Capability to redirect iptable log, in the right containerized
syslog_ns, is now fully working.
Tried my best to have a very minimal foot-print.
BEWARE: patch not done for IPV6 (I can't test it),
you are welcome to do it.
Signed-off-by: Jean-Marc Pigeon <[email protected]>
---
include/linux/netdevice.h | 2 +
include/linux/syslog.h | 8 +++
kernel/syslog.c | 140 +++++++++++++++++++++++++++++++++++++++---
net/core/rtnetlink.c | 3 +
net/ipv4/netfilter/ipt_LOG.c | 11 +++
5 files changed, 154 insertions(+), 10 deletions(-)
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index a3fccc8..c323463 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -918,6 +918,8 @@ struct net_device {
/* Network namespace this network device is inside */
struct net *nd_net;
#endif
+ /* to assign a syslog chanel according device ownership */
+ pid_t syslog_ns_pid;
/* mid-layer private */
void *ml_priv;
diff --git a/include/linux/syslog.h b/include/linux/syslog.h
index 9fbe190..2188f09 100644
--- a/include/linux/syslog.h
+++ b/include/linux/syslog.h
@@ -44,15 +44,23 @@ static inline void put_syslog_ns(struct syslog_ns *ns)
#ifdef CONFIG_PRINTK
extern struct syslog_ns *current_syslog_ns(void);
extern struct syslog_ns *switch_syslog_ns(struct syslog_ns *syslog_ns);
+extern struct syslog_ns *find_syslog_ns_bypid(pid_t pid);
#else
static inline syslog_ns *current_syslog_ns(void)
{
return 0;
}
+
static inline struct syslog_ns *switch_syslog_ns(struct syslog_ns *syslog_ns)
{
return syslog_ns;
}
+
+static inline syslog_ns *find_syslog_ns_bypid(pid_t pid)
+
+{
+ return 0;
+}
#endif
#endif /* _LINUX_SYSLOG_H */
diff --git a/kernel/syslog.c b/kernel/syslog.c
index 9873c06..1ab9bdc 100644
--- a/kernel/syslog.c
+++ b/kernel/syslog.c
@@ -18,7 +18,8 @@
* where container kernel data are redirected, kept and
* managed.
*
- * Containerized syslog is activated by CLONE_SYSLOG
+ * Containerized syslog is activated if CLONE_SYSLOG
+ * condition is true.
*
*/
@@ -47,23 +48,99 @@ struct syslog_ns init_kernel_syslog_ns = {
.buf_len = __LOG_BUF_LEN,
.buf = __log_buf
};
-
struct syslog_ns init_kernel_syslog_ns;
EXPORT_SYMBOL_GPL(init_kernel_syslog_ns);
/*
+ * List of all syslog ns currently allocated
+ * first member of this list (kernel syslog)
+ * can't be removed.
+ */
+struct log_list {
+ spinlock_t list_lock; /*make sure about list access */
+ struct log_list *next; /*next syslog_ns in the list */
+ struct syslog_ns *syslog_ns;
+ } log_list = {
+ .list_lock = __SPIN_LOCK_INITIALIZER(list_lock),
+ .next = (struct log_list *)0,
+ .syslog_ns = &init_kernel_syslog_ns
+ };
+/*
+ * removing a syslog reference from the list
+ *
+ */
+static void removing_syslog_ns(struct syslog_ns *syslog_ns)
+
+{
+ int done;
+ struct log_list *start;
+
+ done = false;
+ start = &log_list;
+ while (start->next != (struct log_list *)0) {
+ struct log_list *check;
+ unsigned long flags;
+
+ spin_lock_irqsave(&(start->next->list_lock), flags);
+ check = start->next;
+ if (check->syslog_ns == syslog_ns) {
+ start->next = check->next;
+ done = true;
+ }
+ spin_unlock_irqrestore(&(check->list_lock), flags);
+ if (done == true) {
+ kfree(check);
+ break;
+ }
+ start = start->next;
+ }
+}
+
+/*
+ * adding a syslog_ns to the list of known syslog
+ *
+ */
+static void adding_syslog_ns(struct syslog_ns *syslog_ns)
+
+{
+ int done;
+ struct log_list *start;
+
+ done = false;
+ start = &log_list;
+ while ((done == false) && (start != (struct log_list *)0)) {
+ unsigned long flags;
+
+ spin_lock_irqsave(&(start->list_lock), flags);
+ if (start->next == (struct log_list *)0) {
+ struct log_list *next;
+
+ next = kzalloc(sizeof(struct log_list), GFP_KERNEL);
+ BUG_ON(!next);
+ spin_lock_init(&(next->list_lock));
+ next->syslog_ns = syslog_ns;
+ start->next = next;
+ done = true;
+ }
+ spin_unlock_irqrestore(&(start->list_lock), flags);
+ start = start->next;
+ }
+}
+
+/*
* Procedure to free all ressources tied to syslog
*
*/
-static struct syslog_ns *free_all_syslog_ns(struct syslog_ns *syslog)
+static struct syslog_ns *free_all_syslog_ns(struct syslog_ns *syslog_ns)
{
- if (syslog != (struct syslog_ns *)0) {
- (void) kfree(syslog->buf);
- (void) kfree(syslog);
- syslog = (struct syslog_ns *)0;
+ if (syslog_ns != (struct syslog_ns *)0) {
+ (void) removing_syslog_ns(syslog_ns);
+ (void) kfree(syslog_ns->buf);
+ (void) kfree(syslog_ns);
+ syslog_ns = (struct syslog_ns *)0;
}
- return syslog;
+ return syslog_ns;
}
/*
@@ -93,9 +170,28 @@ static struct syslog_ns *malloc_syslog_ns(unsigned
container_buf_len)
ns->handle = current->pid;
ns->buf_len = container_buf_len;
ns->buf = buf;
+ (void) adding_syslog_ns(ns);
return ns;
}
/*
+ * Procedure to locate and return a syslog_ns with same handle as pid
submitted.
+ * return a NULL pointer if not found;
+ *
+ */
+static struct syslog_ns *find_its_syslog_ns(pid_t pid)
+
+{
+ struct log_list *start;
+
+ start = &log_list;
+ while (start != (struct log_list *)0) {
+ if (start->syslog_ns->handle == pid)
+ return start->syslog_ns;
+ start = start->next;
+ }
+ return 0;
+}
+/*
* Procedure to ONLY increase syslog buffer size
* If syslog_ns is NULL, assign a brand new syslog_ns
*
@@ -193,7 +289,7 @@ void free_syslog_ns(struct kref *kref)
/*
* Procedure to get the current syslog area linked to a
* container (by CLONE_SYSLOG).
- * if trouble, pin down the problem before it propagate.
+ * if trouble, report host kernel own syslog_ns.
*
*/
struct syslog_ns *current_syslog_ns(void)
@@ -222,7 +318,31 @@ struct syslog_ns *switch_syslog_ns(struct syslog_ns
*syslog_ns)
spin_lock_irqsave(&(current_syslog_ns()->logbuf_lock), flags);
old = current_syslog_ns();
- current->nsproxy->syslog_ns = syslog_ns;
+ if (syslog_ns)
+ current->nsproxy->syslog_ns = syslog_ns;
spin_unlock_irqrestore(&(old->logbuf_lock), flags);
return old;
}
+/*
+ * Procedure to locate the syslog handle own by a given
+ * pid or one of its parents lineage.
+ *
+ */
+struct syslog_ns *find_syslog_ns_bypid(pid_t pid)
+
+{
+ while (pid > 1) {
+ struct syslog_ns *its_ns;
+ struct task_struct *ns_task;
+
+ its_ns = find_its_syslog_ns(pid);
+ if (its_ns)
+ return its_ns;
+ ns_task = find_task_by_vpid(pid);
+ if ((ns_task) && (ns_task->real_parent))
+ pid = ns_task->real_parent->pid;
+ else
+ break;
+ }
+ return log_list.syslog_ns;
+}
diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
index 794bcb8..e41bc93 100644
--- a/net/core/rtnetlink.c
+++ b/net/core/rtnetlink.c
@@ -775,6 +775,9 @@ static int do_setlink(struct net_device *dev, struct
ifinfomsg *ifm,
err = PTR_ERR(net);
goto errout;
}
+ if (dev)
+ dev->syslog_ns_pid = nla_get_u32(tb[IFLA_NET_NS_PID]);
+
err = dev_change_net_namespace(dev, net, ifname);
put_net(net);
if (err)
diff --git a/net/ipv4/netfilter/ipt_LOG.c b/net/ipv4/netfilter/ipt_LOG.c
index ee128ef..7e1467d 100644
--- a/net/ipv4/netfilter/ipt_LOG.c
+++ b/net/ipv4/netfilter/ipt_LOG.c
@@ -14,6 +14,7 @@
#include <linux/spinlock.h>
#include <linux/skbuff.h>
#include <linux/ip.h>
+#include <linux/syslog.h>
#include <net/icmp.h>
#include <net/udp.h>
#include <net/tcp.h>
@@ -382,6 +383,15 @@ ipt_log_packet(u_int8_t pf,
const struct nf_loginfo *loginfo,
const char *prefix)
{
+
+ struct syslog_ns *syslog_ns;
+
+
+ syslog_ns = (struct syslog_ns *)0;
+ if (skb->dev) /*another syslog_ns possible? */
+ syslog_ns = find_syslog_ns_bypid(skb->dev->syslog_ns_pid);
+ syslog_ns = switch_syslog_ns(syslog_ns);
+
if (!loginfo)
loginfo = &default_loginfo;
@@ -422,6 +432,7 @@ ipt_log_packet(u_int8_t pf,
dump_packet(loginfo, skb, 0);
printk("\n");
spin_unlock_bh(&log_lock);
+ (void) switch_syslog_ns(syslog_ns);
}
static unsigned int
--
1.6.6
_______________________________________________
Containers mailing list
[email protected]
https://lists.linux-foundation.org/mailman/listinfo/containers
_______________________________________________
Devel mailing list
[email protected]
https://openvz.org/mailman/listinfo/devel
