Re: [Patch 2/2] ksym_tracer:Handle machine stall when cat trace_pipe for ksym tracer

[Frederic Weisbecker <fweisbec@gmail.com>]

On Wed, Jun 17, 2009 at 04:35:27AM +0530, K.Prasad wrote:
> 'trace_pipe' does not handle a TRACE_TYPE_PARTIAL_LINE well and causes the
> machine to stall.



No, if it stalls here, it means it handles it well :)



> While a TRACE_TYPE_UNHANDLED return causes the tracer to
> output unrelated data, a TRACE_TYPE_HANDLED return presents a clean output
> (minus all partial traces).
> 
> Signed-off-by: K.Prasad <prasad@linux.vnet.ibm.com>
> ---
>  kernel/trace/trace_ksym.c |    6 ++++++
>  1 file changed, 6 insertions(+)
> 
> Index: linux-2.6-tip.hbkpt/kernel/trace/trace_ksym.c
> ===================================================================
> --- linux-2.6-tip.hbkpt.orig/kernel/trace/trace_ksym.c
> +++ linux-2.6-tip.hbkpt/kernel/trace/trace_ksym.c
> @@ -389,6 +389,12 @@ static enum print_line_t ksym_trace_outp
>  
>  	trace_assign_type(field, entry);
>  
> +	/*
> +	 * Return early without any output if we don't have sufficient
> +	 * information
> +	 */
> +	if ((!field->ksym_hbp->info.type) || (!field->ip))
> +		return TRACE_TYPE_HANDLED;



Yeah that seems a good fix. But such silent ignored traces may
hide bugs (current or futures).
Is it a common situation to have a zero ip or an undefined
breakpoint type? How can that happen?

Thanks.



>  	ret = trace_seq_printf(s, "%-15s %-5d %-3d %-20s ", field->p_name,
>  				entry->pid, iter->cpu, field->ksym_name);
>  	if (!ret) {
>