Page MenuHomeFreeBSD
Differential D5872 Diff 15642 sys/netinet/tcp_output.c

Changeset View

Standalone View

View Options

sys/netinet/tcp_output.c

Show First 20 Lines • Show All 1,539 Lines • ▼ Show 20 Lines if (((tp->t_flags & TF_FORCEDATA) == 0 ||
tp->snd_nxt -= len; tp->snd_nxt -= len;
} }
SOCKBUF_UNLOCK_ASSERT(&so->so_snd); /* Check gotos. */ SOCKBUF_UNLOCK_ASSERT(&so->so_snd); /* Check gotos. */
switch (error) { switch (error) {
case EPERM: case EPERM:
tp->t_softerror = error; tp->t_softerror = error;
return (error); return (error);
case ENOBUFS: case ENOBUFS:
if (!tcp_timer_active(tp, TT_REXMT) && if ((len > 0 || (flags & (TH_SYN | TH_FIN))) &&
!tcp_timer_active(tp, TT_REXMT) &&
!tcp_timer_active(tp, TT_PERSIST)) !tcp_timer_active(tp, TT_PERSIST))
tcp_timer_activate(tp, TT_REXMT, tp->t_rxtcur); panic("neither rexmt nor persist timer is set");
jtlUnsubmitted
Not Done
Inline Actions

In my opinion, this does not need to be a panic. A KASSERT() should be sufficient.

Also, this is not the re-usable macro which Lawrence suggested.

jtl: In my opinion, this does //not// need to be a panic. A KASSERT() should be sufficient. Also…
sepherosa_gmail.comAuthorUnsubmitted
Not Done
Inline Actions

It's what I am testing w/, since I need non-INVARIANT kernel to generate enough traffic to make ENOBUFS happen.

We can change it into KASSERT, however, I am wondering should we just use "if (__predict_false(...)) panic" here? Missing timeout here can obviously causing issues for data/SYN/FIN.

I will wrap this code segment into a macro (change it into KASSERT is fine w/ me) after more testing.

I think the code segment is fine now; we just need to inverse the logic, if we choose to use KASSERT.

sepherosa_gmail.com: It's what I am testing w/, since I need non-INVARIANT kernel to generate enough traffic to make…
tp->snd_cwnd = tp->t_maxseg; tp->snd_cwnd = tp->t_maxseg;
return (0); return (0);
case EMSGSIZE: case EMSGSIZE:
/* /*
* For some reason the interface we used initially * For some reason the interface we used initially
* to send segments changed to another or lowered * to send segments changed to another or lowered
* its MTU. * its MTU.
jtlUnsubmitted
Not Done
Inline Actions

You already own a lock on tp. No need to do atomic operations.

jtl: You already own a lock on tp. No need to do atomic operations.
* If TSO was active we either got an interface * If TSO was active we either got an interface
* without TSO capabilits or TSO was turned off. * without TSO capabilits or TSO was turned off.
* If we obtained mtu from ip_output() then update * If we obtained mtu from ip_output() then update
* it and try again. * it and try again.
*/ */
if (tso) if (tso)
tp->t_flags &= ~TF_TSO; tp->t_flags &= ~TF_TSO;
if (mtu != 0) { if (mtu != 0) {
▲ Show 20 Lines • Show All 241 Lines • Show Last 20 Lines