diff mbox series

[01/17] nvme: introduce namespace features flag

Message ID 20200327171545.98970-3-maxg@mellanox.com (mailing list archive)
State Not Applicable
Headers show
Series [01/17] nvme: introduce namespace features flag | expand

Commit Message

Max Gurtovoy March 27, 2020, 5:15 p.m. UTC 
From: Israel Rukshin <israelr@mellanox.com>

Centralize all the metadata checks to one place and make the code more
readable. Introduce a new enum nvme_ns_features for that matter.
The features flag description:
 - NVME_NS_EXT_LBAS - NVMe namespace supports extended LBA format.
 - NVME_NS_MD_HOST_SUPPORTED - NVMe namespace supports getting metadata
   from host's block layer.
 - NVME_NS_MD_CTRL_SUPPORTED - NVMe namespace supports metadata actions
   by the controller (generate/strip).

Suggested-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Max Gurtovoy <maxg@mellanox.com>
Signed-off-by: Israel Rukshin <israelr@mellanox.com>
---
 drivers/nvme/host/core.c | 42 ++++++++++++++++++++++++++++--------------
 drivers/nvme/host/nvme.h |  8 +++++++-
 2 files changed, 35 insertions(+), 15 deletions(-)

Comments

Christoph Hellwig April 21, 2020, 11:59 a.m. UTC | #1 
On Fri, Mar 27, 2020 at 08:15:29PM +0300, Max Gurtovoy wrote:
> From: Israel Rukshin <israelr@mellanox.com>
> 
> Centralize all the metadata checks to one place and make the code more
> readable. Introduce a new enum nvme_ns_features for that matter.
> The features flag description:
>  - NVME_NS_EXT_LBAS - NVMe namespace supports extended LBA format.
>  - NVME_NS_MD_HOST_SUPPORTED - NVMe namespace supports getting metadata
>    from host's block layer.
>  - NVME_NS_MD_CTRL_SUPPORTED - NVMe namespace supports metadata actions
>    by the controller (generate/strip).

So whole I like the ->features flag, the defintion of these two
metadata related features really confuses me.

Here are my vague ideas to improve the situation:

> -static inline bool nvme_ns_has_pi(struct nvme_ns *ns)
> -{
> -	return ns->pi_type && ns->ms == sizeof(struct t10_pi_tuple);
> -}

This function I think is generally useful, I'd rather keep iţ, document
it with a comment and remove the new NVME_NS_MD_CTRL_SUPPORTED
flag.

> -	if (ns->ms && !ns->ext &&
> -	    (ns->ctrl->ops->flags & NVME_F_METADATA_SUPPORTED))
> +	if (ns->features & NVME_NS_MD_HOST_SUPPORTED)
>  		nvme_init_integrity(disk, ns->ms, ns->pi_type);
> -	if ((ns->ms && !nvme_ns_has_pi(ns) && !blk_get_integrity(disk)) ||
> -	    ns->lba_shift > PAGE_SHIFT)
> +
> +	if ((ns->ms && !(ns->features & NVME_NS_MD_CTRL_SUPPORTED) &&
> +	     !(ns->features & NVME_NS_MD_HOST_SUPPORTED) &&
> +	     !blk_get_integrity(disk)) || ns->lba_shift > PAGE_SHIFT)
>  		capacity = 0;

I find this very confusing.  Can we do something like:

	/*
	 * The block layer can't support LBA sizes larger than the page size
	 * yet, so catch this early and don't allow block I/O.
	 */
	if (ns->lba_shift > PAGE_SHIFT)
  		capacity = 0;

	/*
	 * Register a metadata profile for PI, or the plain non-integrity NVMe
	 * metadata masquerading as Typ 0 if supported, otherwise reject block
	 * I/O to namespaces with metadata except when the namespace supports
	 * PI, as it can strip/insert in that case.
	 */
	if (ns->ms) {
		if (IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY) &&
		    (ns->features & NVME_NS_MD_HOST_SUPPORTED))
			nvme_init_integrity(disk, ns->ms, ns->pi_type);
		else if (!nvme_ns_has_pi(ns))
			capacity = 0;
	}

> +	if (ns->ms) {
> +		if (id->flbas & NVME_NS_FLBAS_META_EXT)
> +			ns->features |= NVME_NS_EXT_LBAS;
> +
> +		/*
> +		 * For PCI, Extended logical block will be generated by the
> +		 * controller.
> +		 */
> +		if (ns->ctrl->ops->flags & NVME_F_METADATA_SUPPORTED) {
> +			if (!(ns->features & NVME_NS_EXT_LBAS))
> +				ns->features |= NVME_NS_MD_HOST_SUPPORTED;
> +		}

Maybe:

> +	if (ns->ms) {
> +		if (id->flbas & NVME_NS_FLBAS_META_EXT)
> +			ns->features |= NVME_NS_EXT_LBAS;
> +
> +		/*
> +		 * For PCI, Extended logical block will be generated by the
> +		 * controller.
> +		 */
> +		if (ns->ctrl->ops->flags & NVME_F_METADATA_SUPPORTED) {
> +			if (!(ns->features & NVME_NS_EXT_LBAS))
> +				ns->features |= NVME_NS_MD_HOST_SUPPORTED;
> +		}

This looks a little strange now, but I guess it will make more sense
with the fabrics addition.  I'll take another look later in the series.

> +enum nvme_ns_features {
> +	NVME_NS_EXT_LBAS = 1 << 0,
> +	NVME_NS_MD_HOST_SUPPORTED = 1 << 1,
> +	NVME_NS_MD_CTRL_SUPPORTED = 1 << 2,
> +};

Please document the meaning of each flag.  I also suspect that just
moving ext to a flag first and than adding the NVME_NS_MD_HOST_SUPPORTED
bit might make more sense.  I'd also rename NVME_NS_MD_HOST_SUPPORTED
to NVME_NS_METADATA_SUPPORTED.
James Smart April 21, 2020, 3:53 p.m. UTC | #2 
On 4/21/2020 4:59 AM, Christoph Hellwig wrote:
> On Fri, Mar 27, 2020 at 08:15:29PM +0300, Max Gurtovoy wrote:
>> From: Israel Rukshin <israelr@mellanox.com>
>>
>> Centralize all the metadata checks to one place and make the code more
>> readable. Introduce a new enum nvme_ns_features for that matter.
>> The features flag description:
>>   - NVME_NS_EXT_LBAS - NVMe namespace supports extended LBA format.
>>   - NVME_NS_MD_HOST_SUPPORTED - NVMe namespace supports getting metadata
>>     from host's block layer.
>>   - NVME_NS_MD_CTRL_SUPPORTED - NVMe namespace supports metadata actions
>>     by the controller (generate/strip).
> So whole I like the ->features flag, the defintion of these two
> metadata related features really confuses me.
>
> Here are my vague ideas to improve the situation:
>

Care to look at any of the RFC items I posted on 2/24 - which does 
things a little differently ?   Perhaps find a common ground with Max's 
patches.
http://lists.infradead.org/pipermail/linux-nvme/2020-February/029066.html

Granted I've tweaked what I sent a little as there was no need to make 
nvme_ns_has_pi accessible to the transport.

-- james
Christoph Hellwig April 21, 2020, 6:11 p.m. UTC | #3 
On Tue, Apr 21, 2020 at 08:53:18AM -0700, James Smart wrote:
> Care to look at any of the RFC items I posted on 2/24 - which does things a 
> little differently ?   Perhaps find a common ground with Max's patches.
> http://lists.infradead.org/pipermail/linux-nvme/2020-February/029066.html
>
> Granted I've tweaked what I sent a little as there was no need to make 
> nvme_ns_has_pi accessible to the transport.

I've replied to the series.
diff mbox series

Patch

diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
index 1cab3c6..f3a184f 100644
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c
@@ -203,11 +203,6 @@  static void nvme_delete_ctrl_sync(struct nvme_ctrl *ctrl)
 	nvme_put_ctrl(ctrl);
 }
 
-static inline bool nvme_ns_has_pi(struct nvme_ns *ns)
-{
-	return ns->pi_type && ns->ms == sizeof(struct t10_pi_tuple);
-}
-
 static blk_status_t nvme_error_status(u16 status)
 {
 	switch (status & 0x7ff) {
@@ -706,7 +701,8 @@  static inline blk_status_t nvme_setup_rw(struct nvme_ns *ns,
 		 * namespace capacity to zero to prevent any I/O.
 		 */
 		if (!blk_integrity_rq(req)) {
-			if (WARN_ON_ONCE(!nvme_ns_has_pi(ns)))
+			if (WARN_ON_ONCE(!(ns->features &
+					   NVME_NS_MD_CTRL_SUPPORTED)))
 				return BLK_STS_NOTSUPP;
 			control |= NVME_RW_PRINFO_PRACT;
 		}
@@ -1277,7 +1273,7 @@  static int nvme_submit_io(struct nvme_ns *ns, struct nvme_user_io __user *uio)
 	meta_len = (io.nblocks + 1) * ns->ms;
 	metadata = (void __user *)(uintptr_t)io.metadata;
 
-	if (ns->ext) {
+	if (ns->features & NVME_NS_EXT_LBAS) {
 		length += meta_len;
 		meta_len = 0;
 	} else if (meta_len) {
@@ -1837,11 +1833,12 @@  static void nvme_update_disk_info(struct gendisk *disk,
 	blk_queue_io_min(disk->queue, phys_bs);
 	blk_queue_io_opt(disk->queue, io_opt);
 
-	if (ns->ms && !ns->ext &&
-	    (ns->ctrl->ops->flags & NVME_F_METADATA_SUPPORTED))
+	if (ns->features & NVME_NS_MD_HOST_SUPPORTED)
 		nvme_init_integrity(disk, ns->ms, ns->pi_type);
-	if ((ns->ms && !nvme_ns_has_pi(ns) && !blk_get_integrity(disk)) ||
-	    ns->lba_shift > PAGE_SHIFT)
+
+	if ((ns->ms && !(ns->features & NVME_NS_MD_CTRL_SUPPORTED) &&
+	     !(ns->features & NVME_NS_MD_HOST_SUPPORTED) &&
+	     !blk_get_integrity(disk)) || ns->lba_shift > PAGE_SHIFT)
 		capacity = 0;
 
 	set_capacity(disk, capacity);
@@ -1870,12 +1867,29 @@  static void __nvme_revalidate_disk(struct gendisk *disk, struct nvme_id_ns *id)
 		ns->lba_shift = 9;
 	ns->noiob = le16_to_cpu(id->noiob);
 	ns->ms = le16_to_cpu(id->lbaf[id->flbas & NVME_NS_FLBAS_LBA_MASK].ms);
-	ns->ext = ns->ms && (id->flbas & NVME_NS_FLBAS_META_EXT);
+	ns->features = 0;
 	/* the PI implementation requires metadata equal t10 pi tuple size */
-	if (ns->ms == sizeof(struct t10_pi_tuple))
+	if (ns->ms == sizeof(struct t10_pi_tuple)) {
 		ns->pi_type = id->dps & NVME_NS_DPS_PI_MASK;
-	else
+		if (ns->pi_type)
+			ns->features |= NVME_NS_MD_CTRL_SUPPORTED;
+	} else {
 		ns->pi_type = 0;
+	}
+
+	if (ns->ms) {
+		if (id->flbas & NVME_NS_FLBAS_META_EXT)
+			ns->features |= NVME_NS_EXT_LBAS;
+
+		/*
+		 * For PCI, Extended logical block will be generated by the
+		 * controller.
+		 */
+		if (ns->ctrl->ops->flags & NVME_F_METADATA_SUPPORTED) {
+			if (!(ns->features & NVME_NS_EXT_LBAS))
+				ns->features |= NVME_NS_MD_HOST_SUPPORTED;
+		}
+	}
 
 	if (ns->noiob)
 		nvme_set_chunk_size(ns);
diff --git a/drivers/nvme/host/nvme.h b/drivers/nvme/host/nvme.h
index 2e04a36..83296d0 100644
--- a/drivers/nvme/host/nvme.h
+++ b/drivers/nvme/host/nvme.h
@@ -363,6 +363,12 @@  struct nvme_ns_head {
 #endif
 };
 
+enum nvme_ns_features {
+	NVME_NS_EXT_LBAS = 1 << 0,
+	NVME_NS_MD_HOST_SUPPORTED = 1 << 1,
+	NVME_NS_MD_CTRL_SUPPORTED = 1 << 2,
+};
+
 struct nvme_ns {
 	struct list_head list;
 
@@ -382,8 +388,8 @@  struct nvme_ns {
 	u16 ms;
 	u16 sgs;
 	u32 sws;
-	bool ext;
 	u8 pi_type;
+	unsigned long features;
 	unsigned long flags;
 #define NVME_NS_REMOVING	0
 #define NVME_NS_DEAD     	1