linux-kernel - Re: [PATCH 4.19 29/74] clk: qcom: regmap-mux: fix parent clock lookup

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: <67a77fc5-6db4-ba26-cacb-9758336ad074@linaro.org>
Date:   Wed, 15 Dec 2021 13:37:24 +0300
From:   Dmitry Baryshkov <dmitry.baryshkov@...aro.org>
To:     Pavel Machek <pavel@...x.de>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>
Cc:     linux-kernel@...r.kernel.org, stable@...r.kernel.org,
        Stephen Boyd <sboyd@...nel.org>
Subject: Re: [PATCH 4.19 29/74] clk: qcom: regmap-mux: fix parent clock lookup

On 15/12/2021 12:16, Pavel Machek wrote:
> Hi!
> 
>> The function mux_get_parent() uses qcom_find_src_index() to find the
>> parent clock index, which is incorrect: qcom_find_src_index() uses src
>> enum for the lookup, while mux_get_parent() should use cfg field (which
>> corresponds to the register value). Add qcom_find_cfg_index() function
>> doing this kind of lookup and use it for mux parent lookup.
> 
> This appears to have problems with error handling.
> 
>> +++ b/drivers/clk/qcom/clk-regmap-mux.c
>> @@ -36,7 +36,7 @@ static u8 mux_get_parent(struct clk_hw *
>>   	val &= mask;
>>   
>>   	if (mux->parent_map)
>> -		return qcom_find_src_index(hw, mux->parent_map, val);
>> +		return qcom_find_cfg_index(hw, mux->parent_map, val);
>>   
>>   	return val;
>>   }
> 
> So this returns u8.
> 
>> +int qcom_find_cfg_index(struct clk_hw *hw, const struct parent_map *map, u8 cfg)
>> +{
>> +	int i, num_parents = clk_hw_get_num_parents(hw);
>> +
>> +	for (i = 0; i < num_parents; i++)
>> +		if (cfg == map[i].cfg)
>> +			return i;
>> +
>> +	return -ENOENT;
>> +}
> 
> In case of error, -ENOENT will be cast to u8 in caller. I don't
> believe that is correct.

Unfortunately there is no way to return proper error code from 
clk_ops->get_parent() callback. However returning -ENOENT would 
translate to 254. Then clk_core_get_parent_by_index() would determine 
that there is no such parent and return NULL. A call to clk_set_parent 
would reparent the clock.

Returning some sensible default (e.g. 0) would be much worse, since then 
the clock subsystem would assume that the clock has correct parent. A 
call to clk_set_parent would always result in ops->set_parent() call, 
reparenting the clock correctly.

Most probably it would be correct to make ops->get_parent() return int 
instead of u8 (either an index or an -ERROR). However this was out of 
scope for this patch.

-- 
With best wishes
Dmitry