From patchwork Tue Oct 17 13:05:20 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Adhemerval Zanella X-Patchwork-Id: 78030 X-Patchwork-Delegate: siddhesh@gotplt.org Return-Path: X-Original-To: patchwork@sourceware.org Delivered-To: patchwork@sourceware.org Received: from server2.sourceware.org (localhost [IPv6:::1]) by sourceware.org (Postfix) with ESMTP id C8383382798C for ; Tue, 17 Oct 2023 13:08:18 +0000 (GMT) X-Original-To: libc-alpha@sourceware.org Delivered-To: libc-alpha@sourceware.org Received: from mail-pf1-x434.google.com (mail-pf1-x434.google.com [IPv6:2607:f8b0:4864:20::434]) by sourceware.org (Postfix) with ESMTPS id C0535385B534 for ; Tue, 17 Oct 2023 13:06:41 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.2 sourceware.org C0535385B534 Authentication-Results: sourceware.org; dmarc=pass (p=none dis=none) header.from=linaro.org Authentication-Results: sourceware.org; spf=pass smtp.mailfrom=linaro.org ARC-Filter: OpenARC Filter v1.0.0 sourceware.org C0535385B534 Authentication-Results: server2.sourceware.org; arc=none smtp.remote-ip=2607:f8b0:4864:20::434 ARC-Seal: i=1; a=rsa-sha256; d=sourceware.org; s=key; t=1697548006; cv=none; b=wNJpSJLv1o1ME3kUrYtDHdIQxn/nfTq3Tdh67xQuYNfrmFHHstCO4DBz9fOuiDtWgUkj1iSWur7JAXSBEWX7+Tqi8lgKhkRQz58p47fokbcfxvVoWXSOeoPeps3WDxUdQf482PyRW79gZKIVlqBlKEvVgdYDMYeyvLzI/uHCgek= ARC-Message-Signature: i=1; a=rsa-sha256; d=sourceware.org; s=key; t=1697548006; c=relaxed/simple; bh=7U8/btXGyk9lcXhF8xAUbYQclqnzPW9ncQyeubCaGy4=; h=DKIM-Signature:From:To:Subject:Date:Message-Id:MIME-Version; b=fnj5V7PcP317xJs8yvv1BB/PbdhPbbrajjifFp/+XV5UoyAM7PnXVgVcc6tEeYQXdvsA2mGs7BkiIpBrsMUy+bETcFhkGyZ7waG8PfBY4FooaYd9KDpAcNTlR3FjmL6E0iMpHltDhvQlNE1h06rx7uGxSZ6Iv8wInRWJqWYEjK4= ARC-Authentication-Results: i=1; server2.sourceware.org Received: by mail-pf1-x434.google.com with SMTP id d2e1a72fcca58-6b1d1099a84so3943150b3a.1 for ; Tue, 17 Oct 2023 06:06:41 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=linaro.org; s=google; t=1697548000; x=1698152800; darn=sourceware.org; h=content-transfer-encoding:mime-version:references:in-reply-to :message-id:date:subject:to:from:from:to:cc:subject:date:message-id :reply-to; bh=o8g3niH12Wf/zF2OR8oNKHV32Jg92xyT2wyah0zxCWI=; b=nm6tE0f48P+I6eScxQZZbdA3LQqvnzrYQpduhff+g3bsBqECF/bYzHjkFuzNqwjU8O QNiv8DWY+klws0OPWPCj0RBoGHquhVDgTMMO2aM9LIImK8hfjVFdbocjGKRSTFGZy1YO lCKhpYW1Vhrrv4RoBUyPIpFHbLBTIbnp/p6NsgcbaATdS0WFNG1Py//MNwO4u9hpHh8T aT2ctBp54rIC5+xuKFs5YgCCp41iLXGZSKaPiQ+1tuLqiCCcxf1wIyYLAKDz6v2TKy9S hSyx9dp5G2boLjWBVIi/eeyLr3MFvHYXmSwtn1IjDEMs1Z8/FA9AkwX8l+ceUFzkmeXe FTIw== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20230601; t=1697548000; x=1698152800; h=content-transfer-encoding:mime-version:references:in-reply-to :message-id:date:subject:to:from:x-gm-message-state:from:to:cc :subject:date:message-id:reply-to; bh=o8g3niH12Wf/zF2OR8oNKHV32Jg92xyT2wyah0zxCWI=; b=hDimkhCsnwpJUloui4XNV1W5fudGBJ+kU0mcFHbvwOxTRJT+yDWCxeIz0GZLgj9/FT CEk6Wa2ZSFCfJUp63zuGGgus9WCGVPQZ8jvDUV60eiWU7pXDwxKtGxqk2JhE6zZ9fOi1 mbNCl0haLVxK7asCd8PMxACO98fl7Qgi5GBu9hlb8FN2ZdXfulN0382TgLyuQEviv9Im OMjl4OpRZB7bygNYT1Cp/aSkJd/sBfZw48DtfEVvcjvrCqrwwilg4M4BjY9AXd7Ydq/K MwmDxrVisfk1duSwNDiN0lLuL3Y9yh34r1QUfXfp6evM2V+1KGgTo61FPdV2q4g7PJRE Y/kQ== X-Gm-Message-State: AOJu0Yyc/LQs+YcAIHkcefGasndj0n6DT40pm2o4FzMlLEBfRH6fng1i GIc5U/wRGjHqWszC50mX4frj3budNZuYhNEQ0BsuxQ== X-Google-Smtp-Source: AGHT+IHk2Cby31UwI8ougNcq0+i0rXq9Wpr1x9k7HQoK+wiw+eua0OeHiCJ6wHWOcLettAjavpdC1Q== X-Received: by 2002:a05:6a21:a108:b0:154:a1e4:b676 with SMTP id aq8-20020a056a21a10800b00154a1e4b676mr1943756pzc.4.1697547999733; Tue, 17 Oct 2023 06:06:39 -0700 (PDT) Received: from mandiga.. ([2804:1b3:a7c3:7f2e:11d:92b4:4d78:4197]) by smtp.gmail.com with ESMTPSA id l28-20020a635b5c000000b0056b6d1ac949sm1309788pgm.13.2023.10.17.06.06.38 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Tue, 17 Oct 2023 06:06:39 -0700 (PDT) From: Adhemerval Zanella To: libc-alpha@sourceware.org, Siddhesh Poyarekar Subject: [PATCH v2 13/19] elf: Remove LD_PROFILE for static binaries Date: Tue, 17 Oct 2023 10:05:20 -0300 Message-Id: <20231017130526.2216827-14-adhemerval.zanella@linaro.org> X-Mailer: git-send-email 2.34.1 In-Reply-To: <20231017130526.2216827-1-adhemerval.zanella@linaro.org> References: <20231017130526.2216827-1-adhemerval.zanella@linaro.org> MIME-Version: 1.0 X-Spam-Status: No, score=-13.7 required=5.0 tests=BAYES_00, DKIM_SIGNED, DKIM_VALID, DKIM_VALID_AU, DKIM_VALID_EF, GIT_PATCH_0, RCVD_IN_DNSWL_NONE, SPF_HELO_NONE, SPF_PASS, TXREP autolearn=unavailable autolearn_force=no version=3.4.6 X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on server2.sourceware.org X-BeenThere: libc-alpha@sourceware.org X-Mailman-Version: 2.1.30 Precedence: list List-Id: Libc-alpha mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: libc-alpha-bounces+patchwork=sourceware.org@sourceware.org The _dl_non_dynamic_init does not parse LD_PROFILE, which does not enable profile for dlopen objects. Since dlopen is deprecated for static objects, it is better to remove the support. It also allows to trim down libc.a of profile support. Checked on x86_64-linux-gnu. --- elf/Makefile | 10 ++-- elf/dl-load.c | 10 ++-- elf/dl-runtime.c | 12 ++--- elf/dl-support.c | 9 ---- elf/{dl-profstub.c => libc-dl-profstub.c} | 0 include/dlfcn.h | 5 ++ sysdeps/aarch64/dl-machine.h | 4 +- sysdeps/aarch64/dl-trampoline.S | 3 +- sysdeps/alpha/dl-machine.h | 6 ++- sysdeps/alpha/dl-trampoline.S | 6 +++ sysdeps/arm/dl-machine.h | 4 +- sysdeps/arm/dl-trampoline.S | 2 +- sysdeps/hppa/dl-machine.h | 36 +++++++------ sysdeps/hppa/dl-trampoline.S | 2 + sysdeps/i386/dl-machine.h | 2 + sysdeps/i386/dl-trampoline.S | 2 +- sysdeps/ia64/dl-machine.h | 10 ++-- sysdeps/ia64/dl-trampoline.S | 2 +- sysdeps/loongarch/dl-machine.h | 6 ++- sysdeps/loongarch/dl-trampoline.h | 2 + sysdeps/m68k/dl-machine.h | 4 +- sysdeps/m68k/dl-trampoline.S | 2 + sysdeps/powerpc/powerpc32/dl-machine.c | 2 +- sysdeps/powerpc/powerpc32/dl-machine.h | 10 ++-- sysdeps/powerpc/powerpc32/dl-trampoline.S | 2 +- sysdeps/powerpc/powerpc64/dl-machine.h | 20 ++++--- sysdeps/powerpc/powerpc64/dl-trampoline.S | 2 +- sysdeps/s390/s390-32/dl-machine.h | 8 +-- sysdeps/s390/s390-32/dl-trampoline.h | 2 +- sysdeps/s390/s390-64/dl-machine.h | 8 +-- sysdeps/s390/s390-64/dl-trampoline.h | 2 +- sysdeps/sh/dl-machine.h | 2 + sysdeps/sh/dl-trampoline.S | 2 + sysdeps/sparc/sparc32/dl-machine.h | 4 +- sysdeps/sparc/sparc32/dl-trampoline.S | 2 + sysdeps/sparc/sparc64/dl-machine.h | 4 +- sysdeps/sparc/sparc64/dl-trampoline.S | 2 + sysdeps/x86_64/dl-machine.h | 2 + sysdeps/x86_64/dl-trampoline.S | 64 ++++++++++++----------- 39 files changed, 168 insertions(+), 109 deletions(-) rename elf/{dl-profstub.c => libc-dl-profstub.c} (100%) diff --git a/elf/Makefile b/elf/Makefile index 608bef46f5..e16c373a1c 100644 --- a/elf/Makefile +++ b/elf/Makefile @@ -37,12 +37,12 @@ routines = \ dl-iteratephdr \ dl-libc \ dl-origin \ - dl-profstub \ dl-reloc-static-pie \ dl-support \ dl-sym \ dl-sysdep \ enbl-secure \ + libc-dl-profstub \ libc-dl_find_object \ libc_early_init \ rtld_static_init \ @@ -72,7 +72,6 @@ dl-routines = \ dl-open \ dl-origin \ dl-printf \ - dl-profile \ dl-reloc \ dl-runtime \ dl-scope \ @@ -117,7 +116,11 @@ elide-routines.os = \ # elide-routines.os # These object files are only included in the dynamically-linked libc. -shared-only-routines = libc-dl_find_object +shared-only-routines = \ + libc-dl-profile \ + libc-dl-profstub \ + libc-dl_find_object \ + # shared-only-routines # ld.so uses those routines, plus some special stuff for being the program # interpreter and operating independent of libc. @@ -135,6 +138,7 @@ rtld-routines = \ dl-libc_freeres \ dl-minimal \ dl-mutex \ + dl-profile \ dl-sysdep \ dl-usage \ rtld \ diff --git a/elf/dl-load.c b/elf/dl-load.c index 2923b1141d..7356a4fe48 100644 --- a/elf/dl-load.c +++ b/elf/dl-load.c @@ -1443,11 +1443,6 @@ cannot enable executable stack as shared object requires"); name by which the DSO is actually known. Add that as well. */ if (__glibc_unlikely (origname != NULL)) add_name_to_object (l, origname); -#else - /* Audit modules only exist when linking is dynamic so ORIGNAME - cannot be non-NULL. */ - assert (origname == NULL); -#endif /* When we profile the SONAME might be needed for something else but loading. Add it right away. */ @@ -1455,6 +1450,11 @@ cannot enable executable stack as shared object requires"); && l->l_info[DT_SONAME] != NULL) add_name_to_object (l, ((const char *) D_PTR (l, l_info[DT_STRTAB]) + l->l_info[DT_SONAME]->d_un.d_val)); +#else + /* Audit modules only exist when linking is dynamic so ORIGNAME + cannot be non-NULL. */ + assert (origname == NULL); +#endif /* If we have newly loaded libc.so, update the namespace description. */ diff --git a/elf/dl-runtime.c b/elf/dl-runtime.c index 32a8bfcf74..fe7deda32a 100644 --- a/elf/dl-runtime.c +++ b/elf/dl-runtime.c @@ -162,14 +162,14 @@ _dl_fixup ( return elf_machine_fixup_plt (l, result, refsym, sym, reloc, rel_addr, value); } -#ifndef PROF +#if !defined PROF && defined SHARED DL_FIXUP_VALUE_TYPE __attribute ((noinline)) DL_ARCH_FIXUP_ATTRIBUTE _dl_profile_fixup ( -#ifdef ELF_MACHINE_RUNTIME_FIXUP_ARGS +# ifdef ELF_MACHINE_RUNTIME_FIXUP_ARGS ELF_MACHINE_RUNTIME_FIXUP_ARGS, -#endif +# endif struct link_map *l, ElfW(Word) reloc_arg, ElfW(Addr) retaddr, void *regs, long int *framesizep) { @@ -309,14 +309,12 @@ _dl_profile_fixup ( /* And now perhaps the relocation addend. */ value = elf_machine_plt_value (l, reloc, value); -#ifdef SHARED /* Auditing checkpoint: we have a new binding. Provide the auditing libraries the possibility to change the value and tell us whether further auditing is wanted. */ if (defsym != NULL && GLRO(dl_naudit) > 0) _dl_audit_symbind (l, reloc_result, reloc, defsym, &value, result, true); -#endif /* Store the result for later runs. */ if (__glibc_likely (! GLRO(dl_bind_not))) @@ -335,11 +333,9 @@ _dl_profile_fixup ( long int framesize = -1; -#ifdef SHARED /* Auditing checkpoint: report the PLT entering and allow the auditors to change the value. */ _dl_audit_pltenter (l, reloc_result, &value, regs, &framesize); -#endif /* Store the frame size information. */ *framesizep = framesize; @@ -349,4 +345,4 @@ _dl_profile_fixup ( return value; } -#endif /* PROF */ +#endif /* !defined PROF && defined SHARED */ diff --git a/elf/dl-support.c b/elf/dl-support.c index 44a54dea07..31a608df87 100644 --- a/elf/dl-support.c +++ b/elf/dl-support.c @@ -60,10 +60,6 @@ int _dl_dynamic_weak; /* If nonzero print warnings about problematic situations. */ int _dl_verbose; -/* We never do profiling. */ -const char *_dl_profile; -const char *_dl_profile_output; - /* Names of shared object for which the RUNPATHs and RPATHs should be ignored. */ const char *_dl_inhibit_rpath; @@ -301,11 +297,6 @@ _dl_non_dynamic_init (void) _dl_dynamic_weak = *(getenv ("LD_DYNAMIC_WEAK") ?: "") == '\0'; - _dl_profile_output = getenv ("LD_PROFILE_OUTPUT"); - if (_dl_profile_output == NULL || _dl_profile_output[0] == '\0') - _dl_profile_output - = &"/var/tmp\0/var/profile"[__libc_enable_secure ? 9 : 0]; - if (__libc_enable_secure) { static const char unsecure_envvars[] = diff --git a/elf/dl-profstub.c b/elf/libc-dl-profstub.c similarity index 100% rename from elf/dl-profstub.c rename to elf/libc-dl-profstub.c diff --git a/include/dlfcn.h b/include/dlfcn.h index ae25f05303..a44420fa37 100644 --- a/include/dlfcn.h +++ b/include/dlfcn.h @@ -135,5 +135,10 @@ extern int __dladdr1 (const void *address, Dl_info *info, extern int __dlinfo (void *handle, int request, void *arg); extern char *__dlerror (void); +#ifndef SHARED +# undef DL_CALL_FCT +# define DL_CALL_FCT(fctp, args) ((fctp) args) +#endif + #endif #endif diff --git a/sysdeps/aarch64/dl-machine.h b/sysdeps/aarch64/dl-machine.h index 4170b9269f..a56eb96a79 100644 --- a/sysdeps/aarch64/dl-machine.h +++ b/sysdeps/aarch64/dl-machine.h @@ -68,7 +68,6 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], { ElfW(Addr) *got; extern void _dl_runtime_resolve (ElfW(Word)); - extern void _dl_runtime_profile (ElfW(Word)); got = (ElfW(Addr) *) D_PTR (l, l_info[DT_PLTGOT]); if (got[1]) @@ -83,6 +82,8 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], to intercept the calls to collect information. In this case we don't store the address in the GOT so that all future calls also end in this function. */ +#ifdef SHARED + extern void _dl_runtime_profile (ElfW(Word)); if ( profile) { got[2] = (ElfW(Addr)) &_dl_runtime_profile; @@ -94,6 +95,7 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], GL(dl_profile_map) = l; } else +#endif { /* This function will get called to fix up the GOT entry indicated by the offset on the stack, and then jump to diff --git a/sysdeps/aarch64/dl-trampoline.S b/sysdeps/aarch64/dl-trampoline.S index a3474ba741..d9b3dcd656 100644 --- a/sysdeps/aarch64/dl-trampoline.S +++ b/sysdeps/aarch64/dl-trampoline.S @@ -122,7 +122,8 @@ _dl_runtime_resolve: cfi_endproc .size _dl_runtime_resolve, .-_dl_runtime_resolve -#ifndef PROF + +#if !defined PROF && defined SHARED .globl _dl_runtime_profile .type _dl_runtime_profile, #function cfi_startproc diff --git a/sysdeps/alpha/dl-machine.h b/sysdeps/alpha/dl-machine.h index ed5389e3c5..7fe2afca93 100644 --- a/sysdeps/alpha/dl-machine.h +++ b/sysdeps/alpha/dl-machine.h @@ -75,9 +75,7 @@ elf_machine_runtime_setup (struct link_map *map, struct r_scope_elem *scope[], int lazy, int profile) { extern char _dl_runtime_resolve_new[] attribute_hidden; - extern char _dl_runtime_profile_new[] attribute_hidden; extern char _dl_runtime_resolve_old[] attribute_hidden; - extern char _dl_runtime_profile_old[] attribute_hidden; struct pltgot { char *resolve; @@ -109,6 +107,9 @@ elf_machine_runtime_setup (struct link_map *map, struct r_scope_elem *scope[], else resolve = _dl_runtime_resolve_old; +#ifdef SHARED + extern char _dl_runtime_profile_new[] attribute_hidden; + extern char _dl_runtime_profile_old[] attribute_hidden; if (__builtin_expect (profile, 0)) { if (secureplt) @@ -123,6 +124,7 @@ elf_machine_runtime_setup (struct link_map *map, struct r_scope_elem *scope[], GL(dl_profile_map) = map; } } +#endif pg->resolve = resolve; pg->link = map; diff --git a/sysdeps/alpha/dl-trampoline.S b/sysdeps/alpha/dl-trampoline.S index f8c3d33906..c37600d9d0 100644 --- a/sysdeps/alpha/dl-trampoline.S +++ b/sysdeps/alpha/dl-trampoline.S @@ -89,6 +89,8 @@ _dl_runtime_resolve_new: .globl _dl_runtime_profile_new .type _dl_runtime_profile_new, @function + +#ifdef SHARED #undef FRAMESIZE #define FRAMESIZE 20*8 @@ -207,11 +209,13 @@ _dl_runtime_profile_new: cfi_endproc .size _dl_runtime_profile_new, .-_dl_runtime_profile_new +#endif /* SHARED */ .align 4 .globl _dl_runtime_resolve_old .ent _dl_runtime_resolve_old + #undef FRAMESIZE #define FRAMESIZE 44*8 @@ -340,6 +344,7 @@ _dl_runtime_resolve_old: .usepv _dl_runtime_profile_old, no .type _dl_runtime_profile_old, @function +#ifdef SHARED /* We save the registers in a different order than desired by .mask/.fmask, so we have to use explicit cfi directives. */ cfi_startproc @@ -538,3 +543,4 @@ _dl_runtime_profile_old: cfi_endproc .size _dl_runtime_profile_old, .-_dl_runtime_profile_old +#endif /* SHARED */ diff --git a/sysdeps/arm/dl-machine.h b/sysdeps/arm/dl-machine.h index d720c02c96..a68679e653 100644 --- a/sysdeps/arm/dl-machine.h +++ b/sysdeps/arm/dl-machine.h @@ -65,7 +65,6 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], { Elf32_Addr *got; extern void _dl_runtime_resolve (Elf32_Word); - extern void _dl_runtime_profile (Elf32_Word); if (l->l_info[DT_JMPREL] && lazy) { @@ -88,6 +87,8 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], to intercept the calls to collect information. In this case we don't store the address in the GOT so that all future calls also end in this function. */ +#ifdef SHARED + extern void _dl_runtime_profile (Elf32_Word); if (profile) { got[2] = (Elf32_Addr) &_dl_runtime_profile; @@ -99,6 +100,7 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], GL(dl_profile_map) = l; } else +#endif /* This function will get called to fix up the GOT entry indicated by the offset on the stack, and then jump to the resolved address. */ got[2] = (Elf32_Addr) &_dl_runtime_resolve; diff --git a/sysdeps/arm/dl-trampoline.S b/sysdeps/arm/dl-trampoline.S index 23c2476917..2df5b7ee36 100644 --- a/sysdeps/arm/dl-trampoline.S +++ b/sysdeps/arm/dl-trampoline.S @@ -70,7 +70,7 @@ _dl_runtime_resolve: cfi_endproc .size _dl_runtime_resolve, .-_dl_runtime_resolve -#ifndef PROF +#if !defined PROF && defined SHARED .globl _dl_runtime_profile .type _dl_runtime_profile, #function CFI_SECTIONS diff --git a/sysdeps/hppa/dl-machine.h b/sysdeps/hppa/dl-machine.h index 4e6e70b3c9..993593de5d 100644 --- a/sysdeps/hppa/dl-machine.h +++ b/sysdeps/hppa/dl-machine.h @@ -195,7 +195,6 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], end_jmprel = jmprel + l->l_info[DT_PLTRELSZ]->d_un.d_val; extern void _dl_runtime_resolve (void); - extern void _dl_runtime_profile (void); /* Linking lazily */ if (lazy) @@ -235,22 +234,9 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], got[1] = (Elf32_Addr) l; /* This function will be called to perform the relocation. */ - if (__builtin_expect (!profile, 1)) - { - /* If a static application called us, then _dl_runtime_resolve is not - a function descriptor, but the *real* address of the function... */ - if((unsigned long) &_dl_runtime_resolve & 3) - { - got[-2] = (Elf32_Addr) ((struct fdesc *) - ((unsigned long) &_dl_runtime_resolve & ~3))->ip; - } - else - { - /* Static executable! */ - got[-2] = (Elf32_Addr) &_dl_runtime_resolve; - } - } - else +#ifdef SHARED + extern void _dl_runtime_profile (void); + if (__glibc_unlikely (profile)) { if (GLRO(dl_profile) != NULL && _dl_name_match_p (GLRO(dl_profile), l)) @@ -272,6 +258,22 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], got[-2] = (Elf32_Addr) &_dl_runtime_profile; } } + else +#endif + { + /* If a static application called us, then _dl_runtime_resolve is not + a function descriptor, but the *real* address of the function... */ + if((unsigned long) &_dl_runtime_resolve & 3) + { + got[-2] = (Elf32_Addr) ((struct fdesc *) + ((unsigned long) &_dl_runtime_resolve & ~3))->ip; + } + else + { + /* Static executable! */ + got[-2] = (Elf32_Addr) &_dl_runtime_resolve; + } + } /* Plunk in the gp of this function descriptor so we can make the call to _dl_runtime_xxxxxx */ got[-1] = ltp; diff --git a/sysdeps/hppa/dl-trampoline.S b/sysdeps/hppa/dl-trampoline.S index 689c6e1a40..9e904df3d2 100644 --- a/sysdeps/hppa/dl-trampoline.S +++ b/sysdeps/hppa/dl-trampoline.S @@ -156,6 +156,7 @@ _dl_runtime_resolve: cfi_endproc .size _dl_runtime_resolve, . - _dl_runtime_resolve +#ifdef SHARED .text .global _dl_runtime_profile .type _dl_runtime_profile,@function @@ -359,3 +360,4 @@ L(cont): .PROCEND cfi_endproc .size _dl_runtime_profile, . - _dl_runtime_profile +#endif diff --git a/sysdeps/i386/dl-machine.h b/sysdeps/i386/dl-machine.h index 18749f2ec2..07469e99b0 100644 --- a/sysdeps/i386/dl-machine.h +++ b/sysdeps/i386/dl-machine.h @@ -92,6 +92,7 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], to intercept the calls to collect information. In this case we don't store the address in the GOT so that all future calls also end in this function. */ +#ifdef SHARED if (__glibc_unlikely (profile)) { got[2] = (shstk_enabled @@ -105,6 +106,7 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], GL(dl_profile_map) = l; } else +#endif /* This function will get called to fix up the GOT entry indicated by the offset on the stack, and then jump to the resolved address. */ got[2] = (shstk_enabled diff --git a/sysdeps/i386/dl-trampoline.S b/sysdeps/i386/dl-trampoline.S index 2d55f373b4..3604aabe87 100644 --- a/sysdeps/i386/dl-trampoline.S +++ b/sysdeps/i386/dl-trampoline.S @@ -70,7 +70,7 @@ _dl_runtime_resolve_shstk: cfi_endproc .size _dl_runtime_resolve_shstk, .-_dl_runtime_resolve_shstk -#ifndef PROF +#if !defined PROF && defined SHARED # The SHSTK compatible version. .globl _dl_runtime_profile_shstk .type _dl_runtime_profile_shstk, @function diff --git a/sysdeps/ia64/dl-machine.h b/sysdeps/ia64/dl-machine.h index e1da3dadcb..3ef6b0ef4b 100644 --- a/sysdeps/ia64/dl-machine.h +++ b/sysdeps/ia64/dl-machine.h @@ -121,9 +121,8 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], reserve[0] = (Elf64_Addr) l; /* This function will be called to perform the relocation. */ - if (!profile) - doit = (Elf64_Addr) ELF_PTR_TO_FDESC (&_dl_runtime_resolve)->ip; - else +#ifdef SHARED + if (__glibc_unlikely (profile)) { if (GLRO(dl_profile) != NULL && _dl_name_match_p (GLRO(dl_profile), l)) @@ -134,6 +133,11 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], } doit = (Elf64_Addr) ELF_PTR_TO_FDESC (&_dl_runtime_profile)->ip; } + else +#endif + { + doit = (Elf64_Addr) ELF_PTR_TO_FDESC (&_dl_runtime_resolve)->ip; + } reserve[1] = doit; reserve[2] = gp; diff --git a/sysdeps/ia64/dl-trampoline.S b/sysdeps/ia64/dl-trampoline.S index 54b33c8c02..10d8432c8f 100644 --- a/sysdeps/ia64/dl-trampoline.S +++ b/sysdeps/ia64/dl-trampoline.S @@ -188,7 +188,7 @@ END(_dl_runtime_resolve) #define PLTENTER_FRAME_SIZE (4*8 + 8*8 + 8*16 + 2*8 + 16) #define PLTEXIT_FRAME_SIZE (PLTENTER_FRAME_SIZE + 4*8 + 8*16) -#ifndef PROF +#if !defined PROF && defined SHARED ENTRY(_dl_runtime_profile) { .mii .prologue diff --git a/sysdeps/loongarch/dl-machine.h b/sysdeps/loongarch/dl-machine.h index 57913cefaa..0d17fd21e3 100644 --- a/sysdeps/loongarch/dl-machine.h +++ b/sysdeps/loongarch/dl-machine.h @@ -287,15 +287,16 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], to intercept the calls to collect information. In this case we don't store the address in the GOT so that all future calls also end in this function. */ +#ifdef SHARED if (profile != 0) { -#if !defined __loongarch_soft_float +# if !defined __loongarch_soft_float if (SUPPORT_LASX) gotplt[0] = (ElfW(Addr)) &_dl_runtime_profile_lasx; else if (SUPPORT_LSX) gotplt[0] = (ElfW(Addr)) &_dl_runtime_profile_lsx; else -#endif +# endif gotplt[0] = (ElfW(Addr)) &_dl_runtime_profile; if (GLRO(dl_profile) != NULL @@ -305,6 +306,7 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], GL(dl_profile_map) = l; } else +#endif { /* This function will get called to fix up the GOT entry indicated by the offset on the stack, and then jump to diff --git a/sysdeps/loongarch/dl-trampoline.h b/sysdeps/loongarch/dl-trampoline.h index cb4a287c65..bb6ff47782 100644 --- a/sysdeps/loongarch/dl-trampoline.h +++ b/sysdeps/loongarch/dl-trampoline.h @@ -126,6 +126,7 @@ ENTRY (_dl_runtime_resolve) jirl zero, t1, 0 END (_dl_runtime_resolve) +#ifdef SHARED #include "dl-link.h" ENTRY (_dl_runtime_profile) @@ -367,3 +368,4 @@ ENTRY (_dl_runtime_profile) jirl zero, ra, 0 END (_dl_runtime_profile) +#endif /* SHARED */ diff --git a/sysdeps/m68k/dl-machine.h b/sysdeps/m68k/dl-machine.h index 5ee586b27b..8d7e733e2a 100644 --- a/sysdeps/m68k/dl-machine.h +++ b/sysdeps/m68k/dl-machine.h @@ -75,7 +75,6 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], { Elf32_Addr *got; extern void _dl_runtime_resolve (Elf32_Word); - extern void _dl_runtime_profile (Elf32_Word); if (l->l_info[DT_JMPREL] && lazy) { @@ -93,6 +92,8 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], to intercept the calls to collect information. In this case we don't store the address in the GOT so that all future calls also end in this function. */ +#ifdef SHARED + extern void _dl_runtime_profile (Elf32_Word); if (profile) { got[2] = (Elf32_Addr) &_dl_runtime_profile; @@ -106,6 +107,7 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], } } else +#endif /* This function will get called to fix up the GOT entry indicated by the offset on the stack, and then jump to the resolved address. */ got[2] = (Elf32_Addr) &_dl_runtime_resolve; diff --git a/sysdeps/m68k/dl-trampoline.S b/sysdeps/m68k/dl-trampoline.S index dba3741400..f1b4943868 100644 --- a/sysdeps/m68k/dl-trampoline.S +++ b/sysdeps/m68k/dl-trampoline.S @@ -60,6 +60,7 @@ _dl_runtime_resolve: cfi_endproc .size _dl_runtime_resolve, . - _dl_runtime_resolve +#ifdef SHARED .text .globl _dl_runtime_profile .type _dl_runtime_profile, @function @@ -220,3 +221,4 @@ _dl_runtime_profile: rts cfi_endproc .size _dl_runtime_profile, . - _dl_runtime_profile +#endif /* SHARED */ diff --git a/sysdeps/powerpc/powerpc32/dl-machine.c b/sysdeps/powerpc/powerpc32/dl-machine.c index ef84911ede..e6b603de94 100644 --- a/sysdeps/powerpc/powerpc32/dl-machine.c +++ b/sysdeps/powerpc/powerpc32/dl-machine.c @@ -226,7 +226,7 @@ __elf_machine_runtime_setup (struct link_map *map, int lazy, int profile) Elf32_Word dlrr; Elf32_Word offset; -#ifndef PROF +#if !defined PROF && defined SHARED dlrr = (Elf32_Word) (profile ? _dl_prof_resolve : _dl_runtime_resolve); diff --git a/sysdeps/powerpc/powerpc32/dl-machine.h b/sysdeps/powerpc/powerpc32/dl-machine.h index a4cad7583c..1ff46d5f8a 100644 --- a/sysdeps/powerpc/powerpc32/dl-machine.h +++ b/sysdeps/powerpc/powerpc32/dl-machine.h @@ -188,15 +188,19 @@ elf_machine_runtime_setup (struct link_map *map, struct r_scope_elem *scope[], extern void _dl_runtime_resolve (void); extern void _dl_prof_resolve (void); - if (__glibc_likely (!profile)) - dlrr = _dl_runtime_resolve; - else +#ifdef SHARED + if (__glibc_unlikely (profile)) { if (GLRO(dl_profile) != NULL &&_dl_name_match_p (GLRO(dl_profile), map)) GL(dl_profile_map) = map; dlrr = _dl_prof_resolve; } + else +#endif + { + dlrr = _dl_runtime_resolve; + } got = (Elf32_Addr *) map->l_info[DT_PPC(GOT)]->d_un.d_ptr; glink = got[1]; got[1] = (Elf32_Addr) dlrr; diff --git a/sysdeps/powerpc/powerpc32/dl-trampoline.S b/sysdeps/powerpc/powerpc32/dl-trampoline.S index 93b1673ebb..be8de0e2dc 100644 --- a/sysdeps/powerpc/powerpc32/dl-trampoline.S +++ b/sysdeps/powerpc/powerpc32/dl-trampoline.S @@ -70,7 +70,7 @@ _dl_runtime_resolve: cfi_endproc .size _dl_runtime_resolve,.-_dl_runtime_resolve -#ifndef PROF +#if !defined PROF && defined SHARED .align 2 .globl _dl_prof_resolve .type _dl_prof_resolve,@function diff --git a/sysdeps/powerpc/powerpc64/dl-machine.h b/sysdeps/powerpc/powerpc64/dl-machine.h index 449208e86f..601c3cba9d 100644 --- a/sysdeps/powerpc/powerpc64/dl-machine.h +++ b/sysdeps/powerpc/powerpc64/dl-machine.h @@ -362,13 +362,19 @@ elf_machine_runtime_setup (struct link_map *map, struct r_scope_elem *scope[], Elf64_Word offset; Elf64_Addr dlrr; - dlrr = (Elf64_Addr) (profile ? _dl_profile_resolve - : _dl_runtime_resolve); - if (profile && GLRO(dl_profile) != NULL - && _dl_name_match_p (GLRO(dl_profile), map)) - /* This is the object we are looking for. Say that we really - want profiling and the timers are started. */ - GL(dl_profile_map) = map; +#ifdef SHARED + if (__glibc_unlikely (profile)) + { + dlrr = (Elf64_Addr) _dl_profile_resolve; + if (profile && GLRO(dl_profile) != NULL + && _dl_name_match_p (GLRO(dl_profile), map)) + /* This is the object we are looking for. Say that we really + want profiling and the timers are started. */ + GL(dl_profile_map) = map; + } + else +#endif + dlrr = (Elf64_Addr) _dl_runtime_resolve; #if _CALL_ELF != 2 /* We need to stuff the address/TOC of _dl_runtime_resolve diff --git a/sysdeps/powerpc/powerpc64/dl-trampoline.S b/sysdeps/powerpc/powerpc64/dl-trampoline.S index 1d04ec8109..b2fc2bb133 100644 --- a/sysdeps/powerpc/powerpc64/dl-trampoline.S +++ b/sysdeps/powerpc/powerpc64/dl-trampoline.S @@ -195,7 +195,7 @@ END(_dl_runtime_resolve) and r11 contains the link_map (from PLT0+16). The link_map becomes parm1 (r3) and the index (r0) needs to be converted to an offset (index * 24) in parm2 (r4). */ -#ifndef PROF +#if !defined PROF && defined SHARED .hidden _dl_profile_resolve ENTRY (_dl_profile_resolve, 4) /* Spill r30, r31 to preserve the link_map* and reloc_addr, in case we diff --git a/sysdeps/s390/s390-32/dl-machine.h b/sysdeps/s390/s390-32/dl-machine.h index 100a3e05f6..b8bf2796c7 100644 --- a/sysdeps/s390/s390-32/dl-machine.h +++ b/sysdeps/s390/s390-32/dl-machine.h @@ -124,16 +124,17 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], to intercept the calls to collect information. In this case we don't store the address in the GOT so that all future calls also end in this function. */ +#ifdef SHARED if (__glibc_unlikely (profile)) { -#if defined HAVE_S390_VX_ASM_SUPPORT +# if defined HAVE_S390_VX_ASM_SUPPORT if (GLRO(dl_hwcap) & HWCAP_S390_VX) got[2] = (Elf32_Addr) &_dl_runtime_profile_vx; else got[2] = (Elf32_Addr) &_dl_runtime_profile; -#else +# else got[2] = (Elf32_Addr) &_dl_runtime_profile; -#endif +# endif if (GLRO(dl_profile) != NULL && _dl_name_match_p (GLRO(dl_profile), l)) @@ -142,6 +143,7 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], GL(dl_profile_map) = l; } else +#endif { /* This function will get called to fix up the GOT entry indicated by the offset on the stack, and then jump to the resolved address. */ diff --git a/sysdeps/s390/s390-32/dl-trampoline.h b/sysdeps/s390/s390-32/dl-trampoline.h index 78fdca9d53..8093ab08d3 100644 --- a/sysdeps/s390/s390-32/dl-trampoline.h +++ b/sysdeps/s390/s390-32/dl-trampoline.h @@ -148,7 +148,7 @@ _dl_runtime_resolve: #undef F0_OFF #undef F2_OFF -#ifndef PROF +#if !defined PROF && defined SHARED # define SIZEOF_STRUCT_LA_S390_32_REGS 168 # define REGS_OFF -264 # define R2_OFF -264 diff --git a/sysdeps/s390/s390-64/dl-machine.h b/sysdeps/s390/s390-64/dl-machine.h index 9fabb09750..82259dad64 100644 --- a/sysdeps/s390/s390-64/dl-machine.h +++ b/sysdeps/s390/s390-64/dl-machine.h @@ -111,16 +111,17 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], to intercept the calls to collect information. In this case we don't store the address in the GOT so that all future calls also end in this function. */ +#ifdef SHARED if (__glibc_unlikely (profile)) { -#if defined HAVE_S390_VX_ASM_SUPPORT +# if defined HAVE_S390_VX_ASM_SUPPORT if (GLRO(dl_hwcap) & HWCAP_S390_VX) got[2] = (Elf64_Addr) &_dl_runtime_profile_vx; else got[2] = (Elf64_Addr) &_dl_runtime_profile; -#else +# else got[2] = (Elf64_Addr) &_dl_runtime_profile; -#endif +# endif if (GLRO(dl_profile) != NULL && _dl_name_match_p (GLRO(dl_profile), l)) @@ -129,6 +130,7 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], GL(dl_profile_map) = l; } else +#endif { /* This function will get called to fix up the GOT entry indicated by the offset on the stack, and then jump to the resolved address. */ diff --git a/sysdeps/s390/s390-64/dl-trampoline.h b/sysdeps/s390/s390-64/dl-trampoline.h index 3a7cfc5f92..61188119e6 100644 --- a/sysdeps/s390/s390-64/dl-trampoline.h +++ b/sysdeps/s390/s390-64/dl-trampoline.h @@ -150,7 +150,7 @@ _dl_runtime_resolve: #undef F4_OFF #undef F6_OFF -#ifndef PROF +#if !defined PROF && defined SHARED # define SIZEOF_STRUCT_LA_S390_64_REGS 200 # define REGS_OFF -360 # define R2_OFF -360 diff --git a/sysdeps/sh/dl-machine.h b/sysdeps/sh/dl-machine.h index 0e4eac42c3..e0480eae5a 100644 --- a/sysdeps/sh/dl-machine.h +++ b/sysdeps/sh/dl-machine.h @@ -101,6 +101,7 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], to intercept the calls to collect information. In this case we don't store the address in the GOT so that all future calls also end in this function. */ +#ifdef SHARED if (profile) { got[2] = (Elf32_Addr) &_dl_runtime_profile; @@ -110,6 +111,7 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], GL(dl_profile_map) = l; } else +#endif /* This function will get called to fix up the GOT entry indicated by the offset on the stack, and then jump to the resolved address. */ got[2] = (Elf32_Addr) &_dl_runtime_resolve; diff --git a/sysdeps/sh/dl-trampoline.S b/sysdeps/sh/dl-trampoline.S index 5d703341ed..ecaae34db4 100644 --- a/sysdeps/sh/dl-trampoline.S +++ b/sysdeps/sh/dl-trampoline.S @@ -142,6 +142,7 @@ _dl_runtime_resolve: .size _dl_runtime_resolve, .-_dl_runtime_resolve +#ifdef SHARED .globl _dl_runtime_profile .type _dl_runtime_profile,@function cfi_startproc @@ -428,3 +429,4 @@ _dl_runtime_profile: 8: .long _dl_audit_pltexit #endif .size _dl_runtime_profile, .-_dl_runtime_profile +#endif /* SHARED */ diff --git a/sysdeps/sparc/sparc32/dl-machine.h b/sysdeps/sparc/sparc32/dl-machine.h index 9b57ae1a93..b10e541810 100644 --- a/sysdeps/sparc/sparc32/dl-machine.h +++ b/sysdeps/sparc/sparc32/dl-machine.h @@ -116,7 +116,8 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], bits of %g1 with an offset into the .rela.plt section and jump to the beginning of the PLT. */ plt = (Elf32_Addr *) D_PTR (l, l_info[DT_PLTGOT]); - if (__builtin_expect(profile, 0)) +#ifdef SHARED + if (__glibc_unlikely (profile)) { rfunc = (Elf32_Addr) &_dl_runtime_profile; @@ -125,6 +126,7 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], GL(dl_profile_map) = l; } else +#endif { rfunc = (Elf32_Addr) &_dl_runtime_resolve; } diff --git a/sysdeps/sparc/sparc32/dl-trampoline.S b/sysdeps/sparc/sparc32/dl-trampoline.S index 08ff31b474..5e7d860ae4 100644 --- a/sysdeps/sparc/sparc32/dl-trampoline.S +++ b/sysdeps/sparc/sparc32/dl-trampoline.S @@ -47,6 +47,7 @@ _dl_runtime_resolve: .size _dl_runtime_resolve, .-_dl_runtime_resolve +#ifdef SHARED /* For the profiling cases we pass in our stack frame * as the base of the La_sparc32_regs, so it looks * like: @@ -185,3 +186,4 @@ _dl_runtime_profile: cfi_endproc .size _dl_runtime_profile, .-_dl_runtime_profile +#endif diff --git a/sysdeps/sparc/sparc64/dl-machine.h b/sysdeps/sparc/sparc64/dl-machine.h index 2f04ac550e..98469e7604 100644 --- a/sysdeps/sparc/sparc64/dl-machine.h +++ b/sysdeps/sparc/sparc64/dl-machine.h @@ -136,7 +136,8 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], Elf64_Addr res0_addr, res1_addr; unsigned int *plt = (void *) D_PTR (l, l_info[DT_PLTGOT]); - if (__builtin_expect(profile, 0)) +#ifdef SHARED + if (__glibc_unlikely (profile)) { res0_addr = (Elf64_Addr) &_dl_runtime_profile_0; res1_addr = (Elf64_Addr) &_dl_runtime_profile_1; @@ -146,6 +147,7 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], GL(dl_profile_map) = l; } else +#endif { res0_addr = (Elf64_Addr) &_dl_runtime_resolve_0; res1_addr = (Elf64_Addr) &_dl_runtime_resolve_1; diff --git a/sysdeps/sparc/sparc64/dl-trampoline.S b/sysdeps/sparc/sparc64/dl-trampoline.S index 444690a71e..82b42681dd 100644 --- a/sysdeps/sparc/sparc64/dl-trampoline.S +++ b/sysdeps/sparc/sparc64/dl-trampoline.S @@ -92,6 +92,7 @@ _dl_runtime_resolve_1: .size _dl_runtime_resolve_1, .-_dl_runtime_resolve_1 +#ifdef SHARED /* For the profiling cases we pass in our stack frame * as the base of the La_sparc64_regs, so it looks * like: @@ -323,3 +324,4 @@ _dl_runtime_profile_1: cfi_endproc .size _dl_runtime_resolve_1, .-_dl_runtime_resolve_1 +#endif diff --git a/sysdeps/x86_64/dl-machine.h b/sysdeps/x86_64/dl-machine.h index 9ea2a70837..581a2f1a9e 100644 --- a/sysdeps/x86_64/dl-machine.h +++ b/sysdeps/x86_64/dl-machine.h @@ -89,6 +89,7 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], const struct cpu_features* cpu_features = __get_cpu_features (); +#ifdef SHARED /* The got[2] entry contains the address of a function which gets called to get the address of a so far unresolved function and jump to it. The profiling extension of the dynamic linker allows @@ -111,6 +112,7 @@ elf_machine_runtime_setup (struct link_map *l, struct r_scope_elem *scope[], GL(dl_profile_map) = l; } else +#endif { /* This function will get called to fix up the GOT entry indicated by the offset on the stack, and then jump to diff --git a/sysdeps/x86_64/dl-trampoline.S b/sysdeps/x86_64/dl-trampoline.S index a6b9a1826b..3fd30d58fc 100644 --- a/sysdeps/x86_64/dl-trampoline.S +++ b/sysdeps/x86_64/dl-trampoline.S @@ -53,45 +53,49 @@ #define RESTORE_AVX -#define VEC_SIZE 64 -#define VMOVA vmovdqa64 -#define VEC(i) zmm##i -#define _dl_runtime_profile _dl_runtime_profile_avx512 -# define SECTION(p) p##.evex512 -#include "dl-trampoline.h" -#undef _dl_runtime_profile -#undef VEC -#undef VMOVA -#undef VEC_SIZE -#undef SECTION - -#if MINIMUM_X86_ISA_LEVEL <= AVX_X86_ISA_LEVEL -# define VEC_SIZE 32 -# define VMOVA vmovdqa -# define VEC(i) ymm##i -# define SECTION(p) p##.avx -# define _dl_runtime_profile _dl_runtime_profile_avx +#ifdef SHARED +# define VEC_SIZE 64 +# define VMOVA vmovdqa64 +# define VEC(i) zmm##i +# define _dl_runtime_profile _dl_runtime_profile_avx512 +# define SECTION(p) p##.evex512 # include "dl-trampoline.h" # undef _dl_runtime_profile # undef VEC # undef VMOVA # undef VEC_SIZE # undef SECTION -#endif -#if MINIMUM_X86_ISA_LEVEL < AVX_X86_ISA_LEVEL +# if MINIMUM_X86_ISA_LEVEL <= AVX_X86_ISA_LEVEL +# define VEC_SIZE 32 +# define VMOVA vmovdqa +# define VEC(i) ymm##i +# define SECTION(p) p##.avx +# define _dl_runtime_profile _dl_runtime_profile_avx +# include "dl-trampoline.h" +# undef _dl_runtime_profile +# undef VEC +# undef VMOVA +# undef VEC_SIZE +# undef SECTION +# endif + +# if MINIMUM_X86_ISA_LEVEL < AVX_X86_ISA_LEVEL /* movaps/movups is 1-byte shorter. */ -# define VEC_SIZE 16 -# define VMOVA movaps -# define VEC(i) xmm##i -# define _dl_runtime_profile _dl_runtime_profile_sse -# undef RESTORE_AVX -# include "dl-trampoline.h" -# undef _dl_runtime_profile -# undef VEC -# undef VMOVA -# undef VEC_SIZE +# define VEC_SIZE 16 +# define VMOVA movaps +# define VEC(i) xmm##i +# define _dl_runtime_profile _dl_runtime_profile_sse +# undef RESTORE_AVX +# include "dl-trampoline.h" +# undef _dl_runtime_profile +# undef VEC +# undef VMOVA +# undef VEC_SIZE +# endif +#endif /* SHARED */ +#if MINIMUM_X86_ISA_LEVEL < AVX_X86_ISA_LEVEL # define USE_FXSAVE # define STATE_SAVE_ALIGNMENT 16 # define _dl_runtime_resolve _dl_runtime_resolve_fxsave