d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

- Change the i386's sf_buf implementation so that it never allocates

Browse Source 
more than one sf_buf for one vm_page.  To accomplish this, we add
   a global hash table mapping vm_pages to sf_bufs and a reference
   count to each sf_buf.  (This is similar to the patches for RELENG_4
   at http://www.cs.princeton.edu/~yruan/debox/.)

   For the uninitiated, an sf_buf is nothing more than a kernel virtual
   address that is used for temporary virtual-to-physical mappings by
   sendfile(2) and zero-copy sockets.  As such, there is no reason for
   one vm_page to have several sf_bufs mapping it.  In fact, using more
   than one sf_buf for a single vm_page increases the likelihood that
   sendfile(2) blocks, hurting throughput.
   (See http://www.cs.princeton.edu/~yruan/debox/.)
This commit is contained in:
Alan Cox 2003-11-17 18:22:24 +00:00
parent eb5ed42cfe
commit 0543fa5398
Notes: svn2git 2020-12-20 02:59:44 +00:00 
svn path=/head/; revision=122860
2 changed files with 52 additions and 29 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

78
sys/i386/i386/vm_machdep.c
View File

@ -100,17 +100,24 @@ static volatile u_int cpu_reset_proxy_active;
static void sf_buf_init(void *arg);
SYSINIT(sock_sf, SI_SUB_MBUF, SI_ORDER_ANY, sf_buf_init, NULL)
/*
* Expanded sf_freelist head. Really an SLIST_HEAD() in disguise, with the
* sf_freelist head with the sf_lock mutex.
*/
static struct {
SLIST_HEAD(, sf_buf) sf_head;
struct mtx sf_lock;
} sf_freelist;
LIST_HEAD(sf_head, sf_buf);
/*
* A hash table of active sendfile(2) buffers
*/
static struct sf_head *sf_buf_active;
static u_long sf_buf_hashmask;
#define SF_BUF_HASH(m) (((m) - vm_page_array) & sf_buf_hashmask)
static struct sf_head sf_buf_freelist;
static u_int sf_buf_alloc_want;
/*
* A lock used to synchronize access to the hash table and free list
*/
static struct mtx sf_buf_lock;
extern int _ucodesel, _udatasel;
/*
@ -575,16 +582,17 @@ sf_buf_init(void *arg)
vm_offset_t sf_base;
int i;
mtx_init(&sf_freelist.sf_lock, "sf_bufs list lock", NULL, MTX_DEF);
SLIST_INIT(&sf_freelist.sf_head);
sf_buf_active = hashinit(nsfbufs, M_TEMP, &sf_buf_hashmask);
LIST_INIT(&sf_buf_freelist);
sf_base = kmem_alloc_nofault(kernel_map, nsfbufs * PAGE_SIZE);
sf_bufs = malloc(nsfbufs * sizeof(struct sf_buf), M_TEMP,
M_NOWAIT | M_ZERO);
for (i = 0; i < nsfbufs; i++) {
sf_bufs[i].kva = sf_base + i * PAGE_SIZE;
SLIST_INSERT_HEAD(&sf_freelist.sf_head, &sf_bufs[i], free_list);
LIST_INSERT_HEAD(&sf_buf_freelist, &sf_bufs[i], list_entry);
}
sf_buf_alloc_want = 0;
mtx_init(&sf_buf_lock, "sf_buf", NULL, MTX_DEF);
}
/*
@ -593,13 +601,21 @@ sf_buf_init(void *arg)
struct sf_buf *
sf_buf_alloc(struct vm_page *m)
{
struct sf_head *hash_list;
struct sf_buf *sf;
int error;
mtx_lock(&sf_freelist.sf_lock);
while ((sf = SLIST_FIRST(&sf_freelist.sf_head)) == NULL) {
hash_list = &sf_buf_active[SF_BUF_HASH(m)];
mtx_lock(&sf_buf_lock);
LIST_FOREACH(sf, hash_list, list_entry) {
if (sf->m == m) {
sf->ref_count++;
goto done;
}
}
while ((sf = LIST_FIRST(&sf_buf_freelist)) == NULL) {
sf_buf_alloc_want++;
error = msleep(&sf_freelist, &sf_freelist.sf_lock, PVM|PCATCH,
error = msleep(&sf_buf_freelist, &sf_buf_lock, PVM|PCATCH,
"sfbufa", 0);
sf_buf_alloc_want--;
@ -607,14 +623,15 @@ sf_buf_alloc(struct vm_page *m)
* If we got a signal, don't risk going back to sleep.
*/
if (error)
break;
goto done;
}
if (sf != NULL) {
SLIST_REMOVE_HEAD(&sf_freelist.sf_head, free_list);
sf->m = m;
pmap_qenter(sf->kva, &sf->m, 1);
}
mtx_unlock(&sf_freelist.sf_lock);
LIST_REMOVE(sf, list_entry);
LIST_INSERT_HEAD(hash_list, sf, list_entry);
sf->ref_count = 1;
sf->m = m;
pmap_qenter(sf->kva, &sf->m, 1);
done:
mtx_unlock(&sf_buf_lock);
return (sf);
}
@ -628,8 +645,19 @@ sf_buf_free(void *addr, void *args)
struct vm_page *m;
sf = args;
pmap_qremove((vm_offset_t)addr, 1);
mtx_lock(&sf_buf_lock);
m = sf->m;
sf->ref_count--;
if (sf->ref_count == 0) {
pmap_qremove((vm_offset_t)addr, 1);
sf->m = NULL;
LIST_REMOVE(sf, list_entry);
LIST_INSERT_HEAD(&sf_buf_freelist, sf, list_entry);
if (sf_buf_alloc_want > 0)
wakeup_one(&sf_buf_freelist);
}
mtx_unlock(&sf_buf_lock);
vm_page_lock_queues();
vm_page_unwire(m, 0);
/*
@ -640,12 +668,6 @@ sf_buf_free(void *addr, void *args)
if (m->wire_count == 0 && m->object == NULL)
vm_page_free(m);
vm_page_unlock_queues();
sf->m = NULL;
mtx_lock(&sf_freelist.sf_lock);
SLIST_INSERT_HEAD(&sf_freelist.sf_head, sf, free_list);
if (sf_buf_alloc_want > 0)
wakeup_one(&sf_freelist);
mtx_unlock(&sf_freelist.sf_lock);
}
/*

3
sys/i386/include/sf_buf.h
View File

@ -34,9 +34,10 @@
struct vm_page;
struct sf_buf {
SLIST_ENTRY(sf_buf) free_list; /* list of free buffer slots */
LIST_ENTRY(sf_buf) list_entry; /* list of buffers */
struct vm_page *m; /* currently mapped page */
vm_offset_t kva; /* va of mapping */
int ref_count; /* usage of this mapping */
};
static __inline vm_offset_t