Re: [RFC PATCH v6 24/32] perf record: Compile scriptlets if pass '.c' to --event

From: Wangnan (F)
Date: Tue Jun 09 2015 - 20:13:49 EST




On 2015/6/10 5:48, Alexei Starovoitov wrote:
On 6/8/15 10:50 PM, Wang Nan wrote:
+struct bpf_param bpf_param = {
+ .clang_path = "clang",
+ .llc_path = "llc",
+ .clang_opt = "",
+ .llc_opt = "",
+};

the defaults are ok-ish, but llc is never in PATH.
So most likely it won't work out of the box.
I think the cleanest option is too wait a little bit
until we finish clang integration.
(hopefully few more days)

we've quite been busy on llvm side
and I'm excited to announce that BPF backend finally
graduated from experimental.
So now anyone can just pull the latest llvm and
bpf will be present by default.
No more cmake -DLLVM_EXPERIMENTAL...

Once clang integration is complete. One can pull
upsteam llvm and clang and just use 'clang -O2 -c -march=bpf file.c'


Good news, so llc part should be removed.

Next version I'd like to use following config options:

[llvm]
clang-bpf-cmd-template = "$CLANG_EXEC $CLANG_OPTIONS -c $CLANG_SOURCE -emit-llvm -O2 -o - | /path/to/llc -march=bpf -filetype=obj -o -"
clang-path = "/path/to/clang"
clang-opt = ""

And the default template should be:

$CLANG_EXEC $CLANG_OPTIONS -c "$CLANG_SOURCE" -emit-llvm -O2 -o -

Then with environment variable tricks we make it work with popen.

By this way we can get rid of llc in perf side, and make it work even before clang
integration is complete.

Thank you.

+ err = WEXITSTATUS(pclose(file));
+ if (err || read_sz <= 0) {
+ err = -EINVAL;
+ pr_err("Error:\tsomething went wrong when compiling %s\n",
+ filename);
+ pr_err("Hint:\tTry manually run following command and check:\n");
+ pr_err(" # " CMD_FMT "\n\n",
+ clang_path, bpf_param.clang_opt, filename,
+ llc_path, bpf_param.llc_opt, "/dev/null");

not terribly informative hint.
I think would be good to point to some readme or at least mention
that only the latest clang/llvm support bpf.
Either svn/git trunk or upcoming 3.7
May be a link to llvm.org?



--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/