summaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx10.asm
diff options
context:
space:
mode:
authorDave Airlie <airlied@redhat.com>2022-10-04 09:42:23 +1000
committerDave Airlie <airlied@redhat.com>2022-10-04 09:42:24 +1000
commit65898687cf7392c372ea8d04a88617e2cb794465 (patch)
treeb524feb316f5a12f6f799b2fa14f3678966a9111 /drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx10.asm
parent4ae9f874dc1d662ce7bfdb8144903608bcc3706b (diff)
parentb292cafe2dd02d96a07147e4b160927e8399d5cc (diff)
Merge tag 'amd-drm-next-6.1-2022-09-30' of https://gitlab.freedesktop.org/agd5f/linux into drm-next
amd-drm-next-6.1-2022-09-30: amdgpu: - RLC FW code cleanup - RLC fixes for GC 11.x - SMU 13.x fixes - CP FW code cleanup - SDMA FW code cleanup - GC 11.x fixes - DCN 3.2.x fixes - DCN 3.1.4 fixes - Misc fixes - RAS fixes - SR-IOV fixes - VCN 4.x fixes amdkfd: - GC 11.x fixes - Xnack fixes - UBSAN warning fix Signed-off-by: Dave Airlie <airlied@redhat.com> From: Alex Deucher <alexander.deucher@amd.com> Link: https://patchwork.freedesktop.org/patch/msgid/20220930162012.5823-1-alexander.deucher@amd.com
Diffstat (limited to 'drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx10.asm')
-rw-r--r--drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx10.asm21
1 files changed, 21 insertions, 0 deletions
diff --git a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx10.asm b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx10.asm
index 250ab007399b..0f81670f6f9c 100644
--- a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx10.asm
+++ b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx10.asm
@@ -43,12 +43,14 @@
#define HAVE_XNACK (ASIC_FAMILY < CHIP_SIENNA_CICHLID)
#define HAVE_SENDMSG_RTN (ASIC_FAMILY >= CHIP_PLUM_BONITO)
#define HAVE_BUFFER_LDS_LOAD (ASIC_FAMILY < CHIP_PLUM_BONITO)
+#define SW_SA_TRAP (ASIC_FAMILY >= CHIP_PLUM_BONITO)
var SINGLE_STEP_MISSED_WORKAROUND = 1 //workaround for lost MODE.DEBUG_EN exception when SAVECTX raised
var SQ_WAVE_STATUS_SPI_PRIO_MASK = 0x00000006
var SQ_WAVE_STATUS_HALT_MASK = 0x2000
var SQ_WAVE_STATUS_ECC_ERR_MASK = 0x20000
+var SQ_WAVE_STATUS_TRAP_EN_SHIFT = 6
var SQ_WAVE_LDS_ALLOC_LDS_SIZE_SHIFT = 12
var SQ_WAVE_LDS_ALLOC_LDS_SIZE_SIZE = 9
@@ -183,6 +185,13 @@ L_SKIP_RESTORE:
s_getreg_b32 s_save_trapsts, hwreg(HW_REG_TRAPSTS)
+#if SW_SA_TRAP
+ // If ttmp1[31] is set then trap may occur early.
+ // Spin wait until SAVECTX exception is raised.
+ s_bitcmp1_b32 s_save_pc_hi, 31
+ s_cbranch_scc1 L_CHECK_SAVE
+#endif
+
s_and_b32 ttmp2, s_save_status, SQ_WAVE_STATUS_HALT_MASK
s_cbranch_scc0 L_NOT_HALTED
@@ -1061,8 +1070,20 @@ L_RESTORE_HWREG:
s_and_b32 s_restore_pc_hi, s_restore_pc_hi, 0x0000ffff //pc[47:32] //Do it here in order not to affect STATUS
s_and_b64 exec, exec, exec // Restore STATUS.EXECZ, not writable by s_setreg_b32
s_and_b64 vcc, vcc, vcc // Restore STATUS.VCCZ, not writable by s_setreg_b32
+
+#if SW_SA_TRAP
+ // If traps are enabled then return to the shader with PRIV=0.
+ // Otherwise retain PRIV=1 for subsequent context save requests.
+ s_getreg_b32 s_restore_tmp, hwreg(HW_REG_STATUS)
+ s_bitcmp1_b32 s_restore_tmp, SQ_WAVE_STATUS_TRAP_EN_SHIFT
+ s_cbranch_scc1 L_RETURN_WITHOUT_PRIV
+
s_setreg_b32 hwreg(HW_REG_STATUS), s_restore_status // SCC is included, which is changed by previous salu
+ s_setpc_b64 [s_restore_pc_lo, s_restore_pc_hi]
+L_RETURN_WITHOUT_PRIV:
+#endif
+ s_setreg_b32 hwreg(HW_REG_STATUS), s_restore_status // SCC is included, which is changed by previous salu
s_rfe_b64 s_restore_pc_lo //Return to the main shader program and resume execution
L_END_PGM: