Re: [PATCH 2/4] GenWQE: Add support for EEH error recovery
From: Frank Haverkamp
Date: Wed Jun 04 2014 - 10:06:21 EST
Am Mittwoch, den 04.06.2014, 10:57 -0300 schrieb Kleber Sacilotto de
Souza:
> This patch implements the callbacks and functions necessary to have EEH
> recovery support.
>
> It adds a config option to enable or disable explicit calls to trigger
> platform specific mechanisms on error recovery paths. This option is
> enabled by default only on PPC64 systems and can be overritten via
> debugfs. If this option is enabled, on the error recovery path the
> driver will call pci_channel_offline() to check for error condition and
> issue non-raw MMIO reads to trigger early EEH detection in case of
> hardware failures. This is necessary since the driver MMIO helper
> funtions use raw accessors.
>
> Signed-off-by: Kleber Sacilotto de Souza <klebers@xxxxxxxxxxxxxxxxxx>
> ---
> drivers/misc/genwqe/Kconfig | 6 +++
> drivers/misc/genwqe/card_base.c | 79 ++++++++++++++++++++++++++++++------
> drivers/misc/genwqe/card_base.h | 2 +
> drivers/misc/genwqe/card_ddcb.c | 24 +++++++++--
> drivers/misc/genwqe/card_debugfs.c | 7 +++
> drivers/misc/genwqe/card_dev.c | 5 ++
> drivers/misc/genwqe/card_utils.c | 10 +++++
> 7 files changed, 115 insertions(+), 18 deletions(-)
>
> diff --git a/drivers/misc/genwqe/Kconfig b/drivers/misc/genwqe/Kconfig
> index 6069d8c..4c0a033 100644
> --- a/drivers/misc/genwqe/Kconfig
> +++ b/drivers/misc/genwqe/Kconfig
> @@ -11,3 +11,9 @@ menuconfig GENWQE
> Enables PCIe card driver for IBM GenWQE accelerators.
> The user-space interface is described in
> include/linux/genwqe/genwqe_card.h.
> +
> +config GENWQE_PLATFORM_ERROR_RECOVERY
> + int "Use platform recovery procedures (0=off, 1=on)"
> + depends on GENWQE
> + default 1 if PPC64
> + default 0
> diff --git a/drivers/misc/genwqe/card_base.c b/drivers/misc/genwqe/card_base.c
> index e6cc3e1..87ebaba 100644
> --- a/drivers/misc/genwqe/card_base.c
> +++ b/drivers/misc/genwqe/card_base.c
> @@ -140,6 +140,12 @@ static struct genwqe_dev *genwqe_dev_alloc(void)
> cd->class_genwqe = class_genwqe;
> cd->debugfs_genwqe = debugfs_genwqe;
>
> + /*
> + * This comes from kernel config option and can be overritten via
> + * debugfs.
> + */
> + cd->use_platform_recovery = CONFIG_GENWQE_PLATFORM_ERROR_RECOVERY;
> +
> init_waitqueue_head(&cd->queue_waitq);
>
> spin_lock_init(&cd->file_lock);
> @@ -943,6 +949,19 @@ static int genwqe_health_thread(void *data)
> return 0;
>
> fatal_error:
> + if (cd->use_platform_recovery) {
> + /*
> + * Since we use raw accessors, EEH errors won't be detected
> + * by the platform until we do a non-raw MMIO or config space
> + * read
> + */
> + readq(cd->mmio + IO_SLC_CFGREG_GFIR);
> +
> + /* We do nothing if the card is going over PCI recovery */
> + if (pci_channel_offline(pci_dev))
> + return -EIO;
> + }
> +
> dev_err(&pci_dev->dev,
> "[%s] card unusable. Please trigger unbind!\n", __func__);
>
> @@ -1048,6 +1067,9 @@ static int genwqe_pci_setup(struct genwqe_dev *cd)
> pci_set_master(pci_dev);
> pci_enable_pcie_error_reporting(pci_dev);
>
> + /* EEH recovery requires PCIe fundamental reset */
> + pci_dev->needs_freset = 1;
> +
> /* request complete BAR-0 space (length = 0) */
> cd->mmio_len = pci_resource_len(pci_dev, 0);
> cd->mmio = pci_iomap(pci_dev, 0, 0);
> @@ -1186,23 +1208,40 @@ static pci_ers_result_t genwqe_err_error_detected(struct pci_dev *pci_dev,
>
> dev_err(&pci_dev->dev, "[%s] state=%d\n", __func__, state);
>
> - if (pci_dev == NULL)
> - return PCI_ERS_RESULT_NEED_RESET;
> -
> cd = dev_get_drvdata(&pci_dev->dev);
> if (cd == NULL)
> - return PCI_ERS_RESULT_NEED_RESET;
> + return PCI_ERS_RESULT_DISCONNECT;
>
> - switch (state) {
> - case pci_channel_io_normal:
> - return PCI_ERS_RESULT_CAN_RECOVER;
> - case pci_channel_io_frozen:
> - return PCI_ERS_RESULT_NEED_RESET;
> - case pci_channel_io_perm_failure:
> + /* Stop the card */
> + genwqe_health_check_stop(cd);
> + genwqe_stop(cd);
> +
> + /*
> + * On permanent failure, the PCI code will call device remove
> + * after the return of this function.
> + * genwqe_stop() can be called twice.
> + */
> + if (state == pci_channel_io_perm_failure) {
> return PCI_ERS_RESULT_DISCONNECT;
> + } else {
> + genwqe_pci_remove(cd);
> + return PCI_ERS_RESULT_NEED_RESET;
> }
> +}
> +
> +static pci_ers_result_t genwqe_err_slot_reset(struct pci_dev *pci_dev)
> +{
> + int rc;
> + struct genwqe_dev *cd = dev_get_drvdata(&pci_dev->dev);
>
> - return PCI_ERS_RESULT_NEED_RESET;
> + rc = genwqe_pci_setup(cd);
> + if (!rc) {
> + return PCI_ERS_RESULT_RECOVERED;
> + } else {
> + dev_err(&pci_dev->dev,
> + "err: problems with PCI setup (err=%d)\n", rc);
> + return PCI_ERS_RESULT_DISCONNECT;
> + }
> }
>
> static pci_ers_result_t genwqe_err_result_none(struct pci_dev *dev)
> @@ -1210,8 +1249,22 @@ static pci_ers_result_t genwqe_err_result_none(struct pci_dev *dev)
> return PCI_ERS_RESULT_NONE;
> }
>
> -static void genwqe_err_resume(struct pci_dev *dev)
> +static void genwqe_err_resume(struct pci_dev *pci_dev)
> {
> + int rc;
> + struct genwqe_dev *cd = dev_get_drvdata(&pci_dev->dev);
> +
> + rc = genwqe_start(cd);
> + if (!rc) {
> + rc = genwqe_health_check_start(cd);
> + if (rc)
> + dev_err(&pci_dev->dev,
> + "err: cannot start health checking! (err=%d)\n",
> + rc);
> + } else {
> + dev_err(&pci_dev->dev,
> + "err: cannot start card services! (err=%d)\n", rc);
> + }
> }
>
> static int genwqe_sriov_configure(struct pci_dev *dev, int numvfs)
> @@ -1234,7 +1287,7 @@ static struct pci_error_handlers genwqe_err_handler = {
> .error_detected = genwqe_err_error_detected,
> .mmio_enabled = genwqe_err_result_none,
> .link_reset = genwqe_err_result_none,
> - .slot_reset = genwqe_err_result_none,
> + .slot_reset = genwqe_err_slot_reset,
> .resume = genwqe_err_resume,
> };
>
> diff --git a/drivers/misc/genwqe/card_base.h b/drivers/misc/genwqe/card_base.h
> index 0e608a2..67abd8c 100644
> --- a/drivers/misc/genwqe/card_base.h
> +++ b/drivers/misc/genwqe/card_base.h
> @@ -291,6 +291,8 @@ struct genwqe_dev {
> struct task_struct *health_thread;
> wait_queue_head_t health_waitq;
>
> + int use_platform_recovery; /* use platform recovery mechanisms */
> +
> /* char device */
> dev_t devnum_genwqe; /* major/minor num card */
> struct class *class_genwqe; /* reference to class object */
> diff --git a/drivers/misc/genwqe/card_ddcb.c b/drivers/misc/genwqe/card_ddcb.c
> index c8046db..f0de615 100644
> --- a/drivers/misc/genwqe/card_ddcb.c
> +++ b/drivers/misc/genwqe/card_ddcb.c
> @@ -1118,7 +1118,21 @@ static irqreturn_t genwqe_pf_isr(int irq, void *dev_id)
> * safer, but slower for the good-case ... See above.
> */
> gfir = __genwqe_readq(cd, IO_SLC_CFGREG_GFIR);
> - if ((gfir & GFIR_ERR_TRIGGER) != 0x0) {
> + if (((gfir & GFIR_ERR_TRIGGER) != 0x0) &&
> + !pci_channel_offline(pci_dev)) {
> +
> + if (cd->use_platform_recovery) {
> + /*
> + * Since we use raw accessors, EEH errors won't be
> + * detected by the platform until we do a non-raw
> + * MMIO or config space read
> + */
> + readq(cd->mmio + IO_SLC_CFGREG_GFIR);
> +
> + /* Don't do anything if the PCI channel is frozen */
> + if (pci_channel_offline(pci_dev))
> + goto exit;
> + }
>
> wake_up_interruptible(&cd->health_waitq);
>
> @@ -1126,12 +1140,12 @@ static irqreturn_t genwqe_pf_isr(int irq, void *dev_id)
> * By default GFIRs causes recovery actions. This
> * count is just for debug when recovery is masked.
> */
> - printk_ratelimited(KERN_ERR
> - "%s %s: [%s] GFIR=%016llx\n",
> - GENWQE_DEVNAME, dev_name(&pci_dev->dev),
> - __func__, gfir);
> + dev_err_ratelimited(&pci_dev->dev,
> + "[%s] GFIR=%016llx\n",
> + __func__, gfir);
> }
>
> + exit:
> return IRQ_HANDLED;
> }
>
> diff --git a/drivers/misc/genwqe/card_debugfs.c b/drivers/misc/genwqe/card_debugfs.c
> index 50d2096..2171915 100644
> --- a/drivers/misc/genwqe/card_debugfs.c
> +++ b/drivers/misc/genwqe/card_debugfs.c
> @@ -485,6 +485,13 @@ int genwqe_init_debugfs(struct genwqe_dev *cd)
> goto err1;
> }
>
> + file = debugfs_create_u32("use_platform_recovery", 0666, root,
> + &cd->use_platform_recovery);
> + if (!file) {
> + ret = -ENOMEM;
> + goto err1;
> + }
> +
> cd->debugfs_root = root;
> return 0;
> err1:
> diff --git a/drivers/misc/genwqe/card_dev.c b/drivers/misc/genwqe/card_dev.c
> index 1d2f163..aae4255 100644
> --- a/drivers/misc/genwqe/card_dev.c
> +++ b/drivers/misc/genwqe/card_dev.c
> @@ -1048,10 +1048,15 @@ static long genwqe_ioctl(struct file *filp, unsigned int cmd,
> int rc = 0;
> struct genwqe_file *cfile = (struct genwqe_file *)filp->private_data;
> struct genwqe_dev *cd = cfile->cd;
> + struct pci_dev *pci_dev = cd->pci_dev;
> struct genwqe_reg_io __user *io;
> u64 val;
> u32 reg_offs;
>
> + /* Return -EIO if card hit EEH */
> + if (pci_channel_offline(pci_dev))
> + return -EIO;
> +
> if (_IOC_TYPE(cmd) != GENWQE_IOC_CODE)
> return -EINVAL;
>
> diff --git a/drivers/misc/genwqe/card_utils.c b/drivers/misc/genwqe/card_utils.c
> index d049d27..61846dd 100644
> --- a/drivers/misc/genwqe/card_utils.c
> +++ b/drivers/misc/genwqe/card_utils.c
> @@ -53,12 +53,17 @@
> */
> int __genwqe_writeq(struct genwqe_dev *cd, u64 byte_offs, u64 val)
> {
> + struct pci_dev *pci_dev = cd->pci_dev;
> +
> if (cd->err_inject & GENWQE_INJECT_HARDWARE_FAILURE)
> return -EIO;
>
> if (cd->mmio == NULL)
> return -EIO;
>
> + if (pci_channel_offline(pci_dev))
> + return -EIO;
> +
> __raw_writeq((__force u64)cpu_to_be64(val), cd->mmio + byte_offs);
> return 0;
> }
> @@ -99,12 +104,17 @@ u64 __genwqe_readq(struct genwqe_dev *cd, u64 byte_offs)
> */
> int __genwqe_writel(struct genwqe_dev *cd, u64 byte_offs, u32 val)
> {
> + struct pci_dev *pci_dev = cd->pci_dev;
> +
> if (cd->err_inject & GENWQE_INJECT_HARDWARE_FAILURE)
> return -EIO;
>
> if (cd->mmio == NULL)
> return -EIO;
>
> + if (pci_channel_offline(pci_dev))
> + return -EIO;
> +
> __raw_writel((__force u32)cpu_to_be32(val), cd->mmio + byte_offs);
> return 0;
> }
Thanks for contributing those additions to our driver.
Acked-by: Frank Haverkamp <haver@xxxxxxxxxxxxxxxxxx>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/