net: nfs: add dynamic wait period

This patch tackles the time out problem which leads to break the
boot process, when loading file over nfs. The patch does two things.

First of all, we just ignore messages that arrive with a rpc_id smaller
then the client id. We just interpret this messages as answers to
formaly timed out messages.

Second, when a time out occurs we double the time to wait, so that we
do not stress the server resending the last message.

Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com>
Tested-by: Enric Balletbo i Serra <eballetbo@gmail.com>
This commit is contained in:
Matthias Brugger 2012-12-11 19:14:16 +01:00 committed by Joe Hershberger
parent fb4879b3c7
commit fa84fa708c

View file

@ -37,10 +37,14 @@
# define NFS_TIMEOUT CONFIG_NFS_TIMEOUT
#endif
#define NFS_RPC_ERR 1
#define NFS_RPC_DROP 124
static int fs_mounted;
static unsigned long rpc_id;
static int nfs_offset = -1;
static int nfs_len;
static ulong nfs_timeout = NFS_TIMEOUT;
static char dirfh[NFS_FHSIZE]; /* file handle of directory */
static char filefh[NFS_FHSIZE]; /* file handle of kernel image */
@ -399,8 +403,10 @@ rpc_lookup_reply(int prog, uchar *pkt, unsigned len)
debug("%s\n", __func__);
if (ntohl(rpc_pkt.u.reply.id) != rpc_id)
return -1;
if (ntohl(rpc_pkt.u.reply.id) > rpc_id)
return -NFS_RPC_ERR;
else if (ntohl(rpc_pkt.u.reply.id) < rpc_id)
return -NFS_RPC_DROP;
if (rpc_pkt.u.reply.rstatus ||
rpc_pkt.u.reply.verifier ||
@ -428,8 +434,10 @@ nfs_mount_reply(uchar *pkt, unsigned len)
memcpy((unsigned char *)&rpc_pkt, pkt, len);
if (ntohl(rpc_pkt.u.reply.id) != rpc_id)
return -1;
if (ntohl(rpc_pkt.u.reply.id) > rpc_id)
return -NFS_RPC_ERR;
else if (ntohl(rpc_pkt.u.reply.id) < rpc_id)
return -NFS_RPC_DROP;
if (rpc_pkt.u.reply.rstatus ||
rpc_pkt.u.reply.verifier ||
@ -452,8 +460,10 @@ nfs_umountall_reply(uchar *pkt, unsigned len)
memcpy((unsigned char *)&rpc_pkt, pkt, len);
if (ntohl(rpc_pkt.u.reply.id) != rpc_id)
return -1;
if (ntohl(rpc_pkt.u.reply.id) > rpc_id)
return -NFS_RPC_ERR;
else if (ntohl(rpc_pkt.u.reply.id) < rpc_id)
return -NFS_RPC_DROP;
if (rpc_pkt.u.reply.rstatus ||
rpc_pkt.u.reply.verifier ||
@ -475,8 +485,10 @@ nfs_lookup_reply(uchar *pkt, unsigned len)
memcpy((unsigned char *)&rpc_pkt, pkt, len);
if (ntohl(rpc_pkt.u.reply.id) != rpc_id)
return -1;
if (ntohl(rpc_pkt.u.reply.id) > rpc_id)
return -NFS_RPC_ERR;
else if (ntohl(rpc_pkt.u.reply.id) < rpc_id)
return -NFS_RPC_DROP;
if (rpc_pkt.u.reply.rstatus ||
rpc_pkt.u.reply.verifier ||
@ -499,8 +511,10 @@ nfs_readlink_reply(uchar *pkt, unsigned len)
memcpy((unsigned char *)&rpc_pkt, pkt, len);
if (ntohl(rpc_pkt.u.reply.id) != rpc_id)
return -1;
if (ntohl(rpc_pkt.u.reply.id) > rpc_id)
return -NFS_RPC_ERR;
else if (ntohl(rpc_pkt.u.reply.id) < rpc_id)
return -NFS_RPC_DROP;
if (rpc_pkt.u.reply.rstatus ||
rpc_pkt.u.reply.verifier ||
@ -534,8 +548,10 @@ nfs_read_reply(uchar *pkt, unsigned len)
memcpy((uchar *)&rpc_pkt, pkt, sizeof(rpc_pkt.u.reply));
if (ntohl(rpc_pkt.u.reply.id) != rpc_id)
return -1;
if (ntohl(rpc_pkt.u.reply.id) > rpc_id)
return -NFS_RPC_ERR;
else if (ntohl(rpc_pkt.u.reply.id) < rpc_id)
return -NFS_RPC_DROP;
if (rpc_pkt.u.reply.rstatus ||
rpc_pkt.u.reply.verifier ||
@ -574,7 +590,8 @@ NfsTimeout(void)
NetStartAgain();
} else {
puts("T ");
NetSetTimeout(NFS_TIMEOUT, NfsTimeout);
NetSetTimeout(nfs_timeout + NFS_TIMEOUT * NfsTimeoutCount,
NfsTimeout);
NfsSend();
}
}
@ -583,6 +600,7 @@ static void
NfsHandler(uchar *pkt, unsigned dest, IPaddr_t sip, unsigned src, unsigned len)
{
int rlen;
int reply;
debug("%s\n", __func__);
@ -591,19 +609,24 @@ NfsHandler(uchar *pkt, unsigned dest, IPaddr_t sip, unsigned src, unsigned len)
switch (NfsState) {
case STATE_PRCLOOKUP_PROG_MOUNT_REQ:
rpc_lookup_reply(PROG_MOUNT, pkt, len);
if (rpc_lookup_reply(PROG_MOUNT, pkt, len) == -NFS_RPC_DROP)
break;
NfsState = STATE_PRCLOOKUP_PROG_NFS_REQ;
NfsSend();
break;
case STATE_PRCLOOKUP_PROG_NFS_REQ:
rpc_lookup_reply(PROG_NFS, pkt, len);
if (rpc_lookup_reply(PROG_NFS, pkt, len) == -NFS_RPC_DROP)
break;
NfsState = STATE_MOUNT_REQ;
NfsSend();
break;
case STATE_MOUNT_REQ:
if (nfs_mount_reply(pkt, len)) {
reply = nfs_mount_reply(pkt, len);
if (reply == -NFS_RPC_DROP)
break;
else if (reply == -NFS_RPC_ERR) {
puts("*** ERROR: Cannot mount\n");
/* just to be sure... */
NfsState = STATE_UMOUNT_REQ;
@ -615,7 +638,10 @@ NfsHandler(uchar *pkt, unsigned dest, IPaddr_t sip, unsigned src, unsigned len)
break;
case STATE_UMOUNT_REQ:
if (nfs_umountall_reply(pkt, len)) {
reply = nfs_umountall_reply(pkt, len);
if (reply == -NFS_RPC_DROP)
break;
else if (reply == -NFS_RPC_ERR) {
puts("*** ERROR: Cannot umount\n");
net_set_state(NETLOOP_FAIL);
} else {
@ -625,7 +651,10 @@ NfsHandler(uchar *pkt, unsigned dest, IPaddr_t sip, unsigned src, unsigned len)
break;
case STATE_LOOKUP_REQ:
if (nfs_lookup_reply(pkt, len)) {
reply = nfs_lookup_reply(pkt, len);
if (reply == -NFS_RPC_DROP)
break;
else if (reply == -NFS_RPC_ERR) {
puts("*** ERROR: File lookup fail\n");
NfsState = STATE_UMOUNT_REQ;
NfsSend();
@ -638,7 +667,10 @@ NfsHandler(uchar *pkt, unsigned dest, IPaddr_t sip, unsigned src, unsigned len)
break;
case STATE_READLINK_REQ:
if (nfs_readlink_reply(pkt, len)) {
reply = nfs_readlink_reply(pkt, len);
if (reply == -NFS_RPC_DROP)
break;
else if (reply == -NFS_RPC_ERR) {
puts("*** ERROR: Symlink fail\n");
NfsState = STATE_UMOUNT_REQ;
NfsSend();
@ -654,7 +686,7 @@ NfsHandler(uchar *pkt, unsigned dest, IPaddr_t sip, unsigned src, unsigned len)
case STATE_READ_REQ:
rlen = nfs_read_reply(pkt, len);
NetSetTimeout(NFS_TIMEOUT, NfsTimeout);
NetSetTimeout(nfs_timeout, NfsTimeout);
if (rlen > 0) {
nfs_offset += rlen;
NfsSend();
@ -738,7 +770,7 @@ NfsStart(void)
printf("\nLoad address: 0x%lx\n"
"Loading: *\b", load_addr);
NetSetTimeout(NFS_TIMEOUT, NfsTimeout);
NetSetTimeout(nfs_timeout, NfsTimeout);
net_set_udp_handler(NfsHandler);
NfsTimeoutCount = 0;