linux-kernel - Re: [PATCH v1] drivers/acpi/scan.c: Fixup "acquire device_hotplug_lock in acpi_scan

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <23f28590-7765-bcd9-15f2-94e985b64218@redhat.com>
Date:   Wed, 31 Jul 2019 15:18:42 +0200
From:   David Hildenbrand <david@...hat.com>
To:     Michal Hocko <mhocko@...nel.org>
Cc:     linux-kernel@...r.kernel.org, linux-mm@...ck.org,
        linux-acpi@...r.kernel.org,
        "Rafael J . Wysocki" <rafael.j.wysocki@...el.com>,
        Oscar Salvador <osalvador@...e.de>,
        Andrew Morton <akpm@...ux-foundation.org>
Subject: Re: [PATCH v1] drivers/acpi/scan.c: Fixup "acquire
 device_hotplug_lock in acpi_scan_init()"

On 31.07.19 15:14, Michal Hocko wrote:
> On Wed 31-07-19 15:02:49, David Hildenbrand wrote:
>> On 31.07.19 14:53, Michal Hocko wrote:
>>> On Wed 31-07-19 14:32:01, David Hildenbrand wrote:
>>>> Let's document why we take the lock here. If we're going to overhaul
>>>> memory hotplug locking, we'll have to touch many places - this comment
>>>> will help to clairfy why it was added here.
>>>
>>> And how exactly is "lock for consistency" comment going to help the poor
>>> soul touching that code? How do people know that it is safe to remove it?
>>> I am not going to repeat my arguments how/why I hate "locking for
>>> consistency" (or fun or whatever but a real synchronization reasons)
>>> but if you want to help then just explicitly state what should done to
>>> remove this lock.
>>>
>>
>> I know that you have a different opinion here. To remove the lock,
>> add_memory() locking has to be changed *completely* to the point where
>> we can drop the lock from the documentation of the function (*whoever
>> knows what we have to exactly change* - and I don't have time to do that
>> *right now*).
> 
> Not really. To remove a lock in this particular path it would be
> sufficient to add
> 	/*
> 	 * Although __add_memory used down the road is documented to
> 	 * require lock_device_hotplug, it is not necessary here because
> 	 * this is an early code when userspace or any other code path
> 	 * cannot trigger hotplug operations.
> 	 */

Okay, let me phrase it like this: Are you 100% (!) sure that we don't
need the lock here. I am not -  I only know what I documented back then
and what I found out - could be that we are forgetting something else
the lock protects.

As I already said, I am fine with adding such a comment instead. But I
am not convinced that the absence of the lock is 100% safe. (I am 99.99%
sure ;) ).

-- 

Thanks,

David / dhildenb