[PATCH RESEND 0/1] Validating UAPI backwards compatibility

[John Moon]

+ linux-arm-kernel and Greg KH

Hi all,

The kernel community has rigorously enforced a policy of backwards
compatibility in its UAPI headers for a long time. This has allowed user
applications to enjoy stability across kernel upgrades without
recompiling.

In the vendor driver community (out-of-tree modules), there's been a
lack of discipline when it comes to maintaining UAPI backwards
compatibility. This has been a maintenance burden and limits our options
for long-term support of older devices.

Our goal is to add tooling for vendor driver developers because the
upstream model of expert maintainer code review can be difficult to
replicate in-house. Tools may help developers catch simple UAPI
incompatibilities that could be easily overlooked by in-house review.

We see in the kernel documentation:
"Kernel headers are backwards compatible, but not forwards compatible.
This means that a program built against a C library using older kernel
headers should run on a newer kernel (although it may not have access
to new features), but a program built against newer kernel headers may
not work on an older kernel."[1]

How does the kernel enforce this guarantee? We would be interested to
learn about any tools or methods used by kernel developers to make sure
the above statement remains true.

Could the documentation on UAPI maintenance (from a developer's point of
view) be expanded? Internally, we have a set of guidelines for our kernel
developers regarding UAPI compatibility techniques. If there's interest
in supplying a document on this topic with the kernel, we'd be happy to
submit a draft detailing what we have so far as a jumping off point.

Additionally, I've attached a shell script we've been using internally
to validate changes to our UAPI headers are backwards compatible. The
script uses libabigail's[2] tool abidiff[3] to compare a modified
header's ABI before and after a patch is applied. If an existing UAPI is
modified, the script exits non-zero. We use this script in our CI system
to block changes that fail the check.

Currently, the script works with gcc. It generates output like this when
a backwards-incompatible change is made to a UAPI header:

 !!! ABI differences detected in include/uapi/linux/acct.h (compared to
 file at HEAD^1) !!!

     [C] 'struct acct' changed:
       type size changed from 512 to 544 (in bits)
       1 data member insertion:
         '__u32 new_val', at offset 512 (in bits) at acct.h:71:1

 0/1 UAPI header file changes are backwards compatible
 UAPI header ABI check failed

However, we have not had success with clang. It seems clang is more
aggressive in optimizing dead code away (no matter which options we
pass). Therefore, no ABI differences are found.

We wanted to share with the community to receive feedback and any advice
when it comes to tooling/policy surrounding this issue. Our hope is that
the script will help all kernel UAPI authors (even those that haven't
upstreamed yet) maintain good discipline and avoid breaking userspace.

[1] Documentation/kbuild/headers_install.rst
[2] https://sourceware.org/libabigail/manual/libabigail-overview.html
[3] https://sourceware.org/libabigail/manual/abidiff.html

P.S. While at Qualcomm, Jordan Crouse <jorcrous@xxxxxxxxxx> authored the
original version of the UAPI checker script. Thanks Jordan!

John Moon (1):
  check-uapi: Introduce check-uapi.sh

 scripts/check-uapi.sh | 245 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 245 insertions(+)
 create mode 100755 scripts/check-uapi.sh


base-commit: 033c40a89f55525139fd5b6342281b09b97d05bf
--
2.17.1