Directly use __list_del() to remove 'p->plist' might be unsafe,
as we can see from the __list_del_entry_valid() of 'lib/list_debug.c'
that p->plist->prev or p->plist->next may be the LIST_POISON,
or p->list is not a valid double list.
In that case, __list_del() will be corruption.
Therefore, we suggest that __list_del_entry_valid()
should be added in front of the __list_del() in free_cache.
Fixes: bad5fa6 ("x86, microcode: Move to a proper location")
Signed-off-by: Jiasheng Jiang <[email protected]>
---
arch/x86/kernel/cpu/microcode/amd.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/x86/kernel/cpu/microcode/amd.c b/arch/x86/kernel/cpu/microcode/amd.c
index 3d4a483..e589cff 100644
--- a/arch/x86/kernel/cpu/microcode/amd.c
+++ b/arch/x86/kernel/cpu/microcode/amd.c
@@ -626,6 +626,8 @@ static void free_cache(void)
struct ucode_patch *p, *tmp;
list_for_each_entry_safe(p, tmp, µcode_cache, plist) {
+ if (!__list_del_entry_valid(&p->plist))
+ continue;
__list_del(p->plist.prev, p->plist.next);
kfree(p->data);
kfree(p);
--
2.7.4
On Tue, Sep 28, 2021 at 09:04:44AM +0000, Jiasheng Jiang wrote:
> Directly use __list_del() to remove 'p->plist' might be unsafe,
> as we can see from the __list_del_entry_valid() of 'lib/list_debug.c'
> that p->plist->prev or p->plist->next may be the LIST_POISON,
> or p->list is not a valid double list.
> In that case, __list_del() will be corruption.
> Therefore, we suggest that __list_del_entry_valid()
> should be added in front of the __list_del() in free_cache.
No one should be using list_del directly anyway, so please fix that.
On Tue, Sep 28, 2021 at 22:29:19PM +0000, Christoph Hellwig wrote:
>On Tue, Sep 28, 2021 at 09:04:44AM +0000, Jiasheng Jiang wrote:
>> Directly use __list_del() to remove 'p->plist' might be unsafe,
>> as we can see from the __list_del_entry_valid() of 'lib/list_debug.c'
>> that p->plist->prev or p->plist->next may be the LIST_POISON,
>> or p->list is not a valid double list.
>> In that case, __list_del() will be corruption.
>> Therefore, we suggest that __list_del_entry_valid()
>> should be added in front of the __list_del() in free_cache.
>No one should be using list_del directly anyway, so please fix that.
What the 'fix that' means?