ad5d1122b8
Even if RISC-V has supported the vDSO feature, the latency of the functions for obtaining the system time is still expensive. It is because these functions still trigger a corresponding system call in the process, which slows down the response time. If we want to remove the system call to reduce the latency, the kernel should have the ability to output the system clock information to userspace. This patch introduces the vDSO common flow to enable the kernel to achieve the above feature and uses "rdtime" instruction to obtain the current time in the user space. Under this condition, the latency cost by the ecall from U-mode to S-mode can be eliminated. After applying this patch, the latency of gettimeofday() measured on the HiFive unleashed board can be reduced by %61. Signed-off-by: Vincent Chen <vincent.chen@sifive.com> Reviewed-by: Atish Patra <atish.patra@wdc.com> Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>
100 lines
2.2 KiB
C
100 lines
2.2 KiB
C
// SPDX-License-Identifier: GPL-2.0-only
|
|
/*
|
|
* Copyright (C) 2004 Benjamin Herrenschmidt, IBM Corp.
|
|
* <benh@kernel.crashing.org>
|
|
* Copyright (C) 2012 ARM Limited
|
|
* Copyright (C) 2015 Regents of the University of California
|
|
*/
|
|
|
|
#include <linux/elf.h>
|
|
#include <linux/mm.h>
|
|
#include <linux/slab.h>
|
|
#include <linux/binfmts.h>
|
|
#include <linux/err.h>
|
|
#include <asm/page.h>
|
|
#ifdef GENERIC_TIME_VSYSCALL
|
|
#include <vdso/datapage.h>
|
|
#else
|
|
#include <asm/vdso.h>
|
|
#endif
|
|
|
|
extern char vdso_start[], vdso_end[];
|
|
|
|
static unsigned int vdso_pages;
|
|
static struct page **vdso_pagelist;
|
|
|
|
/*
|
|
* The vDSO data page.
|
|
*/
|
|
static union {
|
|
struct vdso_data data;
|
|
u8 page[PAGE_SIZE];
|
|
} vdso_data_store __page_aligned_data;
|
|
struct vdso_data *vdso_data = &vdso_data_store.data;
|
|
|
|
static int __init vdso_init(void)
|
|
{
|
|
unsigned int i;
|
|
|
|
vdso_pages = (vdso_end - vdso_start) >> PAGE_SHIFT;
|
|
vdso_pagelist =
|
|
kcalloc(vdso_pages + 1, sizeof(struct page *), GFP_KERNEL);
|
|
if (unlikely(vdso_pagelist == NULL)) {
|
|
pr_err("vdso: pagelist allocation failed\n");
|
|
return -ENOMEM;
|
|
}
|
|
|
|
for (i = 0; i < vdso_pages; i++) {
|
|
struct page *pg;
|
|
|
|
pg = virt_to_page(vdso_start + (i << PAGE_SHIFT));
|
|
vdso_pagelist[i] = pg;
|
|
}
|
|
vdso_pagelist[i] = virt_to_page(vdso_data);
|
|
|
|
return 0;
|
|
}
|
|
arch_initcall(vdso_init);
|
|
|
|
int arch_setup_additional_pages(struct linux_binprm *bprm,
|
|
int uses_interp)
|
|
{
|
|
struct mm_struct *mm = current->mm;
|
|
unsigned long vdso_base, vdso_len;
|
|
int ret;
|
|
|
|
vdso_len = (vdso_pages + 1) << PAGE_SHIFT;
|
|
|
|
down_write(&mm->mmap_sem);
|
|
vdso_base = get_unmapped_area(NULL, 0, vdso_len, 0, 0);
|
|
if (IS_ERR_VALUE(vdso_base)) {
|
|
ret = vdso_base;
|
|
goto end;
|
|
}
|
|
|
|
/*
|
|
* Put vDSO base into mm struct. We need to do this before calling
|
|
* install_special_mapping or the perf counter mmap tracking code
|
|
* will fail to recognise it as a vDSO (since arch_vma_name fails).
|
|
*/
|
|
mm->context.vdso = (void *)vdso_base;
|
|
|
|
ret = install_special_mapping(mm, vdso_base, vdso_len,
|
|
(VM_READ | VM_EXEC | VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC),
|
|
vdso_pagelist);
|
|
|
|
if (unlikely(ret))
|
|
mm->context.vdso = NULL;
|
|
|
|
end:
|
|
up_write(&mm->mmap_sem);
|
|
return ret;
|
|
}
|
|
|
|
const char *arch_vma_name(struct vm_area_struct *vma)
|
|
{
|
|
if (vma->vm_mm && (vma->vm_start == (long)vma->vm_mm->context.vdso))
|
|
return "[vdso]";
|
|
return NULL;
|
|
}
|