diff mbox series

[v2,2/2] PCI: designware-ep: Fix the access to DBI/iATU registers before enabling controller

Message ID 1630473361-27198-3-git-send-email-hayashi.kunihiko@socionext.com (mailing list archive)
State Changes Requested
Delegated to: Lorenzo Pieralisi
Headers show
Series PCI: endpoint: Fix core_init_notifier feature | expand

Commit Message

Kunihiko Hayashi Sept. 1, 2021, 5:16 a.m. UTC
The driver using core_init_notifier, e.g. pcie-tegra194.c, runs according
to the following sequence:

    probe()
        dw_pcie_ep_init()

    bind()
        dw_pcie_ep_start()
            enable_irq()

    (interrupt occurred)
    handler()
        [enable controller]
        dw_pcie_ep_init_complete()
        dw_pcie_ep_init_notify()

After receiving an interrupt from RC, the handler enables the controller
and the controller registers can be accessed.
So accessing the registers should do in dw_pcie_ep_init_complete().

Currently dw_pcie_ep_init() has functions dw_iatu_detect() and
dw_pcie_ep_find_capability() that include accesses to DWC registers.
As a result, accessing the registers before enabling the controller,
the access will fail.

The function dw_pcie_ep_init() shouldn't have any access to DWC registers
if the controller is enabled after calling bind(). This moves access codes
to DBI/iATU registers and depending variables from dw_pcie_ep_init() to
dw_pcie_ep_init_complete().

Cc: Xiaowei Bao <xiaowei.bao@nxp.com>
Cc: Vidya Sagar <vidyas@nvidia.com>
Fixes: 6bfc9c3a2c70 ("PCI: designware-ep: Move the function of getting MSI capability forward")
Signed-off-by: Kunihiko Hayashi <hayashi.kunihiko@socionext.com>
Acked-by: Om Prakash Singh <omp@nvidia.com>
Reviewed-by: Vidya Sagar <vidyas@nvidia.com>
---
 drivers/pci/controller/dwc/pcie-designware-ep.c | 81 +++++++++++++------------
 1 file changed, 41 insertions(+), 40 deletions(-)

Comments

Kishon Vijay Abraham I Dec. 3, 2021, 5:06 a.m. UTC | #1
Hi Kunihiko,

On 01/09/21 10:46 am, Kunihiko Hayashi wrote:
> The driver using core_init_notifier, e.g. pcie-tegra194.c, runs according
> to the following sequence:
> 
>     probe()
>         dw_pcie_ep_init()
> 
>     bind()
>         dw_pcie_ep_start()
>             enable_irq()
> 
>     (interrupt occurred)
>     handler()
>         [enable controller]
>         dw_pcie_ep_init_complete()
>         dw_pcie_ep_init_notify()
> 
> After receiving an interrupt from RC, the handler enables the controller
> and the controller registers can be accessed.
> So accessing the registers should do in dw_pcie_ep_init_complete().
> 
> Currently dw_pcie_ep_init() has functions dw_iatu_detect() and
> dw_pcie_ep_find_capability() that include accesses to DWC registers.
> As a result, accessing the registers before enabling the controller,
> the access will fail.
> 
> The function dw_pcie_ep_init() shouldn't have any access to DWC registers
> if the controller is enabled after calling bind(). This moves access codes
> to DBI/iATU registers and depending variables from dw_pcie_ep_init() to
> dw_pcie_ep_init_complete().

Ideally pci_epc_create() should be the last step by the controller driver before
handing the control to the core EPC framework. Since after this step the EPC
framework can start invoking the epc_ops.

Here more stuff is being added to dw_pcie_ep_init_complete() which is required
for epc_ops and this could result in aborts for platforms which does not add
core_init_notifier.

Thanks,
Kishon

> 
> Cc: Xiaowei Bao <xiaowei.bao@nxp.com>
> Cc: Vidya Sagar <vidyas@nvidia.com>
> Fixes: 6bfc9c3a2c70 ("PCI: designware-ep: Move the function of getting MSI capability forward")
> Signed-off-by: Kunihiko Hayashi <hayashi.kunihiko@socionext.com>
> Acked-by: Om Prakash Singh <omp@nvidia.com>
> Reviewed-by: Vidya Sagar <vidyas@nvidia.com>
> ---
>  drivers/pci/controller/dwc/pcie-designware-ep.c | 81 +++++++++++++------------
>  1 file changed, 41 insertions(+), 40 deletions(-)
> 
> diff --git a/drivers/pci/controller/dwc/pcie-designware-ep.c b/drivers/pci/controller/dwc/pcie-designware-ep.c
> index 998b698..00ce83c 100644
> --- a/drivers/pci/controller/dwc/pcie-designware-ep.c
> +++ b/drivers/pci/controller/dwc/pcie-designware-ep.c
> @@ -636,16 +636,56 @@ static unsigned int dw_pcie_ep_find_ext_capability(struct dw_pcie *pci, int cap)
>  int dw_pcie_ep_init_complete(struct dw_pcie_ep *ep)
>  {
>  	struct dw_pcie *pci = to_dw_pcie_from_ep(ep);
> +	struct dw_pcie_ep_func *ep_func;
> +	struct device *dev = pci->dev;
>  	unsigned int offset;
>  	unsigned int nbars;
>  	u8 hdr_type;
> +	u8 func_no;
> +	void *addr;
>  	u32 reg;
>  	int i;
>  
> +	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;
> +
> +	for (func_no = 0; func_no < ep->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);
> +	}
> +
>  	hdr_type = dw_pcie_readb_dbi(pci, PCI_HEADER_TYPE) &
>  		   PCI_HEADER_TYPE_MASK;
>  	if (hdr_type != PCI_HEADER_TYPE_NORMAL) {
> -		dev_err(pci->dev,
> +		dev_err(dev,
>  			"PCIe controller is not set to EP mode (hdr_type:0x%x)!\n",
>  			hdr_type);
>  		return -EIO;
> @@ -674,8 +714,6 @@ 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);
> @@ -683,7 +721,6 @@ 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;
>  
>  	INIT_LIST_HEAD(&ep->func_list);
>  
> @@ -705,8 +742,6 @@ int dw_pcie_ep_init(struct dw_pcie_ep *ep)
>  		}
>  	}
>  
> -	dw_pcie_iatu_detect(pci);
> -
>  	res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "addr_space");
>  	if (!res)
>  		return -EINVAL;
> @@ -714,26 +749,6 @@ 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);
>  
> @@ -750,20 +765,6 @@ 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);
> -
> -		list_add_tail(&ep_func->list, &ep->func_list);
> -	}
> -
>  	if (ep->ops->ep_init)
>  		ep->ops->ep_init(ep);
>  
>
Lorenzo Pieralisi Dec. 6, 2021, 11:23 a.m. UTC | #2
On Fri, Dec 03, 2021 at 10:36:00AM +0530, Kishon Vijay Abraham I wrote:
> Hi Kunihiko,
> 
> On 01/09/21 10:46 am, Kunihiko Hayashi wrote:
> > The driver using core_init_notifier, e.g. pcie-tegra194.c, runs according
> > to the following sequence:
> > 
> >     probe()
> >         dw_pcie_ep_init()
> > 
> >     bind()
> >         dw_pcie_ep_start()
> >             enable_irq()
> > 
> >     (interrupt occurred)
> >     handler()
> >         [enable controller]
> >         dw_pcie_ep_init_complete()
> >         dw_pcie_ep_init_notify()
> > 
> > After receiving an interrupt from RC, the handler enables the controller
> > and the controller registers can be accessed.
> > So accessing the registers should do in dw_pcie_ep_init_complete().
> > 
> > Currently dw_pcie_ep_init() has functions dw_iatu_detect() and
> > dw_pcie_ep_find_capability() that include accesses to DWC registers.
> > As a result, accessing the registers before enabling the controller,
> > the access will fail.
> > 
> > The function dw_pcie_ep_init() shouldn't have any access to DWC registers
> > if the controller is enabled after calling bind(). This moves access codes
> > to DBI/iATU registers and depending variables from dw_pcie_ep_init() to
> > dw_pcie_ep_init_complete().
> 
> Ideally pci_epc_create() should be the last step by the controller
> driver before handing the control to the core EPC framework. Since
> after this step the EPC framework can start invoking the epc_ops.
> 
> Here more stuff is being added to dw_pcie_ep_init_complete() which is
> required for epc_ops and this could result in aborts for platforms
> which does not add core_init_notifier.

This patch needs rework, I will mark the series as "Changes requested".

Lorenzo

> 
> Thanks,
> Kishon
> 
> > 
> > Cc: Xiaowei Bao <xiaowei.bao@nxp.com>
> > Cc: Vidya Sagar <vidyas@nvidia.com>
> > Fixes: 6bfc9c3a2c70 ("PCI: designware-ep: Move the function of getting MSI capability forward")
> > Signed-off-by: Kunihiko Hayashi <hayashi.kunihiko@socionext.com>
> > Acked-by: Om Prakash Singh <omp@nvidia.com>
> > Reviewed-by: Vidya Sagar <vidyas@nvidia.com>
> > ---
> >  drivers/pci/controller/dwc/pcie-designware-ep.c | 81 +++++++++++++------------
> >  1 file changed, 41 insertions(+), 40 deletions(-)
> > 
> > diff --git a/drivers/pci/controller/dwc/pcie-designware-ep.c b/drivers/pci/controller/dwc/pcie-designware-ep.c
> > index 998b698..00ce83c 100644
> > --- a/drivers/pci/controller/dwc/pcie-designware-ep.c
> > +++ b/drivers/pci/controller/dwc/pcie-designware-ep.c
> > @@ -636,16 +636,56 @@ static unsigned int dw_pcie_ep_find_ext_capability(struct dw_pcie *pci, int cap)
> >  int dw_pcie_ep_init_complete(struct dw_pcie_ep *ep)
> >  {
> >  	struct dw_pcie *pci = to_dw_pcie_from_ep(ep);
> > +	struct dw_pcie_ep_func *ep_func;
> > +	struct device *dev = pci->dev;
> >  	unsigned int offset;
> >  	unsigned int nbars;
> >  	u8 hdr_type;
> > +	u8 func_no;
> > +	void *addr;
> >  	u32 reg;
> >  	int i;
> >  
> > +	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;
> > +
> > +	for (func_no = 0; func_no < ep->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);
> > +	}
> > +
> >  	hdr_type = dw_pcie_readb_dbi(pci, PCI_HEADER_TYPE) &
> >  		   PCI_HEADER_TYPE_MASK;
> >  	if (hdr_type != PCI_HEADER_TYPE_NORMAL) {
> > -		dev_err(pci->dev,
> > +		dev_err(dev,
> >  			"PCIe controller is not set to EP mode (hdr_type:0x%x)!\n",
> >  			hdr_type);
> >  		return -EIO;
> > @@ -674,8 +714,6 @@ 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);
> > @@ -683,7 +721,6 @@ 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;
> >  
> >  	INIT_LIST_HEAD(&ep->func_list);
> >  
> > @@ -705,8 +742,6 @@ int dw_pcie_ep_init(struct dw_pcie_ep *ep)
> >  		}
> >  	}
> >  
> > -	dw_pcie_iatu_detect(pci);
> > -
> >  	res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "addr_space");
> >  	if (!res)
> >  		return -EINVAL;
> > @@ -714,26 +749,6 @@ 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);
> >  
> > @@ -750,20 +765,6 @@ 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);
> > -
> > -		list_add_tail(&ep_func->list, &ep->func_list);
> > -	}
> > -
> >  	if (ep->ops->ep_init)
> >  		ep->ops->ep_init(ep);
> >  
> >
Kunihiko Hayashi Jan. 5, 2022, 10:43 a.m. UTC | #3
Hi Kishon, Lorenzo,

Thank you and sorry for late reply.

On 2021/12/06 20:23, Lorenzo Pieralisi wrote:
> On Fri, Dec 03, 2021 at 10:36:00AM +0530, Kishon Vijay Abraham I wrote:
>> Hi Kunihiko,
>>
>> On 01/09/21 10:46 am, Kunihiko Hayashi wrote:
>>> The driver using core_init_notifier, e.g. pcie-tegra194.c, runs
> according
>>> to the following sequence:
>>>
>>>      probe()
>>>          dw_pcie_ep_init()
>>>
>>>      bind()
>>>          dw_pcie_ep_start()
>>>              enable_irq()
>>>
>>>      (interrupt occurred)
>>>      handler()
>>>          [enable controller]
>>>          dw_pcie_ep_init_complete()
>>>          dw_pcie_ep_init_notify()
>>>
>>> After receiving an interrupt from RC, the handler enables the
> controller
>>> and the controller registers can be accessed.
>>> So accessing the registers should do in dw_pcie_ep_init_complete().
>>>
>>> Currently dw_pcie_ep_init() has functions dw_iatu_detect() and
>>> dw_pcie_ep_find_capability() that include accesses to DWC registers.
>>> As a result, accessing the registers before enabling the controller,
>>> the access will fail.
>>>
>>> The function dw_pcie_ep_init() shouldn't have any access to DWC
> registers
>>> if the controller is enabled after calling bind(). This moves access
> codes
>>> to DBI/iATU registers and depending variables from dw_pcie_ep_init()
> to
>>> dw_pcie_ep_init_complete().
>>
>> Ideally pci_epc_create() should be the last step by the controller
>> driver before handing the control to the core EPC framework. Since
>> after this step the EPC framework can start invoking the epc_ops.
>>
>> Here more stuff is being added to dw_pcie_ep_init_complete() which is
>> required for epc_ops and this could result in aborts for platforms
>> which does not add core_init_notifier.
> 
> This patch needs rework, I will mark the series as "Changes requested".

I understand that relocation of dwc register accesses isn't appropriate,
but I couldn't think of any other rework to dwc, and I confirmed
pcie-qcom-ep driver using core_init_notifier.

In pcie-qcom-ep driver, probe() enables clock and deasserts reset first,
and when PERST# interrupt arrives, the handler enables clock and deasserts
reset again. So, dw_pcie_ep_init() can access DBI registers.

In pcie-tegra194 driver, I think the issue will be solved if probe() also
handles clock and reset control. However, the driver has other register
access between core_clk, core_apb_rst, and core_rst controls.
I think that it's appropriate to leave this fix to the developer at this
point.

As this patch series, I'll resend 1/2 patch only and expect pcie-tegra194
driver to be fixed.

Thank you,

---
Best Regards
Kunihiko Hayashi
Manivannan Sadhasivam Jan. 5, 2022, 3:46 p.m. UTC | #4
On Wed, Jan 05, 2022 at 07:43:04PM +0900, Kunihiko Hayashi wrote:
> Hi Kishon, Lorenzo,
> 
> Thank you and sorry for late reply.
> 
> On 2021/12/06 20:23, Lorenzo Pieralisi wrote:
> > On Fri, Dec 03, 2021 at 10:36:00AM +0530, Kishon Vijay Abraham I wrote:
> > > Hi Kunihiko,
> > > 
> > > On 01/09/21 10:46 am, Kunihiko Hayashi wrote:
> > > > The driver using core_init_notifier, e.g. pcie-tegra194.c, runs
> > according
> > > > to the following sequence:
> > > > 
> > > >      probe()
> > > >          dw_pcie_ep_init()
> > > > 
> > > >      bind()
> > > >          dw_pcie_ep_start()
> > > >              enable_irq()
> > > > 
> > > >      (interrupt occurred)
> > > >      handler()
> > > >          [enable controller]
> > > >          dw_pcie_ep_init_complete()
> > > >          dw_pcie_ep_init_notify()
> > > > 
> > > > After receiving an interrupt from RC, the handler enables the
> > controller
> > > > and the controller registers can be accessed.
> > > > So accessing the registers should do in dw_pcie_ep_init_complete().
> > > > 
> > > > Currently dw_pcie_ep_init() has functions dw_iatu_detect() and
> > > > dw_pcie_ep_find_capability() that include accesses to DWC registers.
> > > > As a result, accessing the registers before enabling the controller,
> > > > the access will fail.
> > > > 
> > > > The function dw_pcie_ep_init() shouldn't have any access to DWC
> > registers
> > > > if the controller is enabled after calling bind(). This moves access
> > codes
> > > > to DBI/iATU registers and depending variables from dw_pcie_ep_init()
> > to
> > > > dw_pcie_ep_init_complete().
> > > 
> > > Ideally pci_epc_create() should be the last step by the controller
> > > driver before handing the control to the core EPC framework. Since
> > > after this step the EPC framework can start invoking the epc_ops.
> > > 
> > > Here more stuff is being added to dw_pcie_ep_init_complete() which is
> > > required for epc_ops and this could result in aborts for platforms
> > > which does not add core_init_notifier.
> > 
> > This patch needs rework, I will mark the series as "Changes requested".
> 
> I understand that relocation of dwc register accesses isn't appropriate,
> but I couldn't think of any other rework to dwc, and I confirmed
> pcie-qcom-ep driver using core_init_notifier.
> 
> In pcie-qcom-ep driver, probe() enables clock and deasserts reset first,
> and when PERST# interrupt arrives, the handler enables clock and deasserts
> reset again. So, dw_pcie_ep_init() can access DBI registers.
> 

Yes, only since dw_pcie_ep_init() carries out the DBI accesses, we have enabled
clocks and PHY. Moving the DBI accesses to init_complete() removes the need of
enabling the resources redundantly.

Thanks,
Mani

> In pcie-tegra194 driver, I think the issue will be solved if probe() also
> handles clock and reset control. However, the driver has other register
> access between core_clk, core_apb_rst, and core_rst controls.
> I think that it's appropriate to leave this fix to the developer at this
> point.
> 
> As this patch series, I'll resend 1/2 patch only and expect pcie-tegra194
> driver to be fixed.
> 
> Thank you,
> 
> ---
> Best Regards
> Kunihiko Hayashi
Manivannan Sadhasivam Feb. 10, 2022, 11:02 a.m. UTC | #5
Hi Kishon,

On Fri, Dec 03, 2021 at 10:36:00AM +0530, Kishon Vijay Abraham I wrote:
> Hi Kunihiko,
> 
> On 01/09/21 10:46 am, Kunihiko Hayashi wrote:
> > The driver using core_init_notifier, e.g. pcie-tegra194.c, runs according
> > to the following sequence:
> > 
> >     probe()
> >         dw_pcie_ep_init()
> > 
> >     bind()
> >         dw_pcie_ep_start()
> >             enable_irq()
> > 
> >     (interrupt occurred)
> >     handler()
> >         [enable controller]
> >         dw_pcie_ep_init_complete()
> >         dw_pcie_ep_init_notify()
> > 
> > After receiving an interrupt from RC, the handler enables the controller
> > and the controller registers can be accessed.
> > So accessing the registers should do in dw_pcie_ep_init_complete().
> > 
> > Currently dw_pcie_ep_init() has functions dw_iatu_detect() and
> > dw_pcie_ep_find_capability() that include accesses to DWC registers.
> > As a result, accessing the registers before enabling the controller,
> > the access will fail.
> > 
> > The function dw_pcie_ep_init() shouldn't have any access to DWC registers
> > if the controller is enabled after calling bind(). This moves access codes
> > to DBI/iATU registers and depending variables from dw_pcie_ep_init() to
> > dw_pcie_ep_init_complete().
> 
> Ideally pci_epc_create() should be the last step by the controller driver before
> handing the control to the core EPC framework. Since after this step the EPC
> framework can start invoking the epc_ops.
> 
> Here more stuff is being added to dw_pcie_ep_init_complete() which is required
> for epc_ops and this could result in aborts for platforms which does not add
> core_init_notifier.
> 

Is there a better way to handle this situation? IMO the existing situation is
messy. Assume that if EP is gonna powered separately by an independent power
rail not tied to host PCIe domain (that's the typical endpoint device usecase),
the EP driver will fail to probe due to PHY link not getting stabilized.

So ultimately the board design needs to take care of an extra logic to power the
EP device after powering the host properly and that's not ideal.

Thanks,
Mani

> Thanks,
> Kishon
> 
> > 
> > Cc: Xiaowei Bao <xiaowei.bao@nxp.com>
> > Cc: Vidya Sagar <vidyas@nvidia.com>
> > Fixes: 6bfc9c3a2c70 ("PCI: designware-ep: Move the function of getting MSI capability forward")
> > Signed-off-by: Kunihiko Hayashi <hayashi.kunihiko@socionext.com>
> > Acked-by: Om Prakash Singh <omp@nvidia.com>
> > Reviewed-by: Vidya Sagar <vidyas@nvidia.com>
> > ---
> >  drivers/pci/controller/dwc/pcie-designware-ep.c | 81 +++++++++++++------------
> >  1 file changed, 41 insertions(+), 40 deletions(-)
> > 
> > diff --git a/drivers/pci/controller/dwc/pcie-designware-ep.c b/drivers/pci/controller/dwc/pcie-designware-ep.c
> > index 998b698..00ce83c 100644
> > --- a/drivers/pci/controller/dwc/pcie-designware-ep.c
> > +++ b/drivers/pci/controller/dwc/pcie-designware-ep.c
> > @@ -636,16 +636,56 @@ static unsigned int dw_pcie_ep_find_ext_capability(struct dw_pcie *pci, int cap)
> >  int dw_pcie_ep_init_complete(struct dw_pcie_ep *ep)
> >  {
> >  	struct dw_pcie *pci = to_dw_pcie_from_ep(ep);
> > +	struct dw_pcie_ep_func *ep_func;
> > +	struct device *dev = pci->dev;
> >  	unsigned int offset;
> >  	unsigned int nbars;
> >  	u8 hdr_type;
> > +	u8 func_no;
> > +	void *addr;
> >  	u32 reg;
> >  	int i;
> >  
> > +	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;
> > +
> > +	for (func_no = 0; func_no < ep->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);
> > +	}
> > +
> >  	hdr_type = dw_pcie_readb_dbi(pci, PCI_HEADER_TYPE) &
> >  		   PCI_HEADER_TYPE_MASK;
> >  	if (hdr_type != PCI_HEADER_TYPE_NORMAL) {
> > -		dev_err(pci->dev,
> > +		dev_err(dev,
> >  			"PCIe controller is not set to EP mode (hdr_type:0x%x)!\n",
> >  			hdr_type);
> >  		return -EIO;
> > @@ -674,8 +714,6 @@ 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);
> > @@ -683,7 +721,6 @@ 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;
> >  
> >  	INIT_LIST_HEAD(&ep->func_list);
> >  
> > @@ -705,8 +742,6 @@ int dw_pcie_ep_init(struct dw_pcie_ep *ep)
> >  		}
> >  	}
> >  
> > -	dw_pcie_iatu_detect(pci);
> > -
> >  	res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "addr_space");
> >  	if (!res)
> >  		return -EINVAL;
> > @@ -714,26 +749,6 @@ 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);
> >  
> > @@ -750,20 +765,6 @@ 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);
> > -
> > -		list_add_tail(&ep_func->list, &ep->func_list);
> > -	}
> > -
> >  	if (ep->ops->ep_init)
> >  		ep->ops->ep_init(ep);
> >  
> >
Manivannan Sadhasivam Feb. 10, 2022, 11:04 a.m. UTC | #6
On Wed, Sep 01, 2021 at 02:16:01PM +0900, Kunihiko Hayashi wrote:
> The driver using core_init_notifier, e.g. pcie-tegra194.c, runs according
> to the following sequence:
> 
>     probe()
>         dw_pcie_ep_init()
> 
>     bind()
>         dw_pcie_ep_start()
>             enable_irq()
> 
>     (interrupt occurred)
>     handler()
>         [enable controller]
>         dw_pcie_ep_init_complete()
>         dw_pcie_ep_init_notify()
> 
> After receiving an interrupt from RC, the handler enables the controller
> and the controller registers can be accessed.
> So accessing the registers should do in dw_pcie_ep_init_complete().
> 
> Currently dw_pcie_ep_init() has functions dw_iatu_detect() and
> dw_pcie_ep_find_capability() that include accesses to DWC registers.
> As a result, accessing the registers before enabling the controller,
> the access will fail.
> 
> The function dw_pcie_ep_init() shouldn't have any access to DWC registers
> if the controller is enabled after calling bind(). This moves access codes
> to DBI/iATU registers and depending variables from dw_pcie_ep_init() to
> dw_pcie_ep_init_complete().
> 
> Cc: Xiaowei Bao <xiaowei.bao@nxp.com>
> Cc: Vidya Sagar <vidyas@nvidia.com>
> Fixes: 6bfc9c3a2c70 ("PCI: designware-ep: Move the function of getting MSI capability forward")
> Signed-off-by: Kunihiko Hayashi <hayashi.kunihiko@socionext.com>
> Acked-by: Om Prakash Singh <omp@nvidia.com>
> Reviewed-by: Vidya Sagar <vidyas@nvidia.com>
> ---
>  drivers/pci/controller/dwc/pcie-designware-ep.c | 81 +++++++++++++------------
>  1 file changed, 41 insertions(+), 40 deletions(-)
> 
> diff --git a/drivers/pci/controller/dwc/pcie-designware-ep.c b/drivers/pci/controller/dwc/pcie-designware-ep.c
> index 998b698..00ce83c 100644
> --- a/drivers/pci/controller/dwc/pcie-designware-ep.c
> +++ b/drivers/pci/controller/dwc/pcie-designware-ep.c

[...]

>  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);
> @@ -683,7 +721,6 @@ 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;
>  
>  	INIT_LIST_HEAD(&ep->func_list);
>  
> @@ -705,8 +742,6 @@ int dw_pcie_ep_init(struct dw_pcie_ep *ep)
>  		}
>  	}
>  
> -	dw_pcie_iatu_detect(pci);
> -
>  	res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "addr_space");
>  	if (!res)
>  		return -EINVAL;
> @@ -714,26 +749,6 @@ 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);
>  
> @@ -750,20 +765,6 @@ 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);
> -
> -		list_add_tail(&ep_func->list, &ep->func_list);
> -	}
> -
>  	if (ep->ops->ep_init)
>  		ep->ops->ep_init(ep);

You also need to move ep_init() as it can have DBI access too.

Thanks,
Mani

>  
> -- 
> 2.7.4
>
diff mbox series

Patch

diff --git a/drivers/pci/controller/dwc/pcie-designware-ep.c b/drivers/pci/controller/dwc/pcie-designware-ep.c
index 998b698..00ce83c 100644
--- a/drivers/pci/controller/dwc/pcie-designware-ep.c
+++ b/drivers/pci/controller/dwc/pcie-designware-ep.c
@@ -636,16 +636,56 @@  static unsigned int dw_pcie_ep_find_ext_capability(struct dw_pcie *pci, int cap)
 int dw_pcie_ep_init_complete(struct dw_pcie_ep *ep)
 {
 	struct dw_pcie *pci = to_dw_pcie_from_ep(ep);
+	struct dw_pcie_ep_func *ep_func;
+	struct device *dev = pci->dev;
 	unsigned int offset;
 	unsigned int nbars;
 	u8 hdr_type;
+	u8 func_no;
+	void *addr;
 	u32 reg;
 	int i;
 
+	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;
+
+	for (func_no = 0; func_no < ep->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);
+	}
+
 	hdr_type = dw_pcie_readb_dbi(pci, PCI_HEADER_TYPE) &
 		   PCI_HEADER_TYPE_MASK;
 	if (hdr_type != PCI_HEADER_TYPE_NORMAL) {
-		dev_err(pci->dev,
+		dev_err(dev,
 			"PCIe controller is not set to EP mode (hdr_type:0x%x)!\n",
 			hdr_type);
 		return -EIO;
@@ -674,8 +714,6 @@  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);
@@ -683,7 +721,6 @@  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;
 
 	INIT_LIST_HEAD(&ep->func_list);
 
@@ -705,8 +742,6 @@  int dw_pcie_ep_init(struct dw_pcie_ep *ep)
 		}
 	}
 
-	dw_pcie_iatu_detect(pci);
-
 	res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "addr_space");
 	if (!res)
 		return -EINVAL;
@@ -714,26 +749,6 @@  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);
 
@@ -750,20 +765,6 @@  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);
-
-		list_add_tail(&ep_func->list, &ep->func_list);
-	}
-
 	if (ep->ops->ep_init)
 		ep->ops->ep_init(ep);