misc question about MTD

misc question about MTD

[Jim Zeus]

Hi, folks

when I am reading the code of MTD, I meet some questions :
1.in cfi_probe_chip(cfi_probe.c), what is an 'alias ' of a flash chip? is that the same flash chip mapping in different address?
2.if I send Query command (98h, I mean) twice, would the 'QRY' disappear? and if I send it once, can I read the 'QRY' for several times? (in my deduce , the answer of these two questions should be 'yes')

Thanks

Jim Zeus



_____________________________________________________________
Want a new web-based email account ? ---> http://www.firstlinux.net

_____________________________________________________________
Promote your group and strengthen ties to your members with email@yourgroup.org by Everyone.net  http://www.everyone.net/?btn=tag

Re: misc question about MTD

[David Woodhouse]

zeusj@firstlinux.net said:
> 1.in cfi_probe_chip(cfi_probe.c), what is an 'alias ' of a flash chip?
> is that the same flash chip mapping in different address? 

Yes. Often, people will connect flash chips without all the address lines 
being used, so they can swap in larger chips later. Then you get these 
'aliases'.

> 2.if I send Query command (98h, I mean) twice, would the 'QRY' disappear?

Er, I don't know. I don't think so, but if this _is_ required then it could 
be useful -- currently we have to use a chip-specific command to put chips 
back in read mode when we do the alias detection.

> and if I send it once, can I read the 'QRY' for several times?

Yes, you can.



--
dwmw2

Re: misc question about MTD

[Jim Zeus]

Hi,David

  thank you for your explanation, I have got what you said 

>and if _that_ has 'Q' 'R' 'Y' for data you assume
>they're the same chip, otherwise not.

I think its just "assume" , If the Chip A and Chip B got QRY on the same location for data ,they will be still regarded as same chip,  and that's the "FIXME" in your code means,isnt it:)

   I think before mount the flash should be erased automatically ,but  when I mount my flash (the mtd0 and mtdblock0 are all ok), I got the following message:
>jffs_scan_flash(): Did not find even a single chunk of free space. This is BAD!
>jffs_scan_flash(): Leaving...
>jffs_scan_falsh():Free size accounting screwed
>jfffs_scan_flash():free_chunk_size1 == 0x800000, free_chunk_size2 == 0x0, fmc->f
ree_size == 0x0
Why the flash is still dirty?

and when I was reading code of jffs_scan_flash , I meet the "flipping bits", you know , in the if-else phrase to detect flipping bits:
>if((((__u32)(pos - start)) != 0)){
>....
>}else{
>/* "Flipping bits" detected. This means that our scan for them
>   did not catch this offset. See check_partly_erased_sectors() for
>   more info.
>*/
I analysed all the case ,but I think that's impossible to enter the "else" bracket theoretically 
Did I miss something or there is something special for the hardware?

Thanks

Jim Zeus






_____________________________________________________________
Want a new web-based email account ? ---> http://www.firstlinux.net

_____________________________________________________________
Promote your group and strengthen ties to your members with email@yourgroup.org by Everyone.net  http://www.everyone.net/?btn=tag

Re: misc question about MTD

[David Woodhouse]

zeusj@firstlinux.net said:
>  I think its just "assume" , If the Chip A and Chip B got QRY on the
> same location for data ,they will be still regarded as same chip,  and
> that's the "FIXME" in your code means,isnt it:)

Yes.

>    I think before mount the flash should be erased automatically ,but
> when I mount my flash (the mtd0 and mtdblock0 are all ok), I got the
> following message:

Strange. Is the flash completely erased? It should not be erased 
automatically -- too many people attempt to mount the wrong flash 
partition, and if you mount the partition that contains your bootloader you 
will _not_ thank me for auto-erasing it.

I would strongly recommend you use JFFS2, not JFFS. I haven't looked at JFFS
code for ages now, and I don't think anyone else is doing so either.

> I analysed all the case ,but I think that's impossible to enter the
> "else" bracket theoretically  Did I miss something or there is
> something special for the hardware?

Something special for the hardware. If you power cycle a chip while it's 
erasing, you may find that some bits become undefined -- sometimes they'll 
return a '1', other times a '0'. 


--
dwmw2

Re: misc question about MTD

[Jim Zeus]

Hi, David

--- David Woodhouse <dwmw2@infradead.org> wrote:

>Strange. Is the flash completely erased? It should not be erased 
>automatically -- too many people attempt to mount the wrong flash 
>partition, and if you mount the partition that contains your bootloader you 
>will _not_ thank me for auto-erasing it.

Yes, its ok now , I erase the flash before mount it , then its ok:)


>I would strongly recommend you use JFFS2, not JFFS. I haven't looked at JFFS
>code for ages now, and I don't think anyone else is doing so either.

It is the second time you *strongly* recommend me use JFFS2:) 
ok, I tried it just now: I erased the flash, and I mount it(I do all these just like mount JFFS),but I got the message that I have meet before (in the last board I ported):
jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00020024: 0x2f0c instead
jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00020028: 0x0001 instead
Further such events for this erase block will not be printed
JFFS2: Erase block at 0x00020000 is not formatted. It will be erased

these messages above are my old friend, and I got some new ones:

Cowardly refusing to erase blocks on filesystem with no valid JFFS2 nodes

so what does that means? I am losing the confidence to JFFS2 now:(

And when I am debugging the flash, I find something strange,you know, I wrote some debug message when debugging so when I "cp /usr/somefile /dev/mtd0",I got this:
# cp /usr/somefile /dev/mtd0
MTD_open
ZEUS:begin of mtd_open
ZEUS:begin of get_mtd_device
ZEUS:begin of __get_mtd_device
ZEUS:end of __get_mtd_device
ZEUS:end of get_mtd_device
ZEUS:end of mtd_open
MTD_write
ZEUS:begin of mtd_write
ZEUS:begin of cfi_amdstd_write
ZEUS:end of cfi_amdstd_write
ZEUS:end of mtd_write
MTD_close
ZEUS:begin of mtd_close
ZEUS:begin of put_mtd_device
ZEUS:end of put_mtd_device
ZEUS:end of mtd_close

the "begin of" and "end of" message is printed at the begin and end of functions,
when I delete the printk() in cfi_amdstd_write() to prevent the "ZEUS:begin/end of cfi_amdstd_write" to be shown(I just do it, nothing else), I cant accomplish the same thing that I have done before:
# cp /usr/somefile /dev/mtd0
MTD_open
ZEUS:begin of mtd_open
ZEUS:begin of get_mtd_device
ZEUS:begin of __get_mtd_device
ZEUS:end of __get_mtd_device
ZEUS:end of get_mtd_device
ZEUS:end of mtd_open
MTD_write
ZEUS:begin of mtd_write
Last[3] is 9da5, datum is 6246
/dev/mtd0: I/O error
MTD_close
ZEUS:begin of mtd_close
ZEUS:begin of put_mtd_device
ZEUS:end of put_mtd_device
ZEUS:end of mtd_close
#

I know the "Last[3] is 9da5, datum is 6246" message is to tell me that what should be written is wrongly different from what be read from flash,its a sanity check,but why a printk() can affects the Flash write?

Thanks in advance

Jim Zeus


_____________________________________________________________
Want a new web-based email account ? ---> http://www.firstlinux.net

_____________________________________________________________
Promote your group and strengthen ties to your members with email@yourgroup.org by Everyone.net  http://www.everyone.net/?btn=tag

Re: misc question about MTD

[David Woodhouse]

zeusj@firstlinux.net said:
>  I erased the flash, and I mount it(I do all these just like mount
> JFFS),but I got the message that I have meet before (in the last board
> I ported): 
> jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00020024: 0x2f0c instead 
> jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00020028: 0x0001 instead

The flash is not completely erased. 

> Cowardly refusing to erase blocks on filesystem with no valid JFFS2
> nodes 

The flash is not completely erased, and doesn't contain _anything_ that 
looks like valid JFFS2 data, so we don't erase it and start using it, 
because you might have mounted the wrong partition.

> I know the "Last[3] is 9da5, datum is 6246" message is to tell me that
> what should be written is wrongly different from what be read from
> flash,its a sanity check,but why a printk() can affects the Flash
> write? 

Timing? You definitely have flash driver problems. 




--
dwmw2