| Message ID | 1343165630-21604-3-git-send-email-daniel.vetter@ffwll.ch (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Tue, 24 Jul 2012 23:33:43 +0200 Daniel Vetter <daniel.vetter@ffwll.ch> wrote: > The update_gfx_val function called from mark_busy wasn't taking the > mchdev_lock, as it should have. Also sprinkle a few spinlock asserts > over the code to document things better. > > Things are still rather confusing, especially since a few variables > in dev_priv are used by both the gen6+ rps code and the ilk ips code. > But protected by totally different locks. Follow-on patches will clean > that up. > > v2: Don't add a deadlock ... hence split up update_gfx_val into a > wrapper that grabs the lock and an internal __ variant for callsites > within intel_pm.c that already have taken the lock. > > Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch> > --- > drivers/gpu/drm/i915/intel_pm.c | 50 ++++++++++++++++++++++++++------------- > 1 file changed, 34 insertions(+), 16 deletions(-) > > diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c > index b6e8fbf..21a0088 100644 > --- a/drivers/gpu/drm/i915/intel_pm.c > +++ b/drivers/gpu/drm/i915/intel_pm.c > @@ -2693,6 +2693,21 @@ static const struct cparams { > { 0, 800, 231, 23784 }, > }; > > +/** > + * Lock protecting IPS related data structures > + * - i915_mch_dev > + * - dev_priv->max_delay > + * - dev_priv->min_delay > + * - dev_priv->fmax > + * - dev_priv->gpu_busy > + * - dev_priv->gfx_power > + */ > +static DEFINE_SPINLOCK(mchdev_lock); > + > +/* Global for IPS driver to get at the current i915 device. Protected by > + * mchdev_lock. */ > +static struct drm_i915_private *i915_mch_dev; > + > unsigned long i915_chipset_val(struct drm_i915_private *dev_priv) > { > u64 total_count, diff, ret; > @@ -2700,6 +2715,8 @@ unsigned long i915_chipset_val(struct drm_i915_private *dev_priv) > unsigned long now = jiffies_to_msecs(jiffies), diff1; > int i; > > + assert_spin_locked(&mchdev_lock); > + > diff1 = now - dev_priv->last_time1; > > /* Prevent division-by-zero if we are asking too fast. > @@ -2901,15 +2918,14 @@ static u16 pvid_to_extvid(struct drm_i915_private *dev_priv, u8 pxvid) > return v_table[pxvid].vd; > } > > -void i915_update_gfx_val(struct drm_i915_private *dev_priv) > +void __i915_update_gfx_val(struct drm_i915_private *dev_priv) If I am understanding what you're trying to do, this should be static. > { > struct timespec now, diff1; > u64 diff; > unsigned long diffms; > u32 count; > > - if (dev_priv->info->gen != 5) > - return; > + assert_spin_locked(&mchdev_lock); And if it's static, I think the assertion is unnecessary since all callers are easily checked within the file. NOTE: you've killed a bunch of my assertions previously for similar reasons, so I'm just letting off steam. > > getrawmonotonic(&now); > diff1 = timespec_sub(now, dev_priv->last_time2); > @@ -2937,11 +2953,25 @@ void i915_update_gfx_val(struct drm_i915_private *dev_priv) > dev_priv->gfx_power = diff; > } > > +void i915_update_gfx_val(struct drm_i915_private *dev_priv) > +{ > + if (dev_priv->info->gen != 5) > + return; > + > + spin_lock(&mchdev_lock); > + > + __i915_update_gfx_val(dev_priv); > + > + spin_unlock(&mchdev_lock); > +} > + > unsigned long i915_gfx_val(struct drm_i915_private *dev_priv) > { > unsigned long t, corr, state1, corr2, state2; > u32 pxvid, ext_v; > > + assert_spin_locked(&mchdev_lock); > + > pxvid = I915_READ(PXVFREQ_BASE + (dev_priv->cur_delay * 4)); > pxvid = (pxvid >> 24) & 0x7f; > ext_v = pvid_to_extvid(dev_priv, pxvid); > @@ -2967,23 +2997,11 @@ unsigned long i915_gfx_val(struct drm_i915_private *dev_priv) > state2 = (corr2 * state1) / 10000; > state2 /= 100; /* convert to mW */ > > - i915_update_gfx_val(dev_priv); > + __i915_update_gfx_val(dev_priv); > > return dev_priv->gfx_power + state2; > } > > -/* Global for IPS driver to get at the current i915 device */ > -static struct drm_i915_private *i915_mch_dev; > -/* > - * Lock protecting IPS related data structures > - * - i915_mch_dev > - * - dev_priv->max_delay > - * - dev_priv->min_delay > - * - dev_priv->fmax > - * - dev_priv->gpu_busy > - */ > -static DEFINE_SPINLOCK(mchdev_lock); > - > /** > * i915_read_mch_val - return value for IPS use > *
diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index b6e8fbf..21a0088 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -2693,6 +2693,21 @@ static const struct cparams { { 0, 800, 231, 23784 }, }; +/** + * Lock protecting IPS related data structures + * - i915_mch_dev + * - dev_priv->max_delay + * - dev_priv->min_delay + * - dev_priv->fmax + * - dev_priv->gpu_busy + * - dev_priv->gfx_power + */ +static DEFINE_SPINLOCK(mchdev_lock); + +/* Global for IPS driver to get at the current i915 device. Protected by + * mchdev_lock. */ +static struct drm_i915_private *i915_mch_dev; + unsigned long i915_chipset_val(struct drm_i915_private *dev_priv) { u64 total_count, diff, ret; @@ -2700,6 +2715,8 @@ unsigned long i915_chipset_val(struct drm_i915_private *dev_priv) unsigned long now = jiffies_to_msecs(jiffies), diff1; int i; + assert_spin_locked(&mchdev_lock); + diff1 = now - dev_priv->last_time1; /* Prevent division-by-zero if we are asking too fast. @@ -2901,15 +2918,14 @@ static u16 pvid_to_extvid(struct drm_i915_private *dev_priv, u8 pxvid) return v_table[pxvid].vd; } -void i915_update_gfx_val(struct drm_i915_private *dev_priv) +void __i915_update_gfx_val(struct drm_i915_private *dev_priv) { struct timespec now, diff1; u64 diff; unsigned long diffms; u32 count; - if (dev_priv->info->gen != 5) - return; + assert_spin_locked(&mchdev_lock); getrawmonotonic(&now); diff1 = timespec_sub(now, dev_priv->last_time2); @@ -2937,11 +2953,25 @@ void i915_update_gfx_val(struct drm_i915_private *dev_priv) dev_priv->gfx_power = diff; } +void i915_update_gfx_val(struct drm_i915_private *dev_priv) +{ + if (dev_priv->info->gen != 5) + return; + + spin_lock(&mchdev_lock); + + __i915_update_gfx_val(dev_priv); + + spin_unlock(&mchdev_lock); +} + unsigned long i915_gfx_val(struct drm_i915_private *dev_priv) { unsigned long t, corr, state1, corr2, state2; u32 pxvid, ext_v; + assert_spin_locked(&mchdev_lock); + pxvid = I915_READ(PXVFREQ_BASE + (dev_priv->cur_delay * 4)); pxvid = (pxvid >> 24) & 0x7f; ext_v = pvid_to_extvid(dev_priv, pxvid); @@ -2967,23 +2997,11 @@ unsigned long i915_gfx_val(struct drm_i915_private *dev_priv) state2 = (corr2 * state1) / 10000; state2 /= 100; /* convert to mW */ - i915_update_gfx_val(dev_priv); + __i915_update_gfx_val(dev_priv); return dev_priv->gfx_power + state2; } -/* Global for IPS driver to get at the current i915 device */ -static struct drm_i915_private *i915_mch_dev; -/* - * Lock protecting IPS related data structures - * - i915_mch_dev - * - dev_priv->max_delay - * - dev_priv->min_delay - * - dev_priv->fmax - * - dev_priv->gpu_busy - */ -static DEFINE_SPINLOCK(mchdev_lock); - /** * i915_read_mch_val - return value for IPS use *
The update_gfx_val function called from mark_busy wasn't taking the mchdev_lock, as it should have. Also sprinkle a few spinlock asserts over the code to document things better. Things are still rather confusing, especially since a few variables in dev_priv are used by both the gen6+ rps code and the ilk ips code. But protected by totally different locks. Follow-on patches will clean that up. v2: Don't add a deadlock ... hence split up update_gfx_val into a wrapper that grabs the lock and an internal __ variant for callsites within intel_pm.c that already have taken the lock. Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch> --- drivers/gpu/drm/i915/intel_pm.c | 50 ++++++++++++++++++++++++++------------- 1 file changed, 34 insertions(+), 16 deletions(-)