[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <78433b38-3f34-c97a-ee74-a9b6dee95aa2@fb.com>
Date: Fri, 18 Oct 2019 20:04:44 +0000
From: Yonghong Song <yhs@...com>
To: Andrii Nakryiko <andriin@...com>, Jiri Olsa <jolsa@...nel.org>,
"Alexei Starovoitov" <ast@...nel.org>,
Daniel Borkmann <daniel@...earbox.net>
CC: "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
"bpf@...r.kernel.org" <bpf@...r.kernel.org>,
Martin Lau <kafai@...com>
Subject: Re: [PATCH] bpftool: Try to read btf as raw data if elf read fails
On 10/18/19 9:48 AM, Andrii Nakryiko wrote:
> On 10/18/19 3:34 AM, Jiri Olsa wrote:
>> The bpftool interface stays the same, but now it's possible
>> to run it over BTF raw data, like:
>
> Oh, great, I had similar patch laying around for a while, never got to
> cleaning it up, though, so thanks for picking this up!
>
>>
>> $ bpftool btf dump file /sys/kernel/btf/vmlinux
>> libbpf: failed to get EHDR from /sys/kernel/btf/vmlinux
>
> We should implement this so that we don't get an extra log output with
> errors. I've been thinking about checking first few bytes of the file.
> If that matches BTF_MAGIC, then try to parse it as raw BTF, otherwise
> parse as ELF w/ BTF. Does it make sense?
Agreed, this makes sense. We should not emit errors in such cases.
One minor comment below.
>
>> [1] INT '(anon)' size=4 bits_offset=0 nr_bits=32 encoding=(none)
>> [2] INT 'long unsigned int' size=8 bits_offset=0 nr_bits=64 encoding=(none)
>> [3] CONST '(anon)' type_id=2
>>
>> I'm also adding err init to 0 because I was getting uninitialized
>> warnings from gcc.
>>
>> Signed-off-by: Jiri Olsa <jolsa@...nel.org>
>> ---
>> tools/bpf/bpftool/btf.c | 47 ++++++++++++++++++++++++++++++++++++-----
>> 1 file changed, 42 insertions(+), 5 deletions(-)
>>
>> diff --git a/tools/bpf/bpftool/btf.c b/tools/bpf/bpftool/btf.c
>> index 9a9376d1d3df..100fb7e02329 100644
>> --- a/tools/bpf/bpftool/btf.c
>> +++ b/tools/bpf/bpftool/btf.c
>> @@ -12,6 +12,9 @@
>> #include <libbpf.h>
>> #include <linux/btf.h>
>> #include <linux/hashtable.h>
>> +#include <sys/types.h>
>> +#include <sys/stat.h>
>> +#include <unistd.h>
>>
>> #include "btf.h"
>> #include "json_writer.h"
>> @@ -388,6 +391,35 @@ static int dump_btf_c(const struct btf *btf,
>> return err;
>> }
>>
>> +static struct btf *btf__parse_raw(const char *file)
>> +{
>> + struct btf *btf = ERR_PTR(-EINVAL);
>> + __u8 *buf = NULL;
>> + struct stat st;
>> + FILE *f;
>> +
>> + if (stat(file, &st))
>> + return btf;
>> +
>> + f = fopen(file, "rb");
>> + if (!f)
>> + return btf;
>> +
>> + buf = malloc(st.st_size);
>> + if (!buf)
>> + goto err;
>> +
>> + if ((size_t) st.st_size != fread(buf, 1, st.st_size, f))
>> + goto err;
>> +
>> + btf = btf__new(buf, st.st_size);
>> +
>> +err:
Non error case can also reach here. Let us change
label to a different name, e.g., "done"?
>> + free(buf);
>> + fclose(f);
>> + return btf;
>> +}
>> +
>> static int do_dump(int argc, char **argv)
>> {
>> struct btf *btf = NULL;
>> @@ -397,7 +429,7 @@ static int do_dump(int argc, char **argv)
>> __u32 btf_id = -1;
>> const char *src;
>> int fd = -1;
>> - int err;
>> + int err = 0;
>>
>> if (!REQ_ARGS(2)) {
>> usage();
>> @@ -468,10 +500,15 @@ static int do_dump(int argc, char **argv)
>> btf = btf__parse_elf(*argv, NULL);
>> if (IS_ERR(btf)) {
>> err = PTR_ERR(btf);
>> - btf = NULL;
>> - p_err("failed to load BTF from %s: %s",
>> - *argv, strerror(err));
>> - goto done;
>> + if (err == -LIBBPF_ERRNO__FORMAT)
>> + btf = btf__parse_raw(*argv);
>> + if (IS_ERR(btf)) {
>> + btf = NULL;
>> + /* Display the original error value. */
>> + p_err("failed to load BTF from %s: %s",
>> + *argv, strerror(err));
>> + goto done;
>> + }
>> }
>> NEXT_ARG();
>> } else {
>>
>
Powered by blists - more mailing lists