[PATCH v2 0/2] Provide a stable serial number

[PATCH v2 0/2] Provide a stable serial number

[Johannes Thumshirn]

This pathset is a follow up to
http://lists.infradead.org/pipermail/linux-nvme/2017-July/011934.html.

The 1st patch moves the serial attribute from the contrller to the subsystem,
while the 2nd patch makes it configurable via configfs.

Changes to v1:
* Move generation of random serial to nvmet_subsys_alloc() (Christoph)

Johannes Thumshirn (2):
  nvmet: Move serial number from controller to subsystem
  nvmet: preserve controller serial number between reboots

 drivers/nvme/target/admin-cmd.c |  2 +-
 drivers/nvme/target/configfs.c  | 22 ++++++++++++++++++++++
 drivers/nvme/target/core.c      |  6 +++---
 drivers/nvme/target/nvmet.h     |  2 +-
 4 files changed, 27 insertions(+), 5 deletions(-)

-- 
2.12.3

[PATCH v2 1/2] nvmet: Move serial number from controller to subsystem

[Johannes Thumshirn]

The NVMe specification defines the serial number as:

"Serial Number (SN): Contains the serial number for the NVM subsystem
that is assigned by the vendor as an ASCII string. Refer to section
7.10 for unique identifier requirements. Refer to section 1.5 for ASCII
string requirements"

So move it from the controller to the subsystem, where it belongs.

Signed-off-by: Johannes Thumshirn <jthumshirn@suse.de>
---
 drivers/nvme/target/admin-cmd.c | 2 +-
 drivers/nvme/target/core.c      | 4 +---
 drivers/nvme/target/nvmet.h     | 2 +-
 3 files changed, 3 insertions(+), 5 deletions(-)

diff --git a/drivers/nvme/target/admin-cmd.c b/drivers/nvme/target/admin-cmd.c
index 35f930db3c02..f7ba006d6a65 100644
--- a/drivers/nvme/target/admin-cmd.c
+++ b/drivers/nvme/target/admin-cmd.c
@@ -185,7 +185,7 @@ static void nvmet_execute_identify_ctrl(struct nvmet_req *req)
 	id->ssvid = 0;
 
 	memset(id->sn, ' ', sizeof(id->sn));
-	snprintf(id->sn, sizeof(id->sn), "%llx", ctrl->serial);
+	snprintf(id->sn, sizeof(id->sn), "%llx", ctrl->subsys->serial);
 
 	memset(id->mn, ' ', sizeof(id->mn));
 	strncpy((char *)id->mn, "Linux", sizeof(id->mn));
diff --git a/drivers/nvme/target/core.c b/drivers/nvme/target/core.c
index b5b4ac103748..02efa7113d23 100644
--- a/drivers/nvme/target/core.c
+++ b/drivers/nvme/target/core.c
@@ -767,9 +767,6 @@ u16 nvmet_alloc_ctrl(const char *subsysnqn, const char *hostnqn,
 	memcpy(ctrl->subsysnqn, subsysnqn, NVMF_NQN_SIZE);
 	memcpy(ctrl->hostnqn, hostnqn, NVMF_NQN_SIZE);
 
-	/* generate a random serial number as our controllers are ephemeral: */
-	get_random_bytes(&ctrl->serial, sizeof(ctrl->serial));
-
 	kref_init(&ctrl->ref);
 	ctrl->subsys = subsys;
 
@@ -928,6 +925,7 @@ struct nvmet_subsys *nvmet_subsys_alloc(const char *subsysnqn,
 		return NULL;
 
 	subsys->ver = NVME_VS(1, 3, 0); /* NVMe 1.3.0 */
+	get_random_bytes(&subsys->serial, sizeof(subsys->serial));
 
 	switch (type) {
 	case NVME_NQN_NVME:
diff --git a/drivers/nvme/target/nvmet.h b/drivers/nvme/target/nvmet.h
index 747bbdb4f9c6..e3b244c7e443 100644
--- a/drivers/nvme/target/nvmet.h
+++ b/drivers/nvme/target/nvmet.h
@@ -112,7 +112,6 @@ struct nvmet_ctrl {
 
 	struct mutex		lock;
 	u64			cap;
-	u64			serial;
 	u32			cc;
 	u32			csts;
 
@@ -152,6 +151,7 @@ struct nvmet_subsys {
 	u16			max_qid;
 
 	u64			ver;
+	u64			serial;
 	char			*subsysnqn;
 
 	struct config_group	group;
-- 
2.12.3

Re: [PATCH v2 1/2] nvmet: Move serial number from controller to subsystem

[Christoph Hellwig]

> -	/* generate a random serial number as our controllers are ephemeral: */
> -	get_random_bytes(&ctrl->serial, sizeof(ctrl->serial));
> -
>  	kref_init(&ctrl->ref);
>  	ctrl->subsys = subsys;
>  
> @@ -928,6 +925,7 @@ struct nvmet_subsys *nvmet_subsys_alloc(const char *subsysnqn,
>  		return NULL;
>  
>  	subsys->ver = NVME_VS(1, 3, 0); /* NVMe 1.3.0 */
> +	get_random_bytes(&subsys->serial, sizeof(subsys->serial));

This removes the comment, which I think is fairly useful.

[PATCH v2 2/2] nvmet: preserve controller serial number between reboots

[Johannes Thumshirn]

The NVMe target has no way to preserve controller serial
IDs across reboots which breaks udev scripts doing
SYMLINK+="dev/disk/by-id/nvme-$env{ID_SERIAL}-part%n.

Export the randomly generated serial number via configfs and allow
setting of a serial via configfs to mitigate this breakage.

Signed-off-by: Johannes Thumshirn <jthumshirn@suse.de>
---
 drivers/nvme/target/configfs.c | 22 ++++++++++++++++++++++
 drivers/nvme/target/core.c     |  2 ++
 2 files changed, 24 insertions(+)

diff --git a/drivers/nvme/target/configfs.c b/drivers/nvme/target/configfs.c
index a358ecd93e11..9f19f0bde8f2 100644
--- a/drivers/nvme/target/configfs.c
+++ b/drivers/nvme/target/configfs.c
@@ -686,9 +686,31 @@ static ssize_t nvmet_subsys_version_store(struct config_item *item,
 }
 CONFIGFS_ATTR(nvmet_subsys_, version);
 
+static ssize_t nvmet_subsys_attr_serial_show(struct config_item *item,
+					     char *page)
+{
+	struct nvmet_subsys *subsys = to_subsys(item);
+
+	return snprintf(page, PAGE_SIZE, "%llx\n", subsys->serial);
+}
+
+static ssize_t nvmet_subsys_attr_serial_store(struct config_item *item,
+					      const char *page, size_t count)
+{
+	struct nvmet_subsys *subsys = to_subsys(item);
+
+	down_write(&nvmet_config_sem);
+	sscanf(page, "%llx\n", &subsys->serial);
+	up_write(&nvmet_config_sem);
+
+	return count;
+}
+CONFIGFS_ATTR(nvmet_subsys_, attr_serial);
+
 static struct configfs_attribute *nvmet_subsys_attrs[] = {
 	&nvmet_subsys_attr_attr_allow_any_host,
 	&nvmet_subsys_attr_version,
+	&nvmet_subsys_attr_attr_serial,
 	NULL,
 };
 
diff --git a/drivers/nvme/target/core.c b/drivers/nvme/target/core.c
index 02efa7113d23..8feeed8c709c 100644
--- a/drivers/nvme/target/core.c
+++ b/drivers/nvme/target/core.c
@@ -924,8 +924,10 @@ struct nvmet_subsys *nvmet_subsys_alloc(const char *subsysnqn,
 	if (!subsys)
 		return NULL;
 
+	down_write(&nvmet_config_sem);
 	subsys->ver = NVME_VS(1, 3, 0); /* NVMe 1.3.0 */
 	get_random_bytes(&subsys->serial, sizeof(subsys->serial));
+	up_write(&nvmet_config_sem);
 
 	switch (type) {
 	case NVME_NQN_NVME:
-- 
2.12.3

Re: [PATCH v2 2/2] nvmet: preserve controller serial number between reboots

[Christoph Hellwig]

This patch itself looks good:

Reviewed-by: Christoph Hellwig <hch@lst.de>

>  static struct configfs_attribute *nvmet_subsys_attrs[] = {
>  	&nvmet_subsys_attr_attr_allow_any_host,
>  	&nvmet_subsys_attr_version,
> +	&nvmet_subsys_attr_attr_serial,

but the new version attribute really needs the attr prefix so that
nvmetcli can pick it up.  And we should also add nvmetcli support
for it.  Can you do the rename before your series to ease backporting
as that should go into 3.12 (just like your serial number fix).

Re: [PATCH v2 2/2] nvmet: preserve controller serial number between reboots

[Johannes Thumshirn]

On Fri, Jul 14, 2017 at 02:32:08PM +0200, Christoph Hellwig wrote:
> This patch itself looks good:
> 
> Reviewed-by: Christoph Hellwig <hch@lst.de>
> 
> >  static struct configfs_attribute *nvmet_subsys_attrs[] = {
> >  	&nvmet_subsys_attr_attr_allow_any_host,
> >  	&nvmet_subsys_attr_version,
> > +	&nvmet_subsys_attr_attr_serial,
> 
> but the new version attribute really needs the attr prefix so that
> nvmetcli can pick it up.  And we should also add nvmetcli support
> for it.  Can you do the rename before your series to ease backporting
> as that should go into 3.12 (just like your serial number fix).
> 

Is it OK if I send the patches in the same series? 

-- 
Johannes Thumshirn                                          Storage
jthumshirn@suse.de                                +49 911 74053 689
SUSE LINUX GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: Felix Imendörffer, Jane Smithard, Graham Norton
HRB 21284 (AG Nürnberg)
Key fingerprint = EC38 9CAB C2C4 F25D 8600 D0D0 0393 969D 2D76 0850

Re: [PATCH v2 2/2] nvmet: preserve controller serial number between reboots

[Christoph Hellwig]

On Fri, Jul 14, 2017 at 02:46:32PM +0200, Johannes Thumshirn wrote:
> On Fri, Jul 14, 2017 at 02:32:08PM +0200, Christoph Hellwig wrote:
> > This patch itself looks good:
> > 
> > Reviewed-by: Christoph Hellwig <hch@lst.de>
> > 
> > >  static struct configfs_attribute *nvmet_subsys_attrs[] = {
> > >  	&nvmet_subsys_attr_attr_allow_any_host,
> > >  	&nvmet_subsys_attr_version,
> > > +	&nvmet_subsys_attr_attr_serial,
> > 
> > but the new version attribute really needs the attr prefix so that
> > nvmetcli can pick it up.  And we should also add nvmetcli support
> > for it.  Can you do the rename before your series to ease backporting
> > as that should go into 3.12 (just like your serial number fix).
> > 
> 
> Is it OK if I send the patches in the same series? 

Yes.  Just move the version fix to the beginning, please.