| Chris Metcalf | 867e359 | 2010-05-28 23:09:12 -0400 | [diff] [blame] | 1 | /* | 
 | 2 |  * Copyright 2010 Tilera Corporation. All Rights Reserved. | 
 | 3 |  * | 
 | 4 |  *   This program is free software; you can redistribute it and/or | 
 | 5 |  *   modify it under the terms of the GNU General Public License | 
 | 6 |  *   as published by the Free Software Foundation, version 2. | 
 | 7 |  * | 
 | 8 |  *   This program is distributed in the hope that it will be useful, but | 
 | 9 |  *   WITHOUT ANY WARRANTY; without even the implied warranty of | 
 | 10 |  *   MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or | 
 | 11 |  *   NON INFRINGEMENT.  See the GNU General Public License for | 
 | 12 |  *   more details. | 
 | 13 |  * | 
 | 14 |  * copy new kernel into place and then call hv_reexec | 
 | 15 |  * | 
 | 16 |  */ | 
 | 17 |  | 
 | 18 | #include <linux/linkage.h> | 
 | 19 | #include <arch/chip.h> | 
 | 20 | #include <asm/page.h> | 
 | 21 | #include <hv/hypervisor.h> | 
 | 22 |  | 
 | 23 | #define ___hvb	MEM_SV_INTRPT + HV_GLUE_START_CPA | 
 | 24 |  | 
 | 25 | #define ___hv_dispatch(f) (___hvb + (HV_DISPATCH_ENTRY_SIZE * f)) | 
 | 26 |  | 
 | 27 | #define ___hv_console_putc ___hv_dispatch(HV_DISPATCH_CONSOLE_PUTC) | 
 | 28 | #define ___hv_halt         ___hv_dispatch(HV_DISPATCH_HALT) | 
 | 29 | #define ___hv_reexec       ___hv_dispatch(HV_DISPATCH_REEXEC) | 
 | 30 | #define ___hv_flush_remote ___hv_dispatch(HV_DISPATCH_FLUSH_REMOTE) | 
 | 31 |  | 
 | 32 | #undef RELOCATE_NEW_KERNEL_VERBOSE | 
 | 33 |  | 
 | 34 | STD_ENTRY(relocate_new_kernel) | 
 | 35 |  | 
 | 36 | 	move	r30, r0		/* page list */ | 
 | 37 | 	move	r31, r1		/* address of page we are on */ | 
 | 38 | 	move	r32, r2		/* start address of new kernel */ | 
 | 39 |  | 
 | 40 | 	shri	r1, r1, PAGE_SHIFT | 
 | 41 | 	addi	r1, r1, 1 | 
 | 42 | 	shli	sp, r1, PAGE_SHIFT | 
 | 43 | 	addi	sp, sp, -8 | 
 | 44 | 	/* we now have a stack (whether we need one or not) */ | 
 | 45 |  | 
 | 46 | 	moveli	r40, lo16(___hv_console_putc) | 
 | 47 | 	auli	r40, r40, ha16(___hv_console_putc) | 
 | 48 |  | 
 | 49 | #ifdef RELOCATE_NEW_KERNEL_VERBOSE | 
 | 50 | 	moveli	r0, 'r' | 
 | 51 | 	jalr	r40 | 
 | 52 |  | 
 | 53 | 	moveli	r0, '_' | 
 | 54 | 	jalr	r40 | 
 | 55 |  | 
 | 56 | 	moveli	r0, 'n' | 
 | 57 | 	jalr	r40 | 
 | 58 |  | 
 | 59 | 	moveli	r0, '_' | 
 | 60 | 	jalr	r40 | 
 | 61 |  | 
 | 62 | 	moveli	r0, 'k' | 
 | 63 | 	jalr	r40 | 
 | 64 |  | 
 | 65 | 	moveli	r0, '\n' | 
 | 66 | 	jalr	r40 | 
 | 67 | #endif | 
 | 68 |  | 
 | 69 | 	/* | 
 | 70 | 	 * Throughout this code r30 is pointer to the element of page | 
 | 71 | 	 * list we are working on. | 
 | 72 | 	 * | 
 | 73 | 	 * Normally we get to the next element of the page list by | 
 | 74 | 	 * incrementing r30 by four.  The exception is if the element | 
 | 75 | 	 * on the page list is an IND_INDIRECTION in which case we use | 
 | 76 | 	 * the element with the low bits masked off as the new value | 
 | 77 | 	 * of r30. | 
 | 78 | 	 * | 
 | 79 | 	 * To get this started, we need the value passed to us (which | 
 | 80 | 	 * will always be an IND_INDIRECTION) in memory somewhere with | 
 | 81 | 	 * r30 pointing at it.  To do that, we push the value passed | 
 | 82 | 	 * to us on the stack and make r30 point to it. | 
 | 83 | 	 */ | 
 | 84 |  | 
 | 85 | 	sw	sp, r30 | 
 | 86 | 	move	r30, sp | 
 | 87 | 	addi	sp, sp, -8 | 
 | 88 |  | 
 | 89 | #if CHIP_HAS_CBOX_HOME_MAP() | 
 | 90 | 	/* | 
 | 91 | 	 * On TILEPro, we need to flush all tiles' caches, since we may | 
 | 92 | 	 * have been doing hash-for-home caching there.  Note that we | 
 | 93 | 	 * must do this _after_ we're completely done modifying any memory | 
 | 94 | 	 * other than our output buffer (which we know is locally cached). | 
 | 95 | 	 * We want the caches to be fully clean when we do the reexec, | 
 | 96 | 	 * because the hypervisor is going to do this flush again at that | 
 | 97 | 	 * point, and we don't want that second flush to overwrite any memory. | 
 | 98 | 	 */ | 
 | 99 | 	{ | 
 | 100 | 	 move	r0, zero	 /* cache_pa */ | 
 | 101 | 	 move	r1, zero | 
 | 102 | 	} | 
 | 103 | 	{ | 
 | 104 | 	 auli	r2, zero, ha16(HV_FLUSH_EVICT_L2) /* cache_control */ | 
 | 105 | 	 movei	r3, -1		 /* cache_cpumask; -1 means all client tiles */ | 
 | 106 | 	} | 
 | 107 | 	{ | 
 | 108 | 	 move	r4, zero	 /* tlb_va */ | 
 | 109 | 	 move	r5, zero	 /* tlb_length */ | 
 | 110 | 	} | 
 | 111 | 	{ | 
 | 112 | 	 move	r6, zero	 /* tlb_pgsize */ | 
 | 113 | 	 move	r7, zero	 /* tlb_cpumask */ | 
 | 114 | 	} | 
 | 115 | 	{ | 
 | 116 | 	 move	r8, zero	 /* asids */ | 
 | 117 | 	 moveli	r20, lo16(___hv_flush_remote) | 
 | 118 | 	} | 
 | 119 | 	{ | 
 | 120 | 	 move	r9, zero	 /* asidcount */ | 
 | 121 | 	 auli	r20, r20, ha16(___hv_flush_remote) | 
 | 122 | 	} | 
 | 123 |  | 
 | 124 | 	jalr	r20 | 
 | 125 | #endif | 
 | 126 |  | 
 | 127 | 	/* r33 is destination pointer, default to zero */ | 
 | 128 |  | 
 | 129 | 	moveli	r33, 0 | 
 | 130 |  | 
 | 131 | .Lloop:	lw	r10, r30 | 
 | 132 |  | 
 | 133 | 	andi	r9, r10, 0xf	/* low 4 bits tell us what type it is */ | 
 | 134 | 	xor	r10, r10, r9	/* r10 is now value with low 4 bits stripped */ | 
 | 135 |  | 
 | 136 | 	seqi	r0, r9, 0x1	/* IND_DESTINATION */ | 
 | 137 | 	bzt	r0, .Ltry2 | 
 | 138 |  | 
 | 139 | 	move	r33, r10 | 
 | 140 |  | 
 | 141 | #ifdef RELOCATE_NEW_KERNEL_VERBOSE | 
 | 142 | 	moveli	r0, 'd' | 
 | 143 | 	jalr	r40 | 
 | 144 | #endif | 
 | 145 |  | 
 | 146 | 	addi	r30, r30, 4 | 
 | 147 | 	j	.Lloop | 
 | 148 |  | 
 | 149 | .Ltry2: | 
 | 150 | 	seqi	r0, r9, 0x2	/* IND_INDIRECTION */ | 
 | 151 | 	bzt	r0, .Ltry4 | 
 | 152 |  | 
 | 153 | 	move	r30, r10 | 
 | 154 |  | 
 | 155 | #ifdef RELOCATE_NEW_KERNEL_VERBOSE | 
 | 156 | 	moveli	r0, 'i' | 
 | 157 | 	jalr	r40 | 
 | 158 | #endif | 
 | 159 |  | 
 | 160 | 	j	.Lloop | 
 | 161 |  | 
 | 162 | .Ltry4: | 
 | 163 | 	seqi	r0, r9, 0x4	/* IND_DONE */ | 
 | 164 | 	bzt	r0, .Ltry8 | 
 | 165 |  | 
 | 166 | 	mf | 
 | 167 |  | 
 | 168 | #ifdef RELOCATE_NEW_KERNEL_VERBOSE | 
 | 169 | 	moveli	r0, 'D' | 
 | 170 | 	jalr	r40 | 
 | 171 | 	moveli	r0, '\n' | 
 | 172 | 	jalr	r40 | 
 | 173 | #endif | 
 | 174 |  | 
 | 175 | 	move	r0, r32 | 
 | 176 | 	moveli	r1, 0		/* arg to hv_reexec is 64 bits */ | 
 | 177 |  | 
 | 178 | 	moveli	r41, lo16(___hv_reexec) | 
 | 179 | 	auli	r41, r41, ha16(___hv_reexec) | 
 | 180 |  | 
 | 181 | 	jalr	r41 | 
 | 182 |  | 
 | 183 | 	/* we should not get here */ | 
 | 184 |  | 
 | 185 | 	moveli	r0, '?' | 
 | 186 | 	jalr	r40 | 
 | 187 | 	moveli	r0, '\n' | 
 | 188 | 	jalr	r40 | 
 | 189 |  | 
 | 190 | 	j	.Lhalt | 
 | 191 |  | 
 | 192 | .Ltry8:	seqi	r0, r9, 0x8	/* IND_SOURCE */ | 
 | 193 | 	bz	r0, .Lerr	/* unknown type */ | 
 | 194 |  | 
 | 195 | 	/* copy page at r10 to page at r33 */ | 
 | 196 |  | 
 | 197 | 	move	r11, r33 | 
 | 198 |  | 
 | 199 | 	moveli	r0, lo16(PAGE_SIZE) | 
 | 200 | 	auli	r0, r0, ha16(PAGE_SIZE) | 
 | 201 | 	add	r33, r33, r0 | 
 | 202 |  | 
 | 203 | 	/* copy word at r10 to word at r11 until r11 equals r33 */ | 
 | 204 |  | 
 | 205 | 	/* We know page size must be multiple of 16, so we can unroll | 
 | 206 | 	 * 16 times safely without any edge case checking. | 
 | 207 | 	 * | 
 | 208 | 	 * Issue a flush of the destination every 16 words to avoid | 
 | 209 | 	 * incoherence when starting the new kernel.  (Now this is | 
 | 210 | 	 * just good paranoia because the hv_reexec call will also | 
 | 211 | 	 * take care of this.) | 
 | 212 | 	 */ | 
 | 213 |  | 
 | 214 | 1: | 
 | 215 | 	{ lw	r0, r10; addi	r10, r10, 4 } | 
 | 216 | 	{ sw	r11, r0; addi	r11, r11, 4 } | 
 | 217 | 	{ lw	r0, r10; addi	r10, r10, 4 } | 
 | 218 | 	{ sw	r11, r0; addi	r11, r11, 4 } | 
 | 219 | 	{ lw	r0, r10; addi	r10, r10, 4 } | 
 | 220 | 	{ sw	r11, r0; addi	r11, r11, 4 } | 
 | 221 | 	{ lw	r0, r10; addi	r10, r10, 4 } | 
 | 222 | 	{ sw	r11, r0; addi	r11, r11, 4 } | 
 | 223 | 	{ lw	r0, r10; addi	r10, r10, 4 } | 
 | 224 | 	{ sw	r11, r0; addi	r11, r11, 4 } | 
 | 225 | 	{ lw	r0, r10; addi	r10, r10, 4 } | 
 | 226 | 	{ sw	r11, r0; addi	r11, r11, 4 } | 
 | 227 | 	{ lw	r0, r10; addi	r10, r10, 4 } | 
 | 228 | 	{ sw	r11, r0; addi	r11, r11, 4 } | 
 | 229 | 	{ lw	r0, r10; addi	r10, r10, 4 } | 
 | 230 | 	{ sw	r11, r0; addi	r11, r11, 4 } | 
 | 231 | 	{ lw	r0, r10; addi	r10, r10, 4 } | 
 | 232 | 	{ sw	r11, r0; addi	r11, r11, 4 } | 
 | 233 | 	{ lw	r0, r10; addi	r10, r10, 4 } | 
 | 234 | 	{ sw	r11, r0; addi	r11, r11, 4 } | 
 | 235 | 	{ lw	r0, r10; addi	r10, r10, 4 } | 
 | 236 | 	{ sw	r11, r0; addi	r11, r11, 4 } | 
 | 237 | 	{ lw	r0, r10; addi	r10, r10, 4 } | 
 | 238 | 	{ sw	r11, r0; addi	r11, r11, 4 } | 
 | 239 | 	{ lw	r0, r10; addi	r10, r10, 4 } | 
 | 240 | 	{ sw	r11, r0; addi	r11, r11, 4 } | 
 | 241 | 	{ lw	r0, r10; addi	r10, r10, 4 } | 
 | 242 | 	{ sw	r11, r0; addi	r11, r11, 4 } | 
 | 243 | 	{ lw	r0, r10; addi	r10, r10, 4 } | 
 | 244 | 	{ sw	r11, r0; addi	r11, r11, 4 } | 
 | 245 | 	{ lw	r0, r10; addi	r10, r10, 4 } | 
 | 246 | 	{ sw	r11, r0 } | 
 | 247 | 	{ flush r11    ; addi	r11, r11, 4 } | 
 | 248 |  | 
 | 249 | 	seq	r0, r33, r11 | 
 | 250 | 	bzt	r0, 1b | 
 | 251 |  | 
 | 252 | #ifdef RELOCATE_NEW_KERNEL_VERBOSE | 
 | 253 | 	moveli	r0, 's' | 
 | 254 | 	jalr	r40 | 
 | 255 | #endif | 
 | 256 |  | 
 | 257 | 	addi	r30, r30, 4 | 
 | 258 | 	j	.Lloop | 
 | 259 |  | 
 | 260 |  | 
 | 261 | .Lerr:	moveli	r0, 'e' | 
 | 262 | 	jalr	r40 | 
 | 263 | 	moveli	r0, 'r' | 
 | 264 | 	jalr	r40 | 
 | 265 | 	moveli	r0, 'r' | 
 | 266 | 	jalr	r40 | 
 | 267 | 	moveli	r0, '\n' | 
 | 268 | 	jalr	r40 | 
 | 269 | .Lhalt: | 
 | 270 | 	moveli	r41, lo16(___hv_halt) | 
 | 271 | 	auli	r41, r41, ha16(___hv_halt) | 
 | 272 |  | 
 | 273 | 	jalr	r41 | 
 | 274 | 	STD_ENDPROC(relocate_new_kernel) | 
 | 275 |  | 
 | 276 | 	.section .rodata,"a" | 
 | 277 |  | 
 | 278 | 	.globl relocate_new_kernel_size | 
 | 279 | relocate_new_kernel_size: | 
 | 280 | 	.long .Lend_relocate_new_kernel - relocate_new_kernel |