[PATCH v3] arm64: Add purgatory printing

[Geoff Levand]

Hi Matthias,

On 9/30/20 11:04 AM, Matthias Brugger wrote:
> On 30/09/2020 17:20, Geoff Levand wrote:
>> On 9/30/20 3:36 AM, matthias.bgg at kernel.org wrote:
>>> --- a/kexec/arch/arm64/include/arch/options.h
>>> +++ b/kexec/arch/arm64/include/arch/options.h
>>>
>>> +static uint64_t find_purgatory_sink(const char *console)
>>> +{
>>> +    int fd, ret;
>>> +    char device[255], mem[255];
>>> +    struct stat sb;
>>> +    char buffer[10];
>>> +    uint64_t iomem = 0x0;
>>> +
>>> +    if (!console)
>>> +        return 0;
>>> +
>>> +    ret = snprintf(device, sizeof(device), "/sys/class/tty/%s", console);
>>> +    if (ret < 0 || ret >= sizeof(device)) {
>>> +        fprintf(stderr, "snprintf failed: %s\n", strerror(errno));
>>> +        return 0;
>>> +    }
>>> +
>>> +    if (stat(device, &sb) || !S_ISDIR(sb.st_mode)) {
>>> +        fprintf(stderr, "kexec: %s: No valid console found for %s\n",
>>> +            __func__, device);
>>> +        return 0;
>>> +    }
>>> +
>>> +    snprintf(mem, sizeof(mem), "%s%s", device, "/iomem_base");
>>> +    if (ret < 0 || ret >= sizeof(mem)) {
>>
>> Seems this is is just doing the same check again...
>>
> 
> Well we have to check snprintf outcome here again. Also sizeof(device) == sizeof(mem) we could have the case where the first snprintf fits into the 255 character, while the second, where we add the "/iomem_base" does not.
> 
> Or didn't I understand you correctly?

You forgot to assign ret a new value, so it's just doing the same thing again...

-Geoff