[linux-cifs-client] [PATCH] cifs: Fix insufficient memory allocation for nativeFileSystem field

[Suresh Jayaraman]

Jeff Layton wrote:
> On Tue, 07 Apr 2009 18:45:46 +0530
> Suresh Jayaraman <sjayaraman at suse.de> wrote:
> 
>> Do we still need this conversion again?
>>
> 
> I know this isn't a "real" patch submission yet, but some comments
> below...
> 
>> diff --git a/fs/cifs/cifs_unicode.h b/fs/cifs/cifs_unicode.h
>> index 14eb9a2..0396bdc 100644
>> --- a/fs/cifs/cifs_unicode.h
>> +++ b/fs/cifs/cifs_unicode.h
>> @@ -159,6 +159,23 @@ UniStrnlen(const wchar_t *ucs1, int maxlen)
>>  }
>>
>>  /*
>> + * UniStrnlenBytes: Return the length in bytes of a UTF-8 string
>> + */
>> +static inline size_t
>> +UniStrnlenBytes(const unsigned char *str, int maxlen)
>> +{
>> +       size_t nbytes = 0;
>> +       wchar_t *uni;
> 		^^^^^
> I think you need to allocate actual storage for the character here.
> 

Oh, wait no storage is required I think. We reuse the pointer to wchar_t
and it seems char2uni() returns 1 or -EINVAL
and I was thinking strlen() on a wide char returns length in bytes?

fs/nls/nls_base.c

static int char2uni(const unsigned char *rawstring, int boundlen,
wchar_t *uni)
{
        *uni = charset2uni[*rawstring];
        if (*uni == 0x0000)
                return -EINVAL;
        return 1;
}

but, strnlen() may not work, though..

or am i looking at the wrong place..


>> +
>> +       while (*str++) {
>> +               /* convert each char, find its length and add to nbytes */
>> +               if (char2uni(str, maxlen, uni) > 0)
>> +                       nbytes += strnlen(uni, NLS_MAX_CHARSET_SIZE);
> 
> "uni" is a ptr to a wchar_t, but you're treating it as a string.
> There's no guarantee that it'll be null-terminated. I might be
> mistaken, but doesn't char2uni return the length of the converted
> character in bytes? Tallying up the return from those is probably
> the thing to do.
> 
>> +       }
>> +       return nbytes;
>> +}
>> +
>> +/*
> 
> 


-- 
Suresh Jayaraman