On Fri, Feb 07, 2020 at 02:29:37AM +0000, Damien Le Moal wrote:
> On 2020/02/07 9:29, Dave Chinner wrote:
> > On Thu, Feb 06, 2020 at 02:26:30PM +0900, Damien Le Moal wrote:
> >> zonefs is a very simple file system exposing each zone of a zoned block
> >> device as a file. Unlike a regular file system with zoned block device
> >> support (e.g. f2fs), zonefs does not hide the sequential write
> >> constraint of zoned block devices to the user. Files representing
> >> sequential write zones of the device must be written sequentially
> >> starting from the end of the file (append only writes).
> >
> > ....
> >> + if (flags & IOMAP_WRITE)
> >> + length = zi->i_max_size - offset;
> >> + else
> >> + length = min(length, isize - offset);
> >> + mutex_unlock(&zi->i_truncate_mutex);
> >> +
> >> + iomap->offset = offset & (~sbi->s_blocksize_mask);
> >> + iomap->length = ((offset + length + sbi->s_blocksize_mask) &
> >> + (~sbi->s_blocksize_mask)) - iomap->offset;
> >
> > iomap->length = __ALIGN_MASK(offset + length, sbi->s_blocksize_mask) -
> > iomap->offset;
> >
> > or it could just use ALIGN(..., sb->s_blocksize) and not need
> > pre-calculation of the mask value...
>
> Yes, that is cleaner. Fixed.
>
> >> +static ssize_t zonefs_file_dio_write(struct kiocb *iocb, struct iov_iter *from)
> >> +{
> >> + struct inode *inode = file_inode(iocb->ki_filp);
> >> + struct zonefs_sb_info *sbi = ZONEFS_SB(inode->i_sb);
> >> + struct zonefs_inode_info *zi = ZONEFS_I(inode);
> >> + size_t count;
> >> + ssize_t ret;
> >> +
> >> + /*
> >> + * For async direct IOs to sequential zone files, ignore IOCB_NOWAIT
> >> + * as this can cause write reordering (e.g. the first aio gets EAGAIN
> >> + * on the inode lock but the second goes through but is now unaligned).
> >> + */
> >> + if (zi->i_ztype == ZONEFS_ZTYPE_SEQ && !is_sync_kiocb(iocb)
> >> + && (iocb->ki_flags & IOCB_NOWAIT))
> >> + iocb->ki_flags &= ~IOCB_NOWAIT;
> >
> > Hmmm. I'm wondering if it would be better to return -EOPNOTSUPP here
> > so that the application knows it can't do non-blocking write AIO to
> > this file.
>
> I wondered the same too. In the end, I decided to go with silently ignoring
> the flag (for now) since raw block device accesses do the same (the NOWAIT
> support is not complete and IOs may wait on free tags). I have an idea for
> fixing simply the out-of-order issuing that may result from using nowait. I
> will send a patch for that later and can then remove this.
> But if zonefs does not make it to 5.6 (looking really tight), I will send
> add that patch to a new zonefs series rebased for 5.7.
THat seems reasonable to me.
Cheers,
Dave.
--
Dave Chinner
david@xxxxxxxxxxxxx
