From mboxrd@z Thu Jan 1 00:00:00 1970 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 kanga.kvack.org (kanga.kvack.org [205.233.56.17]) by smtp.lore.kernel.org (Postfix) with ESMTP id D7F73C61DA4 for ; Thu, 16 Feb 2023 08:14:57 +0000 (UTC) Received: by kanga.kvack.org (Postfix) id 15B9F6B0071; Thu, 16 Feb 2023 03:14:57 -0500 (EST) Received: by kanga.kvack.org (Postfix, from userid 40) id 10CE46B0072; Thu, 16 Feb 2023 03:14:57 -0500 (EST) X-Delivered-To: int-list-linux-mm@kvack.org Received: by kanga.kvack.org (Postfix, from userid 63042) id F16DE6B0073; Thu, 16 Feb 2023 03:14:56 -0500 (EST) X-Delivered-To: linux-mm@kvack.org Received: from relay.hostedemail.com (smtprelay0010.hostedemail.com [216.40.44.10]) by kanga.kvack.org (Postfix) with ESMTP id DDAA56B0071 for ; Thu, 16 Feb 2023 03:14:56 -0500 (EST) Received: from smtpin12.hostedemail.com (a10.router.float.18 [10.200.18.1]) by unirelay09.hostedemail.com (Postfix) with ESMTP id 9D7AD80DED for ; Thu, 16 Feb 2023 08:14:56 +0000 (UTC) X-FDA: 80472444192.12.11AF4D0 Received: from relay7-d.mail.gandi.net (relay7-d.mail.gandi.net [217.70.183.200]) by imf17.hostedemail.com (Postfix) with ESMTP id D642240011 for ; Thu, 16 Feb 2023 08:14:53 +0000 (UTC) Authentication-Results: imf17.hostedemail.com; dkim=none; dmarc=none; spf=none (imf17.hostedemail.com: domain of alex@ghiti.fr has no SPF policy when checking 217.70.183.200) smtp.mailfrom=alex@ghiti.fr ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=hostedemail.com; s=arc-20220608; t=1676535294; h=from:from:sender:reply-to:subject:subject:date:date: message-id:message-id:to:to:cc:mime-version:mime-version: content-type:content-type: content-transfer-encoding:content-transfer-encoding: in-reply-to:in-reply-to:references:references; bh=/PzfL7UXBWI/vDfNiElLYlXdxRR+MXb/zq8uspE6kIo=; b=K3lweMmbb3fongfYr2a5cvrwqoQsEhRfdnIWTQd7MHkVCqU7xxgFLEkaQ5pLBBl76aAwLm BlLezMfketizKiVUscTjk5LvQ2zEOudL+viNIL0ytMPg4HbPZNRs/3AJ3flfI+YgXiiJ97 oKmY77NbC4mIxyMXCKPGgzIcxP71NDc= ARC-Authentication-Results: i=1; imf17.hostedemail.com; dkim=none; dmarc=none; spf=none (imf17.hostedemail.com: domain of alex@ghiti.fr has no SPF policy when checking 217.70.183.200) smtp.mailfrom=alex@ghiti.fr ARC-Seal: i=1; s=arc-20220608; d=hostedemail.com; t=1676535294; a=rsa-sha256; cv=none; b=qj+l9pqb/+aB+1xudmkGPnssZJXdvT6eTYkfo+kFaZIB45nlX0FO2MsKFfloAtj7JU+iuV KnFz7m6ut19QdMG60c4fIq1xzcDw20mIXr6KijlRnLZSVBB7KmbLSmokQyxGbEbOIWwGSb QemBgn8J2hUwKY8YAfwkkFPIb/pSRXo= Received: (Authenticated sender: alex@ghiti.fr) by mail.gandi.net (Postfix) with ESMTPSA id 8E0F12000A; Thu, 16 Feb 2023 08:14:49 +0000 (UTC) Message-ID: <2e77208f-78e7-b225-7daa-57f08d703b04@ghiti.fr> Date: Thu, 16 Feb 2023 09:14:49 +0100 MIME-Version: 1.0 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101 Thunderbird/102.7.1 Subject: Re: [PATCH 10/7] riscv: Implement the new page table range API Content-Language: en-US To: "Yin, Fengwei" , "linux-riscv@lists.infradead.org" , "linux-mm@kvack.org" , "paul.walmsley@sifive.com" , "aou@eecs.berkeley.edu" , "palmer@dabbelt.com" , "willy@infradead.org" , "linux-arch@vger.kernel.org" References: <20230211033948.891959-1-willy@infradead.org> <20230215000446.1655635-1-willy@infradead.org> <20230215000446.1655635-2-willy@infradead.org> From: Alexandre Ghiti In-Reply-To: Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 8bit X-Rspamd-Queue-Id: D642240011 X-Rspamd-Server: rspam09 X-Rspam-User: X-Stat-Signature: ep3jqod7ubuddfambnprif6ijhqspuf6 X-HE-Tag: 1676535293-765323 X-HE-Meta: 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 a6QRDmrw 1Z45YRb8bdnvl+H3FDuFGP1qrtnm05pMUeaJZ+DG2E73ZiyqAz/fuyPhWZu4buelrjZq9QJiIeHu7Ry7+Bos6EQrCvIJOFbZCq+0P1jPwaXtFYaNKxigWl5RrHNUzu4eIZ7Cm1J/EWwiBQv0zXgu6rn2YGAz1CW2KpqKoAkRQXPD1U/s= X-Bogosity: Ham, tests=bogofilter, spamicity=0.000000, version=1.2.4 Sender: owner-linux-mm@kvack.org Precedence: bulk X-Loop: owner-majordomo@kvack.org List-ID: Hi Yin, On 2/15/23 09:38, Yin, Fengwei wrote: > On Wed, 2023-02-15 at 00:04 +0000, Matthew Wilcox (Oracle) wrote: >> Add set_ptes(), update_mmu_cache_range() and flush_dcache_folio(). >> >> The PG_dcache_clear flag changes from being a per-page bit to being a >> per-folio bit. >> >> Signed-off-by: Matthew Wilcox (Oracle) >> --- >>  arch/riscv/include/asm/cacheflush.h | 19 +++++++++---------- >>  arch/riscv/include/asm/pgtable.h    | 25 ++++++++++++++++++------- >>  arch/riscv/mm/cacheflush.c          | 11 ++--------- >>  3 files changed, 29 insertions(+), 26 deletions(-) >> >> diff --git a/arch/riscv/include/asm/cacheflush.h >> b/arch/riscv/include/asm/cacheflush.h >> index 03e3b95ae6da..10e5e96f09b5 100644 >> --- a/arch/riscv/include/asm/cacheflush.h >> +++ b/arch/riscv/include/asm/cacheflush.h >> @@ -15,20 +15,19 @@ static inline void local_flush_icache_all(void) >> >>  #define PG_dcache_clean PG_arch_1 >> >> -static inline void flush_dcache_page(struct page *page) >> +static inline void flush_dcache_folio(struct folio *folio) >>  { >> -       /* >> -        * HugeTLB pages are always fully mapped and only head page >> will be >> -        * set PG_dcache_clean (see comments in flush_icache_pte()). >> -        */ >> -       if (PageHuge(page)) >> -               page = compound_head(page); >> - >> -       if (test_bit(PG_dcache_clean, &page->flags)) >> -               clear_bit(PG_dcache_clean, &page->flags); >> +       if (test_bit(PG_dcache_clean, &folio->flags)) >> +               clear_bit(PG_dcache_clean, &folio->flags); >>  } >> +#define flush_dcache_folio flush_dcache_folio >>  #define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 1 >> >> +static inline void flush_dcache_page(struct page *page) >> +{ >> +       flush_dcache_folio(page_folio(page)); >> +} >> + >>  /* >>   * RISC-V doesn't have an instruction to flush parts of the >> instruction cache, >>   * so instead we just flush the whole thing. >> diff --git a/arch/riscv/include/asm/pgtable.h >> b/arch/riscv/include/asm/pgtable.h >> index 13222fd5c4b4..03706c833e70 100644 >> --- a/arch/riscv/include/asm/pgtable.h >> +++ b/arch/riscv/include/asm/pgtable.h >> @@ -405,8 +405,8 @@ static inline pte_t pte_modify(pte_t pte, >> pgprot_t newprot) >> >> >>  /* Commit new configuration to MMU hardware */ >> -static inline void update_mmu_cache(struct vm_area_struct *vma, >> -       unsigned long address, pte_t *ptep) >> +static inline void update_mmu_cache_range(struct vm_area_struct >> *vma, >> +               unsigned long address, pte_t *ptep, unsigned int nr) >>  { >>         /* >>          * The kernel assumes that TLBs don't cache invalid entries, >> but >> @@ -415,8 +415,10 @@ static inline void update_mmu_cache(struct >> vm_area_struct *vma, >>          * Relying on flush_tlb_fix_spurious_fault would suffice, but >>          * the extra traps reduce performance.  So, eagerly >> SFENCE.VMA. >>          */ >> -       flush_tlb_page(vma, address); >> +       flush_tlb_range(vma, address, address + nr * PAGE_SIZE); > The flush_tlb_range() of riscv is a little bit strange to me. It gives > __sbi_tlb_flush_range() stride PAGE_SIZE. That means if (end - start) > is larger than stride, it will trigger flush_tlb_all(). > > So this change could trigger flush_tlb_all() while original > flush_tlb_page() just trigger flush_tlb_page(). Maybe I'm missing something but update_mmu_cache behaviour is not changed here, it will always call flush_tlb_page as nr == 1, right? update_mmu_cache_range though will likely call flush_tlb_all: I have to admit that I'm wondering why we don't only flush the range of pages instead of flushing everything, I'll look into that. Alex > > My understanding is flush_tlb_page() should be better because > flush_pmd_tlb_range() has PMD_SIZE as stride to avoid flush_tlb_all(). > I must miss something here. > > Regards > Yin, Fengwei > >>  } >> +#define update_mmu_cache(vma, addr, ptep) \ >> +       update_mmu_cache_range(vma, addr, ptep, 1) >> >>  #define __HAVE_ARCH_UPDATE_MMU_TLB >>  #define update_mmu_tlb update_mmu_cache >> @@ -456,12 +458,21 @@ static inline void __set_pte_at(struct >> mm_struct *mm, >>         set_pte(ptep, pteval); >>  } >> >> -static inline void set_pte_at(struct mm_struct *mm, >> -       unsigned long addr, pte_t *ptep, pte_t pteval) >> +static inline void set_ptes(struct mm_struct *mm, unsigned long >> addr, >> +               pte_t *ptep, pte_t pteval, unsigned int nr) >>  { >> -       page_table_check_ptes_set(mm, addr, ptep, pteval, 1); >> -       __set_pte_at(mm, addr, ptep, pteval); >> +       page_table_check_ptes_set(mm, addr, ptep, pteval, nr); >> + >> +       for (;;) { >> +               __set_pte_at(mm, addr, ptep, pteval); >> +               if (--nr == 0) >> +                       break; >> +               ptep++; >> +               addr += PAGE_SIZE; >> +               pte_val(pteval) += 1 << _PAGE_PFN_SHIFT; >> +       } >>  } >> +#define set_pte_at(mm, addr, ptep, pte) set_ptes(mm, addr, ptep, >> pte, 1) >> >>  static inline void pte_clear(struct mm_struct *mm, >>         unsigned long addr, pte_t *ptep) >> diff --git a/arch/riscv/mm/cacheflush.c b/arch/riscv/mm/cacheflush.c >> index 3cc07ed45aeb..b725c3f6f57f 100644 >> --- a/arch/riscv/mm/cacheflush.c >> +++ b/arch/riscv/mm/cacheflush.c >> @@ -81,16 +81,9 @@ void flush_icache_mm(struct mm_struct *mm, bool >> local) >>  #ifdef CONFIG_MMU >>  void flush_icache_pte(pte_t pte) >>  { >> -       struct page *page = pte_page(pte); >> +       struct folio *folio = page_folio(pte_page(pte)); >> >> -       /* >> -        * HugeTLB pages are always fully mapped, so only setting >> head page's >> -        * PG_dcache_clean flag is enough. >> -        */ >> -       if (PageHuge(page)) >> -               page = compound_head(page); >> - >> -       if (!test_and_set_bit(PG_dcache_clean, &page->flags)) >> +       if (!test_and_set_bit(PG_dcache_clean, &folio->flags)) >>                 flush_icache_all(); >>  } >>  #endif /* CONFIG_MMU */