2006-05-12 23:31:37

by Greg KH

Subject: sdhci needs card to be present when loading module.

In the 2.6.17-rc4 kernel (and 2.6.17-rc1), on my laptop, if you load the
sdhci driver with no SD card in the slot, it never seems to be able to
detect the insertion of a new card later on.

However, if I load the module with a card present. Removing it and then
plugging it (or another one) in later seems to work just fine.

Is this expected?

Any kernel log messages I can provide to help with this?

thanks,

greg k-h

2006-05-13 15:25:43

by Pierre Ossman

Subject: Re: [Sdhci-devel] sdhci needs card to be present when loading module.

Greg KH wrote:
> In the 2.6.17-rc4 kernel (and 2.6.17-rc1), on my laptop, if you load the
> sdhci driver with no SD card in the slot, it never seems to be able to
> detect the insertion of a new card later on.
>
> However, if I load the module with a card present. Removing it and then
> plugging it (or another one) in later seems to work just fine.
>
> Is this expected?
>

No, this is a bug (hardware probably as I cannot find anything in the
specs that can even be remotely related). The chances of getting any
help from the manufacturer are slim at best. I've already contacted all
the major ones and requested erratas, but so far not a single reply.

> Any kernel log messages I can provide to help with this?

Not really, no. Since I assume this controller works fine in Windows,
I'm guessing it needs a very particular init sequence. It might have
something to do with the ios calls that are made when the slot is empty.

If you have the time, you could try adding some code that will make the
driver ignore ios calls until a card is inserted.

Rgds
Pierre

2006-05-13 15:31:49

by Marcel Holtmann

Subject: Re: [Sdhci-devel] sdhci needs card to be present when loading module.

Hi Greg,

> In the 2.6.17-rc4 kernel (and 2.6.17-rc1), on my laptop, if you load the
> sdhci driver with no SD card in the slot, it never seems to be able to
> detect the insertion of a new card later on.
>
> However, if I load the module with a card present. Removing it and then
> plugging it (or another one) in later seems to work just fine.

I had the same problem with my Ricoh controller (in an IBM X41) and the
latest sdhci driver from mainline. Try the audit patch from Pierre, but
be careful. For me it freezes the machine on eject, because of an
endless loop waiting for some hardware state to change.

Regards

Marcel