From: Dan Carpenter <dan.carpenter@oracle.com>
To: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
Cc: linux-s390@vger.kernel.org, smatch@vger.kernel.org
Subject: Re: smatch and copy_{to,from}_user return values
Date: Wed, 3 Mar 2021 14:20:46 +0300 [thread overview]
Message-ID: <20210303112046.GB2222@kadam> (raw)
In-Reply-To: <b57b4f40-d67c-d57c-c5b2-077b623ed4ed@prevas.dk>
[-- Attachment #1: Type: text/plain, Size: 1983 bytes --]
On Wed, Mar 03, 2021 at 08:50:19AM +0100, Rasmus Villemoes wrote:
> Hi Dan
>
> If you look at vfio_ccw_mdev_ioctl() in drivers/s390/cio/vfio_ccw_ops.c,
> and vfio_ap_mdev_get_device_info() in drivers/s390/crypto/vfio_ap_ops.c,
> there are examples of functions that can both return -Esomething as well
> as may return the return value of a copy_{to,from}_user directly (i.e.,
> in case of error some positive number).
>
> [Those "return copy_to_user();" should probably all be changed to
> "return copy_to_user() ? -EFAULT : 0;" - cc'ing the s390 list in case
> the maintainers want to do that.]
>
> Can smatch detect such cases? I seem to recall it has some concept of
> tagging a function as "returning -Efoo or 0", so it would also need to
> know that copy_{to,from}_user does not return -Efoo. And it also needs
> to follow the control flow, so
>
> ret = copy_to_user();
> if (ret)
> return -EIO;
> something_else;
> return ret; /* this is 0 */
>
> doesn't trigger. And there's gonna be some false positives around signal
> frame setup, which do a lot of "err |= foo(); err |= bar()" where foo()
> report errors as -Exxx and bar can be a copy_to_user(), but in the end
> err is only checked against 0.
>
> Rasmus
Yeah. There is already a check for if you propagate the return from
copy_from_user()... The problem is that this is s390 code and I don't
have a cross compiler set up so this was never reported or fixed.
When I first saw your email, I didn't read it carefully and I thought
you were complaining about code that returns -EIO where -EFAULT is
intended. Anyway, I wrote that check before re-reading the email. LOL.
Attached.
I did a quick "git grep |= copy_" and I see that's mostly used in
signal code where the caller doesn't care about the error code, only
whether it's zero vs non-zero. I considered about excluding "arch/"
from the check but then there are only two instances where this is used
and both are correct.
regards,
dan carpenter
[-- Attachment #2: check_return_efault2.c --]
[-- Type: text/x-csrc, Size: 2096 bytes --]
/*
* Copyright (C) 2021 Oracle.
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version 2
* of the License, or (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, see http://www.gnu.org/copyleft/gpl.txt
*/
#include "smatch.h"
#include "smatch_extra.h"
static int my_id;
static const sval_t ulong_one = { .type = &ulong_ctype, .value = 1 };
static const sval_t ulong_INT_MAX = { .type = &ulong_ctype, .value = INT_MAX };
STATE(copy_failed);
static void match_copy_failed(const char *fn, struct expression *call_expr,
struct expression *expr, void *_unused)
{
set_state(my_id, "path", NULL, ©_failed);
}
static void match_return(struct expression *expr)
{
char *macro = NULL;
sval_t ret;
if (!get_value(expr, &ret))
return;
if (ret.value == -14)
return;
if (get_state(my_id, "path", NULL) != ©_failed)
return;
if (expr->type == EXPR_PREOP && expr->op == '-')
macro = get_macro_name(expr->unop->pos);
if (macro)
sm_warning("return -EFAULT instead of '-%s'", macro);
else
sm_warning("return -EFAULT instead of '%s'", sval_to_str(ret));
}
void check_return_efault2(int id)
{
if (option_project != PROJ_KERNEL)
return;
my_id = id;
return_implies_state_sval("copy_to_user", ulong_one, ulong_INT_MAX, &match_copy_failed, NULL);
return_implies_state_sval("copy_from_user", ulong_one, ulong_INT_MAX, &match_copy_failed, NULL);
return_implies_state_sval("__copy_to_user", ulong_one, ulong_INT_MAX, &match_copy_failed, NULL);
return_implies_state_sval("__copy_from_user", ulong_one, ulong_INT_MAX, &match_copy_failed, NULL);
add_hook(&match_return, RETURN_HOOK);
}
next prev parent reply other threads:[~2021-03-03 22:55 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-03-03 7:50 smatch and copy_{to,from}_user return values Rasmus Villemoes
2021-03-03 11:20 ` Dan Carpenter [this message]
2021-03-05 10:14 ` Dan Carpenter
2021-03-10 10:01 ` Dan Carpenter
2021-03-04 18:35 ` Heiko Carstens
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20210303112046.GB2222@kadam \
--to=dan.carpenter@oracle.com \
--cc=linux-s390@vger.kernel.org \
--cc=rasmus.villemoes@prevas.dk \
--cc=smatch@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox