From: Eric Blake <eblake@redhat.com>
To: Max Reitz <mreitz@redhat.com>, qemu-block@nongnu.org
Cc: qemu-devel@nongnu.org, Markus Armbruster <armbru@redhat.com>,
Alberto Garcia <berto@igalia.com>,
Michael Roth <mdroth@linux.vnet.ibm.com>,
Peter Maydell <peter.maydell@linaro.org>
Subject: Re: [Qemu-devel] [PATCH v3 2/7] qapi: Add qobject_to()
Date: Mon, 19 Mar 2018 14:36:55 -0500 [thread overview]
Message-ID: <24d6fa65-3570-4eff-fce4-23a5b0a5a2f8@redhat.com> (raw)
In-Reply-To: <afd63f12-6b1c-1981-ac3a-52d108725e2d@redhat.com>
On 02/26/2018 05:58 AM, Max Reitz wrote:
> On 2018-02-24 21:57, Eric Blake wrote:
>> On 02/24/2018 09:40 AM, Max Reitz wrote:
>>> This is a dynamic casting macro that, given a QObject type, returns an
>>> object as that type or NULL if the object is of a different type (or
>>> NULL itself).
>>>
>>> +#define qobject_to(obj, type) \
>>> + container_of(qobject_check_type(obj, glue(QTYPE_CAST_TO_, type))
>>> ?: \
>>> + QOBJECT((type *)NULL), \
>>
>> I guess the third (second?) branch of the ternary is written this way,
>> rather than the simpler 'NULL', to ensure that 'type' is still something
>> that can have the QOBJECT() macro applied to it? Should be okay.
>
> It's written this way because of the container_of() around it. We want
> the whole expression to return NULL then, and without the QOBJECT()
> around it, it would only return NULL if offsetof(type, base) == 0 (which
> it is not necessarily).
>
> OTOH, container_of(&((type *)NULL)->base, type, base) is by definition NULL.
>
> (QOBJECT(x) is &(x)->base)
Well, clang's ubsan griped:
https://lists.gnu.org/archive/html/qemu-devel/2018-03/msg05143.html
Practically, all of our qtypes have 'base' at offset 0, which means
(QObject*)addr and (QString*)addr are the same address, even when addr
is NULL. But neither QOBJECT() nor container_of() are currently fit to
run on a NULL pointer, since the 'base' member need not be at offset 0,
at which point, we'd be converting away from the NULL pointer on the
&(x)->base conversion, and then back to the NULL pointer on the
container_of() conversion. So at the end of the day, we get the right
results, but we relied on undefined behavior in the interim.
So here's what I'm squashing in, if you like it (and remembering that I
already swapped argument order to be qobject_to(type, obj) in my pending
pull requests):
diff --git i/include/qapi/qmp/qobject.h w/include/qapi/qmp/qobject.h
index ea9702270e7..e6ce9347ab8 100644
--- i/include/qapi/qmp/qobject.h
+++ w/include/qapi/qmp/qobject.h
@@ -62,9 +62,8 @@ QEMU_BUILD_BUG_MSG(QTYPE__MAX != 7,
"The QTYPE_CAST_TO_* list needs to be extended");
#define qobject_to(type, obj) \
- container_of(qobject_check_type(obj, glue(QTYPE_CAST_TO_, type)) ?: \
- QOBJECT((type *)NULL), \
- type, base)
+ ({ QObject *_tmp = qobject_check_type(obj, glue(QTYPE_CAST_TO_,
type)); \
+ _tmp ? container_of(_tmp, type, base) : (type *)NULL; })
/* Initialize an object to default values */
static inline void qobject_init(QObject *obj, QType type)
--
Eric Blake, Principal Software Engineer
Red Hat, Inc. +1-919-301-3266
Virtualization: qemu.org | libvirt.org
next prev parent reply other threads:[~2018-03-19 19:37 UTC|newest]
Thread overview: 33+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-02-24 15:40 [Qemu-devel] [PATCH v3 0/7] block: Handle null backing link Max Reitz
2018-02-24 15:40 ` [Qemu-devel] [PATCH v3 1/7] compiler: Add QEMU_BUILD_BUG_MSG() macro Max Reitz
2018-02-24 20:08 ` Eric Blake
2018-02-27 13:33 ` Alberto Garcia
2018-02-24 15:40 ` [Qemu-devel] [PATCH v3 2/7] qapi: Add qobject_to() Max Reitz
2018-02-24 20:57 ` Eric Blake
2018-02-26 11:58 ` Max Reitz
2018-03-19 19:36 ` Eric Blake [this message]
2018-03-19 19:38 ` Max Reitz
2018-02-27 13:45 ` Alberto Garcia
2018-02-24 15:40 ` [Qemu-devel] [PATCH v3 3/7] qapi: Replace qobject_to_X(o) by qobject_to(o, X) Max Reitz
2018-02-24 21:04 ` Eric Blake
2018-02-26 12:01 ` Max Reitz
2018-02-27 14:47 ` Eric Blake
2018-02-28 18:08 ` Max Reitz
2018-03-10 21:48 ` Eric Blake
2018-02-24 15:40 ` [Qemu-devel] [PATCH v3 4/7] qapi: Remove qobject_to_X() functions Max Reitz
2018-02-24 21:05 ` Eric Blake
2018-02-24 15:40 ` [Qemu-devel] [PATCH v3 5/7] qapi: Make more of qobject_to() Max Reitz
2018-02-24 21:07 ` Eric Blake
2018-02-24 15:40 ` [Qemu-devel] [PATCH v3 6/7] block: Handle null backing link Max Reitz
2018-02-24 21:09 ` Eric Blake
2018-02-24 15:40 ` [Qemu-devel] [PATCH v3 7/7] block: Deprecate "backing": "" Max Reitz
2018-02-24 18:02 ` [Qemu-devel] [PATCH v3 0/7] block: Handle null backing link no-reply
2018-02-26 11:50 ` Max Reitz
2018-02-24 19:51 ` no-reply
2018-02-24 20:19 ` no-reply
2018-02-24 20:30 ` no-reply
2018-02-25 23:38 ` no-reply
2018-02-26 6:11 ` no-reply
2018-03-09 20:11 ` Eric Blake
2018-03-10 22:34 ` Eric Blake
2018-03-12 12:22 ` Max Reitz
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=24d6fa65-3570-4eff-fce4-23a5b0a5a2f8@redhat.com \
--to=eblake@redhat.com \
--cc=armbru@redhat.com \
--cc=berto@igalia.com \
--cc=mdroth@linux.vnet.ibm.com \
--cc=mreitz@redhat.com \
--cc=peter.maydell@linaro.org \
--cc=qemu-block@nongnu.org \
--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).