Re: [PATCH] linux-user: allow NULL msg in recvfrom

[Laurent Vivier <laurent@vivier.eu>]

Le 26/03/2021 à 14:28, Peter Maydell a écrit :
> On Fri, 26 Mar 2021 at 13:24, Laurent Vivier <laurent@vivier.eu> wrote:
>>
>> Le 26/03/2021 à 05:05, Zach Reizner a écrit :
>>> The kernel allows a NULL msg in recvfrom so that he size of the next
>>> message may be queried before allocating a correctly sized buffer. This
>>> change allows the syscall translator to pass along the NULL msg pointer
>>> instead of returning early with EFAULT.
>>>
>>> Signed-off-by: Zach Reizner <zachr@google.com>
>>> ---
>>>  linux-user/syscall.c | 2 --
>>>  1 file changed, 2 deletions(-)
>>>
>>> diff --git a/linux-user/syscall.c b/linux-user/syscall.c
>>> index 1e508576c7..332544b43c 100644
>>> --- a/linux-user/syscall.c
>>> +++ b/linux-user/syscall.c
>>> @@ -3680,8 +3680,6 @@ static abi_long do_recvfrom(int fd, abi_ulong
>>> msg, size_t len, int flags,
>>>      abi_long ret;
>>>
>>>      host_msg = lock_user(VERIFY_WRITE, msg, len, 0);
>>> -    if (!host_msg)
>>> -        return -TARGET_EFAULT;
>>>      if (target_addr) {
>>>          if (get_user_u32(addrlen, target_addrlen)) {
>>>              ret = -TARGET_EFAULT;
>>>
>>
>> Applied to my linux-user-for-6.0 branch
> 
> Doesn't this mean we'll now incorrectly treat "guest passed
> a bad address" the same as "guest passed NULL" ? lock_user()
> returns NULL for errors, so if you need to handle NULL input
> specially you want something like
> 
>    if (!msg) {
>        host_msg = NULL;
>    } else {
>        host_msg = lock_user(VERIFY_WRITE, msg, len, 0);
>        if (!host_msg) {
>            return -TARGET_EFAULT;
>        }
>    }
> 
> I think ?

Yes, you're right.

Zach, could you update your patch?

Thanks,
Laurent