From patchwork Wed Apr 9 02:45:16 2025 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Samuel Holland X-Patchwork-Id: 14044020 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from bombadil.infradead.org (bombadil.infradead.org [198.137.202.133]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.lore.kernel.org (Postfix) with ESMTPS id E1638C36010 for ; Wed, 9 Apr 2025 02:45:38 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/relaxed; d=lists.infradead.org; s=bombadil.20210309; h=Sender: Content-Transfer-Encoding:Content-Type:List-Subscribe:List-Help:List-Post: List-Archive:List-Unsubscribe:List-Id:MIME-Version:Message-ID:Date:Subject:Cc :To:From:Reply-To:Content-ID:Content-Description:Resent-Date:Resent-From: Resent-Sender:Resent-To:Resent-Cc:Resent-Message-ID:In-Reply-To:References: List-Owner; bh=tynzbZtidgTuBjegEYhC6mPDw+Vr6Qq/kRZHlFqP3I8=; b=xecZEih8AgGqrQ AJb+fFUYw0nIIFQjDzSrfhSD+g+XT7Nc1GdY+kxe85wfMqTcQTtemUvX+nGX4obWCQXmZGufRgMc3 T+ggB4SZM7CfZHdH/ZR5xNotuK8wW0ZdrObKQXr3FqTDY4LkxcG1MMZ6IbkF6p3F8cMWr6tiE+Pnp SwvXxCr1CI3+L+VZRNdD1ZfTFa4TQhfRoDZM8oOkGNdg/5QuOJdrRiSUooExMMx76x3tjoo/rOUfp M7fk0BpP8NPQEr5k51Oxg7RL7gqa/3ujb2EppYmh7FpHK60/i/lc4OGctvP+BOnGfhctn4SjOZswH +6cMoK66926o5CEfK4hQ==; Received: from localhost ([::1] helo=bombadil.infradead.org) by bombadil.infradead.org with esmtp (Exim 4.98.2 #2 (Red Hat Linux)) id 1u2LRT-00000005xoZ-1chZ; Wed, 09 Apr 2025 02:45:31 +0000 Received: from mail-pg1-x52c.google.com ([2607:f8b0:4864:20::52c]) by bombadil.infradead.org with esmtps (Exim 4.98.2 #2 (Red Hat Linux)) id 1u2LRQ-00000005xo3-2Wgg for linux-riscv@lists.infradead.org; Wed, 09 Apr 2025 02:45:29 +0000 Received: by mail-pg1-x52c.google.com with SMTP id 41be03b00d2f7-af5cdf4a2f8so4608294a12.3 for ; Tue, 08 Apr 2025 19:45:27 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=sifive.com; s=google; t=1744166726; x=1744771526; darn=lists.infradead.org; h=content-transfer-encoding:mime-version:message-id:date:subject:cc :to:from:from:to:cc:subject:date:message-id:reply-to; bh=9nsdNo0PEJwcz9AI9dBxmI9czrIQQTsTouDYT/msKPQ=; b=eTop81AkTV/dBOJsCOi4qIO4jY54Mg1TkrYtHVH0G4uOdQCRUsJ0dgb9tENObGfegJ DkUTa9qWWvQLUs09xAlr+w3HWUiW7j5Yz9nDfP/47bZIZtIXcARvvVF1zNca7PhL0/Zu lgaVgQ7FMCcosdDxtfZTLeFABWi71AIsyoULU5RRGBJHgBtEZLbJQgfYAZn2Kn7G6CkE 4lbpnZtM5FisIKaWRmXAYuxmcEjAIdF41i3dEZ6N9nZWV8ky8ctXx5ArJJ7dqyixsl+S 1wlwqpHmWLzGlFz3UYHsEpayh/7KmvoCnncHzL+XVCPJHzuyNSuJoFBQCTIrYOGfufvB 1gKA== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20230601; t=1744166726; x=1744771526; h=content-transfer-encoding:mime-version:message-id:date:subject:cc :to:from:x-gm-message-state:from:to:cc:subject:date:message-id :reply-to; bh=9nsdNo0PEJwcz9AI9dBxmI9czrIQQTsTouDYT/msKPQ=; b=hmwt65mFk2u+Uj9kVnnhW56et5aq2h6feaVZ3nERL02Rz3MyNyPAEmYF6MGBjuQO7r mOTwHUrBaANimaHk5WLX1nvqnyAFvGHFI42dC7BwmuAtJANxXvc8NHeod8WYNNK2Ucf9 h7oFIJdNyVQ/3GCjqjKM32qx+4LonCqytUY5Y1QmGmV+yAfnCXr5SRqtQfiBg9wGG1uQ e9nJHp58dNs7KL5VNP7ssuaSdVLPvy7ZVTugEcJPhrUHf1jiQmfY6zoZt4PLuVkG0TBL luS5e4vACr7QBAC2F1o5kCUDWf/IkoMrUc8w526QHgjFeZqDjVCDU+IdKBblGQ95y1/W uJuQ== X-Forwarded-Encrypted: i=1; AJvYcCV5xQpytXK9pB1cVyq6ujEyMMZZ+3zuxr+b+o1DI3/R58yTpmeuH2poWewv5veQND8sXjdqJQHu6RnSIg==@lists.infradead.org X-Gm-Message-State: AOJu0Yw5BbZnwpyzn77qs6qylU/3f6wERmQ7RQPmhz1qZMrvIEvfXWk4 smUqq5QziM+39WWxiywSnT6vxJGzmxccsH29REXkIvakkCQ089WEH1h/HoVhJjohTHc2NkE5bxk 1 X-Gm-Gg: ASbGncsTDSFaDdlIugdm6+kjzFAWWAd5or/tmLdDwkWaxean0MKfINIkQkUWNzym7vC Ke30vpB1tdqonNqUELPhz5vGJ1M/YojKIGuaEqsBSz2e7cwDz1r9zUG5OTBt5IapiZp4chTckQI B9SyfTtX7ZC8cn4ajzRPAbzihJfcJtKTYYl68jp2c6U6MtJlNxZoHaBkLYasD2ix8PCtGk7tdI5 LqMCnLBNQsHgnz3QhaYMNcxetcJT3Kevy++1curr4U6y77bdc9XnEh1dUIsnxnUPwyTLDF/OLuH cEMU7SqY/iLrwz4Pe2pXrjblJ5hRBfeAWxSnDsKL+Sm3ypU9KIWXlzPq/AiLSAFSZT0= X-Google-Smtp-Source: AGHT+IFLGZtI/XCiZ2AL0Lvh6GdIpSXkAOPiu5RbDBnTMtlfSsuznoWsv8v3/hpYMe1aXDhC1J+4Bg== X-Received: by 2002:a17:90b:2e8e:b0:2fe:9e6c:add9 with SMTP id 98e67ed59e1d1-306dbbc41afmr2426502a91.18.1744166726572; Tue, 08 Apr 2025 19:45:26 -0700 (PDT) Received: from sw06.internal.sifive.com ([4.53.31.132]) by smtp.gmail.com with ESMTPSA id 98e67ed59e1d1-306df401ac8sm175285a91.45.2025.04.08.19.45.25 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Tue, 08 Apr 2025 19:45:26 -0700 (PDT) From: Samuel Holland To: Palmer Dabbelt , Alexandre Ghiti , linux-riscv@lists.infradead.org Cc: Samuel Holland , Albert Ou , Paul Walmsley , linux-kernel@vger.kernel.org Subject: [PATCH] riscv: module: Optimize PLT/GOT entry counting Date: Tue, 8 Apr 2025 19:45:16 -0700 Message-ID: <20250409024519.454828-1-samuel.holland@sifive.com> X-Mailer: git-send-email 2.47.0 MIME-Version: 1.0 X-CRM114-Version: 20100106-BlameMichelson ( TRE 0.8.0 (BSD) ) MR-646709E3 X-CRM114-CacheID: sfid-20250408_194528_653633_3300CC6B X-CRM114-Status: GOOD ( 18.84 ) X-BeenThere: linux-riscv@lists.infradead.org X-Mailman-Version: 2.1.34 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: "linux-riscv" Errors-To: linux-riscv-bounces+linux-riscv=archiver.kernel.org@lists.infradead.org perf reports that 99.63% of the cycles from `modprobe amdgpu` are spent inside module_frob_arch_sections(). This is because amdgpu.ko contains about 300000 relocations in its .rela.text section, and the algorithm in count_max_entries() takes quadratic time. Apply two optimizations from the arm64 code, which together reduce the total execution time by 99.57%. First, sort the relocations so duplicate entries are adjacent. Second, reduce the number of relocations that must be sorted by filtering to only relocations that need PLT/GOT entries, as done in commit d4e0340919fb ("arm64/module: Optimize module load time by optimizing PLT counting"). Unlike the arm64 code, here the filtering and sorting is done in a scratch buffer, because the HI20 relocation search optimization in apply_relocate_add() depends on the original order of the relocations. Signed-off-by: Samuel Holland Reviewed-by: Andrew Jones --- arch/riscv/kernel/module-sections.c | 66 +++++++++++++++++++++++++---- 1 file changed, 57 insertions(+), 9 deletions(-) diff --git a/arch/riscv/kernel/module-sections.c b/arch/riscv/kernel/module-sections.c index e264e59e596e..91d4f0fbd0af 100644 --- a/arch/riscv/kernel/module-sections.c +++ b/arch/riscv/kernel/module-sections.c @@ -9,6 +9,7 @@ #include #include #include +#include unsigned long module_emit_got_entry(struct module *mod, unsigned long val) { @@ -55,19 +56,27 @@ unsigned long module_emit_plt_entry(struct module *mod, unsigned long val) return (unsigned long)&plt[i]; } -static int is_rela_equal(const Elf_Rela *x, const Elf_Rela *y) +#define cmp_3way(a, b) ((a) < (b) ? -1 : (a) > (b)) + +static int cmp_rela(const void *a, const void *b) { - return x->r_info == y->r_info && x->r_addend == y->r_addend; + const Elf_Rela *x = a, *y = b; + int i; + + /* sort by type, symbol index and addend */ + i = cmp_3way(x->r_info, y->r_info); + if (i == 0) + i = cmp_3way(x->r_addend, y->r_addend); + return i; } static bool duplicate_rela(const Elf_Rela *rela, int idx) { - int i; - for (i = 0; i < idx; i++) { - if (is_rela_equal(&rela[i], &rela[idx])) - return true; - } - return false; + /* + * Entries are sorted by type, symbol index and addend. That means + * that, if a duplicate entry exists, it must be in the preceding slot. + */ + return idx > 0 && cmp_rela(rela + idx, rela + idx - 1) == 0; } static void count_max_entries(Elf_Rela *relas, int num, @@ -87,11 +96,33 @@ static void count_max_entries(Elf_Rela *relas, int num, } } +static bool rela_needs_plt_got(const Elf_Rela *rela) +{ + unsigned int type = ELF_R_TYPE(rela->r_info); + + return type == R_RISCV_CALL_PLT || type == R_RISCV_GOT_HI20; +} + +/* Copy PLT and GOT relas to the scratch array. */ +static unsigned int partition_plt_got_relas(const Elf_Rela *relas, Elf_Rela *scratch, + unsigned int num_rela) +{ + int j = 0; + + for (int i = 0; i < num_rela; i++) + if (rela_needs_plt_got(&relas[i])) + scratch[j++] = relas[i]; + + return j; +} + int module_frob_arch_sections(Elf_Ehdr *ehdr, Elf_Shdr *sechdrs, char *secstrings, struct module *mod) { unsigned int num_plts = 0; unsigned int num_gots = 0; + Elf_Rela *scratch = NULL; + size_t scratch_size = 0; int i; /* @@ -132,9 +163,26 @@ int module_frob_arch_sections(Elf_Ehdr *ehdr, Elf_Shdr *sechdrs, if (!(dst_sec->sh_flags & SHF_EXECINSTR)) continue; - count_max_entries(relas, num_rela, &num_plts, &num_gots); + /* + * apply_relocate_add() relies on HI20 and LO12 relocation pairs being + * close together, so sort a copy of the section to avoid interfering. + */ + if (sechdrs[i].sh_size > scratch_size) { + scratch_size = sechdrs[i].sh_size; + scratch = kvrealloc(scratch, scratch_size, GFP_KERNEL); + if (!scratch) + return -ENOMEM; + } + + /* sort relocations requiring a PLT or GOT entry so duplicates are adjacent */ + num_rela = partition_plt_got_relas(relas, scratch, num_rela); + sort(scratch, num_rela, sizeof(Elf_Rela), cmp_rela, NULL); + count_max_entries(scratch, num_rela, &num_plts, &num_gots); } + if (scratch) + kvfree(scratch); + mod->arch.plt.shdr->sh_type = SHT_NOBITS; mod->arch.plt.shdr->sh_flags = SHF_EXECINSTR | SHF_ALLOC; mod->arch.plt.shdr->sh_addralign = L1_CACHE_BYTES;