qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Paolo Bonzini <pbonzini@redhat.com>
To: Richard Henderson <rth@twiddle.net>
Cc: Liu Ping Fan <qemulist@gmail.com>,
	Anthony Liguori <anthony@codemonkey.ws>,
	qemu-devel@nongnu.org
Subject: Re: [Qemu-devel] [PATCH v2 1/2] add a header file for atomic operations
Date: Tue, 18 Jun 2013 17:04:13 +0200	[thread overview]
Message-ID: <51C076ED.8080501@redhat.com> (raw)
In-Reply-To: <51C070E2.4050207@twiddle.net>

Il 18/06/2013 16:38, Richard Henderson ha scritto:
>>>> +#ifndef atomic_read
>>>> +#define atomic_read(ptr)       (*(__typeof__(*ptr) *volatile) (ptr))
>>>>  #endif
>>>>  
>>>> +#ifndef atomic_set
>>>> +#define atomic_set(ptr, i)     ((*(__typeof__(*ptr) *volatile) (ptr)) = (i))
>>>> +#endif
>>>
>>> Use
>>>
>>> __atomic_load(..., __ATOMIC_RELAXED)
>>> __atomic_store(..., __ATOMIC_RELAXED)
>>>
>>> ?
>>
>> Same here, I didn't want proliferation of #ifdefs beyond what is actually required.
> 
> Not knowing exactly where these might be used within the code base, I'd be
> worried about someone applying them to a uint64_t, somewhere a 32-bit host
> might see it.  At which point the above is going to be silently wrong, loaded
> with two 32-bit pieces.
> 
> Given that we're not requiring gcc 4.8, and cannot guarantee use of
> __atomic_load, perhaps we ought to do something like
> 
> #define atomic_read(ptr) \
>   ({ if (sizeof(*(ptr)) > sizeof(ptr)) invalid_atomic_read(); \
>      *(__typeof__(*ptr) *volatile) (ptr)); })
> 
> which should generate a link error when reading a size we can't guarantee will
> Just Work.

Good idea.

>> The FAQ also has an "important note", however:
>>
>>     Important Note: Note that it is important for both threads to access
>>     the same volatile variable in order to properly set up the happens-before
>>     relationship. It is not the case that everything visible to thread A
>>     when it writes volatile field f becomes visible to thread B after it
>>     reads volatile field g. The release and acquire have to "match" (i.e.,
>>     be performed on the same volatile field) to have the right semantics. 
>>
>> Is this final "important note" the difference between ACQ_REL and SEQ_CST?
> 
> Yes, exactly.

I still have some confusion, so I sent another follow-up with the exact
differences in the generated code.  But in any case, I will augment the
documentation with the text from the FAQ, seems safe.

Paolo

  reply	other threads:[~2013-06-18 15:04 UTC|newest]

Thread overview: 41+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2013-06-16 11:21 [Qemu-devel] [PATCH v2 0/2] make AioContext's bh re-entrant Liu Ping Fan
2013-06-16 11:21 ` [Qemu-devel] [PATCH v2 1/2] add a header file for atomic operations Liu Ping Fan
2013-06-17 18:57   ` Richard Henderson
2013-06-18  8:36     ` Paolo Bonzini
2013-06-18 11:03       ` Paolo Bonzini
2013-06-18 14:38       ` Richard Henderson
2013-06-18 15:04         ` Paolo Bonzini [this message]
2013-06-18 13:24     ` [Qemu-devel] Java volatile vs. C11 seq_cst (was Re: [PATCH v2 1/2] add a header file for atomic operations) Paolo Bonzini
2013-06-18 14:50       ` Paul E. McKenney
2013-06-18 15:29         ` Peter Sewell
2013-06-18 15:37         ` Torvald Riegel
2013-06-19  1:53           ` Paul E. McKenney
2013-06-19  7:11             ` Torvald Riegel
2013-06-20 15:18               ` Paul E. McKenney
2013-06-18 16:08         ` Paolo Bonzini
2013-06-18 16:38           ` Torvald Riegel
2013-06-19  1:57             ` Paul E. McKenney
2013-06-19  9:31             ` Paolo Bonzini
2013-06-19 13:15               ` Torvald Riegel
2013-06-19 15:14                 ` Paolo Bonzini
2013-06-19 20:25                   ` Torvald Riegel
2013-06-20  7:53                     ` Paolo Bonzini
2013-06-22 10:55                       ` Torvald Riegel
2013-06-18 15:26       ` Torvald Riegel
2013-06-18 17:38       ` Andrew Haley
2013-06-19  9:30         ` Paolo Bonzini
2013-06-19 15:36           ` Andrew Haley
2013-06-16 11:21 ` [Qemu-devel] [PATCH v2 2/2] QEMUBH: make AioContext's bh re-entrant Liu Ping Fan
2013-06-17 15:28   ` Stefan Hajnoczi
2013-06-17 16:41     ` Paolo Bonzini
2013-06-18  2:19     ` liu ping fan
2013-06-18  9:31       ` Stefan Hajnoczi
2013-06-18 15:14   ` mdroth
2013-06-18 16:19     ` mdroth
2013-06-18 19:20     ` Paolo Bonzini
2013-06-18 22:26       ` mdroth
2013-06-19  9:27         ` Paolo Bonzini
2013-06-20  9:11           ` Stefan Hajnoczi
2013-06-17  7:11 ` [Qemu-devel] [PATCH v2 0/2] " Paolo Bonzini
2013-06-18  2:40   ` liu ping fan
2013-06-18  8:36     ` Paolo Bonzini

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=51C076ED.8080501@redhat.com \
    --to=pbonzini@redhat.com \
    --cc=anthony@codemonkey.ws \
    --cc=qemu-devel@nongnu.org \
    --cc=qemulist@gmail.com \
    --cc=rth@twiddle.net \
    /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).