[Top] [All Lists]

Re: XFS bug with ftruncate, mmap and holes.

To: Stephen Lord <lord@xxxxxxx>
Subject: Re: XFS bug with ftruncate, mmap and holes.
From: Keith Owens <kaos@xxxxxxx>
Date: Tue, 29 Jan 2002 00:23:26 +1100
Cc: linux-xfs@xxxxxxxxxxx, lm@xxxxxxxxxxxx
In-reply-to: Your message of "Mon, 28 Jan 2002 07:11:17 MDT." <3C554DF5.80701@xxxxxxx>
Sender: owner-linux-xfs@xxxxxxxxxxx
On Mon, 28 Jan 2002 07:11:17 -0600, 
Stephen Lord <lord@xxxxxxx> wrote:
>Do we have any indication if this is something which has 'appeared' at 
>some point,
>or has been around for a while?

No idea, I have only just started using mdbm on 2.4.17-xfs.

>Looks like a page dirtied only by mmap 
>is not
>getting flushed by this code. Other user's of mmap (such as the loader) 
>appear not
>to have problems like this.
>I will take a look, but I suspect an msync before unmap will work around 
>this for
>now (not tested yet).

I thought about adding msync but the strange thing is the data is
available in the file after the program has exited but that data is
later corrupted.  The corruption does not appear immediately, only if
the file is left alone for a long time or the system is rebooted does
the corruption appear.  Makes me think something is overwriting the
pages later.

Just added msync(m, size, MS_SYNC) before munmap, bug still exists.

<Prev in Thread] Current Thread [Next in Thread>