Re: [rdma-rdma:wip/jgg-for-next 198/203] drivers/infiniband/ulp/rtrs/rtrs-srv.c:1805 rtrs_rdma_connect() warn: passing zero to 'PTR_ERR'

[Jason Gunthorpe <jgg@nvidia.com>]

[-- Attachment #1: Type: text/plain, Size: 1175 bytes --]

On Mon, Feb 15, 2021 at 04:57:57PM +0300, Dan Carpenter wrote:
> > > d715ff8acbd587 Guoqing Jiang  2020-10-23  1800          /*
> > > d715ff8acbd587 Guoqing Jiang  2020-10-23  1801           * "refcount == 0" happens if a previous thread calls get_or_create_srv
> > > d715ff8acbd587 Guoqing Jiang  2020-10-23  1802           * allocate srv, but chunks of srv are not allocated yet.
> > > d715ff8acbd587 Guoqing Jiang  2020-10-23  1803           */
> > > f0751419d3a15f Md Haris Iqbal 2021-02-12  1804          if (IS_ERR(srv) || refcount_read(&srv->refcount) == 0) {
> > > f0751419d3a15f Md Haris Iqbal 2021-02-12 @1805                  err = PTR_ERR(srv);
> > >
> > > It's not 100% clear to me that an error code is required when
> > > refcount_read() is zero.  Maybe "err = 0;" is okay in that situation?
> > I don't see how srv can be zero
> 
> The cross function database was probably out of date.

Regardless of if it is up to date or not, this sequence:

    if (refcount_read(&srv->refcount) == 0)
         err = PTR_ERR(srv);

Is completely bogus

eg prove srv is a valid pointer by de-refing and then passing a valid
pointer to PTR_ERR

Jason