Re: [PATCH] i386 spinlocks should use the full 32 bits, not only8 bits
From: Eric Dumazet
Date: Tue Oct 11 2005 - 11:37:58 EST
Linus Torvalds a écrit :
On Tue, 11 Oct 2005, Eric Dumazet wrote:
As NR_CPUS might be > 128, and every spining CPU decrements the lock, we need
to use more than 8 bits for a spinlock. The current (i386/x86_64)
implementations have a (theorical) bug in this area.
I don't think there are any x86 machines with > 128 CPU's right now.
The advantage of the byte lock is that a "movb $0" is three bytes shorter
than a "movl $0". And that's the unlock sequence.
1) Would you prefer to change arch/i386/Kconfig
int "Maximum number of CPUs (2-255)"
range 2 255
2) The unlock sequence is not anymore inlined. It appears twice or three times
in the kernel.
3) i386 code is often taken as the base when a port is done. For example
x86_64 has the same problem.
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/