forked from ~ljy/RK356X_SDK_RELEASE
blame | history | patch | commit | commitdiff
hc
2024-01-05 071106ecf68c401173c58808b1cf5f68cc50d390 
 kernel/arch/powerpc/mm/fault.c
....@@ -1,3 +1,4 @@
1
+// SPDX-License-Identifier: GPL-2.0-or-later
12 /*
23  *  PowerPC version
34  *    Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
....@@ -8,11 +9,6 @@
89  *  Modified by Cort Dougan and Paul Mackerras.
910  *
1011  *  Modified for PPC64 by Dave Engebretsen (engebret@ibm.com)
11
- *
12
- *  This program is free software; you can redistribute it and/or
13
- *  modify it under the terms of the GNU General Public License
14
- *  as published by the Free Software Foundation; either version
15
- *  2 of the License, or (at your option) any later version.
1612  */
1713 
1814 #include <linux/signal.h>
....@@ -39,72 +35,20 @@
3935 
4036 #include <asm/firmware.h>
4137 #include <asm/page.h>
42
-#include <asm/pgtable.h>
4338 #include <asm/mmu.h>
4439 #include <asm/mmu_context.h>
4540 #include <asm/siginfo.h>
4641 #include <asm/debug.h>
42
+#include <asm/kup.h>
43
+#include <asm/inst.h>
4744 
48
-static inline bool notify_page_fault(struct pt_regs *regs)
49
-{
50
-	bool ret = false;
5145 
52
-#ifdef CONFIG_KPROBES
53
-	/* kprobe_running() needs smp_processor_id() */
54
-	if (!user_mode(regs)) {
55
-		preempt_disable();
56
-		if (kprobe_running() && kprobe_fault_handler(regs, 11))
57
-			ret = true;
58
-		preempt_enable();
59
-	}
60
-#endif /* CONFIG_KPROBES */
61
-
62
-	if (unlikely(debugger_fault_handler(regs)))
63
-		ret = true;
64
-
65
-	return ret;
66
-}
67
-
68
-/*
69
- * Check whether the instruction inst is a store using
70
- * an update addressing form which will update r1.
71
- */
72
-static bool store_updates_sp(unsigned int inst)
73
-{
74
-	/* check for 1 in the rA field */
75
-	if (((inst >> 16) & 0x1f) != 1)
76
-		return false;
77
-	/* check major opcode */
78
-	switch (inst >> 26) {
79
-	case OP_STWU:
80
-	case OP_STBU:
81
-	case OP_STHU:
82
-	case OP_STFSU:
83
-	case OP_STFDU:
84
-		return true;
85
-	case OP_STD:	/* std or stdu */
86
-		return (inst & 3) == 1;
87
-	case OP_31:
88
-		/* check minor opcode */
89
-		switch ((inst >> 1) & 0x3ff) {
90
-		case OP_31_XOP_STDUX:
91
-		case OP_31_XOP_STWUX:
92
-		case OP_31_XOP_STBUX:
93
-		case OP_31_XOP_STHUX:
94
-		case OP_31_XOP_STFSUX:
95
-		case OP_31_XOP_STFDUX:
96
-			return true;
97
-		}
98
-	}
99
-	return false;
100
-}
10146 /*
10247  * do_page_fault error handling helpers
10348  */
10449 
10550 static int
106
-__bad_area_nosemaphore(struct pt_regs *regs, unsigned long address, int si_code,
107
-		int pkey)
51
+__bad_area_nosemaphore(struct pt_regs *regs, unsigned long address, int si_code)
10852 {
10953 	/*
11054 	 * If we are in kernel mode, bail out with a SEGV, this will
....@@ -114,18 +58,17 @@
11458 	if (!user_mode(regs))
11559 		return SIGSEGV;
11660 
117
-	_exception_pkey(SIGSEGV, regs, si_code, address, pkey);
61
+	_exception(SIGSEGV, regs, si_code, address);
11862 
11963 	return 0;
12064 }
12165 
12266 static noinline int bad_area_nosemaphore(struct pt_regs *regs, unsigned long address)
12367 {
124
-	return __bad_area_nosemaphore(regs, address, SEGV_MAPERR, 0);
68
+	return __bad_area_nosemaphore(regs, address, SEGV_MAPERR);
12569 }
12670 
127
-static int __bad_area(struct pt_regs *regs, unsigned long address, int si_code,
128
-			int pkey)
71
+static int __bad_area(struct pt_regs *regs, unsigned long address, int si_code)
12972 {
13073 	struct mm_struct *mm = current->mm;
13174 
....@@ -133,56 +76,88 @@
13376 	 * Something tried to access memory that isn't in our memory map..
13477 	 * Fix it, but check if it's kernel or user first..
13578 	 */
136
-	up_read(&mm->mmap_sem);
79
+	mmap_read_unlock(mm);
13780 
138
-	return __bad_area_nosemaphore(regs, address, si_code, pkey);
81
+	return __bad_area_nosemaphore(regs, address, si_code);
13982 }
14083 
14184 static noinline int bad_area(struct pt_regs *regs, unsigned long address)
14285 {
143
-	return __bad_area(regs, address, SEGV_MAPERR, 0);
86
+	return __bad_area(regs, address, SEGV_MAPERR);
14487 }
14588 
146
-static int bad_key_fault_exception(struct pt_regs *regs, unsigned long address,
147
-				    int pkey)
89
+#ifdef CONFIG_PPC_MEM_KEYS
90
+static noinline int bad_access_pkey(struct pt_regs *regs, unsigned long address,
91
+				    struct vm_area_struct *vma)
14892 {
149
-	return __bad_area_nosemaphore(regs, address, SEGV_PKUERR, pkey);
93
+	struct mm_struct *mm = current->mm;
94
+	int pkey;
95
+
96
+	/*
97
+	 * We don't try to fetch the pkey from page table because reading
98
+	 * page table without locking doesn't guarantee stable pte value.
99
+	 * Hence the pkey value that we return to userspace can be different
100
+	 * from the pkey that actually caused access error.
101
+	 *
102
+	 * It does *not* guarantee that the VMA we find here
103
+	 * was the one that we faulted on.
104
+	 *
105
+	 * 1. T1   : mprotect_key(foo, PAGE_SIZE, pkey=4);
106
+	 * 2. T1   : set AMR to deny access to pkey=4, touches, page
107
+	 * 3. T1   : faults...
108
+	 * 4.    T2: mprotect_key(foo, PAGE_SIZE, pkey=5);
109
+	 * 5. T1   : enters fault handler, takes mmap_lock, etc...
110
+	 * 6. T1   : reaches here, sees vma_pkey(vma)=5, when we really
111
+	 *	     faulted on a pte with its pkey=4.
112
+	 */
113
+	pkey = vma_pkey(vma);
114
+
115
+	mmap_read_unlock(mm);
116
+
117
+	/*
118
+	 * If we are in kernel mode, bail out with a SEGV, this will
119
+	 * be caught by the assembly which will restore the non-volatile
120
+	 * registers before calling bad_page_fault()
121
+	 */
122
+	if (!user_mode(regs))
123
+		return SIGSEGV;
124
+
125
+	_exception_pkey(regs, address, pkey);
126
+
127
+	return 0;
150128 }
129
+#endif
151130 
152131 static noinline int bad_access(struct pt_regs *regs, unsigned long address)
153132 {
154
-	return __bad_area(regs, address, SEGV_ACCERR, 0);
133
+	return __bad_area(regs, address, SEGV_ACCERR);
155134 }
156135 
157136 static int do_sigbus(struct pt_regs *regs, unsigned long address,
158137 		     vm_fault_t fault)
159138 {
160
-	siginfo_t info;
161
-	unsigned int lsb = 0;
162
-
163139 	if (!user_mode(regs))
164140 		return SIGBUS;
165141 
166142 	current->thread.trap_nr = BUS_ADRERR;
167
-	clear_siginfo(&info);
168
-	info.si_signo = SIGBUS;
169
-	info.si_errno = 0;
170
-	info.si_code = BUS_ADRERR;
171
-	info.si_addr = (void __user *)address;
172143 #ifdef CONFIG_MEMORY_FAILURE
173144 	if (fault & (VM_FAULT_HWPOISON|VM_FAULT_HWPOISON_LARGE)) {
145
+		unsigned int lsb = 0; /* shutup gcc */
146
+
174147 		pr_err("MCE: Killing %s:%d due to hardware memory corruption fault at %lx\n",
175148 			current->comm, current->pid, address);
176
-		info.si_code = BUS_MCEERR_AR;
149
+
150
+		if (fault & VM_FAULT_HWPOISON_LARGE)
151
+			lsb = hstate_index_to_shift(VM_FAULT_GET_HINDEX(fault));
152
+		if (fault & VM_FAULT_HWPOISON)
153
+			lsb = PAGE_SHIFT;
154
+
155
+		force_sig_mceerr(BUS_MCEERR_AR, (void __user *)address, lsb);
156
+		return 0;
177157 	}
178158 
179
-	if (fault & VM_FAULT_HWPOISON_LARGE)
180
-		lsb = hstate_index_to_shift(VM_FAULT_GET_HINDEX(fault));
181
-	if (fault & VM_FAULT_HWPOISON)
182
-		lsb = PAGE_SHIFT;
183159 #endif
184
-	info.si_addr_lsb = lsb;
185
-	force_sig_info(SIGBUS, &info, current);
160
+	force_sig_fault(SIGBUS, BUS_ADRERR, (void __user *)address);
186161 	return 0;
187162 }
188163 
....@@ -218,78 +193,63 @@
218193 }
219194 
220195 /* Is this a bad kernel fault ? */
221
-static bool bad_kernel_fault(bool is_exec, unsigned long error_code,
222
-			     unsigned long address)
196
+static bool bad_kernel_fault(struct pt_regs *regs, unsigned long error_code,
197
+			     unsigned long address, bool is_write)
223198 {
224
-	/* NX faults set DSISR_PROTFAULT on the 8xx, DSISR_NOEXEC_OR_G on others */
225
-	if (is_exec && (error_code & (DSISR_NOEXEC_OR_G | DSISR_KEYFAULT |
226
-				      DSISR_PROTFAULT))) {
227
-		printk_ratelimited(KERN_CRIT "kernel tried to execute"
228
-				   " exec-protected page (%lx) -"
229
-				   "exploit attempt? (uid: %d)\n",
230
-				   address, from_kuid(&init_user_ns,
231
-						      current_uid()));
232
-	}
233
-	return is_exec || (address >= TASK_SIZE);
234
-}
199
+	int is_exec = TRAP(regs) == 0x400;
235200 
236
-// This comes from 64-bit struct rt_sigframe + __SIGNAL_FRAMESIZE
237
-#define SIGFRAME_MAX_SIZE	(4096 + 128)
201
+	if (is_exec) {
202
+		pr_crit_ratelimited("kernel tried to execute %s page (%lx) - exploit attempt? (uid: %d)\n",
203
+				    address >= TASK_SIZE ? "exec-protected" : "user",
204
+				    address,
205
+				    from_kuid(&init_user_ns, current_uid()));
238206 
239
-static bool bad_stack_expansion(struct pt_regs *regs, unsigned long address,
240
-				struct vm_area_struct *vma, unsigned int flags,
241
-				bool *must_retry)
242
-{
243
-	/*
244
-	 * N.B. The POWER/Open ABI allows programs to access up to
245
-	 * 288 bytes below the stack pointer.
246
-	 * The kernel signal delivery code writes a bit over 4KB
247
-	 * below the stack pointer (r1) before decrementing it.
248
-	 * The exec code can write slightly over 640kB to the stack
249
-	 * before setting the user r1.  Thus we allow the stack to
250
-	 * expand to 1MB without further checks.
251
-	 */
252
-	if (address + 0x100000 < vma->vm_end) {
253
-		unsigned int __user *nip = (unsigned int __user *)regs->nip;
254
-		/* get user regs even if this fault is in kernel mode */
255
-		struct pt_regs *uregs = current->thread.regs;
256
-		if (uregs == NULL)
257
-			return true;
258
-
259
-		/*
260
-		 * A user-mode access to an address a long way below
261
-		 * the stack pointer is only valid if the instruction
262
-		 * is one which would update the stack pointer to the
263
-		 * address accessed if the instruction completed,
264
-		 * i.e. either stwu rs,n(r1) or stwux rs,r1,rb
265
-		 * (or the byte, halfword, float or double forms).
266
-		 *
267
-		 * If we don't check this then any write to the area
268
-		 * between the last mapped region and the stack will
269
-		 * expand the stack rather than segfaulting.
270
-		 */
271
-		if (address + SIGFRAME_MAX_SIZE >= uregs->gpr[1])
272
-			return false;
273
-
274
-		if ((flags & FAULT_FLAG_WRITE) && (flags & FAULT_FLAG_USER) &&
275
-		    access_ok(VERIFY_READ, nip, sizeof(*nip))) {
276
-			unsigned int inst;
277
-			int res;
278
-
279
-			pagefault_disable();
280
-			res = __get_user_inatomic(inst, nip);
281
-			pagefault_enable();
282
-			if (!res)
283
-				return !store_updates_sp(inst);
284
-			*must_retry = true;
285
-		}
207
+		// Kernel exec fault is always bad
286208 		return true;
287209 	}
210
+
211
+	if (!is_exec && address < TASK_SIZE && (error_code & DSISR_PROTFAULT) &&
212
+	    !search_exception_tables(regs->nip)) {
213
+		pr_crit_ratelimited("Kernel attempted to access user page (%lx) - exploit attempt? (uid: %d)\n",
214
+				    address,
215
+				    from_kuid(&init_user_ns, current_uid()));
216
+	}
217
+
218
+	// Kernel fault on kernel address is bad
219
+	if (address >= TASK_SIZE)
220
+		return true;
221
+
222
+	// Fault on user outside of certain regions (eg. copy_tofrom_user()) is bad
223
+	if (!search_exception_tables(regs->nip))
224
+		return true;
225
+
226
+	// Read/write fault in a valid region (the exception table search passed
227
+	// above), but blocked by KUAP is bad, it can never succeed.
228
+	if (bad_kuap_fault(regs, address, is_write))
229
+		return true;
230
+
231
+	// What's left? Kernel fault on user in well defined regions (extable
232
+	// matched), and allowed by KUAP in the faulting context.
288233 	return false;
289234 }
290235 
291
-static bool access_error(bool is_write, bool is_exec,
292
-			 struct vm_area_struct *vma)
236
+#ifdef CONFIG_PPC_MEM_KEYS
237
+static bool access_pkey_error(bool is_write, bool is_exec, bool is_pkey,
238
+			      struct vm_area_struct *vma)
239
+{
240
+	/*
241
+	 * Make sure to check the VMA so that we do not perform
242
+	 * faults just to hit a pkey fault as soon as we fill in a
243
+	 * page. Only called for current mm, hence foreign == 0
244
+	 */
245
+	if (!arch_vma_access_permitted(vma, is_write, is_exec, 0))
246
+		return true;
247
+
248
+	return false;
249
+}
250
+#endif
251
+
252
+static bool access_error(bool is_write, bool is_exec, struct vm_area_struct *vma)
293253 {
294254 	/*
295255 	 * Allow execution from readable areas if the MMU does not
....@@ -313,7 +273,7 @@
313273 		return false;
314274 	}
315275 
316
-	if (unlikely(!(vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE))))
276
+	if (unlikely(!vma_is_accessible(vma)))
317277 		return true;
318278 	/*
319279 	 * We should ideally do the vma pkey access check here. But in the
....@@ -341,9 +301,25 @@
341301 static inline void cmo_account_page_fault(void) { }
342302 #endif /* CONFIG_PPC_SMLPAR */
343303 
344
-#ifdef CONFIG_PPC_STD_MMU
345
-static void sanity_check_fault(bool is_write, unsigned long error_code)
304
+static void sanity_check_fault(bool is_write, bool is_user,
305
+			       unsigned long error_code, unsigned long address)
346306 {
307
+	/*
308
+	 * Userspace trying to access kernel address, we get PROTFAULT for that.
309
+	 */
310
+	if (is_user && address >= TASK_SIZE) {
311
+		if ((long)address == -1)
312
+			return;
313
+
314
+		pr_crit_ratelimited("%s[%d]: User access of kernel address (%lx) - exploit attempt? (uid: %d)\n",
315
+				   current->comm, current->pid, address,
316
+				   from_kuid(&init_user_ns, current_uid()));
317
+		return;
318
+	}
319
+
320
+	if (!IS_ENABLED(CONFIG_PPC_BOOK3S))
321
+		return;
322
+
347323 	/*
348324 	 * For hash translation mode, we should never get a
349325 	 * PROTFAULT. Any update to pte to reduce access will result in us
....@@ -373,12 +349,11 @@
373349 	 * For radix, we can get prot fault for autonuma case, because radix
374350 	 * page table will have them marked noaccess for user.
375351 	 */
376
-	if (!radix_enabled() && !is_write)
377
-		WARN_ON_ONCE(error_code & DSISR_PROTFAULT);
352
+	if (radix_enabled() || is_write)
353
+		return;
354
+
355
+	WARN_ON_ONCE(error_code & DSISR_PROTFAULT);
378356 }
379
-#else
380
-static void sanity_check_fault(bool is_write, unsigned long error_code) { }
381
-#endif /* CONFIG_PPC_STD_MMU */
382357 
383358 /*
384359  * Define the correct "is_write" bit in error_code based
....@@ -416,14 +391,14 @@
416391 {
417392 	struct vm_area_struct * vma;
418393 	struct mm_struct *mm = current->mm;
419
-	unsigned int flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
394
+	unsigned int flags = FAULT_FLAG_DEFAULT;
420395  	int is_exec = TRAP(regs) == 0x400;
421396 	int is_user = user_mode(regs);
422397 	int is_write = page_fault_is_write(error_code);
423398 	vm_fault_t fault, major = 0;
424
-	bool must_retry = false;
399
+	bool kprobe_fault = kprobe_page_fault(regs, 11);
425400 
426
-	if (notify_page_fault(regs))
401
+	if (unlikely(debugger_fault_handler(regs) || kprobe_fault))
427402 		return 0;
428403 
429404 	if (unlikely(page_fault_is_bad(error_code))) {
....@@ -435,13 +410,14 @@
435410 	}
436411 
437412 	/* Additional sanity check(s) */
438
-	sanity_check_fault(is_write, error_code);
413
+	sanity_check_fault(is_write, is_user, error_code, address);
439414 
440415 	/*
441416 	 * The kernel should never take an execute fault nor should it
442
-	 * take a page fault to a kernel address.
417
+	 * take a page fault to a kernel address or a page fault to a user
418
+	 * address outside of dedicated places
443419 	 */
444
-	if (unlikely(!is_user && bad_kernel_fault(is_exec, error_code, address)))
420
+	if (unlikely(!is_user && bad_kernel_fault(regs, error_code, address, is_write)))
445421 		return SIGSEGV;
446422 
447423 	/*
....@@ -463,14 +439,10 @@
463439 
464440 	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address);
465441 
466
-	if (error_code & DSISR_KEYFAULT)
467
-		return bad_key_fault_exception(regs, address,
468
-					       get_mm_addr_key(mm, address));
469
-
470442 	/*
471
-	 * We want to do this outside mmap_sem, because reading code around nip
443
+	 * We want to do this outside mmap_lock, because reading code around nip
472444 	 * can result in fault, which will cause a deadlock when called with
473
-	 * mmap_sem held
445
+	 * mmap_lock held
474446 	 */
475447 	if (is_user)
476448 		flags |= FAULT_FLAG_USER;
....@@ -482,7 +454,7 @@
482454 	/* When running in the kernel we expect faults to occur only to
483455 	 * addresses in user space.  All other faults represent errors in the
484456 	 * kernel and should generate an OOPS.  Unfortunately, in the case of an
485
-	 * erroneous fault occurring in a code path which already holds mmap_sem
457
+	 * erroneous fault occurring in a code path which already holds mmap_lock
486458 	 * we will deadlock attempting to validate the fault against the
487459 	 * address space.  Luckily the kernel only validly references user
488460 	 * space from well defined areas of code, which are listed in the
....@@ -494,12 +466,12 @@
494466 	 * source.  If this is invalid we can skip the address space check,
495467 	 * thus avoiding the deadlock.
496468 	 */
497
-	if (unlikely(!down_read_trylock(&mm->mmap_sem))) {
469
+	if (unlikely(!mmap_read_trylock(mm))) {
498470 		if (!is_user && !search_exception_tables(regs->nip))
499471 			return bad_area_nosemaphore(regs, address);
500472 
501473 retry:
502
-		down_read(&mm->mmap_sem);
474
+		mmap_read_lock(mm);
503475 	} else {
504476 		/*
505477 		 * The above down_read_trylock() might have succeeded in
....@@ -512,29 +484,21 @@
512484 	vma = find_vma(mm, address);
513485 	if (unlikely(!vma))
514486 		return bad_area(regs, address);
515
-	if (likely(vma->vm_start <= address))
516
-		goto good_area;
517
-	if (unlikely(!(vma->vm_flags & VM_GROWSDOWN)))
518
-		return bad_area(regs, address);
519487 
520
-	/* The stack is being expanded, check if it's valid */
521
-	if (unlikely(bad_stack_expansion(regs, address, vma, flags,
522
-					 &must_retry))) {
523
-		if (!must_retry)
488
+	if (unlikely(vma->vm_start > address)) {
489
+		if (unlikely(!(vma->vm_flags & VM_GROWSDOWN)))
524490 			return bad_area(regs, address);
525491 
526
-		up_read(&mm->mmap_sem);
527
-		if (fault_in_pages_readable((const char __user *)regs->nip,
528
-					    sizeof(unsigned int)))
529
-			return bad_area_nosemaphore(regs, address);
530
-		goto retry;
492
+		if (unlikely(expand_stack(vma, address)))
493
+			return bad_area(regs, address);
531494 	}
532495 
533
-	/* Try to expand it */
534
-	if (unlikely(expand_stack(vma, address)))
535
-		return bad_area(regs, address);
496
+#ifdef CONFIG_PPC_MEM_KEYS
497
+	if (unlikely(access_pkey_error(is_write, is_exec,
498
+				       (error_code & DSISR_KEYFAULT), vma)))
499
+		return bad_access_pkey(regs, address, vma);
500
+#endif /* CONFIG_PPC_MEM_KEYS */
536501 
537
-good_area:
538502 	if (unlikely(access_error(is_write, is_exec, vma)))
539503 		return bad_access(regs, address);
540504 
....@@ -543,50 +507,25 @@
543507 	 * make sure we exit gracefully rather than endlessly redo
544508 	 * the fault.
545509 	 */
546
-	fault = handle_mm_fault(vma, address, flags);
547
-
548
-#ifdef CONFIG_PPC_MEM_KEYS
549
-	/*
550
-	 * we skipped checking for access error due to key earlier.
551
-	 * Check that using handle_mm_fault error return.
552
-	 */
553
-	if (unlikely(fault & VM_FAULT_SIGSEGV) &&
554
-		!arch_vma_access_permitted(vma, is_write, is_exec, 0)) {
555
-
556
-		int pkey = vma_pkey(vma);
557
-
558
-		up_read(&mm->mmap_sem);
559
-		return bad_key_fault_exception(regs, address, pkey);
560
-	}
561
-#endif /* CONFIG_PPC_MEM_KEYS */
510
+	fault = handle_mm_fault(vma, address, flags, regs);
562511 
563512 	major |= fault & VM_FAULT_MAJOR;
564513 
514
+	if (fault_signal_pending(fault, regs))
515
+		return user_mode(regs) ? 0 : SIGBUS;
516
+
565517 	/*
566
-	 * Handle the retry right now, the mmap_sem has been released in that
518
+	 * Handle the retry right now, the mmap_lock has been released in that
567519 	 * case.
568520 	 */
569521 	if (unlikely(fault & VM_FAULT_RETRY)) {
570
-		/* We retry only once */
571522 		if (flags & FAULT_FLAG_ALLOW_RETRY) {
572
-			/*
573
-			 * Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk
574
-			 * of starvation.
575
-			 */
576
-			flags &= ~FAULT_FLAG_ALLOW_RETRY;
577523 			flags |= FAULT_FLAG_TRIED;
578
-			if (!fatal_signal_pending(current))
579
-				goto retry;
524
+			goto retry;
580525 		}
581
-
582
-		/*
583
-		 * User mode? Just return to handle the fatal exception otherwise
584
-		 * return to bad_page_fault
585
-		 */
586
-		return is_user ? 0 : SIGBUS;
587526 	}
588527 
589
-	up_read(&current->mm->mmap_sem);
528
+	mmap_read_unlock(current->mm);
590529 
591530 	if (unlikely(fault & VM_FAULT_ERROR))
592531 		return mm_fault_error(regs, address, fault);
....@@ -594,14 +533,9 @@
594533 	/*
595534 	 * Major/minor page fault accounting.
596535 	 */
597
-	if (major) {
598
-		current->maj_flt++;
599
-		perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs, address);
536
+	if (major)
600537 		cmo_account_page_fault();
601
-	} else {
602
-		current->min_flt++;
603
-		perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs, address);
604
-	}
538
+
605539 	return 0;
606540 }
607541 NOKPROBE_SYMBOL(__do_page_fault);
....@@ -624,6 +558,7 @@
624558 void bad_page_fault(struct pt_regs *regs, unsigned long address, int sig)
625559 {
626560 	const struct exception_table_entry *entry;
561
+	int is_write = page_fault_is_write(regs->dsisr);
627562 
628563 	/* Are we prepared to handle this fault?  */
629564 	if ((entry = search_exception_tables(regs->nip)) != NULL) {
....@@ -636,9 +571,11 @@
636571 	switch (TRAP(regs)) {
637572 	case 0x300:
638573 	case 0x380:
639
-		pr_alert("BUG: %s at 0x%08lx\n",
574
+	case 0xe00:
575
+		pr_alert("BUG: %s on %s at 0x%08lx\n",
640576 			 regs->dar < PAGE_SIZE ? "Kernel NULL pointer dereference" :
641
-			 "Unable to handle kernel data access", regs->dar);
577
+			 "Unable to handle kernel data access",
578
+			 is_write ? "write" : "read", regs->dar);
642579 		break;
643580 	case 0x400:
644581 	case 0x480: