Re: [PATCH v3 4/5] ACPI/PCI: Add pci_acpi_program_hest_aer_params()

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 





在 2023/8/11 7:30, Bjorn Helgaas 写道:
On Tue, Jul 04, 2023 at 08:05:44PM +0800, LeoLiu-oc wrote:
From: leoliu-oc <leoliu-oc@xxxxxxxxxxx>

The extracted register values from HEST PCI Express AER structures are
written to AER Capabilities.

In the subject, the prevailing style for this file is
(see "git log --oneline drivers/pci/pci-acpi.c"):

   PCI/ACPI: ...

And I'd like the subject to tell users why they might want this patch.
It's obvious from the patch that this adds a function.  What's *not*
obvious is *why* we want this new function.  So the commit log should
tell us what the benefit is, and the subject line should be one-line
summary of that benefit.

This patch adds a function but no caller.  The next patch is one-liner
that adds the caller.  I think these two should be squashed so it's
easier to review (and easier to explain the benefit of *this* patch :))


Ok, I will merge this patch with 5/5 in the next version.

Signed-off-by: leoliu-oc <leoliu-oc@xxxxxxxxxxx>
---
  drivers/pci/pci-acpi.c | 92 ++++++++++++++++++++++++++++++++++++++++++
  drivers/pci/pci.h      |  5 +++
  2 files changed, 97 insertions(+)

diff --git a/drivers/pci/pci-acpi.c b/drivers/pci/pci-acpi.c
index a05350a4e49cb..cff54410e2427 100644
--- a/drivers/pci/pci-acpi.c
+++ b/drivers/pci/pci-acpi.c
@@ -18,6 +18,7 @@
  #include <linux/pm_runtime.h>
  #include <linux/pm_qos.h>
  #include <linux/rwsem.h>
+#include <acpi/apei.h>
  #include "pci.h"
/*
@@ -783,6 +784,97 @@ int pci_acpi_program_hp_params(struct pci_dev *dev)
  	return -ENODEV;
  }
+/*
+ * program_aer_structure_to_aer_registers - Write the AER structure to
+ * the corresponding dev's AER registers.
+ *
+ * @info - the AER structure information
+ *

Remove the spurious blank comment line.


OK.

+ */
+static void program_aer_structure_to_aer_registers(struct acpi_hest_parse_aer_info info)
+{
+	u32 uncorrectable_mask;
+	u32 uncorrectable_severity;
+	u32 correctable_mask;
+	u32 advanced_capabilities;
+	u32 root_error_command;
+	u32 uncorrectable_mask2;
+	u32 uncorrectable_severity2;
+	u32 advanced_capabilities2;
+	int port_type;
+	int pos;
+	struct pci_dev *dev;

Order these declarations in order of use.


OK.

+	dev = info.pci_dev;
+	port_type = pci_pcie_type(dev);
+
+	pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
+	if (!pos)
+		return;
+
+	if (port_type == PCI_EXP_TYPE_ROOT_PORT) {
+		uncorrectable_mask = info.acpi_hest_aer_root_port->uncorrectable_mask;
+		uncorrectable_severity = info.acpi_hest_aer_root_port->uncorrectable_severity;
+		correctable_mask = info.acpi_hest_aer_root_port->correctable_mask;
+		advanced_capabilities = info.acpi_hest_aer_root_port->advanced_capabilities;
+		root_error_command = info.acpi_hest_aer_root_port->root_error_command;

Except for this new code, this file fits in 80 columns, so I'd like
the new code to match.


OK.

+
+		pci_write_config_dword(dev, pos + PCI_ERR_UNCOR_MASK, uncorrectable_mask);

I'm not sure we need to copy everything into local variables.  Maybe
this could be split into three helper functions, which would save a
level of indent and a level of struct traversal (e.g., "rp->" instead
of "info.acpi_hest_aer_root_port->".

   pci_write_config_dword(dev, pos + PCI_ERR_UNCOR_MASK, rp->uncorrectable_mask);

or

   pci_write_config_dword(dev, pos + PCI_ERR_UNCOR_MASK,
                          rp->uncorrectable_mask);

If you have to define a new struct acpi_hest_aer_root_port, you could
make the member names shorter.  But hopefully you *don't* have to do
that, so maybe we're stuck with the long existing member names in
acpi_hest_aer_common.
>> +int pci_acpi_program_hest_aer_params(struct pci_dev *dev)
+{
+	struct acpi_hest_parse_aer_info info = {
+		.pci_dev	= dev,
+		.hest_matched_with_dev	= 0,
+		.acpi_hest_aer_endpoint = NULL,
+		.acpi_hest_aer_root_port = NULL,
+		.acpi_hest_aer_for_bridge = NULL,

Drop the tab from the .pci_dev initialization since the other members
aren't lined up anyway.  I think you can drop the other
initializations completely since they will be initialized to 0 or NULL
pointers by default.


Thanks for your guidance, I will make changes to the code where it fits and does not conform to the specification.

Best Regards.
LeoLiu-oc

+	};
+
+	if (!pci_is_pcie(dev))
+		return -ENODEV;
+
+	apei_hest_parse(apei_hest_parse_aer, &info);
+	if (info.hest_matched_with_dev == 1)
+		program_aer_structure_to_aer_registers(info);
+	else
+		return -ENODEV;
+	return 0;
+}
+
  /**
   * pciehp_is_native - Check whether a hotplug port is handled by the OS
   * @bridge: Hotplug port to check
diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h
index a4c3974340576..37aa4a33eeed2 100644
--- a/drivers/pci/pci.h
+++ b/drivers/pci/pci.h
@@ -713,6 +713,7 @@ void acpi_pci_refresh_power_state(struct pci_dev *dev);
  int acpi_pci_wakeup(struct pci_dev *dev, bool enable);
  bool acpi_pci_need_resume(struct pci_dev *dev);
  pci_power_t acpi_pci_choose_state(struct pci_dev *pdev);
+int pci_acpi_program_hest_aer_params(struct pci_dev *dev);
  #else
  static inline int pci_dev_acpi_reset(struct pci_dev *dev, bool probe)
  {
@@ -752,6 +753,10 @@ static inline pci_power_t acpi_pci_choose_state(struct pci_dev *pdev)
  {
  	return PCI_POWER_ERROR;
  }
+static inline int pci_acpi_program_hest_aer_params(struct pci_dev *dev)
+{
+	return -ENODEV;
+}
  #endif
#ifdef CONFIG_PCIEASPM
--
2.34.1




[Index of Archives]     [DMA Engine]     [Linux Coverity]     [Linux USB]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]     [Greybus]

  Powered by Linux