Implement bpf_probe_read_kernel_str helper, Allow i8* to i8 ArrayType conversion

BCC-Examples/hello_perf_output.py

@@ -1,4 +1,4 @@
-from pythonbpf import bpf, map, struct, section, bpfglobal, compile
+from pythonbpf import bpf, map, struct, section, bpfglobal, BPF, trace_pipe
 from pythonbpf.helper import ktime, pid, comm
 from pythonbpf.maps import PerfEventArray
 
@@ -23,10 +23,9 @@ def events() -> PerfEventArray:
 @section("tracepoint/syscalls/sys_enter_clone")
 def hello(ctx: c_void_p) -> c_int64:
     dataobj = data_t()
-    strobj = "hellohellohello"
     dataobj.pid, dataobj.ts = pid(), ktime()
     comm(dataobj.comm)
-    print(f"clone called at {dataobj.ts} by pid {dataobj.pid}, comm {strobj}")
+    print(f"clone called at {dataobj.ts} by pid {dataobj.pid}, comm {dataobj.comm}")
     events.output(dataobj)
     return 0  # type: ignore [return-value]
 
@@ -37,4 +36,8 @@ def LICENSE() -> str:
     return "GPL"
 
 
-compile()
+# compile
+BPF().load_and_attach()
+
+print("Tracing clone()... Ctrl-C to end")
+trace_pipe()

pythonbpf/allocation_pass.py

@@ -72,6 +72,14 @@ def handle_assign_allocation(builder, stmt, local_sym_tab, structs_sym_tab):
             _allocate_for_constant(builder, var_name, rval, local_sym_tab)
         elif isinstance(rval, ast.BinOp):
             _allocate_for_binop(builder, var_name, local_sym_tab)
+        elif isinstance(rval, ast.Name):
+            # Variable-to-variable assignment (b = a)
+            _allocate_for_name(builder, var_name, rval, local_sym_tab)
+        elif isinstance(rval, ast.Attribute):
+            # Struct field-to-variable assignment (a = dat.fld)
+            _allocate_for_attribute(
+                builder, var_name, rval, local_sym_tab, structs_sym_tab
+            )
         else:
             logger.warning(
                 f"Unsupported assignment value type for {var_name}: {type(rval).__name__}"
@@ -192,3 +200,88 @@ def allocate_temp_pool(builder, max_temps, local_sym_tab):
         temp_var = builder.alloca(ir.IntType(64), name=temp_name)
         temp_var.align = 8
         local_sym_tab[temp_name] = LocalSymbol(temp_var, ir.IntType(64))
+
+
+def _allocate_for_name(builder, var_name, rval, local_sym_tab):
+    """Allocate memory for variable-to-variable assignment (b = a)."""
+    source_var = rval.id
+
+    if source_var not in local_sym_tab:
+        logger.error(f"Source variable '{source_var}' not found in symbol table")
+        return
+
+    # Get type and metadata from source variable
+    source_symbol = local_sym_tab[source_var]
+
+    # Allocate with same type and alignment
+    var = _allocate_with_type(builder, var_name, source_symbol.ir_type)
+    local_sym_tab[var_name] = LocalSymbol(
+        var, source_symbol.ir_type, source_symbol.metadata
+    )
+
+    logger.info(
+        f"Pre-allocated {var_name} from {source_var} with type {source_symbol.ir_type}"
+    )
+
+
+def _allocate_for_attribute(builder, var_name, rval, local_sym_tab, structs_sym_tab):
+    """Allocate memory for struct field-to-variable assignment (a = dat.fld)."""
+    if not isinstance(rval.value, ast.Name):
+        logger.warning(f"Complex attribute access not supported for {var_name}")
+        return
+
+    struct_var = rval.value.id
+    field_name = rval.attr
+
+    # Validate struct and field
+    if struct_var not in local_sym_tab:
+        logger.error(f"Struct variable '{struct_var}' not found")
+        return
+
+    struct_type = local_sym_tab[struct_var].metadata
+    if not struct_type or struct_type not in structs_sym_tab:
+        logger.error(f"Struct type '{struct_type}' not found")
+        return
+
+    struct_info = structs_sym_tab[struct_type]
+    if field_name not in struct_info.fields:
+        logger.error(f"Field '{field_name}' not found in struct '{struct_type}'")
+        return
+
+    # Get field type
+    field_type = struct_info.field_type(field_name)
+
+    # Special case: char array -> allocate as i8* pointer instead
+    if (
+        isinstance(field_type, ir.ArrayType)
+        and isinstance(field_type.element, ir.IntType)
+        and field_type.element.width == 8
+    ):
+        alloc_type = ir.PointerType(ir.IntType(8))
+        logger.info(f"Allocating {var_name} as i8* (pointer to char array)")
+    else:
+        alloc_type = field_type
+
+    var = _allocate_with_type(builder, var_name, alloc_type)
+    local_sym_tab[var_name] = LocalSymbol(var, alloc_type)
+
+    logger.info(
+        f"Pre-allocated {var_name} from {struct_var}.{field_name} with type {alloc_type}"
+    )
+
+
+def _allocate_with_type(builder, var_name, ir_type):
+    """Allocate variable with appropriate alignment for type."""
+    var = builder.alloca(ir_type, name=var_name)
+    var.align = _get_alignment(ir_type)
+    return var
+
+
+def _get_alignment(ir_type):
+    """Get appropriate alignment for IR type."""
+    if isinstance(ir_type, ir.IntType):
+        return ir_type.width // 8
+    elif isinstance(ir_type, ir.ArrayType) and isinstance(ir_type.element, ir.IntType):
+        return ir_type.element.width // 8
+    else:
+        return 8  # Default: pointer size

pythonbpf/assign_pass.py

@@ -2,6 +2,7 @@ import ast
 import logging
 from llvmlite import ir
 from pythonbpf.expr import eval_expr
+from pythonbpf.helper import emit_probe_read_kernel_str_call
 
 logger = logging.getLogger(__name__)
 
@@ -27,27 +28,82 @@ def handle_struct_field_assignment(
 
     # Get field pointer and evaluate value
     field_ptr = struct_info.gep(builder, local_sym_tab[var_name].var, field_name)
-    val = eval_expr(
+    field_type = struct_info.field_type(field_name)
+    val_result = eval_expr(
         func, module, builder, rval, local_sym_tab, map_sym_tab, structs_sym_tab
     )
 
-    if val is None:
+    if val_result is None:
         logger.error(f"Failed to evaluate value for {var_name}.{field_name}")
         return
 
-    # TODO: Handle string assignment to char array (not a priority)
+    val, val_type = val_result
-    field_type = struct_info.field_type(field_name)
+
-    if isinstance(field_type, ir.ArrayType) and val[1] == ir.PointerType(ir.IntType(8)):
+    # Special case: i8* string to [N x i8] char array
-        logger.warning(
+    if _is_char_array(field_type) and _is_i8_ptr(val_type):
-            f"String to char array assignment not implemented for {var_name}.{field_name}"
+        _copy_string_to_char_array(
+            func,
+            module,
+            builder,
+            val,
+            field_ptr,
+            field_type,
+            local_sym_tab,
+            map_sym_tab,
+            structs_sym_tab,
         )
+        logger.info(f"Copied string to char array {var_name}.{field_name}")
         return
 
-    # Store the value
+    # Regular assignment
-    builder.store(val[0], field_ptr)
+    builder.store(val, field_ptr)
     logger.info(f"Assigned to struct field {var_name}.{field_name}")
 
 
+def _copy_string_to_char_array(
+    func,
+    module,
+    builder,
+    src_ptr,
+    dst_ptr,
+    array_type,
+    local_sym_tab,
+    map_sym_tab,
+    struct_sym_tab,
+):
+    """Copy string (i8*) to char array ([N x i8]) using bpf_probe_read_kernel_str"""
+
+    array_size = array_type.count
+
+    # Get pointer to first element: [N x i8]* -> i8*
+    dst_i8_ptr = builder.gep(
+        dst_ptr,
+        [ir.Constant(ir.IntType(32), 0), ir.Constant(ir.IntType(32), 0)],
+        inbounds=True,
+    )
+
+    # Use the shared emitter function
+    emit_probe_read_kernel_str_call(builder, dst_i8_ptr, array_size, src_ptr)
+
+
+def _is_char_array(ir_type):
+    """Check if type is [N x i8]."""
+    return (
+        isinstance(ir_type, ir.ArrayType)
+        and isinstance(ir_type.element, ir.IntType)
+        and ir_type.element.width == 8
+    )
+
+
+def _is_i8_ptr(ir_type):
+    """Check if type is i8*."""
+    return (
+        isinstance(ir_type, ir.PointerType)
+        and isinstance(ir_type.pointee, ir.IntType)
+        and ir_type.pointee.width == 8
+    )
+
+
 def handle_variable_assignment(
     func, module, builder, var_name, rval, local_sym_tab, map_sym_tab, structs_sym_tab
 ):
@@ -71,6 +127,17 @@ def handle_variable_assignment(
             logger.info(f"Initialized struct {struct_name} for variable {var_name}")
             return True
 
+    # Special case: struct field char array -> pointer
+    # Handle this before eval_expr to get the pointer, not the value
+    if isinstance(rval, ast.Attribute) and isinstance(rval.value, ast.Name):
+        converted_val = _try_convert_char_array_to_ptr(
+            rval, var_type, builder, local_sym_tab, structs_sym_tab
+        )
+        if converted_val is not None:
+            builder.store(converted_val, var_ptr)
+            logger.info(f"Assigned char array pointer to {var_name}")
+            return True
+
     val_result = eval_expr(
         func, module, builder, rval, local_sym_tab, map_sym_tab, structs_sym_tab
     )
@@ -106,3 +173,52 @@ def handle_variable_assignment(
     builder.store(val, var_ptr)
     logger.info(f"Assigned value to variable {var_name}")
     return True
+
+
+def _try_convert_char_array_to_ptr(
+    rval, var_type, builder, local_sym_tab, structs_sym_tab
+):
+    """Try to convert char array field to i8* pointer"""
+    # Only convert if target is i8*
+    if not (
+        isinstance(var_type, ir.PointerType)
+        and isinstance(var_type.pointee, ir.IntType)
+        and var_type.pointee.width == 8
+    ):
+        return None
+
+    struct_var = rval.value.id
+    field_name = rval.attr
+
+    # Validate struct
+    if struct_var not in local_sym_tab:
+        return None
+
+    struct_type = local_sym_tab[struct_var].metadata
+    if not struct_type or struct_type not in structs_sym_tab:
+        return None
+
+    struct_info = structs_sym_tab[struct_type]
+    if field_name not in struct_info.fields:
+        return None
+
+    field_type = struct_info.field_type(field_name)
+
+    # Check if it's a char array
+    if not (
+        isinstance(field_type, ir.ArrayType)
+        and isinstance(field_type.element, ir.IntType)
+        and field_type.element.width == 8
+    ):
+        return None
+
+    # Get pointer to struct field
+    struct_ptr = local_sym_tab[struct_var].var
+    field_ptr = struct_info.gep(builder, struct_ptr, field_name)
+
+    # GEP to first element: [N x i8]* -> i8*
+    return builder.gep(
+        field_ptr,
+        [ir.Constant(ir.IntType(32), 0), ir.Constant(ir.IntType(32), 0)],
+        inbounds=True,
+    )

pythonbpf/helper/__init__.py

@@ -1,7 +1,7 @@
 from .helper_registry import HelperHandlerRegistry
 from .helper_utils import reset_scratch_pool
-from .bpf_helper_handler import handle_helper_call
+from .bpf_helper_handler import handle_helper_call, emit_probe_read_kernel_str_call
-from .helpers import ktime, pid, deref, comm, XDP_DROP, XDP_PASS
+from .helpers import ktime, pid, deref, comm, probe_read_str, XDP_DROP, XDP_PASS
 
 
 # Register the helper handler with expr module
@@ -59,10 +59,12 @@ __all__ = [
     "HelperHandlerRegistry",
     "reset_scratch_pool",
     "handle_helper_call",
+    "emit_probe_read_kernel_str_call",
     "ktime",
     "pid",
     "deref",
     "comm",
+    "probe_read_str",
     "XDP_DROP",
     "XDP_PASS",
 ]

pythonbpf/helper/bpf_helper_handler.py

@@ -8,6 +8,8 @@ from .helper_utils import (
     get_flags_val,
     get_data_ptr_and_size,
     get_buffer_ptr_and_size,
+    get_char_array_ptr_and_size,
+    get_ptr_from_arg,
 )
 from .printk_formatter import simple_string_print, handle_fstring_print
 
@@ -26,6 +28,7 @@ class BPFHelperID(Enum):
     BPF_GET_CURRENT_PID_TGID = 14
     BPF_GET_CURRENT_COMM = 16
     BPF_PERF_EVENT_OUTPUT = 25
+    BPF_PROBE_READ_KERNEL_STR = 115
 
 
 @HelperHandlerRegistry.register("ktime")
@@ -368,6 +371,68 @@ def bpf_perf_event_output_handler(
     return result, None
 
 
+def emit_probe_read_kernel_str_call(builder, dst_ptr, dst_size, src_ptr):
+    """Emit LLVM IR call to bpf_probe_read_kernel_str"""
+
+    fn_type = ir.FunctionType(
+        ir.IntType(64),
+        [ir.PointerType(), ir.IntType(32), ir.PointerType()],
+        var_arg=False,
+    )
+    fn_ptr = builder.inttoptr(
+        ir.Constant(ir.IntType(64), BPFHelperID.BPF_PROBE_READ_KERNEL_STR.value),
+        ir.PointerType(fn_type),
+    )
+
+    result = builder.call(
+        fn_ptr,
+        [
+            builder.bitcast(dst_ptr, ir.PointerType()),
+            ir.Constant(ir.IntType(32), dst_size),
+            builder.bitcast(src_ptr, ir.PointerType()),
+        ],
+        tail=False,
+    )
+
+    logger.info(f"Emitted bpf_probe_read_kernel_str (size={dst_size})")
+    return result
+
+
+@HelperHandlerRegistry.register("probe_read_str")
+def bpf_probe_read_kernel_str_emitter(
+    call,
+    map_ptr,
+    module,
+    builder,
+    func,
+    local_sym_tab=None,
+    struct_sym_tab=None,
+    map_sym_tab=None,
+):
+    """Emit LLVM IR for bpf_probe_read_kernel_str helper."""
+
+    if len(call.args) != 2:
+        raise ValueError(
+            f"probe_read_str expects 2 args (dst, src), got {len(call.args)}"
+        )
+
+    # Get destination buffer (char array -> i8*)
+    dst_ptr, dst_size = get_char_array_ptr_and_size(
+        call.args[0], builder, local_sym_tab, struct_sym_tab
+    )
+
+    # Get source pointer (evaluate expression)
+    src_ptr, src_type = get_ptr_from_arg(
+        call.args[1], func, module, builder, local_sym_tab, map_sym_tab, struct_sym_tab
+    )
+
+    # Emit the helper call
+    result = emit_probe_read_kernel_str_call(builder, dst_ptr, dst_size, src_ptr)
+
+    logger.info(f"Emitted bpf_probe_read_kernel_str (size={dst_size})")
+    return result, ir.IntType(64)
+
+
 def handle_helper_call(
     call,
     module,

pythonbpf/helper/helper_utils.py

@@ -4,6 +4,7 @@ import logging
 from llvmlite import ir
 from pythonbpf.expr import (
     get_operand_value,
+    eval_expr,
 )
 
 logger = logging.getLogger(__name__)
@@ -190,3 +191,86 @@ def get_buffer_ptr_and_size(buf_arg, builder, local_sym_tab, struct_sym_tab):
         raise ValueError(
             "comm expects either a struct field (obj.field) or variable name"
         )
+
+
+def get_char_array_ptr_and_size(buf_arg, builder, local_sym_tab, struct_sym_tab):
+    """Get pointer to char array and its size."""
+
+    # Struct field: obj.field
+    if isinstance(buf_arg, ast.Attribute) and isinstance(buf_arg.value, ast.Name):
+        var_name = buf_arg.value.id
+        field_name = buf_arg.attr
+
+        if not (local_sym_tab and var_name in local_sym_tab):
+            raise ValueError(f"Variable '{var_name}' not found")
+
+        struct_type = local_sym_tab[var_name].metadata
+        if not (struct_sym_tab and struct_type in struct_sym_tab):
+            raise ValueError(f"Struct type '{struct_type}' not found")
+
+        struct_info = struct_sym_tab[struct_type]
+        if field_name not in struct_info.fields:
+            raise ValueError(f"Field '{field_name}' not found")
+
+        field_type = struct_info.field_type(field_name)
+        if not _is_char_array(field_type):
+            raise ValueError("Expected char array field")
+
+        struct_ptr = local_sym_tab[var_name].var
+        field_ptr = struct_info.gep(builder, struct_ptr, field_name)
+
+        # GEP to first element: [N x i8]* -> i8*
+        buf_ptr = builder.gep(
+            field_ptr,
+            [ir.Constant(ir.IntType(32), 0), ir.Constant(ir.IntType(32), 0)],
+            inbounds=True,
+        )
+        return buf_ptr, field_type.count
+
+    elif isinstance(buf_arg, ast.Name):
+        # NOTE: We shouldn't be doing this as we can't get size info
+        var_name = buf_arg.id
+        if not (local_sym_tab and var_name in local_sym_tab):
+            raise ValueError(f"Variable '{var_name}' not found")
+
+        var_ptr = local_sym_tab[var_name].var
+        var_type = local_sym_tab[var_name].ir_type
+
+        if not isinstance(var_type, ir.PointerType) and not isinstance(
+            var_type.pointee, ir.IntType(8)
+        ):
+            raise ValueError("Expected str ptr variable")
+
+        return var_ptr, 256  # Size unknown for str ptr, using 256 as default
+
+    else:
+        raise ValueError("Expected struct field or variable name")
+
+
+def _is_char_array(ir_type):
+    """Check if IR type is [N x i8]."""
+    return (
+        isinstance(ir_type, ir.ArrayType)
+        and isinstance(ir_type.element, ir.IntType)
+        and ir_type.element.width == 8
+    )
+
+
+def get_ptr_from_arg(
+    arg, func, module, builder, local_sym_tab, map_sym_tab, struct_sym_tab
+):
+    """Evaluate argument and return pointer value"""
+
+    result = eval_expr(
+        func, module, builder, arg, local_sym_tab, map_sym_tab, struct_sym_tab
+    )
+
+    if not result:
+        raise ValueError("Failed to evaluate argument")
+
+    val, val_type = result
+
+    if not isinstance(val_type, ir.PointerType):
+        raise ValueError(f"Expected pointer type, got {val_type}")
+
+    return val, val_type

pythonbpf/helper/helpers.py

@@ -22,6 +22,11 @@ def comm(buf):
     return ctypes.c_int64(0)
 
 
+def probe_read_str(dst, src):
+    """Safely read a null-terminated string from kernel memory"""
+    return ctypes.c_int64(0)
+
+
 XDP_ABORTED = ctypes.c_int64(0)
 XDP_DROP = ctypes.c_int64(1)
 XDP_PASS = ctypes.c_int64(2)

pythonbpf/helper/printk_formatter.py

@@ -173,6 +173,15 @@ def _populate_fval(ftype, node, fmt_parts, exprs):
             raise NotImplementedError(
                 f"Unsupported pointer target type in f-string: {target}"
             )
+    elif isinstance(ftype, ir.ArrayType):
+        if isinstance(ftype.element, ir.IntType) and ftype.element.width == 8:
+            # Char array
+            fmt_parts.append("%s")
+            exprs.append(node)
+        else:
+            raise NotImplementedError(
+                f"Unsupported array element type in f-string: {ftype.element}"
+            )
     else:
         raise NotImplementedError(f"Unsupported field type in f-string: {ftype}")
 
@@ -197,44 +206,100 @@ def _create_format_string_global(fmt_str, func, module, builder):
 
 def _prepare_expr_args(expr, func, module, builder, local_sym_tab, struct_sym_tab):
     """Evaluate and prepare an expression to use as an arg for bpf_printk."""
-    val, _ = eval_expr(
+
-        func,
+    # Special case: struct field char array needs pointer to first element
-        module,
+    char_array_ptr = _get_struct_char_array_ptr(
-        builder,
+        expr, builder, local_sym_tab, struct_sym_tab
-        expr,
+    )
-        local_sym_tab,
+    if char_array_ptr:
-        None,
+        return char_array_ptr
-        struct_sym_tab,
+
+    # Regular expression evaluation
+    val, _ = eval_expr(func, module, builder, expr, local_sym_tab, None, struct_sym_tab)
+
+    if not val:
+        logger.warning("Failed to evaluate expression for bpf_printk, defaulting to 0")
+        return ir.Constant(ir.IntType(64), 0)
+
+    # Convert value to bpf_printk compatible type
+    if isinstance(val.type, ir.PointerType):
+        return _handle_pointer_arg(val, func, builder)
+    elif isinstance(val.type, ir.IntType):
+        return _handle_int_arg(val, builder)
+    else:
+        logger.warning(f"Unsupported type {val.type} in bpf_printk, defaulting to 0")
+        return ir.Constant(ir.IntType(64), 0)
+
+
+def _get_struct_char_array_ptr(expr, builder, local_sym_tab, struct_sym_tab):
+    """Get pointer to first element of char array in struct field, or None."""
+    if not (isinstance(expr, ast.Attribute) and isinstance(expr.value, ast.Name)):
+        return None
+
+    var_name = expr.value.id
+    field_name = expr.attr
+
+    # Check if it's a valid struct field
+    if not (
+        local_sym_tab
+        and var_name in local_sym_tab
+        and struct_sym_tab
+        and local_sym_tab[var_name].metadata in struct_sym_tab
+    ):
+        return None
+
+    struct_type = local_sym_tab[var_name].metadata
+    struct_info = struct_sym_tab[struct_type]
+
+    if field_name not in struct_info.fields:
+        return None
+
+    field_type = struct_info.field_type(field_name)
+
+    # Check if it's a char array
+    is_char_array = (
+        isinstance(field_type, ir.ArrayType)
+        and isinstance(field_type.element, ir.IntType)
+        and field_type.element.width == 8
     )
 
-    if val:
+    if not is_char_array:
-        if isinstance(val.type, ir.PointerType):
+        return None
-            target, depth = get_base_type_and_depth(val.type)
-            if isinstance(target, ir.IntType):
-                if target.width >= 32:
-                    val = deref_to_depth(func, builder, val, depth)
-                    val = builder.sext(val, ir.IntType(64))
-                elif target.width == 8 and depth == 1:
-                    # NOTE: i8* is string, no need to deref
-                    pass
 
-            else:
+    # Get field pointer and GEP to first element: [N x i8]* -> i8*
-                logger.warning(
+    struct_ptr = local_sym_tab[var_name].var
-                    "Only int and ptr supported in bpf_printk args. Others default to 0."
+    field_ptr = struct_info.gep(builder, struct_ptr, field_name)
-                )
+
-                val = ir.Constant(ir.IntType(64), 0)
+    return builder.gep(
-        elif isinstance(val.type, ir.IntType):
+        field_ptr,
-            if val.type.width < 64:
+        [ir.Constant(ir.IntType(32), 0), ir.Constant(ir.IntType(32), 0)],
-                val = builder.sext(val, ir.IntType(64))
+        inbounds=True,
-        else:
+    )
-            logger.warning(
+
-                "Only int and ptr supported in bpf_printk args. Others default to 0."
+
-            )
+def _handle_pointer_arg(val, func, builder):
-            val = ir.Constant(ir.IntType(64), 0)
+    """Convert pointer type for bpf_printk."""
-        return val
+    target, depth = get_base_type_and_depth(val.type)
-    else:
+
-        logger.warning(
+    if not isinstance(target, ir.IntType):
-            "Failed to evaluate expression for bpf_printk argument. "
+        logger.warning("Only int pointers supported in bpf_printk, defaulting to 0")
-            "It will be converted to 0."
-        )
         return ir.Constant(ir.IntType(64), 0)
+
+    # i8* is string - use as-is
+    if target.width == 8 and depth == 1:
+        return val
+
+    # Integer pointers: dereference and sign-extend to i64
+    if target.width >= 32:
+        val = deref_to_depth(func, builder, val, depth)
+        return builder.sext(val, ir.IntType(64))
+
+    logger.warning("Unsupported pointer width in bpf_printk, defaulting to 0")
+    return ir.Constant(ir.IntType(64), 0)
+
+
+def _handle_int_arg(val, builder):
+    """Convert integer type for bpf_printk (sign-extend to i64)."""
+    if val.type.width < 64:
+        return builder.sext(val, ir.IntType(64))
+    return val