Commit 6a817038 authored by Jay Cornwall's avatar Jay Cornwall Committed by Alex Deucher
Browse files

drm/amdkfd: Add gfx11 trap handler 



Based on gfx10 with following changes:

- GPR_ALLOC.VGPR_SIZE field moved (and size corrected in gfx10)
- s_sendmsg_rtn_b64 replaces some s_sendmsg/s_getreg
- Buffer instructions no longer have direct-to-LDS modifier

Signed-off-by: default avatarJay Cornwall <jay.cornwall@amd.com>
Reviewed-by: default avatarLaurent Morichetti <laurent.morichetti@amd.com>
Signed-off-by: default avatarEric Huang <jinhuieric.huang@amd.com>
Acked-by: default avatarAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: default avatarAlex Deucher <alexander.deucher@amd.com>
parent 5e613723

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

+450 −13
Original line number Diff line number Diff line
@@ -776,7 +776,7 @@ static const uint32_t cwsr_trap_nv1x_hex[] = { 
	0xe0704100, 0x705d0100,

	0xe0704200, 0x705d0200,

	0xe0704300, 0x705d0300,

	0xb9702a05, 0x80708170,

	0xb9703a05, 0x80708170,

	0xbf0d9973, 0xbf850002,

	0x8f708970, 0xbf820001,

	0x8f708a70, 0xb97a1e06,
@@ -855,7 +855,7 @@ static const uint32_t cwsr_trap_nv1x_hex[] = { 
	0x877aff6d, 0x80000000,

	0xbf840040, 0x8f7b867b,

	0x8f7b827b, 0xbef6037b,

	0xb9702a05, 0x80708170,

	0xb9703a05, 0x80708170,

	0xbf0d9973, 0xbf850002,

	0x8f708970, 0xbf820001,

	0x8f708a70, 0xb97a1e06,
@@ -891,7 +891,7 @@ static const uint32_t cwsr_trap_nv1x_hex[] = { 
	0xbef003ff, 0x00000200,

	0xbeff0380, 0xbf820003,

	0xbef003ff, 0x00000400,

	0xbeff03c1, 0xb97b2a05,

	0xbeff03c1, 0xb97b3a05,

	0x807b817b, 0x8f7b827b,

	0x907c9973, 0x877c817c,

	0xbf06817c, 0xbf850017,
@@ -939,7 +939,7 @@ static const uint32_t cwsr_trap_nv1x_hex[] = { 
	0xb96f4306, 0x876fc16f,

	0xbf840029, 0x8f6f866f,

	0x8f6f826f, 0xbef6036f,

	0xb9782a05, 0x80788178,

	0xb9783a05, 0x80788178,

	0xbf0d9972, 0xbf850002,

	0x8f788978, 0xbf820001,

	0x8f788a78, 0xb96e1e06,
@@ -962,7 +962,7 @@ static const uint32_t cwsr_trap_nv1x_hex[] = { 
	0x907c9972, 0x877c817c,

	0xbf06817c, 0xbf850002,

	0xbeff0380, 0xbf820001,

	0xbeff03c1, 0xb96f2a05,

	0xbeff03c1, 0xb96f3a05,

	0x806f816f, 0x8f6f826f,

	0x907c9972, 0x877c817c,

	0xbf06817c, 0xbf850024,
@@ -1010,7 +1010,7 @@ static const uint32_t cwsr_trap_nv1x_hex[] = { 
	0x6e5d0100, 0xe0304200,

	0x6e5d0200, 0xe0304300,

	0x6e5d0300, 0xbf8c3f70,

	0xb9782a05, 0x80788178,

	0xb9783a05, 0x80788178,

	0xbf0d9972, 0xbf850002,

	0x8f788978, 0xbf820001,

	0x8f788a78, 0xb96e1e06,
@@ -1037,7 +1037,7 @@ static const uint32_t cwsr_trap_nv1x_hex[] = { 
	0xbe8c310c, 0xbe8e310e,

	0xbf06807c, 0xbf84fff0,

	0xba80f801, 0x00000000,

	0xbf8a0000, 0xb9782a05,

	0xbf8a0000, 0xb9783a05,

	0x80788178, 0xbf0d9972,

	0xbf850002, 0x8f788978,

	0xbf820001, 0x8f788a78,
@@ -2261,7 +2261,7 @@ static const uint32_t cwsr_trap_gfx10_hex[] = { 
	0xbf8a0000, 0x877aff6d,

	0x80000000, 0xbf840040,

	0x8f7b867b, 0x8f7b827b,

	0xbef6037b, 0xb9702a05,

	0xbef6037b, 0xb9703a05,

	0x80708170, 0xbf0d9973,

	0xbf850002, 0x8f708970,

	0xbf820001, 0x8f708a70,
@@ -2298,7 +2298,7 @@ static const uint32_t cwsr_trap_gfx10_hex[] = { 
	0x00000200, 0xbeff0380,

	0xbf820003, 0xbef003ff,

	0x00000400, 0xbeff03c1,

	0xb97b2a05, 0x807b817b,

	0xb97b3a05, 0x807b817b,

	0x8f7b827b, 0x907c9973,

	0x877c817c, 0xbf06817c,

	0xbf850017, 0xbef603ff,
@@ -2345,7 +2345,7 @@ static const uint32_t cwsr_trap_gfx10_hex[] = { 
	0xbeff03c1, 0xb96f4306,

	0x876fc16f, 0xbf840029,

	0x8f6f866f, 0x8f6f826f,

	0xbef6036f, 0xb9782a05,

	0xbef6036f, 0xb9783a05,

	0x80788178, 0xbf0d9972,

	0xbf850002, 0x8f788978,

	0xbf820001, 0x8f788a78,
@@ -2369,7 +2369,7 @@ static const uint32_t cwsr_trap_gfx10_hex[] = { 
	0x877c817c, 0xbf06817c,

	0xbf850002, 0xbeff0380,

	0xbf820001, 0xbeff03c1,

	0xb96f2a05, 0x806f816f,

	0xb96f3a05, 0x806f816f,

	0x8f6f826f, 0x907c9972,

	0x877c817c, 0xbf06817c,

	0xbf850024, 0xbef603ff,
@@ -2416,7 +2416,7 @@ static const uint32_t cwsr_trap_gfx10_hex[] = { 
	0xe0304100, 0x6e5d0100,

	0xe0304200, 0x6e5d0200,

	0xe0304300, 0x6e5d0300,

	0xbf8c3f70, 0xb9782a05,

	0xbf8c3f70, 0xb9783a05,

	0x80788178, 0xbf0d9972,

	0xbf850002, 0x8f788978,

	0xbf820001, 0x8f788a78,
@@ -2444,7 +2444,7 @@ static const uint32_t cwsr_trap_gfx10_hex[] = { 
	0xbe8e310e, 0xbf06807c,

	0xbf84fff0, 0xba80f801,

	0x00000000, 0xbf8a0000,

	0xb9782a05, 0x80788178,

	0xb9783a05, 0x80788178,

	0xbf0d9972, 0xbf850002,

	0x8f788978, 0xbf820001,

	0x8f788a78, 0xb96e1e06,
@@ -2494,3 +2494,440 @@ static const uint32_t cwsr_trap_gfx10_hex[] = { 
	0xbf9f0000, 0xbf9f0000,

	0xbf9f0000, 0x00000000,

};



static const uint32_t cwsr_trap_gfx11_hex[] = {

	0xbfa00001, 0xbfa0021b,

	0xb0804006, 0xb8f8f802,

	0x91788678, 0xb8fbf803,

	0x8b6eff78, 0x00002000,

	0xbfa10009, 0x8b6eff6d,

	0x00ff0000, 0xbfa2001e,

	0x8b6eff7b, 0x00000400,

	0xbfa20041, 0xbf830010,

	0xb8fbf803, 0xbfa0fffa,

	0x8b6eff7b, 0x00000900,

	0xbfa20015, 0x8b6eff7b,

	0x000071ff, 0xbfa10008,

	0x8b6fff7b, 0x00007080,

	0xbfa10001, 0xbeee1287,

	0xb8eff801, 0x846e8c6e,

	0x8b6e6f6e, 0xbfa2000a,

	0x8b6eff6d, 0x00ff0000,

	0xbfa20007, 0xb8eef801,

	0x8b6eff6e, 0x00000800,

	0xbfa20003, 0x8b6eff7b,

	0x00000400, 0xbfa20026,

	0xbefa4d82, 0xbf89fc07,

	0x84fa887a, 0xf4005bbd,

	0xf8000010, 0xbf89fc07,

	0x846e976e, 0x9177ff77,

	0x00800000, 0x8c776e77,

	0xf4045bbd, 0xf8000000,

	0xbf89fc07, 0xf4045ebd,

	0xf8000008, 0xbf89fc07,

	0x8bee6e6e, 0xbfa10001,

	0xbe80486e, 0x8b6eff6d,

	0x01ff0000, 0xbfa20005,

	0x8c78ff78, 0x00002000,

	0x80ec886c, 0x82ed806d,

	0xbfa00005, 0x8b6eff6d,

	0x01000000, 0xbfa20002,

	0x806c846c, 0x826d806d,

	0x8b6dff6d, 0x0000ffff,

	0x8bfe7e7e, 0x8bea6a6a,

	0xb978f802, 0xbe804a6c,

	0x8b6dff6d, 0x0000ffff,

	0xbefa0080, 0xb97a0283,

	0xbeee007e, 0xbeef007f,

	0xbefe0180, 0xbefe4d84,

	0xbf89fc07, 0x8b7aff7f,

	0x04000000, 0x847a857a,

	0x8c6d7a6d, 0xbefa007e,

	0x8b7bff7f, 0x0000ffff,

	0xbefe00c1, 0xbeff00c1,

	0xdca6c000, 0x007a0000,

	0x7e000280, 0xbefe007a,

	0xbeff007b, 0xb8fb02dc,

	0x847b997b, 0xb8fa3b05,

	0x807a817a, 0xbf0d997b,

	0xbfa20002, 0x847a897a,

	0xbfa00001, 0x847a8a7a,

	0xb8fb1e06, 0x847b8a7b,

	0x807a7b7a, 0x8b7bff7f,

	0x0000ffff, 0x807aff7a,

	0x00000200, 0x807a7e7a,

	0x827b807b, 0xd7610000,

	0x00010870, 0xd7610000,

	0x00010a71, 0xd7610000,

	0x00010c72, 0xd7610000,

	0x00010e73, 0xd7610000,

	0x00011074, 0xd7610000,

	0x00011275, 0xd7610000,

	0x00011476, 0xd7610000,

	0x00011677, 0xd7610000,

	0x00011a79, 0xd7610000,

	0x00011c7e, 0xd7610000,

	0x00011e7f, 0xbefe00ff,

	0x00003fff, 0xbeff0080,

	0xdca6c040, 0x007a0000,

	0xd760007a, 0x00011d00,

	0xd760007b, 0x00011f00,

	0xbefe007a, 0xbeff007b,

	0xbef4007e, 0x8b75ff7f,

	0x0000ffff, 0x8c75ff75,

	0x00040000, 0xbef60080,

	0xbef700ff, 0x10807fac,

	0xbef1007d, 0xbef00080,

	0xb8f302dc, 0x84739973,

	0xbefe00c1, 0x857d9973,

	0x8b7d817d, 0xbf06817d,

	0xbfa20002, 0xbeff0080,

	0xbfa00002, 0xbeff00c1,

	0xbfa00009, 0xbef600ff,

	0x01000000, 0xe0685080,

	0x701d0100, 0xe0685100,

	0x701d0200, 0xe0685180,

	0x701d0300, 0xbfa00008,

	0xbef600ff, 0x01000000,

	0xe0685100, 0x701d0100,

	0xe0685200, 0x701d0200,

	0xe0685300, 0x701d0300,

	0xb8f03b05, 0x80708170,

	0xbf0d9973, 0xbfa20002,

	0x84708970, 0xbfa00001,

	0x84708a70, 0xb8fa1e06,

	0x847a8a7a, 0x80707a70,

	0x8070ff70, 0x00000200,

	0xbef600ff, 0x01000000,

	0x7e000280, 0x7e020280,

	0x7e040280, 0xbefd0080,

	0xd7610002, 0x0000fa71,

	0x807d817d, 0xd7610002,

	0x0000fa6c, 0x807d817d,

	0x917aff6d, 0x80000000,

	0xd7610002, 0x0000fa7a,

	0x807d817d, 0xd7610002,

	0x0000fa6e, 0x807d817d,

	0xd7610002, 0x0000fa6f,

	0x807d817d, 0xd7610002,

	0x0000fa78, 0x807d817d,

	0xb8faf803, 0xd7610002,

	0x0000fa7a, 0x807d817d,

	0xd7610002, 0x0000fa7b,

	0x807d817d, 0xb8f1f801,

	0xd7610002, 0x0000fa71,

	0x807d817d, 0xb8f1f814,

	0xd7610002, 0x0000fa71,

	0x807d817d, 0xb8f1f815,

	0xd7610002, 0x0000fa71,

	0x807d817d, 0xbefe00ff,

	0x0000ffff, 0xbeff0080,

	0xe0685000, 0x701d0200,

	0xbefe00c1, 0xb8f03b05,

	0x80708170, 0xbf0d9973,

	0xbfa20002, 0x84708970,

	0xbfa00001, 0x84708a70,

	0xb8fa1e06, 0x847a8a7a,

	0x80707a70, 0xbef600ff,

	0x01000000, 0xbef90080,

	0xbefd0080, 0xbf800000,

	0xbe804100, 0xbe824102,

	0xbe844104, 0xbe864106,

	0xbe884108, 0xbe8a410a,

	0xbe8c410c, 0xbe8e410e,

	0xd7610002, 0x0000f200,

	0x80798179, 0xd7610002,

	0x0000f201, 0x80798179,

	0xd7610002, 0x0000f202,

	0x80798179, 0xd7610002,

	0x0000f203, 0x80798179,

	0xd7610002, 0x0000f204,

	0x80798179, 0xd7610002,

	0x0000f205, 0x80798179,

	0xd7610002, 0x0000f206,

	0x80798179, 0xd7610002,

	0x0000f207, 0x80798179,

	0xd7610002, 0x0000f208,

	0x80798179, 0xd7610002,

	0x0000f209, 0x80798179,

	0xd7610002, 0x0000f20a,

	0x80798179, 0xd7610002,

	0x0000f20b, 0x80798179,

	0xd7610002, 0x0000f20c,

	0x80798179, 0xd7610002,

	0x0000f20d, 0x80798179,

	0xd7610002, 0x0000f20e,

	0x80798179, 0xd7610002,

	0x0000f20f, 0x80798179,

	0xbf06a079, 0xbfa10006,

	0xe0685000, 0x701d0200,

	0x8070ff70, 0x00000080,

	0xbef90080, 0x7e040280,

	0x807d907d, 0xbf0aff7d,

	0x00000060, 0xbfa2ffbc,

	0xbe804100, 0xbe824102,

	0xbe844104, 0xbe864106,

	0xbe884108, 0xbe8a410a,

	0xd7610002, 0x0000f200,

	0x80798179, 0xd7610002,

	0x0000f201, 0x80798179,

	0xd7610002, 0x0000f202,

	0x80798179, 0xd7610002,

	0x0000f203, 0x80798179,

	0xd7610002, 0x0000f204,

	0x80798179, 0xd7610002,

	0x0000f205, 0x80798179,

	0xd7610002, 0x0000f206,

	0x80798179, 0xd7610002,

	0x0000f207, 0x80798179,

	0xd7610002, 0x0000f208,

	0x80798179, 0xd7610002,

	0x0000f209, 0x80798179,

	0xd7610002, 0x0000f20a,

	0x80798179, 0xd7610002,

	0x0000f20b, 0x80798179,

	0xe0685000, 0x701d0200,

	0xbefe00c1, 0x857d9973,

	0x8b7d817d, 0xbf06817d,

	0xbfa20002, 0xbeff0080,

	0xbfa00001, 0xbeff00c1,

	0xb8fb4306, 0x8b7bc17b,

	0xbfa10044, 0xbfbd0000,

	0x8b7aff6d, 0x80000000,

	0xbfa10040, 0x847b867b,

	0x847b827b, 0xbef6007b,

	0xb8f03b05, 0x80708170,

	0xbf0d9973, 0xbfa20002,

	0x84708970, 0xbfa00001,

	0x84708a70, 0xb8fa1e06,

	0x847a8a7a, 0x80707a70,

	0x8070ff70, 0x00000200,

	0x8070ff70, 0x00000080,

	0xbef600ff, 0x01000000,

	0xd71f0000, 0x000100c1,

	0xd7200000, 0x000200c1,

	0x16000084, 0x857d9973,

	0x8b7d817d, 0xbf06817d,

	0xbefd0080, 0xbfa20012,

	0xbe8300ff, 0x00000080,

	0xbf800000, 0xbf800000,

	0xbf800000, 0xd8d80000,

	0x01000000, 0xbf890000,

	0xe0685000, 0x701d0100,

	0x807d037d, 0x80700370,

	0xd5250000, 0x0001ff00,

	0x00000080, 0xbf0a7b7d,

	0xbfa2fff4, 0xbfa00011,

	0xbe8300ff, 0x00000100,

	0xbf800000, 0xbf800000,

	0xbf800000, 0xd8d80000,

	0x01000000, 0xbf890000,

	0xe0685000, 0x701d0100,

	0x807d037d, 0x80700370,

	0xd5250000, 0x0001ff00,

	0x00000100, 0xbf0a7b7d,

	0xbfa2fff4, 0xbefe00c1,

	0x857d9973, 0x8b7d817d,

	0xbf06817d, 0xbfa20004,

	0xbef000ff, 0x00000200,

	0xbeff0080, 0xbfa00003,

	0xbef000ff, 0x00000400,

	0xbeff00c1, 0xb8fb3b05,

	0x807b817b, 0x847b827b,

	0x857d9973, 0x8b7d817d,

	0xbf06817d, 0xbfa20017,

	0xbef600ff, 0x01000000,

	0xbefd0084, 0xbf0a7b7d,

	0xbfa10037, 0x7e008700,

	0x7e028701, 0x7e048702,

	0x7e068703, 0xe0685000,

	0x701d0000, 0xe0685080,

	0x701d0100, 0xe0685100,

	0x701d0200, 0xe0685180,

	0x701d0300, 0x807d847d,

	0x8070ff70, 0x00000200,

	0xbf0a7b7d, 0xbfa2ffef,

	0xbfa00025, 0xbef600ff,

	0x01000000, 0xbefd0084,

	0xbf0a7b7d, 0xbfa10011,

	0x7e008700, 0x7e028701,

	0x7e048702, 0x7e068703,

	0xe0685000, 0x701d0000,

	0xe0685100, 0x701d0100,

	0xe0685200, 0x701d0200,

	0xe0685300, 0x701d0300,

	0x807d847d, 0x8070ff70,

	0x00000400, 0xbf0a7b7d,

	0xbfa2ffef, 0xb8fb1e06,

	0x8b7bc17b, 0xbfa1000c,

	0x847b837b, 0x807b7d7b,

	0xbefe00c1, 0xbeff0080,

	0x7e008700, 0xe0685000,

	0x701d0000, 0x807d817d,

	0x8070ff70, 0x00000080,

	0xbf0a7b7d, 0xbfa2fff8,

	0xbfa00141, 0xbef4007e,

	0x8b75ff7f, 0x0000ffff,

	0x8c75ff75, 0x00040000,

	0xbef60080, 0xbef700ff,

	0x10807fac, 0xb8f202dc,

	0x84729972, 0x8b6eff7f,

	0x04000000, 0xbfa1003a,

	0xbefe00c1, 0x857d9972,

	0x8b7d817d, 0xbf06817d,

	0xbfa20002, 0xbeff0080,

	0xbfa00001, 0xbeff00c1,

	0xb8ef4306, 0x8b6fc16f,

	0xbfa1002f, 0x846f866f,

	0x846f826f, 0xbef6006f,

	0xb8f83b05, 0x80788178,

	0xbf0d9972, 0xbfa20002,

	0x84788978, 0xbfa00001,

	0x84788a78, 0xb8ee1e06,

	0x846e8a6e, 0x80786e78,

	0x8078ff78, 0x00000200,

	0x8078ff78, 0x00000080,

	0xbef600ff, 0x01000000,

	0x857d9972, 0x8b7d817d,

	0xbf06817d, 0xbefd0080,

	0xbfa2000c, 0xe0500000,

	0x781d0000, 0xbf8903f7,

	0xdac00000, 0x00000000,

	0x807dff7d, 0x00000080,

	0x8078ff78, 0x00000080,

	0xbf0a6f7d, 0xbfa2fff5,

	0xbfa0000b, 0xe0500000,

	0x781d0000, 0xbf8903f7,

	0xdac00000, 0x00000000,

	0x807dff7d, 0x00000100,

	0x8078ff78, 0x00000100,

	0xbf0a6f7d, 0xbfa2fff5,

	0xbef80080, 0xbefe00c1,

	0x857d9972, 0x8b7d817d,

	0xbf06817d, 0xbfa20002,

	0xbeff0080, 0xbfa00001,

	0xbeff00c1, 0xb8ef3b05,

	0x806f816f, 0x846f826f,

	0x857d9972, 0x8b7d817d,

	0xbf06817d, 0xbfa20024,

	0xbef600ff, 0x01000000,

	0xbeee0078, 0x8078ff78,

	0x00000200, 0xbefd0084,

	0xbf0a6f7d, 0xbfa10050,

	0xe0505000, 0x781d0000,

	0xe0505080, 0x781d0100,

	0xe0505100, 0x781d0200,

	0xe0505180, 0x781d0300,

	0xbf8903f7, 0x7e008500,

	0x7e028501, 0x7e048502,

	0x7e068503, 0x807d847d,

	0x8078ff78, 0x00000200,

	0xbf0a6f7d, 0xbfa2ffee,

	0xe0505000, 0x6e1d0000,

	0xe0505080, 0x6e1d0100,

	0xe0505100, 0x6e1d0200,

	0xe0505180, 0x6e1d0300,

	0xbf8903f7, 0xbfa00034,

	0xbef600ff, 0x01000000,

	0xbeee0078, 0x8078ff78,

	0x00000400, 0xbefd0084,

	0xbf0a6f7d, 0xbfa10012,

	0xe0505000, 0x781d0000,

	0xe0505100, 0x781d0100,

	0xe0505200, 0x781d0200,

	0xe0505300, 0x781d0300,

	0xbf8903f7, 0x7e008500,

	0x7e028501, 0x7e048502,

	0x7e068503, 0x807d847d,

	0x8078ff78, 0x00000400,

	0xbf0a6f7d, 0xbfa2ffee,

	0xb8ef1e06, 0x8b6fc16f,

	0xbfa1000e, 0x846f836f,

	0x806f7d6f, 0xbefe00c1,

	0xbeff0080, 0xe0505000,

	0x781d0000, 0xbf8903f7,

	0x7e008500, 0x807d817d,

	0x8078ff78, 0x00000080,

	0xbf0a6f7d, 0xbfa2fff7,

	0xbeff00c1, 0xe0505000,

	0x6e1d0000, 0xe0505100,

	0x6e1d0100, 0xe0505200,

	0x6e1d0200, 0xe0505300,

	0x6e1d0300, 0xbf8903f7,

	0xb8f83b05, 0x80788178,

	0xbf0d9972, 0xbfa20002,

	0x84788978, 0xbfa00001,

	0x84788a78, 0xb8ee1e06,

	0x846e8a6e, 0x80786e78,

	0x8078ff78, 0x00000200,

	0x80f8ff78, 0x00000050,

	0xbef600ff, 0x01000000,

	0xbefd00ff, 0x0000006c,

	0x80f89078, 0xf428403a,

	0xf0000000, 0xbf89fc07,

	0x80fd847d, 0xbf800000,

	0xbe804300, 0xbe824302,

	0x80f8a078, 0xf42c403a,

	0xf0000000, 0xbf89fc07,

	0x80fd887d, 0xbf800000,

	0xbe804300, 0xbe824302,

	0xbe844304, 0xbe864306,

	0x80f8c078, 0xf430403a,

	0xf0000000, 0xbf89fc07,

	0x80fd907d, 0xbf800000,

	0xbe804300, 0xbe824302,

	0xbe844304, 0xbe864306,

	0xbe884308, 0xbe8a430a,

	0xbe8c430c, 0xbe8e430e,

	0xbf06807d, 0xbfa1fff0,

	0xb980f801, 0x00000000,

	0xbfbd0000, 0xb8f83b05,

	0x80788178, 0xbf0d9972,

	0xbfa20002, 0x84788978,

	0xbfa00001, 0x84788a78,

	0xb8ee1e06, 0x846e8a6e,

	0x80786e78, 0x8078ff78,

	0x00000200, 0xbef600ff,

	0x01000000, 0xf4205bfa,

	0xf0000000, 0x80788478,

	0xf4205b3a, 0xf0000000,

	0x80788478, 0xf4205b7a,

	0xf0000000, 0x80788478,

	0xf4205c3a, 0xf0000000,

	0x80788478, 0xf4205c7a,

	0xf0000000, 0x80788478,

	0xf4205eba, 0xf0000000,

	0x80788478, 0xf4205efa,

	0xf0000000, 0x80788478,

	0xf4205e7a, 0xf0000000,

	0x80788478, 0xf4205cfa,

	0xf0000000, 0x80788478,

	0xf4205bba, 0xf0000000,

	0x80788478, 0xbf89fc07,

	0xb96ef814, 0xf4205bba,

	0xf0000000, 0x80788478,

	0xbf89fc07, 0xb96ef815,

	0xbefd006f, 0xbefe0070,

	0xbeff0071, 0x8b6f7bff,

	0x000003ff, 0xb96f4803,

	0x8b6f7bff, 0xfffff800,

	0x856f8b6f, 0xb96fa2c3,

	0xb973f801, 0xb8ee3b05,

	0x806e816e, 0xbf0d9972,

	0xbfa20002, 0x846e896e,

	0xbfa00001, 0x846e8a6e,

	0xb8ef1e06, 0x846f8a6f,

	0x806e6f6e, 0x806eff6e,

	0x00000200, 0x806e746e,

	0x826f8075, 0x8b6fff6f,

	0x0000ffff, 0xf4085c37,

	0xf8000050, 0xf4085d37,

	0xf8000060, 0xf4005e77,

	0xf8000074, 0xbf89fc07,

	0x8b6dff6d, 0x0000ffff,

	0x8bfe7e7e, 0x8bea6a6a,

	0xb97af802, 0xbe804a6c,

	0xbfb00000, 0xbf9f0000,

	0xbf9f0000, 0xbf9f0000,

	0xbf9f0000, 0xbf9f0000,

};

drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx10.asm

+52 −17
Original line number Diff line number Diff line
@@ -23,15 +23,26 @@ 
/* To compile this assembly code:

 *

 * Navi1x:

 *   cpp -DASIC_TARGET_NAVI1X=1 cwsr_trap_handler_gfx10.asm -P -o nv1x.sp3

 *   sp3-nv1x nv1x.sp3 -hex nv1x.hex

 *   cpp -DASIC_FAMILY=CHIP_NAVI10 cwsr_trap_handler_gfx10.asm -P -o nv1x.sp3

 *   sp3 nv1x.sp3 -hex nv1x.hex

 *

 * Others:

 *   cpp -DASIC_TARGET_NAVI1X=0 cwsr_trap_handler_gfx10.asm -P -o gfx10.sp3

 *   sp3-gfx10 gfx10.sp3 -hex gfx10.hex

 * gfx10:

 *   cpp -DASIC_FAMILY=CHIP_SIENNA_CICHLID cwsr_trap_handler_gfx10.asm -P -o gfx10.sp3

 *   sp3 gfx10.sp3 -hex gfx10.hex

 *

 * gfx11:

 *   cpp -DASIC_FAMILY=CHIP_PLUM_BONITO cwsr_trap_handler_gfx10.asm -P -o gfx11.sp3

 *   sp3 gfx11.sp3 -hex gfx11.hex

 */



#define NO_SQC_STORE !ASIC_TARGET_NAVI1X

#define CHIP_NAVI10 26

#define CHIP_SIENNA_CICHLID 30

#define CHIP_PLUM_BONITO 36



#define NO_SQC_STORE (ASIC_FAMILY >= CHIP_SIENNA_CICHLID)

#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)



var SINGLE_STEP_MISSED_WORKAROUND		= 1	//workaround for lost MODE.DEBUG_EN exception when SAVECTX raised
@@ -41,15 +52,18 @@ var SQ_WAVE_STATUS_ECC_ERR_MASK = 0x20000 


var SQ_WAVE_LDS_ALLOC_LDS_SIZE_SHIFT		= 12

var SQ_WAVE_LDS_ALLOC_LDS_SIZE_SIZE		= 9

var SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SHIFT		= 8

var SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SIZE		= 6

var SQ_WAVE_GPR_ALLOC_SGPR_SIZE_SHIFT		= 24

var SQ_WAVE_GPR_ALLOC_SGPR_SIZE_SIZE		= 4

var SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SIZE		= 8

var SQ_WAVE_LDS_ALLOC_VGPR_SHARED_SIZE_SHIFT	= 24

var SQ_WAVE_LDS_ALLOC_VGPR_SHARED_SIZE_SIZE	= 4

var SQ_WAVE_IB_STS2_WAVE64_SHIFT		= 11

var SQ_WAVE_IB_STS2_WAVE64_SIZE			= 1



#if ASIC_FAMILY < CHIP_PLUM_BONITO

var SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SHIFT		= 8

#else

var SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SHIFT		= 12

#endif



var SQ_WAVE_TRAPSTS_SAVECTX_MASK		= 0x400

var SQ_WAVE_TRAPSTS_EXCP_MASK			= 0x1FF

var SQ_WAVE_TRAPSTS_SAVECTX_SHIFT		= 10
@@ -231,15 +245,20 @@ end 
	s_cbranch_scc1	L_SAVE



L_FETCH_2ND_TRAP:

#if ASIC_TARGET_NAVI1X

#if HAVE_XNACK

	save_and_clear_ib_sts(ttmp14, ttmp15)

#endif



	// Read second-level TBA/TMA from first-level TMA and jump if available.

	// ttmp[2:5] and ttmp12 can be used (others hold SPI-initialized debug data)

	// ttmp12 holds SQ_WAVE_STATUS

#if HAVE_SENDMSG_RTN

	s_sendmsg_rtn_b64       [ttmp14, ttmp15], sendmsg(MSG_RTN_GET_TMA)

	s_waitcnt       lgkmcnt(0)

#else

	s_getreg_b32	ttmp14, hwreg(HW_REG_SHADER_TMA_LO)

	s_getreg_b32	ttmp15, hwreg(HW_REG_SHADER_TMA_HI)

#endif

	s_lshl_b64	[ttmp14, ttmp15], [ttmp14, ttmp15], 0x8



	s_load_dword    ttmp2, [ttmp14, ttmp15], 0x10 glc:1			// debug trap enabled flag
@@ -282,7 +301,7 @@ L_TRAP_CASE: 
L_EXIT_TRAP:

	s_and_b32	ttmp1, ttmp1, 0xFFFF



#if ASIC_TARGET_NAVI1X

#if HAVE_XNACK

	restore_ib_sts(ttmp14, ttmp15)

#endif
@@ -298,7 +317,7 @@ L_SAVE: 
	s_mov_b32	s_save_tmp, 0

	s_setreg_b32	hwreg(HW_REG_TRAPSTS, SQ_WAVE_TRAPSTS_SAVECTX_SHIFT, 1), s_save_tmp	//clear saveCtx bit



#if ASIC_TARGET_NAVI1X

#if HAVE_XNACK

	save_and_clear_ib_sts(s_save_tmp, s_save_trapsts)

#endif
@@ -307,9 +326,13 @@ L_SAVE: 
	s_mov_b32	s_save_exec_hi, exec_hi

	s_mov_b64	exec, 0x0						//clear EXEC to get ready to receive



#if HAVE_SENDMSG_RTN

	s_sendmsg_rtn_b64       [exec_lo, exec_hi], sendmsg(MSG_RTN_SAVE_WAVE)

#else

	s_sendmsg	sendmsg(MSG_SAVEWAVE)					//send SPI a message and wait for SPI's write to EXEC

#endif



#if ASIC_TARGET_NAVI1X

#if ASIC_FAMILY < CHIP_SIENNA_CICHLID

L_SLEEP:

	// sleep 1 (64clk) is not enough for 8 waves per SIMD, which will cause

	// SQ hang, since the 7,8th wave could not get arbit to exec inst, while
@@ -389,7 +412,7 @@ L_SLEEP: 
	s_mov_b32	s_save_mem_offset, 0x0

	get_wave_size(s_wave_size)



#if ASIC_TARGET_NAVI1X

#if HAVE_XNACK

	// Save and clear vector XNACK state late to free up SGPRs.

	s_getreg_b32	s_save_xnack_mask, hwreg(HW_REG_SHADER_XNACK_MASK)

	s_setreg_imm32_b32	hwreg(HW_REG_SHADER_XNACK_MASK), 0x0
@@ -777,7 +800,13 @@ L_RESTORE_LDS_NORMAL: 
	s_cbranch_scc1	L_RESTORE_LDS_LOOP_W64



L_RESTORE_LDS_LOOP_W32:

#if HAVE_BUFFER_LDS_LOAD

	buffer_load_dword	v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset lds:1	// first 64DW

#else

	buffer_load_dword       v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset

	s_waitcnt	vmcnt(0)

	ds_store_addtid_b32     v0

#endif

	s_add_u32	m0, m0, 128						// 128 DW

	s_add_u32	s_restore_mem_offset, s_restore_mem_offset, 128		//mem offset increased by 128DW

	s_cmp_lt_u32	m0, s_restore_alloc_size				//scc=(m0 < s_restore_alloc_size) ? 1 : 0
@@ -785,7 +814,13 @@ L_RESTORE_LDS_LOOP_W32: 
	s_branch	L_RESTORE_VGPR



L_RESTORE_LDS_LOOP_W64:

#if HAVE_BUFFER_LDS_LOAD

	buffer_load_dword	v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset lds:1	// first 64DW

#else

	buffer_load_dword       v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset

	s_waitcnt	vmcnt(0)

	ds_store_addtid_b32     v0

#endif

	s_add_u32	m0, m0, 256						// 256 DW

	s_add_u32	s_restore_mem_offset, s_restore_mem_offset, 256		//mem offset increased by 256DW

	s_cmp_lt_u32	m0, s_restore_alloc_size				//scc=(m0 < s_restore_alloc_size) ? 1 : 0
@@ -996,7 +1031,7 @@ L_RESTORE_HWREG: 
	s_and_b32	s_restore_m0, SQ_WAVE_TRAPSTS_PRE_SAVECTX_MASK, s_restore_trapsts

	s_setreg_b32	hwreg(HW_REG_TRAPSTS, SQ_WAVE_TRAPSTS_PRE_SAVECTX_SHIFT, SQ_WAVE_TRAPSTS_PRE_SAVECTX_SIZE), s_restore_m0



#if ASIC_TARGET_NAVI1X

#if HAVE_XNACK

	s_setreg_b32	hwreg(HW_REG_SHADER_XNACK_MASK), s_restore_xnack_mask

#endif
@@ -1019,7 +1054,7 @@ L_RESTORE_HWREG: 
	s_load_dword	ttmp13, [s_restore_ttmps_lo, s_restore_ttmps_hi], 0x74 glc:1

	s_waitcnt	lgkmcnt(0)



#if ASIC_TARGET_NAVI1X

#if HAVE_XNACK

	restore_ib_sts(s_restore_tmp, s_restore_m0)

#endif

drivers/gpu/drm/amd/amdkfd/kfd_device.c

+5 −1
Original line number Diff line number Diff line
@@ -441,10 +441,14 @@ static void kfd_cwsr_init(struct kfd_dev *kfd) 
			BUILD_BUG_ON(sizeof(cwsr_trap_nv1x_hex) > PAGE_SIZE);

			kfd->cwsr_isa = cwsr_trap_nv1x_hex;

			kfd->cwsr_isa_size = sizeof(cwsr_trap_nv1x_hex);

		} else {

		} else if (KFD_GC_VERSION(kfd) < IP_VERSION(11, 0, 0)) {

			BUILD_BUG_ON(sizeof(cwsr_trap_gfx10_hex) > PAGE_SIZE);

			kfd->cwsr_isa = cwsr_trap_gfx10_hex;

			kfd->cwsr_isa_size = sizeof(cwsr_trap_gfx10_hex);

		} else {

			BUILD_BUG_ON(sizeof(cwsr_trap_gfx11_hex) > PAGE_SIZE);

			kfd->cwsr_isa = cwsr_trap_gfx11_hex;

			kfd->cwsr_isa_size = sizeof(cwsr_trap_gfx11_hex);

		}



		kfd->cwsr_enabled = true;