Re: panic in cypto code

Hi Steve,

On 5 Oct 2023, at 17:36, Steve Kargl wrote:
> In case anyone else is using openvpn.
>
> %  pkg info openvpn
> openvpn-2.6.6
> Name           : openvpn
> Version        : 2.6.6
> Installed on   : Tue Sep 19 08:48:55 2023 PDT
> Origin         : security/openvpn
> Architecture   : FreeBSD:15:amd64
>
> % uname -a
> FreeBSD hotrats 15.0-CURRENT #1 main-n265325-9c30461dd25b:
> Thu Sep 14 08:09:18 PDT 2023 kargl@hotrats:$PATH/HOTRATS amd64
>
>
> Fatal double fault
> rip 0xffffffff8099b408 rsp 0xfffffe000e1cc000 rbp 0xfffffe000e1cc010
> rax 0x53749f62934c5349 rdx 0x53749f62934c5349 rbx 0xfffffe000e1cc200
> rcx 0x57bf32fec3cbde70 rsi 0x32e8db2f0591c5da rdi 0x832f0fb1e6d07eb0
> r8 0 r9 0 r10 0
> r11 0x60 r12 0x5af7589946bd13d9 r13 0xbeddd6a808e1dd54
> r14 0xcdf12bbf2708189c r15 0xeb262ae8536a7adf rflags 0x10246
> cs 0x20 ss 0x28 ds 0x3b es 0x3b fs 0x13 gs 0x1b
> fsbase 0x1c02e381d120 gsbase 0xffffffff81a10000 kgsbase 0
> cpuid = 0; apic id = 00
> panic: double fault
> cpuid = 0
> time = 1696512769
> KDB: stack backtrace:
> db_trace_self_wrapper() at db_trace_self_wrapper+0x2b/frame 0xffffffff812b3060
> vpanic() at vpanic+0x132/frame 0xffffffff812b3190
> panic() at panic+0x43/frame 0xffffffff812b31f0
> dblfault_handler() at dblfault_handler+0x1ce/frame 0xffffffff812b32b0
> Xdblfault() at Xdblfault+0xd7/frame 0xffffffff812b32b0
> --- trap 0x17, rip = 0xffffffff8099b408, rsp = 0xfffffe000e1cc000, rbp = 0xfffffe000e1cc010 ---
> gfmultword4() at gfmultword4+0x8/frame 0xfffffe000e1cc010
> gf128_mul4b() at gf128_mul4b+0x63/frame 0xfffffe000e1cc050
> AES_GMAC_Update() at AES_GMAC_Update+0x73/frame 0xfffffe000e1cc0b0
> swcr_gcm() at swcr_gcm+0x660/frame 0xfffffe000e1cc830
> swcr_process() at swcr_process+0x1a/frame 0xfffffe000e1cc850
> crypto_dispatch() at crypto_dispatch+0x42/frame 0xfffffe000e1cc870
> ovpn_transmit_to_peer() at ovpn_transmit_to_peer+0x54e/frame 0xfffffe000e1cc8d0
> ovpn_output() at ovpn_output+0x2a2/frame 0xfffffe000e1cc950
> ip_output() at ip_output+0x11f6/frame 0xfffffe000e1cca40
> ovpn_encap() at ovpn_encap+0x3e7/frame 0xfffffe000e1ccac0
>
> #13 0xffffffff80ae08ce in dblfault_handler (frame=<optimized out>)
>     at /usr/src/sys/amd64/amd64/trap.c:1012
> #14 <signal handler called>
> #15 0xffffffff8099b408 in gfmultword4 (worda=3668422891496654298,
>     wordb=9452791399630012080, wordc=6013606648173318985,
>     wordd=6322828471639465584, x=..., tbl=0xfffffe000e1cc200)
>     at /usr/src/sys/opencrypto/gfmult.c:174
> #16 0xffffffff8099b5d3 in gf128_mul4b (r=...,
>     v=v@entry=0xfffff800076b9a64 "\3156}\373\312w\254iBnD\001ܹ˾\353&*\350Sjz߃/\017\261\346\320~\260Z\367X\231F\275\023\331St\237b\223LSI\276\335֨\b\341\335TW\2772\376\303\313\336pN\265\023\352\2054\002\a/˦9R\321\366p\f\352\204P\360\270\371\250\\\aE?7s\377\253\217b\262%\214\317m",
>     tbl=tbl@entry=0xfffffe000e1cc200) at /usr/src/sys/opencrypto/gfmult.c:268
> #17 0xffffffff8099ab13 in AES_GMAC_Update (ctx=0xfffffe000e1cc200,
>     vdata=<optimized out>, len=144) at /usr/src/sys/opencrypto/gmac.c:94
> #18 0xffffffff80998ae0 in swcr_gcm (ses=0xfffff8020376a048,
>     crp=0xfffff80023386c08) at /usr/src/sys/opencrypto/cryptosoft.c:505
> #19 0xffffffff80997c4a in swcr_process (dev=<optimized out>,
>     crp=0xfffff80023386c08, hint=<optimized out>)
>     at /usr/src/sys/opencrypto/cryptosoft.c:1680
>
Do you have a bit more information about what happened here? As in: can you reproduce this, or do you have any idea what was going on to trigger this? Did anything change in your setup (i.e. is if_ovpn use new, or did you update either kernel or userspace or …?

Do you have the full core dump to poke at?

It might be a bug in the crypto code, but it could also be a bug in the if_ovpn code, so I’d like to work out what caused this.

Best regards,
Kristof