| Message ID | 20190829104928.27404-1-yamada.masahiro@socionext.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [1/3] mmc: sdhci-cadence: enable v4_mode to fix ADMA 64-bit addressing | expand |
On Thu, 29 Aug 2019 at 12:49, Masahiro Yamada <yamada.masahiro@socionext.com> wrote: > > The IP datasheet says this controller is compatible with SD Host > Specification Version v4.00. > > As it turned out, the ADMA of this IP does not work with 64-bit mode > when it is in the Version 3.00 compatible mode; it understands the > old 64-bit descriptor table (as defined in SDHCI v2), but the ADMA > System Address Register (SDHCI_ADMA_ADDRESS) cannot point to the > 64-bit address. > > I noticed this issue only after commit bd2e75633c80 ("dma-contiguous: > use fallback alloc_pages for single pages"). Prior to that commit, > dma_set_mask_and_coherent() returned the dma address that fits in > 32-bit range, at least for the default arm64 configuration > (arch/arm64/configs/defconfig). Now the host->adma_addr exceeds the > 32-bit limit, causing the real problem for the Socionext SoCs. > (As a side-note, I was also able to reproduce the issue for older > kernels by turning off CONFIG_DMA_CMA.) > > Call sdhci_enable_v4_mode() to fix this. > > I think it is better to back-port this, but only possible for v4.20+. > > When this driver was merged (v4.10), the v4 mode support did not exist. > It was added by commit b3f80b434f72 ("mmc: sdhci: Add sd host v4 mode") > i.e. v4.20. > > Cc: <stable@vger.kernel.org> # v4.20+ > Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com> Applied for fixes, by adding below tag, thanks! Fixes: b3f80b434f72 ("mmc: sdhci: Add sd host v4 mode") Kind regards Uffe > --- > > drivers/mmc/host/sdhci-cadence.c | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/drivers/mmc/host/sdhci-cadence.c b/drivers/mmc/host/sdhci-cadence.c > index 163d1cf4367e..44139fceac24 100644 > --- a/drivers/mmc/host/sdhci-cadence.c > +++ b/drivers/mmc/host/sdhci-cadence.c > @@ -369,6 +369,7 @@ static int sdhci_cdns_probe(struct platform_device *pdev) > host->mmc_host_ops.execute_tuning = sdhci_cdns_execute_tuning; > host->mmc_host_ops.hs400_enhanced_strobe = > sdhci_cdns_hs400_enhanced_strobe; > + sdhci_enable_v4_mode(host); > > sdhci_get_of_property(pdev); > > -- > 2.17.1 >
On Thu, Aug 29, 2019 at 8:48 PM Ulf Hansson <ulf.hansson@linaro.org> wrote: > > On Thu, 29 Aug 2019 at 12:49, Masahiro Yamada > <yamada.masahiro@socionext.com> wrote: > > > > The IP datasheet says this controller is compatible with SD Host > > Specification Version v4.00. > > > > As it turned out, the ADMA of this IP does not work with 64-bit mode > > when it is in the Version 3.00 compatible mode; it understands the > > old 64-bit descriptor table (as defined in SDHCI v2), but the ADMA > > System Address Register (SDHCI_ADMA_ADDRESS) cannot point to the > > 64-bit address. > > > > I noticed this issue only after commit bd2e75633c80 ("dma-contiguous: > > use fallback alloc_pages for single pages"). Prior to that commit, > > dma_set_mask_and_coherent() returned the dma address that fits in > > 32-bit range, at least for the default arm64 configuration > > (arch/arm64/configs/defconfig). Now the host->adma_addr exceeds the > > 32-bit limit, causing the real problem for the Socionext SoCs. > > (As a side-note, I was also able to reproduce the issue for older > > kernels by turning off CONFIG_DMA_CMA.) > > > > Call sdhci_enable_v4_mode() to fix this. > > > > I think it is better to back-port this, but only possible for v4.20+. > > > > When this driver was merged (v4.10), the v4 mode support did not exist. > > It was added by commit b3f80b434f72 ("mmc: sdhci: Add sd host v4 mode") > > i.e. v4.20. > > > > Cc: <stable@vger.kernel.org> # v4.20+ > > Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com> > > Applied for fixes, by adding below tag, thanks! > > Fixes: b3f80b434f72 ("mmc: sdhci: Add sd host v4 mode") This is not a bug commit.
On Thu, 29 Aug 2019 at 14:05, Masahiro Yamada <yamada.masahiro@socionext.com> wrote: > > On Thu, Aug 29, 2019 at 8:48 PM Ulf Hansson <ulf.hansson@linaro.org> wrote: > > > > On Thu, 29 Aug 2019 at 12:49, Masahiro Yamada > > <yamada.masahiro@socionext.com> wrote: > > > > > > The IP datasheet says this controller is compatible with SD Host > > > Specification Version v4.00. > > > > > > As it turned out, the ADMA of this IP does not work with 64-bit mode > > > when it is in the Version 3.00 compatible mode; it understands the > > > old 64-bit descriptor table (as defined in SDHCI v2), but the ADMA > > > System Address Register (SDHCI_ADMA_ADDRESS) cannot point to the > > > 64-bit address. > > > > > > I noticed this issue only after commit bd2e75633c80 ("dma-contiguous: > > > use fallback alloc_pages for single pages"). Prior to that commit, > > > dma_set_mask_and_coherent() returned the dma address that fits in > > > 32-bit range, at least for the default arm64 configuration > > > (arch/arm64/configs/defconfig). Now the host->adma_addr exceeds the > > > 32-bit limit, causing the real problem for the Socionext SoCs. > > > (As a side-note, I was also able to reproduce the issue for older > > > kernels by turning off CONFIG_DMA_CMA.) > > > > > > Call sdhci_enable_v4_mode() to fix this. > > > > > > I think it is better to back-port this, but only possible for v4.20+. > > > > > > When this driver was merged (v4.10), the v4 mode support did not exist. > > > It was added by commit b3f80b434f72 ("mmc: sdhci: Add sd host v4 mode") > > > i.e. v4.20. > > > > > > Cc: <stable@vger.kernel.org> # v4.20+ > > > Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com> > > > > Applied for fixes, by adding below tag, thanks! > > > > Fixes: b3f80b434f72 ("mmc: sdhci: Add sd host v4 mode") > > This is not a bug commit. Right, but it can't be applied before this commit, hence why I added it. Not sure that it matters, but I can remove the tag if you insists!? Kind regards Uffe
Hi Ulf, On Thu, Aug 29, 2019 at 10:27 PM Ulf Hansson <ulf.hansson@linaro.org> wrote: > > On Thu, 29 Aug 2019 at 14:05, Masahiro Yamada > <yamada.masahiro@socionext.com> wrote: > > > > On Thu, Aug 29, 2019 at 8:48 PM Ulf Hansson <ulf.hansson@linaro.org> wrote: > > > > > > On Thu, 29 Aug 2019 at 12:49, Masahiro Yamada > > > <yamada.masahiro@socionext.com> wrote: > > > > > > > > The IP datasheet says this controller is compatible with SD Host > > > > Specification Version v4.00. > > > > > > > > As it turned out, the ADMA of this IP does not work with 64-bit mode > > > > when it is in the Version 3.00 compatible mode; it understands the > > > > old 64-bit descriptor table (as defined in SDHCI v2), but the ADMA > > > > System Address Register (SDHCI_ADMA_ADDRESS) cannot point to the > > > > 64-bit address. > > > > > > > > I noticed this issue only after commit bd2e75633c80 ("dma-contiguous: > > > > use fallback alloc_pages for single pages"). Prior to that commit, > > > > dma_set_mask_and_coherent() returned the dma address that fits in > > > > 32-bit range, at least for the default arm64 configuration > > > > (arch/arm64/configs/defconfig). Now the host->adma_addr exceeds the > > > > 32-bit limit, causing the real problem for the Socionext SoCs. > > > > (As a side-note, I was also able to reproduce the issue for older > > > > kernels by turning off CONFIG_DMA_CMA.) > > > > > > > > Call sdhci_enable_v4_mode() to fix this. > > > > > > > > I think it is better to back-port this, but only possible for v4.20+. > > > > > > > > When this driver was merged (v4.10), the v4 mode support did not exist. > > > > It was added by commit b3f80b434f72 ("mmc: sdhci: Add sd host v4 mode") > > > > i.e. v4.20. > > > > > > > > Cc: <stable@vger.kernel.org> # v4.20+ > > > > Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com> > > > > > > Applied for fixes, by adding below tag, thanks! > > > > > > Fixes: b3f80b434f72 ("mmc: sdhci: Add sd host v4 mode") > > > > This is not a bug commit. > > Right, but it can't be applied before this commit, hence why I added > it. Not sure that it matters, but I can remove the tag if you > insists!? I hesitate to add Fixes to the commit that did nothing wrong. I added "Cc: <stable@vger.kernel.org> # v4.20+" so this is enough for the stable kernel maintainers.
On Thu, 29 Aug 2019 at 16:04, Masahiro Yamada <yamada.masahiro@socionext.com> wrote: > > Hi Ulf, > > On Thu, Aug 29, 2019 at 10:27 PM Ulf Hansson <ulf.hansson@linaro.org> wrote: > > > > On Thu, 29 Aug 2019 at 14:05, Masahiro Yamada > > <yamada.masahiro@socionext.com> wrote: > > > > > > On Thu, Aug 29, 2019 at 8:48 PM Ulf Hansson <ulf.hansson@linaro.org> wrote: > > > > > > > > On Thu, 29 Aug 2019 at 12:49, Masahiro Yamada > > > > <yamada.masahiro@socionext.com> wrote: > > > > > > > > > > The IP datasheet says this controller is compatible with SD Host > > > > > Specification Version v4.00. > > > > > > > > > > As it turned out, the ADMA of this IP does not work with 64-bit mode > > > > > when it is in the Version 3.00 compatible mode; it understands the > > > > > old 64-bit descriptor table (as defined in SDHCI v2), but the ADMA > > > > > System Address Register (SDHCI_ADMA_ADDRESS) cannot point to the > > > > > 64-bit address. > > > > > > > > > > I noticed this issue only after commit bd2e75633c80 ("dma-contiguous: > > > > > use fallback alloc_pages for single pages"). Prior to that commit, > > > > > dma_set_mask_and_coherent() returned the dma address that fits in > > > > > 32-bit range, at least for the default arm64 configuration > > > > > (arch/arm64/configs/defconfig). Now the host->adma_addr exceeds the > > > > > 32-bit limit, causing the real problem for the Socionext SoCs. > > > > > (As a side-note, I was also able to reproduce the issue for older > > > > > kernels by turning off CONFIG_DMA_CMA.) > > > > > > > > > > Call sdhci_enable_v4_mode() to fix this. > > > > > > > > > > I think it is better to back-port this, but only possible for v4.20+. > > > > > > > > > > When this driver was merged (v4.10), the v4 mode support did not exist. > > > > > It was added by commit b3f80b434f72 ("mmc: sdhci: Add sd host v4 mode") > > > > > i.e. v4.20. > > > > > > > > > > Cc: <stable@vger.kernel.org> # v4.20+ > > > > > Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com> > > > > > > > > Applied for fixes, by adding below tag, thanks! > > > > > > > > Fixes: b3f80b434f72 ("mmc: sdhci: Add sd host v4 mode") > > > > > > This is not a bug commit. > > > > Right, but it can't be applied before this commit, hence why I added > > it. Not sure that it matters, but I can remove the tag if you > > insists!? > > I hesitate to add Fixes to the commit that > did nothing wrong. > > > I added "Cc: <stable@vger.kernel.org> # v4.20+" > so this is enough for the stable kernel maintainers. Yes, I drop the fixes tag. But as I said, I don't think it's a big deal. Kind regards Uffe
diff --git a/drivers/mmc/host/sdhci-cadence.c b/drivers/mmc/host/sdhci-cadence.c index 163d1cf4367e..44139fceac24 100644 --- a/drivers/mmc/host/sdhci-cadence.c +++ b/drivers/mmc/host/sdhci-cadence.c @@ -369,6 +369,7 @@ static int sdhci_cdns_probe(struct platform_device *pdev) host->mmc_host_ops.execute_tuning = sdhci_cdns_execute_tuning; host->mmc_host_ops.hs400_enhanced_strobe = sdhci_cdns_hs400_enhanced_strobe; + sdhci_enable_v4_mode(host); sdhci_get_of_property(pdev);
The IP datasheet says this controller is compatible with SD Host Specification Version v4.00. As it turned out, the ADMA of this IP does not work with 64-bit mode when it is in the Version 3.00 compatible mode; it understands the old 64-bit descriptor table (as defined in SDHCI v2), but the ADMA System Address Register (SDHCI_ADMA_ADDRESS) cannot point to the 64-bit address. I noticed this issue only after commit bd2e75633c80 ("dma-contiguous: use fallback alloc_pages for single pages"). Prior to that commit, dma_set_mask_and_coherent() returned the dma address that fits in 32-bit range, at least for the default arm64 configuration (arch/arm64/configs/defconfig). Now the host->adma_addr exceeds the 32-bit limit, causing the real problem for the Socionext SoCs. (As a side-note, I was also able to reproduce the issue for older kernels by turning off CONFIG_DMA_CMA.) Call sdhci_enable_v4_mode() to fix this. I think it is better to back-port this, but only possible for v4.20+. When this driver was merged (v4.10), the v4 mode support did not exist. It was added by commit b3f80b434f72 ("mmc: sdhci: Add sd host v4 mode") i.e. v4.20. Cc: <stable@vger.kernel.org> # v4.20+ Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com> --- drivers/mmc/host/sdhci-cadence.c | 1 + 1 file changed, 1 insertion(+)