Message ID | 20240510112740.667445-3-fabio.m.de.francesco@linux.intel.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
Series | Make ELOG log and trace consistently with GHES | expand |
Fabio M. De Francesco wrote: > Currently, extlog_print() (ELOG) only reports CPER PCIe section (UEFI > v2.10, Appendix N.2.7) to the kernel log via print_extlog_rcd(). Instead, > the similar ghes_do_proc() (GHES) prints to kernel log and calls > pci_print_aer() to report via the ftrace infrastructure. > > Add support to report the CPER PCIe Error section also via the ftrace > infrastructure by calling pci_print_aer() to make ELOG act consistently > with GHES. > > Cc: Dan Williams <dan.j.williams@intel.com> > Signed-off-by: Fabio M. De Francesco <fabio.m.de.francesco@linux.intel.com> > --- > drivers/acpi/acpi_extlog.c | 35 +++++++++++++++++++++++++++++++++++ > 1 file changed, 35 insertions(+) > > diff --git a/drivers/acpi/acpi_extlog.c b/drivers/acpi/acpi_extlog.c > index 4e62d7235d33..fb7b0c73f86a 100644 > --- a/drivers/acpi/acpi_extlog.c > +++ b/drivers/acpi/acpi_extlog.c > @@ -131,6 +131,36 @@ static int print_extlog_rcd(const char *pfx, > return 1; > } > > +static void extlog_print_pcie(struct cper_sec_pcie *pcie_err, > + int severity) > +{ > +#ifdef CONFIG_ACPI_APEI_PCIEAER Whenever possible do not use ifdef in C files, see "21) Conditional Compilation" in coding style. I suspect the issue here is that pci_print_aer() is not always defined, but the solution there is something like: diff --git a/include/linux/aer.h b/include/linux/aer.h index 4b97f38f3fcf..6ff54197480d 100644 --- a/include/linux/aer.h +++ b/include/linux/aer.h @@ -42,16 +42,20 @@ int pcie_read_tlp_log(struct pci_dev *dev, int where, struct pcie_tlp_log *log); #if defined(CONFIG_PCIEAER) int pci_aer_clear_nonfatal_status(struct pci_dev *dev); int pcie_aer_is_native(struct pci_dev *dev); +void pci_print_aer(struct pci_dev *dev, int aer_severity, + struct aer_capability_regs *aer); #else static inline int pci_aer_clear_nonfatal_status(struct pci_dev *dev) { return -EINVAL; } static inline int pcie_aer_is_native(struct pci_dev *dev) { return 0; } +static inline void pci_print_aer(struct pci_dev *dev, int aer_severity, + struct aer_capability_regs *aer) +{ +} #endif -void pci_print_aer(struct pci_dev *dev, int aer_severity, - struct aer_capability_regs *aer); int cper_severity_to_aer(int cper_severity); void aer_recover_queue(int domain, unsigned int bus, unsigned int devfn, int severity, struct aer_capability_regs *aer_regs); > + struct aer_capability_regs *aer; > + struct pci_dev *pdev; > + unsigned int devfn; > + unsigned int bus; > + int aer_severity; > + int domain; > + > + if (pcie_err->validation_bits & CPER_PCIE_VALID_DEVICE_ID && > + pcie_err->validation_bits & CPER_PCIE_VALID_AER_INFO) { > + aer_severity = cper_severity_to_aer(severity); > + aer = (struct aer_capability_regs *)pcie_err->aer_info; > + domain = pcie_err->device_id.segment; > + bus = pcie_err->device_id.bus; > + devfn = PCI_DEVFN(pcie_err->device_id.device, > + pcie_err->device_id.function); > + pdev = pci_get_domain_bus_and_slot(domain, bus, devfn); This pci_get_domain_bus_and_slot() takes a refernce on @pdev that reference needs to be dropped after operations on @pdev are complete. > + if (!pdev) { > + pr_err("no pci_dev for %04x:%02x:%02x.%x\n", > + domain, bus, PCI_SLOT(devfn), PCI_FUNC(devfn)); I am not sure this error message is useful. I see this was copied from aer_recover_work_func(). I wonder if a better approach is to just teach pci_print_aer() to be tolerant of a NULL @pdev argument. Most of what pci_print_aer() is just a device name which can be recreated as: sprintf(name, "%04x:%02x:%02x.%d", domain, bus, PCI_SLOT(dev->devfn), PCI_FUNC(dev->devfn)); For now, I would just return if a device cannot be found, and skip the print. > + } > +#endif > +} > + > static int extlog_print(struct notifier_block *nb, unsigned long val, > void *data) > { > @@ -182,6 +212,10 @@ static int extlog_print(struct notifier_block *nb, unsigned long val, > if (gdata->error_data_length >= sizeof(*mem)) > trace_extlog_mem_event(mem, err_seq, fru_id, fru_text, > (u8)gdata->error_severity); > + } else if (guid_equal(sec_type, &CPER_SEC_PCIE)) { > + struct cper_sec_pcie *pcie_err = acpi_hest_get_payload(gdata); > + > + extlog_print_pcie(pcie_err, gdata->error_severity); > } else { > void *err = acpi_hest_get_payload(gdata); > > @@ -331,3 +365,4 @@ module_exit(extlog_exit); > MODULE_AUTHOR("Chen, Gong <gong.chen@intel.com>"); > MODULE_DESCRIPTION("Extended MCA Error Log Driver"); > MODULE_LICENSE("GPL"); > +MODULE_IMPORT_NS(CXL); This is awkward now that pci_print_aer() is consumed by a module that is not "CXL", so perhaps we should either drop the namespacing of the pci_print_aer() symbol or put it in its own namespace. I think the former is the way to go.
diff --git a/drivers/acpi/acpi_extlog.c b/drivers/acpi/acpi_extlog.c index 4e62d7235d33..fb7b0c73f86a 100644 --- a/drivers/acpi/acpi_extlog.c +++ b/drivers/acpi/acpi_extlog.c @@ -131,6 +131,36 @@ static int print_extlog_rcd(const char *pfx, return 1; } +static void extlog_print_pcie(struct cper_sec_pcie *pcie_err, + int severity) +{ +#ifdef CONFIG_ACPI_APEI_PCIEAER + struct aer_capability_regs *aer; + struct pci_dev *pdev; + unsigned int devfn; + unsigned int bus; + int aer_severity; + int domain; + + if (pcie_err->validation_bits & CPER_PCIE_VALID_DEVICE_ID && + pcie_err->validation_bits & CPER_PCIE_VALID_AER_INFO) { + aer_severity = cper_severity_to_aer(severity); + aer = (struct aer_capability_regs *)pcie_err->aer_info; + domain = pcie_err->device_id.segment; + bus = pcie_err->device_id.bus; + devfn = PCI_DEVFN(pcie_err->device_id.device, + pcie_err->device_id.function); + pdev = pci_get_domain_bus_and_slot(domain, bus, devfn); + if (!pdev) { + pr_err("no pci_dev for %04x:%02x:%02x.%x\n", + domain, bus, PCI_SLOT(devfn), PCI_FUNC(devfn)); + return; + } + pci_print_aer(pdev, aer_severity, aer); + } +#endif +} + static int extlog_print(struct notifier_block *nb, unsigned long val, void *data) { @@ -182,6 +212,10 @@ static int extlog_print(struct notifier_block *nb, unsigned long val, if (gdata->error_data_length >= sizeof(*mem)) trace_extlog_mem_event(mem, err_seq, fru_id, fru_text, (u8)gdata->error_severity); + } else if (guid_equal(sec_type, &CPER_SEC_PCIE)) { + struct cper_sec_pcie *pcie_err = acpi_hest_get_payload(gdata); + + extlog_print_pcie(pcie_err, gdata->error_severity); } else { void *err = acpi_hest_get_payload(gdata); @@ -331,3 +365,4 @@ module_exit(extlog_exit); MODULE_AUTHOR("Chen, Gong <gong.chen@intel.com>"); MODULE_DESCRIPTION("Extended MCA Error Log Driver"); MODULE_LICENSE("GPL"); +MODULE_IMPORT_NS(CXL);
Currently, extlog_print() (ELOG) only reports CPER PCIe section (UEFI v2.10, Appendix N.2.7) to the kernel log via print_extlog_rcd(). Instead, the similar ghes_do_proc() (GHES) prints to kernel log and calls pci_print_aer() to report via the ftrace infrastructure. Add support to report the CPER PCIe Error section also via the ftrace infrastructure by calling pci_print_aer() to make ELOG act consistently with GHES. Cc: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Fabio M. De Francesco <fabio.m.de.francesco@linux.intel.com> --- drivers/acpi/acpi_extlog.c | 35 +++++++++++++++++++++++++++++++++++ 1 file changed, 35 insertions(+)