drm/nouveau: Only release VCPI slots on mode changes
authorLyude Paul <lyude@redhat.com>
Thu, 1 Aug 2019 22:02:15 +0000 (18:02 -0400)
committerDave Airlie <airlied@redhat.com>
Thu, 1 Aug 2019 23:49:37 +0000 (09:49 +1000)
Looks like a regression got introduced into nv50_mstc_atomic_check()
that somehow didn't get found until now. If userspace changes
crtc_state->active to false but leaves the CRTC enabled, we end up
calling drm_dp_atomic_find_vcpi_slots() using the PBN calculated in
asyh->dp.pbn. However, if the display is inactive we end up calculating
a PBN of 0, which inadvertently causes us to have an allocation of 0.
>From there, if userspace then disables the CRTC afterwards we end up
accidentally attempting to free the VCPI twice:

WARNING: CPU: 0 PID: 1484 at drivers/gpu/drm/drm_dp_mst_topology.c:3336
drm_dp_atomic_release_vcpi_slots+0x87/0xb0 [drm_kms_helper]
RIP: 0010:drm_dp_atomic_release_vcpi_slots+0x87/0xb0 [drm_kms_helper]
Call Trace:
 drm_atomic_helper_check_modeset+0x3f3/0xa60 [drm_kms_helper]
 ? drm_atomic_check_only+0x43/0x780 [drm]
 drm_atomic_helper_check+0x15/0x90 [drm_kms_helper]
 nv50_disp_atomic_check+0x83/0x1d0 [nouveau]
 drm_atomic_check_only+0x54d/0x780 [drm]
 ? drm_atomic_set_crtc_for_connector+0xec/0x100 [drm]
 drm_atomic_commit+0x13/0x50 [drm]
 drm_atomic_helper_set_config+0x81/0x90 [drm_kms_helper]
 drm_mode_setcrtc+0x194/0x6a0 [drm]
 ? vprintk_emit+0x16a/0x230
 ? drm_ioctl+0x163/0x390 [drm]
 ? drm_mode_getcrtc+0x180/0x180 [drm]
 drm_ioctl_kernel+0xaa/0xf0 [drm]
 drm_ioctl+0x208/0x390 [drm]
 ? drm_mode_getcrtc+0x180/0x180 [drm]
 nouveau_drm_ioctl+0x63/0xb0 [nouveau]
 do_vfs_ioctl+0x405/0x660
 ? recalc_sigpending+0x17/0x50
 ? _copy_from_user+0x37/0x60
 ksys_ioctl+0x5e/0x90
 ? exit_to_usermode_loop+0x92/0xe0
 __x64_sys_ioctl+0x16/0x20
 do_syscall_64+0x59/0x190
 entry_SYSCALL_64_after_hwframe+0x44/0xa9
WARNING: CPU: 0 PID: 1484 at drivers/gpu/drm/drm_dp_mst_topology.c:3336
drm_dp_atomic_release_vcpi_slots+0x87/0xb0 [drm_kms_helper]
---[ end trace 4c395c0c51b1f88d ]---
[drm:drm_dp_atomic_release_vcpi_slots [drm_kms_helper]] *ERROR* no VCPI for
[MST PORT:00000000e288eb7d] found in mst state 000000008e642070

So, fix this by doing what we probably should have done from the start: only
call drm_dp_atomic_find_vcpi_slots() when crtc_state->mode_changed is set, so
that VCPI allocations remain for as long as the CRTC is enabled.

Signed-off-by: Lyude Paul <lyude@redhat.com>
Fixes: 232c9eec417a ("drm/nouveau: Use atomic VCPI helpers for MST")
Cc: Lyude Paul <lyude@redhat.com>
Cc: Ben Skeggs <bskeggs@redhat.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: David Airlie <airlied@redhat.com>
Cc: Jerry Zuo <Jerry.Zuo@amd.com>
Cc: Harry Wentland <harry.wentland@amd.com>
Cc: Juston Li <juston.li@intel.com>
Cc: Karol Herbst <karolherbst@gmail.com>
Cc: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Cc: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: <stable@vger.kernel.org> # v5.1+
Acked-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190801220216.15323-1-lyude@redhat.com
drivers/gpu/drm/nouveau/dispnv50/disp.c

index 8497768f1b4102ac76d7e5d45257aec2bb72a82b..126703816794e77a521437271e09a34a09015f4b 100644 (file)
@@ -780,7 +780,7 @@ nv50_msto_atomic_check(struct drm_encoder *encoder,
                        drm_dp_calc_pbn_mode(crtc_state->adjusted_mode.clock,
                                             connector->display_info.bpc * 3);
 
-       if (drm_atomic_crtc_needs_modeset(crtc_state)) {
+       if (crtc_state->mode_changed) {
                slots = drm_dp_atomic_find_vcpi_slots(state, &mstm->mgr,
                                                      mstc->port,
                                                      asyh->dp.pbn);