Re: 2.6.29 git, resume from ram broken on thinkpad

[Chris Wright]

* Arkadiusz Miskiewicz (a.miskiewicz@xxxxxxxxx) wrote:
> On Friday 03 of April 2009, Chris Wright wrote:
> > * Arkadiusz Miskiewicz (a.miskiewicz@xxxxxxxxx) wrote:
> > > What about 9ea09af3bd3090e8349ca2899ca2011bd94cda85 ?
> > >
> > > stop_machine: introduce stop_machine_create/destroy.
> >
> > That is later fixed in a0e280e0f33f6c859a235fb69a875ed8f3420388.
> >
> > Can you please verify if 2.6.29 works for you? 
> 
> I think that the guilty part is 
> CONFIG_CC_STACKPROTECTOR_ALL=y
> CONFIG_CC_STACKPROTECTOR=y

Indeed, I think you're right.  In fact...this should fix it:

From:  Joseph Cihula <joseph.cihula@xxxxxxxxx>

The __restore_processor_state() fn restores %gs on resume from S3.  As
such, it cannot be protected by the stack-protector guard since %gs will
not be correct on function entry.

There are only a few other fns in this file and it should not negatively
impact kernel security that they will also have the stack-protector
guard removed (and so it's not worth moving them to another file).

Without this change, S3 resume on a kernel built with
CONFIG_CC_STACKPROTECTOR_ALL=y will fail.

Signed-off-by:  Joseph Cihula <joseph.cihula@xxxxxxxxx>

--- ../linux.trees.git/arch/x86/power/Makefile	2009-03-29 
12:12:13.000000000 -0700
+++ arch/x86/power/Makefile	2009-03-30 12:21:19.000000000 -0700
@@ -1,2 +1,7 @@
+# __restore_processor_state() restores %gs after S3 resume and so 
should not
+# itself be stack-protected
+nostackp := $(call cc-option, -fno-stack-protector)
+CFLAGS_cpu_$(BITS).o	:= $(nostackp)
+
obj-$(CONFIG_PM_SLEEP)		+= cpu_$(BITS).o
obj-$(CONFIG_HIBERNATION)	+= hibernate_$(BITS).o hibernate_asm_$(BITS).o



--
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/