Add htmldoc annotation for the newly introduced "head_tail" member
describing it to be a union of the pipe_inode_info's @head and @tail
members.

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Closes: https://lore.kernel.org/lkml/20250305204609.5e64768e@canb.auug.org.au/
Fixes: 3d252160b818 ("fs/pipe: Read pipe->{head,tail} atomically outside pipe->mutex")
Signed-off-by: K Prateek Nayak <kprateek.nayak@amd.com>
---
 include/linux/pipe_fs_i.h | 1 +
 1 file changed, 1 insertion(+)

diff --git a/include/linux/pipe_fs_i.h b/include/linux/pipe_fs_i.h
index 3cc4f8eab853..44b8c346ae2d 100644
--- a/include/linux/pipe_fs_i.h
+++ b/include/linux/pipe_fs_i.h
@@ -65,6 +65,7 @@ union pipe_index {
  *	@wr_wait: writer wait point in case of full pipe
  *	@head: The point of buffer production
  *	@tail: The point of buffer consumption
+ *	@head_tail: unsigned long union of @head and @tail
  *	@note_loss: The next read() should insert a data-lost message
  *	@max_usage: The maximum number of slots that may be used in the ring
  *	@ring_size: total number of buffers (should be a power of 2)

base-commit: 48a5eed9ad584315c30ed35204510536235ce402
-- 
2.34.1

On Wed, 5 Mar 2025 at 01:24, K Prateek Nayak <kprateek.nayak@amd.com> wrote:
>
> Add htmldoc annotation for the newly introduced "head_tail" member
> describing it to be a union of the pipe_inode_info's @head and @tail
> members.

Applied.

I also committed the pipe_occupancy() fix for 32-bit, but did so
without testing - it was ObviouslyCorrect(tm), but considering that
everybody missed it in the original patch, "obvious" is all relative.

And it turns out my worry about pipe_discard_from() was unnecessary.
Yes, the code is buggy. But no, it doesn't matter. Why? There are no
callers of that function ;)

That said, I still hope people will take a look at the code and see if
there's anything else I missed.

Oh, and testing. Testing would be good.

        Linus

Hello Linus,

On 3/5/2025 11:05 PM, Linus Torvalds wrote:
> On Wed, 5 Mar 2025 at 01:24, K Prateek Nayak <kprateek.nayak@amd.com> wrote:
>>
>> Add htmldoc annotation for the newly introduced "head_tail" member
>> describing it to be a union of the pipe_inode_info's @head and @tail
>> members.
> 
> Applied.

Thank you.

> 
> I also committed the pipe_occupancy() fix for 32-bit, but did so
> without testing - it was ObviouslyCorrect(tm), but considering that
> everybody missed it in the original patch, "obvious" is all relative.
> 
> And it turns out my worry about pipe_discard_from() was unnecessary.
> Yes, the code is buggy. But no, it doesn't matter. Why? There are no
> callers of that function ;)

On the topic of pipe_discard_from(), Al had added that helper in commit
c3497fd009ef ("fix short copy handling in copy_mc_pipe_to_iter()") to
discard unused buffer for copy_mc_to_iter()

Its usage in copy_mc_pipe_to_iter() was later removed in commit
8fad7767edcf ("ITER_PIPE: allocate buffers as we go in copy-to-pipe
primitives) in favor of iov_iter_revert()

pipe_advance() started using it in from commit 2c855de93314 ("ITER_PIPE:
clean pipe_advance() up") however all of that was nuked in commit
3fc40265ae2b ("iov_iter: Kill ITER_PIPE")

generic_file_splice_read() started using it from commit 0d9649341363
("splice: stop abusing iov_iter_advance() to flush a pipe") but
generic_file_splice_read() was later removed in commit c6585011bc1d
("splice: Remove generic_file_splice_read()")

I don't see any in kernel user of this helper currently so can the
wrap-around issue be addressed and the helper be removed later?

> 
> That said, I still hope people will take a look at the code and see if
> there's anything else I missed.
> 
> Oh, and testing. Testing would be good.
> 
>          Linus

-- 
Thanks and Regards,
Prateek

On Wed, 5 Mar 2025 at 08:08, K Prateek Nayak <kprateek.nayak@amd.com> wrote:
>
> > And it turns out my worry about pipe_discard_from() was unnecessary.
> > Yes, the code is buggy. But no, it doesn't matter. Why? There are no
> > callers of that function ;)
>
]> I don't see any in kernel user of this helper currently so can the
> wrap-around issue be addressed and the helper be removed later?

Yes. That's what I did. I still would love to see

 - more testing

 - explicit limit checking in pipe_resize_ring()

and then I think we could make CONFIG_64BIT use the 16-bit
pipe_index_t too (because not doing so clearly was a mistake: it meant
that people didn't test the 32-bit case, so the wrap-around bugs
weren't as obvious).

                 Linus

                    Linus