Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Valentin Schneider <vschneid@redhat.com>
To:     Peter Zijlstra <peterz@infradead.org>
Cc:     linux-kernel@vger.kernel.org, linux-trace-kernel@vger.kernel.org,
        linux-doc@vger.kernel.org, kvm@vger.kernel.org, linux-mm@kvack.org,
        bpf@vger.kernel.org, x86@kernel.org,
        Nicolas Saenz Julienne <nsaenzju@redhat.com>,
        Steven Rostedt <rostedt@goodmis.org>,
        Masami Hiramatsu <mhiramat@kernel.org>,
        Jonathan Corbet <corbet@lwn.net>,
        Thomas Gleixner <tglx@linutronix.de>,
        Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
        Dave Hansen <dave.hansen@linux.intel.com>,
        "H. Peter Anvin" <hpa@zytor.com>,
        Paolo Bonzini <pbonzini@redhat.com>,
        Wanpeng Li <wanpengli@tencent.com>,
        Vitaly Kuznetsov <vkuznets@redhat.com>,
        Andy Lutomirski <luto@kernel.org>,
        Frederic Weisbecker <frederic@kernel.org>,
        "Paul E. McKenney" <paulmck@kernel.org>,
        Andrew Morton <akpm@linux-foundation.org>,
        Uladzislau Rezki <urezki@gmail.com>,
        Christoph Hellwig <hch@infradead.org>,
        Lorenzo Stoakes <lstoakes@gmail.com>,
        Josh Poimboeuf <jpoimboe@kernel.org>,
        Kees Cook <keescook@chromium.org>,
        Sami Tolvanen <samitolvanen@google.com>,
        Ard Biesheuvel <ardb@kernel.org>,
        Nicholas Piggin <npiggin@gmail.com>,
        Juerg Haefliger <juerg.haefliger@canonical.com>,
        Nicolas Saenz Julienne <nsaenz@kernel.org>,
        "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>,
        Nadav Amit <namit@vmware.com>,
        Dan Carpenter <error27@gmail.com>,
        Chuang Wang <nashuiliang@gmail.com>,
        Yang Jihong <yangjihong1@huawei.com>,
        Petr Mladek <pmladek@suse.com>,
        "Jason A. Donenfeld" <Jason@zx2c4.com>, Song Liu <song@kernel.org>,
        Julian Pidancet <julian.pidancet@oracle.com>,
        Tom Lendacky <thomas.lendacky@amd.com>,
        Dionna Glaze <dionnaglaze@google.com>,
        Thomas =?utf-8?Q?Wei=C3=9Fschuh?= <linux@weissschuh.net>,
        Juri Lelli <juri.lelli@redhat.com>,
        Daniel Bristot de Oliveira <bristot@redhat.com>,
        Marcelo Tosatti <mtosatti@redhat.com>,
        Yair Podemsky <ypodemsk@redhat.com>
Subject: Re: [RFC PATCH 11/14] context-tracking: Introduce work deferral
 infrastructure
In-Reply-To: <20230705223956.GD2813335@hirez.programming.kicks-ass.net>
References: <20230705181256.3539027-1-vschneid@redhat.com>
 <20230705181256.3539027-12-vschneid@redhat.com>
 <20230705223956.GD2813335@hirez.programming.kicks-ass.net>
Date:   Thu, 06 Jul 2023 12:31:11 +0100
Message-ID: <xhsmhsfa1uvi8.mognet@vschneid.remote.csb>
MIME-Version: 1.0
Content-Type: text/plain
Precedence: bulk

On 06/07/23 00:39, Peter Zijlstra wrote:
> On Wed, Jul 05, 2023 at 07:12:53PM +0100, Valentin Schneider wrote:
>
>> Note: A previous approach by PeterZ [1] used an extra bit in
>> context_tracking.state to flag the presence of deferred callbacks to
>> execute, and the actual callbacks were stored in a separate atomic
>> variable.
>>
>> This meant that the atomic read of context_tracking.state was sufficient to
>> determine whether there are any deferred callbacks to execute.
>> Unfortunately, it presents a race window. Consider the work setting
>> function as:
>>
>>   preempt_disable();
>>   seq = atomic_read(&ct->seq);
>>   if (__context_tracking_seq_in_user(seq)) {
>>        /* ctrl-dep */
>>        atomic_or(work, &ct->work);
>>        ret = atomic_try_cmpxchg(&ct->seq, &seq, seq|CT_SEQ_WORK);
>>   }
>>   preempt_enable();
>>
>>   return ret;
>>
>> Then the following can happen:
>>
>>   CPUx                                             CPUy
>>                                                   CT_SEQ_WORK \in context_tracking.state
>>     atomic_or(WORK_N, &ct->work);
>>                                                    ct_kernel_enter()
>>                                                      ct_state_inc();
>>     atomic_try_cmpxchg(&ct->seq, &seq, seq|CT_SEQ_WORK);
>>
>> The cmpxchg() would fail, ultimately causing an IPI for WORK_N to be
>> sent. Unfortunately, the work bit would remain set, and it can't be sanely
>> cleared in case another CPU set it concurrently - this would ultimately
>> lead to a double execution of the callback, one as a deferred callback and
>> one in the IPI. As not all IPI callbacks are idempotent, this is
>> undesirable.
>
> So adding another atomic is arguably worse.
>
> The thing is, if the NOHZ_FULL CPU is actually doing context transitions
> (SYSCALLs etc..) then everything is fundamentally racy, there is no
> winning that game, we could find the remote CPU is in-kernel, send an
> IPI, the remote CPU does return-to-user and receives the IPI.
>
> And then the USER is upset... because he got an IPI.
>

Yeah, that part is inevitably racy.

The thing I was especially worried about was the potential double
executions (once in IPI, again in deferred work). It's not /too/ bad as the
only two deferred callbacks I'm introducing here are costly-but-stateless,
but IMO is a bad foundation.

But it seems like we can reuse the existing atomic and squeeze some bits in
there, so let's see how that goes :-)