| Message ID | 1506432285-14979-2-git-send-email-sagar.a.kamble@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Quoting Sagar Arun Kamble (2017-09-26 14:24:39) > diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c > index dbe181b..5dcd8c0 100644 > --- a/drivers/gpu/drm/i915/i915_gem.c > +++ b/drivers/gpu/drm/i915/i915_gem.c > @@ -2022,11 +2022,22 @@ int i915_gem_fault(struct vm_fault *vmf) > intel_runtime_pm_put(i915); > } > > +/** > + * i915_gem_runtime_suspend() - Finish GEM suspend > + * @dev_priv: i915 device private > + * > + * This function suspends GuC, removes userspace mappings for all GEM obejcts > + * currently on userfault list and marks fences if any being used as lost. > + * > + * Return: non-zero code on error > + */ > +/** > + * i915_gem_runtime_resume() - Restore GEM state > + * @dev_priv: i915 device private > + * > + * This function inits swizzling, restores fences and resumes GuC. > + * > + * Return: non-zero code on error > + */ > +/** > + * i915_gem_suspend() - Suspend all GT activity. > + * @dev_priv: i915 device private > + * > + * This function disables RPS, flushes all executing context ensuring > + * GEM/GT/Engines idleness, cancels all work that needs GT access and suspends > + * GuC. In the end currently, it also reset the GEM state and GPU HW. > + * > + * Return: non-zero code on error > + */ > +/** > + * i915_gem_resume() - Resume GT activity. > + * @dev_priv: i915 device private > + * > + * This function restores GTT mappings, restores fences and resets the > + * context images and resumes GuC. > + * > + * Return: non-zero code on error > + */ I can very much read what the functions do. These comments blocks are for telling me when to call them, under what pre/post-conditions and the overall intent of the function. Who are these blocks for? This isn't a library interface we expect to be used, these are hooks. -Chris
On 9/26/2017 7:12 PM, Chris Wilson wrote: > Quoting Sagar Arun Kamble (2017-09-26 14:24:39) >> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c >> index dbe181b..5dcd8c0 100644 >> --- a/drivers/gpu/drm/i915/i915_gem.c >> +++ b/drivers/gpu/drm/i915/i915_gem.c >> @@ -2022,11 +2022,22 @@ int i915_gem_fault(struct vm_fault *vmf) >> intel_runtime_pm_put(i915); >> } >> >> +/** >> + * i915_gem_runtime_suspend() - Finish GEM suspend >> + * @dev_priv: i915 device private >> + * >> + * This function suspends GuC, removes userspace mappings for all GEM obejcts >> + * currently on userfault list and marks fences if any being used as lost. >> + * >> + * Return: non-zero code on error >> + */ >> +/** >> + * i915_gem_runtime_resume() - Restore GEM state >> + * @dev_priv: i915 device private >> + * >> + * This function inits swizzling, restores fences and resumes GuC. >> + * >> + * Return: non-zero code on error >> + */ >> +/** >> + * i915_gem_suspend() - Suspend all GT activity. >> + * @dev_priv: i915 device private >> + * >> + * This function disables RPS, flushes all executing context ensuring >> + * GEM/GT/Engines idleness, cancels all work that needs GT access and suspends >> + * GuC. In the end currently, it also reset the GEM state and GPU HW. >> + * >> + * Return: non-zero code on error >> + */ >> +/** >> + * i915_gem_resume() - Resume GT activity. >> + * @dev_priv: i915 device private >> + * >> + * This function restores GTT mappings, restores fences and resets the >> + * context images and resumes GuC. >> + * >> + * Return: non-zero code on error >> + */ > I can very much read what the functions do. These comments blocks are > for telling me when to call them, under what pre/post-conditions and the > overall intent of the function. > > Who are these blocks for? This isn't a library interface we expect to be > used, these are hooks. > -Chris Since these functions are exported from i915_gem.c and are being used from respective drm functions, to make the semantics available in kernel docs I have added these comments.
Quoting Sagar Arun Kamble (2017-09-26 15:24:15) > > > On 9/26/2017 7:12 PM, Chris Wilson wrote: > > Quoting Sagar Arun Kamble (2017-09-26 14:24:39) > >> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c > >> index dbe181b..5dcd8c0 100644 > >> --- a/drivers/gpu/drm/i915/i915_gem.c > >> +++ b/drivers/gpu/drm/i915/i915_gem.c > >> @@ -2022,11 +2022,22 @@ int i915_gem_fault(struct vm_fault *vmf) > >> intel_runtime_pm_put(i915); > >> } > >> > >> +/** > >> + * i915_gem_runtime_suspend() - Finish GEM suspend > >> + * @dev_priv: i915 device private > >> + * > >> + * This function suspends GuC, removes userspace mappings for all GEM obejcts > >> + * currently on userfault list and marks fences if any being used as lost. > >> + * > >> + * Return: non-zero code on error > >> + */ > >> +/** > >> + * i915_gem_runtime_resume() - Restore GEM state > >> + * @dev_priv: i915 device private > >> + * > >> + * This function inits swizzling, restores fences and resumes GuC. > >> + * > >> + * Return: non-zero code on error > >> + */ > >> +/** > >> + * i915_gem_suspend() - Suspend all GT activity. > >> + * @dev_priv: i915 device private > >> + * > >> + * This function disables RPS, flushes all executing context ensuring > >> + * GEM/GT/Engines idleness, cancels all work that needs GT access and suspends > >> + * GuC. In the end currently, it also reset the GEM state and GPU HW. > >> + * > >> + * Return: non-zero code on error > >> + */ > >> +/** > >> + * i915_gem_resume() - Resume GT activity. > >> + * @dev_priv: i915 device private > >> + * > >> + * This function restores GTT mappings, restores fences and resets the > >> + * context images and resumes GuC. > >> + * > >> + * Return: non-zero code on error > >> + */ > > I can very much read what the functions do. These comments blocks are > > for telling me when to call them, under what pre/post-conditions and the > > overall intent of the function. > > > > Who are these blocks for? This isn't a library interface we expect to be > > used, these are hooks. > > -Chris > Since these functions are exported from i915_gem.c and are being used > from respective drm functions, to make > the semantics available in kernel docs I have added these comments. My point is that this isn't functionality that was exported for somebody to call, this was functionality that was pulled from i915_drv.c for a singular purpose. It is driven by that context, and not for general purpose. So who is the consumer? -Chris
Subject includes the word "fences". That was quite surprising as it doesn't involve either dma, sw or guc fences; or the fence registers. -Chris
On 9/26/2017 8:18 PM, Chris Wilson wrote: > Subject includes the word "fences". That was quite surprising as it > doesn't involve either dma, sw or guc fences; or the fence registers. > -Chris This was related to fence registers whose state is updated in i915_gem_restore_fences. With this patch I had moved the this to i915_gem_resume from i915_restore_state.
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index a3bbf18..8d67b8c 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -1698,8 +1698,6 @@ static int i915_drm_resume(struct drm_device *dev) } mutex_unlock(&dev->struct_mutex); - intel_guc_resume(dev_priv); - intel_modeset_init_hw(dev); spin_lock_irq(&dev_priv->irq_lock); @@ -2504,8 +2502,6 @@ static int intel_runtime_suspend(struct device *kdev) return ret; } - intel_guc_suspend(dev_priv); - intel_runtime_pm_disable_interrupts(dev_priv); ret = 0; @@ -2591,8 +2587,6 @@ static int intel_runtime_resume(struct device *kdev) if (intel_uncore_unclaimed_mmio(dev_priv)) DRM_DEBUG_DRIVER("Unclaimed access during suspend, bios?\n"); - intel_guc_resume(dev_priv); - if (IS_GEN9_LP(dev_priv)) { bxt_disable_dc9(dev_priv); bxt_display_core_init(dev_priv, true); diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index dbe181b..5dcd8c0 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -2022,11 +2022,22 @@ int i915_gem_fault(struct vm_fault *vmf) intel_runtime_pm_put(i915); } +/** + * i915_gem_runtime_suspend() - Finish GEM suspend + * @dev_priv: i915 device private + * + * This function suspends GuC, removes userspace mappings for all GEM obejcts + * currently on userfault list and marks fences if any being used as lost. + * + * Return: non-zero code on error + */ int i915_gem_runtime_suspend(struct drm_i915_private *dev_priv) { struct drm_i915_gem_object *obj, *on; int i; + intel_guc_suspend(dev_priv); + /* * Only called during RPM suspend. All users of the userfault_list * must be holding an RPM wakeref to ensure that this can not @@ -2069,6 +2080,14 @@ int i915_gem_runtime_suspend(struct drm_i915_private *dev_priv) return 0; } +/** + * i915_gem_runtime_resume() - Restore GEM state + * @dev_priv: i915 device private + * + * This function inits swizzling, restores fences and resumes GuC. + * + * Return: non-zero code on error + */ int i915_gem_runtime_resume(struct drm_i915_private *dev_priv) { /* @@ -2078,6 +2097,8 @@ int i915_gem_runtime_resume(struct drm_i915_private *dev_priv) i915_gem_init_swizzling(dev_priv); i915_gem_restore_fences(dev_priv); + intel_guc_resume(dev_priv); + return 0; } @@ -4521,6 +4542,16 @@ void i915_gem_sanitize(struct drm_i915_private *i915) } } +/** + * i915_gem_suspend() - Suspend all GT activity. + * @dev_priv: i915 device private + * + * This function disables RPS, flushes all executing context ensuring + * GEM/GT/Engines idleness, cancels all work that needs GT access and suspends + * GuC. In the end currently, it also reset the GEM state and GPU HW. + * + * Return: non-zero code on error + */ int i915_gem_suspend(struct drm_i915_private *dev_priv) { struct drm_device *dev = &dev_priv->drm; @@ -4553,8 +4584,6 @@ int i915_gem_suspend(struct drm_i915_private *dev_priv) i915_gem_contexts_lost(dev_priv); mutex_unlock(&dev->struct_mutex); - intel_guc_suspend(dev_priv); - cancel_delayed_work_sync(&dev_priv->gpu_error.hangcheck_work); cancel_delayed_work_sync(&dev_priv->gt.retire_work); @@ -4571,6 +4600,8 @@ int i915_gem_suspend(struct drm_i915_private *dev_priv) if (WARN_ON(!intel_engines_are_idle(dev_priv))) i915_gem_set_wedged(dev_priv); /* no hope, discard everything */ + intel_guc_suspend(dev_priv); + /* * Neither the BIOS, ourselves or any other kernel * expects the system to be in execlists mode on startup, @@ -4601,6 +4632,15 @@ int i915_gem_suspend(struct drm_i915_private *dev_priv) return ret; } +/** + * i915_gem_resume() - Resume GT activity. + * @dev_priv: i915 device private + * + * This function restores GTT mappings, restores fences and resets the + * context images and resumes GuC. + * + * Return: non-zero code on error + */ int i915_gem_resume(struct drm_i915_private *dev_priv) { struct drm_device *dev = &dev_priv->drm; @@ -4609,13 +4649,14 @@ int i915_gem_resume(struct drm_i915_private *dev_priv) mutex_lock(&dev->struct_mutex); i915_gem_restore_gtt_mappings(dev_priv); + i915_gem_restore_fences(dev_priv); /* As we didn't flush the kernel context before suspend, we cannot * guarantee that the context image is complete. So let's just reset * it and start again. */ dev_priv->gt.resume(dev_priv); - + intel_guc_resume(dev_priv); mutex_unlock(&dev->struct_mutex); return 0; diff --git a/drivers/gpu/drm/i915/i915_suspend.c b/drivers/gpu/drm/i915/i915_suspend.c index 5c86925a..8f3aa4d 100644 --- a/drivers/gpu/drm/i915/i915_suspend.c +++ b/drivers/gpu/drm/i915/i915_suspend.c @@ -108,8 +108,6 @@ int i915_restore_state(struct drm_i915_private *dev_priv) mutex_lock(&dev_priv->drm.struct_mutex); - i915_gem_restore_fences(dev_priv); - if (IS_GEN4(dev_priv)) pci_write_config_word(pdev, GCDGMBUS, dev_priv->regfile.saveGCDGMBUS);
This patch moves GuC suspend/resume handlers to corresponding GEM handlers and orders them properly in the runtime and system suspend/resume flows. It also adds documentation for GEM suspend/resume handlers. i915_gem_restore_fences is GEM resumption task hence it is moved to i915_gem_resume from i915_restore_state. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Imre Deak <imre.deak@intel.com> Cc: Paulo Zanoni <paulo.r.zanoni@intel.com> Cc: Rodrigo Vivi <rodrigo.vivi@intel.com> Cc: Michal Wajdeczko <michal.wajdeczko@intel.com> Cc: Michał Winiarski <michal.winiarski@intel.com> Signed-off-by: Sagar Arun Kamble <sagar.a.kamble@intel.com> --- drivers/gpu/drm/i915/i915_drv.c | 6 ----- drivers/gpu/drm/i915/i915_gem.c | 47 ++++++++++++++++++++++++++++++++++--- drivers/gpu/drm/i915/i915_suspend.c | 2 -- 3 files changed, 44 insertions(+), 11 deletions(-)