LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
* nobh_truncate_page question
@ 2007-02-07 22:45 Dmitriy Monakhov
  2007-02-08  2:11 ` Andrew Morton
  0 siblings, 1 reply; 2+ messages in thread
From: Dmitriy Monakhov @ 2007-02-07 22:45 UTC (permalink / raw)
  To: Linux Kernel

Could some one please explain me what's happens with nobh_truncate_page?

int nobh_truncate_page(struct address_space *mapping, loff_t from)
{
........
 if ((offset & (blocksize - 1)) == 0)
                goto out;

        ret = -ENOMEM;
        page = grab_cache_page(mapping, index);
        if (!page)
                goto out;

        to = (offset + blocksize) & ~(blocksize - 1);
        ret = a_ops->prepare_write(NULL, page, offset, to);
<<<<< This prepare_write() not balanced by appropriate commit_write()
<<<<< Currently this not is fatal because this function used only
<<<<< by ext2 and jfs witch  unbalanced prepare/commit operations
<<<<< not painful.  
        if (ret == 0) {
                kaddr = kmap_atomic(page, KM_USER0);
                memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
                flush_dcache_page(page);
                kunmap_atomic(kaddr, KM_USER0);
                set_page_dirty(page);
        }
        unlock_page(page);
        page_cache_release(page);
out:
        return ret;
}

But we can't just call commit_write after all job done because this result in
i_size increasing till block end. Where are two possible solutions:
a) Call ->commit_write() and then restore i_size to old value.
b) Leave code as it is now, but add warning comment about prepare/commit
   balance violation.


^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: nobh_truncate_page question
  2007-02-07 22:45 nobh_truncate_page question Dmitriy Monakhov
@ 2007-02-08  2:11 ` Andrew Morton
  0 siblings, 0 replies; 2+ messages in thread
From: Andrew Morton @ 2007-02-08  2:11 UTC (permalink / raw)
  To: Dmitriy Monakhov; +Cc: Linux Kernel

On Thu, 08 Feb 2007 01:45:10 +0300
Dmitriy Monakhov <dmonakhov@openvz.org> wrote:

> Could some one please explain me what's happens with nobh_truncate_page?
> 
> int nobh_truncate_page(struct address_space *mapping, loff_t from)
> {
> ........
>  if ((offset & (blocksize - 1)) == 0)
>                 goto out;
> 
>         ret = -ENOMEM;
>         page = grab_cache_page(mapping, index);
>         if (!page)
>                 goto out;
> 
>         to = (offset + blocksize) & ~(blocksize - 1);
>         ret = a_ops->prepare_write(NULL, page, offset, to);
> <<<<< This prepare_write() not balanced by appropriate commit_write()
> <<<<< Currently this not is fatal because this function used only
> <<<<< by ext2 and jfs witch  unbalanced prepare/commit operations
> <<<<< not painful.  

See that comment at the top of nobh_truncate_page()?

>         if (ret == 0) {
>                 kaddr = kmap_atomic(page, KM_USER0);
>                 memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
>                 flush_dcache_page(page);
>                 kunmap_atomic(kaddr, KM_USER0);
>                 set_page_dirty(page);
>         }
>         unlock_page(page);
>         page_cache_release(page);
> out:
>         return ret;
> }
> 
> But we can't just call commit_write after all job done because this result in
> i_size increasing till block end. Where are two possible solutions:
> a) Call ->commit_write() and then restore i_size to old value.
> b) Leave code as it is now, but add warning comment about prepare/commit
>    balance violation.

It just happens that nobh_prepare_write() does all the things which
nobh_truncate_page() wants done.

If you need to change nobh_prepare_write(), you'll need to change
nobh_truncate_page().

Or something.  You haven't told us what you're trying to do.

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2007-02-08  2:11 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2007-02-07 22:45 nobh_truncate_page question Dmitriy Monakhov
2007-02-08  2:11 ` Andrew Morton

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).