Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
Merge tag 'arc-4.6-rc7-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc
Pull ARC fixes from Vineet Gupta:
"Late in the cycle, but this has fixes for couple of issues: a PAE40
boot crash and Arnd spotting lack of barriers in BE io-accessors.
The 3rd patch for enabling highmem in low physical mem ;-) honestly is
more than a "fix" but its been in works for some time, seems to be
stable in testing and enables 2 of our customers to go forward with
4.6 kernel.
- Fix for PTE truncation in PAE40 builds
- Fix for big endian IO accessors lacking IO barrier
- Allow HIGHMEM to work with low physical addresses"
* tag 'arc-4.6-rc7-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc:
ARC: support HIGHMEM even without PAE40
ARC: Fix PAE40 boot failures due to PTE truncation
ARC: Add missing io barriers to io{read,write}{16,32}be()
+130
-35
+13
arch/arc/Kconfig
+13
arch/arc/Kconfig
···
58
58
config RWSEM_GENERIC_SPINLOCK
59
59
def_bool y
60
60
61
+
config ARCH_DISCONTIGMEM_ENABLE
62
+
def_bool y
63
+
61
64
config ARCH_FLATMEM_ENABLE
62
65
def_bool y
63
66
···
350
347
351
348
endchoice
352
349
350
+
config NODES_SHIFT
351
+
int "Maximum NUMA Nodes (as a power of 2)"
352
+
default "1" if !DISCONTIGMEM
353
+
default "2" if DISCONTIGMEM
354
+
depends on NEED_MULTIPLE_NODES
355
+
---help---
356
+
Accessing memory beyond 1GB (with or w/o PAE) requires 2 memory
357
+
zones.
358
+
353
359
if ISA_ARCOMPACT
354
360
355
361
config ARC_COMPACT_IRQ_LEVELS
···
467
455
468
456
config HIGHMEM
469
457
bool "High Memory Support"
458
+
select DISCONTIGMEM
470
459
help
471
460
With ARC 2G:2G address split, only upper 2G is directly addressable by
472
461
kernel. Enable this to potentially allow access to rest of 2G and PAE
+18
-9
arch/arc/include/asm/io.h
+18
-9
arch/arc/include/asm/io.h
···
13
13
#include <asm/byteorder.h>
14
14
#include <asm/page.h>
15
15
16
+
#ifdef CONFIG_ISA_ARCV2
17
+
#include <asm/barrier.h>
18
+
#define __iormb() rmb()
19
+
#define __iowmb() wmb()
20
+
#else
21
+
#define __iormb() do { } while (0)
22
+
#define __iowmb() do { } while (0)
23
+
#endif
24
+
16
25
extern void __iomem *ioremap(phys_addr_t paddr, unsigned long size);
17
26
extern void __iomem *ioremap_prot(phys_addr_t paddr, unsigned long size,
18
27
unsigned long flags);
···
39
30
#define ioremap_nocache(phy, sz) ioremap(phy, sz)
40
31
#define ioremap_wc(phy, sz) ioremap(phy, sz)
41
32
#define ioremap_wt(phy, sz) ioremap(phy, sz)
33
+
34
+
/*
35
+
* io{read,write}{16,32}be() macros
36
+
*/
37
+
#define ioread16be(p) ({ u16 __v = be16_to_cpu((__force __be16)__raw_readw(p)); __iormb(); __v; })
38
+
#define ioread32be(p) ({ u32 __v = be32_to_cpu((__force __be32)__raw_readl(p)); __iormb(); __v; })
39
+
40
+
#define iowrite16be(v,p) ({ __iowmb(); __raw_writew((__force u16)cpu_to_be16(v), p); })
41
+
#define iowrite32be(v,p) ({ __iowmb(); __raw_writel((__force u32)cpu_to_be32(v), p); })
42
42
43
43
/* Change struct page to physical address */
44
44
#define page_to_phys(page) (page_to_pfn(page) << PAGE_SHIFT)
···
125
107
: "memory");
126
108
127
109
}
128
-
129
-
#ifdef CONFIG_ISA_ARCV2
130
-
#include <asm/barrier.h>
131
-
#define __iormb() rmb()
132
-
#define __iowmb() wmb()
133
-
#else
134
-
#define __iormb() do { } while (0)
135
-
#define __iowmb() do { } while (0)
136
-
#endif
137
110
138
111
/*
139
112
* MMIO can also get buffered/optimized in micro-arch, so barriers needed
+43
arch/arc/include/asm/mmzone.h
+43
arch/arc/include/asm/mmzone.h
···
1
+
/*
2
+
* Copyright (C) 2016 Synopsys, Inc. (www.synopsys.com)
3
+
*
4
+
* This program is free software; you can redistribute it and/or modify
5
+
* it under the terms of the GNU General Public License version 2 as
6
+
* published by the Free Software Foundation.
7
+
*/
8
+
9
+
#ifndef _ASM_ARC_MMZONE_H
10
+
#define _ASM_ARC_MMZONE_H
11
+
12
+
#ifdef CONFIG_DISCONTIGMEM
13
+
14
+
extern struct pglist_data node_data[];
15
+
#define NODE_DATA(nid) (&node_data[nid])
16
+
17
+
static inline int pfn_to_nid(unsigned long pfn)
18
+
{
19
+
int is_end_low = 1;
20
+
21
+
if (IS_ENABLED(CONFIG_ARC_HAS_PAE40))
22
+
is_end_low = pfn <= virt_to_pfn(0xFFFFFFFFUL);
23
+
24
+
/*
25
+
* node 0: lowmem: 0x8000_0000 to 0xFFFF_FFFF
26
+
* node 1: HIGHMEM w/o PAE40: 0x0 to 0x7FFF_FFFF
27
+
* HIGHMEM with PAE40: 0x1_0000_0000 to ...
28
+
*/
29
+
if (pfn >= ARCH_PFN_OFFSET && is_end_low)
30
+
return 0;
31
+
32
+
return 1;
33
+
}
34
+
35
+
static inline int pfn_valid(unsigned long pfn)
36
+
{
37
+
int nid = pfn_to_nid(pfn);
38
+
39
+
return (pfn <= node_end_pfn(nid));
40
+
}
41
+
#endif /* CONFIG_DISCONTIGMEM */
42
+
43
+
#endif
+11
-4
arch/arc/include/asm/page.h
+11
-4
arch/arc/include/asm/page.h
···
72
72
73
73
typedef pte_t * pgtable_t;
74
74
75
+
/*
76
+
* Use virt_to_pfn with caution:
77
+
* If used in pte or paddr related macros, it could cause truncation
78
+
* in PAE40 builds
79
+
* As a rule of thumb, only use it in helpers starting with virt_
80
+
* You have been warned !
81
+
*/
75
82
#define virt_to_pfn(kaddr) (__pa(kaddr) >> PAGE_SHIFT)
76
83
77
84
#define ARCH_PFN_OFFSET virt_to_pfn(CONFIG_LINUX_LINK_BASE)
78
85
86
+
#ifdef CONFIG_FLATMEM
79
87
#define pfn_valid(pfn) (((pfn) - ARCH_PFN_OFFSET) < max_mapnr)
88
+
#endif
80
89
81
90
/*
82
91
* __pa, __va, virt_to_page (ALERT: deprecated, don't use them)
···
94
85
* virt here means link-address/program-address as embedded in object code.
95
86
* And for ARC, link-addr = physical address
96
87
*/
97
-
#define __pa(vaddr) ((unsigned long)vaddr)
88
+
#define __pa(vaddr) ((unsigned long)(vaddr))
98
89
#define __va(paddr) ((void *)((unsigned long)(paddr)))
99
90
100
-
#define virt_to_page(kaddr) \
101
-
(mem_map + virt_to_pfn((kaddr) - CONFIG_LINUX_LINK_BASE))
102
-
91
+
#define virt_to_page(kaddr) pfn_to_page(virt_to_pfn(kaddr))
103
92
#define virt_addr_valid(kaddr) pfn_valid(virt_to_pfn(kaddr))
104
93
105
94
/* Default Permissions for stack/heaps pages (Non Executable) */
+6
-7
arch/arc/include/asm/pgtable.h
+6
-7
arch/arc/include/asm/pgtable.h
···
278
278
#define pmd_present(x) (pmd_val(x))
279
279
#define pmd_clear(xp) do { pmd_val(*(xp)) = 0; } while (0)
280
280
281
-
#define pte_page(pte) \
282
-
(mem_map + virt_to_pfn(pte_val(pte) - CONFIG_LINUX_LINK_BASE))
283
-
281
+
#define pte_page(pte) pfn_to_page(pte_pfn(pte))
284
282
#define mk_pte(page, prot) pfn_pte(page_to_pfn(page), prot)
285
-
#define pte_pfn(pte) virt_to_pfn(pte_val(pte))
286
-
#define pfn_pte(pfn, prot) (__pte(((pte_t)(pfn) << PAGE_SHIFT) | \
287
-
pgprot_val(prot)))
288
-
#define __pte_index(addr) (virt_to_pfn(addr) & (PTRS_PER_PTE - 1))
283
+
#define pfn_pte(pfn, prot) (__pte(((pte_t)(pfn) << PAGE_SHIFT) | pgprot_val(prot)))
284
+
285
+
/* Don't use virt_to_pfn for macros below: could cause truncations for PAE40*/
286
+
#define pte_pfn(pte) (pte_val(pte) >> PAGE_SHIFT)
287
+
#define __pte_index(addr) (((addr) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
289
288
290
289
/*
291
290
* pte_offset gets a @ptr to PMD entry (PGD in our 2-tier paging system)
+39
-15
arch/arc/mm/init.c
+39
-15
arch/arc/mm/init.c
···
30
30
static unsigned long low_mem_sz;
31
31
32
32
#ifdef CONFIG_HIGHMEM
33
-
static unsigned long min_high_pfn;
33
+
static unsigned long min_high_pfn, max_high_pfn;
34
34
static u64 high_mem_start;
35
35
static u64 high_mem_sz;
36
+
#endif
37
+
38
+
#ifdef CONFIG_DISCONTIGMEM
39
+
struct pglist_data node_data[MAX_NUMNODES] __read_mostly;
40
+
EXPORT_SYMBOL(node_data);
36
41
#endif
37
42
38
43
/* User can over-ride above with "mem=nnn[KkMm]" in cmdline */
···
114
109
/* Last usable page of low mem */
115
110
max_low_pfn = max_pfn = PFN_DOWN(low_mem_start + low_mem_sz);
116
111
117
-
#ifdef CONFIG_HIGHMEM
118
-
min_high_pfn = PFN_DOWN(high_mem_start);
119
-
max_pfn = PFN_DOWN(high_mem_start + high_mem_sz);
112
+
#ifdef CONFIG_FLATMEM
113
+
/* pfn_valid() uses this */
114
+
max_mapnr = max_low_pfn - min_low_pfn;
120
115
#endif
121
-
122
-
max_mapnr = max_pfn - min_low_pfn;
123
116
124
117
/*------------- bootmem allocator setup -----------------------*/
125
118
···
132
129
* the crash
133
130
*/
134
131
135
-
memblock_add(low_mem_start, low_mem_sz);
132
+
memblock_add_node(low_mem_start, low_mem_sz, 0);
136
133
memblock_reserve(low_mem_start, __pa(_end) - low_mem_start);
137
134
138
135
#ifdef CONFIG_BLK_DEV_INITRD
···
152
149
zones_size[ZONE_NORMAL] = max_low_pfn - min_low_pfn;
153
150
zones_holes[ZONE_NORMAL] = 0;
154
151
155
-
#ifdef CONFIG_HIGHMEM
156
-
zones_size[ZONE_HIGHMEM] = max_pfn - max_low_pfn;
157
-
158
-
/* This handles the peripheral address space hole */
159
-
zones_holes[ZONE_HIGHMEM] = min_high_pfn - max_low_pfn;
160
-
#endif
161
-
162
152
/*
163
153
* We can't use the helper free_area_init(zones[]) because it uses
164
154
* PAGE_OFFSET to compute the @min_low_pfn which would be wrong
···
164
168
zones_holes); /* holes */
165
169
166
170
#ifdef CONFIG_HIGHMEM
171
+
/*
172
+
* Populate a new node with highmem
173
+
*
174
+
* On ARC (w/o PAE) HIGHMEM addresses are actually smaller (0 based)
175
+
* than addresses in normal ala low memory (0x8000_0000 based).
176
+
* Even with PAE, the huge peripheral space hole would waste a lot of
177
+
* mem with single mem_map[]. This warrants a mem_map per region design.
178
+
* Thus HIGHMEM on ARC is imlemented with DISCONTIGMEM.
179
+
*
180
+
* DISCONTIGMEM in turns requires multiple nodes. node 0 above is
181
+
* populated with normal memory zone while node 1 only has highmem
182
+
*/
183
+
node_set_online(1);
184
+
185
+
min_high_pfn = PFN_DOWN(high_mem_start);
186
+
max_high_pfn = PFN_DOWN(high_mem_start + high_mem_sz);
187
+
188
+
zones_size[ZONE_NORMAL] = 0;
189
+
zones_holes[ZONE_NORMAL] = 0;
190
+
191
+
zones_size[ZONE_HIGHMEM] = max_high_pfn - min_high_pfn;
192
+
zones_holes[ZONE_HIGHMEM] = 0;
193
+
194
+
free_area_init_node(1, /* node-id */
195
+
zones_size, /* num pages per zone */
196
+
min_high_pfn, /* first pfn of node */
197
+
zones_holes); /* holes */
198
+
167
199
high_memory = (void *)(min_high_pfn << PAGE_SHIFT);
168
200
kmap_init();
169
201
#endif
···
209
185
unsigned long tmp;
210
186
211
187
reset_all_zones_managed_pages();
212
-
for (tmp = min_high_pfn; tmp < max_pfn; tmp++)
188
+
for (tmp = min_high_pfn; tmp < max_high_pfn; tmp++)
213
189
free_highmem_page(pfn_to_page(tmp));
214
190
#endif
215
191