All of lore.kernel.org
 help / color / mirror / Atom feed
From: "Alex Bennée" <alex.bennee@linaro.org>
To: qemu-devel@nongnu.org
Cc: damien.hedde@greensocs.com, mark.burton@greensocs.com,
	sakisp@xilinx.com, edgari@xilinx.com,
	Paolo Bonzini <pbonzini@redhat.com>,
	Luc Michel <luc.michel@greensocs.com>,
	Richard Henderson <rth@twiddle.net>
Subject: Re: [Qemu-devel] [RFC PATCH] accel/tcg/translator: add tb_enter TCG trace
Date: Fri, 28 Jun 2019 17:33:05 +0100	[thread overview]
Message-ID: <87o92hacvy.fsf@zen.linaroharston> (raw)
In-Reply-To: <20190628113917.15869-1-luc.michel@greensocs.com>


Luc Michel <luc.michel@greensocs.com> writes:

> Add a TCG trace at the begining of a translation block recording the
> first and last (past-the-end) PC values.
>
> Signed-off-by: Luc Michel <luc.michel@greensocs.com>
> ---
> This can be used to trace the execution of the guest quite efficiently.
> It will report each time a TB is entered (using the tb_enter_exec
> trace). The traces arguments give the PC start and past-the-end values.
> It has very little to no performance impact since the trace is actually
> emitted in the generated code only when it is enabled at run time.
>
> It works already quite well on its own to trace guest execution.

You could just remove the disable from:

disable exec_tb(void *tb, uintptr_t pc) "tb:%p pc=0x%"PRIxPTR

And call qemu with -d nochain,trace:exec_tb

> However
> it does not handle the case where a TB is exited in the middle of
> execution. I'm not sure how to properly trace that. A trace could be
> added when `cpu_loop_exit()' is called to report the current PC, but in
> most cases the interesting value (the PC of the instruction that
> caused the exit) is already lost at this stage.
>
> I'm not sure there is a generic (i.e. not target specific) way of
> recovering the last PC executed when cpu_loop_exit() is called. Do you
> think of a better way?
>
> Thanks to the Xilinx's QEMU team who sponsored this work.
> ---
>  accel/tcg/translator.c | 24 ++++++++++++++++++++++++
>  trace-events           |  3 +++
>  2 files changed, 27 insertions(+)
>
> diff --git a/accel/tcg/translator.c b/accel/tcg/translator.c
> index 9226a348a3..c55377aa18 100644
> --- a/accel/tcg/translator.c
> +++ b/accel/tcg/translator.c
> @@ -14,10 +14,11 @@
>  #include "tcg/tcg-op.h"
>  #include "exec/exec-all.h"
>  #include "exec/gen-icount.h"
>  #include "exec/log.h"
>  #include "exec/translator.h"
> +#include "trace-tcg.h"
>
>  /* Pairs with tcg_clear_temp_count.
>     To be called by #TranslatorOps.{translate_insn,tb_stop} if
>     (1) the target is sufficiently clean to support reporting,
>     (2) as and when all temporaries are known to be consumed.
> @@ -28,14 +29,31 @@ void translator_loop_temp_check(DisasContextBase *db)
>          qemu_log("warning: TCG temporary leaks before "
>                   TARGET_FMT_lx "\n", db->pc_next);
>      }
>  }
>
> +static TCGOp *gen_trace_tb_enter(TranslationBlock *tb)
> +{
> +    TCGOp *last_pc_op;
> +
> +    TCGv pc_end = tcg_temp_new();
> +
> +    /* The last PC value is not known yet */
> +    tcg_gen_movi_tl(pc_end, 0xdeadbeef);
> +    last_pc_op = tcg_last_op();
> +
> +    trace_tb_enter_tcg(tcg_ctx->cpu, cpu_env, tb->pc, pc_end);
> +    tcg_temp_free(pc_end);
> +
> +    return last_pc_op;
> +}
> +
>  void translator_loop(const TranslatorOps *ops, DisasContextBase *db,
>                       CPUState *cpu, TranslationBlock *tb, int max_insns)
>  {
>      int bp_insn = 0;
> +    TCGOp *trace_pc_end;
>
>      /* Initialize DisasContext */
>      db->tb = tb;
>      db->pc_first = tb->pc;
>      db->pc_next = db->pc_first;
> @@ -50,10 +68,13 @@ void translator_loop(const TranslatorOps *ops, DisasContextBase *db,
>      /* Reset the temp count so that we can identify leaks */
>      tcg_clear_temp_count();
>
>      /* Start translating.  */
>      gen_tb_start(db->tb);
> +
> +    trace_pc_end = gen_trace_tb_enter(tb);
> +
>      ops->tb_start(db, cpu);
>      tcg_debug_assert(db->is_jmp == DISAS_NEXT);  /* no early exit */
>
>      while (true) {
>          db->num_insns++;
> @@ -110,10 +131,13 @@ void translator_loop(const TranslatorOps *ops, DisasContextBase *db,
>
>      /* Emit code to exit the TB, as indicated by db->is_jmp.  */
>      ops->tb_stop(db, cpu);
>      gen_tb_end(db->tb, db->num_insns - bp_insn);
>
> +    /* Fixup the last PC value in the tb_enter trace now that we know it */
> +    tcg_set_insn_param(trace_pc_end, 1, db->pc_next);
> +
>      /* The disas_log hook may use these values rather than recompute.  */
>      db->tb->size = db->pc_next - db->pc_first;
>      db->tb->icount = db->num_insns;
>
>  #ifdef DEBUG_DISAS
> diff --git a/trace-events b/trace-events
> index aeea3c2bdb..e37fa12ef0 100644
> --- a/trace-events
> +++ b/trace-events
> @@ -157,10 +157,13 @@ vcpu guest_cpu_reset(void)
>  #
>  # Mode: user, softmmu
>  # Targets: TCG(all)
>  vcpu tcg guest_mem_before(TCGv vaddr, uint8_t info) "info=%d", "vaddr=0x%016"PRIx64" info=%d"
>
> +# translator.c
> +vcpu tcg tb_enter(uint64_t pc_start, TCGv pc_end) "pc_start:0x%"PRIx64, "pc:0x%"PRIx64" pc_end:0x%"PRIx64
> +
>  # linux-user/syscall.c
>  # bsd-user/syscall.c
>
>  # @num: System call number.
>  # @arg*: System call argument value.


--
Alex Bennée


      parent reply	other threads:[~2019-06-28 17:32 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-06-28 11:39 [Qemu-devel] [RFC PATCH] accel/tcg/translator: add tb_enter TCG trace Luc Michel
2019-06-28 12:16 ` Richard Henderson
2019-07-01 13:19   ` Edgar E. Iglesias
2019-07-01 13:33     ` Alex Bennée
2019-07-01 13:47       ` Edgar E. Iglesias
2019-06-28 16:33 ` Alex Bennée [this message]

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=87o92hacvy.fsf@zen.linaroharston \
    --to=alex.bennee@linaro.org \
    --cc=damien.hedde@greensocs.com \
    --cc=edgari@xilinx.com \
    --cc=luc.michel@greensocs.com \
    --cc=mark.burton@greensocs.com \
    --cc=pbonzini@redhat.com \
    --cc=qemu-devel@nongnu.org \
    --cc=rth@twiddle.net \
    --cc=sakisp@xilinx.com \
    /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.