RE: [EXT] Re: [PATCH net-next v3 2/7] octeon_ep: poll for control messages

[Veerasenareddy Burru <vburru@marvell.com>]

> -----Original Message-----
> From: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
> Sent: Tuesday, February 14, 2023 9:43 AM
> To: Veerasenareddy Burru <vburru@marvell.com>
> Cc: netdev@vger.kernel.org; linux-kernel@vger.kernel.org; Abhijit Ayarekar
> <aayarekar@marvell.com>; Sathesh B Edara <sedara@marvell.com>;
> Satananda Burla <sburla@marvell.com>; linux-doc@vger.kernel.org; David S.
> Miller <davem@davemloft.net>; Eric Dumazet <edumazet@google.com>;
> Jakub Kicinski <kuba@kernel.org>; Paolo Abeni <pabeni@redhat.com>
> Subject: [EXT] Re: [PATCH net-next v3 2/7] octeon_ep: poll for control
> messages
> 
> External Email
> 
> ----------------------------------------------------------------------
> On Mon, Feb 13, 2023 at 09:14:17PM -0800, Veerasenareddy Burru wrote:
> > Poll for control messages until interrupts are enabled.
> > All the interrupts are enabled in ndo_open().
> > Add ability to listen for notifications from firmware before ndo_open().
> > Once interrupts are enabled, this polling is disabled and all the
> > messages are processed by bottom half of interrupt handler.
> >
> > Signed-off-by: Veerasenareddy Burru <vburru@marvell.com>
> > Signed-off-by: Abhijit Ayarekar <aayarekar@marvell.com>
> 
> small two nits
> 
> > ---
> > v2-> v3:
> >  * resovled review comment; fixed reverse christmas tree.
> >
> > v1 -> v2:
> >  * removed device status oct->status, as it is not required with the
> >    modified implementation in 0001-xxxx.patch
> >
> >  .../marvell/octeon_ep/octep_cn9k_pf.c         | 49 +++++++++----------
> >  .../ethernet/marvell/octeon_ep/octep_main.c   | 35 +++++++++++++
> >  .../ethernet/marvell/octeon_ep/octep_main.h   | 11 ++++-
> >  .../marvell/octeon_ep/octep_regs_cn9k_pf.h    |  4 ++
> >  4 files changed, 71 insertions(+), 28 deletions(-)
> >
> > diff --git a/drivers/net/ethernet/marvell/octeon_ep/octep_cn9k_pf.c
> > b/drivers/net/ethernet/marvell/octeon_ep/octep_cn9k_pf.c
> > index 6ad88d0fe43f..f40ebac15a79 100644
> > --- a/drivers/net/ethernet/marvell/octeon_ep/octep_cn9k_pf.c
> > +++ b/drivers/net/ethernet/marvell/octeon_ep/octep_cn9k_pf.c
> > @@ -352,27 +352,36 @@ static void
> octep_setup_mbox_regs_cn93_pf(struct octep_device *oct, int q_no)
> >  	mbox->mbox_read_reg = oct->mmio[0].hw_addr +
> > CN93_SDP_R_MBOX_VF_PF_DATA(q_no);  }
> >
> > -/* Mailbox Interrupt handler */
> > -static void cn93_handle_pf_mbox_intr(struct octep_device *oct)
> > +/* Process non-ioq interrupts required to keep pf interface running.
> > + * OEI_RINT is needed for control mailbox  */ static int
> > +octep_poll_non_ioq_interrupts_cn93_pf(struct octep_device *oct)
> 
> return bool?
> 

Yes, bool is sufficient. Will make the change in next revision.

> >  {
> > -	u64 mbox_int_val = 0ULL, val = 0ULL, qno = 0ULL;
> > +	int handled = 0;
> > +	u64 reg0;
> >
> > -	mbox_int_val = readq(oct->mbox[0]->mbox_int_reg);
> > -	for (qno = 0; qno < OCTEP_MAX_VF; qno++) {
> > -		val = readq(oct->mbox[qno]->mbox_read_reg);
> > -		dev_dbg(&oct->pdev->dev,
> > -			"PF MBOX READ: val:%llx from VF:%llx\n", val, qno);
> > +	/* Check for OEI INTR */
> > +	reg0 = octep_read_csr64(oct, CN93_SDP_EPF_OEI_RINT);
> > +	if (reg0) {
> > +		dev_info(&oct->pdev->dev,
> > +			 "Received OEI_RINT intr: 0x%llx\n",
> > +			 reg0);
> > +		octep_write_csr64(oct, CN93_SDP_EPF_OEI_RINT, reg0);
> > +		if (reg0 & CN93_SDP_EPF_OEI_RINT_DATA_BIT_MBOX)
> > +			queue_work(octep_wq, &oct->ctrl_mbox_task);
> > +
> > +		handled = 1;
> >  	}
> >
> > -	writeq(mbox_int_val, oct->mbox[0]->mbox_int_reg);
> > +	return handled;
> >  }
> >
> >  /* Interrupts handler for all non-queue generic interrupts. */
> > static irqreturn_t octep_non_ioq_intr_handler_cn93_pf(void *dev)  {
> >  	struct octep_device *oct = (struct octep_device *)dev;
> > -	struct pci_dev *pdev = oct->pdev;
> >  	u64 reg_val = 0;
> > +	struct pci_dev *pdev = oct->pdev;
> 
> why this move of var and rct breakage?
> 

Thank you for the feedback. This change was not necessary. Will revert this change.
I will recheck whole patchset for RCT breakage and fix it in next revision.

> >  	int i = 0;
> >
> >  	/* Check for IRERR INTR */
> > @@ -434,24 +443,9 @@ static irqreturn_t
> octep_non_ioq_intr_handler_cn93_pf(void *dev)
> >  		goto irq_handled;
> >  	}
> >
> > -	/* Check for MBOX INTR */
> > -	reg_val = octep_read_csr64(oct, CN93_SDP_EPF_MBOX_RINT(0));
> > -	if (reg_val) {
> > -		dev_info(&pdev->dev,
> > -			 "Received MBOX_RINT intr: 0x%llx\n", reg_val);
> > -		cn93_handle_pf_mbox_intr(oct);
> > +	/* Check for MBOX INTR and OEI INTR */
> > +	if (octep_poll_non_ioq_interrupts_cn93_pf(oct))
> >  		goto irq_handled;
> > -	}
> > -
> > -	/* Check for OEI INTR */
> > -	reg_val = octep_read_csr64(oct, CN93_SDP_EPF_OEI_RINT);
> > -	if (reg_val) {
> > -		dev_info(&pdev->dev,
> > -			 "Received OEI_EINT intr: 0x%llx\n", reg_val);
> > -		octep_write_csr64(oct, CN93_SDP_EPF_OEI_RINT, reg_val);
> > -		queue_work(octep_wq, &oct->ctrl_mbox_task);
> > -		goto irq_handled;
> > -	}
> >
> >  	/* Check for DMA INTR */
> >  	reg_val = octep_read_csr64(oct, CN93_SDP_EPF_DMA_RINT);
> > @@ -712,6 +706,7 @@ void octep_device_setup_cn93_pf(struct
> octep_device *oct)
> >
> >  	oct->hw_ops.enable_interrupts =
> octep_enable_interrupts_cn93_pf;
> >  	oct->hw_ops.disable_interrupts =
> octep_disable_interrupts_cn93_pf;
> > +	oct->hw_ops.poll_non_ioq_interrupts =
> octep_poll_non_ioq_interrupts_cn93_pf;
> >
> >  	oct->hw_ops.update_iq_read_idx =
> octep_update_iq_read_index_cn93_pf;