slb.c 10.5 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4
/*
 * PowerPC64 SLB support.
 *
 * Copyright (C) 2004 David Gibson <dwg@au.ibm.com>, IBM
5
 * Based on earlier code written by:
L
Linus Torvalds 已提交
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
 * Dave Engebretsen and Mike Corrigan {engebret|mikejc}@us.ibm.com
 *    Copyright (c) 2001 Dave Engebretsen
 * Copyright (C) 2002 Anton Blanchard <anton@au.ibm.com>, IBM
 *
 *
 *      This program is free software; you can redistribute it and/or
 *      modify it under the terms of the GNU General Public License
 *      as published by the Free Software Foundation; either version
 *      2 of the License, or (at your option) any later version.
 */

#include <asm/pgtable.h>
#include <asm/mmu.h>
#include <asm/mmu_context.h>
#include <asm/paca.h>
#include <asm/cputable.h>
22
#include <asm/cacheflush.h>
23 24
#include <asm/smp.h>
#include <linux/compiler.h>
25
#include <asm/udbg.h>
26
#include <asm/code-patching.h>
27

L
Linus Torvalds 已提交
28

29 30 31 32 33 34 35 36 37 38
extern void slb_allocate_realmode(unsigned long ea);
extern void slb_allocate_user(unsigned long ea);

static void slb_allocate(unsigned long ea)
{
	/* Currently, we do real mode for all SLBs including user, but
	 * that will change if we bring back dynamic VSIDs
	 */
	slb_allocate_realmode(ea);
}
L
Linus Torvalds 已提交
39

40 41 42
#define slb_esid_mask(ssize)	\
	(((ssize) == MMU_SEGSIZE_256M)? ESID_MASK: ESID_MASK_1T)

P
Paul Mackerras 已提交
43
static inline unsigned long mk_esid_data(unsigned long ea, int ssize,
44
					 unsigned long entry)
L
Linus Torvalds 已提交
45
{
46
	return (ea & slb_esid_mask(ssize)) | SLB_ESID_V | entry;
L
Linus Torvalds 已提交
47 48
}

P
Paul Mackerras 已提交
49 50
static inline unsigned long mk_vsid_data(unsigned long ea, int ssize,
					 unsigned long flags)
L
Linus Torvalds 已提交
51
{
P
Paul Mackerras 已提交
52 53
	return (get_kernel_vsid(ea, ssize) << slb_vsid_shift(ssize)) | flags |
		((unsigned long) ssize << SLB_VSID_SSIZE_SHIFT);
L
Linus Torvalds 已提交
54 55
}

P
Paul Mackerras 已提交
56
static inline void slb_shadow_update(unsigned long ea, int ssize,
57
				     unsigned long flags,
58
				     unsigned long entry)
L
Linus Torvalds 已提交
59
{
60 61
	/*
	 * Clear the ESID first so the entry is not valid while we are
62 63
	 * updating it.  No write barriers are needed here, provided
	 * we only update the current CPU's SLB shadow buffer.
64 65
	 */
	get_slb_shadow()->save_area[entry].esid = 0;
66 67 68 69
	get_slb_shadow()->save_area[entry].vsid =
				cpu_to_be64(mk_vsid_data(ea, ssize, flags));
	get_slb_shadow()->save_area[entry].esid =
				cpu_to_be64(mk_esid_data(ea, ssize, entry));
70 71
}

72
static inline void slb_shadow_clear(unsigned long entry)
73
{
74
	get_slb_shadow()->save_area[entry].esid = 0;
L
Linus Torvalds 已提交
75 76
}

P
Paul Mackerras 已提交
77 78
static inline void create_shadowed_slbe(unsigned long ea, int ssize,
					unsigned long flags,
79 80 81 82 83 84 85
					unsigned long entry)
{
	/*
	 * Updating the shadow buffer before writing the SLB ensures
	 * we don't get a stale entry here if we get preempted by PHYP
	 * between these two statements.
	 */
P
Paul Mackerras 已提交
86
	slb_shadow_update(ea, ssize, flags, entry);
87 88

	asm volatile("slbmte  %0,%1" :
P
Paul Mackerras 已提交
89 90
		     : "r" (mk_vsid_data(ea, ssize, flags)),
		       "r" (mk_esid_data(ea, ssize, entry))
91 92 93
		     : "memory" );
}

94
static void __slb_flush_and_rebolt(void)
L
Linus Torvalds 已提交
95 96
{
	/* If you change this make sure you change SLB_NUM_BOLTED
97
	 * and PR KVM appropriately too. */
98
	unsigned long linear_llp, vmalloc_llp, lflags, vflags;
P
Paul Mackerras 已提交
99
	unsigned long ksp_esid_data, ksp_vsid_data;
L
Linus Torvalds 已提交
100

101
	linear_llp = mmu_psize_defs[mmu_linear_psize].sllp;
102
	vmalloc_llp = mmu_psize_defs[mmu_vmalloc_psize].sllp;
103
	lflags = SLB_VSID_KERNEL | linear_llp;
104
	vflags = SLB_VSID_KERNEL | vmalloc_llp;
L
Linus Torvalds 已提交
105

P
Paul Mackerras 已提交
106 107
	ksp_esid_data = mk_esid_data(get_paca()->kstack, mmu_kernel_ssize, 2);
	if ((ksp_esid_data & ~0xfffffffUL) <= PAGE_OFFSET) {
L
Linus Torvalds 已提交
108
		ksp_esid_data &= ~SLB_ESID_V;
P
Paul Mackerras 已提交
109
		ksp_vsid_data = 0;
110 111 112
		slb_shadow_clear(2);
	} else {
		/* Update stack entry; others don't change */
P
Paul Mackerras 已提交
113
		slb_shadow_update(get_paca()->kstack, mmu_kernel_ssize, lflags, 2);
114 115
		ksp_vsid_data =
			be64_to_cpu(get_slb_shadow()->save_area[2].vsid);
116
	}
117

L
Linus Torvalds 已提交
118 119 120 121 122 123 124 125 126
	/* We need to do this all in asm, so we're sure we don't touch
	 * the stack between the slbia and rebolting it. */
	asm volatile("isync\n"
		     "slbia\n"
		     /* Slot 1 - first VMALLOC segment */
		     "slbmte	%0,%1\n"
		     /* Slot 2 - kernel stack */
		     "slbmte	%2,%3\n"
		     "isync"
P
Paul Mackerras 已提交
127 128 129
		     :: "r"(mk_vsid_data(VMALLOC_START, mmu_kernel_ssize, vflags)),
		        "r"(mk_esid_data(VMALLOC_START, mmu_kernel_ssize, 1)),
		        "r"(ksp_vsid_data),
L
Linus Torvalds 已提交
130 131 132 133
		        "r"(ksp_esid_data)
		     : "memory");
}

134 135 136 137 138 139 140 141 142 143 144 145 146 147 148
void slb_flush_and_rebolt(void)
{

	WARN_ON(!irqs_disabled());

	/*
	 * We can't take a PMU exception in the following code, so hard
	 * disable interrupts.
	 */
	hard_irq_disable();

	__slb_flush_and_rebolt();
	get_paca()->slb_cache_ptr = 0;
}

149 150 151 152 153
void slb_vmalloc_update(void)
{
	unsigned long vflags;

	vflags = SLB_VSID_KERNEL | mmu_psize_defs[mmu_vmalloc_psize].sllp;
P
Paul Mackerras 已提交
154
	slb_shadow_update(VMALLOC_START, mmu_kernel_ssize, vflags, 1);
155 156 157
	slb_flush_and_rebolt();
}

158 159 160 161 162 163 164 165 166 167 168
/* Helper function to compare esids.  There are four cases to handle.
 * 1. The system is not 1T segment size capable.  Use the GET_ESID compare.
 * 2. The system is 1T capable, both addresses are < 1T, use the GET_ESID compare.
 * 3. The system is 1T capable, only one of the two addresses is > 1T.  This is not a match.
 * 4. The system is 1T capable, both addresses are > 1T, use the GET_ESID_1T macro to compare.
 */
static inline int esids_match(unsigned long addr1, unsigned long addr2)
{
	int esid_1t_count;

	/* System is not 1T segment size capable. */
169
	if (!mmu_has_feature(MMU_FTR_1T_SEGMENT))
170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186
		return (GET_ESID(addr1) == GET_ESID(addr2));

	esid_1t_count = (((addr1 >> SID_SHIFT_1T) != 0) +
				((addr2 >> SID_SHIFT_1T) != 0));

	/* both addresses are < 1T */
	if (esid_1t_count == 0)
		return (GET_ESID(addr1) == GET_ESID(addr2));

	/* One address < 1T, the other > 1T.  Not a match */
	if (esid_1t_count == 1)
		return 0;

	/* Both addresses are > 1T. */
	return (GET_ESID_1T(addr1) == GET_ESID_1T(addr2));
}

L
Linus Torvalds 已提交
187 188 189
/* Flush all user entries from the segment table of the current processor. */
void switch_slb(struct task_struct *tsk, struct mm_struct *mm)
{
190
	unsigned long offset;
P
Paul Mackerras 已提交
191
	unsigned long slbie_data = 0;
L
Linus Torvalds 已提交
192 193
	unsigned long pc = KSTK_EIP(tsk);
	unsigned long stack = KSTK_ESP(tsk);
194
	unsigned long exec_base;
L
Linus Torvalds 已提交
195

196 197 198 199 200 201 202 203
	/*
	 * We need interrupts hard-disabled here, not just soft-disabled,
	 * so that a PMU interrupt can't occur, which might try to access
	 * user memory (to get a stack trace) and possible cause an SLB miss
	 * which would update the slb_cache/slb_cache_ptr fields in the PACA.
	 */
	hard_irq_disable();
	offset = get_paca()->slb_cache_ptr;
204
	if (!mmu_has_feature(MMU_FTR_NO_SLBIE_B) &&
205
	    offset <= SLB_CACHE_ENTRIES) {
L
Linus Torvalds 已提交
206 207 208
		int i;
		asm volatile("isync" : : : "memory");
		for (i = 0; i < offset; i++) {
P
Paul Mackerras 已提交
209 210 211 212 213 214
			slbie_data = (unsigned long)get_paca()->slb_cache[i]
				<< SID_SHIFT; /* EA */
			slbie_data |= user_segment_size(slbie_data)
				<< SLBIE_SSIZE_SHIFT;
			slbie_data |= SLBIE_C; /* C set for user addresses */
			asm volatile("slbie %0" : : "r" (slbie_data));
L
Linus Torvalds 已提交
215 216 217
		}
		asm volatile("isync" : : : "memory");
	} else {
218
		__slb_flush_and_rebolt();
L
Linus Torvalds 已提交
219 220 221 222
	}

	/* Workaround POWER5 < DD2.1 issue */
	if (offset == 1 || offset > SLB_CACHE_ENTRIES)
P
Paul Mackerras 已提交
223
		asm volatile("slbie %0" : : "r" (slbie_data));
L
Linus Torvalds 已提交
224 225 226 227 228 229

	get_paca()->slb_cache_ptr = 0;
	get_paca()->context = mm->context;

	/*
	 * preload some userspace segments into the SLB.
230 231
	 * Almost all 32 and 64bit PowerPC executables are linked at
	 * 0x10000000 so it makes sense to preload this segment.
L
Linus Torvalds 已提交
232
	 */
233
	exec_base = 0x10000000;
L
Linus Torvalds 已提交
234

235
	if (is_kernel_addr(pc) || is_kernel_addr(stack) ||
236
	    is_kernel_addr(exec_base))
L
Linus Torvalds 已提交
237 238
		return;

239
	slb_allocate(pc);
L
Linus Torvalds 已提交
240

241 242
	if (!esids_match(pc, stack))
		slb_allocate(stack);
L
Linus Torvalds 已提交
243

244 245 246
	if (!esids_match(pc, exec_base) &&
	    !esids_match(stack, exec_base))
		slb_allocate(exec_base);
L
Linus Torvalds 已提交
247 248
}

249 250 251
static inline void patch_slb_encoding(unsigned int *insn_addr,
				      unsigned int immed)
{
252 253 254 255 256 257 258 259 260 261 262 263 264 265 266

	/*
	 * This function patches either an li or a cmpldi instruction with
	 * a new immediate value. This relies on the fact that both li
	 * (which is actually addi) and cmpldi both take a 16-bit immediate
	 * value, and it is situated in the same location in the instruction,
	 * ie. bits 16-31 (Big endian bit order) or the lower 16 bits.
	 * The signedness of the immediate operand differs between the two
	 * instructions however this code is only ever patching a small value,
	 * much less than 1 << 15, so we can get away with it.
	 * To patch the value we read the existing instruction, clear the
	 * immediate value, and or in our new value, then write the instruction
	 * back.
	 */
	unsigned int insn = (*insn_addr & 0xffff0000) | immed;
267
	patch_instruction(insn_addr, insn);
268 269
}

270 271 272 273 274
extern u32 slb_miss_kernel_load_linear[];
extern u32 slb_miss_kernel_load_io[];
extern u32 slb_compare_rr_to_size[];
extern u32 slb_miss_kernel_load_vmemmap[];

275 276 277 278 279 280 281 282 283
void slb_set_size(u16 size)
{
	if (mmu_slb_size == size)
		return;

	mmu_slb_size = size;
	patch_slb_encoding(slb_compare_rr_to_size, mmu_slb_size);
}

L
Linus Torvalds 已提交
284 285
void slb_initialize(void)
{
286
	unsigned long linear_llp, vmalloc_llp, io_llp;
287
	unsigned long lflags, vflags;
288
	static int slb_encoding_inited;
289 290 291
#ifdef CONFIG_SPARSEMEM_VMEMMAP
	unsigned long vmemmap_llp;
#endif
292 293 294

	/* Prepare our SLB miss handler based on our page size */
	linear_llp = mmu_psize_defs[mmu_linear_psize].sllp;
295 296 297
	io_llp = mmu_psize_defs[mmu_io_psize].sllp;
	vmalloc_llp = mmu_psize_defs[mmu_vmalloc_psize].sllp;
	get_paca()->vmalloc_sllp = SLB_VSID_KERNEL | vmalloc_llp;
298 299 300
#ifdef CONFIG_SPARSEMEM_VMEMMAP
	vmemmap_llp = mmu_psize_defs[mmu_vmemmap_psize].sllp;
#endif
301 302 303 304
	if (!slb_encoding_inited) {
		slb_encoding_inited = 1;
		patch_slb_encoding(slb_miss_kernel_load_linear,
				   SLB_VSID_KERNEL | linear_llp);
305 306
		patch_slb_encoding(slb_miss_kernel_load_io,
				   SLB_VSID_KERNEL | io_llp);
307 308
		patch_slb_encoding(slb_compare_rr_to_size,
				   mmu_slb_size);
309

310 311
		pr_devel("SLB: linear  LLP = %04lx\n", linear_llp);
		pr_devel("SLB: io      LLP = %04lx\n", io_llp);
312 313 314 315

#ifdef CONFIG_SPARSEMEM_VMEMMAP
		patch_slb_encoding(slb_miss_kernel_load_vmemmap,
				   SLB_VSID_KERNEL | vmemmap_llp);
316
		pr_devel("SLB: vmemmap LLP = %04lx\n", vmemmap_llp);
317
#endif
318 319
	}

320 321
	get_paca()->stab_rr = SLB_NUM_BOLTED;

322
	lflags = SLB_VSID_KERNEL | linear_llp;
323
	vflags = SLB_VSID_KERNEL | vmalloc_llp;
L
Linus Torvalds 已提交
324

325
	/* Invalidate the entire SLB (even entry 0) & all the ERATS */
326 327 328
	asm volatile("isync":::"memory");
	asm volatile("slbmte  %0,%0"::"r" (0) : "memory");
	asm volatile("isync; slbia; isync":::"memory");
P
Paul Mackerras 已提交
329 330
	create_shadowed_slbe(PAGE_OFFSET, mmu_kernel_ssize, lflags, 0);
	create_shadowed_slbe(VMALLOC_START, mmu_kernel_ssize, vflags, 1);
331

332 333 334 335 336
	/* For the boot cpu, we're running on the stack in init_thread_union,
	 * which is in the first segment of the linear mapping, and also
	 * get_paca()->kstack hasn't been initialized yet.
	 * For secondary cpus, we need to bolt the kernel stack entry now.
	 */
337
	slb_shadow_clear(2);
338 339 340 341
	if (raw_smp_processor_id() != boot_cpuid &&
	    (get_paca()->kstack & slb_esid_mask(mmu_kernel_ssize)) > PAGE_OFFSET)
		create_shadowed_slbe(get_paca()->kstack,
				     mmu_kernel_ssize, lflags, 2);
342

343
	asm volatile("isync":::"memory");
L
Linus Torvalds 已提交
344
}