@@ -96,6 +96,7 @@ struct uniphier_cache_data {
void __iomem *ctrl_base;
void __iomem *rev_base;
void __iomem *op_base;
+ void __iomem *way_ctrl_base;
u32 way_present_mask;
u32 way_locked_mask;
u32 nsets;
@@ -256,10 +257,13 @@ static void __init __uniphier_cache_set_locked_ways(
struct uniphier_cache_data *data,
u32 way_mask)
{
+ unsigned int cpu;
+
data->way_locked_mask = way_mask & data->way_present_mask;
- writel_relaxed(~data->way_locked_mask & data->way_present_mask,
- data->ctrl_base + UNIPHIER_SSCLPDAWCR);
+ for_each_possible_cpu(cpu)
+ writel_relaxed(~data->way_locked_mask & data->way_present_mask,
+ data->way_ctrl_base + 4 * cpu);
}
static void uniphier_cache_maint_range(unsigned long start, unsigned long end,
@@ -459,6 +463,8 @@ static int __init __uniphier_cache_init(struct device_node *np,
goto err;
}
+ data->way_ctrl_base = data->ctrl_base + 0xc00;
+
if (*cache_level == 2) {
u32 revision = readl(data->rev_base + UNIPHIER_SSCID);
/*
@@ -467,6 +473,22 @@ static int __init __uniphier_cache_init(struct device_node *np,
*/
if (revision <= 0x16)
data->range_op_max_size = (u32)1 << 22;
+
+ /*
+ * Unfortunatly, the offset address of active way control base
+ * varies from SoC to SoC.
+ */
+ switch (revision) {
+ case 0x11: /* sLD3 */
+ data->way_ctrl_base = data->ctrl_base + 0x870;
+ break;
+ case 0x12: /* LD4 */
+ case 0x16: /* sld8 */
+ data->way_ctrl_base = data->ctrl_base + 0x840;
+ break;
+ default:
+ break;
+ }
}
data->range_op_max_size -= data->line_size;
This outer cache allows to control active ways independently for each CPU, but currently nothing is done for secondary CPUs. In other words, all the ways are locked for secondary CPUs by default. This commit fixes it to fully bring out the performance of this outer cache. There would be two possible ways to achieve this: [1] Each CPU initializes active ways for itself. This can be done via the SSCLPDAWCR register. This is a banked register, so each CPU sees a different instance of the register for its own. [2] The master CPU initializes active ways for all the CPUs. This is available via SSCDAWCARMR(N) registers, where all instances of SSCLPDAWCR are mirrored. They are mapped at the address SSCDAWCARMR + 4 * N, where N is the CPU number. The outer cache frame work does not support a per-CPU init callback. So this commit adopts [2]; the master CPU iterates over possible CPUs setting up SSCDAWCARMR(N) registers. Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com> --- KernelVersion: 4.6-rc4 arch/arm/mm/cache-uniphier.c | 26 ++++++++++++++++++++++++-- 1 file changed, 24 insertions(+), 2 deletions(-)