[BUG REPORT] not ok 1 Derive Kc subkey for camellia128-cts-cmac

[BUG REPORT] not ok 1 Derive Kc subkey for camellia128-cts-cmac

[Nico Pache]

Hi,

One of the RFC 6803 key derivation kunit subtests is failing.

cki-project data warehouse : https://datawarehouse.cki-project.org/issue/2514

Arches: X86_64, ARM64, S390x, ppc64le
First Appeared: ~6.8.rc2

TRACE:
# Subtest: RFC 6803 key derivation
    # RFC 6803 key derivation: ASSERTION FAILED at net/sunrpc/auth_gss/gss_krb5_test.c:63
    Expected err == 0, but
        err == -110 (0xffffffffffffff92)
        not ok 1 Derive Kc subkey for camellia128-cts-cmac
        ok 2 Derive Ke subkey for camellia128-cts-cmac
        ok 3 Derive Ki subkey for camellia128-cts-cmac
        ok 4 Derive Kc subkey for camellia256-cts-cmac
        ok 5 Derive Ke subkey for camellia256-cts-cmac
        ok 6 Derive Ki subkey for camellia256-cts-cmac
    # RFC 6803 key derivation: pass:5 fail:1 skip:0 total:6
    not ok 1 RFC 6803 key derivation
-- 
2.44.0

Re: [BUG REPORT] not ok 1 Derive Kc subkey for camellia128-cts-cmac

[Scott Mayhew]

Hi,

On Thu, 07 Mar 2024, Nico Pache wrote:

> Hi,
> 
> One of the RFC 6803 key derivation kunit subtests is failing.
> 
> cki-project data warehouse : https://datawarehouse.cki-project.org/issue/2514
> 
> Arches: X86_64, ARM64, S390x, ppc64le
> First Appeared: ~6.8.rc2
> 
> TRACE:
> # Subtest: RFC 6803 key derivation
>     # RFC 6803 key derivation: ASSERTION FAILED at net/sunrpc/auth_gss/gss_krb5_test.c:63
>     Expected err == 0, but
>         err == -110 (0xffffffffffffff92)
>         not ok 1 Derive Kc subkey for camellia128-cts-cmac
>         ok 2 Derive Ke subkey for camellia128-cts-cmac
>         ok 3 Derive Ki subkey for camellia128-cts-cmac
>         ok 4 Derive Kc subkey for camellia256-cts-cmac
>         ok 5 Derive Ke subkey for camellia256-cts-cmac
>         ok 6 Derive Ki subkey for camellia256-cts-cmac
>     # RFC 6803 key derivation: pass:5 fail:1 skip:0 total:6
>     not ok 1 RFC 6803 key derivation

This was broken by:
c72a870926c2 kunit: add ability to run tests after boot using debugfs

__kunit_test_suites_init() runs any time a kernel module is loaded, via
the "kunit_mod_nb" notifier_block... even if the kernel module has no
kunit tests.  But now __kunit_test_suites_init() also locks a mutex,
which is a problem if a kunit test itself needs to load a kernel module
(which the gss_krb5_test module does).

This fixes it for me:

---8<---
diff --git a/lib/kunit/test.c b/lib/kunit/test.c
index 088489856db8..18af9453632b 100644
--- a/lib/kunit/test.c
+++ b/lib/kunit/test.c
@@ -707,6 +707,9 @@ int __kunit_test_suites_init(struct kunit_suite * const * const suites, int num_
 {
        unsigned int i;
 
+       if (num_suites == 0)
+               return 0;
+
        if (!kunit_enabled() && num_suites > 0) {
                pr_info("kunit: disabled\n");
                return 0;
---8<---

More detail below:

Here's the modprobe command where I loaded the gss_krb5_test module.  This
process has the "kunit_run_lock" mutex locked:

PID: 1468     TASK: ffff9aed0ac20000  CPU: 0    COMMAND: "modprobe"
 #0 [ffffba974196f6f8] __schedule at ffffffff83fd85f5
 #1 [ffffba974196f7b0] schedule at ffffffff83fd9672
 #2 [ffffba974196f7c8] schedule_timeout at ffffffff83fe0308
 #3 [ffffba974196f818] wait_for_completion_timeout at ffffffff83fda3d4
 #4 [ffffba974196f878] kunit_try_catch_run at ffffffffc0d5e851 [kunit]
 #5 [ffffba974196f8c8] kunit_run_tests at ffffffffc0d5c0ea [kunit]
 #6 [ffffba974196fb78] __kunit_test_suites_init at ffffffffc0d5c9af [kunit]
 #7 [ffffba974196fb98] kunit_module_notify at ffffffffc0d5ba4b [kunit]
 #8 [ffffba974196fc08] notifier_call_chain at ffffffff8314647a
 #9 [ffffba974196fc40] blocking_notifier_call_chain_robust at ffffffff83146565
#10 [ffffba974196fc88] load_module at ffffffff831e1935
#11 [ffffba974196fde8] __do_sys_init_module at ffffffff831e1fba
#12 [ffffba974196fec0] do_syscall_64 at ffffffff83fc3461
#13 [ffffba974196fee8] do_user_addr_fault at ffffffff830979df
#14 [ffffba974196ff28] exc_page_fault at ffffffff83fc9c7f
#15 [ffffba974196ff50] entry_SYSCALL_64_after_hwframe at ffffffff840000ea
    RIP: 00007ff1f272b4ae  RSP: 00007ffd45db8f68  RFLAGS: 00000246
    RAX: ffffffffffffffda  RBX: 000055bf4c0c4b20  RCX: 00007ff1f272b4ae
    RDX: 000055bf4b204e79  RSI: 0000000000099691  RDI: 000055bf4cbfd130
    RBP: 00007ffd45db9020   R8: 000055bf4c0c4010   R9: 0000000000000007
    R10: 0000000000000001  R11: 0000000000000246  R12: 000055bf4b204e79
    R13: 0000000000040000  R14: 000055bf4c0c4c50  R15: 000055bf4c0c4390
    ORIG_RAX: 00000000000000af  CS: 0033  SS: 002b

Here's the kunit test case running.  It's trying to allocate "cmac(camellia)"
via crypto_alloc_shash():

PID: 1508     TASK: ffff9aed155d0000  CPU: 1    COMMAND: "kunit_try_catch"
 #0 [ffffba974194fba0] __schedule at ffffffff83fd85f5
 #1 [ffffba974194fc58] schedule at ffffffff83fd9672
 #2 [ffffba974194fc70] schedule_timeout at ffffffff83fe0308
 #3 [ffffba974194fcc0] wait_for_completion_killable_timeout at ffffffff83fda708
 #4 [ffffba974194fd20] crypto_larval_wait at ffffffff83747fb4
 #5 [ffffba974194fd38] crypto_alg_mod_lookup at ffffffff83748252
 #6 [ffffba974194fd70] crypto_alloc_tfm_node at ffffffff83748492
 #7 [ffffba974194fdb0] krb5_kdf_feedback_cmac at ffffffffc0d76bb2 [rpcsec_gss_krb5]
 #8 [ffffba974194fe30] kdf_case at ffffffffc0d800a8 [gss_krb5_test]
 #9 [ffffba974194fe80] kunit_try_run_case at ffffffffc0d5bb54 [kunit]
#10 [ffffba974194fee8] kunit_generic_run_threadfn_adapter at ffffffffc0d5e797 [kunit]
#11 [ffffba974194fef8] kthread at ffffffff8313eda5
#12 [ffffba974194ff30] ret_from_fork at ffffffff830414a1
#13 [ffffba974194ff50] ret_from_fork_asm at ffffffff830039ab

Here the crypto manager is trying to modprobe the camellia kernel module via a
usermodehelper call:

PID: 1511     TASK: ffff9aed04630000  CPU: 3    COMMAND: "cryptomgr_probe"
 #0 [ffffba974195fb88] __schedule at ffffffff83fd85f5
 #1 [ffffba974195fc40] schedule at ffffffff83fd9672
 #2 [ffffba974195fc58] schedule_timeout at ffffffff83fe03c1
 #3 [ffffba974195fca8] wait_for_completion_state at ffffffff83fdb06d
 #4 [ffffba974195fd18] call_usermodehelper_exec at ffffffff83130313
 #5 [ffffba974195fd68] __request_module at ffffffff831e325d
 #6 [ffffba974195fe28] crypto_alg_mod_lookup at ffffffff83748220
 #7 [ffffba974195fe60] crypto_grab_spawn at ffffffff83749ff7
 #8 [ffffba974195fe98] cmac_create at ffffffff8375c2f0
 #9 [ffffba974195fed8] cryptomgr_probe at ffffffff83754a93
#10 [ffffba974195fef8] kthread at ffffffff8313eda5
#11 [ffffba974195ff30] ret_from_fork at ffffffff830414a1
#12 [ffffba974195ff50] ret_from_fork_asm at ffffffff830039ab

And here's the resulting modprobe command, which is stuck waiting on the
"kunit_run_lock" mutex:

PID: 1512     TASK: ffff9aed143fafc0  CPU: 2    COMMAND: "modprobe"
 #0 [ffffba9741957990] __schedule at ffffffff83fd85f5
 #1 [ffffba9741957a48] schedule at ffffffff83fd9672
 #2 [ffffba9741957a60] schedule_preempt_disabled at ffffffff83fd9cb5
 #3 [ffffba9741957a68] __mutex_lock.constprop.0 at ffffffff83fdc57a
 #4 [ffffba9741957ae8] __kunit_test_suites_init at ffffffffc0d5c95a [kunit]
 #5 [ffffba9741957b08] kunit_module_notify at ffffffffc0d5ba4b [kunit]
 #6 [ffffba9741957b78] notifier_call_chain at ffffffff8314647a
 #7 [ffffba9741957bb0] blocking_notifier_call_chain_robust at ffffffff83146565
 #8 [ffffba9741957bf8] load_module at ffffffff831e1935
 #9 [ffffba9741957d58] __do_sys_init_module at ffffffff831e1fba
#10 [ffffba9741957e30] do_syscall_64 at ffffffff83fc3461
#11 [ffffba9741957e48] __vm_munmap at ffffffff833bcdeb
#12 [ffffba9741957ee8] do_syscall_64 at ffffffff83fc3470
#13 [ffffba9741957f50] entry_SYSCALL_64_after_hwframe at ffffffff840000ea
    RIP: 00007f8ba092b4ae  RSP: 00007ffc771e0378  RFLAGS: 00000246
    RAX: ffffffffffffffda  RBX: 00005572137e6e40  RCX: 00007f8ba092b4ae
    RDX: 0000557211c4de79  RSI: 0000000000080451  RDI: 00007f8b9ff90010
    RBP: 00007ffc771e0430   R8: 00005572137e6010   R9: 0000000000000007
    R10: 0000000000000001  R11: 0000000000000246  R12: 0000557211c4de79
    R13: 0000000000040000  R14: 00005572137e73b0  R15: 00005572137e6400
    ORIG_RAX: 00000000000000af  CS: 0033  SS: 002b

The camellia module doesn't even have any kunit tests, so __kunit_test_suites_init()
is waiting to lock the "kunit_run_lock" mutex for nothing:

crash> module -o | grep num_kunit
  [0x478] int num_kunit_init_suites;
  [0x488] int num_kunit_suites;
crash> mod | grep camellia
ffffffffc0da15c0  camellia_x86_64                ffffffffc0d99000    57344  (not loaded)  [CONFIG_KALLSYMS]
crash> px 0xffffffffc0da15c0+0x478
$1 = 0xffffffffc0da1a38
crash> px 0xffffffffc0da15c0+0x488
$2 = 0xffffffffc0da1a48
crash> rd 0xffffffffc0da1a38
ffffffffc0da1a38:  0000000000000000                    ........
crash> rd 0xffffffffc0da1a48
ffffffffc0da1a48:  0000000000000000                    ........

-Scott
> -- 
> 2.44.0
> 
> 

Re: [BUG REPORT] not ok 1 Derive Kc subkey for camellia128-cts-cmac

[Nico Pache]

On Tue, Mar 19, 2024 at 11:51 AM Scott Mayhew <smayhew@redhat.com> wrote:
>
> Hi,
>
> On Thu, 07 Mar 2024, Nico Pache wrote:
>
> > Hi,
> >
> > One of the RFC 6803 key derivation kunit subtests is failing.
> >
> > cki-project data warehouse : https://datawarehouse.cki-project.org/issue/2514
> >
> > Arches: X86_64, ARM64, S390x, ppc64le
> > First Appeared: ~6.8.rc2
> >
> > TRACE:
> > # Subtest: RFC 6803 key derivation
> >     # RFC 6803 key derivation: ASSERTION FAILED at net/sunrpc/auth_gss/gss_krb5_test.c:63
> >     Expected err == 0, but
> >         err == -110 (0xffffffffffffff92)
> >         not ok 1 Derive Kc subkey for camellia128-cts-cmac
> >         ok 2 Derive Ke subkey for camellia128-cts-cmac
> >         ok 3 Derive Ki subkey for camellia128-cts-cmac
> >         ok 4 Derive Kc subkey for camellia256-cts-cmac
> >         ok 5 Derive Ke subkey for camellia256-cts-cmac
> >         ok 6 Derive Ki subkey for camellia256-cts-cmac
> >     # RFC 6803 key derivation: pass:5 fail:1 skip:0 total:6
> >     not ok 1 RFC 6803 key derivation
>
> This was broken by:
> c72a870926c2 kunit: add ability to run tests after boot using debugfs
>
> __kunit_test_suites_init() runs any time a kernel module is loaded, via
> the "kunit_mod_nb" notifier_block... even if the kernel module has no
> kunit tests.  But now __kunit_test_suites_init() also locks a mutex,
> which is a problem if a kunit test itself needs to load a kernel module
> (which the gss_krb5_test module does).
>
> This fixes it for me:
>
> ---8<---
> diff --git a/lib/kunit/test.c b/lib/kunit/test.c
> index 088489856db8..18af9453632b 100644
> --- a/lib/kunit/test.c
> +++ b/lib/kunit/test.c
> @@ -707,6 +707,9 @@ int __kunit_test_suites_init(struct kunit_suite * const * const suites, int num_
>  {
>         unsigned int i;
>
> +       if (num_suites == 0)
> +               return 0;
> +
>         if (!kunit_enabled() && num_suites > 0) {
>                 pr_info("kunit: disabled\n");
>                 return 0;
> ---8<---
>
Nice find! Would you mind posting a patch?

-- Nico

> More detail below:
>
> Here's the modprobe command where I loaded the gss_krb5_test module.  This
> process has the "kunit_run_lock" mutex locked:
>
> PID: 1468     TASK: ffff9aed0ac20000  CPU: 0    COMMAND: "modprobe"
>  #0 [ffffba974196f6f8] __schedule at ffffffff83fd85f5
>  #1 [ffffba974196f7b0] schedule at ffffffff83fd9672
>  #2 [ffffba974196f7c8] schedule_timeout at ffffffff83fe0308
>  #3 [ffffba974196f818] wait_for_completion_timeout at ffffffff83fda3d4
>  #4 [ffffba974196f878] kunit_try_catch_run at ffffffffc0d5e851 [kunit]
>  #5 [ffffba974196f8c8] kunit_run_tests at ffffffffc0d5c0ea [kunit]
>  #6 [ffffba974196fb78] __kunit_test_suites_init at ffffffffc0d5c9af [kunit]
>  #7 [ffffba974196fb98] kunit_module_notify at ffffffffc0d5ba4b [kunit]
>  #8 [ffffba974196fc08] notifier_call_chain at ffffffff8314647a
>  #9 [ffffba974196fc40] blocking_notifier_call_chain_robust at ffffffff83146565
> #10 [ffffba974196fc88] load_module at ffffffff831e1935
> #11 [ffffba974196fde8] __do_sys_init_module at ffffffff831e1fba
> #12 [ffffba974196fec0] do_syscall_64 at ffffffff83fc3461
> #13 [ffffba974196fee8] do_user_addr_fault at ffffffff830979df
> #14 [ffffba974196ff28] exc_page_fault at ffffffff83fc9c7f
> #15 [ffffba974196ff50] entry_SYSCALL_64_after_hwframe at ffffffff840000ea
>     RIP: 00007ff1f272b4ae  RSP: 00007ffd45db8f68  RFLAGS: 00000246
>     RAX: ffffffffffffffda  RBX: 000055bf4c0c4b20  RCX: 00007ff1f272b4ae
>     RDX: 000055bf4b204e79  RSI: 0000000000099691  RDI: 000055bf4cbfd130
>     RBP: 00007ffd45db9020   R8: 000055bf4c0c4010   R9: 0000000000000007
>     R10: 0000000000000001  R11: 0000000000000246  R12: 000055bf4b204e79
>     R13: 0000000000040000  R14: 000055bf4c0c4c50  R15: 000055bf4c0c4390
>     ORIG_RAX: 00000000000000af  CS: 0033  SS: 002b
>
> Here's the kunit test case running.  It's trying to allocate "cmac(camellia)"
> via crypto_alloc_shash():
>
> PID: 1508     TASK: ffff9aed155d0000  CPU: 1    COMMAND: "kunit_try_catch"
>  #0 [ffffba974194fba0] __schedule at ffffffff83fd85f5
>  #1 [ffffba974194fc58] schedule at ffffffff83fd9672
>  #2 [ffffba974194fc70] schedule_timeout at ffffffff83fe0308
>  #3 [ffffba974194fcc0] wait_for_completion_killable_timeout at ffffffff83fda708
>  #4 [ffffba974194fd20] crypto_larval_wait at ffffffff83747fb4
>  #5 [ffffba974194fd38] crypto_alg_mod_lookup at ffffffff83748252
>  #6 [ffffba974194fd70] crypto_alloc_tfm_node at ffffffff83748492
>  #7 [ffffba974194fdb0] krb5_kdf_feedback_cmac at ffffffffc0d76bb2 [rpcsec_gss_krb5]
>  #8 [ffffba974194fe30] kdf_case at ffffffffc0d800a8 [gss_krb5_test]
>  #9 [ffffba974194fe80] kunit_try_run_case at ffffffffc0d5bb54 [kunit]
> #10 [ffffba974194fee8] kunit_generic_run_threadfn_adapter at ffffffffc0d5e797 [kunit]
> #11 [ffffba974194fef8] kthread at ffffffff8313eda5
> #12 [ffffba974194ff30] ret_from_fork at ffffffff830414a1
> #13 [ffffba974194ff50] ret_from_fork_asm at ffffffff830039ab
>
> Here the crypto manager is trying to modprobe the camellia kernel module via a
> usermodehelper call:
>
> PID: 1511     TASK: ffff9aed04630000  CPU: 3    COMMAND: "cryptomgr_probe"
>  #0 [ffffba974195fb88] __schedule at ffffffff83fd85f5
>  #1 [ffffba974195fc40] schedule at ffffffff83fd9672
>  #2 [ffffba974195fc58] schedule_timeout at ffffffff83fe03c1
>  #3 [ffffba974195fca8] wait_for_completion_state at ffffffff83fdb06d
>  #4 [ffffba974195fd18] call_usermodehelper_exec at ffffffff83130313
>  #5 [ffffba974195fd68] __request_module at ffffffff831e325d
>  #6 [ffffba974195fe28] crypto_alg_mod_lookup at ffffffff83748220
>  #7 [ffffba974195fe60] crypto_grab_spawn at ffffffff83749ff7
>  #8 [ffffba974195fe98] cmac_create at ffffffff8375c2f0
>  #9 [ffffba974195fed8] cryptomgr_probe at ffffffff83754a93
> #10 [ffffba974195fef8] kthread at ffffffff8313eda5
> #11 [ffffba974195ff30] ret_from_fork at ffffffff830414a1
> #12 [ffffba974195ff50] ret_from_fork_asm at ffffffff830039ab
>
> And here's the resulting modprobe command, which is stuck waiting on the
> "kunit_run_lock" mutex:
>
> PID: 1512     TASK: ffff9aed143fafc0  CPU: 2    COMMAND: "modprobe"
>  #0 [ffffba9741957990] __schedule at ffffffff83fd85f5
>  #1 [ffffba9741957a48] schedule at ffffffff83fd9672
>  #2 [ffffba9741957a60] schedule_preempt_disabled at ffffffff83fd9cb5
>  #3 [ffffba9741957a68] __mutex_lock.constprop.0 at ffffffff83fdc57a
>  #4 [ffffba9741957ae8] __kunit_test_suites_init at ffffffffc0d5c95a [kunit]
>  #5 [ffffba9741957b08] kunit_module_notify at ffffffffc0d5ba4b [kunit]
>  #6 [ffffba9741957b78] notifier_call_chain at ffffffff8314647a
>  #7 [ffffba9741957bb0] blocking_notifier_call_chain_robust at ffffffff83146565
>  #8 [ffffba9741957bf8] load_module at ffffffff831e1935
>  #9 [ffffba9741957d58] __do_sys_init_module at ffffffff831e1fba
> #10 [ffffba9741957e30] do_syscall_64 at ffffffff83fc3461
> #11 [ffffba9741957e48] __vm_munmap at ffffffff833bcdeb
> #12 [ffffba9741957ee8] do_syscall_64 at ffffffff83fc3470
> #13 [ffffba9741957f50] entry_SYSCALL_64_after_hwframe at ffffffff840000ea
>     RIP: 00007f8ba092b4ae  RSP: 00007ffc771e0378  RFLAGS: 00000246
>     RAX: ffffffffffffffda  RBX: 00005572137e6e40  RCX: 00007f8ba092b4ae
>     RDX: 0000557211c4de79  RSI: 0000000000080451  RDI: 00007f8b9ff90010
>     RBP: 00007ffc771e0430   R8: 00005572137e6010   R9: 0000000000000007
>     R10: 0000000000000001  R11: 0000000000000246  R12: 0000557211c4de79
>     R13: 0000000000040000  R14: 00005572137e73b0  R15: 00005572137e6400
>     ORIG_RAX: 00000000000000af  CS: 0033  SS: 002b
>
> The camellia module doesn't even have any kunit tests, so __kunit_test_suites_init()
> is waiting to lock the "kunit_run_lock" mutex for nothing:
>
> crash> module -o | grep num_kunit
>   [0x478] int num_kunit_init_suites;
>   [0x488] int num_kunit_suites;
> crash> mod | grep camellia
> ffffffffc0da15c0  camellia_x86_64                ffffffffc0d99000    57344  (not loaded)  [CONFIG_KALLSYMS]
> crash> px 0xffffffffc0da15c0+0x478
> $1 = 0xffffffffc0da1a38
> crash> px 0xffffffffc0da15c0+0x488
> $2 = 0xffffffffc0da1a48
> crash> rd 0xffffffffc0da1a38
> ffffffffc0da1a38:  0000000000000000                    ........
> crash> rd 0xffffffffc0da1a48
> ffffffffc0da1a48:  0000000000000000                    ........
>
> -Scott
> > --
> > 2.44.0
> >
> >
>

Re: [BUG REPORT] not ok 1 Derive Kc subkey for camellia128-cts-cmac

[Scott Mayhew]

On Wed, 20 Mar 2024, Nico Pache wrote:

> On Tue, Mar 19, 2024 at 11:51 AM Scott Mayhew <smayhew@redhat.com> wrote:
> >
> > Hi,
> >
> > On Thu, 07 Mar 2024, Nico Pache wrote:
> >
> > > Hi,
> > >
> > > One of the RFC 6803 key derivation kunit subtests is failing.
> > >
> > > cki-project data warehouse : https://datawarehouse.cki-project.org/issue/2514
> > >
> > > Arches: X86_64, ARM64, S390x, ppc64le
> > > First Appeared: ~6.8.rc2
> > >
> > > TRACE:
> > > # Subtest: RFC 6803 key derivation
> > >     # RFC 6803 key derivation: ASSERTION FAILED at net/sunrpc/auth_gss/gss_krb5_test.c:63
> > >     Expected err == 0, but
> > >         err == -110 (0xffffffffffffff92)
> > >         not ok 1 Derive Kc subkey for camellia128-cts-cmac
> > >         ok 2 Derive Ke subkey for camellia128-cts-cmac
> > >         ok 3 Derive Ki subkey for camellia128-cts-cmac
> > >         ok 4 Derive Kc subkey for camellia256-cts-cmac
> > >         ok 5 Derive Ke subkey for camellia256-cts-cmac
> > >         ok 6 Derive Ki subkey for camellia256-cts-cmac
> > >     # RFC 6803 key derivation: pass:5 fail:1 skip:0 total:6
> > >     not ok 1 RFC 6803 key derivation
> >
> > This was broken by:
> > c72a870926c2 kunit: add ability to run tests after boot using debugfs
> >
> > __kunit_test_suites_init() runs any time a kernel module is loaded, via
> > the "kunit_mod_nb" notifier_block... even if the kernel module has no
> > kunit tests.  But now __kunit_test_suites_init() also locks a mutex,
> > which is a problem if a kunit test itself needs to load a kernel module
> > (which the gss_krb5_test module does).
> >
> > This fixes it for me:
> >
> > ---8<---
> > diff --git a/lib/kunit/test.c b/lib/kunit/test.c
> > index 088489856db8..18af9453632b 100644
> > --- a/lib/kunit/test.c
> > +++ b/lib/kunit/test.c
> > @@ -707,6 +707,9 @@ int __kunit_test_suites_init(struct kunit_suite * const * const suites, int num_
> >  {
> >         unsigned int i;
> >
> > +       if (num_suites == 0)
> > +               return 0;
> > +
> >         if (!kunit_enabled() && num_suites > 0) {
> >                 pr_info("kunit: disabled\n");
> >                 return 0;
> > ---8<---
> >
> Nice find! Would you mind posting a patch?

Yes - while testing I also found an issue with the RFC 8009 encryption
test and I wanted to be sure it wasn't related to this issue before
posting the patch.

-Scott
> 
> -- Nico
> 
> > More detail below:
> >
> > Here's the modprobe command where I loaded the gss_krb5_test module.  This
> > process has the "kunit_run_lock" mutex locked:
> >
> > PID: 1468     TASK: ffff9aed0ac20000  CPU: 0    COMMAND: "modprobe"
> >  #0 [ffffba974196f6f8] __schedule at ffffffff83fd85f5
> >  #1 [ffffba974196f7b0] schedule at ffffffff83fd9672
> >  #2 [ffffba974196f7c8] schedule_timeout at ffffffff83fe0308
> >  #3 [ffffba974196f818] wait_for_completion_timeout at ffffffff83fda3d4
> >  #4 [ffffba974196f878] kunit_try_catch_run at ffffffffc0d5e851 [kunit]
> >  #5 [ffffba974196f8c8] kunit_run_tests at ffffffffc0d5c0ea [kunit]
> >  #6 [ffffba974196fb78] __kunit_test_suites_init at ffffffffc0d5c9af [kunit]
> >  #7 [ffffba974196fb98] kunit_module_notify at ffffffffc0d5ba4b [kunit]
> >  #8 [ffffba974196fc08] notifier_call_chain at ffffffff8314647a
> >  #9 [ffffba974196fc40] blocking_notifier_call_chain_robust at ffffffff83146565
> > #10 [ffffba974196fc88] load_module at ffffffff831e1935
> > #11 [ffffba974196fde8] __do_sys_init_module at ffffffff831e1fba
> > #12 [ffffba974196fec0] do_syscall_64 at ffffffff83fc3461
> > #13 [ffffba974196fee8] do_user_addr_fault at ffffffff830979df
> > #14 [ffffba974196ff28] exc_page_fault at ffffffff83fc9c7f
> > #15 [ffffba974196ff50] entry_SYSCALL_64_after_hwframe at ffffffff840000ea
> >     RIP: 00007ff1f272b4ae  RSP: 00007ffd45db8f68  RFLAGS: 00000246
> >     RAX: ffffffffffffffda  RBX: 000055bf4c0c4b20  RCX: 00007ff1f272b4ae
> >     RDX: 000055bf4b204e79  RSI: 0000000000099691  RDI: 000055bf4cbfd130
> >     RBP: 00007ffd45db9020   R8: 000055bf4c0c4010   R9: 0000000000000007
> >     R10: 0000000000000001  R11: 0000000000000246  R12: 000055bf4b204e79
> >     R13: 0000000000040000  R14: 000055bf4c0c4c50  R15: 000055bf4c0c4390
> >     ORIG_RAX: 00000000000000af  CS: 0033  SS: 002b
> >
> > Here's the kunit test case running.  It's trying to allocate "cmac(camellia)"
> > via crypto_alloc_shash():
> >
> > PID: 1508     TASK: ffff9aed155d0000  CPU: 1    COMMAND: "kunit_try_catch"
> >  #0 [ffffba974194fba0] __schedule at ffffffff83fd85f5
> >  #1 [ffffba974194fc58] schedule at ffffffff83fd9672
> >  #2 [ffffba974194fc70] schedule_timeout at ffffffff83fe0308
> >  #3 [ffffba974194fcc0] wait_for_completion_killable_timeout at ffffffff83fda708
> >  #4 [ffffba974194fd20] crypto_larval_wait at ffffffff83747fb4
> >  #5 [ffffba974194fd38] crypto_alg_mod_lookup at ffffffff83748252
> >  #6 [ffffba974194fd70] crypto_alloc_tfm_node at ffffffff83748492
> >  #7 [ffffba974194fdb0] krb5_kdf_feedback_cmac at ffffffffc0d76bb2 [rpcsec_gss_krb5]
> >  #8 [ffffba974194fe30] kdf_case at ffffffffc0d800a8 [gss_krb5_test]
> >  #9 [ffffba974194fe80] kunit_try_run_case at ffffffffc0d5bb54 [kunit]
> > #10 [ffffba974194fee8] kunit_generic_run_threadfn_adapter at ffffffffc0d5e797 [kunit]
> > #11 [ffffba974194fef8] kthread at ffffffff8313eda5
> > #12 [ffffba974194ff30] ret_from_fork at ffffffff830414a1
> > #13 [ffffba974194ff50] ret_from_fork_asm at ffffffff830039ab
> >
> > Here the crypto manager is trying to modprobe the camellia kernel module via a
> > usermodehelper call:
> >
> > PID: 1511     TASK: ffff9aed04630000  CPU: 3    COMMAND: "cryptomgr_probe"
> >  #0 [ffffba974195fb88] __schedule at ffffffff83fd85f5
> >  #1 [ffffba974195fc40] schedule at ffffffff83fd9672
> >  #2 [ffffba974195fc58] schedule_timeout at ffffffff83fe03c1
> >  #3 [ffffba974195fca8] wait_for_completion_state at ffffffff83fdb06d
> >  #4 [ffffba974195fd18] call_usermodehelper_exec at ffffffff83130313
> >  #5 [ffffba974195fd68] __request_module at ffffffff831e325d
> >  #6 [ffffba974195fe28] crypto_alg_mod_lookup at ffffffff83748220
> >  #7 [ffffba974195fe60] crypto_grab_spawn at ffffffff83749ff7
> >  #8 [ffffba974195fe98] cmac_create at ffffffff8375c2f0
> >  #9 [ffffba974195fed8] cryptomgr_probe at ffffffff83754a93
> > #10 [ffffba974195fef8] kthread at ffffffff8313eda5
> > #11 [ffffba974195ff30] ret_from_fork at ffffffff830414a1
> > #12 [ffffba974195ff50] ret_from_fork_asm at ffffffff830039ab
> >
> > And here's the resulting modprobe command, which is stuck waiting on the
> > "kunit_run_lock" mutex:
> >
> > PID: 1512     TASK: ffff9aed143fafc0  CPU: 2    COMMAND: "modprobe"
> >  #0 [ffffba9741957990] __schedule at ffffffff83fd85f5
> >  #1 [ffffba9741957a48] schedule at ffffffff83fd9672
> >  #2 [ffffba9741957a60] schedule_preempt_disabled at ffffffff83fd9cb5
> >  #3 [ffffba9741957a68] __mutex_lock.constprop.0 at ffffffff83fdc57a
> >  #4 [ffffba9741957ae8] __kunit_test_suites_init at ffffffffc0d5c95a [kunit]
> >  #5 [ffffba9741957b08] kunit_module_notify at ffffffffc0d5ba4b [kunit]
> >  #6 [ffffba9741957b78] notifier_call_chain at ffffffff8314647a
> >  #7 [ffffba9741957bb0] blocking_notifier_call_chain_robust at ffffffff83146565
> >  #8 [ffffba9741957bf8] load_module at ffffffff831e1935
> >  #9 [ffffba9741957d58] __do_sys_init_module at ffffffff831e1fba
> > #10 [ffffba9741957e30] do_syscall_64 at ffffffff83fc3461
> > #11 [ffffba9741957e48] __vm_munmap at ffffffff833bcdeb
> > #12 [ffffba9741957ee8] do_syscall_64 at ffffffff83fc3470
> > #13 [ffffba9741957f50] entry_SYSCALL_64_after_hwframe at ffffffff840000ea
> >     RIP: 00007f8ba092b4ae  RSP: 00007ffc771e0378  RFLAGS: 00000246
> >     RAX: ffffffffffffffda  RBX: 00005572137e6e40  RCX: 00007f8ba092b4ae
> >     RDX: 0000557211c4de79  RSI: 0000000000080451  RDI: 00007f8b9ff90010
> >     RBP: 00007ffc771e0430   R8: 00005572137e6010   R9: 0000000000000007
> >     R10: 0000000000000001  R11: 0000000000000246  R12: 0000557211c4de79
> >     R13: 0000000000040000  R14: 00005572137e73b0  R15: 00005572137e6400
> >     ORIG_RAX: 00000000000000af  CS: 0033  SS: 002b
> >
> > The camellia module doesn't even have any kunit tests, so __kunit_test_suites_init()
> > is waiting to lock the "kunit_run_lock" mutex for nothing:
> >
> > crash> module -o | grep num_kunit
> >   [0x478] int num_kunit_init_suites;
> >   [0x488] int num_kunit_suites;
> > crash> mod | grep camellia
> > ffffffffc0da15c0  camellia_x86_64                ffffffffc0d99000    57344  (not loaded)  [CONFIG_KALLSYMS]
> > crash> px 0xffffffffc0da15c0+0x478
> > $1 = 0xffffffffc0da1a38
> > crash> px 0xffffffffc0da15c0+0x488
> > $2 = 0xffffffffc0da1a48
> > crash> rd 0xffffffffc0da1a38
> > ffffffffc0da1a38:  0000000000000000                    ........
> > crash> rd 0xffffffffc0da1a48
> > ffffffffc0da1a48:  0000000000000000                    ........
> >
> > -Scott
> > > --
> > > 2.44.0
> > >
> > >
> >
> 

[PATCH] kunit: bail out early in __kunit_test_suites_init() if there are no suites to test

[Scott Mayhew]

Commit c72a870926c2 added a mutex to prevent kunit tests from running
concurrently.  Unfortunately that mutex gets locked during module load
regardless of whether the module actually has any kunit tests.  This
causes a problem for kunit tests that might need to load other kernel
modules (e.g. gss_krb5_test loading the camellia module).

So check to see if there are actually any tests to run before locking
the kunit_run_lock mutex.

Fixes: c72a870926c2 ("kunit: add ability to run tests after boot using debugfs")
Reported-by: Nico Pache <npache@redhat.com>
Signed-off-by: Scott Mayhew <smayhew@redhat.com>
---
 lib/kunit/test.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/lib/kunit/test.c b/lib/kunit/test.c
index 1d1475578515..b8514dbb337c 100644
--- a/lib/kunit/test.c
+++ b/lib/kunit/test.c
@@ -712,6 +712,9 @@ int __kunit_test_suites_init(struct kunit_suite * const * const suites, int num_
 {
 	unsigned int i;
 
+	if (num_suites == 0)
+		return 0;
+
 	if (!kunit_enabled() && num_suites > 0) {
 		pr_info("kunit: disabled\n");
 		return 0;
-- 
2.43.0

Re: [PATCH] kunit: bail out early in __kunit_test_suites_init() if there are no suites to test

[Rae Moar]

On Thu, Mar 21, 2024 at 10:32 AM Scott Mayhew <smayhew@redhat.com> wrote:
>
> Commit c72a870926c2 added a mutex to prevent kunit tests from running
> concurrently.  Unfortunately that mutex gets locked during module load
> regardless of whether the module actually has any kunit tests.  This
> causes a problem for kunit tests that might need to load other kernel
> modules (e.g. gss_krb5_test loading the camellia module).
>
> So check to see if there are actually any tests to run before locking
> the kunit_run_lock mutex.
>
> Fixes: c72a870926c2 ("kunit: add ability to run tests after boot using debugfs")
> Reported-by: Nico Pache <npache@redhat.com>
> Signed-off-by: Scott Mayhew <smayhew@redhat.com>

Hi!

Sorry about this bug. Thanks for the patch! We should definitely add this check.

Reviewed-by: Rae Moar <rmoar@google.com>

Thanks!

-Rae

> ---
>  lib/kunit/test.c | 3 +++
>  1 file changed, 3 insertions(+)
>
> diff --git a/lib/kunit/test.c b/lib/kunit/test.c
> index 1d1475578515..b8514dbb337c 100644
> --- a/lib/kunit/test.c
> +++ b/lib/kunit/test.c
> @@ -712,6 +712,9 @@ int __kunit_test_suites_init(struct kunit_suite * const * const suites, int num_
>  {
>         unsigned int i;
>
> +       if (num_suites == 0)
> +               return 0;
> +
>         if (!kunit_enabled() && num_suites > 0) {
>                 pr_info("kunit: disabled\n");
>                 return 0;
> --
> 2.43.0
>

Re: [PATCH] kunit: bail out early in __kunit_test_suites_init() if there are no suites to test

[David Gow]

[-- Attachment #1: Type: text/plain, Size: 1385 bytes --]

On Thu, 21 Mar 2024 at 22:32, Scott Mayhew <smayhew@redhat.com> wrote:
>
> Commit c72a870926c2 added a mutex to prevent kunit tests from running
> concurrently.  Unfortunately that mutex gets locked during module load
> regardless of whether the module actually has any kunit tests.  This
> causes a problem for kunit tests that might need to load other kernel
> modules (e.g. gss_krb5_test loading the camellia module).
>
> So check to see if there are actually any tests to run before locking
> the kunit_run_lock mutex.
>
> Fixes: c72a870926c2 ("kunit: add ability to run tests after boot using debugfs")
> Reported-by: Nico Pache <npache@redhat.com>
> Signed-off-by: Scott Mayhew <smayhew@redhat.com>
> ---

Thanks, this works well here, and is a good idea anyway.

Reviewed-by: David Gow <davidgow@google.com>

Cheers,
-- David

>  lib/kunit/test.c | 3 +++
>  1 file changed, 3 insertions(+)
>
> diff --git a/lib/kunit/test.c b/lib/kunit/test.c
> index 1d1475578515..b8514dbb337c 100644
> --- a/lib/kunit/test.c
> +++ b/lib/kunit/test.c
> @@ -712,6 +712,9 @@ int __kunit_test_suites_init(struct kunit_suite * const * const suites, int num_
>  {
>         unsigned int i;
>
> +       if (num_suites == 0)
> +               return 0;
> +
>         if (!kunit_enabled() && num_suites > 0) {
>                 pr_info("kunit: disabled\n");
>                 return 0;
> --
> 2.43.0
>

[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4014 bytes --]

Re: [PATCH] kunit: bail out early in __kunit_test_suites_init() if there are no suites to test

[Scott Mayhew]

On Sat, 23 Mar 2024, David Gow wrote:

> On Thu, 21 Mar 2024 at 22:32, Scott Mayhew <smayhew@redhat.com> wrote:
> >
> > Commit c72a870926c2 added a mutex to prevent kunit tests from running
> > concurrently.  Unfortunately that mutex gets locked during module load
> > regardless of whether the module actually has any kunit tests.  This
> > causes a problem for kunit tests that might need to load other kernel
> > modules (e.g. gss_krb5_test loading the camellia module).
> >
> > So check to see if there are actually any tests to run before locking
> > the kunit_run_lock mutex.
> >
> > Fixes: c72a870926c2 ("kunit: add ability to run tests after boot using debugfs")
> > Reported-by: Nico Pache <npache@redhat.com>
> > Signed-off-by: Scott Mayhew <smayhew@redhat.com>
> > ---
> 
> Thanks, this works well here, and is a good idea anyway.
> 
> Reviewed-by: David Gow <davidgow@google.com>
> 

Brendan, David,

Is there a reason this patch hasn't been merged?

-Scott

> Cheers,
> -- David
> 
> >  lib/kunit/test.c | 3 +++
> >  1 file changed, 3 insertions(+)
> >
> > diff --git a/lib/kunit/test.c b/lib/kunit/test.c
> > index 1d1475578515..b8514dbb337c 100644
> > --- a/lib/kunit/test.c
> > +++ b/lib/kunit/test.c
> > @@ -712,6 +712,9 @@ int __kunit_test_suites_init(struct kunit_suite * const * const suites, int num_
> >  {
> >         unsigned int i;
> >
> > +       if (num_suites == 0)
> > +               return 0;
> > +
> >         if (!kunit_enabled() && num_suites > 0) {
> >                 pr_info("kunit: disabled\n");
> >                 return 0;
> > --
> > 2.43.0
> >

Re: [PATCH] kunit: bail out early in __kunit_test_suites_init() if there are no suites to test

[David Gow]

[-- Attachment #1: Type: text/plain, Size: 1332 bytes --]

On Tue, 30 Apr 2024 at 21:58, Scott Mayhew <smayhew@redhat.com> wrote:
>
> On Sat, 23 Mar 2024, David Gow wrote:
>
> > On Thu, 21 Mar 2024 at 22:32, Scott Mayhew <smayhew@redhat.com> wrote:
> > >
> > > Commit c72a870926c2 added a mutex to prevent kunit tests from running
> > > concurrently.  Unfortunately that mutex gets locked during module load
> > > regardless of whether the module actually has any kunit tests.  This
> > > causes a problem for kunit tests that might need to load other kernel
> > > modules (e.g. gss_krb5_test loading the camellia module).
> > >
> > > So check to see if there are actually any tests to run before locking
> > > the kunit_run_lock mutex.
> > >
> > > Fixes: c72a870926c2 ("kunit: add ability to run tests after boot using debugfs")
> > > Reported-by: Nico Pache <npache@redhat.com>
> > > Signed-off-by: Scott Mayhew <smayhew@redhat.com>
> > > ---
> >
> > Thanks, this works well here, and is a good idea anyway.
> >
> > Reviewed-by: David Gow <davidgow@google.com>
> >
>
> Brendan, David,
>
> Is there a reason this patch hasn't been merged?
>
> -Scott
>

Sorry: it totally slipped through the net. Thanks for the reminder,
it's merged now:
https://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest.git/commit/?h=kunit&id=2168e528f8679881df7487309f3444a121b2b544

Cheers,
-- David

[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4014 bytes --]