| Message ID | 20230419122131.2167122-1-thierry.reding@gmail.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | phy: mediatek: Avoid floating point constants | expand |
Il 19/04/23 14:21, Thierry Reding ha scritto: > From: Thierry Reding <treding@nvidia.com> > > When building with old versions of GCC (6.3 in this case), the compiler > stumbles over the floating point constants in this driver: > > drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c: In function ‘mtk_hdmi_pll_prepare’: > drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c:331:23: error: ‘-mgeneral-regs-only’ is incompatible with floating-point code > } else if (pixel_clk >= 74.175 * MEGA && pixel_clk <= 300 * MEGA) { > > drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c:361:12: error: ‘-mgeneral-regs-only’ is incompatible with floating-point code > static int mtk_hdmi_pll_prepare(struct clk_hw *hw) > ^~~~~~~~~~~~~~~~~~~~ > drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c:361:12: error: ‘-mgeneral-regs-only’ is incompatible with floating-point code > > Fix this by switching to the KILO macro instead and multiplying the > constants by 1000 to get rid of the floating point. > > Fixes: 45810d486bb4 ("phy: mediatek: add support for phy-mtk-hdmi-mt8195") > Reported-by: Jonathan Hunter <jonathanh@nvidia.com> > Signed-off-by: Thierry Reding <treding@nvidia.com> Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
On Wed, 19 Apr 2023 14:21, Thierry Reding <thierry.reding@gmail.com> wrote: >From: Thierry Reding <treding@nvidia.com> > >When building with old versions of GCC (6.3 in this case), the compiler >stumbles over the floating point constants in this driver: > > drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c: In function ‘mtk_hdmi_pll_prepare’: > drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c:331:23: error: ‘-mgeneral-regs-only’ is incompatible with floating-point code > } else if (pixel_clk >= 74.175 * MEGA && pixel_clk <= 300 * MEGA) { > > drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c:361:12: error: ‘-mgeneral-regs-only’ is incompatible with floating-point code > static int mtk_hdmi_pll_prepare(struct clk_hw *hw) > ^~~~~~~~~~~~~~~~~~~~ > drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c:361:12: error: ‘-mgeneral-regs-only’ is incompatible with floating-point code > >Fix this by switching to the KILO macro instead and multiplying the >constants by 1000 to get rid of the floating point. > >Fixes: 45810d486bb4 ("phy: mediatek: add support for phy-mtk-hdmi-mt8195") >Reported-by: Jonathan Hunter <jonathanh@nvidia.com> >Signed-off-by: Thierry Reding <treding@nvidia.com> >--- > drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c | 8 ++++---- > 1 file changed, 4 insertions(+), 4 deletions(-) > >diff --git a/drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c b/drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c >index abfc077fb0a8..b10af26cad2f 100644 >--- a/drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c >+++ b/drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c >@@ -239,9 +239,9 @@ static int mtk_hdmi_pll_calc(struct mtk_hdmi_phy *hdmi_phy, struct clk_hw *hw, > txposdiv = 8; > else if (tmds_clk >= 54 * MEGA && tmds_clk < 148.35 * MEGA) > txposdiv = 4; >- else if (tmds_clk >= 148.35 * MEGA && tmds_clk < 296.7 * MEGA) >+ else if (tmds_clk >= 148350 * KILO && tmds_clk < 296700 * KILO) > txposdiv = 2; >- else if (tmds_clk >= 296.7 * MEGA && tmds_clk <= 594 * MEGA) >+ else if (tmds_clk >= 296700 * KILO && tmds_clk <= 594 * MEGA) > txposdiv = 1; > else > return -EINVAL; >@@ -328,12 +328,12 @@ static int mtk_hdmi_pll_drv_setting(struct clk_hw *hw) > clk_channel_bias = 0x34; /* 20mA */ > impedance_en = 0xf; > impedance = 0x36; /* 100ohm */ >- } else if (pixel_clk >= 74.175 * MEGA && pixel_clk <= 300 * MEGA) { >+ } else if (pixel_clk >= 74175 * KILO && pixel_clk <= 300 * MEGA) { > data_channel_bias = 0x34; /* 20mA */ > clk_channel_bias = 0x2c; /* 16mA */ > impedance_en = 0xf; > impedance = 0x36; /* 100ohm */ >- } else if (pixel_clk >= 27 * MEGA && pixel_clk < 74.175 * MEGA) { >+ } else if (pixel_clk >= 27 * MEGA && pixel_clk < 74175 * KILO) { > data_channel_bias = 0x14; /* 10mA */ > clk_channel_bias = 0x14; /* 10mA */ > impedance_en = 0x0; >-- >2.40.0 > Thx for the fix! Reviewed-by: Guillaume Ranquet <granquet@baylibre.com>
Hi Vinod, On 19/04/2023 13:21, Thierry Reding wrote: > From: Thierry Reding <treding@nvidia.com> > > When building with old versions of GCC (6.3 in this case), the compiler > stumbles over the floating point constants in this driver: > > : In function ‘mtk_hdmi_pll_prepare’: > drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c:331:23: error: ‘-mgeneral-regs-only’ is incompatible with floating-point code > } else if (pixel_clk >= 74.175 * MEGA && pixel_clk <= 300 * MEGA) { > > drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c:361:12: error: ‘-mgeneral-regs-only’ is incompatible with floating-point code > static int mtk_hdmi_pll_prepare(struct clk_hw *hw) > ^~~~~~~~~~~~~~~~~~~~ > drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c:361:12: error: ‘-mgeneral-regs-only’ is incompatible with floating-point code > > Fix this by switching to the KILO macro instead and multiplying the > constants by 1000 to get rid of the floating point. > > Fixes: 45810d486bb4 ("phy: mediatek: add support for phy-mtk-hdmi-mt8195") > Reported-by: Jonathan Hunter <jonathanh@nvidia.com> > Signed-off-by: Thierry Reding <treding@nvidia.com> > --- > drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c | 8 ++++---- > 1 file changed, 4 insertions(+), 4 deletions(-) > > diff --git a/drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c b/drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c > index abfc077fb0a8..b10af26cad2f 100644 > --- a/drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c > +++ b/drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c > @@ -239,9 +239,9 @@ static int mtk_hdmi_pll_calc(struct mtk_hdmi_phy *hdmi_phy, struct clk_hw *hw, > txposdiv = 8; > else if (tmds_clk >= 54 * MEGA && tmds_clk < 148.35 * MEGA) > txposdiv = 4; > - else if (tmds_clk >= 148.35 * MEGA && tmds_clk < 296.7 * MEGA) > + else if (tmds_clk >= 148350 * KILO && tmds_clk < 296700 * KILO) > txposdiv = 2; > - else if (tmds_clk >= 296.7 * MEGA && tmds_clk <= 594 * MEGA) > + else if (tmds_clk >= 296700 * KILO && tmds_clk <= 594 * MEGA) > txposdiv = 1; > else > return -EINVAL; > @@ -328,12 +328,12 @@ static int mtk_hdmi_pll_drv_setting(struct clk_hw *hw) > clk_channel_bias = 0x34; /* 20mA */ > impedance_en = 0xf; > impedance = 0x36; /* 100ohm */ > - } else if (pixel_clk >= 74.175 * MEGA && pixel_clk <= 300 * MEGA) { > + } else if (pixel_clk >= 74175 * KILO && pixel_clk <= 300 * MEGA) { > data_channel_bias = 0x34; /* 20mA */ > clk_channel_bias = 0x2c; /* 16mA */ > impedance_en = 0xf; > impedance = 0x36; /* 100ohm */ > - } else if (pixel_clk >= 27 * MEGA && pixel_clk < 74.175 * MEGA) { > + } else if (pixel_clk >= 27 * MEGA && pixel_clk < 74175 * KILO) { > data_channel_bias = 0x14; /* 10mA */ > clk_channel_bias = 0x14; /* 10mA */ > impedance_en = 0x0; This is breaking kernel compilation with older compilers such as GCC7. This is now in the mainline and older compilers cannot build the kernel with this driver enabled. The kernel docs say the min version supported is GCC 5.1 currently. Can we get this merged as a fix for v6.4? Reviewed-by: Jon Hunter <jonathanh@nvidia.com> Tested-by: Jon Hunter <jonathanh@nvidia.com> Thanks Jon
On 17-05-23, 12:59, Jon Hunter wrote: > Hi Vinod, > > On 19/04/2023 13:21, Thierry Reding wrote: > > From: Thierry Reding <treding@nvidia.com> > > > > When building with old versions of GCC (6.3 in this case), the compiler > > stumbles over the floating point constants in this driver: > > > > : In function ‘mtk_hdmi_pll_prepare’: > > drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c:331:23: error: ‘-mgeneral-regs-only’ is incompatible with floating-point code > > } else if (pixel_clk >= 74.175 * MEGA && pixel_clk <= 300 * MEGA) { > > > > drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c:361:12: error: ‘-mgeneral-regs-only’ is incompatible with floating-point code > > static int mtk_hdmi_pll_prepare(struct clk_hw *hw) > > ^~~~~~~~~~~~~~~~~~~~ > > drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c:361:12: error: ‘-mgeneral-regs-only’ is incompatible with floating-point code > > > > Fix this by switching to the KILO macro instead and multiplying the > > constants by 1000 to get rid of the floating point. > > > > Fixes: 45810d486bb4 ("phy: mediatek: add support for phy-mtk-hdmi-mt8195") > > Reported-by: Jonathan Hunter <jonathanh@nvidia.com> > > Signed-off-by: Thierry Reding <treding@nvidia.com> > > --- > > drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c | 8 ++++---- > > 1 file changed, 4 insertions(+), 4 deletions(-) > > > > diff --git a/drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c b/drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c > > index abfc077fb0a8..b10af26cad2f 100644 > > --- a/drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c > > +++ b/drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c > > @@ -239,9 +239,9 @@ static int mtk_hdmi_pll_calc(struct mtk_hdmi_phy *hdmi_phy, struct clk_hw *hw, > > txposdiv = 8; > > else if (tmds_clk >= 54 * MEGA && tmds_clk < 148.35 * MEGA) > > txposdiv = 4; > > - else if (tmds_clk >= 148.35 * MEGA && tmds_clk < 296.7 * MEGA) > > + else if (tmds_clk >= 148350 * KILO && tmds_clk < 296700 * KILO) > > txposdiv = 2; > > - else if (tmds_clk >= 296.7 * MEGA && tmds_clk <= 594 * MEGA) > > + else if (tmds_clk >= 296700 * KILO && tmds_clk <= 594 * MEGA) > > txposdiv = 1; > > else > > return -EINVAL; > > @@ -328,12 +328,12 @@ static int mtk_hdmi_pll_drv_setting(struct clk_hw *hw) > > clk_channel_bias = 0x34; /* 20mA */ > > impedance_en = 0xf; > > impedance = 0x36; /* 100ohm */ > > - } else if (pixel_clk >= 74.175 * MEGA && pixel_clk <= 300 * MEGA) { > > + } else if (pixel_clk >= 74175 * KILO && pixel_clk <= 300 * MEGA) { > > data_channel_bias = 0x34; /* 20mA */ > > clk_channel_bias = 0x2c; /* 16mA */ > > impedance_en = 0xf; > > impedance = 0x36; /* 100ohm */ > > - } else if (pixel_clk >= 27 * MEGA && pixel_clk < 74.175 * MEGA) { > > + } else if (pixel_clk >= 27 * MEGA && pixel_clk < 74175 * KILO) { > > data_channel_bias = 0x14; /* 10mA */ > > clk_channel_bias = 0x14; /* 10mA */ > > impedance_en = 0x0; > > > This is breaking kernel compilation with older compilers such as GCC7. This > is now in the mainline and older compilers cannot build the kernel with this > driver enabled. The kernel docs say the min version supported is GCC 5.1 > currently. > > Can we get this merged as a fix for v6.4? Sorry I missed this one but I have already applied 03262a3f5b5b ("phy: mediatek: rework the floating point comparisons to fixed point") and this is in linux-next and should be hopefully in Linus's tree in next rc Thanks
diff --git a/drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c b/drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c index abfc077fb0a8..b10af26cad2f 100644 --- a/drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c +++ b/drivers/phy/mediatek/phy-mtk-hdmi-mt8195.c @@ -239,9 +239,9 @@ static int mtk_hdmi_pll_calc(struct mtk_hdmi_phy *hdmi_phy, struct clk_hw *hw, txposdiv = 8; else if (tmds_clk >= 54 * MEGA && tmds_clk < 148.35 * MEGA) txposdiv = 4; - else if (tmds_clk >= 148.35 * MEGA && tmds_clk < 296.7 * MEGA) + else if (tmds_clk >= 148350 * KILO && tmds_clk < 296700 * KILO) txposdiv = 2; - else if (tmds_clk >= 296.7 * MEGA && tmds_clk <= 594 * MEGA) + else if (tmds_clk >= 296700 * KILO && tmds_clk <= 594 * MEGA) txposdiv = 1; else return -EINVAL; @@ -328,12 +328,12 @@ static int mtk_hdmi_pll_drv_setting(struct clk_hw *hw) clk_channel_bias = 0x34; /* 20mA */ impedance_en = 0xf; impedance = 0x36; /* 100ohm */ - } else if (pixel_clk >= 74.175 * MEGA && pixel_clk <= 300 * MEGA) { + } else if (pixel_clk >= 74175 * KILO && pixel_clk <= 300 * MEGA) { data_channel_bias = 0x34; /* 20mA */ clk_channel_bias = 0x2c; /* 16mA */ impedance_en = 0xf; impedance = 0x36; /* 100ohm */ - } else if (pixel_clk >= 27 * MEGA && pixel_clk < 74.175 * MEGA) { + } else if (pixel_clk >= 27 * MEGA && pixel_clk < 74175 * KILO) { data_channel_bias = 0x14; /* 10mA */ clk_channel_bias = 0x14; /* 10mA */ impedance_en = 0x0;