ASM3242 SuperSpeed+ speed reporting

ASM3242 SuperSpeed+ speed reporting

[James Ettle]

Hello,

I have a 20Gbps USB-C port on a plug-in card that uses an Asmedia 
ASM3242 controller. When I attach devices capable of 10Gbps or 20Gbps, 
lsusb -t reports they are connected at 5Gbps. My test devices are:

* 10Gbps USB hub
* RTL9210B based NVME enclosure, 10Gbps
* ASM236X based NVME enclosure, 20Gbps

Now when I do transfer tests on the NVME enclosures I get speeds too 
high for 5Gbps. In fact everything is consistent with them connecting at 
their best possible SSP speeds. So it looks like just the speed 
reporting is incorrect.

It seems this problem has been encountered before in the following 
thread, a bit over a year ago:

https://lore.kernel.org/all/CAKRa1U6rukBBwWx0vN2wqiHMbUH9zWsc-y1wknOCvhueqWsT7w@mail.gmail.com/

in which Thinh Nguyen posted the following patch:

https://lore.kernel.org/all/6908aa69-469b-8f92-8e19-60685f524f9c@synopsys.com/

I've applied this to 6.2.9 and now the speed are correctly reported in 
lsusb. Can I politely agitate for getting this patch revived and pushed 
upstream?

With thanks,
James Ettle.

Re: ASM3242 SuperSpeed+ speed reporting

[Thinh Nguyen]

Hi,

On Sun, Apr 02, 2023, James Ettle wrote:
> Hello,
> 
> I have a 20Gbps USB-C port on a plug-in card that uses an Asmedia ASM3242
> controller. When I attach devices capable of 10Gbps or 20Gbps, lsusb -t
> reports they are connected at 5Gbps. My test devices are:
> 
> * 10Gbps USB hub
> * RTL9210B based NVME enclosure, 10Gbps
> * ASM236X based NVME enclosure, 20Gbps
> 
> Now when I do transfer tests on the NVME enclosures I get speeds too high
> for 5Gbps. In fact everything is consistent with them connecting at their
> best possible SSP speeds. So it looks like just the speed reporting is
> incorrect.
> 
> It seems this problem has been encountered before in the following thread, a
> bit over a year ago:
> 
> https://urldefense.com/v3/__https://lore.kernel.org/all/CAKRa1U6rukBBwWx0vN2wqiHMbUH9zWsc-y1wknOCvhueqWsT7w@mail.gmail.com/__;!!A4F2R9G_pg!beAeIUnohpkAEs92iPBhg9xFF3wy_neCM8AbKOkCuo4Qo8w6XvED1IcSti7z5AQpMBmNXojdywa7NUorNyrk$
> 
> in which Thinh Nguyen posted the following patch:
> 
> https://urldefense.com/v3/__https://lore.kernel.org/all/6908aa69-469b-8f92-8e19-60685f524f9c@synopsys.com/__;!!A4F2R9G_pg!beAeIUnohpkAEs92iPBhg9xFF3wy_neCM8AbKOkCuo4Qo8w6XvED1IcSti7z5AQpMBmNXojdywa7NT-Vk7Cj$
> 
> I've applied this to 6.2.9 and now the speed are correctly reported in
> lsusb. Can I politely agitate for getting this patch revived and pushed
> upstream?
> 

It was just a testing patch to verify if the workaround works. It's not
kernel ready. A proper patch requires a new xhci quirk matching targeted
Asmedia host. If someone wants to take this diff suggestion and respin
it for a proper patch (with proper testing), I can help review it.

Thanks,
Thinh

Re: ASM3242 SuperSpeed+ speed reporting

[James Ettle]

Hello,

On 04/04/2023 01:07, Thinh Nguyen wrote:
> Hi,
> 
> On Sun, Apr 02, 2023, James Ettle wrote:
>> Hello,
>>
>> I have a 20Gbps USB-C port on a plug-in card that uses an Asmedia ASM3242
>> controller. When I attach devices capable of 10Gbps or 20Gbps, lsusb -t
>> reports they are connected at 5Gbps. My test devices are:
>>
>> * 10Gbps USB hub
>> * RTL9210B based NVME enclosure, 10Gbps
>> * ASM236X based NVME enclosure, 20Gbps
>>
>> Now when I do transfer tests on the NVME enclosures I get speeds too high
>> for 5Gbps. In fact everything is consistent with them connecting at their
>> best possible SSP speeds. So it looks like just the speed reporting is
>> incorrect.
>>
>> It seems this problem has been encountered before in the following thread, a
>> bit over a year ago:
>>
>> https://urldefense.com/v3/__https://lore.kernel.org/all/CAKRa1U6rukBBwWx0vN2wqiHMbUH9zWsc-y1wknOCvhueqWsT7w@mail.gmail.com/__;!!A4F2R9G_pg!beAeIUnohpkAEs92iPBhg9xFF3wy_neCM8AbKOkCuo4Qo8w6XvED1IcSti7z5AQpMBmNXojdywa7NUorNyrk$
>>
>> in which Thinh Nguyen posted the following patch:
>>
>> https://urldefense.com/v3/__https://lore.kernel.org/all/6908aa69-469b-8f92-8e19-60685f524f9c@synopsys.com/__;!!A4F2R9G_pg!beAeIUnohpkAEs92iPBhg9xFF3wy_neCM8AbKOkCuo4Qo8w6XvED1IcSti7z5AQpMBmNXojdywa7NT-Vk7Cj$
>>
>> I've applied this to 6.2.9 and now the speed are correctly reported in
>> lsusb. Can I politely agitate for getting this patch revived and pushed
>> upstream?
>>
> 
> It was just a testing patch to verify if the workaround works. It's not
> kernel ready. A proper patch requires a new xhci quirk matching targeted
> Asmedia host. If someone wants to take this diff suggestion and respin
> it for a proper patch (with proper testing), I can help review it.

Understood. Unfortunately it's beyond my ability to make such a patch 
myself. So this doesn't get lost, should I open a bug in Kernel 
Bugzilla, or is it tracked elsewhere?

Thanks,
-James

Re: ASM3242 SuperSpeed+ speed reporting

[Thinh Nguyen]

On Sat, May 06, 2023, James Ettle wrote:
> Hello,
> 
> On 04/04/2023 01:07, Thinh Nguyen wrote:
> > Hi,
> > 
> > On Sun, Apr 02, 2023, James Ettle wrote:
> > > Hello,
> > > 
> > > I have a 20Gbps USB-C port on a plug-in card that uses an Asmedia ASM3242
> > > controller. When I attach devices capable of 10Gbps or 20Gbps, lsusb -t
> > > reports they are connected at 5Gbps. My test devices are:
> > > 
> > > * 10Gbps USB hub
> > > * RTL9210B based NVME enclosure, 10Gbps
> > > * ASM236X based NVME enclosure, 20Gbps
> > > 
> > > Now when I do transfer tests on the NVME enclosures I get speeds too high
> > > for 5Gbps. In fact everything is consistent with them connecting at their
> > > best possible SSP speeds. So it looks like just the speed reporting is
> > > incorrect.
> > > 
> > > It seems this problem has been encountered before in the following thread, a
> > > bit over a year ago:
> > > 
> > > https://urldefense.com/v3/__https://lore.kernel.org/all/CAKRa1U6rukBBwWx0vN2wqiHMbUH9zWsc-y1wknOCvhueqWsT7w@mail.gmail.com/__;!!A4F2R9G_pg!beAeIUnohpkAEs92iPBhg9xFF3wy_neCM8AbKOkCuo4Qo8w6XvED1IcSti7z5AQpMBmNXojdywa7NUorNyrk$
> > > 
> > > in which Thinh Nguyen posted the following patch:
> > > 
> > > https://urldefense.com/v3/__https://lore.kernel.org/all/6908aa69-469b-8f92-8e19-60685f524f9c@synopsys.com/__;!!A4F2R9G_pg!beAeIUnohpkAEs92iPBhg9xFF3wy_neCM8AbKOkCuo4Qo8w6XvED1IcSti7z5AQpMBmNXojdywa7NT-Vk7Cj$
> > > 
> > > I've applied this to 6.2.9 and now the speed are correctly reported in
> > > lsusb. Can I politely agitate for getting this patch revived and pushed
> > > upstream?
> > > 
> > 
> > It was just a testing patch to verify if the workaround works. It's not
> > kernel ready. A proper patch requires a new xhci quirk matching targeted
> > Asmedia host. If someone wants to take this diff suggestion and respin
> > it for a proper patch (with proper testing), I can help review it.
> 
> Understood. Unfortunately it's beyond my ability to make such a patch
> myself. So this doesn't get lost, should I open a bug in Kernel Bugzilla, or
> is it tracked elsewhere?
> 

Sure. I think it's a good idea to track this somewhere.

Thanks,
Thinh