| Message ID | 20210126115510.53374-1-linmiaohe@huawei.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | mm/hugetlb: Fix use after free when subpool max_hpages accounting is not enabled | expand |
On 1/26/21 3:55 AM, Miaohe Lin wrote: > When subpool max_hpages accounting is not enabled, used_hpages is always 0 > and might lead to release subpool prematurely because it indicates no pages > are used now while there might be. It might be good to say that you need min_hpages accounting (min_size mount option) enabled for this issue to occur. Or, perhaps say this is possible if a hugetlbfs filesystem is created with the min_size option and without the size option. That might better explain the conditions in which a user could see the issue. > In order to fix this issue, we should check used_hpages == 0 iff max_hpages > accounting is enabled. As max_hpages accounting should be enabled in most > common case, this is not worth a Cc stable. I agree that such a combination of mount options is very uncommon. > > Signed-off-by: Hongxiang Lou <louhongxiang@huawei.com> > Signed-off-by: Miaohe Lin <linmiaohe@huawei.com> > --- > mm/hugetlb.c | 16 +++++++++++++--- > 1 file changed, 13 insertions(+), 3 deletions(-) Thanks, Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
On 2021/1/27 8:06, Mike Kravetz wrote: > On 1/26/21 3:55 AM, Miaohe Lin wrote: >> When subpool max_hpages accounting is not enabled, used_hpages is always 0 >> and might lead to release subpool prematurely because it indicates no pages >> are used now while there might be. > > It might be good to say that you need min_hpages accounting (min_size mount > option) enabled for this issue to occur. Or, perhaps say this is possible > if a hugetlbfs filesystem is created with the min_size option and without > the size option. > > That might better explain the conditions in which a user could see the issue. So commit log might looks like this ? """ If a hugetlbfs filesystem is created with the min_size option and without the size option, used_hpages is always 0 and might lead to release subpool prematurely because it indicates no pages are used now while there might be. In order to fix this issue, we should check used_hpages == 0 iff max_hpages accounting is enabled. As max_hpages accounting should be enabled in most common case, this is not worth a Cc stable. """ If so, should I send a V2 or Andrew would kindly do this? Many thanks. > >> In order to fix this issue, we should check used_hpages == 0 iff max_hpages >> accounting is enabled. As max_hpages accounting should be enabled in most >> common case, this is not worth a Cc stable. > > I agree that such a combination of mount options is very uncommon. > >> >> Signed-off-by: Hongxiang Lou <louhongxiang@huawei.com> >> Signed-off-by: Miaohe Lin <linmiaohe@huawei.com> >> --- >> mm/hugetlb.c | 16 +++++++++++++--- >> 1 file changed, 13 insertions(+), 3 deletions(-) > > Thanks, > > Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com> > Many thanks for review.
On Wed, 27 Jan 2021 10:08:22 +0800 Miaohe Lin <linmiaohe@huawei.com> wrote: > On 2021/1/27 8:06, Mike Kravetz wrote: > > On 1/26/21 3:55 AM, Miaohe Lin wrote: > >> When subpool max_hpages accounting is not enabled, used_hpages is always 0 > >> and might lead to release subpool prematurely because it indicates no pages > >> are used now while there might be. > > > > It might be good to say that you need min_hpages accounting (min_size mount > > option) enabled for this issue to occur. Or, perhaps say this is possible > > if a hugetlbfs filesystem is created with the min_size option and without > > the size option. > > > > That might better explain the conditions in which a user could see the issue. > > So commit log might looks like this ? > """ > If a hugetlbfs filesystem is created with the min_size option and without > the size option, used_hpages is always 0 and might lead to release subpool > prematurely because it indicates no pages are used now while there might > be. > > In order to fix this issue, we should check used_hpages == 0 iff max_hpages > accounting is enabled. As max_hpages accounting should be enabled in most > common case, this is not worth a Cc stable. > """ > > If so, should I send a V2 or Andrew would kindly do this? > Many thanks. I made that change, thanks.
Hi: On 2021/1/28 8:17, Andrew Morton wrote: > On Wed, 27 Jan 2021 10:08:22 +0800 Miaohe Lin <linmiaohe@huawei.com> wrote: > >> On 2021/1/27 8:06, Mike Kravetz wrote: >>> On 1/26/21 3:55 AM, Miaohe Lin wrote: >>>> When subpool max_hpages accounting is not enabled, used_hpages is always 0 >>>> and might lead to release subpool prematurely because it indicates no pages >>>> are used now while there might be. >>> >>> It might be good to say that you need min_hpages accounting (min_size mount >>> option) enabled for this issue to occur. Or, perhaps say this is possible >>> if a hugetlbfs filesystem is created with the min_size option and without >>> the size option. >>> >>> That might better explain the conditions in which a user could see the issue. >> >> So commit log might looks like this ? >> """ >> If a hugetlbfs filesystem is created with the min_size option and without >> the size option, used_hpages is always 0 and might lead to release subpool >> prematurely because it indicates no pages are used now while there might >> be. >> >> In order to fix this issue, we should check used_hpages == 0 iff max_hpages >> accounting is enabled. As max_hpages accounting should be enabled in most >> common case, this is not worth a Cc stable. >> """ >> >> If so, should I send a V2 or Andrew would kindly do this? >> Many thanks. > > I made that change, thanks.> It's very kind of you. Many thanks.:) > . >
diff --git a/mm/hugetlb.c b/mm/hugetlb.c index 777bc0e45bf3..53ea65d1c5ab 100644 --- a/mm/hugetlb.c +++ b/mm/hugetlb.c @@ -97,16 +97,26 @@ static inline void ClearPageHugeFreed(struct page *head) /* Forward declaration */ static int hugetlb_acct_memory(struct hstate *h, long delta); -static inline void unlock_or_release_subpool(struct hugepage_subpool *spool) +static inline bool subpool_is_free(struct hugepage_subpool *spool) { - bool free = (spool->count == 0) && (spool->used_hpages == 0); + if (spool->count) + return false; + if (spool->max_hpages != -1) + return spool->used_hpages == 0; + if (spool->min_hpages != -1) + return spool->rsv_hpages == spool->min_hpages; + return true; +} + +static inline void unlock_or_release_subpool(struct hugepage_subpool *spool) +{ spin_unlock(&spool->lock); /* If no pages are used, and no other handles to the subpool * remain, give up any reservations based on minimum size and * free the subpool */ - if (free) { + if (subpool_is_free(spool)) { if (spool->min_hpages != -1) hugetlb_acct_memory(spool->hstate, -spool->min_hpages);