On 7/19/19 6:11 PM, shuah wrote:
On 7/14/19 8:33 AM, Joe Lawrence wrote:
On Sun, Jul 14, 2019 at 10:28:29AM -0400, Joe Lawrence wrote:
>> [ ... snip ... ]
Testing:
Here's the output if modprobe --dry-run doesn't like the modules (not
built, etc.):
TAP version 13
selftests: livepatch: test-livepatch.sh
========================================
SKIP: Failed modprobe --dry-run of module: test_klp_livepatch
not ok 1..1 selftests: livepatch: test-livepatch.sh [SKIP]
selftests: livepatch: test-callbacks.sh
========================================
SKIP: Failed modprobe --dry-run of module: test_klp_callbacks_demo
not ok 1..2 selftests: livepatch: test-callbacks.sh [SKIP]
selftests: livepatch: test-shadow-vars.sh
========================================
SKIP: Failed modprobe --dry-run of module: test_klp_shadow_vars
not ok 1..3 selftests: livepatch: test-shadow-vars.sh [SKIP]
Please refine these messages to say what users should do. In addition
to what failed, also add what is missing - enable config option etc.
Hi Shuah,
Note that v2 was posted [1], but the output remains basically the same,
so your comments still apply.
Off the top of my head, modprobe can fail for many reasons: unprivileged
user, missing .ko files, missing modules.dep entry, kernel vermagic,
interface versions, etc.
What would you think about modifying our skip() function to provide a
catch-all list of CONFIG, environment, etc. requirements? Something
like, "Please ensure that the kernel was build with CONFIG_XYZ and that
the tests are run with foo privileges"? That would let us avoid trying
to figure out exactly why the modprobe failed, but still help out the user.
Alternatively, are there existing modprobe callers in the selftests that
already do a better job?
[1]
https://lore.kernel.org/linux-kselftest/eac825ab-04c2-77cf-671e-1a2a576109b0@xxxxxxxxxxxxxxxxxx/T/#mc2f70095215738cb6a539e2c2e6a415c78a8add6
Thanks,
-- Joe
