linux-kernel - [PATCH RFC bpf-next v2 4/5] bpf: Pin cgroup

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-Id: <20220201205534.1962784-5-haoluo@google.com>
Date:   Tue,  1 Feb 2022 12:55:33 -0800
From:   Hao Luo <haoluo@...gle.com>
To:     Alexei Starovoitov <ast@...nel.org>,
        Andrii Nakryiko <andrii@...nel.org>,
        Daniel Borkmann <daniel@...earbox.net>
Cc:     Martin KaFai Lau <kafai@...com>, Song Liu <songliubraving@...com>,
        Yonghong Song <yhs@...com>, KP Singh <kpsingh@...nel.org>,
        Shakeel Butt <shakeelb@...gle.com>,
        Joe Burton <jevburton.kernel@...il.com>,
        Stanislav Fomichev <sdf@...gle.com>, bpf@...r.kernel.org,
        linux-kernel@...r.kernel.org, Hao Luo <haoluo@...gle.com>
Subject: [PATCH RFC bpf-next v2 4/5] bpf: Pin cgroup_view

When pinning cgroup_view iter into bpffs, the pinning adds a tag to
the parent directory, indicating the underneath directory hierarchy
is a replicate of the cgroup hierarchy. Each of the directory connects
to a cgroup directory. Whenever a subdirectory is created, if there is
a subcgroup of the same name exists, the subdirectory will be populated
with entries holding a list of bpf objects registered in the tag's
inherit list. The inherit list is formed by the objects pinned in the
top level tagged directory. For example,

 bpf_obj_pin(cgroup_view_link, "/sys/fs/bpf/A/obj");

pins a link in A. A becomes tagged and the link object is registered
in the inherit list of A's tag.

 mkdir("/sys/fs/bpf/A/B");

When A/B is created, B inherits the pinned objects in A. B is populated
with objects.

 > ls /sys/fs/bpf/A/B
 obj

Currently, only pinning cgroup_view link can tag a directory. And once
tagged, only rmdir can remove the tag.

Signed-off-by: Hao Luo <haoluo@...gle.com>
---
 kernel/bpf/inode.c | 84 ++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 84 insertions(+)

diff --git a/kernel/bpf/inode.c b/kernel/bpf/inode.c
index 9ae17a2bf779..b71840bf979d 100644
--- a/kernel/bpf/inode.c
+++ b/kernel/bpf/inode.c
@@ -71,6 +71,7 @@ static void free_obj_list(struct kref *kref)
 	list_for_each_entry(e, &list->list, list) {
 		list_del_rcu(&e->list);
 		bpf_any_put(e->obj, e->type);
+		kfree(e->name.name);
 		kfree(e);
 	}
 	kfree(list);
@@ -486,9 +487,20 @@ static int bpf_mkmap(struct dentry *dentry, umode_t mode, void *arg)
 			     &bpffs_map_fops : &bpffs_obj_fops);
 }
 
+static int
+bpf_inherit_object(struct dentry *dentry, umode_t mode, void *obj,
+		   enum bpf_type type);
+
 static int bpf_mklink(struct dentry *dentry, umode_t mode, void *arg)
 {
 	struct bpf_link *link = arg;
+	int err;
+
+	if (bpf_link_support_inherit(link)) {
+		err = bpf_inherit_object(dentry, mode, link, BPF_TYPE_LINK);
+		if (err)
+			return err;
+	}
 
 	return bpf_mkobj_ops(dentry, mode, arg, &bpf_link_iops,
 			     bpf_link_is_iter(link) ?
@@ -586,6 +598,78 @@ static const struct inode_operations bpf_dir_iops = {
 	.unlink		= simple_unlink,
 };
 
+/* bpf_inherit_object - register an object in a diretory tag's inherit list
+ * @dentry: dentry of the location to pin
+ * @mode: mode of created file entry
+ * @obj: bpf object
+ * @type: type of bpf object
+ *
+ * Could be called from bpf_obj_do_pin() or from mkdir().
+ */
+static int bpf_inherit_object(struct dentry *dentry, umode_t mode,
+			      void *obj, enum bpf_type type)
+{
+	struct inode *dir = d_inode(dentry->d_parent);
+	struct obj_list *inherits;
+	struct bpf_inherit_entry *e;
+	struct bpf_dir_tag *tag;
+	const char *name;
+	bool queued = false, new_tag = false;
+
+	/* allocate bpf_dir_tag */
+	tag = inode_tag(dir);
+	if (!tag) {
+		new_tag = true;
+		tag = kzalloc(sizeof(struct bpf_dir_tag), GFP_KERNEL);
+		if (unlikely(!tag))
+			return -ENOMEM;
+
+		tag->type = BPF_DIR_KERNFS_REP;
+		inherits = kzalloc(sizeof(struct obj_list), GFP_KERNEL);
+		if (unlikely(!inherits)) {
+			kfree(tag);
+			return -ENOMEM;
+		}
+
+		kref_init(&inherits->refcnt);
+		INIT_LIST_HEAD(&inherits->list);
+		tag->inherit_objects = inherits;
+		/* initial tag points to the default root cgroup. */
+		tag->private = cgrp_dfl_root.kf_root->kn;
+		dir->i_private = tag;
+	} else {
+		inherits = tag->inherit_objects;
+	}
+
+	list_for_each_entry_rcu(e, &inherits->list, list) {
+		if (!strcmp(dentry->d_name.name, e->name.name)) {
+			queued = true;
+			break;
+		}
+	}
+
+	/* queue in tag's inherit_list. */
+	if (!queued) {
+		e = kzalloc(sizeof(struct bpf_inherit_entry), GFP_KERNEL);
+		if (!e) {
+			if (new_tag) {
+				kfree(tag);
+				kfree(inherits);
+			}
+			return -ENOMEM;
+		}
+
+		INIT_LIST_HEAD(&e->list);
+		e->mode = mode;
+		e->obj = obj;
+		e->type = type;
+		name = kstrdup(dentry->d_name.name, GFP_USER | __GFP_NOWARN);
+		e->name = (struct qstr)QSTR_INIT(name, strlen(name));
+		list_add_rcu(&e->list, &inherits->list);
+	}
+	return 0;
+}
+
 /* pin iterator link into bpffs */
 static int bpf_iter_link_pin_kernel(struct dentry *parent,
 				    const char *name, struct bpf_link *link)
-- 
2.35.0.rc2.247.g8bbb082509-goog