[Ocfs2-tools-devel] [PATCH 3/3] mkfs.ocfs2: Add 64bit journal support.

[Joel Becker]

On Sun, Sep 07, 2008 at 01:53:02PM -0700, Mark Fasheh wrote:
> On Fri, Sep 05, 2008 at 11:36:42PM -0700, Joel Becker wrote:
> > The [no]block32 and [no]block64 journal options are added to mkfs -J.
> > noblock32 and block64 will create 64bit-capable journals.
> 
> Obviously, this patch will have to be updated with any incompat bit we add
> to the scheme.

	Yup.  Any thoughts?  I was just thinking a INCOMPAT_JBD2.  The
kernel only needs to know about the bit (jbd and jbd2 handle the journal
superblock compat checks properly).  Our tools will start supporting
INCOMPAT_JBD2 when they start rejecting unknown journal superblock bits.
Sound good?

> > +/* Returns true if we support these journal features */
> > +static int ocfs2_journal_check_available_features(journal_superblock_t *jsb,
> > +						  ocfs2_fs_options *features)
> > +{
> > +	if (!features->opt_compat && !features->opt_ro_compat &&
> > +	    !features->opt_incompat)
> > +		return 1;
> > +
> > +	if (jsb->s_header.h_blocktype != JBD2_SUPERBLOCK_V2)
> > +		return 0;
> 
> Ahh, this is clearly JFS_SUPERBLOCK_V2 renamed for JBD2.

	Yes,  _V1 didn't have compat bits, _V2 does.  I'm using the JBD2
defines for consistency with the kernel (they're fully compatible).

> So - I take it we've never written a V1 superblock on an Ocfs2 volume before?

	Not that I know of.  If we have, this check fails as it should
(V1 superblocks don't know any features).

> > diff --git a/mkfs.ocfs2/mkfs.h b/mkfs.ocfs2/mkfs.h
> > index b98b705..bd8ac45 100644
> > --- a/mkfs.ocfs2/mkfs.h
> > +++ b/mkfs.ocfs2/mkfs.h
> > @@ -212,6 +212,7 @@ struct _State {
> >  	unsigned int initial_slots;
> >  
> >  	uint64_t journal_size_in_bytes;
> > +	int journal64;
> >  
> >  	char *vol_label;
> >  	char *device_name;
> 
> Don't we need to update check_32bit_blocks()? The function altogether should be skipped
> if we have 64 bit blocks turned on. Also, in case we need to check, and the
> check fails, we can try updating the error text:

	You're right.  I worked on setting the bit without checking
where mkfs might validate the limit.  I was rushing out the quick first
pass before the weekend. :-)

> 	fprintf(stderr, "ERROR: jbd can only store block numbers in 32 bits. "
> 		"%s can hold %"PRIu64" blocks which overflows this limit. If "
> 		"you have a new enough Ocfs2 with JBD2 support, you can try "
> 		"formatting with the \"-Jblock64\" option to turn on support "
> 		"for this size block device.\n"
> 		"Otherwise, consider increasing the block size or decreasing the device "
> 		"size.\n", s->device_name, s->volume_size_in_blocks);

	Sounds good.

Joel

-- 

Life's Little Instruction Book #337

	"Reread your favorite book."

Joel Becker
Principal Software Developer
Oracle
E-mail: joel.becker at oracle.com
Phone: (650) 506-8127