RE: [PATCH v2] sdio: skip initialization on powered resume

[Bing Zhao <bzhao@marvell.com>]

Hi Sahitya,

> -----Original Message-----
> From: Sahitya Tummala [mailto:stummala@codeaurora.org]
> Sent: Tuesday, September 14, 2010 3:16 AM
> To: Bing Zhao
> Cc: Nicolas Pitre; linux-mmc@vger.kernel.org; Michal Miroslaw; Chris Ball; Andrew Morton
> Subject: RE: [PATCH v2] sdio: skip initialization on powered resume
> 
> Hi Bing, Chris,
> 
> On Tue, 2010-09-07 at 19:10 -0700, Bing Zhao wrote:
> > Hi Nicolas,
> >
> > > -----Original Message-----
> > > From: Nicolas Pitre [mailto:nico@fluxnic.net]
> > > Sent: Tuesday, September 07, 2010 6:29 PM
> > > To: Bing Zhao
> > > Cc: linux-mmc@vger.kernel.org; Michal Miroslaw; Chris Ball; Andrew Morton
> > > Subject: RE: [PATCH v2] sdio: skip initialization on powered resume
> > >
> > > On Tue, 7 Sep 2010, Bing Zhao wrote:
> > >
> > > > Thanks for the hint.
> > > >
> > > > The new patch skips reading CCCR, common CIS, and validation of
> > > > vendor/device IDs inside mmc_sdio_init_card() when powered_resume is
> > > > not zero.
> > >
> > > Why do you skip the reading of the CIS and IDs validation?  That's
> > > basically the main reason for still calling mmc_sdio_init_card().  And
> > > that only requires CMD 52 so that should be fine.
> >
> > While the system is suspended, the SDIO card could be in sleep mode (deep sleep or IEEE Power Save)
> as well. Reading CIS or any other CMD52 will fail if the card happens to be in sleep at this moment.
> If we skip re-initialization (including CCCR/CIS and IDs validation), mmc_sdio_init_card() returns
> success. Then the client driver's resume() handler will be called and the card can be woken up by
> client driver.
> 
> In one of your previous patches, "sdio: don't use CMD[357] as part of a
> powered SDIO resume", its mentioned that there is an issue with CMD7 in
> mmc_sdio_init_card() that is called during SDIO resume on Marvell 8686.
> 
> But in mmc_resume_host(), there is still a call to mmc_detect_change()
> which in turn calls mmc_sdio_detect(), thus sending CMD7 to the card.
> Does CMD7 to your card succeed after your client driver resume is
> called?

If CMD7 is sent _before_ client driver's resume handler is called, while 8686 card is in sleep mode, it will fail. If CMD7 is sent _after_ client driver's resume handler is called, it should succeed.

By the way, a patch "mmc: fix all hangs related to mmc/sd card insert/removal during suspend/resume" (4c2ef25fe0b847d2ae818f74758ddb0be1c27d8e by Maxim Levitsky) has removed the call to mmc_detect_change() in mmc_resume_host().

Regards,

Bing

> 
> Thanks,
> Sahitya.
> 
> --
> Sent by a consultant of the Qualcomm Innovation Center, Inc.
> The Qualcomm Innovation Center, Inc. is a member of the Code Aurora
> Forum.
>