Re: [PATCH v18 15/18] s390/vfio-ap: handle config changed and scan complete notification

[Tony Krowiak <akrowiak@linux.ibm.com>]

On 3/24/22 10:09, Jason J. Herne wrote:
> On 2/14/22 19:50, Tony Krowiak wrote:
> ...
>> @@ -790,13 +788,17 @@ static void vfio_ap_unlink_apqn_fr_mdev(struct 
>> ap_matrix_mdev *matrix_mdev,
>>         q = vfio_ap_mdev_get_queue(matrix_mdev, AP_MKQID(apid, apqi));
>>       /* If the queue is assigned to the matrix mdev, unlink it. */
>> -    if (q)
>> +    if (q) {
>>           vfio_ap_unlink_queue_fr_mdev(q);
>>   -    /* If the queue is assigned to the APCB, store it in @qtable. */
>> -    if (test_bit_inv(apid, matrix_mdev->shadow_apcb.apm) &&
>> -        test_bit_inv(apqi, matrix_mdev->shadow_apcb.aqm))
>> -        hash_add(qtable->queues, &q->mdev_qnode, q->apqn);
>> +        /* If the queue is assigned to the APCB, store it in 
>> @qtable. */
>> +        if (qtable) {
>> +            if (test_bit_inv(apid, matrix_mdev->shadow_apcb.apm) &&
>> +                test_bit_inv(apqi, matrix_mdev->shadow_apcb.aqm))
>> +                hash_add(qtable->queues, &q->mdev_qnode,
>> +                     q->apqn);
>> +        }
>> +    }
>>   }
>
> This appears to be an unrelated change. Does this belong in this patch?

No, that does not belong in this patch, it belongs in patch
[PATCH v18 12/18] s390/vfio-ap: reset queues after adapter/domain 
unassignment.

>
>
>>   /**
>> @@ -1271,7 +1273,7 @@ static const struct attribute_group 
>> *vfio_ap_mdev_attr_groups[] = {
>>    * @matrix_mdev: a mediated matrix device
>>    * @kvm: reference to KVM instance
>>    *
>> - * Note: The matrix_dev->lock must be taken prior to calling
>> + * Note: The matrix_dev->mdevs_lock must be taken prior to calling
>
> This also seems to be unrelated.

The change of the name of the matrix_dev->lock to matrix_dev->mdevs_lock
was done in [PATCH v18 09/18] s390/vfio-ap: introduce new mutex to control
access to the KVM pointer; however, this comment is not valid. A commit has
been pushed for upstream merge which includes removal of this comment.

>
>
>>    * this function; however, the lock will be temporarily released 
>> while the
>>    * guest's AP configuration is set to avoid a potential lockdep splat.
>>    * The kvm->lock is taken to set the guest's AP configuration 
>> which, under
>> @@ -1355,7 +1357,7 @@ static int vfio_ap_mdev_iommu_notifier(struct 
>> notifier_block *nb,
>>    * @matrix_mdev: a matrix mediated device
>>    * @kvm: the pointer to the kvm structure being unset.
>>    *
>> - * Note: The matrix_dev->lock must be taken prior to calling
>> + * Note: The matrix_dev->mdevs_lock must be taken prior to calling
>
> Same here.

Ditto

>
>>    * this function; however, the lock will be temporarily released 
>> while the
>>    * guest's AP configuration is cleared to avoid a potential lockdep 
>> splat.
>>    * The kvm->lock is taken to clear the guest's AP configuration 
>> which, under
>> @@ -1708,6 +1710,27 @@ static void vfio_ap_mdev_put_qlocks(struct 
>> ap_matrix_mdev *matrix_mdev)
>>       mutex_unlock(&matrix_dev->guests_lock);
>>   }
>>   +static bool vfio_ap_mdev_do_filter_matrix(struct ap_matrix_mdev 
>> *matrix_mdev,
>> +                      struct vfio_ap_queue *q)
>> +{
>> +    unsigned long apid = AP_QID_CARD(q->apqn);
>> +    unsigned long apqi = AP_QID_QUEUE(q->apqn);
>> +
>> +    /*
>> +     * If the queue is being probed because its APID or APQI is in the
>> +     * process of being added to the host's AP configuration, then 
>> we don't
>> +     * want to filter the matrix now as the filtering will be done 
>> after
>> +     * the driver is notified that the AP bus scan operation has 
>> completed
>> +     * (see the vfio_ap_on_scan_complete callback function).
>> +     */
>> +    if (test_bit_inv(apid, matrix_mdev->apm_add) ||
>> +        test_bit_inv(apqi, matrix_mdev->aqm_add))
>> +        return false;
>> +
>> +
>> +    return true;
>> +}
>> +
>>   int vfio_ap_mdev_probe_queue(struct ap_device *apdev)
>>   {
>>       struct vfio_ap_queue *q;
>> @@ -1725,10 +1748,15 @@ int vfio_ap_mdev_probe_queue(struct ap_device 
>> *apdev)
>>           vfio_ap_mdev_link_queue(matrix_mdev, q);
>>           memset(apm, 0, sizeof(apm));
>>           set_bit_inv(AP_QID_CARD(q->apqn), apm);
>> -        if (vfio_ap_mdev_filter_matrix(apm, q->matrix_mdev->matrix.aqm,
>> -                           q->matrix_mdev))
>> -            vfio_ap_mdev_hotplug_apcb(q->matrix_mdev);
>> +
>> +        if (vfio_ap_mdev_do_filter_matrix(q->matrix_mdev, q)) {
>> +            if (vfio_ap_mdev_filter_matrix(apm,
>> +                        q->matrix_mdev->matrix.aqm,
>> +                        q->matrix_mdev))
>> +                vfio_ap_mdev_hotplug_apcb(q->matrix_mdev);
>> +        }
>>       }
>> +
>>       dev_set_drvdata(&apdev->device, q);
>>       vfio_ap_mdev_put_qlocks(matrix_mdev);
>>   @@ -1783,10 +1811,15 @@ void vfio_ap_mdev_remove_queue(struct 
>> ap_device *apdev)
>>             apid = AP_QID_CARD(q->apqn);
>>           apqi = AP_QID_QUEUE(q->apqn);
>> -        if (test_bit_inv(apid, matrix_mdev->shadow_apcb.apm) &&
>> -            test_bit_inv(apqi, matrix_mdev->shadow_apcb.aqm)) {
>> -            clear_bit_inv(apid, matrix_mdev->shadow_apcb.apm);
>> -            vfio_ap_mdev_hotplug_apcb(matrix_mdev);
>> +
>> +        /*
>> +         * If the queue is assigned to the guest's APCB, then remove
>> +         * the adapter's APID from the APCB and hot it into the guest.
>> +         */
>> +        if (test_bit_inv(apid, q->matrix_mdev->shadow_apcb.apm) &&
>> +            test_bit_inv(apqi, q->matrix_mdev->shadow_apcb.aqm)) {
>> +            clear_bit_inv(apid, q->matrix_mdev->shadow_apcb.apm);
>> +            vfio_ap_mdev_hotplug_apcb(q->matrix_mdev);
>
> It looks like this a bug fix unrelated to this patch...?

It's not a bug fix, but it is unnecessary. Both what was removed and the 
code
added are fine, but the former makes more sense. I have no idea how this
happened other than the code was accidentally changed while merging code
during one of the multitude of rebases undertaken while developing this
series.
>
>>   @@ -1842,3 +1875,267 @@ int vfio_ap_mdev_resource_in_use(unsigned 
>> long *apm, unsigned long *aqm)
>>         return ret;
>>   }
>> +
>> +/**
>> + * vfio_ap_mdev_hot_unplug_cfg - hot unplug the adapters, domains 
>> and control
>> + *                 domains that have been removed from the host's
>> + *                 AP configuration from a guest.
>> + *
>> + * @guest: the guest
>> + * @aprem: the adapters that have been removed from the host's AP 
>> configuration
>> + * @aqrem: the domains that have been removed from the host's AP 
>> configuration
>> + * @cdrem: the control domains that have been removed from the 
>> host's AP
>> + *       configuration.
>> + */
>> +static void vfio_ap_mdev_hot_unplug_cfg(struct ap_matrix_mdev 
>> *matrix_mdev,
>> +                    unsigned long *aprem,
>> +                    unsigned long *aqrem,
>> +                    unsigned long *cdrem)
>> +{
>> +    bool do_hotplug = false;
>
> __bitmap_andnot() returns an int, so I think you should use an int here.

Okay

>
>
>> +    if (!bitmap_empty(aprem, AP_DEVICES)) {
>> +        do_hotplug |= bitmap_andnot(matrix_mdev->shadow_apcb.apm,
>> +                        matrix_mdev->shadow_apcb.apm,
>> +                        aprem, AP_DEVICES);
>
> Also, replace the |= with an = here. This is the first assignment so no
> need to do a logical OR as there is no pre-existing data to preserve.

We don't know which, if any, of the bitmaps (aprem, aqrem, cdrem) passed to
this function contain set bits, so the |= is necessary because this code
block won't be be executed if aprem is empty. That is why the do_hotplug
variable is initialized to false when it is declared (i.e., each of the 
bitmaps
might be empty).

>
>
>> +    }
>> +
>> +    if (!bitmap_empty(aqrem, AP_DOMAINS)) {
>> +        do_hotplug |= bitmap_andnot(matrix_mdev->shadow_apcb.aqm,
>> +                        matrix_mdev->shadow_apcb.aqm,
>> +                        aqrem, AP_DEVICES);
>> +    }
>> +
>> +    if (!bitmap_empty(cdrem, AP_DOMAINS))
>> +        do_hotplug |= bitmap_andnot(matrix_mdev->shadow_apcb.adm,
>> +                        matrix_mdev->shadow_apcb.adm,
>> +                        cdrem, AP_DOMAINS);
>> +
>> +    if (do_hotplug)
>> +        vfio_ap_mdev_hotplug_apcb(matrix_mdev);
>> +}
>> +
>> +/**
>> + * vfio_ap_mdev_cfg_remove - determines which guests are using the 
>> adapters,
>> + *                 domains and control domains that have been removed
>> + *                 from the host AP configuration and unplugs them
>> + *                 from those guests.
>> + *
>> + * @ap_remove:    bitmap specifying which adapters have been removed 
>> from the host
>> + *        config.
>> + * @aq_remove:    bitmap specifying which domains have been removed 
>> from the host
>> + *        config.
>> + * @cd_remove:    bitmap specifying which control domains have been 
>> removed from
>> + *        the host config.
>> + */
> ...
>> +/**
>> + * vfio_ap_mdev_on_cfg_remove - responds to the removal of adapters, 
>> domains and
>> + *                control domains from the host AP configuration
>> + *                by unplugging them from the guests that are
>> + *                using them.
>> + */
>> +static void vfio_ap_mdev_on_cfg_remove(void)
>> +{
>> +    int ap_remove, aq_remove, cd_remove;
>
> These can all be replaced with a single variable, just like you did 
> with do_add.

True, consider it done.

>
>> +    DECLARE_BITMAP(aprem, AP_DEVICES);
>> +    DECLARE_BITMAP(aqrem, AP_DOMAINS);
>> +    DECLARE_BITMAP(cdrem, AP_DOMAINS);
>> +    unsigned long *cur_apm, *cur_aqm, *cur_adm;
>> +    unsigned long *prev_apm, *prev_aqm, *prev_adm;
>> +
>> +    cur_apm = (unsigned long *)matrix_dev->config_info.apm;
>> +    cur_aqm = (unsigned long *)matrix_dev->config_info.aqm;
>> +    cur_adm = (unsigned long *)matrix_dev->config_info.adm;
>> +    prev_apm = (unsigned long *)matrix_dev->config_info_prev.apm;
>> +    prev_aqm = (unsigned long *)matrix_dev->config_info_prev.aqm;
>> +    prev_adm = (unsigned long *)matrix_dev->config_info_prev.adm;
>> +
>> +    ap_remove = bitmap_andnot(aprem, prev_apm, cur_apm, AP_DEVICES);
>> +    aq_remove = bitmap_andnot(aqrem, prev_aqm, cur_aqm, AP_DOMAINS);
>> +    cd_remove = bitmap_andnot(cdrem, prev_adm, cur_adm, AP_DOMAINS);
>> +
>> +    if (ap_remove || aq_remove || cd_remove)
>> +        vfio_ap_mdev_cfg_remove(aprem, aqrem, cdrem);
>> +}
> ...
>> +/**
>> + * vfio_ap_on_cfg_changed - handles notification of changes to the 
>> host AP
>> + *                configuration.
>> + *
>> + * @new_config_info: the new host AP configuration
>> + * @old_config_info: the previous host AP configuration
>> + */
>> +void vfio_ap_on_cfg_changed(struct ap_config_info *new_config_info,
>> +                struct ap_config_info *old_config_info)
>> +{
>> +    mutex_lock(&matrix_dev->guests_lock);
>> +
>> +    memcpy(&matrix_dev->config_info_prev, old_config_info,
>> +               sizeof(struct ap_config_info));
>> +    memcpy(&matrix_dev->config_info, new_config_info,
>> +           sizeof(struct ap_config_info));
>
> Why are we storing old_config_info in the matrix_dev? It appears to only
> be used within the functions called from right here. Why not just pass it
> as an argument? 

At the time I thought it might be valuable to keep around. I suppose it can
be passed to the functions that need it.

>
>
>> +    vfio_ap_mdev_on_cfg_remove();
>> +    vfio_ap_mdev_on_cfg_add();
>> +
>> +    mutex_unlock(&matrix_dev->guests_lock);
>> +}
>
> Here is an idea to restructure things... consider combining logic from
> vfio_ap_mdev_on_cfg_remove and vfio_ap_mdev_on_cfg_add with
> vfio_ap_on_cfg_changed. This makes vfio_ap_on_cfg_changed() longer but
> it eliminates some duplicated code and gets rid of both
> vfio_ap_mdev_on_cfg_remove and vfio_ap_mdev_on_cfg_add.
> note: Untested... :) 

It seems like a lot of busy work for little gain. I prefer shorter 
functions that
perform a specific task. Since we're solely in the realm of personal 
preferences,
I'm going to leave this function as-is.