From: Rusty Russell <rusty@rustcorp.com.au>
To: Tejun Heo <htejun@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>,
Andrew Morton <akpm@linux-foundation.org>,
linux-kernel@vger.kernel.org, Jeff Garzik <jeff@garzik.org>
Subject: Re: [PATCH 0/6] RFC: Typesafe callbacks
Date: Tue, 22 Jan 2008 10:27:14 +1100 [thread overview]
Message-ID: <200801221027.15002.rusty@rustcorp.com.au> (raw)
In-Reply-To: <47949257.1000102@gmail.com>
On Monday 21 January 2008 23:38:47 Tejun Heo wrote:
> Rusty Russell wrote:
> > On Monday 21 January 2008 09:17:30 Rusty Russell wrote:
> >> But it would be cool to allow functions which take an unsigned long.
> >> I'll test this out and see what I can make...
> >
> > I think this comes under "too ugly", but here's an attempt.
>
> Maybe we should be content with pointer type checking. It seems like
> it's going too far and after all the clutter it's not possible to use
> int as argument. :-(
It is possible, but the function has to take an unsigned long.
But I share your dislike of this. Here's my final version, but it's still
ugly.
Rusty.
===
Attempt to create callbacks which take unsigned long as well as
correct pointer types.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
---
include/linux/compiler-gcc.h | 61 +++++++++++++++++++++++++++++++++++++++++
include/linux/compiler-intel.h | 3 ++
include/linux/kernel.h | 21 ++++++++++++++
3 files changed, 85 insertions(+)
diff -r 951ffe1c5238 include/linux/compiler-gcc.h
--- a/include/linux/compiler-gcc.h Tue Jan 22 09:31:56 2008 +1100
+++ b/include/linux/compiler-gcc.h Tue Jan 22 10:15:15 2008 +1100
@@ -53,3 +53,64 @@
#define noinline __attribute__((noinline))
#define __attribute_const__ __attribute__((__const__))
#define __maybe_unused __attribute__((unused))
+
+/* This is not complete: I can't figure out a way to handle enums. */
+#define ulong_compatible(arg) \
+ (__builtin_types_compatible_p(typeof(arg), char) \
+ || __builtin_types_compatible_p(typeof(arg), unsigned char) \
+ || __builtin_types_compatible_p(typeof(arg), signed char) \
+ || __builtin_types_compatible_p(typeof(arg), unsigned short) \
+ || __builtin_types_compatible_p(typeof(arg), short) \
+ || __builtin_types_compatible_p(typeof(arg), unsigned int) \
+ || __builtin_types_compatible_p(typeof(arg), int) \
+ || __builtin_types_compatible_p(typeof(arg), unsigned long) \
+ || __builtin_types_compatible_p(typeof(arg), long))
+
+/**
+ * typesafe_fn_and_arg - cast function type and arg if compatible
+ * @fn: the function or function pointer
+ * @arg: the function argument.
+ * @ulongtype: the function pointer type if arg is an integer
+ * @safetype: the function pointer type which matches typeof(arg)
+ * @voidtype: the function pointer type which takes a void * (to cast to)
+ *
+ * This macro evaluates to two comma separated values: the function pointer,
+ * then the argument.
+ *
+ * Callback functions are usually declared to take a "void *arg"
+ * argument, which stops the compiler from doing type checking. We
+ * can freely cast to function pointer which takes "void *" in two
+ * cases: a function which takes a different pointer type or an
+ * unsigned long.
+ *
+ * Typechecking is done by the caller when they assign or pass these
+ * values. This macro handles the cases where @fn takes an unsigned
+ * long and @arg is compatible with that, and also the case where @fn
+ * and @arg match types. For other cases, @fn is not cast, so using
+ * the results of this macro should cause a warning unless @fn is
+ * already of if type @voidtype.
+ *
+ * Logic looks like this:
+ * if (typeof(@arg) compatible with unsigned long) {
+ * if (typeof(@fn) == @ulongtype)
+ * (@voidtype)@fn and (void *)(unsigned long)@arg; // OK!
+ * else
+ * @fn and @arg; // @fn will warn unless it's of @voidtype already.
+ * } else if (typeof(@fn) == @safetype)
+ * (@voidtype)@fn and @arg; // OK: @arg will warn unless it's a pointer.
+ * else
+ * @fn and @arg; // @fn will warn unless it's of @voidtype already.
+ */
+#define typesafe_fn_and_arg(fn, arg, ulongtype, safetype, voidtype) \
+__builtin_choose_expr((ulong_compatible(arg) \
+ && __builtin_types_compatible_p(typeof(1?(fn):NULL), \
+ ulongtype)) \
+ || (!ulong_compatible(arg) \
+ && __builtin_types_compatible_p(typeof(1?(fn):NULL), \
+ safetype)), \
+ ((voidtype)(fn)), \
+ (fn)), \
+__builtin_choose_expr(__builtin_types_compatible_p(typeof(1?(fn):NULL), \
+ ulongtype), \
+ ((void *)(long)(arg)), (arg))
+
diff -r 951ffe1c5238 include/linux/compiler-intel.h
--- a/include/linux/compiler-intel.h Tue Jan 22 09:31:56 2008 +1100
+++ b/include/linux/compiler-intel.h Tue Jan 22 10:15:15 2008 +1100
@@ -29,3 +29,6 @@
#endif
#define uninitialized_var(x) x
+
+#define typesafe_fn_and_arg(fn, arg, ulongtype, safetype, fntype) \
+ ((fntype)(fn)), ((void *)arg)
diff -r 951ffe1c5238 include/linux/kernel.h
--- a/include/linux/kernel.h Tue Jan 22 09:31:56 2008 +1100
+++ b/include/linux/kernel.h Tue Jan 22 10:15:15 2008 +1100
@@ -379,6 +379,27 @@ static inline int __attribute__ ((format
(void)__tmp; \
})
+/**
+ * callback_and_arg - simple one-argument typesafe callback with argument
+ * @fn: the function or function pointer
+ * @arg: the function argument.
+ * @rettype: the return type of fn
+ *
+ * This macro evaluates to two comma separated values: the function pointer,
+ * then the argument. It will check that @fn and @arg are compatible with
+ * each other and with a void * callback. One of
+ * - fn takes a void * and arg is a pointer
+ * - fn takes a typeof(arg) and arg is a pointer
+ * - fn takes an unsigned long and arg is an integer type.
+ *
+ * See typesafe_fn_and_arg() for the gory details.
+ */
+#define callback_and_arg(fn, arg, rettype) \
+ typesafe_fn_and_arg((fn), (arg), \
+ rettype (*)(unsigned long), \
+ rettype (*)(typeof(arg)), \
+ rettype (*)(void *))
+
struct sysinfo;
extern int do_sysinfo(struct sysinfo *info);
next prev parent reply other threads:[~2008-01-21 23:28 UTC|newest]
Thread overview: 23+ messages / expand[flat|nested] mbox.gz Atom feed top
2008-01-20 9:46 [PATCH 0/6] RFC: Typesafe callbacks Rusty Russell
2008-01-20 9:47 ` [PATCH 1/6] typesafe: Convert stop_machine and callers Rusty Russell
2008-01-20 9:48 ` [PATCH 2/6] typesafe: kthread_create and kthread_run Rusty Russell
2008-01-20 9:50 ` [PATCH 3/6] typesafe: convert kthread users Rusty Russell
2008-01-20 9:51 ` [PATCH 4/6] typesafe: cast_if_type to allow macros functions which take more than one type Rusty Russell
2008-01-20 9:54 ` [PATCH 5/6] typesafe: request_irq and devm_request_irq Rusty Russell
2008-01-20 9:57 ` [PATCH 6/6] typesafe: timers Rusty Russell
2008-01-20 11:25 ` [PATCH 2/6] typesafe: kthread_create and kthread_run Jan Engelhardt
2008-01-20 12:07 ` Bert Wesarg
2008-01-20 16:24 ` Johannes Weiner
2008-01-20 16:43 ` Bert Wesarg
2008-01-20 22:04 ` Rusty Russell
2008-01-21 7:56 ` Bert Wesarg
2008-01-20 12:56 ` [PATCH 0/6] RFC: Typesafe callbacks Tejun Heo
2008-01-20 13:00 ` Tejun Heo
2008-01-20 22:17 ` Rusty Russell
2008-01-21 11:33 ` Rusty Russell
2008-01-21 12:38 ` Tejun Heo
2008-01-21 23:27 ` Rusty Russell [this message]
2008-01-21 23:57 ` Linus Torvalds
2008-01-22 7:16 ` Rusty Russell
2008-01-22 15:53 ` Linus Torvalds
2008-01-22 4:20 ` Andi Kleen
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=200801221027.15002.rusty@rustcorp.com.au \
--to=rusty@rustcorp.com.au \
--cc=akpm@linux-foundation.org \
--cc=htejun@gmail.com \
--cc=jeff@garzik.org \
--cc=linux-kernel@vger.kernel.org \
--cc=torvalds@linux-foundation.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).