Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Sat, 9 Jul 2022 11:34:39 +0530
From:   Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
To:     Kunihiko Hayashi <hayashi.kunihiko@socionext.com>
Cc:     kishon@ti.com, bhelgaas@google.com, robh@kernel.org,
        lorenzo.pieralisi@arm.com, kw@linux.com, linux-pci@vger.kernel.org,
        linux-kernel@vger.kernel.org, Om Prakash Singh <omp@nvidia.com>,
        Vidya Sagar <vidyas@nvidia.com>
Subject: Re: [PATCH] PCI: designware-ep: Move DBI access to init_complete if
 notifier is used
Message-ID: <20220709060439.GA5063@thinkpad>
References: <20220330060515.22328-1-manivannan.sadhasivam@linaro.org>
 <20220427101820.GC2536@thinkpad>
 <20220707165751.GB5458@thinkpad>
 <29c5a103-6f91-e978-6d80-45db6dc4a18b@socionext.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <29c5a103-6f91-e978-6d80-45db6dc4a18b@socionext.com>
Precedence: bulk

On Sat, Jul 09, 2022 at 10:42:32AM +0900, Kunihiko Hayashi wrote:
> Hi Manivannan,
> 
> On 2022/07/08 1:57, Manivannan Sadhasivam wrote:
> > On Wed, Apr 27, 2022 at 03:48:26PM +0530, Manivannan Sadhasivam wrote:
> > > On Wed, Mar 30, 2022 at 11:35:15AM +0530, Manivannan Sadhasivam wrote:
> > > > For controllers supporting the CORE_INIT notifier, the resources are
> > > > supposed to be enabled in the init_complete function. Currently,
> > > > these controllers are enabling the resources during probe time due to
> > > > the DBI access happens in dw_pcie_ep_init().
> > > > 
> > > > This creates the dependency with the host PCIe controller since the
> > > > resource enablement like PHY depends on host PCIe to be up. For the
> > > > standalone endpoint usecase, this would never work. So let's move all
> > > > DBI
> > > > access to init_complete function if CORE_INIT notifier is used. For the
> > > > controllers those doesn't support this notifier, this change is a NO-OP.
> > > > 
> > > > Cc: Kunihiko Hayashi <hayashi.kunihiko@socionext.com>
> > > > Cc: Om Prakash Singh <omp@nvidia.com>
> > > > Cc: Vidya Sagar <vidyas@nvidia.com>
> > > > Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
> > > 
> > > A gentle ping on this patch!
> > > 
> 
> Sorry I couldn't check it.
> I think the original source seems to have changed in the latest, so please
> rebase and fix some conflicts.
> 

Sure, will do.

> > Ping again!
> > 
> > Thanks,
> > Mani
> > 
> > > Thanks,
> > > Mani
> > > 
> > > > ---
> > > >   .../pci/controller/dwc/pcie-designware-ep.c   | 138 ++++++++++++------
> > > >   drivers/pci/controller/dwc/pcie-designware.h  |   1 +
> > > >   2 files changed, 94 insertions(+), 45 deletions(-)
> > > > 
> > > > diff --git a/drivers/pci/controller/dwc/pcie-designware-ep.c
> > > > b/drivers/pci/controller/dwc/pcie-designware-ep.c
> > > > index 0eda8236c125..fb2bf4bf5ba0 100644
> > > > --- a/drivers/pci/controller/dwc/pcie-designware-ep.c
> > > > +++ b/drivers/pci/controller/dwc/pcie-designware-ep.c
> > > > @@ -636,6 +636,63 @@ static unsigned int
> > > > dw_pcie_ep_find_ext_capability(struct dw_pcie *pci, int cap)
> > > >   	return 0;
> > > >   }
> > > > 
> > > > +static int dw_pcie_iatu_config(struct dw_pcie_ep *ep)
> > > > +{
> > > > +	struct dw_pcie *pci = to_dw_pcie_from_ep(ep);
> > > > +	struct device *dev = pci->dev;
> > > > +	void *addr;
> > > > +
> > > > +	dw_pcie_iatu_detect(pci);
> > > > +
> > > > +	ep->ib_window_map = devm_kcalloc(dev,
> > > > +					 BITS_TO_LONGS(pci->num_ib_windows),
> > > > +					 sizeof(long),
> > > > +					 GFP_KERNEL);
> > > > +	if (!ep->ib_window_map)
> > > > +		return -ENOMEM;
> > > > +
> > > > +	ep->ob_window_map = devm_kcalloc(dev,
> > > > +					 BITS_TO_LONGS(pci->num_ob_windows),
> > > > +					 sizeof(long),
> > > > +					 GFP_KERNEL);
> > > > +	if (!ep->ob_window_map)
> > > > +		return -ENOMEM;
> > > > +
> > > > +	addr = devm_kcalloc(dev, pci->num_ob_windows, sizeof(phys_addr_t),
> > > > +			    GFP_KERNEL);
> > > > +	if (!addr)
> > > > +		return -ENOMEM;
> > > > +
> > > > +	ep->outbound_addr = addr;
> > > > +
> > > > +	return 0;
> > > > +}
> > > > +
> > > > +static int dw_pcie_ep_func_init(struct dw_pcie_ep *ep)
> > > > +{
> > > > +	struct dw_pcie *pci = to_dw_pcie_from_ep(ep);
> > > > +	struct dw_pcie_ep_func *ep_func;
> > > > +	struct pci_epc *epc = ep->epc;
> > > > +	struct device *dev = pci->dev;
> > > > +	u8 func_no;
> > > > +
> > > > +	for (func_no = 0; func_no < epc->max_functions; func_no++) {
> > > > +		ep_func = devm_kzalloc(dev, sizeof(*ep_func), GFP_KERNEL);
> > > > +		if (!ep_func)
> > > > +			return -ENOMEM;
> > > > +
> > > > +		ep_func->func_no = func_no;
> > > > +		ep_func->msi_cap = dw_pcie_ep_find_capability(ep, func_no,
> > > > +							      PCI_CAP_ID_MSI);
> > > > +		ep_func->msix_cap = dw_pcie_ep_find_capability(ep, func_no,
> > > > +							       PCI_CAP_ID_MSIX);
> > > > +
> > > > +		list_add_tail(&ep_func->list, &ep->func_list);
> > > > +	}
> > > > +
> > > > +	return 0;
> > > > +}
> > > > +
> > > >   int dw_pcie_ep_init_complete(struct dw_pcie_ep *ep)
> > > >   {
> > > >   	struct dw_pcie *pci = to_dw_pcie_from_ep(ep);
> > > > @@ -643,7 +700,22 @@ int dw_pcie_ep_init_complete(struct dw_pcie_ep *ep)
> > > >   	unsigned int nbars;
> > > >   	u8 hdr_type;
> > > >   	u32 reg;
> > > > -	int i;
> > > > +	int ret, i;
> > > > +
> > > > +	if (ep->core_init_notifier) {
> > > > +		ret = dw_pcie_iatu_config(ep);
> > > > +		if (ret)
> > > > +			return ret;
> > > > +	}
> > > > +
> > > > +	if (ep->core_init_notifier) {
> > > > +		ret = dw_pcie_ep_func_init(ep);
> > > > +		if (ret)
> > > > +			return ret;
> > > > +
> > > > +		if (ep->ops->ep_init)
> > > > +			ep->ops->ep_init(ep);
> > > > +	}
> 
> It isn't necessary to divide function calls under same condition
> (ep->core_init_notifier).
> 

Right. I think one of the part came from previous patch, so I forgot to merge
them.

> > > > 
> > > >   	hdr_type = dw_pcie_readb_dbi(pci, PCI_HEADER_TYPE) &
> > > >   		   PCI_HEADER_TYPE_MASK;
> > > > @@ -677,8 +749,6 @@ EXPORT_SYMBOL_GPL(dw_pcie_ep_init_complete);
> > > >   int dw_pcie_ep_init(struct dw_pcie_ep *ep)
> > > >   {
> > > >   	int ret;
> > > > -	void *addr;
> > > > -	u8 func_no;
> > > >   	struct resource *res;
> > > >   	struct pci_epc *epc;
> > > >   	struct dw_pcie *pci = to_dw_pcie_from_ep(ep);
> > > > @@ -686,7 +756,12 @@ int dw_pcie_ep_init(struct dw_pcie_ep *ep)
> > > >   	struct platform_device *pdev = to_platform_device(dev);
> > > >   	struct device_node *np = dev->of_node;
> > > >   	const struct pci_epc_features *epc_features;
> > > > -	struct dw_pcie_ep_func *ep_func;
> > > > +
> > > > +	if (ep->ops->get_features) {
> > > > +		epc_features = ep->ops->get_features(ep);
> > > > +		if (epc_features->core_init_notifier)
> > > > +			ep->core_init_notifier = true;
> > > > +	}
> > > > 
> > > >   	INIT_LIST_HEAD(&ep->func_list);
> > > > 
> > > > @@ -708,7 +783,11 @@ int dw_pcie_ep_init(struct dw_pcie_ep *ep)
> > > >   		}
> > > >   	}
> > > > 
> > > > -	dw_pcie_iatu_detect(pci);
> > > > +	if (!ep->core_init_notifier) {
> > > > +		ret = dw_pcie_iatu_config(ep);
> > > > +		if (ret)
> > > > +			return ret;
> > > > +	}
> > > > 
> > > >   	res = platform_get_resource_byname(pdev, IORESOURCE_MEM,
> > > > "addr_space");
> > > >   	if (!res)
> > > > @@ -717,26 +796,6 @@ int dw_pcie_ep_init(struct dw_pcie_ep *ep)
> > > >   	ep->phys_base = res->start;
> > > >   	ep->addr_size = resource_size(res);
> > > > 
> > > > -	ep->ib_window_map = devm_kcalloc(dev,
> > > > -					 BITS_TO_LONGS(pci->num_ib_windows),
> > > > -					 sizeof(long),
> > > > -					 GFP_KERNEL);
> > > > -	if (!ep->ib_window_map)
> > > > -		return -ENOMEM;
> > > > -
> > > > -	ep->ob_window_map = devm_kcalloc(dev,
> > > > -					 BITS_TO_LONGS(pci->num_ob_windows),
> > > > -					 sizeof(long),
> > > > -					 GFP_KERNEL);
> > > > -	if (!ep->ob_window_map)
> > > > -		return -ENOMEM;
> > > > -
> > > > -	addr = devm_kcalloc(dev, pci->num_ob_windows, sizeof(phys_addr_t),
> > > > -			    GFP_KERNEL);
> > > > -	if (!addr)
> > > > -		return -ENOMEM;
> > > > -	ep->outbound_addr = addr;
> > > > -
> > > >   	if (pci->link_gen < 1)
> > > >   		pci->link_gen = of_pci_get_max_link_speed(np);
> > > > 
> > > > @@ -753,23 +812,15 @@ int dw_pcie_ep_init(struct dw_pcie_ep *ep)
> > > >   	if (ret < 0)
> > > >   		epc->max_functions = 1;
> > > > 
> > > > -	for (func_no = 0; func_no < epc->max_functions; func_no++) {
> > > > -		ep_func = devm_kzalloc(dev, sizeof(*ep_func), GFP_KERNEL);
> > > > -		if (!ep_func)
> > > > -			return -ENOMEM;
> > > > -
> > > > -		ep_func->func_no = func_no;
> > > > -		ep_func->msi_cap = dw_pcie_ep_find_capability(ep, func_no,
> > > > -							      PCI_CAP_ID_MSI);
> > > > -		ep_func->msix_cap = dw_pcie_ep_find_capability(ep, func_no,
> > > > -							       PCI_CAP_ID_MSIX);
> > > > +	if (!ep->core_init_notifier) {
> > > > +		ret = dw_pcie_ep_func_init(ep);
> > > > +		if (ret)
> > > > +			return ret;
> > > > 
> > > > -		list_add_tail(&ep_func->list, &ep->func_list);
> > > > +		if (ep->ops->ep_init)
> > > > +			ep->ops->ep_init(ep);
> > > >   	}
> > > > 
> > > > -	if (ep->ops->ep_init)
> > > > -		ep->ops->ep_init(ep);
> > > > -
> > > >   	ret = pci_epc_mem_init(epc, ep->phys_base, ep->addr_size,
> > > >   			       ep->page_size);
> > > >   	if (ret < 0) {
> > > > @@ -784,12 +835,9 @@ int dw_pcie_ep_init(struct dw_pcie_ep *ep)
> > > >   		return -ENOMEM;
> > > >   	}
> > > > 
> > > > -	if (ep->ops->get_features) {
> > > > -		epc_features = ep->ops->get_features(ep);
> > > > -		if (epc_features->core_init_notifier)
> > > > -			return 0;
> > > > -	}
> > > > +	if (!ep->core_init_notifier)
> > > > +		return dw_pcie_ep_init_complete(ep);
> > > > 
> > > > -	return dw_pcie_ep_init_complete(ep);
> > > > +	return 0;
> > > >   }
> > > >   EXPORT_SYMBOL_GPL(dw_pcie_ep_init);
> 
> Since devm_pci_epc_create() is called in dw_pcie_ep_init(), EPC functions
> are available at this time, however, the driver can't access any controller
> registers until PERST# signal interrupt occurs.
> 
> I think that EPC functions aren't called until PERST# signal, but
> I'm worried about whether controller access will occur.
> 

Most of the EPC accesses are carried out by EPF and that won't be available
until dw_pcie_ep_init_complete() gets completed for core_init_notifier case.
This is one of the concerns Kishon raised but I think here we are safe.

Thanks,
Mani

> > > > diff --git a/drivers/pci/controller/dwc/pcie-designware.h
> > > > b/drivers/pci/controller/dwc/pcie-designware.h
> > > > index 7d6e9b7576be..aadb14159df7 100644
> > > > --- a/drivers/pci/controller/dwc/pcie-designware.h
> > > > +++ b/drivers/pci/controller/dwc/pcie-designware.h
> > > > @@ -242,6 +242,7 @@ struct dw_pcie_ep {
> > > >   	void __iomem		*msi_mem;
> > > >   	phys_addr_t		msi_mem_phys;
> > > >   	struct pci_epf_bar	*epf_bar[PCI_STD_NUM_BARS];
> > > > +	bool			core_init_notifier;
> > > >   };
> > > > 
> > > >   struct dw_pcie_ops {
> > > > -- 
> > > > 2.25.1
> > > > 
> 
> Thank you,
> 
> ---
> Best Regards
> Kunihiko Hayashi

-- 
மணிவண்ணன் சதாசிவம்