| Message ID | 1312981422-13294-4-git-send-email-lee.jones@linaro.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Wed, Aug 10, 2011 at 02:03:42PM +0100, Lee Jones wrote: > At the request of Arnd Bergmann this patch moves all SoC > platform devices found in sysfs from /sys/devices/platform to > /sys/devices/soc/<SoCNum>/. It is believed as the devices are > SoC specific and a /sys/devices/soc node has recently become > available, that this would be a more appropriate place to > display the data. And what userspace tools did you just break by doing this? greg k-h
On 08/10/2011 05:02 PM, Greg KH wrote: > On Wed, Aug 10, 2011 at 02:03:42PM +0100, Lee Jones wrote: > >> At the request of Arnd Bergmann this patch moves all SoC >> platform devices found in sysfs from /sys/devices/platform to >> /sys/devices/soc/<SoCNum>/. It is believed as the devices are >> SoC specific and a /sys/devices/soc node has recently become >> available, that this would be a more appropriate place to >> display the data. >> > And what userspace tools did you just break by doing this? > I think most of our userspace for this platform poking around in sysfs use /sys/class/* so should be pretty OK. And none of which were properly documented in Documentation/ABI/* anyway, so do we care? grep -r 'devices/platform' Documentation/ABI/ doesn't show anything relating to the ux500. The major breakage would be out-of-tree stuff, if any. Which is out-of-tree due to the absence of a SoC ID framework. Which was nixed by Arnd et al precisely because the above change was not part of it. It's a bit hard for Lee to please everyone with this one I suspect :-D Thanks, Linus Walleij
On Thu, Aug 11, 2011 at 01:57:13PM +0200, Linus Walleij wrote: > On 08/10/2011 05:02 PM, Greg KH wrote: > >On Wed, Aug 10, 2011 at 02:03:42PM +0100, Lee Jones wrote: > >>At the request of Arnd Bergmann this patch moves all SoC > >>platform devices found in sysfs from /sys/devices/platform to > >>/sys/devices/soc/<SoCNum>/. It is believed as the devices are > >>SoC specific and a /sys/devices/soc node has recently become > >>available, that this would be a more appropriate place to > >>display the data. > >And what userspace tools did you just break by doing this? > > I think most of our userspace for this platform poking > around in sysfs use /sys/class/* so should be pretty OK. > > And none of which were properly documented in > Documentation/ABI/* anyway, so do we care? > grep -r 'devices/platform' Documentation/ABI/ > doesn't show anything relating to the ux500. > > The major breakage would be out-of-tree stuff, if > any. > > Which is out-of-tree due to the absence of a SoC ID > framework. > > Which was nixed by Arnd et al precisely because > the above change was not part of it. > > It's a bit hard for Lee to please everyone with this > one I suspect :-D Ok, if you get guarantees from anyone using the in-kernel platform soc code that this move is acceptable, and you guarantee that no userspace tools will break, that's fine. But odds are, you are going to break something... greg k-h
2011/8/11 Greg KH <gregkh@suse.de>: > Ok, if you get guarantees from anyone using the in-kernel platform soc > code that this move is acceptable, and you guarantee that no userspace > tools will break, that's fine. But odds are, you are going to break > something... Yeah I know, but I strongly suspect that these userspace tools are created by the same organization that should've consulted me and the rest of the community about it in the first place, so I'll cope with it and write patches as I get flamed. Thanks, Linus Walleij
On Thursday 11 August 2011, Linus Walleij wrote: > 2011/8/11 Greg KH <gregkh@suse.de>: > > > Ok, if you get guarantees from anyone using the in-kernel platform soc > > code that this move is acceptable, and you guarantee that no userspace > > tools will break, that's fine. But odds are, you are going to break > > something... > > Yeah I know, but I strongly suspect that these userspace tools are > created by the same organization that should've consulted me and > the rest of the community about it in the first place, so I'll cope with > it and write patches as I get flamed. Right. Greg, note that the u8500 platform is still only partially supported upstream, the majority of the necessary device drivers for having a running system is not there and those drivers keep getting changed in order to be merged upstream. Further, with the move of the platform to device tree based probing, the position of all devices will change as well, to resemble the physical HW layout. Arnd
On Wednesday 10 August 2011, Lee Jones wrote: > At the request of Arnd Bergmann this patch moves all SoC > platform devices found in sysfs from /sys/devices/platform to > /sys/devices/soc/<SoCNum>/. It is believed as the devices are > SoC specific and a /sys/devices/soc node has recently become > available, that this would be a more appropriate place to > display the data. > > Signed-off-by: Lee Jones <lee.jones@linaro.org> > --- > arch/arm/mach-ux500/board-mop500.c | 19 +++++++++++++++---- > arch/arm/mach-ux500/board-u5500.c | 6 ++++++ > arch/arm/mach-ux500/cpu-db5500.c | 6 ++++++ > arch/arm/mach-ux500/cpu-db8500.c | 12 +++++++++++- > arch/arm/mach-ux500/devices-common.c | 19 +++++++++++++++---- > arch/arm/mach-ux500/usb.c | 4 ++++ > 6 files changed, 57 insertions(+), 9 deletions(-) > Ok. What does the device tree in /sys/devices look like after this? Does it match the flattened device tree description that you will have? Arnd
diff --git a/arch/arm/mach-ux500/board-mop500.c b/arch/arm/mach-ux500/board-mop500.c index 5b43fed..4c3295d 100644 --- a/arch/arm/mach-ux500/board-mop500.c +++ b/arch/arm/mach-ux500/board-mop500.c @@ -603,6 +603,7 @@ static struct platform_device *snowball_platform_devs[] __initdata = { static void __init mop500_init_machine(void) { int i2c0_devs; + int i; /* * The HREFv60 board removed a GPIO expander and routed @@ -620,12 +621,22 @@ static void __init mop500_init_machine(void) mop500_pins_init(); - if (machine_is_snowball()) + if (machine_is_snowball()) { + if (soc_parent->parent != NULL) + for (i=0; i<ARRAY_SIZE(snowball_platform_devs); i++) + snowball_platform_devs[i]->dev.parent = soc_parent; + platform_add_devices(snowball_platform_devs, - ARRAY_SIZE(snowball_platform_devs)); - else + ARRAY_SIZE(snowball_platform_devs)); + } + else { + if (soc_parent->parent != NULL) + for (i=0; i<ARRAY_SIZE(mop500_platform_devs); i++) + mop500_platform_devs[i]->dev.parent = soc_parent; + platform_add_devices(mop500_platform_devs, - ARRAY_SIZE(mop500_platform_devs)); + ARRAY_SIZE(mop500_platform_devs)); + } mop500_i2c_init(); mop500_sdi_init(); diff --git a/arch/arm/mach-ux500/board-u5500.c b/arch/arm/mach-ux500/board-u5500.c index 81e0475..4866436 100644 --- a/arch/arm/mach-ux500/board-u5500.c +++ b/arch/arm/mach-ux500/board-u5500.c @@ -53,11 +53,17 @@ static void __init u5500_uart_init(void) static void __init u5500_init_machine(void) { + int i; + u5500_init_devices(); u5500_sdi_init(); u5500_uart_init(); + if (soc_parent->parent != NULL) + for (i=0; i<ARRAY_SIZE(u5500_platform_devices); i++) + u5500_platform_devices[i]->dev.parent = soc_parent; + platform_add_devices(u5500_platform_devices, ARRAY_SIZE(u5500_platform_devices)); } diff --git a/arch/arm/mach-ux500/cpu-db5500.c b/arch/arm/mach-ux500/cpu-db5500.c index 7d940a8..52cbbd2 100644 --- a/arch/arm/mach-ux500/cpu-db5500.c +++ b/arch/arm/mach-ux500/cpu-db5500.c @@ -216,11 +216,17 @@ static int usb_db5500_tx_dma_cfg[] = { void __init u5500_init_devices(void) { + int i; + db5500_add_gpios(); db5500_dma_init(); db5500_add_rtc(); db5500_add_usb(usb_db5500_rx_dma_cfg, usb_db5500_tx_dma_cfg); + if (soc_parent->parent != NULL) + for (i=0; i<ARRAY_SIZE(db5500_platform_devs); i++) + db5500_platform_devs[i]->dev.parent = soc_parent; + platform_add_devices(db5500_platform_devs, ARRAY_SIZE(db5500_platform_devs)); } diff --git a/arch/arm/mach-ux500/cpu-db8500.c b/arch/arm/mach-ux500/cpu-db8500.c index 07562e8..cd63363 100644 --- a/arch/arm/mach-ux500/cpu-db8500.c +++ b/arch/arm/mach-ux500/cpu-db8500.c @@ -196,6 +196,8 @@ static int usb_db8500_tx_dma_cfg[] = { */ void __init u8500_init_devices(void) { + int i; + if (cpu_is_u8500ed()) dma40_u8500ed_fixup(); @@ -203,7 +205,15 @@ void __init u8500_init_devices(void) db8500_add_gpios(); db8500_add_usb(usb_db8500_rx_dma_cfg, usb_db8500_tx_dma_cfg); - platform_device_register_simple("cpufreq-u8500", -1, NULL, 0); + + platform_device_register_resndata( + (soc_parent->parent != NULL) ? soc_parent : NULL, + "cpufreq-u8500", -1, NULL, 0, NULL, 0); + + if (soc_parent->parent != NULL) + for (i=0; i<ARRAY_SIZE(platform_devs); i++) + platform_devs[i]->dev.parent = soc_parent; + platform_add_devices(platform_devs, ARRAY_SIZE(platform_devs)); return ; diff --git a/arch/arm/mach-ux500/devices-common.c b/arch/arm/mach-ux500/devices-common.c index 7418552..c8d1399 100644 --- a/arch/arm/mach-ux500/devices-common.c +++ b/arch/arm/mach-ux500/devices-common.c @@ -15,7 +15,7 @@ #include <linux/gpio/nomadik.h> #include <mach/hardware.h> - +#include <mach/setup.h> #include "devices-common.h" struct amba_device * @@ -45,6 +45,9 @@ dbx500_add_amba_device(const char *name, resource_size_t base, dev->dev.platform_data = pdata; + if (soc_parent->parent != NULL) + dev->dev.parent = soc_parent; + ret = amba_device_register(dev, &iomem_resource); if (ret) { kfree(dev); @@ -74,6 +77,9 @@ dbx500_add_platform_device(const char *name, int id, void *pdata, dev->dev.platform_data = pdata; + if (soc_parent->parent != NULL) + dev->dev.parent = soc_parent; + ret = platform_device_add(dev); if (ret) goto out_free; @@ -124,9 +130,14 @@ dbx500_add_gpio(int id, resource_size_t addr, int irq, } }; - return platform_device_register_resndata(NULL, "gpio", id, - resources, ARRAY_SIZE(resources), - pdata, sizeof(*pdata)); + return platform_device_register_resndata( + (soc_parent->parent != NULL) ? soc_parent : NULL, + "gpio", + id, + resources, + ARRAY_SIZE(resources), + pdata, + sizeof(*pdata)); } void dbx500_add_gpios(resource_size_t *base, int num, int irq, diff --git a/arch/arm/mach-ux500/usb.c b/arch/arm/mach-ux500/usb.c index 0a01cbd..29cdd3c 100644 --- a/arch/arm/mach-ux500/usb.c +++ b/arch/arm/mach-ux500/usb.c @@ -9,6 +9,7 @@ #include <linux/dma-mapping.h> #include <plat/ste_dma40.h> #include <mach/hardware.h> +#include <mach/setup.h> #include <mach/usb.h> #define MUSB_DMA40_RX_CH { \ @@ -157,5 +158,8 @@ void ux500_add_usb(resource_size_t base, int irq, int *dma_rx_cfg, ux500_usb_dma_update_rx_ch_config(dma_rx_cfg); ux500_usb_dma_update_tx_ch_config(dma_tx_cfg); + if (soc_parent->parent != NULL) + ux500_musb_device.dev.parent = soc_parent; + platform_device_register(&ux500_musb_device); }
At the request of Arnd Bergmann this patch moves all SoC platform devices found in sysfs from /sys/devices/platform to /sys/devices/soc/<SoCNum>/. It is believed as the devices are SoC specific and a /sys/devices/soc node has recently become available, that this would be a more appropriate place to display the data. Signed-off-by: Lee Jones <lee.jones@linaro.org> --- arch/arm/mach-ux500/board-mop500.c | 19 +++++++++++++++---- arch/arm/mach-ux500/board-u5500.c | 6 ++++++ arch/arm/mach-ux500/cpu-db5500.c | 6 ++++++ arch/arm/mach-ux500/cpu-db8500.c | 12 +++++++++++- arch/arm/mach-ux500/devices-common.c | 19 +++++++++++++++---- arch/arm/mach-ux500/usb.c | 4 ++++ 6 files changed, 57 insertions(+), 9 deletions(-)