I wanted to see how this would behave on my APUs. One gave 0000000,
another ffffffff consistently. But then I had this diff in my tree
and I rebuilt for my workstation. That has
cpu0: AMD Ryzen 9 7950X 16-Core Processor, 4500.00 MHz, 19-61-02
...
ccp0 at pci21 dev 0 function 2 "AMD 17h/90h Crypto" rev 0x00
and it too gives:
ccp: rng ffffffff
ccp: rng ffffffff
ccp: rng ffffffff
ccp: rng ffffffff
Is anyone else seeing non-random output from ccp(4) on non-APU
systems?
Paul
On Wed, Apr 19, 2023 at 05:18:11PM +0200, Christian Weisgerber wrote:
| Christian Weisgerber:
|
| > ccp(4) attaches, so presumably it is used as a source of entropy.
| > Whether the hardware actually provides random output, I don't know.
|
| I built a kernel with an instrumented driver. Unfortunately, no
| entropy is provided:
|
| ccp: rng 00000000
| ccp: rng 00000000
| ccp: rng 00000000
| ccp: rng 00000000
| ccp: rng 00000000
|
| This is with the lastest firmware:
| bios0: vendor coreboot version "v4.19.0.1" date 01/31/2023
|
|
| Index: dev/ic/ccp.c
| ===================================================================
| RCS file: /cvs/src/sys/dev/ic/ccp.c,v
| retrieving revision 1.3
| diff -u -p -r1.3 ccp.c
| --- dev/ic/ccp.c 29 May 2020 04:42:25 -0000 1.3
| +++ dev/ic/ccp.c 19 Apr 2023 15:12:17 -0000
| @@ -56,6 +56,7 @@ ccp_rng(void *arg)
| trng = bus_space_read_4(sc->sc_iot, sc->sc_ioh, CCP_REG_TRNG);
| if (trng != 0)
| enqueue_randomness(trng);
| + printf("ccp: rng %08x\n", trng);
|
| - timeout_add_msec(&sc->sc_tick, 100);
| + timeout_add_msec(&sc->sc_tick, 5000);
| }
| --
| Christian "naddy" Weisgerber [email protected]
|
--
>++++++++[<++++++++++>-]<+++++++.>+++[<------>-]<.>+++[<+
+++++++++++>-]<.>++[<------------>-]<+.--------------.[-]
http://www.weirdnet.nl/
