Re: New TRIM/UNMAP tree published (2009-05-02)

[Hugh Dickins <hugh@veritas.com>]

On Sun, 3 May 2009, Matthew Wilcox wrote:
> On Sat, May 02, 2009 at 11:11:50PM -0700, Matthew Wilcox wrote:
> > > blkdev_issue_discard() and blk_ioctl_discard() has half a page
> > > of common (and changing) code, could be done to use a common
> > > helper that sets policy about bio allocation sizes and such.
> > > 
> > > Just my $0.017
> > 
> > Yes, that works nicely.  Thanks for the suggestion.
> 
> I've pushed out a new git tree:
> 
> http://git.kernel.org/?p=linux/kernel/git/willy/ssd.git;a=shortlog;h=trim-20090502

I'm glad to see this coming alive again, thank you.

I notice OCZ hope to have updated Vertex firmware mid-May,
such that their TRIM support matches up with (y)ours.

One comment below...

> 
> Changes since the last version:
> 
>  - Based on 2.6.30-rc4.
>  - Dropped the three patches which were already merged.
>  - Shuffled 'Make DISCARD_BARRIER and DISCARD_NOBARRIER writes instead of
>    reads' to the front of the queue since it's not controversial and could
>    be merged earlier.
>  - Inserted 'Unify blk_ioctl_discard and blkdev_issue_discard' as patch two.
>  - Updated 'ide: Add support for TRIM' to compile with new IDE code.
> 
> Still to do:
>  - Understand all the changes Bart made to the IDE code; I didn't make all
>    the changes that he did.  I just made it compile for now.
>  - Dave had some objections to the description of 'Make DISCARD_BARRIER and
>    DISCARD_NOBARRIER writes instead of reads', but didn't propose replacement
>    text.
>  - Figure out what to do about memory allocation.

You have GFP_KERNEL allocations in your discard_fn()s.  I believe that
allocations down at that level will need to be GFP_NOIO - the bio mempool
might be empty by this point, you don't want memory allocation to get into
submitting even more I/O to satisfy your discard_fn().  Does it need its
own mempool of reserve pages?  My guess is that you can get away without
that, since a failed discard, though regrettable, loses no data.

Hugh

>  - Handle error returns from discard_prep.
>  - Test the new code still works.
>  - Delve into the latest SCSI spec and see if anything changed in UNMAP
> 
> Given the extensive length of the todo list, I shan't send out the mailbomb.