aboutsummaryrefslogtreecommitdiffstats
path: root/net
diff options
context:
space:
mode:
authorNeilBrown <neilb@suse.de>2006-10-06 03:44:05 -0400
committerLinus Torvalds <torvalds@g5.osdl.org>2006-10-06 11:53:41 -0400
commitc6b0a9f87b82f25fa35206ec04b5160372eabab4 (patch)
tree849ed55115c95a86a3d164d674c178f9504750a5 /net
parent5842730de179405d80649231faa0b3f254477434 (diff)
[PATCH] knfsd: tidy up up meaning of 'buffer size' in nfsd/sunrpc
There is some confusion about the meaning of 'bufsz' for a sunrpc server. In some cases it is the largest message that can be sent or received. In other cases it is the largest 'payload' that can be included in a NFS message. In either case, it is not possible for both the request and the reply to be this large. One of the request or reply may only be one page long, which fits nicely with NFS. So we remove 'bufsz' and replace it with two numbers: 'max_payload' and 'max_mesg'. Max_payload is the size that the server requests. It is used by the server to check the max size allowed on a particular connection: depending on the protocol a lower limit might be used. max_mesg is the largest single message that can be sent or received. It is calculated as the max_payload, rounded up to a multiple of PAGE_SIZE, and with PAGE_SIZE added to overhead. Only one of the request and reply may be this size. The other must be at most one page. Cc: Greg Banks <gnb@sgi.com> Cc: "J. Bruce Fields" <bfields@fieldses.org> Signed-off-by: Neil Brown <neilb@suse.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Diffstat (limited to 'net')
-rw-r--r--net/sunrpc/svc.c17
-rw-r--r--net/sunrpc/svcsock.c28
2 files changed, 24 insertions, 21 deletions
diff --git a/net/sunrpc/svc.c b/net/sunrpc/svc.c
index c2c8bb20d07f..2807fa0eab40 100644
--- a/net/sunrpc/svc.c
+++ b/net/sunrpc/svc.c
@@ -282,7 +282,10 @@ __svc_create(struct svc_program *prog, unsigned int bufsize, int npools,
282 serv->sv_program = prog; 282 serv->sv_program = prog;
283 serv->sv_nrthreads = 1; 283 serv->sv_nrthreads = 1;
284 serv->sv_stats = prog->pg_stats; 284 serv->sv_stats = prog->pg_stats;
285 serv->sv_bufsz = bufsize? bufsize : 4096; 285 if (bufsize > RPCSVC_MAXPAYLOAD)
286 bufsize = RPCSVC_MAXPAYLOAD;
287 serv->sv_max_payload = bufsize? bufsize : 4096;
288 serv->sv_max_mesg = roundup(serv->sv_max_payload + PAGE_SIZE, PAGE_SIZE);
286 serv->sv_shutdown = shutdown; 289 serv->sv_shutdown = shutdown;
287 xdrsize = 0; 290 xdrsize = 0;
288 while (prog) { 291 while (prog) {
@@ -414,9 +417,9 @@ svc_init_buffer(struct svc_rqst *rqstp, unsigned int size)
414 int pages; 417 int pages;
415 int arghi; 418 int arghi;
416 419
417 if (size > RPCSVC_MAXPAYLOAD) 420 pages = size / PAGE_SIZE + 1; /* extra page as we hold both request and reply.
418 size = RPCSVC_MAXPAYLOAD; 421 * We assume one is at most one page
419 pages = 2 + (size+ PAGE_SIZE -1) / PAGE_SIZE; 422 */
420 arghi = 0; 423 arghi = 0;
421 BUG_ON(pages > RPCSVC_MAXPAGES); 424 BUG_ON(pages > RPCSVC_MAXPAGES);
422 while (pages) { 425 while (pages) {
@@ -463,7 +466,7 @@ __svc_create_thread(svc_thread_fn func, struct svc_serv *serv,
463 466
464 if (!(rqstp->rq_argp = kmalloc(serv->sv_xdrsize, GFP_KERNEL)) 467 if (!(rqstp->rq_argp = kmalloc(serv->sv_xdrsize, GFP_KERNEL))
465 || !(rqstp->rq_resp = kmalloc(serv->sv_xdrsize, GFP_KERNEL)) 468 || !(rqstp->rq_resp = kmalloc(serv->sv_xdrsize, GFP_KERNEL))
466 || !svc_init_buffer(rqstp, serv->sv_bufsz)) 469 || !svc_init_buffer(rqstp, serv->sv_max_mesg))
467 goto out_thread; 470 goto out_thread;
468 471
469 serv->sv_nrthreads++; 472 serv->sv_nrthreads++;
@@ -938,8 +941,8 @@ u32 svc_max_payload(const struct svc_rqst *rqstp)
938 941
939 if (rqstp->rq_sock->sk_sock->type == SOCK_DGRAM) 942 if (rqstp->rq_sock->sk_sock->type == SOCK_DGRAM)
940 max = RPCSVC_MAXPAYLOAD_UDP; 943 max = RPCSVC_MAXPAYLOAD_UDP;
941 if (rqstp->rq_server->sv_bufsz < max) 944 if (rqstp->rq_server->sv_max_payload < max)
942 max = rqstp->rq_server->sv_bufsz; 945 max = rqstp->rq_server->sv_max_payload;
943 return max; 946 return max;
944} 947}
945EXPORT_SYMBOL_GPL(svc_max_payload); 948EXPORT_SYMBOL_GPL(svc_max_payload);
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index b39e7e2b648f..61e307cca13d 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -192,13 +192,13 @@ svc_sock_enqueue(struct svc_sock *svsk)
192 svsk->sk_pool = pool; 192 svsk->sk_pool = pool;
193 193
194 set_bit(SOCK_NOSPACE, &svsk->sk_sock->flags); 194 set_bit(SOCK_NOSPACE, &svsk->sk_sock->flags);
195 if (((atomic_read(&svsk->sk_reserved) + serv->sv_bufsz)*2 195 if (((atomic_read(&svsk->sk_reserved) + serv->sv_max_mesg)*2
196 > svc_sock_wspace(svsk)) 196 > svc_sock_wspace(svsk))
197 && !test_bit(SK_CLOSE, &svsk->sk_flags) 197 && !test_bit(SK_CLOSE, &svsk->sk_flags)
198 && !test_bit(SK_CONN, &svsk->sk_flags)) { 198 && !test_bit(SK_CONN, &svsk->sk_flags)) {
199 /* Don't enqueue while not enough space for reply */ 199 /* Don't enqueue while not enough space for reply */
200 dprintk("svc: socket %p no space, %d*2 > %ld, not enqueued\n", 200 dprintk("svc: socket %p no space, %d*2 > %ld, not enqueued\n",
201 svsk->sk_sk, atomic_read(&svsk->sk_reserved)+serv->sv_bufsz, 201 svsk->sk_sk, atomic_read(&svsk->sk_reserved)+serv->sv_max_mesg,
202 svc_sock_wspace(svsk)); 202 svc_sock_wspace(svsk));
203 svsk->sk_pool = NULL; 203 svsk->sk_pool = NULL;
204 clear_bit(SK_BUSY, &svsk->sk_flags); 204 clear_bit(SK_BUSY, &svsk->sk_flags);
@@ -220,7 +220,7 @@ svc_sock_enqueue(struct svc_sock *svsk)
220 rqstp, rqstp->rq_sock); 220 rqstp, rqstp->rq_sock);
221 rqstp->rq_sock = svsk; 221 rqstp->rq_sock = svsk;
222 atomic_inc(&svsk->sk_inuse); 222 atomic_inc(&svsk->sk_inuse);
223 rqstp->rq_reserved = serv->sv_bufsz; 223 rqstp->rq_reserved = serv->sv_max_mesg;
224 atomic_add(rqstp->rq_reserved, &svsk->sk_reserved); 224 atomic_add(rqstp->rq_reserved, &svsk->sk_reserved);
225 BUG_ON(svsk->sk_pool != pool); 225 BUG_ON(svsk->sk_pool != pool);
226 wake_up(&rqstp->rq_wait); 226 wake_up(&rqstp->rq_wait);
@@ -639,8 +639,8 @@ svc_udp_recvfrom(struct svc_rqst *rqstp)
639 * which will access the socket. 639 * which will access the socket.
640 */ 640 */
641 svc_sock_setbufsize(svsk->sk_sock, 641 svc_sock_setbufsize(svsk->sk_sock,
642 (serv->sv_nrthreads+3) * serv->sv_bufsz, 642 (serv->sv_nrthreads+3) * serv->sv_max_mesg,
643 (serv->sv_nrthreads+3) * serv->sv_bufsz); 643 (serv->sv_nrthreads+3) * serv->sv_max_mesg);
644 644
645 if ((rqstp->rq_deferred = svc_deferred_dequeue(svsk))) { 645 if ((rqstp->rq_deferred = svc_deferred_dequeue(svsk))) {
646 svc_sock_received(svsk); 646 svc_sock_received(svsk);
@@ -749,8 +749,8 @@ svc_udp_init(struct svc_sock *svsk)
749 * svc_udp_recvfrom will re-adjust if necessary 749 * svc_udp_recvfrom will re-adjust if necessary
750 */ 750 */
751 svc_sock_setbufsize(svsk->sk_sock, 751 svc_sock_setbufsize(svsk->sk_sock,
752 3 * svsk->sk_server->sv_bufsz, 752 3 * svsk->sk_server->sv_max_mesg,
753 3 * svsk->sk_server->sv_bufsz); 753 3 * svsk->sk_server->sv_max_mesg);
754 754
755 set_bit(SK_DATA, &svsk->sk_flags); /* might have come in before data_ready set up */ 755 set_bit(SK_DATA, &svsk->sk_flags); /* might have come in before data_ready set up */
756 set_bit(SK_CHNGBUF, &svsk->sk_flags); 756 set_bit(SK_CHNGBUF, &svsk->sk_flags);
@@ -993,8 +993,8 @@ svc_tcp_recvfrom(struct svc_rqst *rqstp)
993 * as soon a a complete request arrives. 993 * as soon a a complete request arrives.
994 */ 994 */
995 svc_sock_setbufsize(svsk->sk_sock, 995 svc_sock_setbufsize(svsk->sk_sock,
996 (serv->sv_nrthreads+3) * serv->sv_bufsz, 996 (serv->sv_nrthreads+3) * serv->sv_max_mesg,
997 3 * serv->sv_bufsz); 997 3 * serv->sv_max_mesg);
998 998
999 clear_bit(SK_DATA, &svsk->sk_flags); 999 clear_bit(SK_DATA, &svsk->sk_flags);
1000 1000
@@ -1032,7 +1032,7 @@ svc_tcp_recvfrom(struct svc_rqst *rqstp)
1032 } 1032 }
1033 svsk->sk_reclen &= 0x7fffffff; 1033 svsk->sk_reclen &= 0x7fffffff;
1034 dprintk("svc: TCP record, %d bytes\n", svsk->sk_reclen); 1034 dprintk("svc: TCP record, %d bytes\n", svsk->sk_reclen);
1035 if (svsk->sk_reclen > serv->sv_bufsz) { 1035 if (svsk->sk_reclen > serv->sv_max_mesg) {
1036 printk(KERN_NOTICE "RPC: bad TCP reclen 0x%08lx (large)\n", 1036 printk(KERN_NOTICE "RPC: bad TCP reclen 0x%08lx (large)\n",
1037 (unsigned long) svsk->sk_reclen); 1037 (unsigned long) svsk->sk_reclen);
1038 goto err_delete; 1038 goto err_delete;
@@ -1171,8 +1171,8 @@ svc_tcp_init(struct svc_sock *svsk)
1171 * svc_tcp_recvfrom will re-adjust if necessary 1171 * svc_tcp_recvfrom will re-adjust if necessary
1172 */ 1172 */
1173 svc_sock_setbufsize(svsk->sk_sock, 1173 svc_sock_setbufsize(svsk->sk_sock,
1174 3 * svsk->sk_server->sv_bufsz, 1174 3 * svsk->sk_server->sv_max_mesg,
1175 3 * svsk->sk_server->sv_bufsz); 1175 3 * svsk->sk_server->sv_max_mesg);
1176 1176
1177 set_bit(SK_CHNGBUF, &svsk->sk_flags); 1177 set_bit(SK_CHNGBUF, &svsk->sk_flags);
1178 set_bit(SK_DATA, &svsk->sk_flags); 1178 set_bit(SK_DATA, &svsk->sk_flags);
@@ -1234,7 +1234,7 @@ svc_recv(struct svc_rqst *rqstp, long timeout)
1234 1234
1235 1235
1236 /* now allocate needed pages. If we get a failure, sleep briefly */ 1236 /* now allocate needed pages. If we get a failure, sleep briefly */
1237 pages = 2 + (serv->sv_bufsz + PAGE_SIZE -1) / PAGE_SIZE; 1237 pages = (serv->sv_max_mesg + PAGE_SIZE) / PAGE_SIZE;
1238 for (i=0; i < pages ; i++) 1238 for (i=0; i < pages ; i++)
1239 while (rqstp->rq_pages[i] == NULL) { 1239 while (rqstp->rq_pages[i] == NULL) {
1240 struct page *p = alloc_page(GFP_KERNEL); 1240 struct page *p = alloc_page(GFP_KERNEL);
@@ -1263,7 +1263,7 @@ svc_recv(struct svc_rqst *rqstp, long timeout)
1263 if ((svsk = svc_sock_dequeue(pool)) != NULL) { 1263 if ((svsk = svc_sock_dequeue(pool)) != NULL) {
1264 rqstp->rq_sock = svsk; 1264 rqstp->rq_sock = svsk;
1265 atomic_inc(&svsk->sk_inuse); 1265 atomic_inc(&svsk->sk_inuse);
1266 rqstp->rq_reserved = serv->sv_bufsz; 1266 rqstp->rq_reserved = serv->sv_max_mesg;
1267 atomic_add(rqstp->rq_reserved, &svsk->sk_reserved); 1267 atomic_add(rqstp->rq_reserved, &svsk->sk_reserved);
1268 } else { 1268 } else {
1269 /* No data pending. Go to sleep */ 1269 /* No data pending. Go to sleep */