Merge branch 'refs/heads/upstream-HEAD' into repo-HEAD

.github/workflows/ci.yml

@@ -55,7 +55,7 @@ jobs:
         run: |
           sudo apt-get update -y
           sudo apt-get install -y check dwarves libelf-dev libdw-dev qemu-kvm zstd ${{ matrix.cc == 'clang' && 'libomp-$(clang --version | sed -rn "s/.*clang version ([0-9]+).*/\\1/p")-dev' || '' }}
-          pip install pyroute2 ${USE_PRE_COMMIT/1/pre-commit}
+          pip install pyroute2 setuptools ${USE_PRE_COMMIT/1/pre-commit}
       - name: Generate version.py
         run: python setup.py --version
       - name: Check with mypy

.packit.yaml

@@ -25,39 +25,63 @@ jobs:
   owner: "@meta"
   project: drgn
   targets:
-    - fedora-all-aarch64
-    - fedora-all-i386
-    - fedora-all-ppc64le
-    - fedora-all-s390x
-    - fedora-all-x86_64
-    - fedora-eln-aarch64
-    - fedora-eln-i386
-    - fedora-eln-ppc64le
+    fedora-all-aarch64: {}
+    fedora-all-i386: {}
+    fedora-all-ppc64le: {}
+    fedora-all-s390x: {}
+    fedora-all-x86_64: {}
+    fedora-eln-aarch64: {}
+    fedora-eln-i386: {}
+    fedora-eln-ppc64le: {}
     # Disabled due to fedora-eln/eln#170.
-    # - fedora-eln-s390x
-    - fedora-eln-x86_64
-    - epel-8-aarch64
-    - epel-8-ppc64le
-    - epel-8-s390x
-    - epel-8-x86_64
+    # fedora-eln-s390x: {}
+    fedora-eln-x86_64: {}
+    epel-8-aarch64: {}
+    epel-8-ppc64le: {}
+    epel-8-s390x: {}
+    epel-8-x86_64: {}
+    centos-stream+epel-next-9-aarch64:
+      additional_repos:
+        - https://kojihub.stream.centos.org/kojifiles/repos/c9s-build/latest/aarch64/
+    centos-stream+epel-next-9-ppc64le:
+      additional_repos:
+        - https://kojihub.stream.centos.org/kojifiles/repos/c9s-build/latest/ppc64le/
+    centos-stream+epel-next-9-s390x:
+      additional_repos:
+        - https://kojihub.stream.centos.org/kojifiles/repos/c9s-build/latest/s390x/
+    centos-stream+epel-next-9-x86_64:
+      additional_repos:
+        - https://kojihub.stream.centos.org/kojifiles/repos/c9s-build/latest/x86_64/
 
 - job: copr_build
   trigger: pull_request
   owner: "@meta"
   project: drgn
   targets:
-    - fedora-all-aarch64
-    - fedora-all-i386
-    - fedora-all-ppc64le
-    - fedora-all-s390x
-    - fedora-all-x86_64
-    - fedora-eln-aarch64
-    - fedora-eln-i386
-    - fedora-eln-ppc64le
+    fedora-all-aarch64: {}
+    fedora-all-i386: {}
+    fedora-all-ppc64le: {}
+    fedora-all-s390x: {}
+    fedora-all-x86_64: {}
+    fedora-eln-aarch64: {}
+    fedora-eln-i386: {}
+    fedora-eln-ppc64le: {}
     # Disabled due to fedora-eln/eln#170.
-    # - fedora-eln-s390x
-    - fedora-eln-x86_64
-    - epel-8-aarch64
-    - epel-8-ppc64le
-    - epel-8-s390x
-    - epel-8-x86_64
+    # fedora-eln-s390x: {}
+    fedora-eln-x86_64: {}
+    epel-8-aarch64: {}
+    epel-8-ppc64le: {}
+    epel-8-s390x: {}
+    epel-8-x86_64: {}
+    centos-stream+epel-next-9-aarch64:
+      additional_repos:
+        - https://kojihub.stream.centos.org/kojifiles/repos/c9s-build/latest/aarch64/
+    centos-stream+epel-next-9-ppc64le:
+      additional_repos:
+        - https://kojihub.stream.centos.org/kojifiles/repos/c9s-build/latest/ppc64le/
+    centos-stream+epel-next-9-s390x:
+      additional_repos:
+        - https://kojihub.stream.centos.org/kojifiles/repos/c9s-build/latest/s390x/
+    centos-stream+epel-next-9-x86_64:
+      additional_repos:
+        - https://kojihub.stream.centos.org/kojifiles/repos/c9s-build/latest/x86_64/

_drgn.pyi

@@ -67,14 +67,22 @@ class Program:
     :meth:`[] <.__getitem__>` operator.
     """
 
-    def __init__(self, platform: Optional[Platform] = None) -> None:
+    def __init__(
+        self,
+        platform: Optional[Platform] = None,
+        *,
+        vmcoreinfo: Union[bytes, str, None] = None,
+    ) -> None:
         """
         Create a ``Program`` with no target program. It is usually more
         convenient to use one of the :ref:`api-program-constructors`.
 
         :param platform: The platform of the program, or ``None`` if it should
             be determined automatically when a core dump or symbol file is
             added.
+        :param vmcoreinfo: Optionally provide the ``VMCOREINFO`` note data for
+            Linux kernel core dumps, which will override any detected data. When
+            not provided or ``None``, automatically detect the info.
         """
         ...
     flags: ProgramFlags

contrib/bpf_inspect.py

@@ -4,14 +4,21 @@
 
 """List BPF programs or maps and their properties unavailable via kernel API."""
 
+import sys
 import argparse
 
 from drgn.helpers.common.type import enum_type_to_class
-from drgn.helpers.linux import bpf_map_for_each, bpf_prog_for_each, hlist_for_each_entry
+from drgn.helpers.linux import (
+    bpf_map_for_each,
+    bpf_prog_for_each,
+    bpf_link_for_each,
+    hlist_for_each_entry,
+)
 
 BpfMapType = enum_type_to_class(prog.type("enum bpf_map_type"), "BpfMapType")
 BpfProgType = enum_type_to_class(prog.type("enum bpf_prog_type"), "BpfProgType")
 BpfAttachType = enum_type_to_class(prog.type("enum bpf_attach_type"), "BpfAttachType")
+BpfLinkType = enum_type_to_class(prog.type("enum bpf_link_type"), "BpfLinkType")
 
 
 def get_btf_name(btf, btf_id):
@@ -29,11 +36,11 @@ def get_prog_btf_name(bpf_prog):
     return ""
 
 
-def get_prog_name(bpf_prog):
+def get_bpf_prog_name(bpf_prog):
     return get_prog_btf_name(bpf_prog) or bpf_prog.aux.name.string_().decode()
 
 
-def attach_type_to_tramp(attach_type):
+def bpf_attach_type_to_tramp(attach_type):
     # bpf_tramp_prog_type is available since linux kernel 5.5, this code should
     # be called only after checking for bpf_prog.aux.trampoline to be present
     # though so no error checking here.
@@ -52,21 +59,21 @@ def attach_type_to_tramp(attach_type):
     return BpfProgTrampType.BPF_TRAMP_REPLACE
 
 
-def get_linked_func(bpf_prog):
-    kind = attach_type_to_tramp(bpf_prog.expected_attach_type)
+def get_bpf_linked_func(bpf_prog):
+    kind = bpf_attach_type_to_tramp(bpf_prog.expected_attach_type)
 
     linked_prog = bpf_prog.aux.linked_prog
     linked_prog_id = linked_prog.aux.id.value_()
     linked_btf_id = bpf_prog.aux.attach_btf_id.value_()
     linked_name = (
-        f"{get_prog_name(linked_prog)}->"
+        f"{get_bpf_prog_name(linked_prog)}->"
         f"{get_btf_name(linked_prog.aux.btf, linked_btf_id)}()"
     )
 
     return f"{linked_prog_id}->{linked_btf_id}: {kind.name} {linked_name}"
 
 
-def get_tramp_progs(bpf_prog):
+def get_bpf_tramp_progs(bpf_prog):
     try:
         tr = bpf_prog.aux.member_("trampoline")
     except LookupError:
@@ -88,20 +95,28 @@ def get_tramp_progs(bpf_prog):
                 yield tramp_aux.prog
 
 
-def list_bpf_progs(args):
+def inspect_bpf_prog(bpf_prog):
+    id_ = bpf_prog.aux.id.value_()
+    type_ = BpfProgType(bpf_prog.type).name
+    name = get_bpf_prog_name(bpf_prog)
+
+    linked = ", ".join([get_bpf_linked_func(p) for p in get_bpf_tramp_progs(bpf_prog)])
+    if linked:
+        linked = f" linked:[{linked}]"
+
+    return f"{id_:>6}: {type_:32} {name:32} {linked}"
+
+
+def list_bpf_progs():
     for bpf_prog in bpf_prog_for_each(prog):
-        id_ = bpf_prog.aux.id.value_()
-        type_ = BpfProgType(bpf_prog.type).name
-        name = get_prog_name(bpf_prog)
+        print(inspect_bpf_prog(bpf_prog))
 
-        linked = ", ".join([get_linked_func(p) for p in get_tramp_progs(bpf_prog)])
-        if linked:
-            linked = f" linked:[{linked}]"
 
-        print(f"{id_:>6}: {type_:32} {name:32} {linked}")
+def __list_bpf_progs(args):
+    list_bpf_progs()
 
 
-def list_bpf_maps(args):
+def list_bpf_maps():
     for map_ in bpf_map_for_each(prog):
         id_ = map_.id.value_()
         type_ = BpfMapType(map_.map_type).name
@@ -110,6 +125,46 @@ def list_bpf_maps(args):
         print(f"{id_:>6}: {type_:32} {name}")
 
 
+def __list_bpf_maps(args):
+    list_bpf_maps()
+
+
+def list_bpf_links():
+    for link in bpf_link_for_each(prog):
+        id_ = link.id.value_()
+        type_ = BpfLinkType(link.type).name
+        prog_ = inspect_bpf_prog(link.prog)
+
+        print(f"{id_:>6}: {type_:32} prog: {prog_}")
+
+
+def __list_bpf_links(args):
+    list_bpf_links()
+
+
+def __run_interactive(args):
+    try:
+        from drgn.cli import run_interactive
+    except ImportError:
+        sys.exit("Interactive mode requires drgn 0.0.23+")
+
+    def should_add_to_globals(name):
+        if name.startswith("__"):
+            return False
+        return "bpf" in name or "Bpf" in name or "btf" in name
+
+    globals_keys = globals().keys()
+
+    def globals_func(globals_):
+        for key in globals_keys:
+            if should_add_to_globals(key):
+                globals_[key] = globals()[key]
+
+        return globals_
+
+    run_interactive(prog, globals_func=globals_func)
+
+
 def main():
     parser = argparse.ArgumentParser(
         description="drgn script to list BPF programs or maps and their properties unavailable via kernel API"
@@ -119,10 +174,18 @@ def main():
     subparsers.required = True
 
     prog_parser = subparsers.add_parser("prog", aliases=["p"], help="list BPF programs")
-    prog_parser.set_defaults(func=list_bpf_progs)
+    prog_parser.set_defaults(func=__list_bpf_progs)
 
     map_parser = subparsers.add_parser("map", aliases=["m"], help="list BPF maps")
-    map_parser.set_defaults(func=list_bpf_maps)
+    map_parser.set_defaults(func=__list_bpf_maps)
+
+    link_parser = subparsers.add_parser("link", aliases=["l"], help="list BPF links")
+    link_parser.set_defaults(func=__list_bpf_links)
+
+    interact_parser = subparsers.add_parser(
+        "interact", aliases=["i"], help="start interactive shell, requires 0.0.23+ drgn"
+    )
+    interact_parser.set_defaults(func=__run_interactive)
 
     args = parser.parse_args()
     args.func(args)

contrib/lsmod.py

@@ -6,11 +6,26 @@
 
 from drgn.helpers.linux.list import list_for_each_entry
 
+def module_total_size(mod):
+    # Since Linux kernel commit ac3b43283923 ("module: replace module_layout
+    # with module_memory") (in v6.4), the memory sizes are in the struct
+    # module::mem array. Before that, they are in struct module::init_layout
+    # and struct module::core_layout.
+    try:
+        num_types = mod.prog_["MOD_MEM_NUM_TYPES"]
+    except KeyError:
+        return (mod.init_layout.size + mod.core_layout.size).value_()
+    else:
+        return sum(
+            mod.mem[type].size.value_()
+            for type in range(num_types)
+        )
+
+
 print("Module                  Size  Used by")
 config_module_unload = prog.type("struct module").has_member("refcnt")
 for mod in list_for_each_entry("struct module", prog["modules"].address_of_(), "list"):
     name = mod.name.string_().decode()
-    size = (mod.init_layout.size + mod.core_layout.size).value_()
     if config_module_unload:
         refcnt = mod.refcnt.counter.value_() - 1
         used_by = [
@@ -26,4 +41,4 @@
     used = ",".join(used_by)
     if used:
         used = " " + used
-    print(f"{name:19} {size:>8}  {refcnt}{used}")
+    print(f"{name:19} {module_total_size(mod):>8}  {refcnt}{used}")

drgn/cli.py

@@ -209,6 +209,20 @@ def _main() -> None:
         help="don't load any debugging symbols that were not explicitly added with -s",
     )
 
+    advanced_group = parser.add_argument_group("advanced")
+    advanced_group.add_argument(
+        "--architecture",
+        metavar="ARCH",
+        choices=[a.name for a in drgn.Architecture]
+        + [a.name.lower() for a in drgn.Architecture],
+        help="set the program architecture, in case it can't be auto-detected",
+    )
+    advanced_group.add_argument(
+        "--vmcoreinfo",
+        type=str,
+        metavar="PATH",
+        help="path to vmcoreinfo file (overrides any already present in the file)",
+    )
     parser.add_argument(
         "--log-level",
         choices=["debug", "info", "warning", "error", "critical", "none"],
@@ -258,7 +272,16 @@ def _main() -> None:
     else:
         logger.setLevel(args.log_level.upper())
 
-    prog = drgn.Program()
+    platform = None
+    if args.architecture:
+        platform = drgn.Platform(drgn.Architecture[args.architecture.upper()])
+
+    vmcoreinfo = None
+    if args.vmcoreinfo is not None:
+        with open(args.vmcoreinfo, "rb") as f:
+            vmcoreinfo = f.read()
+
+    prog = drgn.Program(platform=platform, vmcoreinfo=vmcoreinfo)
     try:
         if args.core is not None:
             prog.set_core_dump(args.core)

libdrgn/arch_x86_64.c

@@ -18,6 +18,10 @@
 
 #include "arch_x86_64_defs.inc"
 
+// This is __START_KERNEL_map from the Linux kernel. It has never been modified
+// since the x86_64 architecture was introduced.
+#define START_KERNEL_MAP UINT64_C(0xffffffff80000000)
+
 static const struct drgn_cfi_row default_dwarf_cfi_row_x86_64 = DRGN_CFI_ROW(
 	// The System V psABI defines the CFA as the value of rsp in the calling
 	// frame.
@@ -616,7 +620,18 @@ linux_kernel_pgtable_iterator_next_x86_64(struct drgn_program *prog,
 	for (;; level--) {
 		uint64_t table;
 		bool table_physical;
-		if (level == levels) {
+		if (level == levels && prog->vmcoreinfo.phys_base &&
+		    it->it.pgtable == prog->vmcoreinfo.swapper_pg_dir) {
+			// Avoid recursive address translation on swapper_pg_dir by
+			// directly resolving to a physical address. Don't do
+			// this if phys_base is 0, since that likely means it
+			// was not present in the vmcoreinfo. It has been
+			// present since Linux kernel commit 401721ecd1dc
+			// ("kexec: export the value of phys_base instead of
+			// symbol address") (in v4.10).
+			table = it->it.pgtable + prog->vmcoreinfo.phys_base - START_KERNEL_MAP;
+			table_physical = true;
+		} else if (level == levels) {
 			table = it->it.pgtable;
 			table_physical = false;
 		} else {