Re: [PATCH] libata: Ignore spurious PHY events on LPM policy change

[Tejun Heo]

On Sat, Apr 25, 2015 at 11:34:29AM +0200, Gabriele Mazzotta wrote:
> When the LPM policy is set to ATA_LPM_MAX_POWER, the device might
> generate a spurious PHY event that might cause errors on the link.
> Ignore this event if it occurred within 10s after the policy change.
> 
> The timeout was chosen observing that on a Dell XPS13 9333 these
> spurious events can occur up to roughly 6s after the policy change.

Just a couple things.

Can you please add the following tag?

Link: http://lkml.kernel.org/g/<MSG-ID of the first message in this thread>

> diff --git a/drivers/ata/libahci.c b/drivers/ata/libahci.c
> index 61a9c07..59a2517 100644
> --- a/drivers/ata/libahci.c
> +++ b/drivers/ata/libahci.c
> @@ -1700,6 +1700,9 @@ static void ahci_handle_port_interrupt(struct ata_port *ap,
>  	struct ahci_port_priv *pp = ap->private_data;
>  	struct ahci_host_priv *hpriv = ap->host->private_data;
>  	int resetting = !!(ap->pflags & ATA_PFLAG_RESETTING);
> +	unsigned long lpm_timeout = ap->link.last_lpm_change +
> +				    msecs_to_jiffies(ATA_TMOUT_SPURIOUS_PHY);
> +	bool ignore_event = false;
>  	u32 qc_active = 0;
>  	int rc;
>  
> @@ -1707,8 +1710,16 @@ static void ahci_handle_port_interrupt(struct ata_port *ap,
>  	if (unlikely(resetting))
>  		status &= ~PORT_IRQ_BAD_PMP;
>  
> +	/* ignore the first PHY event after the LPM policy changed
> +	 * as it is might be spurious
> +	 */
> +	if ((ap->link.flags & ATA_LFLAG_CHANGED) &&
> +	    time_before(jiffies, lpm_timeout))
> +		ignore_event = true;

Maybe the following is better?

	ignore_event = (ap->link.flags & ATA_LFLAG_CHANGED) &&
		       time_before(jiffies, lpm_timeout);

> +
>  	/* if LPM is enabled, PHYRDY doesn't mean anything */
> -	if (ap->link.lpm_policy > ATA_LPM_MAX_POWER) {
> +	if (ap->link.lpm_policy > ATA_LPM_MAX_POWER || ignore_event) {
> +		ap->link.flags &= ~ATA_LFLAG_CHANGED;
>  		status &= ~PORT_IRQ_PHYRDY;
>  		ahci_scr_write(&ap->link, SCR_ERROR, SERR_PHYRDY_CHG);
>  	}

I thought more about it and it's weird to have ATA_LFLAG_CHANGED and
last_lpm_change timestamp in libata core side and then open-code the
actual logic in ahci.  How about implementing a generic helper, say,
bool sata_lpm_ignore_phy_events(link) in libata core and use it in
specific drivers?  It can test both the current lpm_policy and the
timeout (preferably a separate patch to introduce the function and
factor out lpm policy testing).

Thanks.

-- 
tejun
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/