Re: [PATCH 2/4] fs/resctrl: Free mon_data structures on rdt_get_tree() failure

[Luck, Tony]

On Fri, May 08, 2026 at 11:21:41AM -0700, Tony Luck wrote:
> If mkdir_mondata_all() succeeds but a subsequent call in rdt_get_tree()
> fails, the mon_data structures allocated by mon_get_kn_priv() are
> leaked. Add mon_put_kn_priv() to the out_mondata error path to free
> them.
> 
> mon_get_kn_priv() and mon_put_kn_priv() moved so defined before used.

My bad, leaving this stale commit comment here. Code rearrangement
in patch 1 covered this.

> Fixes: ee4f0ec938ad ("fs/resctrl: Simplify allocation of mon_data structures")
> Reported-by: Reinette Chatre <reinette.chatre@xxxxxxxxx>
> Assisted-by: GitHub_Copilot_CLI:claude-sonnet-4.6

Hmmm. I don't think I was assisted at all. Actively sabotaged with a
patch that looks plausible, but it actually wrong seems more accurate.

> Signed-off-by: Tony Luck <tony.luck@xxxxxxxxx>
> ---
>  fs/resctrl/rdtgroup.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/fs/resctrl/rdtgroup.c b/fs/resctrl/rdtgroup.c
> index a6376a3fc4c3..0db1a92aefbe 100644
> --- a/fs/resctrl/rdtgroup.c
> +++ b/fs/resctrl/rdtgroup.c
> @@ -3067,8 +3067,10 @@ static int rdt_get_tree(struct fs_context *fc)
>  out_psl:
>  	rdt_pseudo_lock_release();
>  out_mondata:
> -	if (resctrl_arch_mon_capable())
> +	if (resctrl_arch_mon_capable()) {
> +		mon_put_kn_priv();

Claude put this call here ...

>  		kernfs_remove(kn_mondata);
> +	}
>  out_mongrp:
>  	if (resctrl_arch_mon_capable()) {

But it really ought to be here.

>  		rdtgroup_unassign_cntrs(&rdtgroup_default);
> -- 
> 2.54.0

Claude is fired. Replacement patch below.

-Tony