Seriously overengineer the initialization of the env variables we pass
to loader so that we can set MALLOC_PERTURB_ to a random number. Is
this worth it - who can say?
---
loader/init.c | 63 ++++++++++++++++++++++++++++++++++++++++++++++++++ +++++++
1 files changed, 63 insertions(+), 0 deletions(-)

diff --git a/loader/init.c b/loader/init.c
index f5c3dbf..e79691f 100644
--- a/loader/init.c
+++ b/loader/init.c
@@ -455,6 +455,66 @@ static void setupBacktrace(void)
backtrace(&array, 1);
}

+static char *setupMallocPerturb(char *value)
+{
+ FILE *f;
+ unsigned char x;
+ size_t rc;
+ char *ret = NULL;
+
+ f = fopen("/dev/urandom", "r");
+ if (!f)
+ return NULL;
+
+ rc = fread(&x, 1, 1, f);
+ fclose(f);
+ if (rc < 1)
+ return NULL;
+
+ rc = asprintf(&ret, "MALLOC_PERTURB_=%hhu", x);
+ if (rc < 0)
+ return NULL;
+ return ret;
+}
+
+/* these functions return a newly allocated string that never gets freed;
+ * their lifetime is essentially that of main(), and we'd have to track which
+ * are allocated and which aren't, which is pretty pointless... */
+typedef char *(*setupEnvCallback)(char *entry);
+
+static void setupEnv(void)
+{
+ struct {
+ char *name;
+ setupEnvCallback cb;
+ } setupEnvCallbacks[] = {
+ { "MALLOC_PERTURB_", setupMallocPerturb },
+ { NULL, NULL }
+ };
+ int x;
+
+ /* neither array is very big, so this algorithm isn't so bad. If env[]
+ * gets bigger for some reason, we should probably just alphebatize both
+ * (manually) and then only initialize y one time.
+ */
+ for (x = 0; setupEnvCallbacks[x].name != NULL; x++) {
+ int y;
+ int l = strlen(setupEnvCallbacks[x].name) + 1;
+ char cmpstr[l + 1];
+
+ strncpy(cmpstr, setupEnvCallbacks[x].name, l);
+ strcat(cmpstr, "=");
+
+ for (y = 0; env[y] != NULL; y++) {
+ if (!strncmp(env[y], cmpstr, l)) {
+ char *new = setupEnvCallbacks[x].cb(env[y] + l);
+ if (new)
+ env[y] = new;
+ }
+ }
+ }
+}
+
int main(int argc, char **argv) {
pid_t installpid, childpid;
int waitStatus;
@@ -502,6 +562,9 @@ int main(int argc, char **argv) {
/* set up signal handler */
setupBacktrace();

+ /* set up any environment variables that aren't totally static */
+ setupEnv();
+
printstr("
Greetings.
");

printf("anaconda installer init version %s starting
", VERSION);
--
1.7.2.3

_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list

On Fri, Sep 24, 2010 at 04:21:40PM -0400, Peter Jones wrote:
> Seriously overengineer the initialization of the env variables we pass
> to loader so that we can set MALLOC_PERTURB_ to a random number. Is
> this worth it - who can say?
> ---

I really can't speak to its universal worth, but it does look worthy. So
I'll ack it.

--
Brian C. Lane / Anaconda Team
Port Orchard, WA (PST8PDT)
_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list