Re: [Qemu-devel] [PATCH v2 09/11] monitor: Split out monitor/qmp.c

[Markus Armbruster]

Kevin Wolf <address@hidden> writes:

> Am 12.06.2019 um 15:11 hat Markus Armbruster geschrieben:
>> Kevin Wolf <address@hidden> writes:
>> 
>> > Move QMP infrastructure from monitor/misc.c to monitor/qmp.c. This is
>> > code that can be shared for all targets, so compile it only once.
>> 
>> Less code compiled per target, yay!
>> 
>> > The amount of function and particularly extern variables in
>> > monitor_int.h is probably a bit larger than it needs to be, but this way
>> > no non-trivial code modifications are needed. The interfaces between QMP
>> > and the monitor core can be cleaned up later.
>> 
>> That's okay.
>> 
>> I have to admit I naively expected the previous patch moved everything
>> to the new header we need in a header for splitting up monitor/misc.c.
>> How did you decide what to move to the header in which patch?
>
> The previous patch moved only the Monitor{HMP,QMP} data structures and
> their dependencies as I was sure these would be shared. Everything else
> was added to address linker complaints as I was going. I'll clarify this
> in the commit message of the previous patch.
>
>> > Signed-off-by: Kevin Wolf <address@hidden>
>> > Reviewed-by: Dr. David Alan Gilbert <address@hidden>
>> > ---
>> >  include/monitor/monitor.h |   1 +
>> >  monitor/monitor_int.h     |  30 ++-
>> >  monitor/misc.c            | 394 +------------------------------------
>> >  monitor/qmp.c             | 404 ++++++++++++++++++++++++++++++++++++++
>> >  Makefile.objs             |   1 +
>> >  monitor/Makefile.objs     |   1 +
>> >  monitor/trace-events      |   4 +-
>> >  7 files changed, 448 insertions(+), 387 deletions(-)
>> >  create mode 100644 monitor/qmp.c
>> >
>> > diff --git a/include/monitor/monitor.h b/include/monitor/monitor.h
>> > index 1ba354f811..7bbab05320 100644
>> > --- a/include/monitor/monitor.h
>> > +++ b/include/monitor/monitor.h
>> > @@ -21,6 +21,7 @@ bool monitor_cur_is_qmp(void);
>> >  
>> >  void monitor_init_globals(void);
>> >  void monitor_init(Chardev *chr, int flags);
>> > +void monitor_init_qmp(Chardev *chr, int flags);
>> 
>> Why does this one go to the non-internal header?
>
> Most callers already know whether they want QMP or HMP, so they can just
> directly create the right thing instead of going through the
> monitor_init() wrapper.
>
> If you prefer, I can move it to the internal header, though. It's not
> called externally yet.

As is, monitor_init_qmp() and monitor_init_hmp() are awkward interfaces:
what if you pass MONITOR_USE_CONTROL to monitor_init_hmp()?

I can see just one call passing flags that aren't compile-time
constant.  I think a better interface would be

    monitor_init_hmp(Chardev *chr);
    monitor_init_qmp(Chardev *chr, bool pretty);

replacing monitor_init() entirely.  This is my first preference.

My (somewhat distant) second is hiding the awkward interfaces in the
internal header for now, and clean them up later.

Your choice.

>> >  void monitor_cleanup(void);
>> >  
>> >  int monitor_suspend(Monitor *mon);
>> > diff --git a/monitor/monitor_int.h b/monitor/monitor_int.h
>> > index 7122418955..4aabee54e1 100644
>> > --- a/monitor/monitor_int.h
>> > +++ b/monitor/monitor_int.h
>> > @@ -30,10 +30,11 @@
>> >  
>> >  #include "qapi/qmp/qdict.h"
>> >  #include "qapi/qmp/json-parser.h"
>> > -#include "qapi/qapi-commands.h"
>> > +#include "qapi/qmp/dispatch.h"
>> 
>> This part should be squashed into the previous patch.  You'll
>> additionally need qapi/qapi-types-misc.h for QMP_CAPABILITY__MAX there,
>> or keep monitor/monitor.h, even though you need it only here for
>> MONITOR_USE_CONTROL.
>
> Yes, already happened while addressing the comments you had for the
> header.
>
>> >  
>> >  #include "qemu/readline.h"
>> >  #include "chardev/char-fe.h"
>> > +#include "sysemu/iothread.h"
>> 
>> Perhaps IOThread should be typedef'ed in qemu/typedefs.h.  I'm not
>> asking you to do that.
>> 
>> >  
>> >  /*
>> >   * Supported types:
>> > @@ -145,4 +146,31 @@ typedef struct {
>> >      GQueue *qmp_requests;
>> >  } MonitorQMP;
>> >  
>> > +/**
>> > + * Is @mon a QMP monitor?
>> > + */
>> > +static inline bool monitor_is_qmp(const Monitor *mon)
>> > +{
>> > +    return (mon->flags & MONITOR_USE_CONTROL);
>> > +}
>> > +
>> > +typedef QTAILQ_HEAD(MonitorList, Monitor) MonitorList;
>> > +extern IOThread *mon_iothread;
>> > +extern QEMUBH *qmp_dispatcher_bh;
>> > +extern QmpCommandList qmp_commands, qmp_cap_negotiation_commands;
>> > +extern QemuMutex monitor_lock;
>> > +extern MonitorList mon_list;
>> > +extern int mon_refcount;
>> > +
>> > +int monitor_puts(Monitor *mon, const char *str);
>> > +void monitor_data_init(Monitor *mon, int flags, bool skip_flush,
>> > +                       bool use_io_thread);
>> > +int monitor_can_read(void *opaque);
>> > +void monitor_list_append(Monitor *mon);
>> > +void monitor_fdsets_cleanup(void);
>> > +
>> > +void qmp_send_response(MonitorQMP *mon, const QDict *rsp);
>> > +void monitor_data_destroy_qmp(MonitorQMP *mon);
>> > +void monitor_qmp_bh_dispatcher(void *data);
>> > +
>> >  #endif
>> 
>> I trust you these are indeed all needed.
>
> The linker said so. :-)

I've seen people stuff random crap into headers just because, but I
trust you wouldn't do that :)

>> > diff --git a/monitor/qmp.c b/monitor/qmp.c
>> > new file mode 100644
>> > index 0000000000..d425b0f2ba
>> > --- /dev/null
>> > +++ b/monitor/qmp.c
>> > @@ -0,0 +1,404 @@
>> > +/*
>> > + * QEMU monitor
>> > + *
>> > + * Copyright (c) 2003-2004 Fabrice Bellard
>> 
>> I'm pretty confident nothing in this file is actually due to Fabrice.
>> 
>> > + *
>> > + * Permission is hereby granted, free of charge, to any person obtaining 
>> > a copy
>> > + * of this software and associated documentation files (the "Software"), 
>> > to deal
>> > + * in the Software without restriction, including without limitation the 
>> > rights
>> > + * to use, copy, modify, merge, publish, distribute, sublicense, and/or 
>> > sell
>> > + * copies of the Software, and to permit persons to whom the Software is
>> > + * furnished to do so, subject to the following conditions:
>> > + *
>> > + * The above copyright notice and this permission notice shall be 
>> > included in
>> > + * all copies or substantial portions of the Software.
>> > + *
>> > + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, 
>> > EXPRESS OR
>> > + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF 
>> > MERCHANTABILITY,
>> > + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
>> > + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR 
>> > OTHER
>> > + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, 
>> > ARISING FROM,
>> > + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS 
>> > IN
>> > + * THE SOFTWARE.
>> > + */
>> 
>> We inherit the non-standard license regardless.  Lesson learned: do not
>> add substantial code to source files with a non-standard license.
>
> Yes. Nothing I can change here now, though. The legally safe way for
> splitting files is to copy the license header.
>
>> > +
>> > +#include "qemu/osdep.h"
>> > +#include "monitor_int.h"
>> > +
>> > +#include "chardev/char-io.h"
>> > +
>> > +#include "qapi/error.h"
>> > +#include "qapi/qmp/qjson.h"
>> > +#include "qapi/qmp/qstring.h"
>> > +#include "qapi/qmp/qlist.h"
>> > +#include "qapi/qapi-commands-misc.h"
>> > +
>> > +#include "trace.h"
>> 
>> Please sort the includes alphabetically (except qemu/osdep.h, of
>> course).
>
> In other words, move monitor_int.h down and remove the empty lines?
>
> I like separating header files by "category", but that's a matter of
> taste. I'll follow yours if you tell me whether I understood it
> correctly.

Yes.

I hate the quasi-random order of #include we have in QEMU.  The ordering
method you describe would be okay with me, except for one thing: it
takes actual thought.  Alphabetical order can be enforced / fixed up by
a simple machine.

>> > +
>> > +struct QMPRequest {
>> > +    /* Owner of the request */
>> > +    MonitorQMP *mon;
>> > +    /*
>> > +     * Request object to be handled or Error to be reported
>> > +     * (exactly one of them is non-null)
>> > +     */
>> > +    QObject *req;
>> > +    Error *err;
>> > +};
>> > +typedef struct QMPRequest QMPRequest;
>> 
>> Please fuse these two.
>
> This is a code motion patch. Are you sure? [y/n]

Nevermind :)