On Sun, Sep 10, 2006 at 06:45:03AM +0000, YAMAMOTO Takashi wrote:

>  
>  > If you limit dirty buffers to 1/2 of total buffer cache you will
>  > still deadlock. It doesn't matter what limit you put on dirty
>  > buffers. Whenever that limit it reached then vnd cannot process
>  > them because for doing so it would have to exceed that limit. You
>  > must make sure that the filesystem that vnd is calling through
>  > VOP_STRATEGY _can_ allocate new buffers, even those that it
>  > can use for writing.
>  
>  which filesystem are you talking about?

The filesystem where the virtual device resides on, in that
my UFS.


>  > If my patch doesn't make sense, then try to understand why it works.
>  
>  now i've tried your patch.  it didn't work for me.

That's probably because you have a completely different test that
shows a different problem.

>  vndthread just got deadlock on "getnewbuf" without your
>  throttling code triggered at all.
>  
>  	(/bigfile is 512M file w/o holes on ufs.)
>  	vnconfig vnd0 /bigfile
>  	newfs -F /dev/rvnd0d
>  	mount /dev/vnd0d /mnt
>  	dd if=/dev/zero of=/mnt/a bs=1m seek=1m  (ENOSPC)

That creates a sparse file starting at offset 1TB.

>  	vnconfig vnd1 /mnt/a
>  	dd if=/dev/zero of=/dev/vnd1d bs=1m seek=1m  (deadlock)

That would write the sparse file starting from offset 1TB.

>  (i can't follow your recipe in the PR because i don't have 10G disk space.)

You can probably use a smaller file as long as enough data gets
written. mkfs.ext2 just writes the ext2-metadata, so a simple dd
would require a much smaller file.


>  > The writer does never clean blocks. The writer can only create
>  > more dirty blocks. The blocks are "clean" again when they
>  > are processed by the device, i.e. in some iodone routine
>  > which is called in an interrupt. How should a sleeping process
>  > prevent that? In particular, at that time the writer is stopped
>  > in vndstrategy, we definitely know that the vndthread _is_ busy
>  > cleaning buffers, because that is the stop condition. We therefore
>  > make _sure_ that it can do so.
>  
>  i think that our definitions of "writer" are different.
>  mine is "callers of vndstrategy."
>  yours is "callers of write(2)".  right?

Mine is "callers of vndstrategy", for writing the block device
that's the same as "callers of write(2)".

I have a look at your patch.

-- 
                                Michael van Elst
Internet: mlelstv@serpens.de
                                "A potential Snark may lurk in every tree."