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;
Message-ID: <60111450-d66a-a146-1a70-0c093400f3e5@suse.cz>
Date:   Tue, 20 Sep 2022 11:29:40 +0200
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.2.2
Subject: Re: [PATCH] mm: Make failslab writable again
Content-Language: en-US
To:     Alexander Atanasov <alexander.atanasov@virtuozzo.com>,
        Jonathan Corbet <corbet@lwn.net>,
        Christoph Lameter <cl@linux.com>,
        Pekka Enberg <penberg@kernel.org>,
        David Rientjes <rientjes@google.com>,
        Joonsoo Kim <iamjoonsoo.kim@lge.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        Roman Gushchin <roman.gushchin@linux.dev>,
        Hyeonggon Yoo <42.hyeyoo@gmail.com>
Cc:     kernel@openvz.org, Kees Cook <keescook@chromium.org>,
        Roman Gushchin <guro@fb.com>, Jann Horn <jannh@google.com>,
        Vijayanand Jitta <vjitta@codeaurora.org>,
        linux-doc@vger.kernel.org, linux-kernel@vger.kernel.org,
        linux-mm@kvack.org
References: <20220920082033.1727374-1-alexander.atanasov@virtuozzo.com>
 <21646f5c-39ee-a51c-f30c-272de85ee350@suse.cz>
 <6f5a8f71-3d82-97bf-90e1-0f33546bb59b@virtuozzo.com>
From:   Vlastimil Babka <vbabka@suse.cz>
In-Reply-To: <6f5a8f71-3d82-97bf-90e1-0f33546bb59b@virtuozzo.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Precedence: bulk

On 9/20/22 11:17, Alexander Atanasov wrote:
> Hello,
> 
> On 20.09.22 11:42, Vlastimil Babka wrote:
>> On 9/20/22 10:20, Alexander Atanasov wrote:
>>> In (060807f841ac mm, slub: make remaining slub_debug related attributes
>>> read-only failslab) it was made RO.
>>
>> "read-only) failslab was made RO" ?
> 
> Yep.
> 
>>> I think it became a collateral victim to the other two options
>>> (sanity_checks and trace) for which the reasons are perfectly valid.
>>
>> The commit also mentioned that modifying the flags is not protected in any
>> way, see below.
> 
> Yes, indeed.
> 
>>> +static ssize_t failslab_store(struct kmem_cache *s, const char *buf,
>>> +                size_t length)
>>> +{
>>> +    if (s->refcount > 1)
>>> +        return -EINVAL;
>>> +
>>> +    s->flags &= ~SLAB_FAILSLAB;
>>> +    if (buf[0] == '1')
>>> +        s->flags |= SLAB_FAILSLAB;
>>
>> Could we at least use a temporary variable to set up the final value and
>> then do a WRITE_ONCE() to s->flags, so the compiler is not allowed to do
>> some funky stuff? Assuming this is really the only place where we modify
>> s->flags during runtime, so we can't miss other updates due to RMW.
> 
> Since it is set or clear - instead of temporary variable and potentially two
> writes and RMW issues i would suggest this:
> +    if (buf[0] == '1')
> +        s->flags |= SLAB_FAILSLAB;
> +       else
> +        s->flags &= ~SLAB_FAILSLAB;

This way also has RMW issues, and also the compiler is allowed to
temporarily modify s->flags any way it likes; with WRITE_ONCE() it can't.

> If at some point more places need to modify the flags at runtime they can
> switch to atomic bit ops.