Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Fri, 9 Apr 2021 12:08:37 +0200
From:   Oleg Nesterov <oleg@redhat.com>
To:     Jens Axboe <axboe@kernel.dk>
Cc:     Peter Zijlstra <peterz@infradead.org>, linux-block@vger.kernel.org,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH] block: Fix sys_ioprio_set(.which=IOPRIO_WHO_PGRP) task
 iteration
Message-ID: <20210409100836.GB10447@redhat.com>
References: <YG7Q5C4Rb5dx5GFx@hirez.programming.kicks-ass.net>
 <e8579a4e-2456-a89b-b750-892d265ba053@kernel.dk>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <e8579a4e-2456-a89b-b750-892d265ba053@kernel.dk>
User-Agent: Mutt/1.5.24 (2015-08-30)
Precedence: bulk

On 04/08, Jens Axboe wrote:
>
> On 4/8/21 3:46 AM, Peter Zijlstra wrote:
> >
> > do_each_pid_thread() { } while_each_pid_thread() is a double loop and
> > thus break doesn't work as expected. Also, it should be used under
> > tasklist_lock because otherwise we can race against change_pid() for
> > PGID/SID.
>
> Applied, thanks.

Agreed, but can't resist. We can move the "out" label up and avoid the extra
read_unlock(tasklist_lock). IOW, something like below on top of this patch.

Quite possibly this won't change the generated code, gcc is smart enough, but
this makes the code a bit more readable.

Oleg.

--- x/block/ioprio.c~	2021-04-09 12:00:28.066145563 +0200
+++ x/block/ioprio.c	2021-04-09 12:02:01.817849618 +0200
@@ -123,11 +123,10 @@
 			read_lock(&tasklist_lock);
 			do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
 				ret = set_task_ioprio(p, ioprio);
-				if (ret) {
-					read_unlock(&tasklist_lock);
-					goto out;
-				}
+				if (ret)
+					goto out_pgrp;
 			} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
+out_pgrp:
 			read_unlock(&tasklist_lock);
 
 			break;
@@ -159,7 +158,6 @@
 			ret = -EINVAL;
 	}
 
-out:
 	rcu_read_unlock();
 	return ret;
 }