On 8/27/19 1:27 PM, Frediano Ziglio wrote:
From: Victor Toso <me@xxxxxxxxxxxxxx>
Otherwise we get a CLANG_WARNING due accessing garbage.
Covscan report:
> spice-vdagent-0.19.0/src/vdagent/vdagent.c:471:9: warning: 1st function
> call argument is an uninitialized value
> # execvp(orig_argv[0], orig_argv);
> # ^ ~~~~~~~~~~~~
> spice-vdagent-0.19.0/src/vdagent/vdagent.c:421:24: note: Storing
> uninitialized value
> # char **orig_argv = g_memdup(argv, sizeof(char*) * (argc+1));
> # ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> spice-vdagent-0.19.0/src/vdagent/vdagent.c:434:9: note: Assuming 'error'
> is equal to NULL
> # if (error != NULL) {
> # ^~~~~~~~~~~~~
> spice-vdagent-0.19.0/src/vdagent/vdagent.c:434:5: note: Taking false
> branch
> # if (error != NULL) {
> # ^
> spice-vdagent-0.19.0/src/vdagent/vdagent.c:442:9: note: Assuming 'portdev'
> is not equal to NULL
> # if (portdev == NULL)
> # ^~~~~~~~~~~~~~~
> spice-vdagent-0.19.0/src/vdagent/vdagent.c:442:5: note: Taking false
> branch
> # if (portdev == NULL)
> # ^
> spice-vdagent-0.19.0/src/vdagent/vdagent.c:445:9: note: Assuming
> 'vdagentd_socket' is not equal to NULL
> # if (vdagentd_socket == NULL)
> # ^~~~~~~~~~~~~~~~~~~~~~~
> spice-vdagent-0.19.0/src/vdagent/vdagent.c:445:5: note: Taking false
> branch
> # if (vdagentd_socket == NULL)
> # ^
> spice-vdagent-0.19.0/src/vdagent/vdagent.c:448:30: note: Assuming
> 'do_daemonize' is 0
> # openlog("spice-vdagent", do_daemonize ? LOG_PID : (LOG_PID |
> LOG_PERROR),
> # ^~~~~~~~~~~~
> spice-vdagent-0.19.0/src/vdagent/vdagent.c:448:30: note: '?' condition is
> false
> spice-vdagent-0.19.0/src/vdagent/vdagent.c:451:9: note: Assuming the
> condition is false
> # if (!g_file_test(portdev, G_FILE_TEST_EXISTS)) {
> # ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> spice-vdagent-0.19.0/src/vdagent/vdagent.c:451:5: note: Taking false
> branch
> # if (!g_file_test(portdev, G_FILE_TEST_EXISTS)) {
> # ^
> spice-vdagent-0.19.0/src/vdagent/vdagent.c:457:9: note: Assuming
> 'do_daemonize' is 0
> # if (do_daemonize)
> # ^~~~~~~~~~~~
> spice-vdagent-0.19.0/src/vdagent/vdagent.c:457:5: note: Taking false
> branch
> # if (do_daemonize)
> # ^
> spice-vdagent-0.19.0/src/vdagent/vdagent.c:468:9: note: Assuming
> 'version_mismatch' is not equal to 0
> # if (version_mismatch) {
> # ^~~~~~~~~~~~~~~~
> spice-vdagent-0.19.0/src/vdagent/vdagent.c:468:5: note: Taking true branch
> # if (version_mismatch) {
> # ^
> spice-vdagent-0.19.0/src/vdagent/vdagent.c:471:9: note: 1st function call
> argument is an uninitialized value
> # execvp(orig_argv[0], orig_argv);
> # ^ ~~~~~~~~~~~~
> # 469| syslog(LOG_INFO, "Version mismatch, restarting");
> # 470| sleep(1);
> # 471|-> execvp(orig_argv[0], orig_argv);
> # 472| }
> # 473|
Signed-off-by: Victor Toso <victortoso@xxxxxxxxxx>
---
src/vdagent/vdagent.c | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/src/vdagent/vdagent.c b/src/vdagent/vdagent.c
index 0e2e73e..982fc72 100644
--- a/src/vdagent/vdagent.c
+++ b/src/vdagent/vdagent.c
@@ -418,7 +418,11 @@ int main(int argc, char *argv[])
GOptionContext *context;
GError *error = NULL;
VDAgent *agent;
- char **orig_argv = g_memdup(argv, sizeof(char*) * (argc+1));
+ char **orig_argv;
+
+ g_return_val_if_fail(argc > 0 && argv != NULL, -1);
+ orig_argv = g_memdup(argv, sizeof(char*) * (argc+1));
Hi,
I was able to "fix" it by replacing g_memdup with g_malloc + memcpy
- char **orig_argv = g_memdup(argv, sizeof(char*) * (argc+1));
+ char **orig_argv = g_malloc(sizeof(char*) * (argc+1) );
+ memcpy(orig_argv, argv, sizeof(char*) * (argc+1) );
So clang seems to be confused by "side effects" of g_memdup.
Uri.
+ orig_argv[argc] = NULL;
context = g_option_context_new(NULL);
g_option_context_add_main_entries(context, entries, NULL);
I would say better to disable Clang test instead. The code is perfectly
fine. argc is at least 1 (the executable name!) and argv is always terminated
with NULL (that's the standard!).
See https://clang-analyzer.llvm.org/faq.html.
I don't know where this -1 come, but EXIT_FAILURE (which is usually 1) is the standard
return for main function.
Frediano
_______________________________________________
Spice-devel mailing list
Spice-devel@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/spice-devel
_______________________________________________
Spice-devel mailing list
Spice-devel@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/spice-devel