blob: c517c449d0dd3570d9f7fb8b80da200808d7cdb1 [file] [log] [blame]
Gennady Sharapovea2ba7d2006-01-08 01:01:31 -08001/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07002 * Copyright (C) 2000, 2001 Jeff Dike (jdike@karaya.com)
3 * Licensed under the GPL
4 */
5
6#include "linux/kernel.h"
7#include "asm/errno.h"
8#include "linux/sched.h"
9#include "linux/mm.h"
10#include "linux/spinlock.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070011#include "linux/init.h"
12#include "linux/ptrace.h"
13#include "asm/semaphore.h"
14#include "asm/pgtable.h"
15#include "asm/pgalloc.h"
16#include "asm/tlbflush.h"
17#include "asm/a.out.h"
18#include "asm/current.h"
19#include "asm/irq.h"
Paolo 'Blaisorblade' Giarrusso546fe1c2005-09-22 21:44:16 -070020#include "sysdep/sigcontext.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070021#include "kern_util.h"
Jeff Dike4ff83ce2007-05-06 14:51:08 -070022#include "as-layout.h"
Jeff Dikeeb830752007-05-06 14:51:07 -070023#include "arch.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#include "kern.h"
25#include "chan_kern.h"
26#include "mconsole_kern.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include "mem.h"
28#include "mem_kern.h"
Gennady Sharapovc66fdd52006-01-08 01:01:32 -080029#include "sysdep/sigcontext.h"
30#include "sysdep/ptrace.h"
31#include "os.h"
Paolo 'Blaisorblade' Giarrussobe662a12005-09-30 11:58:59 -070032#include "skas.h"
Gennady Sharapovea2ba7d2006-01-08 01:01:31 -080033#include "os.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070034
Paolo 'Blaisorblade' Giarrusso3b521662005-09-03 15:57:26 -070035/* Note this is constrained to return 0, -EFAULT, -EACCESS, -ENOMEM by segv(). */
Jeff Dike1d3468a2006-07-10 04:45:13 -070036int handle_page_fault(unsigned long address, unsigned long ip,
Linus Torvalds1da177e2005-04-16 15:20:36 -070037 int is_write, int is_user, int *code_out)
38{
39 struct mm_struct *mm = current->mm;
40 struct vm_area_struct *vma;
41 pgd_t *pgd;
42 pud_t *pud;
43 pmd_t *pmd;
44 pte_t *pte;
Linus Torvalds1da177e2005-04-16 15:20:36 -070045 int err = -EFAULT;
46
47 *code_out = SEGV_MAPERR;
Paolo 'Blaisorblade' Giarrussofea03cb2005-09-22 21:44:20 -070048
49 /* If the fault was during atomic operation, don't take the fault, just
50 * fail. */
51 if (in_atomic())
52 goto out_nosemaphore;
53
Linus Torvalds1da177e2005-04-16 15:20:36 -070054 down_read(&mm->mmap_sem);
55 vma = find_vma(mm, address);
Jeff Dike1d3468a2006-07-10 04:45:13 -070056 if(!vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -070057 goto out;
Jeff Dike1d3468a2006-07-10 04:45:13 -070058 else if(vma->vm_start <= address)
Linus Torvalds1da177e2005-04-16 15:20:36 -070059 goto good_area;
Jeff Dike1d3468a2006-07-10 04:45:13 -070060 else if(!(vma->vm_flags & VM_GROWSDOWN))
Linus Torvalds1da177e2005-04-16 15:20:36 -070061 goto out;
Jeff Dike2d58cc92005-05-06 21:30:55 -070062 else if(is_user && !ARCH_IS_STACKGROW(address))
Linus Torvalds1da177e2005-04-16 15:20:36 -070063 goto out;
Jeff Dike1d3468a2006-07-10 04:45:13 -070064 else if(expand_stack(vma, address))
Linus Torvalds1da177e2005-04-16 15:20:36 -070065 goto out;
66
Paolo 'Blaisorblade' Giarrusso3b521662005-09-03 15:57:26 -070067good_area:
Linus Torvalds1da177e2005-04-16 15:20:36 -070068 *code_out = SEGV_ACCERR;
Jeff Dike1d3468a2006-07-10 04:45:13 -070069 if(is_write && !(vma->vm_flags & VM_WRITE))
Linus Torvalds1da177e2005-04-16 15:20:36 -070070 goto out;
Jeff Dike13479d52005-05-20 13:59:08 -070071
Paolo 'Blaisorblade' Giarrussod129f312005-09-10 19:44:57 +020072 /* Don't require VM_READ|VM_EXEC for write faults! */
Jeff Dike5d864562007-05-06 14:51:24 -070073 if(!is_write && !(vma->vm_flags & (VM_READ | VM_EXEC)))
74 goto out;
Jeff Dike13479d52005-05-20 13:59:08 -070075
Linus Torvalds1da177e2005-04-16 15:20:36 -070076 do {
Nick Piggin83c54072007-07-19 01:47:05 -070077 int fault;
Paolo 'Blaisorblade' Giarrusso3b521662005-09-03 15:57:26 -070078survive:
Nick Piggin83c54072007-07-19 01:47:05 -070079 fault = handle_mm_fault(mm, vma, address, is_write);
80 if (unlikely(fault & VM_FAULT_ERROR)) {
81 if (fault & VM_FAULT_OOM) {
82 err = -ENOMEM;
83 goto out_of_memory;
84 } else if (fault & VM_FAULT_SIGBUS) {
85 err = -EACCES;
86 goto out;
87 }
Linus Torvalds1da177e2005-04-16 15:20:36 -070088 BUG();
89 }
Nick Piggin83c54072007-07-19 01:47:05 -070090 if (fault & VM_FAULT_MAJOR)
91 current->maj_flt++;
92 else
93 current->min_flt++;
94
Paolo 'Blaisorblade' Giarrusso3b521662005-09-03 15:57:26 -070095 pgd = pgd_offset(mm, address);
96 pud = pud_offset(pgd, address);
97 pmd = pmd_offset(pud, address);
98 pte = pte_offset_kernel(pmd, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -070099 } while(!pte_present(*pte));
100 err = 0;
Paolo 'Blaisorblade' Giarrussocbc24af2005-11-13 16:07:04 -0800101 /* The below warning was added in place of
102 * pte_mkyoung(); if (is_write) pte_mkdirty();
103 * If it's triggered, we'd see normally a hang here (a clean pte is
104 * marked read-only to emulate the dirty bit).
105 * However, the generic code can mark a PTE writable but clean on a
106 * concurrent read fault, triggering this harmlessly. So comment it out.
107 */
108#if 0
Paolo 'Blaisorblade' Giarrusso16b03672005-09-10 19:44:58 +0200109 WARN_ON(!pte_young(*pte) || (is_write && !pte_dirty(*pte)));
Paolo 'Blaisorblade' Giarrussocbc24af2005-11-13 16:07:04 -0800110#endif
Paolo 'Blaisorblade' Giarrusso3b521662005-09-03 15:57:26 -0700111 flush_tlb_page(vma, address);
112out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113 up_read(&mm->mmap_sem);
Paolo 'Blaisorblade' Giarrussofea03cb2005-09-22 21:44:20 -0700114out_nosemaphore:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115 return(err);
116
117/*
118 * We ran out of memory, or some other thing happened to us that made
119 * us unable to handle the page fault gracefully.
120 */
121out_of_memory:
Sukadev Bhattiproluf400e192006-09-29 02:00:07 -0700122 if (is_init(current)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123 up_read(&mm->mmap_sem);
124 yield();
125 down_read(&mm->mmap_sem);
126 goto survive;
127 }
128 goto out;
129}
130
Jeff Dike27aa6ef2007-02-10 01:44:14 -0800131static void bad_segv(struct faultinfo fi, unsigned long ip)
132{
133 struct siginfo si;
134
135 si.si_signo = SIGSEGV;
136 si.si_code = SEGV_ACCERR;
137 si.si_addr = (void __user *) FAULT_ADDRESS(fi);
138 current->thread.arch.faultinfo = fi;
139 force_sig_info(SIGSEGV, &si, current);
140}
141
142static void segv_handler(int sig, union uml_pt_regs *regs)
Gennady Sharapovc66fdd52006-01-08 01:01:32 -0800143{
144 struct faultinfo * fi = UPT_FAULTINFO(regs);
145
146 if(UPT_IS_USER(regs) && !SEGV_IS_FIXABLE(fi)){
147 bad_segv(*fi, UPT_IP(regs));
148 return;
149 }
150 segv(*fi, UPT_IP(regs), UPT_IS_USER(regs), regs);
151}
152
Bodo Stroesserc5784552005-05-05 16:15:31 -0700153/*
154 * We give a *copy* of the faultinfo in the regs to segv.
155 * This must be done, since nesting SEGVs could overwrite
156 * the info in the regs. A pointer to the info then would
157 * give us bad data!
158 */
Jeff Dike5d864562007-05-06 14:51:24 -0700159unsigned long segv(struct faultinfo fi, unsigned long ip, int is_user,
160 union uml_pt_regs *regs)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161{
162 struct siginfo si;
163 void *catcher;
164 int err;
Jeff Dike5d864562007-05-06 14:51:24 -0700165 int is_write = FAULT_WRITE(fi);
166 unsigned long address = FAULT_ADDRESS(fi);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167
Jeff Dike5d864562007-05-06 14:51:24 -0700168 if(!is_user && (address >= start_vm) && (address < end_vm)){
169 flush_tlb_kernel_vm();
170 return 0;
171 }
Jeff Dike377fad32007-05-06 14:51:25 -0700172 else if(current->mm == NULL) {
173 show_regs(container_of(regs, struct pt_regs, regs));
174 panic("Segfault with no mm");
175 }
Paolo 'Blaisorblade' Giarrusso546fe1c2005-09-22 21:44:16 -0700176
Paolo 'Blaisorblade' Giarrussobe662a12005-09-30 11:58:59 -0700177 if (SEGV_IS_FIXABLE(&fi) || SEGV_MAYBE_FIXABLE(&fi))
Paolo 'Blaisorblade' Giarrusso546fe1c2005-09-22 21:44:16 -0700178 err = handle_page_fault(address, ip, is_write, is_user, &si.si_code);
179 else {
180 err = -EFAULT;
181 /* A thread accessed NULL, we get a fault, but CR2 is invalid.
182 * This code is used in __do_copy_from_user() of TT mode. */
183 address = 0;
184 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185
186 catcher = current->thread.fault_catcher;
187 if(!err)
Jeff Dike5d864562007-05-06 14:51:24 -0700188 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189 else if(catcher != NULL){
190 current->thread.fault_addr = (void *) address;
191 do_longjmp(catcher, 1);
Jeff Dike1d3468a2006-07-10 04:45:13 -0700192 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193 else if(current->thread.fault_addr != NULL)
194 panic("fault_addr set but no fault catcher");
Jeff Dike5d864562007-05-06 14:51:24 -0700195 else if(!is_user && arch_fixup(ip, regs))
196 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197
Jeff Dike377fad32007-05-06 14:51:25 -0700198 if(!is_user) {
199 show_regs(container_of(regs, struct pt_regs, regs));
Jeff Dike1d3468a2006-07-10 04:45:13 -0700200 panic("Kernel mode fault at addr 0x%lx, ip 0x%lx",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201 address, ip);
Jeff Dike377fad32007-05-06 14:51:25 -0700202 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203
Paolo 'Blaisorblade' Giarrusso3b521662005-09-03 15:57:26 -0700204 if (err == -EACCES) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205 si.si_signo = SIGBUS;
206 si.si_errno = 0;
207 si.si_code = BUS_ADRERR;
Al Viro4d338e12006-03-31 02:30:15 -0800208 si.si_addr = (void __user *)address;
Jeff Dike5d864562007-05-06 14:51:24 -0700209 current->thread.arch.faultinfo = fi;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210 force_sig_info(SIGBUS, &si, current);
Paolo 'Blaisorblade' Giarrusso3b521662005-09-03 15:57:26 -0700211 } else if (err == -ENOMEM) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212 printk("VM: killing process %s\n", current->comm);
213 do_exit(SIGKILL);
Paolo 'Blaisorblade' Giarrusso3b521662005-09-03 15:57:26 -0700214 } else {
215 BUG_ON(err != -EFAULT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216 si.si_signo = SIGSEGV;
Al Viro4d338e12006-03-31 02:30:15 -0800217 si.si_addr = (void __user *) address;
Jeff Dike5d864562007-05-06 14:51:24 -0700218 current->thread.arch.faultinfo = fi;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219 force_sig_info(SIGSEGV, &si, current);
220 }
Jeff Dike5d864562007-05-06 14:51:24 -0700221 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222}
223
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224void relay_signal(int sig, union uml_pt_regs *regs)
225{
Jeff Dike6edf4282006-09-25 23:33:03 -0700226 if(arch_handle_signal(sig, regs))
227 return;
228
229 if(!UPT_IS_USER(regs)){
230 if(sig == SIGBUS)
Jeff Dike83ff7df2007-05-06 14:51:50 -0700231 printk("Bus error - the host /dev/shm or /tmp mount "
232 "likely just ran out of space\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233 panic("Kernel mode signal %d", sig);
Jeff Dike6edf4282006-09-25 23:33:03 -0700234 }
235
Jeff Dike5d864562007-05-06 14:51:24 -0700236 current->thread.arch.faultinfo = *UPT_FAULTINFO(regs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237 force_sig(sig, current);
238}
239
Jeff Dike27aa6ef2007-02-10 01:44:14 -0800240static void bus_handler(int sig, union uml_pt_regs *regs)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241{
242 if(current->thread.fault_catcher != NULL)
243 do_longjmp(current->thread.fault_catcher, 1);
244 else relay_signal(sig, regs);
245}
246
Jeff Dike27aa6ef2007-02-10 01:44:14 -0800247static void winch(int sig, union uml_pt_regs *regs)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248{
249 do_IRQ(WINCH_IRQ, regs);
250}
251
Jeff Dike53dd2b52006-09-27 01:50:37 -0700252const struct kern_handlers handlinfo_kern = {
253 .relay_signal = relay_signal,
254 .winch = winch,
255 .bus_handler = bus_handler,
256 .page_fault = segv_handler,
257 .sigio_handler = sigio_handler,
258 .timer_handler = timer_handler
259};
260
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261void trap_init(void)
262{
263}