[ovs-dev] [PATCH] ofp-parse: Fix parsing of register values 2**31 and greater.

[Ethan Jackson]

Just reviewed it, looks good, thanks.

Ethan

On Fri, Aug 12, 2011 at 15:02, Ethan Jackson <ethan at nicira.com> wrote:
> That fixes it, thanks.
>
> Ethan
>
> On Fri, Aug 12, 2011 at 15:00, Ben Pfaff <blp at nicira.com> wrote:
>> Reported-by: Ethan Jackson <ethan at nicira.com>
>> ---
>> Ethan, does this fix the problem that you reported?
>>
>> diff --git a/lib/ofp-parse.c b/lib/ofp-parse.c
>> index e6a6af1..5c7feb2 100644
>> --- a/lib/ofp-parse.c
>> +++ b/lib/ofp-parse.c
>> @@ -812,14 +812,20 @@ parse_field_value(struct cls_rule *rule, enum field_index index,
>>  static void
>>  parse_reg_value(struct cls_rule *rule, int reg_idx, const char *value)
>>  {
>> -    uint32_t reg_value, reg_mask;
>> +    /* This uses an oversized destination field (64 bits when 32 bits would do)
>> +     * because some sscanf() implementations truncate the range of %i
>> +     * directives, so that e.g. "%"SCNi16 interprets input of "0xfedc" as a
>> +     * value of 0x7fff.  The other alternatives are to allow only a single
>> +     * radix (e.g. decimal or hexadecimal) or to write more sophisticated
>> +     * parsers. */
>> +    unsigned long long int reg_value, reg_mask;
>>
>>     if (!strcmp(value, "ANY") || !strcmp(value, "*")) {
>>         cls_rule_set_reg_masked(rule, reg_idx, 0, 0);
>> -    } else if (sscanf(value, "%"SCNi32"/%"SCNi32,
>> +    } else if (sscanf(value, "%lli/%lli",
>>                       &reg_value, &reg_mask) == 2) {
>>         cls_rule_set_reg_masked(rule, reg_idx, reg_value, reg_mask);
>> -    } else if (sscanf(value, "%"SCNi32, &reg_value)) {
>> +    } else if (sscanf(value, "%lli", &reg_value)) {
>>         cls_rule_set_reg(rule, reg_idx, reg_value);
>>     } else {
>>         ovs_fatal(0, "register fields must take the form <value> "
>> --
>> 1.7.2.5
>>
>>
>