Write with Stream ID Support

Write with Stream ID Support

[kwan.huen]

The attached patch set enables basic write with stream ID support. 
First patch reads the stream id embedded in the bio and passes to the 
device along with the write command.
Second patch adds two new nvme commands to be used with ioctl 
such that application can do open/close stream and host
initiated garbage collection.

Write with Stream ID Support

[kwan.huen]

The attached patch set enables basic write with stream ID support. 
First patch reads the stream id embedded in the bio and passes to the 
device along with the write command.
Second patch adds two new nvme commands to be used with ioctl 
such that application can do open/close stream and host
initiated garbage collection.

Write with Stream ID Support

[kwan.huen]

The attached patch set enables basic write with stream ID support. 
First patch reads the stream id embedded in the bio and passes to the 
device along with the write command.
Second patch adds two new nvme commands to be used with ioctl 
such that application can do open/close stream and host
initiated garbage collection.

[PATCH 1/2] added stream id write support

[kwan.huen]

---
 drivers/block/nvme-core.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/block/nvme-core.c b/drivers/block/nvme-core.c
index 85b8036..332341a 100644
--- a/drivers/block/nvme-core.c
+++ b/drivers/block/nvme-core.c
@@ -769,6 +769,9 @@ static int nvme_submit_iod(struct nvme_queue *nvmeq, struct nvme_iod *iod,
 	if (req->cmd_flags & REQ_RAHEAD)
 		dsmgmt |= NVME_RW_DSM_FREQ_PREFETCH;
 
+	if (rq_data_dir(req))
+		dsmgmt |= bio_get_streamid(req->bio) << 8;
+
 	cmnd = &nvmeq->sq_cmds[nvmeq->sq_tail];
 	memset(cmnd, 0, sizeof(*cmnd));
 
-- 
1.9.1

[PATCH 1/2] added stream id write support

[kwan.huen]

---
 drivers/block/nvme-core.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/block/nvme-core.c b/drivers/block/nvme-core.c
index 85b8036..332341a 100644
--- a/drivers/block/nvme-core.c
+++ b/drivers/block/nvme-core.c
@@ -769,6 +769,9 @@ static int nvme_submit_iod(struct nvme_queue *nvmeq, struct nvme_iod *iod,
 	if (req->cmd_flags & REQ_RAHEAD)
 		dsmgmt |= NVME_RW_DSM_FREQ_PREFETCH;
 
+	if (rq_data_dir(req))
+		dsmgmt |= bio_get_streamid(req->bio) << 8;
+
 	cmnd = &nvmeq->sq_cmds[nvmeq->sq_tail];
 	memset(cmnd, 0, sizeof(*cmnd));
 
-- 
1.9.1

[PATCH 1/2] added stream id write support

[kwan.huen]

---
 drivers/block/nvme-core.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/block/nvme-core.c b/drivers/block/nvme-core.c
index 85b8036..332341a 100644
--- a/drivers/block/nvme-core.c
+++ b/drivers/block/nvme-core.c
@@ -769,6 +769,9 @@ static int nvme_submit_iod(struct nvme_queue *nvmeq, struct nvme_iod *iod,
 	if (req->cmd_flags & REQ_RAHEAD)
 		dsmgmt |= NVME_RW_DSM_FREQ_PREFETCH;
 
+	if (rq_data_dir(req))
+		dsmgmt |= bio_get_streamid(req->bio) << 8;
+
 	cmnd = &nvmeq->sq_cmds[nvmeq->sq_tail];
 	memset(cmnd, 0, sizeof(*cmnd));
 
-- 
1.9.1

[PATCH 2/2] added two nvme commands for open/close streams and garbage collection

[kwan.huen]

---
 include/uapi/linux/nvme.h | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/include/uapi/linux/nvme.h b/include/uapi/linux/nvme.h
index aef9a81..5025610 100644
--- a/include/uapi/linux/nvme.h
+++ b/include/uapi/linux/nvme.h
@@ -229,6 +229,8 @@ enum nvme_opcode {
 	nvme_cmd_resv_report	= 0x0e,
 	nvme_cmd_resv_acquire	= 0x11,
 	nvme_cmd_resv_release	= 0x15,
+	nvme_cmd_stream_ctrl	= 0x18,
+	nvme_cmd_gc_ctrl	= 0x1c,
 };
 
 struct nvme_common_command {
-- 
1.9.1

[PATCH 2/2] added two nvme commands for open/close streams and garbage collection

[kwan.huen]

---
 include/uapi/linux/nvme.h | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/include/uapi/linux/nvme.h b/include/uapi/linux/nvme.h
index aef9a81..5025610 100644
--- a/include/uapi/linux/nvme.h
+++ b/include/uapi/linux/nvme.h
@@ -229,6 +229,8 @@ enum nvme_opcode {
 	nvme_cmd_resv_report	= 0x0e,
 	nvme_cmd_resv_acquire	= 0x11,
 	nvme_cmd_resv_release	= 0x15,
+	nvme_cmd_stream_ctrl	= 0x18,
+	nvme_cmd_gc_ctrl	= 0x1c,
 };
 
 struct nvme_common_command {
-- 
1.9.1

[PATCH 2/2] added two nvme commands for open/close streams and garbage collection

[kwan.huen]

---
 include/uapi/linux/nvme.h | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/include/uapi/linux/nvme.h b/include/uapi/linux/nvme.h
index aef9a81..5025610 100644
--- a/include/uapi/linux/nvme.h
+++ b/include/uapi/linux/nvme.h
@@ -229,6 +229,8 @@ enum nvme_opcode {
 	nvme_cmd_resv_report	= 0x0e,
 	nvme_cmd_resv_acquire	= 0x11,
 	nvme_cmd_resv_release	= 0x15,
+	nvme_cmd_stream_ctrl	= 0x18,
+	nvme_cmd_gc_ctrl	= 0x1c,
 };
 
 struct nvme_common_command {
-- 
1.9.1

[PATCH 2/2] added two nvme commands for open/close streams and garbage collection

[Wilcox, Matthew R]

Woah, woah, woah!  Those command IDs have *NOT* been assigned by the NVMe working group!  NACK this patch!

-----Original Message-----
From: kwan.huen [mailto:kwan.huen@samsung.com] 
Sent: Monday, April 20, 2015 6:48 PM
To: Wilcox, Matthew R; Busch, Keith; Jens Axboe; kwan.huen at samsung.com; Dimitri John Ledkov
Cc: linux-nvme at lists.infradead.org; linux-kernel at vger.kernel.org; linux-api at vger.kernel.org
Subject: [PATCH 2/2] added two nvme commands for open/close streams and garbage collection

---
 include/uapi/linux/nvme.h | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/include/uapi/linux/nvme.h b/include/uapi/linux/nvme.h
index aef9a81..5025610 100644
--- a/include/uapi/linux/nvme.h
+++ b/include/uapi/linux/nvme.h
@@ -229,6 +229,8 @@ enum nvme_opcode {
 	nvme_cmd_resv_report	= 0x0e,
 	nvme_cmd_resv_acquire	= 0x11,
 	nvme_cmd_resv_release	= 0x15,
+	nvme_cmd_stream_ctrl	= 0x18,
+	nvme_cmd_gc_ctrl	= 0x1c,
 };
 
 struct nvme_common_command {
-- 
1.9.1

Re: [PATCH 1/2] added stream id write support

[Jeff Moyer]

"kwan.huen" <kwan.huen@samsung.com> writes:

> ---
>  drivers/block/nvme-core.c | 3 +++
>  1 file changed, 3 insertions(+)
>
> diff --git a/drivers/block/nvme-core.c b/drivers/block/nvme-core.c
> index 85b8036..332341a 100644
> --- a/drivers/block/nvme-core.c
> +++ b/drivers/block/nvme-core.c
> @@ -769,6 +769,9 @@ static int nvme_submit_iod(struct nvme_queue *nvmeq, struct nvme_iod *iod,
>  	if (req->cmd_flags & REQ_RAHEAD)
>  		dsmgmt |= NVME_RW_DSM_FREQ_PREFETCH;
>  
> +	if (rq_data_dir(req))
> +		dsmgmt |= bio_get_streamid(req->bio) << 8;
> +

There's no public specification for this, yet.  How many bits are set
aside for the stream id?  Do you need to do bounds checking/input
validation?  What happens on adapters with older firmware when these
bits are set?

Cheers,
Jeff

Re: [PATCH 1/2] added stream id write support

[Jeff Moyer]

"kwan.huen" <kwan.huen-Sze3O3UU22JBDgjK7y7TUQ@public.gmane.org> writes:

> ---
>  drivers/block/nvme-core.c | 3 +++
>  1 file changed, 3 insertions(+)
>
> diff --git a/drivers/block/nvme-core.c b/drivers/block/nvme-core.c
> index 85b8036..332341a 100644
> --- a/drivers/block/nvme-core.c
> +++ b/drivers/block/nvme-core.c
> @@ -769,6 +769,9 @@ static int nvme_submit_iod(struct nvme_queue *nvmeq, struct nvme_iod *iod,
>  	if (req->cmd_flags & REQ_RAHEAD)
>  		dsmgmt |= NVME_RW_DSM_FREQ_PREFETCH;
>  
> +	if (rq_data_dir(req))
> +		dsmgmt |= bio_get_streamid(req->bio) << 8;
> +

There's no public specification for this, yet.  How many bits are set
aside for the stream id?  Do you need to do bounds checking/input
validation?  What happens on adapters with older firmware when these
bits are set?

Cheers,
Jeff

[PATCH 1/2] added stream id write support

[Jeff Moyer]

"kwan.huen" <kwan.huen at samsung.com> writes:

> ---
>  drivers/block/nvme-core.c | 3 +++
>  1 file changed, 3 insertions(+)
>
> diff --git a/drivers/block/nvme-core.c b/drivers/block/nvme-core.c
> index 85b8036..332341a 100644
> --- a/drivers/block/nvme-core.c
> +++ b/drivers/block/nvme-core.c
> @@ -769,6 +769,9 @@ static int nvme_submit_iod(struct nvme_queue *nvmeq, struct nvme_iod *iod,
>  	if (req->cmd_flags & REQ_RAHEAD)
>  		dsmgmt |= NVME_RW_DSM_FREQ_PREFETCH;
>  
> +	if (rq_data_dir(req))
> +		dsmgmt |= bio_get_streamid(req->bio) << 8;
> +

There's no public specification for this, yet.  How many bits are set
aside for the stream id?  Do you need to do bounds checking/input
validation?  What happens on adapters with older firmware when these
bits are set?

Cheers,
Jeff

[PATCH 2/2] added two nvme commands for open/close streams and garbage collection

[Kwan (Hingkwan) Huen-SSI]

Thanks Matthew for the review!
Agree that this is still a bit early. We are in the process of getting the spec approved and will push these patches again after it goes on-line.
Thanks,
kwan 
________________________________________
From: Wilcox, Matthew R [matthew.r.wilcox@intel.com]
Sent: Tuesday, April 21, 2015 6:21 AM
To: kwan.huen; Busch, Keith; Jens Axboe; Dimitri John Ledkov
Cc: linux-nvme at lists.infradead.org; linux-kernel at vger.kernel.org; linux-api at vger.kernel.org
Subject: RE: [PATCH 2/2] added two nvme commands for open/close streams and garbage collection

Woah, woah, woah!  Those command IDs have *NOT* been assigned by the NVMe working group!  NACK this patch!

-----Original Message-----
From: kwan.huen [mailto:kwan.huen@samsung.com]
Sent: Monday, April 20, 2015 6:48 PM
To: Wilcox, Matthew R; Busch, Keith; Jens Axboe; kwan.huen at samsung.com; Dimitri John Ledkov
Cc: linux-nvme at lists.infradead.org; linux-kernel at vger.kernel.org; linux-api at vger.kernel.org
Subject: [PATCH 2/2] added two nvme commands for open/close streams and garbage collection

---
 include/uapi/linux/nvme.h | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/include/uapi/linux/nvme.h b/include/uapi/linux/nvme.h
index aef9a81..5025610 100644
--- a/include/uapi/linux/nvme.h
+++ b/include/uapi/linux/nvme.h
@@ -229,6 +229,8 @@ enum nvme_opcode {
        nvme_cmd_resv_report    = 0x0e,
        nvme_cmd_resv_acquire   = 0x11,
        nvme_cmd_resv_release   = 0x15,
+       nvme_cmd_stream_ctrl    = 0x18,
+       nvme_cmd_gc_ctrl        = 0x1c,
 };

 struct nvme_common_command {
--
1.9.1

RE: [PATCH 1/2] added stream id write support

[Hingkwan Huen]

> -----Original Message-----
> From: Jeff Moyer [mailto:jmoyer@redhat.com]
> Sent: Tuesday, April 21, 2015 6:45 AM
> To: kwan.huen
> Cc: Matthew Wilcox; Keith Busch; Jens Axboe; Dimitri John Ledkov; linux-
> nvme@lists.infradead.org; linux-kernel@vger.kernel.org; linux-
> api@vger.kernel.org
> Subject: Re: [PATCH 1/2] added stream id write support
> 
> "kwan.huen" <kwan.huen@samsung.com> writes:
> 
> > ---
> >  drivers/block/nvme-core.c | 3 +++
> >  1 file changed, 3 insertions(+)
> >
> > diff --git a/drivers/block/nvme-core.c b/drivers/block/nvme-core.c
> > index 85b8036..332341a 100644
> > --- a/drivers/block/nvme-core.c
> > +++ b/drivers/block/nvme-core.c
> > @@ -769,6 +769,9 @@ static int nvme_submit_iod(struct nvme_queue
> *nvmeq, struct nvme_iod *iod,
> >  	if (req->cmd_flags & REQ_RAHEAD)
> >  		dsmgmt |= NVME_RW_DSM_FREQ_PREFETCH;
> >
> > +	if (rq_data_dir(req))
> > +		dsmgmt |= bio_get_streamid(req->bio) << 8;
> > +
> 
> There's no public specification for this, yet.  How many bits are set
aside for
> the stream id?  Do you need to do bounds checking/input validation?  What
> happens on adapters with older firmware when these bits are set?
> 
> Cheers,
> Jeff

Thanks Jeff for the review! 
The patch depends on Jens' recent io-streamid patch set, and the new NVMe
spec we are trying to get approved.
The patch is probably still too early and we'll push this again when Jens'
patches are released and the new spec goes public.
Thanks,
kwan

RE: [PATCH 1/2] added stream id write support

[Hingkwan Huen]

> -----Original Message-----
> From: Jeff Moyer [mailto:jmoyer-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org]
> Sent: Tuesday, April 21, 2015 6:45 AM
> To: kwan.huen
> Cc: Matthew Wilcox; Keith Busch; Jens Axboe; Dimitri John Ledkov; linux-
> nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r@public.gmane.org; linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org; linux-
> api-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> Subject: Re: [PATCH 1/2] added stream id write support
> 
> "kwan.huen" <kwan.huen-Sze3O3UU22JBDgjK7y7TUQ@public.gmane.org> writes:
> 
> > ---
> >  drivers/block/nvme-core.c | 3 +++
> >  1 file changed, 3 insertions(+)
> >
> > diff --git a/drivers/block/nvme-core.c b/drivers/block/nvme-core.c
> > index 85b8036..332341a 100644
> > --- a/drivers/block/nvme-core.c
> > +++ b/drivers/block/nvme-core.c
> > @@ -769,6 +769,9 @@ static int nvme_submit_iod(struct nvme_queue
> *nvmeq, struct nvme_iod *iod,
> >  	if (req->cmd_flags & REQ_RAHEAD)
> >  		dsmgmt |= NVME_RW_DSM_FREQ_PREFETCH;
> >
> > +	if (rq_data_dir(req))
> > +		dsmgmt |= bio_get_streamid(req->bio) << 8;
> > +
> 
> There's no public specification for this, yet.  How many bits are set
aside for
> the stream id?  Do you need to do bounds checking/input validation?  What
> happens on adapters with older firmware when these bits are set?
> 
> Cheers,
> Jeff

Thanks Jeff for the review! 
The patch depends on Jens' recent io-streamid patch set, and the new NVMe
spec we are trying to get approved.
The patch is probably still too early and we'll push this again when Jens'
patches are released and the new spec goes public.
Thanks,
kwan

[PATCH 1/2] added stream id write support

[Hingkwan Huen]

> -----Original Message-----
> From: Jeff Moyer [mailto:jmoyer at redhat.com]
> Sent: Tuesday, April 21, 2015 6:45 AM
> To: kwan.huen
> Cc: Matthew Wilcox; Keith Busch; Jens Axboe; Dimitri John Ledkov; linux-
> nvme at lists.infradead.org; linux-kernel at vger.kernel.org; linux-
> api at vger.kernel.org
> Subject: Re: [PATCH 1/2] added stream id write support
> 
> "kwan.huen" <kwan.huen at samsung.com> writes:
> 
> > ---
> >  drivers/block/nvme-core.c | 3 +++
> >  1 file changed, 3 insertions(+)
> >
> > diff --git a/drivers/block/nvme-core.c b/drivers/block/nvme-core.c
> > index 85b8036..332341a 100644
> > --- a/drivers/block/nvme-core.c
> > +++ b/drivers/block/nvme-core.c
> > @@ -769,6 +769,9 @@ static int nvme_submit_iod(struct nvme_queue
> *nvmeq, struct nvme_iod *iod,
> >  	if (req->cmd_flags & REQ_RAHEAD)
> >  		dsmgmt |= NVME_RW_DSM_FREQ_PREFETCH;
> >
> > +	if (rq_data_dir(req))
> > +		dsmgmt |= bio_get_streamid(req->bio) << 8;
> > +
> 
> There's no public specification for this, yet.  How many bits are set
aside for
> the stream id?  Do you need to do bounds checking/input validation?  What
> happens on adapters with older firmware when these bits are set?
> 
> Cheers,
> Jeff

Thanks Jeff for the review! 
The patch depends on Jens' recent io-streamid patch set, and the new NVMe
spec we are trying to get approved.
The patch is probably still too early and we'll push this again when Jens'
patches are released and the new spec goes public.
Thanks,
kwan