The code claims to grab a runtime PM ref when at least one CRTC is active, but that's not actually the case as we grab a runtime PM ref whenever a CRTC is enabled regardless of it's DPMS state. Meaning that we can end up keeping the GPU awake when there are no screens enabled, something we don't really want to do.

Note that we fixed this same issue for nv50 a while ago in:

commit e5d54f193572 ("drm/nouveau/drm/nouveau: Fix runtime PM leak in nv50_disp_atomic_commit()")

Since we're about to remove nouveau_drm->have_disp_power_ref in the next commit, let's also simplify the RPM code here while we're at it: grab a ref during a modeset, grab additional RPM refs for each CRTC enabled by said modeset, and drop an RPM ref for each CRTC disabled by said modeset. This allows us to keep the GPU awake whenever screens are turned on, without needing to use nouveau_drm->have_disp_power_ref.

Signed-off-by: Lyude Paul lyude@redhat.com --- drivers/gpu/drm/nouveau/dispnv04/crtc.c | 18 ++++-------------- 1 file changed, 4 insertions(+), 14 deletions(-)

diff --git a/drivers/gpu/drm/nouveau/dispnv04/crtc.c b/drivers/gpu/drm/nouveau/dispnv04/crtc.c index f22f01020625..08ad8e3b9cd2 100644 --- a/drivers/gpu/drm/nouveau/dispnv04/crtc.c +++ b/drivers/gpu/drm/nouveau/dispnv04/crtc.c @@ -183,6 +183,10 @@ nv_crtc_dpms(struct drm_crtc *crtc, int mode) return;

nv_crtc->last_dpms = mode; + if (mode == DRM_MODE_DPMS_ON) + pm_runtime_get_noresume(dev->dev); + else + pm_runtime_put_noidle(dev->dev);

if (nv_two_heads(dev)) NVSetOwner(dev, nv_crtc->index); @@ -1045,7 +1049,6 @@ nouveau_crtc_set_config(struct drm_mode_set *set,

dev = set->crtc->dev;

- /* get a pm reference here */ ret = pm_runtime_get_sync(dev->dev); if (ret < 0 && ret != -EACCES) return ret; @@ -1061,19 +1064,6 @@ nouveau_crtc_set_config(struct drm_mode_set *set, }

pm_runtime_mark_last_busy(dev->dev); - /* if we have active crtcs and we don't have a power ref, - take the current one */ - if (active && !drm->have_disp_power_ref) { - drm->have_disp_power_ref = true; - return ret; - } - /* if we have no active crtcs, then drop the power ref - we got before */ - if (!active && drm->have_disp_power_ref) { - pm_runtime_put_autosuspend(dev->dev); - drm->have_disp_power_ref = false; - } - /* drop the power reference we got coming in here */ pm_runtime_put_autosuspend(dev->dev); return ret; }

This is something that got noticed a while ago back when I was fixing a large number of runtime PM related issues in nouveau, but never got fixed:

https://patchwork.freedesktop.org/series/46815/#rev7

It's not safe to iterate the entire list of CRTCs in nv50_disp_atomic_commit(), as we could be doing a non-blocking modeset on one CRTC in parallel with one or more other CRTCs. Likewise, this means it's also not safe to do so in order to track runtime PM state. While this code is certainly wrong, so far the only issues I've seen this cause in the wild is the occasional PM ref unbalance after an atomic check failure + module reloading (since the PCI device will outlive nouveau in such scenarios).

So, do this far more elegantly: grab a runtime PM ref across the modeset and commit tail, then grab/put references for each CRTC enable/disable. This also ends up being much simpler then the previous broken solution we had.

Finally, since we've removed all it's users: get rid of nouveau_drm->have_disp_power_ref.

Signed-off-by: Lyude Paul lyude@redhat.com --- drivers/gpu/drm/nouveau/dispnv50/disp.c | 38 +++++++++++-------------- drivers/gpu/drm/nouveau/nouveau_drv.h | 3 -- 2 files changed, 17 insertions(+), 24 deletions(-)

diff --git a/drivers/gpu/drm/nouveau/dispnv50/disp.c b/drivers/gpu/drm/nouveau/dispnv50/disp.c index 126703816794..659e6fa645cb 100644 --- a/drivers/gpu/drm/nouveau/dispnv50/disp.c +++ b/drivers/gpu/drm/nouveau/dispnv50/disp.c @@ -1826,8 +1826,11 @@ nv50_disp_atomic_commit_tail(struct drm_atomic_state *state)

NV_ATOMIC(drm, "%s: clr %04x (set %04x)\n", crtc->name, asyh->clr.mask, asyh->set.mask); - if (old_crtc_state->active && !new_crtc_state->active) + + if (old_crtc_state->active && !new_crtc_state->active) { + pm_runtime_put_noidle(dev->dev); drm_crtc_vblank_off(crtc); + }

if (asyh->clr.mask) { nv50_head_flush_clr(head, asyh, atom->flush_disable); @@ -1913,8 +1916,10 @@ nv50_disp_atomic_commit_tail(struct drm_atomic_state *state) }

if (new_crtc_state->active) { - if (!old_crtc_state->active) + if (!old_crtc_state->active) { drm_crtc_vblank_on(crtc); + pm_runtime_get_noresume(dev->dev); + } if (new_crtc_state->event) drm_crtc_vblank_get(crtc); } @@ -1979,6 +1984,10 @@ nv50_disp_atomic_commit_tail(struct drm_atomic_state *state) drm_atomic_helper_cleanup_planes(dev, state); drm_atomic_helper_commit_cleanup_done(state); drm_atomic_state_put(state); + + /* Drop the RPM ref we got from nv50_disp_atomic_commit() */ + pm_runtime_mark_last_busy(dev->dev); + pm_runtime_put_autosuspend(dev->dev); }

static void @@ -1993,11 +2002,8 @@ static int nv50_disp_atomic_commit(struct drm_device *dev, struct drm_atomic_state *state, bool nonblock) { - struct nouveau_drm *drm = nouveau_drm(dev); struct drm_plane_state *new_plane_state; struct drm_plane *plane; - struct drm_crtc *crtc; - bool active = false; int ret, i;

ret = pm_runtime_get_sync(dev->dev); @@ -2034,27 +2040,17 @@ nv50_disp_atomic_commit(struct drm_device *dev,

drm_atomic_state_get(state);

+ /* + * Grab another RPM ref for the commit tail, which will release the + * ref when it's finished + */ + pm_runtime_get_noresume(dev->dev); + if (nonblock) queue_work(system_unbound_wq, &state->commit_work); else nv50_disp_atomic_commit_tail(state);

- drm_for_each_crtc(crtc, dev) { - if (crtc->state->active) { - if (!drm->have_disp_power_ref) { - drm->have_disp_power_ref = true; - return 0; - } - active = true; - break; - } - } - - if (!active && drm->have_disp_power_ref) { - pm_runtime_put_autosuspend(dev->dev); - drm->have_disp_power_ref = false; - } - err_cleanup: if (ret) drm_atomic_helper_cleanup_planes(dev, state); diff --git a/drivers/gpu/drm/nouveau/nouveau_drv.h b/drivers/gpu/drm/nouveau/nouveau_drv.h index aae035816383..411352dd5390 100644 --- a/drivers/gpu/drm/nouveau/nouveau_drv.h +++ b/drivers/gpu/drm/nouveau/nouveau_drv.h @@ -204,9 +204,6 @@ struct nouveau_drm { /* led management */ struct nouveau_led *led;

- /* display power reference */ - bool have_disp_power_ref; - struct dev_pm_domain vga_pm_domain;

struct nouveau_svm *svm;