__sync_fetch_and_add() doesn't work in bpf

Question

I use __sync_fetch_and_add(&histp->slots[slot], 1) in my bpf program, but it failed on verification, is there somebody know why? the clang and LLVM both are v10. Thanks for your help.

below is the error log:

        ```
        ; if (slot >= MAX_SLOTS)
        214: (bf) r2 = r1
        215: (67) r2 <<= 32
        216: (77) r2 >>= 32
        217: (b7) r3 = 31
        ; __sync_fetch_and_add(&histp->slots[slot], 1);
        218: (2d) if r3 > r2 goto pc+1
        
        from 218 to 220: R0=map_value(id=0,off=0,ks=8,vs=128,imm=0) R1=inv(id=0) R2_w=inv(id=0,umax_value=30,var_off=(0x0; 0x1f)) R3_w=inv31 R4=inv0 R6=inv(id=0,umin_value=1) R7=fp-32 R8=inv(id=0) R9=map_value(id=0,off=0,ks=8,vs=40,imm=0) R10=fp0 fp-8=mmmmmmmm fp-16=mmmmmmmm fp-24=mmmmmmmm fp-32=mmmmmmmm fp-40=mmmmmmmm
        ; __sync_fetch_and_add(&histp->slots[slot], 1);
      ```
      ```
        parent didn't have regs=100 stack=0 marks
        ```
         R0_w=inv(id=0) R6_rw=inv(id=0) R7_rw=ctx(id=0,off=0,imm=0) R8_rw=invP(id=0) R10=fp0 fp-8_w=00000000 fp-16_w=mmmmmmmm fp-24_rw=mmmmmmmm fp-32_rw=00000000
        parent didn't have regs=100 stack=0 marks
     ```
        regs=1 stack=0 before 3: (85) call bpf_ktime_get_ns#5
        224: (b7) r1 = 1
        **; __sync_fetch_and_add(&histp->slots[slot], 1);**
        225: (c3) lock *(u32 *)(r0 +0) += r1
         R0_w=map_value(id=0,off=0,ks=8,vs=128,umax_value=17179869180,var_off=(0x0; 0x3fffffffc),s32_max_value=2147483644,u32_max_value=-4) R1_w=inv1 R2_w=inv(id=0,umax_value=30,var_off=(0x0; 0x1f)) R3_w=inv31 R4=inv0 R6=inv(id=0,umin_value=1) R7=fp-32 R8=inv(id=0) R9=map_value(id=0,off=0,ks=8,vs=40,imm=0) R10=fp0 fp-8=mmmmmmmm fp-16=mmmmmmmm fp-24=mmmmmmmm fp-32=mmmmmmmm fp-40=mmmmmmmm
        R0 unbounded memory access, make sure to bounds check any such access
        processed 145 insns (limit 1000000) max_states_per_insn 0 total_states 11 peak_states 11 mark_read 6
        -- END PROG LOAD LOG --

bpf program code: the hists and hkey are defined as in the biolatency.

        histp = bpf_map_lookup_elem(&scsi_hists, &hkey);                 
        if (!histp) {                                                    
                bpf_map_update_elem(&scsi_hists, &hkey, &initial_hist, 0); 
                histp = bpf_map_lookup_elem(&scsi_hists, &hkey);         
                if (!histp)                                              
                        goto cleanup;                                    
        }                                                                                                               
        slot = log2l(delta);                                             
        if (slot >= MAX_SLOTS)                                           
                slot = MAX_SLOTS - 1;                                    
                                                                                                          
        __sync_fetch_and_add(&histp->slots[slot], 1);