Re: [PATCH 1/3] signal(i386): alternative signal stack wraparoundoccurs

From: Shi Weihua
Date: Sun Nov 18 2007 - 21:18:15 EST


Hi everyone,

If a process uses alternative signal stack by using sigaltstack(),
then that stack overflows and stack wraparound may occur.
Simple explanation:
The accurate esp order is A,B,C,D,...
But now the esp points to A,B,C and A,B,C... dropping into a recursion.

The upper bug and patch about "alternative signal stack wraparound occurs"
has been contributed here at 10/3.
(subject:[PATCH 0/3] signal: alternative signal stack wraparound occurs)
(Please refer to http://lkml.org/lkml/2007/10/3/41).

Now, I renewed the patch and it can stop wraparound.
Can you give me some advice about storing the previous esp?

Signed-off-by: Shi Weihua <shiwh@xxxxxxxxxxxxxx>

---
diff -urpN linux-2.6.24-rc2.orig/arch/x86/kernel/signal_32.c linux-2.6.24-rc2/arch/x86/kernel/signal_32.c
--- linux-2.6.24-rc2.orig/arch/x86/kernel/signal_32.c 2007-11-13 14:30:45.000000000 +0800
+++ linux-2.6.24-rc2/arch/x86/kernel/signal_32.c 2007-11-13 14:38:03.000000000 +0800
@@ -297,7 +297,8 @@ get_sigframe(struct k_sigaction *ka, str

/* This is the X/Open sanctioned signal stack switching. */
if (ka->sa.sa_flags & SA_ONSTACK) {
- if (sas_ss_flags(esp) == 0)
+ if (sas_ss_flags(esp) == 0 &&
+ !on_sig_stack(current->pre_ss_sp))
esp = current->sas_ss_sp + current->sas_ss_size;
}

@@ -330,9 +331,15 @@ static int setup_frame(int sig, struct k

frame = get_sigframe(ka, regs, sizeof(*frame));

+ if ((ka->sa.sa_flags & SA_ONSTACK) &&
+ !sas_ss_flags((unsigned long)frame))
+ goto give_sigsegv;
+
if (!access_ok(VERIFY_WRITE, frame, sizeof(*frame)))
goto give_sigsegv;

+ current->pre_ss_sp = (unsigned long)frame;
+
usig = current_thread_info()->exec_domain
&& current_thread_info()->exec_domain->signal_invmap
&& sig < 32
diff -urpN linux-2.6.24-rc2.orig/include/linux/sched.h linux-2.6.24-rc2/include/linux/sched.h
--- linux-2.6.24-rc2.orig/include/linux/sched.h 2007-11-13 14:29:17.000000000 +0800
+++ linux-2.6.24-rc2/include/linux/sched.h 2007-11-13 14:31:46.000000000 +0800
@@ -1059,6 +1059,7 @@ struct task_struct {
struct sigpending pending;

unsigned long sas_ss_sp;
+ unsigned long pre_ss_sp;
size_t sas_ss_size;
int (*notifier)(void *priv);
void *notifier_data;
diff -urpN linux-2.6.24-rc2.orig/kernel/signal.c linux-2.6.24-rc2/kernel/signal.c
--- linux-2.6.24-rc2.orig/kernel/signal.c 2007-11-13 14:29:16.000000000 +0800
+++ linux-2.6.24-rc2/kernel/signal.c 2007-11-13 14:33:00.000000000 +0800
@@ -2403,6 +2403,9 @@ do_sigaltstack (const stack_t __user *us

current->sas_ss_sp = (unsigned long) ss_sp;
current->sas_ss_size = ss_size;
+
+ /* reset previous sp */
+ current->pre_ss_sp = 0;
}

if (uoss) {


Shi Weihua wrote::
> Fixing alternative signal stack wraparound.
>
> If a process uses alternative signal stack by using sigaltstack()
> and that stack overflow, stack wraparound occurs.
> This patch checks whether the signal frame is on the alternative
> stack. If the frame is not on there, kill a signal SIGSEGV to the
> process forcedly
> then the process will be terminated.
>
> This patch is for i386,version is 2.6.23-rc8.
>
> Signed-off-by: Shi Weihua <shiwh@xxxxxxxxxxxxxx>
>
> diff -pur linux-2.6.23-rc8.orig/arch/i386/kernel/signal.c
> linux-2.6.23-rc8/arch/i386/kernel/signal.c
> --- linux-2.6.23-rc8.orig/arch/i386/kernel/signal.c 2007-09-26
> 09:44:08.000000000 +0900
> +++ linux-2.6.23-rc8/arch/i386/kernel/signal.c 2007-09-26
> 13:14:25.000000000 +0900
> @@ -332,6 +332,10 @@ static int setup_frame(int sig, struct k
>
> frame = get_sigframe(ka, regs, sizeof(*frame));
>
> + if ((ka->sa.sa_flags & SA_ONSTACK) &&
> + !sas_ss_flags((unsigned long)frame))
> + goto give_sigsegv;
> +
> if (!access_ok(VERIFY_WRITE, frame, sizeof(*frame)))
> goto give_sigsegv;
>
> @@ -425,6 +429,10 @@ static int setup_rt_frame(int sig, struc
>
> frame = get_sigframe(ka, regs, sizeof(*frame));
>
> + if ((ka->sa.sa_flags & SA_ONSTACK) &&
> + !sas_ss_flags((unsigned long)frame))
> + goto give_sigsegv;
> +
> if (!access_ok(VERIFY_WRITE, frame, sizeof(*frame)))
> goto give_sigsegv;
>
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
>
>
>

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