Unlocking flash problem with the epxa10db device

Christopher Hoover ch at murgatroid.com
Fri Sep 20 00:52:14 EDT 2002

I wrote cfi_intelext_varsize_frob().  It is behaving as intended.

You have to unlock/lock/erase/etc. an *entire* block/sector at a time.  

There's no way to unlock 4 bytes and leave the other (erasesize - 4)
bytes locked.  So the function returns EINVAL if the size isn't an even
"multiple" of the block size(s).  ("Multiple" here is not meant strictly
in the mathematical sense; consider variably-sized blocks.)

You'll find the other functions behave identically, e.g. erase.


-----Original Message-----
From: linux-arm-kernel-admin at lists.arm.linux.org.uk
[mailto:linux-arm-kernel-admin at lists.arm.linux.org.uk] On Behalf Of
Louie Boczek
Sent: Thursday, September 19, 2002 8:11 PM
To: ARM Linux kernel Mailinglist
Subject: Unlocking flash problem with the epxa10db device

I just started working with the Altera Excalibur board and ran into the 
following problem.

I was not able to write to the flash and found the problem lied in the 
unlocking code.  
In the epxa10db-flash.c:epxa10db_mtd_init() routine it calls:

       mymtd->unlock(mymtd, mymtd->eraseregions[i].offset + j * 
mymtd->eraseregions[i].erasesize, 4);

which eventually calls "cfi_cmdsetc_001.c:cfi_intelext_varsize_frob()" 
in this routine there is a check to make sure that the region being 
unlocked is the whole region with  the check:
        if ((ofs + len) & (regions[i].erasesize -1))
            return -EINVAL;

this fails since the call was only to unlock 4 bytes not erasesize.  So 
I fixed the problem by changing the call to mymtd->unlock()  to pass the

erasesize instead of 4.

But the question is, should "cfi_intelext_varsize_frob()" check to see 
if the whole region is specifed or should it unlock the region that the 
specified location is located.  I think the later.

I am using the linux-2.4.18-rmk7 but I also checked linux-2.4.19 and the

problem is also there.

I am new to this, I hope I gave enough and thorough informartion.


Subscription options:

More information about the linux-mtd mailing list