[Bug 1195768] New: ppc64le: alg: skcipher: failed to allocate transform for p8_aes_xts: -2
https://bugzilla.suse.com/show_bug.cgi?id=1195768 Bug ID: 1195768 Summary: ppc64le: alg: skcipher: failed to allocate transform for p8_aes_xts: -2 Classification: openSUSE Product: openSUSE Tumbleweed Version: Current Hardware: Other OS: Other Status: NEW Severity: Normal Priority: P5 - None Component: Kernel Assignee: kernel-bugs@opensuse.org Reporter: petr.vorel@suse.com QA Contact: qa-bugs@suse.de Found By: --- Blocker: --- There is kernel oops on Tumbleweed ppc64le kernel: [ 0.295853][ T83] Failed to allocate xts(aes) fallback: -2 [ 0.296710][ T83] alg: skcipher: failed to allocate transform for p8_aes_xts: -2 [ 0.297525][ T83] ------------[ cut here ]------------ [ 0.298310][ T83] alg: self-tests for p8_aes_xts (xts(aes)) failed (rc=-2) [ 0.298325][ T83] WARNING: CPU: 2 PID: 83 at crypto/testmgr.c:5681 alg_test+0x8a8/0xd80 [ 0.299990][ T83] Modules linked in: [ 0.300796][ T83] CPU: 2 PID: 83 Comm: cryptomgr_test Not tainted 5.16.4-1-default #1 openSUSE Tumbleweed 514f0c059e014af0dbaa3c399367e9fbb58fffc9 [ 0.302511][ T83] NIP: c00000000077ba08 LR: c00000000077ba04 CTR: c0000000008ea730 [ 0.303410][ T83] REGS: c0000000060bf960 TRAP: 0700 Not tainted (5.16.4-1-default) [ 0.304309][ T83] MSR: 800000000282b033 <SF,VEC,VSX,EE,FP,ME,IR,DR,RI,LE> CR: 28008221 XER: 20000000 [ 0.305243][ T83] CFAR: c00000000015ba50 IRQMASK: 0 [ 0.305243][ T83] GPR00: c00000000077ba04 c0000000060bfc00 c000000002968a00 0000000000000038 [ 0.305243][ T83] GPR04: 00000000ffffbfff c0000000060bf8c0 c0000000060bf8b8 0000000000000000 [ 0.305243][ T83] GPR08: c000000002807148 c000000002647060 c000000002647060 0000000048008221 [ 0.305243][ T83] GPR12: c0000000027a7108 c00000003fffd680 c00000000019a858 c000000002d4cdc0 [ 0.305243][ T83] GPR16: 0000000000000000 0000000000000000 0000000000000000 0000000000000000 [ 0.305243][ T83] GPR20: 0000000000000000 00000000000000a6 c000000000f7dd00 00000000000000a5 [ 0.305243][ T83] GPR24: 00000000000000a5 fffffffffffffffe 0000000000003240 c000000002bff140 [ 0.305243][ T83] GPR28: 0000000000000400 0000000000000105 c000000004b8ca80 c000000004b8ca00 [ 0.313426][ T83] NIP [c00000000077ba08] alg_test+0x8a8/0xd80 [ 0.314280][ T83] LR [c00000000077ba04] alg_test+0x8a4/0xd80 [ 0.315134][ T83] Call Trace: [ 0.316020][ T83] [c0000000060bfc00] [c00000000077ba04] alg_test+0x8a4/0xd80 (unreliable) [ 0.316927][ T83] [c0000000060bfd70] [c000000000777f34] cryptomgr_test+0x44/0x80 [ 0.317823][ T83] [c0000000060bfda0] [c00000000019aa08] kthread+0x1b8/0x1c0 [ 0.318714][ T83] [c0000000060bfe10] [c00000000000cf64] ret_from_kernel_thread+0x5c/0x64 [ 0.319606][ T83] Instruction dump: [ 0.320454][ T83] 3d220029 8929676a 2c090000 40820414 7f3907b4 3c62fe93 7fc5f378 7fe4fb78 [ 0.321391][ T83] 38630dc8 7f26cb78 4b9dffe9 60000000 <0fe00000> fa2100f8 fa410100 fa610108 [ 0.322318][ T83] ---[ end trace 8d7a5ba1cc312459 ]--- System boots and this is just crypto testing API, but tainting kernel just during boot is bad enough to get fixed. I suspect it's upstream issue (haven't tested it on mainline kernel), because we don't have in Tumbleweed these FIPS specific patches from SLES and system booted without fips being enabled. Thus I can report it to linux-crypto ML -- You are receiving this mail because: You are the assignee for the bug.
https://bugzilla.suse.com/show_bug.cgi?id=1195768 https://bugzilla.suse.com/show_bug.cgi?id=1195768#c1 --- Comment #1 from Petr Vorel <petr.vorel@suse.com> --- Created attachment 856029 --> https://bugzilla.suse.com/attachment.cgi?id=856029&action=edit dmesg from affected system (https://openqa.opensuse.org/tests/2176974/file/serial0.txt) dmesg from affected system: https://openqa.opensuse.org/tests/2176974/file/serial0.txt https://openqa.opensuse.org/tests/2176974#step/boot_ltp/117 -- You are receiving this mail because: You are the assignee for the bug.
https://bugzilla.suse.com/show_bug.cgi?id=1195768 https://bugzilla.suse.com/show_bug.cgi?id=1195768#c2 --- Comment #2 from Petr Vorel <petr.vorel@suse.com> --- Related to CONFIG_CRYPTO_DEV_VMX=y, thus ppc64le specific. -- You are receiving this mail because: You are the assignee for the bug.
https://bugzilla.suse.com/show_bug.cgi?id=1195768 https://bugzilla.suse.com/show_bug.cgi?id=1195768#c3 --- Comment #3 from Nicolai Stange <nstange@suse.com> --- (In reply to Petr Vorel from comment #0)
There is kernel oops on Tumbleweed ppc64le kernel:
[ 0.295853][ T83] Failed to allocate xts(aes) fallback: -2 [ 0.296710][ T83] alg: skcipher: failed to allocate transform for p8_aes_xts: -2
Looks like p8_aes_xts fails to allocate the generic fallback implementation (2 == ENOENT). You mentioned CONFIG_CRYPTO_DEV_VMX=y, i.e. it's built-in. I wonder what the settings of CONFIG_CRYPTO_XTS and CONFIG_CRYPTO_AES are each? Is the Tumbleweed kernel getting built from the "stable" branch at https://github.com/openSUSE/kernel-source? If either of these two isn't built-in, then a generic fallback xts(aes) will indeed not be available this early and it's a kernel config issue, I'd say. -- You are receiving this mail because: You are the assignee for the bug.
https://bugzilla.suse.com/show_bug.cgi?id=1195768 https://bugzilla.suse.com/show_bug.cgi?id=1195768#c4 --- Comment #4 from Petr Vorel <petr.vorel@suse.com> --- Created attachment 856034 --> https://bugzilla.suse.com/attachment.cgi?id=856034&action=edit kernel config (5.16.4-1-default - https://openqa.opensuse.org/tests/2176974/file/boot_ltp-config-5.16.4-1-defa...) -- You are receiving this mail because: You are the assignee for the bug.
https://bugzilla.suse.com/show_bug.cgi?id=1195768 https://bugzilla.suse.com/show_bug.cgi?id=1195768#c5 --- Comment #5 from Petr Vorel <petr.vorel@suse.com> --- (In reply to Nicolai Stange from comment #3)
(In reply to Petr Vorel from comment #0)
There is kernel oops on Tumbleweed ppc64le kernel:
[ 0.295853][ T83] Failed to allocate xts(aes) fallback: -2 [ 0.296710][ T83] alg: skcipher: failed to allocate transform for p8_aes_xts: -2
Looks like p8_aes_xts fails to allocate the generic fallback implementation (2 == ENOENT).
You mentioned CONFIG_CRYPTO_DEV_VMX=y, i.e. it's built-in. Yes. The default in kernel mainline defconfig configures it as module (m).
I wonder what the settings of CONFIG_CRYPTO_XTS and CONFIG_CRYPTO_AES are each?
CONFIG_CRYPTO_XTS=m CONFIG_CRYPTO_AES=y CONFIG_CRYPTO_AES_TI=m Also added whole config.
Is the Tumbleweed kernel getting built from the "stable" branch at https://github.com/openSUSE/kernel-source?
Yes, it's a stable kernel https://openqa.opensuse.org/tests/2176974#step/boot_ltp/73 Name : kernel-default Version : 5.16.4 Release : 1.1 Architecture: ppc64le Install Date: Sun Feb 6 15:39:00 2022 Group : System/Kernel Size : 181196396 License : GPL-2.0-only Signature : RSA/SHA256, Wed Feb 2 21:55:01 2022, Key ID b88b2fd43dbdc284 Source RPM : kernel-default-5.16.4-1.1.nosrc.rpm Build Date : Wed Feb 2 21:37:24 2022 Build Host : obs-power8-02 Packager : http://bugs.opensuse.org Vendor : openSUSE URL : http://www.kernel.org/ Summary : The Standard Kernel Description : The standard kernel for both uniprocessor and multiprocessor systems. Source Timestamp: 2022-01-29 12:57:02 +0000 GIT Revision: b1466772bff446593573594227eca436e3a5cbcf GIT Branch: stable Distribution: openSUSE Tumbleweed
If either of these two isn't built-in, then a generic fallback xts(aes) will indeed not be available this early and it's a kernel config issue, I'd say.
CONFIG_CRYPTO_AES=y => so it's our config issue in the end? I guess kernel should not allow invalid configuration => should be mainline issue. -- You are receiving this mail because: You are the assignee for the bug.
https://bugzilla.suse.com/show_bug.cgi?id=1195768 https://bugzilla.suse.com/show_bug.cgi?id=1195768#c6 --- Comment #6 from Nicolai Stange <nstange@suse.com> --- (In reply to Petr Vorel from comment #5)
(In reply to Nicolai Stange from comment #3)
If either of these two isn't built-in, then a generic fallback xts(aes) will indeed not be available this early and it's a kernel config issue, I'd say. CONFIG_CRYPTO_AES=y => so it's our config issue in the end?
Yes, I think so. The culprit is the CONFIG_CRYPTO_XTS=m though: the xts module is not available yet at the time when built-in VMX' crypto algs are getting tested.
I guess kernel should not allow invalid configuration => should be mainline issue.
Agreed, CRYPTO_DEV_VMX should depend on CRYPTO_XTS and CRYPTO_AES, which it currently doesn't AFAICS. -- You are receiving this mail because: You are the assignee for the bug.
https://bugzilla.suse.com/show_bug.cgi?id=1195768 https://bugzilla.suse.com/show_bug.cgi?id=1195768#c7 --- Comment #7 from Petr Vorel <petr.vorel@suse.com> --- (In reply to Nicolai Stange from comment #6) ...
Yes, I think so. The culprit is the CONFIG_CRYPTO_XTS=m though: the xts module is not available yet at the time when built-in VMX' crypto algs are getting tested.
I guess kernel should not allow invalid configuration => should be mainline issue.
Agreed, CRYPTO_DEV_VMX should depend on CRYPTO_XTS and CRYPTO_AES, which it currently doesn't AFAICS.
That was quick, thx! Suppose you're sending patch to linux-crypto, I can verify it. -- You are receiving this mail because: You are the assignee for the bug.
https://bugzilla.suse.com/show_bug.cgi?id=1195768 Petr Vorel <petr.vorel@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Assignee|kernel-bugs@opensuse.org |petr.vorel@suse.com -- You are receiving this mail because: You are the assignee for the bug.
participants (1)
-
bugzilla_noreply@suse.com