Re: [PATCH] mmc: core: Fix struct mmc_gpio corruption

[Marc Zyngier]

On 03/07/18 15:14, Fabio Estevam wrote:
> Hi Marc,
> 
> On Tue, Jul 3, 2018 at 11:11 AM, Marc Zyngier <marc.zyngier@xxxxxxx> wrote:
>> I recently realised that something was amiss with /proc/interrupts
>> on my cubietruck:
>>
>>  51:          0          0     GICv2  41 Level     mv64xxx_i2c
>>  59:       1047          0     GICv2 117 Level     eth0
>>  66:          2          0  sunxi_pio_edge   1 Edge   XXXXX
>>
>> where XXXXX has some non-ASCII characters, while I was expecting:
>>
>>  66:          0          0  sunxi_pio_edge   1 Edge      1c0f000.mmc cd
>>
>> It turns out that the cd_debounce_delay_ms field was recently to
>> struct mmc_gpio, and carefully placed at the one spot where it could
>> cause damage: right *after* the zero-sized array that contains the
>> associated label. This label ends up being passed to the IRQ
>> subsystem, and then overwritten by the cd_debounce_delay_ms field.
>>
>> Funky!
>>
>> Let's move that field to a less crowded location.
>>
>> Fixes: bfd694d5e21c ("mmc: core: Add tunable delay before detecting card after card is inserted")
>> Cc: Shawn Lin <shawn.lin@xxxxxxxxxxxxxx>
>> Cc: Linus Walleij <linus.walleij@xxxxxxxxxx>
>> Cc: Ulf Hansson <ulf.hansson@xxxxxxxxxx>
>> Signed-off-by: Marc Zyngier <marc.zyngier@xxxxxxx>
>> ---
>>  drivers/mmc/core/slot-gpio.c | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/drivers/mmc/core/slot-gpio.c b/drivers/mmc/core/slot-gpio.c
>> index ef05e0039378..1ccf9dc07395 100644
>> --- a/drivers/mmc/core/slot-gpio.c
>> +++ b/drivers/mmc/core/slot-gpio.c
>> @@ -27,8 +27,8 @@ struct mmc_gpio {
>>         bool override_cd_active_level;
>>         irqreturn_t (*cd_gpio_isr)(int irq, void *dev_id);
>>         char *ro_label;
>> -       char cd_label[0];
>>         u32 cd_debounce_delay_ms;
>> +       char cd_label[0];
> 
> Ulf has already applied a fix for this:
> https://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc.git/commit/?h=fixes&id=60feca8f59a0bbdee9545ab52775cad0171151ff
> 
> which is more robust as it will give a gcc error when cd_label gets
> moved in the future.

Ah, good. Thanks for the heads-up.

	M.
-- 
Jazz is not dead. It just smells funny...