/arch/s390/kernel/vdso.c

  1/*
  2 * vdso setup for s390
  3 *
  4 *  Copyright IBM Corp. 2008
  5 *  Author(s): Martin Schwidefsky (schwidefsky@de.ibm.com)
  6 *
  7 * This program is free software; you can redistribute it and/or modify
  8 * it under the terms of the GNU General Public License (version 2 only)
  9 * as published by the Free Software Foundation.
 10 */
 11
 12#include <linux/module.h>
 13#include <linux/errno.h>
 14#include <linux/sched.h>
 15#include <linux/kernel.h>
 16#include <linux/mm.h>
 17#include <linux/smp.h>
 18#include <linux/stddef.h>
 19#include <linux/unistd.h>
 20#include <linux/slab.h>
 21#include <linux/user.h>
 22#include <linux/elf.h>
 23#include <linux/security.h>
 24#include <linux/bootmem.h>
 25#include <linux/compat.h>
 26#include <asm/asm-offsets.h>
 27#include <asm/pgtable.h>
 28#include <asm/system.h>
 29#include <asm/processor.h>
 30#include <asm/mmu.h>
 31#include <asm/mmu_context.h>
 32#include <asm/sections.h>
 33#include <asm/vdso.h>
 34
 35#if defined(CONFIG_32BIT) || defined(CONFIG_COMPAT)
 36extern char vdso32_start, vdso32_end;
 37static void *vdso32_kbase = &vdso32_start;
 38static unsigned int vdso32_pages;
 39static struct page **vdso32_pagelist;
 40#endif
 41
 42#ifdef CONFIG_64BIT
 43extern char vdso64_start, vdso64_end;
 44static void *vdso64_kbase = &vdso64_start;
 45static unsigned int vdso64_pages;
 46static struct page **vdso64_pagelist;
 47#endif /* CONFIG_64BIT */
 48
 49/*
 50 * Should the kernel map a VDSO page into processes and pass its
 51 * address down to glibc upon exec()?
 52 */
 53unsigned int __read_mostly vdso_enabled = 1;
 54
 55static int __init vdso_setup(char *s)
 56{
 57	unsigned long val;
 58	int rc;
 59
 60	rc = 0;
 61	if (strncmp(s, "on", 3) == 0)
 62		vdso_enabled = 1;
 63	else if (strncmp(s, "off", 4) == 0)
 64		vdso_enabled = 0;
 65	else {
 66		rc = strict_strtoul(s, 0, &val);
 67		vdso_enabled = rc ? 0 : !!val;
 68	}
 69	return !rc;
 70}
 71__setup("vdso=", vdso_setup);
 72
 73/*
 74 * The vdso data page
 75 */
 76static union {
 77	struct vdso_data	data;
 78	u8			page[PAGE_SIZE];
 79} vdso_data_store __page_aligned_data;
 80struct vdso_data *vdso_data = &vdso_data_store.data;
 81
 82/*
 83 * Setup vdso data page.
 84 */
 85static void vdso_init_data(struct vdso_data *vd)
 86{
 87	unsigned int facility_list;
 88
 89	facility_list = stfl();
 90	vd->ectg_available =
 91		user_mode != HOME_SPACE_MODE && (facility_list & 1);
 92}
 93
 94#ifdef CONFIG_64BIT
 95/*
 96 * Setup per cpu vdso data page.
 97 */
 98static void vdso_init_per_cpu_data(int cpu, struct vdso_per_cpu_data *vpcd)
 99{
100}
101
102/*
103 * Allocate/free per cpu vdso data.
104 */
105#define SEGMENT_ORDER	2
106
107int vdso_alloc_per_cpu(int cpu, struct _lowcore *lowcore)
108{
109	unsigned long segment_table, page_table, page_frame;
110	u32 *psal, *aste;
111	int i;
112
113	lowcore->vdso_per_cpu_data = __LC_PASTE;
114
115	if (user_mode == HOME_SPACE_MODE || !vdso_enabled)
116		return 0;
117
118	segment_table = __get_free_pages(GFP_KERNEL, SEGMENT_ORDER);
119	page_table = get_zeroed_page(GFP_KERNEL | GFP_DMA);
120	page_frame = get_zeroed_page(GFP_KERNEL);
121	if (!segment_table || !page_table || !page_frame)
122		goto out;
123
124	clear_table((unsigned long *) segment_table, _SEGMENT_ENTRY_EMPTY,
125		    PAGE_SIZE << SEGMENT_ORDER);
126	clear_table((unsigned long *) page_table, _PAGE_TYPE_EMPTY,
127		    256*sizeof(unsigned long));
128
129	*(unsigned long *) segment_table = _SEGMENT_ENTRY + page_table;
130	*(unsigned long *) page_table = _PAGE_RO + page_frame;
131
132	psal = (u32 *) (page_table + 256*sizeof(unsigned long));
133	aste = psal + 32;
134
135	for (i = 4; i < 32; i += 4)
136		psal[i] = 0x80000000;
137
138	lowcore->paste[4] = (u32)(addr_t) psal;
139	psal[0] = 0x20000000;
140	psal[2] = (u32)(addr_t) aste;
141	*(unsigned long *) (aste + 2) = segment_table +
142		_ASCE_TABLE_LENGTH + _ASCE_USER_BITS + _ASCE_TYPE_SEGMENT;
143	aste[4] = (u32)(addr_t) psal;
144	lowcore->vdso_per_cpu_data = page_frame;
145
146	vdso_init_per_cpu_data(cpu, (struct vdso_per_cpu_data *) page_frame);
147	return 0;
148
149out:
150	free_page(page_frame);
151	free_page(page_table);
152	free_pages(segment_table, SEGMENT_ORDER);
153	return -ENOMEM;
154}
155
156void vdso_free_per_cpu(int cpu, struct _lowcore *lowcore)
157{
158	unsigned long segment_table, page_table, page_frame;
159	u32 *psal, *aste;
160
161	if (user_mode == HOME_SPACE_MODE || !vdso_enabled)
162		return;
163
164	psal = (u32 *)(addr_t) lowcore->paste[4];
165	aste = (u32 *)(addr_t) psal[2];
166	segment_table = *(unsigned long *)(aste + 2) & PAGE_MASK;
167	page_table = *(unsigned long *) segment_table;
168	page_frame = *(unsigned long *) page_table;
169
170	free_page(page_frame);
171	free_page(page_table);
172	free_pages(segment_table, SEGMENT_ORDER);
173}
174
175static void __vdso_init_cr5(void *dummy)
176{
177	unsigned long cr5;
178
179	cr5 = offsetof(struct _lowcore, paste);
180	__ctl_load(cr5, 5, 5);
181}
182
183static void vdso_init_cr5(void)
184{
185	if (user_mode != HOME_SPACE_MODE && vdso_enabled)
186		on_each_cpu(__vdso_init_cr5, NULL, 1);
187}
188#endif /* CONFIG_64BIT */
189
190/*
191 * This is called from binfmt_elf, we create the special vma for the
192 * vDSO and insert it into the mm struct tree
193 */
194int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
195{
196	struct mm_struct *mm = current->mm;
197	struct page **vdso_pagelist;
198	unsigned long vdso_pages;
199	unsigned long vdso_base;
200	int rc;
201
202	if (!vdso_enabled)
203		return 0;
204	/*
205	 * Only map the vdso for dynamically linked elf binaries.
206	 */
207	if (!uses_interp)
208		return 0;
209
210	vdso_base = mm->mmap_base;
211#ifdef CONFIG_64BIT
212	vdso_pagelist = vdso64_pagelist;
213	vdso_pages = vdso64_pages;
214#ifdef CONFIG_COMPAT
215	if (is_compat_task()) {
216		vdso_pagelist = vdso32_pagelist;
217		vdso_pages = vdso32_pages;
218	}
219#endif
220#else
221	vdso_pagelist = vdso32_pagelist;
222	vdso_pages = vdso32_pages;
223#endif
224
225	/*
226	 * vDSO has a problem and was disabled, just don't "enable" it for
227	 * the process
228	 */
229	if (vdso_pages == 0)
230		return 0;
231
232	current->mm->context.vdso_base = 0;
233
234	/*
235	 * pick a base address for the vDSO in process space. We try to put
236	 * it at vdso_base which is the "natural" base for it, but we might
237	 * fail and end up putting it elsewhere.
238	 */
239	down_write(&mm->mmap_sem);
240	vdso_base = get_unmapped_area(NULL, vdso_base,
241				      vdso_pages << PAGE_SHIFT, 0, 0);
242	if (IS_ERR_VALUE(vdso_base)) {
243		rc = vdso_base;
244		goto out_up;
245	}
246
247	/*
248	 * Put vDSO base into mm struct. We need to do this before calling
249	 * install_special_mapping or the perf counter mmap tracking code
250	 * will fail to recognise it as a vDSO (since arch_vma_name fails).
251	 */
252	current->mm->context.vdso_base = vdso_base;
253
254	/*
255	 * our vma flags don't have VM_WRITE so by default, the process
256	 * isn't allowed to write those pages.
257	 * gdb can break that with ptrace interface, and thus trigger COW
258	 * on those pages but it's then your responsibility to never do that
259	 * on the "data" page of the vDSO or you'll stop getting kernel
260	 * updates and your nice userland gettimeofday will be totally dead.
261	 * It's fine to use that for setting breakpoints in the vDSO code
262	 * pages though
263	 *
264	 * Make sure the vDSO gets into every core dump.
265	 * Dumping its contents makes post-mortem fully interpretable later
266	 * without matching up the same kernel and hardware config to see
267	 * what PC values meant.
268	 */
269	rc = install_special_mapping(mm, vdso_base, vdso_pages << PAGE_SHIFT,
270				     VM_READ|VM_EXEC|
271				     VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC|
272				     VM_ALWAYSDUMP,
273				     vdso_pagelist);
274	if (rc)
275		current->mm->context.vdso_base = 0;
276out_up:
277	up_write(&mm->mmap_sem);
278	return rc;
279}
280
281const char *arch_vma_name(struct vm_area_struct *vma)
282{
283	if (vma->vm_mm && vma->vm_start == vma->vm_mm->context.vdso_base)
284		return "[vdso]";
285	return NULL;
286}
287
288static int __init vdso_init(void)
289{
290	int i;
291
292	if (!vdso_enabled)
293		return 0;
294	vdso_init_data(vdso_data);
295#if defined(CONFIG_32BIT) || defined(CONFIG_COMPAT)
296	/* Calculate the size of the 32 bit vDSO */
297	vdso32_pages = ((&vdso32_end - &vdso32_start
298			 + PAGE_SIZE - 1) >> PAGE_SHIFT) + 1;
299
300	/* Make sure pages are in the correct state */
301	vdso32_pagelist = kzalloc(sizeof(struct page *) * (vdso32_pages + 1),
302				  GFP_KERNEL);
303	BUG_ON(vdso32_pagelist == NULL);
304	for (i = 0; i < vdso32_pages - 1; i++) {
305		struct page *pg = virt_to_page(vdso32_kbase + i*PAGE_SIZE);
306		ClearPageReserved(pg);
307		get_page(pg);
308		vdso32_pagelist[i] = pg;
309	}
310	vdso32_pagelist[vdso32_pages - 1] = virt_to_page(vdso_data);
311	vdso32_pagelist[vdso32_pages] = NULL;
312#endif
313
314#ifdef CONFIG_64BIT
315	/* Calculate the size of the 64 bit vDSO */
316	vdso64_pages = ((&vdso64_end - &vdso64_start
317			 + PAGE_SIZE - 1) >> PAGE_SHIFT) + 1;
318
319	/* Make sure pages are in the correct state */
320	vdso64_pagelist = kzalloc(sizeof(struct page *) * (vdso64_pages + 1),
321				  GFP_KERNEL);
322	BUG_ON(vdso64_pagelist == NULL);
323	for (i = 0; i < vdso64_pages - 1; i++) {
324		struct page *pg = virt_to_page(vdso64_kbase + i*PAGE_SIZE);
325		ClearPageReserved(pg);
326		get_page(pg);
327		vdso64_pagelist[i] = pg;
328	}
329	vdso64_pagelist[vdso64_pages - 1] = virt_to_page(vdso_data);
330	vdso64_pagelist[vdso64_pages] = NULL;
331#ifndef CONFIG_SMP
332	if (vdso_alloc_per_cpu(0, &S390_lowcore))
333		BUG();
334#endif
335	vdso_init_cr5();
336#endif /* CONFIG_64BIT */
337
338	get_page(virt_to_page(vdso_data));
339
340	smp_wmb();
341
342	return 0;
343}
344arch_initcall(vdso_init);
345
346int in_gate_area_no_task(unsigned long addr)
347{
348	return 0;
349}
350
351int in_gate_area(struct task_struct *task, unsigned long addr)
352{
353	return 0;
354}
355
356struct vm_area_struct *get_gate_vma(struct task_struct *tsk)
357{
358	return NULL;
359}