[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20110803123352.88670c9c.akpm@linux-foundation.org>
Date: Wed, 3 Aug 2011 12:33:52 -0700
From: Andrew Morton <akpm@...ux-foundation.org>
To: Linus Torvalds <torvalds@...ux-foundation.org>
Cc: Vasiliy Kulikov <segoon@...nwall.com>,
Manuel Lauss
<manuel.lauss@...glemail.com>,
linux-kernel@...r.kernel.org, Richard
Weinberger <richard@....at>,
Marc Zyngier <maz@...terjones.org>, Ingo
Molnar <mingo@...e.hu>,
kernel-hardening@...ts.openwall.com,
"Paul E. McKenney" <paul.mckenney@...aro.org>,
Kay Sievers
<kay.sievers@...y.org>
Subject: Re: [PATCH] shm: fix a race between shm_exit() and shm_init()
On Tue, 2 Aug 2011 21:43:23 -1000
Linus Torvalds <torvalds@...ux-foundation.org> wrote:
> On Tue, Aug 2, 2011 at 2:45 AM, Vasiliy Kulikov <segoon@...nwall.com> wrote:
> >
> > From: Vasiliy Kulikov <segoon@...nwall.com>
> > Subject: [PATCH] shm: fix a race between shm_exit() and shm_init()
>
> This patch is disgusting.
>
I think it's buggy too.
>
> > + /*
> > + * For init_ipc_ns shm_ids().rw_mutex is statically initialized
> > + * as kernel threads should be able to use it in do_exit() before
> > + * shm_init(), which is called on do_initcall()
> > + */
> > + if (ns == &init_ipc_ns)
> > + ipc_init_ids(&shm_ids(ns));
> > + else
> > + ipc_init_ids(&shm_ids(ns));
afacit init_ipc_ns.ids[0].rw_mutex and init_ipc_ns.ids[1].rw_mutex
never get initialised with this patch?
Still. It seems that the real bug is that driver_init() is trying to
invoke userspace helpers before the kernel is ready to run userspace.
We could hack around it with something like
--- a/init/main.c~a
+++ a/init/main.c
@@ -108,6 +108,9 @@ bool early_boot_irqs_disabled __read_mos
enum system_states system_state __read_mostly;
EXPORT_SYMBOL(system_state);
+/* This gets set when the kernel is ready to execute usermode code */
+bool usermode_available = false;
+
/*
* Boot command-line arguments
*/
@@ -804,6 +807,8 @@ static int __init kernel_init(void * unu
do_basic_setup();
+ usermode_available = true;
+
/* Open the /dev/console on the rootfs, this should never fail */
if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
printk(KERN_WARNING "Warning: unable to open an initial console.\n");
diff -puN include/linux/init.h~a include/linux/init.h
--- a/include/linux/init.h~a
+++ a/include/linux/init.h
@@ -149,6 +149,7 @@ extern int do_one_initcall(initcall_t fn
extern char __initdata boot_command_line[];
extern char *saved_command_line;
extern unsigned int reset_devices;
+extern bool usermode_available;
/* used by init/main.c */
void setup_arch(char **);
diff -puN kernel/kmod.c~a kernel/kmod.c
--- a/kernel/kmod.c~a
+++ a/kernel/kmod.c
@@ -418,6 +418,9 @@ int call_usermodehelper_exec(struct subp
DECLARE_COMPLETION_ONSTACK(done);
int retval = 0;
+ if (usermode_available == false)
+ return -EINVAL;
+
helper_lock();
if (sub_info->path[0] == '\0')
goto out;
_
(which I think deserves a big fat WARN_ON() when it triggers)
But it would be better to fix driver_init() for real.
Save us, Kay!
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
