Fix handling of long map names (#1328)

Fixes #1312

Signed-off-by: Dave Thaler <dthaler@microsoft.com>

Signed-off-by: Dave Thaler <dthaler@microsoft.com>

libs/api/ebpf_api.cpp

@@ -1848,13 +1848,13 @@ typedef struct _ebpf_pe_context
     const bounded_buffer* data_buffer;
 } ebpf_pe_context_t;
 
-static int
+static int // Returns 0 on success, 1 on error.
 _ebpf_pe_get_map_definitions(
-    void* context,
-    const VA& va,
-    const std::string& section_name,
-    const image_section_header& section_header,
-    const bounded_buffer* buffer) noexcept
+    _Inout_ void* context,
+    _In_ const VA& va,
+    _In_ const std::string& section_name,
+    _In_ const image_section_header& section_header,
+    _In_ const bounded_buffer* buffer) noexcept
 {
     EBPF_LOG_ENTRY();
     UNREFERENCED_PARAMETER(va);
@@ -1863,8 +1863,8 @@ _ebpf_pe_get_map_definitions(
     ebpf_map_t* map = nullptr;
     ebpf_pe_context_t* pe_context = (ebpf_pe_context_t*)context;
     if (section_name == "maps") {
-        // bpf2c generates a section that has map names as strings at the
-        // start of the section.  Skip over them looking for the map_entry_t
+        // bpf2c generates a section that has map names shorter than sizeof(map_entry_t)
+        // at the start of the section.  Skip over them looking for the map_entry_t
         // which starts with an 8-byte-aligned NULL pointer where the previous
         // byte (if any) is also 00, and the following 8 bytes are non-NULL.
         uint32_t map_offset = 0;
@@ -1879,6 +1879,12 @@ _ebpf_pe_get_map_definitions(
             for (int map_index = 0; map_offset + sizeof(map_entry_t) <= section_header.Misc.VirtualSize;
                  map_offset += sizeof(map_entry_t), map_index++) {
                 map_entry_t* entry = (map_entry_t*)(buffer->buf + map_offset);
+                if (entry->address != nullptr) {
+                    // bpf2c generates a section that has map names longer than sizeof(map_entry_t)
+                    // at the end of the section.  This entry seems to be a map name string, so we've
+                    // reached the end of the maps.
+                    break;
+                }
 
                 map = (ebpf_map_t*)calloc(1, sizeof(ebpf_map_t));
                 if (map == nullptr) {

tests/end_to_end/end_to_end.cpp

@@ -492,6 +492,32 @@ divide_by_zero_test_um(ebpf_execution_type_t execution_type)
     bpf_object__close(object);
 }
 
+void
+bad_map_name_um(ebpf_execution_type_t execution_type)
+{
+    _test_helper_end_to_end test_helper;
+
+    int result;
+    const char* error_message = nullptr;
+    bpf_object* object = nullptr;
+    fd_t program_fd;
+
+    single_instance_hook_t hook(EBPF_PROGRAM_TYPE_XDP, EBPF_ATTACH_TYPE_XDP);
+    program_info_provider_t xdp_program_info(EBPF_PROGRAM_TYPE_XDP);
+
+    const char* file_name = (execution_type == EBPF_EXECUTION_NATIVE ? "bad_map_name_um.dll" : "bad_map_name.o");
+
+    result = ebpf_program_load(file_name, BPF_PROG_TYPE_UNSPEC, execution_type, &object, &program_fd, &error_message);
+
+    if (error_message) {
+        printf("ebpf_program_load failed with %s\n", error_message);
+        free((void*)error_message);
+    }
+    REQUIRE(result == -EINVAL);
+
+    bpf_object__close(object);
+}
+
 typedef struct _process_entry
 {
     uint32_t count;
@@ -873,6 +899,7 @@ DECLARE_ALL_TEST_CASES("bindmonitor-tailcall", "[end_to_end]", bindmonitor_tailc
 DECLARE_ALL_TEST_CASES("bindmonitor-ringbuf", "[end_to_end]", bindmonitor_ring_buffer_test);
 DECLARE_ALL_TEST_CASES("utility-helpers", "[end_to_end]", _utility_helper_functions_test);
 DECLARE_ALL_TEST_CASES("map", "[end_to_end]", map_test);
+DECLARE_ALL_TEST_CASES("bad_map_name", "[end_to_end]", bad_map_name_um);
 
 TEST_CASE("enum section", "[end_to_end]")
 {

tests/sample/bad_map_name.c

@@ -0,0 +1,26 @@
+// Copyright (c) Microsoft Corporation
+// SPDX-License-Identifier: MIT
+
+// Whenever this sample program changes, bpf2c_tests will fail unless the
+// expected files in tests\bpf2c_tests\expected are updated. The following
+// script can be used to regenerate the expected files:
+//     generate_expected_bpf2c_output.ps1
+//
+// Usage:
+// .\scripts\generate_expected_bpf2c_output.ps1 <build_output_path>
+// Example:
+// .\scripts\generate_expected_bpf2c_output.ps1 .\x64\Debug\
+
+#include "bpf_helpers.h"
+
+SEC("maps")
+struct bpf_map_def this_map_has_a_name_that_is_longer_than_what_the_ebofcore_driver_can_support = {
+    .type = BPF_MAP_TYPE_HASH, .key_size = sizeof(uint32_t), .value_size = sizeof(uint32_t), .max_entries = 1};
+
+SEC("xdp_prog") int lookup(struct xdp_md* ctx)
+{
+    uint32_t key = 0;
+    void* value =
+        bpf_map_lookup_elem(&this_map_has_a_name_that_is_longer_than_what_the_ebofcore_driver_can_support, &key);
+    return (value == NULL);
+}