From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-eopbgr690042.outbound.protection.outlook.com ([40.107.69.42]:33024 "EHLO NAM04-CO1-obe.outbound.protection.outlook.com" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S933647AbeGIP4O (ORCPT ); Mon, 9 Jul 2018 11:56:14 -0400 From: Alan Douglas Subject: RE: [PATCH 4/5] PCI: cadence: Add Power Management ops for host and EP Date: Mon, 9 Jul 2018 15:56:11 +0000 Message-ID: References: <1529915453-4633-1-git-send-email-adouglas@cadence.com> <1529915453-4633-5-git-send-email-adouglas@cadence.com> <20180709153351.GA16292@red-moon> In-Reply-To: <20180709153351.GA16292@red-moon> Content-Language: en-US Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable MIME-Version: 1.0 Sender: devicetree-owner@vger.kernel.org To: Lorenzo Pieralisi Cc: "bhelgaas@google.com" , "kishon@ti.com" , "linux-pci@vger.kernel.org" , "devicetree@vger.kernel.org" , "robh+dt@kernel.org" List-ID: On 09 July 2018 16:34, Lorenzo Pieralisi wrote: > On Mon, Jun 25, 2018 at 09:30:52AM +0100, Alan Douglas wrote: > > These PM ops will enable/disable the optional PHYs if present. The > > AXI link-down register in the host driver is now cleared in > > cdns_pci_map_bus since the link-down bit will be set if the PHY has > > been disabled. It is not cleared when enabling the PHY, since the > > link will not yet be up. >=20 > It is not entirely clear what you mean here, can you elaborate please ? Thanks for taking a look at this. There is a "Link down indication bit" register in the cadence PCIe IP AXI Configuration Registers, which will be set by the PCIe IP after each link- down occurrence. This bit must be cleared before continuing, or accesses will hang. When the PHY is disabled, the link will go down and this bit wi= ll be set by HW. It will also be set after a secondary bus reset. We cannot simply clear this bit when enabling the PHY, since at this stage the link may not come up (e.g. because the EP PHY is disabled), and so we need to clear it on the first access after link-up. In order to ensure this= , I simply clear it on every config access, in cdns_pci_map_bus, I couldn't thi= nk of a good way to just do it once. Regards, Alan > Thank you, > Lorenzo >=20 > > Signed-off-by: Alan Douglas > > --- > > drivers/pci/controller/pcie-cadence-ep.c | 1 + > > drivers/pci/controller/pcie-cadence-host.c | 3 ++ > > drivers/pci/controller/pcie-cadence.c | 30 ++++++++++++++++++++= ++++++++ > > drivers/pci/controller/pcie-cadence.h | 4 +++ > > 4 files changed, 38 insertions(+), 0 deletions(-) > > > > diff --git a/drivers/pci/controller/pcie-cadence-ep.c b/drivers/pci/con= troller/pcie-cadence-ep.c > > index c02f33d..3eabd99 100644 > > --- a/drivers/pci/controller/pcie-cadence-ep.c > > +++ b/drivers/pci/controller/pcie-cadence-ep.c > > @@ -555,6 +555,7 @@ static void cdns_pcie_ep_shutdown(struct platform_d= evice *pdev) > > .driver =3D { > > .name =3D "cdns-pcie-ep", > > .of_match_table =3D cdns_pcie_ep_of_match, > > + .pm =3D &cdns_pcie_pm_ops, > > }, > > .probe =3D cdns_pcie_ep_probe, > > .shutdown =3D cdns_pcie_ep_shutdown, > > diff --git a/drivers/pci/controller/pcie-cadence-host.c b/drivers/pci/c= ontroller/pcie-cadence-host.c > > index 36f3109..e3e9b7d 100644 > > --- a/drivers/pci/controller/pcie-cadence-host.c > > +++ b/drivers/pci/controller/pcie-cadence-host.c > > @@ -61,6 +61,8 @@ struct cdns_pcie_rc { > > /* Check that the link is up */ > > if (!(cdns_pcie_readl(pcie, CDNS_PCIE_LM_BASE) & 0x1)) > > return NULL; > > + /* Clear AXI link-down status */ > > + cdns_pcie_writel(pcie, CDNS_PCIE_AT_LINKDOWN, 0x0); > > > > /* Update Output registers for AXI region 0. */ > > addr0 =3D CDNS_PCIE_AT_OB_REGION_PCI_ADDR0_NBITS(12) | > > @@ -345,6 +347,7 @@ static int cdns_pcie_host_probe(struct platform_dev= ice *pdev) > > .driver =3D { > > .name =3D "cdns-pcie-host", > > .of_match_table =3D cdns_pcie_host_of_match, > > + .pm =3D &cdns_pcie_pm_ops, > > }, > > .probe =3D cdns_pcie_host_probe, > > }; > > diff --git a/drivers/pci/controller/pcie-cadence.c b/drivers/pci/contro= ller/pcie-cadence.c > > index 2edc126..86f1b00 100644 > > --- a/drivers/pci/controller/pcie-cadence.c > > +++ b/drivers/pci/controller/pcie-cadence.c > > @@ -217,3 +217,33 @@ int cdns_pcie_init_phy(struct device *dev, struct = cdns_pcie *pcie) > > > > return ret; > > } > > + > > +#ifdef CONFIG_PM_SLEEP > > +static int cdns_pcie_suspend_noirq(struct device *dev) > > +{ > > + struct cdns_pcie *pcie =3D dev_get_drvdata(dev); > > + > > + cdns_pcie_disable_phy(pcie); > > + > > + return 0; > > +} > > + > > +static int cdns_pcie_resume_noirq(struct device *dev) > > +{ > > + struct cdns_pcie *pcie =3D dev_get_drvdata(dev); > > + int ret; > > + > > + ret =3D cdns_pcie_enable_phy(pcie); > > + if (ret) { > > + dev_err(dev, "failed to enable phy\n"); > > + return ret; > > + } > > + > > + return 0; > > +} > > +#endif > > + > > +const struct dev_pm_ops cdns_pcie_pm_ops =3D { > > + SET_NOIRQ_SYSTEM_SLEEP_PM_OPS(cdns_pcie_suspend_noirq, > > + cdns_pcie_resume_noirq) > > +}; > > diff --git a/drivers/pci/controller/pcie-cadence.h b/drivers/pci/contro= ller/pcie-cadence.h > > index b342c80..ae6bf2a 100644 > > --- a/drivers/pci/controller/pcie-cadence.h > > +++ b/drivers/pci/controller/pcie-cadence.h > > @@ -166,6 +166,9 @@ > > #define CDNS_PCIE_AT_IB_RP_BAR_ADDR1(bar) \ > > (CDNS_PCIE_AT_BASE + 0x0804 + (bar) * 0x0008) > > > > +/* AXI link down register */ > > +#define CDNS_PCIE_AT_LINKDOWN (CDNS_PCIE_AT_BASE + 0x0824) > > + > > enum cdns_pcie_rp_bar { > > RP_BAR0, > > RP_BAR1, > > @@ -314,5 +317,6 @@ void cdns_pcie_set_outbound_region_for_normal_msg(s= truct cdns_pcie *pcie, u8 fn, > > void cdns_pcie_disable_phy(struct cdns_pcie *pcie); > > int cdns_pcie_enable_phy(struct cdns_pcie *pcie); > > int cdns_pcie_init_phy(struct device *dev, struct cdns_pcie *pcie); > > +extern const struct dev_pm_ops cdns_pcie_pm_ops; > > > > #endif /* _PCIE_CADENCE_H */ > > -- > > 1.7.1 > > From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-eopbgr690042.outbound.protection.outlook.com ([40.107.69.42]:33024 "EHLO NAM04-CO1-obe.outbound.protection.outlook.com" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S933647AbeGIP4O (ORCPT ); Mon, 9 Jul 2018 11:56:14 -0400 From: Alan Douglas To: Lorenzo Pieralisi CC: "bhelgaas@google.com" , "kishon@ti.com" , "linux-pci@vger.kernel.org" , "devicetree@vger.kernel.org" , "robh+dt@kernel.org" Subject: RE: [PATCH 4/5] PCI: cadence: Add Power Management ops for host and EP Date: Mon, 9 Jul 2018 15:56:11 +0000 Message-ID: References: <1529915453-4633-1-git-send-email-adouglas@cadence.com> <1529915453-4633-5-git-send-email-adouglas@cadence.com> <20180709153351.GA16292@red-moon> In-Reply-To: <20180709153351.GA16292@red-moon> Content-Type: text/plain; charset="us-ascii" MIME-Version: 1.0 Sender: linux-pci-owner@vger.kernel.org List-ID: On 09 July 2018 16:34, Lorenzo Pieralisi wrote: > On Mon, Jun 25, 2018 at 09:30:52AM +0100, Alan Douglas wrote: > > These PM ops will enable/disable the optional PHYs if present. The > > AXI link-down register in the host driver is now cleared in > > cdns_pci_map_bus since the link-down bit will be set if the PHY has > > been disabled. It is not cleared when enabling the PHY, since the > > link will not yet be up. >=20 > It is not entirely clear what you mean here, can you elaborate please ? Thanks for taking a look at this. There is a "Link down indication bit" register in the cadence PCIe IP AXI Configuration Registers, which will be set by the PCIe IP after each link- down occurrence. This bit must be cleared before continuing, or accesses will hang. When the PHY is disabled, the link will go down and this bit wi= ll be set by HW. It will also be set after a secondary bus reset. We cannot simply clear this bit when enabling the PHY, since at this stage the link may not come up (e.g. because the EP PHY is disabled), and so we need to clear it on the first access after link-up. In order to ensure this= , I simply clear it on every config access, in cdns_pci_map_bus, I couldn't thi= nk of a good way to just do it once. Regards, Alan > Thank you, > Lorenzo >=20 > > Signed-off-by: Alan Douglas > > --- > > drivers/pci/controller/pcie-cadence-ep.c | 1 + > > drivers/pci/controller/pcie-cadence-host.c | 3 ++ > > drivers/pci/controller/pcie-cadence.c | 30 ++++++++++++++++++++= ++++++++ > > drivers/pci/controller/pcie-cadence.h | 4 +++ > > 4 files changed, 38 insertions(+), 0 deletions(-) > > > > diff --git a/drivers/pci/controller/pcie-cadence-ep.c b/drivers/pci/con= troller/pcie-cadence-ep.c > > index c02f33d..3eabd99 100644 > > --- a/drivers/pci/controller/pcie-cadence-ep.c > > +++ b/drivers/pci/controller/pcie-cadence-ep.c > > @@ -555,6 +555,7 @@ static void cdns_pcie_ep_shutdown(struct platform_d= evice *pdev) > > .driver =3D { > > .name =3D "cdns-pcie-ep", > > .of_match_table =3D cdns_pcie_ep_of_match, > > + .pm =3D &cdns_pcie_pm_ops, > > }, > > .probe =3D cdns_pcie_ep_probe, > > .shutdown =3D cdns_pcie_ep_shutdown, > > diff --git a/drivers/pci/controller/pcie-cadence-host.c b/drivers/pci/c= ontroller/pcie-cadence-host.c > > index 36f3109..e3e9b7d 100644 > > --- a/drivers/pci/controller/pcie-cadence-host.c > > +++ b/drivers/pci/controller/pcie-cadence-host.c > > @@ -61,6 +61,8 @@ struct cdns_pcie_rc { > > /* Check that the link is up */ > > if (!(cdns_pcie_readl(pcie, CDNS_PCIE_LM_BASE) & 0x1)) > > return NULL; > > + /* Clear AXI link-down status */ > > + cdns_pcie_writel(pcie, CDNS_PCIE_AT_LINKDOWN, 0x0); > > > > /* Update Output registers for AXI region 0. */ > > addr0 =3D CDNS_PCIE_AT_OB_REGION_PCI_ADDR0_NBITS(12) | > > @@ -345,6 +347,7 @@ static int cdns_pcie_host_probe(struct platform_dev= ice *pdev) > > .driver =3D { > > .name =3D "cdns-pcie-host", > > .of_match_table =3D cdns_pcie_host_of_match, > > + .pm =3D &cdns_pcie_pm_ops, > > }, > > .probe =3D cdns_pcie_host_probe, > > }; > > diff --git a/drivers/pci/controller/pcie-cadence.c b/drivers/pci/contro= ller/pcie-cadence.c > > index 2edc126..86f1b00 100644 > > --- a/drivers/pci/controller/pcie-cadence.c > > +++ b/drivers/pci/controller/pcie-cadence.c > > @@ -217,3 +217,33 @@ int cdns_pcie_init_phy(struct device *dev, struct = cdns_pcie *pcie) > > > > return ret; > > } > > + > > +#ifdef CONFIG_PM_SLEEP > > +static int cdns_pcie_suspend_noirq(struct device *dev) > > +{ > > + struct cdns_pcie *pcie =3D dev_get_drvdata(dev); > > + > > + cdns_pcie_disable_phy(pcie); > > + > > + return 0; > > +} > > + > > +static int cdns_pcie_resume_noirq(struct device *dev) > > +{ > > + struct cdns_pcie *pcie =3D dev_get_drvdata(dev); > > + int ret; > > + > > + ret =3D cdns_pcie_enable_phy(pcie); > > + if (ret) { > > + dev_err(dev, "failed to enable phy\n"); > > + return ret; > > + } > > + > > + return 0; > > +} > > +#endif > > + > > +const struct dev_pm_ops cdns_pcie_pm_ops =3D { > > + SET_NOIRQ_SYSTEM_SLEEP_PM_OPS(cdns_pcie_suspend_noirq, > > + cdns_pcie_resume_noirq) > > +}; > > diff --git a/drivers/pci/controller/pcie-cadence.h b/drivers/pci/contro= ller/pcie-cadence.h > > index b342c80..ae6bf2a 100644 > > --- a/drivers/pci/controller/pcie-cadence.h > > +++ b/drivers/pci/controller/pcie-cadence.h > > @@ -166,6 +166,9 @@ > > #define CDNS_PCIE_AT_IB_RP_BAR_ADDR1(bar) \ > > (CDNS_PCIE_AT_BASE + 0x0804 + (bar) * 0x0008) > > > > +/* AXI link down register */ > > +#define CDNS_PCIE_AT_LINKDOWN (CDNS_PCIE_AT_BASE + 0x0824) > > + > > enum cdns_pcie_rp_bar { > > RP_BAR0, > > RP_BAR1, > > @@ -314,5 +317,6 @@ void cdns_pcie_set_outbound_region_for_normal_msg(s= truct cdns_pcie *pcie, u8 fn, > > void cdns_pcie_disable_phy(struct cdns_pcie *pcie); > > int cdns_pcie_enable_phy(struct cdns_pcie *pcie); > > int cdns_pcie_init_phy(struct device *dev, struct cdns_pcie *pcie); > > +extern const struct dev_pm_ops cdns_pcie_pm_ops; > > > > #endif /* _PCIE_CADENCE_H */ > > -- > > 1.7.1 > >