tree: https://git.kernel.org/pub/scm/linux/kernel/git/ogabbay/linux.git habanalabs-next head: f3fad4c1da9034b716430407921cbd316d8c76dd commit: a260a3d5bfbee39ef056d45a6c1edd734b6fa375 [10/62] habanalabs: sync stream refactor functions config: s390-allyesconfig (attached as .config) compiler: s390-linux-gcc (GCC) 9.3.0 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://git.kernel.org/pub/scm/linux/kernel/git/ogabbay/linux.git/commit/... git remote add ogabbay https://git.kernel.org/pub/scm/linux/kernel/git/ogabbay/linux.git git fetch --no-tags ogabbay habanalabs-next git checkout a260a3d5bfbee39ef056d45a6c1edd734b6fa375 # save the attached .config to linux build tree COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-9.3.0 make.cross ARCH=s390 If you fix the issue, kindly add following tag as appropriate Reported-by: kernel test robot <lkp(a)intel.com&gt; All warnings (new ones prefixed by >>): drivers/misc/habanalabs/common/command_submission.c: In function 'cs_ioctl_signal_wait': 904 | struct hl_cs_counters_atomic *cntr; | ^~~~ vim +/cntr +904 drivers/misc/habanalabs/common/command_submission.c 894 895 static int cs_ioctl_signal_wait(struct hl_fpriv *hpriv, enum hl_cs_type cs_type, 896 void __user *chunks, u32 num_chunks, 897 u64 *cs_seq) 898 { 899 struct hl_device *hdev = hpriv->hdev; 900 struct hl_ctx *ctx = hpriv->ctx; 901 struct hl_cs_chunk *cs_chunk_array, *chunk; 902 struct hw_queue_properties *hw_queue_prop; 903 struct hl_fence *sig_fence = NULL; 905 struct hl_cs_compl *sig_waitcs_cmpl; 906 struct hl_cs *cs; 907 enum hl_queue_type q_type; 908 u32 size_to_copy, q_idx; 909 u64 signal_seq; 910 int rc; 911 912 *cs_seq = ULLONG_MAX; 913 cntr = &hdev->aggregated_cs_counters; 914 915 if (num_chunks > HL_MAX_JOBS_PER_CS) { 916 dev_err(hdev->dev, 917 "Number of chunks can NOT be larger than %d\n", 918 HL_MAX_JOBS_PER_CS); 919 rc = -EINVAL; 920 goto out; 921 } 922 923 cs_chunk_array = kmalloc_array(num_chunks, sizeof(*cs_chunk_array), 924 GFP_ATOMIC); 925 if (!cs_chunk_array) { 926 rc = -ENOMEM; 927 goto out; 928 } 929 930 size_to_copy = num_chunks * sizeof(struct hl_cs_chunk); 931 if (copy_from_user(cs_chunk_array, chunks, size_to_copy)) { 932 dev_err(hdev->dev, "Failed to copy cs chunk array from user\n"); 933 rc = -EFAULT; 934 goto free_cs_chunk_array; 935 } 936 937 /* currently it is guaranteed to have only one chunk */ 938 chunk = &cs_chunk_array[0]; 939 940 if (chunk->queue_index >= hdev->asic_prop.max_queues) { 941 dev_err(hdev->dev, "Queue index %d is invalid\n", 942 chunk->queue_index); 943 rc = -EINVAL; 944 goto free_cs_chunk_array; 945 } 946 947 q_idx = chunk->queue_index; 948 hw_queue_prop = &hdev->asic_prop.hw_queues_props[q_idx]; 949 q_type = hw_queue_prop->type; 950 951 if ((q_idx >= hdev->asic_prop.max_queues) || 952 (!hw_queue_prop->supports_sync_stream)) { 953 dev_err(hdev->dev, "Queue index %d is invalid\n", q_idx); 954 rc = -EINVAL; 955 goto free_cs_chunk_array; 956 } 957 958 if (cs_type == CS_TYPE_WAIT) { 959 rc = cs_ioctl_extract_signal_seq(hdev, chunk, &signal_seq); 960 if (rc) 961 goto free_cs_chunk_array; 962 963 sig_fence = hl_ctx_get_fence(ctx, signal_seq); 964 if (IS_ERR(sig_fence)) { 965 dev_err(hdev->dev, 966 "Failed to get signal CS with seq 0x%llx\n", 967 signal_seq); 968 rc = PTR_ERR(sig_fence); 969 goto free_cs_chunk_array; 970 } 971 972 if (!sig_fence) { 973 /* signal CS already finished */ 974 rc = 0; 975 goto free_cs_chunk_array; 976 } 977 978 sig_waitcs_cmpl = 979 container_of(sig_fence, struct hl_cs_compl, base_fence); 980 981 if (sig_waitcs_cmpl->type != CS_TYPE_SIGNAL) { 982 dev_err(hdev->dev, 983 "CS seq 0x%llx is not of a signal CS\n", 984 signal_seq); 985 hl_fence_put(sig_fence); 986 rc = -EINVAL; 987 goto free_cs_chunk_array; 988 } 989 990 if (completion_done(&sig_fence->completion)) { 991 /* signal CS already finished */ 992 hl_fence_put(sig_fence); 993 rc = 0; 994 goto free_cs_chunk_array; 995 } 996 } 997 998 /* increment refcnt for context */ 999 hl_ctx_get(hdev, ctx); 1000 1001 rc = allocate_cs(hdev, ctx, cs_type, &cs); 1002 if (rc) { 1003 if (cs_type == CS_TYPE_WAIT) 1004 hl_fence_put(sig_fence); 1005 hl_ctx_put(ctx); 1006 goto free_cs_chunk_array; 1007 } 1008 1009 /* 1010 * Save the signal CS fence for later initialization right before 1011 * hanging the wait CS on the queue. 1012 */ 1013 if (cs_type == CS_TYPE_WAIT) 1014 cs->signal_fence = sig_fence; 1015 1016 hl_debugfs_add_cs(cs); 1017 1018 *cs_seq = cs->sequence; 1019 1020 if (cs_type == CS_TYPE_WAIT || cs_type == CS_TYPE_SIGNAL) 1021 rc = cs_ioctl_signal_wait_create_jobs(hdev, ctx, cs, q_type, 1022 q_idx); 1023 1024 if (rc) 1025 goto put_cs; 1026 1027 1028 /* increment refcount as for external queues we get completion */ 1029 cs_get(cs); 1030 1031 rc = hl_hw_queue_schedule_cs(cs); 1032 if (rc) { 1033 if (rc != -EAGAIN) 1034 dev_err(hdev->dev, 1035 "Failed to submit CS %d.%llu to H/W queues, error %d\n", 1036 ctx->asid, cs->sequence, rc); 1037 goto free_cs_object; 1038 } 1039 1040 rc = HL_CS_STATUS_SUCCESS; 1041 goto put_cs; 1042 1043 free_cs_object: 1044 cs_rollback(hdev, cs); 1045 *cs_seq = ULLONG_MAX; 1046 /* The path below is both for good and erroneous exits */ 1047 put_cs: 1048 /* We finished with the CS in this function, so put the ref */ 1049 cs_put(cs); 1050 free_cs_chunk_array: 1051 kfree(cs_chunk_array); 1052 out: 1053 return rc; 1054 } 1055 --- 0-DAY CI Kernel Test Service, Intel Corporation https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org