Skip to contentSkip to conversation

conntrack: permit updating only a part of the ctmark

PATCH In review Florian Westphal 2 participants 5 messages Jun 15, 2011 netfilter

lore
Share & tools
lore

Florian Westphal

author maintainer
#0 lore raw

cover conntrack: permit updating only a part of the ctmark

Hello Pablo,

here are two patches for conntrack(8).

I want to alter the upper 8 bits of the conntrack mark while keeping the rest intact. At the moment it is only possible to specifiy an absolute value, so i've extended --mark to optionally handle a mask argument. Its possible to either filter by mark (e.g. list conntracks with a particular mark bit set) or to update parts of the mark.

The 2nd patch is not absolutely needed, but it speeds up operation when only a couple of conntracks are affected by the modification.

Thanks,
Florian

author maintainer
#1 lore raw

PATCH 1/2 conntrack: add support for mark mask

Extend --mark option to optionally take a mask, seperated by '/', e.g. --mark 0x80/0xf0.

When used with -L, only test those bits of the mark that are in the mask range (behaves like iptables like -m mark).

When used with -U, zero out those bits indicated by the mask and XOR the new mark into the result (behaves like iptables -j MARK --set-xmark).

---
 conntrack.8     |    8 +++++-
 src/conntrack.c |   68 ++++++++++++++++++++++++++++++++++++++++++++++++++++++-
 2 files changed, 73 insertions(+), 3 deletions(-)
conntrack.8 +6 -2
index 0565907..6525123 100644
@@ -135,8 +135,12 @@ This option is only required in conjunction with "-L, --dump". If this option is
 .BI "-t, --timeout " "TIMEOUT"
 Specify the timeout.
 .TP
-.BI "-m, --mark " "MARK"
-Specify the conntrack mark.
+.BI "-m, --mark " "MARK[/MASK]"
+Specify the conntrack mark.  Optionally, a mask value can be specified.
+In "--update" mode, this mask specifies the bits that should be zeroed before XORing
+the MARK value into the ctmark.
+Otherwise, the mask is logically ANDed with the existing mark before the comparision.
+In "--create" mode, the mask is ignored.
 .TP
 .BI "-c, --secmark " "SECMARK"
 Specify the conntrack selinux security mark.
src/conntrack.c +67 -1

151 lines not on this page · raw message

Signature · 1 line
1.7.3.4
author maintainer
#2 lore raw

PATCH 2/2 conntrack: -U: skip sending conntrack update message if conntrack is unchanged

In reply to Florian Westphal

This speeds up operation when a lot of conntracks exist, but only a few of them have to be altered.

This change is user-visible because the exit message
("%d flow entries have been updated") will now print the number of entries that have been altered instead of the total number of conntracks seen.

---
 src/conntrack.c |    6 ++++++
 1 files changed, 6 insertions(+), 0 deletions(-)
src/conntrack.c +6 -0
index fb133f1..3e1cb11 100644
@@ -1258,6 +1258,12 @@ static int update_cb(enum nf_conntrack_msg_type type,
 	nfct_copy(tmp, obj, NFCT_CP_META);
 	copy_mark(tmp, ct, &tmpl.mark);
 
+	/* do not send NFCT_Q_UPDATE if ct appears unchanged */
+	if (nfct_cmp(tmp, ct, NFCT_CMP_ALL | NFCT_CMP_MASK)) {
+		nfct_destroy(tmp);
+		return NFCT_CB_CONTINUE;
+	}
+
 	res = nfct_query(ith, NFCT_Q_UPDATE, tmp);
 	if (res < 0) {
 		nfct_destroy(tmp);
Signature · 1 line
1.7.3.4
#3 lore raw
In reply to Florian Westphal
Quoted 11 lines · Florian Westphal

Hello Pablo,

here are two patches for conntrack(8).

I want to alter the upper 8 bits of the conntrack mark while keeping the rest intact. At the moment it is only possible to specifiy an absolute value, so i've extended --mark to optionally handle a mask argument. Its possible to either filter by mark (e.g. list conntracks with a particular mark bit set) or to update parts of the mark.

Quoted 2 lines

The 2nd patch is not absolutely needed, but it speeds up operation when only a couple of conntracks are affected by the modification.

I like them both, I've applied them.

Would you send me another patch for the qa/testsuite/ to include some test to check that --mark works with absolute value and masks?

Thank you.

author maintainer
#4 lore raw
Quoted 4 lines · Pablo Neira Ayuso

I like them both, I've applied them.

Would you send me another patch for the qa/testsuite/ to include some test to check that --mark works with absolute value and masks?

Sure, I'll get to this later tonight.

Thanks,
Florian