6 = r1
1: R1=ctx(id=0,off=0,imm=0) R6_w=ctx(id=0,off=0,imm=0) R10=fp0,call_-1
1: (61) r7 = *(u32 *)(r6 +28)
invalid bpf_context access off=28 size=4
Test case: bind4 load with invalid access: src_ip6 .. [PASS]
...
Test case: bind6 allow all .. [PASS]
Summary: 16 PASSED, 0 FAILED
Some test_sock tests are negative tests and verbose verifier
log will be printed out as shown in the above.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Commit 96408c4344 ("tools/bpf: implement libbpf
btf__get_map_kv_tids() API function") refactored
function bpf_map_find_btf_info() and moved bulk of
implementation into btf.c as btf__get_map_kv_tids().
This change introduced a bug such that test_btf will
print out the following warning although the test passed:
BTF libbpf test[2] (test_btf_nokv.o): libbpf: map:btf_map
container_name:____btf_map_btf_map cannot be found
in BTF. Missing BPF_ANNOTATE_KV_PAIR?
Previously, the error message is guarded with pr_debug().
Commit 96408c4344 changed it to pr_warning() and
hence caused the warning.
Restoring to pr_debug() for the message fixed the issue.
Fixes: 96408c4344 ("tools/bpf: implement libbpf btf__get_map_kv_tids() API function")
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Commit 96408c4344 ("tools/bpf: implement libbpf btf__get_map_kv_tids() API function")
added the API function btf__get_map_kv_tids():
btf__get_map_kv_tids(const struct btf *btf, char *map_name, ...)
The parameter map_name has type "char *". This is okay inside libbpf library since
the map_name is from bpf_map->name which also has type "char *".
This will be problematic if the caller for map_name already has attribute "const",
e.g., from C++ string.c_str(). It will result in either a warning or an error.
/home/yhs/work/bcc/src/cc/btf.cc:166:51:
error: invalid conversion from ‘const char*’ to ‘char*’ [-fpermissive]
return btf__get_map_kv_tids(btf_, map_name.c_str()
This patch added "const" attributes to map_name parameter.
Fixes: 96408c4344 ("tools/bpf: implement libbpf btf__get_map_kv_tids() API function")
Signed-off-by: Yonghong Song <yhs@fb.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Commit 9c65112744 ("selftests/btf: add initial BTF dedup tests")
added dedup tests in test_btf.c.
It broke the raw test:
BTF raw test[71] (func proto (Bad arg name_off)):
btf_raw_create:2905:FAIL Error getting string #65535, strs_cnt:1
The test itself encodes invalid func_proto parameter name
offset 0xffffFFFF as a negative test for the kernel.
The above commit changed the meaning of that offset and
resulted in a user space error.
#define NAME_NTH(N) (0xffff0000 | N)
#define IS_NAME_NTH(X) ((X & 0xffff0000) == 0xffff0000)
#define GET_NAME_NTH_IDX(X) (X & 0x0000ffff)
Currently, the kernel permits maximum name offset 0xffff.
Set the test name off as 0x0fffFFFF to trigger the kernel
verification failure.
Cc: Andrii Nakryiko <andriin@fb.com>
Fixes: 9c65112744 ("selftests/btf: add initial BTF dedup tests")
Signed-off-by: Yonghong Song <yhs@fb.com>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Currently, to get map key/value type id's, the macro
BPF_ANNOTATE_KV_PAIR(<map_name>, <key_type>, <value_type>)
needs to be defined in the bpf program for the
corresponding map.
During program/map loading time,
the local static function bpf_map_find_btf_info()
in libbpf.c is implemented to retrieve the key/value
type ids given the map name.
The patch refactored function bpf_map_find_btf_info()
to create an API btf__get_map_kv_tids() which includes
the bulk of implementation for the original function.
The API btf__get_map_kv_tids() can be used by bcc,
a JIT based bpf compilation system, which uses the
same BPF_ANNOTATE_KV_PAIR to record map key/value types.
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The following set of functions, which manipulates .BTF.ext
section, are exposed as API functions:
. btf_ext__new
. btf_ext__free
. btf_ext__reloc_func_info
. btf_ext__reloc_line_info
. btf_ext__func_info_rec_size
. btf_ext__line_info_rec_size
These functions are useful for JIT based bpf codegen, e.g.,
bcc, to manipulate in-memory .BTF.ext sections.
The signature of function btf_ext__reloc_func_info()
is also changed to be the same as its definition in btf.c.
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Currently, the libbpf API function libbpf_set_print()
takes three function pointer parameters for warning, info
and debug printout respectively.
This patch changes the API to have just one function pointer
parameter and the function pointer has one additional
parameter "debugging level". So if in the future, if
the debug level is increased, the function signature
won't change.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Currently, the btf log is allocated and printed out in case
of error at LIBBPF_DEBUG level.
Such logs from kernel are very important for debugging.
For example, bpf syscall BPF_PROG_LOAD command can get
verifier logs back to user space. In function load_program()
of libbpf.c, the log buffer is allocated unconditionally
and printed out at pr_warning() level.
Let us do the similar thing here for btf. Allocate buffer
unconditionally and print out error logs at pr_warning() level.
This can reduce one global function and
optimize for common situations where pr_warning()
is activated either by default or by user supplied
debug output function.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
A global function libbpf_print, which is invisible
outside the shared library, is defined to print based
on levels. The pr_warning, pr_info and pr_debug
macros are moved into the newly created header
common.h. So any .c file including common.h can
use these macros directly.
Currently btf__new and btf_ext__new API has an argument getting
__pr_debug function pointer into btf.c so the debugging information
can be printed there. This patch removed this parameter
from btf__new and btf_ext__new and directly using pr_debug in btf.c.
Another global function libbpf_print_level_available, also
invisible outside the shared library, can test
whether a particular level debug printing is
available or not. It is used in btf.c to
test whether DEBUG level debug printing is availabl or not,
based on which the log buffer will be allocated when loading
btf to the kernel.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Fixed one test_btf raw test such that typedef func_proto
is permitted now.
Fixes: 78a2540e89 ("tools/bpf: Add tests for BTF_KIND_FUNC_PROTO and BTF_KIND_FUNC")
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Current implementation does not allow typedef func_proto.
But it is actually allowed.
-bash-4.4$ cat t.c
typedef int (f) (int);
f *g;
-bash-4.4$ clang -O2 -g -c -target bpf t.c -Xclang -target-feature -Xclang +dwarfris
-bash-4.4$ pahole -JV t.o
File t.o:
[1] PTR (anon) type_id=2
[2] TYPEDEF f type_id=3
[3] FUNC_PROTO (anon) return=4 args=(4 (anon))
[4] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED
-bash-4.4$
This patch related btf verifier to allow such (typedef func_proto)
patterns.
Fixes: 2667a2626f ("bpf: btf: Add BTF_KIND_FUNC and BTF_KIND_FUNC_PROTO")
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
This patch added documentation for BTF (BPF Debug Format).
The document is placed under linux:Documentation/bpf directory.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The bpffs pretty print test is extended to cover int128 types.
Tested on an x64 machine.
$ test_btf -p
......
BTF pretty print array(#3)......OK
PASS:9 SKIP:0 FAIL:0
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
The test_btf pretty print is refactored in order to easily
support multiple map value formats. The next patch will
add __int128 type tests which needs macro guard __SIZEOF_INT128__.
There is no functionality change with this patch.
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Several int128 raw type tests are added to test_btf.
Currently these tests are enabled only for x64 and arm64
for which kernel has CONFIG_ARCH_SUPPORTS_INT128 set.
$ test_btf
......
BTF raw test[106] (128-bit int): OK
BTF raw test[107] (struct, 128-bit int member): OK
BTF raw test[108] (struct, 120-bit int member bitfield): OK
BTF raw test[109] (struct, kind_flag, 128-bit int member): OK
BTF raw test[110] (struct, kind_flag, 120-bit int member bitfield): OK
......
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Currently, btf only supports up to 64-bit integer.
On the other hand, 128bit support for gcc and clang
has existed for a long time. For example, both gcc 4.8
and llvm 3.7 supports types "__int128" and
"unsigned __int128" for virtually all 64bit architectures
including bpf.
The requirement for __int128 support comes from two areas:
. bpf program may use __int128. For example, some bcc tools
(https://github.com/iovisor/bcc/tree/master/tools),
mostly tcp v6 related, tcpstates.py, tcpaccept.py, etc.,
are using __int128 to represent the ipv6 addresses.
. linux itself is using __int128 types. Hence supporting
__int128 type in BTF is required for vmlinux BTF,
which will be used by "compile once and run everywhere"
and other projects.
For 128bit integer, instead of base-10, hex numbers are pretty
printed out as large decimal number is hard to decipher, e.g.,
for ipv6 addresses.
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
x86 compilation has required asm goto support since 4.17.
Since clang does not support asm goto, at 4.17,
Commit b1ae32dbab ("x86/cpufeature: Guard asm_volatile_goto usage
for BPF compilation") worked around the issue by permitting an
alternative implementation without asm goto for clang.
At 5.0, more asm goto usages appeared.
[yhs@148 x86]$ egrep -r asm_volatile_goto
include/asm/cpufeature.h: asm_volatile_goto("1: jmp 6f\n"
include/asm/jump_label.h: asm_volatile_goto("1:"
include/asm/jump_label.h: asm_volatile_goto("1:"
include/asm/rmwcc.h: asm_volatile_goto (fullop "; j" #cc " %l[cc_label]" \
include/asm/uaccess.h: asm_volatile_goto("\n" \
include/asm/uaccess.h: asm_volatile_goto("\n" \
[yhs@148 x86]$
Compiling samples/bpf directories, most bpf programs failed
compilation with error messages like:
In file included from /home/yhs/work/bpf-next/samples/bpf/xdp_sample_pkts_kern.c:2:
In file included from /home/yhs/work/bpf-next/include/linux/ptrace.h:6:
In file included from /home/yhs/work/bpf-next/include/linux/sched.h:15:
In file included from /home/yhs/work/bpf-next/include/linux/sem.h:5:
In file included from /home/yhs/work/bpf-next/include/uapi/linux/sem.h:5:
In file included from /home/yhs/work/bpf-next/include/linux/ipc.h:9:
In file included from /home/yhs/work/bpf-next/include/linux/refcount.h:72:
/home/yhs/work/bpf-next/arch/x86/include/asm/refcount.h:70:9: error: 'asm goto' constructs are not supported yet
return GEN_BINARY_SUFFIXED_RMWcc(LOCK_PREFIX "subl",
^
/home/yhs/work/bpf-next/arch/x86/include/asm/rmwcc.h:67:2: note: expanded from macro 'GEN_BINARY_SUFFIXED_RMWcc'
__GEN_RMWcc(op " %[val], %[var]\n\t" suffix, var, cc, \
^
/home/yhs/work/bpf-next/arch/x86/include/asm/rmwcc.h:21:2: note: expanded from macro '__GEN_RMWcc'
asm_volatile_goto (fullop "; j" #cc " %l[cc_label]" \
^
/home/yhs/work/bpf-next/include/linux/compiler_types.h:188:37: note: expanded from macro 'asm_volatile_goto'
#define asm_volatile_goto(x...) asm goto(x)
Most implementation does not even provide an alternative
implementation. And it is also not practical to make changes
for each call site.
This patch workarounded the asm goto issue by redefining the macro like below:
#define asm_volatile_goto(x...) asm volatile("invalid use of asm_volatile_goto")
If asm_volatile_goto is not used by bpf programs, which is typically the case, nothing bad
will happen. If asm_volatile_goto is used by bpf programs, which is incorrect, the compiler
will issue an error since "invalid use of asm_volatile_goto" is not valid assembly codes.
With this patch, all bpf programs under samples/bpf can pass compilation.
Note that bpf programs under tools/testing/selftests/bpf/ compiled fine as
they do not access kernel internal headers.
Fixes: e769742d35 ("Revert "x86/jump-labels: Macrofy inline assembly code to work around GCC inlining bugs"")
Fixes: 18fe58229d ("x86, asm: change the GEN_*_RMWcc() macros to not quote the condition")
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Commit 8772c8bc09 ("tools: bpftool: support pretty print
with kind_flag set") added bpftool map dump with kind_flag
support. When bitfield_size can be retrieved directly from
btf_member, function btf_dumper_bitfield() is called to
dump the bitfield. The implementation passed the
wrong parameter "bit_offset" to the function. The excepted
value is the bit_offset within a byte while the passed-in
value is the struct member offset.
This commit fixed the bug with passing correct "bit_offset"
with adjusted data pointer.
Fixes: 8772c8bc09 ("tools: bpftool: support pretty print with kind_flag set")
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This patch modified test_btf pretty print test to cover
the bitfield with struct member equal to or greater 256.
Without the previous kernel patch fix, the modified test will fail:
$ test_btf -p
......
BTF pretty print array(#1)......unexpected pprint output
expected: 0: {0,0,0,0x3,0x0,0x3,{0|[0,0,0,0,0,0,0,0]},ENUM_ZERO,4,0x1}
read: 0: {0,0,0,0x3,0x0,0x3,{0|[0,0,0,0,0,0,0,0]},ENUM_ZERO,4,0x0}
BTF pretty print array(#2)......unexpected pprint output
expected: 0: {0,0,0,0x3,0x0,0x3,{0|[0,0,0,0,0,0,0,0]},ENUM_ZERO,4,0x1}
read: 0: {0,0,0,0x3,0x0,0x3,{0|[0,0,0,0,0,0,0,0]},ENUM_ZERO,4,0x0}
PASS:6 SKIP:0 FAIL:2
With the kernel fix, the modified test will succeed:
$ test_btf -p
......
BTF pretty print array(#1)......OK
BTF pretty print array(#2)......OK
PASS:8 SKIP:0 FAIL:0
Fixes: 9d5f9f701b ("bpf: btf: fix struct/union/fwd types with kind_flag")
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Commit 9d5f9f701b ("bpf: btf: fix struct/union/fwd types
with kind_flag") introduced kind_flag and used bitfield_size
in the btf_member to directly pretty print member values.
The commit contained a bug where the incorrect parameters could be
passed to function btf_bitfield_seq_show(). The bits_offset
parameter in the function expects a value less than 8.
Instead, the member offset in the structure is passed.
The below is btf_bitfield_seq_show() func signature:
void btf_bitfield_seq_show(void *data, u8 bits_offset,
u8 nr_bits, struct seq_file *m)
both bits_offset and nr_bits are u8 type. If the bitfield
member offset is greater than 256, incorrect value will
be printed.
This patch fixed the issue by calculating correct proper
data offset and bits_offset similar to non kind_flag case.
Fixes: 9d5f9f701b ("bpf: btf: fix struct/union/fwd types with kind_flag")
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Current btf internal verbose logger logs fwd type as
[2] FWD A type_id=0
where A is the type name.
Commit 9d5f9f701b ("bpf: btf: fix struct/union/fwd types
with kind_flag") introduced kind_flag which can be used
to distinguish whether a forward type is a struct or
union.
Also, "type_id=0" does not carry any meaningful
information for fwd type as btf_type.type = 0 is simply
enforced during btf verification and is not used
anywhere else.
This commit changed the log to
[2] FWD A struct
if kind_flag = 0, or
[2] FWD A union
if kind_flag = 1.
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Current btf func_info, line_info and jited_line are designed to be
extensible. The record sizes for {func,line}_info are passed to kernel,
and the record sizes for {func,line,jited_line}_info are returned to
userspace during bpf_prog_info query.
In bpf selftests test_btf.c, when testing whether kernel returns
a legitimate {func,line, jited_line)_info rec_size, the test only
compares to the minimum allowed size. If the returned rec_size is smaller
than the minimum allowed size, it is considered incorrect.
The minimum allowed size for these three info sizes are equal to
current value of sizeof(struct bpf_func_info), sizeof(struct bpf_line_info)
and sizeof(__u64).
The original thinking was that in the future when rec_size is increased
in kernel, the same test should run correctly. But this sacrificed
the precision of testing under the very kernel the test is shipped with,
and bpf selftest is typically run with the same repo kernel.
So this patch changed the testing of rec_size such that the
kernel returned value should be equal to the size defined by
tools uapi header bpf.h which syncs with kernel uapi header.
Martin discovered a bug in one of rec_size comparisons.
Instead of comparing to minimum func_info rec_size 8, it compares to 4.
This patch fixed that issue as well.
Fixes: 999d82cbc0 ("tools/bpf: enhance test_btf file testing to test func info")
Fixes: 05687352c6 ("bpf: Refactor and bug fix in test_func_type in test_btf.c")
Fixes: 4d6304c763 ("bpf: Add unit tests for bpf_line_info")
Suggested-by: Martin KaFai Lau <kafai@fb.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
The following example shows map pretty print with structures
which include bitfield members.
enum A { A1, A2, A3, A4, A5 };
typedef enum A ___A;
struct tmp_t {
char a1:4;
int a2:4;
int :4;
__u32 a3:4;
int b;
___A b1:4;
enum A b2:4;
};
struct bpf_map_def SEC("maps") tmpmap = {
.type = BPF_MAP_TYPE_ARRAY,
.key_size = sizeof(__u32),
.value_size = sizeof(struct tmp_t),
.max_entries = 1,
};
BPF_ANNOTATE_KV_PAIR(tmpmap, int, struct tmp_t);
and the following map update in the bpf program:
key = 0;
struct tmp_t t = {};
t.a1 = 2;
t.a2 = 4;
t.a3 = 6;
t.b = 7;
t.b1 = 8;
t.b2 = 10;
bpf_map_update_elem(&tmpmap, &key, &t, 0);
With this patch, I am able to print out the map values
correctly with this patch:
bpftool map dump id 187
[{
"key": 0,
"value": {
"a1": 0x2,
"a2": 0x4,
"a3": 0x6,
"b": 7,
"b1": 0x8,
"b2": 0xa
}
}
]
Previously, if a function prototype argument has a typedef
type, the prototype is not printed since
function __btf_dumper_type_only() bailed out with error
if the type is a typedef. This commit corrected this
behavior by printing out typedef properly.
The following example shows forward type and
typedef type can be properly printed in function prototype
with modified test_btf_haskv.c.
struct t;
union u;
__attribute__((noinline))
static int test_long_fname_1(struct dummy_tracepoint_args *arg,
struct t *p1, union u *p2,
__u32 unused)
...
int _dummy_tracepoint(struct dummy_tracepoint_args *arg) {
return test_long_fname_1(arg, 0, 0, 0);
}
$ bpftool p d xlated id 24
...
int test_long_fname_1(struct dummy_tracepoint_args * arg,
struct t * p1, union u * p2,
__u32 unused)
...
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
The core dump funcitonality in btf_dumper_int_bits() is
refactored into a separate function btf_dumper_bitfield()
which will be used by the next patch.
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
The new tests are added to test bpffs map pretty print in kernel with kind_flag
for structure type.
$ test_btf -p
......
BTF pretty print array(#1)......OK
BTF pretty print array(#2)......OK
PASS:8 SKIP:0 FAIL:0
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This patch added unit tests for different types handling
type->info.kind_flag. The following new tests are added:
$ test_btf
...
BTF raw test[82] (invalid int kind_flag): OK
BTF raw test[83] (invalid ptr kind_flag): OK
BTF raw test[84] (invalid array kind_flag): OK
BTF raw test[85] (invalid enum kind_flag): OK
BTF raw test[86] (valid fwd kind_flag): OK
BTF raw test[87] (invalid typedef kind_flag): OK
BTF raw test[88] (invalid volatile kind_flag): OK
BTF raw test[89] (invalid const kind_flag): OK
BTF raw test[90] (invalid restrict kind_flag): OK
BTF raw test[91] (invalid func kind_flag): OK
BTF raw test[92] (invalid func_proto kind_flag): OK
BTF raw test[93] (valid struct kind_flag, bitfield_size = 0): OK
BTF raw test[94] (valid struct kind_flag, int member, bitfield_size != 0): OK
BTF raw test[95] (valid union kind_flag, int member, bitfield_size != 0): OK
BTF raw test[96] (valid struct kind_flag, enum member, bitfield_size != 0): OK
BTF raw test[97] (valid union kind_flag, enum member, bitfield_size != 0): OK
BTF raw test[98] (valid struct kind_flag, typedef member, bitfield_size != 0): OK
BTF raw test[99] (valid union kind_flag, typedef member, bitfield_size != 0): OK
BTF raw test[100] (invalid struct type, bitfield_size greater than struct size): OK
BTF raw test[101] (invalid struct type, kind_flag bitfield base_type int not regular): OK
BTF raw test[102] (invalid struct type, kind_flag base_type int not regular): OK
BTF raw test[103] (invalid union type, bitfield_size greater than struct size): OK
...
PASS:122 SKIP:0 FAIL:0
The second parameter name of macro
BTF_INFO_ENC(kind, root, vlen)
in selftests test_btf.c is also renamed from "root" to "kind_flag".
Note that before this patch "root" is not used and always 0.
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Sync include/uapi/linux/btf.h to tools/include/uapi/linux/btf.h.
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Commit 970289fc0a83 ("bpf: add bpffs pretty print for cgroup
local storage maps") added bpffs pretty print for cgroup
local storage maps. The commit worked for struct without kind_flag
set.
This patch refactored and made pretty print also work
with kind_flag set for the struct.
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This patch fixed two issues with BTF. One is related to
struct/union bitfield encoding and the other is related to
forward type.
Issue #1 and solution:
======================
Current btf encoding of bitfield follows what pahole generates.
For each bitfield, pahole will duplicate the type chain and
put the bitfield size at the final int or enum type.
Since the BTF enum type cannot encode bit size,
pahole workarounds the issue by generating
an int type whenever the enum bit size is not 32.
For example,
-bash-4.4$ cat t.c
typedef int ___int;
enum A { A1, A2, A3 };
struct t {
int a[5];
___int b:4;
volatile enum A c:4;
} g;
-bash-4.4$ gcc -c -O2 -g t.c
The current kernel supports the following BTF encoding:
$ pahole -JV t.o
[1] TYPEDEF ___int type_id=2
[2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED
[3] ENUM A size=4 vlen=3
A1 val=0
A2 val=1
A3 val=2
[4] STRUCT t size=24 vlen=3
a type_id=5 bits_offset=0
b type_id=9 bits_offset=160
c type_id=11 bits_offset=164
[5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5
[6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none)
[7] VOLATILE (anon) type_id=3
[8] INT int size=1 bit_offset=0 nr_bits=4 encoding=(none)
[9] TYPEDEF ___int type_id=8
[10] INT (anon) size=1 bit_offset=0 nr_bits=4 encoding=SIGNED
[11] VOLATILE (anon) type_id=10
Two issues are in the above:
. by changing enum type to int, we lost the original
type information and this will not be ideal later
when we try to convert BTF to a header file.
. the type duplication for bitfields will cause
BTF bloat. Duplicated types cannot be deduplicated
later if the bitfield size is different.
To fix this issue, this patch implemented a compatible
change for BTF struct type encoding:
. the bit 31 of struct_type->info, previously reserved,
now is used to indicate whether bitfield_size is
encoded in btf_member or not.
. if bit 31 of struct_type->info is set,
btf_member->offset will encode like:
bit 0 - 23: bit offset
bit 24 - 31: bitfield size
if bit 31 is not set, the old behavior is preserved:
bit 0 - 31: bit offset
So if the struct contains a bit field, the maximum bit offset
will be reduced to (2^24 - 1) instead of MAX_UINT. The maximum
bitfield size will be 256 which is enough for today as maximum
bitfield in compiler can be 128 where int128 type is supported.
This kernel patch intends to support the new BTF encoding:
$ pahole -JV t.o
[1] TYPEDEF ___int type_id=2
[2] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED
[3] ENUM A size=4 vlen=3
A1 val=0
A2 val=1
A3 val=2
[4] STRUCT t kind_flag=1 size=24 vlen=3
a type_id=5 bitfield_size=0 bits_offset=0
b type_id=1 bitfield_size=4 bits_offset=160
c type_id=7 bitfield_size=4 bits_offset=164
[5] ARRAY (anon) type_id=2 index_type_id=2 nr_elems=5
[6] INT sizetype size=8 bit_offset=0 nr_bits=64 encoding=(none)
[7] VOLATILE (anon) type_id=3
Issue #2 and solution:
======================
Current forward type in BTF does not specify whether the original
type is struct or union. This will not work for type pretty print
and BTF-to-header-file conversion as struct/union must be specified.
$ cat tt.c
struct t;
union u;
int foo(struct t *t, union u *u) { return 0; }
$ gcc -c -g -O2 tt.c
$ pahole -JV tt.o
[1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED
[2] FWD t type_id=0
[3] PTR (anon) type_id=2
[4] FWD u type_id=0
[5] PTR (anon) type_id=4
To fix this issue, similar to issue #1, type->info bit 31
is used. If the bit is set, it is union type. Otherwise, it is
a struct type.
$ pahole -JV tt.o
[1] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED
[2] FWD t kind_flag=0 type_id=0
[3] PTR (anon) kind_flag=0 type_id=2
[4] FWD u kind_flag=1 type_id=0
[5] PTR (anon) kind_flag=0 type_id=4
Pahole/LLVM change:
===================
The new kind_flag functionality has been implemented in pahole
and llvm:
https://github.com/yonghong-song/pahole/tree/bitfieldhttps://github.com/yonghong-song/llvm/tree/bitfield
Note that pahole hasn't implemented func/func_proto kind
and .BTF.ext. So to print function signature with bpftool,
the llvm compiler should be used.
Fixes: 69b693f0ae ("bpf: btf: Introduce BPF Type Format (BTF)")
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Refactor function btf_int_bits_seq_show() by creating
function btf_bitfield_seq_show() which has no dependence
on btf and btf_type. The function btf_bitfield_seq_show()
will be in later patch to directly dump bitfield member values.
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Rename all occurances of *_info_cnt field access
to nr_*_info in tools directory.
The local variables finfo_cnt, linfo_cnt and jited_linfo_cnt
in function do_dump() of tools/bpf/bpftool/prog.c are also
changed to nr_finfo, nr_linfo and nr_jited_linfo to
keep naming convention consistent.
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
In uapi bpf.h, currently we have the following fields in
the struct bpf_prog_info:
__u32 func_info_cnt;
__u32 line_info_cnt;
__u32 jited_line_info_cnt;
The above field names "func_info_cnt" and "line_info_cnt"
also appear in union bpf_attr for program loading.
The original intention is to keep the names the same
between bpf_prog_info and bpf_attr
so it will imply what we returned to user space will be
the same as what the user space passed to the kernel.
Such a naming convention in bpf_prog_info is not consistent
with other fields like:
__u32 nr_jited_ksyms;
__u32 nr_jited_func_lens;
This patch made this adjustment so in bpf_prog_info
newly introduced *_info_cnt becomes nr_*_info.
Acked-by: Song Liu <songliubraving@fb.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
During porting libbpf to bcc, I got some warnings like below:
...
[ 2%] Building C object src/cc/CMakeFiles/bpf-shared.dir/libbpf/src/libbpf.c.o
/home/yhs/work/bcc2/src/cc/libbpf/src/libbpf.c:12:0:
warning: "_GNU_SOURCE" redefined [enabled by default]
#define _GNU_SOURCE
...
[ 3%] Building C object src/cc/CMakeFiles/bpf-shared.dir/libbpf/src/libbpf_errno.c.o
/home/yhs/work/bcc2/src/cc/libbpf/src/libbpf_errno.c: In function ‘libbpf_strerror’:
/home/yhs/work/bcc2/src/cc/libbpf/src/libbpf_errno.c:45:7:
warning: assignment makes integer from pointer without a cast [enabled by default]
ret = strerror_r(err, buf, size);
...
bcc is built with _GNU_SOURCE defined and this caused the above warning.
This patch intends to make libpf _GNU_SOURCE friendly by
. define _GNU_SOURCE in libbpf.c unless it is not defined
. undefine _GNU_SOURCE as non-gnu version of strerror_r is expected.
Signed-off-by: Yonghong Song <yhs@fb.com>
Acked-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Commit b12d6ec097 ("bpf: btf: add btf print functionality")
added btf pretty print functionality to bpftool.
There is a problem though in printing a bitfield whose type
has modifiers.
For example, for a type like
typedef int ___int;
struct tmp_t {
int a:3;
___int b:3;
};
Suppose we have a map
struct bpf_map_def SEC("maps") tmpmap = {
.type = BPF_MAP_TYPE_HASH,
.key_size = sizeof(__u32),
.value_size = sizeof(struct tmp_t),
.max_entries = 1,
};
and the hash table is populated with one element with
key 0 and value (.a = 1 and .b = 2).
In BTF, the struct member "b" will have a type "typedef" which
points to an int type. The current implementation does not
pass the bit offset during transition from typedef to int type,
hence incorrectly print the value as
$ bpftool m d id 79
[{
"key": 0,
"value": {
"a": 0x1,
"b": 0x1
}
}
]
This patch fixed the issue by carrying bit_offset along the type
chain during bit_field print. The correct result can be printed as
$ bpftool m d id 76
[{
"key": 0,
"value": {
"a": 0x1,
"b": 0x2
}
}
]
The kernel pretty print is implemented correctly and does not
have this issue.
Fixes: b12d6ec097 ("bpf: btf: add btf print functionality")
Signed-off-by: Yonghong Song <yhs@fb.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The following additional unit testcases are added to test_btf:
...
BTF raw test[42] (typedef (invalid name, name_off = 0)): OK
BTF raw test[43] (typedef (invalid name, invalid identifier)): OK
BTF raw test[44] (ptr type (invalid name, name_off <> 0)): OK
BTF raw test[45] (volatile type (invalid name, name_off <> 0)): OK
BTF raw test[46] (const type (invalid name, name_off <> 0)): OK
BTF raw test[47] (restrict type (invalid name, name_off <> 0)): OK
BTF raw test[48] (fwd type (invalid name, name_off = 0)): OK
BTF raw test[49] (fwd type (invalid name, invalid identifier)): OK
BTF raw test[50] (array type (invalid name, name_off <> 0)): OK
BTF raw test[51] (struct type (name_off = 0)): OK
BTF raw test[52] (struct type (invalid name, invalid identifier)): OK
BTF raw test[53] (struct member (name_off = 0)): OK
BTF raw test[54] (struct member (invalid name, invalid identifier)): OK
BTF raw test[55] (enum type (name_off = 0)): OK
BTF raw test[56] (enum type (invalid name, invalid identifier)): OK
BTF raw test[57] (enum member (invalid name, name_off = 0)): OK
BTF raw test[58] (enum member (invalid name, invalid identifier)): OK
...
Fixes: c0fa1b6c3e ("bpf: btf: Add BTF tests")
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
This patch added name checking for the following types:
. BTF_KIND_PTR, BTF_KIND_ARRAY, BTF_KIND_VOLATILE,
BTF_KIND_CONST, BTF_KIND_RESTRICT:
the name must be null
. BTF_KIND_STRUCT, BTF_KIND_UNION: the struct/member name
is either null or a valid identifier
. BTF_KIND_ENUM: the enum type name is either null or a valid
identifier; the enumerator name must be a valid identifier.
. BTF_KIND_FWD: the name must be a valid identifier
. BTF_KIND_TYPEDEF: the name must be a valid identifier
For those places a valid name is required, the name must be
a valid C identifier. This can be relaxed later if we found
use cases for a different (non-C) frontend.
Fixes: 69b693f0ae ("bpf: btf: Introduce BPF Type Format (BTF)")
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Function btf_name_valid_identifier() have been implemented in
bpf-next commit 2667a2626f ("bpf: btf: Add BTF_KIND_FUNC and
BTF_KIND_FUNC_PROTO"). Backport this function so later patch
can use it.
Fixes: 69b693f0ae ("bpf: btf: Introduce BPF Type Format (BTF)")
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The selftest test_btf is changed to test both jit and non-jit.
The test result should be the same regardless of whether jit
is enabled or not.
Signed-off-by: Yonghong Song <yhs@fb.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Commit 838e96904f ("bpf: Introduce bpf_func_info")
added bpf func info support. The userspace is able
to get better ksym's for bpf programs with jit, and
is able to print out func prototypes.
For a program containing func-to-func calls, the existing
implementation returns user specified number of function
calls and BTF types if jit is enabled. If the jit is not
enabled, it only returns the type for the main function.
This is undesirable. Interpreter may still be used
and we should keep feature identical regardless of
whether jit is enabled or not.
This patch fixed this discrepancy.
Fixes: 838e96904f ("bpf: Introduce bpf_func_info")
Signed-off-by: Yonghong Song <yhs@fb.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Commit 2993e0515b ("tools/bpf: add support to read .BTF.ext sections")
added support to read .BTF.ext sections from an object file, create
and pass prog_btf_fd and func_info to the kernel.
The program btf_fd (prog->btf_fd) is initialized to be -1 to please
zclose so we do not need special handling dur prog close.
Passing -1 to the kernel, however, will cause loading error.
Passing btf_fd 0 to the kernel if prog->btf_fd is invalid
fixed the problem.
Fixes: 2993e0515b ("tools/bpf: add support to read .BTF.ext sections")
Reported-by: Andrey Ignatov <rdna@fb.com>
Reported-by: Emre Cantimur <haydum@fb.com>
Tested-by: Andrey Ignatov <rdna@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Kernel test robot (lkp@intel.com) reports a compilation error at
https://www.spinics.net/lists/netdev/msg534913.html
introduced by commit 838e96904f ("bpf: Introduce bpf_func_info").
If CONFIG_BPF is defined and CONFIG_BPF_SYSCALL is not defined,
the following error will appear:
kernel/bpf/core.c:414: undefined reference to `btf_type_by_id'
kernel/bpf/core.c:415: undefined reference to `btf_name_by_offset'
When CONFIG_BPF_SYSCALL is not defined,
let us define stub inline functions for btf_type_by_id()
and btf_name_by_offset() in include/linux/btf.h.
This way, the compilation failure can be avoided.
Fixes: 838e96904f ("bpf: Introduce bpf_func_info")
Reported-by: kbuild test robot <lkp@intel.com>
Cc: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Change the bpf programs test_btf_haskv.c and test_btf_nokv.c to
have two sections, and enhance test_btf.c test_file feature
to test btf func_info returned by the kernel.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The function get_btf() is implemented in tools/bpf/bpftool/map.c
to get a btf structure given a map_info. This patch
refactored this function to be function btf_get_from_id()
in tools/lib/bpf so that it can be used later.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Add additional checks in tools/testing/selftests/bpf and
samples/bpf such that if clang/llvm compiler can generate
BTF sections, do not use pahole.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The .BTF section is already available to encode types.
These types can be used for map
pretty print. The whole .BTF will be passed to the
kernel as well for which kernel can verify and return
to the user space for pretty print etc.
The llvm patch at https://reviews.llvm.org/D53736
will generate .BTF section and one more section .BTF.ext.
The .BTF.ext section encodes function type
information and line information. Note that
this patch set only supports function type info.
The functionality is implemented in libbpf.
The .BTF section can be directly loaded into the
kernel, and the .BTF.ext section cannot. The loader
may need to do some relocation and merging,
similar to merging multiple code sections, before
loading into the kernel.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
A two function bpf program is loaded with btf and func_info.
After successful prog load, the bpf_get_info syscall is called
to retrieve prog info to ensure the types returned from the
kernel matches the types passed to the kernel from the
user space.
Several negative tests are also added to test loading/retriving
of func_type info.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The new fields are added for program load in lib/bpf so
application uses api bpf_load_program_xattr() is able
to load program with btf and func_info data.
This functionality will be used in next patch
by bpf selftest test_btf.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The kernel uapi bpf.h is synced to tools directory.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
This patch added interface to load a program with the following
additional information:
. prog_btf_fd
. func_info, func_info_rec_size and func_info_cnt
where func_info will provide function range and type_id
corresponding to each function.
The func_info_rec_size is introduced in the UAPI to specify
struct bpf_func_info size passed from user space. This
intends to make bpf_func_info structure growable in the future.
If the kernel gets a different bpf_func_info size from userspace,
it will try to handle user request with part of bpf_func_info
it can understand. In this patch, kernel can understand
struct bpf_func_info {
__u32 insn_offset;
__u32 type_id;
};
If user passed a bpf func_info record size of 16 bytes, the
kernel can still handle part of records with the above definition.
If verifier agrees with function range provided by the user,
the bpf_prog ksym for each function will use the func name
provided in the type_id, which is supposed to provide better
encoding as it is not limited by 16 bytes program name
limitation and this is better for bpf program which contains
multiple subprograms.
The bpf_prog_info interface is also extended to
return btf_id, func_info, func_info_rec_size and func_info_cnt
to userspace, so userspace can print out the function prototype
for each xlated function. The insn_offset in the returned
func_info corresponds to the insn offset for xlated functions.
With other jit related fields in bpf_prog_info, userspace can also
print out function prototypes for each jited function.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Commit f6f3bac08f ("tools/bpf: bpftool: add net support")
added certain networking support to bpftool.
The implementation relies on a relatively recent uapi header file
linux/tc_act/tc_bpf.h on the host which contains the marco
definition of TCA_ACT_BPF_ID.
Unfortunately, this is not the case for all distributions.
See the email message below where rhel-7.2 does not have
an up-to-date linux/tc_act/tc_bpf.h.
https://www.mail-archive.com/linux-kernel@vger.kernel.org/msg1799211.html
Further investigation found that linux/pkt_cls.h is also needed for macro
TCA_BPF_TAG.
This patch fixed the issue by copying linux/tc_act/tc_bpf.h
and linux/pkt_cls.h from kernel include/uapi directory to
tools/include/uapi directory so building the bpftool does not depend
on host system for these files.
Fixes: f6f3bac08f ("tools/bpf: bpftool: add net support")
Reported-by: kernel test robot <rong.a.chen@intel.com>
Cc: Li Zhijian <zhijianx.li@intel.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
On our test machine, bpf selftest test_flow_dissector.sh failed
with the following error:
# ./test_flow_dissector.sh
bpffs not mounted. Mounting...
libbpf: failed to create map (name: 'jmp_table'): Operation not permitted
libbpf: failed to load object 'bpf_flow.o'
./flow_dissector_load: bpf_prog_load bpf_flow.o
selftests: test_flow_dissector [FAILED]
Let us increase the rlimit to remove the above map
creation failure.
Signed-off-by: Yonghong Song <yhs@fb.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Use __u32 instead u32 in libbpf.c and also use
uapi perf_event.h instead of tools/perf/perf-sys.h.
Signed-off-by: Yonghong Song <yhs@fb.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Currently, helper bpf_get_current_cgroup_id() is not permitted
for CGROUP_DEVICE type of programs. If the helper is used
in such cases, the verifier will log the following error:
0: (bf) r6 = r1
1: (69) r7 = *(u16 *)(r6 +0)
2: (85) call bpf_get_current_cgroup_id#80
unknown func bpf_get_current_cgroup_id#80
The bpf_get_current_cgroup_id() is useful for CGROUP_DEVICE
type of programs in order to customize action based on cgroup id.
This patch added such a support.
Cc: Roman Gushchin <guro@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Roman Gushchin <guro@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
samples/bpf build failed with the following errors:
$ make samples/bpf/
...
HOSTCC samples/bpf/sockex3_user.o
/data/users/yhs/work/net-next/samples/bpf/sockex3_user.c:16:8: error: redefinition of ‘struct bpf_flow_keys’
struct bpf_flow_keys {
^
In file included from /data/users/yhs/work/net-next/samples/bpf/sockex3_user.c:4:0:
./usr/include/linux/bpf.h:2338:9: note: originally defined here
struct bpf_flow_keys *flow_keys;
^
make[3]: *** [samples/bpf/sockex3_user.o] Error 1
Commit d58e468b11 ("flow_dissector: implements flow dissector BPF hook")
introduced struct bpf_flow_keys in include/uapi/linux/bpf.h and hence
caused the naming conflict with samples/bpf/sockex3_user.c.
The fix is to rename struct bpf_flow_keys in samples/bpf/sockex3_user.c
to flow_keys to avoid the conflict.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This is a followup patch for Commit f6f3bac08f
("tools/bpf: bpftool: add net support").
Some improvements are made for the bpftool net output.
Specially, plain output is more concise such that
per attachment should nicely fit in one line.
Compared to previous output, the prog tag is removed
since it can be easily obtained with program id.
Similar to xdp attachments, the device name is added
to tc attachments.
The bpf program attached through shared block
mechanism is supported as well.
$ ip link add dev v1 type veth peer name v2
$ tc qdisc add dev v1 ingress_block 10 egress_block 20 clsact
$ tc qdisc add dev v2 ingress_block 10 egress_block 20 clsact
$ tc filter add block 10 protocol ip prio 25 bpf obj bpf_shared.o sec ingress flowid 1:1
$ tc filter add block 20 protocol ip prio 30 bpf obj bpf_cyclic.o sec classifier flowid 1:1
$ bpftool net
xdp:
tc:
v2(7) clsact/ingress bpf_shared.o:[ingress] id 23
v2(7) clsact/egress bpf_cyclic.o:[classifier] id 24
v1(8) clsact/ingress bpf_shared.o:[ingress] id 23
v1(8) clsact/egress bpf_cyclic.o:[classifier] id 24
The documentation and "bpftool net help" are updated
to make it clear that current implementation only
supports xdp and tc attachments. For programs
attached to cgroups, "bpftool cgroup" can be used
to dump attachments. For other programs e.g.
sk_{filter,skb,msg,reuseport} and lwt/seg6,
iproute2 tools should be used.
The new output:
$ bpftool net
xdp:
eth0(2) driver id 198
tc:
eth0(2) clsact/ingress fbflow_icmp id 335 act [{icmp_action id 336}]
eth0(2) clsact/egress fbflow_egress id 334
$ bpftool -jp net
[{
"xdp": [{
"devname": "eth0",
"ifindex": 2,
"mode": "driver",
"id": 198
}
],
"tc": [{
"devname": "eth0",
"ifindex": 2,
"kind": "clsact/ingress",
"name": "fbflow_icmp",
"id": 335,
"act": [{
"name": "icmp_action",
"id": 336
}
]
},{
"devname": "eth0",
"ifindex": 2,
"kind": "clsact/egress",
"name": "fbflow_egress",
"id": 334
}
]
}
]
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Commit f7010770fb ("tools/bpf: move bpf/lib netlink related
functions into a new file") introduced a while loop for the
netlink recv path. This while loop is needed since the
buffer in recv syscall may not be enough to hold all the
information and in such cases multiple recv calls are needed.
There is a bug introduced by the above commit as
the while loop may block on recv syscall if there is no
more messages are expected. The netlink message header
flag NLM_F_MULTI is used to indicate that more messages
are expected and this patch fixed the bug by doing
further recv syscall only if multipart message is expected.
The patch added another fix regarding to message length of 0.
When netlink recv returns message length of 0, there will be
no more messages for returning data so the while loop
can end.
Fixes: f7010770fb ("tools/bpf: move bpf/lib netlink related functions into a new file")
Reported-by: Björn Töpel <bjorn.topel@intel.com>
Tested-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Currently, prog array map and map of maps are not supported
in bpftool. This patch added the support.
Different from other map types, for prog array map and
map of maps, the key returned bpf_get_next_key() may not
point to a valid value. So for these two map types,
no error will be printed out when such a scenario happens.
The following is the plain and json dump if btf is not available:
$ ./bpftool map dump id 10
key: 08 00 00 00 value: 5c 01 00 00
Found 1 element
$ ./bpftool -jp map dump id 10
[{
"key": ["0x08","0x00","0x00","0x00"
],
"value": ["0x5c","0x01","0x00","0x00"
]
}]
If the BTF is available, the dump looks below:
$ ./bpftool map dump id 2
[{
"key": 0,
"value": 7
}
]
$ ./bpftool -jp map dump id 2
[{
"key": ["0x00","0x00","0x00","0x00"
],
"value": ["0x07","0x00","0x00","0x00"
],
"formatted": {
"key": 0,
"value": 7
}
}]
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Added bpffs pretty print for program array map. For a particular
array index, if the program array points to a valid program,
the "<index>: <prog_id>" will be printed out like
0: 6
which means bpf program with id "6" is installed at index "0".
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Add "bpftool net" support. Networking devices are enumerated
to dump device index/name associated with xdp progs.
For each networking device, tc classes and qdiscs are enumerated
in order to check their bpf filters.
In addition, root handle and clsact ingress/egress are also checked for
bpf filters. Not all filter information is printed out. Only ifindex,
kind, filter name, prog_id and tag are printed out, which are good
enough to show attachment information. If the filter action
is a bpf action, its bpf program id, bpf name and tag will be
printed out as well.
For example,
$ ./bpftool net
xdp [
ifindex 2 devname eth0 prog_id 198
]
tc_filters [
ifindex 2 kind qdisc_htb name prefix_matcher.o:[cls_prefix_matcher_htb]
prog_id 111727 tag d08fe3b4319bc2fd act []
ifindex 2 kind qdisc_clsact_ingress name fbflow_icmp
prog_id 130246 tag 3f265c7f26db62c9 act []
ifindex 2 kind qdisc_clsact_egress name prefix_matcher.o:[cls_prefix_matcher_clsact]
prog_id 111726 tag 99a197826974c876
ifindex 2 kind qdisc_clsact_egress name cls_fg_dscp
prog_id 108619 tag dc4630674fd72dcc act []
ifindex 2 kind qdisc_clsact_egress name fbflow_egress
prog_id 130245 tag 72d2d830d6888d2c
]
$ ./bpftool -jp net
[{
"xdp": [{
"ifindex": 2,
"devname": "eth0",
"prog_id": 198
}
],
"tc_filters": [{
"ifindex": 2,
"kind": "qdisc_htb",
"name": "prefix_matcher.o:[cls_prefix_matcher_htb]",
"prog_id": 111727,
"tag": "d08fe3b4319bc2fd",
"act": []
},{
"ifindex": 2,
"kind": "qdisc_clsact_ingress",
"name": "fbflow_icmp",
"prog_id": 130246,
"tag": "3f265c7f26db62c9",
"act": []
},{
"ifindex": 2,
"kind": "qdisc_clsact_egress",
"name": "prefix_matcher.o:[cls_prefix_matcher_clsact]",
"prog_id": 111726,
"tag": "99a197826974c876"
},{
"ifindex": 2,
"kind": "qdisc_clsact_egress",
"name": "cls_fg_dscp",
"prog_id": 108619,
"tag": "dc4630674fd72dcc",
"act": []
},{
"ifindex": 2,
"kind": "qdisc_clsact_egress",
"name": "fbflow_egress",
"prog_id": 130245,
"tag": "72d2d830d6888d2c"
}
]
}
]
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
This patch added a few netlink attribute parsing functions
and the netlink API functions to query networking links, tc classes,
tc qdiscs and tc filters. For example, the following API is
to get networking links:
int nl_get_link(int sock, unsigned int nl_pid,
dump_nlmsg_t dump_link_nlmsg,
void *cookie);
Note that when the API is called, the user also provided a
callback function with the following signature:
int (*dump_nlmsg_t)(void *cookie, void *msg, struct nlattr **tb);
The "cookie" is the parameter the user passed to the API and will
be available for the callback function.
The "msg" is the information about the result, e.g., ifinfomsg or
tcmsg. The "tb" is the parsed netlink attributes.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
There are no functionality change for this patch.
In the subsequent patches, more netlink related library functions
will be added and a separate file is better than cluttering bpf.c.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Among others, this header will be used later for
bpftool net support.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The bpf selftest test_btf is extended to test bpffs
percpu map pretty print for percpu array, percpu hash and
percpu lru hash.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Added bpffs pretty print for percpu arraymap, percpu hashmap
and percpu lru hashmap.
For each map <key, value> pair, the format is:
<key_value>: {
cpu0: <value_on_cpu0>
cpu1: <value_on_cpu1>
...
cpun: <value_on_cpun>
}
For example, on my VM, there are 4 cpus, and
for test_btf test in the next patch:
cat /sys/fs/bpf/pprint_test_percpu_hash
You may get:
...
43602: {
cpu0: {43602,0,-43602,0x3,0xaa52,0x3,{43602|[82,170,0,0,0,0,0,0]},ENUM_TWO}
cpu1: {43602,0,-43602,0x3,0xaa52,0x3,{43602|[82,170,0,0,0,0,0,0]},ENUM_TWO}
cpu2: {43602,0,-43602,0x3,0xaa52,0x3,{43602|[82,170,0,0,0,0,0,0]},ENUM_TWO}
cpu3: {43602,0,-43602,0x3,0xaa52,0x3,{43602|[82,170,0,0,0,0,0,0]},ENUM_TWO}
}
72847: {
cpu0: {72847,0,-72847,0x3,0x11c8f,0x3,{72847|[143,28,1,0,0,0,0,0]},ENUM_THREE}
cpu1: {72847,0,-72847,0x3,0x11c8f,0x3,{72847|[143,28,1,0,0,0,0,0]},ENUM_THREE}
cpu2: {72847,0,-72847,0x3,0x11c8f,0x3,{72847|[143,28,1,0,0,0,0,0]},ENUM_THREE}
cpu3: {72847,0,-72847,0x3,0x11c8f,0x3,{72847|[143,28,1,0,0,0,0,0]},ENUM_THREE}
}
...
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
The bpf selftest test_cgroup_storage failed in one of
our production test servers.
# sudo ./test_cgroup_storage
Failed to create map: Operation not permitted
It turns out this is due to insufficient locked memory
with system default 16KB.
Similar to other self tests, let us arm the process
with unlimited locked memory. With this change,
the test passed.
# sudo ./test_cgroup_storage
test_cgroup_storage:PASS
Fixes: 68cfa3ac6b ("selftests/bpf: add a cgroup storage test")
Cc: Roman Gushchin <guro@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Acked-by: Roman Gushchin <guro@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Commit 394e40a297 ("bpf: extend bpf_prog_array to store pointers
to the cgroup storage") refactored the bpf_prog_array_copy_core()
to accommodate new structure bpf_prog_array_item which contains
bpf_prog array itself.
In the old code, we had
perf_event_query_prog_array():
mutex_lock(...)
bpf_prog_array_copy_call():
prog = rcu_dereference_check(array, 1)->progs
bpf_prog_array_copy_core(prog, ...)
mutex_unlock(...)
With the above commit, we had
perf_event_query_prog_array():
mutex_lock(...)
bpf_prog_array_copy_call():
bpf_prog_array_copy_core(array, ...):
item = rcu_dereference(array)->items;
...
mutex_unlock(...)
The new code will trigger a lockdep rcu checking warning.
The fix is to change rcu_dereference() to rcu_dereference_check()
to prevent such a warning.
Reported-by: syzbot+6e72317008eef84a216b@syzkaller.appspotmail.com
Fixes: 394e40a297 ("bpf: extend bpf_prog_array to store pointers to the cgroup storage")
Cc: Roman Gushchin <guro@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Roman Gushchin <guro@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Pretty print tests for hash/lru_hash maps are added in test_btf.c.
The btf type blob is the same as pretty print array map test.
The test result:
$ mount -t bpf bpf /sys/fs/bpf
$ ./test_btf -p
BTF pretty print array......OK
BTF pretty print hash......OK
BTF pretty print lru hash......OK
PASS:3 SKIP:0 FAIL:0
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Commit a26ca7c982 ("bpf: btf: Add pretty print support to
the basic arraymap") added pretty print support to array map.
This patch adds pretty print for hash and lru_hash maps.
The following example shows the pretty-print result of
a pinned hashmap:
struct map_value {
int count_a;
int count_b;
};
cat /sys/fs/bpf/pinned_hash_map:
87907: {87907,87908}
57354: {37354,57355}
76625: {76625,76626}
...
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
In function map_seq_next() of kernel/bpf/inode.c,
the first key will be the "0" regardless of the map type.
This works for array. But for hash type, if it happens
key "0" is in the map, the bpffs map show will miss
some items if the key "0" is not the first element of
the first bucket.
This patch fixed the issue by guaranteeing to get
the first element, if the seq_show is just started,
by passing NULL pointer key to map_get_next_key() callback.
This way, no missing elements will occur for
bpffs hash table show even if key "0" is in the map.
Fixes: a26ca7c982 ("bpf: btf: Add pretty print support to the basic arraymap")
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
I hit the following problem when I tried to use bpftool
to dump a percpu array.
$ sudo ./bpftool map show
61: percpu_array name stub flags 0x0
key 4B value 4B max_entries 1 memlock 4096B
...
$ sudo ./bpftool map dump id 61
bpftool: malloc.c:2406: sysmalloc: Assertion
`(old_top == initial_top (av) && old_size == 0) || \
((unsigned long) (old_size) >= MINSIZE && \
prev_inuse (old_top) && \
((unsigned long) old_end & (pagesize - 1)) == 0)'
failed.
Aborted
Further debugging revealed that this is due to
miscommunication between bpftool and kernel.
For example, for the above percpu_array with value size of 4B.
The map info returned to user space has value size of 4B.
In bpftool, the values array for lookup is allocated like:
info->value_size * get_possible_cpus() = 4 * get_possible_cpus()
In kernel (kernel/bpf/syscall.c), the values array size is
rounded up to multiple of 8.
round_up(map->value_size, 8) * num_possible_cpus()
= 8 * num_possible_cpus()
So when kernel copies the values to user buffer, the kernel will
overwrite beyond user buffer boundary.
This patch fixed the issue by allocating and stepping through
percpu map value array properly in bpftool.
Fixes: 71bb428fe2 ("tools: bpf: add bpftool")
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Arnaldo reported the perf build failure with latest llvm/clang compiler
(7.0).
$ make LIBCLANGLLVM=1 -C tools/perf/
<SNIP>
CC /tmp/tmp.t53Qo38zci/tests/kmod-path.o
util/c++/clang.cpp: In function ‘std::unique_ptr<llvm::SmallVectorImpl<char> >
perf::getBPFObjectFromModule(llvm::Module*)’:
util/c++/clang.cpp:150:43: error: no matching function for call to
‘llvm::TargetMachine::addPassesToEmitFile(llvm::legacy::PassManager&,
llvm::raw_svector_ostream&, llvm::TargetMachine::CodeGenFileType)’
TargetMachine::CGFT_ObjectFile)) {
^
In file included from util/c++/clang.cpp:25:0:
/usr/local/include/llvm/Target/TargetMachine.h:254:16: note: candidate:
virtual bool llvm::TargetMachine::addPassesToEmitFile(
llvm::legacy::PassManagerBase&, llvm::raw_pwrite_stream&,
llvm::raw_pwrite_stream*, llvm::TargetMachine::CodeGenFileType, bool,
llvm::MachineModuleInfo*)
virtual bool addPassesToEmitFile(PassManagerBase &, raw_pwrite_stream &,
^~~~~~~~~~~~~~~~~~~
/usr/local/include/llvm/Target/TargetMachine.h:254:16: note:
candidate expects 6 arguments, 3 provided
mv: cannot stat '/tmp/tmp.t53Qo38zci/util/c++/.clang.o.tmp': No such file or directory
make[7]: *** [/home/acme/git/perf/tools/build/Makefile.build:101:
/tmp/tmp.t53Qo38zci/util/c++/clang.o] Error 1
make[6]: *** [/home/acme/git/perf/tools/build/Makefile.build:139: c++] Error 2
make[5]: *** [/home/acme/git/perf/tools/build/Makefile.build:139: util] Error 2
make[5]: *** Waiting for unfinished jobs....
CC /tmp/tmp.t53Qo38zci/tests/thread-map.o
The function addPassesToEmitFile signature changed in llvm 7.0 and such
a change caused the failure. This patch fixed the issue with using
proper function signatures under different compiler versions.
Reported-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Alexei Starovoitov <ast@fb.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Martin KaFai Lau <kafai@fb.com>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/r/20180616174739.1076733-1-yhs@fb.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
On one of our production test machine, when running
bpf selftest test_sockmap, I got the following error:
# sudo ./test_sockmap
libbpf: failed to create map (name: 'sock_map'): Operation not permitted
libbpf: failed to load object 'test_sockmap_kern.o'
libbpf: Can't get the 0th fd from program sk_skb1: only -1 instances
......
load_bpf_file: (-1) Operation not permitted
ERROR: (-1) load bpf failed
The error is due to not-big-enough rlimit
struct rlimit r = {10 * 1024 * 1024, RLIM_INFINITY};
The test already includes "bpf_rlimit.h", which sets current
and max rlimit to RLIM_INFINITY. Let us just use it.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Commit b04df400c3 ("tools/bpftool: add perf subcommand")
introduced bpftool subcommand perf to query bpf program
kuprobe and tracepoint attachments.
The perf subcommand will first test whether bpf subcommand
BPF_TASK_FD_QUERY is supported in kernel or not. It does it
by opening a file with argv[0] and feeds the file descriptor
and current task pid to the kernel for querying.
Such an approach won't work if the argv[0] cannot be opened
successfully in the current directory. This is especially
true when bpftool is accessible through PATH env variable.
The error below reflects the open failure for file argv[0]
at home directory.
[yhs@localhost ~]$ which bpftool
/usr/local/sbin/bpftool
[yhs@localhost ~]$ bpftool perf
Error: perf_query_support: No such file or directory
To fix the issue, let us open root directory ("/")
which exists in every linux system. With the fix, the
error message will correctly reflect the permission issue.
[yhs@localhost ~]$ which bpftool
/usr/local/sbin/bpftool
[yhs@localhost ~]$ bpftool perf
Error: perf_query_support: Operation not permitted
HINT: non root or kernel doesn't support TASK_FD_QUERY
Fixes: b04df400c3 ("tools/bpftool: add perf subcommand")
Reported-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Yonghong Song <yhs@fb.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Commit f269099a7e ("tools/bpf: add a selftest for
bpf_get_current_cgroup_id() helper") added a test
for bpf_get_current_cgroup_id() helper. The bpf program
is attached to tracepoint syscalls/sys_enter_nanosleep
and will record the cgroup id if the tracepoint is hit.
The test program creates a cgroup and attachs itself to
this cgroup and expects that the test program process
cgroup id is the same as the cgroup_id retrieved
by the bpf program.
In a light system where no other processes called
nanosleep syscall, the test case can pass.
In a busy system where many different processes can hit
syscalls/sys_enter_nanosleep tracepoint, the cgroup id
recorded by bpf program may not match the test program
process cgroup_id.
This patch fixed an issue by communicating the test program
pid to bpf program. The bpf program only records
cgroup id if the current task pid is the same as
passed-in pid. This ensures that the recorded cgroup_id
is for the cgroup within which the test program resides.
Fixes: f269099a7e ("tools/bpf: add a selftest for bpf_get_current_cgroup_id() helper")
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Commit bf6fa2c893 ("bpf: implement bpf_get_current_cgroup_id()
helper") introduced a new helper bpf_get_current_cgroup_id().
The helper has a dependency on CONFIG_CGROUPS.
When CONFIG_CGROUPS is not defined, using the helper will result
the following verifier error:
kernel subsystem misconfigured func bpf_get_current_cgroup_id#80
which is hard for users to interpret.
Guarding the reference to bpf_get_current_cgroup_id_proto with
CONFIG_CGROUPS will result in below better message:
unknown func bpf_get_current_cgroup_id#80
Fixes: bf6fa2c893 ("bpf: implement bpf_get_current_cgroup_id() helper")
Suggested-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Syscall name_to_handle_at() can be used to get cgroup id
for a particular cgroup path in user space. The selftest
got cgroup id from both user and kernel, and compare to
ensure they are equal to each other.
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
bpf has been used extensively for tracing. For example, bcc
contains an almost full set of bpf-based tools to trace kernel
and user functions/events. Most tracing tools are currently
either filtered based on pid or system-wide.
Containers have been used quite extensively in industry and
cgroup is often used together to provide resource isolation
and protection. Several processes may run inside the same
container. It is often desirable to get container-level tracing
results as well, e.g. syscall count, function count, I/O
activity, etc.
This patch implements a new helper, bpf_get_current_cgroup_id(),
which will return cgroup id based on the cgroup within which
the current task is running.
The later patch will provide an example to show that
userspace can get the same cgroup id so it could
configure a filter or policy in the bpf program based on
task cgroup id.
The helper is currently implemented for tracing. It can
be added to other program types as well when needed.
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Yonghong Song