Re: [PATCH 2/2] xfs: remove metafile inodes from the active inode stat

["Darrick J. Wong" <djwong@kernel.org>]

On Mon, Feb 02, 2026 at 03:14:32PM +0100, Christoph Hellwig wrote:
> The active inode (or active vnode until recently) stat can get much larger
> than expected on file systems with a lot of metafile inodes like zoned
> file systems on SMR hard disks with 10.000s of rtg rmap inodes.
> 
> Remove all metafile inodes from the active counter to make it more useful
> to track actual workloads and add a separate counter for active metafile
> inodes.
> 
> This fixes xfs/177 on SMR hard drives.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>

Looks ok to me, though I wonder slightly about the atomicity of the
percpu counter inc/decrements.  But it's been that way for a long time
so

Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>

--D

> ---
>  fs/xfs/libxfs/xfs_inode_buf.c |  4 ++++
>  fs/xfs/libxfs/xfs_metafile.c  |  5 +++++
>  fs/xfs/xfs_icache.c           |  5 ++++-
>  fs/xfs/xfs_stats.c            | 11 ++++++++---
>  fs/xfs/xfs_stats.h            |  3 ++-
>  5 files changed, 23 insertions(+), 5 deletions(-)
> 
> diff --git a/fs/xfs/libxfs/xfs_inode_buf.c b/fs/xfs/libxfs/xfs_inode_buf.c
> index dcc9566ef5fe..91a499ced4e8 100644
> --- a/fs/xfs/libxfs/xfs_inode_buf.c
> +++ b/fs/xfs/libxfs/xfs_inode_buf.c
> @@ -268,6 +268,10 @@ xfs_inode_from_disk(
>  	}
>  	if (xfs_is_reflink_inode(ip))
>  		xfs_ifork_init_cow(ip);
> +	if (xfs_is_metadir_inode(ip)) {
> +		XFS_STATS_DEC(ip->i_mount, xs_inodes_active);
> +		XFS_STATS_INC(ip->i_mount, xs_inodes_meta);
> +	}
>  	return 0;
>  
>  out_destroy_data_fork:
> diff --git a/fs/xfs/libxfs/xfs_metafile.c b/fs/xfs/libxfs/xfs_metafile.c
> index cf239f862212..71f004e9dc64 100644
> --- a/fs/xfs/libxfs/xfs_metafile.c
> +++ b/fs/xfs/libxfs/xfs_metafile.c
> @@ -61,6 +61,9 @@ xfs_metafile_set_iflag(
>  	ip->i_diflags2 |= XFS_DIFLAG2_METADATA;
>  	ip->i_metatype = metafile_type;
>  	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
> +
> +	XFS_STATS_DEC(ip->i_mount, xs_inodes_active);
> +	XFS_STATS_INC(ip->i_mount, xs_inodes_meta);
>  }
>  
>  /* Clear the metadata directory inode flag. */
> @@ -74,6 +77,8 @@ xfs_metafile_clear_iflag(
>  
>  	ip->i_diflags2 &= ~XFS_DIFLAG2_METADATA;
>  	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
> +	XFS_STATS_INC(ip->i_mount, xs_inodes_active);
> +	XFS_STATS_DEC(ip->i_mount, xs_inodes_meta);
>  }
>  
>  /*
> diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c
> index f76c6decdaa3..f2d4294efd37 100644
> --- a/fs/xfs/xfs_icache.c
> +++ b/fs/xfs/xfs_icache.c
> @@ -172,7 +172,10 @@ __xfs_inode_free(
>  	/* asserts to verify all state is correct here */
>  	ASSERT(atomic_read(&ip->i_pincount) == 0);
>  	ASSERT(!ip->i_itemp || list_empty(&ip->i_itemp->ili_item.li_bio_list));
> -	XFS_STATS_DEC(ip->i_mount, xs_inodes_active);
> +	if (xfs_is_metadir_inode(ip))
> +		XFS_STATS_DEC(ip->i_mount, xs_inodes_meta);
> +	else
> +		XFS_STATS_DEC(ip->i_mount, xs_inodes_active);
>  
>  	call_rcu(&VFS_I(ip)->i_rcu, xfs_inode_free_callback);
>  }
> diff --git a/fs/xfs/xfs_stats.c b/fs/xfs/xfs_stats.c
> index bc4a5d6dc795..c13d600732c9 100644
> --- a/fs/xfs/xfs_stats.c
> +++ b/fs/xfs/xfs_stats.c
> @@ -59,7 +59,8 @@ int xfs_stats_format(struct xfsstats __percpu *stats, char *buf)
>  		{ "rtrefcntbt",		xfsstats_offset(xs_qm_dqreclaims)},
>  		/* we print both series of quota information together */
>  		{ "qm",			xfsstats_offset(xs_gc_read_calls)},
> -		{ "zoned",		xfsstats_offset(__pad1)},
> +		{ "zoned",		xfsstats_offset(xs_inodes_meta)},
> +		{ "metafile",		xfsstats_offset(xs_xstrat_bytes)},
>  	};
>  
>  	/* Loop over all stats groups */
> @@ -99,16 +100,20 @@ int xfs_stats_format(struct xfsstats __percpu *stats, char *buf)
>  
>  void xfs_stats_clearall(struct xfsstats __percpu *stats)
>  {
> +	uint32_t	xs_inodes_active, xs_inodes_meta;
>  	int		c;
> -	uint32_t	xs_inodes_active;
>  
>  	xfs_notice(NULL, "Clearing xfsstats");
>  	for_each_possible_cpu(c) {
>  		preempt_disable();
> -		/* save xs_inodes_active, it's a universal truth! */
> +		/*
> +		 * Save the active / meta inode counters, as they are stateful.
> +		 */
>  		xs_inodes_active = per_cpu_ptr(stats, c)->s.xs_inodes_active;
> +		xs_inodes_meta = per_cpu_ptr(stats, c)->s.xs_inodes_meta;
>  		memset(per_cpu_ptr(stats, c), 0, sizeof(*stats));
>  		per_cpu_ptr(stats, c)->s.xs_inodes_active = xs_inodes_active;
> +		per_cpu_ptr(stats, c)->s.xs_inodes_meta = xs_inodes_meta;
>  		preempt_enable();
>  	}
>  }
> diff --git a/fs/xfs/xfs_stats.h b/fs/xfs/xfs_stats.h
> index 64bc0cc18126..57c32b86c358 100644
> --- a/fs/xfs/xfs_stats.h
> +++ b/fs/xfs/xfs_stats.h
> @@ -142,7 +142,8 @@ struct __xfsstats {
>  	uint32_t		xs_gc_read_calls;
>  	uint32_t		xs_gc_write_calls;
>  	uint32_t		xs_gc_zone_reset_calls;
> -	uint32_t		__pad1;
> +/* Metafile counters */
> +	uint32_t		xs_inodes_meta;
>  /* Extra precision counters */
>  	uint64_t		xs_xstrat_bytes;
>  	uint64_t		xs_write_bytes;
> -- 
> 2.47.3
> 
>