On Mon, Dec 12, 2016 at 10:33 PM, Ming Lei <tom.leim...@gmail.com> wrote: > On Tue, Dec 13, 2016 at 2:11 PM, Alexei Starovoitov > <alexei.starovoi...@gmail.com> wrote: >> On Mon, Dec 12, 2016 at 9:44 PM, Ming Lei <tom.leim...@gmail.com> wrote: >>> >>>> so you can send me a pre-processed .c file or even better .ll file ? >>>> I don't have arm64 to reproduce it... >>> >>> Please see the whole log of building the prog. >> >> that raw asm outside of any function body comes from >> arch/arm64/include/asm/sysreg.h >> it obviously cannot work with any crosscompiler. >> That's why we have -D__ASM_SYSREG_H in samples/bpf/Makefile > > Yes, passing '-D__ASM_SYSREG_H' just avoids to include > <asm/sysreg.h> when building prog. > > But that workaround doesn't work any more recently, because > some macros defined in sysreg.h are needed by other headers, > please see the building failure in my 2nd email of this thread. > > I can be a headache by just fixing kernel, since old kernel is still > running and old header is still installed. And raw asm() can be > introduced in the future too. > > So could we deal with the issue in bcc?
I guess we can try to filter out all raw asm after clang stage in bcc. Such fix would be better than doing similar filtering in the backend and asking everyone to upgrade to the latest llvm. Brenden, thoughts? _______________________________________________ iovisor-dev mailing list iovisor-dev@lists.iovisor.org https://lists.iovisor.org/mailman/listinfo/iovisor-dev
