Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Thu, 18 Feb 2016 12:04:15 -0800
From: Kees Cook <>
To: Ard Biesheuvel <>
Cc: Arnd Bergmann <>, PaX Team <>, 
	Laura Abbott <>, Greg Kroah-Hartman <>, 
	Mark Rutland <>, Jeremy Linton <>, 
	"" <>, LKML <>
Subject: Re: [PATCH] lkdtm: add test for executing .rodata

On Thu, Feb 18, 2016 at 4:46 AM, Ard Biesheuvel
<> wrote:
> On 18 February 2016 at 13:07, Arnd Bergmann <> wrote:
>> On Thursday 18 February 2016 12:34:50 Ard Biesheuvel wrote:
>>> We have __section() as an alias for __attribute__((__section__())), so
>>> we could use that instead.
>>> However, that does not fix the issue Kees is trying to solve, where a
>>> .rodata section is emitted with the "x" bit set, which causes the
>>> linker to complain:
>>> /tmp/cc50ffWw.s: Assembler messages:
>>> /tmp/cc50ffWw.s:2: Warning: setting incorrect section attributes for
>>> .rodata.text
>>> I wonder if we could get away with doing something like
>>> AFLAGS_lkdtm.o += -Wa,-W
>>> here? This just hides the warnings, but may result in the .rodata
>>> section in the vmlinux file to have X permissions as well. I don't
>>> think anyone uses an ELF loader to load their kernel, but who knows
>>> ...
>> Don't we also get a warning when we link objects with conflicting
>> section attributes?
> I didn't see one
>> Maybe a solution would be to define a separate section for this one
>> function, and then use a linker script to move it into .rodata?
>> Or maybe "objcopy --set-section-flags  --rename-section"?
> I think objcopy is the easiest.

I came to the same conclusion while thinking about it last night. I'm
having a terrible time implementing it in kbuild, though. The
"objcopy" function expects to rename the files, and I can't find a way
to just add it on without breaking the module build.

The flags I'm using that seem to actually do what's needed are (when
using the section name ".text.rodata"):

OBJCOPYFLAGS_lkdtm.o := --set-section-flags .text.rodata=alloc,readonly \
                        --rename-section .text.rodata=.rodata

So the diff against my existing patch looks like this, and works for
CONFIG_LKDTM=y (sorry for whitespace damage...):

diff --git a/drivers/misc/Makefile b/drivers/misc/Makefile
index 537d7f3b78da..55ce014e6b62 100644
--- a/drivers/misc/Makefile
+++ b/drivers/misc/Makefile
@@ -14,7 +14,7 @@ obj-$(CONFIG_BMP085_I2C)      += bmp085-i2c.o
 obj-$(CONFIG_BMP085_SPI)       += bmp085-spi.o
 obj-$(CONFIG_DUMMY_IRQ)                += dummy-irq.o
 obj-$(CONFIG_ICS932S401)       += ics932s401.o
-obj-$(CONFIG_LKDTM)            += lkdtm.o
+obj-$(CONFIG_LKDTM)            += lkdtm_objcopy.o
 obj-$(CONFIG_TIFM_CORE)        += tifm_core.o
 obj-$(CONFIG_TIFM_7XX1)        += tifm_7xx1.o
 obj-$(CONFIG_PHANTOM)          += phantom.o
@@ -56,3 +56,8 @@ obj-$(CONFIG_GENWQE)          += genwqe/
 obj-$(CONFIG_ECHO)             += echo/
 obj-$(CONFIG_VEXPRESS_SYSCFG)  += vexpress-syscfg.o
 obj-$(CONFIG_CXL_BASE)         += cxl/
+OBJCOPYFLAGS_lkdtm_objcopy.o := --set-section-flags
.text.rodata=alloc,readonly \
+                       --rename-section .text.rodata=.rodata
+$(obj)/lkdtm_objcopy.o: $(obj)/lkdtm.o FORCE
+       $(call if_changed,objcopy)
diff --git a/drivers/misc/lkdtm.c b/drivers/misc/lkdtm.c
index b15d08ff71a9..5868e5125fbe 100644
--- a/drivers/misc/lkdtm.c
+++ b/drivers/misc/lkdtm.c
@@ -317,8 +317,7 @@ static int recursive_loop(int remaining)
                return recursive_loop(remaining - 1);

-static void __attribute__((__section__(".rodata,\"a\",\%progbits;//")))
+static void __section(.text.rodata) do_nothing_rodata(void)

But fails in ways I don't understand for CONFIG_LKDTM=m:

./scripts/ warning: overriding commands for target `drivers/m
drivers/misc/Makefile:63: warning: ignoring old commands for target `drivers/mis
make[2]: *** No rule to make target `drivers/misc/lkdtm_objcopy.c', needed by `d
rivers/misc/lkdtm_objcopy.o'.  Stop.
make[2]: *** Waiting for unfinished jobs....
make[1]: *** [drivers/misc] Error 2
make[1]: *** Waiting for unfinished jobs....


Kees Cook
Chrome OS & Brillo Security

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.