[PATCHv3,bpf-next,5/9] selftests/bpf: Add read_buildid function
Commit Message
Adding read_build_id function that parses out build id from
specified binary.
It will replace extract_build_id and also be used in following
changes.
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
tools/testing/selftests/bpf/trace_helpers.c | 86 +++++++++++++++++++++
tools/testing/selftests/bpf/trace_helpers.h | 5 ++
2 files changed, 91 insertions(+)
Comments
On Thu, Mar 16, 2023 at 10:03 AM Jiri Olsa <jolsa@kernel.org> wrote:
>
> Adding read_build_id function that parses out build id from
> specified binary.
>
> It will replace extract_build_id and also be used in following
> changes.
>
> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> ---
> tools/testing/selftests/bpf/trace_helpers.c | 86 +++++++++++++++++++++
> tools/testing/selftests/bpf/trace_helpers.h | 5 ++
> 2 files changed, 91 insertions(+)
>
> diff --git a/tools/testing/selftests/bpf/trace_helpers.c b/tools/testing/selftests/bpf/trace_helpers.c
> index 934bf28fc888..72b38a41f574 100644
> --- a/tools/testing/selftests/bpf/trace_helpers.c
> +++ b/tools/testing/selftests/bpf/trace_helpers.c
> @@ -11,6 +11,9 @@
> #include <linux/perf_event.h>
> #include <sys/mman.h>
> #include "trace_helpers.h"
> +#include <linux/limits.h>
> +#include <libelf.h>
> +#include <gelf.h>
>
> #define TRACEFS_PIPE "/sys/kernel/tracing/trace_pipe"
> #define DEBUGFS_PIPE "/sys/kernel/debug/tracing/trace_pipe"
> @@ -234,3 +237,86 @@ ssize_t get_rel_offset(uintptr_t addr)
> fclose(f);
> return -EINVAL;
> }
> +
> +static int
> +parse_build_id_buf(const void *note_start, Elf32_Word note_size,
> + char *build_id)
nit: single line
should we pass buffer size instead of assuming at least BPF_BUILD_ID_SIZE below?
> +{
> + Elf32_Word note_offs = 0, new_offs;
> +
> + while (note_offs + sizeof(Elf32_Nhdr) < note_size) {
> + Elf32_Nhdr *nhdr = (Elf32_Nhdr *)(note_start + note_offs);
> +
> + if (nhdr->n_type == 3 && nhdr->n_namesz == sizeof("GNU") &&
> + !strcmp((char *)(nhdr + 1), "GNU") && nhdr->n_descsz > 0 &&
> + nhdr->n_descsz <= BPF_BUILD_ID_SIZE) {
> + memcpy(build_id, note_start + note_offs +
> + ALIGN(sizeof("GNU"), 4) + sizeof(Elf32_Nhdr), nhdr->n_descsz);
> + memset(build_id + nhdr->n_descsz, 0, BPF_BUILD_ID_SIZE - nhdr->n_descsz);
> + return (int) nhdr->n_descsz;
> + }
> +
> + new_offs = note_offs + sizeof(Elf32_Nhdr) +
> + ALIGN(nhdr->n_namesz, 4) + ALIGN(nhdr->n_descsz, 4);
> + if (new_offs >= note_size)
> + break;
while condition() above would handle this, so this check appears not necessary?
so just assign note_offs directly?
> + note_offs = new_offs;
> + }
> +
> + return -EINVAL;
nit: -ENOENT or -ESRCH?
> +}
> +
> +/* Reads binary from *path* file and returns it in the *build_id*
> + * which is expected to be at least BPF_BUILD_ID_SIZE bytes.
> + * Returns size of build id on success. On error the error value
> + * is returned.
> + */
> +int read_build_id(const char *path, char *build_id)
> +{
> + int fd, err = -EINVAL;
> + Elf *elf = NULL;
> + GElf_Ehdr ehdr;
> + size_t max, i;
> +
> + fd = open(path, O_RDONLY | O_CLOEXEC);
> + if (fd < 0)
> + return -errno;
> +
> + (void)elf_version(EV_CURRENT);
> +
> + elf = elf_begin(fd, ELF_C_READ, NULL);
ELF_C_READ_MMAP ?
> + if (!elf)
> + goto out;
> + if (elf_kind(elf) != ELF_K_ELF)
> + goto out;
> + if (gelf_getehdr(elf, &ehdr) == NULL)
nit: !gelf_getehdr()
> + goto out;
> + if (ehdr.e_ident[EI_CLASS] != ELFCLASS64)
> + goto out;
does this have to be 64-bit specific?... you are using gelf stuff, you
can be bitness-agnostic here
> +
> + for (i = 0; i < ehdr.e_phnum; i++) {
> + GElf_Phdr mem, *phdr;
> + char *data;
> +
> + phdr = gelf_getphdr(elf, i, &mem);
> + if (!phdr)
> + goto out;
> + if (phdr->p_type != PT_NOTE)
> + continue;
I don't know where ELF + build ID spec is (if at all), but it seems to
always be in the ".note.gnu.build-id" section, so should we check the
name here?
> + data = elf_rawfile(elf, &max);
> + if (!data)
> + goto out;
> + if (phdr->p_offset >= max || (phdr->p_offset + phdr->p_memsz >= max))
`phdr->p_offset + phdr->p_memsz == max` would be fine, no?
> + goto out;
> + err = parse_build_id_buf(data + phdr->p_offset, phdr->p_memsz, build_id);
> + if (err > 0)
> + goto out;
> + err = -EINVAL;
> + }
> +
> +out:
> + if (elf)
> + elf_end(elf);
> + close(fd);
> + return err;
> +}
> diff --git a/tools/testing/selftests/bpf/trace_helpers.h b/tools/testing/selftests/bpf/trace_helpers.h
> index 53efde0e2998..bc3b92057033 100644
> --- a/tools/testing/selftests/bpf/trace_helpers.h
> +++ b/tools/testing/selftests/bpf/trace_helpers.h
> @@ -4,6 +4,9 @@
>
> #include <bpf/libbpf.h>
>
> +#define __ALIGN_MASK(x, mask) (((x)+(mask))&~(mask))
> +#define ALIGN(x, a) __ALIGN_MASK(x, (typeof(x))(a)-1)
> +
> struct ksym {
> long addr;
> char *name;
> @@ -23,4 +26,6 @@ void read_trace_pipe(void);
> ssize_t get_uprobe_offset(const void *addr);
> ssize_t get_rel_offset(uintptr_t addr);
>
> +int read_build_id(const char *path, char *build_id);
> +
> #endif
> --
> 2.39.2
>
On Thu, Mar 16, 2023 at 03:23:03PM -0700, Andrii Nakryiko wrote:
> On Thu, Mar 16, 2023 at 10:03 AM Jiri Olsa <jolsa@kernel.org> wrote:
> >
> > Adding read_build_id function that parses out build id from
> > specified binary.
> >
> > It will replace extract_build_id and also be used in following
> > changes.
> >
> > Signed-off-by: Jiri Olsa <jolsa@kernel.org>
I'll send this separatelly as bpf/selftests fix so doesn't get lost
> > ---
> > tools/testing/selftests/bpf/trace_helpers.c | 86 +++++++++++++++++++++
> > tools/testing/selftests/bpf/trace_helpers.h | 5 ++
> > 2 files changed, 91 insertions(+)
> >
> > diff --git a/tools/testing/selftests/bpf/trace_helpers.c b/tools/testing/selftests/bpf/trace_helpers.c
> > index 934bf28fc888..72b38a41f574 100644
> > --- a/tools/testing/selftests/bpf/trace_helpers.c
> > +++ b/tools/testing/selftests/bpf/trace_helpers.c
> > @@ -11,6 +11,9 @@
> > #include <linux/perf_event.h>
> > #include <sys/mman.h>
> > #include "trace_helpers.h"
> > +#include <linux/limits.h>
> > +#include <libelf.h>
> > +#include <gelf.h>
> >
> > #define TRACEFS_PIPE "/sys/kernel/tracing/trace_pipe"
> > #define DEBUGFS_PIPE "/sys/kernel/debug/tracing/trace_pipe"
> > @@ -234,3 +237,86 @@ ssize_t get_rel_offset(uintptr_t addr)
> > fclose(f);
> > return -EINVAL;
> > }
> > +
> > +static int
> > +parse_build_id_buf(const void *note_start, Elf32_Word note_size,
> > + char *build_id)
>
> nit: single line
ok
>
> should we pass buffer size instead of assuming at least BPF_BUILD_ID_SIZE below?
ok
>
> > +{
> > + Elf32_Word note_offs = 0, new_offs;
> > +
> > + while (note_offs + sizeof(Elf32_Nhdr) < note_size) {
> > + Elf32_Nhdr *nhdr = (Elf32_Nhdr *)(note_start + note_offs);
> > +
> > + if (nhdr->n_type == 3 && nhdr->n_namesz == sizeof("GNU") &&
> > + !strcmp((char *)(nhdr + 1), "GNU") && nhdr->n_descsz > 0 &&
> > + nhdr->n_descsz <= BPF_BUILD_ID_SIZE) {
> > + memcpy(build_id, note_start + note_offs +
> > + ALIGN(sizeof("GNU"), 4) + sizeof(Elf32_Nhdr), nhdr->n_descsz);
> > + memset(build_id + nhdr->n_descsz, 0, BPF_BUILD_ID_SIZE - nhdr->n_descsz);
> > + return (int) nhdr->n_descsz;
> > + }
> > +
> > + new_offs = note_offs + sizeof(Elf32_Nhdr) +
> > + ALIGN(nhdr->n_namesz, 4) + ALIGN(nhdr->n_descsz, 4);
> > + if (new_offs >= note_size)
> > + break;
>
> while condition() above would handle this, so this check appears not necessary?
>
> so just assign note_offs directly?
good idea, it will simplify that
>
>
> > + note_offs = new_offs;
> > + }
> > +
> > + return -EINVAL;
>
> nit: -ENOENT or -ESRCH?
I kept the same error as is in kernel, but ENOENT makes more sense
>
> > +}
> > +
> > +/* Reads binary from *path* file and returns it in the *build_id*
> > + * which is expected to be at least BPF_BUILD_ID_SIZE bytes.
> > + * Returns size of build id on success. On error the error value
> > + * is returned.
> > + */
> > +int read_build_id(const char *path, char *build_id)
> > +{
> > + int fd, err = -EINVAL;
> > + Elf *elf = NULL;
> > + GElf_Ehdr ehdr;
> > + size_t max, i;
> > +
> > + fd = open(path, O_RDONLY | O_CLOEXEC);
> > + if (fd < 0)
> > + return -errno;
> > +
> > + (void)elf_version(EV_CURRENT);
> > +
> > + elf = elf_begin(fd, ELF_C_READ, NULL);
>
> ELF_C_READ_MMAP ?
ok
>
> > + if (!elf)
> > + goto out;
> > + if (elf_kind(elf) != ELF_K_ELF)
> > + goto out;
> > + if (gelf_getehdr(elf, &ehdr) == NULL)
>
> nit: !gelf_getehdr()
ok
>
> > + goto out;
> > + if (ehdr.e_ident[EI_CLASS] != ELFCLASS64)
> > + goto out;
>
> does this have to be 64-bit specific?... you are using gelf stuff, you
> can be bitness-agnostic here
right, I don't think it's needed, will check
>
> > +
> > + for (i = 0; i < ehdr.e_phnum; i++) {
> > + GElf_Phdr mem, *phdr;
> > + char *data;
> > +
> > + phdr = gelf_getphdr(elf, i, &mem);
> > + if (!phdr)
> > + goto out;
> > + if (phdr->p_type != PT_NOTE)
> > + continue;
>
> I don't know where ELF + build ID spec is (if at all), but it seems to
> always be in the ".note.gnu.build-id" section, so should we check the
> name here?
this section name is not manadatory as stated in
https://fedoraproject.org/wiki/RolandMcGrath/BuildID
The new section is canonically called .note.gnu.build-id, but the name is not normative,
and the section can be merged with other SHT_NOTE sections. The ELF note headers give
name "GNU" and type 3 (NT_GNU_BUILD_ID) for a build ID note.
>
>
> > + data = elf_rawfile(elf, &max);
> > + if (!data)
> > + goto out;
> > + if (phdr->p_offset >= max || (phdr->p_offset + phdr->p_memsz >= max))
>
> `phdr->p_offset + phdr->p_memsz == max` would be fine, no?
right, will change
thanks,
jirka
>
> > + goto out;
> > + err = parse_build_id_buf(data + phdr->p_offset, phdr->p_memsz, build_id);
> > + if (err > 0)
> > + goto out;
> > + err = -EINVAL;
> > + }
> > +
> > +out:
> > + if (elf)
> > + elf_end(elf);
> > + close(fd);
> > + return err;
> > +}
> > diff --git a/tools/testing/selftests/bpf/trace_helpers.h b/tools/testing/selftests/bpf/trace_helpers.h
> > index 53efde0e2998..bc3b92057033 100644
> > --- a/tools/testing/selftests/bpf/trace_helpers.h
> > +++ b/tools/testing/selftests/bpf/trace_helpers.h
> > @@ -4,6 +4,9 @@
> >
> > #include <bpf/libbpf.h>
> >
> > +#define __ALIGN_MASK(x, mask) (((x)+(mask))&~(mask))
> > +#define ALIGN(x, a) __ALIGN_MASK(x, (typeof(x))(a)-1)
> > +
> > struct ksym {
> > long addr;
> > char *name;
> > @@ -23,4 +26,6 @@ void read_trace_pipe(void);
> > ssize_t get_uprobe_offset(const void *addr);
> > ssize_t get_rel_offset(uintptr_t addr);
> >
> > +int read_build_id(const char *path, char *build_id);
> > +
> > #endif
> > --
> > 2.39.2
> >
@@ -11,6 +11,9 @@
#include <linux/perf_event.h>
#include <sys/mman.h>
#include "trace_helpers.h"
+#include <linux/limits.h>
+#include <libelf.h>
+#include <gelf.h>
#define TRACEFS_PIPE "/sys/kernel/tracing/trace_pipe"
#define DEBUGFS_PIPE "/sys/kernel/debug/tracing/trace_pipe"
@@ -234,3 +237,86 @@ ssize_t get_rel_offset(uintptr_t addr)
fclose(f);
return -EINVAL;
}
+
+static int
+parse_build_id_buf(const void *note_start, Elf32_Word note_size,
+ char *build_id)
+{
+ Elf32_Word note_offs = 0, new_offs;
+
+ while (note_offs + sizeof(Elf32_Nhdr) < note_size) {
+ Elf32_Nhdr *nhdr = (Elf32_Nhdr *)(note_start + note_offs);
+
+ if (nhdr->n_type == 3 && nhdr->n_namesz == sizeof("GNU") &&
+ !strcmp((char *)(nhdr + 1), "GNU") && nhdr->n_descsz > 0 &&
+ nhdr->n_descsz <= BPF_BUILD_ID_SIZE) {
+ memcpy(build_id, note_start + note_offs +
+ ALIGN(sizeof("GNU"), 4) + sizeof(Elf32_Nhdr), nhdr->n_descsz);
+ memset(build_id + nhdr->n_descsz, 0, BPF_BUILD_ID_SIZE - nhdr->n_descsz);
+ return (int) nhdr->n_descsz;
+ }
+
+ new_offs = note_offs + sizeof(Elf32_Nhdr) +
+ ALIGN(nhdr->n_namesz, 4) + ALIGN(nhdr->n_descsz, 4);
+ if (new_offs >= note_size)
+ break;
+ note_offs = new_offs;
+ }
+
+ return -EINVAL;
+}
+
+/* Reads binary from *path* file and returns it in the *build_id*
+ * which is expected to be at least BPF_BUILD_ID_SIZE bytes.
+ * Returns size of build id on success. On error the error value
+ * is returned.
+ */
+int read_build_id(const char *path, char *build_id)
+{
+ int fd, err = -EINVAL;
+ Elf *elf = NULL;
+ GElf_Ehdr ehdr;
+ size_t max, i;
+
+ fd = open(path, O_RDONLY | O_CLOEXEC);
+ if (fd < 0)
+ return -errno;
+
+ (void)elf_version(EV_CURRENT);
+
+ elf = elf_begin(fd, ELF_C_READ, NULL);
+ if (!elf)
+ goto out;
+ if (elf_kind(elf) != ELF_K_ELF)
+ goto out;
+ if (gelf_getehdr(elf, &ehdr) == NULL)
+ goto out;
+ if (ehdr.e_ident[EI_CLASS] != ELFCLASS64)
+ goto out;
+
+ for (i = 0; i < ehdr.e_phnum; i++) {
+ GElf_Phdr mem, *phdr;
+ char *data;
+
+ phdr = gelf_getphdr(elf, i, &mem);
+ if (!phdr)
+ goto out;
+ if (phdr->p_type != PT_NOTE)
+ continue;
+ data = elf_rawfile(elf, &max);
+ if (!data)
+ goto out;
+ if (phdr->p_offset >= max || (phdr->p_offset + phdr->p_memsz >= max))
+ goto out;
+ err = parse_build_id_buf(data + phdr->p_offset, phdr->p_memsz, build_id);
+ if (err > 0)
+ goto out;
+ err = -EINVAL;
+ }
+
+out:
+ if (elf)
+ elf_end(elf);
+ close(fd);
+ return err;
+}
@@ -4,6 +4,9 @@
#include <bpf/libbpf.h>
+#define __ALIGN_MASK(x, mask) (((x)+(mask))&~(mask))
+#define ALIGN(x, a) __ALIGN_MASK(x, (typeof(x))(a)-1)
+
struct ksym {
long addr;
char *name;
@@ -23,4 +26,6 @@ void read_trace_pipe(void);
ssize_t get_uprobe_offset(const void *addr);
ssize_t get_rel_offset(uintptr_t addr);
+int read_build_id(const char *path, char *build_id);
+
#endif