Message ID | 20191023174515.8892-1-heiko@sntech.de |
---|---|
State | Superseded |
Delegated to: | Kever Yang |
Headers | show |
Series | [U-Boot,1/2] rockchip: clk: rv1108: remove duplicate reset init | expand |
Heiko, On 2019/10/24 上午1:45, Heiko Stuebner wrote: > rockchip_reset_bind() already does the needed init for the reset > registers, only referenced the wrong cru structure. > > So we can get rid of the open-coded reset init and just fix > the correct cru reference. > > Signed-off-by: Heiko Stuebner <heiko@sntech.de> > --- > drivers/clk/rockchip/clk_rv1108.c | 13 +------------ > 1 file changed, 1 insertion(+), 12 deletions(-) > > diff --git a/drivers/clk/rockchip/clk_rv1108.c b/drivers/clk/rockchip/clk_rv1108.c > index 3ebb007fab..5dc31e1eb0 100644 > --- a/drivers/clk/rockchip/clk_rv1108.c > +++ b/drivers/clk/rockchip/clk_rv1108.c > @@ -698,22 +698,11 @@ static int rv1108_clk_bind(struct udevice *dev) > } > > #if CONFIG_IS_ENABLED(CONFIG_RESET_ROCKCHIP) > - ret = offsetof(struct rk3368_cru, softrst_con[0]); > + ret = offsetof(struct rv1108_cru, softrst_con[0]); > ret = rockchip_reset_bind(dev, ret, 13); > if (ret) > debug("Warning: software reset driver bind faile\n"); > #endif > - ret = device_bind_driver_to_node(dev, "rockchip_reset", "reset", > - dev_ofnode(dev), &sf_child); You can't just remove this blob of code, for there is a 'sysreset' driver and a 'reset' driver, they are different, so you should fix this part to be available 'sysreset' driver so that the software can reset the SoC. Thanks, - Kever > - if (ret) { > - debug("Warning: No rockchip reset driver: ret=%d\n", ret); > - } else { > - sf_priv = malloc(sizeof(struct softreset_reg)); > - sf_priv->sf_reset_offset = offsetof(struct rv1108_cru, > - softrst_con[0]); > - sf_priv->sf_reset_num = 13; > - sf_child->priv = sf_priv; > - } > > return 0; > }
Hi Kever, Am Donnerstag, 7. November 2019, 10:03:19 CET schrieb Kever Yang: > On 2019/10/24 上午1:45, Heiko Stuebner wrote: > > rockchip_reset_bind() already does the needed init for the reset > > registers, only referenced the wrong cru structure. > > > > So we can get rid of the open-coded reset init and just fix > > the correct cru reference. > > > > Signed-off-by: Heiko Stuebner <heiko@sntech.de> > > --- > > drivers/clk/rockchip/clk_rv1108.c | 13 +------------ > > 1 file changed, 1 insertion(+), 12 deletions(-) > > > > diff --git a/drivers/clk/rockchip/clk_rv1108.c b/drivers/clk/rockchip/clk_rv1108.c > > index 3ebb007fab..5dc31e1eb0 100644 > > --- a/drivers/clk/rockchip/clk_rv1108.c > > +++ b/drivers/clk/rockchip/clk_rv1108.c > > @@ -698,22 +698,11 @@ static int rv1108_clk_bind(struct udevice *dev) > > } > > > > #if CONFIG_IS_ENABLED(CONFIG_RESET_ROCKCHIP) > > - ret = offsetof(struct rk3368_cru, softrst_con[0]); > > + ret = offsetof(struct rv1108_cru, softrst_con[0]); > > ret = rockchip_reset_bind(dev, ret, 13); > > if (ret) > > debug("Warning: software reset driver bind faile\n"); > > #endif > > - ret = device_bind_driver_to_node(dev, "rockchip_reset", "reset", > > - dev_ofnode(dev), &sf_child); > > You can't just remove this blob of code, for there is a 'sysreset' > driver and a 'reset' driver, they are > > different, so you should fix this part to be available 'sysreset' driver > so that the software can reset the SoC. While there is a sysreset driver, it seems the rv1108 only did init the softrst code here and not the sysreset at all. So rockchip_reset_bind really only does the same as the remove part of code. What I did see was that struct softreset_reg still is defined in the rockchip clock.h and after this patch no-one is using at all anymore. The rockchip_reset driver uses a different struct altogether it seems. Adding an appropriate sysreset driver for rv1108 should likely be a separate patch though. Heiko > > - if (ret) { > > - debug("Warning: No rockchip reset driver: ret=%d\n", ret); > > - } else { > > - sf_priv = malloc(sizeof(struct softreset_reg)); > > - sf_priv->sf_reset_offset = offsetof(struct rv1108_cru, > > - softrst_con[0]); > > - sf_priv->sf_reset_num = 13; > > - sf_child->priv = sf_priv; > > - } > > > > return 0; > > } > > >
Hi Kever, Am Donnerstag, 7. November 2019, 10:13:02 CET schrieb Heiko Stübner: > Am Donnerstag, 7. November 2019, 10:03:19 CET schrieb Kever Yang: > > On 2019/10/24 上午1:45, Heiko Stuebner wrote: > > > rockchip_reset_bind() already does the needed init for the reset > > > registers, only referenced the wrong cru structure. > > > > > > So we can get rid of the open-coded reset init and just fix > > > the correct cru reference. > > > > > > Signed-off-by: Heiko Stuebner <heiko@sntech.de> > > > --- > > > drivers/clk/rockchip/clk_rv1108.c | 13 +------------ > > > 1 file changed, 1 insertion(+), 12 deletions(-) > > > > > > diff --git a/drivers/clk/rockchip/clk_rv1108.c b/drivers/clk/rockchip/clk_rv1108.c > > > index 3ebb007fab..5dc31e1eb0 100644 > > > --- a/drivers/clk/rockchip/clk_rv1108.c > > > +++ b/drivers/clk/rockchip/clk_rv1108.c > > > @@ -698,22 +698,11 @@ static int rv1108_clk_bind(struct udevice *dev) > > > } > > > > > > #if CONFIG_IS_ENABLED(CONFIG_RESET_ROCKCHIP) > > > - ret = offsetof(struct rk3368_cru, softrst_con[0]); > > > + ret = offsetof(struct rv1108_cru, softrst_con[0]); > > > ret = rockchip_reset_bind(dev, ret, 13); > > > if (ret) > > > debug("Warning: software reset driver bind faile\n"); > > > #endif > > > - ret = device_bind_driver_to_node(dev, "rockchip_reset", "reset", > > > - dev_ofnode(dev), &sf_child); > > > > You can't just remove this blob of code, for there is a 'sysreset' > > driver and a 'reset' driver, they are > > > > different, so you should fix this part to be available 'sysreset' driver > > so that the software can reset the SoC. > > While there is a sysreset driver, it seems the rv1108 only did init > the softrst code here and not the sysreset at all. > > So rockchip_reset_bind really only does the same as the remove > part of code. What I did see was that struct softreset_reg > still is defined in the rockchip clock.h and after this patch no-one > is using at all anymore. The rockchip_reset driver uses a different > struct altogether it seems. > > Adding an appropriate sysreset driver for rv1108 > should likely be a separate patch though. I just sent a v2, dropping the unused softreset_reg struct as well. When reviewing please take into account, the rv1108 had the following situation before: -registration of sysreset - registration of old softreset - broken because that driver code is gone - registration of new softreset - broken because of the CONFIG_* from patch2 and also because it was using the wrong struct (rk3368 instead of rv1108) now after the patch (in v2), we have the expected - registration of sysreset - registration of new softreset with correct cru struct Heiko
diff --git a/drivers/clk/rockchip/clk_rv1108.c b/drivers/clk/rockchip/clk_rv1108.c index 3ebb007fab..5dc31e1eb0 100644 --- a/drivers/clk/rockchip/clk_rv1108.c +++ b/drivers/clk/rockchip/clk_rv1108.c @@ -698,22 +698,11 @@ static int rv1108_clk_bind(struct udevice *dev) } #if CONFIG_IS_ENABLED(CONFIG_RESET_ROCKCHIP) - ret = offsetof(struct rk3368_cru, softrst_con[0]); + ret = offsetof(struct rv1108_cru, softrst_con[0]); ret = rockchip_reset_bind(dev, ret, 13); if (ret) debug("Warning: software reset driver bind faile\n"); #endif - ret = device_bind_driver_to_node(dev, "rockchip_reset", "reset", - dev_ofnode(dev), &sf_child); - if (ret) { - debug("Warning: No rockchip reset driver: ret=%d\n", ret); - } else { - sf_priv = malloc(sizeof(struct softreset_reg)); - sf_priv->sf_reset_offset = offsetof(struct rv1108_cru, - softrst_con[0]); - sf_priv->sf_reset_num = 13; - sf_child->priv = sf_priv; - } return 0; }
rockchip_reset_bind() already does the needed init for the reset registers, only referenced the wrong cru structure. So we can get rid of the open-coded reset init and just fix the correct cru reference. Signed-off-by: Heiko Stuebner <heiko@sntech.de> --- drivers/clk/rockchip/clk_rv1108.c | 13 +------------ 1 file changed, 1 insertion(+), 12 deletions(-)