drm/amdkfd: Fix saving the ACC vgprs for Aldebaran

get_num_acc_vgprs does not set status.scc if the number of acc vgprs
is 0, so use an and instruction to set the condition code.

The Aldebaran handler binary was not based on the latest version of
the sources, so this update to the binary is the minimal change only
adding two instructions to set the condition code.

A newer version of the handler should be generated and tested in
another commit.

Signed-off-by: Laurent Morichetti <laurent.morichetti@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

diff --git a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h
index c8453da..475f897 100644 (file)
--- a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h
+++ b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h
@@ -1576,7 +1576,7 @@ static const uint32_t cwsr_trap_arcturus_hex[] = {
 };
 
 static const uint32_t cwsr_trap_aldebaran_hex[] = {
-       0xbf820001, 0xbf8202cd,
+       0xbf820001, 0xbf8202ce,
        0xb8f8f802, 0x89788678,
        0xb8eef801, 0x866eff6e,
        0x00000800, 0xbf840003,
@@ -1873,114 +1873,115 @@ static const uint32_t cwsr_trap_aldebaran_hex[] = {
        0xb8fb2985, 0x807b817b,
        0x8e7b837b, 0xb8fa2b05,
        0x807a817a, 0x8e7a827a,
-       0x80fb7a7b, 0xbf84007a,
-       0x807bff7b, 0x00001000,
-       0xbefc0080, 0xbf11017c,
-       0x867aff78, 0x00400000,
-       0xbf850003, 0xb8faf803,
-       0x897a7aff, 0x10000000,
-       0xbf850059, 0xd3d84000,
-       0x18000100, 0xd3d84001,
-       0x18000101, 0xd3d84002,
-       0x18000102, 0xd3d84003,
-       0x18000103, 0xbe840080,
-       0xd2890000, 0x00000900,
-       0x80048104, 0xd2890001,
-       0x00000900, 0x80048104,
-       0xd2890002, 0x00000900,
-       0x80048104, 0xd2890003,
-       0x00000900, 0x80048104,
-       0xc069003a, 0x00000070,
-       0xbf8cc07f, 0x80709070,
-       0xbf06c004, 0xbf84ffee,
+       0x80fb7a7b, 0x867b7b7b,
+       0xbf84007a, 0x807bff7b,
+       0x00001000, 0xbefc0080,
+       0xbf11017c, 0x867aff78,
+       0x00400000, 0xbf850003,
+       0xb8faf803, 0x897a7aff,
+       0x10000000, 0xbf850059,
+       0xd3d84000, 0x18000100,
+       0xd3d84001, 0x18000101,
+       0xd3d84002, 0x18000102,
+       0xd3d84003, 0x18000103,
        0xbe840080, 0xd2890000,
-       0x00000901, 0x80048104,
-       0xd2890001, 0x00000901,
+       0x00000900, 0x80048104,
+       0xd2890001, 0x00000900,
        0x80048104, 0xd2890002,
-       0x00000901, 0x80048104,
-       0xd2890003, 0x00000901,
+       0x00000900, 0x80048104,
+       0xd2890003, 0x00000900,
        0x80048104, 0xc069003a,
        0x00000070, 0xbf8cc07f,
        0x80709070, 0xbf06c004,
        0xbf84ffee, 0xbe840080,
-       0xd2890000, 0x00000902,
+       0xd2890000, 0x00000901,
        0x80048104, 0xd2890001,
-       0x00000902, 0x80048104,
-       0xd2890002, 0x00000902,
+       0x00000901, 0x80048104,
+       0xd2890002, 0x00000901,
        0x80048104, 0xd2890003,
-       0x00000902, 0x80048104,
+       0x00000901, 0x80048104,
        0xc069003a, 0x00000070,
        0xbf8cc07f, 0x80709070,
        0xbf06c004, 0xbf84ffee,
        0xbe840080, 0xd2890000,
-       0x00000903, 0x80048104,
-       0xd2890001, 0x00000903,
+       0x00000902, 0x80048104,
+       0xd2890001, 0x00000902,
        0x80048104, 0xd2890002,
-       0x00000903, 0x80048104,
-       0xd2890003, 0x00000903,
+       0x00000902, 0x80048104,
+       0xd2890003, 0x00000902,
        0x80048104, 0xc069003a,
        0x00000070, 0xbf8cc07f,
        0x80709070, 0xbf06c004,
-       0xbf84ffee, 0x807c847c,
-       0xbf0a7b7c, 0xbf85ffa9,
-       0xbf9c0000, 0xbf820016,
-       0xd3d84000, 0x18000100,
-       0xd3d84001, 0x18000101,
-       0xd3d84002, 0x18000102,
-       0xd3d84003, 0x18000103,
-       0xe0724000, 0x701d0000,
-       0xe0724100, 0x701d0100,
-       0xe0724200, 0x701d0200,
-       0xe0724300, 0x701d0300,
-       0x807c847c, 0x8070ff70,
-       0x00000400, 0xbf0a7b7c,
-       0xbf85ffeb, 0xbf9c0000,
-       0xbf820100, 0xbef4007e,
-       0x8675ff7f, 0x0000ffff,
-       0x8775ff75, 0x00040000,
-       0xbef60080, 0xbef700ff,
-       0x00807fac, 0x866eff7f,
-       0x08000000, 0x8f6e836e,
-       0x87776e77, 0x866eff7f,
-       0x70000000, 0x8f6e816e,
-       0x87776e77, 0x866eff7f,
-       0x04000000, 0xbf84001f,
+       0xbf84ffee, 0xbe840080,
+       0xd2890000, 0x00000903,
+       0x80048104, 0xd2890001,
+       0x00000903, 0x80048104,
+       0xd2890002, 0x00000903,
+       0x80048104, 0xd2890003,
+       0x00000903, 0x80048104,
+       0xc069003a, 0x00000070,
+       0xbf8cc07f, 0x80709070,
+       0xbf06c004, 0xbf84ffee,
+       0x807c847c, 0xbf0a7b7c,
+       0xbf85ffa9, 0xbf9c0000,
+       0xbf820016, 0xd3d84000,
+       0x18000100, 0xd3d84001,
+       0x18000101, 0xd3d84002,
+       0x18000102, 0xd3d84003,
+       0x18000103, 0xe0724000,
+       0x701d0000, 0xe0724100,
+       0x701d0100, 0xe0724200,
+       0x701d0200, 0xe0724300,
+       0x701d0300, 0x807c847c,
+       0x8070ff70, 0x00000400,
+       0xbf0a7b7c, 0xbf85ffeb,
+       0xbf9c0000, 0xbf820101,
+       0xbef4007e, 0x8675ff7f,
+       0x0000ffff, 0x8775ff75,
+       0x00040000, 0xbef60080,
+       0xbef700ff, 0x00807fac,
+       0x866eff7f, 0x08000000,
+       0x8f6e836e, 0x87776e77,
+       0x866eff7f, 0x70000000,
+       0x8f6e816e, 0x87776e77,
+       0x866eff7f, 0x04000000,
+       0xbf84001f, 0xbefe00c1,
+       0xbeff00c1, 0xb8ef4306,
+       0x866fc16f, 0xbf84001a,
+       0x8e6f866f, 0x8e6f826f,
+       0xbef6006f, 0xb8f82985,
+       0x80788178, 0x8e788a78,
+       0x8e788178, 0xb8ee1605,
+       0x806e816e, 0x8e6e866e,
+       0x80786e78, 0x8078ff78,
+       0x00000080, 0xbef600ff,
+       0x01000000, 0xbefc0080,
+       0xe0510000, 0x781d0000,
+       0xe0510100, 0x781d0000,
+       0x807cff7c, 0x00000200,
+       0x8078ff78, 0x00000200,
+       0xbf0a6f7c, 0xbf85fff6,
        0xbefe00c1, 0xbeff00c1,
-       0xb8ef4306, 0x866fc16f,
-       0xbf84001a, 0x8e6f866f,
-       0x8e6f826f, 0xbef6006f,
-       0xb8f82985, 0x80788178,
-       0x8e788a78, 0x8e788178,
-       0xb8ee1605, 0x806e816e,
-       0x8e6e866e, 0x80786e78,
-       0x8078ff78, 0x00000080,
        0xbef600ff, 0x01000000,
-       0xbefc0080, 0xe0510000,
-       0x781d0000, 0xe0510100,
-       0x781d0000, 0x807cff7c,
-       0x00000200, 0x8078ff78,
-       0x00000200, 0xbf0a6f7c,
-       0xbf85fff6, 0xbefe00c1,
-       0xbeff00c1, 0xbef600ff,
-       0x01000000, 0xb8ef2b05,
-       0x806f816f, 0x8e6f826f,
-       0x806fff6f, 0x00008000,
-       0xbef80080, 0xbeee0078,
-       0x8078ff78, 0x00000400,
-       0xbefc0084, 0xbf11087c,
-       0xe0524000, 0x781d0000,
-       0xe0524100, 0x781d0100,
-       0xe0524200, 0x781d0200,
-       0xe0524300, 0x781d0300,
-       0xbf8c0f70, 0x7e000300,
-       0x7e020301, 0x7e040302,
-       0x7e060303, 0x807c847c,
-       0x8078ff78, 0x00000400,
-       0xbf0a6f7c, 0xbf85ffee,
-       0xb8ef2985, 0x806f816f,
-       0x8e6f836f, 0xb8f92b05,
-       0x80798179, 0x8e798279,
-       0x80ef796f, 0xbf84001a,
+       0xb8ef2b05, 0x806f816f,
+       0x8e6f826f, 0x806fff6f,
+       0x00008000, 0xbef80080,
+       0xbeee0078, 0x8078ff78,
+       0x00000400, 0xbefc0084,
+       0xbf11087c, 0xe0524000,
+       0x781d0000, 0xe0524100,
+       0x781d0100, 0xe0524200,
+       0x781d0200, 0xe0524300,
+       0x781d0300, 0xbf8c0f70,
+       0x7e000300, 0x7e020301,
+       0x7e040302, 0x7e060303,
+       0x807c847c, 0x8078ff78,
+       0x00000400, 0xbf0a6f7c,
+       0xbf85ffee, 0xb8ef2985,
+       0x806f816f, 0x8e6f836f,
+       0xb8f92b05, 0x80798179,
+       0x8e798279, 0x80ef796f,
+       0x866f6f6f, 0xbf84001a,
        0x806fff6f, 0x00008000,
        0xbefc0080, 0xbf11087c,
        0xe0524000, 0x781d0000,

diff --git a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm
index 4b3f076..eed78a0 100644 (file)
--- a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm
+++ b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm
@@ -632,6 +632,7 @@ L_SAVE_VGPR_END:
 #if ASIC_FAMILY >= CHIP_ALDEBARAN
     // ACC VGPR count may differ from ARCH VGPR count.
     get_num_acc_vgprs(s_save_alloc_size, s_save_tmp)
+    s_and_b32       s_save_alloc_size, s_save_alloc_size, s_save_alloc_size
     s_cbranch_scc0  L_SAVE_ACCVGPR_END
     s_add_u32      s_save_alloc_size, s_save_alloc_size, 0x1000                    //add 0x1000 since we compare m0 against it later
 #endif
@@ -769,6 +770,7 @@ L_RESTORE:
 #if ASIC_FAMILY >= CHIP_ALDEBARAN
     // ACC VGPR count may differ from ARCH VGPR count.
     get_num_acc_vgprs(s_restore_alloc_size, s_restore_tmp2)
+    s_and_b32       s_restore_alloc_size, s_restore_alloc_size, s_restore_alloc_size
     s_cbranch_scc0  L_RESTORE_ACCVGPR_END
     s_add_u32      s_restore_alloc_size, s_restore_alloc_size, 0x8000                      //add 0x8000 since we compare m0 against it later
 #endif