On Tue, Jun 15, 2021 at 3:11 AM Lorenz Bauer <lmb@xxxxxxxxxxxxxx> wrote: > > On Tue, 15 Jun 2021 at 00:27, Andrii Nakryiko <andrii.nakryiko@xxxxxxxxx> wrote: > > > > It doesn't seem avoidable. But I'm surprised you are satisfied with > > your patch, it doesn't seem to solve your problem, because you'll > > never trigger those _Pragmas as you'll just fallback to using your > > host architecture. Isn't that right? How did you test your patch? > > I tested the patch by removing -D__TARGET_ARCH_$(SRCARCH) from > BPF_CFLAGS in the Makefile. The pragmas are triggered because the > testsuite compiles with -target bpf. This prevents the "host arch" > fallback from activating. bpf2go specifies -target bpf(el|eb) as well, > so any users will get the _Pragma if they use a new enough > bpf_tracing.h. Oh, I didn't realize -target bpf will prevent host architecture fallback. In that case we don't need a new #define, cool. > > > > > > > Without it we sometimes get an integer cast warning, something about > > > an int to void* cast I think? > > > > hmm.. ok > > This is the error I get: > > progs/lsm.c:166:14: warning: cast to 'void *' from smaller integer > type 'int' [-Wint-to-void-pointer-cast] > void *ptr = (void *)PT_REGS_PARM1(regs); > ^~~~~~~~~~~~~~~~~~~~~~~~~~~ oh, ok, but then those zeros probably best to mark as longs, not long longs (even thought for BPF it's the same), as (unsigned) long is a logical equivalent of a pointer, right? > > -- > Lorenz Bauer | Systems Engineer > 6th Floor, County Hall/The Riverside Building, SE1 7PB, UK > > www.cloudflare.com
