Re: [PATCH][mainline-only] remove useless mnt_want_write call in xfs_write

Re: [PATCH][mainline-only] remove useless mnt_want_write call in xfs_write

[Christoph Hellwig]

Any chance to get this into the git tree for the first 2.6.28 pull?

On Thu, Aug 14, 2008 at 11:25:51PM +0200, Christoph Hellwig wrote:
> When mnt_want_write was introduced a call to it was added around
> xfs_ichgtime, but there is no need for this because a file can't be open
> read/write on a r/o mount, and a mount can't degrade r/o while we still
> have files open for writing.  As the mnt_want_write changes were never
> merged into the CVS tree this patch is for mainline only.
> 
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> 
> --- linux-2.6/fs/xfs/linux-2.6/xfs_lrw.c	2008-08-14 14:52:15.000000000 -0300
> +++ linux-2.6-xfs/fs/xfs/linux-2.6/xfs_lrw.c	2008-08-14 14:54:53.000000000 -0300
> @@ -51,7 +51,6 @@
>  #include "xfs_vnodeops.h"
>  
>  #include <linux/capability.h>
> -#include <linux/mount.h>
>  #include <linux/writeback.h>
>  
>  
> @@ -668,15 +667,8 @@ start:
>  	if (new_size > xip->i_size)
>  		xip->i_new_size = new_size;
>  
> -	/*
> -	 * We're not supposed to change timestamps in readonly-mounted
> -	 * filesystems.  Throw it away if anyone asks us.
> -	 */
> -	if (likely(!(ioflags & IO_INVIS) &&
> -		   !mnt_want_write(file->f_path.mnt))) {
> +	if (likely(!(ioflags & IO_INVIS)))
>  		xfs_ichgtime(xip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
> -		mnt_drop_write(file->f_path.mnt);
> -	}
>  
>  	/*
>  	 * If the offset is beyond the size of the file, we have a couple
---end quoted text---

Re: [PATCH][mainline-only] remove useless mnt_want_write call in xfs_write

[Christoph Hellwig]

ping?

On Mon, Sep 29, 2008 at 09:44:50AM +0200, Christoph Hellwig wrote:
> Any chance to get this into the git tree for the first 2.6.28 pull?
> 
> On Thu, Aug 14, 2008 at 11:25:51PM +0200, Christoph Hellwig wrote:
> > When mnt_want_write was introduced a call to it was added around
> > xfs_ichgtime, but there is no need for this because a file can't be open
> > read/write on a r/o mount, and a mount can't degrade r/o while we still
> > have files open for writing.  As the mnt_want_write changes were never
> > merged into the CVS tree this patch is for mainline only.
> > 
> > 
> > Signed-off-by: Christoph Hellwig <hch@lst.de>
> > 
> > --- linux-2.6/fs/xfs/linux-2.6/xfs_lrw.c	2008-08-14 14:52:15.000000000 -0300
> > +++ linux-2.6-xfs/fs/xfs/linux-2.6/xfs_lrw.c	2008-08-14 14:54:53.000000000 -0300
> > @@ -51,7 +51,6 @@
> >  #include "xfs_vnodeops.h"
> >  
> >  #include <linux/capability.h>
> > -#include <linux/mount.h>
> >  #include <linux/writeback.h>
> >  
> >  
> > @@ -668,15 +667,8 @@ start:
> >  	if (new_size > xip->i_size)
> >  		xip->i_new_size = new_size;
> >  
> > -	/*
> > -	 * We're not supposed to change timestamps in readonly-mounted
> > -	 * filesystems.  Throw it away if anyone asks us.
> > -	 */
> > -	if (likely(!(ioflags & IO_INVIS) &&
> > -		   !mnt_want_write(file->f_path.mnt))) {
> > +	if (likely(!(ioflags & IO_INVIS)))
> >  		xfs_ichgtime(xip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
> > -		mnt_drop_write(file->f_path.mnt);
> > -	}
> >  
> >  	/*
> >  	 * If the offset is beyond the size of the file, we have a couple
> ---end quoted text---
> 
> 
---end quoted text---

Re: [PATCH][mainline-only] remove useless mnt_want_write call in xfs_write

[Christoph Hellwig]

ping^2?

Note that with the new git tree it's also not just mainline only
anymore.

On Mon, Nov 10, 2008 at 08:31:50AM -0500, Christoph Hellwig wrote:
> ping?
> 
> On Mon, Sep 29, 2008 at 09:44:50AM +0200, Christoph Hellwig wrote:
> > Any chance to get this into the git tree for the first 2.6.28 pull?
> > 
> > On Thu, Aug 14, 2008 at 11:25:51PM +0200, Christoph Hellwig wrote:
> > > When mnt_want_write was introduced a call to it was added around
> > > xfs_ichgtime, but there is no need for this because a file can't be open
> > > read/write on a r/o mount, and a mount can't degrade r/o while we still
> > > have files open for writing.  As the mnt_want_write changes were never
> > > merged into the CVS tree this patch is for mainline only.
> > > 
> > > 
> > > Signed-off-by: Christoph Hellwig <hch@lst.de>
> > > 
> > > --- linux-2.6/fs/xfs/linux-2.6/xfs_lrw.c	2008-08-14 14:52:15.000000000 -0300
> > > +++ linux-2.6-xfs/fs/xfs/linux-2.6/xfs_lrw.c	2008-08-14 14:54:53.000000000 -0300
> > > @@ -51,7 +51,6 @@
> > >  #include "xfs_vnodeops.h"
> > >  
> > >  #include <linux/capability.h>
> > > -#include <linux/mount.h>
> > >  #include <linux/writeback.h>
> > >  
> > >  
> > > @@ -668,15 +667,8 @@ start:
> > >  	if (new_size > xip->i_size)
> > >  		xip->i_new_size = new_size;
> > >  
> > > -	/*
> > > -	 * We're not supposed to change timestamps in readonly-mounted
> > > -	 * filesystems.  Throw it away if anyone asks us.
> > > -	 */
> > > -	if (likely(!(ioflags & IO_INVIS) &&
> > > -		   !mnt_want_write(file->f_path.mnt))) {
> > > +	if (likely(!(ioflags & IO_INVIS)))
> > >  		xfs_ichgtime(xip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
> > > -		mnt_drop_write(file->f_path.mnt);
> > > -	}
> > >  
> > >  	/*
> > >  	 * If the offset is beyond the size of the file, we have a couple
> > ---end quoted text---
> > 
> > 
> ---end quoted text---
> 
> 
---end quoted text---

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs