Hi,

This is the 7th version of the patch.

compare with the 6th version, the difference are:

updates per Siddhesh's comments:
1. update the error messages in "handle_counted_by_attribute"
   then update the testing case accordingly;
2. update the error messages in "verify_counted_by_attribute"
   then update the testing case accordingly;
3. update the documentation of "counted_by" in extend.texi
4. for the 3rd argument of ACCESS_WITH_SIZE, change it as following:
+   3rd argument CLASS_OF_SIZE: The size referenced by the REF_TO_SIZE 
represents
+     0: the number of bytes;
+     1: the number of the elements of the object type;

Update all other places accordingly.
5. update the comments of the routine "access_with_size_object_size"
   bail out if (object_size_type & OST_DYNAMIC) == 0 for this routine.
   change the variable name of "type_of_size" to "class_of_size" for 
   consistence.
6. add one more testing case for bound sanitizer to handle the case when
   counted-by field is zero value.


It based on the following original proposal:

https://gcc.gnu.org/pipermail/gcc-patches/2023-November/635884.html
Represent the missing dependence for the "counted_by" attribute and its 
consumers

******The summary of the proposal is:

* Add a new internal function ".ACCESS_WITH_SIZE" to carry the size information 
for every reference to a FAM field;
* In C FE, Replace every reference to a FAM field whose TYPE has the 
"counted_by" attribute with the new internal function ".ACCESS_WITH_SIZE";
* In every consumer of the size information, for example, BDOS or array bound 
sanitizer, query the size information or ACCESS_MODE information from the new 
internal function;
* When expansing to RTL, replace the internal function with the actual 
reference to the FAM field;
* Some adjustment to ipa alias analysis, and other SSA passes to mitigate the 
impact to the optimizer and code generation.


******The new internal function

  .ACCESS_WITH_SIZE (REF_TO_OBJ, REF_TO_SIZE, CLASS_OF_SIZE, TYPE_OF_SIZE, 
ACCESS_MODE, TYPE_OF_REF)

INTERNAL_FN (ACCESS_WITH_SIZE, ECF_LEAF | ECF_NOTHROW, NULL)

which returns the "REF_TO_OBJ" same as the 1st argument;

Both the return type and the type of the first argument of this function have 
been converted from the incomplete array type to the corresponding pointer type.

The call to .ACCESS_WITH_SIZE is wrapped with an INDIRECT_REF, whose type is 
the original imcomplete array type.

Please see the following link for why:
https://gcc.gnu.org/pipermail/gcc-patches/2023-November/638793.html
https://gcc.gnu.org/pipermail/gcc-patches/2023-December/639605.html

1st argument "REF_TO_OBJ": The reference to the object;
2nd argument "REF_TO_SIZE": The reference to the size of the object,
3rd argument "CLASS_OF_SIZE": The size referenced by the REF_TO_SIZE represents
   0: the number of bytes;
   1: the number of the elements of the object type;
4th argument "TYPE_OF_SIZE": A constant 0 with the TYPE of the object
  refed by REF_TO_SIZE
5th argument "ACCESS_MODE":
  -1: Unknown access semantics
   0: none
   1: read_only
   2: write_only
   3: read_write
6th argument "TYPE_OF_REF": A constant 0 with the pointer TYPE to
  to the original flexible array type.

****** The Patch sets included:

1. Provide counted_by attribute to flexible array member field;
      which includes:
      * "counted_by" attribute documentation;
      * C FE handling of the new attribute;
        syntax checking, error reporting;
      * testing cases;

2. Convert "counted_by" attribute to/from .ACCESS_WITH_SIZE.
      which includes:
      * The definition of the new internal function .ACCESS_WITH_SIZE in 
internal-fn.def.
      * C FE converts every reference to a FAM with "counted_by" attribute to a 
call to the internal function .ACCESS_WITH_SIZE.
        (build_component_ref in c_typeck.cc)
        This includes the case when the object is statically allocated and 
initialized.
        In order to make this working, we should update 
initializer_constant_valid_p_1 and output_constant in varasm.cc to include 
calls to .ACCESS_WITH_SIZE.

        However, for the reference inside "offsetof", ignore the "counted_by" 
attribute since it's not useful at all. (c_parser_postfix_expression in 
c/c-parser.cc)
        In addtion to "offsetof", for the reference inside operator "typeof" and
  "alignof", we ignore counted_by attribute too.
        When building ADDR_EXPR for the .ACCESS_WITH_SIZE in C FE,
  replace the call with its first argument.

      * Convert every call to .ACCESS_WITH_SIZE to its first argument.
        (expand_ACCESS_WITH_SIZE in internal-fn.cc)
      * adjust alias analysis to exclude the new internal from clobbering 
anything.
        (ref_maybe_used_by_call_p_1 and call_may_clobber_ref_p_1 in 
tree-ssa-alias.cc)
      * adjust dead code elimination to eliminate the call to .ACCESS_WITH_SIZE 
when
        it's LHS is eliminated as dead code.
        (eliminate_unnecessary_stmts in tree-ssa-dce.cc)
      * Provide the utility routines to check the call is .ACCESS_WITH_SIZE and
        get the reference from the call to .ACCESS_WITH_SIZE.
        (is_access_with_size_p and get_ref_from_access_with_size in tree.cc)
      * testing cases. (for offsetof, static initialization, generation of 
calls to
        .ACCESS_WITH_SIZE, code runs correctly after calls to .ACCESS_WITH_SIZE 
are
        converted back)

3. Use the .ACCESS_WITH_SIZE in builtin object size (sub-object only)
      which includes:
      * use the size info of the .ACCESS_WITH_SIZE for sub-object.
      * when the size is a negative integer, treat it as zero.
      * testing cases. 

4 Use the .ACCESS_WITH_SIZE in bound sanitizer
      which includes:
      * Instrument array_ref with a call to .ACCESS_WITH_SIZE for bound 
sanitizer.
      * when the size is a negative integer, treat it as zero.
      * testing cases. 

5. Add the 6th argument to .ACCESS_WITH_SIZE to carry the TYPE of the flexible 
array.
      which includes:
      * Add the 6th argument to .ACCESS_WITH_SIZE.
      * use the type of the 6th argument of the routine in tree-object-size.cc
      * testing case.

******Remaining works: 

6  Improve __bdos to use the counted_by info in whole-object size for the 
structure with FAM.
7  Emit warnings when the user breaks the requirments for the new counted_by 
attribute
   compilation time: -Wcounted-by
   run time: -fsanitizer=counted-by
      * The initialization to the size field should be done before the first 
reference to the FAM field.
      * the array has at least # of elements specified by the size field all 
the time during the program.

I have bootstrapped and regression tested on both x86 and aarch64, no issue.

Let me know your comments.

thanks.

Qing

Reply via email to