diff mbox series

riscv: setup per-hart stack earlier

Message ID 1683596741-2429007-1-git-send-email-ganboing@gmail.com
State Superseded
Delegated to: Andes
Headers show
Series riscv: setup per-hart stack earlier | expand

Commit Message

Bo Gan May 9, 2023, 1:45 a.m. UTC
Harts need to use per-hart stack before any function call, even if that
function is a simple one. When the callee uses stack for register save/
restore, especially RA, if nested call, concurrent access by multiple
harts on the same stack will cause data-race.

This patch sets up SP before `board_init_f_alloc_reserve`. A side effect
of this is that the memory layout has changed as the following:

+----------------+        +----------------+ <----- SPL_STACK/
|  ......        |        |  hart 0 stack  |        SYS_INIT_SP_ADDR
|  malloc_base   |        +----------------+
+----------------+        |  hart 1 stack  |
|  GD            |        +----------------+ If not SMP, N=1
+----------------+        |  ......        |
|  hart 0 stack  |        +----------------+
+----------------+   ==>  |  hart N-1 stack|
|  hart 1 stack  |        +----------------+
+----------------+        |  ......        |
|  ......        |        |  malloc_base   |
+----------------+        +----------------+
|  hart N-1 stack|        |  GD            |
+----------------+        +----------------+
|                |        |                |

Signed-off-by: Bo Gan <ganboing@gmail.com>
Cc: Rick Chen <rick@andestech.com>
Cc: Leo <ycliang@andestech.com>
---
 arch/riscv/cpu/start.S | 37 ++++++++++++++++++++++++-------------
 1 file changed, 24 insertions(+), 13 deletions(-)

Comments

Rick Chen May 15, 2023, 5:08 a.m. UTC | #1
Hi Bo Gan,

> From: Bo Gan <ganboing@gmail.com>
> Sent: Tuesday, May 09, 2023 9:46 AM
> To: u-boot@lists.denx.de
> Cc: Bo Gan <ganboing@gmail.com>; Rick Jian-Zhi Chen(陳建志) <rick@andestech.com>; Leo Yu-Chi Liang(梁育齊) <ycliang@andestech.com>
> Subject: [PATCH] riscv: setup per-hart stack earlier
>
> Harts need to use per-hart stack before any function call, even if that function is a simple one. When the callee uses stack for register save/ restore, especially RA, if nested call, concurrent access by multiple harts on the same stack will cause data-race.
>
> This patch sets up SP before `board_init_f_alloc_reserve`. A side effect of this is that the memory layout has changed as the following:
>
> +----------------+        +----------------+ <----- SPL_STACK/
> |  ......        |        |  hart 0 stack  |        SYS_INIT_SP_ADDR
> |  malloc_base   |        +----------------+
> +----------------+        |  hart 1 stack  |
> |  GD            |        +----------------+ If not SMP, N=1
> +----------------+        |  ......        |
> |  hart 0 stack  |        +----------------+
> +----------------+   ==>  |  hart N-1 stack|
> |  hart 1 stack  |        +----------------+
> +----------------+        |  ......        |
> |  ......        |        |  malloc_base   |
> +----------------+        +----------------+
> |  hart N-1 stack|        |  GD            |
> +----------------+        +----------------+
> |                |        |                |
>
> Signed-off-by: Bo Gan <ganboing@gmail.com>
> Cc: Rick Chen <rick@andestech.com>
> Cc: Leo <ycliang@andestech.com>
> ---
>  arch/riscv/cpu/start.S | 37 ++++++++++++++++++++++++-------------
>  1 file changed, 24 insertions(+), 13 deletions(-)
>
> diff --git a/arch/riscv/cpu/start.S b/arch/riscv/cpu/start.S index dad22bf..90015c2 100644
> --- a/arch/riscv/cpu/start.S
> +++ b/arch/riscv/cpu/start.S
> @@ -91,16 +91,35 @@ _start:
>   * Set stackpointer in internal/ex RAM to call board_init_f
>   */
>  call_board_init_f:
> -       li      t0, -16
>  #if defined(CONFIG_SPL_BUILD) && defined(CONFIG_SPL_STACK)
> -       li      t1, CONFIG_SPL_STACK
> +       li      t0, CONFIG_SPL_STACK
>  #else
> -       li      t1, SYS_INIT_SP_ADDR
> +       li      t0, CONFIG_SYS_INIT_SP_ADDR

It builds fail as below:

arch/riscv/cpu/start.S:97: Error: illegal operands `li
t0,CONFIG_SYS_INIT_SP_ADDR'

Thanks,
Rick
Bo Gan May 21, 2023, 6:26 a.m. UTC | #2
On 5/14/23 10:08 PM, Rick Chen wrote:
> Hi Bo Gan,
> It builds fail as below:
> 
> arch/riscv/cpu/start.S:97: Error: illegal operands `li
> t0,CONFIG_SYS_INIT_SP_ADDR'
> 
> Thanks,
> Rick
> 

Hi Rick & Leo,

Please help take a look at v2 of the patch:
https://patchwork.ozlabs.org/project/uboot/patch/1684650044-313122-1-git-send-email-ganboing@gmail.com/

I've fixed the macro and also did some testing on my vf2 board.

Thanks!
diff mbox series

Patch

diff --git a/arch/riscv/cpu/start.S b/arch/riscv/cpu/start.S
index dad22bf..90015c2 100644
--- a/arch/riscv/cpu/start.S
+++ b/arch/riscv/cpu/start.S
@@ -91,16 +91,35 @@  _start:
  * Set stackpointer in internal/ex RAM to call board_init_f
  */
 call_board_init_f:
-	li	t0, -16
 #if defined(CONFIG_SPL_BUILD) && defined(CONFIG_SPL_STACK)
-	li	t1, CONFIG_SPL_STACK
+	li	t0, CONFIG_SPL_STACK
 #else
-	li	t1, SYS_INIT_SP_ADDR
+	li	t0, CONFIG_SYS_INIT_SP_ADDR
 #endif
-	and	sp, t1, t0		/* force 16 byte alignment */
+	and	t0, t0, -16		/* force 16 byte alignment */
+
+	/* setup stack */
+#if CONFIG_IS_ENABLED(SMP)
+	/* tp: hart id */
+	slli	t1, tp, CONFIG_STACK_SIZE_SHIFT
+	sub	sp, t0, t1
+#else
+	mv	sp, t0
+#endif
+/*
+ * Now sp points to the right stack belonging to current CPU.
+ * It's essential before any function call, otherwise, we get data-race.
+ */
 
 call_board_init_f_0:
-	mv	a0, sp
+	/* find top of reserve space */
+#if CONFIG_IS_ENABLED(SMP)
+	li	t1, CONFIG_NR_CPUS
+#else
+	li	t1, 1
+#endif
+	slli	t1, t1, CONFIG_STACK_SIZE_SHIFT
+	sub	a0, t0, t1		/* t1 -> size of all CPU stacks */
 	jal	board_init_f_alloc_reserve
 
 	/*
@@ -109,14 +128,6 @@  call_board_init_f_0:
 	 */
 	mv	s0, a0
 
-	/* setup stack */
-#if CONFIG_IS_ENABLED(SMP)
-	/* tp: hart id */
-	slli	t0, tp, CONFIG_STACK_SIZE_SHIFT
-	sub	sp, a0, t0
-#else
-	mv	sp, a0
-#endif
 
 	/* Configure proprietary settings and customized CSRs of harts */
 call_harts_early_init: