Re: [PATCH v2] PCI: Use pci_pcie_find_root_port() to get root port

[Yicong Yang <yangyicong@hisilicon.com>]

On 2020/5/7 3:48, Bjorn Helgaas wrote:
> On Wed, May 06, 2020 at 08:42:56PM +0800, Yicong Yang wrote:
>> Previously we use pcie_find_root_port() to get root port from a pcie
>> device, use pci_find_pcie_root_port() to get root port from a pci
>> device, which increase the complexity.
>>
>> Unify the two functions and use pci_pcie_find_root_port() to get root
>> port from both pci device and pcie device. Then there is not need to
>> distinguish the type of the device.
>>
>> Signed-off-by: Yicong Yang <yangyicong@hisilicon.com>
>> Acked-by: Kalle Valo <kvalo@codeaurora.org> // for wireless
>> Acked-by: Mika Westerberg <mika.westerberg@linux.intel.com> // for thunderbolt
> When I first looked at this, I got distracted because I think many of
> the calls of these functions should go away.  That's out of scope
> right now, but I'll just mention them for the archives:
>
>   - iwl_trans_pcie_dump_regs() dumps AER registers, which should be
>     done by the PCI core, not by the driver.
>
>   - pcie_root_rcb_set() looks up the Root Port's RCB setting, which
>     could easily be inherited.
>
>   - aer_inject().  I'm not convinced anybody uses this, and we have
>     the ACPI EINJ path, which should test a lot more of the path.
>
>   - acpi_pci_bridge_d3() looks up the "HotPlugSupportInD3", which
>     probably could also be inherited.
>
>   - pci_configure_relaxed_ordering() looks up the Root Port's Relaxed
>     Ordering support, which could be inherited.
>
>> ---
>> Change since v1:
>> - Add Mika's Ack for thunderbolt part
>> - Add description for pci_pcie_find_root_port()
>>
>>  drivers/net/wireless/intel/iwlwifi/pcie/trans.c |  2 +-
>>  drivers/pci/pci-acpi.c                          |  4 ++--
>>  drivers/pci/pci.c                               | 24 ------------------------
>>  drivers/pci/pcie/aer_inject.c                   |  2 +-
>>  drivers/pci/probe.c                             |  2 +-
>>  drivers/pci/quirks.c                            |  2 +-
>>  drivers/thunderbolt/switch.c                    |  4 ++--
>>  include/linux/pci.h                             | 17 ++++++++++-------
>>  8 files changed, 18 insertions(+), 39 deletions(-)
>>
>> diff --git a/drivers/net/wireless/intel/iwlwifi/pcie/trans.c b/drivers/net/wireless/intel/iwlwifi/pcie/trans.c
>> index 38d8fe2..556cb8c 100644
>> --- a/drivers/net/wireless/intel/iwlwifi/pcie/trans.c
>> +++ b/drivers/net/wireless/intel/iwlwifi/pcie/trans.c
>> @@ -158,7 +158,7 @@ void iwl_trans_pcie_dump_regs(struct iwl_trans *trans)
>>
>>  	/* Print root port AER registers */
>>  	pos = 0;
>> -	pdev = pcie_find_root_port(pdev);
>> +	pdev = pci_pcie_find_root_port(pdev);
>>  	if (pdev)
>>  		pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_ERR);
>>  	if (pos) {
>> diff --git a/drivers/pci/pci-acpi.c b/drivers/pci/pci-acpi.c
>> index 0c02d50..9316533 100644
>> --- a/drivers/pci/pci-acpi.c
>> +++ b/drivers/pci/pci-acpi.c
>> @@ -246,7 +246,7 @@ static acpi_status decode_type1_hpx_record(union acpi_object *record,
>>
>>  static bool pcie_root_rcb_set(struct pci_dev *dev)
>>  {
>> -	struct pci_dev *rp = pcie_find_root_port(dev);
>> +	struct pci_dev *rp = pci_pcie_find_root_port(dev);
>>  	u16 lnkctl;
>>
>>  	if (!rp)
>> @@ -948,7 +948,7 @@ static bool acpi_pci_bridge_d3(struct pci_dev *dev)
>>  	 * Look for a special _DSD property for the root port and if it
>>  	 * is set we know the hierarchy behind it supports D3 just fine.
>>  	 */
>> -	root = pci_find_pcie_root_port(dev);
>> +	root = pci_pcie_find_root_port(dev);
>>  	if (!root)
>>  		return false;
>>
>> diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
>> index d828ca8..fc5e7b6 100644
>> --- a/drivers/pci/pci.c
>> +++ b/drivers/pci/pci.c
>> @@ -695,30 +695,6 @@ struct resource *pci_find_resource(struct pci_dev *dev, struct resource *res)
>>  EXPORT_SYMBOL(pci_find_resource);
>>
>>  /**
>> - * pci_find_pcie_root_port - return PCIe Root Port
>> - * @dev: PCI device to query
>> - *
>> - * Traverse up the parent chain and return the PCIe Root Port PCI Device
>> - * for a given PCI Device.
>> - */
>> -struct pci_dev *pci_find_pcie_root_port(struct pci_dev *dev)
>> -{
>> -	struct pci_dev *bridge, *highest_pcie_bridge = dev;
>> -
>> -	bridge = pci_upstream_bridge(dev);
>> -	while (bridge && pci_is_pcie(bridge)) {
>> -		highest_pcie_bridge = bridge;
>> -		bridge = pci_upstream_bridge(bridge);
>> -	}
>> -
>> -	if (pci_pcie_type(highest_pcie_bridge) != PCI_EXP_TYPE_ROOT_PORT)
>> -		return NULL;
>> -
>> -	return highest_pcie_bridge;
>> -}
>> -EXPORT_SYMBOL(pci_find_pcie_root_port);
>> -
>> -/**
>>   * pci_wait_for_pending - wait for @mask bit(s) to clear in status word @pos
>>   * @dev: the PCI device to operate on
>>   * @pos: config space offset of status word
>> diff --git a/drivers/pci/pcie/aer_inject.c b/drivers/pci/pcie/aer_inject.c
>> index 6988fe7..c3bfc1b5 100644
>> --- a/drivers/pci/pcie/aer_inject.c
>> +++ b/drivers/pci/pcie/aer_inject.c
>> @@ -332,7 +332,7 @@ static int aer_inject(struct aer_error_inj *einj)
>>  	dev = pci_get_domain_bus_and_slot(einj->domain, einj->bus, devfn);
>>  	if (!dev)
>>  		return -ENODEV;
>> -	rpdev = pcie_find_root_port(dev);
>> +	rpdev = pci_pcie_find_root_port(dev);
>>  	if (!rpdev) {
>>  		pci_err(dev, "Root port not found\n");
>>  		ret = -ENODEV;
>> diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c
>> index 512cb43..50f7733 100644
>> --- a/drivers/pci/probe.c
>> +++ b/drivers/pci/probe.c
>> @@ -2015,7 +2015,7 @@ static void pci_configure_relaxed_ordering(struct pci_dev *dev)
>>  	 * For now, we only deal with Relaxed Ordering issues with Root
>>  	 * Ports. Peer-to-Peer DMA is another can of worms.
>>  	 */
>> -	root = pci_find_pcie_root_port(dev);
>> +	root = pci_pcie_find_root_port(dev);
>>  	if (!root)
>>  		return;
>>
>> diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c
>> index 29f473e..ac62675 100644
>> --- a/drivers/pci/quirks.c
>> +++ b/drivers/pci/quirks.c
>> @@ -4253,7 +4253,7 @@ DECLARE_PCI_FIXUP_CLASS_EARLY(PCI_VENDOR_ID_AMD, 0x1a02, PCI_CLASS_NOT_DEFINED,
>>   */
>>  static void quirk_disable_root_port_attributes(struct pci_dev *pdev)
>>  {
>> -	struct pci_dev *root_port = pci_find_pcie_root_port(pdev);
>> +	struct pci_dev *root_port = pci_pcie_find_root_port(pdev);
>>
>>  	if (!root_port) {
>>  		pci_warn(pdev, "PCIe Completion erratum may cause device errors\n");
>> diff --git a/drivers/thunderbolt/switch.c b/drivers/thunderbolt/switch.c
>> index a2ce990..90844c1 100644
>> --- a/drivers/thunderbolt/switch.c
>> +++ b/drivers/thunderbolt/switch.c
>> @@ -263,7 +263,7 @@ static void nvm_authenticate_start_dma_port(struct tb_switch *sw)
>>  	 * itself. To be on the safe side keep the root port in D0 during
>>  	 * the whole upgrade process.
>>  	 */
>> -	root_port = pci_find_pcie_root_port(sw->tb->nhi->pdev);
>> +	root_port = pci_pcie_find_root_port(sw->tb->nhi->pdev);
>>  	if (root_port)
>>  		pm_runtime_get_noresume(&root_port->dev);
>>  }
>> @@ -272,7 +272,7 @@ static void nvm_authenticate_complete_dma_port(struct tb_switch *sw)
>>  {
>>  	struct pci_dev *root_port;
>>
>> -	root_port = pci_find_pcie_root_port(sw->tb->nhi->pdev);
>> +	root_port = pci_pcie_find_root_port(sw->tb->nhi->pdev);
>>  	if (root_port)
>>  		pm_runtime_put(&root_port->dev);
>>  }
>> diff --git a/include/linux/pci.h b/include/linux/pci.h
>> index 3840a54..51881a1 100644
>> --- a/include/linux/pci.h
>> +++ b/include/linux/pci.h
>> @@ -1011,7 +1011,6 @@ void pci_bus_add_device(struct pci_dev *dev);
>>  void pci_read_bridge_bases(struct pci_bus *child);
>>  struct resource *pci_find_parent_resource(const struct pci_dev *dev,
>>  					  struct resource *res);
>> -struct pci_dev *pci_find_pcie_root_port(struct pci_dev *dev);
>>  u8 pci_swizzle_interrupt_pin(const struct pci_dev *dev, u8 pin);
>>  int pci_get_interrupt_pin(struct pci_dev *dev, struct pci_dev **bridge);
>>  u8 pci_common_swizzle(struct pci_dev *dev, u8 *pinp);
>> @@ -2124,15 +2123,19 @@ static inline int pci_pcie_type(const struct pci_dev *dev)
>>  	return (pcie_caps_reg(dev) & PCI_EXP_FLAGS_TYPE) >> 4;
>>  }
>>
>> -static inline struct pci_dev *pcie_find_root_port(struct pci_dev *dev)
>> +/**
>> + * pci_pcie_find_root_port - Get the PCIe root port device
>> + * @dev: PCI device
>> + *
>> + * Traverse up the parent chain and return the PCIe Root Port PCI Device
>> + * for a given PCI/PCIe Device.
>> + */
>> +static inline struct pci_dev *pci_pcie_find_root_port(struct pci_dev *dev)
> The point of this is to find a Root Port regardless of whether "dev"
> is PCI or PCIe.  So I think pcie_find_root_port() is sufficient; we
> don't need both "pci" and "pcie" in the name.

Okay.


>
>>  {
>> -	while (1) {
>> -		if (!pci_is_pcie(dev))
>> -			break;
>> +	dev = pci_physfn(dev);
>> +	while (dev) {
>>  		if (pci_pcie_type(dev) == PCI_EXP_TYPE_ROOT_PORT)
>>  			return dev;
>> -		if (!dev->bus->self)
>> -			break;
>>  		dev = dev->bus->self;
> Why not use pci_upstream_bridge() here?

We'll judge whether the device is virtual or not every time we call
pci_upstream_bridge(). I think it's unnecessary and we only need to get
the physical function at the beginning of the traverse. It's okay to
use pci_upstream_bridge() here if you suggest so.

Regards,
Yicong


>
>>  	}
>>  	return NULL;
>> --
>> 2.8.1
>>
> .
>