drm/amdkfd: Ensure consistent barrier state saved in gfx12 trap handler

[ Upstream commit d584198 ]

It is possible for some waves in a workgroup to finish their save
sequence before the group leader has had time to capture the workgroup
barrier state.  When this happens, having those waves exit do impact the
barrier state.  As a consequence, the state captured by the group leader
is invalid, and is eventually incorrectly restored.

This patch proposes to have all waves in a workgroup wait for each other
at the end of their save sequence (just before calling s_endpgm_saved).

Signed-off-by: Lancelot SIX <lancelot.six@amd.com>
Reviewed-by: Jay Cornwall <jay.cornwall@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org # 6.12.x
Signed-off-by: Sasha Levin <sashal@kernel.org>

Author: lancesix

drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h

@@ -4117,7 +4117,8 @@ static const uint32_t cwsr_trap_gfx12_hex[] = {
 	0x0000ffff, 0x8bfe7e7e,
 	0x8bea6a6a, 0xb97af804,
 	0xbe804ec2, 0xbf94fffe,
-	0xbe804a6c, 0xbfb10000,
+	0xbe804a6c, 0xbe804ec2,
+	0xbf94fffe, 0xbfb10000,
 	0xbf9f0000, 0xbf9f0000,
 	0xbf9f0000, 0xbf9f0000,
 	0xbf9f0000, 0x00000000,

drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx12.asm

@@ -1049,6 +1049,10 @@ L_SKIP_BARRIER_RESTORE:
 	s_rfe_b64	s_restore_pc_lo						//Return to the main shader program and resume execution
 
 L_END_PGM:
+	// Make sure that no wave of the workgroup can exit the trap handler
+	// before the workgroup barrier state is saved.
+	s_barrier_signal	-2
+	s_barrier_wait	-2
 	s_endpgm_saved
 end