WSL2-Linux-Kernel/arch/sparc/mm/init_64.c

3166 строки
76 KiB
C
Исходник Обычный вид История

License cleanup: add SPDX GPL-2.0 license identifier to files with no license Many source files in the tree are missing licensing information, which makes it harder for compliance tools to determine the correct license. By default all files without license information are under the default license of the kernel, which is GPL version 2. Update the files which contain no license information with the 'GPL-2.0' SPDX license identifier. The SPDX identifier is a legally binding shorthand, which can be used instead of the full boiler plate text. This patch is based on work done by Thomas Gleixner and Kate Stewart and Philippe Ombredanne. How this work was done: Patches were generated and checked against linux-4.14-rc6 for a subset of the use cases: - file had no licensing information it it. - file was a */uapi/* one with no licensing information in it, - file was a */uapi/* one with existing licensing information, Further patches will be generated in subsequent months to fix up cases where non-standard license headers were used, and references to license had to be inferred by heuristics based on keywords. The analysis to determine which SPDX License Identifier to be applied to a file was done in a spreadsheet of side by side results from of the output of two independent scanners (ScanCode & Windriver) producing SPDX tag:value files created by Philippe Ombredanne. Philippe prepared the base worksheet, and did an initial spot review of a few 1000 files. The 4.13 kernel was the starting point of the analysis with 60,537 files assessed. Kate Stewart did a file by file comparison of the scanner results in the spreadsheet to determine which SPDX license identifier(s) to be applied to the file. She confirmed any determination that was not immediately clear with lawyers working with the Linux Foundation. Criteria used to select files for SPDX license identifier tagging was: - Files considered eligible had to be source code files. - Make and config files were included as candidates if they contained >5 lines of source - File already had some variant of a license header in it (even if <5 lines). All documentation files were explicitly excluded. The following heuristics were used to determine which SPDX license identifiers to apply. - when both scanners couldn't find any license traces, file was considered to have no license information in it, and the top level COPYING file license applied. For non */uapi/* files that summary was: SPDX license identifier # files ---------------------------------------------------|------- GPL-2.0 11139 and resulted in the first patch in this series. If that file was a */uapi/* path one, it was "GPL-2.0 WITH Linux-syscall-note" otherwise it was "GPL-2.0". Results of that was: SPDX license identifier # files ---------------------------------------------------|------- GPL-2.0 WITH Linux-syscall-note 930 and resulted in the second patch in this series. - if a file had some form of licensing information in it, and was one of the */uapi/* ones, it was denoted with the Linux-syscall-note if any GPL family license was found in the file or had no licensing in it (per prior point). Results summary: SPDX license identifier # files ---------------------------------------------------|------ GPL-2.0 WITH Linux-syscall-note 270 GPL-2.0+ WITH Linux-syscall-note 169 ((GPL-2.0 WITH Linux-syscall-note) OR BSD-2-Clause) 21 ((GPL-2.0 WITH Linux-syscall-note) OR BSD-3-Clause) 17 LGPL-2.1+ WITH Linux-syscall-note 15 GPL-1.0+ WITH Linux-syscall-note 14 ((GPL-2.0+ WITH Linux-syscall-note) OR BSD-3-Clause) 5 LGPL-2.0+ WITH Linux-syscall-note 4 LGPL-2.1 WITH Linux-syscall-note 3 ((GPL-2.0 WITH Linux-syscall-note) OR MIT) 3 ((GPL-2.0 WITH Linux-syscall-note) AND MIT) 1 and that resulted in the third patch in this series. - when the two scanners agreed on the detected license(s), that became the concluded license(s). - when there was disagreement between the two scanners (one detected a license but the other didn't, or they both detected different licenses) a manual inspection of the file occurred. - In most cases a manual inspection of the information in the file resulted in a clear resolution of the license that should apply (and which scanner probably needed to revisit its heuristics). - When it was not immediately clear, the license identifier was confirmed with lawyers working with the Linux Foundation. - If there was any question as to the appropriate license identifier, the file was flagged for further research and to be revisited later in time. In total, over 70 hours of logged manual review was done on the spreadsheet to determine the SPDX license identifiers to apply to the source files by Kate, Philippe, Thomas and, in some cases, confirmation by lawyers working with the Linux Foundation. Kate also obtained a third independent scan of the 4.13 code base from FOSSology, and compared selected files where the other two scanners disagreed against that SPDX file, to see if there was new insights. The Windriver scanner is based on an older version of FOSSology in part, so they are related. Thomas did random spot checks in about 500 files from the spreadsheets for the uapi headers and agreed with SPDX license identifier in the files he inspected. For the non-uapi files Thomas did random spot checks in about 15000 files. In initial set of patches against 4.14-rc6, 3 files were found to have copy/paste license identifier errors, and have been fixed to reflect the correct identifier. Additionally Philippe spent 10 hours this week doing a detailed manual inspection and review of the 12,461 patched files from the initial patch version early this week with: - a full scancode scan run, collecting the matched texts, detected license ids and scores - reviewing anything where there was a license detected (about 500+ files) to ensure that the applied SPDX license was correct - reviewing anything where there was no detection but the patch license was not GPL-2.0 WITH Linux-syscall-note to ensure that the applied SPDX license was correct This produced a worksheet with 20 files needing minor correction. This worksheet was then exported into 3 different .csv files for the different types of files to be modified. These .csv files were then reviewed by Greg. Thomas wrote a script to parse the csv files and add the proper SPDX tag to the file, in the format that the file expected. This script was further refined by Greg based on the output to detect more types of files automatically and to distinguish between header and source .c files (which need different comment types.) Finally Greg ran the script using the .csv files to generate the patches. Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org> Reviewed-by: Philippe Ombredanne <pombredanne@nexb.com> Reviewed-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-11-01 17:07:57 +03:00
// SPDX-License-Identifier: GPL-2.0
/*
* arch/sparc64/mm/init.c
*
* Copyright (C) 1996-1999 David S. Miller (davem@caip.rutgers.edu)
* Copyright (C) 1997-1999 Jakub Jelinek (jj@sunsite.mff.cuni.cz)
*/
#include <linux/extable.h>
#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/string.h>
#include <linux/init.h>
#include <linux/bootmem.h>
#include <linux/mm.h>
#include <linux/hugetlb.h>
#include <linux/initrd.h>
#include <linux/swap.h>
#include <linux/pagemap.h>
#include <linux/poison.h>
#include <linux/fs.h>
#include <linux/seq_file.h>
#include <linux/kprobes.h>
#include <linux/cache.h>
#include <linux/sort.h>
#include <linux/ioport.h>
#include <linux/percpu.h>
#include <linux/memblock.h>
#include <linux/mmzone.h>
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h percpu.h is included by sched.h and module.h and thus ends up being included when building most .c files. percpu.h includes slab.h which in turn includes gfp.h making everything defined by the two files universally available and complicating inclusion dependencies. percpu.h -> slab.h dependency is about to be removed. Prepare for this change by updating users of gfp and slab facilities include those headers directly instead of assuming availability. As this conversion needs to touch large number of source files, the following script is used as the basis of conversion. http://userweb.kernel.org/~tj/misc/slabh-sweep.py The script does the followings. * Scan files for gfp and slab usages and update includes such that only the necessary includes are there. ie. if only gfp is used, gfp.h, if slab is used, slab.h. * When the script inserts a new include, it looks at the include blocks and try to put the new include such that its order conforms to its surrounding. It's put in the include block which contains core kernel includes, in the same order that the rest are ordered - alphabetical, Christmas tree, rev-Xmas-tree or at the end if there doesn't seem to be any matching order. * If the script can't find a place to put a new include (mostly because the file doesn't have fitting include block), it prints out an error message indicating which .h file needs to be added to the file. The conversion was done in the following steps. 1. The initial automatic conversion of all .c files updated slightly over 4000 files, deleting around 700 includes and adding ~480 gfp.h and ~3000 slab.h inclusions. The script emitted errors for ~400 files. 2. Each error was manually checked. Some didn't need the inclusion, some needed manual addition while adding it to implementation .h or embedding .c file was more appropriate for others. This step added inclusions to around 150 files. 3. The script was run again and the output was compared to the edits from #2 to make sure no file was left behind. 4. Several build tests were done and a couple of problems were fixed. e.g. lib/decompress_*.c used malloc/free() wrappers around slab APIs requiring slab.h to be added manually. 5. The script was run on all .h files but without automatically editing them as sprinkling gfp.h and slab.h inclusions around .h files could easily lead to inclusion dependency hell. Most gfp.h inclusion directives were ignored as stuff from gfp.h was usually wildly available and often used in preprocessor macros. Each slab.h inclusion directive was examined and added manually as necessary. 6. percpu.h was updated not to include slab.h. 7. Build test were done on the following configurations and failures were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my distributed build env didn't work with gcov compiles) and a few more options had to be turned off depending on archs to make things build (like ipr on powerpc/64 which failed due to missing writeq). * x86 and x86_64 UP and SMP allmodconfig and a custom test config. * powerpc and powerpc64 SMP allmodconfig * sparc and sparc64 SMP allmodconfig * ia64 SMP allmodconfig * s390 SMP allmodconfig * alpha SMP allmodconfig * um on x86_64 SMP allmodconfig 8. percpu.h modifications were reverted so that it could be applied as a separate patch and serve as bisection point. Given the fact that I had only a couple of failures from tests on step 6, I'm fairly confident about the coverage of this conversion patch. If there is a breakage, it's likely to be something in one of the arch headers which should be easily discoverable easily on most builds of the specific arch. Signed-off-by: Tejun Heo <tj@kernel.org> Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 11:04:11 +03:00
#include <linux/gfp.h>
#include <asm/head.h>
#include <asm/page.h>
#include <asm/pgalloc.h>
#include <asm/pgtable.h>
#include <asm/oplib.h>
#include <asm/iommu.h>
#include <asm/io.h>
#include <linux/uaccess.h>
#include <asm/mmu_context.h>
#include <asm/tlbflush.h>
#include <asm/dma.h>
#include <asm/starfire.h>
#include <asm/tlb.h>
#include <asm/spitfire.h>
#include <asm/sections.h>
#include <asm/tsb.h>
#include <asm/hypervisor.h>
#include <asm/prom.h>
#include <asm/mdesc.h>
#include <asm/cpudata.h>
#include <asm/setup.h>
#include <asm/irq.h>
#include "init_64.h"
unsigned long kern_linear_pte_xor[4] __read_mostly;
static unsigned long page_cache4v_flag;
/* A bitmap, two bits for every 256MB of physical memory. These two
* bits determine what page size we use for kernel linear
* translations. They form an index into kern_linear_pte_xor[]. The
* value in the indexed slot is XOR'd with the TLB miss virtual
* address to form the resulting TTE. The mapping is:
*
* 0 ==> 4MB
* 1 ==> 256MB
* 2 ==> 2GB
* 3 ==> 16GB
*
* All sun4v chips support 256MB pages. Only SPARC-T4 and later
* support 2GB pages, and hopefully future cpus will support the 16GB
* pages as well. For slots 2 and 3, we encode a 256MB TTE xor there
* if these larger page sizes are not supported by the cpu.
*
* It would be nice to determine this from the machine description
* 'cpu' properties, but we need to have this table setup before the
* MDESC is initialized.
*/
#ifndef CONFIG_DEBUG_PAGEALLOC
/* A special kernel TSB for 4MB, 256MB, 2GB and 16GB linear mappings.
* Space is allocated for this right after the trap table in
* arch/sparc64/kernel/head.S
*/
extern struct tsb swapper_4m_tsb[KERNEL_TSB4M_NENTRIES];
#endif
sparc64: Fix physical memory management regressions with large max_phys_bits. If max_phys_bits needs to be > 43 (f.e. for T4 chips), things like DEBUG_PAGEALLOC stop working because the 3-level page tables only can cover up to 43 bits. Another problem is that when we increased MAX_PHYS_ADDRESS_BITS up to 47, several statically allocated tables became enormous. Compounding this is that we will need to support up to 49 bits of physical addressing for M7 chips. The two tables in question are sparc64_valid_addr_bitmap and kpte_linear_bitmap. The first holds a bitmap, with 1 bit for each 4MB chunk of physical memory, indicating whether that chunk actually exists in the machine and is valid. The second table is a set of 2-bit values which tell how large of a mapping (4MB, 256MB, 2GB, 16GB, respectively) we can use at each 256MB chunk of ram in the system. These tables are huge and take up an enormous amount of the BSS section of the sparc64 kernel image. Specifically, the sparc64_valid_addr_bitmap is 4MB, and the kpte_linear_bitmap is 128K. So let's solve the space wastage and the DEBUG_PAGEALLOC problem at the same time, by using the kernel page tables (as designed) to manage this information. We have to keep using large mappings when DEBUG_PAGEALLOC is disabled, and we do this by encoding huge PMDs and PUDs. On a T4-2 with 256GB of ram the kernel page table takes up 16K with DEBUG_PAGEALLOC disabled and 256MB with it enabled. Furthermore, this memory is dynamically allocated at run time rather than coded statically into the kernel image. Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Bob Picco <bob.picco@oracle.com>
2014-09-25 07:56:11 +04:00
extern struct tsb swapper_tsb[KERNEL_TSB_NENTRIES];
static unsigned long cpu_pgsz_mask;
#define MAX_BANKS 1024
static struct linux_prom64_registers pavail[MAX_BANKS];
static int pavail_ents;
u64 numa_latency[MAX_NUMNODES][MAX_NUMNODES];
static int cmp_p64(const void *a, const void *b)
{
const struct linux_prom64_registers *x = a, *y = b;
if (x->phys_addr > y->phys_addr)
return 1;
if (x->phys_addr < y->phys_addr)
return -1;
return 0;
}
static void __init read_obp_memory(const char *property,
struct linux_prom64_registers *regs,
int *num_ents)
{
phandle node = prom_finddevice("/memory");
int prop_size = prom_getproplen(node, property);
int ents, ret, i;
ents = prop_size / sizeof(struct linux_prom64_registers);
if (ents > MAX_BANKS) {
prom_printf("The machine has more %s property entries than "
"this kernel can support (%d).\n",
property, MAX_BANKS);
prom_halt();
}
ret = prom_getproperty(node, property, (char *) regs, prop_size);
if (ret == -1) {
prom_printf("Couldn't get %s property from /memory.\n",
property);
prom_halt();
}
/* Sanitize what we got from the firmware, by page aligning
* everything.
*/
for (i = 0; i < ents; i++) {
unsigned long base, size;
base = regs[i].phys_addr;
size = regs[i].reg_size;
size &= PAGE_MASK;
if (base & ~PAGE_MASK) {
unsigned long new_base = PAGE_ALIGN(base);
size -= new_base - base;
if ((long) size < 0L)
size = 0UL;
base = new_base;
}
if (size == 0UL) {
/* If it is empty, simply get rid of it.
* This simplifies the logic of the other
* functions that process these arrays.
*/
memmove(&regs[i], &regs[i + 1],
(ents - i - 1) * sizeof(regs[0]));
i--;
ents--;
continue;
}
regs[i].phys_addr = base;
regs[i].reg_size = size;
}
*num_ents = ents;
[SPARC64]: Fix boot failures on SunBlade-150 The sequence to move over to the Linux trap tables from the firmware ones needs to be more air tight. It turns out that to be %100 safe we do need to be able to translate OBP mappings in our TLB miss handlers early. In order not to eat up a lot of kernel image memory with static page tables, just use the translations array in the OBP TLB miss handlers. That solves the bulk of the problem. Furthermore, to make sure the OBP TLB miss path will work even before the fixed MMU globals are loaded, explicitly load %g1 to TLB_SFSR at the beginning of the i-TLB and d-TLB miss handlers. To ease the OBP TLB miss walking of the prom_trans[] array, we sort it then delete all of the non-OBP entries in there (for example, there are entries for the kernel image itself which we're not interested in at all). We also save about 32K of kernel image size with this change. Not a bad side effect :-) There are still some reasons why trampoline.S can't use the setup_trap_table() yet. The most noteworthy are: 1) OBP boots secondary processors with non-bias'd stack for some reason. This is easily fixed by using a small bootup stack in the kernel image explicitly for this purpose. 2) Doing a firmware call via the normal C call prom_set_trap_table() goes through the whole OBP enter/exit sequence that saves and restores OBP and Linux kernel state in the MMUs. This path unfortunately does a "flush %g6" while loading up the OBP locked TLB entries for the firmware call. If we setup the %g6 in the trampoline.S code properly, that is in the PAGE_OFFSET linear mapping, but we're not on the kernel trap table yet so those addresses won't translate properly. One idea is to do a by-hand firmware call like we do in the early bootup code and elsewhere here in trampoline.S But this fails as well, as aparently the secondary processors are not booted with OBP's special locked TLB entries loaded. These are necessary for the firwmare to processes TLB misses correctly up until the point where we take over the trap table. This does need to be resolved at some point. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-10-12 23:22:46 +04:00
sort(regs, ents, sizeof(struct linux_prom64_registers),
cmp_p64, NULL);
}
/* Kernel physical address base and size in bytes. */
unsigned long kern_base __read_mostly;
unsigned long kern_size __read_mostly;
/* Initial ramdisk setup */
extern unsigned long sparc_ramdisk_image64;
extern unsigned int sparc_ramdisk_image;
extern unsigned int sparc_ramdisk_size;
struct page *mem_map_zero __read_mostly;
EXPORT_SYMBOL(mem_map_zero);
unsigned int sparc64_highest_unlocked_tlb_ent __read_mostly;
unsigned long sparc64_kern_pri_context __read_mostly;
unsigned long sparc64_kern_pri_nuc_bits __read_mostly;
unsigned long sparc64_kern_sec_context __read_mostly;
int num_kernel_image_mappings;
#ifdef CONFIG_DEBUG_DCFLUSH
atomic_t dcpage_flushes = ATOMIC_INIT(0);
#ifdef CONFIG_SMP
atomic_t dcpage_flushes_xcall = ATOMIC_INIT(0);
#endif
#endif
inline void flush_dcache_page_impl(struct page *page)
{
BUG_ON(tlb_type == hypervisor);
#ifdef CONFIG_DEBUG_DCFLUSH
atomic_inc(&dcpage_flushes);
#endif
#ifdef DCACHE_ALIASING_POSSIBLE
__flush_dcache_page(page_address(page),
((tlb_type == spitfire) &&
page_mapping(page) != NULL));
#else
if (page_mapping(page) != NULL &&
tlb_type == spitfire)
__flush_icache_page(__pa(page_address(page)));
#endif
}
#define PG_dcache_dirty PG_arch_1
#define PG_dcache_cpu_shift 32UL
#define PG_dcache_cpu_mask \
((1UL<<ilog2(roundup_pow_of_two(NR_CPUS)))-1UL)
#define dcache_dirty_cpu(page) \
(((page)->flags >> PG_dcache_cpu_shift) & PG_dcache_cpu_mask)
static inline void set_dcache_dirty(struct page *page, int this_cpu)
{
unsigned long mask = this_cpu;
unsigned long non_cpu_bits;
non_cpu_bits = ~(PG_dcache_cpu_mask << PG_dcache_cpu_shift);
mask = (mask << PG_dcache_cpu_shift) | (1UL << PG_dcache_dirty);
__asm__ __volatile__("1:\n\t"
"ldx [%2], %%g7\n\t"
"and %%g7, %1, %%g1\n\t"
"or %%g1, %0, %%g1\n\t"
"casx [%2], %%g7, %%g1\n\t"
"cmp %%g7, %%g1\n\t"
"bne,pn %%xcc, 1b\n\t"
" nop"
: /* no outputs */
: "r" (mask), "r" (non_cpu_bits), "r" (&page->flags)
: "g1", "g7");
}
static inline void clear_dcache_dirty_cpu(struct page *page, unsigned long cpu)
{
unsigned long mask = (1UL << PG_dcache_dirty);
__asm__ __volatile__("! test_and_clear_dcache_dirty\n"
"1:\n\t"
"ldx [%2], %%g7\n\t"
"srlx %%g7, %4, %%g1\n\t"
"and %%g1, %3, %%g1\n\t"
"cmp %%g1, %0\n\t"
"bne,pn %%icc, 2f\n\t"
" andn %%g7, %1, %%g1\n\t"
"casx [%2], %%g7, %%g1\n\t"
"cmp %%g7, %%g1\n\t"
"bne,pn %%xcc, 1b\n\t"
" nop\n"
"2:"
: /* no outputs */
: "r" (cpu), "r" (mask), "r" (&page->flags),
"i" (PG_dcache_cpu_mask),
"i" (PG_dcache_cpu_shift)
: "g1", "g7");
}
static inline void tsb_insert(struct tsb *ent, unsigned long tag, unsigned long pte)
{
unsigned long tsb_addr = (unsigned long) ent;
if (tlb_type == cheetah_plus || tlb_type == hypervisor)
tsb_addr = __pa(tsb_addr);
__tsb_insert(tsb_addr, tag, pte);
}
unsigned long _PAGE_ALL_SZ_BITS __read_mostly;
static void flush_dcache(unsigned long pfn)
{
struct page *page;
page = pfn_to_page(pfn);
if (page) {
unsigned long pg_flags;
pg_flags = page->flags;
if (pg_flags & (1UL << PG_dcache_dirty)) {
int cpu = ((pg_flags >> PG_dcache_cpu_shift) &
PG_dcache_cpu_mask);
int this_cpu = get_cpu();
/* This is just to optimize away some function calls
* in the SMP case.
*/
if (cpu == this_cpu)
flush_dcache_page_impl(page);
else
smp_flush_dcache_page_impl(page, cpu);
clear_dcache_dirty_cpu(page, cpu);
put_cpu();
}
}
}
/* mm->context.lock must be held */
static void __update_mmu_tsb_insert(struct mm_struct *mm, unsigned long tsb_index,
unsigned long tsb_hash_shift, unsigned long address,
unsigned long tte)
{
struct tsb *tsb = mm->context.tsb_block[tsb_index].tsb;
unsigned long tag;
if (unlikely(!tsb))
return;
tsb += ((address >> tsb_hash_shift) &
(mm->context.tsb_block[tsb_index].tsb_nentries - 1UL));
tag = (address >> 22UL);
tsb_insert(tsb, tag, tte);
}
#ifdef CONFIG_HUGETLB_PAGE
static void __init add_huge_page_size(unsigned long size)
{
unsigned int order;
if (size_to_hstate(size))
return;
order = ilog2(size) - PAGE_SHIFT;
hugetlb_add_hstate(order);
}
static int __init hugetlbpage_init(void)
{
add_huge_page_size(1UL << HPAGE_64K_SHIFT);
add_huge_page_size(1UL << HPAGE_SHIFT);
add_huge_page_size(1UL << HPAGE_256MB_SHIFT);
add_huge_page_size(1UL << HPAGE_2GB_SHIFT);
return 0;
}
arch_initcall(hugetlbpage_init);
static void __init pud_huge_patch(void)
{
struct pud_huge_patch_entry *p;
unsigned long addr;
p = &__pud_huge_patch;
addr = p->addr;
*(unsigned int *)addr = p->insn;
__asm__ __volatile__("flush %0" : : "r" (addr));
}
static int __init setup_hugepagesz(char *string)
{
unsigned long long hugepage_size;
unsigned int hugepage_shift;
unsigned short hv_pgsz_idx;
unsigned int hv_pgsz_mask;
int rc = 0;
hugepage_size = memparse(string, &string);
hugepage_shift = ilog2(hugepage_size);
switch (hugepage_shift) {
case HPAGE_16GB_SHIFT:
hv_pgsz_mask = HV_PGSZ_MASK_16GB;
hv_pgsz_idx = HV_PGSZ_IDX_16GB;
pud_huge_patch();
break;
case HPAGE_2GB_SHIFT:
hv_pgsz_mask = HV_PGSZ_MASK_2GB;
hv_pgsz_idx = HV_PGSZ_IDX_2GB;
break;
case HPAGE_256MB_SHIFT:
hv_pgsz_mask = HV_PGSZ_MASK_256MB;
hv_pgsz_idx = HV_PGSZ_IDX_256MB;
break;
case HPAGE_SHIFT:
hv_pgsz_mask = HV_PGSZ_MASK_4MB;
hv_pgsz_idx = HV_PGSZ_IDX_4MB;
break;
case HPAGE_64K_SHIFT:
hv_pgsz_mask = HV_PGSZ_MASK_64K;
hv_pgsz_idx = HV_PGSZ_IDX_64K;
break;
default:
hv_pgsz_mask = 0;
}
if ((hv_pgsz_mask & cpu_pgsz_mask) == 0U) {
hugetlb_bad_size();
pr_err("hugepagesz=%llu not supported by MMU.\n",
hugepage_size);
goto out;
}
add_huge_page_size(hugepage_size);
rc = 1;
out:
return rc;
}
__setup("hugepagesz=", setup_hugepagesz);
#endif /* CONFIG_HUGETLB_PAGE */
void update_mmu_cache(struct vm_area_struct *vma, unsigned long address, pte_t *ptep)
{
struct mm_struct *mm;
unsigned long flags;
bool is_huge_tsb;
pte_t pte = *ptep;
if (tlb_type != hypervisor) {
unsigned long pfn = pte_pfn(pte);
if (pfn_valid(pfn))
flush_dcache(pfn);
}
mm = vma->vm_mm;
[SPARC64]: Fix and re-enable dynamic TSB sizing. This is good for up to %50 performance improvement of some test cases. The problem has been the race conditions, and hopefully I've plugged them all up here. 1) There was a serious race in switch_mm() wrt. lazy TLB switching to and from kernel threads. We could erroneously skip a tsb_context_switch() and thus use a stale TSB across a TSB grow event. There is a big comment now in that function describing exactly how it can happen. 2) All code paths that do something with the TSB need to be guarded with the mm->context.lock spinlock. This makes page table flushing paths properly synchronize with both TSB growing and TLB context changes. 3) TSB growing events are moved to the end of successful fault processing. Previously it was in update_mmu_cache() but that is deadlock prone. At the end of do_sparc64_fault() we hold no spinlocks that could deadlock the TSB grow sequence. We also have dropped the address space semaphore. While we're here, add prefetching to the copy_tsb() routine and put it in assembler into the tsb.S file. This piece of code is quite time critical. There are some small negative side effects to this code which can be improved upon. In particular we grab the mm->context.lock even for the tsb insert done by update_mmu_cache() now and that's a bit excessive. We can get rid of that locking, and the same lock taking in flush_tsb_user(), by disabling PSTATE_IE around the whole operation including the capturing of the tsb pointer and tsb_nentries value. That would work because anyone growing the TSB won't free up the old TSB until all cpus respond to the TSB change cross call. I'm not quite so confident in that optimization to put it in right now, but eventually we might be able to and the description is here for reference. This code seems very solid now. It passes several parallel GCC bootstrap builds, and our favorite "nut cruncher" stress test which is a full "make -j8192" build of a "make allmodconfig" kernel. That puts about 256 processes on each cpu's run queue, makes lots of process cpu migrations occur, causes lots of page table and TLB flushing activity, incurs many context version number changes, and it swaps the machine real far out to disk even though there is 16GB of ram on this test system. :-) Signed-off-by: David S. Miller <davem@davemloft.net>
2006-03-16 13:02:32 +03:00
/* Don't insert a non-valid PTE into the TSB, we'll deadlock. */
if (!pte_accessible(mm, pte))
return;
[SPARC64]: Fix and re-enable dynamic TSB sizing. This is good for up to %50 performance improvement of some test cases. The problem has been the race conditions, and hopefully I've plugged them all up here. 1) There was a serious race in switch_mm() wrt. lazy TLB switching to and from kernel threads. We could erroneously skip a tsb_context_switch() and thus use a stale TSB across a TSB grow event. There is a big comment now in that function describing exactly how it can happen. 2) All code paths that do something with the TSB need to be guarded with the mm->context.lock spinlock. This makes page table flushing paths properly synchronize with both TSB growing and TLB context changes. 3) TSB growing events are moved to the end of successful fault processing. Previously it was in update_mmu_cache() but that is deadlock prone. At the end of do_sparc64_fault() we hold no spinlocks that could deadlock the TSB grow sequence. We also have dropped the address space semaphore. While we're here, add prefetching to the copy_tsb() routine and put it in assembler into the tsb.S file. This piece of code is quite time critical. There are some small negative side effects to this code which can be improved upon. In particular we grab the mm->context.lock even for the tsb insert done by update_mmu_cache() now and that's a bit excessive. We can get rid of that locking, and the same lock taking in flush_tsb_user(), by disabling PSTATE_IE around the whole operation including the capturing of the tsb pointer and tsb_nentries value. That would work because anyone growing the TSB won't free up the old TSB until all cpus respond to the TSB change cross call. I'm not quite so confident in that optimization to put it in right now, but eventually we might be able to and the description is here for reference. This code seems very solid now. It passes several parallel GCC bootstrap builds, and our favorite "nut cruncher" stress test which is a full "make -j8192" build of a "make allmodconfig" kernel. That puts about 256 processes on each cpu's run queue, makes lots of process cpu migrations occur, causes lots of page table and TLB flushing activity, incurs many context version number changes, and it swaps the machine real far out to disk even though there is 16GB of ram on this test system. :-) Signed-off-by: David S. Miller <davem@davemloft.net>
2006-03-16 13:02:32 +03:00
spin_lock_irqsave(&mm->context.lock, flags);
is_huge_tsb = false;
#if defined(CONFIG_HUGETLB_PAGE) || defined(CONFIG_TRANSPARENT_HUGEPAGE)
if (mm->context.hugetlb_pte_count || mm->context.thp_pte_count) {
unsigned long hugepage_size = PAGE_SIZE;
if (is_vm_hugetlb_page(vma))
hugepage_size = huge_page_size(hstate_vma(vma));
if (hugepage_size >= PUD_SIZE) {
unsigned long mask = 0x1ffc00000UL;
/* Transfer bits [32:22] from address to resolve
* at 4M granularity.
*/
pte_val(pte) &= ~mask;
pte_val(pte) |= (address & mask);
} else if (hugepage_size >= PMD_SIZE) {
/* We are fabricating 8MB pages using 4MB
* real hw pages.
*/
pte_val(pte) |= (address & (1UL << REAL_HPAGE_SHIFT));
}
if (hugepage_size >= PMD_SIZE) {
__update_mmu_tsb_insert(mm, MM_TSB_HUGE,
REAL_HPAGE_SHIFT, address, pte_val(pte));
is_huge_tsb = true;
}
}
#endif
if (!is_huge_tsb)
__update_mmu_tsb_insert(mm, MM_TSB_BASE, PAGE_SHIFT,
address, pte_val(pte));
[SPARC64]: Fix and re-enable dynamic TSB sizing. This is good for up to %50 performance improvement of some test cases. The problem has been the race conditions, and hopefully I've plugged them all up here. 1) There was a serious race in switch_mm() wrt. lazy TLB switching to and from kernel threads. We could erroneously skip a tsb_context_switch() and thus use a stale TSB across a TSB grow event. There is a big comment now in that function describing exactly how it can happen. 2) All code paths that do something with the TSB need to be guarded with the mm->context.lock spinlock. This makes page table flushing paths properly synchronize with both TSB growing and TLB context changes. 3) TSB growing events are moved to the end of successful fault processing. Previously it was in update_mmu_cache() but that is deadlock prone. At the end of do_sparc64_fault() we hold no spinlocks that could deadlock the TSB grow sequence. We also have dropped the address space semaphore. While we're here, add prefetching to the copy_tsb() routine and put it in assembler into the tsb.S file. This piece of code is quite time critical. There are some small negative side effects to this code which can be improved upon. In particular we grab the mm->context.lock even for the tsb insert done by update_mmu_cache() now and that's a bit excessive. We can get rid of that locking, and the same lock taking in flush_tsb_user(), by disabling PSTATE_IE around the whole operation including the capturing of the tsb pointer and tsb_nentries value. That would work because anyone growing the TSB won't free up the old TSB until all cpus respond to the TSB change cross call. I'm not quite so confident in that optimization to put it in right now, but eventually we might be able to and the description is here for reference. This code seems very solid now. It passes several parallel GCC bootstrap builds, and our favorite "nut cruncher" stress test which is a full "make -j8192" build of a "make allmodconfig" kernel. That puts about 256 processes on each cpu's run queue, makes lots of process cpu migrations occur, causes lots of page table and TLB flushing activity, incurs many context version number changes, and it swaps the machine real far out to disk even though there is 16GB of ram on this test system. :-) Signed-off-by: David S. Miller <davem@davemloft.net>
2006-03-16 13:02:32 +03:00
spin_unlock_irqrestore(&mm->context.lock, flags);
}
void flush_dcache_page(struct page *page)
{
struct address_space *mapping;
int this_cpu;
if (tlb_type == hypervisor)
return;
/* Do not bother with the expensive D-cache flush if it
* is merely the zero page. The 'bigcore' testcase in GDB
* causes this case to run millions of times.
*/
if (page == ZERO_PAGE(0))
return;
this_cpu = get_cpu();
mapping = page_mapping(page);
if (mapping && !mapping_mapped(mapping)) {
int dirty = test_bit(PG_dcache_dirty, &page->flags);
if (dirty) {
int dirty_cpu = dcache_dirty_cpu(page);
if (dirty_cpu == this_cpu)
goto out;
smp_flush_dcache_page_impl(page, dirty_cpu);
}
set_dcache_dirty(page, this_cpu);
} else {
/* We could delay the flush for the !page_mapping
* case too. But that case is for exec env/arg
* pages and those are %99 certainly going to get
* faulted into the tlb (and thus flushed) anyways.
*/
flush_dcache_page_impl(page);
}
out:
put_cpu();
}
EXPORT_SYMBOL(flush_dcache_page);
void __kprobes flush_icache_range(unsigned long start, unsigned long end)
{
/* Cheetah and Hypervisor platform cpus have coherent I-cache. */
if (tlb_type == spitfire) {
unsigned long kaddr;
/* This code only runs on Spitfire cpus so this is
* why we can assume _PAGE_PADDR_4U.
*/
for (kaddr = start; kaddr < end; kaddr += PAGE_SIZE) {
unsigned long paddr, mask = _PAGE_PADDR_4U;
if (kaddr >= PAGE_OFFSET)
paddr = kaddr & mask;
else {
pgd_t *pgdp = pgd_offset_k(kaddr);
pud_t *pudp = pud_offset(pgdp, kaddr);
pmd_t *pmdp = pmd_offset(pudp, kaddr);
pte_t *ptep = pte_offset_kernel(pmdp, kaddr);
paddr = pte_val(*ptep) & mask;
}
__flush_icache_page(paddr);
}
}
}
EXPORT_SYMBOL(flush_icache_range);
void mmu_info(struct seq_file *m)
{
static const char *pgsz_strings[] = {
"8K", "64K", "512K", "4MB", "32MB",
"256MB", "2GB", "16GB",
};
int i, printed;
if (tlb_type == cheetah)
seq_printf(m, "MMU Type\t: Cheetah\n");
else if (tlb_type == cheetah_plus)
seq_printf(m, "MMU Type\t: Cheetah+\n");
else if (tlb_type == spitfire)
seq_printf(m, "MMU Type\t: Spitfire\n");
else if (tlb_type == hypervisor)
seq_printf(m, "MMU Type\t: Hypervisor (sun4v)\n");
else
seq_printf(m, "MMU Type\t: ???\n");
seq_printf(m, "MMU PGSZs\t: ");
printed = 0;
for (i = 0; i < ARRAY_SIZE(pgsz_strings); i++) {
if (cpu_pgsz_mask & (1UL << i)) {
seq_printf(m, "%s%s",
printed ? "," : "", pgsz_strings[i]);
printed++;
}
}
seq_putc(m, '\n');
#ifdef CONFIG_DEBUG_DCFLUSH
seq_printf(m, "DCPageFlushes\t: %d\n",
atomic_read(&dcpage_flushes));
#ifdef CONFIG_SMP
seq_printf(m, "DCPageFlushesXC\t: %d\n",
atomic_read(&dcpage_flushes_xcall));
#endif /* CONFIG_SMP */
#endif /* CONFIG_DEBUG_DCFLUSH */
}
struct linux_prom_translation prom_trans[512] __read_mostly;
unsigned int prom_trans_ents __read_mostly;
unsigned long kern_locked_tte_data;
[SPARC64]: Fix boot failures on SunBlade-150 The sequence to move over to the Linux trap tables from the firmware ones needs to be more air tight. It turns out that to be %100 safe we do need to be able to translate OBP mappings in our TLB miss handlers early. In order not to eat up a lot of kernel image memory with static page tables, just use the translations array in the OBP TLB miss handlers. That solves the bulk of the problem. Furthermore, to make sure the OBP TLB miss path will work even before the fixed MMU globals are loaded, explicitly load %g1 to TLB_SFSR at the beginning of the i-TLB and d-TLB miss handlers. To ease the OBP TLB miss walking of the prom_trans[] array, we sort it then delete all of the non-OBP entries in there (for example, there are entries for the kernel image itself which we're not interested in at all). We also save about 32K of kernel image size with this change. Not a bad side effect :-) There are still some reasons why trampoline.S can't use the setup_trap_table() yet. The most noteworthy are: 1) OBP boots secondary processors with non-bias'd stack for some reason. This is easily fixed by using a small bootup stack in the kernel image explicitly for this purpose. 2) Doing a firmware call via the normal C call prom_set_trap_table() goes through the whole OBP enter/exit sequence that saves and restores OBP and Linux kernel state in the MMUs. This path unfortunately does a "flush %g6" while loading up the OBP locked TLB entries for the firmware call. If we setup the %g6 in the trampoline.S code properly, that is in the PAGE_OFFSET linear mapping, but we're not on the kernel trap table yet so those addresses won't translate properly. One idea is to do a by-hand firmware call like we do in the early bootup code and elsewhere here in trampoline.S But this fails as well, as aparently the secondary processors are not booted with OBP's special locked TLB entries loaded. These are necessary for the firwmare to processes TLB misses correctly up until the point where we take over the trap table. This does need to be resolved at some point. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-10-12 23:22:46 +04:00
/* The obp translations are saved based on 8k pagesize, since obp can
* use a mixture of pagesizes. Misses to the LOW_OBP_ADDRESS ->
* HI_OBP_ADDRESS range are handled in ktlb.S.
[SPARC64]: Fix boot failures on SunBlade-150 The sequence to move over to the Linux trap tables from the firmware ones needs to be more air tight. It turns out that to be %100 safe we do need to be able to translate OBP mappings in our TLB miss handlers early. In order not to eat up a lot of kernel image memory with static page tables, just use the translations array in the OBP TLB miss handlers. That solves the bulk of the problem. Furthermore, to make sure the OBP TLB miss path will work even before the fixed MMU globals are loaded, explicitly load %g1 to TLB_SFSR at the beginning of the i-TLB and d-TLB miss handlers. To ease the OBP TLB miss walking of the prom_trans[] array, we sort it then delete all of the non-OBP entries in there (for example, there are entries for the kernel image itself which we're not interested in at all). We also save about 32K of kernel image size with this change. Not a bad side effect :-) There are still some reasons why trampoline.S can't use the setup_trap_table() yet. The most noteworthy are: 1) OBP boots secondary processors with non-bias'd stack for some reason. This is easily fixed by using a small bootup stack in the kernel image explicitly for this purpose. 2) Doing a firmware call via the normal C call prom_set_trap_table() goes through the whole OBP enter/exit sequence that saves and restores OBP and Linux kernel state in the MMUs. This path unfortunately does a "flush %g6" while loading up the OBP locked TLB entries for the firmware call. If we setup the %g6 in the trampoline.S code properly, that is in the PAGE_OFFSET linear mapping, but we're not on the kernel trap table yet so those addresses won't translate properly. One idea is to do a by-hand firmware call like we do in the early bootup code and elsewhere here in trampoline.S But this fails as well, as aparently the secondary processors are not booted with OBP's special locked TLB entries loaded. These are necessary for the firwmare to processes TLB misses correctly up until the point where we take over the trap table. This does need to be resolved at some point. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-10-12 23:22:46 +04:00
*/
static inline int in_obp_range(unsigned long vaddr)
{
return (vaddr >= LOW_OBP_ADDRESS &&
vaddr < HI_OBP_ADDRESS);
}
[SPARC64]: Fix boot failures on SunBlade-150 The sequence to move over to the Linux trap tables from the firmware ones needs to be more air tight. It turns out that to be %100 safe we do need to be able to translate OBP mappings in our TLB miss handlers early. In order not to eat up a lot of kernel image memory with static page tables, just use the translations array in the OBP TLB miss handlers. That solves the bulk of the problem. Furthermore, to make sure the OBP TLB miss path will work even before the fixed MMU globals are loaded, explicitly load %g1 to TLB_SFSR at the beginning of the i-TLB and d-TLB miss handlers. To ease the OBP TLB miss walking of the prom_trans[] array, we sort it then delete all of the non-OBP entries in there (for example, there are entries for the kernel image itself which we're not interested in at all). We also save about 32K of kernel image size with this change. Not a bad side effect :-) There are still some reasons why trampoline.S can't use the setup_trap_table() yet. The most noteworthy are: 1) OBP boots secondary processors with non-bias'd stack for some reason. This is easily fixed by using a small bootup stack in the kernel image explicitly for this purpose. 2) Doing a firmware call via the normal C call prom_set_trap_table() goes through the whole OBP enter/exit sequence that saves and restores OBP and Linux kernel state in the MMUs. This path unfortunately does a "flush %g6" while loading up the OBP locked TLB entries for the firmware call. If we setup the %g6 in the trampoline.S code properly, that is in the PAGE_OFFSET linear mapping, but we're not on the kernel trap table yet so those addresses won't translate properly. One idea is to do a by-hand firmware call like we do in the early bootup code and elsewhere here in trampoline.S But this fails as well, as aparently the secondary processors are not booted with OBP's special locked TLB entries loaded. These are necessary for the firwmare to processes TLB misses correctly up until the point where we take over the trap table. This does need to be resolved at some point. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-10-12 23:22:46 +04:00
static int cmp_ptrans(const void *a, const void *b)
{
[SPARC64]: Fix boot failures on SunBlade-150 The sequence to move over to the Linux trap tables from the firmware ones needs to be more air tight. It turns out that to be %100 safe we do need to be able to translate OBP mappings in our TLB miss handlers early. In order not to eat up a lot of kernel image memory with static page tables, just use the translations array in the OBP TLB miss handlers. That solves the bulk of the problem. Furthermore, to make sure the OBP TLB miss path will work even before the fixed MMU globals are loaded, explicitly load %g1 to TLB_SFSR at the beginning of the i-TLB and d-TLB miss handlers. To ease the OBP TLB miss walking of the prom_trans[] array, we sort it then delete all of the non-OBP entries in there (for example, there are entries for the kernel image itself which we're not interested in at all). We also save about 32K of kernel image size with this change. Not a bad side effect :-) There are still some reasons why trampoline.S can't use the setup_trap_table() yet. The most noteworthy are: 1) OBP boots secondary processors with non-bias'd stack for some reason. This is easily fixed by using a small bootup stack in the kernel image explicitly for this purpose. 2) Doing a firmware call via the normal C call prom_set_trap_table() goes through the whole OBP enter/exit sequence that saves and restores OBP and Linux kernel state in the MMUs. This path unfortunately does a "flush %g6" while loading up the OBP locked TLB entries for the firmware call. If we setup the %g6 in the trampoline.S code properly, that is in the PAGE_OFFSET linear mapping, but we're not on the kernel trap table yet so those addresses won't translate properly. One idea is to do a by-hand firmware call like we do in the early bootup code and elsewhere here in trampoline.S But this fails as well, as aparently the secondary processors are not booted with OBP's special locked TLB entries loaded. These are necessary for the firwmare to processes TLB misses correctly up until the point where we take over the trap table. This does need to be resolved at some point. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-10-12 23:22:46 +04:00
const struct linux_prom_translation *x = a, *y = b;
[SPARC64]: Fix boot failures on SunBlade-150 The sequence to move over to the Linux trap tables from the firmware ones needs to be more air tight. It turns out that to be %100 safe we do need to be able to translate OBP mappings in our TLB miss handlers early. In order not to eat up a lot of kernel image memory with static page tables, just use the translations array in the OBP TLB miss handlers. That solves the bulk of the problem. Furthermore, to make sure the OBP TLB miss path will work even before the fixed MMU globals are loaded, explicitly load %g1 to TLB_SFSR at the beginning of the i-TLB and d-TLB miss handlers. To ease the OBP TLB miss walking of the prom_trans[] array, we sort it then delete all of the non-OBP entries in there (for example, there are entries for the kernel image itself which we're not interested in at all). We also save about 32K of kernel image size with this change. Not a bad side effect :-) There are still some reasons why trampoline.S can't use the setup_trap_table() yet. The most noteworthy are: 1) OBP boots secondary processors with non-bias'd stack for some reason. This is easily fixed by using a small bootup stack in the kernel image explicitly for this purpose. 2) Doing a firmware call via the normal C call prom_set_trap_table() goes through the whole OBP enter/exit sequence that saves and restores OBP and Linux kernel state in the MMUs. This path unfortunately does a "flush %g6" while loading up the OBP locked TLB entries for the firmware call. If we setup the %g6 in the trampoline.S code properly, that is in the PAGE_OFFSET linear mapping, but we're not on the kernel trap table yet so those addresses won't translate properly. One idea is to do a by-hand firmware call like we do in the early bootup code and elsewhere here in trampoline.S But this fails as well, as aparently the secondary processors are not booted with OBP's special locked TLB entries loaded. These are necessary for the firwmare to processes TLB misses correctly up until the point where we take over the trap table. This does need to be resolved at some point. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-10-12 23:22:46 +04:00
if (x->virt > y->virt)
return 1;
if (x->virt < y->virt)
return -1;
return 0;
}
[SPARC64]: Fix boot failures on SunBlade-150 The sequence to move over to the Linux trap tables from the firmware ones needs to be more air tight. It turns out that to be %100 safe we do need to be able to translate OBP mappings in our TLB miss handlers early. In order not to eat up a lot of kernel image memory with static page tables, just use the translations array in the OBP TLB miss handlers. That solves the bulk of the problem. Furthermore, to make sure the OBP TLB miss path will work even before the fixed MMU globals are loaded, explicitly load %g1 to TLB_SFSR at the beginning of the i-TLB and d-TLB miss handlers. To ease the OBP TLB miss walking of the prom_trans[] array, we sort it then delete all of the non-OBP entries in there (for example, there are entries for the kernel image itself which we're not interested in at all). We also save about 32K of kernel image size with this change. Not a bad side effect :-) There are still some reasons why trampoline.S can't use the setup_trap_table() yet. The most noteworthy are: 1) OBP boots secondary processors with non-bias'd stack for some reason. This is easily fixed by using a small bootup stack in the kernel image explicitly for this purpose. 2) Doing a firmware call via the normal C call prom_set_trap_table() goes through the whole OBP enter/exit sequence that saves and restores OBP and Linux kernel state in the MMUs. This path unfortunately does a "flush %g6" while loading up the OBP locked TLB entries for the firmware call. If we setup the %g6 in the trampoline.S code properly, that is in the PAGE_OFFSET linear mapping, but we're not on the kernel trap table yet so those addresses won't translate properly. One idea is to do a by-hand firmware call like we do in the early bootup code and elsewhere here in trampoline.S But this fails as well, as aparently the secondary processors are not booted with OBP's special locked TLB entries loaded. These are necessary for the firwmare to processes TLB misses correctly up until the point where we take over the trap table. This does need to be resolved at some point. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-10-12 23:22:46 +04:00
/* Read OBP translations property into 'prom_trans[]'. */
static void __init read_obp_translations(void)
{
[SPARC64]: Fix boot failures on SunBlade-150 The sequence to move over to the Linux trap tables from the firmware ones needs to be more air tight. It turns out that to be %100 safe we do need to be able to translate OBP mappings in our TLB miss handlers early. In order not to eat up a lot of kernel image memory with static page tables, just use the translations array in the OBP TLB miss handlers. That solves the bulk of the problem. Furthermore, to make sure the OBP TLB miss path will work even before the fixed MMU globals are loaded, explicitly load %g1 to TLB_SFSR at the beginning of the i-TLB and d-TLB miss handlers. To ease the OBP TLB miss walking of the prom_trans[] array, we sort it then delete all of the non-OBP entries in there (for example, there are entries for the kernel image itself which we're not interested in at all). We also save about 32K of kernel image size with this change. Not a bad side effect :-) There are still some reasons why trampoline.S can't use the setup_trap_table() yet. The most noteworthy are: 1) OBP boots secondary processors with non-bias'd stack for some reason. This is easily fixed by using a small bootup stack in the kernel image explicitly for this purpose. 2) Doing a firmware call via the normal C call prom_set_trap_table() goes through the whole OBP enter/exit sequence that saves and restores OBP and Linux kernel state in the MMUs. This path unfortunately does a "flush %g6" while loading up the OBP locked TLB entries for the firmware call. If we setup the %g6 in the trampoline.S code properly, that is in the PAGE_OFFSET linear mapping, but we're not on the kernel trap table yet so those addresses won't translate properly. One idea is to do a by-hand firmware call like we do in the early bootup code and elsewhere here in trampoline.S But this fails as well, as aparently the secondary processors are not booted with OBP's special locked TLB entries loaded. These are necessary for the firwmare to processes TLB misses correctly up until the point where we take over the trap table. This does need to be resolved at some point. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-10-12 23:22:46 +04:00
int n, node, ents, first, last, i;
node = prom_finddevice("/virtual-memory");
n = prom_getproplen(node, "translations");
if (unlikely(n == 0 || n == -1)) {
prom_printf("prom_mappings: Couldn't get size.\n");
prom_halt();
}
if (unlikely(n > sizeof(prom_trans))) {
prom_printf("prom_mappings: Size %d is too big.\n", n);
prom_halt();
}
if ((n = prom_getproperty(node, "translations",
(char *)&prom_trans[0],
sizeof(prom_trans))) == -1) {
prom_printf("prom_mappings: Couldn't get property.\n");
prom_halt();
}
n = n / sizeof(struct linux_prom_translation);
[SPARC64]: Fix boot failures on SunBlade-150 The sequence to move over to the Linux trap tables from the firmware ones needs to be more air tight. It turns out that to be %100 safe we do need to be able to translate OBP mappings in our TLB miss handlers early. In order not to eat up a lot of kernel image memory with static page tables, just use the translations array in the OBP TLB miss handlers. That solves the bulk of the problem. Furthermore, to make sure the OBP TLB miss path will work even before the fixed MMU globals are loaded, explicitly load %g1 to TLB_SFSR at the beginning of the i-TLB and d-TLB miss handlers. To ease the OBP TLB miss walking of the prom_trans[] array, we sort it then delete all of the non-OBP entries in there (for example, there are entries for the kernel image itself which we're not interested in at all). We also save about 32K of kernel image size with this change. Not a bad side effect :-) There are still some reasons why trampoline.S can't use the setup_trap_table() yet. The most noteworthy are: 1) OBP boots secondary processors with non-bias'd stack for some reason. This is easily fixed by using a small bootup stack in the kernel image explicitly for this purpose. 2) Doing a firmware call via the normal C call prom_set_trap_table() goes through the whole OBP enter/exit sequence that saves and restores OBP and Linux kernel state in the MMUs. This path unfortunately does a "flush %g6" while loading up the OBP locked TLB entries for the firmware call. If we setup the %g6 in the trampoline.S code properly, that is in the PAGE_OFFSET linear mapping, but we're not on the kernel trap table yet so those addresses won't translate properly. One idea is to do a by-hand firmware call like we do in the early bootup code and elsewhere here in trampoline.S But this fails as well, as aparently the secondary processors are not booted with OBP's special locked TLB entries loaded. These are necessary for the firwmare to processes TLB misses correctly up until the point where we take over the trap table. This does need to be resolved at some point. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-10-12 23:22:46 +04:00
ents = n;
sort(prom_trans, ents, sizeof(struct linux_prom_translation),
cmp_ptrans, NULL);
/* Now kick out all the non-OBP entries. */
for (i = 0; i < ents; i++) {
if (in_obp_range(prom_trans[i].virt))
break;
}
first = i;
for (; i < ents; i++) {
if (!in_obp_range(prom_trans[i].virt))
break;
}
last = i;
for (i = 0; i < (last - first); i++) {
struct linux_prom_translation *src = &prom_trans[i + first];
struct linux_prom_translation *dest = &prom_trans[i];
*dest = *src;
}
for (; i < ents; i++) {
struct linux_prom_translation *dest = &prom_trans[i];
dest->virt = dest->size = dest->data = 0x0UL;
}
prom_trans_ents = last - first;
if (tlb_type == spitfire) {
/* Clear diag TTE bits. */
for (i = 0; i < prom_trans_ents; i++)
prom_trans[i].data &= ~0x0003fe0000000000UL;
}
/* Force execute bit on. */
for (i = 0; i < prom_trans_ents; i++)
prom_trans[i].data |= (tlb_type == hypervisor ?
_PAGE_EXEC_4V : _PAGE_EXEC_4U);
}
static void __init hypervisor_tlb_lock(unsigned long vaddr,
unsigned long pte,
unsigned long mmu)
{
unsigned long ret = sun4v_mmu_map_perm_addr(vaddr, 0, pte, mmu);
if (ret != 0) {
prom_printf("hypervisor_tlb_lock[%lx:%x:%lx:%lx]: "
"errors with %lx\n", vaddr, 0, pte, mmu, ret);
prom_halt();
}
}
static unsigned long kern_large_tte(unsigned long paddr);
static void __init remap_kernel(void)
{
unsigned long phys_page, tte_vaddr, tte_data;
int i, tlb_ent = sparc64_highest_locked_tlbent();
tte_vaddr = (unsigned long) KERNBASE;
phys_page = (prom_boot_mapping_phys_low >> ILOG2_4MB) << ILOG2_4MB;
tte_data = kern_large_tte(phys_page);
kern_locked_tte_data = tte_data;
/* Now lock us into the TLBs via Hypervisor or OBP. */
if (tlb_type == hypervisor) {
for (i = 0; i < num_kernel_image_mappings; i++) {
hypervisor_tlb_lock(tte_vaddr, tte_data, HV_MMU_DMMU);
hypervisor_tlb_lock(tte_vaddr, tte_data, HV_MMU_IMMU);
tte_vaddr += 0x400000;
tte_data += 0x400000;
}
} else {
for (i = 0; i < num_kernel_image_mappings; i++) {
prom_dtlb_load(tlb_ent - i, tte_data, tte_vaddr);
prom_itlb_load(tlb_ent - i, tte_data, tte_vaddr);
tte_vaddr += 0x400000;
tte_data += 0x400000;
}
sparc64_highest_unlocked_tlb_ent = tlb_ent - i;
}
if (tlb_type == cheetah_plus) {
sparc64_kern_pri_context = (CTX_CHEETAH_PLUS_CTX0 |
CTX_CHEETAH_PLUS_NUC);
sparc64_kern_pri_nuc_bits = CTX_CHEETAH_PLUS_NUC;
sparc64_kern_sec_context = CTX_CHEETAH_PLUS_CTX0;
}
}
[SPARC64]: Fix boot failures on SunBlade-150 The sequence to move over to the Linux trap tables from the firmware ones needs to be more air tight. It turns out that to be %100 safe we do need to be able to translate OBP mappings in our TLB miss handlers early. In order not to eat up a lot of kernel image memory with static page tables, just use the translations array in the OBP TLB miss handlers. That solves the bulk of the problem. Furthermore, to make sure the OBP TLB miss path will work even before the fixed MMU globals are loaded, explicitly load %g1 to TLB_SFSR at the beginning of the i-TLB and d-TLB miss handlers. To ease the OBP TLB miss walking of the prom_trans[] array, we sort it then delete all of the non-OBP entries in there (for example, there are entries for the kernel image itself which we're not interested in at all). We also save about 32K of kernel image size with this change. Not a bad side effect :-) There are still some reasons why trampoline.S can't use the setup_trap_table() yet. The most noteworthy are: 1) OBP boots secondary processors with non-bias'd stack for some reason. This is easily fixed by using a small bootup stack in the kernel image explicitly for this purpose. 2) Doing a firmware call via the normal C call prom_set_trap_table() goes through the whole OBP enter/exit sequence that saves and restores OBP and Linux kernel state in the MMUs. This path unfortunately does a "flush %g6" while loading up the OBP locked TLB entries for the firmware call. If we setup the %g6 in the trampoline.S code properly, that is in the PAGE_OFFSET linear mapping, but we're not on the kernel trap table yet so those addresses won't translate properly. One idea is to do a by-hand firmware call like we do in the early bootup code and elsewhere here in trampoline.S But this fails as well, as aparently the secondary processors are not booted with OBP's special locked TLB entries loaded. These are necessary for the firwmare to processes TLB misses correctly up until the point where we take over the trap table. This does need to be resolved at some point. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-10-12 23:22:46 +04:00
static void __init inherit_prom_mappings(void)
{
/* Now fixup OBP's idea about where we really are mapped. */
printk("Remapping the kernel... ");
remap_kernel();
printk("done.\n");
}
void prom_world(int enter)
{
if (!enter)
set_fs(get_fs());
__asm__ __volatile__("flushw");
}
void __flush_dcache_range(unsigned long start, unsigned long end)
{
unsigned long va;
if (tlb_type == spitfire) {
int n = 0;
for (va = start; va < end; va += 32) {
spitfire_put_dcache_tag(va & 0x3fe0, 0x0);
if (++n >= 512)
break;
}
} else if (tlb_type == cheetah || tlb_type == cheetah_plus) {
start = __pa(start);
end = __pa(end);
for (va = start; va < end; va += 32)
__asm__ __volatile__("stxa %%g0, [%0] %1\n\t"
"membar #Sync"
: /* no outputs */
: "r" (va),
"i" (ASI_DCACHE_INVALIDATE));
}
}
EXPORT_SYMBOL(__flush_dcache_range);
/* get_new_mmu_context() uses "cache + 1". */
DEFINE_SPINLOCK(ctx_alloc_lock);
unsigned long tlb_context_cache = CTX_FIRST_VERSION;
#define MAX_CTX_NR (1UL << CTX_NR_BITS)
#define CTX_BMAP_SLOTS BITS_TO_LONGS(MAX_CTX_NR)
DECLARE_BITMAP(mmu_context_bmap, MAX_CTX_NR);
DEFINE_PER_CPU(struct mm_struct *, per_cpu_secondary_mm) = {0};
sparc64: new context wrap The current wrap implementation has a race issue: it is called outside of the ctx_alloc_lock, and also does not wait for all CPUs to complete the wrap. This means that a thread can get a new context with a new version and another thread might still be running with the same context. The problem is especially severe on CPUs with shared TLBs, like sun4v. I used the following test to very quickly reproduce the problem: - start over 8K processes (must be more than context IDs) - write and read values at a memory location in every process. Very quickly memory corruptions start happening, and what we read back does not equal what we wrote. Several approaches were explored before settling on this one: Approach 1: Move smp_new_mmu_context_version() inside ctx_alloc_lock, and wait for every process to complete the wrap. (Note: every CPU must WAIT before leaving smp_new_mmu_context_version_client() until every one arrives). This approach ends up with deadlocks, as some threads own locks which other threads are waiting for, and they never receive softint until these threads exit smp_new_mmu_context_version_client(). Since we do not allow the exit, deadlock happens. Approach 2: Handle wrap right during mondo interrupt. Use etrap/rtrap to enter into into C code, and issue new versions to every CPU. This approach adds some overhead to runtime: in switch_mm() we must add some checks to make sure that versions have not changed due to wrap while we were loading the new secondary context. (could be protected by PSTATE_IE but that degrades performance as on M7 and older CPUs as it takes 50 cycles for each access). Also, we still need a global per-cpu array of MMs to know where we need to load new contexts, otherwise we can change context to a thread that is going way (if we received mondo between switch_mm() and switch_to() time). Finally, there are some issues with window registers in rtrap() when context IDs are changed during CPU mondo time. The approach in this patch is the simplest and has almost no impact on runtime. We use the array with mm's where last secondary contexts were loaded onto CPUs and bump their versions to the new generation without changing context IDs. If a new process comes in to get a context ID, it will go through get_new_mmu_context() because of version mismatch. But the running processes do not need to be interrupted. And wrap is quicker as we do not need to xcall and wait for everyone to receive and complete wrap. Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Reviewed-by: Bob Picco <bob.picco@oracle.com> Reviewed-by: Steven Sistare <steven.sistare@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2017-05-31 18:25:24 +03:00
static void mmu_context_wrap(void)
{
unsigned long old_ver = tlb_context_cache & CTX_VERSION_MASK;
unsigned long new_ver, new_ctx, old_ctx;
struct mm_struct *mm;
int cpu;
bitmap_zero(mmu_context_bmap, 1 << CTX_NR_BITS);
/* Reserve kernel context */
set_bit(0, mmu_context_bmap);
new_ver = (tlb_context_cache & CTX_VERSION_MASK) + CTX_FIRST_VERSION;
if (unlikely(new_ver == 0))
new_ver = CTX_FIRST_VERSION;
tlb_context_cache = new_ver;
/*
* Make sure that any new mm that are added into per_cpu_secondary_mm,
* are going to go through get_new_mmu_context() path.
*/
mb();
/*
* Updated versions to current on those CPUs that had valid secondary
* contexts
*/
for_each_online_cpu(cpu) {
/*
* If a new mm is stored after we took this mm from the array,
* it will go into get_new_mmu_context() path, because we
* already bumped the version in tlb_context_cache.
*/
mm = per_cpu(per_cpu_secondary_mm, cpu);
if (unlikely(!mm || mm == &init_mm))
continue;
old_ctx = mm->context.sparc64_ctx_val;
if (likely((old_ctx & CTX_VERSION_MASK) == old_ver)) {
new_ctx = (old_ctx & ~CTX_VERSION_MASK) | new_ver;
set_bit(new_ctx & CTX_NR_MASK, mmu_context_bmap);
mm->context.sparc64_ctx_val = new_ctx;
}
}
}
/* Caller does TLB context flushing on local CPU if necessary.
* The caller also ensures that CTX_VALID(mm->context) is false.
*
* We must be careful about boundary cases so that we never
* let the user have CTX 0 (nucleus) or we ever use a CTX
* version of zero (and thus NO_CONTEXT would not be caught
* by version mis-match tests in mmu_context.h).
*
* Always invoked with interrupts disabled.
*/
void get_new_mmu_context(struct mm_struct *mm)
{
unsigned long ctx, new_ctx;
unsigned long orig_pgsz_bits;
spin_lock(&ctx_alloc_lock);
sparc64: new context wrap The current wrap implementation has a race issue: it is called outside of the ctx_alloc_lock, and also does not wait for all CPUs to complete the wrap. This means that a thread can get a new context with a new version and another thread might still be running with the same context. The problem is especially severe on CPUs with shared TLBs, like sun4v. I used the following test to very quickly reproduce the problem: - start over 8K processes (must be more than context IDs) - write and read values at a memory location in every process. Very quickly memory corruptions start happening, and what we read back does not equal what we wrote. Several approaches were explored before settling on this one: Approach 1: Move smp_new_mmu_context_version() inside ctx_alloc_lock, and wait for every process to complete the wrap. (Note: every CPU must WAIT before leaving smp_new_mmu_context_version_client() until every one arrives). This approach ends up with deadlocks, as some threads own locks which other threads are waiting for, and they never receive softint until these threads exit smp_new_mmu_context_version_client(). Since we do not allow the exit, deadlock happens. Approach 2: Handle wrap right during mondo interrupt. Use etrap/rtrap to enter into into C code, and issue new versions to every CPU. This approach adds some overhead to runtime: in switch_mm() we must add some checks to make sure that versions have not changed due to wrap while we were loading the new secondary context. (could be protected by PSTATE_IE but that degrades performance as on M7 and older CPUs as it takes 50 cycles for each access). Also, we still need a global per-cpu array of MMs to know where we need to load new contexts, otherwise we can change context to a thread that is going way (if we received mondo between switch_mm() and switch_to() time). Finally, there are some issues with window registers in rtrap() when context IDs are changed during CPU mondo time. The approach in this patch is the simplest and has almost no impact on runtime. We use the array with mm's where last secondary contexts were loaded onto CPUs and bump their versions to the new generation without changing context IDs. If a new process comes in to get a context ID, it will go through get_new_mmu_context() because of version mismatch. But the running processes do not need to be interrupted. And wrap is quicker as we do not need to xcall and wait for everyone to receive and complete wrap. Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Reviewed-by: Bob Picco <bob.picco@oracle.com> Reviewed-by: Steven Sistare <steven.sistare@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2017-05-31 18:25:24 +03:00
retry:
/* wrap might have happened, test again if our context became valid */
if (unlikely(CTX_VALID(mm->context)))
goto out;
orig_pgsz_bits = (mm->context.sparc64_ctx_val & CTX_PGSZ_MASK);
ctx = (tlb_context_cache + 1) & CTX_NR_MASK;
new_ctx = find_next_zero_bit(mmu_context_bmap, 1 << CTX_NR_BITS, ctx);
if (new_ctx >= (1 << CTX_NR_BITS)) {
new_ctx = find_next_zero_bit(mmu_context_bmap, ctx, 1);
if (new_ctx >= ctx) {
sparc64: new context wrap The current wrap implementation has a race issue: it is called outside of the ctx_alloc_lock, and also does not wait for all CPUs to complete the wrap. This means that a thread can get a new context with a new version and another thread might still be running with the same context. The problem is especially severe on CPUs with shared TLBs, like sun4v. I used the following test to very quickly reproduce the problem: - start over 8K processes (must be more than context IDs) - write and read values at a memory location in every process. Very quickly memory corruptions start happening, and what we read back does not equal what we wrote. Several approaches were explored before settling on this one: Approach 1: Move smp_new_mmu_context_version() inside ctx_alloc_lock, and wait for every process to complete the wrap. (Note: every CPU must WAIT before leaving smp_new_mmu_context_version_client() until every one arrives). This approach ends up with deadlocks, as some threads own locks which other threads are waiting for, and they never receive softint until these threads exit smp_new_mmu_context_version_client(). Since we do not allow the exit, deadlock happens. Approach 2: Handle wrap right during mondo interrupt. Use etrap/rtrap to enter into into C code, and issue new versions to every CPU. This approach adds some overhead to runtime: in switch_mm() we must add some checks to make sure that versions have not changed due to wrap while we were loading the new secondary context. (could be protected by PSTATE_IE but that degrades performance as on M7 and older CPUs as it takes 50 cycles for each access). Also, we still need a global per-cpu array of MMs to know where we need to load new contexts, otherwise we can change context to a thread that is going way (if we received mondo between switch_mm() and switch_to() time). Finally, there are some issues with window registers in rtrap() when context IDs are changed during CPU mondo time. The approach in this patch is the simplest and has almost no impact on runtime. We use the array with mm's where last secondary contexts were loaded onto CPUs and bump their versions to the new generation without changing context IDs. If a new process comes in to get a context ID, it will go through get_new_mmu_context() because of version mismatch. But the running processes do not need to be interrupted. And wrap is quicker as we do not need to xcall and wait for everyone to receive and complete wrap. Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Reviewed-by: Bob Picco <bob.picco@oracle.com> Reviewed-by: Steven Sistare <steven.sistare@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2017-05-31 18:25:24 +03:00
mmu_context_wrap();
goto retry;
}
}
if (mm->context.sparc64_ctx_val)
cpumask_clear(mm_cpumask(mm));
mmu_context_bmap[new_ctx>>6] |= (1UL << (new_ctx & 63));
new_ctx |= (tlb_context_cache & CTX_VERSION_MASK);
tlb_context_cache = new_ctx;
mm->context.sparc64_ctx_val = new_ctx | orig_pgsz_bits;
sparc64: new context wrap The current wrap implementation has a race issue: it is called outside of the ctx_alloc_lock, and also does not wait for all CPUs to complete the wrap. This means that a thread can get a new context with a new version and another thread might still be running with the same context. The problem is especially severe on CPUs with shared TLBs, like sun4v. I used the following test to very quickly reproduce the problem: - start over 8K processes (must be more than context IDs) - write and read values at a memory location in every process. Very quickly memory corruptions start happening, and what we read back does not equal what we wrote. Several approaches were explored before settling on this one: Approach 1: Move smp_new_mmu_context_version() inside ctx_alloc_lock, and wait for every process to complete the wrap. (Note: every CPU must WAIT before leaving smp_new_mmu_context_version_client() until every one arrives). This approach ends up with deadlocks, as some threads own locks which other threads are waiting for, and they never receive softint until these threads exit smp_new_mmu_context_version_client(). Since we do not allow the exit, deadlock happens. Approach 2: Handle wrap right during mondo interrupt. Use etrap/rtrap to enter into into C code, and issue new versions to every CPU. This approach adds some overhead to runtime: in switch_mm() we must add some checks to make sure that versions have not changed due to wrap while we were loading the new secondary context. (could be protected by PSTATE_IE but that degrades performance as on M7 and older CPUs as it takes 50 cycles for each access). Also, we still need a global per-cpu array of MMs to know where we need to load new contexts, otherwise we can change context to a thread that is going way (if we received mondo between switch_mm() and switch_to() time). Finally, there are some issues with window registers in rtrap() when context IDs are changed during CPU mondo time. The approach in this patch is the simplest and has almost no impact on runtime. We use the array with mm's where last secondary contexts were loaded onto CPUs and bump their versions to the new generation without changing context IDs. If a new process comes in to get a context ID, it will go through get_new_mmu_context() because of version mismatch. But the running processes do not need to be interrupted. And wrap is quicker as we do not need to xcall and wait for everyone to receive and complete wrap. Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com> Reviewed-by: Bob Picco <bob.picco@oracle.com> Reviewed-by: Steven Sistare <steven.sistare@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2017-05-31 18:25:24 +03:00
out:
spin_unlock(&ctx_alloc_lock);
}
static int numa_enabled = 1;
static int numa_debug;
static int __init early_numa(char *p)
{
if (!p)
return 0;
if (strstr(p, "off"))
numa_enabled = 0;
if (strstr(p, "debug"))
numa_debug = 1;
return 0;
}
early_param("numa", early_numa);
#define numadbg(f, a...) \
do { if (numa_debug) \
printk(KERN_INFO f, ## a); \
} while (0)
static void __init find_ramdisk(unsigned long phys_base)
{
#ifdef CONFIG_BLK_DEV_INITRD
if (sparc_ramdisk_image || sparc_ramdisk_image64) {
unsigned long ramdisk_image;
/* Older versions of the bootloader only supported a
* 32-bit physical address for the ramdisk image
* location, stored at sparc_ramdisk_image. Newer
* SILO versions set sparc_ramdisk_image to zero and
* provide a full 64-bit physical address at
* sparc_ramdisk_image64.
*/
ramdisk_image = sparc_ramdisk_image;
if (!ramdisk_image)
ramdisk_image = sparc_ramdisk_image64;
/* Another bootloader quirk. The bootloader normalizes
* the physical address to KERNBASE, so we have to
* factor that back out and add in the lowest valid
* physical page address to get the true physical address.
*/
ramdisk_image -= KERNBASE;
ramdisk_image += phys_base;
numadbg("Found ramdisk at physical address 0x%lx, size %u\n",
ramdisk_image, sparc_ramdisk_size);
initrd_start = ramdisk_image;
initrd_end = ramdisk_image + sparc_ramdisk_size;
memblock_reserve(initrd_start, sparc_ramdisk_size);
initrd_start += PAGE_OFFSET;
initrd_end += PAGE_OFFSET;
}
#endif
}
struct node_mem_mask {
unsigned long mask;
unsigned long match;
};
static struct node_mem_mask node_masks[MAX_NUMNODES];
static int num_node_masks;
#ifdef CONFIG_NEED_MULTIPLE_NODES
struct mdesc_mlgroup {
u64 node;
u64 latency;
u64 match;
u64 mask;
};
static struct mdesc_mlgroup *mlgroups;
static int num_mlgroups;
int numa_cpu_lookup_table[NR_CPUS];
cpumask_t numa_cpumask_lookup_table[MAX_NUMNODES];
struct mdesc_mblock {
u64 base;
u64 size;
u64 offset; /* RA-to-PA */
};
static struct mdesc_mblock *mblocks;
static int num_mblocks;
static struct mdesc_mblock * __init addr_to_mblock(unsigned long addr)
{
struct mdesc_mblock *m = NULL;
int i;
for (i = 0; i < num_mblocks; i++) {
m = &mblocks[i];
if (addr >= m->base &&
addr < (m->base + m->size)) {
break;
}
}
return m;
}
static u64 __init memblock_nid_range_sun4u(u64 start, u64 end, int *nid)
{
int prev_nid, new_nid;
prev_nid = -1;
for ( ; start < end; start += PAGE_SIZE) {
for (new_nid = 0; new_nid < num_node_masks; new_nid++) {
struct node_mem_mask *p = &node_masks[new_nid];
if ((start & p->mask) == p->match) {
if (prev_nid == -1)
prev_nid = new_nid;
break;
}
sparc64: Fix find_node warning if numa node cannot be found When booting up LDOM, find_node() warns that a physical address doesn't match a NUMA node. WARNING: CPU: 0 PID: 0 at arch/sparc/mm/init_64.c:835 find_node+0xf4/0x120 find_node: A physical address doesn't match a NUMA node rule. Some physical memory will be owned by node 0.Modules linked in: CPU: 0 PID: 0 Comm: swapper Not tainted 4.9.0-rc3 #4 Call Trace: [0000000000468ba0] __warn+0xc0/0xe0 [0000000000468c74] warn_slowpath_fmt+0x34/0x60 [00000000004592f4] find_node+0xf4/0x120 [0000000000dd0774] add_node_ranges+0x38/0xe4 [0000000000dd0b1c] numa_parse_mdesc+0x268/0x2e4 [0000000000dd0e9c] bootmem_init+0xb8/0x160 [0000000000dd174c] paging_init+0x808/0x8fc [0000000000dcb0d0] setup_arch+0x2c8/0x2f0 [0000000000dc68a0] start_kernel+0x48/0x424 [0000000000dcb374] start_early_boot+0x27c/0x28c [0000000000a32c08] tlb_fixup_done+0x4c/0x64 [0000000000027f08] 0x27f08 It is because linux use an internal structure node_masks[] to keep the best memory latency node only. However, LDOM mdesc can contain single latency-group with multiple memory latency nodes. If the address doesn't match the best latency node within node_masks[], it should check for an alternative via mdesc. The warning message should only be printed if the address doesn't match any node_masks[] nor within mdesc. To minimize the impact of searching mdesc every time, the last matched mask and index is stored in a variable. Signed-off-by: Thomas Tai <thomas.tai@oracle.com> Reviewed-by: Chris Hyser <chris.hyser@oracle.com> Reviewed-by: Liam Merwick <liam.merwick@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2016-11-03 19:19:01 +03:00
}
if (new_nid == num_node_masks) {
prev_nid = 0;
WARN_ONCE(1, "addr[%Lx] doesn't match a NUMA node rule. Some memory will be owned by node 0.",
start);
break;
}
if (prev_nid != new_nid)
break;
sparc64: Fix find_node warning if numa node cannot be found When booting up LDOM, find_node() warns that a physical address doesn't match a NUMA node. WARNING: CPU: 0 PID: 0 at arch/sparc/mm/init_64.c:835 find_node+0xf4/0x120 find_node: A physical address doesn't match a NUMA node rule. Some physical memory will be owned by node 0.Modules linked in: CPU: 0 PID: 0 Comm: swapper Not tainted 4.9.0-rc3 #4 Call Trace: [0000000000468ba0] __warn+0xc0/0xe0 [0000000000468c74] warn_slowpath_fmt+0x34/0x60 [00000000004592f4] find_node+0xf4/0x120 [0000000000dd0774] add_node_ranges+0x38/0xe4 [0000000000dd0b1c] numa_parse_mdesc+0x268/0x2e4 [0000000000dd0e9c] bootmem_init+0xb8/0x160 [0000000000dd174c] paging_init+0x808/0x8fc [0000000000dcb0d0] setup_arch+0x2c8/0x2f0 [0000000000dc68a0] start_kernel+0x48/0x424 [0000000000dcb374] start_early_boot+0x27c/0x28c [0000000000a32c08] tlb_fixup_done+0x4c/0x64 [0000000000027f08] 0x27f08 It is because linux use an internal structure node_masks[] to keep the best memory latency node only. However, LDOM mdesc can contain single latency-group with multiple memory latency nodes. If the address doesn't match the best latency node within node_masks[], it should check for an alternative via mdesc. The warning message should only be printed if the address doesn't match any node_masks[] nor within mdesc. To minimize the impact of searching mdesc every time, the last matched mask and index is stored in a variable. Signed-off-by: Thomas Tai <thomas.tai@oracle.com> Reviewed-by: Chris Hyser <chris.hyser@oracle.com> Reviewed-by: Liam Merwick <liam.merwick@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2016-11-03 19:19:01 +03:00
}
*nid = prev_nid;
sparc64: Fix find_node warning if numa node cannot be found When booting up LDOM, find_node() warns that a physical address doesn't match a NUMA node. WARNING: CPU: 0 PID: 0 at arch/sparc/mm/init_64.c:835 find_node+0xf4/0x120 find_node: A physical address doesn't match a NUMA node rule. Some physical memory will be owned by node 0.Modules linked in: CPU: 0 PID: 0 Comm: swapper Not tainted 4.9.0-rc3 #4 Call Trace: [0000000000468ba0] __warn+0xc0/0xe0 [0000000000468c74] warn_slowpath_fmt+0x34/0x60 [00000000004592f4] find_node+0xf4/0x120 [0000000000dd0774] add_node_ranges+0x38/0xe4 [0000000000dd0b1c] numa_parse_mdesc+0x268/0x2e4 [0000000000dd0e9c] bootmem_init+0xb8/0x160 [0000000000dd174c] paging_init+0x808/0x8fc [0000000000dcb0d0] setup_arch+0x2c8/0x2f0 [0000000000dc68a0] start_kernel+0x48/0x424 [0000000000dcb374] start_early_boot+0x27c/0x28c [0000000000a32c08] tlb_fixup_done+0x4c/0x64 [0000000000027f08] 0x27f08 It is because linux use an internal structure node_masks[] to keep the best memory latency node only. However, LDOM mdesc can contain single latency-group with multiple memory latency nodes. If the address doesn't match the best latency node within node_masks[], it should check for an alternative via mdesc. The warning message should only be printed if the address doesn't match any node_masks[] nor within mdesc. To minimize the impact of searching mdesc every time, the last matched mask and index is stored in a variable. Signed-off-by: Thomas Tai <thomas.tai@oracle.com> Reviewed-by: Chris Hyser <chris.hyser@oracle.com> Reviewed-by: Liam Merwick <liam.merwick@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2016-11-03 19:19:01 +03:00
return start > end ? end : start;
}
static u64 __init memblock_nid_range(u64 start, u64 end, int *nid)
{
u64 ret_end, pa_start, m_mask, m_match, m_end;
struct mdesc_mblock *mblock;
int _nid, i;
if (tlb_type != hypervisor)
return memblock_nid_range_sun4u(start, end, nid);
mblock = addr_to_mblock(start);
if (!mblock) {
WARN_ONCE(1, "memblock_nid_range: Can't find mblock addr[%Lx]",
start);
_nid = 0;
ret_end = end;
goto done;
}
pa_start = start + mblock->offset;
m_match = 0;
m_mask = 0;
for (_nid = 0; _nid < num_node_masks; _nid++) {
struct node_mem_mask *const m = &node_masks[_nid];
if ((pa_start & m->mask) == m->match) {
m_match = m->match;
m_mask = m->mask;
break;
}
}
if (num_node_masks == _nid) {
/* We could not find NUMA group, so default to 0, but lets
* search for latency group, so we could calculate the correct
* end address that we return
*/
_nid = 0;
for (i = 0; i < num_mlgroups; i++) {
struct mdesc_mlgroup *const m = &mlgroups[i];
if ((pa_start & m->mask) == m->match) {
m_match = m->match;
m_mask = m->mask;
break;
}
}
if (i == num_mlgroups) {
WARN_ONCE(1, "memblock_nid_range: Can't find latency group addr[%Lx]",
start);
ret_end = end;
goto done;
}
}
/*
* Each latency group has match and mask, and each memory block has an
* offset. An address belongs to a latency group if its address matches
* the following formula: ((addr + offset) & mask) == match
* It is, however, slow to check every single page if it matches a
* particular latency group. As optimization we calculate end value by
* using bit arithmetics.
*/
m_end = m_match + (1ul << __ffs(m_mask)) - mblock->offset;
m_end += pa_start & ~((1ul << fls64(m_mask)) - 1);
ret_end = m_end > end ? end : m_end;
done:
*nid = _nid;
return ret_end;
}
#endif
/* This must be invoked after performing all of the necessary
* memblock_set_node() calls for 'nid'. We need to be able to get
* correct data from get_pfn_range_for_nid().
*/
static void __init allocate_node_data(int nid)
{
struct pglist_data *p;
unsigned long start_pfn, end_pfn;
#ifdef CONFIG_NEED_MULTIPLE_NODES
unsigned long paddr;
paddr = memblock_alloc_try_nid(sizeof(struct pglist_data), SMP_CACHE_BYTES, nid);
if (!paddr) {
prom_printf("Cannot allocate pglist_data for nid[%d]\n", nid);
prom_halt();
}
NODE_DATA(nid) = __va(paddr);
memset(NODE_DATA(nid), 0, sizeof(struct pglist_data));
NODE_DATA(nid)->node_id = nid;
#endif
p = NODE_DATA(nid);
get_pfn_range_for_nid(nid, &start_pfn, &end_pfn);
p->node_start_pfn = start_pfn;
p->node_spanned_pages = end_pfn - start_pfn;
}
static void init_node_masks_nonnuma(void)
{
#ifdef CONFIG_NEED_MULTIPLE_NODES
int i;
#endif
numadbg("Initializing tables for non-numa.\n");
node_masks[0].mask = 0;
node_masks[0].match = 0;
num_node_masks = 1;
#ifdef CONFIG_NEED_MULTIPLE_NODES
for (i = 0; i < NR_CPUS; i++)
numa_cpu_lookup_table[i] = 0;
cpumask_setall(&numa_cpumask_lookup_table[0]);
#endif
}
#ifdef CONFIG_NEED_MULTIPLE_NODES
struct pglist_data *node_data[MAX_NUMNODES];
EXPORT_SYMBOL(numa_cpu_lookup_table);
EXPORT_SYMBOL(numa_cpumask_lookup_table);
EXPORT_SYMBOL(node_data);
static int scan_pio_for_cfg_handle(struct mdesc_handle *md, u64 pio,
u32 cfg_handle)
{
u64 arc;
mdesc_for_each_arc(arc, md, pio, MDESC_ARC_TYPE_FWD) {
u64 target = mdesc_arc_target(md, arc);
const u64 *val;
val = mdesc_get_property(md, target,
"cfg-handle", NULL);
if (val && *val == cfg_handle)
return 0;
}
return -ENODEV;
}
static int scan_arcs_for_cfg_handle(struct mdesc_handle *md, u64 grp,
u32 cfg_handle)
{
u64 arc, candidate, best_latency = ~(u64)0;
candidate = MDESC_NODE_NULL;
mdesc_for_each_arc(arc, md, grp, MDESC_ARC_TYPE_FWD) {
u64 target = mdesc_arc_target(md, arc);
const char *name = mdesc_node_name(md, target);
const u64 *val;
if (strcmp(name, "pio-latency-group"))
continue;
val = mdesc_get_property(md, target, "latency", NULL);
if (!val)
continue;
if (*val < best_latency) {
candidate = target;
best_latency = *val;
}
}
if (candidate == MDESC_NODE_NULL)
return -ENODEV;
return scan_pio_for_cfg_handle(md, candidate, cfg_handle);
}
int of_node_to_nid(struct device_node *dp)
{
const struct linux_prom64_registers *regs;
struct mdesc_handle *md;
u32 cfg_handle;
int count, nid;
u64 grp;
/* This is the right thing to do on currently supported
* SUN4U NUMA platforms as well, as the PCI controller does
* not sit behind any particular memory controller.
*/
if (!mlgroups)
return -1;
regs = of_get_property(dp, "reg", NULL);
if (!regs)
return -1;
cfg_handle = (regs->phys_addr >> 32UL) & 0x0fffffff;
md = mdesc_grab();
count = 0;
nid = -1;
mdesc_for_each_node_by_name(md, grp, "group") {
if (!scan_arcs_for_cfg_handle(md, grp, cfg_handle)) {
nid = count;
break;
}
count++;
}
mdesc_release(md);
return nid;
}
static void __init add_node_ranges(void)
{
struct memblock_region *reg;
unsigned long prev_max;
memblock_resized:
prev_max = memblock.memory.max;
for_each_memblock(memory, reg) {
unsigned long size = reg->size;
unsigned long start, end;
start = reg->base;
end = start + size;
while (start < end) {
unsigned long this_end;
int nid;
this_end = memblock_nid_range(start, end, &nid);
numadbg("Setting memblock NUMA node nid[%d] "
"start[%lx] end[%lx]\n",
nid, start, this_end);
memblock_set_node(start, this_end - start,
&memblock.memory, nid);
if (memblock.memory.max != prev_max)
goto memblock_resized;
start = this_end;
}
}
}
static int __init grab_mlgroups(struct mdesc_handle *md)
{
unsigned long paddr;
int count = 0;
u64 node;
mdesc_for_each_node_by_name(md, node, "memory-latency-group")
count++;
if (!count)
return -ENOENT;
paddr = memblock_alloc(count * sizeof(struct mdesc_mlgroup),
SMP_CACHE_BYTES);
if (!paddr)
return -ENOMEM;
mlgroups = __va(paddr);
num_mlgroups = count;
count = 0;
mdesc_for_each_node_by_name(md, node, "memory-latency-group") {
struct mdesc_mlgroup *m = &mlgroups[count++];
const u64 *val;
m->node = node;
val = mdesc_get_property(md, node, "latency", NULL);
m->latency = *val;
val = mdesc_get_property(md, node, "address-match", NULL);
m->match = *val;
val = mdesc_get_property(md, node, "address-mask", NULL);
m->mask = *val;
numadbg("MLGROUP[%d]: node[%llx] latency[%llx] "
"match[%llx] mask[%llx]\n",
count - 1, m->node, m->latency, m->match, m->mask);
}
return 0;
}
static int __init grab_mblocks(struct mdesc_handle *md)
{
unsigned long paddr;
int count = 0;
u64 node;
mdesc_for_each_node_by_name(md, node, "mblock")
count++;
if (!count)
return -ENOENT;
paddr = memblock_alloc(count * sizeof(struct mdesc_mblock),
SMP_CACHE_BYTES);
if (!paddr)
return -ENOMEM;
mblocks = __va(paddr);
num_mblocks = count;
count = 0;
mdesc_for_each_node_by_name(md, node, "mblock") {
struct mdesc_mblock *m = &mblocks[count++];
const u64 *val;
val = mdesc_get_property(md, node, "base", NULL);
m->base = *val;
val = mdesc_get_property(md, node, "size", NULL);
m->size = *val;
val = mdesc_get_property(md, node,
"address-congruence-offset", NULL);
/* The address-congruence-offset property is optional.
* Explicity zero it be identifty this.
*/
if (val)
m->offset = *val;
else
m->offset = 0UL;
numadbg("MBLOCK[%d]: base[%llx] size[%llx] offset[%llx]\n",
count - 1, m->base, m->size, m->offset);
}
return 0;
}
static void __init numa_parse_mdesc_group_cpus(struct mdesc_handle *md,
u64 grp, cpumask_t *mask)
{
u64 arc;
cpumask_clear(mask);
mdesc_for_each_arc(arc, md, grp, MDESC_ARC_TYPE_BACK) {
u64 target = mdesc_arc_target(md, arc);
const char *name = mdesc_node_name(md, target);
const u64 *id;
if (strcmp(name, "cpu"))
continue;
id = mdesc_get_property(md, target, "id", NULL);
if (*id < nr_cpu_ids)
cpumask_set_cpu(*id, mask);
}
}
static struct mdesc_mlgroup * __init find_mlgroup(u64 node)
{
int i;
for (i = 0; i < num_mlgroups; i++) {
struct mdesc_mlgroup *m = &mlgroups[i];
if (m->node == node)
return m;
}
return NULL;
}
int __node_distance(int from, int to)
{
if ((from >= MAX_NUMNODES) || (to >= MAX_NUMNODES)) {
pr_warn("Returning default NUMA distance value for %d->%d\n",
from, to);
return (from == to) ? LOCAL_DISTANCE : REMOTE_DISTANCE;
}
return numa_latency[from][to];
}
static int __init find_best_numa_node_for_mlgroup(struct mdesc_mlgroup *grp)
{
int i;
for (i = 0; i < MAX_NUMNODES; i++) {
struct node_mem_mask *n = &node_masks[i];
if ((grp->mask == n->mask) && (grp->match == n->match))
break;
}
return i;
}
static void __init find_numa_latencies_for_group(struct mdesc_handle *md,
u64 grp, int index)
{
u64 arc;
mdesc_for_each_arc(arc, md, grp, MDESC_ARC_TYPE_FWD) {
int tnode;
u64 target = mdesc_arc_target(md, arc);
struct mdesc_mlgroup *m = find_mlgroup(target);
if (!m)
continue;
tnode = find_best_numa_node_for_mlgroup(m);
if (tnode == MAX_NUMNODES)
continue;
numa_latency[index][tnode] = m->latency;
}
}
static int __init numa_attach_mlgroup(struct mdesc_handle *md, u64 grp,
int index)
{
struct mdesc_mlgroup *candidate = NULL;
u64 arc, best_latency = ~(u64)0;
struct node_mem_mask *n;
mdesc_for_each_arc(arc, md, grp, MDESC_ARC_TYPE_FWD) {
u64 target = mdesc_arc_target(md, arc);
struct mdesc_mlgroup *m = find_mlgroup(target);
if (!m)
continue;
if (m->latency < best_latency) {
candidate = m;
best_latency = m->latency;
}
}
if (!candidate)
return -ENOENT;
if (num_node_masks != index) {
printk(KERN_ERR "Inconsistent NUMA state, "
"index[%d] != num_node_masks[%d]\n",
index, num_node_masks);
return -EINVAL;
}
n = &node_masks[num_node_masks++];
n->mask = candidate->mask;
n->match = candidate->match;
numadbg("NUMA NODE[%d]: mask[%lx] match[%lx] (latency[%llx])\n",
index, n->mask, n->match, candidate->latency);
return 0;
}
static int __init numa_parse_mdesc_group(struct mdesc_handle *md, u64 grp,
int index)
{
cpumask_t mask;
int cpu;
numa_parse_mdesc_group_cpus(md, grp, &mask);
for_each_cpu(cpu, &mask)
numa_cpu_lookup_table[cpu] = index;
cpumask_copy(&numa_cpumask_lookup_table[index], &mask);
if (numa_debug) {
printk(KERN_INFO "NUMA GROUP[%d]: cpus [ ", index);
for_each_cpu(cpu, &mask)
printk("%d ", cpu);
printk("]\n");
}
return numa_attach_mlgroup(md, grp, index);
}
static int __init numa_parse_mdesc(void)
{
struct mdesc_handle *md = mdesc_grab();
int i, j, err, count;
u64 node;
node = mdesc_node_by_name(md, MDESC_NODE_NULL, "latency-groups");
if (node == MDESC_NODE_NULL) {
mdesc_release(md);
return -ENOENT;
}
err = grab_mblocks(md);
if (err < 0)
goto out;
err = grab_mlgroups(md);
if (err < 0)
goto out;
count = 0;
mdesc_for_each_node_by_name(md, node, "group") {
err = numa_parse_mdesc_group(md, node, count);
if (err < 0)
break;
count++;
}
count = 0;
mdesc_for_each_node_by_name(md, node, "group") {
find_numa_latencies_for_group(md, node, count);
count++;
}
/* Normalize numa latency matrix according to ACPI SLIT spec. */
for (i = 0; i < MAX_NUMNODES; i++) {
u64 self_latency = numa_latency[i][i];
for (j = 0; j < MAX_NUMNODES; j++) {
numa_latency[i][j] =
(numa_latency[i][j] * LOCAL_DISTANCE) /
self_latency;
}
}
add_node_ranges();
for (i = 0; i < num_node_masks; i++) {
allocate_node_data(i);
node_set_online(i);
}
err = 0;
out:
mdesc_release(md);
return err;
}
static int __init numa_parse_jbus(void)
{
unsigned long cpu, index;
/* NUMA node id is encoded in bits 36 and higher, and there is
* a 1-to-1 mapping from CPU ID to NUMA node ID.
*/
index = 0;
for_each_present_cpu(cpu) {
numa_cpu_lookup_table[cpu] = index;
cpumask_copy(&numa_cpumask_lookup_table[index], cpumask_of(cpu));
node_masks[index].mask = ~((1UL << 36UL) - 1UL);
node_masks[index].match = cpu << 36UL;
index++;
}
num_node_masks = index;
add_node_ranges();
for (index = 0; index < num_node_masks; index++) {
allocate_node_data(index);
node_set_online(index);
}
return 0;
}
static int __init numa_parse_sun4u(void)
{
if (tlb_type == cheetah || tlb_type == cheetah_plus) {
unsigned long ver;
__asm__ ("rdpr %%ver, %0" : "=r" (ver));
if ((ver >> 32UL) == __JALAPENO_ID ||
(ver >> 32UL) == __SERRANO_ID)
return numa_parse_jbus();
}
return -1;
}
static int __init bootmem_init_numa(void)
{
int i, j;
int err = -1;
numadbg("bootmem_init_numa()\n");
/* Some sane defaults for numa latency values */
for (i = 0; i < MAX_NUMNODES; i++) {
for (j = 0; j < MAX_NUMNODES; j++)
numa_latency[i][j] = (i == j) ?
LOCAL_DISTANCE : REMOTE_DISTANCE;
}
if (numa_enabled) {
if (tlb_type == hypervisor)
err = numa_parse_mdesc();
else
err = numa_parse_sun4u();
}
return err;
}
#else
static int bootmem_init_numa(void)
{
return -1;
}
#endif
static void __init bootmem_init_nonnuma(void)
{
unsigned long top_of_ram = memblock_end_of_DRAM();
unsigned long total_ram = memblock_phys_mem_size();
numadbg("bootmem_init_nonnuma()\n");
printk(KERN_INFO "Top of RAM: 0x%lx, Total RAM: 0x%lx\n",
top_of_ram, total_ram);
printk(KERN_INFO "Memory hole size: %ldMB\n",
(top_of_ram - total_ram) >> 20);
init_node_masks_nonnuma();
memblock_set_node(0, (phys_addr_t)ULLONG_MAX, &memblock.memory, 0);
allocate_node_data(0);
node_set_online(0);
}
static unsigned long __init bootmem_init(unsigned long phys_base)
{
unsigned long end_pfn;
end_pfn = memblock_end_of_DRAM() >> PAGE_SHIFT;
max_pfn = max_low_pfn = end_pfn;
min_low_pfn = (phys_base >> PAGE_SHIFT);
if (bootmem_init_numa() < 0)
bootmem_init_nonnuma();
/* Dump memblock with node info. */
memblock_dump_all();
/* XXX cpu notifier XXX */
sparse_memory_present_with_active_regions(MAX_NUMNODES);
sparse_init();
return end_pfn;
}
static struct linux_prom64_registers pall[MAX_BANKS] __initdata;
static int pall_ents __initdata;
sparc64: Fix physical memory management regressions with large max_phys_bits. If max_phys_bits needs to be > 43 (f.e. for T4 chips), things like DEBUG_PAGEALLOC stop working because the 3-level page tables only can cover up to 43 bits. Another problem is that when we increased MAX_PHYS_ADDRESS_BITS up to 47, several statically allocated tables became enormous. Compounding this is that we will need to support up to 49 bits of physical addressing for M7 chips. The two tables in question are sparc64_valid_addr_bitmap and kpte_linear_bitmap. The first holds a bitmap, with 1 bit for each 4MB chunk of physical memory, indicating whether that chunk actually exists in the machine and is valid. The second table is a set of 2-bit values which tell how large of a mapping (4MB, 256MB, 2GB, 16GB, respectively) we can use at each 256MB chunk of ram in the system. These tables are huge and take up an enormous amount of the BSS section of the sparc64 kernel image. Specifically, the sparc64_valid_addr_bitmap is 4MB, and the kpte_linear_bitmap is 128K. So let's solve the space wastage and the DEBUG_PAGEALLOC problem at the same time, by using the kernel page tables (as designed) to manage this information. We have to keep using large mappings when DEBUG_PAGEALLOC is disabled, and we do this by encoding huge PMDs and PUDs. On a T4-2 with 256GB of ram the kernel page table takes up 16K with DEBUG_PAGEALLOC disabled and 256MB with it enabled. Furthermore, this memory is dynamically allocated at run time rather than coded statically into the kernel image. Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Bob Picco <bob.picco@oracle.com>
2014-09-25 07:56:11 +04:00
static unsigned long max_phys_bits = 40;
bool kern_addr_valid(unsigned long addr)
{
pgd_t *pgd;
pud_t *pud;
pmd_t *pmd;
pte_t *pte;
if ((long)addr < 0L) {
sparc64: Fix physical memory management regressions with large max_phys_bits. If max_phys_bits needs to be > 43 (f.e. for T4 chips), things like DEBUG_PAGEALLOC stop working because the 3-level page tables only can cover up to 43 bits. Another problem is that when we increased MAX_PHYS_ADDRESS_BITS up to 47, several statically allocated tables became enormous. Compounding this is that we will need to support up to 49 bits of physical addressing for M7 chips. The two tables in question are sparc64_valid_addr_bitmap and kpte_linear_bitmap. The first holds a bitmap, with 1 bit for each 4MB chunk of physical memory, indicating whether that chunk actually exists in the machine and is valid. The second table is a set of 2-bit values which tell how large of a mapping (4MB, 256MB, 2GB, 16GB, respectively) we can use at each 256MB chunk of ram in the system. These tables are huge and take up an enormous amount of the BSS section of the sparc64 kernel image. Specifically, the sparc64_valid_addr_bitmap is 4MB, and the kpte_linear_bitmap is 128K. So let's solve the space wastage and the DEBUG_PAGEALLOC problem at the same time, by using the kernel page tables (as designed) to manage this information. We have to keep using large mappings when DEBUG_PAGEALLOC is disabled, and we do this by encoding huge PMDs and PUDs. On a T4-2 with 256GB of ram the kernel page table takes up 16K with DEBUG_PAGEALLOC disabled and 256MB with it enabled. Furthermore, this memory is dynamically allocated at run time rather than coded statically into the kernel image. Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Bob Picco <bob.picco@oracle.com>
2014-09-25 07:56:11 +04:00
unsigned long pa = __pa(addr);
if ((pa >> max_phys_bits) != 0UL)
return false;
sparc64: Fix physical memory management regressions with large max_phys_bits. If max_phys_bits needs to be > 43 (f.e. for T4 chips), things like DEBUG_PAGEALLOC stop working because the 3-level page tables only can cover up to 43 bits. Another problem is that when we increased MAX_PHYS_ADDRESS_BITS up to 47, several statically allocated tables became enormous. Compounding this is that we will need to support up to 49 bits of physical addressing for M7 chips. The two tables in question are sparc64_valid_addr_bitmap and kpte_linear_bitmap. The first holds a bitmap, with 1 bit for each 4MB chunk of physical memory, indicating whether that chunk actually exists in the machine and is valid. The second table is a set of 2-bit values which tell how large of a mapping (4MB, 256MB, 2GB, 16GB, respectively) we can use at each 256MB chunk of ram in the system. These tables are huge and take up an enormous amount of the BSS section of the sparc64 kernel image. Specifically, the sparc64_valid_addr_bitmap is 4MB, and the kpte_linear_bitmap is 128K. So let's solve the space wastage and the DEBUG_PAGEALLOC problem at the same time, by using the kernel page tables (as designed) to manage this information. We have to keep using large mappings when DEBUG_PAGEALLOC is disabled, and we do this by encoding huge PMDs and PUDs. On a T4-2 with 256GB of ram the kernel page table takes up 16K with DEBUG_PAGEALLOC disabled and 256MB with it enabled. Furthermore, this memory is dynamically allocated at run time rather than coded statically into the kernel image. Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Bob Picco <bob.picco@oracle.com>
2014-09-25 07:56:11 +04:00
return pfn_valid(pa >> PAGE_SHIFT);
}
if (addr >= (unsigned long) KERNBASE &&
addr < (unsigned long)&_end)
return true;
sparc64: Fix physical memory management regressions with large max_phys_bits. If max_phys_bits needs to be > 43 (f.e. for T4 chips), things like DEBUG_PAGEALLOC stop working because the 3-level page tables only can cover up to 43 bits. Another problem is that when we increased MAX_PHYS_ADDRESS_BITS up to 47, several statically allocated tables became enormous. Compounding this is that we will need to support up to 49 bits of physical addressing for M7 chips. The two tables in question are sparc64_valid_addr_bitmap and kpte_linear_bitmap. The first holds a bitmap, with 1 bit for each 4MB chunk of physical memory, indicating whether that chunk actually exists in the machine and is valid. The second table is a set of 2-bit values which tell how large of a mapping (4MB, 256MB, 2GB, 16GB, respectively) we can use at each 256MB chunk of ram in the system. These tables are huge and take up an enormous amount of the BSS section of the sparc64 kernel image. Specifically, the sparc64_valid_addr_bitmap is 4MB, and the kpte_linear_bitmap is 128K. So let's solve the space wastage and the DEBUG_PAGEALLOC problem at the same time, by using the kernel page tables (as designed) to manage this information. We have to keep using large mappings when DEBUG_PAGEALLOC is disabled, and we do this by encoding huge PMDs and PUDs. On a T4-2 with 256GB of ram the kernel page table takes up 16K with DEBUG_PAGEALLOC disabled and 256MB with it enabled. Furthermore, this memory is dynamically allocated at run time rather than coded statically into the kernel image. Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Bob Picco <bob.picco@oracle.com>
2014-09-25 07:56:11 +04:00
pgd = pgd_offset_k(addr);
if (pgd_none(*pgd))
return 0;
pud = pud_offset(pgd, addr);
if (pud_none(*pud))
return 0;
if (pud_large(*pud))
return pfn_valid(pud_pfn(*pud));
pmd = pmd_offset(pud, addr);
if (pmd_none(*pmd))
return 0;
if (pmd_large(*pmd))
return pfn_valid(pmd_pfn(*pmd));
pte = pte_offset_kernel(pmd, addr);
if (pte_none(*pte))
return 0;
return pfn_valid(pte_pfn(*pte));
}
EXPORT_SYMBOL(kern_addr_valid);
static unsigned long __ref kernel_map_hugepud(unsigned long vstart,
unsigned long vend,
pud_t *pud)
{
const unsigned long mask16gb = (1UL << 34) - 1UL;
u64 pte_val = vstart;
/* Each PUD is 8GB */
if ((vstart & mask16gb) ||
(vend - vstart <= mask16gb)) {
pte_val ^= kern_linear_pte_xor[2];
pud_val(*pud) = pte_val | _PAGE_PUD_HUGE;
return vstart + PUD_SIZE;
}
pte_val ^= kern_linear_pte_xor[3];
pte_val |= _PAGE_PUD_HUGE;
vend = vstart + mask16gb + 1UL;
while (vstart < vend) {
pud_val(*pud) = pte_val;
pte_val += PUD_SIZE;
vstart += PUD_SIZE;
pud++;
}
return vstart;
}
static bool kernel_can_map_hugepud(unsigned long vstart, unsigned long vend,
bool guard)
{
if (guard && !(vstart & ~PUD_MASK) && (vend - vstart) >= PUD_SIZE)
return true;
return false;
}
static unsigned long __ref kernel_map_hugepmd(unsigned long vstart,
unsigned long vend,
pmd_t *pmd)
{
const unsigned long mask256mb = (1UL << 28) - 1UL;
const unsigned long mask2gb = (1UL << 31) - 1UL;
u64 pte_val = vstart;
/* Each PMD is 8MB */
if ((vstart & mask256mb) ||
(vend - vstart <= mask256mb)) {
pte_val ^= kern_linear_pte_xor[0];
pmd_val(*pmd) = pte_val | _PAGE_PMD_HUGE;
return vstart + PMD_SIZE;
}
if ((vstart & mask2gb) ||
(vend - vstart <= mask2gb)) {
pte_val ^= kern_linear_pte_xor[1];
pte_val |= _PAGE_PMD_HUGE;
vend = vstart + mask256mb + 1UL;
} else {
pte_val ^= kern_linear_pte_xor[2];
pte_val |= _PAGE_PMD_HUGE;
vend = vstart + mask2gb + 1UL;
}
while (vstart < vend) {
pmd_val(*pmd) = pte_val;
pte_val += PMD_SIZE;
vstart += PMD_SIZE;
pmd++;
}
return vstart;
}
static bool kernel_can_map_hugepmd(unsigned long vstart, unsigned long vend,
bool guard)
{
if (guard && !(vstart & ~PMD_MASK) && (vend - vstart) >= PMD_SIZE)
return true;
return false;
}
static unsigned long __ref kernel_map_range(unsigned long pstart,
sparc64: Fix physical memory management regressions with large max_phys_bits. If max_phys_bits needs to be > 43 (f.e. for T4 chips), things like DEBUG_PAGEALLOC stop working because the 3-level page tables only can cover up to 43 bits. Another problem is that when we increased MAX_PHYS_ADDRESS_BITS up to 47, several statically allocated tables became enormous. Compounding this is that we will need to support up to 49 bits of physical addressing for M7 chips. The two tables in question are sparc64_valid_addr_bitmap and kpte_linear_bitmap. The first holds a bitmap, with 1 bit for each 4MB chunk of physical memory, indicating whether that chunk actually exists in the machine and is valid. The second table is a set of 2-bit values which tell how large of a mapping (4MB, 256MB, 2GB, 16GB, respectively) we can use at each 256MB chunk of ram in the system. These tables are huge and take up an enormous amount of the BSS section of the sparc64 kernel image. Specifically, the sparc64_valid_addr_bitmap is 4MB, and the kpte_linear_bitmap is 128K. So let's solve the space wastage and the DEBUG_PAGEALLOC problem at the same time, by using the kernel page tables (as designed) to manage this information. We have to keep using large mappings when DEBUG_PAGEALLOC is disabled, and we do this by encoding huge PMDs and PUDs. On a T4-2 with 256GB of ram the kernel page table takes up 16K with DEBUG_PAGEALLOC disabled and 256MB with it enabled. Furthermore, this memory is dynamically allocated at run time rather than coded statically into the kernel image. Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Bob Picco <bob.picco@oracle.com>
2014-09-25 07:56:11 +04:00
unsigned long pend, pgprot_t prot,
bool use_huge)
{
unsigned long vstart = PAGE_OFFSET + pstart;
unsigned long vend = PAGE_OFFSET + pend;
unsigned long alloc_bytes = 0UL;
if ((vstart & ~PAGE_MASK) || (vend & ~PAGE_MASK)) {
prom_printf("kernel_map: Unaligned physmem[%lx:%lx]\n",
vstart, vend);
prom_halt();
}
while (vstart < vend) {
unsigned long this_end, paddr = __pa(vstart);
pgd_t *pgd = pgd_offset_k(vstart);
pud_t *pud;
pmd_t *pmd;
pte_t *pte;
if (pgd_none(*pgd)) {
pud_t *new;
new = __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, PAGE_SIZE);
alloc_bytes += PAGE_SIZE;
pgd_populate(&init_mm, pgd, new);
}
pud = pud_offset(pgd, vstart);
if (pud_none(*pud)) {
pmd_t *new;
sparc64: Fix physical memory management regressions with large max_phys_bits. If max_phys_bits needs to be > 43 (f.e. for T4 chips), things like DEBUG_PAGEALLOC stop working because the 3-level page tables only can cover up to 43 bits. Another problem is that when we increased MAX_PHYS_ADDRESS_BITS up to 47, several statically allocated tables became enormous. Compounding this is that we will need to support up to 49 bits of physical addressing for M7 chips. The two tables in question are sparc64_valid_addr_bitmap and kpte_linear_bitmap. The first holds a bitmap, with 1 bit for each 4MB chunk of physical memory, indicating whether that chunk actually exists in the machine and is valid. The second table is a set of 2-bit values which tell how large of a mapping (4MB, 256MB, 2GB, 16GB, respectively) we can use at each 256MB chunk of ram in the system. These tables are huge and take up an enormous amount of the BSS section of the sparc64 kernel image. Specifically, the sparc64_valid_addr_bitmap is 4MB, and the kpte_linear_bitmap is 128K. So let's solve the space wastage and the DEBUG_PAGEALLOC problem at the same time, by using the kernel page tables (as designed) to manage this information. We have to keep using large mappings when DEBUG_PAGEALLOC is disabled, and we do this by encoding huge PMDs and PUDs. On a T4-2 with 256GB of ram the kernel page table takes up 16K with DEBUG_PAGEALLOC disabled and 256MB with it enabled. Furthermore, this memory is dynamically allocated at run time rather than coded statically into the kernel image. Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Bob Picco <bob.picco@oracle.com>
2014-09-25 07:56:11 +04:00
if (kernel_can_map_hugepud(vstart, vend, use_huge)) {
vstart = kernel_map_hugepud(vstart, vend, pud);
continue;
}
new = __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, PAGE_SIZE);
alloc_bytes += PAGE_SIZE;
pud_populate(&init_mm, pud, new);
}
pmd = pmd_offset(pud, vstart);
sparc64: Fix physical memory management regressions with large max_phys_bits. If max_phys_bits needs to be > 43 (f.e. for T4 chips), things like DEBUG_PAGEALLOC stop working because the 3-level page tables only can cover up to 43 bits. Another problem is that when we increased MAX_PHYS_ADDRESS_BITS up to 47, several statically allocated tables became enormous. Compounding this is that we will need to support up to 49 bits of physical addressing for M7 chips. The two tables in question are sparc64_valid_addr_bitmap and kpte_linear_bitmap. The first holds a bitmap, with 1 bit for each 4MB chunk of physical memory, indicating whether that chunk actually exists in the machine and is valid. The second table is a set of 2-bit values which tell how large of a mapping (4MB, 256MB, 2GB, 16GB, respectively) we can use at each 256MB chunk of ram in the system. These tables are huge and take up an enormous amount of the BSS section of the sparc64 kernel image. Specifically, the sparc64_valid_addr_bitmap is 4MB, and the kpte_linear_bitmap is 128K. So let's solve the space wastage and the DEBUG_PAGEALLOC problem at the same time, by using the kernel page tables (as designed) to manage this information. We have to keep using large mappings when DEBUG_PAGEALLOC is disabled, and we do this by encoding huge PMDs and PUDs. On a T4-2 with 256GB of ram the kernel page table takes up 16K with DEBUG_PAGEALLOC disabled and 256MB with it enabled. Furthermore, this memory is dynamically allocated at run time rather than coded statically into the kernel image. Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Bob Picco <bob.picco@oracle.com>
2014-09-25 07:56:11 +04:00
if (pmd_none(*pmd)) {
pte_t *new;
sparc64: Fix physical memory management regressions with large max_phys_bits. If max_phys_bits needs to be > 43 (f.e. for T4 chips), things like DEBUG_PAGEALLOC stop working because the 3-level page tables only can cover up to 43 bits. Another problem is that when we increased MAX_PHYS_ADDRESS_BITS up to 47, several statically allocated tables became enormous. Compounding this is that we will need to support up to 49 bits of physical addressing for M7 chips. The two tables in question are sparc64_valid_addr_bitmap and kpte_linear_bitmap. The first holds a bitmap, with 1 bit for each 4MB chunk of physical memory, indicating whether that chunk actually exists in the machine and is valid. The second table is a set of 2-bit values which tell how large of a mapping (4MB, 256MB, 2GB, 16GB, respectively) we can use at each 256MB chunk of ram in the system. These tables are huge and take up an enormous amount of the BSS section of the sparc64 kernel image. Specifically, the sparc64_valid_addr_bitmap is 4MB, and the kpte_linear_bitmap is 128K. So let's solve the space wastage and the DEBUG_PAGEALLOC problem at the same time, by using the kernel page tables (as designed) to manage this information. We have to keep using large mappings when DEBUG_PAGEALLOC is disabled, and we do this by encoding huge PMDs and PUDs. On a T4-2 with 256GB of ram the kernel page table takes up 16K with DEBUG_PAGEALLOC disabled and 256MB with it enabled. Furthermore, this memory is dynamically allocated at run time rather than coded statically into the kernel image. Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Bob Picco <bob.picco@oracle.com>
2014-09-25 07:56:11 +04:00
if (kernel_can_map_hugepmd(vstart, vend, use_huge)) {
vstart = kernel_map_hugepmd(vstart, vend, pmd);
continue;
}
new = __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, PAGE_SIZE);
alloc_bytes += PAGE_SIZE;
pmd_populate_kernel(&init_mm, pmd, new);
}
pte = pte_offset_kernel(pmd, vstart);
this_end = (vstart + PMD_SIZE) & PMD_MASK;
if (this_end > vend)
this_end = vend;
while (vstart < this_end) {
pte_val(*pte) = (paddr | pgprot_val(prot));
vstart += PAGE_SIZE;
paddr += PAGE_SIZE;
pte++;
}
}
return alloc_bytes;
}
sparc64: Fix physical memory management regressions with large max_phys_bits. If max_phys_bits needs to be > 43 (f.e. for T4 chips), things like DEBUG_PAGEALLOC stop working because the 3-level page tables only can cover up to 43 bits. Another problem is that when we increased MAX_PHYS_ADDRESS_BITS up to 47, several statically allocated tables became enormous. Compounding this is that we will need to support up to 49 bits of physical addressing for M7 chips. The two tables in question are sparc64_valid_addr_bitmap and kpte_linear_bitmap. The first holds a bitmap, with 1 bit for each 4MB chunk of physical memory, indicating whether that chunk actually exists in the machine and is valid. The second table is a set of 2-bit values which tell how large of a mapping (4MB, 256MB, 2GB, 16GB, respectively) we can use at each 256MB chunk of ram in the system. These tables are huge and take up an enormous amount of the BSS section of the sparc64 kernel image. Specifically, the sparc64_valid_addr_bitmap is 4MB, and the kpte_linear_bitmap is 128K. So let's solve the space wastage and the DEBUG_PAGEALLOC problem at the same time, by using the kernel page tables (as designed) to manage this information. We have to keep using large mappings when DEBUG_PAGEALLOC is disabled, and we do this by encoding huge PMDs and PUDs. On a T4-2 with 256GB of ram the kernel page table takes up 16K with DEBUG_PAGEALLOC disabled and 256MB with it enabled. Furthermore, this memory is dynamically allocated at run time rather than coded statically into the kernel image. Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Bob Picco <bob.picco@oracle.com>
2014-09-25 07:56:11 +04:00
static void __init flush_all_kernel_tsbs(void)
{
sparc64: Fix physical memory management regressions with large max_phys_bits. If max_phys_bits needs to be > 43 (f.e. for T4 chips), things like DEBUG_PAGEALLOC stop working because the 3-level page tables only can cover up to 43 bits. Another problem is that when we increased MAX_PHYS_ADDRESS_BITS up to 47, several statically allocated tables became enormous. Compounding this is that we will need to support up to 49 bits of physical addressing for M7 chips. The two tables in question are sparc64_valid_addr_bitmap and kpte_linear_bitmap. The first holds a bitmap, with 1 bit for each 4MB chunk of physical memory, indicating whether that chunk actually exists in the machine and is valid. The second table is a set of 2-bit values which tell how large of a mapping (4MB, 256MB, 2GB, 16GB, respectively) we can use at each 256MB chunk of ram in the system. These tables are huge and take up an enormous amount of the BSS section of the sparc64 kernel image. Specifically, the sparc64_valid_addr_bitmap is 4MB, and the kpte_linear_bitmap is 128K. So let's solve the space wastage and the DEBUG_PAGEALLOC problem at the same time, by using the kernel page tables (as designed) to manage this information. We have to keep using large mappings when DEBUG_PAGEALLOC is disabled, and we do this by encoding huge PMDs and PUDs. On a T4-2 with 256GB of ram the kernel page table takes up 16K with DEBUG_PAGEALLOC disabled and 256MB with it enabled. Furthermore, this memory is dynamically allocated at run time rather than coded statically into the kernel image. Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Bob Picco <bob.picco@oracle.com>
2014-09-25 07:56:11 +04:00
int i;
sparc64: Fix physical memory management regressions with large max_phys_bits. If max_phys_bits needs to be > 43 (f.e. for T4 chips), things like DEBUG_PAGEALLOC stop working because the 3-level page tables only can cover up to 43 bits. Another problem is that when we increased MAX_PHYS_ADDRESS_BITS up to 47, several statically allocated tables became enormous. Compounding this is that we will need to support up to 49 bits of physical addressing for M7 chips. The two tables in question are sparc64_valid_addr_bitmap and kpte_linear_bitmap. The first holds a bitmap, with 1 bit for each 4MB chunk of physical memory, indicating whether that chunk actually exists in the machine and is valid. The second table is a set of 2-bit values which tell how large of a mapping (4MB, 256MB, 2GB, 16GB, respectively) we can use at each 256MB chunk of ram in the system. These tables are huge and take up an enormous amount of the BSS section of the sparc64 kernel image. Specifically, the sparc64_valid_addr_bitmap is 4MB, and the kpte_linear_bitmap is 128K. So let's solve the space wastage and the DEBUG_PAGEALLOC problem at the same time, by using the kernel page tables (as designed) to manage this information. We have to keep using large mappings when DEBUG_PAGEALLOC is disabled, and we do this by encoding huge PMDs and PUDs. On a T4-2 with 256GB of ram the kernel page table takes up 16K with DEBUG_PAGEALLOC disabled and 256MB with it enabled. Furthermore, this memory is dynamically allocated at run time rather than coded statically into the kernel image. Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Bob Picco <bob.picco@oracle.com>
2014-09-25 07:56:11 +04:00
for (i = 0; i < KERNEL_TSB_NENTRIES; i++) {
struct tsb *ent = &swapper_tsb[i];
sparc64: Fix physical memory management regressions with large max_phys_bits. If max_phys_bits needs to be > 43 (f.e. for T4 chips), things like DEBUG_PAGEALLOC stop working because the 3-level page tables only can cover up to 43 bits. Another problem is that when we increased MAX_PHYS_ADDRESS_BITS up to 47, several statically allocated tables became enormous. Compounding this is that we will need to support up to 49 bits of physical addressing for M7 chips. The two tables in question are sparc64_valid_addr_bitmap and kpte_linear_bitmap. The first holds a bitmap, with 1 bit for each 4MB chunk of physical memory, indicating whether that chunk actually exists in the machine and is valid. The second table is a set of 2-bit values which tell how large of a mapping (4MB, 256MB, 2GB, 16GB, respectively) we can use at each 256MB chunk of ram in the system. These tables are huge and take up an enormous amount of the BSS section of the sparc64 kernel image. Specifically, the sparc64_valid_addr_bitmap is 4MB, and the kpte_linear_bitmap is 128K. So let's solve the space wastage and the DEBUG_PAGEALLOC problem at the same time, by using the kernel page tables (as designed) to manage this information. We have to keep using large mappings when DEBUG_PAGEALLOC is disabled, and we do this by encoding huge PMDs and PUDs. On a T4-2 with 256GB of ram the kernel page table takes up 16K with DEBUG_PAGEALLOC disabled and 256MB with it enabled. Furthermore, this memory is dynamically allocated at run time rather than coded statically into the kernel image. Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Bob Picco <bob.picco@oracle.com>
2014-09-25 07:56:11 +04:00
ent->tag = (1UL << TSB_TAG_INVALID_BIT);
}
sparc64: Fix physical memory management regressions with large max_phys_bits. If max_phys_bits needs to be > 43 (f.e. for T4 chips), things like DEBUG_PAGEALLOC stop working because the 3-level page tables only can cover up to 43 bits. Another problem is that when we increased MAX_PHYS_ADDRESS_BITS up to 47, several statically allocated tables became enormous. Compounding this is that we will need to support up to 49 bits of physical addressing for M7 chips. The two tables in question are sparc64_valid_addr_bitmap and kpte_linear_bitmap. The first holds a bitmap, with 1 bit for each 4MB chunk of physical memory, indicating whether that chunk actually exists in the machine and is valid. The second table is a set of 2-bit values which tell how large of a mapping (4MB, 256MB, 2GB, 16GB, respectively) we can use at each 256MB chunk of ram in the system. These tables are huge and take up an enormous amount of the BSS section of the sparc64 kernel image. Specifically, the sparc64_valid_addr_bitmap is 4MB, and the kpte_linear_bitmap is 128K. So let's solve the space wastage and the DEBUG_PAGEALLOC problem at the same time, by using the kernel page tables (as designed) to manage this information. We have to keep using large mappings when DEBUG_PAGEALLOC is disabled, and we do this by encoding huge PMDs and PUDs. On a T4-2 with 256GB of ram the kernel page table takes up 16K with DEBUG_PAGEALLOC disabled and 256MB with it enabled. Furthermore, this memory is dynamically allocated at run time rather than coded statically into the kernel image. Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Bob Picco <bob.picco@oracle.com>
2014-09-25 07:56:11 +04:00
#ifndef CONFIG_DEBUG_PAGEALLOC
for (i = 0; i < KERNEL_TSB4M_NENTRIES; i++) {
struct tsb *ent = &swapper_4m_tsb[i];
sparc64: Fix physical memory management regressions with large max_phys_bits. If max_phys_bits needs to be > 43 (f.e. for T4 chips), things like DEBUG_PAGEALLOC stop working because the 3-level page tables only can cover up to 43 bits. Another problem is that when we increased MAX_PHYS_ADDRESS_BITS up to 47, several statically allocated tables became enormous. Compounding this is that we will need to support up to 49 bits of physical addressing for M7 chips. The two tables in question are sparc64_valid_addr_bitmap and kpte_linear_bitmap. The first holds a bitmap, with 1 bit for each 4MB chunk of physical memory, indicating whether that chunk actually exists in the machine and is valid. The second table is a set of 2-bit values which tell how large of a mapping (4MB, 256MB, 2GB, 16GB, respectively) we can use at each 256MB chunk of ram in the system. These tables are huge and take up an enormous amount of the BSS section of the sparc64 kernel image. Specifically, the sparc64_valid_addr_bitmap is 4MB, and the kpte_linear_bitmap is 128K. So let's solve the space wastage and the DEBUG_PAGEALLOC problem at the same time, by using the kernel page tables (as designed) to manage this information. We have to keep using large mappings when DEBUG_PAGEALLOC is disabled, and we do this by encoding huge PMDs and PUDs. On a T4-2 with 256GB of ram the kernel page table takes up 16K with DEBUG_PAGEALLOC disabled and 256MB with it enabled. Furthermore, this memory is dynamically allocated at run time rather than coded statically into the kernel image. Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Bob Picco <bob.picco@oracle.com>
2014-09-25 07:56:11 +04:00
ent->tag = (1UL << TSB_TAG_INVALID_BIT);
}
sparc64: Fix physical memory management regressions with large max_phys_bits. If max_phys_bits needs to be > 43 (f.e. for T4 chips), things like DEBUG_PAGEALLOC stop working because the 3-level page tables only can cover up to 43 bits. Another problem is that when we increased MAX_PHYS_ADDRESS_BITS up to 47, several statically allocated tables became enormous. Compounding this is that we will need to support up to 49 bits of physical addressing for M7 chips. The two tables in question are sparc64_valid_addr_bitmap and kpte_linear_bitmap. The first holds a bitmap, with 1 bit for each 4MB chunk of physical memory, indicating whether that chunk actually exists in the machine and is valid. The second table is a set of 2-bit values which tell how large of a mapping (4MB, 256MB, 2GB, 16GB, respectively) we can use at each 256MB chunk of ram in the system. These tables are huge and take up an enormous amount of the BSS section of the sparc64 kernel image. Specifically, the sparc64_valid_addr_bitmap is 4MB, and the kpte_linear_bitmap is 128K. So let's solve the space wastage and the DEBUG_PAGEALLOC problem at the same time, by using the kernel page tables (as designed) to manage this information. We have to keep using large mappings when DEBUG_PAGEALLOC is disabled, and we do this by encoding huge PMDs and PUDs. On a T4-2 with 256GB of ram the kernel page table takes up 16K with DEBUG_PAGEALLOC disabled and 256MB with it enabled. Furthermore, this memory is dynamically allocated at run time rather than coded statically into the kernel image. Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Bob Picco <bob.picco@oracle.com>
2014-09-25 07:56:11 +04:00
#endif
}
sparc64: Fix physical memory management regressions with large max_phys_bits. If max_phys_bits needs to be > 43 (f.e. for T4 chips), things like DEBUG_PAGEALLOC stop working because the 3-level page tables only can cover up to 43 bits. Another problem is that when we increased MAX_PHYS_ADDRESS_BITS up to 47, several statically allocated tables became enormous. Compounding this is that we will need to support up to 49 bits of physical addressing for M7 chips. The two tables in question are sparc64_valid_addr_bitmap and kpte_linear_bitmap. The first holds a bitmap, with 1 bit for each 4MB chunk of physical memory, indicating whether that chunk actually exists in the machine and is valid. The second table is a set of 2-bit values which tell how large of a mapping (4MB, 256MB, 2GB, 16GB, respectively) we can use at each 256MB chunk of ram in the system. These tables are huge and take up an enormous amount of the BSS section of the sparc64 kernel image. Specifically, the sparc64_valid_addr_bitmap is 4MB, and the kpte_linear_bitmap is 128K. So let's solve the space wastage and the DEBUG_PAGEALLOC problem at the same time, by using the kernel page tables (as designed) to manage this information. We have to keep using large mappings when DEBUG_PAGEALLOC is disabled, and we do this by encoding huge PMDs and PUDs. On a T4-2 with 256GB of ram the kernel page table takes up 16K with DEBUG_PAGEALLOC disabled and 256MB with it enabled. Furthermore, this memory is dynamically allocated at run time rather than coded statically into the kernel image. Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Bob Picco <bob.picco@oracle.com>
2014-09-25 07:56:11 +04:00
extern unsigned int kvmap_linear_patch[1];
static void __init kernel_physical_mapping_init(void)
{
unsigned long i, mem_alloced = 0UL;
sparc64: Fix physical memory management regressions with large max_phys_bits. If max_phys_bits needs to be > 43 (f.e. for T4 chips), things like DEBUG_PAGEALLOC stop working because the 3-level page tables only can cover up to 43 bits. Another problem is that when we increased MAX_PHYS_ADDRESS_BITS up to 47, several statically allocated tables became enormous. Compounding this is that we will need to support up to 49 bits of physical addressing for M7 chips. The two tables in question are sparc64_valid_addr_bitmap and kpte_linear_bitmap. The first holds a bitmap, with 1 bit for each 4MB chunk of physical memory, indicating whether that chunk actually exists in the machine and is valid. The second table is a set of 2-bit values which tell how large of a mapping (4MB, 256MB, 2GB, 16GB, respectively) we can use at each 256MB chunk of ram in the system. These tables are huge and take up an enormous amount of the BSS section of the sparc64 kernel image. Specifically, the sparc64_valid_addr_bitmap is 4MB, and the kpte_linear_bitmap is 128K. So let's solve the space wastage and the DEBUG_PAGEALLOC problem at the same time, by using the kernel page tables (as designed) to manage this information. We have to keep using large mappings when DEBUG_PAGEALLOC is disabled, and we do this by encoding huge PMDs and PUDs. On a T4-2 with 256GB of ram the kernel page table takes up 16K with DEBUG_PAGEALLOC disabled and 256MB with it enabled. Furthermore, this memory is dynamically allocated at run time rather than coded statically into the kernel image. Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Bob Picco <bob.picco@oracle.com>
2014-09-25 07:56:11 +04:00
bool use_huge = true;
sparc64: Fix physical memory management regressions with large max_phys_bits. If max_phys_bits needs to be > 43 (f.e. for T4 chips), things like DEBUG_PAGEALLOC stop working because the 3-level page tables only can cover up to 43 bits. Another problem is that when we increased MAX_PHYS_ADDRESS_BITS up to 47, several statically allocated tables became enormous. Compounding this is that we will need to support up to 49 bits of physical addressing for M7 chips. The two tables in question are sparc64_valid_addr_bitmap and kpte_linear_bitmap. The first holds a bitmap, with 1 bit for each 4MB chunk of physical memory, indicating whether that chunk actually exists in the machine and is valid. The second table is a set of 2-bit values which tell how large of a mapping (4MB, 256MB, 2GB, 16GB, respectively) we can use at each 256MB chunk of ram in the system. These tables are huge and take up an enormous amount of the BSS section of the sparc64 kernel image. Specifically, the sparc64_valid_addr_bitmap is 4MB, and the kpte_linear_bitmap is 128K. So let's solve the space wastage and the DEBUG_PAGEALLOC problem at the same time, by using the kernel page tables (as designed) to manage this information. We have to keep using large mappings when DEBUG_PAGEALLOC is disabled, and we do this by encoding huge PMDs and PUDs. On a T4-2 with 256GB of ram the kernel page table takes up 16K with DEBUG_PAGEALLOC disabled and 256MB with it enabled. Furthermore, this memory is dynamically allocated at run time rather than coded statically into the kernel image. Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Bob Picco <bob.picco@oracle.com>
2014-09-25 07:56:11 +04:00
#ifdef CONFIG_DEBUG_PAGEALLOC
use_huge = false;
#endif
for (i = 0; i < pall_ents; i++) {
unsigned long phys_start, phys_end;
phys_start = pall[i].phys_addr;
phys_end = phys_start + pall[i].reg_size;
mem_alloced += kernel_map_range(phys_start, phys_end,
sparc64: Fix physical memory management regressions with large max_phys_bits. If max_phys_bits needs to be > 43 (f.e. for T4 chips), things like DEBUG_PAGEALLOC stop working because the 3-level page tables only can cover up to 43 bits. Another problem is that when we increased MAX_PHYS_ADDRESS_BITS up to 47, several statically allocated tables became enormous. Compounding this is that we will need to support up to 49 bits of physical addressing for M7 chips. The two tables in question are sparc64_valid_addr_bitmap and kpte_linear_bitmap. The first holds a bitmap, with 1 bit for each 4MB chunk of physical memory, indicating whether that chunk actually exists in the machine and is valid. The second table is a set of 2-bit values which tell how large of a mapping (4MB, 256MB, 2GB, 16GB, respectively) we can use at each 256MB chunk of ram in the system. These tables are huge and take up an enormous amount of the BSS section of the sparc64 kernel image. Specifically, the sparc64_valid_addr_bitmap is 4MB, and the kpte_linear_bitmap is 128K. So let's solve the space wastage and the DEBUG_PAGEALLOC problem at the same time, by using the kernel page tables (as designed) to manage this information. We have to keep using large mappings when DEBUG_PAGEALLOC is disabled, and we do this by encoding huge PMDs and PUDs. On a T4-2 with 256GB of ram the kernel page table takes up 16K with DEBUG_PAGEALLOC disabled and 256MB with it enabled. Furthermore, this memory is dynamically allocated at run time rather than coded statically into the kernel image. Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Bob Picco <bob.picco@oracle.com>
2014-09-25 07:56:11 +04:00
PAGE_KERNEL, use_huge);
}
printk("Allocated %ld bytes for kernel page tables.\n",
mem_alloced);
kvmap_linear_patch[0] = 0x01000000; /* nop */
flushi(&kvmap_linear_patch[0]);
sparc64: Fix physical memory management regressions with large max_phys_bits. If max_phys_bits needs to be > 43 (f.e. for T4 chips), things like DEBUG_PAGEALLOC stop working because the 3-level page tables only can cover up to 43 bits. Another problem is that when we increased MAX_PHYS_ADDRESS_BITS up to 47, several statically allocated tables became enormous. Compounding this is that we will need to support up to 49 bits of physical addressing for M7 chips. The two tables in question are sparc64_valid_addr_bitmap and kpte_linear_bitmap. The first holds a bitmap, with 1 bit for each 4MB chunk of physical memory, indicating whether that chunk actually exists in the machine and is valid. The second table is a set of 2-bit values which tell how large of a mapping (4MB, 256MB, 2GB, 16GB, respectively) we can use at each 256MB chunk of ram in the system. These tables are huge and take up an enormous amount of the BSS section of the sparc64 kernel image. Specifically, the sparc64_valid_addr_bitmap is 4MB, and the kpte_linear_bitmap is 128K. So let's solve the space wastage and the DEBUG_PAGEALLOC problem at the same time, by using the kernel page tables (as designed) to manage this information. We have to keep using large mappings when DEBUG_PAGEALLOC is disabled, and we do this by encoding huge PMDs and PUDs. On a T4-2 with 256GB of ram the kernel page table takes up 16K with DEBUG_PAGEALLOC disabled and 256MB with it enabled. Furthermore, this memory is dynamically allocated at run time rather than coded statically into the kernel image. Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Bob Picco <bob.picco@oracle.com>
2014-09-25 07:56:11 +04:00
flush_all_kernel_tsbs();
__flush_tlb_all();
}
#ifdef CONFIG_DEBUG_PAGEALLOC
void __kernel_map_pages(struct page *page, int numpages, int enable)
{
unsigned long phys_start = page_to_pfn(page) << PAGE_SHIFT;
unsigned long phys_end = phys_start + (numpages * PAGE_SIZE);
kernel_map_range(phys_start, phys_end,
sparc64: Fix physical memory management regressions with large max_phys_bits. If max_phys_bits needs to be > 43 (f.e. for T4 chips), things like DEBUG_PAGEALLOC stop working because the 3-level page tables only can cover up to 43 bits. Another problem is that when we increased MAX_PHYS_ADDRESS_BITS up to 47, several statically allocated tables became enormous. Compounding this is that we will need to support up to 49 bits of physical addressing for M7 chips. The two tables in question are sparc64_valid_addr_bitmap and kpte_linear_bitmap. The first holds a bitmap, with 1 bit for each 4MB chunk of physical memory, indicating whether that chunk actually exists in the machine and is valid. The second table is a set of 2-bit values which tell how large of a mapping (4MB, 256MB, 2GB, 16GB, respectively) we can use at each 256MB chunk of ram in the system. These tables are huge and take up an enormous amount of the BSS section of the sparc64 kernel image. Specifically, the sparc64_valid_addr_bitmap is 4MB, and the kpte_linear_bitmap is 128K. So let's solve the space wastage and the DEBUG_PAGEALLOC problem at the same time, by using the kernel page tables (as designed) to manage this information. We have to keep using large mappings when DEBUG_PAGEALLOC is disabled, and we do this by encoding huge PMDs and PUDs. On a T4-2 with 256GB of ram the kernel page table takes up 16K with DEBUG_PAGEALLOC disabled and 256MB with it enabled. Furthermore, this memory is dynamically allocated at run time rather than coded statically into the kernel image. Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Bob Picco <bob.picco@oracle.com>
2014-09-25 07:56:11 +04:00
(enable ? PAGE_KERNEL : __pgprot(0)), false);
flush_tsb_kernel_range(PAGE_OFFSET + phys_start,
PAGE_OFFSET + phys_end);
/* we should perform an IPI and flush all tlbs,
* but that can deadlock->flush only current cpu.
*/
__flush_tlb_kernel_range(PAGE_OFFSET + phys_start,
PAGE_OFFSET + phys_end);
}
#endif
unsigned long __init find_ecache_flush_span(unsigned long size)
{
int i;
for (i = 0; i < pavail_ents; i++) {
if (pavail[i].reg_size >= size)
return pavail[i].phys_addr;
}
return ~0UL;
}
unsigned long PAGE_OFFSET;
EXPORT_SYMBOL(PAGE_OFFSET);
unsigned long VMALLOC_END = 0x0000010000000000UL;
EXPORT_SYMBOL(VMALLOC_END);
unsigned long sparc64_va_hole_top = 0xfffff80000000000UL;
unsigned long sparc64_va_hole_bottom = 0x0000080000000000UL;
static void __init setup_page_offset(void)
{
if (tlb_type == cheetah || tlb_type == cheetah_plus) {
/* Cheetah/Panther support a full 64-bit virtual
* address, so we can use all that our page tables
* support.
*/
sparc64_va_hole_top = 0xfff0000000000000UL;
sparc64_va_hole_bottom = 0x0010000000000000UL;
max_phys_bits = 42;
} else if (tlb_type == hypervisor) {
switch (sun4v_chip_type) {
case SUN4V_CHIP_NIAGARA1:
case SUN4V_CHIP_NIAGARA2:
/* T1 and T2 support 48-bit virtual addresses. */
sparc64_va_hole_top = 0xffff800000000000UL;
sparc64_va_hole_bottom = 0x0000800000000000UL;
max_phys_bits = 39;
break;
case SUN4V_CHIP_NIAGARA3:
/* T3 supports 48-bit virtual addresses. */
sparc64_va_hole_top = 0xffff800000000000UL;
sparc64_va_hole_bottom = 0x0000800000000000UL;
max_phys_bits = 43;
break;
case SUN4V_CHIP_NIAGARA4:
case SUN4V_CHIP_NIAGARA5:
case SUN4V_CHIP_SPARC64X:
case SUN4V_CHIP_SPARC_M6:
/* T4 and later support 52-bit virtual addresses. */
sparc64_va_hole_top = 0xfff8000000000000UL;
sparc64_va_hole_bottom = 0x0008000000000000UL;
max_phys_bits = 47;
break;
case SUN4V_CHIP_SPARC_M7:
case SUN4V_CHIP_SPARC_SN:
/* M7 and later support 52-bit virtual addresses. */
sparc64_va_hole_top = 0xfff8000000000000UL;
sparc64_va_hole_bottom = 0x0008000000000000UL;
max_phys_bits = 49;
break;
case SUN4V_CHIP_SPARC_M8:
default:
/* M8 and later support 54-bit virtual addresses.
* However, restricting M8 and above VA bits to 53
* as 4-level page table cannot support more than
* 53 VA bits.
*/
sparc64_va_hole_top = 0xfff0000000000000UL;
sparc64_va_hole_bottom = 0x0010000000000000UL;
max_phys_bits = 51;
break;
}
}
if (max_phys_bits > MAX_PHYS_ADDRESS_BITS) {
prom_printf("MAX_PHYS_ADDRESS_BITS is too small, need %lu\n",
max_phys_bits);
prom_halt();
}
PAGE_OFFSET = sparc64_va_hole_top;
VMALLOC_END = ((sparc64_va_hole_bottom >> 1) +
(sparc64_va_hole_bottom >> 2));
pr_info("MM: PAGE_OFFSET is 0x%016lx (max_phys_bits == %lu)\n",
PAGE_OFFSET, max_phys_bits);
pr_info("MM: VMALLOC [0x%016lx --> 0x%016lx]\n",
VMALLOC_START, VMALLOC_END);
pr_info("MM: VMEMMAP [0x%016lx --> 0x%016lx]\n",
VMEMMAP_BASE, VMEMMAP_BASE << 1);
}
static void __init tsb_phys_patch(void)
{
struct tsb_ldquad_phys_patch_entry *pquad;
struct tsb_phys_patch_entry *p;
pquad = &__tsb_ldquad_phys_patch;
while (pquad < &__tsb_ldquad_phys_patch_end) {
unsigned long addr = pquad->addr;
if (tlb_type == hypervisor)
*(unsigned int *) addr = pquad->sun4v_insn;
else
*(unsigned int *) addr = pquad->sun4u_insn;
wmb();
__asm__ __volatile__("flush %0"
: /* no outputs */
: "r" (addr));
pquad++;
}
p = &__tsb_phys_patch;
while (p < &__tsb_phys_patch_end) {
unsigned long addr = p->addr;
*(unsigned int *) addr = p->insn;
wmb();
__asm__ __volatile__("flush %0"
: /* no outputs */
: "r" (addr));
p++;
}
}
/* Don't mark as init, we give this to the Hypervisor. */
#ifndef CONFIG_DEBUG_PAGEALLOC
#define NUM_KTSB_DESCR 2
#else
#define NUM_KTSB_DESCR 1
#endif
static struct hv_tsb_descr ktsb_descr[NUM_KTSB_DESCR];
/* The swapper TSBs are loaded with a base sequence of:
*
* sethi %uhi(SYMBOL), REG1
* sethi %hi(SYMBOL), REG2
* or REG1, %ulo(SYMBOL), REG1
* or REG2, %lo(SYMBOL), REG2
* sllx REG1, 32, REG1
* or REG1, REG2, REG1
*
* When we use physical addressing for the TSB accesses, we patch the
* first four instructions in the above sequence.
*/
static void patch_one_ktsb_phys(unsigned int *start, unsigned int *end, unsigned long pa)
{
unsigned long high_bits, low_bits;
high_bits = (pa >> 32) & 0xffffffff;
low_bits = (pa >> 0) & 0xffffffff;
while (start < end) {
unsigned int *ia = (unsigned int *)(unsigned long)*start;
ia[0] = (ia[0] & ~0x3fffff) | (high_bits >> 10);
__asm__ __volatile__("flush %0" : : "r" (ia));
ia[1] = (ia[1] & ~0x3fffff) | (low_bits >> 10);
__asm__ __volatile__("flush %0" : : "r" (ia + 1));
ia[2] = (ia[2] & ~0x1fff) | (high_bits & 0x3ff);
__asm__ __volatile__("flush %0" : : "r" (ia + 2));
ia[3] = (ia[3] & ~0x1fff) | (low_bits & 0x3ff);
__asm__ __volatile__("flush %0" : : "r" (ia + 3));
start++;
}
}
static void ktsb_phys_patch(void)
{
extern unsigned int __swapper_tsb_phys_patch;
extern unsigned int __swapper_tsb_phys_patch_end;
unsigned long ktsb_pa;
ktsb_pa = kern_base + ((unsigned long)&swapper_tsb[0] - KERNBASE);
patch_one_ktsb_phys(&__swapper_tsb_phys_patch,
&__swapper_tsb_phys_patch_end, ktsb_pa);
#ifndef CONFIG_DEBUG_PAGEALLOC
{
extern unsigned int __swapper_4m_tsb_phys_patch;
extern unsigned int __swapper_4m_tsb_phys_patch_end;
ktsb_pa = (kern_base +
((unsigned long)&swapper_4m_tsb[0] - KERNBASE));
patch_one_ktsb_phys(&__swapper_4m_tsb_phys_patch,
&__swapper_4m_tsb_phys_patch_end, ktsb_pa);
}
#endif
}
static void __init sun4v_ktsb_init(void)
{
unsigned long ktsb_pa;
/* First KTSB for PAGE_SIZE mappings. */
ktsb_pa = kern_base + ((unsigned long)&swapper_tsb[0] - KERNBASE);
switch (PAGE_SIZE) {
case 8 * 1024:
default:
ktsb_descr[0].pgsz_idx = HV_PGSZ_IDX_8K;
ktsb_descr[0].pgsz_mask = HV_PGSZ_MASK_8K;
break;
case 64 * 1024:
ktsb_descr[0].pgsz_idx = HV_PGSZ_IDX_64K;
ktsb_descr[0].pgsz_mask = HV_PGSZ_MASK_64K;
break;
case 512 * 1024:
ktsb_descr[0].pgsz_idx = HV_PGSZ_IDX_512K;
ktsb_descr[0].pgsz_mask = HV_PGSZ_MASK_512K;
break;
case 4 * 1024 * 1024:
ktsb_descr[0].pgsz_idx = HV_PGSZ_IDX_4MB;
ktsb_descr[0].pgsz_mask = HV_PGSZ_MASK_4MB;
break;
}
ktsb_descr[0].assoc = 1;
ktsb_descr[0].num_ttes = KERNEL_TSB_NENTRIES;
ktsb_descr[0].ctx_idx = 0;
ktsb_descr[0].tsb_base = ktsb_pa;
ktsb_descr[0].resv = 0;
#ifndef CONFIG_DEBUG_PAGEALLOC
/* Second KTSB for 4MB/256MB/2GB/16GB mappings. */
ktsb_pa = (kern_base +
((unsigned long)&swapper_4m_tsb[0] - KERNBASE));
ktsb_descr[1].pgsz_idx = HV_PGSZ_IDX_4MB;
ktsb_descr[1].pgsz_mask = ((HV_PGSZ_MASK_4MB |
HV_PGSZ_MASK_256MB |
HV_PGSZ_MASK_2GB |
HV_PGSZ_MASK_16GB) &
cpu_pgsz_mask);
ktsb_descr[1].assoc = 1;
ktsb_descr[1].num_ttes = KERNEL_TSB4M_NENTRIES;
ktsb_descr[1].ctx_idx = 0;
ktsb_descr[1].tsb_base = ktsb_pa;
ktsb_descr[1].resv = 0;
#endif
}
sparc: delete __cpuinit/__CPUINIT usage from all users The __cpuinit type of throwaway sections might have made sense some time ago when RAM was more constrained, but now the savings do not offset the cost and complications. For example, the fix in commit 5e427ec2d0 ("x86: Fix bit corruption at CPU resume time") is a good example of the nasty type of bugs that can be created with improper use of the various __init prefixes. After a discussion on LKML[1] it was decided that cpuinit should go the way of devinit and be phased out. Once all the users are gone, we can then finally remove the macros themselves from linux/init.h. Note that some harmless section mismatch warnings may result, since notify_cpu_starting() and cpu_up() are arch independent (kernel/cpu.c) are flagged as __cpuinit -- so if we remove the __cpuinit from arch specific callers, we will also get section mismatch warnings. As an intermediate step, we intend to turn the linux/init.h cpuinit content into no-ops as early as possible, since that will get rid of these warnings. In any case, they are temporary and harmless. This removes all the arch/sparc uses of the __cpuinit macros from C files and removes __CPUINIT from assembly files. Note that even though arch/sparc/kernel/trampoline_64.S has instances of ".previous" in it, they are all paired off against explicit ".section" directives, and not implicitly paired with __CPUINIT (unlike mips and arm were). [1] https://lkml.org/lkml/2013/5/20/589 Cc: "David S. Miller" <davem@davemloft.net> Cc: sparclinux@vger.kernel.org Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
2013-06-17 23:43:14 +04:00
void sun4v_ktsb_register(void)
{
unsigned long pa, ret;
pa = kern_base + ((unsigned long)&ktsb_descr[0] - KERNBASE);
ret = sun4v_mmu_tsb_ctx0(NUM_KTSB_DESCR, pa);
if (ret != 0) {
prom_printf("hypervisor_mmu_tsb_ctx0[%lx]: "
"errors with %lx\n", pa, ret);
prom_halt();
}
}
static void __init sun4u_linear_pte_xor_finalize(void)
{
#ifndef CONFIG_DEBUG_PAGEALLOC
/* This is where we would add Panther support for
* 32MB and 256MB pages.
*/
#endif
}
static void __init sun4v_linear_pte_xor_finalize(void)
{
unsigned long pagecv_flag;
/* Bit 9 of TTE is no longer CV bit on M7 processor and it instead
* enables MCD error. Do not set bit 9 on M7 processor.
*/
switch (sun4v_chip_type) {
case SUN4V_CHIP_SPARC_M7:
case SUN4V_CHIP_SPARC_M8:
case SUN4V_CHIP_SPARC_SN:
pagecv_flag = 0x00;
break;
default:
pagecv_flag = _PAGE_CV_4V;
break;
}
#ifndef CONFIG_DEBUG_PAGEALLOC
if (cpu_pgsz_mask & HV_PGSZ_MASK_256MB) {
kern_linear_pte_xor[1] = (_PAGE_VALID | _PAGE_SZ256MB_4V) ^
PAGE_OFFSET;
kern_linear_pte_xor[1] |= (_PAGE_CP_4V | pagecv_flag |
_PAGE_P_4V | _PAGE_W_4V);
} else {
kern_linear_pte_xor[1] = kern_linear_pte_xor[0];
}
if (cpu_pgsz_mask & HV_PGSZ_MASK_2GB) {
kern_linear_pte_xor[2] = (_PAGE_VALID | _PAGE_SZ2GB_4V) ^
PAGE_OFFSET;
kern_linear_pte_xor[2] |= (_PAGE_CP_4V | pagecv_flag |
_PAGE_P_4V | _PAGE_W_4V);
} else {
kern_linear_pte_xor[2] = kern_linear_pte_xor[1];
}
if (cpu_pgsz_mask & HV_PGSZ_MASK_16GB) {
kern_linear_pte_xor[3] = (_PAGE_VALID | _PAGE_SZ16GB_4V) ^
PAGE_OFFSET;
kern_linear_pte_xor[3] |= (_PAGE_CP_4V | pagecv_flag |
_PAGE_P_4V | _PAGE_W_4V);
} else {
kern_linear_pte_xor[3] = kern_linear_pte_xor[2];
}
#endif
}
/* paging_init() sets up the page tables */
static unsigned long last_valid_pfn;
static void sun4u_pgprot_init(void);
static void sun4v_pgprot_init(void);
sparc64: mem boot option correction The "mem" boot option can result in many unexpected consequences. This patch attempts to prevent boot hangs which have been experienced on T4-4 and T5-8. Basically the boot loader allocates vmlinuz and initrd higher in available OBP physical memory. For example, on a 2Tb T5-8 it isn't possible to boot with mem=20G. The patch utilizes memblock to avoid reserved regions and trim memory which is only free. Other improvements are possible for a multi-node machine. This is a snippet of the boot log with mem=20G on T5-8 with the patch applied: MEMBLOCK configuration: <- before memory reduction memory size = 0x1ffad6ce000 reserved size = 0xa1adf44 memory.cnt = 0xb memory[0x0] [0x00000030400000-0x00003fdde47fff], 0x3fada48000 bytes memory[0x1] [0x00003fdde4e000-0x00003fdde4ffff], 0x2000 bytes memory[0x2] [0x00080000000000-0x00083fffffffff], 0x4000000000 bytes memory[0x3] [0x00100000000000-0x00103fffffffff], 0x4000000000 bytes memory[0x4] [0x00180000000000-0x00183fffffffff], 0x4000000000 bytes memory[0x5] [0x00200000000000-0x00203fffffffff], 0x4000000000 bytes memory[0x6] [0x00280000000000-0x00283fffffffff], 0x4000000000 bytes memory[0x7] [0x00300000000000-0x00303fffffffff], 0x4000000000 bytes memory[0x8] [0x00380000000000-0x00383fffc71fff], 0x3fffc72000 bytes memory[0x9] [0x00383fffc92000-0x00383fffca1fff], 0x10000 bytes memory[0xa] [0x00383fffcb4000-0x00383fffcb5fff], 0x2000 bytes reserved.cnt = 0x2 reserved[0x0] [0x00380000000000-0x0038000117e7f8], 0x117e7f9 bytes reserved[0x1] [0x00380004000000-0x0038000d02f74a], 0x902f74b bytes ... MEMBLOCK configuration: <- after reduction of memory memory size = 0x50a1adf44 reserved size = 0xa1adf44 memory.cnt = 0x4 memory[0x0] [0x00380000000000-0x0038000117e7f8], 0x117e7f9 bytes memory[0x1] [0x00380004000000-0x0038050d01d74a], 0x50901d74b bytes memory[0x2] [0x00383fffc92000-0x00383fffca1fff], 0x10000 bytes memory[0x3] [0x00383fffcb4000-0x00383fffcb5fff], 0x2000 bytes reserved.cnt = 0x2 reserved[0x0] [0x00380000000000-0x0038000117e7f8], 0x117e7f9 bytes reserved[0x1] [0x00380004000000-0x0038000d02f74a], 0x902f74b bytes ... Early memory node ranges node 7: [mem 0x380000000000-0x38000117dfff] node 7: [mem 0x380004000000-0x380f0d01bfff] node 7: [mem 0x383fffc92000-0x383fffca1fff] node 7: [mem 0x383fffcb4000-0x383fffcb5fff] Could not find start_pfn for node 0 Could not find start_pfn for node 1 Could not find start_pfn for node 2 Could not find start_pfn for node 3 Could not find start_pfn for node 4 Could not find start_pfn for node 5 Could not find start_pfn for node 6 . The patch was tested on T4-1, T5-8 and Jalap?no. Cc: sparclinux@vger.kernel.org Signed-off-by: Bob Picco <bob.picco@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-09-16 17:29:54 +04:00
static phys_addr_t __init available_memory(void)
{
phys_addr_t available = 0ULL;
phys_addr_t pa_start, pa_end;
u64 i;
mm/memblock: add extra "flags" to memblock to allow selection of memory based on attribute Some high end Intel Xeon systems report uncorrectable memory errors as a recoverable machine check. Linux has included code for some time to process these and just signal the affected processes (or even recover completely if the error was in a read only page that can be replaced by reading from disk). But we have no recovery path for errors encountered during kernel code execution. Except for some very specific cases were are unlikely to ever be able to recover. Enter memory mirroring. Actually 3rd generation of memory mirroing. Gen1: All memory is mirrored Pro: No s/w enabling - h/w just gets good data from other side of the mirror Con: Halves effective memory capacity available to OS/applications Gen2: Partial memory mirror - just mirror memory begind some memory controllers Pro: Keep more of the capacity Con: Nightmare to enable. Have to choose between allocating from mirrored memory for safety vs. NUMA local memory for performance Gen3: Address range partial memory mirror - some mirror on each memory controller Pro: Can tune the amount of mirror and keep NUMA performance Con: I have to write memory management code to implement The current plan is just to use mirrored memory for kernel allocations. This has been broken into two phases: 1) This patch series - find the mirrored memory, use it for boot time allocations 2) Wade into mm/page_alloc.c and define a ZONE_MIRROR to pick up the unused mirrored memory from mm/memblock.c and only give it out to select kernel allocations (this is still being scoped because page_alloc.c is scary). This patch (of 3): Add extra "flags" to memblock to allow selection of memory based on attribute. No functional changes Signed-off-by: Tony Luck <tony.luck@intel.com> Cc: Xishi Qiu <qiuxishi@huawei.com> Cc: Hanjun Guo <guohanjun@huawei.com> Cc: Xiexiuqi <xiexiuqi@huawei.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Yinghai Lu <yinghai@kernel.org> Cc: Naoya Horiguchi <nao.horiguchi@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-06-25 02:58:09 +03:00
for_each_free_mem_range(i, NUMA_NO_NODE, MEMBLOCK_NONE, &pa_start,
&pa_end, NULL)
sparc64: mem boot option correction The "mem" boot option can result in many unexpected consequences. This patch attempts to prevent boot hangs which have been experienced on T4-4 and T5-8. Basically the boot loader allocates vmlinuz and initrd higher in available OBP physical memory. For example, on a 2Tb T5-8 it isn't possible to boot with mem=20G. The patch utilizes memblock to avoid reserved regions and trim memory which is only free. Other improvements are possible for a multi-node machine. This is a snippet of the boot log with mem=20G on T5-8 with the patch applied: MEMBLOCK configuration: <- before memory reduction memory size = 0x1ffad6ce000 reserved size = 0xa1adf44 memory.cnt = 0xb memory[0x0] [0x00000030400000-0x00003fdde47fff], 0x3fada48000 bytes memory[0x1] [0x00003fdde4e000-0x00003fdde4ffff], 0x2000 bytes memory[0x2] [0x00080000000000-0x00083fffffffff], 0x4000000000 bytes memory[0x3] [0x00100000000000-0x00103fffffffff], 0x4000000000 bytes memory[0x4] [0x00180000000000-0x00183fffffffff], 0x4000000000 bytes memory[0x5] [0x00200000000000-0x00203fffffffff], 0x4000000000 bytes memory[0x6] [0x00280000000000-0x00283fffffffff], 0x4000000000 bytes memory[0x7] [0x00300000000000-0x00303fffffffff], 0x4000000000 bytes memory[0x8] [0x00380000000000-0x00383fffc71fff], 0x3fffc72000 bytes memory[0x9] [0x00383fffc92000-0x00383fffca1fff], 0x10000 bytes memory[0xa] [0x00383fffcb4000-0x00383fffcb5fff], 0x2000 bytes reserved.cnt = 0x2 reserved[0x0] [0x00380000000000-0x0038000117e7f8], 0x117e7f9 bytes reserved[0x1] [0x00380004000000-0x0038000d02f74a], 0x902f74b bytes ... MEMBLOCK configuration: <- after reduction of memory memory size = 0x50a1adf44 reserved size = 0xa1adf44 memory.cnt = 0x4 memory[0x0] [0x00380000000000-0x0038000117e7f8], 0x117e7f9 bytes memory[0x1] [0x00380004000000-0x0038050d01d74a], 0x50901d74b bytes memory[0x2] [0x00383fffc92000-0x00383fffca1fff], 0x10000 bytes memory[0x3] [0x00383fffcb4000-0x00383fffcb5fff], 0x2000 bytes reserved.cnt = 0x2 reserved[0x0] [0x00380000000000-0x0038000117e7f8], 0x117e7f9 bytes reserved[0x1] [0x00380004000000-0x0038000d02f74a], 0x902f74b bytes ... Early memory node ranges node 7: [mem 0x380000000000-0x38000117dfff] node 7: [mem 0x380004000000-0x380f0d01bfff] node 7: [mem 0x383fffc92000-0x383fffca1fff] node 7: [mem 0x383fffcb4000-0x383fffcb5fff] Could not find start_pfn for node 0 Could not find start_pfn for node 1 Could not find start_pfn for node 2 Could not find start_pfn for node 3 Could not find start_pfn for node 4 Could not find start_pfn for node 5 Could not find start_pfn for node 6 . The patch was tested on T4-1, T5-8 and Jalap?no. Cc: sparclinux@vger.kernel.org Signed-off-by: Bob Picco <bob.picco@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-09-16 17:29:54 +04:00
available = available + (pa_end - pa_start);
return available;
}
#define _PAGE_CACHE_4U (_PAGE_CP_4U | _PAGE_CV_4U)
#define _PAGE_CACHE_4V (_PAGE_CP_4V | _PAGE_CV_4V)
#define __DIRTY_BITS_4U (_PAGE_MODIFIED_4U | _PAGE_WRITE_4U | _PAGE_W_4U)
#define __DIRTY_BITS_4V (_PAGE_MODIFIED_4V | _PAGE_WRITE_4V | _PAGE_W_4V)
#define __ACCESS_BITS_4U (_PAGE_ACCESSED_4U | _PAGE_READ_4U | _PAGE_R)
#define __ACCESS_BITS_4V (_PAGE_ACCESSED_4V | _PAGE_READ_4V | _PAGE_R)
sparc64: mem boot option correction The "mem" boot option can result in many unexpected consequences. This patch attempts to prevent boot hangs which have been experienced on T4-4 and T5-8. Basically the boot loader allocates vmlinuz and initrd higher in available OBP physical memory. For example, on a 2Tb T5-8 it isn't possible to boot with mem=20G. The patch utilizes memblock to avoid reserved regions and trim memory which is only free. Other improvements are possible for a multi-node machine. This is a snippet of the boot log with mem=20G on T5-8 with the patch applied: MEMBLOCK configuration: <- before memory reduction memory size = 0x1ffad6ce000 reserved size = 0xa1adf44 memory.cnt = 0xb memory[0x0] [0x00000030400000-0x00003fdde47fff], 0x3fada48000 bytes memory[0x1] [0x00003fdde4e000-0x00003fdde4ffff], 0x2000 bytes memory[0x2] [0x00080000000000-0x00083fffffffff], 0x4000000000 bytes memory[0x3] [0x00100000000000-0x00103fffffffff], 0x4000000000 bytes memory[0x4] [0x00180000000000-0x00183fffffffff], 0x4000000000 bytes memory[0x5] [0x00200000000000-0x00203fffffffff], 0x4000000000 bytes memory[0x6] [0x00280000000000-0x00283fffffffff], 0x4000000000 bytes memory[0x7] [0x00300000000000-0x00303fffffffff], 0x4000000000 bytes memory[0x8] [0x00380000000000-0x00383fffc71fff], 0x3fffc72000 bytes memory[0x9] [0x00383fffc92000-0x00383fffca1fff], 0x10000 bytes memory[0xa] [0x00383fffcb4000-0x00383fffcb5fff], 0x2000 bytes reserved.cnt = 0x2 reserved[0x0] [0x00380000000000-0x0038000117e7f8], 0x117e7f9 bytes reserved[0x1] [0x00380004000000-0x0038000d02f74a], 0x902f74b bytes ... MEMBLOCK configuration: <- after reduction of memory memory size = 0x50a1adf44 reserved size = 0xa1adf44 memory.cnt = 0x4 memory[0x0] [0x00380000000000-0x0038000117e7f8], 0x117e7f9 bytes memory[0x1] [0x00380004000000-0x0038050d01d74a], 0x50901d74b bytes memory[0x2] [0x00383fffc92000-0x00383fffca1fff], 0x10000 bytes memory[0x3] [0x00383fffcb4000-0x00383fffcb5fff], 0x2000 bytes reserved.cnt = 0x2 reserved[0x0] [0x00380000000000-0x0038000117e7f8], 0x117e7f9 bytes reserved[0x1] [0x00380004000000-0x0038000d02f74a], 0x902f74b bytes ... Early memory node ranges node 7: [mem 0x380000000000-0x38000117dfff] node 7: [mem 0x380004000000-0x380f0d01bfff] node 7: [mem 0x383fffc92000-0x383fffca1fff] node 7: [mem 0x383fffcb4000-0x383fffcb5fff] Could not find start_pfn for node 0 Could not find start_pfn for node 1 Could not find start_pfn for node 2 Could not find start_pfn for node 3 Could not find start_pfn for node 4 Could not find start_pfn for node 5 Could not find start_pfn for node 6 . The patch was tested on T4-1, T5-8 and Jalap?no. Cc: sparclinux@vger.kernel.org Signed-off-by: Bob Picco <bob.picco@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-09-16 17:29:54 +04:00
/* We need to exclude reserved regions. This exclusion will include
* vmlinux and initrd. To be more precise the initrd size could be used to
* compute a new lower limit because it is freed later during initialization.
*/
static void __init reduce_memory(phys_addr_t limit_ram)
{
phys_addr_t avail_ram = available_memory();
phys_addr_t pa_start, pa_end;
u64 i;
if (limit_ram >= avail_ram)
return;
mm/memblock: add extra "flags" to memblock to allow selection of memory based on attribute Some high end Intel Xeon systems report uncorrectable memory errors as a recoverable machine check. Linux has included code for some time to process these and just signal the affected processes (or even recover completely if the error was in a read only page that can be replaced by reading from disk). But we have no recovery path for errors encountered during kernel code execution. Except for some very specific cases were are unlikely to ever be able to recover. Enter memory mirroring. Actually 3rd generation of memory mirroing. Gen1: All memory is mirrored Pro: No s/w enabling - h/w just gets good data from other side of the mirror Con: Halves effective memory capacity available to OS/applications Gen2: Partial memory mirror - just mirror memory begind some memory controllers Pro: Keep more of the capacity Con: Nightmare to enable. Have to choose between allocating from mirrored memory for safety vs. NUMA local memory for performance Gen3: Address range partial memory mirror - some mirror on each memory controller Pro: Can tune the amount of mirror and keep NUMA performance Con: I have to write memory management code to implement The current plan is just to use mirrored memory for kernel allocations. This has been broken into two phases: 1) This patch series - find the mirrored memory, use it for boot time allocations 2) Wade into mm/page_alloc.c and define a ZONE_MIRROR to pick up the unused mirrored memory from mm/memblock.c and only give it out to select kernel allocations (this is still being scoped because page_alloc.c is scary). This patch (of 3): Add extra "flags" to memblock to allow selection of memory based on attribute. No functional changes Signed-off-by: Tony Luck <tony.luck@intel.com> Cc: Xishi Qiu <qiuxishi@huawei.com> Cc: Hanjun Guo <guohanjun@huawei.com> Cc: Xiexiuqi <xiexiuqi@huawei.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Yinghai Lu <yinghai@kernel.org> Cc: Naoya Horiguchi <nao.horiguchi@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-06-25 02:58:09 +03:00
for_each_free_mem_range(i, NUMA_NO_NODE, MEMBLOCK_NONE, &pa_start,
&pa_end, NULL) {
sparc64: mem boot option correction The "mem" boot option can result in many unexpected consequences. This patch attempts to prevent boot hangs which have been experienced on T4-4 and T5-8. Basically the boot loader allocates vmlinuz and initrd higher in available OBP physical memory. For example, on a 2Tb T5-8 it isn't possible to boot with mem=20G. The patch utilizes memblock to avoid reserved regions and trim memory which is only free. Other improvements are possible for a multi-node machine. This is a snippet of the boot log with mem=20G on T5-8 with the patch applied: MEMBLOCK configuration: <- before memory reduction memory size = 0x1ffad6ce000 reserved size = 0xa1adf44 memory.cnt = 0xb memory[0x0] [0x00000030400000-0x00003fdde47fff], 0x3fada48000 bytes memory[0x1] [0x00003fdde4e000-0x00003fdde4ffff], 0x2000 bytes memory[0x2] [0x00080000000000-0x00083fffffffff], 0x4000000000 bytes memory[0x3] [0x00100000000000-0x00103fffffffff], 0x4000000000 bytes memory[0x4] [0x00180000000000-0x00183fffffffff], 0x4000000000 bytes memory[0x5] [0x00200000000000-0x00203fffffffff], 0x4000000000 bytes memory[0x6] [0x00280000000000-0x00283fffffffff], 0x4000000000 bytes memory[0x7] [0x00300000000000-0x00303fffffffff], 0x4000000000 bytes memory[0x8] [0x00380000000000-0x00383fffc71fff], 0x3fffc72000 bytes memory[0x9] [0x00383fffc92000-0x00383fffca1fff], 0x10000 bytes memory[0xa] [0x00383fffcb4000-0x00383fffcb5fff], 0x2000 bytes reserved.cnt = 0x2 reserved[0x0] [0x00380000000000-0x0038000117e7f8], 0x117e7f9 bytes reserved[0x1] [0x00380004000000-0x0038000d02f74a], 0x902f74b bytes ... MEMBLOCK configuration: <- after reduction of memory memory size = 0x50a1adf44 reserved size = 0xa1adf44 memory.cnt = 0x4 memory[0x0] [0x00380000000000-0x0038000117e7f8], 0x117e7f9 bytes memory[0x1] [0x00380004000000-0x0038050d01d74a], 0x50901d74b bytes memory[0x2] [0x00383fffc92000-0x00383fffca1fff], 0x10000 bytes memory[0x3] [0x00383fffcb4000-0x00383fffcb5fff], 0x2000 bytes reserved.cnt = 0x2 reserved[0x0] [0x00380000000000-0x0038000117e7f8], 0x117e7f9 bytes reserved[0x1] [0x00380004000000-0x0038000d02f74a], 0x902f74b bytes ... Early memory node ranges node 7: [mem 0x380000000000-0x38000117dfff] node 7: [mem 0x380004000000-0x380f0d01bfff] node 7: [mem 0x383fffc92000-0x383fffca1fff] node 7: [mem 0x383fffcb4000-0x383fffcb5fff] Could not find start_pfn for node 0 Could not find start_pfn for node 1 Could not find start_pfn for node 2 Could not find start_pfn for node 3 Could not find start_pfn for node 4 Could not find start_pfn for node 5 Could not find start_pfn for node 6 . The patch was tested on T4-1, T5-8 and Jalap?no. Cc: sparclinux@vger.kernel.org Signed-off-by: Bob Picco <bob.picco@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-09-16 17:29:54 +04:00
phys_addr_t region_size = pa_end - pa_start;
phys_addr_t clip_start = pa_start;
avail_ram = avail_ram - region_size;
/* Are we consuming too much? */
if (avail_ram < limit_ram) {
phys_addr_t give_back = limit_ram - avail_ram;
region_size = region_size - give_back;
clip_start = clip_start + give_back;
}
memblock_remove(clip_start, region_size);
if (avail_ram <= limit_ram)
break;
i = 0UL;
}
}
void __init paging_init(void)
{
unsigned long end_pfn, shift, phys_base;
unsigned long real_end, i;
setup_page_offset();
/* These build time checkes make sure that the dcache_dirty_cpu()
* page->flags usage will work.
*
* When a page gets marked as dcache-dirty, we store the
* cpu number starting at bit 32 in the page->flags. Also,
* functions like clear_dcache_dirty_cpu use the cpu mask
* in 13-bit signed-immediate instruction fields.
*/
/*
* Page flags must not reach into upper 32 bits that are used
* for the cpu number
*/
BUILD_BUG_ON(NR_PAGEFLAGS > 32);
/*
* The bit fields placed in the high range must not reach below
* the 32 bit boundary. Otherwise we cannot place the cpu field
* at the 32 bit boundary.
*/
BUILD_BUG_ON(SECTIONS_WIDTH + NODES_WIDTH + ZONES_WIDTH +
ilog2(roundup_pow_of_two(NR_CPUS)) > 32);
BUILD_BUG_ON(NR_CPUS > 4096);
kern_base = (prom_boot_mapping_phys_low >> ILOG2_4MB) << ILOG2_4MB;
kern_size = (unsigned long)&_end - (unsigned long)KERNBASE;
/* Invalidate both kernel TSBs. */
memset(swapper_tsb, 0x40, sizeof(swapper_tsb));
#ifndef CONFIG_DEBUG_PAGEALLOC
memset(swapper_4m_tsb, 0x40, sizeof(swapper_4m_tsb));
#endif
/* TTE.cv bit on sparc v9 occupies the same position as TTE.mcde
* bit on M7 processor. This is a conflicting usage of the same
* bit. Enabling TTE.cv on M7 would turn on Memory Corruption
* Detection error on all pages and this will lead to problems
* later. Kernel does not run with MCD enabled and hence rest
* of the required steps to fully configure memory corruption
* detection are not taken. We need to ensure TTE.mcde is not
* set on M7 processor. Compute the value of cacheability
* flag for use later taking this into consideration.
*/
switch (sun4v_chip_type) {
case SUN4V_CHIP_SPARC_M7:
case SUN4V_CHIP_SPARC_M8:
case SUN4V_CHIP_SPARC_SN:
page_cache4v_flag = _PAGE_CP_4V;
break;
default:
page_cache4v_flag = _PAGE_CACHE_4V;
break;
}
if (tlb_type == hypervisor)
sun4v_pgprot_init();
else
sun4u_pgprot_init();
if (tlb_type == cheetah_plus ||
tlb_type == hypervisor) {
tsb_phys_patch();
ktsb_phys_patch();
}
if (tlb_type == hypervisor)
sun4v_patch_tlb_handlers();
/* Find available physical memory...
*
* Read it twice in order to work around a bug in openfirmware.
* The call to grab this table itself can cause openfirmware to
* allocate memory, which in turn can take away some space from
* the list of available memory. Reading it twice makes sure
* we really do get the final value.
*/
read_obp_translations();
read_obp_memory("reg", &pall[0], &pall_ents);
read_obp_memory("available", &pavail[0], &pavail_ents);
read_obp_memory("available", &pavail[0], &pavail_ents);
phys_base = 0xffffffffffffffffUL;
for (i = 0; i < pavail_ents; i++) {
phys_base = min(phys_base, pavail[i].phys_addr);
memblock_add(pavail[i].phys_addr, pavail[i].reg_size);
}
memblock_reserve(kern_base, kern_size);
find_ramdisk(phys_base);
sparc64: mem boot option correction The "mem" boot option can result in many unexpected consequences. This patch attempts to prevent boot hangs which have been experienced on T4-4 and T5-8. Basically the boot loader allocates vmlinuz and initrd higher in available OBP physical memory. For example, on a 2Tb T5-8 it isn't possible to boot with mem=20G. The patch utilizes memblock to avoid reserved regions and trim memory which is only free. Other improvements are possible for a multi-node machine. This is a snippet of the boot log with mem=20G on T5-8 with the patch applied: MEMBLOCK configuration: <- before memory reduction memory size = 0x1ffad6ce000 reserved size = 0xa1adf44 memory.cnt = 0xb memory[0x0] [0x00000030400000-0x00003fdde47fff], 0x3fada48000 bytes memory[0x1] [0x00003fdde4e000-0x00003fdde4ffff], 0x2000 bytes memory[0x2] [0x00080000000000-0x00083fffffffff], 0x4000000000 bytes memory[0x3] [0x00100000000000-0x00103fffffffff], 0x4000000000 bytes memory[0x4] [0x00180000000000-0x00183fffffffff], 0x4000000000 bytes memory[0x5] [0x00200000000000-0x00203fffffffff], 0x4000000000 bytes memory[0x6] [0x00280000000000-0x00283fffffffff], 0x4000000000 bytes memory[0x7] [0x00300000000000-0x00303fffffffff], 0x4000000000 bytes memory[0x8] [0x00380000000000-0x00383fffc71fff], 0x3fffc72000 bytes memory[0x9] [0x00383fffc92000-0x00383fffca1fff], 0x10000 bytes memory[0xa] [0x00383fffcb4000-0x00383fffcb5fff], 0x2000 bytes reserved.cnt = 0x2 reserved[0x0] [0x00380000000000-0x0038000117e7f8], 0x117e7f9 bytes reserved[0x1] [0x00380004000000-0x0038000d02f74a], 0x902f74b bytes ... MEMBLOCK configuration: <- after reduction of memory memory size = 0x50a1adf44 reserved size = 0xa1adf44 memory.cnt = 0x4 memory[0x0] [0x00380000000000-0x0038000117e7f8], 0x117e7f9 bytes memory[0x1] [0x00380004000000-0x0038050d01d74a], 0x50901d74b bytes memory[0x2] [0x00383fffc92000-0x00383fffca1fff], 0x10000 bytes memory[0x3] [0x00383fffcb4000-0x00383fffcb5fff], 0x2000 bytes reserved.cnt = 0x2 reserved[0x0] [0x00380000000000-0x0038000117e7f8], 0x117e7f9 bytes reserved[0x1] [0x00380004000000-0x0038000d02f74a], 0x902f74b bytes ... Early memory node ranges node 7: [mem 0x380000000000-0x38000117dfff] node 7: [mem 0x380004000000-0x380f0d01bfff] node 7: [mem 0x383fffc92000-0x383fffca1fff] node 7: [mem 0x383fffcb4000-0x383fffcb5fff] Could not find start_pfn for node 0 Could not find start_pfn for node 1 Could not find start_pfn for node 2 Could not find start_pfn for node 3 Could not find start_pfn for node 4 Could not find start_pfn for node 5 Could not find start_pfn for node 6 . The patch was tested on T4-1, T5-8 and Jalap?no. Cc: sparclinux@vger.kernel.org Signed-off-by: Bob Picco <bob.picco@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-09-16 17:29:54 +04:00
if (cmdline_memory_size)
reduce_memory(cmdline_memory_size);
memblock: s/memblock_analyze()/memblock_allow_resize()/ and update users The only function of memblock_analyze() is now allowing resize of memblock region arrays. Rename it to memblock_allow_resize() and update its users. * The following users remain the same other than renaming. arm/mm/init.c::arm_memblock_init() microblaze/kernel/prom.c::early_init_devtree() powerpc/kernel/prom.c::early_init_devtree() openrisc/kernel/prom.c::early_init_devtree() sh/mm/init.c::paging_init() sparc/mm/init_64.c::paging_init() unicore32/mm/init.c::uc32_memblock_init() * In the following users, analyze was used to update total size which is no longer necessary. powerpc/kernel/machine_kexec.c::reserve_crashkernel() powerpc/kernel/prom.c::early_init_devtree() powerpc/mm/init_32.c::MMU_init() powerpc/mm/tlb_nohash.c::__early_init_mmu() powerpc/platforms/ps3/mm.c::ps3_mm_add_memory() powerpc/platforms/embedded6xx/wii.c::wii_memory_fixups() sh/kernel/machine_kexec.c::reserve_crashkernel() * x86/kernel/e820.c::memblock_x86_fill() was directly setting memblock_can_resize before populating memblock and calling analyze afterwards. Call memblock_allow_resize() before start populating. memblock_can_resize is now static inside memblock.c. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Yinghai Lu <yinghai@kernel.org> Cc: Russell King <linux@arm.linux.org.uk> Cc: Michal Simek <monstr@monstr.eu> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Guan Xuetao <gxt@mprc.pku.edu.cn> Cc: "H. Peter Anvin" <hpa@zytor.com>
2011-12-08 22:22:08 +04:00
memblock_allow_resize();
memblock_dump_all();
set_bit(0, mmu_context_bmap);
shift = kern_base + PAGE_OFFSET - ((unsigned long)KERNBASE);
real_end = (unsigned long)_end;
num_kernel_image_mappings = DIV_ROUND_UP(real_end - KERNBASE, 1 << ILOG2_4MB);
printk("Kernel: Using %d locked TLB entries for main kernel image.\n",
num_kernel_image_mappings);
/* Set kernel pgd to upper alias so physical page computations
* work.
*/
init_mm.pgd += ((shift) / (sizeof(pgd_t)));
memset(swapper_pg_dir, 0, sizeof(swapper_pg_dir));
sparc64: Fix physical memory management regressions with large max_phys_bits. If max_phys_bits needs to be > 43 (f.e. for T4 chips), things like DEBUG_PAGEALLOC stop working because the 3-level page tables only can cover up to 43 bits. Another problem is that when we increased MAX_PHYS_ADDRESS_BITS up to 47, several statically allocated tables became enormous. Compounding this is that we will need to support up to 49 bits of physical addressing for M7 chips. The two tables in question are sparc64_valid_addr_bitmap and kpte_linear_bitmap. The first holds a bitmap, with 1 bit for each 4MB chunk of physical memory, indicating whether that chunk actually exists in the machine and is valid. The second table is a set of 2-bit values which tell how large of a mapping (4MB, 256MB, 2GB, 16GB, respectively) we can use at each 256MB chunk of ram in the system. These tables are huge and take up an enormous amount of the BSS section of the sparc64 kernel image. Specifically, the sparc64_valid_addr_bitmap is 4MB, and the kpte_linear_bitmap is 128K. So let's solve the space wastage and the DEBUG_PAGEALLOC problem at the same time, by using the kernel page tables (as designed) to manage this information. We have to keep using large mappings when DEBUG_PAGEALLOC is disabled, and we do this by encoding huge PMDs and PUDs. On a T4-2 with 256GB of ram the kernel page table takes up 16K with DEBUG_PAGEALLOC disabled and 256MB with it enabled. Furthermore, this memory is dynamically allocated at run time rather than coded statically into the kernel image. Signed-off-by: David S. Miller <davem@davemloft.net> Acked-by: Bob Picco <bob.picco@oracle.com>
2014-09-25 07:56:11 +04:00
[SPARC64]: Fix boot failures on SunBlade-150 The sequence to move over to the Linux trap tables from the firmware ones needs to be more air tight. It turns out that to be %100 safe we do need to be able to translate OBP mappings in our TLB miss handlers early. In order not to eat up a lot of kernel image memory with static page tables, just use the translations array in the OBP TLB miss handlers. That solves the bulk of the problem. Furthermore, to make sure the OBP TLB miss path will work even before the fixed MMU globals are loaded, explicitly load %g1 to TLB_SFSR at the beginning of the i-TLB and d-TLB miss handlers. To ease the OBP TLB miss walking of the prom_trans[] array, we sort it then delete all of the non-OBP entries in there (for example, there are entries for the kernel image itself which we're not interested in at all). We also save about 32K of kernel image size with this change. Not a bad side effect :-) There are still some reasons why trampoline.S can't use the setup_trap_table() yet. The most noteworthy are: 1) OBP boots secondary processors with non-bias'd stack for some reason. This is easily fixed by using a small bootup stack in the kernel image explicitly for this purpose. 2) Doing a firmware call via the normal C call prom_set_trap_table() goes through the whole OBP enter/exit sequence that saves and restores OBP and Linux kernel state in the MMUs. This path unfortunately does a "flush %g6" while loading up the OBP locked TLB entries for the firmware call. If we setup the %g6 in the trampoline.S code properly, that is in the PAGE_OFFSET linear mapping, but we're not on the kernel trap table yet so those addresses won't translate properly. One idea is to do a by-hand firmware call like we do in the early bootup code and elsewhere here in trampoline.S But this fails as well, as aparently the secondary processors are not booted with OBP's special locked TLB entries loaded. These are necessary for the firwmare to processes TLB misses correctly up until the point where we take over the trap table. This does need to be resolved at some point. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-10-12 23:22:46 +04:00
inherit_prom_mappings();
/* Ok, we can use our TLB miss and window trap handlers safely. */
setup_tba();
[SPARC64]: Fix boot failures on SunBlade-150 The sequence to move over to the Linux trap tables from the firmware ones needs to be more air tight. It turns out that to be %100 safe we do need to be able to translate OBP mappings in our TLB miss handlers early. In order not to eat up a lot of kernel image memory with static page tables, just use the translations array in the OBP TLB miss handlers. That solves the bulk of the problem. Furthermore, to make sure the OBP TLB miss path will work even before the fixed MMU globals are loaded, explicitly load %g1 to TLB_SFSR at the beginning of the i-TLB and d-TLB miss handlers. To ease the OBP TLB miss walking of the prom_trans[] array, we sort it then delete all of the non-OBP entries in there (for example, there are entries for the kernel image itself which we're not interested in at all). We also save about 32K of kernel image size with this change. Not a bad side effect :-) There are still some reasons why trampoline.S can't use the setup_trap_table() yet. The most noteworthy are: 1) OBP boots secondary processors with non-bias'd stack for some reason. This is easily fixed by using a small bootup stack in the kernel image explicitly for this purpose. 2) Doing a firmware call via the normal C call prom_set_trap_table() goes through the whole OBP enter/exit sequence that saves and restores OBP and Linux kernel state in the MMUs. This path unfortunately does a "flush %g6" while loading up the OBP locked TLB entries for the firmware call. If we setup the %g6 in the trampoline.S code properly, that is in the PAGE_OFFSET linear mapping, but we're not on the kernel trap table yet so those addresses won't translate properly. One idea is to do a by-hand firmware call like we do in the early bootup code and elsewhere here in trampoline.S But this fails as well, as aparently the secondary processors are not booted with OBP's special locked TLB entries loaded. These are necessary for the firwmare to processes TLB misses correctly up until the point where we take over the trap table. This does need to be resolved at some point. Signed-off-by: David S. Miller <davem@davemloft.net>
2005-10-12 23:22:46 +04:00
__flush_tlb_all();
prom_build_devicetree();
of_populate_present_mask();
#ifndef CONFIG_SMP
of_fill_in_cpu_data();
#endif
if (tlb_type == hypervisor) {
sun4v_mdesc_init();
mdesc_populate_present_mask(cpu_all_mask);
#ifndef CONFIG_SMP
mdesc_fill_in_cpu_data(cpu_all_mask);
#endif
mdesc_get_page_sizes(cpu_all_mask, &cpu_pgsz_mask);
sun4v_linear_pte_xor_finalize();
sun4v_ktsb_init();
sun4v_ktsb_register();
} else {
unsigned long impl, ver;
cpu_pgsz_mask = (HV_PGSZ_MASK_8K | HV_PGSZ_MASK_64K |
HV_PGSZ_MASK_512K | HV_PGSZ_MASK_4MB);
__asm__ __volatile__("rdpr %%ver, %0" : "=r" (ver));
impl = ((ver >> 32) & 0xffff);
if (impl == PANTHER_IMPL)
cpu_pgsz_mask |= (HV_PGSZ_MASK_32MB |
HV_PGSZ_MASK_256MB);
sun4u_linear_pte_xor_finalize();
}
/* Flush the TLBs and the 4M TSB so that the updated linear
* pte XOR settings are realized for all mappings.
*/
__flush_tlb_all();
#ifndef CONFIG_DEBUG_PAGEALLOC
memset(swapper_4m_tsb, 0x40, sizeof(swapper_4m_tsb));
#endif
__flush_tlb_all();
/* Setup bootmem... */
last_valid_pfn = end_pfn = bootmem_init(phys_base);
kernel_physical_mapping_init();
{
unsigned long max_zone_pfns[MAX_NR_ZONES];
memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
max_zone_pfns[ZONE_NORMAL] = end_pfn;
free_area_init_nodes(max_zone_pfns);
}
printk("Booting Linux...\n");
}
int page_in_phys_avail(unsigned long paddr)
{
int i;
paddr &= PAGE_MASK;
for (i = 0; i < pavail_ents; i++) {
unsigned long start, end;
start = pavail[i].phys_addr;
end = start + pavail[i].reg_size;
if (paddr >= start && paddr < end)
return 1;
}
if (paddr >= kern_base && paddr < (kern_base + kern_size))
return 1;
#ifdef CONFIG_BLK_DEV_INITRD
if (paddr >= __pa(initrd_start) &&
paddr < __pa(PAGE_ALIGN(initrd_end)))
return 1;
#endif
return 0;
}
static void __init register_page_bootmem_info(void)
{
#ifdef CONFIG_NEED_MULTIPLE_NODES
int i;
for_each_online_node(i)
if (NODE_DATA(i)->node_spanned_pages)
register_page_bootmem_info_node(NODE_DATA(i));
#endif
}
void __init mem_init(void)
{
high_memory = __va(last_valid_pfn << PAGE_SHIFT);
register_page_bootmem_info();
mm: concentrate modification of totalram_pages into the mm core Concentrate code to modify totalram_pages into the mm core, so the arch memory initialized code doesn't need to take care of it. With these changes applied, only following functions from mm core modify global variable totalram_pages: free_bootmem_late(), free_all_bootmem(), free_all_bootmem_node(), adjust_managed_page_count(). With this patch applied, it will be much more easier for us to keep totalram_pages and zone->managed_pages in consistence. Signed-off-by: Jiang Liu <jiang.liu@huawei.com> Acked-by: David Howells <dhowells@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: "Michael S. Tsirkin" <mst@redhat.com> Cc: <sworddragon2@aol.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Catalin Marinas <catalin.marinas@arm.com> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jeremy Fitzhardinge <jeremy@goop.org> Cc: Jianguo Wu <wujianguo@huawei.com> Cc: Joonsoo Kim <js1304@gmail.com> Cc: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Cc: Marek Szyprowski <m.szyprowski@samsung.com> Cc: Mel Gorman <mel@csn.ul.ie> Cc: Michel Lespinasse <walken@google.com> Cc: Minchan Kim <minchan@kernel.org> Cc: Rik van Riel <riel@redhat.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Tang Chen <tangchen@cn.fujitsu.com> Cc: Tejun Heo <tj@kernel.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Wen Congyang <wency@cn.fujitsu.com> Cc: Will Deacon <will.deacon@arm.com> Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com> Cc: Yinghai Lu <yinghai@kernel.org> Cc: Russell King <rmk@arm.linux.org.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-07-04 02:03:24 +04:00
free_all_bootmem();
/*
* Set up the zero page, mark it reserved, so that page count
* is not manipulated when freeing the page from user ptes.
*/
mem_map_zero = alloc_pages(GFP_KERNEL|__GFP_ZERO, 0);
if (mem_map_zero == NULL) {
prom_printf("paging_init: Cannot alloc zero page.\n");
prom_halt();
}
mark_page_reserved(mem_map_zero);
mem_init_print_info(NULL);
if (tlb_type == cheetah || tlb_type == cheetah_plus)
cheetah_ecache_flush_init();
}
void free_initmem(void)
{
unsigned long addr, initend;
int do_free = 1;
/* If the physical memory maps were trimmed by kernel command
* line options, don't even try freeing this initmem stuff up.
* The kernel image could have been in the trimmed out region
* and if so the freeing below will free invalid page structs.
*/
if (cmdline_memory_size)
do_free = 0;
/*
* The init section is aligned to 8k in vmlinux.lds. Page align for >8k pagesizes.
*/
addr = PAGE_ALIGN((unsigned long)(__init_begin));
initend = (unsigned long)(__init_end) & PAGE_MASK;
for (; addr < initend; addr += PAGE_SIZE) {
unsigned long page;
page = (addr +
((unsigned long) __va(kern_base)) -
((unsigned long) KERNBASE));
memset((void *)addr, POISON_FREE_INITMEM, PAGE_SIZE);
if (do_free)
free_reserved_page(virt_to_page(page));
}
}
#ifdef CONFIG_BLK_DEV_INITRD
void free_initrd_mem(unsigned long start, unsigned long end)
{
free_reserved_area((void *)start, (void *)end, POISON_FREE_INITMEM,
"initrd");
}
#endif
pgprot_t PAGE_KERNEL __read_mostly;
EXPORT_SYMBOL(PAGE_KERNEL);
pgprot_t PAGE_KERNEL_LOCKED __read_mostly;
pgprot_t PAGE_COPY __read_mostly;
pgprot_t PAGE_SHARED __read_mostly;
EXPORT_SYMBOL(PAGE_SHARED);
unsigned long pg_iobits __read_mostly;
unsigned long _PAGE_IE __read_mostly;
EXPORT_SYMBOL(_PAGE_IE);
unsigned long _PAGE_E __read_mostly;
EXPORT_SYMBOL(_PAGE_E);
unsigned long _PAGE_CACHE __read_mostly;
EXPORT_SYMBOL(_PAGE_CACHE);
#ifdef CONFIG_SPARSEMEM_VMEMMAP
sparse-vmemmap: specify vmemmap population range in bytes The sparse code, when asking the architecture to populate the vmemmap, specifies the section range as a starting page and a number of pages. This is an awkward interface, because none of the arch-specific code actually thinks of the range in terms of 'struct page' units and always translates it to bytes first. In addition, later patches mix huge page and regular page backing for the vmemmap. For this, they need to call vmemmap_populate_basepages() on sub-section ranges with PAGE_SIZE and PMD_SIZE in mind. But these are not necessarily multiples of the 'struct page' size and so this unit is too coarse. Just translate the section range into bytes once in the generic sparse code, then pass byte ranges down the stack. Signed-off-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Ben Hutchings <ben@decadent.org.uk> Cc: Bernhard Schmidt <Bernhard.Schmidt@lrz.de> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Russell King <rmk@arm.linux.org.uk> Cc: Ingo Molnar <mingo@elte.hu> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: "Luck, Tony" <tony.luck@intel.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Acked-by: David S. Miller <davem@davemloft.net> Tested-by: David S. Miller <davem@davemloft.net> Cc: Wu Fengguang <fengguang.wu@intel.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-04-30 02:07:50 +04:00
int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend,
int node)
{
unsigned long pte_base;
pte_base = (_PAGE_VALID | _PAGE_SZ4MB_4U |
_PAGE_CP_4U | _PAGE_CV_4U |
_PAGE_P_4U | _PAGE_W_4U);
if (tlb_type == hypervisor)
pte_base = (_PAGE_VALID | _PAGE_SZ4MB_4V |
page_cache4v_flag | _PAGE_P_4V | _PAGE_W_4V);
pte_base |= _PAGE_PMD_HUGE;
vstart = vstart & PMD_MASK;
vend = ALIGN(vend, PMD_SIZE);
for (; vstart < vend; vstart += PMD_SIZE) {
pgd_t *pgd = pgd_offset_k(vstart);
unsigned long pte;
pud_t *pud;
pmd_t *pmd;
if (pgd_none(*pgd)) {
pud_t *new = vmemmap_alloc_block(PAGE_SIZE, node);
if (!new)
return -ENOMEM;
pgd_populate(&init_mm, pgd, new);
}
pud = pud_offset(pgd, vstart);
if (pud_none(*pud)) {
pmd_t *new = vmemmap_alloc_block(PAGE_SIZE, node);
if (!new)
return -ENOMEM;
pud_populate(&init_mm, pud, new);
}
pmd = pmd_offset(pud, vstart);
pte = pmd_val(*pmd);
if (!(pte & _PAGE_VALID)) {
void *block = vmemmap_alloc_block(PMD_SIZE, node);
if (!block)
return -ENOMEM;
pmd_val(*pmd) = pte_base | __pa(block);
}
}
return 0;
}
memory-hotplug: implement register_page_bootmem_info_section of sparse-vmemmap For removing memmap region of sparse-vmemmap which is allocated bootmem, memmap region of sparse-vmemmap needs to be registered by get_page_bootmem(). So the patch searches pages of virtual mapping and registers the pages by get_page_bootmem(). NOTE: register_page_bootmem_memmap() is not implemented for ia64, ppc, s390, and sparc. So introduce CONFIG_HAVE_BOOTMEM_INFO_NODE and revert register_page_bootmem_info_node() when platform doesn't support it. It's implemented by adding a new Kconfig option named CONFIG_HAVE_BOOTMEM_INFO_NODE, which will be automatically selected by memory-hotplug feature fully supported archs(currently only on x86_64). Since we have 2 config options called MEMORY_HOTPLUG and MEMORY_HOTREMOVE used for memory hot-add and hot-remove separately, and codes in function register_page_bootmem_info_node() are only used for collecting infomation for hot-remove, so reside it under MEMORY_HOTREMOVE. Besides page_isolation.c selected by MEMORY_ISOLATION under MEMORY_HOTPLUG is also such case, move it too. [mhocko@suse.cz: put register_page_bootmem_memmap inside CONFIG_MEMORY_HOTPLUG_SPARSE] [linfeng@cn.fujitsu.com: introduce CONFIG_HAVE_BOOTMEM_INFO_NODE and revert register_page_bootmem_info_node()] [mhocko@suse.cz: remove the arch specific functions without any implementation] [linfeng@cn.fujitsu.com: mm/Kconfig: move auto selects from MEMORY_HOTPLUG to MEMORY_HOTREMOVE as needed] [rientjes@google.com: fix defined but not used warning] Signed-off-by: Wen Congyang <wency@cn.fujitsu.com> Signed-off-by: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com> Signed-off-by: Tang Chen <tangchen@cn.fujitsu.com> Reviewed-by: Wu Jianguo <wujianguo@huawei.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Jiang Liu <jiang.liu@huawei.com> Cc: Jianguo Wu <wujianguo@huawei.com> Cc: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Cc: Lai Jiangshan <laijs@cn.fujitsu.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: "H. Peter Anvin" <hpa@zytor.com> Signed-off-by: Michal Hocko <mhocko@suse.cz> Signed-off-by: Lin Feng <linfeng@cn.fujitsu.com> Signed-off-by: David Rientjes <rientjes@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-02-23 04:33:00 +04:00
sparse-vmemmap: specify vmemmap population range in bytes The sparse code, when asking the architecture to populate the vmemmap, specifies the section range as a starting page and a number of pages. This is an awkward interface, because none of the arch-specific code actually thinks of the range in terms of 'struct page' units and always translates it to bytes first. In addition, later patches mix huge page and regular page backing for the vmemmap. For this, they need to call vmemmap_populate_basepages() on sub-section ranges with PAGE_SIZE and PMD_SIZE in mind. But these are not necessarily multiples of the 'struct page' size and so this unit is too coarse. Just translate the section range into bytes once in the generic sparse code, then pass byte ranges down the stack. Signed-off-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Ben Hutchings <ben@decadent.org.uk> Cc: Bernhard Schmidt <Bernhard.Schmidt@lrz.de> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Russell King <rmk@arm.linux.org.uk> Cc: Ingo Molnar <mingo@elte.hu> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: "Luck, Tony" <tony.luck@intel.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Acked-by: David S. Miller <davem@davemloft.net> Tested-by: David S. Miller <davem@davemloft.net> Cc: Wu Fengguang <fengguang.wu@intel.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-04-30 02:07:50 +04:00
void vmemmap_free(unsigned long start, unsigned long end)
{
}
#endif /* CONFIG_SPARSEMEM_VMEMMAP */
static void prot_init_common(unsigned long page_none,
unsigned long page_shared,
unsigned long page_copy,
unsigned long page_readonly,
unsigned long page_exec_bit)
{
PAGE_COPY = __pgprot(page_copy);
PAGE_SHARED = __pgprot(page_shared);
protection_map[0x0] = __pgprot(page_none);
protection_map[0x1] = __pgprot(page_readonly & ~page_exec_bit);
protection_map[0x2] = __pgprot(page_copy & ~page_exec_bit);
protection_map[0x3] = __pgprot(page_copy & ~page_exec_bit);
protection_map[0x4] = __pgprot(page_readonly);
protection_map[0x5] = __pgprot(page_readonly);
protection_map[0x6] = __pgprot(page_copy);
protection_map[0x7] = __pgprot(page_copy);
protection_map[0x8] = __pgprot(page_none);
protection_map[0x9] = __pgprot(page_readonly & ~page_exec_bit);
protection_map[0xa] = __pgprot(page_shared & ~page_exec_bit);
protection_map[0xb] = __pgprot(page_shared & ~page_exec_bit);
protection_map[0xc] = __pgprot(page_readonly);
protection_map[0xd] = __pgprot(page_readonly);
protection_map[0xe] = __pgprot(page_shared);
protection_map[0xf] = __pgprot(page_shared);
}
static void __init sun4u_pgprot_init(void)
{
unsigned long page_none, page_shared, page_copy, page_readonly;
unsigned long page_exec_bit;
int i;
PAGE_KERNEL = __pgprot (_PAGE_PRESENT_4U | _PAGE_VALID |
_PAGE_CACHE_4U | _PAGE_P_4U |
__ACCESS_BITS_4U | __DIRTY_BITS_4U |
_PAGE_EXEC_4U);
PAGE_KERNEL_LOCKED = __pgprot (_PAGE_PRESENT_4U | _PAGE_VALID |
_PAGE_CACHE_4U | _PAGE_P_4U |
__ACCESS_BITS_4U | __DIRTY_BITS_4U |
_PAGE_EXEC_4U | _PAGE_L_4U);
_PAGE_IE = _PAGE_IE_4U;
_PAGE_E = _PAGE_E_4U;
_PAGE_CACHE = _PAGE_CACHE_4U;
pg_iobits = (_PAGE_VALID | _PAGE_PRESENT_4U | __DIRTY_BITS_4U |
__ACCESS_BITS_4U | _PAGE_E_4U);
#ifdef CONFIG_DEBUG_PAGEALLOC
kern_linear_pte_xor[0] = _PAGE_VALID ^ PAGE_OFFSET;
#else
kern_linear_pte_xor[0] = (_PAGE_VALID | _PAGE_SZ4MB_4U) ^
PAGE_OFFSET;
#endif
kern_linear_pte_xor[0] |= (_PAGE_CP_4U | _PAGE_CV_4U |
_PAGE_P_4U | _PAGE_W_4U);
for (i = 1; i < 4; i++)
kern_linear_pte_xor[i] = kern_linear_pte_xor[0];
_PAGE_ALL_SZ_BITS = (_PAGE_SZ4MB_4U | _PAGE_SZ512K_4U |
_PAGE_SZ64K_4U | _PAGE_SZ8K_4U |
_PAGE_SZ32MB_4U | _PAGE_SZ256MB_4U);
page_none = _PAGE_PRESENT_4U | _PAGE_ACCESSED_4U | _PAGE_CACHE_4U;
page_shared = (_PAGE_VALID | _PAGE_PRESENT_4U | _PAGE_CACHE_4U |
__ACCESS_BITS_4U | _PAGE_WRITE_4U | _PAGE_EXEC_4U);
page_copy = (_PAGE_VALID | _PAGE_PRESENT_4U | _PAGE_CACHE_4U |
__ACCESS_BITS_4U | _PAGE_EXEC_4U);
page_readonly = (_PAGE_VALID | _PAGE_PRESENT_4U | _PAGE_CACHE_4U |
__ACCESS_BITS_4U | _PAGE_EXEC_4U);
page_exec_bit = _PAGE_EXEC_4U;
prot_init_common(page_none, page_shared, page_copy, page_readonly,
page_exec_bit);
}
static void __init sun4v_pgprot_init(void)
{
unsigned long page_none, page_shared, page_copy, page_readonly;
unsigned long page_exec_bit;
int i;
PAGE_KERNEL = __pgprot (_PAGE_PRESENT_4V | _PAGE_VALID |
page_cache4v_flag | _PAGE_P_4V |
__ACCESS_BITS_4V | __DIRTY_BITS_4V |
_PAGE_EXEC_4V);
PAGE_KERNEL_LOCKED = PAGE_KERNEL;
_PAGE_IE = _PAGE_IE_4V;
_PAGE_E = _PAGE_E_4V;
_PAGE_CACHE = page_cache4v_flag;
#ifdef CONFIG_DEBUG_PAGEALLOC
kern_linear_pte_xor[0] = _PAGE_VALID ^ PAGE_OFFSET;
#else
kern_linear_pte_xor[0] = (_PAGE_VALID | _PAGE_SZ4MB_4V) ^
PAGE_OFFSET;
#endif
kern_linear_pte_xor[0] |= (page_cache4v_flag | _PAGE_P_4V |
_PAGE_W_4V);
for (i = 1; i < 4; i++)
kern_linear_pte_xor[i] = kern_linear_pte_xor[0];
pg_iobits = (_PAGE_VALID | _PAGE_PRESENT_4V | __DIRTY_BITS_4V |
__ACCESS_BITS_4V | _PAGE_E_4V);
_PAGE_ALL_SZ_BITS = (_PAGE_SZ16GB_4V | _PAGE_SZ2GB_4V |
_PAGE_SZ256MB_4V | _PAGE_SZ32MB_4V |
_PAGE_SZ4MB_4V | _PAGE_SZ512K_4V |
_PAGE_SZ64K_4V | _PAGE_SZ8K_4V);
page_none = _PAGE_PRESENT_4V | _PAGE_ACCESSED_4V | page_cache4v_flag;
page_shared = (_PAGE_VALID | _PAGE_PRESENT_4V | page_cache4v_flag |
__ACCESS_BITS_4V | _PAGE_WRITE_4V | _PAGE_EXEC_4V);
page_copy = (_PAGE_VALID | _PAGE_PRESENT_4V | page_cache4v_flag |
__ACCESS_BITS_4V | _PAGE_EXEC_4V);
page_readonly = (_PAGE_VALID | _PAGE_PRESENT_4V | page_cache4v_flag |
__ACCESS_BITS_4V | _PAGE_EXEC_4V);
page_exec_bit = _PAGE_EXEC_4V;
prot_init_common(page_none, page_shared, page_copy, page_readonly,
page_exec_bit);
}
unsigned long pte_sz_bits(unsigned long sz)
{
if (tlb_type == hypervisor) {
switch (sz) {
case 8 * 1024:
default:
return _PAGE_SZ8K_4V;
case 64 * 1024:
return _PAGE_SZ64K_4V;
case 512 * 1024:
return _PAGE_SZ512K_4V;
case 4 * 1024 * 1024:
return _PAGE_SZ4MB_4V;
}
} else {
switch (sz) {
case 8 * 1024:
default:
return _PAGE_SZ8K_4U;
case 64 * 1024:
return _PAGE_SZ64K_4U;
case 512 * 1024:
return _PAGE_SZ512K_4U;
case 4 * 1024 * 1024:
return _PAGE_SZ4MB_4U;
}
}
}
pte_t mk_pte_io(unsigned long page, pgprot_t prot, int space, unsigned long page_size)
{
pte_t pte;
pte_val(pte) = page | pgprot_val(pgprot_noncached(prot));
pte_val(pte) |= (((unsigned long)space) << 32);
pte_val(pte) |= pte_sz_bits(page_size);
return pte;
}
static unsigned long kern_large_tte(unsigned long paddr)
{
unsigned long val;
val = (_PAGE_VALID | _PAGE_SZ4MB_4U |
_PAGE_CP_4U | _PAGE_CV_4U | _PAGE_P_4U |
_PAGE_EXEC_4U | _PAGE_L_4U | _PAGE_W_4U);
if (tlb_type == hypervisor)
val = (_PAGE_VALID | _PAGE_SZ4MB_4V |
page_cache4v_flag | _PAGE_P_4V |
_PAGE_EXEC_4V | _PAGE_W_4V);
return val | paddr;
}
/* If not locked, zap it. */
void __flush_tlb_all(void)
{
unsigned long pstate;
int i;
__asm__ __volatile__("flushw\n\t"
"rdpr %%pstate, %0\n\t"
"wrpr %0, %1, %%pstate"
: "=r" (pstate)
: "i" (PSTATE_IE));
if (tlb_type == hypervisor) {
sun4v_mmu_demap_all();
} else if (tlb_type == spitfire) {
for (i = 0; i < 64; i++) {
/* Spitfire Errata #32 workaround */
/* NOTE: Always runs on spitfire, so no
* cheetah+ page size encodings.
*/
__asm__ __volatile__("stxa %0, [%1] %2\n\t"
"flush %%g6"
: /* No outputs */
: "r" (0),
"r" (PRIMARY_CONTEXT), "i" (ASI_DMMU));
if (!(spitfire_get_dtlb_data(i) & _PAGE_L_4U)) {
__asm__ __volatile__("stxa %%g0, [%0] %1\n\t"
"membar #Sync"
: /* no outputs */
: "r" (TLB_TAG_ACCESS), "i" (ASI_DMMU));
spitfire_put_dtlb_data(i, 0x0UL);
}
/* Spitfire Errata #32 workaround */
/* NOTE: Always runs on spitfire, so no
* cheetah+ page size encodings.
*/
__asm__ __volatile__("stxa %0, [%1] %2\n\t"
"flush %%g6"
: /* No outputs */
: "r" (0),
"r" (PRIMARY_CONTEXT), "i" (ASI_DMMU));
if (!(spitfire_get_itlb_data(i) & _PAGE_L_4U)) {
__asm__ __volatile__("stxa %%g0, [%0] %1\n\t"
"membar #Sync"
: /* no outputs */
: "r" (TLB_TAG_ACCESS), "i" (ASI_IMMU));
spitfire_put_itlb_data(i, 0x0UL);
}
}
} else if (tlb_type == cheetah || tlb_type == cheetah_plus) {
cheetah_flush_dtlb_all();
cheetah_flush_itlb_all();
}
__asm__ __volatile__("wrpr %0, 0, %%pstate"
: : "r" (pstate));
}
pte_t *pte_alloc_one_kernel(struct mm_struct *mm,
unsigned long address)
{
struct page *page = alloc_page(GFP_KERNEL | __GFP_ZERO);
sparc64: Move from 4MB to 8MB huge pages. The impetus for this is that we would like to move to 64-bit PMDs and PGDs, but that would result in only supporting a 42-bit address space with the current page table layout. It'd be nice to support at least 43-bits. The reason we'd end up with only 42-bits after making PMDs and PGDs 64-bit is that we only use half-page sized PTE tables in order to make PMDs line up to 4MB, the hardware huge page size we use. So what we do here is we make huge pages 8MB, and fabricate them using 4MB hw TLB entries. Facilitate this by providing a "REAL_HPAGE_SHIFT" which is used in places that really need to operate on hardware 4MB pages. Use full pages (512 entries) for PTE tables, and adjust PMD_SHIFT, PGD_SHIFT, and the build time CPP test as needed. Use a CPP test to make sure REAL_HPAGE_SHIFT and the _PAGE_SZHUGE_* we use match up. This makes the pgtable cache completely unused, so remove the code managing it and the state used in mm_context_t. Now we have less spinlocks taken in the page table allocation path. The technique we use to fabricate the 8MB pages is to transfer bit 22 from the missing virtual address into the PTEs physical address field. That takes care of the transparent huge pages case. For hugetlb, we fill things in at the PTE level and that code already puts the sub huge page physical bits into the PTEs, based upon the offset, so there is nothing special we need to do. It all just works out. So, a small amount of complexity in the THP case, but this code is about to get much simpler when we move the 64-bit PMDs as we can move away from the fancy 32-bit huge PMD encoding and just put a real PTE value in there. With bug fixes and help from Bob Picco. Signed-off-by: David S. Miller <davem@davemloft.net>
2013-09-26 00:48:49 +04:00
pte_t *pte = NULL;
if (page)
pte = (pte_t *) page_address(page);
return pte;
}
pgtable_t pte_alloc_one(struct mm_struct *mm,
unsigned long address)
{
struct page *page = alloc_page(GFP_KERNEL | __GFP_ZERO);
if (!page)
return NULL;
if (!pgtable_page_ctor(page)) {
free_hot_cold_page(page, 0);
return NULL;
}
return (pte_t *) page_address(page);
}
void pte_free_kernel(struct mm_struct *mm, pte_t *pte)
{
sparc64: Move from 4MB to 8MB huge pages. The impetus for this is that we would like to move to 64-bit PMDs and PGDs, but that would result in only supporting a 42-bit address space with the current page table layout. It'd be nice to support at least 43-bits. The reason we'd end up with only 42-bits after making PMDs and PGDs 64-bit is that we only use half-page sized PTE tables in order to make PMDs line up to 4MB, the hardware huge page size we use. So what we do here is we make huge pages 8MB, and fabricate them using 4MB hw TLB entries. Facilitate this by providing a "REAL_HPAGE_SHIFT" which is used in places that really need to operate on hardware 4MB pages. Use full pages (512 entries) for PTE tables, and adjust PMD_SHIFT, PGD_SHIFT, and the build time CPP test as needed. Use a CPP test to make sure REAL_HPAGE_SHIFT and the _PAGE_SZHUGE_* we use match up. This makes the pgtable cache completely unused, so remove the code managing it and the state used in mm_context_t. Now we have less spinlocks taken in the page table allocation path. The technique we use to fabricate the 8MB pages is to transfer bit 22 from the missing virtual address into the PTEs physical address field. That takes care of the transparent huge pages case. For hugetlb, we fill things in at the PTE level and that code already puts the sub huge page physical bits into the PTEs, based upon the offset, so there is nothing special we need to do. It all just works out. So, a small amount of complexity in the THP case, but this code is about to get much simpler when we move the 64-bit PMDs as we can move away from the fancy 32-bit huge PMD encoding and just put a real PTE value in there. With bug fixes and help from Bob Picco. Signed-off-by: David S. Miller <davem@davemloft.net>
2013-09-26 00:48:49 +04:00
free_page((unsigned long)pte);
}
static void __pte_free(pgtable_t pte)
{
struct page *page = virt_to_page(pte);
sparc64: Move from 4MB to 8MB huge pages. The impetus for this is that we would like to move to 64-bit PMDs and PGDs, but that would result in only supporting a 42-bit address space with the current page table layout. It'd be nice to support at least 43-bits. The reason we'd end up with only 42-bits after making PMDs and PGDs 64-bit is that we only use half-page sized PTE tables in order to make PMDs line up to 4MB, the hardware huge page size we use. So what we do here is we make huge pages 8MB, and fabricate them using 4MB hw TLB entries. Facilitate this by providing a "REAL_HPAGE_SHIFT" which is used in places that really need to operate on hardware 4MB pages. Use full pages (512 entries) for PTE tables, and adjust PMD_SHIFT, PGD_SHIFT, and the build time CPP test as needed. Use a CPP test to make sure REAL_HPAGE_SHIFT and the _PAGE_SZHUGE_* we use match up. This makes the pgtable cache completely unused, so remove the code managing it and the state used in mm_context_t. Now we have less spinlocks taken in the page table allocation path. The technique we use to fabricate the 8MB pages is to transfer bit 22 from the missing virtual address into the PTEs physical address field. That takes care of the transparent huge pages case. For hugetlb, we fill things in at the PTE level and that code already puts the sub huge page physical bits into the PTEs, based upon the offset, so there is nothing special we need to do. It all just works out. So, a small amount of complexity in the THP case, but this code is about to get much simpler when we move the 64-bit PMDs as we can move away from the fancy 32-bit huge PMD encoding and just put a real PTE value in there. With bug fixes and help from Bob Picco. Signed-off-by: David S. Miller <davem@davemloft.net>
2013-09-26 00:48:49 +04:00
pgtable_page_dtor(page);
__free_page(page);
}
void pte_free(struct mm_struct *mm, pgtable_t pte)
{
__pte_free(pte);
}
void pgtable_free(void *table, bool is_page)
{
if (is_page)
__pte_free(table);
else
kmem_cache_free(pgtable_cache, table);
}
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
void update_mmu_cache_pmd(struct vm_area_struct *vma, unsigned long addr,
pmd_t *pmd)
{
unsigned long pte, flags;
struct mm_struct *mm;
pmd_t entry = *pmd;
if (!pmd_large(entry) || !pmd_young(entry))
return;
pte = pmd_val(entry);
/* Don't insert a non-valid PMD into the TSB, we'll deadlock. */
if (!(pte & _PAGE_VALID))
return;
sparc64: Move from 4MB to 8MB huge pages. The impetus for this is that we would like to move to 64-bit PMDs and PGDs, but that would result in only supporting a 42-bit address space with the current page table layout. It'd be nice to support at least 43-bits. The reason we'd end up with only 42-bits after making PMDs and PGDs 64-bit is that we only use half-page sized PTE tables in order to make PMDs line up to 4MB, the hardware huge page size we use. So what we do here is we make huge pages 8MB, and fabricate them using 4MB hw TLB entries. Facilitate this by providing a "REAL_HPAGE_SHIFT" which is used in places that really need to operate on hardware 4MB pages. Use full pages (512 entries) for PTE tables, and adjust PMD_SHIFT, PGD_SHIFT, and the build time CPP test as needed. Use a CPP test to make sure REAL_HPAGE_SHIFT and the _PAGE_SZHUGE_* we use match up. This makes the pgtable cache completely unused, so remove the code managing it and the state used in mm_context_t. Now we have less spinlocks taken in the page table allocation path. The technique we use to fabricate the 8MB pages is to transfer bit 22 from the missing virtual address into the PTEs physical address field. That takes care of the transparent huge pages case. For hugetlb, we fill things in at the PTE level and that code already puts the sub huge page physical bits into the PTEs, based upon the offset, so there is nothing special we need to do. It all just works out. So, a small amount of complexity in the THP case, but this code is about to get much simpler when we move the 64-bit PMDs as we can move away from the fancy 32-bit huge PMD encoding and just put a real PTE value in there. With bug fixes and help from Bob Picco. Signed-off-by: David S. Miller <davem@davemloft.net>
2013-09-26 00:48:49 +04:00
/* We are fabricating 8MB pages using 4MB real hw pages. */
pte |= (addr & (1UL << REAL_HPAGE_SHIFT));
mm = vma->vm_mm;
spin_lock_irqsave(&mm->context.lock, flags);
if (mm->context.tsb_block[MM_TSB_HUGE].tsb != NULL)
sparc64: Move from 4MB to 8MB huge pages. The impetus for this is that we would like to move to 64-bit PMDs and PGDs, but that would result in only supporting a 42-bit address space with the current page table layout. It'd be nice to support at least 43-bits. The reason we'd end up with only 42-bits after making PMDs and PGDs 64-bit is that we only use half-page sized PTE tables in order to make PMDs line up to 4MB, the hardware huge page size we use. So what we do here is we make huge pages 8MB, and fabricate them using 4MB hw TLB entries. Facilitate this by providing a "REAL_HPAGE_SHIFT" which is used in places that really need to operate on hardware 4MB pages. Use full pages (512 entries) for PTE tables, and adjust PMD_SHIFT, PGD_SHIFT, and the build time CPP test as needed. Use a CPP test to make sure REAL_HPAGE_SHIFT and the _PAGE_SZHUGE_* we use match up. This makes the pgtable cache completely unused, so remove the code managing it and the state used in mm_context_t. Now we have less spinlocks taken in the page table allocation path. The technique we use to fabricate the 8MB pages is to transfer bit 22 from the missing virtual address into the PTEs physical address field. That takes care of the transparent huge pages case. For hugetlb, we fill things in at the PTE level and that code already puts the sub huge page physical bits into the PTEs, based upon the offset, so there is nothing special we need to do. It all just works out. So, a small amount of complexity in the THP case, but this code is about to get much simpler when we move the 64-bit PMDs as we can move away from the fancy 32-bit huge PMD encoding and just put a real PTE value in there. With bug fixes and help from Bob Picco. Signed-off-by: David S. Miller <davem@davemloft.net>
2013-09-26 00:48:49 +04:00
__update_mmu_tsb_insert(mm, MM_TSB_HUGE, REAL_HPAGE_SHIFT,
addr, pte);
spin_unlock_irqrestore(&mm->context.lock, flags);
}
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
#if defined(CONFIG_HUGETLB_PAGE) || defined(CONFIG_TRANSPARENT_HUGEPAGE)
static void context_reload(void *__data)
{
struct mm_struct *mm = __data;
if (mm == current->mm)
load_secondary_context(mm);
}
void hugetlb_setup(struct pt_regs *regs)
{
struct mm_struct *mm = current->mm;
struct tsb_config *tp;
mm/fault, arch: Use pagefault_disable() to check for disabled pagefaults in the handler Introduce faulthandler_disabled() and use it to check for irq context and disabled pagefaults (via pagefault_disable()) in the pagefault handlers. Please note that we keep the in_atomic() checks in place - to detect whether in irq context (in which case preemption is always properly disabled). In contrast, preempt_disable() should never be used to disable pagefaults. With !CONFIG_PREEMPT_COUNT, preempt_disable() doesn't modify the preempt counter, and therefore the result of in_atomic() differs. We validate that condition by using might_fault() checks when calling might_sleep(). Therefore, add a comment to faulthandler_disabled(), describing why this is needed. faulthandler_disabled() and pagefault_disable() are defined in linux/uaccess.h, so let's properly add that include to all relevant files. This patch is based on a patch from Thomas Gleixner. Reviewed-and-tested-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: David Hildenbrand <dahi@linux.vnet.ibm.com> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: David.Laight@ACULAB.COM Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: airlied@linux.ie Cc: akpm@linux-foundation.org Cc: benh@kernel.crashing.org Cc: bigeasy@linutronix.de Cc: borntraeger@de.ibm.com Cc: daniel.vetter@intel.com Cc: heiko.carstens@de.ibm.com Cc: herbert@gondor.apana.org.au Cc: hocko@suse.cz Cc: hughd@google.com Cc: mst@redhat.com Cc: paulus@samba.org Cc: ralf@linux-mips.org Cc: schwidefsky@de.ibm.com Cc: yang.shi@windriver.com Link: http://lkml.kernel.org/r/1431359540-32227-7-git-send-email-dahi@linux.vnet.ibm.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2015-05-11 18:52:11 +03:00
if (faulthandler_disabled() || !mm) {
const struct exception_table_entry *entry;
entry = search_exception_tables(regs->tpc);
if (entry) {
regs->tpc = entry->fixup;
regs->tnpc = regs->tpc + 4;
return;
}
pr_alert("Unexpected HugeTLB setup in atomic context.\n");
die_if_kernel("HugeTSB in atomic", regs);
}
tp = &mm->context.tsb_block[MM_TSB_HUGE];
if (likely(tp->tsb == NULL))
tsb_grow(mm, MM_TSB_HUGE, 0);
tsb_context_switch(mm);
smp_tsb_sync(mm);
/* On UltraSPARC-III+ and later, configure the second half of
* the Data-TLB for huge pages.
*/
if (tlb_type == cheetah_plus) {
bool need_context_reload = false;
unsigned long ctx;
spin_lock_irq(&ctx_alloc_lock);
ctx = mm->context.sparc64_ctx_val;
ctx &= ~CTX_PGSZ_MASK;
ctx |= CTX_PGSZ_BASE << CTX_PGSZ0_SHIFT;
ctx |= CTX_PGSZ_HUGE << CTX_PGSZ1_SHIFT;
if (ctx != mm->context.sparc64_ctx_val) {
/* When changing the page size fields, we
* must perform a context flush so that no
* stale entries match. This flush must
* occur with the original context register
* settings.
*/
do_flush_tlb_mm(mm);
/* Reload the context register of all processors
* also executing in this address space.
*/
mm->context.sparc64_ctx_val = ctx;
need_context_reload = true;
}
spin_unlock_irq(&ctx_alloc_lock);
if (need_context_reload)
on_each_cpu(context_reload, mm, 0);
}
}
#endif
static struct resource code_resource = {
.name = "Kernel code",
.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
};
static struct resource data_resource = {
.name = "Kernel data",
.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
};
static struct resource bss_resource = {
.name = "Kernel bss",
.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
};
static inline resource_size_t compute_kern_paddr(void *addr)
{
return (resource_size_t) (addr - KERNBASE + kern_base);
}
static void __init kernel_lds_init(void)
{
code_resource.start = compute_kern_paddr(_text);
code_resource.end = compute_kern_paddr(_etext - 1);
data_resource.start = compute_kern_paddr(_etext);
data_resource.end = compute_kern_paddr(_edata - 1);
bss_resource.start = compute_kern_paddr(__bss_start);
bss_resource.end = compute_kern_paddr(_end - 1);
}
static int __init report_memory(void)
{
int i;
struct resource *res;
kernel_lds_init();
for (i = 0; i < pavail_ents; i++) {
res = kzalloc(sizeof(struct resource), GFP_KERNEL);
if (!res) {
pr_warn("Failed to allocate source.\n");
break;
}
res->name = "System RAM";
res->start = pavail[i].phys_addr;
res->end = pavail[i].phys_addr + pavail[i].reg_size - 1;
res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
if (insert_resource(&iomem_resource, res) < 0) {
pr_warn("Resource insertion failed.\n");
break;
}
insert_resource(res, &code_resource);
insert_resource(res, &data_resource);
insert_resource(res, &bss_resource);
}
return 0;
}
arch_initcall(report_memory);
sparc64: Guard against flushing openfirmware mappings. Based almost entirely upon a patch by Christopher Alexander Tobias Schulze. In commit db64fe02258f1507e13fe5212a989922323685ce ("mm: rewrite vmap layer") lazy VMAP tlb flushing was added to the vmalloc layer. This causes problems on sparc64. Sparc64 has two VMAP mapped regions and they are not contiguous with eachother. First we have the malloc mapping area, then another unrelated region, then the vmalloc region. This "another unrelated region" is where the firmware is mapped. If the lazy TLB flushing logic in the vmalloc code triggers after we've had both a module unload and a vfree or similar, it will pass an address range that goes from somewhere inside the malloc region to somewhere inside the vmalloc region, and thus covering the openfirmware area entirely. The sparc64 kernel learns about openfirmware's dynamic mappings in this region early in the boot, and then services TLB misses in this area. But openfirmware has some locked TLB entries which are not mentioned in those dynamic mappings and we should thus not disturb them. These huge lazy TLB flush ranges causes those openfirmware locked TLB entries to be removed, resulting in all kinds of problems including hard hangs and crashes during reboot/reset. Besides causing problems like this, such huge TLB flush ranges are also incredibly inefficient. A plea has been made with the author of the VMAP lazy TLB flushing code, but for now we'll put a safety guard into our flush_tlb_kernel_range() implementation. Since the implementation has become non-trivial, stop defining it as a macro and instead make it a function in a C source file. Signed-off-by: David S. Miller <davem@davemloft.net>
2014-08-05 07:07:37 +04:00
#ifdef CONFIG_SMP
#define do_flush_tlb_kernel_range smp_flush_tlb_kernel_range
#else
#define do_flush_tlb_kernel_range __flush_tlb_kernel_range
#endif
void flush_tlb_kernel_range(unsigned long start, unsigned long end)
{
if (start < HI_OBP_ADDRESS && end > LOW_OBP_ADDRESS) {
if (start < LOW_OBP_ADDRESS) {
flush_tsb_kernel_range(start, LOW_OBP_ADDRESS);
do_flush_tlb_kernel_range(start, LOW_OBP_ADDRESS);
}
if (end > HI_OBP_ADDRESS) {
sparc64: Fix reversed start/end in flush_tlb_kernel_range() When we have to split up a flush request into multiple pieces (in order to avoid the firmware range) we don't specify the arguments in the right order for the second piece. Fix the order, or else we get hangs as the code tries to flush "a lot" of entries and we get lockups like this: [ 4422.981276] NMI watchdog: BUG: soft lockup - CPU#12 stuck for 23s! [expect:117032] [ 4422.996130] Modules linked in: ipv6 loop usb_storage igb ptp sg sr_mod ehci_pci ehci_hcd pps_core n2_rng rng_core [ 4423.016617] CPU: 12 PID: 117032 Comm: expect Not tainted 3.17.0-rc4+ #1608 [ 4423.030331] task: fff8003cc730e220 ti: fff8003d99d54000 task.ti: fff8003d99d54000 [ 4423.045282] TSTATE: 0000000011001602 TPC: 00000000004521e8 TNPC: 00000000004521ec Y: 00000000 Not tainted [ 4423.064905] TPC: <__flush_tlb_kernel_range+0x28/0x40> [ 4423.074964] g0: 000000000052fd10 g1: 00000001295a8000 g2: ffffff7176ffc000 g3: 0000000000002000 [ 4423.092324] g4: fff8003cc730e220 g5: fff8003dfedcc000 g6: fff8003d99d54000 g7: 0000000000000006 [ 4423.109687] o0: 0000000000000000 o1: 0000000000000000 o2: 0000000000000003 o3: 00000000f0000000 [ 4423.127058] o4: 0000000000000080 o5: 00000001295a8000 sp: fff8003d99d56d01 ret_pc: 000000000052ff54 [ 4423.145121] RPC: <__purge_vmap_area_lazy+0x314/0x3a0> [ 4423.155185] l0: 0000000000000000 l1: 0000000000000000 l2: 0000000000a38040 l3: 0000000000000000 [ 4423.172559] l4: fff8003dae8965e0 l5: ffffffffffffffff l6: 0000000000000000 l7: 00000000f7e2b138 [ 4423.189913] i0: fff8003d99d576a0 i1: fff8003d99d576a8 i2: fff8003d99d575e8 i3: 0000000000000000 [ 4423.207284] i4: 0000000000008008 i5: fff8003d99d575c8 i6: fff8003d99d56df1 i7: 0000000000530c24 [ 4423.224640] I7: <free_vmap_area_noflush+0x64/0x80> [ 4423.234193] Call Trace: [ 4423.239051] [0000000000530c24] free_vmap_area_noflush+0x64/0x80 [ 4423.251029] [0000000000531a7c] remove_vm_area+0x5c/0x80 [ 4423.261628] [0000000000531b80] __vunmap+0x20/0x120 [ 4423.271352] [000000000071cf18] n_tty_close+0x18/0x40 [ 4423.281423] [00000000007222b0] tty_ldisc_close+0x30/0x60 [ 4423.292183] [00000000007225a4] tty_ldisc_reinit+0x24/0xa0 [ 4423.303120] [0000000000722ab4] tty_ldisc_hangup+0xd4/0x1e0 [ 4423.314232] [0000000000719aa0] __tty_hangup+0x280/0x3c0 [ 4423.324835] [0000000000724cb4] pty_close+0x134/0x1a0 [ 4423.334905] [000000000071aa24] tty_release+0x104/0x500 [ 4423.345316] [00000000005511d0] __fput+0x90/0x1e0 [ 4423.354701] [000000000047fa54] task_work_run+0x94/0xe0 [ 4423.365126] [0000000000404b44] __handle_signal+0xc/0x2c Fixes: 4ca9a23765da ("sparc64: Guard against flushing openfirmware mappings.") Signed-off-by: David S. Miller <davem@davemloft.net>
2014-10-05 08:05:14 +04:00
flush_tsb_kernel_range(HI_OBP_ADDRESS, end);
do_flush_tlb_kernel_range(HI_OBP_ADDRESS, end);
sparc64: Guard against flushing openfirmware mappings. Based almost entirely upon a patch by Christopher Alexander Tobias Schulze. In commit db64fe02258f1507e13fe5212a989922323685ce ("mm: rewrite vmap layer") lazy VMAP tlb flushing was added to the vmalloc layer. This causes problems on sparc64. Sparc64 has two VMAP mapped regions and they are not contiguous with eachother. First we have the malloc mapping area, then another unrelated region, then the vmalloc region. This "another unrelated region" is where the firmware is mapped. If the lazy TLB flushing logic in the vmalloc code triggers after we've had both a module unload and a vfree or similar, it will pass an address range that goes from somewhere inside the malloc region to somewhere inside the vmalloc region, and thus covering the openfirmware area entirely. The sparc64 kernel learns about openfirmware's dynamic mappings in this region early in the boot, and then services TLB misses in this area. But openfirmware has some locked TLB entries which are not mentioned in those dynamic mappings and we should thus not disturb them. These huge lazy TLB flush ranges causes those openfirmware locked TLB entries to be removed, resulting in all kinds of problems including hard hangs and crashes during reboot/reset. Besides causing problems like this, such huge TLB flush ranges are also incredibly inefficient. A plea has been made with the author of the VMAP lazy TLB flushing code, but for now we'll put a safety guard into our flush_tlb_kernel_range() implementation. Since the implementation has become non-trivial, stop defining it as a macro and instead make it a function in a C source file. Signed-off-by: David S. Miller <davem@davemloft.net>
2014-08-05 07:07:37 +04:00
}
} else {
flush_tsb_kernel_range(start, end);
do_flush_tlb_kernel_range(start, end);
}
}