drm/etnaviv: fix flush sequence logic

The current logic uses the flush sequence from the current address
space. This is harmless when deducing the flush requirements for the
current submit, as either the incoming address space is the same one
as the currently active one or we switch context, in which case the
flush is unconditional.

However, this sequence is also stored as the current flush sequence
of the GPU. If we switch context the stored flush sequence will no
longer belong to the currently active address space. This incoherency
can then cause missed flushes, resulting in translation errors.

Fixes: 27b6727 ("drm/etnaviv: rework MMU handling")
Signed-off-by: Tomeu Vizoso <[email protected]>
Signed-off-by: Lucas Stach <[email protected]>
Reviewed-by: Christian Gmeiner <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

Author: tomeuv

drivers/gpu/drm/etnaviv/etnaviv_buffer.c

@@ -347,7 +347,7 @@ void etnaviv_buffer_queue(struct etnaviv_gpu *gpu, u32 exec_state,
 	u32 link_target, link_dwords;
 	bool switch_context = gpu->exec_state != exec_state;
 	bool switch_mmu_context = gpu->mmu_context != mmu_context;
-	unsigned int new_flush_seq = READ_ONCE(gpu->mmu_context->flush_seq);
+	unsigned int new_flush_seq = READ_ONCE(mmu_context->flush_seq);
 	bool need_flush = switch_mmu_context || gpu->flush_seq != new_flush_seq;
 	bool has_blt = !!(gpu->identity.minor_features5 &
 			  chipMinorFeatures5_BLT_ENGINE);