/** * DOC: Xe Power Management * * Xe PM implements the main routines for both system level suspend states and * for the opportunistic runtime suspend states. * * System Level Suspend (S-States) - In general this is OS initiated suspend * driven by ACPI for achieving S0ix (a.k.a. S2idle, freeze), S3 (suspend to ram), * S4 (disk). The main functions here are `xe_pm_suspend` and `xe_pm_resume`. They * are the main point for the suspend to and resume from these states. * * PCI Device Suspend (D-States) - This is the opportunistic PCIe device low power * state D3, controlled by the PCI subsystem and ACPI with the help from the * runtime_pm infrastructure. * PCI D3 is special and can mean D3hot, where Vcc power is on for keeping memory * alive and quicker low latency resume or D3Cold where Vcc power is off for * better power savings. * The Vcc control of PCI hierarchy can only be controlled at the PCI root port * level, while the device driver can be behind multiple bridges/switches and * paired with other devices. For this reason, the PCI subsystem cannot perform * the transition towards D3Cold. The lowest runtime PM possible from the PCI * subsystem is D3hot. Then, if all these paired devices in the same root port * are in D3hot, ACPI will assist here and run its own methods (_PR3 and _OFF) * to perform the transition from D3hot to D3cold. Xe may disallow this * transition by calling pci_d3cold_disable(root_pdev) before going to runtime * suspend. It will be based on runtime conditions such as VRAM usage for a * quick and low latency resume for instance. * * Runtime PM - This infrastructure provided by the Linux kernel allows the * device drivers to indicate when the can be runtime suspended, so the device * could be put at D3 (if supported), or allow deeper package sleep states * (PC-states), and/or other low level power states. Xe PM component provides * `xe_pm_runtime_suspend` and `xe_pm_runtime_resume` functions that PCI * subsystem will call before transition to/from runtime suspend. * * Also, Xe PM provides get and put functions that Xe driver will use to * indicate activity. In order to avoid locking complications with the memory * management, whenever possible, these get and put functions needs to be called * from the higher/outer levels. * The main cases that need to be protected from the outer levels are: IOCTL, * sysfs, debugfs, dma-buf sharing, GPU execution. * * This component is not responsible for GT idleness (RC6) nor GT frequency * management (RPS).
*/
/** * xe_rpm_reclaim_safe() - Whether runtime resume can be done from reclaim context * @xe: The xe device. * * Return: true if it is safe to runtime resume from reclaim context. * false otherwise.
*/ bool xe_rpm_reclaim_safe(conststruct xe_device *xe)
{ return !xe->d3cold.capable;
}
/* * Disable the system suspend direct complete optimization. * We need to ensure that the regular device suspend/resume functions * are called since our runtime_pm cannot guarantee local memory * eviction for d3cold. * TODO: Check HDA audio dependencies claimed by i915, and then enforce * this option to integrated graphics as well.
*/ if (IS_DGFX(xe))
dev_pm_set_driver_flags(dev, DPM_FLAG_NO_DIRECT_COMPLETE);
switch (action) { case PM_HIBERNATION_PREPARE: case PM_SUSPEND_PREPARE:
reinit_completion(&xe->pm_block);
xe_pm_runtime_get(xe);
err = xe_bo_evict_all_user(xe); if (err)
drm_dbg(&xe->drm, "Notifier evict user failed (%d)\n", err);
err = xe_bo_notifier_prepare_all_pinned(xe); if (err)
drm_dbg(&xe->drm, "Notifier prepare pin failed (%d)\n", err); /* * Keep the runtime pm reference until post hibernation / post suspend to * avoid a runtime suspend interfering with evicted objects or backup * allocations.
*/ break; case PM_POST_HIBERNATION: case PM_POST_SUSPEND:
complete_all(&xe->pm_block);
xe_pm_wake_rebind_workers(xe);
xe_bo_notifier_unprepare_all_pinned(xe);
xe_pm_runtime_put(xe); break;
}
return NOTIFY_DONE;
}
/** * xe_pm_init - Initialize Xe Power Management * @xe: xe device instance * * This component is responsible for System and Device sleep states. * * Returns 0 for success, negative error code otherwise.
*/ int xe_pm_init(struct xe_device *xe)
{
u32 vram_threshold; int err;
xe->pm_notifier.notifier_call = xe_pm_notifier_callback;
err = register_pm_notifier(&xe->pm_notifier); if (err) return err;
err = drmm_mutex_init(&xe->drm, &xe->rebind_resume_lock); if (err) goto err_unregister;
/* * Just in case it's somehow possible for our writes to be reordered to * the extent that something else re-uses the task written in * pm_callback_task. For example after returning from the callback, but * before the reordered write that resets pm_callback_task back to NULL.
*/
smp_mb(); /* pairs with xe_pm_read_callback_task */
}
/** * xe_pm_runtime_suspended - Check if runtime_pm state is suspended * @xe: xe device instance * * This does not provide any guarantee that the device is going to remain * suspended as it might be racing with the runtime state transitions. * It can be used only as a non-reliable assertion, to ensure that we are not in * the sleep state while trying to access some memory for instance. * * Returns true if PCI device is suspended, false otherwise.
*/ bool xe_pm_runtime_suspended(struct xe_device *xe)
{ return pm_runtime_suspended(xe->drm.dev);
}
/* * The actual xe_pm_runtime_put() is always async underneath, so * exactly where that is called should makes no difference to us. However * we still need to be very careful with the locks that this callback * acquires and the locks that are acquired and held by any callers of * xe_runtime_pm_get(). We already have the matching annotation * on that side, but we also need it here. For example lockdep should be * able to tell us if the following scenario is in theory possible: * * CPU0 | CPU1 (kworker) * lock(A) | * | xe_pm_runtime_suspend() * | lock(A) * xe_pm_runtime_get() | * * This will clearly deadlock since rpm core needs to wait for * xe_pm_runtime_suspend() to complete, but here we are holding lock(A) * on CPU0 which prevents CPU1 making forward progress. With the * annotation here and in xe_pm_runtime_get() lockdep will see * the potential lock inversion and give us a nice splat.
*/
xe_rpm_lockmap_acquire(xe);
err = xe_pxp_pm_suspend(xe->pxp); if (err) goto out;
/* * Applying lock for entire list op as xe_ttm_bo_destroy and xe_bo_move_notify * also checks and deletes bo entry from user fault list.
*/
mutex_lock(&xe->mem_access.vram_userfault.lock);
list_for_each_entry_safe(bo, on,
&xe->mem_access.vram_userfault.list, vram_userfault_link)
xe_bo_runtime_pm_release_mmap_offset(bo);
mutex_unlock(&xe->mem_access.vram_userfault.lock);
xe_display_pm_runtime_suspend(xe);
if (xe->d3cold.allowed) {
err = xe_bo_evict_all(xe); if (err) goto out_resume;
}
/* * For places where resume is synchronous it can be quite easy to deadlock * if we are not careful. Also in practice it might be quite timing * sensitive to ever see the 0 -> 1 transition with the callers locks * held, so deadlocks might exist but are hard for lockdep to ever see. * With this in mind, help lockdep learn about the potentially scary * stuff that can happen inside the runtime_resume callback by acquiring * a dummy lock (it doesn't protect anything and gets compiled out on * non-debug builds). Lockdep then only needs to see the * xe_pm_runtime_xxx_map -> runtime_resume callback once, and then can * hopefully validate all the (callers_locks) -> xe_pm_runtime_xxx_map. * For example if the (callers_locks) are ever grabbed in the * runtime_resume callback, lockdep should give us a nice splat.
*/ staticvoid xe_rpm_might_enter_cb(conststruct xe_device *xe)
{
xe_rpm_lockmap_acquire(xe);
xe_rpm_lockmap_release(xe);
}
/* * Prime the lockdep maps for known locking orders that need to * be supported but that may not always occur on all systems.
*/ staticvoid xe_pm_runtime_lockdep_prime(void)
{ struct dma_resv lockdep_resv;
dma_resv_init(&lockdep_resv);
lock_map_acquire(&xe_pm_runtime_d3cold_map); /* D3Cold takes the dma_resv locks to evict bos */
dma_resv_lock(&lockdep_resv, NULL);
dma_resv_unlock(&lockdep_resv);
lock_map_release(&xe_pm_runtime_d3cold_map);
/* Shrinkers might like to wake up the device under reclaim. */
fs_reclaim_acquire(GFP_KERNEL);
lock_map_acquire(&xe_pm_runtime_nod3cold_map);
lock_map_release(&xe_pm_runtime_nod3cold_map);
fs_reclaim_release(GFP_KERNEL);
}
/** * xe_pm_runtime_get - Get a runtime_pm reference and resume synchronously * @xe: xe device instance
*/ void xe_pm_runtime_get(struct xe_device *xe)
{
trace_xe_pm_runtime_get(xe, __builtin_return_address(0));
pm_runtime_get_noresume(xe->drm.dev);
if (xe_pm_read_callback_task(xe) == current) return;
/** * xe_pm_runtime_put - Put the runtime_pm reference back and mark as idle * @xe: xe device instance
*/ void xe_pm_runtime_put(struct xe_device *xe)
{
trace_xe_pm_runtime_put(xe, __builtin_return_address(0)); if (xe_pm_read_callback_task(xe) == current) {
pm_runtime_put_noidle(xe->drm.dev);
} else {
pm_runtime_mark_last_busy(xe->drm.dev);
pm_runtime_put(xe->drm.dev);
}
}
/** * xe_pm_runtime_get_ioctl - Get a runtime_pm reference before ioctl * @xe: xe device instance * * Returns: Any number greater than or equal to 0 for success, negative error * code otherwise.
*/ int xe_pm_runtime_get_ioctl(struct xe_device *xe)
{
trace_xe_pm_runtime_get_ioctl(xe, __builtin_return_address(0)); if (WARN_ON(xe_pm_read_callback_task(xe) == current)) return -ELOOP;
/** * xe_pm_runtime_get_if_active - Get a runtime_pm reference if device active * @xe: xe device instance * * Return: True if device is awake (regardless the previous number of references) * and a new reference was taken, false otherwise.
*/ bool xe_pm_runtime_get_if_active(struct xe_device *xe)
{ return pm_runtime_get_if_active(xe->drm.dev) > 0;
}
/** * xe_pm_runtime_get_if_in_use - Get a new reference if device is active with previous ref taken * @xe: xe device instance * * Return: True if device is awake, a previous reference had been already taken, * and a new reference was now taken, false otherwise.
*/ bool xe_pm_runtime_get_if_in_use(struct xe_device *xe)
{ if (xe_pm_read_callback_task(xe) == current) { /* The device is awake, grab the ref and move on */
pm_runtime_get_noresume(xe->drm.dev); returntrue;
}
/* * Very unreliable! Should only be used to suppress the false positive case * in the missing outer rpm protection warning.
*/ staticbool xe_pm_suspending_or_resuming(struct xe_device *xe)
{ #ifdef CONFIG_PM struct device *dev = xe->drm.dev;
/** * xe_pm_runtime_get_noresume - Bump runtime PM usage counter without resuming * @xe: xe device instance * * This function should be used in inner places where it is surely already * protected by outer-bound callers of `xe_pm_runtime_get`. * It will warn if not protected. * The reference should be put back after this function regardless, since it * will always bump the usage counter, regardless.
*/ void xe_pm_runtime_get_noresume(struct xe_device *xe)
{ bool ref;
/** * xe_pm_runtime_resume_and_get - Resume, then get a runtime_pm ref if awake. * @xe: xe device instance * * Returns: True if device is awake and the reference was taken, false otherwise.
*/ bool xe_pm_runtime_resume_and_get(struct xe_device *xe)
{ if (xe_pm_read_callback_task(xe) == current) { /* The device is awake, grab the ref and move on */
pm_runtime_get_noresume(xe->drm.dev); returntrue;
}
if (!bridge->driver) {
drm_warn(&xe->drm, "unbounded parent pci bridge, device won't support any PM support.\n");
device_set_pm_not_required(&pdev->dev);
}
}
/** * xe_pm_set_vram_threshold - Set a VRAM threshold for allowing/blocking D3Cold * @xe: xe device instance * @threshold: VRAM size in MiB for the D3cold threshold * * Return: * * 0 - success * * -EINVAL - invalid argument
*/ int xe_pm_set_vram_threshold(struct xe_device *xe, u32 threshold)
{ struct ttm_resource_manager *man;
u32 vram_total_mb = 0; int i;
for (i = XE_PL_VRAM0; i <= XE_PL_VRAM1; ++i) {
man = ttm_manager_type(&xe->ttm, i); if (man)
vram_total_mb += DIV_ROUND_UP_ULL(man->size, 1024 * 1024);
}
Die Informationen auf dieser Webseite wurden
nach bestem Wissen sorgfältig zusammengestellt. Es wird jedoch weder Vollständigkeit, noch Richtigkeit,
noch Qualität der bereit gestellten Informationen zugesichert.
Bemerkung:
Die farbliche Syntaxdarstellung und die Messung sind noch experimentell.