[kernel] Add automatic verification to syscalls

Since we have a DSL for specifying syscalls, we can create a verificaton
method for each syscall that can cover most argument (and eventually
capability) verification instead of doing it piecemeal in each syscall
implementation, which can be more error-prone.

Now a new _syscall_verify_* function exists for every syscall, which
calls the real implementation. The syscall table for the syscall handler
now maps to these verify functions.

Other changes:

- Updated the definition grammar to allow options to have a "key:value"
  style, to eventually support capabilities.
- Added an "optional" option for parameters that says a syscall will
  accept a null value.
- Some bonnibel fixes, as definition file changes weren't always
  properly causing updates in the build dep graph.
- The syscall implementation function signatures are no longer exposed
  in syscall.h. Also, the unused syscall enum has been removed.

src/kernel/kernel.module

@@ -48,6 +48,7 @@ kernel = module("kernel",
         "syscall.cpp.cog",
         "syscall.h.cog",
         "syscall.s",
+        "syscall_verify.cpp.cog",
         "syscalls.inc.cog",
         "syscalls/channel.cpp",
         "syscalls/endpoint.cpp",
@@ -68,10 +69,15 @@ from os.path import join
 
 layout = join(source_root, "definitions/memory_layout.yaml")
 sysconf = join(source_root, "definitions/sysconf.yaml")
-definitions = glob('definitions/**/*.def', recursive=True)
+definitions = glob(join(source_root, 'definitions/**/*.def'), recursive=True)
 
 kernel.add_depends(["memory.h.cog"], [layout])
 kernel.add_depends(["sysconf.h.cog"], [sysconf])
-kernel.add_depends(["syscall.cpp.cog", "syscall.h.cog", "syscalls.inc.cog"], definitions)
+kernel.add_depends([
+        "syscall.cpp.cog",
+        "syscall.h.cog",
+        "syscalls.inc.cog",
+        "syscall_verify.cpp.cog",
+    ], definitions)
 
 kernel.variables['ldflags'] = ["${ldflags}", "-T", "${source_root}/src/kernel/kernel.ld"]

src/kernel/syscall.cpp.cog

@@ -21,9 +21,39 @@ ctx = Context(definitions_path)
 ctx.parse("syscalls.def")
 syscalls = ctx.interfaces['syscalls']
 
-cog.outl(f"constexpr size_t num_syscalls = {len(syscalls.methods)};")
 ]]]*/
 /// [[[end]]]
+
+namespace syscalls
+{
+    /*[[[cog code generation
+
+    last_scope = None
+    for id, scope, method in syscalls.methods:
+        if scope != last_scope:
+            cog.outl()
+            last_scope = scope
+
+        if scope:
+            name = f"{scope.name}_{method.name}"
+        else:
+            name = method.name
+
+        args = []
+        if method.constructor:
+            args.append("j6_handle_t *handle")
+        elif not method.static:
+            args.append("j6_handle_t handle")
+
+        for param in method.params:
+            for type, suffix in param.type.c_names(param.options):
+                args.append(f"{type} {param.name}{suffix}")
+
+        cog.outl(f"""j6_status_t _syscall_verify_{name} ({", ".join(args)});""")
+    ]]]*/
+    //[[[end]]]
+}
+
 uintptr_t syscall_registry[num_syscalls] __attribute__((section(".syscall_registry")));
 
 void
@@ -44,8 +74,8 @@ syscall_initialize()
         else:
             name = method.name
 
-        cog.outl(f"syscall_registry[{id}] = reinterpret_cast<uintptr_t>(syscalls::{name});")
-        cog.outl(f"""log::debug(logs::syscall, "Enabling syscall {id:x} as {name}");""")
+        cog.outl(f"syscall_registry[{id}] = reinterpret_cast<uintptr_t>(syscalls::_syscall_verify_{name});")
+        cog.outl(f"""log::debug(logs::syscall, "Enabling syscall {id:02x} as {name}");""")
         cog.outl("")
     ]]]*/
     //[[[end]]]

src/kernel/syscall.h.cog

@@ -13,48 +13,9 @@ ctx = Context(definitions_path)
 ctx.parse("syscalls.def")
 syscalls = ctx.interfaces["syscalls"]
 
+cog.outl(f"constexpr size_t num_syscalls = {len(syscalls.methods)};")
 ]]]*/
 /// [[[end]]]
 
-enum class syscall : uint64_t
-{
-    /*[[[cog code generation
-
-    for id, scope, method in syscalls.methods:
-        if scope:
-            name = f"{scope.name}_{method.name}"
-        else:
-            name = method.name
-        cog.outl(f"{name:20} = {id},")
-
-    ]]]*/
-    //[[[end]]]
-};
-
 void syscall_initialize();
 extern "C" void syscall_enable();
-
-namespace syscalls
-{
-/*[[[cog code generation
-
-for id, scope, method in syscalls.methods:
-    if scope:
-        name = f"{scope.name}_{method.name}"
-    else:
-        name = method.name
-
-    args = []
-    if method.constructor:
-        args.append("j6_handle_t *handle")
-    elif not method.static:
-        args.append("j6_handle_t handle")
-
-    for param in method.params:
-        for type, suffix in param.type.c_names(param.options):
-            args.append(f"{type} {param.name}{suffix}")
-
-    cog.outl(f"""j6_status_t {name} ({", ".join(args)});""")
-]]]*/
-//[[[end]]]
-}

src/kernel/syscall_verify.cpp.cog

@@ -0,0 +1,73 @@
+// vim: ft=cpp
+
+#include <stdint.h>
+#include <arch/memory.h>
+#include <j6/errors.h>
+#include <j6/types.h>
+
+namespace {
+    template <typename T>
+    __attribute__((always_inline))
+    inline bool check_refparam(const T* param, bool optional) {
+        return reinterpret_cast<uintptr_t>(param) < arch::kernel_offset &&
+            (optional || param);
+    }
+}
+
+namespace syscalls {
+
+/*[[[cog code generation
+from definitions.context import Context
+
+ctx = Context(definitions_path)
+ctx.parse("syscalls.def")
+syscalls = ctx.interfaces["syscalls"]
+
+cbool = {True: "true", False: "false"}
+
+for id, scope, method in syscalls.methods:
+    if scope:
+        name = f"{scope.name}_{method.name}"
+    else:
+        name = f"{method.name}"
+
+    args = []
+    argdefs = []
+    refparams = []
+
+    if method.constructor:
+        argdefs.append("j6_handle_t *handle")
+        args.append("handle")
+        refparams.append(("handle", False))
+    elif not method.static:
+        argdefs.append("j6_handle_t handle")
+        args.append("handle")
+
+    for param in method.params:
+        for type, suffix in param.type.c_names(param.options):
+            arg = f"{param.name}{suffix}"
+            argdefs.append(f"{type} {arg}")
+            args.append(arg)
+
+        if param.refparam:
+            subs = param.type.c_names(param.options)
+            refparams.append((param.name + subs[0][1], param.optional))
+            if param.outparam:
+                for sub in subs[1:]:
+                    refparams.append((param.name + sub[1], param.optional))
+
+    cog.outl(f"""extern j6_status_t {name} ({", ".join(argdefs)});""")
+    cog.outl(f"""j6_status_t _syscall_verify_{name} ({", ".join(argdefs)}) {{""")
+
+    for pname, optional in refparams:
+        cog.outl(f"    if (!check_refparam({pname}, {cbool[optional]}))")
+        cog.outl( "        return j6_err_invalid_arg;")
+        cog.outl()
+
+    cog.outl(f"""    return syscalls::{name}({", ".join(args)});""")
+    cog.outl("}")
+    cog.outl("\n")
+]]]*/
+//[[[end]]]
+
+} // namespace syscalls

src/kernel/syscalls/endpoint.cpp

@@ -29,7 +29,9 @@ endpoint_send(j6_handle_t handle, uint64_t tag, const void * data, size_t data_l
 j6_status_t
 endpoint_receive(j6_handle_t handle, uint64_t * tag, void * data, size_t * data_len, uint64_t timeout)
 {
-    if (!tag || !data_len || (*data_len && !data))
+    // Data is marked optional, but we need the length, and if length > 0,
+    // data is not optional.
+    if (!data_len || (*data_len && !data))
         return j6_err_invalid_arg;
 
     endpoint *e = get_handle<endpoint>(handle);
@@ -46,7 +48,7 @@ endpoint_receive(j6_handle_t handle, uint64_t * tag, void * data, size_t * data_
 j6_status_t
 endpoint_sendrecv(j6_handle_t handle, uint64_t * tag, void * data, size_t * data_len, uint64_t timeout)
 {
-    if (!tag || (*tag & j6_tag_system_flag))
+    if (*tag & j6_tag_system_flag)
         return j6_err_invalid_arg;
 
     endpoint *e = get_handle<endpoint>(handle);

src/kernel/syscalls/object.cpp

@@ -13,9 +13,6 @@ namespace syscalls {
 j6_status_t
 kobject_koid(j6_handle_t handle, j6_koid_t *koid)
 {
-    if (koid == nullptr)
-        return j6_err_invalid_arg;
-
     kobject *obj = get_handle<kobject>(handle);
     if (!obj)
         return j6_err_invalid_arg;

src/kernel/syscalls/system.cpp

@@ -19,9 +19,6 @@ namespace syscalls {
 j6_status_t
 log(const char *message)
 {
-    if (message == nullptr)
-        return j6_err_invalid_arg;
-
     thread &th = thread::current();
     log::info(logs::syscall, "Message[%llx]: %s", th.koid(), message);
     return j6_status_ok;
@@ -36,23 +33,24 @@ noop()
 }
 
 j6_status_t
-system_get_log(j6_handle_t sys, void *buffer, size_t *size)
+system_get_log(j6_handle_t sys, void *buffer, size_t *buffer_len)
 {
-    if (!size || (*size && !buffer))
+    // Buffer is marked optional, but we need the length, and if length > 0,
+    // buffer is not optional.
+    if (!buffer_len || (*buffer_len && !buffer))
         return j6_err_invalid_arg;
 
-    size_t orig_size = *size;
-    *size = g_logger.get_entry(buffer, *size);
+    size_t orig_size = *buffer_len;
+    *buffer_len = g_logger.get_entry(buffer, *buffer_len);
     if (!g_logger.has_log())
         system::get().deassert_signal(j6_signal_system_has_log);
 
-    return (*size > orig_size) ? j6_err_insufficient : j6_status_ok;
+    return (*buffer_len > orig_size) ? j6_err_insufficient : j6_status_ok;
 }
 
 j6_status_t
 system_bind_irq(j6_handle_t sys, j6_handle_t endp, unsigned irq)
 {
-    // TODO: check capabilities on sys handle
     endpoint *e = get_handle<endpoint>(endp);
     if (!e) return j6_err_invalid_arg;
 
@@ -65,9 +63,6 @@ system_bind_irq(j6_handle_t sys, j6_handle_t endp, unsigned irq)
 j6_status_t
 system_map_phys(j6_handle_t handle, j6_handle_t * area, uintptr_t phys, size_t size, uint32_t flags)
 {
-    // TODO: check capabilities on sys handle
-    if (!area) return j6_err_invalid_arg;
-
     // TODO: check to see if frames are already used? How would that collide with
     // the bootloader's allocated pages already being marked used?
     if (!(flags & vm_flags::mmio))
@@ -82,7 +77,6 @@ system_map_phys(j6_handle_t handle, j6_handle_t * area, uintptr_t phys, size_t s
 j6_status_t
 system_request_iopl(j6_handle_t handle, unsigned iopl)
 {
-    // TODO: check capabilities on sys handle
     if (iopl != 0 && iopl != 3)
         return j6_err_invalid_arg;
 

src/kernel/syscalls/vm_area.cpp

@@ -56,9 +56,6 @@ vma_unmap(j6_handle_t handle, j6_handle_t proc)
 j6_status_t
 vma_resize(j6_handle_t handle, size_t *size)
 {
-    if (!size)
-        return j6_err_invalid_arg;
-
     vm_area *a = get_handle<vm_area>(handle);
     if (!a) return j6_err_invalid_arg;