Re: [PATCH v5 perf,bpf 07/15] perf, bpf: save bpf_prog_info information as headers to perf.data

[Jiri Olsa]

On Mon, Mar 04, 2019 at 07:36:14PM +0000, Song Liu wrote:
> 
> 
> > On Mar 4, 2019, at 5:52 AM, Jiri Olsa <jolsa@xxxxxxxxxx> wrote:
> > 
> > On Wed, Feb 27, 2019 at 09:06:35PM -0800, Song Liu wrote:
> > 
> > SNIP
> > 
> >> +static int process_bpf_prog_info(struct feat_fd *ff,
> >> +				 void *data __maybe_unused)
> >> +{
> >> +	struct bpf_prog_info_linear *info_linear;
> >> +	struct bpf_prog_info_node *info_node;
> >> +	struct perf_env *env = &ff->ph->env;
> >> +	u32 count, i;
> >> +	int err = -1;
> >> +
> >> +	if (do_read_u32(ff, &count))
> >> +		return -1;
> >> +
> >> +	down_write(&env->bpf_progs.lock);
> >> +
> >> +	for (i = 0; i < count; ++i) {
> >> +		u32 info_len, data_len;
> >> +
> >> +		info_linear = NULL;
> >> +		info_node = NULL;
> >> +		if (do_read_u32(ff, &info_len))
> >> +			goto out;
> >> +		if (do_read_u32(ff, &data_len))
> >> +			goto out;
> >> +
> >> +		if (info_len > sizeof(struct bpf_prog_info)) {
> >> +			pr_warning("detected invalid bpf_prog_info\n");
> >> +			goto out;
> >> +		}
> >> +
> >> +		info_linear = malloc(sizeof(struct bpf_prog_info_linear) +
> >> +				     data_len);
> >> +		if (!info_linear)
> >> +			goto out;
> >> +		info_linear->info_len = sizeof(struct bpf_prog_info);
> >> +		info_linear->data_len = data_len;
> >> +		if (do_read_u64(ff, (u64 *)(&info_linear->arrays)))
> >> +			goto out;
> >> +		if (__do_read(ff, &info_linear->info, info_len))
> >> +			goto out;
> >> +		if (info_len < sizeof(struct bpf_prog_info))
> >> +			memset(((void *)(&info_linear->info)) + info_len, 0,
> >> +			       sizeof(struct bpf_prog_info) - info_len);
> >> +
> >> +		if (__do_read(ff, info_linear->data, data_len))
> >> +			goto out;
> >> +
> >> +		/* endian mismatch, drop the info, continue */
> >> +		if (ff->ph->needs_swap) {
> >> +			free(info_linear);
> >> +			continue;
> >> +		}
> > 
> > so in this case we can check for needs_swap in the begining
> > of the function and bail out without reading all the data
> 
> If we bail out, perf-record will fail. If we read all the data 
> but ignore them, perf-record will continue without bpf annotation 
> support. I think the latter is a better experience with little 
> effort here. 

i did not mean bail out with error, just return 0, warn and go on

jirka

> 
> > 
> > also please display soem error message saying we don't support
> > ebpf progs data report over the different endianity
> 
> I will add a warning in the next version. 
> 
> Thanks,
> Song
>