Re: [PATCH v2 1/3] erofs: iomap support for non-tailpacking DIO

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Wed, Aug 04, 2021 at 12:30:35PM +0800, Gao Xiang wrote:
> Hi Chao,
> 
> On Wed, Aug 04, 2021 at 10:57:08AM +0800, Chao Yu wrote:
> > On 2021/7/31 3:46, Gao Xiang wrote:
> 
> ...
> 
> > >   }
> > > +static int erofs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
> > > +		unsigned int flags, struct iomap *iomap, struct iomap *srcmap)
> > > +{
> > > +	int ret;
> > > +	struct erofs_map_blocks map;
> > > +
> > > +	map.m_la = offset;
> > > +	map.m_llen = length;
> > > +
> > > +	ret = erofs_map_blocks_flatmode(inode, &map, EROFS_GET_BLOCKS_RAW);
> > > +	if (ret < 0)
> > > +		return ret;
> > > +
> > > +	iomap->bdev = inode->i_sb->s_bdev;
> > > +	iomap->offset = map.m_la;
> > > +	iomap->length = map.m_llen;
> > > +	iomap->flags = 0;
> > > +
> > > +	if (!(map.m_flags & EROFS_MAP_MAPPED)) {
> > > +		iomap->type = IOMAP_HOLE;
> > > +		iomap->addr = IOMAP_NULL_ADDR;
> > > +		if (!iomap->length)
> > > +			iomap->length = length;
> > 
> > This only happens for the case offset exceeds isize?
> 
> Thanks for the review.
> 
> Yeah, this is a convention (length 0 with !EROFS_MAP_MAPPED) for post-EOF
> in erofs_map_blocks_flatmode(), need to follow iomap rule as well.
> 
> > 
> > > +		return 0;
> > > +	}
> > > +
> > > +	/* that shouldn't happen for now */
> > > +	if (map.m_flags & EROFS_MAP_META) {
> > > +		DBG_BUGON(1);
> > > +		return -ENOTBLK;
> > > +	}
> > > +	iomap->type = IOMAP_MAPPED;
> > > +	iomap->addr = map.m_pa;
> > > +	return 0;
> > > +}
> > > +
> > > +const struct iomap_ops erofs_iomap_ops = {
> > > +	.iomap_begin = erofs_iomap_begin,
> > > +};
> > > +
> > > +static int erofs_prepare_dio(struct kiocb *iocb, struct iov_iter *to)
> > > +{
> > > +	struct inode *inode = file_inode(iocb->ki_filp);
> > > +	loff_t align = iocb->ki_pos | iov_iter_count(to) |
> > > +		iov_iter_alignment(to);
> > > +	struct block_device *bdev = inode->i_sb->s_bdev;
> > > +	unsigned int blksize_mask;
> > > +
> > > +	if (bdev)
> > > +		blksize_mask = (1 << ilog2(bdev_logical_block_size(bdev))) - 1;
> > > +	else
> > > +		blksize_mask = (1 << inode->i_blkbits) - 1;
> > > +
> > > +	if (align & blksize_mask)
> > > +		return -EINVAL;
> > > +
> > > +	/*
> > > +	 * Temporarily fall back tail-packing inline to buffered I/O instead
> > > +	 * since tail-packing inline support relies on an iomap core update.
> > > +	 */
> > > +	if (EROFS_I(inode)->datalayout == EROFS_INODE_FLAT_INLINE &&
> > > +	    iocb->ki_pos + iov_iter_count(to) >
> > > +			rounddown(inode->i_size, EROFS_BLKSIZ))
> > > +		return 1;
> > > +	return 0;
> > > +}
> > > +
> > > +static ssize_t erofs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
> > > +{
> > > +	/* no need taking (shared) inode lock since it's a ro filesystem */
> > > +	if (!iov_iter_count(to))
> > > +		return 0;
> > > +
> > > +	if (iocb->ki_flags & IOCB_DIRECT) {
> > > +		int err = erofs_prepare_dio(iocb, to);
> > > +
> > > +		if (!err)
> > > +			return iomap_dio_rw(iocb, to, &erofs_iomap_ops,
> > > +					    NULL, 0);
> > > +		if (err < 0)
> > > +			return err;
> > > +		/*
> > > +		 * Fallback to buffered I/O if the operation being performed on
> > > +		 * the inode is not supported by direct I/O. The IOCB_DIRECT
> > > +		 * flag needs to be cleared here in order to ensure that the
> > > +		 * direct I/O path within generic_file_read_iter() is not
> > > +		 * taken.
> > > +		 */
> > > +		iocb->ki_flags &= ~IOCB_DIRECT;
> > > +	}
> > > +	return generic_file_read_iter(iocb, to);
> > 
> > It looks it's fine to call filemap_read() directly since above codes have
> > covered DIO case, then we don't need to change iocb->ki_flags flag, it's
> > minor though.
> 
> Yeah, we could use filemap_read() here instead. yet IMO, it might be
> better to drop IOCB_DIRECT too to keep iocb consistent with the real
> semantics (even it's not used internally.)

After checking the other users of filemap_read(), I'm fine to leave
IOCB_DIRECT as-is. Will update.

Thanks,
Gao Xiang



[Index of Archives]     [Linux Ext4 Filesystem]     [Union Filesystem]     [Filesystem Testing]     [Ceph Users]     [Ecryptfs]     [AutoFS]     [Kernel Newbies]     [Share Photos]     [Security]     [Netfilter]     [Bugtraq]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux Cachefs]     [Reiser Filesystem]     [Linux RAID]     [Samba]     [Device Mapper]     [CEPH Development]

  Powered by Linux