Re: [PATCH ebpf] bpf: Disallow unprivileged bpf by default

[Pawan Gupta]

On 27.10.2021 23:21, Daniel Borkmann wrote:
> Hello Pawan,
>
> On 10/27/21 10:51 PM, Pawan Gupta wrote:
> > Disabling unprivileged BPF by default would help prevent unprivileged
> > users from creating the conditions required for potential speculative
> > execution side-channel attacks on affected hardware as demonstrated by
> > [1][2][3].
> >
> > This will sync mainline with what most distros are currently applying.
> > An admin can enable this at runtime if necessary.
> >
> > Signed-off-by: Pawan Gupta <pawan.kumar.gupta@xxxxxxxxxxxxxxx>
> >
> > [1] https://access.redhat.com/security/cve/cve-2019-7308
> > [2] https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2021-3490
> > [3] https://bugzilla.redhat.com/show_bug.cgi?id=1672355#c5
>
> Some of your above quoted links are just random ?! For example, [2] has really _zero_ to
> do with what you wrote with regards to speculative execution side-channel attacks ...
>
> We recently did a deep dive on our mitigation work we did in BPF here [0]. This also includes
> an appendix with an extract of the main commits related to the different Spectre variants.
>
> I'd suggest to link to that one instead to avoid confusion on what is related and what not.
>
>  [0] https://ebpf.io/summit-2021-slides/eBPF_Summit_2021-Keynote-Daniel_Borkmann-BPF_and_Spectre.pdf

Sure, I will add reference to this presentation.

> > ---
> >  kernel/bpf/Kconfig | 5 +++++
> >  1 file changed, 5 insertions(+)
> >
> > diff --git a/kernel/bpf/Kconfig b/kernel/bpf/Kconfig
> > index a82d6de86522..73d446294455 100644
> > --- a/kernel/bpf/Kconfig
> > +++ b/kernel/bpf/Kconfig
> > @@ -64,6 +64,7 @@ config BPF_JIT_DEFAULT_ON
> >  config BPF_UNPRIV_DEFAULT_OFF
> >  	bool "Disable unprivileged BPF by default"
> > +	default y
>
> Hm, arm arch has a CPU_SPECTRE Kconfig symbol, see commit c58d237d0852 ("ARM: spectre:
> add Kconfig symbol for CPUs vulnerable to Spectre") that can be selected.
>
> Would be good to generalize it for reuse so archs can select it, and make the above as
> 'default y if CPU_SPECTRE'.

Thanks for your feedback, I will send a v2 soon. I guess below is how
you want it to be:

---
diff --git a/arch/Kconfig b/arch/Kconfig
index 8df1c7102643..6aa856d51cb7 100644
--- a/arch/Kconfig
+++ b/arch/Kconfig
@@ -1091,6 +1091,9 @@ config ARCH_SUPPORTS_RT
 config CPU_NO_EFFICIENT_FFS
 	def_bool n
 
+config CPU_SPECTRE
+	bool
+
 config HAVE_ARCH_VMAP_STACK
 	def_bool n
 	help
diff --git a/arch/arm/mm/Kconfig b/arch/arm/mm/Kconfig
index 8355c3895894..44551465fd03 100644
--- a/arch/arm/mm/Kconfig
+++ b/arch/arm/mm/Kconfig
@@ -828,9 +828,6 @@ config CPU_BPREDICT_DISABLE
 	help
 	  Say Y here to disable branch prediction.  If unsure, say N.
 
-config CPU_SPECTRE
-	bool
-
 config HARDEN_BRANCH_PREDICTOR
 	bool "Harden the branch predictor against aliasing attacks" if EXPERT
 	depends on CPU_SPECTRE
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index d9830e7e1060..769739da67c6 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -124,6 +124,7 @@ config X86
 	select CLKEVT_I8253
 	select CLOCKSOURCE_VALIDATE_LAST_CYCLE
 	select CLOCKSOURCE_WATCHDOG
+	select CPU_SPECTRE
 	select DCACHE_WORD_ACCESS
 	select EDAC_ATOMIC_SCRUB
 	select EDAC_SUPPORT
diff --git a/kernel/bpf/Kconfig b/kernel/bpf/Kconfig
index a82d6de86522..510a5a73f9a2 100644
--- a/kernel/bpf/Kconfig
+++ b/kernel/bpf/Kconfig
@@ -64,6 +64,7 @@ config BPF_JIT_DEFAULT_ON
 
 config BPF_UNPRIV_DEFAULT_OFF
 	bool "Disable unprivileged BPF by default"
+	default y if CPU_SPECTRE
 	depends on BPF_SYSCALL
 	help
 	  Disables unprivileged BPF by default by setting the corresponding
@@ -72,6 +73,10 @@ config BPF_UNPRIV_DEFAULT_OFF
 	  disable it by setting it to 1 (from which no other transition to
 	  0 is possible anymore).
 
+	  Unprivileged BPF can be used to exploit potential speculative
+	  execution side-channel vulnerabilities on affected hardware. If you
+	  are concerned about it, answer Y.
+
 source "kernel/bpf/preload/Kconfig"
 
 config BPF_LSM