2005-04-16 16:20:36 -06:00
|
|
|
/*
|
|
|
|
* Request reply cache. This is currently a global cache, but this may
|
|
|
|
* change in the future and be a per-client cache.
|
|
|
|
*
|
|
|
|
* This code is heavily inspired by the 44BSD implementation, although
|
|
|
|
* it does things a bit differently.
|
|
|
|
*
|
|
|
|
* Copyright (C) 1995, 1996 Olaf Kirch <okir@monad.swb.de>
|
|
|
|
*/
|
|
|
|
|
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 02:04:11 -06:00
|
|
|
#include <linux/slab.h>
|
2016-10-26 05:26:40 -06:00
|
|
|
#include <linux/vmalloc.h>
|
2013-02-04 10:50:00 -07:00
|
|
|
#include <linux/sunrpc/addr.h>
|
2013-02-04 06:18:02 -07:00
|
|
|
#include <linux/highmem.h>
|
2013-03-27 08:15:39 -06:00
|
|
|
#include <linux/log2.h>
|
|
|
|
#include <linux/hash.h>
|
2013-02-04 09:57:27 -07:00
|
|
|
#include <net/checksum.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 02:04:11 -06:00
|
|
|
|
2009-12-03 11:30:56 -07:00
|
|
|
#include "nfsd.h"
|
|
|
|
#include "cache.h"
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2013-02-04 06:18:02 -07:00
|
|
|
#define NFSDDBG_FACILITY NFSDDBG_REPCACHE
|
|
|
|
|
2013-03-27 08:15:39 -06:00
|
|
|
/*
|
|
|
|
* We use this value to determine the number of hash buckets from the max
|
|
|
|
* cache size, the idea being that when the cache is at its maximum number
|
|
|
|
* of entries, then this should be the average number of entries per bucket.
|
|
|
|
*/
|
|
|
|
#define TARGET_BUCKET_SIZE 64
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2014-08-06 11:44:20 -06:00
|
|
|
struct nfsd_drc_bucket {
|
2014-08-06 11:44:21 -06:00
|
|
|
struct list_head lru_head;
|
2014-08-06 11:44:24 -06:00
|
|
|
spinlock_t cache_lock;
|
2014-08-06 11:44:20 -06:00
|
|
|
};
|
|
|
|
|
|
|
|
static struct nfsd_drc_bucket *drc_hashtbl;
|
2013-01-28 12:41:10 -07:00
|
|
|
static struct kmem_cache *drc_slab;
|
2013-03-27 08:15:37 -06:00
|
|
|
|
|
|
|
/* max number of entries allowed in the cache */
|
2013-02-04 06:18:02 -07:00
|
|
|
static unsigned int max_drc_entries;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2013-03-27 08:15:39 -06:00
|
|
|
/* number of significant bits in the hash value */
|
|
|
|
static unsigned int maskbits;
|
2014-08-06 11:44:21 -06:00
|
|
|
static unsigned int drc_hashsize;
|
2013-03-27 08:15:39 -06:00
|
|
|
|
2013-03-27 08:15:37 -06:00
|
|
|
/*
|
|
|
|
* Stats and other tracking of on the duplicate reply cache. All of these and
|
|
|
|
* the "rc" fields in nfsdstats are protected by the cache_lock
|
|
|
|
*/
|
|
|
|
|
|
|
|
/* total number of entries */
|
2014-08-06 11:44:23 -06:00
|
|
|
static atomic_t num_drc_entries;
|
2013-03-27 08:15:37 -06:00
|
|
|
|
|
|
|
/* cache misses due only to checksum comparison failures */
|
|
|
|
static unsigned int payload_misses;
|
|
|
|
|
2013-03-27 08:15:38 -06:00
|
|
|
/* amount of memory (in bytes) currently consumed by the DRC */
|
|
|
|
static unsigned int drc_mem_usage;
|
|
|
|
|
2013-03-27 08:15:39 -06:00
|
|
|
/* longest hash chain seen */
|
|
|
|
static unsigned int longest_chain;
|
|
|
|
|
|
|
|
/* size of cache when we saw the longest hash chain */
|
|
|
|
static unsigned int longest_chain_cachesize;
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
static int nfsd_cache_append(struct svc_rqst *rqstp, struct kvec *vec);
|
2013-08-27 18:18:09 -06:00
|
|
|
static unsigned long nfsd_reply_cache_count(struct shrinker *shrink,
|
|
|
|
struct shrink_control *sc);
|
|
|
|
static unsigned long nfsd_reply_cache_scan(struct shrinker *shrink,
|
|
|
|
struct shrink_control *sc);
|
2013-02-04 06:18:06 -07:00
|
|
|
|
2013-04-05 07:22:39 -06:00
|
|
|
static struct shrinker nfsd_reply_cache_shrinker = {
|
2013-08-27 18:18:09 -06:00
|
|
|
.scan_objects = nfsd_reply_cache_scan,
|
|
|
|
.count_objects = nfsd_reply_cache_count,
|
2013-02-04 06:18:06 -07:00
|
|
|
.seeks = 1,
|
|
|
|
};
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2013-02-04 06:18:02 -07:00
|
|
|
/*
|
|
|
|
* Put a cap on the size of the DRC based on the amount of available
|
|
|
|
* low memory in the machine.
|
|
|
|
*
|
|
|
|
* 64MB: 8192
|
|
|
|
* 128MB: 11585
|
|
|
|
* 256MB: 16384
|
|
|
|
* 512MB: 23170
|
|
|
|
* 1GB: 32768
|
|
|
|
* 2GB: 46340
|
|
|
|
* 4GB: 65536
|
|
|
|
* 8GB: 92681
|
|
|
|
* 16GB: 131072
|
|
|
|
*
|
|
|
|
* ...with a hard cap of 256k entries. In the worst case, each entry will be
|
|
|
|
* ~1k, so the above numbers should give a rough max of the amount of memory
|
|
|
|
* used in k.
|
|
|
|
*/
|
|
|
|
static unsigned int
|
|
|
|
nfsd_cache_size_limit(void)
|
|
|
|
{
|
|
|
|
unsigned int limit;
|
|
|
|
unsigned long low_pages = totalram_pages - totalhigh_pages;
|
|
|
|
|
|
|
|
limit = (16 * int_sqrt(low_pages)) << (PAGE_SHIFT-10);
|
|
|
|
return min_t(unsigned int, limit, 256*1024);
|
|
|
|
}
|
|
|
|
|
2013-03-27 08:15:39 -06:00
|
|
|
/*
|
|
|
|
* Compute the number of hash buckets we need. Divide the max cachesize by
|
|
|
|
* the "target" max bucket size, and round up to next power of two.
|
|
|
|
*/
|
|
|
|
static unsigned int
|
|
|
|
nfsd_hashsize(unsigned int limit)
|
|
|
|
{
|
|
|
|
return roundup_pow_of_two(limit / TARGET_BUCKET_SIZE);
|
|
|
|
}
|
|
|
|
|
2014-08-06 11:44:20 -06:00
|
|
|
static u32
|
|
|
|
nfsd_cache_hash(__be32 xid)
|
|
|
|
{
|
|
|
|
return hash_32(be32_to_cpu(xid), maskbits);
|
|
|
|
}
|
|
|
|
|
2013-01-28 12:41:11 -07:00
|
|
|
static struct svc_cacherep *
|
|
|
|
nfsd_reply_cache_alloc(void)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
|
|
|
struct svc_cacherep *rp;
|
|
|
|
|
2013-01-28 12:41:11 -07:00
|
|
|
rp = kmem_cache_alloc(drc_slab, GFP_KERNEL);
|
|
|
|
if (rp) {
|
2005-04-16 16:20:36 -06:00
|
|
|
rp->c_state = RC_UNUSED;
|
|
|
|
rp->c_type = RC_NOCACHE;
|
2013-01-28 12:41:11 -07:00
|
|
|
INIT_LIST_HEAD(&rp->c_lru);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
2013-01-28 12:41:11 -07:00
|
|
|
return rp;
|
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2013-01-28 12:41:11 -07:00
|
|
|
static void
|
|
|
|
nfsd_reply_cache_free_locked(struct svc_cacherep *rp)
|
|
|
|
{
|
2013-03-27 08:15:38 -06:00
|
|
|
if (rp->c_type == RC_REPLBUFF && rp->c_replvec.iov_base) {
|
|
|
|
drc_mem_usage -= rp->c_replvec.iov_len;
|
2013-01-28 12:41:11 -07:00
|
|
|
kfree(rp->c_replvec.iov_base);
|
2013-03-27 08:15:38 -06:00
|
|
|
}
|
2013-01-28 12:41:11 -07:00
|
|
|
list_del(&rp->c_lru);
|
2014-08-06 11:44:23 -06:00
|
|
|
atomic_dec(&num_drc_entries);
|
2013-03-27 08:15:38 -06:00
|
|
|
drc_mem_usage -= sizeof(*rp);
|
2013-01-28 12:41:11 -07:00
|
|
|
kmem_cache_free(drc_slab, rp);
|
|
|
|
}
|
|
|
|
|
2013-02-04 06:18:04 -07:00
|
|
|
static void
|
2014-08-06 11:44:24 -06:00
|
|
|
nfsd_reply_cache_free(struct nfsd_drc_bucket *b, struct svc_cacherep *rp)
|
2013-02-04 06:18:04 -07:00
|
|
|
{
|
2014-08-06 11:44:24 -06:00
|
|
|
spin_lock(&b->cache_lock);
|
2013-02-04 06:18:04 -07:00
|
|
|
nfsd_reply_cache_free_locked(rp);
|
2014-08-06 11:44:24 -06:00
|
|
|
spin_unlock(&b->cache_lock);
|
2013-02-04 06:18:04 -07:00
|
|
|
}
|
|
|
|
|
2013-01-28 12:41:11 -07:00
|
|
|
int nfsd_reply_cache_init(void)
|
|
|
|
{
|
2013-03-27 08:15:39 -06:00
|
|
|
unsigned int hashsize;
|
2014-08-06 11:44:21 -06:00
|
|
|
unsigned int i;
|
2015-03-19 05:48:31 -06:00
|
|
|
int status = 0;
|
2013-03-27 08:15:39 -06:00
|
|
|
|
2013-03-15 07:16:29 -06:00
|
|
|
max_drc_entries = nfsd_cache_size_limit();
|
2014-08-06 11:44:23 -06:00
|
|
|
atomic_set(&num_drc_entries, 0);
|
2013-03-27 08:15:39 -06:00
|
|
|
hashsize = nfsd_hashsize(max_drc_entries);
|
|
|
|
maskbits = ilog2(hashsize);
|
2013-03-15 07:16:29 -06:00
|
|
|
|
2015-03-19 05:48:31 -06:00
|
|
|
status = register_shrinker(&nfsd_reply_cache_shrinker);
|
|
|
|
if (status)
|
|
|
|
return status;
|
|
|
|
|
2013-01-28 12:41:10 -07:00
|
|
|
drc_slab = kmem_cache_create("nfsd_drc", sizeof(struct svc_cacherep),
|
|
|
|
0, 0, NULL);
|
|
|
|
if (!drc_slab)
|
|
|
|
goto out_nomem;
|
|
|
|
|
2014-08-06 11:44:20 -06:00
|
|
|
drc_hashtbl = kcalloc(hashsize, sizeof(*drc_hashtbl), GFP_KERNEL);
|
2016-10-26 05:26:40 -06:00
|
|
|
if (!drc_hashtbl) {
|
|
|
|
drc_hashtbl = vzalloc(hashsize * sizeof(*drc_hashtbl));
|
|
|
|
if (!drc_hashtbl)
|
|
|
|
goto out_nomem;
|
|
|
|
}
|
|
|
|
|
2014-08-06 11:44:24 -06:00
|
|
|
for (i = 0; i < hashsize; i++) {
|
2014-08-06 11:44:21 -06:00
|
|
|
INIT_LIST_HEAD(&drc_hashtbl[i].lru_head);
|
2014-08-06 11:44:24 -06:00
|
|
|
spin_lock_init(&drc_hashtbl[i].cache_lock);
|
|
|
|
}
|
2014-08-06 11:44:21 -06:00
|
|
|
drc_hashsize = hashsize;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2007-11-09 12:10:56 -07:00
|
|
|
return 0;
|
|
|
|
out_nomem:
|
|
|
|
printk(KERN_ERR "nfsd: failed to allocate reply cache\n");
|
|
|
|
nfsd_reply_cache_shutdown();
|
|
|
|
return -ENOMEM;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2007-11-09 12:10:56 -07:00
|
|
|
void nfsd_reply_cache_shutdown(void)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
|
|
|
struct svc_cacherep *rp;
|
2014-08-06 11:44:21 -06:00
|
|
|
unsigned int i;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2013-02-04 06:18:06 -07:00
|
|
|
unregister_shrinker(&nfsd_reply_cache_shrinker);
|
2013-02-04 06:18:05 -07:00
|
|
|
|
2014-08-06 11:44:21 -06:00
|
|
|
for (i = 0; i < drc_hashsize; i++) {
|
|
|
|
struct list_head *head = &drc_hashtbl[i].lru_head;
|
|
|
|
while (!list_empty(head)) {
|
|
|
|
rp = list_first_entry(head, struct svc_cacherep, c_lru);
|
|
|
|
nfsd_reply_cache_free_locked(rp);
|
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2016-10-26 05:26:40 -06:00
|
|
|
kvfree(drc_hashtbl);
|
2014-08-06 11:44:20 -06:00
|
|
|
drc_hashtbl = NULL;
|
2014-08-06 11:44:21 -06:00
|
|
|
drc_hashsize = 0;
|
2013-01-28 12:41:10 -07:00
|
|
|
|
2015-09-13 06:15:15 -06:00
|
|
|
kmem_cache_destroy(drc_slab);
|
|
|
|
drc_slab = NULL;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2013-02-04 06:18:05 -07:00
|
|
|
* Move cache entry to end of LRU list, and queue the cleaner to run if it's
|
|
|
|
* not already scheduled.
|
2005-04-16 16:20:36 -06:00
|
|
|
*/
|
|
|
|
static void
|
2014-08-06 11:44:21 -06:00
|
|
|
lru_put_end(struct nfsd_drc_bucket *b, struct svc_cacherep *rp)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2013-02-04 06:18:00 -07:00
|
|
|
rp->c_timestamp = jiffies;
|
2014-08-06 11:44:21 -06:00
|
|
|
list_move_tail(&rp->c_lru, &b->lru_head);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2013-08-27 18:18:09 -06:00
|
|
|
static long
|
2014-08-06 11:44:21 -06:00
|
|
|
prune_bucket(struct nfsd_drc_bucket *b)
|
2013-02-04 06:18:05 -07:00
|
|
|
{
|
|
|
|
struct svc_cacherep *rp, *tmp;
|
2013-08-27 18:18:09 -06:00
|
|
|
long freed = 0;
|
2013-02-04 06:18:05 -07:00
|
|
|
|
2014-08-06 11:44:21 -06:00
|
|
|
list_for_each_entry_safe(rp, tmp, &b->lru_head, c_lru) {
|
2014-06-05 07:45:00 -06:00
|
|
|
/*
|
|
|
|
* Don't free entries attached to calls that are still
|
|
|
|
* in-progress, but do keep scanning the list.
|
|
|
|
*/
|
|
|
|
if (rp->c_state == RC_INPROG)
|
|
|
|
continue;
|
2014-08-06 11:44:23 -06:00
|
|
|
if (atomic_read(&num_drc_entries) <= max_drc_entries &&
|
2014-06-05 07:45:00 -06:00
|
|
|
time_before(jiffies, rp->c_timestamp + RC_EXPIRE))
|
2013-02-04 06:18:05 -07:00
|
|
|
break;
|
|
|
|
nfsd_reply_cache_free_locked(rp);
|
2013-08-27 18:18:09 -06:00
|
|
|
freed++;
|
2013-02-04 06:18:05 -07:00
|
|
|
}
|
2014-08-06 11:44:21 -06:00
|
|
|
return freed;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Walk the LRU list and prune off entries that are older than RC_EXPIRE.
|
|
|
|
* Also prune the oldest ones when the total exceeds the max number of entries.
|
|
|
|
*/
|
|
|
|
static long
|
|
|
|
prune_cache_entries(void)
|
|
|
|
{
|
|
|
|
unsigned int i;
|
|
|
|
long freed = 0;
|
|
|
|
|
|
|
|
for (i = 0; i < drc_hashsize; i++) {
|
|
|
|
struct nfsd_drc_bucket *b = &drc_hashtbl[i];
|
|
|
|
|
2014-08-06 11:44:24 -06:00
|
|
|
if (list_empty(&b->lru_head))
|
|
|
|
continue;
|
|
|
|
spin_lock(&b->cache_lock);
|
2014-08-06 11:44:21 -06:00
|
|
|
freed += prune_bucket(b);
|
2014-08-06 11:44:24 -06:00
|
|
|
spin_unlock(&b->cache_lock);
|
2014-08-06 11:44:21 -06:00
|
|
|
}
|
2013-08-27 18:18:09 -06:00
|
|
|
return freed;
|
2013-02-04 06:18:05 -07:00
|
|
|
}
|
|
|
|
|
2013-08-27 18:18:09 -06:00
|
|
|
static unsigned long
|
|
|
|
nfsd_reply_cache_count(struct shrinker *shrink, struct shrink_control *sc)
|
2013-02-04 06:18:06 -07:00
|
|
|
{
|
2014-08-06 11:44:23 -06:00
|
|
|
return atomic_read(&num_drc_entries);
|
2013-02-04 06:18:06 -07:00
|
|
|
}
|
|
|
|
|
2013-08-27 18:18:09 -06:00
|
|
|
static unsigned long
|
|
|
|
nfsd_reply_cache_scan(struct shrinker *shrink, struct shrink_control *sc)
|
|
|
|
{
|
2014-08-06 11:44:24 -06:00
|
|
|
return prune_cache_entries();
|
2013-08-27 18:18:09 -06:00
|
|
|
}
|
2013-02-04 09:57:27 -07:00
|
|
|
/*
|
|
|
|
* Walk an xdr_buf and get a CRC for at most the first RC_CSUMLEN bytes
|
|
|
|
*/
|
|
|
|
static __wsum
|
|
|
|
nfsd_cache_csum(struct svc_rqst *rqstp)
|
|
|
|
{
|
|
|
|
int idx;
|
|
|
|
unsigned int base;
|
|
|
|
__wsum csum;
|
|
|
|
struct xdr_buf *buf = &rqstp->rq_arg;
|
|
|
|
const unsigned char *p = buf->head[0].iov_base;
|
|
|
|
size_t csum_len = min_t(size_t, buf->head[0].iov_len + buf->page_len,
|
|
|
|
RC_CSUMLEN);
|
|
|
|
size_t len = min(buf->head[0].iov_len, csum_len);
|
|
|
|
|
|
|
|
/* rq_arg.head first */
|
|
|
|
csum = csum_partial(p, len, 0);
|
|
|
|
csum_len -= len;
|
|
|
|
|
|
|
|
/* Continue into page array */
|
|
|
|
idx = buf->page_base / PAGE_SIZE;
|
|
|
|
base = buf->page_base & ~PAGE_MASK;
|
|
|
|
while (csum_len) {
|
|
|
|
p = page_address(buf->pages[idx]) + base;
|
2013-02-15 11:36:34 -07:00
|
|
|
len = min_t(size_t, PAGE_SIZE - base, csum_len);
|
2013-02-04 09:57:27 -07:00
|
|
|
csum = csum_partial(p, len, csum);
|
|
|
|
csum_len -= len;
|
|
|
|
base = 0;
|
|
|
|
++idx;
|
|
|
|
}
|
|
|
|
return csum;
|
|
|
|
}
|
|
|
|
|
2013-03-27 08:15:37 -06:00
|
|
|
static bool
|
|
|
|
nfsd_cache_match(struct svc_rqst *rqstp, __wsum csum, struct svc_cacherep *rp)
|
|
|
|
{
|
2014-08-06 11:44:25 -06:00
|
|
|
/* Check RPC XID first */
|
|
|
|
if (rqstp->rq_xid != rp->c_xid)
|
2013-03-27 08:15:37 -06:00
|
|
|
return false;
|
|
|
|
/* compare checksum of NFS data */
|
|
|
|
if (csum != rp->c_csum) {
|
|
|
|
++payload_misses;
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2014-08-06 11:44:25 -06:00
|
|
|
/* Other discriminators */
|
|
|
|
if (rqstp->rq_proc != rp->c_proc ||
|
|
|
|
rqstp->rq_prot != rp->c_prot ||
|
|
|
|
rqstp->rq_vers != rp->c_vers ||
|
|
|
|
rqstp->rq_arg.len != rp->c_len ||
|
|
|
|
!rpc_cmp_addr(svc_addr(rqstp), (struct sockaddr *)&rp->c_addr) ||
|
|
|
|
rpc_get_port(svc_addr(rqstp)) != rpc_get_port((struct sockaddr *)&rp->c_addr))
|
|
|
|
return false;
|
|
|
|
|
2013-03-27 08:15:37 -06:00
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2013-01-28 12:41:14 -07:00
|
|
|
/*
|
|
|
|
* Search the request hash for an entry that matches the given rqstp.
|
|
|
|
* Must be called with cache_lock held. Returns the found entry or
|
|
|
|
* NULL on failure.
|
|
|
|
*/
|
|
|
|
static struct svc_cacherep *
|
2014-08-06 11:44:20 -06:00
|
|
|
nfsd_cache_search(struct nfsd_drc_bucket *b, struct svc_rqst *rqstp,
|
|
|
|
__wsum csum)
|
2013-01-28 12:41:14 -07:00
|
|
|
{
|
2013-03-27 08:15:39 -06:00
|
|
|
struct svc_cacherep *rp, *ret = NULL;
|
2014-08-06 11:44:22 -06:00
|
|
|
struct list_head *rh = &b->lru_head;
|
2013-03-27 08:15:39 -06:00
|
|
|
unsigned int entries = 0;
|
2013-01-28 12:41:14 -07:00
|
|
|
|
2014-08-06 11:44:22 -06:00
|
|
|
list_for_each_entry(rp, rh, c_lru) {
|
2013-03-27 08:15:39 -06:00
|
|
|
++entries;
|
|
|
|
if (nfsd_cache_match(rqstp, csum, rp)) {
|
|
|
|
ret = rp;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* tally hash chain length stats */
|
|
|
|
if (entries > longest_chain) {
|
|
|
|
longest_chain = entries;
|
2014-08-06 11:44:23 -06:00
|
|
|
longest_chain_cachesize = atomic_read(&num_drc_entries);
|
2013-03-27 08:15:39 -06:00
|
|
|
} else if (entries == longest_chain) {
|
|
|
|
/* prefer to keep the smallest cachesize possible here */
|
2014-08-06 11:44:23 -06:00
|
|
|
longest_chain_cachesize = min_t(unsigned int,
|
|
|
|
longest_chain_cachesize,
|
|
|
|
atomic_read(&num_drc_entries));
|
2013-01-28 12:41:14 -07:00
|
|
|
}
|
2013-03-27 08:15:39 -06:00
|
|
|
|
|
|
|
return ret;
|
2013-01-28 12:41:14 -07:00
|
|
|
}
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
/*
|
|
|
|
* Try to find an entry matching the current call in the cache. When none
|
2013-02-14 14:45:13 -07:00
|
|
|
* is found, we try to grab the oldest expired entry off the LRU list. If
|
|
|
|
* a suitable one isn't there, then drop the cache_lock and allocate a
|
|
|
|
* new one, then search again in case one got inserted while this thread
|
|
|
|
* didn't hold the lock.
|
2005-04-16 16:20:36 -06:00
|
|
|
*/
|
|
|
|
int
|
2011-01-24 10:11:02 -07:00
|
|
|
nfsd_cache_lookup(struct svc_rqst *rqstp)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2013-02-04 06:18:02 -07:00
|
|
|
struct svc_cacherep *rp, *found;
|
2006-10-20 00:29:02 -06:00
|
|
|
__be32 xid = rqstp->rq_xid;
|
|
|
|
u32 proto = rqstp->rq_prot,
|
2005-04-16 16:20:36 -06:00
|
|
|
vers = rqstp->rq_vers,
|
|
|
|
proc = rqstp->rq_proc;
|
2013-02-04 09:57:27 -07:00
|
|
|
__wsum csum;
|
2014-08-06 11:44:20 -06:00
|
|
|
u32 hash = nfsd_cache_hash(xid);
|
|
|
|
struct nfsd_drc_bucket *b = &drc_hashtbl[hash];
|
2005-04-16 16:20:36 -06:00
|
|
|
unsigned long age;
|
2011-01-24 10:11:02 -07:00
|
|
|
int type = rqstp->rq_cachetype;
|
2013-03-27 08:15:37 -06:00
|
|
|
int rtn = RC_DOIT;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
rqstp->rq_cacherep = NULL;
|
2013-02-04 06:18:03 -07:00
|
|
|
if (type == RC_NOCACHE) {
|
2005-04-16 16:20:36 -06:00
|
|
|
nfsdstats.rcnocache++;
|
2013-03-27 08:15:37 -06:00
|
|
|
return rtn;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2013-02-04 09:57:27 -07:00
|
|
|
csum = nfsd_cache_csum(rqstp);
|
|
|
|
|
2013-03-27 08:15:37 -06:00
|
|
|
/*
|
|
|
|
* Since the common case is a cache miss followed by an insert,
|
2013-12-05 04:00:51 -07:00
|
|
|
* preallocate an entry.
|
2013-03-27 08:15:37 -06:00
|
|
|
*/
|
2013-02-04 06:18:02 -07:00
|
|
|
rp = nfsd_reply_cache_alloc();
|
2014-08-06 11:44:24 -06:00
|
|
|
spin_lock(&b->cache_lock);
|
2013-03-27 08:15:38 -06:00
|
|
|
if (likely(rp)) {
|
2014-08-06 11:44:23 -06:00
|
|
|
atomic_inc(&num_drc_entries);
|
2013-03-27 08:15:38 -06:00
|
|
|
drc_mem_usage += sizeof(*rp);
|
|
|
|
}
|
2013-02-04 06:18:02 -07:00
|
|
|
|
2013-12-05 04:00:51 -07:00
|
|
|
/* go ahead and prune the cache */
|
2014-08-06 11:44:24 -06:00
|
|
|
prune_bucket(b);
|
2013-12-05 04:00:51 -07:00
|
|
|
|
2014-08-06 11:44:20 -06:00
|
|
|
found = nfsd_cache_search(b, rqstp, csum);
|
2013-02-04 06:18:02 -07:00
|
|
|
if (found) {
|
2013-03-27 08:15:37 -06:00
|
|
|
if (likely(rp))
|
|
|
|
nfsd_reply_cache_free_locked(rp);
|
2013-02-04 06:18:02 -07:00
|
|
|
rp = found;
|
|
|
|
goto found_entry;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2013-03-27 08:15:37 -06:00
|
|
|
if (!rp) {
|
|
|
|
dprintk("nfsd: unable to allocate DRC entry!\n");
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2013-02-04 06:18:02 -07:00
|
|
|
nfsdstats.rcmisses++;
|
2005-04-16 16:20:36 -06:00
|
|
|
rqstp->rq_cacherep = rp;
|
|
|
|
rp->c_state = RC_INPROG;
|
|
|
|
rp->c_xid = xid;
|
|
|
|
rp->c_proc = proc;
|
2013-01-28 12:41:07 -07:00
|
|
|
rpc_copy_addr((struct sockaddr *)&rp->c_addr, svc_addr(rqstp));
|
|
|
|
rpc_set_port((struct sockaddr *)&rp->c_addr, rpc_get_port(svc_addr(rqstp)));
|
2005-04-16 16:20:36 -06:00
|
|
|
rp->c_prot = proto;
|
|
|
|
rp->c_vers = vers;
|
2013-02-04 09:57:27 -07:00
|
|
|
rp->c_len = rqstp->rq_arg.len;
|
|
|
|
rp->c_csum = csum;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2014-08-06 11:44:21 -06:00
|
|
|
lru_put_end(b, rp);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
/* release any buffer */
|
|
|
|
if (rp->c_type == RC_REPLBUFF) {
|
2013-03-27 08:15:38 -06:00
|
|
|
drc_mem_usage -= rp->c_replvec.iov_len;
|
2005-04-16 16:20:36 -06:00
|
|
|
kfree(rp->c_replvec.iov_base);
|
|
|
|
rp->c_replvec.iov_base = NULL;
|
|
|
|
}
|
|
|
|
rp->c_type = RC_NOCACHE;
|
|
|
|
out:
|
2014-08-06 11:44:24 -06:00
|
|
|
spin_unlock(&b->cache_lock);
|
2005-04-16 16:20:36 -06:00
|
|
|
return rtn;
|
|
|
|
|
|
|
|
found_entry:
|
2013-02-04 06:18:02 -07:00
|
|
|
nfsdstats.rchits++;
|
2005-04-16 16:20:36 -06:00
|
|
|
/* We found a matching entry which is either in progress or done. */
|
|
|
|
age = jiffies - rp->c_timestamp;
|
2014-08-06 11:44:21 -06:00
|
|
|
lru_put_end(b, rp);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
rtn = RC_DROPIT;
|
|
|
|
/* Request being processed or excessive rexmits */
|
|
|
|
if (rp->c_state == RC_INPROG || age < RC_DELAY)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
/* From the hall of fame of impractical attacks:
|
|
|
|
* Is this a user who tries to snoop on the cache? */
|
|
|
|
rtn = RC_DOIT;
|
2014-11-19 05:51:14 -07:00
|
|
|
if (!test_bit(RQ_SECURE, &rqstp->rq_flags) && rp->c_secure)
|
2005-04-16 16:20:36 -06:00
|
|
|
goto out;
|
|
|
|
|
|
|
|
/* Compose RPC reply header */
|
|
|
|
switch (rp->c_type) {
|
|
|
|
case RC_NOCACHE:
|
|
|
|
break;
|
|
|
|
case RC_REPLSTAT:
|
|
|
|
svc_putu32(&rqstp->rq_res.head[0], rp->c_replstat);
|
|
|
|
rtn = RC_REPLY;
|
|
|
|
break;
|
|
|
|
case RC_REPLBUFF:
|
|
|
|
if (!nfsd_cache_append(rqstp, &rp->c_replvec))
|
|
|
|
goto out; /* should not happen */
|
|
|
|
rtn = RC_REPLY;
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
printk(KERN_WARNING "nfsd: bad repcache type %d\n", rp->c_type);
|
2013-02-04 06:18:02 -07:00
|
|
|
nfsd_reply_cache_free_locked(rp);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Update a cache entry. This is called from nfsd_dispatch when
|
|
|
|
* the procedure has been executed and the complete reply is in
|
|
|
|
* rqstp->rq_res.
|
|
|
|
*
|
|
|
|
* We're copying around data here rather than swapping buffers because
|
|
|
|
* the toplevel loop requires max-sized buffers, which would be a waste
|
|
|
|
* of memory for a cache with a max reply size of 100 bytes (diropokres).
|
|
|
|
*
|
|
|
|
* If we should start to use different types of cache entries tailored
|
|
|
|
* specifically for attrstat and fh's, we may save even more space.
|
|
|
|
*
|
|
|
|
* Also note that a cachetype of RC_NOCACHE can legally be passed when
|
|
|
|
* nfsd failed to encode a reply that otherwise would have been cached.
|
|
|
|
* In this case, nfsd_cache_update is called with statp == NULL.
|
|
|
|
*/
|
|
|
|
void
|
2006-10-20 00:29:02 -06:00
|
|
|
nfsd_cache_update(struct svc_rqst *rqstp, int cachetype, __be32 *statp)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2013-02-04 06:18:03 -07:00
|
|
|
struct svc_cacherep *rp = rqstp->rq_cacherep;
|
2005-04-16 16:20:36 -06:00
|
|
|
struct kvec *resv = &rqstp->rq_res.head[0], *cachv;
|
2014-08-06 11:44:21 -06:00
|
|
|
u32 hash;
|
|
|
|
struct nfsd_drc_bucket *b;
|
2005-04-16 16:20:36 -06:00
|
|
|
int len;
|
2013-03-27 08:15:38 -06:00
|
|
|
size_t bufsize = 0;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2013-02-04 06:18:03 -07:00
|
|
|
if (!rp)
|
2005-04-16 16:20:36 -06:00
|
|
|
return;
|
|
|
|
|
2014-08-06 11:44:21 -06:00
|
|
|
hash = nfsd_cache_hash(rp->c_xid);
|
|
|
|
b = &drc_hashtbl[hash];
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
len = resv->iov_len - ((char*)statp - (char*)resv->iov_base);
|
|
|
|
len >>= 2;
|
2009-03-31 14:28:13 -06:00
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
/* Don't cache excessive amounts of data and XDR failures */
|
|
|
|
if (!statp || len > (256 >> 2)) {
|
2014-08-06 11:44:24 -06:00
|
|
|
nfsd_reply_cache_free(b, rp);
|
2005-04-16 16:20:36 -06:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
switch (cachetype) {
|
|
|
|
case RC_REPLSTAT:
|
|
|
|
if (len != 1)
|
|
|
|
printk("nfsd: RC_REPLSTAT/reply len %d!\n",len);
|
|
|
|
rp->c_replstat = *statp;
|
|
|
|
break;
|
|
|
|
case RC_REPLBUFF:
|
|
|
|
cachv = &rp->c_replvec;
|
2013-03-27 08:15:38 -06:00
|
|
|
bufsize = len << 2;
|
|
|
|
cachv->iov_base = kmalloc(bufsize, GFP_KERNEL);
|
2005-04-16 16:20:36 -06:00
|
|
|
if (!cachv->iov_base) {
|
2014-08-06 11:44:24 -06:00
|
|
|
nfsd_reply_cache_free(b, rp);
|
2005-04-16 16:20:36 -06:00
|
|
|
return;
|
|
|
|
}
|
2013-03-27 08:15:38 -06:00
|
|
|
cachv->iov_len = bufsize;
|
|
|
|
memcpy(cachv->iov_base, statp, bufsize);
|
2005-04-16 16:20:36 -06:00
|
|
|
break;
|
2013-02-04 06:18:04 -07:00
|
|
|
case RC_NOCACHE:
|
2014-08-06 11:44:24 -06:00
|
|
|
nfsd_reply_cache_free(b, rp);
|
2013-02-04 06:18:04 -07:00
|
|
|
return;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
2014-08-06 11:44:24 -06:00
|
|
|
spin_lock(&b->cache_lock);
|
2013-03-27 08:15:38 -06:00
|
|
|
drc_mem_usage += bufsize;
|
2014-08-06 11:44:21 -06:00
|
|
|
lru_put_end(b, rp);
|
2014-11-19 05:51:14 -07:00
|
|
|
rp->c_secure = test_bit(RQ_SECURE, &rqstp->rq_flags);
|
2005-04-16 16:20:36 -06:00
|
|
|
rp->c_type = cachetype;
|
|
|
|
rp->c_state = RC_DONE;
|
2014-08-06 11:44:24 -06:00
|
|
|
spin_unlock(&b->cache_lock);
|
2005-04-16 16:20:36 -06:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Copy cached reply to current reply buffer. Should always fit.
|
|
|
|
* FIXME as reply is in a page, we should just attach the page, and
|
|
|
|
* keep a refcount....
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
nfsd_cache_append(struct svc_rqst *rqstp, struct kvec *data)
|
|
|
|
{
|
|
|
|
struct kvec *vec = &rqstp->rq_res.head[0];
|
|
|
|
|
|
|
|
if (vec->iov_len + data->iov_len > PAGE_SIZE) {
|
2017-02-27 15:30:02 -07:00
|
|
|
printk(KERN_WARNING "nfsd: cached reply too large (%zd).\n",
|
2005-04-16 16:20:36 -06:00
|
|
|
data->iov_len);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
memcpy((char*)vec->iov_base + vec->iov_len, data->iov_base, data->iov_len);
|
|
|
|
vec->iov_len += data->iov_len;
|
|
|
|
return 1;
|
|
|
|
}
|
2013-03-27 08:15:38 -06:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Note that fields may be added, removed or reordered in the future. Programs
|
|
|
|
* scraping this file for info should test the labels to ensure they're
|
|
|
|
* getting the correct field.
|
|
|
|
*/
|
|
|
|
static int nfsd_reply_cache_stats_show(struct seq_file *m, void *v)
|
|
|
|
{
|
|
|
|
seq_printf(m, "max entries: %u\n", max_drc_entries);
|
2014-08-06 11:44:23 -06:00
|
|
|
seq_printf(m, "num entries: %u\n",
|
|
|
|
atomic_read(&num_drc_entries));
|
2013-03-27 08:15:39 -06:00
|
|
|
seq_printf(m, "hash buckets: %u\n", 1 << maskbits);
|
2013-03-27 08:15:38 -06:00
|
|
|
seq_printf(m, "mem usage: %u\n", drc_mem_usage);
|
|
|
|
seq_printf(m, "cache hits: %u\n", nfsdstats.rchits);
|
|
|
|
seq_printf(m, "cache misses: %u\n", nfsdstats.rcmisses);
|
|
|
|
seq_printf(m, "not cached: %u\n", nfsdstats.rcnocache);
|
|
|
|
seq_printf(m, "payload misses: %u\n", payload_misses);
|
2013-03-27 08:15:39 -06:00
|
|
|
seq_printf(m, "longest chain len: %u\n", longest_chain);
|
|
|
|
seq_printf(m, "cachesize at longest: %u\n", longest_chain_cachesize);
|
2013-03-27 08:15:38 -06:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
int nfsd_reply_cache_stats_open(struct inode *inode, struct file *file)
|
|
|
|
{
|
|
|
|
return single_open(file, nfsd_reply_cache_stats_show, NULL);
|
|
|
|
}
|