From: Ondrej Mosnacek <omosnace@redhat.com> To: Daniel Borkmann <daniel@iogearbox.net> Cc: Paul Moore <paul@paul-moore.com>, Linux Security Module list <linux-security-module@vger.kernel.org>, James Morris <jmorris@namei.org>, Steven Rostedt <rostedt@goodmis.org>, Ingo Molnar <mingo@redhat.com>, Stephen Smalley <stephen.smalley.work@gmail.com>, SElinux list <selinux@vger.kernel.org>, linuxppc-dev@lists.ozlabs.org, Linux FS Devel <linux-fsdevel@vger.kernel.org>, bpf <bpf@vger.kernel.org>, network dev <netdev@vger.kernel.org>, Linux kernel mailing list <linux-kernel@vger.kernel.org>, Casey Schaufler <casey@schaufler-ca.com>, Jiri Olsa <jolsa@redhat.com>, andrii.nakryiko@gmail.com Subject: Re: [PATCH v2] lockdown,selinux: avoid bogus SELinux lockdown permission checks Date: Fri, 28 May 2021 15:42:06 +0200 [thread overview] Message-ID: <CAFqZXNsKf5wSGmspEVEDrm4Ywar-F4kJWbBPBE+_hd1CGQ3jhg@mail.gmail.com> (raw) In-Reply-To: <4fee8c12-194f-3f85-e28b-f7f24ab03c91@iogearbox.net> (I'm off work today and plan to reply also to Paul's comments next week, but for now let me at least share a couple quick thoughts on Daniel's patch.) On Fri, May 28, 2021 at 11:56 AM Daniel Borkmann <daniel@iogearbox.net> wrote: > > On 5/28/21 9:09 AM, Daniel Borkmann wrote: > > On 5/28/21 3:37 AM, Paul Moore wrote: > >> On Mon, May 17, 2021 at 5:22 AM Ondrej Mosnacek <omosnace@redhat.com> wrote: > >>> > >>> Commit 59438b46471a ("security,lockdown,selinux: implement SELinux > >>> lockdown") added an implementation of the locked_down LSM hook to > >>> SELinux, with the aim to restrict which domains are allowed to perform > >>> operations that would breach lockdown. > >>> > >>> However, in several places the security_locked_down() hook is called in > >>> situations where the current task isn't doing any action that would > >>> directly breach lockdown, leading to SELinux checks that are basically > >>> bogus. > >>> > >>> Since in most of these situations converting the callers such that > >>> security_locked_down() is called in a context where the current task > >>> would be meaningful for SELinux is impossible or very non-trivial (and > >>> could lead to TOCTOU issues for the classic Lockdown LSM > >>> implementation), fix this by modifying the hook to accept a struct cred > >>> pointer as argument, where NULL will be interpreted as a request for a > >>> "global", task-independent lockdown decision only. Then modify SELinux > >>> to ignore calls with cred == NULL. > >> > >> I'm not overly excited about skipping the access check when cred is > >> NULL. Based on the description and the little bit that I've dug into > >> thus far it looks like using SECINITSID_KERNEL as the subject would be > >> much more appropriate. *Something* (the kernel in most of the > >> relevant cases it looks like) is requesting that a potentially > >> sensitive disclosure be made, and ignoring it seems like the wrong > >> thing to do. Leaving the access control intact also provides a nice > >> avenue to audit these requests should users want to do that. > > > > I think the rationale/workaround for ignoring calls with cred == NULL (or the previous > > patch with the unimplemented hook) from Ondrej was two-fold, at least speaking for his > > seen tracing cases: > > > > i) The audit events that are triggered due to calls to security_locked_down() > > can OOM kill a machine, see below details [0]. > > > > ii) It seems to be causing a deadlock via slow_avc_audit() -> audit_log_end() > > when presumingly trying to wake up kauditd [1]. Actually, I wasn't aware of the deadlock... But calling an LSM hook [that is backed by a SELinux access check] from within a BPF helper is calling for all kinds of trouble, so I'm not surprised :) > Ondrej / Paul / Jiri: at least for the BPF tracing case specifically (I haven't looked > at the rest but it's also kind of independent), the attached fix should address both > reported issues, please take a look & test. Thanks, I like this solution, although there are a few gotchas: 1. This patch creates a slight "regression" in that if someone flips the Lockdown LSM into lockdown mode on runtime, existing (already loaded) BPF programs will still be able to call the confidentiality-breaching helpers, while before the lockdown would apply also to them. Personally, I don't think it's a big deal (and I bet there are other existing cases where some handle kept from before lockdown could leak data), but I wanted to mention it in case someone thinks the opposite. 2. IIUC. when a BPF program is rejected due to lockdown/SELinux, the kernel will return -EINVAL to userspace (looking at check_helper_call() in kernel/bpf/verifier.c; didn't have time to look at other callers...). It would be nicer if the error code from the security_locked_down() call would be passed through the call chain and eventually returned to the caller. It should be relatively straightforward to convert bpf_base_func_proto() to return a PTR_ERR() instead of NULL on error, but it looks like this would result in quite a big patch updating all the callers (and callers of callers, etc.) with a not-so-small chance of missing some NULL check and introducing a bug... I guess we could live with EINVAL-on-denied in stable kernels and only have the error path refactoring in -next; I'm not sure... 3. This is a bit of a shot-in-the-dark, but I suppose there might be some BPF programs that would be able to do something useful also when the read_kernel helpers return an error, yet the kernel will now outright refuse to load them (when the lockdown hook returns nonzero). I have no idea if such BPF programs realistically exist in practice, but perhaps it would be worth returning some dummy always-error-returning helper function instead of NULL from bpf_base_func_proto() when security_locked_down() returns an error. That would also resolve (2.), basically. (Then there is the question of what error code to use (because Lockdown LSM uses -EPERM, while SELinux -EACCESS), but I think always returning -EPERM from such stub helpers would be a viable choice.) -- Ondrej Mosnacek Software Engineer, Linux Security - SELinux kernel Red Hat, Inc.
WARNING: multiple messages have this Message-ID (diff)
From: Ondrej Mosnacek <omosnace@redhat.com> To: Daniel Borkmann <daniel@iogearbox.net> Cc: Jiri Olsa <jolsa@redhat.com>, Paul Moore <paul@paul-moore.com>, SElinux list <selinux@vger.kernel.org>, network dev <netdev@vger.kernel.org>, Stephen Smalley <stephen.smalley.work@gmail.com>, James Morris <jmorris@namei.org>, Steven Rostedt <rostedt@goodmis.org>, Linux kernel mailing list <linux-kernel@vger.kernel.org>, Casey Schaufler <casey@schaufler-ca.com>, Linux Security Module list <linux-security-module@vger.kernel.org>, Ingo Molnar <mingo@redhat.com>, Linux FS Devel <linux-fsdevel@vger.kernel.org>, bpf <bpf@vger.kernel.org>, andrii.nakryiko@gmail.com, linuxppc-dev@lists.ozlabs.org Subject: Re: [PATCH v2] lockdown,selinux: avoid bogus SELinux lockdown permission checks Date: Fri, 28 May 2021 15:42:06 +0200 [thread overview] Message-ID: <CAFqZXNsKf5wSGmspEVEDrm4Ywar-F4kJWbBPBE+_hd1CGQ3jhg@mail.gmail.com> (raw) In-Reply-To: <4fee8c12-194f-3f85-e28b-f7f24ab03c91@iogearbox.net> (I'm off work today and plan to reply also to Paul's comments next week, but for now let me at least share a couple quick thoughts on Daniel's patch.) On Fri, May 28, 2021 at 11:56 AM Daniel Borkmann <daniel@iogearbox.net> wrote: > > On 5/28/21 9:09 AM, Daniel Borkmann wrote: > > On 5/28/21 3:37 AM, Paul Moore wrote: > >> On Mon, May 17, 2021 at 5:22 AM Ondrej Mosnacek <omosnace@redhat.com> wrote: > >>> > >>> Commit 59438b46471a ("security,lockdown,selinux: implement SELinux > >>> lockdown") added an implementation of the locked_down LSM hook to > >>> SELinux, with the aim to restrict which domains are allowed to perform > >>> operations that would breach lockdown. > >>> > >>> However, in several places the security_locked_down() hook is called in > >>> situations where the current task isn't doing any action that would > >>> directly breach lockdown, leading to SELinux checks that are basically > >>> bogus. > >>> > >>> Since in most of these situations converting the callers such that > >>> security_locked_down() is called in a context where the current task > >>> would be meaningful for SELinux is impossible or very non-trivial (and > >>> could lead to TOCTOU issues for the classic Lockdown LSM > >>> implementation), fix this by modifying the hook to accept a struct cred > >>> pointer as argument, where NULL will be interpreted as a request for a > >>> "global", task-independent lockdown decision only. Then modify SELinux > >>> to ignore calls with cred == NULL. > >> > >> I'm not overly excited about skipping the access check when cred is > >> NULL. Based on the description and the little bit that I've dug into > >> thus far it looks like using SECINITSID_KERNEL as the subject would be > >> much more appropriate. *Something* (the kernel in most of the > >> relevant cases it looks like) is requesting that a potentially > >> sensitive disclosure be made, and ignoring it seems like the wrong > >> thing to do. Leaving the access control intact also provides a nice > >> avenue to audit these requests should users want to do that. > > > > I think the rationale/workaround for ignoring calls with cred == NULL (or the previous > > patch with the unimplemented hook) from Ondrej was two-fold, at least speaking for his > > seen tracing cases: > > > > i) The audit events that are triggered due to calls to security_locked_down() > > can OOM kill a machine, see below details [0]. > > > > ii) It seems to be causing a deadlock via slow_avc_audit() -> audit_log_end() > > when presumingly trying to wake up kauditd [1]. Actually, I wasn't aware of the deadlock... But calling an LSM hook [that is backed by a SELinux access check] from within a BPF helper is calling for all kinds of trouble, so I'm not surprised :) > Ondrej / Paul / Jiri: at least for the BPF tracing case specifically (I haven't looked > at the rest but it's also kind of independent), the attached fix should address both > reported issues, please take a look & test. Thanks, I like this solution, although there are a few gotchas: 1. This patch creates a slight "regression" in that if someone flips the Lockdown LSM into lockdown mode on runtime, existing (already loaded) BPF programs will still be able to call the confidentiality-breaching helpers, while before the lockdown would apply also to them. Personally, I don't think it's a big deal (and I bet there are other existing cases where some handle kept from before lockdown could leak data), but I wanted to mention it in case someone thinks the opposite. 2. IIUC. when a BPF program is rejected due to lockdown/SELinux, the kernel will return -EINVAL to userspace (looking at check_helper_call() in kernel/bpf/verifier.c; didn't have time to look at other callers...). It would be nicer if the error code from the security_locked_down() call would be passed through the call chain and eventually returned to the caller. It should be relatively straightforward to convert bpf_base_func_proto() to return a PTR_ERR() instead of NULL on error, but it looks like this would result in quite a big patch updating all the callers (and callers of callers, etc.) with a not-so-small chance of missing some NULL check and introducing a bug... I guess we could live with EINVAL-on-denied in stable kernels and only have the error path refactoring in -next; I'm not sure... 3. This is a bit of a shot-in-the-dark, but I suppose there might be some BPF programs that would be able to do something useful also when the read_kernel helpers return an error, yet the kernel will now outright refuse to load them (when the lockdown hook returns nonzero). I have no idea if such BPF programs realistically exist in practice, but perhaps it would be worth returning some dummy always-error-returning helper function instead of NULL from bpf_base_func_proto() when security_locked_down() returns an error. That would also resolve (2.), basically. (Then there is the question of what error code to use (because Lockdown LSM uses -EPERM, while SELinux -EACCESS), but I think always returning -EPERM from such stub helpers would be a viable choice.) -- Ondrej Mosnacek Software Engineer, Linux Security - SELinux kernel Red Hat, Inc.
next prev parent reply other threads:[~2021-05-28 13:42 UTC|newest] Thread overview: 74+ messages / expand[flat|nested] mbox.gz Atom feed top 2021-05-17 9:20 [PATCH v2] lockdown,selinux: avoid bogus SELinux lockdown permission checks Ondrej Mosnacek 2021-05-17 9:20 ` [PATCH v2] lockdown, selinux: " Ondrej Mosnacek 2021-05-17 11:00 ` [PATCH v2] lockdown,selinux: " Michael Ellerman 2021-05-17 11:00 ` Michael Ellerman 2021-05-26 11:44 ` Ondrej Mosnacek 2021-05-26 11:44 ` Ondrej Mosnacek 2021-05-27 4:28 ` James Morris 2021-05-27 4:28 ` James Morris 2021-05-27 14:18 ` Paul Moore 2021-05-27 14:18 ` Paul Moore 2021-05-28 1:37 ` Paul Moore 2021-05-28 1:37 ` Paul Moore 2021-05-28 7:09 ` Daniel Borkmann 2021-05-28 7:09 ` Daniel Borkmann 2021-05-28 9:53 ` Jiri Olsa 2021-05-28 9:53 ` Jiri Olsa 2021-05-28 9:56 ` Daniel Borkmann 2021-05-28 9:56 ` Daniel Borkmann 2021-05-28 10:16 ` Jiri Olsa 2021-05-28 10:16 ` Jiri Olsa 2021-05-28 11:47 ` Jiri Olsa 2021-05-28 11:47 ` Jiri Olsa 2021-05-28 11:54 ` Daniel Borkmann 2021-05-28 11:54 ` Daniel Borkmann 2021-05-28 13:42 ` Ondrej Mosnacek [this message] 2021-05-28 13:42 ` Ondrej Mosnacek 2021-05-28 14:20 ` Daniel Borkmann 2021-05-28 14:20 ` Daniel Borkmann 2021-05-28 15:54 ` Paul Moore 2021-05-28 15:54 ` Paul Moore 2021-05-28 15:47 ` Paul Moore 2021-05-28 15:47 ` Paul Moore 2021-05-28 18:10 ` Daniel Borkmann 2021-05-28 18:10 ` Daniel Borkmann 2021-05-28 22:52 ` Paul Moore 2021-05-28 22:52 ` Paul Moore 2021-05-29 18:48 ` Paul Moore 2021-05-29 18:48 ` Paul Moore 2021-05-31 8:24 ` Daniel Borkmann 2021-05-31 8:24 ` Daniel Borkmann 2021-06-01 20:47 ` Paul Moore 2021-06-01 20:47 ` Paul Moore 2021-06-02 12:40 ` Daniel Borkmann 2021-06-02 12:40 ` Daniel Borkmann 2021-06-02 15:13 ` Paul Moore 2021-06-02 15:13 ` Paul Moore 2021-06-03 18:52 ` Daniel Borkmann 2021-06-03 18:52 ` Daniel Borkmann 2021-06-04 4:50 ` Paul Moore 2021-06-04 4:50 ` Paul Moore 2021-06-04 18:02 ` Daniel Borkmann 2021-06-04 18:02 ` Daniel Borkmann 2021-06-04 23:34 ` Paul Moore 2021-06-04 23:34 ` Paul Moore 2021-06-05 0:08 ` Alexei Starovoitov 2021-06-05 0:08 ` Alexei Starovoitov 2021-06-05 18:10 ` Casey Schaufler 2021-06-05 18:10 ` Casey Schaufler 2021-06-05 18:17 ` Linus Torvalds 2021-06-05 18:17 ` Linus Torvalds 2021-06-06 2:11 ` Paul Moore 2021-06-06 2:11 ` Paul Moore 2021-06-06 1:30 ` Paul Moore 2021-06-06 1:30 ` Paul Moore 2021-06-02 13:39 ` Ondrej Mosnacek 2021-06-02 13:39 ` Ondrej Mosnacek 2021-06-03 17:46 ` Paul Moore 2021-06-03 17:46 ` Paul Moore 2021-06-08 11:01 ` Ondrej Mosnacek 2021-06-08 11:01 ` Ondrej Mosnacek 2021-06-09 2:40 ` Paul Moore 2021-06-09 2:40 ` Paul Moore 2021-05-28 13:58 ` Steven Rostedt 2021-05-28 13:58 ` Steven Rostedt
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=CAFqZXNsKf5wSGmspEVEDrm4Ywar-F4kJWbBPBE+_hd1CGQ3jhg@mail.gmail.com \ --to=omosnace@redhat.com \ --cc=andrii.nakryiko@gmail.com \ --cc=bpf@vger.kernel.org \ --cc=casey@schaufler-ca.com \ --cc=daniel@iogearbox.net \ --cc=jmorris@namei.org \ --cc=jolsa@redhat.com \ --cc=linux-fsdevel@vger.kernel.org \ --cc=linux-kernel@vger.kernel.org \ --cc=linux-security-module@vger.kernel.org \ --cc=linuxppc-dev@lists.ozlabs.org \ --cc=mingo@redhat.com \ --cc=netdev@vger.kernel.org \ --cc=paul@paul-moore.com \ --cc=rostedt@goodmis.org \ --cc=selinux@vger.kernel.org \ --cc=stephen.smalley.work@gmail.com \ /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: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.