aboutsummaryrefslogtreecommitdiffstats
path: root/drivers/pci/pcie/aer
diff options
context:
space:
mode:
authorZhang, Yanmin <yanmin.zhang@intel.com>2006-07-31 03:21:33 -0400
committerGreg Kroah-Hartman <gregkh@suse.de>2006-09-26 20:43:53 -0400
commit6c2b374d74857e892080ee726184ec1d15e7d4e4 (patch)
treec107532c288bcede80e45ebc3e46292bfaf0cea2 /drivers/pci/pcie/aer
parent48408157ebf5b2c6dc1e04ba5d258012f6a7f356 (diff)
PCI-Express AER implemetation: AER core and aerdriver
Patch 3 implements the core part of PCI-Express AER and aerdrv port service driver. When a root port service device is probed, the aerdrv will call request_irq to register irq handler for AER error interrupt. When a device sends an PCI-Express error message to the root port, the root port will trigger an interrupt, by either MSI or IO-APIC, then kernel would run the irq handler. The handler collects root error status register and schedules a work. The work will call the core part to process the error based on its type (Correctable/non-fatal/fatal). As for Correctable errors, the patch chooses to just clear the correctable error status register of the device. As for the non-fatal error, the patch follows generic PCI error handler rules to call the error callback functions of the endpoint's driver. If the device is a bridge, the patch chooses to broadcast the error to downstream devices. As for the fatal error, the patch resets the pci-express link and follows generic PCI error handler rules to call the error callback functions of the endpoint's driver. If the device is a bridge, the patch chooses to broadcast the error to downstream devices. Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Diffstat (limited to 'drivers/pci/pcie/aer')
-rw-r--r--drivers/pci/pcie/aer/Kconfig12
-rw-r--r--drivers/pci/pcie/aer/Makefile8
-rw-r--r--drivers/pci/pcie/aer/aerdrv.c346
-rw-r--r--drivers/pci/pcie/aer/aerdrv.h125
-rw-r--r--drivers/pci/pcie/aer/aerdrv_acpi.c68
-rw-r--r--drivers/pci/pcie/aer/aerdrv_core.c757
-rw-r--r--drivers/pci/pcie/aer/aerdrv_errprint.c248
7 files changed, 1564 insertions, 0 deletions
diff --git a/drivers/pci/pcie/aer/Kconfig b/drivers/pci/pcie/aer/Kconfig
new file mode 100644
index 000000000000..3f37a60a6438
--- /dev/null
+++ b/drivers/pci/pcie/aer/Kconfig
@@ -0,0 +1,12 @@
1#
2# PCI Express Root Port Device AER Configuration
3#
4
5config PCIEAER
6 boolean "Root Port Advanced Error Reporting support"
7 depends on PCIEPORTBUS && ACPI
8 default y
9 help
10 This enables PCI Express Root Port Advanced Error Reporting
11 (AER) driver support. Error reporting messages sent to Root
12 Port will be handled by PCI Express AER driver.
diff --git a/drivers/pci/pcie/aer/Makefile b/drivers/pci/pcie/aer/Makefile
new file mode 100644
index 000000000000..15a4f40d520b
--- /dev/null
+++ b/drivers/pci/pcie/aer/Makefile
@@ -0,0 +1,8 @@
1#
2# Makefile for PCI-Express Root Port Advanced Error Reporting Driver
3#
4
5obj-$(CONFIG_PCIEAER) += aerdriver.o
6
7aerdriver-objs := aerdrv_errprint.o aerdrv_core.o aerdrv.o aerdrv_acpi.o
8
diff --git a/drivers/pci/pcie/aer/aerdrv.c b/drivers/pci/pcie/aer/aerdrv.c
new file mode 100644
index 000000000000..0d4ac027d53e
--- /dev/null
+++ b/drivers/pci/pcie/aer/aerdrv.c
@@ -0,0 +1,346 @@
1/*
2 * drivers/pci/pcie/aer/aerdrv.c
3 *
4 * This file is subject to the terms and conditions of the GNU General Public
5 * License. See the file "COPYING" in the main directory of this archive
6 * for more details.
7 *
8 * This file implements the AER root port service driver. The driver will
9 * register an irq handler. When root port triggers an AER interrupt, the irq
10 * handler will collect root port status and schedule a work.
11 *
12 * Copyright (C) 2006 Intel Corp.
13 * Tom Long Nguyen (tom.l.nguyen@intel.com)
14 * Zhang Yanmin (yanmin.zhang@intel.com)
15 *
16 */
17
18#include <linux/module.h>
19#include <linux/pci.h>
20#include <linux/kernel.h>
21#include <linux/errno.h>
22#include <linux/pm.h>
23#include <linux/init.h>
24#include <linux/interrupt.h>
25#include <linux/delay.h>
26#include <linux/pcieport_if.h>
27
28#include "aerdrv.h"
29
30/*
31 * Version Information
32 */
33#define DRIVER_VERSION "v1.0"
34#define DRIVER_AUTHOR "tom.l.nguyen@intel.com"
35#define DRIVER_DESC "Root Port Advanced Error Reporting Driver"
36MODULE_AUTHOR(DRIVER_AUTHOR);
37MODULE_DESCRIPTION(DRIVER_DESC);
38MODULE_LICENSE("GPL");
39
40static int __devinit aer_probe (struct pcie_device *dev,
41 const struct pcie_port_service_id *id );
42static void aer_remove(struct pcie_device *dev);
43static int aer_suspend(struct pcie_device *dev, pm_message_t state)
44{return 0;}
45static int aer_resume(struct pcie_device *dev) {return 0;}
46static pci_ers_result_t aer_error_detected(struct pci_dev *dev,
47 enum pci_channel_state error);
48static void aer_error_resume(struct pci_dev *dev);
49static pci_ers_result_t aer_root_reset(struct pci_dev *dev);
50
51/*
52 * PCI Express bus's AER Root service driver data structure
53 */
54static struct pcie_port_service_id aer_id[] = {
55 {
56 .vendor = PCI_ANY_ID,
57 .device = PCI_ANY_ID,
58 .port_type = PCIE_RC_PORT,
59 .service_type = PCIE_PORT_SERVICE_AER,
60 },
61 { /* end: all zeroes */ }
62};
63
64static struct pci_error_handlers aer_error_handlers = {
65 .error_detected = aer_error_detected,
66 .resume = aer_error_resume,
67};
68
69static struct pcie_port_service_driver aerdrv = {
70 .name = "aer",
71 .id_table = &aer_id[0],
72
73 .probe = aer_probe,
74 .remove = aer_remove,
75
76 .suspend = aer_suspend,
77 .resume = aer_resume,
78
79 .err_handler = &aer_error_handlers,
80
81 .reset_link = aer_root_reset,
82};
83
84/**
85 * aer_irq - Root Port's ISR
86 * @irq: IRQ assigned to Root Port
87 * @context: pointer to Root Port data structure
88 * @r: pointer struct pt_regs
89 *
90 * Invoked when Root Port detects AER messages.
91 **/
92static irqreturn_t aer_irq(int irq, void *context, struct pt_regs * r)
93{
94 unsigned int status, id;
95 struct pcie_device *pdev = (struct pcie_device *)context;
96 struct aer_rpc *rpc = get_service_data(pdev);
97 int next_prod_idx;
98 unsigned long flags;
99 int pos;
100
101 pos = pci_find_aer_capability(pdev->port);
102 /*
103 * Must lock access to Root Error Status Reg, Root Error ID Reg,
104 * and Root error producer/consumer index
105 */
106 spin_lock_irqsave(&rpc->e_lock, flags);
107
108 /* Read error status */
109 pci_read_config_dword(pdev->port, pos + PCI_ERR_ROOT_STATUS, &status);
110 if (!(status & ROOT_ERR_STATUS_MASKS)) {
111 spin_unlock_irqrestore(&rpc->e_lock, flags);
112 return IRQ_NONE;
113 }
114
115 /* Read error source and clear error status */
116 pci_read_config_dword(pdev->port, pos + PCI_ERR_ROOT_COR_SRC, &id);
117 pci_write_config_dword(pdev->port, pos + PCI_ERR_ROOT_STATUS, status);
118
119 /* Store error source for later DPC handler */
120 next_prod_idx = rpc->prod_idx + 1;
121 if (next_prod_idx == AER_ERROR_SOURCES_MAX)
122 next_prod_idx = 0;
123 if (next_prod_idx == rpc->cons_idx) {
124 /*
125 * Error Storm Condition - possibly the same error occurred.
126 * Drop the error.
127 */
128 spin_unlock_irqrestore(&rpc->e_lock, flags);
129 return IRQ_HANDLED;
130 }
131 rpc->e_sources[rpc->prod_idx].status = status;
132 rpc->e_sources[rpc->prod_idx].id = id;
133 rpc->prod_idx = next_prod_idx;
134 spin_unlock_irqrestore(&rpc->e_lock, flags);
135
136 /* Invoke DPC handler */
137 schedule_work(&rpc->dpc_handler);
138
139 return IRQ_HANDLED;
140}
141
142/**
143 * aer_alloc_rpc - allocate Root Port data structure
144 * @dev: pointer to the pcie_dev data structure
145 *
146 * Invoked when Root Port's AER service is loaded.
147 **/
148static struct aer_rpc* aer_alloc_rpc(struct pcie_device *dev)
149{
150 struct aer_rpc *rpc;
151
152 if (!(rpc = (struct aer_rpc *)kmalloc(sizeof(struct aer_rpc),
153 GFP_KERNEL)))
154 return NULL;
155
156 memset(rpc, 0, sizeof(struct aer_rpc));
157 /*
158 * Initialize Root lock access, e_lock, to Root Error Status Reg,
159 * Root Error ID Reg, and Root error producer/consumer index.
160 */
161 rpc->e_lock = SPIN_LOCK_UNLOCKED;
162
163 rpc->rpd = dev;
164 INIT_WORK(&rpc->dpc_handler, aer_isr, (void *)dev);
165 rpc->prod_idx = rpc->cons_idx = 0;
166 mutex_init(&rpc->rpc_mutex);
167 init_waitqueue_head(&rpc->wait_release);
168
169 /* Use PCIE bus function to store rpc into PCIE device */
170 set_service_data(dev, rpc);
171
172 return rpc;
173}
174
175/**
176 * aer_remove - clean up resources
177 * @dev: pointer to the pcie_dev data structure
178 *
179 * Invoked when PCI Express bus unloads or AER probe fails.
180 **/
181static void aer_remove(struct pcie_device *dev)
182{
183 struct aer_rpc *rpc = get_service_data(dev);
184
185 if (rpc) {
186 /* If register interrupt service, it must be free. */
187 if (rpc->isr)
188 free_irq(dev->irq, dev);
189
190 wait_event(rpc->wait_release, rpc->prod_idx == rpc->cons_idx);
191
192 aer_delete_rootport(rpc);
193 set_service_data(dev, NULL);
194 }
195}
196
197/**
198 * aer_probe - initialize resources
199 * @dev: pointer to the pcie_dev data structure
200 * @id: pointer to the service id data structure
201 *
202 * Invoked when PCI Express bus loads AER service driver.
203 **/
204static int __devinit aer_probe (struct pcie_device *dev,
205 const struct pcie_port_service_id *id )
206{
207 int status;
208 struct aer_rpc *rpc;
209 struct device *device = &dev->device;
210
211 /* Init */
212 if ((status = aer_init(dev)))
213 return status;
214
215 /* Alloc rpc data structure */
216 if (!(rpc = aer_alloc_rpc(dev))) {
217 printk(KERN_DEBUG "%s: Alloc rpc fails on PCIE device[%s]\n",
218 __FUNCTION__, device->bus_id);
219 aer_remove(dev);
220 return -ENOMEM;
221 }
222
223 /* Request IRQ ISR */
224 if ((status = request_irq(dev->irq, aer_irq, SA_SHIRQ, "aerdrv",
225 dev))) {
226 printk(KERN_DEBUG "%s: Request ISR fails on PCIE device[%s]\n",
227 __FUNCTION__, device->bus_id);
228 aer_remove(dev);
229 return status;
230 }
231
232 rpc->isr = 1;
233
234 aer_enable_rootport(rpc);
235
236 return status;
237}
238
239/**
240 * aer_root_reset - reset link on Root Port
241 * @dev: pointer to Root Port's pci_dev data structure
242 *
243 * Invoked by Port Bus driver when performing link reset at Root Port.
244 **/
245static pci_ers_result_t aer_root_reset(struct pci_dev *dev)
246{
247 u16 p2p_ctrl;
248 u32 status;
249 int pos;
250
251 pos = pci_find_aer_capability(dev);
252
253 /* Disable Root's interrupt in response to error messages */
254 pci_write_config_dword(dev, pos + PCI_ERR_ROOT_COMMAND, 0);
255
256 /* Assert Secondary Bus Reset */
257 pci_read_config_word(dev, PCI_BRIDGE_CONTROL, &p2p_ctrl);
258 p2p_ctrl |= PCI_CB_BRIDGE_CTL_CB_RESET;
259 pci_write_config_word(dev, PCI_BRIDGE_CONTROL, p2p_ctrl);
260
261 /* De-assert Secondary Bus Reset */
262 p2p_ctrl &= ~PCI_CB_BRIDGE_CTL_CB_RESET;
263 pci_write_config_word(dev, PCI_BRIDGE_CONTROL, p2p_ctrl);
264
265 /*
266 * System software must wait for at least 100ms from the end
267 * of a reset of one or more device before it is permitted
268 * to issue Configuration Requests to those devices.
269 */
270 msleep(200);
271 printk(KERN_DEBUG "Complete link reset at Root[%s]\n", dev->dev.bus_id);
272
273 /* Enable Root Port's interrupt in response to error messages */
274 pci_read_config_dword(dev, pos + PCI_ERR_ROOT_STATUS, &status);
275 pci_write_config_dword(dev, pos + PCI_ERR_ROOT_STATUS, status);
276 pci_write_config_dword(dev,
277 pos + PCI_ERR_ROOT_COMMAND,
278 ROOT_PORT_INTR_ON_MESG_MASK);
279
280 return PCI_ERS_RESULT_RECOVERED;
281}
282
283/**
284 * aer_error_detected - update severity status
285 * @dev: pointer to Root Port's pci_dev data structure
286 * @error: error severity being notified by port bus
287 *
288 * Invoked by Port Bus driver during error recovery.
289 **/
290static pci_ers_result_t aer_error_detected(struct pci_dev *dev,
291 enum pci_channel_state error)
292{
293 /* Root Port has no impact. Always recovers. */
294 return PCI_ERS_RESULT_CAN_RECOVER;
295}
296
297/**
298 * aer_error_resume - clean up corresponding error status bits
299 * @dev: pointer to Root Port's pci_dev data structure
300 *
301 * Invoked by Port Bus driver during nonfatal recovery.
302 **/
303static void aer_error_resume(struct pci_dev *dev)
304{
305 int pos;
306 u32 status, mask;
307 u16 reg16;
308
309 /* Clean up Root device status */
310 pos = pci_find_capability(dev, PCI_CAP_ID_EXP);
311 pci_read_config_word(dev, pos + PCI_EXP_DEVSTA, &reg16);
312 pci_write_config_word(dev, pos + PCI_EXP_DEVSTA, reg16);
313
314 /* Clean AER Root Error Status */
315 pos = pci_find_aer_capability(dev);
316 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, &status);
317 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_SEVER, &mask);
318 if (dev->error_state == pci_channel_io_normal)
319 status &= ~mask; /* Clear corresponding nonfatal bits */
320 else
321 status &= mask; /* Clear corresponding fatal bits */
322 pci_write_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, status);
323}
324
325/**
326 * aer_service_init - register AER root service driver
327 *
328 * Invoked when AER root service driver is loaded.
329 **/
330static int __init aer_service_init(void)
331{
332 return pcie_port_service_register(&aerdrv);
333}
334
335/**
336 * aer_service_exit - unregister AER root service driver
337 *
338 * Invoked when AER root service driver is unloaded.
339 **/
340static void __exit aer_service_exit(void)
341{
342 pcie_port_service_unregister(&aerdrv);
343}
344
345module_init(aer_service_init);
346module_exit(aer_service_exit);
diff --git a/drivers/pci/pcie/aer/aerdrv.h b/drivers/pci/pcie/aer/aerdrv.h
new file mode 100644
index 000000000000..daf0cad88fc8
--- /dev/null
+++ b/drivers/pci/pcie/aer/aerdrv.h
@@ -0,0 +1,125 @@
1/*
2 * Copyright (C) 2006 Intel Corp.
3 * Tom Long Nguyen (tom.l.nguyen@intel.com)
4 * Zhang Yanmin (yanmin.zhang@intel.com)
5 *
6 */
7
8#ifndef _AERDRV_H_
9#define _AERDRV_H_
10
11#include <linux/pcieport_if.h>
12#include <linux/aer.h>
13
14#define AER_NONFATAL 0
15#define AER_FATAL 1
16#define AER_CORRECTABLE 2
17#define AER_UNCORRECTABLE 4
18#define AER_ERROR_MASK 0x001fffff
19#define AER_ERROR(d) (d & AER_ERROR_MASK)
20
21#define OSC_METHOD_RUN_SUCCESS 0
22#define OSC_METHOD_NOT_SUPPORTED 1
23#define OSC_METHOD_RUN_FAILURE 2
24
25/* Root Error Status Register Bits */
26#define ROOT_ERR_STATUS_MASKS 0x0f
27
28#define SYSTEM_ERROR_INTR_ON_MESG_MASK (PCI_EXP_RTCTL_SECEE| \
29 PCI_EXP_RTCTL_SENFEE| \
30 PCI_EXP_RTCTL_SEFEE)
31#define ROOT_PORT_INTR_ON_MESG_MASK (PCI_ERR_ROOT_CMD_COR_EN| \
32 PCI_ERR_ROOT_CMD_NONFATAL_EN| \
33 PCI_ERR_ROOT_CMD_FATAL_EN)
34#define ERR_COR_ID(d) (d & 0xffff)
35#define ERR_UNCOR_ID(d) (d >> 16)
36
37#define AER_SUCCESS 0
38#define AER_UNSUCCESS 1
39#define AER_ERROR_SOURCES_MAX 100
40
41#define AER_LOG_TLP_MASKS (PCI_ERR_UNC_POISON_TLP| \
42 PCI_ERR_UNC_ECRC| \
43 PCI_ERR_UNC_UNSUP| \
44 PCI_ERR_UNC_COMP_ABORT| \
45 PCI_ERR_UNC_UNX_COMP| \
46 PCI_ERR_UNC_MALF_TLP)
47
48/* AER Error Info Flags */
49#define AER_TLP_HEADER_VALID_FLAG 0x00000001
50#define AER_MULTI_ERROR_VALID_FLAG 0x00000002
51
52#define ERR_CORRECTABLE_ERROR_MASK 0x000031c1
53#define ERR_UNCORRECTABLE_ERROR_MASK 0x001ff010
54
55struct header_log_regs {
56 unsigned int dw0;
57 unsigned int dw1;
58 unsigned int dw2;
59 unsigned int dw3;
60};
61
62struct aer_err_info {
63 int severity; /* 0:NONFATAL | 1:FATAL | 2:COR */
64 int flags;
65 unsigned int status; /* COR/UNCOR Error Status */
66 struct header_log_regs tlp; /* TLP Header */
67};
68
69struct aer_err_source {
70 unsigned int status;
71 unsigned int id;
72};
73
74struct aer_rpc {
75 struct pcie_device *rpd; /* Root Port device */
76 struct work_struct dpc_handler;
77 struct aer_err_source e_sources[AER_ERROR_SOURCES_MAX];
78 unsigned short prod_idx; /* Error Producer Index */
79 unsigned short cons_idx; /* Error Consumer Index */
80 int isr;
81 spinlock_t e_lock; /*
82 * Lock access to Error Status/ID Regs
83 * and error producer/consumer index
84 */
85 struct mutex rpc_mutex; /*
86 * only one thread could do
87 * recovery on the same
88 * root port hierachy
89 */
90 wait_queue_head_t wait_release;
91};
92
93struct aer_broadcast_data {
94 enum pci_channel_state state;
95 enum pci_ers_result result;
96};
97
98static inline pci_ers_result_t merge_result(enum pci_ers_result orig,
99 enum pci_ers_result new)
100{
101 switch (orig) {
102 case PCI_ERS_RESULT_CAN_RECOVER:
103 case PCI_ERS_RESULT_RECOVERED:
104 orig = new;
105 break;
106 case PCI_ERS_RESULT_DISCONNECT:
107 if (new == PCI_ERS_RESULT_NEED_RESET)
108 orig = new;
109 break;
110 default:
111 break;
112 }
113
114 return orig;
115}
116
117extern struct bus_type pcie_port_bus_type;
118extern void aer_enable_rootport(struct aer_rpc *rpc);
119extern void aer_delete_rootport(struct aer_rpc *rpc);
120extern int aer_init(struct pcie_device *dev);
121extern void aer_isr(void *context);
122extern void aer_print_error(struct pci_dev *dev, struct aer_err_info *info);
123extern int aer_osc_setup(struct pci_dev *dev);
124
125#endif //_AERDRV_H_
diff --git a/drivers/pci/pcie/aer/aerdrv_acpi.c b/drivers/pci/pcie/aer/aerdrv_acpi.c
new file mode 100644
index 000000000000..fa68e89ebec9
--- /dev/null
+++ b/drivers/pci/pcie/aer/aerdrv_acpi.c
@@ -0,0 +1,68 @@
1/*
2 * Access ACPI _OSC method
3 *
4 * Copyright (C) 2006 Intel Corp.
5 * Tom Long Nguyen (tom.l.nguyen@intel.com)
6 * Zhang Yanmin (yanmin.zhang@intel.com)
7 *
8 */
9
10#include <linux/module.h>
11#include <linux/pci.h>
12#include <linux/kernel.h>
13#include <linux/errno.h>
14#include <linux/pm.h>
15#include <linux/suspend.h>
16#include <linux/acpi.h>
17#include <linux/pci-acpi.h>
18#include <linux/delay.h>
19#include "aerdrv.h"
20
21/**
22 * aer_osc_setup - run ACPI _OSC method
23 *
24 * Return:
25 * Zero if success. Nonzero for otherwise.
26 *
27 * Invoked when PCIE bus loads AER service driver. To avoid conflict with
28 * BIOS AER support requires BIOS to yield AER control to OS native driver.
29 **/
30int aer_osc_setup(struct pci_dev *dev)
31{
32 int retval = OSC_METHOD_RUN_SUCCESS;
33 acpi_status status;
34 acpi_handle handle = DEVICE_ACPI_HANDLE(&dev->dev);
35 struct pci_dev *pdev = dev;
36 struct pci_bus *parent;
37
38 while (!handle) {
39 if (!pdev || !pdev->bus->parent)
40 break;
41 parent = pdev->bus->parent;
42 if (!parent->self)
43 /* Parent must be a host bridge */
44 handle = acpi_get_pci_rootbridge_handle(
45 pci_domain_nr(parent),
46 parent->number);
47 else
48 handle = DEVICE_ACPI_HANDLE(
49 &(parent->self->dev));
50 pdev = parent->self;
51 }
52
53 if (!handle)
54 return OSC_METHOD_NOT_SUPPORTED;
55
56 pci_osc_support_set(OSC_EXT_PCI_CONFIG_SUPPORT);
57 status = pci_osc_control_set(handle, OSC_PCI_EXPRESS_AER_CONTROL |
58 OSC_PCI_EXPRESS_CAP_STRUCTURE_CONTROL);
59 if (ACPI_FAILURE(status)) {
60 if (status == AE_SUPPORT)
61 retval = OSC_METHOD_NOT_SUPPORTED;
62 else
63 retval = OSC_METHOD_RUN_FAILURE;
64 }
65
66 return retval;
67}
68
diff --git a/drivers/pci/pcie/aer/aerdrv_core.c b/drivers/pci/pcie/aer/aerdrv_core.c
new file mode 100644
index 000000000000..5591043acea7
--- /dev/null
+++ b/drivers/pci/pcie/aer/aerdrv_core.c
@@ -0,0 +1,757 @@
1/*
2 * drivers/pci/pcie/aer/aerdrv_core.c
3 *
4 * This file is subject to the terms and conditions of the GNU General Public
5 * License. See the file "COPYING" in the main directory of this archive
6 * for more details.
7 *
8 * This file implements the core part of PCI-Express AER. When an pci-express
9 * error is delivered, an error message will be collected and printed to
10 * console, then, an error recovery procedure will be executed by following
11 * the pci error recovery rules.
12 *
13 * Copyright (C) 2006 Intel Corp.
14 * Tom Long Nguyen (tom.l.nguyen@intel.com)
15 * Zhang Yanmin (yanmin.zhang@intel.com)
16 *
17 */
18
19#include <linux/module.h>
20#include <linux/pci.h>
21#include <linux/kernel.h>
22#include <linux/errno.h>
23#include <linux/pm.h>
24#include <linux/suspend.h>
25#include <linux/acpi.h>
26#include <linux/pci-acpi.h>
27#include <linux/delay.h>
28#include "aerdrv.h"
29
30static int forceload;
31module_param(forceload, bool, 0);
32
33#define PCI_CFG_SPACE_SIZE (0x100)
34int pci_find_aer_capability(struct pci_dev *dev)
35{
36 int pos;
37 u32 reg32 = 0;
38
39 /* Check if it's a pci-express device */
40 pos = pci_find_capability(dev, PCI_CAP_ID_EXP);
41 if (!pos)
42 return 0;
43
44 /* Check if it supports pci-express AER */
45 pos = PCI_CFG_SPACE_SIZE;
46 while (pos) {
47 if (pci_read_config_dword(dev, pos, &reg32))
48 return 0;
49
50 /* some broken boards return ~0 */
51 if (reg32 == 0xffffffff)
52 return 0;
53
54 if (PCI_EXT_CAP_ID(reg32) == PCI_EXT_CAP_ID_ERR)
55 break;
56
57 pos = reg32 >> 20;
58 }
59
60 return pos;
61}
62
63int pci_enable_pcie_error_reporting(struct pci_dev *dev)
64{
65 u16 reg16 = 0;
66 int pos;
67
68 pos = pci_find_capability(dev, PCI_CAP_ID_EXP);
69 if (!pos)
70 return -EIO;
71
72 pci_read_config_word(dev, pos+PCI_EXP_DEVCTL, &reg16);
73 reg16 = reg16 |
74 PCI_EXP_DEVCTL_CERE |
75 PCI_EXP_DEVCTL_NFERE |
76 PCI_EXP_DEVCTL_FERE |
77 PCI_EXP_DEVCTL_URRE;
78 pci_write_config_word(dev, pos+PCI_EXP_DEVCTL,
79 reg16);
80 return 0;
81}
82
83int pci_disable_pcie_error_reporting(struct pci_dev *dev)
84{
85 u16 reg16 = 0;
86 int pos;
87
88 pos = pci_find_capability(dev, PCI_CAP_ID_EXP);
89 if (!pos)
90 return -EIO;
91
92 pci_read_config_word(dev, pos+PCI_EXP_DEVCTL, &reg16);
93 reg16 = reg16 & ~(PCI_EXP_DEVCTL_CERE |
94 PCI_EXP_DEVCTL_NFERE |
95 PCI_EXP_DEVCTL_FERE |
96 PCI_EXP_DEVCTL_URRE);
97 pci_write_config_word(dev, pos+PCI_EXP_DEVCTL,
98 reg16);
99 return 0;
100}
101
102int pci_cleanup_aer_uncorrect_error_status(struct pci_dev *dev)
103{
104 int pos;
105 u32 status, mask;
106
107 pos = pci_find_aer_capability(dev);
108 if (!pos)
109 return -EIO;
110
111 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, &status);
112 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_SEVER, &mask);
113 if (dev->error_state == pci_channel_io_normal)
114 status &= ~mask; /* Clear corresponding nonfatal bits */
115 else
116 status &= mask; /* Clear corresponding fatal bits */
117 pci_write_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, status);
118
119 return 0;
120}
121
122static int find_device_iter(struct device *device, void *data)
123{
124 struct pci_dev *dev;
125 u16 id = *(unsigned long *)data;
126 u8 secondary, subordinate, d_bus = id >> 8;
127
128 if (device->bus == &pci_bus_type) {
129 dev = to_pci_dev(device);
130 if (id == ((dev->bus->number << 8) | dev->devfn)) {
131 /*
132 * Device ID match
133 */
134 *(unsigned long*)data = (unsigned long)device;
135 return 1;
136 }
137
138 /*
139 * If device is P2P, check if it is an upstream?
140 */
141 if (dev->hdr_type & PCI_HEADER_TYPE_BRIDGE) {
142 pci_read_config_byte(dev, PCI_SECONDARY_BUS,
143 &secondary);
144 pci_read_config_byte(dev, PCI_SUBORDINATE_BUS,
145 &subordinate);
146 if (d_bus >= secondary && d_bus <= subordinate) {
147 *(unsigned long*)data = (unsigned long)device;
148 return 1;
149 }
150 }
151 }
152
153 return 0;
154}
155
156/**
157 * find_source_device - search through device hierarchy for source device
158 * @p_dev: pointer to Root Port pci_dev data structure
159 * @id: device ID of agent who sends an error message to this Root Port
160 *
161 * Invoked when error is detected at the Root Port.
162 **/
163static struct device* find_source_device(struct pci_dev *parent, u16 id)
164{
165 struct pci_dev *dev = parent;
166 struct device *device;
167 unsigned long device_addr;
168 int status;
169
170 /* Is Root Port an agent that sends error message? */
171 if (id == ((dev->bus->number << 8) | dev->devfn))
172 return &dev->dev;
173
174 do {
175 device_addr = id;
176 if ((status = device_for_each_child(&dev->dev,
177 &device_addr, find_device_iter))) {
178 device = (struct device*)device_addr;
179 dev = to_pci_dev(device);
180 if (id == ((dev->bus->number << 8) | dev->devfn))
181 return device;
182 }
183 }while (status);
184
185 return NULL;
186}
187
188static void report_error_detected(struct pci_dev *dev, void *data)
189{
190 pci_ers_result_t vote;
191 struct pci_error_handlers *err_handler;
192 struct aer_broadcast_data *result_data;
193 result_data = (struct aer_broadcast_data *) data;
194
195 dev->error_state = result_data->state;
196
197 if (!dev->driver ||
198 !dev->driver->err_handler ||
199 !dev->driver->err_handler->error_detected) {
200 if (result_data->state == pci_channel_io_frozen &&
201 !(dev->hdr_type & PCI_HEADER_TYPE_BRIDGE)) {
202 /*
203 * In case of fatal recovery, if one of down-
204 * stream device has no driver. We might be
205 * unable to recover because a later insmod
206 * of a driver for this device is unaware of
207 * its hw state.
208 */
209 printk(KERN_DEBUG "Device ID[%s] has %s\n",
210 dev->dev.bus_id, (dev->driver) ?
211 "no AER-aware driver" : "no driver");
212 }
213 return;
214 }
215
216 err_handler = dev->driver->err_handler;
217 vote = err_handler->error_detected(dev, result_data->state);
218 result_data->result = merge_result(result_data->result, vote);
219 return;
220}
221
222static void report_mmio_enabled(struct pci_dev *dev, void *data)
223{
224 pci_ers_result_t vote;
225 struct pci_error_handlers *err_handler;
226 struct aer_broadcast_data *result_data;
227 result_data = (struct aer_broadcast_data *) data;
228
229 if (!dev->driver ||
230 !dev->driver->err_handler ||
231 !dev->driver->err_handler->mmio_enabled)
232 return;
233
234 err_handler = dev->driver->err_handler;
235 vote = err_handler->mmio_enabled(dev);
236 result_data->result = merge_result(result_data->result, vote);
237 return;
238}
239
240static void report_slot_reset(struct pci_dev *dev, void *data)
241{
242 pci_ers_result_t vote;
243 struct pci_error_handlers *err_handler;
244 struct aer_broadcast_data *result_data;
245 result_data = (struct aer_broadcast_data *) data;
246
247 if (!dev->driver ||
248 !dev->driver->err_handler ||
249 !dev->driver->err_handler->slot_reset)
250 return;
251
252 err_handler = dev->driver->err_handler;
253 vote = err_handler->slot_reset(dev);
254 result_data->result = merge_result(result_data->result, vote);
255 return;
256}
257
258static void report_resume(struct pci_dev *dev, void *data)
259{
260 struct pci_error_handlers *err_handler;
261
262 dev->error_state = pci_channel_io_normal;
263
264 if (!dev->driver ||
265 !dev->driver->err_handler ||
266 !dev->driver->err_handler->slot_reset)
267 return;
268
269 err_handler = dev->driver->err_handler;
270 err_handler->resume(dev);
271 return;
272}
273
274/**
275 * broadcast_error_message - handle message broadcast to downstream drivers
276 * @device: pointer to from where in a hierarchy message is broadcasted down
277 * @api: callback to be broadcasted
278 * @state: error state
279 *
280 * Invoked during error recovery process. Once being invoked, the content
281 * of error severity will be broadcasted to all downstream drivers in a
282 * hierarchy in question.
283 **/
284static pci_ers_result_t broadcast_error_message(struct pci_dev *dev,
285 enum pci_channel_state state,
286 char *error_mesg,
287 void (*cb)(struct pci_dev *, void *))
288{
289 struct aer_broadcast_data result_data;
290
291 printk(KERN_DEBUG "Broadcast %s message\n", error_mesg);
292 result_data.state = state;
293 if (cb == report_error_detected)
294 result_data.result = PCI_ERS_RESULT_CAN_RECOVER;
295 else
296 result_data.result = PCI_ERS_RESULT_RECOVERED;
297
298 if (dev->hdr_type & PCI_HEADER_TYPE_BRIDGE) {
299 /*
300 * If the error is reported by a bridge, we think this error
301 * is related to the downstream link of the bridge, so we
302 * do error recovery on all subordinates of the bridge instead
303 * of the bridge and clear the error status of the bridge.
304 */
305 if (cb == report_error_detected)
306 dev->error_state = state;
307 pci_walk_bus(dev->subordinate, cb, &result_data);
308 if (cb == report_resume) {
309 pci_cleanup_aer_uncorrect_error_status(dev);
310 dev->error_state = pci_channel_io_normal;
311 }
312 }
313 else {
314 /*
315 * If the error is reported by an end point, we think this
316 * error is related to the upstream link of the end point.
317 */
318 pci_walk_bus(dev->bus, cb, &result_data);
319 }
320
321 return result_data.result;
322}
323
324struct find_aer_service_data {
325 struct pcie_port_service_driver *aer_driver;
326 int is_downstream;
327};
328
329static int find_aer_service_iter(struct device *device, void *data)
330{
331 struct device_driver *driver;
332 struct pcie_port_service_driver *service_driver;
333 struct pcie_device *pcie_dev;
334 struct find_aer_service_data *result;
335
336 result = (struct find_aer_service_data *) data;
337
338 if (device->bus == &pcie_port_bus_type) {
339 pcie_dev = to_pcie_device(device);
340 if (pcie_dev->id.port_type == PCIE_SW_DOWNSTREAM_PORT)
341 result->is_downstream = 1;
342
343 driver = device->driver;
344 if (driver) {
345 service_driver = to_service_driver(driver);
346 if (service_driver->id_table->service_type ==
347 PCIE_PORT_SERVICE_AER) {
348 result->aer_driver = service_driver;
349 return 1;
350 }
351 }
352 }
353
354 return 0;
355}
356
357static void find_aer_service(struct pci_dev *dev,
358 struct find_aer_service_data *data)
359{
360 device_for_each_child(&dev->dev, data, find_aer_service_iter);
361}
362
363static pci_ers_result_t reset_link(struct pcie_device *aerdev,
364 struct pci_dev *dev)
365{
366 struct pci_dev *udev;
367 pci_ers_result_t status;
368 struct find_aer_service_data data;
369
370 if (dev->hdr_type & PCI_HEADER_TYPE_BRIDGE)
371 udev = dev;
372 else
373 udev= dev->bus->self;
374
375 data.is_downstream = 0;
376 data.aer_driver = NULL;
377 find_aer_service(udev, &data);
378
379 /*
380 * Use the aer driver of the error agent firstly.
381 * If it hasn't the aer driver, use the root port's
382 */
383 if (!data.aer_driver || !data.aer_driver->reset_link) {
384 if (data.is_downstream &&
385 aerdev->device.driver &&
386 to_service_driver(aerdev->device.driver)->reset_link) {
387 data.aer_driver =
388 to_service_driver(aerdev->device.driver);
389 } else {
390 printk(KERN_DEBUG "No link-reset support to Device ID"
391 "[%s]\n",
392 dev->dev.bus_id);
393 return PCI_ERS_RESULT_DISCONNECT;
394 }
395 }
396
397 status = data.aer_driver->reset_link(udev);
398 if (status != PCI_ERS_RESULT_RECOVERED) {
399 printk(KERN_DEBUG "Link reset at upstream Device ID"
400 "[%s] failed\n",
401 udev->dev.bus_id);
402 return PCI_ERS_RESULT_DISCONNECT;
403 }
404
405 return status;
406}
407
408/**
409 * do_recovery - handle nonfatal/fatal error recovery process
410 * @aerdev: pointer to a pcie_device data structure of root port
411 * @dev: pointer to a pci_dev data structure of agent detecting an error
412 * @severity: error severity type
413 *
414 * Invoked when an error is nonfatal/fatal. Once being invoked, broadcast
415 * error detected message to all downstream drivers within a hierarchy in
416 * question and return the returned code.
417 **/
418static pci_ers_result_t do_recovery(struct pcie_device *aerdev,
419 struct pci_dev *dev,
420 int severity)
421{
422 pci_ers_result_t status, result = PCI_ERS_RESULT_RECOVERED;
423 enum pci_channel_state state;
424
425 if (severity == AER_FATAL)
426 state = pci_channel_io_frozen;
427 else
428 state = pci_channel_io_normal;
429
430 status = broadcast_error_message(dev,
431 state,
432 "error_detected",
433 report_error_detected);
434
435 if (severity == AER_FATAL) {
436 result = reset_link(aerdev, dev);
437 if (result != PCI_ERS_RESULT_RECOVERED) {
438 /* TODO: Should panic here? */
439 return result;
440 }
441 }
442
443 if (status == PCI_ERS_RESULT_CAN_RECOVER)
444 status = broadcast_error_message(dev,
445 state,
446 "mmio_enabled",
447 report_mmio_enabled);
448
449 if (status == PCI_ERS_RESULT_NEED_RESET) {
450 /*
451 * TODO: Should call platform-specific
452 * functions to reset slot before calling
453 * drivers' slot_reset callbacks?
454 */
455 status = broadcast_error_message(dev,
456 state,
457 "slot_reset",
458 report_slot_reset);
459 }
460
461 if (status == PCI_ERS_RESULT_RECOVERED)
462 broadcast_error_message(dev,
463 state,
464 "resume",
465 report_resume);
466
467 return status;
468}
469
470/**
471 * handle_error_source - handle logging error into an event log
472 * @aerdev: pointer to pcie_device data structure of the root port
473 * @dev: pointer to pci_dev data structure of error source device
474 * @info: comprehensive error information
475 *
476 * Invoked when an error being detected by Root Port.
477 **/
478static void handle_error_source(struct pcie_device * aerdev,
479 struct pci_dev *dev,
480 struct aer_err_info info)
481{
482 pci_ers_result_t status = 0;
483 int pos;
484
485 if (info.severity == AER_CORRECTABLE) {
486 /*
487 * Correctable error does not need software intevention.
488 * No need to go through error recovery process.
489 */
490 pos = pci_find_aer_capability(dev);
491 if (pos)
492 pci_write_config_dword(dev, pos + PCI_ERR_COR_STATUS,
493 info.status);
494 } else {
495 status = do_recovery(aerdev, dev, info.severity);
496 if (status == PCI_ERS_RESULT_RECOVERED) {
497 printk(KERN_DEBUG "AER driver successfully recovered\n");
498 } else {
499 /* TODO: Should kernel panic here? */
500 printk(KERN_DEBUG "AER driver didn't recover\n");
501 }
502 }
503}
504
505/**
506 * aer_enable_rootport - enable Root Port's interrupts when receiving messages
507 * @rpc: pointer to a Root Port data structure
508 *
509 * Invoked when PCIE bus loads AER service driver.
510 **/
511void aer_enable_rootport(struct aer_rpc *rpc)
512{
513 struct pci_dev *pdev = rpc->rpd->port;
514 int pos, aer_pos;
515 u16 reg16;
516 u32 reg32;
517
518 pos = pci_find_capability(pdev, PCI_CAP_ID_EXP);
519 /* Clear PCIE Capability's Device Status */
520 pci_read_config_word(pdev, pos+PCI_EXP_DEVSTA, &reg16);
521 pci_write_config_word(pdev, pos+PCI_EXP_DEVSTA, reg16);
522
523 /* Disable system error generation in response to error messages */
524 pci_read_config_word(pdev, pos + PCI_EXP_RTCTL, &reg16);
525 reg16 &= ~(SYSTEM_ERROR_INTR_ON_MESG_MASK);
526 pci_write_config_word(pdev, pos + PCI_EXP_RTCTL, reg16);
527
528 aer_pos = pci_find_aer_capability(pdev);
529 /* Clear error status */
530 pci_read_config_dword(pdev, aer_pos + PCI_ERR_ROOT_STATUS, &reg32);
531 pci_write_config_dword(pdev, aer_pos + PCI_ERR_ROOT_STATUS, reg32);
532 pci_read_config_dword(pdev, aer_pos + PCI_ERR_COR_STATUS, &reg32);
533 pci_write_config_dword(pdev, aer_pos + PCI_ERR_COR_STATUS, reg32);
534 pci_read_config_dword(pdev, aer_pos + PCI_ERR_UNCOR_STATUS, &reg32);
535 pci_write_config_dword(pdev, aer_pos + PCI_ERR_UNCOR_STATUS, reg32);
536
537 /* Enable Root Port device reporting error itself */
538 pci_read_config_word(pdev, pos+PCI_EXP_DEVCTL, &reg16);
539 reg16 = reg16 |
540 PCI_EXP_DEVCTL_CERE |
541 PCI_EXP_DEVCTL_NFERE |
542 PCI_EXP_DEVCTL_FERE |
543 PCI_EXP_DEVCTL_URRE;
544 pci_write_config_word(pdev, pos+PCI_EXP_DEVCTL,
545 reg16);
546
547 /* Enable Root Port's interrupt in response to error messages */
548 pci_write_config_dword(pdev,
549 aer_pos + PCI_ERR_ROOT_COMMAND,
550 ROOT_PORT_INTR_ON_MESG_MASK);
551}
552
553/**
554 * disable_root_aer - disable Root Port's interrupts when receiving messages
555 * @rpc: pointer to a Root Port data structure
556 *
557 * Invoked when PCIE bus unloads AER service driver.
558 **/
559static void disable_root_aer(struct aer_rpc *rpc)
560{
561 struct pci_dev *pdev = rpc->rpd->port;
562 u32 reg32;
563 int pos;
564
565 pos = pci_find_aer_capability(pdev);
566 /* Disable Root's interrupt in response to error messages */
567 pci_write_config_dword(pdev, pos + PCI_ERR_ROOT_COMMAND, 0);
568
569 /* Clear Root's error status reg */
570 pci_read_config_dword(pdev, pos + PCI_ERR_ROOT_STATUS, &reg32);
571 pci_write_config_dword(pdev, pos + PCI_ERR_ROOT_STATUS, reg32);
572}
573
574/**
575 * get_e_source - retrieve an error source
576 * @rpc: pointer to the root port which holds an error
577 *
578 * Invoked by DPC handler to consume an error.
579 **/
580static struct aer_err_source* get_e_source(struct aer_rpc *rpc)
581{
582 struct aer_err_source *e_source;
583 unsigned long flags;
584
585 /* Lock access to Root error producer/consumer index */
586 spin_lock_irqsave(&rpc->e_lock, flags);
587 if (rpc->prod_idx == rpc->cons_idx) {
588 spin_unlock_irqrestore(&rpc->e_lock, flags);
589 return NULL;
590 }
591 e_source = &rpc->e_sources[rpc->cons_idx];
592 rpc->cons_idx++;
593 if (rpc->cons_idx == AER_ERROR_SOURCES_MAX)
594 rpc->cons_idx = 0;
595 spin_unlock_irqrestore(&rpc->e_lock, flags);
596
597 return e_source;
598}
599
600static int get_device_error_info(struct pci_dev *dev, struct aer_err_info *info)
601{
602 int pos;
603
604 pos = pci_find_aer_capability(dev);
605
606 /* The device might not support AER */
607 if (!pos)
608 return AER_SUCCESS;
609
610 if (info->severity == AER_CORRECTABLE) {
611 pci_read_config_dword(dev, pos + PCI_ERR_COR_STATUS,
612 &info->status);
613 if (!(info->status & ERR_CORRECTABLE_ERROR_MASK))
614 return AER_UNSUCCESS;
615 } else if (dev->hdr_type & PCI_HEADER_TYPE_BRIDGE ||
616 info->severity == AER_NONFATAL) {
617
618 /* Link is still healthy for IO reads */
619 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS,
620 &info->status);
621 if (!(info->status & ERR_UNCORRECTABLE_ERROR_MASK))
622 return AER_UNSUCCESS;
623
624 if (info->status & AER_LOG_TLP_MASKS) {
625 info->flags |= AER_TLP_HEADER_VALID_FLAG;
626 pci_read_config_dword(dev,
627 pos + PCI_ERR_HEADER_LOG, &info->tlp.dw0);
628 pci_read_config_dword(dev,
629 pos + PCI_ERR_HEADER_LOG + 4, &info->tlp.dw1);
630 pci_read_config_dword(dev,
631 pos + PCI_ERR_HEADER_LOG + 8, &info->tlp.dw2);
632 pci_read_config_dword(dev,
633 pos + PCI_ERR_HEADER_LOG + 12, &info->tlp.dw3);
634 }
635 }
636
637 return AER_SUCCESS;
638}
639
640/**
641 * aer_isr_one_error - consume an error detected by root port
642 * @p_device: pointer to error root port service device
643 * @e_src: pointer to an error source
644 **/
645static void aer_isr_one_error(struct pcie_device *p_device,
646 struct aer_err_source *e_src)
647{
648 struct device *s_device;
649 struct aer_err_info e_info = {0, 0, 0,};
650 int i;
651 u16 id;
652
653 /*
654 * There is a possibility that both correctable error and
655 * uncorrectable error being logged. Report correctable error first.
656 */
657 for (i = 1; i & ROOT_ERR_STATUS_MASKS ; i <<= 2) {
658 if (i > 4)
659 break;
660 if (!(e_src->status & i))
661 continue;
662
663 /* Init comprehensive error information */
664 if (i & PCI_ERR_ROOT_COR_RCV) {
665 id = ERR_COR_ID(e_src->id);
666 e_info.severity = AER_CORRECTABLE;
667 } else {
668 id = ERR_UNCOR_ID(e_src->id);
669 e_info.severity = ((e_src->status >> 6) & 1);
670 }
671 if (e_src->status &
672 (PCI_ERR_ROOT_MULTI_COR_RCV |
673 PCI_ERR_ROOT_MULTI_UNCOR_RCV))
674 e_info.flags |= AER_MULTI_ERROR_VALID_FLAG;
675 if (!(s_device = find_source_device(p_device->port, id))) {
676 printk(KERN_DEBUG "%s->can't find device of ID%04x\n",
677 __FUNCTION__, id);
678 continue;
679 }
680 if (get_device_error_info(to_pci_dev(s_device), &e_info) ==
681 AER_SUCCESS) {
682 aer_print_error(to_pci_dev(s_device), &e_info);
683 handle_error_source(p_device,
684 to_pci_dev(s_device),
685 e_info);
686 }
687 }
688}
689
690/**
691 * aer_isr - consume errors detected by root port
692 * @context: pointer to a private data of pcie device
693 *
694 * Invoked, as DPC, when root port records new detected error
695 **/
696void aer_isr(void *context)
697{
698 struct pcie_device *p_device = (struct pcie_device *) context;
699 struct aer_rpc *rpc = get_service_data(p_device);
700 struct aer_err_source *e_src;
701
702 mutex_lock(&rpc->rpc_mutex);
703 e_src = get_e_source(rpc);
704 while (e_src) {
705 aer_isr_one_error(p_device, e_src);
706 e_src = get_e_source(rpc);
707 }
708 mutex_unlock(&rpc->rpc_mutex);
709
710 wake_up(&rpc->wait_release);
711}
712
713/**
714 * aer_delete_rootport - disable root port aer and delete service data
715 * @rpc: pointer to a root port device being deleted
716 *
717 * Invoked when AER service unloaded on a specific Root Port
718 **/
719void aer_delete_rootport(struct aer_rpc *rpc)
720{
721 /* Disable root port AER itself */
722 disable_root_aer(rpc);
723
724 kfree(rpc);
725}
726
727/**
728 * aer_init - provide AER initialization
729 * @dev: pointer to AER pcie device
730 *
731 * Invoked when AER service driver is loaded.
732 **/
733int aer_init(struct pcie_device *dev)
734{
735 int status;
736
737 /* Run _OSC Method */
738 status = aer_osc_setup(dev->port);
739
740 if(status != OSC_METHOD_RUN_SUCCESS) {
741 printk(KERN_DEBUG "%s: AER service init fails - %s\n",
742 __FUNCTION__,
743 (status == OSC_METHOD_NOT_SUPPORTED) ?
744 "No ACPI _OSC support" : "Run ACPI _OSC fails");
745
746 if (!forceload)
747 return status;
748 }
749
750 return AER_SUCCESS;
751}
752
753EXPORT_SYMBOL_GPL(pci_find_aer_capability);
754EXPORT_SYMBOL_GPL(pci_enable_pcie_error_reporting);
755EXPORT_SYMBOL_GPL(pci_disable_pcie_error_reporting);
756EXPORT_SYMBOL_GPL(pci_cleanup_aer_uncorrect_error_status);
757
diff --git a/drivers/pci/pcie/aer/aerdrv_errprint.c b/drivers/pci/pcie/aer/aerdrv_errprint.c
new file mode 100644
index 000000000000..3933d4f30e8c
--- /dev/null
+++ b/drivers/pci/pcie/aer/aerdrv_errprint.c
@@ -0,0 +1,248 @@
1/*
2 * drivers/pci/pcie/aer/aerdrv_errprint.c
3 *
4 * This file is subject to the terms and conditions of the GNU General Public
5 * License. See the file "COPYING" in the main directory of this archive
6 * for more details.
7 *
8 * Format error messages and print them to console.
9 *
10 * Copyright (C) 2006 Intel Corp.
11 * Tom Long Nguyen (tom.l.nguyen@intel.com)
12 * Zhang Yanmin (yanmin.zhang@intel.com)
13 *
14 */
15
16#include <linux/module.h>
17#include <linux/pci.h>
18#include <linux/kernel.h>
19#include <linux/errno.h>
20#include <linux/pm.h>
21#include <linux/suspend.h>
22
23#include "aerdrv.h"
24
25#define AER_AGENT_RECEIVER 0
26#define AER_AGENT_REQUESTER 1
27#define AER_AGENT_COMPLETER 2
28#define AER_AGENT_TRANSMITTER 3
29
30#define AER_AGENT_REQUESTER_MASK (PCI_ERR_UNC_COMP_TIME| \
31 PCI_ERR_UNC_UNSUP)
32
33#define AER_AGENT_COMPLETER_MASK PCI_ERR_UNC_COMP_ABORT
34
35#define AER_AGENT_TRANSMITTER_MASK(t, e) (e & (PCI_ERR_COR_REP_ROLL| \
36 ((t == AER_CORRECTABLE) ? PCI_ERR_COR_REP_TIMER: 0)))
37
38#define AER_GET_AGENT(t, e) \
39 ((e & AER_AGENT_COMPLETER_MASK) ? AER_AGENT_COMPLETER : \
40 (e & AER_AGENT_REQUESTER_MASK) ? AER_AGENT_REQUESTER : \
41 (AER_AGENT_TRANSMITTER_MASK(t, e)) ? AER_AGENT_TRANSMITTER : \
42 AER_AGENT_RECEIVER)
43
44#define AER_PHYSICAL_LAYER_ERROR_MASK PCI_ERR_COR_RCVR
45#define AER_DATA_LINK_LAYER_ERROR_MASK(t, e) \
46 (PCI_ERR_UNC_DLP| \
47 PCI_ERR_COR_BAD_TLP| \
48 PCI_ERR_COR_BAD_DLLP| \
49 PCI_ERR_COR_REP_ROLL| \
50 ((t == AER_CORRECTABLE) ? \
51 PCI_ERR_COR_REP_TIMER: 0))
52
53#define AER_PHYSICAL_LAYER_ERROR 0
54#define AER_DATA_LINK_LAYER_ERROR 1
55#define AER_TRANSACTION_LAYER_ERROR 2
56
57#define AER_GET_LAYER_ERROR(t, e) \
58 ((e & AER_PHYSICAL_LAYER_ERROR_MASK) ? \
59 AER_PHYSICAL_LAYER_ERROR : \
60 (e & AER_DATA_LINK_LAYER_ERROR_MASK(t, e)) ? \
61 AER_DATA_LINK_LAYER_ERROR : \
62 AER_TRANSACTION_LAYER_ERROR)
63
64/*
65 * AER error strings
66 */
67static char* aer_error_severity_string[] = {
68 "Uncorrected (Non-Fatal)",
69 "Uncorrected (Fatal)",
70 "Corrected"
71};
72
73static char* aer_error_layer[] = {
74 "Physical Layer",
75 "Data Link Layer",
76 "Transaction Layer"
77};
78static char* aer_correctable_error_string[] = {
79 "Receiver Error ", /* Bit Position 0 */
80 NULL,
81 NULL,
82 NULL,
83 NULL,
84 NULL,
85 "Bad TLP ", /* Bit Position 6 */
86 "Bad DLLP ", /* Bit Position 7 */
87 "RELAY_NUM Rollover ", /* Bit Position 8 */
88 NULL,
89 NULL,
90 NULL,
91 "Replay Timer Timeout ", /* Bit Position 12 */
92 "Advisory Non-Fatal ", /* Bit Position 13 */
93 NULL,
94 NULL,
95 NULL,
96 NULL,
97 NULL,
98 NULL,
99 NULL,
100 NULL,
101 NULL,
102 NULL,
103 NULL,
104 NULL,
105 NULL,
106 NULL,
107 NULL,
108 NULL,
109 NULL,
110 NULL,
111};
112
113static char* aer_uncorrectable_error_string[] = {
114 NULL,
115 NULL,
116 NULL,
117 NULL,
118 "Data Link Protocol ", /* Bit Position 4 */
119 NULL,
120 NULL,
121 NULL,
122 NULL,
123 NULL,
124 NULL,
125 NULL,
126 "Poisoned TLP ", /* Bit Position 12 */
127 "Flow Control Protocol ", /* Bit Position 13 */
128 "Completion Timeout ", /* Bit Position 14 */
129 "Completer Abort ", /* Bit Position 15 */
130 "Unexpected Completion ", /* Bit Position 16 */
131 "Receiver Overflow ", /* Bit Position 17 */
132 "Malformed TLP ", /* Bit Position 18 */
133 "ECRC ", /* Bit Position 19 */
134 "Unsupported Request ", /* Bit Position 20 */
135 NULL,
136 NULL,
137 NULL,
138 NULL,
139 NULL,
140 NULL,
141 NULL,
142 NULL,
143 NULL,
144 NULL,
145 NULL,
146};
147
148static char* aer_agent_string[] = {
149 "Receiver ID",
150 "Requester ID",
151 "Completer ID",
152 "Transmitter ID"
153};
154
155static char * aer_get_error_source_name(int severity,
156 unsigned int status,
157 char errmsg_buff[])
158{
159 int i;
160 char * errmsg = NULL;
161
162 for (i = 0; i < 32; i++) {
163 if (!(status & (1 << i)))
164 continue;
165
166 if (severity == AER_CORRECTABLE)
167 errmsg = aer_correctable_error_string[i];
168 else
169 errmsg = aer_uncorrectable_error_string[i];
170
171 if (!errmsg) {
172 sprintf(errmsg_buff, "Unknown Error Bit %2d ", i);
173 errmsg = errmsg_buff;
174 }
175
176 break;
177 }
178
179 return errmsg;
180}
181
182static DEFINE_SPINLOCK(logbuf_lock);
183static char errmsg_buff[100];
184void aer_print_error(struct pci_dev *dev, struct aer_err_info *info)
185{
186 char * errmsg;
187 int err_layer, agent;
188 char * loglevel;
189
190 if (info->severity == AER_CORRECTABLE)
191 loglevel = KERN_WARNING;
192 else
193 loglevel = KERN_ERR;
194
195 printk("%s+------ PCI-Express Device Error ------+\n", loglevel);
196 printk("%sError Severity\t\t: %s\n", loglevel,
197 aer_error_severity_string[info->severity]);
198
199 if ( info->status == 0) {
200 printk("%sPCIE Bus Error type\t: (Unaccessible)\n", loglevel);
201 printk("%sUnaccessible Received\t: %s\n", loglevel,
202 info->flags & AER_MULTI_ERROR_VALID_FLAG ?
203 "Multiple" : "First");
204 printk("%sUnregistered Agent ID\t: %04x\n", loglevel,
205 (dev->bus->number << 8) | dev->devfn);
206 } else {
207 err_layer = AER_GET_LAYER_ERROR(info->severity, info->status);
208 printk("%sPCIE Bus Error type\t: %s\n", loglevel,
209 aer_error_layer[err_layer]);
210
211 spin_lock(&logbuf_lock);
212 errmsg = aer_get_error_source_name(info->severity,
213 info->status,
214 errmsg_buff);
215 printk("%s%s\t: %s\n", loglevel, errmsg,
216 info->flags & AER_MULTI_ERROR_VALID_FLAG ?
217 "Multiple" : "First");
218 spin_unlock(&logbuf_lock);
219
220 agent = AER_GET_AGENT(info->severity, info->status);
221 printk("%s%s\t\t: %04x\n", loglevel,
222 aer_agent_string[agent],
223 (dev->bus->number << 8) | dev->devfn);
224
225 printk("%sVendorID=%04xh, DeviceID=%04xh,"
226 " Bus=%02xh, Device=%02xh, Function=%02xh\n",
227 loglevel,
228 dev->vendor,
229 dev->device,
230 dev->bus->number,
231 PCI_SLOT(dev->devfn),
232 PCI_FUNC(dev->devfn));
233
234 if (info->flags & AER_TLP_HEADER_VALID_FLAG) {
235 unsigned char *tlp = (unsigned char *) &info->tlp;
236 printk("%sTLB Header:\n", loglevel);
237 printk("%s%02x%02x%02x%02x %02x%02x%02x%02x"
238 " %02x%02x%02x%02x %02x%02x%02x%02x\n",
239 loglevel,
240 *(tlp + 3), *(tlp + 2), *(tlp + 1), *tlp,
241 *(tlp + 7), *(tlp + 6), *(tlp + 5), *(tlp + 4),
242 *(tlp + 11), *(tlp + 10), *(tlp + 9),
243 *(tlp + 8), *(tlp + 15), *(tlp + 14),
244 *(tlp + 13), *(tlp + 12));
245 }
246 }
247}
248
generated by cgit v1.2.2 (git 2.25.0) at 2024-12-30 16:20:45 -0500