Re: [PATCH 10/11] staging: greybus: change strncpy() to strscpy()

["Arnd Bergmann" <arnd@kernel.org>]

On Thu, Mar 28, 2024, at 16:00, Dan Carpenter wrote:
> On Thu, Mar 28, 2024 at 03:04:54PM +0100, Arnd Bergmann wrote:
>> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
>> ---
>> This is from randconfig testing with random gcc versions, a .config to
>> reproduce is at https://pastebin.com/r13yezkU
>> ---
>>  drivers/staging/greybus/fw-management.c | 4 ++--
>>  1 file changed, 2 insertions(+), 2 deletions(-)
>> 
>> diff --git a/drivers/staging/greybus/fw-management.c b/drivers/staging/greybus/fw-management.c
>> index 3054f084d777..35bfdd5f32d2 100644
>> --- a/drivers/staging/greybus/fw-management.c
>> +++ b/drivers/staging/greybus/fw-management.c
>> @@ -303,13 +303,13 @@ static int fw_mgmt_backend_fw_update_operation(struct fw_mgmt *fw_mgmt,
>>  	struct gb_fw_mgmt_backend_fw_update_request request;
>>  	int ret;
>>  
>> -	strncpy(request.firmware_tag, tag, GB_FIRMWARE_TAG_MAX_SIZE);
>> +	ret = strscpy(request.firmware_tag, tag, GB_FIRMWARE_TAG_MAX_SIZE);
>
> This needs to be strscpy_pad() or it risks an information leak.

Right, I think I misread the code thinking that the strncpy()
destination was user provided, but I see now that this copy is
from user-provided data into the stack, so the padding is indeed
stale stack data.

I could not find out whether this gets copied back to userspace,
but adding the padding is safer indeed.

>>  
>>  	/*
>>  	 * The firmware-tag should be NULL terminated, otherwise throw error and
>                                       ^^^^^^^^^^^^^^^^
> These comments are out of date.
>
>>  	 * fail.
>>  	 */
>> -	if (request.firmware_tag[GB_FIRMWARE_TAG_MAX_SIZE - 1] != '\0') {
>> +	if (ret == -E2BIG) {
>>  		dev_err(fw_mgmt->parent, "backend-update: firmware-tag is not NULL terminated\n");
>                                                           
> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> More out of date prints.

I had thought about changing it when I did the patch, but could
not come up with anything that describes the error condition better:
the cause of the -E2BIG error is still the missing NUL-termination
in the provided string.

Maybe we should instead not print a warning at all? The general
rule is that user triggered operations should not lead to
spamming the kernel logs.

     Arnd