On Tue, Jun 23, 2020 at 11:00:24PM -0400, Kevin O'Connor wrote:
On Sat, Jun 13, 2020 at 02:19:12PM +0300, Roman Bolshakov wrote:
There's a fallback to PIT if TSC is not present but it doesn't work properly. It prevents boot from floppy on isapc and 486 cpu .
SeaBIOS configures PIT in Mode 2. PIT counter is decremented in the mode but timer_adjust_bits() thinks that the counter overflows and increases 32-bit tick counter on each detected "overflow". Invalid overflow detection results in 55ms time advance (1 / 18.2Hz) on each read from PIT counter. So all timers expire much faster and 5-second floppy timeout expires in 83 real microseconds (or just a bit longer).
Provide counter direction to timer_adjust_bits() and normalize the counter to advance ticks in monotonically increasing TimerLast.
Good catch. Could we fix it using the patch below instead though?
--- a/src/hw/timer.c +++ b/src/hw/timer.c @@ -180,7 +180,7 @@ timer_read(void) // Read from PIT. outb(PM_SEL_READBACK | PM_READ_VALUE | PM_READ_COUNTER0, PORT_PIT_MODE); u16 v = inb(PORT_PIT_COUNTER0) | (inb(PORT_PIT_COUNTER0) << 8);
- return timer_adjust_bits(v, 0xffff);
- return timer_adjust_bits(-v, 0xffff);
// Return the TSC value that is 'msecs' time in the future.
I like the approach much more. Initial count value is 0, PIT rearms the timer when 1 is hit, unary negation on unsigned u16 fits perfectly, then timer_adjust_bits recieves 0, 1, 2, ... and timer is rearmed at 0xffff.
Do you want me to send v2 or you plan to apply the fix on your own?
If you plan to apply it on your own, here are the tags: