Skip to contentSkip to conversation

xen/arm: Initial Broadcom A15 platform support

PATCH In review Jon Fraser 3 participants 7 messages Nov 8, 2013 – Nov 22, 2013

lore
Share & tools
lore

Jon Fraser

author
#0 lore raw

PATCH 1/1 xen/arm: Initial Broadcom A15 platform support

Initial platform support for Broadcom A15/B15 platforms.

---
 xen/arch/arm/platforms/Makefile |   1 +
 xen/arch/arm/platforms/brcm.c   | 249 ++++++++++++++++++++++++++++++++++++++++
 2 files changed, 250 insertions(+)
 create mode 100644 xen/arch/arm/platforms/brcm.c
xen/arch/arm/platforms/Makefile +1 -0
index f0dd72c..60635b4 100644
@@ -3,3 +3,4 @@ obj-$(CONFIG_ARM_32) += exynos5.o
 obj-$(CONFIG_ARM_32) += midway.o
 obj-$(CONFIG_ARM_32) += omap5.o
 obj-$(CONFIG_ARM_32) += sunxi.o
+obj-$(CONFIG_ARM_32) += brcm.o
xen/arch/arm/platforms/brcm.c new file +249 -0

249 lines not on this page · raw message

Signature · 1 line
1.7.11.3
#1 lore raw
Quoted 3 lines · Jon Fraser

Initial platform support for Broadcom A15/B15 platforms.

Hello,

Thanks to add support for broadcom processor in Xen.

It seems you use the wrong coding style for this patch. You can read CODING_STYLES at the root of the xen repository.

Quoted 43 lines
---
 xen/arch/arm/platforms/Makefile |   1 +
 xen/arch/arm/platforms/brcm.c   | 249 ++++++++++++++++++++++++++++++++++++++++
 2 files changed, 250 insertions(+)
 create mode 100644 xen/arch/arm/platforms/brcm.c
xen/arch/arm/platforms/Makefile +1 -0
index f0dd72c..60635b4 100644
@@ -3,3 +3,4 @@ obj-$(CONFIG_ARM_32) += exynos5.o
 obj-$(CONFIG_ARM_32) += midway.o
 obj-$(CONFIG_ARM_32) += omap5.o
 obj-$(CONFIG_ARM_32) += sunxi.o
+obj-$(CONFIG_ARM_32) += brcm.o
xen/arch/arm/platforms/brcm.c new file +22 -0
new file mode 100644
index 0000000..b8d8527
@@ -0,0 +1,249 @@
+/*
+ * xen/arch/arm/platform/brcm.c
+ *
+ * Broadcom specific settings
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+#include <asm/platform.h>
+#include <xen/mm.h>
+#include <xen/vmap.h>
+
+
+struct brcm_plat_regs {

xen/arch/arm/platforms/brcm.c line 32

Quoted 10 lines
xen/arch/arm/platforms/brcm.c new file +10 -0
+    uint32_t    hif_mask;
+    uint32_t    hif_cpu_reset_config;
+    uint32_t    hif_boot_continuation;
+    uint32_t    cpu0_pwr_zone_ctrl;
+};
+
+static struct brcm_plat_regs regs;
+
+static int brcm_get_dt_node(char *compat_str, struct dt_device_node **dn,
+                             u32 *reg_base)

spurious space before u32.

xen/arch/arm/platforms/brcm.c line 40

Quoted 8 lines
xen/arch/arm/platforms/brcm.c new file +8 -0
+{
+    struct dt_device_node *dt_dn;
+    u64 reg_base_64;
+    u64 size;
+    int rc;
+
+    dt_dn = dt_find_compatible_node(NULL, NULL, compat_str);
+    if (!dt_dn) {

if ( ... )
{

xen/arch/arm/platforms/brcm.c line 51

Quoted 11 lines
xen/arch/arm/platforms/brcm.c new file +11 -0
+        dprintk(XENLOG_ERR, "%s: missing \"%s\" node\n", __func__, compat_str);
+        return(-EIO);
+    }
+
+    rc = dt_device_get_address(dt_dn, 0, &reg_base_64, &size);
+    if (rc) {
+        dprintk(XENLOG_ERR, "%s: missing \"reg\" prop\n", __func__);
+        return(rc);
+    }
+
+    if (dn)

if ( ... )

xen/arch/arm/platforms/brcm.c line 55

Quoted 4 lines
xen/arch/arm/platforms/brcm.c new file +4 -0
+        *dn = dt_dn;
+
+    if (reg_base)
+        *reg_base = (u32)(reg_base_64 & 0xffffffff);

I think you don't need to use both cast and mask. You can only use one of theses solutions.

xen/arch/arm/platforms/brcm.c line 63

Quoted 8 lines
xen/arch/arm/platforms/brcm.c new file +8 -0
+
+    return(0);
+}
+
+static int brcm_populate_plat_regs(void)
+{
+    int rc;
+    struct dt_device_node *dt_dn;

As you don't modify the node, you should use const.

xen/arch/arm/platforms/brcm.c line 66

Quoted 3 lines
xen/arch/arm/platforms/brcm.c new file +3 -0
+    u32 reg_base;
+    u32 val;
+    char *compat_str;

Same here.

xen/arch/arm/platforms/brcm.c line 73

Quoted 7 lines
xen/arch/arm/platforms/brcm.c new file +7 -0
+
+    compat_str = "brcm,brcmstb-cpu-biu-ctrl";
+    rc = brcm_get_dt_node(compat_str, &dt_dn, &reg_base);
+    if (rc)
+        return(rc);
+
+    dt_property_read_u32(dt_dn, "cpu-reset-config-reg", &val);

You should check the return of dt_property_read_u32

xen/arch/arm/platforms/brcm.c line 76

Quoted 3 lines
xen/arch/arm/platforms/brcm.c new file +3 -0
+    regs.hif_cpu_reset_config = reg_base + val;
+
+    dt_property_read_u32(dt_dn, "cpu0-pwr-zone-ctrl-reg", &val);

Same here.

xen/arch/arm/platforms/brcm.c line 84

Quoted 8 lines
xen/arch/arm/platforms/brcm.c new file +8 -0
+    regs.cpu0_pwr_zone_ctrl = reg_base + val;
+
+    compat_str = "brcm,brcmstb-hif-continuation";
+    rc = brcm_get_dt_node(compat_str, &dt_dn, &reg_base);
+    if (rc)
+        return(rc);
+
+    dt_property_read_u32(dt_dn, "stb-boot-hi-addr0-reg", &val);

Same here.

Quoted 17 lines
xen/arch/arm/platforms/brcm.c new file +17 -0
+    regs.hif_boot_continuation = reg_base + val;
+
+    dprintk(XENLOG_INFO, "hif_cpu_reset_config  : %08xh\n",
+                    regs.hif_cpu_reset_config);
+    dprintk(XENLOG_INFO, "cpu0_pwr_zone_ctrl    : %08xh\n",
+                    regs.cpu0_pwr_zone_ctrl);
+    dprintk(XENLOG_INFO, "hif_boot_continuation : %08xh\n",
+                    regs.hif_boot_continuation);
+
+    return(0);
+}
+
+#define ZONE_PWR_UP_REQ   (1 << 10)
+#define ZONE_PWR_ON_STATE (1 << 26)
+
+static int brcm_cpu_power_on(void __iomem *va, int cpu)
+{

xen/arch/arm/platforms/brcm.c line 111

Quoted 10 lines
xen/arch/arm/platforms/brcm.c new file +10 -0
+    int rc = 0;
+    volatile u32 *reg;
+    u32 tmp;
+
+    reg = (volatile u32 *)(va + (regs.cpu0_pwr_zone_ctrl & ~PAGE_MASK) +
+            (cpu * sizeof(u32)));
+
+    /* request core power on */
+    *reg = ZONE_PWR_UP_REQ;
+    dsb();

You should use writel/readl wrapper to read/write memory mapped with ioremap_*. It will take care of dsb.

xen/arch/arm/platforms/brcm.c line 115

Quoted 4 lines
xen/arch/arm/platforms/brcm.c new file +4 -0
+
+    /* wait for power to be applied to cire */
+    do {
+        tmp = *reg;

Same here.

xen/arch/arm/platforms/brcm.c line 127

Quoted 12 lines
xen/arch/arm/platforms/brcm.c new file +12 -0
+    } while (!(tmp & ZONE_PWR_ON_STATE));
+
+    return rc;
+}
+
+static int brcm_cpu_power_on_all(void)
+{
+    int rc = 0;
+    int cpu;
+    void __iomem *va;
+
+    va = ioremap_nocache(regs.cpu0_pwr_zone_ctrl & PAGE_MASK, PAGE_SIZE);

ioremap_* is able to cooperate with non aligned address. You don't need to align cpu0_pwr_zone_ctrl. You should also check that va is not NULL (it's an error).

xen/arch/arm/platforms/brcm.c line 133

Quoted 6 lines
xen/arch/arm/platforms/brcm.c new file +6 -0
+
+    for (cpu = 1; cpu < 4; cpu++) {
+        rc = brcm_cpu_power_on(va, cpu);
+        if (rc)
+            break;
+    }

I don't like the for loop on the cores. In the future, it's possible that Xen decides to only use 2 cores, so you will power unused core. What about enabling power in brcm_cpu_up, if it's not too late?

Quoted 6 lines
xen/arch/arm/platforms/brcm.c new file +6 -0
+
+    iounmap(va);
+
+    return rc;
+}
+

xen/arch/arm/platforms/brcm.c line 149

Quoted 10 lines
xen/arch/arm/platforms/brcm.c new file +10 -0
+static int __init brcm_smp_init(void)
+{
+    void __iomem *va;
+    volatile u32 *reg;
+    u32 target_pc;
+    dprintk(XENLOG_INFO, "%s\n",__func__);
+
+    if (regs.cpu0_pwr_zone_ctrl) {
+        dprintk(XENLOG_INFO, "applying power to remaining CPU cores\n");
+        brcm_cpu_power_on_all();

You already call brcm_cpu_power_on_all in brcm_init. Why do you need to call it again?

Quoted 6 lines
xen/arch/arm/platforms/brcm.c new file +6 -0
+    }
+
+    va = ioremap_nocache(regs.hif_boot_continuation & PAGE_MASK, PAGE_SIZE);
+    reg = (volatile u32 *)(va + (regs.hif_boot_continuation & ~PAGE_MASK));
+
+

xen/arch/arm/platforms/brcm.c line 165

Quoted 10 lines
xen/arch/arm/platforms/brcm.c new file +10 -0
+    /* set boot continuation registers */
+    dprintk(XENLOG_INFO, "%s: HIF CR at 0x%p\n", __func__, reg);
+    target_pc =__pa(init_secondary);
+    reg[3] = target_pc;
+    reg[5] = target_pc;
+    reg[7] = target_pc;
+
+    iounmap(va);
+
+    va = ioremap_nocache(regs.hif_cpu_reset_config & PAGE_MASK, PAGE_SIZE);

Some comment as the previous ioremap_*

xen/arch/arm/platforms/brcm.c line 169

Quoted 5 lines
xen/arch/arm/platforms/brcm.c new file +4 -0
+    reg = (volatile u32 *)(va + (regs.hif_cpu_reset_config & ~PAGE_MASK));
 
+
+    /* now take the cpus out of reset */
+    reg[0] = 0;

writel?

Quoted 46 lines
xen/arch/arm/platforms/brcm.c new file +46 -0
+
+    iounmap(va);
+
+    return 0;
+}
+
+
+static int brcm_init(void)
+{
+    int rc;
+
+    rc = brcm_populate_plat_regs();
+    if (rc)
+        return(rc);
+
+    /*
+     * All newer 28nm chips will require poking the BPCM prior to starting
+     * the cores.
+     */
+    if (regs.cpu0_pwr_zone_ctrl) {
+        dprintk(XENLOG_INFO, "applying power to remaining CPU cores\n");
+        brcm_cpu_power_on_all();
+    }
+
+    return(0);
+}
+static int __init brcm_cpu_up(int cpu)
+{
+    /*
+     * Nothing to do here, the generic sev() will suffice to kick CPUs
+     * out of either the firmware or our own smp_up_cpu gate,
+     * depending on where they have ended up.
+     */
+
+    return 0;
+}
+
+static int brcm_specific_mapping(struct domain *d)
+{
+    /* Map the entire GISB register space 1:1 */
+    map_mmio_regions(d, 0xf0000000, 0xf0ffffff,
+                     0xf0000000);
+
+    return 0;
+}
+

xen/arch/arm/platforms/brcm.c line 225

Quoted 10 lines
xen/arch/arm/platforms/brcm.c new file +10 -0
+static void brcm_reset(void)
+{
+}
+
+static uint32_t brcm_quirks(void)
+{
+    return PLATFORM_QUIRK_DOM0_MAPPING_11;
+}
+
+static const char const *brcm_dt_compat[] __initdata =

const char * const ... __initconst

Quoted 25 lines
xen/arch/arm/platforms/brcm.c new file +24 -0
+{
+    "arm,brcm",
+    "brcm,brcmstb",
+    NULL
+};
+
+PLATFORM_START(brcm, "Broadcom B15")
+    .compatible     = brcm_dt_compat,
+    .init           = brcm_init,
+    .smp_init       = brcm_smp_init,
+    .cpu_up         = brcm_cpu_up,
+    .quirks         = brcm_quirks,
+    .reset          = brcm_reset,
+    .specific_mapping    = brcm_specific_mapping,
+PLATFORM_END
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
Signature · 1 line
Julien Grall
#2 lore raw
Quoted 10 lines · Julien Grall

xen/arch/arm/platforms/brcm.c line 133

Quoted 6 lines
xen/arch/arm/platforms/brcm.c new file +6 -0
+
+    for (cpu = 1; cpu < 4; cpu++) {
+        rc = brcm_cpu_power_on(va, cpu);
+        if (rc)
+            break;
+    }

I don't like the for loop on the cores. In the future, it's possible that Xen decides to only use 2 cores, so you will power unused core. What about enabling power in brcm_cpu_up, if it's not too late?

Not only in the future, we have command line options for these sorts of things already, which I think the ARM port obeys.

Ian.

#3 lore raw
Quoted 8 lines · Ian Campbell
Quoted 7 lines · Julien Grall
Quoted 6 lines
xen/arch/arm/platforms/brcm.c new file +6 -0
+
+    for (cpu = 1; cpu < 4; cpu++) {
+        rc = brcm_cpu_power_on(va, cpu);
+        if (rc)
+            break;
+    }
Quoted 6 lines
Quoted 3 lines

I don't like the for loop on the cores. In the future, it's possible that Xen decides to only use 2 cores, so you will power unused core. What about enabling power in brcm_cpu_up, if it's not too late?

Not only in the future, we have command line options for these sorts of things already, which I think the ARM port obeys.

Are you talking about "max_cpus" parameter? If so, it's not yet implemented on ARM. We rely on the number of cpus node in the device tree.

Signature · 1 line
Julien Grall
#4 lore raw
Quoted 10 lines · Julien Grall
Quoted 8 lines · Ian Campbell
Quoted 7 lines · Julien Grall
Quoted 6 lines
xen/arch/arm/platforms/brcm.c new file +6 -0
+
+    for (cpu = 1; cpu < 4; cpu++) {
+        rc = brcm_cpu_power_on(va, cpu);
+        if (rc)
+            break;
+    }
Quoted 8 lines
Quoted 6 lines
Quoted 3 lines

I don't like the for loop on the cores. In the future, it's possible that Xen decides to only use 2 cores, so you will power unused core. What about enabling power in brcm_cpu_up, if it's not too late?

Not only in the future, we have command line options for these sorts of things already, which I think the ARM port obeys.

Are you talking about "max_cpus" parameter?

Yes.

Quoted 2 lines

If so, it's not yet
implemented on ARM. We rely on the number of cpus node in the device tree.

I guess I should have checked...

In any case you are right that hardcoding 4 here is not correct and it would be better to do it in one of the provided per-cpu callbacks.

Ian.

author
#5 lore raw
In reply to Julien Grall
Quoted 11 lines · Julien Grall
Quoted 3 lines · Jon Fraser

Initial platform support for Broadcom A15/B15 platforms.

Hello,

Thanks to add support for broadcom processor in Xen.

It seems you use the wrong coding style for this patch. You can read CODING_STYLES at the root of the xen repository.

Habits are hard to break!

Is there a xen equivalent of the linux checkpatch?
Or a set of command line options for indent that is acceptable to the xen community? Switching back and forth between xen and linux kernel, I don't want to keep making formatting mistakes. Well, I know I'll make them, I just want to catch them.

I'll address the rest of the issues and submit a new version as soon as I can.

Thanks,
Jon

#6 lore raw
Quoted 10 lines · Jon Fraser
Quoted 9 lines · Julien Grall
Quoted 3 lines · Jon Fraser

Initial platform support for Broadcom A15/B15 platforms.

Hello,

Thanks to add support for broadcom processor in Xen.

Quoted 7 lines
Quoted 2 lines

It seems you use the wrong coding style for this patch. You can read CODING_STYLES at the root of the xen repository.

Habits are hard to break!

Is there a xen equivalent of the linux checkpatch?

I'm afraid not.

Quoted 2 lines

Or a set of command line options for indent that is acceptable to the xen community?

The closest is the emacs magic blocks at the bottom of most files, but they only really cover the 4-space indentation and not all of the whitespace differences etc.

The bigs things are:
4 space indents
Spaces inside the if expre. eg. "if ( foo == bar )" not "if (foo == bar)"
The { comes on the next line not at the end so:

        if ( foo == bar )
        {
and not
        if ( foo == bar ) {

(similarly for while, for etc)

Quoted 2 lines

Switching back and forth between xen and linux
kernel,

Yeah, I do it all the time as well.

Quoted 11 lines

I don't want to keep making formatting mistakes. Well, I know I'll make them, I just want to catch them.

I'll address the rest of the issues and submit a new version as soon as I can.

Thanks,
Jon