According to BSpec 53998, we should try to
restrict qgv points, which can't provide
enough bandwidth for desired display configuration.
Currently we are just comparing against all of
those and take minimum(worst case).
v2: Fixed wrong PCode reply mask, removed hardcoded
values.
v3: Forbid simultaneous legacy SAGV PCode requests and
restricting qgv points. Put the actual restriction
to commit function, added serialization(thanks to Ville)
to prevent commit being applied out of order in case of
nonblocking and/or nomodeset commits.
v4:
- Minor code refactoring, fixed few typos(thanks to James Ausmus)
- Change the naming of qgv point
masking/unmasking functions(James Ausmus).
- Simplify the masking/unmasking operation itself,
as we don't need to mask only single point per request(James Ausmus)
- Reject and stick to highest bandwidth point if SAGV
can't be enabled(BSpec)
v5:
- Add new mailbox reply codes, which seems to happen during boot
time for TGL and indicate that QGV setting is not yet available.
v6:
- Increase number of supported QGV points to be in sync with BSpec.
v7: - Rebased and resolved conflict to fix build failure.
- Fix NUM_QGV_POINTS to 8 and moved that to header file(James Ausmus)
v8: - Don't report an error if we can't restrict qgv points, as SAGV
can be disabled by BIOS, which is completely legal. So don't
make CI panic. Instead if we detect that there is only 1 QGV
point accessible just analyze if we can fit the required bandwidth
requirements, but no need in restricting.
v9: - Fix wrong QGV transition if we have 0 planes and no SAGV
simultaneously.
v10: - Fix CDCLK corruption, because of global state getting serialized
without modeset, which caused copying of non-calculated cdclk
to be copied to dev_priv(thanks to Ville for the hint).
v11: - Remove unneeded headers and spaces(Matthew Roper)
- Remove unneeded intel_qgv_info qi struct from bw check and zero
out the needed one(Matthew Roper)
- Changed QGV error message to have more clear meaning(Matthew Roper)
- Use state->modeset_set instead of any_ms(Matthew Roper)
- Moved NUM_SAGV_POINTS from i915_reg.h to i915_drv.h where it's used
- Keep using crtc_state->hw.active instead of .enable(Matthew Roper)
- Moved unrelated changes to other patch(using latency as parameter
for plane wm calculation, moved to SAGV refactoring patch)
v12: - Fix rebase conflict with own temporary SAGV/QGV fix.
- Remove unnecessary mask being zero check when unmasking
qgv points as this is completely legal(Matt Roper)
- Check if we are setting the same mask as already being set
in hardware to prevent error from PCode.
- Fix error message when restricting/unrestricting qgv points
to "mask/unmask" which sounds more accurate(Matt Roper)
- Move sagv status setting to icl_get_bw_info from atomic check
as this should be calculated only once.(Matt Roper)
- Edited comments for the case when we can't enable SAGV and
use only 1 QGV point with highest bandwidth to be more
understandable.(Matt Roper)
v13: - Moved max_data_rate in bw check to closer scope(Ville Syrjälä)
- Changed comment for zero new_mask in qgv points masking function
to better reflect reality(Ville Syrjälä)
- Simplified bit mask operation in qgv points masking function
(Ville Syrjälä)
- Moved intel_qgv_points_mask closer to gen11 SAGV disabling,
however this still can't be under modeset condition(Ville Syrjälä)
- Packed qgv_points_mask as u8 and moved closer to pipe_sagv_mask
(Ville Syrjälä)
- Extracted PCode changes to separate patch.(Ville Syrjälä)
- Now treat num_planes 0 same as 1 to avoid confusion and
returning max_bw as 0, which would prevent choosing QGV
point having max bandwidth in case if SAGV is not allowed,
as per BSpec(Ville Syrjälä)
- Do the actual qgv_points_mask swap in the same place as
all other global state parts like cdclk are swapped.
In the next patch, this all will be moved to bw state as
global state, once new global state patch series from Ville
lands
v14: - Now using global state to serialize access to qgv points
- Added global state locking back, otherwise we seem to read
bw state in a wrong way.
v15: - Added TODO comment for near atomic global state locking in
bw code.
Signed-off-by: Stanislav Lisovskiy <stanislav.lisovskiy@xxxxxxxxx>
Cc: Ville Syrjälä <ville.syrjala@xxxxxxxxx>
Cc: James Ausmus <james.ausmus@xxxxxxxxx>
---
drivers/gpu/drm/i915/display/intel_bw.c | 177 ++++++++++++++-----
drivers/gpu/drm/i915/display/intel_bw.h | 9 +
drivers/gpu/drm/i915/display/intel_display.c | 109 ++++++++++++
drivers/gpu/drm/i915/i915_drv.h | 3 +
4 files changed, 249 insertions(+), 49 deletions(-)
diff --git a/drivers/gpu/drm/i915/display/intel_bw.c b/drivers/gpu/drm/i915/display/intel_bw.c
index ff57277e8880..6b7894ef7ea6 100644
--- a/drivers/gpu/drm/i915/display/intel_bw.c
+++ b/drivers/gpu/drm/i915/display/intel_bw.c
@@ -8,6 +8,9 @@
#include "intel_bw.h"
#include "intel_display_types.h"
#include "intel_sideband.h"
+#include "intel_atomic.h"
+#include "intel_pm.h"
+
/* Parameters for Qclk Geyserville (QGV) */
struct intel_qgv_point {
@@ -113,6 +116,26 @@ static int icl_pcode_read_qgv_point_info(struct drm_i915_private *dev_priv,
return 0;
}
+int icl_pcode_restrict_qgv_points(struct drm_i915_private *dev_priv,
+ u32 points_mask)
+{
+ int ret;
+
+ /* bspec says to keep retrying for at least 1 ms */
+ ret = skl_pcode_request(dev_priv, ICL_PCODE_SAGV_DE_MEM_SS_CONFIG,
+ points_mask,
+ GEN11_PCODE_POINTS_RESTRICTED_MASK,
+ GEN11_PCODE_POINTS_RESTRICTED,
+ 1);
+
+ if (ret < 0) {
+ DRM_ERROR("Failed to disable qgv points (%d)\n", ret);
+ return ret;
+ }
+
+ return 0;
+}
+
static int icl_get_qgv_points(struct drm_i915_private *dev_priv,
struct intel_qgv_info *qi)
{
@@ -240,6 +263,16 @@ static int icl_get_bw_info(struct drm_i915_private *dev_priv, const struct intel
break;
}
+ /*
+ * In case if SAGV is disabled in BIOS, we always get 1
+ * SAGV point, but we can't send PCode commands to restrict it
+ * as it will fail and pointless anyway.
+ */
+ if (qi.num_points == 1)
+ dev_priv->sagv_status = I915_SAGV_NOT_CONTROLLED;
+ else
+ dev_priv->sagv_status = I915_SAGV_ENABLED;
+
return 0;
}
@@ -259,7 +292,7 @@ static unsigned int icl_max_bw(struct drm_i915_private *dev_priv,
if (qgv_point >= bi->num_qgv_points)
return UINT_MAX;
- if (num_planes >= bi->num_planes)
+ if (num_planes >= bi->num_planes || !num_planes)
return bi->deratedbw[qgv_point];
}
@@ -277,34 +310,6 @@ void intel_bw_init_hw(struct drm_i915_private *dev_priv)
icl_get_bw_info(dev_priv, &icl_sa_info);
}
-static unsigned int intel_max_data_rate(struct drm_i915_private *dev_priv,
- int num_planes)
-{
- if (INTEL_GEN(dev_priv) >= 11) {
- /*
- * Any bw group has same amount of QGV points
- */
- const struct intel_bw_info *bi =
- &dev_priv->max_bw[0];
- unsigned int min_bw = UINT_MAX;
- int i;
-
- /*
- * FIXME with SAGV disabled maybe we can assume
- * point 1 will always be used? Seems to match
- * the behaviour observed in the wild.
- */
- for (i = 0; i < bi->num_qgv_points; i++) {
- unsigned int bw = icl_max_bw(dev_priv, num_planes, i);
-
- min_bw = min(bw, min_bw);
- }
- return min_bw;
- } else {
- return UINT_MAX;
- }
-}
-
static unsigned int intel_bw_crtc_num_active_planes(const struct intel_crtc_state *crtc_state)
{
/*
@@ -417,11 +422,16 @@ int intel_bw_atomic_check(struct intel_atomic_state *state)
{
struct drm_i915_private *dev_priv = to_i915(state->base.dev);
struct intel_crtc_state *new_crtc_state, *old_crtc_state;
- struct intel_bw_state *bw_state = NULL;
- unsigned int data_rate, max_data_rate;
+ struct intel_bw_state *new_bw_state = NULL;
+ struct intel_bw_state *old_bw_state = NULL;
+ unsigned int data_rate;
unsigned int num_active_planes;
struct intel_crtc *crtc;
int i, ret;
+ u32 allowed_points = 0;
+ unsigned int max_bw_point = 0, max_bw = 0;
+ unsigned int num_qgv_points = dev_priv->max_bw[0].num_qgv_points;
+ u32 mask = (1 << num_qgv_points) - 1;
/* FIXME earlier gens need some checks too */
if (INTEL_GEN(dev_priv) < 11)
@@ -446,41 +456,110 @@ int intel_bw_atomic_check(struct intel_atomic_state *state)
old_active_planes == new_active_planes)
continue;
- bw_state = intel_atomic_get_bw_state(state);
- if (IS_ERR(bw_state))
- return PTR_ERR(bw_state);
+ new_bw_state = intel_atomic_get_bw_state(state);
+ if (IS_ERR(new_bw_state))
+ return PTR_ERR(new_bw_state);
- bw_state->data_rate[crtc->pipe] = new_data_rate;
- bw_state->num_active_planes[crtc->pipe] = new_active_planes;
+ new_bw_state->data_rate[crtc->pipe] = new_data_rate;
+ new_bw_state->num_active_planes[crtc->pipe] = new_active_planes;
drm_dbg_kms(&dev_priv->drm,
"pipe %c data rate %u num active planes %u\n",
pipe_name(crtc->pipe),
- bw_state->data_rate[crtc->pipe],
- bw_state->num_active_planes[crtc->pipe]);
+ new_bw_state->data_rate[crtc->pipe],
+ new_bw_state->num_active_planes[crtc->pipe]);
}
- if (!bw_state)
+ if (!new_bw_state)
return 0;
- ret = intel_atomic_lock_global_state(&bw_state->base);
- if (ret)
+ /*
+ * TODO: Should we just call intel_atomic_serialize_global_state here?
+ * we anyway already have different data rates and this call is
+ * is almost similar, except that it doesn't call duplicate_state
+ * hook.
+ */
+ ret = intel_atomic_lock_global_state(&new_bw_state->base);
+ if (ret) {
+ DRM_DEBUG_KMS("Could not lock global state\n");
return ret;
+ }
- data_rate = intel_bw_data_rate(dev_priv, bw_state);
- num_active_planes = intel_bw_num_active_planes(dev_priv, bw_state);
-
- max_data_rate = intel_max_data_rate(dev_priv, num_active_planes);
+ data_rate = intel_bw_data_rate(dev_priv, new_bw_state);
+ num_active_planes = intel_bw_num_active_planes(dev_priv, new_bw_state);
data_rate = DIV_ROUND_UP(data_rate, 1000);
- if (data_rate > max_data_rate) {
- drm_dbg_kms(&dev_priv->drm,
- "Bandwidth %u MB/s exceeds max available %d MB/s (%d active planes)\n",
- data_rate, max_data_rate, num_active_planes);
+ for (i = 0; i < num_qgv_points; i++) {
+ unsigned int max_data_rate;
+
+ max_data_rate = icl_max_bw(dev_priv, num_active_planes, i);
+ /*
+ * We need to know which qgv point gives us
+ * maximum bandwidth in order to disable SAGV
+ * if we find that we exceed SAGV block time
+ * with watermarks. By that moment we already
+ * have those, as it is calculated earlier in
+ * intel_atomic_check,
+ */
+ if (max_data_rate > max_bw) {
+ max_bw_point = i;
+ max_bw = max_data_rate;
+ }
+ if (max_data_rate >= data_rate)
+ allowed_points |= BIT(i);
+ DRM_DEBUG_KMS("QGV point %d: max bw %d required %d\n",
+ i, max_data_rate, data_rate);
+ }
+
+ /*
+ * BSpec states that we always should have at least one allowed point
+ * left, so if we couldn't - simply reject the configuration for obvious
+ * reasons.
+ */
+ if (allowed_points == 0) {
+ DRM_DEBUG_KMS("No QGV points provide sufficient memory"
+ " bandwidth for display configuration.\n");
return -EINVAL;
}
+ /*
+ * Leave only single point with highest bandwidth, if
+ * we can't enable SAGV due to the increased memory latency it may
+ * cause.
+ */
+ if (!intel_can_enable_sagv_for_state(state)) {
+ allowed_points = 1 << max_bw_point;
+ DRM_DEBUG_KMS("No SAGV, using single QGV point %d\n",
+ max_bw_point);
+ }
+ /*
+ * We store the ones which need to be masked as that is what PCode
+ * actually accepts as a parameter.
+ */
+ new_bw_state->qgv_points_mask = (~allowed_points) & mask;
+
+ DRM_DEBUG_KMS("New state %p qgv mask %x\n",
+ state, new_bw_state->qgv_points_mask);
+
+ old_bw_state = intel_atomic_get_old_bw_state(state);
+ if (IS_ERR(old_bw_state)) {
+ DRM_DEBUG_KMS("Could not get old bw state!\n");
+ return PTR_ERR(old_bw_state);
+ }
+
+ /*
+ * If the actual mask had changed we need to make sure that
+ * the commits are serialized(in case this is a nomodeset, nonblocking)
+ */
+ if (new_bw_state->qgv_points_mask != old_bw_state->qgv_points_mask) {
+ ret = intel_atomic_serialize_global_state(&new_bw_state->base);
+ if (ret) {
+ DRM_DEBUG_KMS("Could not serialize global state\n");
+ return ret;
+ }
+ }
+
return 0;
}
diff --git a/drivers/gpu/drm/i915/display/intel_bw.h b/drivers/gpu/drm/i915/display/intel_bw.h
index fb1760125f9d..071deae3a427 100644
--- a/drivers/gpu/drm/i915/display/intel_bw.h
+++ b/drivers/gpu/drm/i915/display/intel_bw.h
@@ -36,6 +36,13 @@ struct intel_bw_state {
*/
bool can_sagv;
+ /*
+ * Current QGV points mask, which restricts
+ * some particular SAGV states, not to confuse
+ * with pipe_sagv_mask.
+ */
+ u8 qgv_points_mask;
+
unsigned int data_rate[I915_MAX_PIPES];
u8 num_active_planes[I915_MAX_PIPES];
};
@@ -56,5 +63,7 @@ int intel_bw_init(struct drm_i915_private *dev_priv);
int intel_bw_atomic_check(struct intel_atomic_state *state);
void intel_bw_crtc_update(struct intel_bw_state *bw_state,
const struct intel_crtc_state *crtc_state);
+int icl_pcode_restrict_qgv_points(struct drm_i915_private *dev_priv,
+ u32 points_mask);
#endif /* __INTEL_BW_H__ */
diff --git a/drivers/gpu/drm/i915/display/intel_display.c b/drivers/gpu/drm/i915/display/intel_display.c
index 6a4d88e4d41a..cd248bd69794 100644
--- a/drivers/gpu/drm/i915/display/intel_display.c
+++ b/drivers/gpu/drm/i915/display/intel_display.c
@@ -15473,6 +15473,105 @@ static void intel_atomic_cleanup_work(struct work_struct *work)
intel_atomic_helper_free_state(i915);
}
+static void intel_qgv_points_mask(struct intel_atomic_state *state)
+{
+ struct drm_device *dev = state->base.dev;
+ struct drm_i915_private *dev_priv = to_i915(dev);
+ int ret;
+ struct intel_bw_state *new_bw_state = NULL;
+ struct intel_bw_state *old_bw_state = NULL;
+ u32 new_mask = 0;
+ unsigned int num_qgv_points = dev_priv->max_bw[0].num_qgv_points;
+ unsigned int mask = (1 << num_qgv_points) - 1;
+
+ new_bw_state = intel_atomic_get_bw_state(state);
+ if (IS_ERR(new_bw_state)) {
+ WARN(1, "Could not get new bw_state!\n");
+ return;
+ }
+
+ old_bw_state = intel_atomic_get_old_bw_state(state);
+ if (IS_ERR(old_bw_state)) {
+ WARN(1, "Could not get old bw_state!\n");
+ return;
+ }
+
+ new_mask = old_bw_state->qgv_points_mask | new_bw_state->qgv_points_mask;
+
+ /*
+ * If new mask is zero - means there is nothing to mask,
+ * we can only unmask, which should be done in unmask.
+ */
+ if (!new_mask)
+ return;
+
+ WARN_ON(new_mask == mask);
+
+ /*
+ * Just return if we can't control SAGV or don't have it.
+ */
+ if (!intel_has_sagv(dev_priv))
+ return;
+
+ /*
+ * Restrict required qgv points before updating the configuration.
+ * According to BSpec we can't mask and unmask qgv points at the same
+ * time. Also masking should be done before updating the configuration
+ * and unmasking afterwards.
+ */
+ ret = icl_pcode_restrict_qgv_points(dev_priv, new_mask);
+ if (ret < 0)
+ DRM_DEBUG_KMS("Could not mask required qgv points(%d)\n",
+ ret);
+}
+
+static void intel_qgv_points_unmask(struct intel_atomic_state *state)
+{
+ struct drm_device *dev = state->base.dev;
+ struct drm_i915_private *dev_priv = to_i915(dev);
+ int ret;
+ struct intel_bw_state *new_bw_state = NULL;
+ struct intel_bw_state *old_bw_state = NULL;
+ u32 new_mask = 0;
+
+ new_bw_state = intel_atomic_get_bw_state(state);
+ if (IS_ERR(new_bw_state)) {
+ WARN(1, "Could not get new bw_state!\n");
+ return;
+ }
+
+ old_bw_state = intel_atomic_get_old_bw_state(state);
+ if (IS_ERR(old_bw_state)) {
+ WARN(1, "Could not get new bw_state!\n");
+ return;
+ }
+
+ new_mask = new_bw_state->qgv_points_mask;
+
+ /*
+ * Just return if we can't control SAGV or don't have it.
+ */
+ if (!intel_has_sagv(dev_priv))
+ return;
+
+ /*
+ * Nothing to unmask
+ */
+ if (new_mask == old_bw_state->qgv_points_mask)
+ return;
+
+ /*
+ * Allow required qgv points after updating the configuration.
+ * According to BSpec we can't mask and unmask qgv points at the same
+ * time. Also masking should be done before updating the configuration
+ * and unmasking afterwards.
+ */
+ ret = icl_pcode_restrict_qgv_points(dev_priv, new_mask);
+ if (ret < 0)
+ DRM_DEBUG_KMS("Could not unmask required qgv points(%d)\n",
+ ret);
+}
+
static void intel_atomic_commit_tail(struct intel_atomic_state *state)
{
struct drm_device *dev = state->base.dev;
@@ -15506,6 +15605,14 @@ static void intel_atomic_commit_tail(struct intel_atomic_state *state)
for_each_new_intel_crtc_in_state(state, crtc, new_crtc_state, i)
crtc->config = new_crtc_state;
+ /*
+ * Now we need to check if SAGV needs to be disabled(i.e QGV points
+ * modified even, when no modeset is done(for example plane updates
+ * can now trigger that).
+ */
+ if ((INTEL_GEN(dev_priv) >= 11))
+ intel_qgv_points_mask(state);
+
if (state->modeset) {
drm_atomic_helper_update_legacy_modeset_state(dev, &state->base);
@@ -15620,6 +15727,8 @@ static void intel_atomic_commit_tail(struct intel_atomic_state *state)
if (INTEL_GEN(dev_priv) < 11) {
if (state->modeset && intel_can_enable_sagv(dev_priv))
intel_enable_sagv(dev_priv);
+ } else {
+ intel_qgv_points_unmask(state);
}
drm_atomic_helper_commit_hw_done(&state->base);
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 9928d00ea0b1..1143f67bba0a 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -841,6 +841,9 @@ enum intel_pipe_crc_source {
INTEL_PIPE_CRC_SOURCE_MAX,
};
+/* BSpec precisely defines this */
+#define NUM_SAGV_POINTS 8
+
#define INTEL_PIPE_CRC_ENTRIES_NR 128
struct intel_pipe_crc {
spinlock_t lock;
--
2.24.1.485.gad05a3d8e5
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
