2015-11-04 09:38:31

by Zhou Chengming

Subject: [PATCH] livepatch: x86: bugfix about kASLR

When enable KASLR, func->old_addr will be set to zero
and livepatch will find the right old address.
But for reloc, livepatch just verify it using reloc->val
(old addr from user), so verify failed and report
"kernel mismatch" error.

Reported-by: Cyril B. <[email protected]>
Signed-off-by: Zhou Chengming <[email protected]>
---
kernel/livepatch/core.c | 7 ++++++-
1 files changed, 6 insertions(+), 1 deletions(-)

diff --git a/kernel/livepatch/core.c b/kernel/livepatch/core.c
index 53196e2..c8885c6 100644
--- a/kernel/livepatch/core.c
+++ b/kernel/livepatch/core.c
@@ -311,7 +311,12 @@ static int klp_write_object_relocations(struct module *pmod,
return -EINVAL;

for (reloc = obj->relocs; reloc->name; reloc++) {
- if (!klp_is_module(obj)) {
+
+#if defined(CONFIG_RANDOMIZE_BASE)
+ /* KASLR is enabled, disregard old_addr from user */
+ reloc->val = 0;
+#endif
+ if (reloc->val && !klp_is_module(obj)) {
ret = klp_verify_vmlinux_symbol(reloc->name,
reloc->val);
if (ret)
--
1.7.7

2015-11-04 09:45:01

by Jiri Kosina

Subject: Re: [PATCH] livepatch: x86: bugfix about kASLR

On Wed, 4 Nov 2015, Zhou Chengming wrote:

> When enable KASLR, func->old_addr will be set to zero
> and livepatch will find the right old address.
> But for reloc, livepatch just verify it using reloc->val
> (old addr from user), so verify failed and report
> "kernel mismatch" error.
>
> Reported-by: Cyril B. <[email protected]>
> Signed-off-by: Zhou Chengming <[email protected]>
> ---
> kernel/livepatch/core.c | 7 ++++++-
> 1 files changed, 6 insertions(+), 1 deletions(-)
>
> diff --git a/kernel/livepatch/core.c b/kernel/livepatch/core.c
> index 53196e2..c8885c6 100644
> --- a/kernel/livepatch/core.c
> +++ b/kernel/livepatch/core.c
> @@ -311,7 +311,12 @@ static int klp_write_object_relocations(struct module *pmod,
> return -EINVAL;
>
> for (reloc = obj->relocs; reloc->name; reloc++) {
> - if (!klp_is_module(obj)) {
> +
> +#if defined(CONFIG_RANDOMIZE_BASE)

CONFIG_RANDOMIZE_BASE being enabled by itself doesn't directly imply that
kASLR has been actually really enabled during runtime. You need
to check kaslr_enabled() as well.

> + /* KASLR is enabled, disregard old_addr from user */
> + reloc->val = 0;

Is there a reason why to discard it completely? Adding kaslr_offset()
should give us the correct address, right?

Thanks,

--
Jiri Kosina
SUSE Labs

2015-11-04 10:06:08

by Zhou Chengming

Subject: Re: [PATCH] livepatch: x86: bugfix about kASLR

On 2015/11/4 17:44, Jiri Kosina wrote:
> On Wed, 4 Nov 2015, Zhou Chengming wrote:
>
>> When enable KASLR, func->old_addr will be set to zero
>> and livepatch will find the right old address.
>> But for reloc, livepatch just verify it using reloc->val
>> (old addr from user), so verify failed and report
>> "kernel mismatch" error.
>>
>> Reported-by: Cyril B.<[email protected]>
>> Signed-off-by: Zhou Chengming<[email protected]>
>> ---
>> kernel/livepatch/core.c | 7 ++++++-
>> 1 files changed, 6 insertions(+), 1 deletions(-)
>>
>> diff --git a/kernel/livepatch/core.c b/kernel/livepatch/core.c
>> index 53196e2..c8885c6 100644
>> --- a/kernel/livepatch/core.c
>> +++ b/kernel/livepatch/core.c
>> @@ -311,7 +311,12 @@ static int klp_write_object_relocations(struct module *pmod,
>> return -EINVAL;
>>
>> for (reloc = obj->relocs; reloc->name; reloc++) {
>> - if (!klp_is_module(obj)) {
>> +
>> +#if defined(CONFIG_RANDOMIZE_BASE)
> CONFIG_RANDOMIZE_BASE being enabled by itself doesn't directly imply that
> kASLR has been actually really enabled during runtime. You need
> to check kaslr_enabled() as well.

Right.

>> + /* KASLR is enabled, disregard old_addr from user */
>> + reloc->val = 0;
> Is there a reason why to discard it completely? Adding kaslr_offset()
> should give us the correct address, right?
>
> Thanks,
>

Yes, it's definitely better.

Thanks

2015-11-04 16:59:11

by Cyril B.

Subject: Re: [PATCH] livepatch: x86: bugfix about kASLR

Zhou Chengming wrote:
> When enable KASLR, func->old_addr will be set to zero
> and livepatch will find the right old address.
> But for reloc, livepatch just verify it using reloc->val
> (old addr from user), so verify failed and report
> "kernel mismatch" error.
>
> Reported-by: Cyril B.<[email protected]>
> Signed-off-by: Zhou Chengming<[email protected]>
> ---
> kernel/livepatch/core.c | 7 ++++++-
> 1 files changed, 6 insertions(+), 1 deletions(-)
>
> diff --git a/kernel/livepatch/core.c b/kernel/livepatch/core.c
> index 53196e2..c8885c6 100644
> --- a/kernel/livepatch/core.c
> +++ b/kernel/livepatch/core.c
> @@ -311,7 +311,12 @@ static int klp_write_object_relocations(struct module *pmod,
> return -EINVAL;
>
> for (reloc = obj->relocs; reloc->name; reloc++) {
> - if (!klp_is_module(obj)) {
> +
> +#if defined(CONFIG_RANDOMIZE_BASE)
> + /* KASLR is enabled, disregard old_addr from user */
> + reloc->val = 0;
> +#endif
> + if (reloc->val&& !klp_is_module(obj)) {
> ret = klp_verify_vmlinux_symbol(reloc->name,
> reloc->val);
> if (ret)

I still get the 'symbol not found in symbol table' error with that patch
(on top of Linux 4.3).

Cyril