Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
References: <20191208212206.16808-1-repk@triplefau.lt>
User-agent: mu4e 1.3.3; emacs 26.2
From:   Jerome Brunet <jbrunet@baylibre.com>
To:     Remi Pommarel <repk@triplefau.lt>,
        Neil Armstrong <narmstrong@baylibre.com>
Cc:     Michael Turquette <mturquette@baylibre.com>,
        Stephen Boyd <sboyd@kernel.org>,
        Kevin Hilman <khilman@baylibre.com>,
        linux-amlogic@lists.infradead.org, linux-clk@vger.kernel.org,
        linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH] clk: meson: pll: Fix by 0 division in __pll_params_to_rate()
In-reply-to: <20191208212206.16808-1-repk@triplefau.lt>
Date:   Mon, 09 Dec 2019 10:51:13 +0100
Message-ID: <1jo8whesj2.fsf@starbuckisacylon.baylibre.com>
MIME-Version: 1.0
Content-Type: text/plain
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk


On Sun 08 Dec 2019 at 22:22, Remi Pommarel <repk@triplefau.lt> wrote:

> Some meson pll registers can be initialized with 0 as N value, introducing
> the following division by 0 when computing rate :
>
>   UBSAN: Undefined behaviour in drivers/clk/meson/clk-pll.c:75:9
>   division by zero
>   CPU: 0 PID: 1 Comm: swapper/0 Not tainted 5.4.0-rc3-608075-g86c9af8630e1-dirty #400
>   Call trace:
>    dump_backtrace+0x0/0x1c0
>    show_stack+0x14/0x20
>    dump_stack+0xc4/0x100
>    ubsan_epilogue+0x14/0x68
>    __ubsan_handle_divrem_overflow+0x98/0xb8
>    __pll_params_to_rate+0xdc/0x140
>    meson_clk_pll_recalc_rate+0x278/0x3a0
>    __clk_register+0x7c8/0xbb0
>    devm_clk_hw_register+0x54/0xc0
>    meson_eeclkc_probe+0xf4/0x1a0
>    platform_drv_probe+0x54/0xd8
>    really_probe+0x16c/0x438
>    driver_probe_device+0xb0/0xf0
>    device_driver_attach+0x94/0xa0
>    __driver_attach+0x70/0x108
>    bus_for_each_dev+0xd8/0x128
>    driver_attach+0x30/0x40
>    bus_add_driver+0x1b0/0x2d8
>    driver_register+0xbc/0x1d0
>    __platform_driver_register+0x78/0x88
>    axg_driver_init+0x18/0x20
>    do_one_initcall+0xc8/0x24c
>    kernel_init_freeable+0x2b0/0x344
>    kernel_init+0x10/0x128
>    ret_from_fork+0x10/0x18
>
> This checks if N is null before doing the division.

Thanks for reporting this

>
> Fixes: 8289aafa4f36 ("clk: meson: improve pll driver results with
> frac")

In mainline, the commit above went in with sha1 3c4fe763d64d.

Also, this commit is not really responsible for the problem. Having HW
initialized with N = 0 would have failed since the beginning, I believe.

In this case the correct fixes would be:
Fixes: 7a29a869434e ("clk: meson: Add support for Meson clock controller")


> Signed-off-by: Remi Pommarel <repk@triplefau.lt>
> ---
>  drivers/clk/meson/clk-pll.c | 4 ++++
>  1 file changed, 4 insertions(+)
>
> diff --git a/drivers/clk/meson/clk-pll.c b/drivers/clk/meson/clk-pll.c
> index ddb1e5634739..6649659f216a 100644
> --- a/drivers/clk/meson/clk-pll.c
> +++ b/drivers/clk/meson/clk-pll.c
> @@ -66,6 +66,10 @@ static unsigned long __pll_params_to_rate(unsigned long parent_rate,
>  					 (1 << pll->frac.width));
>  	}
>  
> +	/* Avoid by zero division */
> +	if (n == 0)
> +		return 0;

This can only really happen after init, in recalc() rate.

I would much prefer if you could check the n value right after it is
read (meson_parm_read()) in .recalc_rate() and add a comment explaining
that some HW may have this parameter set 0 on init.

> +
>  	return DIV_ROUND_UP_ULL(rate, n);
>  }