xfs
[Top] [All Lists]

Re: [PATCH 10/18] xfs: set sparse inodes feature bit when a sparse chunk

To: Brian Foster <bfoster@xxxxxxxxxx>
Subject: Re: [PATCH 10/18] xfs: set sparse inodes feature bit when a sparse chunk is allocated
From: Dave Chinner <david@xxxxxxxxxxxxx>
Date: Fri, 25 Jul 2014 08:46:28 +1000
Cc: xfs@xxxxxxxxxxx
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <1406211788-63206-11-git-send-email-bfoster@xxxxxxxxxx>
References: <1406211788-63206-1-git-send-email-bfoster@xxxxxxxxxx> <1406211788-63206-11-git-send-email-bfoster@xxxxxxxxxx>
User-agent: Mutt/1.5.21 (2010-09-15)
On Thu, Jul 24, 2014 at 10:23:00AM -0400, Brian Foster wrote:
> The sparse chunk inode record format is backwards compatible with old
> format inobt records as long as full chunks are allocated. The holemask
> field uses higher order bytes of the freecount. While sparse chunks can
> be enabled on previously unsupported fs, older kernel drivers cannot
> parse sparse inode records.
> 
> Set the feature incompatible bit once a sparse inode chunk is allocated
> to prevent older XFS drivers from tripping over the new format.
> 
> Signed-off-by: Brian Foster <bfoster@xxxxxxxxxx>
> ---
>  fs/xfs/libxfs/xfs_ialloc.c | 33 +++++++++++++++++++++++++++++++++
>  1 file changed, 33 insertions(+)
> 
> diff --git a/fs/xfs/libxfs/xfs_ialloc.c b/fs/xfs/libxfs/xfs_ialloc.c
> index 4226b1b..4dd45c2 100644
> --- a/fs/xfs/libxfs/xfs_ialloc.c
> +++ b/fs/xfs/libxfs/xfs_ialloc.c
> @@ -407,6 +407,27 @@ error:
>       return error;
>  }
>  
> +STATIC void
> +xfs_sbversion_add_spinodes(
> +     struct xfs_trans        *tp,
> +     struct xfs_mount        *mp)
> +{
> +     if (xfs_sb_has_incompat_log_feature(&mp->m_sb,
> +                                         XFS_SB_FEAT_INCOMPAT_SPINODES))

wrong function. xfs_sb_has_incompat_feature().

> +             return;
> +
> +     spin_lock(&mp->m_sb_lock);
> +     if (xfs_sb_has_incompat_log_feature(&mp->m_sb,
> +                                         XFS_SB_FEAT_INCOMPAT_SPINODES)) {
> +             spin_unlock(&mp->m_sb_lock);
> +             return;
> +     }
> +
> +     mp->m_sb.sb_features_incompat |= XFS_SB_FEAT_INCOMPAT_SPINODES;
> +     spin_unlock(&mp->m_sb_lock);
> +     xfs_mod_sb(tp, XFS_SB_FEATURES_INCOMPAT);
> +}
> +
>  /*
>   * Allocate new inodes in the allocation group specified by agbp.
>   * Return 0 for success, else error code.
> @@ -631,6 +652,18 @@ xfs_ialloc_ag_alloc(
>               if (error)
>                       return error;
>       }
> +
> +     /*
> +      * Set an incompat feature bit as old drivers can't parse sparse
> +      * records. Pre-sparse inode chunk drivers will include the
> +      * holemask in the higher order freecount bits, resulting in a
> +      * bogus value.
> +      *
> +      * XXX: when is this bit removed?
> +      */
> +     if (xfs_inobt_issparse(&rec))
> +             xfs_sbversion_add_spinodes(tp, args.mp);

I'd like to avoid this dynamic feature bit adding if possible. Do we
really need it?

Cheers,

Dave.
-- 
Dave Chinner
david@xxxxxxxxxxxxx

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