| Message ID | 20230621154314.6976-1-Yunxiang.Li@amd.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | drm/ttm: fix bulk_move corruption when adding a entry | expand |
Am 21.06.23 um 17:42 schrieb Yunxiang Li: > When the resource is the first in the bulk_move range, adding it again > (thus moving it to the tail) will corrupt the list since the first > pointer is not moved. This eventually lead to null pointer deref in > ttm_lru_bulk_move_del() Good catch. > > Fixes: fee2ede15542 ("drm/ttm: rework bulk move handling v5") > Signed-off-by: Yunxiang Li <Yunxiang.Li@amd.com> > --- > drivers/gpu/drm/ttm/ttm_resource.c | 7 ++++++- > 1 file changed, 6 insertions(+), 1 deletion(-) > > diff --git a/drivers/gpu/drm/ttm/ttm_resource.c b/drivers/gpu/drm/ttm/ttm_resource.c > index 7333f7a87a2f..4b84cfb5f916 100644 > --- a/drivers/gpu/drm/ttm/ttm_resource.c > +++ b/drivers/gpu/drm/ttm/ttm_resource.c > @@ -86,6 +86,9 @@ static void ttm_lru_bulk_move_pos_tail(struct ttm_lru_bulk_move_pos *pos, > struct ttm_resource *res) > { > if (pos->last != res) { > + if (pos->first == res) { > + pos->first = list_next_entry(res, lru); > + } Please drop the extra {}. > list_move(&res->lru, &pos->last->lru); > pos->last = res; > } > @@ -111,7 +114,9 @@ static void ttm_lru_bulk_move_del(struct ttm_lru_bulk_move *bulk, > { > struct ttm_lru_bulk_move_pos *pos = ttm_lru_bulk_move_pos(bulk, res); > > - if (unlikely(pos->first == res && pos->last == res)) { > + if (unlikely(!pos->first)) { > + WARN(1, "ttm_lru_bulk_move_pos is corrupt"); Rather do a WARN_ON(!pos->first || !pos->last) before the ifs and change the last if to check for pos->last. Regards, Christian. > + } else if (unlikely(pos->first == res && pos->last == res)) { > pos->first = NULL; > pos->last = NULL; > } else if (pos->first == res) {
diff --git a/drivers/gpu/drm/ttm/ttm_resource.c b/drivers/gpu/drm/ttm/ttm_resource.c index 7333f7a87a2f..4b84cfb5f916 100644 --- a/drivers/gpu/drm/ttm/ttm_resource.c +++ b/drivers/gpu/drm/ttm/ttm_resource.c @@ -86,6 +86,9 @@ static void ttm_lru_bulk_move_pos_tail(struct ttm_lru_bulk_move_pos *pos, struct ttm_resource *res) { if (pos->last != res) { + if (pos->first == res) { + pos->first = list_next_entry(res, lru); + } list_move(&res->lru, &pos->last->lru); pos->last = res; } @@ -111,7 +114,9 @@ static void ttm_lru_bulk_move_del(struct ttm_lru_bulk_move *bulk, { struct ttm_lru_bulk_move_pos *pos = ttm_lru_bulk_move_pos(bulk, res); - if (unlikely(pos->first == res && pos->last == res)) { + if (unlikely(!pos->first)) { + WARN(1, "ttm_lru_bulk_move_pos is corrupt"); + } else if (unlikely(pos->first == res && pos->last == res)) { pos->first = NULL; pos->last = NULL; } else if (pos->first == res) {
When the resource is the first in the bulk_move range, adding it again (thus moving it to the tail) will corrupt the list since the first pointer is not moved. This eventually lead to null pointer deref in ttm_lru_bulk_move_del() Fixes: fee2ede15542 ("drm/ttm: rework bulk move handling v5") Signed-off-by: Yunxiang Li <Yunxiang.Li@amd.com> --- drivers/gpu/drm/ttm/ttm_resource.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)