From: Alan Stern <stern@rowland.harvard.edu>
To: Pavel Machek <pavel@ucw.cz>
Cc: Linus Torvalds <torvalds@linux-foundation.org>,
Alan Cox <alan@lxorguk.ukuu.org.uk>,
"Paul E. McKenney" <paulmck@linux.vnet.ibm.com>,
Andrew Morton <akpm@linux-foundation.org>,
Zdenek Kabelac <zdenek.kabelac@gmail.com>, <davem@davemloft.net>,
"Rafael J. Wysocki" <rjw@sisk.pl>,
Pierre Ossman <drzeus-mmc@drzeus.cx>,
Kernel development list <linux-kernel@vger.kernel.org>,
pm list <linux-pm@lists.linux-foundation.org>
Subject: Re: [patch] Re: using long instead of atomic_t when only set/read is required
Date: Mon, 3 Mar 2008 14:27:12 -0500 (EST) [thread overview]
Message-ID: <Pine.LNX.4.44L0.0803031422080.3611-100000@iolanthe.rowland.org> (raw)
In-Reply-To: <20080303174422.GB13869@elf.ucw.cz>
On Mon, 3 Mar 2008, Pavel Machek wrote:
> Ok... can we get Alan Stern's patch into Documentation/atomic_ops.txt
> , then? I was not aware of this, and there seems to be lot of
> confusion around...
>
> Plus... I really don't think we can "just access" this as normal
> pointers... due to the compiler issues Alan Cox mentioned, and due to
> the ACCESS_ONCE() issue.
Here's an updated version of the patch, including the issue Alan Cox
brought up.
Alan Stern
-----------------------------------------------------------------
Atomicity of reads of write for pointers and integral types (other than
long long) should be documented, along with the limitations imposed by
the compiler.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
---
Index: usb-2.6/Documentation/atomic_ops.txt
===================================================================
--- usb-2.6.orig/Documentation/atomic_ops.txt
+++ usb-2.6/Documentation/atomic_ops.txt
@@ -21,6 +21,24 @@ local_t is very similar to atomic_t. If
updated by one CPU, local_t is probably more appropriate. Please see
Documentation/local_ops.txt for the semantics of local_t.
+For all properly-aligned pointer and integral types other than long
+long, the kernel requires simple reads and writes to be atomic with
+respect to each other. That is, if one CPU reads a pointer value at
+the same time as another CPU overwrites the pointer, it is guaranteed
+that the reader will obtain either the old or the new value of the
+pointer, never some mish-mash combination of the two. Likewise, if
+one CPU writes a long value at the same time as another CPU does, it
+is guaranteed that one or the other of the values will end up stored
+in memory, not some mish-mash combination of bits.
+
+Thus, if all you need is atomicity of reading and writing then you can
+use plain old ints, longs, or pointers; you don't need to use
+atomic_t. But note: This guarantee emphatically does not apply to
+long long values or unaligned values! Note also that gcc does not
+guarantee to compile all C assignment expressions into simple writes.
+For example, a statement like "x = a + b" might cause gcc to emit code
+equivalent to "x = a; x += b", which is decidedly non-atomic.
+
The first operations to implement for atomic_t's are the initializers and
plain reads.
next prev parent reply other threads:[~2008-03-03 19:27 UTC|newest]
Thread overview: 93+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <20080219232338.E9A24108068@picon.linux-foundation.org>
2008-02-20 16:42 ` [Bug 10030] Suspend doesn't work when SD card is inserted Alan Stern
2008-02-20 17:30 ` Pierre Ossman
2008-02-20 19:26 ` Alan Stern
2008-02-20 20:51 ` Pierre Ossman
2008-02-20 20:58 ` Rafael J. Wysocki
2008-02-20 21:06 ` Alan Stern
2008-02-20 22:15 ` Rafael J. Wysocki
2008-02-20 22:24 ` Alan Stern
2008-02-20 22:41 ` Rafael J. Wysocki
2008-02-21 0:02 ` Rafael J. Wysocki
2008-02-21 16:27 ` Alan Stern
2008-02-21 16:38 ` Rafael J. Wysocki
2008-02-21 17:48 ` Alan Stern
2008-02-21 22:47 ` Rafael J. Wysocki
2008-02-21 23:05 ` Alan Stern
2008-02-23 1:30 ` Rafael J. Wysocki
2008-02-23 4:39 ` Alan Stern
2008-02-23 20:16 ` Rafael J. Wysocki
2008-02-23 23:29 ` Alan Stern
2008-02-24 0:19 ` Rafael J. Wysocki
2008-02-24 3:25 ` Alan Stern
2008-02-24 4:26 ` [linux-pm] " Alan Stern
2008-02-24 14:00 ` Rafael J. Wysocki
2008-02-24 15:33 ` Bugs in MMC [was: [Bug 10030] Suspend doesn't work when SD card is inserted] Alan Stern
2008-02-25 17:41 ` Pierre Ossman
2008-02-25 17:58 ` Alan Stern
2008-02-25 18:31 ` Pierre Ossman
2008-02-25 20:00 ` Alan Stern
2008-03-01 14:11 ` Pierre Ossman
2008-03-01 14:36 ` Alan Stern
2008-03-01 14:47 ` Pierre Ossman
2008-02-25 22:51 ` Felipe Balbi
2008-03-03 21:59 ` David Brownell
2008-03-04 6:03 ` Pierre Ossman
2008-03-04 9:44 ` David Brownell
2008-03-04 9:58 ` Pierre Ossman
2008-03-06 21:23 ` David Brownell
2008-03-04 17:53 ` Alan Stern
2008-03-04 18:53 ` David Brownell
2008-03-04 19:51 ` Alan Stern
2008-03-04 20:30 ` David Brownell
2008-03-04 21:00 ` Alan Stern
2008-03-06 15:55 ` Pavel Machek
2008-03-06 20:33 ` Alan Stern
2008-03-06 20:53 ` Zdenek Kabelac
2008-03-06 21:31 ` Rafael J. Wysocki
2008-03-04 17:50 ` Alan Stern
2008-02-24 13:33 ` [Bug 10030] Suspend doesn't work when SD card is inserted Rafael J. Wysocki
2008-02-24 20:25 ` Rafael J. Wysocki
2008-02-24 20:45 ` Alan Stern
2008-02-24 20:56 ` Rafael J. Wysocki
2008-02-24 21:11 ` Alan Stern
2008-02-24 22:18 ` Rafael J. Wysocki
2008-02-24 13:51 ` Rafael J. Wysocki
2008-02-24 19:27 ` Alan Stern
2008-02-24 19:42 ` Zdenek Kabelac
2008-02-24 20:09 ` Rafael J. Wysocki
2008-02-24 18:21 ` Pavel Machek
2008-02-24 19:03 ` Alan Stern
2008-02-24 20:11 ` Pavel Machek
2008-02-24 20:33 ` Alan Stern
2008-02-24 21:42 ` Pavel Machek
2008-02-24 22:21 ` Rafael J. Wysocki
2008-02-25 2:21 ` Alan Stern
2008-02-25 11:41 ` Rafael J. Wysocki
[not found] ` <c4e36d110802250402q7312a488idf752e07db7504e8@mail.gmail.com>
2008-02-25 12:31 ` Rafael J. Wysocki
2008-02-25 2:19 ` Alan Stern
2008-02-25 9:03 ` using long instead of atomic_t when only set/read is required (was Re: [Bug 10030] Suspend doesn't work when SD card is inserted) Pavel Machek
2008-02-25 14:46 ` Alan Stern
2008-03-03 12:08 ` [patch] Re: using long instead of atomic_t when only set/read is required Pavel Machek
2008-03-03 15:42 ` Alan Stern
2008-03-03 15:53 ` Alan Cox
2008-03-03 17:11 ` Alan Stern
2008-03-03 17:26 ` Linus Torvalds
2008-03-03 17:44 ` Pavel Machek
2008-03-03 19:27 ` Alan Stern [this message]
2008-03-06 15:58 ` Mark Lord
2008-03-06 16:11 ` Linus Torvalds
2008-03-06 16:27 ` Mark Lord
2008-03-03 17:16 ` Nick Piggin
2008-03-03 17:31 ` Paul E. McKenney
2008-03-03 17:33 ` Alan Cox
2008-03-03 17:22 ` Paul E. McKenney
2008-03-03 15:48 ` Alan Cox
2008-03-03 17:24 ` Pavel Machek
2008-03-03 20:27 ` Rafael J. Wysocki
2008-03-03 21:12 ` Paul E. McKenney
2008-03-03 22:23 ` Linus Torvalds
2008-03-04 23:32 ` Peter Hartley
2008-03-04 23:54 ` Rafael J. Wysocki
2008-03-05 0:26 ` Paul E. McKenney
2008-02-25 11:40 ` [Bug 10030] Suspend doesn't work when SD card is inserted Rafael J. Wysocki
2008-03-05 17:07 [patch] Re: using long instead of atomic_t when only set/read is required Matti Linnanvuori
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=Pine.LNX.4.44L0.0803031422080.3611-100000@iolanthe.rowland.org \
--to=stern@rowland.harvard.edu \
--cc=akpm@linux-foundation.org \
--cc=alan@lxorguk.ukuu.org.uk \
--cc=davem@davemloft.net \
--cc=drzeus-mmc@drzeus.cx \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-pm@lists.linux-foundation.org \
--cc=paulmck@linux.vnet.ibm.com \
--cc=pavel@ucw.cz \
--cc=rjw@sisk.pl \
--cc=torvalds@linux-foundation.org \
--cc=zdenek.kabelac@gmail.com \
/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).