From: Rusty Russell <rusty@rustcorp.com.au>
To: Chen Gang <gang.chen@asianux.com>
Cc: "linux-kernel\@vger.kernel.org" <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH] kernel: module: strncpy issue, using strlcpy instead of strncpy
In-Reply-To: <516298E0.9000905@asianux.com>
References: <51615AB0.9060502@asianux.com> <87r4ila8sb.fsf@rustcorp.com.au> <516298E0.9000905@asianux.com>
User-Agent: Notmuch/0.14 (http://notmuchmail.org) Emacs/23.4.1 (i686-pc-linux-gnu)
Date: Mon, 08 Apr 2013 23:15:55 +0930
Message-ID: <8738v19lv0.fsf@rustcorp.com.au>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 8BIT
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 2948
Lines: 85

Chen Gang <gang.chen@asianux.com> writes:
> On 2013年04月08日 13:30, Rusty Russell wrote:
>> Chen Gang <gang.chen@asianux.com> writes:
>>> >   ownername and namebuf are all NUL terminated string.
>>> >
>>> >   need always let them ended by '\0'.
>>> >
>>> > Signed-off-by: Chen Gang <gang.chen@asianux.com>
>>> > ---
>>> >  kernel/module.c |    4 ++--
>>> >  1 files changed, 2 insertions(+), 2 deletions(-)
>>> >
>>> > diff --git a/kernel/module.c b/kernel/module.c
>>> > index 3c2c72d..597efd8 100644
>>> > --- a/kernel/module.c
>>> > +++ b/kernel/module.c
>>> > @@ -1283,7 +1283,7 @@ static const struct kernel_symbol *resolve_symbol(struct module *mod,
>>> >  
>>> >  getname:
>>> >  	/* We must make copy under the lock if we failed to get ref. */
>>> > -	strncpy(ownername, module_name(owner), MODULE_NAME_LEN);
>>> > +	strlcpy(ownername, module_name(owner), MODULE_NAME_LEN);
>> This should just be strcpy().
>> 
>
>   for me, either strcpy or strlcpy are ok.
>     strcpy is quicker than strlcpy (in our case, it seems not quite important).
>     strlcpy is more clearer to readers (they do not care about the buffer length again).
>
>   since you prefer strcpy, I need respect your (the original maintainer's) willing.
>   so I need change to strcpy.
>
>   :-)
>
>
>>> >  unlock:
>>> >  	mutex_unlock(&module_mutex);
>>> >  	return sym;
>>> > @@ -3464,7 +3464,7 @@ const char *module_address_lookup(unsigned long addr,
>>> >  	}
>>> >  	/* Make a copy in here where it's safe */
>>> >  	if (ret) {
>>> > -		strncpy(namebuf, ret, KSYM_NAME_LEN - 1);
>>> > +		strlcpy(namebuf, ret, KSYM_NAME_LEN);
>> This isn't a bug, because the caller (kallsyms_lookup) puts a NUL in
>> ret[KSYM_NAME_LEN].
>> 
>
>   originally, it is really not a bug (so subject need delete "strncpy issue").
>   now, I still prefer to set tail '\0' in function module_address_lookup.
>   future, if it is used by others, it is necessary to set tail '\0' in this function.
>
>
>
> and for this patch, is it suitable to send patch v2 ?

Yes, that's fine.

>> However, kallsyms_lookup also calls kallsyms_expand_symbol, which
>> doesn't stop at KSYM_NAME_LEN, so if a name was longer than that we'd
>> have a real bug.
>> 
>> Would you like to take a look at that, too?
>> 
>
>   it looks like a bug.  for me, I prefer to give length check for it.
>
>   but I am sorry, now, I can not be sure whether it is really a bug.

It really is.  We don't export any symbols > 128 characters, but if we
did then kallsyms_expand_symbol() would overflow the buffer handed to
it.

Your suggestion about an explicit length for kallsyms_expand_symbol() is
the correct one.

(This is a separate patch to the cleanup above).

Thanks,
Rusty.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/