Re: [PATCH v2] Fixing directly deferencing a __rcu pointer warning

[Abhinav Singh]

On 10/26/23 20:47, Michael S. Tsirkin wrote:
> On Thu, Oct 26, 2023 at 04:06:24PM +0200, Mateusz Guzik wrote:
> > On 10/26/23, Michael S. Tsirkin <mst@xxxxxxxxxx> wrote:
> > > On Thu, Oct 26, 2023 at 05:46:21PM +0530, Abhinav Singh wrote:
> > > > This patch fixes the warning about directly dereferencing a pointer
> > > > tagged with __rcu annotation.
> > > >
> > > > Dereferencing the pointers tagged with __rcu directly should
> > > > always be avoided according to the docs. There is a rcu helper
> > > > functions rcu_dereference(...) to use when dereferencing a __rcu
> > > > pointer. This functions returns the non __rcu tagged pointer which
> > > > can be dereferenced just like a normal pointers.
> > > >
> > > > Signed-off-by: Abhinav Singh <singhabhinav9051571833@xxxxxxxxx>
> > >
> > > Well yes but these need to be called under rcu_read_lock.
> > > Who does it here?
> > > If no one then maybe you found an actual bug and we need to
> > > fix it not paper over it.
> >
> > There is no bug here.
> >
> > p is the newly created thread, ->real_cred was initialized just prior
> > to this code and there is nobody to whack the creds from under it.
> >
> > Second bit in the patch changes one real_parent deref, but leaves 2
> > others just above it. Once more no bug since the entire thing happens
> > under tasklist_lock, but the patch should either sort all these cases
> > or none.
Sparse reported 3 similar dereferencing warning this patch contains 2 
fixes for 2, but yeah I should fixed all 3 of them.
> > I think it would help if the submitter had shown warnings they see.
The warning message :- warning: dereference of noderef expression
> Yes, and this must be tested under lockdep, which I think would
> spit out warnings for this patch.
Not sure, but I tested this with sparse (make C=2) and after the above 
changes I dont get the warning.
> What should be used here I'm not sure. IIUC rcu_dereference_protected(p, 1)
> is discouraged now?
Not sure but I read that, rcu_dereference should be prefered when 
reading and rcu_dereference_protected should when writing.