[DTrace-devel] [PATCH v2 2/4] dt_impl: bump number of TSLOTS to 8

Alan Maguire alan.maguire at oracle.com
Thu Jul 3 15:18:08 UTC 2025


On 02/07/2025 21:22, Eugene Loh wrote:
> On 7/2/25 10:52, Alan Maguire wrote:
> 
>> On 01/07/2025 19:31, Eugene Loh wrote:
>>> Very simple patch, but one key problem.  The changed line is preceded by
>>> a huge comment block that goes to excruciating pains to explain why the
>>> value should be 4.  So, I'm fine with the change itself and I think the
>>> "Because of the complexity..." sentence in the commit message can be
>>> dropped, but the comment block in the file needs to be updated to
>>> explain (with corresponding pains?) why we need 8 (not 9, not 7).
>>>
>> Ah good catch, forgot to do this. I actually think 6 is enough.
>> The additional seems to be a result of the various complex inet_ntoa*()
>> calls in the tcp.d translators; some have ternary operators with
>> multiple inet_ntoa*()s. The origin appears to be dt_cg_ternary_op() and
>> that has the comment:
>>
>>          /*
>>           * Strings complicate things a bit because dn_left and dn_right
>> might
>>           * actually be temporary strings (tstring) *and* in different
>> slots.
>>           * We need to allocate a new tstring to hold the result, and
>> copy the
>>           * value into the new tstring (and free any tstrings in
>> dn_left and
>>           * dn_right).
>>           */
>>
>> So if we have a left and right ternary and both allocate 3 tstring slots
>> for inet_ntoa*()s that means 3*2 = 6 tstring slots should actually be
>> enough. Additional testing suggests that is so; does the above make
>> sense or are there additional things to consider here?
> 
> Ha!  I do not know;  I prefer NOT to have a headache.  But maybe there
> should also be some tstring stress test using inet_ntoa*() and ternary ops.
>

sure, I can add some tests to operators/tst.ternary.d

>> I'll update the
>> next version with a comment describing the above anyhow. Thanks for the
>> review!
>> Alan
>>
>>
>>> On 6/10/25 09:58, Alan Maguire wrote:
>>>> Because of the complexity of the TCP translators, more tslots are
>>>> needed.
>>>>
>>>> Signed-off-by: Alan Maguire <alan.maguire at oracle.com>
>>>> ---
>>>>    libdtrace/dt_impl.h | 2 +-
>>>>    1 file changed, 1 insertion(+), 1 deletion(-)
>>>>
>>>> diff --git a/libdtrace/dt_impl.h b/libdtrace/dt_impl.h
>>>> index 68fb8ec5..10424f9c 100644
>>>> --- a/libdtrace/dt_impl.h
>>>> +++ b/libdtrace/dt_impl.h
>>>> @@ -218,7 +218,7 @@ typedef struct dt_kern_path {
>>>>     * - cleanpath() holds a prepended '/' char, a string, an appended
>>>> '/' char,
>>>>     *   and a terminating NUL char, or STRSZ + 3 chars altogether
>>>>     */
>>>> -#define DT_TSTRING_SLOTS    4
>>>> +#define DT_TSTRING_SLOTS    8
>>>>    #define DT_TSTRING_SIZE(dtp)    \
>>>>            MAX(P2ROUNDUP((dtp)->dt_options[DTRACEOPT_STRSIZE] + 3,
>>>> 8), \
>>>>                72)




More information about the DTrace-devel mailing list