[PATCH] x86, relocs: Ignore relocations in .notes section

Kees Cook posted 1 patch 1 year, 11 months ago
There is a newer version of this series
arch/x86/tools/relocs.c | 10 ++++++++++
1 file changed, 10 insertions(+)
[PATCH] x86, relocs: Ignore relocations in .notes section
Posted by Kees Cook 1 year, 11 months ago
When building with CONFIG_XEN_PV=y, .text symbols are emitted into the
.notes section so that Xen can find the "startup_xen" entry point. This
information is used prior to booting the kernel, so relocations are not
useful. In fact, performing relocations against the .notes section means
that the KASLR base is exposed since /sys/kernel/notes is world-readable.

To avoid leaking the KASLR base without breaking unprivileged tools that
are expecting to read /sys/kernel/notes, skip performing relocations in
the .notes section. The values readable in .notes are then identical to
those found in System.map.

Reported-by: Guixiong Wei <guixiongwei@gmail.com>
Closes: https://lore.kernel.org/all/20240218073501.54555-1-guixiongwei@gmail.com/
Fixes: 5ead97c84fa7 ("xen: Core Xen implementation")
Fixes: da1a679cde9b ("Add /sys/kernel/notes")
Signed-off-by: Kees Cook <keescook@chromium.org>
---
Cc: Borislav Petkov <bp@alien8.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: x86@kernel.org
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Kristen Carlson Accardi <kristen@linux.intel.com>
Cc: "Jürgen Groß" <jgross@suse.com>
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Stefano Stabellini <sstabellini@kernel.org>
Cc: Oleksandr Tyshchenko <oleksandr_tyshchenko@epam.com>
Cc: Guixiong Wei <weiguixiong@bytedance.com>
Cc: Jann Horn <jannh@google.com>
---
 arch/x86/tools/relocs.c | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/arch/x86/tools/relocs.c b/arch/x86/tools/relocs.c
index a3bae2b24626..0811fff23b9c 100644
--- a/arch/x86/tools/relocs.c
+++ b/arch/x86/tools/relocs.c
@@ -733,6 +733,16 @@ static void walk_relocs(int (*process)(struct section *sec, Elf_Rel *rel,
 		if (sec->shdr.sh_type != SHT_REL_TYPE) {
 			continue;
 		}
+
+		/*
+		 * Do not perform relocations in .notes section; any
+		 * values there are meant for pre-boot consumption (e.g.
+		 * startup_xen).
+		 */
+		if (strcmp(sec_name(sec->shdr.sh_info), ".notes") == 0) {
+			continue;
+		}
+
 		sec_symtab  = sec->link;
 		sec_applies = &secs[sec->shdr.sh_info];
 		if (!(sec_applies->shdr.sh_flags & SHF_ALLOC)) {
-- 
2.34.1

Re: [PATCH] x86, relocs: Ignore relocations in .notes section
Posted by Jürgen Groß 1 year, 11 months ago
On 22.02.24 18:18, Kees Cook wrote:
> When building with CONFIG_XEN_PV=y, .text symbols are emitted into the
> .notes section so that Xen can find the "startup_xen" entry point. This
> information is used prior to booting the kernel, so relocations are not
> useful. In fact, performing relocations against the .notes section means
> that the KASLR base is exposed since /sys/kernel/notes is world-readable.
> 
> To avoid leaking the KASLR base without breaking unprivileged tools that
> are expecting to read /sys/kernel/notes, skip performing relocations in
> the .notes section. The values readable in .notes are then identical to
> those found in System.map.
> 
> Reported-by: Guixiong Wei <guixiongwei@gmail.com>
> Closes: https://lore.kernel.org/all/20240218073501.54555-1-guixiongwei@gmail.com/
> Fixes: 5ead97c84fa7 ("xen: Core Xen implementation")
> Fixes: da1a679cde9b ("Add /sys/kernel/notes")
> Signed-off-by: Kees Cook <keescook@chromium.org>
> ---
> Cc: Borislav Petkov <bp@alien8.de>
> Cc: Thomas Gleixner <tglx@linutronix.de>
> Cc: Ingo Molnar <mingo@redhat.com>
> Cc: Dave Hansen <dave.hansen@linux.intel.com>
> Cc: x86@kernel.org
> Cc: "H. Peter Anvin" <hpa@zytor.com>
> Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
> Cc: Tony Luck <tony.luck@intel.com>
> Cc: Kristen Carlson Accardi <kristen@linux.intel.com>
> Cc: "Jürgen Groß" <jgross@suse.com>
> Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
> Cc: Stefano Stabellini <sstabellini@kernel.org>
> Cc: Oleksandr Tyshchenko <oleksandr_tyshchenko@epam.com>
> Cc: Guixiong Wei <weiguixiong@bytedance.com>
> Cc: Jann Horn <jannh@google.com>
> ---
>   arch/x86/tools/relocs.c | 10 ++++++++++
>   1 file changed, 10 insertions(+)
> 
> diff --git a/arch/x86/tools/relocs.c b/arch/x86/tools/relocs.c
> index a3bae2b24626..0811fff23b9c 100644
> --- a/arch/x86/tools/relocs.c
> +++ b/arch/x86/tools/relocs.c
> @@ -733,6 +733,16 @@ static void walk_relocs(int (*process)(struct section *sec, Elf_Rel *rel,
>   		if (sec->shdr.sh_type != SHT_REL_TYPE) {
>   			continue;
>   		}
> +
> +		/*
> +		 * Do not perform relocations in .notes section; any
> +		 * values there are meant for pre-boot consumption (e.g.
> +		 * startup_xen).
> +		 */
> +		if (strcmp(sec_name(sec->shdr.sh_info), ".notes") == 0) {

Instead of a strcmp(), wouldnt't ...

> +			continue;
> +		}
> +
>   		sec_symtab  = sec->link;
>   		sec_applies = &secs[sec->shdr.sh_info];
>   		if (!(sec_applies->shdr.sh_flags & SHF_ALLOC)) {

... a test of "sec_applies->shdr.sh_type == SHT_NOTE" work as well?

In the end I'm fine with both variants, so:

Reviewed-by: Juergen Gross <jgross@suse.com>


Juergen
Re: [PATCH] x86, relocs: Ignore relocations in .notes section
Posted by Kees Cook 1 year, 11 months ago
On Tue, Feb 27, 2024 at 05:13:35PM +0100, Jürgen Groß wrote:
> ... a test of "sec_applies->shdr.sh_type == SHT_NOTE" work as well?

Oh yeah! Thanks. I'll send a v2.

-- 
Kees Cook
Re: [PATCH] x86, relocs: Ignore relocations in .notes section
Posted by H. Peter Anvin 1 year, 11 months ago
On February 27, 2024 8:13:35 AM PST, "Jürgen Groß" <jgross@suse.com> wrote:
>On 22.02.24 18:18, Kees Cook wrote:
>> When building with CONFIG_XEN_PV=y, .text symbols are emitted into the
>> .notes section so that Xen can find the "startup_xen" entry point. This
>> information is used prior to booting the kernel, so relocations are not
>> useful. In fact, performing relocations against the .notes section means
>> that the KASLR base is exposed since /sys/kernel/notes is world-readable.
>> 
>> To avoid leaking the KASLR base without breaking unprivileged tools that
>> are expecting to read /sys/kernel/notes, skip performing relocations in
>> the .notes section. The values readable in .notes are then identical to
>> those found in System.map.
>> 
>> Reported-by: Guixiong Wei <guixiongwei@gmail.com>
>> Closes: https://lore.kernel.org/all/20240218073501.54555-1-guixiongwei@gmail.com/
>> Fixes: 5ead97c84fa7 ("xen: Core Xen implementation")
>> Fixes: da1a679cde9b ("Add /sys/kernel/notes")
>> Signed-off-by: Kees Cook <keescook@chromium.org>
>> ---
>> Cc: Borislav Petkov <bp@alien8.de>
>> Cc: Thomas Gleixner <tglx@linutronix.de>
>> Cc: Ingo Molnar <mingo@redhat.com>
>> Cc: Dave Hansen <dave.hansen@linux.intel.com>
>> Cc: x86@kernel.org
>> Cc: "H. Peter Anvin" <hpa@zytor.com>
>> Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
>> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
>> Cc: Tony Luck <tony.luck@intel.com>
>> Cc: Kristen Carlson Accardi <kristen@linux.intel.com>
>> Cc: "Jürgen Groß" <jgross@suse.com>
>> Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
>> Cc: Stefano Stabellini <sstabellini@kernel.org>
>> Cc: Oleksandr Tyshchenko <oleksandr_tyshchenko@epam.com>
>> Cc: Guixiong Wei <weiguixiong@bytedance.com>
>> Cc: Jann Horn <jannh@google.com>
>> ---
>>   arch/x86/tools/relocs.c | 10 ++++++++++
>>   1 file changed, 10 insertions(+)
>> 
>> diff --git a/arch/x86/tools/relocs.c b/arch/x86/tools/relocs.c
>> index a3bae2b24626..0811fff23b9c 100644
>> --- a/arch/x86/tools/relocs.c
>> +++ b/arch/x86/tools/relocs.c
>> @@ -733,6 +733,16 @@ static void walk_relocs(int (*process)(struct section *sec, Elf_Rel *rel,
>>   		if (sec->shdr.sh_type != SHT_REL_TYPE) {
>>   			continue;
>>   		}
>> +
>> +		/*
>> +		 * Do not perform relocations in .notes section; any
>> +		 * values there are meant for pre-boot consumption (e.g.
>> +		 * startup_xen).
>> +		 */
>> +		if (strcmp(sec_name(sec->shdr.sh_info), ".notes") == 0) {
>
>Instead of a strcmp(), wouldnt't ...
>
>> +			continue;
>> +		}
>> +
>>   		sec_symtab  = sec->link;
>>   		sec_applies = &secs[sec->shdr.sh_info];
>>   		if (!(sec_applies->shdr.sh_flags & SHF_ALLOC)) {
>
>... a test of "sec_applies->shdr.sh_type == SHT_NOTE" work as well?
>
>In the end I'm fine with both variants, so:
>
>Reviewed-by: Juergen Gross <jgross@suse.com>
>
>
>Juergen

A type check would probably be better...