Re: [PATCH 07/10] ata: ahci: Disallow LPM policy control for external ports

[Niklas Cassel <cassel@kernel.org>]

On Mon, Jun 30, 2025 at 03:26:34PM +0900, Damien Le Moal wrote:
> Commit ae1f3db006b7 ("ata: ahci: do not enable LPM on external ports")
> added an early return in ahci_update_initial_lpm_policy() for all
> ports flagged as external with the ATA_PFLAG_EXTERNAL port flag (e.g.
> eSATA ports or hotplug capable ports) so that the target_lpm_policy of
> these ports is unchanged and set to ATA_LPM_UNKNOWN. thus forcing libata
> EH to not be called for external port. The goal of this change is to
> preserve the initial power management policy to not break the hotplug
> capability of external ports.
> 
> However, this change is incomplete as users or system daemon (e.g.
> systemd-udevd) can still apply the system preferred power management
> policy through sysfs, thus potentially unknowingly breaking the port
> hotplug capability.
> 
> Modify ahci_update_initial_lpm_policy() to flag external ports with
> ATA_FLAG_NO_LPM to prevent changes to the LPM policy by users through
> the sysfs link_power_management_policy host attribute. Also set the
> target_lpm_policy of external ports to ATA_LPM_MAX_POWER to ensure
> that the port is not in a low power state preventing hotplug operations.
> 
> Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
> ---
>  drivers/ata/ahci.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/ata/ahci.c b/drivers/ata/ahci.c
> index 0760fa47d90e..34698ae39f55 100644
> --- a/drivers/ata/ahci.c
> +++ b/drivers/ata/ahci.c
> @@ -1776,7 +1776,9 @@ static void ahci_update_initial_lpm_policy(struct ata_port *ap)
>  	 * LPM if the port advertises itself as an external port.
>  	 */
>  	if (ap->pflags & ATA_PFLAG_EXTERNAL) {
> -		ata_port_dbg(ap, "external port, not enabling LPM\n");
> +		ata_port_warn(ap, "External port, forcing LPM max_power\n");

Should this really be a warning?
External ports are quite common, so it seems like this could easily trigger 32
lines of prints on a controller with 32 ports.

We already have the " ext" print when printing each port during probe,
perhaps that is enough?


Kind regards,
Niklas