Message ID | 1443589243-9521-1-git-send-email-sr@denx.de |
---|---|
State | Accepted |
Delegated to: | Simon Glass |
Headers | show |
On 09/29/2015 11:00 PM, Stefan Roese wrote: > The current "simple" address translation simple_bus_translate() is not > working on some platforms (e.g. MVEBU). As here more complex "ranges" > properties are used in many nodes (multiple tuples etc). This patch > enables the optional use of the common fdt_translate_address() function > which handles this translation correctly. This change makes sense to me, but one comment: > diff --git a/drivers/core/Kconfig b/drivers/core/Kconfig > +config OF_TRANSLATE > + bool "Translate addresses using fdt_translate_address" > + depends on DM && OF_CONTROL > + default y So this is on by default, which I think is correct since applying this technique is required to parse DT correctly. However, ... > diff --git a/drivers/core/device.c b/drivers/core/device.c > + if (CONFIG_IS_ENABLED(OF_TRANSLATE)) { > + /* > + * Use the full-fledged translate function for complex > + * bus setups. > + */ > + return fdt_translate_address((void *)gd->fdt_blob, > + dev->of_offset, reg); fdt_translate_address() is a simple wrapper around __of_translate_address(), and that function calls fdt_parent_offset() which is "slow" per Simon. Surely this patch will receive the same objection as when I added a (single) call to fdt_parent_offset() into the DT address parsing routine (and this patch is worse, since it adds a call to fdt_parent_offset() for each level of DT sub-nodes).
Hi Stephen, On 30.09.2015 18:13, Stephen Warren wrote: > On 09/29/2015 11:00 PM, Stefan Roese wrote: >> The current "simple" address translation simple_bus_translate() is not >> working on some platforms (e.g. MVEBU). As here more complex "ranges" >> properties are used in many nodes (multiple tuples etc). This patch >> enables the optional use of the common fdt_translate_address() function >> which handles this translation correctly. > > This change makes sense to me, but one comment: > >> diff --git a/drivers/core/Kconfig b/drivers/core/Kconfig > >> +config OF_TRANSLATE >> + bool "Translate addresses using fdt_translate_address" >> + depends on DM && OF_CONTROL >> + default y > > So this is on by default, which I think is correct since applying this > technique is required to parse DT correctly. However, ... > >> diff --git a/drivers/core/device.c b/drivers/core/device.c > >> + if (CONFIG_IS_ENABLED(OF_TRANSLATE)) { > >> + /* >> + * Use the full-fledged translate function for complex >> + * bus setups. >> + */ >> + return fdt_translate_address((void *)gd->fdt_blob, >> + dev->of_offset, reg); > > fdt_translate_address() is a simple wrapper around > __of_translate_address(), and that function calls fdt_parent_offset() > which is "slow" per Simon. Surely this patch will receive the same > objection as when I added a (single) call to fdt_parent_offset() into > the DT address parsing routine (and this patch is worse, since it adds a > call to fdt_parent_offset() for each level of DT sub-nodes). Yes, Simon already mentioned his speed related concerns. But I think I convinced him that calling this function once for each driver probe can't be that bad this in the v2 thread: https://patchwork.ozlabs.org/patch/514331/ Simon, is this okay? Will you pull this patch after the v2015.10 release? Thanks, Stefan
Hi Stefan, On 1 October 2015 at 07:59, Stefan Roese <sr@denx.de> wrote: > Hi Stephen, > > > On 30.09.2015 18:13, Stephen Warren wrote: >> >> On 09/29/2015 11:00 PM, Stefan Roese wrote: >>> >>> The current "simple" address translation simple_bus_translate() is not >>> working on some platforms (e.g. MVEBU). As here more complex "ranges" >>> properties are used in many nodes (multiple tuples etc). This patch >>> enables the optional use of the common fdt_translate_address() function >>> which handles this translation correctly. >> >> >> This change makes sense to me, but one comment: >> >>> diff --git a/drivers/core/Kconfig b/drivers/core/Kconfig >> >> >>> +config OF_TRANSLATE >>> + bool "Translate addresses using fdt_translate_address" >>> + depends on DM && OF_CONTROL >>> + default y >> >> >> So this is on by default, which I think is correct since applying this >> technique is required to parse DT correctly. However, ... >> >>> diff --git a/drivers/core/device.c b/drivers/core/device.c >> >> >>> + if (CONFIG_IS_ENABLED(OF_TRANSLATE)) { >> >> >>> + /* >>> + * Use the full-fledged translate function for complex >>> + * bus setups. >>> + */ >>> + return fdt_translate_address((void *)gd->fdt_blob, >>> + dev->of_offset, reg); >> >> >> fdt_translate_address() is a simple wrapper around >> __of_translate_address(), and that function calls fdt_parent_offset() >> which is "slow" per Simon. Surely this patch will receive the same >> objection as when I added a (single) call to fdt_parent_offset() into >> the DT address parsing routine (and this patch is worse, since it adds a >> call to fdt_parent_offset() for each level of DT sub-nodes). > > > Yes, Simon already mentioned his speed related concerns. But I think > I convinced him that calling this function once for each driver probe > can't be that bad this in the v2 thread: > > https://patchwork.ozlabs.org/patch/514331/ > > Simon, is this okay? Will you pull this patch after the v2015.10 > release? It's not great but it is enabled by an option so I think it is OK for now. We talked about the alternatives and they were not too great either. Things like this suggest that we might move to an unflattened device tree post-relocation one day. Regards, Simon
On 29 September 2015 at 23:00, Stefan Roese <sr@denx.de> wrote: > The current "simple" address translation simple_bus_translate() is not > working on some platforms (e.g. MVEBU). As here more complex "ranges" > properties are used in many nodes (multiple tuples etc). This patch > enables the optional use of the common fdt_translate_address() function > which handles this translation correctly. > > Signed-off-by: Stefan Roese <sr@denx.de> > Cc: Simon Glass <sjg@chromium.org> > Cc: Bin Meng <bmeng.cn@gmail.com> > Cc: Marek Vasut <marex@denx.de> > Cc: Masahiro Yamada <yamada.masahiro@socionext.com> > Cc: Stephen Warren <swarren@nvidia.com> > Cc: Lukasz Majewski <l.majewski@samsung.com> > --- > v3: > - Rebased on current U-Boot version > - Added Stephen and Lukasz to Cc > > v2: > - Rework code a bit as suggested by Simon. Also added some comments > to make the use of the code paths more clear. > > drivers/core/Kconfig | 30 ++++++++++++++++++++++++++++++ > drivers/core/device.c | 20 ++++++++++++++++++++ > 2 files changed, 50 insertions(+) Applied to u-boot-dm, thanks!
Hi Stefan, Simon, On Mon, Oct 19, 2015 at 7:16 AM, Simon Glass <sjg@chromium.org> wrote: > On 29 September 2015 at 23:00, Stefan Roese <sr@denx.de> wrote: >> The current "simple" address translation simple_bus_translate() is not >> working on some platforms (e.g. MVEBU). As here more complex "ranges" >> properties are used in many nodes (multiple tuples etc). This patch >> enables the optional use of the common fdt_translate_address() function >> which handles this translation correctly. >> >> Signed-off-by: Stefan Roese <sr@denx.de> >> Cc: Simon Glass <sjg@chromium.org> >> Cc: Bin Meng <bmeng.cn@gmail.com> >> Cc: Marek Vasut <marex@denx.de> >> Cc: Masahiro Yamada <yamada.masahiro@socionext.com> >> Cc: Stephen Warren <swarren@nvidia.com> >> Cc: Lukasz Majewski <l.majewski@samsung.com> >> --- >> v3: >> - Rebased on current U-Boot version >> - Added Stephen and Lukasz to Cc >> >> v2: >> - Rework code a bit as suggested by Simon. Also added some comments >> to make the use of the code paths more clear. >> >> drivers/core/Kconfig | 30 ++++++++++++++++++++++++++++++ >> drivers/core/device.c | 20 ++++++++++++++++++++ >> 2 files changed, 50 insertions(+) > > Applied to u-boot-dm, thanks! When testing Simon's patch [1], I found PCI UART on Intel Crown Bay no longer works. git bisect leads to this commit. Somehow I missed this patch before although I see the commit message get me cc'ed but the email did not bring to my attention. I see this patch introduced OF_TRANSLATE and by default set it to y. This makes the code logic in dev_get_addr() go through fdt_translate_address(), which breaks the things. Should we set OF_TRANSLATE to n by default? If set to y, this requires dts to have complete ranges property everywhere. [1]: http://patchwork.ozlabs.org/patch/549799/ Regards, Bin
Hi Bin, On 03.12.2015 14:34, Bin Meng wrote: > Hi Stefan, Simon, > > On Mon, Oct 19, 2015 at 7:16 AM, Simon Glass <sjg@chromium.org> wrote: >> On 29 September 2015 at 23:00, Stefan Roese <sr@denx.de> wrote: >>> The current "simple" address translation simple_bus_translate() is not >>> working on some platforms (e.g. MVEBU). As here more complex "ranges" >>> properties are used in many nodes (multiple tuples etc). This patch >>> enables the optional use of the common fdt_translate_address() function >>> which handles this translation correctly. >>> >>> Signed-off-by: Stefan Roese <sr@denx.de> >>> Cc: Simon Glass <sjg@chromium.org> >>> Cc: Bin Meng <bmeng.cn@gmail.com> >>> Cc: Marek Vasut <marex@denx.de> >>> Cc: Masahiro Yamada <yamada.masahiro@socionext.com> >>> Cc: Stephen Warren <swarren@nvidia.com> >>> Cc: Lukasz Majewski <l.majewski@samsung.com> >>> --- >>> v3: >>> - Rebased on current U-Boot version >>> - Added Stephen and Lukasz to Cc >>> >>> v2: >>> - Rework code a bit as suggested by Simon. Also added some comments >>> to make the use of the code paths more clear. >>> >>> drivers/core/Kconfig | 30 ++++++++++++++++++++++++++++++ >>> drivers/core/device.c | 20 ++++++++++++++++++++ >>> 2 files changed, 50 insertions(+) >> >> Applied to u-boot-dm, thanks! > > When testing Simon's patch [1], I found PCI UART on Intel Crown Bay no > longer works. git bisect leads to this commit. Somehow I missed this > patch before although I see the commit message get me cc'ed but the > email did not bring to my attention. > > I see this patch introduced OF_TRANSLATE and by default set it to y. > This makes the code logic in dev_get_addr() go through > fdt_translate_address(), which breaks the things. I'm a bit surprised that using the common fdt_translate_address() function instead of the DM internal simple_bus_translate() causes problems on your platform. Are you sure that the ranges are described correctly in your dts? Is the dts a copy from the Linux original one? Ah, probably not, since we're talking about x86 which has no DT support in Linux, right? > Should we set > OF_TRANSLATE to n by default? If set to y, this requires dts to have > complete ranges property everywhere. My understanding here is that x86 is a special case. As it doesn't use the full-blown dts sources from Linux. But most likely some "simple" ones, written exactly for U-Boot / DM. I would still prefer to have this OF_TRANSLATE set to y as default. As its needed for at least some platforms. But if we decide to set it to n, I can live with it as well. Thanks, Stefan
On 12/03/2015 07:12 AM, Stefan Roese wrote: > Hi Bin, > > On 03.12.2015 14:34, Bin Meng wrote: >> Hi Stefan, Simon, >> >> On Mon, Oct 19, 2015 at 7:16 AM, Simon Glass <sjg@chromium.org> wrote: >>> On 29 September 2015 at 23:00, Stefan Roese <sr@denx.de> wrote: >>>> The current "simple" address translation simple_bus_translate() is not >>>> working on some platforms (e.g. MVEBU). As here more complex "ranges" >>>> properties are used in many nodes (multiple tuples etc). This patch >>>> enables the optional use of the common fdt_translate_address() function >>>> which handles this translation correctly. >>>> >>>> Signed-off-by: Stefan Roese <sr@denx.de> >>>> Cc: Simon Glass <sjg@chromium.org> >>>> Cc: Bin Meng <bmeng.cn@gmail.com> >>>> Cc: Marek Vasut <marex@denx.de> >>>> Cc: Masahiro Yamada <yamada.masahiro@socionext.com> >>>> Cc: Stephen Warren <swarren@nvidia.com> >>>> Cc: Lukasz Majewski <l.majewski@samsung.com> >>>> --- >>>> v3: >>>> - Rebased on current U-Boot version >>>> - Added Stephen and Lukasz to Cc >>>> >>>> v2: >>>> - Rework code a bit as suggested by Simon. Also added some comments >>>> to make the use of the code paths more clear. >>>> >>>> drivers/core/Kconfig | 30 ++++++++++++++++++++++++++++++ >>>> drivers/core/device.c | 20 ++++++++++++++++++++ >>>> 2 files changed, 50 insertions(+) >>> >>> Applied to u-boot-dm, thanks! >> >> When testing Simon's patch [1], I found PCI UART on Intel Crown Bay no >> longer works. git bisect leads to this commit. Somehow I missed this >> patch before although I see the commit message get me cc'ed but the >> email did not bring to my attention. >> >> I see this patch introduced OF_TRANSLATE and by default set it to y. >> This makes the code logic in dev_get_addr() go through >> fdt_translate_address(), which breaks the things. > > I'm a bit surprised that using the common fdt_translate_address() > function instead of the DM internal simple_bus_translate() causes > problems on your platform. Are you sure that the ranges are > described correctly in your dts? Is the dts a copy from the Linux > original one? Ah, probably not, since we're talking about x86 > which has no DT support in Linux, right? > >> Should we set >> OF_TRANSLATE to n by default? If set to y, this requires dts to have >> complete ranges property everywhere. > > My understanding here is that x86 is a special case. As it doesn't > use the full-blown dts sources from Linux. But most likely some > "simple" ones, written exactly for U-Boot / DM. > > I would still prefer to have this OF_TRANSLATE set to y as default. > As its needed for at least some platforms. But if we decide to > set it to n, I can live with it as well. Is this the driver that uses U-Boot functions that were intended to parse a standard "reg" property to parse some property other than reg, with different semantics? Actually, I think I'm remembering an issue with SPI on some x86 device, but perhaps the problem here is something similar.
Hi Stefan, On Thu, Dec 3, 2015 at 10:12 PM, Stefan Roese <sr@denx.de> wrote: > Hi Bin, > > > On 03.12.2015 14:34, Bin Meng wrote: >> >> Hi Stefan, Simon, >> >> On Mon, Oct 19, 2015 at 7:16 AM, Simon Glass <sjg@chromium.org> wrote: >>> >>> On 29 September 2015 at 23:00, Stefan Roese <sr@denx.de> wrote: >>>> >>>> The current "simple" address translation simple_bus_translate() is not >>>> working on some platforms (e.g. MVEBU). As here more complex "ranges" >>>> properties are used in many nodes (multiple tuples etc). This patch >>>> enables the optional use of the common fdt_translate_address() function >>>> which handles this translation correctly. >>>> >>>> Signed-off-by: Stefan Roese <sr@denx.de> >>>> Cc: Simon Glass <sjg@chromium.org> >>>> Cc: Bin Meng <bmeng.cn@gmail.com> >>>> Cc: Marek Vasut <marex@denx.de> >>>> Cc: Masahiro Yamada <yamada.masahiro@socionext.com> >>>> Cc: Stephen Warren <swarren@nvidia.com> >>>> Cc: Lukasz Majewski <l.majewski@samsung.com> >>>> --- >>>> v3: >>>> - Rebased on current U-Boot version >>>> - Added Stephen and Lukasz to Cc >>>> >>>> v2: >>>> - Rework code a bit as suggested by Simon. Also added some comments >>>> to make the use of the code paths more clear. >>>> >>>> drivers/core/Kconfig | 30 ++++++++++++++++++++++++++++++ >>>> drivers/core/device.c | 20 ++++++++++++++++++++ >>>> 2 files changed, 50 insertions(+) >>> >>> >>> Applied to u-boot-dm, thanks! >> >> >> When testing Simon's patch [1], I found PCI UART on Intel Crown Bay no >> longer works. git bisect leads to this commit. Somehow I missed this >> patch before although I see the commit message get me cc'ed but the >> email did not bring to my attention. >> >> I see this patch introduced OF_TRANSLATE and by default set it to y. >> This makes the code logic in dev_get_addr() go through >> fdt_translate_address(), which breaks the things. > > > I'm a bit surprised that using the common fdt_translate_address() > function instead of the DM internal simple_bus_translate() causes > problems on your platform. Are you sure that the ranges are > described correctly in your dts? Is the dts a copy from the Linux > original one? Ah, probably not, since we're talking about x86 > which has no DT support in Linux, right? > Is fdt_translate_address() able to handle PCI bus ranges property? PCI has special ranges. The arch/x86/dts/crownbay.dts has something like below: 90 pci { 91 #address-cells = <3>; 92 #size-cells = <2>; 93 compatible = "pci-x86"; 94 u-boot,dm-pre-reloc; 95 ranges = <0x02000000 0x0 0x40000000 0x40000000 0 0x80000000 96 0x42000000 0x0 0xc0000000 0xc0000000 0 0x20000000 97 0x01000000 0x0 0x2000 0x2000 0 0xe000>; 98 99 pcie@17,0 { 100 #address-cells = <3>; 101 #size-cells = <2>; 102 compatible = "pci-bridge"; 103 u-boot,dm-pre-reloc; 104 reg = <0x0000b800 0x0 0x0 0x0 0x0>; >> Should we set >> OF_TRANSLATE to n by default? If set to y, this requires dts to have >> complete ranges property everywhere. > > > My understanding here is that x86 is a special case. As it doesn't > use the full-blown dts sources from Linux. But most likely some > "simple" ones, written exactly for U-Boot / DM. > > I would still prefer to have this OF_TRANSLATE set to y as default. > As its needed for at least some platforms. But if we decide to > set it to n, I can live with it as well. > Regards, Bin
diff --git a/drivers/core/Kconfig b/drivers/core/Kconfig index 41f4e69..15681df 100644 --- a/drivers/core/Kconfig +++ b/drivers/core/Kconfig @@ -120,4 +120,34 @@ config SPL_SIMPLE_BUS Supports the 'simple-bus' driver, which is used on some systems in SPL. +config OF_TRANSLATE + bool "Translate addresses using fdt_translate_address" + depends on DM && OF_CONTROL + default y + help + If this option is enabled, the reg property will be translated + using the fdt_translate_address() function. This is necessary + on some platforms (e.g. MVEBU) using complex "ranges" + properties in many nodes. As this translation is not handled + correctly in the default simple_bus_translate() function. + + If this option is not enabled, simple_bus_translate() will be + used for the address translation. This function is faster and + smaller in size than fdt_translate_address(). + +config SPL_OF_TRANSLATE + bool "Translate addresses using fdt_translate_address" + depends on SPL_DM && SPL_OF_CONTROL + default n + help + If this option is enabled, the reg property will be translated + using the fdt_translate_address() function. This is necessary + on some platforms (e.g. MVEBU) using complex "ranges" + properties in many nodes. As this translation is not handled + correctly in the default simple_bus_translate() function. + + If this option is not enabled, simple_bus_translate() will be + used for the address translation. This function is faster and + smaller in size than fdt_translate_address(). + endmenu diff --git a/drivers/core/device.c b/drivers/core/device.c index 0bc04d4..92fb854 100644 --- a/drivers/core/device.c +++ b/drivers/core/device.c @@ -11,6 +11,7 @@ #include <common.h> #include <fdtdec.h> +#include <fdt_support.h> #include <malloc.h> #include <dm/device.h> #include <dm/device-internal.h> @@ -581,6 +582,25 @@ fdt_addr_t dev_get_addr(struct udevice *dev) #if CONFIG_IS_ENABLED(OF_CONTROL) fdt_addr_t addr; + if (CONFIG_IS_ENABLED(OF_TRANSLATE)) { + const fdt32_t *reg; + + reg = fdt_getprop(gd->fdt_blob, dev->of_offset, "reg", NULL); + if (!reg) + return FDT_ADDR_T_NONE; + + /* + * Use the full-fledged translate function for complex + * bus setups. + */ + return fdt_translate_address((void *)gd->fdt_blob, + dev->of_offset, reg); + } + + /* + * Use the "simple" translate function for less complex + * bus setups. + */ addr = fdtdec_get_addr_size_auto_parent(gd->fdt_blob, dev->parent->of_offset, dev->of_offset, "reg",
The current "simple" address translation simple_bus_translate() is not working on some platforms (e.g. MVEBU). As here more complex "ranges" properties are used in many nodes (multiple tuples etc). This patch enables the optional use of the common fdt_translate_address() function which handles this translation correctly. Signed-off-by: Stefan Roese <sr@denx.de> Cc: Simon Glass <sjg@chromium.org> Cc: Bin Meng <bmeng.cn@gmail.com> Cc: Marek Vasut <marex@denx.de> Cc: Masahiro Yamada <yamada.masahiro@socionext.com> Cc: Stephen Warren <swarren@nvidia.com> Cc: Lukasz Majewski <l.majewski@samsung.com> --- v3: - Rebased on current U-Boot version - Added Stephen and Lukasz to Cc v2: - Rework code a bit as suggested by Simon. Also added some comments to make the use of the code paths more clear. drivers/core/Kconfig | 30 ++++++++++++++++++++++++++++++ drivers/core/device.c | 20 ++++++++++++++++++++ 2 files changed, 50 insertions(+)