On Mon, 28 Mar 2022 12:09:59 +0200, Jiri Slaby wrote:
> On 28. 03. 22, 11:35, Xiaomeng Tong wrote:
> > The bug is here:
> > if (s->len != flen) {
> >
> > The list iterator 's' will point to a bogus position containing
> > HEAD if the list is empty or no element is found.
>
> Could you also explain how that can happen?
>
When list_for_each_entry_* do not early exits (if the list is empty
or no break/goto/return hit inside the loop), it will set pos ('s' here)
with a bogus pointer that point to a invalid struct computed based
on &HEAD using container_of.
#define list_for_each_entry(pos, head, member) \
for (pos = list_first_entry(head, typeof(*pos), member); \
!list_entry_is_head(pos, head, member); \
pos = list_next_entry(pos, member))
> > This case must
> > be checked before any use of the iterator, otherwise it may bpass
> > the 'if (s->len != flen) {' in theory iif s->len's value is flen,
>
> bpass + iif -- others already commented on that and you ignored them.
>
Thank you, i will correct it.
> > or/and lead to an invalid memory access.
> >
> > To fix this bug, use a new variable 'iter' as the list iterator,
> > while using the origin variable 's' as a dedicated pointer to
> > point to the found element. And if the list is empty or no element
> > is found, WARN_ON and return.
> >
> > Cc: [email protected]
> > Fixes: ^1da177e4c3f4 ("Linux-2.6.12-rc2")
>
> That's barely the commit introducing the behavior.
>
So just remove the Fixes tag? or something else? I find this commitID with
git blame.
--
Xiaomeng Tong
On 28. 03. 22, 12:27, Xiaomeng Tong wrote:
> On Mon, 28 Mar 2022 12:09:59 +0200, Jiri Slaby wrote:
>> On 28. 03. 22, 11:35, Xiaomeng Tong wrote:
>>> The bug is here:
>>> if (s->len != flen) {
>>>
>>> The list iterator 's' will point to a bogus position containing
>>> HEAD if the list is empty or no element is found.
>>
>> Could you also explain how that can happen?
>>
>
> When list_for_each_entry_* do not early exits (if the list is empty
> or no break/goto/return hit inside the loop), it will set pos ('s' here)
> with a bogus pointer that point to a invalid struct computed based
> on &HEAD using container_of.
>
> #define list_for_each_entry(pos, head, member) \
> for (pos = list_first_entry(head, typeof(*pos), member); \
> !list_entry_is_head(pos, head, member); \
> pos = list_next_entry(pos, member))
No, I didn't mean what happens on that site on the code level. I think
everyone understands that. Instead, I meant: what circumstances lead to
this _situation_ in reality?
thanks,
--
js
suse labs