Date:   Tue, 28 Feb 2023 11:13:10 -0800
From:   Andrew Morton <akpm@linux-foundation.org>
To:     Jiri Olsa <jolsa@kernel.org>
Cc:     Alexei Starovoitov <ast@kernel.org>,
        Andrii Nakryiko <andrii@kernel.org>,
        Hao Luo <haoluo@google.com>,
        Alexander Viro <viro@zeniv.linux.org.uk>,
        Peter Zijlstra <peterz@infradead.org>,
        Ingo Molnar <mingo@redhat.com>,
        Arnaldo Carvalho de Melo <acme@kernel.org>,
        Matthew Wilcox <willy@infradead.org>, bpf@vger.kernel.org,
        linux-mm@kvack.org, linux-kernel@vger.kernel.org,
        linux-fsdevel@vger.kernel.org, linux-perf-users@vger.kernel.org,
        Martin KaFai Lau <kafai@fb.com>,
        Song Liu <songliubraving@fb.com>, Yonghong Song <yhs@fb.com>,
        John Fastabend <john.fastabend@gmail.com>,
        KP Singh <kpsingh@chromium.org>,
        Stanislav Fomichev <sdf@google.com>,
        Daniel Borkmann <daniel@iogearbox.net>,
        Namhyung Kim <namhyung@gmail.com>
Subject: Re: [PATCH RFC v2 bpf-next 1/9] mm: Store build id in inode object
Message-Id: <20230228111310.05f339a0a1a00e919859ffad@linux-foundation.org>
In-Reply-To: <20230228093206.821563-2-jolsa@kernel.org>
References: <20230228093206.821563-1-jolsa@kernel.org>
        <20230228093206.821563-2-jolsa@kernel.org>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Precedence: bulk

On Tue, 28 Feb 2023 10:31:58 +0100 Jiri Olsa <jolsa@kernel.org> wrote:

> Storing build id in file's inode object for elf executable with build
> id defined. The build id is stored when file is mmaped.
> 
> This is enabled with new config option CONFIG_INODE_BUILD_ID.
> 
> The build id is valid only when the file with given inode is mmap-ed.
> 
> We store either the build id itself or the error we hit during
> the retrieval.
> 
> ...
>
> --- a/include/linux/fs.h
> +++ b/include/linux/fs.h
> @@ -699,6 +700,12 @@ struct inode {
>  	struct fsverity_info	*i_verity_info;
>  #endif
>  
> +#ifdef CONFIG_INODE_BUILD_ID
> +	/* Initialized and valid for executable elf files when mmap-ed. */
> +	struct build_id		*i_build_id;
> +	spinlock_t		i_build_id_lock;
> +#endif
> +

Remember we can have squillions of inodes in memory.  So that's one
costly spinlock!

AFAICT this lock could be removed if mmap_region() were to use an
atomic exchange on inode->i_build_id?

If not, can we use an existing lock?  i_lock would be appropriate
(don't forget to update its comment).

Also, the code in mmap_region() runs build_id_free() inside the locked
region, which seems unnecessary.