Re: [PATCH] mmc: core: Do not hold re-tuning during CMD6 commands

[Chaotian Jing]

On Fri, 2017-03-24 at 09:52 +0200, Adrian Hunter wrote:
> On 24/03/17 08:19, Chaotian Jing wrote:
> > this patch is refine for 'commit c6dbab9cb58f ("mmc: core: Hold re-tuning
> > during switch commands")'
> > Since it has 3 retries at max for CMD6, if the first CMD6 got CRC error,
> > then should do re-tune before the next CMD6 was sent.
> > 
> > Signed-off-by: Chaotian Jing <chaotian.jing@xxxxxxxxxxxx>
> > ---
> >  drivers/mmc/core/mmc_ops.c |    3 +--
> >  1 file changed, 1 insertion(+), 2 deletions(-)
> > 
> > diff --git a/drivers/mmc/core/mmc_ops.c b/drivers/mmc/core/mmc_ops.c
> > index fe80f26..6931927 100644
> > --- a/drivers/mmc/core/mmc_ops.c
> > +++ b/drivers/mmc/core/mmc_ops.c
> > @@ -534,8 +534,6 @@ int __mmc_switch(struct mmc_card *card, u8 set, u8 index, u8 value,
> >  	bool use_r1b_resp = use_busy_signal;
> >  	unsigned char old_timing = host->ios.timing;
> >  
> > -	mmc_retune_hold(host);
> > -
> >  	/*
> >  	 * If the cmd timeout and the max_busy_timeout of the host are both
> >  	 * specified, let's validate them. A failure means we need to prevent
> > @@ -567,6 +565,7 @@ int __mmc_switch(struct mmc_card *card, u8 set, u8 index, u8 value,
> >  		cmd.sanitize_busy = true;
> >  
> >  	err = mmc_wait_for_cmd(host, &cmd, MMC_CMD_RETRIES);
> > +	mmc_retune_hold(host);
> 
> That is not how mmc_retune_hold() works, you need mmc_retune_hold_now() as
> it is here:
> 
> 	https://marc.info/?l=linux-mmc&m=148940903816582
> 
> But using "retries" with commands that have busy-waiting on the data line
> doesn't make much sense anyway.  Particularly with CRC errors, I would
> expect the card is actually busily doing the switch and we need only to wait
> for it.  The same can be true for timeout errors.  For some CMD6 we might
> need to send CMD12 if the card is busy after an error.  I would prefer an
> explicit attempt at recovery from CMD6 errors.
> 

It's the host driver's responsibility to ensure card is not in busy
state before issue the next R1B command, or the MMC core layer needs do
extra check/waiting before issue a R1B command.
I think the purpose of "re-tune" is trying to cover particular case(eg.
voltage fluctuate or EMI or some glitch of host/device which caused CRC
error) , but in such cases, too many cases are disable re-tune function
by mmc_retune_hold(), for example, in this case, if a response CRC error
got then we never have chance to recover it. then cause system cannot
access emmc or suspend/resume fail.