Re: [PATCH v2 08/11] md: add atomic mode switching in RAID 1/10

[Yu Kuai]

Hi,

在 2024/04/18 13:44, tada keisuke 写道:
> This patch depends on patch 07.
>
> All rdevs running in RAID 1/10 switch nr_pending to atomic mode.
> The value of nr_pending is read in a normal operation (choose_best_rdev()).
> Therefore, nr_pending must always be consistent.
>
> Signed-off-by: Keisuke TADA <keisuke1.tada@xxxxxxxxxx>
> Signed-off-by: Toshifumi OHTAKE <toshifumi.ootake@xxxxxxxxxx>
> ---
>   drivers/md/md.h     | 14 ++++++++++++++
>   drivers/md/raid1.c  |  7 +++++++
>   drivers/md/raid10.c |  4 ++++
>   3 files changed, 25 insertions(+)
>
> diff --git a/drivers/md/md.h b/drivers/md/md.h
> index ab09e312c9bb..57b09b567ffa 100644
> --- a/drivers/md/md.h
> +++ b/drivers/md/md.h
> @@ -236,6 +236,20 @@ static inline unsigned long nr_pending_read(struct md_rdev *rdev)
>   	return atomic_long_read(&rdev->nr_pending.data->count);
>   }
>   
> +static inline bool nr_pending_is_percpu_mode(struct md_rdev *rdev)
> +{
> +	unsigned long __percpu *percpu_count;
> +
> +	return __ref_is_percpu(&rdev->nr_pending, &percpu_count);
> +}
> +
> +static inline bool nr_pending_is_atomic_mode(struct md_rdev *rdev)
> +{
> +	unsigned long __percpu *percpu_count;
> +
> +	return !__ref_is_percpu(&rdev->nr_pending, &percpu_count);
> +}
> +
>   static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors,
>   			      sector_t *first_bad, int *bad_sectors)
>   {
> diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c
> index 12318fb15a88..c38ae13aadab 100644
> --- a/drivers/md/raid1.c
> +++ b/drivers/md/raid1.c
> @@ -784,6 +784,7 @@ static int choose_best_rdev(struct r1conf *conf, struct r1bio *r1_bio)
>   		if (ctl.readable_disks++ == 1)
>   			set_bit(R1BIO_FailFast, &r1_bio->state);
>   
> +		WARN_ON_ONCE(nr_pending_is_percpu_mode(rdev));
>   		pending = nr_pending_read(rdev);
>   		dist = abs(r1_bio->sector - conf->mirrors[disk].head_position);
>   
> @@ -1930,6 +1931,7 @@ static int raid1_add_disk(struct mddev *mddev, struct md_rdev *rdev)
>   			if (err)
>   				return err;
>   
> +			percpu_ref_switch_to_atomic_sync(&rdev->nr_pending);
>   			raid1_add_conf(conf, rdev, mirror, false);
>   			/* As all devices are equivalent, we don't need a full recovery
>   			 * if this was recently any drive of the array
> @@ -1949,6 +1951,7 @@ static int raid1_add_disk(struct mddev *mddev, struct md_rdev *rdev)
>   		set_bit(Replacement, &rdev->flags);
>   		raid1_add_conf(conf, rdev, repl_slot, true);
>   		err = 0;
> +		percpu_ref_switch_to_atomic_sync(&rdev->nr_pending);

I don't understand what's the point here, 'nr_pending' will be used when
the rdev issuing IO, and it's always used as atomic mode, there is no
difference.

Consider that 'nr_pending' must be read from IO fast path, use it as
atomic is something we must accept. Unless someone comes up with a plan
to avoid reading 'inflight' counter from fast path like generic block
layer, it's not ok to me to switch to percpu_ref for now.

+CC Paul

HI, Paul, perhaps you RR mode doesn't need such 'inflight' counter
anymore?

Thanks,
Kuai

>   		conf->fullsync = 1;
>   	}
>   
> @@ -3208,6 +3211,7 @@ static void raid1_free(struct mddev *mddev, void *priv);
>   static int raid1_run(struct mddev *mddev)
>   {
>   	struct r1conf *conf;
> +	struct md_rdev *rdev;
>   	int i;
>   	int ret;
>   
> @@ -3269,6 +3273,9 @@ static int raid1_run(struct mddev *mddev)
>   	/*
>   	 * Ok, everything is just fine now
>   	 */
> +	rdev_for_each(rdev, mddev) {
> +		percpu_ref_switch_to_atomic_sync(&rdev->nr_pending);
> +	}
>   	rcu_assign_pointer(mddev->thread, conf->thread);
>   	rcu_assign_pointer(conf->thread, NULL);
>   	mddev->private = conf;
> diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
> index b91dd6c0be5a..66896a1076e1 100644
> --- a/drivers/md/raid10.c
> +++ b/drivers/md/raid10.c
> @@ -808,6 +808,7 @@ static struct md_rdev *read_balance(struct r10conf *conf,
>   
>   		nonrot = bdev_nonrot(rdev->bdev);
>   		has_nonrot_disk |= nonrot;
> +		WARN_ON_ONCE(nr_pending_is_percpu_mode(rdev));
>   		pending = nr_pending_read(rdev);
>   		if (min_pending > pending && nonrot) {
>   			min_pending = pending;
> @@ -2113,6 +2114,7 @@ static int raid10_add_disk(struct mddev *mddev, struct md_rdev *rdev)
>   		p->recovery_disabled = mddev->recovery_disabled - 1;
>   		rdev->raid_disk = mirror;
>   		err = 0;
> +		percpu_ref_switch_to_atomic_sync(&rdev->nr_pending);
>   		if (rdev->saved_raid_disk != mirror)
>   			conf->fullsync = 1;
>   		WRITE_ONCE(p->rdev, rdev);
> @@ -2127,6 +2129,7 @@ static int raid10_add_disk(struct mddev *mddev, struct md_rdev *rdev)
>   		err = mddev_stack_new_rdev(mddev, rdev);
>   		if (err)
>   			return err;
> +		percpu_ref_switch_to_atomic_sync(&rdev->nr_pending);
>   		conf->fullsync = 1;
>   		WRITE_ONCE(p->replacement, rdev);
>   	}
> @@ -4028,6 +4031,7 @@ static int raid10_run(struct mddev *mddev)
>   	rdev_for_each(rdev, mddev) {
>   		long long diff;
>   
> +		percpu_ref_switch_to_atomic_sync(&rdev->nr_pending);
>   		disk_idx = rdev->raid_disk;
>   		if (disk_idx < 0)
>   			continue;