[Patch 1/1] don't suspend erase for erase in cfi_cmdset_0002
Joakim Tjernlund
joakim.tjernlund at transmode.se
Mon Aug 3 08:25:47 EDT 2009
>
>
>
> >
> > >
> > >
> > > Hi,
> > >
> > > came around this problem while stress testing jffs2. From time to time the
> > > block erase failed and the file system overflows. I don't know if there
> > > are any Nor chips out there which allow a new erase to start when in erase
> > > suspend. However, the chips on my board dont't. And even when it doesn't
> > > make much sense to suspend an erase operation for another erase.
> > >
> > > Patch below fixes the problem for me.
> > >
> > > Have fun and take care,
> > >
> > > Peter
> > >
> > > Signed-off-by: Peter Wippich <pewi at gw-instruments.de>
> > > ---
> > > diff -Naur a/drivers/mtd/chips/cfi_cmdset_0002.c
> > > b/drivers/mtd/chips/cfi_cmdset_0002.c
> > > --- a/drivers/mtd/chips/cfi_cmdset_0002.c 2007-07-10 20:56:30.000000000 +0200
> > > +++ b/drivers/mtd/chips/cfi_cmdset_0002.c 2009-08-02 19:55:34.000000000 +0200
> > > @@ -521,6 +521,10 @@
> > > case FL_ERASING:
> > > if (mode == FL_WRITING) /* FIXME: Erase-suspend-program appears broken. */
> > > goto sleep;
> > > + if (mode == FL_ERASING) {
> > > + printk(KERN_INFO "attempt erase suspend with mode FL_ERASING\n") ;
> > > + goto sleep;
> > > + }
> > >
> > > if (!( mode == FL_READY
> > > || mode == FL_POINT
> > >
> >
> > This change looks bogus. You are not supposed to need the extra test.
> > I do think this test is faulty though:
> >
> > if (!( mode == FL_READY
> > || mode == FL_POINT
> > || !cfip
> > || (mode == FL_WRITING && (cfip->EraseSuspend & 0x2))
> > || (mode == FL_WRITING && (cfip->EraseSuspend & 0x1)
> > )))
> >
> > What happens if cfip is NULL?
>
> eeh, and what if erase suspend isn't supported at all?
> Take a hint from cfi_cmdset_0001.c
gaah, cfip->EraseSuspend & 0x1 appears to mean Read only. No wonder
there is a comment /* FIXME: Erase-suspend-program appears broken. */
Note, I don't use cmdset 0002 so I may be full of s...
>
> >
> > Jocke
More information about the linux-mtd
mailing list