LinuxLists.cc - [PATCH] ASoC: da7219: Fix an out-of-bound read in an error handling path

2021-06-23 05:24:50

Subject: [PATCH] ASoC: da7219: Fix an out-of-bound read in an error handling path

If 'of_clk_add_hw_provider()' fails, the previous 'for' loop will have
run completely and 'i' is know to be 'DA7219_DAI_NUM_CLKS'.

In such a case, there will be an out-of-bounds access when using
'da7219->dai_clks_lookup[i]' and '&da7219->dai_clks_hw[i]'.

To avoid that, add a new label, 'err_free_all', which set the expected
value of 'i' in such a case.

Fixes: 78013a1cf297 ("ASoC: da7219: Fix clock handling around codec level probe")
Signed-off-by: Christophe JAILLET <[email protected]>
---
sound/soc/codecs/da7219.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/sound/soc/codecs/da7219.c b/sound/soc/codecs/da7219.c
index 13009d08b09a..1e8b491d1fd3 100644
--- a/sound/soc/codecs/da7219.c
+++ b/sound/soc/codecs/da7219.c
@@ -2204,12 +2204,14 @@ static int da7219_register_dai_clks(struct snd_soc_component *component)
da7219->clk_hw_data);
if (ret) {
dev_err(dev, "Failed to register clock provider\n");
- goto err;
+ goto err_free_all;
}
}

return 0;

+err_free_all:
+ i = DA7219_DAI_NUM_CLKS - 1;
err:
do {
if (da7219->dai_clks_lookup[i])
--
2.30.2

2021-06-23 09:49:53

by Dan Carpenter

[permalink] [raw]

Subject: Re: [PATCH] ASoC: da7219: Fix an out-of-bound read in an error handling path

On Wed, Jun 23, 2021 at 07:22:45AM +0200, Christophe JAILLET wrote:
> If 'of_clk_add_hw_provider()' fails, the previous 'for' loop will have
> run completely and 'i' is know to be 'DA7219_DAI_NUM_CLKS'.
>
> In such a case, there will be an out-of-bounds access when using
> 'da7219->dai_clks_lookup[i]' and '&da7219->dai_clks_hw[i]'.
>
> To avoid that, add a new label, 'err_free_all', which set the expected
> value of 'i' in such a case.
>
> Fixes: 78013a1cf297 ("ASoC: da7219: Fix clock handling around codec level probe")
> Signed-off-by: Christophe JAILLET <[email protected]>
> ---
> sound/soc/codecs/da7219.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/sound/soc/codecs/da7219.c b/sound/soc/codecs/da7219.c
> index 13009d08b09a..1e8b491d1fd3 100644
> --- a/sound/soc/codecs/da7219.c
> +++ b/sound/soc/codecs/da7219.c
> @@ -2204,12 +2204,14 @@ static int da7219_register_dai_clks(struct snd_soc_component *component)
> da7219->clk_hw_data);
> if (ret) {
> dev_err(dev, "Failed to register clock provider\n");
> - goto err;
> + goto err_free_all;
> }
> }
>
> return 0;
>
> +err_free_all:
> + i = DA7219_DAI_NUM_CLKS - 1;
> err:
> do {
> if (da7219->dai_clks_lookup[i])

This do while statement is wrong and it leads to potentially calling
clk_hw_unregister() on clks that haven't been registered.

I think that calling clk_hw_unregister() on unregistered clocks is
supposed to okay but I found a case where it leads to a WARN_ON()
(Nothing else harmful). It's in __clk_register() if the alloc_clk()
fails:

hw->clk = alloc_clk(core, NULL, NULL);
if (IS_ERR(hw->clk)) {
ret = PTR_ERR(hw->clk);
goto fail_create_clk; // <- forgot to set hw->clk = NULL
}

The better way to handle errors from loops is to clean up partial
iterations before doing the goto. So add a clk_hw_unregister() if the
dai_clk_lookup = clkdev_hw_create() assignment fails. Then use a
while (--i >= 0) loop in the unwind section:

err_free_all:
i = DA7219_DAI_NUM_CLKS;
err:
while (--i >= 0) {

regards,
dan carpenter

2021-06-23 21:15:18

by Christophe JAILLET

[permalink] [raw]

Subject: Re: [PATCH] ASoC: da7219: Fix an out-of-bound read in an error handling path

Le 23/06/2021 à 11:46, Dan Carpenter a écrit :
> On Wed, Jun 23, 2021 at 07:22:45AM +0200, Christophe JAILLET wrote:
>> If 'of_clk_add_hw_provider()' fails, the previous 'for' loop will have
>> run completely and 'i' is know to be 'DA7219_DAI_NUM_CLKS'.
>>
>> In such a case, there will be an out-of-bounds access when using
>> 'da7219->dai_clks_lookup[i]' and '&da7219->dai_clks_hw[i]'.
>>
>> To avoid that, add a new label, 'err_free_all', which set the expected
>> value of 'i' in such a case.
>>
>> Fixes: 78013a1cf297 ("ASoC: da7219: Fix clock handling around codec level probe")
>> Signed-off-by: Christophe JAILLET <[email protected]>
>> ---
>> sound/soc/codecs/da7219.c | 4 +++-
>> 1 file changed, 3 insertions(+), 1 deletion(-)
>>
>> diff --git a/sound/soc/codecs/da7219.c b/sound/soc/codecs/da7219.c
>> index 13009d08b09a..1e8b491d1fd3 100644
>> --- a/sound/soc/codecs/da7219.c
>> +++ b/sound/soc/codecs/da7219.c
>> @@ -2204,12 +2204,14 @@ static int da7219_register_dai_clks(struct snd_soc_component *component)
>> da7219->clk_hw_data);
>> if (ret) {
>> dev_err(dev, "Failed to register clock provider\n");
>> - goto err;
>> + goto err_free_all;
>> }
>> }
>>
>> return 0;
>>
>> +err_free_all:
>> + i = DA7219_DAI_NUM_CLKS - 1;
>> err:
>> do {
>> if (da7219->dai_clks_lookup[i])
>
> This do while statement is wrong and it leads to potentially calling
> clk_hw_unregister() on clks that haven't been registered.

Obviously right.

Thanks for the review Dan.
I'll send a v2 in the coming days.

CJ
>
> I think that calling clk_hw_unregister() on unregistered clocks is
> supposed to okay but I found a case where it leads to a WARN_ON()
> (Nothing else harmful). It's in __clk_register() if the alloc_clk()
> fails:
>
> hw->clk = alloc_clk(core, NULL, NULL);
> if (IS_ERR(hw->clk)) {
> ret = PTR_ERR(hw->clk);
> goto fail_create_clk; // <- forgot to set hw->clk = NULL
> }
>
> The better way to handle errors from loops is to clean up partial
> iterations before doing the goto. So add a clk_hw_unregister() if the
> dai_clk_lookup = clkdev_hw_create() assignment fails. Then use a
> while (--i >= 0) loop in the unwind section:
>
> err_free_all:
> i = DA7219_DAI_NUM_CLKS;
> err:
> while (--i >= 0) {
>
> regards,
> dan carpenter
>
>

2021-06-23 21:16:37

by Christophe JAILLET

[permalink] [raw]