1
linux/arch/sparc/mm/sun4c.c

2167 lines
60 KiB
C
Raw Normal View History

/* sun4c.c: Doing in software what should be done in hardware.
*
* Copyright (C) 1996 David S. Miller (davem@davemloft.net)
* Copyright (C) 1996 Eddie C. Dost (ecd@skynet.be)
* Copyright (C) 1996 Andrew Tridgell (Andrew.Tridgell@anu.edu.au)
* Copyright (C) 1997-2000 Anton Blanchard (anton@samba.org)
* Copyright (C) 1998 Jakub Jelinek (jj@sunsite.mff.cuni.cz)
*/
#define NR_TASK_BUCKETS 512
#include <linux/kernel.h>
#include <linux/mm.h>
#include <linux/init.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 01:04:11 -07:00
#include <linux/slab.h>
#include <linux/bootmem.h>
#include <linux/highmem.h>
#include <linux/fs.h>
#include <linux/seq_file.h>
#include <linux/scatterlist.h>
#include <linux/bitmap.h>
#include <asm/sections.h>
#include <asm/page.h>
#include <asm/pgalloc.h>
#include <asm/pgtable.h>
#include <asm/vaddrs.h>
#include <asm/idprom.h>
#include <asm/machines.h>
#include <asm/memreg.h>
#include <asm/processor.h>
#include <asm/auxio.h>
#include <asm/io.h>
#include <asm/oplib.h>
#include <asm/openprom.h>
#include <asm/mmu_context.h>
#include <asm/highmem.h>
#include <asm/btfixup.h>
#include <asm/cacheflush.h>
#include <asm/tlbflush.h>
/* Because of our dynamic kernel TLB miss strategy, and how
* our DVMA mapping allocation works, you _MUST_:
*
* 1) Disable interrupts _and_ not touch any dynamic kernel
* memory while messing with kernel MMU state. By
* dynamic memory I mean any object which is not in
* the kernel image itself or a thread_union (both of
* which are locked into the MMU).
* 2) Disable interrupts while messing with user MMU state.
*/
extern int num_segmaps, num_contexts;
extern unsigned long page_kernel;
/* That's it, we prom_halt() on sun4c if the cache size is something other than 65536.
* So let's save some cycles and just use that everywhere except for that bootup
* sanity check.
*/
#define SUN4C_VAC_SIZE 65536
#define SUN4C_KERNEL_BUCKETS 32
/* Flushing the cache. */
struct sun4c_vac_props sun4c_vacinfo;
unsigned long sun4c_kernel_faults;
/* Invalidate every sun4c cache line tag. */
static void __init sun4c_flush_all(void)
{
unsigned long begin, end;
if (sun4c_vacinfo.on)
panic("SUN4C: AIEEE, trying to invalidate vac while it is on.");
/* Clear 'valid' bit in all cache line tags */
begin = AC_CACHETAGS;
end = (AC_CACHETAGS + SUN4C_VAC_SIZE);
while (begin < end) {
__asm__ __volatile__("sta %%g0, [%0] %1\n\t" : :
"r" (begin), "i" (ASI_CONTROL));
begin += sun4c_vacinfo.linesize;
}
}
static void sun4c_flush_context_hw(void)
{
unsigned long end = SUN4C_VAC_SIZE;
__asm__ __volatile__(
"1: addcc %0, -4096, %0\n\t"
" bne 1b\n\t"
" sta %%g0, [%0] %2"
: "=&r" (end)
: "0" (end), "i" (ASI_HWFLUSHCONTEXT)
: "cc");
}
/* Must be called minimally with IRQs disabled. */
static void sun4c_flush_segment_hw(unsigned long addr)
{
if (sun4c_get_segmap(addr) != invalid_segment) {
unsigned long vac_size = SUN4C_VAC_SIZE;
__asm__ __volatile__(
"1: addcc %0, -4096, %0\n\t"
" bne 1b\n\t"
" sta %%g0, [%2 + %0] %3"
: "=&r" (vac_size)
: "0" (vac_size), "r" (addr), "i" (ASI_HWFLUSHSEG)
: "cc");
}
}
/* File local boot time fixups. */
BTFIXUPDEF_CALL(void, sun4c_flush_page, unsigned long)
BTFIXUPDEF_CALL(void, sun4c_flush_segment, unsigned long)
BTFIXUPDEF_CALL(void, sun4c_flush_context, void)
#define sun4c_flush_page(addr) BTFIXUP_CALL(sun4c_flush_page)(addr)
#define sun4c_flush_segment(addr) BTFIXUP_CALL(sun4c_flush_segment)(addr)
#define sun4c_flush_context() BTFIXUP_CALL(sun4c_flush_context)()
/* Must be called minimally with interrupts disabled. */
static void sun4c_flush_page_hw(unsigned long addr)
{
addr &= PAGE_MASK;
if ((int)sun4c_get_pte(addr) < 0)
__asm__ __volatile__("sta %%g0, [%0] %1"
: : "r" (addr), "i" (ASI_HWFLUSHPAGE));
}
/* Don't inline the software version as it eats too many cache lines if expanded. */
static void sun4c_flush_context_sw(void)
{
unsigned long nbytes = SUN4C_VAC_SIZE;
unsigned long lsize = sun4c_vacinfo.linesize;
__asm__ __volatile__(
"add %2, %2, %%g1\n\t"
"add %2, %%g1, %%g2\n\t"
"add %2, %%g2, %%g3\n\t"
"add %2, %%g3, %%g4\n\t"
"add %2, %%g4, %%g5\n\t"
"add %2, %%g5, %%o4\n\t"
"add %2, %%o4, %%o5\n"
"1:\n\t"
"subcc %0, %%o5, %0\n\t"
"sta %%g0, [%0] %3\n\t"
"sta %%g0, [%0 + %2] %3\n\t"
"sta %%g0, [%0 + %%g1] %3\n\t"
"sta %%g0, [%0 + %%g2] %3\n\t"
"sta %%g0, [%0 + %%g3] %3\n\t"
"sta %%g0, [%0 + %%g4] %3\n\t"
"sta %%g0, [%0 + %%g5] %3\n\t"
"bg 1b\n\t"
" sta %%g0, [%1 + %%o4] %3\n"
: "=&r" (nbytes)
: "0" (nbytes), "r" (lsize), "i" (ASI_FLUSHCTX)
: "g1", "g2", "g3", "g4", "g5", "o4", "o5", "cc");
}
/* Don't inline the software version as it eats too many cache lines if expanded. */
static void sun4c_flush_segment_sw(unsigned long addr)
{
if (sun4c_get_segmap(addr) != invalid_segment) {
unsigned long nbytes = SUN4C_VAC_SIZE;
unsigned long lsize = sun4c_vacinfo.linesize;
__asm__ __volatile__(
"add %2, %2, %%g1\n\t"
"add %2, %%g1, %%g2\n\t"
"add %2, %%g2, %%g3\n\t"
"add %2, %%g3, %%g4\n\t"
"add %2, %%g4, %%g5\n\t"
"add %2, %%g5, %%o4\n\t"
"add %2, %%o4, %%o5\n"
"1:\n\t"
"subcc %1, %%o5, %1\n\t"
"sta %%g0, [%0] %6\n\t"
"sta %%g0, [%0 + %2] %6\n\t"
"sta %%g0, [%0 + %%g1] %6\n\t"
"sta %%g0, [%0 + %%g2] %6\n\t"
"sta %%g0, [%0 + %%g3] %6\n\t"
"sta %%g0, [%0 + %%g4] %6\n\t"
"sta %%g0, [%0 + %%g5] %6\n\t"
"sta %%g0, [%0 + %%o4] %6\n\t"
"bg 1b\n\t"
" add %0, %%o5, %0\n"
: "=&r" (addr), "=&r" (nbytes), "=&r" (lsize)
: "0" (addr), "1" (nbytes), "2" (lsize),
"i" (ASI_FLUSHSEG)
: "g1", "g2", "g3", "g4", "g5", "o4", "o5", "cc");
}
}
/* Don't inline the software version as it eats too many cache lines if expanded. */
static void sun4c_flush_page_sw(unsigned long addr)
{
addr &= PAGE_MASK;
if ((sun4c_get_pte(addr) & (_SUN4C_PAGE_NOCACHE | _SUN4C_PAGE_VALID)) ==
_SUN4C_PAGE_VALID) {
unsigned long left = PAGE_SIZE;
unsigned long lsize = sun4c_vacinfo.linesize;
__asm__ __volatile__(
"add %2, %2, %%g1\n\t"
"add %2, %%g1, %%g2\n\t"
"add %2, %%g2, %%g3\n\t"
"add %2, %%g3, %%g4\n\t"
"add %2, %%g4, %%g5\n\t"
"add %2, %%g5, %%o4\n\t"
"add %2, %%o4, %%o5\n"
"1:\n\t"
"subcc %1, %%o5, %1\n\t"
"sta %%g0, [%0] %6\n\t"
"sta %%g0, [%0 + %2] %6\n\t"
"sta %%g0, [%0 + %%g1] %6\n\t"
"sta %%g0, [%0 + %%g2] %6\n\t"
"sta %%g0, [%0 + %%g3] %6\n\t"
"sta %%g0, [%0 + %%g4] %6\n\t"
"sta %%g0, [%0 + %%g5] %6\n\t"
"sta %%g0, [%0 + %%o4] %6\n\t"
"bg 1b\n\t"
" add %0, %%o5, %0\n"
: "=&r" (addr), "=&r" (left), "=&r" (lsize)
: "0" (addr), "1" (left), "2" (lsize),
"i" (ASI_FLUSHPG)
: "g1", "g2", "g3", "g4", "g5", "o4", "o5", "cc");
}
}
/* The sun4c's do have an on chip store buffer. And the way you
* clear them out isn't so obvious. The only way I can think of
* to accomplish this is to read the current context register,
* store the same value there, then read an external hardware
* register.
*/
void sun4c_complete_all_stores(void)
{
volatile int _unused;
_unused = sun4c_get_context();
sun4c_set_context(_unused);
_unused = get_auxio();
}
/* Bootup utility functions. */
static inline void sun4c_init_clean_segmap(unsigned char pseg)
{
unsigned long vaddr;
sun4c_put_segmap(0, pseg);
for (vaddr = 0; vaddr < SUN4C_REAL_PGDIR_SIZE; vaddr += PAGE_SIZE)
sun4c_put_pte(vaddr, 0);
sun4c_put_segmap(0, invalid_segment);
}
static inline void sun4c_init_clean_mmu(unsigned long kernel_end)
{
unsigned long vaddr;
unsigned char savectx, ctx;
savectx = sun4c_get_context();
for (ctx = 0; ctx < num_contexts; ctx++) {
sun4c_set_context(ctx);
for (vaddr = 0; vaddr < 0x20000000; vaddr += SUN4C_REAL_PGDIR_SIZE)
sun4c_put_segmap(vaddr, invalid_segment);
for (vaddr = 0xe0000000; vaddr < KERNBASE; vaddr += SUN4C_REAL_PGDIR_SIZE)
sun4c_put_segmap(vaddr, invalid_segment);
for (vaddr = kernel_end; vaddr < KADB_DEBUGGER_BEGVM; vaddr += SUN4C_REAL_PGDIR_SIZE)
sun4c_put_segmap(vaddr, invalid_segment);
for (vaddr = LINUX_OPPROM_ENDVM; vaddr; vaddr += SUN4C_REAL_PGDIR_SIZE)
sun4c_put_segmap(vaddr, invalid_segment);
}
sun4c_set_context(savectx);
}
void __init sun4c_probe_vac(void)
{
sun4c_disable_vac();
if ((idprom->id_machtype == (SM_SUN4C | SM_4C_SS1)) ||
(idprom->id_machtype == (SM_SUN4C | SM_4C_SS1PLUS))) {
/* PROM on SS1 lacks this info, to be super safe we
* hard code it here since this arch is cast in stone.
*/
sun4c_vacinfo.num_bytes = 65536;
sun4c_vacinfo.linesize = 16;
} else {
sun4c_vacinfo.num_bytes =
prom_getintdefault(prom_root_node, "vac-size", 65536);
sun4c_vacinfo.linesize =
prom_getintdefault(prom_root_node, "vac-linesize", 16);
}
sun4c_vacinfo.do_hwflushes =
prom_getintdefault(prom_root_node, "vac-hwflush", 0);
if (sun4c_vacinfo.do_hwflushes == 0)
sun4c_vacinfo.do_hwflushes =
prom_getintdefault(prom_root_node, "vac_hwflush", 0);
if (sun4c_vacinfo.num_bytes != 65536) {
prom_printf("WEIRD Sun4C VAC cache size, "
"tell sparclinux@vger.kernel.org");
prom_halt();
}
switch (sun4c_vacinfo.linesize) {
case 16:
sun4c_vacinfo.log2lsize = 4;
break;
case 32:
sun4c_vacinfo.log2lsize = 5;
break;
default:
prom_printf("probe_vac: Didn't expect vac-linesize of %d, halting\n",
sun4c_vacinfo.linesize);
prom_halt();
}
sun4c_flush_all();
sun4c_enable_vac();
}
/* Patch instructions for the low level kernel fault handler. */
extern unsigned long invalid_segment_patch1, invalid_segment_patch1_ff;
extern unsigned long invalid_segment_patch2, invalid_segment_patch2_ff;
extern unsigned long invalid_segment_patch1_1ff, invalid_segment_patch2_1ff;
extern unsigned long num_context_patch1, num_context_patch1_16;
extern unsigned long num_context_patch2_16;
extern unsigned long vac_linesize_patch, vac_linesize_patch_32;
extern unsigned long vac_hwflush_patch1, vac_hwflush_patch1_on;
extern unsigned long vac_hwflush_patch2, vac_hwflush_patch2_on;
#define PATCH_INSN(src, dst) do { \
daddr = &(dst); \
iaddr = &(src); \
*daddr = *iaddr; \
} while (0)
static void __init patch_kernel_fault_handler(void)
{
unsigned long *iaddr, *daddr;
switch (num_segmaps) {
case 128:
/* Default, nothing to do. */
break;
case 256:
PATCH_INSN(invalid_segment_patch1_ff,
invalid_segment_patch1);
PATCH_INSN(invalid_segment_patch2_ff,
invalid_segment_patch2);
break;
case 512:
PATCH_INSN(invalid_segment_patch1_1ff,
invalid_segment_patch1);
PATCH_INSN(invalid_segment_patch2_1ff,
invalid_segment_patch2);
break;
default:
prom_printf("Unhandled number of segmaps: %d\n",
num_segmaps);
prom_halt();
}
switch (num_contexts) {
case 8:
/* Default, nothing to do. */
break;
case 16:
PATCH_INSN(num_context_patch1_16,
num_context_patch1);
break;
default:
prom_printf("Unhandled number of contexts: %d\n",
num_contexts);
prom_halt();
}
if (sun4c_vacinfo.do_hwflushes != 0) {
PATCH_INSN(vac_hwflush_patch1_on, vac_hwflush_patch1);
PATCH_INSN(vac_hwflush_patch2_on, vac_hwflush_patch2);
} else {
switch (sun4c_vacinfo.linesize) {
case 16:
/* Default, nothing to do. */
break;
case 32:
PATCH_INSN(vac_linesize_patch_32, vac_linesize_patch);
break;
default:
prom_printf("Impossible VAC linesize %d, halting...\n",
sun4c_vacinfo.linesize);
prom_halt();
}
}
}
static void __init sun4c_probe_mmu(void)
{
if ((idprom->id_machtype == (SM_SUN4C | SM_4C_SS1)) ||
(idprom->id_machtype == (SM_SUN4C | SM_4C_SS1PLUS))) {
/* Hardcode these just to be safe, PROM on SS1 does
* not have this info available in the root node.
*/
num_segmaps = 128;
num_contexts = 8;
} else {
num_segmaps =
prom_getintdefault(prom_root_node, "mmu-npmg", 128);
num_contexts =
prom_getintdefault(prom_root_node, "mmu-nctx", 0x8);
}
patch_kernel_fault_handler();
}
volatile unsigned long __iomem *sun4c_memerr_reg = NULL;
void __init sun4c_probe_memerr_reg(void)
{
phandle node;
struct linux_prom_registers regs[1];
node = prom_getchild(prom_root_node);
node = prom_searchsiblings(prom_root_node, "memory-error");
if (!node)
return;
if (prom_getproperty(node, "reg", (char *)regs, sizeof(regs)) <= 0)
return;
/* hmm I think regs[0].which_io is zero here anyways */
sun4c_memerr_reg = ioremap(regs[0].phys_addr, regs[0].reg_size);
}
static inline void sun4c_init_ss2_cache_bug(void)
{
if ((idprom->id_machtype == (SM_SUN4C | SM_4C_SS2)) ||
(idprom->id_machtype == (SM_SUN4C | SM_4C_IPX)) ||
(idprom->id_machtype == (SM_SUN4C | SM_4C_ELC))) {
/* Whee.. */
printk("SS2 cache bug detected, uncaching trap table page\n");
sun4c_flush_page((unsigned int) &_start);
sun4c_put_pte(((unsigned long) &_start),
(sun4c_get_pte((unsigned long) &_start) | _SUN4C_PAGE_NOCACHE));
}
}
/* Addr is always aligned on a page boundary for us already. */
static int sun4c_map_dma_area(struct device *dev, dma_addr_t *pba, unsigned long va,
unsigned long addr, int len)
{
unsigned long page, end;
*pba = addr;
end = PAGE_ALIGN((addr + len));
while (addr < end) {
page = va;
sun4c_flush_page(page);
page -= PAGE_OFFSET;
page >>= PAGE_SHIFT;
page |= (_SUN4C_PAGE_VALID | _SUN4C_PAGE_DIRTY |
_SUN4C_PAGE_NOCACHE | _SUN4C_PAGE_PRIV);
sun4c_put_pte(addr, page);
addr += PAGE_SIZE;
va += PAGE_SIZE;
}
return 0;
}
static void sun4c_unmap_dma_area(struct device *dev, unsigned long busa, int len)
{
/* Fortunately for us, bus_addr == uncached_virt in sun4c. */
/* XXX Implement this */
}
/* TLB management. */
/* Don't change this struct without changing entry.S. This is used
* in the in-window kernel fault handler, and you don't want to mess
* with that. (See sun4c_fault in entry.S).
*/
struct sun4c_mmu_entry {
struct sun4c_mmu_entry *next;
struct sun4c_mmu_entry *prev;
unsigned long vaddr;
unsigned char pseg;
unsigned char locked;
/* For user mappings only, and completely hidden from kernel
* TLB miss code.
*/
unsigned char ctx;
struct sun4c_mmu_entry *lru_next;
struct sun4c_mmu_entry *lru_prev;
};
static struct sun4c_mmu_entry mmu_entry_pool[SUN4C_MAX_SEGMAPS];
static void __init sun4c_init_mmu_entry_pool(void)
{
int i;
for (i=0; i < SUN4C_MAX_SEGMAPS; i++) {
mmu_entry_pool[i].pseg = i;
mmu_entry_pool[i].next = NULL;
mmu_entry_pool[i].prev = NULL;
mmu_entry_pool[i].vaddr = 0;
mmu_entry_pool[i].locked = 0;
mmu_entry_pool[i].ctx = 0;
mmu_entry_pool[i].lru_next = NULL;
mmu_entry_pool[i].lru_prev = NULL;
}
mmu_entry_pool[invalid_segment].locked = 1;
}
static inline void fix_permissions(unsigned long vaddr, unsigned long bits_on,
unsigned long bits_off)
{
unsigned long start, end;
end = vaddr + SUN4C_REAL_PGDIR_SIZE;
for (start = vaddr; start < end; start += PAGE_SIZE)
if (sun4c_get_pte(start) & _SUN4C_PAGE_VALID)
sun4c_put_pte(start, (sun4c_get_pte(start) | bits_on) &
~bits_off);
}
static inline void sun4c_init_map_kernelprom(unsigned long kernel_end)
{
unsigned long vaddr;
unsigned char pseg, ctx;
for (vaddr = KADB_DEBUGGER_BEGVM;
vaddr < LINUX_OPPROM_ENDVM;
vaddr += SUN4C_REAL_PGDIR_SIZE) {
pseg = sun4c_get_segmap(vaddr);
if (pseg != invalid_segment) {
mmu_entry_pool[pseg].locked = 1;
for (ctx = 0; ctx < num_contexts; ctx++)
prom_putsegment(ctx, vaddr, pseg);
fix_permissions(vaddr, _SUN4C_PAGE_PRIV, 0);
}
}
for (vaddr = KERNBASE; vaddr < kernel_end; vaddr += SUN4C_REAL_PGDIR_SIZE) {
pseg = sun4c_get_segmap(vaddr);
mmu_entry_pool[pseg].locked = 1;
for (ctx = 0; ctx < num_contexts; ctx++)
prom_putsegment(ctx, vaddr, pseg);
fix_permissions(vaddr, _SUN4C_PAGE_PRIV, _SUN4C_PAGE_NOCACHE);
}
}
static void __init sun4c_init_lock_area(unsigned long start, unsigned long end)
{
int i, ctx;
while (start < end) {
for (i = 0; i < invalid_segment; i++)
if (!mmu_entry_pool[i].locked)
break;
mmu_entry_pool[i].locked = 1;
sun4c_init_clean_segmap(i);
for (ctx = 0; ctx < num_contexts; ctx++)
prom_putsegment(ctx, start, mmu_entry_pool[i].pseg);
start += SUN4C_REAL_PGDIR_SIZE;
}
}
/* Don't change this struct without changing entry.S. This is used
* in the in-window kernel fault handler, and you don't want to mess
* with that. (See sun4c_fault in entry.S).
*/
struct sun4c_mmu_ring {
struct sun4c_mmu_entry ringhd;
int num_entries;
};
static struct sun4c_mmu_ring sun4c_context_ring[SUN4C_MAX_CONTEXTS]; /* used user entries */
static struct sun4c_mmu_ring sun4c_ufree_ring; /* free user entries */
static struct sun4c_mmu_ring sun4c_ulru_ring; /* LRU user entries */
struct sun4c_mmu_ring sun4c_kernel_ring; /* used kernel entries */
struct sun4c_mmu_ring sun4c_kfree_ring; /* free kernel entries */
static inline void sun4c_init_rings(void)
{
int i;
for (i = 0; i < SUN4C_MAX_CONTEXTS; i++) {
sun4c_context_ring[i].ringhd.next =
sun4c_context_ring[i].ringhd.prev =
&sun4c_context_ring[i].ringhd;
sun4c_context_ring[i].num_entries = 0;
}
sun4c_ufree_ring.ringhd.next = sun4c_ufree_ring.ringhd.prev =
&sun4c_ufree_ring.ringhd;
sun4c_ufree_ring.num_entries = 0;
sun4c_ulru_ring.ringhd.lru_next = sun4c_ulru_ring.ringhd.lru_prev =
&sun4c_ulru_ring.ringhd;
sun4c_ulru_ring.num_entries = 0;
sun4c_kernel_ring.ringhd.next = sun4c_kernel_ring.ringhd.prev =
&sun4c_kernel_ring.ringhd;
sun4c_kernel_ring.num_entries = 0;
sun4c_kfree_ring.ringhd.next = sun4c_kfree_ring.ringhd.prev =
&sun4c_kfree_ring.ringhd;
sun4c_kfree_ring.num_entries = 0;
}
static void add_ring(struct sun4c_mmu_ring *ring,
struct sun4c_mmu_entry *entry)
{
struct sun4c_mmu_entry *head = &ring->ringhd;
entry->prev = head;
(entry->next = head->next)->prev = entry;
head->next = entry;
ring->num_entries++;
}
static inline void add_lru(struct sun4c_mmu_entry *entry)
{
struct sun4c_mmu_ring *ring = &sun4c_ulru_ring;
struct sun4c_mmu_entry *head = &ring->ringhd;
entry->lru_next = head;
(entry->lru_prev = head->lru_prev)->lru_next = entry;
head->lru_prev = entry;
}
static void add_ring_ordered(struct sun4c_mmu_ring *ring,
struct sun4c_mmu_entry *entry)
{
struct sun4c_mmu_entry *head = &ring->ringhd;
unsigned long addr = entry->vaddr;
while ((head->next != &ring->ringhd) && (head->next->vaddr < addr))
head = head->next;
entry->prev = head;
(entry->next = head->next)->prev = entry;
head->next = entry;
ring->num_entries++;
add_lru(entry);
}
static inline void remove_ring(struct sun4c_mmu_ring *ring,
struct sun4c_mmu_entry *entry)
{
struct sun4c_mmu_entry *next = entry->next;
(next->prev = entry->prev)->next = next;
ring->num_entries--;
}
static void remove_lru(struct sun4c_mmu_entry *entry)
{
struct sun4c_mmu_entry *next = entry->lru_next;
(next->lru_prev = entry->lru_prev)->lru_next = next;
}
static void free_user_entry(int ctx, struct sun4c_mmu_entry *entry)
{
remove_ring(sun4c_context_ring+ctx, entry);
remove_lru(entry);
add_ring(&sun4c_ufree_ring, entry);
}
static void free_kernel_entry(struct sun4c_mmu_entry *entry,
struct sun4c_mmu_ring *ring)
{
remove_ring(ring, entry);
add_ring(&sun4c_kfree_ring, entry);
}
static void __init sun4c_init_fill_kernel_ring(int howmany)
{
int i;
while (howmany) {
for (i = 0; i < invalid_segment; i++)
if (!mmu_entry_pool[i].locked)
break;
mmu_entry_pool[i].locked = 1;
sun4c_init_clean_segmap(i);
add_ring(&sun4c_kfree_ring, &mmu_entry_pool[i]);
howmany--;
}
}
static void __init sun4c_init_fill_user_ring(void)
{
int i;
for (i = 0; i < invalid_segment; i++) {
if (mmu_entry_pool[i].locked)
continue;
sun4c_init_clean_segmap(i);
add_ring(&sun4c_ufree_ring, &mmu_entry_pool[i]);
}
}
static void sun4c_kernel_unmap(struct sun4c_mmu_entry *kentry)
{
int savectx, ctx;
savectx = sun4c_get_context();
for (ctx = 0; ctx < num_contexts; ctx++) {
sun4c_set_context(ctx);
sun4c_put_segmap(kentry->vaddr, invalid_segment);
}
sun4c_set_context(savectx);
}
static void sun4c_kernel_map(struct sun4c_mmu_entry *kentry)
{
int savectx, ctx;
savectx = sun4c_get_context();
for (ctx = 0; ctx < num_contexts; ctx++) {
sun4c_set_context(ctx);
sun4c_put_segmap(kentry->vaddr, kentry->pseg);
}
sun4c_set_context(savectx);
}
#define sun4c_user_unmap(__entry) \
sun4c_put_segmap((__entry)->vaddr, invalid_segment)
static void sun4c_demap_context(struct sun4c_mmu_ring *crp, unsigned char ctx)
{
struct sun4c_mmu_entry *head = &crp->ringhd;
unsigned long flags;
local_irq_save(flags);
if (head->next != head) {
struct sun4c_mmu_entry *entry = head->next;
int savectx = sun4c_get_context();
flush_user_windows();
sun4c_set_context(ctx);
sun4c_flush_context();
do {
struct sun4c_mmu_entry *next = entry->next;
sun4c_user_unmap(entry);
free_user_entry(ctx, entry);
entry = next;
} while (entry != head);
sun4c_set_context(savectx);
}
local_irq_restore(flags);
}
static int sun4c_user_taken_entries; /* This is how much we have. */
static int max_user_taken_entries; /* This limits us and prevents deadlock. */
static struct sun4c_mmu_entry *sun4c_kernel_strategy(void)
{
struct sun4c_mmu_entry *this_entry;
/* If some are free, return first one. */
if (sun4c_kfree_ring.num_entries) {
this_entry = sun4c_kfree_ring.ringhd.next;
return this_entry;
}
/* Else free one up. */
this_entry = sun4c_kernel_ring.ringhd.prev;
sun4c_flush_segment(this_entry->vaddr);
sun4c_kernel_unmap(this_entry);
free_kernel_entry(this_entry, &sun4c_kernel_ring);
this_entry = sun4c_kfree_ring.ringhd.next;
return this_entry;
}
/* Using this method to free up mmu entries eliminates a lot of
* potential races since we have a kernel that incurs tlb
* replacement faults. There may be performance penalties.
*
* NOTE: Must be called with interrupts disabled.
*/
static struct sun4c_mmu_entry *sun4c_user_strategy(void)
{
struct sun4c_mmu_entry *entry;
unsigned char ctx;
int savectx;
/* If some are free, return first one. */
if (sun4c_ufree_ring.num_entries) {
entry = sun4c_ufree_ring.ringhd.next;
goto unlink_out;
}
if (sun4c_user_taken_entries) {
entry = sun4c_kernel_strategy();
sun4c_user_taken_entries--;
goto kunlink_out;
}
/* Grab from the beginning of the LRU list. */
entry = sun4c_ulru_ring.ringhd.lru_next;
ctx = entry->ctx;
savectx = sun4c_get_context();
flush_user_windows();
sun4c_set_context(ctx);
sun4c_flush_segment(entry->vaddr);
sun4c_user_unmap(entry);
remove_ring(sun4c_context_ring + ctx, entry);
remove_lru(entry);
sun4c_set_context(savectx);
return entry;
unlink_out:
remove_ring(&sun4c_ufree_ring, entry);
return entry;
kunlink_out:
remove_ring(&sun4c_kfree_ring, entry);
return entry;
}
/* NOTE: Must be called with interrupts disabled. */
void sun4c_grow_kernel_ring(void)
{
struct sun4c_mmu_entry *entry;
/* Prevent deadlock condition. */
if (sun4c_user_taken_entries >= max_user_taken_entries)
return;
if (sun4c_ufree_ring.num_entries) {
entry = sun4c_ufree_ring.ringhd.next;
remove_ring(&sun4c_ufree_ring, entry);
add_ring(&sun4c_kfree_ring, entry);
sun4c_user_taken_entries++;
}
}
/* 2 page buckets for task struct and kernel stack allocation.
*
* TASK_STACK_BEGIN
* bucket[0]
* bucket[1]
* [ ... ]
* bucket[NR_TASK_BUCKETS-1]
* TASK_STACK_BEGIN + (sizeof(struct task_bucket) * NR_TASK_BUCKETS)
*
* Each slot looks like:
*
* page 1 -- task struct + beginning of kernel stack
* page 2 -- rest of kernel stack
*/
union task_union *sun4c_bucket[NR_TASK_BUCKETS];
static int sun4c_lowbucket_avail;
#define BUCKET_EMPTY ((union task_union *) 0)
#define BUCKET_SHIFT (PAGE_SHIFT + 1) /* log2(sizeof(struct task_bucket)) */
#define BUCKET_SIZE (1 << BUCKET_SHIFT)
#define BUCKET_NUM(addr) ((((addr) - SUN4C_LOCK_VADDR) >> BUCKET_SHIFT))
#define BUCKET_ADDR(num) (((num) << BUCKET_SHIFT) + SUN4C_LOCK_VADDR)
#define BUCKET_PTE(page) \
((((page) - PAGE_OFFSET) >> PAGE_SHIFT) | pgprot_val(SUN4C_PAGE_KERNEL))
#define BUCKET_PTE_PAGE(pte) \
(PAGE_OFFSET + (((pte) & SUN4C_PFN_MASK) << PAGE_SHIFT))
static void get_locked_segment(unsigned long addr)
{
struct sun4c_mmu_entry *stolen;
unsigned long flags;
local_irq_save(flags);
addr &= SUN4C_REAL_PGDIR_MASK;
stolen = sun4c_user_strategy();
max_user_taken_entries--;
stolen->vaddr = addr;
flush_user_windows();
sun4c_kernel_map(stolen);
local_irq_restore(flags);
}
static void free_locked_segment(unsigned long addr)
{
struct sun4c_mmu_entry *entry;
unsigned long flags;
unsigned char pseg;
local_irq_save(flags);
addr &= SUN4C_REAL_PGDIR_MASK;
pseg = sun4c_get_segmap(addr);
entry = &mmu_entry_pool[pseg];
flush_user_windows();
sun4c_flush_segment(addr);
sun4c_kernel_unmap(entry);
add_ring(&sun4c_ufree_ring, entry);
max_user_taken_entries++;
local_irq_restore(flags);
}
static inline void garbage_collect(int entry)
{
int start, end;
/* 32 buckets per segment... */
entry &= ~31;
start = entry;
for (end = (start + 32); start < end; start++)
if (sun4c_bucket[start] != BUCKET_EMPTY)
return;
/* Entire segment empty, release it. */
free_locked_segment(BUCKET_ADDR(entry));
}
static struct thread_info *sun4c_alloc_thread_info_node(int node)
{
unsigned long addr, pages;
int entry;
pages = __get_free_pages(GFP_KERNEL, THREAD_INFO_ORDER);
if (!pages)
return NULL;
for (entry = sun4c_lowbucket_avail; entry < NR_TASK_BUCKETS; entry++)
if (sun4c_bucket[entry] == BUCKET_EMPTY)
break;
if (entry == NR_TASK_BUCKETS) {
free_pages(pages, THREAD_INFO_ORDER);
return NULL;
}
if (entry >= sun4c_lowbucket_avail)
sun4c_lowbucket_avail = entry + 1;
addr = BUCKET_ADDR(entry);
sun4c_bucket[entry] = (union task_union *) addr;
if(sun4c_get_segmap(addr) == invalid_segment)
get_locked_segment(addr);
/* We are changing the virtual color of the page(s)
* so we must flush the cache to guarantee consistency.
*/
sun4c_flush_page(pages);
sun4c_flush_page(pages + PAGE_SIZE);
sun4c_put_pte(addr, BUCKET_PTE(pages));
sun4c_put_pte(addr + PAGE_SIZE, BUCKET_PTE(pages + PAGE_SIZE));
#ifdef CONFIG_DEBUG_STACK_USAGE
memset((void *)addr, 0, PAGE_SIZE << THREAD_INFO_ORDER);
#endif /* DEBUG_STACK_USAGE */
return (struct thread_info *) addr;
}
static void sun4c_free_thread_info(struct thread_info *ti)
{
unsigned long tiaddr = (unsigned long) ti;
unsigned long pages = BUCKET_PTE_PAGE(sun4c_get_pte(tiaddr));
int entry = BUCKET_NUM(tiaddr);
/* We are deleting a mapping, so the flush here is mandatory. */
sun4c_flush_page(tiaddr);
sun4c_flush_page(tiaddr + PAGE_SIZE);
sun4c_put_pte(tiaddr, 0);
sun4c_put_pte(tiaddr + PAGE_SIZE, 0);
sun4c_bucket[entry] = BUCKET_EMPTY;
if (entry < sun4c_lowbucket_avail)
sun4c_lowbucket_avail = entry;
free_pages(pages, THREAD_INFO_ORDER);
garbage_collect(entry);
}
static void __init sun4c_init_buckets(void)
{
int entry;
if (sizeof(union thread_union) != (PAGE_SIZE << THREAD_INFO_ORDER)) {
extern void thread_info_size_is_bolixed_pete(void);
thread_info_size_is_bolixed_pete();
}
for (entry = 0; entry < NR_TASK_BUCKETS; entry++)
sun4c_bucket[entry] = BUCKET_EMPTY;
sun4c_lowbucket_avail = 0;
}
static unsigned long sun4c_iobuffer_start;
static unsigned long sun4c_iobuffer_end;
static unsigned long sun4c_iobuffer_high;
static unsigned long *sun4c_iobuffer_map;
static int iobuffer_map_size;
/*
* Alias our pages so they do not cause a trap.
* Also one page may be aliased into several I/O areas and we may
* finish these I/O separately.
*/
static char *sun4c_lockarea(char *vaddr, unsigned long size)
{
unsigned long base, scan;
unsigned long npages;
unsigned long vpage;
unsigned long pte;
unsigned long apage;
unsigned long high;
unsigned long flags;
npages = (((unsigned long)vaddr & ~PAGE_MASK) +
size + (PAGE_SIZE-1)) >> PAGE_SHIFT;
local_irq_save(flags);
base = bitmap_find_next_zero_area(sun4c_iobuffer_map, iobuffer_map_size,
0, npages, 0);
if (base >= iobuffer_map_size)
goto abend;
high = ((base + npages) << PAGE_SHIFT) + sun4c_iobuffer_start;
high = SUN4C_REAL_PGDIR_ALIGN(high);
while (high > sun4c_iobuffer_high) {
get_locked_segment(sun4c_iobuffer_high);
sun4c_iobuffer_high += SUN4C_REAL_PGDIR_SIZE;
}
vpage = ((unsigned long) vaddr) & PAGE_MASK;
for (scan = base; scan < base+npages; scan++) {
pte = ((vpage-PAGE_OFFSET) >> PAGE_SHIFT);
pte |= pgprot_val(SUN4C_PAGE_KERNEL);
pte |= _SUN4C_PAGE_NOCACHE;
set_bit(scan, sun4c_iobuffer_map);
apage = (scan << PAGE_SHIFT) + sun4c_iobuffer_start;
/* Flush original mapping so we see the right things later. */
sun4c_flush_page(vpage);
sun4c_put_pte(apage, pte);
vpage += PAGE_SIZE;
}
local_irq_restore(flags);
return (char *) ((base << PAGE_SHIFT) + sun4c_iobuffer_start +
(((unsigned long) vaddr) & ~PAGE_MASK));
abend:
local_irq_restore(flags);
printk("DMA vaddr=0x%p size=%08lx\n", vaddr, size);
panic("Out of iobuffer table");
return NULL;
}
static void sun4c_unlockarea(char *vaddr, unsigned long size)
{
unsigned long vpage, npages;
unsigned long flags;
int scan, high;
vpage = (unsigned long)vaddr & PAGE_MASK;
npages = (((unsigned long)vaddr & ~PAGE_MASK) +
size + (PAGE_SIZE-1)) >> PAGE_SHIFT;
local_irq_save(flags);
while (npages != 0) {
--npages;
/* This mapping is marked non-cachable, no flush necessary. */
sun4c_put_pte(vpage, 0);
clear_bit((vpage - sun4c_iobuffer_start) >> PAGE_SHIFT,
sun4c_iobuffer_map);
vpage += PAGE_SIZE;
}
/* garbage collect */
scan = (sun4c_iobuffer_high - sun4c_iobuffer_start) >> PAGE_SHIFT;
while (scan >= 0 && !sun4c_iobuffer_map[scan >> 5])
scan -= 32;
scan += 32;
high = sun4c_iobuffer_start + (scan << PAGE_SHIFT);
high = SUN4C_REAL_PGDIR_ALIGN(high) + SUN4C_REAL_PGDIR_SIZE;
while (high < sun4c_iobuffer_high) {
sun4c_iobuffer_high -= SUN4C_REAL_PGDIR_SIZE;
free_locked_segment(sun4c_iobuffer_high);
}
local_irq_restore(flags);
}
/* Note the scsi code at init time passes to here buffers
* which sit on the kernel stack, those are already locked
* by implication and fool the page locking code above
* if passed to by mistake.
*/
static __u32 sun4c_get_scsi_one(struct device *dev, char *bufptr, unsigned long len)
{
unsigned long page;
page = ((unsigned long)bufptr) & PAGE_MASK;
if (!virt_addr_valid(page)) {
sun4c_flush_page(page);
return (__u32)bufptr; /* already locked */
}
return (__u32)sun4c_lockarea(bufptr, len);
}
static void sun4c_get_scsi_sgl(struct device *dev, struct scatterlist *sg, int sz)
{
while (sz != 0) {
--sz;
sg->dma_address = (__u32)sun4c_lockarea(sg_virt(sg), sg->length);
sg->dma_length = sg->length;
sg = sg_next(sg);
}
}
static void sun4c_release_scsi_one(struct device *dev, __u32 bufptr, unsigned long len)
{
if (bufptr < sun4c_iobuffer_start)
return; /* On kernel stack or similar, see above */
sun4c_unlockarea((char *)bufptr, len);
}
static void sun4c_release_scsi_sgl(struct device *dev, struct scatterlist *sg, int sz)
{
while (sz != 0) {
--sz;
sun4c_unlockarea((char *)sg->dma_address, sg->length);
sg = sg_next(sg);
}
}
#define TASK_ENTRY_SIZE BUCKET_SIZE /* see above */
#define LONG_ALIGN(x) (((x)+(sizeof(long))-1)&~((sizeof(long))-1))
struct vm_area_struct sun4c_kstack_vma;
static void __init sun4c_init_lock_areas(void)
{
unsigned long sun4c_taskstack_start;
unsigned long sun4c_taskstack_end;
int bitmap_size;
sun4c_init_buckets();
sun4c_taskstack_start = SUN4C_LOCK_VADDR;
sun4c_taskstack_end = (sun4c_taskstack_start +
(TASK_ENTRY_SIZE * NR_TASK_BUCKETS));
if (sun4c_taskstack_end >= SUN4C_LOCK_END) {
prom_printf("Too many tasks, decrease NR_TASK_BUCKETS please.\n");
prom_halt();
}
sun4c_iobuffer_start = sun4c_iobuffer_high =
SUN4C_REAL_PGDIR_ALIGN(sun4c_taskstack_end);
sun4c_iobuffer_end = SUN4C_LOCK_END;
bitmap_size = (sun4c_iobuffer_end - sun4c_iobuffer_start) >> PAGE_SHIFT;
bitmap_size = (bitmap_size + 7) >> 3;
bitmap_size = LONG_ALIGN(bitmap_size);
iobuffer_map_size = bitmap_size << 3;
sun4c_iobuffer_map = __alloc_bootmem(bitmap_size, SMP_CACHE_BYTES, 0UL);
memset((void *) sun4c_iobuffer_map, 0, bitmap_size);
sun4c_kstack_vma.vm_mm = &init_mm;
sun4c_kstack_vma.vm_start = sun4c_taskstack_start;
sun4c_kstack_vma.vm_end = sun4c_taskstack_end;
sun4c_kstack_vma.vm_page_prot = PAGE_SHARED;
sun4c_kstack_vma.vm_flags = VM_READ | VM_WRITE | VM_EXEC;
insert_vm_struct(&init_mm, &sun4c_kstack_vma);
}
/* Cache flushing on the sun4c. */
static void sun4c_flush_cache_all(void)
{
unsigned long begin, end;
flush_user_windows();
begin = (KERNBASE + SUN4C_REAL_PGDIR_SIZE);
end = (begin + SUN4C_VAC_SIZE);
if (sun4c_vacinfo.linesize == 32) {
while (begin < end) {
__asm__ __volatile__(
"ld [%0 + 0x00], %%g0\n\t"
"ld [%0 + 0x20], %%g0\n\t"
"ld [%0 + 0x40], %%g0\n\t"
"ld [%0 + 0x60], %%g0\n\t"
"ld [%0 + 0x80], %%g0\n\t"
"ld [%0 + 0xa0], %%g0\n\t"
"ld [%0 + 0xc0], %%g0\n\t"
"ld [%0 + 0xe0], %%g0\n\t"
"ld [%0 + 0x100], %%g0\n\t"
"ld [%0 + 0x120], %%g0\n\t"
"ld [%0 + 0x140], %%g0\n\t"
"ld [%0 + 0x160], %%g0\n\t"
"ld [%0 + 0x180], %%g0\n\t"
"ld [%0 + 0x1a0], %%g0\n\t"
"ld [%0 + 0x1c0], %%g0\n\t"
"ld [%0 + 0x1e0], %%g0\n"
: : "r" (begin));
begin += 512;
}
} else {
while (begin < end) {
__asm__ __volatile__(
"ld [%0 + 0x00], %%g0\n\t"
"ld [%0 + 0x10], %%g0\n\t"
"ld [%0 + 0x20], %%g0\n\t"
"ld [%0 + 0x30], %%g0\n\t"
"ld [%0 + 0x40], %%g0\n\t"
"ld [%0 + 0x50], %%g0\n\t"
"ld [%0 + 0x60], %%g0\n\t"
"ld [%0 + 0x70], %%g0\n\t"
"ld [%0 + 0x80], %%g0\n\t"
"ld [%0 + 0x90], %%g0\n\t"
"ld [%0 + 0xa0], %%g0\n\t"
"ld [%0 + 0xb0], %%g0\n\t"
"ld [%0 + 0xc0], %%g0\n\t"
"ld [%0 + 0xd0], %%g0\n\t"
"ld [%0 + 0xe0], %%g0\n\t"
"ld [%0 + 0xf0], %%g0\n"
: : "r" (begin));
begin += 256;
}
}
}
static void sun4c_flush_cache_mm(struct mm_struct *mm)
{
int new_ctx = mm->context;
if (new_ctx != NO_CONTEXT) {
flush_user_windows();
if (sun4c_context_ring[new_ctx].num_entries) {
struct sun4c_mmu_entry *head = &sun4c_context_ring[new_ctx].ringhd;
unsigned long flags;
local_irq_save(flags);
if (head->next != head) {
struct sun4c_mmu_entry *entry = head->next;
int savectx = sun4c_get_context();
sun4c_set_context(new_ctx);
sun4c_flush_context();
do {
struct sun4c_mmu_entry *next = entry->next;
sun4c_user_unmap(entry);
free_user_entry(new_ctx, entry);
entry = next;
} while (entry != head);
sun4c_set_context(savectx);
}
local_irq_restore(flags);
}
}
}
static void sun4c_flush_cache_range(struct vm_area_struct *vma, unsigned long start, unsigned long end)
{
struct mm_struct *mm = vma->vm_mm;
int new_ctx = mm->context;
if (new_ctx != NO_CONTEXT) {
struct sun4c_mmu_entry *head = &sun4c_context_ring[new_ctx].ringhd;
struct sun4c_mmu_entry *entry;
unsigned long flags;
flush_user_windows();
local_irq_save(flags);
/* All user segmap chains are ordered on entry->vaddr. */
for (entry = head->next;
(entry != head) && ((entry->vaddr+SUN4C_REAL_PGDIR_SIZE) < start);
entry = entry->next)
;
/* Tracing various job mixtures showed that this conditional
* only passes ~35% of the time for most worse case situations,
* therefore we avoid all of this gross overhead ~65% of the time.
*/
if ((entry != head) && (entry->vaddr < end)) {
int octx = sun4c_get_context();
sun4c_set_context(new_ctx);
/* At this point, always, (start >= entry->vaddr) and
* (entry->vaddr < end), once the latter condition
* ceases to hold, or we hit the end of the list, we
* exit the loop. The ordering of all user allocated
* segmaps makes this all work out so beautifully.
*/
do {
struct sun4c_mmu_entry *next = entry->next;
unsigned long realend;
/* "realstart" is always >= entry->vaddr */
realend = entry->vaddr + SUN4C_REAL_PGDIR_SIZE;
if (end < realend)
realend = end;
if ((realend - entry->vaddr) <= (PAGE_SIZE << 3)) {
unsigned long page = entry->vaddr;
while (page < realend) {
sun4c_flush_page(page);
page += PAGE_SIZE;
}
} else {
sun4c_flush_segment(entry->vaddr);
sun4c_user_unmap(entry);
free_user_entry(new_ctx, entry);
}
entry = next;
} while ((entry != head) && (entry->vaddr < end));
sun4c_set_context(octx);
}
local_irq_restore(flags);
}
}
static void sun4c_flush_cache_page(struct vm_area_struct *vma, unsigned long page)
{
struct mm_struct *mm = vma->vm_mm;
int new_ctx = mm->context;
/* Sun4c has no separate I/D caches so cannot optimize for non
* text page flushes.
*/
if (new_ctx != NO_CONTEXT) {
int octx = sun4c_get_context();
unsigned long flags;
flush_user_windows();
local_irq_save(flags);
sun4c_set_context(new_ctx);
sun4c_flush_page(page);
sun4c_set_context(octx);
local_irq_restore(flags);
}
}
static void sun4c_flush_page_to_ram(unsigned long page)
{
unsigned long flags;
local_irq_save(flags);
sun4c_flush_page(page);
local_irq_restore(flags);
}
/* Sun4c cache is unified, both instructions and data live there, so
* no need to flush the on-stack instructions for new signal handlers.
*/
static void sun4c_flush_sig_insns(struct mm_struct *mm, unsigned long insn_addr)
{
}
/* TLB flushing on the sun4c. These routines count on the cache
* flushing code to flush the user register windows so that we need
* not do so when we get here.
*/
static void sun4c_flush_tlb_all(void)
{
struct sun4c_mmu_entry *this_entry, *next_entry;
unsigned long flags;
int savectx, ctx;
local_irq_save(flags);
this_entry = sun4c_kernel_ring.ringhd.next;
savectx = sun4c_get_context();
flush_user_windows();
while (sun4c_kernel_ring.num_entries) {
next_entry = this_entry->next;
sun4c_flush_segment(this_entry->vaddr);
for (ctx = 0; ctx < num_contexts; ctx++) {
sun4c_set_context(ctx);
sun4c_put_segmap(this_entry->vaddr, invalid_segment);
}
free_kernel_entry(this_entry, &sun4c_kernel_ring);
this_entry = next_entry;
}
sun4c_set_context(savectx);
local_irq_restore(flags);
}
static void sun4c_flush_tlb_mm(struct mm_struct *mm)
{
int new_ctx = mm->context;
if (new_ctx != NO_CONTEXT) {
struct sun4c_mmu_entry *head = &sun4c_context_ring[new_ctx].ringhd;
unsigned long flags;
local_irq_save(flags);
if (head->next != head) {
struct sun4c_mmu_entry *entry = head->next;
int savectx = sun4c_get_context();
sun4c_set_context(new_ctx);
sun4c_flush_context();
do {
struct sun4c_mmu_entry *next = entry->next;
sun4c_user_unmap(entry);
free_user_entry(new_ctx, entry);
entry = next;
} while (entry != head);
sun4c_set_context(savectx);
}
local_irq_restore(flags);
}
}
static void sun4c_flush_tlb_range(struct vm_area_struct *vma, unsigned long start, unsigned long end)
{
struct mm_struct *mm = vma->vm_mm;
int new_ctx = mm->context;
if (new_ctx != NO_CONTEXT) {
struct sun4c_mmu_entry *head = &sun4c_context_ring[new_ctx].ringhd;
struct sun4c_mmu_entry *entry;
unsigned long flags;
local_irq_save(flags);
/* See commentary in sun4c_flush_cache_range(). */
for (entry = head->next;
(entry != head) && ((entry->vaddr+SUN4C_REAL_PGDIR_SIZE) < start);
entry = entry->next)
;
if ((entry != head) && (entry->vaddr < end)) {
int octx = sun4c_get_context();
sun4c_set_context(new_ctx);
do {
struct sun4c_mmu_entry *next = entry->next;
sun4c_flush_segment(entry->vaddr);
sun4c_user_unmap(entry);
free_user_entry(new_ctx, entry);
entry = next;
} while ((entry != head) && (entry->vaddr < end));
sun4c_set_context(octx);
}
local_irq_restore(flags);
}
}
static void sun4c_flush_tlb_page(struct vm_area_struct *vma, unsigned long page)
{
struct mm_struct *mm = vma->vm_mm;
int new_ctx = mm->context;
if (new_ctx != NO_CONTEXT) {
int savectx = sun4c_get_context();
unsigned long flags;
local_irq_save(flags);
sun4c_set_context(new_ctx);
page &= PAGE_MASK;
sun4c_flush_page(page);
sun4c_put_pte(page, 0);
sun4c_set_context(savectx);
local_irq_restore(flags);
}
}
static inline void sun4c_mapioaddr(unsigned long physaddr, unsigned long virt_addr)
{
unsigned long page_entry, pg_iobits;
pg_iobits = _SUN4C_PAGE_PRESENT | _SUN4C_READABLE | _SUN4C_WRITEABLE |
_SUN4C_PAGE_IO | _SUN4C_PAGE_NOCACHE;
page_entry = ((physaddr >> PAGE_SHIFT) & SUN4C_PFN_MASK);
page_entry |= ((pg_iobits | _SUN4C_PAGE_PRIV) & ~(_SUN4C_PAGE_PRESENT));
sun4c_put_pte(virt_addr, page_entry);
}
static void sun4c_mapiorange(unsigned int bus, unsigned long xpa,
unsigned long xva, unsigned int len)
{
while (len != 0) {
len -= PAGE_SIZE;
sun4c_mapioaddr(xpa, xva);
xva += PAGE_SIZE;
xpa += PAGE_SIZE;
}
}
static void sun4c_unmapiorange(unsigned long virt_addr, unsigned int len)
{
while (len != 0) {
len -= PAGE_SIZE;
sun4c_put_pte(virt_addr, 0);
virt_addr += PAGE_SIZE;
}
}
static void sun4c_alloc_context(struct mm_struct *old_mm, struct mm_struct *mm)
{
struct ctx_list *ctxp;
ctxp = ctx_free.next;
if (ctxp != &ctx_free) {
remove_from_ctx_list(ctxp);
add_to_used_ctxlist(ctxp);
mm->context = ctxp->ctx_number;
ctxp->ctx_mm = mm;
return;
}
ctxp = ctx_used.next;
if (ctxp->ctx_mm == old_mm)
ctxp = ctxp->next;
remove_from_ctx_list(ctxp);
add_to_used_ctxlist(ctxp);
ctxp->ctx_mm->context = NO_CONTEXT;
ctxp->ctx_mm = mm;
mm->context = ctxp->ctx_number;
sun4c_demap_context(&sun4c_context_ring[ctxp->ctx_number],
ctxp->ctx_number);
}
/* Switch the current MM context. */
static void sun4c_switch_mm(struct mm_struct *old_mm, struct mm_struct *mm, struct task_struct *tsk, int cpu)
{
struct ctx_list *ctx;
int dirty = 0;
if (mm->context == NO_CONTEXT) {
dirty = 1;
sun4c_alloc_context(old_mm, mm);
} else {
/* Update the LRU ring of contexts. */
ctx = ctx_list_pool + mm->context;
remove_from_ctx_list(ctx);
add_to_used_ctxlist(ctx);
}
if (dirty || old_mm != mm)
sun4c_set_context(mm->context);
}
static void sun4c_destroy_context(struct mm_struct *mm)
{
struct ctx_list *ctx_old;
if (mm->context != NO_CONTEXT) {
sun4c_demap_context(&sun4c_context_ring[mm->context], mm->context);
ctx_old = ctx_list_pool + mm->context;
remove_from_ctx_list(ctx_old);
add_to_free_ctxlist(ctx_old);
mm->context = NO_CONTEXT;
}
}
static void sun4c_mmu_info(struct seq_file *m)
{
int used_user_entries, i;
used_user_entries = 0;
for (i = 0; i < num_contexts; i++)
used_user_entries += sun4c_context_ring[i].num_entries;
seq_printf(m,
"vacsize\t\t: %d bytes\n"
"vachwflush\t: %s\n"
"vaclinesize\t: %d bytes\n"
"mmuctxs\t\t: %d\n"
"mmupsegs\t: %d\n"
"kernelpsegs\t: %d\n"
"kfreepsegs\t: %d\n"
"usedpsegs\t: %d\n"
"ufreepsegs\t: %d\n"
"user_taken\t: %d\n"
"max_taken\t: %d\n",
sun4c_vacinfo.num_bytes,
(sun4c_vacinfo.do_hwflushes ? "yes" : "no"),
sun4c_vacinfo.linesize,
num_contexts,
(invalid_segment + 1),
sun4c_kernel_ring.num_entries,
sun4c_kfree_ring.num_entries,
used_user_entries,
sun4c_ufree_ring.num_entries,
sun4c_user_taken_entries,
max_user_taken_entries);
}
/* Nothing below here should touch the mmu hardware nor the mmu_entry
* data structures.
*/
/* First the functions which the mid-level code uses to directly
* manipulate the software page tables. Some defines since we are
* emulating the i386 page directory layout.
*/
#define PGD_PRESENT 0x001
#define PGD_RW 0x002
#define PGD_USER 0x004
#define PGD_ACCESSED 0x020
#define PGD_DIRTY 0x040
#define PGD_TABLE (PGD_PRESENT | PGD_RW | PGD_USER | PGD_ACCESSED | PGD_DIRTY)
static void sun4c_set_pte(pte_t *ptep, pte_t pte)
{
*ptep = pte;
}
static void sun4c_pgd_set(pgd_t * pgdp, pmd_t * pmdp)
{
}
static void sun4c_pmd_set(pmd_t * pmdp, pte_t * ptep)
{
pmdp->pmdv[0] = PGD_TABLE | (unsigned long) ptep;
}
static void sun4c_pmd_populate(pmd_t * pmdp, struct page * ptep)
{
if (page_address(ptep) == NULL) BUG(); /* No highmem on sun4c */
pmdp->pmdv[0] = PGD_TABLE | (unsigned long) page_address(ptep);
}
static int sun4c_pte_present(pte_t pte)
{
return ((pte_val(pte) & (_SUN4C_PAGE_PRESENT | _SUN4C_PAGE_PRIV)) != 0);
}
static void sun4c_pte_clear(pte_t *ptep) { *ptep = __pte(0); }
static int sun4c_pmd_bad(pmd_t pmd)
{
return (((pmd_val(pmd) & ~PAGE_MASK) != PGD_TABLE) ||
(!virt_addr_valid(pmd_val(pmd))));
}
static int sun4c_pmd_present(pmd_t pmd)
{
return ((pmd_val(pmd) & PGD_PRESENT) != 0);
}
#if 0 /* if PMD takes one word */
static void sun4c_pmd_clear(pmd_t *pmdp) { *pmdp = __pmd(0); }
#else /* if pmd_t is a longish aggregate */
static void sun4c_pmd_clear(pmd_t *pmdp) {
memset((void *)pmdp, 0, sizeof(pmd_t));
}
#endif
static int sun4c_pgd_none(pgd_t pgd) { return 0; }
static int sun4c_pgd_bad(pgd_t pgd) { return 0; }
static int sun4c_pgd_present(pgd_t pgd) { return 1; }
static void sun4c_pgd_clear(pgd_t * pgdp) { }
/*
* The following only work if pte_present() is true.
* Undefined behaviour if not..
*/
static pte_t sun4c_pte_mkwrite(pte_t pte)
{
pte = __pte(pte_val(pte) | _SUN4C_PAGE_WRITE);
if (pte_val(pte) & _SUN4C_PAGE_MODIFIED)
pte = __pte(pte_val(pte) | _SUN4C_PAGE_SILENT_WRITE);
return pte;
}
static pte_t sun4c_pte_mkdirty(pte_t pte)
{
pte = __pte(pte_val(pte) | _SUN4C_PAGE_MODIFIED);
if (pte_val(pte) & _SUN4C_PAGE_WRITE)
pte = __pte(pte_val(pte) | _SUN4C_PAGE_SILENT_WRITE);
return pte;
}
static pte_t sun4c_pte_mkyoung(pte_t pte)
{
pte = __pte(pte_val(pte) | _SUN4C_PAGE_ACCESSED);
if (pte_val(pte) & _SUN4C_PAGE_READ)
pte = __pte(pte_val(pte) | _SUN4C_PAGE_SILENT_READ);
return pte;
}
/*
* Conversion functions: convert a page and protection to a page entry,
* and a page entry and page directory to the page they refer to.
*/
static pte_t sun4c_mk_pte(struct page *page, pgprot_t pgprot)
{
return __pte(page_to_pfn(page) | pgprot_val(pgprot));
}
static pte_t sun4c_mk_pte_phys(unsigned long phys_page, pgprot_t pgprot)
{
return __pte((phys_page >> PAGE_SHIFT) | pgprot_val(pgprot));
}
static pte_t sun4c_mk_pte_io(unsigned long page, pgprot_t pgprot, int space)
{
return __pte(((page - PAGE_OFFSET) >> PAGE_SHIFT) | pgprot_val(pgprot));
}
static unsigned long sun4c_pte_pfn(pte_t pte)
{
return pte_val(pte) & SUN4C_PFN_MASK;
}
static pte_t sun4c_pgoff_to_pte(unsigned long pgoff)
{
return __pte(pgoff | _SUN4C_PAGE_FILE);
}
static unsigned long sun4c_pte_to_pgoff(pte_t pte)
{
return pte_val(pte) & ((1UL << PTE_FILE_MAX_BITS) - 1);
}
static inline unsigned long sun4c_pmd_page_v(pmd_t pmd)
{
return (pmd_val(pmd) & PAGE_MASK);
}
static struct page *sun4c_pmd_page(pmd_t pmd)
{
return virt_to_page(sun4c_pmd_page_v(pmd));
}
static unsigned long sun4c_pgd_page(pgd_t pgd) { return 0; }
/* to find an entry in a page-table-directory */
static inline pgd_t *sun4c_pgd_offset(struct mm_struct * mm, unsigned long address)
{
return mm->pgd + (address >> SUN4C_PGDIR_SHIFT);
}
/* Find an entry in the second-level page table.. */
static pmd_t *sun4c_pmd_offset(pgd_t * dir, unsigned long address)
{
return (pmd_t *) dir;
}
/* Find an entry in the third-level page table.. */
pte_t *sun4c_pte_offset_kernel(pmd_t * dir, unsigned long address)
{
return (pte_t *) sun4c_pmd_page_v(*dir) +
((address >> PAGE_SHIFT) & (SUN4C_PTRS_PER_PTE - 1));
}
static unsigned long sun4c_swp_type(swp_entry_t entry)
{
return (entry.val & SUN4C_SWP_TYPE_MASK);
}
static unsigned long sun4c_swp_offset(swp_entry_t entry)
{
return (entry.val >> SUN4C_SWP_OFF_SHIFT) & SUN4C_SWP_OFF_MASK;
}
static swp_entry_t sun4c_swp_entry(unsigned long type, unsigned long offset)
{
return (swp_entry_t) {
(offset & SUN4C_SWP_OFF_MASK) << SUN4C_SWP_OFF_SHIFT
| (type & SUN4C_SWP_TYPE_MASK) };
}
static void sun4c_free_pte_slow(pte_t *pte)
{
free_page((unsigned long)pte);
}
static void sun4c_free_pgd_slow(pgd_t *pgd)
{
free_page((unsigned long)pgd);
}
static pgd_t *sun4c_get_pgd_fast(void)
{
unsigned long *ret;
if ((ret = pgd_quicklist) != NULL) {
pgd_quicklist = (unsigned long *)(*ret);
ret[0] = ret[1];
pgtable_cache_size--;
} else {
pgd_t *init;
ret = (unsigned long *)__get_free_page(GFP_KERNEL);
memset (ret, 0, (KERNBASE / SUN4C_PGDIR_SIZE) * sizeof(pgd_t));
init = sun4c_pgd_offset(&init_mm, 0);
memcpy (((pgd_t *)ret) + USER_PTRS_PER_PGD, init + USER_PTRS_PER_PGD,
(PTRS_PER_PGD - USER_PTRS_PER_PGD) * sizeof(pgd_t));
}
return (pgd_t *)ret;
}
static void sun4c_free_pgd_fast(pgd_t *pgd)
{
*(unsigned long *)pgd = (unsigned long) pgd_quicklist;
pgd_quicklist = (unsigned long *) pgd;
pgtable_cache_size++;
}
static inline pte_t *
sun4c_pte_alloc_one_fast(struct mm_struct *mm, unsigned long address)
{
unsigned long *ret;
if ((ret = (unsigned long *)pte_quicklist) != NULL) {
pte_quicklist = (unsigned long *)(*ret);
ret[0] = ret[1];
pgtable_cache_size--;
}
return (pte_t *)ret;
}
static pte_t *sun4c_pte_alloc_one_kernel(struct mm_struct *mm, unsigned long address)
{
pte_t *pte;
if ((pte = sun4c_pte_alloc_one_fast(mm, address)) != NULL)
return pte;
pte = (pte_t *)get_zeroed_page(GFP_KERNEL|__GFP_REPEAT);
return pte;
}
CONFIG_HIGHPTE vs. sub-page page tables. Background: I've implemented 1K/2K page tables for s390. These sub-page page tables are required to properly support the s390 virtualization instruction with KVM. The SIE instruction requires that the page tables have 256 page table entries (pte) followed by 256 page status table entries (pgste). The pgstes are only required if the process is using the SIE instruction. The pgstes are updated by the hardware and by the hypervisor for a number of reasons, one of them is dirty and reference bit tracking. To avoid wasting memory the standard pte table allocation should return 1K/2K (31/64 bit) and 2K/4K if the process is using SIE. Problem: Page size on s390 is 4K, page table size is 1K or 2K. That means the s390 version for pte_alloc_one cannot return a pointer to a struct page. Trouble is that with the CONFIG_HIGHPTE feature on x86 pte_alloc_one cannot return a pointer to a pte either, since that would require more than 32 bit for the return value of pte_alloc_one (and the pte * would not be accessible since its not kmapped). Solution: The only solution I found to this dilemma is a new typedef: a pgtable_t. For s390 pgtable_t will be a (pte *) - to be introduced with a later patch. For everybody else it will be a (struct page *). The additional problem with the initialization of the ptl lock and the NR_PAGETABLE accounting is solved with a constructor pgtable_page_ctor and a destructor pgtable_page_dtor. The page table allocation and free functions need to call these two whenever a page table page is allocated or freed. pmd_populate will get a pgtable_t instead of a struct page pointer. To get the pgtable_t back from a pmd entry that has been installed with pmd_populate a new function pmd_pgtable is added. It replaces the pmd_page call in free_pte_range and apply_to_pte_range. Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: <linux-arch@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 05:22:04 -07:00
static pgtable_t sun4c_pte_alloc_one(struct mm_struct *mm, unsigned long address)
{
CONFIG_HIGHPTE vs. sub-page page tables. Background: I've implemented 1K/2K page tables for s390. These sub-page page tables are required to properly support the s390 virtualization instruction with KVM. The SIE instruction requires that the page tables have 256 page table entries (pte) followed by 256 page status table entries (pgste). The pgstes are only required if the process is using the SIE instruction. The pgstes are updated by the hardware and by the hypervisor for a number of reasons, one of them is dirty and reference bit tracking. To avoid wasting memory the standard pte table allocation should return 1K/2K (31/64 bit) and 2K/4K if the process is using SIE. Problem: Page size on s390 is 4K, page table size is 1K or 2K. That means the s390 version for pte_alloc_one cannot return a pointer to a struct page. Trouble is that with the CONFIG_HIGHPTE feature on x86 pte_alloc_one cannot return a pointer to a pte either, since that would require more than 32 bit for the return value of pte_alloc_one (and the pte * would not be accessible since its not kmapped). Solution: The only solution I found to this dilemma is a new typedef: a pgtable_t. For s390 pgtable_t will be a (pte *) - to be introduced with a later patch. For everybody else it will be a (struct page *). The additional problem with the initialization of the ptl lock and the NR_PAGETABLE accounting is solved with a constructor pgtable_page_ctor and a destructor pgtable_page_dtor. The page table allocation and free functions need to call these two whenever a page table page is allocated or freed. pmd_populate will get a pgtable_t instead of a struct page pointer. To get the pgtable_t back from a pmd entry that has been installed with pmd_populate a new function pmd_pgtable is added. It replaces the pmd_page call in free_pte_range and apply_to_pte_range. Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: <linux-arch@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 05:22:04 -07:00
pte_t *pte;
struct page *page;
pte = sun4c_pte_alloc_one_kernel(mm, address);
if (pte == NULL)
return NULL;
CONFIG_HIGHPTE vs. sub-page page tables. Background: I've implemented 1K/2K page tables for s390. These sub-page page tables are required to properly support the s390 virtualization instruction with KVM. The SIE instruction requires that the page tables have 256 page table entries (pte) followed by 256 page status table entries (pgste). The pgstes are only required if the process is using the SIE instruction. The pgstes are updated by the hardware and by the hypervisor for a number of reasons, one of them is dirty and reference bit tracking. To avoid wasting memory the standard pte table allocation should return 1K/2K (31/64 bit) and 2K/4K if the process is using SIE. Problem: Page size on s390 is 4K, page table size is 1K or 2K. That means the s390 version for pte_alloc_one cannot return a pointer to a struct page. Trouble is that with the CONFIG_HIGHPTE feature on x86 pte_alloc_one cannot return a pointer to a pte either, since that would require more than 32 bit for the return value of pte_alloc_one (and the pte * would not be accessible since its not kmapped). Solution: The only solution I found to this dilemma is a new typedef: a pgtable_t. For s390 pgtable_t will be a (pte *) - to be introduced with a later patch. For everybody else it will be a (struct page *). The additional problem with the initialization of the ptl lock and the NR_PAGETABLE accounting is solved with a constructor pgtable_page_ctor and a destructor pgtable_page_dtor. The page table allocation and free functions need to call these two whenever a page table page is allocated or freed. pmd_populate will get a pgtable_t instead of a struct page pointer. To get the pgtable_t back from a pmd entry that has been installed with pmd_populate a new function pmd_pgtable is added. It replaces the pmd_page call in free_pte_range and apply_to_pte_range. Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: <linux-arch@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 05:22:04 -07:00
page = virt_to_page(pte);
pgtable_page_ctor(page);
return page;
}
static inline void sun4c_free_pte_fast(pte_t *pte)
{
*(unsigned long *)pte = (unsigned long) pte_quicklist;
pte_quicklist = (unsigned long *) pte;
pgtable_cache_size++;
}
CONFIG_HIGHPTE vs. sub-page page tables. Background: I've implemented 1K/2K page tables for s390. These sub-page page tables are required to properly support the s390 virtualization instruction with KVM. The SIE instruction requires that the page tables have 256 page table entries (pte) followed by 256 page status table entries (pgste). The pgstes are only required if the process is using the SIE instruction. The pgstes are updated by the hardware and by the hypervisor for a number of reasons, one of them is dirty and reference bit tracking. To avoid wasting memory the standard pte table allocation should return 1K/2K (31/64 bit) and 2K/4K if the process is using SIE. Problem: Page size on s390 is 4K, page table size is 1K or 2K. That means the s390 version for pte_alloc_one cannot return a pointer to a struct page. Trouble is that with the CONFIG_HIGHPTE feature on x86 pte_alloc_one cannot return a pointer to a pte either, since that would require more than 32 bit for the return value of pte_alloc_one (and the pte * would not be accessible since its not kmapped). Solution: The only solution I found to this dilemma is a new typedef: a pgtable_t. For s390 pgtable_t will be a (pte *) - to be introduced with a later patch. For everybody else it will be a (struct page *). The additional problem with the initialization of the ptl lock and the NR_PAGETABLE accounting is solved with a constructor pgtable_page_ctor and a destructor pgtable_page_dtor. The page table allocation and free functions need to call these two whenever a page table page is allocated or freed. pmd_populate will get a pgtable_t instead of a struct page pointer. To get the pgtable_t back from a pmd entry that has been installed with pmd_populate a new function pmd_pgtable is added. It replaces the pmd_page call in free_pte_range and apply_to_pte_range. Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: <linux-arch@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 05:22:04 -07:00
static void sun4c_pte_free(pgtable_t pte)
{
CONFIG_HIGHPTE vs. sub-page page tables. Background: I've implemented 1K/2K page tables for s390. These sub-page page tables are required to properly support the s390 virtualization instruction with KVM. The SIE instruction requires that the page tables have 256 page table entries (pte) followed by 256 page status table entries (pgste). The pgstes are only required if the process is using the SIE instruction. The pgstes are updated by the hardware and by the hypervisor for a number of reasons, one of them is dirty and reference bit tracking. To avoid wasting memory the standard pte table allocation should return 1K/2K (31/64 bit) and 2K/4K if the process is using SIE. Problem: Page size on s390 is 4K, page table size is 1K or 2K. That means the s390 version for pte_alloc_one cannot return a pointer to a struct page. Trouble is that with the CONFIG_HIGHPTE feature on x86 pte_alloc_one cannot return a pointer to a pte either, since that would require more than 32 bit for the return value of pte_alloc_one (and the pte * would not be accessible since its not kmapped). Solution: The only solution I found to this dilemma is a new typedef: a pgtable_t. For s390 pgtable_t will be a (pte *) - to be introduced with a later patch. For everybody else it will be a (struct page *). The additional problem with the initialization of the ptl lock and the NR_PAGETABLE accounting is solved with a constructor pgtable_page_ctor and a destructor pgtable_page_dtor. The page table allocation and free functions need to call these two whenever a page table page is allocated or freed. pmd_populate will get a pgtable_t instead of a struct page pointer. To get the pgtable_t back from a pmd entry that has been installed with pmd_populate a new function pmd_pgtable is added. It replaces the pmd_page call in free_pte_range and apply_to_pte_range. Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: <linux-arch@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 05:22:04 -07:00
pgtable_page_dtor(pte);
sun4c_free_pte_fast(page_address(pte));
}
/*
* allocating and freeing a pmd is trivial: the 1-entry pmd is
* inside the pgd, so has no extra memory associated with it.
*/
static pmd_t *sun4c_pmd_alloc_one(struct mm_struct *mm, unsigned long address)
{
BUG();
return NULL;
}
static void sun4c_free_pmd_fast(pmd_t * pmd) { }
static void sun4c_check_pgt_cache(int low, int high)
{
if (pgtable_cache_size > high) {
do {
if (pgd_quicklist)
sun4c_free_pgd_slow(sun4c_get_pgd_fast());
if (pte_quicklist)
sun4c_free_pte_slow(sun4c_pte_alloc_one_fast(NULL, 0));
} while (pgtable_cache_size > low);
}
}
/* An experiment, turn off by default for now... -DaveM */
#define SUN4C_PRELOAD_PSEG
void sun4c_update_mmu_cache(struct vm_area_struct *vma, unsigned long address, pte_t *ptep)
{
unsigned long flags;
int pseg;
if (vma->vm_mm->context == NO_CONTEXT)
return;
local_irq_save(flags);
address &= PAGE_MASK;
if ((pseg = sun4c_get_segmap(address)) == invalid_segment) {
struct sun4c_mmu_entry *entry = sun4c_user_strategy();
struct mm_struct *mm = vma->vm_mm;
unsigned long start, end;
entry->vaddr = start = (address & SUN4C_REAL_PGDIR_MASK);
entry->ctx = mm->context;
add_ring_ordered(sun4c_context_ring + mm->context, entry);
sun4c_put_segmap(entry->vaddr, entry->pseg);
end = start + SUN4C_REAL_PGDIR_SIZE;
while (start < end) {
#ifdef SUN4C_PRELOAD_PSEG
pgd_t *pgdp = sun4c_pgd_offset(mm, start);
pte_t *ptep;
if (!pgdp)
goto no_mapping;
ptep = sun4c_pte_offset_kernel((pmd_t *) pgdp, start);
if (!ptep || !(pte_val(*ptep) & _SUN4C_PAGE_PRESENT))
goto no_mapping;
sun4c_put_pte(start, pte_val(*ptep));
goto next;
no_mapping:
#endif
sun4c_put_pte(start, 0);
#ifdef SUN4C_PRELOAD_PSEG
next:
#endif
start += PAGE_SIZE;
}
#ifndef SUN4C_PRELOAD_PSEG
sun4c_put_pte(address, pte_val(*ptep));
#endif
local_irq_restore(flags);
return;
} else {
struct sun4c_mmu_entry *entry = &mmu_entry_pool[pseg];
remove_lru(entry);
add_lru(entry);
}
sun4c_put_pte(address, pte_val(*ptep));
local_irq_restore(flags);
}
extern void sparc_context_init(int);
extern unsigned long bootmem_init(unsigned long *pages_avail);
extern unsigned long last_valid_pfn;
void __init sun4c_paging_init(void)
{
int i, cnt;
unsigned long kernel_end, vaddr;
extern struct resource sparc_iomap;
unsigned long end_pfn, pages_avail;
kernel_end = (unsigned long) &_end;
kernel_end = SUN4C_REAL_PGDIR_ALIGN(kernel_end);
pages_avail = 0;
last_valid_pfn = bootmem_init(&pages_avail);
end_pfn = last_valid_pfn;
sun4c_probe_mmu();
invalid_segment = (num_segmaps - 1);
sun4c_init_mmu_entry_pool();
sun4c_init_rings();
sun4c_init_map_kernelprom(kernel_end);
sun4c_init_clean_mmu(kernel_end);
sun4c_init_fill_kernel_ring(SUN4C_KERNEL_BUCKETS);
sun4c_init_lock_area(sparc_iomap.start, IOBASE_END);
sun4c_init_lock_area(DVMA_VADDR, DVMA_END);
sun4c_init_lock_areas();
sun4c_init_fill_user_ring();
sun4c_set_context(0);
memset(swapper_pg_dir, 0, PAGE_SIZE);
memset(pg0, 0, PAGE_SIZE);
memset(pg1, 0, PAGE_SIZE);
memset(pg2, 0, PAGE_SIZE);
memset(pg3, 0, PAGE_SIZE);
/* Save work later. */
vaddr = VMALLOC_START;
swapper_pg_dir[vaddr>>SUN4C_PGDIR_SHIFT] = __pgd(PGD_TABLE | (unsigned long) pg0);
vaddr += SUN4C_PGDIR_SIZE;
swapper_pg_dir[vaddr>>SUN4C_PGDIR_SHIFT] = __pgd(PGD_TABLE | (unsigned long) pg1);
vaddr += SUN4C_PGDIR_SIZE;
swapper_pg_dir[vaddr>>SUN4C_PGDIR_SHIFT] = __pgd(PGD_TABLE | (unsigned long) pg2);
vaddr += SUN4C_PGDIR_SIZE;
swapper_pg_dir[vaddr>>SUN4C_PGDIR_SHIFT] = __pgd(PGD_TABLE | (unsigned long) pg3);
sun4c_init_ss2_cache_bug();
sparc_context_init(num_contexts);
{
unsigned long zones_size[MAX_NR_ZONES];
unsigned long zholes_size[MAX_NR_ZONES];
unsigned long npages;
int znum;
for (znum = 0; znum < MAX_NR_ZONES; znum++)
zones_size[znum] = zholes_size[znum] = 0;
npages = max_low_pfn - pfn_base;
zones_size[ZONE_DMA] = npages;
zholes_size[ZONE_DMA] = npages - pages_avail;
npages = highend_pfn - max_low_pfn;
zones_size[ZONE_HIGHMEM] = npages;
zholes_size[ZONE_HIGHMEM] = npages - calc_highpages();
free_area_init_node(0, zones_size, pfn_base, zholes_size);
}
cnt = 0;
for (i = 0; i < num_segmaps; i++)
if (mmu_entry_pool[i].locked)
cnt++;
max_user_taken_entries = num_segmaps - cnt - 40 - 1;
printk("SUN4C: %d mmu entries for the kernel\n", cnt);
}
static pgprot_t sun4c_pgprot_noncached(pgprot_t prot)
{
prot |= __pgprot(_SUN4C_PAGE_IO | _SUN4C_PAGE_NOCACHE);
return prot;
}
/* Load up routines and constants for sun4c mmu */
void __init ld_mmu_sun4c(void)
{
extern void ___xchg32_sun4c(void);
printk("Loading sun4c MMU routines\n");
/* First the constants */
BTFIXUPSET_SIMM13(pgdir_shift, SUN4C_PGDIR_SHIFT);
BTFIXUPSET_SETHI(pgdir_size, SUN4C_PGDIR_SIZE);
BTFIXUPSET_SETHI(pgdir_mask, SUN4C_PGDIR_MASK);
BTFIXUPSET_SIMM13(ptrs_per_pmd, SUN4C_PTRS_PER_PMD);
BTFIXUPSET_SIMM13(ptrs_per_pgd, SUN4C_PTRS_PER_PGD);
BTFIXUPSET_SIMM13(user_ptrs_per_pgd, KERNBASE / SUN4C_PGDIR_SIZE);
BTFIXUPSET_INT(page_none, pgprot_val(SUN4C_PAGE_NONE));
PAGE_SHARED = pgprot_val(SUN4C_PAGE_SHARED);
BTFIXUPSET_INT(page_copy, pgprot_val(SUN4C_PAGE_COPY));
BTFIXUPSET_INT(page_readonly, pgprot_val(SUN4C_PAGE_READONLY));
BTFIXUPSET_INT(page_kernel, pgprot_val(SUN4C_PAGE_KERNEL));
page_kernel = pgprot_val(SUN4C_PAGE_KERNEL);
/* Functions */
BTFIXUPSET_CALL(pgprot_noncached, sun4c_pgprot_noncached, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(___xchg32, ___xchg32_sun4c, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(do_check_pgt_cache, sun4c_check_pgt_cache, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(flush_cache_all, sun4c_flush_cache_all, BTFIXUPCALL_NORM);
if (sun4c_vacinfo.do_hwflushes) {
BTFIXUPSET_CALL(sun4c_flush_page, sun4c_flush_page_hw, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(sun4c_flush_segment, sun4c_flush_segment_hw, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(sun4c_flush_context, sun4c_flush_context_hw, BTFIXUPCALL_NORM);
} else {
BTFIXUPSET_CALL(sun4c_flush_page, sun4c_flush_page_sw, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(sun4c_flush_segment, sun4c_flush_segment_sw, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(sun4c_flush_context, sun4c_flush_context_sw, BTFIXUPCALL_NORM);
}
BTFIXUPSET_CALL(flush_tlb_mm, sun4c_flush_tlb_mm, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(flush_cache_mm, sun4c_flush_cache_mm, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(destroy_context, sun4c_destroy_context, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(switch_mm, sun4c_switch_mm, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(flush_cache_page, sun4c_flush_cache_page, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(flush_tlb_page, sun4c_flush_tlb_page, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(flush_tlb_range, sun4c_flush_tlb_range, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(flush_cache_range, sun4c_flush_cache_range, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(__flush_page_to_ram, sun4c_flush_page_to_ram, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(flush_tlb_all, sun4c_flush_tlb_all, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(flush_sig_insns, sun4c_flush_sig_insns, BTFIXUPCALL_NOP);
BTFIXUPSET_CALL(set_pte, sun4c_set_pte, BTFIXUPCALL_STO1O0);
BTFIXUPSET_CALL(pte_pfn, sun4c_pte_pfn, BTFIXUPCALL_NORM);
#if 0 /* PAGE_SHIFT <= 12 */ /* Eek. Investigate. XXX */
BTFIXUPSET_CALL(pmd_page, sun4c_pmd_page, BTFIXUPCALL_ANDNINT(PAGE_SIZE - 1));
#else
BTFIXUPSET_CALL(pmd_page, sun4c_pmd_page, BTFIXUPCALL_NORM);
#endif
BTFIXUPSET_CALL(pmd_set, sun4c_pmd_set, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(pmd_populate, sun4c_pmd_populate, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(pte_present, sun4c_pte_present, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(pte_clear, sun4c_pte_clear, BTFIXUPCALL_STG0O0);
BTFIXUPSET_CALL(pmd_bad, sun4c_pmd_bad, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(pmd_present, sun4c_pmd_present, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(pmd_clear, sun4c_pmd_clear, BTFIXUPCALL_STG0O0);
BTFIXUPSET_CALL(pgd_none, sun4c_pgd_none, BTFIXUPCALL_RETINT(0));
BTFIXUPSET_CALL(pgd_bad, sun4c_pgd_bad, BTFIXUPCALL_RETINT(0));
BTFIXUPSET_CALL(pgd_present, sun4c_pgd_present, BTFIXUPCALL_RETINT(1));
BTFIXUPSET_CALL(pgd_clear, sun4c_pgd_clear, BTFIXUPCALL_NOP);
BTFIXUPSET_CALL(mk_pte, sun4c_mk_pte, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(mk_pte_phys, sun4c_mk_pte_phys, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(mk_pte_io, sun4c_mk_pte_io, BTFIXUPCALL_NORM);
BTFIXUPSET_INT(pte_modify_mask, _SUN4C_PAGE_CHG_MASK);
BTFIXUPSET_CALL(pmd_offset, sun4c_pmd_offset, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(pte_offset_kernel, sun4c_pte_offset_kernel, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(free_pte_fast, sun4c_free_pte_fast, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(pte_free, sun4c_pte_free, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(pte_alloc_one_kernel, sun4c_pte_alloc_one_kernel, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(pte_alloc_one, sun4c_pte_alloc_one, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(free_pmd_fast, sun4c_free_pmd_fast, BTFIXUPCALL_NOP);
BTFIXUPSET_CALL(pmd_alloc_one, sun4c_pmd_alloc_one, BTFIXUPCALL_RETO0);
BTFIXUPSET_CALL(free_pgd_fast, sun4c_free_pgd_fast, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(get_pgd_fast, sun4c_get_pgd_fast, BTFIXUPCALL_NORM);
BTFIXUPSET_HALF(pte_writei, _SUN4C_PAGE_WRITE);
BTFIXUPSET_HALF(pte_dirtyi, _SUN4C_PAGE_MODIFIED);
BTFIXUPSET_HALF(pte_youngi, _SUN4C_PAGE_ACCESSED);
BTFIXUPSET_HALF(pte_filei, _SUN4C_PAGE_FILE);
BTFIXUPSET_HALF(pte_wrprotecti, _SUN4C_PAGE_WRITE|_SUN4C_PAGE_SILENT_WRITE);
BTFIXUPSET_HALF(pte_mkcleani, _SUN4C_PAGE_MODIFIED|_SUN4C_PAGE_SILENT_WRITE);
BTFIXUPSET_HALF(pte_mkoldi, _SUN4C_PAGE_ACCESSED|_SUN4C_PAGE_SILENT_READ);
BTFIXUPSET_CALL(pte_mkwrite, sun4c_pte_mkwrite, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(pte_mkdirty, sun4c_pte_mkdirty, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(pte_mkyoung, sun4c_pte_mkyoung, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(update_mmu_cache, sun4c_update_mmu_cache, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(pte_to_pgoff, sun4c_pte_to_pgoff, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(pgoff_to_pte, sun4c_pgoff_to_pte, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(mmu_lockarea, sun4c_lockarea, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(mmu_unlockarea, sun4c_unlockarea, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(mmu_get_scsi_one, sun4c_get_scsi_one, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(mmu_get_scsi_sgl, sun4c_get_scsi_sgl, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(mmu_release_scsi_one, sun4c_release_scsi_one, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(mmu_release_scsi_sgl, sun4c_release_scsi_sgl, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(mmu_map_dma_area, sun4c_map_dma_area, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(mmu_unmap_dma_area, sun4c_unmap_dma_area, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(sparc_mapiorange, sun4c_mapiorange, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(sparc_unmapiorange, sun4c_unmapiorange, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(__swp_type, sun4c_swp_type, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(__swp_offset, sun4c_swp_offset, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(__swp_entry, sun4c_swp_entry, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(alloc_thread_info_node, sun4c_alloc_thread_info_node, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(free_thread_info, sun4c_free_thread_info, BTFIXUPCALL_NORM);
BTFIXUPSET_CALL(mmu_info, sun4c_mmu_info, BTFIXUPCALL_NORM);
/* These should _never_ get called with two level tables. */
BTFIXUPSET_CALL(pgd_set, sun4c_pgd_set, BTFIXUPCALL_NOP);
BTFIXUPSET_CALL(pgd_page_vaddr, sun4c_pgd_page, BTFIXUPCALL_RETO0);
}