On Thu, Dec 9, 2021 at 2:07 PM Christian Göttsche
<cgzones@xxxxxxxxxxxxxx> wrote:
>
> Add a libfuzz[1] based fuzzer testing the reading and parsing of binary policy
> files. This fuzzer will be run within the OSS-Fuzz service.
>
> Handle and reject a variety of edge cases causing crashes or resource leaks.
>
> The fifth patch ("libsepol/fuzz: limit element sizes for fuzzing") limits, to
> avoid oom reports from the fuzzer, caused by huge memory allocations, all
> identifiers to a length of 2^16 for the fuzzer build only.
> A potential limit for the release build needs further discussion.
>
> [1]: https://llvm.org/docs/LibFuzzer.html
>
> v3:
> - Drop RFC status
> - [10] libsepol: add checks for read sizes
> use PERM_SYMTAB_SIZE instead of bare 32 as limit
> - [11] libsepol: enforce avtab item limit
> take zero based numbering of variable items into account
> - [30] libsepol: validate ocontexts
> only check FS and NETIF ocons in selinux policies (not xen)
>
> v2:
> - reorder patches
> 1. oss-fuzz related
> 2. libsepol parsing and other crashesand UB
> 3. enhance policy validation
> - misc changes based on review by James Carter
>
>
> Christian Göttsche (36):
> cifuzz: enable report-unreproducible-crashes
> cifuzz: use the default runtime of 600 seconds
> libsepol/fuzz: silence secilc-fuzzer
> libsepol: add libfuzz based fuzzer for reading binary policies
> libsepol/fuzz: limit element sizes for fuzzing
> libsepol: use logging framework in conditional.c
> libsepol: use logging framework in ebitmap.c
> libsepol: use mallocarray wrapper to avoid overflows
> libsepol: use reallocarray wrapper to avoid overflows
> libsepol: add checks for read sizes
> libsepol: enforce avtab item limit
> libsepol: clean memory on conditional insertion failure
> libsepol: reject abnormal huge sid ids
> libsepol: reject invalid filetrans source type
> libsepol: zero member before potential dereference
> libsepol: use size_t for indexes in strs helpers
> libsepol: do not underflow on short format arguments
> libsepol: do not crash on class gaps
> libsepol: do not crash on user gaps
> libsepol: use correct size for initial string list
> libsepol: do not create a string list with initial size zero
> libsepol: split validation of datum array gaps and entries
> libsepol: validate MLS levels
> libsepol: validate expanded user range and level
> libsepol: validate permission count of classes
> libsepol: resolve log message mismatch
> libsepol: validate avtab and avrule types
> libsepol: validate constraint expression operators and attributes
> libsepol: validate type of avtab type rules
> libsepol: validate ocontexts
> libsepol: validate genfs contexts
> libsepol: validate permissive types
> libsepol: validate policy properties
> libsepol: validate categories
> libsepol: validate fsuse types
> libsepol: validate class default targets
>
> .github/workflows/cifuzz.yml | 3 +-
> libsepol/fuzz/binpolicy-fuzzer.c | 63 ++++
> libsepol/fuzz/policy.bin | Bin 0 -> 1552 bytes
> libsepol/fuzz/secilc-fuzzer.c | 5 +
> libsepol/src/Makefile | 6 +
> libsepol/src/avtab.c | 6 +
> libsepol/src/conditional.c | 53 ++--
> libsepol/src/ebitmap.c | 27 +-
> libsepol/src/expand.c | 4 +-
> libsepol/src/hashtab.c | 4 +-
> libsepol/src/kernel_to_cil.c | 10 +
> libsepol/src/kernel_to_common.c | 23 +-
> libsepol/src/kernel_to_common.h | 4 +-
> libsepol/src/kernel_to_conf.c | 13 +-
> libsepol/src/link.c | 3 +-
> libsepol/src/module.c | 4 +-
> libsepol/src/module_to_cil.c | 13 +-
> libsepol/src/optimize.c | 11 +-
> libsepol/src/policydb.c | 27 +-
> libsepol/src/policydb_validate.c | 477 +++++++++++++++++++++++++++----
> libsepol/src/private.h | 27 +-
> libsepol/src/services.c | 12 +-
> libsepol/src/sidtab.c | 3 +-
> libsepol/src/user_record.c | 8 +-
> libsepol/src/users.c | 12 +-
> libsepol/src/util.c | 11 +-
> libsepol/src/write.c | 2 +-
> scripts/oss-fuzz.sh | 17 +-
> 28 files changed, 686 insertions(+), 162 deletions(-)
> create mode 100644 libsepol/fuzz/binpolicy-fuzzer.c
> create mode 100644 libsepol/fuzz/policy.bin
>
> --
> 2.34.1
>
For the series:
Acked-by: James Carter <jwcart2@xxxxxxxxx>
As I noted, I will move the comment out of the expression when I apply patch 11.
Thanks,
Jim
