[Cake] [sch_cake] Merge branch 'exposeinterval' (5541271)

Sebastian Moeller moeller0 at gmx.de
Mon Oct 12 08:06:51 EDT 2015


Hi Dave,

On Oct 12, 2015, at 14:04 , Dave Taht <dave.taht at gmail.com> wrote:

> sigh.
> 
> In looking this code over again, interval is a 64 bit value in nsec in
> our version of codel, and needs no conversion into usec. It also seems
> to imply that all the conversion values are wrong throughout the code,
> and the api should expose a 64 bit value.

	But tc will only pass in an unsigned worth of microseconds, so there needs to be a conversion to nsecs, no? Isn’t MS2TIME doing that conversion for us, so maybe we need US2TIME to handle time values passed from tc (so interval and potentially target).

Best Regards
	Sebastian

> 
> On Mon, Oct 12, 2015 at 1:49 PM, Sebastian Moeller <moeller0 at gmx.de> wrote:
>> Hi Dave,
>> 
>> 
>> On Oct 12, 2015, at 13:16 , Dave Taht <dave.taht at gmail.com> wrote:
>> 
>>> um, no, interval needs to be kept in usec, and converted more like I did.
>> 
>>        I have not looked at your code yet. In tc (tc_utils.h or so) there is get_time() that knows how to deal with the default time specifiers like sec ms us… I believe from the tc side this should be used to parse the user input, just like Kevin’s code does. Now this needs to be passed to sch_cake.c, where I believe Kevin’s version expects milliseconds (MS2TIME(q->interval)), which does not do the right thing if the user specified “rtf 123usecs” on tic’s cli… So just getting rid of the MS2TIME called should do the trick, no?
>> 
>> Best Regards
>>        Sebastian
>> 
>>> 
>>> On Mon, Oct 12, 2015 at 2:36 AM, Dave Taht <dave.taht at gmail.com> wrote:
>>>> On Mon, Oct 12, 2015 at 11:28 AM, Jonathan Morton <chromatix99 at gmail.com> wrote:
>>>>> 
>>>>>> On 12 Oct, 2015, at 12:24, Dave Taht <dave.taht at gmail.com> wrote:
>>>>>> 
>>>>>> This is incorrect for the DC use case. We should be able to have the
>>>>>> target fall below at least 250us.
>>>>> 
>>>>> Hey, I’m trying to work on this myself, but the code keeps changing under me.
>>>> 
>>>> Tough. Deal with it. :) Once the rtt option works right toke and I can
>>>> finally go test the
>>>> count/2 idea out properly in the testbed, like, today. You otherwise
>>>> have the pen,
>>>> as far as I am concerned.
>>>> 
>>>>> 
>>>>> - Jonathan Morton
>>>>> 
>>>> 
>>>> 
>>>> 
>>>> --
>>>> Dave Täht
>>>> Do you want faster, better, wifi? https://www.patreon.com/dtaht
>>> 
>>> 
>>> 
>>> --
>>> Dave Täht
>>> Do you want faster, better, wifi? https://www.patreon.com/dtaht
>> 
> 
> 
> 
> -- 
> Dave Täht
> Do you want faster, better, wifi? https://www.patreon.com/dtaht




More information about the Cake mailing list