Properties list for Xen

Some functions and macros are found to have properties relevant to the Xen codebase. For this reason, the file docs/properties.json contains all the needed properties.

Here is an example of the properties.json file:

{
   "version": "1.0",
   "content": [
      {
         "description": ""
         "type": "function",       // required
         "value:": "^printk*.$",   // required
         "properties":{
            "pointee_write": "1..2=never",
            "pointee_read": "",
            "taken": ""
            "attribute": ""
         }
      }
   ]
}

Here is an explanation of the fields inside an object of the "content" array:

• description: a brief description of why the properties apply
• type: this is the kind of the element called: it may be either macro or function
• value: must be a regex, starting with ^ and ending with $ and matching function fully qualified name or macro name.

• properties: a list of properties applied to said function. Possible values are:

  • pointee_write: indicate the write use for call arguments that correspond to parameters whose pointee types are non-const
  • pointee_read: indicate the read use for call arguments that correspond to parameters whose pointee types are non-const
  • taken: indicates that the specified address arguments may be stored in objects that persist after the function has ceased to exist (excluding the returned value); address arguments not listed are never taken
  • attribute: attributes a function may have. Possible values are pure, const and noeffect.

  pointee_read and pointee_write use a specific kind of argument, structured as pointee_arg=rw:

  • pointee_arg: argument index for callee. Index 0 refers to the return value, the indices of the arguments start from 1. It can be either a single value or a range.

  • rw: a value that's either always, maybe or never

    • always: for pointee_read: argument pointee is expected to be fully read in the function body, for pointee_write: argument pointee is fully initialized at function exit
    • maybe: for pointee_read: argument pointee may be expected to be read in the function body, for pointee_write: argument pointee may be written by function body
    • never: for pointee_read: argument pointee is not expected to be read in the function body, for pointee_write: argument pointee is never written by function body