Hi John,
In NetworkInterface.c:
162 ret = (MIB_IFROW *) malloc(sizeof(MIB_IFROW));
163 if(ret == NULL)
164 return NULL;
tableP is left dangling if line 164 runs, no?
222 if (ret != NO_ERROR) {
223 if (tableP != NULL) {
224 free(tableP);
225 JNU_ThrowByName(env, "java/lang/Error",
226 "IP Helper Library GetIfTable function failed");
227 JNU_ThrowOutOfMemoryError(env, "Native heap allocation
failure");
228 return -1;
229 }
How do you know it's an OOM here?
289 curr = (netif *)calloc(1, sizeof(netif));
290 if (curr != NULL) {
291 wlen = MultiByteToWideChar(CP_OEMCP, 0, ifrowP->bDescr,
292 ifrowP->dwDescrLen, NULL, 0);
293 if(wlen == 0) {
294 // MultiByteToWideChar should not fail
295 // But in rare case it fails, we allow 'char' to be
displayed
296 curr->displayName = (char *)malloc(ifrowP->dwDescrLen +
1);
297 if(curr->displayName == NULL)
298 return -1;
299
300 } else {
301 curr->displayName = (char
*)malloc(wlen*(sizeof(wchar_t))+1);
302 if(curr->displayName == NULL)
303 return -1;
304 }
Before returning, doesn't curr itself need to be freed? Same thing in this
block:
308 if (curr->name == NULL || curr->displayName == NULL) {
309 if (curr->name) free(curr->name);
310 if (curr->displayName) free(curr->displayName);
311 curr = NULL;
312 }
In NetworkInterface_winXP.c:
154 if(ret == NULL) {
155 JNU_ThrowByName(env, "java/lang/OutOfMemoryError", 0);
156 return NULL;
157 }
adapterInfo needs free here?
I also noticed some ptrs are compared against NULL while others against 0 -
why the inconsistency?
Also, might it be cleaner to use multiple chained goto's to handle cleanup
in functions with multiple allocations? The early returns in some of these
places are a recipe for leaks I think. I'm thinking something like:
r1 = malloc(...);
if (r1 == NULL) goto out1;
r2 = malloc(...);
if (r2 == NULL) goto out2;
out2:
free(r2);
out1:
free(r1);
Thanks
Sent from my phone
On Feb 5, 2013 6:42 PM, "John Zavgren" <john.zavg...@oracle.com> wrote:
> Greetings:
> I modified the use of malloc() and realloc() so that return values are
> checked and potential memory leaks and data corruption problems are
> prevented.
>
> http://cr.openjdk.java.net/~jzavgren/8007606/webrev.01/
>
> Thanks!
> John Zavgren
>
