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: <e77f9820-c818-d7af-185b-b8814313048b@suse.cz>
Date:   Fri, 29 Jul 2022 17:08:31 +0200
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101
 Thunderbird/91.11.0
Subject: Re: [PATCH v3 00/15] common kmalloc v3
Content-Language: en-US
To:     Hyeonggon Yoo <42.hyeyoo@gmail.com>,
        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>,
        Joe Perches <joe@perches.com>,
        Vasily Averin <vasily.averin@linux.dev>,
        Matthew WilCox <willy@infradead.org>
Cc:     linux-kernel@vger.kernel.org, linux-mm@kvack.org
References: <20220712133946.307181-1-42.hyeyoo@gmail.com>
From:   Vlastimil Babka <vbabka@suse.cz>
In-Reply-To: <20220712133946.307181-1-42.hyeyoo@gmail.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 7/12/22 15:39, Hyeonggon Yoo wrote:
> This is v3 of common kmalloc series.
> 
> This series generalize most of kmalloc functions and move its
> implementation into mm/slab_common.c.
> 
> I believe this series give better maintainability of code for SLAB and SLUB.
> Please consider applying.
> 
> This series is based on slab/for-next and also available at
> https://github.com/hygoni/linux/tree/slab-common-v3r0
> 
> Any feedbacks/reviews will be appreciated.

Hi, thanks for all your efforts. It's shaping up nicely so I think the next
version will be ready to be added to -next after the 5.20 merge window.
As I've finished the individual reviews, I'm looking at the result and see a
bit more potential for cleanups, which could be perhaps incorporated to
existing patches, or additionally:

- kmalloc_large_node_notrace() has only one caller, can be removed and the
caller can call __kmalloc_large_node_notrace() directly, especially if it's
not __always_inline as I've IIRC suggested.

- kmem_cache_alloc_trace() and kmem_cache_alloc_node_trace() are weird ones,
they are in fact for kmalloc despite the name. They depend on
CONFIG_TRACING, yet if you look carefully, the !CONFIG_TRACING variant also
goes through a trace_* function. The actual difference seems that
slab_alloc() thus kasan_alloc() and kfence_alloc() don't get the orig_size
that way, which is dubious. It might be worth trying to unify this as well?
E.g.
  - use only the CONFIG_TRACING variant, discard the other
  - declare it in mm/slab.h, this is not for general usage
  - single implementation in mm/slab_common.c that calls
    __kmem_cache_alloc_node() from SLAB/SLUB and does the trace