Re: [PATCH v3 1/5] ACPI/APEI: Add apei_hest_parse_aer()
From: LeoLiu-oc
Date: Mon Sep 11 2023 - 17:19:01 EST
在 2023/8/11 7:17, Bjorn Helgaas 写道:
On Tue, Jul 04, 2023 at 08:05:05PM +0800, LeoLiu-oc wrote:
From: leoliu-oc <leoliu-oc@xxxxxxxxxxx>
apei_hest_parse_aer() is used to parse and extract register values
from HEST PCIe AER structures.
Signed-off-by: leoliu-oc <leoliu-oc@xxxxxxxxxxx>
---
drivers/acpi/apei/hest.c | 114 ++++++++++++++++++++++++++++++++++++++-
include/acpi/actbl1.h | 69 ++++++++++++++++++++++++
include/acpi/apei.h | 7 +++
3 files changed, 189 insertions(+), 1 deletion(-)
diff --git a/drivers/acpi/apei/hest.c b/drivers/acpi/apei/hest.c
index 6aef1ee5e1bdb..fb839a5c480ee 100644
--- a/drivers/acpi/apei/hest.c
+++ b/drivers/acpi/apei/hest.c
@@ -26,6 +26,7 @@
#include <acpi/apei.h>
#include <acpi/ghes.h>
+#include <linux/pci.h>
Should go in the list above along with the other <linux/...> includes.
The trend is to alphabetize the list, so I'd put it here:
#include <linux/io.h>
+#include <linux/pci.h>
#include <linux/platform_device.h>
OK, I will make the change according to your suggestion.
LeoLiu-oc
#include "apei-internal.h"
#define HEST_PFX "HEST: "
@@ -86,7 +87,50 @@ static int hest_esrc_len(struct acpi_hest_header *hest_hdr)
return len;
};
-typedef int (*apei_hest_func_t)(struct acpi_hest_header *hest_hdr, void *data);
I think moving this typedef to include/acpi/apei.h should happen in
the patch where it is needed. I *think* that's the "Remove static
from apei_hest_parse()" patch.
Yes，Agree with you. It's more appropriate to move typedef into "Remove
static from apei_hest_parse()" patch.
LeoLiu-oc
+static bool hest_source_is_pcie_aer(struct acpi_hest_header *hest_hdr)
+{
+ switch (hest_hdr->type) {
+ case ACPI_HEST_TYPE_AER_ROOT_PORT:
+ case ACPI_HEST_TYPE_AER_ENDPOINT:
+ case ACPI_HEST_TYPE_AER_BRIDGE:
+ return true;
+ default:
+ return false;
+ }
+}
+
+static bool hest_match_type(struct acpi_hest_header *hest_hdr,
+ struct pci_dev *dev)
Align "struct pci_dev" under "struct acpi_hest_header", as is typical
(see apei-base.c, erst.c, ghes.c)
OK, I will modify in next version.
LeoLiu-oc
+{
+ u16 hest_type = hest_hdr->type;
+ u8 pcie_type = pci_pcie_type(dev);
+
+ if ((hest_type == ACPI_HEST_TYPE_AER_ROOT_PORT &&
+ pcie_type == PCI_EXP_TYPE_ROOT_PORT) ||
+ (hest_type == ACPI_HEST_TYPE_AER_ENDPOINT &&
+ pcie_type == PCI_EXP_TYPE_ENDPOINT) ||
+ (hest_type == ACPI_HEST_TYPE_AER_BRIDGE &&
+ (pcie_type == PCI_EXP_TYPE_PCI_BRIDGE || pcie_type == PCI_EXP_TYPE_PCIE_BRIDGE)))
+ return true;
+ return false;
+}
+
+static bool hest_match_pci_devfn(struct acpi_hest_aer_common *p,
+ struct pci_dev *pci)
Align parameters.
OK.
+{
+ return ACPI_HEST_SEGMENT(p->bus) == pci_domain_nr(pci->bus) &&
+ ACPI_HEST_BUS(p->bus) == pci->bus->number &&
+ p->device == PCI_SLOT(pci->devfn) &&
+ p->function == PCI_FUNC(pci->devfn);
Remove double space before ACPI_HEST_SEGMENT(). Align the other lines
under ACPI_HEST_SEGMENT() and don't bother indenting the
pci->bus->number, PCI_SLOT, PCI_FUNC, since they don't line up anyway,
i.e.,
return ACPI_HEST_SEGMENT(p->bus) == pci_domain_nr(pci->bus) &&
ACPI_HEST_BUS(p->bus) == pci->bus->number &&
p->device == PCI_SLOT(pci->devfn) &&
p->function == PCI_FUNC(pci->devfn);
The only allowable variable names for a pci_dev pointer are "dev" or
"pdev" :) In this context, I think "dev" is fine (as you used above
for hest_match_type()).
I'm going to unify the variable names for a "pci_dev" pointer.
LeoLiu-oc
+}
+
+static bool hest_match_pci(struct acpi_hest_header *hest_hdr,
+ struct acpi_hest_aer_common *p, struct pci_dev *pci)
+{
+ if (hest_match_type(hest_hdr, pci))
+ return hest_match_pci_devfn(p, pci);
+ return false;
+}
static int apei_hest_parse(apei_hest_func_t func, void *data)
{
@@ -124,6 +168,74 @@ static int apei_hest_parse(apei_hest_func_t func, void *data)
return 0;
}
+/*
+ * apei_hest_parse_aer - Find the AER structure in the HEST and
+ * match it with the PCIe device.
+ *
+ * @hest_hdr: To save the ACPI AER error source in HEST
+ *
+ * Return 1 if the PCIe dev matched with the ACPI AER error source in
+ * HEST, else return 0.
+ */
+int apei_hest_parse_aer(struct acpi_hest_header *hest_hdr, void *data)
Follow the naming pattern of all the other parse functions:
hest_parse_ghes_count
hest_parse_ghes
hest_parse_cmc
E.g., "hest_parse_pcie_aer" or "hest_parse_aer".
OK.
+{
+ struct acpi_hest_parse_aer_info *info = data;
+ struct acpi_hest_aer_endpoint *acpi_hest_aer_endpoint = NULL;
+ struct acpi_hest_aer_root_port *acpi_hest_aer_root_port = NULL;
+ struct acpi_hest_aer_for_bridge *acpi_hest_aer_for_bridge = NULL;
Unnecessary initializations since you always set these below before
using them. But maybe there's a way to do this without these pointers
(see below).
OK.
+ if (!hest_source_is_pcie_aer(hest_hdr))
+ return 0;
+
+ switch (hest_hdr->type) {
+ case ACPI_HEST_TYPE_AER_ROOT_PORT:
+ acpi_hest_aer_root_port = (struct acpi_hest_aer_root_port *)(hest_hdr + 1);
+ if (acpi_hest_aer_root_port->flags & ACPI_HEST_GLOBAL) {
+ if (hest_match_type(hest_hdr, info->pci_dev)) {
+ info->acpi_hest_aer_root_port = acpi_hest_aer_root_port;
+ info->hest_matched_with_dev = 1;
+ }
+ } else if (hest_match_pci(hest_hdr,
+ (struct acpi_hest_aer_common *)acpi_hest_aer_root_port,
+ info->pci_dev)) {
+ info->acpi_hest_aer_root_port = acpi_hest_aer_root_port;
+ info->hest_matched_with_dev = 1;
+ }
+ break;
Could this be refactored along the lines of this?
static bool matches(struct acpi_hest_header *hest_hdr, struct pci_dev *dev)
{
u16 hest_type = hest_hdr->type;
u8 pcie_type = pci_pcie_type(dev);
struct acpi_hest_aer_common *common =
(struct acpi_hest_aer_common *) (hest_hdr + 1);
switch (hest_type) {
case ACPI_HEST_TYPE_AER_ROOT_PORT:
if (pcie_type != PCI_EXP_TYPE_ROOT_PORT)
return false;
break;
case ACPI_HEST_TYPE_AER_ENDPOINT:
if (pcie_type != PCI_EXP_TYPE_ENDPOINT)
return false;
break;
case ACPI_HEST_TYPE_AER_BRIDGE:
if (pcie_type != PCI_EXP_TYPE_BRIDGE &&
pcie_type != PCI_EXP_TYPE_PCIE_BRIDGE)
return false;
break;
default:
return false;
}
if (common->flags & ACPI_HEST_GLOBAL)
return true;
if (hest_match_pci_devfn(common, dev))
return true;
return false;
}
int apei_hest_parse_aer(...)
{
if (!matches(hest_hdr, info->pci_dev))
return 0;
switch (hest_type) {
case ACPI_HEST_TYPE_AER_ROOT_PORT:
info->acpi_hest_aer_root_port =
(struct acpi_hest_aer_root_port *)(hest_hdr + 1);
info->hest_matched_with_dev = 1;
return 1;
case ACPI_HEST_TYPE_AER_ENDPOINT:
...
}
return 0;
}
Yes, your advice makes the code clearer.
LeoLiu-oc
+ case ACPI_HEST_TYPE_AER_ENDPOINT:
+ acpi_hest_aer_endpoint = (struct acpi_hest_aer_endpoint *)(hest_hdr + 1);
+ if (acpi_hest_aer_endpoint->flags & ACPI_HEST_GLOBAL) {
+ if (hest_match_type(hest_hdr, info->pci_dev)) {
+ info->acpi_hest_aer_endpoint = acpi_hest_aer_endpoint;
+ info->hest_matched_with_dev = 1;
+ }
+ } else if (hest_match_pci(hest_hdr,
+ (struct acpi_hest_aer_common *)acpi_hest_aer_endpoint,
+ info->pci_dev)) {
+ info->acpi_hest_aer_endpoint = acpi_hest_aer_endpoint;
+ info->hest_matched_with_dev = 1;
+ }
+ break;
+ case ACPI_HEST_TYPE_AER_BRIDGE:
+ acpi_hest_aer_for_bridge = (struct acpi_hest_aer_for_bridge *)(hest_hdr + 1);
+ if (acpi_hest_aer_for_bridge->flags & ACPI_HEST_GLOBAL) {
+ if (hest_match_type(hest_hdr, info->pci_dev)) {
+ info->acpi_hest_aer_for_bridge = acpi_hest_aer_for_bridge;
+ info->hest_matched_with_dev = 1;
+ }
+ } else if (hest_match_pci(hest_hdr,
+ (struct acpi_hest_aer_common *)acpi_hest_aer_for_bridge,
+ info->pci_dev)) {
+ info->acpi_hest_aer_for_bridge = acpi_hest_aer_for_bridge;
+ info->hest_matched_with_dev = 1;
+ }
+ break;
+ default:
+ break;
+ }
+ return info->hest_matched_with_dev;
+}
+
/*
* Check if firmware advertises firmware first mode. We need FF bit to be set
* along with a set of MC banks which work in FF mode.
diff --git a/include/acpi/actbl1.h b/include/acpi/actbl1.h
index 58b0490a2ad12..31abe06c6e792 100644
--- a/include/acpi/actbl1.h
+++ b/include/acpi/actbl1.h
@@ -1583,6 +1583,75 @@ struct acpi_hest_aer_bridge {
u32 advanced_capabilities2;
};
+struct acpi_hest_parse_aer_info {
+ struct pci_dev *pci_dev;
+ int hest_matched_with_dev;
+ struct acpi_hest_aer_endpoint *acpi_hest_aer_endpoint;
+ struct acpi_hest_aer_root_port *acpi_hest_aer_root_port;
+ struct acpi_hest_aer_for_bridge *acpi_hest_aer_for_bridge;
+};
+
+/* HEST Sub-structure for PCIe Root Port Structure (6) */
+
+struct acpi_hest_aer_root_port {
How are these different from the existing acpi_hest_aer_root,
acpi_hest_aer, and acpi_hest_aer_bridge?
I see that the existing acpi_hest_aer_root includes a struct
acpi_hest_header and the new acpi_hest_aer_root_port does not, but
that doesn't seem like enough reason to duplicate all the rest of the
structure.
Sorry if you've answered this before. If we really need both
acpi_hest_aer_root and acpi_hest_aer_root_port, we should put a hint
about the reason in the commit log and maybe even in a code comment
here.
Agree with you. I will remove that part of the definition and use the
existing definition in the next version.
LeoLiu-oc
+ u16 reserved1;
+ u8 flags;
+ u8 enabled;
+ u32 records_to_preallocate;
+ u32 max_sections_per_record;
+ u32 bus;
+ u16 device;
+ u16 function;
+ u16 device_control;
+ u16 reserved2;
+ u32 uncorrectable_mask;
+ u32 uncorrectable_severity;
+ u32 correctable_mask;
+ u32 advanced_capabilities;
+ u32 root_error_command;
+};
+/* HEST Sub-structure for PCIe/PCI-X Bridge Structure (8) */
+
+struct acpi_hest_aer_for_bridge {
If this struct is needed, s/for_// (the other structs don't include
"for_" and there's no need here).
OK.
Best Regards.
LeoLiu-oc