tcp: Prepare PRR to work with NewReno LossRecovery
ClosedPublic
Actions

Authored by rscheff on Mar 27 2021, 10:26 AM.

Details

Reviewers

rrs
tuexen
jtl
thj
cc
kbowling

Group Reviewers

transport

Commits

rGaf5916594bcb: tcp: Prepare PRR to work with NewReno LossRecovery
rG90cca08e91f5: tcp: Prepare PRR to work with NewReno LossRecovery

Summary

Add proper PRR vnet declarations for consistency.
Also add pointer to tcpopt struct to tcp_do_prr_ack, in preparation
for it to deal with non-SACK window reduction (after loss).

No functional change.

MFC after: 2 weeks

Diff Detail

Repository

rG FreeBSD src repository

Lint

Lint Not Applicable

Unit

Tests Not Applicable

Event Timeline

rscheff created this revision.Mar 27 2021, 10:26 AM

Herald added a reviewer: transport. · View Herald TranscriptMar 27 2021, 10:26 AM

Herald added subscribers: melifaro, imp. · View Herald Transcript

rscheff requested review of this revision.Mar 27 2021, 10:26 AM

Harbormaster completed remote builds in B38107: Diff 86380.Mar 27 2021, 10:26 AM

tuexen accepted this revision.Apr 8 2021, 2:25 PM

This revision is now accepted and ready to land.Apr 8 2021, 2:25 PM

Closed by commit rG90cca08e91f5: tcp: Prepare PRR to work with NewReno LossRecovery (authored by rscheff). · Explain WhyApr 8 2021, 5:18 PM

This revision was automatically updated to reflect the committed changes.

rscheff added a commit: rG90cca08e91f5: tcp: Prepare PRR to work with NewReno LossRecovery.

rscheff added a commit: rGaf5916594bcb: tcp: Prepare PRR to work with NewReno LossRecovery.Apr 22 2021, 6:32 PM

Revision Contents
Changeset List

Path

Size

sys/

netinet/

tcp_input.c

8 lines

tcp_var.h

4 lines

Diff 87053

View Options

sys/netinet/tcp_input.c

Show First 20 Lines • Show All 949 Lines • ▼ Show 20 Lines
}		}

/*		/*
* IP socket option processing.		* IP socket option processing.
*/		*/
int		int
ip_ctloutput(struct socket so, struct sockopt sopt)		ip_ctloutput(struct socket so, struct sockopt sopt)
{		{
struct inpcb *inp = sotoinpcb(so);		struct inpcb *inp;
		struct mbuf *options;
int error, optval;		int error, optval;
		bool rlock = false;
		bool wlock = false;
#ifdef RSS		#ifdef RSS
uint32_t rss_bucket;		uint32_t rss_bucket;
int retval;		int retval;
#endif		#endif

		SOCK_LOCK(so);
		inp = sotoinpcb(so);
		if (inp == NULL) {
		SOCK_UNLOCK(so);
		return ENOTCONN;
		}
		in_pcbref(inp);
		SOCK_UNLOCK(so);

error = optval = 0;		error = optval = 0;
if (sopt->sopt_level != IPPROTO_IP) {		if (sopt->sopt_level != IPPROTO_IP) {
error = EINVAL;		error = EINVAL;

if (sopt->sopt_level == SOL_SOCKET &&		if (sopt->sopt_level == SOL_SOCKET &&
sopt->sopt_dir == SOPT_SET) {		sopt->sopt_dir == SOPT_SET) {
switch (sopt->sopt_name) {		switch (sopt->sopt_name) {
case SO_REUSEADDR:		case SO_REUSEADDR:
INP_WLOCK(inp);		INP_WLOCK(inp);
		wlock = true;
if ((so->so_options & SO_REUSEADDR) != 0)		if ((so->so_options & SO_REUSEADDR) != 0)
inp->inp_flags2 \|= INP_REUSEADDR;		inp->inp_flags2 \|= INP_REUSEADDR;
else		else
inp->inp_flags2 &= ~INP_REUSEADDR;		inp->inp_flags2 &= ~INP_REUSEADDR;
INP_WUNLOCK(inp);
error = 0;		error = 0;
break;		break;
case SO_REUSEPORT:		case SO_REUSEPORT:
INP_WLOCK(inp);		INP_WLOCK(inp);
		wlock = true;
if ((so->so_options & SO_REUSEPORT) != 0)		if ((so->so_options & SO_REUSEPORT) != 0)
inp->inp_flags2 \|= INP_REUSEPORT;		inp->inp_flags2 \|= INP_REUSEPORT;
else		else
inp->inp_flags2 &= ~INP_REUSEPORT;		inp->inp_flags2 &= ~INP_REUSEPORT;
INP_WUNLOCK(inp);
error = 0;		error = 0;
break;		break;
case SO_SETFIB:		case SO_SETFIB:
INP_WLOCK(inp);		INP_WLOCK(inp);
		wlock = true;
inp->inp_inc.inc_fibnum = so->so_fibnum;		inp->inp_inc.inc_fibnum = so->so_fibnum;
INP_WUNLOCK(inp);
error = 0;		error = 0;
break;		break;
case SO_MAX_PACING_RATE:		case SO_MAX_PACING_RATE:
#ifdef RATELIMIT		#ifdef RATELIMIT
INP_WLOCK(inp);		INP_WLOCK(inp);
		wlock = true;
inp->inp_flags2 \|= INP_RATE_LIMIT_CHANGED;		inp->inp_flags2 \|= INP_RATE_LIMIT_CHANGED;
INP_WUNLOCK(inp);
error = 0;		error = 0;
#else		#else
error = EOPNOTSUPP;		error = EOPNOTSUPP;
#endif		#endif
break;		break;
default:		default:
break;		break;
}		}
}		}
		if (!wlock)
		INP_WLOCK(inp);
		if (!in_pcbrele_wlocked(inp))
		INP_WUNLOCK(inp);
return (error);		return (error);
}		}

switch (sopt->sopt_dir) {		switch (sopt->sopt_dir) {
case SOPT_SET:		case SOPT_SET:
switch (sopt->sopt_name) {		switch (sopt->sopt_name) {
case IP_OPTIONS:		case IP_OPTIONS:
#ifdef notyet		#ifdef notyet
Show All 13 Lines	#endif
m->m_len = sopt->sopt_valsize;		m->m_len = sopt->sopt_valsize;
error = sooptcopyin(sopt, mtod(m, char *), m->m_len,		error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
m->m_len);		m->m_len);
if (error) {		if (error) {
m_free(m);		m_free(m);
break;		break;
}		}
INP_WLOCK(inp);		INP_WLOCK(inp);
		wlock = true;
error = ip_pcbopts(inp, sopt->sopt_name, m);		error = ip_pcbopts(inp, sopt->sopt_name, m);
INP_WUNLOCK(inp);		break;
return (error);
}		}

case IP_BINDANY:		case IP_BINDANY:
if (sopt->sopt_td != NULL) {		if (sopt->sopt_td != NULL) {
error = priv_check(sopt->sopt_td,		error = priv_check(sopt->sopt_td,
PRIV_NETINET_BINDANY);		PRIV_NETINET_BINDANY);
if (error)		if (error)
break;		break;
Show All 37 Lines	#endif
if (optval >= 0 && optval <= MAXTTL)		if (optval >= 0 && optval <= MAXTTL)
inp->inp_ip_minttl = optval;		inp->inp_ip_minttl = optval;
else		else
error = EINVAL;		error = EINVAL;
break;		break;

#define OPTSET(bit) do { \		#define OPTSET(bit) do { \
INP_WLOCK(inp); \		INP_WLOCK(inp); \
		wlock = true; \
if (optval) \		if (optval) \
inp->inp_flags \|= bit; \		inp->inp_flags \|= bit; \
else \		else \
inp->inp_flags &= ~bit; \		inp->inp_flags &= ~bit; \
INP_WUNLOCK(inp); \
} while (0)		} while (0)

#define OPTSET2(bit, val) do { \		#define OPTSET2(bit, val) do { \
INP_WLOCK(inp); \		INP_WLOCK(inp); \
		wlock = true; \
if (val) \		if (val) \
inp->inp_flags2 \|= bit; \		inp->inp_flags2 \|= bit; \
else \		else \
inp->inp_flags2 &= ~bit; \		inp->inp_flags2 &= ~bit; \
INP_WUNLOCK(inp); \
} while (0)		} while (0)

case IP_RECVOPTS:		case IP_RECVOPTS:
OPTSET(INP_RECVOPTS);		OPTSET(INP_RECVOPTS);
break;		break;

case IP_RECVRETOPTS:		case IP_RECVRETOPTS:
OPTSET(INP_RECVRETOPTS);		OPTSET(INP_RECVRETOPTS);
▲ Show 20 Lines • Show All 78 Lines • ▼ Show 20 Lines	#undef OPTSET2

case IP_PORTRANGE:		case IP_PORTRANGE:
error = sooptcopyin(sopt, &optval, sizeof optval,		error = sooptcopyin(sopt, &optval, sizeof optval,
sizeof optval);		sizeof optval);
if (error)		if (error)
break;		break;

INP_WLOCK(inp);		INP_WLOCK(inp);
		wlock = true;
switch (optval) {		switch (optval) {
case IP_PORTRANGE_DEFAULT:		case IP_PORTRANGE_DEFAULT:
inp->inp_flags &= ~(INP_LOWPORT);		inp->inp_flags &= ~(INP_LOWPORT);
inp->inp_flags &= ~(INP_HIGHPORT);		inp->inp_flags &= ~(INP_HIGHPORT);
break;		break;

case IP_PORTRANGE_HIGH:		case IP_PORTRANGE_HIGH:
inp->inp_flags &= ~(INP_LOWPORT);		inp->inp_flags &= ~(INP_LOWPORT);
inp->inp_flags \|= INP_HIGHPORT;		inp->inp_flags \|= INP_HIGHPORT;
break;		break;

case IP_PORTRANGE_LOW:		case IP_PORTRANGE_LOW:
inp->inp_flags &= ~(INP_HIGHPORT);		inp->inp_flags &= ~(INP_HIGHPORT);
inp->inp_flags \|= INP_LOWPORT;		inp->inp_flags \|= INP_LOWPORT;
break;		break;

default:		default:
error = EINVAL;		error = EINVAL;
break;		break;
}		}
INP_WUNLOCK(inp);
break;		break;

#if defined(IPSEC) \|\| defined(IPSEC_SUPPORT)		#if defined(IPSEC) \|\| defined(IPSEC_SUPPORT)
case IP_IPSEC_POLICY:		case IP_IPSEC_POLICY:
if (IPSEC_ENABLED(ipv4)) {		if (IPSEC_ENABLED(ipv4)) {
		INP_WLOCK(inp);
		wlock = true;
error = IPSEC_PCBCTL(ipv4, inp, sopt);		error = IPSEC_PCBCTL(ipv4, inp, sopt);
		if (error)
		wlock = false;
break;		break;
}		}
/* FALLTHROUGH */		/* FALLTHROUGH */
#endif /* IPSEC */		#endif /* IPSEC */

default:		default:
error = ENOPROTOOPT;		error = ENOPROTOOPT;
break;		break;
}		}
break;		break;

case SOPT_GET:		case SOPT_GET:
switch (sopt->sopt_name) {		switch (sopt->sopt_name) {
case IP_OPTIONS:		case IP_OPTIONS:
case IP_RETOPTS:		case IP_RETOPTS:
if (inp->inp_options)		if (inp->inp_options) {
		aeUnsubmitted Not Done Inline Actions Is it necessary to acquire wlock for getsockopt? ae: Is it necessary to acquire wlock for getsockopt?
		jason_eggnet.comAuthorUnsubmitted Not Done Inline Actions For some but not all. I'll take a second pass at getsockopt. jason_eggnet.com: For some but not all. I'll take a second pass at getsockopt.
		options = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK\|M_ZERO);
		INP_RLOCK(inp);
		bcopy(inp->inp_options, options, min(inp->inp_options->m_len, sopt->sopt_valsize));
		if (!in_pcbrele_rlocked(inp))
		INP_RUNLOCK(inp);
error = sooptcopyout(sopt,		error = sooptcopyout(sopt,
mtod(inp->inp_options,		mtod(options,
char *),		char *),
inp->inp_options->m_len);		options->m_len);
else		return (error);
		} else
sopt->sopt_valsize = 0;		sopt->sopt_valsize = 0;
break;		break;

case IP_TOS:		case IP_TOS:
case IP_TTL:		case IP_TTL:
case IP_MINTTL:		case IP_MINTTL:
case IP_RECVOPTS:		case IP_RECVOPTS:
case IP_RECVRETOPTS:		case IP_RECVRETOPTS:
▲ Show 20 Lines • Show All 51 Lines • ▼ Show 20 Lines	#define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0)
optval = OPTBIT(INP_RECVTTL);		optval = OPTBIT(INP_RECVTTL);
break;		break;

case IP_RECVIF:		case IP_RECVIF:
optval = OPTBIT(INP_RECVIF);		optval = OPTBIT(INP_RECVIF);
break;		break;

case IP_PORTRANGE:		case IP_PORTRANGE:
		INP_RLOCK(inp);
		rlock = true;
if (inp->inp_flags & INP_HIGHPORT)		if (inp->inp_flags & INP_HIGHPORT)
optval = IP_PORTRANGE_HIGH;		optval = IP_PORTRANGE_HIGH;
else if (inp->inp_flags & INP_LOWPORT)		else if (inp->inp_flags & INP_LOWPORT)
optval = IP_PORTRANGE_LOW;		optval = IP_PORTRANGE_LOW;
else		else
optval = 0;		optval = 0;
break;		break;

Show All 15 Lines	#define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0)
case IP_FLOWTYPE:		case IP_FLOWTYPE:
optval = inp->inp_flowtype;		optval = inp->inp_flowtype;
break;		break;
case IP_RECVFLOWID:		case IP_RECVFLOWID:
optval = OPTBIT2(INP_RECVFLOWID);		optval = OPTBIT2(INP_RECVFLOWID);
break;		break;
#ifdef RSS		#ifdef RSS
case IP_RSSBUCKETID:		case IP_RSSBUCKETID:
		INP_RLOCK(inp);
		rlock = true;
retval = rss_hash2bucket(inp->inp_flowid,		retval = rss_hash2bucket(inp->inp_flowid,
inp->inp_flowtype,		inp->inp_flowtype,
&rss_bucket);		&rss_bucket);
if (retval == 0)		if (retval == 0)
optval = rss_bucket;		optval = rss_bucket;
else		else
error = EINVAL;		error = EINVAL;
break;		break;
case IP_RECVRSSBUCKETID:		case IP_RECVRSSBUCKETID:
optval = OPTBIT2(INP_RECVRSSBUCKETID);		optval = OPTBIT2(INP_RECVRSSBUCKETID);
break;		break;
#endif		#endif
case IP_BINDMULTI:		case IP_BINDMULTI:
optval = OPTBIT2(INP_BINDMULTI);		optval = OPTBIT2(INP_BINDMULTI);
break;		break;
}		}
		if (!rlock)
		INP_RLOCK(inp);
		if (!in_pcbrele_rlocked(inp))
		INP_RUNLOCK(inp);
error = sooptcopyout(sopt, &optval, sizeof optval);		error = sooptcopyout(sopt, &optval, sizeof optval);
break;		return (error);

/*		/*
* Multicast socket options are processed by the in_mcast		* Multicast socket options are processed by the in_mcast
* module.		* module.
*/		*/
case IP_MULTICAST_IF:		case IP_MULTICAST_IF:
case IP_MULTICAST_VIF:		case IP_MULTICAST_VIF:
case IP_MULTICAST_TTL:		case IP_MULTICAST_TTL:
case IP_MULTICAST_LOOP:		case IP_MULTICAST_LOOP:
case IP_MSFILTER:		case IP_MSFILTER:
error = inp_getmoptions(inp, sopt);		error = inp_getmoptions(inp, sopt);
break;		break;

#if defined(IPSEC) \|\| defined(IPSEC_SUPPORT)		#if defined(IPSEC) \|\| defined(IPSEC_SUPPORT)
case IP_IPSEC_POLICY:		case IP_IPSEC_POLICY:
if (IPSEC_ENABLED(ipv4)) {		if (IPSEC_ENABLED(ipv4)) {
		INP_WLOCK(inp);
error = IPSEC_PCBCTL(ipv4, inp, sopt);		error = IPSEC_PCBCTL(ipv4, inp, sopt);
		INP_WUNLOCK(inp);
break;		break;
}		}
/* FALLTHROUGH */		/* FALLTHROUGH */
#endif /* IPSEC */		#endif /* IPSEC */

default:		default:
error = ENOPROTOOPT;		error = ENOPROTOOPT;
break;		break;
}		}
break;		break;
		}
		if (wlock && !in_pcbrele_wlocked(inp))
		INP_WUNLOCK(inp);
		else {
		if (!rlock)
		INP_RLOCK(inp);
		if (!in_pcbrele_rlocked(inp))
		INP_RUNLOCK(inp);
}		}
return (error);		return (error);
}		}

/*		/*
* Routine called from ip_output() to loop back a copy of an IP multicast		* Routine called from ip_output() to loop back a copy of an IP multicast
* packet to the input queue of a specified interface. Note that this		* packet to the input queue of a specified interface. Note that this
* calls the output routine of the loopback "driver", but with an interface		* calls the output routine of the loopback "driver", but with an interface
Show All 35 Lines