[PATCHv3] Update scsi host to use ida for host number

[PATCHv3] Update scsi host to use ida for host number

[Lee Duncan]

Each Scsi_host instance gets a host number starting
at 0, but this is implemented with an atomic integer,
and rollover doesn't seem to have been considered.
Another side-effect of this design is that scsi host
numbers used by iscsi are never reused, thereby  making
rollover more likely. This patch converts Scsi_host
instances to use ida to manage their instance
numbers.

This also means that host instance numbers will be
reused, when available.

Changes from v2:
* Use host_put_index() throughout
* Put host index code together
Changes from v1:
* Added inline host_put_index() and its use

Signed-off-by: Lee Duncan <lduncan@suse.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
---
 drivers/scsi/hosts.c | 47 ++++++++++++++++++++++++++++++++++++++---------
 1 file changed, 38 insertions(+), 9 deletions(-)

diff --git a/drivers/scsi/hosts.c b/drivers/scsi/hosts.c
index 8bb173e01084..a47944867e65 100644
--- a/drivers/scsi/hosts.c
+++ b/drivers/scsi/hosts.c
@@ -33,7 +33,7 @@
 #include <linux/transport_class.h>
 #include <linux/platform_device.h>
 #include <linux/pm_runtime.h>
-
+#include <linux/idr.h>
 #include <scsi/scsi_device.h>
 #include <scsi/scsi_host.h>
 #include <scsi/scsi_transport.h>
@@ -42,8 +42,6 @@
 #include "scsi_logging.h"
 
 
-static atomic_t scsi_host_next_hn = ATOMIC_INIT(0);	/* host_no for next new host */
-
 
 static void scsi_host_cls_release(struct device *dev)
 {
@@ -304,6 +302,31 @@ int scsi_add_host_with_dma(struct Scsi_Host *shost, struct device *dev,
 }
 EXPORT_SYMBOL(scsi_add_host_with_dma);
 
+static DEFINE_SPINLOCK(host_index_lock);
+static DEFINE_IDA(host_index_ida);
+
+static int host_get_index(int *index)
+{
+	int error = -ENOMEM;
+
+	do {
+		if (!ida_pre_get(&host_index_ida, GFP_KERNEL))
+			break;
+		spin_lock(&host_index_lock);
+		error = ida_get_new(&host_index_ida, index);
+		spin_unlock(&host_index_lock);
+	} while (error == -EAGAIN);
+
+	return error;
+}
+
+static inline void host_put_index(int index)
+{
+	spin_lock(&host_index_lock);
+	ida_remove(&host_index_ida, index);
+	spin_unlock(&host_index_lock);
+}
+
 static void scsi_host_dev_release(struct device *dev)
 {
 	struct Scsi_Host *shost = dev_to_shost(dev);
@@ -337,6 +360,8 @@ static void scsi_host_dev_release(struct device *dev)
 
 	kfree(shost->shost_data);
 
+	host_put_index(shost->host_no);
+
 	if (parent)
 		put_device(parent);
 	kfree(shost);
@@ -370,6 +395,8 @@ struct Scsi_Host *scsi_host_alloc(struct scsi_host_template *sht, int privsize)
 {
 	struct Scsi_Host *shost;
 	gfp_t gfp_mask = GFP_KERNEL;
+	int index;
+	int error;
 
 	if (sht->unchecked_isa_dma && privsize)
 		gfp_mask |= __GFP_DMA;
@@ -388,11 +415,11 @@ struct Scsi_Host *scsi_host_alloc(struct scsi_host_template *sht, int privsize)
 	init_waitqueue_head(&shost->host_wait);
 	mutex_init(&shost->scan_mutex);
 
-	/*
-	 * subtract one because we increment first then return, but we need to
-	 * know what the next host number was before increment
-	 */
-	shost->host_no = atomic_inc_return(&scsi_host_next_hn) - 1;
+	error = host_get_index(&index);
+	if (error < 0)
+		goto fail_kfree;
+	shost->host_no = index;
+
 	shost->dma_channel = 0xff;
 
 	/* These three are default values which can be overridden */
@@ -477,7 +504,7 @@ struct Scsi_Host *scsi_host_alloc(struct scsi_host_template *sht, int privsize)
 		shost_printk(KERN_WARNING, shost,
 			"error handler thread failed to spawn, error = %ld\n",
 			PTR_ERR(shost->ehandler));
-		goto fail_kfree;
+		goto fail_free_index;
 	}
 
 	shost->tmf_work_q = alloc_workqueue("scsi_tmf_%d",
@@ -493,6 +520,8 @@ struct Scsi_Host *scsi_host_alloc(struct scsi_host_template *sht, int privsize)
 
  fail_kthread:
 	kthread_stop(shost->ehandler);
+ fail_free_index:
+	host_put_index(index);
  fail_kfree:
 	kfree(shost);
 	return NULL;
-- 
2.1.4

Re: [PATCHv3] Update scsi host to use ida for host number

[Christoph Hellwig]

On Tue, Sep 01, 2015 at 05:03:28PM -0700, Lee Duncan wrote:
> +static int host_get_index(int *index)
> +{
> +	int error = -ENOMEM;
> +
> +	do {
> +		if (!ida_pre_get(&host_index_ida, GFP_KERNEL))
> +			break;
> +		spin_lock(&host_index_lock);
> +		error = ida_get_new(&host_index_ida, index);
> +		spin_unlock(&host_index_lock);
> +	} while (error == -EAGAIN);
> +
> +	return error;
> +}
> +
> +static inline void host_put_index(int index)
> +{
> +	spin_lock(&host_index_lock);
> +	ida_remove(&host_index_ida, index);
> +	spin_unlock(&host_index_lock);
> +}

I really hate how this pattern (and the equivalent for IDA) are
spread all over.  Any chance to have some simple_ida/simple_idr helpers
instead of duplicating it again an again.

Besides that why aren't we using and idr here and use it for host lookup
as well?

Re: [PATCHv3] Update scsi host to use ida for host number

[Lee Duncan]

On 09/01/2015 11:48 PM, Christoph Hellwig wrote:
> On Tue, Sep 01, 2015 at 05:03:28PM -0700, Lee Duncan wrote:
>> +static int host_get_index(int *index)
>> +{
>> +	int error = -ENOMEM;
>> +
>> +	do {
>> +		if (!ida_pre_get(&host_index_ida, GFP_KERNEL))
>> +			break;
>> +		spin_lock(&host_index_lock);
>> +		error = ida_get_new(&host_index_ida, index);
>> +		spin_unlock(&host_index_lock);
>> +	} while (error == -EAGAIN);
>> +
>> +	return error;
>> +}
>> +
>> +static inline void host_put_index(int index)
>> +{
>> +	spin_lock(&host_index_lock);
>> +	ida_remove(&host_index_ida, index);
>> +	spin_unlock(&host_index_lock);
>> +}
> 
> I really hate how this pattern (and the equivalent for IDA) are
> spread all over.  Any chance to have some simple_ida/simple_idr helpers
> instead of duplicating it again an again.
> 
> Besides that why aren't we using and idr here and use it for host lookup
> as well?
> .
> 

Good question. My original goal was just to fix the host_no wrap-around
problem, but I agree, the hosts.c module could benefit from using idr,
making scsi_host_lookup(hostnum) much simpler, and possible faster.

I will submit another patch.

This means I won't need the sequence any longer that you wanted made
into helper functions, though I'd still be glad to create the helper
functions and convert some of the ida users, if you wish.
-- 
Lee Duncan