aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorBenjamin Herrenschmidt <benh@kernel.crashing.org>2012-03-12 19:15:35 -0400
committerBenjamin Herrenschmidt <benh@kernel.crashing.org>2012-03-12 19:15:35 -0400
commitaba0eb84c87928992c021d33ef3ea59c931086b9 (patch)
treebdc323efc6773d0dcce4406ca275e9192eabb9b2
parent7230c5644188cd9e3fb380cc97dde00c464a3ba7 (diff)
parent3780444c4fcec28c96ab7002858bb051215a5fc1 (diff)
Merge branch 'eeh' into next
Diffstat
-rw-r--r--arch/powerpc/include/asm/device.h3
-rw-r--r--arch/powerpc/include/asm/eeh.h134
-rw-r--r--arch/powerpc/include/asm/eeh_event.h33
-rw-r--r--arch/powerpc/include/asm/ppc-pci.h89
-rw-r--r--arch/powerpc/kernel/of_platform.c6
-rw-r--r--arch/powerpc/kernel/rtas_pci.c3
-rw-r--r--arch/powerpc/platforms/pseries/Makefile3
-rw-r--r--arch/powerpc/platforms/pseries/eeh.c1044
-rw-r--r--arch/powerpc/platforms/pseries/eeh_cache.c44
-rw-r--r--arch/powerpc/platforms/pseries/eeh_dev.c102
-rw-r--r--arch/powerpc/platforms/pseries/eeh_driver.c213
-rw-r--r--arch/powerpc/platforms/pseries/eeh_event.c55
-rw-r--r--arch/powerpc/platforms/pseries/eeh_pseries.c565
-rw-r--r--arch/powerpc/platforms/pseries/eeh_sysfs.c25
-rw-r--r--arch/powerpc/platforms/pseries/msi.c2
-rw-r--r--arch/powerpc/platforms/pseries/pci_dlpar.c3
-rw-r--r--arch/powerpc/platforms/pseries/setup.c7
-rw-r--r--include/linux/of.h10
-rw-r--r--include/linux/pci.h7
19 files changed, 1479 insertions, 869 deletions
diff --git a/arch/powerpc/include/asm/device.h b/arch/powerpc/include/asm/device.h
index d57c08acedf..63d5ca49cec 100644
--- a/arch/powerpc/include/asm/device.h
+++ b/arch/powerpc/include/asm/device.h
@@ -31,6 +31,9 @@ struct dev_archdata {
31#ifdef CONFIG_SWIOTLB 31#ifdef CONFIG_SWIOTLB
32 dma_addr_t max_direct_dma_addr; 32 dma_addr_t max_direct_dma_addr;
33#endif 33#endif
34#ifdef CONFIG_EEH
35 struct eeh_dev *edev;
36#endif
34}; 37};
35 38
36struct pdev_archdata { 39struct pdev_archdata {
diff --git a/arch/powerpc/include/asm/eeh.h b/arch/powerpc/include/asm/eeh.h
index 66ea9b8b95c..d60f99814ff 100644
--- a/arch/powerpc/include/asm/eeh.h
+++ b/arch/powerpc/include/asm/eeh.h
@@ -1,6 +1,6 @@
1/* 1/*
2 * eeh.h
3 * Copyright (C) 2001 Dave Engebretsen & Todd Inglett IBM Corporation. 2 * Copyright (C) 2001 Dave Engebretsen & Todd Inglett IBM Corporation.
3 * Copyright 2001-2012 IBM Corporation.
4 * 4 *
5 * This program is free software; you can redistribute it and/or modify 5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by 6 * it under the terms of the GNU General Public License as published by
@@ -31,44 +31,105 @@ struct device_node;
31 31
32#ifdef CONFIG_EEH 32#ifdef CONFIG_EEH
33 33
34extern int eeh_subsystem_enabled; 34/*
35 * The struct is used to trace EEH state for the associated
36 * PCI device node or PCI device. In future, it might
37 * represent PE as well so that the EEH device to form
38 * another tree except the currently existing tree of PCI
39 * buses and PCI devices
40 */
41#define EEH_MODE_SUPPORTED (1<<0) /* EEH supported on the device */
42#define EEH_MODE_NOCHECK (1<<1) /* EEH check should be skipped */
43#define EEH_MODE_ISOLATED (1<<2) /* The device has been isolated */
44#define EEH_MODE_RECOVERING (1<<3) /* Recovering the device */
45#define EEH_MODE_IRQ_DISABLED (1<<4) /* Interrupt disabled */
46
47struct eeh_dev {
48 int mode; /* EEH mode */
49 int class_code; /* Class code of the device */
50 int config_addr; /* Config address */
51 int pe_config_addr; /* PE config address */
52 int check_count; /* Times of ignored error */
53 int freeze_count; /* Times of froze up */
54 int false_positives; /* Times of reported #ff's */
55 u32 config_space[16]; /* Saved PCI config space */
56 struct pci_controller *phb; /* Associated PHB */
57 struct device_node *dn; /* Associated device node */
58 struct pci_dev *pdev; /* Associated PCI device */
59};
60
61static inline struct device_node *eeh_dev_to_of_node(struct eeh_dev *edev)
62{
63 return edev->dn;
64}
65
66static inline struct pci_dev *eeh_dev_to_pci_dev(struct eeh_dev *edev)
67{
68 return edev->pdev;
69}
35 70
36/* Values for eeh_mode bits in device_node */ 71/*
37#define EEH_MODE_SUPPORTED (1<<0) 72 * The struct is used to trace the registered EEH operation
38#define EEH_MODE_NOCHECK (1<<1) 73 * callback functions. Actually, those operation callback
39#define EEH_MODE_ISOLATED (1<<2) 74 * functions are heavily platform dependent. That means the
40#define EEH_MODE_RECOVERING (1<<3) 75 * platform should register its own EEH operation callback
41#define EEH_MODE_IRQ_DISABLED (1<<4) 76 * functions before any EEH further operations.
77 */
78#define EEH_OPT_DISABLE 0 /* EEH disable */
79#define EEH_OPT_ENABLE 1 /* EEH enable */
80#define EEH_OPT_THAW_MMIO 2 /* MMIO enable */
81#define EEH_OPT_THAW_DMA 3 /* DMA enable */
82#define EEH_STATE_UNAVAILABLE (1 << 0) /* State unavailable */
83#define EEH_STATE_NOT_SUPPORT (1 << 1) /* EEH not supported */
84#define EEH_STATE_RESET_ACTIVE (1 << 2) /* Active reset */
85#define EEH_STATE_MMIO_ACTIVE (1 << 3) /* Active MMIO */
86#define EEH_STATE_DMA_ACTIVE (1 << 4) /* Active DMA */
87#define EEH_STATE_MMIO_ENABLED (1 << 5) /* MMIO enabled */
88#define EEH_STATE_DMA_ENABLED (1 << 6) /* DMA enabled */
89#define EEH_RESET_DEACTIVATE 0 /* Deactivate the PE reset */
90#define EEH_RESET_HOT 1 /* Hot reset */
91#define EEH_RESET_FUNDAMENTAL 3 /* Fundamental reset */
92#define EEH_LOG_TEMP 1 /* EEH temporary error log */
93#define EEH_LOG_PERM 2 /* EEH permanent error log */
94
95struct eeh_ops {
96 char *name;
97 int (*init)(void);
98 int (*set_option)(struct device_node *dn, int option);
99 int (*get_pe_addr)(struct device_node *dn);
100 int (*get_state)(struct device_node *dn, int *state);
101 int (*reset)(struct device_node *dn, int option);
102 int (*wait_state)(struct device_node *dn, int max_wait);
103 int (*get_log)(struct device_node *dn, int severity, char *drv_log, unsigned long len);
104 int (*configure_bridge)(struct device_node *dn);
105 int (*read_config)(struct device_node *dn, int where, int size, u32 *val);
106 int (*write_config)(struct device_node *dn, int where, int size, u32 val);
107};
108
109extern struct eeh_ops *eeh_ops;
110extern int eeh_subsystem_enabled;
42 111
43/* Max number of EEH freezes allowed before we consider the device 112/*
44 * to be permanently disabled. */ 113 * Max number of EEH freezes allowed before we consider the device
114 * to be permanently disabled.
115 */
45#define EEH_MAX_ALLOWED_FREEZES 5 116#define EEH_MAX_ALLOWED_FREEZES 5
46 117
118void * __devinit eeh_dev_init(struct device_node *dn, void *data);
119void __devinit eeh_dev_phb_init_dynamic(struct pci_controller *phb);
120void __init eeh_dev_phb_init(void);
47void __init eeh_init(void); 121void __init eeh_init(void);
122#ifdef CONFIG_PPC_PSERIES
123int __init eeh_pseries_init(void);
124#endif
125int __init eeh_ops_register(struct eeh_ops *ops);
126int __exit eeh_ops_unregister(const char *name);
48unsigned long eeh_check_failure(const volatile void __iomem *token, 127unsigned long eeh_check_failure(const volatile void __iomem *token,
49 unsigned long val); 128 unsigned long val);
50int eeh_dn_check_failure(struct device_node *dn, struct pci_dev *dev); 129int eeh_dn_check_failure(struct device_node *dn, struct pci_dev *dev);
51void __init pci_addr_cache_build(void); 130void __init pci_addr_cache_build(void);
52
53/**
54 * eeh_add_device_early
55 * eeh_add_device_late
56 *
57 * Perform eeh initialization for devices added after boot.
58 * Call eeh_add_device_early before doing any i/o to the
59 * device (including config space i/o). Call eeh_add_device_late
60 * to finish the eeh setup for this device.
61 */
62void eeh_add_device_tree_early(struct device_node *); 131void eeh_add_device_tree_early(struct device_node *);
63void eeh_add_device_tree_late(struct pci_bus *); 132void eeh_add_device_tree_late(struct pci_bus *);
64
65/**
66 * eeh_remove_device_recursive - undo EEH for device & children.
67 * @dev: pci device to be removed
68 *
69 * As above, this removes the device; it also removes child
70 * pci devices as well.
71 */
72void eeh_remove_bus_device(struct pci_dev *); 133void eeh_remove_bus_device(struct pci_dev *);
73 134
74/** 135/**
@@ -87,8 +148,25 @@ void eeh_remove_bus_device(struct pci_dev *);
87#define EEH_IO_ERROR_VALUE(size) (~0U >> ((4 - (size)) * 8)) 148#define EEH_IO_ERROR_VALUE(size) (~0U >> ((4 - (size)) * 8))
88 149
89#else /* !CONFIG_EEH */ 150#else /* !CONFIG_EEH */
151
152static inline void *eeh_dev_init(struct device_node *dn, void *data)
153{
154 return NULL;
155}
156
157static inline void eeh_dev_phb_init_dynamic(struct pci_controller *phb) { }
158
159static inline void eeh_dev_phb_init(void) { }
160
90static inline void eeh_init(void) { } 161static inline void eeh_init(void) { }
91 162
163#ifdef CONFIG_PPC_PSERIES
164static inline int eeh_pseries_init(void)
165{
166 return 0;
167}
168#endif /* CONFIG_PPC_PSERIES */
169
92static inline unsigned long eeh_check_failure(const volatile void __iomem *token, unsigned long val) 170static inline unsigned long eeh_check_failure(const volatile void __iomem *token, unsigned long val)
93{ 171{
94 return val; 172 return val;
diff --git a/arch/powerpc/include/asm/eeh_event.h b/arch/powerpc/include/asm/eeh_event.h
index cc3cb04539a..c68b012b779 100644
--- a/arch/powerpc/include/asm/eeh_event.h
+++ b/arch/powerpc/include/asm/eeh_event.h
@@ -1,6 +1,4 @@
1/* 1/*
2 * eeh_event.h
3 *
4 * This program is free software; you can redistribute it and/or modify 2 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License as published by 3 * it under the terms of the GNU General Public License as published by
6 * the Free Software Foundation; either version 2 of the License, or 4 * the Free Software Foundation; either version 2 of the License, or
@@ -22,32 +20,19 @@
22#define ASM_POWERPC_EEH_EVENT_H 20#define ASM_POWERPC_EEH_EVENT_H
23#ifdef __KERNEL__ 21#ifdef __KERNEL__
24 22
25/** EEH event -- structure holding pci controller data that describes 23/*
26 * a change in the isolation status of a PCI slot. A pointer 24 * structure holding pci controller data that describes a
27 * to this struct is passed as the data pointer in a notify callback. 25 * change in the isolation status of a PCI slot. A pointer
26 * to this struct is passed as the data pointer in a notify
27 * callback.
28 */ 28 */
29struct eeh_event { 29struct eeh_event {
30 struct list_head list; 30 struct list_head list; /* to form event queue */
31 struct device_node *dn; /* struct device node */ 31 struct eeh_dev *edev; /* EEH device */
32 struct pci_dev *dev; /* affected device */
33}; 32};
34 33
35/** 34int eeh_send_failure_event(struct eeh_dev *edev);
36 * eeh_send_failure_event - generate a PCI error event 35struct eeh_dev *handle_eeh_events(struct eeh_event *);
37 * @dev pci device
38 *
39 * This routine builds a PCI error event which will be delivered
40 * to all listeners on the eeh_notifier_chain.
41 *
42 * This routine can be called within an interrupt context;
43 * the actual event will be delivered in a normal context
44 * (from a workqueue).
45 */
46int eeh_send_failure_event (struct device_node *dn,
47 struct pci_dev *dev);
48
49/* Main recovery function */
50struct pci_dn * handle_eeh_events (struct eeh_event *);
51 36
52#endif /* __KERNEL__ */ 37#endif /* __KERNEL__ */
53#endif /* ASM_POWERPC_EEH_EVENT_H */ 38#endif /* ASM_POWERPC_EEH_EVENT_H */
diff --git a/arch/powerpc/include/asm/ppc-pci.h b/arch/powerpc/include/asm/ppc-pci.h
index 6d422979eba..e660b37aa7d 100644
--- a/arch/powerpc/include/asm/ppc-pci.h
+++ b/arch/powerpc/include/asm/ppc-pci.h
@@ -47,92 +47,21 @@ extern int rtas_setup_phb(struct pci_controller *phb);
47 47
48extern unsigned long pci_probe_only; 48extern unsigned long pci_probe_only;
49 49
50/* ---- EEH internal-use-only related routines ---- */
51#ifdef CONFIG_EEH 50#ifdef CONFIG_EEH
52 51
52void pci_addr_cache_build(void);
53void pci_addr_cache_insert_device(struct pci_dev *dev); 53void pci_addr_cache_insert_device(struct pci_dev *dev);
54void pci_addr_cache_remove_device(struct pci_dev *dev); 54void pci_addr_cache_remove_device(struct pci_dev *dev);
55void pci_addr_cache_build(void); 55struct pci_dev *pci_addr_cache_get_device(unsigned long addr);
56struct pci_dev *pci_get_device_by_addr(unsigned long addr); 56void eeh_slot_error_detail(struct eeh_dev *edev, int severity);
57 57int eeh_pci_enable(struct eeh_dev *edev, int function);
58/** 58int eeh_reset_pe(struct eeh_dev *);
59 * eeh_slot_error_detail -- record and EEH error condition to the log 59void eeh_restore_bars(struct eeh_dev *);
60 * @pdn: pci device node
61 * @severity: EEH_LOG_TEMP_FAILURE or EEH_LOG_PERM_FAILURE
62 *
63 * Obtains the EEH error details from the RTAS subsystem,
64 * and then logs these details with the RTAS error log system.
65 */
66#define EEH_LOG_TEMP_FAILURE 1
67#define EEH_LOG_PERM_FAILURE 2
68void eeh_slot_error_detail (struct pci_dn *pdn, int severity);
69
70/**
71 * rtas_pci_enable - enable IO transfers for this slot
72 * @pdn: pci device node
73 * @function: either EEH_THAW_MMIO or EEH_THAW_DMA
74 *
75 * Enable I/O transfers to this slot
76 */
77#define EEH_THAW_MMIO 2
78#define EEH_THAW_DMA 3
79int rtas_pci_enable(struct pci_dn *pdn, int function);
80
81/**
82 * rtas_set_slot_reset -- unfreeze a frozen slot
83 * @pdn: pci device node
84 *
85 * Clear the EEH-frozen condition on a slot. This routine
86 * does this by asserting the PCI #RST line for 1/8th of
87 * a second; this routine will sleep while the adapter is
88 * being reset.
89 *
90 * Returns a non-zero value if the reset failed.
91 */
92int rtas_set_slot_reset (struct pci_dn *);
93int eeh_wait_for_slot_status(struct pci_dn *pdn, int max_wait_msecs);
94
95/**
96 * eeh_restore_bars - Restore device configuration info.
97 * @pdn: pci device node
98 *
99 * A reset of a PCI device will clear out its config space.
100 * This routines will restore the config space for this
101 * device, and is children, to values previously obtained
102 * from the firmware.
103 */
104void eeh_restore_bars(struct pci_dn *);
105
106/**
107 * rtas_configure_bridge -- firmware initialization of pci bridge
108 * @pdn: pci device node
109 *
110 * Ask the firmware to configure all PCI bridges devices
111 * located behind the indicated node. Required after a
112 * pci device reset. Does essentially the same hing as
113 * eeh_restore_bars, but for brdges, and lets firmware
114 * do the work.
115 */
116void rtas_configure_bridge(struct pci_dn *);
117
118int rtas_write_config(struct pci_dn *, int where, int size, u32 val); 60int rtas_write_config(struct pci_dn *, int where, int size, u32 val);
119int rtas_read_config(struct pci_dn *, int where, int size, u32 *val); 61int rtas_read_config(struct pci_dn *, int where, int size, u32 *val);
120 62void eeh_mark_slot(struct device_node *dn, int mode_flag);
121/** 63void eeh_clear_slot(struct device_node *dn, int mode_flag);
122 * eeh_mark_slot -- set mode flags for pertition endpoint 64struct device_node *eeh_find_device_pe(struct device_node *dn);
123 * @pdn: pci device node
124 *
125 * mark and clear slots: find "partition endpoint" PE and set or
126 * clear the flags for each subnode of the PE.
127 */
128void eeh_mark_slot (struct device_node *dn, int mode_flag);
129void eeh_clear_slot (struct device_node *dn, int mode_flag);
130
131/**
132 * find_device_pe -- Find the associated "Partiationable Endpoint" PE
133 * @pdn: pci device node
134 */
135struct device_node * find_device_pe(struct device_node *dn);
136 65
137void eeh_sysfs_add_device(struct pci_dev *pdev); 66void eeh_sysfs_add_device(struct pci_dev *pdev);
138void eeh_sysfs_remove_device(struct pci_dev *pdev); 67void eeh_sysfs_remove_device(struct pci_dev *pdev);
diff --git a/arch/powerpc/kernel/of_platform.c b/arch/powerpc/kernel/of_platform.c
index e1612dfb4a9..2049f2d00ff 100644
--- a/arch/powerpc/kernel/of_platform.c
+++ b/arch/powerpc/kernel/of_platform.c
@@ -21,12 +21,13 @@
21#include <linux/of.h> 21#include <linux/of.h>
22#include <linux/of_device.h> 22#include <linux/of_device.h>
23#include <linux/of_platform.h> 23#include <linux/of_platform.h>
24#include <linux/atomic.h>
24 25
25#include <asm/errno.h> 26#include <asm/errno.h>
26#include <asm/topology.h> 27#include <asm/topology.h>
27#include <asm/pci-bridge.h> 28#include <asm/pci-bridge.h>
28#include <asm/ppc-pci.h> 29#include <asm/ppc-pci.h>
29#include <linux/atomic.h> 30#include <asm/eeh.h>
30 31
31#ifdef CONFIG_PPC_OF_PLATFORM_PCI 32#ifdef CONFIG_PPC_OF_PLATFORM_PCI
32 33
@@ -66,6 +67,9 @@ static int __devinit of_pci_phb_probe(struct platform_device *dev)
66 /* Init pci_dn data structures */ 67 /* Init pci_dn data structures */
67 pci_devs_phb_init_dynamic(phb); 68 pci_devs_phb_init_dynamic(phb);
68 69
70 /* Create EEH devices for the PHB */
71 eeh_dev_phb_init_dynamic(phb);
72
69 /* Register devices with EEH */ 73 /* Register devices with EEH */
70#ifdef CONFIG_EEH 74#ifdef CONFIG_EEH
71 if (dev->dev.of_node->child) 75 if (dev->dev.of_node->child)
diff --git a/arch/powerpc/kernel/rtas_pci.c b/arch/powerpc/kernel/rtas_pci.c
index 6cd8f0196b6..517bd86bc3f 100644
--- a/arch/powerpc/kernel/rtas_pci.c
+++ b/arch/powerpc/kernel/rtas_pci.c
@@ -275,6 +275,9 @@ void __init find_and_init_phbs(void)
275 of_node_put(root); 275 of_node_put(root);
276 pci_devs_phb_init(); 276 pci_devs_phb_init();
277 277
278 /* Create EEH devices for all PHBs */
279 eeh_dev_phb_init();
280
278 /* 281 /*
279 * pci_probe_only and pci_assign_all_buses can be set via properties 282 * pci_probe_only and pci_assign_all_buses can be set via properties
280 * in chosen. 283 * in chosen.
diff --git a/arch/powerpc/platforms/pseries/Makefile b/arch/powerpc/platforms/pseries/Makefile
index 67b3e6e0cf3..c222189f5bb 100644
--- a/arch/powerpc/platforms/pseries/Makefile
+++ b/arch/powerpc/platforms/pseries/Makefile
@@ -6,7 +6,8 @@ obj-y := lpar.o hvCall.o nvram.o reconfig.o \
6 firmware.o power.o dlpar.o mobility.o 6 firmware.o power.o dlpar.o mobility.o
7obj-$(CONFIG_SMP) += smp.o 7obj-$(CONFIG_SMP) += smp.o
8obj-$(CONFIG_SCANLOG) += scanlog.o 8obj-$(CONFIG_SCANLOG) += scanlog.o
9obj-$(CONFIG_EEH) += eeh.o eeh_cache.o eeh_driver.o eeh_event.o eeh_sysfs.o 9obj-$(CONFIG_EEH) += eeh.o eeh_dev.o eeh_cache.o eeh_driver.o \
10 eeh_event.o eeh_sysfs.o eeh_pseries.o
10obj-$(CONFIG_KEXEC) += kexec.o 11obj-$(CONFIG_KEXEC) += kexec.o
11obj-$(CONFIG_PCI) += pci.o pci_dlpar.o 12obj-$(CONFIG_PCI) += pci.o pci_dlpar.o
12obj-$(CONFIG_PSERIES_MSI) += msi.o 13obj-$(CONFIG_PSERIES_MSI) += msi.o
diff --git a/arch/powerpc/platforms/pseries/eeh.c b/arch/powerpc/platforms/pseries/eeh.c
index c0b40af4ce4..8011088392d 100644
--- a/arch/powerpc/platforms/pseries/eeh.c
+++ b/arch/powerpc/platforms/pseries/eeh.c
@@ -1,8 +1,8 @@
1/* 1/*
2 * eeh.c
3 * Copyright IBM Corporation 2001, 2005, 2006 2 * Copyright IBM Corporation 2001, 2005, 2006
4 * Copyright Dave Engebretsen & Todd Inglett 2001 3 * Copyright Dave Engebretsen & Todd Inglett 2001
5 * Copyright Linas Vepstas 2005, 2006 4 * Copyright Linas Vepstas 2005, 2006
5 * Copyright 2001-2012 IBM Corporation.
6 * 6 *
7 * This program is free software; you can redistribute it and/or modify 7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by 8 * it under the terms of the GNU General Public License as published by
@@ -22,7 +22,7 @@
22 */ 22 */
23 23
24#include <linux/delay.h> 24#include <linux/delay.h>
25#include <linux/sched.h> /* for init_mm */ 25#include <linux/sched.h>
26#include <linux/init.h> 26#include <linux/init.h>
27#include <linux/list.h> 27#include <linux/list.h>
28#include <linux/pci.h> 28#include <linux/pci.h>
@@ -86,16 +86,8 @@
86/* Time to wait for a PCI slot to report status, in milliseconds */ 86/* Time to wait for a PCI slot to report status, in milliseconds */
87#define PCI_BUS_RESET_WAIT_MSEC (60*1000) 87#define PCI_BUS_RESET_WAIT_MSEC (60*1000)
88 88
89/* RTAS tokens */ 89/* Platform dependent EEH operations */
90static int ibm_set_eeh_option; 90struct eeh_ops *eeh_ops = NULL;
91static int ibm_set_slot_reset;
92static int ibm_read_slot_reset_state;
93static int ibm_read_slot_reset_state2;
94static int ibm_slot_error_detail;
95static int ibm_get_config_addr_info;
96static int ibm_get_config_addr_info2;
97static int ibm_configure_bridge;
98static int ibm_configure_pe;
99 91
100int eeh_subsystem_enabled; 92int eeh_subsystem_enabled;
101EXPORT_SYMBOL(eeh_subsystem_enabled); 93EXPORT_SYMBOL(eeh_subsystem_enabled);
@@ -103,14 +95,6 @@ EXPORT_SYMBOL(eeh_subsystem_enabled);
103/* Lock to avoid races due to multiple reports of an error */ 95/* Lock to avoid races due to multiple reports of an error */
104static DEFINE_RAW_SPINLOCK(confirm_error_lock); 96static DEFINE_RAW_SPINLOCK(confirm_error_lock);
105 97
106/* Buffer for reporting slot-error-detail rtas calls. Its here
107 * in BSS, and not dynamically alloced, so that it ends up in
108 * RMO where RTAS can access it.
109 */
110static unsigned char slot_errbuf[RTAS_ERROR_LOG_MAX];
111static DEFINE_SPINLOCK(slot_errbuf_lock);
112static int eeh_error_buf_size;
113
114/* Buffer for reporting pci register dumps. Its here in BSS, and 98/* Buffer for reporting pci register dumps. Its here in BSS, and
115 * not dynamically alloced, so that it ends up in RMO where RTAS 99 * not dynamically alloced, so that it ends up in RMO where RTAS
116 * can access it. 100 * can access it.
@@ -118,74 +102,50 @@ static int eeh_error_buf_size;
118#define EEH_PCI_REGS_LOG_LEN 4096 102#define EEH_PCI_REGS_LOG_LEN 4096
119static unsigned char pci_regs_buf[EEH_PCI_REGS_LOG_LEN]; 103static unsigned char pci_regs_buf[EEH_PCI_REGS_LOG_LEN];
120 104
121/* System monitoring statistics */ 105/*
122static unsigned long no_device; 106 * The struct is used to maintain the EEH global statistic
123static unsigned long no_dn; 107 * information. Besides, the EEH global statistics will be
124static unsigned long no_cfg_addr; 108 * exported to user space through procfs
125static unsigned long ignored_check; 109 */
126static unsigned long total_mmio_ffs; 110struct eeh_stats {
127static unsigned long false_positives; 111 u64 no_device; /* PCI device not found */
128static unsigned long slot_resets; 112 u64 no_dn; /* OF node not found */
129 113 u64 no_cfg_addr; /* Config address not found */
130#define IS_BRIDGE(class_code) (((class_code)<<16) == PCI_BASE_CLASS_BRIDGE) 114 u64 ignored_check; /* EEH check skipped */
131 115 u64 total_mmio_ffs; /* Total EEH checks */
132/* --------------------------------------------------------------- */ 116 u64 false_positives; /* Unnecessary EEH checks */
133/* Below lies the EEH event infrastructure */ 117 u64 slot_resets; /* PE reset */
118};
134 119
135static void rtas_slot_error_detail(struct pci_dn *pdn, int severity, 120static struct eeh_stats eeh_stats;
136 char *driver_log, size_t loglen)
137{
138 int config_addr;
139 unsigned long flags;
140 int rc;
141 121
142 /* Log the error with the rtas logger */ 122#define IS_BRIDGE(class_code) (((class_code)<<16) == PCI_BASE_CLASS_BRIDGE)
143 spin_lock_irqsave(&slot_errbuf_lock, flags);
144 memset(slot_errbuf, 0, eeh_error_buf_size);
145
146 /* Use PE configuration address, if present */
147 config_addr = pdn->eeh_config_addr;
148 if (pdn->eeh_pe_config_addr)
149 config_addr = pdn->eeh_pe_config_addr;
150
151 rc = rtas_call(ibm_slot_error_detail,
152 8, 1, NULL, config_addr,
153 BUID_HI(pdn->phb->buid),
154 BUID_LO(pdn->phb->buid),
155 virt_to_phys(driver_log), loglen,
156 virt_to_phys(slot_errbuf),
157 eeh_error_buf_size,
158 severity);
159
160 if (rc == 0)
161 log_error(slot_errbuf, ERR_TYPE_RTAS_LOG, 0);
162 spin_unlock_irqrestore(&slot_errbuf_lock, flags);
163}
164 123
165/** 124/**
166 * gather_pci_data - copy assorted PCI config space registers to buff 125 * eeh_gather_pci_data - Copy assorted PCI config space registers to buff
167 * @pdn: device to report data for 126 * @edev: device to report data for
168 * @buf: point to buffer in which to log 127 * @buf: point to buffer in which to log
169 * @len: amount of room in buffer 128 * @len: amount of room in buffer
170 * 129 *
171 * This routine captures assorted PCI configuration space data, 130 * This routine captures assorted PCI configuration space data,
172 * and puts them into a buffer for RTAS error logging. 131 * and puts them into a buffer for RTAS error logging.
173 */ 132 */
174static size_t gather_pci_data(struct pci_dn *pdn, char * buf, size_t len) 133static size_t eeh_gather_pci_data(struct eeh_dev *edev, char * buf, size_t len)
175{ 134{
176 struct pci_dev *dev = pdn->pcidev; 135 struct device_node *dn = eeh_dev_to_of_node(edev);
136 struct pci_dev *dev = eeh_dev_to_pci_dev(edev);
177 u32 cfg; 137 u32 cfg;
178 int cap, i; 138 int cap, i;
179 int n = 0; 139 int n = 0;
180 140
181 n += scnprintf(buf+n, len-n, "%s\n", pdn->node->full_name); 141 n += scnprintf(buf+n, len-n, "%s\n", dn->full_name);
182 printk(KERN_WARNING "EEH: of node=%s\n", pdn->node->full_name); 142 printk(KERN_WARNING "EEH: of node=%s\n", dn->full_name);
183 143
184 rtas_read_config(pdn, PCI_VENDOR_ID, 4, &cfg); 144 eeh_ops->read_config(dn, PCI_VENDOR_ID, 4, &cfg);
185 n += scnprintf(buf+n, len-n, "dev/vend:%08x\n", cfg); 145 n += scnprintf(buf+n, len-n, "dev/vend:%08x\n", cfg);
186 printk(KERN_WARNING "EEH: PCI device/vendor: %08x\n", cfg); 146 printk(KERN_WARNING "EEH: PCI device/vendor: %08x\n", cfg);
187 147
188 rtas_read_config(pdn, PCI_COMMAND, 4, &cfg); 148 eeh_ops->read_config(dn, PCI_COMMAND, 4, &cfg);
189 n += scnprintf(buf+n, len-n, "cmd/stat:%x\n", cfg); 149 n += scnprintf(buf+n, len-n, "cmd/stat:%x\n", cfg);
190 printk(KERN_WARNING "EEH: PCI cmd/status register: %08x\n", cfg); 150 printk(KERN_WARNING "EEH: PCI cmd/status register: %08x\n", cfg);
191 151
@@ -196,11 +156,11 @@ static size_t gather_pci_data(struct pci_dn *pdn, char * buf, size_t len)
196 156
197 /* Gather bridge-specific registers */ 157 /* Gather bridge-specific registers */
198 if (dev->class >> 16 == PCI_BASE_CLASS_BRIDGE) { 158 if (dev->class >> 16 == PCI_BASE_CLASS_BRIDGE) {
199 rtas_read_config(pdn, PCI_SEC_STATUS, 2, &cfg); 159 eeh_ops->read_config(dn, PCI_SEC_STATUS, 2, &cfg);
200 n += scnprintf(buf+n, len-n, "sec stat:%x\n", cfg); 160 n += scnprintf(buf+n, len-n, "sec stat:%x\n", cfg);
201 printk(KERN_WARNING "EEH: Bridge secondary status: %04x\n", cfg); 161 printk(KERN_WARNING "EEH: Bridge secondary status: %04x\n", cfg);
202 162
203 rtas_read_config(pdn, PCI_BRIDGE_CONTROL, 2, &cfg); 163 eeh_ops->read_config(dn, PCI_BRIDGE_CONTROL, 2, &cfg);
204 n += scnprintf(buf+n, len-n, "brdg ctl:%x\n", cfg); 164 n += scnprintf(buf+n, len-n, "brdg ctl:%x\n", cfg);
205 printk(KERN_WARNING "EEH: Bridge control: %04x\n", cfg); 165 printk(KERN_WARNING "EEH: Bridge control: %04x\n", cfg);
206 } 166 }
@@ -208,11 +168,11 @@ static size_t gather_pci_data(struct pci_dn *pdn, char * buf, size_t len)
208 /* Dump out the PCI-X command and status regs */ 168 /* Dump out the PCI-X command and status regs */
209 cap = pci_find_capability(dev, PCI_CAP_ID_PCIX); 169 cap = pci_find_capability(dev, PCI_CAP_ID_PCIX);
210 if (cap) { 170 if (cap) {
211 rtas_read_config(pdn, cap, 4, &cfg); 171 eeh_ops->read_config(dn, cap, 4, &cfg);
212 n += scnprintf(buf+n, len-n, "pcix-cmd:%x\n", cfg); 172 n += scnprintf(buf+n, len-n, "pcix-cmd:%x\n", cfg);
213 printk(KERN_WARNING "EEH: PCI-X cmd: %08x\n", cfg); 173 printk(KERN_WARNING "EEH: PCI-X cmd: %08x\n", cfg);
214 174
215 rtas_read_config(pdn, cap+4, 4, &cfg); 175 eeh_ops->read_config(dn, cap+4, 4, &cfg);
216 n += scnprintf(buf+n, len-n, "pcix-stat:%x\n", cfg); 176 n += scnprintf(buf+n, len-n, "pcix-stat:%x\n", cfg);
217 printk(KERN_WARNING "EEH: PCI-X status: %08x\n", cfg); 177 printk(KERN_WARNING "EEH: PCI-X status: %08x\n", cfg);
218 } 178 }
@@ -225,7 +185,7 @@ static size_t gather_pci_data(struct pci_dn *pdn, char * buf, size_t len)
225 "EEH: PCI-E capabilities and status follow:\n"); 185 "EEH: PCI-E capabilities and status follow:\n");
226 186
227 for (i=0; i<=8; i++) { 187 for (i=0; i<=8; i++) {
228 rtas_read_config(pdn, cap+4*i, 4, &cfg); 188 eeh_ops->read_config(dn, cap+4*i, 4, &cfg);
229 n += scnprintf(buf+n, len-n, "%02x:%x\n", 4*i, cfg); 189 n += scnprintf(buf+n, len-n, "%02x:%x\n", 4*i, cfg);
230 printk(KERN_WARNING "EEH: PCI-E %02x: %08x\n", i, cfg); 190 printk(KERN_WARNING "EEH: PCI-E %02x: %08x\n", i, cfg);
231 } 191 }
@@ -237,7 +197,7 @@ static size_t gather_pci_data(struct pci_dn *pdn, char * buf, size_t len)
237 "EEH: PCI-E AER capability register set follows:\n"); 197 "EEH: PCI-E AER capability register set follows:\n");
238 198
239 for (i=0; i<14; i++) { 199 for (i=0; i<14; i++) {
240 rtas_read_config(pdn, cap+4*i, 4, &cfg); 200 eeh_ops->read_config(dn, cap+4*i, 4, &cfg);
241 n += scnprintf(buf+n, len-n, "%02x:%x\n", 4*i, cfg); 201 n += scnprintf(buf+n, len-n, "%02x:%x\n", 4*i, cfg);
242 printk(KERN_WARNING "EEH: PCI-E AER %02x: %08x\n", i, cfg); 202 printk(KERN_WARNING "EEH: PCI-E AER %02x: %08x\n", i, cfg);
243 } 203 }
@@ -246,111 +206,46 @@ static size_t gather_pci_data(struct pci_dn *pdn, char * buf, size_t len)
246 206
247 /* Gather status on devices under the bridge */ 207 /* Gather status on devices under the bridge */
248 if (dev->class >> 16 == PCI_BASE_CLASS_BRIDGE) { 208 if (dev->class >> 16 == PCI_BASE_CLASS_BRIDGE) {
249 struct device_node *dn; 209 struct device_node *child;
250 210
251 for_each_child_of_node(pdn->node, dn) { 211 for_each_child_of_node(dn, child) {
252 pdn = PCI_DN(dn); 212 if (of_node_to_eeh_dev(child))
253 if (pdn) 213 n += eeh_gather_pci_data(of_node_to_eeh_dev(child), buf+n, len-n);
254 n += gather_pci_data(pdn, buf+n, len-n);
255 } 214 }
256 } 215 }
257 216
258 return n; 217 return n;
259} 218}
260 219
261void eeh_slot_error_detail(struct pci_dn *pdn, int severity)
262{
263 size_t loglen = 0;
264 pci_regs_buf[0] = 0;
265
266 rtas_pci_enable(pdn, EEH_THAW_MMIO);
267 rtas_configure_bridge(pdn);
268 eeh_restore_bars(pdn);
269 loglen = gather_pci_data(pdn, pci_regs_buf, EEH_PCI_REGS_LOG_LEN);
270
271 rtas_slot_error_detail(pdn, severity, pci_regs_buf, loglen);
272}
273
274/** 220/**
275 * read_slot_reset_state - Read the reset state of a device node's slot 221 * eeh_slot_error_detail - Generate combined log including driver log and error log
276 * @dn: device node to read 222 * @edev: device to report error log for
277 * @rets: array to return results in 223 * @severity: temporary or permanent error log
278 */
279static int read_slot_reset_state(struct pci_dn *pdn, int rets[])
280{
281 int token, outputs;
282 int config_addr;
283
284 if (ibm_read_slot_reset_state2 != RTAS_UNKNOWN_SERVICE) {
285 token = ibm_read_slot_reset_state2;
286 outputs = 4;
287 } else {
288 token = ibm_read_slot_reset_state;
289 rets[2] = 0; /* fake PE Unavailable info */
290 outputs = 3;
291 }
292
293 /* Use PE configuration address, if present */
294 config_addr = pdn->eeh_config_addr;
295 if (pdn->eeh_pe_config_addr)
296 config_addr = pdn->eeh_pe_config_addr;
297
298 return rtas_call(token, 3, outputs, rets, config_addr,
299 BUID_HI(pdn->phb->buid), BUID_LO(pdn->phb->buid));
300}
301
302/**
303 * eeh_wait_for_slot_status - returns error status of slot
304 * @pdn pci device node
305 * @max_wait_msecs maximum number to millisecs to wait
306 *
307 * Return negative value if a permanent error, else return
308 * Partition Endpoint (PE) status value.
309 * 224 *
310 * If @max_wait_msecs is positive, then this routine will 225 * This routine should be called to generate the combined log, which
311 * sleep until a valid status can be obtained, or until 226 * is comprised of driver log and error log. The driver log is figured
312 * the max allowed wait time is exceeded, in which case 227 * out from the config space of the corresponding PCI device, while
313 * a -2 is returned. 228 * the error log is fetched through platform dependent function call.
314 */ 229 */
315int 230void eeh_slot_error_detail(struct eeh_dev *edev, int severity)
316eeh_wait_for_slot_status(struct pci_dn *pdn, int max_wait_msecs)
317{ 231{
318 int rc; 232 size_t loglen = 0;
319 int rets[3]; 233 pci_regs_buf[0] = 0;
320 int mwait;
321
322 while (1) {
323 rc = read_slot_reset_state(pdn, rets);
324 if (rc) return rc;
325 if (rets[1] == 0) return -1; /* EEH is not supported */
326
327 if (rets[0] != 5) return rets[0]; /* return actual status */
328
329 if (rets[2] == 0) return -1; /* permanently unavailable */
330 234
331 if (max_wait_msecs <= 0) break; 235 eeh_pci_enable(edev, EEH_OPT_THAW_MMIO);
236 eeh_ops->configure_bridge(eeh_dev_to_of_node(edev));
237 eeh_restore_bars(edev);
238 loglen = eeh_gather_pci_data(edev, pci_regs_buf, EEH_PCI_REGS_LOG_LEN);
332 239
333 mwait = rets[2]; 240 eeh_ops->get_log(eeh_dev_to_of_node(edev), severity, pci_regs_buf, loglen);
334 if (mwait <= 0) {
335 printk (KERN_WARNING
336 "EEH: Firmware returned bad wait value=%d\n", mwait);
337 mwait = 1000;
338 } else if (mwait > 300*1000) {
339 printk (KERN_WARNING
340 "EEH: Firmware is taking too long, time=%d\n", mwait);
341 mwait = 300*1000;
342 }
343 max_wait_msecs -= mwait;
344 msleep (mwait);
345 }
346
347 printk(KERN_WARNING "EEH: Timed out waiting for slot status\n");
348 return -2;
349} 241}
350 242
351/** 243/**
352 * eeh_token_to_phys - convert EEH address token to phys address 244 * eeh_token_to_phys - Convert EEH address token to phys address
353 * @token i/o token, should be address in the form 0xA.... 245 * @token: I/O token, should be address in the form 0xA....
246 *
247 * This routine should be called to convert virtual I/O address
248 * to physical one.
354 */ 249 */
355static inline unsigned long eeh_token_to_phys(unsigned long token) 250static inline unsigned long eeh_token_to_phys(unsigned long token)
356{ 251{
@@ -365,36 +260,43 @@ static inline unsigned long eeh_token_to_phys(unsigned long token)
365 return pa | (token & (PAGE_SIZE-1)); 260 return pa | (token & (PAGE_SIZE-1));
366} 261}
367 262
368/** 263/**
369 * Return the "partitionable endpoint" (pe) under which this device lies 264 * eeh_find_device_pe - Retrieve the PE for the given device
265 * @dn: device node
266 *
267 * Return the PE under which this device lies
370 */ 268 */
371struct device_node * find_device_pe(struct device_node *dn) 269struct device_node *eeh_find_device_pe(struct device_node *dn)
372{ 270{
373 while ((dn->parent) && PCI_DN(dn->parent) && 271 while (dn->parent && of_node_to_eeh_dev(dn->parent) &&
374 (PCI_DN(dn->parent)->eeh_mode & EEH_MODE_SUPPORTED)) { 272 (of_node_to_eeh_dev(dn->parent)->mode & EEH_MODE_SUPPORTED)) {
375 dn = dn->parent; 273 dn = dn->parent;
376 } 274 }
377 return dn; 275 return dn;
378} 276}
379 277
380/** Mark all devices that are children of this device as failed. 278/**
381 * Mark the device driver too, so that it can see the failure 279 * __eeh_mark_slot - Mark all child devices as failed
382 * immediately; this is critical, since some drivers poll 280 * @parent: parent device
383 * status registers in interrupts ... If a driver is polling, 281 * @mode_flag: failure flag
384 * and the slot is frozen, then the driver can deadlock in 282 *
385 * an interrupt context, which is bad. 283 * Mark all devices that are children of this device as failed.
284 * Mark the device driver too, so that it can see the failure
285 * immediately; this is critical, since some drivers poll
286 * status registers in interrupts ... If a driver is polling,
287 * and the slot is frozen, then the driver can deadlock in
288 * an interrupt context, which is bad.
386 */ 289 */
387
388static void __eeh_mark_slot(struct device_node *parent, int mode_flag) 290static void __eeh_mark_slot(struct device_node *parent, int mode_flag)
389{ 291{
390 struct device_node *dn; 292 struct device_node *dn;
391 293
392 for_each_child_of_node(parent, dn) { 294 for_each_child_of_node(parent, dn) {
393 if (PCI_DN(dn)) { 295 if (of_node_to_eeh_dev(dn)) {
394 /* Mark the pci device driver too */ 296 /* Mark the pci device driver too */
395 struct pci_dev *dev = PCI_DN(dn)->pcidev; 297 struct pci_dev *dev = of_node_to_eeh_dev(dn)->pdev;
396 298
397 PCI_DN(dn)->eeh_mode |= mode_flag; 299 of_node_to_eeh_dev(dn)->mode |= mode_flag;
398 300
399 if (dev && dev->driver) 301 if (dev && dev->driver)
400 dev->error_state = pci_channel_io_frozen; 302 dev->error_state = pci_channel_io_frozen;
@@ -404,92 +306,81 @@ static void __eeh_mark_slot(struct device_node *parent, int mode_flag)
404 } 306 }
405} 307}
406 308
407void eeh_mark_slot (struct device_node *dn, int mode_flag) 309/**
310 * eeh_mark_slot - Mark the indicated device and its children as failed
311 * @dn: parent device
312 * @mode_flag: failure flag
313 *
314 * Mark the indicated device and its child devices as failed.
315 * The device drivers are marked as failed as well.
316 */
317void eeh_mark_slot(struct device_node *dn, int mode_flag)
408{ 318{
409 struct pci_dev *dev; 319 struct pci_dev *dev;
410 dn = find_device_pe (dn); 320 dn = eeh_find_device_pe(dn);
411 321
412 /* Back up one, since config addrs might be shared */ 322 /* Back up one, since config addrs might be shared */
413 if (!pcibios_find_pci_bus(dn) && PCI_DN(dn->parent)) 323 if (!pcibios_find_pci_bus(dn) && of_node_to_eeh_dev(dn->parent))
414 dn = dn->parent; 324 dn = dn->parent;
415 325
416 PCI_DN(dn)->eeh_mode |= mode_flag; 326 of_node_to_eeh_dev(dn)->mode |= mode_flag;
417 327
418 /* Mark the pci device too */ 328 /* Mark the pci device too */
419 dev = PCI_DN(dn)->pcidev; 329 dev = of_node_to_eeh_dev(dn)->pdev;
420 if (dev) 330 if (dev)
421 dev->error_state = pci_channel_io_frozen; 331 dev->error_state = pci_channel_io_frozen;
422 332
423 __eeh_mark_slot(dn, mode_flag); 333 __eeh_mark_slot(dn, mode_flag);
424} 334}
425 335
336/**
337 * __eeh_clear_slot - Clear failure flag for the child devices
338 * @parent: parent device
339 * @mode_flag: flag to be cleared
340 *
341 * Clear failure flag for the child devices.
342 */
426static void __eeh_clear_slot(struct device_node *parent, int mode_flag) 343static void __eeh_clear_slot(struct device_node *parent, int mode_flag)
427{ 344{
428 struct device_node *dn; 345 struct device_node *dn;
429 346
430 for_each_child_of_node(parent, dn) { 347 for_each_child_of_node(parent, dn) {
431 if (PCI_DN(dn)) { 348 if (of_node_to_eeh_dev(dn)) {
432 PCI_DN(dn)->eeh_mode &= ~mode_flag; 349 of_node_to_eeh_dev(dn)->mode &= ~mode_flag;
433 PCI_DN(dn)->eeh_check_count = 0; 350 of_node_to_eeh_dev(dn)->check_count = 0;
434 __eeh_clear_slot(dn, mode_flag); 351 __eeh_clear_slot(dn, mode_flag);
435 } 352 }
436 } 353 }
437} 354}
438 355
439void eeh_clear_slot (struct device_node *dn, int mode_flag) 356/**
357 * eeh_clear_slot - Clear failure flag for the indicated device and its children
358 * @dn: parent device
359 * @mode_flag: flag to be cleared
360 *
361 * Clear failure flag for the indicated device and its children.
362 */
363void eeh_clear_slot(struct device_node *dn, int mode_flag)
440{ 364{
441 unsigned long flags; 365 unsigned long flags;
442 raw_spin_lock_irqsave(&confirm_error_lock, flags); 366 raw_spin_lock_irqsave(&confirm_error_lock, flags);
443 367
444 dn = find_device_pe (dn); 368 dn = eeh_find_device_pe(dn);
445 369
446 /* Back up one, since config addrs might be shared */ 370 /* Back up one, since config addrs might be shared */
447 if (!pcibios_find_pci_bus(dn) && PCI_DN(dn->parent)) 371 if (!pcibios_find_pci_bus(dn) && of_node_to_eeh_dev(dn->parent))
448 dn = dn->parent; 372 dn = dn->parent;
449 373
450 PCI_DN(dn)->eeh_mode &= ~mode_flag; 374 of_node_to_eeh_dev(dn)->mode &= ~mode_flag;
451 PCI_DN(dn)->eeh_check_count = 0; 375 of_node_to_eeh_dev(dn)->check_count = 0;
452 __eeh_clear_slot(dn, mode_flag); 376 __eeh_clear_slot(dn, mode_flag);
453 raw_spin_unlock_irqrestore(&confirm_error_lock, flags); 377 raw_spin_unlock_irqrestore(&confirm_error_lock, flags);
454} 378}
455 379
456void __eeh_set_pe_freset(struct device_node *parent, unsigned int *freset)
457{
458 struct device_node *dn;
459
460 for_each_child_of_node(parent, dn) {
461 if (PCI_DN(dn)) {
462
463 struct pci_dev *dev = PCI_DN(dn)->pcidev;
464
465 if (dev && dev->driver)
466 *freset |= dev->needs_freset;
467
468 __eeh_set_pe_freset(dn, freset);
469 }
470 }
471}
472
473void eeh_set_pe_freset(struct device_node *dn, unsigned int *freset)
474{
475 struct pci_dev *dev;
476 dn = find_device_pe(dn);
477
478 /* Back up one, since config addrs might be shared */
479 if (!pcibios_find_pci_bus(dn) && PCI_DN(dn->parent))
480 dn = dn->parent;
481
482 dev = PCI_DN(dn)->pcidev;
483 if (dev)
484 *freset |= dev->needs_freset;
485
486 __eeh_set_pe_freset(dn, freset);
487}
488
489/** 380/**
490 * eeh_dn_check_failure - check if all 1's data is due to EEH slot freeze 381 * eeh_dn_check_failure - Check if all 1's data is due to EEH slot freeze
491 * @dn device node 382 * @dn: device node
492 * @dev pci device, if known 383 * @dev: pci device, if known
493 * 384 *
494 * Check for an EEH failure for the given device node. Call this 385 * Check for an EEH failure for the given device node. Call this
495 * routine if the result of a read was all 0xff's and you want to 386 * routine if the result of a read was all 0xff's and you want to
@@ -504,35 +395,34 @@ void eeh_set_pe_freset(struct device_node *dn, unsigned int *freset)
504int eeh_dn_check_failure(struct device_node *dn, struct pci_dev *dev) 395int eeh_dn_check_failure(struct device_node *dn, struct pci_dev *dev)
505{ 396{
506 int ret; 397 int ret;
507 int rets[3];
508 unsigned long flags; 398 unsigned long flags;
509 struct pci_dn *pdn; 399 struct eeh_dev *edev;
510 int rc = 0; 400 int rc = 0;
511 const char *location; 401 const char *location;
512 402
513 total_mmio_ffs++; 403 eeh_stats.total_mmio_ffs++;
514 404
515 if (!eeh_subsystem_enabled) 405 if (!eeh_subsystem_enabled)
516 return 0; 406 return 0;
517 407
518 if (!dn) { 408 if (!dn) {
519 no_dn++; 409 eeh_stats.no_dn++;
520 return 0; 410 return 0;
521 } 411 }
522 dn = find_device_pe(dn); 412 dn = eeh_find_device_pe(dn);
523 pdn = PCI_DN(dn); 413 edev = of_node_to_eeh_dev(dn);
524 414
525 /* Access to IO BARs might get this far and still not want checking. */ 415 /* Access to IO BARs might get this far and still not want checking. */
526 if (!(pdn->eeh_mode & EEH_MODE_SUPPORTED) || 416 if (!(edev->mode & EEH_MODE_SUPPORTED) ||
527 pdn->eeh_mode & EEH_MODE_NOCHECK) { 417 edev->mode & EEH_MODE_NOCHECK) {
528 ignored_check++; 418 eeh_stats.ignored_check++;
529 pr_debug("EEH: Ignored check (%x) for %s %s\n", 419 pr_debug("EEH: Ignored check (%x) for %s %s\n",
530 pdn->eeh_mode, eeh_pci_name(dev), dn->full_name); 420 edev->mode, eeh_pci_name(dev), dn->full_name);
531 return 0; 421 return 0;
532 } 422 }
533 423
534 if (!pdn->eeh_config_addr && !pdn->eeh_pe_config_addr) { 424 if (!edev->config_addr && !edev->pe_config_addr) {
535 no_cfg_addr++; 425 eeh_stats.no_cfg_addr++;
536 return 0; 426 return 0;
537 } 427 }
538 428
@@ -544,15 +434,15 @@ int eeh_dn_check_failure(struct device_node *dn, struct pci_dev *dev)
544 */ 434 */
545 raw_spin_lock_irqsave(&confirm_error_lock, flags); 435 raw_spin_lock_irqsave(&confirm_error_lock, flags);
546 rc = 1; 436 rc = 1;
547 if (pdn->eeh_mode & EEH_MODE_ISOLATED) { 437 if (edev->mode & EEH_MODE_ISOLATED) {
548 pdn->eeh_check_count ++; 438 edev->check_count++;
549 if (pdn->eeh_check_count % EEH_MAX_FAILS == 0) { 439 if (edev->check_count % EEH_MAX_FAILS == 0) {
550 location = of_get_property(dn, "ibm,loc-code", NULL); 440 location = of_get_property(dn, "ibm,loc-code", NULL);
551 printk (KERN_ERR "EEH: %d reads ignored for recovering device at " 441 printk(KERN_ERR "EEH: %d reads ignored for recovering device at "
552 "location=%s driver=%s pci addr=%s\n", 442 "location=%s driver=%s pci addr=%s\n",
553 pdn->eeh_check_count, location, 443 edev->check_count, location,
554 eeh_driver_name(dev), eeh_pci_name(dev)); 444 eeh_driver_name(dev), eeh_pci_name(dev));
555 printk (KERN_ERR "EEH: Might be infinite loop in %s driver\n", 445 printk(KERN_ERR "EEH: Might be infinite loop in %s driver\n",
556 eeh_driver_name(dev)); 446 eeh_driver_name(dev));
557 dump_stack(); 447 dump_stack();
558 } 448 }
@@ -566,58 +456,39 @@ int eeh_dn_check_failure(struct device_node *dn, struct pci_dev *dev)
566 * function zero of a multi-function device. 456 * function zero of a multi-function device.
567 * In any case they must share a common PHB. 457 * In any case they must share a common PHB.
568 */ 458 */
569 ret = read_slot_reset_state(pdn, rets); 459 ret = eeh_ops->get_state(dn, NULL);
570
571 /* If the call to firmware failed, punt */
572 if (ret != 0) {
573 printk(KERN_WARNING "EEH: read_slot_reset_state() failed; rc=%d dn=%s\n",
574 ret, dn->full_name);
575 false_positives++;
576 pdn->eeh_false_positives ++;
577 rc = 0;
578 goto dn_unlock;
579 }
580 460
581 /* Note that config-io to empty slots may fail; 461 /* Note that config-io to empty slots may fail;
582 * they are empty when they don't have children. */ 462 * they are empty when they don't have children.
583 if ((rets[0] == 5) && (rets[2] == 0) && (dn->child == NULL)) { 463 * We will punt with the following conditions: Failure to get
584 false_positives++; 464 * PE's state, EEH not support and Permanently unavailable
585 pdn->eeh_false_positives ++; 465 * state, PE is in good state.
586 rc = 0; 466 */
587 goto dn_unlock; 467 if ((ret < 0) ||
588 } 468 (ret == EEH_STATE_NOT_SUPPORT) ||
589 469 (ret & (EEH_STATE_MMIO_ACTIVE | EEH_STATE_DMA_ACTIVE)) ==
590 /* If EEH is not supported on this device, punt. */ 470 (EEH_STATE_MMIO_ACTIVE | EEH_STATE_DMA_ACTIVE)) {
591 if (rets[1] != 1) { 471 eeh_stats.false_positives++;
592 printk(KERN_WARNING "EEH: event on unsupported device, rc=%d dn=%s\n", 472 edev->false_positives ++;
593 ret, dn->full_name);
594 false_positives++;
595 pdn->eeh_false_positives ++;
596 rc = 0;
597 goto dn_unlock;
598 }
599
600 /* If not the kind of error we know about, punt. */
601 if (rets[0] != 1 && rets[0] != 2 && rets[0] != 4 && rets[0] != 5) {
602 false_positives++;
603 pdn->eeh_false_positives ++;
604 rc = 0; 473 rc = 0;
605 goto dn_unlock; 474 goto dn_unlock;
606 } 475 }
607 476
608 slot_resets++; 477 eeh_stats.slot_resets++;
609 478
610 /* Avoid repeated reports of this failure, including problems 479 /* Avoid repeated reports of this failure, including problems
611 * with other functions on this device, and functions under 480 * with other functions on this device, and functions under
612 * bridges. */ 481 * bridges.
613 eeh_mark_slot (dn, EEH_MODE_ISOLATED); 482 */
483 eeh_mark_slot(dn, EEH_MODE_ISOLATED);
614 raw_spin_unlock_irqrestore(&confirm_error_lock, flags); 484 raw_spin_unlock_irqrestore(&confirm_error_lock, flags);
615 485
616 eeh_send_failure_event (dn, dev); 486 eeh_send_failure_event(edev);
617 487
618 /* Most EEH events are due to device driver bugs. Having 488 /* Most EEH events are due to device driver bugs. Having
619 * a stack trace will help the device-driver authors figure 489 * a stack trace will help the device-driver authors figure
620 * out what happened. So print that out. */ 490 * out what happened. So print that out.
491 */
621 dump_stack(); 492 dump_stack();
622 return 1; 493 return 1;
623 494
@@ -629,9 +500,9 @@ dn_unlock:
629EXPORT_SYMBOL_GPL(eeh_dn_check_failure); 500EXPORT_SYMBOL_GPL(eeh_dn_check_failure);
630 501
631/** 502/**
632 * eeh_check_failure - check if all 1's data is due to EEH slot freeze 503 * eeh_check_failure - Check if all 1's data is due to EEH slot freeze
633 * @token i/o token, should be address in the form 0xA.... 504 * @token: I/O token, should be address in the form 0xA....
634 * @val value, should be all 1's (XXX why do we need this arg??) 505 * @val: value, should be all 1's (XXX why do we need this arg??)
635 * 506 *
636 * Check for an EEH failure at the given token address. Call this 507 * Check for an EEH failure at the given token address. Call this
637 * routine if the result of a read was all 0xff's and you want to 508 * routine if the result of a read was all 0xff's and you want to
@@ -648,14 +519,14 @@ unsigned long eeh_check_failure(const volatile void __iomem *token, unsigned lon
648 519
649 /* Finding the phys addr + pci device; this is pretty quick. */ 520 /* Finding the phys addr + pci device; this is pretty quick. */
650 addr = eeh_token_to_phys((unsigned long __force) token); 521 addr = eeh_token_to_phys((unsigned long __force) token);
651 dev = pci_get_device_by_addr(addr); 522 dev = pci_addr_cache_get_device(addr);
652 if (!dev) { 523 if (!dev) {
653 no_device++; 524 eeh_stats.no_device++;
654 return val; 525 return val;
655 } 526 }
656 527
657 dn = pci_device_to_OF_node(dev); 528 dn = pci_device_to_OF_node(dev);
658 eeh_dn_check_failure (dn, dev); 529 eeh_dn_check_failure(dn, dev);
659 530
660 pci_dev_put(dev); 531 pci_dev_put(dev);
661 return val; 532 return val;
@@ -663,115 +534,54 @@ unsigned long eeh_check_failure(const volatile void __iomem *token, unsigned lon
663 534
664EXPORT_SYMBOL(eeh_check_failure); 535EXPORT_SYMBOL(eeh_check_failure);
665 536
666/* ------------------------------------------------------------- */
667/* The code below deals with error recovery */
668 537
669/** 538/**
670 * rtas_pci_enable - enable MMIO or DMA transfers for this slot 539 * eeh_pci_enable - Enable MMIO or DMA transfers for this slot
671 * @pdn pci device node 540 * @edev: pci device node
541 *
542 * This routine should be called to reenable frozen MMIO or DMA
543 * so that it would work correctly again. It's useful while doing
544 * recovery or log collection on the indicated device.
672 */ 545 */
673 546int eeh_pci_enable(struct eeh_dev *edev, int function)
674int
675rtas_pci_enable(struct pci_dn *pdn, int function)
676{ 547{
677 int config_addr;
678 int rc; 548 int rc;
549 struct device_node *dn = eeh_dev_to_of_node(edev);
679 550
680 /* Use PE configuration address, if present */ 551 rc = eeh_ops->set_option(dn, function);
681 config_addr = pdn->eeh_config_addr;
682 if (pdn->eeh_pe_config_addr)
683 config_addr = pdn->eeh_pe_config_addr;
684
685 rc = rtas_call(ibm_set_eeh_option, 4, 1, NULL,
686 config_addr,
687 BUID_HI(pdn->phb->buid),
688 BUID_LO(pdn->phb->buid),
689 function);
690
691 if (rc) 552 if (rc)
692 printk(KERN_WARNING "EEH: Unexpected state change %d, err=%d dn=%s\n", 553 printk(KERN_WARNING "EEH: Unexpected state change %d, err=%d dn=%s\n",
693 function, rc, pdn->node->full_name); 554 function, rc, dn->full_name);
694 555
695 rc = eeh_wait_for_slot_status (pdn, PCI_BUS_RESET_WAIT_MSEC); 556 rc = eeh_ops->wait_state(dn, PCI_BUS_RESET_WAIT_MSEC);
696 if ((rc == 4) && (function == EEH_THAW_MMIO)) 557 if (rc > 0 && (rc & EEH_STATE_MMIO_ENABLED) &&
558 (function == EEH_OPT_THAW_MMIO))
697 return 0; 559 return 0;
698 560
699 return rc; 561 return rc;
700} 562}
701 563
702/** 564/**
703 * rtas_pci_slot_reset - raises/lowers the pci #RST line
704 * @pdn pci device node
705 * @state: 1/0 to raise/lower the #RST
706 *
707 * Clear the EEH-frozen condition on a slot. This routine
708 * asserts the PCI #RST line if the 'state' argument is '1',
709 * and drops the #RST line if 'state is '0'. This routine is
710 * safe to call in an interrupt context.
711 *
712 */
713
714static void
715rtas_pci_slot_reset(struct pci_dn *pdn, int state)
716{
717 int config_addr;
718 int rc;
719
720 BUG_ON (pdn==NULL);
721
722 if (!pdn->phb) {
723 printk (KERN_WARNING "EEH: in slot reset, device node %s has no phb\n",
724 pdn->node->full_name);
725 return;
726 }
727
728 /* Use PE configuration address, if present */
729 config_addr = pdn->eeh_config_addr;
730 if (pdn->eeh_pe_config_addr)
731 config_addr = pdn->eeh_pe_config_addr;
732
733 rc = rtas_call(ibm_set_slot_reset, 4, 1, NULL,
734 config_addr,
735 BUID_HI(pdn->phb->buid),
736 BUID_LO(pdn->phb->buid),
737 state);
738
739 /* Fundamental-reset not supported on this PE, try hot-reset */
740 if (rc == -8 && state == 3) {
741 rc = rtas_call(ibm_set_slot_reset, 4, 1, NULL,
742 config_addr,
743 BUID_HI(pdn->phb->buid),
744 BUID_LO(pdn->phb->buid), 1);
745 if (rc)
746 printk(KERN_WARNING
747 "EEH: Unable to reset the failed slot,"
748 " #RST=%d dn=%s\n",
749 rc, pdn->node->full_name);
750 }
751}
752
753/**
754 * pcibios_set_pcie_slot_reset - Set PCI-E reset state 565 * pcibios_set_pcie_slot_reset - Set PCI-E reset state
755 * @dev: pci device struct 566 * @dev: pci device struct
756 * @state: reset state to enter 567 * @state: reset state to enter
757 * 568 *
758 * Return value: 569 * Return value:
759 * 0 if success 570 * 0 if success
760 **/ 571 */
761int pcibios_set_pcie_reset_state(struct pci_dev *dev, enum pcie_reset_state state) 572int pcibios_set_pcie_reset_state(struct pci_dev *dev, enum pcie_reset_state state)
762{ 573{
763 struct device_node *dn = pci_device_to_OF_node(dev); 574 struct device_node *dn = pci_device_to_OF_node(dev);
764 struct pci_dn *pdn = PCI_DN(dn);
765 575
766 switch (state) { 576 switch (state) {
767 case pcie_deassert_reset: 577 case pcie_deassert_reset:
768 rtas_pci_slot_reset(pdn, 0); 578 eeh_ops->reset(dn, EEH_RESET_DEACTIVATE);
769 break; 579 break;
770 case pcie_hot_reset: 580 case pcie_hot_reset:
771 rtas_pci_slot_reset(pdn, 1); 581 eeh_ops->reset(dn, EEH_RESET_HOT);
772 break; 582 break;
773 case pcie_warm_reset: 583 case pcie_warm_reset:
774 rtas_pci_slot_reset(pdn, 3); 584 eeh_ops->reset(dn, EEH_RESET_FUNDAMENTAL);
775 break; 585 break;
776 default: 586 default:
777 return -EINVAL; 587 return -EINVAL;
@@ -781,13 +591,66 @@ int pcibios_set_pcie_reset_state(struct pci_dev *dev, enum pcie_reset_state stat
781} 591}
782 592
783/** 593/**
784 * rtas_set_slot_reset -- assert the pci #RST line for 1/4 second 594 * __eeh_set_pe_freset - Check the required reset for child devices
785 * @pdn: pci device node to be reset. 595 * @parent: parent device
596 * @freset: return value
597 *
598 * Each device might have its preferred reset type: fundamental or
599 * hot reset. The routine is used to collect the information from
600 * the child devices so that they could be reset accordingly.
601 */
602void __eeh_set_pe_freset(struct device_node *parent, unsigned int *freset)
603{
604 struct device_node *dn;
605
606 for_each_child_of_node(parent, dn) {
607 if (of_node_to_eeh_dev(dn)) {
608 struct pci_dev *dev = of_node_to_eeh_dev(dn)->pdev;
609
610 if (dev && dev->driver)
611 *freset |= dev->needs_freset;
612
613 __eeh_set_pe_freset(dn, freset);
614 }
615 }
616}
617
618/**
619 * eeh_set_pe_freset - Check the required reset for the indicated device and its children
620 * @dn: parent device
621 * @freset: return value
622 *
623 * Each device might have its preferred reset type: fundamental or
624 * hot reset. The routine is used to collected the information for
625 * the indicated device and its children so that the bunch of the
626 * devices could be reset properly.
786 */ 627 */
628void eeh_set_pe_freset(struct device_node *dn, unsigned int *freset)
629{
630 struct pci_dev *dev;
631 dn = eeh_find_device_pe(dn);
632
633 /* Back up one, since config addrs might be shared */
634 if (!pcibios_find_pci_bus(dn) && of_node_to_eeh_dev(dn->parent))
635 dn = dn->parent;
636
637 dev = of_node_to_eeh_dev(dn)->pdev;
638 if (dev)
639 *freset |= dev->needs_freset;
787 640
788static void __rtas_set_slot_reset(struct pci_dn *pdn) 641 __eeh_set_pe_freset(dn, freset);
642}
643
644/**
645 * eeh_reset_pe_once - Assert the pci #RST line for 1/4 second
646 * @edev: pci device node to be reset.
647 *
648 * Assert the PCI #RST line for 1/4 second.
649 */
650static void eeh_reset_pe_once(struct eeh_dev *edev)
789{ 651{
790 unsigned int freset = 0; 652 unsigned int freset = 0;
653 struct device_node *dn = eeh_dev_to_of_node(edev);
791 654
792 /* Determine type of EEH reset required for 655 /* Determine type of EEH reset required for
793 * Partitionable Endpoint, a hot-reset (1) 656 * Partitionable Endpoint, a hot-reset (1)
@@ -795,58 +658,68 @@ static void __rtas_set_slot_reset(struct pci_dn *pdn)
795 * A fundamental reset required by any device under 658 * A fundamental reset required by any device under
796 * Partitionable Endpoint trumps hot-reset. 659 * Partitionable Endpoint trumps hot-reset.
797 */ 660 */
798 eeh_set_pe_freset(pdn->node, &freset); 661 eeh_set_pe_freset(dn, &freset);
799 662
800 if (freset) 663 if (freset)
801 rtas_pci_slot_reset(pdn, 3); 664 eeh_ops->reset(dn, EEH_RESET_FUNDAMENTAL);
802 else 665 else
803 rtas_pci_slot_reset(pdn, 1); 666 eeh_ops->reset(dn, EEH_RESET_HOT);
804 667
805 /* The PCI bus requires that the reset be held high for at least 668 /* The PCI bus requires that the reset be held high for at least
806 * a 100 milliseconds. We wait a bit longer 'just in case'. */ 669 * a 100 milliseconds. We wait a bit longer 'just in case'.
807 670 */
808#define PCI_BUS_RST_HOLD_TIME_MSEC 250 671#define PCI_BUS_RST_HOLD_TIME_MSEC 250
809 msleep (PCI_BUS_RST_HOLD_TIME_MSEC); 672 msleep(PCI_BUS_RST_HOLD_TIME_MSEC);
810 673
811 /* We might get hit with another EEH freeze as soon as the 674 /* We might get hit with another EEH freeze as soon as the
812 * pci slot reset line is dropped. Make sure we don't miss 675 * pci slot reset line is dropped. Make sure we don't miss
813 * these, and clear the flag now. */ 676 * these, and clear the flag now.
814 eeh_clear_slot (pdn->node, EEH_MODE_ISOLATED); 677 */
678 eeh_clear_slot(dn, EEH_MODE_ISOLATED);
815 679
816 rtas_pci_slot_reset (pdn, 0); 680 eeh_ops->reset(dn, EEH_RESET_DEACTIVATE);
817 681
818 /* After a PCI slot has been reset, the PCI Express spec requires 682 /* After a PCI slot has been reset, the PCI Express spec requires
819 * a 1.5 second idle time for the bus to stabilize, before starting 683 * a 1.5 second idle time for the bus to stabilize, before starting
820 * up traffic. */ 684 * up traffic.
685 */
821#define PCI_BUS_SETTLE_TIME_MSEC 1800 686#define PCI_BUS_SETTLE_TIME_MSEC 1800
822 msleep (PCI_BUS_SETTLE_TIME_MSEC); 687 msleep(PCI_BUS_SETTLE_TIME_MSEC);
823} 688}
824 689
825int rtas_set_slot_reset(struct pci_dn *pdn) 690/**
691 * eeh_reset_pe - Reset the indicated PE
692 * @edev: PCI device associated EEH device
693 *
694 * This routine should be called to reset indicated device, including
695 * PE. A PE might include multiple PCI devices and sometimes PCI bridges
696 * might be involved as well.
697 */
698int eeh_reset_pe(struct eeh_dev *edev)
826{ 699{
827 int i, rc; 700 int i, rc;
701 struct device_node *dn = eeh_dev_to_of_node(edev);
828 702
829 /* Take three shots at resetting the bus */ 703 /* Take three shots at resetting the bus */
830 for (i=0; i<3; i++) { 704 for (i=0; i<3; i++) {
831 __rtas_set_slot_reset(pdn); 705 eeh_reset_pe_once(edev);
832 706
833 rc = eeh_wait_for_slot_status(pdn, PCI_BUS_RESET_WAIT_MSEC); 707 rc = eeh_ops->wait_state(dn, PCI_BUS_RESET_WAIT_MSEC);
834 if (rc == 0) 708 if (rc == (EEH_STATE_MMIO_ACTIVE | EEH_STATE_DMA_ACTIVE))
835 return 0; 709 return 0;
836 710
837 if (rc < 0) { 711 if (rc < 0) {
838 printk(KERN_ERR "EEH: unrecoverable slot failure %s\n", 712 printk(KERN_ERR "EEH: unrecoverable slot failure %s\n",
839 pdn->node->full_name); 713 dn->full_name);
840 return -1; 714 return -1;
841 } 715 }
842 printk(KERN_ERR "EEH: bus reset %d failed on slot %s, rc=%d\n", 716 printk(KERN_ERR "EEH: bus reset %d failed on slot %s, rc=%d\n",
843 i+1, pdn->node->full_name, rc); 717 i+1, dn->full_name, rc);
844 } 718 }
845 719
846 return -1; 720 return -1;
847} 721}
848 722
849/* ------------------------------------------------------- */
850/** Save and restore of PCI BARs 723/** Save and restore of PCI BARs
851 * 724 *
852 * Although firmware will set up BARs during boot, it doesn't 725 * Although firmware will set up BARs during boot, it doesn't
@@ -856,181 +729,122 @@ int rtas_set_slot_reset(struct pci_dn *pdn)
856 */ 729 */
857 730
858/** 731/**
859 * __restore_bars - Restore the Base Address Registers 732 * eeh_restore_one_device_bars - Restore the Base Address Registers for one device
860 * @pdn: pci device node 733 * @edev: PCI device associated EEH device
861 * 734 *
862 * Loads the PCI configuration space base address registers, 735 * Loads the PCI configuration space base address registers,
863 * the expansion ROM base address, the latency timer, and etc. 736 * the expansion ROM base address, the latency timer, and etc.
864 * from the saved values in the device node. 737 * from the saved values in the device node.
865 */ 738 */
866static inline void __restore_bars (struct pci_dn *pdn) 739static inline void eeh_restore_one_device_bars(struct eeh_dev *edev)
867{ 740{
868 int i; 741 int i;
869 u32 cmd; 742 u32 cmd;
743 struct device_node *dn = eeh_dev_to_of_node(edev);
744
745 if (!edev->phb)
746 return;
870 747
871 if (NULL==pdn->phb) return;
872 for (i=4; i<10; i++) { 748 for (i=4; i<10; i++) {
873 rtas_write_config(pdn, i*4, 4, pdn->config_space[i]); 749 eeh_ops->write_config(dn, i*4, 4, edev->config_space[i]);
874 } 750 }
875 751
876 /* 12 == Expansion ROM Address */ 752 /* 12 == Expansion ROM Address */
877 rtas_write_config(pdn, 12*4, 4, pdn->config_space[12]); 753 eeh_ops->write_config(dn, 12*4, 4, edev->config_space[12]);
878 754
879#define BYTE_SWAP(OFF) (8*((OFF)/4)+3-(OFF)) 755#define BYTE_SWAP(OFF) (8*((OFF)/4)+3-(OFF))
880#define SAVED_BYTE(OFF) (((u8 *)(pdn->config_space))[BYTE_SWAP(OFF)]) 756#define SAVED_BYTE(OFF) (((u8 *)(edev->config_space))[BYTE_SWAP(OFF)])
881 757
882 rtas_write_config (pdn, PCI_CACHE_LINE_SIZE, 1, 758 eeh_ops->write_config(dn, PCI_CACHE_LINE_SIZE, 1,
883 SAVED_BYTE(PCI_CACHE_LINE_SIZE)); 759 SAVED_BYTE(PCI_CACHE_LINE_SIZE));
884 760
885 rtas_write_config (pdn, PCI_LATENCY_TIMER, 1, 761 eeh_ops->write_config(dn, PCI_LATENCY_TIMER, 1,
886 SAVED_BYTE(PCI_LATENCY_TIMER)); 762 SAVED_BYTE(PCI_LATENCY_TIMER));
887 763
888 /* max latency, min grant, interrupt pin and line */ 764 /* max latency, min grant, interrupt pin and line */
889 rtas_write_config(pdn, 15*4, 4, pdn->config_space[15]); 765 eeh_ops->write_config(dn, 15*4, 4, edev->config_space[15]);
890 766
891 /* Restore PERR & SERR bits, some devices require it, 767 /* Restore PERR & SERR bits, some devices require it,
892 don't touch the other command bits */ 768 * don't touch the other command bits
893 rtas_read_config(pdn, PCI_COMMAND, 4, &cmd); 769 */
894 if (pdn->config_space[1] & PCI_COMMAND_PARITY) 770 eeh_ops->read_config(dn, PCI_COMMAND, 4, &cmd);
771 if (edev->config_space[1] & PCI_COMMAND_PARITY)
895 cmd |= PCI_COMMAND_PARITY; 772 cmd |= PCI_COMMAND_PARITY;
896 else 773 else
897 cmd &= ~PCI_COMMAND_PARITY; 774 cmd &= ~PCI_COMMAND_PARITY;
898 if (pdn->config_space[1] & PCI_COMMAND_SERR) 775 if (edev->config_space[1] & PCI_COMMAND_SERR)
899 cmd |= PCI_COMMAND_SERR; 776 cmd |= PCI_COMMAND_SERR;
900 else 777 else
901 cmd &= ~PCI_COMMAND_SERR; 778 cmd &= ~PCI_COMMAND_SERR;
902 rtas_write_config(pdn, PCI_COMMAND, 4, cmd); 779 eeh_ops->write_config(dn, PCI_COMMAND, 4, cmd);
903} 780}
904 781
905/** 782/**
906 * eeh_restore_bars - restore the PCI config space info 783 * eeh_restore_bars - Restore the PCI config space info
784 * @edev: EEH device
907 * 785 *
908 * This routine performs a recursive walk to the children 786 * This routine performs a recursive walk to the children
909 * of this device as well. 787 * of this device as well.
910 */ 788 */
911void eeh_restore_bars(struct pci_dn *pdn) 789void eeh_restore_bars(struct eeh_dev *edev)
912{ 790{
913 struct device_node *dn; 791 struct device_node *dn;
914 if (!pdn) 792 if (!edev)
915 return; 793 return;
916 794
917 if ((pdn->eeh_mode & EEH_MODE_SUPPORTED) && !IS_BRIDGE(pdn->class_code)) 795 if ((edev->mode & EEH_MODE_SUPPORTED) && !IS_BRIDGE(edev->class_code))
918 __restore_bars (pdn); 796 eeh_restore_one_device_bars(edev);
919 797
920 for_each_child_of_node(pdn->node, dn) 798 for_each_child_of_node(eeh_dev_to_of_node(edev), dn)
921 eeh_restore_bars (PCI_DN(dn)); 799 eeh_restore_bars(of_node_to_eeh_dev(dn));
922} 800}
923 801
924/** 802/**
925 * eeh_save_bars - save device bars 803 * eeh_save_bars - Save device bars
804 * @edev: PCI device associated EEH device
926 * 805 *
927 * Save the values of the device bars. Unlike the restore 806 * Save the values of the device bars. Unlike the restore
928 * routine, this routine is *not* recursive. This is because 807 * routine, this routine is *not* recursive. This is because
929 * PCI devices are added individually; but, for the restore, 808 * PCI devices are added individually; but, for the restore,
930 * an entire slot is reset at a time. 809 * an entire slot is reset at a time.
931 */ 810 */
932static void eeh_save_bars(struct pci_dn *pdn) 811static void eeh_save_bars(struct eeh_dev *edev)
933{ 812{
934 int i; 813 int i;
814 struct device_node *dn;
935 815
936 if (!pdn ) 816 if (!edev)
937 return; 817 return;
818 dn = eeh_dev_to_of_node(edev);
938 819
939 for (i = 0; i < 16; i++) 820 for (i = 0; i < 16; i++)
940 rtas_read_config(pdn, i * 4, 4, &pdn->config_space[i]); 821 eeh_ops->read_config(dn, i * 4, 4, &edev->config_space[i]);
941}
942
943void
944rtas_configure_bridge(struct pci_dn *pdn)
945{
946 int config_addr;
947 int rc;
948 int token;
949
950 /* Use PE configuration address, if present */
951 config_addr = pdn->eeh_config_addr;
952 if (pdn->eeh_pe_config_addr)
953 config_addr = pdn->eeh_pe_config_addr;
954
955 /* Use new configure-pe function, if supported */
956 if (ibm_configure_pe != RTAS_UNKNOWN_SERVICE)
957 token = ibm_configure_pe;
958 else
959 token = ibm_configure_bridge;
960
961 rc = rtas_call(token, 3, 1, NULL,
962 config_addr,
963 BUID_HI(pdn->phb->buid),
964 BUID_LO(pdn->phb->buid));
965 if (rc) {
966 printk (KERN_WARNING "EEH: Unable to configure device bridge (%d) for %s\n",
967 rc, pdn->node->full_name);
968 }
969} 822}
970 823
971/* ------------------------------------------------------------- */ 824/**
972/* The code below deals with enabling EEH for devices during the 825 * eeh_early_enable - Early enable EEH on the indicated device
973 * early boot sequence. EEH must be enabled before any PCI probing 826 * @dn: device node
974 * can be done. 827 * @data: BUID
828 *
829 * Enable EEH functionality on the specified PCI device. The function
830 * is expected to be called before real PCI probing is done. However,
831 * the PHBs have been initialized at this point.
975 */ 832 */
976 833static void *eeh_early_enable(struct device_node *dn, void *data)
977#define EEH_ENABLE 1
978
979struct eeh_early_enable_info {
980 unsigned int buid_hi;
981 unsigned int buid_lo;
982};
983
984static int get_pe_addr (int config_addr,
985 struct eeh_early_enable_info *info)
986{ 834{
987 unsigned int rets[3];
988 int ret;
989
990 /* Use latest config-addr token on power6 */
991 if (ibm_get_config_addr_info2 != RTAS_UNKNOWN_SERVICE) {
992 /* Make sure we have a PE in hand */
993 ret = rtas_call (ibm_get_config_addr_info2, 4, 2, rets,
994 config_addr, info->buid_hi, info->buid_lo, 1);
995 if (ret || (rets[0]==0))
996 return 0;
997
998 ret = rtas_call (ibm_get_config_addr_info2, 4, 2, rets,
999 config_addr, info->buid_hi, info->buid_lo, 0);
1000 if (ret)
1001 return 0;
1002 return rets[0];
1003 }
1004
1005 /* Use older config-addr token on power5 */
1006 if (ibm_get_config_addr_info != RTAS_UNKNOWN_SERVICE) {
1007 ret = rtas_call (ibm_get_config_addr_info, 4, 2, rets,
1008 config_addr, info->buid_hi, info->buid_lo, 0);
1009 if (ret)
1010 return 0;
1011 return rets[0];
1012 }
1013 return 0;
1014}
1015
1016/* Enable eeh for the given device node. */
1017static void *early_enable_eeh(struct device_node *dn, void *data)
1018{
1019 unsigned int rets[3];
1020 struct eeh_early_enable_info *info = data;
1021 int ret; 835 int ret;
1022 const u32 *class_code = of_get_property(dn, "class-code", NULL); 836 const u32 *class_code = of_get_property(dn, "class-code", NULL);
1023 const u32 *vendor_id = of_get_property(dn, "vendor-id", NULL); 837 const u32 *vendor_id = of_get_property(dn, "vendor-id", NULL);
1024 const u32 *device_id = of_get_property(dn, "device-id", NULL); 838 const u32 *device_id = of_get_property(dn, "device-id", NULL);
1025 const u32 *regs; 839 const u32 *regs;
1026 int enable; 840 int enable;
1027 struct pci_dn *pdn = PCI_DN(dn); 841 struct eeh_dev *edev = of_node_to_eeh_dev(dn);
1028 842
1029 pdn->class_code = 0; 843 edev->class_code = 0;
1030 pdn->eeh_mode = 0; 844 edev->mode = 0;
1031 pdn->eeh_check_count = 0; 845 edev->check_count = 0;
1032 pdn->eeh_freeze_count = 0; 846 edev->freeze_count = 0;
1033 pdn->eeh_false_positives = 0; 847 edev->false_positives = 0;
1034 848
1035 if (!of_device_is_available(dn)) 849 if (!of_device_is_available(dn))
1036 return NULL; 850 return NULL;
@@ -1041,54 +855,56 @@ static void *early_enable_eeh(struct device_node *dn, void *data)
1041 855
1042 /* There is nothing to check on PCI to ISA bridges */ 856 /* There is nothing to check on PCI to ISA bridges */
1043 if (dn->type && !strcmp(dn->type, "isa")) { 857 if (dn->type && !strcmp(dn->type, "isa")) {
1044 pdn->eeh_mode |= EEH_MODE_NOCHECK; 858 edev->mode |= EEH_MODE_NOCHECK;
1045 return NULL; 859 return NULL;
1046 } 860 }
1047 pdn->class_code = *class_code; 861 edev->class_code = *class_code;
1048 862
1049 /* Ok... see if this device supports EEH. Some do, some don't, 863 /* Ok... see if this device supports EEH. Some do, some don't,
1050 * and the only way to find out is to check each and every one. */ 864 * and the only way to find out is to check each and every one.
865 */
1051 regs = of_get_property(dn, "reg", NULL); 866 regs = of_get_property(dn, "reg", NULL);
1052 if (regs) { 867 if (regs) {
1053 /* First register entry is addr (00BBSS00) */ 868 /* First register entry is addr (00BBSS00) */
1054 /* Try to enable eeh */ 869 /* Try to enable eeh */
1055 ret = rtas_call(ibm_set_eeh_option, 4, 1, NULL, 870 ret = eeh_ops->set_option(dn, EEH_OPT_ENABLE);
1056 regs[0], info->buid_hi, info->buid_lo,
1057 EEH_ENABLE);
1058 871
1059 enable = 0; 872 enable = 0;
1060 if (ret == 0) { 873 if (ret == 0) {
1061 pdn->eeh_config_addr = regs[0]; 874 edev->config_addr = regs[0];
1062 875
1063 /* If the newer, better, ibm,get-config-addr-info is supported, 876 /* If the newer, better, ibm,get-config-addr-info is supported,
1064 * then use that instead. */ 877 * then use that instead.
1065 pdn->eeh_pe_config_addr = get_pe_addr(pdn->eeh_config_addr, info); 878 */
879 edev->pe_config_addr = eeh_ops->get_pe_addr(dn);
1066 880
1067 /* Some older systems (Power4) allow the 881 /* Some older systems (Power4) allow the
1068 * ibm,set-eeh-option call to succeed even on nodes 882 * ibm,set-eeh-option call to succeed even on nodes
1069 * where EEH is not supported. Verify support 883 * where EEH is not supported. Verify support
1070 * explicitly. */ 884 * explicitly.
1071 ret = read_slot_reset_state(pdn, rets); 885 */
1072 if ((ret == 0) && (rets[1] == 1)) 886 ret = eeh_ops->get_state(dn, NULL);
887 if (ret > 0 && ret != EEH_STATE_NOT_SUPPORT)
1073 enable = 1; 888 enable = 1;
1074 } 889 }
1075 890
1076 if (enable) { 891 if (enable) {
1077 eeh_subsystem_enabled = 1; 892 eeh_subsystem_enabled = 1;
1078 pdn->eeh_mode |= EEH_MODE_SUPPORTED; 893 edev->mode |= EEH_MODE_SUPPORTED;
1079 894
1080 pr_debug("EEH: %s: eeh enabled, config=%x pe_config=%x\n", 895 pr_debug("EEH: %s: eeh enabled, config=%x pe_config=%x\n",
1081 dn->full_name, pdn->eeh_config_addr, 896 dn->full_name, edev->config_addr,
1082 pdn->eeh_pe_config_addr); 897 edev->pe_config_addr);
1083 } else { 898 } else {
1084 899
1085 /* This device doesn't support EEH, but it may have an 900 /* This device doesn't support EEH, but it may have an
1086 * EEH parent, in which case we mark it as supported. */ 901 * EEH parent, in which case we mark it as supported.
1087 if (dn->parent && PCI_DN(dn->parent) 902 */
1088 && (PCI_DN(dn->parent)->eeh_mode & EEH_MODE_SUPPORTED)) { 903 if (dn->parent && of_node_to_eeh_dev(dn->parent) &&
904 (of_node_to_eeh_dev(dn->parent)->mode & EEH_MODE_SUPPORTED)) {
1089 /* Parent supports EEH. */ 905 /* Parent supports EEH. */
1090 pdn->eeh_mode |= EEH_MODE_SUPPORTED; 906 edev->mode |= EEH_MODE_SUPPORTED;
1091 pdn->eeh_config_addr = PCI_DN(dn->parent)->eeh_config_addr; 907 edev->config_addr = of_node_to_eeh_dev(dn->parent)->config_addr;
1092 return NULL; 908 return NULL;
1093 } 909 }
1094 } 910 }
@@ -1097,11 +913,63 @@ static void *early_enable_eeh(struct device_node *dn, void *data)
1097 dn->full_name); 913 dn->full_name);
1098 } 914 }
1099 915
1100 eeh_save_bars(pdn); 916 eeh_save_bars(edev);
1101 return NULL; 917 return NULL;
1102} 918}
1103 919
1104/* 920/**
921 * eeh_ops_register - Register platform dependent EEH operations
922 * @ops: platform dependent EEH operations
923 *
924 * Register the platform dependent EEH operation callback
925 * functions. The platform should call this function before
926 * any other EEH operations.
927 */
928int __init eeh_ops_register(struct eeh_ops *ops)
929{
930 if (!ops->name) {
931 pr_warning("%s: Invalid EEH ops name for %p\n",
932 __func__, ops);
933 return -EINVAL;
934 }
935
936 if (eeh_ops && eeh_ops != ops) {
937 pr_warning("%s: EEH ops of platform %s already existing (%s)\n",
938 __func__, eeh_ops->name, ops->name);
939 return -EEXIST;
940 }
941
942 eeh_ops = ops;
943
944 return 0;
945}
946
947/**
948 * eeh_ops_unregister - Unreigster platform dependent EEH operations
949 * @name: name of EEH platform operations
950 *
951 * Unregister the platform dependent EEH operation callback
952 * functions.
953 */
954int __exit eeh_ops_unregister(const char *name)
955{
956 if (!name || !strlen(name)) {
957 pr_warning("%s: Invalid EEH ops name\n",
958 __func__);
959 return -EINVAL;
960 }
961
962 if (eeh_ops && !strcmp(eeh_ops->name, name)) {
963 eeh_ops = NULL;
964 return 0;
965 }
966
967 return -EEXIST;
968}
969
970/**
971 * eeh_init - EEH initialization
972 *
1105 * Initialize EEH by trying to enable it for all of the adapters in the system. 973 * Initialize EEH by trying to enable it for all of the adapters in the system.
1106 * As a side effect we can determine here if eeh is supported at all. 974 * As a side effect we can determine here if eeh is supported at all.
1107 * Note that we leave EEH on so failed config cycles won't cause a machine 975 * Note that we leave EEH on so failed config cycles won't cause a machine
@@ -1117,50 +985,35 @@ static void *early_enable_eeh(struct device_node *dn, void *data)
1117void __init eeh_init(void) 985void __init eeh_init(void)
1118{ 986{
1119 struct device_node *phb, *np; 987 struct device_node *phb, *np;
1120 struct eeh_early_enable_info info; 988 int ret;
989
990 /* call platform initialization function */
991 if (!eeh_ops) {
992 pr_warning("%s: Platform EEH operation not found\n",
993 __func__);
994 return;
995 } else if ((ret = eeh_ops->init())) {
996 pr_warning("%s: Failed to call platform init function (%d)\n",
997 __func__, ret);
998 return;
999 }
1121 1000
1122 raw_spin_lock_init(&confirm_error_lock); 1001 raw_spin_lock_init(&confirm_error_lock);
1123 spin_lock_init(&slot_errbuf_lock);
1124 1002
1125 np = of_find_node_by_path("/rtas"); 1003 np = of_find_node_by_path("/rtas");
1126 if (np == NULL) 1004 if (np == NULL)
1127 return; 1005 return;
1128 1006
1129 ibm_set_eeh_option = rtas_token("ibm,set-eeh-option");
1130 ibm_set_slot_reset = rtas_token("ibm,set-slot-reset");
1131 ibm_read_slot_reset_state2 = rtas_token("ibm,read-slot-reset-state2");
1132 ibm_read_slot_reset_state = rtas_token("ibm,read-slot-reset-state");
1133 ibm_slot_error_detail = rtas_token("ibm,slot-error-detail");
1134 ibm_get_config_addr_info = rtas_token("ibm,get-config-addr-info");
1135 ibm_get_config_addr_info2 = rtas_token("ibm,get-config-addr-info2");
1136 ibm_configure_bridge = rtas_token ("ibm,configure-bridge");
1137 ibm_configure_pe = rtas_token("ibm,configure-pe");
1138
1139 if (ibm_set_eeh_option == RTAS_UNKNOWN_SERVICE)
1140 return;
1141
1142 eeh_error_buf_size = rtas_token("rtas-error-log-max");
1143 if (eeh_error_buf_size == RTAS_UNKNOWN_SERVICE) {
1144 eeh_error_buf_size = 1024;
1145 }
1146 if (eeh_error_buf_size > RTAS_ERROR_LOG_MAX) {
1147 printk(KERN_WARNING "EEH: rtas-error-log-max is bigger than allocated "
1148 "buffer ! (%d vs %d)", eeh_error_buf_size, RTAS_ERROR_LOG_MAX);
1149 eeh_error_buf_size = RTAS_ERROR_LOG_MAX;
1150 }
1151
1152 /* Enable EEH for all adapters. Note that eeh requires buid's */ 1007 /* Enable EEH for all adapters. Note that eeh requires buid's */
1153 for (phb = of_find_node_by_name(NULL, "pci"); phb; 1008 for (phb = of_find_node_by_name(NULL, "pci"); phb;
1154 phb = of_find_node_by_name(phb, "pci")) { 1009 phb = of_find_node_by_name(phb, "pci")) {
1155 unsigned long buid; 1010 unsigned long buid;
1156 1011
1157 buid = get_phb_buid(phb); 1012 buid = get_phb_buid(phb);
1158 if (buid == 0 || PCI_DN(phb) == NULL) 1013 if (buid == 0 || !of_node_to_eeh_dev(phb))
1159 continue; 1014 continue;
1160 1015
1161 info.buid_lo = BUID_LO(buid); 1016 traverse_pci_devices(phb, eeh_early_enable, NULL);
1162 info.buid_hi = BUID_HI(buid);
1163 traverse_pci_devices(phb, early_enable_eeh, &info);
1164 } 1017 }
1165 1018
1166 if (eeh_subsystem_enabled) 1019 if (eeh_subsystem_enabled)
@@ -1170,7 +1023,7 @@ void __init eeh_init(void)
1170} 1023}
1171 1024
1172/** 1025/**
1173 * eeh_add_device_early - enable EEH for the indicated device_node 1026 * eeh_add_device_early - Enable EEH for the indicated device_node
1174 * @dn: device node for which to set up EEH 1027 * @dn: device node for which to set up EEH
1175 * 1028 *
1176 * This routine must be used to perform EEH initialization for PCI 1029 * This routine must be used to perform EEH initialization for PCI
@@ -1184,21 +1037,26 @@ void __init eeh_init(void)
1184static void eeh_add_device_early(struct device_node *dn) 1037static void eeh_add_device_early(struct device_node *dn)
1185{ 1038{
1186 struct pci_controller *phb; 1039 struct pci_controller *phb;
1187 struct eeh_early_enable_info info;
1188 1040
1189 if (!dn || !PCI_DN(dn)) 1041 if (!dn || !of_node_to_eeh_dev(dn))
1190 return; 1042 return;
1191 phb = PCI_DN(dn)->phb; 1043 phb = of_node_to_eeh_dev(dn)->phb;
1192 1044
1193 /* USB Bus children of PCI devices will not have BUID's */ 1045 /* USB Bus children of PCI devices will not have BUID's */
1194 if (NULL == phb || 0 == phb->buid) 1046 if (NULL == phb || 0 == phb->buid)
1195 return; 1047 return;
1196 1048
1197 info.buid_hi = BUID_HI(phb->buid); 1049 eeh_early_enable(dn, NULL);
1198 info.buid_lo = BUID_LO(phb->buid);
1199 early_enable_eeh(dn, &info);
1200} 1050}
1201 1051
1052/**
1053 * eeh_add_device_tree_early - Enable EEH for the indicated device
1054 * @dn: device node
1055 *
1056 * This routine must be used to perform EEH initialization for the
1057 * indicated PCI device that was added after system boot (e.g.
1058 * hotplug, dlpar).
1059 */
1202void eeh_add_device_tree_early(struct device_node *dn) 1060void eeh_add_device_tree_early(struct device_node *dn)
1203{ 1061{
1204 struct device_node *sib; 1062 struct device_node *sib;
@@ -1210,7 +1068,7 @@ void eeh_add_device_tree_early(struct device_node *dn)
1210EXPORT_SYMBOL_GPL(eeh_add_device_tree_early); 1068EXPORT_SYMBOL_GPL(eeh_add_device_tree_early);
1211 1069
1212/** 1070/**
1213 * eeh_add_device_late - perform EEH initialization for the indicated pci device 1071 * eeh_add_device_late - Perform EEH initialization for the indicated pci device
1214 * @dev: pci device for which to set up EEH 1072 * @dev: pci device for which to set up EEH
1215 * 1073 *
1216 * This routine must be used to complete EEH initialization for PCI 1074 * This routine must be used to complete EEH initialization for PCI
@@ -1219,7 +1077,7 @@ EXPORT_SYMBOL_GPL(eeh_add_device_tree_early);
1219static void eeh_add_device_late(struct pci_dev *dev) 1077static void eeh_add_device_late(struct pci_dev *dev)
1220{ 1078{
1221 struct device_node *dn; 1079 struct device_node *dn;
1222 struct pci_dn *pdn; 1080 struct eeh_dev *edev;
1223 1081
1224 if (!dev || !eeh_subsystem_enabled) 1082 if (!dev || !eeh_subsystem_enabled)
1225 return; 1083 return;
@@ -1227,20 +1085,29 @@ static void eeh_add_device_late(struct pci_dev *dev)
1227 pr_debug("EEH: Adding device %s\n", pci_name(dev)); 1085 pr_debug("EEH: Adding device %s\n", pci_name(dev));
1228 1086
1229 dn = pci_device_to_OF_node(dev); 1087 dn = pci_device_to_OF_node(dev);
1230 pdn = PCI_DN(dn); 1088 edev = pci_dev_to_eeh_dev(dev);
1231 if (pdn->pcidev == dev) { 1089 if (edev->pdev == dev) {
1232 pr_debug("EEH: Already referenced !\n"); 1090 pr_debug("EEH: Already referenced !\n");
1233 return; 1091 return;
1234 } 1092 }
1235 WARN_ON(pdn->pcidev); 1093 WARN_ON(edev->pdev);
1236 1094
1237 pci_dev_get (dev); 1095 pci_dev_get(dev);
1238 pdn->pcidev = dev; 1096 edev->pdev = dev;
1097 dev->dev.archdata.edev = edev;
1239 1098
1240 pci_addr_cache_insert_device(dev); 1099 pci_addr_cache_insert_device(dev);
1241 eeh_sysfs_add_device(dev); 1100 eeh_sysfs_add_device(dev);
1242} 1101}
1243 1102
1103/**
1104 * eeh_add_device_tree_late - Perform EEH initialization for the indicated PCI bus
1105 * @bus: PCI bus
1106 *
1107 * This routine must be used to perform EEH initialization for PCI
1108 * devices which are attached to the indicated PCI bus. The PCI bus
1109 * is added after system boot through hotplug or dlpar.
1110 */
1244void eeh_add_device_tree_late(struct pci_bus *bus) 1111void eeh_add_device_tree_late(struct pci_bus *bus)
1245{ 1112{
1246 struct pci_dev *dev; 1113 struct pci_dev *dev;
@@ -1257,7 +1124,7 @@ void eeh_add_device_tree_late(struct pci_bus *bus)
1257EXPORT_SYMBOL_GPL(eeh_add_device_tree_late); 1124EXPORT_SYMBOL_GPL(eeh_add_device_tree_late);
1258 1125
1259/** 1126/**
1260 * eeh_remove_device - undo EEH setup for the indicated pci device 1127 * eeh_remove_device - Undo EEH setup for the indicated pci device
1261 * @dev: pci device to be removed 1128 * @dev: pci device to be removed
1262 * 1129 *
1263 * This routine should be called when a device is removed from 1130 * This routine should be called when a device is removed from
@@ -1268,25 +1135,35 @@ EXPORT_SYMBOL_GPL(eeh_add_device_tree_late);
1268 */ 1135 */
1269static void eeh_remove_device(struct pci_dev *dev) 1136static void eeh_remove_device(struct pci_dev *dev)
1270{ 1137{
1271 struct device_node *dn; 1138 struct eeh_dev *edev;
1139
1272 if (!dev || !eeh_subsystem_enabled) 1140 if (!dev || !eeh_subsystem_enabled)
1273 return; 1141 return;
1142 edev = pci_dev_to_eeh_dev(dev);
1274 1143
1275 /* Unregister the device with the EEH/PCI address search system */ 1144 /* Unregister the device with the EEH/PCI address search system */
1276 pr_debug("EEH: Removing device %s\n", pci_name(dev)); 1145 pr_debug("EEH: Removing device %s\n", pci_name(dev));
1277 1146
1278 dn = pci_device_to_OF_node(dev); 1147 if (!edev || !edev->pdev) {
1279 if (PCI_DN(dn)->pcidev == NULL) {
1280 pr_debug("EEH: Not referenced !\n"); 1148 pr_debug("EEH: Not referenced !\n");
1281 return; 1149 return;
1282 } 1150 }
1283 PCI_DN(dn)->pcidev = NULL; 1151 edev->pdev = NULL;
1284 pci_dev_put (dev); 1152 dev->dev.archdata.edev = NULL;
1153 pci_dev_put(dev);
1285 1154
1286 pci_addr_cache_remove_device(dev); 1155 pci_addr_cache_remove_device(dev);
1287 eeh_sysfs_remove_device(dev); 1156 eeh_sysfs_remove_device(dev);
1288} 1157}
1289 1158
1159/**
1160 * eeh_remove_bus_device - Undo EEH setup for the indicated PCI device
1161 * @dev: PCI device
1162 *
1163 * This routine must be called when a device is removed from the
1164 * running system through hotplug or dlpar. The corresponding
1165 * PCI address cache will be removed.
1166 */
1290void eeh_remove_bus_device(struct pci_dev *dev) 1167void eeh_remove_bus_device(struct pci_dev *dev)
1291{ 1168{
1292 struct pci_bus *bus = dev->subordinate; 1169 struct pci_bus *bus = dev->subordinate;
@@ -1305,21 +1182,24 @@ static int proc_eeh_show(struct seq_file *m, void *v)
1305{ 1182{
1306 if (0 == eeh_subsystem_enabled) { 1183 if (0 == eeh_subsystem_enabled) {
1307 seq_printf(m, "EEH Subsystem is globally disabled\n"); 1184 seq_printf(m, "EEH Subsystem is globally disabled\n");
1308 seq_printf(m, "eeh_total_mmio_ffs=%ld\n", total_mmio_ffs); 1185 seq_printf(m, "eeh_total_mmio_ffs=%llu\n", eeh_stats.total_mmio_ffs);
1309 } else { 1186 } else {
1310 seq_printf(m, "EEH Subsystem is enabled\n"); 1187 seq_printf(m, "EEH Subsystem is enabled\n");
1311 seq_printf(m, 1188 seq_printf(m,
1312 "no device=%ld\n" 1189 "no device=%llu\n"
1313 "no device node=%ld\n" 1190 "no device node=%llu\n"
1314 "no config address=%ld\n" 1191 "no config address=%llu\n"
1315 "check not wanted=%ld\n" 1192 "check not wanted=%llu\n"
1316 "eeh_total_mmio_ffs=%ld\n" 1193 "eeh_total_mmio_ffs=%llu\n"
1317 "eeh_false_positives=%ld\n" 1194 "eeh_false_positives=%llu\n"
1318 "eeh_slot_resets=%ld\n", 1195 "eeh_slot_resets=%llu\n",
1319 no_device, no_dn, no_cfg_addr, 1196 eeh_stats.no_device,
1320 ignored_check, total_mmio_ffs, 1197 eeh_stats.no_dn,
1321 false_positives, 1198 eeh_stats.no_cfg_addr,
1322 slot_resets); 1199 eeh_stats.ignored_check,
1200 eeh_stats.total_mmio_ffs,
1201 eeh_stats.false_positives,
1202 eeh_stats.slot_resets);
1323 } 1203 }
1324 1204
1325 return 0; 1205 return 0;
diff --git a/arch/powerpc/platforms/pseries/eeh_cache.c b/arch/powerpc/platforms/pseries/eeh_cache.c
index fc5ae767989..e5ae1c687c6 100644
--- a/arch/powerpc/platforms/pseries/eeh_cache.c
+++ b/arch/powerpc/platforms/pseries/eeh_cache.c
@@ -1,5 +1,4 @@
1/* 1/*
2 * eeh_cache.c
3 * PCI address cache; allows the lookup of PCI devices based on I/O address 2 * PCI address cache; allows the lookup of PCI devices based on I/O address
4 * 3 *
5 * Copyright IBM Corporation 2004 4 * Copyright IBM Corporation 2004
@@ -47,8 +46,7 @@
47 * than any hash algo I could think of for this problem, even 46 * than any hash algo I could think of for this problem, even
48 * with the penalty of slow pointer chases for d-cache misses). 47 * with the penalty of slow pointer chases for d-cache misses).
49 */ 48 */
50struct pci_io_addr_range 49struct pci_io_addr_range {
51{
52 struct rb_node rb_node; 50 struct rb_node rb_node;
53 unsigned long addr_lo; 51 unsigned long addr_lo;
54 unsigned long addr_hi; 52 unsigned long addr_hi;
@@ -56,13 +54,12 @@ struct pci_io_addr_range
56 unsigned int flags; 54 unsigned int flags;
57}; 55};
58 56
59static struct pci_io_addr_cache 57static struct pci_io_addr_cache {
60{
61 struct rb_root rb_root; 58 struct rb_root rb_root;
62 spinlock_t piar_lock; 59 spinlock_t piar_lock;
63} pci_io_addr_cache_root; 60} pci_io_addr_cache_root;
64 61
65static inline struct pci_dev *__pci_get_device_by_addr(unsigned long addr) 62static inline struct pci_dev *__pci_addr_cache_get_device(unsigned long addr)
66{ 63{
67 struct rb_node *n = pci_io_addr_cache_root.rb_root.rb_node; 64 struct rb_node *n = pci_io_addr_cache_root.rb_root.rb_node;
68 65
@@ -86,7 +83,7 @@ static inline struct pci_dev *__pci_get_device_by_addr(unsigned long addr)
86} 83}
87 84
88/** 85/**
89 * pci_get_device_by_addr - Get device, given only address 86 * pci_addr_cache_get_device - Get device, given only address
90 * @addr: mmio (PIO) phys address or i/o port number 87 * @addr: mmio (PIO) phys address or i/o port number
91 * 88 *
92 * Given an mmio phys address, or a port number, find a pci device 89 * Given an mmio phys address, or a port number, find a pci device
@@ -95,13 +92,13 @@ static inline struct pci_dev *__pci_get_device_by_addr(unsigned long addr)
95 * from zero (that is, they do *not* have pci_io_addr added in). 92 * from zero (that is, they do *not* have pci_io_addr added in).
96 * It is safe to call this function within an interrupt. 93 * It is safe to call this function within an interrupt.
97 */ 94 */
98struct pci_dev *pci_get_device_by_addr(unsigned long addr) 95struct pci_dev *pci_addr_cache_get_device(unsigned long addr)
99{ 96{
100 struct pci_dev *dev; 97 struct pci_dev *dev;
101 unsigned long flags; 98 unsigned long flags;
102 99
103 spin_lock_irqsave(&pci_io_addr_cache_root.piar_lock, flags); 100 spin_lock_irqsave(&pci_io_addr_cache_root.piar_lock, flags);
104 dev = __pci_get_device_by_addr(addr); 101 dev = __pci_addr_cache_get_device(addr);
105 spin_unlock_irqrestore(&pci_io_addr_cache_root.piar_lock, flags); 102 spin_unlock_irqrestore(&pci_io_addr_cache_root.piar_lock, flags);
106 return dev; 103 return dev;
107} 104}
@@ -166,7 +163,7 @@ pci_addr_cache_insert(struct pci_dev *dev, unsigned long alo,
166 163
167#ifdef DEBUG 164#ifdef DEBUG
168 printk(KERN_DEBUG "PIAR: insert range=[%lx:%lx] dev=%s\n", 165 printk(KERN_DEBUG "PIAR: insert range=[%lx:%lx] dev=%s\n",
169 alo, ahi, pci_name (dev)); 166 alo, ahi, pci_name(dev));
170#endif 167#endif
171 168
172 rb_link_node(&piar->rb_node, parent, p); 169 rb_link_node(&piar->rb_node, parent, p);
@@ -178,7 +175,7 @@ pci_addr_cache_insert(struct pci_dev *dev, unsigned long alo,
178static void __pci_addr_cache_insert_device(struct pci_dev *dev) 175static void __pci_addr_cache_insert_device(struct pci_dev *dev)
179{ 176{
180 struct device_node *dn; 177 struct device_node *dn;
181 struct pci_dn *pdn; 178 struct eeh_dev *edev;
182 int i; 179 int i;
183 180
184 dn = pci_device_to_OF_node(dev); 181 dn = pci_device_to_OF_node(dev);
@@ -187,13 +184,19 @@ static void __pci_addr_cache_insert_device(struct pci_dev *dev)
187 return; 184 return;
188 } 185 }
189 186
187 edev = of_node_to_eeh_dev(dn);
188 if (!edev) {
189 pr_warning("PCI: no EEH dev found for dn=%s\n",
190 dn->full_name);
191 return;
192 }
193
190 /* Skip any devices for which EEH is not enabled. */ 194 /* Skip any devices for which EEH is not enabled. */
191 pdn = PCI_DN(dn); 195 if (!(edev->mode & EEH_MODE_SUPPORTED) ||
192 if (!(pdn->eeh_mode & EEH_MODE_SUPPORTED) || 196 edev->mode & EEH_MODE_NOCHECK) {
193 pdn->eeh_mode & EEH_MODE_NOCHECK) {
194#ifdef DEBUG 197#ifdef DEBUG
195 printk(KERN_INFO "PCI: skip building address cache for=%s - %s\n", 198 pr_info("PCI: skip building address cache for=%s - %s\n",
196 pci_name(dev), pdn->node->full_name); 199 pci_name(dev), dn->full_name);
197#endif 200#endif
198 return; 201 return;
199 } 202 }
@@ -284,6 +287,7 @@ void pci_addr_cache_remove_device(struct pci_dev *dev)
284void __init pci_addr_cache_build(void) 287void __init pci_addr_cache_build(void)
285{ 288{
286 struct device_node *dn; 289 struct device_node *dn;
290 struct eeh_dev *edev;
287 struct pci_dev *dev = NULL; 291 struct pci_dev *dev = NULL;
288 292
289 spin_lock_init(&pci_io_addr_cache_root.piar_lock); 293 spin_lock_init(&pci_io_addr_cache_root.piar_lock);
@@ -294,8 +298,14 @@ void __init pci_addr_cache_build(void)
294 dn = pci_device_to_OF_node(dev); 298 dn = pci_device_to_OF_node(dev);
295 if (!dn) 299 if (!dn)
296 continue; 300 continue;
301
302 edev = of_node_to_eeh_dev(dn);
303 if (!edev)
304 continue;
305
297 pci_dev_get(dev); /* matching put is in eeh_remove_device() */ 306 pci_dev_get(dev); /* matching put is in eeh_remove_device() */
298 PCI_DN(dn)->pcidev = dev; 307 dev->dev.archdata.edev = edev;
308 edev->pdev = dev;
299 309
300 eeh_sysfs_add_device(dev); 310 eeh_sysfs_add_device(dev);
301 } 311 }
diff --git a/arch/powerpc/platforms/pseries/eeh_dev.c b/arch/powerpc/platforms/pseries/eeh_dev.c
new file mode 100644
index 00000000000..f3aed7dcae9
--- /dev/null
+++ b/arch/powerpc/platforms/pseries/eeh_dev.c
@@ -0,0 +1,102 @@
1/*
2 * The file intends to implement dynamic creation of EEH device, which will
3 * be bound with OF node and PCI device simutaneously. The EEH devices would
4 * be foundamental information for EEH core components to work proerly. Besides,
5 * We have to support multiple situations where dynamic creation of EEH device
6 * is required:
7 *
8 * 1) Before PCI emunation starts, we need create EEH devices according to the
9 * PCI sensitive OF nodes.
10 * 2) When PCI emunation is done, we need do the binding between PCI device and
11 * the associated EEH device.
12 * 3) DR (Dynamic Reconfiguration) would create PCI sensitive OF node. EEH device
13 * will be created while PCI sensitive OF node is detected from DR.
14 * 4) PCI hotplug needs redoing the binding between PCI device and EEH device. If
15 * PHB is newly inserted, we also need create EEH devices accordingly.
16 *
17 * Copyright Benjamin Herrenschmidt & Gavin Shan, IBM Corporation 2012.
18 *
19 * This program is free software; you can redistribute it and/or modify
20 * it under the terms of the GNU General Public License as published by
21 * the Free Software Foundation; either version 2 of the License, or
22 * (at your option) any later version.
23 *
24 * This program is distributed in the hope that it will be useful,
25 * but WITHOUT ANY WARRANTY; without even the implied warranty of
26 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
27 * GNU General Public License for more details.
28 *
29 * You should have received a copy of the GNU General Public License
30 * along with this program; if not, write to the Free Software
31 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
32 */
33
34#include <linux/export.h>
35#include <linux/gfp.h>
36#include <linux/init.h>
37#include <linux/kernel.h>
38#include <linux/pci.h>
39#include <linux/string.h>
40
41#include <asm/pci-bridge.h>
42#include <asm/ppc-pci.h>
43
44/**
45 * eeh_dev_init - Create EEH device according to OF node
46 * @dn: device node
47 * @data: PHB
48 *
49 * It will create EEH device according to the given OF node. The function
50 * might be called by PCI emunation, DR, PHB hotplug.
51 */
52void * __devinit eeh_dev_init(struct device_node *dn, void *data)
53{
54 struct pci_controller *phb = data;
55 struct eeh_dev *edev;
56
57 /* Allocate EEH device */
58 edev = zalloc_maybe_bootmem(sizeof(*edev), GFP_KERNEL);
59 if (!edev) {
60 pr_warning("%s: out of memory\n", __func__);
61 return NULL;
62 }
63
64 /* Associate EEH device with OF node */
65 dn->edev = edev;
66 edev->dn = dn;
67 edev->phb = phb;
68
69 return NULL;
70}
71
72/**
73 * eeh_dev_phb_init_dynamic - Create EEH devices for devices included in PHB
74 * @phb: PHB
75 *
76 * Scan the PHB OF node and its child association, then create the
77 * EEH devices accordingly
78 */
79void __devinit eeh_dev_phb_init_dynamic(struct pci_controller *phb)
80{
81 struct device_node *dn = phb->dn;
82
83 /* EEH device for PHB */
84 eeh_dev_init(dn, phb);
85
86 /* EEH devices for children OF nodes */
87 traverse_pci_devices(dn, eeh_dev_init, phb);
88}
89
90/**
91 * eeh_dev_phb_init - Create EEH devices for devices included in existing PHBs
92 *
93 * Scan all the existing PHBs and create EEH devices for their OF
94 * nodes and their children OF nodes
95 */
96void __init eeh_dev_phb_init(void)
97{
98 struct pci_controller *phb, *tmp;
99
100 list_for_each_entry_safe(phb, tmp, &hose_list, list_node)
101 eeh_dev_phb_init_dynamic(phb);
102}
diff --git a/arch/powerpc/platforms/pseries/eeh_driver.c b/arch/powerpc/platforms/pseries/eeh_driver.c
index 1b6cb10589e..baf92cd9dfa 100644
--- a/arch/powerpc/platforms/pseries/eeh_driver.c
+++ b/arch/powerpc/platforms/pseries/eeh_driver.c
@@ -33,8 +33,14 @@
33#include <asm/prom.h> 33#include <asm/prom.h>
34#include <asm/rtas.h> 34#include <asm/rtas.h>
35 35
36 36/**
37static inline const char * pcid_name (struct pci_dev *pdev) 37 * eeh_pcid_name - Retrieve name of PCI device driver
38 * @pdev: PCI device
39 *
40 * This routine is used to retrieve the name of PCI device driver
41 * if that's valid.
42 */
43static inline const char *eeh_pcid_name(struct pci_dev *pdev)
38{ 44{
39 if (pdev && pdev->dev.driver) 45 if (pdev && pdev->dev.driver)
40 return pdev->dev.driver->name; 46 return pdev->dev.driver->name;
@@ -64,48 +70,59 @@ static void print_device_node_tree(struct pci_dn *pdn, int dent)
64#endif 70#endif
65 71
66/** 72/**
67 * eeh_disable_irq - disable interrupt for the recovering device 73 * eeh_disable_irq - Disable interrupt for the recovering device
74 * @dev: PCI device
75 *
76 * This routine must be called when reporting temporary or permanent
77 * error to the particular PCI device to disable interrupt of that
78 * device. If the device has enabled MSI or MSI-X interrupt, we needn't
79 * do real work because EEH should freeze DMA transfers for those PCI
80 * devices encountering EEH errors, which includes MSI or MSI-X.
68 */ 81 */
69static void eeh_disable_irq(struct pci_dev *dev) 82static void eeh_disable_irq(struct pci_dev *dev)
70{ 83{
71 struct device_node *dn = pci_device_to_OF_node(dev); 84 struct eeh_dev *edev = pci_dev_to_eeh_dev(dev);
72 85
73 /* Don't disable MSI and MSI-X interrupts. They are 86 /* Don't disable MSI and MSI-X interrupts. They are
74 * effectively disabled by the DMA Stopped state 87 * effectively disabled by the DMA Stopped state
75 * when an EEH error occurs. 88 * when an EEH error occurs.
76 */ 89 */
77 if (dev->msi_enabled || dev->msix_enabled) 90 if (dev->msi_enabled || dev->msix_enabled)
78 return; 91 return;
79 92
80 if (!irq_has_action(dev->irq)) 93 if (!irq_has_action(dev->irq))
81 return; 94 return;
82 95
83 PCI_DN(dn)->eeh_mode |= EEH_MODE_IRQ_DISABLED; 96 edev->mode |= EEH_MODE_IRQ_DISABLED;
84 disable_irq_nosync(dev->irq); 97 disable_irq_nosync(dev->irq);
85} 98}
86 99
87/** 100/**
88 * eeh_enable_irq - enable interrupt for the recovering device 101 * eeh_enable_irq - Enable interrupt for the recovering device
102 * @dev: PCI device
103 *
104 * This routine must be called to enable interrupt while failed
105 * device could be resumed.
89 */ 106 */
90static void eeh_enable_irq(struct pci_dev *dev) 107static void eeh_enable_irq(struct pci_dev *dev)
91{ 108{
92 struct device_node *dn = pci_device_to_OF_node(dev); 109 struct eeh_dev *edev = pci_dev_to_eeh_dev(dev);
93 110
94 if ((PCI_DN(dn)->eeh_mode) & EEH_MODE_IRQ_DISABLED) { 111 if ((edev->mode) & EEH_MODE_IRQ_DISABLED) {
95 PCI_DN(dn)->eeh_mode &= ~EEH_MODE_IRQ_DISABLED; 112 edev->mode &= ~EEH_MODE_IRQ_DISABLED;
96 enable_irq(dev->irq); 113 enable_irq(dev->irq);
97 } 114 }
98} 115}
99 116
100/* ------------------------------------------------------- */
101/** 117/**
102 * eeh_report_error - report pci error to each device driver 118 * eeh_report_error - Report pci error to each device driver
119 * @dev: PCI device
120 * @userdata: return value
103 * 121 *
104 * Report an EEH error to each device driver, collect up and 122 * Report an EEH error to each device driver, collect up and
105 * merge the device driver responses. Cumulative response 123 * merge the device driver responses. Cumulative response
106 * passed back in "userdata". 124 * passed back in "userdata".
107 */ 125 */
108
109static int eeh_report_error(struct pci_dev *dev, void *userdata) 126static int eeh_report_error(struct pci_dev *dev, void *userdata)
110{ 127{
111 enum pci_ers_result rc, *res = userdata; 128 enum pci_ers_result rc, *res = userdata;
@@ -122,7 +139,7 @@ static int eeh_report_error(struct pci_dev *dev, void *userdata)
122 !driver->err_handler->error_detected) 139 !driver->err_handler->error_detected)
123 return 0; 140 return 0;
124 141
125 rc = driver->err_handler->error_detected (dev, pci_channel_io_frozen); 142 rc = driver->err_handler->error_detected(dev, pci_channel_io_frozen);
126 143
127 /* A driver that needs a reset trumps all others */ 144 /* A driver that needs a reset trumps all others */
128 if (rc == PCI_ERS_RESULT_NEED_RESET) *res = rc; 145 if (rc == PCI_ERS_RESULT_NEED_RESET) *res = rc;
@@ -132,13 +149,14 @@ static int eeh_report_error(struct pci_dev *dev, void *userdata)
132} 149}
133 150
134/** 151/**
135 * eeh_report_mmio_enabled - tell drivers that MMIO has been enabled 152 * eeh_report_mmio_enabled - Tell drivers that MMIO has been enabled
153 * @dev: PCI device
154 * @userdata: return value
136 * 155 *
137 * Tells each device driver that IO ports, MMIO and config space I/O 156 * Tells each device driver that IO ports, MMIO and config space I/O
138 * are now enabled. Collects up and merges the device driver responses. 157 * are now enabled. Collects up and merges the device driver responses.
139 * Cumulative response passed back in "userdata". 158 * Cumulative response passed back in "userdata".
140 */ 159 */
141
142static int eeh_report_mmio_enabled(struct pci_dev *dev, void *userdata) 160static int eeh_report_mmio_enabled(struct pci_dev *dev, void *userdata)
143{ 161{
144 enum pci_ers_result rc, *res = userdata; 162 enum pci_ers_result rc, *res = userdata;
@@ -149,7 +167,7 @@ static int eeh_report_mmio_enabled(struct pci_dev *dev, void *userdata)
149 !driver->err_handler->mmio_enabled) 167 !driver->err_handler->mmio_enabled)
150 return 0; 168 return 0;
151 169
152 rc = driver->err_handler->mmio_enabled (dev); 170 rc = driver->err_handler->mmio_enabled(dev);
153 171
154 /* A driver that needs a reset trumps all others */ 172 /* A driver that needs a reset trumps all others */
155 if (rc == PCI_ERS_RESULT_NEED_RESET) *res = rc; 173 if (rc == PCI_ERS_RESULT_NEED_RESET) *res = rc;
@@ -159,9 +177,15 @@ static int eeh_report_mmio_enabled(struct pci_dev *dev, void *userdata)
159} 177}
160 178
161/** 179/**
162 * eeh_report_reset - tell device that slot has been reset 180 * eeh_report_reset - Tell device that slot has been reset
181 * @dev: PCI device
182 * @userdata: return value
183 *
184 * This routine must be called while EEH tries to reset particular
185 * PCI device so that the associated PCI device driver could take
186 * some actions, usually to save data the driver needs so that the
187 * driver can work again while the device is recovered.
163 */ 188 */
164
165static int eeh_report_reset(struct pci_dev *dev, void *userdata) 189static int eeh_report_reset(struct pci_dev *dev, void *userdata)
166{ 190{
167 enum pci_ers_result rc, *res = userdata; 191 enum pci_ers_result rc, *res = userdata;
@@ -188,9 +212,14 @@ static int eeh_report_reset(struct pci_dev *dev, void *userdata)
188} 212}
189 213
190/** 214/**
191 * eeh_report_resume - tell device to resume normal operations 215 * eeh_report_resume - Tell device to resume normal operations
216 * @dev: PCI device
217 * @userdata: return value
218 *
219 * This routine must be called to notify the device driver that it
220 * could resume so that the device driver can do some initialization
221 * to make the recovered device work again.
192 */ 222 */
193
194static int eeh_report_resume(struct pci_dev *dev, void *userdata) 223static int eeh_report_resume(struct pci_dev *dev, void *userdata)
195{ 224{
196 struct pci_driver *driver = dev->driver; 225 struct pci_driver *driver = dev->driver;
@@ -212,12 +241,13 @@ static int eeh_report_resume(struct pci_dev *dev, void *userdata)
212} 241}
213 242
214/** 243/**
215 * eeh_report_failure - tell device driver that device is dead. 244 * eeh_report_failure - Tell device driver that device is dead.
245 * @dev: PCI device
246 * @userdata: return value
216 * 247 *
217 * This informs the device driver that the device is permanently 248 * This informs the device driver that the device is permanently
218 * dead, and that no further recovery attempts will be made on it. 249 * dead, and that no further recovery attempts will be made on it.
219 */ 250 */
220
221static int eeh_report_failure(struct pci_dev *dev, void *userdata) 251static int eeh_report_failure(struct pci_dev *dev, void *userdata)
222{ 252{
223 struct pci_driver *driver = dev->driver; 253 struct pci_driver *driver = dev->driver;
@@ -238,65 +268,46 @@ static int eeh_report_failure(struct pci_dev *dev, void *userdata)
238 return 0; 268 return 0;
239} 269}
240 270
241/* ------------------------------------------------------- */
242/** 271/**
243 * handle_eeh_events -- reset a PCI device after hard lockup. 272 * eeh_reset_device - Perform actual reset of a pci slot
244 * 273 * @edev: PE associated EEH device
245 * pSeries systems will isolate a PCI slot if the PCI-Host 274 * @bus: PCI bus corresponding to the isolcated slot
246 * bridge detects address or data parity errors, DMA's
247 * occurring to wild addresses (which usually happen due to
248 * bugs in device drivers or in PCI adapter firmware).
249 * Slot isolations also occur if #SERR, #PERR or other misc
250 * PCI-related errors are detected.
251 * 275 *
252 * Recovery process consists of unplugging the device driver 276 * This routine must be called to do reset on the indicated PE.
253 * (which generated hotplug events to userspace), then issuing 277 * During the reset, udev might be invoked because those affected
254 * a PCI #RST to the device, then reconfiguring the PCI config 278 * PCI devices will be removed and then added.
255 * space for all bridges & devices under this slot, and then
256 * finally restarting the device drivers (which cause a second
257 * set of hotplug events to go out to userspace).
258 */ 279 */
259 280static int eeh_reset_device(struct eeh_dev *edev, struct pci_bus *bus)
260/**
261 * eeh_reset_device() -- perform actual reset of a pci slot
262 * @bus: pointer to the pci bus structure corresponding
263 * to the isolated slot. A non-null value will
264 * cause all devices under the bus to be removed
265 * and then re-added.
266 * @pe_dn: pointer to a "Partionable Endpoint" device node.
267 * This is the top-level structure on which pci
268 * bus resets can be performed.
269 */
270
271static int eeh_reset_device (struct pci_dn *pe_dn, struct pci_bus *bus)
272{ 281{
273 struct device_node *dn; 282 struct device_node *dn;
274 int cnt, rc; 283 int cnt, rc;
275 284
276 /* pcibios will clear the counter; save the value */ 285 /* pcibios will clear the counter; save the value */
277 cnt = pe_dn->eeh_freeze_count; 286 cnt = edev->freeze_count;
278 287
279 if (bus) 288 if (bus)
280 pcibios_remove_pci_devices(bus); 289 pcibios_remove_pci_devices(bus);
281 290
282 /* Reset the pci controller. (Asserts RST#; resets config space). 291 /* Reset the pci controller. (Asserts RST#; resets config space).
283 * Reconfigure bridges and devices. Don't try to bring the system 292 * Reconfigure bridges and devices. Don't try to bring the system
284 * up if the reset failed for some reason. */ 293 * up if the reset failed for some reason.
285 rc = rtas_set_slot_reset(pe_dn); 294 */
295 rc = eeh_reset_pe(edev);
286 if (rc) 296 if (rc)
287 return rc; 297 return rc;
288 298
289 /* Walk over all functions on this device. */ 299 /* Walk over all functions on this device. */
290 dn = pe_dn->node; 300 dn = eeh_dev_to_of_node(edev);
291 if (!pcibios_find_pci_bus(dn) && PCI_DN(dn->parent)) 301 if (!pcibios_find_pci_bus(dn) && of_node_to_eeh_dev(dn->parent))
292 dn = dn->parent->child; 302 dn = dn->parent->child;
293 303
294 while (dn) { 304 while (dn) {
295 struct pci_dn *ppe = PCI_DN(dn); 305 struct eeh_dev *pedev = of_node_to_eeh_dev(dn);
306
296 /* On Power4, always true because eeh_pe_config_addr=0 */ 307 /* On Power4, always true because eeh_pe_config_addr=0 */
297 if (pe_dn->eeh_pe_config_addr == ppe->eeh_pe_config_addr) { 308 if (edev->pe_config_addr == pedev->pe_config_addr) {
298 rtas_configure_bridge(ppe); 309 eeh_ops->configure_bridge(dn);
299 eeh_restore_bars(ppe); 310 eeh_restore_bars(pedev);
300 } 311 }
301 dn = dn->sibling; 312 dn = dn->sibling;
302 } 313 }
@@ -308,10 +319,10 @@ static int eeh_reset_device (struct pci_dn *pe_dn, struct pci_bus *bus)
308 * potentially weird things happen. 319 * potentially weird things happen.
309 */ 320 */
310 if (bus) { 321 if (bus) {
311 ssleep (5); 322 ssleep(5);
312 pcibios_add_pci_devices(bus); 323 pcibios_add_pci_devices(bus);
313 } 324 }
314 pe_dn->eeh_freeze_count = cnt; 325 edev->freeze_count = cnt;
315 326
316 return 0; 327 return 0;
317} 328}
@@ -321,23 +332,39 @@ static int eeh_reset_device (struct pci_dn *pe_dn, struct pci_bus *bus)
321 */ 332 */
322#define MAX_WAIT_FOR_RECOVERY 150 333#define MAX_WAIT_FOR_RECOVERY 150
323 334
324struct pci_dn * handle_eeh_events (struct eeh_event *event) 335/**
336 * eeh_handle_event - Reset a PCI device after hard lockup.
337 * @event: EEH event
338 *
339 * While PHB detects address or data parity errors on particular PCI
340 * slot, the associated PE will be frozen. Besides, DMA's occurring
341 * to wild addresses (which usually happen due to bugs in device
342 * drivers or in PCI adapter firmware) can cause EEH error. #SERR,
343 * #PERR or other misc PCI-related errors also can trigger EEH errors.
344 *
345 * Recovery process consists of unplugging the device driver (which
346 * generated hotplug events to userspace), then issuing a PCI #RST to
347 * the device, then reconfiguring the PCI config space for all bridges
348 * & devices under this slot, and then finally restarting the device
349 * drivers (which cause a second set of hotplug events to go out to
350 * userspace).
351 */
352struct eeh_dev *handle_eeh_events(struct eeh_event *event)
325{ 353{
326 struct device_node *frozen_dn; 354 struct device_node *frozen_dn;
327 struct pci_dn *frozen_pdn; 355 struct eeh_dev *frozen_edev;
328 struct pci_bus *frozen_bus; 356 struct pci_bus *frozen_bus;
329 int rc = 0; 357 int rc = 0;
330 enum pci_ers_result result = PCI_ERS_RESULT_NONE; 358 enum pci_ers_result result = PCI_ERS_RESULT_NONE;
331 const char *location, *pci_str, *drv_str, *bus_pci_str, *bus_drv_str; 359 const char *location, *pci_str, *drv_str, *bus_pci_str, *bus_drv_str;
332 360
333 frozen_dn = find_device_pe(event->dn); 361 frozen_dn = eeh_find_device_pe(eeh_dev_to_of_node(event->edev));
334 if (!frozen_dn) { 362 if (!frozen_dn) {
335 363 location = of_get_property(eeh_dev_to_of_node(event->edev), "ibm,loc-code", NULL);
336 location = of_get_property(event->dn, "ibm,loc-code", NULL);
337 location = location ? location : "unknown"; 364 location = location ? location : "unknown";
338 printk(KERN_ERR "EEH: Error: Cannot find partition endpoint " 365 printk(KERN_ERR "EEH: Error: Cannot find partition endpoint "
339 "for location=%s pci addr=%s\n", 366 "for location=%s pci addr=%s\n",
340 location, eeh_pci_name(event->dev)); 367 location, eeh_pci_name(eeh_dev_to_pci_dev(event->edev)));
341 return NULL; 368 return NULL;
342 } 369 }
343 370
@@ -350,9 +377,10 @@ struct pci_dn * handle_eeh_events (struct eeh_event *event)
350 * which was always an EADS pci bridge. In the new style, 377 * which was always an EADS pci bridge. In the new style,
351 * there might not be any EADS bridges, and even when there are, 378 * there might not be any EADS bridges, and even when there are,
352 * the firmware marks them as "EEH incapable". So another 379 * the firmware marks them as "EEH incapable". So another
353 * two-step is needed to find the pci bus.. */ 380 * two-step is needed to find the pci bus..
381 */
354 if (!frozen_bus) 382 if (!frozen_bus)
355 frozen_bus = pcibios_find_pci_bus (frozen_dn->parent); 383 frozen_bus = pcibios_find_pci_bus(frozen_dn->parent);
356 384
357 if (!frozen_bus) { 385 if (!frozen_bus) {
358 printk(KERN_ERR "EEH: Cannot find PCI bus " 386 printk(KERN_ERR "EEH: Cannot find PCI bus "
@@ -361,22 +389,21 @@ struct pci_dn * handle_eeh_events (struct eeh_event *event)
361 return NULL; 389 return NULL;
362 } 390 }
363 391
364 frozen_pdn = PCI_DN(frozen_dn); 392 frozen_edev = of_node_to_eeh_dev(frozen_dn);
365 frozen_pdn->eeh_freeze_count++; 393 frozen_edev->freeze_count++;
394 pci_str = eeh_pci_name(eeh_dev_to_pci_dev(event->edev));
395 drv_str = eeh_pcid_name(eeh_dev_to_pci_dev(event->edev));
366 396
367 pci_str = eeh_pci_name(event->dev); 397 if (frozen_edev->freeze_count > EEH_MAX_ALLOWED_FREEZES)
368 drv_str = pcid_name(event->dev);
369
370 if (frozen_pdn->eeh_freeze_count > EEH_MAX_ALLOWED_FREEZES)
371 goto excess_failures; 398 goto excess_failures;
372 399
373 printk(KERN_WARNING 400 printk(KERN_WARNING
374 "EEH: This PCI device has failed %d times in the last hour:\n", 401 "EEH: This PCI device has failed %d times in the last hour:\n",
375 frozen_pdn->eeh_freeze_count); 402 frozen_edev->freeze_count);
376 403
377 if (frozen_pdn->pcidev) { 404 if (frozen_edev->pdev) {
378 bus_pci_str = pci_name(frozen_pdn->pcidev); 405 bus_pci_str = pci_name(frozen_edev->pdev);
379 bus_drv_str = pcid_name(frozen_pdn->pcidev); 406 bus_drv_str = eeh_pcid_name(frozen_edev->pdev);
380 printk(KERN_WARNING 407 printk(KERN_WARNING
381 "EEH: Bus location=%s driver=%s pci addr=%s\n", 408 "EEH: Bus location=%s driver=%s pci addr=%s\n",
382 location, bus_drv_str, bus_pci_str); 409 location, bus_drv_str, bus_pci_str);
@@ -395,9 +422,10 @@ struct pci_dn * handle_eeh_events (struct eeh_event *event)
395 pci_walk_bus(frozen_bus, eeh_report_error, &result); 422 pci_walk_bus(frozen_bus, eeh_report_error, &result);
396 423
397 /* Get the current PCI slot state. This can take a long time, 424 /* Get the current PCI slot state. This can take a long time,
398 * sometimes over 3 seconds for certain systems. */ 425 * sometimes over 3 seconds for certain systems.
399 rc = eeh_wait_for_slot_status (frozen_pdn, MAX_WAIT_FOR_RECOVERY*1000); 426 */
400 if (rc < 0) { 427 rc = eeh_ops->wait_state(eeh_dev_to_of_node(frozen_edev), MAX_WAIT_FOR_RECOVERY*1000);
428 if (rc < 0 || rc == EEH_STATE_NOT_SUPPORT) {
401 printk(KERN_WARNING "EEH: Permanent failure\n"); 429 printk(KERN_WARNING "EEH: Permanent failure\n");
402 goto hard_fail; 430 goto hard_fail;
403 } 431 }
@@ -406,14 +434,14 @@ struct pci_dn * handle_eeh_events (struct eeh_event *event)
406 * don't post the error log until after all dev drivers 434 * don't post the error log until after all dev drivers
407 * have been informed. 435 * have been informed.
408 */ 436 */
409 eeh_slot_error_detail(frozen_pdn, EEH_LOG_TEMP_FAILURE); 437 eeh_slot_error_detail(frozen_edev, EEH_LOG_TEMP);
410 438
411 /* If all device drivers were EEH-unaware, then shut 439 /* If all device drivers were EEH-unaware, then shut
412 * down all of the device drivers, and hope they 440 * down all of the device drivers, and hope they
413 * go down willingly, without panicing the system. 441 * go down willingly, without panicing the system.
414 */ 442 */
415 if (result == PCI_ERS_RESULT_NONE) { 443 if (result == PCI_ERS_RESULT_NONE) {
416 rc = eeh_reset_device(frozen_pdn, frozen_bus); 444 rc = eeh_reset_device(frozen_edev, frozen_bus);
417 if (rc) { 445 if (rc) {
418 printk(KERN_WARNING "EEH: Unable to reset, rc=%d\n", rc); 446 printk(KERN_WARNING "EEH: Unable to reset, rc=%d\n", rc);
419 goto hard_fail; 447 goto hard_fail;
@@ -422,7 +450,7 @@ struct pci_dn * handle_eeh_events (struct eeh_event *event)
422 450
423 /* If all devices reported they can proceed, then re-enable MMIO */ 451 /* If all devices reported they can proceed, then re-enable MMIO */
424 if (result == PCI_ERS_RESULT_CAN_RECOVER) { 452 if (result == PCI_ERS_RESULT_CAN_RECOVER) {
425 rc = rtas_pci_enable(frozen_pdn, EEH_THAW_MMIO); 453 rc = eeh_pci_enable(frozen_edev, EEH_OPT_THAW_MMIO);
426 454
427 if (rc < 0) 455 if (rc < 0)
428 goto hard_fail; 456 goto hard_fail;
@@ -436,7 +464,7 @@ struct pci_dn * handle_eeh_events (struct eeh_event *event)
436 464
437 /* If all devices reported they can proceed, then re-enable DMA */ 465 /* If all devices reported they can proceed, then re-enable DMA */
438 if (result == PCI_ERS_RESULT_CAN_RECOVER) { 466 if (result == PCI_ERS_RESULT_CAN_RECOVER) {
439 rc = rtas_pci_enable(frozen_pdn, EEH_THAW_DMA); 467 rc = eeh_pci_enable(frozen_edev, EEH_OPT_THAW_DMA);
440 468
441 if (rc < 0) 469 if (rc < 0)
442 goto hard_fail; 470 goto hard_fail;
@@ -454,7 +482,7 @@ struct pci_dn * handle_eeh_events (struct eeh_event *event)
454 482
455 /* If any device called out for a reset, then reset the slot */ 483 /* If any device called out for a reset, then reset the slot */
456 if (result == PCI_ERS_RESULT_NEED_RESET) { 484 if (result == PCI_ERS_RESULT_NEED_RESET) {
457 rc = eeh_reset_device(frozen_pdn, NULL); 485 rc = eeh_reset_device(frozen_edev, NULL);
458 if (rc) { 486 if (rc) {
459 printk(KERN_WARNING "EEH: Cannot reset, rc=%d\n", rc); 487 printk(KERN_WARNING "EEH: Cannot reset, rc=%d\n", rc);
460 goto hard_fail; 488 goto hard_fail;
@@ -473,7 +501,7 @@ struct pci_dn * handle_eeh_events (struct eeh_event *event)
473 /* Tell all device drivers that they can resume operations */ 501 /* Tell all device drivers that they can resume operations */
474 pci_walk_bus(frozen_bus, eeh_report_resume, NULL); 502 pci_walk_bus(frozen_bus, eeh_report_resume, NULL);
475 503
476 return frozen_pdn; 504 return frozen_edev;
477 505
478excess_failures: 506excess_failures:
479 /* 507 /*
@@ -486,7 +514,7 @@ excess_failures:
486 "has failed %d times in the last hour " 514 "has failed %d times in the last hour "
487 "and has been permanently disabled.\n" 515 "and has been permanently disabled.\n"
488 "Please try reseating this device or replacing it.\n", 516 "Please try reseating this device or replacing it.\n",
489 location, drv_str, pci_str, frozen_pdn->eeh_freeze_count); 517 location, drv_str, pci_str, frozen_edev->freeze_count);
490 goto perm_error; 518 goto perm_error;
491 519
492hard_fail: 520hard_fail:
@@ -497,7 +525,7 @@ hard_fail:
497 location, drv_str, pci_str); 525 location, drv_str, pci_str);
498 526
499perm_error: 527perm_error:
500 eeh_slot_error_detail(frozen_pdn, EEH_LOG_PERM_FAILURE); 528 eeh_slot_error_detail(frozen_edev, EEH_LOG_PERM);
501 529
502 /* Notify all devices that they're about to go down. */ 530 /* Notify all devices that they're about to go down. */
503 pci_walk_bus(frozen_bus, eeh_report_failure, NULL); 531 pci_walk_bus(frozen_bus, eeh_report_failure, NULL);
@@ -508,4 +536,3 @@ perm_error:
508 return NULL; 536 return NULL;
509} 537}
510 538
511/* ---------- end of file ---------- */
diff --git a/arch/powerpc/platforms/pseries/eeh_event.c b/arch/powerpc/platforms/pseries/eeh_event.c
index d2383cfb6df..4a475256585 100644
--- a/arch/powerpc/platforms/pseries/eeh_event.c
+++ b/arch/powerpc/platforms/pseries/eeh_event.c
@@ -1,6 +1,4 @@
1/* 1/*
2 * eeh_event.c
3 *
4 * This program is free software; you can redistribute it and/or modify 2 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License as published by 3 * it under the terms of the GNU General Public License as published by
6 * the Free Software Foundation; either version 2 of the License, or 4 * the Free Software Foundation; either version 2 of the License, or
@@ -46,7 +44,7 @@ DECLARE_WORK(eeh_event_wq, eeh_thread_launcher);
46DEFINE_MUTEX(eeh_event_mutex); 44DEFINE_MUTEX(eeh_event_mutex);
47 45
48/** 46/**
49 * eeh_event_handler - dispatch EEH events. 47 * eeh_event_handler - Dispatch EEH events.
50 * @dummy - unused 48 * @dummy - unused
51 * 49 *
52 * The detection of a frozen slot can occur inside an interrupt, 50 * The detection of a frozen slot can occur inside an interrupt,
@@ -58,10 +56,10 @@ DEFINE_MUTEX(eeh_event_mutex);
58static int eeh_event_handler(void * dummy) 56static int eeh_event_handler(void * dummy)
59{ 57{
60 unsigned long flags; 58 unsigned long flags;
61 struct eeh_event *event; 59 struct eeh_event *event;
62 struct pci_dn *pdn; 60 struct eeh_dev *edev;
63 61
64 daemonize ("eehd"); 62 daemonize("eehd");
65 set_current_state(TASK_INTERRUPTIBLE); 63 set_current_state(TASK_INTERRUPTIBLE);
66 64
67 spin_lock_irqsave(&eeh_eventlist_lock, flags); 65 spin_lock_irqsave(&eeh_eventlist_lock, flags);
@@ -79,31 +77,37 @@ static int eeh_event_handler(void * dummy)
79 77
80 /* Serialize processing of EEH events */ 78 /* Serialize processing of EEH events */
81 mutex_lock(&eeh_event_mutex); 79 mutex_lock(&eeh_event_mutex);
82 eeh_mark_slot(event->dn, EEH_MODE_RECOVERING); 80 edev = event->edev;
81 eeh_mark_slot(eeh_dev_to_of_node(edev), EEH_MODE_RECOVERING);
83 82
84 printk(KERN_INFO "EEH: Detected PCI bus error on device %s\n", 83 printk(KERN_INFO "EEH: Detected PCI bus error on device %s\n",
85 eeh_pci_name(event->dev)); 84 eeh_pci_name(edev->pdev));
85
86 edev = handle_eeh_events(event);
86 87
87 pdn = handle_eeh_events(event); 88 eeh_clear_slot(eeh_dev_to_of_node(edev), EEH_MODE_RECOVERING);
89 pci_dev_put(edev->pdev);
88 90
89 eeh_clear_slot(event->dn, EEH_MODE_RECOVERING);
90 pci_dev_put(event->dev);
91 kfree(event); 91 kfree(event);
92 mutex_unlock(&eeh_event_mutex); 92 mutex_unlock(&eeh_event_mutex);
93 93
94 /* If there are no new errors after an hour, clear the counter. */ 94 /* If there are no new errors after an hour, clear the counter. */
95 if (pdn && pdn->eeh_freeze_count>0) { 95 if (edev && edev->freeze_count>0) {
96 msleep_interruptible (3600*1000); 96 msleep_interruptible(3600*1000);
97 if (pdn->eeh_freeze_count>0) 97 if (edev->freeze_count>0)
98 pdn->eeh_freeze_count--; 98 edev->freeze_count--;
99
99 } 100 }
100 101
101 return 0; 102 return 0;
102} 103}
103 104
104/** 105/**
105 * eeh_thread_launcher 106 * eeh_thread_launcher - Start kernel thread to handle EEH events
106 * @dummy - unused 107 * @dummy - unused
108 *
109 * This routine is called to start the kernel thread for processing
110 * EEH event.
107 */ 111 */
108static void eeh_thread_launcher(struct work_struct *dummy) 112static void eeh_thread_launcher(struct work_struct *dummy)
109{ 113{
@@ -112,18 +116,18 @@ static void eeh_thread_launcher(struct work_struct *dummy)
112} 116}
113 117
114/** 118/**
115 * eeh_send_failure_event - generate a PCI error event 119 * eeh_send_failure_event - Generate a PCI error event
116 * @dev pci device 120 * @edev: EEH device
117 * 121 *
118 * This routine can be called within an interrupt context; 122 * This routine can be called within an interrupt context;
119 * the actual event will be delivered in a normal context 123 * the actual event will be delivered in a normal context
120 * (from a workqueue). 124 * (from a workqueue).
121 */ 125 */
122int eeh_send_failure_event (struct device_node *dn, 126int eeh_send_failure_event(struct eeh_dev *edev)
123 struct pci_dev *dev)
124{ 127{
125 unsigned long flags; 128 unsigned long flags;
126 struct eeh_event *event; 129 struct eeh_event *event;
130 struct device_node *dn = eeh_dev_to_of_node(edev);
127 const char *location; 131 const char *location;
128 132
129 if (!mem_init_done) { 133 if (!mem_init_done) {
@@ -135,15 +139,14 @@ int eeh_send_failure_event (struct device_node *dn,
135 } 139 }
136 event = kmalloc(sizeof(*event), GFP_ATOMIC); 140 event = kmalloc(sizeof(*event), GFP_ATOMIC);
137 if (event == NULL) { 141 if (event == NULL) {
138 printk (KERN_ERR "EEH: out of memory, event not handled\n"); 142 printk(KERN_ERR "EEH: out of memory, event not handled\n");
139 return 1; 143 return 1;
140 } 144 }
141 145
142 if (dev) 146 if (edev->pdev)
143 pci_dev_get(dev); 147 pci_dev_get(edev->pdev);
144 148
145 event->dn = dn; 149 event->edev = edev;
146 event->dev = dev;
147 150
148 /* We may or may not be called in an interrupt context */ 151 /* We may or may not be called in an interrupt context */
149 spin_lock_irqsave(&eeh_eventlist_lock, flags); 152 spin_lock_irqsave(&eeh_eventlist_lock, flags);
@@ -154,5 +157,3 @@ int eeh_send_failure_event (struct device_node *dn,
154 157
155 return 0; 158 return 0;
156} 159}
157
158/********************** END OF FILE ******************************/
diff --git a/arch/powerpc/platforms/pseries/eeh_pseries.c b/arch/powerpc/platforms/pseries/eeh_pseries.c
new file mode 100644
index 00000000000..8752f79a6af
--- /dev/null
+++ b/arch/powerpc/platforms/pseries/eeh_pseries.c
@@ -0,0 +1,565 @@
1/*
2 * The file intends to implement the platform dependent EEH operations on pseries.
3 * Actually, the pseries platform is built based on RTAS heavily. That means the
4 * pseries platform dependent EEH operations will be built on RTAS calls. The functions
5 * are devired from arch/powerpc/platforms/pseries/eeh.c and necessary cleanup has
6 * been done.
7 *
8 * Copyright Benjamin Herrenschmidt & Gavin Shan, IBM Corporation 2011.
9 * Copyright IBM Corporation 2001, 2005, 2006
10 * Copyright Dave Engebretsen & Todd Inglett 2001
11 * Copyright Linas Vepstas 2005, 2006
12 *
13 * This program is free software; you can redistribute it and/or modify
14 * it under the terms of the GNU General Public License as published by
15 * the Free Software Foundation; either version 2 of the License, or
16 * (at your option) any later version.
17 *
18 * This program is distributed in the hope that it will be useful,
19 * but WITHOUT ANY WARRANTY; without even the implied warranty of
20 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21 * GNU General Public License for more details.
22 *
23 * You should have received a copy of the GNU General Public License
24 * along with this program; if not, write to the Free Software
25 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
26 */
27
28#include <linux/atomic.h>
29#include <linux/delay.h>
30#include <linux/export.h>
31#include <linux/init.h>
32#include <linux/list.h>
33#include <linux/of.h>
34#include <linux/pci.h>
35#include <linux/proc_fs.h>
36#include <linux/rbtree.h>
37#include <linux/sched.h>
38#include <linux/seq_file.h>
39#include <linux/spinlock.h>
40
41#include <asm/eeh.h>
42#include <asm/eeh_event.h>
43#include <asm/io.h>
44#include <asm/machdep.h>
45#include <asm/ppc-pci.h>
46#include <asm/rtas.h>
47
48/* RTAS tokens */
49static int ibm_set_eeh_option;
50static int ibm_set_slot_reset;
51static int ibm_read_slot_reset_state;
52static int ibm_read_slot_reset_state2;
53static int ibm_slot_error_detail;
54static int ibm_get_config_addr_info;
55static int ibm_get_config_addr_info2;
56static int ibm_configure_bridge;
57static int ibm_configure_pe;
58
59/*
60 * Buffer for reporting slot-error-detail rtas calls. Its here
61 * in BSS, and not dynamically alloced, so that it ends up in
62 * RMO where RTAS can access it.
63 */
64static unsigned char slot_errbuf[RTAS_ERROR_LOG_MAX];
65static DEFINE_SPINLOCK(slot_errbuf_lock);
66static int eeh_error_buf_size;
67
68/**
69 * pseries_eeh_init - EEH platform dependent initialization
70 *
71 * EEH platform dependent initialization on pseries.
72 */
73static int pseries_eeh_init(void)
74{
75 /* figure out EEH RTAS function call tokens */
76 ibm_set_eeh_option = rtas_token("ibm,set-eeh-option");
77 ibm_set_slot_reset = rtas_token("ibm,set-slot-reset");
78 ibm_read_slot_reset_state2 = rtas_token("ibm,read-slot-reset-state2");
79 ibm_read_slot_reset_state = rtas_token("ibm,read-slot-reset-state");
80 ibm_slot_error_detail = rtas_token("ibm,slot-error-detail");
81 ibm_get_config_addr_info2 = rtas_token("ibm,get-config-addr-info2");
82 ibm_get_config_addr_info = rtas_token("ibm,get-config-addr-info");
83 ibm_configure_pe = rtas_token("ibm,configure-pe");
84 ibm_configure_bridge = rtas_token ("ibm,configure-bridge");
85
86 /* necessary sanity check */
87 if (ibm_set_eeh_option == RTAS_UNKNOWN_SERVICE) {
88 pr_warning("%s: RTAS service <ibm,set-eeh-option> invalid\n",
89 __func__);
90 return -EINVAL;
91 } else if (ibm_set_slot_reset == RTAS_UNKNOWN_SERVICE) {
92 pr_warning("%s: RTAS service <ibm, set-slot-reset> invalid\n",
93 __func__);
94 return -EINVAL;
95 } else if (ibm_read_slot_reset_state2 == RTAS_UNKNOWN_SERVICE &&
96 ibm_read_slot_reset_state == RTAS_UNKNOWN_SERVICE) {
97 pr_warning("%s: RTAS service <ibm,read-slot-reset-state2> and "
98 "<ibm,read-slot-reset-state> invalid\n",
99 __func__);
100 return -EINVAL;
101 } else if (ibm_slot_error_detail == RTAS_UNKNOWN_SERVICE) {
102 pr_warning("%s: RTAS service <ibm,slot-error-detail> invalid\n",
103 __func__);
104 return -EINVAL;
105 } else if (ibm_get_config_addr_info2 == RTAS_UNKNOWN_SERVICE &&
106 ibm_get_config_addr_info == RTAS_UNKNOWN_SERVICE) {
107 pr_warning("%s: RTAS service <ibm,get-config-addr-info2> and "
108 "<ibm,get-config-addr-info> invalid\n",
109 __func__);
110 return -EINVAL;
111 } else if (ibm_configure_pe == RTAS_UNKNOWN_SERVICE &&
112 ibm_configure_bridge == RTAS_UNKNOWN_SERVICE) {
113 pr_warning("%s: RTAS service <ibm,configure-pe> and "
114 "<ibm,configure-bridge> invalid\n",
115 __func__);
116 return -EINVAL;
117 }
118
119 /* Initialize error log lock and size */
120 spin_lock_init(&slot_errbuf_lock);
121 eeh_error_buf_size = rtas_token("rtas-error-log-max");
122 if (eeh_error_buf_size == RTAS_UNKNOWN_SERVICE) {
123 pr_warning("%s: unknown EEH error log size\n",
124 __func__);
125 eeh_error_buf_size = 1024;
126 } else if (eeh_error_buf_size > RTAS_ERROR_LOG_MAX) {
127 pr_warning("%s: EEH error log size %d exceeds the maximal %d\n",
128 __func__, eeh_error_buf_size, RTAS_ERROR_LOG_MAX);
129 eeh_error_buf_size = RTAS_ERROR_LOG_MAX;
130 }
131
132 return 0;
133}
134
135/**
136 * pseries_eeh_set_option - Initialize EEH or MMIO/DMA reenable
137 * @dn: device node
138 * @option: operation to be issued
139 *
140 * The function is used to control the EEH functionality globally.
141 * Currently, following options are support according to PAPR:
142 * Enable EEH, Disable EEH, Enable MMIO and Enable DMA
143 */
144static int pseries_eeh_set_option(struct device_node *dn, int option)
145{
146 int ret = 0;
147 struct eeh_dev *edev;
148 const u32 *reg;
149 int config_addr;
150
151 edev = of_node_to_eeh_dev(dn);
152
153 /*
154 * When we're enabling or disabling EEH functioality on
155 * the particular PE, the PE config address is possibly
156 * unavailable. Therefore, we have to figure it out from
157 * the FDT node.
158 */
159 switch (option) {
160 case EEH_OPT_DISABLE:
161 case EEH_OPT_ENABLE:
162 reg = of_get_property(dn, "reg", NULL);
163 config_addr = reg[0];
164 break;
165
166 case EEH_OPT_THAW_MMIO:
167 case EEH_OPT_THAW_DMA:
168 config_addr = edev->config_addr;
169 if (edev->pe_config_addr)
170 config_addr = edev->pe_config_addr;
171 break;
172
173 default:
174 pr_err("%s: Invalid option %d\n",
175 __func__, option);
176 return -EINVAL;
177 }
178
179 ret = rtas_call(ibm_set_eeh_option, 4, 1, NULL,
180 config_addr, BUID_HI(edev->phb->buid),
181 BUID_LO(edev->phb->buid), option);
182
183 return ret;
184}
185
186/**
187 * pseries_eeh_get_pe_addr - Retrieve PE address
188 * @dn: device node
189 *
190 * Retrieve the assocated PE address. Actually, there're 2 RTAS
191 * function calls dedicated for the purpose. We need implement
192 * it through the new function and then the old one. Besides,
193 * you should make sure the config address is figured out from
194 * FDT node before calling the function.
195 *
196 * It's notable that zero'ed return value means invalid PE config
197 * address.
198 */
199static int pseries_eeh_get_pe_addr(struct device_node *dn)
200{
201 struct eeh_dev *edev;
202 int ret = 0;
203 int rets[3];
204
205 edev = of_node_to_eeh_dev(dn);
206
207 if (ibm_get_config_addr_info2 != RTAS_UNKNOWN_SERVICE) {
208 /*
209 * First of all, we need to make sure there has one PE
210 * associated with the device. Otherwise, PE address is
211 * meaningless.
212 */
213 ret = rtas_call(ibm_get_config_addr_info2, 4, 2, rets,
214 edev->config_addr, BUID_HI(edev->phb->buid),
215 BUID_LO(edev->phb->buid), 1);
216 if (ret || (rets[0] == 0))
217 return 0;
218
219 /* Retrieve the associated PE config address */
220 ret = rtas_call(ibm_get_config_addr_info2, 4, 2, rets,
221 edev->config_addr, BUID_HI(edev->phb->buid),
222 BUID_LO(edev->phb->buid), 0);
223 if (ret) {
224 pr_warning("%s: Failed to get PE address for %s\n",
225 __func__, dn->full_name);
226 return 0;
227 }
228
229 return rets[0];
230 }
231
232 if (ibm_get_config_addr_info != RTAS_UNKNOWN_SERVICE) {
233 ret = rtas_call(ibm_get_config_addr_info, 4, 2, rets,
234 edev->config_addr, BUID_HI(edev->phb->buid),
235 BUID_LO(edev->phb->buid), 0);
236 if (ret) {
237 pr_warning("%s: Failed to get PE address for %s\n",
238 __func__, dn->full_name);
239 return 0;
240 }
241
242 return rets[0];
243 }
244
245 return ret;
246}
247
248/**
249 * pseries_eeh_get_state - Retrieve PE state
250 * @dn: PE associated device node
251 * @state: return value
252 *
253 * Retrieve the state of the specified PE. On RTAS compliant
254 * pseries platform, there already has one dedicated RTAS function
255 * for the purpose. It's notable that the associated PE config address
256 * might be ready when calling the function. Therefore, endeavour to
257 * use the PE config address if possible. Further more, there're 2
258 * RTAS calls for the purpose, we need to try the new one and back
259 * to the old one if the new one couldn't work properly.
260 */
261static int pseries_eeh_get_state(struct device_node *dn, int *state)
262{
263 struct eeh_dev *edev;
264 int config_addr;
265 int ret;
266 int rets[4];
267 int result;
268
269 /* Figure out PE config address if possible */
270 edev = of_node_to_eeh_dev(dn);
271 config_addr = edev->config_addr;
272 if (edev->pe_config_addr)
273 config_addr = edev->pe_config_addr;
274
275 if (ibm_read_slot_reset_state2 != RTAS_UNKNOWN_SERVICE) {
276 ret = rtas_call(ibm_read_slot_reset_state2, 3, 4, rets,
277 config_addr, BUID_HI(edev->phb->buid),
278 BUID_LO(edev->phb->buid));
279 } else if (ibm_read_slot_reset_state != RTAS_UNKNOWN_SERVICE) {
280 /* Fake PE unavailable info */
281 rets[2] = 0;
282 ret = rtas_call(ibm_read_slot_reset_state, 3, 3, rets,
283 config_addr, BUID_HI(edev->phb->buid),
284 BUID_LO(edev->phb->buid));
285 } else {
286 return EEH_STATE_NOT_SUPPORT;
287 }
288
289 if (ret)
290 return ret;
291
292 /* Parse the result out */
293 result = 0;
294 if (rets[1]) {
295 switch(rets[0]) {
296 case 0:
297 result &= ~EEH_STATE_RESET_ACTIVE;
298 result |= EEH_STATE_MMIO_ACTIVE;
299 result |= EEH_STATE_DMA_ACTIVE;
300 break;
301 case 1:
302 result |= EEH_STATE_RESET_ACTIVE;
303 result |= EEH_STATE_MMIO_ACTIVE;
304 result |= EEH_STATE_DMA_ACTIVE;
305 break;
306 case 2:
307 result &= ~EEH_STATE_RESET_ACTIVE;
308 result &= ~EEH_STATE_MMIO_ACTIVE;
309 result &= ~EEH_STATE_DMA_ACTIVE;
310 break;
311 case 4:
312 result &= ~EEH_STATE_RESET_ACTIVE;
313 result &= ~EEH_STATE_MMIO_ACTIVE;
314 result &= ~EEH_STATE_DMA_ACTIVE;
315 result |= EEH_STATE_MMIO_ENABLED;
316 break;
317 case 5:
318 if (rets[2]) {
319 if (state) *state = rets[2];
320 result = EEH_STATE_UNAVAILABLE;
321 } else {
322 result = EEH_STATE_NOT_SUPPORT;
323 }
324 default:
325 result = EEH_STATE_NOT_SUPPORT;
326 }
327 } else {
328 result = EEH_STATE_NOT_SUPPORT;
329 }
330
331 return result;
332}
333
334/**
335 * pseries_eeh_reset - Reset the specified PE
336 * @dn: PE associated device node
337 * @option: reset option
338 *
339 * Reset the specified PE
340 */
341static int pseries_eeh_reset(struct device_node *dn, int option)
342{
343 struct eeh_dev *edev;
344 int config_addr;
345 int ret;
346
347 /* Figure out PE address */
348 edev = of_node_to_eeh_dev(dn);
349 config_addr = edev->config_addr;
350 if (edev->pe_config_addr)
351 config_addr = edev->pe_config_addr;
352
353 /* Reset PE through RTAS call */
354 ret = rtas_call(ibm_set_slot_reset, 4, 1, NULL,
355 config_addr, BUID_HI(edev->phb->buid),
356 BUID_LO(edev->phb->buid), option);
357
358 /* If fundamental-reset not supported, try hot-reset */
359 if (option == EEH_RESET_FUNDAMENTAL &&
360 ret == -8) {
361 ret = rtas_call(ibm_set_slot_reset, 4, 1, NULL,
362 config_addr, BUID_HI(edev->phb->buid),
363 BUID_LO(edev->phb->buid), EEH_RESET_HOT);
364 }
365
366 return ret;
367}
368
369/**
370 * pseries_eeh_wait_state - Wait for PE state
371 * @dn: PE associated device node
372 * @max_wait: maximal period in microsecond
373 *
374 * Wait for the state of associated PE. It might take some time
375 * to retrieve the PE's state.
376 */
377static int pseries_eeh_wait_state(struct device_node *dn, int max_wait)
378{
379 int ret;
380 int mwait;
381
382 /*
383 * According to PAPR, the state of PE might be temporarily
384 * unavailable. Under the circumstance, we have to wait
385 * for indicated time determined by firmware. The maximal
386 * wait time is 5 minutes, which is acquired from the original
387 * EEH implementation. Also, the original implementation
388 * also defined the minimal wait time as 1 second.
389 */
390#define EEH_STATE_MIN_WAIT_TIME (1000)
391#define EEH_STATE_MAX_WAIT_TIME (300 * 1000)
392
393 while (1) {
394 ret = pseries_eeh_get_state(dn, &mwait);
395
396 /*
397 * If the PE's state is temporarily unavailable,
398 * we have to wait for the specified time. Otherwise,
399 * the PE's state will be returned immediately.
400 */
401 if (ret != EEH_STATE_UNAVAILABLE)
402 return ret;
403
404 if (max_wait <= 0) {
405 pr_warning("%s: Timeout when getting PE's state (%d)\n",
406 __func__, max_wait);
407 return EEH_STATE_NOT_SUPPORT;
408 }
409
410 if (mwait <= 0) {
411 pr_warning("%s: Firmware returned bad wait value %d\n",
412 __func__, mwait);
413 mwait = EEH_STATE_MIN_WAIT_TIME;
414 } else if (mwait > EEH_STATE_MAX_WAIT_TIME) {
415 pr_warning("%s: Firmware returned too long wait value %d\n",
416 __func__, mwait);
417 mwait = EEH_STATE_MAX_WAIT_TIME;
418 }
419
420 max_wait -= mwait;
421 msleep(mwait);
422 }
423
424 return EEH_STATE_NOT_SUPPORT;
425}
426
427/**
428 * pseries_eeh_get_log - Retrieve error log
429 * @dn: device node
430 * @severity: temporary or permanent error log
431 * @drv_log: driver log to be combined with retrieved error log
432 * @len: length of driver log
433 *
434 * Retrieve the temporary or permanent error from the PE.
435 * Actually, the error will be retrieved through the dedicated
436 * RTAS call.
437 */
438static int pseries_eeh_get_log(struct device_node *dn, int severity, char *drv_log, unsigned long len)
439{
440 struct eeh_dev *edev;
441 int config_addr;
442 unsigned long flags;
443 int ret;
444
445 edev = of_node_to_eeh_dev(dn);
446 spin_lock_irqsave(&slot_errbuf_lock, flags);
447 memset(slot_errbuf, 0, eeh_error_buf_size);
448
449 /* Figure out the PE address */
450 config_addr = edev->config_addr;
451 if (edev->pe_config_addr)
452 config_addr = edev->pe_config_addr;
453
454 ret = rtas_call(ibm_slot_error_detail, 8, 1, NULL, config_addr,
455 BUID_HI(edev->phb->buid), BUID_LO(edev->phb->buid),
456 virt_to_phys(drv_log), len,
457 virt_to_phys(slot_errbuf), eeh_error_buf_size,
458 severity);
459 if (!ret)
460 log_error(slot_errbuf, ERR_TYPE_RTAS_LOG, 0);
461 spin_unlock_irqrestore(&slot_errbuf_lock, flags);
462
463 return ret;
464}
465
466/**
467 * pseries_eeh_configure_bridge - Configure PCI bridges in the indicated PE
468 * @dn: PE associated device node
469 *
470 * The function will be called to reconfigure the bridges included
471 * in the specified PE so that the mulfunctional PE would be recovered
472 * again.
473 */
474static int pseries_eeh_configure_bridge(struct device_node *dn)
475{
476 struct eeh_dev *edev;
477 int config_addr;
478 int ret;
479
480 /* Figure out the PE address */
481 edev = of_node_to_eeh_dev(dn);
482 config_addr = edev->config_addr;
483 if (edev->pe_config_addr)
484 config_addr = edev->pe_config_addr;
485
486 /* Use new configure-pe function, if supported */
487 if (ibm_configure_pe != RTAS_UNKNOWN_SERVICE) {
488 ret = rtas_call(ibm_configure_pe, 3, 1, NULL,
489 config_addr, BUID_HI(edev->phb->buid),
490 BUID_LO(edev->phb->buid));
491 } else if (ibm_configure_bridge != RTAS_UNKNOWN_SERVICE) {
492 ret = rtas_call(ibm_configure_bridge, 3, 1, NULL,
493 config_addr, BUID_HI(edev->phb->buid),
494 BUID_LO(edev->phb->buid));
495 } else {
496 return -EFAULT;
497 }
498
499 if (ret)
500 pr_warning("%s: Unable to configure bridge %d for %s\n",
501 __func__, ret, dn->full_name);
502
503 return ret;
504}
505
506/**
507 * pseries_eeh_read_config - Read PCI config space
508 * @dn: device node
509 * @where: PCI address
510 * @size: size to read
511 * @val: return value
512 *
513 * Read config space from the speicifed device
514 */
515static int pseries_eeh_read_config(struct device_node *dn, int where, int size, u32 *val)
516{
517 struct pci_dn *pdn;
518
519 pdn = PCI_DN(dn);
520
521 return rtas_read_config(pdn, where, size, val);
522}
523
524/**
525 * pseries_eeh_write_config - Write PCI config space
526 * @dn: device node
527 * @where: PCI address
528 * @size: size to write
529 * @val: value to be written
530 *
531 * Write config space to the specified device
532 */
533static int pseries_eeh_write_config(struct device_node *dn, int where, int size, u32 val)
534{
535 struct pci_dn *pdn;
536
537 pdn = PCI_DN(dn);
538
539 return rtas_write_config(pdn, where, size, val);
540}
541
542static struct eeh_ops pseries_eeh_ops = {
543 .name = "pseries",
544 .init = pseries_eeh_init,
545 .set_option = pseries_eeh_set_option,
546 .get_pe_addr = pseries_eeh_get_pe_addr,
547 .get_state = pseries_eeh_get_state,
548 .reset = pseries_eeh_reset,
549 .wait_state = pseries_eeh_wait_state,
550 .get_log = pseries_eeh_get_log,
551 .configure_bridge = pseries_eeh_configure_bridge,
552 .read_config = pseries_eeh_read_config,
553 .write_config = pseries_eeh_write_config
554};
555
556/**
557 * eeh_pseries_init - Register platform dependent EEH operations
558 *
559 * EEH initialization on pseries platform. This function should be
560 * called before any EEH related functions.
561 */
562int __init eeh_pseries_init(void)
563{
564 return eeh_ops_register(&pseries_eeh_ops);
565}
diff --git a/arch/powerpc/platforms/pseries/eeh_sysfs.c b/arch/powerpc/platforms/pseries/eeh_sysfs.c
index eb744ee234d..243b3510d70 100644
--- a/arch/powerpc/platforms/pseries/eeh_sysfs.c
+++ b/arch/powerpc/platforms/pseries/eeh_sysfs.c
@@ -28,7 +28,7 @@
28#include <asm/pci-bridge.h> 28#include <asm/pci-bridge.h>
29 29
30/** 30/**
31 * EEH_SHOW_ATTR -- create sysfs entry for eeh statistic 31 * EEH_SHOW_ATTR -- Create sysfs entry for eeh statistic
32 * @_name: name of file in sysfs directory 32 * @_name: name of file in sysfs directory
33 * @_memb: name of member in struct pci_dn to access 33 * @_memb: name of member in struct pci_dn to access
34 * @_format: printf format for display 34 * @_format: printf format for display
@@ -41,24 +41,21 @@ static ssize_t eeh_show_##_name(struct device *dev, \
41 struct device_attribute *attr, char *buf) \ 41 struct device_attribute *attr, char *buf) \
42{ \ 42{ \
43 struct pci_dev *pdev = to_pci_dev(dev); \ 43 struct pci_dev *pdev = to_pci_dev(dev); \
44 struct device_node *dn = pci_device_to_OF_node(pdev); \ 44 struct eeh_dev *edev = pci_dev_to_eeh_dev(pdev); \
45 struct pci_dn *pdn; \
46 \ 45 \
47 if (!dn || PCI_DN(dn) == NULL) \ 46 if (!edev) \
48 return 0; \ 47 return 0; \
49 \ 48 \
50 pdn = PCI_DN(dn); \ 49 return sprintf(buf, _format "\n", edev->_memb); \
51 return sprintf(buf, _format "\n", pdn->_memb); \
52} \ 50} \
53static DEVICE_ATTR(_name, S_IRUGO, eeh_show_##_name, NULL); 51static DEVICE_ATTR(_name, S_IRUGO, eeh_show_##_name, NULL);
54 52
55 53EEH_SHOW_ATTR(eeh_mode, mode, "0x%x");
56EEH_SHOW_ATTR(eeh_mode, eeh_mode, "0x%x"); 54EEH_SHOW_ATTR(eeh_config_addr, config_addr, "0x%x");
57EEH_SHOW_ATTR(eeh_config_addr, eeh_config_addr, "0x%x"); 55EEH_SHOW_ATTR(eeh_pe_config_addr, pe_config_addr, "0x%x");
58EEH_SHOW_ATTR(eeh_pe_config_addr, eeh_pe_config_addr, "0x%x"); 56EEH_SHOW_ATTR(eeh_check_count, check_count, "%d" );
59EEH_SHOW_ATTR(eeh_check_count, eeh_check_count, "%d"); 57EEH_SHOW_ATTR(eeh_freeze_count, freeze_count, "%d" );
60EEH_SHOW_ATTR(eeh_freeze_count, eeh_freeze_count, "%d"); 58EEH_SHOW_ATTR(eeh_false_positives, false_positives, "%d" );
61EEH_SHOW_ATTR(eeh_false_positives, eeh_false_positives, "%d");
62 59
63void eeh_sysfs_add_device(struct pci_dev *pdev) 60void eeh_sysfs_add_device(struct pci_dev *pdev)
64{ 61{
diff --git a/arch/powerpc/platforms/pseries/msi.c b/arch/powerpc/platforms/pseries/msi.c
index 38d24e7e7bb..109fdb75578 100644
--- a/arch/powerpc/platforms/pseries/msi.c
+++ b/arch/powerpc/platforms/pseries/msi.c
@@ -217,7 +217,7 @@ static struct device_node *find_pe_dn(struct pci_dev *dev, int *total)
217 if (!dn) 217 if (!dn)
218 return NULL; 218 return NULL;
219 219
220 dn = find_device_pe(dn); 220 dn = eeh_find_device_pe(dn);
221 if (!dn) 221 if (!dn)
222 return NULL; 222 return NULL;
223 223
diff --git a/arch/powerpc/platforms/pseries/pci_dlpar.c b/arch/powerpc/platforms/pseries/pci_dlpar.c
index 55d4ec1bd1a..fbb21fc3080 100644
--- a/arch/powerpc/platforms/pseries/pci_dlpar.c
+++ b/arch/powerpc/platforms/pseries/pci_dlpar.c
@@ -147,6 +147,9 @@ struct pci_controller * __devinit init_phb_dynamic(struct device_node *dn)
147 147
148 pci_devs_phb_init_dynamic(phb); 148 pci_devs_phb_init_dynamic(phb);
149 149
150 /* Create EEH devices for the PHB */
151 eeh_dev_phb_init_dynamic(phb);
152
150 if (dn->child) 153 if (dn->child)
151 eeh_add_device_tree_early(dn); 154 eeh_add_device_tree_early(dn);
152 155
diff --git a/arch/powerpc/platforms/pseries/setup.c b/arch/powerpc/platforms/pseries/setup.c
index d928412cfb3..8f137af616a 100644
--- a/arch/powerpc/platforms/pseries/setup.c
+++ b/arch/powerpc/platforms/pseries/setup.c
@@ -260,8 +260,12 @@ static int pci_dn_reconfig_notifier(struct notifier_block *nb, unsigned long act
260 switch (action) { 260 switch (action) {
261 case PSERIES_RECONFIG_ADD: 261 case PSERIES_RECONFIG_ADD:
262 pci = np->parent->data; 262 pci = np->parent->data;
263 if (pci) 263 if (pci) {
264 update_dn_pci_info(np, pci->phb); 264 update_dn_pci_info(np, pci->phb);
265
266 /* Create EEH device for the OF node */
267 eeh_dev_init(np, pci->phb);
268 }
265 break; 269 break;
266 default: 270 default:
267 err = NOTIFY_DONE; 271 err = NOTIFY_DONE;
@@ -381,6 +385,7 @@ static void __init pSeries_setup_arch(void)
381 385
382 /* Find and initialize PCI host bridges */ 386 /* Find and initialize PCI host bridges */
383 init_pci_config_tokens(); 387 init_pci_config_tokens();
388 eeh_pseries_init();
384 find_and_init_phbs(); 389 find_and_init_phbs();
385 pSeries_reconfig_notifier_register(&pci_dn_reconfig_nb); 390 pSeries_reconfig_notifier_register(&pci_dn_reconfig_nb);
386 eeh_init(); 391 eeh_init();
diff --git a/include/linux/of.h b/include/linux/of.h
index a75a831e205..3e710d87808 100644
--- a/include/linux/of.h
+++ b/include/linux/of.h
@@ -58,6 +58,9 @@ struct device_node {
58 struct kref kref; 58 struct kref kref;
59 unsigned long _flags; 59 unsigned long _flags;
60 void *data; 60 void *data;
61#if defined(CONFIG_EEH)
62 struct eeh_dev *edev;
63#endif
61#if defined(CONFIG_SPARC) 64#if defined(CONFIG_SPARC)
62 char *path_component_name; 65 char *path_component_name;
63 unsigned int unique_id; 66 unsigned int unique_id;
@@ -72,6 +75,13 @@ struct of_phandle_args {
72 uint32_t args[MAX_PHANDLE_ARGS]; 75 uint32_t args[MAX_PHANDLE_ARGS];
73}; 76};
74 77
78#if defined(CONFIG_EEH)
79static inline struct eeh_dev *of_node_to_eeh_dev(struct device_node *dn)
80{
81 return dn->edev;
82}
83#endif
84
75#if defined(CONFIG_SPARC) || !defined(CONFIG_OF) 85#if defined(CONFIG_SPARC) || !defined(CONFIG_OF)
76/* Dummy ref counting routines - to be implemented later */ 86/* Dummy ref counting routines - to be implemented later */
77static inline struct device_node *of_node_get(struct device_node *node) 87static inline struct device_node *of_node_get(struct device_node *node)
diff --git a/include/linux/pci.h b/include/linux/pci.h
index a16b1df3def..cfeee2a6a5a 100644
--- a/include/linux/pci.h
+++ b/include/linux/pci.h
@@ -1647,6 +1647,13 @@ static inline void pci_set_bus_of_node(struct pci_bus *bus) { }
1647static inline void pci_release_bus_of_node(struct pci_bus *bus) { } 1647static inline void pci_release_bus_of_node(struct pci_bus *bus) { }
1648#endif /* CONFIG_OF */ 1648#endif /* CONFIG_OF */
1649 1649
1650#ifdef CONFIG_EEH
1651static inline struct eeh_dev *pci_dev_to_eeh_dev(struct pci_dev *pdev)
1652{
1653 return pdev->dev.archdata.edev;
1654}
1655#endif
1656
1650/** 1657/**
1651 * pci_find_upstream_pcie_bridge - find upstream PCIe-to-PCI bridge of a device 1658 * pci_find_upstream_pcie_bridge - find upstream PCIe-to-PCI bridge of a device
1652 * @pdev: the PCI device 1659 * @pdev: the PCI device
generated by cgit v1.2.2 (git 2.25.0) at 2024-10-22 16:53:01 -0400