minix/minix/servers/vm/cache.c
David van Moolenbroek e321f65582 libfsdriver: support mmap on FSes with no device
This patch adds (very limited) support for memory-mapping pages on
file systems that are mounted on the special "none" device and that
do not implement PEEK support by themselves.  This includes hgfs,
vbfs, and procfs.

The solution is implemented in libvtreefs, and consists of allocating
pages, filling them with content by calling the file system's READ
functionality, passing the pages to VM, and freeing them again.  A new
VM flag is used to indicate that these pages should be mapped in only
once, and thus not cached beyond their single use.  This prevents
stale data from getting mapped in without the involvement of the file
system, which would be problematic on file systems where file contents
may become outdated at any time.  No VM caching means no sharing and
poor performance, but mmap no longer fails on these file systems.

Compared to a libc-based approach, this patch retains the on-demand
nature of mmap.  Especially tail(1) is known to map in a large file
area only to use a small portion of it.

All file systems now need to be given permission for the SETCACHEPAGE
and CLEARCACHE calls to VM.

A very basic regression test is added to test74.

Change-Id: I17afc4cb97315b515cad1542521b98f293b6b559
2014-11-15 11:19:52 +00:00

333 lines
7 KiB
C

/* File that implements the data structure, insert, lookup and remove
* functions for file system cache blocks.
*
* Cache blocks can be mapped into the memory of processes by the
* 'cache' and 'file' memory types.
*/
#include <assert.h>
#include <string.h>
#include <minix/hash.h>
#include "proto.h"
#include "vm.h"
#include "region.h"
#include "glo.h"
#include "cache.h"
/* cache datastructure */
#define HASHSIZE 65536
static struct cached_page *cache_hash_bydev[HASHSIZE];
static struct cached_page *cache_hash_byino[HASHSIZE];
static struct cached_page *lru_oldest = NULL, *lru_newest = NULL;
static u32_t cached_pages = 0;
static void lru_rm(struct cached_page *hb)
{
struct cached_page *newer = hb->newer, *older = hb->older;
assert(lru_newest);
assert(lru_oldest);
if(newer) {
assert(newer->older == hb);
newer->older = older;
}
if(older) {
assert(older->newer == hb);
older->newer = newer;
}
if(lru_newest == hb) { assert(!newer); lru_newest = older; }
if(lru_oldest == hb) { assert(!older); lru_oldest = newer; }
if(lru_newest) assert(lru_newest->newer == NULL);
if(lru_oldest) assert(lru_oldest->older == NULL);
cached_pages--;
}
static void lru_add(struct cached_page *hb)
{
if(lru_newest) {
assert(lru_oldest);
assert(!lru_newest->newer);
lru_newest->newer = hb;
} else {
assert(!lru_oldest);
lru_oldest = hb;
}
hb->older = lru_newest;
hb->newer = NULL;
lru_newest = hb;
cached_pages++;
}
void cache_lru_touch(struct cached_page *hb)
{
lru_rm(hb);
lru_add(hb);
}
static __inline u32_t makehash(u32_t p1, u64_t p2)
{
u32_t offlo = ex64lo(p2), offhi = ex64hi(p2),
v = 0x12345678;
hash_mix(p1, offlo, offhi);
hash_final(offlo, offhi, v);
return v % HASHSIZE;
}
#if CACHE_SANITY
void cache_sanitycheck_internal(void)
{
int h;
int n = 0;
int byino = 0;
int withino = 0;
int bydev_total = 0, lru_total = 0;
struct cached_page *cp;
for(h = 0; h < HASHSIZE; h++) {
for(cp = cache_hash_bydev[h]; cp; cp = cp->hash_next_dev) {
assert(cp->dev != NO_DEV);
assert(h == makehash(cp->dev, cp->dev_offset));
assert(cp == find_cached_page_bydev(cp->dev, cp->dev_offset, cp->ino, cp->ino_offset));
if(cp->ino != VMC_NO_INODE) withino++;
bydev_total++;
n++;
assert(n < 1500000);
}
for(cp = cache_hash_byino[h]; cp; cp = cp->hash_next_ino) {
assert(cp->dev != NO_DEV);
assert(cp->ino != VMC_NO_INODE);
assert(h == makehash(cp->ino, cp->ino_offset));
byino++;
n++;
assert(n < 1500000);
}
}
assert(byino == withino);
if(lru_newest) {
assert(lru_oldest);
assert(!lru_newest->newer);
assert(!lru_oldest->older);
} else {
assert(!lru_oldest);
}
for(cp = lru_oldest; cp; cp = cp->newer) {
struct cached_page *newer = cp->newer,
*older = cp->older;
if(newer) assert(newer->older == cp);
if(older) assert(older->newer == cp);
lru_total++;
}
assert(lru_total == bydev_total);
assert(lru_total == cached_pages);
}
#endif
#define rmhash_f(fname, nextfield) \
static void \
fname(struct cached_page *cp, struct cached_page **head) \
{ \
struct cached_page *hb; \
if(*head == cp) { *head = cp->nextfield; return; } \
for(hb = *head; hb && cp != hb->nextfield; hb = hb->nextfield) ; \
assert(hb); assert(hb->nextfield == cp); \
hb->nextfield = cp->nextfield; \
return; \
}
rmhash_f(rmhash_byino, hash_next_ino)
rmhash_f(rmhash_bydev, hash_next_dev)
static void addcache_byino(struct cached_page *hb)
{
int hv_ino = makehash(hb->ino, hb->ino_offset);
assert(hb->ino != VMC_NO_INODE);
hb->hash_next_ino = cache_hash_byino[hv_ino];
cache_hash_byino[hv_ino] = hb;
}
static void
update_inohash(struct cached_page *hb, ino_t ino, u64_t ino_off)
{
assert(ino != VMC_NO_INODE);
if(hb->ino != VMC_NO_INODE) {
int h = makehash(hb->ino, hb->ino_offset);
rmhash_byino(hb, &cache_hash_byino[h]);
}
hb->ino = ino;
hb->ino_offset = ino_off;
addcache_byino(hb);
}
struct cached_page *
find_cached_page_bydev(dev_t dev, u64_t dev_off, ino_t ino, u64_t ino_off, int touchlru)
{
struct cached_page *hb;
for(hb = cache_hash_bydev[makehash(dev, dev_off)]; hb; hb=hb->hash_next_dev) {
if(hb->dev == dev && hb->dev_offset == dev_off) {
if(ino != VMC_NO_INODE) {
if(hb->ino != ino || hb->ino_offset != ino_off) {
update_inohash(hb, ino, ino_off);
}
}
if(touchlru) cache_lru_touch(hb);
return hb;
}
}
return NULL;
}
struct cached_page *find_cached_page_byino(dev_t dev, ino_t ino, u64_t ino_off, int touchlru)
{
struct cached_page *hb;
assert(ino != VMC_NO_INODE);
assert(dev != NO_DEV);
for(hb = cache_hash_byino[makehash(ino, ino_off)]; hb; hb=hb->hash_next_ino) {
if(hb->dev == dev && hb->ino == ino && hb->ino_offset == ino_off) {
if(touchlru) cache_lru_touch(hb);
return hb;
}
}
return NULL;
}
int addcache(dev_t dev, u64_t dev_off, ino_t ino, u64_t ino_off, int flags,
struct phys_block *pb)
{
int hv_dev;
struct cached_page *hb;
if(pb->flags & PBF_INCACHE) {
printf("VM: already in cache\n");
return EINVAL;
}
if(!SLABALLOC(hb)) {
printf("VM: no memory for cache node\n");
return ENOMEM;
}
assert(dev != NO_DEV);
#if CACHE_SANITY
assert(!find_cached_page_bydev(dev, dev_off, ino, ino_off));
#endif
hb->dev = dev;
hb->dev_offset = dev_off;
hb->ino = ino;
hb->ino_offset = ino_off;
hb->flags = flags & VMSF_ONCE;
hb->page = pb;
hb->page->refcount++; /* block also referenced by cache now */
hb->page->flags |= PBF_INCACHE;
hv_dev = makehash(dev, dev_off);
hb->hash_next_dev = cache_hash_bydev[hv_dev];
cache_hash_bydev[hv_dev] = hb;
if(hb->ino != VMC_NO_INODE)
addcache_byino(hb);
lru_add(hb);
return OK;
}
void rmcache(struct cached_page *cp)
{
struct phys_block *pb = cp->page;
int hv_dev = makehash(cp->dev, cp->dev_offset);
assert(cp->page->flags & PBF_INCACHE);
cp->page->flags &= ~PBF_INCACHE;
rmhash_bydev(cp, &cache_hash_bydev[hv_dev]);
if(cp->ino != VMC_NO_INODE) {
int hv_ino = makehash(cp->ino, cp->ino_offset);
rmhash_byino(cp, &cache_hash_byino[hv_ino]);
}
assert(cp->page->refcount >= 1);
cp->page->refcount--;
lru_rm(cp);
if(pb->refcount == 0) {
assert(pb->phys != MAP_NONE);
free_mem(ABS2CLICK(pb->phys), 1);
SLABFREE(pb);
}
SLABFREE(cp);
}
int cache_freepages(int pages)
{
struct cached_page *cp, *newercp;
int freed = 0;
int oldsteps = 0;
int skips = 0;
for(cp = lru_oldest; cp && freed < pages; cp = newercp) {
newercp = cp->newer;
assert(cp->page->refcount >= 1);
if(cp->page->refcount == 1) {
rmcache(cp);
freed++;
skips = 0;
} else skips++;
oldsteps++;
}
return freed;
}
/*
* Remove all pages that are associated with the given device.
*/
void
clear_cache_bydev(dev_t dev)
{
struct cached_page *cp, *ncp;
int h;
for (h = 0; h < HASHSIZE; h++) {
for (cp = cache_hash_bydev[h]; cp != NULL; cp = ncp) {
ncp = cp->hash_next_dev;
if (cp->dev == dev)
rmcache(cp);
}
}
}
void get_stats_info(struct vm_stats_info *vsi)
{
vsi->vsi_cached = cached_pages;
}