On Sun, 18 May 2014, Mark Cave-Ayland wrote:
On 18/05/14 00:18, BALATON Zoltan wrote:
- fword("boot-device");
- boot_device = pop_fstr_copy();
- if (strcmp(boot_device, "disk") == 0) {
What is boot-device normally set to if no override is specified? pop_fstr_copy() has a little gotcha in that a zero length string becomes a NULL on conversion, so generally I tend to use a pattern like this:
I've tested it and the user can specify an empty string by not giving any value "-prom-env boot-device=" but this does not lead to a crash because the NULL value is only used in strcmp where it's not equal and in free where it does nothing. So you could apply this patch as is (with deobfuscating my email address in the commit message) unless you want me to do some other changes.
Regards, BALATON Zoltan
if (boot_device && strcmp(boot_device, "disk") == 0) { ... }
I'm just wondering if we shouldn't always assume that a value exists and is non-zero...
switch (fw_cfg_read_i16(FW_CFG_BOOT_DEVICE)) {case 'c':boot_path = "hd";break;default:case 'd':boot_path = "cd";break;}snprintf(buf, sizeof(buf), "%s:,\\\\:tbxi %s:,\\ppc\\bootinfo.txt%s:,%%BOOT", boot_path, boot_path, boot_path);
push_str(buf);fword("encode-string");push_str("boot-device");fword("property"); }- free(boot_device);