Re: [Qemu-devel] [PATCH 5/5] HMP: Introduce console command

[Luiz Capitulino]

On Fri, 26 Oct 2012 03:43:10 +0800
Lei Li <address@hidden> wrote:

> On 10/24/2012 08:55 PM, Luiz Capitulino wrote:
> > On Wed, 24 Oct 2012 15:17:21 +0800
> > Lei Li <address@hidden> wrote:
> >
> >> On 10/23/2012 02:59 AM, Luiz Capitulino wrote:
> >>> On Mon, 22 Oct 2012 00:48:01 +0800
> >>> Lei Li <address@hidden> wrote:
> >>>
> >>>> Signed-off-by: Lei Li <address@hidden>
> >>>> ---
> >>>>    hmp-commands.hx |   23 +++++++++++++++++++++++
> >>>>    hmp.c           |   53 
> >>>> +++++++++++++++++++++++++++++++++++++++++++++++++++++
> >>>>    hmp.h           |    1 +
> >>>>    monitor.c       |   15 +++++++++++++++
> >>>>    monitor.h       |    3 +++
> >>>>    5 files changed, 95 insertions(+), 0 deletions(-)
> >>>>
> >>>> diff --git a/hmp-commands.hx b/hmp-commands.hx
> >>>> index 5f91428..f862a53 100644
> >>>> --- a/hmp-commands.hx
> >>>> +++ b/hmp-commands.hx
> >>>> @@ -868,7 +868,30 @@ char device and return @var{size} of the data.
> >>>>    that specifies behavior when the queue is full/empty. By default is
> >>>>    'drop'. Note that the 'block' option is not supported now.
> >>>>            -b for 'block' option. None for 'drop' option.
> >>>> +ETEXI
> >>>> +
> >>>> +    {
> >>>> +        .name       = "console",
> >>>> +        .args_type  = "chardev:s",
> >>>> +        .params     = "chardev",
> >>>> +        .help       = "Connect to the serial console from within the"
> >>>> +                      "monitor, allow to write data to memchardev"
> >>>> +                      "'chardev'. Exit from the console and return back"
> >>>> +                      "to monitor by typing 'ctrl-]'",
> >>>> +        .mhandler.cmd = hmp_console,
> >>>> +    },
> >>>>    
> >>>> +STEXI
> >>>> address@hidden console @var{device}
> >>>> address@hidden console
> >>>> +
> >>>> +Connect to the serial console from within the monitor, allow to write 
> >>>> data
> >>>> +to memchardev @var{chardev}. Exit from the console and return back to
> >>>> +monitor by typing 'ctrl-]'.
> >>>> address@hidden
> >>>> +(qemu) console foo
> >>>> +foo: data string...
> >>>> address@hidden example
> >>>>    ETEXI
> >>>>    
> >>>>        {
> >>>> diff --git a/hmp.c b/hmp.c
> >>>> index fa858c4..bc245f4 100644
> >>>> --- a/hmp.c
> >>>> +++ b/hmp.c
> >>>> @@ -1245,3 +1245,56 @@ void hmp_screen_dump(Monitor *mon, const QDict 
> >>>> *qdict)
> >>>>        qmp_screendump(filename, &err);
> >>>>        hmp_handle_error(mon, &err);
> >>>>    }
> >>>> +
> >>>> +enum escape_char
> >>>> +{
> >>>> +    ESCAPE_CHAR_CTRL_GS = 0x1d  /* ctrl-] used for escape */
> >>>> +};
> >>>> +
> >>>> +static void hmp_read_console(Monitor *mon, const char *data,
> >>>> +                             void *opaque)
> >>>> +{
> >>>> +    CharDriverState *chr = opaque;
> >>>> +    uint32_t size = strlen(data);
> >>>> +    enum DataFormat format = DATA_FORMAT_UTF8;
> >>>> +    enum CongestionControl control = CONGESTION_CONTROL_DROP;
> >>>> +    enum escape_char console_escape = ESCAPE_CHAR_CTRL_GS;
> >>>> +
> >>>> +    Error *err = NULL;
> >>>> +
> >>>> +    if (*data == console_escape) {
> >>>> +        monitor_resume(mon);
> >>>> +        return;
> >>>> +    }
> >>>> +
> >>>> +    qmp_memchar_write(chr->label, size, data, 0, format,
> >>>> +                      0, control, &err);
> >>>> +    if (err) {
> >>>> +        monitor_printf(mon, "%s\n", error_get_pretty(err));
> >>>> +        error_free(err);
> >>>> +        return;
> >>>> +    }
> >>> Shouldn't you also read from the device?
> >> The use-case for this console command is just allow the user to
> >> write data to each memchar device as in a signal terminal. Then
> >> type escape sequences to take you back to the monitor. So I don't
> >> think it is also need to read from the device in this command.
> >>
> >> BTW, we can read from the device by hmp_memchar_read. :)
> > And how is the console command better than the hmp_memchar_write one?
> >
> > Could you please give examples?
> 
> Sure. The "console" command behaves like:
> 
> (qemu) console foo
> foo: hello world
> (qemu)
> 
> you can input data, then put enter or ctrl-] to return back to
> the monitor. The data input from console would be written to the
> CirMemCharDriver backend, which provide a more friendly and nice
> HMP command as Anthony suggested.

Maybe I'm missing something, but I don't understand how this is
better than the hmp_memchar_write command.

> >>>> +
> >>>> +    monitor_read_command(mon, 1);
> >>>> +}
> >>>> +
> >>>> +void hmp_console(Monitor *mon, const QDict *qdict)
> >>>> +{
> >>>> +    const char *device = qdict_get_str(qdict, "chardev");
> >>>> +    CharDriverState *chr;
> >>>> +    Error *err = NULL;
> >>>> +
> >>>> +    chr = qemu_chr_find(device);
> >>>> +
> >>>> +    if (!chr) {
> >>>> +        error_set(&err, QERR_DEVICE_NOT_FOUND, device);
> >>>> +        goto out;
> >>>> +    }
> >>> No need to do this here as the QMP command will do it too.
> >>>
> >> I think we should do this check here, otherwise would
> >> cause core dump when 'console' to a chardev that does not
> >> exist. Wepass parameterchr->label by qmp_memchar_write()
> >> in the handler hmp_read_console.
> > Do you need the chr object? Why don't you just pass 'device' to
> > monitor_read_console()?
> 
> Yes, we need 'device' and 'chr' both here. 'device' pass to 
> monitor_read_console
> for the friendly prompt. 'chr' pass to handler hmp_read_console, the reason 
> did not
> pass 'device' directly to it is the limitation of argument for the handler
> hmp_read_console.

What limitation? It takes 'data' and 'opaque' (which is a bit weird).

> Also it's better to give the prompt before the user enter the input
> console if the device not found, the use-case as I post above.

Why?

> 
> >>>> +
> >>>> +    if (monitor_read_console(mon, device, hmp_read_console, chr) < 0) {
> >>>> +        monitor_printf(mon, "Connect to console %s failed\n", device);
> >>>> +    }
> >>>> +
> >>>> +out:
> >>>> +    hmp_handle_error(mon, &err);
> >>>> +}
> >>>> diff --git a/hmp.h b/hmp.h
> >>>> index a5a0cfe..5b54a79 100644
> >>>> --- a/hmp.h
> >>>> +++ b/hmp.h
> >>>> @@ -77,5 +77,6 @@ void hmp_getfd(Monitor *mon, const QDict *qdict);
> >>>>    void hmp_closefd(Monitor *mon, const QDict *qdict);
> >>>>    void hmp_send_key(Monitor *mon, const QDict *qdict);
> >>>>    void hmp_screen_dump(Monitor *mon, const QDict *qdict);
> >>>> +void hmp_console(Monitor *mon, const QDict *qdict);
> >>>>    
> >>>>    #endif
> >>>> diff --git a/monitor.c b/monitor.c
> >>>> index 131b325..453c084 100644
> >>>> --- a/monitor.c
> >>>> +++ b/monitor.c
> >>>> @@ -256,6 +256,21 @@ int monitor_read_password(Monitor *mon, 
> >>>> ReadLineFunc *readline_func,
> >>>>        }
> >>>>    }
> >>>>    
> >>>> +int monitor_read_console(Monitor *mon, const char *device,
> >>>> +                         ReadLineFunc *readline_func, void *opaque)
> >>>> +{
> >>>> +    char prompt[60];
> >>>> +
> >>>> +    if (!mon->rs) {
> >>>> +        return -1;
> >>>> +    }
> >>>> +
> >>>> +    snprintf(prompt, sizeof(prompt), "%s: ", device);
> >>>> +    readline_start(mon->rs, prompt, 0, readline_func, opaque);
> >>>> +
> >>>> +    return 0;
> >>>> +}
> >>>> +
> >>>>    void monitor_flush(Monitor *mon)
> >>>>    {
> >>>>        if (mon && mon->outbuf_index != 0 && !mon->mux_out) {
> >>>> diff --git a/monitor.h b/monitor.h
> >>>> index b6e7d95..735bd1b 100644
> >>>> --- a/monitor.h
> >>>> +++ b/monitor.h
> >>>> @@ -86,6 +86,9 @@ ReadLineState *monitor_get_rs(Monitor *mon);
> >>>>    int monitor_read_password(Monitor *mon, ReadLineFunc *readline_func,
> >>>>                              void *opaque);
> >>>>    
> >>>> +int monitor_read_console(Monitor *mon, const char *device,
> >>>> +                         ReadLineFunc *readline_func, void *opaque);
> >>>> +
> >>>>    int qmp_qom_set(Monitor *mon, const QDict *qdict, QObject **ret);
> >>>>    
> >>>>    int qmp_qom_get(Monitor *mon, const QDict *qdict, QObject **ret);
> >>
> >
> 
>