Using SCODB on SVR5

Function entry tracing

The entry of functions can be traced using the trace function entry (tfe) command:

   tfe [depth]	/* Trace function entry */

Function entry tracing is used to determine flow of control in the kernel, and is implemented using the instruction trace capability of the 80386.

At each function entry, a line is printed showing the function called, and execution is suspended until the user gives an input character. If the user enters a quit character, the debugger prompt will appear. Otherwise, execution proceeds until the next function entry.

The following example illustrates setting a breakpoint at a function and then using tfe to trace function calls from that point:

   debug0:1> bp open
   debug0:2> q
   $ cat /etc/passwd
   BREAKPOINT:
          D007E053 open+3               open
   debug0:3> tfe
   >-- open() <Space>
   >--- copen(1, 7FFFFE94) <Space>
   >---- ufalloc(0) <Space>
   >---- namei(D0011034, 0) <Space>
   >----- upath(D0011034, 0, E0001148) <Space>
   >------ rcopyfault+7C() <Space>
   ...

Each line in the output is printed in the following format:

   >dashes function name ( arguments )

The current function is always printed first before tracing function entry. The ``>'' is printed to show that the line is output of tfe.

A number of dashes are printed, to show the depth of the function call in the stack. In the above example, copen( ) is at depth 3 in the stack; it was called by the open( ) function at depth 2, which was called by the systrap( ) function at depth 1, which was called by the sys_call( ) function at depth -. Any function that copen( ) calls would be a depth 4 function, and so would have four dashes, and so on.

Using the optional depth argument to tfe causes tfe to print only those functions of a level less than or equal to (current level + depth).

Note that execution is greatly slowed while using tfe, and that no breakpoints are in effect while using tfe.