qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: "Edgar E. Iglesias" <edgar.iglesias@gmail.com>
To: Richard Henderson <richard.henderson@linaro.org>
Cc: qemu-devel@nongnu.org
Subject: Re: [PATCH 62/77] target/microblaze: Try to keep imm and delay slot together
Date: Thu, 27 Aug 2020 21:17:22 +0200	[thread overview]
Message-ID: <20200827191722.GP2954729@toto> (raw)
In-Reply-To: <20200825205950.730499-63-richard.henderson@linaro.org>

On Tue, Aug 25, 2020 at 01:59:35PM -0700, Richard Henderson wrote:
> If the last insn on a page is imm, or a branch with delay slot,
> then end a tb early if this has not begun the tb.  If it has
> begun the tb, then we can allow the tb to span two pages as if
> the imm plus its consumer, or branch plus delay, or imm plus
> branch plus delay, are a single insn.
> 
> If the insn in the delay slot faults, then the exception handler
> will have reset the PC to the beginning of this sequence anyway,
> via the stored D_FLAG and BIMM_FLAG bits.
> 
> Disable all of this when single-stepping.


Hi Richard,

We've got a Linux boot that fails after applying this patch.
It goes from always working to only working like 1 out of 3 times.
It fails deep in user-space so I don't have a good log for it.

I guess we'll have to review this patch more.

I can share images but the machine is (DTB) dynamic so it only runs with
Xilinx QEMU 5.1 branch (I've backported your series for testing).

Cheers,
Edgar

> 
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> ---
>  target/microblaze/translate.c | 65 ++++++++++++++++++++++++++++++-----
>  1 file changed, 56 insertions(+), 9 deletions(-)
> 
> diff --git a/target/microblaze/translate.c b/target/microblaze/translate.c
> index 4675326083..fcfc1ac184 100644
> --- a/target/microblaze/translate.c
> +++ b/target/microblaze/translate.c
> @@ -530,11 +530,50 @@ static void gen_idivu(TCGv_i32 out, TCGv_i32 ina, TCGv_i32 inb)
>  DO_TYPEA_CFG(idiv, use_div, true, gen_idiv)
>  DO_TYPEA_CFG(idivu, use_div, true, gen_idivu)
>  
> +/*
> + * Try to keep the current instruction with the one following.
> + * So if this insn is the last in the TB, and is not the first
> + * in the TB, and we are not singlestepping, then back up and
> + * exit the current TB.
> + */
> +static bool wait_for_next_tb(DisasContext *dc)
> +{
> +    if (dc->base.num_insns >= dc->base.max_insns
> +        && !dc->base.singlestep_enabled) {
> +        /* Also consider if this insn (e.g. brid) itself uses an imm. */
> +        int ninsns = (dc->tb_flags & IMM_FLAG ? 2 : 1);
> +
> +        /*
> +         * If this is not the first insn in the TB, back up and
> +         * start again with a new TB.
> +         */
> +        if (dc->base.num_insns > ninsns) {
> +            dc->base.pc_next -= ninsns * 4;
> +            dc->base.num_insns -= ninsns;
> +            dc->base.is_jmp = DISAS_TOO_MANY;
> +            return true;
> +        }
> +
> +        /*
> +         * Correspondingly, if this is the first insn of the TB,
> +         * then extend the TB as necessary to keep it with the
> +         * next insn.  Do this by *reducing* the number of insns
> +         * processed by this TB so that icount does fail an assertion.
> +         */
> +        if (dc->base.num_insns == ninsns) {
> +            dc->base.num_insns = 0;
> +        }
> +    }
> +    return false;
> +}
> +
>  static bool trans_imm(DisasContext *dc, arg_imm *arg)
>  {
> -    dc->ext_imm = arg->imm << 16;
> -    tcg_gen_movi_i32(cpu_imm, dc->ext_imm);
> -    dc->tb_flags_to_set = IMM_FLAG;
> +    if (!wait_for_next_tb(dc)) {
> +        dc->ext_imm = arg->imm << 16;
> +        tcg_gen_movi_i32(cpu_imm, dc->ext_imm);
> +        dc->tb_flags_to_set = IMM_FLAG;
> +    }
>      return true;
>  }
>  
> @@ -1311,12 +1350,17 @@ static void eval_cond_jmp(DisasContext *dc, TCGv_i32 pc_true, TCGv_i32 pc_false)
>      tcg_temp_free_i32(zero);
>  }
>  
> -static void dec_setup_dslot(DisasContext *dc)
> +static bool dec_setup_dslot(DisasContext *dc)
>  {
> +    if (wait_for_next_tb(dc)) {
> +        return true;
> +    }
> +
>      dc->tb_flags_to_set |= D_FLAG;
>      if (dc->type_b && (dc->tb_flags & IMM_FLAG)) {
>          dc->tb_flags_to_set |= BIMM_FLAG;
>      }
> +    return false;
>  }
>  
>  static void dec_bcc(DisasContext *dc)
> @@ -1327,8 +1371,8 @@ static void dec_bcc(DisasContext *dc)
>      cc = EXTRACT_FIELD(dc->ir, 21, 23);
>      dslot = dc->ir & (1 << 25);
>  
> -    if (dslot) {
> -        dec_setup_dslot(dc);
> +    if (dslot && dec_setup_dslot(dc)) {
> +        return;
>      }
>  
>      if (dc->type_b) {
> @@ -1402,9 +1446,10 @@ static void dec_br(DisasContext *dc)
>          }
>      }
>  
> -    if (dslot) {
> -        dec_setup_dslot(dc);
> +    if (dslot && dec_setup_dslot(dc)) {
> +        return;
>      }
> +
>      if (link && dc->rd) {
>          tcg_gen_movi_i32(cpu_R[dc->rd], dc->base.pc_next);
>      }
> @@ -1513,7 +1558,9 @@ static void dec_rts(DisasContext *dc)
>          return;
>      }
>  
> -    dec_setup_dslot(dc);
> +    if (dec_setup_dslot(dc)) {
> +        return;
> +    }
>  
>      if (i_bit) {
>          dc->tb_flags |= DRTI_FLAG;
> -- 
> 2.25.1
> 


  reply	other threads:[~2020-08-27 19:18 UTC|newest]

Thread overview: 102+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-08-25 20:58 [PATCH 00/77] target/microblaze improvements Richard Henderson
2020-08-25 20:58 ` [PATCH 01/77] tests/tcg: Add microblaze to arches filter Richard Henderson
2020-08-26 16:20   ` Edgar E. Iglesias
2020-08-25 20:58 ` [PATCH 02/77] tests/tcg: Do not require FE_TOWARDZERO Richard Henderson
2020-08-26 16:20   ` Edgar E. Iglesias
2020-08-25 20:58 ` [PATCH 03/77] tests/tcg: Do not require FE_* exception bits Richard Henderson
2020-08-26 16:21   ` Edgar E. Iglesias
2020-08-25 20:58 ` [PATCH 04/77] target/microblaze: Tidy gdbstub Richard Henderson
2020-08-25 20:58 ` [PATCH 05/77] target/microblaze: Split out PC from env->sregs Richard Henderson
2020-08-25 20:58 ` [PATCH 06/77] target/microblaze: Split out MSR " Richard Henderson
2020-08-25 20:58 ` [PATCH 07/77] target/microblaze: Split out EAR " Richard Henderson
2020-08-25 20:58 ` [PATCH 08/77] target/microblaze: Split out ESR " Richard Henderson
2020-08-25 20:58 ` [PATCH 09/77] target/microblaze: Split out FSR " Richard Henderson
2020-08-25 20:58 ` [PATCH 10/77] target/microblaze: Split out BTR " Richard Henderson
2020-08-25 20:58 ` [PATCH 11/77] target/microblaze: Split out EDR " Richard Henderson
2020-08-25 20:58 ` [PATCH 12/77] target/microblaze: Split the cpu_SR array Richard Henderson
2020-08-25 20:58 ` [PATCH 13/77] target/microblaze: Fix width of PC and BTARGET Richard Henderson
2020-08-25 20:58 ` [PATCH 14/77] target/microblaze: Fix width of MSR Richard Henderson
2020-08-25 20:58 ` [PATCH 15/77] target/microblaze: Fix width of ESR Richard Henderson
2020-08-25 20:58 ` [PATCH 16/77] target/microblaze: Fix width of FSR Richard Henderson
2020-08-25 20:58 ` [PATCH 17/77] target/microblaze: Fix width of BTR Richard Henderson
2020-08-25 20:58 ` [PATCH 18/77] target/microblaze: Fix width of EDR Richard Henderson
2020-08-25 20:58 ` [PATCH 19/77] target/microblaze: Remove cpu_ear Richard Henderson
2020-08-25 20:58 ` [PATCH 20/77] target/microblaze: Tidy raising of exceptions Richard Henderson
2020-08-25 20:58 ` [PATCH 21/77] target/microblaze: Mark raise_exception as noreturn Richard Henderson
2020-08-25 20:58 ` [PATCH 22/77] target/microblaze: Remove helper_debug and env->debug Richard Henderson
2020-08-25 20:58 ` [PATCH 23/77] target/microblaze: Rename env_* tcg variables to cpu_* Richard Henderson
2020-08-25 20:58 ` [PATCH 24/77] target/microblaze: Tidy mb_tcg_init Richard Henderson
2020-08-25 20:58 ` [PATCH 25/77] target/microblaze: Split out MSR[C] to its own variable Richard Henderson
2020-08-25 20:58 ` [PATCH 26/77] target/microblaze: Use DISAS_NORETURN Richard Henderson
2020-08-25 20:59 ` [PATCH 27/77] target/microblaze: Check singlestep_enabled in gen_goto_tb Richard Henderson
2020-08-25 20:59 ` [PATCH 28/77] target/microblaze: Convert to DisasContextBase Richard Henderson
2020-08-25 20:59 ` [PATCH 29/77] target/microblaze: Convert to translator_loop Richard Henderson
2020-08-25 20:59 ` [PATCH 30/77] target/microblaze: Remove SIM_COMPAT Richard Henderson
2020-08-25 20:59 ` [PATCH 31/77] target/microblaze: Remove DISAS_GNU Richard Henderson
2020-08-25 20:59 ` [PATCH 32/77] target/microblaze: Remove empty D macros Richard Henderson
2020-08-25 20:59 ` [PATCH 33/77] target/microblaze: Remove LOG_DIS Richard Henderson
2020-08-25 20:59 ` [PATCH 34/77] target/microblaze: Ensure imm constant is always available Richard Henderson
2020-08-25 20:59 ` [PATCH 35/77] target/microblaze: Add decodetree infrastructure Richard Henderson
2020-08-25 20:59 ` [PATCH 36/77] target/microblaze: Convert dec_add to decodetree Richard Henderson
2020-08-25 20:59 ` [PATCH 37/77] target/microblaze: Convert dec_sub " Richard Henderson
2020-08-25 20:59 ` [PATCH 38/77] target/microblaze: Implement cmp and cmpu inline Richard Henderson
2020-08-25 20:59 ` [PATCH 39/77] target/microblaze: Convert dec_pattern to decodetree Richard Henderson
2020-08-25 20:59 ` [PATCH 40/77] target/microblaze: Convert dec_and, dec_or, dec_xor " Richard Henderson
2020-08-25 20:59 ` [PATCH 41/77] target/microblaze: Convert dec_mul " Richard Henderson
2020-08-25 20:59 ` [PATCH 42/77] target/microblaze: Convert dec_div " Richard Henderson
2020-08-25 20:59 ` [PATCH 43/77] target/microblaze: Unwind properly when raising divide-by-zero Richard Henderson
2020-08-25 20:59 ` [PATCH 44/77] target/microblaze: Convert dec_bit to decodetree Richard Henderson
2020-08-25 20:59 ` [PATCH 45/77] target/microblaze: Convert dec_barrel " Richard Henderson
2020-08-25 20:59 ` [PATCH 46/77] target/microblaze: Convert dec_imm " Richard Henderson
2020-08-25 20:59 ` [PATCH 47/77] target/microblaze: Convert dec_fpu " Richard Henderson
2020-08-25 20:59 ` [PATCH 48/77] target/microblaze: Fix cpu unwind for fpu exceptions Richard Henderson
2020-08-25 20:59 ` [PATCH 49/77] target/microblaze: Mark fpu helpers TCG_CALL_NO_WG Richard Henderson
2020-08-25 20:59 ` [PATCH 50/77] target/microblaze: Replace MSR_EE_FLAG with MSR_EE Richard Henderson
2020-08-25 20:59 ` [PATCH 51/77] target/microblaze: Cache mem_index in DisasContext Richard Henderson
2020-08-25 20:59 ` [PATCH 52/77] target/microblaze: Fix cpu unwind for stackprot Richard Henderson
2020-08-25 20:59 ` [PATCH 53/77] target/microblaze: Convert dec_load and dec_store to decodetree Richard Henderson
2020-08-25 20:59 ` [PATCH 54/77] target/microblaze: Assert no overlap in flags making up tb_flags Richard Henderson
2020-08-25 20:59 ` [PATCH 55/77] target/microblaze: Move bimm to BIMM_FLAG Richard Henderson
2020-08-25 20:59 ` [PATCH 56/77] target/microblaze: Store "current" iflags in insn_start Richard Henderson
2020-08-25 20:59 ` [PATCH 57/77] tcg: Add tcg_get_insn_start_param Richard Henderson
2020-08-25 20:59 ` [PATCH 58/77] target/microblaze: Use cc->do_unaligned_access Richard Henderson
2020-08-25 20:59 ` [PATCH 59/77] target/microblaze: Replace clear_imm with tb_flags_to_set Richard Henderson
2020-08-25 20:59 ` [PATCH 60/77] target/microblaze: Replace delayed_branch " Richard Henderson
2020-08-25 20:59 ` [PATCH 61/77] target/microblaze: Tidy mb_cpu_dump_state Richard Henderson
2020-08-25 20:59 ` [PATCH 62/77] target/microblaze: Try to keep imm and delay slot together Richard Henderson
2020-08-27 19:17   ` Edgar E. Iglesias [this message]
2020-08-27 21:33     ` Richard Henderson
2020-08-29 15:27     ` Richard Henderson
2020-08-25 20:59 ` [PATCH 63/77] target/microblaze: Convert brk and brki to decodetree Richard Henderson
2020-08-25 20:59 ` [PATCH 64/77] target/microblaze: Convert mbar " Richard Henderson
2020-08-27  9:24   ` Edgar E. Iglesias
2020-08-27  9:58     ` Richard Henderson
2020-08-27 10:08       ` Edgar E. Iglesias
2020-08-27 11:11         ` Richard Henderson
2020-08-25 20:59 ` [PATCH 65/77] target/microblaze: Reorganize branching Richard Henderson
2020-08-25 20:59 ` [PATCH 66/77] target/microblaze: Use tcg_gen_lookup_and_goto_ptr Richard Henderson
2020-08-28  6:33   ` Edgar E. Iglesias
2020-08-28 13:32     ` Richard Henderson
2020-08-25 20:59 ` [PATCH 67/77] target/microblaze: Convert dec_br to decodetree Richard Henderson
2020-08-25 20:59 ` [PATCH 68/77] target/microblaze: Convert dec_bcc " Richard Henderson
2020-08-25 20:59 ` [PATCH 69/77] target/microblaze: Convert dec_rts " Richard Henderson
2020-08-25 20:59 ` [PATCH 70/77] target/microblaze: Tidy do_rti, do_rtb, do_rte Richard Henderson
2020-08-25 20:59 ` [PATCH 71/77] target/microblaze: Convert msrclr, msrset to decodetree Richard Henderson
2020-08-25 20:59 ` [PATCH 72/77] target/microblaze: Convert dec_msr " Richard Henderson
2020-08-25 20:59 ` [PATCH 73/77] target/microblaze: Convert dec_stream " Richard Henderson
2020-08-27 21:10   ` Edgar E. Iglesias
2020-08-27 21:12     ` Richard Henderson
2020-08-25 20:59 ` [PATCH 74/77] target/microblaze: Remove last of old decoder Richard Henderson
2020-08-25 20:59 ` [PATCH 75/77] target/microblaze: Remove cpu_R[0] Richard Henderson
2020-08-25 20:59 ` [PATCH 76/77] target/microblaze: Add flags markup to some helpers Richard Henderson
2020-08-25 20:59 ` [PATCH 77/77] target/microblaze: Reduce linux-user address space to 32-bit Richard Henderson
2020-08-26 15:27 ` [PATCH 00/77] target/microblaze improvements Edgar E. Iglesias
2020-08-26 18:07 ` Edgar E. Iglesias
2020-08-27  9:11 ` Edgar E. Iglesias
2020-08-27 10:01   ` Richard Henderson
2020-08-27 10:22     ` Edgar E. Iglesias
2020-08-27 11:19       ` Richard Henderson
2020-08-27 17:09         ` Edgar E. Iglesias
2020-08-28 13:36           ` Richard Henderson
2020-08-28 14:04             ` Edgar E. Iglesias
2020-08-28 13:19 ` Edgar E. Iglesias

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=20200827191722.GP2954729@toto \
    --to=edgar.iglesias@gmail.com \
    --cc=qemu-devel@nongnu.org \
    --cc=richard.henderson@linaro.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).