[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <CA1141EF-76A8-47A9-97B9-3CB2FC246B1A@oracle.com>
Date: Tue, 13 Apr 2021 17:01:37 -0400
From: Alex Kogan <alex.kogan@...cle.com>
To: Andi Kleen <ak@...ux.intel.com>
Cc: linux@...linux.org.uk, Peter Zijlstra <peterz@...radead.org>,
Ingo Molnar <mingo@...hat.com>, will.deacon@....com,
arnd@...db.de, longman@...hat.com, linux-arch@...r.kernel.org,
linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
tglx@...utronix.de, bp@...en8.de, hpa@...or.com, x86@...nel.org,
guohanjun@...wei.com, jglauber@...vell.com,
steven.sistare@...cle.com, daniel.m.jordan@...cle.com,
dave.dice@...cle.com
Subject: Re: [External] : Re: [PATCH v14 4/6] locking/qspinlock: Introduce
starvation avoidance into CNA
Hi, Andi.
Thanks for your comments!
> On Apr 13, 2021, at 2:03 AM, Andi Kleen <ak@...ux.intel.com> wrote:
>
> Alex Kogan <alex.kogan@...cle.com> writes:
>>
>> + numa_spinlock_threshold= [NUMA, PV_OPS]
>> + Set the time threshold in milliseconds for the
>> + number of intra-node lock hand-offs before the
>> + NUMA-aware spinlock is forced to be passed to
>> + a thread on another NUMA node. Valid values
>> + are in the [1..100] range. Smaller values result
>> + in a more fair, but less performant spinlock,
>> + and vice versa. The default value is 10.
>
> ms granularity seems very coarse grained for this. Surely
> at some point of spinning you can afford a ktime_get? But ok.
We are reading time when we are at the head of the (main) queue, but
don’t have the lock yet. Not sure about the latency of ktime_get(), but
anything reasonably fast but not necessarily precise should work.
> Could you turn that into a moduleparm which can be changed at runtime?
> Would be strange to have to reboot just to play with this parameter
Yes, good suggestion, thanks.
> This would also make the code a lot shorter I guess.
So you don’t think we need the command-line parameter, just the module_param?
Regards,
— Alex
Powered by blists - more mailing lists