qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Richard Henderson <richard.henderson@linaro.org>
To: "Philippe Mathieu-Daudé" <f4bug@amsat.org>, qemu-devel@nongnu.org
Cc: Peter Maydell <peter.maydell@linaro.org>,
	Luc Michel <luc@lmichel.fr>,
	Andrew Baumann <Andrew.Baumann@microsoft.com>,
	Paul Zimmerman <pauldzim@gmail.com>,
	Niek Linnenbank <nieklinnenbank@gmail.com>,
	qemu-arm@nongnu.org
Subject: Re: [PATCH v3 3/4] hw/timer/bcm2835: Support the timer COMPARE registers
Date: Sat, 3 Oct 2020 12:17:52 -0500	[thread overview]
Message-ID: <c47c8248-0bfe-5747-1ec1-073fb755f039@linaro.org> (raw)
In-Reply-To: <20201002164216.1741110-4-f4bug@amsat.org>

On 10/2/20 11:42 AM, Philippe Mathieu-Daudé wrote:
> @@ -78,16 +71,29 @@ static void bcm2835_systmr_write(void *opaque, hwaddr offset,
>                                   uint64_t value, unsigned size)
>  {
>      BCM2835SystemTimerState *s = BCM2835_SYSTIMER(opaque);
> +    int index;
> +    uint64_t now;
> +    uint64_t triggers_delay_us;
>  
>      trace_bcm2835_systmr_write(offset, value);
>      switch (offset) {
>      case A_CTRL_STATUS:
>          s->reg.ctrl_status &= ~value; /* Ack */
> -        bcm2835_systmr_update_irq(s);
> +        for (index = 0; index < ARRAY_SIZE(s->tmr); index++) {
> +            if (extract32(value, index, 1)) {
> +                trace_bcm2835_systmr_irq_ack(index);
> +                qemu_set_irq(s->tmr[index].irq, 0);
> +            }

I think it might be instructive to have the parameter be uint64_t value64, and
the immediately do

    uint32_t value = value64;

That matches up better with extract32, the trace arguments...

> +        }
>          break;
>      case A_COMPARE0 ... A_COMPARE3:
> -        s->reg.compare[(offset - A_COMPARE0) >> 2] = value;
> -        bcm2835_systmr_update_compare(s, (offset - A_COMPARE0) >> 2);
> +        index = (offset - A_COMPARE0) >> 2;
> +        s->reg.compare[index] = value;
> +        now = qemu_clock_get_us(QEMU_CLOCK_VIRTUAL);
> +        /* Compare lower 32-bits of the free-running counter. */
> +        triggers_delay_us = value - (now & UINT32_MAX);
> +        trace_bcm2835_systmr_run(index, triggers_delay_us);
> +        timer_mod(&s->tmr[index].timer, now + triggers_delay_us);

... and here.

Also, the arithmetic looks off.

Consider when you want a long timeout, and pass in a value slightly below now.
 So, e.g.

  now   = 0xabcdffffffff;
  value = 0x0000fffffffe;

since triggers_delay_us is uint64_t, that comparison becomes

  triggers_delay_us = 0x0000fffffffe - 0xffffffff;
                    = 0xffffffffffffffff;

Then you add back in now, and do *not* get a value in the future:

    now + triggers_delay_us
  = 0xabcdffffffff + 0xffffffffffffffff
  = 0xabcdfffffffe

What I think you want is

  uint32_t triggers_delay_us = value - now
                             = 0xffffffff;

which then zero-extends when you add back to now to get

    now + triggers_delay_us
  = 0xabcdffffffff + 0xffffffff
  = 0xabcefffffffe

which is indeed in the future.


r~


  reply	other threads:[~2020-10-03 17:20 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-10-02 16:42 [PATCH v3 0/4] hw/arm/raspi: Fix SYS_timer to unbrick Linux kernels v3.7+ Philippe Mathieu-Daudé
2020-10-02 16:42 ` [PATCH v3 1/4] hw/timer/bcm2835: Introduce BCM2835_SYSTIMER_COUNT definition Philippe Mathieu-Daudé
2020-10-03 16:42   ` Richard Henderson
2020-10-02 16:42 ` [PATCH v3 2/4] hw/timer/bcm2835: Rename variable holding CTRL_STATUS register Philippe Mathieu-Daudé
2020-10-03 16:42   ` Richard Henderson
2020-10-02 16:42 ` [PATCH v3 3/4] hw/timer/bcm2835: Support the timer COMPARE registers Philippe Mathieu-Daudé
2020-10-03 17:17   ` Richard Henderson [this message]
2020-10-10 20:15     ` Philippe Mathieu-Daudé
2020-10-02 16:42 ` [PATCH v3 4/4] hw/arm/bcm2835_peripherals: Correctly wire the SYS_timer IRQs Philippe Mathieu-Daudé
2020-10-03 17:18   ` Richard Henderson

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=c47c8248-0bfe-5747-1ec1-073fb755f039@linaro.org \
    --to=richard.henderson@linaro.org \
    --cc=Andrew.Baumann@microsoft.com \
    --cc=f4bug@amsat.org \
    --cc=luc@lmichel.fr \
    --cc=nieklinnenbank@gmail.com \
    --cc=pauldzim@gmail.com \
    --cc=peter.maydell@linaro.org \
    --cc=qemu-arm@nongnu.org \
    --cc=qemu-devel@nongnu.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).