From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-pf1-x42c.google.com (mail-pf1-x42c.google.com [IPv6:2607:f8b0:4864:20::42c]) by sourceware.org (Postfix) with ESMTPS id C8BE43858D37 for ; Fri, 21 Jan 2022 01:29:32 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.1 sourceware.org C8BE43858D37 Authentication-Results: sourceware.org; dmarc=none (p=none dis=none) header.from=sifive.com Authentication-Results: sourceware.org; spf=pass smtp.mailfrom=sifive.com Received: by mail-pf1-x42c.google.com with SMTP id x37so3479016pfh.8 for ; Thu, 20 Jan 2022 17:29:32 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=sifive.com; s=google; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=7oPjbepZXroqHna1336tgkSdkEwJQcj0KL+ybgTAnJc=; b=BoJdAJZXySi+xxR94/qa9TDZuCjcRlqfYushEcARrMAQq+EFafuArHjuDLeJemXIsK hdk8ZOeEXjvFh2FKqs8IYuoz3hNylqU6mTI84nD+jS4NQ+of0XObkAePm4wDq/lszDLd 5LcFah3AtdqFHeJyq8liknMBG9b1RqkXAjhThMrL01mEFrWaaV+28P7hx714Xcch2N5Y kuyl1uj7wwu4uKsvpnizMJFQ0OGKdQevMkE3fYYX0jwvOOeMvikp0wo6SJhrX405Qrta AyhhfaYaM+EIxpIN+wc2+vSkw4Ve5UH8GF10BCzQU9pgaRrDXXsF1DeVP9wBbrl2V1DV DIaw== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=7oPjbepZXroqHna1336tgkSdkEwJQcj0KL+ybgTAnJc=; b=ZD/Px2/TlypAktjTO55OTfCKvbvV5vAqNVo3/elSESkdzLC8+Y7qYIkmao6MxjuUIM Wsy6xQE4LsI8zkag/QP3WlJK1/UBM2ci4ihCIG039Io6VQQt6jM7IirJgoIeaKmpZ06N 8DIhtfgWjY8t6e8gZDkj7JNhNswYSqB5gWLhxKt2s/LEnKea8cUzF5pC/PteeSG/FHCn KRm5I2HqXardhuXvHLZ31t+cusIirhYU9uVla7vDiHl3fcPPfpb5LXqubONRp1D4NJeV bT6y04pnR3Rl7PyJSU3a6l1uw2sMlFPOrVHw1Tar7H1MMh/sBqbgvn8s2YuxbI48Wdlk cz8A== X-Gm-Message-State: AOAM531kAjXSSLysznao/C2EPk8IkHoGZ9kezDsxklSo3QWJd3nPycUb Rt0JgkFXPK5NV96IVLfIzqVqGZ5y0VLxx1JZzqWmvZJtcOSLBg== X-Google-Smtp-Source: ABdhPJyMqhdT+Uhf+Mb2PV0gVPNAzWcao1/7vunVBno6hjwohcxNuDkMgFdGCr4w78uuRdEbl6EqLBT/Qh6AoIV7RnQ= X-Received: by 2002:a63:497:: with SMTP id 145mr1189479pge.127.1642728571808; Thu, 20 Jan 2022 17:29:31 -0800 (PST) MIME-Version: 1.0 References: <20220118043159.27521-2-vincent.chen@sifive.com> In-Reply-To: From: Vincent Chen Date: Fri, 21 Jan 2022 09:29:20 +0800 Message-ID: Subject: Re: [PATCH v2 1/2] RISC-V: remove riscv-specific sigcontext.h To: Kito Cheng Cc: Palmer Dabbelt , GNU C Library , Darius Rad , Andrew Waterman , DJ Delorie , Greentime Hu Content-Type: text/plain; charset="UTF-8" X-Spam-Status: No, score=-10.4 required=5.0 tests=BAYES_00, DKIM_SIGNED, DKIM_VALID, DKIM_VALID_AU, DKIM_VALID_EF, GIT_PATCH_0, KAM_SHORT, RCVD_IN_DNSWL_NONE, SPF_HELO_NONE, SPF_PASS, TXREP autolearn=ham autolearn_force=no version=3.4.4 X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on server2.sourceware.org X-BeenThere: libc-alpha@sourceware.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Libc-alpha mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 21 Jan 2022 01:29:34 -0000 On Thu, Jan 20, 2022 at 10:47 AM Kito Cheng wrote: > > Just provide a tricky workaround from GCC side: > > +#ifdef _ASM_RISCV_SIGCONTEXT_H > +#define SIGCONTEXT_PC(SC) (SC)->sc_regs.pc > +#else > +#define SIGCONTEXT_PC(SC) (SC)->gregs[0] > +#endif > > https://gcc.gnu.org/pipermail/gcc-patches/2022-January/588682.html > > > On Thu, Jan 20, 2022 at 10:36 AM Palmer Dabbelt wrote: > > > > On Mon, 17 Jan 2022 20:31:58 PST (-0800), vincent.chen@sifive.com wrote: > > > Remove riscv-specific sigcontext.h so that Glibc can directly use > > > sigcontext.h provided by the kernel to reduce synchronization work > > > when new extension support is introduced. > > > --- > > > .../unix/sysv/linux/riscv/bits/sigcontext.h | 31 ------------------- > > > 1 file changed, 31 deletions(-) > > > delete mode 100644 sysdeps/unix/sysv/linux/riscv/bits/sigcontext.h > > > > > > diff --git a/sysdeps/unix/sysv/linux/riscv/bits/sigcontext.h b/sysdeps/unix/sysv/linux/riscv/bits/sigcontext.h > > > deleted file mode 100644 > > > index b6e15b5f62..0000000000 > > > --- a/sysdeps/unix/sysv/linux/riscv/bits/sigcontext.h > > > +++ /dev/null > > > @@ -1,31 +0,0 @@ > > > -/* Machine-dependent signal context structure for Linux. RISC-V version. > > > - Copyright (C) 1996-2022 Free Software Foundation, Inc. This file is part of the GNU C Library. > > > - > > > - The GNU C Library is free software; you can redistribute it and/or > > > - modify it under the terms of the GNU Lesser General Public > > > - License as published by the Free Software Foundation; either > > > - version 2.1 of the License, or (at your option) any later version. > > > - > > > - The GNU C Library is distributed in the hope that it will be useful, > > > - but WITHOUT ANY WARRANTY; without even the implied warranty of > > > - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > > > - Lesser General Public License for more details. > > > - > > > - You should have received a copy of the GNU Lesser General Public > > > - License along with the GNU C Library. If not, see > > > - . */ > > > - > > > -#ifndef _BITS_SIGCONTEXT_H > > > -#define _BITS_SIGCONTEXT_H 1 > > > - > > > -#if !defined _SIGNAL_H && !defined _SYS_UCONTEXT_H > > > -# error "Never use directly; include instead." > > > -#endif > > > - > > > -struct sigcontext { > > > - /* gregs[0] holds the program counter. */ > > > - unsigned long int gregs[32]; > > > - unsigned long long int fpregs[66] __attribute__ ((__aligned__ (16))); > > > -}; > > > - > > > -#endif > > > > This will definitely break API compatibility (the fields have > > different names) but should be fine for ABI compatibility. IIUC that's > > within the rules, but I'm not sure it's a desirable outcome. Probably > > would have been better to get this right the first time around, but I'm > > not sure it's worth fixing -- essentially we're making a bunch of users > > change things so we don't have to. That said, it's pretty ugly to have > > two different definitions of a structure with the same name and layout. > > > > Maybe there's some sort of macro-related trick we can use? ie, provide > > the current definition unless users opt into the Linux one (presumably > > so they can talk about the V state). There's going to be some hoops to > > jump through there to maintain ABI compatibility either way, so it's > > possible we could tie these two together? I can understand what you are worried about. Therefore, I also tried to build multiple Yocto images, such as core-image-full-cmdline, core-image-x11, core-image-sato, and core-image-base, to evaluate the impact. After applying Kito's solution to GCC, I didn't get any build errors. According to the results, maybe we can have a quick conclusion about the impact of this patch. The new version Glibc is not compatible with the old version GCC (The old Glibc is still compatible with the new version GCC due to Kito's patch) Some public programs that use struct sigcontext are not covered by this test. (If someone can tell me which program I'm missing, I'm willing to fix it) Some in-house programs use struct sigcontext_t to access signal stack. IMO, the impact seems not severe at this moment. Directly using the kernel's sigcontext can get us away from the pain if we need to add new registers to the signal context for a new extension or vendor customized extension. In addition, I was keeping to find a suitable solution to solve it as you described. But if I still cannot come up with a solution, do you mind that bits/sigcontext.h has a duplicate data struct related to V extension? Thank you