Re: [RFC PATCH v2 2/2] mtd: rawnand: meson: add support for Amlogic NAND flash controller

[Boris Brezillon]

On Wed, 29 Aug 2018 18:29:05 +0800
Liang Yang <liang.yang@xxxxxxxxxxx> wrote:

> On 8/29/2018 6:08 PM, Liang Yang wrote:
> > 
> > On 8/28/2018 9:26 PM, Boris Brezillon wrote:  
> >> On Tue, 28 Aug 2018 21:21:48 +0800
> >> Liang Yang <liang.yang@xxxxxxxxxxx> wrote:
> >>  
> >>> Hi Boris,
> >>>
> >>> On 8/24/2018 8:48 PM, Boris Brezillon wrote:  
> >>>> On Wed, 22 Aug 2018 22:08:42 +0800
> >>>> Liang Yang <liang.yang@xxxxxxxxxxx> wrote:  
> >>>>>> You have to wait tWB, that's for sure.  
> >>>>> we have a maximum 32 commands fifo. when command is written into
> >>>>> NFC_REG_CMD, it doesn't mean that command is executing right now, 
> >>>>> maybe
> >>>>> it is buffering on the queue.Assume one ERASE operation, when 2nd
> >>>>> command(0xd0) is written into NFC_REG_CMD and then come into
> >>>>> NAND_OP_WAITRDY_INSTR, if I read the RB status by register, it may be
> >>>>> wrong because 0xd0 may not being executed. it is unusual unless
> >>>>> buffering two many command.  
> >>>>
> >>>> You should flush the queue and wait for it to empty at the end of  
> >>>> ->exec_op().
> >>>>> so it seems that i still need to use nand_soft_waitrdy or wait cmd is
> >>>>> executed somewhere.  
> >>>>
> >>>> Don't you have a WAIT_FOR_RB instruction? What is NFC_CMD_RB for? Also,
> >>>> NFC_CMD_IDLE seems to allow you to add an arbitrary delay, and that's
> >>>> probably what you should use for tWB.
> >>>>
> >>>> em, I can wait for RB by reading the status from register now. but when  
> >>> calling nand_soft_waitrdy, i really met a problem. One *jiffies* is
> >>> about 4ms. When programming, it pass 1ms to
> >>> instr->ctx.waitrdy.timeout_ms and nand_soft_waitrdy will be only one
> >>> *jiffies* to reach timeout. And then calling nand_soft_waitrdy maybe at
> >>> the tail of 4ms interval, it may only wait 100us and next jiffies
> >>> arrive. Is it correct?  
> >>
> >> Hm, no. If you initialize the time you compare to (using time_before()
> >> or time_after()) correctly it should not happen. Anyway, I keep thinking
> >> this is not how it should be done. Did you try NFC_CMD_RB? Did you ask
> >> HW designers what it was created for?
> >>  
> > I am using NFC_CMD_RB and checking with irq. it is ok now.  
> there are two usages for NFC_CMD_RB. One reads the data status 
> continuously by hardware after sending 0x70 command; the other checks 
> the r/b IO status continuously.both can send irq when r/b is ready.

Both should do what you expect, so I guess you're good.