Use gimli::write::Dwarf::convert_with_filter for DWARF transform (#12428)

Replace parts of the transform code with `gimli`'s generic DWARF
transformation support. Most of the remaining code is specific to the
needs of Wasmtime.

The overall behaviour is the same as the previous implementation. We
build a graph of the DIE dependencies, and prune DIEs that don't have
valid code ranges. Then we traverse the DIE tree again and transform the
DIEs. However, there are some differences.

Previously, unresolved references to DIEs were stored in `PendingUnitRefs`
and `PendingDebugInfoRefs`, then at the end of the transformation we
went back and fixed up these references. The new behaviour reserves IDs
for all of the DIEs in the dependency tree before transformation, which
avoids the need to fix up references later. It also allows gimli to
correctly handle references in DWARF expressions, although that doesn't
currently matter for Wasmtime because it doesn't handle references in
expressions yet. The visible effect of this in Wasmtime is that the
order of transformed attributes will now always match the original DWARF.

The DIE tree pruning is slightly different. We no longer add back-edges
pointing to namespace DIEs or the root DIE, which previously caused some
DIEs to be reachable when they should not have been. In particular,
this affects DW_TAG_variable DIEs for global variables. Wasmtime can't
currently translate the location for globals, so they were added without
a valid DW_AT_location. These DIEs are now omitted from the transformed
DWARF. In the future when global variables can be translated correctly,
they can be included by calling `ReserveUnitSection::require_entry`.

Performance measurements for `wasmtime compile -D debug-info=y` show
minimal change.

Author: philipc

crates/cranelift/src/debug/gc.rs

@@ -1,154 +1,25 @@
 use crate::debug::Reader;
 use crate::debug::transform::AddressTransform;
-use gimli::UnitSectionOffset;
 use gimli::constants;
 use gimli::read;
-use std::collections::{HashMap, HashSet};
-
-#[derive(Debug)]
-pub struct Dependencies {
-    edges: HashMap<UnitSectionOffset, HashSet<UnitSectionOffset>>,
-    roots: HashSet<UnitSectionOffset>,
-}
-
-impl Dependencies {
-    fn new() -> Dependencies {
-        Dependencies {
-            edges: HashMap::new(),
-            roots: HashSet::new(),
-        }
-    }
-
-    fn add_edge(&mut self, a: UnitSectionOffset, b: UnitSectionOffset) {
-        use std::collections::hash_map::Entry;
-        match self.edges.entry(a) {
-            Entry::Occupied(mut o) => {
-                o.get_mut().insert(b);
-            }
-            Entry::Vacant(v) => {
-                let mut set = HashSet::new();
-                set.insert(b);
-                v.insert(set);
-            }
-        }
-    }
-
-    fn add_root(&mut self, root: UnitSectionOffset) {
-        self.roots.insert(root);
-    }
-
-    pub fn get_reachable(&self) -> HashSet<UnitSectionOffset> {
-        let mut reachable = self.roots.clone();
-        let mut queue = Vec::new();
-        for i in self.roots.iter() {
-            if let Some(deps) = self.edges.get(i) {
-                for j in deps {
-                    if reachable.contains(j) {
-                        continue;
-                    }
-                    reachable.insert(*j);
-                    queue.push(*j);
-                }
-            }
-        }
-        while let Some(i) = queue.pop() {
-            if let Some(deps) = self.edges.get(&i) {
-                for j in deps {
-                    if reachable.contains(j) {
-                        continue;
-                    }
-                    reachable.insert(*j);
-                    queue.push(*j);
-                }
-            }
-        }
-        reachable
-    }
-}
+use gimli::write;
 
 pub fn build_dependencies(
-    dwarf: &read::Dwarf<Reader<'_>>,
+    filter: &mut write::FilterUnitSection<'_, Reader<'_>>,
     at: &AddressTransform,
-) -> read::Result<Dependencies> {
-    let mut deps = Dependencies::new();
-    let mut units = dwarf.units();
-    while let Some(unit) = units.next()? {
-        build_unit_dependencies(unit, dwarf, at, &mut deps)?;
+) -> write::ConvertResult<()> {
+    while let Some(mut unit) = filter.read_unit()? {
+        build_die_dependencies(&mut unit, at)?;
     }
-    Ok(deps)
-}
-
-fn build_unit_dependencies(
-    header: read::UnitHeader<Reader<'_>>,
-    dwarf: &read::Dwarf<Reader<'_>>,
-    at: &AddressTransform,
-    deps: &mut Dependencies,
-) -> read::Result<()> {
-    let unit = dwarf.unit(header)?;
-    let unit = unit.unit_ref(dwarf);
-    let mut tree = unit.entries_tree(None)?;
-    let root = tree.root()?;
-    build_die_dependencies(root, unit, at, deps)?;
     Ok(())
 }
 
-fn has_die_back_edge(die: &read::DebuggingInformationEntry<Reader<'_>>) -> read::Result<bool> {
-    // DIEs can be broadly divided into three categories:
-    // 1. Extensions of their parents; effectively attributes: DW_TAG_variable, DW_TAG_member, etc.
-    // 2. Standalone entities referred to by other DIEs via 'reference' class attributes: types.
-    // 3. Structural entities that organize how the above relate to each other: namespaces.
-    // Here, we must make sure to return 'true' for DIEs in the first category since stripping them,
-    // provided their parent is alive, is always wrong. To be conservatively correct in the face
-    // of new/vendor tags, we maintain a "(mostly) known good" list of tags of the latter categories.
-    let result = match die.tag() {
-        constants::DW_TAG_array_type
-        | constants::DW_TAG_atomic_type
-        | constants::DW_TAG_base_type
-        | constants::DW_TAG_class_type
-        | constants::DW_TAG_const_type
-        | constants::DW_TAG_dwarf_procedure
-        | constants::DW_TAG_entry_point
-        | constants::DW_TAG_enumeration_type
-        | constants::DW_TAG_pointer_type
-        | constants::DW_TAG_ptr_to_member_type
-        | constants::DW_TAG_reference_type
-        | constants::DW_TAG_restrict_type
-        | constants::DW_TAG_rvalue_reference_type
-        | constants::DW_TAG_string_type
-        | constants::DW_TAG_structure_type
-        | constants::DW_TAG_typedef
-        | constants::DW_TAG_union_type
-        | constants::DW_TAG_unspecified_type
-        | constants::DW_TAG_volatile_type
-        | constants::DW_TAG_coarray_type
-        | constants::DW_TAG_common_block
-        | constants::DW_TAG_dynamic_type
-        | constants::DW_TAG_file_type
-        | constants::DW_TAG_immutable_type
-        | constants::DW_TAG_interface_type
-        | constants::DW_TAG_set_type
-        | constants::DW_TAG_shared_type
-        | constants::DW_TAG_subroutine_type
-        | constants::DW_TAG_packed_type
-        | constants::DW_TAG_template_alias
-        | constants::DW_TAG_namelist
-        | constants::DW_TAG_namespace
-        | constants::DW_TAG_imported_unit
-        | constants::DW_TAG_imported_declaration
-        | constants::DW_TAG_imported_module
-        | constants::DW_TAG_module => false,
-        constants::DW_TAG_subprogram => die.attr(constants::DW_AT_declaration).is_some(),
-        _ => true,
-    };
-    Ok(result)
-}
-
 fn has_valid_code_range(
-    die: &read::DebuggingInformationEntry<Reader<'_>>,
-    unit: read::UnitRef<Reader<'_>>,
+    die: &write::FilterUnitEntry<'_, Reader<'_>>,
     at: &AddressTransform,
 ) -> read::Result<bool> {
-    match die.tag() {
+    let unit = die.read_unit;
+    match die.tag {
         constants::DW_TAG_subprogram => {
             if let Some(ranges_attr) = die.attr_value(constants::DW_AT_ranges) {
                 let offset = match ranges_attr {
@@ -215,50 +86,14 @@ fn has_valid_code_range(
 }
 
 fn build_die_dependencies(
-    die: read::EntriesTreeNode<Reader<'_>>,
-    unit: read::UnitRef<Reader<'_>>,
+    unit: &mut write::FilterUnit<'_, Reader<'_>>,
     at: &AddressTransform,
-    deps: &mut Dependencies,
-) -> read::Result<()> {
-    let entry = die.entry();
-    let offset = entry.offset().to_unit_section_offset(&unit);
-    for attr in entry.attrs() {
-        build_attr_dependencies(attr, offset, unit, at, deps)?;
-    }
-
-    let mut children = die.children();
-    while let Some(child) = children.next()? {
-        let child_entry = child.entry();
-        let child_offset = child_entry.offset().to_unit_section_offset(&unit);
-        deps.add_edge(child_offset, offset);
-        if has_die_back_edge(child_entry)? {
-            deps.add_edge(offset, child_offset);
-        }
-        if has_valid_code_range(child_entry, unit, at)? {
-            deps.add_root(child_offset);
+) -> write::ConvertResult<()> {
+    let mut die = write::FilterUnitEntry::null(unit.read_unit);
+    while unit.read_entry(&mut die)? {
+        if has_valid_code_range(&die, at)? {
+            unit.require_entry(die.offset);
         }
-        build_die_dependencies(child, unit, at, deps)?;
-    }
-    Ok(())
-}
-
-fn build_attr_dependencies(
-    attr: &read::Attribute<Reader<'_>>,
-    offset: UnitSectionOffset,
-    unit: read::UnitRef<Reader<'_>>,
-    _at: &AddressTransform,
-    deps: &mut Dependencies,
-) -> read::Result<()> {
-    match attr.value() {
-        read::AttributeValue::UnitRef(val) => {
-            let ref_offset = val.to_unit_section_offset(&unit);
-            deps.add_edge(offset, ref_offset);
-        }
-        read::AttributeValue::DebugInfoRef(val) => {
-            let ref_offset = UnitSectionOffset(val.0);
-            deps.add_edge(offset, ref_offset);
-        }
-        _ => (),
     }
     Ok(())
 }