Skip to contentSkip to conversation

i8254 and TSC related code

DISCUSSION Stanislav Maslovski 2 participants 3 messages Jan 4, 2006 – Jan 9, 2006

lore
Share & tools
lore
#0 lore raw

Hello linux-kernel,

I have been reading sources in arch/i386/kernel/timers/ (in 2.6.14.5) and arch/i386/kernel/time.c (in 2.4.32) for some time and I have got a few questions. It would be nice if somebody could clarify the things for me a little.

The first question is about the use of i8254's mode 2 in channel 0. The datasheet on i82C54 says that in this mode the counter is used as a programmable divider of the CLK frequency. The corresponding divisor value (LATCH) is first loaded into the counter at the initialization. After the init the channel generates OUT pulses at the frequency CLK / LATCH. The specification also states that the counter is decremented down to 1 (not to 0) after which the LATCH value is reloaded. For instance, if LATCH is 3, then the count sequence is 3 2 1 3 2 1 3 2 1 and so on, which exactly corresponds to division of the CLK frequency by 3.

It seems that linux assumes a different behaviour in mode 2. For example, delay_at_last_interrupt is calculated as

        count = ((LATCH-1) - count) * TICK_SIZE;
        delay_at_last_interrupt = (count + LATCH/2) / LATCH;

The first line suggests that "count" is assumed to change in the [ LATCH-1; 0 ] range, does not it? In 2.4.32 there is even a piece of code in timer.c:timer_interrupt() that checks for the condition "count == LATCH" and does "count--" if this condition is true. The comment to this piece of code says:

/* Some i8253 clones hold the LATCH value visible
momentarily as they flip back to zero */

which explicitly contradicts the Intel docs.

The next question is about the piece of code in mark_offset_tsc() in timer_tsc.c from 2.6.14 which treats the case of pit_latch_buggy. The code replaces the most recent read value of the counter with the middle of 3 most recently read values. Those last reads are separated in time by at least one timer interrupt each and therefore are not related at all! Depending on interrupt latencies and the load those values can be anything from LATCH to 1 even in a non-buggy PIT. I am pretty sure that this code cannot help in the case of pit_latch_buggy either, however I could not find any information on this partucular bug. In its present form the code will likely produce irregular time jumps, the frequency of which will depend on the system load, etc. There will be jumps back in time too because the system time is calculated based on both TSC timestamps (which are monotonic) and "delay_at_last_interrupt" which changes randomly in this case. I think this code is bogus and should be removed. By the way, there is no such code in 2.4.32. I did a simple
check: I made the "if(pit_latch_buggy)" in the code be always true and then ran the modified kernel. My machine has a normal PIT. The following program was started from the user space:

===================== time_mon.c ====================== #include <stdio.h>
#include <sys/time.h>
#include <time.h>
struct timeval tv1, tv2;
struct timezone tz;
main()

{
	int cnt = 10;
	puts("Starting...");
	while (cnt--) {
		do {
			tv1 = tv2;
			gettimeofday(&tv2, &tz);
		} while (tv2.tv_sec > tv1.tv_sec || tv2.tv_usec >= tv1.tv_usec);
		printf("t1 = %lu.%6lu; t2 = %lu.%6lu\n",
			(unsigned long)tv1.tv_sec, (unsigned long)tv1.tv_usec,
			(unsigned long)tv2.tv_sec, (unsigned long)tv2.tv_usec);
	}
	puts("Done.");
}

=======================================================

Here are the values I received from the output of it:

t1 = 1136142051.862198; t2 = 1136142051.861307
t1 = 1136142256.233167; t2 = 1136142256.232274
t1 = 1136142403.500171; t2 = 1136142403.499295
t1 = 1136142607.871141; t2 = 1136142607.870262
t1 = 1136144390.113896; t2 = 1136144390.113019
t1 = 1136144594.484868; t2 = 1136144594.483985
t1 = 1136144946.123843; t2 = 1136144946.122972
t1 = 1136146229.457609; t2 = 1136146229.456722
t1 = 1136146728.363602; t2 = 1136146728.362730

The other load of the machine under the test was minimal:
I read some docs on the net, etc.

Next question is about TSC used in conjunction with PIT for timekeeping. The idea of using TSC to detect lost timer interrupts is nice. The idea of getting finer time resolution with the help of TSC is nice too, but I have some questions still. First, why in timer_tsc.c:mark_offset_tsc(), the TSC is read before latching the PIT count? Especially that in the recent kernels there is a spin_lock() call in between the TSC and the PIT reads?
For the values read from TSC and PIT to be corresponding the reads must happen almost
at the same time (even the comment in the code says so ;-))

I would place read_tsc() immediately after the latching operation "outb_p(0x00, PIT_MODE);"
It is better to have it immediately after it and not immediately before it also because the actual write to a register of an i/o device happens at the end of the bus cycle, but not at the beginning. I think that such a change will also make redundant the "corner case correction" done at the end of mark_offset_tsc():

       if (lost && abs(delay - delay_at_last_interrupt) > (900000/HZ))
               jiffies_64++;

By the way (and this is my last question for now) the test for this correction is buggy because in the current implementation "lost" does not account for the lost ticks but for the really_lost + 1. This is seen from the code some lines above:

       lost = delta/(1000000/HZ);                                 
       delay = delta%(1000000/HZ);
       if (lost >= 2) {
               jiffies_64 += lost-1;

At the beginning of this code delta is normally slightly above 1000000/HZ so that lost == 1 even if there was not a really lost tick. This was taken into account in the increment for jiffies but was not in the "corner case" correction...

I do realize that the presenly avaliable time keeping code looks like a big mess (sorry, developers, but that is a fact :( ) and that there will soon (probably) be an alternative timekeeping system avaliable, but for some time this old timekeeping code will stay in the kernel, so it still deserves fixing, IMHO.

Please comment.

--
Stanislav

#1 lore raw
Quoted 6 lines · Stanislav Maslovski

Hello linux-kernel,

I have been reading sources in arch/i386/kernel/timers/ (in 2.6.14.5) and arch/i386/kernel/time.c (in 2.4.32) for some time and I have got a few questions. It would be nice if somebody could clarify the things for me a little.

I'm glad to see your interest in timekeeping! Its a area that could use more review.

Quoted 23 lines

The first question is about the use of i8254's mode 2 in channel 0. The datasheet on i82C54 says that in this mode the counter is used as a programmable divider of the CLK frequency. The corresponding divisor value (LATCH) is first loaded into the counter at the initialization. After the init the channel generates OUT pulses at the frequency CLK / LATCH. The specification also states that the counter is decremented down to 1 (not to 0) after which the LATCH value is reloaded. For instance, if LATCH is 3, then the count sequence is 3 2 1 3 2 1 3 2 1 and so on, which exactly corresponds to division of the CLK frequency by 3.

It seems that linux assumes a different behaviour in mode 2. For example, delay_at_last_interrupt is calculated as

        count = ((LATCH-1) - count) * TICK_SIZE;
        delay_at_last_interrupt = (count + LATCH/2) / LATCH;

The first line suggests that "count" is assumed to change in the [ LATCH-1; 0 ] range, does not it? In 2.4.32 there is even a piece of code in timer.c:timer_interrupt() that checks for the condition "count == LATCH" and does "count--" if this condition is true. The comment to this piece of code says:

Quoted 4 lines

/* Some i8253 clones hold the LATCH value visible
momentarily as they flip back to zero */

which explicitly contradicts the Intel docs.

Hmmm. That logic is as old as the hills, but from your description it does seem that if the PIT goes from [LATCH, 1] the logic should be (LATCH - count)*...

Thoughts from anyone else?

Quoted 55 lines

The next question is about the piece of code in mark_offset_tsc() in timer_tsc.c from 2.6.14 which treats the case of pit_latch_buggy. The code replaces the most recent read value of the counter with the middle of 3 most recently read values. Those last reads are separated in time by at least one timer interrupt each and therefore are not related at all! Depending on interrupt latencies and the load those values can be anything from LATCH to 1 even in a non-buggy PIT. I am pretty sure that this code cannot help in the case of pit_latch_buggy either, however I could not find any information on this partucular bug. In its present form the code will likely produce irregular time jumps, the frequency of which will depend on the system load, etc. There will be jumps back in time too because the system time is calculated based on both TSC timestamps (which are monotonic) and "delay_at_last_interrupt" which changes randomly in this case. I think this code is bogus and should be removed. By the way, there is no such code in 2.4.32. I did a simple
check: I made the "if(pit_latch_buggy)" in the code be always true and then ran the modified kernel. My machine has a normal PIT. The following program was started from the user space:

===================== time_mon.c ====================== #include <stdio.h>
#include <sys/time.h>
#include <time.h>
struct timeval tv1, tv2;
struct timezone tz;
main()

{
	int cnt = 10;
	puts("Starting...");
	while (cnt--) {
		do {
			tv1 = tv2;
			gettimeofday(&tv2, &tz);
		} while (tv2.tv_sec > tv1.tv_sec || tv2.tv_usec >= tv1.tv_usec);
		printf("t1 = %lu.%6lu; t2 = %lu.%6lu\n",
			(unsigned long)tv1.tv_sec, (unsigned long)tv1.tv_usec,
			(unsigned long)tv2.tv_sec, (unsigned long)tv2.tv_usec);
	}
	puts("Done.");
}

=======================================================

Here are the values I received from the output of it:

t1 = 1136142051.862198; t2 = 1136142051.861307
t1 = 1136142256.233167; t2 = 1136142256.232274
t1 = 1136142403.500171; t2 = 1136142403.499295
t1 = 1136142607.871141; t2 = 1136142607.870262
t1 = 1136144390.113896; t2 = 1136144390.113019
t1 = 1136144594.484868; t2 = 1136144594.483985
t1 = 1136144946.123843; t2 = 1136144946.122972
t1 = 1136146229.457609; t2 = 1136146229.456722
t1 = 1136146728.363602; t2 = 1136146728.362730

Quoted 2 lines

The other load of the machine under the test was minimal:
I read some docs on the net, etc.

It does seem odd. While the notion that it takes the middle of the last three latched PIT values, to insure you have a sane value. The value selected may not have much to do with the current interrupt handler latency (which is why we touch the PIT at all here).

Someone familiar w/ the Cyrix 5510/5520 should comment here since that is what it has been added for.

Quoted 12 lines

Next question is about TSC used in conjunction with PIT for timekeeping. The idea of using TSC to detect lost timer interrupts is nice. The idea of getting finer time resolution with the help of TSC is nice too, but I have some questions still. First, why in timer_tsc.c:mark_offset_tsc(), the TSC is read before latching the PIT count? Especially that in the recent kernels there is a spin_lock() call in between the TSC and the PIT reads?
For the values read from TSC and PIT to be corresponding the reads must happen almost
at the same time (even the comment in the code says so ;-))

Quoted 6 lines

I would place read_tsc() immediately after the latching operation "outb_p(0x00, PIT_MODE);"
It is better to have it immediately after it and not immediately before it also because the actual write to a register of an i/o device happens at the end of the bus cycle, but not at the beginning.

I'd not be opposed to this, or at least pulling the rdtsc inside the spinlock should be fine.

Quoted 6 lines

I think that such a
change will also make redundant the "corner case correction" done at the end of mark_offset_tsc():

       if (lost && abs(delay - delay_at_last_interrupt) > (900000/HZ))
               jiffies_64++;

I need to think it through a bit more, but it seems there is the potential for the inverse of this corner case, where the PIT is latched just before it triggers another interrupt, and the TSC is read after the interrupt signal with a value greater then one interrupt. So while this check might be made unnecessary, additional logic could be needed.

Quoted 10 lines

By the way (and this is my last question for now) the test for this correction is buggy because in the current implementation "lost" does not account for the lost ticks but for the really_lost + 1. This is seen from the code some lines above:

       lost = delta/(1000000/HZ);                                 
       delay = delta%(1000000/HZ);
       if (lost >= 2) {
               jiffies_64 += lost-1;
Quoted 4 lines

At the beginning of this code delta is normally slightly above 1000000/HZ so that lost == 1 even if there was not a really lost tick. This was taken into account in the increment for jiffies but was not in the "corner case" correction...

Actually, that should be ok since we're trying to avoid doing any math if the interrupt is early, although I admit it is confusing.

So imagine the following:

0      1      2      3      4      5      6      7
|-+----|--+---|+-----|-----+|------|------|+----+|

A B C D E F

Where the numbered "|"s are the interrupt signals and lettered "+"s are the interrupt handlers.

So, at at interrupt handler B:

Quoted 1 line

From the last interrupt handler A, we've stored the time between 0 and A

as delay_at_last_interrupt.

We enter mark_offset_tsc() and calculate the time from B to A using the TSC, saving it in delta. We also store count from the PIT which we will use later.

We add delay_at_last_interrupt to delta to give us the full interval from 0 to B. We divide this interval by the interrupt frequency and save that as lost (which should store the value 1).

We use an estimate of where 1 should be to calculate the distance from 1 to B and store this in delay.

Since lost is less then 2, no lost tick compensation occurs.

We use the value count, stored earlier from the PIT, to calculate the interval from 1 to B and store it delay_at_last_interrupt.

Then assuming a full tick has occurred (ie: lost is 1), we double check our math for the interval 1 to B so the PIT based delay_at_last interrupt is close enough to the mixed PIT/TSC value delay calculated earlier.

That's the easy case (no early or late ticks) but you can go head and work out the others in your head.

It is quite ugly, so I might be missing your point. Please clarify if so.

Quoted 5 lines

I do realize that the presenly avaliable time keeping code looks like a big mess (sorry, developers, but that is a fact :( ) and that there will soon (probably) be an alternative timekeeping system avaliable, but for some time this old timekeeping code will stay in the kernel, so it still deserves fixing, IMHO.

Oh, I quite agree with you. I am mainly to blame, since when I initially created the timer_opts code I was very hesitant about changing behaviour so I preserved alot of the odd code that could probably have used a cleanup.

Most of the bugs and uglyness here comes from the fact that we're mixing time values generated from different hardware (ie: the PIT and the TSC). So we have to have some sanity checking so that they aren't totally off, and where they are just a touch off (due to calibration error or whatever), we just pray no one notices.

After trying to find the right balance of "trusting the TSC unless we should trust the PIT but double check and shrug if its wrong" and still finding corner cases, I got fed up and started work on the generic timekeeping infrastructure which allows for a simple linear mapping (well, ignoring NTP corrections) from the hardware counter value to the system time. This greatly cleans up the timekeeping logic and avoids this mixing hardware when generating time values.

The patches are currently in the -rt tree as well as 2.6.15-mm1. So if your careful eyes want to read over something else, it would be greatly appreciated.

Additionally, your point that the upcoming rework should not keep bugs in the current code from being fix is absolutely right! So please feel free to submit patches to lkml and Andrew fixing any current issues and I'll be sure to look them over.

thanks
-john

#2 lore raw

Hello,

Quoted 2 lines

I'm glad to see your interest in timekeeping! Its a area that could use more review.

Thank you for the reply! My comments follow below.

Quoted 15 lines
Quoted 15 lines

It seems that linux assumes a different behaviour in mode 2. For example, delay_at_last_interrupt is calculated as

        count = ((LATCH-1) - count) * TICK_SIZE;
        delay_at_last_interrupt = (count + LATCH/2) / LATCH;

The first line suggests that "count" is assumed to change in the [ LATCH-1; 0 ] range, does not it? In 2.4.32 there is even a piece of code in timer.c:timer_interrupt() that checks for the condition "count == LATCH" and does "count--" if this condition is true. The comment to this piece of code says:

/* Some i8253 clones hold the LATCH value visible
momentarily as they flip back to zero */

Quoted 7 lines
Quoted 1 line

which explicitly contradicts the Intel docs.

Hmmm. That logic is as old as the hills, but from your description it does seem that if the PIT goes from [LATCH, 1] the logic should be (LATCH - count)*...

Thoughts from anyone else?

While waiting for an answer from someone else, I programmed a simple bootable tester for the PIT (attached). It can do 3 tests: range test, latency test and sequence test. The range test polls PIT for some time to determine the range of count values. The latency test prints PIT readouts obtained immediately on entering the interrupt handler. The sequence test polls PIT as fast as it can around the point of underflow when the LATCH value gets reloaded (you should run it several times to see what happens).

You will need nasm to compile the code (see the file for instructions) and a diskette to boot it from.

On my two machines, for LATCH=1000h the results are (hex):
range: 0001 1000
latency: 0FFE, 0FFD ... about these values
LATCH and 1 can be seen in the sequence, but never 0 or anything > LATCH.

[ skipped some discussion of pit_latch_buggy ]

Quoted 7 lines

It does seem odd. While the notion that it takes the middle of the last three latched PIT values, to insure you have a sane value. The value selected may not have much to do with the current interrupt handler latency (which is why we touch the PIT at all here).

Someone familiar w/ the Cyrix 5510/5520 should comment here since that is what it has been added for.

It would be nice if somebody could run my test on this platform.

[ skipped ]

Quoted 12 lines
Quoted 10 lines

For the values read from TSC and PIT to be corresponding the reads must happen almost
at the same time (even the comment in the code says so ;-))

I would place read_tsc() immediately after the latching operation "outb_p(0x00, PIT_MODE);"
It is better to have it immediately after it and not immediately before it also because the actual write to a register of an i/o device happens at the end of the bus cycle, but not at the beginning.

Quoted 2 lines

I'd not be opposed to this, or at least pulling the rdtsc inside the spinlock should be fine.

After investigating it for some more time I found that it is rather better to have rdtsc operation immediately before PIT latching, inside the spin lock. My original reasoning did not take into account a delaying "out" which is added on i386 after the the latching "out" (in outb_p()).

Quoted 4 lines
Quoted 4 lines

I think that such a
change will also make redundant the "corner case correction" done at the end of mark_offset_tsc():

Quoted 9 lines
Quoted 2 lines
       if (lost && abs(delay - delay_at_last_interrupt) > (900000/HZ))
               jiffies_64++;

I need to think it through a bit more, but it seems there is the potential for the inverse of this corner case, where the PIT is latched just before it triggers another interrupt, and the TSC is read after the interrupt signal with a value greater then one interrupt. So while this check might be made unnecessary, additional logic could be needed.

I see. Below I will give some more thoughts on the corner case.

Quoted 40 lines
Quoted 10 lines

By the way (and this is my last question for now) the test for this correction is buggy because in the current implementation "lost" does not account for the lost ticks but for the really_lost + 1. This is seen from the code some lines above:

       lost = delta/(1000000/HZ);
       delay = delta%(1000000/HZ);
       if (lost >= 2) {
               jiffies_64 += lost-1;
Quoted 4 lines

At the beginning of this code delta is normally slightly above 1000000/HZ so that lost == 1 even if there was not a really lost tick. This was taken into account in the increment for jiffies but was not in the "corner case" correction...

Actually, that should be ok since we're trying to avoid doing any math if the interrupt is early, although I admit it is confusing.

So imagine the following:

0      1      2      3      4      5      6      7
|-+----|--+---|+-----|-----+|------|------|+----+|
 A       B    C           D               E    F

Where the numbered "|"s are the interrupt signals and lettered "+"s are the interrupt handlers.

So, at at interrupt handler B:

Quoted 1 line

From the last interrupt handler A, we've stored the time between 0 and A

as delay_at_last_interrupt.

We enter mark_offset_tsc() and calculate the time from B to A using the TSC, saving it in delta. We also store count from the PIT which we will use later.

We add delay_at_last_interrupt to delta to give us the full interval from 0 to B. We divide this interval by the interrupt frequency and save that as lost (which should store the value 1).

Quoted 7 lines

We use an estimate of where 1 should be to calculate the distance from 1 to B and store this in delay.

Since lost is less then 2, no lost tick compensation occurs.

We use the value count, stored earlier from the PIT, to calculate the interval from 1 to B and store it delay_at_last_interrupt.

Up to this point my understanding of the code fully agrees with yours ;)

Quoted 10 lines

Then assuming a full tick has occurred (ie: lost is 1), we double check our math for the interval 1 to B so the PIT based delay_at_last interrupt is close enough to the mixed PIT/TSC value delay calculated earlier.

That's the easy case (no early or late ticks) but you can go head and work out the others in your head.

It is quite ugly, so I might be missing your point. Please clarify if so.

Mmm. Is a hardware interrupt really lost when your lost == 1 and we are in the corner case? Do we need to increment jiffies here?

It seems that the low-level linux interrupt handler ACKs the interrups early enough. If I am not mistaken the following (i removed some levels of indirection) sequence of calls happens when satisfying IRQ 0:
(-> mean calls, {} mean call contexts, kernel 2.6.15)

gate in arch/i386/kernel/entry.S
`->__do_IRQ() in kernel/irq/handle.c

       {
           acknowlegde_hw_irq;
           ->timer_interrupt() in arch/i386/kernel/time.c
              {
                 ->mark_offset_tsc()
              1:
                 /* after this point code essentialy does */
                 ++jiffies;
                 update_times;
              }
       }

This means that if we see the corner case inside mark_offset_tsc() then most likely the hw IRQ that has started us has been ACKed in time. In this case the interrupt controller will be ready to take the IRQ generated at the "corner". Immediately after leaving the interrupt handler a new interrupt will fire at us so that jiffies will be incremented twice.

Please correct me if I am wrong.

By the way, in arch/i386/kernel/timer.c there is another ACK done when CONFIG_X86_IO_APIC is defined. On my diagram it happens at point 1. What is the purpose of it?

Quoted 3 lines

The patches are currently in the -rt tree as well as 2.6.15-mm1. So if your careful eyes want to read over something else, it would be greatly appreciated.

I am willing to do it.

Quoted 4 lines

Additionally, your point that the upcoming rework should not keep bugs in the current code from being fix is absolutely right! So please feel free to submit patches to lkml and Andrew fixing any current issues and I'll be sure to look them over.

Yup. When I feel I understand the matter enough well ;)

Below goes the test program.

Yours,
Stanislav

; pit.asm: A simple bootable PIT tester.
; Copyright (c) 2006 Stanislav Maslovski <[email protected]>
;

;   Compile with nasm, and then put the binary into the boot sector of a diskette:
;   # nasm pit.asm
;   # cat pit >/dev/fd0
;   # reboot
;   When booted use PF keys F1-F3 for tests.
;
;   This program is free software; you can redistribute it and/or modify
;   it under the terms of the GNU General Public License as published by
;   the Free Software Foundation; either version 2 of the License, or
;   (at your option) any later version.
;
;   This program is distributed in the hope that it will be useful,
;   but WITHOUT ANY WARRANTY; without even the implied warranty of
;   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
;   GNU General Public License for more details.
;
;   You should have received a copy of the GNU General Public License
;   along with this program; if not, write to the Free Software
;   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
;
	cpu	8086
%define delay	jmp short $+2
; *****************************
	section .text
	org	7C00h

v_off equ 4*08h
v_seg equ v_off+2
TRIES equ 16
LATCH equ 1000h
tbl equ 8000h
; here CS:IP = 0000:7C00; 64K of low memory is more than enough for us start:
; set segregs and stack
mov ax, cs
mov ds, ax
mov es, ax
cli
mov ss, ax
mov sp, 7C00h; expands below the code
sti
; print init message
call dnl
mov si, init_msg
call puts
; reprogram irq 0 vector

	cli
	mov	ax, [v_off]
	mov	[old_vec+0], ax
	mov	ax, [v_seg]
	mov	[old_vec+2], ax
	mov	word [v_off], irq_h
	mov	[v_seg], cs	
; set channel 0 mode 2 r/w 2 bytes, LATCH
	mov	al, 34h
	mov	dx, LATCH
	call	pit_reset
	sti
; catch several interrupts
	mov	cx, 5
	mov	[cnt1], cx
.l1:
	cmp	[cnt1], cx
	jae	.l1
	loop	.l1
	jmp	short main

; main loop
main_dnl:
call dnl
main:
; print banner
mov si, banner_msg
call puts
; wait for a function key pressed
.l2:
xor ah, ah
int 16h
or al, al
jnz .l2
mov cx, TRIES
mov di, tbl
; select the test

	sub	ah, 3Bh
	jz	range
	dec	ah
	jz	latency
	dec	ah
	jz	sequence
	jmp	short .l2

dump:
; dumps the table of values
mov cx, TRIES
mov si, tbl
.l3:

	lodsw
	mov	dx, ax
	call	printw
	loop	.l3

%if (TRIES % 16)
jmp short main_dnl
%else
jmp short main
%endif

; *****************************
; latency test
; *****************************
latency:

	mov	[cnt1], cx
.l1:
	cmp	[cnt1], cx
	jae	.l1
	mov	ax, [ch0]
	stosw
	loop	.l1
	jmp	short dump

; *****************************
; sequence test
; *****************************
sequence:
cli
; catch this value or less

	mov	dx, [seq_st]
.l2
	call	pit_latch
	cmp	ax, dx
	ja	.l2

; read sequence into the table
.l4:
stosw
.l3:

	call	pit_latch	
	cmp	ax, dx
	je	.l3
	mov	dx, ax
	loop	.l4
; correct the start value (useful at the next test)
	mov	ax, [tbl]
	sub	ax, dx
	add	ax, LATCH
	shr	ax, 1
	mov	[seq_st], ax
	sti
	jmp	short dump

; *****************************
; range test
; *****************************
range:
cli
xor cx, cx
mov bx, cx
mov dx, cx
dec dx
.l3:
call pit_latch
cmp ax, dx
jae .l1
mov dx, ax
.l1:
cmp ax, bx
jbe .l2
mov bx, ax
.l2:

	loop	.l3
	sti
	call	printw
	mov	dx, bx
	call	printw
	jmp	main_dnl

; *****************************
irq_h:
push ax
; read the PIT count

	call	pit_latch
	mov	[cs:ch0], ax
; update counters
	dec	word [cs:cnt1]
	dec	word [cs:cnt2]
	jnz	.l1
	mov	word [cs:cnt2], 10000h/LATCH

; jump to bios isr
pop ax
jmp far [cs:old_vec]
.l1:
; ack the interrupt and leave

	mov	al, 20h
	out	20h, al
	pop	ax
	iret
; *****************************
putch:
	push	cx
	mov	cx, 1
	mov	ah, 0Eh
	int	10h
	pop	cx
	ret
; *****************************
dnl:
	mov	al, 0Dh
	call	putch
	mov	al, 0Ah
	call	putch
	ret

; *****************************
puts:
lodsb
or al, al
jz .l1
call putch
jmp short puts
.l1:
call dnl
ret
; *****************************
printw:

	push	cx
	mov	cx, 4
	call	print_hex
	mov	al, ' '
	call	putch
	pop	cx
	ret
; *****************************
print_hex:
.l1:
	rol	dx, 1
	rol	dx, 1
	rol	dx, 1
	rol	dx, 1
	mov	al, dl
	and	al, 0Fh
	add	al, '0'
	cmp	al, '9'
	jbe	.l2
	add	al, 'A'-'9'-1
.l2:
	call	putch
	loop	.l1
	ret
; *****************************
pit_latch:
	xor	al, al
	out	43h, al
	delay
pit_inw:
	in	al, 40h
	delay
	xchg	al, ah
	in	al, 40h
	delay
	xchg	al, ah
	ret
; *****************************
pit_reset:
	out	43h, al
	delay
pit_outw:
	mov	al, dl
	out	40h, al
	delay
	mov	al, dh
	out	40h, al
	delay
	ret
; *****************************
cnt1:
	dw	0
cnt2:
	dw	10000h/LATCH
ch0:
	dw	0
; *****************************
init_msg:
	db	'PIT init: Ch 0 Mod 2 rw 2b LATCH=1000h', 0
banner_msg:
	db	'Test: F1=Range F2=Latency F3=Sequence', 0
old_vec:
	dw	0, 0

seq_st:
dw TRIES
; *****************************
; signature

	times 510-($-$$) db 0FFh
	db	55h, 0AAh

; *****************************