Re: [PATCH] block: replace __getblk_slow misfix by grow_dev_page fix

[Jeff Moyer <jmoyer@redhat.com>]

Hugh Dickins <hughd@google.com> writes:

> On Thu, 23 Aug 2012, Jeff Moyer wrote:
>> Hugh Dickins <hughd@google.com> writes:
>> 
>> > [PATCH] block: replace __getblk_slow misfix by grow_dev_page fix
>> >
>> > Commit 91f68c89d8f3 ("block: fix infinite loop in __getblk_slow")
>> > is not good: a successful call to grow_buffers() cannot guarantee
>> > that the page won't be reclaimed before the immediate next call to
>> > __find_get_block(), which is why there was always a loop there.
>> [snip]
>> > Revert 91f68c89d8f3, restoring __getblk_slow() to how it was (plus
>> > a checkpatch nitfix).  Simplify the interface between grow_buffers()
>> > and grow_dev_page(), and avoid the infinite loop beyond end of device
>> > by instead checking init_page_buffers()'s end_block there (I presume
>> > that's more efficient than a repeated call to blkdev_max_block()),
>> > returning -ENXIO to __getblk_slow() in that case.
>> >
>> > And remove akpm's ten-year-old "__getblk() cannot fail ... weird"
>> > comment, but that is worrying: are all users of __getblk() really
>> > now prepared for a NULL bh beyond end of device, or will some oops??
>> 
>> Hi, Hugh,
>> 
>> Thanks for digging into this.  I had wondered whether that loop was
>> required for other purposes, and you've verified that it was.  I mostly
>> like your fix.  Returning end_block from init_page_buffers is a little
>> strange,
>
> It is a little strange, I agree.  I wrestled a lot with the way block
> and end_block were known at opposite ends of the callstack, and needed
> to be brought together for the check.
>
> If you think it would be better to move the blkdev_max_block()
> lookup up a level into grow_dev_page(), then pass end_block down to
> init_page_buffers(), that could work as well; though we'd then still
> need to report "failure" from init_page_buffers().

I have no strong opinion;  I can live with it as-is.

> Great, thanks a lot.  It (but equally its incorrect earlier version)
> have been running fine at home under my swapping load.  I'm confident
> that it fixes the issues I had been seeing, although, strictly speaking,
> it'll take another two or three days to reach bisection confidence level.

Well, I bought your analysis, so I'm at least ready to see this get
pushed. ;-)

Cheers,
Jeff