All of lore.kernel.org
 help / color / mirror / Atom feed
From: Markus Armbruster <armbru@redhat.com>
To: Eric Blake <eblake@redhat.com>
Cc: qemu-devel@nongnu.org, Michael Roth <mdroth@linux.vnet.ibm.com>
Subject: Re: [Qemu-devel] [PATCH v15 02/23] qapi: Guarantee NULL obj on input visitor callback error
Date: Thu, 28 Apr 2016 14:24:52 +0200	[thread overview]
Message-ID: <87pot9or8r.fsf@dusky.pond.sub.org> (raw)
In-Reply-To: <1461801715-24307-3-git-send-email-eblake@redhat.com> (Eric Blake's message of "Wed, 27 Apr 2016 18:01:34 -0600")

Eric Blake <eblake@redhat.com> writes:

> Our existing input visitors were not very consistent on errors
> in a function taking 'TYPE **obj' (that is, start_struct(),
> start_alternate(), next_list(), type_str(), and type_any()).
> While all of them set '*obj' to allocated storage on success,
> it was not obvious whether '*obj' was guaranteed safe on failure,
> or whether it was left uninitialized.  But a future patch wants
> to guarantee that visit_type_FOO() does not leak a partially-
> constructed obj back to the caller; it is easier to implement
> this if we can reliably state that '*obj' is assigned on exit,
> even on failures.  Add assertions to enforce it.

I had to read this several times, because by now I've forgotten that
we're talking about input visitors only.  Easy enough to avoid: ... that
input visitors assign to *obj regardless of success or failure.

Begs the question what is assigned to it on failure, though.

>
> The opts-visitor start_struct() doesn't set an error, but it
> also was doing a weird check for 0 size; all callers pass in
> non-zero size if obj is non-NULL.
>
> The testsuite has at least one spot where we no longer need
> to pre-initialize a variable prior to a visit; valgrind confirms
> that the test is still fine with the cleanup.
>
> A later patch will document the design constraint implemented
> here.
>
> Signed-off-by: Eric Blake <eblake@redhat.com>
>
> ---
> v15: enhance commit message, hoist assertions from later in series
> v14: no change
> v13: no change
> v12: new patch
> ---
>  qapi/qapi-visit-core.c        | 34 ++++++++++++++++++++++++++++++----
>  qapi/opts-visitor.c           |  3 ++-
>  qapi/qmp-input-visitor.c      |  4 ++++
>  qapi/string-input-visitor.c   |  1 +
>  tests/test-qmp-input-strict.c |  2 +-
>  5 files changed, 38 insertions(+), 6 deletions(-)
>
> diff --git a/qapi/qapi-visit-core.c b/qapi/qapi-visit-core.c
> index 3cd7edc..3a131ce 100644
> --- a/qapi/qapi-visit-core.c
> +++ b/qapi/qapi-visit-core.c
> @@ -23,7 +23,13 @@
>  void visit_start_struct(Visitor *v, const char *name, void **obj,
>                          size_t size, Error **errp)
>  {
> -    v->start_struct(v, name, obj, size, errp);
> +    Error *err = NULL;
> +
> +    v->start_struct(v, name, obj, size, &err);
> +    if (obj && v->type == VISITOR_INPUT) {
> +        assert(err || *obj);
> +    }
> +    error_propagate(errp, err);
>  }

The commit message claims you're adding assertions to enforce input
visitors assign *obj even on failure.  This assertion doesn't do that.
It enforces "on success, *obj is non-null".  Is that what you want?  Or
do you actually want something like "either err or *obj are non-null"?
I.e.

           assert(!err != !*obj);

>
>  void visit_end_struct(Visitor *v, Error **errp)
> @@ -51,9 +57,15 @@ void visit_start_alternate(Visitor *v, const char *name,
>                             GenericAlternate **obj, size_t size,
>                             bool promote_int, Error **errp)
>  {
> +    Error *err = NULL;
> +
>      assert(obj && size >= sizeof(GenericAlternate));
>      if (v->start_alternate) {
> -        v->start_alternate(v, name, obj, size, promote_int, errp);
> +        v->start_alternate(v, name, obj, size, promote_int, &err);
> +        if (v->type == VISITOR_INPUT) {
> +            assert(err || *obj);
> +        }
> +        error_propagate(errp, err);
>      }
>  }
>

Hmm, you check the postcondition only when v implements
start_alternate().  Shouldn't it hold regardless of v?  If yes, then
let's check it regardless of v:

       if (v->start_alternate) {
           v->start_alternate(v, name, obj, size, promote_int, &err);
       }
       if (v->type == VISITOR_INPUT) {
           assert(err || *obj);
       }
       error_propagate(errp, err);

But that makes it pretty obvious that the postcondition won't hold when
!v->start_alternate.  May v->start_alternate() be null for an input
visitor?  According to visitor-impl.h, it may not.  Okay.

> @@ -188,7 +200,14 @@ void visit_type_bool(Visitor *v, const char *name, bool *obj, Error **errp)
>
>  void visit_type_str(Visitor *v, const char *name, char **obj, Error **errp)
>  {
> -    v->type_str(v, name, obj, errp);
> +    Error *err = NULL;
> +
> +    assert(obj);
> +    v->type_str(v, name, obj, &err);
> +    if (v->type == VISITOR_INPUT) {
> +        assert(err || *obj);
> +    }
> +    error_propagate(errp, err);
>  }
>
>  void visit_type_number(Visitor *v, const char *name, double *obj,
> @@ -199,7 +218,14 @@ void visit_type_number(Visitor *v, const char *name, double *obj,
>
>  void visit_type_any(Visitor *v, const char *name, QObject **obj, Error **errp)
>  {
> -    v->type_any(v, name, obj, errp);
> +    Error *err = NULL;
> +
> +    assert(obj);
> +    v->type_any(v, name, obj, &err);
> +    if (v->type == VISITOR_INPUT) {
> +        assert(err || *obj);
> +    }
> +    error_propagate(errp, err);
>  }
>

The commit message lists start_struct(), start_alternate(), next_list(),
type_str(), and type_any().  You cover them except for next_list().  Why
is that missing?

>  static void output_type_enum(Visitor *v, const char *name, int *obj,
> diff --git a/qapi/opts-visitor.c b/qapi/opts-visitor.c
> index 66aeaed..4cb6436 100644
> --- a/qapi/opts-visitor.c
> +++ b/qapi/opts-visitor.c
> @@ -133,7 +133,7 @@ opts_start_struct(Visitor *v, const char *name, void **obj,
>      const QemuOpt *opt;
>
>      if (obj) {
> -        *obj = g_malloc0(size > 0 ? size : 1);
> +        *obj = g_malloc0(size);
>      }
>      if (ov->depth++ > 0) {
>          return;
> @@ -314,6 +314,7 @@ opts_type_str(Visitor *v, const char *name, char **obj, Error **errp)
>
>      opt = lookup_scalar(ov, name, errp);
>      if (!opt) {
> +        *obj = NULL;
>          return;
>      }
>      *obj = g_strdup(opt->str ? opt->str : "");
> diff --git a/qapi/qmp-input-visitor.c b/qapi/qmp-input-visitor.c
> index 02d4233..77cce8b 100644
> --- a/qapi/qmp-input-visitor.c
> +++ b/qapi/qmp-input-visitor.c
> @@ -120,6 +120,9 @@ static void qmp_input_start_struct(Visitor *v, const char *name, void **obj,
>      QObject *qobj = qmp_input_get_object(qiv, name, true);
>      Error *err = NULL;
>
> +    if (obj) {
> +        *obj = NULL;
> +    }
>      if (!qobj || qobject_type(qobj) != QTYPE_QDICT) {
>          error_setg(errp, QERR_INVALID_PARAMETER_TYPE, name ? name : "null",
>                     "QDict");
> @@ -267,6 +270,7 @@ static void qmp_input_type_str(Visitor *v, const char *name, char **obj,
>      QString *qstr = qobject_to_qstring(qmp_input_get_object(qiv, name, true));
>
>      if (!qstr) {
> +        *obj = NULL;
>          error_setg(errp, QERR_INVALID_PARAMETER_TYPE, name ? name : "null",
>                     "string");
>          return;
> diff --git a/qapi/string-input-visitor.c b/qapi/string-input-visitor.c
> index d604575..797973a 100644
> --- a/qapi/string-input-visitor.c
> +++ b/qapi/string-input-visitor.c
> @@ -293,6 +293,7 @@ static void parse_type_str(Visitor *v, const char *name, char **obj,
>      if (siv->string) {
>          *obj = g_strdup(siv->string);
>      } else {
> +        *obj = NULL;
>          error_setg(errp, QERR_INVALID_PARAMETER_TYPE, name ? name : "null",
>                     "string");
>      }
> diff --git a/tests/test-qmp-input-strict.c b/tests/test-qmp-input-strict.c
> index d71727e..d5f80ec 100644
> --- a/tests/test-qmp-input-strict.c
> +++ b/tests/test-qmp-input-strict.c
> @@ -263,7 +263,7 @@ static void test_validate_fail_union_flat_no_discrim(TestInputVisitorData *data,
>  static void test_validate_fail_alternate(TestInputVisitorData *data,
>                                           const void *unused)
>  {
> -    UserDefAlternate *tmp = NULL;
> +    UserDefAlternate *tmp;
>      Visitor *v;
>      Error *err = NULL;

  reply	other threads:[~2016-04-28 12:25 UTC|newest]

Thread overview: 49+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-04-28  0:01 [Qemu-devel] [PATCH v15 00/23] qapi visitor cleanups (post-introspection cleanups subset E) Eric Blake
2016-04-28  0:01 ` [Qemu-devel] [PATCH v15 01/23] qapi-visit: Add visitor.type classification Eric Blake
2016-04-28  0:01 ` [Qemu-devel] [PATCH v15 02/23] qapi: Guarantee NULL obj on input visitor callback error Eric Blake
2016-04-28 12:24   ` Markus Armbruster [this message]
2016-04-28 13:00     ` Eric Blake
2016-04-28 15:41       ` Eric Blake
2016-04-28 16:02   ` [Qemu-devel] [PATCH v15 02A/23] fixup! " Eric Blake
2016-04-28  0:01 ` [Qemu-devel] [PATCH v15 03/23] qmp: Drop dead command->type Eric Blake
2016-04-28  0:01 ` [Qemu-devel] [PATCH v15 04/23] qmp-input: Clean up stack handling Eric Blake
2016-04-28  0:01 ` [Qemu-devel] [PATCH v15 05/23] qapi: Use strict QMP input visitor in more places Eric Blake
2016-04-28 13:06   ` Markus Armbruster
2016-04-28 14:28     ` Eric Blake
2016-04-28  0:01 ` [Qemu-devel] [PATCH v15 06/23] qmp-input: Don't consume input when checking has_member Eric Blake
2016-04-28  0:01 ` [Qemu-devel] [PATCH v15 07/23] qapi-commands: Wrap argument visit in visit_start_struct Eric Blake
2016-04-28  0:01 ` [Qemu-devel] [PATCH v15 08/23] monitor: Let generated code validate arguments Eric Blake
2016-04-28 14:09   ` Markus Armbruster
2016-04-28 14:39     ` Marc-André Lureau
2016-04-28 18:00       ` Markus Armbruster
2016-04-28 18:58         ` Eric Blake
2016-04-28 14:47     ` Eric Blake
2016-04-28  0:01 ` [Qemu-devel] [PATCH v15 09/23] qom: Wrap prop visit in visit_start_struct Eric Blake
2016-04-28 14:46   ` Markus Armbruster
2016-04-28 15:14     ` Eric Blake
2016-04-28  0:01 ` [Qemu-devel] [PATCH v15 10/23] qmp-input: Require struct push to visit members of top dict Eric Blake
2016-04-28 15:00   ` Markus Armbruster
2016-04-28 15:04     ` Eric Blake
2016-04-28  0:01 ` [Qemu-devel] [PATCH v15 11/23] qmp-input: Refactor when list is advanced Eric Blake
2016-04-28 15:19   ` Markus Armbruster
2016-04-28  0:01 ` [Qemu-devel] [PATCH v15 12/23] qapi: Document visitor interfaces, add assertions Eric Blake
2016-04-28 16:34   ` Markus Armbruster
2016-04-28 19:02     ` Eric Blake
2016-04-28  0:01 ` [Qemu-devel] [PATCH v15 13/23] tests: Add check-qnull Eric Blake
2016-04-28  0:01 ` [Qemu-devel] [PATCH v15 14/23] qapi: Add visit_type_null() visitor Eric Blake
2016-04-28 16:40   ` Markus Armbruster
2016-04-28  0:01 ` [Qemu-devel] [PATCH v15 15/23] qmp: Support explicit null during visits Eric Blake
2016-04-28 16:50   ` Markus Armbruster
2016-04-28 19:07     ` Eric Blake
2016-04-28  0:01 ` [Qemu-devel] [PATCH v15 16/23] spapr_drc: Expose 'null' in qom-get when there is no fdt Eric Blake
2016-04-28  0:01 ` [Qemu-devel] [PATCH v15 17/23] qmp: Add qmp_output_visitor_reset() Eric Blake
2016-04-28  0:01 ` [Qemu-devel] [PATCH v15 18/23] qmp: Tighten output visitor rules Eric Blake
2016-04-28  0:01 ` [Qemu-devel] [PATCH v15 19/23] qapi: Split visit_end_struct() into pieces Eric Blake
2016-04-28  0:01 ` [Qemu-devel] [PATCH v15 20/23] tests/string-input-visitor: Add negative integer tests Eric Blake
2016-04-28  0:01 ` [Qemu-devel] [PATCH v15 21/23] qapi: Fix string input visitor handling of invalid list Eric Blake
2016-04-28 17:18   ` Markus Armbruster
2016-04-28  0:01 ` [Qemu-devel] [PATCH v15 22/23] qapi: Simplify semantics of visit_next_list() Eric Blake
2016-04-28 15:44   ` Eric Blake
2016-04-28  0:01 ` [Qemu-devel] [PATCH v15 23/23] qapi: Change visit_type_FOO() to no longer return partial objects Eric Blake
2016-04-28 17:42   ` Markus Armbruster
2016-04-28 18:03 ` [Qemu-devel] [PATCH v15 00/23] qapi visitor cleanups (post-introspection cleanups subset E) Markus Armbruster

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=87pot9or8r.fsf@dusky.pond.sub.org \
    --to=armbru@redhat.com \
    --cc=eblake@redhat.com \
    --cc=mdroth@linux.vnet.ibm.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.