powerpc/eeh: Trace EEH state based on PE

Since we've introduced dedicated struct to trace individual PEs,
it's reasonable to trace its state through the dedicated struct
instead of using "eeh_dev" any more.

The patches implements the state tracing based on PE. It's notable
that the PE state will be applied to the specified PE as well as
its child PEs. That complies with the rule that problematic parent
PE will prevent those child PEs from working properly.

Signed-off-by: Gavin Shan <shangw@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

diff --git a/arch/powerpc/include/asm/eeh.h b/arch/powerpc/include/asm/eeh.h
index 250ae2718bb8ef2ba12fce3b72aabe3251aedadc..f86a85f5f049bc1ab6e4daee8aad7e04fbe9fed4 100644 (file)
--- a/arch/powerpc/include/asm/eeh.h
+++ b/arch/powerpc/include/asm/eeh.h
@@ -67,6 +67,9 @@ struct eeh_pe {
        struct list_head child;         /* Child PEs                    */
 };
 
+#define eeh_pe_for_each_dev(pe, edev) \
+               list_for_each_entry(edev, &pe->edevs, list)
+
 /*
  * The struct is used to trace EEH state for the associated
  * PCI device node or PCI device. In future, it might

diff --git a/arch/powerpc/include/asm/ppc-pci.h b/arch/powerpc/include/asm/ppc-pci.h
index 80fa704d410fe5105bd7776776a5af9a257147ac..c7e5bd60bca16cf0c091fa07232d18decde0be4c 100644 (file)
--- a/arch/powerpc/include/asm/ppc-pci.h
+++ b/arch/powerpc/include/asm/ppc-pci.h
@@ -57,8 +57,8 @@ int eeh_reset_pe(struct eeh_dev *);
 void eeh_restore_bars(struct eeh_dev *);
 int rtas_write_config(struct pci_dn *, int where, int size, u32 val);
 int rtas_read_config(struct pci_dn *, int where, int size, u32 *val);
-void eeh_mark_slot(struct device_node *dn, int mode_flag);
-void eeh_clear_slot(struct device_node *dn, int mode_flag);
+void eeh_pe_state_mark(struct eeh_pe *pe, int state);
+void eeh_pe_state_clear(struct eeh_pe *pe, int state);
 struct device_node *eeh_find_device_pe(struct device_node *dn);
 
 void eeh_sysfs_add_device(struct pci_dev *pdev);

diff --git a/arch/powerpc/platforms/pseries/eeh.c b/arch/powerpc/platforms/pseries/eeh.c
index b60863b184c363ce815d73001eece96f73c20f0c..9c623c23057fa1ea7074ba46c68a7dd5086dcdf6 100644 (file)
--- a/arch/powerpc/platforms/pseries/eeh.c
+++ b/arch/powerpc/platforms/pseries/eeh.c
@@ -278,108 +278,6 @@ struct device_node *eeh_find_device_pe(struct device_node *dn)
        return dn;
 }
 
-/**
- * __eeh_mark_slot - Mark all child devices as failed
- * @parent: parent device
- * @mode_flag: failure flag
- *
- * Mark all devices that are children of this device as failed.
- * Mark the device driver too, so that it can see the failure
- * immediately; this is critical, since some drivers poll
- * status registers in interrupts ... If a driver is polling,
- * and the slot is frozen, then the driver can deadlock in
- * an interrupt context, which is bad.
- */
-static void __eeh_mark_slot(struct device_node *parent, int mode_flag)
-{
-       struct device_node *dn;
-
-       for_each_child_of_node(parent, dn) {
-               if (of_node_to_eeh_dev(dn)) {
-                       /* Mark the pci device driver too */
-                       struct pci_dev *dev = of_node_to_eeh_dev(dn)->pdev;
-
-                       of_node_to_eeh_dev(dn)->mode |= mode_flag;
-
-                       if (dev && dev->driver)
-                               dev->error_state = pci_channel_io_frozen;
-
-                       __eeh_mark_slot(dn, mode_flag);
-               }
-       }
-}
-
-/**
- * eeh_mark_slot - Mark the indicated device and its children as failed
- * @dn: parent device
- * @mode_flag: failure flag
- *
- * Mark the indicated device and its child devices as failed.
- * The device drivers are marked as failed as well.
- */
-void eeh_mark_slot(struct device_node *dn, int mode_flag)
-{
-       struct pci_dev *dev;
-       dn = eeh_find_device_pe(dn);
-
-       /* Back up one, since config addrs might be shared */
-       if (!pcibios_find_pci_bus(dn) && of_node_to_eeh_dev(dn->parent))
-               dn = dn->parent;
-
-       of_node_to_eeh_dev(dn)->mode |= mode_flag;
-
-       /* Mark the pci device too */
-       dev = of_node_to_eeh_dev(dn)->pdev;
-       if (dev)
-               dev->error_state = pci_channel_io_frozen;
-
-       __eeh_mark_slot(dn, mode_flag);
-}
-
-/**
- * __eeh_clear_slot - Clear failure flag for the child devices
- * @parent: parent device
- * @mode_flag: flag to be cleared
- *
- * Clear failure flag for the child devices.
- */
-static void __eeh_clear_slot(struct device_node *parent, int mode_flag)
-{
-       struct device_node *dn;
-
-       for_each_child_of_node(parent, dn) {
-               if (of_node_to_eeh_dev(dn)) {
-                       of_node_to_eeh_dev(dn)->mode &= ~mode_flag;
-                       of_node_to_eeh_dev(dn)->check_count = 0;
-                       __eeh_clear_slot(dn, mode_flag);
-               }
-       }
-}
-
-/**
- * eeh_clear_slot - Clear failure flag for the indicated device and its children
- * @dn: parent device
- * @mode_flag: flag to be cleared
- *
- * Clear failure flag for the indicated device and its children.
- */
-void eeh_clear_slot(struct device_node *dn, int mode_flag)
-{
-       unsigned long flags;
-       raw_spin_lock_irqsave(&confirm_error_lock, flags);
-       
-       dn = eeh_find_device_pe(dn);
-       
-       /* Back up one, since config addrs might be shared */
-       if (!pcibios_find_pci_bus(dn) && of_node_to_eeh_dev(dn->parent))
-               dn = dn->parent;
-
-       of_node_to_eeh_dev(dn)->mode &= ~mode_flag;
-       of_node_to_eeh_dev(dn)->check_count = 0;
-       __eeh_clear_slot(dn, mode_flag);
-       raw_spin_unlock_irqrestore(&confirm_error_lock, flags);
-}
-
 /**
  * eeh_dn_check_failure - Check if all 1's data is due to EEH slot freeze
  * @dn: device node

diff --git a/arch/powerpc/platforms/pseries/eeh_pe.c b/arch/powerpc/platforms/pseries/eeh_pe.c
index e941e6315fa6b83e86e455893b3b962441ab1a3b..0dac1b69582d4867714c655956cc7d9c92821582 100644 (file)
--- a/arch/powerpc/platforms/pseries/eeh_pe.c
+++ b/arch/powerpc/platforms/pseries/eeh_pe.c
@@ -388,3 +388,82 @@ int eeh_rmv_from_parent_pe(struct eeh_dev *edev)
 
        return 0;
 }
+
+/**
+ * __eeh_pe_state_mark - Mark the state for the PE
+ * @data: EEH PE
+ * @flag: state
+ *
+ * The function is used to mark the indicated state for the given
+ * PE. Also, the associated PCI devices will be put into IO frozen
+ * state as well.
+ */
+static void *__eeh_pe_state_mark(void *data, void *flag)
+{
+       struct eeh_pe *pe = (struct eeh_pe *)data;
+       int state = *((int *)flag);
+       struct eeh_dev *tmp;
+       struct pci_dev *pdev;
+
+       /*
+        * Mark the PE with the indicated state. Also,
+        * the associated PCI device will be put into
+        * I/O frozen state to avoid I/O accesses from
+        * the PCI device driver.
+        */
+       pe->state |= state;
+       eeh_pe_for_each_dev(pe, tmp) {
+               pdev = eeh_dev_to_pci_dev(tmp);
+               if (pdev)
+                       pdev->error_state = pci_channel_io_frozen;
+       }
+
+       return NULL;
+}
+
+/**
+ * eeh_pe_state_mark - Mark specified state for PE and its associated device
+ * @pe: EEH PE
+ *
+ * EEH error affects the current PE and its child PEs. The function
+ * is used to mark appropriate state for the affected PEs and the
+ * associated devices.
+ */
+void eeh_pe_state_mark(struct eeh_pe *pe, int state)
+{
+       eeh_pe_traverse(pe, __eeh_pe_state_mark, &state);
+}
+
+/**
+ * __eeh_pe_state_clear - Clear state for the PE
+ * @data: EEH PE
+ * @flag: state
+ *
+ * The function is used to clear the indicated state from the
+ * given PE. Besides, we also clear the check count of the PE
+ * as well.
+ */
+static void *__eeh_pe_state_clear(void *data, void *flag)
+{
+       struct eeh_pe *pe = (struct eeh_pe *)data;
+       int state = *((int *)flag);
+
+       pe->state &= ~state;
+       pe->check_count = 0;
+
+       return NULL;
+}
+
+/**
+ * eeh_pe_state_clear - Clear state for the PE and its children
+ * @pe: PE
+ * @state: state to be cleared
+ *
+ * When the PE and its children has been recovered from error,
+ * we need clear the error state for that. The function is used
+ * for the purpose.
+ */
+void eeh_pe_state_clear(struct eeh_pe *pe, int state)
+{
+       eeh_pe_traverse(pe, __eeh_pe_state_clear, &state);
+}