1 // SPDX-License-Identifier: MIT
27  * Survivability Mode is a software based workflow for recovering a system in a failed boot state
28  * Here system recoverability is concerned with recovering the firmware responsible for boot.
30  * Boot Survivability
33  * Boot Survivability is implemented by loading the driver with bare minimum (no drm card) to allow
35  * modified such that it enters survivability mode when pcode initialization is incomplete and boot
41  * In-Field-Repair (IFR) to repair a single card without impacting the other cards in a node.
51  * Survivability mode is indicated by the below admin-only readable sysfs which provides additional
57  *	Provides boot status
69  * (:ref:`xe-device-wedging`) requiring a firmware flash to restore normal operation.
97 	struct xe_survivability *survivability = &xe->survivability;  in populate_survivability_info()
98 	struct xe_survivability_info *info = survivability->info;  in populate_survivability_info()
132 	struct xe_survivability *survivability = &xe->survivability;  in log_survivability_info()
133 	struct xe_survivability_info *info = survivability->info;  in log_survivability_info()
136 	dev_info(&pdev->dev, "Survivability Boot Status : Critical Failure (%d)\n",  in log_survivability_info()
137 		 survivability->boot_status);  in log_survivability_info()
140 			dev_info(&pdev->dev, "%s: 0x%x - 0x%x\n", info[id].name,  in log_survivability_info()
147 	struct xe_survivability *survivability = &xe->survivability;  in check_boot_failure()
149 	return survivability->boot_status == NON_CRITICAL_FAILURE ||  in check_boot_failure()
150 		survivability->boot_status == CRITICAL_FAILURE;  in check_boot_failure()
158 	struct xe_survivability *survivability = &xe->survivability;  in survivability_mode_show()
159 	struct xe_survivability_info *info = survivability->info;  in survivability_mode_show()
163 			       survivability->type ? "Runtime" : "Boot");  in survivability_mode_show()
170 			count += sysfs_emit_at(buff, count, "%s: 0x%x - 0x%x\n", info[index].name,  in survivability_mode_show()
182 	struct pci_dev *pdev = to_pci_dev(xe->drm.dev);  in xe_survivability_mode_fini()
183 	struct device *dev = &pdev->dev;  in xe_survivability_mode_fini()
185 	sysfs_remove_file(&dev->kobj, &dev_attr_survivability_mode.attr);  in xe_survivability_mode_fini()
190 	struct device *dev = &pdev->dev;  in create_survivability_sysfs()
195 	ret = sysfs_create_file(&dev->kobj, &dev_attr_survivability_mode.attr);  in create_survivability_sysfs()
201 	ret = devm_add_action_or_reset(xe->drm.dev,  in create_survivability_sysfs()
211 	struct device *dev = &pdev->dev;  in enable_boot_survivability_mode()
213 	struct xe_survivability *survivability = &xe->survivability;  in enable_boot_survivability_mode()
221 	survivability->mode = true;  in enable_boot_survivability_mode()
238 	survivability->mode = false;  in enable_boot_survivability_mode()
244 	struct xe_survivability *survivability = &xe->survivability;  in init_survivability_mode()
247 	survivability->size = MAX_SCRATCH_MMIO;  in init_survivability_mode()
249 	info = devm_kcalloc(xe->drm.dev, survivability->size, sizeof(*info),  in init_survivability_mode()
252 		return -ENOMEM;  in init_survivability_mode()
254 	survivability->info = info;  in init_survivability_mode()
262  * xe_survivability_mode_is_boot_enabled- check if boot survivability mode is enabled
265  * Returns true if in boot survivability mode of type, else false
269 	struct xe_survivability *survivability = &xe->survivability;  in xe_survivability_mode_is_boot_enabled()
271 	return survivability->mode && survivability->type == XE_SURVIVABILITY_TYPE_BOOT;  in xe_survivability_mode_is_boot_enabled()
275  * xe_survivability_mode_is_requested - check if it's possible to enable survivability
279  * This function reads configfs and  boot status from Pcode.
281  * Return: true if platform support is available and boot status indicates
286 	struct xe_survivability *survivability = &xe->survivability;  in xe_survivability_mode_is_requested()
288 	struct pci_dev *pdev = to_pci_dev(xe->drm.dev);  in xe_survivability_mode_is_requested()
292 	if (!IS_DGFX(xe) || IS_SRIOV_VF(xe) || xe->info.platform < XE_BATTLEMAGE)  in xe_survivability_mode_is_requested()
301 	survivability->boot_status = REG_FIELD_GET(BOOT_STATUS, data);  in xe_survivability_mode_is_requested()
307  * xe_survivability_mode_runtime_enable - Initialize and enable runtime survivability mode
311  * Runtime survivability mode is enabled when certain errors cause the device to be
312  * in non-recoverable state. The device is declared wedged with the appropriate
315  * Return: 0 if runtime survivability mode is enabled, negative error code otherwise.
319 	struct xe_survivability *survivability = &xe->survivability;  in xe_survivability_mode_runtime_enable()
320 	struct pci_dev *pdev = to_pci_dev(xe->drm.dev);  in xe_survivability_mode_runtime_enable()
323 	if (!IS_DGFX(xe) || IS_SRIOV_VF(xe) || xe->info.platform < XE_BATTLEMAGE) {  in xe_survivability_mode_runtime_enable()
324 		dev_err(&pdev->dev, "Runtime Survivability Mode not supported\n");  in xe_survivability_mode_runtime_enable()
325 		return -EINVAL;  in xe_survivability_mode_runtime_enable()
334 		dev_err(&pdev->dev, "Failed to create survivability mode sysfs\n");  in xe_survivability_mode_runtime_enable()
336 	survivability->type = XE_SURVIVABILITY_TYPE_RUNTIME;  in xe_survivability_mode_runtime_enable()
337 	dev_err(&pdev->dev, "Runtime Survivability mode enabled\n");  in xe_survivability_mode_runtime_enable()
341 …dev_err(&pdev->dev, "Firmware flash required, Please refer to the userspace documentation for more…  in xe_survivability_mode_runtime_enable()
347  * xe_survivability_mode_boot_enable - Initialize and enable boot survivability mode
350  * Initialize survivability information and enable boot survivability mode
352  * Return: 0 if boot survivability mode is enabled or not requested, negative error
357 	struct xe_survivability *survivability = &xe->survivability;  in xe_survivability_mode_boot_enable()
358 	struct pci_dev *pdev = to_pci_dev(xe->drm.dev);  in xe_survivability_mode_boot_enable()
368 	/* Log breadcrumbs but do not enter survivability mode for Critical boot errors */  in xe_survivability_mode_boot_enable()
369 	if (survivability->boot_status == CRITICAL_FAILURE) {  in xe_survivability_mode_boot_enable()
371 		return -ENXIO;  in xe_survivability_mode_boot_enable()
374 	survivability->type = XE_SURVIVABILITY_TYPE_BOOT;  in xe_survivability_mode_boot_enable()