All of lore.kernel.org
 help / color / mirror / Atom feed
From: "Alex Bennée" <alex.bennee@linaro.org>
To: Pavel Dovgalyuk <dovgaluk@ispras.ru>
Cc: rth@twiddle.net, pbonzini@redhat.com, peter.maydell@linaro.org,
	qemu-devel@nongnu.org, mttcg@listserver.greensocs.com,
	fred.konrad@greensocs.com, a.rigo@virtualopensystems.com,
	cota@braap.org, bobby.prani@gmail.com, nikunj@linux.vnet.ibm.com,
	'Peter Crosthwaite' <crosthwaite.peter@gmail.com>
Subject: Re: [Qemu-devel] [RFC PATCH v1 7/9] cpus: move icount preparation out of tcg_exec_cpu
Date: Tue, 04 Apr 2017 09:56:44 +0100	[thread overview]
Message-ID: <8760ikblf7.fsf@linaro.org> (raw)
In-Reply-To: <000201d2ad05$e28626d0$a7927470$@ru>


Pavel Dovgalyuk <dovgaluk@ispras.ru> writes:

> I guess you are trying to fix the sympthoms of the case
> when iothread is trying to access instruction count.

In theory the main-loop should be sequenced before or after vCPU events
because of the BQL. I'm not sure why this is not currently the case.

> Maybe the solution is providing access to current_cpu for the iothread
> coupled with your patch 8?

Providing cross-thread access to CPU structures brings its own
challenges.

But it does occur to me we should probably ensure
timer_state.qemu_icount has appropriate barriers. This should be ensured
by the BQL but if it is ever accessed by 2 threads without a BQL
transition in-between then it is potentially racey.

>
> Pavel Dovgalyuk
>
>
>> -----Original Message-----
>> From: Alex Bennée [mailto:alex.bennee@linaro.org]
>> Sent: Monday, April 03, 2017 3:45 PM
>> To: dovgaluk@ispras.ru; rth@twiddle.net; pbonzini@redhat.com
>> Cc: peter.maydell@linaro.org; qemu-devel@nongnu.org; mttcg@listserver.greensocs.com;
>> fred.konrad@greensocs.com; a.rigo@virtualopensystems.com; cota@braap.org;
>> bobby.prani@gmail.com; nikunj@linux.vnet.ibm.com; Alex Bennée; Peter Crosthwaite
>> Subject: [RFC PATCH v1 7/9] cpus: move icount preparation out of tcg_exec_cpu
>>
>> As icount is only supported for single-threaded execution due to the
>> requirement for determinism let's remove it from the common
>> tcg_exec_cpu path.
>>
>> Also remove the additional fiddling which shouldn't be required as the
>> icount counters should all be rectified as you enter the loop.
>>
>> Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
>> ---
>>  cpus.c | 67 +++++++++++++++++++++++++++++++++++++++++++++---------------------
>>  1 file changed, 46 insertions(+), 21 deletions(-)
>>
>> diff --git a/cpus.c b/cpus.c
>> index 18b1746770..87638a75d2 100644
>> --- a/cpus.c
>> +++ b/cpus.c
>> @@ -1179,47 +1179,66 @@ static void handle_icount_deadline(void)
>>      }
>>  }
>>
>> -static int tcg_cpu_exec(CPUState *cpu)
>> +static void prepare_icount_for_run(CPUState *cpu)
>>  {
>> -    int ret;
>> -#ifdef CONFIG_PROFILER
>> -    int64_t ti;
>> -#endif
>> -
>> -#ifdef CONFIG_PROFILER
>> -    ti = profile_getclock();
>> -#endif
>>      if (use_icount) {
>>          int64_t count;
>>          int decr;
>> -        timers_state.qemu_icount -= (cpu->icount_decr.u16.low
>> -                                    + cpu->icount_extra);
>> -        cpu->icount_decr.u16.low = 0;
>> -        cpu->icount_extra = 0;
>> +
>> +        /* These should always be cleared by process_icount_data after
>> +         * each vCPU execution. However u16.high can be raised
>> +         * asynchronously by cpu_exit/cpu_interrupt/tcg_handle_interrupt
>> +         */
>> +        g_assert(cpu->icount_decr.u16.low == 0);
>> +        g_assert(cpu->icount_extra == 0);
>> +
>> +
>>          count = tcg_get_icount_limit();
>> +
>>          timers_state.qemu_icount += count;
>>          decr = (count > 0xffff) ? 0xffff : count;
>>          count -= decr;
>>          cpu->icount_decr.u16.low = decr;
>>          cpu->icount_extra = count;
>>      }
>> -    qemu_mutex_unlock_iothread();
>> -    cpu_exec_start(cpu);
>> -    ret = cpu_exec(cpu);
>> -    cpu_exec_end(cpu);
>> -    qemu_mutex_lock_iothread();
>> -#ifdef CONFIG_PROFILER
>> -    tcg_time += profile_getclock() - ti;
>> -#endif
>> +}
>> +
>> +static void process_icount_data(CPUState *cpu)
>> +{
>>      if (use_icount) {
>>          /* Fold pending instructions back into the
>>             instruction counter, and clear the interrupt flag.  */
>>          timers_state.qemu_icount -= (cpu->icount_decr.u16.low
>>                          + cpu->icount_extra);
>> +
>> +        /* We must be under BQL here as cpu_exit can tweak
>> +           icount_decr.u32 */
>> +        g_assert(qemu_mutex_iothread_locked());
>>          cpu->icount_decr.u32 = 0;
>>          cpu->icount_extra = 0;
>>          replay_account_executed_instructions();
>>      }
>> +}
>> +
>> +
>> +static int tcg_cpu_exec(CPUState *cpu)
>> +{
>> +    int ret;
>> +#ifdef CONFIG_PROFILER
>> +    int64_t ti;
>> +#endif
>> +
>> +#ifdef CONFIG_PROFILER
>> +    ti = profile_getclock();
>> +#endif
>> +    qemu_mutex_unlock_iothread();
>> +    cpu_exec_start(cpu);
>> +    ret = cpu_exec(cpu);
>> +    cpu_exec_end(cpu);
>> +    qemu_mutex_lock_iothread();
>> +#ifdef CONFIG_PROFILER
>> +    tcg_time += profile_getclock() - ti;
>> +#endif
>>      return ret;
>>  }
>>
>> @@ -1306,7 +1325,13 @@ static void *qemu_tcg_rr_cpu_thread_fn(void *arg)
>>
>>              if (cpu_can_run(cpu)) {
>>                  int r;
>> +
>> +                prepare_icount_for_run(cpu);
>> +
>>                  r = tcg_cpu_exec(cpu);
>> +
>> +                process_icount_data(cpu);
>> +
>>                  if (r == EXCP_DEBUG) {
>>                      cpu_handle_guest_debug(cpu);
>>                      break;
>> --
>> 2.11.0


--
Alex Bennée

  reply	other threads:[~2017-04-04  8:56 UTC|newest]

Thread overview: 32+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-04-03 12:45 [Qemu-devel] [RFC PATCH v1 0/9] MTTCG and record/replay fixes for rc3 Alex Bennée
2017-04-03 12:45 ` [Qemu-devel] [RFC PATCH v1 1/9] scripts/qemugdb/mtree.py: fix up mtree dump Alex Bennée
2017-04-03 12:45 ` [Qemu-devel] [RFC PATCH v1 2/9] scripts/qemu-gdb/timers.py: new helper to dump timer state Alex Bennée
2017-04-03 14:02   ` Philippe Mathieu-Daudé
2017-04-03 12:45 ` [Qemu-devel] [RFC PATCH v1 3/9] scripts/replay-dump.py: replay log dumper Alex Bennée
2017-04-03 12:45 ` [Qemu-devel] [RFC PATCH v1 4/9] target/i386/misc_helper: wrap BQL around another IRQ generator Alex Bennée
2017-04-04 16:53   ` Richard Henderson
2017-04-04 17:36     ` Eduardo Habkost
2017-04-03 12:45 ` [Qemu-devel] [RFC PATCH v1 5/9] cpus: remove icount handling from qemu_tcg_cpu_thread_fn Alex Bennée
2017-04-04 16:53   ` Richard Henderson
2017-04-03 12:45 ` [Qemu-devel] [RFC PATCH v1 6/9] cpus: check cpu->running in cpu_get_icount_raw() Alex Bennée
2017-04-03 14:00   ` Philippe Mathieu-Daudé
2017-04-04 16:54   ` Richard Henderson
2017-04-03 12:45 ` [Qemu-devel] [RFC PATCH v1 7/9] cpus: move icount preparation out of tcg_exec_cpu Alex Bennée
2017-04-04  5:39   ` Pavel Dovgalyuk
2017-04-04  8:56     ` Alex Bennée [this message]
2017-04-04 10:46       ` Alex Bennée
2017-04-04 10:53         ` Paolo Bonzini
2017-04-04 12:31           ` Alex Bennée
2017-04-04 12:37             ` Paolo Bonzini
2017-04-04 13:29               ` Alex Bennée
2017-04-05 10:44                 ` Pavel Dovgalyuk
2017-04-05 11:18                   ` Alex Bennée
2017-04-03 12:45 ` [Qemu-devel] [RFC PATCH v1 8/9] cpus: don't credit executed instructions before they have run Alex Bennée
2017-04-03 17:04   ` Paolo Bonzini
2017-04-04  5:37   ` Pavel Dovgalyuk
2017-04-04 10:13     ` Paolo Bonzini
2017-04-07 11:27       ` Pavel Dovgalyuk
2017-04-04 14:39   ` Paolo Bonzini
2017-04-03 12:45 ` [Qemu-devel] [RFC PATCH v1 9/9] replay: gracefully handle backward time events Alex Bennée
2017-04-03 17:03 ` [Qemu-devel] [RFC PATCH v1 0/9] MTTCG and record/replay fixes for rc3 Paolo Bonzini
2017-04-04  8:50   ` Alex Bennée

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=8760ikblf7.fsf@linaro.org \
    --to=alex.bennee@linaro.org \
    --cc=a.rigo@virtualopensystems.com \
    --cc=bobby.prani@gmail.com \
    --cc=cota@braap.org \
    --cc=crosthwaite.peter@gmail.com \
    --cc=dovgaluk@ispras.ru \
    --cc=fred.konrad@greensocs.com \
    --cc=mttcg@listserver.greensocs.com \
    --cc=nikunj@linux.vnet.ibm.com \
    --cc=pbonzini@redhat.com \
    --cc=peter.maydell@linaro.org \
    --cc=qemu-devel@nongnu.org \
    --cc=rth@twiddle.net \
    /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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.