Date:   Wed, 30 Sep 2020 17:06:49 +0000
From:   Athani Nadeem Ladkhan <nadeem@...ence.com>
To:     Bjorn Helgaas <helgaas@...nel.org>
CC:     Tom Joseph <tjoseph@...ence.com>,
        "lorenzo.pieralisi@....com" <lorenzo.pieralisi@....com>,
        "robh@...nel.org" <robh@...nel.org>,
        "bhelgaas@...gle.com" <bhelgaas@...gle.com>,
        "linux-pci@...r.kernel.org" <linux-pci@...r.kernel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "kishon@...com" <kishon@...com>, Milind Parab <mparab@...ence.com>,
        Swapnil Kashinath Jakhade <sjakhade@...ence.com>
Subject: RE: [PATCH v2] PCI: Cadence: Add quirk for Gen2 controller to do
 autonomous speed change.

Hi Bjorn,

> -----Original Message-----
> From: Bjorn Helgaas <helgaas@...nel.org>
> Sent: Thursday, September 24, 2020 2:08 AM
> To: Athani Nadeem Ladkhan <nadeem@...ence.com>
> Cc: Tom Joseph <tjoseph@...ence.com>; lorenzo.pieralisi@....com;
> robh@...nel.org; bhelgaas@...gle.com; linux-pci@...r.kernel.org; linux-
> kernel@...r.kernel.org; kishon@...com; Milind Parab
> <mparab@...ence.com>; Swapnil Kashinath Jakhade
> <sjakhade@...ence.com>
> Subject: Re: [PATCH v2] PCI: Cadence: Add quirk for Gen2 controller to do
> autonomous speed change.
> 
> EXTERNAL MAIL
> 
> 
> Something like:
> 
>   PCI: cadence: Retrain Link to work around Gen2 training defect
> 
> to match history (see "git log --oneline drivers/pci/controller/cadence/pcie-
> cadence-host.c").
This will be corrected in patch version 3.
> 
> On Wed, Sep 23, 2020 at 08:34:27PM +0200, Nadeem Athani wrote:
> > Cadence controller will not initiate autonomous speed change if
> > strapped as Gen2. The Retrain bit is set as a quirk to trigger this
> > speed change.
> 
> To match the spec terminology:
> 
>   Set the Retrain Link bit ...
This will be corrected in patch version 3.
> 
> Obviously I don't know the details of your device or even how PCIe works at
> this level.  But IIUC a link always comes up at 2.5 GT/s first and then the
> upstream and downstream components negotiate the highest speed they
> both support.  It sounds like your controller doesn't actually do this
> negotiation unless you set the Retrain Link bit?
Some Gen2 controller variants doesn't support this negotiation. Hence required to set the Retrain Link bit.
This will be handled in much better way in patch version 3.
> 
> Is cdns_pcie_host_init_root_port() the only time this needs to be done?  We
> don't have to worry about doing this again after a reset, hot-add event, etc?
Yes. Only during init.
> 
> > Signed-off-by: Nadeem Athani <nadeem@...ence.com>
> > ---
> >  drivers/pci/controller/cadence/pcie-cadence-host.c | 14 ++++++++++++++
> >  drivers/pci/controller/cadence/pcie-cadence.h      | 15 +++++++++++++++
> >  2 files changed, 29 insertions(+)
> >
> > diff --git a/drivers/pci/controller/cadence/pcie-cadence-host.c
> > b/drivers/pci/controller/cadence/pcie-cadence-host.c
> > index 4550e0d469ca..a2317614268d 100644
> > --- a/drivers/pci/controller/cadence/pcie-cadence-host.c
> > +++ b/drivers/pci/controller/cadence/pcie-cadence-host.c
> > @@ -83,6 +83,9 @@ static int cdns_pcie_host_init_root_port(struct
> cdns_pcie_rc *rc)
> >  	struct cdns_pcie *pcie = &rc->pcie;
> >  	u32 value, ctrl;
> >  	u32 id;
> > +	u32 link_cap = CDNS_PCIE_LINK_CAP_OFFSET;
> 
> This is not actually the link cap offset.  Based on the usage, this appears to be
> the offset of the PCIe Capability.
This will be corrected in patch version 3.
> 
> > +	u8 sls;
> > +	u16 lnk_ctl;
> >
> >  	/*
> >  	 * Set the root complex BAR configuration register:
> > @@ -111,6 +114,17 @@ static int cdns_pcie_host_init_root_port(struct
> cdns_pcie_rc *rc)
> >  	if (rc->device_id != 0xffff)
> >  		cdns_pcie_rp_writew(pcie, PCI_DEVICE_ID, rc->device_id);
> >
> > +	/* Quirk to enable autonomous speed change for GEN2 controller */
> > +	/* Reading Supported Link Speed value */
> > +	sls = PCI_EXP_LNKCAP_SLS &
> > +		cdns_pcie_rp_readb(pcie, link_cap + PCI_EXP_LNKCAP);
> 
> The conventional way to write this would be
> 
>   sls = cdns_pcie_rp_readb(pcie, link_cap + PCI_EXP_LNKCAP) &
>     PCI_EXP_LNKCAP_SLS;
This will be corrected in patch version 3.
> 
> > +	if (sls == PCI_EXP_LNKCAP_SLS_5_0GB) {
> > +		/* Since this a Gen2 controller, set Retrain Link(RL) bit */
> > +		lnk_ctl = cdns_pcie_rp_readw(pcie, link_cap +
> PCI_EXP_LNKCTL);
> > +		lnk_ctl |= PCI_EXP_LNKCTL_RL;
> > +		cdns_pcie_rp_writew(pcie, link_cap + PCI_EXP_LNKCTL,
> lnk_ctl);
> > +	}
> > +
> >  	cdns_pcie_rp_writeb(pcie, PCI_CLASS_REVISION, 0);
> >  	cdns_pcie_rp_writeb(pcie, PCI_CLASS_PROG, 0);
> >  	cdns_pcie_rp_writew(pcie, PCI_CLASS_DEVICE,
> PCI_CLASS_BRIDGE_PCI);
> > diff --git a/drivers/pci/controller/cadence/pcie-cadence.h
> > b/drivers/pci/controller/cadence/pcie-cadence.h
> > index feed1e3038f4..fe560480c573 100644
> > --- a/drivers/pci/controller/cadence/pcie-cadence.h
> > +++ b/drivers/pci/controller/cadence/pcie-cadence.h
> > @@ -120,6 +120,7 @@
> >   */
> >  #define CDNS_PCIE_RP_BASE	0x00200000
> >
> > +#define CDNS_PCIE_LINK_CAP_OFFSET 0xC0
> 
> Use lower-case in hex as the rest of the file does.
This will be corrected in patch version 3.
> 
> >  /*
> >   * Address Translation Registers
> > @@ -413,6 +414,20 @@ static inline void cdns_pcie_rp_writew(struct
> cdns_pcie *pcie,
> >  	cdns_pcie_write_sz(addr, 0x2, value);  }
> >
> > +static inline u8 cdns_pcie_rp_readb(struct cdns_pcie *pcie, u32 reg)
> > +{
> > +	void __iomem *addr = pcie->reg_base + CDNS_PCIE_RP_BASE + reg;
> > +
> > +	return cdns_pcie_read_sz(addr, 0x1); }
> > +
> > +static inline u16 cdns_pcie_rp_readw(struct cdns_pcie *pcie, u32 reg)
> > +{
> > +	void __iomem *addr = pcie->reg_base + CDNS_PCIE_RP_BASE + reg;
> > +
> > +	return cdns_pcie_read_sz(addr, 0x2); }
> > +
> >  /* Endpoint Function register access */  static inline void
> > cdns_pcie_ep_fn_writeb(struct cdns_pcie *pcie, u8 fn,
> >  					  u32 reg, u8 value)
> > --
> > 2.15.0
> >

Powered by blists - more mailing lists