On Wed, 2010-10-13 at 13:19 +0200, Christoph Hellwig wrote:
> > +static int iblock_do_discard(struct se_task *task, enum blk_discard_type type)
> > +{
> > + struct iblock_dev *ibd = task->se_dev->dev_ptr;
> > + struct block_device *bd = ibd->ibd_bd;
> > + struct se_cmd *cmd = TASK_CMD(task);
> > +
> > + if (type == DISCARD_UNMAP)
> > + return transport_generic_unmap(cmd, bd);
> > + else if (type == DISCARD_WRITE_SAME_UNMAP)
> > + return iblock_emulate_write_same_unmap(task);
> > + else {
> > + printk(KERN_ERR "Unsupported discard_type_t: %d\n", type);
> > + return -ENOSYS;
> > + }
> > +
> > + return -ENOSYS;
> > +}
> > +
>
> I don't think the discard code is quite, nor nicely structured.
>
> The parsing of the WRITE SAME and UNMAP CDBs is something the generic
> CDB parsing code should do,
Ok, so you are thinking about a seperate transport_emulate_write_same()
and transport_emulate_unmap() called from
transport_emulate_control_cdb(), right..?
> and just give a range of lists of lba/len
> pairs to the ->discard method in the backed.
Yes, these are already available from the passed struct
se_task->task_lba and ->task_size values.
> Also your generic
> discard helpers aren't actually generic - they require a block device
> and thus should be only in iblock.c. While your hack in the file
> backend to use it if we're using a block device as backing file
> works it's rather gross. Having the file backend general enough to
> work with a block devices is fine, but adding special hacks that
> only work on block device while having a fully working bio based backed
> is a bit gross.
Yes, so the problem of trying to make this code generic (eg: outside of
TCM subsystem plugins) is that blk_issue_discard() takes struct
block_device, which means we the subsystem plugin has to locate struct
block_device inside of non generic cide.
So, then the main issue becomes FILEIO + block level discard and how to
issue an blk_issue_discard() from struct fileio in the most sane way.
If there is no sane way then I will just drop this bit, or just do the
file level 'hole punch' that you are speaking about.
> Btw, at least on XFS you can implement discard using
> hole punch operations, although that can lead to quite bad fragmentation
> in cases. Just as block-level discards can lead to quite bad
> performance - I'd suggest to not enable them by default.
>
Ok, I will disable these by default for IBLOCK and FILEIO, and require
an explict override from user with the emulate_tpu and emulate_tpws
values in /sys/kernel/config/target/core/$HBA/$DEV/attrib/
> One other thing I noticed is that you use igrab a lot. In general
> drivers have absolutely no need for a igrab. If you have a reference
> to the file behind an inode you keep the inode in core and there's no
> need at all to grab a second reference to it.
The igrab() and iput() usage in FILEIO code are used for locating the
struct block_device for blk_issue_discard(). If this is the case then I
will remove these from FILEIO.
Thanks!
--nab
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
