Kernel oops after sync command in jffs2

[Frederic Giasson]

Hi,

I think I found a possible source for the oops I reported previously this
week.

When the oops happen, the do_erase_oneblock() function if cfi_cmdset_0002.c
complains that DQ5 has been asserted to "1" while the flash was still
erasing, meaning that a program command has been attempted.  

So I examined the chip driver to see if there were a gap in the mutual
exclusion of the chip access.  I found that function cfi_amdstd_write() send
unlock bypass command to the flash without verifying if the flash is ready
first, if cfi->fast_prog is true.  I tryied to disable fast_prog in
cfi_probe.c, and the oops did not happen again.  It sound like it were the
problem, although I am not 100% sure yet.

I will send you the oops trace along with the callback trace if it ever
shows up again.

Frédéric Giasson

|-----Original Message-----
|From: David Woodhouse [mailto:dwmw2 at infradead.org]
|Sent: Thursday, May 16, 2002 1:35 PM
|To: Frederic Giasson
|Cc: 'linux-mtd at lists.infradead.org'
|Subject: Re: Kernel oops after sync command in jffs2 
|
|
|
|Show oops? Decoded.
|
|--
|dwmw2
|
|