drm/dp_mst: correct the shifting in DP_REMOTE_I2C_READ
diff mbox series

Message ID 20191230070516.4760-1-Wayne.Lin@amd.com
State New
Headers show
Series
  • drm/dp_mst: correct the shifting in DP_REMOTE_I2C_READ
Related show

Commit Message

Lin, Wayne Dec. 30, 2019, 7:05 a.m. UTC
[Why]
According to DP spec, it should shift left 4 digits for NO_STOP_BIT
in REMOTE_I2C_READ message. Not 5 digits.

[How]
Correct the shifting value of NO_STOP_BIT for DP_REMOTE_I2C_READ case in
drm_dp_encode_sideband_req().

Signed-off-by: Wayne Lin <Wayne.Lin@amd.com>
---
 drivers/gpu/drm/drm_dp_mst_topology.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Comments

Jani Nikula Dec. 30, 2019, 11:15 a.m. UTC | #1
On Mon, 30 Dec 2019, Wayne Lin <Wayne.Lin@amd.com> wrote:
> [Why]
> According to DP spec, it should shift left 4 digits for NO_STOP_BIT
> in REMOTE_I2C_READ message. Not 5 digits.
>
> [How]
> Correct the shifting value of NO_STOP_BIT for DP_REMOTE_I2C_READ case in
> drm_dp_encode_sideband_req().

Which commit introduced the issue? Fixes: tag. Does it need a stable
backport? Does this fix a user visible bug?

BR,
Jani.

> Signed-off-by: Wayne Lin <Wayne.Lin@amd.com>
> ---
>  drivers/gpu/drm/drm_dp_mst_topology.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/drm_dp_mst_topology.c b/drivers/gpu/drm/drm_dp_mst_topology.c
> index 1d1bfa49ca2b..0557e225ff67 100644
> --- a/drivers/gpu/drm/drm_dp_mst_topology.c
> +++ b/drivers/gpu/drm/drm_dp_mst_topology.c
> @@ -393,7 +393,7 @@ drm_dp_encode_sideband_req(const struct drm_dp_sideband_msg_req_body *req,
>  			memcpy(&buf[idx], req->u.i2c_read.transactions[i].bytes, req->u.i2c_read.transactions[i].num_bytes);
>  			idx += req->u.i2c_read.transactions[i].num_bytes;
>  
> -			buf[idx] = (req->u.i2c_read.transactions[i].no_stop_bit & 0x1) << 5;
> +			buf[idx] = (req->u.i2c_read.transactions[i].no_stop_bit & 0x1) << 4;
>  			buf[idx] |= (req->u.i2c_read.transactions[i].i2c_transaction_delay & 0xf);
>  			idx++;
>  		}
Harry Wentland Dec. 30, 2019, 3:26 p.m. UTC | #2
On 2019-12-30 2:05 a.m., Wayne Lin wrote:
> [Why]
> According to DP spec, it should shift left 4 digits for NO_STOP_BIT
> in REMOTE_I2C_READ message. Not 5 digits.
> 
> [How]
> Correct the shifting value of NO_STOP_BIT for DP_REMOTE_I2C_READ case in
> drm_dp_encode_sideband_req().
> 
> Signed-off-by: Wayne Lin <Wayne.Lin@amd.com>

Good catch. Looks like this has been there since the beginning of MST in
the kernel. How did you find this? Did this cause bad EDID reads or some
other problem? If so the commit message should probably mention it.

Harry

> ---
>  drivers/gpu/drm/drm_dp_mst_topology.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/drm_dp_mst_topology.c b/drivers/gpu/drm/drm_dp_mst_topology.c
> index 1d1bfa49ca2b..0557e225ff67 100644
> --- a/drivers/gpu/drm/drm_dp_mst_topology.c
> +++ b/drivers/gpu/drm/drm_dp_mst_topology.c
> @@ -393,7 +393,7 @@ drm_dp_encode_sideband_req(const struct drm_dp_sideband_msg_req_body *req,
>  			memcpy(&buf[idx], req->u.i2c_read.transactions[i].bytes, req->u.i2c_read.transactions[i].num_bytes);
>  			idx += req->u.i2c_read.transactions[i].num_bytes;
>  
> -			buf[idx] = (req->u.i2c_read.transactions[i].no_stop_bit & 0x1) << 5;
> +			buf[idx] = (req->u.i2c_read.transactions[i].no_stop_bit & 0x1) << 4;
>  			buf[idx] |= (req->u.i2c_read.transactions[i].i2c_transaction_delay & 0xf);
>  			idx++;
>  		}
>
Lin, Wayne Dec. 31, 2019, 3:30 a.m. UTC | #3
[AMD Official Use Only - Internal Distribution Only]

> ________________________________________
> From: Jani Nikula <jani.nikula@linux.intel.com>
> Sent: Monday, December 30, 2019 19:15
> To: Lin, Wayne; dri-devel@lists.freedesktop.org; amd-gfx@lists.freedesktop.org
> Cc: Zuo, Jerry; Kazlauskas, Nicholas; Lin, Wayne
> Subject: Re: [PATCH] drm/dp_mst: correct the shifting in DP_REMOTE_I2C_READ
>
> On Mon, 30 Dec 2019, Wayne Lin <Wayne.Lin@amd.com> wrote:
> > [Why]
> > According to DP spec, it should shift left 4 digits for NO_STOP_BIT
> > in REMOTE_I2C_READ message. Not 5 digits.
> >
> > [How]
> > Correct the shifting value of NO_STOP_BIT for DP_REMOTE_I2C_READ case in
> > drm_dp_encode_sideband_req().
>
> Which commit introduced the issue? Fixes: tag. Does it need a stable
> backport? Does this fix a user visible bug?
>
> BR,
> Jani.
>
Thanks for your time and reminder.

It seems like the issue has been there since very beginning.(commit: ad7f8a1).
It doesn't introduce user visible bug under my test cases, but this affects the I2C signal
between I2C master and I2C slave. Not pretty sure if there is any eeprom will reset
the written offset once received I2C stop. If so, that might cause wrongly reading EDID.
I will Cc to stable. Thanks.
> > Signed-off-by: Wayne Lin <Wayne.Lin@amd.com>
> > ---
> >  drivers/gpu/drm/drm_dp_mst_topology.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/gpu/drm/drm_dp_mst_topology.c b/drivers/gpu/drm/drm_dp_mst_topology.c
> > index 1d1bfa49ca2b..0557e225ff67 100644
> > --- a/drivers/gpu/drm/drm_dp_mst_topology.c
> > +++ b/drivers/gpu/drm/drm_dp_mst_topology.c
> > @@ -393,7 +393,7 @@ drm_dp_encode_sideband_req(const struct drm_dp_sideband_msg_req_body *req,
> >                       memcpy(&buf[idx], req->u.i2c_read.transactions[i].bytes, req->u.i2c_read.transactions[i].num_bytes);
> >                       idx += req->u.i2c_read.transactions[i].num_bytes;
> >
> > -                     buf[idx] = (req->u.i2c_read.transactions[i].no_stop_bit & 0x1) << 5;
> > +                     buf[idx] = (req->u.i2c_read.transactions[i].no_stop_bit & 0x1) << 4;
> >                       buf[idx] |= (req->u.i2c_read.transactions[i].i2c_transaction_delay & 0xf);
> >                       idx++;
> >               }
>
> --
> Jani Nikula, Intel Open Source Graphics Center
--
Wayne Lin
Lin, Wayne Dec. 31, 2019, 4 a.m. UTC | #4
[AMD Official Use Only - Internal Distribution Only]

> ________________________________________
> From: Wentland, Harry <Harry.Wentland@amd.com>
> Sent: Monday, December 30, 2019 23:26
> To: Lin, Wayne; dri-devel@lists.freedesktop.org; amd-gfx@lists.freedesktop.org
> Cc: lyude@redhat.com; Zuo, Jerry; Kazlauskas, Nicholas; Wentland, Harry
> Subject: Re: [PATCH] drm/dp_mst: correct the shifting in DP_REMOTE_I2C_READ
>
> On 2019-12-30 2:05 a.m., Wayne Lin wrote:
> > [Why]
> > According to DP spec, it should shift left 4 digits for NO_STOP_BIT
> > in REMOTE_I2C_READ message. Not 5 digits.
> >
> > [How]
> > Correct the shifting value of NO_STOP_BIT for DP_REMOTE_I2C_READ case in
> > drm_dp_encode_sideband_req().
> >
> > Signed-off-by: Wayne Lin <Wayne.Lin@amd.com>
>
> Good catch. Looks like this has been there since the beginning of MST in
> the kernel. How did you find this? Did this cause bad EDID reads or some
> other problem? If so the commit message should probably mention it.
>
> Harry
Thanks for your time.

I found this while I was trying to debug MST issues by using AUX monitor.
So far, I don't observe problems that relate to this under my test cases and environment.
However, this bit might affect the I2C signal generated by I2C master, I will mention more
in the commit message.

Thanks.
>
> > ---
> >  drivers/gpu/drm/drm_dp_mst_topology.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/gpu/drm/drm_dp_mst_topology.c b/drivers/gpu/drm/drm_dp_mst_topology.c
> > index 1d1bfa49ca2b..0557e225ff67 100644
> > --- a/drivers/gpu/drm/drm_dp_mst_topology.c
> > +++ b/drivers/gpu/drm/drm_dp_mst_topology.c
> > @@ -393,7 +393,7 @@ drm_dp_encode_sideband_req(const struct drm_dp_sideband_msg_req_body *req,
> >                       memcpy(&buf[idx], req->u.i2c_read.transactions[i].bytes, req->u.i2c_read.transactions[i].num_bytes);
> >                       idx += req->u.i2c_read.transactions[i].num_bytes;
> >
> > -                     buf[idx] = (req->u.i2c_read.transactions[i].no_stop_bit & 0x1) << 5;
> > +                     buf[idx] = (req->u.i2c_read.transactions[i].no_stop_bit & 0x1) << 4;
> >                       buf[idx] |= (req->u.i2c_read.transactions[i].i2c_transaction_delay & 0xf);
> >                       idx++;
> >               }
> >
--
Wayne Lin
Harry Wentland Dec. 31, 2019, 2:40 p.m. UTC | #5
On 2019-12-30 11:00 p.m., Lin, Wayne wrote:
> [AMD Official Use Only - Internal Distribution Only]
> 

Make sure to select the correct AIP designation for public emails. See
my email on that from yesterday. :)

>> ________________________________________
>> From: Wentland, Harry <Harry.Wentland@amd.com>
>> Sent: Monday, December 30, 2019 23:26
>> To: Lin, Wayne; dri-devel@lists.freedesktop.org; amd-gfx@lists.freedesktop.org
>> Cc: lyude@redhat.com; Zuo, Jerry; Kazlauskas, Nicholas; Wentland, Harry
>> Subject: Re: [PATCH] drm/dp_mst: correct the shifting in DP_REMOTE_I2C_READ
>>
>> On 2019-12-30 2:05 a.m., Wayne Lin wrote:
>>> [Why]
>>> According to DP spec, it should shift left 4 digits for NO_STOP_BIT
>>> in REMOTE_I2C_READ message. Not 5 digits.
>>>
>>> [How]
>>> Correct the shifting value of NO_STOP_BIT for DP_REMOTE_I2C_READ case in
>>> drm_dp_encode_sideband_req().
>>>
>>> Signed-off-by: Wayne Lin <Wayne.Lin@amd.com>
>>
>> Good catch. Looks like this has been there since the beginning of MST in
>> the kernel. How did you find this? Did this cause bad EDID reads or some
>> other problem? If so the commit message should probably mention it.
>>
>> Harry
> Thanks for your time.
> 
> I found this while I was trying to debug MST issues by using AUX monitor.
> So far, I don't observe problems that relate to this under my test cases and environment.
> However, this bit might affect the I2C signal generated by I2C master, I will mention more
> in the commit message.
> 

Thanks. Was just curious if you saw any real issues. Good find
nevertheless and good patch.

You can add my
Reviewed-by: Harry Wentland <harry.wentland@amd.com>

Harry

> Thanks.
>>
>>> ---
>>>  drivers/gpu/drm/drm_dp_mst_topology.c | 2 +-
>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/gpu/drm/drm_dp_mst_topology.c b/drivers/gpu/drm/drm_dp_mst_topology.c
>>> index 1d1bfa49ca2b..0557e225ff67 100644
>>> --- a/drivers/gpu/drm/drm_dp_mst_topology.c
>>> +++ b/drivers/gpu/drm/drm_dp_mst_topology.c
>>> @@ -393,7 +393,7 @@ drm_dp_encode_sideband_req(const struct drm_dp_sideband_msg_req_body *req,
>>>                       memcpy(&buf[idx], req->u.i2c_read.transactions[i].bytes, req->u.i2c_read.transactions[i].num_bytes);
>>>                       idx += req->u.i2c_read.transactions[i].num_bytes;
>>>
>>> -                     buf[idx] = (req->u.i2c_read.transactions[i].no_stop_bit & 0x1) << 5;
>>> +                     buf[idx] = (req->u.i2c_read.transactions[i].no_stop_bit & 0x1) << 4;
>>>                       buf[idx] |= (req->u.i2c_read.transactions[i].i2c_transaction_delay & 0xf);
>>>                       idx++;
>>>               }
>>>
> --
> Wayne Lin
>
Ville Syrjala Jan. 7, 2020, 6:57 p.m. UTC | #6
On Tue, Dec 31, 2019 at 03:30:47AM +0000, Lin, Wayne wrote:
> [AMD Official Use Only - Internal Distribution Only]
> 
> > ________________________________________
> > From: Jani Nikula <jani.nikula@linux.intel.com>
> > Sent: Monday, December 30, 2019 19:15
> > To: Lin, Wayne; dri-devel@lists.freedesktop.org; amd-gfx@lists.freedesktop.org
> > Cc: Zuo, Jerry; Kazlauskas, Nicholas; Lin, Wayne
> > Subject: Re: [PATCH] drm/dp_mst: correct the shifting in DP_REMOTE_I2C_READ
> >
> > On Mon, 30 Dec 2019, Wayne Lin <Wayne.Lin@amd.com> wrote:
> > > [Why]
> > > According to DP spec, it should shift left 4 digits for NO_STOP_BIT
> > > in REMOTE_I2C_READ message. Not 5 digits.
> > >
> > > [How]
> > > Correct the shifting value of NO_STOP_BIT for DP_REMOTE_I2C_READ case in
> > > drm_dp_encode_sideband_req().
> >
> > Which commit introduced the issue? Fixes: tag. Does it need a stable
> > backport? Does this fix a user visible bug?
> >
> > BR,
> > Jani.
> >
> Thanks for your time and reminder.
> 
> It seems like the issue has been there since very beginning.(commit: ad7f8a1).
> It doesn't introduce user visible bug under my test cases, but this affects the I2C signal
> between I2C master and I2C slave. Not pretty sure if there is any eeprom will reset
> the written offset once received I2C stop. If so, that might cause wrongly reading EDID.
> I will Cc to stable. Thanks.

The segment address should be reset on STOP. So large EDIDs should
fail. IIRC we had a bug report of some sort about this which I tried
to fix by confgiuring .no_stop_bit correctly, but apparently I failed
to double check that the bit get stuffed onto the wire correctly.

Ah yes, https://bugs.freedesktop.org/show_bug.cgi?id=108081
So you may have just fixed that one, although we seem to have closed
it already.

> > > Signed-off-by: Wayne Lin <Wayne.Lin@amd.com>
> > > ---
> > >  drivers/gpu/drm/drm_dp_mst_topology.c | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/gpu/drm/drm_dp_mst_topology.c b/drivers/gpu/drm/drm_dp_mst_topology.c
> > > index 1d1bfa49ca2b..0557e225ff67 100644
> > > --- a/drivers/gpu/drm/drm_dp_mst_topology.c
> > > +++ b/drivers/gpu/drm/drm_dp_mst_topology.c
> > > @@ -393,7 +393,7 @@ drm_dp_encode_sideband_req(const struct drm_dp_sideband_msg_req_body *req,
> > >                       memcpy(&buf[idx], req->u.i2c_read.transactions[i].bytes, req->u.i2c_read.transactions[i].num_bytes);
> > >                       idx += req->u.i2c_read.transactions[i].num_bytes;
> > >
> > > -                     buf[idx] = (req->u.i2c_read.transactions[i].no_stop_bit & 0x1) << 5;
> > > +                     buf[idx] = (req->u.i2c_read.transactions[i].no_stop_bit & 0x1) << 4;
> > >                       buf[idx] |= (req->u.i2c_read.transactions[i].i2c_transaction_delay & 0xf);
> > >                       idx++;
> > >               }
> >
> > --
> > Jani Nikula, Intel Open Source Graphics Center
> --
> Wayne Lin
> _______________________________________________
> dri-devel mailing list
> dri-devel@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/dri-devel
Lin, Wayne Jan. 8, 2020, 2:34 a.m. UTC | #7
[AMD Public Use]



> -----Original Message-----
> From: Ville Syrjälä <ville.syrjala@linux.intel.com>
> Sent: Wednesday, January 8, 2020 2:57 AM
> To: Lin, Wayne <Wayne.Lin@amd.com>
> Cc: Jani Nikula <jani.nikula@linux.intel.com>; dri-
> devel@lists.freedesktop.org; amd-gfx@lists.freedesktop.org; Zuo, Jerry
> <Jerry.Zuo@amd.com>; Kazlauskas, Nicholas
> <Nicholas.Kazlauskas@amd.com>
> Subject: Re: [PATCH] drm/dp_mst: correct the shifting in
> DP_REMOTE_I2C_READ
> 
> On Tue, Dec 31, 2019 at 03:30:47AM +0000, Lin, Wayne wrote:
> > [AMD Official Use Only - Internal Distribution Only]
> >
> > > ________________________________________
> > > From: Jani Nikula <jani.nikula@linux.intel.com>
> > > Sent: Monday, December 30, 2019 19:15
> > > To: Lin, Wayne; dri-devel@lists.freedesktop.org;
> > > amd-gfx@lists.freedesktop.org
> > > Cc: Zuo, Jerry; Kazlauskas, Nicholas; Lin, Wayne
> > > Subject: Re: [PATCH] drm/dp_mst: correct the shifting in
> > > DP_REMOTE_I2C_READ
> > >
> > > On Mon, 30 Dec 2019, Wayne Lin <Wayne.Lin@amd.com> wrote:
> > > > [Why]
> > > > According to DP spec, it should shift left 4 digits for
> > > > NO_STOP_BIT in REMOTE_I2C_READ message. Not 5 digits.
> > > >
> > > > [How]
> > > > Correct the shifting value of NO_STOP_BIT for DP_REMOTE_I2C_READ
> > > > case in drm_dp_encode_sideband_req().
> > >
> > > Which commit introduced the issue? Fixes: tag. Does it need a stable
> > > backport? Does this fix a user visible bug?
> > >
> > > BR,
> > > Jani.
> > >
> > Thanks for your time and reminder.
> >
> > It seems like the issue has been there since very beginning.(commit:
> ad7f8a1).
> > It doesn't introduce user visible bug under my test cases, but this
> > affects the I2C signal between I2C master and I2C slave. Not pretty
> > sure if there is any eeprom will reset the written offset once received I2C
> stop. If so, that might cause wrongly reading EDID.
> > I will Cc to stable. Thanks.
> 
> The segment address should be reset on STOP. So large EDIDs should fail.
> IIRC we had a bug report of some sort about this which I tried to fix by
> confgiuring .no_stop_bit correctly, but apparently I failed to double check
> that the bit get stuffed onto the wire correctly.
> 
> Ah yes,
> https://bugs.freedesktop.org/show_bug.cgi?id=108081
> So you may have just fixed that one, although we seem to have closed it
> already.

Thanks for your time and the explanation.
> 
> > > > Signed-off-by: Wayne Lin <Wayne.Lin@amd.com>
> > > > ---
> > > >  drivers/gpu/drm/drm_dp_mst_topology.c | 2 +-
> > > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > >
> > > > diff --git a/drivers/gpu/drm/drm_dp_mst_topology.c
> > > > b/drivers/gpu/drm/drm_dp_mst_topology.c
> > > > index 1d1bfa49ca2b..0557e225ff67 100644
> > > > --- a/drivers/gpu/drm/drm_dp_mst_topology.c
> > > > +++ b/drivers/gpu/drm/drm_dp_mst_topology.c
> > > > @@ -393,7 +393,7 @@ drm_dp_encode_sideband_req(const struct
> drm_dp_sideband_msg_req_body *req,
> > > >                       memcpy(&buf[idx], req->u.i2c_read.transactions[i].bytes,
> req->u.i2c_read.transactions[i].num_bytes);
> > > >                       idx +=
> > > > req->u.i2c_read.transactions[i].num_bytes;
> > > >
> > > > -                     buf[idx] = (req->u.i2c_read.transactions[i].no_stop_bit &
> 0x1) << 5;
> > > > +                     buf[idx] =
> > > > + (req->u.i2c_read.transactions[i].no_stop_bit & 0x1) << 4;
> > > >                       buf[idx] |= (req-
> >u.i2c_read.transactions[i].i2c_transaction_delay & 0xf);
> > > >                       idx++;
> > > >               }
> > >
> > > --
> > > Jani Nikula, Intel Open Source Graphics Center
> > --
> > Wayne Lin
> > _______________________________________________
> > dri-devel mailing list
> > dri-devel@lists.freedesktop.org
> > https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flist
> > s.freedesktop.org%2Fmailman%2Flistinfo%2Fdri-
> devel&amp;data=02%7C01%7C
> >
> Wayne.Lin%40amd.com%7C7a299e0e845242312acb08d793a36e63%7C3dd896
> 1fe4884
> >
> e608e11a82d994e183d%7C0%7C0%7C637140202457938159&amp;sdata=yK4I
> 7fgenrR
> > f%2FXrs5jKXv%2BmOZdjV7dl%2BiNUJcsxnXG0%3D&amp;reserved=0
> 
> --
> Ville Syrjälä
> Intel
--
Best regards,
Wayne Lin

Patch
diff mbox series

diff --git a/drivers/gpu/drm/drm_dp_mst_topology.c b/drivers/gpu/drm/drm_dp_mst_topology.c
index 1d1bfa49ca2b..0557e225ff67 100644
--- a/drivers/gpu/drm/drm_dp_mst_topology.c
+++ b/drivers/gpu/drm/drm_dp_mst_topology.c
@@ -393,7 +393,7 @@  drm_dp_encode_sideband_req(const struct drm_dp_sideband_msg_req_body *req,
 			memcpy(&buf[idx], req->u.i2c_read.transactions[i].bytes, req->u.i2c_read.transactions[i].num_bytes);
 			idx += req->u.i2c_read.transactions[i].num_bytes;
 
-			buf[idx] = (req->u.i2c_read.transactions[i].no_stop_bit & 0x1) << 5;
+			buf[idx] = (req->u.i2c_read.transactions[i].no_stop_bit & 0x1) << 4;
 			buf[idx] |= (req->u.i2c_read.transactions[i].i2c_transaction_delay & 0xf);
 			idx++;
 		}