Date:   Thu, 14 Jun 2018 14:18:06 -0300
From:   Arnaldo Carvalho de Melo <acme@...nel.org>
To:     Martin KaFai Lau <kafai@...com>
Cc:     netdev@...r.kernel.org, Alexei Starovoitov <ast@...com>,
        Daniel Borkmann <daniel@...earbox.net>, kernel-team@...com,
        Wang Nan <wangnan0@...wei.com>, Jiri Olsa <jolsa@...nel.org>,
        Namhyung Kim <namhyung@...nel.org>,
        Ingo Molnar <mingo@...nel.org>
Subject: Re: [PATCH bpf-next v5 00/10] BTF: BPF Type Format

Em Thu, Jun 14, 2018 at 09:22:27AM -0700, Martin KaFai Lau escreveu:
> On Thu, Jun 14, 2018 at 12:03:34PM -0300, Arnaldo Carvalho de Melo wrote:
> 
> > > > > > 1. The tools/testing/selftests/bpf/Makefile has the CLANG_FLAGS and
> > > > > >    LLC_FLAGS needed to compile the bpf prog.  It requires a new
> > > > > >    "-mattr=dwarf" llc option which was added to the future
> > > > > >    llvm 7.0.
> 
> [ ... ]
> 
> > I tried it, but it didn't work, see:
> > 
> > [root@...et bpf]# cat hello.c 
> > #include "stdio.h"
> > 
> > int syscall_enter(openat)(void *ctx)
> > {
> > 	puts("Hello, world\n");
> > 	return 0;
> > }
> > [root@...et bpf]# trace -e openat,hello.c touch /tmp/kafai
> > clang-6.0: error: unknown argument: '-mattr=dwarf'
> "-mattr=dwarf" is currently a llc only option.
> 
> tools/testing/selftests/bpf/Makefile has example on how to pipe clang to llc.
 
> e.g.:
> clang -g -O2 -target bpf -emit-llvm -c hello.c -o - | llc -march=bpf -mcpu=generic -mattr=dwarfris -filetype=obj -o hello.o

Ok, so I'll probably add a llvm.opts .perfconfig entry that, if present
will tell tools/perf/util/llvm-utils.c that piping the output of clang
to llvm, so that we can use llvm specific options, needs to be done.

Probably, for the time being I'll check for -g in llvm.clang-opt and if
it is there, set up the piping...

Just out of curiosity, is there any plan to have this as a clang option?

Just to finish this thing here, lemme try a slightly modified version of
your command line:

[root@...et bpf]# clang -D__KERNEL__ -D__NR_CPUS__=4 -DLINUX_VERSION_CODE=0x41100 -g -nostdinc -isystem /usr/lib/gcc/x86_64-redhat-linux/7/include -I/home/acme/git/linux/arch/x86/include -I./arch/x86/include/generated  -I/home/acme/git/linux/include -I./include -I/home/acme/git/linux/arch/x86/include/uapi -I./arch/x86/include/generated/uapi -I/home/acme/git/linux/include/uapi -I./include/generated/uapi -include /home/acme/git/linux/include/linux/kconfig.h  -I/home/acme/lib/include/perf/bpf -Wno-unused-value -Wno-pointer-sign -working-directory /lib/modules/4.17.0-rc5/build -c /home/acme/bpf/hello.c -target bpf -emit-llvm -O2 -o - | llc -march=bpf -mcpu=generic -mattr=dwarfris -filetype=obj -o hello2.o
[root@...et bpf]# 

[root@...et bpf]# file hello2.o
hello2.o: ELF 64-bit LSB relocatable, *unknown arch 0xf7* version 1 (SYSV), with debug_info, not stripped
[root@...et bpf]# pahole hello2.o
struct bpf_map_def {
	unsigned int               type;                 /*     0     4 */
	unsigned int               key_size;             /*     4     4 */
	unsigned int               value_size;           /*     8     4 */
	unsigned int               max_entries;          /*    12     4 */

	/* size: 16, cachelines: 1, members: 4 */
	/* last cacheline: 16 bytes */
};
[root@...et bpf]#

Finally works, thanks.

Thanks,

- Arnaldo
 
> > ERROR:	unable to compile hello.c
> > Hint:	Check error message shown above.
> > Hint:	You can also pre-compile it into .o using:
> >      		clang -target bpf -O2 -c hello.c
> >      	with proper -I and -D options.
> > event syntax error: 'hello.c'
> >                      \___ Failed to load hello.c from source: Error when compiling BPF scriptlet
> > 
> > (add -v to see detail)
> > Run 'perf list' for a list of valid events
> > 
> >  Usage: perf trace [<options>] [<command>]
> >     or: perf trace [<options>] -- <command> [<options>]
> >     or: perf trace record [<options>] [<command>]
> >     or: perf trace record [<options>] -- <command> [<options>]
> > 
> >     -e, --event <event>   event/syscall selector. use 'perf list' to list available events
> > [root@...et bpf]#
> > 
> > The full command line with that is:
> > 
> > [root@...et bpf]# trace -v -e openat,hello.c touch /tmp/kafai |& grep mattr
> > set env: CLANG_OPTIONS=-g -mattr=dwarf
> > llvm compiling command : /usr/local/bin/clang -D__KERNEL__ -D__NR_CPUS__=4 -DLINUX_VERSION_CODE=0x41100 -g -mattr=dwarf  -nostdinc -isystem /usr/lib/gcc/x86_64-redhat-linux/7/include -I/home/acme/git/linux/arch/x86/include -I./arch/x86/include/generated  -I/home/acme/git/linux/include -I./include -I/home/acme/git/linux/arch/x86/include/uapi -I./arch/x86/include/generated/uapi -I/home/acme/git/linux/include/uapi -I./include/generated/uapi -include /home/acme/git/linux/include/linux/kconfig.h  -I/home/acme/lib/include/perf/bpf -Wno-unused-value -Wno-pointer-sign -working-directory /lib/modules/4.17.0-rc5/build -c /home/acme/bpf/hello.c -target bpf -O2 -o -
> > clang-6.0: error: unknown argument: '-mattr=dwarf'
> > [root@...et bpf]#
> > 
> > This is with these llvm and clang trees:
> > 
> > [root@...et llvm]# git log --oneline -5
> > 98c78e82f54 (HEAD -> master, origin/master, origin/HEAD) [asan] Instrument comdat globals on COFF targets
> > 6ad988b5998 [DAGCombiner] clean up comments; NFC
> > a735ba5b795 [X86][SSE] Support v8i16/v16i16 rotations
> > 1503b9f6fe8 [x86] add tests for node-level FMF; NFC
> > 4a49826736f [x86] regenerate test checks; NFC
> > [root@...et llvm]#
> > 
> > [root@...et llvm]# cd tools/clang/
> > [root@...et clang]# git log --oneline -5
> > 8c873daccc (HEAD -> master, origin/master, origin/HEAD) [X86] Add builtins for vpermq/vpermpd instructions to enable target feature checking.
> > a344be6ba4 [X86] Change immediate type for some builtins from char to int.
> > dcdd53793e [CUDA] Fix emission of constant strings in sections
> > a90c85acaf [X86] Add builtins for shufps and shufpd to enable target feature and immediate range checking.
> > ff71c0eccc [X86] Add builtins for pshufd, pshuflw, and pshufhw to enable target feature and immediate range checking.
> > [root@...et clang]#
> > 
> > [root@...et clang]# git log | grep mattr=dwarf
> > [root@...et clang]# cd -
> > /home/acme/git.tmp/git/llvm
> > [root@...et llvm]# git log | grep mattr=dwarf
> >     bpf: introduce -mattr=dwarfris to disable DwarfUsesRelocationsAcrossSections
> >     This patch introduces a new flag -mattr=dwarfris
> > [root@...et llvm]#
> > 
> > Humm, so its -mattr=dwarfris and not -attr=dwarf?
> > 
> > Didn't help :-\
> > 
> > commit 0e0047f8c9ada2f0fe0c5f01579a80e2455b8df5
> > Author: Yonghong Song <yhs@...com>
> > Date:   Thu Mar 1 23:04:59 2018 +0000
> > 
> >     bpf: introduce -mattr=dwarfris to disable DwarfUsesRelocationsAcrossSections
> >     
> >     Commit e4507fb8c94b ("bpf: disable DwarfUsesRelocationsAcrossSections")
> >     disables MCAsmInfo DwarfUsesRelocationsAcrossSections unconditionally
> >     so that dwarf will not use cross section (between dwarf and symbol table)
> >     relocations. This new debug format enables pahole to dump structures
> >     correctly as libdwarves.so does not have BPF backend support yet.
> >     
> >     This new debug format, however, breaks bcc (https://github.com/iovisor/bcc)
> >     source debug output as llvm in-memory Dwarf support has some issues to
> >     handle it. More specifically, with DwarfUsesRelocationsAcrossSections
> >     disabled, JIT compiler does not generate .debug_abbrev and Dwarf
> >     DIE (debug info entry) processing is not happy about this.
> >     
> >     This patch introduces a new flag -mattr=dwarfris
> >     (dwarf relocation in section) to disable DwarfUsesRelocationsAcrossSections.
> >     DwarfUsesRelocationsAcrossSections is true by default.
> >     
> >     Signed-off-by: Yonghong Song <yhs@...com>
> >     
> >     git-svn-id: https://urldefense.proofpoint.com/v2/url?u=https-3A__llvm.org_svn_llvm-2Dproject_llvm_trunk-40326505&d=DwIBAg&c=5VD0RTtNlTh3ycd41b3MUw&r=VQnoQ7LvghIj0gVEaiQSUw&m=LO28-RE-2ZJTXto_gff4BgnxXkbUq8d2CEz1jD_wDl4&s=VCR3pGVfY54-OsZ3BqRsOr3FF5JVyltwbnbzu30_4EY&e= 91177308-0d34-0410-b5e6-96231b3b80d8
> > 
> > 

Powered by blists - more mailing lists