From: Vineet Gupta <vgupta@kernel.org>
To: linux-snps-arc@lists.infradead.org
Cc: linux-kernel@vger.kernel.org, linux-mm@kvack.org,
Anshuman Khandual <anshuman.khandual@arm.com>,
Mike Rapoport <rppt@kernel.org>, Vineet Gupta <vgupta@kernel.org>
Subject: [PATCH 15/18] ARC: mm: support 3 levels of page tables
Date: Tue, 10 Aug 2021 17:42:55 -0700 [thread overview]
Message-ID: <20210811004258.138075-16-vgupta@kernel.org> (raw)
In-Reply-To: <20210811004258.138075-1-vgupta@kernel.org>
ARCv2 MMU is software walked and Linux implements 2 levels of paging: pgd/pte.
Forthcoming hw will have multiple levels, so this change preps mm code
for same. It is also fun to try multi levels even on soft-walked code to
ensure generic mm code is robust to handle.
overview
________
2 levels {pgd, pte} : pmd is folded but pmd_* macros are valid and operate on pgd
3 levels {pgd, pmd, pte}:
- pud is folded and pud_* macros point to pgd
- pmd_* macros operate on actual pmd
code changes
____________
1. #include <asm-generic/pgtable-nopud.h>
2. Define CONFIG_PGTABLE_LEVELS 3
3a. Define PMD_SHIFT, PMD_SIZE, PMD_MASK, pmd_t
3b. Define pmd_val() which actually deals with pmd
(pmd_offset(), pmd_index() are provided by generic code)
3c. Define pmd_alloc_one() and pmd_free() to allocate pmd
(pmd_populate/pmd_free already exist)
4. Define pud_none(), pud_bad() macros based on generic pud_val() which
internally pertains to pgd now.
4b. define pud_populate() to just setup pgd
Signed-off-by: Vineet Gupta <vgupta@kernel.org>
---
arch/arc/Kconfig | 4 ++
arch/arc/include/asm/page.h | 11 +++++
arch/arc/include/asm/pgalloc.h | 22 ++++++++++
arch/arc/include/asm/pgtable-levels.h | 63 ++++++++++++++++++++++++---
arch/arc/include/asm/processor.h | 2 +-
arch/arc/mm/fault.c | 4 ++
arch/arc/mm/tlb.c | 4 +-
arch/arc/mm/tlbex.S | 9 ++++
8 files changed, 111 insertions(+), 8 deletions(-)
diff --git a/arch/arc/Kconfig b/arch/arc/Kconfig
index 59d5b2a179f6..43cb8aaf57a2 100644
--- a/arch/arc/Kconfig
+++ b/arch/arc/Kconfig
@@ -314,6 +314,10 @@ config ARC_HUGEPAGE_16M
endchoice
+config PGTABLE_LEVELS
+ int "Number of Page table levels"
+ default 2
+
config ARC_COMPACT_IRQ_LEVELS
depends on ISA_ARCOMPACT
bool "Setup Timer IRQ as high Priority"
diff --git a/arch/arc/include/asm/page.h b/arch/arc/include/asm/page.h
index 313e6f543d2d..df3cc154ae4a 100644
--- a/arch/arc/include/asm/page.h
+++ b/arch/arc/include/asm/page.h
@@ -41,6 +41,17 @@ typedef struct {
#define pgd_val(x) ((x).pgd)
#define __pgd(x) ((pgd_t) { (x) })
+#if CONFIG_PGTABLE_LEVELS > 2
+
+typedef struct {
+ unsigned long pmd;
+} pmd_t;
+
+#define pmd_val(x) ((x).pmd)
+#define __pmd(x) ((pmd_t) { (x) })
+
+#endif
+
typedef struct {
#ifdef CONFIG_ARC_HAS_PAE40
unsigned long long pte;
diff --git a/arch/arc/include/asm/pgalloc.h b/arch/arc/include/asm/pgalloc.h
index 0cf73431eb89..01c2d84418ed 100644
--- a/arch/arc/include/asm/pgalloc.h
+++ b/arch/arc/include/asm/pgalloc.h
@@ -86,6 +86,28 @@ static inline void pgd_free(struct mm_struct *mm, pgd_t *pgd)
}
+#if CONFIG_PGTABLE_LEVELS > 2
+
+static inline void pud_populate(struct mm_struct *mm, pud_t *pudp, pmd_t *pmdp)
+{
+ set_pud(pudp, __pud((unsigned long)pmdp));
+}
+
+static inline pmd_t *pmd_alloc_one(struct mm_struct *mm, unsigned long addr)
+{
+ return (pmd_t *)__get_free_page(
+ GFP_KERNEL | __GFP_RETRY_MAYFAIL | __GFP_ZERO);
+}
+
+static inline void pmd_free(struct mm_struct *mm, pmd_t *pmd)
+{
+ free_page((unsigned long)pmd);
+}
+
+#define __pmd_free_tlb(tlb, pmd, addr) pmd_free((tlb)->mm, pmd)
+
+#endif
+
/*
* With software-only page-tables, addr-split for traversal is tweakable and
* that directly governs how big tables would be at each level.
diff --git a/arch/arc/include/asm/pgtable-levels.h b/arch/arc/include/asm/pgtable-levels.h
index 8ece75335bb5..1c2f022d4ad0 100644
--- a/arch/arc/include/asm/pgtable-levels.h
+++ b/arch/arc/include/asm/pgtable-levels.h
@@ -10,6 +10,8 @@
#ifndef _ASM_ARC_PGTABLE_LEVELS_H
#define _ASM_ARC_PGTABLE_LEVELS_H
+#if CONFIG_PGTABLE_LEVELS == 2
+
/*
* 2 level paging setup for software walked MMUv3 (ARC700) and MMUv4 (HS)
*
@@ -37,16 +39,38 @@
#define PGDIR_SHIFT 21
#endif
-#define PGDIR_SIZE BIT(PGDIR_SHIFT) /* vaddr span, not PDG sz */
-#define PGDIR_MASK (~(PGDIR_SIZE - 1))
+#else
+
+/*
+ * A default 3 level paging testing setup in software walked MMU
+ * MMUv4 (8K page): <4> : <7> : <8> : <13>
+ */
+#define PGDIR_SHIFT 28
+#if CONFIG_PGTABLE_LEVELS > 2
+#define PMD_SHIFT 21
+#endif
+
+#endif
+#define PGDIR_SIZE BIT(PGDIR_SHIFT)
+#define PGDIR_MASK (~(PGDIR_SIZE - 1))
#define PTRS_PER_PGD BIT(32 - PGDIR_SHIFT)
-#define PTRS_PER_PTE BIT(PGDIR_SHIFT - PAGE_SHIFT)
+#if CONFIG_PGTABLE_LEVELS > 2
+#define PMD_SIZE BIT(PMD_SHIFT)
+#define PMD_MASK (~(PMD_SIZE - 1))
+#define PTRS_PER_PMD BIT(PGDIR_SHIFT - PMD_SHIFT)
+#endif
+
+#define PTRS_PER_PTE BIT(PMD_SHIFT - PAGE_SHIFT)
#ifndef __ASSEMBLY__
+#if CONFIG_PGTABLE_LEVELS > 2
+#include <asm-generic/pgtable-nopud.h>
+#else
#include <asm-generic/pgtable-nopmd.h>
+#endif
/*
* 1st level paging: pgd
@@ -57,9 +81,35 @@
#define pgd_ERROR(e) \
pr_crit("%s:%d: bad pgd %08lx.\n", __FILE__, __LINE__, pgd_val(e))
+#if CONFIG_PGTABLE_LEVELS > 2
+
+/* In 3 level paging, pud_* macros work on pgd */
+#define pud_none(x) (!pud_val(x))
+#define pud_bad(x) ((pud_val(x) & ~PAGE_MASK))
+#define pud_present(x) (pud_val(x))
+#define pud_clear(xp) do { pud_val(*(xp)) = 0; } while (0)
+#define pud_pgtable(pud) ((pmd_t *)(pud_val(pud) & PAGE_MASK))
+#define pud_page(pud) virt_to_page(pud_pgtable(pud))
+#define set_pud(pudp, pud) (*(pudp) = pud)
+
+/*
+ * 2nd level paging: pmd
+ */
+#define pmd_ERROR(e) \
+ pr_crit("%s:%d: bad pmd %08lx.\n", __FILE__, __LINE__, pmd_val(e))
+
+#define pmd_pfn(pmd) ((pmd_val(pmd) & PMD_MASK) >> PAGE_SHIFT)
+#define pfn_pmd(pfn,prot) __pmd(((pfn) << PAGE_SHIFT) | pgprot_val(prot))
+#define mk_pmd(page,prot) pfn_pmd(page_to_pfn(page),prot)
+
+#endif
+
/*
- * Due to the strange way generic pgtable level folding works, in a 2 level
- * setup, pmd_val() returns pgd, so these pmd_* macros actually work on pgd
+ * Due to the strange way generic pgtable level folding works, the pmd_* macros
+ * - are valid even for 2 levels (which supposedly only has pgd - pte)
+ * - behave differently for 2 vs. 3
+ * In 2 level paging (pgd -> pte), pmd_* macros work on pgd
+ * In 3+ level paging (pgd -> pmd -> pte), pmd_* macros work on pmd
*/
#define pmd_none(x) (!pmd_val(x))
#define pmd_bad(x) ((pmd_val(x) & ~PAGE_MASK))
@@ -70,6 +120,9 @@
#define set_pmd(pmdp, pmd) (*(pmdp) = pmd)
#define pmd_pgtable(pmd) ((pgtable_t) pmd_page_vaddr(pmd))
+/*
+ * 3rd level paging: pte
+ */
#define pte_ERROR(e) \
pr_crit("%s:%d: bad pte %08lx.\n", __FILE__, __LINE__, pte_val(e))
diff --git a/arch/arc/include/asm/processor.h b/arch/arc/include/asm/processor.h
index e4031ecd3c8c..f28afcf5c6d1 100644
--- a/arch/arc/include/asm/processor.h
+++ b/arch/arc/include/asm/processor.h
@@ -93,7 +93,7 @@ extern unsigned int get_wchan(struct task_struct *p);
#define VMALLOC_START (PAGE_OFFSET - (CONFIG_ARC_KVADDR_SIZE << 20))
/* 1 PGDIR_SIZE each for fixmap/pkmap, 2 PGDIR_SIZE gutter (see asm/highmem.h) */
-#define VMALLOC_SIZE ((CONFIG_ARC_KVADDR_SIZE << 20) - PGDIR_SIZE * 4)
+#define VMALLOC_SIZE ((CONFIG_ARC_KVADDR_SIZE << 20) - PMD_SIZE * 4)
#define VMALLOC_END (VMALLOC_START + VMALLOC_SIZE)
diff --git a/arch/arc/mm/fault.c b/arch/arc/mm/fault.c
index 41f154320964..8da2f0ad8c69 100644
--- a/arch/arc/mm/fault.c
+++ b/arch/arc/mm/fault.c
@@ -39,6 +39,8 @@ noinline static int handle_kernel_vaddr_fault(unsigned long address)
if (!pgd_present(*pgd_k))
goto bad_area;
+ set_pgd(pgd, *pgd_k);
+
p4d = p4d_offset(pgd, address);
p4d_k = p4d_offset(pgd_k, address);
if (!p4d_present(*p4d_k))
@@ -49,6 +51,8 @@ noinline static int handle_kernel_vaddr_fault(unsigned long address)
if (!pud_present(*pud_k))
goto bad_area;
+ set_pud(pud, *pud_k);
+
pmd = pmd_offset(pud, address);
pmd_k = pmd_offset(pud_k, address);
if (!pmd_present(*pmd_k))
diff --git a/arch/arc/mm/tlb.c b/arch/arc/mm/tlb.c
index 34f16e0b41e6..77da83569b36 100644
--- a/arch/arc/mm/tlb.c
+++ b/arch/arc/mm/tlb.c
@@ -658,8 +658,8 @@ char *arc_mmu_mumbojumbo(int cpu_id, char *buf, int len)
IS_USED_CFG(CONFIG_TRANSPARENT_HUGEPAGE));
n += scnprintf(buf + n, len - n,
- "MMU [v%x]\t: %dk PAGE, %sJTLB %d (%dx%d), uDTLB %d, uITLB %d%s%s\n",
- p_mmu->ver, p_mmu->pg_sz_k, super_pg,
+ "MMU [v%x]\t: %dk PAGE, %s, swalk %d lvl, JTLB %d (%dx%d), uDTLB %d, uITLB %d%s%s\n",
+ p_mmu->ver, p_mmu->pg_sz_k, super_pg, CONFIG_PGTABLE_LEVELS,
p_mmu->sets * p_mmu->ways, p_mmu->sets, p_mmu->ways,
p_mmu->u_dtlb, p_mmu->u_itlb,
IS_AVAIL2(p_mmu->pae, ", PAE40 ", CONFIG_ARC_HAS_PAE40));
diff --git a/arch/arc/mm/tlbex.S b/arch/arc/mm/tlbex.S
index d08bd09a0afc..5f6bfdfda1be 100644
--- a/arch/arc/mm/tlbex.S
+++ b/arch/arc/mm/tlbex.S
@@ -173,6 +173,15 @@ ex_saved_reg1:
tst r3, r3
bz do_slow_path_pf ; if no Page Table, do page fault
+#if CONFIG_PGTABLE_LEVELS > 2
+ lsr r0, r2, PMD_SHIFT ; Bits for indexing into PMD
+ and r0, r0, (PTRS_PER_PMD - 1)
+ ld.as r1, [r3, r0] ; PMD entry
+ tst r1, r1
+ bz do_slow_path_pf
+ mov r3, r1
+#endif
+
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
and.f 0, r3, _PAGE_HW_SZ ; Is this Huge PMD (thp)
add2.nz r1, r1, r0
--
2.25.1
next prev parent reply other threads:[~2021-08-11 0:43 UTC|newest]
Thread overview: 37+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-08-11 0:42 [PATCH 00/18] ARC mm updates to support 3 or 4 levels of paging Vineet Gupta
2021-08-11 0:42 ` [PATCH 01/18] ARC: mm: simplify mmu scratch register assingment to mmu needs Vineet Gupta
2021-08-11 0:42 ` [PATCH 02/18] ARC: mm: remove tlb paranoid code Vineet Gupta
2021-08-11 0:42 ` [PATCH 03/18] ARC: mm: move mmu/cache externs out to setup.h Vineet Gupta
2021-08-11 5:10 ` Mike Rapoport
2021-08-11 18:46 ` Vineet Gupta
2021-08-11 0:42 ` [PATCH 04/18] ARC: mm: remove pgd_offset_fast Vineet Gupta
2021-08-11 5:12 ` Mike Rapoport
2021-08-11 18:54 ` Vineet Gupta
2021-08-11 0:42 ` [PATCH 05/18] ARC: mm: Fixes to allow STRICT_MM_TYPECHECKS Vineet Gupta
2021-08-11 0:42 ` [PATCH 06/18] ARC: mm: Enable STRICT_MM_TYPECHECKS Vineet Gupta
2021-08-11 12:04 ` Mike Rapoport
2021-08-11 19:01 ` Vineet Gupta
2021-08-11 0:42 ` [PATCH 07/18] ARC: ioremap: use more commonly used PAGE_KERNEL based uncached flag Vineet Gupta
2021-08-11 5:18 ` Mike Rapoport
2021-08-11 18:58 ` Vineet Gupta
2021-08-11 0:42 ` [PATCH 08/18] ARC: mm: pmd_populate* to use the canonical set_pmd (and drop pmd_set) Vineet Gupta
2021-08-11 0:42 ` [PATCH 09/18] ARC: mm: non-functional code cleanup ahead of 3 levels Vineet Gupta
2021-08-11 12:31 ` Mike Rapoport
2021-08-12 1:37 ` Vineet Gupta
2021-08-12 6:18 ` Mike Rapoport
2021-08-12 18:58 ` Vineet Gupta
2021-08-11 0:42 ` [PATCH 10/18] ARC: mm: move MMU specific bits out of ASID allocator Vineet Gupta
2021-08-11 0:42 ` [PATCH 11/18] ARC: mm: move MMU specific bits out of entry code Vineet Gupta
2021-08-11 12:15 ` Mike Rapoport
2021-08-11 19:30 ` Vineet Gupta
2021-08-11 0:42 ` [PATCH 12/18] ARC: mm: disintegrate mmu.h (arcv2 bits out) Vineet Gupta
2021-08-11 0:42 ` [PATCH 13/18] ARC: mm: disintegrate pgtable.h into levels and flags Vineet Gupta
2021-08-11 0:42 ` [PATCH 14/18] ARC: mm: hack to allow 2 level build with 4 level code Vineet Gupta
2021-08-11 0:42 ` Vineet Gupta [this message]
2021-08-11 12:24 ` [PATCH 15/18] ARC: mm: support 3 levels of page tables Mike Rapoport
2021-08-11 22:15 ` Vineet Gupta
2021-08-11 0:42 ` [PATCH 16/18] ARC: mm: support 4 " Vineet Gupta
2021-08-11 12:28 ` Mike Rapoport
2021-08-11 22:17 ` Vineet Gupta
2021-08-11 0:42 ` [PATCH 17/18] ARC: mm: vmalloc sync from kernel to user table to update PMD Vineet Gupta
2021-08-11 0:42 ` [PATCH 18/18] ARC: mm: introduce _PAGE_TABLE to explicitly link pgd,pud,pmd entries Vineet Gupta
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20210811004258.138075-16-vgupta@kernel.org \
--to=vgupta@kernel.org \
--cc=anshuman.khandual@arm.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=linux-snps-arc@lists.infradead.org \
--cc=rppt@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).