Re: [PATCH] dm mpath: delay retry activate_path on SCSI_DH_RETRY

[Mike Christie <mchristi@redhat.com>]

Chandra Seetharaman wrote:
> Resubmitting the patch with 2 changes:
>  1. pg_init_delay_secs was used inconsistently (jiffies and seconds).
>     Fixed the problem
>  2. Moved the #define to dm_mpath.c from scsi_dh.h
> -----------------------
> From: Chandra Seetharaman <sekharan@us.ibm.com>
> 
> SCSI Device Handlers return SCSI_DH_IMM_RETRY if we could retry
> immediately and SCSI_DH_RETRY in cases where it is better to retry
> after some delay.
> 
> Currently we retry immediately irrespective of SCSI_DH_IMM_RETRY and
> SCSI_DH_RETRY. This patch adds a user configurable attribute
> pg_init_delay_secs which specifies the number of seconds to delay
> before retrying scsi_dh_activate, when it returned SCSI_DH_RETRY.
> 
> Default for this attribute is set to 2 seconds.
> 
> Signed-off-by: Nikanth Karthikesan <knikanth@suse.de>
> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com>
> 
> ---
>  drivers/md/dm-mpath.c |   44 +++++++++++++++++++++++++++++++++++---------
>  1 file changed, 35 insertions(+), 9 deletions(-)
> 
> Index: linux-2.6.29/drivers/md/dm-mpath.c
> ===================================================================
> --- linux-2.6.29.orig/drivers/md/dm-mpath.c
> +++ linux-2.6.29/drivers/md/dm-mpath.c
> @@ -24,6 +24,7 @@
>  
>  #define DM_MSG_PREFIX "multipath"
>  #define MESG_STR(x) x, sizeof(x)
> +#define DM_PG_INIT_RETRY_DELAY 2
>  
>  /* Path properties */
>  struct pgpath {
> @@ -35,7 +36,7 @@ struct pgpath {
>  
>  	struct dm_path path;
>  	struct work_struct deactivate_path;
> -	struct work_struct activate_path;
> +	struct delayed_work activate_path;
>  };
>  
>  #define path_to_pgpath(__pgp) container_of((__pgp), struct pgpath, path)
> @@ -69,6 +70,7 @@ struct multipath {
>  	struct list_head priority_groups;
>  	unsigned pg_init_required;	/* pg_init needs calling? */
>  	unsigned pg_init_in_progress;	/* Only one pg_init allowed at once */
> +	unsigned pg_init_delay;		/* To delay or not to delay */
>  
>  	unsigned nr_valid_paths;	/* Total number of usable paths */
>  	struct pgpath *current_pgpath;
> @@ -81,6 +83,7 @@ struct multipath {
>  	unsigned saved_queue_if_no_path;/* Saved state during suspension */
>  	unsigned pg_init_retries;	/* Number of times to retry pg_init */
>  	unsigned pg_init_count;		/* Number of times pg_init called */
> +	unsigned pg_init_delay_secs;	/* Delay in seconds before retry */
>  
>  	struct work_struct process_queued_ios;
>  	struct bio_list queued_ios;
> @@ -127,7 +130,7 @@ static struct pgpath *alloc_pgpath(void)
>  	if (pgpath) {
>  		pgpath->is_active = 1;
>  		INIT_WORK(&pgpath->deactivate_path, deactivate_path);
> -		INIT_WORK(&pgpath->activate_path, activate_path);
> +		INIT_DELAYED_WORK(&pgpath->activate_path, activate_path);
>  	}
>  
>  	return pgpath;
> @@ -195,6 +198,7 @@ static struct multipath *alloc_multipath
>  		INIT_LIST_HEAD(&m->priority_groups);
>  		spin_lock_init(&m->lock);
>  		m->queue_io = 1;
> +		m->pg_init_delay_secs = DM_PG_INIT_RETRY_DELAY;
>  		INIT_WORK(&m->process_queued_ios, process_queued_ios);
>  		INIT_WORK(&m->trigger_event, trigger_event);
>  		m->mpio_pool = mempool_create_slab_pool(MIN_IOS, _mpio_cache);
> @@ -443,9 +447,12 @@ static void process_queued_ios(struct wo
>  		m->pg_init_count++;
>  		m->pg_init_required = 0;
>  		list_for_each_entry(tmp, &pgpath->pg->pgpaths, list) {
> -			queue_work(kmpath_handlerd, &tmp->activate_path);
> +			queue_delayed_work(kmpath_handlerd, &tmp->activate_path,
> +				m->pg_init_delay ?
> +					m->pg_init_delay_secs * HZ : 0);
>  			m->pg_init_in_progress++;
>  		}
> +		m->pg_init_delay = 0;
>  	}
>  out:
>  	spin_unlock_irqrestore(&m->lock, flags);
> @@ -714,8 +721,9 @@ static int parse_features(struct arg_set
>  	const char *param_name;
>  
>  	static struct param _params[] = {
> -		{0, 3, "invalid number of feature args"},
> +		{0, 5, "invalid number of feature args"},
>  		{1, 50, "pg_init_retries must be between 1 and 50"},
> +		{1, 50, "pg_init_delay_secs must be between 1 and 50"},
>  	};
>  
>  	r = read_param(_params, shift(as), &argc, &ti->error);
> @@ -742,6 +750,14 @@ static int parse_features(struct arg_set
>  			continue;
>  		}
>  
> +		if (!strnicmp(param_name, MESG_STR("pg_init_delay_secs")) &&
> +		    (argc >= 1)) {
> +			r = read_param(_params + 1, shift(as),
> +				       &m->pg_init_delay_secs, &ti->error);
> +			argc--;
> +			continue;
> +		}
> +
>  		ti->error = "Unrecognised multipath feature request";
>  		r = -EINVAL;
>  	} while (argc && !r);
> @@ -919,7 +935,7 @@ static int reinstate_path(struct pgpath 
>  		queue_work(kmultipathd, &m->process_queued_ios);
>  	} else if (m->hw_handler_name && (m->current_pg == pgpath->pg)) {
>   		m->pg_init_in_progress++;
> -		queue_work(kmpath_handlerd, &pgpath->activate_path);
> +		queue_delayed_work(kmpath_handlerd, &pgpath->activate_path, 0);
>  	}
>  
>  	dm_path_uevent(DM_UEVENT_PATH_REINSTATED, m->ti,
> @@ -1053,6 +1069,7 @@ static void pg_init_done(struct dm_path 
>  	struct priority_group *pg = pgpath->pg;
>  	struct multipath *m = pg->m;
>  	unsigned long flags;
> +	unsigned int delay = 0;
>  
>  	/* device or driver problems */
>  	switch (errors) {
> @@ -1077,8 +1094,11 @@ static void pg_init_done(struct dm_path 
>  		 */
>  		bypass_pg(m, pg, 1);
>  		break;
> -	/* TODO: For SCSI_DH_RETRY we should wait a couple seconds */
> +	/*
> +	 * For SCSI_DH_RETRY we wait before retrying.
> +	 */
>  	case SCSI_DH_RETRY:
> +		delay = 1;
>  	case SCSI_DH_IMM_RETRY:
>  	case SCSI_DH_RES_TEMP_UNAVAIL:
>  		if (pg_init_limit_reached(m, pgpath))
> @@ -1107,8 +1127,10 @@ static void pg_init_done(struct dm_path 
>  	}
>  
>  	m->pg_init_in_progress--;
> -	if (!m->pg_init_in_progress)
> +	if (!m->pg_init_in_progress) {
> +		m->pg_init_delay = delay;
>  		queue_work(kmultipathd, &m->process_queued_ios);
> +	}
>  	spin_unlock_irqrestore(&m->lock, flags);
>  }
>  
> @@ -1116,7 +1138,7 @@ static void activate_path(struct work_st
>  {
>  	int ret;
>  	struct pgpath *pgpath =
> -		container_of(work, struct pgpath, activate_path);
> +		container_of(work, struct pgpath, activate_path.work);
>  
>  	ret = scsi_dh_activate(bdev_get_queue(pgpath->path.dev->bdev));
>  	pg_init_done(&pgpath->path, ret);
> @@ -1252,11 +1274,15 @@ static int multipath_status(struct dm_ta
>  		DMEMIT("2 %u %u ", m->queue_size, m->pg_init_count);
>  	else {
>  		DMEMIT("%u ", m->queue_if_no_path +
> -			      (m->pg_init_retries > 0) * 2);
> +			      (m->pg_init_retries > 0) * 2 +
> +			      (m->pg_init_delay_secs !=
> +						DM_PG_INIT_RETRY_DELAY) * 2);
>  		if (m->queue_if_no_path)
>  			DMEMIT("queue_if_no_path ");
>  		if (m->pg_init_retries)
>  			DMEMIT("pg_init_retries %u ", m->pg_init_retries);
> +		if (m->pg_init_delay_secs != DM_PG_INIT_RETRY_DELAY)
> +			DMEMIT("pg_init_delay_secs %u ", m->pg_init_delay_secs);
>  	}
>  
>  	if (!m->hw_handler_name || type == STATUSTYPE_INFO)
> 
> 
> 

Looks ok.