Re: [PATCH 05/27] DDF: Implement store_super_ddf

[NeilBrown <neilb@suse.de>]

[-- Attachment #1: Type: text/plain, Size: 4106 bytes --]

On Wed, 10 Jul 2013 19:37:24 +0200 Martin Wilck <mwilck@arcor.de> wrote:

> Hi Neil,
> 
> >> This patch implements the previously unsupported case where
> >> store_super_ddf is called with a non-empty superblock.
> >>
> >> For DDF, writing meta data to just one disk makes no sense.
> >> We would run the risk of writing inconsistent meta data
> >> to the devices. So just call __write_init_super_ddf and
> >> write to all devices, including the one passed by the caller.
> >>
> >> This patch assumes that the device to store the superblock on
> >> has already been added to the DDF structure. Otherwise, an
> >> error message will be emitted.
> >>
> >> Signed-off-by: Martin Wilck <mwilck@arcor.de>
> > 
> > I'm not sure I really like this.  If mdadm is calling ->store_super, then it
> > wants to write to just one device.  It will quite possibly call
> > ->store_super on all devices, and with your change that will write to all
> > devices multiple times.
> > So maybe a different interface is needed?
> 
> The point is the sequence number handling. If the sequence number
> increases, it would be dangerous to write only one superblock.

->store_super should not change the sequence number (I think).  Its purpose
is to fix up the metadata in some way, typically so that a failed array
doesn't look like it is failed any more, so it can be assembled.
It should really just write back exactly the same superblock except for
whatever changes were explicitly requested via "update_super" (plus checksums
of course).

> 
> But perhaps it can be done otherwise, I need to examine that further.
> First I need a clear understanding what this API is used for.
> 
> > What is the big-picture effect of this patch?  What mdadm function now works
> > that didn't work before?
> 
> To be honest, I don't know. I came up with this patch when scanning
> through the code looking for possible problems. This is not necessary to
> make mdadm -e ddf -C -l10 work.

I've reverted it then - I really don't think it is wanted.

Thanks,
NeilBrown


> 
> Martin
> 
> > 
> > Thanks,
> > NeilBrown
> > 
> > 
> > 
> >> ---
> >>  super-ddf.c |   41 +++++++++++++++++++++++++++++++++++------
> >>  1 files changed, 35 insertions(+), 6 deletions(-)
> >>
> >> diff --git a/super-ddf.c b/super-ddf.c
> >> index b806949..b3c846d 100644
> >> --- a/super-ddf.c
> >> +++ b/super-ddf.c
> >> @@ -3471,15 +3471,44 @@ static int store_super_ddf(struct supertype *st, int fd)
> >>  	if (!ddf)
> >>  		return 1;
> >>  
> >> -	/* ->dlist and ->conflist will be set for updates, currently not
> >> -	 * supported
> >> -	 */
> >> -	if (ddf->dlist || ddf->conflist)
> >> -		return 1;
> >> -
> >>  	if (!get_dev_size(fd, NULL, &dsize))
> >>  		return 1;
> >>  
> >> +	if (ddf->dlist || ddf->conflist) {
> >> +		struct stat sta;
> >> +		struct dl *dl;
> >> +		int ofd, ret;
> >> +
> >> +		if (fstat(fd, &sta) == -1 || !S_ISBLK(sta.st_mode)) {
> >> +			pr_err("%s: file descriptor for invalid device\n",
> >> +			       __func__);
> >> +			return 1;
> >> +		}
> >> +		for (dl = ddf->dlist; dl; dl = dl->next)
> >> +			if (dl->major == (int)major(sta.st_rdev) &&
> >> +			    dl->minor == (int)minor(sta.st_rdev))
> >> +				break;
> >> +		if (!dl) {
> >> +			pr_err("%s: couldn't find disk %d/%d\n", __func__,
> >> +			       (int)major(sta.st_rdev),
> >> +			       (int)minor(sta.st_rdev));
> >> +			return 1;
> >> +		}
> >> +		/*
> >> +		   For DDF, writing to just one disk makes no sense.
> >> +		   We would run the risk of writing inconsistent meta data
> >> +		   to the devices. So just call __write_init_super_ddf and
> >> +		   write to all devices, including this one.
> >> +		   Use the fd passed to this function, just in case dl->fd
> >> +		   is invalid.
> >> +		 */
> >> +		ofd = dl->fd;
> >> +		dl->fd = fd;
> >> +		ret =  __write_init_super_ddf(st);
> >> +		dl->fd = ofd;
> >> +		return ret;
> >> +	}
> >> +
> >>  	if (posix_memalign(&buf, 512, 512) != 0)
> >>  		return 1;
> >>  	memset(buf, 0, 512);
> > 


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 828 bytes --]