Re: [PATCH] docs: kref: Clarify the use of two kref_put() in example code
From: Manivannan Sadhasivam
Date: Wed Feb 19 2020 - 06:11:05 EST
On Wed, Feb 19, 2020 at 03:58:18AM -0700, Jonathan Corbet wrote:
> On Thu, 13 Feb 2020 18:23:11 +0530
> Manivannan Sadhasivam <manivannan.sadhasivam@xxxxxxxxxx> wrote:
> > Eventhough the current documentation explains that the reference count
> > gets incremented by both kref_init() and kref_get(), it is often
> > misunderstood that only one instance of kref_put() is needed in the
> > example code. So let's clarify that a bit.
> > Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@xxxxxxxxxx>
> > ---
> > Documentation/kref.txt | 4 ++++
> > 1 file changed, 4 insertions(+)
> > diff --git a/Documentation/kref.txt b/Documentation/kref.txt
> > index 3af384156d7e..c61eea6f1bf2 100644
> > --- a/Documentation/kref.txt
> > +++ b/Documentation/kref.txt
> > @@ -128,6 +128,10 @@ since we already have a valid pointer that we own a refcount for. The
> > put needs no lock because nothing tries to get the data without
> > already holding a pointer.
> > +In the above example, kref_put() will be called 2 times in both success
> > +and error paths. This is necessary because the reference count got
> > +incremented 2 times by kref_init() and kref_get().
> Out of curiosity, where have you seen this misunderstanding happening?
> I'm not really opposed to this change, but I don't understand why it's
> really needed.
Jakub mistakenly spotted one refcounting issue in one of my patchset:
Then I tried to show him the kernel doc for kref and that's where I got this
example code slightly confusing. And while looking into the log, I noticed that
someone deleted the kref_put in error path mistakenly and that commit got
reverted after that. This issue was even discussed in stack overflow.
So I thought about making it more clear of why the kref_put is needed in error