Add full cache description to DTS to avoid:
1. "Early cacheinfo failed" warnings,
2. Cache topology detection which leads to early memory allocations and
   "BUG: sleeping function called from invalid context" on PREEMPT_RT
   kernel:

  smp: Bringing up secondary CPUs ...
  Detected VIPT I-cache on CPU1
  BUG: sleeping function called from invalid context at kernel/locking/spinlock_rt.c:46
  in_atomic(): 1, irqs_disabled(): 128, non_block: 0, pid: 0, name: swapper/1
  preempt_count: 1, expected: 0
  RCU nest depth: 1, expected: 1
  3 locks held by swapper/1/0:
   #0: ffff5e337eee5f18 (&pcp->lock){+.+.}-{3:3}, at: get_page_from_freelist+0x20c/0xffc
   #1: ffffa9e24a900b18 (rcu_read_lock){....}-{1:3}, at: rt_spin_trylock+0x40/0xe4
   #2: ffff5e337efc8918 (&zone->lock){+.+.}-{3:3}, at: rmqueue_bulk+0x54/0x720
  irq event stamp: 0
  Call trace:
   __might_resched+0x17c/0x214
   rt_spin_lock+0x5c/0x100
   rmqueue_bulk+0x54/0x720
   get_page_from_freelist+0xcfc/0xffc
   __alloc_pages+0xec/0x1150
   alloc_page_interleave+0x1c/0xd0
   alloc_pages+0xec/0x160
   new_slab+0x330/0x454
   ___slab_alloc+0x5b8/0xba0
   __kmem_cache_alloc_node+0xf4/0x20c
   __kmalloc+0x60/0x100
   detect_cache_attributes+0x2a8/0x5a0
   update_siblings_masks+0x28/0x300
   store_cpu_topology+0x58/0x70
   secondary_start_kernel+0xc8/0x154

Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>

---

The patch extends and depends in context on:
https://lore.kernel.org/r/20221107155825.1644604-17-pierre.gondois@arm.com

Cache sizes from publicly available data:
https://en.wikichip.org/wiki/qualcomm/snapdragon_800/865
---
 arch/arm64/boot/dts/qcom/sm8250.dtsi | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)

diff --git a/arch/arm64/boot/dts/qcom/sm8250.dtsi b/arch/arm64/boot/dts/qcom/sm8250.dtsi
index 4478bf0e7b22..4555345ee42e 100644
--- a/arch/arm64/boot/dts/qcom/sm8250.dtsi
+++ b/arch/arm64/boot/dts/qcom/sm8250.dtsi
@@ -111,10 +111,14 @@ CPU0: cpu@0 {
 			L2_0: l2-cache {
 				compatible = "cache";
 				cache-level = <2>;
+				cache-size = <0x20000>;
+				cache-unified;
 				next-level-cache = <&L3_0>;
 				L3_0: l3-cache {
 					compatible = "cache";
 					cache-level = <3>;
+					cache-size = <0x400000>;
+					cache-unified;
 				};
 			};
 		};
@@ -137,6 +141,8 @@ CPU1: cpu@100 {
 			L2_100: l2-cache {
 				compatible = "cache";
 				cache-level = <2>;
+				cache-size = <0x20000>;
+				cache-unified;
 				next-level-cache = <&L3_0>;
 			};
 		};
@@ -159,6 +165,8 @@ CPU2: cpu@200 {
 			L2_200: l2-cache {
 				compatible = "cache";
 				cache-level = <2>;
+				cache-size = <0x20000>;
+				cache-unified;
 				next-level-cache = <&L3_0>;
 			};
 		};
@@ -181,6 +189,8 @@ CPU3: cpu@300 {
 			L2_300: l2-cache {
 				compatible = "cache";
 				cache-level = <2>;
+				cache-size = <0x20000>;
+				cache-unified;
 				next-level-cache = <&L3_0>;
 			};
 		};
@@ -203,6 +213,8 @@ CPU4: cpu@400 {
 			L2_400: l2-cache {
 				compatible = "cache";
 				cache-level = <2>;
+				cache-size = <0x40000>;
+				cache-unified;
 				next-level-cache = <&L3_0>;
 			};
 		};
@@ -225,6 +237,8 @@ CPU5: cpu@500 {
 			L2_500: l2-cache {
 				compatible = "cache";
 				cache-level = <2>;
+				cache-size = <0x40000>;
+				cache-unified;
 				next-level-cache = <&L3_0>;
 			};
 
@@ -248,6 +262,8 @@ CPU6: cpu@600 {
 			L2_600: l2-cache {
 				compatible = "cache";
 				cache-level = <2>;
+				cache-size = <0x40000>;
+				cache-unified;
 				next-level-cache = <&L3_0>;
 			};
 		};
@@ -270,6 +286,8 @@ CPU7: cpu@700 {
 			L2_700: l2-cache {
 				compatible = "cache";
 				cache-level = <2>;
+				cache-size = <0x80000>;
+				cache-unified;
 				next-level-cache = <&L3_0>;
 			};
 		};
-- 
2.34.1

On Thu, 29 Dec 2022 14:27:31 +0100, Krzysztof Kozlowski wrote:
> Add full cache description to DTS to avoid:
> 1. "Early cacheinfo failed" warnings,
> 2. Cache topology detection which leads to early memory allocations and
>    "BUG: sleeping function called from invalid context" on PREEMPT_RT
>    kernel:
> 
>   smp: Bringing up secondary CPUs ...
>   Detected VIPT I-cache on CPU1
>   BUG: sleeping function called from invalid context at kernel/locking/spinlock_rt.c:46
>   in_atomic(): 1, irqs_disabled(): 128, non_block: 0, pid: 0, name: swapper/1
>   preempt_count: 1, expected: 0
>   RCU nest depth: 1, expected: 1
>   3 locks held by swapper/1/0:
>    #0: ffff5e337eee5f18 (&pcp->lock){+.+.}-{3:3}, at: get_page_from_freelist+0x20c/0xffc
>    #1: ffffa9e24a900b18 (rcu_read_lock){....}-{1:3}, at: rt_spin_trylock+0x40/0xe4
>    #2: ffff5e337efc8918 (&zone->lock){+.+.}-{3:3}, at: rmqueue_bulk+0x54/0x720
>   irq event stamp: 0
>   Call trace:
>    __might_resched+0x17c/0x214
>    rt_spin_lock+0x5c/0x100
>    rmqueue_bulk+0x54/0x720
>    get_page_from_freelist+0xcfc/0xffc
>    __alloc_pages+0xec/0x1150
>    alloc_page_interleave+0x1c/0xd0
>    alloc_pages+0xec/0x160
>    new_slab+0x330/0x454
>    ___slab_alloc+0x5b8/0xba0
>    __kmem_cache_alloc_node+0xf4/0x20c
>    __kmalloc+0x60/0x100
>    detect_cache_attributes+0x2a8/0x5a0
>    update_siblings_masks+0x28/0x300
>    store_cpu_topology+0x58/0x70
>    secondary_start_kernel+0xc8/0x154
> 
> [...]

Applied, thanks!

[1/1] arm64: dts: qcom: sm8250: add cache size
      commit: ac1d8a8e2eb5bd67e266e3121bb6b39b7f28a9ec

Best regards,
-- 
Bjorn Andersson <andersson@kernel.org>