~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/arch/um/kernel/trap.c

Version: ~ [ linux-5.8 ] ~ [ linux-5.7.14 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.57 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.138 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.193 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.232 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.232 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.85 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 // SPDX-License-Identifier: GPL-2.0
  2 /*
  3  * Copyright (C) 2000 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
  4  */
  5 
  6 #include <linux/mm.h>
  7 #include <linux/sched/signal.h>
  8 #include <linux/hardirq.h>
  9 #include <linux/module.h>
 10 #include <linux/uaccess.h>
 11 #include <linux/sched/debug.h>
 12 #include <asm/current.h>
 13 #include <asm/tlbflush.h>
 14 #include <arch.h>
 15 #include <as-layout.h>
 16 #include <kern_util.h>
 17 #include <os.h>
 18 #include <skas.h>
 19 
 20 /*
 21  * Note this is constrained to return 0, -EFAULT, -EACCES, -ENOMEM by
 22  * segv().
 23  */
 24 int handle_page_fault(unsigned long address, unsigned long ip,
 25                       int is_write, int is_user, int *code_out)
 26 {
 27         struct mm_struct *mm = current->mm;
 28         struct vm_area_struct *vma;
 29         pmd_t *pmd;
 30         pte_t *pte;
 31         int err = -EFAULT;
 32         unsigned int flags = FAULT_FLAG_DEFAULT;
 33 
 34         *code_out = SEGV_MAPERR;
 35 
 36         /*
 37          * If the fault was with pagefaults disabled, don't take the fault, just
 38          * fail.
 39          */
 40         if (faulthandler_disabled())
 41                 goto out_nosemaphore;
 42 
 43         if (is_user)
 44                 flags |= FAULT_FLAG_USER;
 45 retry:
 46         mmap_read_lock(mm);
 47         vma = find_vma(mm, address);
 48         if (!vma)
 49                 goto out;
 50         else if (vma->vm_start <= address)
 51                 goto good_area;
 52         else if (!(vma->vm_flags & VM_GROWSDOWN))
 53                 goto out;
 54         else if (is_user && !ARCH_IS_STACKGROW(address))
 55                 goto out;
 56         else if (expand_stack(vma, address))
 57                 goto out;
 58 
 59 good_area:
 60         *code_out = SEGV_ACCERR;
 61         if (is_write) {
 62                 if (!(vma->vm_flags & VM_WRITE))
 63                         goto out;
 64                 flags |= FAULT_FLAG_WRITE;
 65         } else {
 66                 /* Don't require VM_READ|VM_EXEC for write faults! */
 67                 if (!(vma->vm_flags & (VM_READ | VM_EXEC)))
 68                         goto out;
 69         }
 70 
 71         do {
 72                 vm_fault_t fault;
 73 
 74                 fault = handle_mm_fault(vma, address, flags);
 75 
 76                 if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current))
 77                         goto out_nosemaphore;
 78 
 79                 if (unlikely(fault & VM_FAULT_ERROR)) {
 80                         if (fault & VM_FAULT_OOM) {
 81                                 goto out_of_memory;
 82                         } else if (fault & VM_FAULT_SIGSEGV) {
 83                                 goto out;
 84                         } else if (fault & VM_FAULT_SIGBUS) {
 85                                 err = -EACCES;
 86                                 goto out;
 87                         }
 88                         BUG();
 89                 }
 90                 if (flags & FAULT_FLAG_ALLOW_RETRY) {
 91                         if (fault & VM_FAULT_MAJOR)
 92                                 current->maj_flt++;
 93                         else
 94                                 current->min_flt++;
 95                         if (fault & VM_FAULT_RETRY) {
 96                                 flags |= FAULT_FLAG_TRIED;
 97 
 98                                 goto retry;
 99                         }
100                 }
101 
102                 pmd = pmd_off(mm, address);
103                 pte = pte_offset_kernel(pmd, address);
104         } while (!pte_present(*pte));
105         err = 0;
106         /*
107          * The below warning was added in place of
108          *      pte_mkyoung(); if (is_write) pte_mkdirty();
109          * If it's triggered, we'd see normally a hang here (a clean pte is
110          * marked read-only to emulate the dirty bit).
111          * However, the generic code can mark a PTE writable but clean on a
112          * concurrent read fault, triggering this harmlessly. So comment it out.
113          */
114 #if 0
115         WARN_ON(!pte_young(*pte) || (is_write && !pte_dirty(*pte)));
116 #endif
117         flush_tlb_page(vma, address);
118 out:
119         mmap_read_unlock(mm);
120 out_nosemaphore:
121         return err;
122 
123 out_of_memory:
124         /*
125          * We ran out of memory, call the OOM killer, and return the userspace
126          * (which will retry the fault, or kill us if we got oom-killed).
127          */
128         mmap_read_unlock(mm);
129         if (!is_user)
130                 goto out_nosemaphore;
131         pagefault_out_of_memory();
132         return 0;
133 }
134 EXPORT_SYMBOL(handle_page_fault);
135 
136 static void show_segv_info(struct uml_pt_regs *regs)
137 {
138         struct task_struct *tsk = current;
139         struct faultinfo *fi = UPT_FAULTINFO(regs);
140 
141         if (!unhandled_signal(tsk, SIGSEGV))
142                 return;
143 
144         if (!printk_ratelimit())
145                 return;
146 
147         printk("%s%s[%d]: segfault at %lx ip %px sp %px error %x",
148                 task_pid_nr(tsk) > 1 ? KERN_INFO : KERN_EMERG,
149                 tsk->comm, task_pid_nr(tsk), FAULT_ADDRESS(*fi),
150                 (void *)UPT_IP(regs), (void *)UPT_SP(regs),
151                 fi->error_code);
152 
153         print_vma_addr(KERN_CONT " in ", UPT_IP(regs));
154         printk(KERN_CONT "\n");
155 }
156 
157 static void bad_segv(struct faultinfo fi, unsigned long ip)
158 {
159         current->thread.arch.faultinfo = fi;
160         force_sig_fault(SIGSEGV, SEGV_ACCERR, (void __user *) FAULT_ADDRESS(fi));
161 }
162 
163 void fatal_sigsegv(void)
164 {
165         force_sigsegv(SIGSEGV);
166         do_signal(&current->thread.regs);
167         /*
168          * This is to tell gcc that we're not returning - do_signal
169          * can, in general, return, but in this case, it's not, since
170          * we just got a fatal SIGSEGV queued.
171          */
172         os_dump_core();
173 }
174 
175 /**
176  * segv_handler() - the SIGSEGV handler
177  * @sig:        the signal number
178  * @unused_si:  the signal info struct; unused in this handler
179  * @regs:       the ptrace register information
180  *
181  * The handler first extracts the faultinfo from the UML ptrace regs struct.
182  * If the userfault did not happen in an UML userspace process, bad_segv is called.
183  * Otherwise the signal did happen in a cloned userspace process, handle it.
184  */
185 void segv_handler(int sig, struct siginfo *unused_si, struct uml_pt_regs *regs)
186 {
187         struct faultinfo * fi = UPT_FAULTINFO(regs);
188 
189         if (UPT_IS_USER(regs) && !SEGV_IS_FIXABLE(fi)) {
190                 show_segv_info(regs);
191                 bad_segv(*fi, UPT_IP(regs));
192                 return;
193         }
194         segv(*fi, UPT_IP(regs), UPT_IS_USER(regs), regs);
195 }
196 
197 /*
198  * We give a *copy* of the faultinfo in the regs to segv.
199  * This must be done, since nesting SEGVs could overwrite
200  * the info in the regs. A pointer to the info then would
201  * give us bad data!
202  */
203 unsigned long segv(struct faultinfo fi, unsigned long ip, int is_user,
204                    struct uml_pt_regs *regs)
205 {
206         jmp_buf *catcher;
207         int si_code;
208         int err;
209         int is_write = FAULT_WRITE(fi);
210         unsigned long address = FAULT_ADDRESS(fi);
211 
212         if (!is_user && regs)
213                 current->thread.segv_regs = container_of(regs, struct pt_regs, regs);
214 
215         if (!is_user && (address >= start_vm) && (address < end_vm)) {
216                 flush_tlb_kernel_vm();
217                 goto out;
218         }
219         else if (current->mm == NULL) {
220                 show_regs(container_of(regs, struct pt_regs, regs));
221                 panic("Segfault with no mm");
222         }
223         else if (!is_user && address > PAGE_SIZE && address < TASK_SIZE) {
224                 show_regs(container_of(regs, struct pt_regs, regs));
225                 panic("Kernel tried to access user memory at addr 0x%lx, ip 0x%lx",
226                        address, ip);
227         }
228 
229         if (SEGV_IS_FIXABLE(&fi))
230                 err = handle_page_fault(address, ip, is_write, is_user,
231                                         &si_code);
232         else {
233                 err = -EFAULT;
234                 /*
235                  * A thread accessed NULL, we get a fault, but CR2 is invalid.
236                  * This code is used in __do_copy_from_user() of TT mode.
237                  * XXX tt mode is gone, so maybe this isn't needed any more
238                  */
239                 address = 0;
240         }
241 
242         catcher = current->thread.fault_catcher;
243         if (!err)
244                 goto out;
245         else if (catcher != NULL) {
246                 current->thread.fault_addr = (void *) address;
247                 UML_LONGJMP(catcher, 1);
248         }
249         else if (current->thread.fault_addr != NULL)
250                 panic("fault_addr set but no fault catcher");
251         else if (!is_user && arch_fixup(ip, regs))
252                 goto out;
253 
254         if (!is_user) {
255                 show_regs(container_of(regs, struct pt_regs, regs));
256                 panic("Kernel mode fault at addr 0x%lx, ip 0x%lx",
257                       address, ip);
258         }
259 
260         show_segv_info(regs);
261 
262         if (err == -EACCES) {
263                 current->thread.arch.faultinfo = fi;
264                 force_sig_fault(SIGBUS, BUS_ADRERR, (void __user *)address);
265         } else {
266                 BUG_ON(err != -EFAULT);
267                 current->thread.arch.faultinfo = fi;
268                 force_sig_fault(SIGSEGV, si_code, (void __user *) address);
269         }
270 
271 out:
272         if (regs)
273                 current->thread.segv_regs = NULL;
274 
275         return 0;
276 }
277 
278 void relay_signal(int sig, struct siginfo *si, struct uml_pt_regs *regs)
279 {
280         int code, err;
281         if (!UPT_IS_USER(regs)) {
282                 if (sig == SIGBUS)
283                         printk(KERN_ERR "Bus error - the host /dev/shm or /tmp "
284                                "mount likely just ran out of space\n");
285                 panic("Kernel mode signal %d", sig);
286         }
287 
288         arch_examine_signal(sig, regs);
289 
290         /* Is the signal layout for the signal known?
291          * Signal data must be scrubbed to prevent information leaks.
292          */
293         code = si->si_code;
294         err = si->si_errno;
295         if ((err == 0) && (siginfo_layout(sig, code) == SIL_FAULT)) {
296                 struct faultinfo *fi = UPT_FAULTINFO(regs);
297                 current->thread.arch.faultinfo = *fi;
298                 force_sig_fault(sig, code, (void __user *)FAULT_ADDRESS(*fi));
299         } else {
300                 printk(KERN_ERR "Attempted to relay unknown signal %d (si_code = %d) with errno %d\n",
301                        sig, code, err);
302                 force_sig(sig);
303         }
304 }
305 
306 void bus_handler(int sig, struct siginfo *si, struct uml_pt_regs *regs)
307 {
308         if (current->thread.fault_catcher != NULL)
309                 UML_LONGJMP(current->thread.fault_catcher, 1);
310         else
311                 relay_signal(sig, si, regs);
312 }
313 
314 void winch(int sig, struct siginfo *unused_si, struct uml_pt_regs *regs)
315 {
316         do_IRQ(WINCH_IRQ, regs);
317 }
318 
319 void trap_init(void)
320 {
321 }
322 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp