On 2024-11-14 18:50, Luiz Augusto von Dentz wrote:
> Hi David,
>
> On Thu, Nov 14, 2024 at 9:30 PM David Wei <dw@xxxxxxxxxxx> wrote:
>>
>> On 2024-11-14 18:15, Luiz Augusto von Dentz wrote:
>>> Hi David,
>>>
>>> On Thu, Nov 14, 2024 at 7:42 PM David Wei <dw@xxxxxxxxxxx> wrote:
>>>>
>>>> On 2024-11-14 15:27, Michal Luczaj wrote:
>>>>> The bt_copy_from_sockptr() return value is being misinterpreted by most
>>>>> users: a non-zero result is mistakenly assumed to represent an error code,
>>>>> but actually indicates the number of bytes that could not be copied.
>>>>>
>>>>> Remove bt_copy_from_sockptr() and adapt callers to use
>>>>> copy_safe_from_sockptr().
>>>>>
>>>>> For sco_sock_setsockopt() (case BT_CODEC) use copy_struct_from_sockptr() to
>>>>> scrub parts of uninitialized buffer.
>>>>>
>>>>> Opportunistically, rename `len` to `optlen` in hci_sock_setsockopt_old()
>>>>> and hci_sock_setsockopt().
>>>>>
>>>>> Fixes: 51eda36d33e4 ("Bluetooth: SCO: Fix not validating setsockopt user input")
>>>>> Fixes: a97de7bff13b ("Bluetooth: RFCOMM: Fix not validating setsockopt user input")
>>>>> Fixes: 4f3951242ace ("Bluetooth: L2CAP: Fix not validating setsockopt user input")
>>>>> Fixes: 9e8742cdfc4b ("Bluetooth: ISO: Fix not validating setsockopt user input")
>>>>> Fixes: b2186061d604 ("Bluetooth: hci_sock: Fix not validating setsockopt user input")
>>>>> Signed-off-by: Michal Luczaj <mhal@xxxxxxx>
>>>>> ---
>>>>> include/net/bluetooth/bluetooth.h | 9 ---------
>>>>> net/bluetooth/hci_sock.c | 14 +++++++-------
>>>>> net/bluetooth/iso.c | 10 +++++-----
>>>>> net/bluetooth/l2cap_sock.c | 20 +++++++++++---------
>>>>> net/bluetooth/rfcomm/sock.c | 9 ++++-----
>>>>> net/bluetooth/sco.c | 11 ++++++-----
>>>>> 6 files changed, 33 insertions(+), 40 deletions(-)
>>>>>
>>>> ...
>>>>> diff --git a/net/bluetooth/rfcomm/sock.c b/net/bluetooth/rfcomm/sock.c
>>>>> index f48250e3f2e103c75d5937e1608e43c123aa3297..1001fb4cc21c0ecc7bcdd3ea9041770ede4f27b8 100644
>>>>> --- a/net/bluetooth/rfcomm/sock.c
>>>>> +++ b/net/bluetooth/rfcomm/sock.c
>>>>> @@ -629,10 +629,9 @@ static int rfcomm_sock_setsockopt_old(struct socket *sock, int optname,
>>>>>
>>>>> switch (optname) {
>>>>> case RFCOMM_LM:
>>>>> - if (bt_copy_from_sockptr(&opt, sizeof(opt), optval, optlen)) {
>>>>> - err = -EFAULT;
>>>>> + err = copy_safe_from_sockptr(&opt, sizeof(opt), optval, optlen);
>>>>> + if (err)
>>>>> break;
>>>>> - }
>>>>
>>>> This will return a positive integer if copy_safe_from_sockptr() fails.
>>>
>>> What are you talking about copy_safe_from_sockptr never returns a
>>> positive value:
>>>
>>> * Returns:
>>> * * -EINVAL: @optlen < @ksize
>>> * * -EFAULT: access to userspace failed.
>>> * * 0 : @ksize bytes were copied
>>
>> Isn't this what this series is about? copy_from_sockptr() returns 0 on
>> success, or a positive integer for number of bytes NOT copied on error.
>> Patch 4 even updates the docs for copy_from_sockptr().
>>
>> copy_safe_from_sockptr()
>> -> copy_from_sockptr()
>> -> copy_from_sockptr_offset()
>> -> memcpy() for kernel to kernel OR
>> -> copy_from_user() otherwise
>
> Well except the safe version does check what would otherwise cause a
> positive return by the likes of copy_from_user and returns -EINVAL
> instead, otherwise the documentation of copy_safe_from_sockptr is just
> wrong and shall state that it could return positive as well but I
> guess that would just make it as inconvenient so we might as well
> detect when a positive value would be returned just return -EFAULT
> instead.
Yes it checks and returns EINVAL, but not EFAULT which is what my
comment on the original patch is about. Most of the calls to
bt_copy_from_sockptr() that Michal replaced with
copy_safe_from_sockptr() remain incorrect because it is assumed that
EFAULT is returned. Only rfcomm_sock_setsockopt_old() was vaguely doing
the right thing and the patch changed it back to the incorrect pattern:
err = copy_safe_from_sockptr(...);
if (err)
break;
But I do agree that making copy_safe_from_sockptr() do the right thing
and EFAULT will be easier and prevent future problems given that
copy_from_sockptr() is meant to be deprecated anyhow.
>
>> And copy_from_user() follows the same 0 for success or N > 0 for
>> failure. It does not EFAULT on its own AFAIK.
>>
>> The docs for copy_safe_from_sockptr() that you've linked contains the
>> exact misunderstanding that Michal is correcting.
>>
>>>
>>>> Shouldn't this be:
>>>>
>>>> err = -EFAULT;
>>>> if (copy_safe_from_sockptr(&opt, sizeof(opt), optval, optlen))
>>>> break;
>>>
>>>
>>>
>
>
>
