aboutsummaryrefslogtreecommitdiffstats
path: root/arch/ppc64/kernel
diff options
context:
space:
mode:
Diffstat (limited to 'arch/ppc64/kernel')
-rw-r--r--arch/ppc64/kernel/LparData.c37
-rw-r--r--arch/ppc64/kernel/Makefile7
-rw-r--r--arch/ppc64/kernel/asm-offsets.c3
-rw-r--r--arch/ppc64/kernel/cputable.c40
-rw-r--r--arch/ppc64/kernel/firmware.c47
-rw-r--r--arch/ppc64/kernel/head.S523
-rw-r--r--arch/ppc64/kernel/iSeries_htab.c5
-rw-r--r--arch/ppc64/kernel/iSeries_setup.c30
-rw-r--r--arch/ppc64/kernel/iSeries_vio.c155
-rw-r--r--arch/ppc64/kernel/lmb.c151
-rw-r--r--arch/ppc64/kernel/lparcfg.c7
-rw-r--r--arch/ppc64/kernel/misc.S98
-rw-r--r--arch/ppc64/kernel/of_device.c2
-rw-r--r--arch/ppc64/kernel/pSeries_iommu.c3
-rw-r--r--arch/ppc64/kernel/pSeries_lpar.c4
-rw-r--r--arch/ppc64/kernel/pSeries_setup.c39
-rw-r--r--arch/ppc64/kernel/pSeries_smp.c3
-rw-r--r--arch/ppc64/kernel/pSeries_vio.c273
-rw-r--r--arch/ppc64/kernel/pacaData.c4
-rw-r--r--arch/ppc64/kernel/pmac_setup.c2
-rw-r--r--arch/ppc64/kernel/pmc.c21
-rw-r--r--arch/ppc64/kernel/process.c12
-rw-r--r--arch/ppc64/kernel/prom.c184
-rw-r--r--arch/ppc64/kernel/prom_init.c93
-rw-r--r--arch/ppc64/kernel/rtas_pci.c19
-rw-r--r--arch/ppc64/kernel/setup.c30
-rw-r--r--arch/ppc64/kernel/signal.c5
-rw-r--r--arch/ppc64/kernel/signal32.c5
-rw-r--r--arch/ppc64/kernel/sysfs.c57
-rw-r--r--arch/ppc64/kernel/time.c7
-rw-r--r--arch/ppc64/kernel/vio.c444
31 files changed, 1231 insertions, 1079 deletions
diff --git a/arch/ppc64/kernel/LparData.c b/arch/ppc64/kernel/LparData.c
index 1c11031c838e..0a9c23ca2f0c 100644
--- a/arch/ppc64/kernel/LparData.c
+++ b/arch/ppc64/kernel/LparData.c
@@ -51,6 +51,17 @@ struct HvReleaseData hvReleaseData = {
51 0xf4, 0x4b, 0xf6, 0xf4 }, 51 0xf4, 0x4b, 0xf6, 0xf4 },
52}; 52};
53 53
54/*
55 * The NACA. The first dword of the naca is required by the iSeries
56 * hypervisor to point to itVpdAreas. The hypervisor finds the NACA
57 * through the pointer in hvReleaseData.
58 */
59struct naca_struct naca = {
60 .xItVpdAreas = &itVpdAreas,
61 .xRamDisk = 0,
62 .xRamDiskSize = 0,
63};
64
54extern void system_reset_iSeries(void); 65extern void system_reset_iSeries(void);
55extern void machine_check_iSeries(void); 66extern void machine_check_iSeries(void);
56extern void data_access_iSeries(void); 67extern void data_access_iSeries(void);
@@ -214,29 +225,3 @@ struct ItVpdAreas itVpdAreas = {
214 0,0 225 0,0
215 } 226 }
216}; 227};
217
218struct msChunks msChunks;
219EXPORT_SYMBOL(msChunks);
220
221/* Depending on whether this is called from iSeries or pSeries setup
222 * code, the location of the msChunks struct may or may not have
223 * to be reloc'd, so we force the caller to do that for us by passing
224 * in a pointer to the structure.
225 */
226unsigned long
227msChunks_alloc(unsigned long mem, unsigned long num_chunks, unsigned long chunk_size)
228{
229 unsigned long offset = reloc_offset();
230 struct msChunks *_msChunks = PTRRELOC(&msChunks);
231
232 _msChunks->num_chunks = num_chunks;
233 _msChunks->chunk_size = chunk_size;
234 _msChunks->chunk_shift = __ilog2(chunk_size);
235 _msChunks->chunk_mask = (1UL<<_msChunks->chunk_shift)-1;
236
237 mem = _ALIGN(mem, sizeof(msChunks_entry));
238 _msChunks->abs = (msChunks_entry *)(mem + offset);
239 mem += num_chunks * sizeof(msChunks_entry);
240
241 return mem;
242}
diff --git a/arch/ppc64/kernel/Makefile b/arch/ppc64/kernel/Makefile
index 2ecccb6b4f8c..f4b3bfcc109d 100644
--- a/arch/ppc64/kernel/Makefile
+++ b/arch/ppc64/kernel/Makefile
@@ -11,7 +11,7 @@ obj-y := setup.o entry.o traps.o irq.o idle.o dma.o \
11 udbg.o binfmt_elf32.o sys_ppc32.o ioctl32.o \ 11 udbg.o binfmt_elf32.o sys_ppc32.o ioctl32.o \
12 ptrace32.o signal32.o rtc.o init_task.o \ 12 ptrace32.o signal32.o rtc.o init_task.o \
13 lmb.o cputable.o cpu_setup_power4.o idle_power4.o \ 13 lmb.o cputable.o cpu_setup_power4.o idle_power4.o \
14 iommu.o sysfs.o vdso.o pmc.o 14 iommu.o sysfs.o vdso.o pmc.o firmware.o
15obj-y += vdso32/ vdso64/ 15obj-y += vdso32/ vdso64/
16 16
17obj-$(CONFIG_PPC_OF) += of_device.o 17obj-$(CONFIG_PPC_OF) += of_device.o
@@ -50,7 +50,10 @@ obj-$(CONFIG_LPARCFG) += lparcfg.o
50obj-$(CONFIG_HVC_CONSOLE) += hvconsole.o 50obj-$(CONFIG_HVC_CONSOLE) += hvconsole.o
51obj-$(CONFIG_BOOTX_TEXT) += btext.o 51obj-$(CONFIG_BOOTX_TEXT) += btext.o
52obj-$(CONFIG_HVCS) += hvcserver.o 52obj-$(CONFIG_HVCS) += hvcserver.o
53obj-$(CONFIG_IBMVIO) += vio.o 53
54vio-obj-$(CONFIG_PPC_PSERIES) += pSeries_vio.o
55vio-obj-$(CONFIG_PPC_ISERIES) += iSeries_vio.o
56obj-$(CONFIG_IBMVIO) += vio.o $(vio-obj-y)
54obj-$(CONFIG_XICS) += xics.o 57obj-$(CONFIG_XICS) += xics.o
55obj-$(CONFIG_MPIC) += mpic.o 58obj-$(CONFIG_MPIC) += mpic.o
56 59
diff --git a/arch/ppc64/kernel/asm-offsets.c b/arch/ppc64/kernel/asm-offsets.c
index abb9e5b5da03..17e35d0fed09 100644
--- a/arch/ppc64/kernel/asm-offsets.c
+++ b/arch/ppc64/kernel/asm-offsets.c
@@ -94,7 +94,8 @@ int main(void)
94 DEFINE(PACASLBCACHEPTR, offsetof(struct paca_struct, slb_cache_ptr)); 94 DEFINE(PACASLBCACHEPTR, offsetof(struct paca_struct, slb_cache_ptr));
95 DEFINE(PACACONTEXTID, offsetof(struct paca_struct, context.id)); 95 DEFINE(PACACONTEXTID, offsetof(struct paca_struct, context.id));
96#ifdef CONFIG_HUGETLB_PAGE 96#ifdef CONFIG_HUGETLB_PAGE
97 DEFINE(PACAHTLBSEGS, offsetof(struct paca_struct, context.htlb_segs)); 97 DEFINE(PACALOWHTLBAREAS, offsetof(struct paca_struct, context.low_htlb_areas));
98 DEFINE(PACAHIGHHTLBAREAS, offsetof(struct paca_struct, context.high_htlb_areas));
98#endif /* CONFIG_HUGETLB_PAGE */ 99#endif /* CONFIG_HUGETLB_PAGE */
99 DEFINE(PACADEFAULTDECR, offsetof(struct paca_struct, default_decr)); 100 DEFINE(PACADEFAULTDECR, offsetof(struct paca_struct, default_decr));
100 DEFINE(PACA_EXGEN, offsetof(struct paca_struct, exgen)); 101 DEFINE(PACA_EXGEN, offsetof(struct paca_struct, exgen));
diff --git a/arch/ppc64/kernel/cputable.c b/arch/ppc64/kernel/cputable.c
index 77cec42f9525..4847f2ac8c9f 100644
--- a/arch/ppc64/kernel/cputable.c
+++ b/arch/ppc64/kernel/cputable.c
@@ -5,7 +5,7 @@
5 * 5 *
6 * Modifications for ppc64: 6 * Modifications for ppc64:
7 * Copyright (C) 2003 Dave Engebretsen <engebret@us.ibm.com> 7 * Copyright (C) 2003 Dave Engebretsen <engebret@us.ibm.com>
8 * 8 *
9 * This program is free software; you can redistribute it and/or 9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License 10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version 11 * as published by the Free Software Foundation; either version
@@ -60,7 +60,6 @@ struct cpu_spec cpu_specs[] = {
60 .icache_bsize = 128, 60 .icache_bsize = 128,
61 .dcache_bsize = 128, 61 .dcache_bsize = 128,
62 .cpu_setup = __setup_cpu_power3, 62 .cpu_setup = __setup_cpu_power3,
63 .firmware_features = COMMON_PPC64_FW,
64 }, 63 },
65 { /* Power3+ */ 64 { /* Power3+ */
66 .pvr_mask = 0xffff0000, 65 .pvr_mask = 0xffff0000,
@@ -73,7 +72,6 @@ struct cpu_spec cpu_specs[] = {
73 .icache_bsize = 128, 72 .icache_bsize = 128,
74 .dcache_bsize = 128, 73 .dcache_bsize = 128,
75 .cpu_setup = __setup_cpu_power3, 74 .cpu_setup = __setup_cpu_power3,
76 .firmware_features = COMMON_PPC64_FW,
77 }, 75 },
78 { /* Northstar */ 76 { /* Northstar */
79 .pvr_mask = 0xffff0000, 77 .pvr_mask = 0xffff0000,
@@ -86,7 +84,6 @@ struct cpu_spec cpu_specs[] = {
86 .icache_bsize = 128, 84 .icache_bsize = 128,
87 .dcache_bsize = 128, 85 .dcache_bsize = 128,
88 .cpu_setup = __setup_cpu_power3, 86 .cpu_setup = __setup_cpu_power3,
89 .firmware_features = COMMON_PPC64_FW,
90 }, 87 },
91 { /* Pulsar */ 88 { /* Pulsar */
92 .pvr_mask = 0xffff0000, 89 .pvr_mask = 0xffff0000,
@@ -99,7 +96,6 @@ struct cpu_spec cpu_specs[] = {
99 .icache_bsize = 128, 96 .icache_bsize = 128,
100 .dcache_bsize = 128, 97 .dcache_bsize = 128,
101 .cpu_setup = __setup_cpu_power3, 98 .cpu_setup = __setup_cpu_power3,
102 .firmware_features = COMMON_PPC64_FW,
103 }, 99 },
104 { /* I-star */ 100 { /* I-star */
105 .pvr_mask = 0xffff0000, 101 .pvr_mask = 0xffff0000,
@@ -112,7 +108,6 @@ struct cpu_spec cpu_specs[] = {
112 .icache_bsize = 128, 108 .icache_bsize = 128,
113 .dcache_bsize = 128, 109 .dcache_bsize = 128,
114 .cpu_setup = __setup_cpu_power3, 110 .cpu_setup = __setup_cpu_power3,
115 .firmware_features = COMMON_PPC64_FW,
116 }, 111 },
117 { /* S-star */ 112 { /* S-star */
118 .pvr_mask = 0xffff0000, 113 .pvr_mask = 0xffff0000,
@@ -125,7 +120,6 @@ struct cpu_spec cpu_specs[] = {
125 .icache_bsize = 128, 120 .icache_bsize = 128,
126 .dcache_bsize = 128, 121 .dcache_bsize = 128,
127 .cpu_setup = __setup_cpu_power3, 122 .cpu_setup = __setup_cpu_power3,
128 .firmware_features = COMMON_PPC64_FW,
129 }, 123 },
130 { /* Power4 */ 124 { /* Power4 */
131 .pvr_mask = 0xffff0000, 125 .pvr_mask = 0xffff0000,
@@ -138,7 +132,6 @@ struct cpu_spec cpu_specs[] = {
138 .icache_bsize = 128, 132 .icache_bsize = 128,
139 .dcache_bsize = 128, 133 .dcache_bsize = 128,
140 .cpu_setup = __setup_cpu_power4, 134 .cpu_setup = __setup_cpu_power4,
141 .firmware_features = COMMON_PPC64_FW,
142 }, 135 },
143 { /* Power4+ */ 136 { /* Power4+ */
144 .pvr_mask = 0xffff0000, 137 .pvr_mask = 0xffff0000,
@@ -151,7 +144,6 @@ struct cpu_spec cpu_specs[] = {
151 .icache_bsize = 128, 144 .icache_bsize = 128,
152 .dcache_bsize = 128, 145 .dcache_bsize = 128,
153 .cpu_setup = __setup_cpu_power4, 146 .cpu_setup = __setup_cpu_power4,
154 .firmware_features = COMMON_PPC64_FW,
155 }, 147 },
156 { /* PPC970 */ 148 { /* PPC970 */
157 .pvr_mask = 0xffff0000, 149 .pvr_mask = 0xffff0000,
@@ -166,7 +158,6 @@ struct cpu_spec cpu_specs[] = {
166 .icache_bsize = 128, 158 .icache_bsize = 128,
167 .dcache_bsize = 128, 159 .dcache_bsize = 128,
168 .cpu_setup = __setup_cpu_ppc970, 160 .cpu_setup = __setup_cpu_ppc970,
169 .firmware_features = COMMON_PPC64_FW,
170 }, 161 },
171 { /* PPC970FX */ 162 { /* PPC970FX */
172 .pvr_mask = 0xffff0000, 163 .pvr_mask = 0xffff0000,
@@ -181,7 +172,6 @@ struct cpu_spec cpu_specs[] = {
181 .icache_bsize = 128, 172 .icache_bsize = 128,
182 .dcache_bsize = 128, 173 .dcache_bsize = 128,
183 .cpu_setup = __setup_cpu_ppc970, 174 .cpu_setup = __setup_cpu_ppc970,
184 .firmware_features = COMMON_PPC64_FW,
185 }, 175 },
186 { /* PPC970MP */ 176 { /* PPC970MP */
187 .pvr_mask = 0xffff0000, 177 .pvr_mask = 0xffff0000,
@@ -196,7 +186,6 @@ struct cpu_spec cpu_specs[] = {
196 .icache_bsize = 128, 186 .icache_bsize = 128,
197 .dcache_bsize = 128, 187 .dcache_bsize = 128,
198 .cpu_setup = __setup_cpu_ppc970, 188 .cpu_setup = __setup_cpu_ppc970,
199 .firmware_features = COMMON_PPC64_FW,
200 }, 189 },
201 { /* Power5 */ 190 { /* Power5 */
202 .pvr_mask = 0xffff0000, 191 .pvr_mask = 0xffff0000,
@@ -211,7 +200,6 @@ struct cpu_spec cpu_specs[] = {
211 .icache_bsize = 128, 200 .icache_bsize = 128,
212 .dcache_bsize = 128, 201 .dcache_bsize = 128,
213 .cpu_setup = __setup_cpu_power4, 202 .cpu_setup = __setup_cpu_power4,
214 .firmware_features = COMMON_PPC64_FW,
215 }, 203 },
216 { /* Power5 */ 204 { /* Power5 */
217 .pvr_mask = 0xffff0000, 205 .pvr_mask = 0xffff0000,
@@ -226,7 +214,6 @@ struct cpu_spec cpu_specs[] = {
226 .icache_bsize = 128, 214 .icache_bsize = 128,
227 .dcache_bsize = 128, 215 .dcache_bsize = 128,
228 .cpu_setup = __setup_cpu_power4, 216 .cpu_setup = __setup_cpu_power4,
229 .firmware_features = COMMON_PPC64_FW,
230 }, 217 },
231 { /* BE DD1.x */ 218 { /* BE DD1.x */
232 .pvr_mask = 0xffff0000, 219 .pvr_mask = 0xffff0000,
@@ -241,7 +228,6 @@ struct cpu_spec cpu_specs[] = {
241 .icache_bsize = 128, 228 .icache_bsize = 128,
242 .dcache_bsize = 128, 229 .dcache_bsize = 128,
243 .cpu_setup = __setup_cpu_be, 230 .cpu_setup = __setup_cpu_be,
244 .firmware_features = COMMON_PPC64_FW,
245 }, 231 },
246 { /* default match */ 232 { /* default match */
247 .pvr_mask = 0x00000000, 233 .pvr_mask = 0x00000000,
@@ -254,29 +240,5 @@ struct cpu_spec cpu_specs[] = {
254 .icache_bsize = 128, 240 .icache_bsize = 128,
255 .dcache_bsize = 128, 241 .dcache_bsize = 128,
256 .cpu_setup = __setup_cpu_power4, 242 .cpu_setup = __setup_cpu_power4,
257 .firmware_features = COMMON_PPC64_FW,
258 } 243 }
259}; 244};
260
261firmware_feature_t firmware_features_table[FIRMWARE_MAX_FEATURES] = {
262 {FW_FEATURE_PFT, "hcall-pft"},
263 {FW_FEATURE_TCE, "hcall-tce"},
264 {FW_FEATURE_SPRG0, "hcall-sprg0"},
265 {FW_FEATURE_DABR, "hcall-dabr"},
266 {FW_FEATURE_COPY, "hcall-copy"},
267 {FW_FEATURE_ASR, "hcall-asr"},
268 {FW_FEATURE_DEBUG, "hcall-debug"},
269 {FW_FEATURE_PERF, "hcall-perf"},
270 {FW_FEATURE_DUMP, "hcall-dump"},
271 {FW_FEATURE_INTERRUPT, "hcall-interrupt"},
272 {FW_FEATURE_MIGRATE, "hcall-migrate"},
273 {FW_FEATURE_PERFMON, "hcall-perfmon"},
274 {FW_FEATURE_CRQ, "hcall-crq"},
275 {FW_FEATURE_VIO, "hcall-vio"},
276 {FW_FEATURE_RDMA, "hcall-rdma"},
277 {FW_FEATURE_LLAN, "hcall-lLAN"},
278 {FW_FEATURE_BULK, "hcall-bulk"},
279 {FW_FEATURE_XDABR, "hcall-xdabr"},
280 {FW_FEATURE_MULTITCE, "hcall-multi-tce"},
281 {FW_FEATURE_SPLPAR, "hcall-splpar"},
282};
diff --git a/arch/ppc64/kernel/firmware.c b/arch/ppc64/kernel/firmware.c
new file mode 100644
index 000000000000..d8432c0fb27d
--- /dev/null
+++ b/arch/ppc64/kernel/firmware.c
@@ -0,0 +1,47 @@
1/*
2 * arch/ppc64/kernel/firmware.c
3 *
4 * Extracted from cputable.c
5 *
6 * Copyright (C) 2001 Ben. Herrenschmidt (benh@kernel.crashing.org)
7 *
8 * Modifications for ppc64:
9 * Copyright (C) 2003 Dave Engebretsen <engebret@us.ibm.com>
10 * Copyright (C) 2005 Stephen Rothwell, IBM Corporation
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.
16 */
17
18#include <linux/config.h>
19
20#include <asm/firmware.h>
21
22unsigned long ppc64_firmware_features;
23
24#ifdef CONFIG_PPC_PSERIES
25firmware_feature_t firmware_features_table[FIRMWARE_MAX_FEATURES] = {
26 {FW_FEATURE_PFT, "hcall-pft"},
27 {FW_FEATURE_TCE, "hcall-tce"},
28 {FW_FEATURE_SPRG0, "hcall-sprg0"},
29 {FW_FEATURE_DABR, "hcall-dabr"},
30 {FW_FEATURE_COPY, "hcall-copy"},
31 {FW_FEATURE_ASR, "hcall-asr"},
32 {FW_FEATURE_DEBUG, "hcall-debug"},
33 {FW_FEATURE_PERF, "hcall-perf"},
34 {FW_FEATURE_DUMP, "hcall-dump"},
35 {FW_FEATURE_INTERRUPT, "hcall-interrupt"},
36 {FW_FEATURE_MIGRATE, "hcall-migrate"},
37 {FW_FEATURE_PERFMON, "hcall-perfmon"},
38 {FW_FEATURE_CRQ, "hcall-crq"},
39 {FW_FEATURE_VIO, "hcall-vio"},
40 {FW_FEATURE_RDMA, "hcall-rdma"},
41 {FW_FEATURE_LLAN, "hcall-lLAN"},
42 {FW_FEATURE_BULK, "hcall-bulk"},
43 {FW_FEATURE_XDABR, "hcall-xdabr"},
44 {FW_FEATURE_MULTITCE, "hcall-multi-tce"},
45 {FW_FEATURE_SPLPAR, "hcall-splpar"},
46};
47#endif
diff --git a/arch/ppc64/kernel/head.S b/arch/ppc64/kernel/head.S
index accaa052d31f..036959775623 100644
--- a/arch/ppc64/kernel/head.S
+++ b/arch/ppc64/kernel/head.S
@@ -23,14 +23,11 @@
23 * 2 of the License, or (at your option) any later version. 23 * 2 of the License, or (at your option) any later version.
24 */ 24 */
25 25
26#define SECONDARY_PROCESSORS
27
28#include <linux/config.h> 26#include <linux/config.h>
29#include <linux/threads.h> 27#include <linux/threads.h>
30#include <asm/processor.h> 28#include <asm/processor.h>
31#include <asm/page.h> 29#include <asm/page.h>
32#include <asm/mmu.h> 30#include <asm/mmu.h>
33#include <asm/naca.h>
34#include <asm/systemcfg.h> 31#include <asm/systemcfg.h>
35#include <asm/ppc_asm.h> 32#include <asm/ppc_asm.h>
36#include <asm/offsets.h> 33#include <asm/offsets.h>
@@ -45,18 +42,13 @@
45#endif 42#endif
46 43
47/* 44/*
48 * hcall interface to pSeries LPAR
49 */
50#define H_SET_ASR 0x30
51
52/*
53 * We layout physical memory as follows: 45 * We layout physical memory as follows:
54 * 0x0000 - 0x00ff : Secondary processor spin code 46 * 0x0000 - 0x00ff : Secondary processor spin code
55 * 0x0100 - 0x2fff : pSeries Interrupt prologs 47 * 0x0100 - 0x2fff : pSeries Interrupt prologs
56 * 0x3000 - 0x3fff : Interrupt support 48 * 0x3000 - 0x5fff : interrupt support, iSeries and common interrupt prologs
57 * 0x4000 - 0x4fff : NACA 49 * 0x6000 - 0x6fff : Initial (CPU0) segment table
58 * 0x6000 : iSeries and common interrupt prologs 50 * 0x7000 - 0x7fff : FWNMI data area
59 * 0x9000 - 0x9fff : Initial segment table 51 * 0x8000 - : Early init and support code
60 */ 52 */
61 53
62/* 54/*
@@ -94,6 +86,7 @@ END_FTR_SECTION(0, 1)
94 86
95 /* Catch branch to 0 in real mode */ 87 /* Catch branch to 0 in real mode */
96 trap 88 trap
89
97#ifdef CONFIG_PPC_ISERIES 90#ifdef CONFIG_PPC_ISERIES
98 /* 91 /*
99 * At offset 0x20, there is a pointer to iSeries LPAR data. 92 * At offset 0x20, there is a pointer to iSeries LPAR data.
@@ -103,12 +96,12 @@ END_FTR_SECTION(0, 1)
103 .llong hvReleaseData-KERNELBASE 96 .llong hvReleaseData-KERNELBASE
104 97
105 /* 98 /*
106 * At offset 0x28 and 0x30 are offsets to the msChunks 99 * At offset 0x28 and 0x30 are offsets to the mschunks_map
107 * array (used by the iSeries LPAR debugger to do translation 100 * array (used by the iSeries LPAR debugger to do translation
108 * between physical addresses and absolute addresses) and 101 * between physical addresses and absolute addresses) and
109 * to the pidhash table (also used by the debugger) 102 * to the pidhash table (also used by the debugger)
110 */ 103 */
111 .llong msChunks-KERNELBASE 104 .llong mschunks_map-KERNELBASE
112 .llong 0 /* pidhash-KERNELBASE SFRXXX */ 105 .llong 0 /* pidhash-KERNELBASE SFRXXX */
113 106
114 /* Offset 0x38 - Pointer to start of embedded System.map */ 107 /* Offset 0x38 - Pointer to start of embedded System.map */
@@ -120,7 +113,7 @@ embedded_sysmap_start:
120embedded_sysmap_end: 113embedded_sysmap_end:
121 .llong 0 114 .llong 0
122 115
123#else /* CONFIG_PPC_ISERIES */ 116#endif /* CONFIG_PPC_ISERIES */
124 117
125 /* Secondary processors spin on this value until it goes to 1. */ 118 /* Secondary processors spin on this value until it goes to 1. */
126 .globl __secondary_hold_spinloop 119 .globl __secondary_hold_spinloop
@@ -155,7 +148,7 @@ _GLOBAL(__secondary_hold)
155 std r24,__secondary_hold_acknowledge@l(0) 148 std r24,__secondary_hold_acknowledge@l(0)
156 sync 149 sync
157 150
158 /* All secondary cpu's wait here until told to start. */ 151 /* All secondary cpus wait here until told to start. */
159100: ld r4,__secondary_hold_spinloop@l(0) 152100: ld r4,__secondary_hold_spinloop@l(0)
160 cmpdi 0,r4,1 153 cmpdi 0,r4,1
161 bne 100b 154 bne 100b
@@ -170,7 +163,6 @@ _GLOBAL(__secondary_hold)
170 BUG_OPCODE 163 BUG_OPCODE
171#endif 164#endif
172#endif 165#endif
173#endif
174 166
175/* This value is used to mark exception frames on the stack. */ 167/* This value is used to mark exception frames on the stack. */
176 .section ".toc","aw" 168 .section ".toc","aw"
@@ -502,33 +494,37 @@ system_call_pSeries:
502 STD_EXCEPTION_PSERIES(0x1300, instruction_breakpoint) 494 STD_EXCEPTION_PSERIES(0x1300, instruction_breakpoint)
503 STD_EXCEPTION_PSERIES(0x1700, altivec_assist) 495 STD_EXCEPTION_PSERIES(0x1700, altivec_assist)
504 496
497 . = 0x3000
498
499/*** pSeries interrupt support ***/
500
505 /* moved from 0xf00 */ 501 /* moved from 0xf00 */
506 STD_EXCEPTION_PSERIES(0x3000, performance_monitor) 502 STD_EXCEPTION_PSERIES(., performance_monitor)
507 503
508 . = 0x3100 504 .align 7
509_GLOBAL(do_stab_bolted_pSeries) 505_GLOBAL(do_stab_bolted_pSeries)
510 mtcrf 0x80,r12 506 mtcrf 0x80,r12
511 mfspr r12,SPRG2 507 mfspr r12,SPRG2
512 EXCEPTION_PROLOG_PSERIES(PACA_EXSLB, .do_stab_bolted) 508 EXCEPTION_PROLOG_PSERIES(PACA_EXSLB, .do_stab_bolted)
513 509
514 510/*
515 /* Space for the naca. Architected to be located at real address 511 * Vectors for the FWNMI option. Share common code.
516 * NACA_PHYS_ADDR. Various tools rely on this location being fixed. 512 */
517 * The first dword of the naca is required by iSeries LPAR to 513 .globl system_reset_fwnmi
518 * point to itVpdAreas. On pSeries native, this value is not used. 514system_reset_fwnmi:
519 */ 515 HMT_MEDIUM
520 . = NACA_PHYS_ADDR 516 mtspr SPRG1,r13 /* save r13 */
521 .globl __end_interrupts 517 RUNLATCH_ON(r13)
522__end_interrupts: 518 EXCEPTION_PROLOG_PSERIES(PACA_EXGEN, system_reset_common)
523#ifdef CONFIG_PPC_ISERIES
524 .globl naca
525naca:
526 .llong itVpdAreas
527 .llong 0 /* xRamDisk */
528 .llong 0 /* xRamDiskSize */
529 519
530 . = 0x6100 520 .globl machine_check_fwnmi
521machine_check_fwnmi:
522 HMT_MEDIUM
523 mtspr SPRG1,r13 /* save r13 */
524 RUNLATCH_ON(r13)
525 EXCEPTION_PROLOG_PSERIES(PACA_EXMC, machine_check_common)
531 526
527#ifdef CONFIG_PPC_ISERIES
532/*** ISeries-LPAR interrupt handlers ***/ 528/*** ISeries-LPAR interrupt handlers ***/
533 529
534 STD_EXCEPTION_ISERIES(0x200, machine_check, PACA_EXMC) 530 STD_EXCEPTION_ISERIES(0x200, machine_check, PACA_EXMC)
@@ -626,9 +622,7 @@ system_reset_iSeries:
626 622
627 cmpwi 0,r23,0 623 cmpwi 0,r23,0
628 beq iSeries_secondary_smp_loop /* Loop until told to go */ 624 beq iSeries_secondary_smp_loop /* Loop until told to go */
629#ifdef SECONDARY_PROCESSORS
630 bne .__secondary_start /* Loop until told to go */ 625 bne .__secondary_start /* Loop until told to go */
631#endif
632iSeries_secondary_smp_loop: 626iSeries_secondary_smp_loop:
633 /* Let the Hypervisor know we are alive */ 627 /* Let the Hypervisor know we are alive */
634 /* 8002 is a call to HvCallCfg::getLps, a harmless Hypervisor function */ 628 /* 8002 is a call to HvCallCfg::getLps, a harmless Hypervisor function */
@@ -671,51 +665,8 @@ hardware_interrupt_iSeries_masked:
671 ld r13,PACA_EXGEN+EX_R13(r13) 665 ld r13,PACA_EXGEN+EX_R13(r13)
672 rfid 666 rfid
673 b . /* prevent speculative execution */ 667 b . /* prevent speculative execution */
674#endif
675
676/*
677 * Data area reserved for FWNMI option.
678 */
679 .= 0x7000
680 .globl fwnmi_data_area
681fwnmi_data_area:
682
683#ifdef CONFIG_PPC_ISERIES
684 . = LPARMAP_PHYS
685#include "lparmap.s"
686#endif /* CONFIG_PPC_ISERIES */ 668#endif /* CONFIG_PPC_ISERIES */
687 669
688/*
689 * Vectors for the FWNMI option. Share common code.
690 */
691 . = 0x8000
692 .globl system_reset_fwnmi
693system_reset_fwnmi:
694 HMT_MEDIUM
695 mtspr SPRG1,r13 /* save r13 */
696 RUNLATCH_ON(r13)
697 EXCEPTION_PROLOG_PSERIES(PACA_EXGEN, system_reset_common)
698 .globl machine_check_fwnmi
699machine_check_fwnmi:
700 HMT_MEDIUM
701 mtspr SPRG1,r13 /* save r13 */
702 RUNLATCH_ON(r13)
703 EXCEPTION_PROLOG_PSERIES(PACA_EXMC, machine_check_common)
704
705 /*
706 * Space for the initial segment table
707 * For LPAR, the hypervisor must fill in at least one entry
708 * before we get control (with relocate on)
709 */
710 . = STAB0_PHYS_ADDR
711 .globl __start_stab
712__start_stab:
713
714 . = (STAB0_PHYS_ADDR + PAGE_SIZE)
715 .globl __end_stab
716__end_stab:
717
718
719/*** Common interrupt handlers ***/ 670/*** Common interrupt handlers ***/
720 671
721 STD_EXCEPTION_COMMON(0x100, system_reset, .system_reset_exception) 672 STD_EXCEPTION_COMMON(0x100, system_reset, .system_reset_exception)
@@ -752,8 +703,8 @@ machine_check_common:
752 * R9 contains the saved CR, r13 points to the paca, 703 * R9 contains the saved CR, r13 points to the paca,
753 * r10 contains the (bad) kernel stack pointer, 704 * r10 contains the (bad) kernel stack pointer,
754 * r11 and r12 contain the saved SRR0 and SRR1. 705 * r11 and r12 contain the saved SRR0 and SRR1.
755 * We switch to using the paca guard page as an emergency stack, 706 * We switch to using an emergency stack, save the registers there,
756 * save the registers there, and call kernel_bad_stack(), which panics. 707 * and call kernel_bad_stack(), which panics.
757 */ 708 */
758bad_stack: 709bad_stack:
759 ld r1,PACAEMERGSP(r13) 710 ld r1,PACAEMERGSP(r13)
@@ -906,6 +857,62 @@ fp_unavailable_common:
906 bl .kernel_fp_unavailable_exception 857 bl .kernel_fp_unavailable_exception
907 BUG_OPCODE 858 BUG_OPCODE
908 859
860/*
861 * load_up_fpu(unused, unused, tsk)
862 * Disable FP for the task which had the FPU previously,
863 * and save its floating-point registers in its thread_struct.
864 * Enables the FPU for use in the kernel on return.
865 * On SMP we know the fpu is free, since we give it up every
866 * switch (ie, no lazy save of the FP registers).
867 * On entry: r13 == 'current' && last_task_used_math != 'current'
868 */
869_STATIC(load_up_fpu)
870 mfmsr r5 /* grab the current MSR */
871 ori r5,r5,MSR_FP
872 mtmsrd r5 /* enable use of fpu now */
873 isync
874/*
875 * For SMP, we don't do lazy FPU switching because it just gets too
876 * horrendously complex, especially when a task switches from one CPU
877 * to another. Instead we call giveup_fpu in switch_to.
878 *
879 */
880#ifndef CONFIG_SMP
881 ld r3,last_task_used_math@got(r2)
882 ld r4,0(r3)
883 cmpdi 0,r4,0
884 beq 1f
885 /* Save FP state to last_task_used_math's THREAD struct */
886 addi r4,r4,THREAD
887 SAVE_32FPRS(0, r4)
888 mffs fr0
889 stfd fr0,THREAD_FPSCR(r4)
890 /* Disable FP for last_task_used_math */
891 ld r5,PT_REGS(r4)
892 ld r4,_MSR-STACK_FRAME_OVERHEAD(r5)
893 li r6,MSR_FP|MSR_FE0|MSR_FE1
894 andc r4,r4,r6
895 std r4,_MSR-STACK_FRAME_OVERHEAD(r5)
8961:
897#endif /* CONFIG_SMP */
898 /* enable use of FP after return */
899 ld r4,PACACURRENT(r13)
900 addi r5,r4,THREAD /* Get THREAD */
901 ld r4,THREAD_FPEXC_MODE(r5)
902 ori r12,r12,MSR_FP
903 or r12,r12,r4
904 std r12,_MSR(r1)
905 lfd fr0,THREAD_FPSCR(r5)
906 mtfsf 0xff,fr0
907 REST_32FPRS(0, r5)
908#ifndef CONFIG_SMP
909 /* Update last_task_used_math to 'current' */
910 subi r4,r5,THREAD /* Back to 'current' */
911 std r4,0(r3)
912#endif /* CONFIG_SMP */
913 /* restore registers and return */
914 b fast_exception_return
915
909 .align 7 916 .align 7
910 .globl altivec_unavailable_common 917 .globl altivec_unavailable_common
911altivec_unavailable_common: 918altivec_unavailable_common:
@@ -921,6 +928,80 @@ END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
921 bl .altivec_unavailable_exception 928 bl .altivec_unavailable_exception
922 b .ret_from_except 929 b .ret_from_except
923 930
931#ifdef CONFIG_ALTIVEC
932/*
933 * load_up_altivec(unused, unused, tsk)
934 * Disable VMX for the task which had it previously,
935 * and save its vector registers in its thread_struct.
936 * Enables the VMX for use in the kernel on return.
937 * On SMP we know the VMX is free, since we give it up every
938 * switch (ie, no lazy save of the vector registers).
939 * On entry: r13 == 'current' && last_task_used_altivec != 'current'
940 */
941_STATIC(load_up_altivec)
942 mfmsr r5 /* grab the current MSR */
943 oris r5,r5,MSR_VEC@h
944 mtmsrd r5 /* enable use of VMX now */
945 isync
946
947/*
948 * For SMP, we don't do lazy VMX switching because it just gets too
949 * horrendously complex, especially when a task switches from one CPU
950 * to another. Instead we call giveup_altvec in switch_to.
951 * VRSAVE isn't dealt with here, that is done in the normal context
952 * switch code. Note that we could rely on vrsave value to eventually
953 * avoid saving all of the VREGs here...
954 */
955#ifndef CONFIG_SMP
956 ld r3,last_task_used_altivec@got(r2)
957 ld r4,0(r3)
958 cmpdi 0,r4,0
959 beq 1f
960 /* Save VMX state to last_task_used_altivec's THREAD struct */
961 addi r4,r4,THREAD
962 SAVE_32VRS(0,r5,r4)
963 mfvscr vr0
964 li r10,THREAD_VSCR
965 stvx vr0,r10,r4
966 /* Disable VMX for last_task_used_altivec */
967 ld r5,PT_REGS(r4)
968 ld r4,_MSR-STACK_FRAME_OVERHEAD(r5)
969 lis r6,MSR_VEC@h
970 andc r4,r4,r6
971 std r4,_MSR-STACK_FRAME_OVERHEAD(r5)
9721:
973#endif /* CONFIG_SMP */
974 /* Hack: if we get an altivec unavailable trap with VRSAVE
975 * set to all zeros, we assume this is a broken application
976 * that fails to set it properly, and thus we switch it to
977 * all 1's
978 */
979 mfspr r4,SPRN_VRSAVE
980 cmpdi 0,r4,0
981 bne+ 1f
982 li r4,-1
983 mtspr SPRN_VRSAVE,r4
9841:
985 /* enable use of VMX after return */
986 ld r4,PACACURRENT(r13)
987 addi r5,r4,THREAD /* Get THREAD */
988 oris r12,r12,MSR_VEC@h
989 std r12,_MSR(r1)
990 li r4,1
991 li r10,THREAD_VSCR
992 stw r4,THREAD_USED_VR(r5)
993 lvx vr0,r10,r5
994 mtvscr vr0
995 REST_32VRS(0,r4,r5)
996#ifndef CONFIG_SMP
997 /* Update last_task_used_math to 'current' */
998 subi r4,r5,THREAD /* Back to 'current' */
999 std r4,0(r3)
1000#endif /* CONFIG_SMP */
1001 /* restore registers and return */
1002 b fast_exception_return
1003#endif /* CONFIG_ALTIVEC */
1004
924/* 1005/*
925 * Hash table stuff 1006 * Hash table stuff
926 */ 1007 */
@@ -1167,6 +1248,42 @@ unrecov_slb:
1167 bl .unrecoverable_exception 1248 bl .unrecoverable_exception
1168 b 1b 1249 b 1b
1169 1250
1251/*
1252 * Space for CPU0's segment table.
1253 *
1254 * On iSeries, the hypervisor must fill in at least one entry before
1255 * we get control (with relocate on). The address is give to the hv
1256 * as a page number (see xLparMap in LparData.c), so this must be at a
1257 * fixed address (the linker can't compute (u64)&initial_stab >>
1258 * PAGE_SHIFT).
1259 */
1260 . = STAB0_PHYS_ADDR /* 0x6000 */
1261 .globl initial_stab
1262initial_stab:
1263 .space 4096
1264
1265/*
1266 * Data area reserved for FWNMI option.
1267 * This address (0x7000) is fixed by the RPA.
1268 */
1269 .= 0x7000
1270 .globl fwnmi_data_area
1271fwnmi_data_area:
1272
1273 /* iSeries does not use the FWNMI stuff, so it is safe to put
1274 * this here, even if we later allow kernels that will boot on
1275 * both pSeries and iSeries */
1276#ifdef CONFIG_PPC_ISERIES
1277 . = LPARMAP_PHYS
1278#include "lparmap.s"
1279/*
1280 * This ".text" is here for old compilers that generate a trailing
1281 * .note section when compiling .c files to .s
1282 */
1283 .text
1284#endif /* CONFIG_PPC_ISERIES */
1285
1286 . = 0x8000
1170 1287
1171/* 1288/*
1172 * On pSeries, secondary processors spin in the following code. 1289 * On pSeries, secondary processors spin in the following code.
@@ -1200,7 +1317,7 @@ _GLOBAL(pSeries_secondary_smp_init)
1200 b .kexec_wait /* next kernel might do better */ 1317 b .kexec_wait /* next kernel might do better */
1201 1318
12022: mtspr SPRG3,r13 /* Save vaddr of paca in SPRG3 */ 13192: mtspr SPRG3,r13 /* Save vaddr of paca in SPRG3 */
1203 /* From now on, r24 is expected to be logica cpuid */ 1320 /* From now on, r24 is expected to be logical cpuid */
1204 mr r24,r5 1321 mr r24,r5
12053: HMT_LOW 13223: HMT_LOW
1206 lbz r23,PACAPROCSTART(r13) /* Test if this processor should */ 1323 lbz r23,PACAPROCSTART(r13) /* Test if this processor should */
@@ -1213,10 +1330,8 @@ _GLOBAL(pSeries_secondary_smp_init)
1213 1330
1214 cmpwi 0,r23,0 1331 cmpwi 0,r23,0
1215#ifdef CONFIG_SMP 1332#ifdef CONFIG_SMP
1216#ifdef SECONDARY_PROCESSORS
1217 bne .__secondary_start 1333 bne .__secondary_start
1218#endif 1334#endif
1219#endif
1220 b 3b /* Loop until told to go */ 1335 b 3b /* Loop until told to go */
1221 1336
1222#ifdef CONFIG_PPC_ISERIES 1337#ifdef CONFIG_PPC_ISERIES
@@ -1430,228 +1545,6 @@ _GLOBAL(copy_and_flush)
1430.align 8 1545.align 8
1431copy_to_here: 1546copy_to_here:
1432 1547
1433/*
1434 * load_up_fpu(unused, unused, tsk)
1435 * Disable FP for the task which had the FPU previously,
1436 * and save its floating-point registers in its thread_struct.
1437 * Enables the FPU for use in the kernel on return.
1438 * On SMP we know the fpu is free, since we give it up every
1439 * switch (ie, no lazy save of the FP registers).
1440 * On entry: r13 == 'current' && last_task_used_math != 'current'
1441 */
1442_STATIC(load_up_fpu)
1443 mfmsr r5 /* grab the current MSR */
1444 ori r5,r5,MSR_FP
1445 mtmsrd r5 /* enable use of fpu now */
1446 isync
1447/*
1448 * For SMP, we don't do lazy FPU switching because it just gets too
1449 * horrendously complex, especially when a task switches from one CPU
1450 * to another. Instead we call giveup_fpu in switch_to.
1451 *
1452 */
1453#ifndef CONFIG_SMP
1454 ld r3,last_task_used_math@got(r2)
1455 ld r4,0(r3)
1456 cmpdi 0,r4,0
1457 beq 1f
1458 /* Save FP state to last_task_used_math's THREAD struct */
1459 addi r4,r4,THREAD
1460 SAVE_32FPRS(0, r4)
1461 mffs fr0
1462 stfd fr0,THREAD_FPSCR(r4)
1463 /* Disable FP for last_task_used_math */
1464 ld r5,PT_REGS(r4)
1465 ld r4,_MSR-STACK_FRAME_OVERHEAD(r5)
1466 li r6,MSR_FP|MSR_FE0|MSR_FE1
1467 andc r4,r4,r6
1468 std r4,_MSR-STACK_FRAME_OVERHEAD(r5)
14691:
1470#endif /* CONFIG_SMP */
1471 /* enable use of FP after return */
1472 ld r4,PACACURRENT(r13)
1473 addi r5,r4,THREAD /* Get THREAD */
1474 ld r4,THREAD_FPEXC_MODE(r5)
1475 ori r12,r12,MSR_FP
1476 or r12,r12,r4
1477 std r12,_MSR(r1)
1478 lfd fr0,THREAD_FPSCR(r5)
1479 mtfsf 0xff,fr0
1480 REST_32FPRS(0, r5)
1481#ifndef CONFIG_SMP
1482 /* Update last_task_used_math to 'current' */
1483 subi r4,r5,THREAD /* Back to 'current' */
1484 std r4,0(r3)
1485#endif /* CONFIG_SMP */
1486 /* restore registers and return */
1487 b fast_exception_return
1488
1489/*
1490 * disable_kernel_fp()
1491 * Disable the FPU.
1492 */
1493_GLOBAL(disable_kernel_fp)
1494 mfmsr r3
1495 rldicl r0,r3,(63-MSR_FP_LG),1
1496 rldicl r3,r0,(MSR_FP_LG+1),0
1497 mtmsrd r3 /* disable use of fpu now */
1498 isync
1499 blr
1500
1501/*
1502 * giveup_fpu(tsk)
1503 * Disable FP for the task given as the argument,
1504 * and save the floating-point registers in its thread_struct.
1505 * Enables the FPU for use in the kernel on return.
1506 */
1507_GLOBAL(giveup_fpu)
1508 mfmsr r5
1509 ori r5,r5,MSR_FP
1510 mtmsrd r5 /* enable use of fpu now */
1511 isync
1512 cmpdi 0,r3,0
1513 beqlr- /* if no previous owner, done */
1514 addi r3,r3,THREAD /* want THREAD of task */
1515 ld r5,PT_REGS(r3)
1516 cmpdi 0,r5,0
1517 SAVE_32FPRS(0, r3)
1518 mffs fr0
1519 stfd fr0,THREAD_FPSCR(r3)
1520 beq 1f
1521 ld r4,_MSR-STACK_FRAME_OVERHEAD(r5)
1522 li r3,MSR_FP|MSR_FE0|MSR_FE1
1523 andc r4,r4,r3 /* disable FP for previous task */
1524 std r4,_MSR-STACK_FRAME_OVERHEAD(r5)
15251:
1526#ifndef CONFIG_SMP
1527 li r5,0
1528 ld r4,last_task_used_math@got(r2)
1529 std r5,0(r4)
1530#endif /* CONFIG_SMP */
1531 blr
1532
1533
1534#ifdef CONFIG_ALTIVEC
1535
1536/*
1537 * load_up_altivec(unused, unused, tsk)
1538 * Disable VMX for the task which had it previously,
1539 * and save its vector registers in its thread_struct.
1540 * Enables the VMX for use in the kernel on return.
1541 * On SMP we know the VMX is free, since we give it up every
1542 * switch (ie, no lazy save of the vector registers).
1543 * On entry: r13 == 'current' && last_task_used_altivec != 'current'
1544 */
1545_STATIC(load_up_altivec)
1546 mfmsr r5 /* grab the current MSR */
1547 oris r5,r5,MSR_VEC@h
1548 mtmsrd r5 /* enable use of VMX now */
1549 isync
1550
1551/*
1552 * For SMP, we don't do lazy VMX switching because it just gets too
1553 * horrendously complex, especially when a task switches from one CPU
1554 * to another. Instead we call giveup_altvec in switch_to.
1555 * VRSAVE isn't dealt with here, that is done in the normal context
1556 * switch code. Note that we could rely on vrsave value to eventually
1557 * avoid saving all of the VREGs here...
1558 */
1559#ifndef CONFIG_SMP
1560 ld r3,last_task_used_altivec@got(r2)
1561 ld r4,0(r3)
1562 cmpdi 0,r4,0
1563 beq 1f
1564 /* Save VMX state to last_task_used_altivec's THREAD struct */
1565 addi r4,r4,THREAD
1566 SAVE_32VRS(0,r5,r4)
1567 mfvscr vr0
1568 li r10,THREAD_VSCR
1569 stvx vr0,r10,r4
1570 /* Disable VMX for last_task_used_altivec */
1571 ld r5,PT_REGS(r4)
1572 ld r4,_MSR-STACK_FRAME_OVERHEAD(r5)
1573 lis r6,MSR_VEC@h
1574 andc r4,r4,r6
1575 std r4,_MSR-STACK_FRAME_OVERHEAD(r5)
15761:
1577#endif /* CONFIG_SMP */
1578 /* Hack: if we get an altivec unavailable trap with VRSAVE
1579 * set to all zeros, we assume this is a broken application
1580 * that fails to set it properly, and thus we switch it to
1581 * all 1's
1582 */
1583 mfspr r4,SPRN_VRSAVE
1584 cmpdi 0,r4,0
1585 bne+ 1f
1586 li r4,-1
1587 mtspr SPRN_VRSAVE,r4
15881:
1589 /* enable use of VMX after return */
1590 ld r4,PACACURRENT(r13)
1591 addi r5,r4,THREAD /* Get THREAD */
1592 oris r12,r12,MSR_VEC@h
1593 std r12,_MSR(r1)
1594 li r4,1
1595 li r10,THREAD_VSCR
1596 stw r4,THREAD_USED_VR(r5)
1597 lvx vr0,r10,r5
1598 mtvscr vr0
1599 REST_32VRS(0,r4,r5)
1600#ifndef CONFIG_SMP
1601 /* Update last_task_used_math to 'current' */
1602 subi r4,r5,THREAD /* Back to 'current' */
1603 std r4,0(r3)
1604#endif /* CONFIG_SMP */
1605 /* restore registers and return */
1606 b fast_exception_return
1607
1608/*
1609 * disable_kernel_altivec()
1610 * Disable the VMX.
1611 */
1612_GLOBAL(disable_kernel_altivec)
1613 mfmsr r3
1614 rldicl r0,r3,(63-MSR_VEC_LG),1
1615 rldicl r3,r0,(MSR_VEC_LG+1),0
1616 mtmsrd r3 /* disable use of VMX now */
1617 isync
1618 blr
1619
1620/*
1621 * giveup_altivec(tsk)
1622 * Disable VMX for the task given as the argument,
1623 * and save the vector registers in its thread_struct.
1624 * Enables the VMX for use in the kernel on return.
1625 */
1626_GLOBAL(giveup_altivec)
1627 mfmsr r5
1628 oris r5,r5,MSR_VEC@h
1629 mtmsrd r5 /* enable use of VMX now */
1630 isync
1631 cmpdi 0,r3,0
1632 beqlr- /* if no previous owner, done */
1633 addi r3,r3,THREAD /* want THREAD of task */
1634 ld r5,PT_REGS(r3)
1635 cmpdi 0,r5,0
1636 SAVE_32VRS(0,r4,r3)
1637 mfvscr vr0
1638 li r4,THREAD_VSCR
1639 stvx vr0,r4,r3
1640 beq 1f
1641 ld r4,_MSR-STACK_FRAME_OVERHEAD(r5)
1642 lis r3,MSR_VEC@h
1643 andc r4,r4,r3 /* disable FP for previous task */
1644 std r4,_MSR-STACK_FRAME_OVERHEAD(r5)
16451:
1646#ifndef CONFIG_SMP
1647 li r5,0
1648 ld r4,last_task_used_altivec@got(r2)
1649 std r5,0(r4)
1650#endif /* CONFIG_SMP */
1651 blr
1652
1653#endif /* CONFIG_ALTIVEC */
1654
1655#ifdef CONFIG_SMP 1548#ifdef CONFIG_SMP
1656#ifdef CONFIG_PPC_PMAC 1549#ifdef CONFIG_PPC_PMAC
1657/* 1550/*
@@ -2002,9 +1895,6 @@ _STATIC(start_here_common)
2002 1895
2003 bl .start_kernel 1896 bl .start_kernel
2004 1897
2005_GLOBAL(__setup_cpu_power3)
2006 blr
2007
2008_GLOBAL(hmt_init) 1898_GLOBAL(hmt_init)
2009#ifdef CONFIG_HMT 1899#ifdef CONFIG_HMT
2010 LOADADDR(r5, hmt_thread_data) 1900 LOADADDR(r5, hmt_thread_data)
@@ -2095,20 +1985,19 @@ _GLOBAL(smp_release_cpus)
2095 1985
2096/* 1986/*
2097 * We put a few things here that have to be page-aligned. 1987 * We put a few things here that have to be page-aligned.
2098 * This stuff goes at the beginning of the data segment, 1988 * This stuff goes at the beginning of the bss, which is page-aligned.
2099 * which is page-aligned.
2100 */ 1989 */
2101 .data 1990 .section ".bss"
1991
2102 .align 12 1992 .align 12
2103 .globl sdata 1993
2104sdata:
2105 .globl empty_zero_page 1994 .globl empty_zero_page
2106empty_zero_page: 1995empty_zero_page:
2107 .space 4096 1996 .space PAGE_SIZE
2108 1997
2109 .globl swapper_pg_dir 1998 .globl swapper_pg_dir
2110swapper_pg_dir: 1999swapper_pg_dir:
2111 .space 4096 2000 .space PAGE_SIZE
2112 2001
2113/* 2002/*
2114 * This space gets a copy of optional info passed to us by the bootstrap 2003 * This space gets a copy of optional info passed to us by the bootstrap
diff --git a/arch/ppc64/kernel/iSeries_htab.c b/arch/ppc64/kernel/iSeries_htab.c
index b0250ae4a72a..2192055a90a0 100644
--- a/arch/ppc64/kernel/iSeries_htab.c
+++ b/arch/ppc64/kernel/iSeries_htab.c
@@ -41,6 +41,7 @@ static long iSeries_hpte_insert(unsigned long hpte_group, unsigned long va,
41 unsigned long prpn, unsigned long vflags, 41 unsigned long prpn, unsigned long vflags,
42 unsigned long rflags) 42 unsigned long rflags)
43{ 43{
44 unsigned long arpn;
44 long slot; 45 long slot;
45 hpte_t lhpte; 46 hpte_t lhpte;
46 int secondary = 0; 47 int secondary = 0;
@@ -70,8 +71,10 @@ static long iSeries_hpte_insert(unsigned long hpte_group, unsigned long va,
70 slot &= 0x7fffffffffffffff; 71 slot &= 0x7fffffffffffffff;
71 } 72 }
72 73
74 arpn = phys_to_abs(prpn << PAGE_SHIFT) >> PAGE_SHIFT;
75
73 lhpte.v = (va >> 23) << HPTE_V_AVPN_SHIFT | vflags | HPTE_V_VALID; 76 lhpte.v = (va >> 23) << HPTE_V_AVPN_SHIFT | vflags | HPTE_V_VALID;
74 lhpte.r = (physRpn_to_absRpn(prpn) << HPTE_R_RPN_SHIFT) | rflags; 77 lhpte.r = (arpn << HPTE_R_RPN_SHIFT) | rflags;
75 78
76 /* Now fill in the actual HPTE */ 79 /* Now fill in the actual HPTE */
77 HvCallHpt_addValidate(slot, secondary, &lhpte); 80 HvCallHpt_addValidate(slot, secondary, &lhpte);
diff --git a/arch/ppc64/kernel/iSeries_setup.c b/arch/ppc64/kernel/iSeries_setup.c
index a649edbb23b6..3ffefbbc6623 100644
--- a/arch/ppc64/kernel/iSeries_setup.c
+++ b/arch/ppc64/kernel/iSeries_setup.c
@@ -39,6 +39,7 @@
39#include <asm/cputable.h> 39#include <asm/cputable.h>
40#include <asm/sections.h> 40#include <asm/sections.h>
41#include <asm/iommu.h> 41#include <asm/iommu.h>
42#include <asm/firmware.h>
42 43
43#include <asm/time.h> 44#include <asm/time.h>
44#include "iSeries_setup.h" 45#include "iSeries_setup.h"
@@ -314,6 +315,8 @@ static void __init iSeries_init_early(void)
314 315
315 DBG(" -> iSeries_init_early()\n"); 316 DBG(" -> iSeries_init_early()\n");
316 317
318 ppc64_firmware_features = FW_FEATURE_ISERIES;
319
317 ppcdbg_initialize(); 320 ppcdbg_initialize();
318 321
319#if defined(CONFIG_BLK_DEV_INITRD) 322#if defined(CONFIG_BLK_DEV_INITRD)
@@ -412,6 +415,22 @@ static void __init iSeries_init_early(void)
412 DBG(" <- iSeries_init_early()\n"); 415 DBG(" <- iSeries_init_early()\n");
413} 416}
414 417
418struct mschunks_map mschunks_map = {
419 /* XXX We don't use these, but Piranha might need them. */
420 .chunk_size = MSCHUNKS_CHUNK_SIZE,
421 .chunk_shift = MSCHUNKS_CHUNK_SHIFT,
422 .chunk_mask = MSCHUNKS_OFFSET_MASK,
423};
424EXPORT_SYMBOL(mschunks_map);
425
426void mschunks_alloc(unsigned long num_chunks)
427{
428 klimit = _ALIGN(klimit, sizeof(u32));
429 mschunks_map.mapping = (u32 *)klimit;
430 klimit += num_chunks * sizeof(u32);
431 mschunks_map.num_chunks = num_chunks;
432}
433
415/* 434/*
416 * The iSeries may have very large memories ( > 128 GB ) and a partition 435 * The iSeries may have very large memories ( > 128 GB ) and a partition
417 * may get memory in "chunks" that may be anywhere in the 2**52 real 436 * may get memory in "chunks" that may be anywhere in the 2**52 real
@@ -449,7 +468,7 @@ static void __init build_iSeries_Memory_Map(void)
449 468
450 /* Chunk size on iSeries is 256K bytes */ 469 /* Chunk size on iSeries is 256K bytes */
451 totalChunks = (u32)HvLpConfig_getMsChunks(); 470 totalChunks = (u32)HvLpConfig_getMsChunks();
452 klimit = msChunks_alloc(klimit, totalChunks, 1UL << 18); 471 mschunks_alloc(totalChunks);
453 472
454 /* 473 /*
455 * Get absolute address of our load area 474 * Get absolute address of our load area
@@ -486,7 +505,7 @@ static void __init build_iSeries_Memory_Map(void)
486 printk("Load area size %dK\n", loadAreaSize * 256); 505 printk("Load area size %dK\n", loadAreaSize * 256);
487 506
488 for (nextPhysChunk = 0; nextPhysChunk < loadAreaSize; ++nextPhysChunk) 507 for (nextPhysChunk = 0; nextPhysChunk < loadAreaSize; ++nextPhysChunk)
489 msChunks.abs[nextPhysChunk] = 508 mschunks_map.mapping[nextPhysChunk] =
490 loadAreaFirstChunk + nextPhysChunk; 509 loadAreaFirstChunk + nextPhysChunk;
491 510
492 /* 511 /*
@@ -495,7 +514,7 @@ static void __init build_iSeries_Memory_Map(void)
495 */ 514 */
496 hptFirstChunk = (u32)addr_to_chunk(HvCallHpt_getHptAddress()); 515 hptFirstChunk = (u32)addr_to_chunk(HvCallHpt_getHptAddress());
497 hptSizePages = (u32)HvCallHpt_getHptPages(); 516 hptSizePages = (u32)HvCallHpt_getHptPages();
498 hptSizeChunks = hptSizePages >> (msChunks.chunk_shift - PAGE_SHIFT); 517 hptSizeChunks = hptSizePages >> (MSCHUNKS_CHUNK_SHIFT - PAGE_SHIFT);
499 hptLastChunk = hptFirstChunk + hptSizeChunks - 1; 518 hptLastChunk = hptFirstChunk + hptSizeChunks - 1;
500 519
501 printk("HPT absolute addr = %016lx, size = %dK\n", 520 printk("HPT absolute addr = %016lx, size = %dK\n",
@@ -552,7 +571,8 @@ static void __init build_iSeries_Memory_Map(void)
552 (absChunk > hptLastChunk)) && 571 (absChunk > hptLastChunk)) &&
553 ((absChunk < loadAreaFirstChunk) || 572 ((absChunk < loadAreaFirstChunk) ||
554 (absChunk > loadAreaLastChunk))) { 573 (absChunk > loadAreaLastChunk))) {
555 msChunks.abs[nextPhysChunk] = absChunk; 574 mschunks_map.mapping[nextPhysChunk] =
575 absChunk;
556 ++nextPhysChunk; 576 ++nextPhysChunk;
557 } 577 }
558 } 578 }
@@ -944,6 +964,8 @@ void __init iSeries_early_setup(void)
944 ppc_md.calibrate_decr = iSeries_calibrate_decr; 964 ppc_md.calibrate_decr = iSeries_calibrate_decr;
945 ppc_md.progress = iSeries_progress; 965 ppc_md.progress = iSeries_progress;
946 966
967 /* XXX Implement enable_pmcs for iSeries */
968
947 if (get_paca()->lppaca.shared_proc) { 969 if (get_paca()->lppaca.shared_proc) {
948 ppc_md.idle_loop = iseries_shared_idle; 970 ppc_md.idle_loop = iseries_shared_idle;
949 printk(KERN_INFO "Using shared processor idle loop\n"); 971 printk(KERN_INFO "Using shared processor idle loop\n");
diff --git a/arch/ppc64/kernel/iSeries_vio.c b/arch/ppc64/kernel/iSeries_vio.c
new file mode 100644
index 000000000000..6b754b0c8344
--- /dev/null
+++ b/arch/ppc64/kernel/iSeries_vio.c
@@ -0,0 +1,155 @@
1/*
2 * IBM PowerPC iSeries Virtual I/O Infrastructure Support.
3 *
4 * Copyright (c) 2005 Stephen Rothwell, IBM Corp.
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
10 */
11#include <linux/types.h>
12#include <linux/device.h>
13#include <linux/init.h>
14
15#include <asm/vio.h>
16#include <asm/iommu.h>
17#include <asm/abs_addr.h>
18#include <asm/page.h>
19#include <asm/iSeries/vio.h>
20#include <asm/iSeries/HvTypes.h>
21#include <asm/iSeries/HvLpConfig.h>
22#include <asm/iSeries/HvCallXm.h>
23
24struct device *iSeries_vio_dev = &vio_bus_device.dev;
25EXPORT_SYMBOL(iSeries_vio_dev);
26
27static struct iommu_table veth_iommu_table;
28static struct iommu_table vio_iommu_table;
29
30static void __init iommu_vio_init(void)
31{
32 struct iommu_table *t;
33 struct iommu_table_cb cb;
34 unsigned long cbp;
35 unsigned long itc_entries;
36
37 cb.itc_busno = 255; /* Bus 255 is the virtual bus */
38 cb.itc_virtbus = 0xff; /* Ask for virtual bus */
39
40 cbp = virt_to_abs(&cb);
41 HvCallXm_getTceTableParms(cbp);
42
43 itc_entries = cb.itc_size * PAGE_SIZE / sizeof(union tce_entry);
44 veth_iommu_table.it_size = itc_entries / 2;
45 veth_iommu_table.it_busno = cb.itc_busno;
46 veth_iommu_table.it_offset = cb.itc_offset;
47 veth_iommu_table.it_index = cb.itc_index;
48 veth_iommu_table.it_type = TCE_VB;
49 veth_iommu_table.it_blocksize = 1;
50
51 t = iommu_init_table(&veth_iommu_table);
52
53 if (!t)
54 printk("Virtual Bus VETH TCE table failed.\n");
55
56 vio_iommu_table.it_size = itc_entries - veth_iommu_table.it_size;
57 vio_iommu_table.it_busno = cb.itc_busno;
58 vio_iommu_table.it_offset = cb.itc_offset +
59 veth_iommu_table.it_size;
60 vio_iommu_table.it_index = cb.itc_index;
61 vio_iommu_table.it_type = TCE_VB;
62 vio_iommu_table.it_blocksize = 1;
63
64 t = iommu_init_table(&vio_iommu_table);
65
66 if (!t)
67 printk("Virtual Bus VIO TCE table failed.\n");
68}
69
70/**
71 * vio_register_device_iseries: - Register a new iSeries vio device.
72 * @voidev: The device to register.
73 */
74static struct vio_dev *__init vio_register_device_iseries(char *type,
75 uint32_t unit_num)
76{
77 struct vio_dev *viodev;
78
79 /* allocate a vio_dev for this device */
80 viodev = kmalloc(sizeof(struct vio_dev), GFP_KERNEL);
81 if (!viodev)
82 return NULL;
83 memset(viodev, 0, sizeof(struct vio_dev));
84
85 snprintf(viodev->dev.bus_id, BUS_ID_SIZE, "%s%d", type, unit_num);
86
87 viodev->name = viodev->dev.bus_id;
88 viodev->type = type;
89 viodev->unit_address = unit_num;
90 viodev->iommu_table = &vio_iommu_table;
91 if (vio_register_device(viodev) == NULL) {
92 kfree(viodev);
93 return NULL;
94 }
95 return viodev;
96}
97
98void __init probe_bus_iseries(void)
99{
100 HvLpIndexMap vlan_map;
101 struct vio_dev *viodev;
102 int i;
103
104 /* there is only one of each of these */
105 vio_register_device_iseries("viocons", 0);
106 vio_register_device_iseries("vscsi", 0);
107
108 vlan_map = HvLpConfig_getVirtualLanIndexMap();
109 for (i = 0; i < HVMAXARCHITECTEDVIRTUALLANS; i++) {
110 if ((vlan_map & (0x8000 >> i)) == 0)
111 continue;
112 viodev = vio_register_device_iseries("vlan", i);
113 /* veth is special and has it own iommu_table */
114 viodev->iommu_table = &veth_iommu_table;
115 }
116 for (i = 0; i < HVMAXARCHITECTEDVIRTUALDISKS; i++)
117 vio_register_device_iseries("viodasd", i);
118 for (i = 0; i < HVMAXARCHITECTEDVIRTUALCDROMS; i++)
119 vio_register_device_iseries("viocd", i);
120 for (i = 0; i < HVMAXARCHITECTEDVIRTUALTAPES; i++)
121 vio_register_device_iseries("viotape", i);
122}
123
124/**
125 * vio_match_device_iseries: - Tell if a iSeries VIO device matches a
126 * vio_device_id
127 */
128static int vio_match_device_iseries(const struct vio_device_id *id,
129 const struct vio_dev *dev)
130{
131 return strncmp(dev->type, id->type, strlen(id->type)) == 0;
132}
133
134static struct vio_bus_ops vio_bus_ops_iseries = {
135 .match = vio_match_device_iseries,
136};
137
138/**
139 * vio_bus_init_iseries: - Initialize the iSeries virtual IO bus
140 */
141static int __init vio_bus_init_iseries(void)
142{
143 int err;
144
145 err = vio_bus_init(&vio_bus_ops_iseries);
146 if (err == 0) {
147 iommu_vio_init();
148 vio_bus_device.iommu_table = &vio_iommu_table;
149 iSeries_vio_dev = &vio_bus_device.dev;
150 probe_bus_iseries();
151 }
152 return err;
153}
154
155__initcall(vio_bus_init_iseries);
diff --git a/arch/ppc64/kernel/lmb.c b/arch/ppc64/kernel/lmb.c
index d6c6bd03d2a4..5adaca2ddc9d 100644
--- a/arch/ppc64/kernel/lmb.c
+++ b/arch/ppc64/kernel/lmb.c
@@ -28,33 +28,28 @@ void lmb_dump_all(void)
28{ 28{
29#ifdef DEBUG 29#ifdef DEBUG
30 unsigned long i; 30 unsigned long i;
31 struct lmb *_lmb = &lmb;
32 31
33 udbg_printf("lmb_dump_all:\n"); 32 udbg_printf("lmb_dump_all:\n");
34 udbg_printf(" memory.cnt = 0x%lx\n", 33 udbg_printf(" memory.cnt = 0x%lx\n",
35 _lmb->memory.cnt); 34 lmb.memory.cnt);
36 udbg_printf(" memory.size = 0x%lx\n", 35 udbg_printf(" memory.size = 0x%lx\n",
37 _lmb->memory.size); 36 lmb.memory.size);
38 for (i=0; i < _lmb->memory.cnt ;i++) { 37 for (i=0; i < lmb.memory.cnt ;i++) {
39 udbg_printf(" memory.region[0x%x].base = 0x%lx\n", 38 udbg_printf(" memory.region[0x%x].base = 0x%lx\n",
40 i, _lmb->memory.region[i].base); 39 i, lmb.memory.region[i].base);
41 udbg_printf(" .physbase = 0x%lx\n",
42 _lmb->memory.region[i].physbase);
43 udbg_printf(" .size = 0x%lx\n", 40 udbg_printf(" .size = 0x%lx\n",
44 _lmb->memory.region[i].size); 41 lmb.memory.region[i].size);
45 } 42 }
46 43
47 udbg_printf("\n reserved.cnt = 0x%lx\n", 44 udbg_printf("\n reserved.cnt = 0x%lx\n",
48 _lmb->reserved.cnt); 45 lmb.reserved.cnt);
49 udbg_printf(" reserved.size = 0x%lx\n", 46 udbg_printf(" reserved.size = 0x%lx\n",
50 _lmb->reserved.size); 47 lmb.reserved.size);
51 for (i=0; i < _lmb->reserved.cnt ;i++) { 48 for (i=0; i < lmb.reserved.cnt ;i++) {
52 udbg_printf(" reserved.region[0x%x].base = 0x%lx\n", 49 udbg_printf(" reserved.region[0x%x].base = 0x%lx\n",
53 i, _lmb->reserved.region[i].base); 50 i, lmb.reserved.region[i].base);
54 udbg_printf(" .physbase = 0x%lx\n",
55 _lmb->reserved.region[i].physbase);
56 udbg_printf(" .size = 0x%lx\n", 51 udbg_printf(" .size = 0x%lx\n",
57 _lmb->reserved.region[i].size); 52 lmb.reserved.region[i].size);
58 } 53 }
59#endif /* DEBUG */ 54#endif /* DEBUG */
60} 55}
@@ -98,7 +93,6 @@ lmb_coalesce_regions(struct lmb_region *rgn, unsigned long r1, unsigned long r2)
98 rgn->region[r1].size += rgn->region[r2].size; 93 rgn->region[r1].size += rgn->region[r2].size;
99 for (i=r2; i < rgn->cnt-1; i++) { 94 for (i=r2; i < rgn->cnt-1; i++) {
100 rgn->region[i].base = rgn->region[i+1].base; 95 rgn->region[i].base = rgn->region[i+1].base;
101 rgn->region[i].physbase = rgn->region[i+1].physbase;
102 rgn->region[i].size = rgn->region[i+1].size; 96 rgn->region[i].size = rgn->region[i+1].size;
103 } 97 }
104 rgn->cnt--; 98 rgn->cnt--;
@@ -108,49 +102,29 @@ lmb_coalesce_regions(struct lmb_region *rgn, unsigned long r1, unsigned long r2)
108void __init 102void __init
109lmb_init(void) 103lmb_init(void)
110{ 104{
111 struct lmb *_lmb = &lmb;
112
113 /* Create a dummy zero size LMB which will get coalesced away later. 105 /* Create a dummy zero size LMB which will get coalesced away later.
114 * This simplifies the lmb_add() code below... 106 * This simplifies the lmb_add() code below...
115 */ 107 */
116 _lmb->memory.region[0].base = 0; 108 lmb.memory.region[0].base = 0;
117 _lmb->memory.region[0].size = 0; 109 lmb.memory.region[0].size = 0;
118 _lmb->memory.cnt = 1; 110 lmb.memory.cnt = 1;
119 111
120 /* Ditto. */ 112 /* Ditto. */
121 _lmb->reserved.region[0].base = 0; 113 lmb.reserved.region[0].base = 0;
122 _lmb->reserved.region[0].size = 0; 114 lmb.reserved.region[0].size = 0;
123 _lmb->reserved.cnt = 1; 115 lmb.reserved.cnt = 1;
124} 116}
125 117
126/* This routine called with relocation disabled. */ 118/* This routine called with relocation disabled. */
127void __init 119void __init
128lmb_analyze(void) 120lmb_analyze(void)
129{ 121{
130 unsigned long i; 122 int i;
131 unsigned long mem_size = 0; 123
132 unsigned long size_mask = 0; 124 lmb.memory.size = 0;
133 struct lmb *_lmb = &lmb;
134#ifdef CONFIG_MSCHUNKS
135 unsigned long physbase = 0;
136#endif
137
138 for (i=0; i < _lmb->memory.cnt; i++) {
139 unsigned long lmb_size;
140
141 lmb_size = _lmb->memory.region[i].size;
142
143#ifdef CONFIG_MSCHUNKS
144 _lmb->memory.region[i].physbase = physbase;
145 physbase += lmb_size;
146#else
147 _lmb->memory.region[i].physbase = _lmb->memory.region[i].base;
148#endif
149 mem_size += lmb_size;
150 size_mask |= lmb_size;
151 }
152 125
153 _lmb->memory.size = mem_size; 126 for (i = 0; i < lmb.memory.cnt; i++)
127 lmb.memory.size += lmb.memory.region[i].size;
154} 128}
155 129
156/* This routine called with relocation disabled. */ 130/* This routine called with relocation disabled. */
@@ -168,7 +142,6 @@ lmb_add_region(struct lmb_region *rgn, unsigned long base, unsigned long size)
168 adjacent = lmb_addrs_adjacent(base,size,rgnbase,rgnsize); 142 adjacent = lmb_addrs_adjacent(base,size,rgnbase,rgnsize);
169 if ( adjacent > 0 ) { 143 if ( adjacent > 0 ) {
170 rgn->region[i].base -= size; 144 rgn->region[i].base -= size;
171 rgn->region[i].physbase -= size;
172 rgn->region[i].size += size; 145 rgn->region[i].size += size;
173 coalesced++; 146 coalesced++;
174 break; 147 break;
@@ -195,11 +168,9 @@ lmb_add_region(struct lmb_region *rgn, unsigned long base, unsigned long size)
195 for (i=rgn->cnt-1; i >= 0; i--) { 168 for (i=rgn->cnt-1; i >= 0; i--) {
196 if (base < rgn->region[i].base) { 169 if (base < rgn->region[i].base) {
197 rgn->region[i+1].base = rgn->region[i].base; 170 rgn->region[i+1].base = rgn->region[i].base;
198 rgn->region[i+1].physbase = rgn->region[i].physbase;
199 rgn->region[i+1].size = rgn->region[i].size; 171 rgn->region[i+1].size = rgn->region[i].size;
200 } else { 172 } else {
201 rgn->region[i+1].base = base; 173 rgn->region[i+1].base = base;
202 rgn->region[i+1].physbase = lmb_abs_to_phys(base);
203 rgn->region[i+1].size = size; 174 rgn->region[i+1].size = size;
204 break; 175 break;
205 } 176 }
@@ -213,12 +184,11 @@ lmb_add_region(struct lmb_region *rgn, unsigned long base, unsigned long size)
213long __init 184long __init
214lmb_add(unsigned long base, unsigned long size) 185lmb_add(unsigned long base, unsigned long size)
215{ 186{
216 struct lmb *_lmb = &lmb; 187 struct lmb_region *_rgn = &(lmb.memory);
217 struct lmb_region *_rgn = &(_lmb->memory);
218 188
219 /* On pSeries LPAR systems, the first LMB is our RMO region. */ 189 /* On pSeries LPAR systems, the first LMB is our RMO region. */
220 if ( base == 0 ) 190 if ( base == 0 )
221 _lmb->rmo_size = size; 191 lmb.rmo_size = size;
222 192
223 return lmb_add_region(_rgn, base, size); 193 return lmb_add_region(_rgn, base, size);
224 194
@@ -227,8 +197,7 @@ lmb_add(unsigned long base, unsigned long size)
227long __init 197long __init
228lmb_reserve(unsigned long base, unsigned long size) 198lmb_reserve(unsigned long base, unsigned long size)
229{ 199{
230 struct lmb *_lmb = &lmb; 200 struct lmb_region *_rgn = &(lmb.reserved);
231 struct lmb_region *_rgn = &(_lmb->reserved);
232 201
233 return lmb_add_region(_rgn, base, size); 202 return lmb_add_region(_rgn, base, size);
234} 203}
@@ -260,13 +229,10 @@ lmb_alloc_base(unsigned long size, unsigned long align, unsigned long max_addr)
260{ 229{
261 long i, j; 230 long i, j;
262 unsigned long base = 0; 231 unsigned long base = 0;
263 struct lmb *_lmb = &lmb;
264 struct lmb_region *_mem = &(_lmb->memory);
265 struct lmb_region *_rsv = &(_lmb->reserved);
266 232
267 for (i=_mem->cnt-1; i >= 0; i--) { 233 for (i=lmb.memory.cnt-1; i >= 0; i--) {
268 unsigned long lmbbase = _mem->region[i].base; 234 unsigned long lmbbase = lmb.memory.region[i].base;
269 unsigned long lmbsize = _mem->region[i].size; 235 unsigned long lmbsize = lmb.memory.region[i].size;
270 236
271 if ( max_addr == LMB_ALLOC_ANYWHERE ) 237 if ( max_addr == LMB_ALLOC_ANYWHERE )
272 base = _ALIGN_DOWN(lmbbase+lmbsize-size, align); 238 base = _ALIGN_DOWN(lmbbase+lmbsize-size, align);
@@ -276,8 +242,8 @@ lmb_alloc_base(unsigned long size, unsigned long align, unsigned long max_addr)
276 continue; 242 continue;
277 243
278 while ( (lmbbase <= base) && 244 while ( (lmbbase <= base) &&
279 ((j = lmb_overlaps_region(_rsv,base,size)) >= 0) ) { 245 ((j = lmb_overlaps_region(&lmb.reserved,base,size)) >= 0) ) {
280 base = _ALIGN_DOWN(_rsv->region[j].base-size, align); 246 base = _ALIGN_DOWN(lmb.reserved.region[j].base-size, align);
281 } 247 }
282 248
283 if ( (base != 0) && (lmbbase <= base) ) 249 if ( (base != 0) && (lmbbase <= base) )
@@ -287,62 +253,24 @@ lmb_alloc_base(unsigned long size, unsigned long align, unsigned long max_addr)
287 if ( i < 0 ) 253 if ( i < 0 )
288 return 0; 254 return 0;
289 255
290 lmb_add_region(_rsv, base, size); 256 lmb_add_region(&lmb.reserved, base, size);
291 257
292 return base; 258 return base;
293} 259}
294 260
261/* You must call lmb_analyze() before this. */
295unsigned long __init 262unsigned long __init
296lmb_phys_mem_size(void) 263lmb_phys_mem_size(void)
297{ 264{
298 struct lmb *_lmb = &lmb; 265 return lmb.memory.size;
299#ifdef CONFIG_MSCHUNKS
300 return _lmb->memory.size;
301#else
302 struct lmb_region *_mem = &(_lmb->memory);
303 unsigned long total = 0;
304 int i;
305
306 /* add all physical memory to the bootmem map */
307 for (i=0; i < _mem->cnt; i++)
308 total += _mem->region[i].size;
309 return total;
310#endif /* CONFIG_MSCHUNKS */
311} 266}
312 267
313unsigned long __init 268unsigned long __init
314lmb_end_of_DRAM(void) 269lmb_end_of_DRAM(void)
315{ 270{
316 struct lmb *_lmb = &lmb; 271 int idx = lmb.memory.cnt - 1;
317 struct lmb_region *_mem = &(_lmb->memory);
318 int idx = _mem->cnt - 1;
319
320#ifdef CONFIG_MSCHUNKS
321 return (_mem->region[idx].physbase + _mem->region[idx].size);
322#else
323 return (_mem->region[idx].base + _mem->region[idx].size);
324#endif /* CONFIG_MSCHUNKS */
325
326 return 0;
327}
328
329unsigned long __init
330lmb_abs_to_phys(unsigned long aa)
331{
332 unsigned long i, pa = aa;
333 struct lmb *_lmb = &lmb;
334 struct lmb_region *_mem = &(_lmb->memory);
335
336 for (i=0; i < _mem->cnt; i++) {
337 unsigned long lmbbase = _mem->region[i].base;
338 unsigned long lmbsize = _mem->region[i].size;
339 if ( lmb_addrs_overlap(aa,1,lmbbase,lmbsize) ) {
340 pa = _mem->region[i].physbase + (aa - lmbbase);
341 break;
342 }
343 }
344 272
345 return pa; 273 return (lmb.memory.region[idx].base + lmb.memory.region[idx].size);
346} 274}
347 275
348/* 276/*
@@ -353,20 +281,19 @@ void __init lmb_enforce_memory_limit(void)
353{ 281{
354 extern unsigned long memory_limit; 282 extern unsigned long memory_limit;
355 unsigned long i, limit; 283 unsigned long i, limit;
356 struct lmb_region *mem = &(lmb.memory);
357 284
358 if (! memory_limit) 285 if (! memory_limit)
359 return; 286 return;
360 287
361 limit = memory_limit; 288 limit = memory_limit;
362 for (i = 0; i < mem->cnt; i++) { 289 for (i = 0; i < lmb.memory.cnt; i++) {
363 if (limit > mem->region[i].size) { 290 if (limit > lmb.memory.region[i].size) {
364 limit -= mem->region[i].size; 291 limit -= lmb.memory.region[i].size;
365 continue; 292 continue;
366 } 293 }
367 294
368 mem->region[i].size = limit; 295 lmb.memory.region[i].size = limit;
369 mem->cnt = i + 1; 296 lmb.memory.cnt = i + 1;
370 break; 297 break;
371 } 298 }
372} 299}
diff --git a/arch/ppc64/kernel/lparcfg.c b/arch/ppc64/kernel/lparcfg.c
index 02e96627fa66..edad361a8db0 100644
--- a/arch/ppc64/kernel/lparcfg.c
+++ b/arch/ppc64/kernel/lparcfg.c
@@ -29,7 +29,7 @@
29#include <asm/iSeries/HvLpConfig.h> 29#include <asm/iSeries/HvLpConfig.h>
30#include <asm/lppaca.h> 30#include <asm/lppaca.h>
31#include <asm/hvcall.h> 31#include <asm/hvcall.h>
32#include <asm/cputable.h> 32#include <asm/firmware.h>
33#include <asm/rtas.h> 33#include <asm/rtas.h>
34#include <asm/system.h> 34#include <asm/system.h>
35#include <asm/time.h> 35#include <asm/time.h>
@@ -273,6 +273,7 @@ static void parse_system_parameter_string(struct seq_file *m)
273 if (!workbuffer) { 273 if (!workbuffer) {
274 printk(KERN_ERR "%s %s kmalloc failure at line %d \n", 274 printk(KERN_ERR "%s %s kmalloc failure at line %d \n",
275 __FILE__, __FUNCTION__, __LINE__); 275 __FILE__, __FUNCTION__, __LINE__);
276 kfree(local_buffer);
276 return; 277 return;
277 } 278 }
278#ifdef LPARCFG_DEBUG 279#ifdef LPARCFG_DEBUG
@@ -377,7 +378,7 @@ static int lparcfg_data(struct seq_file *m, void *v)
377 378
378 partition_active_processors = lparcfg_count_active_processors(); 379 partition_active_processors = lparcfg_count_active_processors();
379 380
380 if (cur_cpu_spec->firmware_features & FW_FEATURE_SPLPAR) { 381 if (firmware_has_feature(FW_FEATURE_SPLPAR)) {
381 unsigned long h_entitled, h_unallocated; 382 unsigned long h_entitled, h_unallocated;
382 unsigned long h_aggregation, h_resource; 383 unsigned long h_aggregation, h_resource;
383 unsigned long pool_idle_time, pool_procs; 384 unsigned long pool_idle_time, pool_procs;
@@ -571,7 +572,7 @@ int __init lparcfg_init(void)
571 mode_t mode = S_IRUSR; 572 mode_t mode = S_IRUSR;
572 573
573 /* Allow writing if we have FW_FEATURE_SPLPAR */ 574 /* Allow writing if we have FW_FEATURE_SPLPAR */
574 if (cur_cpu_spec->firmware_features & FW_FEATURE_SPLPAR) { 575 if (firmware_has_feature(FW_FEATURE_SPLPAR)) {
575 lparcfg_fops.write = lparcfg_write; 576 lparcfg_fops.write = lparcfg_write;
576 mode |= S_IWUSR; 577 mode |= S_IWUSR;
577 } 578 }
diff --git a/arch/ppc64/kernel/misc.S b/arch/ppc64/kernel/misc.S
index a05b50b738e9..474df0a862bf 100644
--- a/arch/ppc64/kernel/misc.S
+++ b/arch/ppc64/kernel/misc.S
@@ -680,6 +680,104 @@ _GLOBAL(kernel_thread)
680 ld r30,-16(r1) 680 ld r30,-16(r1)
681 blr 681 blr
682 682
683/*
684 * disable_kernel_fp()
685 * Disable the FPU.
686 */
687_GLOBAL(disable_kernel_fp)
688 mfmsr r3
689 rldicl r0,r3,(63-MSR_FP_LG),1
690 rldicl r3,r0,(MSR_FP_LG+1),0
691 mtmsrd r3 /* disable use of fpu now */
692 isync
693 blr
694
695/*
696 * giveup_fpu(tsk)
697 * Disable FP for the task given as the argument,
698 * and save the floating-point registers in its thread_struct.
699 * Enables the FPU for use in the kernel on return.
700 */
701_GLOBAL(giveup_fpu)
702 mfmsr r5
703 ori r5,r5,MSR_FP
704 mtmsrd r5 /* enable use of fpu now */
705 isync
706 cmpdi 0,r3,0
707 beqlr- /* if no previous owner, done */
708 addi r3,r3,THREAD /* want THREAD of task */
709 ld r5,PT_REGS(r3)
710 cmpdi 0,r5,0
711 SAVE_32FPRS(0, r3)
712 mffs fr0
713 stfd fr0,THREAD_FPSCR(r3)
714 beq 1f
715 ld r4,_MSR-STACK_FRAME_OVERHEAD(r5)
716 li r3,MSR_FP|MSR_FE0|MSR_FE1
717 andc r4,r4,r3 /* disable FP for previous task */
718 std r4,_MSR-STACK_FRAME_OVERHEAD(r5)
7191:
720#ifndef CONFIG_SMP
721 li r5,0
722 ld r4,last_task_used_math@got(r2)
723 std r5,0(r4)
724#endif /* CONFIG_SMP */
725 blr
726
727#ifdef CONFIG_ALTIVEC
728
729#if 0 /* this has no callers for now */
730/*
731 * disable_kernel_altivec()
732 * Disable the VMX.
733 */
734_GLOBAL(disable_kernel_altivec)
735 mfmsr r3
736 rldicl r0,r3,(63-MSR_VEC_LG),1
737 rldicl r3,r0,(MSR_VEC_LG+1),0
738 mtmsrd r3 /* disable use of VMX now */
739 isync
740 blr
741#endif /* 0 */
742
743/*
744 * giveup_altivec(tsk)
745 * Disable VMX for the task given as the argument,
746 * and save the vector registers in its thread_struct.
747 * Enables the VMX for use in the kernel on return.
748 */
749_GLOBAL(giveup_altivec)
750 mfmsr r5
751 oris r5,r5,MSR_VEC@h
752 mtmsrd r5 /* enable use of VMX now */
753 isync
754 cmpdi 0,r3,0
755 beqlr- /* if no previous owner, done */
756 addi r3,r3,THREAD /* want THREAD of task */
757 ld r5,PT_REGS(r3)
758 cmpdi 0,r5,0
759 SAVE_32VRS(0,r4,r3)
760 mfvscr vr0
761 li r4,THREAD_VSCR
762 stvx vr0,r4,r3
763 beq 1f
764 ld r4,_MSR-STACK_FRAME_OVERHEAD(r5)
765 lis r3,MSR_VEC@h
766 andc r4,r4,r3 /* disable FP for previous task */
767 std r4,_MSR-STACK_FRAME_OVERHEAD(r5)
7681:
769#ifndef CONFIG_SMP
770 li r5,0
771 ld r4,last_task_used_altivec@got(r2)
772 std r5,0(r4)
773#endif /* CONFIG_SMP */
774 blr
775
776#endif /* CONFIG_ALTIVEC */
777
778_GLOBAL(__setup_cpu_power3)
779 blr
780
683/* kexec_wait(phys_cpu) 781/* kexec_wait(phys_cpu)
684 * 782 *
685 * wait for the flag to change, indicating this kernel is going away but 783 * wait for the flag to change, indicating this kernel is going away but
diff --git a/arch/ppc64/kernel/of_device.c b/arch/ppc64/kernel/of_device.c
index b80e81984ba8..da580812ddfe 100644
--- a/arch/ppc64/kernel/of_device.c
+++ b/arch/ppc64/kernel/of_device.c
@@ -236,7 +236,6 @@ void of_device_unregister(struct of_device *ofdev)
236struct of_device* of_platform_device_create(struct device_node *np, const char *bus_id) 236struct of_device* of_platform_device_create(struct device_node *np, const char *bus_id)
237{ 237{
238 struct of_device *dev; 238 struct of_device *dev;
239 u32 *reg;
240 239
241 dev = kmalloc(sizeof(*dev), GFP_KERNEL); 240 dev = kmalloc(sizeof(*dev), GFP_KERNEL);
242 if (!dev) 241 if (!dev)
@@ -250,7 +249,6 @@ struct of_device* of_platform_device_create(struct device_node *np, const char *
250 dev->dev.bus = &of_platform_bus_type; 249 dev->dev.bus = &of_platform_bus_type;
251 dev->dev.release = of_release_dev; 250 dev->dev.release = of_release_dev;
252 251
253 reg = (u32 *)get_property(np, "reg", NULL);
254 strlcpy(dev->dev.bus_id, bus_id, BUS_ID_SIZE); 252 strlcpy(dev->dev.bus_id, bus_id, BUS_ID_SIZE);
255 253
256 if (of_device_register(dev) != 0) { 254 if (of_device_register(dev) != 0) {
diff --git a/arch/ppc64/kernel/pSeries_iommu.c b/arch/ppc64/kernel/pSeries_iommu.c
index 69130522a87e..9d5e1e7fc389 100644
--- a/arch/ppc64/kernel/pSeries_iommu.c
+++ b/arch/ppc64/kernel/pSeries_iommu.c
@@ -45,6 +45,7 @@
45#include <asm/plpar_wrappers.h> 45#include <asm/plpar_wrappers.h>
46#include <asm/pSeries_reconfig.h> 46#include <asm/pSeries_reconfig.h>
47#include <asm/systemcfg.h> 47#include <asm/systemcfg.h>
48#include <asm/firmware.h>
48#include "pci.h" 49#include "pci.h"
49 50
50#define DBG(fmt...) 51#define DBG(fmt...)
@@ -546,7 +547,7 @@ void iommu_init_early_pSeries(void)
546 } 547 }
547 548
548 if (systemcfg->platform & PLATFORM_LPAR) { 549 if (systemcfg->platform & PLATFORM_LPAR) {
549 if (cur_cpu_spec->firmware_features & FW_FEATURE_MULTITCE) { 550 if (firmware_has_feature(FW_FEATURE_MULTITCE)) {
550 ppc_md.tce_build = tce_buildmulti_pSeriesLP; 551 ppc_md.tce_build = tce_buildmulti_pSeriesLP;
551 ppc_md.tce_free = tce_freemulti_pSeriesLP; 552 ppc_md.tce_free = tce_freemulti_pSeriesLP;
552 } else { 553 } else {
diff --git a/arch/ppc64/kernel/pSeries_lpar.c b/arch/ppc64/kernel/pSeries_lpar.c
index 74dd144dcce8..0a3ddc9227c5 100644
--- a/arch/ppc64/kernel/pSeries_lpar.c
+++ b/arch/ppc64/kernel/pSeries_lpar.c
@@ -52,7 +52,6 @@ EXPORT_SYMBOL(plpar_hcall_4out);
52EXPORT_SYMBOL(plpar_hcall_norets); 52EXPORT_SYMBOL(plpar_hcall_norets);
53EXPORT_SYMBOL(plpar_hcall_8arg_2ret); 53EXPORT_SYMBOL(plpar_hcall_8arg_2ret);
54 54
55extern void fw_feature_init(void);
56extern void pSeries_find_serial_port(void); 55extern void pSeries_find_serial_port(void);
57 56
58 57
@@ -279,7 +278,6 @@ long pSeries_lpar_hpte_insert(unsigned long hpte_group,
279 unsigned long va, unsigned long prpn, 278 unsigned long va, unsigned long prpn,
280 unsigned long vflags, unsigned long rflags) 279 unsigned long vflags, unsigned long rflags)
281{ 280{
282 unsigned long arpn = physRpn_to_absRpn(prpn);
283 unsigned long lpar_rc; 281 unsigned long lpar_rc;
284 unsigned long flags; 282 unsigned long flags;
285 unsigned long slot; 283 unsigned long slot;
@@ -290,7 +288,7 @@ long pSeries_lpar_hpte_insert(unsigned long hpte_group,
290 if (vflags & HPTE_V_LARGE) 288 if (vflags & HPTE_V_LARGE)
291 hpte_v &= ~(1UL << HPTE_V_AVPN_SHIFT); 289 hpte_v &= ~(1UL << HPTE_V_AVPN_SHIFT);
292 290
293 hpte_r = (arpn << HPTE_R_RPN_SHIFT) | rflags; 291 hpte_r = (prpn << HPTE_R_RPN_SHIFT) | rflags;
294 292
295 /* Now fill in the actual HPTE */ 293 /* Now fill in the actual HPTE */
296 /* Set CEC cookie to 0 */ 294 /* Set CEC cookie to 0 */
diff --git a/arch/ppc64/kernel/pSeries_setup.c b/arch/ppc64/kernel/pSeries_setup.c
index 5bec956e44a0..f0f0630cf07c 100644
--- a/arch/ppc64/kernel/pSeries_setup.c
+++ b/arch/ppc64/kernel/pSeries_setup.c
@@ -60,7 +60,8 @@
60#include <asm/nvram.h> 60#include <asm/nvram.h>
61#include <asm/plpar_wrappers.h> 61#include <asm/plpar_wrappers.h>
62#include <asm/xics.h> 62#include <asm/xics.h>
63#include <asm/cputable.h> 63#include <asm/firmware.h>
64#include <asm/pmc.h>
64 65
65#include "i8259.h" 66#include "i8259.h"
66#include "mpic.h" 67#include "mpic.h"
@@ -187,6 +188,21 @@ static void __init pSeries_setup_mpic(void)
187 " MPIC "); 188 " MPIC ");
188} 189}
189 190
191static void pseries_lpar_enable_pmcs(void)
192{
193 unsigned long set, reset;
194
195 power4_enable_pmcs();
196
197 set = 1UL << 63;
198 reset = 0;
199 plpar_hcall_norets(H_PERFMON, set, reset);
200
201 /* instruct hypervisor to maintain PMCs */
202 if (firmware_has_feature(FW_FEATURE_SPLPAR))
203 get_paca()->lppaca.pmcregs_in_use = 1;
204}
205
190static void __init pSeries_setup_arch(void) 206static void __init pSeries_setup_arch(void)
191{ 207{
192 /* Fixup ppc_md depending on the type of interrupt controller */ 208 /* Fixup ppc_md depending on the type of interrupt controller */
@@ -231,11 +247,9 @@ static void __init pSeries_setup_arch(void)
231 247
232 pSeries_nvram_init(); 248 pSeries_nvram_init();
233 249
234 if (cur_cpu_spec->firmware_features & FW_FEATURE_SPLPAR)
235 vpa_init(boot_cpuid);
236
237 /* Choose an idle loop */ 250 /* Choose an idle loop */
238 if (cur_cpu_spec->firmware_features & FW_FEATURE_SPLPAR) { 251 if (firmware_has_feature(FW_FEATURE_SPLPAR)) {
252 vpa_init(boot_cpuid);
239 if (get_paca()->lppaca.shared_proc) { 253 if (get_paca()->lppaca.shared_proc) {
240 printk(KERN_INFO "Using shared processor idle loop\n"); 254 printk(KERN_INFO "Using shared processor idle loop\n");
241 ppc_md.idle_loop = pseries_shared_idle; 255 ppc_md.idle_loop = pseries_shared_idle;
@@ -247,6 +261,11 @@ static void __init pSeries_setup_arch(void)
247 printk(KERN_INFO "Using default idle loop\n"); 261 printk(KERN_INFO "Using default idle loop\n");
248 ppc_md.idle_loop = default_idle; 262 ppc_md.idle_loop = default_idle;
249 } 263 }
264
265 if (systemcfg->platform & PLATFORM_LPAR)
266 ppc_md.enable_pmcs = pseries_lpar_enable_pmcs;
267 else
268 ppc_md.enable_pmcs = power4_enable_pmcs;
250} 269}
251 270
252static int __init pSeries_init_panel(void) 271static int __init pSeries_init_panel(void)
@@ -260,11 +279,11 @@ static int __init pSeries_init_panel(void)
260arch_initcall(pSeries_init_panel); 279arch_initcall(pSeries_init_panel);
261 280
262 281
263/* Build up the firmware_features bitmask field 282/* Build up the ppc64_firmware_features bitmask field
264 * using contents of device-tree/ibm,hypertas-functions. 283 * using contents of device-tree/ibm,hypertas-functions.
265 * Ultimately this functionality may be moved into prom.c prom_init(). 284 * Ultimately this functionality may be moved into prom.c prom_init().
266 */ 285 */
267void __init fw_feature_init(void) 286static void __init fw_feature_init(void)
268{ 287{
269 struct device_node * dn; 288 struct device_node * dn;
270 char * hypertas; 289 char * hypertas;
@@ -272,7 +291,7 @@ void __init fw_feature_init(void)
272 291
273 DBG(" -> fw_feature_init()\n"); 292 DBG(" -> fw_feature_init()\n");
274 293
275 cur_cpu_spec->firmware_features = 0; 294 ppc64_firmware_features = 0;
276 dn = of_find_node_by_path("/rtas"); 295 dn = of_find_node_by_path("/rtas");
277 if (dn == NULL) { 296 if (dn == NULL) {
278 printk(KERN_ERR "WARNING ! Cannot find RTAS in device-tree !\n"); 297 printk(KERN_ERR "WARNING ! Cannot find RTAS in device-tree !\n");
@@ -288,7 +307,7 @@ void __init fw_feature_init(void)
288 if ((firmware_features_table[i].name) && 307 if ((firmware_features_table[i].name) &&
289 (strcmp(firmware_features_table[i].name,hypertas))==0) { 308 (strcmp(firmware_features_table[i].name,hypertas))==0) {
290 /* we have a match */ 309 /* we have a match */
291 cur_cpu_spec->firmware_features |= 310 ppc64_firmware_features |=
292 (firmware_features_table[i].val); 311 (firmware_features_table[i].val);
293 break; 312 break;
294 } 313 }
@@ -302,7 +321,7 @@ void __init fw_feature_init(void)
302 of_node_put(dn); 321 of_node_put(dn);
303 no_rtas: 322 no_rtas:
304 printk(KERN_INFO "firmware_features = 0x%lx\n", 323 printk(KERN_INFO "firmware_features = 0x%lx\n",
305 cur_cpu_spec->firmware_features); 324 ppc64_firmware_features);
306 325
307 DBG(" <- fw_feature_init()\n"); 326 DBG(" <- fw_feature_init()\n");
308} 327}
diff --git a/arch/ppc64/kernel/pSeries_smp.c b/arch/ppc64/kernel/pSeries_smp.c
index 62c55a123560..79c7f3223665 100644
--- a/arch/ppc64/kernel/pSeries_smp.c
+++ b/arch/ppc64/kernel/pSeries_smp.c
@@ -41,6 +41,7 @@
41#include <asm/machdep.h> 41#include <asm/machdep.h>
42#include <asm/xics.h> 42#include <asm/xics.h>
43#include <asm/cputable.h> 43#include <asm/cputable.h>
44#include <asm/firmware.h>
44#include <asm/system.h> 45#include <asm/system.h>
45#include <asm/rtas.h> 46#include <asm/rtas.h>
46#include <asm/plpar_wrappers.h> 47#include <asm/plpar_wrappers.h>
@@ -326,7 +327,7 @@ static void __devinit smp_xics_setup_cpu(int cpu)
326 if (cpu != boot_cpuid) 327 if (cpu != boot_cpuid)
327 xics_setup_cpu(); 328 xics_setup_cpu();
328 329
329 if (cur_cpu_spec->firmware_features & FW_FEATURE_SPLPAR) 330 if (firmware_has_feature(FW_FEATURE_SPLPAR))
330 vpa_init(cpu); 331 vpa_init(cpu);
331 332
332 cpu_clear(cpu, of_spin_map); 333 cpu_clear(cpu, of_spin_map);
diff --git a/arch/ppc64/kernel/pSeries_vio.c b/arch/ppc64/kernel/pSeries_vio.c
new file mode 100644
index 000000000000..e0ae06f58f86
--- /dev/null
+++ b/arch/ppc64/kernel/pSeries_vio.c
@@ -0,0 +1,273 @@
1/*
2 * IBM PowerPC pSeries Virtual I/O Infrastructure Support.
3 *
4 * Copyright (c) 2003-2005 IBM Corp.
5 * Dave Engebretsen engebret@us.ibm.com
6 * Santiago Leon santil@us.ibm.com
7 * Hollis Blanchard <hollisb@us.ibm.com>
8 * Stephen Rothwell
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
14 */
15
16#include <linux/init.h>
17#include <linux/module.h>
18#include <linux/mm.h>
19#include <linux/kobject.h>
20#include <asm/iommu.h>
21#include <asm/dma.h>
22#include <asm/prom.h>
23#include <asm/vio.h>
24#include <asm/hvcall.h>
25
26extern struct subsystem devices_subsys; /* needed for vio_find_name() */
27
28static void probe_bus_pseries(void)
29{
30 struct device_node *node_vroot, *of_node;
31
32 node_vroot = find_devices("vdevice");
33 if ((node_vroot == NULL) || (node_vroot->child == NULL))
34 /* this machine doesn't do virtual IO, and that's ok */
35 return;
36
37 /*
38 * Create struct vio_devices for each virtual device in the device tree.
39 * Drivers will associate with them later.
40 */
41 for (of_node = node_vroot->child; of_node != NULL;
42 of_node = of_node->sibling) {
43 printk(KERN_DEBUG "%s: processing %p\n", __FUNCTION__, of_node);
44 vio_register_device_node(of_node);
45 }
46}
47
48/**
49 * vio_match_device_pseries: - Tell if a pSeries VIO device matches a
50 * vio_device_id
51 */
52static int vio_match_device_pseries(const struct vio_device_id *id,
53 const struct vio_dev *dev)
54{
55 return (strncmp(dev->type, id->type, strlen(id->type)) == 0) &&
56 device_is_compatible(dev->dev.platform_data, id->compat);
57}
58
59static void vio_release_device_pseries(struct device *dev)
60{
61 /* XXX free TCE table */
62 of_node_put(dev->platform_data);
63}
64
65static ssize_t viodev_show_devspec(struct device *dev,
66 struct device_attribute *attr, char *buf)
67{
68 struct device_node *of_node = dev->platform_data;
69
70 return sprintf(buf, "%s\n", of_node->full_name);
71}
72DEVICE_ATTR(devspec, S_IRUSR | S_IRGRP | S_IROTH, viodev_show_devspec, NULL);
73
74static void vio_unregister_device_pseries(struct vio_dev *viodev)
75{
76 device_remove_file(&viodev->dev, &dev_attr_devspec);
77}
78
79static struct vio_bus_ops vio_bus_ops_pseries = {
80 .match = vio_match_device_pseries,
81 .unregister_device = vio_unregister_device_pseries,
82 .release_device = vio_release_device_pseries,
83};
84
85/**
86 * vio_bus_init_pseries: - Initialize the pSeries virtual IO bus
87 */
88static int __init vio_bus_init_pseries(void)
89{
90 int err;
91
92 err = vio_bus_init(&vio_bus_ops_pseries);
93 if (err == 0)
94 probe_bus_pseries();
95 return err;
96}
97
98__initcall(vio_bus_init_pseries);
99
100/**
101 * vio_build_iommu_table: - gets the dma information from OF and
102 * builds the TCE tree.
103 * @dev: the virtual device.
104 *
105 * Returns a pointer to the built tce tree, or NULL if it can't
106 * find property.
107*/
108static struct iommu_table *vio_build_iommu_table(struct vio_dev *dev)
109{
110 unsigned int *dma_window;
111 struct iommu_table *newTceTable;
112 unsigned long offset;
113 int dma_window_property_size;
114
115 dma_window = (unsigned int *) get_property(dev->dev.platform_data, "ibm,my-dma-window", &dma_window_property_size);
116 if(!dma_window) {
117 return NULL;
118 }
119
120 newTceTable = (struct iommu_table *) kmalloc(sizeof(struct iommu_table), GFP_KERNEL);
121
122 /* There should be some code to extract the phys-encoded offset
123 using prom_n_addr_cells(). However, according to a comment
124 on earlier versions, it's always zero, so we don't bother */
125 offset = dma_window[1] >> PAGE_SHIFT;
126
127 /* TCE table size - measured in tce entries */
128 newTceTable->it_size = dma_window[4] >> PAGE_SHIFT;
129 /* offset for VIO should always be 0 */
130 newTceTable->it_offset = offset;
131 newTceTable->it_busno = 0;
132 newTceTable->it_index = (unsigned long)dma_window[0];
133 newTceTable->it_type = TCE_VB;
134
135 return iommu_init_table(newTceTable);
136}
137
138/**
139 * vio_register_device_node: - Register a new vio device.
140 * @of_node: The OF node for this device.
141 *
142 * Creates and initializes a vio_dev structure from the data in
143 * of_node (dev.platform_data) and adds it to the list of virtual devices.
144 * Returns a pointer to the created vio_dev or NULL if node has
145 * NULL device_type or compatible fields.
146 */
147struct vio_dev * __devinit vio_register_device_node(struct device_node *of_node)
148{
149 struct vio_dev *viodev;
150 unsigned int *unit_address;
151 unsigned int *irq_p;
152
153 /* we need the 'device_type' property, in order to match with drivers */
154 if ((NULL == of_node->type)) {
155 printk(KERN_WARNING
156 "%s: node %s missing 'device_type'\n", __FUNCTION__,
157 of_node->name ? of_node->name : "<unknown>");
158 return NULL;
159 }
160
161 unit_address = (unsigned int *)get_property(of_node, "reg", NULL);
162 if (!unit_address) {
163 printk(KERN_WARNING "%s: node %s missing 'reg'\n", __FUNCTION__,
164 of_node->name ? of_node->name : "<unknown>");
165 return NULL;
166 }
167
168 /* allocate a vio_dev for this node */
169 viodev = kmalloc(sizeof(struct vio_dev), GFP_KERNEL);
170 if (!viodev) {
171 return NULL;
172 }
173 memset(viodev, 0, sizeof(struct vio_dev));
174
175 viodev->dev.platform_data = of_node_get(of_node);
176
177 viodev->irq = NO_IRQ;
178 irq_p = (unsigned int *)get_property(of_node, "interrupts", NULL);
179 if (irq_p) {
180 int virq = virt_irq_create_mapping(*irq_p);
181 if (virq == NO_IRQ) {
182 printk(KERN_ERR "Unable to allocate interrupt "
183 "number for %s\n", of_node->full_name);
184 } else
185 viodev->irq = irq_offset_up(virq);
186 }
187
188 snprintf(viodev->dev.bus_id, BUS_ID_SIZE, "%x", *unit_address);
189 viodev->name = of_node->name;
190 viodev->type = of_node->type;
191 viodev->unit_address = *unit_address;
192 viodev->iommu_table = vio_build_iommu_table(viodev);
193
194 /* register with generic device framework */
195 if (vio_register_device(viodev) == NULL) {
196 /* XXX free TCE table */
197 kfree(viodev);
198 return NULL;
199 }
200 device_create_file(&viodev->dev, &dev_attr_devspec);
201
202 return viodev;
203}
204EXPORT_SYMBOL(vio_register_device_node);
205
206/**
207 * vio_get_attribute: - get attribute for virtual device
208 * @vdev: The vio device to get property.
209 * @which: The property/attribute to be extracted.
210 * @length: Pointer to length of returned data size (unused if NULL).
211 *
212 * Calls prom.c's get_property() to return the value of the
213 * attribute specified by the preprocessor constant @which
214*/
215const void * vio_get_attribute(struct vio_dev *vdev, void* which, int* length)
216{
217 return get_property(vdev->dev.platform_data, (char*)which, length);
218}
219EXPORT_SYMBOL(vio_get_attribute);
220
221/* vio_find_name() - internal because only vio.c knows how we formatted the
222 * kobject name
223 * XXX once vio_bus_type.devices is actually used as a kset in
224 * drivers/base/bus.c, this function should be removed in favor of
225 * "device_find(kobj_name, &vio_bus_type)"
226 */
227static struct vio_dev *vio_find_name(const char *kobj_name)
228{
229 struct kobject *found;
230
231 found = kset_find_obj(&devices_subsys.kset, kobj_name);
232 if (!found)
233 return NULL;
234
235 return to_vio_dev(container_of(found, struct device, kobj));
236}
237
238/**
239 * vio_find_node - find an already-registered vio_dev
240 * @vnode: device_node of the virtual device we're looking for
241 */
242struct vio_dev *vio_find_node(struct device_node *vnode)
243{
244 uint32_t *unit_address;
245 char kobj_name[BUS_ID_SIZE];
246
247 /* construct the kobject name from the device node */
248 unit_address = (uint32_t *)get_property(vnode, "reg", NULL);
249 if (!unit_address)
250 return NULL;
251 snprintf(kobj_name, BUS_ID_SIZE, "%x", *unit_address);
252
253 return vio_find_name(kobj_name);
254}
255EXPORT_SYMBOL(vio_find_node);
256
257int vio_enable_interrupts(struct vio_dev *dev)
258{
259 int rc = h_vio_signal(dev->unit_address, VIO_IRQ_ENABLE);
260 if (rc != H_Success)
261 printk(KERN_ERR "vio: Error 0x%x enabling interrupts\n", rc);
262 return rc;
263}
264EXPORT_SYMBOL(vio_enable_interrupts);
265
266int vio_disable_interrupts(struct vio_dev *dev)
267{
268 int rc = h_vio_signal(dev->unit_address, VIO_IRQ_DISABLE);
269 if (rc != H_Success)
270 printk(KERN_ERR "vio: Error 0x%x disabling interrupts\n", rc);
271 return rc;
272}
273EXPORT_SYMBOL(vio_disable_interrupts);
diff --git a/arch/ppc64/kernel/pacaData.c b/arch/ppc64/kernel/pacaData.c
index 6316188737b6..6182a2cd90a5 100644
--- a/arch/ppc64/kernel/pacaData.c
+++ b/arch/ppc64/kernel/pacaData.c
@@ -78,7 +78,7 @@ extern unsigned long __toc_start;
78 78
79#define BOOTCPU_PACA_INIT(number) \ 79#define BOOTCPU_PACA_INIT(number) \
80{ \ 80{ \
81 PACA_INIT_COMMON(number, 1, 0, STAB0_VIRT_ADDR) \ 81 PACA_INIT_COMMON(number, 1, 0, (u64)&initial_stab) \
82 PACA_INIT_ISERIES(number) \ 82 PACA_INIT_ISERIES(number) \
83} 83}
84 84
@@ -90,7 +90,7 @@ extern unsigned long __toc_start;
90 90
91#define BOOTCPU_PACA_INIT(number) \ 91#define BOOTCPU_PACA_INIT(number) \
92{ \ 92{ \
93 PACA_INIT_COMMON(number, 1, STAB0_PHYS_ADDR, STAB0_VIRT_ADDR) \ 93 PACA_INIT_COMMON(number, 1, STAB0_PHYS_ADDR, (u64)&initial_stab) \
94} 94}
95#endif 95#endif
96 96
diff --git a/arch/ppc64/kernel/pmac_setup.c b/arch/ppc64/kernel/pmac_setup.c
index e40877fa67cd..8ff86a766cdf 100644
--- a/arch/ppc64/kernel/pmac_setup.c
+++ b/arch/ppc64/kernel/pmac_setup.c
@@ -71,6 +71,7 @@
71#include <asm/of_device.h> 71#include <asm/of_device.h>
72#include <asm/lmb.h> 72#include <asm/lmb.h>
73#include <asm/smu.h> 73#include <asm/smu.h>
74#include <asm/pmc.h>
74 75
75#include "pmac.h" 76#include "pmac.h"
76#include "mpic.h" 77#include "mpic.h"
@@ -511,4 +512,5 @@ struct machdep_calls __initdata pmac_md = {
511 .progress = pmac_progress, 512 .progress = pmac_progress,
512 .check_legacy_ioport = pmac_check_legacy_ioport, 513 .check_legacy_ioport = pmac_check_legacy_ioport,
513 .idle_loop = native_idle, 514 .idle_loop = native_idle,
515 .enable_pmcs = power4_enable_pmcs,
514}; 516};
diff --git a/arch/ppc64/kernel/pmc.c b/arch/ppc64/kernel/pmc.c
index 67be773f9c00..cdfec7438d01 100644
--- a/arch/ppc64/kernel/pmc.c
+++ b/arch/ppc64/kernel/pmc.c
@@ -65,3 +65,24 @@ void release_pmc_hardware(void)
65 spin_unlock(&pmc_owner_lock); 65 spin_unlock(&pmc_owner_lock);
66} 66}
67EXPORT_SYMBOL_GPL(release_pmc_hardware); 67EXPORT_SYMBOL_GPL(release_pmc_hardware);
68
69void power4_enable_pmcs(void)
70{
71 unsigned long hid0;
72
73 hid0 = mfspr(HID0);
74 hid0 |= 1UL << (63 - 20);
75
76 /* POWER4 requires the following sequence */
77 asm volatile(
78 "sync\n"
79 "mtspr %1, %0\n"
80 "mfspr %0, %1\n"
81 "mfspr %0, %1\n"
82 "mfspr %0, %1\n"
83 "mfspr %0, %1\n"
84 "mfspr %0, %1\n"
85 "mfspr %0, %1\n"
86 "isync" : "=&r" (hid0) : "i" (HID0), "0" (hid0):
87 "memory");
88}
diff --git a/arch/ppc64/kernel/process.c b/arch/ppc64/kernel/process.c
index f7cae05e40fb..7a7e027653ad 100644
--- a/arch/ppc64/kernel/process.c
+++ b/arch/ppc64/kernel/process.c
@@ -50,6 +50,7 @@
50#include <asm/machdep.h> 50#include <asm/machdep.h>
51#include <asm/iSeries/HvCallHpt.h> 51#include <asm/iSeries/HvCallHpt.h>
52#include <asm/cputable.h> 52#include <asm/cputable.h>
53#include <asm/firmware.h>
53#include <asm/sections.h> 54#include <asm/sections.h>
54#include <asm/tlbflush.h> 55#include <asm/tlbflush.h>
55#include <asm/time.h> 56#include <asm/time.h>
@@ -202,11 +203,10 @@ struct task_struct *__switch_to(struct task_struct *prev,
202 new_thread = &new->thread; 203 new_thread = &new->thread;
203 old_thread = &current->thread; 204 old_thread = &current->thread;
204 205
205/* Collect purr utilization data per process and per processor wise */ 206 /* Collect purr utilization data per process and per processor
206/* purr is nothing but processor time base */ 207 * wise purr is nothing but processor time base
207 208 */
208#if defined(CONFIG_PPC_PSERIES) 209 if (firmware_has_feature(FW_FEATURE_SPLPAR)) {
209 if (cur_cpu_spec->firmware_features & FW_FEATURE_SPLPAR) {
210 struct cpu_usage *cu = &__get_cpu_var(cpu_usage_array); 210 struct cpu_usage *cu = &__get_cpu_var(cpu_usage_array);
211 long unsigned start_tb, current_tb; 211 long unsigned start_tb, current_tb;
212 start_tb = old_thread->start_tb; 212 start_tb = old_thread->start_tb;
@@ -214,8 +214,6 @@ struct task_struct *__switch_to(struct task_struct *prev,
214 old_thread->accum_tb += (current_tb - start_tb); 214 old_thread->accum_tb += (current_tb - start_tb);
215 new_thread->start_tb = current_tb; 215 new_thread->start_tb = current_tb;
216 } 216 }
217#endif
218
219 217
220 local_irq_save(flags); 218 local_irq_save(flags);
221 last = _switch(old_thread, new_thread); 219 last = _switch(old_thread, new_thread);
diff --git a/arch/ppc64/kernel/prom.c b/arch/ppc64/kernel/prom.c
index 5aca01ddd81f..b21848826791 100644
--- a/arch/ppc64/kernel/prom.c
+++ b/arch/ppc64/kernel/prom.c
@@ -625,8 +625,8 @@ void __init finish_device_tree(void)
625 625
626static inline char *find_flat_dt_string(u32 offset) 626static inline char *find_flat_dt_string(u32 offset)
627{ 627{
628 return ((char *)initial_boot_params) + initial_boot_params->off_dt_strings 628 return ((char *)initial_boot_params) +
629 + offset; 629 initial_boot_params->off_dt_strings + offset;
630} 630}
631 631
632/** 632/**
@@ -635,26 +635,33 @@ static inline char *find_flat_dt_string(u32 offset)
635 * unflatten the tree 635 * unflatten the tree
636 */ 636 */
637static int __init scan_flat_dt(int (*it)(unsigned long node, 637static int __init scan_flat_dt(int (*it)(unsigned long node,
638 const char *full_path, void *data), 638 const char *uname, int depth,
639 void *data),
639 void *data) 640 void *data)
640{ 641{
641 unsigned long p = ((unsigned long)initial_boot_params) + 642 unsigned long p = ((unsigned long)initial_boot_params) +
642 initial_boot_params->off_dt_struct; 643 initial_boot_params->off_dt_struct;
643 int rc = 0; 644 int rc = 0;
645 int depth = -1;
644 646
645 do { 647 do {
646 u32 tag = *((u32 *)p); 648 u32 tag = *((u32 *)p);
647 char *pathp; 649 char *pathp;
648 650
649 p += 4; 651 p += 4;
650 if (tag == OF_DT_END_NODE) 652 if (tag == OF_DT_END_NODE) {
653 depth --;
654 continue;
655 }
656 if (tag == OF_DT_NOP)
651 continue; 657 continue;
652 if (tag == OF_DT_END) 658 if (tag == OF_DT_END)
653 break; 659 break;
654 if (tag == OF_DT_PROP) { 660 if (tag == OF_DT_PROP) {
655 u32 sz = *((u32 *)p); 661 u32 sz = *((u32 *)p);
656 p += 8; 662 p += 8;
657 p = _ALIGN(p, sz >= 8 ? 8 : 4); 663 if (initial_boot_params->version < 0x10)
664 p = _ALIGN(p, sz >= 8 ? 8 : 4);
658 p += sz; 665 p += sz;
659 p = _ALIGN(p, 4); 666 p = _ALIGN(p, 4);
660 continue; 667 continue;
@@ -664,9 +671,18 @@ static int __init scan_flat_dt(int (*it)(unsigned long node,
664 " device tree !\n", tag); 671 " device tree !\n", tag);
665 return -EINVAL; 672 return -EINVAL;
666 } 673 }
674 depth++;
667 pathp = (char *)p; 675 pathp = (char *)p;
668 p = _ALIGN(p + strlen(pathp) + 1, 4); 676 p = _ALIGN(p + strlen(pathp) + 1, 4);
669 rc = it(p, pathp, data); 677 if ((*pathp) == '/') {
678 char *lp, *np;
679 for (lp = NULL, np = pathp; *np; np++)
680 if ((*np) == '/')
681 lp = np+1;
682 if (lp != NULL)
683 pathp = lp;
684 }
685 rc = it(p, pathp, depth, data);
670 if (rc != 0) 686 if (rc != 0)
671 break; 687 break;
672 } while(1); 688 } while(1);
@@ -689,17 +705,21 @@ static void* __init get_flat_dt_prop(unsigned long node, const char *name,
689 const char *nstr; 705 const char *nstr;
690 706
691 p += 4; 707 p += 4;
708 if (tag == OF_DT_NOP)
709 continue;
692 if (tag != OF_DT_PROP) 710 if (tag != OF_DT_PROP)
693 return NULL; 711 return NULL;
694 712
695 sz = *((u32 *)p); 713 sz = *((u32 *)p);
696 noff = *((u32 *)(p + 4)); 714 noff = *((u32 *)(p + 4));
697 p += 8; 715 p += 8;
698 p = _ALIGN(p, sz >= 8 ? 8 : 4); 716 if (initial_boot_params->version < 0x10)
717 p = _ALIGN(p, sz >= 8 ? 8 : 4);
699 718
700 nstr = find_flat_dt_string(noff); 719 nstr = find_flat_dt_string(noff);
701 if (nstr == NULL) { 720 if (nstr == NULL) {
702 printk(KERN_WARNING "Can't find property index name !\n"); 721 printk(KERN_WARNING "Can't find property index"
722 " name !\n");
703 return NULL; 723 return NULL;
704 } 724 }
705 if (strcmp(name, nstr) == 0) { 725 if (strcmp(name, nstr) == 0) {
@@ -713,7 +733,7 @@ static void* __init get_flat_dt_prop(unsigned long node, const char *name,
713} 733}
714 734
715static void *__init unflatten_dt_alloc(unsigned long *mem, unsigned long size, 735static void *__init unflatten_dt_alloc(unsigned long *mem, unsigned long size,
716 unsigned long align) 736 unsigned long align)
717{ 737{
718 void *res; 738 void *res;
719 739
@@ -727,13 +747,16 @@ static void *__init unflatten_dt_alloc(unsigned long *mem, unsigned long size,
727static unsigned long __init unflatten_dt_node(unsigned long mem, 747static unsigned long __init unflatten_dt_node(unsigned long mem,
728 unsigned long *p, 748 unsigned long *p,
729 struct device_node *dad, 749 struct device_node *dad,
730 struct device_node ***allnextpp) 750 struct device_node ***allnextpp,
751 unsigned long fpsize)
731{ 752{
732 struct device_node *np; 753 struct device_node *np;
733 struct property *pp, **prev_pp = NULL; 754 struct property *pp, **prev_pp = NULL;
734 char *pathp; 755 char *pathp;
735 u32 tag; 756 u32 tag;
736 unsigned int l; 757 unsigned int l, allocl;
758 int has_name = 0;
759 int new_format = 0;
737 760
738 tag = *((u32 *)(*p)); 761 tag = *((u32 *)(*p));
739 if (tag != OF_DT_BEGIN_NODE) { 762 if (tag != OF_DT_BEGIN_NODE) {
@@ -742,21 +765,62 @@ static unsigned long __init unflatten_dt_node(unsigned long mem,
742 } 765 }
743 *p += 4; 766 *p += 4;
744 pathp = (char *)*p; 767 pathp = (char *)*p;
745 l = strlen(pathp) + 1; 768 l = allocl = strlen(pathp) + 1;
746 *p = _ALIGN(*p + l, 4); 769 *p = _ALIGN(*p + l, 4);
747 770
748 np = unflatten_dt_alloc(&mem, sizeof(struct device_node) + l, 771 /* version 0x10 has a more compact unit name here instead of the full
772 * path. we accumulate the full path size using "fpsize", we'll rebuild
773 * it later. We detect this because the first character of the name is
774 * not '/'.
775 */
776 if ((*pathp) != '/') {
777 new_format = 1;
778 if (fpsize == 0) {
779 /* root node: special case. fpsize accounts for path
780 * plus terminating zero. root node only has '/', so
781 * fpsize should be 2, but we want to avoid the first
782 * level nodes to have two '/' so we use fpsize 1 here
783 */
784 fpsize = 1;
785 allocl = 2;
786 } else {
787 /* account for '/' and path size minus terminal 0
788 * already in 'l'
789 */
790 fpsize += l;
791 allocl = fpsize;
792 }
793 }
794
795
796 np = unflatten_dt_alloc(&mem, sizeof(struct device_node) + allocl,
749 __alignof__(struct device_node)); 797 __alignof__(struct device_node));
750 if (allnextpp) { 798 if (allnextpp) {
751 memset(np, 0, sizeof(*np)); 799 memset(np, 0, sizeof(*np));
752 np->full_name = ((char*)np) + sizeof(struct device_node); 800 np->full_name = ((char*)np) + sizeof(struct device_node);
753 memcpy(np->full_name, pathp, l); 801 if (new_format) {
802 char *p = np->full_name;
803 /* rebuild full path for new format */
804 if (dad && dad->parent) {
805 strcpy(p, dad->full_name);
806#ifdef DEBUG
807 if ((strlen(p) + l + 1) != allocl) {
808 DBG("%s: p: %d, l: %d, a: %d\n",
809 pathp, strlen(p), l, allocl);
810 }
811#endif
812 p += strlen(p);
813 }
814 *(p++) = '/';
815 memcpy(p, pathp, l);
816 } else
817 memcpy(np->full_name, pathp, l);
754 prev_pp = &np->properties; 818 prev_pp = &np->properties;
755 **allnextpp = np; 819 **allnextpp = np;
756 *allnextpp = &np->allnext; 820 *allnextpp = &np->allnext;
757 if (dad != NULL) { 821 if (dad != NULL) {
758 np->parent = dad; 822 np->parent = dad;
759 /* we temporarily use the `next' field as `last_child'. */ 823 /* we temporarily use the next field as `last_child'*/
760 if (dad->next == 0) 824 if (dad->next == 0)
761 dad->child = np; 825 dad->child = np;
762 else 826 else
@@ -770,18 +834,26 @@ static unsigned long __init unflatten_dt_node(unsigned long mem,
770 char *pname; 834 char *pname;
771 835
772 tag = *((u32 *)(*p)); 836 tag = *((u32 *)(*p));
837 if (tag == OF_DT_NOP) {
838 *p += 4;
839 continue;
840 }
773 if (tag != OF_DT_PROP) 841 if (tag != OF_DT_PROP)
774 break; 842 break;
775 *p += 4; 843 *p += 4;
776 sz = *((u32 *)(*p)); 844 sz = *((u32 *)(*p));
777 noff = *((u32 *)((*p) + 4)); 845 noff = *((u32 *)((*p) + 4));
778 *p = _ALIGN((*p) + 8, sz >= 8 ? 8 : 4); 846 *p += 8;
847 if (initial_boot_params->version < 0x10)
848 *p = _ALIGN(*p, sz >= 8 ? 8 : 4);
779 849
780 pname = find_flat_dt_string(noff); 850 pname = find_flat_dt_string(noff);
781 if (pname == NULL) { 851 if (pname == NULL) {
782 printk("Can't find property name in list !\n"); 852 printk("Can't find property name in list !\n");
783 break; 853 break;
784 } 854 }
855 if (strcmp(pname, "name") == 0)
856 has_name = 1;
785 l = strlen(pname) + 1; 857 l = strlen(pname) + 1;
786 pp = unflatten_dt_alloc(&mem, sizeof(struct property), 858 pp = unflatten_dt_alloc(&mem, sizeof(struct property),
787 __alignof__(struct property)); 859 __alignof__(struct property));
@@ -801,6 +873,36 @@ static unsigned long __init unflatten_dt_node(unsigned long mem,
801 } 873 }
802 *p = _ALIGN((*p) + sz, 4); 874 *p = _ALIGN((*p) + sz, 4);
803 } 875 }
876 /* with version 0x10 we may not have the name property, recreate
877 * it here from the unit name if absent
878 */
879 if (!has_name) {
880 char *p = pathp, *ps = pathp, *pa = NULL;
881 int sz;
882
883 while (*p) {
884 if ((*p) == '@')
885 pa = p;
886 if ((*p) == '/')
887 ps = p + 1;
888 p++;
889 }
890 if (pa < ps)
891 pa = p;
892 sz = (pa - ps) + 1;
893 pp = unflatten_dt_alloc(&mem, sizeof(struct property) + sz,
894 __alignof__(struct property));
895 if (allnextpp) {
896 pp->name = "name";
897 pp->length = sz;
898 pp->value = (unsigned char *)(pp + 1);
899 *prev_pp = pp;
900 prev_pp = &pp->next;
901 memcpy(pp->value, ps, sz - 1);
902 ((char *)pp->value)[sz - 1] = 0;
903 DBG("fixed up name for %s -> %s\n", pathp, pp->value);
904 }
905 }
804 if (allnextpp) { 906 if (allnextpp) {
805 *prev_pp = NULL; 907 *prev_pp = NULL;
806 np->name = get_property(np, "name", NULL); 908 np->name = get_property(np, "name", NULL);
@@ -812,11 +914,11 @@ static unsigned long __init unflatten_dt_node(unsigned long mem,
812 np->type = "<NULL>"; 914 np->type = "<NULL>";
813 } 915 }
814 while (tag == OF_DT_BEGIN_NODE) { 916 while (tag == OF_DT_BEGIN_NODE) {
815 mem = unflatten_dt_node(mem, p, np, allnextpp); 917 mem = unflatten_dt_node(mem, p, np, allnextpp, fpsize);
816 tag = *((u32 *)(*p)); 918 tag = *((u32 *)(*p));
817 } 919 }
818 if (tag != OF_DT_END_NODE) { 920 if (tag != OF_DT_END_NODE) {
819 printk("Weird tag at start of node: %x\n", tag); 921 printk("Weird tag at end of node: %x\n", tag);
820 return mem; 922 return mem;
821 } 923 }
822 *p += 4; 924 *p += 4;
@@ -842,21 +944,32 @@ void __init unflatten_device_tree(void)
842 /* First pass, scan for size */ 944 /* First pass, scan for size */
843 start = ((unsigned long)initial_boot_params) + 945 start = ((unsigned long)initial_boot_params) +
844 initial_boot_params->off_dt_struct; 946 initial_boot_params->off_dt_struct;
845 size = unflatten_dt_node(0, &start, NULL, NULL); 947 size = unflatten_dt_node(0, &start, NULL, NULL, 0);
948 size = (size | 3) + 1;
846 949
847 DBG(" size is %lx, allocating...\n", size); 950 DBG(" size is %lx, allocating...\n", size);
848 951
849 /* Allocate memory for the expanded device tree */ 952 /* Allocate memory for the expanded device tree */
850 mem = (unsigned long)abs_to_virt(lmb_alloc(size, 953 mem = lmb_alloc(size + 4, __alignof__(struct device_node));
851 __alignof__(struct device_node))); 954 if (!mem) {
955 DBG("Couldn't allocate memory with lmb_alloc()!\n");
956 panic("Couldn't allocate memory with lmb_alloc()!\n");
957 }
958 mem = (unsigned long)abs_to_virt(mem);
959
960 ((u32 *)mem)[size / 4] = 0xdeadbeef;
961
852 DBG(" unflattening...\n", mem); 962 DBG(" unflattening...\n", mem);
853 963
854 /* Second pass, do actual unflattening */ 964 /* Second pass, do actual unflattening */
855 start = ((unsigned long)initial_boot_params) + 965 start = ((unsigned long)initial_boot_params) +
856 initial_boot_params->off_dt_struct; 966 initial_boot_params->off_dt_struct;
857 unflatten_dt_node(mem, &start, NULL, &allnextp); 967 unflatten_dt_node(mem, &start, NULL, &allnextp, 0);
858 if (*((u32 *)start) != OF_DT_END) 968 if (*((u32 *)start) != OF_DT_END)
859 printk(KERN_WARNING "Weird tag at end of tree: %x\n", *((u32 *)start)); 969 printk(KERN_WARNING "Weird tag at end of tree: %08x\n", *((u32 *)start));
970 if (((u32 *)mem)[size / 4] != 0xdeadbeef)
971 printk(KERN_WARNING "End of tree marker overwritten: %08x\n",
972 ((u32 *)mem)[size / 4] );
860 *allnextp = NULL; 973 *allnextp = NULL;
861 974
862 /* Get pointer to OF "/chosen" node for use everywhere */ 975 /* Get pointer to OF "/chosen" node for use everywhere */
@@ -880,7 +993,7 @@ void __init unflatten_device_tree(void)
880 993
881 994
882static int __init early_init_dt_scan_cpus(unsigned long node, 995static int __init early_init_dt_scan_cpus(unsigned long node,
883 const char *full_path, void *data) 996 const char *uname, int depth, void *data)
884{ 997{
885 char *type = get_flat_dt_prop(node, "device_type", NULL); 998 char *type = get_flat_dt_prop(node, "device_type", NULL);
886 u32 *prop; 999 u32 *prop;
@@ -947,13 +1060,15 @@ static int __init early_init_dt_scan_cpus(unsigned long node,
947} 1060}
948 1061
949static int __init early_init_dt_scan_chosen(unsigned long node, 1062static int __init early_init_dt_scan_chosen(unsigned long node,
950 const char *full_path, void *data) 1063 const char *uname, int depth, void *data)
951{ 1064{
952 u32 *prop; 1065 u32 *prop;
953 u64 *prop64; 1066 u64 *prop64;
954 extern unsigned long memory_limit, tce_alloc_start, tce_alloc_end; 1067 extern unsigned long memory_limit, tce_alloc_start, tce_alloc_end;
955 1068
956 if (strcmp(full_path, "/chosen") != 0) 1069 DBG("search \"chosen\", depth: %d, uname: %s\n", depth, uname);
1070
1071 if (depth != 1 || strcmp(uname, "chosen") != 0)
957 return 0; 1072 return 0;
958 1073
959 /* get platform type */ 1074 /* get platform type */
@@ -1003,18 +1118,20 @@ static int __init early_init_dt_scan_chosen(unsigned long node,
1003} 1118}
1004 1119
1005static int __init early_init_dt_scan_root(unsigned long node, 1120static int __init early_init_dt_scan_root(unsigned long node,
1006 const char *full_path, void *data) 1121 const char *uname, int depth, void *data)
1007{ 1122{
1008 u32 *prop; 1123 u32 *prop;
1009 1124
1010 if (strcmp(full_path, "/") != 0) 1125 if (depth != 0)
1011 return 0; 1126 return 0;
1012 1127
1013 prop = (u32 *)get_flat_dt_prop(node, "#size-cells", NULL); 1128 prop = (u32 *)get_flat_dt_prop(node, "#size-cells", NULL);
1014 dt_root_size_cells = (prop == NULL) ? 1 : *prop; 1129 dt_root_size_cells = (prop == NULL) ? 1 : *prop;
1015 1130 DBG("dt_root_size_cells = %x\n", dt_root_size_cells);
1131
1016 prop = (u32 *)get_flat_dt_prop(node, "#address-cells", NULL); 1132 prop = (u32 *)get_flat_dt_prop(node, "#address-cells", NULL);
1017 dt_root_addr_cells = (prop == NULL) ? 2 : *prop; 1133 dt_root_addr_cells = (prop == NULL) ? 2 : *prop;
1134 DBG("dt_root_addr_cells = %x\n", dt_root_addr_cells);
1018 1135
1019 /* break now */ 1136 /* break now */
1020 return 1; 1137 return 1;
@@ -1042,7 +1159,7 @@ static unsigned long __init dt_mem_next_cell(int s, cell_t **cellp)
1042 1159
1043 1160
1044static int __init early_init_dt_scan_memory(unsigned long node, 1161static int __init early_init_dt_scan_memory(unsigned long node,
1045 const char *full_path, void *data) 1162 const char *uname, int depth, void *data)
1046{ 1163{
1047 char *type = get_flat_dt_prop(node, "device_type", NULL); 1164 char *type = get_flat_dt_prop(node, "device_type", NULL);
1048 cell_t *reg, *endp; 1165 cell_t *reg, *endp;
@@ -1058,7 +1175,9 @@ static int __init early_init_dt_scan_memory(unsigned long node,
1058 1175
1059 endp = reg + (l / sizeof(cell_t)); 1176 endp = reg + (l / sizeof(cell_t));
1060 1177
1061 DBG("memory scan node %s ...\n", full_path); 1178 DBG("memory scan node %s ..., reg size %ld, data: %x %x %x %x, ...\n",
1179 uname, l, reg[0], reg[1], reg[2], reg[3]);
1180
1062 while ((endp - reg) >= (dt_root_addr_cells + dt_root_size_cells)) { 1181 while ((endp - reg) >= (dt_root_addr_cells + dt_root_size_cells)) {
1063 unsigned long base, size; 1182 unsigned long base, size;
1064 1183
@@ -1469,10 +1588,11 @@ struct device_node *of_find_node_by_path(const char *path)
1469 struct device_node *np = allnodes; 1588 struct device_node *np = allnodes;
1470 1589
1471 read_lock(&devtree_lock); 1590 read_lock(&devtree_lock);
1472 for (; np != 0; np = np->allnext) 1591 for (; np != 0; np = np->allnext) {
1473 if (np->full_name != 0 && strcasecmp(np->full_name, path) == 0 1592 if (np->full_name != 0 && strcasecmp(np->full_name, path) == 0
1474 && of_node_get(np)) 1593 && of_node_get(np))
1475 break; 1594 break;
1595 }
1476 read_unlock(&devtree_lock); 1596 read_unlock(&devtree_lock);
1477 return np; 1597 return np;
1478} 1598}
diff --git a/arch/ppc64/kernel/prom_init.c b/arch/ppc64/kernel/prom_init.c
index dbbe6c79d8da..122283a1d39a 100644
--- a/arch/ppc64/kernel/prom_init.c
+++ b/arch/ppc64/kernel/prom_init.c
@@ -892,7 +892,10 @@ static void __init prom_init_mem(void)
892 if ( RELOC(of_platform) == PLATFORM_PSERIES_LPAR ) 892 if ( RELOC(of_platform) == PLATFORM_PSERIES_LPAR )
893 RELOC(alloc_top) = RELOC(rmo_top); 893 RELOC(alloc_top) = RELOC(rmo_top);
894 else 894 else
895 RELOC(alloc_top) = RELOC(rmo_top) = min(0x40000000ul, RELOC(ram_top)); 895 /* Some RS64 machines have buggy firmware where claims up at 1GB
896 * fails. Cap at 768MB as a workaround. Still plenty of room.
897 */
898 RELOC(alloc_top) = RELOC(rmo_top) = min(0x30000000ul, RELOC(ram_top));
896 899
897 prom_printf("memory layout at init:\n"); 900 prom_printf("memory layout at init:\n");
898 prom_printf(" memory_limit : %x (16 MB aligned)\n", RELOC(prom_memory_limit)); 901 prom_printf(" memory_limit : %x (16 MB aligned)\n", RELOC(prom_memory_limit));
@@ -1534,7 +1537,8 @@ static unsigned long __init dt_find_string(char *str)
1534 */ 1537 */
1535#define MAX_PROPERTY_NAME 64 1538#define MAX_PROPERTY_NAME 64
1536 1539
1537static void __init scan_dt_build_strings(phandle node, unsigned long *mem_start, 1540static void __init scan_dt_build_strings(phandle node,
1541 unsigned long *mem_start,
1538 unsigned long *mem_end) 1542 unsigned long *mem_end)
1539{ 1543{
1540 unsigned long offset = reloc_offset(); 1544 unsigned long offset = reloc_offset();
@@ -1547,16 +1551,21 @@ static void __init scan_dt_build_strings(phandle node, unsigned long *mem_start,
1547 /* get and store all property names */ 1551 /* get and store all property names */
1548 prev_name = RELOC(""); 1552 prev_name = RELOC("");
1549 for (;;) { 1553 for (;;) {
1550 int rc;
1551
1552 /* 64 is max len of name including nul. */ 1554 /* 64 is max len of name including nul. */
1553 namep = make_room(mem_start, mem_end, MAX_PROPERTY_NAME, 1); 1555 namep = make_room(mem_start, mem_end, MAX_PROPERTY_NAME, 1);
1554 rc = call_prom("nextprop", 3, 1, node, prev_name, namep); 1556 if (call_prom("nextprop", 3, 1, node, prev_name, namep) != 1) {
1555 if (rc != 1) {
1556 /* No more nodes: unwind alloc */ 1557 /* No more nodes: unwind alloc */
1557 *mem_start = (unsigned long)namep; 1558 *mem_start = (unsigned long)namep;
1558 break; 1559 break;
1559 } 1560 }
1561
1562 /* skip "name" */
1563 if (strcmp(namep, RELOC("name")) == 0) {
1564 *mem_start = (unsigned long)namep;
1565 prev_name = RELOC("name");
1566 continue;
1567 }
1568 /* get/create string entry */
1560 soff = dt_find_string(namep); 1569 soff = dt_find_string(namep);
1561 if (soff != 0) { 1570 if (soff != 0) {
1562 *mem_start = (unsigned long)namep; 1571 *mem_start = (unsigned long)namep;
@@ -1571,7 +1580,7 @@ static void __init scan_dt_build_strings(phandle node, unsigned long *mem_start,
1571 1580
1572 /* do all our children */ 1581 /* do all our children */
1573 child = call_prom("child", 1, 1, node); 1582 child = call_prom("child", 1, 1, node);
1574 while (child != (phandle)0) { 1583 while (child != 0) {
1575 scan_dt_build_strings(child, mem_start, mem_end); 1584 scan_dt_build_strings(child, mem_start, mem_end);
1576 child = call_prom("peer", 1, 1, child); 1585 child = call_prom("peer", 1, 1, child);
1577 } 1586 }
@@ -1580,16 +1589,13 @@ static void __init scan_dt_build_strings(phandle node, unsigned long *mem_start,
1580static void __init scan_dt_build_struct(phandle node, unsigned long *mem_start, 1589static void __init scan_dt_build_struct(phandle node, unsigned long *mem_start,
1581 unsigned long *mem_end) 1590 unsigned long *mem_end)
1582{ 1591{
1583 int l, align;
1584 phandle child; 1592 phandle child;
1585 char *namep, *prev_name, *sstart, *p, *ep; 1593 char *namep, *prev_name, *sstart, *p, *ep, *lp, *path;
1586 unsigned long soff; 1594 unsigned long soff;
1587 unsigned char *valp; 1595 unsigned char *valp;
1588 unsigned long offset = reloc_offset(); 1596 unsigned long offset = reloc_offset();
1589 char pname[MAX_PROPERTY_NAME]; 1597 static char pname[MAX_PROPERTY_NAME];
1590 char *path; 1598 int l;
1591
1592 path = RELOC(prom_scratch);
1593 1599
1594 dt_push_token(OF_DT_BEGIN_NODE, mem_start, mem_end); 1600 dt_push_token(OF_DT_BEGIN_NODE, mem_start, mem_end);
1595 1601
@@ -1599,23 +1605,33 @@ static void __init scan_dt_build_struct(phandle node, unsigned long *mem_start,
1599 namep, *mem_end - *mem_start); 1605 namep, *mem_end - *mem_start);
1600 if (l >= 0) { 1606 if (l >= 0) {
1601 /* Didn't fit? Get more room. */ 1607 /* Didn't fit? Get more room. */
1602 if (l+1 > *mem_end - *mem_start) { 1608 if ((l+1) > (*mem_end - *mem_start)) {
1603 namep = make_room(mem_start, mem_end, l+1, 1); 1609 namep = make_room(mem_start, mem_end, l+1, 1);
1604 call_prom("package-to-path", 3, 1, node, namep, l); 1610 call_prom("package-to-path", 3, 1, node, namep, l);
1605 } 1611 }
1606 namep[l] = '\0'; 1612 namep[l] = '\0';
1613
1607 /* Fixup an Apple bug where they have bogus \0 chars in the 1614 /* Fixup an Apple bug where they have bogus \0 chars in the
1608 * middle of the path in some properties 1615 * middle of the path in some properties
1609 */ 1616 */
1610 for (p = namep, ep = namep + l; p < ep; p++) 1617 for (p = namep, ep = namep + l; p < ep; p++)
1611 if (*p == '\0') { 1618 if (*p == '\0') {
1612 memmove(p, p+1, ep - p); 1619 memmove(p, p+1, ep - p);
1613 ep--; l--; 1620 ep--; l--; p--;
1614 } 1621 }
1615 *mem_start = _ALIGN(((unsigned long) namep) + strlen(namep) + 1, 4); 1622
1623 /* now try to extract the unit name in that mess */
1624 for (p = namep, lp = NULL; *p; p++)
1625 if (*p == '/')
1626 lp = p + 1;
1627 if (lp != NULL)
1628 memmove(namep, lp, strlen(lp) + 1);
1629 *mem_start = _ALIGN(((unsigned long) namep) +
1630 strlen(namep) + 1, 4);
1616 } 1631 }
1617 1632
1618 /* get it again for debugging */ 1633 /* get it again for debugging */
1634 path = RELOC(prom_scratch);
1619 memset(path, 0, PROM_SCRATCH_SIZE); 1635 memset(path, 0, PROM_SCRATCH_SIZE);
1620 call_prom("package-to-path", 3, 1, node, path, PROM_SCRATCH_SIZE-1); 1636 call_prom("package-to-path", 3, 1, node, path, PROM_SCRATCH_SIZE-1);
1621 1637
@@ -1623,23 +1639,27 @@ static void __init scan_dt_build_struct(phandle node, unsigned long *mem_start,
1623 prev_name = RELOC(""); 1639 prev_name = RELOC("");
1624 sstart = (char *)RELOC(dt_string_start); 1640 sstart = (char *)RELOC(dt_string_start);
1625 for (;;) { 1641 for (;;) {
1626 int rc; 1642 if (call_prom("nextprop", 3, 1, node, prev_name,
1627 1643 RELOC(pname)) != 1)
1628 rc = call_prom("nextprop", 3, 1, node, prev_name, pname);
1629 if (rc != 1)
1630 break; 1644 break;
1631 1645
1646 /* skip "name" */
1647 if (strcmp(RELOC(pname), RELOC("name")) == 0) {
1648 prev_name = RELOC("name");
1649 continue;
1650 }
1651
1632 /* find string offset */ 1652 /* find string offset */
1633 soff = dt_find_string(pname); 1653 soff = dt_find_string(RELOC(pname));
1634 if (soff == 0) { 1654 if (soff == 0) {
1635 prom_printf("WARNING: Can't find string index for <%s>, node %s\n", 1655 prom_printf("WARNING: Can't find string index for"
1636 pname, path); 1656 " <%s>, node %s\n", RELOC(pname), path);
1637 break; 1657 break;
1638 } 1658 }
1639 prev_name = sstart + soff; 1659 prev_name = sstart + soff;
1640 1660
1641 /* get length */ 1661 /* get length */
1642 l = call_prom("getproplen", 2, 1, node, pname); 1662 l = call_prom("getproplen", 2, 1, node, RELOC(pname));
1643 1663
1644 /* sanity checks */ 1664 /* sanity checks */
1645 if (l == PROM_ERROR) 1665 if (l == PROM_ERROR)
@@ -1648,7 +1668,7 @@ static void __init scan_dt_build_struct(phandle node, unsigned long *mem_start,
1648 prom_printf("WARNING: ignoring large property "); 1668 prom_printf("WARNING: ignoring large property ");
1649 /* It seems OF doesn't null-terminate the path :-( */ 1669 /* It seems OF doesn't null-terminate the path :-( */
1650 prom_printf("[%s] ", path); 1670 prom_printf("[%s] ", path);
1651 prom_printf("%s length 0x%x\n", pname, l); 1671 prom_printf("%s length 0x%x\n", RELOC(pname), l);
1652 continue; 1672 continue;
1653 } 1673 }
1654 1674
@@ -1658,17 +1678,16 @@ static void __init scan_dt_build_struct(phandle node, unsigned long *mem_start,
1658 dt_push_token(soff, mem_start, mem_end); 1678 dt_push_token(soff, mem_start, mem_end);
1659 1679
1660 /* push property content */ 1680 /* push property content */
1661 align = (l >= 8) ? 8 : 4; 1681 valp = make_room(mem_start, mem_end, l, 4);
1662 valp = make_room(mem_start, mem_end, l, align); 1682 call_prom("getprop", 4, 1, node, RELOC(pname), valp, l);
1663 call_prom("getprop", 4, 1, node, pname, valp, l);
1664 *mem_start = _ALIGN(*mem_start, 4); 1683 *mem_start = _ALIGN(*mem_start, 4);
1665 } 1684 }
1666 1685
1667 /* Add a "linux,phandle" property. */ 1686 /* Add a "linux,phandle" property. */
1668 soff = dt_find_string(RELOC("linux,phandle")); 1687 soff = dt_find_string(RELOC("linux,phandle"));
1669 if (soff == 0) 1688 if (soff == 0)
1670 prom_printf("WARNING: Can't find string index for <linux-phandle>" 1689 prom_printf("WARNING: Can't find string index for"
1671 " node %s\n", path); 1690 " <linux-phandle> node %s\n", path);
1672 else { 1691 else {
1673 dt_push_token(OF_DT_PROP, mem_start, mem_end); 1692 dt_push_token(OF_DT_PROP, mem_start, mem_end);
1674 dt_push_token(4, mem_start, mem_end); 1693 dt_push_token(4, mem_start, mem_end);
@@ -1679,7 +1698,7 @@ static void __init scan_dt_build_struct(phandle node, unsigned long *mem_start,
1679 1698
1680 /* do all our children */ 1699 /* do all our children */
1681 child = call_prom("child", 1, 1, node); 1700 child = call_prom("child", 1, 1, node);
1682 while (child != (phandle)0) { 1701 while (child != 0) {
1683 scan_dt_build_struct(child, mem_start, mem_end); 1702 scan_dt_build_struct(child, mem_start, mem_end);
1684 child = call_prom("peer", 1, 1, child); 1703 child = call_prom("peer", 1, 1, child);
1685 } 1704 }
@@ -1718,7 +1737,8 @@ static void __init flatten_device_tree(void)
1718 1737
1719 /* Build header and make room for mem rsv map */ 1738 /* Build header and make room for mem rsv map */
1720 mem_start = _ALIGN(mem_start, 4); 1739 mem_start = _ALIGN(mem_start, 4);
1721 hdr = make_room(&mem_start, &mem_end, sizeof(struct boot_param_header), 4); 1740 hdr = make_room(&mem_start, &mem_end,
1741 sizeof(struct boot_param_header), 4);
1722 RELOC(dt_header_start) = (unsigned long)hdr; 1742 RELOC(dt_header_start) = (unsigned long)hdr;
1723 rsvmap = make_room(&mem_start, &mem_end, sizeof(mem_reserve_map), 8); 1743 rsvmap = make_room(&mem_start, &mem_end, sizeof(mem_reserve_map), 8);
1724 1744
@@ -1731,11 +1751,11 @@ static void __init flatten_device_tree(void)
1731 namep = make_room(&mem_start, &mem_end, 16, 1); 1751 namep = make_room(&mem_start, &mem_end, 16, 1);
1732 strcpy(namep, RELOC("linux,phandle")); 1752 strcpy(namep, RELOC("linux,phandle"));
1733 mem_start = (unsigned long)namep + strlen(namep) + 1; 1753 mem_start = (unsigned long)namep + strlen(namep) + 1;
1734 RELOC(dt_string_end) = mem_start;
1735 1754
1736 /* Build string array */ 1755 /* Build string array */
1737 prom_printf("Building dt strings...\n"); 1756 prom_printf("Building dt strings...\n");
1738 scan_dt_build_strings(root, &mem_start, &mem_end); 1757 scan_dt_build_strings(root, &mem_start, &mem_end);
1758 RELOC(dt_string_end) = mem_start;
1739 1759
1740 /* Build structure */ 1760 /* Build structure */
1741 mem_start = PAGE_ALIGN(mem_start); 1761 mem_start = PAGE_ALIGN(mem_start);
@@ -1750,9 +1770,11 @@ static void __init flatten_device_tree(void)
1750 hdr->totalsize = RELOC(dt_struct_end) - RELOC(dt_header_start); 1770 hdr->totalsize = RELOC(dt_struct_end) - RELOC(dt_header_start);
1751 hdr->off_dt_struct = RELOC(dt_struct_start) - RELOC(dt_header_start); 1771 hdr->off_dt_struct = RELOC(dt_struct_start) - RELOC(dt_header_start);
1752 hdr->off_dt_strings = RELOC(dt_string_start) - RELOC(dt_header_start); 1772 hdr->off_dt_strings = RELOC(dt_string_start) - RELOC(dt_header_start);
1773 hdr->dt_strings_size = RELOC(dt_string_end) - RELOC(dt_string_start);
1753 hdr->off_mem_rsvmap = ((unsigned long)rsvmap) - RELOC(dt_header_start); 1774 hdr->off_mem_rsvmap = ((unsigned long)rsvmap) - RELOC(dt_header_start);
1754 hdr->version = OF_DT_VERSION; 1775 hdr->version = OF_DT_VERSION;
1755 hdr->last_comp_version = 1; 1776 /* Version 16 is not backward compatible */
1777 hdr->last_comp_version = 0x10;
1756 1778
1757 /* Reserve the whole thing and copy the reserve map in, we 1779 /* Reserve the whole thing and copy the reserve map in, we
1758 * also bump mem_reserve_cnt to cause further reservations to 1780 * also bump mem_reserve_cnt to cause further reservations to
@@ -1808,6 +1830,9 @@ static void __init fixup_device_tree(void)
1808 /* does it need fixup ? */ 1830 /* does it need fixup ? */
1809 if (prom_getproplen(i2c, "interrupts") > 0) 1831 if (prom_getproplen(i2c, "interrupts") > 0)
1810 return; 1832 return;
1833
1834 prom_printf("fixing up bogus interrupts for u3 i2c...\n");
1835
1811 /* interrupt on this revision of u3 is number 0 and level */ 1836 /* interrupt on this revision of u3 is number 0 and level */
1812 interrupts[0] = 0; 1837 interrupts[0] = 0;
1813 interrupts[1] = 1; 1838 interrupts[1] = 1;
diff --git a/arch/ppc64/kernel/rtas_pci.c b/arch/ppc64/kernel/rtas_pci.c
index 1048817befb8..1dccadaddd1d 100644
--- a/arch/ppc64/kernel/rtas_pci.c
+++ b/arch/ppc64/kernel/rtas_pci.c
@@ -58,6 +58,21 @@ static int config_access_valid(struct device_node *dn, int where)
58 return 0; 58 return 0;
59} 59}
60 60
61static int of_device_available(struct device_node * dn)
62{
63 char * status;
64
65 status = get_property(dn, "status", NULL);
66
67 if (!status)
68 return 1;
69
70 if (!strcmp(status, "okay"))
71 return 1;
72
73 return 0;
74}
75
61static int rtas_read_config(struct device_node *dn, int where, int size, u32 *val) 76static int rtas_read_config(struct device_node *dn, int where, int size, u32 *val)
62{ 77{
63 int returnval = -1; 78 int returnval = -1;
@@ -103,7 +118,7 @@ static int rtas_pci_read_config(struct pci_bus *bus,
103 118
104 /* Search only direct children of the bus */ 119 /* Search only direct children of the bus */
105 for (dn = busdn->child; dn; dn = dn->sibling) 120 for (dn = busdn->child; dn; dn = dn->sibling)
106 if (dn->devfn == devfn) 121 if (dn->devfn == devfn && of_device_available(dn))
107 return rtas_read_config(dn, where, size, val); 122 return rtas_read_config(dn, where, size, val);
108 return PCIBIOS_DEVICE_NOT_FOUND; 123 return PCIBIOS_DEVICE_NOT_FOUND;
109} 124}
@@ -146,7 +161,7 @@ static int rtas_pci_write_config(struct pci_bus *bus,
146 161
147 /* Search only direct children of the bus */ 162 /* Search only direct children of the bus */
148 for (dn = busdn->child; dn; dn = dn->sibling) 163 for (dn = busdn->child; dn; dn = dn->sibling)
149 if (dn->devfn == devfn) 164 if (dn->devfn == devfn && of_device_available(dn))
150 return rtas_write_config(dn, where, size, val); 165 return rtas_write_config(dn, where, size, val);
151 return PCIBIOS_DEVICE_NOT_FOUND; 166 return PCIBIOS_DEVICE_NOT_FOUND;
152} 167}
diff --git a/arch/ppc64/kernel/setup.c b/arch/ppc64/kernel/setup.c
index 687e85595208..ee3b20de2e7a 100644
--- a/arch/ppc64/kernel/setup.c
+++ b/arch/ppc64/kernel/setup.c
@@ -536,15 +536,19 @@ static void __init check_for_initrd(void)
536 536
537 DBG(" -> check_for_initrd()\n"); 537 DBG(" -> check_for_initrd()\n");
538 538
539 prop = (u64 *)get_property(of_chosen, "linux,initrd-start", NULL); 539 if (of_chosen) {
540 if (prop != NULL) { 540 prop = (u64 *)get_property(of_chosen,
541 initrd_start = (unsigned long)__va(*prop); 541 "linux,initrd-start", NULL);
542 prop = (u64 *)get_property(of_chosen, "linux,initrd-end", NULL);
543 if (prop != NULL) { 542 if (prop != NULL) {
544 initrd_end = (unsigned long)__va(*prop); 543 initrd_start = (unsigned long)__va(*prop);
545 initrd_below_start_ok = 1; 544 prop = (u64 *)get_property(of_chosen,
546 } else 545 "linux,initrd-end", NULL);
547 initrd_start = 0; 546 if (prop != NULL) {
547 initrd_end = (unsigned long)__va(*prop);
548 initrd_below_start_ok = 1;
549 } else
550 initrd_start = 0;
551 }
548 } 552 }
549 553
550 /* If we were passed an initrd, set the ROOT_DEV properly if the values 554 /* If we were passed an initrd, set the ROOT_DEV properly if the values
@@ -627,7 +631,7 @@ void __init setup_system(void)
627 * Initialize xmon 631 * Initialize xmon
628 */ 632 */
629#ifdef CONFIG_XMON_DEFAULT 633#ifdef CONFIG_XMON_DEFAULT
630 xmon_init(); 634 xmon_init(1);
631#endif 635#endif
632 /* 636 /*
633 * Register early console 637 * Register early console
@@ -706,6 +710,8 @@ void machine_power_off(void)
706 local_irq_disable(); 710 local_irq_disable();
707 while (1) ; 711 while (1) ;
708} 712}
713/* Used by the G5 thermal driver */
714EXPORT_SYMBOL_GPL(machine_power_off);
709 715
710void machine_halt(void) 716void machine_halt(void)
711{ 717{
@@ -1341,11 +1347,13 @@ static int __init early_xmon(char *p)
1341 /* ensure xmon is enabled */ 1347 /* ensure xmon is enabled */
1342 if (p) { 1348 if (p) {
1343 if (strncmp(p, "on", 2) == 0) 1349 if (strncmp(p, "on", 2) == 0)
1344 xmon_init(); 1350 xmon_init(1);
1351 if (strncmp(p, "off", 3) == 0)
1352 xmon_init(0);
1345 if (strncmp(p, "early", 5) != 0) 1353 if (strncmp(p, "early", 5) != 0)
1346 return 0; 1354 return 0;
1347 } 1355 }
1348 xmon_init(); 1356 xmon_init(1);
1349 debugger(NULL); 1357 debugger(NULL);
1350 1358
1351 return 0; 1359 return 0;
diff --git a/arch/ppc64/kernel/signal.c b/arch/ppc64/kernel/signal.c
index bf782276984c..49a79a55c32d 100644
--- a/arch/ppc64/kernel/signal.c
+++ b/arch/ppc64/kernel/signal.c
@@ -481,10 +481,11 @@ static int handle_signal(unsigned long sig, struct k_sigaction *ka,
481 /* Set up Signal Frame */ 481 /* Set up Signal Frame */
482 ret = setup_rt_frame(sig, ka, info, oldset, regs); 482 ret = setup_rt_frame(sig, ka, info, oldset, regs);
483 483
484 if (ret && !(ka->sa.sa_flags & SA_NODEFER)) { 484 if (ret) {
485 spin_lock_irq(&current->sighand->siglock); 485 spin_lock_irq(&current->sighand->siglock);
486 sigorsets(&current->blocked, &current->blocked, &ka->sa.sa_mask); 486 sigorsets(&current->blocked, &current->blocked, &ka->sa.sa_mask);
487 sigaddset(&current->blocked,sig); 487 if (!(ka->sa.sa_flags & SA_NODEFER))
488 sigaddset(&current->blocked,sig);
488 recalc_sigpending(); 489 recalc_sigpending();
489 spin_unlock_irq(&current->sighand->siglock); 490 spin_unlock_irq(&current->sighand->siglock);
490 } 491 }
diff --git a/arch/ppc64/kernel/signal32.c b/arch/ppc64/kernel/signal32.c
index 3c2fa5c284c0..46f4d6cc7fc9 100644
--- a/arch/ppc64/kernel/signal32.c
+++ b/arch/ppc64/kernel/signal32.c
@@ -976,11 +976,12 @@ int do_signal32(sigset_t *oldset, struct pt_regs *regs)
976 else 976 else
977 ret = handle_signal32(signr, &ka, &info, oldset, regs, newsp); 977 ret = handle_signal32(signr, &ka, &info, oldset, regs, newsp);
978 978
979 if (ret && !(ka.sa.sa_flags & SA_NODEFER)) { 979 if (ret) {
980 spin_lock_irq(&current->sighand->siglock); 980 spin_lock_irq(&current->sighand->siglock);
981 sigorsets(&current->blocked, &current->blocked, 981 sigorsets(&current->blocked, &current->blocked,
982 &ka.sa.sa_mask); 982 &ka.sa.sa_mask);
983 sigaddset(&current->blocked, signr); 983 if (!(ka.sa.sa_flags & SA_NODEFER))
984 sigaddset(&current->blocked, signr);
984 recalc_sigpending(); 985 recalc_sigpending();
985 spin_unlock_irq(&current->sighand->siglock); 986 spin_unlock_irq(&current->sighand->siglock);
986 } 987 }
diff --git a/arch/ppc64/kernel/sysfs.c b/arch/ppc64/kernel/sysfs.c
index 02b8ac4e0168..f311ee7c0070 100644
--- a/arch/ppc64/kernel/sysfs.c
+++ b/arch/ppc64/kernel/sysfs.c
@@ -13,6 +13,7 @@
13#include <asm/current.h> 13#include <asm/current.h>
14#include <asm/processor.h> 14#include <asm/processor.h>
15#include <asm/cputable.h> 15#include <asm/cputable.h>
16#include <asm/firmware.h>
16#include <asm/hvcall.h> 17#include <asm/hvcall.h>
17#include <asm/prom.h> 18#include <asm/prom.h>
18#include <asm/systemcfg.h> 19#include <asm/systemcfg.h>
@@ -100,6 +101,8 @@ static int __init setup_smt_snooze_delay(char *str)
100} 101}
101__setup("smt-snooze-delay=", setup_smt_snooze_delay); 102__setup("smt-snooze-delay=", setup_smt_snooze_delay);
102 103
104#endif /* CONFIG_PPC_MULTIPLATFORM */
105
103/* 106/*
104 * Enabling PMCs will slow partition context switch times so we only do 107 * Enabling PMCs will slow partition context switch times so we only do
105 * it the first time we write to the PMCs. 108 * it the first time we write to the PMCs.
@@ -109,65 +112,15 @@ static DEFINE_PER_CPU(char, pmcs_enabled);
109 112
110void ppc64_enable_pmcs(void) 113void ppc64_enable_pmcs(void)
111{ 114{
112 unsigned long hid0;
113#ifdef CONFIG_PPC_PSERIES
114 unsigned long set, reset;
115#endif /* CONFIG_PPC_PSERIES */
116
117 /* Only need to enable them once */ 115 /* Only need to enable them once */
118 if (__get_cpu_var(pmcs_enabled)) 116 if (__get_cpu_var(pmcs_enabled))
119 return; 117 return;
120 118
121 __get_cpu_var(pmcs_enabled) = 1; 119 __get_cpu_var(pmcs_enabled) = 1;
122 120
123 switch (systemcfg->platform) { 121 if (ppc_md.enable_pmcs)
124 case PLATFORM_PSERIES: 122 ppc_md.enable_pmcs();
125 case PLATFORM_POWERMAC:
126 hid0 = mfspr(HID0);
127 hid0 |= 1UL << (63 - 20);
128
129 /* POWER4 requires the following sequence */
130 asm volatile(
131 "sync\n"
132 "mtspr %1, %0\n"
133 "mfspr %0, %1\n"
134 "mfspr %0, %1\n"
135 "mfspr %0, %1\n"
136 "mfspr %0, %1\n"
137 "mfspr %0, %1\n"
138 "mfspr %0, %1\n"
139 "isync" : "=&r" (hid0) : "i" (HID0), "0" (hid0):
140 "memory");
141 break;
142
143#ifdef CONFIG_PPC_PSERIES
144 case PLATFORM_PSERIES_LPAR:
145 set = 1UL << 63;
146 reset = 0;
147 plpar_hcall_norets(H_PERFMON, set, reset);
148 break;
149#endif /* CONFIG_PPC_PSERIES */
150
151 default:
152 break;
153 }
154
155#ifdef CONFIG_PPC_PSERIES
156 /* instruct hypervisor to maintain PMCs */
157 if (cur_cpu_spec->firmware_features & FW_FEATURE_SPLPAR)
158 get_paca()->lppaca.pmcregs_in_use = 1;
159#endif /* CONFIG_PPC_PSERIES */
160} 123}
161
162#else
163
164/* PMC stuff */
165void ppc64_enable_pmcs(void)
166{
167 /* XXX Implement for iseries */
168}
169#endif /* CONFIG_PPC_MULTIPLATFORM */
170
171EXPORT_SYMBOL(ppc64_enable_pmcs); 124EXPORT_SYMBOL(ppc64_enable_pmcs);
172 125
173/* XXX convert to rusty's on_one_cpu */ 126/* XXX convert to rusty's on_one_cpu */
diff --git a/arch/ppc64/kernel/time.c b/arch/ppc64/kernel/time.c
index 909462e1adea..1696e1b05bb9 100644
--- a/arch/ppc64/kernel/time.c
+++ b/arch/ppc64/kernel/time.c
@@ -67,6 +67,7 @@
67#include <asm/prom.h> 67#include <asm/prom.h>
68#include <asm/sections.h> 68#include <asm/sections.h>
69#include <asm/systemcfg.h> 69#include <asm/systemcfg.h>
70#include <asm/firmware.h>
70 71
71u64 jiffies_64 __cacheline_aligned_in_smp = INITIAL_JIFFIES; 72u64 jiffies_64 __cacheline_aligned_in_smp = INITIAL_JIFFIES;
72 73
@@ -370,13 +371,11 @@ int timer_interrupt(struct pt_regs * regs)
370 process_hvlpevents(regs); 371 process_hvlpevents(regs);
371#endif 372#endif
372 373
373/* collect purr register values often, for accurate calculations */ 374 /* collect purr register values often, for accurate calculations */
374#if defined(CONFIG_PPC_PSERIES) 375 if (firmware_has_feature(FW_FEATURE_SPLPAR)) {
375 if (cur_cpu_spec->firmware_features & FW_FEATURE_SPLPAR) {
376 struct cpu_usage *cu = &__get_cpu_var(cpu_usage_array); 376 struct cpu_usage *cu = &__get_cpu_var(cpu_usage_array);
377 cu->current_tb = mfspr(SPRN_PURR); 377 cu->current_tb = mfspr(SPRN_PURR);
378 } 378 }
379#endif
380 379
381 irq_exit(); 380 irq_exit();
382 381
diff --git a/arch/ppc64/kernel/vio.c b/arch/ppc64/kernel/vio.c
index 0c0ba71ac0e8..c90e1dd875ce 100644
--- a/arch/ppc64/kernel/vio.c
+++ b/arch/ppc64/kernel/vio.c
@@ -1,10 +1,11 @@
1/* 1/*
2 * IBM PowerPC Virtual I/O Infrastructure Support. 2 * IBM PowerPC Virtual I/O Infrastructure Support.
3 * 3 *
4 * Copyright (c) 2003 IBM Corp. 4 * Copyright (c) 2003-2005 IBM Corp.
5 * Dave Engebretsen engebret@us.ibm.com 5 * Dave Engebretsen engebret@us.ibm.com
6 * Santiago Leon santil@us.ibm.com 6 * Santiago Leon santil@us.ibm.com
7 * Hollis Blanchard <hollisb@us.ibm.com> 7 * Hollis Blanchard <hollisb@us.ibm.com>
8 * Stephen Rothwell
8 * 9 *
9 * This program is free software; you can redistribute it and/or 10 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License 11 * modify it under the terms of the GNU General Public License
@@ -14,61 +15,30 @@
14 15
15#include <linux/init.h> 16#include <linux/init.h>
16#include <linux/console.h> 17#include <linux/console.h>
17#include <linux/version.h>
18#include <linux/module.h> 18#include <linux/module.h>
19#include <linux/kobject.h>
20#include <linux/mm.h> 19#include <linux/mm.h>
21#include <linux/dma-mapping.h> 20#include <linux/dma-mapping.h>
22#include <asm/rtas.h>
23#include <asm/iommu.h> 21#include <asm/iommu.h>
24#include <asm/dma.h> 22#include <asm/dma.h>
25#include <asm/ppcdebug.h>
26#include <asm/vio.h> 23#include <asm/vio.h>
27#include <asm/hvcall.h>
28#include <asm/iSeries/vio.h>
29#include <asm/iSeries/HvTypes.h>
30#include <asm/iSeries/HvCallXm.h>
31#include <asm/iSeries/HvLpConfig.h>
32
33#define DBGENTER() pr_debug("%s entered\n", __FUNCTION__)
34
35extern struct subsystem devices_subsys; /* needed for vio_find_name() */
36 24
37static const struct vio_device_id *vio_match_device( 25static const struct vio_device_id *vio_match_device(
38 const struct vio_device_id *, const struct vio_dev *); 26 const struct vio_device_id *, const struct vio_dev *);
39 27
40#ifdef CONFIG_PPC_PSERIES 28struct vio_dev vio_bus_device = { /* fake "parent" device */
41static struct iommu_table *vio_build_iommu_table(struct vio_dev *);
42static int vio_num_address_cells;
43#endif
44#ifdef CONFIG_PPC_ISERIES
45static struct iommu_table veth_iommu_table;
46static struct iommu_table vio_iommu_table;
47#endif
48static struct vio_dev vio_bus_device = { /* fake "parent" device */
49 .name = vio_bus_device.dev.bus_id, 29 .name = vio_bus_device.dev.bus_id,
50 .type = "", 30 .type = "",
51#ifdef CONFIG_PPC_ISERIES
52 .iommu_table = &vio_iommu_table,
53#endif
54 .dev.bus_id = "vio", 31 .dev.bus_id = "vio",
55 .dev.bus = &vio_bus_type, 32 .dev.bus = &vio_bus_type,
56}; 33};
57 34
58#ifdef CONFIG_PPC_ISERIES 35static struct vio_bus_ops vio_bus_ops;
59static struct vio_dev *__init vio_register_device_iseries(char *type,
60 uint32_t unit_num);
61
62struct device *iSeries_vio_dev = &vio_bus_device.dev;
63EXPORT_SYMBOL(iSeries_vio_dev);
64
65#define device_is_compatible(a, b) 1
66 36
67#endif 37/*
68 38 * Convert from struct device to struct vio_dev and pass to driver.
69/* convert from struct device to struct vio_dev and pass to driver.
70 * dev->driver has already been set by generic code because vio_bus_match 39 * dev->driver has already been set by generic code because vio_bus_match
71 * succeeded. */ 40 * succeeded.
41 */
72static int vio_bus_probe(struct device *dev) 42static int vio_bus_probe(struct device *dev)
73{ 43{
74 struct vio_dev *viodev = to_vio_dev(dev); 44 struct vio_dev *viodev = to_vio_dev(dev);
@@ -76,15 +46,12 @@ static int vio_bus_probe(struct device *dev)
76 const struct vio_device_id *id; 46 const struct vio_device_id *id;
77 int error = -ENODEV; 47 int error = -ENODEV;
78 48
79 DBGENTER();
80
81 if (!viodrv->probe) 49 if (!viodrv->probe)
82 return error; 50 return error;
83 51
84 id = vio_match_device(viodrv->id_table, viodev); 52 id = vio_match_device(viodrv->id_table, viodev);
85 if (id) { 53 if (id)
86 error = viodrv->probe(viodev, id); 54 error = viodrv->probe(viodev, id);
87 }
88 55
89 return error; 56 return error;
90} 57}
@@ -95,11 +62,8 @@ static int vio_bus_remove(struct device *dev)
95 struct vio_dev *viodev = to_vio_dev(dev); 62 struct vio_dev *viodev = to_vio_dev(dev);
96 struct vio_driver *viodrv = to_vio_driver(dev->driver); 63 struct vio_driver *viodrv = to_vio_driver(dev->driver);
97 64
98 DBGENTER(); 65 if (viodrv->remove)
99
100 if (viodrv->remove) {
101 return viodrv->remove(viodev); 66 return viodrv->remove(viodev);
102 }
103 67
104 /* driver can't remove */ 68 /* driver can't remove */
105 return 1; 69 return 1;
@@ -135,193 +99,72 @@ void vio_unregister_driver(struct vio_driver *viodrv)
135EXPORT_SYMBOL(vio_unregister_driver); 99EXPORT_SYMBOL(vio_unregister_driver);
136 100
137/** 101/**
138 * vio_match_device: - Tell if a VIO device has a matching VIO device id structure. 102 * vio_match_device: - Tell if a VIO device has a matching
139 * @ids: array of VIO device id structures to search in 103 * VIO device id structure.
140 * @dev: the VIO device structure to match against 104 * @ids: array of VIO device id structures to search in
105 * @dev: the VIO device structure to match against
141 * 106 *
142 * Used by a driver to check whether a VIO device present in the 107 * Used by a driver to check whether a VIO device present in the
143 * system is in its list of supported devices. Returns the matching 108 * system is in its list of supported devices. Returns the matching
144 * vio_device_id structure or NULL if there is no match. 109 * vio_device_id structure or NULL if there is no match.
145 */ 110 */
146static const struct vio_device_id * vio_match_device(const struct vio_device_id *ids, 111static const struct vio_device_id *vio_match_device(
147 const struct vio_dev *dev) 112 const struct vio_device_id *ids, const struct vio_dev *dev)
148{ 113{
149 DBGENTER(); 114 while (ids->type[0] != '\0') {
150 115 if (vio_bus_ops.match(ids, dev))
151 while (ids->type) {
152 if ((strncmp(dev->type, ids->type, strlen(ids->type)) == 0) &&
153 device_is_compatible(dev->dev.platform_data, ids->compat))
154 return ids; 116 return ids;
155 ids++; 117 ids++;
156 } 118 }
157 return NULL; 119 return NULL;
158} 120}
159 121
160#ifdef CONFIG_PPC_ISERIES
161void __init iommu_vio_init(void)
162{
163 struct iommu_table *t;
164 struct iommu_table_cb cb;
165 unsigned long cbp;
166 unsigned long itc_entries;
167
168 cb.itc_busno = 255; /* Bus 255 is the virtual bus */
169 cb.itc_virtbus = 0xff; /* Ask for virtual bus */
170
171 cbp = virt_to_abs(&cb);
172 HvCallXm_getTceTableParms(cbp);
173
174 itc_entries = cb.itc_size * PAGE_SIZE / sizeof(union tce_entry);
175 veth_iommu_table.it_size = itc_entries / 2;
176 veth_iommu_table.it_busno = cb.itc_busno;
177 veth_iommu_table.it_offset = cb.itc_offset;
178 veth_iommu_table.it_index = cb.itc_index;
179 veth_iommu_table.it_type = TCE_VB;
180 veth_iommu_table.it_blocksize = 1;
181
182 t = iommu_init_table(&veth_iommu_table);
183
184 if (!t)
185 printk("Virtual Bus VETH TCE table failed.\n");
186
187 vio_iommu_table.it_size = itc_entries - veth_iommu_table.it_size;
188 vio_iommu_table.it_busno = cb.itc_busno;
189 vio_iommu_table.it_offset = cb.itc_offset +
190 veth_iommu_table.it_size;
191 vio_iommu_table.it_index = cb.itc_index;
192 vio_iommu_table.it_type = TCE_VB;
193 vio_iommu_table.it_blocksize = 1;
194
195 t = iommu_init_table(&vio_iommu_table);
196
197 if (!t)
198 printk("Virtual Bus VIO TCE table failed.\n");
199}
200#endif
201
202#ifdef CONFIG_PPC_PSERIES
203static void probe_bus_pseries(void)
204{
205 struct device_node *node_vroot, *of_node;
206
207 node_vroot = find_devices("vdevice");
208 if ((node_vroot == NULL) || (node_vroot->child == NULL))
209 /* this machine doesn't do virtual IO, and that's ok */
210 return;
211
212 vio_num_address_cells = prom_n_addr_cells(node_vroot->child);
213
214 /*
215 * Create struct vio_devices for each virtual device in the device tree.
216 * Drivers will associate with them later.
217 */
218 for (of_node = node_vroot->child; of_node != NULL;
219 of_node = of_node->sibling) {
220 printk(KERN_DEBUG "%s: processing %p\n", __FUNCTION__, of_node);
221 vio_register_device_node(of_node);
222 }
223}
224#endif
225
226#ifdef CONFIG_PPC_ISERIES
227static void probe_bus_iseries(void)
228{
229 HvLpIndexMap vlan_map = HvLpConfig_getVirtualLanIndexMap();
230 struct vio_dev *viodev;
231 int i;
232
233 /* there is only one of each of these */
234 vio_register_device_iseries("viocons", 0);
235 vio_register_device_iseries("vscsi", 0);
236
237 vlan_map = HvLpConfig_getVirtualLanIndexMap();
238 for (i = 0; i < HVMAXARCHITECTEDVIRTUALLANS; i++) {
239 if ((vlan_map & (0x8000 >> i)) == 0)
240 continue;
241 viodev = vio_register_device_iseries("vlan", i);
242 /* veth is special and has it own iommu_table */
243 viodev->iommu_table = &veth_iommu_table;
244 }
245 for (i = 0; i < HVMAXARCHITECTEDVIRTUALDISKS; i++)
246 vio_register_device_iseries("viodasd", i);
247 for (i = 0; i < HVMAXARCHITECTEDVIRTUALCDROMS; i++)
248 vio_register_device_iseries("viocd", i);
249 for (i = 0; i < HVMAXARCHITECTEDVIRTUALTAPES; i++)
250 vio_register_device_iseries("viotape", i);
251}
252#endif
253
254/** 122/**
255 * vio_bus_init: - Initialize the virtual IO bus 123 * vio_bus_init: - Initialize the virtual IO bus
256 */ 124 */
257static int __init vio_bus_init(void) 125int __init vio_bus_init(struct vio_bus_ops *ops)
258{ 126{
259 int err; 127 int err;
260 128
129 vio_bus_ops = *ops;
130
261 err = bus_register(&vio_bus_type); 131 err = bus_register(&vio_bus_type);
262 if (err) { 132 if (err) {
263 printk(KERN_ERR "failed to register VIO bus\n"); 133 printk(KERN_ERR "failed to register VIO bus\n");
264 return err; 134 return err;
265 } 135 }
266 136
267 /* the fake parent of all vio devices, just to give us a nice directory */ 137 /*
138 * The fake parent of all vio devices, just to give us
139 * a nice directory
140 */
268 err = device_register(&vio_bus_device.dev); 141 err = device_register(&vio_bus_device.dev);
269 if (err) { 142 if (err) {
270 printk(KERN_WARNING "%s: device_register returned %i\n", __FUNCTION__, 143 printk(KERN_WARNING "%s: device_register returned %i\n",
271 err); 144 __FUNCTION__, err);
272 return err; 145 return err;
273 } 146 }
274 147
275#ifdef CONFIG_PPC_PSERIES
276 probe_bus_pseries();
277#endif
278#ifdef CONFIG_PPC_ISERIES
279 probe_bus_iseries();
280#endif
281
282 return 0; 148 return 0;
283} 149}
284 150
285__initcall(vio_bus_init);
286
287/* vio_dev refcount hit 0 */ 151/* vio_dev refcount hit 0 */
288static void __devinit vio_dev_release(struct device *dev) 152static void __devinit vio_dev_release(struct device *dev)
289{ 153{
290 DBGENTER(); 154 if (vio_bus_ops.release_device)
291 155 vio_bus_ops.release_device(dev);
292#ifdef CONFIG_PPC_PSERIES
293 /* XXX free TCE table */
294 of_node_put(dev->platform_data);
295#endif
296 kfree(to_vio_dev(dev)); 156 kfree(to_vio_dev(dev));
297} 157}
298 158
299#ifdef CONFIG_PPC_PSERIES 159static ssize_t viodev_show_name(struct device *dev,
300static ssize_t viodev_show_devspec(struct device *dev, struct device_attribute *attr, char *buf) 160 struct device_attribute *attr, char *buf)
301{
302 struct device_node *of_node = dev->platform_data;
303
304 return sprintf(buf, "%s\n", of_node->full_name);
305}
306DEVICE_ATTR(devspec, S_IRUSR | S_IRGRP | S_IROTH, viodev_show_devspec, NULL);
307#endif
308
309static ssize_t viodev_show_name(struct device *dev, struct device_attribute *attr, char *buf)
310{ 161{
311 return sprintf(buf, "%s\n", to_vio_dev(dev)->name); 162 return sprintf(buf, "%s\n", to_vio_dev(dev)->name);
312} 163}
313DEVICE_ATTR(name, S_IRUSR | S_IRGRP | S_IROTH, viodev_show_name, NULL); 164DEVICE_ATTR(name, S_IRUSR | S_IRGRP | S_IROTH, viodev_show_name, NULL);
314 165
315static struct vio_dev * __devinit vio_register_device_common( 166struct vio_dev * __devinit vio_register_device(struct vio_dev *viodev)
316 struct vio_dev *viodev, char *name, char *type,
317 uint32_t unit_address, struct iommu_table *iommu_table)
318{ 167{
319 DBGENTER();
320
321 viodev->name = name;
322 viodev->type = type;
323 viodev->unit_address = unit_address;
324 viodev->iommu_table = iommu_table;
325 /* init generic 'struct device' fields: */ 168 /* init generic 'struct device' fields: */
326 viodev->dev.parent = &vio_bus_device.dev; 169 viodev->dev.parent = &vio_bus_device.dev;
327 viodev->dev.bus = &vio_bus_type; 170 viodev->dev.bus = &vio_bus_type;
@@ -338,222 +181,15 @@ static struct vio_dev * __devinit vio_register_device_common(
338 return viodev; 181 return viodev;
339} 182}
340 183
341#ifdef CONFIG_PPC_PSERIES
342/**
343 * vio_register_device_node: - Register a new vio device.
344 * @of_node: The OF node for this device.
345 *
346 * Creates and initializes a vio_dev structure from the data in
347 * of_node (dev.platform_data) and adds it to the list of virtual devices.
348 * Returns a pointer to the created vio_dev or NULL if node has
349 * NULL device_type or compatible fields.
350 */
351struct vio_dev * __devinit vio_register_device_node(struct device_node *of_node)
352{
353 struct vio_dev *viodev;
354 unsigned int *unit_address;
355 unsigned int *irq_p;
356
357 DBGENTER();
358
359 /* we need the 'device_type' property, in order to match with drivers */
360 if ((NULL == of_node->type)) {
361 printk(KERN_WARNING
362 "%s: node %s missing 'device_type'\n", __FUNCTION__,
363 of_node->name ? of_node->name : "<unknown>");
364 return NULL;
365 }
366
367 unit_address = (unsigned int *)get_property(of_node, "reg", NULL);
368 if (!unit_address) {
369 printk(KERN_WARNING "%s: node %s missing 'reg'\n", __FUNCTION__,
370 of_node->name ? of_node->name : "<unknown>");
371 return NULL;
372 }
373
374 /* allocate a vio_dev for this node */
375 viodev = kmalloc(sizeof(struct vio_dev), GFP_KERNEL);
376 if (!viodev) {
377 return NULL;
378 }
379 memset(viodev, 0, sizeof(struct vio_dev));
380
381 viodev->dev.platform_data = of_node_get(of_node);
382
383 viodev->irq = NO_IRQ;
384 irq_p = (unsigned int *)get_property(of_node, "interrupts", NULL);
385 if (irq_p) {
386 int virq = virt_irq_create_mapping(*irq_p);
387 if (virq == NO_IRQ) {
388 printk(KERN_ERR "Unable to allocate interrupt "
389 "number for %s\n", of_node->full_name);
390 } else
391 viodev->irq = irq_offset_up(virq);
392 }
393
394 snprintf(viodev->dev.bus_id, BUS_ID_SIZE, "%x", *unit_address);
395
396 /* register with generic device framework */
397 if (vio_register_device_common(viodev, of_node->name, of_node->type,
398 *unit_address, vio_build_iommu_table(viodev))
399 == NULL) {
400 /* XXX free TCE table */
401 kfree(viodev);
402 return NULL;
403 }
404 device_create_file(&viodev->dev, &dev_attr_devspec);
405
406 return viodev;
407}
408EXPORT_SYMBOL(vio_register_device_node);
409#endif
410
411#ifdef CONFIG_PPC_ISERIES
412/**
413 * vio_register_device: - Register a new vio device.
414 * @voidev: The device to register.
415 */
416static struct vio_dev *__init vio_register_device_iseries(char *type,
417 uint32_t unit_num)
418{
419 struct vio_dev *viodev;
420
421 DBGENTER();
422
423 /* allocate a vio_dev for this node */
424 viodev = kmalloc(sizeof(struct vio_dev), GFP_KERNEL);
425 if (!viodev)
426 return NULL;
427 memset(viodev, 0, sizeof(struct vio_dev));
428
429 snprintf(viodev->dev.bus_id, BUS_ID_SIZE, "%s%d", type, unit_num);
430
431 return vio_register_device_common(viodev, viodev->dev.bus_id, type,
432 unit_num, &vio_iommu_table);
433}
434#endif
435
436void __devinit vio_unregister_device(struct vio_dev *viodev) 184void __devinit vio_unregister_device(struct vio_dev *viodev)
437{ 185{
438 DBGENTER(); 186 if (vio_bus_ops.unregister_device)
439#ifdef CONFIG_PPC_PSERIES 187 vio_bus_ops.unregister_device(viodev);
440 device_remove_file(&viodev->dev, &dev_attr_devspec);
441#endif
442 device_remove_file(&viodev->dev, &dev_attr_name); 188 device_remove_file(&viodev->dev, &dev_attr_name);
443 device_unregister(&viodev->dev); 189 device_unregister(&viodev->dev);
444} 190}
445EXPORT_SYMBOL(vio_unregister_device); 191EXPORT_SYMBOL(vio_unregister_device);
446 192
447#ifdef CONFIG_PPC_PSERIES
448/**
449 * vio_get_attribute: - get attribute for virtual device
450 * @vdev: The vio device to get property.
451 * @which: The property/attribute to be extracted.
452 * @length: Pointer to length of returned data size (unused if NULL).
453 *
454 * Calls prom.c's get_property() to return the value of the
455 * attribute specified by the preprocessor constant @which
456*/
457const void * vio_get_attribute(struct vio_dev *vdev, void* which, int* length)
458{
459 return get_property(vdev->dev.platform_data, (char*)which, length);
460}
461EXPORT_SYMBOL(vio_get_attribute);
462
463/* vio_find_name() - internal because only vio.c knows how we formatted the
464 * kobject name
465 * XXX once vio_bus_type.devices is actually used as a kset in
466 * drivers/base/bus.c, this function should be removed in favor of
467 * "device_find(kobj_name, &vio_bus_type)"
468 */
469static struct vio_dev *vio_find_name(const char *kobj_name)
470{
471 struct kobject *found;
472
473 found = kset_find_obj(&devices_subsys.kset, kobj_name);
474 if (!found)
475 return NULL;
476
477 return to_vio_dev(container_of(found, struct device, kobj));
478}
479
480/**
481 * vio_find_node - find an already-registered vio_dev
482 * @vnode: device_node of the virtual device we're looking for
483 */
484struct vio_dev *vio_find_node(struct device_node *vnode)
485{
486 uint32_t *unit_address;
487 char kobj_name[BUS_ID_SIZE];
488
489 /* construct the kobject name from the device node */
490 unit_address = (uint32_t *)get_property(vnode, "reg", NULL);
491 if (!unit_address)
492 return NULL;
493 snprintf(kobj_name, BUS_ID_SIZE, "%x", *unit_address);
494
495 return vio_find_name(kobj_name);
496}
497EXPORT_SYMBOL(vio_find_node);
498
499/**
500 * vio_build_iommu_table: - gets the dma information from OF and builds the TCE tree.
501 * @dev: the virtual device.
502 *
503 * Returns a pointer to the built tce tree, or NULL if it can't
504 * find property.
505*/
506static struct iommu_table * vio_build_iommu_table(struct vio_dev *dev)
507{
508 unsigned int *dma_window;
509 struct iommu_table *newTceTable;
510 unsigned long offset;
511 int dma_window_property_size;
512
513 dma_window = (unsigned int *) get_property(dev->dev.platform_data, "ibm,my-dma-window", &dma_window_property_size);
514 if(!dma_window) {
515 return NULL;
516 }
517
518 newTceTable = (struct iommu_table *) kmalloc(sizeof(struct iommu_table), GFP_KERNEL);
519
520 /* There should be some code to extract the phys-encoded offset
521 using prom_n_addr_cells(). However, according to a comment
522 on earlier versions, it's always zero, so we don't bother */
523 offset = dma_window[1] >> PAGE_SHIFT;
524
525 /* TCE table size - measured in tce entries */
526 newTceTable->it_size = dma_window[4] >> PAGE_SHIFT;
527 /* offset for VIO should always be 0 */
528 newTceTable->it_offset = offset;
529 newTceTable->it_busno = 0;
530 newTceTable->it_index = (unsigned long)dma_window[0];
531 newTceTable->it_type = TCE_VB;
532
533 return iommu_init_table(newTceTable);
534}
535
536int vio_enable_interrupts(struct vio_dev *dev)
537{
538 int rc = h_vio_signal(dev->unit_address, VIO_IRQ_ENABLE);
539 if (rc != H_Success) {
540 printk(KERN_ERR "vio: Error 0x%x enabling interrupts\n", rc);
541 }
542 return rc;
543}
544EXPORT_SYMBOL(vio_enable_interrupts);
545
546int vio_disable_interrupts(struct vio_dev *dev)
547{
548 int rc = h_vio_signal(dev->unit_address, VIO_IRQ_DISABLE);
549 if (rc != H_Success) {
550 printk(KERN_ERR "vio: Error 0x%x disabling interrupts\n", rc);
551 }
552 return rc;
553}
554EXPORT_SYMBOL(vio_disable_interrupts);
555#endif
556
557static dma_addr_t vio_map_single(struct device *dev, void *vaddr, 193static dma_addr_t vio_map_single(struct device *dev, void *vaddr,
558 size_t size, enum dma_data_direction direction) 194 size_t size, enum dma_data_direction direction)
559{ 195{
@@ -615,18 +251,8 @@ static int vio_bus_match(struct device *dev, struct device_driver *drv)
615 const struct vio_dev *vio_dev = to_vio_dev(dev); 251 const struct vio_dev *vio_dev = to_vio_dev(dev);
616 struct vio_driver *vio_drv = to_vio_driver(drv); 252 struct vio_driver *vio_drv = to_vio_driver(drv);
617 const struct vio_device_id *ids = vio_drv->id_table; 253 const struct vio_device_id *ids = vio_drv->id_table;
618 const struct vio_device_id *found_id;
619
620 DBGENTER();
621 254
622 if (!ids) 255 return (ids != NULL) && (vio_match_device(ids, vio_dev) != NULL);
623 return 0;
624
625 found_id = vio_match_device(ids, vio_dev);
626 if (found_id)
627 return 1;
628
629 return 0;
630} 256}
631 257
632struct bus_type vio_bus_type = { 258struct bus_type vio_bus_type = {