| Message ID | 1385001613-19098-1-git-send-email-acourbot@nvidia.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 11/20/2013 07:40 PM, Alexandre Courbot wrote: > FUSE clock is enabled by most bootloaders, but we cannot expect it to be > on in all contexts (e.g. kexec). > > Ensure the FUSE clock is enabled before any of its registers is touched. > Since FUSE is touched very early during system boot (before the clock > devices are registered), directly manipulate the clock register bit in > case the clock device cannot be acquired. This looks reasonable to me. I'll apply it soon after -rc1.
On 11/22/2013 05:30 AM, Stephen Warren wrote: > On 11/20/2013 07:40 PM, Alexandre Courbot wrote: >> FUSE clock is enabled by most bootloaders, but we cannot expect it to be >> on in all contexts (e.g. kexec). >> >> Ensure the FUSE clock is enabled before any of its registers is touched. >> Since FUSE is touched very early during system boot (before the clock >> devices are registered), directly manipulate the clock register bit in >> case the clock device cannot be acquired. > > This looks reasonable to me. I'll apply it soon after -rc1. Thanks. Be careful as I noticed I misformatted my commit message. The part after the "--" will not be stripped by git am as I intended it to be. I understood what I did wrong and will hopefully not make that mistake again. Sorry for the inconvenience.
Hi Stephen, On Fri, Nov 22, 2013 at 10:35 AM, Alex Courbot <acourbot@nvidia.com> wrote: > On 11/22/2013 05:30 AM, Stephen Warren wrote: >> >> On 11/20/2013 07:40 PM, Alexandre Courbot wrote: >>> >>> FUSE clock is enabled by most bootloaders, but we cannot expect it to be >>> on in all contexts (e.g. kexec). >>> >>> Ensure the FUSE clock is enabled before any of its registers is touched. >>> Since FUSE is touched very early during system boot (before the clock >>> devices are registered), directly manipulate the clock register bit in >>> case the clock device cannot be acquired. >> >> >> This looks reasonable to me. I'll apply it soon after -rc1. > > > Thanks. Be careful as I noticed I misformatted my commit message. The part > after the "--" will not be stripped by git am as I intended it to be. I > understood what I did wrong and will hopefully not make that mistake again. > > Sorry for the inconvenience. I am not seeing this in your tree, have you applied it somewhere already?
On 12/04/2013 12:09 AM, Alexandre Courbot wrote: > Hi Stephen, > > On Fri, Nov 22, 2013 at 10:35 AM, Alex Courbot <acourbot@nvidia.com> wrote: >> On 11/22/2013 05:30 AM, Stephen Warren wrote: >>> >>> On 11/20/2013 07:40 PM, Alexandre Courbot wrote: >>>> >>>> FUSE clock is enabled by most bootloaders, but we cannot expect it to be >>>> on in all contexts (e.g. kexec). >>>> >>>> Ensure the FUSE clock is enabled before any of its registers is touched. >>>> Since FUSE is touched very early during system boot (before the clock >>>> devices are registered), directly manipulate the clock register bit in >>>> case the clock device cannot be acquired. >>> >>> >>> This looks reasonable to me. I'll apply it soon after -rc1. >> >> >> Thanks. Be careful as I noticed I misformatted my commit message. The part >> after the "--" will not be stripped by git am as I intended it to be. I >> understood what I did wrong and will hopefully not make that mistake again. >> >> Sorry for the inconvenience. > > I am not seeing this in your tree, have you applied it somewhere already? I haven't applied anything for 3.14 yet. I was hoping to get the DMA/clock/reset DT binding rework in first, but it's been dragging on for while, so I'll probably take a look at what can be applied without conflicting with that soon.
On 11/20/2013 07:40 PM, Alexandre Courbot wrote: > FUSE clock is enabled by most bootloaders, but we cannot expect it to be > on in all contexts (e.g. kexec). > > Ensure the FUSE clock is enabled before any of its registers is touched. > Since FUSE is touched very early during system boot (before the clock > devices are registered), directly manipulate the clock register bit in > case the clock device cannot be acquired. I've applied this to Tegra's for-3.14/soc branch.
diff --git a/arch/arm/mach-tegra/fuse.c b/arch/arm/mach-tegra/fuse.c index 9a4e910c3796..2adc2ab081e2 100644 --- a/arch/arm/mach-tegra/fuse.c +++ b/arch/arm/mach-tegra/fuse.c @@ -22,6 +22,7 @@ #include <linux/io.h> #include <linux/export.h> #include <linux/random.h> +#include <linux/clk.h> #include <linux/tegra-soc.h> #include "fuse.h" @@ -54,6 +55,7 @@ int tegra_cpu_speedo_id; /* only exist in Tegra30 and later */ int tegra_soc_speedo_id; enum tegra_revision tegra_revision; +static struct clk *fuse_clk; static int tegra_fuse_spare_bit; static void (*tegra_init_speedo_data)(void); @@ -77,6 +79,22 @@ static const char *tegra_revision_name[TEGRA_REVISION_MAX] = { [TEGRA_REVISION_A04] = "A04", }; +static void tegra_fuse_enable_clk(void) +{ + if (IS_ERR(fuse_clk)) + fuse_clk = clk_get_sys(NULL, "fuse"); + if (IS_ERR(fuse_clk)) + return; + clk_prepare_enable(fuse_clk); +} + +static void tegra_fuse_disable_clk(void) +{ + if (IS_ERR(fuse_clk)) + return; + clk_disable_unprepare(fuse_clk); +} + u32 tegra_fuse_readl(unsigned long offset) { return tegra_apb_readl(TEGRA_FUSE_BASE + offset); @@ -84,7 +102,15 @@ u32 tegra_fuse_readl(unsigned long offset) bool tegra_spare_fuse(int bit) { - return tegra_fuse_readl(tegra_fuse_spare_bit + bit * 4); + bool ret; + + tegra_fuse_enable_clk(); + + ret = tegra_fuse_readl(tegra_fuse_spare_bit + bit * 4); + + tegra_fuse_disable_clk(); + + return ret; } static enum tegra_revision tegra_get_revision(u32 id) @@ -113,10 +139,14 @@ static void tegra_get_process_id(void) { u32 reg; + tegra_fuse_enable_clk(); + reg = tegra_fuse_readl(tegra_fuse_spare_bit); tegra_cpu_process_id = (reg >> 6) & 3; reg = tegra_fuse_readl(tegra_fuse_spare_bit); tegra_core_process_id = (reg >> 12) & 3; + + tegra_fuse_disable_clk(); } u32 tegra_read_chipid(void) @@ -159,6 +189,15 @@ void __init tegra_init_fuse(void) reg |= 1 << 28; writel(reg, IO_ADDRESS(TEGRA_CLK_RESET_BASE + 0x48)); + /* + * Enable FUSE clock. This needs to be hardcoded because the clock + * subsystem is not active during early boot. + */ + reg = readl(IO_ADDRESS(TEGRA_CLK_RESET_BASE + 0x14)); + reg |= 1 << 7; + writel(reg, IO_ADDRESS(TEGRA_CLK_RESET_BASE + 0x14)); + fuse_clk = ERR_PTR(-EINVAL); + reg = tegra_fuse_readl(FUSE_SKU_INFO); randomness[0] = reg; tegra_sku_id = reg & 0xFF;
FUSE clock is enabled by most bootloaders, but we cannot expect it to be on in all contexts (e.g. kexec). Ensure the FUSE clock is enabled before any of its registers is touched. Since FUSE is touched very early during system boot (before the clock devices are registered), directly manipulate the clock register bit in case the clock device cannot be acquired. Signed-off-by: Alexandre Courbot <acourbot@nvidia.com> -- Changelog since v1: - Moved clock device definition into another patch - Kept the initial idea of enabling the fuse clock at init and keeping it that way until the clock framework takes it over. The reason is that exported functions might be called from tegra_init_fuse() that acquire the clock (e.g. tegra_get_revision()). Since we don't have any enable counter here, the clock would be disabled earlier than we want if we were to allow disabling of the clock before the clock framework is available. --- arch/arm/mach-tegra/fuse.c | 41 ++++++++++++++++++++++++++++++++++++++++- 1 file changed, 40 insertions(+), 1 deletion(-)