On 12/14/25 8:10 AM, H.J. Lu wrote:
On Sat, Dec 13, 2025 at 5:33 PM Stefan Schulze Frielinghaus
<[email protected]> wrote:
On Sun, Oct 26, 2025 at 10:29:30AM +0800, H.J. Lu wrote:
On Sat, Oct 25, 2025 at 4:42 PM H.J. Lu <[email protected]> wrote:
...
Here is the v4 patch.
Add a new target hook, stack_protect_guard_symbol, to support the user
provided stack protection guard as an internal symbol. If the hook is
true,
1. Make __stack_chk_guard an internal C/C++ symbol.
2. Declare __stack_chk_guard as a size_t variable if size_t has the same
size as pointer so that it can be initialized as an integer.
3. If the user declared variable matches __stack_chk_guard, merge it
with __stack_chk_guard, including its visibility attribute.
4. Define the __stack_protection_guard_is_internal_symbol__ macro to
indicate that __stack_chk_guard is an internal symbol.
gcc/
PR c/121911
* target.def (stack_protect_guard_symbol): New target hook.
* targhooks.cc (default_stack_protect_guard): If
targetm.stack_protect_guard_symbol is true, make it an internal
symbol and use size_type_node if it has the same size as
ptr_type_node.
* tree.cc (build_common_tree_nodes): If stack protector is
enabled and targetm.stack_protect_guard_symbol is true, call
targetm.stack_protect_guard.
* config/i386/i386.cc (TARGET_STACK_PROTECT_GUARD_SYMBOL): New.
* doc/tm.texi: Regenerated.
* doc/tm.texi.in (TARGET_STACK_PROTECT_GUARD_SYMBOL): New.
gcc/c-family/
PR c/121911
* c-common.cc (matching_stack_protect_guard_decl_p): New.
* c-common.h (matching_stack_protect_guard_decl_p): Likewise.
* c-cppbuiltin.cc (c_cpp_builtins): Define the
__stack_protection_guard_is_internal_symbol__ macro if
targetm.stack_protect_guard_symbol is true.
gcc/c/
PR c/121911
* c-decl.cc: Include "c-family/c-common.h".
(merge_decls): Add a bool argument to keep the old type and copy
the old type if true.
(duplicate_decls): If the old decl is the stack protect guard and
the new decl matches the old decl, keep the type of the old decl.
gcc/cp/
PR c/121911
* decl.cc: Include "c-family/c-common.h".
(duplicate_decls): 2 types match if the old decl is the stack
protect guard and the new decl matches the old decl.
gcc/testsuite/
PR c/121911
* g++.target/i386/ssp-global-1.C: New test.
* g++.target/i386/ssp-global-2.C: Likewise.
* g++.target/i386/ssp-global-3.C: Likewise.
* g++.target/i386/ssp-global-4.C: Likewise.
* g++.target/i386/ssp-global-hidden-1.C: Likewise.
* g++.target/i386/ssp-global-hidden-2.C: Likewise.
* g++.target/i386/ssp-global-hidden-3.C: Likewise.
* gcc.target/i386/ssp-global-2.c: Likewise.
* gcc.target/i386/ssp-global-3.c: Likewise.
* gcc.target/i386/ssp-global-4.c: Likewise.
* gcc.target/i386/ssp-global-hidden-1.c: Likewise.
* gcc.target/i386/ssp-global-hidden-2.c: Likewise.
* gcc.target/i386/ssp-global-hidden-3.c: Likewise.
Here is the v5 patch. The main difference is that __stack_chk_guard
is an internal symbol with C linkage now.
Here is the v6 patch to rename the target hook, stack_protect_guard_symbol,
to stack_protect_guard_symbol_p.
Add a new target hook, stack_protect_guard_symbol_p, to support the user
provided stack protection guard as an internal symbol. If the hook
returns true,
1. Make __stack_chk_guard an internal C/C++ symbol.
2. Declare __stack_chk_guard as a size_t variable if size_t has the same
size as pointer so that it can be initialized as an integer.
3. If the user declared variable matches __stack_chk_guard, merge it
with __stack_chk_guard, including its visibility attribute.
4. Define the __stack_protection_guard_is_internal_symbol__ macro to
indicate that __stack_chk_guard is an internal symbol.
gcc/
PR c/121911
* target.def (stack_protect_guard_symbol_p): New target hook.
* targhooks.cc (default_stack_protect_guard): If the
stack_protect_guard_symbol_p hook returns true, use size_type_node
if it has the same size as ptr_type_node.
* config/i386/i386.cc (ix86_stack_protect_guard_symbol_p): New.
(TARGET_STACK_PROTECT_GUARD_SYMBOL_P): Likewise.
* doc/tm.texi: Regenerated.
* doc/tm.texi.in (TARGET_STACK_PROTECT_GUARD_SYMBOL_P): New.
gcc/c-family/
PR c/121911
* c-common.cc (stack_protect_guard_decl): New.
(pushing_stack_protect_guard_decl): Likewise.
(c_stack_protect_guard_decl_p): Likewise.
(duplicate_stack_protect_guard_decl_p): Likewise.
(c_common_nodes_and_builtins): If the stack_protect_guard_symbol_p
hook returns true, call targetm.stack_protect_guard.
* c-common.h (c_stack_protect_guard_decl_p): New.
(duplicate_stack_protect_guard_decl_p): Likewise.
* c-cppbuiltin.cc (c_cpp_builtins): Define the
__stack_protection_guard_is_internal_symbol__ macro if the
stack_protect_guard_symbol_p hook returns.
gcc/c/
PR c/121911
* c-decl.cc: Include "c-family/c-common.h".
(merge_decls): Add a bool argument to keep the old type and copy
the old type if true.
(duplicate_decls): If the old decl is the stack protect guard and
the new decl matches the old decl, keep the type of the old decl.
gcc/cp/
PR c/121911
* cp-objcp-common.cc (cp_pushdecl): Change the stack protection
guard symbol to C linkage.
* decl.cc: Include "c-family/c-common.h".
(duplicate_decls): 2 types match if the old decl is the stack
protect guard and the new decl matches the old decl.
gcc/testsuite/
PR c/121911
* g++.target/i386/ssp-global-1.C: New test.
* g++.target/i386/ssp-global-2.C: Likewise.
* g++.target/i386/ssp-global-3.C: Likewise.
* g++.target/i386/ssp-global-4.C: Likewise.
* g++.target/i386/ssp-global-hidden-1.C: Likewise.
* g++.target/i386/ssp-global-hidden-2.C: Likewise.
* g++.target/i386/ssp-global-hidden-3.C: Likewise.
* gcc.target/i386/ssp-global-2.c: Likewise.
* gcc.target/i386/ssp-global-3.c: Likewise.
* gcc.target/i386/ssp-global-4.c: Likewise.
* gcc.target/i386/ssp-global-hidden-1.c: Likewise.
* gcc.target/i386/ssp-global-hidden-2.c: Likewise.
* gcc.target/i386/ssp-global-hidden-3.c: Likewise.
+/* The stack protection guard. */
+static tree stack_protect_guard_decl;
I don't see why we need a local copy of targetm.stack_protect_guard()?
+/* True if the stack protection guard is being pushed. */
+static bool pushing_stack_protect_guard_decl;
...or a flag in addition to comparing with stack_protect_guard().
It seems like a lot of the complication of this patch is to allow
declarations with a different type, contrary to every other declaration.
This is quite novel, and it's not clear to me what the motivation is.
Why can't we just say it needs to be uintptr_t?
Jason
