From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([209.51.188.92]:35118) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1hGJVX-0002sd-UY for qemu-devel@nongnu.org; Tue, 16 Apr 2019 04:32:01 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1hGJVW-0006Oi-Lw for qemu-devel@nongnu.org; Tue, 16 Apr 2019 04:31:59 -0400 Received: from mail-pl1-x642.google.com ([2607:f8b0:4864:20::642]:46578) by eggs.gnu.org with esmtps (TLS1.0:RSA_AES_128_CBC_SHA1:16) (Exim 4.71) (envelope-from ) id 1hGJVW-0006Mj-Aw for qemu-devel@nongnu.org; Tue, 16 Apr 2019 04:31:58 -0400 Received: by mail-pl1-x642.google.com with SMTP id y6so9945366pll.13 for ; Tue, 16 Apr 2019 01:31:58 -0700 (PDT) From: Richard Henderson Date: Mon, 15 Apr 2019 22:31:50 -1000 Message-Id: <20190416083150.19649-3-richard.henderson@linaro.org> In-Reply-To: <20190416083150.19649-1-richard.henderson@linaro.org> References: <20190416083150.19649-1-richard.henderson@linaro.org> Subject: [Qemu-devel] [PATCH 2/2] tcg: Restart after TB code generation overflow List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-devel@nongnu.org Cc: peter.maydell@linaro.org If a TB generates too much code, try again with fewer insns. Fixes: https://bugs.launchpad.net/bugs/1824853 Signed-off-by: Richard Henderson --- accel/tcg/translate-all.c | 39 +++++++++++++++++++++++++++++++++------ tcg/tcg.c | 4 ++++ 2 files changed, 37 insertions(+), 6 deletions(-) diff --git a/accel/tcg/translate-all.c b/accel/tcg/translate-all.c index fcdfe6c0ec..7086d5fd2a 100644 --- a/accel/tcg/translate-all.c +++ b/accel/tcg/translate-all.c @@ -1721,6 +1721,7 @@ TranslationBlock *tb_gen_code(CPUState *cpu, tb->cflags = cflags; tb->trace_vcpu_dstate = *cpu->trace_dstate; tcg_ctx->tb_cflags = cflags; + tb_overflow: #ifdef CONFIG_PROFILER /* includes aborted translations because of exceptions */ @@ -1754,14 +1755,40 @@ TranslationBlock *tb_gen_code(CPUState *cpu, ti = profile_getclock(); #endif - /* ??? Overflow could be handled better here. In particular, we - don't need to re-do gen_intermediate_code, nor should we re-do - the tcg optimization currently hidden inside tcg_gen_code. All - that should be required is to flush the TBs, allocate a new TB, - re-initialize it per above, and re-do the actual code generation. */ gen_code_size = tcg_gen_code(tcg_ctx, tb); if (unlikely(gen_code_size < 0)) { - goto buffer_overflow; + switch (gen_code_size) { + case -1: + /* + * Overflow of code_gen_buffer, or the current slice of it. + * + * TODO: We don't need to re-do gen_intermediate_code, nor + * should we re-do the tcg optimization currently hidden + * inside tcg_gen_code. All that should be required is to + * flush the TBs, allocate a new TB, re-initialize it per + * above, and re-do the actual code generation. + */ + goto buffer_overflow; + + case -2: + /* + * The code generated for the TranslationBlock is too large. + * The maximum size allowed by the unwind info is 64k. + * There may be stricter constraints from relocations + * in the tcg backend. + * + * Try again with half as many insns as we attempted this time. + * If a single insn overflows, there's a bug somewhere... + */ + max_insns = tb->icount; +qemu_log("TB overflow: pc=%lx insns=%d\n", (unsigned long)pc, max_insns); + assert(max_insns > 1); + max_insns /= 2; + goto tb_overflow; + + default: + g_assert_not_reached(); + } } search_size = encode_search(tb, (void *)gen_code_buf + gen_code_size); if (unlikely(search_size < 0)) { diff --git a/tcg/tcg.c b/tcg/tcg.c index 9b2bf7f439..aa0e94521b 100644 --- a/tcg/tcg.c +++ b/tcg/tcg.c @@ -3991,6 +3991,10 @@ int tcg_gen_code(TCGContext *s, TranslationBlock *tb) if (unlikely((void *)s->code_ptr > s->code_gen_highwater)) { return -1; } + /* Test for TB overflow, as seen by gen_insn_end_off. */ + if (unlikely(tcg_current_code_size(s) > UINT16_MAX)) { + return -2; + } } tcg_debug_assert(num_insns >= 0); s->gen_insn_end_off[num_insns] = tcg_current_code_size(s); -- 2.17.1