On 12/08/2017 04:46 PM, Ahmed Mansour wrote:
This patch adds changes necessary to move functionality present in PowerPC folders with ARM architectures that have DPAA1 QBMan hardware
- Create new board/freescale/common/fsl_portals.c to house shared device tree fixups for DPAA1 devices with ARM and PowerPC cores
- Add new header file to top includes directory to allow files in both architectures to grab the function prototypes
- Port inhibit_portals() from PowerPC to ARM. This function is used in setup to disable interrupts on all QMan and BMan portals. It is needed because the interrupts are enabled by default for all portals including unused/uninitialised portals. When the kernel attempts to go to deep sleep the unused portals prevent it from doing so
Signed-off-by: Ahmed Mansour ahmed.mansour@nxp.com
Changes in v2:
- Add get_qman_freq() to replace get_sys_info() for readability
- Correct the copyright year in new files
- Replace !ARM with PPC to wall off PowerPC SOCs specific qman setup
- Rename portals.c -> fsl_portals.c for clarity
arch/arm/cpu/armv8/fsl-layerscape/cpu.c | 4 + arch/arm/cpu/armv8/fsl-layerscape/fdt.c | 9 + .../arm/cpu/armv8/fsl-layerscape/fsl_lsch2_speed.c | 11 + .../include/asm/arch-fsl-layerscape/immap_lsch2.h | 29 ++ arch/arm/include/asm/arch-fsl-layerscape/speed.h | 1 + arch/powerpc/cpu/mpc85xx/cpu_init.c | 3 +- arch/powerpc/cpu/mpc85xx/fdt.c | 1 + arch/powerpc/cpu/mpc85xx/portals.c | 281 ------------------- arch/powerpc/include/asm/fsl_liodn.h | 7 +- arch/powerpc/include/asm/fsl_portals.h | 4 - arch/powerpc/include/asm/immap_85xx.h | 60 ---- drivers/misc/Makefile | 1 + drivers/misc/fsl_portals.c | 304 +++++++++++++++++++++ include/configs/ls1043a_common.h | 2 + include/fsl_qbman.h | 75 +++++ 15 files changed, 444 insertions(+), 348 deletions(-) create mode 100644 drivers/misc/fsl_portals.c create mode 100644 include/fsl_qbman.h
diff --git a/arch/arm/cpu/armv8/fsl-layerscape/cpu.c b/arch/arm/cpu/armv8/fsl-layerscape/cpu.c index d082629..3fd352f 100644 --- a/arch/arm/cpu/armv8/fsl-layerscape/cpu.c +++ b/arch/arm/cpu/armv8/fsl-layerscape/cpu.c @@ -30,6 +30,7 @@ #endif #include <asm/arch/clock.h> #include <hwconfig.h> +#include <fsl_qbman.h>
DECLARE_GLOBAL_DATA_PTR;
@@ -565,6 +566,9 @@ int arch_early_init_r(void) #ifdef CONFIG_FMAN_ENET fman_enet_init(); #endif +#ifdef CONFIG_SYS_DPAA_QBMAN
- setup_qbman_portals();
+#endif return 0; }
diff --git a/arch/arm/cpu/armv8/fsl-layerscape/fdt.c b/arch/arm/cpu/armv8/fsl-layerscape/fdt.c index cae59da..382bf3d 100644 --- a/arch/arm/cpu/armv8/fsl-layerscape/fdt.c +++ b/arch/arm/cpu/armv8/fsl-layerscape/fdt.c @@ -26,6 +26,8 @@ #ifdef CONFIG_ARMV8_SEC_FIRMWARE_SUPPORT #include <asm/armv8/sec_firmware.h> #endif +#include <asm/arch/speed.h> +#include <fsl_qbman.h>
int fdt_fixup_phy_connection(void *blob, int offset, phy_interface_t phyc) { @@ -415,6 +417,13 @@ void ft_cpu_setup(void *blob, bd_t *bd) fdt_fixup_esdhc(blob, bd); #endif
+#ifdef CONFIG_SYS_DPAA_QBMAN
- fdt_fixup_bportals(blob);
- fdt_fixup_qportals(blob);
- do_fixup_by_compat_u32(blob, "fsl,qman",
"clock-frequency", get_qman_freq(), 1);+#endif
#ifdef CONFIG_SYS_DPAA_FMAN fdt_fixup_fman_firmware(blob); #endif diff --git a/arch/arm/cpu/armv8/fsl-layerscape/fsl_lsch2_speed.c b/arch/arm/cpu/armv8/fsl-layerscape/fsl_lsch2_speed.c index 2d7775e..f707205 100644 --- a/arch/arm/cpu/armv8/fsl-layerscape/fsl_lsch2_speed.c +++ b/arch/arm/cpu/armv8/fsl-layerscape/fsl_lsch2_speed.c @@ -155,6 +155,17 @@ void get_sys_info(struct sys_info *sys_info) sys_info->freq_localbus = sys_info->freq_systembus / CONFIG_SYS_FSL_IFC_CLK_DIV; #endif +#ifdef CONFIG_SYS_DPAA_QBMAN
- sys_info->freq_qman = sys_info->freq_systembus;
+#endif +}
+unsigned long get_qman_freq(void) +{
- struct sys_info sys_info;
- get_sys_info(&sys_info);
nitpick Insert a blank line before return.
- return sys_info.freq_systembus;
}
If you think this clock will be used multiple times, you can consider to use the same way sdhc_clk is handled. For this patch, I only see one place it is called.
<snip>
diff --git a/drivers/misc/fsl_portals.c b/drivers/misc/fsl_portals.c new file mode 100644 index 0000000..8f8503f --- /dev/null +++ b/drivers/misc/fsl_portals.c @@ -0,0 +1,304 @@ +/*
- Copyright 2017 NXP
Shouldn't we carry the old license year and copyright for Freescale because this file is created by moving those code here? I am not a lawyer though.
Let's take a look from a different angle. I know you use patman to generate the patches. Sometimes we need to think what the best way is. Patman is a good tool. But I don't think it turns on the switch "--find-copies". This switch can detect copies and renames. As you stated in the change log, you renamed portals.c to fsl_portals.c. Your patch didn't show this. You can manually run "git format-patch -M -C" to generate the patches. For this case, the "-C" switch makes a difference. It detects the renaming so the patch has less insertions, and easier to review.
15 files changed, 198 insertions(+), 458 deletions(-)
York