From: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Date: Mon, 9 Feb 2009 17:55:09 +0000 (-0500)
Subject: Raise the number of nested readers limit 2^16 and 2^32 on 32 and 64-bits arch.
X-Git-Tag: v0.1~303
X-Git-Url: https://git.lttng.org/?p=urcu.git;a=commitdiff_plain;h=4917a879125c1fca08ccb4328104701e83108556

Raise the number of nested readers limit 2^16 and 2^32 on 32 and 64-bits arch.

> I don't know of any code in the Linux kernel that nests rcu_read_lock()
> anywhere near that deep.  And if someone does find such a case, it is
> pretty easy to use 15 bits rather than 8 to hold the nesting depth, just
> by changing the definition of RCU_GP_CTR_BIT.
>

You know what ? Changing RCU_GP_CTR_BIT to 16 uses a
testw %ax, %ax instead of a testb %al, %al. The trick here is that
RCU_GP_CTR_BIT must be a multiple of 8 so we can use a full 8-bits,
16-bits or 32-bits bitmask for the lower order bits.

On 64-bits, using a RCU_GP_CTR_BIT of 32 is also ok. It uses a testl.

To provide 32-bits compability and allow the deepest nesting possible, I
think it makes sense to use

/* Use the amount of bits equal to half of the architecture long size */

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
---

diff --git a/urcu.h b/urcu.h
index 2e8c042..b8f6d9b 100644
--- a/urcu.h
+++ b/urcu.h
@@ -162,11 +162,12 @@ static inline void debug_yield_init(void)
 #endif
 
 /*
- * Limiting the nesting level to 256 to keep instructions small in the read
- * fast-path.
+ * The trick here is that RCU_GP_CTR_BIT must be a multiple of 8 so we can use a
+ * full 8-bits, 16-bits or 32-bits bitmask for the lower order bits.
  */
 #define RCU_GP_COUNT		(1U << 0)
-#define RCU_GP_CTR_BIT		(1U << 8)
+/* Use the amount of bits equal to half of the architecture long size */
+#define RCU_GP_CTR_BIT		(sizeof(long) << 2)
 #define RCU_GP_CTR_NEST_MASK	(RCU_GP_CTR_BIT - 1)
 
 /*