On Fri, Feb 22, 2019 at 4:55 AM Jason Gunthorpe <jgg@xxxxxxxx> wrote:
>
> On Wed, Feb 20, 2019 at 04:21:02PM -0800, Matthew Wilcox wrote:
> > This driver doesn't look up the pointer that it stores, so this can
> > just be an IDA and save some memory.
> >
> > Signed-off-by: Matthew Wilcox <willy@xxxxxxxxxxxxx>
> > ---
> > drivers/infiniband/hw/ocrdma/ocrdma_main.c | 16 +++++++---------
> > 1 file changed, 7 insertions(+), 9 deletions(-)
> >
> > diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_main.c b/drivers/infiniband/hw/ocrdma/ocrdma_main.c
> > index 1f393842453a..b014eeffbcce 100644
> > --- a/drivers/infiniband/hw/ocrdma/ocrdma_main.c
> > +++ b/drivers/infiniband/hw/ocrdma/ocrdma_main.c
> > @@ -62,7 +62,7 @@ MODULE_DESCRIPTION(OCRDMA_ROCE_DRV_DESC " " OCRDMA_ROCE_DRV_VERSION);
> > MODULE_AUTHOR("Emulex Corporation");
> > MODULE_LICENSE("Dual BSD/GPL");
> >
> > -static DEFINE_IDR(ocrdma_dev_id);
> > +static DEFINE_IDA(ocrdma_dev_id);
> >
> > void ocrdma_get_guid(struct ocrdma_dev *dev, u8 *guid)
> > {
> > @@ -302,12 +302,12 @@ static struct ocrdma_dev *ocrdma_add(struct be_dev_info *dev_info)
> > }
> > dev->mbx_cmd = kzalloc(sizeof(struct ocrdma_mqe_emb_cmd), GFP_KERNEL);
> > if (!dev->mbx_cmd)
> > - goto idr_err;
> > + goto ida_err;
> >
> > memcpy(&dev->nic_info, dev_info, sizeof(*dev_info));
> > - dev->id = idr_alloc(&ocrdma_dev_id, NULL, 0, 0, GFP_KERNEL);
> > + dev->id = ida_alloc(&ocrdma_dev_id, GFP_KERNEL);
>
> WTF.. There are only three places where this id is used:
> - A whole bunch of pr_* calls which *really* should be dev_* calls
> with no ID (grrrr)
> - It is passed to userspace in the ucontext udata, and userpace
> ignores it (double grrrrr)
> - It is used to form a name for request_irq (use the pci BDF)
>
> We should delete this thing :(
>
Will send a patch to remove this.
> Jason
