[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20220425093509.GG2731@worktop.programming.kicks-ass.net>
Date: Mon, 25 Apr 2022 11:35:09 +0200
From: Peter Zijlstra <peterz@...radead.org>
To: Donghai Qiao <dqiao@...hat.com>
Cc: akpm@...ux-foundation.org, sfr@...b.auug.org.au, arnd@...db.de,
heying24@...wei.com, andriy.shevchenko@...ux.intel.com,
axboe@...nel.dk, rdunlap@...radead.org, tglx@...utronix.de,
gor@...ux.ibm.com, donghai.w.qiao@...il.com,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2 05/11] smp: replace smp_call_function_single_async()
with smp_call_private()
On Fri, Apr 22, 2022 at 04:00:34PM -0400, Donghai Qiao wrote:
> Replaced smp_call_function_single_async() with smp_call_private()
> and also extended smp_call_private() to support one CPU synchronous
> call with preallocated csd structures.
>
> Ideally, the new interface smp_call() should be able to do what
> smp_call_function_single_async() does. Because the csd is provided
> and maintained by the callers, it exposes the risk of corrupting
> the call_single_queue[cpu] linked list if the clents menipulate
> their csd inappropriately. On the other hand, there should have no
> noticeable performance advantage to provide preallocated csd for
> cross call kernel consumers. Thus, in the long run, the consumers
> should change to not use this type of preallocated csd.
You're completely missing the point of this interface. *please* look at
it more carefully.
Powered by blists - more mailing lists