[rtems commit] Add a way to control the level of checks in the code
Sebastian Huber
sebh at rtems.org
Mon Mar 2 06:52:34 UTC 2020
Module: rtems
Branch: master
Commit: b44c574d8527355e862a5a67ccf8f333cdb9840b
Changeset: http://git.rtems.org/rtems/commit/?id=b44c574d8527355e862a5a67ccf8f333cdb9840b
Author: Simon Glass <sjg at chromium.org>
Date: Thu Feb 20 14:45:51 2020 -0700
Add a way to control the level of checks in the code
Add a new ASSUME_MASK option, which allows for some control over the
checks used in libfdt. With all assumptions enabled, libfdt assumes that
the input data and parameters are all correct and that internal errors
cannot happen.
By default no assumptions are made and all checks are enabled.
Signed-off-by: Simon Glass <sjg at chromium.org>
Message-Id: <20200220214557.176528-3-sjg at chromium.org>
Signed-off-by: David Gibson <david at gibson.dropbear.id.au>
---
cpukit/dtc/libfdt/libfdt_internal.h | 104 ++++++++++++++++++++++++++++++++++++
1 file changed, 104 insertions(+)
diff --git a/cpukit/dtc/libfdt/libfdt_internal.h b/cpukit/dtc/libfdt/libfdt_internal.h
index 058c735..e9913cd 100644
--- a/cpukit/dtc/libfdt/libfdt_internal.h
+++ b/cpukit/dtc/libfdt/libfdt_internal.h
@@ -48,4 +48,108 @@ static inline struct fdt_reserve_entry *fdt_mem_rsv_w_(void *fdt, int n)
#define FDT_SW_MAGIC (~FDT_MAGIC)
+/**********************************************************************/
+/* Checking controls */
+/**********************************************************************/
+
+#ifndef FDT_ASSUME_MASK
+#define FDT_ASSUME_MASK 0
+#endif
+
+/*
+ * Defines assumptions which can be enabled. Each of these can be enabled
+ * individually. For maximum saftey, don't enable any assumptions!
+ *
+ * For minimal code size and no safety, use ASSUME_PERFECT at your own risk.
+ * You should have another method of validating the device tree, such as a
+ * signature or hash check before using libfdt.
+ *
+ * For situations where security is not a concern it may be safe to enable
+ * ASSUME_SANE.
+ */
+enum {
+ /*
+ * This does essentially no checks. Only the latest device-tree
+ * version is correctly handled. Inconsistencies or errors in the device
+ * tree may cause undefined behaviour or crashes. Invalid parameters
+ * passed to libfdt may do the same.
+ *
+ * If an error occurs when modifying the tree it may leave the tree in
+ * an intermediate (but valid) state. As an example, adding a property
+ * where there is insufficient space may result in the property name
+ * being added to the string table even though the property itself is
+ * not added to the struct section.
+ *
+ * Only use this if you have a fully validated device tree with
+ * the latest supported version and wish to minimise code size.
+ */
+ ASSUME_PERFECT = 0xff,
+
+ /*
+ * This assumes that the device tree is sane. i.e. header metadata
+ * and basic hierarchy are correct.
+ *
+ * With this assumption enabled, normal device trees produced by libfdt
+ * and the compiler should be handled safely. Malicious device trees and
+ * complete garbage may cause libfdt to behave badly or crash.
+ *
+ * Note: Only checks that relate exclusively to the device tree itself
+ * (not the parameters passed to libfdt) are disabled by this
+ * assumption. This includes checking headers, tags and the like.
+ */
+ ASSUME_VALID_DTB = 1 << 0,
+
+ /*
+ * This builds on ASSUME_VALID_DTB and further assumes that libfdt
+ * functions are called with valid parameters, i.e. not trigger
+ * FDT_ERR_BADOFFSET or offsets that are out of bounds. It disables any
+ * extensive checking of parameters and the device tree, making various
+ * assumptions about correctness.
+ *
+ * It doesn't make sense to enable this assumption unless
+ * ASSUME_VALID_DTB is also enabled.
+ */
+ ASSUME_VALID_INPUT = 1 << 1,
+
+ /*
+ * This disables checks for device-tree version and removes all code
+ * which handles older versions.
+ *
+ * Only enable this if you know you have a device tree with the latest
+ * version.
+ */
+ ASSUME_LATEST = 1 << 2,
+
+ /*
+ * This assume that it is OK for a failed additional to the device tree
+ * due to lack of space or some other problem can skip any rollback
+ * steps (such as dropping the property name from the string table).
+ * This is safe to enable in most circumstances, even though it may
+ * leave the tree in a sub-optimal state.
+ */
+ ASSUME_NO_ROLLBACK = 1 << 3,
+
+ /*
+ * This assumes that the device tree components appear in the correct
+ * order. As such it disables a check in fdt_open_into() and removes the
+ * ability to fix the problem there. This is safe if you know that the
+ * device tree is correctly ordered. See fdt_blocks_misordered_().
+ */
+ ASSUME_LIBFDT_ORDER = 1 << 4,
+};
+
+/**
+ * can_assume_() - check if a particular assumption is enabled
+ *
+ * @mask: Mask to check (ASSUME_...)
+ * @return true if that assumption is enabled, else false
+ */
+static inline bool can_assume_(int mask)
+{
+ return FDT_ASSUME_MASK & mask;
+}
+
+/** helper macros for checking assumptions */
+#define can_assume(_assume) can_assume_(ASSUME_ ## _assume)
+
#endif /* LIBFDT_INTERNAL_H */
More information about the vc
mailing list