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:   Thomas Gleixner <tglx@linutronix.de>
To:     Peter Zijlstra <peterz@infradead.org>, Jann Horn <jannh@google.com>
Cc:     Ingo Molnar <mingo@redhat.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        Darren Hart <dvhart@infradead.org>,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH] threads: Update PID limit comment according to futex
 UAPI change
In-Reply-To: <87mt2bc9wz.ffs@tglx>
References: <20200302112939.8068-1-jannh@google.com>
 <20200303102540.GC2579@hirez.programming.kicks-ass.net>
 <8736apede7.fsf@nanos.tec.linutronix.de> <87mt2bc9wz.ffs@tglx>
Date:   Thu, 11 May 2023 15:26:22 +0200
Message-ID: <87jzxfc83l.ffs@tglx>
MIME-Version: 1.0
Content-Type: text/plain
Precedence: bulk

On Thu, May 11 2023 at 14:47, Thomas Gleixner wrote:
>> #define FUTEX_TID_MASK          0x3fffffff
>
> That futex mask is irrelevant because there is another limitiation for
> the PID space namely posix CPU timers:
>
> static inline clockid_t make_process_cpuclock(const unsigned int pid,
> 		                              const clockid_t clock)
> {
> 	return ((~pid) << 3) | clock;
> }
>
> That existed even before we fiddled with robust futexes and puts a hard
> limit of 29 bits on the PID space.

Actually it's worse than this. The limit is 28 bits because posix CPU
timers rely on this:

static const struct k_clock *clockid_to_kclock(const clockid_t id)
{
	clockid_t idx = id;

	if (id < 0) {
		return (id & CLOCKFD_MASK) == CLOCKFD ?
			&clock_posix_dynamic : &clock_posix_cpu;
	}
        ....

So TID must be smaller than 2^28 because if bit 28 would be set, then
the resulting clock would be positive after shift.

Sigh...