[net PATCH v3 1/3] net: netprio: fix files lock and remove useless d_path bits

Add lock to prevent a race with a file closing and also remove
useless and ugly sscanf code. The extra code was never needed
and the case it supposedly protected against is in fact handled
correctly by sock_from_file as pointed out by Al Viro.

CC: Neil Horman <nhorman@tuxdriver.com>
Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: John Fastabend <redacted>
---

 net/core/netprio_cgroup.c |   22 ++++------------------
 1 files changed, 4 insertions(+), 18 deletions(-)

diff --git a/net/core/netprio_cgroup.c b/net/core/netprio_cgroup.c
index ed0c043..f65dba3 100644
--- a/net/core/netprio_cgroup.c
+++ b/net/core/netprio_cgroup.c

@@ -277,12 +277,6 @@ out_free_devname:
 void net_prio_attach(struct cgroup *cgrp, struct cgroup_taskset *tset)
 {
 	struct task_struct *p;
-	char *tmp = kzalloc(sizeof(char) * PATH_MAX, GFP_KERNEL);
-
-	if (!tmp) {
-		pr_warn("Unable to attach cgrp due to alloc failure!\n");
-		return;
-	}
 
 	cgroup_taskset_for_each(p, cgrp, tset) {
 		unsigned int fd;

@@ -296,32 +290,24 @@ void net_prio_attach(struct cgroup *cgrp, struct cgroup_taskset *tset)
 			continue;
 		}
 
-		rcu_read_lock();
+		spin_lock(&files->file_lock);
 		fdt = files_fdtable(files);
 		for (fd = 0; fd < fdt->max_fds; fd++) {
-			char *path;
 			struct file *file;
 			struct socket *sock;
-			unsigned long s;
-			int rv, err = 0;
+			int err;
 
 			file = fcheck_files(files, fd);
 			if (!file)
 				continue;
 
-			path = d_path(&file->f_path, tmp, PAGE_SIZE);
-			rv = sscanf(path, "socket:[%lu]", &s);
-			if (rv <= 0)
-				continue;
-
 			sock = sock_from_file(file, &err);
-			if (!err)
+			if (sock)
 				sock_update_netprioidx(sock->sk, p);
 		}
-		rcu_read_unlock();
+		spin_unlock(&files->file_lock);
 		task_unlock(p);
 	}
-	kfree(tmp);
 }
 
 static struct cftype ss_files[] = {

A socket fd passed in a SCM_RIGHTS datagram was not getting
updated with the new tasks cgrp prioidx. This leaves IO on
the socket tagged with the old tasks priority.

To fix this add a check in the scm recvmsg path to update the
sock cgrp prioidx with the new tasks value.

Thanks to Al Viro for catching this.

CC: Neil Horman <nhorman@tuxdriver.com>
Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: John Fastabend <redacted>
---

 net/core/scm.c |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)

diff --git a/net/core/scm.c b/net/core/scm.c
index 8f6ccfd..040cebe 100644
--- a/net/core/scm.c
+++ b/net/core/scm.c

@@ -265,6 +265,7 @@ void scm_detach_fds(struct msghdr *msg, struct scm_cookie *scm)
 	for (i=0, cmfptr=(__force int __user *)CMSG_DATA(cm); i<fdmax;
 	     i++, cmfptr++)
 	{
+		struct socket *sock;
 		int new_fd;
 		err = security_file_receive(fp[i]);
 		if (err)

@@ -281,6 +282,9 @@ void scm_detach_fds(struct msghdr *msg, struct scm_cookie *scm)
 		}
 		/* Bump the usage count and install the file. */
 		get_file(fp[i]);
+		sock = sock_from_file(fp[i], &err);
+		if (sock)
+			sock_update_netprioidx(sock->sk, current);
 		fd_install(new_fd, fp[i]);
 	}
 

A race exists where creating cgroups and also updating the priomap
may result in losing a priomap update. This is because priomap
writers are not protected by rtnl_lock.

Move priority writer into rtnl_lock()/rtnl_unlock().

CC: Neil Horman <nhorman@tuxdriver.com>
Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: John Fastabend <redacted>
---

 net/core/netprio_cgroup.c |    8 +++-----
 1 files changed, 3 insertions(+), 5 deletions(-)

diff --git a/net/core/netprio_cgroup.c b/net/core/netprio_cgroup.c
index f65dba3..c75e3f9 100644
--- a/net/core/netprio_cgroup.c
+++ b/net/core/netprio_cgroup.c

@@ -101,12 +101,10 @@ static int write_update_netdev_table(struct net_device *dev)
 	u32 max_len;
 	struct netprio_map *map;
 
-	rtnl_lock();
 	max_len = atomic_read(&max_prioidx) + 1;
 	map = rtnl_dereference(dev->priomap);
 	if (!map || map->priomap_len < max_len)
 		ret = extend_netdev_table(dev, max_len);
-	rtnl_unlock();
 
 	return ret;
 }

@@ -256,17 +254,17 @@ static int write_priomap(struct cgroup *cgrp, struct cftype *cft,
 	if (!dev)
 		goto out_free_devname;
 
+	rtnl_lock();
 	ret = write_update_netdev_table(dev);
 	if (ret < 0)
 		goto out_put_dev;
 
-	rcu_read_lock();
-	map = rcu_dereference(dev->priomap);
+	map = rtnl_dereference(dev->priomap);
 	if (map)
 		map->priomap[prioidx] = priority;
-	rcu_read_unlock();
 
 out_put_dev:
+	rtnl_unlock();
 	dev_put(dev);
 
 out_free_devname:

On Tue, Aug 14, 2012 at 03:34:24PM -0700, John Fastabend wrote:

Acked-by: Neil Horman <nhorman@tuxdriver.com>

It looks good to me.  Al, could you please lend your review here too?

On Tue, Aug 14, 2012 at 03:34:30PM -0700, John Fastabend wrote:

Acked-by: Neil Horman <nhorman@tuxdriver.com>

On Tue, Aug 14, 2012 at 03:34:35PM -0700, John Fastabend wrote:

Acked-by: Neil Horman <nhorman@tuxdriver.com>

On Wed, Aug 15, 2012 at 06:52:06AM -0400, Neil Horman wrote:

On Tue, Aug 14, 2012 at 03:34:24PM -0700, John Fastabend wrote:

quoted

Add lock to prevent a race with a file closing and also remove
useless and ugly sscanf code. The extra code was never needed
and the case it supposedly protected against is in fact handled
correctly by sock_from_file as pointed out by Al Viro.

CC: Neil Horman <nhorman@tuxdriver.com>
Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: John Fastabend <redacted>
---

 net/core/netprio_cgroup.c |   22 ++++------------------
 1 files changed, 4 insertions(+), 18 deletions(-)

diff --git a/net/core/netprio_cgroup.c b/net/core/netprio_cgroup.c
index ed0c043..f65dba3 100644
--- a/net/core/netprio_cgroup.c
+++ b/net/core/netprio_cgroup.c

@@ -277,12 +277,6 @@ out_free_devname:
 void net_prio_attach(struct cgroup *cgrp, struct cgroup_taskset *tset)
 {
 	struct task_struct *p;
-	char *tmp = kzalloc(sizeof(char) * PATH_MAX, GFP_KERNEL);
-
-	if (!tmp) {
-		pr_warn("Unable to attach cgrp due to alloc failure!\n");
-		return;
-	}
 
 	cgroup_taskset_for_each(p, cgrp, tset) {
 		unsigned int fd;

@@ -296,32 +290,24 @@ void net_prio_attach(struct cgroup *cgrp, struct cgroup_taskset *tset)
 			continue;
 		}
 
-		rcu_read_lock();
+		spin_lock(&files->file_lock);
 		fdt = files_fdtable(files);
 		for (fd = 0; fd < fdt->max_fds; fd++) {
-			char *path;
 			struct file *file;
 			struct socket *sock;
-			unsigned long s;
-			int rv, err = 0;
+			int err;
 
 			file = fcheck_files(files, fd);
 			if (!file)
 				continue;
 
-			path = d_path(&file->f_path, tmp, PAGE_SIZE);
-			rv = sscanf(path, "socket:[%lu]", &s);
-			if (rv <= 0)
-				continue;
-
 			sock = sock_from_file(file, &err);
-			if (!err)
+			if (sock)
 				sock_update_netprioidx(sock->sk, p);
 		}
-		rcu_read_unlock();
+		spin_unlock(&files->file_lock);
 		task_unlock(p);
 	}
-	kfree(tmp);
 }
 
 static struct cftype ss_files[] = {

Acked-by: Neil Horman <nhorman@tuxdriver.com>

It looks good to me.  Al, could you please lend your review here too?

Tolerable...  I still don't like the idea of iterating through the
descriptor tables, but at least that variant is safe wrt locking and
lifetime rules.

From: John Fastabend <redacted>
Date: Tue, 14 Aug 2012 15:34:24 -0700

Add lock to prevent a race with a file closing and also remove
useless and ugly sscanf code. The extra code was never needed
and the case it supposedly protected against is in fact handled
correctly by sock_from_file as pointed out by Al Viro.

CC: Neil Horman <nhorman@tuxdriver.com>
Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: John Fastabend <redacted>

Applied.

From: John Fastabend <redacted>
Date: Tue, 14 Aug 2012 15:34:30 -0700

A socket fd passed in a SCM_RIGHTS datagram was not getting
updated with the new tasks cgrp prioidx. This leaves IO on
the socket tagged with the old tasks priority.

To fix this add a check in the scm recvmsg path to update the
sock cgrp prioidx with the new tasks value.

Thanks to Al Viro for catching this.

CC: Neil Horman <nhorman@tuxdriver.com>
Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: John Fastabend <redacted>

Applied.

From: John Fastabend <redacted>
Date: Tue, 14 Aug 2012 15:34:35 -0700

A race exists where creating cgroups and also updating the priomap
may result in losing a priomap update. This is because priomap
writers are not protected by rtnl_lock.

Move priority writer into rtnl_lock()/rtnl_unlock().

CC: Neil Horman <nhorman@tuxdriver.com>
Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: John Fastabend <redacted>

Applied.