diff mbox

[RFC,1/5] new helper: iov_iter_rw()

Message ID 34dc78b262546e9343e0ed872232a97f5eaa5f15.1426502566.git.osandov@osandov.com (mailing list archive)
State New, archived
Headers show

Commit Message

Omar Sandoval March 16, 2015, 11:33 a.m. UTC
Get either READ or WRITE out of iter->type.

Signed-off-by: Omar Sandoval <osandov@osandov.com>
---
 include/linux/uio.h | 10 ++++++++++
 1 file changed, 10 insertions(+)

Comments

Al Viro March 16, 2015, 5:36 p.m. UTC | #1
On Mon, Mar 16, 2015 at 04:33:49AM -0700, Omar Sandoval wrote:
> Get either READ or WRITE out of iter->type.

Umm...  

> + * Get one of READ or WRITE out of iter->type without any other flags OR'd in
> + * with it.
> + */
> +static inline int iov_iter_rw(const struct iov_iter *i)
> +{
> +	return i->type & RW_MASK;
> +}

TBH, I would turn that into a macro.  Reason: indirect includes.

How about

#define iov_iter_rw(i) ((0 ? (struct iov_iter *)0 : (i))->type & RW_MASK)

Should do you all the type safety of inline function and avoids the need
to include fs.h in uio.h; _users_ of iov_iter_rw() obviously still need
fs.h, but such places always used to...
--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
David Sterba March 17, 2015, 9:31 a.m. UTC | #2
On Mon, Mar 16, 2015 at 05:36:05PM +0000, Al Viro wrote:
> On Mon, Mar 16, 2015 at 04:33:49AM -0700, Omar Sandoval wrote:
> > Get either READ or WRITE out of iter->type.
> 
> Umm...  
> 
> > + * Get one of READ or WRITE out of iter->type without any other flags OR'd in
> > + * with it.
> > + */
> > +static inline int iov_iter_rw(const struct iov_iter *i)
> > +{
> > +	return i->type & RW_MASK;
> > +}
> 
> TBH, I would turn that into a macro.  Reason: indirect includes.

Agreed, but the proposed define is rather cryptic and I was not able to
understand the meaning on the first glance.

> #define iov_iter_rw(i) ((0 ? (struct iov_iter *)0 : (i))->type & RW_MASK)

This worked for me, does not compile with anything else than
'struct iov_iter*' as i:

#define iov_iter_rw(i)	({			\
	struct iov_iter __iter = *(i);		\
	(i)->type & RW_MASK;			\
	})

The assignment is optimized out.
--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Omar Sandoval March 17, 2015, 10:18 a.m. UTC | #3
On Tue, Mar 17, 2015 at 10:31:51AM +0100, David Sterba wrote:
> On Mon, Mar 16, 2015 at 05:36:05PM +0000, Al Viro wrote:
> > On Mon, Mar 16, 2015 at 04:33:49AM -0700, Omar Sandoval wrote:
> > > Get either READ or WRITE out of iter->type.
> > 
> > Umm...  
> > 
> > > + * Get one of READ or WRITE out of iter->type without any other flags OR'd in
> > > + * with it.
> > > + */
> > > +static inline int iov_iter_rw(const struct iov_iter *i)
> > > +{
> > > +	return i->type & RW_MASK;
> > > +}
> > 
> > TBH, I would turn that into a macro.  Reason: indirect includes.
> 
> Agreed, but the proposed define is rather cryptic and I was not able to
> understand the meaning on the first glance.
> 
> > #define iov_iter_rw(i) ((0 ? (struct iov_iter *)0 : (i))->type & RW_MASK)
> 
> This worked for me, does not compile with anything else than
> 'struct iov_iter*' as i:
> 
> #define iov_iter_rw(i)	({			\
> 	struct iov_iter __iter = *(i);		\
> 	(i)->type & RW_MASK;			\
> 	})
> 
> The assignment is optimized out.

[-cc individual fs maintainers to avoid all of these email bounces,
should've looked a bit closer at that get_maintainer.pl output...]

I agree that this is a bit more readable, but it evaluates i twice.
That's an easy fix, just do __iter.type instead of (i)->type, but
there's still the possibility of someone passing in something called
__iter as i, and the fix for that tends to be "add more underscores". At
the very least, Al's macro could probably use a comment explaining
what's going on there, though.
Al Viro March 17, 2015, 6:19 p.m. UTC | #4
On Tue, Mar 17, 2015 at 10:31:51AM +0100, David Sterba wrote:

> Agreed, but the proposed define is rather cryptic and I was not able to
> understand the meaning on the first glance.
> 
> > #define iov_iter_rw(i) ((0 ? (struct iov_iter *)0 : (i))->type & RW_MASK)
> 
> This worked for me, does not compile with anything else than
> 'struct iov_iter*' as i:
> 
> #define iov_iter_rw(i)	({			\
> 	struct iov_iter __iter = *(i);		\
> 	(i)->type & RW_MASK;			\
> 	})
> 
> The assignment is optimized out.

... and you are getting
	a) use of rather lousy gccism when plain C would do
	b) double evaluation since you've got it wrong (should've been
__iter.type & RW_MASK, if you do it that way).  As it is, if argument has
any side effects, your variant will trigger those twice - even if the
destination of the assignment is never used, the side effects remain.

I agree that it could use /* use ?: for typechecking */, but let's not go into
({...}) land unless we absolutely have to.
--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
diff mbox

Patch

diff --git a/include/linux/uio.h b/include/linux/uio.h
index 7188029..87a47b3 100644
--- a/include/linux/uio.h
+++ b/include/linux/uio.h
@@ -10,6 +10,7 @@ 
 #define __LINUX_UIO_H
 
 #include <linux/kernel.h>
+#include <linux/fs.h>
 #include <uapi/linux/uio.h>
 
 struct page;
@@ -111,6 +112,15 @@  static inline bool iter_is_iovec(struct iov_iter *i)
 }
 
 /*
+ * Get one of READ or WRITE out of iter->type without any other flags OR'd in
+ * with it.
+ */
+static inline int iov_iter_rw(const struct iov_iter *i)
+{
+	return i->type & RW_MASK;
+}
+
+/*
  * Cap the iov_iter by given limit; note that the second argument is
  * *not* the new size - it's upper limit for such.  Passing it a value
  * greater than the amount of data in iov_iter is fine - it'll just do