linux-kernel - Re: [PATCH v3 5/5] fs: unicode: Add utf8 module and a unicode layer

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <YFpPxCQiMLqctIuS@gmail.com>
Date:   Tue, 23 Mar 2021 13:29:56 -0700
From:   Eric Biggers <ebiggers@...nel.org>
To:     Gabriel Krisman Bertazi <krisman@...labora.com>
Cc:     Shreeya Patel <shreeya.patel@...labora.com>, tytso@....edu,
        adilger.kernel@...ger.ca, jaegeuk@...nel.org, chao@...nel.org,
        drosen@...gle.com, yuchao0@...wei.com, linux-ext4@...r.kernel.org,
        linux-kernel@...r.kernel.org,
        linux-f2fs-devel@...ts.sourceforge.net,
        linux-fsdevel@...r.kernel.org, kernel@...labora.com,
        andre.almeida@...labora.com
Subject: Re: [PATCH v3 5/5] fs: unicode: Add utf8 module and a unicode layer

On Tue, Mar 23, 2021 at 03:51:44PM -0400, Gabriel Krisman Bertazi wrote:
> > -int unicode_validate(const struct unicode_map *um, const struct qstr *str)
> > -{
> > -	const struct utf8data *data = utf8nfdi(um->version);
> > -
> > -	if (utf8nlen(data, str->name, str->len) < 0)
> > -		return -1;
> > -	return 0;
> > -}
> > +struct unicode_ops *utf8_ops;
> > +EXPORT_SYMBOL(utf8_ops);
> > +
> > +int _utf8_validate(const struct unicode_map *um, const struct qstr *str)
> > +{
> > +	return 0;
> > +}
> > -EXPORT_SYMBOL(unicode_validate);
> 
> I think that any calls to the default static calls should return errors
> instead of succeeding without doing anything.
> 
> In fact, are the default calls really necessary?  If someone gets here,
> there is a bug elsewhere, so WARN_ON and maybe -EIO.  
> 
> int unicode_validate_default_static_call(...)
> {
>    WARN_ON(1);
>    return -EIO;
> }
> 
> Or just have a NULL default, as I mentioned below, if that is possible.
> 
[...]
> > +DEFINE_STATIC_CALL(utf8_validate, _utf8_validate);
> > +DEFINE_STATIC_CALL(utf8_strncmp, _utf8_strncmp);
> > +DEFINE_STATIC_CALL(utf8_strncasecmp, _utf8_strncasecmp);
> > +DEFINE_STATIC_CALL(utf8_strncasecmp_folded, _utf8_strncasecmp_folded);
> > +DEFINE_STATIC_CALL(utf8_normalize, _utf8_normalize);
> > +DEFINE_STATIC_CALL(utf8_casefold, _utf8_casefold);
> > +DEFINE_STATIC_CALL(utf8_casefold_hash, _utf8_casefold_hash);
> > +DEFINE_STATIC_CALL(utf8_load, _utf8_load);
> > +DEFINE_STATIC_CALL_NULL(utf8_unload, _utf8_unload);
> > +EXPORT_STATIC_CALL(utf8_strncmp);
> > +EXPORT_STATIC_CALL(utf8_strncasecmp);
> > +EXPORT_STATIC_CALL(utf8_strncasecmp_folded);
> 
> I'm having a hard time understanding why some use
> DEFINE_STATIC_CALL_NULL, while other use DEFINE_STATIC_CALL.  This new
> static call API is new to me :).  None of this can be called if the
> module is not loaded anyway, so perhaps the default function can just be
> NULL, per the documentation of include/linux/static_call.h?
> 
> Anyway, Aren't utf8_{validate,casefold,normalize} missing the
> equivalent EXPORT_STATIC_CALL?
> 

The static_call API is fairly new to me too.  But the intent of this patch seems
to be that none of the utf8 functions are called without the utf8 module loaded.
If they are called, it's a kernel bug.  So there are two options for what to do
if it happens anyway:

  1. call a "null" static call, which does nothing

*or*

  2. call a default function which does WARN_ON_ONCE() and returns an error if
     possible.

(or 3. don't use static calls and instead dereference a NULL utf8_ops like
previous versions of this patch did.)

It shouldn't really matter which of these approaches you take, but please be
consistent and use the same one everywhere.

> + void unicode_unregister(void)
> + {
> +         spin_lock(&utf8ops_lock);
> +         utf8_ops = NULL;
> +         spin_unlock(&utf8ops_lock);
> + }
> + EXPORT_SYMBOL(unicode_unregister);

This should restore the static calls to their default values (either NULL or the
default functions, depending on what you decide).

Also, it's weird to still have the utf8_ops structure when using static calls.
It seems it should be one way or the other: static calls *or* utf8_ops.

The static calls could be exported, and the module could be responsible for
updating them.  That would eliminate the need for utf8_ops.

- Eric