From: Andy Lutomirski <email@example.com> To: Andy Lutomirski <firstname.lastname@example.org> Cc: Daniel Borkmann <email@example.com>, Alexei Starovoitov <firstname.lastname@example.org>, Song Liu <email@example.com>, Kees Cook <firstname.lastname@example.org>, Networking <email@example.com>, bpf <firstname.lastname@example.org>, Alexei Starovoitov <email@example.com>, Kernel Team <Kernelfirstname.lastname@example.org>, Lorenz Bauer <email@example.com>, Jann Horn <firstname.lastname@example.org>, Greg KH <email@example.com>, Linux API <firstname.lastname@example.org>, LSM List <email@example.com>, Chenbo Feng <firstname.lastname@example.org> Subject: RFC: very rough draft of a bpf permission model Date: Thu, 22 Aug 2019 08:17:54 -0700 Message-ID: <CALCETrWU4xJh4UBg0BboCwdGrgj+dUShsH5ETpiRgEpXJTEfQA@mail.gmail.com> (raw) In-Reply-To: <CALCETrUWQbPK3Pc6P5i_UqHPXJmZVyvuYXfq+VRtD6A3emaRhw@mail.gmail.com> BPF security strawman, v0.1 This is very rough. Most of this, especially the API details, needs work before it's ready to implement. The whole concept also needs review. = Goals = The overall goal is to make it possible to use eBPF without having what is effectively administrator access. For example, an eBPF user should not be able to directly tamper with other processes (unless this permission is explicitly granted) and should not be able to read or write other users' eBPF maps. It should be possible to use eBPF inside a user namespace without breaking the userns security model. Due to the risk of speculation attacks and such being carried out via eBPF, it should not become possible to use too much of eBPF without the administrator's permission. (NB: it is already possible to use *classic* BPF without any permission, and classic BPF is translated internally to eBPF, so this goal can only be met to a limited extent.) = Definitions = Global capability: A capability bit in the caller's effective mask, so long as the caller is in the root user namespace. Tasks in non-root user namespaces never have global capabilibies. This is what capable() checks. Namespace capability: A capability over a specific user namespace. Tasks in a user namespace have all the capabilities in their effective mask over their user namespace. A namespace capability generally indicates that the capability applies to the user namespace itself and to all non-user namespaces that live in the user namespace. For example, CAP_NET_ADMIN means that you can configure all networks namespaces in the current user namespace. This is what ns_capable() checks. Anything that requires a global capability will not work in a non-root user namespace. = unprivileged_bpf_disabled = Nothing in here supercedes unprivileged_bpf_disabled. If unprivileged_bpf_disabled = 1, then these proposals should not allow anything that is disallowed today. The idea is to make unprivileged_bpf_disabled=0 both safer and more useful. = Test runs = Global CAP_SYS_ADMIN is needed to test-run a program. Test-running a program exposes its own attack surface. It's also the only way to run a program at all if you merely have permission to load the program but not to attach it anywhere. Some of the proposed changes below will make it possible to load most program types without = Access to programs and maps = There are two basic security concerns when accessing programs and maps: the attack surface against the kernel and the ability to access other people's maps. Unprivileged processes may read a map if they have an FMODE_READ descriptor for the map. Unprivileged processes may write a map if they have an FMODE_WRITE descriptor to the map. Unprivileged processes may open a persistent map with a mode consistent with the permissions in bpffs. Unprivileged processes may create a bpffs inode for an existing map if the have an RW file descriptor for the map. (This is a change to current behavior. Daniel, Alexei thought the current behavior was intentional. Do you recall whether this is the case?) The _BY_ID map APIs inherently have no concept of ownership of maps. These APIs will continue to require global CAP_SYS_ADMIN. The small number of things that currently require the _BY_ID APIs, e.g., reading maps of maps, can be addressed if needed with new APIs that return fds instead of ids. Otherwise using them will continue to require global capabilities. Unprivileged processes may create exactly the set of maps that they can create today. Future proposals may extend this by a variety of means; this current proposal makes no changes. = Program loading = Loading a program carries the following risks: - It exposes the attack surface in the program verifier itself. That is possible, although unlikely, that merely verifying a malicious program could crash or otherwise cause a kernel malfunction. - It exposes the attack surface of insufficient checks in the verifier. That is, a verifier bug could allow a malicious program that is dangerous when run. - It exposes all of the functions that the program type can call. Some functions, e.g. bpf_probe_read(), should require privilege to call. - It exposes resource attacks. Currently, privileged users can load programs that use more resources than unprivileged users can load. - It exposes pointer-to-integer conversions. This requires global capabilities. - The program could contain speculation attack gadgets. - Loading a program is a prerequisite to attaching the program. I propose the following: Flag functions that require privilege as such. Loading a program that calls such a function will require a global capability. The privileged functions are mainly used for tracing, I think, and kernel tracing should require global capabilities. Loading a program that uses privileged verifier features (function calls or pointer-to-integer-conversions) will continue to require privileges. Loading a function that uses excessive resources can continue to require global capabilities or it could use a new set of cgroup settings that adjust the bpf complexity limits. Loading a function that bypasses the various speculation attack hardening features (e.g. constant blinding) requires global capabilities. Other than this, bpf program types can have a new flag set to allow them to be loaded without any privileges. Some bpf program types may need additional care, e.g. perf bpf events. They can be attached without privilege even in current kernels, and this might need to change. (optional) Add an API to load a program where the program source comes from a file specified by id instead of in memory. This would allow LSMs to require that bpf() programs be appropriately labeled. If LSMs require use of this API, it will make it much harder to exploit the verifier or speculation bugs. As a possible future extension, a way to selectively grant the ability to use specific program types without privilege could be useful. This could be done with a cgroup option, for example. = Cgroup attach = Cgroups have their own hierarchy that does not necessarily follow the namespace hierarchy. Unless cgroups integrate with namespaces in ways that they currently don't, namespace capabilites cannot be used to grant permission to operate on cgroups. I propose that attaching and detaching bpf programs to cgroups use a permission model similar to the model for changing non-bpf cgroup settings. In particular, each bpf_attach_type will get a new file in a cgroup directory. So there will be /sys/fs/cgroup/cgroup_name/bpf.inet_ingress, bpf.inet_egress, etc. A new API will be added to bpf() to attach and detach programs. The new API will take an fd to the bpf.attach_type file instead of to the cgroup directory. It will require FMODE_WRITE. This API will *not* require any capability. To prevent anyone with a delegated cgroup from automatically being able to use all bpf program types, the new bpf.attach_type files will be opt-in as part of the hierarchy. This could be done by writing "+bpf.*" or "+bpf.inet_ingress" to cgroup.subtree_control to make all the bpf.attach_type files or just bpf.inet_ingress available in descendents of the cgroup in question. This could alternatively be a new bpf.subtree_control file if that seems better. The result of these changes will be that root can use the old attach API or the new attach API. Unprivileged programs cannot use the old attach API. Unprivileged programs can use the new attach API if they are explicitly granted permission by all their ancestor cgroup managers. = Additional mitigations = Optional: there may be cases where a user can load a bpf program but can't attach or otherwise execute it. Nonetheless, it's plausible that such a program could be speculatively executed. The kernel could mitigate this by only marking a JITted bpf program executable when it is first attached or test-run.
next prev parent reply index Thread overview: 61+ messages / expand[flat|nested] mbox.gz Atom feed top [not found] <email@example.com> [not found] ` <firstname.lastname@example.org> [not found] ` <email@example.com> [not found] ` <3C595328-3ABE-4421-9772-8D41094A4F57@fb.com> [not found] ` <CALCETrWBnH4Q43POU8cQ7YMjb9LioK28FDEQf7aHZbdf1eBZWg@mail.gmail.com> [not found] ` <0DE7F23E-9CD2-4F03-82B5-835506B59056@fb.com> [not found] ` <CALCETrWBWbNFJvsTCeUchu3BZJ3SH3dvtXLUB2EhnPrzFfsLNA@mail.gmail.com> [not found] ` <201907021115.DCD56BBABB@keescook> [not found] ` <CALCETrXTta26CTtEDnzvtd03-WOGdXcnsAogP8JjLkcj4-mHvg@mail.gmail.com> [not found] ` <4A7A225A-6C23-4C0F-9A95-7C6C56B281ED@fb.com> [not found] ` <CALCETrX2bMnwC6_t4b_G-hzJSfMPrkK4YKs5ebcecv2LJ0rt3w@mail.gmail.com> [not found] ` <514D5453-0AEE-420F-AEB6-3F4F58C62E7E@fb.com> [not found] ` <1DE886F3-3982-45DE-B545-67AD6A4871AB@amacapital.net> [not found] ` <7F51F8B8-CF4C-4D82-AAE1-F0F28951DB7F@fb.com> [not found] ` <77354A95-4107-41A7-8936-D144F01C3CA4@fb.com> [not found] ` <369476A8-4CE1-43DA-9239-06437C0384C7@fb.com> 2019-07-30 20:24 ` [PATCH v2 bpf-next 1/4] bpf: unprivileged BPF access via /dev/bpf Andy Lutomirski 2019-07-31 8:10 ` Song Liu 2019-07-31 19:09 ` Andy Lutomirski 2019-08-02 7:21 ` Song Liu 2019-08-04 22:16 ` Andy Lutomirski 2019-08-05 0:08 ` Andy Lutomirski 2019-08-05 5:47 ` Andy Lutomirski 2019-08-05 7:36 ` Song Liu 2019-08-05 17:23 ` Andy Lutomirski 2019-08-05 19:21 ` Alexei Starovoitov 2019-08-05 21:25 ` Andy Lutomirski 2019-08-05 22:21 ` Andy Lutomirski 2019-08-06 1:11 ` Alexei Starovoitov 2019-08-07 5:24 ` Andy Lutomirski 2019-08-07 9:03 ` Lorenz Bauer 2019-08-07 13:52 ` Andy Lutomirski 2019-08-13 21:58 ` Alexei Starovoitov 2019-08-13 22:26 ` Daniel Colascione 2019-08-13 23:24 ` Andy Lutomirski 2019-08-13 23:06 ` Andy Lutomirski 2019-08-14 0:57 ` Alexei Starovoitov 2019-08-14 17:51 ` Andy Lutomirski 2019-08-14 22:05 ` Alexei Starovoitov 2019-08-14 22:30 ` Andy Lutomirski 2019-08-14 23:33 ` Alexei Starovoitov 2019-08-14 23:59 ` Andy Lutomirski 2019-08-15 0:36 ` Alexei Starovoitov 2019-08-15 11:24 ` Jordan Glover 2019-08-15 17:28 ` Alexei Starovoitov 2019-08-15 18:36 ` Andy Lutomirski 2019-08-15 23:08 ` Alexei Starovoitov 2019-08-16 9:34 ` Jordan Glover 2019-08-16 9:59 ` Thomas Gleixner 2019-08-16 11:33 ` Jordan Glover 2019-08-16 19:52 ` Alexei Starovoitov 2019-08-16 20:28 ` Thomas Gleixner 2019-08-17 15:02 ` Alexei Starovoitov 2019-08-17 15:44 ` Andy Lutomirski 2019-08-19 9:15 ` Thomas Gleixner 2019-08-19 17:27 ` Alexei Starovoitov 2019-08-19 17:38 ` Andy Lutomirski 2019-08-15 18:43 ` Jordan Glover 2019-08-15 19:46 ` Kees Cook 2019-08-15 23:46 ` Alexei Starovoitov 2019-08-16 0:54 ` Andy Lutomirski 2019-08-16 5:56 ` Song Liu 2019-08-16 21:45 ` Alexei Starovoitov 2019-08-16 22:22 ` Christian Brauner 2019-08-17 15:08 ` Alexei Starovoitov 2019-08-17 15:16 ` Christian Brauner 2019-08-17 15:36 ` Alexei Starovoitov 2019-08-17 15:42 ` Christian Brauner 2019-08-22 14:17 ` Daniel Borkmann 2019-08-22 15:16 ` Andy Lutomirski 2019-08-22 15:17 ` Andy Lutomirski [this message] 2019-08-22 23:26 ` RFC: very rough draft of a bpf permission model Alexei Starovoitov 2019-08-23 23:09 ` Andy Lutomirski 2019-08-26 22:36 ` Alexei Starovoitov 2019-08-27 0:05 ` Andy Lutomirski 2019-08-27 0:34 ` Alexei Starovoitov 2019-08-22 22:48 ` [PATCH v2 bpf-next 1/4] bpf: unprivileged BPF access via /dev/bpf Alexei Starovoitov
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=CALCETrWU4xJh4UBg0BboCwdGrgj+dUShsH5ETpiRgEpXJTEfQA@mail.gmail.com \ --firstname.lastname@example.org \ --cc=Kernelemail@example.com \ --firstname.lastname@example.org \ --email@example.com \ --firstname.lastname@example.org \ --email@example.com \ --firstname.lastname@example.org \ --email@example.com \ --firstname.lastname@example.org \ --email@example.com \ --firstname.lastname@example.org \ --email@example.com \ --firstname.lastname@example.org \ --email@example.com \ --firstname.lastname@example.org \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: link
Linux-Security-Module Archive on lore.kernel.org Archives are clonable: git clone --mirror https://lore.kernel.org/linux-security-module/0 linux-security-module/git/0.git # If you have public-inbox 1.1+ installed, you may # initialize and index your mirror using the following commands: public-inbox-init -V2 linux-security-module linux-security-module/ https://lore.kernel.org/linux-security-module \ email@example.com public-inbox-index linux-security-module Example config snippet for mirrors Newsgroup available over NNTP: nntp://nntp.lore.kernel.org/org.kernel.vger.linux-security-module AGPL code for this site: git clone https://public-inbox.org/public-inbox.git