Re: [PATCH 0/5] x86/speculation: Disable IBRS when idle

[Waiman Long <longman@redhat.com>]

On 6/18/23 21:18, Waiman Long wrote:
> On 6/17/23 08:21, Peter Zijlstra wrote:
>> On Fri, Jun 16, 2023 at 10:14:52PM +0200, Robin Jarry wrote:
>>> Waiman Long, Jun 16, 2023 at 21:59:
>>>> For Intel processors that need to turn on IBRS to protect against
>>>> Spectre v2 and Retbleed, the IBRS bit in the SPEC_CTRL MSR affects
>>>> the performance of the whole core even if only one thread is turning
>>>> it on when running in the kernel. For user space heavy applications,
>>>> the performance impact of occasionally turning IBRS on during syscalls
>>>> shouldn't be significant. Unfortunately, that is not the case when the
>>>> sibling thread is idling in the kernel. In that case, the performance
>>>> impact can be significant.
>>>>
>>>> When DPDK is running on an isolated CPU thread processing network 
>>>> packets
>>>> in user space while its sibling thread is idle. The performance of the
>>>> busy DPDK thread with IBRS on and off in the sibling idle thread are:
>>>>
>>>>                                  IBRS on               IBRS off
>>>>                                  ------- --------
>>>>    packets/second:                  7.8M 10.4M
>>>>    avg tsc cycles/packet:         282.26 209.86
>>>>
>>>> This is a 25% performance degradation. The test system is a Intel Xeon
>>>> 4114 CPU @ 2.20GHz.
>>>>
>>>> This patch series turns off IBRS when in various idle mode to 
>>>> eliminate
>>>> the performance impact of the idling thread on its busy sibling 
>>>> thread.
>>> Hi Longman,
>>>
>>> thanks a lot for the quick turnaround on this issue.
>>>
>>> Tested-by: Robin Jarry <rjarry@redhat.com>
>> I can't see the patches -- they didn't arrive in my mailbox nor can I
>> find them in the archive, in fact this here mail is the only evidence
>> they exist at all.
>
> I got a rebound message from your mail server about incorrect message 
> format. It is probably caused by some problem in my end.
>
>
>> However, did you all see intel_idle_ibrs() and how that is selected for
>> C6 and up?
>>
>> What exactly isn't working there?
>
> We were testing on the RHEL9.2 kernel which doesn't have your 
> intel_idle_ibrs() patch yet. My preliminary testing does indicate your 
> patch will likely work. I will ask Jerry to test a newer RHEL9.3 
> kernel with the intel_idle_ibrs() patch to see if it helps.

We may need to extend your current solution to cover more cases. Perhaps 
adding a module parameter (e.g. idle_no_ibrs) to force the use of 
intel_idle_ibrs(). BTW, is it really the case that we can't disable IBRS 
when irq is enabled? The idle thread does not really interact with any 
user applications. I don't think there is any risk of information 
leakage even if we disable IBRS with interrupt enabled. Is my assumption 
incorrect?

Thanks,
Longman