linux-kernel - Re: [PATCH] staging: android: logger: fix kuid/uid in logger

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]

Message-ID: <3175901.1lz7PhpNtK@wuerfel>
Date:	Thu, 30 Oct 2014 08:04:44 +0100
From:	Arnd Bergmann <arnd@...db.de>
To:	Xiong Zhou <jencce.kernel@...il.com>
Cc:	linux-kernel@...r.kernel.org, kernel-team@...roid.com,
	nnk@...gle.com, john.stultz@...aro.org, gregkh@...uxfoundation.org,
	devel@...verdev.osuosl.org
Subject: Re: [PATCH] staging: android: logger: fix kuid/uid in logger_entry

On Thursday 30 October 2014 10:23:31 Xiong Zhou wrote:
> From: Xiong Zhou <jencce.kernel@...il.com>
> 
> struct logger_entry can be returned to userspace via ioctl,
> so it is wrong to have a kuid_t member, fixing it to uid_t.
> This was introduced by commit bd471258f2, to pass uidguid
> type checks : UIDGID_STRICT_TYPE_CHECKS, which has been
> removed from kernel in commit 261000a56b6.
> 
> Fixes: bd471258f2 (logger: use kuid_t instead of uid_t)
> Signed-off-by: Xiong Zhou <jencce.kernel@...il.com>

Sorry, but this isn't good.

You had an earlier patch that introduced a bug by hiding a warning,
and now you send a new patch to hide the bug better instead of
fixing it?

> diff --git a/drivers/staging/android/logger.c b/drivers/staging/android/logger.c
> index 28b93d3..fb06bf2 100644
> --- a/drivers/staging/android/logger.c
> +++ b/drivers/staging/android/logger.c
> @@ -252,7 +252,7 @@ static size_t get_next_entry_by_uid(struct logger_log *log,
> 
>                 entry = get_entry_header(log, off, &scratch);
> 
> -               if (uid_eq(entry->euid, euid))
> +               if (entry->euid == __kuid_val(euid))
>                         return off;
> 
>                 next_len = sizeof(struct logger_entry) + entry->len;

The code above was correct, you are just breaking it more.

> @@ -430,7 +430,7 @@ static ssize_t logger_write_iter(struct kiocb *iocb, struct iov_iter *from)
>         header.tid = current->pid;
>         header.sec = now.tv_sec;
>         header.nsec = now.tv_nsec;
> -       header.euid = current_euid();
> +       header.euid = __kuid_val(current_euid());
>         header.len = count;
>         header.hdr_size = sizeof(struct logger_entry);
> 

This writes to the internal data structure, so here you should still use
kuid_t. The problem is that you are later leaking this to user space,
so a user in one container can read entries from a user with the same
uid in another container, and that a process calling read() with the
r_all bit set will get entries with uids from all containers, using
the kernel-internal uid values instead of the ones valid for the current
container.

	Arnd
--
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/