Re: [RFC PATCH v2 1/5] mtd: cfi_cmdset_0002: Add support for polling status register

[Joakim Tjernlund <Joakim.Tjernlund@xxxxxxxxxxxx>]

On Mon, 2019-03-25 at 22:36 +0530, Vignesh Raghavendra wrote:
> 
> On 25/03/19 7:21 PM, Joakim Tjernlund wrote:
> > On Mon, 2019-03-25 at 18:27 +0530, Vignesh Raghavendra wrote:
> > > CAUTION: This email originated from outside of the organization. Do not click links or open attachments unless you recognize the sender and know the content is safe.
> > > 
> > > 
> > > Hi,
> > > 
> > > On 21/03/19 11:41 PM, Joakim Tjernlund wrote:
> > > > On Thu, 2019-03-21 at 23:15 +0530, Vignesh Raghavendra wrote:
> > > > > HyperFlash devices are compliant with CFI AMD/Fujitsu Extended Command
> > > > > Set(0x0002) for flash operations, therefore drivers/mtd/chips/cfi_cmdset_0002.c
> > > > > can be use as is. But these devices do not support DQ polling method of
> > > > > determining chip ready/good status. These flashes provide Status
> > > > > Register whose bits can be polled to know status of flash operation.
> > > > > 
> > > > > Cypress HyperFlash datasheet here[1], talks about CFI Amd/Fujitsu
> > > > > Extended Query version 1.5. Bit 0 of "Software Features supported" field
> > > > > of CFI Primary Vendor-Specific Extended Query table indicates
> > > > > presence/absence of status register and Bit 1 indicates whether or not
> > > > > DQ polling is supported. Using these bits, its possible to determine
> > > > > whether flash supports DQ polling or need to use Status Register.
> > > > > 
> > > > > Add support for polling status register to know device ready/status of
> > > > > erase/write operations when DQ polling is not supported.
> > > > 
> > > > Isn't this new Status scheme just a copy of Intels(cmdset_0001)?
> > > 
> > > Yes, but with one difference: At the end of program/erase operation,
> > > device directly enters status register mode and  starts reflecting
> > > status register content at any address.
> > > The device remains in the read status register state until another
> > > command is written to the device. Therefore there is notion of device is
> > > in "status register read mode" (FL_STATUS) state
> > 
> > That seems to vary and long time ago RMK added this:
> > 		/* If the flash has finished erasing, then 'erase suspend'
> > 		 * appears to make some (28F320) flash devices switch to
> > 		 * 'read' mode.  Make sure that we switch to 'read status'
> > 		 * mode so we get the right data. --rmk
> > 		 */
> > 		map_write(map, CMD(0x70), chip->in_progress_block_addr);
> > 
> 
> This behavior is expected with cmdset_0001. Because "The device remains
> in the read status register state until another command is written",
> therefore "erase suspend' command after erase completion will switch
> device to read mode. And therefore read status is safe thing to do for
> cmdset_0001.
> 
> But in case of cmdset_0002 erase completion will not put device to read
> status mode and therefore no special status tracking is required.
> 
> > > But in case of cfi_cmdset_0002, once program/erase operation is
> > > complete, device returns to previous address space overlay from which
> > > operation was started from (mostly read mode)
> > 
> > I hope you can do the same as Intel here, issue an explicit Status CMD or you will be in trouble.
> 
> Even if we issue Read Status command to enter read status mode, any
> single subsequent read will put device back to read mode. So, sending
> explicit Status CMD is of not much use.
> 
> As long as cmdset_0002 driver ensures sending Read Status cmd and next
> single read can be done in one go (ie. mutex held), I don't see any
> trouble here. This is already take care off.

Ouch, a non sticky Status sounds borken. Are you sure that nothing can change the
chip between you issue the Status CMD and read out of status bits?
Like if an erase/suspend/resume completes just after Status CMD but before Status readout?

 Jocke

> 
> > Also, I think you need to use the various map_word_xxx as in:
> > status = map_read(map, chip->in_progress_block_addr);
> > if (map_word_andequal(map, status, status_OK, status_OK))
> > 	break;
> 
> Yes, I will fixup this patch to use map_word_* wherever necessary in the
> next revision.
> 
> 
> > otherwise you will break interleaved setups(like two 8-bit flashes in parallel to
> > form one 16 bit bus). Maybe this is not supported for CMDSET 0002 ?
> > 
> 
> Interleaved is indeed supported by cmdset_0002. Thanks for pointing that
> out!
> 
> >  Jocke
> > 
> > > In order to enter status register overlay mode, Read Status command is
> > > to be written to addr_unlock1(0x555) address. The overlay is in effect
> > > for one read access, specifically the next read access that follows the
> > > Status Register Read command
> > > Therefore code around FL_STATUS state in cfi_cmdset_0001 is not
> > > applicable to cfi_cmdset_0002 as is.
> > > 
> > > 
> > > > If so I think the new status impl. in 0002 should borrow from 0001 as this is a
> > > > hardened and battle tested impl.
> > > > 
> > > 
> > > In case of cfi_cmdset_0001.c, program/erase is followed by
> > > inval_cache_and_wait_for_operation() to poll ready bit and based on
> > > status register value, success or the error handling is done.
> > > 
> > > Most of the code corresponding to inval_cache_and_wait_for_operation()
> > > is already in cfi_cmdset_0002.c. So, whats missing in this patch is
> > > handling and reporting of errors as reflected in status register after
> > > write/erase failures. I will add that in the next version.
> > > 
> > > But, I don't see much to borrow apart from error handling sequence.
> > > Please, let me know if I missed something.
> > > 
> > > > I know other modern 0002 chips supports both old and new impl. of Status and I world
> > > > guess that we will see more chips with new Status only.
> > > > 
> > > 
> > > Agreed. Newer devices would mostly be CFI 1.5.
> > > 
> > > --
> > > Regards
> > > Vignesh

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/