Re: [PATCH 5/7] hfsplus: stop using ioctl_by_bdev

[Christoph Hellwig <hch@lst.de>]

On Mon, May 04, 2020 at 10:16:40AM -0600, Jens Axboe wrote:
> On 4/25/20 1:57 AM, Christoph Hellwig wrote:
> >  	if (HFSPLUS_SB(sb)->session >= 0) {
> > +		struct cdrom_tocentry te;
> > +
> > +		if (!cdi)
> > +			return -EINVAL;
> > +
> >  		te.cdte_track = HFSPLUS_SB(sb)->session;
> >  		te.cdte_format = CDROM_LBA;
> > -		res = ioctl_by_bdev(sb->s_bdev,
> > -			CDROMREADTOCENTRY, (unsigned long)&te);
> > -		if (!res && (te.cdte_ctrl & CDROM_DATA_TRACK) == 4) {
> > -			*start = (sector_t)te.cdte_addr.lba << 2;
> > -			return 0;
> > +		if (cdrom_read_tocentry(cdi, &te) ||
> > +		    (te.cdte_ctrl & CDROM_DATA_TRACK) != 4) {
> > +			pr_err("invalid session number or type of track\n");
> > +			return -EINVAL;
> >  		}
> 
> I must be missing something obvious from just looking over the patches,
> but how does this work if cdrom is modular and hfsplus is builtin?

In that case disk_to_cdi will return NULL as it uses IS_REACHABLE
and the file systems won't query the CD-ROM specific information.