Aren
/
NetBSD
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Alter the strategy for reading/writing a bit. First of all, use 

PUFFS_KFLAG_WTCACHE.  Second, create separate fids for reading and
writing.  If opening for read, open a read-only fid and for write
a write-only fid; use these for reading and writing.  When the
open-count for a node drops to zero, clunk both.  This avoids hitting
the fid limit when accessing large directory hierarchies.

Two problems remain:
* does not take credentials into account, although we can only mount
  the remote 9P file server with one set of credentials, so not a
  huge worry
* doesn't work for the open/mmap/close/access_memory_window case, but
  that will require some further kernel changes
This commit is contained in:
pooka 2007-04-22 18:10:48 +00:00
parent 2878c03588
commit 436cfd0df4
3 changed files with 44 additions and 22 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
usr.sbin/puffs/mount_9p/ninepuffs.c
View File

@ -1,4 +1,4 @@
/* $NetBSD: ninepuffs.c,v 1.1 2007/04/21 14:21:43 pooka Exp $ */
/* $NetBSD: ninepuffs.c,v 1.2 2007/04/22 18:10:48 pooka Exp $ */
/*
* Copyright (c) 2007 Antti Kantee. All Rights Reserved.
@ -31,7 +31,7 @@
#include <sys/cdefs.h>
#ifndef lint
__RCSID("$NetBSD: ninepuffs.c,v 1.1 2007/04/21 14:21:43 pooka Exp $");
__RCSID("$NetBSD: ninepuffs.c,v 1.2 2007/04/22 18:10:48 pooka Exp $");
#endif /* !lint */
#include <sys/types.h>
@ -147,6 +147,7 @@ main(int argc, char *argv[])
if (pflags & PUFFS_FLAG_OPDUMP)
detach = 0;
pflags |= PUFFS_KFLAG_WTCACHE;
PUFFSOP_INIT(pops);

5
usr.sbin/puffs/mount_9p/ninepuffs.h
View File

@ -1,4 +1,4 @@
/* $NetBSD: ninepuffs.h,v 1.1 2007/04/21 14:21:43 pooka Exp $ */
/* $NetBSD: ninepuffs.h,v 1.2 2007/04/22 18:10:48 pooka Exp $ */
/*
* Copyright (c) 2007 Antti Kantee. All Rights Reserved.
@ -144,7 +144,8 @@ struct dirfid {
struct p9pnode {
p9pfid_t fid_base;
p9pfid_t fid_open;
p9pfid_t fid_read;
p9pfid_t fid_write;
int opencount;
LIST_HEAD(,dirfid) dir_openlist;

56
usr.sbin/puffs/mount_9p/node.c
View File

@ -1,4 +1,4 @@
/* $NetBSD: node.c,v 1.1 2007/04/21 14:21:44 pooka Exp $ */
/* $NetBSD: node.c,v 1.2 2007/04/22 18:10:48 pooka Exp $ */
/*
* Copyright (c) 2007 Antti Kantee. All Rights Reserved.
@ -27,7 +27,7 @@
#include <sys/cdefs.h>
#ifndef lint
__RCSID("$NetBSD: node.c,v 1.1 2007/04/21 14:21:44 pooka Exp $");
__RCSID("$NetBSD: node.c,v 1.2 2007/04/22 18:10:48 pooka Exp $");
#endif /* !lint */
#include <assert.h>
@ -220,12 +220,11 @@ puffs9p_node_setattr(struct puffs_cc *pcc, void *opc,
* wait until readdir, since it's probable we need to be able to
* open a directory there in any case.
*
* If it's a regular file, open it with full permissions here.
* Let the upper layers of the kernel worry about permission
* control.
* If it's a regular file, open it here with whatever credentials
* we happen to have. Let the upper layers of the kernel worry
* about permission control.
*
* XXX: this doesn't work too well due to us hitting the open file
* limit pretty soon
* XXX: this does not work fully for the mmap case
*/
int
puffs9p_node_open(struct puffs_cc *pcc, void *opc, int mode,
@ -238,13 +237,23 @@ puffs9p_node_open(struct puffs_cc *pcc, void *opc, int mode,
int error = 0;
p9n->opencount++;
if (pn->pn_va.va_type != VDIR && p9n->fid_open == P9P_INVALFID) {
nfid = NEXTFID(p9p);
error = proto_cc_open(pcc, p9n->fid_base, nfid,
P9PROTO_OMODE_RDWR);
if (error)
return error;
p9n->fid_open = nfid;
if (pn->pn_va.va_type != VDIR) {
if (mode & FREAD && p9n->fid_read == P9P_INVALFID) {
nfid = NEXTFID(p9p);
error = proto_cc_open(pcc, p9n->fid_base, nfid,
P9PROTO_OMODE_READ);
if (error)
return error;
p9n->fid_read = nfid;
}
if (mode & FWRITE && p9n->fid_write == P9P_INVALFID) {
nfid = NEXTFID(p9p);
error = proto_cc_open(pcc, p9n->fid_base, nfid,
P9PROTO_OMODE_WRITE);
if (error)
return error;
p9n->fid_write = nfid;
}
}
return 0;
@ -257,9 +266,20 @@ puffs9p_node_close(struct puffs_cc *pcc, void *opc, int flags,
struct puffs_node *pn = opc;
struct p9pnode *p9n = pn->pn_data;
if (--p9n->opencount == 0)
if (pn->pn_va.va_type == VDIR)
if (--p9n->opencount == 0) {
if (pn->pn_va.va_type == VDIR) {
nukealldf(pcc, p9n);
} else {
if (p9n->fid_read != P9P_INVALFID) {
proto_cc_clunkfid(pcc, p9n->fid_read, 0);
p9n->fid_read = P9P_INVALFID;
}
if (p9n->fid_write != P9P_INVALFID) {
proto_cc_clunkfid(pcc, p9n->fid_write, 0);
p9n->fid_write = P9P_INVALFID;
}
}
}
return 0;
}
@ -279,7 +299,7 @@ puffs9p_node_read(struct puffs_cc *pcc, void *opc, uint8_t *buf,
while (*resid > 0) {
p9pbuf_put_1(pb, P9PROTO_T_READ);
p9pbuf_put_2(pb, tag);
p9pbuf_put_4(pb, p9n->fid_open);
p9pbuf_put_4(pb, p9n->fid_read);
p9pbuf_put_8(pb, offset+nread);
p9pbuf_put_4(pb, MIN((uint32_t)*resid,p9p->maxreq-24));
outbuf_enqueue(p9p, pb, pcc, tag);
@ -325,7 +345,7 @@ puffs9p_node_write(struct puffs_cc *pcc, void *opc, uint8_t *buf,
p9pbuf_put_1(pb, P9PROTO_T_WRITE);
p9pbuf_put_2(pb, tag);
p9pbuf_put_4(pb, p9n->fid_open);
p9pbuf_put_4(pb, p9n->fid_write);
p9pbuf_put_8(pb, offset+nwrite);
p9pbuf_put_4(pb, chunk);
p9pbuf_write_data(pb, buf+nwrite, chunk);