Re: tsc: Fast TSC calibration failed with sever AMD Ryzen processor (2200G, 2400G, Ryzen 7 1700)

From: Paul Menzel
Date: Mon Jan 28 2019 - 12:18:08 EST

Dear Tom,


On 01/24/19 00:33, Lendacky, Thomas wrote:
> On 1/23/19 6:56 AM, Paul Menzel wrote:

>> On 01/22/19 21:24, Lendacky, Thomas wrote:
>>> On 1/22/19 10:53 AM, Paul Menzel wrote:
>>>> [Adding Tom to CC]
>>
>>>> On 01/14/19 11:09, Paul Menzel wrote:
>>>>
>>>>> On 01/11/19 21:43, Thomas Gleixner wrote:
>>>>>
>>>>>> On Mon, 7 Jan 2019, Paul Menzel wrote:
>>>>>>> On 01/07/19 16:24, Thomas Gleixner wrote:
>>>>>>>>> Linux 4.19.13 from Debian Sid/unstable logs the message below on the board MSI
>>>>>>>>> MS-7A37/B350M MORTAR with the processor AMD Ryzen 3 2200G.
>>>>>>>>>
>>>>>>>>> As a result, the early time stamps do not seem to be working.
>>>>>>>>
>>>>>>>>>> [ 0.000000] DMI: Micro-Star International Co., Ltd. MS-7A37/B350M MORTAR (MS-7A37), BIOS 1.I0 11/06/2018
>>>>>>>>>> [ 0.000000] tsc: Fast TSC calibration failed
>>>>>>>>
>>>>>>>> And the further boot log says:
>>>>>>>>
>>>>>>>> [ 0.036000] tsc: Unable to calibrate against PIT
>>>>>>>> [ 0.036000] tsc: using HPET reference calibration
>>>>>>>> [ 0.036000] tsc: Detected 3500.117 MHz processor
>>>>>>>>
>>>>>>>> So the quick calibration in early boot fails because the PIT seems not to
>>>>>>>> do what the kernel expects. Nothing we can cure :(
>>>>>>>
>>>>>>> I see. Can AMD confirm that this is the expected behavior? If yes, should
>>>>>>> the fast TSC calibration be skipped on these devices?
>>
>>> It's not expected behavior. All of the systems that I have access to do
>>> not exhibit this issue. Having said that, I have a limited number of
>>> systems available to me.
>>
>> But as a data point, what Ryzen systems did you test with? Just to know, if
>> there are configurations where the same processor behaves inconsistently.
>>
>> Can you request one of the failing systems mentioned below to reproduce the
>> problem?
>>
>>> I don't have much experience in this area, but if it is something that
>>> consistently occurs, you might try to see if you can better identify why
>>> it fails. The message is issued in pit_hpet_ptimer_calibrate_cpu() in file
>>> arch/x86/kernel/tsc.c.
>>
>> With the attached patch applied, I get:
>>
>> [ 0.000000] tsc: quick_pit_calibrate: break in if !pit_expect_msb, i = 42
>> [ 0.000000] tsc: Fast TSC calibration failed, i = 42
>> [ 0.000000] tsc: Using PIT calibration value
>>
>> The functions `pit_verify_msb()` and `pit_expect_msb()` are:
>>
>> ```
>> static inline int pit_verify_msb(unsigned char val)
>> {
>> /* Ignore LSB */
>> inb(0x42);
>> return inb(0x42) == val;
>> }
>>
>> static inline int pit_expect_msb(unsigned char val, u64 *tscp, unsigned long *deltap)
>> {
>> int count;
>> u64 tsc = 0, prev_tsc = 0;
>>
>> for (count = 0; count < 50000; count++) {
>> if (!pit_verify_msb(val))
>> break;
>> prev_tsc = tsc;
>> tsc = get_cycles();
>> }
>> *deltap = get_cycles() - prev_tsc;
>> *tscp = tsc;
>>
>> /*
>> * We require _some_ success, but the quality control
>> * will be based on the error terms on the TSC values.
>> */
>> return count > 5;
>> }
>> ```
>>
>> So count is smaller than or equal to 5, and `pit_verify_msb(val)` failed early,
>> right?
>
> Depends on what you mean by "failed early." The loop iterated 42 times
> as it tried to reach the acceptable error rate, so pit_expect_msb() had
> succeeded a number of times before failing.

The maximum is 233, and in my tests the counter `i` never surpassed 55 or so.

> One of the things I noticed while searching for info on the PIT is that
> one of the specs [1] mentioned that just reading the counter values (as
> opposed to using the counter latch command or the read-back command) could
> return undefined values if the counters are being updated while being
> read. I'm not sure if that is what is occurring or if that matters in
> this day and age, I'm not familiar with this area.
>
> There's also talk of SMIs messing with this calibration. Could a long
> running SMI (something close to 214us) run that results in keeping "count"
> under 6? Again, just speculation on my part.

I printed the values read from `inb(0x42)`, and log them, if they are unequal.
Please find the patch attached.

Here are the results of five boots.

[ 0.000000] DMI: HP HP EliteDesk 705 G4 MT/83E7, BIOS Q06 Ver. 02.04.01 09/14/2018
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = ff != 0 = val
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fe != ff = val
[ 0.000000] tsc: pit_expect_msb: val = 0xff: count = 52
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fd != fe = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfe: count = 51
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fc != fd = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfd: count = 47
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fb != fc = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfc: count = 47
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fa != fb = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfb: count = 47
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f9 != fa = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfa: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f8 != f9 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf9: count = 47
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f7 != f8 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf8: count = 47
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f6 != f7 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf7: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f5 != f6 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf6: count = 47
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f4 != f5 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf5: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f3 != f4 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf4: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f2 != f3 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf3: count = 47
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f1 != f2 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf2: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f0 != f1 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf1: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = ef != f0 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf0: count = 47
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = ee != ef = val
[ 0.000000] tsc: pit_expect_msb: val = 0xef: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = ed != ee = val
[ 0.000000] tsc: pit_expect_msb: val = 0xee: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = ec != ed = val
[ 0.000000] tsc: pit_expect_msb: val = 0xed: count = 47
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = eb != ec = val
[ 0.000000] tsc: pit_expect_msb: val = 0xec: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = ea != eb = val
[ 0.000000] tsc: pit_expect_msb: val = 0xeb: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = e9 != ea = val
[ 0.000000] tsc: pit_expect_msb: val = 0xea: count = 47
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = e8 != e9 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xe9: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = e7 != e8 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xe8: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = e6 != e7 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xe7: count = 47
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = e5 != e6 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xe6: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = e4 != e5 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xe5: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = e3 != e4 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xe4: count = 47
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = e2 != e3 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xe3: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = e1 != e2 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xe2: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = e0 != e1 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xe1: count = 47
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = df != e0 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xe0: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = de != df = val
[ 0.000000] tsc: pit_expect_msb: val = 0xdf: count = 47
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = dd != de = val
[ 0.000000] tsc: pit_expect_msb: val = 0xde: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = dc != dd = val
[ 0.000000] tsc: pit_expect_msb: val = 0xdd: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = da != dc = val
[ 0.000000] tsc: pit_expect_msb: val = 0xdc: count = 4
[ 0.000000] tsc: quick_pit_calibrate: break in if !pit_expect_msb, i = 35
[ 0.000000] tsc: Fast TSC calibration failed, i = 35 from 233

[ 0.000000] DMI: HP HP EliteDesk 705 G4 MT/83E7, BIOS Q06 Ver. 02.04.01 09/14/2018
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = ff != 0 = val
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fe != ff = val
[ 0.000000] tsc: pit_expect_msb: val = 0xff: count = 53
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fd != fe = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfe: count = 53
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fc != fd = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfd: count = 52
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fb != fc = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfc: count = 51
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fa != fb = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfb: count = 50
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f9 != fa = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfa: count = 45
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f8 != f9 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf9: count = 46
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f7 != f8 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf8: count = 45
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f6 != f7 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf7: count = 45
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f5 != f6 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf6: count = 46
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f4 != f5 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf5: count = 45
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f3 != f4 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf4: count = 45
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f2 != f3 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf3: count = 46
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f1 != f2 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf2: count = 45
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f0 != f1 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf1: count = 45
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = ef != f0 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf0: count = 46
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = ee != ef = val
[ 0.000000] tsc: pit_expect_msb: val = 0xef: count = 45
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = ed != ee = val
[ 0.000000] tsc: pit_expect_msb: val = 0xee: count = 45
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = ec != ed = val
[ 0.000000] tsc: pit_expect_msb: val = 0xed: count = 46
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = eb != ec = val
[ 0.000000] tsc: pit_expect_msb: val = 0xec: count = 45
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = ea != eb = val
[ 0.000000] tsc: pit_expect_msb: val = 0xeb: count = 45
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = e9 != ea = val
[ 0.000000] tsc: pit_expect_msb: val = 0xea: count = 46
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = e8 != e9 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xe9: count = 45
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = e7 != e8 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xe8: count = 45
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = e5 != e7 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xe7: count = 14
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = e5 != e6 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xe6: count = 0
[ 0.000000] tsc: quick_pit_calibrate: break in if !pit_expect_msb, i = 25
[ 0.000000] tsc: Fast TSC calibration failed, i = 25 from 233

[ 0.000000] DMI: HP HP EliteDesk 705 G4 MT/83E7, BIOS Q06 Ver. 02.04.01 09/14/2018
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = ff != 0 = val
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fe != ff = val
[ 0.000000] tsc: pit_expect_msb: val = 0xff: count = 53
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fd != fe = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfe: count = 53
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fc != fd = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfd: count = 52
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fb != fc = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfc: count = 44
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fa != fb = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfb: count = 51
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f7 != fa = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfa: count = 13
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f7 != f9 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf9: count = 0
[ 0.000000] tsc: quick_pit_calibrate: break in if !pit_expect_msb, i = 6
[ 0.000000] tsc: Fast TSC calibration failed, i = 6 from 233

[ 0.000000] DMI: HP HP EliteDesk 705 G4 MT/83E7, BIOS Q06 Ver. 02.04.01 09/14/2018
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = ff != 0 = val
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fe != ff = val
[ 0.000000] tsc: pit_expect_msb: val = 0xff: count = 39
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fd != fe = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfe: count = 39
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fc != fd = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfd: count = 39
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fb != fc = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfc: count = 39
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fa != fb = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfb: count = 39
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f9 != fa = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfa: count = 39
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f8 != f9 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf9: count = 40
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f7 != f8 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf8: count = 39
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f6 != f7 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf7: count = 39
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f5 != f6 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf6: count = 39
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f4 != f5 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf5: count = 39
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f3 != f4 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf4: count = 39
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f2 != f3 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf3: count = 39
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f1 != f2 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf2: count = 40
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f0 != f1 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf1: count = 39
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = ef != f0 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf0: count = 39
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = ed != ef = val
[ 0.000000] tsc: pit_expect_msb: val = 0xef: count = 13
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = ed != ee = val
[ 0.000000] tsc: pit_expect_msb: val = 0xee: count = 0
[ 0.000000] tsc: quick_pit_calibrate: break in if !pit_expect_msb, i = 17
[ 0.000000] tsc: Fast TSC calibration failed, i = 17 from 233

[ 0.000000] DMI: HP HP EliteDesk 705 G4 MT/83E7, BIOS Q06 Ver. 02.04.01 09/14/2018
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = ff != 0 = val
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fe != ff = val
[ 0.000000] tsc: pit_expect_msb: val = 0xff: count = 52
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fd != fe = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfe: count = 51
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fc != fd = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfd: count = 44
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fb != fc = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfc: count = 46
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = fa != fb = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfb: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f9 != fa = val
[ 0.000000] tsc: pit_expect_msb: val = 0xfa: count = 47
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f8 != f9 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf9: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f7 != f8 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf8: count = 47
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f6 != f7 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf7: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f5 != f6 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf6: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f4 != f5 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf5: count = 47
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f3 != f4 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf4: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f2 != f3 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf3: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f1 != f2 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf2: count = 47
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = f0 != f1 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf1: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = ef != f0 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xf0: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = ee != ef = val
[ 0.000000] tsc: pit_expect_msb: val = 0xef: count = 47
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = ed != ee = val
[ 0.000000] tsc: pit_expect_msb: val = 0xee: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = ec != ed = val
[ 0.000000] tsc: pit_expect_msb: val = 0xed: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = eb != ec = val
[ 0.000000] tsc: pit_expect_msb: val = 0xec: count = 47
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = ea != eb = val
[ 0.000000] tsc: pit_expect_msb: val = 0xeb: count = 48
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = e9 != ea = val
[ 0.000000] tsc: pit_expect_msb: val = 0xea: count = 47
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = e6 != e9 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xe9: count = 42
[ 0.000000] tsc: pit_verify_msb: inb(0x42) = e6 != e8 = val
[ 0.000000] tsc: pit_expect_msb: val = 0xe8: count = 0
[ 0.000000] tsc: quick_pit_calibrate: break in if !pit_expect_msb, i = 23
[ 0.000000] tsc: Fast TSC calibration failed, i = 23 from 233

So you can see, that, if my patch does not have any side effects, the
read value is in the end smaller than the expected value and decreases
not by one, but by more than one.

So in the last boot from e9 to e6, and the code decreases by one, so
expects e8 in the next iteration.

Of course your explanation that an SMI could cause this, could still
apply. Another explanation would be, that the PIT(?) does not decrease
by a constant rate.

I remembered that there was also a TSC related problem in coreboot,
that Google fixed by the commit below [2].

> soc/amd/stoneyridge: remove dependence on TSC
>
> The TSC rate is empirically swinging during early boot. That
> leaves timestamps and udelay()s to not be correct. To rectify this
> stop using TSC for all of these time sources. Instead use the
> performance TSC which is at a fixed 100MHz clock. That provides
> stable time sources and legit timestamps.
>
> BUG=b:72378235,b:72170796

Aaron replied in #coreboot@xxxxxxxxxxxxxxx:

> I don't know that the lkml report applies to this ryzen, but my
> recollection of that bug on stoney was that TSC wasn't exactly constant
> rate until deeper into the boot flow. I think there's something that
> happens in SMU at a point in the boot that stabilizes the clock rate.
> Ryzen could be the same thing or a completely different bug all
> together.

The HP and MSI firmwares take more than eight seconds though, compared
to the less than one second by coreboot, so it might be something else.

Unfortunately, I have neither resources nor knowledge to look into this,
and it looks like a hardware issue to me, only AMD can debug. Tom, could
you forward that to the appropriate departments in AMD?


Kind regards,

Paul


> [1] http://www.scs.stanford.edu/10wi-cs140/pintos/specs/8254.pdf
[2] https://review.coreboot.org/23424

>>>>>> It should work and we really don't want to add cpu family/model based
>>>>>> decisions whether we invoke something or not. Those tables are stale before
>>>>>> they hit mainline.
>>>>>
>>>>> Understood. If itâs supposed to work, any hints on how to debug this?
>>>>>
>>>>> Does some Linux kernel developers have an AMD Ryzen system, and can reproduce
>>>>> the issue?
>>>>>
>>>>> It seems to fail with an AMD Ryzen 2400G too [1].
>>>>
>>>> We now have an HP EliteDesk 705 G4 MT with that processsor, showing the same
>>>> problem.
>>>>
>>>> ```
>>>> [ 0.000000] Linux version 4.20.0.mx64.238 (root@xxxxxxxxxxxxxxxxxxxxxxx) (gcc version 7.3.0 (GCC)) #1 SMP Mon Dec 24 14:50:00 CET 2018
>>>> [â]
>>>> [ 0.000000] NX (Execute Disable) protection: active
>>>> [ 0.000000] SMBIOS 3.1 present.
>>>> [ 0.000000] DMI: HP HP EliteDesk 705 G4 MT/83E7, BIOS Q06 Ver. 02.04.01 09/14/2018
>>>> [ 0.000000] tsc: Fast TSC calibration failed
>>>> [ 0.000000] e820: update [mem 0x00000000-0x00000fff] usable ==> reserved
>>>> [â]
>>>> [ 0.017860] smpboot: CPU0: AMD Ryzen 5 PRO 2400G with Radeon Vega Graphics (family: 0x17, model: 0x11, stepping: 0x0)
>>>> [â]
>>>> ```
>>>>
>>>>> It also fails on an AMD Ryzen 7 1700 [2].
>>>>>
>>>>> ```
>>>>> [ 0.000000] Linux version 4.15.0-kali3-amd64 (devel@xxxxxxxx) (gcc version 7.3.0 (Debian 7.3.0-16)) #1 SMP Debian 4.15.17-1kali1 (2018-04-25)
>>>>> [â]
>>>>> [ 0.008000] ..TIMER: vector=0x30 apic1=0 pin1=2 apic2=-1 pin2=-1
>>>>> [ 0.028000] tsc: Fast TSC calibration failed
>>>>> [ 0.032000] tsc: PIT calibration matches HPET. 1 loops
>>>>> [ 0.032000] tsc: Detected 2994.246 MHz processor
>>>>> [â]
>>>>> [ 0.044000] smpboot: CPU0: AMD Ryzen 7 1700 Eight-Core Processor (family: 0x17, model: 0x1, stepping: 0x1)
>>>>> ```
>>>>>
>>>>> It *works* here on one system with AMD Ryzen 5 PRO 1500 and Linux 4.14.87.
>>>>>
>>>>> ```
>>>>> [ 0.000000] Linux version 4.14.87.mx64.236 (root@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx) (gcc version 7.3.0 (GCC)) #1 SMP Mon Dec 10 09:48:57 CET 2018
>>>>> [â]
>>>>> [ 0.000000] tsc: Fast TSC calibration using PIT
>>>>> [â]
>>>>> [ 0.035000] smpboot: CPU0: AMD Ryzen 5 PRO 1500 Quad-Core Processor (family: 0x17, model: 0x1, stepping: 0x1)
>>>>> ```
>>>>
>>>> How to continue from here? Is documentation for that available from AMD?
>>>> I didnât find a BKDG (Bios Kernel Developer Guide) at [3].
>>
>>
>> Kind regards,
>>
>> Paul
>>
>>
>>>>> [1]: https://bbs.archlinux.org/viewtopic.php?pid=1781282#p1781282
>>>>> [2]: https://forums.kali.org/showthread.php?40444-error-loading-amdgpu-drivers-AMD-RX580-driver[3]: https://developer.amd.com/resources/developer-guides-manuals
From 7720e8dbc578da02b020d7e8f5441e74b3726004 Mon Sep 17 00:00:00 2001
From: Paul Menzel <pmenzel@xxxxxxxxxxxxx>
Date: Wed, 23 Jan 2019 00:24:37 +0100
Subject: [PATCH] x86/kernel/tsc: Debug early TSC calibration

Signed-off-by: Paul Menzel <pmenzel@xxxxxxxxxxxxx>
---
arch/x86/kernel/tsc.c | 20 ++++++++++++++++----
1 file changed, 16 insertions(+), 4 deletions(-)

diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c
index 3fae23834069..4696ef7326b6 100644
--- a/arch/x86/kernel/tsc.c
+++ b/arch/x86/kernel/tsc.c
@@ -478,9 +478,15 @@ static unsigned long pit_calibrate_tsc(u32 latch, unsigned long ms, int loopmin)
*/
static inline int pit_verify_msb(unsigned char val)
{
+ unsigned char val1;
+
/* Ignore LSB */
inb(0x42);
- return inb(0x42) == val;
+ val1 = inb(0x42);
+
+ if (val1 != val)
+ pr_err("%s: inb(0x42) = %x != %x = val", __func__, val1, val);
+ return val1 == val;
}

static inline int pit_expect_msb(unsigned char val, u64 *tscp, unsigned long *deltap)
@@ -497,6 +503,8 @@ static inline int pit_expect_msb(unsigned char val, u64 *tscp, unsigned long *de
*deltap = get_cycles() - prev_tsc;
*tscp = tsc;

+ pr_err("%s: val = 0x%x: count = %d\n", __func__, val, count);
+
/*
* We require _some_ success, but the quality control
* will be based on the error terms on the TSC values.
@@ -550,8 +558,10 @@ static unsigned long quick_pit_calibrate(void)

if (pit_expect_msb(0xff, &tsc, &d1)) {
for (i = 1; i <= MAX_QUICK_PIT_ITERATIONS; i++) {
- if (!pit_expect_msb(0xff-i, &delta, &d2))
+ if (!pit_expect_msb(0xff-i, &delta, &d2)) {
+ pr_err("%s: break in if !pit_expect_msb, i = %d\n", __func__, i);
break;
+ }

delta -= tsc;

@@ -576,12 +586,14 @@ static unsigned long quick_pit_calibrate(void)
* This also guarantees serialization of the
* last cycle read ('d2') in pit_expect_msb.
*/
- if (!pit_verify_msb(0xfe - i))
+ if (!pit_verify_msb(0xfe - i)) {
+ pr_err("%s: break in if !pit_verify_msb\n", __func__);
break;
+ }
goto success;
}
}
- pr_info("Fast TSC calibration failed\n");
+ pr_err("Fast TSC calibration failed, i = %u from %u\n", i, MAX_QUICK_PIT_ITERATIONS);
return 0;

success:
--
2.17.1

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature