Re: [PATCH v2] irqchip/gic-v3: workaround for ASR8601 when reading mpidr
From: Marc Zyngier
Date: Mon May 29 2023 - 11:02:23 EST
On Mon, 22 May 2023 12:06:43 +0100,
zhengyan <zhengyan@xxxxxxxxxxxx> wrote:
>
> This patch add workaround for ASR8601, which uses an armv8.2
> processor with a gic-500. But gic-500 is incompatible with
> ARMv8.2 implementations from ARM.
>
> ARMv8.2 from ARM implementation uses Multiprocessor Affinity
> Register to identify the logical address of the core by
> | cluster | core | thread |.
> However, gic-500 only supports topologies with
> affinity levels less than 2 as
> | cluster | core|.
>
> So we need this patch as workaround to shift the MPIDR values
> to ensure proper functionality
>
> Signed-off-by: zhengyan <zhengyan@xxxxxxxxxxxx>
> ---
> Documentation/arm64/silicon-errata.rst | 4 ++++
> drivers/irqchip/irq-gic-v3.c | 30 ++++++++++++++++++++++++++
> 2 files changed, 34 insertions(+)
>
> diff --git a/Documentation/arm64/silicon-errata.rst b/Documentation/arm64/silicon-errata.rst
> index 9e311bc43e05..d6430ade349d 100644
> --- a/Documentation/arm64/silicon-errata.rst
> +++ b/Documentation/arm64/silicon-errata.rst
> @@ -214,3 +214,7 @@ stable kernels.
> +----------------+-----------------+-----------------+-----------------------------+
> | Fujitsu | A64FX | E#010001 | FUJITSU_ERRATUM_010001 |
> +----------------+-----------------+-----------------+-----------------------------+
> +
> ++----------------+-----------------+-----------------+-----------------------------+
> +| ASR | ASR8601 | #8601001 | N/A |
> ++----------------+-----------------+-----------------+-----------------------------+
> diff --git a/drivers/irqchip/irq-gic-v3.c b/drivers/irqchip/irq-gic-v3.c
> index 6fcee221f201..cf64783dfe70 100644
> --- a/drivers/irqchip/irq-gic-v3.c
> +++ b/drivers/irqchip/irq-gic-v3.c
> @@ -39,6 +39,9 @@
>
> #define FLAGS_WORKAROUND_GICR_WAKER_MSM8996 (1ULL << 0)
> #define FLAGS_WORKAROUND_CAVIUM_ERRATUM_38539 (1ULL << 1)
> +#define FLAGS_WORKAROUND_ASR_ERRATUM_8601001 (1ULL << 2)
> +
> +#define ASR8601_AFF_QUIRK(aff) (aff >> 8)
This is wrong. There are more than just affinity bits in MPIDR, and
you're making a mess of the result.
>
> #define GIC_IRQ_TYPE_PARTITION (GIC_IRQ_TYPE_LPI + 1)
>
> @@ -659,6 +662,9 @@ static u64 gic_mpidr_to_affinity(unsigned long mpidr)
> {
> u64 aff;
>
> + if (gic_data.flags & FLAGS_WORKAROUND_ASR_ERRATUM_8601001)
> + mpidr = ASR8601_AFF_QUIRK(mpidr);
> +
> aff = ((u64)MPIDR_AFFINITY_LEVEL(mpidr, 3) << 32 |
> MPIDR_AFFINITY_LEVEL(mpidr, 2) << 16 |
> MPIDR_AFFINITY_LEVEL(mpidr, 1) << 8 |
> @@ -970,6 +976,9 @@ static int __gic_populate_rdist(struct redist_region *region, void __iomem *ptr)
> * Convert affinity to a 32bit value that can be matched to
> * GICR_TYPER bits [63:32].
> */
> + if (gic_data.flags & FLAGS_WORKAROUND_ASR_ERRATUM_8601001)
> + mpidr = ASR8601_AFF_QUIRK(mpidr);
It really wasn't what I had in mind when I asked you to place this
inside a helper. The whole thing looks horrible, and I really don't
want to have to maintain anything like it.
I came up with the following patch, which keeps the workaround *in a
single spot*.
Let me know if that works for you.
M.
diff --git a/drivers/irqchip/irq-gic-v3.c b/drivers/irqchip/irq-gic-v3.c
index 6fcee221f201..b7d69ef4da9f 100644
--- a/drivers/irqchip/irq-gic-v3.c
+++ b/drivers/irqchip/irq-gic-v3.c
@@ -39,6 +39,7 @@
#define FLAGS_WORKAROUND_GICR_WAKER_MSM8996 (1ULL << 0)
#define FLAGS_WORKAROUND_CAVIUM_ERRATUM_38539 (1ULL << 1)
+#define FLAGS_WORKAROUND_ASR_ERRATUM_8601001 (1ULL << 2)
#define GIC_IRQ_TYPE_PARTITION (GIC_IRQ_TYPE_LPI + 1)
@@ -655,10 +656,16 @@ static int gic_irq_set_vcpu_affinity(struct irq_data *d, void *vcpu)
return 0;
}
-static u64 gic_mpidr_to_affinity(unsigned long mpidr)
+static u64 gic_cpu_to_affinity(int cpu)
{
+ u64 mpidr = cpu_logical_map(cpu);
u64 aff;
+ /* ASR8601 needs to have its affinities shifted down... */
+ if (unlikely(gic_data.flags & FLAGS_WORKAROUND_ASR_ERRATUM_8601001))
+ mpidr = (MPIDR_AFFINITY_LEVEL(mpidr, 1) |
+ (MPIDR_AFFINITY_LEVEL(mpidr, 2) << 8));
+
aff = ((u64)MPIDR_AFFINITY_LEVEL(mpidr, 3) << 32 |
MPIDR_AFFINITY_LEVEL(mpidr, 2) << 16 |
MPIDR_AFFINITY_LEVEL(mpidr, 1) << 8 |
@@ -913,7 +920,7 @@ static void __init gic_dist_init(void)
* Set all global interrupts to the boot CPU only. ARE must be
* enabled.
*/
- affinity = gic_mpidr_to_affinity(cpu_logical_map(smp_processor_id()));
+ affinity = gic_cpu_to_affinity(smp_processor_id());
for (i = 32; i < GIC_LINE_NR; i++)
gic_write_irouter(affinity, base + GICD_IROUTER + i * 8);
@@ -962,7 +969,7 @@ static int gic_iterate_rdists(int (*fn)(struct redist_region *, void __iomem *))
static int __gic_populate_rdist(struct redist_region *region, void __iomem *ptr)
{
- unsigned long mpidr = cpu_logical_map(smp_processor_id());
+ unsigned long mpidr;
u64 typer;
u32 aff;
@@ -970,6 +977,8 @@ static int __gic_populate_rdist(struct redist_region *region, void __iomem *ptr)
* Convert affinity to a 32bit value that can be matched to
* GICR_TYPER bits [63:32].
*/
+ mpidr = gic_cpu_to_affinity(smp_processor_id());
+
aff = (MPIDR_AFFINITY_LEVEL(mpidr, 3) << 24 |
MPIDR_AFFINITY_LEVEL(mpidr, 2) << 16 |
MPIDR_AFFINITY_LEVEL(mpidr, 1) << 8 |
@@ -1262,9 +1271,11 @@ static u16 gic_compute_target_list(int *base_cpu, const struct cpumask *mask,
unsigned long cluster_id)
{
int next_cpu, cpu = *base_cpu;
- unsigned long mpidr = cpu_logical_map(cpu);
+ unsigned long mpidr;
u16 tlist = 0;
+ mpidr = gic_cpu_to_affinity(cpu);
+
while (cpu < nr_cpu_ids) {
tlist |= 1 << (mpidr & 0xf);
@@ -1273,8 +1284,7 @@ static u16 gic_compute_target_list(int *base_cpu, const struct cpumask *mask,
goto out;
cpu = next_cpu;
- mpidr = cpu_logical_map(cpu);
-
+ mpidr = gic_cpu_to_affinity(cpu);
if (cluster_id != MPIDR_TO_SGI_CLUSTER_ID(mpidr)) {
cpu--;
goto out;
@@ -1318,7 +1328,7 @@ static void gic_ipi_send_mask(struct irq_data *d, const struct cpumask *mask)
dsb(ishst);
for_each_cpu(cpu, mask) {
- u64 cluster_id = MPIDR_TO_SGI_CLUSTER_ID(cpu_logical_map(cpu));
+ u64 cluster_id = MPIDR_TO_SGI_CLUSTER_ID(gic_cpu_to_affinity(cpu));
u16 tlist;
tlist = gic_compute_target_list(&cpu, mask, cluster_id);
@@ -1376,7 +1386,7 @@ static int gic_set_affinity(struct irq_data *d, const struct cpumask *mask_val,
offset = convert_offset_index(d, GICD_IROUTER, &index);
reg = gic_dist_base(d) + offset + (index * 8);
- val = gic_mpidr_to_affinity(cpu_logical_map(cpu));
+ val = gic_cpu_to_affinity(cpu);
gic_write_irouter(val, reg);
@@ -1786,6 +1796,15 @@ static bool gic_enable_quirk_nvidia_t241(void *data)
return true;
}
+static bool gic_enable_quirk_asr8601(void *data)
+{
+ struct gic_chip_data *d = data;
+
+ d->flags |= FLAGS_WORKAROUND_ASR_ERRATUM_8601001;
+
+ return true;
+}
+
static const struct gic_quirk gic_quirks[] = {
{
.desc = "GICv3: Qualcomm MSM8996 broken firmware",
@@ -1823,6 +1842,11 @@ static const struct gic_quirk gic_quirks[] = {
.mask = 0xffffffff,
.init = gic_enable_quirk_nvidia_t241,
},
+ {
+ .desc = "GICv3: ASR erratum 8601001",
+ .compatible = "asr,asr8601-gic-v3",
+ .init = gic_enable_quirk_asr8601,
+ },
{
}
};
--
Without deviation from the norm, progress is not possible.