Thread (10 messages) 10 messages, 5 authors, 2023-02-01

Broken Address Dependency in mm/ksm.c::cmp_and_merge_page()

From: Paul Heidekrüger <hidden>
Date: 2022-04-22 10:37:18
Also in: lkml, llvm

Hi all, 

My dependency checker is flagging yet another broken dependency. For
context, see [1].

Thankfully, it is fairly straight-forward to explain this time.
stable_node = page_stable_node(page);
Line 2032 in mm/ksm.c::cmp_and_merge_page() sees the return value of a
call to "page_stable_node()", which can depend on a "READ_ONCE()", being
assigned to "stable_node".
if (stable_node) {
        if (stable_node->head != &migrate_nodes &&
            get_kpfn_nid(READ_ONCE(stable_node->kpfn)) != 
            NUMA(stable_node->nid)) {
                stable_node_dup_del(stable_node); ‣dup: stable_node
                stable_node->head = &migrate_nodes;
                list_add(&stable_node->list, stable_node->head);
The dependency chain then runs into the two following if's, through an
assignment of "migrate_nodes" to "stable_node->head" (line 2038) and
finally reaches a call to "list_add()" (line 2039) where
"stable_node->head" gets passed as the second function argument. 
        }
}

static inline void list_add(struct list_head *new, struct list_head *head)
{
        __list_add(new, head, head->next);
}

static inline void __list_add(struct list_head *new,
                              struct list_head *prev,
                              struct list_head *next)
{
        if (!__list_add_valid(new, prev, next))
                return;

        next->prev = new;
        new->next = next;
        new->prev = prev;
        WRITE_ONCE(prev->next, new);
}
By being passed into "list_add()" via "stable_node->head", the
dependency chain eventually reaches a "WRITE_ONCE()" in "__list_add()"
whose destination address, "stable_node->head->next", is part of the
dependency chain and therefore carries an address dependency. 

However, as a result of the assignment in line 2038, Clang knows that
"stable_node->head" is "migrate_nodes" and replaces it, thereby breaking
the dependency chain. 

What do you think?

Many thanks,
Paul

--
[1]: https://lore.kernel.org/all/Yk7%2FT8BJITwz+Og1@Pauls-MacBook-Pro.local/ (local)
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help