On 13.2.2019 23.13, Ivan Mironov wrote:
> Hi all,
> 
> There is a known problem on some MacBooks: internal card reader
> disappears after the first suspend/resume and all subsequent attempts
> to suspend laptop are failing.
> 
> This was reported[1][2] and even discussed in the mailing lists[3],
> without any real solution. After trying various things[4], including
> some existing quirks, I discovered that switching link state to DISABLED
> before suspend both fixes the card reader device and allows any subsequent
> suspend to succeed.
> 
> First patch adds code for this new quirk, and second patch enables this
> quirk for card reader device which is used in my macbook.
> 
> I'm not really familiar with either USB standards or kernel code to
> support them, so this patch series is RFC. I'm especially unsure with the
> "resume" part, because I implemented it by trial and error mostly.
> However, I'm using kernel with these patches and it works for me.
> 
> Also, feel free to suggest other kernel patches or existing quirks or
> quirk combinations to fix the same problem.
> 
> Oh, and by the way: I've checked schematics of various macbooks available
> on the internet. It seems that the actual chip is Genesys Logic GL3219,
> probably just with the custom ID. What I found curious, is that USB 2.0
> pins of this chip (D+ and D-) are not really connected anywhere, but
> instead shorted through the resistor. Could it be possible that this
> somehow messes up some logic inside the device, host controller or
> linux kernel?
> 

This card reader prevents second system suspend on latest kernels, see thread:
https://marc.info/?l=linux-usb&m=154816680816246&w=2

In that case the card reader fails to resume from usb3 U3 suspend state,
and ends up stuck in USB3 polling state, which now prevents suspend

Could you try a testpatch (attached) to see if it helps?

Thanks
-Mathias