[lkml]   [2011]   [Jun]   [12]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [PATCH] x86, vsyscall: Fix build warning in vsyscall_64.c
On Sun, Jun 12, 2011 at 1:12 AM, Rakib Mullick <> wrote:
> On Sat, Jun 11, 2011 at 5:01 PM, Andrew Lutomirski <> wrote:
>> On Sat, Jun 11, 2011 at 3:31 AM, Rakib Mullick <> wrote:
>>> Due to commit 5cec93c216db77 (x86-64: Emulate legacy vsyscalls), we get the following warning:
>>>   arch/x86/kernel/vsyscall_64.c: In function ‘do_emulate_vsyscall’:
>>>   arch/x86/kernel/vsyscall_64.c:111:7: warning: ‘ret’ may be used uninitialized in this function
>> What's the code path that uses ret without initializing it?
> In case of, vsyscall_nr is default it might gets uninitialized. And
> current code already treats it as a bug.
>>> -       if (ret == -EFAULT) {
>>> +       if (ret == -EFAULT || ret == -EINVAL) {
>>>                /*
>>>                 * Bad news -- userspace fed a bad pointer to a vsyscall.
>>>                 *
>> EINVAL doesn't seem like grounds to fault.  (I'm not sure how to get
>> EINVAL from time, gettimeofday, or getcpu, but in case there is, we
>> should return it back to userspace.)
> If ret = EINVAL, then it means vsyscall_nr doesn't any of
> gettimeofday, time or getcpu. So, I grounds it into fault. In case of
> gettimeofday, EINVAL may gets return. But, maybe not in case of time
> or getcpu. So, maybe we need to check EINVAL in case of gettimeofday
> and maybe should separate EINVAL and EFAULT.

I think there are three separate issues here:

1. Can ret be used uninitialized? I say no, even as seen by the
compiler. If vsyscall_nr is 0, 1, or 2, then ret is initialized. If
vsyscall_nr is 3, then the BUG gets hit. BUG is defined as some
assembly magic followed by unreachable(), and the compiler is supposed
to know that code after unreachable() is qunreachable. So how can ret
be used uninitialized?

What version of gcc do you have? gcc (GCC) 4.6.0 20110530 (Red Hat
4.6.0-9) does not produce this warning.

2. Is the BUG correct? I say yes. vsyscall_nr can only be 0, 1, 2,
or 3 (see the function that generates it), and the only way that 3
could happen is if regs->ip == 0xffffffffff600c02. That can't happen
because the instruction at ...601 is int3.

3. Should the test for EFAULT be changed to EINVAL? I can't see why.
We need to preserve userspace ABI, and userspace expects vsyscalls
that fail for reasons other than a fault to return an error, not
segfault the caller.

Note that regs->as *is* the return value, so we're not ignoring errors.

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
More majordomo info at
Please read the FAQ at

 \ /
  Last update: 2011-06-13 04:55    [W:0.122 / U:4.752 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site