From: David Howells <dhowells@redhat.com>
To: linux-am33-list@redhat.com
Cc: linux-kernel@vger.kernel.org,
Akira Takeuchi <takeuchi.akr@jp.panasonic.com>,
Kiyoshi Owada <owada.kiyoshi@jp.panasonic.com>
Subject: [PATCH 29/43] MN10300: SMP TLB flushing
Date: Tue, 26 Oct 2010 03:55:36 +0100 [thread overview]
Message-ID: <20101026025536.23512.19652.stgit@warthog.procyon.org.uk> (raw)
In-Reply-To: <20101026025301.23512.24525.stgit@warthog.procyon.org.uk>
From: Akira Takeuchi <takeuchi.akr@jp.panasonic.com>
Implement global TLB flushing for MN10300. This will be used by the AM34 which
is SMP capable.
Signed-off-by: Akira Takeuchi <takeuchi.akr@jp.panasonic.com>
Signed-off-by: Kiyoshi Owada <owada.kiyoshi@jp.panasonic.com>
Signed-off-by: David Howells <dhowells@redhat.com>
---
arch/mn10300/include/asm/mmu_context.h | 22 +++
arch/mn10300/include/asm/tlbflush.h | 81 ++++++++----
arch/mn10300/mm/Makefile | 2
arch/mn10300/mm/tlb-smp.c | 214 ++++++++++++++++++++++++++++++++
4 files changed, 290 insertions(+), 29 deletions(-)
create mode 100644 arch/mn10300/mm/tlb-smp.c
diff --git a/arch/mn10300/include/asm/mmu_context.h b/arch/mn10300/include/asm/mmu_context.h
index 5fb3648..c8f6c82 100644
--- a/arch/mn10300/include/asm/mmu_context.h
+++ b/arch/mn10300/include/asm/mmu_context.h
@@ -36,6 +36,22 @@
#define enter_lazy_tlb(mm, tsk) do {} while (0)
+static inline void cpu_ran_vm(int cpu, struct mm_struct *mm)
+{
+#ifdef CONFIG_SMP
+ cpumask_set_cpu(cpu, mm_cpumask(mm));
+#endif
+}
+
+static inline bool cpu_maybe_ran_vm(int cpu, struct mm_struct *mm)
+{
+#ifdef CONFIG_SMP
+ return cpumask_test_and_set_cpu(cpu, mm_cpumask(mm));
+#else
+ return true;
+#endif
+}
+
#ifdef CONFIG_MN10300_TLB_USE_PIDR
extern unsigned long mmu_context_cache[NR_CPUS];
#define mm_context(mm) (mm->context.tlbpid[smp_processor_id()])
@@ -127,7 +143,13 @@ static inline void activate_context(struct mm_struct *mm)
static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
struct task_struct *tsk)
{
+ int cpu = smp_processor_id();
+
if (prev != next) {
+#ifdef CONFIG_SMP
+ per_cpu(cpu_tlbstate, cpu).active_mm = next;
+#endif
+ cpu_ran_vm(cpu, next);
PTBR = (unsigned long) next->pgd;
activate_context(next);
}
diff --git a/arch/mn10300/include/asm/tlbflush.h b/arch/mn10300/include/asm/tlbflush.h
index c3c194d..efddd6e 100644
--- a/arch/mn10300/include/asm/tlbflush.h
+++ b/arch/mn10300/include/asm/tlbflush.h
@@ -11,6 +11,7 @@
#ifndef _ASM_TLBFLUSH_H
#define _ASM_TLBFLUSH_H
+#include <linux/mm.h>
#include <asm/processor.h>
struct tlb_state {
@@ -93,39 +94,61 @@ void local_flush_tlb_page(struct mm_struct *mm, unsigned long addr)
* - flush_tlb_range(mm, start, end) flushes a range of pages
* - flush_tlb_pgtables(mm, start, end) flushes a range of page tables
*/
-#define flush_tlb_all() \
-do { \
- preempt_disable(); \
- local_flush_tlb_all(); \
- preempt_enable(); \
-} while (0)
-
-#define flush_tlb_mm(mm) \
-do { \
- preempt_disable(); \
- local_flush_tlb_all(); \
- preempt_enable(); \
-} while (0)
-
-#define flush_tlb_range(vma, start, end) \
-do { \
- unsigned long __s __attribute__((unused)) = (start); \
- unsigned long __e __attribute__((unused)) = (end); \
- preempt_disable(); \
- local_flush_tlb_all(); \
- preempt_enable(); \
-} while (0)
+#ifdef CONFIG_SMP
+
+#include <asm/smp.h>
+
+extern void flush_tlb_all(void);
+extern void flush_tlb_current_task(void);
+extern void flush_tlb_mm(struct mm_struct *);
+extern void flush_tlb_page(struct vm_area_struct *, unsigned long);
+
+#define flush_tlb() flush_tlb_current_task()
+
+static inline void flush_tlb_range(struct vm_area_struct *vma,
+ unsigned long start, unsigned long end)
+{
+ flush_tlb_mm(vma->vm_mm);
+}
+
+#else /* CONFIG_SMP */
+
+static inline void flush_tlb_all(void)
+{
+ preempt_disable();
+ local_flush_tlb_all();
+ preempt_enable();
+}
+
+static inline void flush_tlb_mm(struct mm_struct *mm)
+{
+ preempt_disable();
+ local_flush_tlb_all();
+ preempt_enable();
+}
+
+static inline void flush_tlb_range(struct vm_area_struct *vma,
+ unsigned long start, unsigned long end)
+{
+ preempt_disable();
+ local_flush_tlb_all();
+ preempt_enable();
+}
#define flush_tlb_page(vma, addr) local_flush_tlb_page((vma)->vm_mm, addr)
#define flush_tlb() flush_tlb_all()
-#define flush_tlb_kernel_range(start, end) \
-do { \
- unsigned long __s __attribute__((unused)) = (start); \
- unsigned long __e __attribute__((unused)) = (end); \
- flush_tlb_all(); \
-} while (0)
+#endif /* CONFIG_SMP */
-#define flush_tlb_pgtables(mm, start, end) do {} while (0)
+static inline void flush_tlb_kernel_range(unsigned long start,
+ unsigned long end)
+{
+ flush_tlb_all();
+}
+
+static inline void flush_tlb_pgtables(struct mm_struct *mm,
+ unsigned long start, unsigned long end)
+{
+}
#endif /* _ASM_TLBFLUSH_H */
diff --git a/arch/mn10300/mm/Makefile b/arch/mn10300/mm/Makefile
index 184745f..203fee2 100644
--- a/arch/mn10300/mm/Makefile
+++ b/arch/mn10300/mm/Makefile
@@ -18,3 +18,5 @@ cacheflush-$(CONFIG_MN10300_CACHE_DISABLED) := cache-disabled.o
obj-y := \
init.o fault.o pgtable.o extable.o tlb-mn10300.o mmu-context.o \
misalignment.o dma-alloc.o $(cacheflush-y)
+
+obj-$(CONFIG_SMP) += tlb-smp.o
diff --git a/arch/mn10300/mm/tlb-smp.c b/arch/mn10300/mm/tlb-smp.c
new file mode 100644
index 0000000..0b6a5ad
--- /dev/null
+++ b/arch/mn10300/mm/tlb-smp.c
@@ -0,0 +1,214 @@
+/* SMP TLB support routines.
+ *
+ * Copyright (C) 2006-2008 Panasonic Corporation
+ * All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * version 2 as published by the Free Software Foundation.
+ *
+ * This program 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 General Public License for more details.
+ */
+#include <linux/interrupt.h>
+#include <linux/spinlock.h>
+#include <linux/init.h>
+#include <linux/jiffies.h>
+#include <linux/cpumask.h>
+#include <linux/err.h>
+#include <linux/kernel.h>
+#include <linux/delay.h>
+#include <linux/sched.h>
+#include <linux/profile.h>
+#include <linux/smp.h>
+#include <asm/tlbflush.h>
+#include <asm/system.h>
+#include <asm/bitops.h>
+#include <asm/processor.h>
+#include <asm/bug.h>
+#include <asm/exceptions.h>
+#include <asm/hardirq.h>
+#include <asm/fpu.h>
+#include <asm/mmu_context.h>
+#include <asm/thread_info.h>
+#include <asm/cpu-regs.h>
+#include <asm/intctl-regs.h>
+
+/*
+ * For flush TLB
+ */
+#define FLUSH_ALL 0xffffffff
+
+static cpumask_t flush_cpumask;
+static struct mm_struct *flush_mm;
+static unsigned long flush_va;
+static DEFINE_SPINLOCK(tlbstate_lock);
+
+DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate) = {
+ &init_mm, 0
+};
+
+static void flush_tlb_others(cpumask_t cpumask, struct mm_struct *mm,
+ unsigned long va);
+static void do_flush_tlb_all(void *info);
+
+/**
+ * smp_flush_tlb - Callback to invalidate the TLB.
+ * @unused: Callback context (ignored).
+ */
+void smp_flush_tlb(void *unused)
+{
+ unsigned long cpu_id;
+
+ cpu_id = get_cpu();
+
+ if (!cpu_isset(cpu_id, flush_cpumask))
+ /* This was a BUG() but until someone can quote me the line
+ * from the intel manual that guarantees an IPI to multiple
+ * CPUs is retried _only_ on the erroring CPUs its staying as a
+ * return
+ *
+ * BUG();
+ */
+ goto out;
+
+ if (flush_va == FLUSH_ALL)
+ local_flush_tlb();
+ else
+ local_flush_tlb_page(flush_mm, flush_va);
+
+ smp_mb__before_clear_bit();
+ cpu_clear(cpu_id, flush_cpumask);
+ smp_mb__after_clear_bit();
+out:
+ put_cpu();
+}
+
+/**
+ * flush_tlb_others - Tell the specified CPUs to invalidate their TLBs
+ * @cpumask: The list of CPUs to target.
+ * @mm: The VM context to flush from (if va!=FLUSH_ALL).
+ * @va: Virtual address to flush or FLUSH_ALL to flush everything.
+ */
+static void flush_tlb_others(cpumask_t cpumask, struct mm_struct *mm,
+ unsigned long va)
+{
+ cpumask_t tmp;
+
+ /* A couple of sanity checks (to be removed):
+ * - mask must not be empty
+ * - current CPU must not be in mask
+ * - we do not send IPIs to as-yet unbooted CPUs.
+ */
+ BUG_ON(!mm);
+ BUG_ON(cpus_empty(cpumask));
+ BUG_ON(cpu_isset(smp_processor_id(), cpumask));
+
+ cpus_and(tmp, cpumask, cpu_online_map);
+ BUG_ON(!cpus_equal(cpumask, tmp));
+
+ /* I'm not happy about this global shared spinlock in the MM hot path,
+ * but we'll see how contended it is.
+ *
+ * Temporarily this turns IRQs off, so that lockups are detected by the
+ * NMI watchdog.
+ */
+ spin_lock(&tlbstate_lock);
+
+ flush_mm = mm;
+ flush_va = va;
+#if NR_CPUS <= BITS_PER_LONG
+ atomic_set_mask(cpumask.bits[0], &flush_cpumask.bits[0]);
+#else
+#error Not supported.
+#endif
+
+ /* FIXME: if NR_CPUS>=3, change send_IPI_mask */
+ smp_call_function(smp_flush_tlb, NULL, 1);
+
+ while (!cpus_empty(flush_cpumask))
+ /* Lockup detection does not belong here */
+ smp_mb();
+
+ flush_mm = NULL;
+ flush_va = 0;
+ spin_unlock(&tlbstate_lock);
+}
+
+/**
+ * flush_tlb_mm - Invalidate TLB of specified VM context
+ * @mm: The VM context to invalidate.
+ */
+void flush_tlb_mm(struct mm_struct *mm)
+{
+ cpumask_t cpu_mask;
+
+ preempt_disable();
+ cpu_mask = mm->cpu_vm_mask;
+ cpu_clear(smp_processor_id(), cpu_mask);
+
+ local_flush_tlb();
+ if (!cpus_empty(cpu_mask))
+ flush_tlb_others(cpu_mask, mm, FLUSH_ALL);
+
+ preempt_enable();
+}
+
+/**
+ * flush_tlb_current_task - Invalidate TLB of current task
+ */
+void flush_tlb_current_task(void)
+{
+ struct mm_struct *mm = current->mm;
+ cpumask_t cpu_mask;
+
+ preempt_disable();
+ cpu_mask = mm->cpu_vm_mask;
+ cpu_clear(smp_processor_id(), cpu_mask);
+
+ local_flush_tlb();
+ if (!cpus_empty(cpu_mask))
+ flush_tlb_others(cpu_mask, mm, FLUSH_ALL);
+
+ preempt_enable();
+}
+
+/**
+ * flush_tlb_page - Invalidate TLB of page
+ * @vma: The VM context to invalidate the page for.
+ * @va: The virtual address of the page to invalidate.
+ */
+void flush_tlb_page(struct vm_area_struct *vma, unsigned long va)
+{
+ struct mm_struct *mm = vma->vm_mm;
+ cpumask_t cpu_mask;
+
+ preempt_disable();
+ cpu_mask = mm->cpu_vm_mask;
+ cpu_clear(smp_processor_id(), cpu_mask);
+
+ local_flush_tlb_page(mm, va);
+ if (!cpus_empty(cpu_mask))
+ flush_tlb_others(cpu_mask, mm, va);
+
+ preempt_enable();
+}
+
+/**
+ * do_flush_tlb_all - Callback to completely invalidate a TLB
+ * @unused: Callback context (ignored).
+ */
+static void do_flush_tlb_all(void *unused)
+{
+ local_flush_tlb_all();
+}
+
+/**
+ * flush_tlb_all - Completely invalidate TLBs on all CPUs
+ */
+void flush_tlb_all(void)
+{
+ on_each_cpu(do_flush_tlb_all, 0, 1);
+}
next prev parent reply other threads:[~2010-10-26 2:55 UTC|newest]
Thread overview: 47+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-10-26 2:53 [PATCH 00/43] MN10300: Add SMP support and AM34 CPU core support David Howells
2010-10-26 2:53 ` [PATCH 01/43] Partially revert patch that encloses asm-offset.h numbers in brackets David Howells
2010-10-26 2:53 ` [PATCH 02/43] MN10300: BUG to BUG_ON changes David Howells
2010-10-26 2:53 ` [PATCH 03/43] bitops: Change the bitmap index from int to unsigned long [mn10300] David Howells
2010-10-26 2:53 ` [PATCH 04/43] mn10300: Use pci_claim_resource David Howells
2010-10-26 2:53 ` [PATCH 05/43] MN10300: Prevent cnt32_to_63() from being preempted in sched_clock() David Howells
2010-10-26 2:53 ` [PATCH 06/43] Typedef SMP call function pointer David Howells
2010-10-26 2:53 ` [PATCH 07/43] MN10300: Don't cast away the volatile in test_bit() David Howells
2010-10-26 2:53 ` [PATCH 08/43] MN10300: Add reads[bwl]() and writes[bwl]() David Howells
2010-10-26 2:53 ` [PATCH 09/43] MN10300: Provide the functions to fully disable maskable interrupts David Howells
2010-10-26 2:53 ` [PATCH 10/43] MN10300: Differentiate AM33_2 and AM33_3 in config David Howells
2010-10-26 2:54 ` [PATCH 11/43] MN10300: Move DMA engine control reg defs to MN103E010 processor directory David Howells
2010-10-26 2:54 ` [PATCH 12/43] MN10300: Don't hard code the cacheline size in register defs David Howells
2010-10-26 2:54 ` [PATCH 13/43] MN10300: Add CPU register bits for AM34 David Howells
2010-10-26 2:54 ` [PATCH 14/43] MN10300: Remove monitor/JTAG functions David Howells
2010-10-26 2:54 ` [PATCH 15/43] MN10300: Cache: Split cache bits out of arch Kconfig David Howells
2010-10-26 2:54 ` [PATCH 16/43] MN10300: Provide a MN10300_CACHE_ENABLED config option David Howells
2010-10-26 2:54 ` [PATCH 17/43] MN10300: Reorder asm/cacheflush.h to put primitives first David Howells
2010-10-26 2:54 ` [PATCH 18/43] MN10300: AM34: The current cacheflush routines operate by controlling tag regs David Howells
2010-10-26 2:54 ` [PATCH 19/43] MN10300: Cacheflush functions should take unsigned long addresses David Howells
2010-10-26 2:54 ` [PATCH 20/43] MN10300: SMP: Differentiate local cache flushing David Howells
2010-10-26 2:54 ` [PATCH 21/43] MN10300: AM34: Add cacheflushing by using the AM34 purge registers David Howells
2010-10-26 2:54 ` [PATCH 22/43] MN10300: Allow some cacheflushes to be avoided if cache snooping is available David Howells
2010-10-26 2:55 ` [PATCH 23/43] MN10300: Cache: Implement SMP global cache flushing David Howells
2010-10-26 2:55 ` [PATCH 24/43] MN10300: Make the boot wrapper able to use writeback caching David Howells
2010-10-26 2:55 ` [PATCH 25/43] MN10300: AM34 erratum requires MMUCTR read and write on exception entry David Howells
2010-10-26 2:55 ` [PATCH 26/43] MN10300: Rename __flush_tlb*() to local_flush_tlb*() David Howells
2010-10-26 2:55 ` [PATCH 27/43] MN10300: Make the use of PIDR to mark TLB entries controllable David Howells
2010-10-26 2:55 ` [PATCH 28/43] MN10300: Use the [ID]PTEL2 registers rather than [ID]PTEL for TLB control David Howells
2010-10-26 2:55 ` David Howells [this message]
2010-10-26 2:55 ` [PATCH 30/43] MN10300: Make the FPU operate in non-lazy mode under SMP David Howells
2010-10-26 2:55 ` [PATCH 31/43] MN10300: Implement atomic ops using atomic ops unit David Howells
2010-10-26 2:55 ` [PATCH 32/43] MN10300: Optimise do_csum() David Howells
2010-10-26 2:55 ` [PATCH 33/43] MN10300: Make various interrupt priority settings configurable David Howells
2010-10-26 2:56 ` [PATCH 34/43] MN10300: Delete idle_timestamp from irq_cpustat_t David Howells
2010-10-26 2:56 ` [PATCH 35/43] MN10300: And Panasonic AM34 subarch and implement SMP David Howells
2010-10-27 13:31 ` [Linux-am33-list] " Suzuki Takashi
2010-10-27 14:33 ` David Howells
2010-10-26 2:56 ` [PATCH 36/43] MN10300: ASB2364: Add support for SMSC911X and SMC911X David Howells
2010-10-26 2:56 ` [PATCH 37/43] MN10300: Map userspace atomic op regs as a vmalloc page David Howells
2010-10-26 2:56 ` [PATCH 38/43] MN10300: Specify an ELF HWCAP flag for MN10300 Atomic Operations Unit support David Howells
2010-10-26 2:56 ` [PATCH 39/43] MN10300: Generic time support David Howells
2010-10-26 2:56 ` [PATCH 40/43] MN10300: Update the ASB2303 defconfig David Howells
2010-10-26 2:56 ` [PATCH 41/43] MN10300: Create a defconfig for the ASB2364 board David Howells
2010-10-26 2:56 ` [PATCH 42/43] MN10300: Change "Matsushita" to "Panasonic" David Howells
2010-10-26 2:56 ` [PATCH 43/43] MN10300: Save frame pointer in thread_info struct rather than global var David Howells
2010-10-26 3:07 ` [PATCH 00/43] MN10300: Add SMP support and AM34 CPU core support David Howells
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=20101026025536.23512.19652.stgit@warthog.procyon.org.uk \
--to=dhowells@redhat.com \
--cc=linux-am33-list@redhat.com \
--cc=linux-kernel@vger.kernel.org \
--cc=owada.kiyoshi@jp.panasonic.com \
--cc=takeuchi.akr@jp.panasonic.com \
/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
all inboxes | Powered by JetHome®