| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Date: Wed, 15 Jun 2022 06:41:51 +0800
From: kernel test robot <lkp@...el.com>
To: Rob Clark <robdclark@...omium.org>
Cc: llvm@...ts.linux.dev, kbuild-all@...ts.01.org,
linux-kernel@...r.kernel.org, Erico Nunes <nunes.erico@...il.com>,
Eric Anholt <eric@...olt.net>
Subject: [anholt:5.17-for-mesa-ci 2375/2390]
drivers/gpu/drm/msm/adreno/a6xx_gpu.c:1115:4: error: implicit declaration of
function 'writeq'
tree: https://github.com/anholt/linux 5.17-for-mesa-ci
head: bc830a9e22faa411feb4b2ef4b25f7e6d04d36fe
commit: 07aeca39b9491d3455451da0923b19dc0c480e41 [2375/2390] drm/msm/a6xx: make GPUs SMMU context bank available in it's aperture.
config: hexagon-randconfig-r045-20220613 (https://download.01.org/0day-ci/archive/20220615/202206150657.egHDOJhU-lkp@intel.com/config)
compiler: clang version 15.0.0 (https://github.com/llvm/llvm-project c97436f8b6e2718286e8496faf53a2c800e281cf)
reproduce (this is a W=1 build):
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
chmod +x ~/bin/make.cross
# https://github.com/anholt/linux/commit/07aeca39b9491d3455451da0923b19dc0c480e41
git remote add anholt https://github.com/anholt/linux
git fetch --no-tags anholt 5.17-for-mesa-ci
git checkout 07aeca39b9491d3455451da0923b19dc0c480e41
# save the config file
mkdir build_dir && cp config build_dir/.config
COMPILER_INSTALL_PATH=$HOME/0day COMPILER=clang make.cross W=1 O=build_dir ARCH=hexagon SHELL=/bin/bash drivers/gpu/drm/msm/
If you fix the issue, kindly add following tag where applicable
Reported-by: kernel test robot <lkp@...el.com>
All errors (new ones prefixed by >>):
>> drivers/gpu/drm/msm/adreno/a6xx_gpu.c:1115:4: error: implicit declaration of function 'writeq' [-Werror,-Wimplicit-function-declaration]
writeq(0x48000, reg); /* offset of cb0 from gpu's base */
^
1 error generated.
vim +/writeq +1115 drivers/gpu/drm/msm/adreno/a6xx_gpu.c
872
873 #define A6XX_INT_MASK (A6XX_RBBM_INT_0_MASK_CP_AHB_ERROR | \
874 A6XX_RBBM_INT_0_MASK_RBBM_ATB_ASYNCFIFO_OVERFLOW | \
875 A6XX_RBBM_INT_0_MASK_CP_HW_ERROR | \
876 A6XX_RBBM_INT_0_MASK_CP_IB2 | \
877 A6XX_RBBM_INT_0_MASK_CP_IB1 | \
878 A6XX_RBBM_INT_0_MASK_CP_RB | \
879 A6XX_RBBM_INT_0_MASK_CP_CACHE_FLUSH_TS | \
880 A6XX_RBBM_INT_0_MASK_RBBM_ATB_BUS_OVERFLOW | \
881 A6XX_RBBM_INT_0_MASK_RBBM_HANG_DETECT | \
882 A6XX_RBBM_INT_0_MASK_UCHE_OOB_ACCESS | \
883 A6XX_RBBM_INT_0_MASK_UCHE_TRAP_INTR)
884
885 static int hw_init(struct msm_gpu *gpu)
886 {
887 struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
888 struct a6xx_gpu *a6xx_gpu = to_a6xx_gpu(adreno_gpu);
889 int ret;
890
891 /* Make sure the GMU keeps the GPU on while we set it up */
892 a6xx_gmu_set_oob(&a6xx_gpu->gmu, GMU_OOB_GPU_SET);
893
894 gpu_write(gpu, REG_A6XX_RBBM_SECVID_TSB_CNTL, 0);
895
896 /*
897 * Disable the trusted memory range - we don't actually supported secure
898 * memory rendering at this point in time and we don't want to block off
899 * part of the virtual memory space.
900 */
901 gpu_write64(gpu, REG_A6XX_RBBM_SECVID_TSB_TRUSTED_BASE_LO,
902 REG_A6XX_RBBM_SECVID_TSB_TRUSTED_BASE_HI, 0x00000000);
903 gpu_write(gpu, REG_A6XX_RBBM_SECVID_TSB_TRUSTED_SIZE, 0x00000000);
904
905 /* Turn on 64 bit addressing for all blocks */
906 gpu_write(gpu, REG_A6XX_CP_ADDR_MODE_CNTL, 0x1);
907 gpu_write(gpu, REG_A6XX_VSC_ADDR_MODE_CNTL, 0x1);
908 gpu_write(gpu, REG_A6XX_GRAS_ADDR_MODE_CNTL, 0x1);
909 gpu_write(gpu, REG_A6XX_RB_ADDR_MODE_CNTL, 0x1);
910 gpu_write(gpu, REG_A6XX_PC_ADDR_MODE_CNTL, 0x1);
911 gpu_write(gpu, REG_A6XX_HLSQ_ADDR_MODE_CNTL, 0x1);
912 gpu_write(gpu, REG_A6XX_VFD_ADDR_MODE_CNTL, 0x1);
913 gpu_write(gpu, REG_A6XX_VPC_ADDR_MODE_CNTL, 0x1);
914 gpu_write(gpu, REG_A6XX_UCHE_ADDR_MODE_CNTL, 0x1);
915 gpu_write(gpu, REG_A6XX_SP_ADDR_MODE_CNTL, 0x1);
916 gpu_write(gpu, REG_A6XX_TPL1_ADDR_MODE_CNTL, 0x1);
917 gpu_write(gpu, REG_A6XX_RBBM_SECVID_TSB_ADDR_MODE_CNTL, 0x1);
918
919 /* enable hardware clockgating */
920 a6xx_set_hwcg(gpu, true);
921
922 /* VBIF/GBIF start*/
923 if (adreno_is_a640_family(adreno_gpu) ||
924 adreno_is_a650_family(adreno_gpu)) {
925 gpu_write(gpu, REG_A6XX_GBIF_QSB_SIDE0, 0x00071620);
926 gpu_write(gpu, REG_A6XX_GBIF_QSB_SIDE1, 0x00071620);
927 gpu_write(gpu, REG_A6XX_GBIF_QSB_SIDE2, 0x00071620);
928 gpu_write(gpu, REG_A6XX_GBIF_QSB_SIDE3, 0x00071620);
929 gpu_write(gpu, REG_A6XX_GBIF_QSB_SIDE3, 0x00071620);
930 gpu_write(gpu, REG_A6XX_RBBM_GBIF_CLIENT_QOS_CNTL, 0x3);
931 } else {
932 gpu_write(gpu, REG_A6XX_RBBM_VBIF_CLIENT_QOS_CNTL, 0x3);
933 }
934
935 if (adreno_is_a630(adreno_gpu))
936 gpu_write(gpu, REG_A6XX_VBIF_GATE_OFF_WRREQ_EN, 0x00000009);
937
938 /* Make all blocks contribute to the GPU BUSY perf counter */
939 gpu_write(gpu, REG_A6XX_RBBM_PERFCTR_GPU_BUSY_MASKED, 0xffffffff);
940
941 /* Disable L2 bypass in the UCHE */
942 gpu_write(gpu, REG_A6XX_UCHE_WRITE_RANGE_MAX_LO, 0xffffffc0);
943 gpu_write(gpu, REG_A6XX_UCHE_WRITE_RANGE_MAX_HI, 0x0001ffff);
944 gpu_write(gpu, REG_A6XX_UCHE_TRAP_BASE_LO, 0xfffff000);
945 gpu_write(gpu, REG_A6XX_UCHE_TRAP_BASE_HI, 0x0001ffff);
946 gpu_write(gpu, REG_A6XX_UCHE_WRITE_THRU_BASE_LO, 0xfffff000);
947 gpu_write(gpu, REG_A6XX_UCHE_WRITE_THRU_BASE_HI, 0x0001ffff);
948
949 if (!adreno_is_a650_family(adreno_gpu)) {
950 /* Set the GMEM VA range [0x100000:0x100000 + gpu->gmem - 1] */
951 gpu_write64(gpu, REG_A6XX_UCHE_GMEM_RANGE_MIN_LO,
952 REG_A6XX_UCHE_GMEM_RANGE_MIN_HI, 0x00100000);
953
954 gpu_write64(gpu, REG_A6XX_UCHE_GMEM_RANGE_MAX_LO,
955 REG_A6XX_UCHE_GMEM_RANGE_MAX_HI,
956 0x00100000 + adreno_gpu->gmem - 1);
957 }
958
959 gpu_write(gpu, REG_A6XX_UCHE_FILTER_CNTL, 0x804);
960 gpu_write(gpu, REG_A6XX_UCHE_CACHE_WAYS, 0x4);
961
962 if (adreno_is_a640_family(adreno_gpu) ||
963 adreno_is_a650_family(adreno_gpu))
964 gpu_write(gpu, REG_A6XX_CP_ROQ_THRESHOLDS_2, 0x02000140);
965 else
966 gpu_write(gpu, REG_A6XX_CP_ROQ_THRESHOLDS_2, 0x010000c0);
967 gpu_write(gpu, REG_A6XX_CP_ROQ_THRESHOLDS_1, 0x8040362c);
968
969 if (adreno_is_a660_family(adreno_gpu))
970 gpu_write(gpu, REG_A6XX_CP_LPAC_PROG_FIFO_SIZE, 0x00000020);
971
972 /* Setting the mem pool size */
973 gpu_write(gpu, REG_A6XX_CP_MEM_POOL_SIZE, 128);
974
975 /* Setting the primFifo thresholds default values,
976 * and vccCacheSkipDis=1 bit (0x200) for A640 and newer
977 */
978 if (adreno_is_a650(adreno_gpu) || adreno_is_a660(adreno_gpu))
979 gpu_write(gpu, REG_A6XX_PC_DBG_ECO_CNTL, 0x00300200);
980 else if (adreno_is_a640_family(adreno_gpu) || adreno_is_7c3(adreno_gpu))
981 gpu_write(gpu, REG_A6XX_PC_DBG_ECO_CNTL, 0x00200200);
982 else if (adreno_is_a650(adreno_gpu) || adreno_is_a660(adreno_gpu))
983 gpu_write(gpu, REG_A6XX_PC_DBG_ECO_CNTL, 0x00300200);
984 else
985 gpu_write(gpu, REG_A6XX_PC_DBG_ECO_CNTL, 0x00180000);
986
987 /* Set the AHB default slave response to "ERROR" */
988 gpu_write(gpu, REG_A6XX_CP_AHB_CNTL, 0x1);
989
990 /* Turn on performance counters */
991 gpu_write(gpu, REG_A6XX_RBBM_PERFCTR_CNTL, 0x1);
992
993 /* Select CP0 to always count cycles */
994 gpu_write(gpu, REG_A6XX_CP_PERFCTR_CP_SEL(0), PERF_CP_ALWAYS_COUNT);
995
996 a6xx_set_ubwc_config(gpu);
997
998 /* Enable fault detection */
999 gpu_write(gpu, REG_A6XX_RBBM_INTERFACE_HANG_INT_CNTL,
1000 (1 << 30) | 0x1fffff);
1001
1002 gpu_write(gpu, REG_A6XX_UCHE_CLIENT_PF, 1);
1003
1004 /* Set weights for bicubic filtering */
1005 if (adreno_is_a650_family(adreno_gpu)) {
1006 gpu_write(gpu, REG_A6XX_TPL1_BICUBIC_WEIGHTS_TABLE_0, 0);
1007 gpu_write(gpu, REG_A6XX_TPL1_BICUBIC_WEIGHTS_TABLE_1,
1008 0x3fe05ff4);
1009 gpu_write(gpu, REG_A6XX_TPL1_BICUBIC_WEIGHTS_TABLE_2,
1010 0x3fa0ebee);
1011 gpu_write(gpu, REG_A6XX_TPL1_BICUBIC_WEIGHTS_TABLE_3,
1012 0x3f5193ed);
1013 gpu_write(gpu, REG_A6XX_TPL1_BICUBIC_WEIGHTS_TABLE_4,
1014 0x3f0243f0);
1015 }
1016
1017 /* Protect registers from the CP */
1018 a6xx_set_cp_protect(gpu);
1019
1020 if (adreno_is_a660_family(adreno_gpu)) {
1021 gpu_write(gpu, REG_A6XX_CP_CHICKEN_DBG, 0x1);
1022 gpu_write(gpu, REG_A6XX_RBBM_GBIF_CLIENT_QOS_CNTL, 0x0);
1023 }
1024
1025 /* Set dualQ + disable afull for A660 GPU */
1026 if (adreno_is_a660(adreno_gpu))
1027 gpu_write(gpu, REG_A6XX_UCHE_CMDQ_CONFIG, 0x66906);
1028
1029 /* Enable expanded apriv for targets that support it */
1030 if (gpu->hw_apriv) {
1031 gpu_write(gpu, REG_A6XX_CP_APRIV_CNTL,
1032 (1 << 6) | (1 << 5) | (1 << 3) | (1 << 2) | (1 << 1));
1033 }
1034
1035 /* Enable interrupts */
1036 gpu_write(gpu, REG_A6XX_RBBM_INT_0_MASK, A6XX_INT_MASK);
1037
1038 ret = adreno_hw_init(gpu);
1039 if (ret)
1040 goto out;
1041
1042 ret = a6xx_ucode_init(gpu);
1043 if (ret)
1044 goto out;
1045
1046 /* Set the ringbuffer address */
1047 gpu_write64(gpu, REG_A6XX_CP_RB_BASE, REG_A6XX_CP_RB_BASE_HI,
1048 gpu->rb[0]->iova);
1049
1050 /* Targets that support extended APRIV can use the RPTR shadow from
1051 * hardware but all the other ones need to disable the feature. Targets
1052 * that support the WHERE_AM_I opcode can use that instead
1053 */
1054 if (adreno_gpu->base.hw_apriv)
1055 gpu_write(gpu, REG_A6XX_CP_RB_CNTL, MSM_GPU_RB_CNTL_DEFAULT);
1056 else
1057 gpu_write(gpu, REG_A6XX_CP_RB_CNTL,
1058 MSM_GPU_RB_CNTL_DEFAULT | AXXX_CP_RB_CNTL_NO_UPDATE);
1059
1060 /*
1061 * Expanded APRIV and targets that support WHERE_AM_I both need a
1062 * privileged buffer to store the RPTR shadow
1063 */
1064
1065 if (adreno_gpu->base.hw_apriv || a6xx_gpu->has_whereami) {
1066 if (!a6xx_gpu->shadow_bo) {
1067 a6xx_gpu->shadow = msm_gem_kernel_new(gpu->dev,
1068 sizeof(u32) * gpu->nr_rings,
1069 MSM_BO_WC | MSM_BO_MAP_PRIV,
1070 gpu->aspace, &a6xx_gpu->shadow_bo,
1071 &a6xx_gpu->shadow_iova);
1072
1073 if (IS_ERR(a6xx_gpu->shadow))
1074 return PTR_ERR(a6xx_gpu->shadow);
1075
1076 msm_gem_object_set_name(a6xx_gpu->shadow_bo, "shadow");
1077 }
1078
1079 gpu_write64(gpu, REG_A6XX_CP_RB_RPTR_ADDR_LO,
1080 REG_A6XX_CP_RB_RPTR_ADDR_HI,
1081 shadowptr(a6xx_gpu, gpu->rb[0]));
1082 }
1083
1084 /* Always come up on rb 0 */
1085 a6xx_gpu->cur_ring = gpu->rb[0];
1086
1087 gpu->cur_ctx_seqno = 0;
1088
1089 /* Enable the SQE_to start the CP engine */
1090 gpu_write(gpu, REG_A6XX_CP_SQE_CNTL, 1);
1091
1092 ret = a6xx_cp_init(gpu);
1093 if (ret)
1094 goto out;
1095
1096 /*
1097 * Try to load a zap shader into the secure world. If successful
1098 * we can use the CP to switch out of secure mode. If not then we
1099 * have no resource but to try to switch ourselves out manually. If we
1100 * guessed wrong then access to the RBBM_SECVID_TRUST_CNTL register will
1101 * be blocked and a permissions violation will soon follow.
1102 */
1103 ret = a6xx_zap_shader_init(gpu);
1104 if (!ret) {
1105 OUT_PKT7(gpu->rb[0], CP_SET_SECURE_MODE, 1);
1106 OUT_RING(gpu->rb[0], 0x00000000);
1107
1108 a6xx_flush(gpu, gpu->rb[0]);
1109 if (!a6xx_idle(gpu, gpu->rb[0]))
1110 return -EINVAL;
1111 } else if (ret == -ENODEV) {
1112 static bool first = true;
1113 if (first) {
1114 void __iomem *reg = ioremap(0x05060000, 0x1000);
> 1115 writeq(0x48000, reg); /* offset of cb0 from gpu's base */
1116 iounmap(reg);
1117 }
1118 /*
1119 * This device does not use zap shader (but print a warning
1120 * just in case someone got their dt wrong.. hopefully they
1121 * have a debug UART to realize the error of their ways...
1122 * if you mess this up you are about to crash horribly)
1123 */
1124 dev_warn_once(gpu->dev->dev,
1125 "Zap shader not enabled - using SECVID_TRUST_CNTL instead\n");
1126 gpu_write(gpu, REG_A6XX_RBBM_SECVID_TRUST_CNTL, 0x0);
1127 ret = 0;
1128 } else {
1129 return ret;
1130 }
1131
1132 out:
1133 /*
1134 * Tell the GMU that we are done touching the GPU and it can start power
1135 * management
1136 */
1137 a6xx_gmu_clear_oob(&a6xx_gpu->gmu, GMU_OOB_GPU_SET);
1138
1139 if (a6xx_gpu->gmu.legacy) {
1140 /* Take the GMU out of its special boot mode */
1141 a6xx_gmu_clear_oob(&a6xx_gpu->gmu, GMU_OOB_BOOT_SLUMBER);
1142 }
1143
1144 return ret;
1145 }
1146
--
0-DAY CI Kernel Test Service
https://01.org/lkp
Powered by blists - more mailing lists