[PATCH 3.16-stable 0/5] Add PVCLOCK_FIXMAP user mapping

[PATCH 3.16-stable 0/5] Add PVCLOCK_FIXMAP user mapping

[Juerg Haefliger]

Patch 3 fixes the userspace segfaults caused by the PVCLOCK_FIXMAP user
mapping (which I've copied from the 3.2 kaiser patchset). I don't claim I fully
understand this so the fix might be too broad.

Andrea Arcangeli (1):
  x86/mm/kaiser: remove paravirt clock warning

Juerg Haefliger (3):
  Revert "x86: kvmclock: Disable use from vDSO if KPTI is enabled"
  x86/kaiser: Add PVCLOCK_FIXMAP user mapping
  x86/kaiser: Fix segfaults caused by the PVCLOCK_FIXMAP user mapping

Marcelo Tosatti (1):
  kvmclock: export kvmclock clocksource and data pointers

 arch/x86/include/asm/kvmclock.h |  6 ++++++
 arch/x86/kernel/kvmclock.c      |  9 +++------
 arch/x86/mm/kaiser.c            | 12 +++++++++++-
 3 files changed, 20 insertions(+), 7 deletions(-)
 create mode 100644 arch/x86/include/asm/kvmclock.h

-- 
2.14.1

[PATCH 3.16-stable 1/5] Revert "x86: kvmclock: Disable use from vDSO if KPTI is enabled"

[Juerg Haefliger]

This reverts commit abe3029e4febfa18e4a9562a792465182b3992a0.

Signed-off-by: Juerg Haefliger <juerg.haefliger@canonical.com>
---
 arch/x86/kernel/kvmclock.c | 5 -----
 arch/x86/mm/kaiser.c       | 2 +-
 2 files changed, 1 insertion(+), 6 deletions(-)

diff --git a/arch/x86/kernel/kvmclock.c b/arch/x86/kernel/kvmclock.c
index 206aa5a2afe0..a2de9bc7ac0b 100644
--- a/arch/x86/kernel/kvmclock.c
+++ b/arch/x86/kernel/kvmclock.c
@@ -24,7 +24,6 @@
 #include <linux/percpu.h>
 #include <linux/hardirq.h>
 #include <linux/memblock.h>
-#include <linux/kaiser.h>
 
 #include <asm/x86_init.h>
 #include <asm/reboot.h>
@@ -282,10 +281,6 @@ int __init kvm_setup_vsyscall_timeinfo(void)
 	if (!hv_clock)
 		return 0;
 
-	/* FIXME: Need to add pvclock pages to user-space page tables */
-	if (kaiser_enabled)
-		return 0;
-
 	size = PAGE_ALIGN(sizeof(struct pvclock_vsyscall_time_info)*NR_CPUS);
 
 	preempt_disable();
diff --git a/arch/x86/mm/kaiser.c b/arch/x86/mm/kaiser.c
index 8d1019d176e2..8b5939649088 100644
--- a/arch/x86/mm/kaiser.c
+++ b/arch/x86/mm/kaiser.c
@@ -322,7 +322,7 @@ silent_disable:
  */
 void __init kaiser_init(void)
 {
-	int cpu, idx;
+	int cpu;
 
 	if (!kaiser_enabled)
 		return;
-- 
2.14.1

[PATCH 3.16-stable 2/5] x86/kaiser: Add PVCLOCK_FIXMAP user mapping

[Juerg Haefliger]

Code lifted from patch 'x86/mm/kaiser: re-enable vsyscalls', courtesy of
Andrea Arcangeli <aarcange@redhat.com>.

Signed-off-by: Juerg Haefliger <juerg.haefliger@canonical.com>
---
 arch/x86/mm/kaiser.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/arch/x86/mm/kaiser.c b/arch/x86/mm/kaiser.c
index 8b5939649088..fc50dee484f6 100644
--- a/arch/x86/mm/kaiser.c
+++ b/arch/x86/mm/kaiser.c
@@ -322,7 +322,7 @@ silent_disable:
  */
 void __init kaiser_init(void)
 {
-	int cpu;
+	int cpu, idx;
 
 	if (!kaiser_enabled)
 		return;
@@ -371,6 +371,14 @@ void __init kaiser_init(void)
 		kaiser_add_user_map_early((void *)VSYSCALL_ADDR, PAGE_SIZE,
 					  vsyscall_pgprot);
 
+#ifdef CONFIG_PARAVIRT_CLOCK
+	for (idx = 0; idx <= (PVCLOCK_FIXMAP_END-PVCLOCK_FIXMAP_BEGIN); idx++) {
+		kaiser_add_user_map_early((void *)__fix_to_virt(PVCLOCK_FIXMAP_BEGIN + idx),
+					  PAGE_SIZE,
+					  __PAGE_KERNEL_VVAR | _PAGE_GLOBAL);
+	}
+#endif
+
 	pr_info("enabled\n");
 }
 
-- 
2.14.1

[PATCH 3.16-stable 3/5] x86/kaiser: Fix segfaults caused by the PVCLOCK_FIXMAP user mapping

[Juerg Haefliger]

Without this, we'll get early userspace segfaults like this in the
PVCLOCK_FIXMAP area:

[    0.602244] init[1]: segfault at ffffffffff5ff020 ip 00007fff7154afc1 sp 00007fff71540c08 error 5
[    0.603026] Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b
[    0.603026]
[    0.603624] CPU: 0 PID: 1 Comm: init Not tainted 3.16.53-rc2-stable+ #240
[    0.604002] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1ubuntu1 04/01/2014
[    0.604002]  0000000000000000 ffff88003e31bc80 ffffffff81741db3 ffffffff81a60558
[    0.604002]  ffff88003d222380 ffff88003e31bcf8 ffffffff8173da0e 0000000000000010
[    0.604002]  ffff88003e31bd08 ffff88003e31bca8 0000000000000000 000000000000000b
[    0.604002] Call Trace:
[    0.604002]  [<ffffffff81741db3>] dump_stack+0x64/0x82
[    0.604002]  [<ffffffff8173da0e>] panic+0xc8/0x203
[    0.604002]  [<ffffffff8109c7f0>] do_exit+0xb00/0xb00
[    0.604002]  [<ffffffff8109d5ff>] do_group_exit+0x3f/0xa0
[    0.604002]  [<ffffffff810ac9c2>] get_signal_to_deliver+0x1c2/0x5e0
[    0.604002]  [<ffffffff810494f8>] do_signal+0x48/0x730
[    0.604002]  [<ffffffff8173d70c>] ? __bad_area_nosemaphore+0x1bd/0x1ca
[    0.604002]  [<ffffffff8108cad4>] ? __do_page_fault+0x84/0x400
[    0.604002]  [<ffffffff81049c45>] do_notify_resume+0x65/0x80
[    0.604002]  [<ffffffff8174c002>] retint_signal+0x48/0x86
[    0.604002] Kernel Offset: 0x0 from 0xffffffff81000000 (relocation range: 0xffffffff80000000-0xffffffff9fffffff)
[    0.604002] ---[ end Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b

This is probably too big of a hammer and needs some polishing but it's
working :-)

Signed-off-by: Juerg Haefliger <juerg.haefliger@canonical.com>
---
 arch/x86/mm/kaiser.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/x86/mm/kaiser.c b/arch/x86/mm/kaiser.c
index fc50dee484f6..3f0b0b41badc 100644
--- a/arch/x86/mm/kaiser.c
+++ b/arch/x86/mm/kaiser.c
@@ -202,7 +202,7 @@ static int kaiser_add_user_map(const void *__start_addr, unsigned long size,
 			ret = -EIO;
 			break;
 		}
-		pte = kaiser_pagetable_walk(address, flags & _PAGE_USER);
+		pte = kaiser_pagetable_walk(address, true);
 		if (!pte) {
 			ret = -ENOMEM;
 			break;
-- 
2.14.1

[PATCH 3.16-stable 4/5] kvmclock: export kvmclock clocksource and data pointers

[Juerg Haefliger]

From: Marcelo Tosatti <mtosatti@redhat.com>

To be used by KVM PTP driver.

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit f4066c2bc4d0de4e5dcbff21dae41e89fe8f38c0)
Signed-off-by: Juerg Haefliger <juerg.haefliger@canonical.com>
---
 arch/x86/include/asm/kvmclock.h | 6 ++++++
 arch/x86/kernel/kvmclock.c      | 4 +++-
 2 files changed, 9 insertions(+), 1 deletion(-)
 create mode 100644 arch/x86/include/asm/kvmclock.h

diff --git a/arch/x86/include/asm/kvmclock.h b/arch/x86/include/asm/kvmclock.h
new file mode 100644
index 000000000000..f260bef63591
--- /dev/null
+++ b/arch/x86/include/asm/kvmclock.h
@@ -0,0 +1,6 @@
+#ifndef _ASM_X86_KVM_CLOCK_H
+#define _ASM_X86_KVM_CLOCK_H
+
+extern struct clocksource kvm_clock;
+
+#endif /* _ASM_X86_KVM_CLOCK_H */
diff --git a/arch/x86/kernel/kvmclock.c b/arch/x86/kernel/kvmclock.c
index a2de9bc7ac0b..321c2acf0025 100644
--- a/arch/x86/kernel/kvmclock.c
+++ b/arch/x86/kernel/kvmclock.c
@@ -27,6 +27,7 @@
 
 #include <asm/x86_init.h>
 #include <asm/reboot.h>
+#include <asm/kvmclock.h>
 
 static int kvmclock = 1;
 static int msr_kvm_system_time = MSR_KVM_SYSTEM_TIME;
@@ -146,13 +147,14 @@ bool kvm_check_and_clear_guest_paused(void)
 	return ret;
 }
 
-static struct clocksource kvm_clock = {
+struct clocksource kvm_clock = {
 	.name = "kvm-clock",
 	.read = kvm_clock_get_cycles,
 	.rating = 400,
 	.mask = CLOCKSOURCE_MASK(64),
 	.flags = CLOCK_SOURCE_IS_CONTINUOUS,
 };
+EXPORT_SYMBOL_GPL(kvm_clock);
 
 int kvm_register_clock(char *txt)
 {
-- 
2.14.1

[PATCH 3.16-stable 5/5] x86/mm/kaiser: remove paravirt clock warning

[Juerg Haefliger]

From: Andrea Arcangeli <aarcange@redhat.com>

Adding shadow pgd mappings for unmapped kernel-side areas triggers
harmless boot warnings. To reduce the boot warnings (if booted on bare
metal) make the fixmap mapping of pvclock region conditional to
pvclock being actived at boot. KAISER is initialized after pvlock.

Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Juerg Haefliger <juerg.haefliger@canonical.com>
---
 arch/x86/mm/kaiser.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/arch/x86/mm/kaiser.c b/arch/x86/mm/kaiser.c
index 3f0b0b41badc..14c332caab60 100644
--- a/arch/x86/mm/kaiser.c
+++ b/arch/x86/mm/kaiser.c
@@ -24,6 +24,7 @@ extern struct mm_struct init_mm;
 #include <asm/desc.h>
 #include <asm/vsyscall.h>
 #include <asm/cmdline.h>
+#include <asm/kvmclock.h>
 
 int kaiser_enabled __read_mostly = 1;
 EXPORT_SYMBOL(kaiser_enabled);	/* for inlined TLB flush functions */
@@ -372,7 +373,8 @@ void __init kaiser_init(void)
 					  vsyscall_pgprot);
 
 #ifdef CONFIG_PARAVIRT_CLOCK
-	for (idx = 0; idx <= (PVCLOCK_FIXMAP_END-PVCLOCK_FIXMAP_BEGIN); idx++) {
+	for (idx = 0; kvm_clock.archdata.vclock_mode == VCLOCK_PVCLOCK &&
+		     idx <= (PVCLOCK_FIXMAP_END-PVCLOCK_FIXMAP_BEGIN); idx++) {
 		kaiser_add_user_map_early((void *)__fix_to_virt(PVCLOCK_FIXMAP_BEGIN + idx),
 					  PAGE_SIZE,
 					  __PAGE_KERNEL_VVAR | _PAGE_GLOBAL);
-- 
2.14.1

Re: [PATCH 3.16-stable 3/5] x86/kaiser: Fix segfaults caused by the PVCLOCK_FIXMAP user mapping

[Ben Hutchings]

[-- Attachment #1: Type: text/plain, Size: 2876 bytes --]

On Wed, 2018-01-17 at 17:22 +0100, Juerg Haefliger wrote:
> Without this, we'll get early userspace segfaults like this in the
> PVCLOCK_FIXMAP area:
> 
> [    0.602244] init[1]: segfault at ffffffffff5ff020 ip 00007fff7154afc1 sp 00007fff71540c08 error 5
> [    0.603026] Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b
> [    0.603026]
> [    0.603624] CPU: 0 PID: 1 Comm: init Not tainted 3.16.53-rc2-stable+ #240
> [    0.604002] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1ubuntu1 04/01/2014
> [    0.604002]  0000000000000000 ffff88003e31bc80 ffffffff81741db3 ffffffff81a60558
> [    0.604002]  ffff88003d222380 ffff88003e31bcf8 ffffffff8173da0e 0000000000000010
> [    0.604002]  ffff88003e31bd08 ffff88003e31bca8 0000000000000000 000000000000000b
> [    0.604002] Call Trace:
> [    0.604002]  [<ffffffff81741db3>] dump_stack+0x64/0x82
> [    0.604002]  [<ffffffff8173da0e>] panic+0xc8/0x203
> [    0.604002]  [<ffffffff8109c7f0>] do_exit+0xb00/0xb00
> [    0.604002]  [<ffffffff8109d5ff>] do_group_exit+0x3f/0xa0
> [    0.604002]  [<ffffffff810ac9c2>] get_signal_to_deliver+0x1c2/0x5e0
> [    0.604002]  [<ffffffff810494f8>] do_signal+0x48/0x730
> [    0.604002]  [<ffffffff8173d70c>] ? __bad_area_nosemaphore+0x1bd/0x1ca
> [    0.604002]  [<ffffffff8108cad4>] ? __do_page_fault+0x84/0x400
> [    0.604002]  [<ffffffff81049c45>] do_notify_resume+0x65/0x80
> [    0.604002]  [<ffffffff8174c002>] retint_signal+0x48/0x86
> [    0.604002] Kernel Offset: 0x0 from 0xffffffff81000000 (relocation range: 0xffffffff80000000-0xffffffff9fffffff)
> [    0.604002] ---[ end Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b
> 
> This is probably too big of a hammer and needs some polishing but it's
> working :-)

It certainly is too big a hammer.

I think what's going wrong is that the pmd containing the fixmap area
is already mapped with _KERNPG_TABLE, i.e. excluding _PAGE_USER.  Then
when we map the ptes for the pvclock area we don't update the pmd
protections.

I'll look at some other branches to see how this was addressed there.

Ben.

> Signed-off-by: Juerg Haefliger <juerg.haefliger@canonical.com>
> ---
>  arch/x86/mm/kaiser.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/arch/x86/mm/kaiser.c b/arch/x86/mm/kaiser.c
> index fc50dee484f6..3f0b0b41badc 100644
> --- a/arch/x86/mm/kaiser.c
> +++ b/arch/x86/mm/kaiser.c
> @@ -202,7 +202,7 @@ static int kaiser_add_user_map(const void *__start_addr, unsigned long size,
>  			ret = -EIO;
>  			break;
>  		}
> -		pte = kaiser_pagetable_walk(address, flags & _PAGE_USER);
> +		pte = kaiser_pagetable_walk(address, true);
>  		if (!pte) {
>  			ret = -ENOMEM;
>  			break;
-- 
Ben Hutchings
Beware of programmers who carry screwdrivers. - Leonard Brandwein


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [PATCH 3.16-stable 3/5] x86/kaiser: Fix segfaults caused by the PVCLOCK_FIXMAP user mapping

[Ben Hutchings]

[-- Attachment #1: Type: text/plain, Size: 1083 bytes --]

On Mon, 2018-02-05 at 15:03 +0000, Ben Hutchings wrote:
[...]
> It certainly is too big a hammer.
> 
> I think what's going wrong is that the pmd containing the fixmap area
> is already mapped with _KERNPG_TABLE, i.e. excluding _PAGE_USER.  Then
> when we map the ptes for the pvclock area we don't update the pmd
> protections.
> 
> I'll look at some other branches to see how this was addressed there.
[...]

I ended up backporting most of the vDSO pvclock changes up to 4.5, as
was done in 4.4-stable.  This removes the need for the vDSO to access
the pvclock fixmap.  I picked these commits:

e6577a7ce99a x86, vdso: Move the vvar area before the vdso text
73459e2a1ada x86: pvclock: Really remove the sched notifier for cross-cpu migrations
6b078f5de7fc x86, vdso, pvclock: Simplify and speed up the vdso pvclock reader
dac16fba6fc5 x86/vdso: Get pvclock data from the vvar VMA instead of the fixmap

and then applied your revert to re-enable use of pvclock.

Ben.

-- 
Ben Hutchings
Beware of programmers who carry screwdrivers. - Leonard Brandwein

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 833 bytes --]