This works for me. I just installed the latest snapshot, synced the
source, applied your patch, built and installed and it works just
fine.
Hate Gmail for patches, though.
Thanks. Hope I replied in time to make the 5.3 cut. :)
Kent.
On Mon, Feb 11, 2013 at 2:34 AM, Martin Pieuchot <mpieuc...@nolizard.org> wrote:
> Hi Kent,
>
> On 14/01/13(Mon) 10:05, Kent Fritz wrote:
>> On Fri, Jan 11, 2013 at 5:29 PM, Stefan Sperling <s...@openbsd.org> wrote:
>> > I see. So this is happening during pms_probe() which runs before the
>> > protocol is selected. Maybe fix it like this? I think the code should
>> > cope with hardware that returns unrecognizable garbage. But I don't
>> > know very much about PS/2.
>> >
>> > Thanks for pinning down the problem!
>> >
>> > Index: pckbc.c
>> > ===================================================================
>> > RCS file: /cvs/src/sys/dev/ic/pckbc.c,v
>> > retrieving revision 1.31
>> > diff -u -p -r1.31 pckbc.c
>> > --- pckbc.c 17 Oct 2012 19:16:10 -0000 1.31
>> > +++ pckbc.c 12 Jan 2013 01:25:41 -0000
>> > @@ -620,6 +620,11 @@ pckbc_poll_cmd1(struct pckbc_internal *t
>> > #ifdef PCKBCDEBUG
>> > printf("pckbc_cmd: lost 0x%x\n", c);
>> > #endif
>> > + /* Don't retry cmd forever. */
>> > + if (cmd->retries++ >= 5) {
>> > + cmd->status = EIO;
>> > + return;
>> > + }
>> > }
>> >
>> > while (cmd->responseidx < cmd->responselen) {
>>
>> That patch works fine. Tested on i386 on nT-i1250. Thanks for
>> pointing me in the right direction!
>
> Could you try the diff below and tell me if it also fix your problem? I
> believe it's better to handle the bat failure core (0xfc) like we
> already do with the bat completion code rather than exiting for any
> value..
>
> M.
>
> Index: pckbc.c
> ===================================================================
> RCS file: /home/ncvs/src/sys/dev/ic/pckbc.c,v
> retrieving revision 1.31
> diff -u -p -r1.31 pckbc.c
> --- pckbc.c 17 Oct 2012 19:16:10 -0000 1.31
> +++ pckbc.c 11 Feb 2013 10:24:49 -0000
> @@ -48,6 +48,12 @@
> #include <dev/pckbc/pckbdvar.h>
> #endif
>
> +#ifdef PCKBCDEBUG
> +#define DPRINTF(x...) do { printf(x); } while (0);
> +#else
> +#define DPRINTF(x...)
> +#endif
> +
> /* descriptor for one device command */
> struct pckbc_devcmd {
> TAILQ_ENTRY(pckbc_devcmd) next;
> @@ -102,9 +108,10 @@ int pckbcintr_internal(struct pckbc_inte
>
> const char *pckbc_slot_names[] = { "kbd", "aux" };
>
> -#define KBC_DEVCMD_ACK 0xfa
> -#define KBC_DEVCMD_RESEND 0xfe
> -#define KBC_DEVCMD_BAT 0xaa
> +#define KBC_DEVCMD_ACK 0xfa
> +#define KBC_DEVCMD_RESEND 0xfe
> +#define KBC_DEVCMD_BAT_DONE 0xaa
> +#define KBC_DEVCMD_BAT_FAIL 0xfc
>
> #define KBD_DELAY DELAY(8)
>
> @@ -587,39 +594,32 @@ pckbc_poll_cmd1(struct pckbc_internal *t
> break;
> }
>
> - if (c == KBC_DEVCMD_ACK) {
> + switch (c) {
> + case KBC_DEVCMD_ACK:
> cmd->cmdidx++;
> continue;
> - }
> /*
> * Some legacy free PCs keep returning Basic Assurance Test
> * (BAT) instead of something usable, so fail gracefully.
> */
> - if (c == KBC_DEVCMD_RESEND || c == KBC_DEVCMD_BAT) {
> -#ifdef PCKBCDEBUG
> - printf("pckbc_cmd: %s\n",
> + case KBC_DEVCMD_RESEND:
> + case KBC_DEVCMD_BAT_DONE:
> + case KBC_DEVCMD_BAT_FAIL:
> + DPRINTF("pckbc_cmd: %s\n",
> c == KBC_DEVCMD_RESEND ? "RESEND": "BAT");
> -#endif
> if (cmd->retries++ < 5)
> continue;
> - else {
> -#ifdef PCKBCDEBUG
> - printf("pckbc: cmd failed\n");
> -#endif
> - cmd->status = ENXIO;
> - return;
> - }
> - }
> - if (c == -1) {
> -#ifdef PCKBCDEBUG
> - printf("pckbc_cmd: timeout\n");
> -#endif
> +
> + DPRINTF("pckbc_cmd: cmd failed\n");
> + cmd->status = ENXIO;
> + return;
> + case -1:
> + DPRINTF("pckbc_cmd: timeout\n");
> cmd->status = EIO;
> return;
> + default:
> + DPRINTF("pckbc_cmd: lost 0x%x\n", c);
> }
> -#ifdef PCKBCDEBUG
> - printf("pckbc_cmd: lost 0x%x\n", c);
> -#endif
> }
>
> while (cmd->responseidx < cmd->responselen) {
