Message ID | 576EFE30.4060501@bluematt.me |
---|---|
State | Changes Requested |
Delegated to: | Simon Glass |
Headers | show |
Hi Matt, On 25 June 2016 at 14:57, Matt Corallo <linux@bluematt.me> wrote: > Signed-off-by: Matt Corallo <git@bluematt.me> > --- > drivers/mmc/mmc.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/drivers/mmc/mmc.c b/drivers/mmc/mmc.c > index aabfc71..eba20f0 100644 > --- a/drivers/mmc/mmc.c > +++ b/drivers/mmc/mmc.c > @@ -1734,7 +1734,8 @@ int mmc_init(struct mmc *mmc) > #ifdef CONFIG_DM_MMC > struct mmc_uclass_priv *upriv = dev_get_uclass_priv(mmc->dev); > > - upriv->mmc = mmc; > + if (upriv) > + upriv->mmc = mmc; > #endif > if (mmc->has_init) > return 0; > -- > 2.1.4 Can you please add a commit message explaining why this is needed and what it fixes? How can mmc_init() be called before the MMC device is there? Is this related to this patch? http://patchwork.ozlabs.org/patch/640735/ Regards, Simon
On 29.06.2016 05:27, Simon Glass wrote: > Hi Matt, > > On 25 June 2016 at 14:57, Matt Corallo <linux@bluematt.me> wrote: >> Signed-off-by: Matt Corallo <git@bluematt.me> >> --- >> drivers/mmc/mmc.c | 3 ++- >> 1 file changed, 2 insertions(+), 1 deletion(-) >> >> diff --git a/drivers/mmc/mmc.c b/drivers/mmc/mmc.c >> index aabfc71..eba20f0 100644 >> --- a/drivers/mmc/mmc.c >> +++ b/drivers/mmc/mmc.c >> @@ -1734,7 +1734,8 @@ int mmc_init(struct mmc *mmc) >> #ifdef CONFIG_DM_MMC >> struct mmc_uclass_priv *upriv = dev_get_uclass_priv(mmc->dev); >> >> - upriv->mmc = mmc; >> + if (upriv) >> + upriv->mmc = mmc; >> #endif >> if (mmc->has_init) >> return 0; >> -- >> 2.1.4 > > Can you please add a commit message explaining why this is needed and > what it fixes? How can mmc_init() be called before the MMC device is > there? Is this related to this patch? > > http://patchwork.ozlabs.org/patch/640735/ It's related in a way that it fixed crash of dragonboard before I submitted my patch :) Not sure if it meant to fix dragonboard or some other board. Although - imho - it's nice to make check like that.. or at least some kind of assert. Mateusz
Hi Meteusz, On 30 June 2016 at 12:18, Mateusz Kulikowski <mateusz.kulikowski@gmail.com> wrote: > On 29.06.2016 05:27, Simon Glass wrote: >> Hi Matt, >> >> On 25 June 2016 at 14:57, Matt Corallo <linux@bluematt.me> wrote: >>> Signed-off-by: Matt Corallo <git@bluematt.me> >>> --- >>> drivers/mmc/mmc.c | 3 ++- >>> 1 file changed, 2 insertions(+), 1 deletion(-) >>> >>> diff --git a/drivers/mmc/mmc.c b/drivers/mmc/mmc.c >>> index aabfc71..eba20f0 100644 >>> --- a/drivers/mmc/mmc.c >>> +++ b/drivers/mmc/mmc.c >>> @@ -1734,7 +1734,8 @@ int mmc_init(struct mmc *mmc) >>> #ifdef CONFIG_DM_MMC >>> struct mmc_uclass_priv *upriv = dev_get_uclass_priv(mmc->dev); >>> >>> - upriv->mmc = mmc; >>> + if (upriv) >>> + upriv->mmc = mmc; >>> #endif >>> if (mmc->has_init) >>> return 0; >>> -- >>> 2.1.4 >> >> Can you please add a commit message explaining why this is needed and >> what it fixes? How can mmc_init() be called before the MMC device is >> there? Is this related to this patch? >> >> http://patchwork.ozlabs.org/patch/640735/ > > It's related in a way that it fixed crash of dragonboard before I submitted my patch :) > > Not sure if it meant to fix dragonboard or some other board. > > Although - imho - it's nice to make check like that.. or at least some kind of assert. An assert() would be fine with me. Regards, Simon
Sorry, this was related to https://patchwork.ozlabs.org/patch/624614/, not dragonboard, and yes, is the same issue that was addressed by the patch you linked. With this patch the MMC worked fine, but, indeed, as with dragonboard, its better to assign mmc instead of removing this (this seems to be the only place where mmc is actually used during boot for me, so maybe not?). Matt On 06/30/16 19:28, Simon Glass wrote: > Hi Meteusz, > > On 30 June 2016 at 12:18, Mateusz Kulikowski > <mateusz.kulikowski@gmail.com> wrote: >> On 29.06.2016 05:27, Simon Glass wrote: >>> Hi Matt, >>> >>> On 25 June 2016 at 14:57, Matt Corallo <linux@bluematt.me> wrote: >>>> Signed-off-by: Matt Corallo <git@bluematt.me> >>>> --- >>>> drivers/mmc/mmc.c | 3 ++- >>>> 1 file changed, 2 insertions(+), 1 deletion(-) >>>> >>>> diff --git a/drivers/mmc/mmc.c b/drivers/mmc/mmc.c >>>> index aabfc71..eba20f0 100644 >>>> --- a/drivers/mmc/mmc.c >>>> +++ b/drivers/mmc/mmc.c >>>> @@ -1734,7 +1734,8 @@ int mmc_init(struct mmc *mmc) >>>> #ifdef CONFIG_DM_MMC >>>> struct mmc_uclass_priv *upriv = dev_get_uclass_priv(mmc->dev); >>>> >>>> - upriv->mmc = mmc; >>>> + if (upriv) >>>> + upriv->mmc = mmc; >>>> #endif >>>> if (mmc->has_init) >>>> return 0; >>>> -- >>>> 2.1.4 >>> >>> Can you please add a commit message explaining why this is needed and >>> what it fixes? How can mmc_init() be called before the MMC device is >>> there? Is this related to this patch? >>> >>> http://patchwork.ozlabs.org/patch/640735/ >> >> It's related in a way that it fixed crash of dragonboard before I submitted my patch :) >> >> Not sure if it meant to fix dragonboard or some other board. >> >> Although - imho - it's nice to make check like that.. or at least some kind of assert. > > An assert() would be fine with me. > > Regards, > Simon >
On 07/01/2016 04:28 AM, Simon Glass wrote: > Hi Meteusz, > > On 30 June 2016 at 12:18, Mateusz Kulikowski > <mateusz.kulikowski@gmail.com> wrote: >> On 29.06.2016 05:27, Simon Glass wrote: >>> Hi Matt, >>> >>> On 25 June 2016 at 14:57, Matt Corallo <linux@bluematt.me> wrote: >>>> Signed-off-by: Matt Corallo <git@bluematt.me> >>>> --- >>>> drivers/mmc/mmc.c | 3 ++- >>>> 1 file changed, 2 insertions(+), 1 deletion(-) >>>> >>>> diff --git a/drivers/mmc/mmc.c b/drivers/mmc/mmc.c >>>> index aabfc71..eba20f0 100644 >>>> --- a/drivers/mmc/mmc.c >>>> +++ b/drivers/mmc/mmc.c >>>> @@ -1734,7 +1734,8 @@ int mmc_init(struct mmc *mmc) >>>> #ifdef CONFIG_DM_MMC >>>> struct mmc_uclass_priv *upriv = dev_get_uclass_priv(mmc->dev); >>>> >>>> - upriv->mmc = mmc; >>>> + if (upriv) >>>> + upriv->mmc = mmc; >>>> #endif >>>> if (mmc->has_init) >>>> return 0; >>>> -- >>>> 2.1.4 >>> >>> Can you please add a commit message explaining why this is needed and >>> what it fixes? How can mmc_init() be called before the MMC device is >>> there? Is this related to this patch? >>> >>> http://patchwork.ozlabs.org/patch/640735/ >> >> It's related in a way that it fixed crash of dragonboard before I submitted my patch :) >> >> Not sure if it meant to fix dragonboard or some other board. >> >> Although - imho - it's nice to make check like that.. or at least some kind of assert. I have just one question..Maybe it's not related with this patch.. But i saw other driver did that upriv->mmc is assigned to mmc in each drivers. What's difference with "upriv->mmc = mmc" in mmc.c? ./drivers/mmc/fsl_esdhc.c:972: upriv->mmc = priv->mmc; ./drivers/mmc/mmc.c:1606: upriv->mmc = mmc; ./drivers/mmc/mmc-uclass.c:91: return upriv->mmc; ./drivers/mmc/msm_sdhci.c:147: upriv->mmc = host->mmc; ./drivers/mmc/omap_hsmmc.c:830: upriv->mmc = mmc; ./drivers/mmc/rockchip_dw_mmc.c:114: upriv->mmc = host->mmc; ./drivers/mmc/socfpga_dw_mmc.c:110: upriv->mmc = host->mmc; ./drivers/mmc/uniphier-sd.c:727: upriv->mmc = priv->mmc; ./drivers/mmc/zynq_sdhci.c:52: upriv->mmc = host->mmc; Is it duplicated? (I didn't see in more detail.) Best Regards, Jaehoon Chung > > An assert() would be fine with me. > > Regards, > Simon > _______________________________________________ > U-Boot mailing list > U-Boot@lists.denx.de > http://lists.denx.de/mailman/listinfo/u-boot > > >
Hi, On 1 July 2016 at 04:37, Jaehoon Chung <jh80.chung@samsung.com> wrote: > On 07/01/2016 04:28 AM, Simon Glass wrote: >> Hi Meteusz, >> >> On 30 June 2016 at 12:18, Mateusz Kulikowski >> <mateusz.kulikowski@gmail.com> wrote: >>> On 29.06.2016 05:27, Simon Glass wrote: >>>> Hi Matt, >>>> >>>> On 25 June 2016 at 14:57, Matt Corallo <linux@bluematt.me> wrote: >>>>> Signed-off-by: Matt Corallo <git@bluematt.me> >>>>> --- >>>>> drivers/mmc/mmc.c | 3 ++- >>>>> 1 file changed, 2 insertions(+), 1 deletion(-) >>>>> >>>>> diff --git a/drivers/mmc/mmc.c b/drivers/mmc/mmc.c >>>>> index aabfc71..eba20f0 100644 >>>>> --- a/drivers/mmc/mmc.c >>>>> +++ b/drivers/mmc/mmc.c >>>>> @@ -1734,7 +1734,8 @@ int mmc_init(struct mmc *mmc) >>>>> #ifdef CONFIG_DM_MMC >>>>> struct mmc_uclass_priv *upriv = dev_get_uclass_priv(mmc->dev); >>>>> >>>>> - upriv->mmc = mmc; >>>>> + if (upriv) >>>>> + upriv->mmc = mmc; >>>>> #endif >>>>> if (mmc->has_init) >>>>> return 0; >>>>> -- >>>>> 2.1.4 >>>> >>>> Can you please add a commit message explaining why this is needed and >>>> what it fixes? How can mmc_init() be called before the MMC device is >>>> there? Is this related to this patch? >>>> >>>> http://patchwork.ozlabs.org/patch/640735/ >>> >>> It's related in a way that it fixed crash of dragonboard before I submitted my patch :) >>> >>> Not sure if it meant to fix dragonboard or some other board. >>> >>> Although - imho - it's nice to make check like that.. or at least some kind of assert. > > I have just one question..Maybe it's not related with this patch.. > But i saw other driver did that upriv->mmc is assigned to mmc in each drivers. > > What's difference with "upriv->mmc = mmc" in mmc.c? > > ./drivers/mmc/fsl_esdhc.c:972: upriv->mmc = priv->mmc; > ./drivers/mmc/mmc.c:1606: upriv->mmc = mmc; > ./drivers/mmc/mmc-uclass.c:91: return upriv->mmc; > ./drivers/mmc/msm_sdhci.c:147: upriv->mmc = host->mmc; > ./drivers/mmc/omap_hsmmc.c:830: upriv->mmc = mmc; > ./drivers/mmc/rockchip_dw_mmc.c:114: upriv->mmc = host->mmc; > ./drivers/mmc/socfpga_dw_mmc.c:110: upriv->mmc = host->mmc; > ./drivers/mmc/uniphier-sd.c:727: upriv->mmc = priv->mmc; > ./drivers/mmc/zynq_sdhci.c:52: upriv->mmc = host->mmc; > > Is it duplicated? > (I didn't see in more detail.) > > Best Regards, > Jaehoon Chung I am actually not happy with how this is at present. It should be handled by a function call. If you'd like to look at this please do, otherwise I will take a pass at it. It should be clear at this point what we need. > >> >> An assert() would be fine with me. >> >> Regards, >> Simon >> _______________________________________________ >> U-Boot mailing list >> U-Boot@lists.denx.de >> http://lists.denx.de/mailman/listinfo/u-boot >> >> >> > Regards, Simon
On 07/02/2016 03:17 AM, Simon Glass wrote: > Hi, > > On 1 July 2016 at 04:37, Jaehoon Chung <jh80.chung@samsung.com> wrote: >> On 07/01/2016 04:28 AM, Simon Glass wrote: >>> Hi Meteusz, >>> >>> On 30 June 2016 at 12:18, Mateusz Kulikowski >>> <mateusz.kulikowski@gmail.com> wrote: >>>> On 29.06.2016 05:27, Simon Glass wrote: >>>>> Hi Matt, >>>>> >>>>> On 25 June 2016 at 14:57, Matt Corallo <linux@bluematt.me> wrote: >>>>>> Signed-off-by: Matt Corallo <git@bluematt.me> >>>>>> --- >>>>>> drivers/mmc/mmc.c | 3 ++- >>>>>> 1 file changed, 2 insertions(+), 1 deletion(-) >>>>>> >>>>>> diff --git a/drivers/mmc/mmc.c b/drivers/mmc/mmc.c >>>>>> index aabfc71..eba20f0 100644 >>>>>> --- a/drivers/mmc/mmc.c >>>>>> +++ b/drivers/mmc/mmc.c >>>>>> @@ -1734,7 +1734,8 @@ int mmc_init(struct mmc *mmc) >>>>>> #ifdef CONFIG_DM_MMC >>>>>> struct mmc_uclass_priv *upriv = dev_get_uclass_priv(mmc->dev); >>>>>> >>>>>> - upriv->mmc = mmc; >>>>>> + if (upriv) >>>>>> + upriv->mmc = mmc; >>>>>> #endif >>>>>> if (mmc->has_init) >>>>>> return 0; >>>>>> -- >>>>>> 2.1.4 >>>>> >>>>> Can you please add a commit message explaining why this is needed and >>>>> what it fixes? How can mmc_init() be called before the MMC device is >>>>> there? Is this related to this patch? >>>>> >>>>> http://patchwork.ozlabs.org/patch/640735/ >>>> >>>> It's related in a way that it fixed crash of dragonboard before I submitted my patch :) >>>> >>>> Not sure if it meant to fix dragonboard or some other board. >>>> >>>> Although - imho - it's nice to make check like that.. or at least some kind of assert. >> >> I have just one question..Maybe it's not related with this patch.. >> But i saw other driver did that upriv->mmc is assigned to mmc in each drivers. >> >> What's difference with "upriv->mmc = mmc" in mmc.c? >> >> ./drivers/mmc/fsl_esdhc.c:972: upriv->mmc = priv->mmc; >> ./drivers/mmc/mmc.c:1606: upriv->mmc = mmc; >> ./drivers/mmc/mmc-uclass.c:91: return upriv->mmc; >> ./drivers/mmc/msm_sdhci.c:147: upriv->mmc = host->mmc; >> ./drivers/mmc/omap_hsmmc.c:830: upriv->mmc = mmc; >> ./drivers/mmc/rockchip_dw_mmc.c:114: upriv->mmc = host->mmc; >> ./drivers/mmc/socfpga_dw_mmc.c:110: upriv->mmc = host->mmc; >> ./drivers/mmc/uniphier-sd.c:727: upriv->mmc = priv->mmc; >> ./drivers/mmc/zynq_sdhci.c:52: upriv->mmc = host->mmc; >> >> Is it duplicated? >> (I didn't see in more detail.) >> >> Best Regards, >> Jaehoon Chung > > I am actually not happy with how this is at present. It should be > handled by a function call. If you'd like to look at this please do, > otherwise I will take a pass at it. It should be clear at this point > what we need. I will send the patch for cleaning. Best Regards, Jaehoon Chung > >> >>> >>> An assert() would be fine with me. >>> >>> Regards, >>> Simon >>> _______________________________________________ >>> U-Boot mailing list >>> U-Boot@lists.denx.de >>> http://lists.denx.de/mailman/listinfo/u-boot >>> >>> >>> >> > > > Regards, > Simon > > >
diff --git a/drivers/mmc/mmc.c b/drivers/mmc/mmc.c index aabfc71..eba20f0 100644 --- a/drivers/mmc/mmc.c +++ b/drivers/mmc/mmc.c @@ -1734,7 +1734,8 @@ int mmc_init(struct mmc *mmc) #ifdef CONFIG_DM_MMC struct mmc_uclass_priv *upriv = dev_get_uclass_priv(mmc->dev); - upriv->mmc = mmc; + if (upriv) + upriv->mmc = mmc; #endif if (mmc->has_init) return 0;
Signed-off-by: Matt Corallo <git@bluematt.me> --- drivers/mmc/mmc.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)