Comme=
nt # 90
on bug 11067=
4
from Tom =
B
I'm not sure this is helpful but I managed to somewhat test th=
e race condition
theory.
If you follow the callstack:
vega20_set_fclk_to_highest_dpm_level -> smum_send_msg_to_smc_with_parame=
ter ->
vega20_send_msg_to_smc_with_parameter -> vega20_wait_for_response ->
phm_wait_for_register_unequal you find this code in smu_helper.c:
int phm_wait_on_register(struct pp_hwmgr *hwmgr, uint32_t index,
uint32_t value, uint32_t mask)
{
uint32_t i;
uint32_t cur_value;
if (hwmgr =3D=3D NULL || hwmgr->device =3D=3D NULL) {
pr_err("Invalid Hardware Manager!");
return -EINVAL;
}
for (i =3D 0; i < hwmgr->usec_timeout; i++) {
cur_value =3D cgs_read_register(hwmgr->device, index);
if ((cur_value & mask) =3D=3D (value & mask))
break;
udelay(1);
}
/* timeout means wrong logic*/
if (i =3D=3D hwmgr->usec_timeout)
return -1;
return 0;
}
The timeout there is interesting. I increased it.
for (i =3D 0; i < hwmgr->usec_timeout*10; i++) {
cur_value =3D cgs_read_register(hwmgr->device, index);
if ((cur_value & mask) =3D=3D (value & mask))
break;
udelay(1);
}
The PC takes significantly longer to boot (10 or so seconds when it's usual=
ly
instant) and the error still occurs. So I'm not sure it's just a matter of
waiting.