frank-w
mtk_crypto crashes
FYI there is some issues with something. Most time it was reproducible during something goes to swap, during high load on SATA and network (likely IRQ related race).
Just in case you'll see new paches about mtk_crypto, put links here.
[ 1895.593658] Unable to handle kernel paging request at virtual address 37b868a6
[ 1895.600882] pgd = dce90000
[ 1895.603645] [37b868a6] *pgd=00000000
[ 1895.607231] Internal error: Oops: 5 [#1] SMP ARM
[ 1895.660832] PC is at mtk_sha_append_sg.part.1+0x44/0xfc [mtk_crypto]
[ 1895.667149] LR is at 0xffffffe0
[ 1895.670270] pc : [<bf1ee9f0>] lr : [<ffffffe0>] psr: a0070113
[ 1895.676492] sp : dbb23e28 ip : 00000000 fp : dbb23e4c
[ 1895.681680] r10: 00000004 r9 : ffffe000 r8 : c0e9f340
[ 1895.686888] r7 : dcb1e390 r6 : 00000000 r5 : 00001000 r4 : da856e40
[ 1895.693375] r3 : fca378e3 r2 : fca378e3 r1 : 37b8689e r0 : da856e40
[ 1895.699859] Flags: NzCv IRQs on FIQs on Mode SVC_32 ISA ARM Segment none
[ 1895.706948] Control: 10c5387d Table: 9ce9006a DAC: 00000051
[ 1895.844795] [<bf1ee9f0>] (mtk_sha_append_sg.part.1 [mtk_crypto]) from [<bf1eead4>] (mtk_sha_update_slow+0x2c/0x108 [mtk_crypto])
[ 1895.856354] [<bf1eead4>] (mtk_sha_update_slow [mtk_crypto]) from [<bf1eecec>] (mtk_sha_update_start+0x13c/0x380 [mtk_crypto])
[ 1895.867612] [<bf1eecec>] (mtk_sha_update_start [mtk_crypto]) from [<bf1ef5c4>] (mtk_sha_done_task+0x124/0x188 [mtk_crypto])
[ 1895.878719] [<bf1ef5c4>] (mtk_sha_done_task [mtk_crypto]) from [<c0142828>] (tasklet_action+0x70/0x104)
[ 1895.888096] [<c0142828>] (tasklet_action) from [<c010161c>] (__do_softirq+0x13c/0x36c)
[ 1895.896010] [<c010161c>] (__do_softirq) from [<c01421f8>] (irq_exit+0xac/0x154)
[ 1895.903321] [<c01421f8>] (irq_exit) from [<c01949d0>] (__handle_domain_irq+0x70/0xc4)
[ 1895.911150] [<c01949d0>] (__handle_domain_irq) from [<c0101498>] (gic_handle_irq+0x5c/0xa0)
[ 1895.919490] [<c0101498>] (gic_handle_irq) from [<c09fcfd4>] (__irq_usr+0x54/0x80)
Which kernel do you use here? Have you tried a more recent one (5.0 rc1 maybe)? imho it's better to post as bug to kernel-mailing list
https://www.kernel.org/doc/html/latest/admin-guide/reporting-bugs.html
4.14. I use havily mtkhnat, and it also have some issues. Looks like these things are not very popular, as I can't find any recent work on this matter
mtk-hnat does not use crypto and i have it ported from 4.9 to 4.14. there are only a few infos about it so had not tested it deeply...only looked on interrupts and debugfs. But it shoild not trigger crypto-oops. You can disable it to try....i guess it's a buffer-problem