UFS storage support for Tensor/GS101?

UFS storage support for Tensor/GS101?

[Eric Biggers]

Hi Peter!  Thanks for all the hard work you've been doing on upstreaming support
for Tensor/GS101.  I've tested that the latest upstream kernel is now able to
boot to a serial console on Pixel 6.

I'm wondering if you have any plans to upstream support for the UFS storage on
that platform?  It should be able to use the already-upstream
drivers/ufs/host/ufs-exynos.c driver, but changes may be required; I'm not sure
how many.  I'm interested in this mainly because I'd like to upstream some
patches related to inline encryption, and that depends on UFS storage working.

I'm interested in helping with this, but I wanted to check to see if you know
about any existing plans first.

Thanks!

- Eric

Re: UFS storage support for Tensor/GS101?

[Peter Griffin]

Hi Eric,

+ Will

On Tue, 6 Feb 2024 at 03:45, Eric Biggers <ebiggers@kernel.org> wrote:
>
> Hi Peter!  Thanks for all the hard work you've been doing on upstreaming support
> for Tensor/GS101.  I've tested that the latest upstream kernel is now able to
> boot to a serial console on Pixel 6.

Thanks for testing :)
>
> I'm wondering if you have any plans to upstream support for the UFS storage on
> that platform?  It should be able to use the already-upstream
> drivers/ufs/host/ufs-exynos.c driver, but changes may be required; I'm not sure
> how many.  I'm interested in this mainly because I'd like to upstream some
> patches related to inline encryption, and that depends on UFS storage working.
>
> I'm interested in helping with this, but I wanted to check to see if you know
> about any existing plans first.

We certainly want to add support for UFS upstream, and if you can help
out with enabling it that would be great! It would free us up to work
on enabling something else in parallel. From a first pass analysis of
the downstream UFS code it looks like ufs phy driver likely needs the
most attention.

The last UFS dependency (that I'm aware of) that is missing currently
is clock support for cmu_hsi2 (which I was planning on working on as
my next task). Also the UFS phy accesses PMU registers so that will
require my exynos-pmu [1] series but that is hopefully close to being
merged.

So I think we are quite close to having all the UFS dependencies in
place. Shall I ping you once I have some code for cmu_hsi2, and
hopefully that should be enough for you to start working on the UFS
driver and phy?

Thanks,

Peter

[1] https://lkml.org/lkml/2024/2/2/795

Re: UFS storage support for Tensor/GS101?

[Eric Biggers]

Hi Peter,

On Tue, Feb 06, 2024 at 05:09:13PM +0000, Peter Griffin wrote:
> Hi Eric,
> 
> + Will
> 
> On Tue, 6 Feb 2024 at 03:45, Eric Biggers <ebiggers@kernel.org> wrote:
> >
> > Hi Peter!  Thanks for all the hard work you've been doing on upstreaming support
> > for Tensor/GS101.  I've tested that the latest upstream kernel is now able to
> > boot to a serial console on Pixel 6.
> 
> Thanks for testing :)
> >
> > I'm wondering if you have any plans to upstream support for the UFS storage on
> > that platform?  It should be able to use the already-upstream
> > drivers/ufs/host/ufs-exynos.c driver, but changes may be required; I'm not sure
> > how many.  I'm interested in this mainly because I'd like to upstream some
> > patches related to inline encryption, and that depends on UFS storage working.
> >
> > I'm interested in helping with this, but I wanted to check to see if you know
> > about any existing plans first.
> 
> We certainly want to add support for UFS upstream, and if you can help
> out with enabling it that would be great! It would free us up to work
> on enabling something else in parallel. From a first pass analysis of
> the downstream UFS code it looks like ufs phy driver likely needs the
> most attention.
> 
> The last UFS dependency (that I'm aware of) that is missing currently
> is clock support for cmu_hsi2 (which I was planning on working on as
> my next task). Also the UFS phy accesses PMU registers so that will
> require my exynos-pmu [1] series but that is hopefully close to being
> merged.
> 
> So I think we are quite close to having all the UFS dependencies in
> place. Shall I ping you once I have some code for cmu_hsi2, and
> hopefully that should be enough for you to start working on the UFS
> driver and phy?
> 
> Thanks,
> 
> Peter

Yes, let me know as soon as you have some code for cmu_hsi2, and I'll try to get
UFS working.  I don't know how hard it will be, though, so it's possible I'll
need help from others.

Thanks,

- Eric

Re: UFS storage support for Tensor/GS101?

[William McVicker]

On 02/06/2024, Eric Biggers wrote:
> Hi Peter,
> 
> On Tue, Feb 06, 2024 at 05:09:13PM +0000, Peter Griffin wrote:
> > Hi Eric,
> > 
> > + Will
> > 
> > On Tue, 6 Feb 2024 at 03:45, Eric Biggers <ebiggers@kernel.org> wrote:
> > >
> > > Hi Peter!  Thanks for all the hard work you've been doing on upstreaming support
> > > for Tensor/GS101.  I've tested that the latest upstream kernel is now able to
> > > boot to a serial console on Pixel 6.
> > 
> > Thanks for testing :)
> > >
> > > I'm wondering if you have any plans to upstream support for the UFS storage on
> > > that platform?  It should be able to use the already-upstream
> > > drivers/ufs/host/ufs-exynos.c driver, but changes may be required; I'm not sure
> > > how many.  I'm interested in this mainly because I'd like to upstream some
> > > patches related to inline encryption, and that depends on UFS storage working.
> > >
> > > I'm interested in helping with this, but I wanted to check to see if you know
> > > about any existing plans first.
> > 
> > We certainly want to add support for UFS upstream, and if you can help
> > out with enabling it that would be great! It would free us up to work
> > on enabling something else in parallel. From a first pass analysis of
> > the downstream UFS code it looks like ufs phy driver likely needs the
> > most attention.
> > 
> > The last UFS dependency (that I'm aware of) that is missing currently
> > is clock support for cmu_hsi2 (which I was planning on working on as
> > my next task). Also the UFS phy accesses PMU registers so that will
> > require my exynos-pmu [1] series but that is hopefully close to being
> > merged.
> > 
> > So I think we are quite close to having all the UFS dependencies in
> > place. Shall I ping you once I have some code for cmu_hsi2, and
> > hopefully that should be enough for you to start working on the UFS
> > driver and phy?
> > 
> > Thanks,
> > 
> > Peter
> 
> Yes, let me know as soon as you have some code for cmu_hsi2, and I'll try to get
> UFS working.  I don't know how hard it will be, though, so it's possible I'll
> need help from others.
> 
> Thanks,
> 
> - Eric

+Bart

Hi Eric,

Thanks for reaching out! I was planning on sending a message to you and Bart
once the UFS dependencies that Peter mentioned were close to landing. We would
love to have your help!

Bart,
I know you mentiond in the past that you looked into upstreaming the Pixel UFS
driver but were blocked on it's dependencies. Have you had a chance to see what
has landed for gs101 upstream thus far? It would be great if you could provide
feedback on your findings! And please let us know if you're interested in
collaborating with us on upstreaming the gs101 UFS bits.

Thanks,
Will

Re: UFS storage support for Tensor/GS101?

[Bart Van Assche]

On 2/7/24 10:04, William McVicker wrote:
> I know you mentiond in the past that you looked into upstreaming the Pixel UFS
> driver but were blocked on it's dependencies. Have you had a chance to see what
> has landed for gs101 upstream thus far? It would be great if you could provide
> feedback on your findings! And please let us know if you're interested in
> collaborating with us on upstreaming the gs101 UFS bits.

The question about the dependencies should be answered by Samsung. They know
much more about the dependencies of the Exynos UFS driver than I do. I can
help with reviewing any patches they post.

Thanks,

Bart.

Re: UFS storage support for Tensor/GS101?

[Peter Griffin]

Hi Eric,

On Wed, 7 Feb 2024 at 01:52, Eric Biggers <ebiggers@kernel.org> wrote:
>
> Hi Peter,
>
> On Tue, Feb 06, 2024 at 05:09:13PM +0000, Peter Griffin wrote:
> > Hi Eric,
> >
> > + Will
> >
> > On Tue, 6 Feb 2024 at 03:45, Eric Biggers <ebiggers@kernel.org> wrote:
> > >
> > > Hi Peter!  Thanks for all the hard work you've been doing on upstreaming support
> > > for Tensor/GS101.  I've tested that the latest upstream kernel is now able to
> > > boot to a serial console on Pixel 6.
> >
> > Thanks for testing :)
> > >
> > > I'm wondering if you have any plans to upstream support for the UFS storage on
> > > that platform?  It should be able to use the already-upstream
> > > drivers/ufs/host/ufs-exynos.c driver, but changes may be required; I'm not sure
> > > how many.  I'm interested in this mainly because I'd like to upstream some
> > > patches related to inline encryption, and that depends on UFS storage working.
> > >
> > > I'm interested in helping with this, but I wanted to check to see if you know
> > > about any existing plans first.
> >
> > We certainly want to add support for UFS upstream, and if you can help
> > out with enabling it that would be great! It would free us up to work
> > on enabling something else in parallel. From a first pass analysis of
> > the downstream UFS code it looks like ufs phy driver likely needs the
> > most attention.
> >
> > The last UFS dependency (that I'm aware of) that is missing currently
> > is clock support for cmu_hsi2 (which I was planning on working on as
> > my next task). Also the UFS phy accesses PMU registers so that will
> > require my exynos-pmu [1] series but that is hopefully close to being
> > merged.
> >
> > So I think we are quite close to having all the UFS dependencies in
> > place. Shall I ping you once I have some code for cmu_hsi2, and
> > hopefully that should be enough for you to start working on the UFS
> > driver and phy?
> >
> > Thanks,
> >
> > Peter
>
> Yes, let me know as soon as you have some code for cmu_hsi2, and I'll try to get
> UFS working.  I don't know how hard it will be, though, so it's possible I'll
> need help from others.

I just posted an initial series that adds the cmu_hsi2 clocks, UFS
controller glue
and UFS phy for GS101 (your on cc). With that series applied, UFS link comes up
and you can mount flash partitions etc.

Hopefully that is enough to unblock you so you can start on the inline
encryption
feature. Let me know how you get on :)

Thanks,

Peter.