[elrepo] CentOS/Kernel 3.18/Thunderbolt2

Israel Brewster israel at ravnalaska.net
Mon Jan 26 18:22:49 EST 2015


On Jan 26, 2015, at 2:10 PM, Alan Bartlett <ajb at elrepo.org> wrote:

> On 26 January 2015 at 22:54, Israel Brewster <israel at ravnalaska.net> wrote:
>> On Jan 26, 2015, at 11:58 AM, Alan Bartlett <ajb at elrepo.org> wrote:
>>> 
>>> Hmm . . . I've taken a look at the Kconfig file and the section
>>> relevant to the kernel configuration that references "Thunderbolt"
>>> reads as follows --
>>> 
>>> [quote]
>>> menuconfig THUNDERBOLT
>>>   tristate "Thunderbolt support for Apple devices"
>>>   depends on PCI
>>>   select CRC32
>>>   help
>>>     Cactus Ridge Thunderbolt Controller driver
>>>     This driver is required if you want to hotplug Thunderbolt devices on
>>>     Apple hardware.
>>> 
>>>     Device chaining is currently not supported.
>>> 
>>>     To compile this driver a module, choose M here. The module will be
>>>     called thunderbolt.
>>> [/quote]
>>> 
>>> The references to "Apple devices" and "Apple hardware" makes me think
>>> that is not the correct configuration choice.
>>> 
>>> I can certainly "turn it on", as a module but I have my doubts as to
>>> whether it will help.
>> 
>> It looks like you might be right, but perhaps not for the reasons you were thinking. The key phrase in this information, to my mind, is: "Cactus Ridge Thunderbolt Controller driver". Cactus Ridge is the code name for Intel's Thunderbolt controllers that accompanied Ivy Bridge systems (http://ark.intel.com/products/codename/57097/Cactus-Ridge). It's *possible* that that specific chip was only used in Apple hardware (certainly Apple hardware is the primary user of thunderbolt controllers), but I don't know of any reason why that would be the case *necessarily*.
>> 
>> That said, the chip in my system (and, in fact, any thunderbolt 2 system I am aware of at the moment) is a "Falcon Ridge" controller, released Q3 2013 according to the Wikipedia article. As such, I also have my doubts that the "Cactus Ridge" controller driver will help. It's possible the Linux kernel simply doesn't support thunderbolt 2 (yet?).
>> 
>> On the other hand, there is this: http://permalink.gmane.org/gmane.linux.kernel/1717419 which seems to indicate a patch to make things work with the Falcon Ridge controller was available back in June of last year. I think. It also references Apple hardware, but the controller is the same. And this: http://www.gossamer-threads.com/lists/linux/kernel/1935522 which seems to indicate that simply adding some ID's to the Cactus Ridge driver makes it work with thunderbolt2 (Falcon Ridge) controllers - so perhaps the help simply hasn't been updated?
>> 
>> I suppose, however, that it is also possible that if I want to use the thunderbolt card, I'll have to run windows. Which would stink. Maybe there is a Linux kernel mailing list I should be asking on?
> 
> The more I look, the more it seems to be a somewhat "oddball" device
> interface. A short while ago I was passed a link to a web page [4] by
> Greg Kroah-Hartman.

Yeah, I read that article a while ago, and pretty much dismissed it. For one thing, it completely misses the point of Thunderbolt as essentially external PCI - way more capable than just hard drives (although, admittedly, that is the primary use it has been put to). For another, the benchmarks he refers to only show the limitations in performance of a single drive, not of the interface - there is actually a note at the end stating this. Once you start using a RAID (with a fast enough controller) or multiple drives at the same time, thunderbolt -especially thunderbolt2 - can easily surpass USB3 performance. See a little later in the article where they admitted that to make the dual USB 3.0 RAID 0 keep up with the thunderbolt RAID 0, they had to use two independent USB 3.0 ports, which of course gave the USB 3 interfaces a combined equivalent throughput of the single thunderbolt 1 interface.

But, of course, that is all beside the point. Either the Linux kernel can work with thunderbolt, or it can't. Unfortunately it seems difficult to find the answer :-)

> 
> As for a mailing list for all things kernel, I can point you to the
> Linux Kernel Mailing List. [5]

Thanks.

-----------------------------------------------
Israel Brewster
Systems Analyst II
Ravn Alaska
5245 Airport Industrial Rd
Fairbanks, AK 99709
(907) 450-7293
-----------------------------------------------

> 
> Alan.
> 
> [4] http://www.kroah.com/log/linux/
> [5] https://lkml.org/
> _______________________________________________
> elrepo mailing list
> elrepo at lists.elrepo.org
> http://lists.elrepo.org/mailman/listinfo/elrepo



More information about the elrepo mailing list