Skip to contentSkip to conversation

rbd: set the 'device' link in sysfs

PATCH In review Hannes Reinecke 2 participants 4 messages Jan 23, 2020 – Jan 27, 2020

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 rbd: set the 'device' link in sysfs

The rbd driver already provides additional information in sysfs under /sys/bus/rbd, so we should set the 'device' link in the block device to reference this information.

Cc David Disseldorp <[email protected]>
---
 drivers/block/rbd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/block/rbd.c +1 -1
index 9f1f8689e316..3240b7744aeb 100644
@@ -6938,7 +6938,7 @@ static ssize_t do_rbd_add(struct bus_type *bus,
 	if (rc)
 		goto err_out_image_lock;
 
-	add_disk(rbd_dev->disk);
+	device_add_disk(&rbd_dev->dev, rbd_dev->disk, NULL);
 	/* see rbd_init_disk() */
 	blk_put_queue(rbd_dev->disk->queue);
 
Signature · 1 line
2.16.4
maintainer
#1 lore raw
Quoted 16 lines · Hannes Reinecke

The rbd driver already provides additional information in sysfs under /sys/bus/rbd, so we should set the 'device' link in the block device to reference this information.

Cc David Disseldorp <[email protected]>
---
 drivers/block/rbd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/block/rbd.c +0 -0
index 9f1f8689e316..3240b7744aeb 100644

drivers/block/rbd.c static ssize_t do_rbd_add(struct bus_type *bus, · line 6943

Quoted 7 lines
drivers/block/rbd.c +1 -1
@@ -6938,7 +6938,7 @@ static ssize_t do_rbd_add(struct bus_type *bus,
        if (rc)
                goto err_out_image_lock;
 
-       add_disk(rbd_dev->disk);
+       device_add_disk(&rbd_dev->dev, rbd_dev->disk, NULL);
        /* see rbd_init_disk() */
        blk_put_queue(rbd_dev->disk->queue);

Hi Hannes,

I looked at this a while ago and didn't go through with the patch because I wasn't sure whether this symlink can point to something arbitrary. IIRC it usually points a couple of levels up, to some parent. In the rbd case, this would be a completely different tree:
/sys/devices/virtual -> /sys/bus/rbd.

Do you know if there is precedent for this in some other driver? Are you sure it's not going to break any assumptions?

How did this came up? I'm curious because rbd(8) is only utility that I know of that (somewhat) cares.

Thanks,

Ilya

#2 lore raw
Quoted 27 lines · Ilya Dryomov
Quoted 16 lines · Hannes Reinecke

The rbd driver already provides additional information in sysfs under /sys/bus/rbd, so we should set the 'device' link in the block device to reference this information.

Cc David Disseldorp <[email protected]>
---
 drivers/block/rbd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/block/rbd.c +0 -0
index 9f1f8689e316..3240b7744aeb 100644

drivers/block/rbd.c static ssize_t do_rbd_add(struct bus_type *bus, · line 6943

Quoted 7 lines
drivers/block/rbd.c +1 -1
@@ -6938,7 +6938,7 @@ static ssize_t do_rbd_add(struct bus_type *bus,
        if (rc)
                goto err_out_image_lock;
 
-       add_disk(rbd_dev->disk);
+       device_add_disk(&rbd_dev->dev, rbd_dev->disk, NULL);
        /* see rbd_init_disk() */
        blk_put_queue(rbd_dev->disk->queue);

Hi Hannes,

Quoted 6 lines

I looked at this a while ago and didn't go through with the patch because I wasn't sure whether this symlink can point to something arbitrary. IIRC it usually points a couple of levels up, to some parent. In the rbd case, this would be a completely different tree:
/sys/devices/virtual -> /sys/bus/rbd.

Yes, this is expected.
The 'device' link will _always_ point into a different tree; the accessor via /sys/block or /sys/bus are assumed to be virtual entries, with the 'device' link pointing to the underlying device.
In our case the underlying device is also a virtual entity, but that's okay.

Quoted 3 lines

Do you know if there is precedent for this in some other driver? Are you sure it's not going to break any assumptions?

Things like iscsi do the very same thing.
And no, it doesn't break assumptions; quite the contrary.

Quoted 3 lines

How did this came up? I'm curious because rbd(8) is only utility that I know of that (somewhat) cares.

That's me digging through the rbd driver and see if I could hook up nvme-over-fabrics as a frontend to ceph/rbd.
As such I needed to find some information about the rbd device, and found that while the driver provides additional information under /sys/devices/rbd/X, this information is not linked to /sys/block/rbdX. And this patch just links these to bits of information together, so that now you can get the driver specific informations by following the 'device' link like all the other drivers do, too.

Cheers,

Hannes

Signature · 4 lines
Dr. Hannes Reinecke		      Teamlead Storage & Networking
[email protected]			                  +49 911 74053 688
SUSE Software Solutions Germany GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), GF: Felix Imendörffer
maintainer
#3 lore raw
Quoted 41 lines · Hannes Reinecke
Quoted 27 lines · Ilya Dryomov
Quoted 16 lines · Hannes Reinecke

The rbd driver already provides additional information in sysfs under /sys/bus/rbd, so we should set the 'device' link in the block device to reference this information.

Cc David Disseldorp <[email protected]>
---
 drivers/block/rbd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/block/rbd.c +0 -0
index 9f1f8689e316..3240b7744aeb 100644

drivers/block/rbd.c static ssize_t do_rbd_add(struct bus_type *bus, · line 6943

Quoted 7 lines
drivers/block/rbd.c +1 -1
@@ -6938,7 +6938,7 @@ static ssize_t do_rbd_add(struct bus_type *bus,
        if (rc)
                goto err_out_image_lock;
 
-       add_disk(rbd_dev->disk);
+       device_add_disk(&rbd_dev->dev, rbd_dev->disk, NULL);
        /* see rbd_init_disk() */
        blk_put_queue(rbd_dev->disk->queue);

Hi Hannes,

Quoted 6 lines

I looked at this a while ago and didn't go through with the patch because I wasn't sure whether this symlink can point to something arbitrary. IIRC it usually points a couple of levels up, to some parent. In the rbd case, this would be a completely different tree:
/sys/devices/virtual -> /sys/bus/rbd.

Yes, this is expected.
The 'device' link will _always_ point into a different tree; the accessor via /sys/block or /sys/bus are assumed to be virtual entries, with the 'device' link pointing to the underlying device.
In our case the underlying device is also a virtual entity, but that's okay.

Quoted 5 lines
Quoted 3 lines

Do you know if there is precedent for this in some other driver? Are you sure it's not going to break any assumptions?

Things like iscsi do the very same thing.
And no, it doesn't break assumptions; quite the contrary.

I see, thanks for the explanation.

Queued up for 5.6.

Thanks,

Ilya