linux-kernel - Re: [PATCH] fanotify: prove that structure size is correct

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] [day] [month] [year] [list]

Date:	Thu, 02 Oct 2014 20:56:37 +0200
From:	Heinrich Schuchardt <xypron.glpk@....de>
To:	Kees Cook <keescook@...omium.org>, linux-kernel@...r.kernel.org
CC:	Eric Paris <eparis@...hat.com>, Jan Kara <jack@...e.cz>
Subject: Re: [PATCH] fanotify: prove that structure size is correct


On 29.09.2014 01:35, Kees Cook wrote:
> When building with CONFIG_DEBUG_STRICT_USER_COPY_CHECKS, this copy_to_user
> wasn't provably correct (the length was in a variable but the structure
> wasn't). The compiler wasn't able to check that the length was being
> statically assigned, so include a BUG_ON to notice if this ever changes.
>
> Signed-off-by: Kees Cook <keescook@...omium.org>
> ---
>   fs/notify/fanotify/fanotify_user.c | 1 +
>   1 file changed, 1 insertion(+)
>
> diff --git a/fs/notify/fanotify/fanotify_user.c b/fs/notify/fanotify/fanotify_user.c
> index b13992a41bd9..7bdb43222875 100644
> --- a/fs/notify/fanotify/fanotify_user.c
> +++ b/fs/notify/fanotify/fanotify_user.c
> @@ -216,6 +216,7 @@ static ssize_t copy_event_to_user(struct fsnotify_group *group,
>
>   	fd = fanotify_event_metadata.fd;
>   	ret = -EFAULT;
> +	BUG_ON(fanotify_event_metadata.event_len != FAN_EVENT_METADATA_LEN);


Thank you for indicating an inconsistency in fanotify_user.c

In fanotify_user.c a mismatch exists between the assumptions made in 
get_one_event() and copy_event_to_user().

According to the fanotify.7 man page:
"FAN_EVENT_METADATA_LEN - This is the minimum size (and currently the 
only size) of any event metadata."

get_one_event() assumes that the event metadata has a length of 
FAN_EVENT_METADATA_LEN and does not care about the value of event_len.

copy_event_to_user() assumes that event_len specifies the length of the 
event metadata and does not care about FAN_EVENT_METADATA_LEN.

Due to these different assumptions it cannot be proven inside 
fanotify_user.c that copy_to_user() is called correctly.

If we want such a prove inside fanotify_user.c, we should change 
get_one_event() to use the following:

// Get the event without removing it from the group.
fsnotify_event *evt = fsnotify_peek_notify_event(group);
// Check we have enough user memory to copy the event.
if (evt->event_len > count)
     return ERR_PTR(-EINVAL);
// Remove the event from the group.
fsnotify_remove_notify_event(group);
// Return the pointer we tested.
// fanotify_remove_notify_event() should return the same pointer.
return evt;

If after this change you still want a BUG_ON, pass count to 
copy_event_to_user() and write
BUG_ON(fanotify_event_metadata.event_len > count);

Best regards

Heinrich Schuchardt

>   	if (copy_to_user(buf, &fanotify_event_metadata,
>   			 fanotify_event_metadata.event_len))
>   		goto out_close_fd;
>

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