[lvc-project] [PATCH v4 1/2] xfs: skip inode inactivation on a shut down mount

[Christoph Hellwig]

On Sun, Jun 07, 2026 at 09:30:25PM +0300, Mikhail Lobanov wrote:
> v4: brace both branches of the if/else.  Without CONFIG_XFS_QUOTA
>     xfs_qm_dqdetach() expands to nothing, leaving the else with an empty
>     body, which trips -Wempty-body on a W=1 build (i386-allnoconfig,
>     reported by the kernel test robot).

We should really turn these stubs from macros into inlines, but..

> @@ -1940,10 +1940,26 @@ static int
>  xfs_inodegc_inactivate(
>  	struct xfs_inode	*ip)
>  {
> -	int			error;
> +	int			error = 0;
> 
>  	trace_xfs_inode_inactivating(ip);
> -	error = xfs_inactive(ip);
> +
> +	/*
> +	 * If the filesystem has been shut down - for example a mount that
> +	 * failed after background inactivation was enabled - do not
> +	 * inactivate the inode.  Inactivation modifies persistent metadata,
> +	 * its transactions cannot complete on a shut down mount, and the
> +	 * subsystems it relies on (e.g. quota, mp->m_quotainfo) may not be
> +	 * set up.  Drop any attached dquots and make the inode reclaimable,
> +	 * the same way xfs_inode_mark_reclaimable() does when it sends an
> +	 * inode straight to reclaim.
> +	 */
> +	if (!xfs_is_shutdown(ip->i_mount)) {
> +		error = xfs_inactive(ip);
> +	} else {
> +		/* Going straight to reclaim, so drop the dquots. */
> +		xfs_qm_dqdetach(ip);
> +	}

It might be easier to just move the shutdown check and xfs_qm_dqdetach
cal into the beginning of xfs_inactive as:

	if (xfs_is_shutdown(mp)
		goto out;

after initializing mp at declaration time at the top of the function.

But this is also still missing my other comment:  xfs_qm_dqdetach
will still cause a NULL pointer dereference if quotas haven't been
set up yet or torn down already.  So we'll probably want another
patch to exist early in xfs_qm_dqdetach for that case.