Skip to contentSkip to conversation

iov_iter: Declare new iterator direction symbols

PATCH In review David Howells 2 participants 2 messages Nov 4, 2022

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 iov_iter: Declare new iterator direction symbols

Hi Linus, Al,

If we're going to go with Al's changes to switch to using ITER_SOURCE and ITER_DEST instead of READ/WRITE, can we put just the new symbols into mainline now, even if we leave the rest for the next merge window?

Thanks,
David
---
From: Al Viro <[email protected]>

iov_iter: Declare new iterator direction symbols

READ/WRITE proved to be actively confusing - the meanings are "data destination, as used with read(2)" and "data source, as used with write(2)", but people keep interpreting those as "we read data from it" and "we write data to it", i.e. exactly the wrong way.

Call them ITER_DEST and ITER_SOURCE - at least that is harder to misinterpret...

[dhowells] Declare the symbols for later use and change to an enum. If READ/WRITE are switched to an enum also, I think the compiler should generate a warning if they're mixed.

---
 include/linux/uio.h |    5 +++++
 1 file changed, 5 insertions(+)
include/linux/uio.h +5 -0
index 2e3134b14ffd..7c1317b34c57 100644
@@ -29,6 +29,11 @@ enum iter_type {
 	ITER_UBUF,
 };
 
+enum iov_iter_direction {
+	ITER_DEST	= 0,	/* Iterator is a destination buffer (== READ) */
+	ITER_SOURCE	= 1,	/* Iterator is a source buffer (== WRITE) */
+};
+
 struct iov_iter_state {
 	size_t iov_offset;
 	size_t count;
#1 lore raw
Quoted 4 lines · David Howells

If we're going to go with Al's changes to switch to using ITER_SOURCE and ITER_DEST instead of READ/WRITE, can we put just the new symbols into mainline now, even if we leave the rest for the next merge window?

No, I really don't want to have mixed-used stuff in the kernel.

Continue to use the old names until/if the conversion happens, at which point it's the conversion code that does it.

No "one branch uses new names, another uses old names" mess.

Linus