Skip to contentSkip to conversation

Fix Xen boot with CONFIG_CC_STACKPROTECTOR

GIT PULL Jeremy Fitzhardinge 3 participants 5 messages Aug 20, 2009 – Aug 21, 2009

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

Hi Ingo,

Here's a couple of little changes which prevents an early boot crash under Xen, due to the stackprotector preamble using segment registers before they're set up. I only need to disable stackprotector for a handful of functions, but unfortunately it looks like it can only be disabled on a per-file basis.

Thanks,
J

The following changes since commit 64f1607ffbbc772685733ea63e6f7f4183df1b16:
Linus Torvalds (1):

        Linux 2.6.31-rc6

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/jeremy/xen.git bugfix

Jeremy Fitzhardinge (2):
x86: make sure load_percpu_segment has no stackprotector
xen: rearrange things to fix stackprotector

 arch/x86/kernel/cpu/Makefile |    4 ++++
 arch/x86/xen/Makefile        |    4 ++++
 arch/x86/xen/enlighten.c     |   22 ++++++++++------------
 3 files changed, 18 insertions(+), 12 deletions(-)
arch/x86/kernel/cpu/Makefile +4 -0
index 4e242f9..8b5b9b6 100644
@@ -7,6 +7,10 @@ ifdef CONFIG_FUNCTION_TRACER
 CFLAGS_REMOVE_common.o = -pg
 endif
 
+# Make sure load_percpu_segment has no stackprotector
+nostackp := $(call cc-option, -fno-stack-protector)
+CFLAGS_common.o		:= $(nostackp)
+
 obj-y			:= intel_cacheinfo.o addon_cpuid_features.o
 obj-y			+= proc.o capflags.o powerflags.o common.o
 obj-y			+= vmware.o hypervisor.o
arch/x86/xen/Makefile +4 -0
index 3b767d0..a5b9288 100644
@@ -5,6 +5,10 @@ CFLAGS_REMOVE_time.o = -pg
 CFLAGS_REMOVE_irq.o = -pg
 endif
 
+# Make sure early boot has no stackprotector
+nostackp := $(call cc-option, -fno-stack-protector)
+CFLAGS_enlighten.o		:= $(nostackp)
+
 obj-y		:= enlighten.o setup.o multicalls.o mmu.o irq.o \
 			time.o xen-asm.o xen-asm_$(BITS).o \
 			grant-table.o suspend.o
arch/x86/xen/enlighten.c +10 -12
index f09e8c3..edcf72a 100644
@@ -925,10 +925,6 @@ asmlinkage void __init xen_start_kernel(void)
 
 	xen_domain_type = XEN_PV_DOMAIN;
 
-	BUG_ON(memcmp(xen_start_info->magic, "xen-3", 5) != 0);
-
-	xen_setup_features();
-
 	/* Install Xen paravirt ops */
 	pv_info = xen_info;
 	pv_init_ops = xen_init_ops;
@@ -937,8 +933,15 @@ asmlinkage void __init xen_start_kernel(void)
 	pv_apic_ops = xen_apic_ops;
 	pv_mmu_ops = xen_mmu_ops;
 
-	xen_init_irq_ops();
+#ifdef CONFIG_X86_64
+	/*
+	 * Setup percpu state.  We only need to do this for 64-bit
+	 * because 32-bit already has %fs set properly.
+	 */
+	load_percpu_segment(0);
+#endif
 
+	xen_init_irq_ops();
 	xen_init_cpuid_mask();
 
 #ifdef CONFIG_X86_LOCAL_APIC
@@ -948,6 +951,8 @@ asmlinkage void __init xen_start_kernel(void)
 	set_xen_basic_apic_ops();
 #endif
 
+	xen_setup_features();
+
 	if (xen_feature(XENFEAT_mmu_pt_update_preserve_ad)) {
 		pv_mmu_ops.ptep_modify_prot_start = xen_ptep_modify_prot_start;
 		pv_mmu_ops.ptep_modify_prot_commit = xen_ptep_modify_prot_commit;
@@ -955,13 +960,6 @@ asmlinkage void __init xen_start_kernel(void)
 
 	machine_ops = xen_machine_ops;
 
-#ifdef CONFIG_X86_64
-	/*
-	 * Setup percpu state.  We only need to do this for 64-bit
-	 * because 32-bit already has %fs set properly.
-	 */
-	load_percpu_segment(0);
-#endif
 	/*
 	 * The only reliable way to retain the initial address of the
 	 * percpu gdt_page is to remember it here, so we can go and
#1 lore raw
Quoted 19 lines · Jeremy Fitzhardinge

Hi Ingo,

Here's a couple of little changes which prevents an early boot crash under Xen, due to the stackprotector preamble using segment registers before they're set up. I only need to disable stackprotector for a handful of functions, but unfortunately it looks like it can only be disabled on a per-file basis.

Thanks,
J

The following changes since commit 64f1607ffbbc772685733ea63e6f7f4183df1b16:
Linus Torvalds (1):

        Linux 2.6.31-rc6

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/jeremy/xen.git bugfix

Quoted 8 lines

Jeremy Fitzhardinge (2):
x86: make sure load_percpu_segment has no stackprotector
xen: rearrange things to fix stackprotector

 arch/x86/kernel/cpu/Makefile |    4 ++++
 arch/x86/xen/Makefile        |    4 ++++
 arch/x86/xen/enlighten.c     |   22 ++++++++++------------
 3 files changed, 18 insertions(+), 12 deletions(-)

Pulled into tip:x86/urgent, thanks a lot Jeremy!

Btw., is there any way to turn off stackprotector on a per function basis, a GCC attribute perhaps? That would be preferable to turning it off for all of common.o.

Ingo

#2 lore raw
Quoted 7 lines · Ingo Molnar

Pulled into tip:x86/urgent, thanks a lot Jeremy!

Btw., is there any way to turn off stackprotector on a per function basis, a GCC attribute perhaps? That would be preferable to turning it off for all of common.o.

Jeremy asked me that, I researched it, and there isn't one -- the only option is to segregate functions into separate files.

-hpa

Signature · 2 lines
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel.  I don't speak on their behalf.
#3 lore raw
In reply to Ingo Molnar
Quoted 8 lines · Ingo Molnar
Quoted 3 lines

I only need to disable stackprotector for a
handful of functions, but unfortunately it looks like it can only be disabled on a per-file basis.

Pulled into tip:x86/urgent, thanks a lot Jeremy!

Btw., is there any way to turn off stackprotector on a per function basis, a GCC attribute perhaps? That would be preferable to turning it off for all of common.o.

Yeah, its awkward. But it looks like this patch is not sufficient, so drop it for now. I'll send a replacement shortly (though it doesn't hurt to keep it; I can just generate a delta if that's better).

J

#4 lore raw
Quoted 6 lines · Jeremy Fitzhardinge
Quoted 5 lines · Ingo Molnar
Quoted 3 lines

I only need to disable stackprotector for a
handful of functions, but unfortunately it looks like it can only be disabled on a per-file basis.

Pulled into tip:x86/urgent, thanks a lot Jeremy!

Quoted 8 lines
Quoted 3 lines

Btw., is there any way to turn off stackprotector on a per function basis, a GCC attribute perhaps? That would be preferable to turning it off for all of common.o.

Yeah, its awkward. But it looks like this patch is not sufficient, so drop it for now. I'll send a replacement shortly (though it doesn't hurt to keep it; I can just generate a delta if that's better).

I already pulled it as the fixes looked right so please send a delta based on top of tip:x86/urgent.

Thanks,

Ingo