Re: [Intel-wired-lan] [net-next] irdma: change name of interrupts

[Michal Swiatkowski <michal.swiatkowski@linux.intel.com>]

On Mon, Nov 14, 2022 at 04:19:20PM +0000, Saleem, Shiraz wrote:
> > Subject: [net-next] irdma: change name of interrupts
> > 
> 
> This cant go through net-next. It needs to go through the rdma tree. I will co-ordinate w. Tony on it.
> 
> Use "RDMA/irdma: change name of interrupts" for the subject.
>

I will remember, is "rdma" the name of tree?

> > Add more information in interrupt names.
> > 
> > Before this patch it was:
> > irdma
> > CEQ
> > CEQ
> > ...
> > 
> > Now:
> > irdma-ice-AEQ
> > irdma-ice-CEQ-0
> > irdma-ice-CEQ-1
> 
> This is better for sure. But we need to add the device name. Not sure if adding "ice" is adding any value.
> 
> Ideally, for the RDMA drivers, this should be ib_device name. But there is a limitation in RDMA subsystem that doesn't
> set the name till after the ib device is registered which happens later in the drv.probe() flow.
> 
> So I think our best option is to use the PCI device name.
> 

Sure, will change it, thanks.

> > ...
> > 
> > Signed-off-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
> > Suggested-by: Piotr Raczynski <piotr.raczynski@intel.com>
> > ---
> >  drivers/infiniband/hw/irdma/defs.h |  2 ++
> >  drivers/infiniband/hw/irdma/hw.c   | 14 +++++++++++---
> >  drivers/infiniband/hw/irdma/main.h |  1 +
> >  3 files changed, 14 insertions(+), 3 deletions(-)
> > 
> > diff --git a/drivers/infiniband/hw/irdma/defs.h b/drivers/infiniband/hw/irdma/defs.h
> > index c1906cab5c8a..09e4aba31b64 100644
> > --- a/drivers/infiniband/hw/irdma/defs.h
> > +++ b/drivers/infiniband/hw/irdma/defs.h
> > @@ -20,6 +20,8 @@
> >  #define IRDMA_IRD_HW_SIZE_128	3
> >  #define IRDMA_IRD_HW_SIZE_256	4
> > 
> > +#define IRDMA_INT_NAME_STR_LEN (32)
> 
> INT to IRQ which is more used in this driver to reflect interrupt. Drop the (). And move to this main.h where the irdma_msix_vector struct is defined.
>

Done

> > +
> >  enum irdma_protocol_used {
> >  	IRDMA_ANY_PROTOCOL = 0,
> >  	IRDMA_IWARP_PROTOCOL_ONLY = 1,
> > diff --git a/drivers/infiniband/hw/irdma/hw.c b/drivers/infiniband/hw/irdma/hw.c
> > index ab246447520b..46665624e515 100644
> > --- a/drivers/infiniband/hw/irdma/hw.c
> > +++ b/drivers/infiniband/hw/irdma/hw.c
> > @@ -1087,14 +1087,20 @@ static int irdma_cfg_ceq_vector(struct irdma_pci_f *rf,
> > struct irdma_ceq *iwceq,
> >  	int status;
> > 
> >  	if (rf->msix_shared && !ceq_id) {
> > +		snprintf(msix_vec->name, sizeof(msix_vec->name) - 1,
> > +			 "irdma-%s-AEQCEQ-%d",
> > +			 dev_driver_string(&rf->pcidev->dev), ceq_id);
> 
> As per previous comment, I think dev_name((&rf->pcidev->dev) is better. Given this should we make the msix_vec.name be a 64b variable?
>

Yeah, it looks better, thanks. I changed name to 64b.

> Shiraz
> 
> 
_______________________________________________
Intel-wired-lan mailing list
Intel-wired-lan@osuosl.org
https://lists.osuosl.org/mailman/listinfo/intel-wired-lan