Re: [PATCH v3 1/2] ubi: provide a way to skip CRC checks

[Richard Weinberger]

Am Montag, 2. Juli 2018, 09:51:21 CEST schrieb Boris Brezillon:
> Hi Artem,
> 
> On Mon, 02 Jul 2018 10:30:25 +0300
> Artem Bityutskiy <dedekind1@xxxxxxxxx> wrote:
> 
> > Hi,
> > 
> > On Thu, 2018-06-28 at 09:40 +0200, Quentin Schulz wrote:
> > > diff --git a/drivers/mtd/ubi/kapi.c b/drivers/mtd/ubi/kapi.c
> > > index d4b2e87..e9e9ecb 100644
> > > --- a/drivers/mtd/ubi/kapi.c
> > > +++ b/drivers/mtd/ubi/kapi.c
> > > @@ -202,7 +202,7 @@ struct ubi_volume_desc *ubi_open_volume(int ubi_num, int vol_id, int mode)
> > >  	desc->mode = mode;
> > >  
> > >  	mutex_lock(&ubi->ckvol_mutex);
> > > -	if (!vol->checked) {
> > > +	if (!vol->checked && !vol->skip_check) {
> > >  		/* This is the first open - check the volume */
> > >  		err = ubi_check_volume(ubi, vol_id);
> > >  		if (err < 0) {  
> > 
> > Did you deliberately did not add a similar check to 'vol_cdev_write()' ?
> > You want to skip checking on load but do have the checking after volume update ?
> 
> Yep, it's on purpose, I asked Quentin to keep the test on the update
> volume path.
> 
> > Looks a bit inconsistent to me. At the very least deserves a comment in
> > 'vol_cdev_write()' about why 'skip_check' flag is ignored there.
> 
> Well, I thought checking the CRC just after updating the volume made
> sense, just to make sure things were written correctly on the medium.
> Let's add a comment explaining why we keep the check here, unless you
> see a strong reason to get rid of this check in the update path.

+1

I also vote for keeping the check.
vol->skip_check is really just "skip the check upon volume open".

Thanks,
//richard