From 17215b36b22d309a58a3b7bd08123f06e89657c9 Mon Sep 17 00:00:00 2001 From: Andrew Bartlett Date: Tue, 26 Nov 2019 15:44:32 +1300 Subject: CVE-2019-14902 dsdb: Explain that descriptor_sd_propagation_recursive() is proctected by a transaction This means we can trust the DB did not change between the two search requests. BUG: https://bugzilla.samba.org/show_bug.cgi?id=12497 Signed-off-by: Andrew Bartlett --- source4/dsdb/samdb/ldb_modules/descriptor.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/source4/dsdb/samdb/ldb_modules/descriptor.c b/source4/dsdb/samdb/ldb_modules/descriptor.c index 9018b750ab5..fb2854438e1 100644 --- a/source4/dsdb/samdb/ldb_modules/descriptor.c +++ b/source4/dsdb/samdb/ldb_modules/descriptor.c @@ -1199,6 +1199,9 @@ static int descriptor_sd_propagation_recursive(struct ldb_module *module, * LDB_SCOPE_SUBTREE searches are expensive. * * Note: that we do not search for deleted/recycled objects + * + * We know this is safe against a rename race as we are in the + * prepare_commit(), so must be in a transaction. */ ret = dsdb_module_search(module, change, -- cgit v1.2.1