[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <d3e5ab37839047a483212c8b3fc667e5@AcuMS.aculab.com>
Date: Sun, 17 Dec 2023 17:40:19 +0000
From: David Laight <David.Laight@...LAB.COM>
To: 'Linus Torvalds' <torvalds@...ux-foundation.org>, Steven Rostedt
<rostedt@...dmis.org>
CC: "linux-arch@...r.kernel.org" <linux-arch@...r.kernel.org>, LKML
<linux-kernel@...r.kernel.org>, Linux Trace Kernel
<linux-trace-kernel@...r.kernel.org>, Masami Hiramatsu <mhiramat@...nel.org>,
Mark Rutland <mark.rutland@....com>, Mathieu Desnoyers
<mathieu.desnoyers@...icios.com>
Subject: RE: [PATCH] ring-buffer: Remove 32bit timestamp logic
...
> My guess is that *most* 32-bit architectures do not have a 64-bit
> cmpxchg - not even the irq-safe one.
Does any sparc32 even have a 32-bit cmpxchg?
The original versions (which were definitely SMP capable)
only had a byte sized atomic exchange that always wrote 0xff.
Sparc32 does have 'load/store double' (two 32bit registers)
but 32bit cpu like nios2 and (I think) RISCV (and probably
anything else loosely based on MIPS) only have single register
load/store instructions.
They'll mainly be UP only, I've not looked at RISCV enough to
see what it has when supporting SMP.
> For the UP case you can do your own, of course.
A generic version of the soft interrupt disable code would help.
Then it would just be an inc/dec of memory rather than having
to save the current interrupt enable state.
Especially for code that only disables interrupts for a few
instructions - so the costs of deferring the interrupt don't
happen often.
David
-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)
Powered by blists - more mailing lists