[dm-devel] Re: [PATCH] Fixes to EMC scsi device handler

Chandra Seetharaman sekharan at us.ibm.com
Tue May 6 01:34:18 UTC 2008


On Mon, 2008-05-05 at 13:55 +0200, Hannes Reinecke wrote:
> Hi Chandra,
> 
> based on your latest patchset there are some issues which
> should be fixed:
> 
> - When SYSFS_DEPRECATED is not set also scsi_target and
>   scsi_host devices will be attached to the scsi bus.
>   So we have to check with 'scsi_is_sdev_device' before
>   processing the notifier call

This need to be done for all the handlers, right ?
>  
> - Newer EMC flarecode also provide for a 'VRAID' device
>   model; this should be added to the device map
> - The name should be 'emc', not 'emc_clariion', for
>   integration with existing multipath-tools
> - It's spelled 'detached' :-)
> 
> Patch attached.

Looks good to me.
> 
> Some issues which I'd like to see eventually:
> 
> - We should have a way of adding/modifying the internal
>   device map. Currently we have to recompile the module
>   every time a new device definition is found.

Yeah, very true. it will become an annoyance. Is this the reason why
you were asking for the sysfs interface (below) ?

> - The ->activate function should be able to be triggered
>   manually via sysfs.

Why would you need this ? 
> 
> James, seeing that the scsi_dh pointer is already present
> in the sdev structure, would you accept a patch which adds
> scsi_dh attributes to drivers/scsi/scsi_sysfs.c?
> Having them encapsulated into drivers/scsi/device_handler
> somewhere is really painful due to the asynchronous nature
> of this whole thing ...
> 
> Apart from this: Great work!
> I'll check the other device handlers, too; already noticed
> that the device map for HP is wrong :-)

Can you please help fix it. 
> 
> Cheers,
> 
> hannes
> plain text document attachment (scsi-dh-emc-fixes)
> Fixes to EMC scsi device handler
> 
> Several fixes to the EMC scsi device handler:
> - The name should be 'emc' for seamless integration with multipathing
> - Newer EMC flarecode also has a 'VRAID' type
> - Check if this is really a sdev before processing the notifier call
> 
> Signed-off-by: Hannes Reinecke <hare at suse.de>
> 
> diff --git a/drivers/scsi/device_handler/scsi_dh_emc.c b/drivers/scsi/device_handler/scsi_dh_emc.c
> index 1b18989..0aded96 100644
> --- a/drivers/scsi/device_handler/scsi_dh_emc.c
> +++ b/drivers/scsi/device_handler/scsi_dh_emc.c
> @@ -25,7 +25,7 @@
>  #include <scsi/scsi_dh.h>
>  #include <scsi/scsi_device.h>
> 
> -#define CLARIION_NAME			"emc_clariion"
> +#define CLARIION_NAME			"emc"
> 
>  #define CLARIION_TRESPASS_PAGE		0x22
>  #define CLARIION_BUFFER_SIZE		0x80
> @@ -390,12 +390,13 @@ static int clariion_check_sense(struct scsi_device *sdev,
>  	return SUCCESS;
>  }
> 
> -static const struct {
> +static const struct clariion_dev_list_t {
>  	char *vendor;
>  	char *model;
>  } clariion_dev_list[] = {
>  	{"DGC", "RAID"},
>  	{"DGC", "DISK"},
> +	{"DGC", "VRAID"},
>  	{NULL, NULL},
>  };
> 
> @@ -422,6 +423,9 @@ static int clariion_bus_notify(struct notifier_block *nb,
>  	int i, found = 0;
>  	unsigned long flags;
> 
> +	if (!scsi_is_sdev_device(dev))
> +		return 0;
> +
>  	if (action == BUS_NOTIFY_ADD_DEVICE) {
>  		for (i = 0; clariion_dev_list[i].vendor; i++) {
>  			if (!strncmp(sdev->vendor, clariion_dev_list[i].vendor,
> @@ -465,7 +469,7 @@ static int clariion_bus_notify(struct notifier_block *nb,
>  		sdev->scsi_dh_data = NULL;
>  		spin_unlock_irqrestore(sdev->request_queue->queue_lock, flags);
> 
> -		sdev_printk(KERN_NOTICE, sdev, "Dettached %s.\n",
> +		sdev_printk(KERN_NOTICE, sdev, "Detached %s.\n",
>  			    CLARIION_NAME);
> 
>  		kfree(scsi_dh_data);




More information about the dm-devel mailing list