lua: Properly support high-range 64 addresses

Lua's native Number type is a 64-bit double with 52-bit precision, which
was causing rounding errors when storing and working with high-range
memory addresses (namely, addresses from the kernel, which are all in
the `0xffffffff........` range).

To work around this, we've made sure to never call `tonumber` on any
variables that represent memory addresses, and instead continue
operating on them with their native types: LuaJIT can work with the
underlying `uint64_t` type for these values and transparently perform
all kinds of numeric operations.

The only limitation of working with native 64-bit types in LuaJIT is
that they cannot be printed with the language's default `string.format`
API. To give better UX to probe writers, these APIs have been
monkeypatched so the `%p` format specifier will now properly handle
64-bit addresses and print them in an appropriate format.

examples/lua/memleak.lua

@@ -148,7 +148,7 @@ return function(BPF, utils)
     if args.pid == nil then
       sym = sym .. " [kernel]"
     end
-    return string.format("%s (%x)", sym, addr)
+    return string.format("%s (%p)", sym, addr)
   end
 
   local function print_outstanding()
@@ -174,7 +174,7 @@ return function(BPF, utils)
         end
 
         if args.show_allocs then
-          print("\taddr = %x size = %s" % {tonumber(address), tonumber(info.size)})
+          print("\taddr = %p size = %s" % {address, tonumber(info.size)})
         end
       end
     end

examples/lua/offcputime.lua

@@ -107,7 +107,7 @@ return function(BPF, utils)
 
   for k, v in counts:items() do
     for addr in stack_traces:walk(tonumber(k.stack_id)) do
-      print("    %-16x %s" % {addr, ksym:lookup(addr)})
+      print("    %-16p %s" % {addr, ksym:lookup(addr)})
     end
     print("    %-16s %s" % {"-", ffi.string(k.name)})
     print("        %d\n" % tonumber(v))

src/lua/bcc/bpf.lua

@@ -188,8 +188,8 @@ function Bpf:attach_uprobe(args)
   local path, addr = LD.check_path_symbol(args.name, args.sym, args.addr)
   local fn = self:load_func(args.fn_name, 'BPF_PROG_TYPE_KPROBE')
   local ptype = args.retprobe and "r" or "p"
-  local ev_name = string.format("%s_%s_0x%x", ptype, path:gsub("[^%a%d]", "_"), addr)
-  local desc = string.format("%s:uprobes/%s %s:0x%x", ptype, ev_name, path, addr)
+  local ev_name = string.format("%s_%s_0x%p", ptype, path:gsub("[^%a%d]", "_"), addr)
+  local desc = string.format("%s:uprobes/%s %s:0x%p", ptype, ev_name, path, addr)
 
   log.info(desc)
 

src/lua/bcc/ld.lua

@@ -14,6 +14,7 @@ See the License for the specific language governing permissions and
 limitations under the License.
 ]]
 local ffi = require("ffi")
+local posix = require("bcc.vendor.posix")
 
 local _find_library_cache = {}
 local function _find_library(name)
@@ -60,7 +61,7 @@ local function _find_load_address(path)
     path)
 
   if addr then
-    addr = tonumber(addr, 16)
+    addr = posix.tonumber64(addr, 16)
     _find_load_address_cache[path] = addr
   end
 
@@ -85,7 +86,7 @@ local function _find_symbol(path, sym)
     path, sym)
 
   if addr then
-    addr = tonumber(addr, 16)
+    addr = posix.tonumber64(addr, 16)
     symbols[sym] = addr
   end
 

src/lua/bcc/sym.lua

@@ -13,6 +13,7 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License.
 ]]
+local posix = require("bcc.vendor.posix")
 local ProcSymbols = class("ProcSymbols")
 
 function ProcSymbols:initialize(pid)
@@ -46,7 +47,7 @@ function ProcSymbols:_get_code_ranges()
       local range = parts[1]:split("-", true)
       assert(#range == 2)
 
-      ranges[binary] = {tonumber(range[1], 16), tonumber(range[2], 16)}
+      ranges[binary] = {posix.tonumber64(range[1], 16), posix.tonumber64(range[2], 16)}
     end
   end
 
@@ -83,8 +84,8 @@ function ProcSymbols:_get_sym_ranges(binary)
   for line in proc:lines() do
     local parts = line:split()
     if is_function_sym(parts) then
-      local sym_start = tonumber(parts[1], 16)
-      local sym_len = tonumber(parts[5], 16)
+      local sym_start = posix.tonumber64(parts[1], 16)
+      local sym_len = posix.tonumber64(parts[5], 16)
       local sym_name = parts[6]
       sym_ranges[sym_name] = {sym_start, sym_len}
     end
@@ -102,10 +103,10 @@ function ProcSymbols:_decode_sym(binary, offset)
     local start = range[1]
     local length = range[2]
     if offset >= start and offset <= (start + length) then
-      return string.format("%s+0x%x", name, offset - start)
+      return string.format("%s+0x%p", name, offset - start)
     end
   end
-  return string.format("%x", offset)
+  return string.format("%p", offset)
 end
 
 function ProcSymbols:lookup(addr)
@@ -121,7 +122,7 @@ function ProcSymbols:lookup(addr)
     end
   end
 
-  return string.format("%x", addr)
+  return string.format("%p", addr)
 end
 
 local KSymbols = class("KSymbols")
@@ -142,7 +143,7 @@ function KSymbols:_load()
     if not first_line then
       local cols = line:split()
       local name = cols[3]
-      local addr = tonumber(cols[1], 16)
+      local addr = posix.tonumber64(cols[1], 16)
       table.insert(self.ksyms, {name, addr})
       self.ksym_names[name] = #self.ksyms
     end
@@ -164,7 +165,7 @@ function KSymbols:lookup(addr, with_offset)
 
   if with_offset then
     local offset = addr - self.ksyms[idx][2]
-    return "%s %x" % {self.ksyms[idx][1], offset}
+    return "%s %p" % {self.ksyms[idx][1], offset}
   else
     return self.ksyms[idx][1]
   end

src/lua/bcc/table.lua

@@ -286,7 +286,7 @@ function StackTrace:walk(id)
       return nil
     end
 
-    local addr = tonumber(pstack[0].ip[i])
+    local addr = pstack[0].ip[i]
     if addr == 0 then
       return nil
     end

src/lua/bcc/vendor/helpers.lua

+do
+  local ffi = require("ffi")
+  local ptrtype = ffi.typeof("uint64_t")
+  local strformat = string.format
+  function string.format(format, ...)
+    local args = {...}
+    local match_no = 1
+    local newfmt, count = string.gsub(format, "()%%(.-)(%a)",
+      function(_, mods, t)
+        local n = match_no
+        match_no = match_no + 1
+        if t == 'p' and ffi.istype(ptrtype, args[n]) then
+          local lo = tonumber(args[n] % 4294967296ULL)
+          local hi = tonumber(args[n] / 4294967296ULL)
+          args[n] = (hi == 0) and strformat("%x", lo) or strformat("%x%08x", hi, lo)
+          return "%"..mods.."s"
+        end
+      end)
+    if count == 0 then
+      return strformat(format, ...)
+    else
+      return strformat(newfmt, unpack(args,1,select('#',...)))
+    end
+  end
+end
+
 function string.starts(s, p)
   return string.sub(s, 1, string.len(p)) == p
 end
@@ -66,7 +92,7 @@ function table.bsearch(list, value, mkval)
       return mid
     end
   end
-  return nil
+  return low - 1
 end
 
 function table.join(a, b)

src/lua/bcc/vendor/posix.lua

@@ -29,6 +29,8 @@ int clock_nanosleep(clockid_t clock_id, int flags,
   const struct timespec *request, struct timespec *remain);
 
 int get_nprocs(void);
+
+uint64_t strtoull(const char *nptr, char **endptr, int base);
 ]]
 
 local CLOCK = {
@@ -62,9 +64,15 @@ local function cpu_count()
   return tonumber(ffi.C.get_nprocs())
 end
 
+local function tonumber64(n, base)
+  assert(type(n) == "string")
+  return ffi.C.strtoull(n, nil, base or 10)
+end
+
 return {
   time_ns=time_ns,
   sleep=sleep,
   CLOCK=CLOCK,
   cpu_count=cpu_count,
+  tonumber64=tonumber64,
 }