[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAJZ5v0hV2AFEgiuxxbDFUWLa0ZthSz3a=-9U4pjXm-GmmSgexw@mail.gmail.com>
Date: Fri, 28 Oct 2022 17:46:22 +0200
From: "Rafael J. Wysocki" <rafael@...nel.org>
To: Chen Zhongjin <chenzhongjin@...wei.com>, robert.moore@...el.com
Cc: linux-kernel@...r.kernel.org, linux-acpi@...r.kernel.org,
devel@...ica.org, rafael.j.wysocki@...el.com, lenb@...nel.org,
lv.zheng@...el.com
Subject: Re: [PATCH] ACPICA: Fix use-after-free in acpi_ps_parse_aml()
On Wed, Oct 19, 2022 at 9:38 AM Chen Zhongjin <chenzhongjin@...wei.com> wrote:
>
> KASAN reports a use-after-free problem and causes kernel panic
> triggered by: modprobe acpiphp_ibm
>
> BUG: KASAN:
> use-after-free in acpi_ds_dump_method_stack (drivers/acpi/acpica/dsdebug.c:145)
> Read of size 8 at addr ffff888002f843f0 by task modprobe/519
>
> CPU: 2 PID: 519 Comm: modprobe Not tainted 6.0.0+
> Hardware name: QEMU Standard PC (i440FX + PIIX, 1996)
> Call Trace:
> <TASK>
> acpi_ds_dump_method_stack (drivers/acpi/acpica/dsdebug.c:145)
> acpi_ds_method_error (drivers/acpi/acpica/dsmethod.c:232)
> acpi_ps_parse_aml (drivers/acpi/acpica/psparse.c:607)
> ...
> </TASK>
>
> Allocated by task 519:
> ...
> __kasan_kmalloc (mm/kasan/common.c:526)
> acpi_ds_create_walk_state (drivers/acpi/acpica/dswstate.c:519)
> acpi_ds_call_control_method (drivers/acpi/acpica/dsmethod.c:498)
> acpi_ps_parse_aml (drivers/acpi/acpica/psparse.c:607)
> ...
>
> Freed by task 519:
> ...
> __kmem_cache_free+0xb6/0x3c0
> acpi_ds_delete_walk_state (drivers/acpi/acpica/dswstate.c:722)
> acpi_ds_call_control_method (drivers/acpi/acpica/dsmethod.c:586)
> acpi_ps_parse_aml (drivers/acpi/acpica/psparse.c:607)
> ...
> ---[ end Kernel panic - not syncing: Fatal exception ]---
>
> In the error path in acpi_ps_parse_aml():
>
> acpi_ds_call_control_method()
> acpi_ds_create_walk_state()
> acpi_ds_push_walk_state()
> # thread->walk_state_list = walk_state
>
> acpi_ds_init_aml_walk # *fail*
> goto cleanup:
> acpi_ds_delete_walk_state() # ACPI_FREE(walk_state)
>
> acpi_ds_method_error()
> acpi_ds_dump_method_stack()
> # using freed thread->walk_state_list
>
> Briefly, the walk_state is pushed to thread, and freed without being poped.
> Then it is used in acpi_ds_dump_method_stack() and causes use-after-free.
>
> Add acpi_ds_pop_walk_state(thread) to the error path to fix the problem.
>
> Fixes: 0bac4295526c ("ACPICA: Dispatcher: Move stack traversal code to dispatcher")
>
> Signed-off-by: Chen Zhongjin <chenzhongjin@...wei.com>
This should be submitted to the upstream project on GitHub, but it
looks bad enough, so I'll take care of this.
Applied as 6.1-rc material, thanks!
> ---
> drivers/acpi/acpica/dsmethod.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/drivers/acpi/acpica/dsmethod.c b/drivers/acpi/acpica/dsmethod.c
> index ae2e768830bf..19da7fc73186 100644
> --- a/drivers/acpi/acpica/dsmethod.c
> +++ b/drivers/acpi/acpica/dsmethod.c
> @@ -581,6 +581,7 @@ acpi_ds_call_control_method(struct acpi_thread_state *thread,
>
> acpi_ds_terminate_control_method(obj_desc, next_walk_state);
> acpi_ds_delete_walk_state(next_walk_state);
> + acpi_ds_pop_walk_state(thread);
>
> return_ACPI_STATUS(status);
> }
> --
Bob, this looks correct to me, but I may be missing something in which
case please let me know.
Powered by blists - more mailing lists