freebsd-skq/sys/netncp/ncp_rq.c
Jeff Roberson ace8398da0 Refactor select to reduce contention and hide internal implementation
details from consumers.

 - Track individual selecters on a per-descriptor basis such that there
   are no longer collisions and after sleeping for events only those
   descriptors which triggered events must be rescaned.
 - Protect the selinfo (per descriptor) structure with a mtx pool mutex.
   mtx pool mutexes were chosen to preserve api compatibility with
   existing code which does nothing but bzero() to setup selinfo
   structures.
 - Use a per-thread wait channel rather than a global wait channel.
 - Hide select implementation details in a seltd structure which is
   opaque to the rest of the kernel.
 - Provide a 'selsocket' interface for those kernel consumers who wish to
   select on a socket when they have no fd so they no longer have to
   be aware of select implementation details.

Tested by:	kris
Reviewed on:	arch
2007-12-16 06:21:20 +00:00

487 lines
12 KiB
C

/*-
* Copyright (c) 1999-2001 Boris Popov
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by Boris Popov.
* 4. Neither the name of the author nor the names of any co-contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* Routines to prepare request and fetch reply
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/errno.h>
#include <sys/kernel.h>
#include <sys/malloc.h>
#include <sys/mbuf.h>
#include <sys/poll.h>
#include <sys/proc.h>
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/uio.h>
#include <netncp/ncp.h>
#include <netncp/ncp_conn.h>
#include <netncp/ncp_rq.h>
#include <netncp/ncp_subr.h>
#include <netncp/ncp_ncp.h>
#include <netncp/ncp_sock.h>
#include <netncp/ncp_nls.h>
static MALLOC_DEFINE(M_NCPRQ, "NCPRQ", "NCP request");
static int ncp_sign_packet(struct ncp_conn *conn, struct ncp_rq *rqp, int *size);
int
ncp_rq_alloc_any(u_int32_t ptype, u_int8_t fn, struct ncp_conn *ncp,
struct thread *td, struct ucred *cred,
struct ncp_rq **rqpp)
{
struct ncp_rq *rqp;
int error;
MALLOC(rqp, struct ncp_rq *, sizeof(*rqp), M_NCPRQ, M_WAITOK);
error = ncp_rq_init_any(rqp, ptype, fn, ncp, td, cred);
rqp->nr_flags |= NCPR_ALLOCED;
if (error) {
ncp_rq_done(rqp);
return error;
}
*rqpp = rqp;
return 0;
}
int
ncp_rq_alloc(u_int8_t fn, struct ncp_conn *ncp,
struct thread *td, struct ucred *cred, struct ncp_rq **rqpp)
{
return ncp_rq_alloc_any(NCP_REQUEST, fn, ncp, td, cred, rqpp);
}
int
ncp_rq_alloc_subfn(u_int8_t fn, u_int8_t subfn, struct ncp_conn *ncp,
struct thread *td, struct ucred *cred, struct ncp_rq **rqpp)
{
struct ncp_rq *rqp;
int error;
error = ncp_rq_alloc_any(NCP_REQUEST, fn, ncp, td, cred, &rqp);
if (error)
return error;
mb_reserve(&rqp->rq, 2);
mb_put_uint8(&rqp->rq, subfn);
*rqpp = rqp;
return 0;
}
int
ncp_rq_init_any(struct ncp_rq *rqp, u_int32_t ptype, u_int8_t fn,
struct ncp_conn *ncp,
struct thread *td, struct ucred *cred)
{
struct ncp_rqhdr *rq;
struct ncp_bursthdr *brq;
struct mbchain *mbp;
int error;
bzero(rqp, sizeof(*rqp));
error = ncp_conn_access(ncp, cred, NCPM_EXECUTE);
if (error)
return error;
rqp->nr_td = td;
rqp->nr_cred = cred;
rqp->nr_conn = ncp;
mbp = &rqp->rq;
if (mb_init(mbp) != 0)
return ENOBUFS;
switch(ptype) {
case NCP_PACKET_BURST:
brq = (struct ncp_bursthdr*)mb_reserve(mbp, sizeof(*brq));
brq->bh_type = ptype;
brq->bh_streamtype = 0x2;
break;
default:
rq = (struct ncp_rqhdr*)mb_reserve(mbp, sizeof(*rq));
rq->type = ptype;
rq->seq = 0; /* filled later */
rq->fn = fn;
break;
}
rqp->nr_minrplen = -1;
return 0;
}
void
ncp_rq_done(struct ncp_rq *rqp)
{
mb_done(&rqp->rq);
md_done(&rqp->rp);
if (rqp->nr_flags & NCPR_ALLOCED)
free(rqp, M_NCPRQ);
return;
}
/*
* Routines to fill the request
*/
static int
ncp_rq_pathstrhelp(struct mbchain *mbp, c_caddr_t src, caddr_t dst,
size_t *srclen, size_t *dstlen)
{
int len;
if (*srclen < *dstlen) {
*dstlen = *srclen;
len = (int)*srclen;
} else {
*srclen = *dstlen;
len = (int)*dstlen;
}
ncp_pathcopy(src, dst, len, mbp->mb_udata);
return 0;
}
int
ncp_rq_pathstring(struct ncp_rq *rqp, int size, const char *name,
struct ncp_nlstables *nt)
{
struct mbchain *mbp = &rqp->rq;
mb_put_uint8(mbp, size);
mbp->mb_copy = ncp_rq_pathstrhelp;
mbp->mb_udata = nt;
return mb_put_mem(mbp, (c_caddr_t)name, size, MB_MCUSTOM);
}
int
ncp_rq_pstring(struct ncp_rq *rqp, const char *s)
{
u_int len = strlen(s);
int error;
if (len > 255)
return EINVAL;
error = mb_put_uint8(&rqp->rq, len);
if (error)
return error;
return mb_put_mem(&rqp->rq, s, len, MB_MSYSTEM);
}
int
ncp_rq_dbase_path(struct ncp_rq *rqp, u_int8_t vol_num, u_int32_t dir_base,
int namelen, u_char *path, struct ncp_nlstables *nt)
{
struct mbchain *mbp = &rqp->rq;
int complen;
mb_put_uint8(mbp, vol_num);
mb_put_mem(mbp, (c_caddr_t)&dir_base, sizeof(dir_base), MB_MSYSTEM);
mb_put_uint8(mbp, 1); /* with dirbase */
if (path != NULL && path[0]) {
if (namelen < 0) {
namelen = *path++;
mb_put_uint8(mbp, namelen);
for(; namelen; namelen--) {
complen = *path++;
mb_put_uint8(mbp, complen);
mb_put_mem(mbp, path, complen, MB_MSYSTEM);
path += complen;
}
} else {
mb_put_uint8(mbp, 1); /* 1 component */
ncp_rq_pathstring(rqp, namelen, path, nt);
}
} else {
mb_put_uint8(mbp, 0);
mb_put_uint8(mbp, 0);
}
return 0;
}
/*
* Make a signature for the current packet and add it at the end of the
* packet.
*/
static int
ncp_sign_packet(struct ncp_conn *conn, struct ncp_rq *rqp, int *size)
{
u_char data[64];
int error;
bzero(data, sizeof(data));
bcopy(conn->sign_root, data, 8);
setdle(data, 8, *size);
m_copydata(rqp->rq.mb_top, sizeof(struct ncp_rqhdr) - 1,
min((*size) - sizeof(struct ncp_rqhdr)+1, 52), data + 12);
ncp_sign(conn->sign_state, data, conn->sign_state);
error = mb_put_mem(&rqp->rq, (caddr_t)conn->sign_state, 8, MB_MSYSTEM);
if (error)
return error;
(*size) += 8;
return 0;
}
/*
* Low level send rpc, here we do not attempt to restore any connection,
* Connection expected to be locked
*/
int
ncp_request_int(struct ncp_rq *rqp)
{
struct ncp_conn *conn = rqp->nr_conn;
struct thread *td = conn->td;
struct socket *so = conn->ncp_so;
struct ncp_rqhdr *rq;
struct ncp_rphdr *rp=NULL;
struct timeval tv;
struct mbuf *m, *mreply = NULL;
struct mbchain *mbp;
int error, len, dosend, plen = 0, gotpacket;
if (so == NULL) {
printf("%s: ncp_so is NULL !\n",__func__);
ncp_conn_invalidate(conn);
return ENOTCONN;
}
if (td == NULL)
td = curthread; /* XXX maybe procpage ? */
/*
* Flush out replies on previous reqs
*/
tv.tv_sec = 0;
tv.tv_usec = 0;
while (selsocket(so, POLLIN, &tv, td) == 0) {
if (ncp_sock_recv(so, &m, &len) != 0)
break;
m_freem(m);
}
mbp = &rqp->rq;
len = mb_fixhdr(mbp);
rq = mtod(mbp->mb_top, struct ncp_rqhdr *);
rq->seq = conn->seq;
m = rqp->rq.mb_top;
switch (rq->fn) {
case 0x15: case 0x16: case 0x17: case 0x23:
*(u_int16_t*)(rq + 1) = htons(len - 2 - sizeof(*rq));
break;
}
if (conn->flags & NCPFL_SIGNACTIVE) {
error = ncp_sign_packet(conn, rqp, &len);
if (error)
return error;
mbp->mb_top->m_pkthdr.len = len;
}
rq->conn_low = conn->connid & 0xff;
/* rq->task = p->p_pgrp->pg_id & 0xff; */ /*p->p_pid*/
/* XXX: this is temporary fix till I find a better solution */
rq->task = rq->conn_low;
rq->conn_high = conn->connid >> 8;
rqp->rexmit = conn->li.retry_count;
error = 0;
for(dosend = 1;;) {
if (rqp->rexmit-- == 0) {
error = ETIMEDOUT;
break;
}
error = 0;
if (dosend) {
NCPSDEBUG("send:%04x f=%02x c=%d l=%d s=%d t=%d\n",rq->type, rq->fn, (rq->conn_high << 8) + rq->conn_low,
mbp->mb_top->m_pkthdr.len, rq->seq, rq->task
);
error = ncp_sock_send(so, mbp->mb_top, rqp);
if (error)
break;
}
tv.tv_sec = conn->li.timeout;
tv.tv_usec = 0;
error = selsocket(so, POLLIN, &tv, td);
if (error == EWOULDBLOCK ) /* timeout expired */
continue;
error = ncp_chkintr(conn, td);
if (error)
break;
/*
* At this point it is possible to get more than one
* reply from server. In general, last reply should be for
* current request, but not always. So, we loop through
* all replies to find the right answer and flush others.
*/
gotpacket = 0; /* nothing good found */
dosend = 1; /* resend rq if error */
for (;;) {
error = 0;
tv.tv_sec = 0;
tv.tv_usec = 0;
if (selsocket(so, POLLIN, &tv, td) != 0)
break;
/* if (so->so_rcv.sb_cc == 0) {
break;
}*/
error = ncp_sock_recv(so, &m, &len);
if (error)
break; /* must be more checks !!! */
if (m->m_len < sizeof(*rp)) {
m = m_pullup(m, sizeof(*rp));
if (m == NULL) {
printf("%s: reply too short\n",__func__);
continue;
}
}
rp = mtod(m, struct ncp_rphdr*);
if (len == sizeof(*rp) && rp->type == NCP_POSITIVE_ACK) {
NCPSDEBUG("got positive acknowledge\n");
m_freem(m);
rqp->rexmit = conn->li.retry_count;
dosend = 0; /* server just busy and will reply ASAP */
continue;
}
NCPSDEBUG("recv:%04x c=%d l=%d s=%d t=%d cc=%02x cs=%02x\n",rp->type,
(rp->conn_high << 8) + rp->conn_low, len, rp->seq, rp->task,
rp->completion_code, rp->connection_state);
NCPDDEBUG(m);
if ( (rp->type == NCP_REPLY) &&
((rq->type == NCP_ALLOC_SLOT) ||
((rp->conn_low == rq->conn_low) &&
(rp->conn_high == rq->conn_high)
))) {
if (rq->seq > rp->seq || (rq->seq == 0 && rp->seq == 0xff)) {
dosend = 1;
}
if (rp->seq == rq->seq) {
if (gotpacket) {
m_freem(m);
} else {
gotpacket = 1;
mreply = m;
plen = len;
}
continue; /* look up other for other packets */
}
}
m_freem(m);
NCPSDEBUG("reply mismatch\n");
} /* for receive */
if (error || gotpacket)
break;
/* try to resend, or just wait */
}
conn->seq++;
if (error) {
NCPSDEBUG("error=%d\n", error);
/*
* Any error except interruped call means that we have
* to reconnect. So, eliminate future timeouts by invalidating
* connection now.
*/
if (error != EINTR)
ncp_conn_invalidate(conn);
return (error);
}
if (conn->flags & NCPFL_SIGNACTIVE) {
/* XXX: check reply signature */
m_adj(mreply, -8);
plen -= 8;
}
rp = mtod(mreply, struct ncp_rphdr*);
md_initm(&rqp->rp, mreply);
rqp->nr_rpsize = plen - sizeof(*rp);
rqp->nr_cc = error = rp->completion_code;
if (error)
error |= 0x8900; /* server error */
rqp->nr_cs = rp->connection_state;
if (rqp->nr_cs & (NCP_CS_BAD_CONN | NCP_CS_SERVER_DOWN)) {
NCPSDEBUG("server drop us\n");
ncp_conn_invalidate(conn);
error = ECONNRESET;
}
md_get_mem(&rqp->rp, NULL, sizeof(*rp), MB_MSYSTEM);
return error;
}
/*
* Here we will try to restore any loggedin & dropped connection,
* connection should be locked on entry
*/
static __inline int
ncp_restore_login(struct ncp_conn *conn)
{
int error;
printf("ncprq: Restoring connection, flags = %x\n", conn->flags);
conn->flags |= NCPFL_RESTORING;
error = ncp_conn_reconnect(conn);
if (!error && (conn->flags & NCPFL_WASLOGGED))
error = ncp_conn_login(conn, conn->td, conn->ucred);
if (error)
ncp_ncp_disconnect(conn);
conn->flags &= ~NCPFL_RESTORING;
return error;
}
int
ncp_request(struct ncp_rq *rqp)
{
struct ncp_conn *ncp = rqp->nr_conn;
int error, rcnt;
error = ncp_conn_lock(ncp, rqp->nr_td, rqp->nr_cred, NCPM_EXECUTE);
if (error)
goto out;
rcnt = NCP_RESTORE_COUNT;
for(;;) {
if (ncp->flags & NCPFL_ATTACHED) {
error = ncp_request_int(rqp);
if (ncp->flags & NCPFL_ATTACHED)
break;
}
if (rcnt-- == 0) {
error = ECONNRESET;
break;
}
/*
* Do not attempt to restore connection recursively
*/
if (ncp->flags & NCPFL_RESTORING) {
error = ENOTCONN;
break;
}
error = ncp_restore_login(ncp);
if (error)
continue;
}
ncp_conn_unlock(ncp, rqp->nr_td);
out:
if (error && (rqp->nr_flags & NCPR_DONTFREEONERR) == 0)
ncp_rq_done(rqp);
return error;
}