Julien Cristau wrote:
> On Mon, Nov 29, 2010 at 20:57:46 -0800, Alan Coopersmith wrote:
>
>> diff --git a/hw/xfree86/loader/loadmod.c b/hw/xfree86/loader/loadmod.c
>> index 06d082b..751fec6 100644
>> --- a/hw/xfree86/loader/loadmod.c
>> +++ b/hw/xfree86/loader/loadmod.c
>> @@ -406,22 +406,22 @@ FindModuleInSubdir(const char *dirpath, const char
>> *module)
>>
>> snprintf(tmpBuf, PATH_MAX, "lib%s.so", module);
>> if (strcmp(direntry->d_name, tmpBuf) == 0) {
>> - ret = malloc(strlen(tmpBuf) + strlen(dirpath) + 1);
>> - sprintf(ret, "%s%s", dirpath, tmpBuf);
>> + if (Xasprintf(&ret, "%s%s", dirpath, tmpBuf) == -1)
>> + ret = NULL;
>> break;
>> }
>>
>> snprintf(tmpBuf, PATH_MAX, "%s_drv.so", module);
>> if (strcmp(direntry->d_name, tmpBuf) == 0) {
>> - ret = malloc(strlen(tmpBuf) + strlen(dirpath) + 1);
>> - sprintf(ret, "%s%s", dirpath, tmpBuf);
>> + if (Xasprintf(&ret, "%s%s", dirpath, tmpBuf) == -1)
>> + ret = NULL;
>> break;
>> }
>>
>> snprintf(tmpBuf, PATH_MAX, "%s.so", module);
>> if (strcmp(direntry->d_name, tmpBuf) == 0) {
>> - ret = malloc(strlen(tmpBuf) + strlen(dirpath) + 1);
>> - sprintf(ret, "%s%s", dirpath, tmpBuf);
>> + if (Xasprintf(&ret, "%s%s", dirpath, tmpBuf) == -1)
>> + ret = NULL;
>> break;
>> }
>> }
>
> Should these use XNFasprintf?
Certainly the previous code effectively did that (though by just SEGV on
writing to the NULL value returned by malloc, instead of an orderly abort),
but since the callers handle NULL return values already, it seemed like
returning failure was better than killing the server. Not that you can
get much farther into initialization if you run out of memory for module
path strings while just loading the modules.
--
-Alan Coopersmith- [email protected]
Oracle Solaris Platform Engineering: X Window System
_______________________________________________
[email protected]: X.Org development
Archives: http://lists.x.org/archives/xorg-devel
Info: http://lists.x.org/mailman/listinfo/xorg-devel
