qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Markus Armbruster <armbru@redhat.com>
To: Markus Armbruster <armbru@redhat.com>
Cc: kwolf@redhat.com, Paolo Bonzini <pbonzini@redhat.com>,
	qemu-devel@nongnu.org, imammedo@redhat.com
Subject: Re: [PATCH 08/25] hmp: replace "O" parser with keyval
Date: Mon, 01 Mar 2021 11:43:38 +0100	[thread overview]
Message-ID: <87k0qrjgrp.fsf@dusky.pond.sub.org> (raw)
In-Reply-To: <87zh0xo0fw.fsf@dusky.pond.sub.org> (Markus Armbruster's message of "Mon, 25 Jan 2021 10:00:51 +0100")

Markus Armbruster <armbru@redhat.com> writes:

> Paolo Bonzini <pbonzini@redhat.com> writes:
>
>> HMP is using QemuOpts to parse free-form commands device_add,
>> netdev_add and object_add.  However, none of these need QemuOpts
>> for validation (these three QemuOptsLists are all of the catch-all
>> kind), and keyval is already able to parse into QDict.  So use
>> keyval directly, avoiding the detour from
>> string to QemuOpts to QDict.
>>
>> The args_type now stores the implied key.  This arguably makes more
>> sense than storing the QemuOptsList name; at least, it _is_ a key
>> that might end up in the arguments QDict.
>>
>> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
>
> Switching from QemuOpts to keyval changes the accepted language.  We may
> change it, because HMP is not a stable interface.  The commit message
> should point out the change, though.  Maybe even release notes, not
> sure.
>
> Let's recap the differences briefly:
>
> * Boolean sugar: deprecated in QemuOpts, nonexistent in keyval
>
> * QemuOpts accepts a number of more or less crazy corner cases keyval
>   rejects: invalid key, long key (silently truncated), first rather than
>   last id= wins (unlike other keys), implied key with empty value.
>
> * QemuOpts rejects anti-social ID such as id=666, keyval leaves this to
>   the caller, because key "id" is not special in keyval.
>
>   Are these still rejected with your patch?
>
>> ---
>>  hmp-commands.hx |  6 +++---
>>  monitor/hmp.c   | 20 +++++++++-----------
>>  2 files changed, 12 insertions(+), 14 deletions(-)
>>
>> diff --git a/hmp-commands.hx b/hmp-commands.hx
>> index 73e0832ea1..6ee746b53e 100644
>> --- a/hmp-commands.hx
>> +++ b/hmp-commands.hx
>> @@ -669,7 +669,7 @@ ERST
>>  
>>      {
>>          .name       = "device_add",
>> -        .args_type  = "device:O",
>> +        .args_type  = "driver:O",
>>          .params     = "driver[,prop=value][,...]",
>>          .help       = "add device, like -device on the command line",
>>          .cmd        = hmp_device_add,
>> @@ -1315,7 +1315,7 @@ ERST
>>  
>>      {
>>          .name       = "netdev_add",
>> -        .args_type  = "netdev:O",
>> +        .args_type  = "type:O",
>>          .params     = "[user|tap|socket|vde|bridge|hubport|netmap|vhost-user],id=str[,prop=value][,...]",
>>          .help       = "add host network device",
>>          .cmd        = hmp_netdev_add,
>> @@ -1343,7 +1343,7 @@ ERST
>>  
>>      {
>>          .name       = "object_add",
>> -        .args_type  = "object:O",
>> +        .args_type  = "qom-type:O",
>>          .params     = "[qom-type=]type,id=str[,prop=value][,...]",
>>          .help       = "create QOM object",
>>          .cmd        = hmp_object_add,
>> diff --git a/monitor/hmp.c b/monitor/hmp.c
>> index 6c0b33a0b1..d2cb886da5 100644
>> --- a/monitor/hmp.c
>> +++ b/monitor/hmp.c
>> @@ -744,13 +744,9 @@ static QDict *monitor_parse_arguments(Monitor *mon,
>>              break;
>>          case 'O':
>>              {
>> -                QemuOptsList *opts_list;
>> -                QemuOpts *opts;
>> +                Error *errp;

Missing initializer.  This is what causes the assertion failure reported
below.

>> +                bool help;
>>  
>> -                opts_list = qemu_find_opts(key);
>> -                if (!opts_list || opts_list->desc->name) {
>> -                    goto bad_type;
>> -                }
>>                  while (qemu_isspace(*p)) {
>>                      p++;
>>                  }
>> @@ -760,12 +756,14 @@ static QDict *monitor_parse_arguments(Monitor *mon,
>>                  if (get_str(buf, sizeof(buf), &p) < 0) {
>>                      goto fail;
>>                  }
>> -                opts = qemu_opts_parse_noisily(opts_list, buf, true);
>> -                if (!opts) {
>> -                    goto fail;
>> +                keyval_parse_into(qdict, buf, key, &help, &errp);
>> +                if (help) {
>
> Uh...
>
>> +                    if (qdict_haskey(qdict, key)) {
>
> If we parsed a value for the implied key (sugared or not),
>
>> +                        qdict_put_bool(qdict, "help", true);
>
> then encode the help request by mapping key "help" to true,
>
>> +                    } else {
>> +                        qdict_put_str(qdict, key, "help");
>
> else by mapping the implied key to "help".
>
>> +                    }
>
> Test cases:
>
> * device_add help
>
>   @qdict before the patch:
>
>     {
>         "driver": "help"
>     }
>
>   No change.
>
> * device_add e1000,help
>
>   @qdict before the patch:
>
>     {
>         "driver": "e1000",
>         "help": "on"
>     }
>
>   Afterwards:
>
>     {
>         "driver": "e1000",
>         "help": true
>     }
>
>   If this is okay, the commit message should explain it.
>
> * device_add help,e1000
>
>     {
>         "e1000": "on",
>         "driver": "help"
>     }
>
>   Afterwards:
>   upstream-qemu: ../util/error.c:59: error_setv: Assertion `*errp == NULL' failed.

Optimization masks this crash for me.

With proper initialization, I get

    {
        "driver": "help"
    }

instead.  If this change is okay, the commit message should explain it.

>
>>                  }
>> -                qemu_opts_to_qdict(opts, qdict);
>> -                qemu_opts_del(opts);
>>              }
>>              break;
>>          case '/':



  parent reply	other threads:[~2021-03-01 10:44 UTC|newest]

Thread overview: 67+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-01-18 16:30 [PATCH 00/25] qemu-option, keyval, vl: switch -object/-M/-accel to keyval parsing Paolo Bonzini
2021-01-18 16:30 ` [PATCH 01/25] qemu-option: clean up id vs. list->merge_lists Paolo Bonzini
2021-01-19 12:33   ` Kevin Wolf
2021-01-19 13:58   ` Markus Armbruster
2021-01-19 14:20     ` Paolo Bonzini
2021-01-20  8:03       ` Markus Armbruster
2021-01-20 12:37         ` Paolo Bonzini
2021-01-20 12:50           ` Markus Armbruster
2021-01-18 16:30 ` [PATCH 02/25] qemu-option: move help handling to get_opt_name_value Paolo Bonzini
2021-01-19 15:10   ` Markus Armbruster
2021-01-18 16:30 ` [PATCH 03/25] qemu-option: warn for short-form boolean options Paolo Bonzini
2021-01-19 15:56   ` Markus Armbruster
2021-01-19 17:04     ` Paolo Bonzini
2021-01-20  8:42       ` Markus Armbruster
2021-01-20 12:40         ` Paolo Bonzini
2021-01-20 12:59           ` Markus Armbruster
2021-01-20 14:05             ` Paolo Bonzini
2021-01-18 16:30 ` [PATCH 04/25] keyval: accept escaped commas in implied option Paolo Bonzini
2021-01-21 12:58   ` Markus Armbruster
2021-01-22  8:39   ` Markus Armbruster
2021-01-18 16:30 ` [PATCH 05/25] keyval: simplify keyval_parse_one Paolo Bonzini
2021-01-22 13:48   ` Markus Armbruster
2021-01-22 15:00     ` Paolo Bonzini
2021-01-22 15:44       ` Markus Armbruster
2021-01-18 16:30 ` [PATCH 06/25] tests: convert check-qom-proplist to keyval Paolo Bonzini
2021-01-22 14:14   ` Markus Armbruster
2021-01-22 14:38     ` Paolo Bonzini
2021-01-22 14:48     ` Paolo Bonzini
2021-01-18 16:30 ` [PATCH 07/25] keyval: introduce keyval_parse_into Paolo Bonzini
2021-01-22 14:22   ` Markus Armbruster
2021-01-22 14:30     ` Paolo Bonzini
2021-01-18 16:30 ` [PATCH 08/25] hmp: replace "O" parser with keyval Paolo Bonzini
2021-01-25  9:00   ` Markus Armbruster
2021-02-26 11:25     ` Paolo Bonzini
2021-03-01 10:14       ` Markus Armbruster
2021-03-01 10:23         ` Paolo Bonzini
2021-03-01 13:35           ` Markus Armbruster
2021-03-01 10:43     ` Markus Armbruster [this message]
2021-03-01 11:54       ` Paolo Bonzini
2021-01-18 16:30 ` [PATCH 09/25] qom: use qemu_printf to print help for user-creatable objects Paolo Bonzini
2021-01-25 12:47   ` Markus Armbruster
2021-01-18 16:30 ` [PATCH 10/25] hmp: special case help options for object_add Paolo Bonzini
2021-01-25 12:48   ` Markus Armbruster
2021-01-25 12:49     ` Paolo Bonzini
2021-01-25 14:02       ` Markus Armbruster
2021-01-18 16:30 ` [PATCH 11/25] remove -writeconfig Paolo Bonzini
2021-01-25 12:53   ` Markus Armbruster
2021-01-25 14:01     ` Paolo Bonzini
2021-01-25 14:12       ` Daniel P. Berrangé
2021-01-18 16:31 ` [PATCH 12/25] qemu-config: add error propagation to qemu_config_parse Paolo Bonzini
2021-01-25 13:54   ` Markus Armbruster
2021-01-18 16:31 ` [PATCH 13/25] qemu-option: support accept-any QemuOptsList in qemu_opts_absorb_qdict Paolo Bonzini
2021-01-18 16:31 ` [PATCH 14/25] qemu-config: parse configuration files to a QDict Paolo Bonzini
2021-01-18 16:31 ` [PATCH 15/25] vl: plumb keyval-based options into -set and -readconfig Paolo Bonzini
2021-01-25 11:48   ` Markus Armbruster
2021-01-25 13:59     ` Paolo Bonzini
2021-01-18 16:31 ` [PATCH 16/25] qom: do not modify QDict argument in user_creatable_add_dict Paolo Bonzini
2021-01-18 16:31 ` [PATCH 17/25] qemu-io: use keyval for -object parsing Paolo Bonzini
2021-01-18 16:31 ` [PATCH 18/25] qemu-nbd: " Paolo Bonzini
2021-01-18 16:31 ` [PATCH 19/25] qemu-img: " Paolo Bonzini
2021-01-18 16:31 ` [PATCH 20/25] qemu: " Paolo Bonzini
2021-01-18 16:31 ` [PATCH 21/25] storage-daemon: do not register the "object" group with QemuOpts Paolo Bonzini
2021-01-18 16:31 ` [PATCH 22/25] qom: export more functions for use with non-UserCreatable objects Paolo Bonzini
2021-01-18 16:31 ` [PATCH 23/25] vl: switch -M parsing to keyval Paolo Bonzini
2021-01-18 16:31 ` [PATCH 24/25] qemu-option: remove now-dead code Paolo Bonzini
2021-01-18 16:31 ` [PATCH 25/25] vl: switch -accel parsing to keyval Paolo Bonzini
2021-01-18 17:18 ` [PATCH 00/25] qemu-option, keyval, vl: switch -object/-M/-accel to keyval parsing no-reply

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=87k0qrjgrp.fsf@dusky.pond.sub.org \
    --to=armbru@redhat.com \
    --cc=imammedo@redhat.com \
    --cc=kwolf@redhat.com \
    --cc=pbonzini@redhat.com \
    --cc=qemu-devel@nongnu.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).