Re: [PATCH v5 1/5] PCI/AER: Define and allocate aer_stats structure for AER capable devices

[Rajat Jain]

On Thu, Jun 21, 2018 at 6:17 AM, Bjorn Helgaas <helgaas@xxxxxxxxxx> wrote:
> On Wed, Jun 20, 2018 at 04:41:43PM -0700, Rajat Jain wrote:
>> Define a structure to hold the AER statistics. There are 2 groups
>> of statistics: dev_* counters that are to be collected for all AER
>> capable devices and rootport_* counters that are collected for all
>> (AER capable) rootports only. Allocate and free this structure when
>> device is added or released (thus counters survive the lifetime of the
>> device).
>>
>> Signed-off-by: Rajat Jain <rajatja@xxxxxxxxxx>
>> ---
>> v5: Same as v4
>> v4: Same as v3
>> v3: Merge everything in aer.c
>>
>>  drivers/pci/pcie/aer.c | 60 ++++++++++++++++++++++++++++++++++++++++++
>>  drivers/pci/probe.c    |  1 +
>>  include/linux/pci.h    |  3 +++
>>  3 files changed, 64 insertions(+)
>>
>> diff --git a/drivers/pci/pcie/aer.c b/drivers/pci/pcie/aer.c
>> index a2e88386af28..f9fa994b6c33 100644
>> --- a/drivers/pci/pcie/aer.c
>> +++ b/drivers/pci/pcie/aer.c
>> @@ -33,6 +33,9 @@
>>  #define AER_ERROR_SOURCES_MAX                100
>>  #define AER_MAX_MULTI_ERR_DEVICES    5       /* Not likely to have more */
>>
>> +#define AER_MAX_TYPEOF_CORRECTABLE_ERRS 16   /* as per PCI_ERR_COR_STATUS */
>> +#define AER_MAX_TYPEOF_UNCORRECTABLE_ERRS 26 /* as per PCI_ERR_UNCOR_STATUS*/
>> +
>>  struct aer_err_info {
>>       struct pci_dev *dev[AER_MAX_MULTI_ERR_DEVICES];
>>       int error_dev_num;
>> @@ -76,6 +79,40 @@ struct aer_rpc {
>>                                        */
>>  };
>>
>> +/* AER stats for the device */
>> +struct aer_stats {
>> +
>> +     /*
>> +      * Fields for all AER capable devices. They indicate the errors
>> +      * "as seen by this device". Note that this may mean that if an
>> +      * end point is causing problems, the AER counters may increment
>> +      * at its link partner (e.g. root port) because the errors will be
>> +      * "seen" by the link partner and not the the problematic end point
>> +      * itself (which may report all counters as 0 as it never saw any
>> +      * problems).
>> +      */
>> +     /* Individual counters for different type of correctable errors */
>> +     u64 dev_cor_errs[AER_MAX_TYPEOF_CORRECTABLE_ERRS];
>> +     /* Individual counters for different type of uncorrectable errors */
>> +     u64 dev_uncor_errs[AER_MAX_TYPEOF_UNCORRECTABLE_ERRS];
>> +     /* Total number of correctable errors seen by this device */
>> +     u64 dev_total_cor_errs;
>> +     /* Total number of fatal uncorrectable errors seen by this device */
>> +     u64 dev_total_fatal_errs;
>> +     /* Total number of fatal uncorrectable errors seen by this device */
>> +     u64 dev_total_nonfatal_errs;
>> +
>> +     /*
>> +      * Fields for Root ports only, these indicate the total number of
>> +      * ERR_COR, ERR_FATAL, and ERR_NONFATAL messages received by the
>> +      * rootport, INCLUDING the ones that are generated internally (by
>> +      * the rootport itself)
>
> Strictly speaking, I think these are applicable for both root ports
> and root complex event collectors, right?

Correct, I will reword this comment to state this.

>
>> +      */
>> +     u64 rootport_total_cor_errs;
>> +     u64 rootport_total_fatal_errs;
>> +     u64 rootport_total_nonfatal_errs;
>> +};
>> +
>>  #define AER_LOG_TLP_MASKS            (PCI_ERR_UNC_POISON_TLP|        \
>>                                       PCI_ERR_UNC_ECRC|               \
>>                                       PCI_ERR_UNC_UNSUP|              \
>> @@ -402,12 +439,35 @@ int pci_cleanup_aer_error_status_regs(struct pci_dev *dev)
>>       return 0;
>>  }
>>
>> +static int pci_aer_stats_init(struct pci_dev *pdev)
>> +{
>> +     pdev->aer_stats = kzalloc(sizeof(struct aer_stats), GFP_KERNEL);
>> +     if (!pdev->aer_stats) {
>> +             dev_err(&pdev->dev, "No memory for aer_stats\n");
>
> pci_err(), if we need the message at all.
>
> Based on c7abb2352c29 ("PCI: Remove unnecessary messages for memory
> allocation failures"), I'd be inclined to drop the message completely.

Will do.

>
>> +             return -ENOMEM;
>> +     }
>> +     return 0;
>> +}
>> +
>> +static void pci_aer_stats_exit(struct pci_dev *pdev)
>> +{
>> +     kfree(pdev->aer_stats);
>> +     pdev->aer_stats = NULL;
>> +}
>> +
>>  int pci_aer_init(struct pci_dev *dev)
>>  {
>>       dev->aer_cap = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
>> +     if (!dev->aer_cap || pci_aer_stats_init(dev))
>> +             return -EIO;
>
> This skips pci_cleanup_aer_error_status_regs() if the kzalloc() fails.
> I think we should still do pci_cleanup_aer_error_status_regs(), even
> if the alloc fails.

Will do.

>
> Nobody checks the return value of pci_aer_init(), so I think you can
> simplify this by making these functions void.

Will do.

>
> Maybe even squash them together, i.e., do the kzalloc() directly in
> pci_aer_init() and the kfree() directly in pci_aer_exit()?

Will do.

>
>>       return pci_cleanup_aer_error_status_regs(dev);
>>  }
>>
>> +void pci_aer_exit(struct pci_dev *dev)
>> +{
>> +     pci_aer_stats_exit(dev);
>> +}
>> +
>>  #define AER_AGENT_RECEIVER           0
>>  #define AER_AGENT_REQUESTER          1
>>  #define AER_AGENT_COMPLETER          2
>> diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c
>> index ac876e32de4b..48edd0c9e4bc 100644
>> --- a/drivers/pci/probe.c
>> +++ b/drivers/pci/probe.c
>> @@ -2064,6 +2064,7 @@ static void pci_configure_device(struct pci_dev *dev)
>>
>>  static void pci_release_capabilities(struct pci_dev *dev)
>>  {
>> +     pci_aer_exit(dev);
>>       pci_vpd_release(dev);
>>       pci_iov_release(dev);
>>       pci_free_cap_save_buffers(dev);
>> diff --git a/include/linux/pci.h b/include/linux/pci.h
>> index 340029b2fb38..8d59c6c19a19 100644
>> --- a/include/linux/pci.h
>> +++ b/include/linux/pci.h
>> @@ -299,6 +299,7 @@ struct pci_dev {
>>       u8              hdr_type;       /* PCI header type (`multi' flag masked out) */
>>  #ifdef CONFIG_PCIEAER
>>       u16             aer_cap;        /* AER capability offset */
>> +     struct aer_stats *aer_stats;    /* AER stats for this device */
>>  #endif
>>       u8              pcie_cap;       /* PCIe capability offset */
>>       u8              msi_cap;        /* MSI capability offset */
>> @@ -1471,10 +1472,12 @@ static inline bool pcie_aspm_support_enabled(void) { return false; }
>>  void pci_no_aer(void);
>>  bool pci_aer_available(void);
>>  int pci_aer_init(struct pci_dev *dev);
>> +void pci_aer_exit(struct pci_dev *dev);
>
> With the exception of pci_aer_available(), these are only used inside
> drivers/pci.  This might be a good opportunity to move those private
> things to drivers/pci/pci.h (in a separate patch, of course).


Will do.

>
>>  #else
>>  static inline void pci_no_aer(void) { }
>>  static inline bool pci_aer_available(void) { return false; }
>>  static inline int pci_aer_init(struct pci_dev *d) { return -ENODEV; }
>> +static inline void pci_aer_exit(struct pci_dev *d) { }
>>  #endif
>>
>>  #ifdef CONFIG_PCIE_ECRC
>> --
>> 2.18.0.rc1.244.gcf134e6275-goog
>>