Re: [Xen-devel] [PATCH v6 5/5] xen: add debugtrace entry when entry count is wrapping

[Juergen Gross <jgross@xxxxxxxx>]

On 09.09.19 10:19, Jan Beulich wrote:
> On 09.09.2019 09:35, Juergen Gross wrote:
> > @@ -173,10 +177,23 @@ void debugtrace_printk(const char *fmt, ...)
> >       nr = vsnprintf(buf, sizeof(buf), fmt, args);
> >       va_end(args);
> >   
> > +    if ( count == DEBUGTRACE_COUNT_WRAP )
> > +    {
> > +        count = 0;
> > +        wrap_cnt++;
> > +    }
> > +
> >       if ( debugtrace_send_to_console )
> >       {
> > -        unsigned int n = scnprintf(cntbuf, sizeof(cntbuf), "%u ", ++count);
> > +        unsigned int n;
> > +
> > +        if ( count == 0 )
> > +        {
> > +            n = scnprintf(cntbuf, sizeof(cntbuf), "wrap: %lu\n", wrap_cnt);
> > +            console_serial_puts(cntbuf, n);
> > +        }
> >   
> > +        n = scnprintf(cntbuf, sizeof(cntbuf), "%u ", ++count);
> >           console_serial_puts(cntbuf, n);
> >           console_serial_puts(buf, nr);
> >       }
> Just one minor remark: In this case you issue a "wrap" message at
> start of day, ...
>
> > @@ -184,8 +201,16 @@ void debugtrace_printk(const char *fmt, ...)
> >       {
> >           unsigned int cpu = debugtrace_per_cpu ? smp_processor_id() : 0;
> >   
> > -        if ( debugtrace_buf_empty || strcmp(buf, last_buf) )
> > +        if ( debugtrace_buf_empty || wrap_cnt != data->wrap_cnt ||
> > +             strcmp(buf, last_buf) )
> >           {
> > +            if ( wrap_cnt != data->wrap_cnt )
> > +            {
> > +                snprintf(cntbuf, sizeof(cntbuf), "wrap: %lu->%lu\n",
> > +                         data->wrap_cnt, wrap_cnt);
> > +                debugtrace_add_to_buf(cntbuf);
> > +                data->wrap_cnt = wrap_cnt;
> > +            }
> >               debugtrace_buf_empty = false;
> >               last_prd = data->prd;
> >               last_count = ++count;
> ... whereas here one gets issued (as I'd expect it) only after the
> first wrapping of "count".
Yes, I know. The reason I have done it this way was the possibility to
support debugtrace via console even in case of a zero-sized buffer in
future. I can point that out in the commit message.


Juergen

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel